xref: /openbsd/usr.sbin/ripd/kroute.c (revision 5b133f3f)
1 /*	$OpenBSD: kroute.c,v 1.36 2023/03/08 04:43:14 guenther Exp $ */
2 
3 /*
4  * Copyright (c) 2004 Esben Norby <norby@openbsd.org>
5  * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
6  *
7  * Permission to use, copy, modify, and distribute this software for any
8  * purpose with or without fee is hereby granted, provided that the above
9  * copyright notice and this permission notice appear in all copies.
10  *
11  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
14  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18  */
19 
20 #include <sys/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 	u_int8_t		fib_prio;
48 	int			fd;
49 	struct event		ev;
50 	u_int			rdomain;
51 } kr_state;
52 
53 struct kroute_node {
54 	RB_ENTRY(kroute_node)	 entry;
55 	struct kroute		 r;
56 };
57 
58 struct kif_node {
59 	RB_ENTRY(kif_node)	 entry;
60 	struct kif		 k;
61 };
62 
63 void	kr_redistribute(int, struct kroute *);
64 int	kroute_compare(struct kroute_node *, struct kroute_node *);
65 int	kif_compare(struct kif_node *, struct kif_node *);
66 int	kr_change_fib(struct kroute_node *, struct kroute *, int);
67 
68 struct kroute_node	*kroute_find(in_addr_t, in_addr_t, u_int8_t);
69 int			 kroute_insert(struct kroute_node *);
70 int			 kroute_remove(struct kroute_node *);
71 void			 kroute_clear(void);
72 
73 struct kif_node		*kif_find(int);
74 int			 kif_insert(struct kif_node *);
75 int			 kif_remove(struct kif_node *);
76 void			 kif_clear(void);
77 int			 kif_validate(int);
78 
79 struct kroute_node	*kroute_match(in_addr_t);
80 
81 int		protect_lo(void);
82 u_int8_t	prefixlen_classful(in_addr_t);
83 void		get_rtaddrs(int, struct sockaddr *, struct sockaddr **);
84 void		if_change(u_short, int, struct if_data *);
85 void		if_announce(void *);
86 
87 int		send_rtmsg(int, int, struct kroute *);
88 int		dispatch_rtmsg(void);
89 int		fetchtable(void);
90 int		fetchifs(int);
91 
92 RB_HEAD(kroute_tree, kroute_node)	krt;
93 RB_PROTOTYPE(kroute_tree, kroute_node, entry, kroute_compare)
94 RB_GENERATE(kroute_tree, kroute_node, entry, kroute_compare)
95 
96 RB_HEAD(kif_tree, kif_node)		kit;
RB_PROTOTYPE(kif_tree,kif_node,entry,kif_compare)97 RB_PROTOTYPE(kif_tree, kif_node, entry, kif_compare)
98 RB_GENERATE(kif_tree, kif_node, entry, kif_compare)
99 
100 int
101 kif_init(void)
102 {
103 	RB_INIT(&kit);
104 
105 	if (fetchifs(0) == -1)
106 		return (-1);
107 
108 	return (0);
109 }
110 
111 int
kr_init(int fs,u_int rdomain,u_int8_t fib_prio)112 kr_init(int fs, u_int rdomain, u_int8_t fib_prio)
113 {
114 	int		opt = 0, rcvbuf, default_rcvbuf;
115 	socklen_t	optlen;
116 
117 	if ((kr_state.fd = socket(AF_ROUTE,
118 	    SOCK_RAW | SOCK_CLOEXEC | SOCK_NONBLOCK, 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 	kr_state.fib_prio = fib_prio;
144 
145 	RB_INIT(&krt);
146 
147 	if (fetchtable() == -1)
148 		return (-1);
149 
150 	if (protect_lo() == -1)
151 		return (-1);
152 
153 	kr_state.fib_sync = fs; /* now set correct sync mode */
154 	kr_state.rdomain = rdomain;
155 
156 	event_set(&kr_state.ev, kr_state.fd, EV_READ | EV_PERSIST,
157 	    kr_dispatch_msg, NULL);
158 	event_add(&kr_state.ev, NULL);
159 
160 	return (0);
161 }
162 
163 int
kr_change_fib(struct kroute_node * kr,struct kroute * kroute,int action)164 kr_change_fib(struct kroute_node *kr, struct kroute *kroute, int action)
165 {
166 	/* nexthop within 127/8 -> ignore silently */
167 	if ((kroute->nexthop.s_addr & htonl(IN_CLASSA_NET)) ==
168 	    htonl(INADDR_LOOPBACK & IN_CLASSA_NET))
169 		return (0);
170 
171 	if (send_rtmsg(kr_state.fd, action, kroute) == -1)
172 		return (-1);
173 
174 	if (action == RTM_ADD) {
175 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL)
176 			fatal("kr_change_fib");
177 
178 		kr->r.prefix.s_addr = kroute->prefix.s_addr;
179 		kr->r.netmask.s_addr = kroute->netmask.s_addr;
180 		kr->r.nexthop.s_addr = kroute->nexthop.s_addr;
181 		kr->r.flags = kroute->flags |= F_RIPD_INSERTED;
182 		kr->r.priority = kr_state.fib_prio;
183 
184 		if (kroute_insert(kr) == -1) {
185 			log_debug("kr_update_fib: cannot insert %s",
186 			    inet_ntoa(kroute->nexthop));
187 		}
188 	} else
189 		kr->r.nexthop.s_addr = kroute->nexthop.s_addr;
190 
191 	return (0);
192 }
193 
194 int
kr_change(struct kroute * kroute)195 kr_change(struct kroute *kroute)
196 {
197 	struct kroute_node	*kr;
198 	int			 action = RTM_ADD;
199 
200 	kr = kroute_find(kroute->prefix.s_addr, kroute->netmask.s_addr,
201 	    kr_state.fib_prio);
202 	if (kr != NULL)
203 		action = RTM_CHANGE;
204 
205 	return (kr_change_fib(kr, kroute, action));
206 }
207 
208 int
kr_delete(struct kroute * kroute)209 kr_delete(struct kroute *kroute)
210 {
211 	struct kroute_node	*kr;
212 
213 	kr = kroute_find(kroute->prefix.s_addr, kroute->netmask.s_addr,
214 	    kr_state.fib_prio);
215 	if (kr == NULL)
216 		return (0);
217 
218 	if (kr->r.priority != kr_state.fib_prio)
219 		log_warn("kr_delete_fib: %s/%d has wrong priority %d",
220 		    inet_ntoa(kr->r.prefix), mask2prefixlen(kr->r.netmask.s_addr),
221 		    kr->r.priority);
222 
223 	if (send_rtmsg(kr_state.fd, RTM_DELETE, kroute) == -1)
224 		return (-1);
225 
226 	if (kroute_remove(kr) == -1)
227 		return (-1);
228 
229 	return (0);
230 }
231 
232 void
kr_shutdown(void)233 kr_shutdown(void)
234 {
235 	kr_fib_decouple();
236 
237 	kroute_clear();
238 	kif_clear();
239 }
240 
241 void
kr_fib_couple(void)242 kr_fib_couple(void)
243 {
244 	struct kroute_node	*kr;
245 
246 	if (kr_state.fib_sync == 1)	/* already coupled */
247 		return;
248 
249 	kr_state.fib_sync = 1;
250 
251 	RB_FOREACH(kr, kroute_tree, &krt)
252 		if (kr->r.priority == kr_state.fib_prio)
253 			send_rtmsg(kr_state.fd, RTM_ADD, &kr->r);
254 
255 	log_info("kernel routing table coupled");
256 }
257 
258 void
kr_fib_decouple(void)259 kr_fib_decouple(void)
260 {
261 	struct kroute_node	*kr;
262 
263 	if (kr_state.fib_sync == 0)	/* already decoupled */
264 		return;
265 
266 	RB_FOREACH(kr, kroute_tree, &krt)
267 		if (kr->r.priority == kr_state.fib_prio)
268 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
269 
270 	kr_state.fib_sync = 0;
271 
272 	log_info("kernel routing table decoupled");
273 }
274 
275 void
kr_dispatch_msg(int fd,short event,void * bula)276 kr_dispatch_msg(int fd, short event, void *bula)
277 {
278 	dispatch_rtmsg();
279 }
280 
281 void
kr_show_route(struct imsg * imsg)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
kr_ifinfo(char * ifname,pid_t pid)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
kr_redistribute(int type,struct kroute * kr)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 and multicast addresses
360 	 * as not redistributable.
361 	 */
362 	a = ntohl(kr->prefix.s_addr);
363 	if (IN_MULTICAST(a) || (a >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
364 		return;
365 	/*
366 	 * Consider networks with nexthop loopback as not redistributable
367 	 * unless it is a reject or blackhole route.
368 	 */
369 	if (kr->nexthop.s_addr == htonl(INADDR_LOOPBACK) &&
370 	    !(kr->flags & (F_BLACKHOLE|F_REJECT)))
371 		return;
372 
373 	/* Should we redistribute this route? */
374 	if (!rip_redistribute(kr))
375 		goto dont_redistribute;
376 
377 	/* Does not matter if we resend the kr, the RDE will cope. */
378 	kr->flags |= F_REDISTRIBUTED;
379 	main_imsg_compose_rde(type, 0, kr, sizeof(struct kroute));
380 }
381 
382 /* rb-tree compare */
383 int
kroute_compare(struct kroute_node * a,struct kroute_node * b)384 kroute_compare(struct kroute_node *a, struct kroute_node *b)
385 {
386 	if (ntohl(a->r.prefix.s_addr) < ntohl(b->r.prefix.s_addr))
387 		return (-1);
388 	if (ntohl(a->r.prefix.s_addr) > ntohl(b->r.prefix.s_addr))
389 		return (1);
390 	if (ntohl(a->r.netmask.s_addr) < ntohl(b->r.netmask.s_addr))
391 		return (-1);
392 	if (ntohl(a->r.netmask.s_addr) > ntohl(b->r.netmask.s_addr))
393 		return (1);
394 
395 	/* if the priority is RTP_ANY finish on the first address hit */
396 	if (a->r.priority == RTP_ANY || b->r.priority == RTP_ANY)
397 		return (0);
398 	if (a->r.priority < b->r.priority)
399 		return (-1);
400 	if (a->r.priority > b->r.priority)
401 		return (1);
402 
403 	return (0);
404 }
405 
406 int
kif_compare(struct kif_node * a,struct kif_node * b)407 kif_compare(struct kif_node *a, struct kif_node *b)
408 {
409 	return (b->k.ifindex - a->k.ifindex);
410 }
411 
412 /* tree management */
413 struct kroute_node *
kroute_find(in_addr_t prefix,in_addr_t netmask,u_int8_t prio)414 kroute_find(in_addr_t prefix, in_addr_t netmask, u_int8_t prio)
415 {
416 	struct kroute_node	s, *kn, *tmp;
417 
418 	s.r.prefix.s_addr = prefix;
419 	s.r.netmask.s_addr = netmask;
420 	s.r.priority = prio;
421 
422 	kn = RB_FIND(kroute_tree, &krt, &s);
423 	if (kn && prio == RTP_ANY) {
424 		tmp = RB_PREV(kroute_tree, &krt, kn);
425 		while (tmp) {
426 			if (kroute_compare(&s, tmp) == 0)
427 				kn = tmp;
428 			else
429 				break;
430 			tmp = RB_PREV(kroute_tree, &krt, kn);
431 		}
432 	}
433 
434 	return (kn);
435 }
436 
437 int
kroute_insert(struct kroute_node * kr)438 kroute_insert(struct kroute_node *kr)
439 {
440 	if (RB_INSERT(kroute_tree, &krt, kr) != NULL) {
441 		log_warnx("kroute_insert failed for %s/%u",
442 		    inet_ntoa(kr->r.prefix),
443 		    mask2prefixlen(kr->r.netmask.s_addr));
444 		free(kr);
445 		return (-1);
446 	}
447 
448 	if (!(kr->r.flags & F_KERNEL)) {
449 		/* don't validate or redistribute rip route */
450 		kr->r.flags &= ~F_DOWN;
451 		return (0);
452 	}
453 
454 	if (kif_validate(kr->r.ifindex))
455 		kr->r.flags &= ~F_DOWN;
456 	else
457 		kr->r.flags |= F_DOWN;
458 
459 	kr_redistribute(IMSG_NETWORK_ADD, &kr->r);
460 
461 	return (0);
462 }
463 
464 int
kroute_remove(struct kroute_node * kr)465 kroute_remove(struct kroute_node *kr)
466 {
467 	if (RB_REMOVE(kroute_tree, &krt, kr) == NULL) {
468 		log_warnx("kroute_remove failed for %s/%u",
469 		    inet_ntoa(kr->r.prefix),
470 		    mask2prefixlen(kr->r.netmask.s_addr));
471 		return (-1);
472 	}
473 
474 	kr_redistribute(IMSG_NETWORK_DEL, &kr->r);
475 	rtlabel_unref(kr->r.rtlabel);
476 
477 	free(kr);
478 	return (0);
479 }
480 
481 void
kroute_clear(void)482 kroute_clear(void)
483 {
484 	struct kroute_node	*kr;
485 
486 	while ((kr = RB_MIN(kroute_tree, &krt)) != NULL)
487 		kroute_remove(kr);
488 }
489 
490 struct kif_node *
kif_find(int ifindex)491 kif_find(int ifindex)
492 {
493 	struct kif_node	s;
494 
495 	bzero(&s, sizeof(s));
496 	s.k.ifindex = ifindex;
497 
498 	return (RB_FIND(kif_tree, &kit, &s));
499 }
500 
501 struct kif *
kif_findname(char * ifname)502 kif_findname(char *ifname)
503 {
504 	struct kif_node	*kif;
505 
506 	RB_FOREACH(kif, kif_tree, &kit)
507 		if (!strcmp(ifname, kif->k.ifname))
508 			return (&kif->k);
509 
510 	return (NULL);
511 }
512 
513 int
kif_insert(struct kif_node * kif)514 kif_insert(struct kif_node *kif)
515 {
516 	if (RB_INSERT(kif_tree, &kit, kif) != NULL) {
517 		log_warnx("RB_INSERT(kif_tree, &kit, kif)");
518 		free(kif);
519 		return (-1);
520 	}
521 
522 	return (0);
523 }
524 
525 int
kif_remove(struct kif_node * kif)526 kif_remove(struct kif_node *kif)
527 {
528 	if (RB_REMOVE(kif_tree, &kit, kif) == NULL) {
529 		log_warnx("RB_REMOVE(kif_tree, &kit, kif)");
530 		return (-1);
531 	}
532 
533 	free(kif);
534 	return (0);
535 }
536 
537 void
kif_clear(void)538 kif_clear(void)
539 {
540 	struct kif_node	*kif;
541 
542 	while ((kif = RB_MIN(kif_tree, &kit)) != NULL)
543 		kif_remove(kif);
544 }
545 
546 int
kif_validate(int ifindex)547 kif_validate(int ifindex)
548 {
549 	struct kif_node		*kif;
550 
551 	if ((kif = kif_find(ifindex)) == NULL) {
552 		log_warnx("interface with index %u not found", ifindex);
553 		return (1);
554 	}
555 
556 	return (kif->k.nh_reachable);
557 }
558 
559 struct kroute_node *
kroute_match(in_addr_t key)560 kroute_match(in_addr_t key)
561 {
562 	u_int8_t		 i;
563 	struct kroute_node	*kr;
564 
565 	/* we will never match the default route */
566 	for (i = 32; i > 0; i--)
567 		if ((kr = kroute_find(key & prefixlen2mask(i),
568 		    prefixlen2mask(i), RTP_ANY)) != NULL)
569 			return (kr);
570 
571 	/* if we don't have a match yet, try to find a default route */
572 	if ((kr = kroute_find(0, 0, RTP_ANY)) != NULL)
573 			return (kr);
574 
575 	return (NULL);
576 }
577 
578 /* misc */
579 int
protect_lo(void)580 protect_lo(void)
581 {
582 	struct kroute_node	*kr;
583 
584 	/* special protection for 127/8 */
585 	if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
586 		log_warn("protect_lo");
587 		return (-1);
588 	}
589 	kr->r.prefix.s_addr = htonl(INADDR_LOOPBACK);
590 	kr->r.netmask.s_addr = htonl(IN_CLASSA_NET);
591 	kr->r.flags = F_KERNEL|F_CONNECTED;
592 
593 	if (RB_INSERT(kroute_tree, &krt, kr) != NULL)
594 		free(kr);	/* kernel route already there, no problem */
595 
596 	return (0);
597 }
598 
599 u_int8_t
prefixlen_classful(in_addr_t ina)600 prefixlen_classful(in_addr_t ina)
601 {
602 	/* it hurt to write this. */
603 
604 	if (ina >= 0xf0000000U)		/* class E */
605 		return (32);
606 	else if (ina >= 0xe0000000U)	/* class D */
607 		return (4);
608 	else if (ina >= 0xc0000000U)	/* class C */
609 		return (24);
610 	else if (ina >= 0x80000000U)	/* class B */
611 		return (16);
612 	else				/* class A */
613 		return (8);
614 }
615 
616 u_int8_t
mask2prefixlen(in_addr_t ina)617 mask2prefixlen(in_addr_t ina)
618 {
619 	if (ina == 0)
620 		return (0);
621 	else
622 		return (33 - ffs(ntohl(ina)));
623 }
624 
625 in_addr_t
prefixlen2mask(u_int8_t prefixlen)626 prefixlen2mask(u_int8_t prefixlen)
627 {
628 	if (prefixlen == 0)
629 		return (0);
630 
631 	return (htonl(0xffffffff << (32 - prefixlen)));
632 }
633 
634 #define ROUNDUP(a) \
635 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
636 
637 void
get_rtaddrs(int addrs,struct sockaddr * sa,struct sockaddr ** rti_info)638 get_rtaddrs(int addrs, struct sockaddr *sa, struct sockaddr **rti_info)
639 {
640 	int	i;
641 
642 	for (i = 0; i < RTAX_MAX; i++) {
643 		if (addrs & (1 << i)) {
644 			rti_info[i] = sa;
645 			sa = (struct sockaddr *)((char *)(sa) +
646 			    ROUNDUP(sa->sa_len));
647 		} else
648 			rti_info[i] = NULL;
649 	}
650 }
651 
652 void
if_change(u_short ifindex,int flags,struct if_data * ifd)653 if_change(u_short ifindex, int flags, struct if_data *ifd)
654 {
655 	struct kif_node		*kif;
656 	struct kroute_node	*kr;
657 	int			 type;
658 	u_int8_t		 reachable;
659 
660 	if ((kif = kif_find(ifindex)) == NULL) {
661 		log_warnx("interface with index %u not found", ifindex);
662 		return;
663 	}
664 
665 	kif->k.flags = flags;
666 	kif->k.link_state = ifd->ifi_link_state;
667 	kif->k.if_type = ifd->ifi_type;
668 	kif->k.baudrate = ifd->ifi_baudrate;
669 
670 	if ((reachable = (flags & IFF_UP) &&
671 	    LINK_STATE_IS_UP(ifd->ifi_link_state)) == kif->k.nh_reachable)
672 		return;		/* nothing changed wrt nexthop validity */
673 
674 	kif->k.nh_reachable = reachable;
675 	type = reachable ? IMSG_NETWORK_ADD : IMSG_NETWORK_DEL;
676 
677 	/* notify ripe about interface link state */
678 	main_imsg_compose_ripe(IMSG_IFINFO, 0, &kif->k, sizeof(kif->k));
679 
680 	/* update redistribute list */
681 	RB_FOREACH(kr, kroute_tree, &krt)
682 		if (kr->r.ifindex == ifindex) {
683 			if (reachable)
684 				kr->r.flags &= ~F_DOWN;
685 			else
686 				kr->r.flags |= F_DOWN;
687 
688 			kr_redistribute(type, &kr->r);
689 		}
690 }
691 
692 void
if_announce(void * msg)693 if_announce(void *msg)
694 {
695 	struct if_announcemsghdr	*ifan;
696 	struct kif_node			*kif;
697 
698 	ifan = msg;
699 
700 	switch (ifan->ifan_what) {
701 	case IFAN_ARRIVAL:
702 		if ((kif = calloc(1, sizeof(struct kif_node))) == NULL) {
703 			log_warn("if_announce");
704 			return;
705 		}
706 
707 		kif->k.ifindex = ifan->ifan_index;
708 		strlcpy(kif->k.ifname, ifan->ifan_name, sizeof(kif->k.ifname));
709 		kif_insert(kif);
710 		break;
711 	case IFAN_DEPARTURE:
712 		kif = kif_find(ifan->ifan_index);
713 		kif_remove(kif);
714 		break;
715 	}
716 }
717 
718 /* rtsock */
719 int
send_rtmsg(int fd,int action,struct kroute * kroute)720 send_rtmsg(int fd, int action, struct kroute *kroute)
721 {
722 	struct iovec		iov[4];
723 	struct rt_msghdr	hdr;
724 	struct sockaddr_in	prefix;
725 	struct sockaddr_in	nexthop;
726 	struct sockaddr_in	mask;
727 	int			iovcnt = 0;
728 
729 	if (kr_state.fib_sync == 0)
730 		return (0);
731 
732 	/* initialize header */
733 	bzero(&hdr, sizeof(hdr));
734 	hdr.rtm_version = RTM_VERSION;
735 	hdr.rtm_type = action;
736 	hdr.rtm_priority = kr_state.fib_prio;
737 	hdr.rtm_tableid = kr_state.rdomain;
738 	if (action == RTM_CHANGE)
739 		hdr.rtm_fmask = RTF_REJECT|RTF_BLACKHOLE;
740 	hdr.rtm_seq = kr_state.rtseq++;	/* overflow doesn't matter */
741 	hdr.rtm_msglen = sizeof(hdr);
742 	/* adjust iovec */
743 	iov[iovcnt].iov_base = &hdr;
744 	iov[iovcnt++].iov_len = sizeof(hdr);
745 
746 	bzero(&prefix, sizeof(prefix));
747 	prefix.sin_len = sizeof(prefix);
748 	prefix.sin_family = AF_INET;
749 	prefix.sin_addr.s_addr = kroute->prefix.s_addr;
750 	/* adjust header */
751 	hdr.rtm_addrs |= RTA_DST;
752 	hdr.rtm_msglen += sizeof(prefix);
753 	/* adjust iovec */
754 	iov[iovcnt].iov_base = &prefix;
755 	iov[iovcnt++].iov_len = sizeof(prefix);
756 
757 	if (kroute->nexthop.s_addr != 0) {
758 		bzero(&nexthop, sizeof(nexthop));
759 		nexthop.sin_len = sizeof(nexthop);
760 		nexthop.sin_family = AF_INET;
761 		nexthop.sin_addr.s_addr = kroute->nexthop.s_addr;
762 		/* adjust header */
763 		hdr.rtm_flags |= RTF_GATEWAY;
764 		hdr.rtm_addrs |= RTA_GATEWAY;
765 		hdr.rtm_msglen += sizeof(nexthop);
766 		/* adjust iovec */
767 		iov[iovcnt].iov_base = &nexthop;
768 		iov[iovcnt++].iov_len = sizeof(nexthop);
769 	}
770 
771 	bzero(&mask, sizeof(mask));
772 	mask.sin_len = sizeof(mask);
773 	mask.sin_family = AF_INET;
774 	mask.sin_addr.s_addr = kroute->netmask.s_addr;
775 	/* adjust header */
776 	hdr.rtm_addrs |= RTA_NETMASK;
777 	hdr.rtm_msglen += sizeof(mask);
778 	/* adjust iovec */
779 	iov[iovcnt].iov_base = &mask;
780 	iov[iovcnt++].iov_len = sizeof(mask);
781 
782 
783 retry:
784 	if (writev(fd, iov, iovcnt) == -1) {
785 		if (errno == ESRCH) {
786 			if (hdr.rtm_type == RTM_CHANGE) {
787 				hdr.rtm_type = RTM_ADD;
788 				goto retry;
789 			} else if (hdr.rtm_type == RTM_DELETE) {
790 				log_info("route %s/%u vanished before delete",
791 				    inet_ntoa(kroute->prefix),
792 				    mask2prefixlen(kroute->netmask.s_addr));
793 				return (0);
794 			}
795 		}
796 		log_warn("send_rtmsg: action %u, prefix %s/%u",
797 		    hdr.rtm_type, inet_ntoa(kroute->prefix),
798 		    mask2prefixlen(kroute->netmask.s_addr));
799 		return (0);
800 	}
801 
802 	return (0);
803 }
804 
805 int
fetchtable(void)806 fetchtable(void)
807 {
808 	size_t			 len;
809 	int			 mib[7];
810 	char			*buf, *next, *lim;
811 	struct rt_msghdr	*rtm;
812 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
813 	struct sockaddr_in	*sa_in;
814 	struct sockaddr_rtlabel	*label;
815 	struct kroute_node	*kr;
816 	struct iface		*iface = NULL;
817 
818 	mib[0] = CTL_NET;
819 	mib[1] = PF_ROUTE;
820 	mib[2] = 0;
821 	mib[3] = AF_INET;
822 	mib[4] = NET_RT_DUMP;
823 	mib[5] = 0;
824 	mib[6] = kr_state.rdomain;	/* rtableid */
825 
826 	if (sysctl(mib, 7, NULL, &len, NULL, 0) == -1) {
827 		log_warn("sysctl");
828 		return (-1);
829 	}
830 	if ((buf = malloc(len)) == NULL) {
831 		log_warn("fetchtable");
832 		return (-1);
833 	}
834 	if (sysctl(mib, 7, buf, &len, NULL, 0) == -1) {
835 		log_warn("sysctl");
836 		free(buf);
837 		return (-1);
838 	}
839 
840 	lim = buf + len;
841 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
842 		rtm = (struct rt_msghdr *)next;
843 		if (rtm->rtm_version != RTM_VERSION)
844 			continue;
845 		sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
846 		get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
847 
848 		if ((sa = rti_info[RTAX_DST]) == NULL)
849 			continue;
850 
851 		/* Skip ARP/ND cache and broadcast routes. */
852 		if (rtm->rtm_flags & (RTF_LLINFO|RTF_BROADCAST))
853 			continue;
854 
855 #ifdef RTF_MPATH
856 		if (rtm->rtm_flags & RTF_MPATH)		/* multipath */
857 			continue;
858 #endif
859 
860 		if ((kr = calloc(1, sizeof(struct kroute_node))) == NULL) {
861 			log_warn("fetchtable");
862 			free(buf);
863 			return (-1);
864 		}
865 
866 		kr->r.flags = F_KERNEL;
867 		kr->r.priority = rtm->rtm_priority;
868 
869 		switch (sa->sa_family) {
870 		case AF_INET:
871 			kr->r.prefix.s_addr =
872 			    ((struct sockaddr_in *)sa)->sin_addr.s_addr;
873 			sa_in = (struct sockaddr_in *)rti_info[RTAX_NETMASK];
874 			if (rtm->rtm_flags & RTF_STATIC)
875 				kr->r.flags |= F_STATIC;
876 			if (rtm->rtm_flags & RTF_BLACKHOLE)
877 				kr->r.flags |= F_BLACKHOLE;
878 			if (rtm->rtm_flags & RTF_REJECT)
879 				kr->r.flags |= F_REJECT;
880 			if (rtm->rtm_flags & RTF_DYNAMIC)
881 				kr->r.flags |= F_DYNAMIC;
882 			if (sa_in != NULL) {
883 				if (sa_in->sin_len == 0)
884 					break;
885 				kr->r.netmask.s_addr =
886 				    sa_in->sin_addr.s_addr;
887 			} else if (rtm->rtm_flags & RTF_HOST)
888 				kr->r.netmask.s_addr = prefixlen2mask(32);
889 			else
890 				kr->r.netmask.s_addr =
891 				    prefixlen2mask(prefixlen_classful
892 					(kr->r.prefix.s_addr));
893 			break;
894 		default:
895 			free(kr);
896 			continue;
897 		}
898 
899 		kr->r.ifindex = rtm->rtm_index;
900 
901 		iface = if_find_index(rtm->rtm_index);
902 		if (iface != NULL)
903 			kr->r.metric = iface->cost;
904 		else
905 			kr->r.metric = DEFAULT_COST;
906 
907 		if ((sa = rti_info[RTAX_GATEWAY]) != NULL)
908 			switch (sa->sa_family) {
909 			case AF_INET:
910 				if (rtm->rtm_flags & RTF_CONNECTED) {
911 					kr->r.flags |= F_CONNECTED;
912 					break;
913 				}
914 
915 				kr->r.nexthop.s_addr =
916 				    ((struct sockaddr_in *)sa)->sin_addr.s_addr;
917 				break;
918 			case AF_LINK:
919 				/*
920 				 * Traditional BSD connected routes have
921 				 * a gateway of type AF_LINK.
922 				 */
923 				kr->r.flags |= F_CONNECTED;
924 				break;
925 			}
926 
927 		if (rtm->rtm_priority == kr_state.fib_prio) {
928 			send_rtmsg(kr_state.fd, RTM_DELETE, &kr->r);
929 			free(kr);
930 		} else {
931 			if ((label = (struct sockaddr_rtlabel *)
932 			    rti_info[RTAX_LABEL]) != NULL)
933 				kr->r.rtlabel =
934 				    rtlabel_name2id(label->sr_label);
935 			kroute_insert(kr);
936 		}
937 
938 	}
939 	free(buf);
940 	return (0);
941 }
942 
943 int
fetchifs(int ifindex)944 fetchifs(int ifindex)
945 {
946 	size_t			 len;
947 	int			 mib[6];
948 	char			*buf, *next, *lim;
949 	struct if_msghdr	 ifm;
950 	struct kif_node		*kif;
951 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
952 	struct sockaddr_dl	*sdl;
953 
954 	mib[0] = CTL_NET;
955 	mib[1] = PF_ROUTE;
956 	mib[2] = 0;
957 	mib[3] = AF_INET;
958 	mib[4] = NET_RT_IFLIST;
959 	mib[5] = ifindex;
960 
961 	if (sysctl(mib, 6, NULL, &len, NULL, 0) == -1) {
962 		log_warn("sysctl");
963 		return (-1);
964 	}
965 	if ((buf = malloc(len)) == NULL) {
966 		log_warn("fetchif");
967 		return (-1);
968 	}
969 	if (sysctl(mib, 6, buf, &len, NULL, 0) == -1) {
970 		log_warn("sysctl");
971 		free(buf);
972 		return (-1);
973 	}
974 
975 	lim = buf + len;
976 	for (next = buf; next < lim; next += ifm.ifm_msglen) {
977 		memcpy(&ifm, next, sizeof(ifm));
978 		if (ifm.ifm_version != RTM_VERSION)
979 			continue;
980 		if (ifm.ifm_type != RTM_IFINFO)
981 			continue;
982 
983 		sa = (struct sockaddr *)(next + sizeof(ifm));
984 		get_rtaddrs(ifm.ifm_addrs, sa, rti_info);
985 
986 		if ((kif = calloc(1, sizeof(struct kif_node))) == NULL) {
987 			log_warn("fetchifs");
988 			free(buf);
989 			return (-1);
990 		}
991 
992 		kif->k.ifindex = ifm.ifm_index;
993 		kif->k.flags = ifm.ifm_flags;
994 		kif->k.link_state = ifm.ifm_data.ifi_link_state;
995 		kif->k.if_type = ifm.ifm_data.ifi_type;
996 		kif->k.baudrate = ifm.ifm_data.ifi_baudrate;
997 		kif->k.mtu = ifm.ifm_data.ifi_mtu;
998 		kif->k.nh_reachable = (kif->k.flags & IFF_UP) &&
999 		    LINK_STATE_IS_UP(ifm.ifm_data.ifi_link_state);
1000 		if ((sa = rti_info[RTAX_IFP]) != NULL)
1001 			if (sa->sa_family == AF_LINK) {
1002 				sdl = (struct sockaddr_dl *)sa;
1003 				if (sdl->sdl_nlen >= sizeof(kif->k.ifname))
1004 					memcpy(kif->k.ifname, sdl->sdl_data,
1005 					    sizeof(kif->k.ifname) - 1);
1006 				else if (sdl->sdl_nlen > 0)
1007 					memcpy(kif->k.ifname, sdl->sdl_data,
1008 					    sdl->sdl_nlen);
1009 				/* string already terminated via calloc() */
1010 			}
1011 
1012 		kif_insert(kif);
1013 	}
1014 	free(buf);
1015 	return (0);
1016 }
1017 
1018 int
dispatch_rtmsg(void)1019 dispatch_rtmsg(void)
1020 {
1021 	char			 buf[RT_BUF_SIZE];
1022 	ssize_t			 n;
1023 	char			*next, *lim;
1024 	struct rt_msghdr	*rtm;
1025 	struct if_msghdr	 ifm;
1026 	struct sockaddr		*sa, *rti_info[RTAX_MAX];
1027 	struct sockaddr_in	*sa_in;
1028 	struct sockaddr_rtlabel	*label;
1029 	struct kroute_node	*kr;
1030 	struct in_addr		 prefix, nexthop, netmask;
1031 	struct iface		*iface = NULL;
1032 	int			 flags;
1033 	u_short			 ifindex = 0;
1034 	u_int8_t		 metric, prio;
1035 
1036 	if ((n = read(kr_state.fd, &buf, sizeof(buf))) == -1) {
1037 		if (errno == EAGAIN || errno == EINTR)
1038 			return (0);
1039 		log_warn("dispatch_rtmsg: read error");
1040 		return (-1);
1041 	}
1042 
1043 	if (n == 0) {
1044 		log_warnx("routing socket closed");
1045 		return (-1);
1046 	}
1047 
1048 	lim = buf + n;
1049 	for (next = buf; next < lim; next += rtm->rtm_msglen) {
1050 		rtm = (struct rt_msghdr *)next;
1051 		if (lim < next + sizeof(u_short) ||
1052 		    lim < next + rtm->rtm_msglen)
1053 			fatalx("dispatch_rtmsg: partial rtm in buffer");
1054 		if (rtm->rtm_version != RTM_VERSION)
1055 			continue;
1056 
1057 		prefix.s_addr = 0;
1058 		netmask.s_addr = 0;
1059 		flags = F_KERNEL;
1060 		nexthop.s_addr = 0;
1061 		prio = 0;
1062 
1063 		if (rtm->rtm_type == RTM_ADD || rtm->rtm_type == RTM_CHANGE ||
1064 		    rtm->rtm_type == RTM_DELETE) {
1065 			sa = (struct sockaddr *)(next + rtm->rtm_hdrlen);
1066 			get_rtaddrs(rtm->rtm_addrs, sa, rti_info);
1067 
1068 			if (rtm->rtm_tableid != kr_state.rdomain)
1069 				continue;
1070 
1071 			if (rtm->rtm_pid == kr_state.pid)	/* cause by us */
1072 				continue;
1073 
1074 			if (rtm->rtm_errno)			/* failed attempts... */
1075 				continue;
1076 
1077 			/* Skip ARP/ND cache and broadcast routes. */
1078 			if (rtm->rtm_flags & (RTF_LLINFO|RTF_BROADCAST))
1079 				continue;
1080 
1081 			prio = rtm->rtm_priority;
1082 
1083 			switch (sa->sa_family) {
1084 			case AF_INET:
1085 				prefix.s_addr =
1086 				    ((struct sockaddr_in *)sa)->sin_addr.s_addr;
1087 				sa_in = (struct sockaddr_in *)
1088 				    rti_info[RTAX_NETMASK];
1089 				if (sa_in != NULL) {
1090 					if (sa_in->sin_len != 0)
1091 						netmask.s_addr =
1092 						    sa_in->sin_addr.s_addr;
1093 				} else if (rtm->rtm_flags & RTF_HOST)
1094 					netmask.s_addr = prefixlen2mask(32);
1095 				else
1096 					netmask.s_addr =
1097 					    prefixlen2mask(prefixlen_classful(
1098 						prefix.s_addr));
1099 				if (rtm->rtm_flags & RTF_STATIC)
1100 					flags |= F_STATIC;
1101 				if (rtm->rtm_flags & RTF_BLACKHOLE)
1102 					flags |= F_BLACKHOLE;
1103 				if (rtm->rtm_flags & RTF_REJECT)
1104 					flags |= F_REJECT;
1105 				if (rtm->rtm_flags & RTF_DYNAMIC)
1106 					flags |= F_DYNAMIC;
1107 				break;
1108 			default:
1109 				continue;
1110 			}
1111 
1112 			ifindex = rtm->rtm_index;
1113 			if ((sa = rti_info[RTAX_GATEWAY]) != NULL) {
1114 				switch (sa->sa_family) {
1115 				case AF_INET:
1116 					nexthop.s_addr = ((struct
1117 					    sockaddr_in *)sa)->sin_addr.s_addr;
1118 					break;
1119 				case AF_LINK:
1120 					flags |= F_CONNECTED;
1121 					break;
1122 				}
1123 			}
1124 		}
1125 
1126 		switch (rtm->rtm_type) {
1127 		case RTM_ADD:
1128 		case RTM_CHANGE:
1129 			if (nexthop.s_addr == 0 && !(flags & F_CONNECTED)) {
1130 				log_warnx("dispatch_rtmsg no nexthop for %s/%u",
1131 				    inet_ntoa(prefix),
1132 				    mask2prefixlen(netmask.s_addr));
1133 				continue;
1134 			}
1135 
1136 			if ((kr = kroute_find(prefix.s_addr, netmask.s_addr,
1137 			    prio)) != NULL) {
1138 				if (kr->r.flags & F_REDISTRIBUTED)
1139 					flags |= F_REDISTRIBUTED;
1140 				kr->r.nexthop.s_addr = nexthop.s_addr;
1141 				kr->r.flags = flags;
1142 				kr->r.ifindex = ifindex;
1143 				kr->r.priority = prio;
1144 
1145 				rtlabel_unref(kr->r.rtlabel);
1146 				kr->r.rtlabel = 0;
1147 				if ((label = (struct sockaddr_rtlabel *)
1148 				    rti_info[RTAX_LABEL]) != NULL)
1149 					kr->r.rtlabel =
1150 					    rtlabel_name2id(label->sr_label);
1151 
1152 				if (kif_validate(kr->r.ifindex))
1153 					kr->r.flags &= ~F_DOWN;
1154 				else
1155 					kr->r.flags |= F_DOWN;
1156 
1157 				/* just readd, the RDE will care */
1158 				kr_redistribute(IMSG_NETWORK_ADD, &kr->r);
1159 			} else {
1160 				if ((kr = calloc(1,
1161 				    sizeof(struct kroute_node))) == NULL) {
1162 					log_warn("dispatch_rtmsg");
1163 					return (-1);
1164 				}
1165 
1166 				iface = if_find_index(rtm->rtm_index);
1167 				if (iface != NULL)
1168 					metric = iface->cost;
1169 				else
1170 					metric = DEFAULT_COST;
1171 
1172 				kr->r.prefix.s_addr = prefix.s_addr;
1173 				kr->r.netmask.s_addr = netmask.s_addr;
1174 				kr->r.nexthop.s_addr = nexthop.s_addr;
1175 				kr->r.metric = metric;
1176 				kr->r.flags = flags;
1177 				kr->r.ifindex = ifindex;
1178 
1179 				if ((label = (struct sockaddr_rtlabel *)
1180 				    rti_info[RTAX_LABEL]) != NULL)
1181 					kr->r.rtlabel =
1182 					    rtlabel_name2id(label->sr_label);
1183 
1184 				kroute_insert(kr);
1185 			}
1186 			break;
1187 		case RTM_DELETE:
1188 			if ((kr = kroute_find(prefix.s_addr, netmask.s_addr,
1189 			    prio)) == NULL)
1190 				continue;
1191 			if (!(kr->r.flags & F_KERNEL))
1192 				continue;
1193 			if (kroute_remove(kr) == -1)
1194 				return (-1);
1195 			break;
1196 		case RTM_IFINFO:
1197 			memcpy(&ifm, next, sizeof(ifm));
1198 			if_change(ifm.ifm_index, ifm.ifm_flags,
1199 			    &ifm.ifm_data);
1200 			break;
1201 		case RTM_IFANNOUNCE:
1202 			if_announce(next);
1203 			break;
1204 		default:
1205 			/* ignore for now */
1206 			break;
1207 		}
1208 	}
1209 	return (0);
1210 }
1211