xref: /openbsd-src/usr.sbin/ospf6d/kroute.c (revision 50b7afb2c2c0993b0894d4e34bf857cb13ed9c80)
1 /*	$OpenBSD: kroute.c,v 1.44 2014/06/23 03:46:17 guenther 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 	 * unless it is a reject or blackhole route.
398 	 */
399 	if (IN6_IS_ADDR_LOOPBACK(&kr->nexthop) &&
400 	    !(kr->flags & (F_BLACKHOLE|F_REJECT)))
401 		goto dont_redistribute;
402 
403 	/* Should we redistribute this route? */
404 	if (!ospf_redistribute(kr, &metric))
405 		goto dont_redistribute;
406 
407 	/* prefix should be redistributed */
408 	kr->flags |= F_REDISTRIBUTED;
409 	/*
410 	 * only one of all multipath routes can be redistributed so
411 	 * redistribute the best one.
412 	 */
413 	if (rr->metric > metric) {
414 		rr->kr = *kr;
415 		rr->metric = metric;
416 	}
417 
418 	return (1);
419 
420 dont_redistribute:
421 	/* was the route redistributed? */
422 	if ((kr->flags & F_REDISTRIBUTED) == 0)
423 		return (0);
424 
425 	kr->flags &= ~F_REDISTRIBUTED;
426 	return (1);
427 }
428 
429 void
430 kr_redistribute(struct kroute_node *kh)
431 {
432 	struct kroute_node	*kn;
433 	struct rroute		 rr;
434 	int			 redistribute = 0;
435 
436 	bzero(&rr, sizeof(rr));
437 	rr.metric = UINT_MAX;
438 	for (kn = kh; kn; kn = kn->next)
439 		if (kr_redist_eval(&kn->r, &rr))
440 			redistribute = 1;
441 
442 	if (!redistribute)
443 		return;
444 
445 	if (rr.kr.flags & F_REDISTRIBUTED) {
446 		main_imsg_compose_rde(IMSG_NETWORK_ADD, 0, &rr,
447 		    sizeof(struct rroute));
448 	} else {
449 		rr.metric = DEFAULT_REDIST_METRIC;	/* some dummy value */
450 		rr.kr = kh->r;
451 		main_imsg_compose_rde(IMSG_NETWORK_DEL, 0, &rr,
452 		    sizeof(struct rroute));
453 	}
454 }
455 
456 void
457 kr_reload(void)
458 {
459 	struct kroute_node	*kr, *kn;
460 	u_int32_t		 dummy;
461 	int			 r;
462 
463 	RB_FOREACH(kr, kroute_tree, &krt) {
464 		for (kn = kr; kn; kn = kn->next) {
465 			r = ospf_redistribute(&kn->r, &dummy);
466 			/*
467 			 * if it is redistributed, redistribute again metric
468 			 * may have changed.
469 			 */
470 			if ((kn->r.flags & F_REDISTRIBUTED && !r) || r)
471 				break;
472 		}
473 		if (kn) {
474 			/*
475 			 * kr_redistribute copes with removes and RDE with
476 			 * duplicates
477 			 */
478 			kr_redistribute(kr);
479 		}
480 	}
481 }
482 
483 /* rb-tree compare */
484 int
485 kroute_compare(struct kroute_node *a, struct kroute_node *b)
486 {
487 	int	i;
488 
489 	/* XXX maybe switch a & b */
490 	i = memcmp(&a->r.prefix, &b->r.prefix, sizeof(a->r.prefix));
491 	if (i)
492 		return (i);
493 	if (a->r.prefixlen < b->r.prefixlen)
494 		return (-1);
495 	if (a->r.prefixlen > b->r.prefixlen)
496 		return (1);
497 	return (0);
498 }
499 
500 /* tree management */
501 struct kroute_node *
502 kroute_find(const struct in6_addr *prefix, u_int8_t prefixlen)
503 {
504 	struct kroute_node	s;
505 
506 	s.r.prefix = *prefix;
507 	s.r.prefixlen = prefixlen;
508 
509 	return (RB_FIND(kroute_tree, &krt, &s));
510 }
511 
512 struct kroute_node *
513 kroute_matchgw(struct kroute_node *kr, struct in6_addr *nh, unsigned int scope)
514 {
515 	while (kr) {
516 		if (scope == kr->r.scope &&
517 		    IN6_ARE_ADDR_EQUAL(&kr->r.nexthop, nh))
518 			return (kr);
519 		kr = kr->next;
520 	}
521 
522 	return (NULL);
523 }
524 
525 int
526 kroute_insert(struct kroute_node *kr)
527 {
528 	struct kroute_node	*krm, *krh;
529 
530 	if ((krh = RB_INSERT(kroute_tree, &krt, kr)) != NULL) {
531 		/*
532 		 * Multipath route, add at end of list and clone the
533 		 * ospfd inserted flag.
534 		 */
535 		krm = krh;
536 		kr->r.flags |= krm->r.flags & F_OSPFD_INSERTED;
537 		while (krm->next != NULL)
538 			krm = krm->next;
539 		krm->next = kr;
540 		kr->next = NULL; /* to be sure */
541 	} else
542 		krh = kr;
543 
544 	if (!(kr->r.flags & F_KERNEL)) {
545 		/* don't validate or redistribute ospf route */
546 		kr->r.flags &= ~F_DOWN;
547 		return (0);
548 	}
549 
550 	if (kif_validate(kr->r.ifindex))
551 		kr->r.flags &= ~F_DOWN;
552 	else
553 		kr->r.flags |= F_DOWN;
554 
555 	kr_redistribute(krh);
556 	return (0);
557 }
558 
559 int
560 kroute_remove(struct kroute_node *kr)
561 {
562 	struct kroute_node	*krm;
563 
564 	if ((krm = RB_FIND(kroute_tree, &krt, kr)) == NULL) {
565 		log_warnx("kroute_remove failed to find %s/%u",
566 		    log_in6addr(&kr->r.prefix), kr->r.prefixlen);
567 		return (-1);
568 	}
569 
570 	if (krm == kr) {
571 		/* head element */
572 		if (RB_REMOVE(kroute_tree, &krt, kr) == NULL) {
573 			log_warnx("kroute_remove failed for %s/%u",
574 			    log_in6addr(&kr->r.prefix), kr->r.prefixlen);
575 			return (-1);
576 		}
577 		if (kr->next != NULL) {
578 			if (RB_INSERT(kroute_tree, &krt, kr->next) != NULL) {
579 				log_warnx("kroute_remove failed to add %s/%u",
580 				    log_in6addr(&kr->r.prefix),
581 				    kr->r.prefixlen);
582 				return (-1);
583 			}
584 		}
585 	} else {
586 		/* somewhere in the list */
587 		while (krm->next != kr && krm->next != NULL)
588 			krm = krm->next;
589 		if (krm->next == NULL) {
590 			log_warnx("kroute_remove multipath list corrupted "
591 			    "for %s/%u", log_in6addr(&kr->r.prefix),
592 			    kr->r.prefixlen);
593 			return (-1);
594 		}
595 		krm->next = kr->next;
596 	}
597 
598 	kr_redist_remove(krm, kr);
599 	rtlabel_unref(kr->r.rtlabel);
600 
601 	free(kr);
602 	return (0);
603 }
604 
605 void
606 kroute_clear(void)
607 {
608 	struct kroute_node	*kr;
609 
610 	while ((kr = RB_MIN(kroute_tree, &krt)) != NULL)
611 		kroute_remove(kr);
612 }
613 
614 struct iface *
615 kif_update(u_short ifindex, int flags, struct if_data *ifd,
616     struct sockaddr_dl *sdl)
617 {
618 	struct iface	*iface;
619 	char		 ifname[IF_NAMESIZE];
620 
621 	if ((iface = if_find(ifindex)) == NULL) {
622 		bzero(ifname, sizeof(ifname));
623 		if (sdl && sdl->sdl_family == AF_LINK) {
624 			if (sdl->sdl_nlen >= sizeof(ifname))
625 				memcpy(ifname, sdl->sdl_data,
626 				    sizeof(ifname) - 1);
627 			else if (sdl->sdl_nlen > 0)
628 				memcpy(ifname, sdl->sdl_data, sdl->sdl_nlen);
629 			else
630 				return (NULL);
631 		} else
632 			return (NULL);
633 		if ((iface = if_new(ifindex, ifname)) == NULL)
634 			return (NULL);
635 		iface->cflags |= F_IFACE_AVAIL;
636 	}
637 
638 	if_update(iface, ifd->ifi_mtu, flags, ifd->ifi_type,
639 	    ifd->ifi_link_state, ifd->ifi_baudrate);
640 
641 	return (iface);
642 }
643 
644 int
645 kif_validate(u_short ifindex)
646 {
647 	struct iface	*iface;
648 
649 	if ((iface = if_find(ifindex)) == NULL) {
650 		log_warnx("interface with index %u not found", ifindex);
651 		return (-1);
652 	}
653 
654 	return ((iface->flags & IFF_UP) && LINK_STATE_IS_UP(iface->linkstate));
655 }
656 
657 struct kroute_node *
658 kroute_match(struct in6_addr *key)
659 {
660 	int			 i;
661 	struct kroute_node	*kr;
662 	struct in6_addr		 ina;
663 
664 	/* we will never match the default route */
665 	for (i = 128; i > 0; i--) {
666 		inet6applymask(&ina, key, i);
667 		if ((kr = kroute_find(&ina, i)) != NULL)
668 			return (kr);
669 	}
670 
671 	/* if we don't have a match yet, try to find a default route */
672 	if ((kr = kroute_find(&in6addr_any, 0)) != NULL)
673 			return (kr);
674 
675 	return (NULL);
676 }
677 
678 /* misc */
679 int
680 protect_lo(void)
681 {
682 	struct kroute_node	*kr;
683 
684 	/* special protection for loopback */
685 	if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
686 		log_warn("protect_lo");
687 		return (-1);
688 	}
689 	memcpy(&kr->r.prefix, &in6addr_loopback, sizeof(kr->r.prefix));
690 	kr->r.prefixlen = 128;
691 	kr->r.flags = F_KERNEL|F_CONNECTED;
692 
693 	if (RB_INSERT(kroute_tree, &krt, kr) != NULL)
694 		free(kr);	/* kernel route already there, no problem */
695 
696 	return (0);
697 }
698 
699 #define ROUNDUP(a) \
700 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
701 
702 void
703 get_rtaddrs(int addrs, struct sockaddr *sa, struct sockaddr **rti_info)
704 {
705 	int	i;
706 
707 	for (i = 0; i < RTAX_MAX; i++) {
708 		if (addrs & (1 << i)) {
709 			rti_info[i] = sa;
710 			sa = (struct sockaddr *)((char *)(sa) +
711 			    ROUNDUP(sa->sa_len));
712 		} else
713 			rti_info[i] = NULL;
714 	}
715 }
716 
717 void
718 if_change(u_short ifindex, int flags, struct if_data *ifd)
719 {
720 	struct kroute_node	*kr, *tkr;
721 	struct iface		*iface;
722 	u_int8_t		 wasvalid, isvalid;
723 
724 	wasvalid = kif_validate(ifindex);
725 
726 	if ((iface = kif_update(ifindex, flags, ifd, NULL)) == NULL) {
727 		log_warn("if_change: kif_update(%u)", ifindex);
728 		return;
729 	}
730 
731 	isvalid = (iface->flags & IFF_UP) &&
732 	    LINK_STATE_IS_UP(iface->linkstate);
733 
734 	if (wasvalid == isvalid)
735 		return;		/* nothing changed wrt validity */
736 
737 	/* inform engine and rde about state change if interface is used */
738 	if (iface->cflags & F_IFACE_CONFIGURED) {
739 		main_imsg_compose_ospfe(IMSG_IFINFO, 0, iface,
740 		    sizeof(struct iface));
741 		main_imsg_compose_rde(IMSG_IFINFO, 0, iface,
742 		    sizeof(struct iface));
743 	}
744 
745 	/* update redistribute list */
746 	RB_FOREACH(kr, kroute_tree, &krt) {
747 		for (tkr = kr; tkr != NULL; tkr = tkr->next) {
748 			if (tkr->r.ifindex == ifindex) {
749 				if (isvalid)
750 					tkr->r.flags &= ~F_DOWN;
751 				else
752 					tkr->r.flags |= F_DOWN;
753 
754 			}
755 		}
756 		kr_redistribute(kr);
757 	}
758 }
759 
760 void
761 if_newaddr(u_short ifindex, struct sockaddr_in6 *ifa, struct sockaddr_in6 *mask,
762     struct sockaddr_in6 *brd)
763 {
764 	struct iface		*iface;
765 	struct iface_addr	*ia;
766 	struct ifaddrchange	 ifc;
767 
768 	if (ifa == NULL || ifa->sin6_family != AF_INET6)
769 		return;
770 	if ((iface = if_find(ifindex)) == NULL) {
771 		log_warnx("if_newaddr: corresponding if %d not found", ifindex);
772 		return;
773 	}
774 
775 	/* We only care about link-local and global-scope. */
776 	if (IN6_IS_ADDR_UNSPECIFIED(&ifa->sin6_addr) ||
777 	    IN6_IS_ADDR_LOOPBACK(&ifa->sin6_addr) ||
778 	    IN6_IS_ADDR_MULTICAST(&ifa->sin6_addr) ||
779 	    IN6_IS_ADDR_SITELOCAL(&ifa->sin6_addr) ||
780 	    IN6_IS_ADDR_V4MAPPED(&ifa->sin6_addr) ||
781 	    IN6_IS_ADDR_V4COMPAT(&ifa->sin6_addr))
782 		return;
783 
784 	clearscope(&ifa->sin6_addr);
785 
786 	if (IN6_IS_ADDR_LINKLOCAL(&ifa->sin6_addr) ||
787 	    iface->flags & IFF_LOOPBACK)
788 		iface->addr = ifa->sin6_addr;
789 
790 	if ((ia = calloc(1, sizeof(struct iface_addr))) == NULL)
791 		fatal("if_newaddr");
792 
793 	ia->addr = ifa->sin6_addr;
794 
795 	if (mask)
796 		ia->prefixlen = mask2prefixlen(mask);
797 	else
798 		ia->prefixlen = 0;
799 	if (brd && brd->sin6_family == AF_INET6)
800 		ia->dstbrd = brd->sin6_addr;
801 	else
802 		bzero(&ia->dstbrd, sizeof(ia->dstbrd));
803 
804 	switch (iface->type) {
805 	case IF_TYPE_BROADCAST:
806 	case IF_TYPE_NBMA:
807 		log_debug("if_newaddr: ifindex %u, addr %s/%d",
808 		    ifindex, log_in6addr(&ia->addr), ia->prefixlen);
809 		break;
810 	case IF_TYPE_VIRTUALLINK:	/* FIXME */
811 		break;
812 	case IF_TYPE_POINTOPOINT:
813 	case IF_TYPE_POINTOMULTIPOINT:
814 		log_debug("if_newaddr: ifindex %u, addr %s/%d, "
815 		    "dest %s", ifindex, log_in6addr(&ia->addr),
816 		    ia->prefixlen, log_in6addr(&ia->dstbrd));
817 		break;
818 	default:
819 		fatalx("if_newaddr: unknown interface type");
820 	}
821 
822 	TAILQ_INSERT_TAIL(&iface->ifa_list, ia, entry);
823 	/* inform engine and rde if interface is used */
824 	if (iface->cflags & F_IFACE_CONFIGURED) {
825 		ifc.addr = ia->addr;
826 		ifc.dstbrd = ia->dstbrd;
827 		ifc.prefixlen = ia->prefixlen;
828 		ifc.ifindex = ifindex;
829 		main_imsg_compose_ospfe(IMSG_IFADDRNEW, 0, &ifc, sizeof(ifc));
830 		main_imsg_compose_rde(IMSG_IFADDRNEW, 0, &ifc, sizeof(ifc));
831 	}
832 }
833 
834 void
835 if_deladdr(u_short ifindex, struct sockaddr_in6 *ifa, struct sockaddr_in6 *mask,
836     struct sockaddr_in6 *brd)
837 {
838 	struct iface		*iface;
839 	struct iface_addr	*ia, *nia;
840 	struct ifaddrchange	 ifc;
841 
842 	if (ifa == NULL || ifa->sin6_family != AF_INET6)
843 		return;
844 	if ((iface = if_find(ifindex)) == NULL) {
845 		log_warnx("if_deladdr: corresponding if %d not found", ifindex);
846 		return;
847 	}
848 
849 	/* We only care about link-local and global-scope. */
850 	if (IN6_IS_ADDR_UNSPECIFIED(&ifa->sin6_addr) ||
851 	    IN6_IS_ADDR_LOOPBACK(&ifa->sin6_addr) ||
852 	    IN6_IS_ADDR_MULTICAST(&ifa->sin6_addr) ||
853 	    IN6_IS_ADDR_SITELOCAL(&ifa->sin6_addr) ||
854 	    IN6_IS_ADDR_V4MAPPED(&ifa->sin6_addr) ||
855 	    IN6_IS_ADDR_V4COMPAT(&ifa->sin6_addr))
856 		return;
857 
858 	clearscope(&ifa->sin6_addr);
859 
860 	for (ia = TAILQ_FIRST(&iface->ifa_list); ia != NULL; ia = nia) {
861 		nia = TAILQ_NEXT(ia, entry);
862 
863 		if (IN6_ARE_ADDR_EQUAL(&ia->addr, &ifa->sin6_addr)) {
864 			log_debug("if_deladdr: ifindex %u, addr %s/%d",
865 			    ifindex, log_in6addr(&ia->addr), ia->prefixlen);
866 			TAILQ_REMOVE(&iface->ifa_list, ia, entry);
867 			/* inform engine and rde if interface is used */
868 			if (iface->cflags & F_IFACE_CONFIGURED) {
869 				ifc.addr = ia->addr;
870 				ifc.dstbrd = ia->dstbrd;
871 				ifc.prefixlen = ia->prefixlen;
872 				ifc.ifindex = ifindex;
873 				main_imsg_compose_ospfe(IMSG_IFADDRDEL, 0, &ifc,
874 				    sizeof(ifc));
875 				main_imsg_compose_rde(IMSG_IFADDRDEL, 0, &ifc,
876 				    sizeof(ifc));
877 			}
878 			free(ia);
879 			return;
880 		}
881 	}
882 }
883 
884 void
885 if_announce(void *msg)
886 {
887 	struct if_announcemsghdr	*ifan;
888 	struct iface			*iface;
889 
890 	ifan = msg;
891 
892 	switch (ifan->ifan_what) {
893 	case IFAN_ARRIVAL:
894 		if ((iface = if_new(ifan->ifan_index, ifan->ifan_name)) == NULL)
895 			fatal("if_announce failed");
896 		iface->cflags |= F_IFACE_AVAIL;
897 		break;
898 	case IFAN_DEPARTURE:
899 		iface = if_find(ifan->ifan_index);
900 		if (iface->cflags & F_IFACE_CONFIGURED) {
901 			main_imsg_compose_rde(IMSG_IFDELETE, 0,
902 			    &iface->ifindex, sizeof(iface->ifindex));
903 			main_imsg_compose_ospfe(IMSG_IFDELETE, 0,
904 			    &iface->ifindex, sizeof(iface->ifindex));
905 		}
906 		if_del(iface);
907 		break;
908 	}
909 }
910 
911 /* rtsock */
912 int
913 send_rtmsg(int fd, int action, struct kroute *kroute)
914 {
915 	struct iovec		iov[5];
916 	struct rt_msghdr	hdr;
917 	struct pad {
918 		struct sockaddr_in6	addr;
919 		char			pad[sizeof(long)]; /* thank you IPv6 */
920 	} prefix, nexthop, mask;
921 	struct {
922 		struct sockaddr_dl	addr;
923 		char			pad[sizeof(long)];
924 	} ifp;
925 	struct sockaddr_rtlabel	sa_rl;
926 	int			iovcnt = 0;
927 	const char		*label;
928 
929 	if (kr_state.fib_sync == 0)
930 		return (0);
931 
932 	/* initialize header */
933 	bzero(&hdr, sizeof(hdr));
934 	hdr.rtm_version = RTM_VERSION;
935 	hdr.rtm_type = action;
936 	hdr.rtm_flags = RTF_UP;
937 	hdr.rtm_priority = RTP_OSPF;
938 	if (action == RTM_CHANGE)
939 		hdr.rtm_fmask = RTF_REJECT|RTF_BLACKHOLE;
940 	hdr.rtm_seq = kr_state.rtseq++;	/* overflow doesn't matter */
941 	hdr.rtm_hdrlen = sizeof(hdr);
942 	hdr.rtm_msglen = sizeof(hdr);
943 	/* adjust iovec */
944 	iov[iovcnt].iov_base = &hdr;
945 	iov[iovcnt++].iov_len = sizeof(hdr);
946 
947 	bzero(&prefix, sizeof(prefix));
948 	prefix.addr.sin6_len = sizeof(struct sockaddr_in6);
949 	prefix.addr.sin6_family = AF_INET6;
950 	prefix.addr.sin6_addr = kroute->prefix;
951 	/* adjust header */
952 	hdr.rtm_addrs |= RTA_DST;
953 	hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
954 	/* adjust iovec */
955 	iov[iovcnt].iov_base = &prefix;
956 	iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
957 
958 	if (!IN6_IS_ADDR_UNSPECIFIED(&kroute->nexthop)) {
959 		bzero(&nexthop, sizeof(nexthop));
960 		nexthop.addr.sin6_len = sizeof(struct sockaddr_in6);
961 		nexthop.addr.sin6_family = AF_INET6;
962 		nexthop.addr.sin6_addr = kroute->nexthop;
963 		nexthop.addr.sin6_scope_id = kroute->scope;
964 		/*
965 		 * XXX we should set the sin6_scope_id but the kernel
966 		 * XXX does not expect it that way. It must be fiddled
967 		 * XXX into the sin6_addr. Welcome to the typical
968 		 * XXX IPv6 insanity and all without wine bottles.
969 		 */
970 		embedscope(&nexthop.addr);
971 
972 		/* adjust header */
973 		hdr.rtm_flags |= RTF_GATEWAY;
974 		hdr.rtm_addrs |= RTA_GATEWAY;
975 		hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
976 		/* adjust iovec */
977 		iov[iovcnt].iov_base = &nexthop;
978 		iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
979 	} else if (kroute->ifindex) {
980 		/*
981 		 * We don't have an interface address in that network,
982 		 * so we install a cloning route.  The kernel will then
983 		 * do neigbor discovery.
984 		 */
985 		bzero(&ifp, sizeof(ifp));
986 		ifp.addr.sdl_len = sizeof(struct sockaddr_dl);
987 		ifp.addr.sdl_family = AF_LINK;
988 
989 		ifp.addr.sdl_index  = kroute->ifindex;
990 		/* adjust header */
991 		hdr.rtm_flags |= RTF_CLONING;
992 		hdr.rtm_addrs |= RTA_GATEWAY;
993 		hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_dl));
994 		/* adjust iovec */
995 		iov[iovcnt].iov_base = &ifp;
996 		iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_dl));
997 	}
998 
999 	bzero(&mask, sizeof(mask));
1000 	mask.addr.sin6_len = sizeof(struct sockaddr_in6);
1001 	mask.addr.sin6_family = AF_INET6;
1002 	mask.addr.sin6_addr = *prefixlen2mask(kroute->prefixlen);
1003 	/* adjust header */
1004 	if (kroute->prefixlen == 128)
1005 		hdr.rtm_flags |= RTF_HOST;
1006 	hdr.rtm_addrs |= RTA_NETMASK;
1007 	hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
1008 	/* adjust iovec */
1009 	iov[iovcnt].iov_base = &mask;
1010 	iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
1011 
1012 	if (kroute->rtlabel != 0) {
1013 		sa_rl.sr_len = sizeof(sa_rl);
1014 		sa_rl.sr_family = AF_UNSPEC;
1015 		label = rtlabel_id2name(kroute->rtlabel);
1016 		if (strlcpy(sa_rl.sr_label, label,
1017 		    sizeof(sa_rl.sr_label)) >= sizeof(sa_rl.sr_label)) {
1018 			log_warnx("send_rtmsg: invalid rtlabel");
1019 			return (-1);
1020 		}
1021 		/* adjust header */
1022 		hdr.rtm_addrs |= RTA_LABEL;
1023 		hdr.rtm_msglen += sizeof(sa_rl);
1024 		/* adjust iovec */
1025 		iov[iovcnt].iov_base = &sa_rl;
1026 		iov[iovcnt++].iov_len = sizeof(sa_rl);
1027 	}
1028 
1029 retry:
1030 	if (writev(fd, iov, iovcnt) == -1) {
1031 		if (errno == ESRCH) {
1032 			if (hdr.rtm_type == RTM_CHANGE) {
1033 				hdr.rtm_type = RTM_ADD;
1034 				goto retry;
1035 			} else if (hdr.rtm_type == RTM_DELETE) {
1036 				log_info("route %s/%u vanished before delete",
1037 				    log_sockaddr(&prefix), kroute->prefixlen);
1038 				return (0);
1039 			}
1040 		}
1041 		log_warn("send_rtmsg: action %u, prefix %s/%u", hdr.rtm_type,
1042 		    log_sockaddr(&prefix), kroute->prefixlen);
1043 		return (0);
1044 	}
1045 
1046 	return (0);
1047 }
1048 
1049 int
1050 fetchtable(void)
1051 {
1052 	size_t			 len;
1053 	int			 mib[7];
1054 	char			*buf, *next, *lim;
1055 	struct rt_msghdr	*rtm;
1056 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1057 	struct sockaddr_in6	*sa_in6;
1058 	struct sockaddr_rtlabel	*label;
1059 	struct kroute_node	*kr;
1060 
1061 	mib[0] = CTL_NET;
1062 	mib[1] = PF_ROUTE;
1063 	mib[2] = 0;
1064 	mib[3] = AF_INET6;
1065 	mib[4] = NET_RT_DUMP;
1066 	mib[5] = 0;
1067 	mib[6] = 0;	/* rtableid */
1068 
1069 	if (sysctl(mib, 7, NULL, &len, NULL, 0) == -1) {
1070 		log_warn("sysctl");
1071 		return (-1);
1072 	}
1073 	if ((buf = malloc(len)) == NULL) {
1074 		log_warn("fetchtable");
1075 		return (-1);
1076 	}
1077 	if (sysctl(mib, 7, buf, &len, NULL, 0) == -1) {
1078 		log_warn("sysctl");
1079 		free(buf);
1080 		return (-1);
1081 	}
1082 
1083 	lim = buf + len;
1084 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1085 		rtm = (struct rt_msghdr *)next;
1086 		if (rtm->rtm_version != RTM_VERSION)
1087 			continue;
1088 		sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
1089 		get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
1090 
1091 		if ((sa = rti_info[RTAX_DST]) == NULL)
1092 			continue;
1093 
1094 		if (rtm->rtm_flags & RTF_LLINFO)	/* arp cache */
1095 			continue;
1096 
1097 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
1098 			log_warn("fetchtable");
1099 			free(buf);
1100 			return (-1);
1101 		}
1102 
1103 		kr->r.flags = F_KERNEL;
1104 
1105 		switch (sa->sa_family) {
1106 		case AF_INET6:
1107 			kr->r.prefix =
1108 			    ((struct sockaddr_in6 *)sa)->sin6_addr;
1109 			sa_in6 = (struct sockaddr_in6 *)rti_info[RTAX_NETMASK];
1110 			if (rtm->rtm_flags & RTF_STATIC)
1111 				kr->r.flags |= F_STATIC;
1112 			if (rtm->rtm_flags & RTF_BLACKHOLE)
1113 				kr->r.flags |= F_BLACKHOLE;
1114 			if (rtm->rtm_flags & RTF_REJECT)
1115 				kr->r.flags |= F_REJECT;
1116 			if (rtm->rtm_flags & RTF_DYNAMIC)
1117 				kr->r.flags |= F_DYNAMIC;
1118 			if (rtm->rtm_flags & RTF_PROTO1)
1119 				kr->r.flags |= F_BGPD_INSERTED;
1120 			if (sa_in6 != NULL) {
1121 				if (sa_in6->sin6_len == 0)
1122 					break;
1123 				kr->r.prefixlen =
1124 				    mask2prefixlen(sa_in6);
1125 			} else if (rtm->rtm_flags & RTF_HOST)
1126 				kr->r.prefixlen = 128;
1127 			else
1128 				fatalx("classful IPv6 route?!!");
1129 			break;
1130 		default:
1131 			free(kr);
1132 			continue;
1133 		}
1134 
1135 		kr->r.ifindex = rtm->rtm_index;
1136 		if ((sa = rti_info[RTAX_GATEWAY]) != NULL)
1137 			switch (sa->sa_family) {
1138 			case AF_INET6:
1139 				sa_in6 = (struct sockaddr_in6 *)sa;
1140 				/*
1141 				 * XXX The kernel provides the scope via the
1142 				 * XXX kame hack instead of the scope_id field.
1143 				 */
1144 				recoverscope(sa_in6);
1145 				kr->r.nexthop = sa_in6->sin6_addr;
1146 				kr->r.scope = sa_in6->sin6_scope_id;
1147 				break;
1148 			case AF_LINK:
1149 				kr->r.flags |= F_CONNECTED;
1150 				break;
1151 			}
1152 
1153 		if (rtm->rtm_flags & RTF_PROTO2)  {
1154 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
1155 			free(kr);
1156 		} else {
1157 			if ((label = (struct sockaddr_rtlabel *)
1158 			    rti_info[RTAX_LABEL]) != NULL) {
1159 				kr->r.rtlabel =
1160 				    rtlabel_name2id(label->sr_label);
1161 				kr->r.ext_tag =
1162 				    rtlabel_id2tag(kr->r.rtlabel);
1163 			}
1164 			kroute_insert(kr);
1165 		}
1166 
1167 	}
1168 	free(buf);
1169 	return (0);
1170 }
1171 
1172 int
1173 fetchifs(u_short ifindex)
1174 {
1175 	size_t			 len;
1176 	int			 mib[6];
1177 	char			*buf, *next, *lim;
1178 	struct rt_msghdr	*rtm;
1179 	struct if_msghdr	 ifm;
1180 	struct ifa_msghdr	*ifam;
1181 	struct iface		*iface;
1182 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
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_IFLIST;
1189 	mib[5] = ifindex;
1190 
1191 	if (sysctl(mib, 6, NULL, &len, NULL, 0) == -1) {
1192 		log_warn("sysctl");
1193 		return (-1);
1194 	}
1195 	if ((buf = malloc(len)) == NULL) {
1196 		log_warn("fetchifs");
1197 		return (-1);
1198 	}
1199 	if (sysctl(mib, 6, buf, &len, NULL, 0) == -1) {
1200 		log_warn("sysctl");
1201 		free(buf);
1202 		return (-1);
1203 	}
1204 
1205 	lim = buf + len;
1206 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1207 		rtm = (struct rt_msghdr *)next;
1208 		if (rtm->rtm_version != RTM_VERSION)
1209 			continue;
1210 		switch (rtm->rtm_type) {
1211 		case RTM_IFINFO:
1212 			bcopy(rtm, &ifm, sizeof ifm);
1213 			sa = (struct sockaddr *)(next + sizeof(ifm));
1214 			get_rtaddrs(ifm.ifm_addrs, sa, rti_info);
1215 
1216 			if ((iface = kif_update(ifm.ifm_index,
1217 			    ifm.ifm_flags, &ifm.ifm_data,
1218 			    (struct sockaddr_dl *)rti_info[RTAX_IFP])) == NULL)
1219 			break;
1220 		case RTM_NEWADDR:
1221 			ifam = (struct ifa_msghdr *)rtm;
1222 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1223 			    RTA_BRD)) == 0)
1224 				break;
1225 			sa = (struct sockaddr *)(ifam + 1);
1226 			get_rtaddrs(ifam->ifam_addrs, sa, rti_info);
1227 
1228 			if_newaddr(ifam->ifam_index,
1229 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1230 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1231 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1232 			break;
1233 		}
1234 	}
1235 	free(buf);
1236 	return (0);
1237 }
1238 
1239 int
1240 dispatch_rtmsg(void)
1241 {
1242 	char			 buf[RT_BUF_SIZE];
1243 	ssize_t			 n;
1244 	char			*next, *lim;
1245 	struct rt_msghdr	*rtm;
1246 	struct if_msghdr	 ifm;
1247 	struct ifa_msghdr	*ifam;
1248 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1249 	struct sockaddr_in6	*sa_in6;
1250 	struct sockaddr_rtlabel	*label;
1251 	struct kroute_node	*kr, *okr;
1252 	struct in6_addr		 prefix, nexthop;
1253 	u_int8_t		 prefixlen;
1254 	int			 flags, mpath;
1255 	unsigned int		 scope;
1256 	u_short			 ifindex = 0;
1257 
1258 	if ((n = read(kr_state.fd, &buf, sizeof(buf))) == -1) {
1259 		log_warn("dispatch_rtmsg: read error");
1260 		return (-1);
1261 	}
1262 
1263 	if (n == 0) {
1264 		log_warnx("routing socket closed");
1265 		return (-1);
1266 	}
1267 
1268 	lim = buf + n;
1269 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1270 		rtm = (struct rt_msghdr *)next;
1271 		if (rtm->rtm_version != RTM_VERSION)
1272 			continue;
1273 
1274 		bzero(&prefix, sizeof(prefix));
1275 		bzero(&nexthop, sizeof(nexthop));
1276 		scope = 0;
1277 		prefixlen = 0;
1278 		flags = F_KERNEL;
1279 		mpath = 0;
1280 
1281 		if (rtm->rtm_type == RTM_ADD || rtm->rtm_type == RTM_CHANGE ||
1282 		    rtm->rtm_type == RTM_DELETE) {
1283 			sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
1284 			get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
1285 
1286 			if (rtm->rtm_tableid != 0)
1287 				continue;
1288 
1289 			if (rtm->rtm_pid == kr_state.pid) /* caused by us */
1290 				continue;
1291 
1292 			if (rtm->rtm_errno)		/* failed attempts... */
1293 				continue;
1294 
1295 			if (rtm->rtm_flags & RTF_LLINFO)	/* arp cache */
1296 				continue;
1297 
1298 #ifdef RTF_MPATH
1299 			if (rtm->rtm_flags & RTF_MPATH)
1300 				mpath = 1;
1301 #endif
1302 			switch (sa->sa_family) {
1303 			case AF_INET6:
1304 				prefix =
1305 				    ((struct sockaddr_in6 *)sa)->sin6_addr;
1306 				sa_in6 = (struct sockaddr_in6 *)
1307 				    rti_info[RTAX_NETMASK];
1308 				if (sa_in6 != NULL) {
1309 					if (sa_in6->sin6_len != 0)
1310 						prefixlen = mask2prefixlen(
1311 						    sa_in6);
1312 				} else if (rtm->rtm_flags & RTF_HOST)
1313 					prefixlen = 128;
1314 				else
1315 					fatalx("classful IPv6 address?!!");
1316 				if (rtm->rtm_flags & RTF_STATIC)
1317 					flags |= F_STATIC;
1318 				if (rtm->rtm_flags & RTF_BLACKHOLE)
1319 					flags |= F_BLACKHOLE;
1320 				if (rtm->rtm_flags & RTF_REJECT)
1321 					flags |= F_REJECT;
1322 				if (rtm->rtm_flags & RTF_DYNAMIC)
1323 					flags |= F_DYNAMIC;
1324 				if (rtm->rtm_flags & RTF_PROTO1)
1325 					flags |= F_BGPD_INSERTED;
1326 				break;
1327 			default:
1328 				continue;
1329 			}
1330 
1331 			ifindex = rtm->rtm_index;
1332 			if ((sa = rti_info[RTAX_GATEWAY]) != NULL) {
1333 				switch (sa->sa_family) {
1334 				case AF_INET6:
1335 					sa_in6 = (struct sockaddr_in6 *)sa;
1336 					/*
1337 					 * XXX The kernel provides the scope
1338 					 * XXX via the kame hack instead of
1339 					 * XXX the scope_id field.
1340 					 */
1341 					recoverscope(sa_in6);
1342 					nexthop = sa_in6->sin6_addr;
1343 					scope = sa_in6->sin6_scope_id;
1344 					break;
1345 				case AF_LINK:
1346 					flags |= F_CONNECTED;
1347 					break;
1348 				}
1349 			}
1350 		}
1351 
1352 		switch (rtm->rtm_type) {
1353 		case RTM_ADD:
1354 		case RTM_CHANGE:
1355 			if (IN6_IS_ADDR_UNSPECIFIED(&nexthop) &&
1356 			    !(flags & F_CONNECTED)) {
1357 				log_warnx("dispatch_rtmsg no nexthop for %s/%u",
1358 				    log_in6addr(&prefix), prefixlen);
1359 				continue;
1360 			}
1361 
1362 			if ((okr = kroute_find(&prefix, prefixlen)) !=
1363 			    NULL) {
1364 				/* just add new multipath routes */
1365 				if (mpath && rtm->rtm_type == RTM_ADD)
1366 					goto add;
1367 				/* get the correct route */
1368 				kr = okr;
1369 				if (mpath && (kr = kroute_matchgw(okr,
1370 				    &nexthop, scope)) == NULL) {
1371 					log_warnx("dispatch_rtmsg mpath route"
1372 					    " not found");
1373 					/* add routes we missed out earlier */
1374 					goto add;
1375 				}
1376 
1377 				/*
1378 				 * ospf route overridden by kernel. Preference
1379 				 * of the route is not checked because this is
1380 				 * forced -- most probably by a user.
1381 				 */
1382 				if (kr->r.flags & F_OSPFD_INSERTED)
1383 					flags |= F_OSPFD_INSERTED;
1384 				if (kr->r.flags & F_REDISTRIBUTED)
1385 					flags |= F_REDISTRIBUTED;
1386 				kr->r.nexthop = nexthop;
1387 				kr->r.scope = scope;
1388 				kr->r.flags = flags;
1389 				kr->r.ifindex = ifindex;
1390 
1391 				rtlabel_unref(kr->r.rtlabel);
1392 				kr->r.rtlabel = 0;
1393 				kr->r.ext_tag = 0;
1394 				if ((label = (struct sockaddr_rtlabel *)
1395 				    rti_info[RTAX_LABEL]) != NULL) {
1396 					kr->r.rtlabel =
1397 					    rtlabel_name2id(label->sr_label);
1398 					kr->r.ext_tag =
1399 					    rtlabel_id2tag(kr->r.rtlabel);
1400 				}
1401 
1402 				if (kif_validate(kr->r.ifindex))
1403 					kr->r.flags &= ~F_DOWN;
1404 				else
1405 					kr->r.flags |= F_DOWN;
1406 
1407 				/* just readd, the RDE will care */
1408 				kr_redistribute(okr);
1409 			} else {
1410 add:
1411 				if ((kr = calloc(1,
1412 				    sizeof(struct kroute_node))) == NULL) {
1413 					log_warn("dispatch_rtmsg");
1414 					return (-1);
1415 				}
1416 				kr->r.prefix = prefix;
1417 				kr->r.prefixlen = prefixlen;
1418 				kr->r.nexthop = nexthop;
1419 				kr->r.scope = scope;
1420 				kr->r.flags = flags;
1421 				kr->r.ifindex = ifindex;
1422 
1423 				if ((label = (struct sockaddr_rtlabel *)
1424 				    rti_info[RTAX_LABEL]) != NULL) {
1425 					kr->r.rtlabel =
1426 					    rtlabel_name2id(label->sr_label);
1427 					kr->r.ext_tag =
1428 					    rtlabel_id2tag(kr->r.rtlabel);
1429 				}
1430 
1431 				kroute_insert(kr);
1432 			}
1433 			break;
1434 		case RTM_DELETE:
1435 			if ((kr = kroute_find(&prefix, prefixlen)) ==
1436 			    NULL)
1437 				continue;
1438 			if (!(kr->r.flags & F_KERNEL))
1439 				continue;
1440 			/* get the correct route */
1441 			okr = kr;
1442 			if (mpath && (kr = kroute_matchgw(kr, &nexthop,
1443 			    scope)) == NULL) {
1444 				log_warnx("dispatch_rtmsg mpath route"
1445 				    " not found");
1446 				return (-1);
1447 			}
1448 			/*
1449 			 * last route is getting removed request the
1450 			 * ospf route from the RDE to insert instead
1451 			 */
1452 			if (okr == kr && kr->next == NULL &&
1453 			    kr->r.flags & F_OSPFD_INSERTED)
1454 				main_imsg_compose_rde(IMSG_KROUTE_GET, 0,
1455 				    &kr->r, sizeof(struct kroute));
1456 			if (kroute_remove(kr) == -1)
1457 				return (-1);
1458 			break;
1459 		case RTM_IFINFO:
1460 			memcpy(&ifm, next, sizeof(ifm));
1461 			if_change(ifm.ifm_index, ifm.ifm_flags,
1462 			    &ifm.ifm_data);
1463 			break;
1464 		case RTM_NEWADDR:
1465 			ifam = (struct ifa_msghdr *)rtm;
1466 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1467 			    RTA_BRD)) == 0)
1468 				break;
1469 			sa = (struct sockaddr *)(ifam + 1);
1470 			get_rtaddrs(ifam->ifam_addrs, sa, rti_info);
1471 
1472 			if_newaddr(ifam->ifam_index,
1473 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1474 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1475 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1476 			break;
1477 		case RTM_DELADDR:
1478 			ifam = (struct ifa_msghdr *)rtm;
1479 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1480 			    RTA_BRD)) == 0)
1481 				break;
1482 			sa = (struct sockaddr *)(ifam + 1);
1483 			get_rtaddrs(ifam->ifam_addrs, sa, rti_info);
1484 
1485 			if_deladdr(ifam->ifam_index,
1486 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1487 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1488 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1489 			break;
1490 		case RTM_IFANNOUNCE:
1491 			if_announce(next);
1492 			break;
1493 		default:
1494 			/* ignore for now */
1495 			break;
1496 		}
1497 	}
1498 	return (0);
1499 }
1500