xref: /freebsd/sys/net/rtsock.c (revision e39e854e27f53a784c3982cbeb68f4ad1cfd9162)
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 #include <netinet/ip_carp.h>
67 #ifdef INET6
68 #include <netinet6/scope6_var.h>
69 #endif
70 
71 #if defined(INET) || defined(INET6)
72 #ifdef SCTP
73 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
74 #endif /* SCTP */
75 #endif
76 
77 #ifdef COMPAT_FREEBSD32
78 #include <sys/mount.h>
79 #include <compat/freebsd32/freebsd32.h>
80 
81 struct if_data32 {
82 	uint8_t	ifi_type;
83 	uint8_t	ifi_physical;
84 	uint8_t	ifi_addrlen;
85 	uint8_t	ifi_hdrlen;
86 	uint8_t	ifi_link_state;
87 	uint8_t	ifi_vhid;
88 	uint8_t	ifi_spare_char2;
89 	uint8_t	ifi_datalen;
90 	uint32_t ifi_mtu;
91 	uint32_t ifi_metric;
92 	uint32_t ifi_baudrate;
93 	uint32_t ifi_ipackets;
94 	uint32_t ifi_ierrors;
95 	uint32_t ifi_opackets;
96 	uint32_t ifi_oerrors;
97 	uint32_t ifi_collisions;
98 	uint32_t ifi_ibytes;
99 	uint32_t ifi_obytes;
100 	uint32_t ifi_imcasts;
101 	uint32_t ifi_omcasts;
102 	uint32_t ifi_iqdrops;
103 	uint32_t ifi_noproto;
104 	uint32_t ifi_hwassist;
105 	int32_t	ifi_epoch;
106 	struct	timeval32 ifi_lastchange;
107 };
108 
109 struct if_msghdr32 {
110 	uint16_t ifm_msglen;
111 	uint8_t	ifm_version;
112 	uint8_t	ifm_type;
113 	int32_t	ifm_addrs;
114 	int32_t	ifm_flags;
115 	uint16_t ifm_index;
116 	struct	if_data32 ifm_data;
117 };
118 
119 struct if_msghdrl32 {
120 	uint16_t ifm_msglen;
121 	uint8_t	ifm_version;
122 	uint8_t	ifm_type;
123 	int32_t	ifm_addrs;
124 	int32_t	ifm_flags;
125 	uint16_t ifm_index;
126 	uint16_t _ifm_spare1;
127 	uint16_t ifm_len;
128 	uint16_t ifm_data_off;
129 	struct	if_data32 ifm_data;
130 };
131 
132 struct ifa_msghdrl32 {
133 	uint16_t ifam_msglen;
134 	uint8_t	ifam_version;
135 	uint8_t	ifam_type;
136 	int32_t	ifam_addrs;
137 	int32_t	ifam_flags;
138 	uint16_t ifam_index;
139 	uint16_t _ifam_spare1;
140 	uint16_t ifam_len;
141 	uint16_t ifam_data_off;
142 	int32_t	ifam_metric;
143 	struct	if_data32 ifam_data;
144 };
145 #endif /* COMPAT_FREEBSD32 */
146 
147 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
148 
149 /* NB: these are not modified */
150 static struct	sockaddr route_src = { 2, PF_ROUTE, };
151 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
152 
153 /* These are external hooks for CARP. */
154 int	(*carp_get_vhid_p)(struct ifaddr *);
155 
156 /*
157  * Used by rtsock/raw_input callback code to decide whether to filter the update
158  * notification to a socket bound to a particular FIB.
159  */
160 #define	RTS_FILTER_FIB	M_PROTO8
161 #define	RTS_ALLFIBS	-1
162 
163 static struct {
164 	int	ip_count;	/* attached w/ AF_INET */
165 	int	ip6_count;	/* attached w/ AF_INET6 */
166 	int	ipx_count;	/* attached w/ AF_IPX */
167 	int	any_count;	/* total attached */
168 } route_cb;
169 
170 struct mtx rtsock_mtx;
171 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
172 
173 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
174 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
175 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
176 
177 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
178 
179 struct walkarg {
180 	int	w_tmemsize;
181 	int	w_op, w_arg;
182 	caddr_t	w_tmem;
183 	struct sysctl_req *w_req;
184 };
185 
186 static void	rts_input(struct mbuf *m);
187 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
188 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
189 			caddr_t cp, struct walkarg *w);
190 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
191 			struct rt_addrinfo *rtinfo);
192 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
193 static int	sysctl_iflist(int af, struct walkarg *w);
194 static int	sysctl_ifmalist(int af, struct walkarg *w);
195 static int	route_output(struct mbuf *m, struct socket *so);
196 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
197 			struct rt_metrics_lite *out);
198 static void	rt_getmetrics(const struct rt_metrics_lite *in,
199 			struct rt_metrics *out);
200 static void	rt_dispatch(struct mbuf *, sa_family_t);
201 
202 static struct netisr_handler rtsock_nh = {
203 	.nh_name = "rtsock",
204 	.nh_handler = rts_input,
205 	.nh_proto = NETISR_ROUTE,
206 	.nh_policy = NETISR_POLICY_SOURCE,
207 };
208 
209 static int
210 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
211 {
212 	int error, qlimit;
213 
214 	netisr_getqlimit(&rtsock_nh, &qlimit);
215 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
216         if (error || !req->newptr)
217                 return (error);
218 	if (qlimit < 1)
219 		return (EINVAL);
220 	return (netisr_setqlimit(&rtsock_nh, qlimit));
221 }
222 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
223     0, 0, sysctl_route_netisr_maxqlen, "I",
224     "maximum routing socket dispatch queue length");
225 
226 static void
227 rts_init(void)
228 {
229 	int tmp;
230 
231 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
232 		rtsock_nh.nh_qlimit = tmp;
233 	netisr_register(&rtsock_nh);
234 }
235 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
236 
237 static int
238 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
239     struct rawcb *rp)
240 {
241 	int fibnum;
242 
243 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
244 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
245 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
246 
247 	/* No filtering requested. */
248 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
249 		return (0);
250 
251 	/* Check if it is a rts and the fib matches the one of the socket. */
252 	fibnum = M_GETFIB(m);
253 	if (proto->sp_family != PF_ROUTE ||
254 	    rp->rcb_socket == NULL ||
255 	    rp->rcb_socket->so_fibnum == fibnum)
256 		return (0);
257 
258 	/* Filtering requested and no match, the socket shall be skipped. */
259 	return (1);
260 }
261 
262 static void
263 rts_input(struct mbuf *m)
264 {
265 	struct sockproto route_proto;
266 	unsigned short *family;
267 	struct m_tag *tag;
268 
269 	route_proto.sp_family = PF_ROUTE;
270 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
271 	if (tag != NULL) {
272 		family = (unsigned short *)(tag + 1);
273 		route_proto.sp_protocol = *family;
274 		m_tag_delete(m, tag);
275 	} else
276 		route_proto.sp_protocol = 0;
277 
278 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
279 }
280 
281 /*
282  * It really doesn't make any sense at all for this code to share much
283  * with raw_usrreq.c, since its functionality is so restricted.  XXX
284  */
285 static void
286 rts_abort(struct socket *so)
287 {
288 
289 	raw_usrreqs.pru_abort(so);
290 }
291 
292 static void
293 rts_close(struct socket *so)
294 {
295 
296 	raw_usrreqs.pru_close(so);
297 }
298 
299 /* pru_accept is EOPNOTSUPP */
300 
301 static int
302 rts_attach(struct socket *so, int proto, struct thread *td)
303 {
304 	struct rawcb *rp;
305 	int s, error;
306 
307 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
308 
309 	/* XXX */
310 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
311 	if (rp == NULL)
312 		return ENOBUFS;
313 
314 	/*
315 	 * The splnet() is necessary to block protocols from sending
316 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
317 	 * this PCB is extant but incompletely initialized.
318 	 * Probably we should try to do more of this work beforehand and
319 	 * eliminate the spl.
320 	 */
321 	s = splnet();
322 	so->so_pcb = (caddr_t)rp;
323 	so->so_fibnum = td->td_proc->p_fibnum;
324 	error = raw_attach(so, proto);
325 	rp = sotorawcb(so);
326 	if (error) {
327 		splx(s);
328 		so->so_pcb = NULL;
329 		free(rp, M_PCB);
330 		return error;
331 	}
332 	RTSOCK_LOCK();
333 	switch(rp->rcb_proto.sp_protocol) {
334 	case AF_INET:
335 		route_cb.ip_count++;
336 		break;
337 	case AF_INET6:
338 		route_cb.ip6_count++;
339 		break;
340 	case AF_IPX:
341 		route_cb.ipx_count++;
342 		break;
343 	}
344 	route_cb.any_count++;
345 	RTSOCK_UNLOCK();
346 	soisconnected(so);
347 	so->so_options |= SO_USELOOPBACK;
348 	splx(s);
349 	return 0;
350 }
351 
352 static int
353 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
354 {
355 
356 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
357 }
358 
359 static int
360 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
361 {
362 
363 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
364 }
365 
366 /* pru_connect2 is EOPNOTSUPP */
367 /* pru_control is EOPNOTSUPP */
368 
369 static void
370 rts_detach(struct socket *so)
371 {
372 	struct rawcb *rp = sotorawcb(so);
373 
374 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
375 
376 	RTSOCK_LOCK();
377 	switch(rp->rcb_proto.sp_protocol) {
378 	case AF_INET:
379 		route_cb.ip_count--;
380 		break;
381 	case AF_INET6:
382 		route_cb.ip6_count--;
383 		break;
384 	case AF_IPX:
385 		route_cb.ipx_count--;
386 		break;
387 	}
388 	route_cb.any_count--;
389 	RTSOCK_UNLOCK();
390 	raw_usrreqs.pru_detach(so);
391 }
392 
393 static int
394 rts_disconnect(struct socket *so)
395 {
396 
397 	return (raw_usrreqs.pru_disconnect(so));
398 }
399 
400 /* pru_listen is EOPNOTSUPP */
401 
402 static int
403 rts_peeraddr(struct socket *so, struct sockaddr **nam)
404 {
405 
406 	return (raw_usrreqs.pru_peeraddr(so, nam));
407 }
408 
409 /* pru_rcvd is EOPNOTSUPP */
410 /* pru_rcvoob is EOPNOTSUPP */
411 
412 static int
413 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
414 	 struct mbuf *control, struct thread *td)
415 {
416 
417 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
418 }
419 
420 /* pru_sense is null */
421 
422 static int
423 rts_shutdown(struct socket *so)
424 {
425 
426 	return (raw_usrreqs.pru_shutdown(so));
427 }
428 
429 static int
430 rts_sockaddr(struct socket *so, struct sockaddr **nam)
431 {
432 
433 	return (raw_usrreqs.pru_sockaddr(so, nam));
434 }
435 
436 static struct pr_usrreqs route_usrreqs = {
437 	.pru_abort =		rts_abort,
438 	.pru_attach =		rts_attach,
439 	.pru_bind =		rts_bind,
440 	.pru_connect =		rts_connect,
441 	.pru_detach =		rts_detach,
442 	.pru_disconnect =	rts_disconnect,
443 	.pru_peeraddr =		rts_peeraddr,
444 	.pru_send =		rts_send,
445 	.pru_shutdown =		rts_shutdown,
446 	.pru_sockaddr =		rts_sockaddr,
447 	.pru_close =		rts_close,
448 };
449 
450 #ifndef _SOCKADDR_UNION_DEFINED
451 #define	_SOCKADDR_UNION_DEFINED
452 /*
453  * The union of all possible address formats we handle.
454  */
455 union sockaddr_union {
456 	struct sockaddr		sa;
457 	struct sockaddr_in	sin;
458 	struct sockaddr_in6	sin6;
459 };
460 #endif /* _SOCKADDR_UNION_DEFINED */
461 
462 static int
463 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
464     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
465 {
466 
467 	/* First, see if the returned address is part of the jail. */
468 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
469 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
470 		return (0);
471 	}
472 
473 	switch (info->rti_info[RTAX_DST]->sa_family) {
474 #ifdef INET
475 	case AF_INET:
476 	{
477 		struct in_addr ia;
478 		struct ifaddr *ifa;
479 		int found;
480 
481 		found = 0;
482 		/*
483 		 * Try to find an address on the given outgoing interface
484 		 * that belongs to the jail.
485 		 */
486 		IF_ADDR_RLOCK(ifp);
487 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
488 			struct sockaddr *sa;
489 			sa = ifa->ifa_addr;
490 			if (sa->sa_family != AF_INET)
491 				continue;
492 			ia = ((struct sockaddr_in *)sa)->sin_addr;
493 			if (prison_check_ip4(cred, &ia) == 0) {
494 				found = 1;
495 				break;
496 			}
497 		}
498 		IF_ADDR_RUNLOCK(ifp);
499 		if (!found) {
500 			/*
501 			 * As a last resort return the 'default' jail address.
502 			 */
503 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
504 			    sin_addr;
505 			if (prison_get_ip4(cred, &ia) != 0)
506 				return (ESRCH);
507 		}
508 		bzero(&saun->sin, sizeof(struct sockaddr_in));
509 		saun->sin.sin_len = sizeof(struct sockaddr_in);
510 		saun->sin.sin_family = AF_INET;
511 		saun->sin.sin_addr.s_addr = ia.s_addr;
512 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
513 		break;
514 	}
515 #endif
516 #ifdef INET6
517 	case AF_INET6:
518 	{
519 		struct in6_addr ia6;
520 		struct ifaddr *ifa;
521 		int found;
522 
523 		found = 0;
524 		/*
525 		 * Try to find an address on the given outgoing interface
526 		 * that belongs to the jail.
527 		 */
528 		IF_ADDR_RLOCK(ifp);
529 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
530 			struct sockaddr *sa;
531 			sa = ifa->ifa_addr;
532 			if (sa->sa_family != AF_INET6)
533 				continue;
534 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
535 			    &ia6, sizeof(struct in6_addr));
536 			if (prison_check_ip6(cred, &ia6) == 0) {
537 				found = 1;
538 				break;
539 			}
540 		}
541 		IF_ADDR_RUNLOCK(ifp);
542 		if (!found) {
543 			/*
544 			 * As a last resort return the 'default' jail address.
545 			 */
546 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
547 			    sin6_addr;
548 			if (prison_get_ip6(cred, &ia6) != 0)
549 				return (ESRCH);
550 		}
551 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
552 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
553 		saun->sin6.sin6_family = AF_INET6;
554 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
555 		if (sa6_recoverscope(&saun->sin6) != 0)
556 			return (ESRCH);
557 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
558 		break;
559 	}
560 #endif
561 	default:
562 		return (ESRCH);
563 	}
564 	return (0);
565 }
566 
567 /*ARGSUSED*/
568 static int
569 route_output(struct mbuf *m, struct socket *so)
570 {
571 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
572 	struct rt_msghdr *rtm = NULL;
573 	struct rtentry *rt = NULL;
574 	struct radix_node_head *rnh;
575 	struct rt_addrinfo info;
576 	int len, error = 0;
577 	struct ifnet *ifp = NULL;
578 	union sockaddr_union saun;
579 	sa_family_t saf = AF_UNSPEC;
580 
581 #define senderr(e) { error = e; goto flush;}
582 	if (m == NULL || ((m->m_len < sizeof(long)) &&
583 		       (m = m_pullup(m, sizeof(long))) == NULL))
584 		return (ENOBUFS);
585 	if ((m->m_flags & M_PKTHDR) == 0)
586 		panic("route_output");
587 	len = m->m_pkthdr.len;
588 	if (len < sizeof(*rtm) ||
589 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
590 		info.rti_info[RTAX_DST] = NULL;
591 		senderr(EINVAL);
592 	}
593 	R_Malloc(rtm, struct rt_msghdr *, len);
594 	if (rtm == NULL) {
595 		info.rti_info[RTAX_DST] = NULL;
596 		senderr(ENOBUFS);
597 	}
598 	m_copydata(m, 0, len, (caddr_t)rtm);
599 	if (rtm->rtm_version != RTM_VERSION) {
600 		info.rti_info[RTAX_DST] = NULL;
601 		senderr(EPROTONOSUPPORT);
602 	}
603 	rtm->rtm_pid = curproc->p_pid;
604 	bzero(&info, sizeof(info));
605 	info.rti_addrs = rtm->rtm_addrs;
606 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
607 		info.rti_info[RTAX_DST] = NULL;
608 		senderr(EINVAL);
609 	}
610 	info.rti_flags = rtm->rtm_flags;
611 	if (info.rti_info[RTAX_DST] == NULL ||
612 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
613 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
614 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
615 		senderr(EINVAL);
616 	saf = info.rti_info[RTAX_DST]->sa_family;
617 	/*
618 	 * Verify that the caller has the appropriate privilege; RTM_GET
619 	 * is the only operation the non-superuser is allowed.
620 	 */
621 	if (rtm->rtm_type != RTM_GET) {
622 		error = priv_check(curthread, PRIV_NET_ROUTE);
623 		if (error)
624 			senderr(error);
625 	}
626 
627 	/*
628 	 * The given gateway address may be an interface address.
629 	 * For example, issuing a "route change" command on a route
630 	 * entry that was created from a tunnel, and the gateway
631 	 * address given is the local end point. In this case the
632 	 * RTF_GATEWAY flag must be cleared or the destination will
633 	 * not be reachable even though there is no error message.
634 	 */
635 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
636 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
637 		struct route gw_ro;
638 
639 		bzero(&gw_ro, sizeof(gw_ro));
640 		gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
641 		rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
642 		/*
643 		 * A host route through the loopback interface is
644 		 * installed for each interface adddress. In pre 8.0
645 		 * releases the interface address of a PPP link type
646 		 * is not reachable locally. This behavior is fixed as
647 		 * part of the new L2/L3 redesign and rewrite work. The
648 		 * signature of this interface address route is the
649 		 * AF_LINK sa_family type of the rt_gateway, and the
650 		 * rt_ifp has the IFF_LOOPBACK flag set.
651 		 */
652 		if (gw_ro.ro_rt != NULL &&
653 		    gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
654 		    gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)
655 			info.rti_flags &= ~RTF_GATEWAY;
656 		if (gw_ro.ro_rt != NULL)
657 			RTFREE(gw_ro.ro_rt);
658 	}
659 
660 	switch (rtm->rtm_type) {
661 		struct rtentry *saved_nrt;
662 
663 	case RTM_ADD:
664 		if (info.rti_info[RTAX_GATEWAY] == NULL)
665 			senderr(EINVAL);
666 		saved_nrt = NULL;
667 
668 		/* support for new ARP code */
669 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
670 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
671 			error = lla_rt_output(rtm, &info);
672 			break;
673 		}
674 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
675 		    so->so_fibnum);
676 		if (error == 0 && saved_nrt) {
677 			RT_LOCK(saved_nrt);
678 			rt_setmetrics(rtm->rtm_inits,
679 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
680 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
681 			RT_REMREF(saved_nrt);
682 			RT_UNLOCK(saved_nrt);
683 		}
684 		break;
685 
686 	case RTM_DELETE:
687 		saved_nrt = NULL;
688 		/* support for new ARP code */
689 		if (info.rti_info[RTAX_GATEWAY] &&
690 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
691 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
692 			error = lla_rt_output(rtm, &info);
693 			break;
694 		}
695 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
696 		    so->so_fibnum);
697 		if (error == 0) {
698 			RT_LOCK(saved_nrt);
699 			rt = saved_nrt;
700 			goto report;
701 		}
702 		break;
703 
704 	case RTM_GET:
705 	case RTM_CHANGE:
706 	case RTM_LOCK:
707 		rnh = rt_tables_get_rnh(so->so_fibnum,
708 		    info.rti_info[RTAX_DST]->sa_family);
709 		if (rnh == NULL)
710 			senderr(EAFNOSUPPORT);
711 		RADIX_NODE_HEAD_RLOCK(rnh);
712 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
713 			info.rti_info[RTAX_NETMASK], rnh);
714 		if (rt == NULL) {	/* XXX looks bogus */
715 			RADIX_NODE_HEAD_RUNLOCK(rnh);
716 			senderr(ESRCH);
717 		}
718 #ifdef RADIX_MPATH
719 		/*
720 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
721 		 * we require users to specify a matching RTAX_GATEWAY.
722 		 *
723 		 * for RTM_GET, gate is optional even with multipath.
724 		 * if gate == NULL the first match is returned.
725 		 * (no need to call rt_mpath_matchgate if gate == NULL)
726 		 */
727 		if (rn_mpath_capable(rnh) &&
728 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
729 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
730 			if (!rt) {
731 				RADIX_NODE_HEAD_RUNLOCK(rnh);
732 				senderr(ESRCH);
733 			}
734 		}
735 #endif
736 		/*
737 		 * If performing proxied L2 entry insertion, and
738 		 * the actual PPP host entry is found, perform
739 		 * another search to retrieve the prefix route of
740 		 * the local end point of the PPP link.
741 		 */
742 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
743 			struct sockaddr laddr;
744 
745 			if (rt->rt_ifp != NULL &&
746 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
747 				struct ifaddr *ifa;
748 
749 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
750 				if (ifa != NULL)
751 					rt_maskedcopy(ifa->ifa_addr,
752 						      &laddr,
753 						      ifa->ifa_netmask);
754 			} else
755 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
756 					      &laddr,
757 					      rt->rt_ifa->ifa_netmask);
758 			/*
759 			 * refactor rt and no lock operation necessary
760 			 */
761 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
762 			if (rt == NULL) {
763 				RADIX_NODE_HEAD_RUNLOCK(rnh);
764 				senderr(ESRCH);
765 			}
766 		}
767 		RT_LOCK(rt);
768 		RT_ADDREF(rt);
769 		RADIX_NODE_HEAD_RUNLOCK(rnh);
770 
771 		/*
772 		 * Fix for PR: 82974
773 		 *
774 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
775 		 * returns a perfect match in case a netmask is
776 		 * specified.  For host routes only a longest prefix
777 		 * match is returned so it is necessary to compare the
778 		 * existence of the netmask.  If both have a netmask
779 		 * rnh_lookup() did a perfect match and if none of them
780 		 * have a netmask both are host routes which is also a
781 		 * perfect match.
782 		 */
783 
784 		if (rtm->rtm_type != RTM_GET &&
785 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
786 			RT_UNLOCK(rt);
787 			senderr(ESRCH);
788 		}
789 
790 		switch(rtm->rtm_type) {
791 
792 		case RTM_GET:
793 		report:
794 			RT_LOCK_ASSERT(rt);
795 			if ((rt->rt_flags & RTF_HOST) == 0
796 			    ? jailed_without_vnet(curthread->td_ucred)
797 			    : prison_if(curthread->td_ucred,
798 			    rt_key(rt)) != 0) {
799 				RT_UNLOCK(rt);
800 				senderr(ESRCH);
801 			}
802 			info.rti_info[RTAX_DST] = rt_key(rt);
803 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
804 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
805 			info.rti_info[RTAX_GENMASK] = 0;
806 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
807 				ifp = rt->rt_ifp;
808 				if (ifp) {
809 					info.rti_info[RTAX_IFP] =
810 					    ifp->if_addr->ifa_addr;
811 					error = rtm_get_jailed(&info, ifp, rt,
812 					    &saun, curthread->td_ucred);
813 					if (error != 0) {
814 						RT_UNLOCK(rt);
815 						senderr(error);
816 					}
817 					if (ifp->if_flags & IFF_POINTOPOINT)
818 						info.rti_info[RTAX_BRD] =
819 						    rt->rt_ifa->ifa_dstaddr;
820 					rtm->rtm_index = ifp->if_index;
821 				} else {
822 					info.rti_info[RTAX_IFP] = NULL;
823 					info.rti_info[RTAX_IFA] = NULL;
824 				}
825 			} else if ((ifp = rt->rt_ifp) != NULL) {
826 				rtm->rtm_index = ifp->if_index;
827 			}
828 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
829 			if (len > rtm->rtm_msglen) {
830 				struct rt_msghdr *new_rtm;
831 				R_Malloc(new_rtm, struct rt_msghdr *, len);
832 				if (new_rtm == NULL) {
833 					RT_UNLOCK(rt);
834 					senderr(ENOBUFS);
835 				}
836 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
837 				Free(rtm); rtm = new_rtm;
838 			}
839 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
840 			rtm->rtm_flags = rt->rt_flags;
841 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
842 			rtm->rtm_addrs = info.rti_addrs;
843 			break;
844 
845 		case RTM_CHANGE:
846 			/*
847 			 * New gateway could require new ifaddr, ifp;
848 			 * flags may also be different; ifp may be specified
849 			 * by ll sockaddr when protocol address is ambiguous
850 			 */
851 			if (((rt->rt_flags & RTF_GATEWAY) &&
852 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
853 			    info.rti_info[RTAX_IFP] != NULL ||
854 			    (info.rti_info[RTAX_IFA] != NULL &&
855 			     !sa_equal(info.rti_info[RTAX_IFA],
856 				       rt->rt_ifa->ifa_addr))) {
857 				RT_UNLOCK(rt);
858 				RADIX_NODE_HEAD_LOCK(rnh);
859 				error = rt_getifa_fib(&info, rt->rt_fibnum);
860 				/*
861 				 * XXXRW: Really we should release this
862 				 * reference later, but this maintains
863 				 * historical behavior.
864 				 */
865 				if (info.rti_ifa != NULL)
866 					ifa_free(info.rti_ifa);
867 				RADIX_NODE_HEAD_UNLOCK(rnh);
868 				if (error != 0)
869 					senderr(error);
870 				RT_LOCK(rt);
871 			}
872 			if (info.rti_ifa != NULL &&
873 			    info.rti_ifa != rt->rt_ifa &&
874 			    rt->rt_ifa != NULL &&
875 			    rt->rt_ifa->ifa_rtrequest != NULL) {
876 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
877 				    &info);
878 				ifa_free(rt->rt_ifa);
879 			}
880 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
881 				RT_UNLOCK(rt);
882 				RADIX_NODE_HEAD_LOCK(rnh);
883 				RT_LOCK(rt);
884 
885 				error = rt_setgate(rt, rt_key(rt),
886 				    info.rti_info[RTAX_GATEWAY]);
887 				RADIX_NODE_HEAD_UNLOCK(rnh);
888 				if (error != 0) {
889 					RT_UNLOCK(rt);
890 					senderr(error);
891 				}
892 				rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
893 			}
894 			if (info.rti_ifa != NULL &&
895 			    info.rti_ifa != rt->rt_ifa) {
896 				ifa_ref(info.rti_ifa);
897 				rt->rt_ifa = info.rti_ifa;
898 				rt->rt_ifp = info.rti_ifp;
899 			}
900 			/* Allow some flags to be toggled on change. */
901 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
902 				    (rtm->rtm_flags & RTF_FMASK);
903 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
904 					&rt->rt_rmx);
905 			rtm->rtm_index = rt->rt_ifp->if_index;
906 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
907 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
908 			/* FALLTHROUGH */
909 		case RTM_LOCK:
910 			/* We don't support locks anymore */
911 			break;
912 		}
913 		RT_UNLOCK(rt);
914 		break;
915 
916 	default:
917 		senderr(EOPNOTSUPP);
918 	}
919 
920 flush:
921 	if (rtm) {
922 		if (error)
923 			rtm->rtm_errno = error;
924 		else
925 			rtm->rtm_flags |= RTF_DONE;
926 	}
927 	if (rt)		/* XXX can this be true? */
928 		RTFREE(rt);
929     {
930 	struct rawcb *rp = NULL;
931 	/*
932 	 * Check to see if we don't want our own messages.
933 	 */
934 	if ((so->so_options & SO_USELOOPBACK) == 0) {
935 		if (route_cb.any_count <= 1) {
936 			if (rtm)
937 				Free(rtm);
938 			m_freem(m);
939 			return (error);
940 		}
941 		/* There is another listener, so construct message */
942 		rp = sotorawcb(so);
943 	}
944 	if (rtm) {
945 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
946 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
947 			m_freem(m);
948 			m = NULL;
949 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
950 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
951 	}
952 	if (m) {
953 		M_SETFIB(m, so->so_fibnum);
954 		m->m_flags |= RTS_FILTER_FIB;
955 		if (rp) {
956 			/*
957 			 * XXX insure we don't get a copy by
958 			 * invalidating our protocol
959 			 */
960 			unsigned short family = rp->rcb_proto.sp_family;
961 			rp->rcb_proto.sp_family = 0;
962 			rt_dispatch(m, saf);
963 			rp->rcb_proto.sp_family = family;
964 		} else
965 			rt_dispatch(m, saf);
966 	}
967 	/* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
968 	if (rtm)
969 		Free(rtm);
970     }
971 	return (error);
972 #undef	sa_equal
973 }
974 
975 static void
976 rt_setmetrics(u_long which, const struct rt_metrics *in,
977 	struct rt_metrics_lite *out)
978 {
979 #define metric(f, e) if (which & (f)) out->e = in->e;
980 	/*
981 	 * Only these are stored in the routing entry since introduction
982 	 * of tcp hostcache. The rest is ignored.
983 	 */
984 	metric(RTV_MTU, rmx_mtu);
985 	metric(RTV_WEIGHT, rmx_weight);
986 	/* Userland -> kernel timebase conversion. */
987 	if (which & RTV_EXPIRE)
988 		out->rmx_expire = in->rmx_expire ?
989 		    in->rmx_expire - time_second + time_uptime : 0;
990 #undef metric
991 }
992 
993 static void
994 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
995 {
996 #define metric(e) out->e = in->e;
997 	bzero(out, sizeof(*out));
998 	metric(rmx_mtu);
999 	metric(rmx_weight);
1000 	/* Kernel -> userland timebase conversion. */
1001 	out->rmx_expire = in->rmx_expire ?
1002 	    in->rmx_expire - time_uptime + time_second : 0;
1003 #undef metric
1004 }
1005 
1006 /*
1007  * Extract the addresses of the passed sockaddrs.
1008  * Do a little sanity checking so as to avoid bad memory references.
1009  * This data is derived straight from userland.
1010  */
1011 static int
1012 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1013 {
1014 	struct sockaddr *sa;
1015 	int i;
1016 
1017 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1018 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1019 			continue;
1020 		sa = (struct sockaddr *)cp;
1021 		/*
1022 		 * It won't fit.
1023 		 */
1024 		if (cp + sa->sa_len > cplim)
1025 			return (EINVAL);
1026 		/*
1027 		 * there are no more.. quit now
1028 		 * If there are more bits, they are in error.
1029 		 * I've seen this. route(1) can evidently generate these.
1030 		 * This causes kernel to core dump.
1031 		 * for compatibility, If we see this, point to a safe address.
1032 		 */
1033 		if (sa->sa_len == 0) {
1034 			rtinfo->rti_info[i] = &sa_zero;
1035 			return (0); /* should be EINVAL but for compat */
1036 		}
1037 		/* accept it */
1038 		rtinfo->rti_info[i] = sa;
1039 		cp += SA_SIZE(sa);
1040 	}
1041 	return (0);
1042 }
1043 
1044 /*
1045  * Used by the routing socket.
1046  */
1047 static struct mbuf *
1048 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1049 {
1050 	struct rt_msghdr *rtm;
1051 	struct mbuf *m;
1052 	int i;
1053 	struct sockaddr *sa;
1054 	int len, dlen;
1055 
1056 	switch (type) {
1057 
1058 	case RTM_DELADDR:
1059 	case RTM_NEWADDR:
1060 		len = sizeof(struct ifa_msghdr);
1061 		break;
1062 
1063 	case RTM_DELMADDR:
1064 	case RTM_NEWMADDR:
1065 		len = sizeof(struct ifma_msghdr);
1066 		break;
1067 
1068 	case RTM_IFINFO:
1069 		len = sizeof(struct if_msghdr);
1070 		break;
1071 
1072 	case RTM_IFANNOUNCE:
1073 	case RTM_IEEE80211:
1074 		len = sizeof(struct if_announcemsghdr);
1075 		break;
1076 
1077 	default:
1078 		len = sizeof(struct rt_msghdr);
1079 	}
1080 	if (len > MCLBYTES)
1081 		panic("rt_msg1");
1082 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1083 	if (m && len > MHLEN) {
1084 		MCLGET(m, M_DONTWAIT);
1085 		if ((m->m_flags & M_EXT) == 0) {
1086 			m_free(m);
1087 			m = NULL;
1088 		}
1089 	}
1090 	if (m == NULL)
1091 		return (m);
1092 	m->m_pkthdr.len = m->m_len = len;
1093 	m->m_pkthdr.rcvif = NULL;
1094 	rtm = mtod(m, struct rt_msghdr *);
1095 	bzero((caddr_t)rtm, len);
1096 	for (i = 0; i < RTAX_MAX; i++) {
1097 		if ((sa = rtinfo->rti_info[i]) == NULL)
1098 			continue;
1099 		rtinfo->rti_addrs |= (1 << i);
1100 		dlen = SA_SIZE(sa);
1101 		m_copyback(m, len, dlen, (caddr_t)sa);
1102 		len += dlen;
1103 	}
1104 	if (m->m_pkthdr.len != len) {
1105 		m_freem(m);
1106 		return (NULL);
1107 	}
1108 	rtm->rtm_msglen = len;
1109 	rtm->rtm_version = RTM_VERSION;
1110 	rtm->rtm_type = type;
1111 	return (m);
1112 }
1113 
1114 /*
1115  * Used by the sysctl code and routing socket.
1116  */
1117 static int
1118 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1119 {
1120 	int i;
1121 	int len, dlen, second_time = 0;
1122 	caddr_t cp0;
1123 
1124 	rtinfo->rti_addrs = 0;
1125 again:
1126 	switch (type) {
1127 
1128 	case RTM_DELADDR:
1129 	case RTM_NEWADDR:
1130 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1131 #ifdef COMPAT_FREEBSD32
1132 			if (w->w_req->flags & SCTL_MASK32)
1133 				len = sizeof(struct ifa_msghdrl32);
1134 			else
1135 #endif
1136 				len = sizeof(struct ifa_msghdrl);
1137 		} else
1138 			len = sizeof(struct ifa_msghdr);
1139 		break;
1140 
1141 	case RTM_IFINFO:
1142 #ifdef COMPAT_FREEBSD32
1143 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1144 			if (w->w_op == NET_RT_IFLISTL)
1145 				len = sizeof(struct if_msghdrl32);
1146 			else
1147 				len = sizeof(struct if_msghdr32);
1148 			break;
1149 		}
1150 #endif
1151 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1152 			len = sizeof(struct if_msghdrl);
1153 		else
1154 			len = sizeof(struct if_msghdr);
1155 		break;
1156 
1157 	case RTM_NEWMADDR:
1158 		len = sizeof(struct ifma_msghdr);
1159 		break;
1160 
1161 	default:
1162 		len = sizeof(struct rt_msghdr);
1163 	}
1164 	cp0 = cp;
1165 	if (cp0)
1166 		cp += len;
1167 	for (i = 0; i < RTAX_MAX; i++) {
1168 		struct sockaddr *sa;
1169 
1170 		if ((sa = rtinfo->rti_info[i]) == NULL)
1171 			continue;
1172 		rtinfo->rti_addrs |= (1 << i);
1173 		dlen = SA_SIZE(sa);
1174 		if (cp) {
1175 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1176 			cp += dlen;
1177 		}
1178 		len += dlen;
1179 	}
1180 	len = ALIGN(len);
1181 	if (cp == NULL && w != NULL && !second_time) {
1182 		struct walkarg *rw = w;
1183 
1184 		if (rw->w_req) {
1185 			if (rw->w_tmemsize < len) {
1186 				if (rw->w_tmem)
1187 					free(rw->w_tmem, M_RTABLE);
1188 				rw->w_tmem = (caddr_t)
1189 					malloc(len, M_RTABLE, M_NOWAIT);
1190 				if (rw->w_tmem)
1191 					rw->w_tmemsize = len;
1192 			}
1193 			if (rw->w_tmem) {
1194 				cp = rw->w_tmem;
1195 				second_time = 1;
1196 				goto again;
1197 			}
1198 		}
1199 	}
1200 	if (cp) {
1201 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1202 
1203 		rtm->rtm_version = RTM_VERSION;
1204 		rtm->rtm_type = type;
1205 		rtm->rtm_msglen = len;
1206 	}
1207 	return (len);
1208 }
1209 
1210 /*
1211  * This routine is called to generate a message from the routing
1212  * socket indicating that a redirect has occured, a routing lookup
1213  * has failed, or that a protocol has detected timeouts to a particular
1214  * destination.
1215  */
1216 void
1217 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1218     int fibnum)
1219 {
1220 	struct rt_msghdr *rtm;
1221 	struct mbuf *m;
1222 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1223 
1224 	if (route_cb.any_count == 0)
1225 		return;
1226 	m = rt_msg1(type, rtinfo);
1227 	if (m == NULL)
1228 		return;
1229 
1230 	if (fibnum != RTS_ALLFIBS) {
1231 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1232 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1233 		M_SETFIB(m, fibnum);
1234 		m->m_flags |= RTS_FILTER_FIB;
1235 	}
1236 
1237 	rtm = mtod(m, struct rt_msghdr *);
1238 	rtm->rtm_flags = RTF_DONE | flags;
1239 	rtm->rtm_errno = error;
1240 	rtm->rtm_addrs = rtinfo->rti_addrs;
1241 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1242 }
1243 
1244 void
1245 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1246 {
1247 
1248 	rt_missmsg_fib(type, rtinfo, flags, error, RTS_ALLFIBS);
1249 }
1250 
1251 /*
1252  * This routine is called to generate a message from the routing
1253  * socket indicating that the status of a network interface has changed.
1254  */
1255 void
1256 rt_ifmsg(struct ifnet *ifp)
1257 {
1258 	struct if_msghdr *ifm;
1259 	struct mbuf *m;
1260 	struct rt_addrinfo info;
1261 
1262 	if (route_cb.any_count == 0)
1263 		return;
1264 	bzero((caddr_t)&info, sizeof(info));
1265 	m = rt_msg1(RTM_IFINFO, &info);
1266 	if (m == NULL)
1267 		return;
1268 	ifm = mtod(m, struct if_msghdr *);
1269 	ifm->ifm_index = ifp->if_index;
1270 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1271 	ifm->ifm_data = ifp->if_data;
1272 	ifm->ifm_addrs = 0;
1273 	rt_dispatch(m, AF_UNSPEC);
1274 }
1275 
1276 /*
1277  * This is called to generate messages from the routing socket
1278  * indicating a network interface has had addresses associated with it.
1279  * if we ever reverse the logic and replace messages TO the routing
1280  * socket indicate a request to configure interfaces, then it will
1281  * be unnecessary as the routing socket will automatically generate
1282  * copies of it.
1283  */
1284 void
1285 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
1286     int fibnum)
1287 {
1288 	struct rt_addrinfo info;
1289 	struct sockaddr *sa = NULL;
1290 	int pass;
1291 	struct mbuf *m = NULL;
1292 	struct ifnet *ifp = ifa->ifa_ifp;
1293 
1294 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1295 		("unexpected cmd %u", cmd));
1296 #if defined(INET) || defined(INET6)
1297 #ifdef SCTP
1298 	/*
1299 	 * notify the SCTP stack
1300 	 * this will only get called when an address is added/deleted
1301 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1302 	 */
1303 	sctp_addr_change(ifa, cmd);
1304 #endif /* SCTP */
1305 #endif
1306 	if (route_cb.any_count == 0)
1307 		return;
1308 	for (pass = 1; pass < 3; pass++) {
1309 		bzero((caddr_t)&info, sizeof(info));
1310 		if ((cmd == RTM_ADD && pass == 1) ||
1311 		    (cmd == RTM_DELETE && pass == 2)) {
1312 			struct ifa_msghdr *ifam;
1313 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1314 
1315 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1316 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1317 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1318 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1319 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1320 				continue;
1321 			ifam = mtod(m, struct ifa_msghdr *);
1322 			ifam->ifam_index = ifp->if_index;
1323 			ifam->ifam_metric = ifa->ifa_metric;
1324 			ifam->ifam_flags = ifa->ifa_flags;
1325 			ifam->ifam_addrs = info.rti_addrs;
1326 		}
1327 		if ((cmd == RTM_ADD && pass == 2) ||
1328 		    (cmd == RTM_DELETE && pass == 1)) {
1329 			struct rt_msghdr *rtm;
1330 
1331 			if (rt == NULL)
1332 				continue;
1333 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1334 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1335 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1336 			if ((m = rt_msg1(cmd, &info)) == NULL)
1337 				continue;
1338 			rtm = mtod(m, struct rt_msghdr *);
1339 			rtm->rtm_index = ifp->if_index;
1340 			rtm->rtm_flags |= rt->rt_flags;
1341 			rtm->rtm_errno = error;
1342 			rtm->rtm_addrs = info.rti_addrs;
1343 		}
1344 		if (fibnum != RTS_ALLFIBS) {
1345 			KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: "
1346 			    "fibnum out of range 0 <= %d < %d", __func__,
1347 			     fibnum, rt_numfibs));
1348 			M_SETFIB(m, fibnum);
1349 			m->m_flags |= RTS_FILTER_FIB;
1350 		}
1351 		rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1352 	}
1353 }
1354 
1355 void
1356 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1357 {
1358 
1359 	rt_newaddrmsg_fib(cmd, ifa, error, rt, RTS_ALLFIBS);
1360 }
1361 
1362 /*
1363  * This is the analogue to the rt_newaddrmsg which performs the same
1364  * function but for multicast group memberhips.  This is easier since
1365  * there is no route state to worry about.
1366  */
1367 void
1368 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1369 {
1370 	struct rt_addrinfo info;
1371 	struct mbuf *m = NULL;
1372 	struct ifnet *ifp = ifma->ifma_ifp;
1373 	struct ifma_msghdr *ifmam;
1374 
1375 	if (route_cb.any_count == 0)
1376 		return;
1377 
1378 	bzero((caddr_t)&info, sizeof(info));
1379 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1380 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1381 	/*
1382 	 * If a link-layer address is present, present it as a ``gateway''
1383 	 * (similarly to how ARP entries, e.g., are presented).
1384 	 */
1385 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1386 	m = rt_msg1(cmd, &info);
1387 	if (m == NULL)
1388 		return;
1389 	ifmam = mtod(m, struct ifma_msghdr *);
1390 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1391 	    __func__));
1392 	ifmam->ifmam_index = ifp->if_index;
1393 	ifmam->ifmam_addrs = info.rti_addrs;
1394 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1395 }
1396 
1397 static struct mbuf *
1398 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1399 	struct rt_addrinfo *info)
1400 {
1401 	struct if_announcemsghdr *ifan;
1402 	struct mbuf *m;
1403 
1404 	if (route_cb.any_count == 0)
1405 		return NULL;
1406 	bzero((caddr_t)info, sizeof(*info));
1407 	m = rt_msg1(type, info);
1408 	if (m != NULL) {
1409 		ifan = mtod(m, struct if_announcemsghdr *);
1410 		ifan->ifan_index = ifp->if_index;
1411 		strlcpy(ifan->ifan_name, ifp->if_xname,
1412 			sizeof(ifan->ifan_name));
1413 		ifan->ifan_what = what;
1414 	}
1415 	return m;
1416 }
1417 
1418 /*
1419  * This is called to generate routing socket messages indicating
1420  * IEEE80211 wireless events.
1421  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1422  */
1423 void
1424 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1425 {
1426 	struct mbuf *m;
1427 	struct rt_addrinfo info;
1428 
1429 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1430 	if (m != NULL) {
1431 		/*
1432 		 * Append the ieee80211 data.  Try to stick it in the
1433 		 * mbuf containing the ifannounce msg; otherwise allocate
1434 		 * a new mbuf and append.
1435 		 *
1436 		 * NB: we assume m is a single mbuf.
1437 		 */
1438 		if (data_len > M_TRAILINGSPACE(m)) {
1439 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1440 			if (n == NULL) {
1441 				m_freem(m);
1442 				return;
1443 			}
1444 			bcopy(data, mtod(n, void *), data_len);
1445 			n->m_len = data_len;
1446 			m->m_next = n;
1447 		} else if (data_len > 0) {
1448 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1449 			m->m_len += data_len;
1450 		}
1451 		if (m->m_flags & M_PKTHDR)
1452 			m->m_pkthdr.len += data_len;
1453 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1454 		rt_dispatch(m, AF_UNSPEC);
1455 	}
1456 }
1457 
1458 /*
1459  * This is called to generate routing socket messages indicating
1460  * network interface arrival and departure.
1461  */
1462 void
1463 rt_ifannouncemsg(struct ifnet *ifp, int what)
1464 {
1465 	struct mbuf *m;
1466 	struct rt_addrinfo info;
1467 
1468 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1469 	if (m != NULL)
1470 		rt_dispatch(m, AF_UNSPEC);
1471 }
1472 
1473 static void
1474 rt_dispatch(struct mbuf *m, sa_family_t saf)
1475 {
1476 	struct m_tag *tag;
1477 
1478 	/*
1479 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1480 	 * use when injecting the mbuf into the routing socket buffer from
1481 	 * the netisr.
1482 	 */
1483 	if (saf != AF_UNSPEC) {
1484 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1485 		    M_NOWAIT);
1486 		if (tag == NULL) {
1487 			m_freem(m);
1488 			return;
1489 		}
1490 		*(unsigned short *)(tag + 1) = saf;
1491 		m_tag_prepend(m, tag);
1492 	}
1493 #ifdef VIMAGE
1494 	if (V_loif)
1495 		m->m_pkthdr.rcvif = V_loif;
1496 	else {
1497 		m_freem(m);
1498 		return;
1499 	}
1500 #endif
1501 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1502 }
1503 
1504 /*
1505  * This is used in dumping the kernel table via sysctl().
1506  */
1507 static int
1508 sysctl_dumpentry(struct radix_node *rn, void *vw)
1509 {
1510 	struct walkarg *w = vw;
1511 	struct rtentry *rt = (struct rtentry *)rn;
1512 	int error = 0, size;
1513 	struct rt_addrinfo info;
1514 
1515 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1516 		return 0;
1517 	if ((rt->rt_flags & RTF_HOST) == 0
1518 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1519 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1520 		return (0);
1521 	bzero((caddr_t)&info, sizeof(info));
1522 	info.rti_info[RTAX_DST] = rt_key(rt);
1523 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1524 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1525 	info.rti_info[RTAX_GENMASK] = 0;
1526 	if (rt->rt_ifp) {
1527 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1528 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1529 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1530 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1531 	}
1532 	size = rt_msg2(RTM_GET, &info, NULL, w);
1533 	if (w->w_req && w->w_tmem) {
1534 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1535 
1536 		rtm->rtm_flags = rt->rt_flags;
1537 		/*
1538 		 * let's be honest about this being a retarded hack
1539 		 */
1540 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1541 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1542 		rtm->rtm_index = rt->rt_ifp->if_index;
1543 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1544 		rtm->rtm_addrs = info.rti_addrs;
1545 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1546 		return (error);
1547 	}
1548 	return (error);
1549 }
1550 
1551 #ifdef COMPAT_FREEBSD32
1552 static void
1553 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1554 {
1555 
1556 	bzero(dst, sizeof(*dst));
1557 	CP(*src, *dst, ifi_type);
1558 	CP(*src, *dst, ifi_physical);
1559 	CP(*src, *dst, ifi_addrlen);
1560 	CP(*src, *dst, ifi_hdrlen);
1561 	CP(*src, *dst, ifi_link_state);
1562 	CP(*src, *dst, ifi_vhid);
1563 	dst->ifi_datalen = sizeof(struct if_data32);
1564 	CP(*src, *dst, ifi_mtu);
1565 	CP(*src, *dst, ifi_metric);
1566 	CP(*src, *dst, ifi_baudrate);
1567 	CP(*src, *dst, ifi_ipackets);
1568 	CP(*src, *dst, ifi_ierrors);
1569 	CP(*src, *dst, ifi_opackets);
1570 	CP(*src, *dst, ifi_oerrors);
1571 	CP(*src, *dst, ifi_collisions);
1572 	CP(*src, *dst, ifi_ibytes);
1573 	CP(*src, *dst, ifi_obytes);
1574 	CP(*src, *dst, ifi_imcasts);
1575 	CP(*src, *dst, ifi_omcasts);
1576 	CP(*src, *dst, ifi_iqdrops);
1577 	CP(*src, *dst, ifi_noproto);
1578 	CP(*src, *dst, ifi_hwassist);
1579 	CP(*src, *dst, ifi_epoch);
1580 	TV_CP(*src, *dst, ifi_lastchange);
1581 }
1582 #endif
1583 
1584 static int
1585 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info,
1586     struct walkarg *w, int len)
1587 {
1588 	struct if_msghdrl *ifm;
1589 
1590 #ifdef COMPAT_FREEBSD32
1591 	if (w->w_req->flags & SCTL_MASK32) {
1592 		struct if_msghdrl32 *ifm32;
1593 
1594 		ifm32 = (struct if_msghdrl32 *)w->w_tmem;
1595 		ifm32->ifm_addrs = info->rti_addrs;
1596 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1597 		ifm32->ifm_index = ifp->if_index;
1598 		ifm32->_ifm_spare1 = 0;
1599 		ifm32->ifm_len = sizeof(*ifm32);
1600 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1601 
1602 		copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1603 		/* Fixup if_data carp(4) vhid. */
1604 		if (carp_get_vhid_p != NULL)
1605 			ifm32->ifm_data.ifi_vhid =
1606 			    (*carp_get_vhid_p)(ifp->if_addr);
1607 
1608 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1609 	}
1610 #endif
1611 	ifm = (struct if_msghdrl *)w->w_tmem;
1612 	ifm->ifm_addrs = info->rti_addrs;
1613 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1614 	ifm->ifm_index = ifp->if_index;
1615 	ifm->_ifm_spare1 = 0;
1616 	ifm->ifm_len = sizeof(*ifm);
1617 	ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1618 
1619 	ifm->ifm_data = ifp->if_data;
1620 	/* Fixup if_data carp(4) vhid. */
1621 	if (carp_get_vhid_p != NULL)
1622 		ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1623 
1624 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1625 }
1626 
1627 static int
1628 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info,
1629     struct walkarg *w, int len)
1630 {
1631 	struct if_msghdr *ifm;
1632 
1633 #ifdef COMPAT_FREEBSD32
1634 	if (w->w_req->flags & SCTL_MASK32) {
1635 		struct if_msghdr32 *ifm32;
1636 
1637 		ifm32 = (struct if_msghdr32 *)w->w_tmem;
1638 		ifm32->ifm_addrs = info->rti_addrs;
1639 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1640 		ifm32->ifm_index = ifp->if_index;
1641 
1642 		copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1643 		/* Fixup if_data carp(4) vhid. */
1644 		if (carp_get_vhid_p != NULL)
1645 			ifm32->ifm_data.ifi_vhid =
1646 			    (*carp_get_vhid_p)(ifp->if_addr);
1647 
1648 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1649 	}
1650 #endif
1651 	ifm = (struct if_msghdr *)w->w_tmem;
1652 	ifm->ifm_addrs = info->rti_addrs;
1653 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1654 	ifm->ifm_index = ifp->if_index;
1655 
1656 	ifm->ifm_data = ifp->if_data;
1657 	/* Fixup if_data carp(4) vhid. */
1658 	if (carp_get_vhid_p != NULL)
1659 		ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1660 
1661 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1662 }
1663 
1664 static int
1665 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1666     struct walkarg *w, int len)
1667 {
1668 	struct ifa_msghdrl *ifam;
1669 
1670 #ifdef COMPAT_FREEBSD32
1671 	if (w->w_req->flags & SCTL_MASK32) {
1672 		struct ifa_msghdrl32 *ifam32;
1673 
1674 		ifam32 = (struct ifa_msghdrl32 *)w->w_tmem;
1675 		ifam32->ifam_addrs = info->rti_addrs;
1676 		ifam32->ifam_flags = ifa->ifa_flags;
1677 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1678 		ifam32->_ifam_spare1 = 0;
1679 		ifam32->ifam_len = sizeof(*ifam32);
1680 		ifam32->ifam_data_off =
1681 		    offsetof(struct ifa_msghdrl32, ifam_data);
1682 		ifam32->ifam_metric = ifa->ifa_metric;
1683 
1684 		copy_ifdata32(&ifa->ifa_ifp->if_data, &ifam32->ifam_data);
1685 		/* Fixup if_data carp(4) vhid. */
1686 		if (carp_get_vhid_p != NULL)
1687 			ifam32->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1688 
1689 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifam32, len));
1690 	}
1691 #endif
1692 
1693 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1694 	ifam->ifam_addrs = info->rti_addrs;
1695 	ifam->ifam_flags = ifa->ifa_flags;
1696 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1697 	ifam->_ifam_spare1 = 0;
1698 	ifam->ifam_len = sizeof(*ifam);
1699 	ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1700 	ifam->ifam_metric = ifa->ifa_metric;
1701 
1702 	ifam->ifam_data = ifa->if_data;
1703 	/* Fixup if_data carp(4) vhid. */
1704 	if (carp_get_vhid_p != NULL)
1705 		ifam->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1706 
1707 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1708 }
1709 
1710 static int
1711 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1712     struct walkarg *w, int len)
1713 {
1714 	struct ifa_msghdr *ifam;
1715 
1716 	ifam = (struct ifa_msghdr *)w->w_tmem;
1717 	ifam->ifam_addrs = info->rti_addrs;
1718 	ifam->ifam_flags = ifa->ifa_flags;
1719 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1720 	ifam->ifam_metric = ifa->ifa_metric;
1721 
1722 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1723 }
1724 
1725 static int
1726 sysctl_iflist(int af, struct walkarg *w)
1727 {
1728 	struct ifnet *ifp;
1729 	struct ifaddr *ifa;
1730 	struct rt_addrinfo info;
1731 	int len, error = 0;
1732 
1733 	bzero((caddr_t)&info, sizeof(info));
1734 	IFNET_RLOCK();
1735 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1736 		if (w->w_arg && w->w_arg != ifp->if_index)
1737 			continue;
1738 		IF_ADDR_RLOCK(ifp);
1739 		ifa = ifp->if_addr;
1740 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1741 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1742 		info.rti_info[RTAX_IFP] = NULL;
1743 		if (w->w_req && w->w_tmem) {
1744 			if (w->w_op == NET_RT_IFLISTL)
1745 				error = sysctl_iflist_ifml(ifp, &info, w, len);
1746 			else
1747 				error = sysctl_iflist_ifm(ifp, &info, w, len);
1748 			if (error)
1749 				goto done;
1750 		}
1751 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1752 			if (af && af != ifa->ifa_addr->sa_family)
1753 				continue;
1754 			if (prison_if(w->w_req->td->td_ucred,
1755 			    ifa->ifa_addr) != 0)
1756 				continue;
1757 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1758 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1759 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1760 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1761 			if (w->w_req && w->w_tmem) {
1762 				if (w->w_op == NET_RT_IFLISTL)
1763 					error = sysctl_iflist_ifaml(ifa, &info,
1764 					    w, len);
1765 				else
1766 					error = sysctl_iflist_ifam(ifa, &info,
1767 					    w, len);
1768 				if (error)
1769 					goto done;
1770 			}
1771 		}
1772 		IF_ADDR_RUNLOCK(ifp);
1773 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1774 			info.rti_info[RTAX_BRD] = NULL;
1775 	}
1776 done:
1777 	if (ifp != NULL)
1778 		IF_ADDR_RUNLOCK(ifp);
1779 	IFNET_RUNLOCK();
1780 	return (error);
1781 }
1782 
1783 static int
1784 sysctl_ifmalist(int af, struct walkarg *w)
1785 {
1786 	struct ifnet *ifp;
1787 	struct ifmultiaddr *ifma;
1788 	struct	rt_addrinfo info;
1789 	int	len, error = 0;
1790 	struct ifaddr *ifa;
1791 
1792 	bzero((caddr_t)&info, sizeof(info));
1793 	IFNET_RLOCK();
1794 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1795 		if (w->w_arg && w->w_arg != ifp->if_index)
1796 			continue;
1797 		ifa = ifp->if_addr;
1798 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1799 		IF_ADDR_RLOCK(ifp);
1800 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1801 			if (af && af != ifma->ifma_addr->sa_family)
1802 				continue;
1803 			if (prison_if(w->w_req->td->td_ucred,
1804 			    ifma->ifma_addr) != 0)
1805 				continue;
1806 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1807 			info.rti_info[RTAX_GATEWAY] =
1808 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1809 			    ifma->ifma_lladdr : NULL;
1810 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1811 			if (w->w_req && w->w_tmem) {
1812 				struct ifma_msghdr *ifmam;
1813 
1814 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1815 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1816 				ifmam->ifmam_flags = 0;
1817 				ifmam->ifmam_addrs = info.rti_addrs;
1818 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1819 				if (error) {
1820 					IF_ADDR_RUNLOCK(ifp);
1821 					goto done;
1822 				}
1823 			}
1824 		}
1825 		IF_ADDR_RUNLOCK(ifp);
1826 	}
1827 done:
1828 	IFNET_RUNLOCK();
1829 	return (error);
1830 }
1831 
1832 static int
1833 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1834 {
1835 	int	*name = (int *)arg1;
1836 	u_int	namelen = arg2;
1837 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1838 	int	i, lim, error = EINVAL;
1839 	u_char	af;
1840 	struct	walkarg w;
1841 
1842 	name ++;
1843 	namelen--;
1844 	if (req->newptr)
1845 		return (EPERM);
1846 	if (namelen != 3)
1847 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1848 	af = name[0];
1849 	if (af > AF_MAX)
1850 		return (EINVAL);
1851 	bzero(&w, sizeof(w));
1852 	w.w_op = name[1];
1853 	w.w_arg = name[2];
1854 	w.w_req = req;
1855 
1856 	error = sysctl_wire_old_buffer(req, 0);
1857 	if (error)
1858 		return (error);
1859 	switch (w.w_op) {
1860 
1861 	case NET_RT_DUMP:
1862 	case NET_RT_FLAGS:
1863 		if (af == 0) {			/* dump all tables */
1864 			i = 1;
1865 			lim = AF_MAX;
1866 		} else				/* dump only one table */
1867 			i = lim = af;
1868 
1869 		/*
1870 		 * take care of llinfo entries, the caller must
1871 		 * specify an AF
1872 		 */
1873 		if (w.w_op == NET_RT_FLAGS &&
1874 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1875 			if (af != 0)
1876 				error = lltable_sysctl_dumparp(af, w.w_req);
1877 			else
1878 				error = EINVAL;
1879 			break;
1880 		}
1881 		/*
1882 		 * take care of routing entries
1883 		 */
1884 		for (error = 0; error == 0 && i <= lim; i++) {
1885 			rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1886 			if (rnh != NULL) {
1887 				RADIX_NODE_HEAD_RLOCK(rnh);
1888 			    	error = rnh->rnh_walktree(rnh,
1889 				    sysctl_dumpentry, &w);
1890 				RADIX_NODE_HEAD_RUNLOCK(rnh);
1891 			} else if (af != 0)
1892 				error = EAFNOSUPPORT;
1893 		}
1894 		break;
1895 
1896 	case NET_RT_IFLIST:
1897 	case NET_RT_IFLISTL:
1898 		error = sysctl_iflist(af, &w);
1899 		break;
1900 
1901 	case NET_RT_IFMALIST:
1902 		error = sysctl_ifmalist(af, &w);
1903 		break;
1904 	}
1905 	if (w.w_tmem)
1906 		free(w.w_tmem, M_RTABLE);
1907 	return (error);
1908 }
1909 
1910 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1911 
1912 /*
1913  * Definitions of protocols supported in the ROUTE domain.
1914  */
1915 
1916 static struct domain routedomain;		/* or at least forward */
1917 
1918 static struct protosw routesw[] = {
1919 {
1920 	.pr_type =		SOCK_RAW,
1921 	.pr_domain =		&routedomain,
1922 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1923 	.pr_output =		route_output,
1924 	.pr_ctlinput =		raw_ctlinput,
1925 	.pr_init =		raw_init,
1926 	.pr_usrreqs =		&route_usrreqs
1927 }
1928 };
1929 
1930 static struct domain routedomain = {
1931 	.dom_family =		PF_ROUTE,
1932 	.dom_name =		 "route",
1933 	.dom_protosw =		routesw,
1934 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1935 };
1936 
1937 VNET_DOMAIN_SET(route);
1938