xref: /freebsd/sys/net/rtsock.c (revision c321779676d0722f5d0a65c30e8b23b8a6547364)
1 /*-
2  * Copyright (c) 1988, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
30  * $FreeBSD$
31  */
32 #include "opt_compat.h"
33 #include "opt_mpath.h"
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 
37 #include <sys/param.h>
38 #include <sys/jail.h>
39 #include <sys/kernel.h>
40 #include <sys/domain.h>
41 #include <sys/lock.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/priv.h>
45 #include <sys/proc.h>
46 #include <sys/protosw.h>
47 #include <sys/rwlock.h>
48 #include <sys/signalvar.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/sysctl.h>
52 #include <sys/systm.h>
53 
54 #include <net/if.h>
55 #include <net/if_var.h>
56 #include <net/if_dl.h>
57 #include <net/if_llatbl.h>
58 #include <net/if_types.h>
59 #include <net/netisr.h>
60 #include <net/raw_cb.h>
61 #include <net/route.h>
62 #include <net/route_var.h>
63 #include <net/vnet.h>
64 
65 #include <netinet/in.h>
66 #include <netinet/if_ether.h>
67 #include <netinet/ip_carp.h>
68 #ifdef INET6
69 #include <netinet6/ip6_var.h>
70 #include <netinet6/scope6_var.h>
71 #endif
72 
73 #ifdef COMPAT_FREEBSD32
74 #include <sys/mount.h>
75 #include <compat/freebsd32/freebsd32.h>
76 
77 struct if_msghdr32 {
78 	uint16_t ifm_msglen;
79 	uint8_t	ifm_version;
80 	uint8_t	ifm_type;
81 	int32_t	ifm_addrs;
82 	int32_t	ifm_flags;
83 	uint16_t ifm_index;
84 	struct	if_data ifm_data;
85 };
86 
87 struct if_msghdrl32 {
88 	uint16_t ifm_msglen;
89 	uint8_t	ifm_version;
90 	uint8_t	ifm_type;
91 	int32_t	ifm_addrs;
92 	int32_t	ifm_flags;
93 	uint16_t ifm_index;
94 	uint16_t _ifm_spare1;
95 	uint16_t ifm_len;
96 	uint16_t ifm_data_off;
97 	struct	if_data ifm_data;
98 };
99 
100 struct ifa_msghdrl32 {
101 	uint16_t ifam_msglen;
102 	uint8_t	ifam_version;
103 	uint8_t	ifam_type;
104 	int32_t	ifam_addrs;
105 	int32_t	ifam_flags;
106 	uint16_t ifam_index;
107 	uint16_t _ifam_spare1;
108 	uint16_t ifam_len;
109 	uint16_t ifam_data_off;
110 	int32_t	ifam_metric;
111 	struct	if_data ifam_data;
112 };
113 #endif /* COMPAT_FREEBSD32 */
114 
115 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
116 
117 /* NB: these are not modified */
118 static struct	sockaddr route_src = { 2, PF_ROUTE, };
119 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
120 
121 /* These are external hooks for CARP. */
122 int	(*carp_get_vhid_p)(struct ifaddr *);
123 
124 /*
125  * Used by rtsock/raw_input callback code to decide whether to filter the update
126  * notification to a socket bound to a particular FIB.
127  */
128 #define	RTS_FILTER_FIB	M_PROTO8
129 
130 typedef struct {
131 	int	ip_count;	/* attached w/ AF_INET */
132 	int	ip6_count;	/* attached w/ AF_INET6 */
133 	int	any_count;	/* total attached */
134 } route_cb_t;
135 static VNET_DEFINE(route_cb_t, route_cb);
136 #define	V_route_cb VNET(route_cb)
137 
138 struct mtx rtsock_mtx;
139 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
140 
141 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
142 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
143 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
144 
145 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
146 
147 struct walkarg {
148 	int	w_tmemsize;
149 	int	w_op, w_arg;
150 	caddr_t	w_tmem;
151 	struct sysctl_req *w_req;
152 };
153 
154 static void	rts_input(struct mbuf *m);
155 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
156 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
157 			struct walkarg *w, int *plen);
158 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
159 			struct rt_addrinfo *rtinfo);
160 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
161 static int	sysctl_iflist(int af, struct walkarg *w);
162 static int	sysctl_ifmalist(int af, struct walkarg *w);
163 static int	route_output(struct mbuf *m, struct socket *so, ...);
164 static void	rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
165 static void	rt_dispatch(struct mbuf *, sa_family_t);
166 static struct sockaddr	*rtsock_fix_netmask(struct sockaddr *dst,
167 			struct sockaddr *smask, struct sockaddr_storage *dmask);
168 
169 static struct netisr_handler rtsock_nh = {
170 	.nh_name = "rtsock",
171 	.nh_handler = rts_input,
172 	.nh_proto = NETISR_ROUTE,
173 	.nh_policy = NETISR_POLICY_SOURCE,
174 };
175 
176 static int
177 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
178 {
179 	int error, qlimit;
180 
181 	netisr_getqlimit(&rtsock_nh, &qlimit);
182 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
183         if (error || !req->newptr)
184                 return (error);
185 	if (qlimit < 1)
186 		return (EINVAL);
187 	return (netisr_setqlimit(&rtsock_nh, qlimit));
188 }
189 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
190     0, 0, sysctl_route_netisr_maxqlen, "I",
191     "maximum routing socket dispatch queue length");
192 
193 static void
194 vnet_rts_init(void)
195 {
196 	int tmp;
197 
198 	if (IS_DEFAULT_VNET(curvnet)) {
199 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
200 			rtsock_nh.nh_qlimit = tmp;
201 		netisr_register(&rtsock_nh);
202 	}
203 #ifdef VIMAGE
204 	 else
205 		netisr_register_vnet(&rtsock_nh);
206 #endif
207 }
208 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
209     vnet_rts_init, 0);
210 
211 #ifdef VIMAGE
212 static void
213 vnet_rts_uninit(void)
214 {
215 
216 	netisr_unregister_vnet(&rtsock_nh);
217 }
218 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
219     vnet_rts_uninit, 0);
220 #endif
221 
222 static int
223 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
224     struct rawcb *rp)
225 {
226 	int fibnum;
227 
228 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
229 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
230 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
231 
232 	/* No filtering requested. */
233 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
234 		return (0);
235 
236 	/* Check if it is a rts and the fib matches the one of the socket. */
237 	fibnum = M_GETFIB(m);
238 	if (proto->sp_family != PF_ROUTE ||
239 	    rp->rcb_socket == NULL ||
240 	    rp->rcb_socket->so_fibnum == fibnum)
241 		return (0);
242 
243 	/* Filtering requested and no match, the socket shall be skipped. */
244 	return (1);
245 }
246 
247 static void
248 rts_input(struct mbuf *m)
249 {
250 	struct sockproto route_proto;
251 	unsigned short *family;
252 	struct m_tag *tag;
253 
254 	route_proto.sp_family = PF_ROUTE;
255 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
256 	if (tag != NULL) {
257 		family = (unsigned short *)(tag + 1);
258 		route_proto.sp_protocol = *family;
259 		m_tag_delete(m, tag);
260 	} else
261 		route_proto.sp_protocol = 0;
262 
263 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
264 }
265 
266 /*
267  * It really doesn't make any sense at all for this code to share much
268  * with raw_usrreq.c, since its functionality is so restricted.  XXX
269  */
270 static void
271 rts_abort(struct socket *so)
272 {
273 
274 	raw_usrreqs.pru_abort(so);
275 }
276 
277 static void
278 rts_close(struct socket *so)
279 {
280 
281 	raw_usrreqs.pru_close(so);
282 }
283 
284 /* pru_accept is EOPNOTSUPP */
285 
286 static int
287 rts_attach(struct socket *so, int proto, struct thread *td)
288 {
289 	struct rawcb *rp;
290 	int error;
291 
292 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
293 
294 	/* XXX */
295 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
296 
297 	so->so_pcb = (caddr_t)rp;
298 	so->so_fibnum = td->td_proc->p_fibnum;
299 	error = raw_attach(so, proto);
300 	rp = sotorawcb(so);
301 	if (error) {
302 		so->so_pcb = NULL;
303 		free(rp, M_PCB);
304 		return error;
305 	}
306 	RTSOCK_LOCK();
307 	switch(rp->rcb_proto.sp_protocol) {
308 	case AF_INET:
309 		V_route_cb.ip_count++;
310 		break;
311 	case AF_INET6:
312 		V_route_cb.ip6_count++;
313 		break;
314 	}
315 	V_route_cb.any_count++;
316 	RTSOCK_UNLOCK();
317 	soisconnected(so);
318 	so->so_options |= SO_USELOOPBACK;
319 	return 0;
320 }
321 
322 static int
323 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
324 {
325 
326 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
327 }
328 
329 static int
330 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
331 {
332 
333 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
334 }
335 
336 /* pru_connect2 is EOPNOTSUPP */
337 /* pru_control is EOPNOTSUPP */
338 
339 static void
340 rts_detach(struct socket *so)
341 {
342 	struct rawcb *rp = sotorawcb(so);
343 
344 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
345 
346 	RTSOCK_LOCK();
347 	switch(rp->rcb_proto.sp_protocol) {
348 	case AF_INET:
349 		V_route_cb.ip_count--;
350 		break;
351 	case AF_INET6:
352 		V_route_cb.ip6_count--;
353 		break;
354 	}
355 	V_route_cb.any_count--;
356 	RTSOCK_UNLOCK();
357 	raw_usrreqs.pru_detach(so);
358 }
359 
360 static int
361 rts_disconnect(struct socket *so)
362 {
363 
364 	return (raw_usrreqs.pru_disconnect(so));
365 }
366 
367 /* pru_listen is EOPNOTSUPP */
368 
369 static int
370 rts_peeraddr(struct socket *so, struct sockaddr **nam)
371 {
372 
373 	return (raw_usrreqs.pru_peeraddr(so, nam));
374 }
375 
376 /* pru_rcvd is EOPNOTSUPP */
377 /* pru_rcvoob is EOPNOTSUPP */
378 
379 static int
380 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
381 	 struct mbuf *control, struct thread *td)
382 {
383 
384 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
385 }
386 
387 /* pru_sense is null */
388 
389 static int
390 rts_shutdown(struct socket *so)
391 {
392 
393 	return (raw_usrreqs.pru_shutdown(so));
394 }
395 
396 static int
397 rts_sockaddr(struct socket *so, struct sockaddr **nam)
398 {
399 
400 	return (raw_usrreqs.pru_sockaddr(so, nam));
401 }
402 
403 static struct pr_usrreqs route_usrreqs = {
404 	.pru_abort =		rts_abort,
405 	.pru_attach =		rts_attach,
406 	.pru_bind =		rts_bind,
407 	.pru_connect =		rts_connect,
408 	.pru_detach =		rts_detach,
409 	.pru_disconnect =	rts_disconnect,
410 	.pru_peeraddr =		rts_peeraddr,
411 	.pru_send =		rts_send,
412 	.pru_shutdown =		rts_shutdown,
413 	.pru_sockaddr =		rts_sockaddr,
414 	.pru_close =		rts_close,
415 };
416 
417 #ifndef _SOCKADDR_UNION_DEFINED
418 #define	_SOCKADDR_UNION_DEFINED
419 /*
420  * The union of all possible address formats we handle.
421  */
422 union sockaddr_union {
423 	struct sockaddr		sa;
424 	struct sockaddr_in	sin;
425 	struct sockaddr_in6	sin6;
426 };
427 #endif /* _SOCKADDR_UNION_DEFINED */
428 
429 static int
430 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
431     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
432 {
433 
434 	/* First, see if the returned address is part of the jail. */
435 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
436 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
437 		return (0);
438 	}
439 
440 	switch (info->rti_info[RTAX_DST]->sa_family) {
441 #ifdef INET
442 	case AF_INET:
443 	{
444 		struct in_addr ia;
445 		struct ifaddr *ifa;
446 		int found;
447 
448 		found = 0;
449 		/*
450 		 * Try to find an address on the given outgoing interface
451 		 * that belongs to the jail.
452 		 */
453 		IF_ADDR_RLOCK(ifp);
454 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
455 			struct sockaddr *sa;
456 			sa = ifa->ifa_addr;
457 			if (sa->sa_family != AF_INET)
458 				continue;
459 			ia = ((struct sockaddr_in *)sa)->sin_addr;
460 			if (prison_check_ip4(cred, &ia) == 0) {
461 				found = 1;
462 				break;
463 			}
464 		}
465 		IF_ADDR_RUNLOCK(ifp);
466 		if (!found) {
467 			/*
468 			 * As a last resort return the 'default' jail address.
469 			 */
470 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
471 			    sin_addr;
472 			if (prison_get_ip4(cred, &ia) != 0)
473 				return (ESRCH);
474 		}
475 		bzero(&saun->sin, sizeof(struct sockaddr_in));
476 		saun->sin.sin_len = sizeof(struct sockaddr_in);
477 		saun->sin.sin_family = AF_INET;
478 		saun->sin.sin_addr.s_addr = ia.s_addr;
479 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
480 		break;
481 	}
482 #endif
483 #ifdef INET6
484 	case AF_INET6:
485 	{
486 		struct in6_addr ia6;
487 		struct ifaddr *ifa;
488 		int found;
489 
490 		found = 0;
491 		/*
492 		 * Try to find an address on the given outgoing interface
493 		 * that belongs to the jail.
494 		 */
495 		IF_ADDR_RLOCK(ifp);
496 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
497 			struct sockaddr *sa;
498 			sa = ifa->ifa_addr;
499 			if (sa->sa_family != AF_INET6)
500 				continue;
501 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
502 			    &ia6, sizeof(struct in6_addr));
503 			if (prison_check_ip6(cred, &ia6) == 0) {
504 				found = 1;
505 				break;
506 			}
507 		}
508 		IF_ADDR_RUNLOCK(ifp);
509 		if (!found) {
510 			/*
511 			 * As a last resort return the 'default' jail address.
512 			 */
513 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
514 			    sin6_addr;
515 			if (prison_get_ip6(cred, &ia6) != 0)
516 				return (ESRCH);
517 		}
518 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
519 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
520 		saun->sin6.sin6_family = AF_INET6;
521 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
522 		if (sa6_recoverscope(&saun->sin6) != 0)
523 			return (ESRCH);
524 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
525 		break;
526 	}
527 #endif
528 	default:
529 		return (ESRCH);
530 	}
531 	return (0);
532 }
533 
534 /*ARGSUSED*/
535 static int
536 route_output(struct mbuf *m, struct socket *so, ...)
537 {
538 	struct rt_msghdr *rtm = NULL;
539 	struct rtentry *rt = NULL;
540 	struct rib_head *rnh;
541 	struct rt_addrinfo info;
542 	struct sockaddr_storage ss;
543 #ifdef INET6
544 	struct sockaddr_in6 *sin6;
545 	int i, rti_need_deembed = 0;
546 #endif
547 	int alloc_len = 0, len, error = 0, fibnum;
548 	struct ifnet *ifp = NULL;
549 	union sockaddr_union saun;
550 	sa_family_t saf = AF_UNSPEC;
551 	struct rawcb *rp = NULL;
552 	struct walkarg w;
553 
554 	fibnum = so->so_fibnum;
555 
556 #define senderr(e) { error = e; goto flush;}
557 	if (m == NULL || ((m->m_len < sizeof(long)) &&
558 		       (m = m_pullup(m, sizeof(long))) == NULL))
559 		return (ENOBUFS);
560 	if ((m->m_flags & M_PKTHDR) == 0)
561 		panic("route_output");
562 	len = m->m_pkthdr.len;
563 	if (len < sizeof(*rtm) ||
564 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
565 		senderr(EINVAL);
566 
567 	/*
568 	 * Most of current messages are in range 200-240 bytes,
569 	 * minimize possible re-allocation on reply using larger size
570 	 * buffer aligned on 1k boundaty.
571 	 */
572 	alloc_len = roundup2(len, 1024);
573 	if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
574 		senderr(ENOBUFS);
575 
576 	m_copydata(m, 0, len, (caddr_t)rtm);
577 	bzero(&info, sizeof(info));
578 	bzero(&w, sizeof(w));
579 
580 	if (rtm->rtm_version != RTM_VERSION) {
581 		/* Do not touch message since format is unknown */
582 		free(rtm, M_TEMP);
583 		rtm = NULL;
584 		senderr(EPROTONOSUPPORT);
585 	}
586 
587 	/*
588 	 * Starting from here, it is possible
589 	 * to alter original message and insert
590 	 * caller PID and error value.
591 	 */
592 
593 	rtm->rtm_pid = curproc->p_pid;
594 	info.rti_addrs = rtm->rtm_addrs;
595 
596 	info.rti_mflags = rtm->rtm_inits;
597 	info.rti_rmx = &rtm->rtm_rmx;
598 
599 	/*
600 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
601 	 * link-local address because rtrequest requires addresses with
602 	 * embedded scope id.
603 	 */
604 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
605 		senderr(EINVAL);
606 
607 	info.rti_flags = rtm->rtm_flags;
608 	if (info.rti_info[RTAX_DST] == NULL ||
609 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
610 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
611 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
612 		senderr(EINVAL);
613 	saf = info.rti_info[RTAX_DST]->sa_family;
614 	/*
615 	 * Verify that the caller has the appropriate privilege; RTM_GET
616 	 * is the only operation the non-superuser is allowed.
617 	 */
618 	if (rtm->rtm_type != RTM_GET) {
619 		error = priv_check(curthread, PRIV_NET_ROUTE);
620 		if (error)
621 			senderr(error);
622 	}
623 
624 	/*
625 	 * The given gateway address may be an interface address.
626 	 * For example, issuing a "route change" command on a route
627 	 * entry that was created from a tunnel, and the gateway
628 	 * address given is the local end point. In this case the
629 	 * RTF_GATEWAY flag must be cleared or the destination will
630 	 * not be reachable even though there is no error message.
631 	 */
632 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
633 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
634 		struct rt_addrinfo ginfo;
635 		struct sockaddr *gdst;
636 
637 		bzero(&ginfo, sizeof(ginfo));
638 		bzero(&ss, sizeof(ss));
639 		ss.ss_len = sizeof(ss);
640 
641 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
642 		gdst = info.rti_info[RTAX_GATEWAY];
643 
644 		/*
645 		 * A host route through the loopback interface is
646 		 * installed for each interface adddress. In pre 8.0
647 		 * releases the interface address of a PPP link type
648 		 * is not reachable locally. This behavior is fixed as
649 		 * part of the new L2/L3 redesign and rewrite work. The
650 		 * signature of this interface address route is the
651 		 * AF_LINK sa_family type of the rt_gateway, and the
652 		 * rt_ifp has the IFF_LOOPBACK flag set.
653 		 */
654 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
655 			if (ss.ss_family == AF_LINK &&
656 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
657 				info.rti_flags &= ~RTF_GATEWAY;
658 				info.rti_flags |= RTF_GWFLAG_COMPAT;
659 			}
660 			rib_free_info(&ginfo);
661 		}
662 	}
663 
664 	switch (rtm->rtm_type) {
665 		struct rtentry *saved_nrt;
666 
667 	case RTM_ADD:
668 	case RTM_CHANGE:
669 		if (info.rti_info[RTAX_GATEWAY] == NULL)
670 			senderr(EINVAL);
671 		saved_nrt = NULL;
672 
673 		/* support for new ARP code */
674 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
675 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
676 			error = lla_rt_output(rtm, &info);
677 #ifdef INET6
678 			if (error == 0)
679 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
680 #endif
681 			break;
682 		}
683 		error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
684 		    fibnum);
685 		if (error == 0 && saved_nrt != NULL) {
686 #ifdef INET6
687 			rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
688 #endif
689 			RT_LOCK(saved_nrt);
690 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
691 			RT_REMREF(saved_nrt);
692 			RT_UNLOCK(saved_nrt);
693 		}
694 		break;
695 
696 	case RTM_DELETE:
697 		saved_nrt = NULL;
698 		/* support for new ARP code */
699 		if (info.rti_info[RTAX_GATEWAY] &&
700 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
701 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
702 			error = lla_rt_output(rtm, &info);
703 #ifdef INET6
704 			if (error == 0)
705 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
706 #endif
707 			break;
708 		}
709 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
710 		if (error == 0) {
711 			RT_LOCK(saved_nrt);
712 			rt = saved_nrt;
713 			goto report;
714 		}
715 #ifdef INET6
716 		/* rt_msg2() will not be used when RTM_DELETE fails. */
717 		rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
718 #endif
719 		break;
720 
721 	case RTM_GET:
722 		rnh = rt_tables_get_rnh(fibnum, saf);
723 		if (rnh == NULL)
724 			senderr(EAFNOSUPPORT);
725 
726 		RIB_RLOCK(rnh);
727 
728 		if (info.rti_info[RTAX_NETMASK] == NULL &&
729 		    rtm->rtm_type == RTM_GET) {
730 			/*
731 			 * Provide longest prefix match for
732 			 * address lookup (no mask).
733 			 * 'route -n get addr'
734 			 */
735 			rt = (struct rtentry *) rnh->rnh_matchaddr(
736 			    info.rti_info[RTAX_DST], &rnh->head);
737 		} else
738 			rt = (struct rtentry *) rnh->rnh_lookup(
739 			    info.rti_info[RTAX_DST],
740 			    info.rti_info[RTAX_NETMASK], &rnh->head);
741 
742 		if (rt == NULL) {
743 			RIB_RUNLOCK(rnh);
744 			senderr(ESRCH);
745 		}
746 #ifdef RADIX_MPATH
747 		/*
748 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
749 		 * we require users to specify a matching RTAX_GATEWAY.
750 		 *
751 		 * for RTM_GET, gate is optional even with multipath.
752 		 * if gate == NULL the first match is returned.
753 		 * (no need to call rt_mpath_matchgate if gate == NULL)
754 		 */
755 		if (rt_mpath_capable(rnh) &&
756 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
757 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
758 			if (!rt) {
759 				RIB_RUNLOCK(rnh);
760 				senderr(ESRCH);
761 			}
762 		}
763 #endif
764 		/*
765 		 * If performing proxied L2 entry insertion, and
766 		 * the actual PPP host entry is found, perform
767 		 * another search to retrieve the prefix route of
768 		 * the local end point of the PPP link.
769 		 */
770 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
771 			struct sockaddr laddr;
772 
773 			if (rt->rt_ifp != NULL &&
774 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
775 				struct ifaddr *ifa;
776 
777 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1,
778 						RT_ALL_FIBS);
779 				if (ifa != NULL)
780 					rt_maskedcopy(ifa->ifa_addr,
781 						      &laddr,
782 						      ifa->ifa_netmask);
783 			} else
784 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
785 					      &laddr,
786 					      rt->rt_ifa->ifa_netmask);
787 			/*
788 			 * refactor rt and no lock operation necessary
789 			 */
790 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
791 			    &rnh->head);
792 			if (rt == NULL) {
793 				RIB_RUNLOCK(rnh);
794 				senderr(ESRCH);
795 			}
796 		}
797 		RT_LOCK(rt);
798 		RT_ADDREF(rt);
799 		RIB_RUNLOCK(rnh);
800 
801 report:
802 		RT_LOCK_ASSERT(rt);
803 		if ((rt->rt_flags & RTF_HOST) == 0
804 		    ? jailed_without_vnet(curthread->td_ucred)
805 		    : prison_if(curthread->td_ucred,
806 		    rt_key(rt)) != 0) {
807 			RT_UNLOCK(rt);
808 			senderr(ESRCH);
809 		}
810 		info.rti_info[RTAX_DST] = rt_key(rt);
811 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
812 		info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
813 		    rt_mask(rt), &ss);
814 		info.rti_info[RTAX_GENMASK] = 0;
815 		if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
816 			ifp = rt->rt_ifp;
817 			if (ifp) {
818 				info.rti_info[RTAX_IFP] =
819 				    ifp->if_addr->ifa_addr;
820 				error = rtm_get_jailed(&info, ifp, rt,
821 				    &saun, curthread->td_ucred);
822 				if (error != 0) {
823 					RT_UNLOCK(rt);
824 					senderr(error);
825 				}
826 				if (ifp->if_flags & IFF_POINTOPOINT)
827 					info.rti_info[RTAX_BRD] =
828 					    rt->rt_ifa->ifa_dstaddr;
829 				rtm->rtm_index = ifp->if_index;
830 			} else {
831 				info.rti_info[RTAX_IFP] = NULL;
832 				info.rti_info[RTAX_IFA] = NULL;
833 			}
834 		} else if ((ifp = rt->rt_ifp) != NULL) {
835 			rtm->rtm_index = ifp->if_index;
836 		}
837 
838 		/* Check if we need to realloc storage */
839 		rtsock_msg_buffer(rtm->rtm_type, &info, NULL, &len);
840 		if (len > alloc_len) {
841 			struct rt_msghdr *new_rtm;
842 			new_rtm = malloc(len, M_TEMP, M_NOWAIT);
843 			if (new_rtm == NULL) {
844 				RT_UNLOCK(rt);
845 				senderr(ENOBUFS);
846 			}
847 			bcopy(rtm, new_rtm, rtm->rtm_msglen);
848 			free(rtm, M_TEMP);
849 			rtm = new_rtm;
850 			alloc_len = len;
851 		}
852 
853 		w.w_tmem = (caddr_t)rtm;
854 		w.w_tmemsize = alloc_len;
855 		rtsock_msg_buffer(rtm->rtm_type, &info, &w, &len);
856 
857 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
858 			rtm->rtm_flags = RTF_GATEWAY |
859 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
860 		else
861 			rtm->rtm_flags = rt->rt_flags;
862 		rt_getmetrics(rt, &rtm->rtm_rmx);
863 		rtm->rtm_addrs = info.rti_addrs;
864 
865 		RT_UNLOCK(rt);
866 		break;
867 
868 	default:
869 		senderr(EOPNOTSUPP);
870 	}
871 
872 flush:
873 	if (rt != NULL)
874 		RTFREE(rt);
875 	/*
876 	 * Check to see if we don't want our own messages.
877 	 */
878 	if ((so->so_options & SO_USELOOPBACK) == 0) {
879 		if (V_route_cb.any_count <= 1) {
880 			if (rtm != NULL)
881 				free(rtm, M_TEMP);
882 			m_freem(m);
883 			return (error);
884 		}
885 		/* There is another listener, so construct message */
886 		rp = sotorawcb(so);
887 	}
888 
889 	if (rtm != NULL) {
890 #ifdef INET6
891 		if (rti_need_deembed) {
892 			/* sin6_scope_id is recovered before sending rtm. */
893 			sin6 = (struct sockaddr_in6 *)&ss;
894 			for (i = 0; i < RTAX_MAX; i++) {
895 				if (info.rti_info[i] == NULL)
896 					continue;
897 				if (info.rti_info[i]->sa_family != AF_INET6)
898 					continue;
899 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
900 				if (sa6_recoverscope(sin6) == 0)
901 					bcopy(sin6, info.rti_info[i],
902 						    sizeof(*sin6));
903 			}
904 		}
905 #endif
906 		if (error != 0)
907 			rtm->rtm_errno = error;
908 		else
909 			rtm->rtm_flags |= RTF_DONE;
910 
911 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
912 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
913 			m_freem(m);
914 			m = NULL;
915 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
916 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
917 
918 		free(rtm, M_TEMP);
919 	}
920 	if (m != NULL) {
921 		M_SETFIB(m, fibnum);
922 		m->m_flags |= RTS_FILTER_FIB;
923 		if (rp) {
924 			/*
925 			 * XXX insure we don't get a copy by
926 			 * invalidating our protocol
927 			 */
928 			unsigned short family = rp->rcb_proto.sp_family;
929 			rp->rcb_proto.sp_family = 0;
930 			rt_dispatch(m, saf);
931 			rp->rcb_proto.sp_family = family;
932 		} else
933 			rt_dispatch(m, saf);
934 	}
935 
936 	return (error);
937 }
938 
939 static void
940 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
941 {
942 
943 	bzero(out, sizeof(*out));
944 	out->rmx_mtu = rt->rt_mtu;
945 	out->rmx_weight = rt->rt_weight;
946 	out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
947 	/* Kernel -> userland timebase conversion. */
948 	out->rmx_expire = rt->rt_expire ?
949 	    rt->rt_expire - time_uptime + time_second : 0;
950 }
951 
952 /*
953  * Extract the addresses of the passed sockaddrs.
954  * Do a little sanity checking so as to avoid bad memory references.
955  * This data is derived straight from userland.
956  */
957 static int
958 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
959 {
960 	struct sockaddr *sa;
961 	int i;
962 
963 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
964 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
965 			continue;
966 		sa = (struct sockaddr *)cp;
967 		/*
968 		 * It won't fit.
969 		 */
970 		if (cp + sa->sa_len > cplim)
971 			return (EINVAL);
972 		/*
973 		 * there are no more.. quit now
974 		 * If there are more bits, they are in error.
975 		 * I've seen this. route(1) can evidently generate these.
976 		 * This causes kernel to core dump.
977 		 * for compatibility, If we see this, point to a safe address.
978 		 */
979 		if (sa->sa_len == 0) {
980 			rtinfo->rti_info[i] = &sa_zero;
981 			return (0); /* should be EINVAL but for compat */
982 		}
983 		/* accept it */
984 #ifdef INET6
985 		if (sa->sa_family == AF_INET6)
986 			sa6_embedscope((struct sockaddr_in6 *)sa,
987 			    V_ip6_use_defzone);
988 #endif
989 		rtinfo->rti_info[i] = sa;
990 		cp += SA_SIZE(sa);
991 	}
992 	return (0);
993 }
994 
995 /*
996  * Fill in @dmask with valid netmask leaving original @smask
997  * intact. Mostly used with radix netmasks.
998  */
999 static struct sockaddr *
1000 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
1001     struct sockaddr_storage *dmask)
1002 {
1003 	if (dst == NULL || smask == NULL)
1004 		return (NULL);
1005 
1006 	memset(dmask, 0, dst->sa_len);
1007 	memcpy(dmask, smask, smask->sa_len);
1008 	dmask->ss_len = dst->sa_len;
1009 	dmask->ss_family = dst->sa_family;
1010 
1011 	return ((struct sockaddr *)dmask);
1012 }
1013 
1014 /*
1015  * Writes information related to @rtinfo object to newly-allocated mbuf.
1016  * Assumes MCLBYTES is enough to construct any message.
1017  * Used for OS notifications of vaious events (if/ifa announces,etc)
1018  *
1019  * Returns allocated mbuf or NULL on failure.
1020  */
1021 static struct mbuf *
1022 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1023 {
1024 	struct rt_msghdr *rtm;
1025 	struct mbuf *m;
1026 	int i;
1027 	struct sockaddr *sa;
1028 #ifdef INET6
1029 	struct sockaddr_storage ss;
1030 	struct sockaddr_in6 *sin6;
1031 #endif
1032 	int len, dlen;
1033 
1034 	switch (type) {
1035 
1036 	case RTM_DELADDR:
1037 	case RTM_NEWADDR:
1038 		len = sizeof(struct ifa_msghdr);
1039 		break;
1040 
1041 	case RTM_DELMADDR:
1042 	case RTM_NEWMADDR:
1043 		len = sizeof(struct ifma_msghdr);
1044 		break;
1045 
1046 	case RTM_IFINFO:
1047 		len = sizeof(struct if_msghdr);
1048 		break;
1049 
1050 	case RTM_IFANNOUNCE:
1051 	case RTM_IEEE80211:
1052 		len = sizeof(struct if_announcemsghdr);
1053 		break;
1054 
1055 	default:
1056 		len = sizeof(struct rt_msghdr);
1057 	}
1058 
1059 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1060 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1061 	if (len > MHLEN)
1062 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1063 	else
1064 		m = m_gethdr(M_NOWAIT, MT_DATA);
1065 	if (m == NULL)
1066 		return (m);
1067 
1068 	m->m_pkthdr.len = m->m_len = len;
1069 	rtm = mtod(m, struct rt_msghdr *);
1070 	bzero((caddr_t)rtm, len);
1071 	for (i = 0; i < RTAX_MAX; i++) {
1072 		if ((sa = rtinfo->rti_info[i]) == NULL)
1073 			continue;
1074 		rtinfo->rti_addrs |= (1 << i);
1075 		dlen = SA_SIZE(sa);
1076 #ifdef INET6
1077 		if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1078 			sin6 = (struct sockaddr_in6 *)&ss;
1079 			bcopy(sa, sin6, sizeof(*sin6));
1080 			if (sa6_recoverscope(sin6) == 0)
1081 				sa = (struct sockaddr *)sin6;
1082 		}
1083 #endif
1084 		m_copyback(m, len, dlen, (caddr_t)sa);
1085 		len += dlen;
1086 	}
1087 	if (m->m_pkthdr.len != len) {
1088 		m_freem(m);
1089 		return (NULL);
1090 	}
1091 	rtm->rtm_msglen = len;
1092 	rtm->rtm_version = RTM_VERSION;
1093 	rtm->rtm_type = type;
1094 	return (m);
1095 }
1096 
1097 /*
1098  * Writes information related to @rtinfo object to preallocated buffer.
1099  * Stores needed size in @plen. If @w is NULL, calculates size without
1100  * writing.
1101  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1102  *
1103  * Returns 0 on success.
1104  *
1105  */
1106 static int
1107 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1108 {
1109 	int i;
1110 	int len, buflen = 0, dlen;
1111 	caddr_t cp = NULL;
1112 	struct rt_msghdr *rtm = NULL;
1113 #ifdef INET6
1114 	struct sockaddr_storage ss;
1115 	struct sockaddr_in6 *sin6;
1116 #endif
1117 
1118 	switch (type) {
1119 
1120 	case RTM_DELADDR:
1121 	case RTM_NEWADDR:
1122 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1123 #ifdef COMPAT_FREEBSD32
1124 			if (w->w_req->flags & SCTL_MASK32)
1125 				len = sizeof(struct ifa_msghdrl32);
1126 			else
1127 #endif
1128 				len = sizeof(struct ifa_msghdrl);
1129 		} else
1130 			len = sizeof(struct ifa_msghdr);
1131 		break;
1132 
1133 	case RTM_IFINFO:
1134 #ifdef COMPAT_FREEBSD32
1135 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1136 			if (w->w_op == NET_RT_IFLISTL)
1137 				len = sizeof(struct if_msghdrl32);
1138 			else
1139 				len = sizeof(struct if_msghdr32);
1140 			break;
1141 		}
1142 #endif
1143 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1144 			len = sizeof(struct if_msghdrl);
1145 		else
1146 			len = sizeof(struct if_msghdr);
1147 		break;
1148 
1149 	case RTM_NEWMADDR:
1150 		len = sizeof(struct ifma_msghdr);
1151 		break;
1152 
1153 	default:
1154 		len = sizeof(struct rt_msghdr);
1155 	}
1156 
1157 	if (w != NULL) {
1158 		rtm = (struct rt_msghdr *)w->w_tmem;
1159 		buflen = w->w_tmemsize - len;
1160 		cp = (caddr_t)w->w_tmem + len;
1161 	}
1162 
1163 	rtinfo->rti_addrs = 0;
1164 	for (i = 0; i < RTAX_MAX; i++) {
1165 		struct sockaddr *sa;
1166 
1167 		if ((sa = rtinfo->rti_info[i]) == NULL)
1168 			continue;
1169 		rtinfo->rti_addrs |= (1 << i);
1170 		dlen = SA_SIZE(sa);
1171 		if (cp != NULL && buflen >= dlen) {
1172 #ifdef INET6
1173 			if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1174 				sin6 = (struct sockaddr_in6 *)&ss;
1175 				bcopy(sa, sin6, sizeof(*sin6));
1176 				if (sa6_recoverscope(sin6) == 0)
1177 					sa = (struct sockaddr *)sin6;
1178 			}
1179 #endif
1180 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1181 			cp += dlen;
1182 			buflen -= dlen;
1183 		} else if (cp != NULL) {
1184 			/*
1185 			 * Buffer too small. Count needed size
1186 			 * and return with error.
1187 			 */
1188 			cp = NULL;
1189 		}
1190 
1191 		len += dlen;
1192 	}
1193 
1194 	if (cp != NULL) {
1195 		dlen = ALIGN(len) - len;
1196 		if (buflen < dlen)
1197 			cp = NULL;
1198 		else
1199 			buflen -= dlen;
1200 	}
1201 	len = ALIGN(len);
1202 
1203 	if (cp != NULL) {
1204 		/* fill header iff buffer is large enough */
1205 		rtm->rtm_version = RTM_VERSION;
1206 		rtm->rtm_type = type;
1207 		rtm->rtm_msglen = len;
1208 	}
1209 
1210 	*plen = len;
1211 
1212 	if (w != NULL && cp == NULL)
1213 		return (ENOBUFS);
1214 
1215 	return (0);
1216 }
1217 
1218 /*
1219  * This routine is called to generate a message from the routing
1220  * socket indicating that a redirect has occurred, a routing lookup
1221  * has failed, or that a protocol has detected timeouts to a particular
1222  * destination.
1223  */
1224 void
1225 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1226     int fibnum)
1227 {
1228 	struct rt_msghdr *rtm;
1229 	struct mbuf *m;
1230 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1231 
1232 	if (V_route_cb.any_count == 0)
1233 		return;
1234 	m = rtsock_msg_mbuf(type, rtinfo);
1235 	if (m == NULL)
1236 		return;
1237 
1238 	if (fibnum != RT_ALL_FIBS) {
1239 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1240 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1241 		M_SETFIB(m, fibnum);
1242 		m->m_flags |= RTS_FILTER_FIB;
1243 	}
1244 
1245 	rtm = mtod(m, struct rt_msghdr *);
1246 	rtm->rtm_flags = RTF_DONE | flags;
1247 	rtm->rtm_errno = error;
1248 	rtm->rtm_addrs = rtinfo->rti_addrs;
1249 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1250 }
1251 
1252 void
1253 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1254 {
1255 
1256 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1257 }
1258 
1259 /*
1260  * This routine is called to generate a message from the routing
1261  * socket indicating that the status of a network interface has changed.
1262  */
1263 void
1264 rt_ifmsg(struct ifnet *ifp)
1265 {
1266 	struct if_msghdr *ifm;
1267 	struct mbuf *m;
1268 	struct rt_addrinfo info;
1269 
1270 	if (V_route_cb.any_count == 0)
1271 		return;
1272 	bzero((caddr_t)&info, sizeof(info));
1273 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1274 	if (m == NULL)
1275 		return;
1276 	ifm = mtod(m, struct if_msghdr *);
1277 	ifm->ifm_index = ifp->if_index;
1278 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1279 	if_data_copy(ifp, &ifm->ifm_data);
1280 	ifm->ifm_addrs = 0;
1281 	rt_dispatch(m, AF_UNSPEC);
1282 }
1283 
1284 /*
1285  * Announce interface address arrival/withdraw.
1286  * Please do not call directly, use rt_addrmsg().
1287  * Assume input data to be valid.
1288  * Returns 0 on success.
1289  */
1290 int
1291 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1292 {
1293 	struct rt_addrinfo info;
1294 	struct sockaddr *sa;
1295 	int ncmd;
1296 	struct mbuf *m;
1297 	struct ifa_msghdr *ifam;
1298 	struct ifnet *ifp = ifa->ifa_ifp;
1299 	struct sockaddr_storage ss;
1300 
1301 	if (V_route_cb.any_count == 0)
1302 		return (0);
1303 
1304 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1305 
1306 	bzero((caddr_t)&info, sizeof(info));
1307 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1308 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1309 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1310 	    info.rti_info[RTAX_IFP], ifa->ifa_netmask, &ss);
1311 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1312 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1313 		return (ENOBUFS);
1314 	ifam = mtod(m, struct ifa_msghdr *);
1315 	ifam->ifam_index = ifp->if_index;
1316 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1317 	ifam->ifam_flags = ifa->ifa_flags;
1318 	ifam->ifam_addrs = info.rti_addrs;
1319 
1320 	if (fibnum != RT_ALL_FIBS) {
1321 		M_SETFIB(m, fibnum);
1322 		m->m_flags |= RTS_FILTER_FIB;
1323 	}
1324 
1325 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1326 
1327 	return (0);
1328 }
1329 
1330 /*
1331  * Announce route addition/removal.
1332  * Please do not call directly, use rt_routemsg().
1333  * Note that @rt data MAY be inconsistent/invalid:
1334  * if some userland app sends us "invalid" route message (invalid mask,
1335  * no dst, wrong address families, etc...) we need to pass it back
1336  * to app (and any other rtsock consumers) with rtm_errno field set to
1337  * non-zero value.
1338  *
1339  * Returns 0 on success.
1340  */
1341 int
1342 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1343     int fibnum)
1344 {
1345 	struct rt_addrinfo info;
1346 	struct sockaddr *sa;
1347 	struct mbuf *m;
1348 	struct rt_msghdr *rtm;
1349 	struct sockaddr_storage ss;
1350 
1351 	if (V_route_cb.any_count == 0)
1352 		return (0);
1353 
1354 	bzero((caddr_t)&info, sizeof(info));
1355 	info.rti_info[RTAX_DST] = sa = rt_key(rt);
1356 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(sa, rt_mask(rt), &ss);
1357 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1358 	if ((m = rtsock_msg_mbuf(cmd, &info)) == NULL)
1359 		return (ENOBUFS);
1360 	rtm = mtod(m, struct rt_msghdr *);
1361 	rtm->rtm_index = ifp->if_index;
1362 	rtm->rtm_flags |= rt->rt_flags;
1363 	rtm->rtm_errno = error;
1364 	rtm->rtm_addrs = info.rti_addrs;
1365 
1366 	if (fibnum != RT_ALL_FIBS) {
1367 		M_SETFIB(m, fibnum);
1368 		m->m_flags |= RTS_FILTER_FIB;
1369 	}
1370 
1371 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1372 
1373 	return (0);
1374 }
1375 
1376 /*
1377  * This is the analogue to the rt_newaddrmsg which performs the same
1378  * function but for multicast group memberhips.  This is easier since
1379  * there is no route state to worry about.
1380  */
1381 void
1382 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1383 {
1384 	struct rt_addrinfo info;
1385 	struct mbuf *m = NULL;
1386 	struct ifnet *ifp = ifma->ifma_ifp;
1387 	struct ifma_msghdr *ifmam;
1388 
1389 	if (V_route_cb.any_count == 0)
1390 		return;
1391 
1392 	bzero((caddr_t)&info, sizeof(info));
1393 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1394 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1395 	/*
1396 	 * If a link-layer address is present, present it as a ``gateway''
1397 	 * (similarly to how ARP entries, e.g., are presented).
1398 	 */
1399 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1400 	m = rtsock_msg_mbuf(cmd, &info);
1401 	if (m == NULL)
1402 		return;
1403 	ifmam = mtod(m, struct ifma_msghdr *);
1404 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1405 	    __func__));
1406 	ifmam->ifmam_index = ifp->if_index;
1407 	ifmam->ifmam_addrs = info.rti_addrs;
1408 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1409 }
1410 
1411 static struct mbuf *
1412 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1413 	struct rt_addrinfo *info)
1414 {
1415 	struct if_announcemsghdr *ifan;
1416 	struct mbuf *m;
1417 
1418 	if (V_route_cb.any_count == 0)
1419 		return NULL;
1420 	bzero((caddr_t)info, sizeof(*info));
1421 	m = rtsock_msg_mbuf(type, info);
1422 	if (m != NULL) {
1423 		ifan = mtod(m, struct if_announcemsghdr *);
1424 		ifan->ifan_index = ifp->if_index;
1425 		strlcpy(ifan->ifan_name, ifp->if_xname,
1426 			sizeof(ifan->ifan_name));
1427 		ifan->ifan_what = what;
1428 	}
1429 	return m;
1430 }
1431 
1432 /*
1433  * This is called to generate routing socket messages indicating
1434  * IEEE80211 wireless events.
1435  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1436  */
1437 void
1438 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1439 {
1440 	struct mbuf *m;
1441 	struct rt_addrinfo info;
1442 
1443 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1444 	if (m != NULL) {
1445 		/*
1446 		 * Append the ieee80211 data.  Try to stick it in the
1447 		 * mbuf containing the ifannounce msg; otherwise allocate
1448 		 * a new mbuf and append.
1449 		 *
1450 		 * NB: we assume m is a single mbuf.
1451 		 */
1452 		if (data_len > M_TRAILINGSPACE(m)) {
1453 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1454 			if (n == NULL) {
1455 				m_freem(m);
1456 				return;
1457 			}
1458 			bcopy(data, mtod(n, void *), data_len);
1459 			n->m_len = data_len;
1460 			m->m_next = n;
1461 		} else if (data_len > 0) {
1462 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1463 			m->m_len += data_len;
1464 		}
1465 		if (m->m_flags & M_PKTHDR)
1466 			m->m_pkthdr.len += data_len;
1467 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1468 		rt_dispatch(m, AF_UNSPEC);
1469 	}
1470 }
1471 
1472 /*
1473  * This is called to generate routing socket messages indicating
1474  * network interface arrival and departure.
1475  */
1476 void
1477 rt_ifannouncemsg(struct ifnet *ifp, int what)
1478 {
1479 	struct mbuf *m;
1480 	struct rt_addrinfo info;
1481 
1482 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1483 	if (m != NULL)
1484 		rt_dispatch(m, AF_UNSPEC);
1485 }
1486 
1487 static void
1488 rt_dispatch(struct mbuf *m, sa_family_t saf)
1489 {
1490 	struct m_tag *tag;
1491 
1492 	/*
1493 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1494 	 * use when injecting the mbuf into the routing socket buffer from
1495 	 * the netisr.
1496 	 */
1497 	if (saf != AF_UNSPEC) {
1498 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1499 		    M_NOWAIT);
1500 		if (tag == NULL) {
1501 			m_freem(m);
1502 			return;
1503 		}
1504 		*(unsigned short *)(tag + 1) = saf;
1505 		m_tag_prepend(m, tag);
1506 	}
1507 #ifdef VIMAGE
1508 	if (V_loif)
1509 		m->m_pkthdr.rcvif = V_loif;
1510 	else {
1511 		m_freem(m);
1512 		return;
1513 	}
1514 #endif
1515 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1516 }
1517 
1518 /*
1519  * This is used in dumping the kernel table via sysctl().
1520  */
1521 static int
1522 sysctl_dumpentry(struct radix_node *rn, void *vw)
1523 {
1524 	struct walkarg *w = vw;
1525 	struct rtentry *rt = (struct rtentry *)rn;
1526 	int error = 0, size;
1527 	struct rt_addrinfo info;
1528 	struct sockaddr_storage ss;
1529 
1530 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1531 		return 0;
1532 	if ((rt->rt_flags & RTF_HOST) == 0
1533 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1534 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1535 		return (0);
1536 	bzero((caddr_t)&info, sizeof(info));
1537 	info.rti_info[RTAX_DST] = rt_key(rt);
1538 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1539 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1540 	    rt_mask(rt), &ss);
1541 	info.rti_info[RTAX_GENMASK] = 0;
1542 	if (rt->rt_ifp) {
1543 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1544 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1545 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1546 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1547 	}
1548 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1549 		return (error);
1550 	if (w->w_req && w->w_tmem) {
1551 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1552 
1553 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1554 			rtm->rtm_flags = RTF_GATEWAY |
1555 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1556 		else
1557 			rtm->rtm_flags = rt->rt_flags;
1558 		rt_getmetrics(rt, &rtm->rtm_rmx);
1559 		rtm->rtm_index = rt->rt_ifp->if_index;
1560 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1561 		rtm->rtm_addrs = info.rti_addrs;
1562 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1563 		return (error);
1564 	}
1565 	return (error);
1566 }
1567 
1568 static int
1569 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1570     struct rt_addrinfo *info, struct walkarg *w, int len)
1571 {
1572 	struct if_msghdrl *ifm;
1573 	struct if_data *ifd;
1574 
1575 	ifm = (struct if_msghdrl *)w->w_tmem;
1576 
1577 #ifdef COMPAT_FREEBSD32
1578 	if (w->w_req->flags & SCTL_MASK32) {
1579 		struct if_msghdrl32 *ifm32;
1580 
1581 		ifm32 = (struct if_msghdrl32 *)ifm;
1582 		ifm32->ifm_addrs = info->rti_addrs;
1583 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1584 		ifm32->ifm_index = ifp->if_index;
1585 		ifm32->_ifm_spare1 = 0;
1586 		ifm32->ifm_len = sizeof(*ifm32);
1587 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1588 		ifd = &ifm32->ifm_data;
1589 	} else
1590 #endif
1591 	{
1592 		ifm->ifm_addrs = info->rti_addrs;
1593 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1594 		ifm->ifm_index = ifp->if_index;
1595 		ifm->_ifm_spare1 = 0;
1596 		ifm->ifm_len = sizeof(*ifm);
1597 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1598 		ifd = &ifm->ifm_data;
1599 	}
1600 
1601 	memcpy(ifd, src_ifd, sizeof(*ifd));
1602 
1603 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1604 }
1605 
1606 static int
1607 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1608     struct rt_addrinfo *info, struct walkarg *w, int len)
1609 {
1610 	struct if_msghdr *ifm;
1611 	struct if_data *ifd;
1612 
1613 	ifm = (struct if_msghdr *)w->w_tmem;
1614 
1615 #ifdef COMPAT_FREEBSD32
1616 	if (w->w_req->flags & SCTL_MASK32) {
1617 		struct if_msghdr32 *ifm32;
1618 
1619 		ifm32 = (struct if_msghdr32 *)ifm;
1620 		ifm32->ifm_addrs = info->rti_addrs;
1621 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1622 		ifm32->ifm_index = ifp->if_index;
1623 		ifd = &ifm32->ifm_data;
1624 	} else
1625 #endif
1626 	{
1627 		ifm->ifm_addrs = info->rti_addrs;
1628 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1629 		ifm->ifm_index = ifp->if_index;
1630 		ifd = &ifm->ifm_data;
1631 	}
1632 
1633 	memcpy(ifd, src_ifd, sizeof(*ifd));
1634 
1635 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1636 }
1637 
1638 static int
1639 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1640     struct walkarg *w, int len)
1641 {
1642 	struct ifa_msghdrl *ifam;
1643 	struct if_data *ifd;
1644 
1645 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1646 
1647 #ifdef COMPAT_FREEBSD32
1648 	if (w->w_req->flags & SCTL_MASK32) {
1649 		struct ifa_msghdrl32 *ifam32;
1650 
1651 		ifam32 = (struct ifa_msghdrl32 *)ifam;
1652 		ifam32->ifam_addrs = info->rti_addrs;
1653 		ifam32->ifam_flags = ifa->ifa_flags;
1654 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1655 		ifam32->_ifam_spare1 = 0;
1656 		ifam32->ifam_len = sizeof(*ifam32);
1657 		ifam32->ifam_data_off =
1658 		    offsetof(struct ifa_msghdrl32, ifam_data);
1659 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1660 		ifd = &ifam32->ifam_data;
1661 	} else
1662 #endif
1663 	{
1664 		ifam->ifam_addrs = info->rti_addrs;
1665 		ifam->ifam_flags = ifa->ifa_flags;
1666 		ifam->ifam_index = ifa->ifa_ifp->if_index;
1667 		ifam->_ifam_spare1 = 0;
1668 		ifam->ifam_len = sizeof(*ifam);
1669 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1670 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1671 		ifd = &ifam->ifam_data;
1672 	}
1673 
1674 	bzero(ifd, sizeof(*ifd));
1675 	ifd->ifi_datalen = sizeof(struct if_data);
1676 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1677 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1678 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1679 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1680 
1681 	/* Fixup if_data carp(4) vhid. */
1682 	if (carp_get_vhid_p != NULL)
1683 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1684 
1685 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1686 }
1687 
1688 static int
1689 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1690     struct walkarg *w, int len)
1691 {
1692 	struct ifa_msghdr *ifam;
1693 
1694 	ifam = (struct ifa_msghdr *)w->w_tmem;
1695 	ifam->ifam_addrs = info->rti_addrs;
1696 	ifam->ifam_flags = ifa->ifa_flags;
1697 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1698 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1699 
1700 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1701 }
1702 
1703 static int
1704 sysctl_iflist(int af, struct walkarg *w)
1705 {
1706 	struct ifnet *ifp;
1707 	struct ifaddr *ifa;
1708 	struct if_data ifd;
1709 	struct rt_addrinfo info;
1710 	int len, error = 0;
1711 	struct sockaddr_storage ss;
1712 
1713 	bzero((caddr_t)&info, sizeof(info));
1714 	bzero(&ifd, sizeof(ifd));
1715 	IFNET_RLOCK_NOSLEEP();
1716 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1717 		if (w->w_arg && w->w_arg != ifp->if_index)
1718 			continue;
1719 		if_data_copy(ifp, &ifd);
1720 		IF_ADDR_RLOCK(ifp);
1721 		ifa = ifp->if_addr;
1722 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1723 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1724 		if (error != 0)
1725 			goto done;
1726 		info.rti_info[RTAX_IFP] = NULL;
1727 		if (w->w_req && w->w_tmem) {
1728 			if (w->w_op == NET_RT_IFLISTL)
1729 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1730 				    len);
1731 			else
1732 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1733 				    len);
1734 			if (error)
1735 				goto done;
1736 		}
1737 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1738 			if (af && af != ifa->ifa_addr->sa_family)
1739 				continue;
1740 			if (prison_if(w->w_req->td->td_ucred,
1741 			    ifa->ifa_addr) != 0)
1742 				continue;
1743 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1744 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1745 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
1746 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1747 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1748 			if (error != 0)
1749 				goto done;
1750 			if (w->w_req && w->w_tmem) {
1751 				if (w->w_op == NET_RT_IFLISTL)
1752 					error = sysctl_iflist_ifaml(ifa, &info,
1753 					    w, len);
1754 				else
1755 					error = sysctl_iflist_ifam(ifa, &info,
1756 					    w, len);
1757 				if (error)
1758 					goto done;
1759 			}
1760 		}
1761 		IF_ADDR_RUNLOCK(ifp);
1762 		info.rti_info[RTAX_IFA] = NULL;
1763 		info.rti_info[RTAX_NETMASK] = NULL;
1764 		info.rti_info[RTAX_BRD] = NULL;
1765 	}
1766 done:
1767 	if (ifp != NULL)
1768 		IF_ADDR_RUNLOCK(ifp);
1769 	IFNET_RUNLOCK_NOSLEEP();
1770 	return (error);
1771 }
1772 
1773 static int
1774 sysctl_ifmalist(int af, struct walkarg *w)
1775 {
1776 	struct rt_addrinfo info;
1777 	struct ifaddr *ifa;
1778 	struct ifmultiaddr *ifma;
1779 	struct ifnet *ifp;
1780 	int error, len;
1781 
1782 	error = 0;
1783 	bzero((caddr_t)&info, sizeof(info));
1784 
1785 	IFNET_RLOCK_NOSLEEP();
1786 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1787 		if (w->w_arg && w->w_arg != ifp->if_index)
1788 			continue;
1789 		ifa = ifp->if_addr;
1790 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1791 		IF_ADDR_RLOCK(ifp);
1792 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1793 			if (af && af != ifma->ifma_addr->sa_family)
1794 				continue;
1795 			if (prison_if(w->w_req->td->td_ucred,
1796 			    ifma->ifma_addr) != 0)
1797 				continue;
1798 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1799 			info.rti_info[RTAX_GATEWAY] =
1800 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1801 			    ifma->ifma_lladdr : NULL;
1802 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1803 			if (error != 0)
1804 				break;
1805 			if (w->w_req && w->w_tmem) {
1806 				struct ifma_msghdr *ifmam;
1807 
1808 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1809 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1810 				ifmam->ifmam_flags = 0;
1811 				ifmam->ifmam_addrs = info.rti_addrs;
1812 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1813 				if (error != 0)
1814 					break;
1815 			}
1816 		}
1817 		IF_ADDR_RUNLOCK(ifp);
1818 		if (error != 0)
1819 			break;
1820 	}
1821 	IFNET_RUNLOCK_NOSLEEP();
1822 	return (error);
1823 }
1824 
1825 static int
1826 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1827 {
1828 	int	*name = (int *)arg1;
1829 	u_int	namelen = arg2;
1830 	struct rib_head *rnh = NULL; /* silence compiler. */
1831 	int	i, lim, error = EINVAL;
1832 	int	fib = 0;
1833 	u_char	af;
1834 	struct	walkarg w;
1835 
1836 	name ++;
1837 	namelen--;
1838 	if (req->newptr)
1839 		return (EPERM);
1840 	if (name[1] == NET_RT_DUMP) {
1841 		if (namelen == 3)
1842 			fib = req->td->td_proc->p_fibnum;
1843 		else if (namelen == 4)
1844 			fib = (name[3] == RT_ALL_FIBS) ?
1845 			    req->td->td_proc->p_fibnum : name[3];
1846 		else
1847 			return ((namelen < 3) ? EISDIR : ENOTDIR);
1848 		if (fib < 0 || fib >= rt_numfibs)
1849 			return (EINVAL);
1850 	} else if (namelen != 3)
1851 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1852 	af = name[0];
1853 	if (af > AF_MAX)
1854 		return (EINVAL);
1855 	bzero(&w, sizeof(w));
1856 	w.w_op = name[1];
1857 	w.w_arg = name[2];
1858 	w.w_req = req;
1859 
1860 	error = sysctl_wire_old_buffer(req, 0);
1861 	if (error)
1862 		return (error);
1863 
1864 	/*
1865 	 * Allocate reply buffer in advance.
1866 	 * All rtsock messages has maximum length of u_short.
1867 	 */
1868 	w.w_tmemsize = 65536;
1869 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
1870 
1871 	switch (w.w_op) {
1872 
1873 	case NET_RT_DUMP:
1874 	case NET_RT_FLAGS:
1875 		if (af == 0) {			/* dump all tables */
1876 			i = 1;
1877 			lim = AF_MAX;
1878 		} else				/* dump only one table */
1879 			i = lim = af;
1880 
1881 		/*
1882 		 * take care of llinfo entries, the caller must
1883 		 * specify an AF
1884 		 */
1885 		if (w.w_op == NET_RT_FLAGS &&
1886 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1887 			if (af != 0)
1888 				error = lltable_sysctl_dumparp(af, w.w_req);
1889 			else
1890 				error = EINVAL;
1891 			break;
1892 		}
1893 		/*
1894 		 * take care of routing entries
1895 		 */
1896 		for (error = 0; error == 0 && i <= lim; i++) {
1897 			rnh = rt_tables_get_rnh(fib, i);
1898 			if (rnh != NULL) {
1899 				RIB_RLOCK(rnh);
1900 			    	error = rnh->rnh_walktree(&rnh->head,
1901 				    sysctl_dumpentry, &w);
1902 				RIB_RUNLOCK(rnh);
1903 			} else if (af != 0)
1904 				error = EAFNOSUPPORT;
1905 		}
1906 		break;
1907 
1908 	case NET_RT_IFLIST:
1909 	case NET_RT_IFLISTL:
1910 		error = sysctl_iflist(af, &w);
1911 		break;
1912 
1913 	case NET_RT_IFMALIST:
1914 		error = sysctl_ifmalist(af, &w);
1915 		break;
1916 	}
1917 
1918 	free(w.w_tmem, M_TEMP);
1919 	return (error);
1920 }
1921 
1922 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1923 
1924 /*
1925  * Definitions of protocols supported in the ROUTE domain.
1926  */
1927 
1928 static struct domain routedomain;		/* or at least forward */
1929 
1930 static struct protosw routesw[] = {
1931 {
1932 	.pr_type =		SOCK_RAW,
1933 	.pr_domain =		&routedomain,
1934 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1935 	.pr_output =		route_output,
1936 	.pr_ctlinput =		raw_ctlinput,
1937 	.pr_init =		raw_init,
1938 	.pr_usrreqs =		&route_usrreqs
1939 }
1940 };
1941 
1942 static struct domain routedomain = {
1943 	.dom_family =		PF_ROUTE,
1944 	.dom_name =		 "route",
1945 	.dom_protosw =		routesw,
1946 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
1947 };
1948 
1949 VNET_DOMAIN_SET(route);
1950