xref: /openbsd-src/usr.sbin/ospf6d/kroute.c (revision d13be5d47e4149db2549a9828e244d59dbc43f15)
1 /*	$OpenBSD: kroute.c,v 1.36 2011/07/07 18:39:11 claudio 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/param.h>
21 #include <sys/types.h>
22 #include <sys/socket.h>
23 #include <sys/sysctl.h>
24 #include <sys/tree.h>
25 #include <sys/uio.h>
26 #include <netinet/in.h>
27 #include <arpa/inet.h>
28 #include <net/if.h>
29 #include <net/if_dl.h>
30 #include <net/if_types.h>
31 #include <net/route.h>
32 #include <err.h>
33 #include <errno.h>
34 #include <fcntl.h>
35 #include <stdio.h>
36 #include <stdlib.h>
37 #include <string.h>
38 #include <unistd.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 	int			fd;
49 	struct event		ev;
50 } kr_state;
51 
52 struct kroute_node {
53 	RB_ENTRY(kroute_node)	 entry;
54 	struct kroute		 r;
55 	struct kroute_node	*next;
56 };
57 
58 void	kr_redist_remove(struct kroute_node *, struct kroute_node *);
59 int	kr_redist_eval(struct kroute *, struct rroute *);
60 void	kr_redistribute(struct kroute_node *);
61 int	kroute_compare(struct kroute_node *, struct kroute_node *);
62 
63 struct kroute_node	*kroute_find(const struct in6_addr *, u_int8_t);
64 struct kroute_node	*kroute_matchgw(struct kroute_node *,
65 			    struct in6_addr *, unsigned int);
66 int			 kroute_insert(struct kroute_node *);
67 int			 kroute_remove(struct kroute_node *);
68 void			 kroute_clear(void);
69 
70 struct iface		*kif_update(u_short, int, struct if_data *,
71 			   struct sockaddr_dl *);
72 int			 kif_validate(u_short);
73 
74 struct kroute_node	*kroute_match(struct in6_addr *);
75 
76 int		protect_lo(void);
77 void		get_rtaddrs(int, struct sockaddr *, struct sockaddr **);
78 void		if_change(u_short, int, struct if_data *);
79 void		if_newaddr(u_short, struct sockaddr_in6 *,
80 		    struct sockaddr_in6 *, struct sockaddr_in6 *);
81 void		if_deladdr(u_short, struct sockaddr_in6 *,
82 		    struct sockaddr_in6 *, struct sockaddr_in6 *);
83 void		if_announce(void *);
84 
85 int		send_rtmsg(int, int, struct kroute *);
86 int		dispatch_rtmsg(void);
87 int		fetchtable(void);
88 
89 RB_HEAD(kroute_tree, kroute_node)	krt;
90 RB_PROTOTYPE(kroute_tree, kroute_node, entry, kroute_compare)
91 RB_GENERATE(kroute_tree, kroute_node, entry, kroute_compare)
92 
93 int
94 kr_init(int fs)
95 {
96 	int		opt = 0, rcvbuf, default_rcvbuf;
97 	socklen_t	optlen;
98 
99 	kr_state.fib_sync = fs;
100 
101 	if ((kr_state.fd = socket(AF_ROUTE, SOCK_RAW, 0)) == -1) {
102 		log_warn("kr_init: socket");
103 		return (-1);
104 	}
105 
106 	/* not interested in my own messages */
107 	if (setsockopt(kr_state.fd, SOL_SOCKET, SO_USELOOPBACK,
108 	    &opt, sizeof(opt)) == -1)
109 		log_warn("kr_init: setsockopt");	/* not fatal */
110 
111 	/* grow receive buffer, don't wanna miss messages */
112 	optlen = sizeof(default_rcvbuf);
113 	if (getsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF,
114 	    &default_rcvbuf, &optlen) == -1)
115 		log_warn("kr_init getsockopt SOL_SOCKET SO_RCVBUF");
116 	else
117 		for (rcvbuf = MAX_RTSOCK_BUF;
118 		    rcvbuf > default_rcvbuf &&
119 		    setsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF,
120 		    &rcvbuf, sizeof(rcvbuf)) == -1 && errno == ENOBUFS;
121 		    rcvbuf /= 2)
122 			;	/* nothing */
123 
124 	kr_state.pid = getpid();
125 	kr_state.rtseq = 1;
126 
127 	RB_INIT(&krt);
128 
129 	if (fetchtable() == -1)
130 		return (-1);
131 
132 	if (protect_lo() == -1)
133 		return (-1);
134 
135 	event_set(&kr_state.ev, kr_state.fd, EV_READ | EV_PERSIST,
136 	    kr_dispatch_msg, NULL);
137 	event_add(&kr_state.ev, NULL);
138 
139 	return (0);
140 }
141 
142 int
143 kr_change(struct kroute *kroute)
144 {
145 	struct kroute_node	*kr;
146 	int			 action = RTM_ADD;
147 
148 	kroute->rtlabel = rtlabel_tag2id(kroute->ext_tag);
149 
150 	if ((kr = kroute_find(&kroute->prefix, kroute->prefixlen)) !=
151 	    NULL) {
152 		if (!(kr->r.flags & F_KERNEL))
153 			action = RTM_CHANGE;
154 		else {	/* a non-ospf route already exists. not a problem */
155 			if (!(kr->r.flags & F_BGPD_INSERTED)) {
156 				do {
157 					kr->r.flags |= F_OSPFD_INSERTED;
158 					kr = kr->next;
159 				} while (kr);
160 				return (0);
161 			}
162 			/*
163 			 * XXX as long as there is no multipath support in
164 			 * bgpd this is safe else we end up in a bad situation.
165 			 */
166 			/*
167 			 * ospf route has higher pref
168 			 * - reset flags to the ospf ones
169 			 * - use RTM_CHANGE
170 			 * - zero out ifindex (this is no longer relevant)
171 			 */
172 			action = RTM_CHANGE;
173 			kr->r.flags = kroute->flags | F_OSPFD_INSERTED;
174 			kr->r.ifindex = 0;
175 			rtlabel_unref(kr->r.rtlabel);
176 			kr->r.ext_tag = kroute->ext_tag;
177 			kr->r.rtlabel = kroute->rtlabel;
178 		}
179 	}
180 
181 	/* nexthop within 127/8 -> ignore silently */
182 	if (kr && IN6_IS_ADDR_LOOPBACK(&kr->r.nexthop))
183 		return (0);
184 
185 	/*
186 	 * Ingnore updates that did not change the route.
187 	 * Currently only the nexthop can change.
188 	 */
189 	if (kr && kr->r.scope == kroute->scope &&
190 	    IN6_ARE_ADDR_EQUAL(&kr->r.nexthop, &kroute->nexthop))
191 		return (0);
192 
193 	if (send_rtmsg(kr_state.fd, action, kroute) == -1)
194 		return (-1);
195 
196 	if (action == RTM_ADD) {
197 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
198 			log_warn("kr_change");
199 			return (-1);
200 		}
201 		kr->r.prefix = kroute->prefix;
202 		kr->r.prefixlen = kroute->prefixlen;
203 		kr->r.nexthop = kroute->nexthop;
204 		kr->r.scope = kroute->scope;
205 		kr->r.flags = kroute->flags | F_OSPFD_INSERTED;
206 		kr->r.ext_tag = kroute->ext_tag;
207 		kr->r.rtlabel = kroute->rtlabel;
208 
209 		if (kroute_insert(kr) == -1)
210 			free(kr);
211 	} else if (kr) {
212 		kr->r.nexthop = kroute->nexthop;
213 		kr->r.scope = kroute->scope;
214 	}
215 
216 	return (0);
217 }
218 
219 int
220 kr_delete(struct kroute *kroute)
221 {
222 	struct kroute_node	*kr;
223 
224 	if ((kr = kroute_find(&kroute->prefix, kroute->prefixlen)) ==
225 	    NULL)
226 		return (0);
227 
228 	if (!(kr->r.flags & F_OSPFD_INSERTED))
229 		return (0);
230 
231 	if (kr->r.flags & F_KERNEL) {
232 		/* remove F_OSPFD_INSERTED flag, route still exists in kernel */
233 		do {
234 			kr->r.flags &= ~F_OSPFD_INSERTED;
235 			kr = kr->next;
236 		} while (kr);
237 		return (0);
238 	}
239 
240 	if (send_rtmsg(kr_state.fd, RTM_DELETE, kroute) == -1)
241 		return (-1);
242 
243 	if (kroute_remove(kr) == -1)
244 		return (-1);
245 
246 	return (0);
247 }
248 
249 void
250 kr_shutdown(void)
251 {
252 	kr_fib_decouple();
253 	kroute_clear();
254 }
255 
256 void
257 kr_fib_couple(void)
258 {
259 	struct kroute_node	*kr;
260 
261 	if (kr_state.fib_sync == 1)	/* already coupled */
262 		return;
263 
264 	kr_state.fib_sync = 1;
265 
266 	RB_FOREACH(kr, kroute_tree, &krt)
267 		if (!(kr->r.flags & F_KERNEL))
268 			send_rtmsg(kr_state.fd, RTM_ADD, &kr->r);
269 
270 	log_info("kernel routing table coupled");
271 }
272 
273 void
274 kr_fib_decouple(void)
275 {
276 	struct kroute_node	*kr;
277 
278 	if (kr_state.fib_sync == 0)	/* already decoupled */
279 		return;
280 
281 	RB_FOREACH(kr, kroute_tree, &krt)
282 		if (!(kr->r.flags & F_KERNEL))
283 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
284 
285 	kr_state.fib_sync = 0;
286 
287 	log_info("kernel routing table decoupled");
288 }
289 
290 /* ARGSUSED */
291 void
292 kr_dispatch_msg(int fd, short event, void *bula)
293 {
294 	dispatch_rtmsg();
295 }
296 
297 void
298 kr_show_route(struct imsg *imsg)
299 {
300 	struct kroute_node	*kr;
301 	struct kroute_node	*kn;
302 	int			 flags;
303 	struct in6_addr		 addr;
304 
305 	switch (imsg->hdr.type) {
306 	case IMSG_CTL_KROUTE:
307 		if (imsg->hdr.len != IMSG_HEADER_SIZE + sizeof(flags)) {
308 			log_warnx("kr_show_route: wrong imsg len");
309 			return;
310 		}
311 		memcpy(&flags, imsg->data, sizeof(flags));
312 		RB_FOREACH(kr, kroute_tree, &krt)
313 			if (!flags || kr->r.flags & flags) {
314 				kn = kr;
315 				do {
316 					main_imsg_compose_ospfe(IMSG_CTL_KROUTE,
317 					    imsg->hdr.pid,
318 					    &kn->r, sizeof(kn->r));
319 				} while ((kn = kn->next) != NULL);
320 			}
321 		break;
322 	case IMSG_CTL_KROUTE_ADDR:
323 		if (imsg->hdr.len != IMSG_HEADER_SIZE +
324 		    sizeof(struct in6_addr)) {
325 			log_warnx("kr_show_route: wrong imsg len");
326 			return;
327 		}
328 		memcpy(&addr, imsg->data, sizeof(addr));
329 		kr = NULL;
330 		kr = kroute_match(&addr);
331 		if (kr != NULL)
332 			main_imsg_compose_ospfe(IMSG_CTL_KROUTE, imsg->hdr.pid,
333 			    &kr->r, sizeof(kr->r));
334 		break;
335 	default:
336 		log_debug("kr_show_route: error handling imsg");
337 		break;
338 	}
339 
340 	main_imsg_compose_ospfe(IMSG_CTL_END, imsg->hdr.pid, NULL, 0);
341 }
342 
343 void
344 kr_redist_remove(struct kroute_node *kh, struct kroute_node *kn)
345 {
346 	struct rroute	 rr;
347 
348 	/* was the route redistributed? */
349 	if ((kn->r.flags & F_REDISTRIBUTED) == 0)
350 		return;
351 
352 	/* remove redistributed flag */
353 	kn->r.flags &= ~F_REDISTRIBUTED;
354 	rr.kr = kn->r;
355 	rr.metric = DEFAULT_REDIST_METRIC;	/* some dummy value */
356 
357 	/* probably inform the RDE (check if no other path is redistributed) */
358 	for (kn = kh; kn; kn = kn->next)
359 		if (kn->r.flags & F_REDISTRIBUTED)
360 			break;
361 
362 	if (kn == NULL)
363 		main_imsg_compose_rde(IMSG_NETWORK_DEL, 0, &rr,
364 		    sizeof(struct rroute));
365 }
366 
367 int
368 kr_redist_eval(struct kroute *kr, struct rroute *rr)
369 {
370 	u_int32_t	 metric = 0;
371 
372 	/* Only non-ospfd routes are considered for redistribution. */
373 	if (!(kr->flags & F_KERNEL))
374 		goto dont_redistribute;
375 
376 	/* Dynamic routes are not redistributable. */
377 	if (kr->flags & F_DYNAMIC)
378 		goto dont_redistribute;
379 
380 	/* interface is not up and running so don't announce */
381 	if (kr->flags & F_DOWN)
382 		goto dont_redistribute;
383 
384 	/*
385 	 * We consider loopback, multicast, link- and site-local,
386 	 * IPv4 mapped and IPv4 compatible addresses as not redistributable.
387 	 */
388 	if (IN6_IS_ADDR_LOOPBACK(&kr->prefix) ||
389 	    IN6_IS_ADDR_MULTICAST(&kr->prefix) ||
390 	    IN6_IS_ADDR_LINKLOCAL(&kr->prefix) ||
391 	    IN6_IS_ADDR_SITELOCAL(&kr->prefix) ||
392 	    IN6_IS_ADDR_V4MAPPED(&kr->prefix) ||
393 	    IN6_IS_ADDR_V4COMPAT(&kr->prefix))
394 		goto dont_redistribute;
395 	/*
396 	 * Consider networks with nexthop loopback as not redistributable.
397 	 */
398 	if (IN6_IS_ADDR_LOOPBACK(&kr->nexthop))
399 		goto dont_redistribute;
400 
401 	/* Should we redistrubute this route? */
402 	if (!ospf_redistribute(kr, &metric))
403 		goto dont_redistribute;
404 
405 	/* prefix should be redistributed */
406 	kr->flags |= F_REDISTRIBUTED;
407 	/*
408 	 * only on of all multipath routes can be redistributed so
409 	 * redistribute the best one.
410 	 */
411 	if (rr->metric > metric) {
412 		rr->kr = *kr;
413 		rr->metric = metric;
414 	}
415 	return (1);
416 
417 dont_redistribute:
418 	/* was the route redistributed? */
419 	if ((kr->flags & F_REDISTRIBUTED) == 0)
420 		return (0);
421 
422 	kr->flags &= ~F_REDISTRIBUTED;
423 	return (1);
424 }
425 
426 void
427 kr_redistribute(struct kroute_node *kh)
428 {
429 	struct kroute_node	*kn;
430 	struct rroute		 rr;
431 	int			 redistribute = 0;
432 
433 	bzero(&rr, sizeof(rr));
434 	rr.metric = UINT_MAX;
435 	for (kn = kh; kn; kn = kn->next)
436 		if (kr_redist_eval(&kn->r, &rr))
437 			redistribute = 1;
438 
439 	if (!redistribute)
440 		return;
441 
442 	if (rr.kr.flags & F_REDISTRIBUTED) {
443 		main_imsg_compose_rde(IMSG_NETWORK_ADD, 0, &rr,
444 		    sizeof(struct rroute));
445 	} else {
446 		rr.metric = DEFAULT_REDIST_METRIC;	/* some dummy value */
447 		rr.kr = kh->r;
448 		main_imsg_compose_rde(IMSG_NETWORK_DEL, 0, &rr,
449 		    sizeof(struct rroute));
450 	}
451 }
452 
453 void
454 kr_reload(void)
455 {
456 	struct kroute_node	*kr, *kn;
457 	u_int32_t		 dummy;
458 	int			 r;
459 
460 	RB_FOREACH(kr, kroute_tree, &krt) {
461 		for (kn = kr; kn; kn = kn->next) {
462 			r = ospf_redistribute(&kn->r, &dummy);
463 			/*
464 			 * if it is redistributed, redistribute again metric
465 			 * may have changed.
466 			 */
467 			if ((kn->r.flags & F_REDISTRIBUTED && !r) || r)
468 				break;
469 		}
470 		if (kn) {
471 			/*
472 			 * kr_redistribute copes with removes and RDE with
473 			 * duplicates
474 			 */
475 			kr_redistribute(kr);
476 		}
477 	}
478 }
479 
480 /* rb-tree compare */
481 int
482 kroute_compare(struct kroute_node *a, struct kroute_node *b)
483 {
484 	int	i;
485 
486 	/* XXX maybe switch a & b */
487 	i = memcmp(&a->r.prefix, &b->r.prefix, sizeof(a->r.prefix));
488 	if (i)
489 		return (i);
490 	if (a->r.prefixlen < b->r.prefixlen)
491 		return (-1);
492 	if (a->r.prefixlen > b->r.prefixlen)
493 		return (1);
494 	return (0);
495 }
496 
497 /* tree management */
498 struct kroute_node *
499 kroute_find(const struct in6_addr *prefix, u_int8_t prefixlen)
500 {
501 	struct kroute_node	s;
502 
503 	s.r.prefix = *prefix;
504 	s.r.prefixlen = prefixlen;
505 
506 	return (RB_FIND(kroute_tree, &krt, &s));
507 }
508 
509 struct kroute_node *
510 kroute_matchgw(struct kroute_node *kr, struct in6_addr *nh, unsigned int scope)
511 {
512 	while (kr) {
513 		if (scope == kr->r.scope &&
514 		    IN6_ARE_ADDR_EQUAL(&kr->r.nexthop, nh))
515 			return (kr);
516 		kr = kr->next;
517 	}
518 
519 	return (NULL);
520 }
521 
522 int
523 kroute_insert(struct kroute_node *kr)
524 {
525 	struct kroute_node	*krm, *krh;
526 
527 	if ((krh = RB_INSERT(kroute_tree, &krt, kr)) != NULL) {
528 		/*
529 		 * Multipath route, add at end of list and clone the
530 		 * ospfd inserted flag.
531 		 */
532 		krm = krh;
533 		kr->r.flags |= krm->r.flags & F_OSPFD_INSERTED;
534 		while (krm->next != NULL)
535 			krm = krm->next;
536 		krm->next = kr;
537 		kr->next = NULL; /* to be sure */
538 	} else
539 		krh = kr;
540 
541 	if (!(kr->r.flags & F_KERNEL)) {
542 		/* don't validate or redistribute ospf route */
543 		kr->r.flags &= ~F_DOWN;
544 		return (0);
545 	}
546 
547 	if (kif_validate(kr->r.ifindex))
548 		kr->r.flags &= ~F_DOWN;
549 	else
550 		kr->r.flags |= F_DOWN;
551 
552 	kr_redistribute(krh);
553 	return (0);
554 }
555 
556 int
557 kroute_remove(struct kroute_node *kr)
558 {
559 	struct kroute_node	*krm;
560 
561 	if ((krm = RB_FIND(kroute_tree, &krt, kr)) == NULL) {
562 		log_warnx("kroute_remove failed to find %s/%u",
563 		    log_in6addr(&kr->r.prefix), kr->r.prefixlen);
564 		return (-1);
565 	}
566 
567 	if (krm == kr) {
568 		/* head element */
569 		if (RB_REMOVE(kroute_tree, &krt, kr) == NULL) {
570 			log_warnx("kroute_remove failed for %s/%u",
571 			    log_in6addr(&kr->r.prefix), kr->r.prefixlen);
572 			return (-1);
573 		}
574 		if (kr->next != NULL) {
575 			if (RB_INSERT(kroute_tree, &krt, kr->next) != NULL) {
576 				log_warnx("kroute_remove failed to add %s/%u",
577 				    log_in6addr(&kr->r.prefix),
578 				    kr->r.prefixlen);
579 				return (-1);
580 			}
581 		}
582 	} else {
583 		/* somewhere in the list */
584 		while (krm->next != kr && krm->next != NULL)
585 			krm = krm->next;
586 		if (krm->next == NULL) {
587 			log_warnx("kroute_remove multipath list corrupted "
588 			    "for %s/%u", log_in6addr(&kr->r.prefix),
589 			    kr->r.prefixlen);
590 			return (-1);
591 		}
592 		krm->next = kr->next;
593 	}
594 
595 	kr_redist_remove(krm, kr);
596 	rtlabel_unref(kr->r.rtlabel);
597 
598 	free(kr);
599 	return (0);
600 }
601 
602 void
603 kroute_clear(void)
604 {
605 	struct kroute_node	*kr;
606 
607 	while ((kr = RB_MIN(kroute_tree, &krt)) != NULL)
608 		kroute_remove(kr);
609 }
610 
611 struct iface *
612 kif_update(u_short ifindex, int flags, struct if_data *ifd,
613     struct sockaddr_dl *sdl)
614 {
615 	struct iface	*iface;
616 	char		 ifname[IF_NAMESIZE];
617 
618 	if ((iface = if_find(ifindex)) == NULL) {
619 		bzero(ifname, sizeof(ifname));
620 		if (sdl && sdl->sdl_family == AF_LINK) {
621 			if (sdl->sdl_nlen >= sizeof(ifname))
622 				memcpy(ifname, sdl->sdl_data,
623 				    sizeof(ifname) - 1);
624 			else if (sdl->sdl_nlen > 0)
625 				memcpy(ifname, sdl->sdl_data, sdl->sdl_nlen);
626 			else
627 				return (NULL);
628 		} else
629 			return (NULL);
630 		if ((iface = if_new(ifindex, ifname)) == NULL)
631 			return (NULL);
632 		iface->cflags |= F_IFACE_AVAIL;
633 	}
634 
635 	if_update(iface, ifd->ifi_mtu, flags, ifd->ifi_type,
636 	    ifd->ifi_link_state, ifd->ifi_baudrate);
637 
638 	return (iface);
639 }
640 
641 int
642 kif_validate(u_short ifindex)
643 {
644 	struct iface	*iface;
645 
646 	if ((iface = if_find(ifindex)) == NULL) {
647 		log_warnx("interface with index %u not found", ifindex);
648 		return (-1);
649 	}
650 
651 	return ((iface->flags & IFF_UP) && LINK_STATE_IS_UP(iface->linkstate));
652 }
653 
654 struct kroute_node *
655 kroute_match(struct in6_addr *key)
656 {
657 	int			 i;
658 	struct kroute_node	*kr;
659 	struct in6_addr		 ina;
660 
661 	/* we will never match the default route */
662 	for (i = 128; i > 0; i--) {
663 		inet6applymask(&ina, key, i);
664 		if ((kr = kroute_find(&ina, i)) != NULL)
665 			return (kr);
666 	}
667 
668 	/* if we don't have a match yet, try to find a default route */
669 	if ((kr = kroute_find(&in6addr_any, 0)) != NULL)
670 			return (kr);
671 
672 	return (NULL);
673 }
674 
675 /* misc */
676 int
677 protect_lo(void)
678 {
679 	struct kroute_node	*kr;
680 
681 	/* special protection for loopback */
682 	if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
683 		log_warn("protect_lo");
684 		return (-1);
685 	}
686 	memcpy(&kr->r.prefix, &in6addr_loopback, sizeof(kr->r.prefix));
687 	kr->r.prefixlen = 128;
688 	kr->r.flags = F_KERNEL|F_CONNECTED;
689 
690 	if (RB_INSERT(kroute_tree, &krt, kr) != NULL)
691 		free(kr);	/* kernel route already there, no problem */
692 
693 	return (0);
694 }
695 
696 u_int8_t
697 mask2prefixlen(struct sockaddr_in6 *sa_in6)
698 {
699 	u_int8_t	l = 0, *ap, *ep;
700 
701 	/*
702 	 * sin6_len is the size of the sockaddr so substract the offset of
703 	 * the possibly truncated sin6_addr struct.
704 	 */
705 	ap = (u_int8_t *)&sa_in6->sin6_addr;
706 	ep = (u_int8_t *)sa_in6 + sa_in6->sin6_len;
707 	for (; ap < ep; ap++) {
708 		/* this "beauty" is adopted from sbin/route/show.c ... */
709 		switch (*ap) {
710 		case 0xff:
711 			l += 8;
712 			break;
713 		case 0xfe:
714 			l += 7;
715 			return (l);
716 		case 0xfc:
717 			l += 6;
718 			return (l);
719 		case 0xf8:
720 			l += 5;
721 			return (l);
722 		case 0xf0:
723 			l += 4;
724 			return (l);
725 		case 0xe0:
726 			l += 3;
727 			return (l);
728 		case 0xc0:
729 			l += 2;
730 			return (l);
731 		case 0x80:
732 			l += 1;
733 			return (l);
734 		case 0x00:
735 			return (l);
736 		default:
737 			fatalx("non contiguous inet6 netmask");
738 		}
739 	}
740 
741 	return (l);
742 }
743 
744 struct in6_addr *
745 prefixlen2mask(u_int8_t prefixlen)
746 {
747 	static struct in6_addr	mask;
748 	int			i;
749 
750 	bzero(&mask, sizeof(mask));
751 	for (i = 0; i < prefixlen / 8; i++)
752 		mask.s6_addr[i] = 0xff;
753 	i = prefixlen % 8;
754 	if (i)
755 		mask.s6_addr[prefixlen / 8] = 0xff00 >> i;
756 
757 	return (&mask);
758 }
759 
760 void
761 inet6applymask(struct in6_addr *dest, const struct in6_addr *src, int prefixlen)
762 {
763 	struct in6_addr	mask;
764 	int		i;
765 
766 	bzero(&mask, sizeof(mask));
767 	for (i = 0; i < prefixlen / 8; i++)
768 		mask.s6_addr[i] = 0xff;
769 	i = prefixlen % 8;
770 	if (i)
771 		mask.s6_addr[prefixlen / 8] = 0xff00 >> i;
772 
773 	for (i = 0; i < 16; i++)
774 		dest->s6_addr[i] = src->s6_addr[i] & mask.s6_addr[i];
775 }
776 
777 #define ROUNDUP(a) \
778 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
779 
780 void
781 get_rtaddrs(int addrs, struct sockaddr *sa, struct sockaddr **rti_info)
782 {
783 	int	i;
784 
785 	for (i = 0; i < RTAX_MAX; i++) {
786 		if (addrs & (1 << i)) {
787 			rti_info[i] = sa;
788 			sa = (struct sockaddr *)((char *)(sa) +
789 			    ROUNDUP(sa->sa_len));
790 		} else
791 			rti_info[i] = NULL;
792 	}
793 }
794 
795 void
796 if_change(u_short ifindex, int flags, struct if_data *ifd)
797 {
798 	struct kroute_node	*kr, *tkr;
799 	struct iface		*iface;
800 	u_int8_t		 wasvalid, isvalid;
801 
802 	wasvalid = kif_validate(ifindex);
803 
804 	if ((iface = kif_update(ifindex, flags, ifd, NULL)) == NULL) {
805 		log_warn("if_change: kif_update(%u)", ifindex);
806 		return;
807 	}
808 
809 	isvalid = (iface->flags & IFF_UP) &&
810 	    LINK_STATE_IS_UP(iface->linkstate);
811 
812 	if (wasvalid == isvalid)
813 		return;		/* nothing changed wrt validity */
814 
815 	/* inform engine and rde about state change if interface is used */
816 	if (iface->cflags & F_IFACE_CONFIGURED) {
817 		main_imsg_compose_ospfe(IMSG_IFINFO, 0, iface,
818 		    sizeof(struct iface));
819 		main_imsg_compose_rde(IMSG_IFINFO, 0, iface,
820 		    sizeof(struct iface));
821 	}
822 
823 	/* update redistribute list */
824 	RB_FOREACH(kr, kroute_tree, &krt) {
825 		for (tkr = kr; tkr != NULL; tkr = tkr->next) {
826 			if (tkr->r.ifindex == ifindex) {
827 				if (isvalid)
828 					tkr->r.flags &= ~F_DOWN;
829 				else
830 					tkr->r.flags |= F_DOWN;
831 
832 			}
833 		}
834 		kr_redistribute(kr);
835 	}
836 }
837 
838 void
839 if_newaddr(u_short ifindex, struct sockaddr_in6 *ifa, struct sockaddr_in6 *mask,
840     struct sockaddr_in6 *brd)
841 {
842 	struct iface		*iface;
843 	struct iface_addr	*ia;
844 	struct ifaddrchange	 ifc;
845 
846 	if (ifa == NULL || ifa->sin6_family != AF_INET6)
847 		return;
848 	if ((iface = if_find(ifindex)) == NULL) {
849 		log_warnx("if_newaddr: corresponding if %i not found", ifindex);
850 		return;
851 	}
852 
853 	/* We only care about link-local and global-scope. */
854 	if (IN6_IS_ADDR_UNSPECIFIED(&ifa->sin6_addr) ||
855 	    IN6_IS_ADDR_LOOPBACK(&ifa->sin6_addr) ||
856 	    IN6_IS_ADDR_MULTICAST(&ifa->sin6_addr) ||
857 	    IN6_IS_ADDR_SITELOCAL(&ifa->sin6_addr) ||
858 	    IN6_IS_ADDR_V4MAPPED(&ifa->sin6_addr) ||
859 	    IN6_IS_ADDR_V4COMPAT(&ifa->sin6_addr))
860 		return;
861 
862 	/* XXX thanks, KAME, for this ugliness... adopted from route/show.c */
863 	if (IN6_IS_ADDR_LINKLOCAL(&ifa->sin6_addr)) {
864 		ifa->sin6_addr.s6_addr[2] = 0;
865 		ifa->sin6_addr.s6_addr[3] = 0;
866 	}
867 
868 	if (IN6_IS_ADDR_LINKLOCAL(&ifa->sin6_addr) ||
869 	    iface->flags & IFF_LOOPBACK)
870 		iface->addr = ifa->sin6_addr;
871 
872 	if ((ia = calloc(1, sizeof(struct iface_addr))) == NULL)
873 		fatal("if_newaddr");
874 
875 	ia->addr = ifa->sin6_addr;
876 
877 	if (mask)
878 		ia->prefixlen = mask2prefixlen(mask);
879 	else
880 		ia->prefixlen = 0;
881 	if (brd && brd->sin6_family == AF_INET6)
882 		ia->dstbrd = brd->sin6_addr;
883 	else
884 		bzero(&ia->dstbrd, sizeof(ia->dstbrd));
885 
886 	switch (iface->type) {
887 	case IF_TYPE_BROADCAST:
888 	case IF_TYPE_NBMA:
889 		log_debug("if_newaddr: ifindex %u, addr %s/%d",
890 		    ifindex, log_in6addr(&ia->addr), ia->prefixlen);
891 		break;
892 	case IF_TYPE_VIRTUALLINK:	/* FIXME */
893 		break;
894 	case IF_TYPE_POINTOPOINT:
895 	case IF_TYPE_POINTOMULTIPOINT:
896 		log_debug("if_newaddr: ifindex %u, addr %s/%d, "
897 		    "dest %s", ifindex, log_in6addr(&ia->addr),
898 		    ia->prefixlen, log_in6addr(&ia->dstbrd));
899 		break;
900 	default:
901 		fatalx("if_newaddr: unknown interface type");
902 	}
903 
904 	TAILQ_INSERT_TAIL(&iface->ifa_list, ia, entry);
905 	/* inform engine and rde if interface is used */
906 	if (iface->cflags & F_IFACE_CONFIGURED) {
907 		ifc.addr = ia->addr;
908 		ifc.dstbrd = ia->dstbrd;
909 		ifc.prefixlen = ia->prefixlen;
910 		ifc.ifindex = ifindex;
911 		main_imsg_compose_ospfe(IMSG_IFADDRNEW, 0, &ifc, sizeof(ifc));
912 		main_imsg_compose_rde(IMSG_IFADDRNEW, 0, &ifc, sizeof(ifc));
913 	}
914 }
915 
916 void
917 if_deladdr(u_short ifindex, struct sockaddr_in6 *ifa, struct sockaddr_in6 *mask,
918     struct sockaddr_in6 *brd)
919 {
920 	struct iface		*iface;
921 	struct iface_addr	*ia, *nia;
922 	struct ifaddrchange	 ifc;
923 
924 	if (ifa == NULL || ifa->sin6_family != AF_INET6)
925 		return;
926 	if ((iface = if_find(ifindex)) == NULL) {
927 		log_warnx("if_deladdr: corresponding if %i not found", ifindex);
928 		return;
929 	}
930 
931 	/* We only care about link-local and global-scope. */
932 	if (IN6_IS_ADDR_UNSPECIFIED(&ifa->sin6_addr) ||
933 	    IN6_IS_ADDR_LOOPBACK(&ifa->sin6_addr) ||
934 	    IN6_IS_ADDR_MULTICAST(&ifa->sin6_addr) ||
935 	    IN6_IS_ADDR_SITELOCAL(&ifa->sin6_addr) ||
936 	    IN6_IS_ADDR_V4MAPPED(&ifa->sin6_addr) ||
937 	    IN6_IS_ADDR_V4COMPAT(&ifa->sin6_addr))
938 		return;
939 
940 	/* XXX thanks, KAME, for this ugliness... adopted from route/show.c */
941 	if (IN6_IS_ADDR_LINKLOCAL(&ifa->sin6_addr)) {
942 		ifa->sin6_addr.s6_addr[2] = 0;
943 		ifa->sin6_addr.s6_addr[3] = 0;
944 	}
945 
946 	for (ia = TAILQ_FIRST(&iface->ifa_list); ia != NULL; ia = nia) {
947 		nia = TAILQ_NEXT(ia, entry);
948 
949 		if (IN6_ARE_ADDR_EQUAL(&ia->addr, &ifa->sin6_addr)) {
950 			log_debug("if_deladdr: ifindex %u, addr %s/%d",
951 			    ifindex, log_in6addr(&ia->addr), ia->prefixlen);
952 			TAILQ_REMOVE(&iface->ifa_list, ia, entry);
953 			/* inform engine and rde if interface is used */
954 			if (iface->cflags & F_IFACE_CONFIGURED) {
955 				ifc.addr = ia->addr;
956 				ifc.dstbrd = ia->dstbrd;
957 				ifc.prefixlen = ia->prefixlen;
958 				ifc.ifindex = ifindex;
959 				main_imsg_compose_ospfe(IMSG_IFADDRDEL, 0, &ifc,
960 				    sizeof(ifc));
961 				main_imsg_compose_rde(IMSG_IFADDRDEL, 0, &ifc,
962 				    sizeof(ifc));
963 			}
964 			free(ia);
965 			return;
966 		}
967 	}
968 }
969 
970 void
971 if_announce(void *msg)
972 {
973 	struct if_announcemsghdr	*ifan;
974 	struct iface			*iface;
975 
976 	ifan = msg;
977 
978 	switch (ifan->ifan_what) {
979 	case IFAN_ARRIVAL:
980 		if ((iface = if_new(ifan->ifan_index, ifan->ifan_name)) == NULL)
981 			fatal("if_announce failed");
982 		iface->cflags |= F_IFACE_AVAIL;
983 		break;
984 	case IFAN_DEPARTURE:
985 		iface = if_find(ifan->ifan_index);
986 		if (iface->cflags & F_IFACE_CONFIGURED) {
987 			main_imsg_compose_rde(IMSG_IFDELETE, 0,
988 			    &iface->ifindex, sizeof(iface->ifindex));
989 			main_imsg_compose_ospfe(IMSG_IFDELETE, 0,
990 			    &iface->ifindex, sizeof(iface->ifindex));
991 		}
992 		if_del(iface);
993 		break;
994 	}
995 }
996 
997 /* rtsock */
998 int
999 send_rtmsg(int fd, int action, struct kroute *kroute)
1000 {
1001 	struct iovec		iov[5];
1002 	struct rt_msghdr	hdr;
1003 	struct pad {
1004 		struct sockaddr_in6	addr;
1005 		char			pad[sizeof(long)]; /* thank you IPv6 */
1006 	} prefix, nexthop, mask;
1007 	struct {
1008 		struct sockaddr_dl	addr;
1009 		char			pad[sizeof(long)];
1010 	} ifp;
1011 	struct sockaddr_rtlabel	sa_rl;
1012 	int			iovcnt = 0;
1013 	const char		*label;
1014 
1015 	if (kr_state.fib_sync == 0)
1016 		return (0);
1017 
1018 	/* initialize header */
1019 	bzero(&hdr, sizeof(hdr));
1020 	hdr.rtm_version = RTM_VERSION;
1021 	hdr.rtm_type = action;
1022 	hdr.rtm_flags = RTF_UP;
1023 	hdr.rtm_priority = RTP_OSPF;
1024 	if (action == RTM_CHANGE)
1025 		hdr.rtm_fmask = RTF_REJECT|RTF_BLACKHOLE;
1026 	hdr.rtm_seq = kr_state.rtseq++;	/* overflow doesn't matter */
1027 	hdr.rtm_hdrlen = sizeof(hdr);
1028 	hdr.rtm_msglen = sizeof(hdr);
1029 	/* adjust iovec */
1030 	iov[iovcnt].iov_base = &hdr;
1031 	iov[iovcnt++].iov_len = sizeof(hdr);
1032 
1033 	bzero(&prefix, sizeof(prefix));
1034 	prefix.addr.sin6_len = sizeof(struct sockaddr_in6);
1035 	prefix.addr.sin6_family = AF_INET6;
1036 	prefix.addr.sin6_addr = kroute->prefix;
1037 	/* adjust header */
1038 	hdr.rtm_addrs |= RTA_DST;
1039 	hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
1040 	/* adjust iovec */
1041 	iov[iovcnt].iov_base = &prefix;
1042 	iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
1043 
1044 	if (!IN6_IS_ADDR_UNSPECIFIED(&kroute->nexthop)) {
1045 		bzero(&nexthop, sizeof(nexthop));
1046 		nexthop.addr.sin6_len = sizeof(struct sockaddr_in6);
1047 		nexthop.addr.sin6_family = AF_INET6;
1048 		nexthop.addr.sin6_addr = kroute->nexthop;
1049 		/*
1050 		 * XXX we should set the sin6_scope_id but the kernel
1051 		 * XXX does not expect it that way. It must be fiddled
1052 		 * XXX into the sin6_addr. Welcome to the typical
1053 		 * XXX IPv6 insanity and all without wine bottles.
1054 		 */
1055 		if (IN6_IS_ADDR_LINKLOCAL(&nexthop.addr.sin6_addr)) {
1056 			/* nexthop.addr.sin6_scope_id = kroute->scope; */
1057 			nexthop.addr.sin6_addr.s6_addr[2] =
1058 			    (kroute->scope >> 8) & 0xff;
1059 			nexthop.addr.sin6_addr.s6_addr[3] =
1060 			    kroute->scope & 0xff;
1061 		}
1062 		/* adjust header */
1063 		hdr.rtm_flags |= RTF_GATEWAY;
1064 		hdr.rtm_addrs |= RTA_GATEWAY;
1065 		hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
1066 		/* adjust iovec */
1067 		iov[iovcnt].iov_base = &nexthop;
1068 		iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
1069 	} else if (kroute->ifindex) {
1070 		/*
1071 		 * We don't have an interface address in that network,
1072 		 * so we install a cloning route.  The kernel will then
1073 		 * do neigbor discovery.
1074 		 */
1075 		bzero(&ifp, sizeof(ifp));
1076 		ifp.addr.sdl_len = sizeof(struct sockaddr_dl);
1077 		ifp.addr.sdl_family = AF_LINK;
1078 
1079 		ifp.addr.sdl_index  = kroute->ifindex;
1080 		/* adjust header */
1081 		hdr.rtm_flags |= RTF_CLONING;
1082 		hdr.rtm_addrs |= RTA_GATEWAY;
1083 		hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_dl));
1084 		/* adjust iovec */
1085 		iov[iovcnt].iov_base = &ifp;
1086 		iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_dl));
1087 	}
1088 
1089 	bzero(&mask, sizeof(mask));
1090 	mask.addr.sin6_len = sizeof(struct sockaddr_in6);
1091 	mask.addr.sin6_family = AF_INET6;
1092 	mask.addr.sin6_addr = *prefixlen2mask(kroute->prefixlen);
1093 	/* adjust header */
1094 	if (kroute->prefixlen == 128)
1095 		hdr.rtm_flags |= RTF_HOST;
1096 	hdr.rtm_addrs |= RTA_NETMASK;
1097 	hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
1098 	/* adjust iovec */
1099 	iov[iovcnt].iov_base = &mask;
1100 	iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
1101 
1102 	if (kroute->rtlabel != 0) {
1103 		sa_rl.sr_len = sizeof(sa_rl);
1104 		sa_rl.sr_family = AF_UNSPEC;
1105 		label = rtlabel_id2name(kroute->rtlabel);
1106 		if (strlcpy(sa_rl.sr_label, label,
1107 		    sizeof(sa_rl.sr_label)) >= sizeof(sa_rl.sr_label)) {
1108 			log_warnx("send_rtmsg: invalid rtlabel");
1109 			return (-1);
1110 		}
1111 		/* adjust header */
1112 		hdr.rtm_addrs |= RTA_LABEL;
1113 		hdr.rtm_msglen += sizeof(sa_rl);
1114 		/* adjust iovec */
1115 		iov[iovcnt].iov_base = &sa_rl;
1116 		iov[iovcnt++].iov_len = sizeof(sa_rl);
1117 	}
1118 
1119 retry:
1120 	if (writev(fd, iov, iovcnt) == -1) {
1121 		if (errno == ESRCH) {
1122 			if (hdr.rtm_type == RTM_CHANGE) {
1123 				hdr.rtm_type = RTM_ADD;
1124 				goto retry;
1125 			} else if (hdr.rtm_type == RTM_DELETE) {
1126 				log_info("route %s/%u vanished before delete",
1127 				    log_sockaddr(&prefix), kroute->prefixlen);
1128 				return (0);
1129 			}
1130 		}
1131 		log_warn("send_rtmsg: action %u, prefix %s/%u", hdr.rtm_type,
1132 		    log_sockaddr(&prefix), kroute->prefixlen);
1133 		return (0);
1134 	}
1135 
1136 	return (0);
1137 }
1138 
1139 int
1140 fetchtable(void)
1141 {
1142 	size_t			 len;
1143 	int			 mib[7];
1144 	char			*buf, *next, *lim;
1145 	struct rt_msghdr	*rtm;
1146 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1147 	struct sockaddr_in6	*sa_in6;
1148 	struct sockaddr_rtlabel	*label;
1149 	struct kroute_node	*kr;
1150 
1151 	mib[0] = CTL_NET;
1152 	mib[1] = AF_ROUTE;
1153 	mib[2] = 0;
1154 	mib[3] = AF_INET6;
1155 	mib[4] = NET_RT_DUMP;
1156 	mib[5] = 0;
1157 	mib[6] = 0;	/* rtableid */
1158 
1159 	if (sysctl(mib, 7, NULL, &len, NULL, 0) == -1) {
1160 		log_warn("sysctl");
1161 		return (-1);
1162 	}
1163 	if ((buf = malloc(len)) == NULL) {
1164 		log_warn("fetchtable");
1165 		return (-1);
1166 	}
1167 	if (sysctl(mib, 7, buf, &len, NULL, 0) == -1) {
1168 		log_warn("sysctl");
1169 		free(buf);
1170 		return (-1);
1171 	}
1172 
1173 	lim = buf + len;
1174 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1175 		rtm = (struct rt_msghdr *)next;
1176 		if (rtm->rtm_version != RTM_VERSION)
1177 			continue;
1178 		sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
1179 		get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
1180 
1181 		if ((sa = rti_info[RTAX_DST]) == NULL)
1182 			continue;
1183 
1184 		if (rtm->rtm_flags & RTF_LLINFO)	/* arp cache */
1185 			continue;
1186 
1187 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
1188 			log_warn("fetchtable");
1189 			free(buf);
1190 			return (-1);
1191 		}
1192 
1193 		kr->r.flags = F_KERNEL;
1194 
1195 		switch (sa->sa_family) {
1196 		case AF_INET6:
1197 			kr->r.prefix =
1198 			    ((struct sockaddr_in6 *)sa)->sin6_addr;
1199 			sa_in6 = (struct sockaddr_in6 *)rti_info[RTAX_NETMASK];
1200 			if (rtm->rtm_flags & RTF_STATIC)
1201 				kr->r.flags |= F_STATIC;
1202 			if (rtm->rtm_flags & RTF_DYNAMIC)
1203 				kr->r.flags |= F_DYNAMIC;
1204 			if (rtm->rtm_flags & RTF_PROTO1)
1205 				kr->r.flags |= F_BGPD_INSERTED;
1206 			if (sa_in6 != NULL) {
1207 				if (sa_in6->sin6_len == 0)
1208 					break;
1209 				kr->r.prefixlen =
1210 				    mask2prefixlen(sa_in6);
1211 			} else if (rtm->rtm_flags & RTF_HOST)
1212 				kr->r.prefixlen = 128;
1213 			else
1214 				fatalx("classful IPv6 route?!!");
1215 			break;
1216 		default:
1217 			free(kr);
1218 			continue;
1219 		}
1220 
1221 		kr->r.ifindex = rtm->rtm_index;
1222 		if ((sa = rti_info[RTAX_GATEWAY]) != NULL)
1223 			switch (sa->sa_family) {
1224 			case AF_INET6:
1225 				kr->r.nexthop =
1226 				    ((struct sockaddr_in6 *)sa)->sin6_addr;
1227 				kr->r.scope =
1228 				    ((struct sockaddr_in6 *)sa)->sin6_scope_id;
1229 				break;
1230 			case AF_LINK:
1231 				kr->r.flags |= F_CONNECTED;
1232 				break;
1233 			}
1234 
1235 		if (rtm->rtm_flags & RTF_PROTO2)  {
1236 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
1237 			free(kr);
1238 		} else {
1239 			if ((label = (struct sockaddr_rtlabel *)
1240 			    rti_info[RTAX_LABEL]) != NULL) {
1241 				kr->r.rtlabel =
1242 				    rtlabel_name2id(label->sr_label);
1243 				kr->r.ext_tag =
1244 				    rtlabel_id2tag(kr->r.rtlabel);
1245 			}
1246 			kroute_insert(kr);
1247 		}
1248 
1249 	}
1250 	free(buf);
1251 	return (0);
1252 }
1253 
1254 int
1255 fetchifs(u_short ifindex)
1256 {
1257 	size_t			 len;
1258 	int			 mib[6];
1259 	char			*buf, *next, *lim;
1260 	struct rt_msghdr	*rtm;
1261 	struct if_msghdr	 ifm;
1262 	struct ifa_msghdr	*ifam;
1263 	struct iface		*iface;
1264 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1265 
1266 	mib[0] = CTL_NET;
1267 	mib[1] = AF_ROUTE;
1268 	mib[2] = 0;
1269 	mib[3] = AF_INET6;
1270 	mib[4] = NET_RT_IFLIST;
1271 	mib[5] = ifindex;
1272 
1273 	if (sysctl(mib, 6, NULL, &len, NULL, 0) == -1) {
1274 		log_warn("sysctl");
1275 		return (-1);
1276 	}
1277 	if ((buf = malloc(len)) == NULL) {
1278 		log_warn("fetchifs");
1279 		return (-1);
1280 	}
1281 	if (sysctl(mib, 6, buf, &len, NULL, 0) == -1) {
1282 		log_warn("sysctl");
1283 		free(buf);
1284 		return (-1);
1285 	}
1286 
1287 	lim = buf + len;
1288 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1289 		rtm = (struct rt_msghdr *)next;
1290 		if (rtm->rtm_version != RTM_VERSION)
1291 			continue;
1292 		switch (rtm->rtm_type) {
1293 		case RTM_IFINFO:
1294 			bcopy(rtm, &ifm, sizeof ifm);
1295 			sa = (struct sockaddr *)(next + sizeof(ifm));
1296 			get_rtaddrs(ifm.ifm_addrs, sa, rti_info);
1297 
1298 			if ((iface = kif_update(ifm.ifm_index,
1299 			    ifm.ifm_flags, &ifm.ifm_data,
1300 			    (struct sockaddr_dl *)rti_info[RTAX_IFP])) == NULL)
1301 			break;
1302 		case RTM_NEWADDR:
1303 			ifam = (struct ifa_msghdr *)rtm;
1304 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1305 			    RTA_BRD)) == 0)
1306 				break;
1307 			sa = (struct sockaddr *)(ifam + 1);
1308 			get_rtaddrs(ifam->ifam_addrs, sa, rti_info);
1309 
1310 			if_newaddr(ifam->ifam_index,
1311 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1312 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1313 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1314 			break;
1315 		}
1316 	}
1317 	free(buf);
1318 	return (0);
1319 }
1320 
1321 int
1322 dispatch_rtmsg(void)
1323 {
1324 	char			 buf[RT_BUF_SIZE];
1325 	ssize_t			 n;
1326 	char			*next, *lim;
1327 	struct rt_msghdr	*rtm;
1328 	struct if_msghdr	 ifm;
1329 	struct ifa_msghdr	*ifam;
1330 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1331 	struct sockaddr_in6	*sa_in6;
1332 	struct sockaddr_rtlabel	*label;
1333 	struct kroute_node	*kr, *okr;
1334 	struct in6_addr		 prefix, nexthop;
1335 	u_int8_t		 prefixlen;
1336 	int			 flags, mpath;
1337 	unsigned int		 scope;
1338 	u_short			 ifindex = 0;
1339 
1340 	if ((n = read(kr_state.fd, &buf, sizeof(buf))) == -1) {
1341 		log_warn("dispatch_rtmsg: read error");
1342 		return (-1);
1343 	}
1344 
1345 	if (n == 0) {
1346 		log_warnx("routing socket closed");
1347 		return (-1);
1348 	}
1349 
1350 	lim = buf + n;
1351 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1352 		rtm = (struct rt_msghdr *)next;
1353 		if (rtm->rtm_version != RTM_VERSION)
1354 			continue;
1355 
1356 		bzero(&prefix, sizeof(prefix));
1357 		bzero(&nexthop, sizeof(nexthop));
1358 		scope = 0;
1359 		prefixlen = 0;
1360 		flags = F_KERNEL;
1361 		mpath = 0;
1362 
1363 		if (rtm->rtm_type == RTM_ADD || rtm->rtm_type == RTM_CHANGE ||
1364 		    rtm->rtm_type == RTM_DELETE) {
1365 			sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
1366 			get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
1367 
1368 			if (rtm->rtm_tableid != 0)
1369 				continue;
1370 
1371 			if (rtm->rtm_pid == kr_state.pid) /* caused by us */
1372 				continue;
1373 
1374 			if (rtm->rtm_errno)		/* failed attempts... */
1375 				continue;
1376 
1377 			if (rtm->rtm_flags & RTF_LLINFO)	/* arp cache */
1378 				continue;
1379 
1380 #ifdef RTF_MPATH
1381 			if (rtm->rtm_flags & RTF_MPATH)
1382 				mpath = 1;
1383 #endif
1384 			switch (sa->sa_family) {
1385 			case AF_INET6:
1386 				prefix =
1387 				    ((struct sockaddr_in6 *)sa)->sin6_addr;
1388 				sa_in6 = (struct sockaddr_in6 *)
1389 				    rti_info[RTAX_NETMASK];
1390 				if (sa_in6 != NULL) {
1391 					if (sa_in6->sin6_len != 0)
1392 						prefixlen = mask2prefixlen(
1393 						    sa_in6);
1394 				} else if (rtm->rtm_flags & RTF_HOST)
1395 					prefixlen = 128;
1396 				else
1397 					fatalx("classful IPv6 address?!!");
1398 				if (rtm->rtm_flags & RTF_STATIC)
1399 					flags |= F_STATIC;
1400 				if (rtm->rtm_flags & RTF_DYNAMIC)
1401 					flags |= F_DYNAMIC;
1402 				if (rtm->rtm_flags & RTF_PROTO1)
1403 					flags |= F_BGPD_INSERTED;
1404 				break;
1405 			default:
1406 				continue;
1407 			}
1408 
1409 			ifindex = rtm->rtm_index;
1410 			if ((sa = rti_info[RTAX_GATEWAY]) != NULL) {
1411 				switch (sa->sa_family) {
1412 				case AF_INET6:
1413 					nexthop = ((struct sockaddr_in6 *)
1414 					    sa)->sin6_addr;
1415 					scope = ((struct sockaddr_in6 *)
1416 					    sa)->sin6_scope_id;
1417 					break;
1418 				case AF_LINK:
1419 					flags |= F_CONNECTED;
1420 					break;
1421 				}
1422 			}
1423 		}
1424 
1425 		switch (rtm->rtm_type) {
1426 		case RTM_ADD:
1427 		case RTM_CHANGE:
1428 			if (IN6_IS_ADDR_UNSPECIFIED(&nexthop) &&
1429 			    !(flags & F_CONNECTED)) {
1430 				log_warnx("dispatch_rtmsg no nexthop for %s/%u",
1431 				    log_in6addr(&prefix), prefixlen);
1432 				continue;
1433 			}
1434 
1435 			if ((okr = kroute_find(&prefix, prefixlen)) !=
1436 			    NULL) {
1437 				/* just add new multipath routes */
1438 				if (mpath && rtm->rtm_type == RTM_ADD)
1439 					goto add;
1440 				/* get the correct route */
1441 				kr = okr;
1442 				if (mpath && (kr = kroute_matchgw(okr,
1443 				    &nexthop, scope)) == NULL) {
1444 					log_warnx("dispatch_rtmsg mpath route"
1445 					    " not found");
1446 					/* add routes we missed out earlier */
1447 					goto add;
1448 				}
1449 
1450 				/*
1451 				 * ospf route overridden by kernel. Preference
1452 				 * of the route is not checked because this is
1453 				 * forced -- most probably by a user.
1454 				 */
1455 				if (kr->r.flags & F_OSPFD_INSERTED)
1456 					flags |= F_OSPFD_INSERTED;
1457 				if (kr->r.flags & F_REDISTRIBUTED)
1458 					flags |= F_REDISTRIBUTED;
1459 				kr->r.nexthop = nexthop;
1460 				kr->r.scope = scope;
1461 				kr->r.flags = flags;
1462 				kr->r.ifindex = ifindex;
1463 
1464 				rtlabel_unref(kr->r.rtlabel);
1465 				kr->r.rtlabel = 0;
1466 				kr->r.ext_tag = 0;
1467 				if ((label = (struct sockaddr_rtlabel *)
1468 				    rti_info[RTAX_LABEL]) != NULL) {
1469 					kr->r.rtlabel =
1470 					    rtlabel_name2id(label->sr_label);
1471 					kr->r.ext_tag =
1472 					    rtlabel_id2tag(kr->r.rtlabel);
1473 				}
1474 
1475 				if (kif_validate(kr->r.ifindex))
1476 					kr->r.flags &= ~F_DOWN;
1477 				else
1478 					kr->r.flags |= F_DOWN;
1479 
1480 				/* just readd, the RDE will care */
1481 				kr_redistribute(okr);
1482 			} else {
1483 add:
1484 				if ((kr = calloc(1,
1485 				    sizeof(struct kroute_node))) == NULL) {
1486 					log_warn("dispatch_rtmsg");
1487 					return (-1);
1488 				}
1489 				kr->r.prefix = prefix;
1490 				kr->r.prefixlen = prefixlen;
1491 				kr->r.nexthop = nexthop;
1492 				kr->r.scope = scope;
1493 				kr->r.flags = flags;
1494 				kr->r.ifindex = ifindex;
1495 
1496 				if ((label = (struct sockaddr_rtlabel *)
1497 				    rti_info[RTAX_LABEL]) != NULL) {
1498 					kr->r.rtlabel =
1499 					    rtlabel_name2id(label->sr_label);
1500 					kr->r.ext_tag =
1501 					    rtlabel_id2tag(kr->r.rtlabel);
1502 				}
1503 
1504 				kroute_insert(kr);
1505 			}
1506 			break;
1507 		case RTM_DELETE:
1508 			if ((kr = kroute_find(&prefix, prefixlen)) ==
1509 			    NULL)
1510 				continue;
1511 			if (!(kr->r.flags & F_KERNEL))
1512 				continue;
1513 			/* get the correct route */
1514 			okr = kr;
1515 			if (mpath && (kr = kroute_matchgw(kr, &nexthop,
1516 			    scope)) == NULL) {
1517 				log_warnx("dispatch_rtmsg mpath route"
1518 				    " not found");
1519 				return (-1);
1520 			}
1521 			/*
1522 			 * last route is getting removed request the
1523 			 * ospf route from the RDE to insert instead
1524 			 */
1525 			if (okr == kr && kr->next == NULL &&
1526 			    kr->r.flags & F_OSPFD_INSERTED)
1527 				main_imsg_compose_rde(IMSG_KROUTE_GET, 0,
1528 				    &kr->r, sizeof(struct kroute));
1529 			if (kroute_remove(kr) == -1)
1530 				return (-1);
1531 			break;
1532 		case RTM_IFINFO:
1533 			memcpy(&ifm, next, sizeof(ifm));
1534 			if_change(ifm.ifm_index, ifm.ifm_flags,
1535 			    &ifm.ifm_data);
1536 			break;
1537 		case RTM_NEWADDR:
1538 			ifam = (struct ifa_msghdr *)rtm;
1539 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1540 			    RTA_BRD)) == 0)
1541 				break;
1542 			sa = (struct sockaddr *)(ifam + 1);
1543 			get_rtaddrs(ifam->ifam_addrs, sa, rti_info);
1544 
1545 			if_newaddr(ifam->ifam_index,
1546 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1547 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1548 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1549 			break;
1550 		case RTM_DELADDR:
1551 			ifam = (struct ifa_msghdr *)rtm;
1552 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1553 			    RTA_BRD)) == 0)
1554 				break;
1555 			sa = (struct sockaddr *)(ifam + 1);
1556 			get_rtaddrs(ifam->ifam_addrs, sa, rti_info);
1557 
1558 			if_deladdr(ifam->ifam_index,
1559 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1560 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1561 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1562 			break;
1563 		case RTM_IFANNOUNCE:
1564 			if_announce(next);
1565 			break;
1566 		default:
1567 			/* ignore for now */
1568 			break;
1569 		}
1570 	}
1571 	return (0);
1572 }
1573