xref: /freebsd/sys/net/rtsock.c (revision 43d1e6ee299ad4e143d90d3ad374d1c24bd3306f)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1988, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
32  * $FreeBSD$
33  */
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/rmlock.h>
49 #include <sys/rwlock.h>
50 #include <sys/signalvar.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/sysctl.h>
54 #include <sys/systm.h>
55 
56 #include <net/if.h>
57 #include <net/if_var.h>
58 #include <net/if_dl.h>
59 #include <net/if_llatbl.h>
60 #include <net/if_types.h>
61 #include <net/netisr.h>
62 #include <net/raw_cb.h>
63 #include <net/route.h>
64 #include <net/route_var.h>
65 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/if_ether.h>
69 #include <netinet/ip_carp.h>
70 #ifdef INET6
71 #include <netinet6/ip6_var.h>
72 #include <netinet6/scope6_var.h>
73 #endif
74 
75 #ifdef COMPAT_FREEBSD32
76 #include <sys/mount.h>
77 #include <compat/freebsd32/freebsd32.h>
78 
79 struct if_msghdr32 {
80 	uint16_t ifm_msglen;
81 	uint8_t	ifm_version;
82 	uint8_t	ifm_type;
83 	int32_t	ifm_addrs;
84 	int32_t	ifm_flags;
85 	uint16_t ifm_index;
86 	uint16_t _ifm_spare1;
87 	struct	if_data ifm_data;
88 };
89 
90 struct if_msghdrl32 {
91 	uint16_t ifm_msglen;
92 	uint8_t	ifm_version;
93 	uint8_t	ifm_type;
94 	int32_t	ifm_addrs;
95 	int32_t	ifm_flags;
96 	uint16_t ifm_index;
97 	uint16_t _ifm_spare1;
98 	uint16_t ifm_len;
99 	uint16_t ifm_data_off;
100 	uint32_t _ifm_spare2;
101 	struct	if_data ifm_data;
102 };
103 
104 struct ifa_msghdrl32 {
105 	uint16_t ifam_msglen;
106 	uint8_t	ifam_version;
107 	uint8_t	ifam_type;
108 	int32_t	ifam_addrs;
109 	int32_t	ifam_flags;
110 	uint16_t ifam_index;
111 	uint16_t _ifam_spare1;
112 	uint16_t ifam_len;
113 	uint16_t ifam_data_off;
114 	int32_t	ifam_metric;
115 	struct	if_data ifam_data;
116 };
117 
118 #define SA_SIZE32(sa)						\
119     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
120 	sizeof(int)		:				\
121 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
122 
123 #endif /* COMPAT_FREEBSD32 */
124 
125 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
126 
127 /* NB: these are not modified */
128 static struct	sockaddr route_src = { 2, PF_ROUTE, };
129 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
130 
131 /* These are external hooks for CARP. */
132 int	(*carp_get_vhid_p)(struct ifaddr *);
133 
134 /*
135  * Used by rtsock/raw_input callback code to decide whether to filter the update
136  * notification to a socket bound to a particular FIB.
137  */
138 #define	RTS_FILTER_FIB	M_PROTO8
139 
140 typedef struct {
141 	int	ip_count;	/* attached w/ AF_INET */
142 	int	ip6_count;	/* attached w/ AF_INET6 */
143 	int	any_count;	/* total attached */
144 } route_cb_t;
145 VNET_DEFINE_STATIC(route_cb_t, route_cb);
146 #define	V_route_cb VNET(route_cb)
147 
148 struct mtx rtsock_mtx;
149 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
150 
151 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
152 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
153 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
154 
155 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
156 
157 struct walkarg {
158 	int	w_tmemsize;
159 	int	w_op, w_arg;
160 	caddr_t	w_tmem;
161 	struct sysctl_req *w_req;
162 };
163 
164 static void	rts_input(struct mbuf *m);
165 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
166 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
167 			struct walkarg *w, int *plen);
168 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
169 			struct rt_addrinfo *rtinfo);
170 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
171 static int	sysctl_iflist(int af, struct walkarg *w);
172 static int	sysctl_ifmalist(int af, struct walkarg *w);
173 static int	route_output(struct mbuf *m, struct socket *so, ...);
174 static void	rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
175 static void	rt_dispatch(struct mbuf *, sa_family_t);
176 static struct sockaddr	*rtsock_fix_netmask(struct sockaddr *dst,
177 			struct sockaddr *smask, struct sockaddr_storage *dmask);
178 
179 static struct netisr_handler rtsock_nh = {
180 	.nh_name = "rtsock",
181 	.nh_handler = rts_input,
182 	.nh_proto = NETISR_ROUTE,
183 	.nh_policy = NETISR_POLICY_SOURCE,
184 };
185 
186 static int
187 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
188 {
189 	int error, qlimit;
190 
191 	netisr_getqlimit(&rtsock_nh, &qlimit);
192 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
193         if (error || !req->newptr)
194                 return (error);
195 	if (qlimit < 1)
196 		return (EINVAL);
197 	return (netisr_setqlimit(&rtsock_nh, qlimit));
198 }
199 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
200     0, 0, sysctl_route_netisr_maxqlen, "I",
201     "maximum routing socket dispatch queue length");
202 
203 static void
204 vnet_rts_init(void)
205 {
206 	int tmp;
207 
208 	if (IS_DEFAULT_VNET(curvnet)) {
209 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
210 			rtsock_nh.nh_qlimit = tmp;
211 		netisr_register(&rtsock_nh);
212 	}
213 #ifdef VIMAGE
214 	 else
215 		netisr_register_vnet(&rtsock_nh);
216 #endif
217 }
218 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
219     vnet_rts_init, 0);
220 
221 #ifdef VIMAGE
222 static void
223 vnet_rts_uninit(void)
224 {
225 
226 	netisr_unregister_vnet(&rtsock_nh);
227 }
228 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
229     vnet_rts_uninit, 0);
230 #endif
231 
232 static int
233 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
234     struct rawcb *rp)
235 {
236 	int fibnum;
237 
238 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
239 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
240 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
241 
242 	/* No filtering requested. */
243 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
244 		return (0);
245 
246 	/* Check if it is a rts and the fib matches the one of the socket. */
247 	fibnum = M_GETFIB(m);
248 	if (proto->sp_family != PF_ROUTE ||
249 	    rp->rcb_socket == NULL ||
250 	    rp->rcb_socket->so_fibnum == fibnum)
251 		return (0);
252 
253 	/* Filtering requested and no match, the socket shall be skipped. */
254 	return (1);
255 }
256 
257 static void
258 rts_input(struct mbuf *m)
259 {
260 	struct sockproto route_proto;
261 	unsigned short *family;
262 	struct m_tag *tag;
263 
264 	route_proto.sp_family = PF_ROUTE;
265 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
266 	if (tag != NULL) {
267 		family = (unsigned short *)(tag + 1);
268 		route_proto.sp_protocol = *family;
269 		m_tag_delete(m, tag);
270 	} else
271 		route_proto.sp_protocol = 0;
272 
273 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
274 }
275 
276 /*
277  * It really doesn't make any sense at all for this code to share much
278  * with raw_usrreq.c, since its functionality is so restricted.  XXX
279  */
280 static void
281 rts_abort(struct socket *so)
282 {
283 
284 	raw_usrreqs.pru_abort(so);
285 }
286 
287 static void
288 rts_close(struct socket *so)
289 {
290 
291 	raw_usrreqs.pru_close(so);
292 }
293 
294 /* pru_accept is EOPNOTSUPP */
295 
296 static int
297 rts_attach(struct socket *so, int proto, struct thread *td)
298 {
299 	struct rawcb *rp;
300 	int error;
301 
302 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
303 
304 	/* XXX */
305 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
306 
307 	so->so_pcb = (caddr_t)rp;
308 	so->so_fibnum = td->td_proc->p_fibnum;
309 	error = raw_attach(so, proto);
310 	rp = sotorawcb(so);
311 	if (error) {
312 		so->so_pcb = NULL;
313 		free(rp, M_PCB);
314 		return error;
315 	}
316 	RTSOCK_LOCK();
317 	switch(rp->rcb_proto.sp_protocol) {
318 	case AF_INET:
319 		V_route_cb.ip_count++;
320 		break;
321 	case AF_INET6:
322 		V_route_cb.ip6_count++;
323 		break;
324 	}
325 	V_route_cb.any_count++;
326 	RTSOCK_UNLOCK();
327 	soisconnected(so);
328 	so->so_options |= SO_USELOOPBACK;
329 	return 0;
330 }
331 
332 static int
333 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
334 {
335 
336 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
337 }
338 
339 static int
340 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
341 {
342 
343 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
344 }
345 
346 /* pru_connect2 is EOPNOTSUPP */
347 /* pru_control is EOPNOTSUPP */
348 
349 static void
350 rts_detach(struct socket *so)
351 {
352 	struct rawcb *rp = sotorawcb(so);
353 
354 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
355 
356 	RTSOCK_LOCK();
357 	switch(rp->rcb_proto.sp_protocol) {
358 	case AF_INET:
359 		V_route_cb.ip_count--;
360 		break;
361 	case AF_INET6:
362 		V_route_cb.ip6_count--;
363 		break;
364 	}
365 	V_route_cb.any_count--;
366 	RTSOCK_UNLOCK();
367 	raw_usrreqs.pru_detach(so);
368 }
369 
370 static int
371 rts_disconnect(struct socket *so)
372 {
373 
374 	return (raw_usrreqs.pru_disconnect(so));
375 }
376 
377 /* pru_listen is EOPNOTSUPP */
378 
379 static int
380 rts_peeraddr(struct socket *so, struct sockaddr **nam)
381 {
382 
383 	return (raw_usrreqs.pru_peeraddr(so, nam));
384 }
385 
386 /* pru_rcvd is EOPNOTSUPP */
387 /* pru_rcvoob is EOPNOTSUPP */
388 
389 static int
390 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
391 	 struct mbuf *control, struct thread *td)
392 {
393 
394 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
395 }
396 
397 /* pru_sense is null */
398 
399 static int
400 rts_shutdown(struct socket *so)
401 {
402 
403 	return (raw_usrreqs.pru_shutdown(so));
404 }
405 
406 static int
407 rts_sockaddr(struct socket *so, struct sockaddr **nam)
408 {
409 
410 	return (raw_usrreqs.pru_sockaddr(so, nam));
411 }
412 
413 static struct pr_usrreqs route_usrreqs = {
414 	.pru_abort =		rts_abort,
415 	.pru_attach =		rts_attach,
416 	.pru_bind =		rts_bind,
417 	.pru_connect =		rts_connect,
418 	.pru_detach =		rts_detach,
419 	.pru_disconnect =	rts_disconnect,
420 	.pru_peeraddr =		rts_peeraddr,
421 	.pru_send =		rts_send,
422 	.pru_shutdown =		rts_shutdown,
423 	.pru_sockaddr =		rts_sockaddr,
424 	.pru_close =		rts_close,
425 };
426 
427 #ifndef _SOCKADDR_UNION_DEFINED
428 #define	_SOCKADDR_UNION_DEFINED
429 /*
430  * The union of all possible address formats we handle.
431  */
432 union sockaddr_union {
433 	struct sockaddr		sa;
434 	struct sockaddr_in	sin;
435 	struct sockaddr_in6	sin6;
436 };
437 #endif /* _SOCKADDR_UNION_DEFINED */
438 
439 static int
440 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
441     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
442 {
443 #if defined(INET) || defined(INET6)
444 	struct epoch_tracker et;
445 #endif
446 
447 	/* First, see if the returned address is part of the jail. */
448 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
449 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
450 		return (0);
451 	}
452 
453 	switch (info->rti_info[RTAX_DST]->sa_family) {
454 #ifdef INET
455 	case AF_INET:
456 	{
457 		struct in_addr ia;
458 		struct ifaddr *ifa;
459 		int found;
460 
461 		found = 0;
462 		/*
463 		 * Try to find an address on the given outgoing interface
464 		 * that belongs to the jail.
465 		 */
466 		NET_EPOCH_ENTER(et);
467 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
468 			struct sockaddr *sa;
469 			sa = ifa->ifa_addr;
470 			if (sa->sa_family != AF_INET)
471 				continue;
472 			ia = ((struct sockaddr_in *)sa)->sin_addr;
473 			if (prison_check_ip4(cred, &ia) == 0) {
474 				found = 1;
475 				break;
476 			}
477 		}
478 		NET_EPOCH_EXIT(et);
479 		if (!found) {
480 			/*
481 			 * As a last resort return the 'default' jail address.
482 			 */
483 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
484 			    sin_addr;
485 			if (prison_get_ip4(cred, &ia) != 0)
486 				return (ESRCH);
487 		}
488 		bzero(&saun->sin, sizeof(struct sockaddr_in));
489 		saun->sin.sin_len = sizeof(struct sockaddr_in);
490 		saun->sin.sin_family = AF_INET;
491 		saun->sin.sin_addr.s_addr = ia.s_addr;
492 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
493 		break;
494 	}
495 #endif
496 #ifdef INET6
497 	case AF_INET6:
498 	{
499 		struct in6_addr ia6;
500 		struct ifaddr *ifa;
501 		int found;
502 
503 		found = 0;
504 		/*
505 		 * Try to find an address on the given outgoing interface
506 		 * that belongs to the jail.
507 		 */
508 		NET_EPOCH_ENTER(et);
509 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
510 			struct sockaddr *sa;
511 			sa = ifa->ifa_addr;
512 			if (sa->sa_family != AF_INET6)
513 				continue;
514 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
515 			    &ia6, sizeof(struct in6_addr));
516 			if (prison_check_ip6(cred, &ia6) == 0) {
517 				found = 1;
518 				break;
519 			}
520 		}
521 		NET_EPOCH_EXIT(et);
522 		if (!found) {
523 			/*
524 			 * As a last resort return the 'default' jail address.
525 			 */
526 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
527 			    sin6_addr;
528 			if (prison_get_ip6(cred, &ia6) != 0)
529 				return (ESRCH);
530 		}
531 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
532 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
533 		saun->sin6.sin6_family = AF_INET6;
534 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
535 		if (sa6_recoverscope(&saun->sin6) != 0)
536 			return (ESRCH);
537 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
538 		break;
539 	}
540 #endif
541 	default:
542 		return (ESRCH);
543 	}
544 	return (0);
545 }
546 
547 /*ARGSUSED*/
548 static int
549 route_output(struct mbuf *m, struct socket *so, ...)
550 {
551 	RIB_RLOCK_TRACKER;
552 	struct rt_msghdr *rtm = NULL;
553 	struct rtentry *rt = NULL;
554 	struct rib_head *rnh;
555 	struct rt_addrinfo info;
556 	struct sockaddr_storage ss;
557 #ifdef INET6
558 	struct sockaddr_in6 *sin6;
559 	int i, rti_need_deembed = 0;
560 #endif
561 	int alloc_len = 0, len, error = 0, fibnum;
562 	struct ifnet *ifp = NULL;
563 	union sockaddr_union saun;
564 	sa_family_t saf = AF_UNSPEC;
565 	struct rawcb *rp = NULL;
566 	struct walkarg w;
567 
568 	fibnum = so->so_fibnum;
569 
570 #define senderr(e) { error = e; goto flush;}
571 	if (m == NULL || ((m->m_len < sizeof(long)) &&
572 		       (m = m_pullup(m, sizeof(long))) == NULL))
573 		return (ENOBUFS);
574 	if ((m->m_flags & M_PKTHDR) == 0)
575 		panic("route_output");
576 	len = m->m_pkthdr.len;
577 	if (len < sizeof(*rtm) ||
578 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
579 		senderr(EINVAL);
580 
581 	/*
582 	 * Most of current messages are in range 200-240 bytes,
583 	 * minimize possible re-allocation on reply using larger size
584 	 * buffer aligned on 1k boundaty.
585 	 */
586 	alloc_len = roundup2(len, 1024);
587 	if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
588 		senderr(ENOBUFS);
589 
590 	m_copydata(m, 0, len, (caddr_t)rtm);
591 	bzero(&info, sizeof(info));
592 	bzero(&w, sizeof(w));
593 
594 	if (rtm->rtm_version != RTM_VERSION) {
595 		/* Do not touch message since format is unknown */
596 		free(rtm, M_TEMP);
597 		rtm = NULL;
598 		senderr(EPROTONOSUPPORT);
599 	}
600 
601 	/*
602 	 * Starting from here, it is possible
603 	 * to alter original message and insert
604 	 * caller PID and error value.
605 	 */
606 
607 	rtm->rtm_pid = curproc->p_pid;
608 	info.rti_addrs = rtm->rtm_addrs;
609 
610 	info.rti_mflags = rtm->rtm_inits;
611 	info.rti_rmx = &rtm->rtm_rmx;
612 
613 	/*
614 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
615 	 * link-local address because rtrequest requires addresses with
616 	 * embedded scope id.
617 	 */
618 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
619 		senderr(EINVAL);
620 
621 	if (rtm->rtm_flags & RTF_RNH_LOCKED)
622 		senderr(EINVAL);
623 	info.rti_flags = rtm->rtm_flags;
624 	if (info.rti_info[RTAX_DST] == NULL ||
625 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
626 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
627 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
628 		senderr(EINVAL);
629 	saf = info.rti_info[RTAX_DST]->sa_family;
630 	/*
631 	 * Verify that the caller has the appropriate privilege; RTM_GET
632 	 * is the only operation the non-superuser is allowed.
633 	 */
634 	if (rtm->rtm_type != RTM_GET) {
635 		error = priv_check(curthread, PRIV_NET_ROUTE);
636 		if (error)
637 			senderr(error);
638 	}
639 
640 	/*
641 	 * The given gateway address may be an interface address.
642 	 * For example, issuing a "route change" command on a route
643 	 * entry that was created from a tunnel, and the gateway
644 	 * address given is the local end point. In this case the
645 	 * RTF_GATEWAY flag must be cleared or the destination will
646 	 * not be reachable even though there is no error message.
647 	 */
648 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
649 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
650 		struct rt_addrinfo ginfo;
651 		struct sockaddr *gdst;
652 
653 		bzero(&ginfo, sizeof(ginfo));
654 		bzero(&ss, sizeof(ss));
655 		ss.ss_len = sizeof(ss);
656 
657 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
658 		gdst = info.rti_info[RTAX_GATEWAY];
659 
660 		/*
661 		 * A host route through the loopback interface is
662 		 * installed for each interface adddress. In pre 8.0
663 		 * releases the interface address of a PPP link type
664 		 * is not reachable locally. This behavior is fixed as
665 		 * part of the new L2/L3 redesign and rewrite work. The
666 		 * signature of this interface address route is the
667 		 * AF_LINK sa_family type of the rt_gateway, and the
668 		 * rt_ifp has the IFF_LOOPBACK flag set.
669 		 */
670 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
671 			if (ss.ss_family == AF_LINK &&
672 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
673 				info.rti_flags &= ~RTF_GATEWAY;
674 				info.rti_flags |= RTF_GWFLAG_COMPAT;
675 			}
676 			rib_free_info(&ginfo);
677 		}
678 	}
679 
680 	switch (rtm->rtm_type) {
681 		struct rtentry *saved_nrt;
682 
683 	case RTM_ADD:
684 	case RTM_CHANGE:
685 		if (rtm->rtm_type == RTM_ADD) {
686 			if (info.rti_info[RTAX_GATEWAY] == NULL)
687 				senderr(EINVAL);
688 		}
689 		saved_nrt = NULL;
690 
691 		/* support for new ARP code */
692 		if (info.rti_info[RTAX_GATEWAY] != NULL &&
693 		    info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
694 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
695 			error = lla_rt_output(rtm, &info);
696 #ifdef INET6
697 			if (error == 0)
698 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
699 #endif
700 			break;
701 		}
702 		error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
703 		    fibnum);
704 		if (error == 0 && saved_nrt != NULL) {
705 #ifdef INET6
706 			rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
707 #endif
708 			RT_LOCK(saved_nrt);
709 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
710 			RT_REMREF(saved_nrt);
711 			RT_UNLOCK(saved_nrt);
712 		}
713 		break;
714 
715 	case RTM_DELETE:
716 		saved_nrt = NULL;
717 		/* support for new ARP code */
718 		if (info.rti_info[RTAX_GATEWAY] &&
719 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
720 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
721 			error = lla_rt_output(rtm, &info);
722 #ifdef INET6
723 			if (error == 0)
724 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
725 #endif
726 			break;
727 		}
728 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
729 		if (error == 0) {
730 			RT_LOCK(saved_nrt);
731 			rt = saved_nrt;
732 			goto report;
733 		}
734 #ifdef INET6
735 		/* rt_msg2() will not be used when RTM_DELETE fails. */
736 		rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
737 #endif
738 		break;
739 
740 	case RTM_GET:
741 		rnh = rt_tables_get_rnh(fibnum, saf);
742 		if (rnh == NULL)
743 			senderr(EAFNOSUPPORT);
744 
745 		RIB_RLOCK(rnh);
746 
747 		if (info.rti_info[RTAX_NETMASK] == NULL &&
748 		    rtm->rtm_type == RTM_GET) {
749 			/*
750 			 * Provide longest prefix match for
751 			 * address lookup (no mask).
752 			 * 'route -n get addr'
753 			 */
754 			rt = (struct rtentry *) rnh->rnh_matchaddr(
755 			    info.rti_info[RTAX_DST], &rnh->head);
756 		} else
757 			rt = (struct rtentry *) rnh->rnh_lookup(
758 			    info.rti_info[RTAX_DST],
759 			    info.rti_info[RTAX_NETMASK], &rnh->head);
760 
761 		if (rt == NULL) {
762 			RIB_RUNLOCK(rnh);
763 			senderr(ESRCH);
764 		}
765 #ifdef RADIX_MPATH
766 		/*
767 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
768 		 * we require users to specify a matching RTAX_GATEWAY.
769 		 *
770 		 * for RTM_GET, gate is optional even with multipath.
771 		 * if gate == NULL the first match is returned.
772 		 * (no need to call rt_mpath_matchgate if gate == NULL)
773 		 */
774 		if (rt_mpath_capable(rnh) &&
775 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
776 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
777 			if (!rt) {
778 				RIB_RUNLOCK(rnh);
779 				senderr(ESRCH);
780 			}
781 		}
782 #endif
783 		/*
784 		 * If performing proxied L2 entry insertion, and
785 		 * the actual PPP host entry is found, perform
786 		 * another search to retrieve the prefix route of
787 		 * the local end point of the PPP link.
788 		 */
789 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
790 			struct sockaddr laddr;
791 
792 			if (rt->rt_ifp != NULL &&
793 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
794 				struct epoch_tracker et;
795 				struct ifaddr *ifa;
796 
797 				NET_EPOCH_ENTER(et);
798 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1,
799 						RT_ALL_FIBS);
800 				if (ifa != NULL)
801 					rt_maskedcopy(ifa->ifa_addr,
802 						      &laddr,
803 						      ifa->ifa_netmask);
804 				NET_EPOCH_EXIT(et);
805 			} else
806 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
807 					      &laddr,
808 					      rt->rt_ifa->ifa_netmask);
809 			/*
810 			 * refactor rt and no lock operation necessary
811 			 */
812 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
813 			    &rnh->head);
814 			if (rt == NULL) {
815 				RIB_RUNLOCK(rnh);
816 				senderr(ESRCH);
817 			}
818 		}
819 		RT_LOCK(rt);
820 		RT_ADDREF(rt);
821 		RIB_RUNLOCK(rnh);
822 
823 report:
824 		RT_LOCK_ASSERT(rt);
825 		if ((rt->rt_flags & RTF_HOST) == 0
826 		    ? jailed_without_vnet(curthread->td_ucred)
827 		    : prison_if(curthread->td_ucred,
828 		    rt_key(rt)) != 0) {
829 			RT_UNLOCK(rt);
830 			senderr(ESRCH);
831 		}
832 		info.rti_info[RTAX_DST] = rt_key(rt);
833 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
834 		info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
835 		    rt_mask(rt), &ss);
836 		info.rti_info[RTAX_GENMASK] = 0;
837 		if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
838 			ifp = rt->rt_ifp;
839 			if (ifp) {
840 				info.rti_info[RTAX_IFP] =
841 				    ifp->if_addr->ifa_addr;
842 				error = rtm_get_jailed(&info, ifp, rt,
843 				    &saun, curthread->td_ucred);
844 				if (error != 0) {
845 					RT_UNLOCK(rt);
846 					senderr(error);
847 				}
848 				if (ifp->if_flags & IFF_POINTOPOINT)
849 					info.rti_info[RTAX_BRD] =
850 					    rt->rt_ifa->ifa_dstaddr;
851 				rtm->rtm_index = ifp->if_index;
852 			} else {
853 				info.rti_info[RTAX_IFP] = NULL;
854 				info.rti_info[RTAX_IFA] = NULL;
855 			}
856 		} else if ((ifp = rt->rt_ifp) != NULL) {
857 			rtm->rtm_index = ifp->if_index;
858 		}
859 
860 		/* Check if we need to realloc storage */
861 		rtsock_msg_buffer(rtm->rtm_type, &info, NULL, &len);
862 		if (len > alloc_len) {
863 			struct rt_msghdr *new_rtm;
864 			new_rtm = malloc(len, M_TEMP, M_NOWAIT);
865 			if (new_rtm == NULL) {
866 				RT_UNLOCK(rt);
867 				senderr(ENOBUFS);
868 			}
869 			bcopy(rtm, new_rtm, rtm->rtm_msglen);
870 			free(rtm, M_TEMP);
871 			rtm = new_rtm;
872 			alloc_len = len;
873 		}
874 
875 		w.w_tmem = (caddr_t)rtm;
876 		w.w_tmemsize = alloc_len;
877 		rtsock_msg_buffer(rtm->rtm_type, &info, &w, &len);
878 
879 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
880 			rtm->rtm_flags = RTF_GATEWAY |
881 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
882 		else
883 			rtm->rtm_flags = rt->rt_flags;
884 		rt_getmetrics(rt, &rtm->rtm_rmx);
885 		rtm->rtm_addrs = info.rti_addrs;
886 
887 		RT_UNLOCK(rt);
888 		break;
889 
890 	default:
891 		senderr(EOPNOTSUPP);
892 	}
893 
894 flush:
895 	if (rt != NULL)
896 		RTFREE(rt);
897 	/*
898 	 * Check to see if we don't want our own messages.
899 	 */
900 	if ((so->so_options & SO_USELOOPBACK) == 0) {
901 		if (V_route_cb.any_count <= 1) {
902 			if (rtm != NULL)
903 				free(rtm, M_TEMP);
904 			m_freem(m);
905 			return (error);
906 		}
907 		/* There is another listener, so construct message */
908 		rp = sotorawcb(so);
909 	}
910 
911 	if (rtm != NULL) {
912 #ifdef INET6
913 		if (rti_need_deembed) {
914 			/* sin6_scope_id is recovered before sending rtm. */
915 			sin6 = (struct sockaddr_in6 *)&ss;
916 			for (i = 0; i < RTAX_MAX; i++) {
917 				if (info.rti_info[i] == NULL)
918 					continue;
919 				if (info.rti_info[i]->sa_family != AF_INET6)
920 					continue;
921 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
922 				if (sa6_recoverscope(sin6) == 0)
923 					bcopy(sin6, info.rti_info[i],
924 						    sizeof(*sin6));
925 			}
926 		}
927 #endif
928 		if (error != 0)
929 			rtm->rtm_errno = error;
930 		else
931 			rtm->rtm_flags |= RTF_DONE;
932 
933 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
934 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
935 			m_freem(m);
936 			m = NULL;
937 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
938 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
939 
940 		free(rtm, M_TEMP);
941 	}
942 	if (m != NULL) {
943 		M_SETFIB(m, fibnum);
944 		m->m_flags |= RTS_FILTER_FIB;
945 		if (rp) {
946 			/*
947 			 * XXX insure we don't get a copy by
948 			 * invalidating our protocol
949 			 */
950 			unsigned short family = rp->rcb_proto.sp_family;
951 			rp->rcb_proto.sp_family = 0;
952 			rt_dispatch(m, saf);
953 			rp->rcb_proto.sp_family = family;
954 		} else
955 			rt_dispatch(m, saf);
956 	}
957 
958 	return (error);
959 }
960 
961 static void
962 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
963 {
964 
965 	bzero(out, sizeof(*out));
966 	out->rmx_mtu = rt->rt_mtu;
967 	out->rmx_weight = rt->rt_weight;
968 	out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
969 	/* Kernel -> userland timebase conversion. */
970 	out->rmx_expire = rt->rt_expire ?
971 	    rt->rt_expire - time_uptime + time_second : 0;
972 }
973 
974 /*
975  * Extract the addresses of the passed sockaddrs.
976  * Do a little sanity checking so as to avoid bad memory references.
977  * This data is derived straight from userland.
978  */
979 static int
980 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
981 {
982 	struct sockaddr *sa;
983 	int i;
984 
985 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
986 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
987 			continue;
988 		sa = (struct sockaddr *)cp;
989 		/*
990 		 * It won't fit.
991 		 */
992 		if (cp + sa->sa_len > cplim)
993 			return (EINVAL);
994 		/*
995 		 * there are no more.. quit now
996 		 * If there are more bits, they are in error.
997 		 * I've seen this. route(1) can evidently generate these.
998 		 * This causes kernel to core dump.
999 		 * for compatibility, If we see this, point to a safe address.
1000 		 */
1001 		if (sa->sa_len == 0) {
1002 			rtinfo->rti_info[i] = &sa_zero;
1003 			return (0); /* should be EINVAL but for compat */
1004 		}
1005 		/* accept it */
1006 #ifdef INET6
1007 		if (sa->sa_family == AF_INET6)
1008 			sa6_embedscope((struct sockaddr_in6 *)sa,
1009 			    V_ip6_use_defzone);
1010 #endif
1011 		rtinfo->rti_info[i] = sa;
1012 		cp += SA_SIZE(sa);
1013 	}
1014 	return (0);
1015 }
1016 
1017 /*
1018  * Fill in @dmask with valid netmask leaving original @smask
1019  * intact. Mostly used with radix netmasks.
1020  */
1021 static struct sockaddr *
1022 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
1023     struct sockaddr_storage *dmask)
1024 {
1025 	if (dst == NULL || smask == NULL)
1026 		return (NULL);
1027 
1028 	memset(dmask, 0, dst->sa_len);
1029 	memcpy(dmask, smask, smask->sa_len);
1030 	dmask->ss_len = dst->sa_len;
1031 	dmask->ss_family = dst->sa_family;
1032 
1033 	return ((struct sockaddr *)dmask);
1034 }
1035 
1036 /*
1037  * Writes information related to @rtinfo object to newly-allocated mbuf.
1038  * Assumes MCLBYTES is enough to construct any message.
1039  * Used for OS notifications of vaious events (if/ifa announces,etc)
1040  *
1041  * Returns allocated mbuf or NULL on failure.
1042  */
1043 static struct mbuf *
1044 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1045 {
1046 	struct rt_msghdr *rtm;
1047 	struct mbuf *m;
1048 	int i;
1049 	struct sockaddr *sa;
1050 #ifdef INET6
1051 	struct sockaddr_storage ss;
1052 	struct sockaddr_in6 *sin6;
1053 #endif
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 
1081 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1082 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1083 	if (len > MHLEN)
1084 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1085 	else
1086 		m = m_gethdr(M_NOWAIT, MT_DATA);
1087 	if (m == NULL)
1088 		return (m);
1089 
1090 	m->m_pkthdr.len = m->m_len = len;
1091 	rtm = mtod(m, struct rt_msghdr *);
1092 	bzero((caddr_t)rtm, len);
1093 	for (i = 0; i < RTAX_MAX; i++) {
1094 		if ((sa = rtinfo->rti_info[i]) == NULL)
1095 			continue;
1096 		rtinfo->rti_addrs |= (1 << i);
1097 		dlen = SA_SIZE(sa);
1098 #ifdef INET6
1099 		if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1100 			sin6 = (struct sockaddr_in6 *)&ss;
1101 			bcopy(sa, sin6, sizeof(*sin6));
1102 			if (sa6_recoverscope(sin6) == 0)
1103 				sa = (struct sockaddr *)sin6;
1104 		}
1105 #endif
1106 		m_copyback(m, len, dlen, (caddr_t)sa);
1107 		len += dlen;
1108 	}
1109 	if (m->m_pkthdr.len != len) {
1110 		m_freem(m);
1111 		return (NULL);
1112 	}
1113 	rtm->rtm_msglen = len;
1114 	rtm->rtm_version = RTM_VERSION;
1115 	rtm->rtm_type = type;
1116 	return (m);
1117 }
1118 
1119 /*
1120  * Writes information related to @rtinfo object to preallocated buffer.
1121  * Stores needed size in @plen. If @w is NULL, calculates size without
1122  * writing.
1123  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1124  *
1125  * Returns 0 on success.
1126  *
1127  */
1128 static int
1129 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1130 {
1131 	int i;
1132 	int len, buflen = 0, dlen;
1133 	caddr_t cp = NULL;
1134 	struct rt_msghdr *rtm = NULL;
1135 #ifdef INET6
1136 	struct sockaddr_storage ss;
1137 	struct sockaddr_in6 *sin6;
1138 #endif
1139 #ifdef COMPAT_FREEBSD32
1140 	bool compat32 = false;
1141 #endif
1142 
1143 	switch (type) {
1144 
1145 	case RTM_DELADDR:
1146 	case RTM_NEWADDR:
1147 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1148 #ifdef COMPAT_FREEBSD32
1149 			if (w->w_req->flags & SCTL_MASK32) {
1150 				len = sizeof(struct ifa_msghdrl32);
1151 				compat32 = true;
1152 			} else
1153 #endif
1154 				len = sizeof(struct ifa_msghdrl);
1155 		} else
1156 			len = sizeof(struct ifa_msghdr);
1157 		break;
1158 
1159 	case RTM_IFINFO:
1160 #ifdef COMPAT_FREEBSD32
1161 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1162 			if (w->w_op == NET_RT_IFLISTL)
1163 				len = sizeof(struct if_msghdrl32);
1164 			else
1165 				len = sizeof(struct if_msghdr32);
1166 			compat32 = true;
1167 			break;
1168 		}
1169 #endif
1170 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1171 			len = sizeof(struct if_msghdrl);
1172 		else
1173 			len = sizeof(struct if_msghdr);
1174 		break;
1175 
1176 	case RTM_NEWMADDR:
1177 		len = sizeof(struct ifma_msghdr);
1178 		break;
1179 
1180 	default:
1181 		len = sizeof(struct rt_msghdr);
1182 	}
1183 
1184 	if (w != NULL) {
1185 		rtm = (struct rt_msghdr *)w->w_tmem;
1186 		buflen = w->w_tmemsize - len;
1187 		cp = (caddr_t)w->w_tmem + len;
1188 	}
1189 
1190 	rtinfo->rti_addrs = 0;
1191 	for (i = 0; i < RTAX_MAX; i++) {
1192 		struct sockaddr *sa;
1193 
1194 		if ((sa = rtinfo->rti_info[i]) == NULL)
1195 			continue;
1196 		rtinfo->rti_addrs |= (1 << i);
1197 #ifdef COMPAT_FREEBSD32
1198 		if (compat32)
1199 			dlen = SA_SIZE32(sa);
1200 		else
1201 #endif
1202 			dlen = SA_SIZE(sa);
1203 		if (cp != NULL && buflen >= dlen) {
1204 #ifdef INET6
1205 			if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1206 				sin6 = (struct sockaddr_in6 *)&ss;
1207 				bcopy(sa, sin6, sizeof(*sin6));
1208 				if (sa6_recoverscope(sin6) == 0)
1209 					sa = (struct sockaddr *)sin6;
1210 			}
1211 #endif
1212 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1213 			cp += dlen;
1214 			buflen -= dlen;
1215 		} else if (cp != NULL) {
1216 			/*
1217 			 * Buffer too small. Count needed size
1218 			 * and return with error.
1219 			 */
1220 			cp = NULL;
1221 		}
1222 
1223 		len += dlen;
1224 	}
1225 
1226 	if (cp != NULL) {
1227 		dlen = ALIGN(len) - len;
1228 		if (buflen < dlen)
1229 			cp = NULL;
1230 		else {
1231 			bzero(cp, dlen);
1232 			cp += dlen;
1233 			buflen -= dlen;
1234 		}
1235 	}
1236 	len = ALIGN(len);
1237 
1238 	if (cp != NULL) {
1239 		/* fill header iff buffer is large enough */
1240 		rtm->rtm_version = RTM_VERSION;
1241 		rtm->rtm_type = type;
1242 		rtm->rtm_msglen = len;
1243 	}
1244 
1245 	*plen = len;
1246 
1247 	if (w != NULL && cp == NULL)
1248 		return (ENOBUFS);
1249 
1250 	return (0);
1251 }
1252 
1253 /*
1254  * This routine is called to generate a message from the routing
1255  * socket indicating that a redirect has occurred, a routing lookup
1256  * has failed, or that a protocol has detected timeouts to a particular
1257  * destination.
1258  */
1259 void
1260 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1261     int fibnum)
1262 {
1263 	struct rt_msghdr *rtm;
1264 	struct mbuf *m;
1265 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1266 
1267 	if (V_route_cb.any_count == 0)
1268 		return;
1269 	m = rtsock_msg_mbuf(type, rtinfo);
1270 	if (m == NULL)
1271 		return;
1272 
1273 	if (fibnum != RT_ALL_FIBS) {
1274 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1275 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1276 		M_SETFIB(m, fibnum);
1277 		m->m_flags |= RTS_FILTER_FIB;
1278 	}
1279 
1280 	rtm = mtod(m, struct rt_msghdr *);
1281 	rtm->rtm_flags = RTF_DONE | flags;
1282 	rtm->rtm_errno = error;
1283 	rtm->rtm_addrs = rtinfo->rti_addrs;
1284 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1285 }
1286 
1287 void
1288 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1289 {
1290 
1291 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1292 }
1293 
1294 /*
1295  * This routine is called to generate a message from the routing
1296  * socket indicating that the status of a network interface has changed.
1297  */
1298 void
1299 rt_ifmsg(struct ifnet *ifp)
1300 {
1301 	struct if_msghdr *ifm;
1302 	struct mbuf *m;
1303 	struct rt_addrinfo info;
1304 
1305 	if (V_route_cb.any_count == 0)
1306 		return;
1307 	bzero((caddr_t)&info, sizeof(info));
1308 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1309 	if (m == NULL)
1310 		return;
1311 	ifm = mtod(m, struct if_msghdr *);
1312 	ifm->ifm_index = ifp->if_index;
1313 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1314 	if_data_copy(ifp, &ifm->ifm_data);
1315 	ifm->ifm_addrs = 0;
1316 	rt_dispatch(m, AF_UNSPEC);
1317 }
1318 
1319 /*
1320  * Announce interface address arrival/withdraw.
1321  * Please do not call directly, use rt_addrmsg().
1322  * Assume input data to be valid.
1323  * Returns 0 on success.
1324  */
1325 int
1326 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1327 {
1328 	struct rt_addrinfo info;
1329 	struct sockaddr *sa;
1330 	int ncmd;
1331 	struct mbuf *m;
1332 	struct ifa_msghdr *ifam;
1333 	struct ifnet *ifp = ifa->ifa_ifp;
1334 	struct sockaddr_storage ss;
1335 
1336 	if (V_route_cb.any_count == 0)
1337 		return (0);
1338 
1339 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1340 
1341 	bzero((caddr_t)&info, sizeof(info));
1342 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1343 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1344 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1345 	    info.rti_info[RTAX_IFP], ifa->ifa_netmask, &ss);
1346 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1347 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1348 		return (ENOBUFS);
1349 	ifam = mtod(m, struct ifa_msghdr *);
1350 	ifam->ifam_index = ifp->if_index;
1351 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1352 	ifam->ifam_flags = ifa->ifa_flags;
1353 	ifam->ifam_addrs = info.rti_addrs;
1354 
1355 	if (fibnum != RT_ALL_FIBS) {
1356 		M_SETFIB(m, fibnum);
1357 		m->m_flags |= RTS_FILTER_FIB;
1358 	}
1359 
1360 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1361 
1362 	return (0);
1363 }
1364 
1365 /*
1366  * Announce route addition/removal.
1367  * Please do not call directly, use rt_routemsg().
1368  * Note that @rt data MAY be inconsistent/invalid:
1369  * if some userland app sends us "invalid" route message (invalid mask,
1370  * no dst, wrong address families, etc...) we need to pass it back
1371  * to app (and any other rtsock consumers) with rtm_errno field set to
1372  * non-zero value.
1373  *
1374  * Returns 0 on success.
1375  */
1376 int
1377 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1378     int fibnum)
1379 {
1380 	struct rt_addrinfo info;
1381 	struct sockaddr *sa;
1382 	struct mbuf *m;
1383 	struct rt_msghdr *rtm;
1384 	struct sockaddr_storage ss;
1385 
1386 	if (V_route_cb.any_count == 0)
1387 		return (0);
1388 
1389 	bzero((caddr_t)&info, sizeof(info));
1390 	info.rti_info[RTAX_DST] = sa = rt_key(rt);
1391 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(sa, rt_mask(rt), &ss);
1392 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1393 	if ((m = rtsock_msg_mbuf(cmd, &info)) == NULL)
1394 		return (ENOBUFS);
1395 	rtm = mtod(m, struct rt_msghdr *);
1396 	rtm->rtm_index = ifp->if_index;
1397 	rtm->rtm_flags |= rt->rt_flags;
1398 	rtm->rtm_errno = error;
1399 	rtm->rtm_addrs = info.rti_addrs;
1400 
1401 	if (fibnum != RT_ALL_FIBS) {
1402 		M_SETFIB(m, fibnum);
1403 		m->m_flags |= RTS_FILTER_FIB;
1404 	}
1405 
1406 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1407 
1408 	return (0);
1409 }
1410 
1411 /*
1412  * This is the analogue to the rt_newaddrmsg which performs the same
1413  * function but for multicast group memberhips.  This is easier since
1414  * there is no route state to worry about.
1415  */
1416 void
1417 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1418 {
1419 	struct rt_addrinfo info;
1420 	struct mbuf *m = NULL;
1421 	struct ifnet *ifp = ifma->ifma_ifp;
1422 	struct ifma_msghdr *ifmam;
1423 
1424 	if (V_route_cb.any_count == 0)
1425 		return;
1426 
1427 	bzero((caddr_t)&info, sizeof(info));
1428 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1429 	if (ifp && ifp->if_addr)
1430 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1431 	else
1432 		info.rti_info[RTAX_IFP] = NULL;
1433 	/*
1434 	 * If a link-layer address is present, present it as a ``gateway''
1435 	 * (similarly to how ARP entries, e.g., are presented).
1436 	 */
1437 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1438 	m = rtsock_msg_mbuf(cmd, &info);
1439 	if (m == NULL)
1440 		return;
1441 	ifmam = mtod(m, struct ifma_msghdr *);
1442 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1443 	    __func__));
1444 	ifmam->ifmam_index = ifp->if_index;
1445 	ifmam->ifmam_addrs = info.rti_addrs;
1446 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1447 }
1448 
1449 static struct mbuf *
1450 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1451 	struct rt_addrinfo *info)
1452 {
1453 	struct if_announcemsghdr *ifan;
1454 	struct mbuf *m;
1455 
1456 	if (V_route_cb.any_count == 0)
1457 		return NULL;
1458 	bzero((caddr_t)info, sizeof(*info));
1459 	m = rtsock_msg_mbuf(type, info);
1460 	if (m != NULL) {
1461 		ifan = mtod(m, struct if_announcemsghdr *);
1462 		ifan->ifan_index = ifp->if_index;
1463 		strlcpy(ifan->ifan_name, ifp->if_xname,
1464 			sizeof(ifan->ifan_name));
1465 		ifan->ifan_what = what;
1466 	}
1467 	return m;
1468 }
1469 
1470 /*
1471  * This is called to generate routing socket messages indicating
1472  * IEEE80211 wireless events.
1473  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1474  */
1475 void
1476 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1477 {
1478 	struct mbuf *m;
1479 	struct rt_addrinfo info;
1480 
1481 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1482 	if (m != NULL) {
1483 		/*
1484 		 * Append the ieee80211 data.  Try to stick it in the
1485 		 * mbuf containing the ifannounce msg; otherwise allocate
1486 		 * a new mbuf and append.
1487 		 *
1488 		 * NB: we assume m is a single mbuf.
1489 		 */
1490 		if (data_len > M_TRAILINGSPACE(m)) {
1491 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1492 			if (n == NULL) {
1493 				m_freem(m);
1494 				return;
1495 			}
1496 			bcopy(data, mtod(n, void *), data_len);
1497 			n->m_len = data_len;
1498 			m->m_next = n;
1499 		} else if (data_len > 0) {
1500 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1501 			m->m_len += data_len;
1502 		}
1503 		if (m->m_flags & M_PKTHDR)
1504 			m->m_pkthdr.len += data_len;
1505 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1506 		rt_dispatch(m, AF_UNSPEC);
1507 	}
1508 }
1509 
1510 /*
1511  * This is called to generate routing socket messages indicating
1512  * network interface arrival and departure.
1513  */
1514 void
1515 rt_ifannouncemsg(struct ifnet *ifp, int what)
1516 {
1517 	struct mbuf *m;
1518 	struct rt_addrinfo info;
1519 
1520 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1521 	if (m != NULL)
1522 		rt_dispatch(m, AF_UNSPEC);
1523 }
1524 
1525 static void
1526 rt_dispatch(struct mbuf *m, sa_family_t saf)
1527 {
1528 	struct m_tag *tag;
1529 
1530 	/*
1531 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1532 	 * use when injecting the mbuf into the routing socket buffer from
1533 	 * the netisr.
1534 	 */
1535 	if (saf != AF_UNSPEC) {
1536 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1537 		    M_NOWAIT);
1538 		if (tag == NULL) {
1539 			m_freem(m);
1540 			return;
1541 		}
1542 		*(unsigned short *)(tag + 1) = saf;
1543 		m_tag_prepend(m, tag);
1544 	}
1545 #ifdef VIMAGE
1546 	if (V_loif)
1547 		m->m_pkthdr.rcvif = V_loif;
1548 	else {
1549 		m_freem(m);
1550 		return;
1551 	}
1552 #endif
1553 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1554 }
1555 
1556 /*
1557  * This is used in dumping the kernel table via sysctl().
1558  */
1559 static int
1560 sysctl_dumpentry(struct radix_node *rn, void *vw)
1561 {
1562 	struct walkarg *w = vw;
1563 	struct rtentry *rt = (struct rtentry *)rn;
1564 	int error = 0, size;
1565 	struct rt_addrinfo info;
1566 	struct sockaddr_storage ss;
1567 
1568 	NET_EPOCH_ASSERT();
1569 
1570 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1571 		return 0;
1572 	if ((rt->rt_flags & RTF_HOST) == 0
1573 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1574 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1575 		return (0);
1576 	bzero((caddr_t)&info, sizeof(info));
1577 	info.rti_info[RTAX_DST] = rt_key(rt);
1578 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1579 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1580 	    rt_mask(rt), &ss);
1581 	info.rti_info[RTAX_GENMASK] = 0;
1582 	if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) {
1583 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1584 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1585 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1586 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1587 	}
1588 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1589 		return (error);
1590 	if (w->w_req && w->w_tmem) {
1591 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1592 
1593 		bzero(&rtm->rtm_index,
1594 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1595 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1596 			rtm->rtm_flags = RTF_GATEWAY |
1597 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1598 		else
1599 			rtm->rtm_flags = rt->rt_flags;
1600 		rt_getmetrics(rt, &rtm->rtm_rmx);
1601 		rtm->rtm_index = rt->rt_ifp->if_index;
1602 		rtm->rtm_addrs = info.rti_addrs;
1603 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1604 		return (error);
1605 	}
1606 	return (error);
1607 }
1608 
1609 static int
1610 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1611     struct rt_addrinfo *info, struct walkarg *w, int len)
1612 {
1613 	struct if_msghdrl *ifm;
1614 	struct if_data *ifd;
1615 
1616 	ifm = (struct if_msghdrl *)w->w_tmem;
1617 
1618 #ifdef COMPAT_FREEBSD32
1619 	if (w->w_req->flags & SCTL_MASK32) {
1620 		struct if_msghdrl32 *ifm32;
1621 
1622 		ifm32 = (struct if_msghdrl32 *)ifm;
1623 		ifm32->ifm_addrs = info->rti_addrs;
1624 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1625 		ifm32->ifm_index = ifp->if_index;
1626 		ifm32->_ifm_spare1 = 0;
1627 		ifm32->ifm_len = sizeof(*ifm32);
1628 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1629 		ifm32->_ifm_spare2 = 0;
1630 		ifd = &ifm32->ifm_data;
1631 	} else
1632 #endif
1633 	{
1634 		ifm->ifm_addrs = info->rti_addrs;
1635 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1636 		ifm->ifm_index = ifp->if_index;
1637 		ifm->_ifm_spare1 = 0;
1638 		ifm->ifm_len = sizeof(*ifm);
1639 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1640 		ifm->_ifm_spare2 = 0;
1641 		ifd = &ifm->ifm_data;
1642 	}
1643 
1644 	memcpy(ifd, src_ifd, sizeof(*ifd));
1645 
1646 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1647 }
1648 
1649 static int
1650 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1651     struct rt_addrinfo *info, struct walkarg *w, int len)
1652 {
1653 	struct if_msghdr *ifm;
1654 	struct if_data *ifd;
1655 
1656 	ifm = (struct if_msghdr *)w->w_tmem;
1657 
1658 #ifdef COMPAT_FREEBSD32
1659 	if (w->w_req->flags & SCTL_MASK32) {
1660 		struct if_msghdr32 *ifm32;
1661 
1662 		ifm32 = (struct if_msghdr32 *)ifm;
1663 		ifm32->ifm_addrs = info->rti_addrs;
1664 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1665 		ifm32->ifm_index = ifp->if_index;
1666 		ifm32->_ifm_spare1 = 0;
1667 		ifd = &ifm32->ifm_data;
1668 	} else
1669 #endif
1670 	{
1671 		ifm->ifm_addrs = info->rti_addrs;
1672 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1673 		ifm->ifm_index = ifp->if_index;
1674 		ifm->_ifm_spare1 = 0;
1675 		ifd = &ifm->ifm_data;
1676 	}
1677 
1678 	memcpy(ifd, src_ifd, sizeof(*ifd));
1679 
1680 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1681 }
1682 
1683 static int
1684 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1685     struct walkarg *w, int len)
1686 {
1687 	struct ifa_msghdrl *ifam;
1688 	struct if_data *ifd;
1689 
1690 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1691 
1692 #ifdef COMPAT_FREEBSD32
1693 	if (w->w_req->flags & SCTL_MASK32) {
1694 		struct ifa_msghdrl32 *ifam32;
1695 
1696 		ifam32 = (struct ifa_msghdrl32 *)ifam;
1697 		ifam32->ifam_addrs = info->rti_addrs;
1698 		ifam32->ifam_flags = ifa->ifa_flags;
1699 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1700 		ifam32->_ifam_spare1 = 0;
1701 		ifam32->ifam_len = sizeof(*ifam32);
1702 		ifam32->ifam_data_off =
1703 		    offsetof(struct ifa_msghdrl32, ifam_data);
1704 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1705 		ifd = &ifam32->ifam_data;
1706 	} else
1707 #endif
1708 	{
1709 		ifam->ifam_addrs = info->rti_addrs;
1710 		ifam->ifam_flags = ifa->ifa_flags;
1711 		ifam->ifam_index = ifa->ifa_ifp->if_index;
1712 		ifam->_ifam_spare1 = 0;
1713 		ifam->ifam_len = sizeof(*ifam);
1714 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1715 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1716 		ifd = &ifam->ifam_data;
1717 	}
1718 
1719 	bzero(ifd, sizeof(*ifd));
1720 	ifd->ifi_datalen = sizeof(struct if_data);
1721 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1722 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1723 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1724 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1725 
1726 	/* Fixup if_data carp(4) vhid. */
1727 	if (carp_get_vhid_p != NULL)
1728 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1729 
1730 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1731 }
1732 
1733 static int
1734 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1735     struct walkarg *w, int len)
1736 {
1737 	struct ifa_msghdr *ifam;
1738 
1739 	ifam = (struct ifa_msghdr *)w->w_tmem;
1740 	ifam->ifam_addrs = info->rti_addrs;
1741 	ifam->ifam_flags = ifa->ifa_flags;
1742 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1743 	ifam->_ifam_spare1 = 0;
1744 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1745 
1746 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1747 }
1748 
1749 static int
1750 sysctl_iflist(int af, struct walkarg *w)
1751 {
1752 	struct ifnet *ifp;
1753 	struct ifaddr *ifa;
1754 	struct if_data ifd;
1755 	struct rt_addrinfo info;
1756 	int len, error = 0;
1757 	struct sockaddr_storage ss;
1758 	struct epoch_tracker et;
1759 
1760 	bzero((caddr_t)&info, sizeof(info));
1761 	bzero(&ifd, sizeof(ifd));
1762 	NET_EPOCH_ENTER(et);
1763 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1764 		if (w->w_arg && w->w_arg != ifp->if_index)
1765 			continue;
1766 		if_data_copy(ifp, &ifd);
1767 		ifa = ifp->if_addr;
1768 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1769 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1770 		if (error != 0)
1771 			goto done;
1772 		info.rti_info[RTAX_IFP] = NULL;
1773 		if (w->w_req && w->w_tmem) {
1774 			if (w->w_op == NET_RT_IFLISTL)
1775 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1776 				    len);
1777 			else
1778 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1779 				    len);
1780 			if (error)
1781 				goto done;
1782 		}
1783 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1784 			if (af && af != ifa->ifa_addr->sa_family)
1785 				continue;
1786 			if (prison_if(w->w_req->td->td_ucred,
1787 			    ifa->ifa_addr) != 0)
1788 				continue;
1789 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1790 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1791 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
1792 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1793 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1794 			if (error != 0)
1795 				goto done;
1796 			if (w->w_req && w->w_tmem) {
1797 				if (w->w_op == NET_RT_IFLISTL)
1798 					error = sysctl_iflist_ifaml(ifa, &info,
1799 					    w, len);
1800 				else
1801 					error = sysctl_iflist_ifam(ifa, &info,
1802 					    w, len);
1803 				if (error)
1804 					goto done;
1805 			}
1806 		}
1807 		info.rti_info[RTAX_IFA] = NULL;
1808 		info.rti_info[RTAX_NETMASK] = NULL;
1809 		info.rti_info[RTAX_BRD] = NULL;
1810 	}
1811 done:
1812 	NET_EPOCH_EXIT(et);
1813 	return (error);
1814 }
1815 
1816 static int
1817 sysctl_ifmalist(int af, struct walkarg *w)
1818 {
1819 	struct rt_addrinfo info;
1820 	struct epoch_tracker et;
1821 	struct ifaddr *ifa;
1822 	struct ifmultiaddr *ifma;
1823 	struct ifnet *ifp;
1824 	int error, len;
1825 
1826 	error = 0;
1827 	bzero((caddr_t)&info, sizeof(info));
1828 
1829 	NET_EPOCH_ENTER(et);
1830 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1831 		if (w->w_arg && w->w_arg != ifp->if_index)
1832 			continue;
1833 		ifa = ifp->if_addr;
1834 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1835 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1836 			if (af && af != ifma->ifma_addr->sa_family)
1837 				continue;
1838 			if (prison_if(w->w_req->td->td_ucred,
1839 			    ifma->ifma_addr) != 0)
1840 				continue;
1841 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1842 			info.rti_info[RTAX_GATEWAY] =
1843 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1844 			    ifma->ifma_lladdr : NULL;
1845 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1846 			if (error != 0)
1847 				break;
1848 			if (w->w_req && w->w_tmem) {
1849 				struct ifma_msghdr *ifmam;
1850 
1851 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1852 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1853 				ifmam->ifmam_flags = 0;
1854 				ifmam->ifmam_addrs = info.rti_addrs;
1855 				ifmam->_ifmam_spare1 = 0;
1856 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1857 				if (error != 0)
1858 					break;
1859 			}
1860 		}
1861 		if (error != 0)
1862 			break;
1863 	}
1864 	NET_EPOCH_EXIT(et);
1865 	return (error);
1866 }
1867 
1868 static int
1869 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1870 {
1871 	RIB_RLOCK_TRACKER;
1872 	int	*name = (int *)arg1;
1873 	u_int	namelen = arg2;
1874 	struct rib_head *rnh = NULL; /* silence compiler. */
1875 	int	i, lim, error = EINVAL;
1876 	int	fib = 0;
1877 	u_char	af;
1878 	struct	walkarg w;
1879 
1880 	name ++;
1881 	namelen--;
1882 	if (req->newptr)
1883 		return (EPERM);
1884 	if (name[1] == NET_RT_DUMP) {
1885 		if (namelen == 3)
1886 			fib = req->td->td_proc->p_fibnum;
1887 		else if (namelen == 4)
1888 			fib = (name[3] == RT_ALL_FIBS) ?
1889 			    req->td->td_proc->p_fibnum : name[3];
1890 		else
1891 			return ((namelen < 3) ? EISDIR : ENOTDIR);
1892 		if (fib < 0 || fib >= rt_numfibs)
1893 			return (EINVAL);
1894 	} else if (namelen != 3)
1895 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1896 	af = name[0];
1897 	if (af > AF_MAX)
1898 		return (EINVAL);
1899 	bzero(&w, sizeof(w));
1900 	w.w_op = name[1];
1901 	w.w_arg = name[2];
1902 	w.w_req = req;
1903 
1904 	error = sysctl_wire_old_buffer(req, 0);
1905 	if (error)
1906 		return (error);
1907 
1908 	/*
1909 	 * Allocate reply buffer in advance.
1910 	 * All rtsock messages has maximum length of u_short.
1911 	 */
1912 	w.w_tmemsize = 65536;
1913 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
1914 
1915 	switch (w.w_op) {
1916 
1917 	case NET_RT_DUMP:
1918 	case NET_RT_FLAGS:
1919 		if (af == 0) {			/* dump all tables */
1920 			i = 1;
1921 			lim = AF_MAX;
1922 		} else				/* dump only one table */
1923 			i = lim = af;
1924 
1925 		/*
1926 		 * take care of llinfo entries, the caller must
1927 		 * specify an AF
1928 		 */
1929 		if (w.w_op == NET_RT_FLAGS &&
1930 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1931 			if (af != 0)
1932 				error = lltable_sysctl_dumparp(af, w.w_req);
1933 			else
1934 				error = EINVAL;
1935 			break;
1936 		}
1937 		/*
1938 		 * take care of routing entries
1939 		 */
1940 		for (error = 0; error == 0 && i <= lim; i++) {
1941 			rnh = rt_tables_get_rnh(fib, i);
1942 			if (rnh != NULL) {
1943 				struct epoch_tracker et;
1944 
1945 				RIB_RLOCK(rnh);
1946 				NET_EPOCH_ENTER(et);
1947 			    	error = rnh->rnh_walktree(&rnh->head,
1948 				    sysctl_dumpentry, &w);
1949 				NET_EPOCH_EXIT(et);
1950 				RIB_RUNLOCK(rnh);
1951 			} else if (af != 0)
1952 				error = EAFNOSUPPORT;
1953 		}
1954 		break;
1955 
1956 	case NET_RT_IFLIST:
1957 	case NET_RT_IFLISTL:
1958 		error = sysctl_iflist(af, &w);
1959 		break;
1960 
1961 	case NET_RT_IFMALIST:
1962 		error = sysctl_ifmalist(af, &w);
1963 		break;
1964 	}
1965 
1966 	free(w.w_tmem, M_TEMP);
1967 	return (error);
1968 }
1969 
1970 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1971 
1972 /*
1973  * Definitions of protocols supported in the ROUTE domain.
1974  */
1975 
1976 static struct domain routedomain;		/* or at least forward */
1977 
1978 static struct protosw routesw[] = {
1979 {
1980 	.pr_type =		SOCK_RAW,
1981 	.pr_domain =		&routedomain,
1982 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1983 	.pr_output =		route_output,
1984 	.pr_ctlinput =		raw_ctlinput,
1985 	.pr_init =		raw_init,
1986 	.pr_usrreqs =		&route_usrreqs
1987 }
1988 };
1989 
1990 static struct domain routedomain = {
1991 	.dom_family =		PF_ROUTE,
1992 	.dom_name =		 "route",
1993 	.dom_protosw =		routesw,
1994 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
1995 };
1996 
1997 VNET_DOMAIN_SET(route);
1998