xref: /freebsd/sys/net/rtsock.c (revision 33644623554bb0fc57ed3c7d874193a498679b22)
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/domain.h>
39 #include <sys/jail.h>
40 #include <sys/kernel.h>
41 #include <sys/malloc.h>
42 #include <sys/mbuf.h>
43 #include <sys/priv.h>
44 #include <sys/proc.h>
45 #include <sys/protosw.h>
46 #include <sys/signalvar.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/sysctl.h>
50 #include <sys/systm.h>
51 #include <sys/vimage.h>
52 
53 #include <net/if.h>
54 #include <net/netisr.h>
55 #include <net/raw_cb.h>
56 #include <net/route.h>
57 
58 #include <netinet/in.h>
59 #ifdef INET6
60 #include <netinet6/scope6_var.h>
61 #endif
62 
63 #ifdef SCTP
64 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
65 #endif /* SCTP */
66 
67 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
68 
69 /* NB: these are not modified */
70 static struct	sockaddr route_src = { 2, PF_ROUTE, };
71 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
72 
73 static struct {
74 	int	ip_count;	/* attached w/ AF_INET */
75 	int	ip6_count;	/* attached w/ AF_INET6 */
76 	int	ipx_count;	/* attached w/ AF_IPX */
77 	int	any_count;	/* total attached */
78 } route_cb;
79 
80 struct mtx rtsock_mtx;
81 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
82 
83 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
84 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
85 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
86 
87 static struct	ifqueue rtsintrq;
88 
89 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
90 SYSCTL_INT(_net_route, OID_AUTO, netisr_maxqlen, CTLFLAG_RW,
91     &rtsintrq.ifq_maxlen, 0, "maximum routing socket dispatch queue length");
92 
93 struct walkarg {
94 	int	w_tmemsize;
95 	int	w_op, w_arg;
96 	caddr_t	w_tmem;
97 	struct sysctl_req *w_req;
98 };
99 
100 static void	rts_input(struct mbuf *m);
101 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
102 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
103 			caddr_t cp, struct walkarg *w);
104 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
105 			struct rt_addrinfo *rtinfo);
106 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
107 static int	sysctl_iflist(int af, struct walkarg *w);
108 static int	sysctl_ifmalist(int af, struct walkarg *w);
109 static int	route_output(struct mbuf *m, struct socket *so);
110 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
111 			struct rt_metrics_lite *out);
112 static void	rt_getmetrics(const struct rt_metrics_lite *in,
113 			struct rt_metrics *out);
114 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
115 
116 static void
117 rts_init(void)
118 {
119 	int tmp;
120 
121 	rtsintrq.ifq_maxlen = 256;
122 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
123 		rtsintrq.ifq_maxlen = tmp;
124 	mtx_init(&rtsintrq.ifq_mtx, "rts_inq", NULL, MTX_DEF);
125 	netisr_register(NETISR_ROUTE, rts_input, &rtsintrq, 0);
126 }
127 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
128 
129 static void
130 rts_input(struct mbuf *m)
131 {
132 	struct sockproto route_proto;
133 	unsigned short *family;
134 	struct m_tag *tag;
135 
136 	route_proto.sp_family = PF_ROUTE;
137 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
138 	if (tag != NULL) {
139 		family = (unsigned short *)(tag + 1);
140 		route_proto.sp_protocol = *family;
141 		m_tag_delete(m, tag);
142 	} else
143 		route_proto.sp_protocol = 0;
144 
145 	raw_input(m, &route_proto, &route_src);
146 }
147 
148 /*
149  * It really doesn't make any sense at all for this code to share much
150  * with raw_usrreq.c, since its functionality is so restricted.  XXX
151  */
152 static void
153 rts_abort(struct socket *so)
154 {
155 
156 	raw_usrreqs.pru_abort(so);
157 }
158 
159 static void
160 rts_close(struct socket *so)
161 {
162 
163 	raw_usrreqs.pru_close(so);
164 }
165 
166 /* pru_accept is EOPNOTSUPP */
167 
168 static int
169 rts_attach(struct socket *so, int proto, struct thread *td)
170 {
171 	struct rawcb *rp;
172 	int s, error;
173 
174 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
175 
176 	/* XXX */
177 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
178 	if (rp == NULL)
179 		return ENOBUFS;
180 
181 	/*
182 	 * The splnet() is necessary to block protocols from sending
183 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
184 	 * this PCB is extant but incompletely initialized.
185 	 * Probably we should try to do more of this work beforehand and
186 	 * eliminate the spl.
187 	 */
188 	s = splnet();
189 	so->so_pcb = (caddr_t)rp;
190 	so->so_fibnum = td->td_proc->p_fibnum;
191 	error = raw_attach(so, proto);
192 	rp = sotorawcb(so);
193 	if (error) {
194 		splx(s);
195 		so->so_pcb = NULL;
196 		free(rp, M_PCB);
197 		return error;
198 	}
199 	RTSOCK_LOCK();
200 	switch(rp->rcb_proto.sp_protocol) {
201 	case AF_INET:
202 		route_cb.ip_count++;
203 		break;
204 	case AF_INET6:
205 		route_cb.ip6_count++;
206 		break;
207 	case AF_IPX:
208 		route_cb.ipx_count++;
209 		break;
210 	}
211 	route_cb.any_count++;
212 	RTSOCK_UNLOCK();
213 	soisconnected(so);
214 	so->so_options |= SO_USELOOPBACK;
215 	splx(s);
216 	return 0;
217 }
218 
219 static int
220 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
221 {
222 
223 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
224 }
225 
226 static int
227 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
228 {
229 
230 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
231 }
232 
233 /* pru_connect2 is EOPNOTSUPP */
234 /* pru_control is EOPNOTSUPP */
235 
236 static void
237 rts_detach(struct socket *so)
238 {
239 	struct rawcb *rp = sotorawcb(so);
240 
241 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
242 
243 	RTSOCK_LOCK();
244 	switch(rp->rcb_proto.sp_protocol) {
245 	case AF_INET:
246 		route_cb.ip_count--;
247 		break;
248 	case AF_INET6:
249 		route_cb.ip6_count--;
250 		break;
251 	case AF_IPX:
252 		route_cb.ipx_count--;
253 		break;
254 	}
255 	route_cb.any_count--;
256 	RTSOCK_UNLOCK();
257 	raw_usrreqs.pru_detach(so);
258 }
259 
260 static int
261 rts_disconnect(struct socket *so)
262 {
263 
264 	return (raw_usrreqs.pru_disconnect(so));
265 }
266 
267 /* pru_listen is EOPNOTSUPP */
268 
269 static int
270 rts_peeraddr(struct socket *so, struct sockaddr **nam)
271 {
272 
273 	return (raw_usrreqs.pru_peeraddr(so, nam));
274 }
275 
276 /* pru_rcvd is EOPNOTSUPP */
277 /* pru_rcvoob is EOPNOTSUPP */
278 
279 static int
280 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
281 	 struct mbuf *control, struct thread *td)
282 {
283 
284 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
285 }
286 
287 /* pru_sense is null */
288 
289 static int
290 rts_shutdown(struct socket *so)
291 {
292 
293 	return (raw_usrreqs.pru_shutdown(so));
294 }
295 
296 static int
297 rts_sockaddr(struct socket *so, struct sockaddr **nam)
298 {
299 
300 	return (raw_usrreqs.pru_sockaddr(so, nam));
301 }
302 
303 static struct pr_usrreqs route_usrreqs = {
304 	.pru_abort =		rts_abort,
305 	.pru_attach =		rts_attach,
306 	.pru_bind =		rts_bind,
307 	.pru_connect =		rts_connect,
308 	.pru_detach =		rts_detach,
309 	.pru_disconnect =	rts_disconnect,
310 	.pru_peeraddr =		rts_peeraddr,
311 	.pru_send =		rts_send,
312 	.pru_shutdown =		rts_shutdown,
313 	.pru_sockaddr =		rts_sockaddr,
314 	.pru_close =		rts_close,
315 };
316 
317 #ifndef _SOCKADDR_UNION_DEFINED
318 #define	_SOCKADDR_UNION_DEFINED
319 /*
320  * The union of all possible address formats we handle.
321  */
322 union sockaddr_union {
323 	struct sockaddr		sa;
324 	struct sockaddr_in	sin;
325 	struct sockaddr_in6	sin6;
326 };
327 #endif /* _SOCKADDR_UNION_DEFINED */
328 
329 static int
330 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
331     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
332 {
333 
334 	switch (info->rti_info[RTAX_DST]->sa_family) {
335 #ifdef INET
336 	case AF_INET:
337 	{
338 		struct in_addr ia;
339 
340 		/*
341 		 * 1. Check if the returned address is part of the jail.
342 		 */
343 		ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->sin_addr;
344 		if (prison_check_ip4(cred, &ia) != 0) {
345 			info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
346 
347 		} else {
348 			struct ifaddr *ifa;
349 			int found;
350 
351 			found = 0;
352 
353 			/*
354 			 * 2. Try to find an address on the given outgoing
355 			 *    interface that belongs to the jail.
356 			 */
357 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
358 				struct sockaddr *sa;
359 				sa = ifa->ifa_addr;
360 				if (sa->sa_family != AF_INET)
361 					continue;
362 				ia = ((struct sockaddr_in *)sa)->sin_addr;
363 				if (prison_check_ip4(cred, &ia) != 0) {
364 					found = 1;
365 					break;
366 				}
367 			}
368 			if (!found) {
369 				/*
370 				 * 3. As a last resort return the 'default'
371 				 * jail address.
372 				 */
373 				if (prison_getip4(cred, &ia) != 0)
374 					return (ESRCH);
375 			}
376 			bzero(&saun->sin, sizeof(struct sockaddr_in));
377 			saun->sin.sin_len = sizeof(struct sockaddr_in);
378 			saun->sin.sin_family = AF_INET;
379 			saun->sin.sin_addr.s_addr = ia.s_addr;
380 			info->rti_info[RTAX_IFA] =
381 			    (struct sockaddr *)&saun->sin;
382 		}
383 		break;
384 	}
385 #endif
386 #ifdef INET6
387 	case AF_INET6:
388 	{
389 		struct in6_addr ia6;
390 
391 		/*
392 		 * 1. Check if the returned address is part of the jail.
393 		 */
394 		bcopy(&((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->sin6_addr,
395 		    &ia6, sizeof(struct in6_addr));
396 		if (prison_check_ip6(cred, &ia6) != 0) {
397 			info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
398 		} else {
399 			struct ifaddr *ifa;
400 			int found;
401 
402 			found = 0;
403 
404 			/*
405 			 * 2. Try to find an address on the given outgoing
406 			 *    interface that belongs to the jail.
407 			 */
408 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
409 				struct sockaddr *sa;
410 				sa = ifa->ifa_addr;
411 				if (sa->sa_family != AF_INET6)
412 					continue;
413 				bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
414 				    &ia6, sizeof(struct in6_addr));
415 				if (prison_check_ip6(cred, &ia6) != 0) {
416 					found = 1;
417 					break;
418 				}
419 			}
420 			if (!found) {
421 				/*
422 				 * 3. As a last resort return the 'default'
423 				 * jail address.
424 				 */
425 				if (prison_getip6(cred, &ia6) != 0)
426 					return (ESRCH);
427 			}
428 			bzero(&saun->sin6, sizeof(struct sockaddr_in6));
429 			saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
430 			saun->sin6.sin6_family = AF_INET6;
431 			bcopy(&ia6, &saun->sin6.sin6_addr,
432 			    sizeof(struct in6_addr));
433 			if (sa6_recoverscope(&saun->sin6) != 0)
434 				return (ESRCH);
435 			info->rti_info[RTAX_IFA] =
436 			    (struct sockaddr *)&saun->sin6;
437 		}
438 		break;
439 	}
440 #endif
441 	default:
442 		return (ESRCH);
443 	}
444 	return (0);
445 }
446 
447 /*ARGSUSED*/
448 static int
449 route_output(struct mbuf *m, struct socket *so)
450 {
451 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
452 	INIT_VNET_NET(so->so_vnet);
453 	struct rt_msghdr *rtm = NULL;
454 	struct rtentry *rt = NULL;
455 	struct radix_node_head *rnh;
456 	struct rt_addrinfo info;
457 	int len, error = 0;
458 	struct ifnet *ifp = NULL;
459 	union sockaddr_union saun;
460 
461 #define senderr(e) { error = e; goto flush;}
462 	if (m == NULL || ((m->m_len < sizeof(long)) &&
463 		       (m = m_pullup(m, sizeof(long))) == NULL))
464 		return (ENOBUFS);
465 	if ((m->m_flags & M_PKTHDR) == 0)
466 		panic("route_output");
467 	len = m->m_pkthdr.len;
468 	if (len < sizeof(*rtm) ||
469 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
470 		info.rti_info[RTAX_DST] = NULL;
471 		senderr(EINVAL);
472 	}
473 	R_Malloc(rtm, struct rt_msghdr *, len);
474 	if (rtm == NULL) {
475 		info.rti_info[RTAX_DST] = NULL;
476 		senderr(ENOBUFS);
477 	}
478 	m_copydata(m, 0, len, (caddr_t)rtm);
479 	if (rtm->rtm_version != RTM_VERSION) {
480 		info.rti_info[RTAX_DST] = NULL;
481 		senderr(EPROTONOSUPPORT);
482 	}
483 	rtm->rtm_pid = curproc->p_pid;
484 	bzero(&info, sizeof(info));
485 	info.rti_addrs = rtm->rtm_addrs;
486 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
487 		info.rti_info[RTAX_DST] = NULL;
488 		senderr(EINVAL);
489 	}
490 	info.rti_flags = rtm->rtm_flags;
491 	if (info.rti_info[RTAX_DST] == NULL ||
492 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
493 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
494 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
495 		senderr(EINVAL);
496 	if (info.rti_info[RTAX_GENMASK]) {
497 		struct radix_node *t;
498 		t = rn_addmask((caddr_t) info.rti_info[RTAX_GENMASK], 0, 1);
499 		if (t != NULL &&
500 		    bcmp((char *)(void *)info.rti_info[RTAX_GENMASK] + 1,
501 		    (char *)(void *)t->rn_key + 1,
502 		    ((struct sockaddr *)t->rn_key)->sa_len - 1) == 0)
503 			info.rti_info[RTAX_GENMASK] =
504 			    (struct sockaddr *)t->rn_key;
505 		else
506 			senderr(ENOBUFS);
507 	}
508 
509 	/*
510 	 * Verify that the caller has the appropriate privilege; RTM_GET
511 	 * is the only operation the non-superuser is allowed.
512 	 */
513 	if (rtm->rtm_type != RTM_GET) {
514 		error = priv_check(curthread, PRIV_NET_ROUTE);
515 		if (error)
516 			senderr(error);
517 	}
518 
519 	switch (rtm->rtm_type) {
520 		struct rtentry *saved_nrt;
521 
522 	case RTM_ADD:
523 		if (info.rti_info[RTAX_GATEWAY] == NULL)
524 			senderr(EINVAL);
525 		saved_nrt = NULL;
526 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
527 		    so->so_fibnum);
528 		if (error == 0 && saved_nrt) {
529 			RT_LOCK(saved_nrt);
530 			rt_setmetrics(rtm->rtm_inits,
531 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
532 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
533 			RT_REMREF(saved_nrt);
534 			saved_nrt->rt_genmask = info.rti_info[RTAX_GENMASK];
535 			RT_UNLOCK(saved_nrt);
536 		}
537 		break;
538 
539 	case RTM_DELETE:
540 		saved_nrt = NULL;
541 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
542 		    so->so_fibnum);
543 		if (error == 0) {
544 			RT_LOCK(saved_nrt);
545 			rt = saved_nrt;
546 			goto report;
547 		}
548 		break;
549 
550 	case RTM_GET:
551 	case RTM_CHANGE:
552 	case RTM_LOCK:
553 		rnh = V_rt_tables[so->so_fibnum][info.rti_info[RTAX_DST]->sa_family];
554 		if (rnh == NULL)
555 			senderr(EAFNOSUPPORT);
556 		RADIX_NODE_HEAD_LOCK(rnh);
557 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
558 			info.rti_info[RTAX_NETMASK], rnh);
559 		if (rt == NULL) {	/* XXX looks bogus */
560 			RADIX_NODE_HEAD_UNLOCK(rnh);
561 			senderr(ESRCH);
562 		}
563 #ifdef RADIX_MPATH
564 		/*
565 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
566 		 * we require users to specify a matching RTAX_GATEWAY.
567 		 *
568 		 * for RTM_GET, gate is optional even with multipath.
569 		 * if gate == NULL the first match is returned.
570 		 * (no need to call rt_mpath_matchgate if gate == NULL)
571 		 */
572 		if (rn_mpath_capable(rnh) &&
573 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
574 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
575 			if (!rt) {
576 				RADIX_NODE_HEAD_UNLOCK(rnh);
577 				senderr(ESRCH);
578 			}
579 		}
580 #endif
581 		RT_LOCK(rt);
582 		RT_ADDREF(rt);
583 		RADIX_NODE_HEAD_UNLOCK(rnh);
584 
585 		/*
586 		 * Fix for PR: 82974
587 		 *
588 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
589 		 * returns a perfect match in case a netmask is
590 		 * specified.  For host routes only a longest prefix
591 		 * match is returned so it is necessary to compare the
592 		 * existence of the netmask.  If both have a netmask
593 		 * rnh_lookup() did a perfect match and if none of them
594 		 * have a netmask both are host routes which is also a
595 		 * perfect match.
596 		 */
597 
598 		if (rtm->rtm_type != RTM_GET &&
599 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
600 			RT_UNLOCK(rt);
601 			senderr(ESRCH);
602 		}
603 
604 		switch(rtm->rtm_type) {
605 
606 		case RTM_GET:
607 		report:
608 			RT_LOCK_ASSERT(rt);
609 			info.rti_info[RTAX_DST] = rt_key(rt);
610 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
611 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
612 			info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
613 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
614 				ifp = rt->rt_ifp;
615 				if (ifp) {
616 					info.rti_info[RTAX_IFP] =
617 					    ifp->if_addr->ifa_addr;
618 					if (jailed(so->so_cred)) {
619 						error = rtm_get_jailed(
620 						    &info, ifp, rt, &saun,
621 						    so->so_cred);
622 						if (error != 0) {
623 							RT_UNLOCK(rt);
624 							senderr(ESRCH);
625 						}
626 					} else {
627 						info.rti_info[RTAX_IFA] =
628 						    rt->rt_ifa->ifa_addr;
629 					}
630 					if (ifp->if_flags & IFF_POINTOPOINT)
631 						info.rti_info[RTAX_BRD] =
632 						    rt->rt_ifa->ifa_dstaddr;
633 					rtm->rtm_index = ifp->if_index;
634 				} else {
635 					info.rti_info[RTAX_IFP] = NULL;
636 					info.rti_info[RTAX_IFA] = NULL;
637 				}
638 			} else if ((ifp = rt->rt_ifp) != NULL) {
639 				rtm->rtm_index = ifp->if_index;
640 			}
641 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
642 			if (len > rtm->rtm_msglen) {
643 				struct rt_msghdr *new_rtm;
644 				R_Malloc(new_rtm, struct rt_msghdr *, len);
645 				if (new_rtm == NULL) {
646 					RT_UNLOCK(rt);
647 					senderr(ENOBUFS);
648 				}
649 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
650 				Free(rtm); rtm = new_rtm;
651 			}
652 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
653 			rtm->rtm_flags = rt->rt_flags;
654 			rtm->rtm_use = 0;
655 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
656 			rtm->rtm_addrs = info.rti_addrs;
657 			break;
658 
659 		case RTM_CHANGE:
660 			/*
661 			 * New gateway could require new ifaddr, ifp;
662 			 * flags may also be different; ifp may be specified
663 			 * by ll sockaddr when protocol address is ambiguous
664 			 */
665 			if (((rt->rt_flags & RTF_GATEWAY) &&
666 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
667 			    info.rti_info[RTAX_IFP] != NULL ||
668 			    (info.rti_info[RTAX_IFA] != NULL &&
669 			     !sa_equal(info.rti_info[RTAX_IFA],
670 				       rt->rt_ifa->ifa_addr))) {
671 				RT_UNLOCK(rt);
672 				if ((error = rt_getifa_fib(&info,
673 				    rt->rt_fibnum)) != 0)
674 					senderr(error);
675 				RT_LOCK(rt);
676 			}
677 			if (info.rti_ifa != NULL &&
678 			    info.rti_ifa != rt->rt_ifa &&
679 			    rt->rt_ifa != NULL &&
680 			    rt->rt_ifa->ifa_rtrequest != NULL) {
681 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
682 				    &info);
683 				IFAFREE(rt->rt_ifa);
684 			}
685 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
686 				if ((error = rt_setgate(rt, rt_key(rt),
687 					info.rti_info[RTAX_GATEWAY])) != 0) {
688 					RT_UNLOCK(rt);
689 					senderr(error);
690 				}
691 				if (!(rt->rt_flags & RTF_LLINFO))
692 					rt->rt_flags |= RTF_GATEWAY;
693 			}
694 			if (info.rti_ifa != NULL &&
695 			    info.rti_ifa != rt->rt_ifa) {
696 				IFAREF(info.rti_ifa);
697 				rt->rt_ifa = info.rti_ifa;
698 				rt->rt_ifp = info.rti_ifp;
699 			}
700 			/* Allow some flags to be toggled on change. */
701 			if (rtm->rtm_fmask & RTF_FMASK)
702 				rt->rt_flags = (rt->rt_flags &
703 				    ~rtm->rtm_fmask) |
704 				    (rtm->rtm_flags & rtm->rtm_fmask);
705 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
706 					&rt->rt_rmx);
707 			rtm->rtm_index = rt->rt_ifp->if_index;
708 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
709 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
710 			if (info.rti_info[RTAX_GENMASK])
711 				rt->rt_genmask = info.rti_info[RTAX_GENMASK];
712 			/* FALLTHROUGH */
713 		case RTM_LOCK:
714 			/* We don't support locks anymore */
715 			break;
716 		}
717 		RT_UNLOCK(rt);
718 		break;
719 
720 	default:
721 		senderr(EOPNOTSUPP);
722 	}
723 
724 flush:
725 	if (rtm) {
726 		if (error)
727 			rtm->rtm_errno = error;
728 		else
729 			rtm->rtm_flags |= RTF_DONE;
730 	}
731 	if (rt)		/* XXX can this be true? */
732 		RTFREE(rt);
733     {
734 	struct rawcb *rp = NULL;
735 	/*
736 	 * Check to see if we don't want our own messages.
737 	 */
738 	if ((so->so_options & SO_USELOOPBACK) == 0) {
739 		if (route_cb.any_count <= 1) {
740 			if (rtm)
741 				Free(rtm);
742 			m_freem(m);
743 			return (error);
744 		}
745 		/* There is another listener, so construct message */
746 		rp = sotorawcb(so);
747 	}
748 	if (rtm) {
749 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
750 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
751 			m_freem(m);
752 			m = NULL;
753 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
754 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
755 		Free(rtm);
756 	}
757 	if (m) {
758 		if (rp) {
759 			/*
760 			 * XXX insure we don't get a copy by
761 			 * invalidating our protocol
762 			 */
763 			unsigned short family = rp->rcb_proto.sp_family;
764 			rp->rcb_proto.sp_family = 0;
765 			rt_dispatch(m, info.rti_info[RTAX_DST]);
766 			rp->rcb_proto.sp_family = family;
767 		} else
768 			rt_dispatch(m, info.rti_info[RTAX_DST]);
769 	}
770     }
771 	return (error);
772 #undef	sa_equal
773 }
774 
775 static void
776 rt_setmetrics(u_long which, const struct rt_metrics *in,
777 	struct rt_metrics_lite *out)
778 {
779 #define metric(f, e) if (which & (f)) out->e = in->e;
780 	/*
781 	 * Only these are stored in the routing entry since introduction
782 	 * of tcp hostcache. The rest is ignored.
783 	 */
784 	metric(RTV_MTU, rmx_mtu);
785 	/* Userland -> kernel timebase conversion. */
786 	if (which & RTV_EXPIRE)
787 		out->rmx_expire = in->rmx_expire ?
788 		    in->rmx_expire - time_second + time_uptime : 0;
789 #undef metric
790 }
791 
792 static void
793 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
794 {
795 #define metric(e) out->e = in->e;
796 	bzero(out, sizeof(*out));
797 	metric(rmx_mtu);
798 	/* Kernel -> userland timebase conversion. */
799 	out->rmx_expire = in->rmx_expire ?
800 	    in->rmx_expire - time_uptime + time_second : 0;
801 #undef metric
802 }
803 
804 /*
805  * Extract the addresses of the passed sockaddrs.
806  * Do a little sanity checking so as to avoid bad memory references.
807  * This data is derived straight from userland.
808  */
809 static int
810 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
811 {
812 	struct sockaddr *sa;
813 	int i;
814 
815 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
816 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
817 			continue;
818 		sa = (struct sockaddr *)cp;
819 		/*
820 		 * It won't fit.
821 		 */
822 		if (cp + sa->sa_len > cplim)
823 			return (EINVAL);
824 		/*
825 		 * there are no more.. quit now
826 		 * If there are more bits, they are in error.
827 		 * I've seen this. route(1) can evidently generate these.
828 		 * This causes kernel to core dump.
829 		 * for compatibility, If we see this, point to a safe address.
830 		 */
831 		if (sa->sa_len == 0) {
832 			rtinfo->rti_info[i] = &sa_zero;
833 			return (0); /* should be EINVAL but for compat */
834 		}
835 		/* accept it */
836 		rtinfo->rti_info[i] = sa;
837 		cp += SA_SIZE(sa);
838 	}
839 	return (0);
840 }
841 
842 static struct mbuf *
843 rt_msg1(int type, struct rt_addrinfo *rtinfo)
844 {
845 	struct rt_msghdr *rtm;
846 	struct mbuf *m;
847 	int i;
848 	struct sockaddr *sa;
849 	int len, dlen;
850 
851 	switch (type) {
852 
853 	case RTM_DELADDR:
854 	case RTM_NEWADDR:
855 		len = sizeof(struct ifa_msghdr);
856 		break;
857 
858 	case RTM_DELMADDR:
859 	case RTM_NEWMADDR:
860 		len = sizeof(struct ifma_msghdr);
861 		break;
862 
863 	case RTM_IFINFO:
864 		len = sizeof(struct if_msghdr);
865 		break;
866 
867 	case RTM_IFANNOUNCE:
868 	case RTM_IEEE80211:
869 		len = sizeof(struct if_announcemsghdr);
870 		break;
871 
872 	default:
873 		len = sizeof(struct rt_msghdr);
874 	}
875 	if (len > MCLBYTES)
876 		panic("rt_msg1");
877 	m = m_gethdr(M_DONTWAIT, MT_DATA);
878 	if (m && len > MHLEN) {
879 		MCLGET(m, M_DONTWAIT);
880 		if ((m->m_flags & M_EXT) == 0) {
881 			m_free(m);
882 			m = NULL;
883 		}
884 	}
885 	if (m == NULL)
886 		return (m);
887 	m->m_pkthdr.len = m->m_len = len;
888 	m->m_pkthdr.rcvif = NULL;
889 	rtm = mtod(m, struct rt_msghdr *);
890 	bzero((caddr_t)rtm, len);
891 	for (i = 0; i < RTAX_MAX; i++) {
892 		if ((sa = rtinfo->rti_info[i]) == NULL)
893 			continue;
894 		rtinfo->rti_addrs |= (1 << i);
895 		dlen = SA_SIZE(sa);
896 		m_copyback(m, len, dlen, (caddr_t)sa);
897 		len += dlen;
898 	}
899 	if (m->m_pkthdr.len != len) {
900 		m_freem(m);
901 		return (NULL);
902 	}
903 	rtm->rtm_msglen = len;
904 	rtm->rtm_version = RTM_VERSION;
905 	rtm->rtm_type = type;
906 	return (m);
907 }
908 
909 static int
910 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
911 {
912 	int i;
913 	int len, dlen, second_time = 0;
914 	caddr_t cp0;
915 
916 	rtinfo->rti_addrs = 0;
917 again:
918 	switch (type) {
919 
920 	case RTM_DELADDR:
921 	case RTM_NEWADDR:
922 		len = sizeof(struct ifa_msghdr);
923 		break;
924 
925 	case RTM_IFINFO:
926 		len = sizeof(struct if_msghdr);
927 		break;
928 
929 	case RTM_NEWMADDR:
930 		len = sizeof(struct ifma_msghdr);
931 		break;
932 
933 	default:
934 		len = sizeof(struct rt_msghdr);
935 	}
936 	cp0 = cp;
937 	if (cp0)
938 		cp += len;
939 	for (i = 0; i < RTAX_MAX; i++) {
940 		struct sockaddr *sa;
941 
942 		if ((sa = rtinfo->rti_info[i]) == NULL)
943 			continue;
944 		rtinfo->rti_addrs |= (1 << i);
945 		dlen = SA_SIZE(sa);
946 		if (cp) {
947 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
948 			cp += dlen;
949 		}
950 		len += dlen;
951 	}
952 	len = ALIGN(len);
953 	if (cp == NULL && w != NULL && !second_time) {
954 		struct walkarg *rw = w;
955 
956 		if (rw->w_req) {
957 			if (rw->w_tmemsize < len) {
958 				if (rw->w_tmem)
959 					free(rw->w_tmem, M_RTABLE);
960 				rw->w_tmem = (caddr_t)
961 					malloc(len, M_RTABLE, M_NOWAIT);
962 				if (rw->w_tmem)
963 					rw->w_tmemsize = len;
964 			}
965 			if (rw->w_tmem) {
966 				cp = rw->w_tmem;
967 				second_time = 1;
968 				goto again;
969 			}
970 		}
971 	}
972 	if (cp) {
973 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
974 
975 		rtm->rtm_version = RTM_VERSION;
976 		rtm->rtm_type = type;
977 		rtm->rtm_msglen = len;
978 	}
979 	return (len);
980 }
981 
982 /*
983  * This routine is called to generate a message from the routing
984  * socket indicating that a redirect has occured, a routing lookup
985  * has failed, or that a protocol has detected timeouts to a particular
986  * destination.
987  */
988 void
989 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
990 {
991 	struct rt_msghdr *rtm;
992 	struct mbuf *m;
993 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
994 
995 	if (route_cb.any_count == 0)
996 		return;
997 	m = rt_msg1(type, rtinfo);
998 	if (m == NULL)
999 		return;
1000 	rtm = mtod(m, struct rt_msghdr *);
1001 	rtm->rtm_flags = RTF_DONE | flags;
1002 	rtm->rtm_errno = error;
1003 	rtm->rtm_addrs = rtinfo->rti_addrs;
1004 	rt_dispatch(m, sa);
1005 }
1006 
1007 /*
1008  * This routine is called to generate a message from the routing
1009  * socket indicating that the status of a network interface has changed.
1010  */
1011 void
1012 rt_ifmsg(struct ifnet *ifp)
1013 {
1014 	struct if_msghdr *ifm;
1015 	struct mbuf *m;
1016 	struct rt_addrinfo info;
1017 
1018 	if (route_cb.any_count == 0)
1019 		return;
1020 	bzero((caddr_t)&info, sizeof(info));
1021 	m = rt_msg1(RTM_IFINFO, &info);
1022 	if (m == NULL)
1023 		return;
1024 	ifm = mtod(m, struct if_msghdr *);
1025 	ifm->ifm_index = ifp->if_index;
1026 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1027 	ifm->ifm_data = ifp->if_data;
1028 	ifm->ifm_addrs = 0;
1029 	rt_dispatch(m, NULL);
1030 }
1031 
1032 /*
1033  * This is called to generate messages from the routing socket
1034  * indicating a network interface has had addresses associated with it.
1035  * if we ever reverse the logic and replace messages TO the routing
1036  * socket indicate a request to configure interfaces, then it will
1037  * be unnecessary as the routing socket will automatically generate
1038  * copies of it.
1039  */
1040 void
1041 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1042 {
1043 	struct rt_addrinfo info;
1044 	struct sockaddr *sa = NULL;
1045 	int pass;
1046 	struct mbuf *m = NULL;
1047 	struct ifnet *ifp = ifa->ifa_ifp;
1048 
1049 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1050 		("unexpected cmd %u", cmd));
1051 #ifdef SCTP
1052 	/*
1053 	 * notify the SCTP stack
1054 	 * this will only get called when an address is added/deleted
1055 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1056 	 */
1057 	sctp_addr_change(ifa, cmd);
1058 #endif /* SCTP */
1059 	if (route_cb.any_count == 0)
1060 		return;
1061 	for (pass = 1; pass < 3; pass++) {
1062 		bzero((caddr_t)&info, sizeof(info));
1063 		if ((cmd == RTM_ADD && pass == 1) ||
1064 		    (cmd == RTM_DELETE && pass == 2)) {
1065 			struct ifa_msghdr *ifam;
1066 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1067 
1068 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1069 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1070 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1071 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1072 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1073 				continue;
1074 			ifam = mtod(m, struct ifa_msghdr *);
1075 			ifam->ifam_index = ifp->if_index;
1076 			ifam->ifam_metric = ifa->ifa_metric;
1077 			ifam->ifam_flags = ifa->ifa_flags;
1078 			ifam->ifam_addrs = info.rti_addrs;
1079 		}
1080 		if ((cmd == RTM_ADD && pass == 2) ||
1081 		    (cmd == RTM_DELETE && pass == 1)) {
1082 			struct rt_msghdr *rtm;
1083 
1084 			if (rt == NULL)
1085 				continue;
1086 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1087 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1088 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1089 			if ((m = rt_msg1(cmd, &info)) == NULL)
1090 				continue;
1091 			rtm = mtod(m, struct rt_msghdr *);
1092 			rtm->rtm_index = ifp->if_index;
1093 			rtm->rtm_flags |= rt->rt_flags;
1094 			rtm->rtm_errno = error;
1095 			rtm->rtm_addrs = info.rti_addrs;
1096 		}
1097 		rt_dispatch(m, sa);
1098 	}
1099 }
1100 
1101 /*
1102  * This is the analogue to the rt_newaddrmsg which performs the same
1103  * function but for multicast group memberhips.  This is easier since
1104  * there is no route state to worry about.
1105  */
1106 void
1107 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1108 {
1109 	struct rt_addrinfo info;
1110 	struct mbuf *m = NULL;
1111 	struct ifnet *ifp = ifma->ifma_ifp;
1112 	struct ifma_msghdr *ifmam;
1113 
1114 	if (route_cb.any_count == 0)
1115 		return;
1116 
1117 	bzero((caddr_t)&info, sizeof(info));
1118 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1119 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1120 	/*
1121 	 * If a link-layer address is present, present it as a ``gateway''
1122 	 * (similarly to how ARP entries, e.g., are presented).
1123 	 */
1124 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1125 	m = rt_msg1(cmd, &info);
1126 	if (m == NULL)
1127 		return;
1128 	ifmam = mtod(m, struct ifma_msghdr *);
1129 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1130 	    __func__));
1131 	ifmam->ifmam_index = ifp->if_index;
1132 	ifmam->ifmam_addrs = info.rti_addrs;
1133 	rt_dispatch(m, ifma->ifma_addr);
1134 }
1135 
1136 static struct mbuf *
1137 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1138 	struct rt_addrinfo *info)
1139 {
1140 	struct if_announcemsghdr *ifan;
1141 	struct mbuf *m;
1142 
1143 	if (route_cb.any_count == 0)
1144 		return NULL;
1145 	bzero((caddr_t)info, sizeof(*info));
1146 	m = rt_msg1(type, info);
1147 	if (m != NULL) {
1148 		ifan = mtod(m, struct if_announcemsghdr *);
1149 		ifan->ifan_index = ifp->if_index;
1150 		strlcpy(ifan->ifan_name, ifp->if_xname,
1151 			sizeof(ifan->ifan_name));
1152 		ifan->ifan_what = what;
1153 	}
1154 	return m;
1155 }
1156 
1157 /*
1158  * This is called to generate routing socket messages indicating
1159  * IEEE80211 wireless events.
1160  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1161  */
1162 void
1163 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1164 {
1165 	struct mbuf *m;
1166 	struct rt_addrinfo info;
1167 
1168 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1169 	if (m != NULL) {
1170 		/*
1171 		 * Append the ieee80211 data.  Try to stick it in the
1172 		 * mbuf containing the ifannounce msg; otherwise allocate
1173 		 * a new mbuf and append.
1174 		 *
1175 		 * NB: we assume m is a single mbuf.
1176 		 */
1177 		if (data_len > M_TRAILINGSPACE(m)) {
1178 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1179 			if (n == NULL) {
1180 				m_freem(m);
1181 				return;
1182 			}
1183 			bcopy(data, mtod(n, void *), data_len);
1184 			n->m_len = data_len;
1185 			m->m_next = n;
1186 		} else if (data_len > 0) {
1187 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1188 			m->m_len += data_len;
1189 		}
1190 		if (m->m_flags & M_PKTHDR)
1191 			m->m_pkthdr.len += data_len;
1192 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1193 		rt_dispatch(m, NULL);
1194 	}
1195 }
1196 
1197 /*
1198  * This is called to generate routing socket messages indicating
1199  * network interface arrival and departure.
1200  */
1201 void
1202 rt_ifannouncemsg(struct ifnet *ifp, int what)
1203 {
1204 	struct mbuf *m;
1205 	struct rt_addrinfo info;
1206 
1207 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1208 	if (m != NULL)
1209 		rt_dispatch(m, NULL);
1210 }
1211 
1212 static void
1213 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1214 {
1215 	INIT_VNET_NET(curvnet);
1216 	struct m_tag *tag;
1217 
1218 	/*
1219 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1220 	 * use when injecting the mbuf into the routing socket buffer from
1221 	 * the netisr.
1222 	 */
1223 	if (sa != NULL) {
1224 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1225 		    M_NOWAIT);
1226 		if (tag == NULL) {
1227 			m_freem(m);
1228 			return;
1229 		}
1230 		*(unsigned short *)(tag + 1) = sa->sa_family;
1231 		m_tag_prepend(m, tag);
1232 	}
1233 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1234 }
1235 
1236 /*
1237  * This is used in dumping the kernel table via sysctl().
1238  */
1239 static int
1240 sysctl_dumpentry(struct radix_node *rn, void *vw)
1241 {
1242 	struct walkarg *w = vw;
1243 	struct rtentry *rt = (struct rtentry *)rn;
1244 	int error = 0, size;
1245 	struct rt_addrinfo info;
1246 
1247 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1248 		return 0;
1249 	bzero((caddr_t)&info, sizeof(info));
1250 	info.rti_info[RTAX_DST] = rt_key(rt);
1251 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1252 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1253 	info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1254 	if (rt->rt_ifp) {
1255 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1256 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1257 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1258 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1259 	}
1260 	size = rt_msg2(RTM_GET, &info, NULL, w);
1261 	if (w->w_req && w->w_tmem) {
1262 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1263 
1264 		rtm->rtm_flags = rt->rt_flags;
1265 		rtm->rtm_use = rt->rt_rmx.rmx_pksent;
1266 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1267 		rtm->rtm_index = rt->rt_ifp->if_index;
1268 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1269 		rtm->rtm_addrs = info.rti_addrs;
1270 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1271 		return (error);
1272 	}
1273 	return (error);
1274 }
1275 
1276 static int
1277 sysctl_iflist(int af, struct walkarg *w)
1278 {
1279 	INIT_VNET_NET(curvnet);
1280 	struct ifnet *ifp;
1281 	struct ifaddr *ifa;
1282 	struct rt_addrinfo info;
1283 	int len, error = 0;
1284 
1285 	bzero((caddr_t)&info, sizeof(info));
1286 	IFNET_RLOCK();
1287 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1288 		if (w->w_arg && w->w_arg != ifp->if_index)
1289 			continue;
1290 		ifa = ifp->if_addr;
1291 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1292 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1293 		info.rti_info[RTAX_IFP] = NULL;
1294 		if (w->w_req && w->w_tmem) {
1295 			struct if_msghdr *ifm;
1296 
1297 			ifm = (struct if_msghdr *)w->w_tmem;
1298 			ifm->ifm_index = ifp->if_index;
1299 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1300 			ifm->ifm_data = ifp->if_data;
1301 			ifm->ifm_addrs = info.rti_addrs;
1302 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
1303 			if (error)
1304 				goto done;
1305 		}
1306 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1307 			if (af && af != ifa->ifa_addr->sa_family)
1308 				continue;
1309 			if (jailed(curthread->td_ucred) &&
1310 			    !prison_if(curthread->td_ucred, ifa->ifa_addr))
1311 				continue;
1312 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1313 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1314 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1315 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1316 			if (w->w_req && w->w_tmem) {
1317 				struct ifa_msghdr *ifam;
1318 
1319 				ifam = (struct ifa_msghdr *)w->w_tmem;
1320 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1321 				ifam->ifam_flags = ifa->ifa_flags;
1322 				ifam->ifam_metric = ifa->ifa_metric;
1323 				ifam->ifam_addrs = info.rti_addrs;
1324 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1325 				if (error)
1326 					goto done;
1327 			}
1328 		}
1329 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1330 			info.rti_info[RTAX_BRD] = NULL;
1331 	}
1332 done:
1333 	IFNET_RUNLOCK();
1334 	return (error);
1335 }
1336 
1337 int
1338 sysctl_ifmalist(int af, struct walkarg *w)
1339 {
1340 	INIT_VNET_NET(curvnet);
1341 	struct ifnet *ifp;
1342 	struct ifmultiaddr *ifma;
1343 	struct	rt_addrinfo info;
1344 	int	len, error = 0;
1345 	struct ifaddr *ifa;
1346 
1347 	bzero((caddr_t)&info, sizeof(info));
1348 	IFNET_RLOCK();
1349 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1350 		if (w->w_arg && w->w_arg != ifp->if_index)
1351 			continue;
1352 		ifa = ifp->if_addr;
1353 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1354 		IF_ADDR_LOCK(ifp);
1355 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1356 			if (af && af != ifma->ifma_addr->sa_family)
1357 				continue;
1358 			if (jailed(curproc->p_ucred) &&
1359 			    !prison_if(curproc->p_ucred, ifma->ifma_addr))
1360 				continue;
1361 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1362 			info.rti_info[RTAX_GATEWAY] =
1363 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1364 			    ifma->ifma_lladdr : NULL;
1365 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1366 			if (w->w_req && w->w_tmem) {
1367 				struct ifma_msghdr *ifmam;
1368 
1369 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1370 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1371 				ifmam->ifmam_flags = 0;
1372 				ifmam->ifmam_addrs = info.rti_addrs;
1373 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1374 				if (error) {
1375 					IF_ADDR_UNLOCK(ifp);
1376 					goto done;
1377 				}
1378 			}
1379 		}
1380 		IF_ADDR_UNLOCK(ifp);
1381 	}
1382 done:
1383 	IFNET_RUNLOCK();
1384 	return (error);
1385 }
1386 
1387 static int
1388 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1389 {
1390 	INIT_VNET_NET(curvnet);
1391 	int	*name = (int *)arg1;
1392 	u_int	namelen = arg2;
1393 	struct radix_node_head *rnh;
1394 	int	i, lim, error = EINVAL;
1395 	u_char	af;
1396 	struct	walkarg w;
1397 
1398 	name ++;
1399 	namelen--;
1400 	if (req->newptr)
1401 		return (EPERM);
1402 	if (namelen != 3)
1403 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1404 	af = name[0];
1405 	if (af > AF_MAX)
1406 		return (EINVAL);
1407 	bzero(&w, sizeof(w));
1408 	w.w_op = name[1];
1409 	w.w_arg = name[2];
1410 	w.w_req = req;
1411 
1412 	error = sysctl_wire_old_buffer(req, 0);
1413 	if (error)
1414 		return (error);
1415 	switch (w.w_op) {
1416 
1417 	case NET_RT_DUMP:
1418 	case NET_RT_FLAGS:
1419 		if (af == 0) {			/* dump all tables */
1420 			i = 1;
1421 			lim = AF_MAX;
1422 		} else				/* dump only one table */
1423 			i = lim = af;
1424 		for (error = 0; error == 0 && i <= lim; i++)
1425 			if ((rnh = V_rt_tables[curthread->td_proc->p_fibnum][i]) != NULL) {
1426 				RADIX_NODE_HEAD_LOCK(rnh);
1427 			    	error = rnh->rnh_walktree(rnh,
1428 				    sysctl_dumpentry, &w);
1429 				RADIX_NODE_HEAD_UNLOCK(rnh);
1430 			} else if (af != 0)
1431 				error = EAFNOSUPPORT;
1432 		break;
1433 
1434 	case NET_RT_IFLIST:
1435 		error = sysctl_iflist(af, &w);
1436 		break;
1437 
1438 	case NET_RT_IFMALIST:
1439 		error = sysctl_ifmalist(af, &w);
1440 		break;
1441 	}
1442 	if (w.w_tmem)
1443 		free(w.w_tmem, M_RTABLE);
1444 	return (error);
1445 }
1446 
1447 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1448 
1449 /*
1450  * Definitions of protocols supported in the ROUTE domain.
1451  */
1452 
1453 static struct domain routedomain;		/* or at least forward */
1454 
1455 static struct protosw routesw[] = {
1456 {
1457 	.pr_type =		SOCK_RAW,
1458 	.pr_domain =		&routedomain,
1459 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1460 	.pr_output =		route_output,
1461 	.pr_ctlinput =		raw_ctlinput,
1462 	.pr_init =		raw_init,
1463 	.pr_usrreqs =		&route_usrreqs
1464 }
1465 };
1466 
1467 static struct domain routedomain = {
1468 	.dom_family =		PF_ROUTE,
1469 	.dom_name =		 "route",
1470 	.dom_protosw =		routesw,
1471 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1472 };
1473 
1474 DOMAIN_SET(route);
1475