xref: /freebsd/sys/net/rtsock.c (revision 10b59a9b4add0320d52c15ce057dd697261e7dfc)
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_compat.h"
33 #include "opt_sctp.h"
34 #include "opt_mpath.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/jail.h>
40 #include <sys/kernel.h>
41 #include <sys/domain.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 
55 #include <net/if.h>
56 #include <net/if_dl.h>
57 #include <net/if_llatbl.h>
58 #include <net/if_types.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 #include <netinet/if_ether.h>
66 #ifdef INET6
67 #include <netinet6/scope6_var.h>
68 #endif
69 
70 #if defined(INET) || defined(INET6)
71 #ifdef SCTP
72 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
73 #endif /* SCTP */
74 #endif
75 
76 #ifdef COMPAT_FREEBSD32
77 #include <sys/mount.h>
78 #include <compat/freebsd32/freebsd32.h>
79 
80 struct if_data32 {
81 	uint8_t	ifi_type;
82 	uint8_t	ifi_physical;
83 	uint8_t	ifi_addrlen;
84 	uint8_t	ifi_hdrlen;
85 	uint8_t	ifi_link_state;
86 	uint8_t	ifi_spare_char1;
87 	uint8_t	ifi_spare_char2;
88 	uint8_t	ifi_datalen;
89 	uint32_t ifi_mtu;
90 	uint32_t ifi_metric;
91 	uint32_t ifi_baudrate;
92 	uint32_t ifi_ipackets;
93 	uint32_t ifi_ierrors;
94 	uint32_t ifi_opackets;
95 	uint32_t ifi_oerrors;
96 	uint32_t ifi_collisions;
97 	uint32_t ifi_ibytes;
98 	uint32_t ifi_obytes;
99 	uint32_t ifi_imcasts;
100 	uint32_t ifi_omcasts;
101 	uint32_t ifi_iqdrops;
102 	uint32_t ifi_noproto;
103 	uint32_t ifi_hwassist;
104 	int32_t	ifi_epoch;
105 	struct	timeval32 ifi_lastchange;
106 };
107 
108 struct if_msghdr32 {
109 	uint16_t ifm_msglen;
110 	uint8_t	ifm_version;
111 	uint8_t	ifm_type;
112 	int32_t	ifm_addrs;
113 	int32_t	ifm_flags;
114 	uint16_t ifm_index;
115 	struct	if_data32 ifm_data;
116 };
117 #endif
118 
119 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
120 
121 /* NB: these are not modified */
122 static struct	sockaddr route_src = { 2, PF_ROUTE, };
123 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
124 
125 /*
126  * Used by rtsock/raw_input callback code to decide whether to filter the update
127  * notification to a socket bound to a particular FIB.
128  */
129 #define	RTS_FILTER_FIB	M_PROTO8
130 #define	RTS_ALLFIBS	-1
131 
132 static struct {
133 	int	ip_count;	/* attached w/ AF_INET */
134 	int	ip6_count;	/* attached w/ AF_INET6 */
135 	int	ipx_count;	/* attached w/ AF_IPX */
136 	int	any_count;	/* total attached */
137 } route_cb;
138 
139 struct mtx rtsock_mtx;
140 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
141 
142 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
143 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
144 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
145 
146 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
147 
148 struct walkarg {
149 	int	w_tmemsize;
150 	int	w_op, w_arg;
151 	caddr_t	w_tmem;
152 	struct sysctl_req *w_req;
153 };
154 
155 static void	rts_input(struct mbuf *m);
156 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
157 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
158 			caddr_t cp, struct walkarg *w);
159 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
160 			struct rt_addrinfo *rtinfo);
161 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
162 static int	sysctl_iflist(int af, struct walkarg *w);
163 static int	sysctl_ifmalist(int af, struct walkarg *w);
164 static int	route_output(struct mbuf *m, struct socket *so);
165 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
166 			struct rt_metrics_lite *out);
167 static void	rt_getmetrics(const struct rt_metrics_lite *in,
168 			struct rt_metrics *out);
169 static void	rt_dispatch(struct mbuf *, sa_family_t);
170 
171 static struct netisr_handler rtsock_nh = {
172 	.nh_name = "rtsock",
173 	.nh_handler = rts_input,
174 	.nh_proto = NETISR_ROUTE,
175 	.nh_policy = NETISR_POLICY_SOURCE,
176 };
177 
178 static int
179 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
180 {
181 	int error, qlimit;
182 
183 	netisr_getqlimit(&rtsock_nh, &qlimit);
184 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
185         if (error || !req->newptr)
186                 return (error);
187 	if (qlimit < 1)
188 		return (EINVAL);
189 	return (netisr_setqlimit(&rtsock_nh, qlimit));
190 }
191 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
192     0, 0, sysctl_route_netisr_maxqlen, "I",
193     "maximum routing socket dispatch queue length");
194 
195 static void
196 rts_init(void)
197 {
198 	int tmp;
199 
200 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
201 		rtsock_nh.nh_qlimit = tmp;
202 	netisr_register(&rtsock_nh);
203 }
204 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
205 
206 static int
207 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
208     struct rawcb *rp)
209 {
210 	int fibnum;
211 
212 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
213 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
214 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
215 
216 	/* No filtering requested. */
217 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
218 		return (0);
219 
220 	/* Check if it is a rts and the fib matches the one of the socket. */
221 	fibnum = M_GETFIB(m);
222 	if (proto->sp_family != PF_ROUTE ||
223 	    rp->rcb_socket == NULL ||
224 	    rp->rcb_socket->so_fibnum == fibnum)
225 		return (0);
226 
227 	/* Filtering requested and no match, the socket shall be skipped. */
228 	return (1);
229 }
230 
231 static void
232 rts_input(struct mbuf *m)
233 {
234 	struct sockproto route_proto;
235 	unsigned short *family;
236 	struct m_tag *tag;
237 
238 	route_proto.sp_family = PF_ROUTE;
239 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
240 	if (tag != NULL) {
241 		family = (unsigned short *)(tag + 1);
242 		route_proto.sp_protocol = *family;
243 		m_tag_delete(m, tag);
244 	} else
245 		route_proto.sp_protocol = 0;
246 
247 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
248 }
249 
250 /*
251  * It really doesn't make any sense at all for this code to share much
252  * with raw_usrreq.c, since its functionality is so restricted.  XXX
253  */
254 static void
255 rts_abort(struct socket *so)
256 {
257 
258 	raw_usrreqs.pru_abort(so);
259 }
260 
261 static void
262 rts_close(struct socket *so)
263 {
264 
265 	raw_usrreqs.pru_close(so);
266 }
267 
268 /* pru_accept is EOPNOTSUPP */
269 
270 static int
271 rts_attach(struct socket *so, int proto, struct thread *td)
272 {
273 	struct rawcb *rp;
274 	int s, error;
275 
276 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
277 
278 	/* XXX */
279 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
280 	if (rp == NULL)
281 		return ENOBUFS;
282 
283 	/*
284 	 * The splnet() is necessary to block protocols from sending
285 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
286 	 * this PCB is extant but incompletely initialized.
287 	 * Probably we should try to do more of this work beforehand and
288 	 * eliminate the spl.
289 	 */
290 	s = splnet();
291 	so->so_pcb = (caddr_t)rp;
292 	so->so_fibnum = td->td_proc->p_fibnum;
293 	error = raw_attach(so, proto);
294 	rp = sotorawcb(so);
295 	if (error) {
296 		splx(s);
297 		so->so_pcb = NULL;
298 		free(rp, M_PCB);
299 		return error;
300 	}
301 	RTSOCK_LOCK();
302 	switch(rp->rcb_proto.sp_protocol) {
303 	case AF_INET:
304 		route_cb.ip_count++;
305 		break;
306 	case AF_INET6:
307 		route_cb.ip6_count++;
308 		break;
309 	case AF_IPX:
310 		route_cb.ipx_count++;
311 		break;
312 	}
313 	route_cb.any_count++;
314 	RTSOCK_UNLOCK();
315 	soisconnected(so);
316 	so->so_options |= SO_USELOOPBACK;
317 	splx(s);
318 	return 0;
319 }
320 
321 static int
322 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
323 {
324 
325 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
326 }
327 
328 static int
329 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
330 {
331 
332 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
333 }
334 
335 /* pru_connect2 is EOPNOTSUPP */
336 /* pru_control is EOPNOTSUPP */
337 
338 static void
339 rts_detach(struct socket *so)
340 {
341 	struct rawcb *rp = sotorawcb(so);
342 
343 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
344 
345 	RTSOCK_LOCK();
346 	switch(rp->rcb_proto.sp_protocol) {
347 	case AF_INET:
348 		route_cb.ip_count--;
349 		break;
350 	case AF_INET6:
351 		route_cb.ip6_count--;
352 		break;
353 	case AF_IPX:
354 		route_cb.ipx_count--;
355 		break;
356 	}
357 	route_cb.any_count--;
358 	RTSOCK_UNLOCK();
359 	raw_usrreqs.pru_detach(so);
360 }
361 
362 static int
363 rts_disconnect(struct socket *so)
364 {
365 
366 	return (raw_usrreqs.pru_disconnect(so));
367 }
368 
369 /* pru_listen is EOPNOTSUPP */
370 
371 static int
372 rts_peeraddr(struct socket *so, struct sockaddr **nam)
373 {
374 
375 	return (raw_usrreqs.pru_peeraddr(so, nam));
376 }
377 
378 /* pru_rcvd is EOPNOTSUPP */
379 /* pru_rcvoob is EOPNOTSUPP */
380 
381 static int
382 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
383 	 struct mbuf *control, struct thread *td)
384 {
385 
386 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
387 }
388 
389 /* pru_sense is null */
390 
391 static int
392 rts_shutdown(struct socket *so)
393 {
394 
395 	return (raw_usrreqs.pru_shutdown(so));
396 }
397 
398 static int
399 rts_sockaddr(struct socket *so, struct sockaddr **nam)
400 {
401 
402 	return (raw_usrreqs.pru_sockaddr(so, nam));
403 }
404 
405 static struct pr_usrreqs route_usrreqs = {
406 	.pru_abort =		rts_abort,
407 	.pru_attach =		rts_attach,
408 	.pru_bind =		rts_bind,
409 	.pru_connect =		rts_connect,
410 	.pru_detach =		rts_detach,
411 	.pru_disconnect =	rts_disconnect,
412 	.pru_peeraddr =		rts_peeraddr,
413 	.pru_send =		rts_send,
414 	.pru_shutdown =		rts_shutdown,
415 	.pru_sockaddr =		rts_sockaddr,
416 	.pru_close =		rts_close,
417 };
418 
419 #ifndef _SOCKADDR_UNION_DEFINED
420 #define	_SOCKADDR_UNION_DEFINED
421 /*
422  * The union of all possible address formats we handle.
423  */
424 union sockaddr_union {
425 	struct sockaddr		sa;
426 	struct sockaddr_in	sin;
427 	struct sockaddr_in6	sin6;
428 };
429 #endif /* _SOCKADDR_UNION_DEFINED */
430 
431 static int
432 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
433     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
434 {
435 
436 	/* First, see if the returned address is part of the jail. */
437 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
438 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
439 		return (0);
440 	}
441 
442 	switch (info->rti_info[RTAX_DST]->sa_family) {
443 #ifdef INET
444 	case AF_INET:
445 	{
446 		struct in_addr ia;
447 		struct ifaddr *ifa;
448 		int found;
449 
450 		found = 0;
451 		/*
452 		 * Try to find an address on the given outgoing interface
453 		 * that belongs to the jail.
454 		 */
455 		IF_ADDR_LOCK(ifp);
456 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
457 			struct sockaddr *sa;
458 			sa = ifa->ifa_addr;
459 			if (sa->sa_family != AF_INET)
460 				continue;
461 			ia = ((struct sockaddr_in *)sa)->sin_addr;
462 			if (prison_check_ip4(cred, &ia) == 0) {
463 				found = 1;
464 				break;
465 			}
466 		}
467 		IF_ADDR_UNLOCK(ifp);
468 		if (!found) {
469 			/*
470 			 * As a last resort return the 'default' jail address.
471 			 */
472 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
473 			    sin_addr;
474 			if (prison_get_ip4(cred, &ia) != 0)
475 				return (ESRCH);
476 		}
477 		bzero(&saun->sin, sizeof(struct sockaddr_in));
478 		saun->sin.sin_len = sizeof(struct sockaddr_in);
479 		saun->sin.sin_family = AF_INET;
480 		saun->sin.sin_addr.s_addr = ia.s_addr;
481 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
482 		break;
483 	}
484 #endif
485 #ifdef INET6
486 	case AF_INET6:
487 	{
488 		struct in6_addr ia6;
489 		struct ifaddr *ifa;
490 		int found;
491 
492 		found = 0;
493 		/*
494 		 * Try to find an address on the given outgoing interface
495 		 * that belongs to the jail.
496 		 */
497 		IF_ADDR_LOCK(ifp);
498 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
499 			struct sockaddr *sa;
500 			sa = ifa->ifa_addr;
501 			if (sa->sa_family != AF_INET6)
502 				continue;
503 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
504 			    &ia6, sizeof(struct in6_addr));
505 			if (prison_check_ip6(cred, &ia6) == 0) {
506 				found = 1;
507 				break;
508 			}
509 		}
510 		IF_ADDR_UNLOCK(ifp);
511 		if (!found) {
512 			/*
513 			 * As a last resort return the 'default' jail address.
514 			 */
515 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
516 			    sin6_addr;
517 			if (prison_get_ip6(cred, &ia6) != 0)
518 				return (ESRCH);
519 		}
520 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
521 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
522 		saun->sin6.sin6_family = AF_INET6;
523 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
524 		if (sa6_recoverscope(&saun->sin6) != 0)
525 			return (ESRCH);
526 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
527 		break;
528 	}
529 #endif
530 	default:
531 		return (ESRCH);
532 	}
533 	return (0);
534 }
535 
536 /*ARGSUSED*/
537 static int
538 route_output(struct mbuf *m, struct socket *so)
539 {
540 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
541 	struct rt_msghdr *rtm = NULL;
542 	struct rtentry *rt = NULL;
543 	struct radix_node_head *rnh;
544 	struct rt_addrinfo info;
545 	int len, error = 0;
546 	struct ifnet *ifp = NULL;
547 	union sockaddr_union saun;
548 	sa_family_t saf = AF_UNSPEC;
549 
550 #define senderr(e) { error = e; goto flush;}
551 	if (m == NULL || ((m->m_len < sizeof(long)) &&
552 		       (m = m_pullup(m, sizeof(long))) == NULL))
553 		return (ENOBUFS);
554 	if ((m->m_flags & M_PKTHDR) == 0)
555 		panic("route_output");
556 	len = m->m_pkthdr.len;
557 	if (len < sizeof(*rtm) ||
558 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
559 		info.rti_info[RTAX_DST] = NULL;
560 		senderr(EINVAL);
561 	}
562 	R_Malloc(rtm, struct rt_msghdr *, len);
563 	if (rtm == NULL) {
564 		info.rti_info[RTAX_DST] = NULL;
565 		senderr(ENOBUFS);
566 	}
567 	m_copydata(m, 0, len, (caddr_t)rtm);
568 	if (rtm->rtm_version != RTM_VERSION) {
569 		info.rti_info[RTAX_DST] = NULL;
570 		senderr(EPROTONOSUPPORT);
571 	}
572 	rtm->rtm_pid = curproc->p_pid;
573 	bzero(&info, sizeof(info));
574 	info.rti_addrs = rtm->rtm_addrs;
575 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
576 		info.rti_info[RTAX_DST] = NULL;
577 		senderr(EINVAL);
578 	}
579 	info.rti_flags = rtm->rtm_flags;
580 	if (info.rti_info[RTAX_DST] == NULL ||
581 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
582 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
583 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
584 		senderr(EINVAL);
585 	saf = info.rti_info[RTAX_DST]->sa_family;
586 	/*
587 	 * Verify that the caller has the appropriate privilege; RTM_GET
588 	 * is the only operation the non-superuser is allowed.
589 	 */
590 	if (rtm->rtm_type != RTM_GET) {
591 		error = priv_check(curthread, PRIV_NET_ROUTE);
592 		if (error)
593 			senderr(error);
594 	}
595 
596 	/*
597 	 * The given gateway address may be an interface address.
598 	 * For example, issuing a "route change" command on a route
599 	 * entry that was created from a tunnel, and the gateway
600 	 * address given is the local end point. In this case the
601 	 * RTF_GATEWAY flag must be cleared or the destination will
602 	 * not be reachable even though there is no error message.
603 	 */
604 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
605 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
606 		struct route gw_ro;
607 
608 		bzero(&gw_ro, sizeof(gw_ro));
609 		gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
610 		rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
611 		/*
612 		 * A host route through the loopback interface is
613 		 * installed for each interface adddress. In pre 8.0
614 		 * releases the interface address of a PPP link type
615 		 * is not reachable locally. This behavior is fixed as
616 		 * part of the new L2/L3 redesign and rewrite work. The
617 		 * signature of this interface address route is the
618 		 * AF_LINK sa_family type of the rt_gateway, and the
619 		 * rt_ifp has the IFF_LOOPBACK flag set.
620 		 */
621 		if (gw_ro.ro_rt != NULL &&
622 		    gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
623 		    gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)
624 			info.rti_flags &= ~RTF_GATEWAY;
625 		if (gw_ro.ro_rt != NULL)
626 			RTFREE(gw_ro.ro_rt);
627 	}
628 
629 	switch (rtm->rtm_type) {
630 		struct rtentry *saved_nrt;
631 
632 	case RTM_ADD:
633 		if (info.rti_info[RTAX_GATEWAY] == NULL)
634 			senderr(EINVAL);
635 		saved_nrt = NULL;
636 
637 		/* support for new ARP code */
638 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
639 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
640 			error = lla_rt_output(rtm, &info);
641 			break;
642 		}
643 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
644 		    so->so_fibnum);
645 		if (error == 0 && saved_nrt) {
646 			RT_LOCK(saved_nrt);
647 			rt_setmetrics(rtm->rtm_inits,
648 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
649 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
650 			RT_REMREF(saved_nrt);
651 			RT_UNLOCK(saved_nrt);
652 		}
653 		break;
654 
655 	case RTM_DELETE:
656 		saved_nrt = NULL;
657 		/* support for new ARP code */
658 		if (info.rti_info[RTAX_GATEWAY] &&
659 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
660 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
661 			error = lla_rt_output(rtm, &info);
662 			break;
663 		}
664 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
665 		    so->so_fibnum);
666 		if (error == 0) {
667 			RT_LOCK(saved_nrt);
668 			rt = saved_nrt;
669 			goto report;
670 		}
671 		break;
672 
673 	case RTM_GET:
674 	case RTM_CHANGE:
675 	case RTM_LOCK:
676 		rnh = rt_tables_get_rnh(so->so_fibnum,
677 		    info.rti_info[RTAX_DST]->sa_family);
678 		if (rnh == NULL)
679 			senderr(EAFNOSUPPORT);
680 		RADIX_NODE_HEAD_RLOCK(rnh);
681 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
682 			info.rti_info[RTAX_NETMASK], rnh);
683 		if (rt == NULL) {	/* XXX looks bogus */
684 			RADIX_NODE_HEAD_RUNLOCK(rnh);
685 			senderr(ESRCH);
686 		}
687 #ifdef RADIX_MPATH
688 		/*
689 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
690 		 * we require users to specify a matching RTAX_GATEWAY.
691 		 *
692 		 * for RTM_GET, gate is optional even with multipath.
693 		 * if gate == NULL the first match is returned.
694 		 * (no need to call rt_mpath_matchgate if gate == NULL)
695 		 */
696 		if (rn_mpath_capable(rnh) &&
697 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
698 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
699 			if (!rt) {
700 				RADIX_NODE_HEAD_RUNLOCK(rnh);
701 				senderr(ESRCH);
702 			}
703 		}
704 #endif
705 		/*
706 		 * If performing proxied L2 entry insertion, and
707 		 * the actual PPP host entry is found, perform
708 		 * another search to retrieve the prefix route of
709 		 * the local end point of the PPP link.
710 		 */
711 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
712 			struct sockaddr laddr;
713 
714 			if (rt->rt_ifp != NULL &&
715 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
716 				struct ifaddr *ifa;
717 
718 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
719 				if (ifa != NULL)
720 					rt_maskedcopy(ifa->ifa_addr,
721 						      &laddr,
722 						      ifa->ifa_netmask);
723 			} else
724 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
725 					      &laddr,
726 					      rt->rt_ifa->ifa_netmask);
727 			/*
728 			 * refactor rt and no lock operation necessary
729 			 */
730 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
731 			if (rt == NULL) {
732 				RADIX_NODE_HEAD_RUNLOCK(rnh);
733 				senderr(ESRCH);
734 			}
735 		}
736 		RT_LOCK(rt);
737 		RT_ADDREF(rt);
738 		RADIX_NODE_HEAD_RUNLOCK(rnh);
739 
740 		/*
741 		 * Fix for PR: 82974
742 		 *
743 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
744 		 * returns a perfect match in case a netmask is
745 		 * specified.  For host routes only a longest prefix
746 		 * match is returned so it is necessary to compare the
747 		 * existence of the netmask.  If both have a netmask
748 		 * rnh_lookup() did a perfect match and if none of them
749 		 * have a netmask both are host routes which is also a
750 		 * perfect match.
751 		 */
752 
753 		if (rtm->rtm_type != RTM_GET &&
754 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
755 			RT_UNLOCK(rt);
756 			senderr(ESRCH);
757 		}
758 
759 		switch(rtm->rtm_type) {
760 
761 		case RTM_GET:
762 		report:
763 			RT_LOCK_ASSERT(rt);
764 			if ((rt->rt_flags & RTF_HOST) == 0
765 			    ? jailed_without_vnet(curthread->td_ucred)
766 			    : prison_if(curthread->td_ucred,
767 			    rt_key(rt)) != 0) {
768 				RT_UNLOCK(rt);
769 				senderr(ESRCH);
770 			}
771 			info.rti_info[RTAX_DST] = rt_key(rt);
772 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
773 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
774 			info.rti_info[RTAX_GENMASK] = 0;
775 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
776 				ifp = rt->rt_ifp;
777 				if (ifp) {
778 					info.rti_info[RTAX_IFP] =
779 					    ifp->if_addr->ifa_addr;
780 					error = rtm_get_jailed(&info, ifp, rt,
781 					    &saun, curthread->td_ucred);
782 					if (error != 0) {
783 						RT_UNLOCK(rt);
784 						senderr(error);
785 					}
786 					if (ifp->if_flags & IFF_POINTOPOINT)
787 						info.rti_info[RTAX_BRD] =
788 						    rt->rt_ifa->ifa_dstaddr;
789 					rtm->rtm_index = ifp->if_index;
790 				} else {
791 					info.rti_info[RTAX_IFP] = NULL;
792 					info.rti_info[RTAX_IFA] = NULL;
793 				}
794 			} else if ((ifp = rt->rt_ifp) != NULL) {
795 				rtm->rtm_index = ifp->if_index;
796 			}
797 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
798 			if (len > rtm->rtm_msglen) {
799 				struct rt_msghdr *new_rtm;
800 				R_Malloc(new_rtm, struct rt_msghdr *, len);
801 				if (new_rtm == NULL) {
802 					RT_UNLOCK(rt);
803 					senderr(ENOBUFS);
804 				}
805 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
806 				Free(rtm); rtm = new_rtm;
807 			}
808 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
809 			rtm->rtm_flags = rt->rt_flags;
810 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
811 			rtm->rtm_addrs = info.rti_addrs;
812 			break;
813 
814 		case RTM_CHANGE:
815 			/*
816 			 * New gateway could require new ifaddr, ifp;
817 			 * flags may also be different; ifp may be specified
818 			 * by ll sockaddr when protocol address is ambiguous
819 			 */
820 			if (((rt->rt_flags & RTF_GATEWAY) &&
821 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
822 			    info.rti_info[RTAX_IFP] != NULL ||
823 			    (info.rti_info[RTAX_IFA] != NULL &&
824 			     !sa_equal(info.rti_info[RTAX_IFA],
825 				       rt->rt_ifa->ifa_addr))) {
826 				RT_UNLOCK(rt);
827 				RADIX_NODE_HEAD_LOCK(rnh);
828 				error = rt_getifa_fib(&info, rt->rt_fibnum);
829 				/*
830 				 * XXXRW: Really we should release this
831 				 * reference later, but this maintains
832 				 * historical behavior.
833 				 */
834 				if (info.rti_ifa != NULL)
835 					ifa_free(info.rti_ifa);
836 				RADIX_NODE_HEAD_UNLOCK(rnh);
837 				if (error != 0)
838 					senderr(error);
839 				RT_LOCK(rt);
840 			}
841 			if (info.rti_ifa != NULL &&
842 			    info.rti_ifa != rt->rt_ifa &&
843 			    rt->rt_ifa != NULL &&
844 			    rt->rt_ifa->ifa_rtrequest != NULL) {
845 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
846 				    &info);
847 				ifa_free(rt->rt_ifa);
848 			}
849 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
850 				RT_UNLOCK(rt);
851 				RADIX_NODE_HEAD_LOCK(rnh);
852 				RT_LOCK(rt);
853 
854 				error = rt_setgate(rt, rt_key(rt),
855 				    info.rti_info[RTAX_GATEWAY]);
856 				RADIX_NODE_HEAD_UNLOCK(rnh);
857 				if (error != 0) {
858 					RT_UNLOCK(rt);
859 					senderr(error);
860 				}
861 				rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
862 			}
863 			if (info.rti_ifa != NULL &&
864 			    info.rti_ifa != rt->rt_ifa) {
865 				ifa_ref(info.rti_ifa);
866 				rt->rt_ifa = info.rti_ifa;
867 				rt->rt_ifp = info.rti_ifp;
868 			}
869 			/* Allow some flags to be toggled on change. */
870 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
871 				    (rtm->rtm_flags & RTF_FMASK);
872 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
873 					&rt->rt_rmx);
874 			rtm->rtm_index = rt->rt_ifp->if_index;
875 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
876 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
877 			/* FALLTHROUGH */
878 		case RTM_LOCK:
879 			/* We don't support locks anymore */
880 			break;
881 		}
882 		RT_UNLOCK(rt);
883 		break;
884 
885 	default:
886 		senderr(EOPNOTSUPP);
887 	}
888 
889 flush:
890 	if (rtm) {
891 		if (error)
892 			rtm->rtm_errno = error;
893 		else
894 			rtm->rtm_flags |= RTF_DONE;
895 	}
896 	if (rt)		/* XXX can this be true? */
897 		RTFREE(rt);
898     {
899 	struct rawcb *rp = NULL;
900 	/*
901 	 * Check to see if we don't want our own messages.
902 	 */
903 	if ((so->so_options & SO_USELOOPBACK) == 0) {
904 		if (route_cb.any_count <= 1) {
905 			if (rtm)
906 				Free(rtm);
907 			m_freem(m);
908 			return (error);
909 		}
910 		/* There is another listener, so construct message */
911 		rp = sotorawcb(so);
912 	}
913 	if (rtm) {
914 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
915 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
916 			m_freem(m);
917 			m = NULL;
918 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
919 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
920 	}
921 	if (m) {
922 		M_SETFIB(m, so->so_fibnum);
923 		m->m_flags |= RTS_FILTER_FIB;
924 		if (rp) {
925 			/*
926 			 * XXX insure we don't get a copy by
927 			 * invalidating our protocol
928 			 */
929 			unsigned short family = rp->rcb_proto.sp_family;
930 			rp->rcb_proto.sp_family = 0;
931 			rt_dispatch(m, saf);
932 			rp->rcb_proto.sp_family = family;
933 		} else
934 			rt_dispatch(m, saf);
935 	}
936 	/* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
937 	if (rtm)
938 		Free(rtm);
939     }
940 	return (error);
941 #undef	sa_equal
942 }
943 
944 static void
945 rt_setmetrics(u_long which, const struct rt_metrics *in,
946 	struct rt_metrics_lite *out)
947 {
948 #define metric(f, e) if (which & (f)) out->e = in->e;
949 	/*
950 	 * Only these are stored in the routing entry since introduction
951 	 * of tcp hostcache. The rest is ignored.
952 	 */
953 	metric(RTV_MTU, rmx_mtu);
954 	metric(RTV_WEIGHT, rmx_weight);
955 	/* Userland -> kernel timebase conversion. */
956 	if (which & RTV_EXPIRE)
957 		out->rmx_expire = in->rmx_expire ?
958 		    in->rmx_expire - time_second + time_uptime : 0;
959 #undef metric
960 }
961 
962 static void
963 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
964 {
965 #define metric(e) out->e = in->e;
966 	bzero(out, sizeof(*out));
967 	metric(rmx_mtu);
968 	metric(rmx_weight);
969 	/* Kernel -> userland timebase conversion. */
970 	out->rmx_expire = in->rmx_expire ?
971 	    in->rmx_expire - time_uptime + time_second : 0;
972 #undef metric
973 }
974 
975 /*
976  * Extract the addresses of the passed sockaddrs.
977  * Do a little sanity checking so as to avoid bad memory references.
978  * This data is derived straight from userland.
979  */
980 static int
981 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
982 {
983 	struct sockaddr *sa;
984 	int i;
985 
986 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
987 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
988 			continue;
989 		sa = (struct sockaddr *)cp;
990 		/*
991 		 * It won't fit.
992 		 */
993 		if (cp + sa->sa_len > cplim)
994 			return (EINVAL);
995 		/*
996 		 * there are no more.. quit now
997 		 * If there are more bits, they are in error.
998 		 * I've seen this. route(1) can evidently generate these.
999 		 * This causes kernel to core dump.
1000 		 * for compatibility, If we see this, point to a safe address.
1001 		 */
1002 		if (sa->sa_len == 0) {
1003 			rtinfo->rti_info[i] = &sa_zero;
1004 			return (0); /* should be EINVAL but for compat */
1005 		}
1006 		/* accept it */
1007 		rtinfo->rti_info[i] = sa;
1008 		cp += SA_SIZE(sa);
1009 	}
1010 	return (0);
1011 }
1012 
1013 static struct mbuf *
1014 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1015 {
1016 	struct rt_msghdr *rtm;
1017 	struct mbuf *m;
1018 	int i;
1019 	struct sockaddr *sa;
1020 	int len, dlen;
1021 
1022 	switch (type) {
1023 
1024 	case RTM_DELADDR:
1025 	case RTM_NEWADDR:
1026 		len = sizeof(struct ifa_msghdr);
1027 		break;
1028 
1029 	case RTM_DELMADDR:
1030 	case RTM_NEWMADDR:
1031 		len = sizeof(struct ifma_msghdr);
1032 		break;
1033 
1034 	case RTM_IFINFO:
1035 		len = sizeof(struct if_msghdr);
1036 		break;
1037 
1038 	case RTM_IFANNOUNCE:
1039 	case RTM_IEEE80211:
1040 		len = sizeof(struct if_announcemsghdr);
1041 		break;
1042 
1043 	default:
1044 		len = sizeof(struct rt_msghdr);
1045 	}
1046 	if (len > MCLBYTES)
1047 		panic("rt_msg1");
1048 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1049 	if (m && len > MHLEN) {
1050 		MCLGET(m, M_DONTWAIT);
1051 		if ((m->m_flags & M_EXT) == 0) {
1052 			m_free(m);
1053 			m = NULL;
1054 		}
1055 	}
1056 	if (m == NULL)
1057 		return (m);
1058 	m->m_pkthdr.len = m->m_len = len;
1059 	m->m_pkthdr.rcvif = NULL;
1060 	rtm = mtod(m, struct rt_msghdr *);
1061 	bzero((caddr_t)rtm, len);
1062 	for (i = 0; i < RTAX_MAX; i++) {
1063 		if ((sa = rtinfo->rti_info[i]) == NULL)
1064 			continue;
1065 		rtinfo->rti_addrs |= (1 << i);
1066 		dlen = SA_SIZE(sa);
1067 		m_copyback(m, len, dlen, (caddr_t)sa);
1068 		len += dlen;
1069 	}
1070 	if (m->m_pkthdr.len != len) {
1071 		m_freem(m);
1072 		return (NULL);
1073 	}
1074 	rtm->rtm_msglen = len;
1075 	rtm->rtm_version = RTM_VERSION;
1076 	rtm->rtm_type = type;
1077 	return (m);
1078 }
1079 
1080 static int
1081 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1082 {
1083 	int i;
1084 	int len, dlen, second_time = 0;
1085 	caddr_t cp0;
1086 
1087 	rtinfo->rti_addrs = 0;
1088 again:
1089 	switch (type) {
1090 
1091 	case RTM_DELADDR:
1092 	case RTM_NEWADDR:
1093 		len = sizeof(struct ifa_msghdr);
1094 		break;
1095 
1096 	case RTM_IFINFO:
1097 #ifdef COMPAT_FREEBSD32
1098 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1099 			len = sizeof(struct if_msghdr32);
1100 			break;
1101 		}
1102 #endif
1103 		len = sizeof(struct if_msghdr);
1104 		break;
1105 
1106 	case RTM_NEWMADDR:
1107 		len = sizeof(struct ifma_msghdr);
1108 		break;
1109 
1110 	default:
1111 		len = sizeof(struct rt_msghdr);
1112 	}
1113 	cp0 = cp;
1114 	if (cp0)
1115 		cp += len;
1116 	for (i = 0; i < RTAX_MAX; i++) {
1117 		struct sockaddr *sa;
1118 
1119 		if ((sa = rtinfo->rti_info[i]) == NULL)
1120 			continue;
1121 		rtinfo->rti_addrs |= (1 << i);
1122 		dlen = SA_SIZE(sa);
1123 		if (cp) {
1124 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1125 			cp += dlen;
1126 		}
1127 		len += dlen;
1128 	}
1129 	len = ALIGN(len);
1130 	if (cp == NULL && w != NULL && !second_time) {
1131 		struct walkarg *rw = w;
1132 
1133 		if (rw->w_req) {
1134 			if (rw->w_tmemsize < len) {
1135 				if (rw->w_tmem)
1136 					free(rw->w_tmem, M_RTABLE);
1137 				rw->w_tmem = (caddr_t)
1138 					malloc(len, M_RTABLE, M_NOWAIT);
1139 				if (rw->w_tmem)
1140 					rw->w_tmemsize = len;
1141 			}
1142 			if (rw->w_tmem) {
1143 				cp = rw->w_tmem;
1144 				second_time = 1;
1145 				goto again;
1146 			}
1147 		}
1148 	}
1149 	if (cp) {
1150 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1151 
1152 		rtm->rtm_version = RTM_VERSION;
1153 		rtm->rtm_type = type;
1154 		rtm->rtm_msglen = len;
1155 	}
1156 	return (len);
1157 }
1158 
1159 /*
1160  * This routine is called to generate a message from the routing
1161  * socket indicating that a redirect has occured, a routing lookup
1162  * has failed, or that a protocol has detected timeouts to a particular
1163  * destination.
1164  */
1165 void
1166 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1167     int fibnum)
1168 {
1169 	struct rt_msghdr *rtm;
1170 	struct mbuf *m;
1171 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1172 
1173 	if (route_cb.any_count == 0)
1174 		return;
1175 	m = rt_msg1(type, rtinfo);
1176 	if (m == NULL)
1177 		return;
1178 
1179 	if (fibnum != RTS_ALLFIBS) {
1180 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1181 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1182 		M_SETFIB(m, fibnum);
1183 		m->m_flags |= RTS_FILTER_FIB;
1184 	}
1185 
1186 	rtm = mtod(m, struct rt_msghdr *);
1187 	rtm->rtm_flags = RTF_DONE | flags;
1188 	rtm->rtm_errno = error;
1189 	rtm->rtm_addrs = rtinfo->rti_addrs;
1190 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1191 }
1192 
1193 void
1194 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1195 {
1196 
1197 	rt_missmsg_fib(type, rtinfo, flags, error, RTS_ALLFIBS);
1198 }
1199 
1200 /*
1201  * This routine is called to generate a message from the routing
1202  * socket indicating that the status of a network interface has changed.
1203  */
1204 void
1205 rt_ifmsg(struct ifnet *ifp)
1206 {
1207 	struct if_msghdr *ifm;
1208 	struct mbuf *m;
1209 	struct rt_addrinfo info;
1210 
1211 	if (route_cb.any_count == 0)
1212 		return;
1213 	bzero((caddr_t)&info, sizeof(info));
1214 	m = rt_msg1(RTM_IFINFO, &info);
1215 	if (m == NULL)
1216 		return;
1217 	ifm = mtod(m, struct if_msghdr *);
1218 	ifm->ifm_index = ifp->if_index;
1219 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1220 	ifm->ifm_data = ifp->if_data;
1221 	ifm->ifm_addrs = 0;
1222 	rt_dispatch(m, AF_UNSPEC);
1223 }
1224 
1225 /*
1226  * This is called to generate messages from the routing socket
1227  * indicating a network interface has had addresses associated with it.
1228  * if we ever reverse the logic and replace messages TO the routing
1229  * socket indicate a request to configure interfaces, then it will
1230  * be unnecessary as the routing socket will automatically generate
1231  * copies of it.
1232  */
1233 void
1234 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
1235     int fibnum)
1236 {
1237 	struct rt_addrinfo info;
1238 	struct sockaddr *sa = NULL;
1239 	int pass;
1240 	struct mbuf *m = NULL;
1241 	struct ifnet *ifp = ifa->ifa_ifp;
1242 
1243 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1244 		("unexpected cmd %u", cmd));
1245 #if defined(INET) || defined(INET6)
1246 #ifdef SCTP
1247 	/*
1248 	 * notify the SCTP stack
1249 	 * this will only get called when an address is added/deleted
1250 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1251 	 */
1252 	sctp_addr_change(ifa, cmd);
1253 #endif /* SCTP */
1254 #endif
1255 	if (route_cb.any_count == 0)
1256 		return;
1257 	for (pass = 1; pass < 3; pass++) {
1258 		bzero((caddr_t)&info, sizeof(info));
1259 		if ((cmd == RTM_ADD && pass == 1) ||
1260 		    (cmd == RTM_DELETE && pass == 2)) {
1261 			struct ifa_msghdr *ifam;
1262 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1263 
1264 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1265 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1266 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1267 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1268 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1269 				continue;
1270 			ifam = mtod(m, struct ifa_msghdr *);
1271 			ifam->ifam_index = ifp->if_index;
1272 			ifam->ifam_metric = ifa->ifa_metric;
1273 			ifam->ifam_flags = ifa->ifa_flags;
1274 			ifam->ifam_addrs = info.rti_addrs;
1275 		}
1276 		if ((cmd == RTM_ADD && pass == 2) ||
1277 		    (cmd == RTM_DELETE && pass == 1)) {
1278 			struct rt_msghdr *rtm;
1279 
1280 			if (rt == NULL)
1281 				continue;
1282 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1283 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1284 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1285 			if ((m = rt_msg1(cmd, &info)) == NULL)
1286 				continue;
1287 			rtm = mtod(m, struct rt_msghdr *);
1288 			rtm->rtm_index = ifp->if_index;
1289 			rtm->rtm_flags |= rt->rt_flags;
1290 			rtm->rtm_errno = error;
1291 			rtm->rtm_addrs = info.rti_addrs;
1292 		}
1293 		if (fibnum != RTS_ALLFIBS) {
1294 			KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: "
1295 			    "fibnum out of range 0 <= %d < %d", __func__,
1296 			     fibnum, rt_numfibs));
1297 			M_SETFIB(m, fibnum);
1298 			m->m_flags |= RTS_FILTER_FIB;
1299 		}
1300 		rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1301 	}
1302 }
1303 
1304 void
1305 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1306 {
1307 
1308 	rt_newaddrmsg_fib(cmd, ifa, error, rt, RTS_ALLFIBS);
1309 }
1310 
1311 /*
1312  * This is the analogue to the rt_newaddrmsg which performs the same
1313  * function but for multicast group memberhips.  This is easier since
1314  * there is no route state to worry about.
1315  */
1316 void
1317 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1318 {
1319 	struct rt_addrinfo info;
1320 	struct mbuf *m = NULL;
1321 	struct ifnet *ifp = ifma->ifma_ifp;
1322 	struct ifma_msghdr *ifmam;
1323 
1324 	if (route_cb.any_count == 0)
1325 		return;
1326 
1327 	bzero((caddr_t)&info, sizeof(info));
1328 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1329 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1330 	/*
1331 	 * If a link-layer address is present, present it as a ``gateway''
1332 	 * (similarly to how ARP entries, e.g., are presented).
1333 	 */
1334 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1335 	m = rt_msg1(cmd, &info);
1336 	if (m == NULL)
1337 		return;
1338 	ifmam = mtod(m, struct ifma_msghdr *);
1339 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1340 	    __func__));
1341 	ifmam->ifmam_index = ifp->if_index;
1342 	ifmam->ifmam_addrs = info.rti_addrs;
1343 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1344 }
1345 
1346 static struct mbuf *
1347 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1348 	struct rt_addrinfo *info)
1349 {
1350 	struct if_announcemsghdr *ifan;
1351 	struct mbuf *m;
1352 
1353 	if (route_cb.any_count == 0)
1354 		return NULL;
1355 	bzero((caddr_t)info, sizeof(*info));
1356 	m = rt_msg1(type, info);
1357 	if (m != NULL) {
1358 		ifan = mtod(m, struct if_announcemsghdr *);
1359 		ifan->ifan_index = ifp->if_index;
1360 		strlcpy(ifan->ifan_name, ifp->if_xname,
1361 			sizeof(ifan->ifan_name));
1362 		ifan->ifan_what = what;
1363 	}
1364 	return m;
1365 }
1366 
1367 /*
1368  * This is called to generate routing socket messages indicating
1369  * IEEE80211 wireless events.
1370  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1371  */
1372 void
1373 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1374 {
1375 	struct mbuf *m;
1376 	struct rt_addrinfo info;
1377 
1378 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1379 	if (m != NULL) {
1380 		/*
1381 		 * Append the ieee80211 data.  Try to stick it in the
1382 		 * mbuf containing the ifannounce msg; otherwise allocate
1383 		 * a new mbuf and append.
1384 		 *
1385 		 * NB: we assume m is a single mbuf.
1386 		 */
1387 		if (data_len > M_TRAILINGSPACE(m)) {
1388 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1389 			if (n == NULL) {
1390 				m_freem(m);
1391 				return;
1392 			}
1393 			bcopy(data, mtod(n, void *), data_len);
1394 			n->m_len = data_len;
1395 			m->m_next = n;
1396 		} else if (data_len > 0) {
1397 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1398 			m->m_len += data_len;
1399 		}
1400 		if (m->m_flags & M_PKTHDR)
1401 			m->m_pkthdr.len += data_len;
1402 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1403 		rt_dispatch(m, AF_UNSPEC);
1404 	}
1405 }
1406 
1407 /*
1408  * This is called to generate routing socket messages indicating
1409  * network interface arrival and departure.
1410  */
1411 void
1412 rt_ifannouncemsg(struct ifnet *ifp, int what)
1413 {
1414 	struct mbuf *m;
1415 	struct rt_addrinfo info;
1416 
1417 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1418 	if (m != NULL)
1419 		rt_dispatch(m, AF_UNSPEC);
1420 }
1421 
1422 static void
1423 rt_dispatch(struct mbuf *m, sa_family_t saf)
1424 {
1425 	struct m_tag *tag;
1426 
1427 	/*
1428 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1429 	 * use when injecting the mbuf into the routing socket buffer from
1430 	 * the netisr.
1431 	 */
1432 	if (saf != AF_UNSPEC) {
1433 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1434 		    M_NOWAIT);
1435 		if (tag == NULL) {
1436 			m_freem(m);
1437 			return;
1438 		}
1439 		*(unsigned short *)(tag + 1) = saf;
1440 		m_tag_prepend(m, tag);
1441 	}
1442 #ifdef VIMAGE
1443 	if (V_loif)
1444 		m->m_pkthdr.rcvif = V_loif;
1445 	else {
1446 		m_freem(m);
1447 		return;
1448 	}
1449 #endif
1450 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1451 }
1452 
1453 /*
1454  * This is used in dumping the kernel table via sysctl().
1455  */
1456 static int
1457 sysctl_dumpentry(struct radix_node *rn, void *vw)
1458 {
1459 	struct walkarg *w = vw;
1460 	struct rtentry *rt = (struct rtentry *)rn;
1461 	int error = 0, size;
1462 	struct rt_addrinfo info;
1463 
1464 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1465 		return 0;
1466 	if ((rt->rt_flags & RTF_HOST) == 0
1467 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1468 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1469 		return (0);
1470 	bzero((caddr_t)&info, sizeof(info));
1471 	info.rti_info[RTAX_DST] = rt_key(rt);
1472 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1473 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1474 	info.rti_info[RTAX_GENMASK] = 0;
1475 	if (rt->rt_ifp) {
1476 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1477 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1478 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1479 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1480 	}
1481 	size = rt_msg2(RTM_GET, &info, NULL, w);
1482 	if (w->w_req && w->w_tmem) {
1483 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1484 
1485 		rtm->rtm_flags = rt->rt_flags;
1486 		/*
1487 		 * let's be honest about this being a retarded hack
1488 		 */
1489 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1490 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1491 		rtm->rtm_index = rt->rt_ifp->if_index;
1492 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1493 		rtm->rtm_addrs = info.rti_addrs;
1494 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1495 		return (error);
1496 	}
1497 	return (error);
1498 }
1499 
1500 #ifdef COMPAT_FREEBSD32
1501 static void
1502 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1503 {
1504 
1505 	bzero(dst, sizeof(*dst));
1506 	CP(*src, *dst, ifi_type);
1507 	CP(*src, *dst, ifi_physical);
1508 	CP(*src, *dst, ifi_addrlen);
1509 	CP(*src, *dst, ifi_hdrlen);
1510 	CP(*src, *dst, ifi_link_state);
1511 	dst->ifi_datalen = sizeof(struct if_data32);
1512 	CP(*src, *dst, ifi_mtu);
1513 	CP(*src, *dst, ifi_metric);
1514 	CP(*src, *dst, ifi_baudrate);
1515 	CP(*src, *dst, ifi_ipackets);
1516 	CP(*src, *dst, ifi_ierrors);
1517 	CP(*src, *dst, ifi_opackets);
1518 	CP(*src, *dst, ifi_oerrors);
1519 	CP(*src, *dst, ifi_collisions);
1520 	CP(*src, *dst, ifi_ibytes);
1521 	CP(*src, *dst, ifi_obytes);
1522 	CP(*src, *dst, ifi_imcasts);
1523 	CP(*src, *dst, ifi_omcasts);
1524 	CP(*src, *dst, ifi_iqdrops);
1525 	CP(*src, *dst, ifi_noproto);
1526 	CP(*src, *dst, ifi_hwassist);
1527 	CP(*src, *dst, ifi_epoch);
1528 	TV_CP(*src, *dst, ifi_lastchange);
1529 }
1530 #endif
1531 
1532 static int
1533 sysctl_iflist(int af, struct walkarg *w)
1534 {
1535 	struct ifnet *ifp;
1536 	struct ifaddr *ifa;
1537 	struct rt_addrinfo info;
1538 	int len, error = 0;
1539 
1540 	bzero((caddr_t)&info, sizeof(info));
1541 	IFNET_RLOCK();
1542 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1543 		if (w->w_arg && w->w_arg != ifp->if_index)
1544 			continue;
1545 		IF_ADDR_LOCK(ifp);
1546 		ifa = ifp->if_addr;
1547 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1548 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1549 		info.rti_info[RTAX_IFP] = NULL;
1550 		if (w->w_req && w->w_tmem) {
1551 			struct if_msghdr *ifm;
1552 
1553 #ifdef COMPAT_FREEBSD32
1554 			if (w->w_req->flags & SCTL_MASK32) {
1555 				struct if_msghdr32 *ifm32;
1556 
1557 				ifm32 = (struct if_msghdr32 *)w->w_tmem;
1558 				ifm32->ifm_index = ifp->if_index;
1559 				ifm32->ifm_flags = ifp->if_flags |
1560 				    ifp->if_drv_flags;
1561 				copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1562 				ifm32->ifm_addrs = info.rti_addrs;
1563 				error = SYSCTL_OUT(w->w_req, (caddr_t)ifm32,
1564 				    len);
1565 				goto sysctl_out;
1566 			}
1567 #endif
1568 			ifm = (struct if_msghdr *)w->w_tmem;
1569 			ifm->ifm_index = ifp->if_index;
1570 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1571 			ifm->ifm_data = ifp->if_data;
1572 			ifm->ifm_addrs = info.rti_addrs;
1573 			error = SYSCTL_OUT(w->w_req, (caddr_t)ifm, len);
1574 #ifdef COMPAT_FREEBSD32
1575 		sysctl_out:
1576 #endif
1577 			if (error)
1578 				goto done;
1579 		}
1580 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1581 			if (af && af != ifa->ifa_addr->sa_family)
1582 				continue;
1583 			if (prison_if(w->w_req->td->td_ucred,
1584 			    ifa->ifa_addr) != 0)
1585 				continue;
1586 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1587 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1588 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1589 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1590 			if (w->w_req && w->w_tmem) {
1591 				struct ifa_msghdr *ifam;
1592 
1593 				ifam = (struct ifa_msghdr *)w->w_tmem;
1594 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1595 				ifam->ifam_flags = ifa->ifa_flags;
1596 				ifam->ifam_metric = ifa->ifa_metric;
1597 				ifam->ifam_addrs = info.rti_addrs;
1598 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1599 				if (error)
1600 					goto done;
1601 			}
1602 		}
1603 		IF_ADDR_UNLOCK(ifp);
1604 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1605 			info.rti_info[RTAX_BRD] = NULL;
1606 	}
1607 done:
1608 	if (ifp != NULL)
1609 		IF_ADDR_UNLOCK(ifp);
1610 	IFNET_RUNLOCK();
1611 	return (error);
1612 }
1613 
1614 static int
1615 sysctl_ifmalist(int af, struct walkarg *w)
1616 {
1617 	struct ifnet *ifp;
1618 	struct ifmultiaddr *ifma;
1619 	struct	rt_addrinfo info;
1620 	int	len, error = 0;
1621 	struct ifaddr *ifa;
1622 
1623 	bzero((caddr_t)&info, sizeof(info));
1624 	IFNET_RLOCK();
1625 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1626 		if (w->w_arg && w->w_arg != ifp->if_index)
1627 			continue;
1628 		ifa = ifp->if_addr;
1629 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1630 		IF_ADDR_LOCK(ifp);
1631 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1632 			if (af && af != ifma->ifma_addr->sa_family)
1633 				continue;
1634 			if (prison_if(w->w_req->td->td_ucred,
1635 			    ifma->ifma_addr) != 0)
1636 				continue;
1637 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1638 			info.rti_info[RTAX_GATEWAY] =
1639 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1640 			    ifma->ifma_lladdr : NULL;
1641 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1642 			if (w->w_req && w->w_tmem) {
1643 				struct ifma_msghdr *ifmam;
1644 
1645 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1646 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1647 				ifmam->ifmam_flags = 0;
1648 				ifmam->ifmam_addrs = info.rti_addrs;
1649 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1650 				if (error) {
1651 					IF_ADDR_UNLOCK(ifp);
1652 					goto done;
1653 				}
1654 			}
1655 		}
1656 		IF_ADDR_UNLOCK(ifp);
1657 	}
1658 done:
1659 	IFNET_RUNLOCK();
1660 	return (error);
1661 }
1662 
1663 static int
1664 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1665 {
1666 	int	*name = (int *)arg1;
1667 	u_int	namelen = arg2;
1668 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1669 	int	i, lim, error = EINVAL;
1670 	u_char	af;
1671 	struct	walkarg w;
1672 
1673 	name ++;
1674 	namelen--;
1675 	if (req->newptr)
1676 		return (EPERM);
1677 	if (namelen != 3)
1678 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1679 	af = name[0];
1680 	if (af > AF_MAX)
1681 		return (EINVAL);
1682 	bzero(&w, sizeof(w));
1683 	w.w_op = name[1];
1684 	w.w_arg = name[2];
1685 	w.w_req = req;
1686 
1687 	error = sysctl_wire_old_buffer(req, 0);
1688 	if (error)
1689 		return (error);
1690 	switch (w.w_op) {
1691 
1692 	case NET_RT_DUMP:
1693 	case NET_RT_FLAGS:
1694 		if (af == 0) {			/* dump all tables */
1695 			i = 1;
1696 			lim = AF_MAX;
1697 		} else				/* dump only one table */
1698 			i = lim = af;
1699 
1700 		/*
1701 		 * take care of llinfo entries, the caller must
1702 		 * specify an AF
1703 		 */
1704 		if (w.w_op == NET_RT_FLAGS &&
1705 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1706 			if (af != 0)
1707 				error = lltable_sysctl_dumparp(af, w.w_req);
1708 			else
1709 				error = EINVAL;
1710 			break;
1711 		}
1712 		/*
1713 		 * take care of routing entries
1714 		 */
1715 		for (error = 0; error == 0 && i <= lim; i++) {
1716 			rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1717 			if (rnh != NULL) {
1718 				RADIX_NODE_HEAD_LOCK(rnh);
1719 			    	error = rnh->rnh_walktree(rnh,
1720 				    sysctl_dumpentry, &w);
1721 				RADIX_NODE_HEAD_UNLOCK(rnh);
1722 			} else if (af != 0)
1723 				error = EAFNOSUPPORT;
1724 		}
1725 		break;
1726 
1727 	case NET_RT_IFLIST:
1728 		error = sysctl_iflist(af, &w);
1729 		break;
1730 
1731 	case NET_RT_IFMALIST:
1732 		error = sysctl_ifmalist(af, &w);
1733 		break;
1734 	}
1735 	if (w.w_tmem)
1736 		free(w.w_tmem, M_RTABLE);
1737 	return (error);
1738 }
1739 
1740 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1741 
1742 /*
1743  * Definitions of protocols supported in the ROUTE domain.
1744  */
1745 
1746 static struct domain routedomain;		/* or at least forward */
1747 
1748 static struct protosw routesw[] = {
1749 {
1750 	.pr_type =		SOCK_RAW,
1751 	.pr_domain =		&routedomain,
1752 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1753 	.pr_output =		route_output,
1754 	.pr_ctlinput =		raw_ctlinput,
1755 	.pr_init =		raw_init,
1756 	.pr_usrreqs =		&route_usrreqs
1757 }
1758 };
1759 
1760 static struct domain routedomain = {
1761 	.dom_family =		PF_ROUTE,
1762 	.dom_name =		 "route",
1763 	.dom_protosw =		routesw,
1764 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1765 };
1766 
1767 VNET_DOMAIN_SET(route);
1768