xref: /freebsd/sys/net/rtsock.c (revision eb6d21b4ca6d668cf89afd99eef7baeafa712197)
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 	/*
517 	 * The given gateway address may be an interface address.
518 	 * For example, issuing a "route change" command on a route
519 	 * entry that was created from a tunnel, and the gateway
520 	 * address given is the local end point. In this case the
521 	 * RTF_GATEWAY flag must be cleared or the destination will
522 	 * not be reachable even though there is no error message.
523 	 */
524 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
525 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
526 		struct route gw_ro;
527 
528 		bzero(&gw_ro, sizeof(gw_ro));
529 		gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
530 		rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
531 		/*
532 		 * A host route through the loopback interface is
533 		 * installed for each interface adddress. In pre 8.0
534 		 * releases the interface address of a PPP link type
535 		 * is not reachable locally. This behavior is fixed as
536 		 * part of the new L2/L3 redesign and rewrite work. The
537 		 * signature of this interface address route is the
538 		 * AF_LINK sa_family type of the rt_gateway, and the
539 		 * rt_ifp has the IFF_LOOPBACK flag set.
540 		 */
541 		if (gw_ro.ro_rt != NULL &&
542 		    gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
543 		    gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)
544 			info.rti_flags &= ~RTF_GATEWAY;
545 		if (gw_ro.ro_rt != NULL)
546 			RTFREE(gw_ro.ro_rt);
547 	}
548 
549 	switch (rtm->rtm_type) {
550 		struct rtentry *saved_nrt;
551 
552 	case RTM_ADD:
553 		if (info.rti_info[RTAX_GATEWAY] == NULL)
554 			senderr(EINVAL);
555 		saved_nrt = NULL;
556 
557 		/* support for new ARP code */
558 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
559 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
560 			error = lla_rt_output(rtm, &info);
561 			break;
562 		}
563 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
564 		    so->so_fibnum);
565 		if (error == 0 && saved_nrt) {
566 			RT_LOCK(saved_nrt);
567 			rt_setmetrics(rtm->rtm_inits,
568 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
569 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
570 			RT_REMREF(saved_nrt);
571 			RT_UNLOCK(saved_nrt);
572 		}
573 		break;
574 
575 	case RTM_DELETE:
576 		saved_nrt = NULL;
577 		/* support for new ARP code */
578 		if (info.rti_info[RTAX_GATEWAY] &&
579 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
580 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
581 			error = lla_rt_output(rtm, &info);
582 			break;
583 		}
584 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
585 		    so->so_fibnum);
586 		if (error == 0) {
587 			RT_LOCK(saved_nrt);
588 			rt = saved_nrt;
589 			goto report;
590 		}
591 		break;
592 
593 	case RTM_GET:
594 	case RTM_CHANGE:
595 	case RTM_LOCK:
596 		rnh = rt_tables_get_rnh(so->so_fibnum,
597 		    info.rti_info[RTAX_DST]->sa_family);
598 		if (rnh == NULL)
599 			senderr(EAFNOSUPPORT);
600 		RADIX_NODE_HEAD_RLOCK(rnh);
601 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
602 			info.rti_info[RTAX_NETMASK], rnh);
603 		if (rt == NULL) {	/* XXX looks bogus */
604 			RADIX_NODE_HEAD_RUNLOCK(rnh);
605 			senderr(ESRCH);
606 		}
607 #ifdef RADIX_MPATH
608 		/*
609 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
610 		 * we require users to specify a matching RTAX_GATEWAY.
611 		 *
612 		 * for RTM_GET, gate is optional even with multipath.
613 		 * if gate == NULL the first match is returned.
614 		 * (no need to call rt_mpath_matchgate if gate == NULL)
615 		 */
616 		if (rn_mpath_capable(rnh) &&
617 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
618 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
619 			if (!rt) {
620 				RADIX_NODE_HEAD_RUNLOCK(rnh);
621 				senderr(ESRCH);
622 			}
623 		}
624 #endif
625 		RT_LOCK(rt);
626 		RT_ADDREF(rt);
627 		RADIX_NODE_HEAD_RUNLOCK(rnh);
628 
629 		/*
630 		 * Fix for PR: 82974
631 		 *
632 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
633 		 * returns a perfect match in case a netmask is
634 		 * specified.  For host routes only a longest prefix
635 		 * match is returned so it is necessary to compare the
636 		 * existence of the netmask.  If both have a netmask
637 		 * rnh_lookup() did a perfect match and if none of them
638 		 * have a netmask both are host routes which is also a
639 		 * perfect match.
640 		 */
641 
642 		if (rtm->rtm_type != RTM_GET &&
643 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
644 			RT_UNLOCK(rt);
645 			senderr(ESRCH);
646 		}
647 
648 		switch(rtm->rtm_type) {
649 
650 		case RTM_GET:
651 		report:
652 			RT_LOCK_ASSERT(rt);
653 			if ((rt->rt_flags & RTF_HOST) == 0
654 			    ? jailed_without_vnet(curthread->td_ucred)
655 			    : prison_if(curthread->td_ucred,
656 			    rt_key(rt)) != 0) {
657 				RT_UNLOCK(rt);
658 				senderr(ESRCH);
659 			}
660 			info.rti_info[RTAX_DST] = rt_key(rt);
661 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
662 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
663 			info.rti_info[RTAX_GENMASK] = 0;
664 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
665 				ifp = rt->rt_ifp;
666 				if (ifp) {
667 					info.rti_info[RTAX_IFP] =
668 					    ifp->if_addr->ifa_addr;
669 					error = rtm_get_jailed(&info, ifp, rt,
670 					    &saun, curthread->td_ucred);
671 					if (error != 0) {
672 						RT_UNLOCK(rt);
673 						senderr(error);
674 					}
675 					if (ifp->if_flags & IFF_POINTOPOINT)
676 						info.rti_info[RTAX_BRD] =
677 						    rt->rt_ifa->ifa_dstaddr;
678 					rtm->rtm_index = ifp->if_index;
679 				} else {
680 					info.rti_info[RTAX_IFP] = NULL;
681 					info.rti_info[RTAX_IFA] = NULL;
682 				}
683 			} else if ((ifp = rt->rt_ifp) != NULL) {
684 				rtm->rtm_index = ifp->if_index;
685 			}
686 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
687 			if (len > rtm->rtm_msglen) {
688 				struct rt_msghdr *new_rtm;
689 				R_Malloc(new_rtm, struct rt_msghdr *, len);
690 				if (new_rtm == NULL) {
691 					RT_UNLOCK(rt);
692 					senderr(ENOBUFS);
693 				}
694 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
695 				Free(rtm); rtm = new_rtm;
696 			}
697 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
698 			rtm->rtm_flags = rt->rt_flags;
699 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
700 			rtm->rtm_addrs = info.rti_addrs;
701 			break;
702 
703 		case RTM_CHANGE:
704 			/*
705 			 * New gateway could require new ifaddr, ifp;
706 			 * flags may also be different; ifp may be specified
707 			 * by ll sockaddr when protocol address is ambiguous
708 			 */
709 			if (((rt->rt_flags & RTF_GATEWAY) &&
710 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
711 			    info.rti_info[RTAX_IFP] != NULL ||
712 			    (info.rti_info[RTAX_IFA] != NULL &&
713 			     !sa_equal(info.rti_info[RTAX_IFA],
714 				       rt->rt_ifa->ifa_addr))) {
715 				RT_UNLOCK(rt);
716 				RADIX_NODE_HEAD_LOCK(rnh);
717 				error = rt_getifa_fib(&info, rt->rt_fibnum);
718 				/*
719 				 * XXXRW: Really we should release this
720 				 * reference later, but this maintains
721 				 * historical behavior.
722 				 */
723 				if (info.rti_ifa != NULL)
724 					ifa_free(info.rti_ifa);
725 				RADIX_NODE_HEAD_UNLOCK(rnh);
726 				if (error != 0)
727 					senderr(error);
728 				RT_LOCK(rt);
729 			}
730 			if (info.rti_ifa != NULL &&
731 			    info.rti_ifa != rt->rt_ifa &&
732 			    rt->rt_ifa != NULL &&
733 			    rt->rt_ifa->ifa_rtrequest != NULL) {
734 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
735 				    &info);
736 				ifa_free(rt->rt_ifa);
737 			}
738 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
739 				RT_UNLOCK(rt);
740 				RADIX_NODE_HEAD_LOCK(rnh);
741 				RT_LOCK(rt);
742 
743 				error = rt_setgate(rt, rt_key(rt),
744 				    info.rti_info[RTAX_GATEWAY]);
745 				RADIX_NODE_HEAD_UNLOCK(rnh);
746 				if (error != 0) {
747 					RT_UNLOCK(rt);
748 					senderr(error);
749 				}
750 				rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
751 			}
752 			if (info.rti_ifa != NULL &&
753 			    info.rti_ifa != rt->rt_ifa) {
754 				ifa_ref(info.rti_ifa);
755 				rt->rt_ifa = info.rti_ifa;
756 				rt->rt_ifp = info.rti_ifp;
757 			}
758 			/* Allow some flags to be toggled on change. */
759 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
760 				    (rtm->rtm_flags & RTF_FMASK);
761 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
762 					&rt->rt_rmx);
763 			rtm->rtm_index = rt->rt_ifp->if_index;
764 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
765 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
766 			/* FALLTHROUGH */
767 		case RTM_LOCK:
768 			/* We don't support locks anymore */
769 			break;
770 		}
771 		RT_UNLOCK(rt);
772 		break;
773 
774 	default:
775 		senderr(EOPNOTSUPP);
776 	}
777 
778 flush:
779 	if (rtm) {
780 		if (error)
781 			rtm->rtm_errno = error;
782 		else
783 			rtm->rtm_flags |= RTF_DONE;
784 	}
785 	if (rt)		/* XXX can this be true? */
786 		RTFREE(rt);
787     {
788 	struct rawcb *rp = NULL;
789 	/*
790 	 * Check to see if we don't want our own messages.
791 	 */
792 	if ((so->so_options & SO_USELOOPBACK) == 0) {
793 		if (route_cb.any_count <= 1) {
794 			if (rtm)
795 				Free(rtm);
796 			m_freem(m);
797 			return (error);
798 		}
799 		/* There is another listener, so construct message */
800 		rp = sotorawcb(so);
801 	}
802 	if (rtm) {
803 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
804 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
805 			m_freem(m);
806 			m = NULL;
807 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
808 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
809 		Free(rtm);
810 	}
811 	if (m) {
812 		if (rp) {
813 			/*
814 			 * XXX insure we don't get a copy by
815 			 * invalidating our protocol
816 			 */
817 			unsigned short family = rp->rcb_proto.sp_family;
818 			rp->rcb_proto.sp_family = 0;
819 			rt_dispatch(m, info.rti_info[RTAX_DST]);
820 			rp->rcb_proto.sp_family = family;
821 		} else
822 			rt_dispatch(m, info.rti_info[RTAX_DST]);
823 	}
824     }
825 	return (error);
826 #undef	sa_equal
827 }
828 
829 static void
830 rt_setmetrics(u_long which, const struct rt_metrics *in,
831 	struct rt_metrics_lite *out)
832 {
833 #define metric(f, e) if (which & (f)) out->e = in->e;
834 	/*
835 	 * Only these are stored in the routing entry since introduction
836 	 * of tcp hostcache. The rest is ignored.
837 	 */
838 	metric(RTV_MTU, rmx_mtu);
839 	metric(RTV_WEIGHT, rmx_weight);
840 	/* Userland -> kernel timebase conversion. */
841 	if (which & RTV_EXPIRE)
842 		out->rmx_expire = in->rmx_expire ?
843 		    in->rmx_expire - time_second + time_uptime : 0;
844 #undef metric
845 }
846 
847 static void
848 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
849 {
850 #define metric(e) out->e = in->e;
851 	bzero(out, sizeof(*out));
852 	metric(rmx_mtu);
853 	metric(rmx_weight);
854 	/* Kernel -> userland timebase conversion. */
855 	out->rmx_expire = in->rmx_expire ?
856 	    in->rmx_expire - time_uptime + time_second : 0;
857 #undef metric
858 }
859 
860 /*
861  * Extract the addresses of the passed sockaddrs.
862  * Do a little sanity checking so as to avoid bad memory references.
863  * This data is derived straight from userland.
864  */
865 static int
866 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
867 {
868 	struct sockaddr *sa;
869 	int i;
870 
871 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
872 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
873 			continue;
874 		sa = (struct sockaddr *)cp;
875 		/*
876 		 * It won't fit.
877 		 */
878 		if (cp + sa->sa_len > cplim)
879 			return (EINVAL);
880 		/*
881 		 * there are no more.. quit now
882 		 * If there are more bits, they are in error.
883 		 * I've seen this. route(1) can evidently generate these.
884 		 * This causes kernel to core dump.
885 		 * for compatibility, If we see this, point to a safe address.
886 		 */
887 		if (sa->sa_len == 0) {
888 			rtinfo->rti_info[i] = &sa_zero;
889 			return (0); /* should be EINVAL but for compat */
890 		}
891 		/* accept it */
892 		rtinfo->rti_info[i] = sa;
893 		cp += SA_SIZE(sa);
894 	}
895 	return (0);
896 }
897 
898 static struct mbuf *
899 rt_msg1(int type, struct rt_addrinfo *rtinfo)
900 {
901 	struct rt_msghdr *rtm;
902 	struct mbuf *m;
903 	int i;
904 	struct sockaddr *sa;
905 	int len, dlen;
906 
907 	switch (type) {
908 
909 	case RTM_DELADDR:
910 	case RTM_NEWADDR:
911 		len = sizeof(struct ifa_msghdr);
912 		break;
913 
914 	case RTM_DELMADDR:
915 	case RTM_NEWMADDR:
916 		len = sizeof(struct ifma_msghdr);
917 		break;
918 
919 	case RTM_IFINFO:
920 		len = sizeof(struct if_msghdr);
921 		break;
922 
923 	case RTM_IFANNOUNCE:
924 	case RTM_IEEE80211:
925 		len = sizeof(struct if_announcemsghdr);
926 		break;
927 
928 	default:
929 		len = sizeof(struct rt_msghdr);
930 	}
931 	if (len > MCLBYTES)
932 		panic("rt_msg1");
933 	m = m_gethdr(M_DONTWAIT, MT_DATA);
934 	if (m && len > MHLEN) {
935 		MCLGET(m, M_DONTWAIT);
936 		if ((m->m_flags & M_EXT) == 0) {
937 			m_free(m);
938 			m = NULL;
939 		}
940 	}
941 	if (m == NULL)
942 		return (m);
943 	m->m_pkthdr.len = m->m_len = len;
944 	m->m_pkthdr.rcvif = NULL;
945 	rtm = mtod(m, struct rt_msghdr *);
946 	bzero((caddr_t)rtm, len);
947 	for (i = 0; i < RTAX_MAX; i++) {
948 		if ((sa = rtinfo->rti_info[i]) == NULL)
949 			continue;
950 		rtinfo->rti_addrs |= (1 << i);
951 		dlen = SA_SIZE(sa);
952 		m_copyback(m, len, dlen, (caddr_t)sa);
953 		len += dlen;
954 	}
955 	if (m->m_pkthdr.len != len) {
956 		m_freem(m);
957 		return (NULL);
958 	}
959 	rtm->rtm_msglen = len;
960 	rtm->rtm_version = RTM_VERSION;
961 	rtm->rtm_type = type;
962 	return (m);
963 }
964 
965 static int
966 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
967 {
968 	int i;
969 	int len, dlen, second_time = 0;
970 	caddr_t cp0;
971 
972 	rtinfo->rti_addrs = 0;
973 again:
974 	switch (type) {
975 
976 	case RTM_DELADDR:
977 	case RTM_NEWADDR:
978 		len = sizeof(struct ifa_msghdr);
979 		break;
980 
981 	case RTM_IFINFO:
982 		len = sizeof(struct if_msghdr);
983 		break;
984 
985 	case RTM_NEWMADDR:
986 		len = sizeof(struct ifma_msghdr);
987 		break;
988 
989 	default:
990 		len = sizeof(struct rt_msghdr);
991 	}
992 	cp0 = cp;
993 	if (cp0)
994 		cp += len;
995 	for (i = 0; i < RTAX_MAX; i++) {
996 		struct sockaddr *sa;
997 
998 		if ((sa = rtinfo->rti_info[i]) == NULL)
999 			continue;
1000 		rtinfo->rti_addrs |= (1 << i);
1001 		dlen = SA_SIZE(sa);
1002 		if (cp) {
1003 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1004 			cp += dlen;
1005 		}
1006 		len += dlen;
1007 	}
1008 	len = ALIGN(len);
1009 	if (cp == NULL && w != NULL && !second_time) {
1010 		struct walkarg *rw = w;
1011 
1012 		if (rw->w_req) {
1013 			if (rw->w_tmemsize < len) {
1014 				if (rw->w_tmem)
1015 					free(rw->w_tmem, M_RTABLE);
1016 				rw->w_tmem = (caddr_t)
1017 					malloc(len, M_RTABLE, M_NOWAIT);
1018 				if (rw->w_tmem)
1019 					rw->w_tmemsize = len;
1020 			}
1021 			if (rw->w_tmem) {
1022 				cp = rw->w_tmem;
1023 				second_time = 1;
1024 				goto again;
1025 			}
1026 		}
1027 	}
1028 	if (cp) {
1029 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1030 
1031 		rtm->rtm_version = RTM_VERSION;
1032 		rtm->rtm_type = type;
1033 		rtm->rtm_msglen = len;
1034 	}
1035 	return (len);
1036 }
1037 
1038 /*
1039  * This routine is called to generate a message from the routing
1040  * socket indicating that a redirect has occured, a routing lookup
1041  * has failed, or that a protocol has detected timeouts to a particular
1042  * destination.
1043  */
1044 void
1045 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1046 {
1047 	struct rt_msghdr *rtm;
1048 	struct mbuf *m;
1049 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1050 
1051 	if (route_cb.any_count == 0)
1052 		return;
1053 	m = rt_msg1(type, rtinfo);
1054 	if (m == NULL)
1055 		return;
1056 	rtm = mtod(m, struct rt_msghdr *);
1057 	rtm->rtm_flags = RTF_DONE | flags;
1058 	rtm->rtm_errno = error;
1059 	rtm->rtm_addrs = rtinfo->rti_addrs;
1060 	rt_dispatch(m, sa);
1061 }
1062 
1063 /*
1064  * This routine is called to generate a message from the routing
1065  * socket indicating that the status of a network interface has changed.
1066  */
1067 void
1068 rt_ifmsg(struct ifnet *ifp)
1069 {
1070 	struct if_msghdr *ifm;
1071 	struct mbuf *m;
1072 	struct rt_addrinfo info;
1073 
1074 	if (route_cb.any_count == 0)
1075 		return;
1076 	bzero((caddr_t)&info, sizeof(info));
1077 	m = rt_msg1(RTM_IFINFO, &info);
1078 	if (m == NULL)
1079 		return;
1080 	ifm = mtod(m, struct if_msghdr *);
1081 	ifm->ifm_index = ifp->if_index;
1082 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1083 	ifm->ifm_data = ifp->if_data;
1084 	ifm->ifm_addrs = 0;
1085 	rt_dispatch(m, NULL);
1086 }
1087 
1088 /*
1089  * This is called to generate messages from the routing socket
1090  * indicating a network interface has had addresses associated with it.
1091  * if we ever reverse the logic and replace messages TO the routing
1092  * socket indicate a request to configure interfaces, then it will
1093  * be unnecessary as the routing socket will automatically generate
1094  * copies of it.
1095  */
1096 void
1097 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1098 {
1099 	struct rt_addrinfo info;
1100 	struct sockaddr *sa = NULL;
1101 	int pass;
1102 	struct mbuf *m = NULL;
1103 	struct ifnet *ifp = ifa->ifa_ifp;
1104 
1105 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1106 		("unexpected cmd %u", cmd));
1107 #if defined(INET) || defined(INET6)
1108 #ifdef SCTP
1109 	/*
1110 	 * notify the SCTP stack
1111 	 * this will only get called when an address is added/deleted
1112 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1113 	 */
1114 	sctp_addr_change(ifa, cmd);
1115 #endif /* SCTP */
1116 #endif
1117 	if (route_cb.any_count == 0)
1118 		return;
1119 	for (pass = 1; pass < 3; pass++) {
1120 		bzero((caddr_t)&info, sizeof(info));
1121 		if ((cmd == RTM_ADD && pass == 1) ||
1122 		    (cmd == RTM_DELETE && pass == 2)) {
1123 			struct ifa_msghdr *ifam;
1124 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1125 
1126 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1127 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1128 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1129 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1130 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1131 				continue;
1132 			ifam = mtod(m, struct ifa_msghdr *);
1133 			ifam->ifam_index = ifp->if_index;
1134 			ifam->ifam_metric = ifa->ifa_metric;
1135 			ifam->ifam_flags = ifa->ifa_flags;
1136 			ifam->ifam_addrs = info.rti_addrs;
1137 		}
1138 		if ((cmd == RTM_ADD && pass == 2) ||
1139 		    (cmd == RTM_DELETE && pass == 1)) {
1140 			struct rt_msghdr *rtm;
1141 
1142 			if (rt == NULL)
1143 				continue;
1144 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1145 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1146 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1147 			if ((m = rt_msg1(cmd, &info)) == NULL)
1148 				continue;
1149 			rtm = mtod(m, struct rt_msghdr *);
1150 			rtm->rtm_index = ifp->if_index;
1151 			rtm->rtm_flags |= rt->rt_flags;
1152 			rtm->rtm_errno = error;
1153 			rtm->rtm_addrs = info.rti_addrs;
1154 		}
1155 		rt_dispatch(m, sa);
1156 	}
1157 }
1158 
1159 /*
1160  * This is the analogue to the rt_newaddrmsg which performs the same
1161  * function but for multicast group memberhips.  This is easier since
1162  * there is no route state to worry about.
1163  */
1164 void
1165 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1166 {
1167 	struct rt_addrinfo info;
1168 	struct mbuf *m = NULL;
1169 	struct ifnet *ifp = ifma->ifma_ifp;
1170 	struct ifma_msghdr *ifmam;
1171 
1172 	if (route_cb.any_count == 0)
1173 		return;
1174 
1175 	bzero((caddr_t)&info, sizeof(info));
1176 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1177 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1178 	/*
1179 	 * If a link-layer address is present, present it as a ``gateway''
1180 	 * (similarly to how ARP entries, e.g., are presented).
1181 	 */
1182 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1183 	m = rt_msg1(cmd, &info);
1184 	if (m == NULL)
1185 		return;
1186 	ifmam = mtod(m, struct ifma_msghdr *);
1187 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1188 	    __func__));
1189 	ifmam->ifmam_index = ifp->if_index;
1190 	ifmam->ifmam_addrs = info.rti_addrs;
1191 	rt_dispatch(m, ifma->ifma_addr);
1192 }
1193 
1194 static struct mbuf *
1195 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1196 	struct rt_addrinfo *info)
1197 {
1198 	struct if_announcemsghdr *ifan;
1199 	struct mbuf *m;
1200 
1201 	if (route_cb.any_count == 0)
1202 		return NULL;
1203 	bzero((caddr_t)info, sizeof(*info));
1204 	m = rt_msg1(type, info);
1205 	if (m != NULL) {
1206 		ifan = mtod(m, struct if_announcemsghdr *);
1207 		ifan->ifan_index = ifp->if_index;
1208 		strlcpy(ifan->ifan_name, ifp->if_xname,
1209 			sizeof(ifan->ifan_name));
1210 		ifan->ifan_what = what;
1211 	}
1212 	return m;
1213 }
1214 
1215 /*
1216  * This is called to generate routing socket messages indicating
1217  * IEEE80211 wireless events.
1218  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1219  */
1220 void
1221 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1222 {
1223 	struct mbuf *m;
1224 	struct rt_addrinfo info;
1225 
1226 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1227 	if (m != NULL) {
1228 		/*
1229 		 * Append the ieee80211 data.  Try to stick it in the
1230 		 * mbuf containing the ifannounce msg; otherwise allocate
1231 		 * a new mbuf and append.
1232 		 *
1233 		 * NB: we assume m is a single mbuf.
1234 		 */
1235 		if (data_len > M_TRAILINGSPACE(m)) {
1236 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1237 			if (n == NULL) {
1238 				m_freem(m);
1239 				return;
1240 			}
1241 			bcopy(data, mtod(n, void *), data_len);
1242 			n->m_len = data_len;
1243 			m->m_next = n;
1244 		} else if (data_len > 0) {
1245 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1246 			m->m_len += data_len;
1247 		}
1248 		if (m->m_flags & M_PKTHDR)
1249 			m->m_pkthdr.len += data_len;
1250 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1251 		rt_dispatch(m, NULL);
1252 	}
1253 }
1254 
1255 /*
1256  * This is called to generate routing socket messages indicating
1257  * network interface arrival and departure.
1258  */
1259 void
1260 rt_ifannouncemsg(struct ifnet *ifp, int what)
1261 {
1262 	struct mbuf *m;
1263 	struct rt_addrinfo info;
1264 
1265 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1266 	if (m != NULL)
1267 		rt_dispatch(m, NULL);
1268 }
1269 
1270 static void
1271 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1272 {
1273 	struct m_tag *tag;
1274 
1275 	/*
1276 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1277 	 * use when injecting the mbuf into the routing socket buffer from
1278 	 * the netisr.
1279 	 */
1280 	if (sa != NULL) {
1281 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1282 		    M_NOWAIT);
1283 		if (tag == NULL) {
1284 			m_freem(m);
1285 			return;
1286 		}
1287 		*(unsigned short *)(tag + 1) = sa->sa_family;
1288 		m_tag_prepend(m, tag);
1289 	}
1290 #ifdef VIMAGE
1291 	if (V_loif)
1292 		m->m_pkthdr.rcvif = V_loif;
1293 	else {
1294 		m_freem(m);
1295 		return;
1296 	}
1297 #endif
1298 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1299 }
1300 
1301 /*
1302  * This is used in dumping the kernel table via sysctl().
1303  */
1304 static int
1305 sysctl_dumpentry(struct radix_node *rn, void *vw)
1306 {
1307 	struct walkarg *w = vw;
1308 	struct rtentry *rt = (struct rtentry *)rn;
1309 	int error = 0, size;
1310 	struct rt_addrinfo info;
1311 
1312 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1313 		return 0;
1314 	if ((rt->rt_flags & RTF_HOST) == 0
1315 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1316 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1317 		return (0);
1318 	bzero((caddr_t)&info, sizeof(info));
1319 	info.rti_info[RTAX_DST] = rt_key(rt);
1320 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1321 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1322 	info.rti_info[RTAX_GENMASK] = 0;
1323 	if (rt->rt_ifp) {
1324 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1325 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1326 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1327 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1328 	}
1329 	size = rt_msg2(RTM_GET, &info, NULL, w);
1330 	if (w->w_req && w->w_tmem) {
1331 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1332 
1333 		rtm->rtm_flags = rt->rt_flags;
1334 		/*
1335 		 * let's be honest about this being a retarded hack
1336 		 */
1337 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1338 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1339 		rtm->rtm_index = rt->rt_ifp->if_index;
1340 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1341 		rtm->rtm_addrs = info.rti_addrs;
1342 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1343 		return (error);
1344 	}
1345 	return (error);
1346 }
1347 
1348 static int
1349 sysctl_iflist(int af, struct walkarg *w)
1350 {
1351 	struct ifnet *ifp;
1352 	struct ifaddr *ifa;
1353 	struct rt_addrinfo info;
1354 	int len, error = 0;
1355 
1356 	bzero((caddr_t)&info, sizeof(info));
1357 	IFNET_RLOCK();
1358 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1359 		if (w->w_arg && w->w_arg != ifp->if_index)
1360 			continue;
1361 		ifa = ifp->if_addr;
1362 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1363 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1364 		info.rti_info[RTAX_IFP] = NULL;
1365 		if (w->w_req && w->w_tmem) {
1366 			struct if_msghdr *ifm;
1367 
1368 			ifm = (struct if_msghdr *)w->w_tmem;
1369 			ifm->ifm_index = ifp->if_index;
1370 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1371 			ifm->ifm_data = ifp->if_data;
1372 			ifm->ifm_addrs = info.rti_addrs;
1373 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
1374 			if (error)
1375 				goto done;
1376 		}
1377 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1378 			if (af && af != ifa->ifa_addr->sa_family)
1379 				continue;
1380 			if (prison_if(w->w_req->td->td_ucred,
1381 			    ifa->ifa_addr) != 0)
1382 				continue;
1383 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1384 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1385 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1386 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1387 			if (w->w_req && w->w_tmem) {
1388 				struct ifa_msghdr *ifam;
1389 
1390 				ifam = (struct ifa_msghdr *)w->w_tmem;
1391 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1392 				ifam->ifam_flags = ifa->ifa_flags;
1393 				ifam->ifam_metric = ifa->ifa_metric;
1394 				ifam->ifam_addrs = info.rti_addrs;
1395 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1396 				if (error)
1397 					goto done;
1398 			}
1399 		}
1400 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1401 			info.rti_info[RTAX_BRD] = NULL;
1402 	}
1403 done:
1404 	IFNET_RUNLOCK();
1405 	return (error);
1406 }
1407 
1408 static int
1409 sysctl_ifmalist(int af, struct walkarg *w)
1410 {
1411 	struct ifnet *ifp;
1412 	struct ifmultiaddr *ifma;
1413 	struct	rt_addrinfo info;
1414 	int	len, error = 0;
1415 	struct ifaddr *ifa;
1416 
1417 	bzero((caddr_t)&info, sizeof(info));
1418 	IFNET_RLOCK();
1419 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1420 		if (w->w_arg && w->w_arg != ifp->if_index)
1421 			continue;
1422 		ifa = ifp->if_addr;
1423 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1424 		IF_ADDR_LOCK(ifp);
1425 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1426 			if (af && af != ifma->ifma_addr->sa_family)
1427 				continue;
1428 			if (prison_if(w->w_req->td->td_ucred,
1429 			    ifma->ifma_addr) != 0)
1430 				continue;
1431 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1432 			info.rti_info[RTAX_GATEWAY] =
1433 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1434 			    ifma->ifma_lladdr : NULL;
1435 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1436 			if (w->w_req && w->w_tmem) {
1437 				struct ifma_msghdr *ifmam;
1438 
1439 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1440 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1441 				ifmam->ifmam_flags = 0;
1442 				ifmam->ifmam_addrs = info.rti_addrs;
1443 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1444 				if (error) {
1445 					IF_ADDR_UNLOCK(ifp);
1446 					goto done;
1447 				}
1448 			}
1449 		}
1450 		IF_ADDR_UNLOCK(ifp);
1451 	}
1452 done:
1453 	IFNET_RUNLOCK();
1454 	return (error);
1455 }
1456 
1457 static int
1458 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1459 {
1460 	int	*name = (int *)arg1;
1461 	u_int	namelen = arg2;
1462 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1463 	int	i, lim, error = EINVAL;
1464 	u_char	af;
1465 	struct	walkarg w;
1466 
1467 	name ++;
1468 	namelen--;
1469 	if (req->newptr)
1470 		return (EPERM);
1471 	if (namelen != 3)
1472 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1473 	af = name[0];
1474 	if (af > AF_MAX)
1475 		return (EINVAL);
1476 	bzero(&w, sizeof(w));
1477 	w.w_op = name[1];
1478 	w.w_arg = name[2];
1479 	w.w_req = req;
1480 
1481 	error = sysctl_wire_old_buffer(req, 0);
1482 	if (error)
1483 		return (error);
1484 	switch (w.w_op) {
1485 
1486 	case NET_RT_DUMP:
1487 	case NET_RT_FLAGS:
1488 		if (af == 0) {			/* dump all tables */
1489 			i = 1;
1490 			lim = AF_MAX;
1491 		} else				/* dump only one table */
1492 			i = lim = af;
1493 
1494 		/*
1495 		 * take care of llinfo entries, the caller must
1496 		 * specify an AF
1497 		 */
1498 		if (w.w_op == NET_RT_FLAGS &&
1499 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1500 			if (af != 0)
1501 				error = lltable_sysctl_dumparp(af, w.w_req);
1502 			else
1503 				error = EINVAL;
1504 			break;
1505 		}
1506 		/*
1507 		 * take care of routing entries
1508 		 */
1509 		for (error = 0; error == 0 && i <= lim; i++) {
1510 			rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1511 			if (rnh != NULL) {
1512 				RADIX_NODE_HEAD_LOCK(rnh);
1513 			    	error = rnh->rnh_walktree(rnh,
1514 				    sysctl_dumpentry, &w);
1515 				RADIX_NODE_HEAD_UNLOCK(rnh);
1516 			} else if (af != 0)
1517 				error = EAFNOSUPPORT;
1518 		}
1519 		break;
1520 
1521 	case NET_RT_IFLIST:
1522 		error = sysctl_iflist(af, &w);
1523 		break;
1524 
1525 	case NET_RT_IFMALIST:
1526 		error = sysctl_ifmalist(af, &w);
1527 		break;
1528 	}
1529 	if (w.w_tmem)
1530 		free(w.w_tmem, M_RTABLE);
1531 	return (error);
1532 }
1533 
1534 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1535 
1536 /*
1537  * Definitions of protocols supported in the ROUTE domain.
1538  */
1539 
1540 static struct domain routedomain;		/* or at least forward */
1541 
1542 static struct protosw routesw[] = {
1543 {
1544 	.pr_type =		SOCK_RAW,
1545 	.pr_domain =		&routedomain,
1546 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1547 	.pr_output =		route_output,
1548 	.pr_ctlinput =		raw_ctlinput,
1549 	.pr_init =		raw_init,
1550 	.pr_usrreqs =		&route_usrreqs
1551 }
1552 };
1553 
1554 static struct domain routedomain = {
1555 	.dom_family =		PF_ROUTE,
1556 	.dom_name =		 "route",
1557 	.dom_protosw =		routesw,
1558 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1559 };
1560 
1561 VNET_DOMAIN_SET(route);
1562