xref: /freebsd/sys/net/rtsock.c (revision c74c7b73a005e689b922dfcfe5b94804669b595b)
1 /*-
2  * Copyright (c) 1988, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
30  * $FreeBSD$
31  */
32 #include "opt_sctp.h"
33 #include "opt_mpath.h"
34 #include "opt_route.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/domain.h>
40 #include <sys/jail.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/rwlock.h>
49 #include <sys/signalvar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/systm.h>
54 #include <sys/vimage.h>
55 
56 #include <net/if.h>
57 #include <net/if_dl.h>
58 #include <net/if_llatbl.h>
59 #include <net/netisr.h>
60 #include <net/raw_cb.h>
61 #include <net/route.h>
62 #include <net/vnet.h>
63 
64 #include <netinet/in.h>
65 #ifdef INET6
66 #include <netinet6/scope6_var.h>
67 #endif
68 
69 #ifdef SCTP
70 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
71 #endif /* SCTP */
72 
73 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
74 
75 /* NB: these are not modified */
76 static struct	sockaddr route_src = { 2, PF_ROUTE, };
77 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
78 
79 static struct {
80 	int	ip_count;	/* attached w/ AF_INET */
81 	int	ip6_count;	/* attached w/ AF_INET6 */
82 	int	ipx_count;	/* attached w/ AF_IPX */
83 	int	any_count;	/* total attached */
84 } route_cb;
85 
86 struct mtx rtsock_mtx;
87 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
88 
89 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
90 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
91 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
92 
93 static struct	ifqueue rtsintrq;
94 
95 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
96 SYSCTL_INT(_net_route, OID_AUTO, netisr_maxqlen, CTLFLAG_RW,
97     &rtsintrq.ifq_maxlen, 0, "maximum routing socket dispatch queue length");
98 
99 struct walkarg {
100 	int	w_tmemsize;
101 	int	w_op, w_arg;
102 	caddr_t	w_tmem;
103 	struct sysctl_req *w_req;
104 };
105 
106 static void	rts_input(struct mbuf *m);
107 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
108 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
109 			caddr_t cp, struct walkarg *w);
110 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
111 			struct rt_addrinfo *rtinfo);
112 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
113 static int	sysctl_iflist(int af, struct walkarg *w);
114 static int	sysctl_ifmalist(int af, struct walkarg *w);
115 static int	route_output(struct mbuf *m, struct socket *so);
116 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
117 			struct rt_metrics_lite *out);
118 static void	rt_getmetrics(const struct rt_metrics_lite *in,
119 			struct rt_metrics *out);
120 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
121 
122 static void
123 rts_init(void)
124 {
125 	int tmp;
126 
127 	rtsintrq.ifq_maxlen = 256;
128 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
129 		rtsintrq.ifq_maxlen = tmp;
130 	mtx_init(&rtsintrq.ifq_mtx, "rts_inq", NULL, MTX_DEF);
131 	netisr_register(NETISR_ROUTE, rts_input, &rtsintrq, 0);
132 }
133 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
134 
135 static void
136 rts_input(struct mbuf *m)
137 {
138 	struct sockproto route_proto;
139 	unsigned short *family;
140 	struct m_tag *tag;
141 
142 	route_proto.sp_family = PF_ROUTE;
143 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
144 	if (tag != NULL) {
145 		family = (unsigned short *)(tag + 1);
146 		route_proto.sp_protocol = *family;
147 		m_tag_delete(m, tag);
148 	} else
149 		route_proto.sp_protocol = 0;
150 
151 	raw_input(m, &route_proto, &route_src);
152 }
153 
154 /*
155  * It really doesn't make any sense at all for this code to share much
156  * with raw_usrreq.c, since its functionality is so restricted.  XXX
157  */
158 static void
159 rts_abort(struct socket *so)
160 {
161 
162 	raw_usrreqs.pru_abort(so);
163 }
164 
165 static void
166 rts_close(struct socket *so)
167 {
168 
169 	raw_usrreqs.pru_close(so);
170 }
171 
172 /* pru_accept is EOPNOTSUPP */
173 
174 static int
175 rts_attach(struct socket *so, int proto, struct thread *td)
176 {
177 	struct rawcb *rp;
178 	int s, error;
179 
180 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
181 
182 	/* XXX */
183 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
184 	if (rp == NULL)
185 		return ENOBUFS;
186 
187 	/*
188 	 * The splnet() is necessary to block protocols from sending
189 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
190 	 * this PCB is extant but incompletely initialized.
191 	 * Probably we should try to do more of this work beforehand and
192 	 * eliminate the spl.
193 	 */
194 	s = splnet();
195 	so->so_pcb = (caddr_t)rp;
196 	so->so_fibnum = td->td_proc->p_fibnum;
197 	error = raw_attach(so, proto);
198 	rp = sotorawcb(so);
199 	if (error) {
200 		splx(s);
201 		so->so_pcb = NULL;
202 		free(rp, M_PCB);
203 		return error;
204 	}
205 	RTSOCK_LOCK();
206 	switch(rp->rcb_proto.sp_protocol) {
207 	case AF_INET:
208 		route_cb.ip_count++;
209 		break;
210 	case AF_INET6:
211 		route_cb.ip6_count++;
212 		break;
213 	case AF_IPX:
214 		route_cb.ipx_count++;
215 		break;
216 	}
217 	route_cb.any_count++;
218 	RTSOCK_UNLOCK();
219 	soisconnected(so);
220 	so->so_options |= SO_USELOOPBACK;
221 	splx(s);
222 	return 0;
223 }
224 
225 static int
226 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
227 {
228 
229 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
230 }
231 
232 static int
233 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
234 {
235 
236 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
237 }
238 
239 /* pru_connect2 is EOPNOTSUPP */
240 /* pru_control is EOPNOTSUPP */
241 
242 static void
243 rts_detach(struct socket *so)
244 {
245 	struct rawcb *rp = sotorawcb(so);
246 
247 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
248 
249 	RTSOCK_LOCK();
250 	switch(rp->rcb_proto.sp_protocol) {
251 	case AF_INET:
252 		route_cb.ip_count--;
253 		break;
254 	case AF_INET6:
255 		route_cb.ip6_count--;
256 		break;
257 	case AF_IPX:
258 		route_cb.ipx_count--;
259 		break;
260 	}
261 	route_cb.any_count--;
262 	RTSOCK_UNLOCK();
263 	raw_usrreqs.pru_detach(so);
264 }
265 
266 static int
267 rts_disconnect(struct socket *so)
268 {
269 
270 	return (raw_usrreqs.pru_disconnect(so));
271 }
272 
273 /* pru_listen is EOPNOTSUPP */
274 
275 static int
276 rts_peeraddr(struct socket *so, struct sockaddr **nam)
277 {
278 
279 	return (raw_usrreqs.pru_peeraddr(so, nam));
280 }
281 
282 /* pru_rcvd is EOPNOTSUPP */
283 /* pru_rcvoob is EOPNOTSUPP */
284 
285 static int
286 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
287 	 struct mbuf *control, struct thread *td)
288 {
289 
290 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
291 }
292 
293 /* pru_sense is null */
294 
295 static int
296 rts_shutdown(struct socket *so)
297 {
298 
299 	return (raw_usrreqs.pru_shutdown(so));
300 }
301 
302 static int
303 rts_sockaddr(struct socket *so, struct sockaddr **nam)
304 {
305 
306 	return (raw_usrreqs.pru_sockaddr(so, nam));
307 }
308 
309 static struct pr_usrreqs route_usrreqs = {
310 	.pru_abort =		rts_abort,
311 	.pru_attach =		rts_attach,
312 	.pru_bind =		rts_bind,
313 	.pru_connect =		rts_connect,
314 	.pru_detach =		rts_detach,
315 	.pru_disconnect =	rts_disconnect,
316 	.pru_peeraddr =		rts_peeraddr,
317 	.pru_send =		rts_send,
318 	.pru_shutdown =		rts_shutdown,
319 	.pru_sockaddr =		rts_sockaddr,
320 	.pru_close =		rts_close,
321 };
322 
323 #ifndef _SOCKADDR_UNION_DEFINED
324 #define	_SOCKADDR_UNION_DEFINED
325 /*
326  * The union of all possible address formats we handle.
327  */
328 union sockaddr_union {
329 	struct sockaddr		sa;
330 	struct sockaddr_in	sin;
331 	struct sockaddr_in6	sin6;
332 };
333 #endif /* _SOCKADDR_UNION_DEFINED */
334 
335 static int
336 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
337     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
338 {
339 
340 	switch (info->rti_info[RTAX_DST]->sa_family) {
341 #ifdef INET
342 	case AF_INET:
343 	{
344 		struct in_addr ia;
345 
346 		/*
347 		 * 1. Check if the returned address is part of the jail.
348 		 */
349 		ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->sin_addr;
350 		if (prison_check_ip4(cred, &ia) != 0) {
351 			info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
352 
353 		} else {
354 			struct ifaddr *ifa;
355 			int found;
356 
357 			found = 0;
358 
359 			/*
360 			 * 2. Try to find an address on the given outgoing
361 			 *    interface that belongs to the jail.
362 			 */
363 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
364 				struct sockaddr *sa;
365 				sa = ifa->ifa_addr;
366 				if (sa->sa_family != AF_INET)
367 					continue;
368 				ia = ((struct sockaddr_in *)sa)->sin_addr;
369 				if (prison_check_ip4(cred, &ia) != 0) {
370 					found = 1;
371 					break;
372 				}
373 			}
374 			if (!found) {
375 				/*
376 				 * 3. As a last resort return the 'default'
377 				 * jail address.
378 				 */
379 				if (prison_getip4(cred, &ia) != 0)
380 					return (ESRCH);
381 			}
382 			bzero(&saun->sin, sizeof(struct sockaddr_in));
383 			saun->sin.sin_len = sizeof(struct sockaddr_in);
384 			saun->sin.sin_family = AF_INET;
385 			saun->sin.sin_addr.s_addr = ia.s_addr;
386 			info->rti_info[RTAX_IFA] =
387 			    (struct sockaddr *)&saun->sin;
388 		}
389 		break;
390 	}
391 #endif
392 #ifdef INET6
393 	case AF_INET6:
394 	{
395 		struct in6_addr ia6;
396 
397 		/*
398 		 * 1. Check if the returned address is part of the jail.
399 		 */
400 		bcopy(&((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->sin6_addr,
401 		    &ia6, sizeof(struct in6_addr));
402 		if (prison_check_ip6(cred, &ia6) != 0) {
403 			info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
404 		} else {
405 			struct ifaddr *ifa;
406 			int found;
407 
408 			found = 0;
409 
410 			/*
411 			 * 2. Try to find an address on the given outgoing
412 			 *    interface that belongs to the jail.
413 			 */
414 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
415 				struct sockaddr *sa;
416 				sa = ifa->ifa_addr;
417 				if (sa->sa_family != AF_INET6)
418 					continue;
419 				bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
420 				    &ia6, sizeof(struct in6_addr));
421 				if (prison_check_ip6(cred, &ia6) != 0) {
422 					found = 1;
423 					break;
424 				}
425 			}
426 			if (!found) {
427 				/*
428 				 * 3. As a last resort return the 'default'
429 				 * jail address.
430 				 */
431 				if (prison_getip6(cred, &ia6) != 0)
432 					return (ESRCH);
433 			}
434 			bzero(&saun->sin6, sizeof(struct sockaddr_in6));
435 			saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
436 			saun->sin6.sin6_family = AF_INET6;
437 			bcopy(&ia6, &saun->sin6.sin6_addr,
438 			    sizeof(struct in6_addr));
439 			if (sa6_recoverscope(&saun->sin6) != 0)
440 				return (ESRCH);
441 			info->rti_info[RTAX_IFA] =
442 			    (struct sockaddr *)&saun->sin6;
443 		}
444 		break;
445 	}
446 #endif
447 	default:
448 		return (ESRCH);
449 	}
450 	return (0);
451 }
452 
453 /*ARGSUSED*/
454 static int
455 route_output(struct mbuf *m, struct socket *so)
456 {
457 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
458 	INIT_VNET_NET(so->so_vnet);
459 	struct rt_msghdr *rtm = NULL;
460 	struct rtentry *rt = NULL;
461 	struct radix_node_head *rnh;
462 	struct rt_addrinfo info;
463 	int len, error = 0;
464 	struct ifnet *ifp = NULL;
465 	union sockaddr_union saun;
466 
467 #define senderr(e) { error = e; goto flush;}
468 	if (m == NULL || ((m->m_len < sizeof(long)) &&
469 		       (m = m_pullup(m, sizeof(long))) == NULL))
470 		return (ENOBUFS);
471 	if ((m->m_flags & M_PKTHDR) == 0)
472 		panic("route_output");
473 	len = m->m_pkthdr.len;
474 	if (len < sizeof(*rtm) ||
475 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
476 		info.rti_info[RTAX_DST] = NULL;
477 		senderr(EINVAL);
478 	}
479 	R_Malloc(rtm, struct rt_msghdr *, len);
480 	if (rtm == NULL) {
481 		info.rti_info[RTAX_DST] = NULL;
482 		senderr(ENOBUFS);
483 	}
484 	m_copydata(m, 0, len, (caddr_t)rtm);
485 	if (rtm->rtm_version != RTM_VERSION) {
486 		info.rti_info[RTAX_DST] = NULL;
487 		senderr(EPROTONOSUPPORT);
488 	}
489 	rtm->rtm_pid = curproc->p_pid;
490 	bzero(&info, sizeof(info));
491 	info.rti_addrs = rtm->rtm_addrs;
492 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
493 		info.rti_info[RTAX_DST] = NULL;
494 		senderr(EINVAL);
495 	}
496 	info.rti_flags = rtm->rtm_flags;
497 	if (info.rti_info[RTAX_DST] == NULL ||
498 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
499 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
500 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
501 		senderr(EINVAL);
502 	/*
503 	 * Verify that the caller has the appropriate privilege; RTM_GET
504 	 * is the only operation the non-superuser is allowed.
505 	 */
506 	if (rtm->rtm_type != RTM_GET) {
507 		error = priv_check(curthread, PRIV_NET_ROUTE);
508 		if (error)
509 			senderr(error);
510 	}
511 
512 	switch (rtm->rtm_type) {
513 		struct rtentry *saved_nrt;
514 
515 	case RTM_ADD:
516 		if (info.rti_info[RTAX_GATEWAY] == NULL)
517 			senderr(EINVAL);
518 		saved_nrt = NULL;
519 
520 		/* support for new ARP code */
521 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
522 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
523 			error = lla_rt_output(rtm, &info);
524 			break;
525 		}
526 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
527 		    so->so_fibnum);
528 		if (error == 0 && saved_nrt) {
529 			RT_LOCK(saved_nrt);
530 			rt_setmetrics(rtm->rtm_inits,
531 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
532 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
533 			RT_REMREF(saved_nrt);
534 			RT_UNLOCK(saved_nrt);
535 		}
536 		break;
537 
538 	case RTM_DELETE:
539 		saved_nrt = NULL;
540 		/* support for new ARP code */
541 		if (info.rti_info[RTAX_GATEWAY] &&
542 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
543 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
544 			error = lla_rt_output(rtm, &info);
545 			break;
546 		}
547 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
548 		    so->so_fibnum);
549 		if (error == 0) {
550 			RT_LOCK(saved_nrt);
551 			rt = saved_nrt;
552 			goto report;
553 		}
554 		break;
555 
556 	case RTM_GET:
557 	case RTM_CHANGE:
558 	case RTM_LOCK:
559 		rnh = V_rt_tables[so->so_fibnum][info.rti_info[RTAX_DST]->sa_family];
560 		if (rnh == NULL)
561 			senderr(EAFNOSUPPORT);
562 		RADIX_NODE_HEAD_RLOCK(rnh);
563 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
564 			info.rti_info[RTAX_NETMASK], rnh);
565 		if (rt == NULL) {	/* XXX looks bogus */
566 			RADIX_NODE_HEAD_RUNLOCK(rnh);
567 			senderr(ESRCH);
568 		}
569 #ifdef RADIX_MPATH
570 		/*
571 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
572 		 * we require users to specify a matching RTAX_GATEWAY.
573 		 *
574 		 * for RTM_GET, gate is optional even with multipath.
575 		 * if gate == NULL the first match is returned.
576 		 * (no need to call rt_mpath_matchgate if gate == NULL)
577 		 */
578 		if (rn_mpath_capable(rnh) &&
579 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
580 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
581 			if (!rt) {
582 				RADIX_NODE_HEAD_RUNLOCK(rnh);
583 				senderr(ESRCH);
584 			}
585 		}
586 #endif
587 		RT_LOCK(rt);
588 		RT_ADDREF(rt);
589 		RADIX_NODE_HEAD_RUNLOCK(rnh);
590 
591 		/*
592 		 * Fix for PR: 82974
593 		 *
594 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
595 		 * returns a perfect match in case a netmask is
596 		 * specified.  For host routes only a longest prefix
597 		 * match is returned so it is necessary to compare the
598 		 * existence of the netmask.  If both have a netmask
599 		 * rnh_lookup() did a perfect match and if none of them
600 		 * have a netmask both are host routes which is also a
601 		 * perfect match.
602 		 */
603 
604 		if (rtm->rtm_type != RTM_GET &&
605 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
606 			RT_UNLOCK(rt);
607 			senderr(ESRCH);
608 		}
609 
610 		switch(rtm->rtm_type) {
611 
612 		case RTM_GET:
613 		report:
614 			RT_LOCK_ASSERT(rt);
615 			if (jailed(curthread->td_ucred) &&
616 			    ((rt->rt_flags & RTF_HOST) == 0 ||
617 			    !prison_if(curthread->td_ucred, rt_key(rt)))) {
618 				RT_UNLOCK(rt);
619 				senderr(ESRCH);
620 			}
621 			info.rti_info[RTAX_DST] = rt_key(rt);
622 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
623 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
624 			info.rti_info[RTAX_GENMASK] = 0;
625 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
626 				ifp = rt->rt_ifp;
627 				if (ifp) {
628 					info.rti_info[RTAX_IFP] =
629 					    ifp->if_addr->ifa_addr;
630 					if (jailed(curthread->td_ucred)) {
631 						error = rtm_get_jailed(
632 						    &info, ifp, rt, &saun,
633 						    curthread->td_ucred);
634 						if (error != 0) {
635 							RT_UNLOCK(rt);
636 							senderr(ESRCH);
637 						}
638 					} else {
639 						info.rti_info[RTAX_IFA] =
640 						    rt->rt_ifa->ifa_addr;
641 					}
642 					if (ifp->if_flags & IFF_POINTOPOINT)
643 						info.rti_info[RTAX_BRD] =
644 						    rt->rt_ifa->ifa_dstaddr;
645 					rtm->rtm_index = ifp->if_index;
646 				} else {
647 					info.rti_info[RTAX_IFP] = NULL;
648 					info.rti_info[RTAX_IFA] = NULL;
649 				}
650 			} else if ((ifp = rt->rt_ifp) != NULL) {
651 				rtm->rtm_index = ifp->if_index;
652 			}
653 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
654 			if (len > rtm->rtm_msglen) {
655 				struct rt_msghdr *new_rtm;
656 				R_Malloc(new_rtm, struct rt_msghdr *, len);
657 				if (new_rtm == NULL) {
658 					RT_UNLOCK(rt);
659 					senderr(ENOBUFS);
660 				}
661 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
662 				Free(rtm); rtm = new_rtm;
663 			}
664 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
665 			rtm->rtm_flags = rt->rt_flags;
666 			rtm->rtm_use = 0;
667 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
668 			rtm->rtm_addrs = info.rti_addrs;
669 			break;
670 
671 		case RTM_CHANGE:
672 			/*
673 			 * New gateway could require new ifaddr, ifp;
674 			 * flags may also be different; ifp may be specified
675 			 * by ll sockaddr when protocol address is ambiguous
676 			 */
677 			if (((rt->rt_flags & RTF_GATEWAY) &&
678 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
679 			    info.rti_info[RTAX_IFP] != NULL ||
680 			    (info.rti_info[RTAX_IFA] != NULL &&
681 			     !sa_equal(info.rti_info[RTAX_IFA],
682 				       rt->rt_ifa->ifa_addr))) {
683 				RT_UNLOCK(rt);
684 				RADIX_NODE_HEAD_LOCK(rnh);
685 				error = rt_getifa_fib(&info, rt->rt_fibnum);
686 				RADIX_NODE_HEAD_UNLOCK(rnh);
687 				if (error != 0)
688 					senderr(error);
689 				RT_LOCK(rt);
690 			}
691 			if (info.rti_ifa != NULL &&
692 			    info.rti_ifa != rt->rt_ifa &&
693 			    rt->rt_ifa != NULL &&
694 			    rt->rt_ifa->ifa_rtrequest != NULL) {
695 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
696 				    &info);
697 				IFAFREE(rt->rt_ifa);
698 			}
699 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
700 				RT_UNLOCK(rt);
701 				RADIX_NODE_HEAD_LOCK(rnh);
702 				RT_LOCK(rt);
703 
704 				error = rt_setgate(rt, rt_key(rt),
705 				    info.rti_info[RTAX_GATEWAY]);
706 				RADIX_NODE_HEAD_UNLOCK(rnh);
707 				if (error != 0) {
708 					RT_UNLOCK(rt);
709 					senderr(error);
710 				}
711 				rt->rt_flags |= RTF_GATEWAY;
712 			}
713 			if (info.rti_ifa != NULL &&
714 			    info.rti_ifa != rt->rt_ifa) {
715 				IFAREF(info.rti_ifa);
716 				rt->rt_ifa = info.rti_ifa;
717 				rt->rt_ifp = info.rti_ifp;
718 			}
719 			/* Allow some flags to be toggled on change. */
720 			if (rtm->rtm_fmask & RTF_FMASK)
721 				rt->rt_flags = (rt->rt_flags &
722 				    ~rtm->rtm_fmask) |
723 				    (rtm->rtm_flags & rtm->rtm_fmask);
724 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
725 					&rt->rt_rmx);
726 			rtm->rtm_index = rt->rt_ifp->if_index;
727 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
728 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
729 			/* FALLTHROUGH */
730 		case RTM_LOCK:
731 			/* We don't support locks anymore */
732 			break;
733 		}
734 		RT_UNLOCK(rt);
735 		break;
736 
737 	default:
738 		senderr(EOPNOTSUPP);
739 	}
740 
741 flush:
742 	if (rtm) {
743 		if (error)
744 			rtm->rtm_errno = error;
745 		else
746 			rtm->rtm_flags |= RTF_DONE;
747 	}
748 	if (rt)		/* XXX can this be true? */
749 		RTFREE(rt);
750     {
751 	struct rawcb *rp = NULL;
752 	/*
753 	 * Check to see if we don't want our own messages.
754 	 */
755 	if ((so->so_options & SO_USELOOPBACK) == 0) {
756 		if (route_cb.any_count <= 1) {
757 			if (rtm)
758 				Free(rtm);
759 			m_freem(m);
760 			return (error);
761 		}
762 		/* There is another listener, so construct message */
763 		rp = sotorawcb(so);
764 	}
765 	if (rtm) {
766 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
767 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
768 			m_freem(m);
769 			m = NULL;
770 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
771 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
772 		Free(rtm);
773 	}
774 	if (m) {
775 		if (rp) {
776 			/*
777 			 * XXX insure we don't get a copy by
778 			 * invalidating our protocol
779 			 */
780 			unsigned short family = rp->rcb_proto.sp_family;
781 			rp->rcb_proto.sp_family = 0;
782 			rt_dispatch(m, info.rti_info[RTAX_DST]);
783 			rp->rcb_proto.sp_family = family;
784 		} else
785 			rt_dispatch(m, info.rti_info[RTAX_DST]);
786 	}
787     }
788 	return (error);
789 #undef	sa_equal
790 }
791 
792 static void
793 rt_setmetrics(u_long which, const struct rt_metrics *in,
794 	struct rt_metrics_lite *out)
795 {
796 #define metric(f, e) if (which & (f)) out->e = in->e;
797 	/*
798 	 * Only these are stored in the routing entry since introduction
799 	 * of tcp hostcache. The rest is ignored.
800 	 */
801 	metric(RTV_MTU, rmx_mtu);
802 	/* Userland -> kernel timebase conversion. */
803 	if (which & RTV_EXPIRE)
804 		out->rmx_expire = in->rmx_expire ?
805 		    in->rmx_expire - time_second + time_uptime : 0;
806 #undef metric
807 }
808 
809 static void
810 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
811 {
812 #define metric(e) out->e = in->e;
813 	bzero(out, sizeof(*out));
814 	metric(rmx_mtu);
815 	/* Kernel -> userland timebase conversion. */
816 	out->rmx_expire = in->rmx_expire ?
817 	    in->rmx_expire - time_uptime + time_second : 0;
818 #undef metric
819 }
820 
821 /*
822  * Extract the addresses of the passed sockaddrs.
823  * Do a little sanity checking so as to avoid bad memory references.
824  * This data is derived straight from userland.
825  */
826 static int
827 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
828 {
829 	struct sockaddr *sa;
830 	int i;
831 
832 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
833 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
834 			continue;
835 		sa = (struct sockaddr *)cp;
836 		/*
837 		 * It won't fit.
838 		 */
839 		if (cp + sa->sa_len > cplim)
840 			return (EINVAL);
841 		/*
842 		 * there are no more.. quit now
843 		 * If there are more bits, they are in error.
844 		 * I've seen this. route(1) can evidently generate these.
845 		 * This causes kernel to core dump.
846 		 * for compatibility, If we see this, point to a safe address.
847 		 */
848 		if (sa->sa_len == 0) {
849 			rtinfo->rti_info[i] = &sa_zero;
850 			return (0); /* should be EINVAL but for compat */
851 		}
852 		/* accept it */
853 		rtinfo->rti_info[i] = sa;
854 		cp += SA_SIZE(sa);
855 	}
856 	return (0);
857 }
858 
859 static struct mbuf *
860 rt_msg1(int type, struct rt_addrinfo *rtinfo)
861 {
862 	struct rt_msghdr *rtm;
863 	struct mbuf *m;
864 	int i;
865 	struct sockaddr *sa;
866 	int len, dlen;
867 
868 	switch (type) {
869 
870 	case RTM_DELADDR:
871 	case RTM_NEWADDR:
872 		len = sizeof(struct ifa_msghdr);
873 		break;
874 
875 	case RTM_DELMADDR:
876 	case RTM_NEWMADDR:
877 		len = sizeof(struct ifma_msghdr);
878 		break;
879 
880 	case RTM_IFINFO:
881 		len = sizeof(struct if_msghdr);
882 		break;
883 
884 	case RTM_IFANNOUNCE:
885 	case RTM_IEEE80211:
886 		len = sizeof(struct if_announcemsghdr);
887 		break;
888 
889 	default:
890 		len = sizeof(struct rt_msghdr);
891 	}
892 	if (len > MCLBYTES)
893 		panic("rt_msg1");
894 	m = m_gethdr(M_DONTWAIT, MT_DATA);
895 	if (m && len > MHLEN) {
896 		MCLGET(m, M_DONTWAIT);
897 		if ((m->m_flags & M_EXT) == 0) {
898 			m_free(m);
899 			m = NULL;
900 		}
901 	}
902 	if (m == NULL)
903 		return (m);
904 	m->m_pkthdr.len = m->m_len = len;
905 	m->m_pkthdr.rcvif = NULL;
906 	rtm = mtod(m, struct rt_msghdr *);
907 	bzero((caddr_t)rtm, len);
908 	for (i = 0; i < RTAX_MAX; i++) {
909 		if ((sa = rtinfo->rti_info[i]) == NULL)
910 			continue;
911 		rtinfo->rti_addrs |= (1 << i);
912 		dlen = SA_SIZE(sa);
913 		m_copyback(m, len, dlen, (caddr_t)sa);
914 		len += dlen;
915 	}
916 	if (m->m_pkthdr.len != len) {
917 		m_freem(m);
918 		return (NULL);
919 	}
920 	rtm->rtm_msglen = len;
921 	rtm->rtm_version = RTM_VERSION;
922 	rtm->rtm_type = type;
923 	return (m);
924 }
925 
926 static int
927 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
928 {
929 	int i;
930 	int len, dlen, second_time = 0;
931 	caddr_t cp0;
932 
933 	rtinfo->rti_addrs = 0;
934 again:
935 	switch (type) {
936 
937 	case RTM_DELADDR:
938 	case RTM_NEWADDR:
939 		len = sizeof(struct ifa_msghdr);
940 		break;
941 
942 	case RTM_IFINFO:
943 		len = sizeof(struct if_msghdr);
944 		break;
945 
946 	case RTM_NEWMADDR:
947 		len = sizeof(struct ifma_msghdr);
948 		break;
949 
950 	default:
951 		len = sizeof(struct rt_msghdr);
952 	}
953 	cp0 = cp;
954 	if (cp0)
955 		cp += len;
956 	for (i = 0; i < RTAX_MAX; i++) {
957 		struct sockaddr *sa;
958 
959 		if ((sa = rtinfo->rti_info[i]) == NULL)
960 			continue;
961 		rtinfo->rti_addrs |= (1 << i);
962 		dlen = SA_SIZE(sa);
963 		if (cp) {
964 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
965 			cp += dlen;
966 		}
967 		len += dlen;
968 	}
969 	len = ALIGN(len);
970 	if (cp == NULL && w != NULL && !second_time) {
971 		struct walkarg *rw = w;
972 
973 		if (rw->w_req) {
974 			if (rw->w_tmemsize < len) {
975 				if (rw->w_tmem)
976 					free(rw->w_tmem, M_RTABLE);
977 				rw->w_tmem = (caddr_t)
978 					malloc(len, M_RTABLE, M_NOWAIT);
979 				if (rw->w_tmem)
980 					rw->w_tmemsize = len;
981 			}
982 			if (rw->w_tmem) {
983 				cp = rw->w_tmem;
984 				second_time = 1;
985 				goto again;
986 			}
987 		}
988 	}
989 	if (cp) {
990 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
991 
992 		rtm->rtm_version = RTM_VERSION;
993 		rtm->rtm_type = type;
994 		rtm->rtm_msglen = len;
995 	}
996 	return (len);
997 }
998 
999 /*
1000  * This routine is called to generate a message from the routing
1001  * socket indicating that a redirect has occured, a routing lookup
1002  * has failed, or that a protocol has detected timeouts to a particular
1003  * destination.
1004  */
1005 void
1006 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1007 {
1008 	struct rt_msghdr *rtm;
1009 	struct mbuf *m;
1010 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1011 
1012 	if (route_cb.any_count == 0)
1013 		return;
1014 	m = rt_msg1(type, rtinfo);
1015 	if (m == NULL)
1016 		return;
1017 	rtm = mtod(m, struct rt_msghdr *);
1018 	rtm->rtm_flags = RTF_DONE | flags;
1019 	rtm->rtm_errno = error;
1020 	rtm->rtm_addrs = rtinfo->rti_addrs;
1021 	rt_dispatch(m, sa);
1022 }
1023 
1024 /*
1025  * This routine is called to generate a message from the routing
1026  * socket indicating that the status of a network interface has changed.
1027  */
1028 void
1029 rt_ifmsg(struct ifnet *ifp)
1030 {
1031 	struct if_msghdr *ifm;
1032 	struct mbuf *m;
1033 	struct rt_addrinfo info;
1034 
1035 	if (route_cb.any_count == 0)
1036 		return;
1037 	bzero((caddr_t)&info, sizeof(info));
1038 	m = rt_msg1(RTM_IFINFO, &info);
1039 	if (m == NULL)
1040 		return;
1041 	ifm = mtod(m, struct if_msghdr *);
1042 	ifm->ifm_index = ifp->if_index;
1043 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1044 	ifm->ifm_data = ifp->if_data;
1045 	ifm->ifm_addrs = 0;
1046 	rt_dispatch(m, NULL);
1047 }
1048 
1049 /*
1050  * This is called to generate messages from the routing socket
1051  * indicating a network interface has had addresses associated with it.
1052  * if we ever reverse the logic and replace messages TO the routing
1053  * socket indicate a request to configure interfaces, then it will
1054  * be unnecessary as the routing socket will automatically generate
1055  * copies of it.
1056  */
1057 void
1058 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1059 {
1060 	struct rt_addrinfo info;
1061 	struct sockaddr *sa = NULL;
1062 	int pass;
1063 	struct mbuf *m = NULL;
1064 	struct ifnet *ifp = ifa->ifa_ifp;
1065 
1066 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1067 		("unexpected cmd %u", cmd));
1068 #ifdef SCTP
1069 	/*
1070 	 * notify the SCTP stack
1071 	 * this will only get called when an address is added/deleted
1072 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1073 	 */
1074 	sctp_addr_change(ifa, cmd);
1075 #endif /* SCTP */
1076 	if (route_cb.any_count == 0)
1077 		return;
1078 	for (pass = 1; pass < 3; pass++) {
1079 		bzero((caddr_t)&info, sizeof(info));
1080 		if ((cmd == RTM_ADD && pass == 1) ||
1081 		    (cmd == RTM_DELETE && pass == 2)) {
1082 			struct ifa_msghdr *ifam;
1083 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1084 
1085 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1086 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1087 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1088 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1089 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1090 				continue;
1091 			ifam = mtod(m, struct ifa_msghdr *);
1092 			ifam->ifam_index = ifp->if_index;
1093 			ifam->ifam_metric = ifa->ifa_metric;
1094 			ifam->ifam_flags = ifa->ifa_flags;
1095 			ifam->ifam_addrs = info.rti_addrs;
1096 		}
1097 		if ((cmd == RTM_ADD && pass == 2) ||
1098 		    (cmd == RTM_DELETE && pass == 1)) {
1099 			struct rt_msghdr *rtm;
1100 
1101 			if (rt == NULL)
1102 				continue;
1103 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1104 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1105 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1106 			if ((m = rt_msg1(cmd, &info)) == NULL)
1107 				continue;
1108 			rtm = mtod(m, struct rt_msghdr *);
1109 			rtm->rtm_index = ifp->if_index;
1110 			rtm->rtm_flags |= rt->rt_flags;
1111 			rtm->rtm_errno = error;
1112 			rtm->rtm_addrs = info.rti_addrs;
1113 		}
1114 		rt_dispatch(m, sa);
1115 	}
1116 }
1117 
1118 /*
1119  * This is the analogue to the rt_newaddrmsg which performs the same
1120  * function but for multicast group memberhips.  This is easier since
1121  * there is no route state to worry about.
1122  */
1123 void
1124 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1125 {
1126 	struct rt_addrinfo info;
1127 	struct mbuf *m = NULL;
1128 	struct ifnet *ifp = ifma->ifma_ifp;
1129 	struct ifma_msghdr *ifmam;
1130 
1131 	if (route_cb.any_count == 0)
1132 		return;
1133 
1134 	bzero((caddr_t)&info, sizeof(info));
1135 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1136 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1137 	/*
1138 	 * If a link-layer address is present, present it as a ``gateway''
1139 	 * (similarly to how ARP entries, e.g., are presented).
1140 	 */
1141 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1142 	m = rt_msg1(cmd, &info);
1143 	if (m == NULL)
1144 		return;
1145 	ifmam = mtod(m, struct ifma_msghdr *);
1146 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1147 	    __func__));
1148 	ifmam->ifmam_index = ifp->if_index;
1149 	ifmam->ifmam_addrs = info.rti_addrs;
1150 	rt_dispatch(m, ifma->ifma_addr);
1151 }
1152 
1153 static struct mbuf *
1154 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1155 	struct rt_addrinfo *info)
1156 {
1157 	struct if_announcemsghdr *ifan;
1158 	struct mbuf *m;
1159 
1160 	if (route_cb.any_count == 0)
1161 		return NULL;
1162 	bzero((caddr_t)info, sizeof(*info));
1163 	m = rt_msg1(type, info);
1164 	if (m != NULL) {
1165 		ifan = mtod(m, struct if_announcemsghdr *);
1166 		ifan->ifan_index = ifp->if_index;
1167 		strlcpy(ifan->ifan_name, ifp->if_xname,
1168 			sizeof(ifan->ifan_name));
1169 		ifan->ifan_what = what;
1170 	}
1171 	return m;
1172 }
1173 
1174 /*
1175  * This is called to generate routing socket messages indicating
1176  * IEEE80211 wireless events.
1177  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1178  */
1179 void
1180 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1181 {
1182 	struct mbuf *m;
1183 	struct rt_addrinfo info;
1184 
1185 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1186 	if (m != NULL) {
1187 		/*
1188 		 * Append the ieee80211 data.  Try to stick it in the
1189 		 * mbuf containing the ifannounce msg; otherwise allocate
1190 		 * a new mbuf and append.
1191 		 *
1192 		 * NB: we assume m is a single mbuf.
1193 		 */
1194 		if (data_len > M_TRAILINGSPACE(m)) {
1195 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1196 			if (n == NULL) {
1197 				m_freem(m);
1198 				return;
1199 			}
1200 			bcopy(data, mtod(n, void *), data_len);
1201 			n->m_len = data_len;
1202 			m->m_next = n;
1203 		} else if (data_len > 0) {
1204 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1205 			m->m_len += data_len;
1206 		}
1207 		if (m->m_flags & M_PKTHDR)
1208 			m->m_pkthdr.len += data_len;
1209 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1210 		rt_dispatch(m, NULL);
1211 	}
1212 }
1213 
1214 /*
1215  * This is called to generate routing socket messages indicating
1216  * network interface arrival and departure.
1217  */
1218 void
1219 rt_ifannouncemsg(struct ifnet *ifp, int what)
1220 {
1221 	struct mbuf *m;
1222 	struct rt_addrinfo info;
1223 
1224 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1225 	if (m != NULL)
1226 		rt_dispatch(m, NULL);
1227 }
1228 
1229 static void
1230 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1231 {
1232 	INIT_VNET_NET(curvnet);
1233 	struct m_tag *tag;
1234 
1235 	/*
1236 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1237 	 * use when injecting the mbuf into the routing socket buffer from
1238 	 * the netisr.
1239 	 */
1240 	if (sa != NULL) {
1241 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1242 		    M_NOWAIT);
1243 		if (tag == NULL) {
1244 			m_freem(m);
1245 			return;
1246 		}
1247 		*(unsigned short *)(tag + 1) = sa->sa_family;
1248 		m_tag_prepend(m, tag);
1249 	}
1250 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1251 }
1252 
1253 /*
1254  * This is used in dumping the kernel table via sysctl().
1255  */
1256 static int
1257 sysctl_dumpentry(struct radix_node *rn, void *vw)
1258 {
1259 	struct walkarg *w = vw;
1260 	struct rtentry *rt = (struct rtentry *)rn;
1261 	int error = 0, size;
1262 	struct rt_addrinfo info;
1263 
1264 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1265 		return 0;
1266 	if (jailed(w->w_req->td->td_ucred) &&
1267 	    ((rt->rt_flags & RTF_HOST) == 0 ||
1268 	    !prison_if(w->w_req->td->td_ucred, rt_key(rt))))
1269 		return (0);
1270 	bzero((caddr_t)&info, sizeof(info));
1271 	info.rti_info[RTAX_DST] = rt_key(rt);
1272 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1273 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1274 	info.rti_info[RTAX_GENMASK] = 0;
1275 	if (rt->rt_ifp) {
1276 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1277 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1278 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1279 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1280 	}
1281 	size = rt_msg2(RTM_GET, &info, NULL, w);
1282 	if (w->w_req && w->w_tmem) {
1283 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1284 
1285 		rtm->rtm_flags = rt->rt_flags;
1286 		rtm->rtm_use = rt->rt_rmx.rmx_pksent;
1287 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1288 		rtm->rtm_index = rt->rt_ifp->if_index;
1289 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1290 		rtm->rtm_addrs = info.rti_addrs;
1291 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1292 		return (error);
1293 	}
1294 	return (error);
1295 }
1296 
1297 static int
1298 sysctl_iflist(int af, struct walkarg *w)
1299 {
1300 	INIT_VNET_NET(curvnet);
1301 	struct ifnet *ifp;
1302 	struct ifaddr *ifa;
1303 	struct rt_addrinfo info;
1304 	int len, error = 0;
1305 
1306 	bzero((caddr_t)&info, sizeof(info));
1307 	IFNET_RLOCK();
1308 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1309 		if (w->w_arg && w->w_arg != ifp->if_index)
1310 			continue;
1311 		ifa = ifp->if_addr;
1312 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1313 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1314 		info.rti_info[RTAX_IFP] = NULL;
1315 		if (w->w_req && w->w_tmem) {
1316 			struct if_msghdr *ifm;
1317 
1318 			ifm = (struct if_msghdr *)w->w_tmem;
1319 			ifm->ifm_index = ifp->if_index;
1320 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1321 			ifm->ifm_data = ifp->if_data;
1322 			ifm->ifm_addrs = info.rti_addrs;
1323 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
1324 			if (error)
1325 				goto done;
1326 		}
1327 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1328 			if (af && af != ifa->ifa_addr->sa_family)
1329 				continue;
1330 			if (jailed(w->w_req->td->td_ucred) &&
1331 			    !prison_if(w->w_req->td->td_ucred, ifa->ifa_addr))
1332 				continue;
1333 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1334 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1335 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1336 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1337 			if (w->w_req && w->w_tmem) {
1338 				struct ifa_msghdr *ifam;
1339 
1340 				ifam = (struct ifa_msghdr *)w->w_tmem;
1341 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1342 				ifam->ifam_flags = ifa->ifa_flags;
1343 				ifam->ifam_metric = ifa->ifa_metric;
1344 				ifam->ifam_addrs = info.rti_addrs;
1345 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1346 				if (error)
1347 					goto done;
1348 			}
1349 		}
1350 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1351 			info.rti_info[RTAX_BRD] = NULL;
1352 	}
1353 done:
1354 	IFNET_RUNLOCK();
1355 	return (error);
1356 }
1357 
1358 static int
1359 sysctl_ifmalist(int af, struct walkarg *w)
1360 {
1361 	INIT_VNET_NET(curvnet);
1362 	struct ifnet *ifp;
1363 	struct ifmultiaddr *ifma;
1364 	struct	rt_addrinfo info;
1365 	int	len, error = 0;
1366 	struct ifaddr *ifa;
1367 
1368 	bzero((caddr_t)&info, sizeof(info));
1369 	IFNET_RLOCK();
1370 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1371 		if (w->w_arg && w->w_arg != ifp->if_index)
1372 			continue;
1373 		ifa = ifp->if_addr;
1374 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1375 		IF_ADDR_LOCK(ifp);
1376 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1377 			if (af && af != ifma->ifma_addr->sa_family)
1378 				continue;
1379 			if (jailed(w->w_req->td->td_ucred) &&
1380 			    !prison_if(w->w_req->td->td_ucred, ifma->ifma_addr))
1381 				continue;
1382 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1383 			info.rti_info[RTAX_GATEWAY] =
1384 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1385 			    ifma->ifma_lladdr : NULL;
1386 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1387 			if (w->w_req && w->w_tmem) {
1388 				struct ifma_msghdr *ifmam;
1389 
1390 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1391 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1392 				ifmam->ifmam_flags = 0;
1393 				ifmam->ifmam_addrs = info.rti_addrs;
1394 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1395 				if (error) {
1396 					IF_ADDR_UNLOCK(ifp);
1397 					goto done;
1398 				}
1399 			}
1400 		}
1401 		IF_ADDR_UNLOCK(ifp);
1402 	}
1403 done:
1404 	IFNET_RUNLOCK();
1405 	return (error);
1406 }
1407 
1408 static int
1409 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1410 {
1411 	INIT_VNET_NET(curvnet);
1412 	int	*name = (int *)arg1;
1413 	u_int	namelen = arg2;
1414 	struct radix_node_head *rnh;
1415 	int	i, lim, error = EINVAL;
1416 	u_char	af;
1417 	struct	walkarg w;
1418 
1419 	name ++;
1420 	namelen--;
1421 	if (req->newptr)
1422 		return (EPERM);
1423 	if (namelen != 3)
1424 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1425 	af = name[0];
1426 	if (af > AF_MAX)
1427 		return (EINVAL);
1428 	bzero(&w, sizeof(w));
1429 	w.w_op = name[1];
1430 	w.w_arg = name[2];
1431 	w.w_req = req;
1432 
1433 	error = sysctl_wire_old_buffer(req, 0);
1434 	if (error)
1435 		return (error);
1436 	switch (w.w_op) {
1437 
1438 	case NET_RT_DUMP:
1439 	case NET_RT_FLAGS:
1440 		if (af == 0) {			/* dump all tables */
1441 			i = 1;
1442 			lim = AF_MAX;
1443 		} else				/* dump only one table */
1444 			i = lim = af;
1445 
1446 		/*
1447 		 * take care of llinfo entries, the caller must
1448 		 * specify an AF
1449 		 */
1450 		if (w.w_op == NET_RT_FLAGS &&
1451 #if defined(COMPAT_ROUTE_FLAGS)
1452 		    (w.w_arg & RTF_LLINFO)) {
1453 #else
1454 		    w.w_arg == 0) {
1455 #endif
1456 			if (af != 0)
1457 				error = lltable_sysctl_dumparp(af, w.w_req);
1458 			else
1459 				error = EINVAL;
1460 			break;
1461 		}
1462 		/*
1463 		 * take care of routing entries
1464 		 */
1465 		for (error = 0; error == 0 && i <= lim; i++)
1466 			if ((rnh = V_rt_tables[req->td->td_proc->p_fibnum][i]) != NULL) {
1467 				RADIX_NODE_HEAD_LOCK(rnh);
1468 			    	error = rnh->rnh_walktree(rnh,
1469 				    sysctl_dumpentry, &w);
1470 				RADIX_NODE_HEAD_UNLOCK(rnh);
1471 			} else if (af != 0)
1472 				error = EAFNOSUPPORT;
1473 		break;
1474 
1475 	case NET_RT_IFLIST:
1476 		error = sysctl_iflist(af, &w);
1477 		break;
1478 
1479 	case NET_RT_IFMALIST:
1480 		error = sysctl_ifmalist(af, &w);
1481 		break;
1482 	}
1483 	if (w.w_tmem)
1484 		free(w.w_tmem, M_RTABLE);
1485 	return (error);
1486 }
1487 
1488 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1489 
1490 /*
1491  * Definitions of protocols supported in the ROUTE domain.
1492  */
1493 
1494 static struct domain routedomain;		/* or at least forward */
1495 
1496 static struct protosw routesw[] = {
1497 {
1498 	.pr_type =		SOCK_RAW,
1499 	.pr_domain =		&routedomain,
1500 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1501 	.pr_output =		route_output,
1502 	.pr_ctlinput =		raw_ctlinput,
1503 	.pr_init =		raw_init,
1504 	.pr_usrreqs =		&route_usrreqs
1505 }
1506 };
1507 
1508 static struct domain routedomain = {
1509 	.dom_family =		PF_ROUTE,
1510 	.dom_name =		 "route",
1511 	.dom_protosw =		routesw,
1512 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1513 };
1514 
1515 DOMAIN_SET(route);
1516