xref: /openbsd-src/usr.sbin/ospf6d/kroute.c (revision f2da64fbbbf1b03f09f390ab01267c93dfd77c4c)
1 /*	$OpenBSD: kroute.c,v 1.48 2015/07/17 20:12:38 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/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 	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,
102 	    SOCK_RAW | SOCK_CLOEXEC | SOCK_NONBLOCK, 0)) == -1) {
103 		log_warn("kr_init: socket");
104 		return (-1);
105 	}
106 
107 	/* not interested in my own messages */
108 	if (setsockopt(kr_state.fd, SOL_SOCKET, SO_USELOOPBACK,
109 	    &opt, sizeof(opt)) == -1)
110 		log_warn("kr_init: setsockopt");	/* not fatal */
111 
112 	/* grow receive buffer, don't wanna miss messages */
113 	optlen = sizeof(default_rcvbuf);
114 	if (getsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF,
115 	    &default_rcvbuf, &optlen) == -1)
116 		log_warn("kr_init getsockopt SOL_SOCKET SO_RCVBUF");
117 	else
118 		for (rcvbuf = MAX_RTSOCK_BUF;
119 		    rcvbuf > default_rcvbuf &&
120 		    setsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF,
121 		    &rcvbuf, sizeof(rcvbuf)) == -1 && errno == ENOBUFS;
122 		    rcvbuf /= 2)
123 			;	/* nothing */
124 
125 	kr_state.pid = getpid();
126 	kr_state.rtseq = 1;
127 
128 	RB_INIT(&krt);
129 
130 	if (fetchtable() == -1)
131 		return (-1);
132 
133 	if (protect_lo() == -1)
134 		return (-1);
135 
136 	event_set(&kr_state.ev, kr_state.fd, EV_READ | EV_PERSIST,
137 	    kr_dispatch_msg, NULL);
138 	event_add(&kr_state.ev, NULL);
139 
140 	return (0);
141 }
142 
143 int
144 kr_change(struct kroute *kroute)
145 {
146 	struct kroute_node	*kr;
147 	int			 action = RTM_ADD;
148 
149 	kroute->rtlabel = rtlabel_tag2id(kroute->ext_tag);
150 
151 	if ((kr = kroute_find(&kroute->prefix, kroute->prefixlen)) !=
152 	    NULL) {
153 		if (!(kr->r.flags & F_KERNEL))
154 			action = RTM_CHANGE;
155 		else {	/* a non-ospf route already exists. not a problem */
156 			if (!(kr->r.flags & F_BGPD_INSERTED)) {
157 				do {
158 					kr->r.flags |= F_OSPFD_INSERTED;
159 					kr = kr->next;
160 				} while (kr);
161 				return (0);
162 			}
163 			/*
164 			 * XXX as long as there is no multipath support in
165 			 * bgpd this is safe else we end up in a bad situation.
166 			 */
167 			/*
168 			 * ospf route has higher pref
169 			 * - reset flags to the ospf ones
170 			 * - use RTM_CHANGE
171 			 * - zero out ifindex (this is no longer relevant)
172 			 */
173 			action = RTM_CHANGE;
174 			kr->r.flags = kroute->flags | F_OSPFD_INSERTED;
175 			kr->r.ifindex = 0;
176 			rtlabel_unref(kr->r.rtlabel);
177 			kr->r.ext_tag = kroute->ext_tag;
178 			kr->r.rtlabel = kroute->rtlabel;
179 		}
180 	}
181 
182 	/* nexthop within 127/8 -> ignore silently */
183 	if (kr && IN6_IS_ADDR_LOOPBACK(&kr->r.nexthop))
184 		return (0);
185 
186 	/*
187 	 * Ingnore updates that did not change the route.
188 	 * Currently only the nexthop can change.
189 	 */
190 	if (kr && kr->r.scope == kroute->scope &&
191 	    IN6_ARE_ADDR_EQUAL(&kr->r.nexthop, &kroute->nexthop))
192 		return (0);
193 
194 	if (send_rtmsg(kr_state.fd, action, kroute) == -1)
195 		return (-1);
196 
197 	if (action == RTM_ADD) {
198 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
199 			log_warn("kr_change");
200 			return (-1);
201 		}
202 		kr->r.prefix = kroute->prefix;
203 		kr->r.prefixlen = kroute->prefixlen;
204 		kr->r.nexthop = kroute->nexthop;
205 		kr->r.scope = kroute->scope;
206 		kr->r.flags = kroute->flags | F_OSPFD_INSERTED;
207 		kr->r.ext_tag = kroute->ext_tag;
208 		kr->r.rtlabel = kroute->rtlabel;
209 
210 		if (kroute_insert(kr) == -1)
211 			free(kr);
212 	} else if (kr) {
213 		kr->r.nexthop = kroute->nexthop;
214 		kr->r.scope = kroute->scope;
215 	}
216 
217 	return (0);
218 }
219 
220 int
221 kr_delete(struct kroute *kroute)
222 {
223 	struct kroute_node	*kr;
224 
225 	if ((kr = kroute_find(&kroute->prefix, kroute->prefixlen)) ==
226 	    NULL)
227 		return (0);
228 
229 	if (!(kr->r.flags & F_OSPFD_INSERTED))
230 		return (0);
231 
232 	if (kr->r.flags & F_KERNEL) {
233 		/* remove F_OSPFD_INSERTED flag, route still exists in kernel */
234 		do {
235 			kr->r.flags &= ~F_OSPFD_INSERTED;
236 			kr = kr->next;
237 		} while (kr);
238 		return (0);
239 	}
240 
241 	if (send_rtmsg(kr_state.fd, RTM_DELETE, kroute) == -1)
242 		return (-1);
243 
244 	if (kroute_remove(kr) == -1)
245 		return (-1);
246 
247 	return (0);
248 }
249 
250 void
251 kr_shutdown(void)
252 {
253 	kr_fib_decouple();
254 	kroute_clear();
255 }
256 
257 void
258 kr_fib_couple(void)
259 {
260 	struct kroute_node	*kr;
261 
262 	if (kr_state.fib_sync == 1)	/* already coupled */
263 		return;
264 
265 	kr_state.fib_sync = 1;
266 
267 	RB_FOREACH(kr, kroute_tree, &krt)
268 		if (!(kr->r.flags & F_KERNEL))
269 			send_rtmsg(kr_state.fd, RTM_ADD, &kr->r);
270 
271 	log_info("kernel routing table coupled");
272 }
273 
274 void
275 kr_fib_decouple(void)
276 {
277 	struct kroute_node	*kr;
278 
279 	if (kr_state.fib_sync == 0)	/* already decoupled */
280 		return;
281 
282 	RB_FOREACH(kr, kroute_tree, &krt)
283 		if (!(kr->r.flags & F_KERNEL))
284 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
285 
286 	kr_state.fib_sync = 0;
287 
288 	log_info("kernel routing table decoupled");
289 }
290 
291 /* ARGSUSED */
292 void
293 kr_dispatch_msg(int fd, short event, void *bula)
294 {
295 	dispatch_rtmsg();
296 }
297 
298 void
299 kr_show_route(struct imsg *imsg)
300 {
301 	struct kroute_node	*kr;
302 	struct kroute_node	*kn;
303 	int			 flags;
304 	struct in6_addr		 addr;
305 
306 	switch (imsg->hdr.type) {
307 	case IMSG_CTL_KROUTE:
308 		if (imsg->hdr.len != IMSG_HEADER_SIZE + sizeof(flags)) {
309 			log_warnx("kr_show_route: wrong imsg len");
310 			return;
311 		}
312 		memcpy(&flags, imsg->data, sizeof(flags));
313 		RB_FOREACH(kr, kroute_tree, &krt)
314 			if (!flags || kr->r.flags & flags) {
315 				kn = kr;
316 				do {
317 					main_imsg_compose_ospfe(IMSG_CTL_KROUTE,
318 					    imsg->hdr.pid,
319 					    &kn->r, sizeof(kn->r));
320 				} while ((kn = kn->next) != NULL);
321 			}
322 		break;
323 	case IMSG_CTL_KROUTE_ADDR:
324 		if (imsg->hdr.len != IMSG_HEADER_SIZE +
325 		    sizeof(struct in6_addr)) {
326 			log_warnx("kr_show_route: wrong imsg len");
327 			return;
328 		}
329 		memcpy(&addr, imsg->data, sizeof(addr));
330 		kr = NULL;
331 		kr = kroute_match(&addr);
332 		if (kr != NULL)
333 			main_imsg_compose_ospfe(IMSG_CTL_KROUTE, imsg->hdr.pid,
334 			    &kr->r, sizeof(kr->r));
335 		break;
336 	default:
337 		log_debug("kr_show_route: error handling imsg");
338 		break;
339 	}
340 
341 	main_imsg_compose_ospfe(IMSG_CTL_END, imsg->hdr.pid, NULL, 0);
342 }
343 
344 void
345 kr_redist_remove(struct kroute_node *kh, struct kroute_node *kn)
346 {
347 	struct rroute	 rr;
348 
349 	/* was the route redistributed? */
350 	if ((kn->r.flags & F_REDISTRIBUTED) == 0)
351 		return;
352 
353 	/* remove redistributed flag */
354 	kn->r.flags &= ~F_REDISTRIBUTED;
355 	rr.kr = kn->r;
356 	rr.metric = DEFAULT_REDIST_METRIC;	/* some dummy value */
357 
358 	/* probably inform the RDE (check if no other path is redistributed) */
359 	for (kn = kh; kn; kn = kn->next)
360 		if (kn->r.flags & F_REDISTRIBUTED)
361 			break;
362 
363 	if (kn == NULL)
364 		main_imsg_compose_rde(IMSG_NETWORK_DEL, 0, &rr,
365 		    sizeof(struct rroute));
366 }
367 
368 int
369 kr_redist_eval(struct kroute *kr, struct rroute *rr)
370 {
371 	u_int32_t	 metric = 0;
372 
373 	/* Only non-ospfd routes are considered for redistribution. */
374 	if (!(kr->flags & F_KERNEL))
375 		goto dont_redistribute;
376 
377 	/* Dynamic routes are not redistributable. */
378 	if (kr->flags & F_DYNAMIC)
379 		goto dont_redistribute;
380 
381 	/* interface is not up and running so don't announce */
382 	if (kr->flags & F_DOWN)
383 		goto dont_redistribute;
384 
385 	/*
386 	 * We consider loopback, multicast, link- and site-local,
387 	 * IPv4 mapped and IPv4 compatible addresses as not redistributable.
388 	 */
389 	if (IN6_IS_ADDR_LOOPBACK(&kr->prefix) ||
390 	    IN6_IS_ADDR_MULTICAST(&kr->prefix) ||
391 	    IN6_IS_ADDR_LINKLOCAL(&kr->prefix) ||
392 	    IN6_IS_ADDR_SITELOCAL(&kr->prefix) ||
393 	    IN6_IS_ADDR_V4MAPPED(&kr->prefix) ||
394 	    IN6_IS_ADDR_V4COMPAT(&kr->prefix))
395 		goto dont_redistribute;
396 	/*
397 	 * Consider networks with nexthop loopback as not redistributable
398 	 * unless it is a reject or blackhole route.
399 	 */
400 	if (IN6_IS_ADDR_LOOPBACK(&kr->nexthop) &&
401 	    !(kr->flags & (F_BLACKHOLE|F_REJECT)))
402 		goto dont_redistribute;
403 
404 	/* Should we redistribute this route? */
405 	if (!ospf_redistribute(kr, &metric))
406 		goto dont_redistribute;
407 
408 	/* prefix should be redistributed */
409 	kr->flags |= F_REDISTRIBUTED;
410 	/*
411 	 * only one of all multipath routes can be redistributed so
412 	 * redistribute the best one.
413 	 */
414 	if (rr->metric > metric) {
415 		rr->kr = *kr;
416 		rr->metric = metric;
417 	}
418 
419 	return (1);
420 
421 dont_redistribute:
422 	/* was the route redistributed? */
423 	if ((kr->flags & F_REDISTRIBUTED) == 0)
424 		return (0);
425 
426 	kr->flags &= ~F_REDISTRIBUTED;
427 	return (1);
428 }
429 
430 void
431 kr_redistribute(struct kroute_node *kh)
432 {
433 	struct kroute_node	*kn;
434 	struct rroute		 rr;
435 	int			 redistribute = 0;
436 
437 	bzero(&rr, sizeof(rr));
438 	rr.metric = UINT_MAX;
439 	for (kn = kh; kn; kn = kn->next)
440 		if (kr_redist_eval(&kn->r, &rr))
441 			redistribute = 1;
442 
443 	if (!redistribute)
444 		return;
445 
446 	if (rr.kr.flags & F_REDISTRIBUTED) {
447 		main_imsg_compose_rde(IMSG_NETWORK_ADD, 0, &rr,
448 		    sizeof(struct rroute));
449 	} else {
450 		rr.metric = DEFAULT_REDIST_METRIC;	/* some dummy value */
451 		rr.kr = kh->r;
452 		main_imsg_compose_rde(IMSG_NETWORK_DEL, 0, &rr,
453 		    sizeof(struct rroute));
454 	}
455 }
456 
457 void
458 kr_reload(void)
459 {
460 	struct kroute_node	*kr, *kn;
461 	u_int32_t		 dummy;
462 	int			 r;
463 
464 	RB_FOREACH(kr, kroute_tree, &krt) {
465 		for (kn = kr; kn; kn = kn->next) {
466 			r = ospf_redistribute(&kn->r, &dummy);
467 			/*
468 			 * if it is redistributed, redistribute again metric
469 			 * may have changed.
470 			 */
471 			if ((kn->r.flags & F_REDISTRIBUTED && !r) || r)
472 				break;
473 		}
474 		if (kn) {
475 			/*
476 			 * kr_redistribute copes with removes and RDE with
477 			 * duplicates
478 			 */
479 			kr_redistribute(kr);
480 		}
481 	}
482 }
483 
484 /* rb-tree compare */
485 int
486 kroute_compare(struct kroute_node *a, struct kroute_node *b)
487 {
488 	int	i;
489 
490 	/* XXX maybe switch a & b */
491 	i = memcmp(&a->r.prefix, &b->r.prefix, sizeof(a->r.prefix));
492 	if (i)
493 		return (i);
494 	if (a->r.prefixlen < b->r.prefixlen)
495 		return (-1);
496 	if (a->r.prefixlen > b->r.prefixlen)
497 		return (1);
498 	return (0);
499 }
500 
501 /* tree management */
502 struct kroute_node *
503 kroute_find(const struct in6_addr *prefix, u_int8_t prefixlen)
504 {
505 	struct kroute_node	s;
506 
507 	s.r.prefix = *prefix;
508 	s.r.prefixlen = prefixlen;
509 
510 	return (RB_FIND(kroute_tree, &krt, &s));
511 }
512 
513 struct kroute_node *
514 kroute_matchgw(struct kroute_node *kr, struct in6_addr *nh, unsigned int scope)
515 {
516 	while (kr) {
517 		if (scope == kr->r.scope &&
518 		    IN6_ARE_ADDR_EQUAL(&kr->r.nexthop, nh))
519 			return (kr);
520 		kr = kr->next;
521 	}
522 
523 	return (NULL);
524 }
525 
526 int
527 kroute_insert(struct kroute_node *kr)
528 {
529 	struct kroute_node	*krm, *krh;
530 
531 	if ((krh = RB_INSERT(kroute_tree, &krt, kr)) != NULL) {
532 		/*
533 		 * Multipath route, add at end of list and clone the
534 		 * ospfd inserted flag.
535 		 */
536 		krm = krh;
537 		kr->r.flags |= krm->r.flags & F_OSPFD_INSERTED;
538 		while (krm->next != NULL)
539 			krm = krm->next;
540 		krm->next = kr;
541 		kr->next = NULL; /* to be sure */
542 	} else
543 		krh = kr;
544 
545 	if (!(kr->r.flags & F_KERNEL)) {
546 		/* don't validate or redistribute ospf route */
547 		kr->r.flags &= ~F_DOWN;
548 		return (0);
549 	}
550 
551 	if (kif_validate(kr->r.ifindex))
552 		kr->r.flags &= ~F_DOWN;
553 	else
554 		kr->r.flags |= F_DOWN;
555 
556 	kr_redistribute(krh);
557 	return (0);
558 }
559 
560 int
561 kroute_remove(struct kroute_node *kr)
562 {
563 	struct kroute_node	*krm;
564 
565 	if ((krm = RB_FIND(kroute_tree, &krt, kr)) == NULL) {
566 		log_warnx("kroute_remove failed to find %s/%u",
567 		    log_in6addr(&kr->r.prefix), kr->r.prefixlen);
568 		return (-1);
569 	}
570 
571 	if (krm == kr) {
572 		/* head element */
573 		if (RB_REMOVE(kroute_tree, &krt, kr) == NULL) {
574 			log_warnx("kroute_remove failed for %s/%u",
575 			    log_in6addr(&kr->r.prefix), kr->r.prefixlen);
576 			return (-1);
577 		}
578 		if (kr->next != NULL) {
579 			if (RB_INSERT(kroute_tree, &krt, kr->next) != NULL) {
580 				log_warnx("kroute_remove failed to add %s/%u",
581 				    log_in6addr(&kr->r.prefix),
582 				    kr->r.prefixlen);
583 				return (-1);
584 			}
585 		}
586 	} else {
587 		/* somewhere in the list */
588 		while (krm->next != kr && krm->next != NULL)
589 			krm = krm->next;
590 		if (krm->next == NULL) {
591 			log_warnx("kroute_remove multipath list corrupted "
592 			    "for %s/%u", log_in6addr(&kr->r.prefix),
593 			    kr->r.prefixlen);
594 			return (-1);
595 		}
596 		krm->next = kr->next;
597 	}
598 
599 	kr_redist_remove(krm, kr);
600 	rtlabel_unref(kr->r.rtlabel);
601 
602 	free(kr);
603 	return (0);
604 }
605 
606 void
607 kroute_clear(void)
608 {
609 	struct kroute_node	*kr;
610 
611 	while ((kr = RB_MIN(kroute_tree, &krt)) != NULL)
612 		kroute_remove(kr);
613 }
614 
615 struct iface *
616 kif_update(u_short ifindex, int flags, struct if_data *ifd,
617     struct sockaddr_dl *sdl)
618 {
619 	struct iface	*iface;
620 	char		 ifname[IF_NAMESIZE];
621 
622 	if ((iface = if_find(ifindex)) == NULL) {
623 		bzero(ifname, sizeof(ifname));
624 		if (sdl && sdl->sdl_family == AF_LINK) {
625 			if (sdl->sdl_nlen >= sizeof(ifname))
626 				memcpy(ifname, sdl->sdl_data,
627 				    sizeof(ifname) - 1);
628 			else if (sdl->sdl_nlen > 0)
629 				memcpy(ifname, sdl->sdl_data, sdl->sdl_nlen);
630 			else
631 				return (NULL);
632 		} else
633 			return (NULL);
634 		if ((iface = if_new(ifindex, ifname)) == NULL)
635 			return (NULL);
636 		iface->cflags |= F_IFACE_AVAIL;
637 	}
638 
639 	if_update(iface, ifd->ifi_mtu, flags, ifd->ifi_type,
640 	    ifd->ifi_link_state, ifd->ifi_baudrate);
641 
642 	return (iface);
643 }
644 
645 int
646 kif_validate(u_short ifindex)
647 {
648 	struct iface	*iface;
649 
650 	if ((iface = if_find(ifindex)) == NULL) {
651 		log_warnx("interface with index %u not found", ifindex);
652 		return (-1);
653 	}
654 
655 	return ((iface->flags & IFF_UP) && LINK_STATE_IS_UP(iface->linkstate));
656 }
657 
658 struct kroute_node *
659 kroute_match(struct in6_addr *key)
660 {
661 	int			 i;
662 	struct kroute_node	*kr;
663 	struct in6_addr		 ina;
664 
665 	/* we will never match the default route */
666 	for (i = 128; i > 0; i--) {
667 		inet6applymask(&ina, key, i);
668 		if ((kr = kroute_find(&ina, i)) != NULL)
669 			return (kr);
670 	}
671 
672 	/* if we don't have a match yet, try to find a default route */
673 	if ((kr = kroute_find(&in6addr_any, 0)) != NULL)
674 			return (kr);
675 
676 	return (NULL);
677 }
678 
679 /* misc */
680 int
681 protect_lo(void)
682 {
683 	struct kroute_node	*kr;
684 
685 	/* special protection for loopback */
686 	if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
687 		log_warn("protect_lo");
688 		return (-1);
689 	}
690 	memcpy(&kr->r.prefix, &in6addr_loopback, sizeof(kr->r.prefix));
691 	kr->r.prefixlen = 128;
692 	kr->r.flags = F_KERNEL|F_CONNECTED;
693 
694 	if (RB_INSERT(kroute_tree, &krt, kr) != NULL)
695 		free(kr);	/* kernel route already there, no problem */
696 
697 	return (0);
698 }
699 
700 #define ROUNDUP(a) \
701 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
702 
703 void
704 get_rtaddrs(int addrs, struct sockaddr *sa, struct sockaddr **rti_info)
705 {
706 	int	i;
707 
708 	for (i = 0; i < RTAX_MAX; i++) {
709 		if (addrs & (1 << i)) {
710 			rti_info[i] = sa;
711 			sa = (struct sockaddr *)((char *)(sa) +
712 			    ROUNDUP(sa->sa_len));
713 		} else
714 			rti_info[i] = NULL;
715 	}
716 }
717 
718 void
719 if_change(u_short ifindex, int flags, struct if_data *ifd)
720 {
721 	struct kroute_node	*kr, *tkr;
722 	struct iface		*iface;
723 	u_int8_t		 wasvalid, isvalid;
724 
725 	wasvalid = kif_validate(ifindex);
726 
727 	if ((iface = kif_update(ifindex, flags, ifd, NULL)) == NULL) {
728 		log_warn("if_change: kif_update(%u)", ifindex);
729 		return;
730 	}
731 
732 	isvalid = (iface->flags & IFF_UP) &&
733 	    LINK_STATE_IS_UP(iface->linkstate);
734 
735 	if (wasvalid == isvalid)
736 		return;		/* nothing changed wrt validity */
737 
738 	/* inform engine and rde about state change if interface is used */
739 	if (iface->cflags & F_IFACE_CONFIGURED) {
740 		main_imsg_compose_ospfe(IMSG_IFINFO, 0, iface,
741 		    sizeof(struct iface));
742 		main_imsg_compose_rde(IMSG_IFINFO, 0, iface,
743 		    sizeof(struct iface));
744 	}
745 
746 	/* update redistribute list */
747 	RB_FOREACH(kr, kroute_tree, &krt) {
748 		for (tkr = kr; tkr != NULL; tkr = tkr->next) {
749 			if (tkr->r.ifindex == ifindex) {
750 				if (isvalid)
751 					tkr->r.flags &= ~F_DOWN;
752 				else
753 					tkr->r.flags |= F_DOWN;
754 
755 			}
756 		}
757 		kr_redistribute(kr);
758 	}
759 }
760 
761 void
762 if_newaddr(u_short ifindex, struct sockaddr_in6 *ifa, struct sockaddr_in6 *mask,
763     struct sockaddr_in6 *brd)
764 {
765 	struct iface		*iface;
766 	struct iface_addr	*ia;
767 	struct ifaddrchange	 ifc;
768 
769 	if (ifa == NULL || ifa->sin6_family != AF_INET6)
770 		return;
771 	if ((iface = if_find(ifindex)) == NULL) {
772 		log_warnx("if_newaddr: corresponding if %d not found", ifindex);
773 		return;
774 	}
775 
776 	/* We only care about link-local and global-scope. */
777 	if (IN6_IS_ADDR_UNSPECIFIED(&ifa->sin6_addr) ||
778 	    IN6_IS_ADDR_LOOPBACK(&ifa->sin6_addr) ||
779 	    IN6_IS_ADDR_MULTICAST(&ifa->sin6_addr) ||
780 	    IN6_IS_ADDR_SITELOCAL(&ifa->sin6_addr) ||
781 	    IN6_IS_ADDR_V4MAPPED(&ifa->sin6_addr) ||
782 	    IN6_IS_ADDR_V4COMPAT(&ifa->sin6_addr))
783 		return;
784 
785 	clearscope(&ifa->sin6_addr);
786 
787 	if (IN6_IS_ADDR_LINKLOCAL(&ifa->sin6_addr) ||
788 	    iface->flags & IFF_LOOPBACK)
789 		iface->addr = ifa->sin6_addr;
790 
791 	if ((ia = calloc(1, sizeof(struct iface_addr))) == NULL)
792 		fatal("if_newaddr");
793 
794 	ia->addr = ifa->sin6_addr;
795 
796 	if (mask)
797 		ia->prefixlen = mask2prefixlen(mask);
798 	else
799 		ia->prefixlen = 0;
800 	if (brd && brd->sin6_family == AF_INET6)
801 		ia->dstbrd = brd->sin6_addr;
802 	else
803 		bzero(&ia->dstbrd, sizeof(ia->dstbrd));
804 
805 	switch (iface->type) {
806 	case IF_TYPE_BROADCAST:
807 	case IF_TYPE_NBMA:
808 		log_debug("if_newaddr: ifindex %u, addr %s/%d",
809 		    ifindex, log_in6addr(&ia->addr), ia->prefixlen);
810 		break;
811 	case IF_TYPE_VIRTUALLINK:	/* FIXME */
812 		break;
813 	case IF_TYPE_POINTOPOINT:
814 	case IF_TYPE_POINTOMULTIPOINT:
815 		log_debug("if_newaddr: ifindex %u, addr %s/%d, "
816 		    "dest %s", ifindex, log_in6addr(&ia->addr),
817 		    ia->prefixlen, log_in6addr(&ia->dstbrd));
818 		break;
819 	default:
820 		fatalx("if_newaddr: unknown interface type");
821 	}
822 
823 	TAILQ_INSERT_TAIL(&iface->ifa_list, ia, entry);
824 	/* inform engine and rde if interface is used */
825 	if (iface->cflags & F_IFACE_CONFIGURED) {
826 		ifc.addr = ia->addr;
827 		ifc.dstbrd = ia->dstbrd;
828 		ifc.prefixlen = ia->prefixlen;
829 		ifc.ifindex = ifindex;
830 		main_imsg_compose_ospfe(IMSG_IFADDRNEW, 0, &ifc, sizeof(ifc));
831 		main_imsg_compose_rde(IMSG_IFADDRNEW, 0, &ifc, sizeof(ifc));
832 	}
833 }
834 
835 void
836 if_deladdr(u_short ifindex, struct sockaddr_in6 *ifa, struct sockaddr_in6 *mask,
837     struct sockaddr_in6 *brd)
838 {
839 	struct iface		*iface;
840 	struct iface_addr	*ia, *nia;
841 	struct ifaddrchange	 ifc;
842 
843 	if (ifa == NULL || ifa->sin6_family != AF_INET6)
844 		return;
845 	if ((iface = if_find(ifindex)) == NULL) {
846 		log_warnx("if_deladdr: corresponding if %d not found", ifindex);
847 		return;
848 	}
849 
850 	/* We only care about link-local and global-scope. */
851 	if (IN6_IS_ADDR_UNSPECIFIED(&ifa->sin6_addr) ||
852 	    IN6_IS_ADDR_LOOPBACK(&ifa->sin6_addr) ||
853 	    IN6_IS_ADDR_MULTICAST(&ifa->sin6_addr) ||
854 	    IN6_IS_ADDR_SITELOCAL(&ifa->sin6_addr) ||
855 	    IN6_IS_ADDR_V4MAPPED(&ifa->sin6_addr) ||
856 	    IN6_IS_ADDR_V4COMPAT(&ifa->sin6_addr))
857 		return;
858 
859 	clearscope(&ifa->sin6_addr);
860 
861 	for (ia = TAILQ_FIRST(&iface->ifa_list); ia != NULL; ia = nia) {
862 		nia = TAILQ_NEXT(ia, entry);
863 
864 		if (IN6_ARE_ADDR_EQUAL(&ia->addr, &ifa->sin6_addr)) {
865 			log_debug("if_deladdr: ifindex %u, addr %s/%d",
866 			    ifindex, log_in6addr(&ia->addr), ia->prefixlen);
867 			TAILQ_REMOVE(&iface->ifa_list, ia, entry);
868 			/* inform engine and rde if interface is used */
869 			if (iface->cflags & F_IFACE_CONFIGURED) {
870 				ifc.addr = ia->addr;
871 				ifc.dstbrd = ia->dstbrd;
872 				ifc.prefixlen = ia->prefixlen;
873 				ifc.ifindex = ifindex;
874 				main_imsg_compose_ospfe(IMSG_IFADDRDEL, 0, &ifc,
875 				    sizeof(ifc));
876 				main_imsg_compose_rde(IMSG_IFADDRDEL, 0, &ifc,
877 				    sizeof(ifc));
878 			}
879 			free(ia);
880 			return;
881 		}
882 	}
883 }
884 
885 void
886 if_announce(void *msg)
887 {
888 	struct if_announcemsghdr	*ifan;
889 	struct iface			*iface;
890 
891 	ifan = msg;
892 
893 	switch (ifan->ifan_what) {
894 	case IFAN_ARRIVAL:
895 		if ((iface = if_new(ifan->ifan_index, ifan->ifan_name)) == NULL)
896 			fatal("if_announce failed");
897 		iface->cflags |= F_IFACE_AVAIL;
898 		break;
899 	case IFAN_DEPARTURE:
900 		iface = if_find(ifan->ifan_index);
901 		if (iface->cflags & F_IFACE_CONFIGURED) {
902 			main_imsg_compose_rde(IMSG_IFDELETE, 0,
903 			    &iface->ifindex, sizeof(iface->ifindex));
904 			main_imsg_compose_ospfe(IMSG_IFDELETE, 0,
905 			    &iface->ifindex, sizeof(iface->ifindex));
906 		}
907 		if_del(iface);
908 		break;
909 	}
910 }
911 
912 /* rtsock */
913 int
914 send_rtmsg(int fd, int action, struct kroute *kroute)
915 {
916 	struct iovec		iov[5];
917 	struct rt_msghdr	hdr;
918 	struct pad {
919 		struct sockaddr_in6	addr;
920 		char			pad[sizeof(long)]; /* thank you IPv6 */
921 	} prefix, nexthop, mask;
922 	struct {
923 		struct sockaddr_dl	addr;
924 		char			pad[sizeof(long)];
925 	} ifp;
926 	struct sockaddr_rtlabel	sa_rl;
927 	int			iovcnt = 0;
928 	const char		*label;
929 
930 	if (kr_state.fib_sync == 0)
931 		return (0);
932 
933 	/* initialize header */
934 	bzero(&hdr, sizeof(hdr));
935 	hdr.rtm_version = RTM_VERSION;
936 	hdr.rtm_type = action;
937 	hdr.rtm_flags = RTF_UP;
938 	hdr.rtm_priority = RTP_OSPF;
939 	if (action == RTM_CHANGE)
940 		hdr.rtm_fmask = RTF_REJECT|RTF_BLACKHOLE;
941 	hdr.rtm_seq = kr_state.rtseq++;	/* overflow doesn't matter */
942 	hdr.rtm_hdrlen = sizeof(hdr);
943 	hdr.rtm_msglen = sizeof(hdr);
944 	/* adjust iovec */
945 	iov[iovcnt].iov_base = &hdr;
946 	iov[iovcnt++].iov_len = sizeof(hdr);
947 
948 	bzero(&prefix, sizeof(prefix));
949 	prefix.addr.sin6_len = sizeof(struct sockaddr_in6);
950 	prefix.addr.sin6_family = AF_INET6;
951 	prefix.addr.sin6_addr = kroute->prefix;
952 	/* adjust header */
953 	hdr.rtm_addrs |= RTA_DST;
954 	hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
955 	/* adjust iovec */
956 	iov[iovcnt].iov_base = &prefix;
957 	iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
958 
959 	if (!IN6_IS_ADDR_UNSPECIFIED(&kroute->nexthop)) {
960 		bzero(&nexthop, sizeof(nexthop));
961 		nexthop.addr.sin6_len = sizeof(struct sockaddr_in6);
962 		nexthop.addr.sin6_family = AF_INET6;
963 		nexthop.addr.sin6_addr = kroute->nexthop;
964 		nexthop.addr.sin6_scope_id = kroute->scope;
965 		/*
966 		 * XXX we should set the sin6_scope_id but the kernel
967 		 * XXX does not expect it that way. It must be fiddled
968 		 * XXX into the sin6_addr. Welcome to the typical
969 		 * XXX IPv6 insanity and all without wine bottles.
970 		 */
971 		embedscope(&nexthop.addr);
972 
973 		/* adjust header */
974 		hdr.rtm_flags |= RTF_GATEWAY;
975 		hdr.rtm_addrs |= RTA_GATEWAY;
976 		hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
977 		/* adjust iovec */
978 		iov[iovcnt].iov_base = &nexthop;
979 		iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
980 	} else if (kroute->ifindex) {
981 		/*
982 		 * We don't have an interface address in that network,
983 		 * so we install a cloning route.  The kernel will then
984 		 * do neigbor discovery.
985 		 */
986 		bzero(&ifp, sizeof(ifp));
987 		ifp.addr.sdl_len = sizeof(struct sockaddr_dl);
988 		ifp.addr.sdl_family = AF_LINK;
989 
990 		ifp.addr.sdl_index  = kroute->ifindex;
991 		/* adjust header */
992 		hdr.rtm_flags |= RTF_CLONING;
993 		hdr.rtm_addrs |= RTA_GATEWAY;
994 		hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_dl));
995 		/* adjust iovec */
996 		iov[iovcnt].iov_base = &ifp;
997 		iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_dl));
998 	}
999 
1000 	bzero(&mask, sizeof(mask));
1001 	mask.addr.sin6_len = sizeof(struct sockaddr_in6);
1002 	mask.addr.sin6_family = AF_INET6;
1003 	mask.addr.sin6_addr = *prefixlen2mask(kroute->prefixlen);
1004 	/* adjust header */
1005 	if (kroute->prefixlen == 128)
1006 		hdr.rtm_flags |= RTF_HOST;
1007 	hdr.rtm_addrs |= RTA_NETMASK;
1008 	hdr.rtm_msglen += ROUNDUP(sizeof(struct sockaddr_in6));
1009 	/* adjust iovec */
1010 	iov[iovcnt].iov_base = &mask;
1011 	iov[iovcnt++].iov_len = ROUNDUP(sizeof(struct sockaddr_in6));
1012 
1013 	if (kroute->rtlabel != 0) {
1014 		sa_rl.sr_len = sizeof(sa_rl);
1015 		sa_rl.sr_family = AF_UNSPEC;
1016 		label = rtlabel_id2name(kroute->rtlabel);
1017 		if (strlcpy(sa_rl.sr_label, label,
1018 		    sizeof(sa_rl.sr_label)) >= sizeof(sa_rl.sr_label)) {
1019 			log_warnx("send_rtmsg: invalid rtlabel");
1020 			return (-1);
1021 		}
1022 		/* adjust header */
1023 		hdr.rtm_addrs |= RTA_LABEL;
1024 		hdr.rtm_msglen += sizeof(sa_rl);
1025 		/* adjust iovec */
1026 		iov[iovcnt].iov_base = &sa_rl;
1027 		iov[iovcnt++].iov_len = sizeof(sa_rl);
1028 	}
1029 
1030 retry:
1031 	if (writev(fd, iov, iovcnt) == -1) {
1032 		if (errno == ESRCH) {
1033 			if (hdr.rtm_type == RTM_CHANGE) {
1034 				hdr.rtm_type = RTM_ADD;
1035 				goto retry;
1036 			} else if (hdr.rtm_type == RTM_DELETE) {
1037 				log_info("route %s/%u vanished before delete",
1038 				    log_sockaddr(&prefix), kroute->prefixlen);
1039 				return (0);
1040 			}
1041 		}
1042 		log_warn("send_rtmsg: action %u, prefix %s/%u", hdr.rtm_type,
1043 		    log_sockaddr(&prefix), kroute->prefixlen);
1044 		return (0);
1045 	}
1046 
1047 	return (0);
1048 }
1049 
1050 int
1051 fetchtable(void)
1052 {
1053 	size_t			 len;
1054 	int			 mib[7];
1055 	char			*buf, *next, *lim;
1056 	struct rt_msghdr	*rtm;
1057 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1058 	struct sockaddr_in6	*sa_in6;
1059 	struct sockaddr_rtlabel	*label;
1060 	struct kroute_node	*kr;
1061 
1062 	mib[0] = CTL_NET;
1063 	mib[1] = PF_ROUTE;
1064 	mib[2] = 0;
1065 	mib[3] = AF_INET6;
1066 	mib[4] = NET_RT_DUMP;
1067 	mib[5] = 0;
1068 	mib[6] = 0;	/* rtableid */
1069 
1070 	if (sysctl(mib, 7, NULL, &len, NULL, 0) == -1) {
1071 		log_warn("sysctl");
1072 		return (-1);
1073 	}
1074 	if ((buf = malloc(len)) == NULL) {
1075 		log_warn("fetchtable");
1076 		return (-1);
1077 	}
1078 	if (sysctl(mib, 7, buf, &len, NULL, 0) == -1) {
1079 		log_warn("sysctl");
1080 		free(buf);
1081 		return (-1);
1082 	}
1083 
1084 	lim = buf + len;
1085 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1086 		rtm = (struct rt_msghdr *)next;
1087 		if (rtm->rtm_version != RTM_VERSION)
1088 			continue;
1089 		sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
1090 		get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
1091 
1092 		if ((sa = rti_info[RTAX_DST]) == NULL)
1093 			continue;
1094 
1095 		/* Skip ARP/ND cache and broadcast routes. */
1096 		if (rtm->rtm_flags & (RTF_LLINFO|RTF_BROADCAST))
1097 			continue;
1098 
1099 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
1100 			log_warn("fetchtable");
1101 			free(buf);
1102 			return (-1);
1103 		}
1104 
1105 		kr->r.flags = F_KERNEL;
1106 
1107 		switch (sa->sa_family) {
1108 		case AF_INET6:
1109 			kr->r.prefix =
1110 			    ((struct sockaddr_in6 *)sa)->sin6_addr;
1111 			sa_in6 = (struct sockaddr_in6 *)rti_info[RTAX_NETMASK];
1112 			if (rtm->rtm_flags & RTF_STATIC)
1113 				kr->r.flags |= F_STATIC;
1114 			if (rtm->rtm_flags & RTF_BLACKHOLE)
1115 				kr->r.flags |= F_BLACKHOLE;
1116 			if (rtm->rtm_flags & RTF_REJECT)
1117 				kr->r.flags |= F_REJECT;
1118 			if (rtm->rtm_flags & RTF_DYNAMIC)
1119 				kr->r.flags |= F_DYNAMIC;
1120 			if (rtm->rtm_flags & RTF_PROTO1)
1121 				kr->r.flags |= F_BGPD_INSERTED;
1122 			if (sa_in6 != NULL) {
1123 				if (sa_in6->sin6_len == 0)
1124 					break;
1125 				kr->r.prefixlen =
1126 				    mask2prefixlen(sa_in6);
1127 			} else if (rtm->rtm_flags & RTF_HOST)
1128 				kr->r.prefixlen = 128;
1129 			else
1130 				fatalx("classful IPv6 route?!!");
1131 			break;
1132 		default:
1133 			free(kr);
1134 			continue;
1135 		}
1136 
1137 		kr->r.ifindex = rtm->rtm_index;
1138 		if ((sa = rti_info[RTAX_GATEWAY]) != NULL)
1139 			switch (sa->sa_family) {
1140 			case AF_INET6:
1141 				if (rtm->rtm_flags & RTF_CONNECTED) {
1142 					kr->r.flags |= F_CONNECTED;
1143 					break;
1144 				}
1145 
1146 				sa_in6 = (struct sockaddr_in6 *)sa;
1147 				/*
1148 				 * XXX The kernel provides the scope via the
1149 				 * XXX kame hack instead of the scope_id field.
1150 				 */
1151 				recoverscope(sa_in6);
1152 				kr->r.nexthop = sa_in6->sin6_addr;
1153 				kr->r.scope = sa_in6->sin6_scope_id;
1154 				break;
1155 			case AF_LINK:
1156 				/*
1157 				 * Traditional BSD connected routes have
1158 				 * a gateway of type AF_LINK.
1159 				 */
1160 				kr->r.flags |= F_CONNECTED;
1161 				break;
1162 			}
1163 
1164 		if (rtm->rtm_flags & RTF_PROTO2)  {
1165 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
1166 			free(kr);
1167 		} else {
1168 			if ((label = (struct sockaddr_rtlabel *)
1169 			    rti_info[RTAX_LABEL]) != NULL) {
1170 				kr->r.rtlabel =
1171 				    rtlabel_name2id(label->sr_label);
1172 				kr->r.ext_tag =
1173 				    rtlabel_id2tag(kr->r.rtlabel);
1174 			}
1175 			kroute_insert(kr);
1176 		}
1177 
1178 	}
1179 	free(buf);
1180 	return (0);
1181 }
1182 
1183 int
1184 fetchifs(u_short ifindex)
1185 {
1186 	size_t			 len;
1187 	int			 mib[6];
1188 	char			*buf, *next, *lim;
1189 	struct rt_msghdr	*rtm;
1190 	struct if_msghdr	 ifm;
1191 	struct ifa_msghdr	*ifam;
1192 	struct iface		*iface;
1193 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1194 
1195 	mib[0] = CTL_NET;
1196 	mib[1] = PF_ROUTE;
1197 	mib[2] = 0;
1198 	mib[3] = AF_INET6;
1199 	mib[4] = NET_RT_IFLIST;
1200 	mib[5] = ifindex;
1201 
1202 	if (sysctl(mib, 6, NULL, &len, NULL, 0) == -1) {
1203 		log_warn("sysctl");
1204 		return (-1);
1205 	}
1206 	if ((buf = malloc(len)) == NULL) {
1207 		log_warn("fetchifs");
1208 		return (-1);
1209 	}
1210 	if (sysctl(mib, 6, buf, &len, NULL, 0) == -1) {
1211 		log_warn("sysctl");
1212 		free(buf);
1213 		return (-1);
1214 	}
1215 
1216 	lim = buf + len;
1217 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1218 		rtm = (struct rt_msghdr *)next;
1219 		if (rtm->rtm_version != RTM_VERSION)
1220 			continue;
1221 		switch (rtm->rtm_type) {
1222 		case RTM_IFINFO:
1223 			bcopy(rtm, &ifm, sizeof ifm);
1224 			sa = (struct sockaddr *)(next + sizeof(ifm));
1225 			get_rtaddrs(ifm.ifm_addrs, sa, rti_info);
1226 
1227 			if ((iface = kif_update(ifm.ifm_index,
1228 			    ifm.ifm_flags, &ifm.ifm_data,
1229 			    (struct sockaddr_dl *)rti_info[RTAX_IFP])) == NULL)
1230 			break;
1231 		case RTM_NEWADDR:
1232 			ifam = (struct ifa_msghdr *)rtm;
1233 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1234 			    RTA_BRD)) == 0)
1235 				break;
1236 			sa = (struct sockaddr *)(ifam + 1);
1237 			get_rtaddrs(ifam->ifam_addrs, sa, rti_info);
1238 
1239 			if_newaddr(ifam->ifam_index,
1240 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1241 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1242 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1243 			break;
1244 		}
1245 	}
1246 	free(buf);
1247 	return (0);
1248 }
1249 
1250 int
1251 dispatch_rtmsg(void)
1252 {
1253 	char			 buf[RT_BUF_SIZE];
1254 	ssize_t			 n;
1255 	char			*next, *lim;
1256 	struct rt_msghdr	*rtm;
1257 	struct if_msghdr	 ifm;
1258 	struct ifa_msghdr	*ifam;
1259 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1260 	struct sockaddr_in6	*sa_in6;
1261 	struct sockaddr_rtlabel	*label;
1262 	struct kroute_node	*kr, *okr;
1263 	struct in6_addr		 prefix, nexthop;
1264 	u_int8_t		 prefixlen;
1265 	int			 flags, mpath;
1266 	unsigned int		 scope;
1267 	u_short			 ifindex = 0;
1268 
1269 	if ((n = read(kr_state.fd, &buf, sizeof(buf))) == -1) {
1270 		if (errno == EAGAIN || errno == EINTR)
1271 			return (0);
1272 		log_warn("dispatch_rtmsg: read error");
1273 		return (-1);
1274 	}
1275 
1276 	if (n == 0) {
1277 		log_warnx("routing socket closed");
1278 		return (-1);
1279 	}
1280 
1281 	lim = buf + n;
1282 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1283 		rtm = (struct rt_msghdr *)next;
1284 		if (lim < next + sizeof(u_short) ||
1285 		    lim < next + rtm->rtm_msglen)
1286 			fatalx("dispatch_rtmsg: partial rtm in buffer");
1287 		if (rtm->rtm_version != RTM_VERSION)
1288 			continue;
1289 
1290 		bzero(&prefix, sizeof(prefix));
1291 		bzero(&nexthop, sizeof(nexthop));
1292 		scope = 0;
1293 		prefixlen = 0;
1294 		flags = F_KERNEL;
1295 		mpath = 0;
1296 
1297 		if (rtm->rtm_type == RTM_ADD || rtm->rtm_type == RTM_CHANGE ||
1298 		    rtm->rtm_type == RTM_DELETE) {
1299 			sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
1300 			get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
1301 
1302 			if (rtm->rtm_tableid != 0)
1303 				continue;
1304 
1305 			if (rtm->rtm_pid == kr_state.pid) /* caused by us */
1306 				continue;
1307 
1308 			if (rtm->rtm_errno)		/* failed attempts... */
1309 				continue;
1310 
1311 			/* Skip ARP/ND cache and broadcast routes. */
1312 			if (rtm->rtm_flags & (RTF_LLINFO|RTF_BROADCAST))
1313 				continue;
1314 
1315 #ifdef RTF_MPATH
1316 			if (rtm->rtm_flags & RTF_MPATH)
1317 				mpath = 1;
1318 #endif
1319 			switch (sa->sa_family) {
1320 			case AF_INET6:
1321 				prefix =
1322 				    ((struct sockaddr_in6 *)sa)->sin6_addr;
1323 				sa_in6 = (struct sockaddr_in6 *)
1324 				    rti_info[RTAX_NETMASK];
1325 				if (sa_in6 != NULL) {
1326 					if (sa_in6->sin6_len != 0)
1327 						prefixlen = mask2prefixlen(
1328 						    sa_in6);
1329 				} else if (rtm->rtm_flags & RTF_HOST)
1330 					prefixlen = 128;
1331 				else
1332 					fatalx("classful IPv6 address?!!");
1333 				if (rtm->rtm_flags & RTF_STATIC)
1334 					flags |= F_STATIC;
1335 				if (rtm->rtm_flags & RTF_BLACKHOLE)
1336 					flags |= F_BLACKHOLE;
1337 				if (rtm->rtm_flags & RTF_REJECT)
1338 					flags |= F_REJECT;
1339 				if (rtm->rtm_flags & RTF_DYNAMIC)
1340 					flags |= F_DYNAMIC;
1341 				if (rtm->rtm_flags & RTF_PROTO1)
1342 					flags |= F_BGPD_INSERTED;
1343 				break;
1344 			default:
1345 				continue;
1346 			}
1347 
1348 			ifindex = rtm->rtm_index;
1349 			if ((sa = rti_info[RTAX_GATEWAY]) != NULL) {
1350 				switch (sa->sa_family) {
1351 				case AF_INET6:
1352 					sa_in6 = (struct sockaddr_in6 *)sa;
1353 					/*
1354 					 * XXX The kernel provides the scope
1355 					 * XXX via the kame hack instead of
1356 					 * XXX the scope_id field.
1357 					 */
1358 					recoverscope(sa_in6);
1359 					nexthop = sa_in6->sin6_addr;
1360 					scope = sa_in6->sin6_scope_id;
1361 					break;
1362 				case AF_LINK:
1363 					flags |= F_CONNECTED;
1364 					break;
1365 				}
1366 			}
1367 		}
1368 
1369 		switch (rtm->rtm_type) {
1370 		case RTM_ADD:
1371 		case RTM_CHANGE:
1372 			if (IN6_IS_ADDR_UNSPECIFIED(&nexthop) &&
1373 			    !(flags & F_CONNECTED)) {
1374 				log_warnx("dispatch_rtmsg no nexthop for %s/%u",
1375 				    log_in6addr(&prefix), prefixlen);
1376 				continue;
1377 			}
1378 
1379 			if ((okr = kroute_find(&prefix, prefixlen)) !=
1380 			    NULL) {
1381 				/* just add new multipath routes */
1382 				if (mpath && rtm->rtm_type == RTM_ADD)
1383 					goto add;
1384 				/* get the correct route */
1385 				kr = okr;
1386 				if (mpath && (kr = kroute_matchgw(okr,
1387 				    &nexthop, scope)) == NULL) {
1388 					log_warnx("dispatch_rtmsg mpath route"
1389 					    " not found");
1390 					/* add routes we missed out earlier */
1391 					goto add;
1392 				}
1393 
1394 				/*
1395 				 * ospf route overridden by kernel. Preference
1396 				 * of the route is not checked because this is
1397 				 * forced -- most probably by a user.
1398 				 */
1399 				if (kr->r.flags & F_OSPFD_INSERTED)
1400 					flags |= F_OSPFD_INSERTED;
1401 				if (kr->r.flags & F_REDISTRIBUTED)
1402 					flags |= F_REDISTRIBUTED;
1403 				kr->r.nexthop = nexthop;
1404 				kr->r.scope = scope;
1405 				kr->r.flags = flags;
1406 				kr->r.ifindex = ifindex;
1407 
1408 				rtlabel_unref(kr->r.rtlabel);
1409 				kr->r.rtlabel = 0;
1410 				kr->r.ext_tag = 0;
1411 				if ((label = (struct sockaddr_rtlabel *)
1412 				    rti_info[RTAX_LABEL]) != NULL) {
1413 					kr->r.rtlabel =
1414 					    rtlabel_name2id(label->sr_label);
1415 					kr->r.ext_tag =
1416 					    rtlabel_id2tag(kr->r.rtlabel);
1417 				}
1418 
1419 				if (kif_validate(kr->r.ifindex))
1420 					kr->r.flags &= ~F_DOWN;
1421 				else
1422 					kr->r.flags |= F_DOWN;
1423 
1424 				/* just readd, the RDE will care */
1425 				kr_redistribute(okr);
1426 			} else {
1427 add:
1428 				if ((kr = calloc(1,
1429 				    sizeof(struct kroute_node))) == NULL) {
1430 					log_warn("dispatch_rtmsg");
1431 					return (-1);
1432 				}
1433 				kr->r.prefix = prefix;
1434 				kr->r.prefixlen = prefixlen;
1435 				kr->r.nexthop = nexthop;
1436 				kr->r.scope = scope;
1437 				kr->r.flags = flags;
1438 				kr->r.ifindex = ifindex;
1439 
1440 				if ((label = (struct sockaddr_rtlabel *)
1441 				    rti_info[RTAX_LABEL]) != NULL) {
1442 					kr->r.rtlabel =
1443 					    rtlabel_name2id(label->sr_label);
1444 					kr->r.ext_tag =
1445 					    rtlabel_id2tag(kr->r.rtlabel);
1446 				}
1447 
1448 				kroute_insert(kr);
1449 			}
1450 			break;
1451 		case RTM_DELETE:
1452 			if ((kr = kroute_find(&prefix, prefixlen)) ==
1453 			    NULL)
1454 				continue;
1455 			if (!(kr->r.flags & F_KERNEL))
1456 				continue;
1457 			/* get the correct route */
1458 			okr = kr;
1459 			if (mpath && (kr = kroute_matchgw(kr, &nexthop,
1460 			    scope)) == NULL) {
1461 				log_warnx("dispatch_rtmsg mpath route"
1462 				    " not found");
1463 				return (-1);
1464 			}
1465 			/*
1466 			 * last route is getting removed request the
1467 			 * ospf route from the RDE to insert instead
1468 			 */
1469 			if (okr == kr && kr->next == NULL &&
1470 			    kr->r.flags & F_OSPFD_INSERTED)
1471 				main_imsg_compose_rde(IMSG_KROUTE_GET, 0,
1472 				    &kr->r, sizeof(struct kroute));
1473 			if (kroute_remove(kr) == -1)
1474 				return (-1);
1475 			break;
1476 		case RTM_IFINFO:
1477 			memcpy(&ifm, next, sizeof(ifm));
1478 			if_change(ifm.ifm_index, ifm.ifm_flags,
1479 			    &ifm.ifm_data);
1480 			break;
1481 		case RTM_NEWADDR:
1482 			ifam = (struct ifa_msghdr *)rtm;
1483 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1484 			    RTA_BRD)) == 0)
1485 				break;
1486 			sa = (struct sockaddr *)(ifam + 1);
1487 			get_rtaddrs(ifam->ifam_addrs, sa, rti_info);
1488 
1489 			if_newaddr(ifam->ifam_index,
1490 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1491 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1492 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1493 			break;
1494 		case RTM_DELADDR:
1495 			ifam = (struct ifa_msghdr *)rtm;
1496 			if ((ifam->ifam_addrs & (RTA_NETMASK | RTA_IFA |
1497 			    RTA_BRD)) == 0)
1498 				break;
1499 			sa = (struct sockaddr *)(ifam + 1);
1500 			get_rtaddrs(ifam->ifam_addrs, sa, rti_info);
1501 
1502 			if_deladdr(ifam->ifam_index,
1503 			    (struct sockaddr_in6 *)rti_info[RTAX_IFA],
1504 			    (struct sockaddr_in6 *)rti_info[RTAX_NETMASK],
1505 			    (struct sockaddr_in6 *)rti_info[RTAX_BRD]);
1506 			break;
1507 		case RTM_IFANNOUNCE:
1508 			if_announce(next);
1509 			break;
1510 		default:
1511 			/* ignore for now */
1512 			break;
1513 		}
1514 	}
1515 	return (0);
1516 }
1517