xref: /freebsd/sys/net/rtsock.c (revision 9517e866259191fcd39434a97ad849a9b59b9b9f)
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_inet.h"
35 #include "opt_inet6.h"
36 
37 #include <sys/param.h>
38 #include <sys/jail.h>
39 #include <sys/kernel.h>
40 #include <sys/domain.h>
41 #include <sys/lock.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/priv.h>
45 #include <sys/proc.h>
46 #include <sys/protosw.h>
47 #include <sys/rwlock.h>
48 #include <sys/signalvar.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/sysctl.h>
52 #include <sys/systm.h>
53 
54 #include <net/if.h>
55 #include <net/if_dl.h>
56 #include <net/if_llatbl.h>
57 #include <net/netisr.h>
58 #include <net/raw_cb.h>
59 #include <net/route.h>
60 #include <net/vnet.h>
61 
62 #include <netinet/in.h>
63 #ifdef INET6
64 #include <netinet6/scope6_var.h>
65 #endif
66 
67 #if defined(INET) || defined(INET6)
68 #ifdef SCTP
69 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
70 #endif /* SCTP */
71 #endif
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 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
94 
95 struct walkarg {
96 	int	w_tmemsize;
97 	int	w_op, w_arg;
98 	caddr_t	w_tmem;
99 	struct sysctl_req *w_req;
100 };
101 
102 static void	rts_input(struct mbuf *m);
103 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
104 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
105 			caddr_t cp, struct walkarg *w);
106 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
107 			struct rt_addrinfo *rtinfo);
108 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
109 static int	sysctl_iflist(int af, struct walkarg *w);
110 static int	sysctl_ifmalist(int af, struct walkarg *w);
111 static int	route_output(struct mbuf *m, struct socket *so);
112 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
113 			struct rt_metrics_lite *out);
114 static void	rt_getmetrics(const struct rt_metrics_lite *in,
115 			struct rt_metrics *out);
116 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
117 
118 static struct netisr_handler rtsock_nh = {
119 	.nh_name = "rtsock",
120 	.nh_handler = rts_input,
121 	.nh_proto = NETISR_ROUTE,
122 	.nh_policy = NETISR_POLICY_SOURCE,
123 };
124 
125 static int
126 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
127 {
128 	int error, qlimit;
129 
130 	netisr_getqlimit(&rtsock_nh, &qlimit);
131 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
132         if (error || !req->newptr)
133                 return (error);
134 	if (qlimit < 1)
135 		return (EINVAL);
136 	return (netisr_setqlimit(&rtsock_nh, qlimit));
137 }
138 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
139     0, 0, sysctl_route_netisr_maxqlen, "I",
140     "maximum routing socket dispatch queue length");
141 
142 static void
143 rts_init(void)
144 {
145 	int tmp;
146 
147 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
148 		rtsock_nh.nh_qlimit = tmp;
149 	netisr_register(&rtsock_nh);
150 }
151 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
152 
153 static void
154 rts_input(struct mbuf *m)
155 {
156 	struct sockproto route_proto;
157 	unsigned short *family;
158 	struct m_tag *tag;
159 
160 	route_proto.sp_family = PF_ROUTE;
161 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
162 	if (tag != NULL) {
163 		family = (unsigned short *)(tag + 1);
164 		route_proto.sp_protocol = *family;
165 		m_tag_delete(m, tag);
166 	} else
167 		route_proto.sp_protocol = 0;
168 
169 	raw_input(m, &route_proto, &route_src);
170 }
171 
172 /*
173  * It really doesn't make any sense at all for this code to share much
174  * with raw_usrreq.c, since its functionality is so restricted.  XXX
175  */
176 static void
177 rts_abort(struct socket *so)
178 {
179 
180 	raw_usrreqs.pru_abort(so);
181 }
182 
183 static void
184 rts_close(struct socket *so)
185 {
186 
187 	raw_usrreqs.pru_close(so);
188 }
189 
190 /* pru_accept is EOPNOTSUPP */
191 
192 static int
193 rts_attach(struct socket *so, int proto, struct thread *td)
194 {
195 	struct rawcb *rp;
196 	int s, error;
197 
198 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
199 
200 	/* XXX */
201 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
202 	if (rp == NULL)
203 		return ENOBUFS;
204 
205 	/*
206 	 * The splnet() is necessary to block protocols from sending
207 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
208 	 * this PCB is extant but incompletely initialized.
209 	 * Probably we should try to do more of this work beforehand and
210 	 * eliminate the spl.
211 	 */
212 	s = splnet();
213 	so->so_pcb = (caddr_t)rp;
214 	so->so_fibnum = td->td_proc->p_fibnum;
215 	error = raw_attach(so, proto);
216 	rp = sotorawcb(so);
217 	if (error) {
218 		splx(s);
219 		so->so_pcb = NULL;
220 		free(rp, M_PCB);
221 		return error;
222 	}
223 	RTSOCK_LOCK();
224 	switch(rp->rcb_proto.sp_protocol) {
225 	case AF_INET:
226 		route_cb.ip_count++;
227 		break;
228 	case AF_INET6:
229 		route_cb.ip6_count++;
230 		break;
231 	case AF_IPX:
232 		route_cb.ipx_count++;
233 		break;
234 	}
235 	route_cb.any_count++;
236 	RTSOCK_UNLOCK();
237 	soisconnected(so);
238 	so->so_options |= SO_USELOOPBACK;
239 	splx(s);
240 	return 0;
241 }
242 
243 static int
244 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
245 {
246 
247 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
248 }
249 
250 static int
251 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
252 {
253 
254 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
255 }
256 
257 /* pru_connect2 is EOPNOTSUPP */
258 /* pru_control is EOPNOTSUPP */
259 
260 static void
261 rts_detach(struct socket *so)
262 {
263 	struct rawcb *rp = sotorawcb(so);
264 
265 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
266 
267 	RTSOCK_LOCK();
268 	switch(rp->rcb_proto.sp_protocol) {
269 	case AF_INET:
270 		route_cb.ip_count--;
271 		break;
272 	case AF_INET6:
273 		route_cb.ip6_count--;
274 		break;
275 	case AF_IPX:
276 		route_cb.ipx_count--;
277 		break;
278 	}
279 	route_cb.any_count--;
280 	RTSOCK_UNLOCK();
281 	raw_usrreqs.pru_detach(so);
282 }
283 
284 static int
285 rts_disconnect(struct socket *so)
286 {
287 
288 	return (raw_usrreqs.pru_disconnect(so));
289 }
290 
291 /* pru_listen is EOPNOTSUPP */
292 
293 static int
294 rts_peeraddr(struct socket *so, struct sockaddr **nam)
295 {
296 
297 	return (raw_usrreqs.pru_peeraddr(so, nam));
298 }
299 
300 /* pru_rcvd is EOPNOTSUPP */
301 /* pru_rcvoob is EOPNOTSUPP */
302 
303 static int
304 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
305 	 struct mbuf *control, struct thread *td)
306 {
307 
308 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
309 }
310 
311 /* pru_sense is null */
312 
313 static int
314 rts_shutdown(struct socket *so)
315 {
316 
317 	return (raw_usrreqs.pru_shutdown(so));
318 }
319 
320 static int
321 rts_sockaddr(struct socket *so, struct sockaddr **nam)
322 {
323 
324 	return (raw_usrreqs.pru_sockaddr(so, nam));
325 }
326 
327 static struct pr_usrreqs route_usrreqs = {
328 	.pru_abort =		rts_abort,
329 	.pru_attach =		rts_attach,
330 	.pru_bind =		rts_bind,
331 	.pru_connect =		rts_connect,
332 	.pru_detach =		rts_detach,
333 	.pru_disconnect =	rts_disconnect,
334 	.pru_peeraddr =		rts_peeraddr,
335 	.pru_send =		rts_send,
336 	.pru_shutdown =		rts_shutdown,
337 	.pru_sockaddr =		rts_sockaddr,
338 	.pru_close =		rts_close,
339 };
340 
341 #ifndef _SOCKADDR_UNION_DEFINED
342 #define	_SOCKADDR_UNION_DEFINED
343 /*
344  * The union of all possible address formats we handle.
345  */
346 union sockaddr_union {
347 	struct sockaddr		sa;
348 	struct sockaddr_in	sin;
349 	struct sockaddr_in6	sin6;
350 };
351 #endif /* _SOCKADDR_UNION_DEFINED */
352 
353 static int
354 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
355     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
356 {
357 
358 	/* First, see if the returned address is part of the jail. */
359 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
360 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
361 		return (0);
362 	}
363 
364 	switch (info->rti_info[RTAX_DST]->sa_family) {
365 #ifdef INET
366 	case AF_INET:
367 	{
368 		struct in_addr ia;
369 		struct ifaddr *ifa;
370 		int found;
371 
372 		found = 0;
373 		/*
374 		 * Try to find an address on the given outgoing interface
375 		 * that belongs to the jail.
376 		 */
377 		IF_ADDR_LOCK(ifp);
378 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
379 			struct sockaddr *sa;
380 			sa = ifa->ifa_addr;
381 			if (sa->sa_family != AF_INET)
382 				continue;
383 			ia = ((struct sockaddr_in *)sa)->sin_addr;
384 			if (prison_check_ip4(cred, &ia) == 0) {
385 				found = 1;
386 				break;
387 			}
388 		}
389 		IF_ADDR_UNLOCK(ifp);
390 		if (!found) {
391 			/*
392 			 * As a last resort return the 'default' jail address.
393 			 */
394 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
395 			    sin_addr;
396 			if (prison_get_ip4(cred, &ia) != 0)
397 				return (ESRCH);
398 		}
399 		bzero(&saun->sin, sizeof(struct sockaddr_in));
400 		saun->sin.sin_len = sizeof(struct sockaddr_in);
401 		saun->sin.sin_family = AF_INET;
402 		saun->sin.sin_addr.s_addr = ia.s_addr;
403 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
404 		break;
405 	}
406 #endif
407 #ifdef INET6
408 	case AF_INET6:
409 	{
410 		struct in6_addr ia6;
411 		struct ifaddr *ifa;
412 		int found;
413 
414 		found = 0;
415 		/*
416 		 * Try to find an address on the given outgoing interface
417 		 * that belongs to the jail.
418 		 */
419 		IF_ADDR_LOCK(ifp);
420 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
421 			struct sockaddr *sa;
422 			sa = ifa->ifa_addr;
423 			if (sa->sa_family != AF_INET6)
424 				continue;
425 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
426 			    &ia6, sizeof(struct in6_addr));
427 			if (prison_check_ip6(cred, &ia6) == 0) {
428 				found = 1;
429 				break;
430 			}
431 		}
432 		IF_ADDR_UNLOCK(ifp);
433 		if (!found) {
434 			/*
435 			 * As a last resort return the 'default' jail address.
436 			 */
437 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
438 			    sin6_addr;
439 			if (prison_get_ip6(cred, &ia6) != 0)
440 				return (ESRCH);
441 		}
442 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
443 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
444 		saun->sin6.sin6_family = AF_INET6;
445 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
446 		if (sa6_recoverscope(&saun->sin6) != 0)
447 			return (ESRCH);
448 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
449 		break;
450 	}
451 #endif
452 	default:
453 		return (ESRCH);
454 	}
455 	return (0);
456 }
457 
458 /*ARGSUSED*/
459 static int
460 route_output(struct mbuf *m, struct socket *so)
461 {
462 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
463 	struct rt_msghdr *rtm = NULL;
464 	struct rtentry *rt = NULL;
465 	struct radix_node_head *rnh;
466 	struct rt_addrinfo info;
467 	int len, error = 0;
468 	struct ifnet *ifp = NULL;
469 	union sockaddr_union saun;
470 
471 #define senderr(e) { error = e; goto flush;}
472 	if (m == NULL || ((m->m_len < sizeof(long)) &&
473 		       (m = m_pullup(m, sizeof(long))) == NULL))
474 		return (ENOBUFS);
475 	if ((m->m_flags & M_PKTHDR) == 0)
476 		panic("route_output");
477 	len = m->m_pkthdr.len;
478 	if (len < sizeof(*rtm) ||
479 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
480 		info.rti_info[RTAX_DST] = NULL;
481 		senderr(EINVAL);
482 	}
483 	R_Malloc(rtm, struct rt_msghdr *, len);
484 	if (rtm == NULL) {
485 		info.rti_info[RTAX_DST] = NULL;
486 		senderr(ENOBUFS);
487 	}
488 	m_copydata(m, 0, len, (caddr_t)rtm);
489 	if (rtm->rtm_version != RTM_VERSION) {
490 		info.rti_info[RTAX_DST] = NULL;
491 		senderr(EPROTONOSUPPORT);
492 	}
493 	rtm->rtm_pid = curproc->p_pid;
494 	bzero(&info, sizeof(info));
495 	info.rti_addrs = rtm->rtm_addrs;
496 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
497 		info.rti_info[RTAX_DST] = NULL;
498 		senderr(EINVAL);
499 	}
500 	info.rti_flags = rtm->rtm_flags;
501 	if (info.rti_info[RTAX_DST] == NULL ||
502 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
503 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
504 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
505 		senderr(EINVAL);
506 	/*
507 	 * Verify that the caller has the appropriate privilege; RTM_GET
508 	 * is the only operation the non-superuser is allowed.
509 	 */
510 	if (rtm->rtm_type != RTM_GET) {
511 		error = priv_check(curthread, PRIV_NET_ROUTE);
512 		if (error)
513 			senderr(error);
514 	}
515 
516 	switch (rtm->rtm_type) {
517 		struct rtentry *saved_nrt;
518 
519 	case RTM_ADD:
520 		if (info.rti_info[RTAX_GATEWAY] == NULL)
521 			senderr(EINVAL);
522 		saved_nrt = NULL;
523 
524 		/* support for new ARP code */
525 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
526 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
527 			error = lla_rt_output(rtm, &info);
528 			break;
529 		}
530 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
531 		    so->so_fibnum);
532 		if (error == 0 && saved_nrt) {
533 			RT_LOCK(saved_nrt);
534 			rt_setmetrics(rtm->rtm_inits,
535 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
536 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
537 			RT_REMREF(saved_nrt);
538 			RT_UNLOCK(saved_nrt);
539 		}
540 		break;
541 
542 	case RTM_DELETE:
543 		saved_nrt = NULL;
544 		/* support for new ARP code */
545 		if (info.rti_info[RTAX_GATEWAY] &&
546 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
547 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
548 			error = lla_rt_output(rtm, &info);
549 			break;
550 		}
551 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
552 		    so->so_fibnum);
553 		if (error == 0) {
554 			RT_LOCK(saved_nrt);
555 			rt = saved_nrt;
556 			goto report;
557 		}
558 		break;
559 
560 	case RTM_GET:
561 	case RTM_CHANGE:
562 	case RTM_LOCK:
563 		rnh = rt_tables_get_rnh(so->so_fibnum,
564 		    info.rti_info[RTAX_DST]->sa_family);
565 		if (rnh == NULL)
566 			senderr(EAFNOSUPPORT);
567 		RADIX_NODE_HEAD_RLOCK(rnh);
568 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
569 			info.rti_info[RTAX_NETMASK], rnh);
570 		if (rt == NULL) {	/* XXX looks bogus */
571 			RADIX_NODE_HEAD_RUNLOCK(rnh);
572 			senderr(ESRCH);
573 		}
574 #ifdef RADIX_MPATH
575 		/*
576 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
577 		 * we require users to specify a matching RTAX_GATEWAY.
578 		 *
579 		 * for RTM_GET, gate is optional even with multipath.
580 		 * if gate == NULL the first match is returned.
581 		 * (no need to call rt_mpath_matchgate if gate == NULL)
582 		 */
583 		if (rn_mpath_capable(rnh) &&
584 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
585 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
586 			if (!rt) {
587 				RADIX_NODE_HEAD_RUNLOCK(rnh);
588 				senderr(ESRCH);
589 			}
590 		}
591 #endif
592 		RT_LOCK(rt);
593 		RT_ADDREF(rt);
594 		RADIX_NODE_HEAD_RUNLOCK(rnh);
595 
596 		/*
597 		 * Fix for PR: 82974
598 		 *
599 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
600 		 * returns a perfect match in case a netmask is
601 		 * specified.  For host routes only a longest prefix
602 		 * match is returned so it is necessary to compare the
603 		 * existence of the netmask.  If both have a netmask
604 		 * rnh_lookup() did a perfect match and if none of them
605 		 * have a netmask both are host routes which is also a
606 		 * perfect match.
607 		 */
608 
609 		if (rtm->rtm_type != RTM_GET &&
610 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
611 			RT_UNLOCK(rt);
612 			senderr(ESRCH);
613 		}
614 
615 		switch(rtm->rtm_type) {
616 
617 		case RTM_GET:
618 		report:
619 			RT_LOCK_ASSERT(rt);
620 			if ((rt->rt_flags & RTF_HOST) == 0
621 			    ? jailed(curthread->td_ucred)
622 			    : prison_if(curthread->td_ucred,
623 			    rt_key(rt)) != 0) {
624 				RT_UNLOCK(rt);
625 				senderr(ESRCH);
626 			}
627 			info.rti_info[RTAX_DST] = rt_key(rt);
628 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
629 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
630 			info.rti_info[RTAX_GENMASK] = 0;
631 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
632 				ifp = rt->rt_ifp;
633 				if (ifp) {
634 					info.rti_info[RTAX_IFP] =
635 					    ifp->if_addr->ifa_addr;
636 					error = rtm_get_jailed(&info, ifp, rt,
637 					    &saun, curthread->td_ucred);
638 					if (error != 0) {
639 						RT_UNLOCK(rt);
640 						senderr(error);
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 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
667 			rtm->rtm_addrs = info.rti_addrs;
668 			break;
669 
670 		case RTM_CHANGE:
671 			/*
672 			 * New gateway could require new ifaddr, ifp;
673 			 * flags may also be different; ifp may be specified
674 			 * by ll sockaddr when protocol address is ambiguous
675 			 */
676 			if (((rt->rt_flags & RTF_GATEWAY) &&
677 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
678 			    info.rti_info[RTAX_IFP] != NULL ||
679 			    (info.rti_info[RTAX_IFA] != NULL &&
680 			     !sa_equal(info.rti_info[RTAX_IFA],
681 				       rt->rt_ifa->ifa_addr))) {
682 				RT_UNLOCK(rt);
683 				RADIX_NODE_HEAD_LOCK(rnh);
684 				error = rt_getifa_fib(&info, rt->rt_fibnum);
685 				/*
686 				 * XXXRW: Really we should release this
687 				 * reference later, but this maintains
688 				 * historical behavior.
689 				 */
690 				if (info.rti_ifa != NULL)
691 					ifa_free(info.rti_ifa);
692 				RADIX_NODE_HEAD_UNLOCK(rnh);
693 				if (error != 0)
694 					senderr(error);
695 				RT_LOCK(rt);
696 			}
697 			if (info.rti_ifa != NULL &&
698 			    info.rti_ifa != rt->rt_ifa &&
699 			    rt->rt_ifa != NULL &&
700 			    rt->rt_ifa->ifa_rtrequest != NULL) {
701 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
702 				    &info);
703 				ifa_free(rt->rt_ifa);
704 			}
705 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
706 				RT_UNLOCK(rt);
707 				RADIX_NODE_HEAD_LOCK(rnh);
708 				RT_LOCK(rt);
709 
710 				error = rt_setgate(rt, rt_key(rt),
711 				    info.rti_info[RTAX_GATEWAY]);
712 				RADIX_NODE_HEAD_UNLOCK(rnh);
713 				if (error != 0) {
714 					RT_UNLOCK(rt);
715 					senderr(error);
716 				}
717 				rt->rt_flags |= RTF_GATEWAY;
718 			}
719 			if (info.rti_ifa != NULL &&
720 			    info.rti_ifa != rt->rt_ifa) {
721 				ifa_ref(info.rti_ifa);
722 				rt->rt_ifa = info.rti_ifa;
723 				rt->rt_ifp = info.rti_ifp;
724 			}
725 			/* Allow some flags to be toggled on change. */
726 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
727 				    (rtm->rtm_flags & RTF_FMASK);
728 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
729 					&rt->rt_rmx);
730 			rtm->rtm_index = rt->rt_ifp->if_index;
731 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
732 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
733 			/* FALLTHROUGH */
734 		case RTM_LOCK:
735 			/* We don't support locks anymore */
736 			break;
737 		}
738 		RT_UNLOCK(rt);
739 		break;
740 
741 	default:
742 		senderr(EOPNOTSUPP);
743 	}
744 
745 flush:
746 	if (rtm) {
747 		if (error)
748 			rtm->rtm_errno = error;
749 		else
750 			rtm->rtm_flags |= RTF_DONE;
751 	}
752 	if (rt)		/* XXX can this be true? */
753 		RTFREE(rt);
754     {
755 	struct rawcb *rp = NULL;
756 	/*
757 	 * Check to see if we don't want our own messages.
758 	 */
759 	if ((so->so_options & SO_USELOOPBACK) == 0) {
760 		if (route_cb.any_count <= 1) {
761 			if (rtm)
762 				Free(rtm);
763 			m_freem(m);
764 			return (error);
765 		}
766 		/* There is another listener, so construct message */
767 		rp = sotorawcb(so);
768 	}
769 	if (rtm) {
770 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
771 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
772 			m_freem(m);
773 			m = NULL;
774 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
775 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
776 		Free(rtm);
777 	}
778 	if (m) {
779 		if (rp) {
780 			/*
781 			 * XXX insure we don't get a copy by
782 			 * invalidating our protocol
783 			 */
784 			unsigned short family = rp->rcb_proto.sp_family;
785 			rp->rcb_proto.sp_family = 0;
786 			rt_dispatch(m, info.rti_info[RTAX_DST]);
787 			rp->rcb_proto.sp_family = family;
788 		} else
789 			rt_dispatch(m, info.rti_info[RTAX_DST]);
790 	}
791     }
792 	return (error);
793 #undef	sa_equal
794 }
795 
796 static void
797 rt_setmetrics(u_long which, const struct rt_metrics *in,
798 	struct rt_metrics_lite *out)
799 {
800 #define metric(f, e) if (which & (f)) out->e = in->e;
801 	/*
802 	 * Only these are stored in the routing entry since introduction
803 	 * of tcp hostcache. The rest is ignored.
804 	 */
805 	metric(RTV_MTU, rmx_mtu);
806 	metric(RTV_WEIGHT, rmx_weight);
807 	/* Userland -> kernel timebase conversion. */
808 	if (which & RTV_EXPIRE)
809 		out->rmx_expire = in->rmx_expire ?
810 		    in->rmx_expire - time_second + time_uptime : 0;
811 #undef metric
812 }
813 
814 static void
815 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
816 {
817 #define metric(e) out->e = in->e;
818 	bzero(out, sizeof(*out));
819 	metric(rmx_mtu);
820 	metric(rmx_weight);
821 	/* Kernel -> userland timebase conversion. */
822 	out->rmx_expire = in->rmx_expire ?
823 	    in->rmx_expire - time_uptime + time_second : 0;
824 #undef metric
825 }
826 
827 /*
828  * Extract the addresses of the passed sockaddrs.
829  * Do a little sanity checking so as to avoid bad memory references.
830  * This data is derived straight from userland.
831  */
832 static int
833 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
834 {
835 	struct sockaddr *sa;
836 	int i;
837 
838 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
839 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
840 			continue;
841 		sa = (struct sockaddr *)cp;
842 		/*
843 		 * It won't fit.
844 		 */
845 		if (cp + sa->sa_len > cplim)
846 			return (EINVAL);
847 		/*
848 		 * there are no more.. quit now
849 		 * If there are more bits, they are in error.
850 		 * I've seen this. route(1) can evidently generate these.
851 		 * This causes kernel to core dump.
852 		 * for compatibility, If we see this, point to a safe address.
853 		 */
854 		if (sa->sa_len == 0) {
855 			rtinfo->rti_info[i] = &sa_zero;
856 			return (0); /* should be EINVAL but for compat */
857 		}
858 		/* accept it */
859 		rtinfo->rti_info[i] = sa;
860 		cp += SA_SIZE(sa);
861 	}
862 	return (0);
863 }
864 
865 static struct mbuf *
866 rt_msg1(int type, struct rt_addrinfo *rtinfo)
867 {
868 	struct rt_msghdr *rtm;
869 	struct mbuf *m;
870 	int i;
871 	struct sockaddr *sa;
872 	int len, dlen;
873 
874 	switch (type) {
875 
876 	case RTM_DELADDR:
877 	case RTM_NEWADDR:
878 		len = sizeof(struct ifa_msghdr);
879 		break;
880 
881 	case RTM_DELMADDR:
882 	case RTM_NEWMADDR:
883 		len = sizeof(struct ifma_msghdr);
884 		break;
885 
886 	case RTM_IFINFO:
887 		len = sizeof(struct if_msghdr);
888 		break;
889 
890 	case RTM_IFANNOUNCE:
891 	case RTM_IEEE80211:
892 		len = sizeof(struct if_announcemsghdr);
893 		break;
894 
895 	default:
896 		len = sizeof(struct rt_msghdr);
897 	}
898 	if (len > MCLBYTES)
899 		panic("rt_msg1");
900 	m = m_gethdr(M_DONTWAIT, MT_DATA);
901 	if (m && len > MHLEN) {
902 		MCLGET(m, M_DONTWAIT);
903 		if ((m->m_flags & M_EXT) == 0) {
904 			m_free(m);
905 			m = NULL;
906 		}
907 	}
908 	if (m == NULL)
909 		return (m);
910 	m->m_pkthdr.len = m->m_len = len;
911 	m->m_pkthdr.rcvif = NULL;
912 	rtm = mtod(m, struct rt_msghdr *);
913 	bzero((caddr_t)rtm, len);
914 	for (i = 0; i < RTAX_MAX; i++) {
915 		if ((sa = rtinfo->rti_info[i]) == NULL)
916 			continue;
917 		rtinfo->rti_addrs |= (1 << i);
918 		dlen = SA_SIZE(sa);
919 		m_copyback(m, len, dlen, (caddr_t)sa);
920 		len += dlen;
921 	}
922 	if (m->m_pkthdr.len != len) {
923 		m_freem(m);
924 		return (NULL);
925 	}
926 	rtm->rtm_msglen = len;
927 	rtm->rtm_version = RTM_VERSION;
928 	rtm->rtm_type = type;
929 	return (m);
930 }
931 
932 static int
933 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
934 {
935 	int i;
936 	int len, dlen, second_time = 0;
937 	caddr_t cp0;
938 
939 	rtinfo->rti_addrs = 0;
940 again:
941 	switch (type) {
942 
943 	case RTM_DELADDR:
944 	case RTM_NEWADDR:
945 		len = sizeof(struct ifa_msghdr);
946 		break;
947 
948 	case RTM_IFINFO:
949 		len = sizeof(struct if_msghdr);
950 		break;
951 
952 	case RTM_NEWMADDR:
953 		len = sizeof(struct ifma_msghdr);
954 		break;
955 
956 	default:
957 		len = sizeof(struct rt_msghdr);
958 	}
959 	cp0 = cp;
960 	if (cp0)
961 		cp += len;
962 	for (i = 0; i < RTAX_MAX; i++) {
963 		struct sockaddr *sa;
964 
965 		if ((sa = rtinfo->rti_info[i]) == NULL)
966 			continue;
967 		rtinfo->rti_addrs |= (1 << i);
968 		dlen = SA_SIZE(sa);
969 		if (cp) {
970 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
971 			cp += dlen;
972 		}
973 		len += dlen;
974 	}
975 	len = ALIGN(len);
976 	if (cp == NULL && w != NULL && !second_time) {
977 		struct walkarg *rw = w;
978 
979 		if (rw->w_req) {
980 			if (rw->w_tmemsize < len) {
981 				if (rw->w_tmem)
982 					free(rw->w_tmem, M_RTABLE);
983 				rw->w_tmem = (caddr_t)
984 					malloc(len, M_RTABLE, M_NOWAIT);
985 				if (rw->w_tmem)
986 					rw->w_tmemsize = len;
987 			}
988 			if (rw->w_tmem) {
989 				cp = rw->w_tmem;
990 				second_time = 1;
991 				goto again;
992 			}
993 		}
994 	}
995 	if (cp) {
996 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
997 
998 		rtm->rtm_version = RTM_VERSION;
999 		rtm->rtm_type = type;
1000 		rtm->rtm_msglen = len;
1001 	}
1002 	return (len);
1003 }
1004 
1005 /*
1006  * This routine is called to generate a message from the routing
1007  * socket indicating that a redirect has occured, a routing lookup
1008  * has failed, or that a protocol has detected timeouts to a particular
1009  * destination.
1010  */
1011 void
1012 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1013 {
1014 	struct rt_msghdr *rtm;
1015 	struct mbuf *m;
1016 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1017 
1018 	if (route_cb.any_count == 0)
1019 		return;
1020 	m = rt_msg1(type, rtinfo);
1021 	if (m == NULL)
1022 		return;
1023 	rtm = mtod(m, struct rt_msghdr *);
1024 	rtm->rtm_flags = RTF_DONE | flags;
1025 	rtm->rtm_errno = error;
1026 	rtm->rtm_addrs = rtinfo->rti_addrs;
1027 	rt_dispatch(m, sa);
1028 }
1029 
1030 /*
1031  * This routine is called to generate a message from the routing
1032  * socket indicating that the status of a network interface has changed.
1033  */
1034 void
1035 rt_ifmsg(struct ifnet *ifp)
1036 {
1037 	struct if_msghdr *ifm;
1038 	struct mbuf *m;
1039 	struct rt_addrinfo info;
1040 
1041 	if (route_cb.any_count == 0)
1042 		return;
1043 	bzero((caddr_t)&info, sizeof(info));
1044 	m = rt_msg1(RTM_IFINFO, &info);
1045 	if (m == NULL)
1046 		return;
1047 	ifm = mtod(m, struct if_msghdr *);
1048 	ifm->ifm_index = ifp->if_index;
1049 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1050 	ifm->ifm_data = ifp->if_data;
1051 	ifm->ifm_addrs = 0;
1052 	rt_dispatch(m, NULL);
1053 }
1054 
1055 /*
1056  * This is called to generate messages from the routing socket
1057  * indicating a network interface has had addresses associated with it.
1058  * if we ever reverse the logic and replace messages TO the routing
1059  * socket indicate a request to configure interfaces, then it will
1060  * be unnecessary as the routing socket will automatically generate
1061  * copies of it.
1062  */
1063 void
1064 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1065 {
1066 	struct rt_addrinfo info;
1067 	struct sockaddr *sa = NULL;
1068 	int pass;
1069 	struct mbuf *m = NULL;
1070 	struct ifnet *ifp = ifa->ifa_ifp;
1071 
1072 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1073 		("unexpected cmd %u", cmd));
1074 #if defined(INET) || defined(INET6)
1075 #ifdef SCTP
1076 	/*
1077 	 * notify the SCTP stack
1078 	 * this will only get called when an address is added/deleted
1079 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1080 	 */
1081 	sctp_addr_change(ifa, cmd);
1082 #endif /* SCTP */
1083 #endif
1084 	if (route_cb.any_count == 0)
1085 		return;
1086 	for (pass = 1; pass < 3; pass++) {
1087 		bzero((caddr_t)&info, sizeof(info));
1088 		if ((cmd == RTM_ADD && pass == 1) ||
1089 		    (cmd == RTM_DELETE && pass == 2)) {
1090 			struct ifa_msghdr *ifam;
1091 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1092 
1093 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1094 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1095 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1096 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1097 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1098 				continue;
1099 			ifam = mtod(m, struct ifa_msghdr *);
1100 			ifam->ifam_index = ifp->if_index;
1101 			ifam->ifam_metric = ifa->ifa_metric;
1102 			ifam->ifam_flags = ifa->ifa_flags;
1103 			ifam->ifam_addrs = info.rti_addrs;
1104 		}
1105 		if ((cmd == RTM_ADD && pass == 2) ||
1106 		    (cmd == RTM_DELETE && pass == 1)) {
1107 			struct rt_msghdr *rtm;
1108 
1109 			if (rt == NULL)
1110 				continue;
1111 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1112 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1113 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1114 			if ((m = rt_msg1(cmd, &info)) == NULL)
1115 				continue;
1116 			rtm = mtod(m, struct rt_msghdr *);
1117 			rtm->rtm_index = ifp->if_index;
1118 			rtm->rtm_flags |= rt->rt_flags;
1119 			rtm->rtm_errno = error;
1120 			rtm->rtm_addrs = info.rti_addrs;
1121 		}
1122 		rt_dispatch(m, sa);
1123 	}
1124 }
1125 
1126 /*
1127  * This is the analogue to the rt_newaddrmsg which performs the same
1128  * function but for multicast group memberhips.  This is easier since
1129  * there is no route state to worry about.
1130  */
1131 void
1132 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1133 {
1134 	struct rt_addrinfo info;
1135 	struct mbuf *m = NULL;
1136 	struct ifnet *ifp = ifma->ifma_ifp;
1137 	struct ifma_msghdr *ifmam;
1138 
1139 	if (route_cb.any_count == 0)
1140 		return;
1141 
1142 	bzero((caddr_t)&info, sizeof(info));
1143 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1144 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1145 	/*
1146 	 * If a link-layer address is present, present it as a ``gateway''
1147 	 * (similarly to how ARP entries, e.g., are presented).
1148 	 */
1149 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1150 	m = rt_msg1(cmd, &info);
1151 	if (m == NULL)
1152 		return;
1153 	ifmam = mtod(m, struct ifma_msghdr *);
1154 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1155 	    __func__));
1156 	ifmam->ifmam_index = ifp->if_index;
1157 	ifmam->ifmam_addrs = info.rti_addrs;
1158 	rt_dispatch(m, ifma->ifma_addr);
1159 }
1160 
1161 static struct mbuf *
1162 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1163 	struct rt_addrinfo *info)
1164 {
1165 	struct if_announcemsghdr *ifan;
1166 	struct mbuf *m;
1167 
1168 	if (route_cb.any_count == 0)
1169 		return NULL;
1170 	bzero((caddr_t)info, sizeof(*info));
1171 	m = rt_msg1(type, info);
1172 	if (m != NULL) {
1173 		ifan = mtod(m, struct if_announcemsghdr *);
1174 		ifan->ifan_index = ifp->if_index;
1175 		strlcpy(ifan->ifan_name, ifp->if_xname,
1176 			sizeof(ifan->ifan_name));
1177 		ifan->ifan_what = what;
1178 	}
1179 	return m;
1180 }
1181 
1182 /*
1183  * This is called to generate routing socket messages indicating
1184  * IEEE80211 wireless events.
1185  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1186  */
1187 void
1188 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1189 {
1190 	struct mbuf *m;
1191 	struct rt_addrinfo info;
1192 
1193 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1194 	if (m != NULL) {
1195 		/*
1196 		 * Append the ieee80211 data.  Try to stick it in the
1197 		 * mbuf containing the ifannounce msg; otherwise allocate
1198 		 * a new mbuf and append.
1199 		 *
1200 		 * NB: we assume m is a single mbuf.
1201 		 */
1202 		if (data_len > M_TRAILINGSPACE(m)) {
1203 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1204 			if (n == NULL) {
1205 				m_freem(m);
1206 				return;
1207 			}
1208 			bcopy(data, mtod(n, void *), data_len);
1209 			n->m_len = data_len;
1210 			m->m_next = n;
1211 		} else if (data_len > 0) {
1212 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1213 			m->m_len += data_len;
1214 		}
1215 		if (m->m_flags & M_PKTHDR)
1216 			m->m_pkthdr.len += data_len;
1217 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1218 		rt_dispatch(m, NULL);
1219 	}
1220 }
1221 
1222 /*
1223  * This is called to generate routing socket messages indicating
1224  * network interface arrival and departure.
1225  */
1226 void
1227 rt_ifannouncemsg(struct ifnet *ifp, int what)
1228 {
1229 	struct mbuf *m;
1230 	struct rt_addrinfo info;
1231 
1232 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1233 	if (m != NULL)
1234 		rt_dispatch(m, NULL);
1235 }
1236 
1237 static void
1238 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1239 {
1240 	struct m_tag *tag;
1241 
1242 	/*
1243 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1244 	 * use when injecting the mbuf into the routing socket buffer from
1245 	 * the netisr.
1246 	 */
1247 	if (sa != NULL) {
1248 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1249 		    M_NOWAIT);
1250 		if (tag == NULL) {
1251 			m_freem(m);
1252 			return;
1253 		}
1254 		*(unsigned short *)(tag + 1) = sa->sa_family;
1255 		m_tag_prepend(m, tag);
1256 	}
1257 #ifdef VIMAGE
1258 	if (V_loif)
1259 		m->m_pkthdr.rcvif = V_loif;
1260 	else {
1261 		m_freem(m);
1262 		return;
1263 	}
1264 #endif
1265 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1266 }
1267 
1268 /*
1269  * This is used in dumping the kernel table via sysctl().
1270  */
1271 static int
1272 sysctl_dumpentry(struct radix_node *rn, void *vw)
1273 {
1274 	struct walkarg *w = vw;
1275 	struct rtentry *rt = (struct rtentry *)rn;
1276 	int error = 0, size;
1277 	struct rt_addrinfo info;
1278 
1279 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1280 		return 0;
1281 	if ((rt->rt_flags & RTF_HOST) == 0
1282 	    ? jailed(w->w_req->td->td_ucred)
1283 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1284 		return (0);
1285 	bzero((caddr_t)&info, sizeof(info));
1286 	info.rti_info[RTAX_DST] = rt_key(rt);
1287 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1288 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1289 	info.rti_info[RTAX_GENMASK] = 0;
1290 	if (rt->rt_ifp) {
1291 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1292 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1293 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1294 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1295 	}
1296 	size = rt_msg2(RTM_GET, &info, NULL, w);
1297 	if (w->w_req && w->w_tmem) {
1298 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1299 
1300 		rtm->rtm_flags = rt->rt_flags;
1301 		/*
1302 		 * let's be honest about this being a retarded hack
1303 		 */
1304 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1305 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1306 		rtm->rtm_index = rt->rt_ifp->if_index;
1307 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1308 		rtm->rtm_addrs = info.rti_addrs;
1309 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1310 		return (error);
1311 	}
1312 	return (error);
1313 }
1314 
1315 static int
1316 sysctl_iflist(int af, struct walkarg *w)
1317 {
1318 	struct ifnet *ifp;
1319 	struct ifaddr *ifa;
1320 	struct rt_addrinfo info;
1321 	int len, error = 0;
1322 
1323 	bzero((caddr_t)&info, sizeof(info));
1324 	IFNET_RLOCK();
1325 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1326 		if (w->w_arg && w->w_arg != ifp->if_index)
1327 			continue;
1328 		ifa = ifp->if_addr;
1329 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1330 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1331 		info.rti_info[RTAX_IFP] = NULL;
1332 		if (w->w_req && w->w_tmem) {
1333 			struct if_msghdr *ifm;
1334 
1335 			ifm = (struct if_msghdr *)w->w_tmem;
1336 			ifm->ifm_index = ifp->if_index;
1337 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1338 			ifm->ifm_data = ifp->if_data;
1339 			ifm->ifm_addrs = info.rti_addrs;
1340 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
1341 			if (error)
1342 				goto done;
1343 		}
1344 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1345 			if (af && af != ifa->ifa_addr->sa_family)
1346 				continue;
1347 			if (prison_if(w->w_req->td->td_ucred,
1348 			    ifa->ifa_addr) != 0)
1349 				continue;
1350 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1351 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1352 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1353 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1354 			if (w->w_req && w->w_tmem) {
1355 				struct ifa_msghdr *ifam;
1356 
1357 				ifam = (struct ifa_msghdr *)w->w_tmem;
1358 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1359 				ifam->ifam_flags = ifa->ifa_flags;
1360 				ifam->ifam_metric = ifa->ifa_metric;
1361 				ifam->ifam_addrs = info.rti_addrs;
1362 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1363 				if (error)
1364 					goto done;
1365 			}
1366 		}
1367 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1368 			info.rti_info[RTAX_BRD] = NULL;
1369 	}
1370 done:
1371 	IFNET_RUNLOCK();
1372 	return (error);
1373 }
1374 
1375 static int
1376 sysctl_ifmalist(int af, struct walkarg *w)
1377 {
1378 	struct ifnet *ifp;
1379 	struct ifmultiaddr *ifma;
1380 	struct	rt_addrinfo info;
1381 	int	len, error = 0;
1382 	struct ifaddr *ifa;
1383 
1384 	bzero((caddr_t)&info, sizeof(info));
1385 	IFNET_RLOCK();
1386 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1387 		if (w->w_arg && w->w_arg != ifp->if_index)
1388 			continue;
1389 		ifa = ifp->if_addr;
1390 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1391 		IF_ADDR_LOCK(ifp);
1392 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1393 			if (af && af != ifma->ifma_addr->sa_family)
1394 				continue;
1395 			if (prison_if(w->w_req->td->td_ucred,
1396 			    ifma->ifma_addr) != 0)
1397 				continue;
1398 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1399 			info.rti_info[RTAX_GATEWAY] =
1400 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1401 			    ifma->ifma_lladdr : NULL;
1402 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1403 			if (w->w_req && w->w_tmem) {
1404 				struct ifma_msghdr *ifmam;
1405 
1406 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1407 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1408 				ifmam->ifmam_flags = 0;
1409 				ifmam->ifmam_addrs = info.rti_addrs;
1410 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1411 				if (error) {
1412 					IF_ADDR_UNLOCK(ifp);
1413 					goto done;
1414 				}
1415 			}
1416 		}
1417 		IF_ADDR_UNLOCK(ifp);
1418 	}
1419 done:
1420 	IFNET_RUNLOCK();
1421 	return (error);
1422 }
1423 
1424 static int
1425 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1426 {
1427 	int	*name = (int *)arg1;
1428 	u_int	namelen = arg2;
1429 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1430 	int	i, lim, error = EINVAL;
1431 	u_char	af;
1432 	struct	walkarg w;
1433 
1434 	name ++;
1435 	namelen--;
1436 	if (req->newptr)
1437 		return (EPERM);
1438 	if (namelen != 3)
1439 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1440 	af = name[0];
1441 	if (af > AF_MAX)
1442 		return (EINVAL);
1443 	bzero(&w, sizeof(w));
1444 	w.w_op = name[1];
1445 	w.w_arg = name[2];
1446 	w.w_req = req;
1447 
1448 	error = sysctl_wire_old_buffer(req, 0);
1449 	if (error)
1450 		return (error);
1451 	switch (w.w_op) {
1452 
1453 	case NET_RT_DUMP:
1454 	case NET_RT_FLAGS:
1455 		if (af == 0) {			/* dump all tables */
1456 			i = 1;
1457 			lim = AF_MAX;
1458 		} else				/* dump only one table */
1459 			i = lim = af;
1460 
1461 		/*
1462 		 * take care of llinfo entries, the caller must
1463 		 * specify an AF
1464 		 */
1465 		if (w.w_op == NET_RT_FLAGS &&
1466 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1467 			if (af != 0)
1468 				error = lltable_sysctl_dumparp(af, w.w_req);
1469 			else
1470 				error = EINVAL;
1471 			break;
1472 		}
1473 		/*
1474 		 * take care of routing entries
1475 		 */
1476 		for (error = 0; error == 0 && i <= lim; i++)
1477 			rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1478 			if (rnh != NULL) {
1479 				RADIX_NODE_HEAD_LOCK(rnh);
1480 			    	error = rnh->rnh_walktree(rnh,
1481 				    sysctl_dumpentry, &w);
1482 				RADIX_NODE_HEAD_UNLOCK(rnh);
1483 			} else if (af != 0)
1484 				error = EAFNOSUPPORT;
1485 		break;
1486 
1487 	case NET_RT_IFLIST:
1488 		error = sysctl_iflist(af, &w);
1489 		break;
1490 
1491 	case NET_RT_IFMALIST:
1492 		error = sysctl_ifmalist(af, &w);
1493 		break;
1494 	}
1495 	if (w.w_tmem)
1496 		free(w.w_tmem, M_RTABLE);
1497 	return (error);
1498 }
1499 
1500 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1501 
1502 /*
1503  * Definitions of protocols supported in the ROUTE domain.
1504  */
1505 
1506 static struct domain routedomain;		/* or at least forward */
1507 
1508 static struct protosw routesw[] = {
1509 {
1510 	.pr_type =		SOCK_RAW,
1511 	.pr_domain =		&routedomain,
1512 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1513 	.pr_output =		route_output,
1514 	.pr_ctlinput =		raw_ctlinput,
1515 	.pr_init =		raw_init,
1516 	.pr_usrreqs =		&route_usrreqs
1517 }
1518 };
1519 
1520 static struct domain routedomain = {
1521 	.dom_family =		PF_ROUTE,
1522 	.dom_name =		 "route",
1523 	.dom_protosw =		routesw,
1524 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1525 };
1526 
1527 VNET_DOMAIN_SET(route);
1528