xref: /openbsd-src/usr.sbin/ripd/kroute.c (revision 2b0358df1d88d06ef4139321dd05bd5e05d91eaf)
1 /*	$OpenBSD: kroute.c,v 1.14 2009/03/31 09:11:45 michele 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 "rip.h"
41 #include "ripd.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 };
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 
66 struct kroute_node	*kroute_find(in_addr_t, in_addr_t);
67 int			 kroute_insert(struct kroute_node *);
68 int			 kroute_remove(struct kroute_node *);
69 void			 kroute_clear(void);
70 
71 struct kif_node		*kif_find(int);
72 int			 kif_insert(struct kif_node *);
73 int			 kif_remove(struct kif_node *);
74 void			 kif_clear(void);
75 int			 kif_validate(int);
76 
77 struct kroute_node	*kroute_match(in_addr_t);
78 
79 int		protect_lo(void);
80 u_int8_t	prefixlen_classful(in_addr_t);
81 void		get_rtaddrs(int, struct sockaddr *, struct sockaddr **);
82 void		if_change(u_short, int, struct if_data *);
83 void		if_announce(void *);
84 
85 int		send_rtmsg(int, int, struct kroute *);
86 int		dispatch_rtmsg(void);
87 int		fetchtable(void);
88 int		fetchifs(int);
89 
90 RB_HEAD(kroute_tree, kroute_node)	krt;
91 RB_PROTOTYPE(kroute_tree, kroute_node, entry, kroute_compare)
92 RB_GENERATE(kroute_tree, kroute_node, entry, kroute_compare)
93 
94 RB_HEAD(kif_tree, kif_node)		kit;
95 RB_PROTOTYPE(kif_tree, kif_node, entry, kif_compare)
96 RB_GENERATE(kif_tree, kif_node, entry, kif_compare)
97 
98 struct kroute kr_all_rip_routers;
99 int	flag_all_rip_routers = 0;
100 
101 int
102 kif_init(void)
103 {
104 	RB_INIT(&kit);
105 
106 	if (fetchifs(0) == -1)
107 		return (-1);
108 
109 	return (0);
110 }
111 
112 int
113 kr_init(int fs)
114 {
115 	int		opt = 0, rcvbuf, default_rcvbuf;
116 	socklen_t	optlen;
117 
118 	if ((kr_state.fd = socket(AF_ROUTE, SOCK_RAW, 0)) == -1) {
119 		log_warn("kr_init: socket");
120 		return (-1);
121 	}
122 
123 	/* not interested in my own messages */
124 	if (setsockopt(kr_state.fd, SOL_SOCKET, SO_USELOOPBACK,
125 	    &opt, sizeof(opt)) == -1)
126 		log_warn("kr_init: setsockopt");	/* not fatal */
127 
128 	/* grow receive buffer, don't wanna miss messages */
129 	optlen = sizeof(default_rcvbuf);
130 	if (getsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF,
131 	    &default_rcvbuf, &optlen) == -1)
132 		log_warn("kr_init getsockopt SOL_SOCKET SO_RCVBUF");
133 	else
134 		for (rcvbuf = MAX_RTSOCK_BUF;
135 		    rcvbuf > default_rcvbuf &&
136 		    setsockopt(kr_state.fd, SOL_SOCKET, SO_RCVBUF,
137 		    &rcvbuf, sizeof(rcvbuf)) == -1 && errno == ENOBUFS;
138 		    rcvbuf /= 2)
139 			;	/* nothing */
140 
141 	kr_state.pid = getpid();
142 	kr_state.rtseq = 1;
143 
144 	RB_INIT(&krt);
145 
146 	if (fetchtable() == -1)
147 		return (-1);
148 
149 	if (protect_lo() == -1)
150 		return (-1);
151 
152 	kr_all_rip_routers.prefix.s_addr = inet_addr(ALL_RIP_ROUTERS);
153 	kr_all_rip_routers.netmask.s_addr = htonl(INADDR_BROADCAST);
154 	kr_all_rip_routers.nexthop.s_addr = htonl(INADDR_LOOPBACK);
155 
156 	kr_state.fib_sync = 1; /* force addition of multicast route */
157 	if (send_rtmsg(kr_state.fd, RTM_ADD, &kr_all_rip_routers) != -1)
158 		flag_all_rip_routers = 1;
159 
160 	kr_state.fib_sync = fs; /* now set correct sync mode */
161 
162 	event_set(&kr_state.ev, kr_state.fd, EV_READ | EV_PERSIST,
163 	    kr_dispatch_msg, NULL);
164 	event_add(&kr_state.ev, NULL);
165 
166 	return (0);
167 }
168 
169 int
170 kr_change(struct kroute *kroute)
171 {
172 	struct kroute_node	*kr;
173 	int			 action = RTM_ADD;
174 
175 	if ((kr = kroute_find(kroute->prefix.s_addr, kroute->netmask.s_addr)) !=
176 	    NULL) {
177 		if (!(kr->r.flags & F_KERNEL))
178 			action = RTM_CHANGE;
179 		else {	/* a non-rip route already exists. not a problem */
180 			if (!(kr->r.flags & (F_BGPD_INSERTED|
181 			    F_OSPFD_INSERTED))) {
182 				kr->r.flags |= F_RIPD_INSERTED;
183 				return (0);
184 			}
185 			/*
186 			 * rip route has higher pref
187 			 * - reset flags to the rip ones
188 			 * - use RTM_CHANGE
189 			 * - zero out ifindex (this is no longer relevant)
190 			 */
191 			action = RTM_CHANGE;
192 			kr->r.flags = kroute->flags | F_RIPD_INSERTED;
193 			kr->r.ifindex = 0;
194 			rtlabel_unref(kr->r.rtlabel);
195 			kr->r.rtlabel = 0;
196 		}
197 	}
198 
199 	/* nexthop within 127/8 -> ignore silently */
200 	if ((kroute->nexthop.s_addr & htonl(IN_CLASSA_NET)) ==
201 	    htonl(INADDR_LOOPBACK & IN_CLASSA_NET))
202 		return (0);
203 
204 	if (send_rtmsg(kr_state.fd, action, kroute) == -1)
205 		return (-1);
206 
207 	if (action == RTM_ADD) {
208 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
209 			log_warn("kr_change");
210 			return (-1);
211 		}
212 		kr->r.prefix.s_addr = kroute->prefix.s_addr;
213 		kr->r.netmask.s_addr = kroute->netmask.s_addr;
214 		kr->r.nexthop.s_addr = kroute->nexthop.s_addr;
215 		kr->r.flags = kroute->flags |= F_RIPD_INSERTED;
216 
217 		if (kroute_insert(kr) == -1)
218 			free(kr);
219 	} else
220 		kr->r.nexthop.s_addr = kroute->nexthop.s_addr;
221 
222 	return (0);
223 }
224 
225 int
226 kr_delete(struct kroute *kroute)
227 {
228 	struct kroute_node	*kr;
229 
230 	if ((kr = kroute_find(kroute->prefix.s_addr, kroute->netmask.s_addr)) ==
231 	    NULL)
232 		return (0);
233 
234 	if (!(kr->r.flags & F_RIPD_INSERTED))
235 		return (0);
236 
237 	if (kr->r.flags & F_KERNEL) {
238 		/* remove F_RIPD_INSERTED flag, route still exists in kernel */
239 		kr->r.flags &= ~F_RIPD_INSERTED;
240 		return (0);
241 	}
242 
243 	if (send_rtmsg(kr_state.fd, RTM_DELETE, kroute) == -1)
244 		return (-1);
245 
246 	if (kroute_remove(kr) == -1)
247 		return (-1);
248 
249 	return (0);
250 }
251 
252 void
253 kr_shutdown(void)
254 {
255 	kr_fib_decouple();
256 
257 	if (flag_all_rip_routers) {
258 		kr_state.fib_sync = 1; /* force removal of mulitcast route */
259 		(void)send_rtmsg(kr_state.fd, RTM_DELETE, &kr_all_rip_routers);
260 	}
261 
262 	kroute_clear();
263 	kif_clear();
264 }
265 
266 void
267 kr_fib_couple(void)
268 {
269 	struct kroute_node	*kr;
270 
271 	if (kr_state.fib_sync == 1)	/* already coupled */
272 		return;
273 
274 	kr_state.fib_sync = 1;
275 
276 	RB_FOREACH(kr, kroute_tree, &krt)
277 		if (!(kr->r.flags & F_KERNEL))
278 			send_rtmsg(kr_state.fd, RTM_ADD, &kr->r);
279 
280 	log_info("kernel routing table coupled");
281 }
282 
283 void
284 kr_fib_decouple(void)
285 {
286 	struct kroute_node	*kr;
287 
288 	if (kr_state.fib_sync == 0)	/* already decoupled */
289 		return;
290 
291 	RB_FOREACH(kr, kroute_tree, &krt)
292 		if (!(kr->r.flags & F_KERNEL))
293 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
294 
295 	kr_state.fib_sync = 0;
296 
297 	log_info("kernel routing table decoupled");
298 }
299 
300 /* ARGSUSED */
301 void
302 kr_dispatch_msg(int fd, short event, void *bula)
303 {
304 	dispatch_rtmsg();
305 }
306 
307 void
308 kr_show_route(struct imsg *imsg)
309 {
310 	struct kroute_node	*kr;
311 	int			 flags;
312 	struct in_addr		 addr;
313 
314 	switch (imsg->hdr.type) {
315 	case IMSG_CTL_KROUTE:
316 		if (imsg->hdr.len != IMSG_HEADER_SIZE + sizeof(flags)) {
317 			log_warnx("kr_show_route: wrong imsg len");
318 			return;
319 		}
320 		memcpy(&flags, imsg->data, sizeof(flags));
321 		RB_FOREACH(kr, kroute_tree, &krt)
322 			if (!flags || kr->r.flags & flags) {
323 				main_imsg_compose_ripe(IMSG_CTL_KROUTE,
324 				    imsg->hdr.pid, &kr->r, sizeof(kr->r));
325 			}
326 		break;
327 	case IMSG_CTL_KROUTE_ADDR:
328 		if (imsg->hdr.len != IMSG_HEADER_SIZE +
329 		    sizeof(struct in_addr)) {
330 			log_warnx("kr_show_route: wrong imsg len");
331 			return;
332 		}
333 		memcpy(&addr, imsg->data, sizeof(addr));
334 		kr = NULL;
335 		kr = kroute_match(addr.s_addr);
336 		if (kr != NULL)
337 			main_imsg_compose_ripe(IMSG_CTL_KROUTE, imsg->hdr.pid,
338 			    &kr->r, sizeof(kr->r));
339 		break;
340 	default:
341 		log_debug("kr_show_route: error handling imsg");
342 		break;
343 	}
344 
345 	main_imsg_compose_ripe(IMSG_CTL_END, imsg->hdr.pid, NULL, 0);
346 }
347 
348 void
349 kr_ifinfo(char *ifname, pid_t pid)
350 {
351 	struct kif_node	*kif;
352 
353 	RB_FOREACH(kif, kif_tree, &kit)
354 		if (ifname == NULL || !strcmp(ifname, kif->k.ifname)) {
355 			main_imsg_compose_ripe(IMSG_CTL_IFINFO,
356 			    pid, &kif->k, sizeof(kif->k));
357 		}
358 
359 	main_imsg_compose_ripe(IMSG_CTL_END, pid, NULL, 0);
360 }
361 
362 void
363 kr_redistribute(int type, struct kroute *kr)
364 {
365 	u_int32_t	a;
366 
367 
368 	if (type == IMSG_NETWORK_DEL) {
369 dont_redistribute:
370 		/* was the route redistributed? */
371 		if (kr->flags & F_REDISTRIBUTED) {
372 			/* remove redistributed flag */
373 			kr->flags &= ~F_REDISTRIBUTED;
374 			main_imsg_compose_rde(type, 0, kr,
375 			    sizeof(struct kroute));
376 		}
377 		return;
378 	}
379 
380 	/* interface is not up and running so don't announce */
381 	if (kr->flags & F_DOWN)
382 		return;
383 
384 	/*
385 	 * We consider the loopback net, multicast and experimental addresses
386 	 * as not redistributable.
387 	 */
388 	a = ntohl(kr->prefix.s_addr);
389 	if (IN_MULTICAST(a) || IN_BADCLASS(a) ||
390 	    (a >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
391 		return;
392 	/*
393 	 * Consider networks with nexthop loopback as not redistributable.
394 	 */
395 	if (kr->nexthop.s_addr == htonl(INADDR_LOOPBACK))
396 		return;
397 
398 	/* Should we redistribute this route? */
399 	if (!rip_redistribute(kr))
400 		goto dont_redistribute;
401 
402 	/* Does not matter if we resend the kr, the RDE will cope. */
403 	kr->flags |= F_REDISTRIBUTED;
404 	main_imsg_compose_rde(type, 0, kr, sizeof(struct kroute));
405 }
406 
407 /* rb-tree compare */
408 int
409 kroute_compare(struct kroute_node *a, struct kroute_node *b)
410 {
411 	if (ntohl(a->r.prefix.s_addr) < ntohl(b->r.prefix.s_addr))
412 		return (-1);
413 	if (ntohl(a->r.prefix.s_addr) > ntohl(b->r.prefix.s_addr))
414 		return (1);
415 	if (ntohl(a->r.netmask.s_addr) < ntohl(b->r.netmask.s_addr))
416 		return (-1);
417 	if (ntohl(a->r.netmask.s_addr) > ntohl(b->r.netmask.s_addr))
418 		return (1);
419 	return (0);
420 }
421 
422 int
423 kif_compare(struct kif_node *a, struct kif_node *b)
424 {
425 	return (b->k.ifindex - a->k.ifindex);
426 }
427 
428 /* tree management */
429 struct kroute_node *
430 kroute_find(in_addr_t prefix, in_addr_t netmask)
431 {
432 	struct kroute_node	s;
433 
434 	s.r.prefix.s_addr = prefix;
435 	s.r.netmask.s_addr = netmask;
436 
437 	return (RB_FIND(kroute_tree, &krt, &s));
438 }
439 
440 int
441 kroute_insert(struct kroute_node *kr)
442 {
443 	if (RB_INSERT(kroute_tree, &krt, kr) != NULL) {
444 		log_warnx("kroute_insert failed for %s/%u",
445 		    inet_ntoa(kr->r.prefix),
446 		    mask2prefixlen(kr->r.netmask.s_addr));
447 		free(kr);
448 		return (-1);
449 	}
450 
451 	if (!(kr->r.flags & F_KERNEL)) {
452 		/* don't validate or redistribute rip route */
453 		kr->r.flags &= ~F_DOWN;
454 		return (0);
455 	}
456 
457 	if (kif_validate(kr->r.ifindex))
458 		kr->r.flags &= ~F_DOWN;
459 	else
460 		kr->r.flags |= F_DOWN;
461 
462 	kr_redistribute(IMSG_NETWORK_ADD, &kr->r);
463 
464 	return (0);
465 }
466 
467 int
468 kroute_remove(struct kroute_node *kr)
469 {
470 	if (RB_REMOVE(kroute_tree, &krt, kr) == NULL) {
471 		log_warnx("kroute_remove failed for %s/%u",
472 		    inet_ntoa(kr->r.prefix),
473 		    mask2prefixlen(kr->r.netmask.s_addr));
474 		return (-1);
475 	}
476 
477 	kr_redistribute(IMSG_NETWORK_DEL, &kr->r);
478 	rtlabel_unref(kr->r.rtlabel);
479 
480 	free(kr);
481 	return (0);
482 }
483 
484 void
485 kroute_clear(void)
486 {
487 	struct kroute_node	*kr;
488 
489 	while ((kr = RB_MIN(kroute_tree, &krt)) != NULL)
490 		kroute_remove(kr);
491 }
492 
493 struct kif_node *
494 kif_find(int ifindex)
495 {
496 	struct kif_node	s;
497 
498 	bzero(&s, sizeof(s));
499 	s.k.ifindex = ifindex;
500 
501 	return (RB_FIND(kif_tree, &kit, &s));
502 }
503 
504 struct kif *
505 kif_findname(char *ifname)
506 {
507 	struct kif_node	*kif;
508 
509 	RB_FOREACH(kif, kif_tree, &kit)
510 		if (!strcmp(ifname, kif->k.ifname))
511 			return (&kif->k);
512 
513 	return (NULL);
514 }
515 
516 int
517 kif_insert(struct kif_node *kif)
518 {
519 	if (RB_INSERT(kif_tree, &kit, kif) != NULL) {
520 		log_warnx("RB_INSERT(kif_tree, &kit, kif)");
521 		free(kif);
522 		return (-1);
523 	}
524 
525 	return (0);
526 }
527 
528 int
529 kif_remove(struct kif_node *kif)
530 {
531 	if (RB_REMOVE(kif_tree, &kit, kif) == NULL) {
532 		log_warnx("RB_REMOVE(kif_tree, &kit, kif)");
533 		return (-1);
534 	}
535 
536 	free(kif);
537 	return (0);
538 }
539 
540 void
541 kif_clear(void)
542 {
543 	struct kif_node	*kif;
544 
545 	while ((kif = RB_MIN(kif_tree, &kit)) != NULL)
546 		kif_remove(kif);
547 }
548 
549 int
550 kif_validate(int ifindex)
551 {
552 	struct kif_node		*kif;
553 
554 	if ((kif = kif_find(ifindex)) == NULL) {
555 		log_warnx("interface with index %u not found", ifindex);
556 		return (1);
557 	}
558 
559 	return (kif->k.nh_reachable);
560 }
561 
562 struct kroute_node *
563 kroute_match(in_addr_t key)
564 {
565 	u_int8_t		 i;
566 	struct kroute_node	*kr;
567 
568 	/* we will never match the default route */
569 	for (i = 32; i > 0; i--)
570 		if ((kr = kroute_find(key & prefixlen2mask(i),
571 		    prefixlen2mask(i))) != NULL)
572 			return (kr);
573 
574 	/* if we don't have a match yet, try to find a default route */
575 	if ((kr = kroute_find(0, 0)) != NULL)
576 			return (kr);
577 
578 	return (NULL);
579 }
580 
581 /* misc */
582 int
583 protect_lo(void)
584 {
585 	struct kroute_node	*kr;
586 
587 	/* special protection for 127/8 */
588 	if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
589 		log_warn("protect_lo");
590 		return (-1);
591 	}
592 	kr->r.prefix.s_addr = htonl(INADDR_LOOPBACK);
593 	kr->r.netmask.s_addr = htonl(IN_CLASSA_NET);
594 	kr->r.flags = F_KERNEL|F_CONNECTED;
595 
596 	if (RB_INSERT(kroute_tree, &krt, kr) != NULL)
597 		free(kr);	/* kernel route already there, no problem */
598 
599 	return (0);
600 }
601 
602 u_int8_t
603 prefixlen_classful(in_addr_t ina)
604 {
605 	/* it hurt to write this. */
606 
607 	if (ina >= 0xf0000000U)		/* class E */
608 		return (32);
609 	else if (ina >= 0xe0000000U)	/* class D */
610 		return (4);
611 	else if (ina >= 0xc0000000U)	/* class C */
612 		return (24);
613 	else if (ina >= 0x80000000U)	/* class B */
614 		return (16);
615 	else				/* class A */
616 		return (8);
617 }
618 
619 u_int8_t
620 mask2prefixlen(in_addr_t ina)
621 {
622 	if (ina == 0)
623 		return (0);
624 	else
625 		return (33 - ffs(ntohl(ina)));
626 }
627 
628 in_addr_t
629 prefixlen2mask(u_int8_t prefixlen)
630 {
631 	if (prefixlen == 0)
632 		return (0);
633 
634 	return (htonl(0xffffffff << (32 - prefixlen)));
635 }
636 
637 #define	ROUNDUP(a, size)	\
638     (((a) & ((size) - 1)) ? (1 + ((a) | ((size) - 1))) : (a))
639 
640 void
641 get_rtaddrs(int addrs, struct sockaddr *sa, struct sockaddr **rti_info)
642 {
643 	int	i;
644 
645 	for (i = 0; i < RTAX_MAX; i++) {
646 		if (addrs & (1 << i)) {
647 			rti_info[i] = sa;
648 			sa = (struct sockaddr *)((char *)(sa) +
649 			    ROUNDUP(sa->sa_len, sizeof(long)));
650 		} else
651 			rti_info[i] = NULL;
652 	}
653 }
654 
655 void
656 if_change(u_short ifindex, int flags, struct if_data *ifd)
657 {
658 	struct kif_node		*kif;
659 	struct kroute_node	*kr;
660 	int			 type;
661 	u_int8_t		 reachable;
662 
663 	if ((kif = kif_find(ifindex)) == NULL) {
664 		log_warnx("interface with index %u not found", ifindex);
665 		return;
666 	}
667 
668 	kif->k.flags = flags;
669 	kif->k.link_state = ifd->ifi_link_state;
670 	kif->k.media_type = ifd->ifi_type;
671 	kif->k.baudrate = ifd->ifi_baudrate;
672 
673 	if ((reachable = (flags & IFF_UP) &&
674 	    (LINK_STATE_IS_UP(ifd->ifi_link_state) ||
675 	    (ifd->ifi_link_state == LINK_STATE_UNKNOWN &&
676 	    ifd->ifi_type != IFT_CARP))) == kif->k.nh_reachable)
677 		return;		/* nothing changed wrt nexthop validity */
678 
679 	kif->k.nh_reachable = reachable;
680 	type = reachable ? IMSG_NETWORK_ADD : IMSG_NETWORK_DEL;
681 
682 	/* notify ripe about interface link state */
683 	main_imsg_compose_ripe(IMSG_IFINFO, 0, &kif->k, sizeof(kif->k));
684 
685 	/* update redistribute list */
686 	RB_FOREACH(kr, kroute_tree, &krt)
687 		if (kr->r.ifindex == ifindex) {
688 			if (reachable)
689 				kr->r.flags &= ~F_DOWN;
690 			else
691 				kr->r.flags |= F_DOWN;
692 
693 			kr_redistribute(type, &kr->r);
694 		}
695 }
696 
697 void
698 if_announce(void *msg)
699 {
700 	struct if_announcemsghdr	*ifan;
701 	struct kif_node			*kif;
702 
703 	ifan = msg;
704 
705 	switch (ifan->ifan_what) {
706 	case IFAN_ARRIVAL:
707 		if ((kif = calloc(1, sizeof(struct kif_node))) == NULL) {
708 			log_warn("if_announce");
709 			return;
710 		}
711 
712 		kif->k.ifindex = ifan->ifan_index;
713 		strlcpy(kif->k.ifname, ifan->ifan_name, sizeof(kif->k.ifname));
714 		kif_insert(kif);
715 		break;
716 	case IFAN_DEPARTURE:
717 		kif = kif_find(ifan->ifan_index);
718 		kif_remove(kif);
719 		break;
720 	}
721 }
722 
723 /* rtsock */
724 int
725 send_rtmsg(int fd, int action, struct kroute *kroute)
726 {
727 	struct iovec		iov[4];
728 	struct rt_msghdr	hdr;
729 	struct sockaddr_in	prefix;
730 	struct sockaddr_in	nexthop;
731 	struct sockaddr_in	mask;
732 	int			iovcnt = 0;
733 
734 	if (kr_state.fib_sync == 0)
735 		return (0);
736 
737 	/* initialize header */
738 	bzero(&hdr, sizeof(hdr));
739 	hdr.rtm_version = RTM_VERSION;
740 	hdr.rtm_type = action;
741 	hdr.rtm_flags = RTF_PROTO3;
742 	hdr.rtm_priority = RTP_RIP;
743 	if (action == RTM_CHANGE)	/* force PROTO3 reset the other flags */
744 		hdr.rtm_fmask =
745 		    RTF_PROTO3|RTF_PROTO2|RTF_PROTO1|RTF_REJECT|RTF_BLACKHOLE;
746 	hdr.rtm_seq = kr_state.rtseq++;	/* overflow doesn't matter */
747 	hdr.rtm_msglen = sizeof(hdr);
748 	/* adjust iovec */
749 	iov[iovcnt].iov_base = &hdr;
750 	iov[iovcnt++].iov_len = sizeof(hdr);
751 
752 	bzero(&prefix, sizeof(prefix));
753 	prefix.sin_len = sizeof(prefix);
754 	prefix.sin_family = AF_INET;
755 	prefix.sin_addr.s_addr = kroute->prefix.s_addr;
756 	/* adjust header */
757 	hdr.rtm_addrs |= RTA_DST;
758 	hdr.rtm_msglen += sizeof(prefix);
759 	/* adjust iovec */
760 	iov[iovcnt].iov_base = &prefix;
761 	iov[iovcnt++].iov_len = sizeof(prefix);
762 
763 	if (kroute->nexthop.s_addr != 0) {
764 		bzero(&nexthop, sizeof(nexthop));
765 		nexthop.sin_len = sizeof(nexthop);
766 		nexthop.sin_family = AF_INET;
767 		nexthop.sin_addr.s_addr = kroute->nexthop.s_addr;
768 		/* adjust header */
769 		hdr.rtm_flags |= RTF_GATEWAY;
770 		hdr.rtm_addrs |= RTA_GATEWAY;
771 		hdr.rtm_msglen += sizeof(nexthop);
772 		/* adjust iovec */
773 		iov[iovcnt].iov_base = &nexthop;
774 		iov[iovcnt++].iov_len = sizeof(nexthop);
775 	}
776 
777 	bzero(&mask, sizeof(mask));
778 	mask.sin_len = sizeof(mask);
779 	mask.sin_family = AF_INET;
780 	mask.sin_addr.s_addr = kroute->netmask.s_addr;
781 	/* adjust header */
782 	hdr.rtm_addrs |= RTA_NETMASK;
783 	hdr.rtm_msglen += sizeof(mask);
784 	/* adjust iovec */
785 	iov[iovcnt].iov_base = &mask;
786 	iov[iovcnt++].iov_len = sizeof(mask);
787 
788 
789 retry:
790 	if (writev(fd, iov, iovcnt) == -1) {
791 		switch (errno) {
792 		case ESRCH:
793 			if (hdr.rtm_type == RTM_CHANGE) {
794 				hdr.rtm_type = RTM_ADD;
795 				goto retry;
796 			} else if (hdr.rtm_type == RTM_DELETE) {
797 				log_info("route %s/%u vanished before delete",
798 				    inet_ntoa(kroute->prefix),
799 				    mask2prefixlen(kroute->netmask.s_addr));
800 				return (0);
801 			} else {
802 				log_warnx("send_rtmsg: action %u, "
803 				    "prefix %s/%u: %s", hdr.rtm_type,
804 				    inet_ntoa(kroute->prefix),
805 				    mask2prefixlen(kroute->netmask.s_addr),
806 				    strerror(errno));
807 				return (0);
808 			}
809 			break;
810 		default:
811 			log_warnx("send_rtmsg: action %u, prefix %s/%u: %s",
812 			    hdr.rtm_type, inet_ntoa(kroute->prefix),
813 			    mask2prefixlen(kroute->netmask.s_addr),
814 			    strerror(errno));
815 			return (0);
816 		}
817 	}
818 
819 	return (0);
820 }
821 
822 int
823 fetchtable(void)
824 {
825 	size_t			 len;
826 	int			 mib[7];
827 	char			*buf, *next, *lim;
828 	struct rt_msghdr	*rtm;
829 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
830 	struct sockaddr_in	*sa_in;
831 	struct sockaddr_rtlabel	*label;
832 	struct kroute_node	*kr;
833 	struct iface		*iface = NULL;
834 
835 	mib[0] = CTL_NET;
836 	mib[1] = AF_ROUTE;
837 	mib[2] = 0;
838 	mib[3] = AF_INET;
839 	mib[4] = NET_RT_DUMP;
840 	mib[5] = 0;
841 	mib[6] = 0;	/* rtableid */
842 
843 	if (sysctl(mib, 7, NULL, &len, NULL, 0) == -1) {
844 		log_warn("sysctl");
845 		return (-1);
846 	}
847 	if ((buf = malloc(len)) == NULL) {
848 		log_warn("fetchtable");
849 		return (-1);
850 	}
851 	if (sysctl(mib, 7, buf, &len, NULL, 0) == -1) {
852 		log_warn("sysctl");
853 		free(buf);
854 		return (-1);
855 	}
856 
857 	lim = buf + len;
858 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
859 		rtm = (struct rt_msghdr *)next;
860 		if (rtm->rtm_version != RTM_VERSION)
861 			continue;
862 		sa = (struct sockaddr *)(rtm + 1);
863 		get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
864 
865 		if ((sa = rti_info[RTAX_DST]) == NULL)
866 			continue;
867 
868 		if (rtm->rtm_flags & RTF_LLINFO)	/* arp cache */
869 			continue;
870 
871 #ifdef RTF_MPATH
872 		if (rtm->rtm_flags & RTF_MPATH)		/* multipath */
873 			continue;
874 #endif
875 
876 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
877 			log_warn("fetchtable");
878 			free(buf);
879 			return (-1);
880 		}
881 
882 		kr->r.flags = F_KERNEL;
883 
884 		switch (sa->sa_family) {
885 		case AF_INET:
886 			kr->r.prefix.s_addr =
887 			    ((struct sockaddr_in *)sa)->sin_addr.s_addr;
888 			sa_in = (struct sockaddr_in *)rti_info[RTAX_NETMASK];
889 			if (rtm->rtm_flags & RTF_STATIC)
890 				kr->r.flags |= F_STATIC;
891 			if (rtm->rtm_flags & RTF_DYNAMIC)
892 				kr->r.flags |= F_DYNAMIC;
893 			if (rtm->rtm_flags & RTF_PROTO1)
894 				kr->r.flags |= F_BGPD_INSERTED;
895 			if (rtm->rtm_flags & RTF_PROTO2)
896 				kr->r.flags |= F_OSPFD_INSERTED;
897 			if (sa_in != NULL) {
898 				if (sa_in->sin_len == 0)
899 					break;
900 				kr->r.netmask.s_addr =
901 				    sa_in->sin_addr.s_addr;
902 			} else if (rtm->rtm_flags & RTF_HOST)
903 				kr->r.netmask.s_addr = prefixlen2mask(32);
904 			else
905 				kr->r.netmask.s_addr =
906 				    prefixlen2mask(prefixlen_classful
907 					(kr->r.prefix.s_addr));
908 			break;
909 		default:
910 			free(kr);
911 			continue;
912 		}
913 
914 		kr->r.ifindex = rtm->rtm_index;
915 
916 		iface = if_find_index(rtm->rtm_index);
917 		if (iface != NULL)
918 			kr->r.metric = iface->cost;
919 		else
920 			kr->r.metric = DEFAULT_COST;
921 
922 		if ((sa = rti_info[RTAX_GATEWAY]) != NULL)
923 			switch (sa->sa_family) {
924 			case AF_INET:
925 				kr->r.nexthop.s_addr =
926 				    ((struct sockaddr_in *)sa)->sin_addr.s_addr;
927 				break;
928 			case AF_LINK:
929 				kr->r.flags |= F_CONNECTED;
930 				break;
931 			}
932 
933 		if (rtm->rtm_flags & RTF_PROTO3) {
934 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
935 			free(kr);
936 		} else {
937 			if ((label = (struct sockaddr_rtlabel *)
938 			    rti_info[RTAX_LABEL]) != NULL)
939 				kr->r.rtlabel =
940 				    rtlabel_name2id(label->sr_label);
941 			kroute_insert(kr);
942 		}
943 
944 	}
945 	free(buf);
946 	return (0);
947 }
948 
949 int
950 fetchifs(int ifindex)
951 {
952 	size_t			 len;
953 	int			 mib[6];
954 	char			*buf, *next, *lim;
955 	struct if_msghdr	 ifm;
956 	struct kif_node		*kif;
957 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
958 	struct sockaddr_dl	*sdl;
959 
960 	mib[0] = CTL_NET;
961 	mib[1] = AF_ROUTE;
962 	mib[2] = 0;
963 	mib[3] = AF_INET;
964 	mib[4] = NET_RT_IFLIST;
965 	mib[5] = ifindex;
966 
967 	if (sysctl(mib, 6, NULL, &len, NULL, 0) == -1) {
968 		log_warn("sysctl");
969 		return (-1);
970 	}
971 	if ((buf = malloc(len)) == NULL) {
972 		log_warn("fetchif");
973 		return (-1);
974 	}
975 	if (sysctl(mib, 6, buf, &len, NULL, 0) == -1) {
976 		log_warn("sysctl");
977 		free(buf);
978 		return (-1);
979 	}
980 
981 	lim = buf + len;
982 	for (next = buf; next < lim; next += ifm.ifm_msglen) {
983 		memcpy(&ifm, next, sizeof(ifm));
984 		if (ifm.ifm_version != RTM_VERSION)
985 			continue;
986 		if (ifm.ifm_type != RTM_IFINFO)
987 			continue;
988 
989 		sa = (struct sockaddr *)(next + sizeof(ifm));
990 		get_rtaddrs(ifm.ifm_addrs, sa, rti_info);
991 
992 		if ((kif = calloc(1, sizeof(struct kif_node))) == NULL) {
993 			log_warn("fetchifs");
994 			free(buf);
995 			return (-1);
996 		}
997 
998 		kif->k.ifindex = ifm.ifm_index;
999 		kif->k.flags = ifm.ifm_flags;
1000 		kif->k.link_state = ifm.ifm_data.ifi_link_state;
1001 		kif->k.media_type = ifm.ifm_data.ifi_type;
1002 		kif->k.baudrate = ifm.ifm_data.ifi_baudrate;
1003 		kif->k.mtu = ifm.ifm_data.ifi_mtu;
1004 		kif->k.nh_reachable = (kif->k.flags & IFF_UP) &&
1005 		    (LINK_STATE_IS_UP(ifm.ifm_data.ifi_link_state) ||
1006 		    (ifm.ifm_data.ifi_link_state == LINK_STATE_UNKNOWN &&
1007 		    ifm.ifm_data.ifi_type != IFT_CARP));
1008 		if ((sa = rti_info[RTAX_IFP]) != NULL)
1009 			if (sa->sa_family == AF_LINK) {
1010 				sdl = (struct sockaddr_dl *)sa;
1011 				if (sdl->sdl_nlen >= sizeof(kif->k.ifname))
1012 					memcpy(kif->k.ifname, sdl->sdl_data,
1013 					    sizeof(kif->k.ifname) - 1);
1014 				else if (sdl->sdl_nlen > 0)
1015 					memcpy(kif->k.ifname, sdl->sdl_data,
1016 					    sdl->sdl_nlen);
1017 				/* string already terminated via calloc() */
1018 			}
1019 
1020 		kif_insert(kif);
1021 	}
1022 	free(buf);
1023 	return (0);
1024 }
1025 
1026 int
1027 dispatch_rtmsg(void)
1028 {
1029 	char			 buf[RT_BUF_SIZE];
1030 	ssize_t			 n;
1031 	char			*next, *lim;
1032 	struct rt_msghdr	*rtm;
1033 	struct if_msghdr	 ifm;
1034 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1035 	struct sockaddr_in	*sa_in;
1036 	struct sockaddr_rtlabel	*label;
1037 	struct kroute_node	*kr;
1038 	struct in_addr		 prefix, nexthop, netmask;
1039 	struct iface		*iface = NULL;
1040 	int			 flags;
1041 	u_short			 ifindex = 0;
1042 	u_int8_t		 metric;
1043 
1044 	if ((n = read(kr_state.fd, &buf, sizeof(buf))) == -1) {
1045 		log_warn("dispatch_rtmsg: read error");
1046 		return (-1);
1047 	}
1048 
1049 	if (n == 0) {
1050 		log_warnx("routing socket closed");
1051 		return (-1);
1052 	}
1053 
1054 	lim = buf + n;
1055 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1056 		rtm = (struct rt_msghdr *)next;
1057 		if (rtm->rtm_version != RTM_VERSION)
1058 			continue;
1059 
1060 		prefix.s_addr = 0;
1061 		netmask.s_addr = 0;
1062 		flags = F_KERNEL;
1063 		nexthop.s_addr = 0;
1064 
1065 		if (rtm->rtm_type == RTM_ADD || rtm->rtm_type == RTM_CHANGE ||
1066 		    rtm->rtm_type == RTM_DELETE) {
1067 			sa = (struct sockaddr *)(rtm + 1);
1068 			get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
1069 
1070 			if (rtm->rtm_tableid != 0)
1071 				continue;
1072 
1073 			if (rtm->rtm_pid == kr_state.pid)	/* cause by us */
1074 				continue;
1075 
1076 			if (rtm->rtm_errno)			/* failed attempts... */
1077 				continue;
1078 
1079 			if (rtm->rtm_flags & RTF_LLINFO)	/* arp cache */
1080 				continue;
1081 
1082 			switch (sa->sa_family) {
1083 			case AF_INET:
1084 				prefix.s_addr =
1085 				    ((struct sockaddr_in *)sa)->sin_addr.s_addr;
1086 				sa_in = (struct sockaddr_in *)
1087 				    rti_info[RTAX_NETMASK];
1088 				if (sa_in != NULL) {
1089 					if (sa_in->sin_len != 0)
1090 						netmask.s_addr =
1091 						    sa_in->sin_addr.s_addr;
1092 				} else if (rtm->rtm_flags & RTF_HOST)
1093 					netmask.s_addr = prefixlen2mask(32);
1094 				else
1095 					netmask.s_addr =
1096 					    prefixlen2mask(prefixlen_classful(
1097 						prefix.s_addr));
1098 				if (rtm->rtm_flags & RTF_STATIC)
1099 					flags |= F_STATIC;
1100 				if (rtm->rtm_flags & RTF_DYNAMIC)
1101 					flags |= F_DYNAMIC;
1102 				if (rtm->rtm_flags & RTF_PROTO1)
1103 					flags |= F_BGPD_INSERTED;
1104 				if (rtm->rtm_flags & RTF_PROTO2)
1105 					flags |= F_OSPFD_INSERTED;
1106 				break;
1107 			default:
1108 				continue;
1109 			}
1110 
1111 			ifindex = rtm->rtm_index;
1112 			if ((sa = rti_info[RTAX_GATEWAY]) != NULL) {
1113 				switch (sa->sa_family) {
1114 				case AF_INET:
1115 					nexthop.s_addr = ((struct
1116 					    sockaddr_in *)sa)->sin_addr.s_addr;
1117 					break;
1118 				case AF_LINK:
1119 					flags |= F_CONNECTED;
1120 					break;
1121 				}
1122 			}
1123 		}
1124 
1125 		switch (rtm->rtm_type) {
1126 		case RTM_ADD:
1127 		case RTM_CHANGE:
1128 			if (nexthop.s_addr == 0 && !(flags & F_CONNECTED)) {
1129 				log_warnx("dispatch_rtmsg no nexthop for %s/%u",
1130 				    inet_ntoa(prefix),
1131 				    mask2prefixlen(netmask.s_addr));
1132 				continue;
1133 			}
1134 
1135 			if ((kr = kroute_find(prefix.s_addr, netmask.s_addr)) !=
1136 			    NULL) {
1137 				/* rip route overridden by kernel */
1138 				/* pref is not checked because this is forced */
1139 				if (kr->r.flags & F_RIPD_INSERTED)
1140 					flags |= F_RIPD_INSERTED;
1141 				if (kr->r.flags & F_REDISTRIBUTED)
1142 					flags |= F_REDISTRIBUTED;
1143 				kr->r.nexthop.s_addr = nexthop.s_addr;
1144 				kr->r.flags = flags;
1145 				kr->r.ifindex = ifindex;
1146 
1147 				rtlabel_unref(kr->r.rtlabel);
1148 				kr->r.rtlabel = 0;
1149 				if ((label = (struct sockaddr_rtlabel *)
1150 				    rti_info[RTAX_LABEL]) != NULL)
1151 					kr->r.rtlabel =
1152 					    rtlabel_name2id(label->sr_label);
1153 
1154 				if (kif_validate(kr->r.ifindex))
1155 					kr->r.flags &= ~F_DOWN;
1156 				else
1157 					kr->r.flags |= F_DOWN;
1158 
1159 				/* just readd, the RDE will care */
1160 				kr_redistribute(IMSG_NETWORK_ADD, &kr->r);
1161 			} else {
1162 				if ((kr = calloc(1,
1163 				    sizeof(struct kroute_node))) == NULL) {
1164 					log_warn("dispatch_rtmsg");
1165 					return (-1);
1166 				}
1167 
1168 				iface = if_find_index(rtm->rtm_index);
1169 				if (iface != NULL)
1170 					metric = iface->cost;
1171 				else
1172 					metric = DEFAULT_COST;
1173 
1174 				kr->r.prefix.s_addr = prefix.s_addr;
1175 				kr->r.netmask.s_addr = netmask.s_addr;
1176 				kr->r.nexthop.s_addr = nexthop.s_addr;
1177 				kr->r.metric = metric;
1178 				kr->r.flags = flags;
1179 				kr->r.ifindex = ifindex;
1180 
1181 				if ((label = (struct sockaddr_rtlabel *)
1182 				    rti_info[RTAX_LABEL]) != NULL)
1183 					kr->r.rtlabel =
1184 					    rtlabel_name2id(label->sr_label);
1185 
1186 				kroute_insert(kr);
1187 			}
1188 			break;
1189 		case RTM_DELETE:
1190 			if ((kr = kroute_find(prefix.s_addr, netmask.s_addr)) ==
1191 			    NULL)
1192 				continue;
1193 			if (!(kr->r.flags & F_KERNEL))
1194 				continue;
1195 			if (kr->r.flags & F_RIPD_INSERTED)
1196 				main_imsg_compose_rde(IMSG_KROUTE_GET, 0,
1197 				    &kr->r, sizeof(struct kroute));
1198 			if (kroute_remove(kr) == -1)
1199 				return (-1);
1200 			break;
1201 		case RTM_IFINFO:
1202 			memcpy(&ifm, next, sizeof(ifm));
1203 			if_change(ifm.ifm_index, ifm.ifm_flags,
1204 			    &ifm.ifm_data);
1205 			break;
1206 		case RTM_IFANNOUNCE:
1207 			if_announce(next);
1208 			break;
1209 		default:
1210 			/* ignore for now */
1211 			break;
1212 		}
1213 	}
1214 	return (0);
1215 }
1216