xref: /openbsd-src/sys/net/rtsock.c (revision 505ee9ea3b177e2387d907a91ca7da069f3f14d8)
1 /*	$OpenBSD: rtsock.c,v 1.299 2020/06/24 22:03:42 cheloha 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 *, int);
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 struct rtpcb {
142 	struct socket		*rop_socket;
143 
144 	SRPL_ENTRY(rtpcb)	rop_list;
145 	struct refcnt		rop_refcnt;
146 	struct timeout		rop_timeout;
147 	unsigned int		rop_msgfilter;
148 	unsigned int		rop_flags;
149 	u_int			rop_rtableid;
150 	unsigned short		rop_proto;
151 	u_char			rop_priority;
152 };
153 #define	sotortpcb(so)	((struct rtpcb *)(so)->so_pcb)
154 
155 struct rtptable {
156 	SRPL_HEAD(, rtpcb)	rtp_list;
157 	struct srpl_rc		rtp_rc;
158 	struct rwlock		rtp_lk;
159 	unsigned int		rtp_count;
160 };
161 
162 struct pool rtpcb_pool;
163 struct rtptable rtptable;
164 
165 /*
166  * These flags and timeout are used for indicating to userland (via a
167  * RTM_DESYNC msg) when the route socket has overflowed and messages
168  * have been lost.
169  */
170 #define ROUTECB_FLAG_DESYNC	0x1	/* Route socket out of memory */
171 #define ROUTECB_FLAG_FLUSH	0x2	/* Wait until socket is empty before
172 					   queueing more packets */
173 
174 #define ROUTE_DESYNC_RESEND_TIMEOUT	200	/* In ms */
175 
176 void
177 route_prinit(void)
178 {
179 	srpl_rc_init(&rtptable.rtp_rc, rcb_ref, rcb_unref, NULL);
180 	rw_init(&rtptable.rtp_lk, "rtsock");
181 	SRPL_INIT(&rtptable.rtp_list);
182 	pool_init(&rtpcb_pool, sizeof(struct rtpcb), 0,
183 	    IPL_NONE, PR_WAITOK, "rtpcb", NULL);
184 }
185 
186 void
187 rcb_ref(void *null, void *v)
188 {
189 	struct rtpcb *rop = v;
190 
191 	refcnt_take(&rop->rop_refcnt);
192 }
193 
194 void
195 rcb_unref(void *null, void *v)
196 {
197 	struct rtpcb *rop = v;
198 
199 	refcnt_rele_wake(&rop->rop_refcnt);
200 }
201 
202 int
203 route_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
204     struct mbuf *control, struct proc *p)
205 {
206 	struct rtpcb	*rop;
207 	int		 error = 0;
208 
209 	if (req == PRU_CONTROL)
210 		return (EOPNOTSUPP);
211 
212 	soassertlocked(so);
213 
214 	if (control && control->m_len) {
215 		error = EOPNOTSUPP;
216 		goto release;
217 	}
218 
219 	rop = sotortpcb(so);
220 	if (rop == NULL) {
221 		error = EINVAL;
222 		goto release;
223 	}
224 
225 	switch (req) {
226 	/* no connect, bind, accept. Socket is connected from the start */
227 	case PRU_CONNECT:
228 	case PRU_BIND:
229 	case PRU_CONNECT2:
230 	case PRU_LISTEN:
231 	case PRU_ACCEPT:
232 		error = EOPNOTSUPP;
233 		break;
234 
235 	case PRU_DISCONNECT:
236 	case PRU_ABORT:
237 		soisdisconnected(so);
238 		break;
239 	case PRU_SHUTDOWN:
240 		socantsendmore(so);
241 		break;
242 	case PRU_SENSE:
243 		/* stat: don't bother with a blocksize. */
244 		break;
245 
246 	/* minimal support, just implement a fake peer address */
247 	case PRU_SOCKADDR:
248 		error = EINVAL;
249 		break;
250 	case PRU_PEERADDR:
251 		bcopy(&route_src, mtod(nam, caddr_t), route_src.sa_len);
252 		nam->m_len = route_src.sa_len;
253 		break;
254 
255 	case PRU_RCVD:
256 		/*
257 		 * If we are in a FLUSH state, check if the buffer is
258 		 * empty so that we can clear the flag.
259 		 */
260 		if (((rop->rop_flags & ROUTECB_FLAG_FLUSH) != 0) &&
261 		    ((sbspace(rop->rop_socket, &rop->rop_socket->so_rcv) ==
262 		    rop->rop_socket->so_rcv.sb_hiwat)))
263 			rop->rop_flags &= ~ROUTECB_FLAG_FLUSH;
264 		break;
265 
266 	case PRU_RCVOOB:
267 	case PRU_SENDOOB:
268 		error = EOPNOTSUPP;
269 		break;
270 	case PRU_SEND:
271 		if (nam) {
272 			error = EISCONN;
273 			break;
274 		}
275 		error = (*so->so_proto->pr_output)(m, so, NULL, NULL);
276 		m = NULL;
277 		break;
278 	default:
279 		panic("route_usrreq");
280 	}
281 
282  release:
283 	if (req != PRU_RCVD && req != PRU_RCVOOB && req != PRU_SENSE) {
284 		m_freem(control);
285 		m_freem(m);
286 	}
287 	return (error);
288 }
289 
290 int
291 route_attach(struct socket *so, int proto)
292 {
293 	struct rtpcb	*rop;
294 	int		 error;
295 
296 	/*
297 	 * use the rawcb but allocate a rtpcb, this
298 	 * code does not care about the additional fields
299 	 * and works directly on the raw socket.
300 	 */
301 	rop = pool_get(&rtpcb_pool, PR_WAITOK|PR_ZERO);
302 	so->so_pcb = rop;
303 	/* Init the timeout structure */
304 	timeout_set(&rop->rop_timeout, rtm_senddesync_timer, so);
305 	refcnt_init(&rop->rop_refcnt);
306 
307 	if (curproc == NULL)
308 		error = EACCES;
309 	else
310 		error = soreserve(so, ROUTESNDQ, ROUTERCVQ);
311 	if (error) {
312 		pool_put(&rtpcb_pool, rop);
313 		return (error);
314 	}
315 
316 	rop->rop_socket = so;
317 	rop->rop_proto = proto;
318 
319 	rop->rop_rtableid = curproc->p_p->ps_rtableid;
320 
321 	soisconnected(so);
322 	so->so_options |= SO_USELOOPBACK;
323 
324 	rw_enter(&rtptable.rtp_lk, RW_WRITE);
325 	SRPL_INSERT_HEAD_LOCKED(&rtptable.rtp_rc, &rtptable.rtp_list, rop,
326 	    rop_list);
327 	rtptable.rtp_count++;
328 	rw_exit(&rtptable.rtp_lk);
329 
330 	return (0);
331 }
332 
333 int
334 route_detach(struct socket *so)
335 {
336 	struct rtpcb	*rop;
337 
338 	soassertlocked(so);
339 
340 	rop = sotortpcb(so);
341 	if (rop == NULL)
342 		return (EINVAL);
343 
344 	rw_enter(&rtptable.rtp_lk, RW_WRITE);
345 
346 	timeout_del(&rop->rop_timeout);
347 	rtptable.rtp_count--;
348 
349 	SRPL_REMOVE_LOCKED(&rtptable.rtp_rc, &rtptable.rtp_list, rop, rtpcb,
350 	    rop_list);
351 	rw_exit(&rtptable.rtp_lk);
352 
353 	/* wait for all references to drop */
354 	refcnt_finalize(&rop->rop_refcnt, "rtsockrefs");
355 
356 	so->so_pcb = NULL;
357 	KASSERT((so->so_state & SS_NOFDREF) == 0);
358 	pool_put(&rtpcb_pool, rop);
359 
360 	return (0);
361 }
362 
363 int
364 route_ctloutput(int op, struct socket *so, int level, int optname,
365     struct mbuf *m)
366 {
367 	struct rtpcb *rop = sotortpcb(so);
368 	int error = 0;
369 	unsigned int tid, prio;
370 
371 	if (level != AF_ROUTE)
372 		return (EINVAL);
373 
374 	switch (op) {
375 	case PRCO_SETOPT:
376 		switch (optname) {
377 		case ROUTE_MSGFILTER:
378 			if (m == NULL || m->m_len != sizeof(unsigned int))
379 				error = EINVAL;
380 			else
381 				rop->rop_msgfilter = *mtod(m, unsigned int *);
382 			break;
383 		case ROUTE_TABLEFILTER:
384 			if (m == NULL || m->m_len != sizeof(unsigned int)) {
385 				error = EINVAL;
386 				break;
387 			}
388 			tid = *mtod(m, unsigned int *);
389 			if (tid != RTABLE_ANY && !rtable_exists(tid))
390 				error = ENOENT;
391 			else
392 				rop->rop_rtableid = tid;
393 			break;
394 		case ROUTE_PRIOFILTER:
395 			if (m == NULL || m->m_len != sizeof(unsigned int)) {
396 				error = EINVAL;
397 				break;
398 			}
399 			prio = *mtod(m, unsigned int *);
400 			if (prio > RTP_MAX)
401 				error = EINVAL;
402 			else
403 				rop->rop_priority = prio;
404 			break;
405 		default:
406 			error = ENOPROTOOPT;
407 			break;
408 		}
409 		break;
410 	case PRCO_GETOPT:
411 		switch (optname) {
412 		case ROUTE_MSGFILTER:
413 			m->m_len = sizeof(unsigned int);
414 			*mtod(m, unsigned int *) = rop->rop_msgfilter;
415 			break;
416 		case ROUTE_TABLEFILTER:
417 			m->m_len = sizeof(unsigned int);
418 			*mtod(m, unsigned int *) = rop->rop_rtableid;
419 			break;
420 		case ROUTE_PRIOFILTER:
421 			m->m_len = sizeof(unsigned int);
422 			*mtod(m, unsigned int *) = rop->rop_priority;
423 			break;
424 		default:
425 			error = ENOPROTOOPT;
426 			break;
427 		}
428 	}
429 	return (error);
430 }
431 
432 void
433 rtm_senddesync_timer(void *xso)
434 {
435 	struct socket	*so = xso;
436 	int		 s;
437 
438 	s = solock(so);
439 	rtm_senddesync(so);
440 	sounlock(so, s);
441 }
442 
443 void
444 rtm_senddesync(struct socket *so)
445 {
446 	struct rtpcb	*rop = sotortpcb(so);
447 	struct mbuf	*desync_mbuf;
448 
449 	soassertlocked(so);
450 
451 	/* If we are in a DESYNC state, try to send a RTM_DESYNC packet */
452 	if ((rop->rop_flags & ROUTECB_FLAG_DESYNC) == 0)
453 		return;
454 
455 	/*
456 	 * If we fail to alloc memory or if sbappendaddr()
457 	 * fails, re-add timeout and try again.
458 	 */
459 	desync_mbuf = rtm_msg1(RTM_DESYNC, NULL);
460 	if (desync_mbuf != NULL) {
461 		if (sbappendaddr(so, &so->so_rcv, &route_src,
462 		    desync_mbuf, NULL) != 0) {
463 			rop->rop_flags &= ~ROUTECB_FLAG_DESYNC;
464 			sorwakeup(rop->rop_socket);
465 			return;
466 		}
467 		m_freem(desync_mbuf);
468 	}
469 	/* Re-add timeout to try sending msg again */
470 	timeout_add_msec(&rop->rop_timeout, ROUTE_DESYNC_RESEND_TIMEOUT);
471 }
472 
473 void
474 route_input(struct mbuf *m0, struct socket *so0, sa_family_t sa_family)
475 {
476 	struct socket *so;
477 	struct rtpcb *rop;
478 	struct rt_msghdr *rtm;
479 	struct mbuf *m = m0;
480 	struct socket *last = NULL;
481 	struct srp_ref sr;
482 	int s;
483 
484 	/* ensure that we can access the rtm_type via mtod() */
485 	if (m->m_len < offsetof(struct rt_msghdr, rtm_type) + 1) {
486 		m_freem(m);
487 		return;
488 	}
489 
490 	SRPL_FOREACH(rop, &sr, &rtptable.rtp_list, rop_list) {
491 		/*
492 		 * If route socket is bound to an address family only send
493 		 * messages that match the address family. Address family
494 		 * agnostic messages are always sent.
495 		 */
496 		if (sa_family != AF_UNSPEC && rop->rop_proto != AF_UNSPEC &&
497 		    rop->rop_proto != sa_family)
498 			continue;
499 
500 
501 		so = rop->rop_socket;
502 		s = solock(so);
503 
504 		/*
505 		 * Check to see if we don't want our own messages and
506 		 * if we can receive anything.
507 		 */
508 		if ((so0 == so && !(so0->so_options & SO_USELOOPBACK)) ||
509 		    !(so->so_state & SS_ISCONNECTED) ||
510 		    (so->so_state & SS_CANTRCVMORE)) {
511 next:
512 			sounlock(so, s);
513 			continue;
514 		}
515 
516 		/* filter messages that the process does not want */
517 		rtm = mtod(m, struct rt_msghdr *);
518 		/* but RTM_DESYNC can't be filtered */
519 		if (rtm->rtm_type != RTM_DESYNC && rop->rop_msgfilter != 0 &&
520 		    !(rop->rop_msgfilter & (1 << rtm->rtm_type)))
521 			goto next;
522 		switch (rtm->rtm_type) {
523 		case RTM_IFANNOUNCE:
524 		case RTM_DESYNC:
525 			/* no tableid */
526 			break;
527 		case RTM_RESOLVE:
528 		case RTM_NEWADDR:
529 		case RTM_DELADDR:
530 		case RTM_IFINFO:
531 		case RTM_80211INFO:
532 		case RTM_BFD:
533 			/* check against rdomain id */
534 			if (rop->rop_rtableid != RTABLE_ANY &&
535 			    rtable_l2(rop->rop_rtableid) != rtm->rtm_tableid)
536 				goto next;
537 			break;
538 		default:
539 			if (rop->rop_priority != 0 &&
540 			    rop->rop_priority < rtm->rtm_priority)
541 				goto next;
542 			/* check against rtable id */
543 			if (rop->rop_rtableid != RTABLE_ANY &&
544 			    rop->rop_rtableid != rtm->rtm_tableid)
545 				goto next;
546 			break;
547 		}
548 
549 		/*
550 		 * Check to see if the flush flag is set. If so, don't queue
551 		 * any more messages until the flag is cleared.
552 		 */
553 		if ((rop->rop_flags & ROUTECB_FLAG_FLUSH) != 0)
554 			goto next;
555 		sounlock(so, s);
556 
557 		if (last) {
558 			s = solock(last);
559 			rtm_sendup(last, m, 1);
560 			sounlock(last, s);
561 			refcnt_rele_wake(&sotortpcb(last)->rop_refcnt);
562 		}
563 		/* keep a reference for last */
564 		refcnt_take(&rop->rop_refcnt);
565 		last = rop->rop_socket;
566 	}
567 	SRPL_LEAVE(&sr);
568 
569 	if (last) {
570 		s = solock(last);
571 		rtm_sendup(last, m, 0);
572 		sounlock(last, s);
573 		refcnt_rele_wake(&sotortpcb(last)->rop_refcnt);
574 	} else
575 		m_freem(m);
576 }
577 
578 int
579 rtm_sendup(struct socket *so, struct mbuf *m0, int more)
580 {
581 	struct rtpcb *rop = sotortpcb(so);
582 	struct mbuf *m;
583 
584 	soassertlocked(so);
585 
586 	if (more) {
587 		m = m_copym(m0, 0, M_COPYALL, M_NOWAIT);
588 		if (m == NULL)
589 			return (ENOMEM);
590 	} else
591 		m = m0;
592 
593 	if (sbspace(so, &so->so_rcv) < (2 * MSIZE) ||
594 	    sbappendaddr(so, &so->so_rcv, &route_src, m, NULL) == 0) {
595 		/* Flag socket as desync'ed and flush required */
596 		rop->rop_flags |= ROUTECB_FLAG_DESYNC | ROUTECB_FLAG_FLUSH;
597 		rtm_senddesync(so);
598 		m_freem(m);
599 		return (ENOBUFS);
600 	}
601 
602 	sorwakeup(so);
603 	return (0);
604 }
605 
606 struct rt_msghdr *
607 rtm_report(struct rtentry *rt, u_char type, int seq, int tableid)
608 {
609 	struct rt_msghdr	*rtm;
610 	struct rt_addrinfo	 info;
611 	struct sockaddr_rtlabel	 sa_rl;
612 	struct sockaddr_in6	 sa_mask;
613 #ifdef BFD
614 	struct sockaddr_bfd	 sa_bfd;
615 #endif
616 	struct ifnet		*ifp = NULL;
617 	int			 len;
618 
619 	bzero(&info, sizeof(info));
620 	info.rti_info[RTAX_DST] = rt_key(rt);
621 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
622 	info.rti_info[RTAX_NETMASK] = rt_plen2mask(rt, &sa_mask);
623 	info.rti_info[RTAX_LABEL] = rtlabel_id2sa(rt->rt_labelid, &sa_rl);
624 #ifdef BFD
625 	if (rt->rt_flags & RTF_BFD)
626 		info.rti_info[RTAX_BFD] = bfd2sa(rt, &sa_bfd);
627 #endif
628 #ifdef MPLS
629 	if (rt->rt_flags & RTF_MPLS) {
630 		struct sockaddr_mpls	 sa_mpls;
631 
632 		bzero(&sa_mpls, sizeof(sa_mpls));
633 		sa_mpls.smpls_family = AF_MPLS;
634 		sa_mpls.smpls_len = sizeof(sa_mpls);
635 		sa_mpls.smpls_label = ((struct rt_mpls *)
636 		    rt->rt_llinfo)->mpls_label;
637 		info.rti_info[RTAX_SRC] = (struct sockaddr *)&sa_mpls;
638 		info.rti_mpls = ((struct rt_mpls *)
639 		    rt->rt_llinfo)->mpls_operation;
640 	}
641 #endif
642 	ifp = if_get(rt->rt_ifidx);
643 	if (ifp != NULL) {
644 		info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
645 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
646 		if (ifp->if_flags & IFF_POINTOPOINT)
647 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
648 	}
649 	if_put(ifp);
650 	/* RTAX_GENMASK, RTAX_AUTHOR, RTAX_SRCMASK ignored */
651 
652 	/* build new route message */
653 	len = rtm_msg2(type, RTM_VERSION, &info, NULL, NULL);
654 	rtm = malloc(len, M_RTABLE, M_WAITOK | M_ZERO);
655 
656 	rtm_msg2(type, RTM_VERSION, &info, (caddr_t)rtm, NULL);
657 	rtm->rtm_type = type;
658 	rtm->rtm_index = rt->rt_ifidx;
659 	rtm->rtm_tableid = tableid;
660 	rtm->rtm_priority = rt->rt_priority & RTP_MASK;
661 	rtm->rtm_flags = rt->rt_flags;
662 	rtm->rtm_pid = curproc->p_p->ps_pid;
663 	rtm->rtm_seq = seq;
664 	rtm_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
665 	rtm->rtm_addrs = info.rti_addrs;
666 #ifdef MPLS
667 	rtm->rtm_mpls = info.rti_mpls;
668 #endif
669 	return rtm;
670 }
671 
672 int
673 route_output(struct mbuf *m, struct socket *so, struct sockaddr *dstaddr,
674     struct mbuf *control)
675 {
676 	struct rt_msghdr	*rtm = NULL;
677 	struct rtentry		*rt = NULL;
678 	struct rt_addrinfo	 info;
679 	int			 len, seq, error = 0;
680 	u_int			 tableid;
681 	u_int8_t		 prio;
682 	u_char			 vers, type;
683 
684 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
685 	    (m = m_pullup(m, sizeof(int32_t))) == 0))
686 		return (ENOBUFS);
687 	if ((m->m_flags & M_PKTHDR) == 0)
688 		panic("route_output");
689 	len = m->m_pkthdr.len;
690 	if (len < offsetof(struct rt_msghdr, rtm_hdrlen) + 1 ||
691 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
692 		error = EINVAL;
693 		goto fail;
694 	}
695 	vers = mtod(m, struct rt_msghdr *)->rtm_version;
696 	switch (vers) {
697 	case RTM_VERSION:
698 		if (len < sizeof(struct rt_msghdr)) {
699 			error = EINVAL;
700 			goto fail;
701 		}
702 		if (len > RTM_MAXSIZE) {
703 			error = EMSGSIZE;
704 			goto fail;
705 		}
706 		rtm = malloc(len, M_RTABLE, M_WAITOK);
707 		m_copydata(m, 0, len, (caddr_t)rtm);
708 		break;
709 	default:
710 		error = EPROTONOSUPPORT;
711 		goto fail;
712 	}
713 
714 	/* Verify that the caller is sending an appropriate message early */
715 	switch (rtm->rtm_type) {
716 	case RTM_ADD:
717 	case RTM_DELETE:
718 	case RTM_GET:
719 	case RTM_CHANGE:
720 	case RTM_PROPOSAL:
721 		break;
722 	default:
723 		error = EOPNOTSUPP;
724 		goto fail;
725 	}
726 	/*
727 	 * Verify that the header length is valid.
728 	 * All messages from userland start with a struct rt_msghdr.
729 	 */
730 	if (rtm->rtm_hdrlen == 0)	/* old client */
731 		rtm->rtm_hdrlen = sizeof(struct rt_msghdr);
732 	if (rtm->rtm_hdrlen < sizeof(struct rt_msghdr) ||
733 	    len < rtm->rtm_hdrlen) {
734 		error = EINVAL;
735 		goto fail;
736 	}
737 
738 	rtm->rtm_pid = curproc->p_p->ps_pid;
739 
740 	/*
741 	 * Verify that the caller has the appropriate privilege; RTM_GET
742 	 * is the only operation the non-superuser is allowed.
743 	 */
744 	if (rtm->rtm_type != RTM_GET && suser(curproc) != 0) {
745 		error = EACCES;
746 		goto fail;
747 	}
748 	tableid = rtm->rtm_tableid;
749 	if (!rtable_exists(tableid)) {
750 		if (rtm->rtm_type == RTM_ADD) {
751 			if ((error = rtable_add(tableid)) != 0)
752 				goto fail;
753 		} else {
754 			error = EINVAL;
755 			goto fail;
756 		}
757 	}
758 
759 
760 	/* Do not let userland play with kernel-only flags. */
761 	if ((rtm->rtm_flags & (RTF_LOCAL|RTF_BROADCAST)) != 0) {
762 		error = EINVAL;
763 		goto fail;
764 	}
765 
766 	/* make sure that kernel-only bits are not set */
767 	rtm->rtm_priority &= RTP_MASK;
768 	rtm->rtm_flags &= ~(RTF_DONE|RTF_CLONED|RTF_CACHED);
769 	rtm->rtm_fmask &= RTF_FMASK;
770 
771 	if (rtm->rtm_priority != 0) {
772 		if (rtm->rtm_priority > RTP_MAX ||
773 		    rtm->rtm_priority == RTP_LOCAL) {
774 			error = EINVAL;
775 			goto fail;
776 		}
777 		prio = rtm->rtm_priority;
778 	} else if (rtm->rtm_type != RTM_ADD)
779 		prio = RTP_ANY;
780 	else if (rtm->rtm_flags & RTF_STATIC)
781 		prio = 0;
782 	else
783 		prio = RTP_DEFAULT;
784 
785 	bzero(&info, sizeof(info));
786 	info.rti_addrs = rtm->rtm_addrs;
787 	if ((error = rtm_xaddrs(rtm->rtm_hdrlen + (caddr_t)rtm,
788 	    len + (caddr_t)rtm, &info)) != 0)
789 		goto fail;
790 	info.rti_flags = rtm->rtm_flags;
791 	if (rtm->rtm_type != RTM_PROPOSAL &&
792 	   (info.rti_info[RTAX_DST] == NULL ||
793 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
794 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
795 	    info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX) ||
796 	    info.rti_info[RTAX_GENMASK] != NULL)) {
797 		error = EINVAL;
798 		goto fail;
799 	}
800 #ifdef MPLS
801 	info.rti_mpls = rtm->rtm_mpls;
802 #endif
803 
804 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
805 	    info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
806 	    (info.rti_flags & RTF_CLONING) == 0) {
807 		info.rti_flags |= RTF_LLINFO;
808 	}
809 
810 	/*
811 	 * Validate RTM_PROPOSAL and pass it along or error out.
812 	 */
813 	if (rtm->rtm_type == RTM_PROPOSAL) {
814 		if (rtm_validate_proposal(&info) == -1) {
815 			error = EINVAL;
816 			goto fail;
817 		}
818 		/*
819 		 * If this is a solicitation proposal forward request to
820 		 * all interfaces. Most handlers will ignore it but at least
821 		 * umb(4) will send a response to this event.
822 		 */
823 		if (rtm->rtm_priority == RTP_PROPOSAL_SOLICIT) {
824 			struct ifnet *ifp;
825 			NET_LOCK();
826 			TAILQ_FOREACH(ifp, &ifnet, if_list) {
827 				ifp->if_rtrequest(ifp, RTM_PROPOSAL, NULL);
828 			}
829 			NET_UNLOCK();
830 		}
831 	} else {
832 		error = rtm_output(rtm, &rt, &info, prio, tableid);
833 		if (!error) {
834 			type = rtm->rtm_type;
835 			seq = rtm->rtm_seq;
836 			free(rtm, M_RTABLE, len);
837 			rtm = rtm_report(rt, type, seq, tableid);
838 			len = rtm->rtm_msglen;
839 		}
840 	}
841 
842 	rtfree(rt);
843 	if (error) {
844 		rtm->rtm_errno = error;
845 	} else {
846 		rtm->rtm_flags |= RTF_DONE;
847 	}
848 
849 	/*
850 	 * Check to see if we don't want our own messages.
851 	 */
852 	if (!(so->so_options & SO_USELOOPBACK)) {
853 		if (rtptable.rtp_count <= 1) {
854 			/* no other listener and no loopback of messages */
855 fail:
856 			free(rtm, M_RTABLE, len);
857 			m_freem(m);
858 			return (error);
859 		}
860 	}
861 	if (m_copyback(m, 0, len, rtm, M_NOWAIT)) {
862 		m_freem(m);
863 		m = NULL;
864 	} else if (m->m_pkthdr.len > len)
865 		m_adj(m, len - m->m_pkthdr.len);
866 	free(rtm, M_RTABLE, len);
867 	if (m)
868 		route_input(m, so, info.rti_info[RTAX_DST] ?
869 		    info.rti_info[RTAX_DST]->sa_family : AF_UNSPEC);
870 
871 	return (error);
872 }
873 
874 int
875 rtm_output(struct rt_msghdr *rtm, struct rtentry **prt,
876     struct rt_addrinfo *info, uint8_t prio, unsigned int tableid)
877 {
878 	struct rtentry		*rt = *prt;
879 	struct ifnet		*ifp = NULL;
880 	int			 plen, newgate = 0, error = 0;
881 
882 	switch (rtm->rtm_type) {
883 	case RTM_ADD:
884 		if (info->rti_info[RTAX_GATEWAY] == NULL) {
885 			error = EINVAL;
886 			break;
887 		}
888 
889 		rt = rtable_match(tableid, info->rti_info[RTAX_DST], NULL);
890 		if ((error = route_arp_conflict(rt, info))) {
891 			rtfree(rt);
892 			rt = NULL;
893 			break;
894 		}
895 
896 		/*
897 		 * We cannot go through a delete/create/insert cycle for
898 		 * cached route because this can lead to races in the
899 		 * receive path.  Instead we update the L2 cache.
900 		 */
901 		if ((rt != NULL) && ISSET(rt->rt_flags, RTF_CACHED))
902 			goto change;
903 
904 		rtfree(rt);
905 		rt = NULL;
906 
907 		NET_LOCK();
908 		if ((error = rtm_getifa(info, tableid)) != 0) {
909 			NET_UNLOCK();
910 			break;
911 		}
912 		error = rtrequest(RTM_ADD, info, prio, &rt, tableid);
913 		NET_UNLOCK();
914 		if (error == 0)
915 			rtm_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
916 			    &rt->rt_rmx);
917 		break;
918 	case RTM_DELETE:
919 		rt = rtable_lookup(tableid, info->rti_info[RTAX_DST],
920 		    info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY],
921 		    prio);
922 		if (rt == NULL) {
923 			error = ESRCH;
924 			break;
925 		}
926 
927 		/*
928 		 * If we got multipath routes, we require users to specify
929 		 * a matching gateway.
930 		 */
931 		if (ISSET(rt->rt_flags, RTF_MPATH) &&
932 		    info->rti_info[RTAX_GATEWAY] == NULL) {
933 			error = ESRCH;
934 			break;
935 		}
936 
937 		/* Detaching an interface requires the KERNEL_LOCK(). */
938 		ifp = if_get(rt->rt_ifidx);
939 		KASSERT(ifp != NULL);
940 
941 		/*
942 		 * Invalidate the cache of automagically created and
943 		 * referenced L2 entries to make sure that ``rt_gwroute''
944 		 * pointer stays valid for other CPUs.
945 		 */
946 		if ((ISSET(rt->rt_flags, RTF_CACHED))) {
947 			NET_LOCK();
948 			ifp->if_rtrequest(ifp, RTM_INVALIDATE, rt);
949 			/* Reset the MTU of the gateway route. */
950 			rtable_walk(tableid, rt_key(rt)->sa_family, NULL,
951 			    route_cleargateway, rt);
952 			NET_UNLOCK();
953 			if_put(ifp);
954 			break;
955 		}
956 
957 		/*
958 		 * Make sure that local routes are only modified by the
959 		 * kernel.
960 		 */
961 		if (ISSET(rt->rt_flags, RTF_LOCAL|RTF_BROADCAST)) {
962 			if_put(ifp);
963 			error = EINVAL;
964 			break;
965 		}
966 
967 		rtfree(rt);
968 		rt = NULL;
969 
970 		NET_LOCK();
971 		error = rtrequest_delete(info, prio, ifp, &rt, tableid);
972 		NET_UNLOCK();
973 		if_put(ifp);
974 		break;
975 	case RTM_CHANGE:
976 		rt = rtable_lookup(tableid, info->rti_info[RTAX_DST],
977 		    info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY],
978 		    prio);
979 		/*
980 		 * If we got multipath routes, we require users to specify
981 		 * a matching gateway.
982 		 */
983 		if ((rt != NULL) && ISSET(rt->rt_flags, RTF_MPATH) &&
984 		    (info->rti_info[RTAX_GATEWAY] == NULL)) {
985 			rtfree(rt);
986 			rt = NULL;
987 		}
988 		/*
989 		 * If RTAX_GATEWAY is the argument we're trying to
990 		 * change, try to find a compatible route.
991 		 */
992 		if ((rt == NULL) && (info->rti_info[RTAX_GATEWAY] != NULL)) {
993 			rt = rtable_lookup(tableid, info->rti_info[RTAX_DST],
994 			    info->rti_info[RTAX_NETMASK], NULL, prio);
995 			/* Ensure we don't pick a multipath one. */
996 			if ((rt != NULL) && ISSET(rt->rt_flags, RTF_MPATH)) {
997 				rtfree(rt);
998 				rt = NULL;
999 			}
1000 		}
1001 
1002 		if (rt == NULL) {
1003 			error = ESRCH;
1004 			break;
1005 		}
1006 
1007 		/*
1008 		 * Make sure that local routes are only modified by the
1009 		 * kernel.
1010 		 */
1011 		if (ISSET(rt->rt_flags, RTF_LOCAL|RTF_BROADCAST)) {
1012 			error = EINVAL;
1013 			break;
1014 		}
1015 
1016 		/*
1017 		 * RTM_CHANGE needs a perfect match.
1018 		 */
1019 		plen = rtable_satoplen(info->rti_info[RTAX_DST]->sa_family,
1020 		    info->rti_info[RTAX_NETMASK]);
1021 		if (rt_plen(rt) != plen) {
1022 			error = ESRCH;
1023 			break;
1024 		}
1025 
1026 		if (info->rti_info[RTAX_GATEWAY] != NULL)
1027 			if (rt->rt_gateway == NULL ||
1028 			    bcmp(rt->rt_gateway,
1029 			    info->rti_info[RTAX_GATEWAY],
1030 			    info->rti_info[RTAX_GATEWAY]->sa_len)) {
1031 				newgate = 1;
1032 			}
1033 		/*
1034 		 * Check reachable gateway before changing the route.
1035 		 * New gateway could require new ifaddr, ifp;
1036 		 * flags may also be different; ifp may be specified
1037 		 * by ll sockaddr when protocol address is ambiguous.
1038 		 */
1039 		if (newgate || info->rti_info[RTAX_IFP] != NULL ||
1040 		    info->rti_info[RTAX_IFA] != NULL) {
1041 			struct ifaddr	*ifa = NULL;
1042 
1043 			NET_LOCK();
1044 			if ((error = rtm_getifa(info, tableid)) != 0) {
1045 				NET_UNLOCK();
1046 				break;
1047 			}
1048 			ifa = info->rti_ifa;
1049 			if (rt->rt_ifa != ifa) {
1050 				ifp = if_get(rt->rt_ifidx);
1051 				KASSERT(ifp != NULL);
1052 				ifp->if_rtrequest(ifp, RTM_DELETE, rt);
1053 				ifafree(rt->rt_ifa);
1054 				if_put(ifp);
1055 
1056 				ifa->ifa_refcnt++;
1057 				rt->rt_ifa = ifa;
1058 				rt->rt_ifidx = ifa->ifa_ifp->if_index;
1059 				/* recheck link state after ifp change */
1060 				rt_if_linkstate_change(rt, ifa->ifa_ifp,
1061 				    tableid);
1062 			}
1063 			NET_UNLOCK();
1064 		}
1065 change:
1066 		if (info->rti_info[RTAX_GATEWAY] != NULL) {
1067 			/* When updating the gateway, make sure it is valid. */
1068 			if (!newgate && rt->rt_gateway->sa_family !=
1069 			    info->rti_info[RTAX_GATEWAY]->sa_family) {
1070 				error = EINVAL;
1071 				break;
1072 			}
1073 
1074 			NET_LOCK();
1075 			error = rt_setgate(rt,
1076 			    info->rti_info[RTAX_GATEWAY], tableid);
1077 			NET_UNLOCK();
1078 			if (error)
1079 				break;
1080 		}
1081 #ifdef MPLS
1082 		if (rtm->rtm_flags & RTF_MPLS) {
1083 			NET_LOCK();
1084 			error = rt_mpls_set(rt,
1085 			    info->rti_info[RTAX_SRC], info->rti_mpls);
1086 			NET_UNLOCK();
1087 			if (error)
1088 				break;
1089 		} else if (newgate || (rtm->rtm_fmask & RTF_MPLS)) {
1090 			NET_LOCK();
1091 			/* if gateway changed remove MPLS information */
1092 			rt_mpls_clear(rt);
1093 			NET_UNLOCK();
1094 		}
1095 #endif
1096 
1097 #ifdef BFD
1098 		if (ISSET(rtm->rtm_flags, RTF_BFD)) {
1099 			if ((error = bfdset(rt)))
1100 				break;
1101 		} else if (!ISSET(rtm->rtm_flags, RTF_BFD) &&
1102 		    ISSET(rtm->rtm_fmask, RTF_BFD)) {
1103 			bfdclear(rt);
1104 		}
1105 #endif
1106 
1107 		NET_LOCK();
1108 		/* Hack to allow some flags to be toggled */
1109 		if (rtm->rtm_fmask) {
1110 			/* MPLS flag it is set by rt_mpls_set() */
1111 			rtm->rtm_fmask &= ~RTF_MPLS;
1112 			rtm->rtm_flags &= ~RTF_MPLS;
1113 			rt->rt_flags =
1114 			    (rt->rt_flags & ~rtm->rtm_fmask) |
1115 			    (rtm->rtm_flags & rtm->rtm_fmask);
1116 		}
1117 		rtm_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, &rt->rt_rmx);
1118 
1119 		ifp = if_get(rt->rt_ifidx);
1120 		KASSERT(ifp != NULL);
1121 		ifp->if_rtrequest(ifp, RTM_ADD, rt);
1122 		if_put(ifp);
1123 
1124 		if (info->rti_info[RTAX_LABEL] != NULL) {
1125 			char *rtlabel = ((struct sockaddr_rtlabel *)
1126 			    info->rti_info[RTAX_LABEL])->sr_label;
1127 			rtlabel_unref(rt->rt_labelid);
1128 			rt->rt_labelid = rtlabel_name2id(rtlabel);
1129 		}
1130 		if_group_routechange(info->rti_info[RTAX_DST],
1131 		    info->rti_info[RTAX_NETMASK]);
1132 		rt->rt_locks &= ~(rtm->rtm_inits);
1133 		rt->rt_locks |= (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
1134 		NET_UNLOCK();
1135 		break;
1136 	case RTM_GET:
1137 		rt = rtable_lookup(tableid, info->rti_info[RTAX_DST],
1138 		    info->rti_info[RTAX_NETMASK], info->rti_info[RTAX_GATEWAY],
1139 		    prio);
1140 		if (rt == NULL)
1141 			error = ESRCH;
1142 		break;
1143 	}
1144 
1145 	*prt = rt;
1146 	return (error);
1147 }
1148 
1149 struct ifaddr *
1150 ifa_ifwithroute(int flags, struct sockaddr *dst, struct sockaddr *gateway,
1151     unsigned int rtableid)
1152 {
1153 	struct ifaddr	*ifa;
1154 
1155 	if ((flags & RTF_GATEWAY) == 0) {
1156 		/*
1157 		 * If we are adding a route to an interface,
1158 		 * and the interface is a pt to pt link
1159 		 * we should search for the destination
1160 		 * as our clue to the interface.  Otherwise
1161 		 * we can use the local address.
1162 		 */
1163 		ifa = NULL;
1164 		if (flags & RTF_HOST)
1165 			ifa = ifa_ifwithdstaddr(dst, rtableid);
1166 		if (ifa == NULL)
1167 			ifa = ifa_ifwithaddr(gateway, rtableid);
1168 	} else {
1169 		/*
1170 		 * If we are adding a route to a remote net
1171 		 * or host, the gateway may still be on the
1172 		 * other end of a pt to pt link.
1173 		 */
1174 		ifa = ifa_ifwithdstaddr(gateway, rtableid);
1175 	}
1176 	if (ifa == NULL) {
1177 		if (gateway->sa_family == AF_LINK) {
1178 			struct sockaddr_dl *sdl = satosdl(gateway);
1179 			struct ifnet *ifp = if_get(sdl->sdl_index);
1180 
1181 			if (ifp != NULL)
1182 				ifa = ifaof_ifpforaddr(dst, ifp);
1183 			if_put(ifp);
1184 		} else {
1185 			struct rtentry *rt;
1186 
1187 			rt = rtalloc(gateway, RT_RESOLVE, rtable_l2(rtableid));
1188 			if (rt != NULL)
1189 				ifa = rt->rt_ifa;
1190 			rtfree(rt);
1191 		}
1192 	}
1193 	if (ifa == NULL)
1194 		return (NULL);
1195 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
1196 		struct ifaddr	*oifa = ifa;
1197 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
1198 		if (ifa == NULL)
1199 			ifa = oifa;
1200 	}
1201 	return (ifa);
1202 }
1203 
1204 int
1205 rtm_getifa(struct rt_addrinfo *info, unsigned int rtid)
1206 {
1207 	struct ifnet	*ifp = NULL;
1208 
1209 	/*
1210 	 * The "returned" `ifa' is guaranteed to be alive only if
1211 	 * the NET_LOCK() is held.
1212 	 */
1213 	NET_ASSERT_LOCKED();
1214 
1215 	/*
1216 	 * ifp may be specified by sockaddr_dl when protocol address
1217 	 * is ambiguous
1218 	 */
1219 	if (info->rti_info[RTAX_IFP] != NULL) {
1220 		struct sockaddr_dl *sdl;
1221 
1222 		sdl = satosdl(info->rti_info[RTAX_IFP]);
1223 		ifp = if_get(sdl->sdl_index);
1224 	}
1225 
1226 #ifdef IPSEC
1227 	/*
1228 	 * If the destination is a PF_KEY address, we'll look
1229 	 * for the existence of a encap interface number or address
1230 	 * in the options list of the gateway. By default, we'll return
1231 	 * enc0.
1232 	 */
1233 	if (info->rti_info[RTAX_DST] &&
1234 	    info->rti_info[RTAX_DST]->sa_family == PF_KEY)
1235 		info->rti_ifa = enc_getifa(rtid, 0);
1236 #endif
1237 
1238 	if (info->rti_ifa == NULL && info->rti_info[RTAX_IFA] != NULL)
1239 		info->rti_ifa = ifa_ifwithaddr(info->rti_info[RTAX_IFA], rtid);
1240 
1241 	if (info->rti_ifa == NULL) {
1242 		struct sockaddr	*sa;
1243 
1244 		if ((sa = info->rti_info[RTAX_IFA]) == NULL)
1245 			if ((sa = info->rti_info[RTAX_GATEWAY]) == NULL)
1246 				sa = info->rti_info[RTAX_DST];
1247 
1248 		if (sa != NULL && ifp != NULL)
1249 			info->rti_ifa = ifaof_ifpforaddr(sa, ifp);
1250 		else if (info->rti_info[RTAX_DST] != NULL &&
1251 		    info->rti_info[RTAX_GATEWAY] != NULL)
1252 			info->rti_ifa = ifa_ifwithroute(info->rti_flags,
1253 			    info->rti_info[RTAX_DST],
1254 			    info->rti_info[RTAX_GATEWAY],
1255 			    rtid);
1256 		else if (sa != NULL)
1257 			info->rti_ifa = ifa_ifwithroute(info->rti_flags,
1258 			    sa, sa, rtid);
1259 	}
1260 
1261 	if_put(ifp);
1262 
1263 	if (info->rti_ifa == NULL)
1264 		return (ENETUNREACH);
1265 
1266 	return (0);
1267 }
1268 
1269 int
1270 route_cleargateway(struct rtentry *rt, void *arg, unsigned int rtableid)
1271 {
1272 	struct rtentry *nhrt = arg;
1273 
1274 	if (ISSET(rt->rt_flags, RTF_GATEWAY) && rt->rt_gwroute == nhrt &&
1275 	    !ISSET(rt->rt_locks, RTV_MTU))
1276 		rt->rt_mtu = 0;
1277 
1278 	return (0);
1279 }
1280 
1281 /*
1282  * Check if the user request to insert an ARP entry does not conflict
1283  * with existing ones.
1284  *
1285  * Only two entries are allowed for a given IP address: a private one
1286  * (priv) and a public one (pub).
1287  */
1288 int
1289 route_arp_conflict(struct rtentry *rt, struct rt_addrinfo *info)
1290 {
1291 	int		 proxy = (info->rti_flags & RTF_ANNOUNCE);
1292 
1293 	if ((info->rti_flags & RTF_LLINFO) == 0 ||
1294 	    (info->rti_info[RTAX_DST]->sa_family != AF_INET))
1295 		return (0);
1296 
1297 	if (rt == NULL || !ISSET(rt->rt_flags, RTF_LLINFO))
1298 		return (0);
1299 
1300 	/* If the entry is cached, it can be updated. */
1301 	if (ISSET(rt->rt_flags, RTF_CACHED))
1302 		return (0);
1303 
1304 	/*
1305 	 * Same destination, not cached and both "priv" or "pub" conflict.
1306 	 * If a second entry exists, it always conflict.
1307 	 */
1308 	if ((ISSET(rt->rt_flags, RTF_ANNOUNCE) == proxy) ||
1309 	    ISSET(rt->rt_flags, RTF_MPATH))
1310 		return (EEXIST);
1311 
1312 	/* No conflict but an entry exist so we need to force mpath. */
1313 	info->rti_flags |= RTF_MPATH;
1314 	return (0);
1315 }
1316 
1317 void
1318 rtm_setmetrics(u_long which, const struct rt_metrics *in,
1319     struct rt_kmetrics *out)
1320 {
1321 	int64_t expire;
1322 
1323 	if (which & RTV_MTU)
1324 		out->rmx_mtu = in->rmx_mtu;
1325 	if (which & RTV_EXPIRE) {
1326 		expire = in->rmx_expire;
1327 		if (expire != 0) {
1328 			expire -= gettime();
1329 			expire += getuptime();
1330 		}
1331 
1332 		out->rmx_expire = expire;
1333 	}
1334 }
1335 
1336 void
1337 rtm_getmetrics(const struct rt_kmetrics *in, struct rt_metrics *out)
1338 {
1339 	int64_t expire;
1340 
1341 	expire = in->rmx_expire;
1342 	if (expire != 0) {
1343 		expire -= getuptime();
1344 		expire += gettime();
1345 	}
1346 
1347 	bzero(out, sizeof(*out));
1348 	out->rmx_locks = in->rmx_locks;
1349 	out->rmx_mtu = in->rmx_mtu;
1350 	out->rmx_expire = expire;
1351 	out->rmx_pksent = in->rmx_pksent;
1352 }
1353 
1354 #define ROUNDUP(a) \
1355 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
1356 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
1357 
1358 int
1359 rtm_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1360 {
1361 	struct sockaddr	*sa;
1362 	int		 i;
1363 
1364 	/*
1365 	 * Parse address bits, split address storage in chunks, and
1366 	 * set info pointers.  Use sa_len for traversing the memory
1367 	 * and check that we stay within in the limit.
1368 	 */
1369 	bzero(rtinfo->rti_info, sizeof(rtinfo->rti_info));
1370 	for (i = 0; i < sizeof(rtinfo->rti_addrs) * 8; i++) {
1371 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1372 			continue;
1373 		if (i >= RTAX_MAX || cp + sizeof(socklen_t) > cplim)
1374 			return (EINVAL);
1375 		sa = (struct sockaddr *)cp;
1376 		if (cp + sa->sa_len > cplim)
1377 			return (EINVAL);
1378 		rtinfo->rti_info[i] = sa;
1379 		ADVANCE(cp, sa);
1380 	}
1381 	/*
1382 	 * Check that the address family is suitable for the route address
1383 	 * type.  Check that each address has a size that fits its family
1384 	 * and its length is within the size.  Strings within addresses must
1385 	 * be NUL terminated.
1386 	 */
1387 	for (i = 0; i < RTAX_MAX; i++) {
1388 		size_t len, maxlen, size;
1389 
1390 		sa = rtinfo->rti_info[i];
1391 		if (sa == NULL)
1392 			continue;
1393 		maxlen = size = 0;
1394 		switch (i) {
1395 		case RTAX_DST:
1396 		case RTAX_GATEWAY:
1397 		case RTAX_SRC:
1398 			switch (sa->sa_family) {
1399 			case AF_INET:
1400 				size = sizeof(struct sockaddr_in);
1401 				break;
1402 			case AF_LINK:
1403 				size = sizeof(struct sockaddr_dl);
1404 				break;
1405 #ifdef INET6
1406 			case AF_INET6:
1407 				size = sizeof(struct sockaddr_in6);
1408 				break;
1409 #endif
1410 #ifdef MPLS
1411 			case AF_MPLS:
1412 				size = sizeof(struct sockaddr_mpls);
1413 				break;
1414 #endif
1415 			}
1416 			break;
1417 		case RTAX_IFP:
1418 			if (sa->sa_family != AF_LINK)
1419 				return (EAFNOSUPPORT);
1420 			/*
1421 			 * XXX Should be sizeof(struct sockaddr_dl), but
1422 			 * route(8) has a bug and provides less memory.
1423 			 * arp(8) has another bug and uses sizeof pointer.
1424 			 */
1425 			size = 4;
1426 			break;
1427 		case RTAX_IFA:
1428 			switch (sa->sa_family) {
1429 			case AF_INET:
1430 				size = sizeof(struct sockaddr_in);
1431 				break;
1432 #ifdef INET6
1433 			case AF_INET6:
1434 				size = sizeof(struct sockaddr_in6);
1435 				break;
1436 #endif
1437 			default:
1438 				return (EAFNOSUPPORT);
1439 			}
1440 			break;
1441 		case RTAX_LABEL:
1442 			sa->sa_family = AF_UNSPEC;
1443 			maxlen = RTLABEL_LEN;
1444 			size = sizeof(struct sockaddr_rtlabel);
1445 			break;
1446 #ifdef BFD
1447 		case RTAX_BFD:
1448 			sa->sa_family = AF_UNSPEC;
1449 			size = sizeof(struct sockaddr_bfd);
1450 			break;
1451 #endif
1452 		case RTAX_DNS:
1453 			/* more validation in rtm_validate_proposal */
1454 			if (sa->sa_len > sizeof(struct sockaddr_rtdns))
1455 				return (EINVAL);
1456 			if (sa->sa_len < offsetof(struct sockaddr_rtdns,
1457 			    sr_dns))
1458 				return (EINVAL);
1459 			switch (sa->sa_family) {
1460 			case AF_INET:
1461 #ifdef INET6
1462 			case AF_INET6:
1463 #endif
1464 				break;
1465 			default:
1466 				return (EAFNOSUPPORT);
1467 			}
1468 			break;
1469 		case RTAX_STATIC:
1470 			sa->sa_family = AF_UNSPEC;
1471 			maxlen = RTSTATIC_LEN;
1472 			size = sizeof(struct sockaddr_rtstatic);
1473 			break;
1474 		case RTAX_SEARCH:
1475 			sa->sa_family = AF_UNSPEC;
1476 			maxlen = RTSEARCH_LEN;
1477 			size = sizeof(struct sockaddr_rtsearch);
1478 			break;
1479 		}
1480 		if (size) {
1481 			/* memory for the full struct must be provided */
1482 			if (sa->sa_len < size)
1483 				return (EINVAL);
1484 		}
1485 		if (maxlen) {
1486 			/* this should not happen */
1487 			if (2 + maxlen > size)
1488 				return (EINVAL);
1489 			/* strings must be NUL terminated within the struct */
1490 			len = strnlen(sa->sa_data, maxlen);
1491 			if (len >= maxlen || 2 + len >= sa->sa_len)
1492 				return (EINVAL);
1493 			break;
1494 		}
1495 	}
1496 	return (0);
1497 }
1498 
1499 struct mbuf *
1500 rtm_msg1(int type, struct rt_addrinfo *rtinfo)
1501 {
1502 	struct rt_msghdr	*rtm;
1503 	struct mbuf		*m;
1504 	int			 i;
1505 	struct sockaddr		*sa;
1506 	int			 len, dlen, hlen;
1507 
1508 	switch (type) {
1509 	case RTM_DELADDR:
1510 	case RTM_NEWADDR:
1511 		len = sizeof(struct ifa_msghdr);
1512 		break;
1513 	case RTM_IFINFO:
1514 		len = sizeof(struct if_msghdr);
1515 		break;
1516 	case RTM_IFANNOUNCE:
1517 		len = sizeof(struct if_announcemsghdr);
1518 		break;
1519 #ifdef BFD
1520 	case RTM_BFD:
1521 		len = sizeof(struct bfd_msghdr);
1522 		break;
1523 #endif
1524 	case RTM_80211INFO:
1525 		len = sizeof(struct if_ieee80211_msghdr);
1526 		break;
1527 	default:
1528 		len = sizeof(struct rt_msghdr);
1529 		break;
1530 	}
1531 	if (len > MCLBYTES)
1532 		panic("rtm_msg1");
1533 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1534 	if (m && len > MHLEN) {
1535 		MCLGET(m, M_DONTWAIT);
1536 		if ((m->m_flags & M_EXT) == 0) {
1537 			m_free(m);
1538 			m = NULL;
1539 		}
1540 	}
1541 	if (m == NULL)
1542 		return (m);
1543 	m->m_pkthdr.len = m->m_len = hlen = len;
1544 	m->m_pkthdr.ph_ifidx = 0;
1545 	rtm = mtod(m, struct rt_msghdr *);
1546 	bzero(rtm, len);
1547 	for (i = 0; i < RTAX_MAX; i++) {
1548 		if (rtinfo == NULL || (sa = rtinfo->rti_info[i]) == NULL)
1549 			continue;
1550 		rtinfo->rti_addrs |= (1 << i);
1551 		dlen = ROUNDUP(sa->sa_len);
1552 		if (m_copyback(m, len, dlen, sa, M_NOWAIT)) {
1553 			m_freem(m);
1554 			return (NULL);
1555 		}
1556 		len += dlen;
1557 	}
1558 	rtm->rtm_msglen = len;
1559 	rtm->rtm_hdrlen = hlen;
1560 	rtm->rtm_version = RTM_VERSION;
1561 	rtm->rtm_type = type;
1562 	return (m);
1563 }
1564 
1565 int
1566 rtm_msg2(int type, int vers, struct rt_addrinfo *rtinfo, caddr_t cp,
1567     struct walkarg *w)
1568 {
1569 	int		i;
1570 	int		len, dlen, hlen, second_time = 0;
1571 	caddr_t		cp0;
1572 
1573 	rtinfo->rti_addrs = 0;
1574 again:
1575 	switch (type) {
1576 	case RTM_DELADDR:
1577 	case RTM_NEWADDR:
1578 		len = sizeof(struct ifa_msghdr);
1579 		break;
1580 	case RTM_IFINFO:
1581 		len = sizeof(struct if_msghdr);
1582 		break;
1583 	default:
1584 		len = sizeof(struct rt_msghdr);
1585 		break;
1586 	}
1587 	hlen = len;
1588 	if ((cp0 = cp) != NULL)
1589 		cp += len;
1590 	for (i = 0; i < RTAX_MAX; i++) {
1591 		struct sockaddr *sa;
1592 
1593 		if ((sa = rtinfo->rti_info[i]) == NULL)
1594 			continue;
1595 		rtinfo->rti_addrs |= (1 << i);
1596 		dlen = ROUNDUP(sa->sa_len);
1597 		if (cp) {
1598 			bcopy(sa, cp, (size_t)dlen);
1599 			cp += dlen;
1600 		}
1601 		len += dlen;
1602 	}
1603 	/* align message length to the next natural boundary */
1604 	len = ALIGN(len);
1605 	if (cp == 0 && w != NULL && !second_time) {
1606 		w->w_needed += len;
1607 		if (w->w_needed <= 0 && w->w_where) {
1608 			if (w->w_tmemsize < len) {
1609 				free(w->w_tmem, M_RTABLE, w->w_tmemsize);
1610 				w->w_tmem = malloc(len, M_RTABLE,
1611 				    M_NOWAIT | M_ZERO);
1612 				if (w->w_tmem)
1613 					w->w_tmemsize = len;
1614 			}
1615 			if (w->w_tmem) {
1616 				cp = w->w_tmem;
1617 				second_time = 1;
1618 				goto again;
1619 			} else
1620 				w->w_where = 0;
1621 		}
1622 	}
1623 	if (cp && w)		/* clear the message header */
1624 		bzero(cp0, hlen);
1625 
1626 	if (cp) {
1627 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1628 
1629 		rtm->rtm_version = RTM_VERSION;
1630 		rtm->rtm_type = type;
1631 		rtm->rtm_msglen = len;
1632 		rtm->rtm_hdrlen = hlen;
1633 	}
1634 	return (len);
1635 }
1636 
1637 void
1638 rtm_send(struct rtentry *rt, int cmd, int error, unsigned int rtableid)
1639 {
1640 	struct rt_addrinfo	 info;
1641 	struct ifnet		*ifp;
1642 	struct sockaddr_rtlabel	 sa_rl;
1643 	struct sockaddr_in6	 sa_mask;
1644 
1645 	memset(&info, 0, sizeof(info));
1646 	info.rti_info[RTAX_DST] = rt_key(rt);
1647 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1648 	if (!ISSET(rt->rt_flags, RTF_HOST))
1649 		info.rti_info[RTAX_NETMASK] = rt_plen2mask(rt, &sa_mask);
1650 	info.rti_info[RTAX_LABEL] = rtlabel_id2sa(rt->rt_labelid, &sa_rl);
1651 	ifp = if_get(rt->rt_ifidx);
1652 	if (ifp != NULL) {
1653 		info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
1654 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1655 	}
1656 
1657 	rtm_miss(cmd, &info, rt->rt_flags, rt->rt_priority, rt->rt_ifidx, error,
1658 	    rtableid);
1659 	if_put(ifp);
1660 }
1661 
1662 /*
1663  * This routine is called to generate a message from the routing
1664  * socket indicating that a redirect has occurred, a routing lookup
1665  * has failed, or that a protocol has detected timeouts to a particular
1666  * destination.
1667  */
1668 void
1669 rtm_miss(int type, struct rt_addrinfo *rtinfo, int flags, uint8_t prio,
1670     u_int ifidx, int error, u_int tableid)
1671 {
1672 	struct rt_msghdr	*rtm;
1673 	struct mbuf		*m;
1674 	struct sockaddr		*sa = rtinfo->rti_info[RTAX_DST];
1675 
1676 	if (rtptable.rtp_count == 0)
1677 		return;
1678 	m = rtm_msg1(type, rtinfo);
1679 	if (m == NULL)
1680 		return;
1681 	rtm = mtod(m, struct rt_msghdr *);
1682 	rtm->rtm_flags = RTF_DONE | flags;
1683 	rtm->rtm_priority = prio;
1684 	rtm->rtm_errno = error;
1685 	rtm->rtm_tableid = tableid;
1686 	rtm->rtm_addrs = rtinfo->rti_addrs;
1687 	rtm->rtm_index = ifidx;
1688 	route_input(m, NULL, sa ? sa->sa_family : AF_UNSPEC);
1689 }
1690 
1691 /*
1692  * This routine is called to generate a message from the routing
1693  * socket indicating that the status of a network interface has changed.
1694  */
1695 void
1696 rtm_ifchg(struct ifnet *ifp)
1697 {
1698 	struct if_msghdr	*ifm;
1699 	struct mbuf		*m;
1700 
1701 	if (rtptable.rtp_count == 0)
1702 		return;
1703 	m = rtm_msg1(RTM_IFINFO, NULL);
1704 	if (m == NULL)
1705 		return;
1706 	ifm = mtod(m, struct if_msghdr *);
1707 	ifm->ifm_index = ifp->if_index;
1708 	ifm->ifm_tableid = ifp->if_rdomain;
1709 	ifm->ifm_flags = ifp->if_flags;
1710 	ifm->ifm_xflags = ifp->if_xflags;
1711 	if_getdata(ifp, &ifm->ifm_data);
1712 	ifm->ifm_addrs = 0;
1713 	route_input(m, NULL, AF_UNSPEC);
1714 }
1715 
1716 /*
1717  * This is called to generate messages from the routing socket
1718  * indicating a network interface has had addresses associated with it.
1719  * if we ever reverse the logic and replace messages TO the routing
1720  * socket indicate a request to configure interfaces, then it will
1721  * be unnecessary as the routing socket will automatically generate
1722  * copies of it.
1723  */
1724 void
1725 rtm_addr(int cmd, struct ifaddr *ifa)
1726 {
1727 	struct ifnet		*ifp = ifa->ifa_ifp;
1728 	struct mbuf		*m;
1729 	struct rt_addrinfo	 info;
1730 	struct ifa_msghdr	*ifam;
1731 
1732 	if (rtptable.rtp_count == 0)
1733 		return;
1734 
1735 	memset(&info, 0, sizeof(info));
1736 	info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1737 	info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
1738 	info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1739 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1740 	if ((m = rtm_msg1(cmd, &info)) == NULL)
1741 		return;
1742 	ifam = mtod(m, struct ifa_msghdr *);
1743 	ifam->ifam_index = ifp->if_index;
1744 	ifam->ifam_metric = ifa->ifa_metric;
1745 	ifam->ifam_flags = ifa->ifa_flags;
1746 	ifam->ifam_addrs = info.rti_addrs;
1747 	ifam->ifam_tableid = ifp->if_rdomain;
1748 
1749 	route_input(m, NULL,
1750 	    ifa->ifa_addr ? ifa->ifa_addr->sa_family : AF_UNSPEC);
1751 }
1752 
1753 /*
1754  * This is called to generate routing socket messages indicating
1755  * network interface arrival and departure.
1756  */
1757 void
1758 rtm_ifannounce(struct ifnet *ifp, int what)
1759 {
1760 	struct if_announcemsghdr	*ifan;
1761 	struct mbuf			*m;
1762 
1763 	if (rtptable.rtp_count == 0)
1764 		return;
1765 	m = rtm_msg1(RTM_IFANNOUNCE, NULL);
1766 	if (m == NULL)
1767 		return;
1768 	ifan = mtod(m, struct if_announcemsghdr *);
1769 	ifan->ifan_index = ifp->if_index;
1770 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof(ifan->ifan_name));
1771 	ifan->ifan_what = what;
1772 	route_input(m, NULL, AF_UNSPEC);
1773 }
1774 
1775 #ifdef BFD
1776 /*
1777  * This is used to generate routing socket messages indicating
1778  * the state of a BFD session.
1779  */
1780 void
1781 rtm_bfd(struct bfd_config *bfd)
1782 {
1783 	struct bfd_msghdr	*bfdm;
1784 	struct sockaddr_bfd	 sa_bfd;
1785 	struct mbuf		*m;
1786 	struct rt_addrinfo	 info;
1787 
1788 	if (rtptable.rtp_count == 0)
1789 		return;
1790 	memset(&info, 0, sizeof(info));
1791 	info.rti_info[RTAX_DST] = rt_key(bfd->bc_rt);
1792 	info.rti_info[RTAX_IFA] = bfd->bc_rt->rt_ifa->ifa_addr;
1793 
1794 	m = rtm_msg1(RTM_BFD, &info);
1795 	if (m == NULL)
1796 		return;
1797 	bfdm = mtod(m, struct bfd_msghdr *);
1798 	bfdm->bm_addrs = info.rti_addrs;
1799 
1800 	bfd2sa(bfd->bc_rt, &sa_bfd);
1801 	memcpy(&bfdm->bm_sa, &sa_bfd, sizeof(sa_bfd));
1802 
1803 	route_input(m, NULL, info.rti_info[RTAX_DST]->sa_family);
1804 }
1805 #endif /* BFD */
1806 
1807 /*
1808  * This is used to generate routing socket messages indicating
1809  * the state of an ieee80211 interface.
1810  */
1811 void
1812 rtm_80211info(struct ifnet *ifp, struct if_ieee80211_data *ifie)
1813 {
1814 	struct if_ieee80211_msghdr	*ifim;
1815 	struct mbuf			*m;
1816 
1817 	if (rtptable.rtp_count == 0)
1818 		return;
1819 	m = rtm_msg1(RTM_80211INFO, NULL);
1820 	if (m == NULL)
1821 		return;
1822 	ifim = mtod(m, struct if_ieee80211_msghdr *);
1823 	ifim->ifim_index = ifp->if_index;
1824 	ifim->ifim_tableid = ifp->if_rdomain;
1825 
1826 	memcpy(&ifim->ifim_ifie, ifie, sizeof(ifim->ifim_ifie));
1827 	route_input(m, NULL, AF_UNSPEC);
1828 }
1829 
1830 /*
1831  * This is used to generate routing socket messages indicating
1832  * the address selection proposal from an interface.
1833  */
1834 void
1835 rtm_proposal(struct ifnet *ifp, struct rt_addrinfo *rtinfo, int flags,
1836     uint8_t prio)
1837 {
1838 	struct rt_msghdr	*rtm;
1839 	struct mbuf		*m;
1840 
1841 	m = rtm_msg1(RTM_PROPOSAL, rtinfo);
1842 	if (m == NULL)
1843 		return;
1844 	rtm = mtod(m, struct rt_msghdr *);
1845 	rtm->rtm_flags = RTF_DONE | flags;
1846 	rtm->rtm_priority = prio;
1847 	rtm->rtm_tableid = ifp->if_rdomain;
1848 	rtm->rtm_index = ifp->if_index;
1849 	rtm->rtm_addrs = rtinfo->rti_addrs;
1850 
1851 	route_input(m, NULL, rtinfo->rti_info[RTAX_DNS]->sa_family);
1852 }
1853 
1854 /*
1855  * This is used in dumping the kernel table via sysctl().
1856  */
1857 int
1858 sysctl_dumpentry(struct rtentry *rt, void *v, unsigned int id)
1859 {
1860 	struct walkarg		*w = v;
1861 	int			 error = 0, size;
1862 	struct rt_addrinfo	 info;
1863 	struct ifnet		*ifp;
1864 #ifdef BFD
1865 	struct sockaddr_bfd	 sa_bfd;
1866 #endif
1867 	struct sockaddr_rtlabel	 sa_rl;
1868 	struct sockaddr_in6	 sa_mask;
1869 
1870 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1871 		return 0;
1872 	if (w->w_op == NET_RT_DUMP && w->w_arg) {
1873 		u_int8_t prio = w->w_arg & RTP_MASK;
1874 		if (w->w_arg < 0) {
1875 			prio = (-w->w_arg) & RTP_MASK;
1876 			/* Show all routes that are not this priority */
1877 			if (prio == (rt->rt_priority & RTP_MASK))
1878 				return 0;
1879 		} else {
1880 			if (prio != (rt->rt_priority & RTP_MASK) &&
1881 			    prio != RTP_ANY)
1882 				return 0;
1883 		}
1884 	}
1885 	bzero(&info, sizeof(info));
1886 	info.rti_info[RTAX_DST] = rt_key(rt);
1887 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1888 	info.rti_info[RTAX_NETMASK] = rt_plen2mask(rt, &sa_mask);
1889 	ifp = if_get(rt->rt_ifidx);
1890 	if (ifp != NULL) {
1891 		info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
1892 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1893 		if (ifp->if_flags & IFF_POINTOPOINT)
1894 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1895 	}
1896 	if_put(ifp);
1897 	info.rti_info[RTAX_LABEL] = rtlabel_id2sa(rt->rt_labelid, &sa_rl);
1898 #ifdef BFD
1899 	if (rt->rt_flags & RTF_BFD)
1900 		info.rti_info[RTAX_BFD] = bfd2sa(rt, &sa_bfd);
1901 #endif
1902 #ifdef MPLS
1903 	if (rt->rt_flags & RTF_MPLS) {
1904 		struct sockaddr_mpls	 sa_mpls;
1905 
1906 		bzero(&sa_mpls, sizeof(sa_mpls));
1907 		sa_mpls.smpls_family = AF_MPLS;
1908 		sa_mpls.smpls_len = sizeof(sa_mpls);
1909 		sa_mpls.smpls_label = ((struct rt_mpls *)
1910 		    rt->rt_llinfo)->mpls_label;
1911 		info.rti_info[RTAX_SRC] = (struct sockaddr *)&sa_mpls;
1912 		info.rti_mpls = ((struct rt_mpls *)
1913 		    rt->rt_llinfo)->mpls_operation;
1914 	}
1915 #endif
1916 
1917 	size = rtm_msg2(RTM_GET, RTM_VERSION, &info, NULL, w);
1918 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1919 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1920 
1921 		rtm->rtm_pid = curproc->p_p->ps_pid;
1922 		rtm->rtm_flags = rt->rt_flags;
1923 		rtm->rtm_priority = rt->rt_priority & RTP_MASK;
1924 		rtm_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1925 		/* Do not account the routing table's reference. */
1926 		rtm->rtm_rmx.rmx_refcnt = rt->rt_refcnt - 1;
1927 		rtm->rtm_index = rt->rt_ifidx;
1928 		rtm->rtm_addrs = info.rti_addrs;
1929 		rtm->rtm_tableid = id;
1930 #ifdef MPLS
1931 		rtm->rtm_mpls = info.rti_mpls;
1932 #endif
1933 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1934 			w->w_where = NULL;
1935 		else
1936 			w->w_where += size;
1937 	}
1938 	return (error);
1939 }
1940 
1941 int
1942 sysctl_iflist(int af, struct walkarg *w)
1943 {
1944 	struct ifnet		*ifp;
1945 	struct ifaddr		*ifa;
1946 	struct rt_addrinfo	 info;
1947 	int			 len, error = 0;
1948 
1949 	bzero(&info, sizeof(info));
1950 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1951 		if (w->w_arg && w->w_arg != ifp->if_index)
1952 			continue;
1953 		/* Copy the link-layer address first */
1954 		info.rti_info[RTAX_IFP] = sdltosa(ifp->if_sadl);
1955 		len = rtm_msg2(RTM_IFINFO, RTM_VERSION, &info, 0, w);
1956 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1957 			struct if_msghdr *ifm;
1958 
1959 			ifm = (struct if_msghdr *)w->w_tmem;
1960 			ifm->ifm_index = ifp->if_index;
1961 			ifm->ifm_tableid = ifp->if_rdomain;
1962 			ifm->ifm_flags = ifp->if_flags;
1963 			if_getdata(ifp, &ifm->ifm_data);
1964 			ifm->ifm_addrs = info.rti_addrs;
1965 			error = copyout(ifm, w->w_where, len);
1966 			if (error)
1967 				return (error);
1968 			w->w_where += len;
1969 		}
1970 		info.rti_info[RTAX_IFP] = NULL;
1971 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1972 			KASSERT(ifa->ifa_addr->sa_family != AF_LINK);
1973 			if (af && af != ifa->ifa_addr->sa_family)
1974 				continue;
1975 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1976 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1977 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1978 			len = rtm_msg2(RTM_NEWADDR, RTM_VERSION, &info, 0, w);
1979 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1980 				struct ifa_msghdr *ifam;
1981 
1982 				ifam = (struct ifa_msghdr *)w->w_tmem;
1983 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1984 				ifam->ifam_flags = ifa->ifa_flags;
1985 				ifam->ifam_metric = ifa->ifa_metric;
1986 				ifam->ifam_addrs = info.rti_addrs;
1987 				error = copyout(w->w_tmem, w->w_where, len);
1988 				if (error)
1989 					return (error);
1990 				w->w_where += len;
1991 			}
1992 		}
1993 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1994 		    info.rti_info[RTAX_BRD] = NULL;
1995 	}
1996 	return (0);
1997 }
1998 
1999 int
2000 sysctl_ifnames(struct walkarg *w)
2001 {
2002 	struct if_nameindex_msg ifn;
2003 	struct ifnet *ifp;
2004 	int error = 0;
2005 
2006 	/* XXX ignore tableid for now */
2007 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
2008 		if (w->w_arg && w->w_arg != ifp->if_index)
2009 			continue;
2010 		w->w_needed += sizeof(ifn);
2011 		if (w->w_where && w->w_needed <= 0) {
2012 
2013 			memset(&ifn, 0, sizeof(ifn));
2014 			ifn.if_index = ifp->if_index;
2015 			strlcpy(ifn.if_name, ifp->if_xname,
2016 			    sizeof(ifn.if_name));
2017 			error = copyout(&ifn, w->w_where, sizeof(ifn));
2018 			if (error)
2019 				return (error);
2020 			w->w_where += sizeof(ifn);
2021 		}
2022 	}
2023 
2024 	return (0);
2025 }
2026 
2027 int
2028 sysctl_rtable(int *name, u_int namelen, void *where, size_t *given, void *new,
2029     size_t newlen)
2030 {
2031 	int			 i, error = EINVAL;
2032 	u_char			 af;
2033 	struct walkarg		 w;
2034 	struct rt_tableinfo	 tableinfo;
2035 	u_int			 tableid = 0;
2036 
2037 	if (new)
2038 		return (EPERM);
2039 	if (namelen < 3 || namelen > 4)
2040 		return (EINVAL);
2041 	af = name[0];
2042 	bzero(&w, sizeof(w));
2043 	w.w_where = where;
2044 	w.w_given = *given;
2045 	w.w_needed = 0 - w.w_given;
2046 	w.w_op = name[1];
2047 	w.w_arg = name[2];
2048 
2049 	if (namelen == 4) {
2050 		tableid = name[3];
2051 		if (!rtable_exists(tableid))
2052 			return (ENOENT);
2053 	} else
2054 		tableid = curproc->p_p->ps_rtableid;
2055 
2056 	switch (w.w_op) {
2057 	case NET_RT_DUMP:
2058 	case NET_RT_FLAGS:
2059 		NET_LOCK();
2060 		for (i = 1; i <= AF_MAX; i++) {
2061 			if (af != 0 && af != i)
2062 				continue;
2063 
2064 			error = rtable_walk(tableid, i, NULL, sysctl_dumpentry,
2065 			    &w);
2066 			if (error == EAFNOSUPPORT)
2067 				error = 0;
2068 			if (error)
2069 				break;
2070 		}
2071 		NET_UNLOCK();
2072 		break;
2073 
2074 	case NET_RT_IFLIST:
2075 		NET_LOCK();
2076 		error = sysctl_iflist(af, &w);
2077 		NET_UNLOCK();
2078 		break;
2079 
2080 	case NET_RT_STATS:
2081 		return (sysctl_rtable_rtstat(where, given, new));
2082 	case NET_RT_TABLE:
2083 		tableid = w.w_arg;
2084 		if (!rtable_exists(tableid))
2085 			return (ENOENT);
2086 		memset(&tableinfo, 0, sizeof tableinfo);
2087 		tableinfo.rti_tableid = tableid;
2088 		tableinfo.rti_domainid = rtable_l2(tableid);
2089 		error = sysctl_rdstruct(where, given, new,
2090 		    &tableinfo, sizeof(tableinfo));
2091 		return (error);
2092 	case NET_RT_IFNAMES:
2093 		NET_LOCK();
2094 		error = sysctl_ifnames(&w);
2095 		NET_UNLOCK();
2096 		break;
2097 	}
2098 	free(w.w_tmem, M_RTABLE, w.w_tmemsize);
2099 	w.w_needed += w.w_given;
2100 	if (where) {
2101 		*given = w.w_where - (caddr_t)where;
2102 		if (*given < w.w_needed)
2103 			return (ENOMEM);
2104 	} else
2105 		*given = (11 * w.w_needed) / 10;
2106 
2107 	return (error);
2108 }
2109 
2110 int
2111 sysctl_rtable_rtstat(void *oldp, size_t *oldlenp, void *newp)
2112 {
2113 	extern struct cpumem *rtcounters;
2114 	uint64_t counters[rts_ncounters];
2115 	struct rtstat rtstat;
2116 	uint32_t *words = (uint32_t *)&rtstat;
2117 	int i;
2118 
2119 	CTASSERT(sizeof(rtstat) == (nitems(counters) * sizeof(uint32_t)));
2120 	memset(&rtstat, 0, sizeof rtstat);
2121 	counters_read(rtcounters, counters, nitems(counters));
2122 
2123 	for (i = 0; i < nitems(counters); i++)
2124 		words[i] = (uint32_t)counters[i];
2125 
2126 	return (sysctl_rdstruct(oldp, oldlenp, newp, &rtstat, sizeof(rtstat)));
2127 }
2128 
2129 int
2130 rtm_validate_proposal(struct rt_addrinfo *info)
2131 {
2132 	if (info->rti_addrs & ~(RTA_NETMASK | RTA_IFA | RTA_DNS | RTA_STATIC |
2133 	    RTA_SEARCH)) {
2134 		return -1;
2135 	}
2136 
2137 	if (ISSET(info->rti_addrs, RTA_NETMASK)) {
2138 		struct sockaddr *sa = info->rti_info[RTAX_NETMASK];
2139 		if (sa == NULL)
2140 			return -1;
2141 		switch (sa->sa_family) {
2142 		case AF_INET:
2143 			if (sa->sa_len != sizeof(struct sockaddr_in))
2144 				return -1;
2145 			break;
2146 		case AF_INET6:
2147 			if (sa->sa_len != sizeof(struct sockaddr_in6))
2148 				return -1;
2149 			break;
2150 		default:
2151 			return -1;
2152 		}
2153 	}
2154 
2155 	if (ISSET(info->rti_addrs, RTA_IFA)) {
2156 		struct sockaddr *sa = info->rti_info[RTAX_IFA];
2157 		if (sa == NULL)
2158 			return -1;
2159 		switch (sa->sa_family) {
2160 		case AF_INET:
2161 			if (sa->sa_len != sizeof(struct sockaddr_in))
2162 				return -1;
2163 			break;
2164 		case AF_INET6:
2165 			if (sa->sa_len != sizeof(struct sockaddr_in6))
2166 				return -1;
2167 			break;
2168 		default:
2169 			return -1;
2170 		}
2171 	}
2172 
2173 	if (ISSET(info->rti_addrs, RTA_DNS)) {
2174 		struct sockaddr_rtdns *rtdns =
2175 		    (struct sockaddr_rtdns *)info->rti_info[RTAX_DNS];
2176 		if (rtdns == NULL)
2177 			return -1;
2178 		if (rtdns->sr_len > sizeof(*rtdns))
2179 			return -1;
2180 		if (rtdns->sr_len < offsetof(struct sockaddr_rtdns, sr_dns))
2181 			return -1;
2182 		switch (rtdns->sr_family) {
2183 		case AF_INET:
2184 			if ((rtdns->sr_len - offsetof(struct sockaddr_rtdns,
2185 			    sr_dns)) % sizeof(struct in_addr) != 0)
2186 				return -1;
2187 			break;
2188 #ifdef INET6
2189 		case AF_INET6:
2190 			if ((rtdns->sr_len - offsetof(struct sockaddr_rtdns,
2191 			    sr_dns)) % sizeof(struct in6_addr) != 0)
2192 				return -1;
2193 			break;
2194 #endif
2195 		default:
2196 			return -1;
2197 		}
2198 	}
2199 
2200 	if (ISSET(info->rti_addrs, RTA_STATIC)) {
2201 		struct sockaddr_rtstatic *rtstatic =
2202 		    (struct sockaddr_rtstatic *)info->rti_info[RTAX_STATIC];
2203 		if (rtstatic == NULL)
2204 			return -1;
2205 		if (rtstatic->sr_len > sizeof(*rtstatic))
2206 			return -1;
2207 		if (rtstatic->sr_len <=
2208 		    offsetof(struct sockaddr_rtstatic, sr_static))
2209 			return -1;
2210 	}
2211 
2212 	if (ISSET(info->rti_addrs, RTA_SEARCH)) {
2213 		struct sockaddr_rtsearch *rtsearch =
2214 		    (struct sockaddr_rtsearch *)info->rti_info[RTAX_SEARCH];
2215 		if (rtsearch == NULL)
2216 			return -1;
2217 		if (rtsearch->sr_len > sizeof(*rtsearch))
2218 			return -1;
2219 		if (rtsearch->sr_len <=
2220 		    offsetof(struct sockaddr_rtsearch, sr_search))
2221 			return -1;
2222 	}
2223 
2224 	return 0;
2225 }
2226 
2227 /*
2228  * Definitions of protocols supported in the ROUTE domain.
2229  */
2230 
2231 extern	struct domain routedomain;		/* or at least forward */
2232 
2233 struct protosw routesw[] = {
2234 {
2235   .pr_type	= SOCK_RAW,
2236   .pr_domain	= &routedomain,
2237   .pr_flags	= PR_ATOMIC|PR_ADDR|PR_WANTRCVD,
2238   .pr_output	= route_output,
2239   .pr_ctloutput	= route_ctloutput,
2240   .pr_usrreq	= route_usrreq,
2241   .pr_attach	= route_attach,
2242   .pr_detach	= route_detach,
2243   .pr_init	= route_prinit,
2244   .pr_sysctl	= sysctl_rtable
2245 }
2246 };
2247 
2248 struct domain routedomain = {
2249   .dom_family = PF_ROUTE,
2250   .dom_name = "route",
2251   .dom_init = route_init,
2252   .dom_protosw = routesw,
2253   .dom_protoswNPROTOSW = &routesw[nitems(routesw)]
2254 };
2255