xref: /freebsd/sys/net/rtsock.c (revision a6578a04e440f79f3b913660221caa9cde3e722c)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1988, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
32  * $FreeBSD$
33  */
34 #include "opt_mpath.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/jail.h>
40 #include <sys/kernel.h>
41 #include <sys/domain.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/rmlock.h>
49 #include <sys/rwlock.h>
50 #include <sys/signalvar.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/sysctl.h>
54 #include <sys/systm.h>
55 
56 #include <net/if.h>
57 #include <net/if_var.h>
58 #include <net/if_dl.h>
59 #include <net/if_llatbl.h>
60 #include <net/if_types.h>
61 #include <net/netisr.h>
62 #include <net/raw_cb.h>
63 #include <net/route.h>
64 #include <net/route_var.h>
65 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/if_ether.h>
69 #include <netinet/ip_carp.h>
70 #ifdef INET6
71 #include <netinet6/ip6_var.h>
72 #include <netinet6/scope6_var.h>
73 #endif
74 
75 #ifdef COMPAT_FREEBSD32
76 #include <sys/mount.h>
77 #include <compat/freebsd32/freebsd32.h>
78 
79 struct if_msghdr32 {
80 	uint16_t ifm_msglen;
81 	uint8_t	ifm_version;
82 	uint8_t	ifm_type;
83 	int32_t	ifm_addrs;
84 	int32_t	ifm_flags;
85 	uint16_t ifm_index;
86 	uint16_t _ifm_spare1;
87 	struct	if_data ifm_data;
88 };
89 
90 struct if_msghdrl32 {
91 	uint16_t ifm_msglen;
92 	uint8_t	ifm_version;
93 	uint8_t	ifm_type;
94 	int32_t	ifm_addrs;
95 	int32_t	ifm_flags;
96 	uint16_t ifm_index;
97 	uint16_t _ifm_spare1;
98 	uint16_t ifm_len;
99 	uint16_t ifm_data_off;
100 	uint32_t _ifm_spare2;
101 	struct	if_data ifm_data;
102 };
103 
104 struct ifa_msghdrl32 {
105 	uint16_t ifam_msglen;
106 	uint8_t	ifam_version;
107 	uint8_t	ifam_type;
108 	int32_t	ifam_addrs;
109 	int32_t	ifam_flags;
110 	uint16_t ifam_index;
111 	uint16_t _ifam_spare1;
112 	uint16_t ifam_len;
113 	uint16_t ifam_data_off;
114 	int32_t	ifam_metric;
115 	struct	if_data ifam_data;
116 };
117 
118 #define SA_SIZE32(sa)						\
119     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
120 	sizeof(int)		:				\
121 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
122 
123 #endif /* COMPAT_FREEBSD32 */
124 
125 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
126 
127 /* NB: these are not modified */
128 static struct	sockaddr route_src = { 2, PF_ROUTE, };
129 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
130 
131 /* These are external hooks for CARP. */
132 int	(*carp_get_vhid_p)(struct ifaddr *);
133 
134 /*
135  * Used by rtsock/raw_input callback code to decide whether to filter the update
136  * notification to a socket bound to a particular FIB.
137  */
138 #define	RTS_FILTER_FIB	M_PROTO8
139 
140 typedef struct {
141 	int	ip_count;	/* attached w/ AF_INET */
142 	int	ip6_count;	/* attached w/ AF_INET6 */
143 	int	any_count;	/* total attached */
144 } route_cb_t;
145 VNET_DEFINE_STATIC(route_cb_t, route_cb);
146 #define	V_route_cb VNET(route_cb)
147 
148 struct mtx rtsock_mtx;
149 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
150 
151 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
152 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
153 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
154 
155 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
156 
157 struct walkarg {
158 	int	w_tmemsize;
159 	int	w_op, w_arg;
160 	caddr_t	w_tmem;
161 	struct sysctl_req *w_req;
162 };
163 
164 static void	rts_input(struct mbuf *m);
165 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
166 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
167 			struct walkarg *w, int *plen);
168 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
169 			struct rt_addrinfo *rtinfo);
170 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
171 static int	sysctl_iflist(int af, struct walkarg *w);
172 static int	sysctl_ifmalist(int af, struct walkarg *w);
173 static int	route_output(struct mbuf *m, struct socket *so, ...);
174 static void	rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
175 static void	rt_dispatch(struct mbuf *, sa_family_t);
176 static struct sockaddr	*rtsock_fix_netmask(struct sockaddr *dst,
177 			struct sockaddr *smask, struct sockaddr_storage *dmask);
178 
179 static struct netisr_handler rtsock_nh = {
180 	.nh_name = "rtsock",
181 	.nh_handler = rts_input,
182 	.nh_proto = NETISR_ROUTE,
183 	.nh_policy = NETISR_POLICY_SOURCE,
184 };
185 
186 static int
187 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
188 {
189 	int error, qlimit;
190 
191 	netisr_getqlimit(&rtsock_nh, &qlimit);
192 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
193         if (error || !req->newptr)
194                 return (error);
195 	if (qlimit < 1)
196 		return (EINVAL);
197 	return (netisr_setqlimit(&rtsock_nh, qlimit));
198 }
199 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
200     0, 0, sysctl_route_netisr_maxqlen, "I",
201     "maximum routing socket dispatch queue length");
202 
203 static void
204 vnet_rts_init(void)
205 {
206 	int tmp;
207 
208 	if (IS_DEFAULT_VNET(curvnet)) {
209 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
210 			rtsock_nh.nh_qlimit = tmp;
211 		netisr_register(&rtsock_nh);
212 	}
213 #ifdef VIMAGE
214 	 else
215 		netisr_register_vnet(&rtsock_nh);
216 #endif
217 }
218 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
219     vnet_rts_init, 0);
220 
221 #ifdef VIMAGE
222 static void
223 vnet_rts_uninit(void)
224 {
225 
226 	netisr_unregister_vnet(&rtsock_nh);
227 }
228 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
229     vnet_rts_uninit, 0);
230 #endif
231 
232 static int
233 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
234     struct rawcb *rp)
235 {
236 	int fibnum;
237 
238 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
239 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
240 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
241 
242 	/* No filtering requested. */
243 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
244 		return (0);
245 
246 	/* Check if it is a rts and the fib matches the one of the socket. */
247 	fibnum = M_GETFIB(m);
248 	if (proto->sp_family != PF_ROUTE ||
249 	    rp->rcb_socket == NULL ||
250 	    rp->rcb_socket->so_fibnum == fibnum)
251 		return (0);
252 
253 	/* Filtering requested and no match, the socket shall be skipped. */
254 	return (1);
255 }
256 
257 static void
258 rts_input(struct mbuf *m)
259 {
260 	struct sockproto route_proto;
261 	unsigned short *family;
262 	struct m_tag *tag;
263 
264 	route_proto.sp_family = PF_ROUTE;
265 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
266 	if (tag != NULL) {
267 		family = (unsigned short *)(tag + 1);
268 		route_proto.sp_protocol = *family;
269 		m_tag_delete(m, tag);
270 	} else
271 		route_proto.sp_protocol = 0;
272 
273 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
274 }
275 
276 /*
277  * It really doesn't make any sense at all for this code to share much
278  * with raw_usrreq.c, since its functionality is so restricted.  XXX
279  */
280 static void
281 rts_abort(struct socket *so)
282 {
283 
284 	raw_usrreqs.pru_abort(so);
285 }
286 
287 static void
288 rts_close(struct socket *so)
289 {
290 
291 	raw_usrreqs.pru_close(so);
292 }
293 
294 /* pru_accept is EOPNOTSUPP */
295 
296 static int
297 rts_attach(struct socket *so, int proto, struct thread *td)
298 {
299 	struct rawcb *rp;
300 	int error;
301 
302 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
303 
304 	/* XXX */
305 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
306 
307 	so->so_pcb = (caddr_t)rp;
308 	so->so_fibnum = td->td_proc->p_fibnum;
309 	error = raw_attach(so, proto);
310 	rp = sotorawcb(so);
311 	if (error) {
312 		so->so_pcb = NULL;
313 		free(rp, M_PCB);
314 		return error;
315 	}
316 	RTSOCK_LOCK();
317 	switch(rp->rcb_proto.sp_protocol) {
318 	case AF_INET:
319 		V_route_cb.ip_count++;
320 		break;
321 	case AF_INET6:
322 		V_route_cb.ip6_count++;
323 		break;
324 	}
325 	V_route_cb.any_count++;
326 	RTSOCK_UNLOCK();
327 	soisconnected(so);
328 	so->so_options |= SO_USELOOPBACK;
329 	return 0;
330 }
331 
332 static int
333 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
334 {
335 
336 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
337 }
338 
339 static int
340 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
341 {
342 
343 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
344 }
345 
346 /* pru_connect2 is EOPNOTSUPP */
347 /* pru_control is EOPNOTSUPP */
348 
349 static void
350 rts_detach(struct socket *so)
351 {
352 	struct rawcb *rp = sotorawcb(so);
353 
354 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
355 
356 	RTSOCK_LOCK();
357 	switch(rp->rcb_proto.sp_protocol) {
358 	case AF_INET:
359 		V_route_cb.ip_count--;
360 		break;
361 	case AF_INET6:
362 		V_route_cb.ip6_count--;
363 		break;
364 	}
365 	V_route_cb.any_count--;
366 	RTSOCK_UNLOCK();
367 	raw_usrreqs.pru_detach(so);
368 }
369 
370 static int
371 rts_disconnect(struct socket *so)
372 {
373 
374 	return (raw_usrreqs.pru_disconnect(so));
375 }
376 
377 /* pru_listen is EOPNOTSUPP */
378 
379 static int
380 rts_peeraddr(struct socket *so, struct sockaddr **nam)
381 {
382 
383 	return (raw_usrreqs.pru_peeraddr(so, nam));
384 }
385 
386 /* pru_rcvd is EOPNOTSUPP */
387 /* pru_rcvoob is EOPNOTSUPP */
388 
389 static int
390 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
391 	 struct mbuf *control, struct thread *td)
392 {
393 
394 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
395 }
396 
397 /* pru_sense is null */
398 
399 static int
400 rts_shutdown(struct socket *so)
401 {
402 
403 	return (raw_usrreqs.pru_shutdown(so));
404 }
405 
406 static int
407 rts_sockaddr(struct socket *so, struct sockaddr **nam)
408 {
409 
410 	return (raw_usrreqs.pru_sockaddr(so, nam));
411 }
412 
413 static struct pr_usrreqs route_usrreqs = {
414 	.pru_abort =		rts_abort,
415 	.pru_attach =		rts_attach,
416 	.pru_bind =		rts_bind,
417 	.pru_connect =		rts_connect,
418 	.pru_detach =		rts_detach,
419 	.pru_disconnect =	rts_disconnect,
420 	.pru_peeraddr =		rts_peeraddr,
421 	.pru_send =		rts_send,
422 	.pru_shutdown =		rts_shutdown,
423 	.pru_sockaddr =		rts_sockaddr,
424 	.pru_close =		rts_close,
425 };
426 
427 #ifndef _SOCKADDR_UNION_DEFINED
428 #define	_SOCKADDR_UNION_DEFINED
429 /*
430  * The union of all possible address formats we handle.
431  */
432 union sockaddr_union {
433 	struct sockaddr		sa;
434 	struct sockaddr_in	sin;
435 	struct sockaddr_in6	sin6;
436 };
437 #endif /* _SOCKADDR_UNION_DEFINED */
438 
439 static int
440 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
441     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
442 {
443 
444 	/* First, see if the returned address is part of the jail. */
445 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
446 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
447 		return (0);
448 	}
449 
450 	switch (info->rti_info[RTAX_DST]->sa_family) {
451 #ifdef INET
452 	case AF_INET:
453 	{
454 		struct in_addr ia;
455 		struct ifaddr *ifa;
456 		int found;
457 
458 		found = 0;
459 		/*
460 		 * Try to find an address on the given outgoing interface
461 		 * that belongs to the jail.
462 		 */
463 		IF_ADDR_RLOCK(ifp);
464 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
465 			struct sockaddr *sa;
466 			sa = ifa->ifa_addr;
467 			if (sa->sa_family != AF_INET)
468 				continue;
469 			ia = ((struct sockaddr_in *)sa)->sin_addr;
470 			if (prison_check_ip4(cred, &ia) == 0) {
471 				found = 1;
472 				break;
473 			}
474 		}
475 		IF_ADDR_RUNLOCK(ifp);
476 		if (!found) {
477 			/*
478 			 * As a last resort return the 'default' jail address.
479 			 */
480 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
481 			    sin_addr;
482 			if (prison_get_ip4(cred, &ia) != 0)
483 				return (ESRCH);
484 		}
485 		bzero(&saun->sin, sizeof(struct sockaddr_in));
486 		saun->sin.sin_len = sizeof(struct sockaddr_in);
487 		saun->sin.sin_family = AF_INET;
488 		saun->sin.sin_addr.s_addr = ia.s_addr;
489 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
490 		break;
491 	}
492 #endif
493 #ifdef INET6
494 	case AF_INET6:
495 	{
496 		struct in6_addr ia6;
497 		struct ifaddr *ifa;
498 		int found;
499 
500 		found = 0;
501 		/*
502 		 * Try to find an address on the given outgoing interface
503 		 * that belongs to the jail.
504 		 */
505 		IF_ADDR_RLOCK(ifp);
506 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
507 			struct sockaddr *sa;
508 			sa = ifa->ifa_addr;
509 			if (sa->sa_family != AF_INET6)
510 				continue;
511 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
512 			    &ia6, sizeof(struct in6_addr));
513 			if (prison_check_ip6(cred, &ia6) == 0) {
514 				found = 1;
515 				break;
516 			}
517 		}
518 		IF_ADDR_RUNLOCK(ifp);
519 		if (!found) {
520 			/*
521 			 * As a last resort return the 'default' jail address.
522 			 */
523 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
524 			    sin6_addr;
525 			if (prison_get_ip6(cred, &ia6) != 0)
526 				return (ESRCH);
527 		}
528 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
529 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
530 		saun->sin6.sin6_family = AF_INET6;
531 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
532 		if (sa6_recoverscope(&saun->sin6) != 0)
533 			return (ESRCH);
534 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
535 		break;
536 	}
537 #endif
538 	default:
539 		return (ESRCH);
540 	}
541 	return (0);
542 }
543 
544 /*ARGSUSED*/
545 static int
546 route_output(struct mbuf *m, struct socket *so, ...)
547 {
548 	RIB_RLOCK_TRACKER;
549 	struct rt_msghdr *rtm = NULL;
550 	struct rtentry *rt = NULL;
551 	struct rib_head *rnh;
552 	struct rt_addrinfo info;
553 	struct sockaddr_storage ss;
554 #ifdef INET6
555 	struct sockaddr_in6 *sin6;
556 	int i, rti_need_deembed = 0;
557 #endif
558 	int alloc_len = 0, len, error = 0, fibnum;
559 	struct ifnet *ifp = NULL;
560 	union sockaddr_union saun;
561 	sa_family_t saf = AF_UNSPEC;
562 	struct rawcb *rp = NULL;
563 	struct walkarg w;
564 
565 	fibnum = so->so_fibnum;
566 
567 #define senderr(e) { error = e; goto flush;}
568 	if (m == NULL || ((m->m_len < sizeof(long)) &&
569 		       (m = m_pullup(m, sizeof(long))) == NULL))
570 		return (ENOBUFS);
571 	if ((m->m_flags & M_PKTHDR) == 0)
572 		panic("route_output");
573 	len = m->m_pkthdr.len;
574 	if (len < sizeof(*rtm) ||
575 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
576 		senderr(EINVAL);
577 
578 	/*
579 	 * Most of current messages are in range 200-240 bytes,
580 	 * minimize possible re-allocation on reply using larger size
581 	 * buffer aligned on 1k boundaty.
582 	 */
583 	alloc_len = roundup2(len, 1024);
584 	if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
585 		senderr(ENOBUFS);
586 
587 	m_copydata(m, 0, len, (caddr_t)rtm);
588 	bzero(&info, sizeof(info));
589 	bzero(&w, sizeof(w));
590 
591 	if (rtm->rtm_version != RTM_VERSION) {
592 		/* Do not touch message since format is unknown */
593 		free(rtm, M_TEMP);
594 		rtm = NULL;
595 		senderr(EPROTONOSUPPORT);
596 	}
597 
598 	/*
599 	 * Starting from here, it is possible
600 	 * to alter original message and insert
601 	 * caller PID and error value.
602 	 */
603 
604 	rtm->rtm_pid = curproc->p_pid;
605 	info.rti_addrs = rtm->rtm_addrs;
606 
607 	info.rti_mflags = rtm->rtm_inits;
608 	info.rti_rmx = &rtm->rtm_rmx;
609 
610 	/*
611 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
612 	 * link-local address because rtrequest requires addresses with
613 	 * embedded scope id.
614 	 */
615 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
616 		senderr(EINVAL);
617 
618 	info.rti_flags = rtm->rtm_flags;
619 	if (info.rti_info[RTAX_DST] == NULL ||
620 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
621 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
622 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
623 		senderr(EINVAL);
624 	saf = info.rti_info[RTAX_DST]->sa_family;
625 	/*
626 	 * Verify that the caller has the appropriate privilege; RTM_GET
627 	 * is the only operation the non-superuser is allowed.
628 	 */
629 	if (rtm->rtm_type != RTM_GET) {
630 		error = priv_check(curthread, PRIV_NET_ROUTE);
631 		if (error)
632 			senderr(error);
633 	}
634 
635 	/*
636 	 * The given gateway address may be an interface address.
637 	 * For example, issuing a "route change" command on a route
638 	 * entry that was created from a tunnel, and the gateway
639 	 * address given is the local end point. In this case the
640 	 * RTF_GATEWAY flag must be cleared or the destination will
641 	 * not be reachable even though there is no error message.
642 	 */
643 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
644 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
645 		struct rt_addrinfo ginfo;
646 		struct sockaddr *gdst;
647 
648 		bzero(&ginfo, sizeof(ginfo));
649 		bzero(&ss, sizeof(ss));
650 		ss.ss_len = sizeof(ss);
651 
652 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
653 		gdst = info.rti_info[RTAX_GATEWAY];
654 
655 		/*
656 		 * A host route through the loopback interface is
657 		 * installed for each interface adddress. In pre 8.0
658 		 * releases the interface address of a PPP link type
659 		 * is not reachable locally. This behavior is fixed as
660 		 * part of the new L2/L3 redesign and rewrite work. The
661 		 * signature of this interface address route is the
662 		 * AF_LINK sa_family type of the rt_gateway, and the
663 		 * rt_ifp has the IFF_LOOPBACK flag set.
664 		 */
665 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
666 			if (ss.ss_family == AF_LINK &&
667 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
668 				info.rti_flags &= ~RTF_GATEWAY;
669 				info.rti_flags |= RTF_GWFLAG_COMPAT;
670 			}
671 			rib_free_info(&ginfo);
672 		}
673 	}
674 
675 	switch (rtm->rtm_type) {
676 		struct rtentry *saved_nrt;
677 
678 	case RTM_ADD:
679 	case RTM_CHANGE:
680 		if (rtm->rtm_type == RTM_ADD) {
681 			if (info.rti_info[RTAX_GATEWAY] == NULL)
682 				senderr(EINVAL);
683 		}
684 		saved_nrt = NULL;
685 
686 		/* support for new ARP code */
687 		if (info.rti_info[RTAX_GATEWAY] != NULL &&
688 		    info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
689 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
690 			error = lla_rt_output(rtm, &info);
691 #ifdef INET6
692 			if (error == 0)
693 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
694 #endif
695 			break;
696 		}
697 		error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
698 		    fibnum);
699 		if (error == 0 && saved_nrt != NULL) {
700 #ifdef INET6
701 			rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
702 #endif
703 			RT_LOCK(saved_nrt);
704 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
705 			RT_REMREF(saved_nrt);
706 			RT_UNLOCK(saved_nrt);
707 		}
708 		break;
709 
710 	case RTM_DELETE:
711 		saved_nrt = NULL;
712 		/* support for new ARP code */
713 		if (info.rti_info[RTAX_GATEWAY] &&
714 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
715 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
716 			error = lla_rt_output(rtm, &info);
717 #ifdef INET6
718 			if (error == 0)
719 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
720 #endif
721 			break;
722 		}
723 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
724 		if (error == 0) {
725 			RT_LOCK(saved_nrt);
726 			rt = saved_nrt;
727 			goto report;
728 		}
729 #ifdef INET6
730 		/* rt_msg2() will not be used when RTM_DELETE fails. */
731 		rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
732 #endif
733 		break;
734 
735 	case RTM_GET:
736 		rnh = rt_tables_get_rnh(fibnum, saf);
737 		if (rnh == NULL)
738 			senderr(EAFNOSUPPORT);
739 
740 		RIB_RLOCK(rnh);
741 
742 		if (info.rti_info[RTAX_NETMASK] == NULL &&
743 		    rtm->rtm_type == RTM_GET) {
744 			/*
745 			 * Provide longest prefix match for
746 			 * address lookup (no mask).
747 			 * 'route -n get addr'
748 			 */
749 			rt = (struct rtentry *) rnh->rnh_matchaddr(
750 			    info.rti_info[RTAX_DST], &rnh->head);
751 		} else
752 			rt = (struct rtentry *) rnh->rnh_lookup(
753 			    info.rti_info[RTAX_DST],
754 			    info.rti_info[RTAX_NETMASK], &rnh->head);
755 
756 		if (rt == NULL) {
757 			RIB_RUNLOCK(rnh);
758 			senderr(ESRCH);
759 		}
760 #ifdef RADIX_MPATH
761 		/*
762 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
763 		 * we require users to specify a matching RTAX_GATEWAY.
764 		 *
765 		 * for RTM_GET, gate is optional even with multipath.
766 		 * if gate == NULL the first match is returned.
767 		 * (no need to call rt_mpath_matchgate if gate == NULL)
768 		 */
769 		if (rt_mpath_capable(rnh) &&
770 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
771 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
772 			if (!rt) {
773 				RIB_RUNLOCK(rnh);
774 				senderr(ESRCH);
775 			}
776 		}
777 #endif
778 		/*
779 		 * If performing proxied L2 entry insertion, and
780 		 * the actual PPP host entry is found, perform
781 		 * another search to retrieve the prefix route of
782 		 * the local end point of the PPP link.
783 		 */
784 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
785 			struct sockaddr laddr;
786 
787 			if (rt->rt_ifp != NULL &&
788 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
789 				struct ifaddr *ifa;
790 
791 				NET_EPOCH_ENTER();
792 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1,
793 						RT_ALL_FIBS);
794 				if (ifa != NULL)
795 					rt_maskedcopy(ifa->ifa_addr,
796 						      &laddr,
797 						      ifa->ifa_netmask);
798 				NET_EPOCH_EXIT();
799 			} else
800 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
801 					      &laddr,
802 					      rt->rt_ifa->ifa_netmask);
803 			/*
804 			 * refactor rt and no lock operation necessary
805 			 */
806 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
807 			    &rnh->head);
808 			if (rt == NULL) {
809 				RIB_RUNLOCK(rnh);
810 				senderr(ESRCH);
811 			}
812 		}
813 		RT_LOCK(rt);
814 		RT_ADDREF(rt);
815 		RIB_RUNLOCK(rnh);
816 
817 report:
818 		RT_LOCK_ASSERT(rt);
819 		if ((rt->rt_flags & RTF_HOST) == 0
820 		    ? jailed_without_vnet(curthread->td_ucred)
821 		    : prison_if(curthread->td_ucred,
822 		    rt_key(rt)) != 0) {
823 			RT_UNLOCK(rt);
824 			senderr(ESRCH);
825 		}
826 		info.rti_info[RTAX_DST] = rt_key(rt);
827 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
828 		info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
829 		    rt_mask(rt), &ss);
830 		info.rti_info[RTAX_GENMASK] = 0;
831 		if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
832 			ifp = rt->rt_ifp;
833 			if (ifp) {
834 				info.rti_info[RTAX_IFP] =
835 				    ifp->if_addr->ifa_addr;
836 				error = rtm_get_jailed(&info, ifp, rt,
837 				    &saun, curthread->td_ucred);
838 				if (error != 0) {
839 					RT_UNLOCK(rt);
840 					senderr(error);
841 				}
842 				if (ifp->if_flags & IFF_POINTOPOINT)
843 					info.rti_info[RTAX_BRD] =
844 					    rt->rt_ifa->ifa_dstaddr;
845 				rtm->rtm_index = ifp->if_index;
846 			} else {
847 				info.rti_info[RTAX_IFP] = NULL;
848 				info.rti_info[RTAX_IFA] = NULL;
849 			}
850 		} else if ((ifp = rt->rt_ifp) != NULL) {
851 			rtm->rtm_index = ifp->if_index;
852 		}
853 
854 		/* Check if we need to realloc storage */
855 		rtsock_msg_buffer(rtm->rtm_type, &info, NULL, &len);
856 		if (len > alloc_len) {
857 			struct rt_msghdr *new_rtm;
858 			new_rtm = malloc(len, M_TEMP, M_NOWAIT);
859 			if (new_rtm == NULL) {
860 				RT_UNLOCK(rt);
861 				senderr(ENOBUFS);
862 			}
863 			bcopy(rtm, new_rtm, rtm->rtm_msglen);
864 			free(rtm, M_TEMP);
865 			rtm = new_rtm;
866 			alloc_len = len;
867 		}
868 
869 		w.w_tmem = (caddr_t)rtm;
870 		w.w_tmemsize = alloc_len;
871 		rtsock_msg_buffer(rtm->rtm_type, &info, &w, &len);
872 
873 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
874 			rtm->rtm_flags = RTF_GATEWAY |
875 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
876 		else
877 			rtm->rtm_flags = rt->rt_flags;
878 		rt_getmetrics(rt, &rtm->rtm_rmx);
879 		rtm->rtm_addrs = info.rti_addrs;
880 
881 		RT_UNLOCK(rt);
882 		break;
883 
884 	default:
885 		senderr(EOPNOTSUPP);
886 	}
887 
888 flush:
889 	if (rt != NULL)
890 		RTFREE(rt);
891 	/*
892 	 * Check to see if we don't want our own messages.
893 	 */
894 	if ((so->so_options & SO_USELOOPBACK) == 0) {
895 		if (V_route_cb.any_count <= 1) {
896 			if (rtm != NULL)
897 				free(rtm, M_TEMP);
898 			m_freem(m);
899 			return (error);
900 		}
901 		/* There is another listener, so construct message */
902 		rp = sotorawcb(so);
903 	}
904 
905 	if (rtm != NULL) {
906 #ifdef INET6
907 		if (rti_need_deembed) {
908 			/* sin6_scope_id is recovered before sending rtm. */
909 			sin6 = (struct sockaddr_in6 *)&ss;
910 			for (i = 0; i < RTAX_MAX; i++) {
911 				if (info.rti_info[i] == NULL)
912 					continue;
913 				if (info.rti_info[i]->sa_family != AF_INET6)
914 					continue;
915 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
916 				if (sa6_recoverscope(sin6) == 0)
917 					bcopy(sin6, info.rti_info[i],
918 						    sizeof(*sin6));
919 			}
920 		}
921 #endif
922 		if (error != 0)
923 			rtm->rtm_errno = error;
924 		else
925 			rtm->rtm_flags |= RTF_DONE;
926 
927 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
928 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
929 			m_freem(m);
930 			m = NULL;
931 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
932 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
933 
934 		free(rtm, M_TEMP);
935 	}
936 	if (m != NULL) {
937 		M_SETFIB(m, fibnum);
938 		m->m_flags |= RTS_FILTER_FIB;
939 		if (rp) {
940 			/*
941 			 * XXX insure we don't get a copy by
942 			 * invalidating our protocol
943 			 */
944 			unsigned short family = rp->rcb_proto.sp_family;
945 			rp->rcb_proto.sp_family = 0;
946 			rt_dispatch(m, saf);
947 			rp->rcb_proto.sp_family = family;
948 		} else
949 			rt_dispatch(m, saf);
950 	}
951 
952 	return (error);
953 }
954 
955 static void
956 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
957 {
958 
959 	bzero(out, sizeof(*out));
960 	out->rmx_mtu = rt->rt_mtu;
961 	out->rmx_weight = rt->rt_weight;
962 	out->rmx_pksent = counter_u64_fetch(rt->rt_pksent);
963 	/* Kernel -> userland timebase conversion. */
964 	out->rmx_expire = rt->rt_expire ?
965 	    rt->rt_expire - time_uptime + time_second : 0;
966 }
967 
968 /*
969  * Extract the addresses of the passed sockaddrs.
970  * Do a little sanity checking so as to avoid bad memory references.
971  * This data is derived straight from userland.
972  */
973 static int
974 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
975 {
976 	struct sockaddr *sa;
977 	int i;
978 
979 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
980 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
981 			continue;
982 		sa = (struct sockaddr *)cp;
983 		/*
984 		 * It won't fit.
985 		 */
986 		if (cp + sa->sa_len > cplim)
987 			return (EINVAL);
988 		/*
989 		 * there are no more.. quit now
990 		 * If there are more bits, they are in error.
991 		 * I've seen this. route(1) can evidently generate these.
992 		 * This causes kernel to core dump.
993 		 * for compatibility, If we see this, point to a safe address.
994 		 */
995 		if (sa->sa_len == 0) {
996 			rtinfo->rti_info[i] = &sa_zero;
997 			return (0); /* should be EINVAL but for compat */
998 		}
999 		/* accept it */
1000 #ifdef INET6
1001 		if (sa->sa_family == AF_INET6)
1002 			sa6_embedscope((struct sockaddr_in6 *)sa,
1003 			    V_ip6_use_defzone);
1004 #endif
1005 		rtinfo->rti_info[i] = sa;
1006 		cp += SA_SIZE(sa);
1007 	}
1008 	return (0);
1009 }
1010 
1011 /*
1012  * Fill in @dmask with valid netmask leaving original @smask
1013  * intact. Mostly used with radix netmasks.
1014  */
1015 static struct sockaddr *
1016 rtsock_fix_netmask(struct sockaddr *dst, struct sockaddr *smask,
1017     struct sockaddr_storage *dmask)
1018 {
1019 	if (dst == NULL || smask == NULL)
1020 		return (NULL);
1021 
1022 	memset(dmask, 0, dst->sa_len);
1023 	memcpy(dmask, smask, smask->sa_len);
1024 	dmask->ss_len = dst->sa_len;
1025 	dmask->ss_family = dst->sa_family;
1026 
1027 	return ((struct sockaddr *)dmask);
1028 }
1029 
1030 /*
1031  * Writes information related to @rtinfo object to newly-allocated mbuf.
1032  * Assumes MCLBYTES is enough to construct any message.
1033  * Used for OS notifications of vaious events (if/ifa announces,etc)
1034  *
1035  * Returns allocated mbuf or NULL on failure.
1036  */
1037 static struct mbuf *
1038 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1039 {
1040 	struct rt_msghdr *rtm;
1041 	struct mbuf *m;
1042 	int i;
1043 	struct sockaddr *sa;
1044 #ifdef INET6
1045 	struct sockaddr_storage ss;
1046 	struct sockaddr_in6 *sin6;
1047 #endif
1048 	int len, dlen;
1049 
1050 	switch (type) {
1051 
1052 	case RTM_DELADDR:
1053 	case RTM_NEWADDR:
1054 		len = sizeof(struct ifa_msghdr);
1055 		break;
1056 
1057 	case RTM_DELMADDR:
1058 	case RTM_NEWMADDR:
1059 		len = sizeof(struct ifma_msghdr);
1060 		break;
1061 
1062 	case RTM_IFINFO:
1063 		len = sizeof(struct if_msghdr);
1064 		break;
1065 
1066 	case RTM_IFANNOUNCE:
1067 	case RTM_IEEE80211:
1068 		len = sizeof(struct if_announcemsghdr);
1069 		break;
1070 
1071 	default:
1072 		len = sizeof(struct rt_msghdr);
1073 	}
1074 
1075 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1076 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1077 	if (len > MHLEN)
1078 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1079 	else
1080 		m = m_gethdr(M_NOWAIT, MT_DATA);
1081 	if (m == NULL)
1082 		return (m);
1083 
1084 	m->m_pkthdr.len = m->m_len = len;
1085 	rtm = mtod(m, struct rt_msghdr *);
1086 	bzero((caddr_t)rtm, len);
1087 	for (i = 0; i < RTAX_MAX; i++) {
1088 		if ((sa = rtinfo->rti_info[i]) == NULL)
1089 			continue;
1090 		rtinfo->rti_addrs |= (1 << i);
1091 		dlen = SA_SIZE(sa);
1092 #ifdef INET6
1093 		if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1094 			sin6 = (struct sockaddr_in6 *)&ss;
1095 			bcopy(sa, sin6, sizeof(*sin6));
1096 			if (sa6_recoverscope(sin6) == 0)
1097 				sa = (struct sockaddr *)sin6;
1098 		}
1099 #endif
1100 		m_copyback(m, len, dlen, (caddr_t)sa);
1101 		len += dlen;
1102 	}
1103 	if (m->m_pkthdr.len != len) {
1104 		m_freem(m);
1105 		return (NULL);
1106 	}
1107 	rtm->rtm_msglen = len;
1108 	rtm->rtm_version = RTM_VERSION;
1109 	rtm->rtm_type = type;
1110 	return (m);
1111 }
1112 
1113 /*
1114  * Writes information related to @rtinfo object to preallocated buffer.
1115  * Stores needed size in @plen. If @w is NULL, calculates size without
1116  * writing.
1117  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1118  *
1119  * Returns 0 on success.
1120  *
1121  */
1122 static int
1123 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1124 {
1125 	int i;
1126 	int len, buflen = 0, dlen;
1127 	caddr_t cp = NULL;
1128 	struct rt_msghdr *rtm = NULL;
1129 #ifdef INET6
1130 	struct sockaddr_storage ss;
1131 	struct sockaddr_in6 *sin6;
1132 #endif
1133 #ifdef COMPAT_FREEBSD32
1134 	bool compat32 = false;
1135 #endif
1136 
1137 	switch (type) {
1138 
1139 	case RTM_DELADDR:
1140 	case RTM_NEWADDR:
1141 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1142 #ifdef COMPAT_FREEBSD32
1143 			if (w->w_req->flags & SCTL_MASK32) {
1144 				len = sizeof(struct ifa_msghdrl32);
1145 				compat32 = true;
1146 			} else
1147 #endif
1148 				len = sizeof(struct ifa_msghdrl);
1149 		} else
1150 			len = sizeof(struct ifa_msghdr);
1151 		break;
1152 
1153 	case RTM_IFINFO:
1154 #ifdef COMPAT_FREEBSD32
1155 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1156 			if (w->w_op == NET_RT_IFLISTL)
1157 				len = sizeof(struct if_msghdrl32);
1158 			else
1159 				len = sizeof(struct if_msghdr32);
1160 			compat32 = true;
1161 			break;
1162 		}
1163 #endif
1164 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1165 			len = sizeof(struct if_msghdrl);
1166 		else
1167 			len = sizeof(struct if_msghdr);
1168 		break;
1169 
1170 	case RTM_NEWMADDR:
1171 		len = sizeof(struct ifma_msghdr);
1172 		break;
1173 
1174 	default:
1175 		len = sizeof(struct rt_msghdr);
1176 	}
1177 
1178 	if (w != NULL) {
1179 		rtm = (struct rt_msghdr *)w->w_tmem;
1180 		buflen = w->w_tmemsize - len;
1181 		cp = (caddr_t)w->w_tmem + len;
1182 	}
1183 
1184 	rtinfo->rti_addrs = 0;
1185 	for (i = 0; i < RTAX_MAX; i++) {
1186 		struct sockaddr *sa;
1187 
1188 		if ((sa = rtinfo->rti_info[i]) == NULL)
1189 			continue;
1190 		rtinfo->rti_addrs |= (1 << i);
1191 #ifdef COMPAT_FREEBSD32
1192 		if (compat32)
1193 			dlen = SA_SIZE32(sa);
1194 		else
1195 #endif
1196 			dlen = SA_SIZE(sa);
1197 		if (cp != NULL && buflen >= dlen) {
1198 #ifdef INET6
1199 			if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1200 				sin6 = (struct sockaddr_in6 *)&ss;
1201 				bcopy(sa, sin6, sizeof(*sin6));
1202 				if (sa6_recoverscope(sin6) == 0)
1203 					sa = (struct sockaddr *)sin6;
1204 			}
1205 #endif
1206 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1207 			cp += dlen;
1208 			buflen -= dlen;
1209 		} else if (cp != NULL) {
1210 			/*
1211 			 * Buffer too small. Count needed size
1212 			 * and return with error.
1213 			 */
1214 			cp = NULL;
1215 		}
1216 
1217 		len += dlen;
1218 	}
1219 
1220 	if (cp != NULL) {
1221 		dlen = ALIGN(len) - len;
1222 		if (buflen < dlen)
1223 			cp = NULL;
1224 		else {
1225 			bzero(cp, dlen);
1226 			cp += dlen;
1227 			buflen -= dlen;
1228 		}
1229 	}
1230 	len = ALIGN(len);
1231 
1232 	if (cp != NULL) {
1233 		/* fill header iff buffer is large enough */
1234 		rtm->rtm_version = RTM_VERSION;
1235 		rtm->rtm_type = type;
1236 		rtm->rtm_msglen = len;
1237 	}
1238 
1239 	*plen = len;
1240 
1241 	if (w != NULL && cp == NULL)
1242 		return (ENOBUFS);
1243 
1244 	return (0);
1245 }
1246 
1247 /*
1248  * This routine is called to generate a message from the routing
1249  * socket indicating that a redirect has occurred, a routing lookup
1250  * has failed, or that a protocol has detected timeouts to a particular
1251  * destination.
1252  */
1253 void
1254 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1255     int fibnum)
1256 {
1257 	struct rt_msghdr *rtm;
1258 	struct mbuf *m;
1259 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1260 
1261 	if (V_route_cb.any_count == 0)
1262 		return;
1263 	m = rtsock_msg_mbuf(type, rtinfo);
1264 	if (m == NULL)
1265 		return;
1266 
1267 	if (fibnum != RT_ALL_FIBS) {
1268 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1269 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1270 		M_SETFIB(m, fibnum);
1271 		m->m_flags |= RTS_FILTER_FIB;
1272 	}
1273 
1274 	rtm = mtod(m, struct rt_msghdr *);
1275 	rtm->rtm_flags = RTF_DONE | flags;
1276 	rtm->rtm_errno = error;
1277 	rtm->rtm_addrs = rtinfo->rti_addrs;
1278 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1279 }
1280 
1281 void
1282 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1283 {
1284 
1285 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1286 }
1287 
1288 /*
1289  * This routine is called to generate a message from the routing
1290  * socket indicating that the status of a network interface has changed.
1291  */
1292 void
1293 rt_ifmsg(struct ifnet *ifp)
1294 {
1295 	struct if_msghdr *ifm;
1296 	struct mbuf *m;
1297 	struct rt_addrinfo info;
1298 
1299 	if (V_route_cb.any_count == 0)
1300 		return;
1301 	bzero((caddr_t)&info, sizeof(info));
1302 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1303 	if (m == NULL)
1304 		return;
1305 	ifm = mtod(m, struct if_msghdr *);
1306 	ifm->ifm_index = ifp->if_index;
1307 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1308 	if_data_copy(ifp, &ifm->ifm_data);
1309 	ifm->ifm_addrs = 0;
1310 	rt_dispatch(m, AF_UNSPEC);
1311 }
1312 
1313 /*
1314  * Announce interface address arrival/withdraw.
1315  * Please do not call directly, use rt_addrmsg().
1316  * Assume input data to be valid.
1317  * Returns 0 on success.
1318  */
1319 int
1320 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1321 {
1322 	struct rt_addrinfo info;
1323 	struct sockaddr *sa;
1324 	int ncmd;
1325 	struct mbuf *m;
1326 	struct ifa_msghdr *ifam;
1327 	struct ifnet *ifp = ifa->ifa_ifp;
1328 	struct sockaddr_storage ss;
1329 
1330 	if (V_route_cb.any_count == 0)
1331 		return (0);
1332 
1333 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1334 
1335 	bzero((caddr_t)&info, sizeof(info));
1336 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1337 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1338 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1339 	    info.rti_info[RTAX_IFP], ifa->ifa_netmask, &ss);
1340 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1341 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1342 		return (ENOBUFS);
1343 	ifam = mtod(m, struct ifa_msghdr *);
1344 	ifam->ifam_index = ifp->if_index;
1345 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1346 	ifam->ifam_flags = ifa->ifa_flags;
1347 	ifam->ifam_addrs = info.rti_addrs;
1348 
1349 	if (fibnum != RT_ALL_FIBS) {
1350 		M_SETFIB(m, fibnum);
1351 		m->m_flags |= RTS_FILTER_FIB;
1352 	}
1353 
1354 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1355 
1356 	return (0);
1357 }
1358 
1359 /*
1360  * Announce route addition/removal.
1361  * Please do not call directly, use rt_routemsg().
1362  * Note that @rt data MAY be inconsistent/invalid:
1363  * if some userland app sends us "invalid" route message (invalid mask,
1364  * no dst, wrong address families, etc...) we need to pass it back
1365  * to app (and any other rtsock consumers) with rtm_errno field set to
1366  * non-zero value.
1367  *
1368  * Returns 0 on success.
1369  */
1370 int
1371 rtsock_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
1372     int fibnum)
1373 {
1374 	struct rt_addrinfo info;
1375 	struct sockaddr *sa;
1376 	struct mbuf *m;
1377 	struct rt_msghdr *rtm;
1378 	struct sockaddr_storage ss;
1379 
1380 	if (V_route_cb.any_count == 0)
1381 		return (0);
1382 
1383 	bzero((caddr_t)&info, sizeof(info));
1384 	info.rti_info[RTAX_DST] = sa = rt_key(rt);
1385 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(sa, rt_mask(rt), &ss);
1386 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1387 	if ((m = rtsock_msg_mbuf(cmd, &info)) == NULL)
1388 		return (ENOBUFS);
1389 	rtm = mtod(m, struct rt_msghdr *);
1390 	rtm->rtm_index = ifp->if_index;
1391 	rtm->rtm_flags |= rt->rt_flags;
1392 	rtm->rtm_errno = error;
1393 	rtm->rtm_addrs = info.rti_addrs;
1394 
1395 	if (fibnum != RT_ALL_FIBS) {
1396 		M_SETFIB(m, fibnum);
1397 		m->m_flags |= RTS_FILTER_FIB;
1398 	}
1399 
1400 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1401 
1402 	return (0);
1403 }
1404 
1405 /*
1406  * This is the analogue to the rt_newaddrmsg which performs the same
1407  * function but for multicast group memberhips.  This is easier since
1408  * there is no route state to worry about.
1409  */
1410 void
1411 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1412 {
1413 	struct rt_addrinfo info;
1414 	struct mbuf *m = NULL;
1415 	struct ifnet *ifp = ifma->ifma_ifp;
1416 	struct ifma_msghdr *ifmam;
1417 
1418 	if (V_route_cb.any_count == 0)
1419 		return;
1420 
1421 	bzero((caddr_t)&info, sizeof(info));
1422 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1423 	if (ifp && ifp->if_addr)
1424 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1425 	else
1426 		info.rti_info[RTAX_IFP] = NULL;
1427 	/*
1428 	 * If a link-layer address is present, present it as a ``gateway''
1429 	 * (similarly to how ARP entries, e.g., are presented).
1430 	 */
1431 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1432 	m = rtsock_msg_mbuf(cmd, &info);
1433 	if (m == NULL)
1434 		return;
1435 	ifmam = mtod(m, struct ifma_msghdr *);
1436 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1437 	    __func__));
1438 	ifmam->ifmam_index = ifp->if_index;
1439 	ifmam->ifmam_addrs = info.rti_addrs;
1440 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1441 }
1442 
1443 static struct mbuf *
1444 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1445 	struct rt_addrinfo *info)
1446 {
1447 	struct if_announcemsghdr *ifan;
1448 	struct mbuf *m;
1449 
1450 	if (V_route_cb.any_count == 0)
1451 		return NULL;
1452 	bzero((caddr_t)info, sizeof(*info));
1453 	m = rtsock_msg_mbuf(type, info);
1454 	if (m != NULL) {
1455 		ifan = mtod(m, struct if_announcemsghdr *);
1456 		ifan->ifan_index = ifp->if_index;
1457 		strlcpy(ifan->ifan_name, ifp->if_xname,
1458 			sizeof(ifan->ifan_name));
1459 		ifan->ifan_what = what;
1460 	}
1461 	return m;
1462 }
1463 
1464 /*
1465  * This is called to generate routing socket messages indicating
1466  * IEEE80211 wireless events.
1467  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1468  */
1469 void
1470 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1471 {
1472 	struct mbuf *m;
1473 	struct rt_addrinfo info;
1474 
1475 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1476 	if (m != NULL) {
1477 		/*
1478 		 * Append the ieee80211 data.  Try to stick it in the
1479 		 * mbuf containing the ifannounce msg; otherwise allocate
1480 		 * a new mbuf and append.
1481 		 *
1482 		 * NB: we assume m is a single mbuf.
1483 		 */
1484 		if (data_len > M_TRAILINGSPACE(m)) {
1485 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1486 			if (n == NULL) {
1487 				m_freem(m);
1488 				return;
1489 			}
1490 			bcopy(data, mtod(n, void *), data_len);
1491 			n->m_len = data_len;
1492 			m->m_next = n;
1493 		} else if (data_len > 0) {
1494 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1495 			m->m_len += data_len;
1496 		}
1497 		if (m->m_flags & M_PKTHDR)
1498 			m->m_pkthdr.len += data_len;
1499 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1500 		rt_dispatch(m, AF_UNSPEC);
1501 	}
1502 }
1503 
1504 /*
1505  * This is called to generate routing socket messages indicating
1506  * network interface arrival and departure.
1507  */
1508 void
1509 rt_ifannouncemsg(struct ifnet *ifp, int what)
1510 {
1511 	struct mbuf *m;
1512 	struct rt_addrinfo info;
1513 
1514 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1515 	if (m != NULL)
1516 		rt_dispatch(m, AF_UNSPEC);
1517 }
1518 
1519 static void
1520 rt_dispatch(struct mbuf *m, sa_family_t saf)
1521 {
1522 	struct m_tag *tag;
1523 
1524 	/*
1525 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1526 	 * use when injecting the mbuf into the routing socket buffer from
1527 	 * the netisr.
1528 	 */
1529 	if (saf != AF_UNSPEC) {
1530 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1531 		    M_NOWAIT);
1532 		if (tag == NULL) {
1533 			m_freem(m);
1534 			return;
1535 		}
1536 		*(unsigned short *)(tag + 1) = saf;
1537 		m_tag_prepend(m, tag);
1538 	}
1539 #ifdef VIMAGE
1540 	if (V_loif)
1541 		m->m_pkthdr.rcvif = V_loif;
1542 	else {
1543 		m_freem(m);
1544 		return;
1545 	}
1546 #endif
1547 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1548 }
1549 
1550 /*
1551  * This is used in dumping the kernel table via sysctl().
1552  */
1553 static int
1554 sysctl_dumpentry(struct radix_node *rn, void *vw)
1555 {
1556 	struct walkarg *w = vw;
1557 	struct rtentry *rt = (struct rtentry *)rn;
1558 	int error = 0, size;
1559 	struct rt_addrinfo info;
1560 	struct sockaddr_storage ss;
1561 
1562 	IFNET_RLOCK_NOSLEEP_ASSERT();
1563 
1564 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1565 		return 0;
1566 	if ((rt->rt_flags & RTF_HOST) == 0
1567 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1568 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1569 		return (0);
1570 	bzero((caddr_t)&info, sizeof(info));
1571 	info.rti_info[RTAX_DST] = rt_key(rt);
1572 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1573 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1574 	    rt_mask(rt), &ss);
1575 	info.rti_info[RTAX_GENMASK] = 0;
1576 	if (rt->rt_ifp && !(rt->rt_ifp->if_flags & IFF_DYING)) {
1577 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1578 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1579 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1580 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1581 	}
1582 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1583 		return (error);
1584 	if (w->w_req && w->w_tmem) {
1585 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1586 
1587 		bzero(&rtm->rtm_index,
1588 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1589 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1590 			rtm->rtm_flags = RTF_GATEWAY |
1591 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1592 		else
1593 			rtm->rtm_flags = rt->rt_flags;
1594 		rt_getmetrics(rt, &rtm->rtm_rmx);
1595 		rtm->rtm_index = rt->rt_ifp->if_index;
1596 		rtm->rtm_addrs = info.rti_addrs;
1597 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1598 		return (error);
1599 	}
1600 	return (error);
1601 }
1602 
1603 static int
1604 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1605     struct rt_addrinfo *info, struct walkarg *w, int len)
1606 {
1607 	struct if_msghdrl *ifm;
1608 	struct if_data *ifd;
1609 
1610 	ifm = (struct if_msghdrl *)w->w_tmem;
1611 
1612 #ifdef COMPAT_FREEBSD32
1613 	if (w->w_req->flags & SCTL_MASK32) {
1614 		struct if_msghdrl32 *ifm32;
1615 
1616 		ifm32 = (struct if_msghdrl32 *)ifm;
1617 		ifm32->ifm_addrs = info->rti_addrs;
1618 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1619 		ifm32->ifm_index = ifp->if_index;
1620 		ifm32->_ifm_spare1 = 0;
1621 		ifm32->ifm_len = sizeof(*ifm32);
1622 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1623 		ifm32->_ifm_spare2 = 0;
1624 		ifd = &ifm32->ifm_data;
1625 	} else
1626 #endif
1627 	{
1628 		ifm->ifm_addrs = info->rti_addrs;
1629 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1630 		ifm->ifm_index = ifp->if_index;
1631 		ifm->_ifm_spare1 = 0;
1632 		ifm->ifm_len = sizeof(*ifm);
1633 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1634 		ifm->_ifm_spare2 = 0;
1635 		ifd = &ifm->ifm_data;
1636 	}
1637 
1638 	memcpy(ifd, src_ifd, sizeof(*ifd));
1639 
1640 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1641 }
1642 
1643 static int
1644 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1645     struct rt_addrinfo *info, struct walkarg *w, int len)
1646 {
1647 	struct if_msghdr *ifm;
1648 	struct if_data *ifd;
1649 
1650 	ifm = (struct if_msghdr *)w->w_tmem;
1651 
1652 #ifdef COMPAT_FREEBSD32
1653 	if (w->w_req->flags & SCTL_MASK32) {
1654 		struct if_msghdr32 *ifm32;
1655 
1656 		ifm32 = (struct if_msghdr32 *)ifm;
1657 		ifm32->ifm_addrs = info->rti_addrs;
1658 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1659 		ifm32->ifm_index = ifp->if_index;
1660 		ifm32->_ifm_spare1 = 0;
1661 		ifd = &ifm32->ifm_data;
1662 	} else
1663 #endif
1664 	{
1665 		ifm->ifm_addrs = info->rti_addrs;
1666 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1667 		ifm->ifm_index = ifp->if_index;
1668 		ifm->_ifm_spare1 = 0;
1669 		ifd = &ifm->ifm_data;
1670 	}
1671 
1672 	memcpy(ifd, src_ifd, sizeof(*ifd));
1673 
1674 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1675 }
1676 
1677 static int
1678 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1679     struct walkarg *w, int len)
1680 {
1681 	struct ifa_msghdrl *ifam;
1682 	struct if_data *ifd;
1683 
1684 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1685 
1686 #ifdef COMPAT_FREEBSD32
1687 	if (w->w_req->flags & SCTL_MASK32) {
1688 		struct ifa_msghdrl32 *ifam32;
1689 
1690 		ifam32 = (struct ifa_msghdrl32 *)ifam;
1691 		ifam32->ifam_addrs = info->rti_addrs;
1692 		ifam32->ifam_flags = ifa->ifa_flags;
1693 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1694 		ifam32->_ifam_spare1 = 0;
1695 		ifam32->ifam_len = sizeof(*ifam32);
1696 		ifam32->ifam_data_off =
1697 		    offsetof(struct ifa_msghdrl32, ifam_data);
1698 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1699 		ifd = &ifam32->ifam_data;
1700 	} else
1701 #endif
1702 	{
1703 		ifam->ifam_addrs = info->rti_addrs;
1704 		ifam->ifam_flags = ifa->ifa_flags;
1705 		ifam->ifam_index = ifa->ifa_ifp->if_index;
1706 		ifam->_ifam_spare1 = 0;
1707 		ifam->ifam_len = sizeof(*ifam);
1708 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1709 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1710 		ifd = &ifam->ifam_data;
1711 	}
1712 
1713 	bzero(ifd, sizeof(*ifd));
1714 	ifd->ifi_datalen = sizeof(struct if_data);
1715 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1716 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1717 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1718 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1719 
1720 	/* Fixup if_data carp(4) vhid. */
1721 	if (carp_get_vhid_p != NULL)
1722 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1723 
1724 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1725 }
1726 
1727 static int
1728 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1729     struct walkarg *w, int len)
1730 {
1731 	struct ifa_msghdr *ifam;
1732 
1733 	ifam = (struct ifa_msghdr *)w->w_tmem;
1734 	ifam->ifam_addrs = info->rti_addrs;
1735 	ifam->ifam_flags = ifa->ifa_flags;
1736 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1737 	ifam->_ifam_spare1 = 0;
1738 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1739 
1740 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1741 }
1742 
1743 static int
1744 sysctl_iflist(int af, struct walkarg *w)
1745 {
1746 	struct ifnet *ifp;
1747 	struct ifaddr *ifa;
1748 	struct if_data ifd;
1749 	struct rt_addrinfo info;
1750 	int len, error = 0;
1751 	struct sockaddr_storage ss;
1752 	struct epoch_tracker et;
1753 
1754 	bzero((caddr_t)&info, sizeof(info));
1755 	bzero(&ifd, sizeof(ifd));
1756 	NET_EPOCH_ENTER_ET(et);
1757 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1758 		if (w->w_arg && w->w_arg != ifp->if_index)
1759 			continue;
1760 		if_data_copy(ifp, &ifd);
1761 		ifa = ifp->if_addr;
1762 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1763 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1764 		if (error != 0)
1765 			goto done;
1766 		info.rti_info[RTAX_IFP] = NULL;
1767 		if (w->w_req && w->w_tmem) {
1768 			if (w->w_op == NET_RT_IFLISTL)
1769 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1770 				    len);
1771 			else
1772 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1773 				    len);
1774 			if (error)
1775 				goto done;
1776 		}
1777 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1778 			if (af && af != ifa->ifa_addr->sa_family)
1779 				continue;
1780 			if (prison_if(w->w_req->td->td_ucred,
1781 			    ifa->ifa_addr) != 0)
1782 				continue;
1783 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1784 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1785 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
1786 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1787 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1788 			if (error != 0)
1789 				goto done;
1790 			if (w->w_req && w->w_tmem) {
1791 				if (w->w_op == NET_RT_IFLISTL)
1792 					error = sysctl_iflist_ifaml(ifa, &info,
1793 					    w, len);
1794 				else
1795 					error = sysctl_iflist_ifam(ifa, &info,
1796 					    w, len);
1797 				if (error)
1798 					goto done;
1799 			}
1800 		}
1801 		info.rti_info[RTAX_IFA] = NULL;
1802 		info.rti_info[RTAX_NETMASK] = NULL;
1803 		info.rti_info[RTAX_BRD] = NULL;
1804 	}
1805 done:
1806 	NET_EPOCH_EXIT_ET(et);
1807 	return (error);
1808 }
1809 
1810 static int
1811 sysctl_ifmalist(int af, struct walkarg *w)
1812 {
1813 	struct rt_addrinfo info;
1814 	struct ifaddr *ifa;
1815 	struct ifmultiaddr *ifma;
1816 	struct ifnet *ifp;
1817 	int error, len;
1818 
1819 	error = 0;
1820 	bzero((caddr_t)&info, sizeof(info));
1821 
1822 	IFNET_RLOCK_NOSLEEP();
1823 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1824 		if (w->w_arg && w->w_arg != ifp->if_index)
1825 			continue;
1826 		ifa = ifp->if_addr;
1827 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1828 		IF_ADDR_RLOCK(ifp);
1829 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1830 			if (af && af != ifma->ifma_addr->sa_family)
1831 				continue;
1832 			if (prison_if(w->w_req->td->td_ucred,
1833 			    ifma->ifma_addr) != 0)
1834 				continue;
1835 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1836 			info.rti_info[RTAX_GATEWAY] =
1837 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1838 			    ifma->ifma_lladdr : NULL;
1839 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1840 			if (error != 0)
1841 				break;
1842 			if (w->w_req && w->w_tmem) {
1843 				struct ifma_msghdr *ifmam;
1844 
1845 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1846 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1847 				ifmam->ifmam_flags = 0;
1848 				ifmam->ifmam_addrs = info.rti_addrs;
1849 				ifmam->_ifmam_spare1 = 0;
1850 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1851 				if (error != 0)
1852 					break;
1853 			}
1854 		}
1855 		IF_ADDR_RUNLOCK(ifp);
1856 		if (error != 0)
1857 			break;
1858 	}
1859 	IFNET_RUNLOCK_NOSLEEP();
1860 	return (error);
1861 }
1862 
1863 static int
1864 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1865 {
1866 	RIB_RLOCK_TRACKER;
1867 	int	*name = (int *)arg1;
1868 	u_int	namelen = arg2;
1869 	struct rib_head *rnh = NULL; /* silence compiler. */
1870 	int	i, lim, error = EINVAL;
1871 	int	fib = 0;
1872 	u_char	af;
1873 	struct	walkarg w;
1874 
1875 	name ++;
1876 	namelen--;
1877 	if (req->newptr)
1878 		return (EPERM);
1879 	if (name[1] == NET_RT_DUMP) {
1880 		if (namelen == 3)
1881 			fib = req->td->td_proc->p_fibnum;
1882 		else if (namelen == 4)
1883 			fib = (name[3] == RT_ALL_FIBS) ?
1884 			    req->td->td_proc->p_fibnum : name[3];
1885 		else
1886 			return ((namelen < 3) ? EISDIR : ENOTDIR);
1887 		if (fib < 0 || fib >= rt_numfibs)
1888 			return (EINVAL);
1889 	} else if (namelen != 3)
1890 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1891 	af = name[0];
1892 	if (af > AF_MAX)
1893 		return (EINVAL);
1894 	bzero(&w, sizeof(w));
1895 	w.w_op = name[1];
1896 	w.w_arg = name[2];
1897 	w.w_req = req;
1898 
1899 	error = sysctl_wire_old_buffer(req, 0);
1900 	if (error)
1901 		return (error);
1902 
1903 	/*
1904 	 * Allocate reply buffer in advance.
1905 	 * All rtsock messages has maximum length of u_short.
1906 	 */
1907 	w.w_tmemsize = 65536;
1908 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
1909 
1910 	switch (w.w_op) {
1911 
1912 	case NET_RT_DUMP:
1913 	case NET_RT_FLAGS:
1914 		if (af == 0) {			/* dump all tables */
1915 			i = 1;
1916 			lim = AF_MAX;
1917 		} else				/* dump only one table */
1918 			i = lim = af;
1919 
1920 		/*
1921 		 * take care of llinfo entries, the caller must
1922 		 * specify an AF
1923 		 */
1924 		if (w.w_op == NET_RT_FLAGS &&
1925 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1926 			if (af != 0)
1927 				error = lltable_sysctl_dumparp(af, w.w_req);
1928 			else
1929 				error = EINVAL;
1930 			break;
1931 		}
1932 		/*
1933 		 * take care of routing entries
1934 		 */
1935 		for (error = 0; error == 0 && i <= lim; i++) {
1936 			rnh = rt_tables_get_rnh(fib, i);
1937 			if (rnh != NULL) {
1938 				RIB_RLOCK(rnh);
1939 				IFNET_RLOCK_NOSLEEP();
1940 			    	error = rnh->rnh_walktree(&rnh->head,
1941 				    sysctl_dumpentry, &w);
1942 				IFNET_RUNLOCK_NOSLEEP();
1943 				RIB_RUNLOCK(rnh);
1944 			} else if (af != 0)
1945 				error = EAFNOSUPPORT;
1946 		}
1947 		break;
1948 
1949 	case NET_RT_IFLIST:
1950 	case NET_RT_IFLISTL:
1951 		error = sysctl_iflist(af, &w);
1952 		break;
1953 
1954 	case NET_RT_IFMALIST:
1955 		error = sysctl_ifmalist(af, &w);
1956 		break;
1957 	}
1958 
1959 	free(w.w_tmem, M_TEMP);
1960 	return (error);
1961 }
1962 
1963 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1964 
1965 /*
1966  * Definitions of protocols supported in the ROUTE domain.
1967  */
1968 
1969 static struct domain routedomain;		/* or at least forward */
1970 
1971 static struct protosw routesw[] = {
1972 {
1973 	.pr_type =		SOCK_RAW,
1974 	.pr_domain =		&routedomain,
1975 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1976 	.pr_output =		route_output,
1977 	.pr_ctlinput =		raw_ctlinput,
1978 	.pr_init =		raw_init,
1979 	.pr_usrreqs =		&route_usrreqs
1980 }
1981 };
1982 
1983 static struct domain routedomain = {
1984 	.dom_family =		PF_ROUTE,
1985 	.dom_name =		 "route",
1986 	.dom_protosw =		routesw,
1987 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
1988 };
1989 
1990 VNET_DOMAIN_SET(route);
1991