xref: /openbsd-src/usr.sbin/ripd/kroute.c (revision f2da64fbbbf1b03f09f390ab01267c93dfd77c4c)
1 /*	$OpenBSD: kroute.c,v 1.31 2015/09/27 17:32:36 stsp 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 
39 #include "rip.h"
40 #include "ripd.h"
41 #include "log.h"
42 
43 struct {
44 	u_int32_t		rtseq;
45 	pid_t			pid;
46 	int			fib_sync;
47 	int			fd;
48 	struct event		ev;
49 	u_int			rdomain;
50 } kr_state;
51 
52 struct kroute_node {
53 	RB_ENTRY(kroute_node)	 entry;
54 	struct kroute		 r;
55 };
56 
57 struct kif_node {
58 	RB_ENTRY(kif_node)	 entry;
59 	struct kif		 k;
60 };
61 
62 void	kr_redistribute(int, struct kroute *);
63 int	kroute_compare(struct kroute_node *, struct kroute_node *);
64 int	kif_compare(struct kif_node *, struct kif_node *);
65 int	kr_change_fib(struct kroute_node *, struct kroute *, int);
66 
67 struct kroute_node	*kroute_find(in_addr_t, in_addr_t, u_int8_t);
68 int			 kroute_insert(struct kroute_node *);
69 int			 kroute_remove(struct kroute_node *);
70 void			 kroute_clear(void);
71 
72 struct kif_node		*kif_find(int);
73 int			 kif_insert(struct kif_node *);
74 int			 kif_remove(struct kif_node *);
75 void			 kif_clear(void);
76 int			 kif_validate(int);
77 
78 struct kroute_node	*kroute_match(in_addr_t);
79 
80 int		protect_lo(void);
81 u_int8_t	prefixlen_classful(in_addr_t);
82 void		get_rtaddrs(int, struct sockaddr *, struct sockaddr **);
83 void		if_change(u_short, int, struct if_data *);
84 void		if_announce(void *);
85 
86 int		send_rtmsg(int, int, struct kroute *);
87 int		dispatch_rtmsg(void);
88 int		fetchtable(void);
89 int		fetchifs(int);
90 
91 RB_HEAD(kroute_tree, kroute_node)	krt;
92 RB_PROTOTYPE(kroute_tree, kroute_node, entry, kroute_compare)
93 RB_GENERATE(kroute_tree, kroute_node, entry, kroute_compare)
94 
95 RB_HEAD(kif_tree, kif_node)		kit;
96 RB_PROTOTYPE(kif_tree, kif_node, entry, kif_compare)
97 RB_GENERATE(kif_tree, kif_node, entry, kif_compare)
98 
99 int
100 kif_init(void)
101 {
102 	RB_INIT(&kit);
103 
104 	if (fetchifs(0) == -1)
105 		return (-1);
106 
107 	return (0);
108 }
109 
110 int
111 kr_init(int fs, u_int rdomain)
112 {
113 	int		opt = 0, rcvbuf, default_rcvbuf;
114 	socklen_t	optlen;
115 
116 	if ((kr_state.fd = socket(AF_ROUTE,
117 	    SOCK_RAW | SOCK_CLOEXEC | SOCK_NONBLOCK, 0)) == -1) {
118 		log_warn("kr_init: socket");
119 		return (-1);
120 	}
121 
122 	/* not interested in my own messages */
123 	if (setsockopt(kr_state.fd, SOL_SOCKET, SO_USELOOPBACK,
124 	    &opt, sizeof(opt)) == -1)
125 		log_warn("kr_init: setsockopt");	/* not fatal */
126 
127 	/* grow receive buffer, don't wanna miss messages */
128 	optlen = sizeof(default_rcvbuf);
129 	if (getsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF,
130 	    &default_rcvbuf, &optlen) == -1)
131 		log_warn("kr_init getsockopt SOL_SOCKET SO_RCVBUF");
132 	else
133 		for (rcvbuf = MAX_RTSOCK_BUF;
134 		    rcvbuf > default_rcvbuf &&
135 		    setsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF,
136 		    &rcvbuf, sizeof(rcvbuf)) == -1 && errno == ENOBUFS;
137 		    rcvbuf /= 2)
138 			;	/* nothing */
139 
140 	kr_state.pid = getpid();
141 	kr_state.rtseq = 1;
142 
143 	RB_INIT(&krt);
144 
145 	if (fetchtable() == -1)
146 		return (-1);
147 
148 	if (protect_lo() == -1)
149 		return (-1);
150 
151 	kr_state.fib_sync = fs; /* now set correct sync mode */
152 	kr_state.rdomain = rdomain;
153 
154 	event_set(&kr_state.ev, kr_state.fd, EV_READ | EV_PERSIST,
155 	    kr_dispatch_msg, NULL);
156 	event_add(&kr_state.ev, NULL);
157 
158 	return (0);
159 }
160 
161 int
162 kr_change_fib(struct kroute_node *kr, struct kroute *kroute, int action)
163 {
164 	/* nexthop within 127/8 -> ignore silently */
165 	if ((kroute->nexthop.s_addr & htonl(IN_CLASSA_NET)) ==
166 	    htonl(INADDR_LOOPBACK & IN_CLASSA_NET))
167 		return (0);
168 
169 	if (send_rtmsg(kr_state.fd, action, kroute) == -1)
170 		return (-1);
171 
172 	if (action == RTM_ADD) {
173 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL)
174 			fatal("kr_change_fib");
175 
176 		kr->r.prefix.s_addr = kroute->prefix.s_addr;
177 		kr->r.netmask.s_addr = kroute->netmask.s_addr;
178 		kr->r.nexthop.s_addr = kroute->nexthop.s_addr;
179 		kr->r.flags = kroute->flags |= F_RIPD_INSERTED;
180 		kr->r.priority = RTP_RIP;
181 
182 		if (kroute_insert(kr) == -1) {
183 			log_debug("kr_update_fib: cannot insert %s",
184 			    inet_ntoa(kr->r.nexthop));
185 			free(kr);
186 		}
187 	} else
188 		kr->r.nexthop.s_addr = kroute->nexthop.s_addr;
189 
190 	return (0);
191 }
192 
193 int
194 kr_change(struct kroute *kroute)
195 {
196 	struct kroute_node	*kr;
197 	int			 action = RTM_ADD;
198 
199 	kr = kroute_find(kroute->prefix.s_addr, kroute->netmask.s_addr,
200 	    RTP_RIP);
201 	if (kr != NULL)
202 		action = RTM_CHANGE;
203 
204 	return (kr_change_fib(kr, kroute, action));
205 }
206 
207 int
208 kr_delete(struct kroute *kroute)
209 {
210 	struct kroute_node	*kr;
211 
212 	kr = kroute_find(kroute->prefix.s_addr, kroute->netmask.s_addr,
213 	    RTP_RIP);
214 	if (kr == NULL)
215 		return (0);
216 
217 	if (kr->r.priority != RTP_RIP)
218 		log_warn("kr_delete_fib: %s/%d has wrong priority %d",
219 		    inet_ntoa(kr->r.prefix), mask2prefixlen(kr->r.netmask.s_addr),
220 		    kr->r.priority);
221 
222 	if (send_rtmsg(kr_state.fd, RTM_DELETE, kroute) == -1)
223 		return (-1);
224 
225 	if (kroute_remove(kr) == -1)
226 		return (-1);
227 
228 	return (0);
229 }
230 
231 void
232 kr_shutdown(void)
233 {
234 	kr_fib_decouple();
235 
236 	kroute_clear();
237 	kif_clear();
238 }
239 
240 void
241 kr_fib_couple(void)
242 {
243 	struct kroute_node	*kr;
244 
245 	if (kr_state.fib_sync == 1)	/* already coupled */
246 		return;
247 
248 	kr_state.fib_sync = 1;
249 
250 	RB_FOREACH(kr, kroute_tree, &krt)
251 		if (kr->r.priority == RTP_RIP)
252 			send_rtmsg(kr_state.fd, RTM_ADD, &kr->r);
253 
254 	log_info("kernel routing table coupled");
255 }
256 
257 void
258 kr_fib_decouple(void)
259 {
260 	struct kroute_node	*kr;
261 
262 	if (kr_state.fib_sync == 0)	/* already decoupled */
263 		return;
264 
265 	RB_FOREACH(kr, kroute_tree, &krt)
266 		if (kr->r.priority == RTP_RIP)
267 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
268 
269 	kr_state.fib_sync = 0;
270 
271 	log_info("kernel routing table decoupled");
272 }
273 
274 /* ARGSUSED */
275 void
276 kr_dispatch_msg(int fd, short event, void *bula)
277 {
278 	dispatch_rtmsg();
279 }
280 
281 void
282 kr_show_route(struct imsg *imsg)
283 {
284 	struct kroute_node	*kr;
285 	int			 flags;
286 	struct in_addr		 addr;
287 
288 	switch (imsg->hdr.type) {
289 	case IMSG_CTL_KROUTE:
290 		if (imsg->hdr.len != IMSG_HEADER_SIZE + sizeof(flags)) {
291 			log_warnx("kr_show_route: wrong imsg len");
292 			return;
293 		}
294 		memcpy(&flags, imsg->data, sizeof(flags));
295 		RB_FOREACH(kr, kroute_tree, &krt)
296 			if (!flags || kr->r.flags & flags) {
297 				main_imsg_compose_ripe(IMSG_CTL_KROUTE,
298 				    imsg->hdr.pid, &kr->r, sizeof(kr->r));
299 			}
300 		break;
301 	case IMSG_CTL_KROUTE_ADDR:
302 		if (imsg->hdr.len != IMSG_HEADER_SIZE +
303 		    sizeof(struct in_addr)) {
304 			log_warnx("kr_show_route: wrong imsg len");
305 			return;
306 		}
307 		memcpy(&addr, imsg->data, sizeof(addr));
308 		kr = NULL;
309 		kr = kroute_match(addr.s_addr);
310 		if (kr != NULL)
311 			main_imsg_compose_ripe(IMSG_CTL_KROUTE, imsg->hdr.pid,
312 			    &kr->r, sizeof(kr->r));
313 		break;
314 	default:
315 		log_debug("kr_show_route: error handling imsg");
316 		break;
317 	}
318 
319 	main_imsg_compose_ripe(IMSG_CTL_END, imsg->hdr.pid, NULL, 0);
320 }
321 
322 void
323 kr_ifinfo(char *ifname, pid_t pid)
324 {
325 	struct kif_node	*kif;
326 
327 	RB_FOREACH(kif, kif_tree, &kit)
328 		if (ifname == NULL || !strcmp(ifname, kif->k.ifname)) {
329 			main_imsg_compose_ripe(IMSG_CTL_IFINFO,
330 			    pid, &kif->k, sizeof(kif->k));
331 		}
332 
333 	main_imsg_compose_ripe(IMSG_CTL_END, pid, NULL, 0);
334 }
335 
336 void
337 kr_redistribute(int type, struct kroute *kr)
338 {
339 	u_int32_t	a;
340 
341 
342 	if (type == IMSG_NETWORK_DEL) {
343 dont_redistribute:
344 		/* was the route redistributed? */
345 		if (kr->flags & F_REDISTRIBUTED) {
346 			/* remove redistributed flag */
347 			kr->flags &= ~F_REDISTRIBUTED;
348 			main_imsg_compose_rde(type, 0, kr,
349 			    sizeof(struct kroute));
350 		}
351 		return;
352 	}
353 
354 	/* interface is not up and running so don't announce */
355 	if (kr->flags & F_DOWN)
356 		return;
357 
358 	/*
359 	 * We consider the loopback net, multicast and experimental addresses
360 	 * as not redistributable.
361 	 */
362 	a = ntohl(kr->prefix.s_addr);
363 	if (IN_MULTICAST(a) || IN_BADCLASS(a) ||
364 	    (a >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
365 		return;
366 	/*
367 	 * Consider networks with nexthop loopback as not redistributable
368 	 * unless it is a reject or blackhole route.
369 	 */
370 	if (kr->nexthop.s_addr == htonl(INADDR_LOOPBACK) &&
371 	    !(kr->flags & (F_BLACKHOLE|F_REJECT)))
372 		return;
373 
374 	/* Should we redistribute this route? */
375 	if (!rip_redistribute(kr))
376 		goto dont_redistribute;
377 
378 	/* Does not matter if we resend the kr, the RDE will cope. */
379 	kr->flags |= F_REDISTRIBUTED;
380 	main_imsg_compose_rde(type, 0, kr, sizeof(struct kroute));
381 }
382 
383 /* rb-tree compare */
384 int
385 kroute_compare(struct kroute_node *a, struct kroute_node *b)
386 {
387 	if (ntohl(a->r.prefix.s_addr) < ntohl(b->r.prefix.s_addr))
388 		return (-1);
389 	if (ntohl(a->r.prefix.s_addr) > ntohl(b->r.prefix.s_addr))
390 		return (1);
391 	if (ntohl(a->r.netmask.s_addr) < ntohl(b->r.netmask.s_addr))
392 		return (-1);
393 	if (ntohl(a->r.netmask.s_addr) > ntohl(b->r.netmask.s_addr))
394 		return (1);
395 
396 	/* if the priority is RTP_ANY finish on the first address hit */
397 	if (a->r.priority == RTP_ANY || b->r.priority == RTP_ANY)
398 		return (0);
399 	if (a->r.priority < b->r.priority)
400 		return (-1);
401 	if (a->r.priority > b->r.priority)
402 		return (1);
403 
404 	return (0);
405 }
406 
407 int
408 kif_compare(struct kif_node *a, struct kif_node *b)
409 {
410 	return (b->k.ifindex - a->k.ifindex);
411 }
412 
413 /* tree management */
414 struct kroute_node *
415 kroute_find(in_addr_t prefix, in_addr_t netmask, u_int8_t prio)
416 {
417 	struct kroute_node	s, *kn, *tmp;
418 
419 	s.r.prefix.s_addr = prefix;
420 	s.r.netmask.s_addr = netmask;
421 	s.r.priority = prio;
422 
423 	kn = RB_FIND(kroute_tree, &krt, &s);
424 	if (kn && prio == RTP_ANY) {
425 		tmp = RB_PREV(kroute_tree, &krt, kn);
426 		while (tmp) {
427 			if (kroute_compare(&s, tmp) == 0)
428 				kn = tmp;
429 			else
430 				break;
431 			tmp = RB_PREV(kroute_tree, &krt, kn);
432 		}
433 	}
434 
435 	return (kn);
436 }
437 
438 int
439 kroute_insert(struct kroute_node *kr)
440 {
441 	if (RB_INSERT(kroute_tree, &krt, kr) != NULL) {
442 		log_warnx("kroute_insert failed for %s/%u",
443 		    inet_ntoa(kr->r.prefix),
444 		    mask2prefixlen(kr->r.netmask.s_addr));
445 		free(kr);
446 		return (-1);
447 	}
448 
449 	if (!(kr->r.flags & F_KERNEL)) {
450 		/* don't validate or redistribute rip route */
451 		kr->r.flags &= ~F_DOWN;
452 		return (0);
453 	}
454 
455 	if (kif_validate(kr->r.ifindex))
456 		kr->r.flags &= ~F_DOWN;
457 	else
458 		kr->r.flags |= F_DOWN;
459 
460 	kr_redistribute(IMSG_NETWORK_ADD, &kr->r);
461 
462 	return (0);
463 }
464 
465 int
466 kroute_remove(struct kroute_node *kr)
467 {
468 	if (RB_REMOVE(kroute_tree, &krt, kr) == NULL) {
469 		log_warnx("kroute_remove failed for %s/%u",
470 		    inet_ntoa(kr->r.prefix),
471 		    mask2prefixlen(kr->r.netmask.s_addr));
472 		return (-1);
473 	}
474 
475 	kr_redistribute(IMSG_NETWORK_DEL, &kr->r);
476 	rtlabel_unref(kr->r.rtlabel);
477 
478 	free(kr);
479 	return (0);
480 }
481 
482 void
483 kroute_clear(void)
484 {
485 	struct kroute_node	*kr;
486 
487 	while ((kr = RB_MIN(kroute_tree, &krt)) != NULL)
488 		kroute_remove(kr);
489 }
490 
491 struct kif_node *
492 kif_find(int ifindex)
493 {
494 	struct kif_node	s;
495 
496 	bzero(&s, sizeof(s));
497 	s.k.ifindex = ifindex;
498 
499 	return (RB_FIND(kif_tree, &kit, &s));
500 }
501 
502 struct kif *
503 kif_findname(char *ifname)
504 {
505 	struct kif_node	*kif;
506 
507 	RB_FOREACH(kif, kif_tree, &kit)
508 		if (!strcmp(ifname, kif->k.ifname))
509 			return (&kif->k);
510 
511 	return (NULL);
512 }
513 
514 int
515 kif_insert(struct kif_node *kif)
516 {
517 	if (RB_INSERT(kif_tree, &kit, kif) != NULL) {
518 		log_warnx("RB_INSERT(kif_tree, &kit, kif)");
519 		free(kif);
520 		return (-1);
521 	}
522 
523 	return (0);
524 }
525 
526 int
527 kif_remove(struct kif_node *kif)
528 {
529 	if (RB_REMOVE(kif_tree, &kit, kif) == NULL) {
530 		log_warnx("RB_REMOVE(kif_tree, &kit, kif)");
531 		return (-1);
532 	}
533 
534 	free(kif);
535 	return (0);
536 }
537 
538 void
539 kif_clear(void)
540 {
541 	struct kif_node	*kif;
542 
543 	while ((kif = RB_MIN(kif_tree, &kit)) != NULL)
544 		kif_remove(kif);
545 }
546 
547 int
548 kif_validate(int ifindex)
549 {
550 	struct kif_node		*kif;
551 
552 	if ((kif = kif_find(ifindex)) == NULL) {
553 		log_warnx("interface with index %u not found", ifindex);
554 		return (1);
555 	}
556 
557 	return (kif->k.nh_reachable);
558 }
559 
560 struct kroute_node *
561 kroute_match(in_addr_t key)
562 {
563 	u_int8_t		 i;
564 	struct kroute_node	*kr;
565 
566 	/* we will never match the default route */
567 	for (i = 32; i > 0; i--)
568 		if ((kr = kroute_find(key & prefixlen2mask(i),
569 		    prefixlen2mask(i), RTP_ANY)) != NULL)
570 			return (kr);
571 
572 	/* if we don't have a match yet, try to find a default route */
573 	if ((kr = kroute_find(0, 0, RTP_ANY)) != NULL)
574 			return (kr);
575 
576 	return (NULL);
577 }
578 
579 /* misc */
580 int
581 protect_lo(void)
582 {
583 	struct kroute_node	*kr;
584 
585 	/* special protection for 127/8 */
586 	if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
587 		log_warn("protect_lo");
588 		return (-1);
589 	}
590 	kr->r.prefix.s_addr = htonl(INADDR_LOOPBACK);
591 	kr->r.netmask.s_addr = htonl(IN_CLASSA_NET);
592 	kr->r.flags = F_KERNEL|F_CONNECTED;
593 
594 	if (RB_INSERT(kroute_tree, &krt, kr) != NULL)
595 		free(kr);	/* kernel route already there, no problem */
596 
597 	return (0);
598 }
599 
600 u_int8_t
601 prefixlen_classful(in_addr_t ina)
602 {
603 	/* it hurt to write this. */
604 
605 	if (ina >= 0xf0000000U)		/* class E */
606 		return (32);
607 	else if (ina >= 0xe0000000U)	/* class D */
608 		return (4);
609 	else if (ina >= 0xc0000000U)	/* class C */
610 		return (24);
611 	else if (ina >= 0x80000000U)	/* class B */
612 		return (16);
613 	else				/* class A */
614 		return (8);
615 }
616 
617 u_int8_t
618 mask2prefixlen(in_addr_t ina)
619 {
620 	if (ina == 0)
621 		return (0);
622 	else
623 		return (33 - ffs(ntohl(ina)));
624 }
625 
626 in_addr_t
627 prefixlen2mask(u_int8_t prefixlen)
628 {
629 	if (prefixlen == 0)
630 		return (0);
631 
632 	return (htonl(0xffffffff << (32 - prefixlen)));
633 }
634 
635 #define	ROUNDUP(a, size)	\
636     (((a) & ((size) - 1)) ? (1 + ((a) | ((size) - 1))) : (a))
637 
638 void
639 get_rtaddrs(int addrs, struct sockaddr *sa, struct sockaddr **rti_info)
640 {
641 	int	i;
642 
643 	for (i = 0; i < RTAX_MAX; i++) {
644 		if (addrs & (1 << i)) {
645 			rti_info[i] = sa;
646 			sa = (struct sockaddr *)((char *)(sa) +
647 			    ROUNDUP(sa->sa_len, sizeof(long)));
648 		} else
649 			rti_info[i] = NULL;
650 	}
651 }
652 
653 void
654 if_change(u_short ifindex, int flags, struct if_data *ifd)
655 {
656 	struct kif_node		*kif;
657 	struct kroute_node	*kr;
658 	int			 type;
659 	u_int8_t		 reachable;
660 
661 	if ((kif = kif_find(ifindex)) == NULL) {
662 		log_warnx("interface with index %u not found", ifindex);
663 		return;
664 	}
665 
666 	kif->k.flags = flags;
667 	kif->k.link_state = ifd->ifi_link_state;
668 	kif->k.if_type = ifd->ifi_type;
669 	kif->k.baudrate = ifd->ifi_baudrate;
670 
671 	if ((reachable = (flags & IFF_UP) &&
672 	    LINK_STATE_IS_UP(ifd->ifi_link_state)) == kif->k.nh_reachable)
673 		return;		/* nothing changed wrt nexthop validity */
674 
675 	kif->k.nh_reachable = reachable;
676 	type = reachable ? IMSG_NETWORK_ADD : IMSG_NETWORK_DEL;
677 
678 	/* notify ripe about interface link state */
679 	main_imsg_compose_ripe(IMSG_IFINFO, 0, &kif->k, sizeof(kif->k));
680 
681 	/* update redistribute list */
682 	RB_FOREACH(kr, kroute_tree, &krt)
683 		if (kr->r.ifindex == ifindex) {
684 			if (reachable)
685 				kr->r.flags &= ~F_DOWN;
686 			else
687 				kr->r.flags |= F_DOWN;
688 
689 			kr_redistribute(type, &kr->r);
690 		}
691 }
692 
693 void
694 if_announce(void *msg)
695 {
696 	struct if_announcemsghdr	*ifan;
697 	struct kif_node			*kif;
698 
699 	ifan = msg;
700 
701 	switch (ifan->ifan_what) {
702 	case IFAN_ARRIVAL:
703 		if ((kif = calloc(1, sizeof(struct kif_node))) == NULL) {
704 			log_warn("if_announce");
705 			return;
706 		}
707 
708 		kif->k.ifindex = ifan->ifan_index;
709 		strlcpy(kif->k.ifname, ifan->ifan_name, sizeof(kif->k.ifname));
710 		kif_insert(kif);
711 		break;
712 	case IFAN_DEPARTURE:
713 		kif = kif_find(ifan->ifan_index);
714 		kif_remove(kif);
715 		break;
716 	}
717 }
718 
719 /* rtsock */
720 int
721 send_rtmsg(int fd, int action, struct kroute *kroute)
722 {
723 	struct iovec		iov[4];
724 	struct rt_msghdr	hdr;
725 	struct sockaddr_in	prefix;
726 	struct sockaddr_in	nexthop;
727 	struct sockaddr_in	mask;
728 	int			iovcnt = 0;
729 
730 	if (kr_state.fib_sync == 0)
731 		return (0);
732 
733 	/* initialize header */
734 	bzero(&hdr, sizeof(hdr));
735 	hdr.rtm_version = RTM_VERSION;
736 	hdr.rtm_type = action;
737 	hdr.rtm_priority = RTP_RIP;
738 	hdr.rtm_tableid = kr_state.rdomain;
739 	if (action == RTM_CHANGE)
740 		hdr.rtm_fmask = RTF_REJECT|RTF_BLACKHOLE;
741 	hdr.rtm_seq = kr_state.rtseq++;	/* overflow doesn't matter */
742 	hdr.rtm_msglen = sizeof(hdr);
743 	/* adjust iovec */
744 	iov[iovcnt].iov_base = &hdr;
745 	iov[iovcnt++].iov_len = sizeof(hdr);
746 
747 	bzero(&prefix, sizeof(prefix));
748 	prefix.sin_len = sizeof(prefix);
749 	prefix.sin_family = AF_INET;
750 	prefix.sin_addr.s_addr = kroute->prefix.s_addr;
751 	/* adjust header */
752 	hdr.rtm_addrs |= RTA_DST;
753 	hdr.rtm_msglen += sizeof(prefix);
754 	/* adjust iovec */
755 	iov[iovcnt].iov_base = &prefix;
756 	iov[iovcnt++].iov_len = sizeof(prefix);
757 
758 	if (kroute->nexthop.s_addr != 0) {
759 		bzero(&nexthop, sizeof(nexthop));
760 		nexthop.sin_len = sizeof(nexthop);
761 		nexthop.sin_family = AF_INET;
762 		nexthop.sin_addr.s_addr = kroute->nexthop.s_addr;
763 		/* adjust header */
764 		hdr.rtm_flags |= RTF_GATEWAY;
765 		hdr.rtm_addrs |= RTA_GATEWAY;
766 		hdr.rtm_msglen += sizeof(nexthop);
767 		/* adjust iovec */
768 		iov[iovcnt].iov_base = &nexthop;
769 		iov[iovcnt++].iov_len = sizeof(nexthop);
770 	}
771 
772 	bzero(&mask, sizeof(mask));
773 	mask.sin_len = sizeof(mask);
774 	mask.sin_family = AF_INET;
775 	mask.sin_addr.s_addr = kroute->netmask.s_addr;
776 	/* adjust header */
777 	hdr.rtm_addrs |= RTA_NETMASK;
778 	hdr.rtm_msglen += sizeof(mask);
779 	/* adjust iovec */
780 	iov[iovcnt].iov_base = &mask;
781 	iov[iovcnt++].iov_len = sizeof(mask);
782 
783 
784 retry:
785 	if (writev(fd, iov, iovcnt) == -1) {
786 		if (errno == ESRCH) {
787 			if (hdr.rtm_type == RTM_CHANGE) {
788 				hdr.rtm_type = RTM_ADD;
789 				goto retry;
790 			} else if (hdr.rtm_type == RTM_DELETE) {
791 				log_info("route %s/%u vanished before delete",
792 				    inet_ntoa(kroute->prefix),
793 				    mask2prefixlen(kroute->netmask.s_addr));
794 				return (0);
795 			}
796 		}
797 		log_warn("send_rtmsg: action %u, prefix %s/%u",
798 		    hdr.rtm_type, inet_ntoa(kroute->prefix),
799 		    mask2prefixlen(kroute->netmask.s_addr));
800 		return (0);
801 	}
802 
803 	return (0);
804 }
805 
806 int
807 fetchtable(void)
808 {
809 	size_t			 len;
810 	int			 mib[7];
811 	char			*buf, *next, *lim;
812 	struct rt_msghdr	*rtm;
813 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
814 	struct sockaddr_in	*sa_in;
815 	struct sockaddr_rtlabel	*label;
816 	struct kroute_node	*kr;
817 	struct iface		*iface = NULL;
818 
819 	mib[0] = CTL_NET;
820 	mib[1] = PF_ROUTE;
821 	mib[2] = 0;
822 	mib[3] = AF_INET;
823 	mib[4] = NET_RT_DUMP;
824 	mib[5] = 0;
825 	mib[6] = kr_state.rdomain;	/* rtableid */
826 
827 	if (sysctl(mib, 7, NULL, &len, NULL, 0) == -1) {
828 		log_warn("sysctl");
829 		return (-1);
830 	}
831 	if ((buf = malloc(len)) == NULL) {
832 		log_warn("fetchtable");
833 		return (-1);
834 	}
835 	if (sysctl(mib, 7, buf, &len, NULL, 0) == -1) {
836 		log_warn("sysctl");
837 		free(buf);
838 		return (-1);
839 	}
840 
841 	lim = buf + len;
842 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
843 		rtm = (struct rt_msghdr *)next;
844 		if (rtm->rtm_version != RTM_VERSION)
845 			continue;
846 		sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
847 		get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
848 
849 		if ((sa = rti_info[RTAX_DST]) == NULL)
850 			continue;
851 
852 		/* Skip ARP/ND cache and broadcast routes. */
853 		if (rtm->rtm_flags & (RTF_LLINFO|RTF_BROADCAST))
854 			continue;
855 
856 #ifdef RTF_MPATH
857 		if (rtm->rtm_flags & RTF_MPATH)		/* multipath */
858 			continue;
859 #endif
860 
861 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
862 			log_warn("fetchtable");
863 			free(buf);
864 			return (-1);
865 		}
866 
867 		kr->r.flags = F_KERNEL;
868 		kr->r.priority = rtm->rtm_priority;
869 
870 		switch (sa->sa_family) {
871 		case AF_INET:
872 			kr->r.prefix.s_addr =
873 			    ((struct sockaddr_in *)sa)->sin_addr.s_addr;
874 			sa_in = (struct sockaddr_in *)rti_info[RTAX_NETMASK];
875 			if (rtm->rtm_flags & RTF_STATIC)
876 				kr->r.flags |= F_STATIC;
877 			if (rtm->rtm_flags & RTF_BLACKHOLE)
878 				kr->r.flags |= F_BLACKHOLE;
879 			if (rtm->rtm_flags & RTF_REJECT)
880 				kr->r.flags |= F_REJECT;
881 			if (rtm->rtm_flags & RTF_DYNAMIC)
882 				kr->r.flags |= F_DYNAMIC;
883 			if (sa_in != NULL) {
884 				if (sa_in->sin_len == 0)
885 					break;
886 				kr->r.netmask.s_addr =
887 				    sa_in->sin_addr.s_addr;
888 			} else if (rtm->rtm_flags & RTF_HOST)
889 				kr->r.netmask.s_addr = prefixlen2mask(32);
890 			else
891 				kr->r.netmask.s_addr =
892 				    prefixlen2mask(prefixlen_classful
893 					(kr->r.prefix.s_addr));
894 			break;
895 		default:
896 			free(kr);
897 			continue;
898 		}
899 
900 		kr->r.ifindex = rtm->rtm_index;
901 
902 		iface = if_find_index(rtm->rtm_index);
903 		if (iface != NULL)
904 			kr->r.metric = iface->cost;
905 		else
906 			kr->r.metric = DEFAULT_COST;
907 
908 		if ((sa = rti_info[RTAX_GATEWAY]) != NULL)
909 			switch (sa->sa_family) {
910 			case AF_INET:
911 				if (rtm->rtm_flags & RTF_CONNECTED) {
912 					kr->r.flags |= F_CONNECTED;
913 					break;
914 				}
915 
916 				kr->r.nexthop.s_addr =
917 				    ((struct sockaddr_in *)sa)->sin_addr.s_addr;
918 				break;
919 			case AF_LINK:
920 				/*
921 				 * Traditional BSD connected routes have
922 				 * a gateway of type AF_LINK.
923 				 */
924 				kr->r.flags |= F_CONNECTED;
925 				break;
926 			}
927 
928 		if (rtm->rtm_priority == RTP_RIP) {
929 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
930 			free(kr);
931 		} else {
932 			if ((label = (struct sockaddr_rtlabel *)
933 			    rti_info[RTAX_LABEL]) != NULL)
934 				kr->r.rtlabel =
935 				    rtlabel_name2id(label->sr_label);
936 			kroute_insert(kr);
937 		}
938 
939 	}
940 	free(buf);
941 	return (0);
942 }
943 
944 int
945 fetchifs(int ifindex)
946 {
947 	size_t			 len;
948 	int			 mib[6];
949 	char			*buf, *next, *lim;
950 	struct if_msghdr	 ifm;
951 	struct kif_node		*kif;
952 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
953 	struct sockaddr_dl	*sdl;
954 
955 	mib[0] = CTL_NET;
956 	mib[1] = PF_ROUTE;
957 	mib[2] = 0;
958 	mib[3] = AF_INET;
959 	mib[4] = NET_RT_IFLIST;
960 	mib[5] = ifindex;
961 
962 	if (sysctl(mib, 6, NULL, &len, NULL, 0) == -1) {
963 		log_warn("sysctl");
964 		return (-1);
965 	}
966 	if ((buf = malloc(len)) == NULL) {
967 		log_warn("fetchif");
968 		return (-1);
969 	}
970 	if (sysctl(mib, 6, buf, &len, NULL, 0) == -1) {
971 		log_warn("sysctl");
972 		free(buf);
973 		return (-1);
974 	}
975 
976 	lim = buf + len;
977 	for (next = buf; next < lim; next += ifm.ifm_msglen) {
978 		memcpy(&ifm, next, sizeof(ifm));
979 		if (ifm.ifm_version != RTM_VERSION)
980 			continue;
981 		if (ifm.ifm_type != RTM_IFINFO)
982 			continue;
983 
984 		sa = (struct sockaddr *)(next + sizeof(ifm));
985 		get_rtaddrs(ifm.ifm_addrs, sa, rti_info);
986 
987 		if ((kif = calloc(1, sizeof(struct kif_node))) == NULL) {
988 			log_warn("fetchifs");
989 			free(buf);
990 			return (-1);
991 		}
992 
993 		kif->k.ifindex = ifm.ifm_index;
994 		kif->k.flags = ifm.ifm_flags;
995 		kif->k.link_state = ifm.ifm_data.ifi_link_state;
996 		kif->k.if_type = ifm.ifm_data.ifi_type;
997 		kif->k.baudrate = ifm.ifm_data.ifi_baudrate;
998 		kif->k.mtu = ifm.ifm_data.ifi_mtu;
999 		kif->k.nh_reachable = (kif->k.flags & IFF_UP) &&
1000 		    LINK_STATE_IS_UP(ifm.ifm_data.ifi_link_state);
1001 		if ((sa = rti_info[RTAX_IFP]) != NULL)
1002 			if (sa->sa_family == AF_LINK) {
1003 				sdl = (struct sockaddr_dl *)sa;
1004 				if (sdl->sdl_nlen >= sizeof(kif->k.ifname))
1005 					memcpy(kif->k.ifname, sdl->sdl_data,
1006 					    sizeof(kif->k.ifname) - 1);
1007 				else if (sdl->sdl_nlen > 0)
1008 					memcpy(kif->k.ifname, sdl->sdl_data,
1009 					    sdl->sdl_nlen);
1010 				/* string already terminated via calloc() */
1011 			}
1012 
1013 		kif_insert(kif);
1014 	}
1015 	free(buf);
1016 	return (0);
1017 }
1018 
1019 int
1020 dispatch_rtmsg(void)
1021 {
1022 	char			 buf[RT_BUF_SIZE];
1023 	ssize_t			 n;
1024 	char			*next, *lim;
1025 	struct rt_msghdr	*rtm;
1026 	struct if_msghdr	 ifm;
1027 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1028 	struct sockaddr_in	*sa_in;
1029 	struct sockaddr_rtlabel	*label;
1030 	struct kroute_node	*kr;
1031 	struct in_addr		 prefix, nexthop, netmask;
1032 	struct iface		*iface = NULL;
1033 	int			 flags;
1034 	u_short			 ifindex = 0;
1035 	u_int8_t		 metric, prio;
1036 
1037 	if ((n = read(kr_state.fd, &buf, sizeof(buf))) == -1) {
1038 		if (errno == EAGAIN || errno == EINTR)
1039 			return (0);
1040 		log_warn("dispatch_rtmsg: read error");
1041 		return (-1);
1042 	}
1043 
1044 	if (n == 0) {
1045 		log_warnx("routing socket closed");
1046 		return (-1);
1047 	}
1048 
1049 	lim = buf + n;
1050 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1051 		rtm = (struct rt_msghdr *)next;
1052 		if (lim < next + sizeof(u_short) ||
1053 		    lim < next + rtm->rtm_msglen)
1054 			fatalx("dispatch_rtmsg: partial rtm in buffer");
1055 		if (rtm->rtm_version != RTM_VERSION)
1056 			continue;
1057 
1058 		prefix.s_addr = 0;
1059 		netmask.s_addr = 0;
1060 		flags = F_KERNEL;
1061 		nexthop.s_addr = 0;
1062 		prio = 0;
1063 
1064 		if (rtm->rtm_type == RTM_ADD || rtm->rtm_type == RTM_CHANGE ||
1065 		    rtm->rtm_type == RTM_DELETE) {
1066 			sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
1067 			get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
1068 
1069 			if (rtm->rtm_tableid != kr_state.rdomain)
1070 				continue;
1071 
1072 			if (rtm->rtm_pid == kr_state.pid)	/* cause by us */
1073 				continue;
1074 
1075 			if (rtm->rtm_errno)			/* failed attempts... */
1076 				continue;
1077 
1078 			/* Skip ARP/ND cache and broadcast routes. */
1079 			if (rtm->rtm_flags & (RTF_LLINFO|RTF_BROADCAST))
1080 				continue;
1081 
1082 			prio = rtm->rtm_priority;
1083 
1084 			switch (sa->sa_family) {
1085 			case AF_INET:
1086 				prefix.s_addr =
1087 				    ((struct sockaddr_in *)sa)->sin_addr.s_addr;
1088 				sa_in = (struct sockaddr_in *)
1089 				    rti_info[RTAX_NETMASK];
1090 				if (sa_in != NULL) {
1091 					if (sa_in->sin_len != 0)
1092 						netmask.s_addr =
1093 						    sa_in->sin_addr.s_addr;
1094 				} else if (rtm->rtm_flags & RTF_HOST)
1095 					netmask.s_addr = prefixlen2mask(32);
1096 				else
1097 					netmask.s_addr =
1098 					    prefixlen2mask(prefixlen_classful(
1099 						prefix.s_addr));
1100 				if (rtm->rtm_flags & RTF_STATIC)
1101 					flags |= F_STATIC;
1102 				if (rtm->rtm_flags & RTF_BLACKHOLE)
1103 					flags |= F_BLACKHOLE;
1104 				if (rtm->rtm_flags & RTF_REJECT)
1105 					flags |= F_REJECT;
1106 				if (rtm->rtm_flags & RTF_DYNAMIC)
1107 					flags |= F_DYNAMIC;
1108 				break;
1109 			default:
1110 				continue;
1111 			}
1112 
1113 			ifindex = rtm->rtm_index;
1114 			if ((sa = rti_info[RTAX_GATEWAY]) != NULL) {
1115 				switch (sa->sa_family) {
1116 				case AF_INET:
1117 					nexthop.s_addr = ((struct
1118 					    sockaddr_in *)sa)->sin_addr.s_addr;
1119 					break;
1120 				case AF_LINK:
1121 					flags |= F_CONNECTED;
1122 					break;
1123 				}
1124 			}
1125 		}
1126 
1127 		switch (rtm->rtm_type) {
1128 		case RTM_ADD:
1129 		case RTM_CHANGE:
1130 			if (nexthop.s_addr == 0 && !(flags & F_CONNECTED)) {
1131 				log_warnx("dispatch_rtmsg no nexthop for %s/%u",
1132 				    inet_ntoa(prefix),
1133 				    mask2prefixlen(netmask.s_addr));
1134 				continue;
1135 			}
1136 
1137 			if ((kr = kroute_find(prefix.s_addr, netmask.s_addr,
1138 			    prio)) != NULL) {
1139 				if (kr->r.flags & F_REDISTRIBUTED)
1140 					flags |= F_REDISTRIBUTED;
1141 				kr->r.nexthop.s_addr = nexthop.s_addr;
1142 				kr->r.flags = flags;
1143 				kr->r.ifindex = ifindex;
1144 				kr->r.priority = prio;
1145 
1146 				rtlabel_unref(kr->r.rtlabel);
1147 				kr->r.rtlabel = 0;
1148 				if ((label = (struct sockaddr_rtlabel *)
1149 				    rti_info[RTAX_LABEL]) != NULL)
1150 					kr->r.rtlabel =
1151 					    rtlabel_name2id(label->sr_label);
1152 
1153 				if (kif_validate(kr->r.ifindex))
1154 					kr->r.flags &= ~F_DOWN;
1155 				else
1156 					kr->r.flags |= F_DOWN;
1157 
1158 				/* just readd, the RDE will care */
1159 				kr_redistribute(IMSG_NETWORK_ADD, &kr->r);
1160 			} else {
1161 				if ((kr = calloc(1,
1162 				    sizeof(struct kroute_node))) == NULL) {
1163 					log_warn("dispatch_rtmsg");
1164 					return (-1);
1165 				}
1166 
1167 				iface = if_find_index(rtm->rtm_index);
1168 				if (iface != NULL)
1169 					metric = iface->cost;
1170 				else
1171 					metric = DEFAULT_COST;
1172 
1173 				kr->r.prefix.s_addr = prefix.s_addr;
1174 				kr->r.netmask.s_addr = netmask.s_addr;
1175 				kr->r.nexthop.s_addr = nexthop.s_addr;
1176 				kr->r.metric = metric;
1177 				kr->r.flags = flags;
1178 				kr->r.ifindex = ifindex;
1179 
1180 				if ((label = (struct sockaddr_rtlabel *)
1181 				    rti_info[RTAX_LABEL]) != NULL)
1182 					kr->r.rtlabel =
1183 					    rtlabel_name2id(label->sr_label);
1184 
1185 				kroute_insert(kr);
1186 			}
1187 			break;
1188 		case RTM_DELETE:
1189 			if ((kr = kroute_find(prefix.s_addr, netmask.s_addr,
1190 			    prio)) == NULL)
1191 				continue;
1192 			if (!(kr->r.flags & F_KERNEL))
1193 				continue;
1194 			if (kroute_remove(kr) == -1)
1195 				return (-1);
1196 			break;
1197 		case RTM_IFINFO:
1198 			memcpy(&ifm, next, sizeof(ifm));
1199 			if_change(ifm.ifm_index, ifm.ifm_flags,
1200 			    &ifm.ifm_data);
1201 			break;
1202 		case RTM_IFANNOUNCE:
1203 			if_announce(next);
1204 			break;
1205 		default:
1206 			/* ignore for now */
1207 			break;
1208 		}
1209 	}
1210 	return (0);
1211 }
1212