xref: /openbsd-src/sys/net/rtsock.c (revision a2d98599e3c69439489c0bfafa67f0ebef23661b)
1 /*	$OpenBSD: rtsock.c,v 1.321 2021/09/07 16:07:46 mvs Exp $	*/
2 /*	$NetBSD: rtsock.c,v 1.18 1996/03/29 00:32:10 cgd Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1988, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)rtsock.c	8.6 (Berkeley) 2/11/95
62  */
63 
64 #include <sys/param.h>
65 #include <sys/systm.h>
66 #include <sys/proc.h>
67 #include <sys/sysctl.h>
68 #include <sys/mbuf.h>
69 #include <sys/socket.h>
70 #include <sys/socketvar.h>
71 #include <sys/domain.h>
72 #include <sys/pool.h>
73 #include <sys/protosw.h>
74 #include <sys/srp.h>
75 
76 #include <net/if.h>
77 #include <net/if_dl.h>
78 #include <net/if_var.h>
79 #include <net/route.h>
80 
81 #include <netinet/in.h>
82 
83 #ifdef MPLS
84 #include <netmpls/mpls.h>
85 #endif
86 #ifdef IPSEC
87 #include <netinet/ip_ipsp.h>
88 #include <net/if_enc.h>
89 #endif
90 #ifdef BFD
91 #include <net/bfd.h>
92 #endif
93 
94 #include <sys/stdarg.h>
95 #include <sys/kernel.h>
96 #include <sys/timeout.h>
97 
98 #define	ROUTESNDQ	8192
99 #define	ROUTERCVQ	8192
100 
101 const struct sockaddr route_src = { 2, PF_ROUTE, };
102 
103 struct walkarg {
104 	int	w_op, w_arg, w_given, w_needed, w_tmemsize;
105 	caddr_t	w_where, w_tmem;
106 };
107 
108 void	route_prinit(void);
109 void	rcb_ref(void *, void *);
110 void	rcb_unref(void *, void *);
111 int	route_output(struct mbuf *, struct socket *, struct sockaddr *,
112 	    struct mbuf *);
113 int	route_ctloutput(int, struct socket *, int, int, struct mbuf *);
114 int	route_usrreq(struct socket *, int, struct mbuf *, struct mbuf *,
115 	    struct mbuf *, struct proc *);
116 void	route_input(struct mbuf *m0, struct socket *, sa_family_t);
117 int	route_arp_conflict(struct rtentry *, struct rt_addrinfo *);
118 int	route_cleargateway(struct rtentry *, void *, unsigned int);
119 void	rtm_senddesync_timer(void *);
120 void	rtm_senddesync(struct socket *);
121 int	rtm_sendup(struct socket *, struct mbuf *);
122 
123 int	rtm_getifa(struct rt_addrinfo *, unsigned int);
124 int	rtm_output(struct rt_msghdr *, struct rtentry **, struct rt_addrinfo *,
125 	    uint8_t, unsigned int);
126 struct rt_msghdr *rtm_report(struct rtentry *, u_char, int, int);
127 struct mbuf	*rtm_msg1(int, struct rt_addrinfo *);
128 int		 rtm_msg2(int, int, struct rt_addrinfo *, caddr_t,
129 		     struct walkarg *);
130 int		 rtm_xaddrs(caddr_t, caddr_t, struct rt_addrinfo *);
131 int		 rtm_validate_proposal(struct rt_addrinfo *);
132 void		 rtm_setmetrics(u_long, const struct rt_metrics *,
133 		     struct rt_kmetrics *);
134 void		 rtm_getmetrics(const struct rt_kmetrics *,
135 		     struct rt_metrics *);
136 
137 int		 sysctl_iflist(int, struct walkarg *);
138 int		 sysctl_ifnames(struct walkarg *);
139 int		 sysctl_rtable_rtstat(void *, size_t *, void *);
140 
141 int		 rt_setsource(unsigned int, struct sockaddr *);
142 
143 /*
144  * Locks used to protect struct members
145  *       I       immutable after creation
146  *       s       solock
147  */
148 struct rtpcb {
149 	struct socket		*rop_socket;		/* [I] */
150 
151 	SRPL_ENTRY(rtpcb)	rop_list;
152 	struct refcnt		rop_refcnt;
153 	struct timeout		rop_timeout;
154 	unsigned int		rop_msgfilter;		/* [s] */
155 	unsigned int		rop_flagfilter;		/* [s] */
156 	unsigned int		rop_flags;		/* [s] */
157 	u_int			rop_rtableid;		/* [s] */
158 	unsigned short		rop_proto;		/* [I] */
159 	u_char			rop_priority;		/* [s] */
160 };
161 #define	sotortpcb(so)	((struct rtpcb *)(so)->so_pcb)
162 
163 struct rtptable {
164 	SRPL_HEAD(, rtpcb)	rtp_list;
165 	struct srpl_rc		rtp_rc;
166 	struct rwlock		rtp_lk;
167 	unsigned int		rtp_count;
168 };
169 
170 struct pool rtpcb_pool;
171 struct rtptable rtptable;
172 
173 /*
174  * These flags and timeout are used for indicating to userland (via a
175  * RTM_DESYNC msg) when the route socket has overflowed and messages
176  * have been lost.
177  */
178 #define ROUTECB_FLAG_DESYNC	0x1	/* Route socket out of memory */
179 #define ROUTECB_FLAG_FLUSH	0x2	/* Wait until socket is empty before
180 					   queueing more packets */
181 
182 #define ROUTE_DESYNC_RESEND_TIMEOUT	200	/* In ms */
183 
184 void
185 route_prinit(void)
186 {
187 	srpl_rc_init(&rtptable.rtp_rc, rcb_ref, rcb_unref, NULL);
188 	rw_init(&rtptable.rtp_lk, "rtsock");
189 	SRPL_INIT(&rtptable.rtp_list);
190 	pool_init(&rtpcb_pool, sizeof(struct rtpcb), 0,
191 	    IPL_SOFTNET, PR_WAITOK, "rtpcb", NULL);
192 }
193 
194 void
195 rcb_ref(void *null, void *v)
196 {
197 	struct rtpcb *rop = v;
198 
199 	refcnt_take(&rop->rop_refcnt);
200 }
201 
202 void
203 rcb_unref(void *null, void *v)
204 {
205 	struct rtpcb *rop = v;
206 
207 	refcnt_rele_wake(&rop->rop_refcnt);
208 }
209 
210 int
211 route_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
212     struct mbuf *control, struct proc *p)
213 {
214 	struct rtpcb	*rop;
215 	int		 error = 0;
216 
217 	if (req == PRU_CONTROL)
218 		return (EOPNOTSUPP);
219 
220 	soassertlocked(so);
221 
222 	if (control && control->m_len) {
223 		error = EOPNOTSUPP;
224 		goto release;
225 	}
226 
227 	rop = sotortpcb(so);
228 	if (rop == NULL) {
229 		error = EINVAL;
230 		goto release;
231 	}
232 
233 	switch (req) {
234 	/* no connect, bind, accept. Socket is connected from the start */
235 	case PRU_CONNECT:
236 	case PRU_BIND:
237 	case PRU_CONNECT2:
238 	case PRU_LISTEN:
239 	case PRU_ACCEPT:
240 		error = EOPNOTSUPP;
241 		break;
242 
243 	case PRU_DISCONNECT:
244 	case PRU_ABORT:
245 		soisdisconnected(so);
246 		break;
247 	case PRU_SHUTDOWN:
248 		socantsendmore(so);
249 		break;
250 	case PRU_SENSE:
251 		/* stat: don't bother with a blocksize. */
252 		break;
253 
254 	/* minimal support, just implement a fake peer address */
255 	case PRU_SOCKADDR:
256 		error = EINVAL;
257 		break;
258 	case PRU_PEERADDR:
259 		bcopy(&route_src, mtod(nam, caddr_t), route_src.sa_len);
260 		nam->m_len = route_src.sa_len;
261 		break;
262 
263 	case PRU_RCVD:
264 		/*
265 		 * If we are in a FLUSH state, check if the buffer is
266 		 * empty so that we can clear the flag.
267 		 */
268 		if (((rop->rop_flags & ROUTECB_FLAG_FLUSH) != 0) &&
269 		    ((sbspace(rop->rop_socket, &rop->rop_socket->so_rcv) ==
270 		    rop->rop_socket->so_rcv.sb_hiwat)))
271 			rop->rop_flags &= ~ROUTECB_FLAG_FLUSH;
272 		break;
273 
274 	case PRU_RCVOOB:
275 	case PRU_SENDOOB:
276 		error = EOPNOTSUPP;
277 		break;
278 	case PRU_SEND:
279 		if (nam) {
280 			error = EISCONN;
281 			break;
282 		}
283 		error = (*so->so_proto->pr_output)(m, so, NULL, NULL);
284 		m = NULL;
285 		break;
286 	default:
287 		panic("route_usrreq");
288 	}
289 
290  release:
291 	if (req != PRU_RCVD && req != PRU_RCVOOB && req != PRU_SENSE) {
292 		m_freem(control);
293 		m_freem(m);
294 	}
295 	return (error);
296 }
297 
298 int
299 route_attach(struct socket *so, int proto)
300 {
301 	struct rtpcb	*rop;
302 	int		 error;
303 
304 	error = soreserve(so, ROUTESNDQ, ROUTERCVQ);
305 	if (error)
306 		return (error);
307 	/*
308 	 * use the rawcb but allocate a rtpcb, this
309 	 * code does not care about the additional fields
310 	 * and works directly on the raw socket.
311 	 */
312 	rop = pool_get(&rtpcb_pool, PR_WAITOK|PR_ZERO);
313 	so->so_pcb = rop;
314 	/* Init the timeout structure */
315 	timeout_set_proc(&rop->rop_timeout, rtm_senddesync_timer, so);
316 	refcnt_init(&rop->rop_refcnt);
317 
318 	rop->rop_socket = so;
319 	rop->rop_proto = proto;
320 
321 	rop->rop_rtableid = curproc->p_p->ps_rtableid;
322 
323 	soisconnected(so);
324 	so->so_options |= SO_USELOOPBACK;
325 
326 	rw_enter(&rtptable.rtp_lk, RW_WRITE);
327 	SRPL_INSERT_HEAD_LOCKED(&rtptable.rtp_rc, &rtptable.rtp_list, rop,
328 	    rop_list);
329 	rtptable.rtp_count++;
330 	rw_exit(&rtptable.rtp_lk);
331 
332 	return (0);
333 }
334 
335 int
336 route_detach(struct socket *so)
337 {
338 	struct rtpcb	*rop;
339 
340 	soassertlocked(so);
341 
342 	rop = sotortpcb(so);
343 	if (rop == NULL)
344 		return (EINVAL);
345 
346 	rw_enter(&rtptable.rtp_lk, RW_WRITE);
347 
348 	rtptable.rtp_count--;
349 	SRPL_REMOVE_LOCKED(&rtptable.rtp_rc, &rtptable.rtp_list, rop, rtpcb,
350 	    rop_list);
351 	rw_exit(&rtptable.rtp_lk);
352 
353 	sounlock(so, SL_LOCKED);
354 
355 	/* wait for all references to drop */
356 	refcnt_finalize(&rop->rop_refcnt, "rtsockrefs");
357 	timeout_del_barrier(&rop->rop_timeout);
358 
359 	solock(so);
360 
361 	so->so_pcb = NULL;
362 	KASSERT((so->so_state & SS_NOFDREF) == 0);
363 	pool_put(&rtpcb_pool, rop);
364 
365 	return (0);
366 }
367 
368 int
369 route_ctloutput(int op, struct socket *so, int level, int optname,
370     struct mbuf *m)
371 {
372 	struct rtpcb *rop = sotortpcb(so);
373 	int error = 0;
374 	unsigned int tid, prio;
375 
376 	if (level != AF_ROUTE)
377 		return (EINVAL);
378 
379 	switch (op) {
380 	case PRCO_SETOPT:
381 		switch (optname) {
382 		case ROUTE_MSGFILTER:
383 			if (m == NULL || m->m_len != sizeof(unsigned int))
384 				error = EINVAL;
385 			else
386 				rop->rop_msgfilter = *mtod(m, unsigned int *);
387 			break;
388 		case ROUTE_TABLEFILTER:
389 			if (m == NULL || m->m_len != sizeof(unsigned int)) {
390 				error = EINVAL;
391 				break;
392 			}
393 			tid = *mtod(m, unsigned int *);
394 			if (tid != RTABLE_ANY && !rtable_exists(tid))
395 				error = ENOENT;
396 			else
397 				rop->rop_rtableid = tid;
398 			break;
399 		case ROUTE_PRIOFILTER:
400 			if (m == NULL || m->m_len != sizeof(unsigned int)) {
401 				error = EINVAL;
402 				break;
403 			}
404 			prio = *mtod(m, unsigned int *);
405 			if (prio > RTP_MAX)
406 				error = EINVAL;
407 			else
408 				rop->rop_priority = prio;
409 			break;
410 		case ROUTE_FLAGFILTER:
411 			if (m == NULL || m->m_len != sizeof(unsigned int))
412 				error = EINVAL;
413 			else
414 				rop->rop_flagfilter = *mtod(m, unsigned int *);
415 			break;
416 		default:
417 			error = ENOPROTOOPT;
418 			break;
419 		}
420 		break;
421 	case PRCO_GETOPT:
422 		switch (optname) {
423 		case ROUTE_MSGFILTER:
424 			m->m_len = sizeof(unsigned int);
425 			*mtod(m, unsigned int *) = rop->rop_msgfilter;
426 			break;
427 		case ROUTE_TABLEFILTER:
428 			m->m_len = sizeof(unsigned int);
429 			*mtod(m, unsigned int *) = rop->rop_rtableid;
430 			break;
431 		case ROUTE_PRIOFILTER:
432 			m->m_len = sizeof(unsigned int);
433 			*mtod(m, unsigned int *) = rop->rop_priority;
434 			break;
435 		case ROUTE_FLAGFILTER:
436 			m->m_len = sizeof(unsigned int);
437 			*mtod(m, unsigned int *) = rop->rop_flagfilter;
438 			break;
439 		default:
440 			error = ENOPROTOOPT;
441 			break;
442 		}
443 	}
444 	return (error);
445 }
446 
447 void
448 rtm_senddesync_timer(void *xso)
449 {
450 	struct socket	*so = xso;
451 	int		 s;
452 
453 	s = solock(so);
454 	rtm_senddesync(so);
455 	sounlock(so, s);
456 }
457 
458 void
459 rtm_senddesync(struct socket *so)
460 {
461 	struct rtpcb	*rop = sotortpcb(so);
462 	struct mbuf	*desync_mbuf;
463 
464 	soassertlocked(so);
465 
466 	/*
467 	 * Dying socket is disconnected by upper layer and there is
468 	 * no reason to send packet. Also we shouldn't reschedule
469 	 * timeout(9), otherwise timeout_del_barrier(9) can't help us.
470 	 */
471 	if ((so->so_state & SS_ISCONNECTED) == 0 ||
472 	    (so->so_state & SS_CANTRCVMORE))
473 		return;
474 
475 	/* If we are in a DESYNC state, try to send a RTM_DESYNC packet */
476 	if ((rop->rop_flags & ROUTECB_FLAG_DESYNC) == 0)
477 		return;
478 
479 	/*
480 	 * If we fail to alloc memory or if sbappendaddr()
481 	 * fails, re-add timeout and try again.
482 	 */
483 	desync_mbuf = rtm_msg1(RTM_DESYNC, NULL);
484 	if (desync_mbuf != NULL) {
485 		if (sbappendaddr(so, &so->so_rcv, &route_src,
486 		    desync_mbuf, NULL) != 0) {
487 			rop->rop_flags &= ~ROUTECB_FLAG_DESYNC;
488 			sorwakeup(rop->rop_socket);
489 			return;
490 		}
491 		m_freem(desync_mbuf);
492 	}
493 	/* Re-add timeout to try sending msg again */
494 	timeout_add_msec(&rop->rop_timeout, ROUTE_DESYNC_RESEND_TIMEOUT);
495 }
496 
497 void
498 route_input(struct mbuf *m0, struct socket *so0, sa_family_t sa_family)
499 {
500 	struct socket *so;
501 	struct rtpcb *rop;
502 	struct rt_msghdr *rtm;
503 	struct mbuf *m = m0;
504 	struct srp_ref sr;
505 	int s;
506 
507 	/* ensure that we can access the rtm_type via mtod() */
508 	if (m->m_len < offsetof(struct rt_msghdr, rtm_type) + 1) {
509 		m_freem(m);
510 		return;
511 	}
512 
513 	SRPL_FOREACH(rop, &sr, &rtptable.rtp_list, rop_list) {
514 		/*
515 		 * If route socket is bound to an address family only send
516 		 * messages that match the address family. Address family
517 		 * agnostic messages are always sent.
518 		 */
519 		if (sa_family != AF_UNSPEC && rop->rop_proto != AF_UNSPEC &&
520 		    rop->rop_proto != sa_family)
521 			continue;
522 
523 
524 		so = rop->rop_socket;
525 		s = solock(so);
526 
527 		/*
528 		 * Check to see if we don't want our own messages and
529 		 * if we can receive anything.
530 		 */
531 		if ((so0 == so && !(so0->so_options & SO_USELOOPBACK)) ||
532 		    !(so->so_state & SS_ISCONNECTED) ||
533 		    (so->so_state & SS_CANTRCVMORE))
534 			goto next;
535 
536 		/* filter messages that the process does not want */
537 		rtm = mtod(m, struct rt_msghdr *);
538 		/* but RTM_DESYNC can't be filtered */
539 		if (rtm->rtm_type != RTM_DESYNC) {
540 			if (rop->rop_msgfilter != 0 &&
541 			    !(rop->rop_msgfilter & (1 << rtm->rtm_type)))
542 				goto next;
543 			if (ISSET(rop->rop_flagfilter, rtm->rtm_flags))
544 				goto next;
545 		}
546 		switch (rtm->rtm_type) {
547 		case RTM_IFANNOUNCE:
548 		case RTM_DESYNC:
549 			/* no tableid */
550 			break;
551 		case RTM_RESOLVE:
552 		case RTM_NEWADDR:
553 		case RTM_DELADDR:
554 		case RTM_IFINFO:
555 		case RTM_80211INFO:
556 		case RTM_BFD:
557 			/* check against rdomain id */
558 			if (rop->rop_rtableid != RTABLE_ANY &&
559 			    rtable_l2(rop->rop_rtableid) != rtm->rtm_tableid)
560 				goto next;
561 			break;
562 		default:
563 			if (rop->rop_priority != 0 &&
564 			    rop->rop_priority < rtm->rtm_priority)
565 				goto next;
566 			/* check against rtable id */
567 			if (rop->rop_rtableid != RTABLE_ANY &&
568 			    rop->rop_rtableid != rtm->rtm_tableid)
569 				goto next;
570 			break;
571 		}
572 
573 		/*
574 		 * Check to see if the flush flag is set. If so, don't queue
575 		 * any more messages until the flag is cleared.
576 		 */
577 		if ((rop->rop_flags & ROUTECB_FLAG_FLUSH) != 0)
578 			goto next;
579 
580 		rtm_sendup(so, m);
581 next:
582 		sounlock(so, s);
583 	}
584 	SRPL_LEAVE(&sr);
585 
586 	m_freem(m);
587 }
588 
589 int
590 rtm_sendup(struct socket *so, struct mbuf *m0)
591 {
592 	struct rtpcb *rop = sotortpcb(so);
593 	struct mbuf *m;
594 
595 	soassertlocked(so);
596 
597 	m = m_copym(m0, 0, M_COPYALL, M_NOWAIT);
598 	if (m == NULL)
599 		return (ENOMEM);
600 
601 	if (sbspace(so, &so->so_rcv) < (2 * MSIZE) ||
602 	    sbappendaddr(so, &so->so_rcv, &route_src, m, NULL) == 0) {
603 		/* Flag socket as desync'ed and flush required */
604 		rop->rop_flags |= ROUTECB_FLAG_DESYNC | ROUTECB_FLAG_FLUSH;
605 		rtm_senddesync(so);
606 		m_freem(m);
607 		return (ENOBUFS);
608 	}
609 
610 	sorwakeup(so);
611 	return (0);
612 }
613 
614 struct rt_msghdr *
615 rtm_report(struct rtentry *rt, u_char type, int seq, int tableid)
616 {
617 	struct rt_msghdr	*rtm;
618 	struct rt_addrinfo	 info;
619 	struct sockaddr_rtlabel	 sa_rl;
620 	struct sockaddr_in6	 sa_mask;
621 #ifdef BFD
622 	struct sockaddr_bfd	 sa_bfd;
623 #endif
624 	struct ifnet		*ifp = NULL;
625 	int			 len;
626 
627 	bzero(&info, sizeof(info));
628 	info.rti_info[RTAX_DST] = rt_key(rt);
629 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
630 	info.rti_info[RTAX_NETMASK] = rt_plen2mask(rt, &sa_mask);
631 	info.rti_info[RTAX_LABEL] = rtlabel_id2sa(rt->rt_labelid, &sa_rl);
632 #ifdef BFD
633 	if (rt->rt_flags & RTF_BFD)
634 		info.rti_info[RTAX_BFD] = bfd2sa(rt, &sa_bfd);
635 #endif
636 #ifdef MPLS
637 	if (rt->rt_flags & RTF_MPLS) {
638 		struct sockaddr_mpls	 sa_mpls;
639 
640 		bzero(&sa_mpls, sizeof(sa_mpls));
641 		sa_mpls.smpls_family = AF_MPLS;
642 		sa_mpls.smpls_len = sizeof(sa_mpls);
643 		sa_mpls.smpls_label = ((struct rt_mpls *)
644 		    rt->rt_llinfo)->mpls_label;
645 		info.rti_info[RTAX_SRC] = (struct sockaddr *)&sa_mpls;
646 		info.rti_mpls = ((struct rt_mpls *)
647 		    rt->rt_llinfo)->mpls_operation;
648 	}
649 #endif
650 	ifp = if_get(rt->rt_ifidx);
651 	if (ifp != NULL) {
652 		info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
653 		info.rti_info[RTAX_IFA] =
654 		    rtable_getsource(tableid, info.rti_info[RTAX_DST]->sa_family);
655 		if (info.rti_info[RTAX_IFA] == NULL)
656 			info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
657 		if (ifp->if_flags & IFF_POINTOPOINT)
658 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
659 	}
660 	if_put(ifp);
661 	/* RTAX_GENMASK, RTAX_AUTHOR, RTAX_SRCMASK ignored */
662 
663 	/* build new route message */
664 	len = rtm_msg2(type, RTM_VERSION, &info, NULL, NULL);
665 	rtm = malloc(len, M_RTABLE, M_WAITOK | M_ZERO);
666 
667 	rtm_msg2(type, RTM_VERSION, &info, (caddr_t)rtm, NULL);
668 	rtm->rtm_type = type;
669 	rtm->rtm_index = rt->rt_ifidx;
670 	rtm->rtm_tableid = tableid;
671 	rtm->rtm_priority = rt->rt_priority & RTP_MASK;
672 	rtm->rtm_flags = rt->rt_flags;
673 	rtm->rtm_pid = curproc->p_p->ps_pid;
674 	rtm->rtm_seq = seq;
675 	rtm_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
676 	rtm->rtm_addrs = info.rti_addrs;
677 #ifdef MPLS
678 	rtm->rtm_mpls = info.rti_mpls;
679 #endif
680 	return rtm;
681 }
682 
683 int
684 route_output(struct mbuf *m, struct socket *so, struct sockaddr *dstaddr,
685     struct mbuf *control)
686 {
687 	struct rt_msghdr	*rtm = NULL;
688 	struct rtentry		*rt = NULL;
689 	struct rt_addrinfo	 info;
690 	struct ifnet		*ifp;
691 	int			 len, seq, useloopback, error = 0;
692 	u_int			 tableid;
693 	u_int8_t		 prio;
694 	u_char			 vers, type;
695 
696 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
697 	    (m = m_pullup(m, sizeof(int32_t))) == 0))
698 		return (ENOBUFS);
699 	if ((m->m_flags & M_PKTHDR) == 0)
700 		panic("route_output");
701 
702 	useloopback = so->so_options & SO_USELOOPBACK;
703 
704 	/*
705 	 * The socket can't be closed concurrently because the file
706 	 * descriptor reference is still held.
707 	 */
708 
709 	sounlock(so, SL_LOCKED);
710 
711 	len = m->m_pkthdr.len;
712 	if (len < offsetof(struct rt_msghdr, rtm_hdrlen) + 1 ||
713 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
714 		error = EINVAL;
715 		goto fail;
716 	}
717 	vers = mtod(m, struct rt_msghdr *)->rtm_version;
718 	switch (vers) {
719 	case RTM_VERSION:
720 		if (len < sizeof(struct rt_msghdr)) {
721 			error = EINVAL;
722 			goto fail;
723 		}
724 		if (len > RTM_MAXSIZE) {
725 			error = EMSGSIZE;
726 			goto fail;
727 		}
728 		rtm = malloc(len, M_RTABLE, M_WAITOK);
729 		m_copydata(m, 0, len, rtm);
730 		break;
731 	default:
732 		error = EPROTONOSUPPORT;
733 		goto fail;
734 	}
735 
736 	/* Verify that the caller is sending an appropriate message early */
737 	switch (rtm->rtm_type) {
738 	case RTM_ADD:
739 	case RTM_DELETE:
740 	case RTM_GET:
741 	case RTM_CHANGE:
742 	case RTM_PROPOSAL:
743 	case RTM_SOURCE:
744 		break;
745 	default:
746 		error = EOPNOTSUPP;
747 		goto fail;
748 	}
749 	/*
750 	 * Verify that the header length is valid.
751 	 * All messages from userland start with a struct rt_msghdr.
752 	 */
753 	if (rtm->rtm_hdrlen == 0)	/* old client */
754 		rtm->rtm_hdrlen = sizeof(struct rt_msghdr);
755 	if (rtm->rtm_hdrlen < sizeof(struct rt_msghdr) ||
756 	    len < rtm->rtm_hdrlen) {
757 		error = EINVAL;
758 		goto fail;
759 	}
760 
761 	rtm->rtm_pid = curproc->p_p->ps_pid;
762 
763 	/*
764 	 * Verify that the caller has the appropriate privilege; RTM_GET
765 	 * is the only operation the non-superuser is allowed.
766 	 */
767 	if (rtm->rtm_type != RTM_GET && suser(curproc) != 0) {
768 		error = EACCES;
769 		goto fail;
770 	}
771 	tableid = rtm->rtm_tableid;
772 	if (!rtable_exists(tableid)) {
773 		if (rtm->rtm_type == RTM_ADD) {
774 			if ((error = rtable_add(tableid)) != 0)
775 				goto fail;
776 		} else {
777 			error = EINVAL;
778 			goto fail;
779 		}
780 	}
781 
782 	/* Do not let userland play with kernel-only flags. */
783 	if ((rtm->rtm_flags & (RTF_LOCAL|RTF_BROADCAST)) != 0) {
784 		error = EINVAL;
785 		goto fail;
786 	}
787 
788 	/* make sure that kernel-only bits are not set */
789 	rtm->rtm_priority &= RTP_MASK;
790 	rtm->rtm_flags &= ~(RTF_DONE|RTF_CLONED|RTF_CACHED);
791 	rtm->rtm_fmask &= RTF_FMASK;
792 
793 	if (rtm->rtm_priority != 0) {
794 		if (rtm->rtm_priority > RTP_MAX ||
795 		    rtm->rtm_priority == RTP_LOCAL) {
796 			error = EINVAL;
797 			goto fail;
798 		}
799 		prio = rtm->rtm_priority;
800 	} else if (rtm->rtm_type != RTM_ADD)
801 		prio = RTP_ANY;
802 	else if (rtm->rtm_flags & RTF_STATIC)
803 		prio = 0;
804 	else
805 		prio = RTP_DEFAULT;
806 
807 	bzero(&info, sizeof(info));
808 	info.rti_addrs = rtm->rtm_addrs;
809 	if ((error = rtm_xaddrs(rtm->rtm_hdrlen + (caddr_t)rtm,
810 	    len + (caddr_t)rtm, &info)) != 0)
811 		goto fail;
812 
813 	info.rti_flags = rtm->rtm_flags;
814 
815 	if (rtm->rtm_type != RTM_SOURCE &&
816 	    rtm->rtm_type != RTM_PROPOSAL &&
817 	    (info.rti_info[RTAX_DST] == NULL ||
818 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
819 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
820 	    info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX) ||
821 	    info.rti_info[RTAX_GENMASK] != NULL)) {
822 		error = EINVAL;
823 		goto fail;
824 	}
825 #ifdef MPLS
826 	info.rti_mpls = rtm->rtm_mpls;
827 #endif
828 
829 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
830 	    info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
831 	    (info.rti_flags & RTF_CLONING) == 0) {
832 		info.rti_flags |= RTF_LLINFO;
833 	}
834 
835 	/*
836 	 * Validate RTM_PROPOSAL and pass it along or error out.
837 	 */
838 	if (rtm->rtm_type == RTM_PROPOSAL) {
839 		if (rtm_validate_proposal(&info) == -1) {
840 			error = EINVAL;
841 			goto fail;
842 		}
843 		/*
844 		 * If this is a solicitation proposal forward request to
845 		 * all interfaces. Most handlers will ignore it but at least
846 		 * umb(4) will send a response to this event.
847 		 */
848 		if (rtm->rtm_priority == RTP_PROPOSAL_SOLICIT) {
849 			NET_LOCK();
850 			TAILQ_FOREACH(ifp, &ifnet, if_list) {
851 				ifp->if_rtrequest(ifp, RTM_PROPOSAL, NULL);
852 			}
853 			NET_UNLOCK();
854 		}
855 	} else if (rtm->rtm_type == RTM_SOURCE) {
856 		if (info.rti_info[RTAX_IFA] == NULL) {
857 			error = EINVAL;
858 			goto fail;
859 		}
860 		if ((error =
861 		    rt_setsource(tableid, info.rti_info[RTAX_IFA])) != 0)
862 			goto fail;
863 	} else {
864 		error = rtm_output(rtm, &rt, &info, prio, tableid);
865 		if (!error) {
866 			type = rtm->rtm_type;
867 			seq = rtm->rtm_seq;
868 			free(rtm, M_RTABLE, len);
869 			rtm = rtm_report(rt, type, seq, tableid);
870 			len = rtm->rtm_msglen;
871 		}
872 	}
873 
874 	rtfree(rt);
875 	if (error) {
876 		rtm->rtm_errno = error;
877 	} else {
878 		rtm->rtm_flags |= RTF_DONE;
879 	}
880 
881 	/*
882 	 * Check to see if we don't want our own messages.
883 	 */
884 	if (!useloopback) {
885 		if (rtptable.rtp_count == 0) {
886 			/* no other listener and no loopback of messages */
887 			goto fail;
888 		}
889 	}
890 	if (m_copyback(m, 0, len, rtm, M_NOWAIT)) {
891 		m_freem(m);
892 		m = NULL;
893 	} else if (m->m_pkthdr.len > len)
894 		m_adj(m, len - m->m_pkthdr.len);
895 	free(rtm, M_RTABLE, len);
896 	if (m)
897 		route_input(m, so, info.rti_info[RTAX_DST] ?
898 		    info.rti_info[RTAX_DST]->sa_family : AF_UNSPEC);
899 	solock(so);
900 
901 	return (error);
902 fail:
903 	free(rtm, M_RTABLE, len);
904 	m_freem(m);
905 	solock(so);
906 
907 	return (error);
908 }
909 
910 int
911 rtm_output(struct rt_msghdr *rtm, struct rtentry **prt,
912     struct rt_addrinfo *info, uint8_t prio, unsigned int tableid)
913 {
914 	struct rtentry		*rt = *prt;
915 	struct ifnet		*ifp = NULL;
916 	int			 plen, newgate = 0, error = 0;
917 
918 	switch (rtm->rtm_type) {
919 	case RTM_ADD:
920 		if (info->rti_info[RTAX_GATEWAY] == NULL) {
921 			error = EINVAL;
922 			break;
923 		}
924 
925 		rt = rtable_match(tableid, info->rti_info[RTAX_DST], NULL);
926 		if ((error = route_arp_conflict(rt, info))) {
927 			rtfree(rt);
928 			rt = NULL;
929 			break;
930 		}
931 
932 		/*
933 		 * We cannot go through a delete/create/insert cycle for
934 		 * cached route because this can lead to races in the
935 		 * receive path.  Instead we update the L2 cache.
936 		 */
937 		if ((rt != NULL) && ISSET(rt->rt_flags, RTF_CACHED)) {
938 			ifp = if_get(rt->rt_ifidx);
939 			if (ifp == NULL) {
940 				rtfree(rt);
941 				rt = NULL;
942 				error = ESRCH;
943 				break;
944 			}
945 
946 			goto change;
947 		}
948 
949 		rtfree(rt);
950 		rt = NULL;
951 
952 		NET_LOCK();
953 		if ((error = rtm_getifa(info, tableid)) != 0) {
954 			NET_UNLOCK();
955 			break;
956 		}
957 		error = rtrequest(RTM_ADD, info, prio, &rt, tableid);
958 		NET_UNLOCK();
959 		if (error == 0)
960 			rtm_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
961 			    &rt->rt_rmx);
962 		break;
963 	case RTM_DELETE:
964 		rt = rtable_lookup(tableid, info->rti_info[RTAX_DST],
965 		    info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY],
966 		    prio);
967 		if (rt == NULL) {
968 			error = ESRCH;
969 			break;
970 		}
971 
972 		/*
973 		 * If we got multipath routes, we require users to specify
974 		 * a matching gateway.
975 		 */
976 		if (ISSET(rt->rt_flags, RTF_MPATH) &&
977 		    info->rti_info[RTAX_GATEWAY] == NULL) {
978 			error = ESRCH;
979 			break;
980 		}
981 
982 		ifp = if_get(rt->rt_ifidx);
983 		if (ifp == NULL) {
984 			rtfree(rt);
985 			rt = NULL;
986 			error = ESRCH;
987 			break;
988 		}
989 
990 		/*
991 		 * Invalidate the cache of automagically created and
992 		 * referenced L2 entries to make sure that ``rt_gwroute''
993 		 * pointer stays valid for other CPUs.
994 		 */
995 		if ((ISSET(rt->rt_flags, RTF_CACHED))) {
996 			NET_LOCK();
997 			ifp->if_rtrequest(ifp, RTM_INVALIDATE, rt);
998 			/* Reset the MTU of the gateway route. */
999 			rtable_walk(tableid, rt_key(rt)->sa_family, NULL,
1000 			    route_cleargateway, rt);
1001 			NET_UNLOCK();
1002 			break;
1003 		}
1004 
1005 		/*
1006 		 * Make sure that local routes are only modified by the
1007 		 * kernel.
1008 		 */
1009 		if (ISSET(rt->rt_flags, RTF_LOCAL|RTF_BROADCAST)) {
1010 			error = EINVAL;
1011 			break;
1012 		}
1013 
1014 		rtfree(rt);
1015 		rt = NULL;
1016 
1017 		NET_LOCK();
1018 		error = rtrequest_delete(info, prio, ifp, &rt, tableid);
1019 		NET_UNLOCK();
1020 		break;
1021 	case RTM_CHANGE:
1022 		rt = rtable_lookup(tableid, info->rti_info[RTAX_DST],
1023 		    info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY],
1024 		    prio);
1025 		/*
1026 		 * If we got multipath routes, we require users to specify
1027 		 * a matching gateway.
1028 		 */
1029 		if ((rt != NULL) && ISSET(rt->rt_flags, RTF_MPATH) &&
1030 		    (info->rti_info[RTAX_GATEWAY] == NULL)) {
1031 			rtfree(rt);
1032 			rt = NULL;
1033 		}
1034 
1035 		/*
1036 		 * If RTAX_GATEWAY is the argument we're trying to
1037 		 * change, try to find a compatible route.
1038 		 */
1039 		if ((rt == NULL) && (info->rti_info[RTAX_GATEWAY] != NULL)) {
1040 			rt = rtable_lookup(tableid, info->rti_info[RTAX_DST],
1041 			    info->rti_info[RTAX_NETMASK], NULL, prio);
1042 			/* Ensure we don't pick a multipath one. */
1043 			if ((rt != NULL) && ISSET(rt->rt_flags, RTF_MPATH)) {
1044 				rtfree(rt);
1045 				rt = NULL;
1046 			}
1047 		}
1048 
1049 		if (rt == NULL) {
1050 			error = ESRCH;
1051 			break;
1052 		}
1053 
1054 		/*
1055 		 * Make sure that local routes are only modified by the
1056 		 * kernel.
1057 		 */
1058 		if (ISSET(rt->rt_flags, RTF_LOCAL|RTF_BROADCAST)) {
1059 			error = EINVAL;
1060 			break;
1061 		}
1062 
1063 		ifp = if_get(rt->rt_ifidx);
1064 		if (ifp == NULL) {
1065 			rtfree(rt);
1066 			rt = NULL;
1067 			error = ESRCH;
1068 			break;
1069 		}
1070 
1071 		/*
1072 		 * RTM_CHANGE needs a perfect match.
1073 		 */
1074 		plen = rtable_satoplen(info->rti_info[RTAX_DST]->sa_family,
1075 		    info->rti_info[RTAX_NETMASK]);
1076 		if (rt_plen(rt) != plen) {
1077 			error = ESRCH;
1078 			break;
1079 		}
1080 
1081 		if (info->rti_info[RTAX_GATEWAY] != NULL)
1082 			if (rt->rt_gateway == NULL ||
1083 			    bcmp(rt->rt_gateway,
1084 			    info->rti_info[RTAX_GATEWAY],
1085 			    info->rti_info[RTAX_GATEWAY]->sa_len)) {
1086 				newgate = 1;
1087 			}
1088 		/*
1089 		 * Check reachable gateway before changing the route.
1090 		 * New gateway could require new ifaddr, ifp;
1091 		 * flags may also be different; ifp may be specified
1092 		 * by ll sockaddr when protocol address is ambiguous.
1093 		 */
1094 		if (newgate || info->rti_info[RTAX_IFP] != NULL ||
1095 		    info->rti_info[RTAX_IFA] != NULL) {
1096 			struct ifaddr	*ifa = NULL;
1097 
1098 			NET_LOCK();
1099 			if ((error = rtm_getifa(info, tableid)) != 0) {
1100 				NET_UNLOCK();
1101 				break;
1102 			}
1103 			ifa = info->rti_ifa;
1104 			if (rt->rt_ifa != ifa) {
1105 				ifp->if_rtrequest(ifp, RTM_DELETE, rt);
1106 				ifafree(rt->rt_ifa);
1107 
1108 				ifa->ifa_refcnt++;
1109 				rt->rt_ifa = ifa;
1110 				rt->rt_ifidx = ifa->ifa_ifp->if_index;
1111 				/* recheck link state after ifp change */
1112 				rt_if_linkstate_change(rt, ifa->ifa_ifp,
1113 				    tableid);
1114 			}
1115 			NET_UNLOCK();
1116 		}
1117 change:
1118 		if (info->rti_info[RTAX_GATEWAY] != NULL) {
1119 			/* When updating the gateway, make sure it is valid. */
1120 			if (!newgate && rt->rt_gateway->sa_family !=
1121 			    info->rti_info[RTAX_GATEWAY]->sa_family) {
1122 				error = EINVAL;
1123 				break;
1124 			}
1125 
1126 			NET_LOCK();
1127 			error = rt_setgate(rt,
1128 			    info->rti_info[RTAX_GATEWAY], tableid);
1129 			NET_UNLOCK();
1130 			if (error)
1131 				break;
1132 		}
1133 #ifdef MPLS
1134 		if (rtm->rtm_flags & RTF_MPLS) {
1135 			NET_LOCK();
1136 			error = rt_mpls_set(rt,
1137 			    info->rti_info[RTAX_SRC], info->rti_mpls);
1138 			NET_UNLOCK();
1139 			if (error)
1140 				break;
1141 		} else if (newgate || (rtm->rtm_fmask & RTF_MPLS)) {
1142 			NET_LOCK();
1143 			/* if gateway changed remove MPLS information */
1144 			rt_mpls_clear(rt);
1145 			NET_UNLOCK();
1146 		}
1147 #endif
1148 
1149 #ifdef BFD
1150 		if (ISSET(rtm->rtm_flags, RTF_BFD)) {
1151 			if ((error = bfdset(rt)))
1152 				break;
1153 		} else if (!ISSET(rtm->rtm_flags, RTF_BFD) &&
1154 		    ISSET(rtm->rtm_fmask, RTF_BFD)) {
1155 			bfdclear(rt);
1156 		}
1157 #endif
1158 
1159 		NET_LOCK();
1160 		/* Hack to allow some flags to be toggled */
1161 		if (rtm->rtm_fmask) {
1162 			/* MPLS flag it is set by rt_mpls_set() */
1163 			rtm->rtm_fmask &= ~RTF_MPLS;
1164 			rtm->rtm_flags &= ~RTF_MPLS;
1165 			rt->rt_flags =
1166 			    (rt->rt_flags & ~rtm->rtm_fmask) |
1167 			    (rtm->rtm_flags & rtm->rtm_fmask);
1168 		}
1169 		rtm_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, &rt->rt_rmx);
1170 
1171 		ifp->if_rtrequest(ifp, RTM_ADD, rt);
1172 
1173 		if (info->rti_info[RTAX_LABEL] != NULL) {
1174 			char *rtlabel = ((struct sockaddr_rtlabel *)
1175 			    info->rti_info[RTAX_LABEL])->sr_label;
1176 			rtlabel_unref(rt->rt_labelid);
1177 			rt->rt_labelid = rtlabel_name2id(rtlabel);
1178 		}
1179 		if_group_routechange(info->rti_info[RTAX_DST],
1180 		    info->rti_info[RTAX_NETMASK]);
1181 		rt->rt_locks &= ~(rtm->rtm_inits);
1182 		rt->rt_locks |= (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
1183 		NET_UNLOCK();
1184 		break;
1185 	case RTM_GET:
1186 		rt = rtable_lookup(tableid, info->rti_info[RTAX_DST],
1187 		    info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY],
1188 		    prio);
1189 		if (rt == NULL)
1190 			error = ESRCH;
1191 		break;
1192 	}
1193 
1194 	if_put(ifp);
1195 	*prt = rt;
1196 	return (error);
1197 }
1198 
1199 struct ifaddr *
1200 ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway,
1201     unsigned int rtableid)
1202 {
1203 	struct ifaddr	*ifa;
1204 
1205 	if ((flags & RTF_GATEWAY) == 0) {
1206 		/*
1207 		 * If we are adding a route to an interface,
1208 		 * and the interface is a pt to pt link
1209 		 * we should search for the destination
1210 		 * as our clue to the interface.  Otherwise
1211 		 * we can use the local address.
1212 		 */
1213 		ifa = NULL;
1214 		if (flags & RTF_HOST)
1215 			ifa = ifa_ifwithdstaddr(dst, rtableid);
1216 		if (ifa == NULL)
1217 			ifa = ifa_ifwithaddr(gateway, rtableid);
1218 	} else {
1219 		/*
1220 		 * If we are adding a route to a remote net
1221 		 * or host, the gateway may still be on the
1222 		 * other end of a pt to pt link.
1223 		 */
1224 		ifa = ifa_ifwithdstaddr(gateway, rtableid);
1225 	}
1226 	if (ifa == NULL) {
1227 		if (gateway->sa_family == AF_LINK) {
1228 			struct sockaddr_dl *sdl = satosdl(gateway);
1229 			struct ifnet *ifp = if_get(sdl->sdl_index);
1230 
1231 			if (ifp != NULL)
1232 				ifa = ifaof_ifpforaddr(dst, ifp);
1233 			if_put(ifp);
1234 		} else {
1235 			struct rtentry *rt;
1236 
1237 			rt = rtalloc(gateway, RT_RESOLVE, rtable_l2(rtableid));
1238 			if (rt != NULL)
1239 				ifa = rt->rt_ifa;
1240 			rtfree(rt);
1241 		}
1242 	}
1243 	if (ifa == NULL)
1244 		return (NULL);
1245 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
1246 		struct ifaddr	*oifa = ifa;
1247 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
1248 		if (ifa == NULL)
1249 			ifa = oifa;
1250 	}
1251 	return (ifa);
1252 }
1253 
1254 int
1255 rtm_getifa(struct rt_addrinfo *info, unsigned int rtid)
1256 {
1257 	struct ifnet	*ifp = NULL;
1258 
1259 	/*
1260 	 * The "returned" `ifa' is guaranteed to be alive only if
1261 	 * the NET_LOCK() is held.
1262 	 */
1263 	NET_ASSERT_LOCKED();
1264 
1265 	/*
1266 	 * ifp may be specified by sockaddr_dl when protocol address
1267 	 * is ambiguous
1268 	 */
1269 	if (info->rti_info[RTAX_IFP] != NULL) {
1270 		struct sockaddr_dl *sdl;
1271 
1272 		sdl = satosdl(info->rti_info[RTAX_IFP]);
1273 		ifp = if_get(sdl->sdl_index);
1274 	}
1275 
1276 #ifdef IPSEC
1277 	/*
1278 	 * If the destination is a PF_KEY address, we'll look
1279 	 * for the existence of a encap interface number or address
1280 	 * in the options list of the gateway. By default, we'll return
1281 	 * enc0.
1282 	 */
1283 	if (info->rti_info[RTAX_DST] &&
1284 	    info->rti_info[RTAX_DST]->sa_family == PF_KEY)
1285 		info->rti_ifa = enc_getifa(rtid, 0);
1286 #endif
1287 
1288 	if (info->rti_ifa == NULL && info->rti_info[RTAX_IFA] != NULL)
1289 		info->rti_ifa = ifa_ifwithaddr(info->rti_info[RTAX_IFA], rtid);
1290 
1291 	if (info->rti_ifa == NULL) {
1292 		struct sockaddr	*sa;
1293 
1294 		if ((sa = info->rti_info[RTAX_IFA]) == NULL)
1295 			if ((sa = info->rti_info[RTAX_GATEWAY]) == NULL)
1296 				sa = info->rti_info[RTAX_DST];
1297 
1298 		if (sa != NULL && ifp != NULL)
1299 			info->rti_ifa = ifaof_ifpforaddr(sa, ifp);
1300 		else if (info->rti_info[RTAX_DST] != NULL &&
1301 		    info->rti_info[RTAX_GATEWAY] != NULL)
1302 			info->rti_ifa = ifa_ifwithroute(info->rti_flags,
1303 			    info->rti_info[RTAX_DST],
1304 			    info->rti_info[RTAX_GATEWAY],
1305 			    rtid);
1306 		else if (sa != NULL)
1307 			info->rti_ifa = ifa_ifwithroute(info->rti_flags,
1308 			    sa, sa, rtid);
1309 	}
1310 
1311 	if_put(ifp);
1312 
1313 	if (info->rti_ifa == NULL)
1314 		return (ENETUNREACH);
1315 
1316 	return (0);
1317 }
1318 
1319 int
1320 route_cleargateway(struct rtentry *rt, void *arg, unsigned int rtableid)
1321 {
1322 	struct rtentry *nhrt = arg;
1323 
1324 	if (ISSET(rt->rt_flags, RTF_GATEWAY) && rt->rt_gwroute == nhrt &&
1325 	    !ISSET(rt->rt_locks, RTV_MTU))
1326 		rt->rt_mtu = 0;
1327 
1328 	return (0);
1329 }
1330 
1331 /*
1332  * Check if the user request to insert an ARP entry does not conflict
1333  * with existing ones.
1334  *
1335  * Only two entries are allowed for a given IP address: a private one
1336  * (priv) and a public one (pub).
1337  */
1338 int
1339 route_arp_conflict(struct rtentry *rt, struct rt_addrinfo *info)
1340 {
1341 	int		 proxy = (info->rti_flags & RTF_ANNOUNCE);
1342 
1343 	if ((info->rti_flags & RTF_LLINFO) == 0 ||
1344 	    (info->rti_info[RTAX_DST]->sa_family != AF_INET))
1345 		return (0);
1346 
1347 	if (rt == NULL || !ISSET(rt->rt_flags, RTF_LLINFO))
1348 		return (0);
1349 
1350 	/* If the entry is cached, it can be updated. */
1351 	if (ISSET(rt->rt_flags, RTF_CACHED))
1352 		return (0);
1353 
1354 	/*
1355 	 * Same destination, not cached and both "priv" or "pub" conflict.
1356 	 * If a second entry exists, it always conflict.
1357 	 */
1358 	if ((ISSET(rt->rt_flags, RTF_ANNOUNCE) == proxy) ||
1359 	    ISSET(rt->rt_flags, RTF_MPATH))
1360 		return (EEXIST);
1361 
1362 	/* No conflict but an entry exist so we need to force mpath. */
1363 	info->rti_flags |= RTF_MPATH;
1364 	return (0);
1365 }
1366 
1367 void
1368 rtm_setmetrics(u_long which, const struct rt_metrics *in,
1369     struct rt_kmetrics *out)
1370 {
1371 	int64_t expire;
1372 
1373 	if (which & RTV_MTU)
1374 		out->rmx_mtu = in->rmx_mtu;
1375 	if (which & RTV_EXPIRE) {
1376 		expire = in->rmx_expire;
1377 		if (expire != 0) {
1378 			expire -= gettime();
1379 			expire += getuptime();
1380 		}
1381 
1382 		out->rmx_expire = expire;
1383 	}
1384 }
1385 
1386 void
1387 rtm_getmetrics(const struct rt_kmetrics *in, struct rt_metrics *out)
1388 {
1389 	int64_t expire;
1390 
1391 	expire = in->rmx_expire;
1392 	if (expire != 0) {
1393 		expire -= getuptime();
1394 		expire += gettime();
1395 	}
1396 
1397 	bzero(out, sizeof(*out));
1398 	out->rmx_locks = in->rmx_locks;
1399 	out->rmx_mtu = in->rmx_mtu;
1400 	out->rmx_expire = expire;
1401 	out->rmx_pksent = in->rmx_pksent;
1402 }
1403 
1404 #define ROUNDUP(a) \
1405 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
1406 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
1407 
1408 int
1409 rtm_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1410 {
1411 	struct sockaddr	*sa;
1412 	int		 i;
1413 
1414 	/*
1415 	 * Parse address bits, split address storage in chunks, and
1416 	 * set info pointers.  Use sa_len for traversing the memory
1417 	 * and check that we stay within in the limit.
1418 	 */
1419 	bzero(rtinfo->rti_info, sizeof(rtinfo->rti_info));
1420 	for (i = 0; i < sizeof(rtinfo->rti_addrs) * 8; i++) {
1421 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1422 			continue;
1423 		if (i >= RTAX_MAX || cp + sizeof(socklen_t) > cplim)
1424 			return (EINVAL);
1425 		sa = (struct sockaddr *)cp;
1426 		if (cp + sa->sa_len > cplim)
1427 			return (EINVAL);
1428 		rtinfo->rti_info[i] = sa;
1429 		ADVANCE(cp, sa);
1430 	}
1431 	/*
1432 	 * Check that the address family is suitable for the route address
1433 	 * type.  Check that each address has a size that fits its family
1434 	 * and its length is within the size.  Strings within addresses must
1435 	 * be NUL terminated.
1436 	 */
1437 	for (i = 0; i < RTAX_MAX; i++) {
1438 		size_t len, maxlen, size;
1439 
1440 		sa = rtinfo->rti_info[i];
1441 		if (sa == NULL)
1442 			continue;
1443 		maxlen = size = 0;
1444 		switch (i) {
1445 		case RTAX_DST:
1446 		case RTAX_GATEWAY:
1447 		case RTAX_SRC:
1448 			switch (sa->sa_family) {
1449 			case AF_INET:
1450 				size = sizeof(struct sockaddr_in);
1451 				break;
1452 			case AF_LINK:
1453 				size = sizeof(struct sockaddr_dl);
1454 				break;
1455 #ifdef INET6
1456 			case AF_INET6:
1457 				size = sizeof(struct sockaddr_in6);
1458 				break;
1459 #endif
1460 #ifdef MPLS
1461 			case AF_MPLS:
1462 				size = sizeof(struct sockaddr_mpls);
1463 				break;
1464 #endif
1465 			}
1466 			break;
1467 		case RTAX_IFP:
1468 			if (sa->sa_family != AF_LINK)
1469 				return (EAFNOSUPPORT);
1470 			/*
1471 			 * XXX Should be sizeof(struct sockaddr_dl), but
1472 			 * route(8) has a bug and provides less memory.
1473 			 * arp(8) has another bug and uses sizeof pointer.
1474 			 */
1475 			size = 4;
1476 			break;
1477 		case RTAX_IFA:
1478 			switch (sa->sa_family) {
1479 			case AF_INET:
1480 				size = sizeof(struct sockaddr_in);
1481 				break;
1482 #ifdef INET6
1483 			case AF_INET6:
1484 				size = sizeof(struct sockaddr_in6);
1485 				break;
1486 #endif
1487 			default:
1488 				return (EAFNOSUPPORT);
1489 			}
1490 			break;
1491 		case RTAX_LABEL:
1492 			sa->sa_family = AF_UNSPEC;
1493 			maxlen = RTLABEL_LEN;
1494 			size = sizeof(struct sockaddr_rtlabel);
1495 			break;
1496 #ifdef BFD
1497 		case RTAX_BFD:
1498 			sa->sa_family = AF_UNSPEC;
1499 			size = sizeof(struct sockaddr_bfd);
1500 			break;
1501 #endif
1502 		case RTAX_DNS:
1503 			/* more validation in rtm_validate_proposal */
1504 			if (sa->sa_len > sizeof(struct sockaddr_rtdns))
1505 				return (EINVAL);
1506 			if (sa->sa_len < offsetof(struct sockaddr_rtdns,
1507 			    sr_dns))
1508 				return (EINVAL);
1509 			switch (sa->sa_family) {
1510 			case AF_INET:
1511 #ifdef INET6
1512 			case AF_INET6:
1513 #endif
1514 				break;
1515 			default:
1516 				return (EAFNOSUPPORT);
1517 			}
1518 			break;
1519 		case RTAX_STATIC:
1520 			sa->sa_family = AF_UNSPEC;
1521 			maxlen = RTSTATIC_LEN;
1522 			size = sizeof(struct sockaddr_rtstatic);
1523 			break;
1524 		case RTAX_SEARCH:
1525 			sa->sa_family = AF_UNSPEC;
1526 			maxlen = RTSEARCH_LEN;
1527 			size = sizeof(struct sockaddr_rtsearch);
1528 			break;
1529 		}
1530 		if (size) {
1531 			/* memory for the full struct must be provided */
1532 			if (sa->sa_len < size)
1533 				return (EINVAL);
1534 		}
1535 		if (maxlen) {
1536 			/* this should not happen */
1537 			if (2 + maxlen > size)
1538 				return (EINVAL);
1539 			/* strings must be NUL terminated within the struct */
1540 			len = strnlen(sa->sa_data, maxlen);
1541 			if (len >= maxlen || 2 + len >= sa->sa_len)
1542 				return (EINVAL);
1543 			break;
1544 		}
1545 	}
1546 	return (0);
1547 }
1548 
1549 struct mbuf *
1550 rtm_msg1(int type, struct rt_addrinfo *rtinfo)
1551 {
1552 	struct rt_msghdr	*rtm;
1553 	struct mbuf		*m;
1554 	int			 i;
1555 	struct sockaddr		*sa;
1556 	int			 len, dlen, hlen;
1557 
1558 	switch (type) {
1559 	case RTM_DELADDR:
1560 	case RTM_NEWADDR:
1561 		len = sizeof(struct ifa_msghdr);
1562 		break;
1563 	case RTM_IFINFO:
1564 		len = sizeof(struct if_msghdr);
1565 		break;
1566 	case RTM_IFANNOUNCE:
1567 		len = sizeof(struct if_announcemsghdr);
1568 		break;
1569 #ifdef BFD
1570 	case RTM_BFD:
1571 		len = sizeof(struct bfd_msghdr);
1572 		break;
1573 #endif
1574 	case RTM_80211INFO:
1575 		len = sizeof(struct if_ieee80211_msghdr);
1576 		break;
1577 	default:
1578 		len = sizeof(struct rt_msghdr);
1579 		break;
1580 	}
1581 	if (len > MCLBYTES)
1582 		panic("rtm_msg1");
1583 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1584 	if (m && len > MHLEN) {
1585 		MCLGET(m, M_DONTWAIT);
1586 		if ((m->m_flags & M_EXT) == 0) {
1587 			m_free(m);
1588 			m = NULL;
1589 		}
1590 	}
1591 	if (m == NULL)
1592 		return (m);
1593 	m->m_pkthdr.len = m->m_len = hlen = len;
1594 	m->m_pkthdr.ph_ifidx = 0;
1595 	rtm = mtod(m, struct rt_msghdr *);
1596 	bzero(rtm, len);
1597 	for (i = 0; i < RTAX_MAX; i++) {
1598 		if (rtinfo == NULL || (sa = rtinfo->rti_info[i]) == NULL)
1599 			continue;
1600 		rtinfo->rti_addrs |= (1 << i);
1601 		dlen = ROUNDUP(sa->sa_len);
1602 		if (m_copyback(m, len, dlen, sa, M_NOWAIT)) {
1603 			m_freem(m);
1604 			return (NULL);
1605 		}
1606 		len += dlen;
1607 	}
1608 	rtm->rtm_msglen = len;
1609 	rtm->rtm_hdrlen = hlen;
1610 	rtm->rtm_version = RTM_VERSION;
1611 	rtm->rtm_type = type;
1612 	return (m);
1613 }
1614 
1615 int
1616 rtm_msg2(int type, int vers, struct rt_addrinfo *rtinfo, caddr_t cp,
1617     struct walkarg *w)
1618 {
1619 	int		i;
1620 	int		len, dlen, hlen, second_time = 0;
1621 	caddr_t		cp0;
1622 
1623 	rtinfo->rti_addrs = 0;
1624 again:
1625 	switch (type) {
1626 	case RTM_DELADDR:
1627 	case RTM_NEWADDR:
1628 		len = sizeof(struct ifa_msghdr);
1629 		break;
1630 	case RTM_IFINFO:
1631 		len = sizeof(struct if_msghdr);
1632 		break;
1633 	default:
1634 		len = sizeof(struct rt_msghdr);
1635 		break;
1636 	}
1637 	hlen = len;
1638 	if ((cp0 = cp) != NULL)
1639 		cp += len;
1640 	for (i = 0; i < RTAX_MAX; i++) {
1641 		struct sockaddr *sa;
1642 
1643 		if ((sa = rtinfo->rti_info[i]) == NULL)
1644 			continue;
1645 		rtinfo->rti_addrs |= (1 << i);
1646 		dlen = ROUNDUP(sa->sa_len);
1647 		if (cp) {
1648 			bcopy(sa, cp, (size_t)dlen);
1649 			cp += dlen;
1650 		}
1651 		len += dlen;
1652 	}
1653 	/* align message length to the next natural boundary */
1654 	len = ALIGN(len);
1655 	if (cp == 0 && w != NULL && !second_time) {
1656 		w->w_needed += len;
1657 		if (w->w_needed <= 0 && w->w_where) {
1658 			if (w->w_tmemsize < len) {
1659 				free(w->w_tmem, M_RTABLE, w->w_tmemsize);
1660 				w->w_tmem = malloc(len, M_RTABLE,
1661 				    M_NOWAIT | M_ZERO);
1662 				if (w->w_tmem)
1663 					w->w_tmemsize = len;
1664 			}
1665 			if (w->w_tmem) {
1666 				cp = w->w_tmem;
1667 				second_time = 1;
1668 				goto again;
1669 			} else
1670 				w->w_where = 0;
1671 		}
1672 	}
1673 	if (cp && w)		/* clear the message header */
1674 		bzero(cp0, hlen);
1675 
1676 	if (cp) {
1677 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1678 
1679 		rtm->rtm_version = RTM_VERSION;
1680 		rtm->rtm_type = type;
1681 		rtm->rtm_msglen = len;
1682 		rtm->rtm_hdrlen = hlen;
1683 	}
1684 	return (len);
1685 }
1686 
1687 void
1688 rtm_send(struct rtentry *rt, int cmd, int error, unsigned int rtableid)
1689 {
1690 	struct rt_addrinfo	 info;
1691 	struct ifnet		*ifp;
1692 	struct sockaddr_rtlabel	 sa_rl;
1693 	struct sockaddr_in6	 sa_mask;
1694 
1695 	memset(&info, 0, sizeof(info));
1696 	info.rti_info[RTAX_DST] = rt_key(rt);
1697 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1698 	if (!ISSET(rt->rt_flags, RTF_HOST))
1699 		info.rti_info[RTAX_NETMASK] = rt_plen2mask(rt, &sa_mask);
1700 	info.rti_info[RTAX_LABEL] = rtlabel_id2sa(rt->rt_labelid, &sa_rl);
1701 	ifp = if_get(rt->rt_ifidx);
1702 	if (ifp != NULL) {
1703 		info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
1704 		info.rti_info[RTAX_IFA] =
1705 		    rtable_getsource(rtableid, info.rti_info[RTAX_DST]->sa_family);
1706 		if (info.rti_info[RTAX_IFA] == NULL)
1707 			info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1708 	}
1709 
1710 	rtm_miss(cmd, &info, rt->rt_flags, rt->rt_priority, rt->rt_ifidx, error,
1711 	    rtableid);
1712 	if_put(ifp);
1713 }
1714 
1715 /*
1716  * This routine is called to generate a message from the routing
1717  * socket indicating that a redirect has occurred, a routing lookup
1718  * has failed, or that a protocol has detected timeouts to a particular
1719  * destination.
1720  */
1721 void
1722 rtm_miss(int type, struct rt_addrinfo *rtinfo, int flags, uint8_t prio,
1723     u_int ifidx, int error, u_int tableid)
1724 {
1725 	struct rt_msghdr	*rtm;
1726 	struct mbuf		*m;
1727 	struct sockaddr		*sa = rtinfo->rti_info[RTAX_DST];
1728 
1729 	if (rtptable.rtp_count == 0)
1730 		return;
1731 	m = rtm_msg1(type, rtinfo);
1732 	if (m == NULL)
1733 		return;
1734 	rtm = mtod(m, struct rt_msghdr *);
1735 	rtm->rtm_flags = RTF_DONE | flags;
1736 	rtm->rtm_priority = prio;
1737 	rtm->rtm_errno = error;
1738 	rtm->rtm_tableid = tableid;
1739 	rtm->rtm_addrs = rtinfo->rti_addrs;
1740 	rtm->rtm_index = ifidx;
1741 	route_input(m, NULL, sa ? sa->sa_family : AF_UNSPEC);
1742 }
1743 
1744 /*
1745  * This routine is called to generate a message from the routing
1746  * socket indicating that the status of a network interface has changed.
1747  */
1748 void
1749 rtm_ifchg(struct ifnet *ifp)
1750 {
1751 	struct rt_addrinfo	 info;
1752 	struct if_msghdr	*ifm;
1753 	struct mbuf		*m;
1754 
1755 	if (rtptable.rtp_count == 0)
1756 		return;
1757 	memset(&info, 0, sizeof(info));
1758 	info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
1759 	m = rtm_msg1(RTM_IFINFO, &info);
1760 	if (m == NULL)
1761 		return;
1762 	ifm = mtod(m, struct if_msghdr *);
1763 	ifm->ifm_index = ifp->if_index;
1764 	ifm->ifm_tableid = ifp->if_rdomain;
1765 	ifm->ifm_flags = ifp->if_flags;
1766 	ifm->ifm_xflags = ifp->if_xflags;
1767 	if_getdata(ifp, &ifm->ifm_data);
1768 	ifm->ifm_addrs = info.rti_addrs;
1769 	route_input(m, NULL, AF_UNSPEC);
1770 }
1771 
1772 /*
1773  * This is called to generate messages from the routing socket
1774  * indicating a network interface has had addresses associated with it.
1775  * if we ever reverse the logic and replace messages TO the routing
1776  * socket indicate a request to configure interfaces, then it will
1777  * be unnecessary as the routing socket will automatically generate
1778  * copies of it.
1779  */
1780 void
1781 rtm_addr(int cmd, struct ifaddr *ifa)
1782 {
1783 	struct ifnet		*ifp = ifa->ifa_ifp;
1784 	struct mbuf		*m;
1785 	struct rt_addrinfo	 info;
1786 	struct ifa_msghdr	*ifam;
1787 
1788 	if (rtptable.rtp_count == 0)
1789 		return;
1790 
1791 	memset(&info, 0, sizeof(info));
1792 	info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1793 	info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
1794 	info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1795 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1796 	if ((m = rtm_msg1(cmd, &info)) == NULL)
1797 		return;
1798 	ifam = mtod(m, struct ifa_msghdr *);
1799 	ifam->ifam_index = ifp->if_index;
1800 	ifam->ifam_metric = ifa->ifa_metric;
1801 	ifam->ifam_flags = ifa->ifa_flags;
1802 	ifam->ifam_addrs = info.rti_addrs;
1803 	ifam->ifam_tableid = ifp->if_rdomain;
1804 
1805 	route_input(m, NULL,
1806 	    ifa->ifa_addr ? ifa->ifa_addr->sa_family : AF_UNSPEC);
1807 }
1808 
1809 /*
1810  * This is called to generate routing socket messages indicating
1811  * network interface arrival and departure.
1812  */
1813 void
1814 rtm_ifannounce(struct ifnet *ifp, int what)
1815 {
1816 	struct if_announcemsghdr	*ifan;
1817 	struct mbuf			*m;
1818 
1819 	if (rtptable.rtp_count == 0)
1820 		return;
1821 	m = rtm_msg1(RTM_IFANNOUNCE, NULL);
1822 	if (m == NULL)
1823 		return;
1824 	ifan = mtod(m, struct if_announcemsghdr *);
1825 	ifan->ifan_index = ifp->if_index;
1826 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof(ifan->ifan_name));
1827 	ifan->ifan_what = what;
1828 	route_input(m, NULL, AF_UNSPEC);
1829 }
1830 
1831 #ifdef BFD
1832 /*
1833  * This is used to generate routing socket messages indicating
1834  * the state of a BFD session.
1835  */
1836 void
1837 rtm_bfd(struct bfd_config *bfd)
1838 {
1839 	struct bfd_msghdr	*bfdm;
1840 	struct sockaddr_bfd	 sa_bfd;
1841 	struct mbuf		*m;
1842 	struct rt_addrinfo	 info;
1843 
1844 	if (rtptable.rtp_count == 0)
1845 		return;
1846 	memset(&info, 0, sizeof(info));
1847 	info.rti_info[RTAX_DST] = rt_key(bfd->bc_rt);
1848 	info.rti_info[RTAX_IFA] = bfd->bc_rt->rt_ifa->ifa_addr;
1849 
1850 	m = rtm_msg1(RTM_BFD, &info);
1851 	if (m == NULL)
1852 		return;
1853 	bfdm = mtod(m, struct bfd_msghdr *);
1854 	bfdm->bm_addrs = info.rti_addrs;
1855 
1856 	bfd2sa(bfd->bc_rt, &sa_bfd);
1857 	memcpy(&bfdm->bm_sa, &sa_bfd, sizeof(sa_bfd));
1858 
1859 	route_input(m, NULL, info.rti_info[RTAX_DST]->sa_family);
1860 }
1861 #endif /* BFD */
1862 
1863 /*
1864  * This is used to generate routing socket messages indicating
1865  * the state of an ieee80211 interface.
1866  */
1867 void
1868 rtm_80211info(struct ifnet *ifp, struct if_ieee80211_data *ifie)
1869 {
1870 	struct if_ieee80211_msghdr	*ifim;
1871 	struct mbuf			*m;
1872 
1873 	if (rtptable.rtp_count == 0)
1874 		return;
1875 	m = rtm_msg1(RTM_80211INFO, NULL);
1876 	if (m == NULL)
1877 		return;
1878 	ifim = mtod(m, struct if_ieee80211_msghdr *);
1879 	ifim->ifim_index = ifp->if_index;
1880 	ifim->ifim_tableid = ifp->if_rdomain;
1881 
1882 	memcpy(&ifim->ifim_ifie, ifie, sizeof(ifim->ifim_ifie));
1883 	route_input(m, NULL, AF_UNSPEC);
1884 }
1885 
1886 /*
1887  * This is used to generate routing socket messages indicating
1888  * the address selection proposal from an interface.
1889  */
1890 void
1891 rtm_proposal(struct ifnet *ifp, struct rt_addrinfo *rtinfo, int flags,
1892     uint8_t prio)
1893 {
1894 	struct rt_msghdr	*rtm;
1895 	struct mbuf		*m;
1896 
1897 	m = rtm_msg1(RTM_PROPOSAL, rtinfo);
1898 	if (m == NULL)
1899 		return;
1900 	rtm = mtod(m, struct rt_msghdr *);
1901 	rtm->rtm_flags = RTF_DONE | flags;
1902 	rtm->rtm_priority = prio;
1903 	rtm->rtm_tableid = ifp->if_rdomain;
1904 	rtm->rtm_index = ifp->if_index;
1905 	rtm->rtm_addrs = rtinfo->rti_addrs;
1906 
1907 	route_input(m, NULL, rtinfo->rti_info[RTAX_DNS]->sa_family);
1908 }
1909 
1910 /*
1911  * This is used in dumping the kernel table via sysctl().
1912  */
1913 int
1914 sysctl_dumpentry(struct rtentry *rt, void *v, unsigned int id)
1915 {
1916 	struct walkarg		*w = v;
1917 	int			 error = 0, size;
1918 	struct rt_addrinfo	 info;
1919 	struct ifnet		*ifp;
1920 #ifdef BFD
1921 	struct sockaddr_bfd	 sa_bfd;
1922 #endif
1923 	struct sockaddr_rtlabel	 sa_rl;
1924 	struct sockaddr_in6	 sa_mask;
1925 
1926 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1927 		return 0;
1928 	if (w->w_op == NET_RT_DUMP && w->w_arg) {
1929 		u_int8_t prio = w->w_arg & RTP_MASK;
1930 		if (w->w_arg < 0) {
1931 			prio = (-w->w_arg) & RTP_MASK;
1932 			/* Show all routes that are not this priority */
1933 			if (prio == (rt->rt_priority & RTP_MASK))
1934 				return 0;
1935 		} else {
1936 			if (prio != (rt->rt_priority & RTP_MASK) &&
1937 			    prio != RTP_ANY)
1938 				return 0;
1939 		}
1940 	}
1941 	bzero(&info, sizeof(info));
1942 	info.rti_info[RTAX_DST] = rt_key(rt);
1943 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1944 	info.rti_info[RTAX_NETMASK] = rt_plen2mask(rt, &sa_mask);
1945 	ifp = if_get(rt->rt_ifidx);
1946 	if (ifp != NULL) {
1947 		info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
1948 		info.rti_info[RTAX_IFA] =
1949 		    rtable_getsource(id, info.rti_info[RTAX_DST]->sa_family);
1950 		if (info.rti_info[RTAX_IFA] == NULL)
1951 			info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1952 		if (ifp->if_flags & IFF_POINTOPOINT)
1953 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1954 	}
1955 	if_put(ifp);
1956 	info.rti_info[RTAX_LABEL] = rtlabel_id2sa(rt->rt_labelid, &sa_rl);
1957 #ifdef BFD
1958 	if (rt->rt_flags & RTF_BFD)
1959 		info.rti_info[RTAX_BFD] = bfd2sa(rt, &sa_bfd);
1960 #endif
1961 #ifdef MPLS
1962 	if (rt->rt_flags & RTF_MPLS) {
1963 		struct sockaddr_mpls	 sa_mpls;
1964 
1965 		bzero(&sa_mpls, sizeof(sa_mpls));
1966 		sa_mpls.smpls_family = AF_MPLS;
1967 		sa_mpls.smpls_len = sizeof(sa_mpls);
1968 		sa_mpls.smpls_label = ((struct rt_mpls *)
1969 		    rt->rt_llinfo)->mpls_label;
1970 		info.rti_info[RTAX_SRC] = (struct sockaddr *)&sa_mpls;
1971 		info.rti_mpls = ((struct rt_mpls *)
1972 		    rt->rt_llinfo)->mpls_operation;
1973 	}
1974 #endif
1975 
1976 	size = rtm_msg2(RTM_GET, RTM_VERSION, &info, NULL, w);
1977 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1978 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1979 
1980 		rtm->rtm_pid = curproc->p_p->ps_pid;
1981 		rtm->rtm_flags = rt->rt_flags;
1982 		rtm->rtm_priority = rt->rt_priority & RTP_MASK;
1983 		rtm_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1984 		/* Do not account the routing table's reference. */
1985 		rtm->rtm_rmx.rmx_refcnt = rt->rt_refcnt - 1;
1986 		rtm->rtm_index = rt->rt_ifidx;
1987 		rtm->rtm_addrs = info.rti_addrs;
1988 		rtm->rtm_tableid = id;
1989 #ifdef MPLS
1990 		rtm->rtm_mpls = info.rti_mpls;
1991 #endif
1992 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1993 			w->w_where = NULL;
1994 		else
1995 			w->w_where += size;
1996 	}
1997 	return (error);
1998 }
1999 
2000 int
2001 sysctl_iflist(int af, struct walkarg *w)
2002 {
2003 	struct ifnet		*ifp;
2004 	struct ifaddr		*ifa;
2005 	struct rt_addrinfo	 info;
2006 	int			 len, error = 0;
2007 
2008 	bzero(&info, sizeof(info));
2009 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
2010 		if (w->w_arg && w->w_arg != ifp->if_index)
2011 			continue;
2012 		/* Copy the link-layer address first */
2013 		info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
2014 		len = rtm_msg2(RTM_IFINFO, RTM_VERSION, &info, 0, w);
2015 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
2016 			struct if_msghdr *ifm;
2017 
2018 			ifm = (struct if_msghdr *)w->w_tmem;
2019 			ifm->ifm_index = ifp->if_index;
2020 			ifm->ifm_tableid = ifp->if_rdomain;
2021 			ifm->ifm_flags = ifp->if_flags;
2022 			if_getdata(ifp, &ifm->ifm_data);
2023 			ifm->ifm_addrs = info.rti_addrs;
2024 			error = copyout(ifm, w->w_where, len);
2025 			if (error)
2026 				return (error);
2027 			w->w_where += len;
2028 		}
2029 		info.rti_info[RTAX_IFP] = NULL;
2030 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2031 			KASSERT(ifa->ifa_addr->sa_family != AF_LINK);
2032 			if (af && af != ifa->ifa_addr->sa_family)
2033 				continue;
2034 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2035 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
2036 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2037 			len = rtm_msg2(RTM_NEWADDR, RTM_VERSION, &info, 0, w);
2038 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
2039 				struct ifa_msghdr *ifam;
2040 
2041 				ifam = (struct ifa_msghdr *)w->w_tmem;
2042 				ifam->ifam_index = ifa->ifa_ifp->if_index;
2043 				ifam->ifam_flags = ifa->ifa_flags;
2044 				ifam->ifam_metric = ifa->ifa_metric;
2045 				ifam->ifam_addrs = info.rti_addrs;
2046 				error = copyout(w->w_tmem, w->w_where, len);
2047 				if (error)
2048 					return (error);
2049 				w->w_where += len;
2050 			}
2051 		}
2052 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
2053 		    info.rti_info[RTAX_BRD] = NULL;
2054 	}
2055 	return (0);
2056 }
2057 
2058 int
2059 sysctl_ifnames(struct walkarg *w)
2060 {
2061 	struct if_nameindex_msg ifn;
2062 	struct ifnet *ifp;
2063 	int error = 0;
2064 
2065 	/* XXX ignore tableid for now */
2066 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
2067 		if (w->w_arg && w->w_arg != ifp->if_index)
2068 			continue;
2069 		w->w_needed += sizeof(ifn);
2070 		if (w->w_where && w->w_needed <= 0) {
2071 
2072 			memset(&ifn, 0, sizeof(ifn));
2073 			ifn.if_index = ifp->if_index;
2074 			strlcpy(ifn.if_name, ifp->if_xname,
2075 			    sizeof(ifn.if_name));
2076 			error = copyout(&ifn, w->w_where, sizeof(ifn));
2077 			if (error)
2078 				return (error);
2079 			w->w_where += sizeof(ifn);
2080 		}
2081 	}
2082 
2083 	return (0);
2084 }
2085 
2086 int
2087 sysctl_source(int af, u_int tableid, struct walkarg *w)
2088 {
2089 	struct sockaddr	*sa;
2090 	int		 size, error = 0;
2091 
2092 	sa = rtable_getsource(tableid, af);
2093 	if (sa) {
2094 		switch (sa->sa_family) {
2095 		case AF_INET:
2096 			size = sizeof(struct sockaddr_in);
2097 			break;
2098 #ifdef INET6
2099 		case AF_INET6:
2100 			size = sizeof(struct sockaddr_in6);
2101 			break;
2102 #endif
2103 		default:
2104 			return (0);
2105 		}
2106 		w->w_needed += size;
2107 		if (w->w_where && w->w_needed <= 0) {
2108 			if ((error = copyout(sa, w->w_where, size)))
2109 				return (error);
2110 			w->w_where += size;
2111 		}
2112 	}
2113 	return (0);
2114 }
2115 
2116 int
2117 sysctl_rtable(int *name, u_int namelen, void *where, size_t *given, void *new,
2118     size_t newlen)
2119 {
2120 	int			 i, error = EINVAL;
2121 	u_char			 af;
2122 	struct walkarg		 w;
2123 	struct rt_tableinfo	 tableinfo;
2124 	u_int			 tableid = 0;
2125 
2126 	if (new)
2127 		return (EPERM);
2128 	if (namelen < 3 || namelen > 4)
2129 		return (EINVAL);
2130 	af = name[0];
2131 	bzero(&w, sizeof(w));
2132 	w.w_where = where;
2133 	w.w_given = *given;
2134 	w.w_needed = 0 - w.w_given;
2135 	w.w_op = name[1];
2136 	w.w_arg = name[2];
2137 
2138 	if (namelen == 4) {
2139 		tableid = name[3];
2140 		if (!rtable_exists(tableid))
2141 			return (ENOENT);
2142 	} else
2143 		tableid = curproc->p_p->ps_rtableid;
2144 
2145 	switch (w.w_op) {
2146 	case NET_RT_DUMP:
2147 	case NET_RT_FLAGS:
2148 		NET_LOCK();
2149 		for (i = 1; i <= AF_MAX; i++) {
2150 			if (af != 0 && af != i)
2151 				continue;
2152 
2153 			error = rtable_walk(tableid, i, NULL, sysctl_dumpentry,
2154 			    &w);
2155 			if (error == EAFNOSUPPORT)
2156 				error = 0;
2157 			if (error)
2158 				break;
2159 		}
2160 		NET_UNLOCK();
2161 		break;
2162 
2163 	case NET_RT_IFLIST:
2164 		NET_LOCK();
2165 		error = sysctl_iflist(af, &w);
2166 		NET_UNLOCK();
2167 		break;
2168 
2169 	case NET_RT_STATS:
2170 		return (sysctl_rtable_rtstat(where, given, new));
2171 	case NET_RT_TABLE:
2172 		tableid = w.w_arg;
2173 		if (!rtable_exists(tableid))
2174 			return (ENOENT);
2175 		memset(&tableinfo, 0, sizeof tableinfo);
2176 		tableinfo.rti_tableid = tableid;
2177 		tableinfo.rti_domainid = rtable_l2(tableid);
2178 		error = sysctl_rdstruct(where, given, new,
2179 		    &tableinfo, sizeof(tableinfo));
2180 		return (error);
2181 	case NET_RT_IFNAMES:
2182 		NET_LOCK();
2183 		error = sysctl_ifnames(&w);
2184 		NET_UNLOCK();
2185 		break;
2186 	case NET_RT_SOURCE:
2187 		tableid = w.w_arg;
2188 		if (!rtable_exists(tableid))
2189 			return (ENOENT);
2190 		NET_LOCK();
2191 		for (i = 1; i <= AF_MAX; i++) {
2192 			if (af != 0 && af != i)
2193 				continue;
2194 
2195 			error = sysctl_source(i, tableid, &w);
2196 			if (error == EAFNOSUPPORT)
2197 				error = 0;
2198 			if (error)
2199 				break;
2200 		}
2201 		NET_UNLOCK();
2202 		break;
2203 	}
2204 	free(w.w_tmem, M_RTABLE, w.w_tmemsize);
2205 	w.w_needed += w.w_given;
2206 	if (where) {
2207 		*given = w.w_where - (caddr_t)where;
2208 		if (*given < w.w_needed)
2209 			return (ENOMEM);
2210 	} else
2211 		*given = (11 * w.w_needed) / 10;
2212 
2213 	return (error);
2214 }
2215 
2216 int
2217 sysctl_rtable_rtstat(void *oldp, size_t *oldlenp, void *newp)
2218 {
2219 	extern struct cpumem *rtcounters;
2220 	uint64_t counters[rts_ncounters];
2221 	struct rtstat rtstat;
2222 	uint32_t *words = (uint32_t *)&rtstat;
2223 	int i;
2224 
2225 	CTASSERT(sizeof(rtstat) == (nitems(counters) * sizeof(uint32_t)));
2226 	memset(&rtstat, 0, sizeof rtstat);
2227 	counters_read(rtcounters, counters, nitems(counters));
2228 
2229 	for (i = 0; i < nitems(counters); i++)
2230 		words[i] = (uint32_t)counters[i];
2231 
2232 	return (sysctl_rdstruct(oldp, oldlenp, newp, &rtstat, sizeof(rtstat)));
2233 }
2234 
2235 int
2236 rtm_validate_proposal(struct rt_addrinfo *info)
2237 {
2238 	if (info->rti_addrs & ~(RTA_NETMASK | RTA_IFA | RTA_DNS | RTA_STATIC |
2239 	    RTA_SEARCH)) {
2240 		return -1;
2241 	}
2242 
2243 	if (ISSET(info->rti_addrs, RTA_NETMASK)) {
2244 		struct sockaddr *sa = info->rti_info[RTAX_NETMASK];
2245 		if (sa == NULL)
2246 			return -1;
2247 		switch (sa->sa_family) {
2248 		case AF_INET:
2249 			if (sa->sa_len != sizeof(struct sockaddr_in))
2250 				return -1;
2251 			break;
2252 		case AF_INET6:
2253 			if (sa->sa_len != sizeof(struct sockaddr_in6))
2254 				return -1;
2255 			break;
2256 		default:
2257 			return -1;
2258 		}
2259 	}
2260 
2261 	if (ISSET(info->rti_addrs, RTA_IFA)) {
2262 		struct sockaddr *sa = info->rti_info[RTAX_IFA];
2263 		if (sa == NULL)
2264 			return -1;
2265 		switch (sa->sa_family) {
2266 		case AF_INET:
2267 			if (sa->sa_len != sizeof(struct sockaddr_in))
2268 				return -1;
2269 			break;
2270 		case AF_INET6:
2271 			if (sa->sa_len != sizeof(struct sockaddr_in6))
2272 				return -1;
2273 			break;
2274 		default:
2275 			return -1;
2276 		}
2277 	}
2278 
2279 	if (ISSET(info->rti_addrs, RTA_DNS)) {
2280 		struct sockaddr_rtdns *rtdns =
2281 		    (struct sockaddr_rtdns *)info->rti_info[RTAX_DNS];
2282 		if (rtdns == NULL)
2283 			return -1;
2284 		if (rtdns->sr_len > sizeof(*rtdns))
2285 			return -1;
2286 		if (rtdns->sr_len < offsetof(struct sockaddr_rtdns, sr_dns))
2287 			return -1;
2288 		switch (rtdns->sr_family) {
2289 		case AF_INET:
2290 			if ((rtdns->sr_len - offsetof(struct sockaddr_rtdns,
2291 			    sr_dns)) % sizeof(struct in_addr) != 0)
2292 				return -1;
2293 			break;
2294 #ifdef INET6
2295 		case AF_INET6:
2296 			if ((rtdns->sr_len - offsetof(struct sockaddr_rtdns,
2297 			    sr_dns)) % sizeof(struct in6_addr) != 0)
2298 				return -1;
2299 			break;
2300 #endif
2301 		default:
2302 			return -1;
2303 		}
2304 	}
2305 
2306 	if (ISSET(info->rti_addrs, RTA_STATIC)) {
2307 		struct sockaddr_rtstatic *rtstatic =
2308 		    (struct sockaddr_rtstatic *)info->rti_info[RTAX_STATIC];
2309 		if (rtstatic == NULL)
2310 			return -1;
2311 		if (rtstatic->sr_len > sizeof(*rtstatic))
2312 			return -1;
2313 		if (rtstatic->sr_len <=
2314 		    offsetof(struct sockaddr_rtstatic, sr_static))
2315 			return -1;
2316 	}
2317 
2318 	if (ISSET(info->rti_addrs, RTA_SEARCH)) {
2319 		struct sockaddr_rtsearch *rtsearch =
2320 		    (struct sockaddr_rtsearch *)info->rti_info[RTAX_SEARCH];
2321 		if (rtsearch == NULL)
2322 			return -1;
2323 		if (rtsearch->sr_len > sizeof(*rtsearch))
2324 			return -1;
2325 		if (rtsearch->sr_len <=
2326 		    offsetof(struct sockaddr_rtsearch, sr_search))
2327 			return -1;
2328 	}
2329 
2330 	return 0;
2331 }
2332 
2333 int
2334 rt_setsource(unsigned int rtableid, struct sockaddr *src)
2335 {
2336 	struct ifaddr	*ifa;
2337 	int		error;
2338 	/*
2339 	 * If source address is 0.0.0.0 or ::
2340 	 * use automatic source selection
2341 	 */
2342 	switch(src->sa_family) {
2343 	case AF_INET:
2344 		if(satosin(src)->sin_addr.s_addr == INADDR_ANY) {
2345 			rtable_setsource(rtableid, AF_INET, NULL);
2346 			return (0);
2347 		}
2348 		break;
2349 #ifdef INET6
2350 	case AF_INET6:
2351 		if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(src)->sin6_addr)) {
2352 			rtable_setsource(rtableid, AF_INET6, NULL);
2353 			return (0);
2354 		}
2355 		break;
2356 #endif
2357 	default:
2358 		return (EAFNOSUPPORT);
2359 	}
2360 
2361 	KERNEL_LOCK();
2362 	/*
2363 	 * Check if source address is assigned to an interface in the
2364 	 * same rdomain
2365 	 */
2366 	if ((ifa = ifa_ifwithaddr(src, rtableid)) == NULL) {
2367 		KERNEL_UNLOCK();
2368 		return (EINVAL);
2369 	}
2370 
2371 	error = rtable_setsource(rtableid, src->sa_family, ifa->ifa_addr);
2372 	KERNEL_UNLOCK();
2373 
2374 	return (error);
2375 }
2376 
2377 /*
2378  * Definitions of protocols supported in the ROUTE domain.
2379  */
2380 
2381 const struct protosw routesw[] = {
2382 {
2383   .pr_type	= SOCK_RAW,
2384   .pr_domain	= &routedomain,
2385   .pr_flags	= PR_ATOMIC|PR_ADDR|PR_WANTRCVD,
2386   .pr_output	= route_output,
2387   .pr_ctloutput	= route_ctloutput,
2388   .pr_usrreq	= route_usrreq,
2389   .pr_attach	= route_attach,
2390   .pr_detach	= route_detach,
2391   .pr_init	= route_prinit,
2392   .pr_sysctl	= sysctl_rtable
2393 }
2394 };
2395 
2396 const struct domain routedomain = {
2397   .dom_family = PF_ROUTE,
2398   .dom_name = "route",
2399   .dom_init = route_init,
2400   .dom_protosw = routesw,
2401   .dom_protoswNPROTOSW = &routesw[nitems(routesw)]
2402 };
2403