xref: /freebsd/sys/net/rtsock.c (revision bdafb02fcb88389fd1ab684cfe734cb429d35618)
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 	struct	if_data ifm_data;
87 };
88 
89 struct if_msghdrl32 {
90 	uint16_t ifm_msglen;
91 	uint8_t	ifm_version;
92 	uint8_t	ifm_type;
93 	int32_t	ifm_addrs;
94 	int32_t	ifm_flags;
95 	uint16_t ifm_index;
96 	uint16_t _ifm_spare1;
97 	uint16_t ifm_len;
98 	uint16_t ifm_data_off;
99 	struct	if_data ifm_data;
100 };
101 
102 struct ifa_msghdrl32 {
103 	uint16_t ifam_msglen;
104 	uint8_t	ifam_version;
105 	uint8_t	ifam_type;
106 	int32_t	ifam_addrs;
107 	int32_t	ifam_flags;
108 	uint16_t ifam_index;
109 	uint16_t _ifam_spare1;
110 	uint16_t ifam_len;
111 	uint16_t ifam_data_off;
112 	int32_t	ifam_metric;
113 	struct	if_data ifam_data;
114 };
115 
116 #define SA_SIZE32(sa)						\
117     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
118 	sizeof(int)		:				\
119 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
120 
121 #endif /* COMPAT_FREEBSD32 */
122 
123 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
124 
125 /* NB: these are not modified */
126 static struct	sockaddr route_src = { 2, PF_ROUTE, };
127 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
128 
129 /* These are external hooks for CARP. */
130 int	(*carp_get_vhid_p)(struct ifaddr *);
131 
132 /*
133  * Used by rtsock/raw_input callback code to decide whether to filter the update
134  * notification to a socket bound to a particular FIB.
135  */
136 #define	RTS_FILTER_FIB	M_PROTO8
137 
138 typedef struct {
139 	int	ip_count;	/* attached w/ AF_INET */
140 	int	ip6_count;	/* attached w/ AF_INET6 */
141 	int	any_count;	/* total attached */
142 } route_cb_t;
143 VNET_DEFINE_STATIC(route_cb_t, route_cb);
144 #define	V_route_cb VNET(route_cb)
145 
146 struct mtx rtsock_mtx;
147 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
148 
149 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
150 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
151 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
152 
153 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
154 
155 struct walkarg {
156 	int	w_tmemsize;
157 	int	w_op, w_arg;
158 	caddr_t	w_tmem;
159 	struct sysctl_req *w_req;
160 };
161 
162 static void	rts_input(struct mbuf *m);
163 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
164 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
165 			struct walkarg *w, int *plen);
166 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
167 			struct rt_addrinfo *rtinfo);
168 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
169 static int	sysctl_iflist(int af, struct walkarg *w);
170 static int	sysctl_ifmalist(int af, struct walkarg *w);
171 static int	route_output(struct mbuf *m, struct socket *so, ...);
172 static void	rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
173 static void	rt_dispatch(struct mbuf *, sa_family_t);
174 static struct sockaddr	*rtsock_fix_netmask(struct sockaddr *dst,
175 			struct sockaddr *smask, struct sockaddr_storage *dmask);
176 
177 static struct netisr_handler rtsock_nh = {
178 	.nh_name = "rtsock",
179 	.nh_handler = rts_input,
180 	.nh_proto = NETISR_ROUTE,
181 	.nh_policy = NETISR_POLICY_SOURCE,
182 };
183 
184 static int
185 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
186 {
187 	int error, qlimit;
188 
189 	netisr_getqlimit(&rtsock_nh, &qlimit);
190 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
191         if (error || !req->newptr)
192                 return (error);
193 	if (qlimit < 1)
194 		return (EINVAL);
195 	return (netisr_setqlimit(&rtsock_nh, qlimit));
196 }
197 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
198     0, 0, sysctl_route_netisr_maxqlen, "I",
199     "maximum routing socket dispatch queue length");
200 
201 static void
202 vnet_rts_init(void)
203 {
204 	int tmp;
205 
206 	if (IS_DEFAULT_VNET(curvnet)) {
207 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
208 			rtsock_nh.nh_qlimit = tmp;
209 		netisr_register(&rtsock_nh);
210 	}
211 #ifdef VIMAGE
212 	 else
213 		netisr_register_vnet(&rtsock_nh);
214 #endif
215 }
216 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
217     vnet_rts_init, 0);
218 
219 #ifdef VIMAGE
220 static void
221 vnet_rts_uninit(void)
222 {
223 
224 	netisr_unregister_vnet(&rtsock_nh);
225 }
226 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
227     vnet_rts_uninit, 0);
228 #endif
229 
230 static int
231 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
232     struct rawcb *rp)
233 {
234 	int fibnum;
235 
236 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
237 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
238 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
239 
240 	/* No filtering requested. */
241 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
242 		return (0);
243 
244 	/* Check if it is a rts and the fib matches the one of the socket. */
245 	fibnum = M_GETFIB(m);
246 	if (proto->sp_family != PF_ROUTE ||
247 	    rp->rcb_socket == NULL ||
248 	    rp->rcb_socket->so_fibnum == fibnum)
249 		return (0);
250 
251 	/* Filtering requested and no match, the socket shall be skipped. */
252 	return (1);
253 }
254 
255 static void
256 rts_input(struct mbuf *m)
257 {
258 	struct sockproto route_proto;
259 	unsigned short *family;
260 	struct m_tag *tag;
261 
262 	route_proto.sp_family = PF_ROUTE;
263 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
264 	if (tag != NULL) {
265 		family = (unsigned short *)(tag + 1);
266 		route_proto.sp_protocol = *family;
267 		m_tag_delete(m, tag);
268 	} else
269 		route_proto.sp_protocol = 0;
270 
271 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
272 }
273 
274 /*
275  * It really doesn't make any sense at all for this code to share much
276  * with raw_usrreq.c, since its functionality is so restricted.  XXX
277  */
278 static void
279 rts_abort(struct socket *so)
280 {
281 
282 	raw_usrreqs.pru_abort(so);
283 }
284 
285 static void
286 rts_close(struct socket *so)
287 {
288 
289 	raw_usrreqs.pru_close(so);
290 }
291 
292 /* pru_accept is EOPNOTSUPP */
293 
294 static int
295 rts_attach(struct socket *so, int proto, struct thread *td)
296 {
297 	struct rawcb *rp;
298 	int error;
299 
300 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
301 
302 	/* XXX */
303 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
304 
305 	so->so_pcb = (caddr_t)rp;
306 	so->so_fibnum = td->td_proc->p_fibnum;
307 	error = raw_attach(so, proto);
308 	rp = sotorawcb(so);
309 	if (error) {
310 		so->so_pcb = NULL;
311 		free(rp, M_PCB);
312 		return error;
313 	}
314 	RTSOCK_LOCK();
315 	switch(rp->rcb_proto.sp_protocol) {
316 	case AF_INET:
317 		V_route_cb.ip_count++;
318 		break;
319 	case AF_INET6:
320 		V_route_cb.ip6_count++;
321 		break;
322 	}
323 	V_route_cb.any_count++;
324 	RTSOCK_UNLOCK();
325 	soisconnected(so);
326 	so->so_options |= SO_USELOOPBACK;
327 	return 0;
328 }
329 
330 static int
331 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
332 {
333 
334 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
335 }
336 
337 static int
338 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
339 {
340 
341 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
342 }
343 
344 /* pru_connect2 is EOPNOTSUPP */
345 /* pru_control is EOPNOTSUPP */
346 
347 static void
348 rts_detach(struct socket *so)
349 {
350 	struct rawcb *rp = sotorawcb(so);
351 
352 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
353 
354 	RTSOCK_LOCK();
355 	switch(rp->rcb_proto.sp_protocol) {
356 	case AF_INET:
357 		V_route_cb.ip_count--;
358 		break;
359 	case AF_INET6:
360 		V_route_cb.ip6_count--;
361 		break;
362 	}
363 	V_route_cb.any_count--;
364 	RTSOCK_UNLOCK();
365 	raw_usrreqs.pru_detach(so);
366 }
367 
368 static int
369 rts_disconnect(struct socket *so)
370 {
371 
372 	return (raw_usrreqs.pru_disconnect(so));
373 }
374 
375 /* pru_listen is EOPNOTSUPP */
376 
377 static int
378 rts_peeraddr(struct socket *so, struct sockaddr **nam)
379 {
380 
381 	return (raw_usrreqs.pru_peeraddr(so, nam));
382 }
383 
384 /* pru_rcvd is EOPNOTSUPP */
385 /* pru_rcvoob is EOPNOTSUPP */
386 
387 static int
388 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
389 	 struct mbuf *control, struct thread *td)
390 {
391 
392 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
393 }
394 
395 /* pru_sense is null */
396 
397 static int
398 rts_shutdown(struct socket *so)
399 {
400 
401 	return (raw_usrreqs.pru_shutdown(so));
402 }
403 
404 static int
405 rts_sockaddr(struct socket *so, struct sockaddr **nam)
406 {
407 
408 	return (raw_usrreqs.pru_sockaddr(so, nam));
409 }
410 
411 static struct pr_usrreqs route_usrreqs = {
412 	.pru_abort =		rts_abort,
413 	.pru_attach =		rts_attach,
414 	.pru_bind =		rts_bind,
415 	.pru_connect =		rts_connect,
416 	.pru_detach =		rts_detach,
417 	.pru_disconnect =	rts_disconnect,
418 	.pru_peeraddr =		rts_peeraddr,
419 	.pru_send =		rts_send,
420 	.pru_shutdown =		rts_shutdown,
421 	.pru_sockaddr =		rts_sockaddr,
422 	.pru_close =		rts_close,
423 };
424 
425 #ifndef _SOCKADDR_UNION_DEFINED
426 #define	_SOCKADDR_UNION_DEFINED
427 /*
428  * The union of all possible address formats we handle.
429  */
430 union sockaddr_union {
431 	struct sockaddr		sa;
432 	struct sockaddr_in	sin;
433 	struct sockaddr_in6	sin6;
434 };
435 #endif /* _SOCKADDR_UNION_DEFINED */
436 
437 static int
438 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
439     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
440 {
441 
442 	/* First, see if the returned address is part of the jail. */
443 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
444 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
445 		return (0);
446 	}
447 
448 	switch (info->rti_info[RTAX_DST]->sa_family) {
449 #ifdef INET
450 	case AF_INET:
451 	{
452 		struct in_addr ia;
453 		struct ifaddr *ifa;
454 		int found;
455 
456 		found = 0;
457 		/*
458 		 * Try to find an address on the given outgoing interface
459 		 * that belongs to the jail.
460 		 */
461 		IF_ADDR_RLOCK(ifp);
462 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
463 			struct sockaddr *sa;
464 			sa = ifa->ifa_addr;
465 			if (sa->sa_family != AF_INET)
466 				continue;
467 			ia = ((struct sockaddr_in *)sa)->sin_addr;
468 			if (prison_check_ip4(cred, &ia) == 0) {
469 				found = 1;
470 				break;
471 			}
472 		}
473 		IF_ADDR_RUNLOCK(ifp);
474 		if (!found) {
475 			/*
476 			 * As a last resort return the 'default' jail address.
477 			 */
478 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
479 			    sin_addr;
480 			if (prison_get_ip4(cred, &ia) != 0)
481 				return (ESRCH);
482 		}
483 		bzero(&saun->sin, sizeof(struct sockaddr_in));
484 		saun->sin.sin_len = sizeof(struct sockaddr_in);
485 		saun->sin.sin_family = AF_INET;
486 		saun->sin.sin_addr.s_addr = ia.s_addr;
487 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
488 		break;
489 	}
490 #endif
491 #ifdef INET6
492 	case AF_INET6:
493 	{
494 		struct in6_addr ia6;
495 		struct ifaddr *ifa;
496 		int found;
497 
498 		found = 0;
499 		/*
500 		 * Try to find an address on the given outgoing interface
501 		 * that belongs to the jail.
502 		 */
503 		IF_ADDR_RLOCK(ifp);
504 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
505 			struct sockaddr *sa;
506 			sa = ifa->ifa_addr;
507 			if (sa->sa_family != AF_INET6)
508 				continue;
509 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
510 			    &ia6, sizeof(struct in6_addr));
511 			if (prison_check_ip6(cred, &ia6) == 0) {
512 				found = 1;
513 				break;
514 			}
515 		}
516 		IF_ADDR_RUNLOCK(ifp);
517 		if (!found) {
518 			/*
519 			 * As a last resort return the 'default' jail address.
520 			 */
521 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
522 			    sin6_addr;
523 			if (prison_get_ip6(cred, &ia6) != 0)
524 				return (ESRCH);
525 		}
526 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
527 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
528 		saun->sin6.sin6_family = AF_INET6;
529 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
530 		if (sa6_recoverscope(&saun->sin6) != 0)
531 			return (ESRCH);
532 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
533 		break;
534 	}
535 #endif
536 	default:
537 		return (ESRCH);
538 	}
539 	return (0);
540 }
541 
542 /*ARGSUSED*/
543 static int
544 route_output(struct mbuf *m, struct socket *so, ...)
545 {
546 	RIB_RLOCK_TRACKER;
547 	struct rt_msghdr *rtm = NULL;
548 	struct rtentry *rt = NULL;
549 	struct rib_head *rnh;
550 	struct rt_addrinfo info;
551 	struct sockaddr_storage ss;
552 #ifdef INET6
553 	struct sockaddr_in6 *sin6;
554 	int i, rti_need_deembed = 0;
555 #endif
556 	int alloc_len = 0, len, error = 0, fibnum;
557 	struct ifnet *ifp = NULL;
558 	union sockaddr_union saun;
559 	sa_family_t saf = AF_UNSPEC;
560 	struct rawcb *rp = NULL;
561 	struct walkarg w;
562 
563 	fibnum = so->so_fibnum;
564 
565 #define senderr(e) { error = e; goto flush;}
566 	if (m == NULL || ((m->m_len < sizeof(long)) &&
567 		       (m = m_pullup(m, sizeof(long))) == NULL))
568 		return (ENOBUFS);
569 	if ((m->m_flags & M_PKTHDR) == 0)
570 		panic("route_output");
571 	len = m->m_pkthdr.len;
572 	if (len < sizeof(*rtm) ||
573 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
574 		senderr(EINVAL);
575 
576 	/*
577 	 * Most of current messages are in range 200-240 bytes,
578 	 * minimize possible re-allocation on reply using larger size
579 	 * buffer aligned on 1k boundaty.
580 	 */
581 	alloc_len = roundup2(len, 1024);
582 	if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
583 		senderr(ENOBUFS);
584 
585 	m_copydata(m, 0, len, (caddr_t)rtm);
586 	bzero(&info, sizeof(info));
587 	bzero(&w, sizeof(w));
588 
589 	if (rtm->rtm_version != RTM_VERSION) {
590 		/* Do not touch message since format is unknown */
591 		free(rtm, M_TEMP);
592 		rtm = NULL;
593 		senderr(EPROTONOSUPPORT);
594 	}
595 
596 	/*
597 	 * Starting from here, it is possible
598 	 * to alter original message and insert
599 	 * caller PID and error value.
600 	 */
601 
602 	rtm->rtm_pid = curproc->p_pid;
603 	info.rti_addrs = rtm->rtm_addrs;
604 
605 	info.rti_mflags = rtm->rtm_inits;
606 	info.rti_rmx = &rtm->rtm_rmx;
607 
608 	/*
609 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
610 	 * link-local address because rtrequest requires addresses with
611 	 * embedded scope id.
612 	 */
613 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
614 		senderr(EINVAL);
615 
616 	info.rti_flags = rtm->rtm_flags;
617 	if (info.rti_info[RTAX_DST] == NULL ||
618 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
619 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
620 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
621 		senderr(EINVAL);
622 	saf = info.rti_info[RTAX_DST]->sa_family;
623 	/*
624 	 * Verify that the caller has the appropriate privilege; RTM_GET
625 	 * is the only operation the non-superuser is allowed.
626 	 */
627 	if (rtm->rtm_type != RTM_GET) {
628 		error = priv_check(curthread, PRIV_NET_ROUTE);
629 		if (error)
630 			senderr(error);
631 	}
632 
633 	/*
634 	 * The given gateway address may be an interface address.
635 	 * For example, issuing a "route change" command on a route
636 	 * entry that was created from a tunnel, and the gateway
637 	 * address given is the local end point. In this case the
638 	 * RTF_GATEWAY flag must be cleared or the destination will
639 	 * not be reachable even though there is no error message.
640 	 */
641 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
642 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
643 		struct rt_addrinfo ginfo;
644 		struct sockaddr *gdst;
645 
646 		bzero(&ginfo, sizeof(ginfo));
647 		bzero(&ss, sizeof(ss));
648 		ss.ss_len = sizeof(ss);
649 
650 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
651 		gdst = info.rti_info[RTAX_GATEWAY];
652 
653 		/*
654 		 * A host route through the loopback interface is
655 		 * installed for each interface adddress. In pre 8.0
656 		 * releases the interface address of a PPP link type
657 		 * is not reachable locally. This behavior is fixed as
658 		 * part of the new L2/L3 redesign and rewrite work. The
659 		 * signature of this interface address route is the
660 		 * AF_LINK sa_family type of the rt_gateway, and the
661 		 * rt_ifp has the IFF_LOOPBACK flag set.
662 		 */
663 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
664 			if (ss.ss_family == AF_LINK &&
665 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
666 				info.rti_flags &= ~RTF_GATEWAY;
667 				info.rti_flags |= RTF_GWFLAG_COMPAT;
668 			}
669 			rib_free_info(&ginfo);
670 		}
671 	}
672 
673 	switch (rtm->rtm_type) {
674 		struct rtentry *saved_nrt;
675 
676 	case RTM_ADD:
677 	case RTM_CHANGE:
678 		if (rtm->rtm_type == RTM_ADD) {
679 			if (info.rti_info[RTAX_GATEWAY] == NULL)
680 				senderr(EINVAL);
681 		}
682 		saved_nrt = NULL;
683 
684 		/* support for new ARP code */
685 		if (info.rti_info[RTAX_GATEWAY] != NULL &&
686 		    info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
687 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
688 			error = lla_rt_output(rtm, &info);
689 #ifdef INET6
690 			if (error == 0)
691 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
692 #endif
693 			break;
694 		}
695 		error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
696 		    fibnum);
697 		if (error == 0 && saved_nrt != NULL) {
698 #ifdef INET6
699 			rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
700 #endif
701 			RT_LOCK(saved_nrt);
702 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
703 			RT_REMREF(saved_nrt);
704 			RT_UNLOCK(saved_nrt);
705 		}
706 		break;
707 
708 	case RTM_DELETE:
709 		saved_nrt = NULL;
710 		/* support for new ARP code */
711 		if (info.rti_info[RTAX_GATEWAY] &&
712 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
713 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
714 			error = lla_rt_output(rtm, &info);
715 #ifdef INET6
716 			if (error == 0)
717 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
718 #endif
719 			break;
720 		}
721 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
722 		if (error == 0) {
723 			RT_LOCK(saved_nrt);
724 			rt = saved_nrt;
725 			goto report;
726 		}
727 #ifdef INET6
728 		/* rt_msg2() will not be used when RTM_DELETE fails. */
729 		rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
730 #endif
731 		break;
732 
733 	case RTM_GET:
734 		rnh = rt_tables_get_rnh(fibnum, saf);
735 		if (rnh == NULL)
736 			senderr(EAFNOSUPPORT);
737 
738 		RIB_RLOCK(rnh);
739 
740 		if (info.rti_info[RTAX_NETMASK] == NULL &&
741 		    rtm->rtm_type == RTM_GET) {
742 			/*
743 			 * Provide longest prefix match for
744 			 * address lookup (no mask).
745 			 * 'route -n get addr'
746 			 */
747 			rt = (struct rtentry *) rnh->rnh_matchaddr(
748 			    info.rti_info[RTAX_DST], &rnh->head);
749 		} else
750 			rt = (struct rtentry *) rnh->rnh_lookup(
751 			    info.rti_info[RTAX_DST],
752 			    info.rti_info[RTAX_NETMASK], &rnh->head);
753 
754 		if (rt == NULL) {
755 			RIB_RUNLOCK(rnh);
756 			senderr(ESRCH);
757 		}
758 #ifdef RADIX_MPATH
759 		/*
760 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
761 		 * we require users to specify a matching RTAX_GATEWAY.
762 		 *
763 		 * for RTM_GET, gate is optional even with multipath.
764 		 * if gate == NULL the first match is returned.
765 		 * (no need to call rt_mpath_matchgate if gate == NULL)
766 		 */
767 		if (rt_mpath_capable(rnh) &&
768 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
769 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
770 			if (!rt) {
771 				RIB_RUNLOCK(rnh);
772 				senderr(ESRCH);
773 			}
774 		}
775 #endif
776 		/*
777 		 * If performing proxied L2 entry insertion, and
778 		 * the actual PPP host entry is found, perform
779 		 * another search to retrieve the prefix route of
780 		 * the local end point of the PPP link.
781 		 */
782 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
783 			struct sockaddr laddr;
784 
785 			if (rt->rt_ifp != NULL &&
786 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
787 				struct ifaddr *ifa;
788 
789 				NET_EPOCH_ENTER();
790 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1,
791 						RT_ALL_FIBS);
792 				if (ifa != NULL)
793 					rt_maskedcopy(ifa->ifa_addr,
794 						      &laddr,
795 						      ifa->ifa_netmask);
796 				NET_EPOCH_EXIT();
797 			} else
798 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
799 					      &laddr,
800 					      rt->rt_ifa->ifa_netmask);
801 			/*
802 			 * refactor rt and no lock operation necessary
803 			 */
804 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
805 			    &rnh->head);
806 			if (rt == NULL) {
807 				RIB_RUNLOCK(rnh);
808 				senderr(ESRCH);
809 			}
810 		}
811 		RT_LOCK(rt);
812 		RT_ADDREF(rt);
813 		RIB_RUNLOCK(rnh);
814 
815 report:
816 		RT_LOCK_ASSERT(rt);
817 		if ((rt->rt_flags & RTF_HOST) == 0
818 		    ? jailed_without_vnet(curthread->td_ucred)
819 		    : prison_if(curthread->td_ucred,
820 		    rt_key(rt)) != 0) {
821 			RT_UNLOCK(rt);
822 			senderr(ESRCH);
823 		}
824 		info.rti_info[RTAX_DST] = rt_key(rt);
825 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
826 		info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
827 		    rt_mask(rt), &ss);
828 		info.rti_info[RTAX_GENMASK] = 0;
829 		if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
830 			ifp = rt->rt_ifp;
831 			if (ifp) {
832 				info.rti_info[RTAX_IFP] =
833 				    ifp->if_addr->ifa_addr;
834 				error = rtm_get_jailed(&info, ifp, rt,
835 				    &saun, curthread->td_ucred);
836 				if (error != 0) {
837 					RT_UNLOCK(rt);
838 					senderr(error);
839 				}
840 				if (ifp->if_flags & IFF_POINTOPOINT)
841 					info.rti_info[RTAX_BRD] =
842 					    rt->rt_ifa->ifa_dstaddr;
843 				rtm->rtm_index = ifp->if_index;
844 			} else {
845 				info.rti_info[RTAX_IFP] = NULL;
846 				info.rti_info[RTAX_IFA] = NULL;
847 			}
848 		} else if ((ifp = rt->rt_ifp) != NULL) {
849 			rtm->rtm_index = ifp->if_index;
850 		}
851 
852 		/* Check if we need to realloc storage */
853 		rtsock_msg_buffer(rtm->rtm_type, &info, NULL, &len);
854 		if (len > alloc_len) {
855 			struct rt_msghdr *new_rtm;
856 			new_rtm = malloc(len, M_TEMP, M_NOWAIT);
857 			if (new_rtm == NULL) {
858 				RT_UNLOCK(rt);
859 				senderr(ENOBUFS);
860 			}
861 			bcopy(rtm, new_rtm, rtm->rtm_msglen);
862 			free(rtm, M_TEMP);
863 			rtm = new_rtm;
864 			alloc_len = len;
865 		}
866 
867 		w.w_tmem = (caddr_t)rtm;
868 		w.w_tmemsize = alloc_len;
869 		rtsock_msg_buffer(rtm->rtm_type, &info, &w, &len);
870 
871 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
872 			rtm->rtm_flags = RTF_GATEWAY |
873 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
874 		else
875 			rtm->rtm_flags = rt->rt_flags;
876 		rt_getmetrics(rt, &rtm->rtm_rmx);
877 		rtm->rtm_addrs = info.rti_addrs;
878 
879 		RT_UNLOCK(rt);
880 		break;
881 
882 	default:
883 		senderr(EOPNOTSUPP);
884 	}
885 
886 flush:
887 	if (rt != NULL)
888 		RTFREE(rt);
889 	/*
890 	 * Check to see if we don't want our own messages.
891 	 */
892 	if ((so->so_options & SO_USELOOPBACK) == 0) {
893 		if (V_route_cb.any_count <= 1) {
894 			if (rtm != NULL)
895 				free(rtm, M_TEMP);
896 			m_freem(m);
897 			return (error);
898 		}
899 		/* There is another listener, so construct message */
900 		rp = sotorawcb(so);
901 	}
902 
903 	if (rtm != NULL) {
904 #ifdef INET6
905 		if (rti_need_deembed) {
906 			/* sin6_scope_id is recovered before sending rtm. */
907 			sin6 = (struct sockaddr_in6 *)&ss;
908 			for (i = 0; i < RTAX_MAX; i++) {
909 				if (info.rti_info[i] == NULL)
910 					continue;
911 				if (info.rti_info[i]->sa_family != AF_INET6)
912 					continue;
913 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
914 				if (sa6_recoverscope(sin6) == 0)
915 					bcopy(sin6, info.rti_info[i],
916 						    sizeof(*sin6));
917 			}
918 		}
919 #endif
920 		if (error != 0)
921 			rtm->rtm_errno = error;
922 		else
923 			rtm->rtm_flags |= RTF_DONE;
924 
925 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
926 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
927 			m_freem(m);
928 			m = NULL;
929 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
930 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
931 
932 		free(rtm, M_TEMP);
933 	}
934 	if (m != NULL) {
935 		M_SETFIB(m, fibnum);
936 		m->m_flags |= RTS_FILTER_FIB;
937 		if (rp) {
938 			/*
939 			 * XXX insure we don't get a copy by
940 			 * invalidating our protocol
941 			 */
942 			unsigned short family = rp->rcb_proto.sp_family;
943 			rp->rcb_proto.sp_family = 0;
944 			rt_dispatch(m, saf);
945 			rp->rcb_proto.sp_family = family;
946 		} else
947 			rt_dispatch(m, saf);
948 	}
949 
950 	return (error);
951 }
952 
953 static void
954 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
955 {
956 
957 	bzero(out, sizeof(*out));
958 	out->rmx_mtu = rt->rt_mtu;
959 	out->rmx_weight = rt->rt_weight;
960 	out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
961 	/* Kernel -> userland timebase conversion. */
962 	out->rmx_expire = rt->rt_expire ?
963 	    rt->rt_expire - time_uptime + time_second : 0;
964 }
965 
966 /*
967  * Extract the addresses of the passed sockaddrs.
968  * Do a little sanity checking so as to avoid bad memory references.
969  * This data is derived straight from userland.
970  */
971 static int
972 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
973 {
974 	struct sockaddr *sa;
975 	int i;
976 
977 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
978 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
979 			continue;
980 		sa = (struct sockaddr *)cp;
981 		/*
982 		 * It won't fit.
983 		 */
984 		if (cp + sa->sa_len > cplim)
985 			return (EINVAL);
986 		/*
987 		 * there are no more.. quit now
988 		 * If there are more bits, they are in error.
989 		 * I've seen this. route(1) can evidently generate these.
990 		 * This causes kernel to core dump.
991 		 * for compatibility, If we see this, point to a safe address.
992 		 */
993 		if (sa->sa_len == 0) {
994 			rtinfo->rti_info[i] = &sa_zero;
995 			return (0); /* should be EINVAL but for compat */
996 		}
997 		/* accept it */
998 #ifdef INET6
999 		if (sa->sa_family == AF_INET6)
1000 			sa6_embedscope((struct sockaddr_in6 *)sa,
1001 			    V_ip6_use_defzone);
1002 #endif
1003 		rtinfo->rti_info[i] = sa;
1004 		cp += SA_SIZE(sa);
1005 	}
1006 	return (0);
1007 }
1008 
1009 /*
1010  * Fill in @dmask with valid netmask leaving original @smask
1011  * intact. Mostly used with radix netmasks.
1012  */
1013 static struct sockaddr *
1014 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
1015     struct sockaddr_storage *dmask)
1016 {
1017 	if (dst == NULL || smask == NULL)
1018 		return (NULL);
1019 
1020 	memset(dmask, 0, dst->sa_len);
1021 	memcpy(dmask, smask, smask->sa_len);
1022 	dmask->ss_len = dst->sa_len;
1023 	dmask->ss_family = dst->sa_family;
1024 
1025 	return ((struct sockaddr *)dmask);
1026 }
1027 
1028 /*
1029  * Writes information related to @rtinfo object to newly-allocated mbuf.
1030  * Assumes MCLBYTES is enough to construct any message.
1031  * Used for OS notifications of vaious events (if/ifa announces,etc)
1032  *
1033  * Returns allocated mbuf or NULL on failure.
1034  */
1035 static struct mbuf *
1036 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1037 {
1038 	struct rt_msghdr *rtm;
1039 	struct mbuf *m;
1040 	int i;
1041 	struct sockaddr *sa;
1042 #ifdef INET6
1043 	struct sockaddr_storage ss;
1044 	struct sockaddr_in6 *sin6;
1045 #endif
1046 	int len, dlen;
1047 
1048 	switch (type) {
1049 
1050 	case RTM_DELADDR:
1051 	case RTM_NEWADDR:
1052 		len = sizeof(struct ifa_msghdr);
1053 		break;
1054 
1055 	case RTM_DELMADDR:
1056 	case RTM_NEWMADDR:
1057 		len = sizeof(struct ifma_msghdr);
1058 		break;
1059 
1060 	case RTM_IFINFO:
1061 		len = sizeof(struct if_msghdr);
1062 		break;
1063 
1064 	case RTM_IFANNOUNCE:
1065 	case RTM_IEEE80211:
1066 		len = sizeof(struct if_announcemsghdr);
1067 		break;
1068 
1069 	default:
1070 		len = sizeof(struct rt_msghdr);
1071 	}
1072 
1073 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1074 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1075 	if (len > MHLEN)
1076 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1077 	else
1078 		m = m_gethdr(M_NOWAIT, MT_DATA);
1079 	if (m == NULL)
1080 		return (m);
1081 
1082 	m->m_pkthdr.len = m->m_len = len;
1083 	rtm = mtod(m, struct rt_msghdr *);
1084 	bzero((caddr_t)rtm, len);
1085 	for (i = 0; i < RTAX_MAX; i++) {
1086 		if ((sa = rtinfo->rti_info[i]) == NULL)
1087 			continue;
1088 		rtinfo->rti_addrs |= (1 << i);
1089 		dlen = SA_SIZE(sa);
1090 #ifdef INET6
1091 		if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1092 			sin6 = (struct sockaddr_in6 *)&ss;
1093 			bcopy(sa, sin6, sizeof(*sin6));
1094 			if (sa6_recoverscope(sin6) == 0)
1095 				sa = (struct sockaddr *)sin6;
1096 		}
1097 #endif
1098 		m_copyback(m, len, dlen, (caddr_t)sa);
1099 		len += dlen;
1100 	}
1101 	if (m->m_pkthdr.len != len) {
1102 		m_freem(m);
1103 		return (NULL);
1104 	}
1105 	rtm->rtm_msglen = len;
1106 	rtm->rtm_version = RTM_VERSION;
1107 	rtm->rtm_type = type;
1108 	return (m);
1109 }
1110 
1111 /*
1112  * Writes information related to @rtinfo object to preallocated buffer.
1113  * Stores needed size in @plen. If @w is NULL, calculates size without
1114  * writing.
1115  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1116  *
1117  * Returns 0 on success.
1118  *
1119  */
1120 static int
1121 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1122 {
1123 	int i;
1124 	int len, buflen = 0, dlen;
1125 	caddr_t cp = NULL;
1126 	struct rt_msghdr *rtm = NULL;
1127 #ifdef INET6
1128 	struct sockaddr_storage ss;
1129 	struct sockaddr_in6 *sin6;
1130 #endif
1131 #ifdef COMPAT_FREEBSD32
1132 	bool compat32 = false;
1133 #endif
1134 
1135 	switch (type) {
1136 
1137 	case RTM_DELADDR:
1138 	case RTM_NEWADDR:
1139 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1140 #ifdef COMPAT_FREEBSD32
1141 			if (w->w_req->flags & SCTL_MASK32) {
1142 				len = sizeof(struct ifa_msghdrl32);
1143 				compat32 = true;
1144 			} else
1145 #endif
1146 				len = sizeof(struct ifa_msghdrl);
1147 		} else
1148 			len = sizeof(struct ifa_msghdr);
1149 		break;
1150 
1151 	case RTM_IFINFO:
1152 #ifdef COMPAT_FREEBSD32
1153 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1154 			if (w->w_op == NET_RT_IFLISTL)
1155 				len = sizeof(struct if_msghdrl32);
1156 			else
1157 				len = sizeof(struct if_msghdr32);
1158 			compat32 = true;
1159 			break;
1160 		}
1161 #endif
1162 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1163 			len = sizeof(struct if_msghdrl);
1164 		else
1165 			len = sizeof(struct if_msghdr);
1166 		break;
1167 
1168 	case RTM_NEWMADDR:
1169 		len = sizeof(struct ifma_msghdr);
1170 		break;
1171 
1172 	default:
1173 		len = sizeof(struct rt_msghdr);
1174 	}
1175 
1176 	if (w != NULL) {
1177 		rtm = (struct rt_msghdr *)w->w_tmem;
1178 		buflen = w->w_tmemsize - len;
1179 		cp = (caddr_t)w->w_tmem + len;
1180 	}
1181 
1182 	rtinfo->rti_addrs = 0;
1183 	for (i = 0; i < RTAX_MAX; i++) {
1184 		struct sockaddr *sa;
1185 
1186 		if ((sa = rtinfo->rti_info[i]) == NULL)
1187 			continue;
1188 		rtinfo->rti_addrs |= (1 << i);
1189 #ifdef COMPAT_FREEBSD32
1190 		if (compat32)
1191 			dlen = SA_SIZE32(sa);
1192 		else
1193 #endif
1194 			dlen = SA_SIZE(sa);
1195 		if (cp != NULL && buflen >= dlen) {
1196 #ifdef INET6
1197 			if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1198 				sin6 = (struct sockaddr_in6 *)&ss;
1199 				bcopy(sa, sin6, sizeof(*sin6));
1200 				if (sa6_recoverscope(sin6) == 0)
1201 					sa = (struct sockaddr *)sin6;
1202 			}
1203 #endif
1204 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1205 			cp += dlen;
1206 			buflen -= dlen;
1207 		} else if (cp != NULL) {
1208 			/*
1209 			 * Buffer too small. Count needed size
1210 			 * and return with error.
1211 			 */
1212 			cp = NULL;
1213 		}
1214 
1215 		len += dlen;
1216 	}
1217 
1218 	if (cp != NULL) {
1219 		dlen = ALIGN(len) - len;
1220 		if (buflen < dlen)
1221 			cp = NULL;
1222 		else
1223 			buflen -= dlen;
1224 	}
1225 	len = ALIGN(len);
1226 
1227 	if (cp != NULL) {
1228 		/* fill header iff buffer is large enough */
1229 		rtm->rtm_version = RTM_VERSION;
1230 		rtm->rtm_type = type;
1231 		rtm->rtm_msglen = len;
1232 	}
1233 
1234 	*plen = len;
1235 
1236 	if (w != NULL && cp == NULL)
1237 		return (ENOBUFS);
1238 
1239 	return (0);
1240 }
1241 
1242 /*
1243  * This routine is called to generate a message from the routing
1244  * socket indicating that a redirect has occurred, a routing lookup
1245  * has failed, or that a protocol has detected timeouts to a particular
1246  * destination.
1247  */
1248 void
1249 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1250     int fibnum)
1251 {
1252 	struct rt_msghdr *rtm;
1253 	struct mbuf *m;
1254 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1255 
1256 	if (V_route_cb.any_count == 0)
1257 		return;
1258 	m = rtsock_msg_mbuf(type, rtinfo);
1259 	if (m == NULL)
1260 		return;
1261 
1262 	if (fibnum != RT_ALL_FIBS) {
1263 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1264 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1265 		M_SETFIB(m, fibnum);
1266 		m->m_flags |= RTS_FILTER_FIB;
1267 	}
1268 
1269 	rtm = mtod(m, struct rt_msghdr *);
1270 	rtm->rtm_flags = RTF_DONE | flags;
1271 	rtm->rtm_errno = error;
1272 	rtm->rtm_addrs = rtinfo->rti_addrs;
1273 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1274 }
1275 
1276 void
1277 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1278 {
1279 
1280 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1281 }
1282 
1283 /*
1284  * This routine is called to generate a message from the routing
1285  * socket indicating that the status of a network interface has changed.
1286  */
1287 void
1288 rt_ifmsg(struct ifnet *ifp)
1289 {
1290 	struct if_msghdr *ifm;
1291 	struct mbuf *m;
1292 	struct rt_addrinfo info;
1293 
1294 	if (V_route_cb.any_count == 0)
1295 		return;
1296 	bzero((caddr_t)&info, sizeof(info));
1297 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1298 	if (m == NULL)
1299 		return;
1300 	ifm = mtod(m, struct if_msghdr *);
1301 	ifm->ifm_index = ifp->if_index;
1302 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1303 	if_data_copy(ifp, &ifm->ifm_data);
1304 	ifm->ifm_addrs = 0;
1305 	rt_dispatch(m, AF_UNSPEC);
1306 }
1307 
1308 /*
1309  * Announce interface address arrival/withdraw.
1310  * Please do not call directly, use rt_addrmsg().
1311  * Assume input data to be valid.
1312  * Returns 0 on success.
1313  */
1314 int
1315 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1316 {
1317 	struct rt_addrinfo info;
1318 	struct sockaddr *sa;
1319 	int ncmd;
1320 	struct mbuf *m;
1321 	struct ifa_msghdr *ifam;
1322 	struct ifnet *ifp = ifa->ifa_ifp;
1323 	struct sockaddr_storage ss;
1324 
1325 	if (V_route_cb.any_count == 0)
1326 		return (0);
1327 
1328 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1329 
1330 	bzero((caddr_t)&info, sizeof(info));
1331 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1332 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1333 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1334 	    info.rti_info[RTAX_IFP], ifa->ifa_netmask, &ss);
1335 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1336 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1337 		return (ENOBUFS);
1338 	ifam = mtod(m, struct ifa_msghdr *);
1339 	ifam->ifam_index = ifp->if_index;
1340 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1341 	ifam->ifam_flags = ifa->ifa_flags;
1342 	ifam->ifam_addrs = info.rti_addrs;
1343 
1344 	if (fibnum != RT_ALL_FIBS) {
1345 		M_SETFIB(m, fibnum);
1346 		m->m_flags |= RTS_FILTER_FIB;
1347 	}
1348 
1349 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1350 
1351 	return (0);
1352 }
1353 
1354 /*
1355  * Announce route addition/removal.
1356  * Please do not call directly, use rt_routemsg().
1357  * Note that @rt data MAY be inconsistent/invalid:
1358  * if some userland app sends us "invalid" route message (invalid mask,
1359  * no dst, wrong address families, etc...) we need to pass it back
1360  * to app (and any other rtsock consumers) with rtm_errno field set to
1361  * non-zero value.
1362  *
1363  * Returns 0 on success.
1364  */
1365 int
1366 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1367     int fibnum)
1368 {
1369 	struct rt_addrinfo info;
1370 	struct sockaddr *sa;
1371 	struct mbuf *m;
1372 	struct rt_msghdr *rtm;
1373 	struct sockaddr_storage ss;
1374 
1375 	if (V_route_cb.any_count == 0)
1376 		return (0);
1377 
1378 	bzero((caddr_t)&info, sizeof(info));
1379 	info.rti_info[RTAX_DST] = sa = rt_key(rt);
1380 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(sa, rt_mask(rt), &ss);
1381 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1382 	if ((m = rtsock_msg_mbuf(cmd, &info)) == NULL)
1383 		return (ENOBUFS);
1384 	rtm = mtod(m, struct rt_msghdr *);
1385 	rtm->rtm_index = ifp->if_index;
1386 	rtm->rtm_flags |= rt->rt_flags;
1387 	rtm->rtm_errno = error;
1388 	rtm->rtm_addrs = info.rti_addrs;
1389 
1390 	if (fibnum != RT_ALL_FIBS) {
1391 		M_SETFIB(m, fibnum);
1392 		m->m_flags |= RTS_FILTER_FIB;
1393 	}
1394 
1395 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1396 
1397 	return (0);
1398 }
1399 
1400 /*
1401  * This is the analogue to the rt_newaddrmsg which performs the same
1402  * function but for multicast group memberhips.  This is easier since
1403  * there is no route state to worry about.
1404  */
1405 void
1406 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1407 {
1408 	struct rt_addrinfo info;
1409 	struct mbuf *m = NULL;
1410 	struct ifnet *ifp = ifma->ifma_ifp;
1411 	struct ifma_msghdr *ifmam;
1412 
1413 	if (V_route_cb.any_count == 0)
1414 		return;
1415 
1416 	bzero((caddr_t)&info, sizeof(info));
1417 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1418 	if (ifp && ifp->if_addr)
1419 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1420 	else
1421 		info.rti_info[RTAX_IFP] = NULL;
1422 	/*
1423 	 * If a link-layer address is present, present it as a ``gateway''
1424 	 * (similarly to how ARP entries, e.g., are presented).
1425 	 */
1426 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1427 	m = rtsock_msg_mbuf(cmd, &info);
1428 	if (m == NULL)
1429 		return;
1430 	ifmam = mtod(m, struct ifma_msghdr *);
1431 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1432 	    __func__));
1433 	ifmam->ifmam_index = ifp->if_index;
1434 	ifmam->ifmam_addrs = info.rti_addrs;
1435 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1436 }
1437 
1438 static struct mbuf *
1439 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1440 	struct rt_addrinfo *info)
1441 {
1442 	struct if_announcemsghdr *ifan;
1443 	struct mbuf *m;
1444 
1445 	if (V_route_cb.any_count == 0)
1446 		return NULL;
1447 	bzero((caddr_t)info, sizeof(*info));
1448 	m = rtsock_msg_mbuf(type, info);
1449 	if (m != NULL) {
1450 		ifan = mtod(m, struct if_announcemsghdr *);
1451 		ifan->ifan_index = ifp->if_index;
1452 		strlcpy(ifan->ifan_name, ifp->if_xname,
1453 			sizeof(ifan->ifan_name));
1454 		ifan->ifan_what = what;
1455 	}
1456 	return m;
1457 }
1458 
1459 /*
1460  * This is called to generate routing socket messages indicating
1461  * IEEE80211 wireless events.
1462  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1463  */
1464 void
1465 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1466 {
1467 	struct mbuf *m;
1468 	struct rt_addrinfo info;
1469 
1470 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1471 	if (m != NULL) {
1472 		/*
1473 		 * Append the ieee80211 data.  Try to stick it in the
1474 		 * mbuf containing the ifannounce msg; otherwise allocate
1475 		 * a new mbuf and append.
1476 		 *
1477 		 * NB: we assume m is a single mbuf.
1478 		 */
1479 		if (data_len > M_TRAILINGSPACE(m)) {
1480 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1481 			if (n == NULL) {
1482 				m_freem(m);
1483 				return;
1484 			}
1485 			bcopy(data, mtod(n, void *), data_len);
1486 			n->m_len = data_len;
1487 			m->m_next = n;
1488 		} else if (data_len > 0) {
1489 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1490 			m->m_len += data_len;
1491 		}
1492 		if (m->m_flags & M_PKTHDR)
1493 			m->m_pkthdr.len += data_len;
1494 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1495 		rt_dispatch(m, AF_UNSPEC);
1496 	}
1497 }
1498 
1499 /*
1500  * This is called to generate routing socket messages indicating
1501  * network interface arrival and departure.
1502  */
1503 void
1504 rt_ifannouncemsg(struct ifnet *ifp, int what)
1505 {
1506 	struct mbuf *m;
1507 	struct rt_addrinfo info;
1508 
1509 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1510 	if (m != NULL)
1511 		rt_dispatch(m, AF_UNSPEC);
1512 }
1513 
1514 static void
1515 rt_dispatch(struct mbuf *m, sa_family_t saf)
1516 {
1517 	struct m_tag *tag;
1518 
1519 	/*
1520 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1521 	 * use when injecting the mbuf into the routing socket buffer from
1522 	 * the netisr.
1523 	 */
1524 	if (saf != AF_UNSPEC) {
1525 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1526 		    M_NOWAIT);
1527 		if (tag == NULL) {
1528 			m_freem(m);
1529 			return;
1530 		}
1531 		*(unsigned short *)(tag + 1) = saf;
1532 		m_tag_prepend(m, tag);
1533 	}
1534 #ifdef VIMAGE
1535 	if (V_loif)
1536 		m->m_pkthdr.rcvif = V_loif;
1537 	else {
1538 		m_freem(m);
1539 		return;
1540 	}
1541 #endif
1542 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1543 }
1544 
1545 /*
1546  * This is used in dumping the kernel table via sysctl().
1547  */
1548 static int
1549 sysctl_dumpentry(struct radix_node *rn, void *vw)
1550 {
1551 	struct walkarg *w = vw;
1552 	struct rtentry *rt = (struct rtentry *)rn;
1553 	int error = 0, size;
1554 	struct rt_addrinfo info;
1555 	struct sockaddr_storage ss;
1556 
1557 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1558 		return 0;
1559 	if ((rt->rt_flags & RTF_HOST) == 0
1560 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1561 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1562 		return (0);
1563 	bzero((caddr_t)&info, sizeof(info));
1564 	info.rti_info[RTAX_DST] = rt_key(rt);
1565 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1566 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1567 	    rt_mask(rt), &ss);
1568 	info.rti_info[RTAX_GENMASK] = 0;
1569 	if (rt->rt_ifp) {
1570 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1571 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1572 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1573 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1574 	}
1575 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1576 		return (error);
1577 	if (w->w_req && w->w_tmem) {
1578 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1579 
1580 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1581 			rtm->rtm_flags = RTF_GATEWAY |
1582 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1583 		else
1584 			rtm->rtm_flags = rt->rt_flags;
1585 		rt_getmetrics(rt, &rtm->rtm_rmx);
1586 		rtm->rtm_index = rt->rt_ifp->if_index;
1587 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1588 		rtm->rtm_addrs = info.rti_addrs;
1589 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1590 		return (error);
1591 	}
1592 	return (error);
1593 }
1594 
1595 static int
1596 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1597     struct rt_addrinfo *info, struct walkarg *w, int len)
1598 {
1599 	struct if_msghdrl *ifm;
1600 	struct if_data *ifd;
1601 
1602 	ifm = (struct if_msghdrl *)w->w_tmem;
1603 
1604 #ifdef COMPAT_FREEBSD32
1605 	if (w->w_req->flags & SCTL_MASK32) {
1606 		struct if_msghdrl32 *ifm32;
1607 
1608 		ifm32 = (struct if_msghdrl32 *)ifm;
1609 		ifm32->ifm_addrs = info->rti_addrs;
1610 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1611 		ifm32->ifm_index = ifp->if_index;
1612 		ifm32->_ifm_spare1 = 0;
1613 		ifm32->ifm_len = sizeof(*ifm32);
1614 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1615 		ifd = &ifm32->ifm_data;
1616 	} else
1617 #endif
1618 	{
1619 		ifm->ifm_addrs = info->rti_addrs;
1620 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1621 		ifm->ifm_index = ifp->if_index;
1622 		ifm->_ifm_spare1 = 0;
1623 		ifm->ifm_len = sizeof(*ifm);
1624 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1625 		ifd = &ifm->ifm_data;
1626 	}
1627 
1628 	memcpy(ifd, src_ifd, sizeof(*ifd));
1629 
1630 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1631 }
1632 
1633 static int
1634 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1635     struct rt_addrinfo *info, struct walkarg *w, int len)
1636 {
1637 	struct if_msghdr *ifm;
1638 	struct if_data *ifd;
1639 
1640 	ifm = (struct if_msghdr *)w->w_tmem;
1641 
1642 #ifdef COMPAT_FREEBSD32
1643 	if (w->w_req->flags & SCTL_MASK32) {
1644 		struct if_msghdr32 *ifm32;
1645 
1646 		ifm32 = (struct if_msghdr32 *)ifm;
1647 		ifm32->ifm_addrs = info->rti_addrs;
1648 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1649 		ifm32->ifm_index = ifp->if_index;
1650 		ifd = &ifm32->ifm_data;
1651 	} else
1652 #endif
1653 	{
1654 		ifm->ifm_addrs = info->rti_addrs;
1655 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1656 		ifm->ifm_index = ifp->if_index;
1657 		ifd = &ifm->ifm_data;
1658 	}
1659 
1660 	memcpy(ifd, src_ifd, sizeof(*ifd));
1661 
1662 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1663 }
1664 
1665 static int
1666 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1667     struct walkarg *w, int len)
1668 {
1669 	struct ifa_msghdrl *ifam;
1670 	struct if_data *ifd;
1671 
1672 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1673 
1674 #ifdef COMPAT_FREEBSD32
1675 	if (w->w_req->flags & SCTL_MASK32) {
1676 		struct ifa_msghdrl32 *ifam32;
1677 
1678 		ifam32 = (struct ifa_msghdrl32 *)ifam;
1679 		ifam32->ifam_addrs = info->rti_addrs;
1680 		ifam32->ifam_flags = ifa->ifa_flags;
1681 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1682 		ifam32->_ifam_spare1 = 0;
1683 		ifam32->ifam_len = sizeof(*ifam32);
1684 		ifam32->ifam_data_off =
1685 		    offsetof(struct ifa_msghdrl32, ifam_data);
1686 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1687 		ifd = &ifam32->ifam_data;
1688 	} else
1689 #endif
1690 	{
1691 		ifam->ifam_addrs = info->rti_addrs;
1692 		ifam->ifam_flags = ifa->ifa_flags;
1693 		ifam->ifam_index = ifa->ifa_ifp->if_index;
1694 		ifam->_ifam_spare1 = 0;
1695 		ifam->ifam_len = sizeof(*ifam);
1696 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1697 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1698 		ifd = &ifam->ifam_data;
1699 	}
1700 
1701 	bzero(ifd, sizeof(*ifd));
1702 	ifd->ifi_datalen = sizeof(struct if_data);
1703 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1704 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1705 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1706 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1707 
1708 	/* Fixup if_data carp(4) vhid. */
1709 	if (carp_get_vhid_p != NULL)
1710 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1711 
1712 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1713 }
1714 
1715 static int
1716 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1717     struct walkarg *w, int len)
1718 {
1719 	struct ifa_msghdr *ifam;
1720 
1721 	ifam = (struct ifa_msghdr *)w->w_tmem;
1722 	ifam->ifam_addrs = info->rti_addrs;
1723 	ifam->ifam_flags = ifa->ifa_flags;
1724 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1725 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1726 
1727 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1728 }
1729 
1730 static int
1731 sysctl_iflist(int af, struct walkarg *w)
1732 {
1733 	struct ifnet *ifp;
1734 	struct ifaddr *ifa;
1735 	struct if_data ifd;
1736 	struct rt_addrinfo info;
1737 	int len, error = 0;
1738 	struct sockaddr_storage ss;
1739 	struct epoch_tracker et;
1740 
1741 	bzero((caddr_t)&info, sizeof(info));
1742 	bzero(&ifd, sizeof(ifd));
1743 	NET_EPOCH_ENTER_ET(et);
1744 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1745 		if (w->w_arg && w->w_arg != ifp->if_index)
1746 			continue;
1747 		if_data_copy(ifp, &ifd);
1748 		ifa = ifp->if_addr;
1749 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1750 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1751 		if (error != 0)
1752 			goto done;
1753 		info.rti_info[RTAX_IFP] = NULL;
1754 		if (w->w_req && w->w_tmem) {
1755 			if (w->w_op == NET_RT_IFLISTL)
1756 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1757 				    len);
1758 			else
1759 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1760 				    len);
1761 			if (error)
1762 				goto done;
1763 		}
1764 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1765 			if (af && af != ifa->ifa_addr->sa_family)
1766 				continue;
1767 			if (prison_if(w->w_req->td->td_ucred,
1768 			    ifa->ifa_addr) != 0)
1769 				continue;
1770 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1771 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1772 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
1773 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1774 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1775 			if (error != 0)
1776 				goto done;
1777 			if (w->w_req && w->w_tmem) {
1778 				if (w->w_op == NET_RT_IFLISTL)
1779 					error = sysctl_iflist_ifaml(ifa, &info,
1780 					    w, len);
1781 				else
1782 					error = sysctl_iflist_ifam(ifa, &info,
1783 					    w, len);
1784 				if (error)
1785 					goto done;
1786 			}
1787 		}
1788 		info.rti_info[RTAX_IFA] = NULL;
1789 		info.rti_info[RTAX_NETMASK] = NULL;
1790 		info.rti_info[RTAX_BRD] = NULL;
1791 	}
1792 done:
1793 	NET_EPOCH_EXIT_ET(et);
1794 	return (error);
1795 }
1796 
1797 static int
1798 sysctl_ifmalist(int af, struct walkarg *w)
1799 {
1800 	struct rt_addrinfo info;
1801 	struct ifaddr *ifa;
1802 	struct ifmultiaddr *ifma;
1803 	struct ifnet *ifp;
1804 	int error, len;
1805 
1806 	error = 0;
1807 	bzero((caddr_t)&info, sizeof(info));
1808 
1809 	IFNET_RLOCK_NOSLEEP();
1810 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1811 		if (w->w_arg && w->w_arg != ifp->if_index)
1812 			continue;
1813 		ifa = ifp->if_addr;
1814 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1815 		IF_ADDR_RLOCK(ifp);
1816 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1817 			if (af && af != ifma->ifma_addr->sa_family)
1818 				continue;
1819 			if (prison_if(w->w_req->td->td_ucred,
1820 			    ifma->ifma_addr) != 0)
1821 				continue;
1822 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1823 			info.rti_info[RTAX_GATEWAY] =
1824 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1825 			    ifma->ifma_lladdr : NULL;
1826 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1827 			if (error != 0)
1828 				break;
1829 			if (w->w_req && w->w_tmem) {
1830 				struct ifma_msghdr *ifmam;
1831 
1832 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1833 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1834 				ifmam->ifmam_flags = 0;
1835 				ifmam->ifmam_addrs = info.rti_addrs;
1836 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1837 				if (error != 0)
1838 					break;
1839 			}
1840 		}
1841 		IF_ADDR_RUNLOCK(ifp);
1842 		if (error != 0)
1843 			break;
1844 	}
1845 	IFNET_RUNLOCK_NOSLEEP();
1846 	return (error);
1847 }
1848 
1849 static int
1850 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1851 {
1852 	RIB_RLOCK_TRACKER;
1853 	int	*name = (int *)arg1;
1854 	u_int	namelen = arg2;
1855 	struct rib_head *rnh = NULL; /* silence compiler. */
1856 	int	i, lim, error = EINVAL;
1857 	int	fib = 0;
1858 	u_char	af;
1859 	struct	walkarg w;
1860 
1861 	name ++;
1862 	namelen--;
1863 	if (req->newptr)
1864 		return (EPERM);
1865 	if (name[1] == NET_RT_DUMP) {
1866 		if (namelen == 3)
1867 			fib = req->td->td_proc->p_fibnum;
1868 		else if (namelen == 4)
1869 			fib = (name[3] == RT_ALL_FIBS) ?
1870 			    req->td->td_proc->p_fibnum : name[3];
1871 		else
1872 			return ((namelen < 3) ? EISDIR : ENOTDIR);
1873 		if (fib < 0 || fib >= rt_numfibs)
1874 			return (EINVAL);
1875 	} else if (namelen != 3)
1876 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1877 	af = name[0];
1878 	if (af > AF_MAX)
1879 		return (EINVAL);
1880 	bzero(&w, sizeof(w));
1881 	w.w_op = name[1];
1882 	w.w_arg = name[2];
1883 	w.w_req = req;
1884 
1885 	error = sysctl_wire_old_buffer(req, 0);
1886 	if (error)
1887 		return (error);
1888 
1889 	/*
1890 	 * Allocate reply buffer in advance.
1891 	 * All rtsock messages has maximum length of u_short.
1892 	 */
1893 	w.w_tmemsize = 65536;
1894 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
1895 
1896 	switch (w.w_op) {
1897 
1898 	case NET_RT_DUMP:
1899 	case NET_RT_FLAGS:
1900 		if (af == 0) {			/* dump all tables */
1901 			i = 1;
1902 			lim = AF_MAX;
1903 		} else				/* dump only one table */
1904 			i = lim = af;
1905 
1906 		/*
1907 		 * take care of llinfo entries, the caller must
1908 		 * specify an AF
1909 		 */
1910 		if (w.w_op == NET_RT_FLAGS &&
1911 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1912 			if (af != 0)
1913 				error = lltable_sysctl_dumparp(af, w.w_req);
1914 			else
1915 				error = EINVAL;
1916 			break;
1917 		}
1918 		/*
1919 		 * take care of routing entries
1920 		 */
1921 		for (error = 0; error == 0 && i <= lim; i++) {
1922 			rnh = rt_tables_get_rnh(fib, i);
1923 			if (rnh != NULL) {
1924 				RIB_RLOCK(rnh);
1925 			    	error = rnh->rnh_walktree(&rnh->head,
1926 				    sysctl_dumpentry, &w);
1927 				RIB_RUNLOCK(rnh);
1928 			} else if (af != 0)
1929 				error = EAFNOSUPPORT;
1930 		}
1931 		break;
1932 
1933 	case NET_RT_IFLIST:
1934 	case NET_RT_IFLISTL:
1935 		error = sysctl_iflist(af, &w);
1936 		break;
1937 
1938 	case NET_RT_IFMALIST:
1939 		error = sysctl_ifmalist(af, &w);
1940 		break;
1941 	}
1942 
1943 	free(w.w_tmem, M_TEMP);
1944 	return (error);
1945 }
1946 
1947 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1948 
1949 /*
1950  * Definitions of protocols supported in the ROUTE domain.
1951  */
1952 
1953 static struct domain routedomain;		/* or at least forward */
1954 
1955 static struct protosw routesw[] = {
1956 {
1957 	.pr_type =		SOCK_RAW,
1958 	.pr_domain =		&routedomain,
1959 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1960 	.pr_output =		route_output,
1961 	.pr_ctlinput =		raw_ctlinput,
1962 	.pr_init =		raw_init,
1963 	.pr_usrreqs =		&route_usrreqs
1964 }
1965 };
1966 
1967 static struct domain routedomain = {
1968 	.dom_family =		PF_ROUTE,
1969 	.dom_name =		 "route",
1970 	.dom_protosw =		routesw,
1971 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
1972 };
1973 
1974 VNET_DOMAIN_SET(route);
1975