xref: /freebsd/sys/net/rtsock.c (revision 82431678fce5c893ef9c7418ad6d998ad4187de6)
1 /*-
2  * Copyright (c) 1988, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
30  * $FreeBSD$
31  */
32 #include "opt_sctp.h"
33 #include "opt_mpath.h"
34 #include "opt_route.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/domain.h>
40 #include <sys/jail.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/rwlock.h>
49 #include <sys/signalvar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/systm.h>
54 #include <sys/vimage.h>
55 
56 #include <net/if.h>
57 #include <net/if_dl.h>
58 #include <net/if_llatbl.h>
59 #include <net/netisr.h>
60 #include <net/raw_cb.h>
61 #include <net/route.h>
62 #include <net/vnet.h>
63 
64 #include <netinet/in.h>
65 #ifdef INET6
66 #include <netinet6/scope6_var.h>
67 #endif
68 
69 #ifdef SCTP
70 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
71 #endif /* SCTP */
72 
73 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
74 
75 /* NB: these are not modified */
76 static struct	sockaddr route_src = { 2, PF_ROUTE, };
77 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
78 
79 static struct {
80 	int	ip_count;	/* attached w/ AF_INET */
81 	int	ip6_count;	/* attached w/ AF_INET6 */
82 	int	ipx_count;	/* attached w/ AF_IPX */
83 	int	any_count;	/* total attached */
84 } route_cb;
85 
86 struct mtx rtsock_mtx;
87 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
88 
89 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
90 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
91 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
92 
93 static struct	ifqueue rtsintrq;
94 
95 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
96 SYSCTL_INT(_net_route, OID_AUTO, netisr_maxqlen, CTLFLAG_RW,
97     &rtsintrq.ifq_maxlen, 0, "maximum routing socket dispatch queue length");
98 
99 struct walkarg {
100 	int	w_tmemsize;
101 	int	w_op, w_arg;
102 	caddr_t	w_tmem;
103 	struct sysctl_req *w_req;
104 };
105 
106 static void	rts_input(struct mbuf *m);
107 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
108 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
109 			caddr_t cp, struct walkarg *w);
110 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
111 			struct rt_addrinfo *rtinfo);
112 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
113 static int	sysctl_iflist(int af, struct walkarg *w);
114 static int	sysctl_ifmalist(int af, struct walkarg *w);
115 static int	route_output(struct mbuf *m, struct socket *so);
116 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
117 			struct rt_metrics_lite *out);
118 static void	rt_getmetrics(const struct rt_metrics_lite *in,
119 			struct rt_metrics *out);
120 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
121 
122 static void
123 rts_init(void)
124 {
125 	int tmp;
126 
127 	rtsintrq.ifq_maxlen = 256;
128 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
129 		rtsintrq.ifq_maxlen = tmp;
130 	mtx_init(&rtsintrq.ifq_mtx, "rts_inq", NULL, MTX_DEF);
131 	netisr_register(NETISR_ROUTE, rts_input, &rtsintrq, 0);
132 }
133 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
134 
135 static void
136 rts_input(struct mbuf *m)
137 {
138 	struct sockproto route_proto;
139 	unsigned short *family;
140 	struct m_tag *tag;
141 
142 	route_proto.sp_family = PF_ROUTE;
143 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
144 	if (tag != NULL) {
145 		family = (unsigned short *)(tag + 1);
146 		route_proto.sp_protocol = *family;
147 		m_tag_delete(m, tag);
148 	} else
149 		route_proto.sp_protocol = 0;
150 
151 	raw_input(m, &route_proto, &route_src);
152 }
153 
154 /*
155  * It really doesn't make any sense at all for this code to share much
156  * with raw_usrreq.c, since its functionality is so restricted.  XXX
157  */
158 static void
159 rts_abort(struct socket *so)
160 {
161 
162 	raw_usrreqs.pru_abort(so);
163 }
164 
165 static void
166 rts_close(struct socket *so)
167 {
168 
169 	raw_usrreqs.pru_close(so);
170 }
171 
172 /* pru_accept is EOPNOTSUPP */
173 
174 static int
175 rts_attach(struct socket *so, int proto, struct thread *td)
176 {
177 	struct rawcb *rp;
178 	int s, error;
179 
180 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
181 
182 	/* XXX */
183 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
184 	if (rp == NULL)
185 		return ENOBUFS;
186 
187 	/*
188 	 * The splnet() is necessary to block protocols from sending
189 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
190 	 * this PCB is extant but incompletely initialized.
191 	 * Probably we should try to do more of this work beforehand and
192 	 * eliminate the spl.
193 	 */
194 	s = splnet();
195 	so->so_pcb = (caddr_t)rp;
196 	so->so_fibnum = td->td_proc->p_fibnum;
197 	error = raw_attach(so, proto);
198 	rp = sotorawcb(so);
199 	if (error) {
200 		splx(s);
201 		so->so_pcb = NULL;
202 		free(rp, M_PCB);
203 		return error;
204 	}
205 	RTSOCK_LOCK();
206 	switch(rp->rcb_proto.sp_protocol) {
207 	case AF_INET:
208 		route_cb.ip_count++;
209 		break;
210 	case AF_INET6:
211 		route_cb.ip6_count++;
212 		break;
213 	case AF_IPX:
214 		route_cb.ipx_count++;
215 		break;
216 	}
217 	route_cb.any_count++;
218 	RTSOCK_UNLOCK();
219 	soisconnected(so);
220 	so->so_options |= SO_USELOOPBACK;
221 	splx(s);
222 	return 0;
223 }
224 
225 static int
226 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
227 {
228 
229 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
230 }
231 
232 static int
233 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
234 {
235 
236 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
237 }
238 
239 /* pru_connect2 is EOPNOTSUPP */
240 /* pru_control is EOPNOTSUPP */
241 
242 static void
243 rts_detach(struct socket *so)
244 {
245 	struct rawcb *rp = sotorawcb(so);
246 
247 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
248 
249 	RTSOCK_LOCK();
250 	switch(rp->rcb_proto.sp_protocol) {
251 	case AF_INET:
252 		route_cb.ip_count--;
253 		break;
254 	case AF_INET6:
255 		route_cb.ip6_count--;
256 		break;
257 	case AF_IPX:
258 		route_cb.ipx_count--;
259 		break;
260 	}
261 	route_cb.any_count--;
262 	RTSOCK_UNLOCK();
263 	raw_usrreqs.pru_detach(so);
264 }
265 
266 static int
267 rts_disconnect(struct socket *so)
268 {
269 
270 	return (raw_usrreqs.pru_disconnect(so));
271 }
272 
273 /* pru_listen is EOPNOTSUPP */
274 
275 static int
276 rts_peeraddr(struct socket *so, struct sockaddr **nam)
277 {
278 
279 	return (raw_usrreqs.pru_peeraddr(so, nam));
280 }
281 
282 /* pru_rcvd is EOPNOTSUPP */
283 /* pru_rcvoob is EOPNOTSUPP */
284 
285 static int
286 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
287 	 struct mbuf *control, struct thread *td)
288 {
289 
290 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
291 }
292 
293 /* pru_sense is null */
294 
295 static int
296 rts_shutdown(struct socket *so)
297 {
298 
299 	return (raw_usrreqs.pru_shutdown(so));
300 }
301 
302 static int
303 rts_sockaddr(struct socket *so, struct sockaddr **nam)
304 {
305 
306 	return (raw_usrreqs.pru_sockaddr(so, nam));
307 }
308 
309 static struct pr_usrreqs route_usrreqs = {
310 	.pru_abort =		rts_abort,
311 	.pru_attach =		rts_attach,
312 	.pru_bind =		rts_bind,
313 	.pru_connect =		rts_connect,
314 	.pru_detach =		rts_detach,
315 	.pru_disconnect =	rts_disconnect,
316 	.pru_peeraddr =		rts_peeraddr,
317 	.pru_send =		rts_send,
318 	.pru_shutdown =		rts_shutdown,
319 	.pru_sockaddr =		rts_sockaddr,
320 	.pru_close =		rts_close,
321 };
322 
323 #ifndef _SOCKADDR_UNION_DEFINED
324 #define	_SOCKADDR_UNION_DEFINED
325 /*
326  * The union of all possible address formats we handle.
327  */
328 union sockaddr_union {
329 	struct sockaddr		sa;
330 	struct sockaddr_in	sin;
331 	struct sockaddr_in6	sin6;
332 };
333 #endif /* _SOCKADDR_UNION_DEFINED */
334 
335 static int
336 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
337     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
338 {
339 
340 	/* First, see if the returned address is part of the jail. */
341 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
342 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
343 		return (0);
344 	}
345 
346 	switch (info->rti_info[RTAX_DST]->sa_family) {
347 #ifdef INET
348 	case AF_INET:
349 	{
350 		struct in_addr ia;
351 		struct ifaddr *ifa;
352 		int found;
353 
354 		found = 0;
355 		/*
356 		 * Try to find an address on the given outgoing interface
357 		 * that belongs to the jail.
358 		 */
359 		IF_ADDR_LOCK(ifp);
360 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
361 			struct sockaddr *sa;
362 			sa = ifa->ifa_addr;
363 			if (sa->sa_family != AF_INET)
364 				continue;
365 			ia = ((struct sockaddr_in *)sa)->sin_addr;
366 			if (prison_check_ip4(cred, &ia) == 0) {
367 				found = 1;
368 				break;
369 			}
370 		}
371 		IF_ADDR_UNLOCK(ifp);
372 		if (!found) {
373 			/*
374 			 * As a last resort return the 'default' jail address.
375 			 */
376 			if (prison_get_ip4(cred, &ia) != 0)
377 				return (ESRCH);
378 		}
379 		bzero(&saun->sin, sizeof(struct sockaddr_in));
380 		saun->sin.sin_len = sizeof(struct sockaddr_in);
381 		saun->sin.sin_family = AF_INET;
382 		saun->sin.sin_addr.s_addr = ia.s_addr;
383 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
384 		break;
385 	}
386 #endif
387 #ifdef INET6
388 	case AF_INET6:
389 	{
390 		struct in6_addr ia6;
391 		struct ifaddr *ifa;
392 		int found;
393 
394 		found = 0;
395 		/*
396 		 * Try to find an address on the given outgoing interface
397 		 * that belongs to the jail.
398 		 */
399 		IF_ADDR_LOCK(ifp);
400 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
401 			struct sockaddr *sa;
402 			sa = ifa->ifa_addr;
403 			if (sa->sa_family != AF_INET6)
404 				continue;
405 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
406 			    &ia6, sizeof(struct in6_addr));
407 			if (prison_check_ip6(cred, &ia6) == 0) {
408 				found = 1;
409 				break;
410 			}
411 		}
412 		IF_ADDR_UNLOCK(ifp);
413 		if (!found) {
414 			/*
415 			 * As a last resort return the 'default' jail address.
416 			 */
417 			if (prison_get_ip6(cred, &ia6) != 0)
418 				return (ESRCH);
419 		}
420 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
421 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
422 		saun->sin6.sin6_family = AF_INET6;
423 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
424 		if (sa6_recoverscope(&saun->sin6) != 0)
425 			return (ESRCH);
426 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
427 		break;
428 	}
429 #endif
430 	default:
431 		return (ESRCH);
432 	}
433 	return (0);
434 }
435 
436 /*ARGSUSED*/
437 static int
438 route_output(struct mbuf *m, struct socket *so)
439 {
440 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
441 	INIT_VNET_NET(so->so_vnet);
442 	struct rt_msghdr *rtm = NULL;
443 	struct rtentry *rt = NULL;
444 	struct radix_node_head *rnh;
445 	struct rt_addrinfo info;
446 	int len, error = 0;
447 	struct ifnet *ifp = NULL;
448 	union sockaddr_union saun;
449 
450 #define senderr(e) { error = e; goto flush;}
451 	if (m == NULL || ((m->m_len < sizeof(long)) &&
452 		       (m = m_pullup(m, sizeof(long))) == NULL))
453 		return (ENOBUFS);
454 	if ((m->m_flags & M_PKTHDR) == 0)
455 		panic("route_output");
456 	len = m->m_pkthdr.len;
457 	if (len < sizeof(*rtm) ||
458 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
459 		info.rti_info[RTAX_DST] = NULL;
460 		senderr(EINVAL);
461 	}
462 	R_Malloc(rtm, struct rt_msghdr *, len);
463 	if (rtm == NULL) {
464 		info.rti_info[RTAX_DST] = NULL;
465 		senderr(ENOBUFS);
466 	}
467 	m_copydata(m, 0, len, (caddr_t)rtm);
468 	if (rtm->rtm_version != RTM_VERSION) {
469 		info.rti_info[RTAX_DST] = NULL;
470 		senderr(EPROTONOSUPPORT);
471 	}
472 	rtm->rtm_pid = curproc->p_pid;
473 	bzero(&info, sizeof(info));
474 	info.rti_addrs = rtm->rtm_addrs;
475 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
476 		info.rti_info[RTAX_DST] = NULL;
477 		senderr(EINVAL);
478 	}
479 	info.rti_flags = rtm->rtm_flags;
480 	if (info.rti_info[RTAX_DST] == NULL ||
481 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
482 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
483 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
484 		senderr(EINVAL);
485 	/*
486 	 * Verify that the caller has the appropriate privilege; RTM_GET
487 	 * is the only operation the non-superuser is allowed.
488 	 */
489 	if (rtm->rtm_type != RTM_GET) {
490 		error = priv_check(curthread, PRIV_NET_ROUTE);
491 		if (error)
492 			senderr(error);
493 	}
494 
495 	switch (rtm->rtm_type) {
496 		struct rtentry *saved_nrt;
497 
498 	case RTM_ADD:
499 		if (info.rti_info[RTAX_GATEWAY] == NULL)
500 			senderr(EINVAL);
501 		saved_nrt = NULL;
502 
503 		/* support for new ARP code */
504 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
505 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
506 			error = lla_rt_output(rtm, &info);
507 			break;
508 		}
509 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
510 		    so->so_fibnum);
511 		if (error == 0 && saved_nrt) {
512 			RT_LOCK(saved_nrt);
513 			rt_setmetrics(rtm->rtm_inits,
514 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
515 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
516 			RT_REMREF(saved_nrt);
517 			RT_UNLOCK(saved_nrt);
518 		}
519 		break;
520 
521 	case RTM_DELETE:
522 		saved_nrt = NULL;
523 		/* support for new ARP code */
524 		if (info.rti_info[RTAX_GATEWAY] &&
525 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
526 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
527 			error = lla_rt_output(rtm, &info);
528 			break;
529 		}
530 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
531 		    so->so_fibnum);
532 		if (error == 0) {
533 			RT_LOCK(saved_nrt);
534 			rt = saved_nrt;
535 			goto report;
536 		}
537 		break;
538 
539 	case RTM_GET:
540 	case RTM_CHANGE:
541 	case RTM_LOCK:
542 		rnh = V_rt_tables[so->so_fibnum][info.rti_info[RTAX_DST]->sa_family];
543 		if (rnh == NULL)
544 			senderr(EAFNOSUPPORT);
545 		RADIX_NODE_HEAD_RLOCK(rnh);
546 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
547 			info.rti_info[RTAX_NETMASK], rnh);
548 		if (rt == NULL) {	/* XXX looks bogus */
549 			RADIX_NODE_HEAD_RUNLOCK(rnh);
550 			senderr(ESRCH);
551 		}
552 #ifdef RADIX_MPATH
553 		/*
554 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
555 		 * we require users to specify a matching RTAX_GATEWAY.
556 		 *
557 		 * for RTM_GET, gate is optional even with multipath.
558 		 * if gate == NULL the first match is returned.
559 		 * (no need to call rt_mpath_matchgate if gate == NULL)
560 		 */
561 		if (rn_mpath_capable(rnh) &&
562 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
563 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
564 			if (!rt) {
565 				RADIX_NODE_HEAD_RUNLOCK(rnh);
566 				senderr(ESRCH);
567 			}
568 		}
569 #endif
570 		RT_LOCK(rt);
571 		RT_ADDREF(rt);
572 		RADIX_NODE_HEAD_RUNLOCK(rnh);
573 
574 		/*
575 		 * Fix for PR: 82974
576 		 *
577 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
578 		 * returns a perfect match in case a netmask is
579 		 * specified.  For host routes only a longest prefix
580 		 * match is returned so it is necessary to compare the
581 		 * existence of the netmask.  If both have a netmask
582 		 * rnh_lookup() did a perfect match and if none of them
583 		 * have a netmask both are host routes which is also a
584 		 * perfect match.
585 		 */
586 
587 		if (rtm->rtm_type != RTM_GET &&
588 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
589 			RT_UNLOCK(rt);
590 			senderr(ESRCH);
591 		}
592 
593 		switch(rtm->rtm_type) {
594 
595 		case RTM_GET:
596 		report:
597 			RT_LOCK_ASSERT(rt);
598 			if ((rt->rt_flags & RTF_HOST) == 0
599 			    ? jailed(curthread->td_ucred)
600 			    : prison_if(curthread->td_ucred,
601 			    rt_key(rt)) != 0) {
602 				RT_UNLOCK(rt);
603 				senderr(ESRCH);
604 			}
605 			info.rti_info[RTAX_DST] = rt_key(rt);
606 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
607 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
608 			info.rti_info[RTAX_GENMASK] = 0;
609 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
610 				ifp = rt->rt_ifp;
611 				if (ifp) {
612 					info.rti_info[RTAX_IFP] =
613 					    ifp->if_addr->ifa_addr;
614 					error = rtm_get_jailed(&info, ifp, rt,
615 					    &saun, curthread->td_ucred);
616 					if (error != 0) {
617 						RT_UNLOCK(rt);
618 						senderr(error);
619 					}
620 					if (ifp->if_flags & IFF_POINTOPOINT)
621 						info.rti_info[RTAX_BRD] =
622 						    rt->rt_ifa->ifa_dstaddr;
623 					rtm->rtm_index = ifp->if_index;
624 				} else {
625 					info.rti_info[RTAX_IFP] = NULL;
626 					info.rti_info[RTAX_IFA] = NULL;
627 				}
628 			} else if ((ifp = rt->rt_ifp) != NULL) {
629 				rtm->rtm_index = ifp->if_index;
630 			}
631 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
632 			if (len > rtm->rtm_msglen) {
633 				struct rt_msghdr *new_rtm;
634 				R_Malloc(new_rtm, struct rt_msghdr *, len);
635 				if (new_rtm == NULL) {
636 					RT_UNLOCK(rt);
637 					senderr(ENOBUFS);
638 				}
639 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
640 				Free(rtm); rtm = new_rtm;
641 			}
642 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
643 			rtm->rtm_flags = rt->rt_flags;
644 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
645 			rtm->rtm_addrs = info.rti_addrs;
646 			break;
647 
648 		case RTM_CHANGE:
649 			/*
650 			 * New gateway could require new ifaddr, ifp;
651 			 * flags may also be different; ifp may be specified
652 			 * by ll sockaddr when protocol address is ambiguous
653 			 */
654 			if (((rt->rt_flags & RTF_GATEWAY) &&
655 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
656 			    info.rti_info[RTAX_IFP] != NULL ||
657 			    (info.rti_info[RTAX_IFA] != NULL &&
658 			     !sa_equal(info.rti_info[RTAX_IFA],
659 				       rt->rt_ifa->ifa_addr))) {
660 				RT_UNLOCK(rt);
661 				RADIX_NODE_HEAD_LOCK(rnh);
662 				error = rt_getifa_fib(&info, rt->rt_fibnum);
663 				RADIX_NODE_HEAD_UNLOCK(rnh);
664 				if (error != 0)
665 					senderr(error);
666 				RT_LOCK(rt);
667 			}
668 			if (info.rti_ifa != NULL &&
669 			    info.rti_ifa != rt->rt_ifa &&
670 			    rt->rt_ifa != NULL &&
671 			    rt->rt_ifa->ifa_rtrequest != NULL) {
672 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
673 				    &info);
674 				IFAFREE(rt->rt_ifa);
675 			}
676 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
677 				RT_UNLOCK(rt);
678 				RADIX_NODE_HEAD_LOCK(rnh);
679 				RT_LOCK(rt);
680 
681 				error = rt_setgate(rt, rt_key(rt),
682 				    info.rti_info[RTAX_GATEWAY]);
683 				RADIX_NODE_HEAD_UNLOCK(rnh);
684 				if (error != 0) {
685 					RT_UNLOCK(rt);
686 					senderr(error);
687 				}
688 				rt->rt_flags |= RTF_GATEWAY;
689 			}
690 			if (info.rti_ifa != NULL &&
691 			    info.rti_ifa != rt->rt_ifa) {
692 				IFAREF(info.rti_ifa);
693 				rt->rt_ifa = info.rti_ifa;
694 				rt->rt_ifp = info.rti_ifp;
695 			}
696 			/* Allow some flags to be toggled on change. */
697 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
698 				    (rtm->rtm_flags & RTF_FMASK);
699 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
700 					&rt->rt_rmx);
701 			rtm->rtm_index = rt->rt_ifp->if_index;
702 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
703 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
704 			/* FALLTHROUGH */
705 		case RTM_LOCK:
706 			/* We don't support locks anymore */
707 			break;
708 		}
709 		RT_UNLOCK(rt);
710 		break;
711 
712 	default:
713 		senderr(EOPNOTSUPP);
714 	}
715 
716 flush:
717 	if (rtm) {
718 		if (error)
719 			rtm->rtm_errno = error;
720 		else
721 			rtm->rtm_flags |= RTF_DONE;
722 	}
723 	if (rt)		/* XXX can this be true? */
724 		RTFREE(rt);
725     {
726 	struct rawcb *rp = NULL;
727 	/*
728 	 * Check to see if we don't want our own messages.
729 	 */
730 	if ((so->so_options & SO_USELOOPBACK) == 0) {
731 		if (route_cb.any_count <= 1) {
732 			if (rtm)
733 				Free(rtm);
734 			m_freem(m);
735 			return (error);
736 		}
737 		/* There is another listener, so construct message */
738 		rp = sotorawcb(so);
739 	}
740 	if (rtm) {
741 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
742 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
743 			m_freem(m);
744 			m = NULL;
745 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
746 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
747 		Free(rtm);
748 	}
749 	if (m) {
750 		if (rp) {
751 			/*
752 			 * XXX insure we don't get a copy by
753 			 * invalidating our protocol
754 			 */
755 			unsigned short family = rp->rcb_proto.sp_family;
756 			rp->rcb_proto.sp_family = 0;
757 			rt_dispatch(m, info.rti_info[RTAX_DST]);
758 			rp->rcb_proto.sp_family = family;
759 		} else
760 			rt_dispatch(m, info.rti_info[RTAX_DST]);
761 	}
762     }
763 	return (error);
764 #undef	sa_equal
765 }
766 
767 static void
768 rt_setmetrics(u_long which, const struct rt_metrics *in,
769 	struct rt_metrics_lite *out)
770 {
771 #define metric(f, e) if (which & (f)) out->e = in->e;
772 	/*
773 	 * Only these are stored in the routing entry since introduction
774 	 * of tcp hostcache. The rest is ignored.
775 	 */
776 	metric(RTV_MTU, rmx_mtu);
777 	metric(RTV_WEIGHT, rmx_weight);
778 	/* Userland -> kernel timebase conversion. */
779 	if (which & RTV_EXPIRE)
780 		out->rmx_expire = in->rmx_expire ?
781 		    in->rmx_expire - time_second + time_uptime : 0;
782 #undef metric
783 }
784 
785 static void
786 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
787 {
788 #define metric(e) out->e = in->e;
789 	bzero(out, sizeof(*out));
790 	metric(rmx_mtu);
791 	metric(rmx_weight);
792 	/* Kernel -> userland timebase conversion. */
793 	out->rmx_expire = in->rmx_expire ?
794 	    in->rmx_expire - time_uptime + time_second : 0;
795 #undef metric
796 }
797 
798 /*
799  * Extract the addresses of the passed sockaddrs.
800  * Do a little sanity checking so as to avoid bad memory references.
801  * This data is derived straight from userland.
802  */
803 static int
804 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
805 {
806 	struct sockaddr *sa;
807 	int i;
808 
809 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
810 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
811 			continue;
812 		sa = (struct sockaddr *)cp;
813 		/*
814 		 * It won't fit.
815 		 */
816 		if (cp + sa->sa_len > cplim)
817 			return (EINVAL);
818 		/*
819 		 * there are no more.. quit now
820 		 * If there are more bits, they are in error.
821 		 * I've seen this. route(1) can evidently generate these.
822 		 * This causes kernel to core dump.
823 		 * for compatibility, If we see this, point to a safe address.
824 		 */
825 		if (sa->sa_len == 0) {
826 			rtinfo->rti_info[i] = &sa_zero;
827 			return (0); /* should be EINVAL but for compat */
828 		}
829 		/* accept it */
830 		rtinfo->rti_info[i] = sa;
831 		cp += SA_SIZE(sa);
832 	}
833 	return (0);
834 }
835 
836 static struct mbuf *
837 rt_msg1(int type, struct rt_addrinfo *rtinfo)
838 {
839 	struct rt_msghdr *rtm;
840 	struct mbuf *m;
841 	int i;
842 	struct sockaddr *sa;
843 	int len, dlen;
844 
845 	switch (type) {
846 
847 	case RTM_DELADDR:
848 	case RTM_NEWADDR:
849 		len = sizeof(struct ifa_msghdr);
850 		break;
851 
852 	case RTM_DELMADDR:
853 	case RTM_NEWMADDR:
854 		len = sizeof(struct ifma_msghdr);
855 		break;
856 
857 	case RTM_IFINFO:
858 		len = sizeof(struct if_msghdr);
859 		break;
860 
861 	case RTM_IFANNOUNCE:
862 	case RTM_IEEE80211:
863 		len = sizeof(struct if_announcemsghdr);
864 		break;
865 
866 	default:
867 		len = sizeof(struct rt_msghdr);
868 	}
869 	if (len > MCLBYTES)
870 		panic("rt_msg1");
871 	m = m_gethdr(M_DONTWAIT, MT_DATA);
872 	if (m && len > MHLEN) {
873 		MCLGET(m, M_DONTWAIT);
874 		if ((m->m_flags & M_EXT) == 0) {
875 			m_free(m);
876 			m = NULL;
877 		}
878 	}
879 	if (m == NULL)
880 		return (m);
881 	m->m_pkthdr.len = m->m_len = len;
882 	m->m_pkthdr.rcvif = NULL;
883 	rtm = mtod(m, struct rt_msghdr *);
884 	bzero((caddr_t)rtm, len);
885 	for (i = 0; i < RTAX_MAX; i++) {
886 		if ((sa = rtinfo->rti_info[i]) == NULL)
887 			continue;
888 		rtinfo->rti_addrs |= (1 << i);
889 		dlen = SA_SIZE(sa);
890 		m_copyback(m, len, dlen, (caddr_t)sa);
891 		len += dlen;
892 	}
893 	if (m->m_pkthdr.len != len) {
894 		m_freem(m);
895 		return (NULL);
896 	}
897 	rtm->rtm_msglen = len;
898 	rtm->rtm_version = RTM_VERSION;
899 	rtm->rtm_type = type;
900 	return (m);
901 }
902 
903 static int
904 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
905 {
906 	int i;
907 	int len, dlen, second_time = 0;
908 	caddr_t cp0;
909 
910 	rtinfo->rti_addrs = 0;
911 again:
912 	switch (type) {
913 
914 	case RTM_DELADDR:
915 	case RTM_NEWADDR:
916 		len = sizeof(struct ifa_msghdr);
917 		break;
918 
919 	case RTM_IFINFO:
920 		len = sizeof(struct if_msghdr);
921 		break;
922 
923 	case RTM_NEWMADDR:
924 		len = sizeof(struct ifma_msghdr);
925 		break;
926 
927 	default:
928 		len = sizeof(struct rt_msghdr);
929 	}
930 	cp0 = cp;
931 	if (cp0)
932 		cp += len;
933 	for (i = 0; i < RTAX_MAX; i++) {
934 		struct sockaddr *sa;
935 
936 		if ((sa = rtinfo->rti_info[i]) == NULL)
937 			continue;
938 		rtinfo->rti_addrs |= (1 << i);
939 		dlen = SA_SIZE(sa);
940 		if (cp) {
941 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
942 			cp += dlen;
943 		}
944 		len += dlen;
945 	}
946 	len = ALIGN(len);
947 	if (cp == NULL && w != NULL && !second_time) {
948 		struct walkarg *rw = w;
949 
950 		if (rw->w_req) {
951 			if (rw->w_tmemsize < len) {
952 				if (rw->w_tmem)
953 					free(rw->w_tmem, M_RTABLE);
954 				rw->w_tmem = (caddr_t)
955 					malloc(len, M_RTABLE, M_NOWAIT);
956 				if (rw->w_tmem)
957 					rw->w_tmemsize = len;
958 			}
959 			if (rw->w_tmem) {
960 				cp = rw->w_tmem;
961 				second_time = 1;
962 				goto again;
963 			}
964 		}
965 	}
966 	if (cp) {
967 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
968 
969 		rtm->rtm_version = RTM_VERSION;
970 		rtm->rtm_type = type;
971 		rtm->rtm_msglen = len;
972 	}
973 	return (len);
974 }
975 
976 /*
977  * This routine is called to generate a message from the routing
978  * socket indicating that a redirect has occured, a routing lookup
979  * has failed, or that a protocol has detected timeouts to a particular
980  * destination.
981  */
982 void
983 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
984 {
985 	struct rt_msghdr *rtm;
986 	struct mbuf *m;
987 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
988 
989 	if (route_cb.any_count == 0)
990 		return;
991 	m = rt_msg1(type, rtinfo);
992 	if (m == NULL)
993 		return;
994 	rtm = mtod(m, struct rt_msghdr *);
995 	rtm->rtm_flags = RTF_DONE | flags;
996 	rtm->rtm_errno = error;
997 	rtm->rtm_addrs = rtinfo->rti_addrs;
998 	rt_dispatch(m, sa);
999 }
1000 
1001 /*
1002  * This routine is called to generate a message from the routing
1003  * socket indicating that the status of a network interface has changed.
1004  */
1005 void
1006 rt_ifmsg(struct ifnet *ifp)
1007 {
1008 	struct if_msghdr *ifm;
1009 	struct mbuf *m;
1010 	struct rt_addrinfo info;
1011 
1012 	if (route_cb.any_count == 0)
1013 		return;
1014 	bzero((caddr_t)&info, sizeof(info));
1015 	m = rt_msg1(RTM_IFINFO, &info);
1016 	if (m == NULL)
1017 		return;
1018 	ifm = mtod(m, struct if_msghdr *);
1019 	ifm->ifm_index = ifp->if_index;
1020 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1021 	ifm->ifm_data = ifp->if_data;
1022 	ifm->ifm_addrs = 0;
1023 	rt_dispatch(m, NULL);
1024 }
1025 
1026 /*
1027  * This is called to generate messages from the routing socket
1028  * indicating a network interface has had addresses associated with it.
1029  * if we ever reverse the logic and replace messages TO the routing
1030  * socket indicate a request to configure interfaces, then it will
1031  * be unnecessary as the routing socket will automatically generate
1032  * copies of it.
1033  */
1034 void
1035 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1036 {
1037 	struct rt_addrinfo info;
1038 	struct sockaddr *sa = NULL;
1039 	int pass;
1040 	struct mbuf *m = NULL;
1041 	struct ifnet *ifp = ifa->ifa_ifp;
1042 
1043 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1044 		("unexpected cmd %u", cmd));
1045 #ifdef SCTP
1046 	/*
1047 	 * notify the SCTP stack
1048 	 * this will only get called when an address is added/deleted
1049 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1050 	 */
1051 	sctp_addr_change(ifa, cmd);
1052 #endif /* SCTP */
1053 	if (route_cb.any_count == 0)
1054 		return;
1055 	for (pass = 1; pass < 3; pass++) {
1056 		bzero((caddr_t)&info, sizeof(info));
1057 		if ((cmd == RTM_ADD && pass == 1) ||
1058 		    (cmd == RTM_DELETE && pass == 2)) {
1059 			struct ifa_msghdr *ifam;
1060 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1061 
1062 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1063 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1064 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1065 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1066 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1067 				continue;
1068 			ifam = mtod(m, struct ifa_msghdr *);
1069 			ifam->ifam_index = ifp->if_index;
1070 			ifam->ifam_metric = ifa->ifa_metric;
1071 			ifam->ifam_flags = ifa->ifa_flags;
1072 			ifam->ifam_addrs = info.rti_addrs;
1073 		}
1074 		if ((cmd == RTM_ADD && pass == 2) ||
1075 		    (cmd == RTM_DELETE && pass == 1)) {
1076 			struct rt_msghdr *rtm;
1077 
1078 			if (rt == NULL)
1079 				continue;
1080 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1081 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1082 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1083 			if ((m = rt_msg1(cmd, &info)) == NULL)
1084 				continue;
1085 			rtm = mtod(m, struct rt_msghdr *);
1086 			rtm->rtm_index = ifp->if_index;
1087 			rtm->rtm_flags |= rt->rt_flags;
1088 			rtm->rtm_errno = error;
1089 			rtm->rtm_addrs = info.rti_addrs;
1090 		}
1091 		rt_dispatch(m, sa);
1092 	}
1093 }
1094 
1095 /*
1096  * This is the analogue to the rt_newaddrmsg which performs the same
1097  * function but for multicast group memberhips.  This is easier since
1098  * there is no route state to worry about.
1099  */
1100 void
1101 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1102 {
1103 	struct rt_addrinfo info;
1104 	struct mbuf *m = NULL;
1105 	struct ifnet *ifp = ifma->ifma_ifp;
1106 	struct ifma_msghdr *ifmam;
1107 
1108 	if (route_cb.any_count == 0)
1109 		return;
1110 
1111 	bzero((caddr_t)&info, sizeof(info));
1112 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1113 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1114 	/*
1115 	 * If a link-layer address is present, present it as a ``gateway''
1116 	 * (similarly to how ARP entries, e.g., are presented).
1117 	 */
1118 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1119 	m = rt_msg1(cmd, &info);
1120 	if (m == NULL)
1121 		return;
1122 	ifmam = mtod(m, struct ifma_msghdr *);
1123 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1124 	    __func__));
1125 	ifmam->ifmam_index = ifp->if_index;
1126 	ifmam->ifmam_addrs = info.rti_addrs;
1127 	rt_dispatch(m, ifma->ifma_addr);
1128 }
1129 
1130 static struct mbuf *
1131 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1132 	struct rt_addrinfo *info)
1133 {
1134 	struct if_announcemsghdr *ifan;
1135 	struct mbuf *m;
1136 
1137 	if (route_cb.any_count == 0)
1138 		return NULL;
1139 	bzero((caddr_t)info, sizeof(*info));
1140 	m = rt_msg1(type, info);
1141 	if (m != NULL) {
1142 		ifan = mtod(m, struct if_announcemsghdr *);
1143 		ifan->ifan_index = ifp->if_index;
1144 		strlcpy(ifan->ifan_name, ifp->if_xname,
1145 			sizeof(ifan->ifan_name));
1146 		ifan->ifan_what = what;
1147 	}
1148 	return m;
1149 }
1150 
1151 /*
1152  * This is called to generate routing socket messages indicating
1153  * IEEE80211 wireless events.
1154  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1155  */
1156 void
1157 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1158 {
1159 	struct mbuf *m;
1160 	struct rt_addrinfo info;
1161 
1162 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1163 	if (m != NULL) {
1164 		/*
1165 		 * Append the ieee80211 data.  Try to stick it in the
1166 		 * mbuf containing the ifannounce msg; otherwise allocate
1167 		 * a new mbuf and append.
1168 		 *
1169 		 * NB: we assume m is a single mbuf.
1170 		 */
1171 		if (data_len > M_TRAILINGSPACE(m)) {
1172 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1173 			if (n == NULL) {
1174 				m_freem(m);
1175 				return;
1176 			}
1177 			bcopy(data, mtod(n, void *), data_len);
1178 			n->m_len = data_len;
1179 			m->m_next = n;
1180 		} else if (data_len > 0) {
1181 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1182 			m->m_len += data_len;
1183 		}
1184 		if (m->m_flags & M_PKTHDR)
1185 			m->m_pkthdr.len += data_len;
1186 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1187 		rt_dispatch(m, NULL);
1188 	}
1189 }
1190 
1191 /*
1192  * This is called to generate routing socket messages indicating
1193  * network interface arrival and departure.
1194  */
1195 void
1196 rt_ifannouncemsg(struct ifnet *ifp, int what)
1197 {
1198 	struct mbuf *m;
1199 	struct rt_addrinfo info;
1200 
1201 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1202 	if (m != NULL)
1203 		rt_dispatch(m, NULL);
1204 }
1205 
1206 static void
1207 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1208 {
1209 	INIT_VNET_NET(curvnet);
1210 	struct m_tag *tag;
1211 
1212 	/*
1213 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1214 	 * use when injecting the mbuf into the routing socket buffer from
1215 	 * the netisr.
1216 	 */
1217 	if (sa != NULL) {
1218 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1219 		    M_NOWAIT);
1220 		if (tag == NULL) {
1221 			m_freem(m);
1222 			return;
1223 		}
1224 		*(unsigned short *)(tag + 1) = sa->sa_family;
1225 		m_tag_prepend(m, tag);
1226 	}
1227 #ifdef VIMAGE
1228 	if (V_loif)
1229 		m->m_pkthdr.rcvif = V_loif;
1230 	else {
1231 		m_freem(m);
1232 		return;
1233 	}
1234 #endif
1235 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1236 }
1237 
1238 /*
1239  * This is used in dumping the kernel table via sysctl().
1240  */
1241 static int
1242 sysctl_dumpentry(struct radix_node *rn, void *vw)
1243 {
1244 	struct walkarg *w = vw;
1245 	struct rtentry *rt = (struct rtentry *)rn;
1246 	int error = 0, size;
1247 	struct rt_addrinfo info;
1248 
1249 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1250 		return 0;
1251 	if ((rt->rt_flags & RTF_HOST) == 0
1252 	    ? jailed(w->w_req->td->td_ucred)
1253 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1254 		return (0);
1255 	bzero((caddr_t)&info, sizeof(info));
1256 	info.rti_info[RTAX_DST] = rt_key(rt);
1257 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1258 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1259 	info.rti_info[RTAX_GENMASK] = 0;
1260 	if (rt->rt_ifp) {
1261 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1262 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1263 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1264 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1265 	}
1266 	size = rt_msg2(RTM_GET, &info, NULL, w);
1267 	if (w->w_req && w->w_tmem) {
1268 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1269 
1270 		rtm->rtm_flags = rt->rt_flags;
1271 		/*
1272 		 * let's be honest about this being a retarded hack
1273 		 */
1274 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1275 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1276 		rtm->rtm_index = rt->rt_ifp->if_index;
1277 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1278 		rtm->rtm_addrs = info.rti_addrs;
1279 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1280 		return (error);
1281 	}
1282 	return (error);
1283 }
1284 
1285 static int
1286 sysctl_iflist(int af, struct walkarg *w)
1287 {
1288 	INIT_VNET_NET(curvnet);
1289 	struct ifnet *ifp;
1290 	struct ifaddr *ifa;
1291 	struct rt_addrinfo info;
1292 	int len, error = 0;
1293 
1294 	bzero((caddr_t)&info, sizeof(info));
1295 	IFNET_RLOCK();
1296 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1297 		if (w->w_arg && w->w_arg != ifp->if_index)
1298 			continue;
1299 		ifa = ifp->if_addr;
1300 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1301 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1302 		info.rti_info[RTAX_IFP] = NULL;
1303 		if (w->w_req && w->w_tmem) {
1304 			struct if_msghdr *ifm;
1305 
1306 			ifm = (struct if_msghdr *)w->w_tmem;
1307 			ifm->ifm_index = ifp->if_index;
1308 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1309 			ifm->ifm_data = ifp->if_data;
1310 			ifm->ifm_addrs = info.rti_addrs;
1311 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
1312 			if (error)
1313 				goto done;
1314 		}
1315 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1316 			if (af && af != ifa->ifa_addr->sa_family)
1317 				continue;
1318 			if (prison_if(w->w_req->td->td_ucred,
1319 			    ifa->ifa_addr) != 0)
1320 				continue;
1321 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1322 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1323 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1324 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1325 			if (w->w_req && w->w_tmem) {
1326 				struct ifa_msghdr *ifam;
1327 
1328 				ifam = (struct ifa_msghdr *)w->w_tmem;
1329 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1330 				ifam->ifam_flags = ifa->ifa_flags;
1331 				ifam->ifam_metric = ifa->ifa_metric;
1332 				ifam->ifam_addrs = info.rti_addrs;
1333 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1334 				if (error)
1335 					goto done;
1336 			}
1337 		}
1338 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1339 			info.rti_info[RTAX_BRD] = NULL;
1340 	}
1341 done:
1342 	IFNET_RUNLOCK();
1343 	return (error);
1344 }
1345 
1346 static int
1347 sysctl_ifmalist(int af, struct walkarg *w)
1348 {
1349 	INIT_VNET_NET(curvnet);
1350 	struct ifnet *ifp;
1351 	struct ifmultiaddr *ifma;
1352 	struct	rt_addrinfo info;
1353 	int	len, error = 0;
1354 	struct ifaddr *ifa;
1355 
1356 	bzero((caddr_t)&info, sizeof(info));
1357 	IFNET_RLOCK();
1358 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1359 		if (w->w_arg && w->w_arg != ifp->if_index)
1360 			continue;
1361 		ifa = ifp->if_addr;
1362 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1363 		IF_ADDR_LOCK(ifp);
1364 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1365 			if (af && af != ifma->ifma_addr->sa_family)
1366 				continue;
1367 			if (prison_if(w->w_req->td->td_ucred,
1368 			    ifma->ifma_addr) != 0)
1369 				continue;
1370 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1371 			info.rti_info[RTAX_GATEWAY] =
1372 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1373 			    ifma->ifma_lladdr : NULL;
1374 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1375 			if (w->w_req && w->w_tmem) {
1376 				struct ifma_msghdr *ifmam;
1377 
1378 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1379 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1380 				ifmam->ifmam_flags = 0;
1381 				ifmam->ifmam_addrs = info.rti_addrs;
1382 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1383 				if (error) {
1384 					IF_ADDR_UNLOCK(ifp);
1385 					goto done;
1386 				}
1387 			}
1388 		}
1389 		IF_ADDR_UNLOCK(ifp);
1390 	}
1391 done:
1392 	IFNET_RUNLOCK();
1393 	return (error);
1394 }
1395 
1396 static int
1397 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1398 {
1399 	INIT_VNET_NET(curvnet);
1400 	int	*name = (int *)arg1;
1401 	u_int	namelen = arg2;
1402 	struct radix_node_head *rnh;
1403 	int	i, lim, error = EINVAL;
1404 	u_char	af;
1405 	struct	walkarg w;
1406 
1407 	name ++;
1408 	namelen--;
1409 	if (req->newptr)
1410 		return (EPERM);
1411 	if (namelen != 3)
1412 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1413 	af = name[0];
1414 	if (af > AF_MAX)
1415 		return (EINVAL);
1416 	bzero(&w, sizeof(w));
1417 	w.w_op = name[1];
1418 	w.w_arg = name[2];
1419 	w.w_req = req;
1420 
1421 	error = sysctl_wire_old_buffer(req, 0);
1422 	if (error)
1423 		return (error);
1424 	switch (w.w_op) {
1425 
1426 	case NET_RT_DUMP:
1427 	case NET_RT_FLAGS:
1428 		if (af == 0) {			/* dump all tables */
1429 			i = 1;
1430 			lim = AF_MAX;
1431 		} else				/* dump only one table */
1432 			i = lim = af;
1433 
1434 		/*
1435 		 * take care of llinfo entries, the caller must
1436 		 * specify an AF
1437 		 */
1438 		if (w.w_op == NET_RT_FLAGS &&
1439 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1440 			if (af != 0)
1441 				error = lltable_sysctl_dumparp(af, w.w_req);
1442 			else
1443 				error = EINVAL;
1444 			break;
1445 		}
1446 		/*
1447 		 * take care of routing entries
1448 		 */
1449 		for (error = 0; error == 0 && i <= lim; i++)
1450 			if ((rnh = V_rt_tables[req->td->td_proc->p_fibnum][i]) != NULL) {
1451 				RADIX_NODE_HEAD_LOCK(rnh);
1452 			    	error = rnh->rnh_walktree(rnh,
1453 				    sysctl_dumpentry, &w);
1454 				RADIX_NODE_HEAD_UNLOCK(rnh);
1455 			} else if (af != 0)
1456 				error = EAFNOSUPPORT;
1457 		break;
1458 
1459 	case NET_RT_IFLIST:
1460 		error = sysctl_iflist(af, &w);
1461 		break;
1462 
1463 	case NET_RT_IFMALIST:
1464 		error = sysctl_ifmalist(af, &w);
1465 		break;
1466 	}
1467 	if (w.w_tmem)
1468 		free(w.w_tmem, M_RTABLE);
1469 	return (error);
1470 }
1471 
1472 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1473 
1474 /*
1475  * Definitions of protocols supported in the ROUTE domain.
1476  */
1477 
1478 static struct domain routedomain;		/* or at least forward */
1479 
1480 static struct protosw routesw[] = {
1481 {
1482 	.pr_type =		SOCK_RAW,
1483 	.pr_domain =		&routedomain,
1484 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1485 	.pr_output =		route_output,
1486 	.pr_ctlinput =		raw_ctlinput,
1487 	.pr_init =		raw_init,
1488 	.pr_usrreqs =		&route_usrreqs
1489 }
1490 };
1491 
1492 static struct domain routedomain = {
1493 	.dom_family =		PF_ROUTE,
1494 	.dom_name =		 "route",
1495 	.dom_protosw =		routesw,
1496 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1497 };
1498 
1499 DOMAIN_SET(route);
1500