xref: /openbsd-src/sys/net/rtsock.c (revision 50b7afb2c2c0993b0894d4e34bf857cb13ed9c80)
1 /*	$OpenBSD: rtsock.c,v 1.149 2014/07/12 18:44:22 tedu 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/protosw.h>
73 
74 #include <net/if.h>
75 #include <net/route.h>
76 #include <net/raw_cb.h>
77 
78 #ifdef MPLS
79 #include <netmpls/mpls.h>
80 #endif
81 
82 #include <sys/stdarg.h>
83 #include <sys/kernel.h>
84 #include <sys/timeout.h>
85 
86 struct sockaddr		route_dst = { 2, PF_ROUTE, };
87 struct sockaddr		route_src = { 2, PF_ROUTE, };
88 struct sockproto	route_proto = { PF_ROUTE, };
89 
90 struct walkarg {
91 	int	w_op, w_arg, w_given, w_needed, w_tmemsize;
92 	caddr_t	w_where, w_tmem;
93 };
94 
95 int	route_ctloutput(int, struct socket *, int, int, struct mbuf **);
96 void	route_input(struct mbuf *m0, ...);
97 
98 struct mbuf	*rt_msg1(int, struct rt_addrinfo *);
99 int		 rt_msg2(int, int, struct rt_addrinfo *, caddr_t,
100 		     struct walkarg *);
101 void		 rt_xaddrs(caddr_t, caddr_t, struct rt_addrinfo *);
102 
103 struct routecb {
104 	struct rawcb	rcb;
105 	struct timeout	timeout;
106 	unsigned int	msgfilter;
107 	unsigned int	flags;
108 	u_int		rtableid;
109 };
110 #define	sotoroutecb(so)	((struct routecb *)(so)->so_pcb)
111 
112 /*
113  * These flags and timeout are used for indicating to userland (via a
114  * RTM_DESYNC msg) when the route socket has overflowed and messages
115  * have been lost.
116  */
117 #define ROUTECB_FLAG_DESYNC	0x1	/* Route socket out of memory */
118 #define ROUTECB_FLAG_FLUSH	0x2	/* Wait until socket is empty before
119 					   queueing more packets */
120 
121 #define ROUTE_DESYNC_RESEND_TIMEOUT	(hz / 5)	/* In hz */
122 
123 void	rt_senddesync(void *);
124 
125 int
126 route_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
127     struct mbuf *control, struct proc *p)
128 {
129 	struct rawcb	*rp;
130 	struct routecb	*rop;
131 	int		 s, af;
132 	int		 error = 0;
133 
134 	s = splsoftnet();
135 	rp = sotorawcb(so);
136 
137 	switch (req) {
138 	case PRU_ATTACH:
139 		/*
140 		 * use the rawcb but allocate a routecb, this
141 		 * code does not care about the additional fields
142 		 * and works directly on the raw socket.
143 		 */
144 		rop = malloc(sizeof(struct routecb), M_PCB, M_WAITOK|M_ZERO);
145 		rp = &rop->rcb;
146 		so->so_pcb = rp;
147 		/* Init the timeout structure */
148 		timeout_set(&((struct routecb *)rp)->timeout, rt_senddesync, rp);
149 		/*
150 		 * Don't call raw_usrreq() in the attach case, because
151 		 * we want to allow non-privileged processes to listen
152 		 * on and send "safe" commands to the routing socket.
153 		 */
154 		if (curproc == 0)
155 			error = EACCES;
156 		else
157 			error = raw_attach(so, (int)(long)nam);
158 		if (error) {
159 			free(rp, M_PCB, 0);
160 			splx(s);
161 			return (error);
162 		}
163 		rop->rtableid = curproc->p_p->ps_rtableid;
164 		af = rp->rcb_proto.sp_protocol;
165 		if (af == AF_INET)
166 			route_cb.ip_count++;
167 		else if (af == AF_INET6)
168 			route_cb.ip6_count++;
169 #ifdef MPLS
170 		else if (af == AF_MPLS)
171 			route_cb.mpls_count++;
172 #endif
173 		rp->rcb_faddr = &route_src;
174 		route_cb.any_count++;
175 		soisconnected(so);
176 		so->so_options |= SO_USELOOPBACK;
177 		break;
178 
179 	case PRU_RCVD:
180 		rop = (struct routecb *)rp;
181 
182 		/*
183 		 * If we are in a FLUSH state, check if the buffer is
184 		 * empty so that we can clear the flag.
185 		 */
186 		if (((rop->flags & ROUTECB_FLAG_FLUSH) != 0) &&
187 		    ((sbspace(&rp->rcb_socket->so_rcv) ==
188 		    rp->rcb_socket->so_rcv.sb_hiwat)))
189 			rop->flags &= ~ROUTECB_FLAG_FLUSH;
190 		break;
191 
192 	case PRU_DETACH:
193 		if (rp) {
194 			timeout_del(&((struct routecb *)rp)->timeout);
195 			af = rp->rcb_proto.sp_protocol;
196 			if (af == AF_INET)
197 				route_cb.ip_count--;
198 			else if (af == AF_INET6)
199 				route_cb.ip6_count--;
200 #ifdef MPLS
201 			else if (af == AF_MPLS)
202 				route_cb.mpls_count--;
203 #endif
204 			route_cb.any_count--;
205 		}
206 		/* FALLTHROUGH */
207 	default:
208 		error = raw_usrreq(so, req, m, nam, control, p);
209 	}
210 
211 	splx(s);
212 	return (error);
213 }
214 
215 int
216 route_ctloutput(int op, struct socket *so, int level, int optname,
217     struct mbuf **mp)
218 {
219 	struct routecb *rop = sotoroutecb(so);
220 	struct mbuf *m = *mp;
221 	int error = 0;
222 	unsigned int tid;
223 
224 	if (level != AF_ROUTE) {
225 		error = EINVAL;
226 		if (op == PRCO_SETOPT && *mp)
227 			m_free(*mp);
228 		return (error);
229 	}
230 
231 	switch (op) {
232 	case PRCO_SETOPT:
233 		switch (optname) {
234 		case ROUTE_MSGFILTER:
235 			if (m == NULL || m->m_len != sizeof(unsigned int))
236 				error = EINVAL;
237 			else
238 				rop->msgfilter = *mtod(m, unsigned int *);
239 			break;
240 		case ROUTE_TABLEFILTER:
241 			if (m == NULL || m->m_len != sizeof(unsigned int)) {
242 				error = EINVAL;
243 				break;
244 			}
245 			tid = *mtod(m, unsigned int *);
246 			if (tid != RTABLE_ANY && !rtable_exists(tid))
247 				error = ENOENT;
248 			else
249 				rop->rtableid = tid;
250 			break;
251 		default:
252 			error = ENOPROTOOPT;
253 			break;
254 		}
255 		if (m)
256 			m_free(m);
257 		break;
258 	case PRCO_GETOPT:
259 		switch (optname) {
260 		case ROUTE_MSGFILTER:
261 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
262 			m->m_len = sizeof(unsigned int);
263 			*mtod(m, unsigned int *) = rop->msgfilter;
264 			break;
265 		case ROUTE_TABLEFILTER:
266 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
267 			m->m_len = sizeof(unsigned int);
268 			*mtod(m, unsigned int *) = rop->rtableid;
269 			break;
270 		default:
271 			error = ENOPROTOOPT;
272 			break;
273 		}
274 	}
275 	return (error);
276 }
277 
278 void
279 rt_senddesync(void *data)
280 {
281 	struct rawcb	*rp;
282 	struct routecb	*rop;
283 	struct mbuf	*desync_mbuf;
284 
285 	rp = (struct rawcb *)data;
286 	rop = (struct routecb *)rp;
287 
288 	/* If we are in a DESYNC state, try to send a RTM_DESYNC packet */
289 	if ((rop->flags & ROUTECB_FLAG_DESYNC) != 0) {
290 		/*
291 		 * If we fail to alloc memory or if sbappendaddr()
292 		 * fails, re-add timeout and try again.
293 		 */
294 		desync_mbuf = rt_msg1(RTM_DESYNC, NULL);
295 		if ((desync_mbuf != NULL) &&
296 		    (sbappendaddr(&rp->rcb_socket->so_rcv, &route_src,
297 		    desync_mbuf, (struct mbuf *)0) != 0)) {
298 			rop->flags &= ~ROUTECB_FLAG_DESYNC;
299 			sorwakeup(rp->rcb_socket);
300 		} else {
301 			if (desync_mbuf)
302 				m_freem(desync_mbuf);
303 			/* Re-add timeout to try sending msg again */
304 			timeout_add(&rop->timeout, ROUTE_DESYNC_RESEND_TIMEOUT);
305 		}
306 	}
307 }
308 
309 void
310 route_input(struct mbuf *m0, ...)
311 {
312 	struct rawcb *rp;
313 	struct routecb *rop;
314 	struct rt_msghdr *rtm;
315 	struct mbuf *m = m0;
316 	int sockets = 0;
317 	struct socket *last = NULL;
318 	va_list ap;
319 	struct sockproto *proto;
320 	struct sockaddr *sosrc, *sodst;
321 
322 	va_start(ap, m0);
323 	proto = va_arg(ap, struct sockproto *);
324 	sosrc = va_arg(ap, struct sockaddr *);
325 	sodst = va_arg(ap, struct sockaddr *);
326 	va_end(ap);
327 
328 	/* ensure that we can access the rtm_type via mtod() */
329 	if (m->m_len < offsetof(struct rt_msghdr, rtm_type) + 1) {
330 		m_freem(m);
331 		return;
332 	}
333 
334 	LIST_FOREACH(rp, &rawcb, rcb_list) {
335 		if (rp->rcb_socket->so_state & SS_CANTRCVMORE)
336 			continue;
337 		if (rp->rcb_proto.sp_family != proto->sp_family)
338 			continue;
339 		if (rp->rcb_proto.sp_protocol && proto->sp_protocol &&
340 		    rp->rcb_proto.sp_protocol != proto->sp_protocol)
341 			continue;
342 		/*
343 		 * We assume the lower level routines have
344 		 * placed the address in a canonical format
345 		 * suitable for a structure comparison.
346 		 *
347 		 * Note that if the lengths are not the same
348 		 * the comparison will fail at the first byte.
349 		 */
350 #define	equal(a1, a2) \
351   (bcmp((caddr_t)(a1), (caddr_t)(a2), a1->sa_len) == 0)
352 		if (rp->rcb_laddr && !equal(rp->rcb_laddr, sodst))
353 			continue;
354 		if (rp->rcb_faddr && !equal(rp->rcb_faddr, sosrc))
355 			continue;
356 
357 		/* filter messages that the process does not want */
358 		rop = (struct routecb *)rp;
359 		rtm = mtod(m, struct rt_msghdr *);
360 		/* but RTM_DESYNC can't be filtered */
361 		if (rtm->rtm_type != RTM_DESYNC && rop->msgfilter != 0 &&
362 		    !(rop->msgfilter & (1 << rtm->rtm_type)))
363 			continue;
364 		switch (rtm->rtm_type) {
365 		case RTM_IFANNOUNCE:
366 		case RTM_DESYNC:
367 			/* no tableid */
368 			break;
369 		case RTM_RESOLVE:
370 		case RTM_NEWADDR:
371 		case RTM_DELADDR:
372 		case RTM_IFINFO:
373 			/* check against rdomain id */
374 			if (rop->rtableid != RTABLE_ANY &&
375 			    rtable_l2(rop->rtableid) != rtm->rtm_tableid)
376 				continue;
377 			break;
378 		default:
379 			/* check against rtable id */
380 			if (rop->rtableid != RTABLE_ANY &&
381 			    rop->rtableid != rtm->rtm_tableid)
382 				continue;
383 			break;
384 		}
385 
386 		/*
387 		 * Check to see if the flush flag is set. If so, don't queue
388 		 * any more messages until the flag is cleared.
389 		 */
390 		if ((rop->flags & ROUTECB_FLAG_FLUSH) != 0)
391 			continue;
392 
393 		if (last) {
394 			struct mbuf *n;
395 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
396 				if (sbspace(&last->so_rcv) < (2 * MSIZE) ||
397 				    sbappendaddr(&last->so_rcv, sosrc,
398 				    n, (struct mbuf *)0) == 0) {
399 					/*
400 					 * Flag socket as desync'ed and
401 					 * flush required
402 					 */
403 					sotoroutecb(last)->flags |=
404 					    ROUTECB_FLAG_DESYNC |
405 					    ROUTECB_FLAG_FLUSH;
406 					rt_senddesync((void *) sotorawcb(last));
407 					m_freem(n);
408 				} else {
409 					sorwakeup(last);
410 					sockets++;
411 				}
412 			}
413 		}
414 		last = rp->rcb_socket;
415 	}
416 	if (last) {
417 		if (sbspace(&last->so_rcv) < (2 * MSIZE) ||
418 		    sbappendaddr(&last->so_rcv, sosrc,
419 		    m, (struct mbuf *)0) == 0) {
420 			/* Flag socket as desync'ed and flush required */
421 			sotoroutecb(last)->flags |=
422 			    ROUTECB_FLAG_DESYNC | ROUTECB_FLAG_FLUSH;
423 			rt_senddesync((void *) sotorawcb(last));
424 			m_freem(m);
425 		} else {
426 			sorwakeup(last);
427 			sockets++;
428 		}
429 	} else
430 		m_freem(m);
431 }
432 
433 int
434 route_output(struct mbuf *m, ...)
435 {
436 	struct rt_msghdr	*rtm = NULL;
437 	struct radix_node	*rn = NULL;
438 	struct rtentry		*rt = NULL;
439 	struct rtentry		*saved_nrt = NULL;
440 	struct radix_node_head	*rnh;
441 	struct rt_addrinfo	 info;
442 	int			 len, newgate, error = 0;
443 	struct ifnet		*ifp = NULL;
444 	struct ifaddr		*ifa = NULL;
445 	struct socket		*so;
446 	struct rawcb		*rp = NULL;
447 	struct sockaddr_rtlabel	 sa_rl;
448 #ifdef MPLS
449 	struct sockaddr_mpls	 sa_mpls, *psa_mpls;
450 #endif
451 	va_list			 ap;
452 	u_int			 tableid;
453 	u_int8_t		 prio;
454 	u_char			 vers;
455 
456 	va_start(ap, m);
457 	so = va_arg(ap, struct socket *);
458 	va_end(ap);
459 
460 	info.rti_info[RTAX_DST] = NULL;	/* for error handling (goto flush) */
461 	if (m == 0 || ((m->m_len < sizeof(int32_t)) &&
462 	    (m = m_pullup(m, sizeof(int32_t))) == 0))
463 		return (ENOBUFS);
464 	if ((m->m_flags & M_PKTHDR) == 0)
465 		panic("route_output");
466 	len = m->m_pkthdr.len;
467 	if (len < offsetof(struct rt_msghdr, rtm_type) + 1 ||
468 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
469 		error = EINVAL;
470 		goto fail;
471 	}
472 	vers = mtod(m, struct rt_msghdr *)->rtm_version;
473 	switch (vers) {
474 	case RTM_VERSION:
475 		if (len < sizeof(struct rt_msghdr)) {
476 			error = EINVAL;
477 			goto fail;
478 		}
479 		if (len > RTM_MAXSIZE) {
480 			error = EMSGSIZE;
481 			goto fail;
482 		}
483 		rtm = malloc(len, M_RTABLE, M_NOWAIT);
484 		if (rtm == NULL) {
485 			error = ENOBUFS;
486 			goto fail;
487 		}
488 		m_copydata(m, 0, len, (caddr_t)rtm);
489 		break;
490 	default:
491 		error = EPROTONOSUPPORT;
492 		goto fail;
493 	}
494 	rtm->rtm_pid = curproc->p_p->ps_pid;
495 	if (rtm->rtm_hdrlen == 0)	/* old client */
496 		rtm->rtm_hdrlen = sizeof(struct rt_msghdr);
497 	if (len < rtm->rtm_hdrlen) {
498 		error = EINVAL;
499 		goto fail;
500 	}
501 
502 	/* Verify that the caller is sending an appropriate message early */
503 	switch (rtm->rtm_type) {
504 	case RTM_ADD:
505 	case RTM_DELETE:
506 	case RTM_GET:
507 	case RTM_CHANGE:
508 	case RTM_LOCK:
509 		break;
510 	default:
511 		error = EOPNOTSUPP;
512 		goto fail;
513 	}
514 
515 	/*
516 	 * Verify that the caller has the appropriate privilege; RTM_GET
517 	 * is the only operation the non-superuser is allowed.
518 	 */
519 	if (rtm->rtm_type != RTM_GET && suser(curproc, 0) != 0) {
520 		error = EACCES;
521 		goto fail;
522 	}
523 
524 	tableid = rtm->rtm_tableid;
525 	if (!rtable_exists(tableid)) {
526 		if (rtm->rtm_type == RTM_ADD) {
527 			if ((error = rtable_add(tableid)) != 0)
528 				goto flush;
529 		} else {
530 			error = EINVAL;
531 			goto flush;
532 		}
533 	}
534 
535 
536 	/* Do not let userland play with kernel-only flags. */
537 	if ((rtm->rtm_flags & (RTF_LOCAL|RTF_BROADCAST)) != 0) {
538 		error = EINVAL;
539 		goto fail;
540 	}
541 
542 	/* make sure that kernel-only bits are not set */
543 	rtm->rtm_priority &= RTP_MASK;
544 	rtm->rtm_flags &= ~(RTF_DONE|RTF_CLONED);
545 	rtm->rtm_fmask &= RTF_FMASK;
546 
547 	if (rtm->rtm_priority != 0) {
548 		if (rtm->rtm_priority > RTP_MAX ||
549 		    rtm->rtm_priority == RTP_LOCAL) {
550 			error = EINVAL;
551 			goto fail;
552 		}
553 		prio = rtm->rtm_priority;
554 	} else if (rtm->rtm_type != RTM_ADD)
555 		prio = RTP_ANY;
556 	else if (rtm->rtm_flags & RTF_STATIC)
557 		prio = 0;
558 	else
559 		prio = RTP_DEFAULT;
560 
561 	bzero(&info, sizeof(info));
562 	info.rti_addrs = rtm->rtm_addrs;
563 	rt_xaddrs(rtm->rtm_hdrlen + (caddr_t)rtm, len + (caddr_t)rtm, &info);
564 	info.rti_flags = rtm->rtm_flags;
565 	if (info.rti_info[RTAX_DST] == NULL ||
566 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
567 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
568 	    info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX) ||
569 	    info.rti_info[RTAX_GENMASK] != NULL) {
570 		error = EINVAL;
571 		goto flush;
572 	}
573 #ifdef MPLS
574 	info.rti_mpls = rtm->rtm_mpls;
575 #endif
576 
577 	switch (rtm->rtm_type) {
578 	case RTM_ADD:
579 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
580 			error = EINVAL;
581 			goto flush;
582 		}
583 		error = rtrequest1(rtm->rtm_type, &info, prio, &saved_nrt,
584 		    tableid);
585 		if (error == 0 && saved_nrt) {
586 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
587 			    &saved_nrt->rt_rmx);
588 			saved_nrt->rt_refcnt--;
589 			/* write back the priority the kernel used */
590 			rtm->rtm_priority = saved_nrt->rt_priority & RTP_MASK;
591 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
592 			rtm->rtm_flags = saved_nrt->rt_flags;
593 		}
594 		break;
595 	case RTM_DELETE:
596 		error = rtrequest1(rtm->rtm_type, &info, prio, &saved_nrt,
597 		    tableid);
598 		if (error == 0) {
599 			(rt = saved_nrt)->rt_refcnt++;
600 			goto report;
601 		}
602 		break;
603 	case RTM_GET:
604 	case RTM_CHANGE:
605 	case RTM_LOCK:
606 		rnh = rtable_get(tableid, info.rti_info[RTAX_DST]->sa_family);
607 		if (rnh == NULL) {
608 			error = EAFNOSUPPORT;
609 			goto flush;
610 		}
611 		rt = rt_lookup(info.rti_info[RTAX_DST],
612 		    info.rti_info[RTAX_NETMASK], tableid);
613 		rn = (struct radix_node *)rt;
614 		if (rn == NULL || (rn->rn_flags & RNF_ROOT) != 0) {
615 			error = ESRCH;
616 			rt = NULL;
617 			goto flush;
618 		}
619 #ifndef SMALL_KERNEL
620 		if (rn_mpath_capable(rnh)) {
621 			/* first find the right priority */
622 			rt = rt_mpath_matchgate(rt, NULL, prio);
623 			if (!rt) {
624 				error = ESRCH;
625 				goto flush;
626 			}
627 			/*
628 			 * For RTM_CHANGE/LOCK, if we got multipath routes,
629 			 * a matching RTAX_GATEWAY is required.
630 			 * OR
631 			 * If a gateway is specified then RTM_GET and
632 			 * RTM_LOCK must match the gateway no matter
633 			 * what even in the non multipath case.
634 			 */
635 			if ((rt->rt_flags & RTF_MPATH) ||
636 			    (info.rti_info[RTAX_GATEWAY] && rtm->rtm_type !=
637 			    RTM_CHANGE)) {
638 				rt = rt_mpath_matchgate(rt,
639 				    info.rti_info[RTAX_GATEWAY], prio);
640 				if (!rt) {
641 					error = ESRCH;
642 					goto flush;
643 				}
644 				/*
645 				 * only RTM_GET may use an empty gateway
646 				 * on multipath routes
647 				 */
648 				if (!info.rti_info[RTAX_GATEWAY] &&
649 				    rtm->rtm_type != RTM_GET) {
650 					rt = NULL;
651 					error = ESRCH;
652 					goto flush;
653 				}
654 			}
655 			rn = (struct radix_node *)rt;
656 		}
657 #endif
658 		rt->rt_refcnt++;
659 
660 		/*
661 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
662 		 * returns a perfect match in case a netmask is specified.
663 		 * For host routes only a longest prefix match is returned
664 		 * so it is necessary to compare the existence of the netmaks.
665 		 * If both have a netmask rn_lookup() did a perfect match and
666 		 * if none of them have a netmask both are host routes which is
667 		 * also a perfect match.
668 		 */
669 		if (rtm->rtm_type != RTM_GET &&
670 		    !rt_mask(rt) != !info.rti_info[RTAX_NETMASK]) {
671 			error = ESRCH;
672 			goto flush;
673 		}
674 
675 		switch (rtm->rtm_type) {
676 		case RTM_GET:
677 report:
678 			info.rti_info[RTAX_DST] = rt_key(rt);
679 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
680 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
681 			info.rti_info[RTAX_LABEL] =
682 			    rtlabel_id2sa(rt->rt_labelid, &sa_rl);
683 #ifdef MPLS
684 			if (rt->rt_flags & RTF_MPLS) {
685 				bzero(&sa_mpls, sizeof(sa_mpls));
686 				sa_mpls.smpls_family = AF_MPLS;
687 				sa_mpls.smpls_len = sizeof(sa_mpls);
688 				sa_mpls.smpls_label = ((struct rt_mpls *)
689 				    rt->rt_llinfo)->mpls_label;
690 				info.rti_info[RTAX_SRC] =
691 				    (struct sockaddr *)&sa_mpls;
692 				info.rti_mpls = ((struct rt_mpls *)
693 				    rt->rt_llinfo)->mpls_operation;
694 				rtm->rtm_mpls = info.rti_mpls;
695 			}
696 #endif
697 			info.rti_info[RTAX_IFP] = NULL;
698 			info.rti_info[RTAX_IFA] = NULL;
699 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA) &&
700 			    (ifp = rt->rt_ifp) != NULL) {
701 				info.rti_info[RTAX_IFP] =
702 					(struct sockaddr *)ifp->if_sadl;
703 				info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
704 				if (ifp->if_flags & IFF_POINTOPOINT)
705 					info.rti_info[RTAX_BRD] =
706 					    rt->rt_ifa->ifa_dstaddr;
707 				else
708 					info.rti_info[RTAX_BRD] = NULL;
709 				rtm->rtm_index = ifp->if_index;
710 			}
711 			len = rt_msg2(rtm->rtm_type, RTM_VERSION, &info, NULL,
712 			    NULL);
713 			if (len > rtm->rtm_msglen) {
714 				struct rt_msghdr	*new_rtm;
715 				new_rtm = malloc(len, M_RTABLE, M_NOWAIT);
716 				if (new_rtm == NULL) {
717 					error = ENOBUFS;
718 					goto flush;
719 				}
720 				memcpy(new_rtm, rtm, rtm->rtm_msglen);
721 				free(rtm, M_RTABLE, 0);
722 				rtm = new_rtm;
723 			}
724 			rt_msg2(rtm->rtm_type, RTM_VERSION, &info, (caddr_t)rtm,
725 			    NULL);
726 			rtm->rtm_flags = rt->rt_flags;
727 			rtm->rtm_use = 0;
728 			rtm->rtm_priority = rt->rt_priority & RTP_MASK;
729 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
730 			rtm->rtm_addrs = info.rti_addrs;
731 			break;
732 
733 		case RTM_CHANGE:
734 			/*
735 			 * new gateway could require new ifaddr, ifp;
736 			 * flags may also be different; ifp may be specified
737 			 * by ll sockaddr when protocol address is ambiguous
738 			 */
739 			if ((error = rt_getifa(&info, tableid)) != 0)
740 				goto flush;
741 			newgate = 0;
742 			if (info.rti_info[RTAX_GATEWAY] != NULL)
743 				if (rt->rt_gateway == NULL ||
744 				    bcmp(rt->rt_gateway,
745 				    info.rti_info[RTAX_GATEWAY],
746 				    info.rti_info[RTAX_GATEWAY]->sa_len)) {
747 					newgate = 1;
748 				}
749 			if (info.rti_info[RTAX_GATEWAY] != NULL &&
750 			    rt_setgate(rt, rt_key(rt),
751 			    info.rti_info[RTAX_GATEWAY], tableid)) {
752 				error = EDQUOT;
753 				goto flush;
754 			}
755 			ifa = info.rti_ifa;
756 			if (ifa) {
757 				if (rt->rt_ifa != ifa) {
758 					if (rt->rt_ifa->ifa_rtrequest)
759 						rt->rt_ifa->ifa_rtrequest(
760 						    RTM_DELETE, rt);
761 					ifafree(rt->rt_ifa);
762 					rt->rt_ifa = ifa;
763 					ifa->ifa_refcnt++;
764 					rt->rt_ifp = ifa->ifa_ifp;
765 #ifndef SMALL_KERNEL
766 					/* recheck link state after ifp change*/
767 					rt_if_linkstate_change(
768 					    (struct radix_node *)rt, rt->rt_ifp,
769 					    tableid);
770 #endif
771 				}
772 			}
773 #ifdef MPLS
774 			if ((rtm->rtm_flags & RTF_MPLS) &&
775 			    info.rti_info[RTAX_SRC] != NULL) {
776 				struct rt_mpls *rt_mpls;
777 
778 				psa_mpls = (struct sockaddr_mpls *)
779 				    info.rti_info[RTAX_SRC];
780 
781 				if (rt->rt_llinfo == NULL) {
782 					rt->rt_llinfo =
783 					    malloc(sizeof(struct rt_mpls),
784 					    M_TEMP, M_NOWAIT|M_ZERO);
785 				}
786 				if (rt->rt_llinfo == NULL) {
787 					error = ENOMEM;
788 					goto flush;
789 				}
790 
791 				rt_mpls = (struct rt_mpls *)rt->rt_llinfo;
792 
793 				if (psa_mpls != NULL) {
794 					rt_mpls->mpls_label =
795 					    psa_mpls->smpls_label;
796 				}
797 
798 				rt_mpls->mpls_operation = info.rti_mpls;
799 
800 				/* XXX: set experimental bits */
801 
802 				rt->rt_flags |= RTF_MPLS;
803 			} else if (newgate || ((rtm->rtm_fmask & RTF_MPLS) &&
804 			    !(rtm->rtm_flags & RTF_MPLS))) {
805 				/* if gateway changed remove MPLS information */
806 				if (rt->rt_llinfo != NULL &&
807 				    rt->rt_flags & RTF_MPLS) {
808 					free(rt->rt_llinfo, M_TEMP, 0);
809 					rt->rt_llinfo = NULL;
810 					rt->rt_flags &= ~RTF_MPLS;
811 				}
812 			}
813 #endif
814 			/* Hack to allow some flags to be toggled */
815 			if (rtm->rtm_fmask)
816 				rt->rt_flags =
817 				    (rt->rt_flags & ~rtm->rtm_fmask) |
818 				    (rtm->rtm_flags & rtm->rtm_fmask);
819 
820 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
821 			    &rt->rt_rmx);
822 			rtm->rtm_index = rt->rt_ifp->if_index;
823 			rtm->rtm_priority = rt->rt_priority & RTP_MASK;
824 			rtm->rtm_flags = rt->rt_flags;
825 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
826 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt);
827 			if (info.rti_info[RTAX_LABEL] != NULL) {
828 				char *rtlabel = ((struct sockaddr_rtlabel *)
829 				    info.rti_info[RTAX_LABEL])->sr_label;
830 				rtlabel_unref(rt->rt_labelid);
831 				rt->rt_labelid =
832 				    rtlabel_name2id(rtlabel);
833 			}
834 			if_group_routechange(info.rti_info[RTAX_DST],
835 			    info.rti_info[RTAX_NETMASK]);
836 			/* FALLTHROUGH */
837 		case RTM_LOCK:
838 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
839 			rt->rt_rmx.rmx_locks |=
840 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
841 			rtm->rtm_priority = rt->rt_priority & RTP_MASK;
842 			break;
843 		}
844 		break;
845 	}
846 
847 flush:
848 	if (rtm) {
849 		if (error)
850 			rtm->rtm_errno = error;
851 		else {
852 			rtm->rtm_flags |= RTF_DONE;
853 		}
854 	}
855 	if (info.rti_info[RTAX_DST])
856 		route_proto.sp_protocol = info.rti_info[RTAX_DST]->sa_family;
857 	if (rt)
858 		rtfree(rt);
859 
860 	/*
861 	 * Check to see if we don't want our own messages.
862 	 */
863 	if (!(so->so_options & SO_USELOOPBACK)) {
864 		if (route_cb.any_count <= 1) {
865 fail:
866 			free(rtm, M_RTABLE, 0);
867 			m_freem(m);
868 			return (error);
869 		}
870 		/* There is another listener, so construct message */
871 		rp = sotorawcb(so);
872 	}
873 	if (rp)
874 		rp->rcb_proto.sp_family = 0; /* Avoid us */
875 	if (rtm) {
876 		if (m_copyback(m, 0, rtm->rtm_msglen, rtm, M_NOWAIT)) {
877 			m_freem(m);
878 			m = NULL;
879 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
880 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
881 		free(rtm, M_RTABLE, 0);
882 	}
883 	if (m)
884 		route_input(m, &route_proto, &route_src, &route_dst);
885 	if (rp)
886 		rp->rcb_proto.sp_family = PF_ROUTE;
887 
888 	return (error);
889 }
890 
891 void
892 rt_setmetrics(u_long which, struct rt_metrics *in, struct rt_kmetrics *out)
893 {
894 	if (which & RTV_MTU)
895 		out->rmx_mtu = in->rmx_mtu;
896 	if (which & RTV_EXPIRE)
897 		out->rmx_expire = in->rmx_expire;
898 	/* RTV_PRIORITY handled before */
899 }
900 
901 void
902 rt_getmetrics(struct rt_kmetrics *in, struct rt_metrics *out)
903 {
904 	bzero(out, sizeof(*out));
905 	out->rmx_locks = in->rmx_locks;
906 	out->rmx_mtu = in->rmx_mtu;
907 	out->rmx_expire = in->rmx_expire;
908 	out->rmx_pksent = in->rmx_pksent;
909 }
910 
911 #define ROUNDUP(a) \
912 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
913 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
914 
915 void
916 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
917 {
918 	struct sockaddr	*sa;
919 	int		 i;
920 
921 	bzero(rtinfo->rti_info, sizeof(rtinfo->rti_info));
922 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
923 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
924 			continue;
925 		rtinfo->rti_info[i] = sa = (struct sockaddr *)cp;
926 		ADVANCE(cp, sa);
927 	}
928 }
929 
930 struct mbuf *
931 rt_msg1(int type, struct rt_addrinfo *rtinfo)
932 {
933 	struct rt_msghdr	*rtm;
934 	struct mbuf		*m;
935 	int			 i;
936 	struct sockaddr		*sa;
937 	int			 len, dlen, hlen;
938 
939 	switch (type) {
940 	case RTM_DELADDR:
941 	case RTM_NEWADDR:
942 		len = sizeof(struct ifa_msghdr);
943 		break;
944 	case RTM_IFINFO:
945 		len = sizeof(struct if_msghdr);
946 		break;
947 	case RTM_IFANNOUNCE:
948 		len = sizeof(struct if_announcemsghdr);
949 		break;
950 	default:
951 		len = sizeof(struct rt_msghdr);
952 		break;
953 	}
954 	if (len > MCLBYTES)
955 		panic("rt_msg1");
956 	m = m_gethdr(M_DONTWAIT, MT_DATA);
957 	if (m && len > MHLEN) {
958 		MCLGET(m, M_DONTWAIT);
959 		if ((m->m_flags & M_EXT) == 0) {
960 			m_free(m);
961 			m = NULL;
962 		}
963 	}
964 	if (m == 0)
965 		return (m);
966 	m->m_pkthdr.len = m->m_len = hlen = len;
967 	m->m_pkthdr.rcvif = NULL;
968 	rtm = mtod(m, struct rt_msghdr *);
969 	bzero(rtm, len);
970 	for (i = 0; i < RTAX_MAX; i++) {
971 		if (rtinfo == NULL || (sa = rtinfo->rti_info[i]) == NULL)
972 			continue;
973 		rtinfo->rti_addrs |= (1 << i);
974 		dlen = ROUNDUP(sa->sa_len);
975 		if (m_copyback(m, len, dlen, sa, M_NOWAIT)) {
976 			m_freem(m);
977 			return (NULL);
978 		}
979 		len += dlen;
980 	}
981 	rtm->rtm_msglen = len;
982 	rtm->rtm_hdrlen = hlen;
983 	rtm->rtm_version = RTM_VERSION;
984 	rtm->rtm_type = type;
985 	return (m);
986 }
987 
988 int
989 rt_msg2(int type, int vers, struct rt_addrinfo *rtinfo, caddr_t cp,
990     struct walkarg *w)
991 {
992 	int		i;
993 	int		len, dlen, hlen, second_time = 0;
994 	caddr_t		cp0;
995 
996 	rtinfo->rti_addrs = 0;
997 again:
998 	switch (type) {
999 	case RTM_DELADDR:
1000 	case RTM_NEWADDR:
1001 		len = sizeof(struct ifa_msghdr);
1002 		break;
1003 	case RTM_IFINFO:
1004 		len = sizeof(struct if_msghdr);
1005 		break;
1006 	default:
1007 		len = sizeof(struct rt_msghdr);
1008 		break;
1009 	}
1010 	hlen = len;
1011 	if ((cp0 = cp) != NULL)
1012 		cp += len;
1013 	for (i = 0; i < RTAX_MAX; i++) {
1014 		struct sockaddr *sa;
1015 
1016 		if ((sa = rtinfo->rti_info[i]) == NULL)
1017 			continue;
1018 		rtinfo->rti_addrs |= (1 << i);
1019 		dlen = ROUNDUP(sa->sa_len);
1020 		if (cp) {
1021 			bcopy(sa, cp, (size_t)dlen);
1022 			cp += dlen;
1023 		}
1024 		len += dlen;
1025 	}
1026 	/* align message length to the next natural boundary */
1027 	len = ALIGN(len);
1028 	if (cp == 0 && w != NULL && !second_time) {
1029 		struct walkarg *rw = w;
1030 
1031 		rw->w_needed += len;
1032 		if (rw->w_needed <= 0 && rw->w_where) {
1033 			if (rw->w_tmemsize < len) {
1034 				free(rw->w_tmem, M_RTABLE, 0);
1035 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
1036 				if (rw->w_tmem)
1037 					rw->w_tmemsize = len;
1038 			}
1039 			if (rw->w_tmem) {
1040 				cp = rw->w_tmem;
1041 				second_time = 1;
1042 				goto again;
1043 			} else
1044 				rw->w_where = 0;
1045 		}
1046 	}
1047 	if (cp && w)		/* clear the message header */
1048 		bzero(cp0, hlen);
1049 
1050 	if (cp) {
1051 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1052 
1053 		rtm->rtm_version = RTM_VERSION;
1054 		rtm->rtm_type = type;
1055 		rtm->rtm_msglen = len;
1056 		rtm->rtm_hdrlen = hlen;
1057 	}
1058 	return (len);
1059 }
1060 
1061 /*
1062  * This routine is called to generate a message from the routing
1063  * socket indicating that a redirect has occurred, a routing lookup
1064  * has failed, or that a protocol has detected timeouts to a particular
1065  * destination.
1066  */
1067 void
1068 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags,
1069     struct ifnet *ifp, int error, u_int tableid)
1070 {
1071 	struct rt_msghdr	*rtm;
1072 	struct mbuf		*m;
1073 	struct sockaddr		*sa = rtinfo->rti_info[RTAX_DST];
1074 
1075 	if (route_cb.any_count == 0)
1076 		return;
1077 	m = rt_msg1(type, rtinfo);
1078 	if (m == 0)
1079 		return;
1080 	rtm = mtod(m, struct rt_msghdr *);
1081 	rtm->rtm_flags = RTF_DONE | flags;
1082 	rtm->rtm_errno = error;
1083 	rtm->rtm_tableid = tableid;
1084 	rtm->rtm_addrs = rtinfo->rti_addrs;
1085 	if (ifp != NULL)
1086 		rtm->rtm_index = ifp->if_index;
1087 	if (sa == NULL)
1088 		route_proto.sp_protocol = 0;
1089 	else
1090 		route_proto.sp_protocol = sa->sa_family;
1091 	route_input(m, &route_proto, &route_src, &route_dst);
1092 }
1093 
1094 /*
1095  * This routine is called to generate a message from the routing
1096  * socket indicating that the status of a network interface has changed.
1097  */
1098 void
1099 rt_ifmsg(struct ifnet *ifp)
1100 {
1101 	struct if_msghdr	*ifm;
1102 	struct mbuf		*m;
1103 
1104 	if (route_cb.any_count == 0)
1105 		return;
1106 	m = rt_msg1(RTM_IFINFO, NULL);
1107 	if (m == 0)
1108 		return;
1109 	ifm = mtod(m, struct if_msghdr *);
1110 	ifm->ifm_index = ifp->if_index;
1111 	ifm->ifm_tableid = ifp->if_rdomain;
1112 	ifm->ifm_flags = ifp->if_flags;
1113 	ifm->ifm_xflags = ifp->if_xflags;
1114 	ifm->ifm_data = ifp->if_data;
1115 	ifm->ifm_addrs = 0;
1116 	route_proto.sp_protocol = 0;
1117 	route_input(m, &route_proto, &route_src, &route_dst);
1118 }
1119 
1120 /*
1121  * This is called to generate messages from the routing socket
1122  * indicating a network interface has had addresses associated with it.
1123  * if we ever reverse the logic and replace messages TO the routing
1124  * socket indicate a request to configure interfaces, then it will
1125  * be unnecessary as the routing socket will automatically generate
1126  * copies of it.
1127  */
1128 void
1129 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1130 {
1131 	struct rt_addrinfo	 info;
1132 	struct sockaddr		*sa = NULL;
1133 	int			 pass;
1134 	struct mbuf		*m = NULL;
1135 	struct ifnet		*ifp = ifa->ifa_ifp;
1136 
1137 	if (route_cb.any_count == 0)
1138 		return;
1139 	for (pass = 1; pass < 3; pass++) {
1140 		bzero(&info, sizeof(info));
1141 		if ((cmd == RTM_ADD && pass == 1) ||
1142 		    (cmd == RTM_DELETE && pass == 2)) {
1143 			struct ifa_msghdr	*ifam;
1144 			int			 ncmd;
1145 
1146 			if (cmd == RTM_ADD)
1147 				ncmd = RTM_NEWADDR;
1148 			else
1149 				ncmd = RTM_DELADDR;
1150 
1151 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1152 			info.rti_info[RTAX_IFP] =
1153 			    (struct sockaddr *)ifp->if_sadl;
1154 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1155 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1156 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1157 				continue;
1158 			ifam = mtod(m, struct ifa_msghdr *);
1159 			ifam->ifam_index = ifp->if_index;
1160 			ifam->ifam_metric = ifa->ifa_metric;
1161 			ifam->ifam_flags = ifa->ifa_flags;
1162 			ifam->ifam_addrs = info.rti_addrs;
1163 			ifam->ifam_tableid = ifp->if_rdomain;
1164 		}
1165 		if ((cmd == RTM_ADD && pass == 2) ||
1166 		    (cmd == RTM_DELETE && pass == 1)) {
1167 			struct rt_msghdr *rtm;
1168 			struct sockaddr_rtlabel sa_rl;
1169 
1170 			if (rt == 0)
1171 				continue;
1172 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1173 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1174 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1175 			info.rti_info[RTAX_LABEL] =
1176 			    rtlabel_id2sa(rt->rt_labelid, &sa_rl);
1177 			if ((m = rt_msg1(cmd, &info)) == NULL)
1178 				continue;
1179 			rtm = mtod(m, struct rt_msghdr *);
1180 			rtm->rtm_index = ifp->if_index;
1181 			rtm->rtm_flags |= rt->rt_flags;
1182 			rtm->rtm_priority = rt->rt_priority & RTP_MASK;
1183 			rtm->rtm_errno = error;
1184 			rtm->rtm_addrs = info.rti_addrs;
1185 			rtm->rtm_tableid = ifp->if_rdomain;
1186 		}
1187 		if (sa == NULL)
1188 			route_proto.sp_protocol = 0;
1189 		else
1190 			route_proto.sp_protocol = sa->sa_family;
1191 		route_input(m, &route_proto, &route_src, &route_dst);
1192 	}
1193 }
1194 
1195 /*
1196  * This is called to generate routing socket messages indicating
1197  * network interface arrival and departure.
1198  */
1199 void
1200 rt_ifannouncemsg(struct ifnet *ifp, int what)
1201 {
1202 	struct if_announcemsghdr	*ifan;
1203 	struct mbuf			*m;
1204 
1205 	if (route_cb.any_count == 0)
1206 		return;
1207 	m = rt_msg1(RTM_IFANNOUNCE, NULL);
1208 	if (m == 0)
1209 		return;
1210 	ifan = mtod(m, struct if_announcemsghdr *);
1211 	ifan->ifan_index = ifp->if_index;
1212 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof(ifan->ifan_name));
1213 	ifan->ifan_what = what;
1214 	route_proto.sp_protocol = 0;
1215 	route_input(m, &route_proto, &route_src, &route_dst);
1216 }
1217 
1218 /*
1219  * This is used in dumping the kernel table via sysctl().
1220  */
1221 int
1222 sysctl_dumpentry(struct radix_node *rn, void *v, u_int id)
1223 {
1224 	struct walkarg		*w = v;
1225 	struct rtentry		*rt = (struct rtentry *)rn;
1226 	int			 error = 0, size;
1227 	struct rt_addrinfo	 info;
1228 #ifdef MPLS
1229 	struct sockaddr_mpls	 sa_mpls;
1230 #endif
1231 	struct sockaddr_rtlabel	 sa_rl;
1232 
1233 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1234 		return 0;
1235 	bzero(&info, sizeof(info));
1236 	info.rti_info[RTAX_DST] = rt_key(rt);
1237 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1238 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1239 	if (rt->rt_ifp) {
1240 		info.rti_info[RTAX_IFP] =
1241 		    (struct sockaddr *)rt->rt_ifp->if_sadl;
1242 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1243 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1244 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1245 	}
1246 	info.rti_info[RTAX_LABEL] = rtlabel_id2sa(rt->rt_labelid, &sa_rl);
1247 #ifdef MPLS
1248 	if (rt->rt_flags & RTF_MPLS) {
1249 		bzero(&sa_mpls, sizeof(sa_mpls));
1250 		sa_mpls.smpls_family = AF_MPLS;
1251 		sa_mpls.smpls_len = sizeof(sa_mpls);
1252 		sa_mpls.smpls_label = ((struct rt_mpls *)
1253 		    rt->rt_llinfo)->mpls_label;
1254 		info.rti_info[RTAX_SRC] = (struct sockaddr *)&sa_mpls;
1255 		info.rti_mpls = ((struct rt_mpls *)
1256 		    rt->rt_llinfo)->mpls_operation;
1257 	}
1258 #endif
1259 
1260 	size = rt_msg2(RTM_GET, RTM_VERSION, &info, NULL, w);
1261 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1262 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1263 
1264 		rtm->rtm_flags = rt->rt_flags;
1265 		rtm->rtm_priority = rt->rt_priority & RTP_MASK;
1266 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1267 		rtm->rtm_rmx.rmx_refcnt = rt->rt_refcnt;
1268 		rtm->rtm_index = rt->rt_ifp->if_index;
1269 		rtm->rtm_addrs = info.rti_addrs;
1270 		rtm->rtm_tableid = id;
1271 #ifdef MPLS
1272 		rtm->rtm_mpls = info.rti_mpls;
1273 #endif
1274 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1275 			w->w_where = NULL;
1276 		else
1277 			w->w_where += size;
1278 	}
1279 	return (error);
1280 }
1281 
1282 int
1283 sysctl_iflist(int af, struct walkarg *w)
1284 {
1285 	struct ifnet		*ifp;
1286 	struct ifaddr		*ifa;
1287 	struct rt_addrinfo	 info;
1288 	int			 len, error = 0;
1289 
1290 	bzero(&info, sizeof(info));
1291 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1292 		if (w->w_arg && w->w_arg != ifp->if_index)
1293 			continue;
1294 		/* Copy the link-layer address first */
1295 		info.rti_info[RTAX_IFP] = (struct sockaddr *)ifp->if_sadl;
1296 		len = rt_msg2(RTM_IFINFO, RTM_VERSION, &info, 0, w);
1297 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1298 			struct if_msghdr *ifm;
1299 
1300 			ifm = (struct if_msghdr *)w->w_tmem;
1301 			ifm->ifm_index = ifp->if_index;
1302 			ifm->ifm_tableid = ifp->if_rdomain;
1303 			ifm->ifm_flags = ifp->if_flags;
1304 			ifm->ifm_data = ifp->if_data;
1305 			ifm->ifm_addrs = info.rti_addrs;
1306 			error = copyout(ifm, w->w_where, len);
1307 			if (error)
1308 				return (error);
1309 			w->w_where += len;
1310 		}
1311 		info.rti_info[RTAX_IFP] = NULL;
1312 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1313 			KASSERT(ifa->ifa_addr->sa_family != AF_LINK);
1314 			if (af && af != ifa->ifa_addr->sa_family)
1315 				continue;
1316 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1317 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1318 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1319 			len = rt_msg2(RTM_NEWADDR, RTM_VERSION, &info, 0, w);
1320 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1321 				struct ifa_msghdr *ifam;
1322 
1323 				ifam = (struct ifa_msghdr *)w->w_tmem;
1324 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1325 				ifam->ifam_flags = ifa->ifa_flags;
1326 				ifam->ifam_metric = ifa->ifa_metric;
1327 				ifam->ifam_addrs = info.rti_addrs;
1328 				error = copyout(w->w_tmem, w->w_where, len);
1329 				if (error)
1330 					return (error);
1331 				w->w_where += len;
1332 			}
1333 		}
1334 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1335 		    info.rti_info[RTAX_BRD] = NULL;
1336 	}
1337 	return (0);
1338 }
1339 
1340 int
1341 sysctl_rtable(int *name, u_int namelen, void *where, size_t *given, void *new,
1342     size_t newlen)
1343 {
1344 	struct radix_node_head	*rnh;
1345 	int			 i, s, error = EINVAL;
1346 	u_char  		 af;
1347 	struct walkarg		 w;
1348 	struct rt_tableinfo	 tableinfo;
1349 	u_int			 tableid = 0;
1350 
1351 	if (new)
1352 		return (EPERM);
1353 	if (namelen < 3 || namelen > 4)
1354 		return (EINVAL);
1355 	af = name[0];
1356 	bzero(&w, sizeof(w));
1357 	w.w_where = where;
1358 	w.w_given = *given;
1359 	w.w_needed = 0 - w.w_given;
1360 	w.w_op = name[1];
1361 	w.w_arg = name[2];
1362 
1363 	if (namelen == 4) {
1364 		tableid = name[3];
1365 		if (!rtable_exists(tableid))
1366 			return (ENOENT);
1367 	} else
1368 		tableid = curproc->p_p->ps_rtableid;
1369 
1370 	s = splsoftnet();
1371 	switch (w.w_op) {
1372 
1373 	case NET_RT_DUMP:
1374 	case NET_RT_FLAGS:
1375 		for (i = 1; i <= AF_MAX; i++)
1376 			if ((rnh = rtable_get(tableid, i)) != NULL &&
1377 			    (af == 0 || af == i) &&
1378 			    (error = (*rnh->rnh_walktree)(rnh,
1379 			    sysctl_dumpentry, &w)))
1380 				break;
1381 		break;
1382 
1383 	case NET_RT_IFLIST:
1384 		error = sysctl_iflist(af, &w);
1385 		break;
1386 
1387 	case NET_RT_STATS:
1388 		error = sysctl_rdstruct(where, given, new,
1389 		    &rtstat, sizeof(rtstat));
1390 		splx(s);
1391 		return (error);
1392 	case NET_RT_TABLE:
1393 		tableid = w.w_arg;
1394 		if (!rtable_exists(tableid)) {
1395 			splx(s);
1396 			return (ENOENT);
1397 		}
1398 		tableinfo.rti_tableid = tableid;
1399 		tableinfo.rti_domainid = rtable_l2(tableid);
1400 		error = sysctl_rdstruct(where, given, new,
1401 		    &tableinfo, sizeof(tableinfo));
1402 		splx(s);
1403 		return (error);
1404 	}
1405 	splx(s);
1406 	free(w.w_tmem, M_RTABLE, 0);
1407 	w.w_needed += w.w_given;
1408 	if (where) {
1409 		*given = w.w_where - (caddr_t)where;
1410 		if (*given < w.w_needed)
1411 			return (ENOMEM);
1412 	} else
1413 		*given = (11 * w.w_needed) / 10;
1414 
1415 	return (error);
1416 }
1417 
1418 /*
1419  * Definitions of protocols supported in the ROUTE domain.
1420  */
1421 
1422 extern	struct domain routedomain;		/* or at least forward */
1423 
1424 struct protosw routesw[] = {
1425 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR|PR_WANTRCVD,
1426   route_input,	route_output,	raw_ctlinput,	route_ctloutput,
1427   route_usrreq,
1428   raw_init,	0,		0,		0,
1429   sysctl_rtable,
1430 }
1431 };
1432 
1433 struct domain routedomain =
1434     { PF_ROUTE, "route", route_init, 0, 0,
1435       routesw, &routesw[nitems(routesw)] };
1436