xref: /freebsd/sys/net/rtsock.c (revision 4e99f45480598189d49d45a825533a6c9e12f02c)
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_ddb.h"
35 #include "opt_mpath.h"
36 #include "opt_inet.h"
37 #include "opt_inet6.h"
38 
39 #include <sys/param.h>
40 #include <sys/jail.h>
41 #include <sys/kernel.h>
42 #include <sys/domain.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/protosw.h>
49 #include <sys/rmlock.h>
50 #include <sys/rwlock.h>
51 #include <sys/signalvar.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/systm.h>
56 
57 #include <net/if.h>
58 #include <net/if_var.h>
59 #include <net/if_dl.h>
60 #include <net/if_llatbl.h>
61 #include <net/if_types.h>
62 #include <net/netisr.h>
63 #include <net/raw_cb.h>
64 #include <net/route.h>
65 #include <net/route/route_var.h>
66 #ifdef RADIX_MPATH
67 #include <net/radix_mpath.h>
68 #endif
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/if_ether.h>
73 #include <netinet/ip_carp.h>
74 #ifdef INET6
75 #include <netinet6/ip6_var.h>
76 #include <netinet6/scope6_var.h>
77 #endif
78 #include <net/route/nhop.h>
79 #include <net/route/shared.h>
80 
81 #ifdef COMPAT_FREEBSD32
82 #include <sys/mount.h>
83 #include <compat/freebsd32/freebsd32.h>
84 
85 struct if_msghdr32 {
86 	uint16_t ifm_msglen;
87 	uint8_t	ifm_version;
88 	uint8_t	ifm_type;
89 	int32_t	ifm_addrs;
90 	int32_t	ifm_flags;
91 	uint16_t ifm_index;
92 	uint16_t _ifm_spare1;
93 	struct	if_data ifm_data;
94 };
95 
96 struct if_msghdrl32 {
97 	uint16_t ifm_msglen;
98 	uint8_t	ifm_version;
99 	uint8_t	ifm_type;
100 	int32_t	ifm_addrs;
101 	int32_t	ifm_flags;
102 	uint16_t ifm_index;
103 	uint16_t _ifm_spare1;
104 	uint16_t ifm_len;
105 	uint16_t ifm_data_off;
106 	uint32_t _ifm_spare2;
107 	struct	if_data ifm_data;
108 };
109 
110 struct ifa_msghdrl32 {
111 	uint16_t ifam_msglen;
112 	uint8_t	ifam_version;
113 	uint8_t	ifam_type;
114 	int32_t	ifam_addrs;
115 	int32_t	ifam_flags;
116 	uint16_t ifam_index;
117 	uint16_t _ifam_spare1;
118 	uint16_t ifam_len;
119 	uint16_t ifam_data_off;
120 	int32_t	ifam_metric;
121 	struct	if_data ifam_data;
122 };
123 
124 #define SA_SIZE32(sa)						\
125     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
126 	sizeof(int)		:				\
127 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
128 
129 #endif /* COMPAT_FREEBSD32 */
130 
131 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
132 
133 /* NB: these are not modified */
134 static struct	sockaddr route_src = { 2, PF_ROUTE, };
135 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
136 
137 /* These are external hooks for CARP. */
138 int	(*carp_get_vhid_p)(struct ifaddr *);
139 
140 /*
141  * Used by rtsock/raw_input callback code to decide whether to filter the update
142  * notification to a socket bound to a particular FIB.
143  */
144 #define	RTS_FILTER_FIB	M_PROTO8
145 
146 typedef struct {
147 	int	ip_count;	/* attached w/ AF_INET */
148 	int	ip6_count;	/* attached w/ AF_INET6 */
149 	int	any_count;	/* total attached */
150 } route_cb_t;
151 VNET_DEFINE_STATIC(route_cb_t, route_cb);
152 #define	V_route_cb VNET(route_cb)
153 
154 struct mtx rtsock_mtx;
155 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
156 
157 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
158 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
159 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
160 
161 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
162     "");
163 
164 struct walkarg {
165 	int	w_tmemsize;
166 	int	w_op, w_arg;
167 	caddr_t	w_tmem;
168 	struct sysctl_req *w_req;
169 };
170 
171 static void	rts_input(struct mbuf *m);
172 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
173 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
174 			struct walkarg *w, int *plen);
175 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
176 			struct rt_addrinfo *rtinfo);
177 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
178 static int	sysctl_iflist(int af, struct walkarg *w);
179 static int	sysctl_ifmalist(int af, struct walkarg *w);
180 static int	route_output(struct mbuf *m, struct socket *so, ...);
181 static void	rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out);
182 static void	rt_dispatch(struct mbuf *, sa_family_t);
183 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
184 			struct rt_msghdr *rtm, struct rtentry **ret_nrt);
185 static int	update_rtm_from_rte(struct rt_addrinfo *info,
186 			struct rt_msghdr **prtm, int alloc_len,
187 			struct rtentry *rt);
188 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
189 			struct mbuf *m, sa_family_t saf, u_int fibnum,
190 			int rtm_errno);
191 static int	can_export_rte(struct ucred *td_ucred, const struct rtentry *rt);
192 
193 static struct netisr_handler rtsock_nh = {
194 	.nh_name = "rtsock",
195 	.nh_handler = rts_input,
196 	.nh_proto = NETISR_ROUTE,
197 	.nh_policy = NETISR_POLICY_SOURCE,
198 };
199 
200 static int
201 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
202 {
203 	int error, qlimit;
204 
205 	netisr_getqlimit(&rtsock_nh, &qlimit);
206 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
207         if (error || !req->newptr)
208                 return (error);
209 	if (qlimit < 1)
210 		return (EINVAL);
211 	return (netisr_setqlimit(&rtsock_nh, qlimit));
212 }
213 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
214     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
215     0, 0, sysctl_route_netisr_maxqlen, "I",
216     "maximum routing socket dispatch queue length");
217 
218 static void
219 vnet_rts_init(void)
220 {
221 	int tmp;
222 
223 	if (IS_DEFAULT_VNET(curvnet)) {
224 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
225 			rtsock_nh.nh_qlimit = tmp;
226 		netisr_register(&rtsock_nh);
227 	}
228 #ifdef VIMAGE
229 	 else
230 		netisr_register_vnet(&rtsock_nh);
231 #endif
232 }
233 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
234     vnet_rts_init, 0);
235 
236 #ifdef VIMAGE
237 static void
238 vnet_rts_uninit(void)
239 {
240 
241 	netisr_unregister_vnet(&rtsock_nh);
242 }
243 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
244     vnet_rts_uninit, 0);
245 #endif
246 
247 static int
248 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
249     struct rawcb *rp)
250 {
251 	int fibnum;
252 
253 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
254 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
255 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
256 
257 	/* No filtering requested. */
258 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
259 		return (0);
260 
261 	/* Check if it is a rts and the fib matches the one of the socket. */
262 	fibnum = M_GETFIB(m);
263 	if (proto->sp_family != PF_ROUTE ||
264 	    rp->rcb_socket == NULL ||
265 	    rp->rcb_socket->so_fibnum == fibnum)
266 		return (0);
267 
268 	/* Filtering requested and no match, the socket shall be skipped. */
269 	return (1);
270 }
271 
272 static void
273 rts_input(struct mbuf *m)
274 {
275 	struct sockproto route_proto;
276 	unsigned short *family;
277 	struct m_tag *tag;
278 
279 	route_proto.sp_family = PF_ROUTE;
280 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
281 	if (tag != NULL) {
282 		family = (unsigned short *)(tag + 1);
283 		route_proto.sp_protocol = *family;
284 		m_tag_delete(m, tag);
285 	} else
286 		route_proto.sp_protocol = 0;
287 
288 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
289 }
290 
291 /*
292  * It really doesn't make any sense at all for this code to share much
293  * with raw_usrreq.c, since its functionality is so restricted.  XXX
294  */
295 static void
296 rts_abort(struct socket *so)
297 {
298 
299 	raw_usrreqs.pru_abort(so);
300 }
301 
302 static void
303 rts_close(struct socket *so)
304 {
305 
306 	raw_usrreqs.pru_close(so);
307 }
308 
309 /* pru_accept is EOPNOTSUPP */
310 
311 static int
312 rts_attach(struct socket *so, int proto, struct thread *td)
313 {
314 	struct rawcb *rp;
315 	int error;
316 
317 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
318 
319 	/* XXX */
320 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
321 
322 	so->so_pcb = (caddr_t)rp;
323 	so->so_fibnum = td->td_proc->p_fibnum;
324 	error = raw_attach(so, proto);
325 	rp = sotorawcb(so);
326 	if (error) {
327 		so->so_pcb = NULL;
328 		free(rp, M_PCB);
329 		return error;
330 	}
331 	RTSOCK_LOCK();
332 	switch(rp->rcb_proto.sp_protocol) {
333 	case AF_INET:
334 		V_route_cb.ip_count++;
335 		break;
336 	case AF_INET6:
337 		V_route_cb.ip6_count++;
338 		break;
339 	}
340 	V_route_cb.any_count++;
341 	RTSOCK_UNLOCK();
342 	soisconnected(so);
343 	so->so_options |= SO_USELOOPBACK;
344 	return 0;
345 }
346 
347 static int
348 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
349 {
350 
351 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
352 }
353 
354 static int
355 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
356 {
357 
358 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
359 }
360 
361 /* pru_connect2 is EOPNOTSUPP */
362 /* pru_control is EOPNOTSUPP */
363 
364 static void
365 rts_detach(struct socket *so)
366 {
367 	struct rawcb *rp = sotorawcb(so);
368 
369 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
370 
371 	RTSOCK_LOCK();
372 	switch(rp->rcb_proto.sp_protocol) {
373 	case AF_INET:
374 		V_route_cb.ip_count--;
375 		break;
376 	case AF_INET6:
377 		V_route_cb.ip6_count--;
378 		break;
379 	}
380 	V_route_cb.any_count--;
381 	RTSOCK_UNLOCK();
382 	raw_usrreqs.pru_detach(so);
383 }
384 
385 static int
386 rts_disconnect(struct socket *so)
387 {
388 
389 	return (raw_usrreqs.pru_disconnect(so));
390 }
391 
392 /* pru_listen is EOPNOTSUPP */
393 
394 static int
395 rts_peeraddr(struct socket *so, struct sockaddr **nam)
396 {
397 
398 	return (raw_usrreqs.pru_peeraddr(so, nam));
399 }
400 
401 /* pru_rcvd is EOPNOTSUPP */
402 /* pru_rcvoob is EOPNOTSUPP */
403 
404 static int
405 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
406 	 struct mbuf *control, struct thread *td)
407 {
408 
409 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
410 }
411 
412 /* pru_sense is null */
413 
414 static int
415 rts_shutdown(struct socket *so)
416 {
417 
418 	return (raw_usrreqs.pru_shutdown(so));
419 }
420 
421 static int
422 rts_sockaddr(struct socket *so, struct sockaddr **nam)
423 {
424 
425 	return (raw_usrreqs.pru_sockaddr(so, nam));
426 }
427 
428 static struct pr_usrreqs route_usrreqs = {
429 	.pru_abort =		rts_abort,
430 	.pru_attach =		rts_attach,
431 	.pru_bind =		rts_bind,
432 	.pru_connect =		rts_connect,
433 	.pru_detach =		rts_detach,
434 	.pru_disconnect =	rts_disconnect,
435 	.pru_peeraddr =		rts_peeraddr,
436 	.pru_send =		rts_send,
437 	.pru_shutdown =		rts_shutdown,
438 	.pru_sockaddr =		rts_sockaddr,
439 	.pru_close =		rts_close,
440 };
441 
442 #ifndef _SOCKADDR_UNION_DEFINED
443 #define	_SOCKADDR_UNION_DEFINED
444 /*
445  * The union of all possible address formats we handle.
446  */
447 union sockaddr_union {
448 	struct sockaddr		sa;
449 	struct sockaddr_in	sin;
450 	struct sockaddr_in6	sin6;
451 };
452 #endif /* _SOCKADDR_UNION_DEFINED */
453 
454 static int
455 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
456     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
457 {
458 #if defined(INET) || defined(INET6)
459 	struct epoch_tracker et;
460 #endif
461 
462 	/* First, see if the returned address is part of the jail. */
463 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
464 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
465 		return (0);
466 	}
467 
468 	switch (info->rti_info[RTAX_DST]->sa_family) {
469 #ifdef INET
470 	case AF_INET:
471 	{
472 		struct in_addr ia;
473 		struct ifaddr *ifa;
474 		int found;
475 
476 		found = 0;
477 		/*
478 		 * Try to find an address on the given outgoing interface
479 		 * that belongs to the jail.
480 		 */
481 		NET_EPOCH_ENTER(et);
482 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
483 			struct sockaddr *sa;
484 			sa = ifa->ifa_addr;
485 			if (sa->sa_family != AF_INET)
486 				continue;
487 			ia = ((struct sockaddr_in *)sa)->sin_addr;
488 			if (prison_check_ip4(cred, &ia) == 0) {
489 				found = 1;
490 				break;
491 			}
492 		}
493 		NET_EPOCH_EXIT(et);
494 		if (!found) {
495 			/*
496 			 * As a last resort return the 'default' jail address.
497 			 */
498 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
499 			    sin_addr;
500 			if (prison_get_ip4(cred, &ia) != 0)
501 				return (ESRCH);
502 		}
503 		bzero(&saun->sin, sizeof(struct sockaddr_in));
504 		saun->sin.sin_len = sizeof(struct sockaddr_in);
505 		saun->sin.sin_family = AF_INET;
506 		saun->sin.sin_addr.s_addr = ia.s_addr;
507 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
508 		break;
509 	}
510 #endif
511 #ifdef INET6
512 	case AF_INET6:
513 	{
514 		struct in6_addr ia6;
515 		struct ifaddr *ifa;
516 		int found;
517 
518 		found = 0;
519 		/*
520 		 * Try to find an address on the given outgoing interface
521 		 * that belongs to the jail.
522 		 */
523 		NET_EPOCH_ENTER(et);
524 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
525 			struct sockaddr *sa;
526 			sa = ifa->ifa_addr;
527 			if (sa->sa_family != AF_INET6)
528 				continue;
529 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
530 			    &ia6, sizeof(struct in6_addr));
531 			if (prison_check_ip6(cred, &ia6) == 0) {
532 				found = 1;
533 				break;
534 			}
535 		}
536 		NET_EPOCH_EXIT(et);
537 		if (!found) {
538 			/*
539 			 * As a last resort return the 'default' jail address.
540 			 */
541 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
542 			    sin6_addr;
543 			if (prison_get_ip6(cred, &ia6) != 0)
544 				return (ESRCH);
545 		}
546 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
547 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
548 		saun->sin6.sin6_family = AF_INET6;
549 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
550 		if (sa6_recoverscope(&saun->sin6) != 0)
551 			return (ESRCH);
552 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
553 		break;
554 	}
555 #endif
556 	default:
557 		return (ESRCH);
558 	}
559 	return (0);
560 }
561 
562 /*
563  * Fills in @info based on userland-provided @rtm message.
564  *
565  * Returns 0 on success.
566  */
567 static int
568 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
569 {
570 	int error;
571 	sa_family_t saf;
572 
573 	rtm->rtm_pid = curproc->p_pid;
574 	info->rti_addrs = rtm->rtm_addrs;
575 
576 	info->rti_mflags = rtm->rtm_inits;
577 	info->rti_rmx = &rtm->rtm_rmx;
578 
579 	/*
580 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
581 	 * link-local address because rtrequest requires addresses with
582 	 * embedded scope id.
583 	 */
584 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
585 		return (EINVAL);
586 
587 	if (rtm->rtm_flags & RTF_RNH_LOCKED)
588 		return (EINVAL);
589 	info->rti_flags = rtm->rtm_flags;
590 	if (info->rti_info[RTAX_DST] == NULL ||
591 	    info->rti_info[RTAX_DST]->sa_family >= AF_MAX ||
592 	    (info->rti_info[RTAX_GATEWAY] != NULL &&
593 	     info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
594 		return (EINVAL);
595 	saf = info->rti_info[RTAX_DST]->sa_family;
596 	/*
597 	 * Verify that the caller has the appropriate privilege; RTM_GET
598 	 * is the only operation the non-superuser is allowed.
599 	 */
600 	if (rtm->rtm_type != RTM_GET) {
601 		error = priv_check(curthread, PRIV_NET_ROUTE);
602 		if (error != 0)
603 			return (error);
604 	}
605 
606 	/*
607 	 * The given gateway address may be an interface address.
608 	 * For example, issuing a "route change" command on a route
609 	 * entry that was created from a tunnel, and the gateway
610 	 * address given is the local end point. In this case the
611 	 * RTF_GATEWAY flag must be cleared or the destination will
612 	 * not be reachable even though there is no error message.
613 	 */
614 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
615 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
616 		struct rt_addrinfo ginfo;
617 		struct sockaddr *gdst;
618 		struct sockaddr_storage ss;
619 
620 		bzero(&ginfo, sizeof(ginfo));
621 		bzero(&ss, sizeof(ss));
622 		ss.ss_len = sizeof(ss);
623 
624 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
625 		gdst = info->rti_info[RTAX_GATEWAY];
626 
627 		/*
628 		 * A host route through the loopback interface is
629 		 * installed for each interface adddress. In pre 8.0
630 		 * releases the interface address of a PPP link type
631 		 * is not reachable locally. This behavior is fixed as
632 		 * part of the new L2/L3 redesign and rewrite work. The
633 		 * signature of this interface address route is the
634 		 * AF_LINK sa_family type of the gateway, and the
635 		 * rt_ifp has the IFF_LOOPBACK flag set.
636 		 */
637 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
638 			if (ss.ss_family == AF_LINK &&
639 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
640 				info->rti_flags &= ~RTF_GATEWAY;
641 				info->rti_flags |= RTF_GWFLAG_COMPAT;
642 			}
643 			rib_free_info(&ginfo);
644 		}
645 	}
646 
647 	return (0);
648 }
649 
650 /*
651  * Handles RTM_GET message from routing socket, returning matching rt.
652  *
653  * Returns:
654  * 0 on success, with locked and referenced matching rt in @rt_nrt
655  * errno of failure
656  */
657 static int
658 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
659     struct rt_msghdr *rtm, struct rtentry **ret_nrt)
660 {
661 	RIB_RLOCK_TRACKER;
662 	struct rtentry *rt;
663 	struct rib_head *rnh;
664 	sa_family_t saf;
665 
666 	saf = info->rti_info[RTAX_DST]->sa_family;
667 
668 	rnh = rt_tables_get_rnh(fibnum, saf);
669 	if (rnh == NULL)
670 		return (EAFNOSUPPORT);
671 
672 	RIB_RLOCK(rnh);
673 
674 	if (info->rti_info[RTAX_NETMASK] == NULL) {
675 		/*
676 		 * Provide longest prefix match for
677 		 * address lookup (no mask).
678 		 * 'route -n get addr'
679 		 */
680 		rt = (struct rtentry *) rnh->rnh_matchaddr(
681 		    info->rti_info[RTAX_DST], &rnh->head);
682 	} else
683 		rt = (struct rtentry *) rnh->rnh_lookup(
684 		    info->rti_info[RTAX_DST],
685 		    info->rti_info[RTAX_NETMASK], &rnh->head);
686 
687 	if (rt == NULL) {
688 		RIB_RUNLOCK(rnh);
689 		return (ESRCH);
690 	}
691 #ifdef RADIX_MPATH
692 	/*
693 	 * for RTM_GET, gate is optional even with multipath.
694 	 * if gate == NULL the first match is returned.
695 	 * (no need to call rt_mpath_matchgate if gate == NULL)
696 	 */
697 	if (rt_mpath_capable(rnh) && info->rti_info[RTAX_GATEWAY]) {
698 		rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
699 		if (!rt) {
700 			RIB_RUNLOCK(rnh);
701 			return (ESRCH);
702 		}
703 	}
704 #endif
705 	/*
706 	 * If performing proxied L2 entry insertion, and
707 	 * the actual PPP host entry is found, perform
708 	 * another search to retrieve the prefix route of
709 	 * the local end point of the PPP link.
710 	 * TODO: move this logic to userland.
711 	 */
712 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
713 		struct sockaddr laddr;
714 		struct nhop_object *nh;
715 
716 		nh = rt->rt_nhop;
717 		if (nh->nh_ifp != NULL &&
718 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
719 			struct epoch_tracker et;
720 			struct ifaddr *ifa;
721 
722 			NET_EPOCH_ENTER(et);
723 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
724 					RT_ALL_FIBS);
725 			NET_EPOCH_EXIT(et);
726 			if (ifa != NULL)
727 				rt_maskedcopy(ifa->ifa_addr,
728 					      &laddr,
729 					      ifa->ifa_netmask);
730 		} else
731 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
732 				      &laddr,
733 				      nh->nh_ifa->ifa_netmask);
734 		/*
735 		 * refactor rt and no lock operation necessary
736 		 */
737 		rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
738 		    &rnh->head);
739 		if (rt == NULL) {
740 			RIB_RUNLOCK(rnh);
741 			return (ESRCH);
742 		}
743 	}
744 	RT_LOCK(rt);
745 	RIB_RUNLOCK(rnh);
746 
747 	*ret_nrt = rt;
748 
749 	return (0);
750 }
751 
752 /*
753  * Update sockaddrs, flags, etc in @prtm based on @rt data.
754  * Assumes @rt is locked.
755  * rtm can be reallocated.
756  *
757  * Returns 0 on success, along with pointer to (potentially reallocated)
758  *  rtm.
759  *
760  */
761 static int
762 update_rtm_from_rte(struct rt_addrinfo *info, struct rt_msghdr **prtm,
763     int alloc_len, struct rtentry *rt)
764 {
765 	struct sockaddr_storage netmask_ss;
766 	struct walkarg w;
767 	union sockaddr_union saun;
768 	struct rt_msghdr *rtm, *orig_rtm = NULL;
769 	struct nhop_object *nh;
770 	struct ifnet *ifp;
771 	int error, len;
772 
773 	RT_LOCK_ASSERT(rt);
774 
775 	rtm = *prtm;
776 
777 	nh = rt->rt_nhop;
778 	info->rti_info[RTAX_DST] = rt_key(rt);
779 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
780 	info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
781 	    rt_mask(rt), &netmask_ss);
782 	info->rti_info[RTAX_GENMASK] = 0;
783 	ifp = nh->nh_ifp;
784 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
785 		if (ifp) {
786 			info->rti_info[RTAX_IFP] =
787 			    ifp->if_addr->ifa_addr;
788 			error = rtm_get_jailed(info, ifp, nh,
789 			    &saun, curthread->td_ucred);
790 			if (error != 0)
791 				return (error);
792 			if (ifp->if_flags & IFF_POINTOPOINT)
793 				info->rti_info[RTAX_BRD] =
794 				    nh->nh_ifa->ifa_dstaddr;
795 			rtm->rtm_index = ifp->if_index;
796 		} else {
797 			info->rti_info[RTAX_IFP] = NULL;
798 			info->rti_info[RTAX_IFA] = NULL;
799 		}
800 	} else if (ifp != NULL)
801 		rtm->rtm_index = ifp->if_index;
802 
803 	/* Check if we need to realloc storage */
804 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
805 	if (len > alloc_len) {
806 		struct rt_msghdr *tmp_rtm;
807 
808 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
809 		if (tmp_rtm == NULL)
810 			return (ENOBUFS);
811 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
812 		orig_rtm = rtm;
813 		rtm = tmp_rtm;
814 		alloc_len = len;
815 
816 		/*
817 		 * Delay freeing original rtm as info contains
818 		 * data referencing it.
819 		 */
820 	}
821 
822 	w.w_tmem = (caddr_t)rtm;
823 	w.w_tmemsize = alloc_len;
824 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
825 
826 	if (rt->rt_flags & RTF_GWFLAG_COMPAT)
827 		rtm->rtm_flags = RTF_GATEWAY |
828 			(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
829 	else
830 		rtm->rtm_flags = rt->rt_flags;
831 	rt_getmetrics(rt, &rtm->rtm_rmx);
832 	rtm->rtm_addrs = info->rti_addrs;
833 
834 	if (orig_rtm != NULL)
835 		free(orig_rtm, M_TEMP);
836 	*prtm = rtm;
837 
838 	return (0);
839 }
840 
841 /*ARGSUSED*/
842 static int
843 route_output(struct mbuf *m, struct socket *so, ...)
844 {
845 	struct rt_msghdr *rtm = NULL;
846 	struct rtentry *rt = NULL;
847 	struct rt_addrinfo info;
848 	struct epoch_tracker et;
849 #ifdef INET6
850 	struct sockaddr_storage ss;
851 	struct sockaddr_in6 *sin6;
852 	int i, rti_need_deembed = 0;
853 #endif
854 	int alloc_len = 0, len, error = 0, fibnum;
855 	sa_family_t saf = AF_UNSPEC;
856 	struct walkarg w;
857 
858 	fibnum = so->so_fibnum;
859 
860 #define senderr(e) { error = e; goto flush;}
861 	if (m == NULL || ((m->m_len < sizeof(long)) &&
862 		       (m = m_pullup(m, sizeof(long))) == NULL))
863 		return (ENOBUFS);
864 	if ((m->m_flags & M_PKTHDR) == 0)
865 		panic("route_output");
866 	NET_EPOCH_ENTER(et);
867 	len = m->m_pkthdr.len;
868 	if (len < sizeof(*rtm) ||
869 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
870 		senderr(EINVAL);
871 
872 	/*
873 	 * Most of current messages are in range 200-240 bytes,
874 	 * minimize possible re-allocation on reply using larger size
875 	 * buffer aligned on 1k boundaty.
876 	 */
877 	alloc_len = roundup2(len, 1024);
878 	if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
879 		senderr(ENOBUFS);
880 
881 	m_copydata(m, 0, len, (caddr_t)rtm);
882 	bzero(&info, sizeof(info));
883 	bzero(&w, sizeof(w));
884 
885 	if (rtm->rtm_version != RTM_VERSION) {
886 		/* Do not touch message since format is unknown */
887 		free(rtm, M_TEMP);
888 		rtm = NULL;
889 		senderr(EPROTONOSUPPORT);
890 	}
891 
892 	/*
893 	 * Starting from here, it is possible
894 	 * to alter original message and insert
895 	 * caller PID and error value.
896 	 */
897 
898 	if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) {
899 		senderr(error);
900 	}
901 
902 	saf = info.rti_info[RTAX_DST]->sa_family;
903 
904 	/* support for new ARP code */
905 	if (rtm->rtm_flags & RTF_LLDATA) {
906 		error = lla_rt_output(rtm, &info);
907 #ifdef INET6
908 		if (error == 0)
909 			rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
910 #endif
911 		goto flush;
912 	}
913 
914 	switch (rtm->rtm_type) {
915 		struct rtentry *saved_nrt;
916 
917 	case RTM_ADD:
918 	case RTM_CHANGE:
919 		if (rtm->rtm_type == RTM_ADD) {
920 			if (info.rti_info[RTAX_GATEWAY] == NULL)
921 				senderr(EINVAL);
922 		}
923 		saved_nrt = NULL;
924 		error = rtrequest1_fib(rtm->rtm_type, &info, &saved_nrt,
925 		    fibnum);
926 		if (error == 0 && saved_nrt != NULL) {
927 #ifdef INET6
928 			rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
929 #endif
930 			RT_LOCK(saved_nrt);
931 			rtm->rtm_index = saved_nrt->rt_nhop->nh_ifp->if_index;
932 			RT_UNLOCK(saved_nrt);
933 		}
934 		break;
935 
936 	case RTM_DELETE:
937 		saved_nrt = NULL;
938 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, fibnum);
939 		if (error == 0) {
940 			RT_LOCK(saved_nrt);
941 			rt = saved_nrt;
942 			goto report;
943 		}
944 #ifdef INET6
945 		/* rt_msg2() will not be used when RTM_DELETE fails. */
946 		rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
947 #endif
948 		break;
949 
950 	case RTM_GET:
951 		error = handle_rtm_get(&info, fibnum, rtm, &rt);
952 		if (error != 0)
953 			senderr(error);
954 
955 report:
956 		RT_LOCK_ASSERT(rt);
957 		if (!can_export_rte(curthread->td_ucred, rt)) {
958 			RT_UNLOCK(rt);
959 			senderr(ESRCH);
960 		}
961 		error = update_rtm_from_rte(&info, &rtm, alloc_len, rt);
962 		/*
963 		 * Note that some sockaddr pointers may have changed to
964 		 * point to memory outsize @rtm. Some may be pointing
965 		 * to the on-stack variables.
966 		 * Given that, any pointer in @info CANNOT BE USED.
967 		 */
968 
969 		/*
970 		 * scopeid deembedding has been performed while
971 		 * writing updated rtm in rtsock_msg_buffer().
972 		 * With that in mind, skip deembedding procedure below.
973 		 */
974 #ifdef INET6
975 		rti_need_deembed = 0;
976 #endif
977 		RT_UNLOCK(rt);
978 		if (error != 0)
979 			senderr(error);
980 		break;
981 
982 	default:
983 		senderr(EOPNOTSUPP);
984 	}
985 
986 flush:
987 	NET_EPOCH_EXIT(et);
988 	rt = NULL;
989 
990 #ifdef INET6
991 	if (rtm != NULL) {
992 		if (rti_need_deembed) {
993 			/* sin6_scope_id is recovered before sending rtm. */
994 			sin6 = (struct sockaddr_in6 *)&ss;
995 			for (i = 0; i < RTAX_MAX; i++) {
996 				if (info.rti_info[i] == NULL)
997 					continue;
998 				if (info.rti_info[i]->sa_family != AF_INET6)
999 					continue;
1000 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1001 				if (sa6_recoverscope(sin6) == 0)
1002 					bcopy(sin6, info.rti_info[i],
1003 						    sizeof(*sin6));
1004 			}
1005 		}
1006 	}
1007 #endif
1008 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1009 
1010 	return (error);
1011 }
1012 
1013 /*
1014  * Sends the prepared reply message in @rtm to all rtsock clients.
1015  * Frees @m and @rtm.
1016  *
1017  */
1018 static void
1019 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1020     sa_family_t saf, u_int fibnum, int rtm_errno)
1021 {
1022 	struct rawcb *rp = NULL;
1023 
1024 	/*
1025 	 * Check to see if we don't want our own messages.
1026 	 */
1027 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1028 		if (V_route_cb.any_count <= 1) {
1029 			if (rtm != NULL)
1030 				free(rtm, M_TEMP);
1031 			m_freem(m);
1032 			return;
1033 		}
1034 		/* There is another listener, so construct message */
1035 		rp = sotorawcb(so);
1036 	}
1037 
1038 	if (rtm != NULL) {
1039 		if (rtm_errno!= 0)
1040 			rtm->rtm_errno = rtm_errno;
1041 		else
1042 			rtm->rtm_flags |= RTF_DONE;
1043 
1044 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1045 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1046 			m_freem(m);
1047 			m = NULL;
1048 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1049 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1050 
1051 		free(rtm, M_TEMP);
1052 	}
1053 	if (m != NULL) {
1054 		M_SETFIB(m, fibnum);
1055 		m->m_flags |= RTS_FILTER_FIB;
1056 		if (rp) {
1057 			/*
1058 			 * XXX insure we don't get a copy by
1059 			 * invalidating our protocol
1060 			 */
1061 			unsigned short family = rp->rcb_proto.sp_family;
1062 			rp->rcb_proto.sp_family = 0;
1063 			rt_dispatch(m, saf);
1064 			rp->rcb_proto.sp_family = family;
1065 		} else
1066 			rt_dispatch(m, saf);
1067 	}
1068 }
1069 
1070 
1071 static void
1072 rt_getmetrics(const struct rtentry *rt, struct rt_metrics *out)
1073 {
1074 
1075 	bzero(out, sizeof(*out));
1076 	out->rmx_mtu = rt->rt_nhop->nh_mtu;
1077 	out->rmx_weight = rt->rt_weight;
1078 	out->rmx_nhidx = nhop_get_idx(rt->rt_nhop);
1079 	/* Kernel -> userland timebase conversion. */
1080 	out->rmx_expire = rt->rt_expire ?
1081 	    rt->rt_expire - time_uptime + time_second : 0;
1082 }
1083 
1084 /*
1085  * Extract the addresses of the passed sockaddrs.
1086  * Do a little sanity checking so as to avoid bad memory references.
1087  * This data is derived straight from userland.
1088  */
1089 static int
1090 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1091 {
1092 	struct sockaddr *sa;
1093 	int i;
1094 
1095 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1096 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1097 			continue;
1098 		sa = (struct sockaddr *)cp;
1099 		/*
1100 		 * It won't fit.
1101 		 */
1102 		if (cp + sa->sa_len > cplim)
1103 			return (EINVAL);
1104 		/*
1105 		 * there are no more.. quit now
1106 		 * If there are more bits, they are in error.
1107 		 * I've seen this. route(1) can evidently generate these.
1108 		 * This causes kernel to core dump.
1109 		 * for compatibility, If we see this, point to a safe address.
1110 		 */
1111 		if (sa->sa_len == 0) {
1112 			rtinfo->rti_info[i] = &sa_zero;
1113 			return (0); /* should be EINVAL but for compat */
1114 		}
1115 		/* accept it */
1116 #ifdef INET6
1117 		if (sa->sa_family == AF_INET6)
1118 			sa6_embedscope((struct sockaddr_in6 *)sa,
1119 			    V_ip6_use_defzone);
1120 #endif
1121 		rtinfo->rti_info[i] = sa;
1122 		cp += SA_SIZE(sa);
1123 	}
1124 	return (0);
1125 }
1126 
1127 /*
1128  * Fill in @dmask with valid netmask leaving original @smask
1129  * intact. Mostly used with radix netmasks.
1130  */
1131 struct sockaddr *
1132 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1133     struct sockaddr_storage *dmask)
1134 {
1135 	if (dst == NULL || smask == NULL)
1136 		return (NULL);
1137 
1138 	memset(dmask, 0, dst->sa_len);
1139 	memcpy(dmask, smask, smask->sa_len);
1140 	dmask->ss_len = dst->sa_len;
1141 	dmask->ss_family = dst->sa_family;
1142 
1143 	return ((struct sockaddr *)dmask);
1144 }
1145 
1146 /*
1147  * Writes information related to @rtinfo object to newly-allocated mbuf.
1148  * Assumes MCLBYTES is enough to construct any message.
1149  * Used for OS notifications of vaious events (if/ifa announces,etc)
1150  *
1151  * Returns allocated mbuf or NULL on failure.
1152  */
1153 static struct mbuf *
1154 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1155 {
1156 	struct rt_msghdr *rtm;
1157 	struct mbuf *m;
1158 	int i;
1159 	struct sockaddr *sa;
1160 #ifdef INET6
1161 	struct sockaddr_storage ss;
1162 	struct sockaddr_in6 *sin6;
1163 #endif
1164 	int len, dlen;
1165 
1166 	switch (type) {
1167 
1168 	case RTM_DELADDR:
1169 	case RTM_NEWADDR:
1170 		len = sizeof(struct ifa_msghdr);
1171 		break;
1172 
1173 	case RTM_DELMADDR:
1174 	case RTM_NEWMADDR:
1175 		len = sizeof(struct ifma_msghdr);
1176 		break;
1177 
1178 	case RTM_IFINFO:
1179 		len = sizeof(struct if_msghdr);
1180 		break;
1181 
1182 	case RTM_IFANNOUNCE:
1183 	case RTM_IEEE80211:
1184 		len = sizeof(struct if_announcemsghdr);
1185 		break;
1186 
1187 	default:
1188 		len = sizeof(struct rt_msghdr);
1189 	}
1190 
1191 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1192 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1193 	if (len > MHLEN)
1194 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1195 	else
1196 		m = m_gethdr(M_NOWAIT, MT_DATA);
1197 	if (m == NULL)
1198 		return (m);
1199 
1200 	m->m_pkthdr.len = m->m_len = len;
1201 	rtm = mtod(m, struct rt_msghdr *);
1202 	bzero((caddr_t)rtm, len);
1203 	for (i = 0; i < RTAX_MAX; i++) {
1204 		if ((sa = rtinfo->rti_info[i]) == NULL)
1205 			continue;
1206 		rtinfo->rti_addrs |= (1 << i);
1207 		dlen = SA_SIZE(sa);
1208 #ifdef INET6
1209 		if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1210 			sin6 = (struct sockaddr_in6 *)&ss;
1211 			bcopy(sa, sin6, sizeof(*sin6));
1212 			if (sa6_recoverscope(sin6) == 0)
1213 				sa = (struct sockaddr *)sin6;
1214 		}
1215 #endif
1216 		m_copyback(m, len, dlen, (caddr_t)sa);
1217 		len += dlen;
1218 	}
1219 	if (m->m_pkthdr.len != len) {
1220 		m_freem(m);
1221 		return (NULL);
1222 	}
1223 	rtm->rtm_msglen = len;
1224 	rtm->rtm_version = RTM_VERSION;
1225 	rtm->rtm_type = type;
1226 	return (m);
1227 }
1228 
1229 /*
1230  * Writes information related to @rtinfo object to preallocated buffer.
1231  * Stores needed size in @plen. If @w is NULL, calculates size without
1232  * writing.
1233  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1234  *
1235  * Returns 0 on success.
1236  *
1237  */
1238 static int
1239 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1240 {
1241 	int i;
1242 	int len, buflen = 0, dlen;
1243 	caddr_t cp = NULL;
1244 	struct rt_msghdr *rtm = NULL;
1245 #ifdef INET6
1246 	struct sockaddr_storage ss;
1247 	struct sockaddr_in6 *sin6;
1248 #endif
1249 #ifdef COMPAT_FREEBSD32
1250 	bool compat32 = false;
1251 #endif
1252 
1253 	switch (type) {
1254 
1255 	case RTM_DELADDR:
1256 	case RTM_NEWADDR:
1257 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1258 #ifdef COMPAT_FREEBSD32
1259 			if (w->w_req->flags & SCTL_MASK32) {
1260 				len = sizeof(struct ifa_msghdrl32);
1261 				compat32 = true;
1262 			} else
1263 #endif
1264 				len = sizeof(struct ifa_msghdrl);
1265 		} else
1266 			len = sizeof(struct ifa_msghdr);
1267 		break;
1268 
1269 	case RTM_IFINFO:
1270 #ifdef COMPAT_FREEBSD32
1271 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1272 			if (w->w_op == NET_RT_IFLISTL)
1273 				len = sizeof(struct if_msghdrl32);
1274 			else
1275 				len = sizeof(struct if_msghdr32);
1276 			compat32 = true;
1277 			break;
1278 		}
1279 #endif
1280 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1281 			len = sizeof(struct if_msghdrl);
1282 		else
1283 			len = sizeof(struct if_msghdr);
1284 		break;
1285 
1286 	case RTM_NEWMADDR:
1287 		len = sizeof(struct ifma_msghdr);
1288 		break;
1289 
1290 	default:
1291 		len = sizeof(struct rt_msghdr);
1292 	}
1293 
1294 	if (w != NULL) {
1295 		rtm = (struct rt_msghdr *)w->w_tmem;
1296 		buflen = w->w_tmemsize - len;
1297 		cp = (caddr_t)w->w_tmem + len;
1298 	}
1299 
1300 	rtinfo->rti_addrs = 0;
1301 	for (i = 0; i < RTAX_MAX; i++) {
1302 		struct sockaddr *sa;
1303 
1304 		if ((sa = rtinfo->rti_info[i]) == NULL)
1305 			continue;
1306 		rtinfo->rti_addrs |= (1 << i);
1307 #ifdef COMPAT_FREEBSD32
1308 		if (compat32)
1309 			dlen = SA_SIZE32(sa);
1310 		else
1311 #endif
1312 			dlen = SA_SIZE(sa);
1313 		if (cp != NULL && buflen >= dlen) {
1314 #ifdef INET6
1315 			if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1316 				sin6 = (struct sockaddr_in6 *)&ss;
1317 				bcopy(sa, sin6, sizeof(*sin6));
1318 				if (sa6_recoverscope(sin6) == 0)
1319 					sa = (struct sockaddr *)sin6;
1320 			}
1321 #endif
1322 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1323 			cp += dlen;
1324 			buflen -= dlen;
1325 		} else if (cp != NULL) {
1326 			/*
1327 			 * Buffer too small. Count needed size
1328 			 * and return with error.
1329 			 */
1330 			cp = NULL;
1331 		}
1332 
1333 		len += dlen;
1334 	}
1335 
1336 	if (cp != NULL) {
1337 		dlen = ALIGN(len) - len;
1338 		if (buflen < dlen)
1339 			cp = NULL;
1340 		else {
1341 			bzero(cp, dlen);
1342 			cp += dlen;
1343 			buflen -= dlen;
1344 		}
1345 	}
1346 	len = ALIGN(len);
1347 
1348 	if (cp != NULL) {
1349 		/* fill header iff buffer is large enough */
1350 		rtm->rtm_version = RTM_VERSION;
1351 		rtm->rtm_type = type;
1352 		rtm->rtm_msglen = len;
1353 	}
1354 
1355 	*plen = len;
1356 
1357 	if (w != NULL && cp == NULL)
1358 		return (ENOBUFS);
1359 
1360 	return (0);
1361 }
1362 
1363 /*
1364  * This routine is called to generate a message from the routing
1365  * socket indicating that a redirect has occurred, a routing lookup
1366  * has failed, or that a protocol has detected timeouts to a particular
1367  * destination.
1368  */
1369 void
1370 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1371     int fibnum)
1372 {
1373 	struct rt_msghdr *rtm;
1374 	struct mbuf *m;
1375 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1376 
1377 	if (V_route_cb.any_count == 0)
1378 		return;
1379 	m = rtsock_msg_mbuf(type, rtinfo);
1380 	if (m == NULL)
1381 		return;
1382 
1383 	if (fibnum != RT_ALL_FIBS) {
1384 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1385 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1386 		M_SETFIB(m, fibnum);
1387 		m->m_flags |= RTS_FILTER_FIB;
1388 	}
1389 
1390 	rtm = mtod(m, struct rt_msghdr *);
1391 	rtm->rtm_flags = RTF_DONE | flags;
1392 	rtm->rtm_errno = error;
1393 	rtm->rtm_addrs = rtinfo->rti_addrs;
1394 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1395 }
1396 
1397 void
1398 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1399 {
1400 
1401 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1402 }
1403 
1404 /*
1405  * This routine is called to generate a message from the routing
1406  * socket indicating that the status of a network interface has changed.
1407  */
1408 void
1409 rt_ifmsg(struct ifnet *ifp)
1410 {
1411 	struct if_msghdr *ifm;
1412 	struct mbuf *m;
1413 	struct rt_addrinfo info;
1414 
1415 	if (V_route_cb.any_count == 0)
1416 		return;
1417 	bzero((caddr_t)&info, sizeof(info));
1418 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1419 	if (m == NULL)
1420 		return;
1421 	ifm = mtod(m, struct if_msghdr *);
1422 	ifm->ifm_index = ifp->if_index;
1423 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1424 	if_data_copy(ifp, &ifm->ifm_data);
1425 	ifm->ifm_addrs = 0;
1426 	rt_dispatch(m, AF_UNSPEC);
1427 }
1428 
1429 /*
1430  * Announce interface address arrival/withdraw.
1431  * Please do not call directly, use rt_addrmsg().
1432  * Assume input data to be valid.
1433  * Returns 0 on success.
1434  */
1435 int
1436 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1437 {
1438 	struct rt_addrinfo info;
1439 	struct sockaddr *sa;
1440 	int ncmd;
1441 	struct mbuf *m;
1442 	struct ifa_msghdr *ifam;
1443 	struct ifnet *ifp = ifa->ifa_ifp;
1444 	struct sockaddr_storage ss;
1445 
1446 	if (V_route_cb.any_count == 0)
1447 		return (0);
1448 
1449 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1450 
1451 	bzero((caddr_t)&info, sizeof(info));
1452 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1453 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1454 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1455 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1456 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1457 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1458 		return (ENOBUFS);
1459 	ifam = mtod(m, struct ifa_msghdr *);
1460 	ifam->ifam_index = ifp->if_index;
1461 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1462 	ifam->ifam_flags = ifa->ifa_flags;
1463 	ifam->ifam_addrs = info.rti_addrs;
1464 
1465 	if (fibnum != RT_ALL_FIBS) {
1466 		M_SETFIB(m, fibnum);
1467 		m->m_flags |= RTS_FILTER_FIB;
1468 	}
1469 
1470 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1471 
1472 	return (0);
1473 }
1474 
1475 /*
1476  * Announce route addition/removal to rtsock based on @rt data.
1477  * Callers are advives to use rt_routemsg() instead of using this
1478  *  function directly.
1479  * Assume @rt data is consistent.
1480  *
1481  * Returns 0 on success.
1482  */
1483 int
1484 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs,
1485     int fibnum)
1486 {
1487 	struct sockaddr_storage ss;
1488 	struct rt_addrinfo info;
1489 
1490 	if (V_route_cb.any_count == 0)
1491 		return (0);
1492 
1493 	bzero((caddr_t)&info, sizeof(info));
1494 	info.rti_info[RTAX_DST] = rt_key(rt);
1495 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss);
1496 	info.rti_info[RTAX_GATEWAY] = &rt->rt_nhop->gw_sa;
1497 	info.rti_flags = rt->rt_flags;
1498 	info.rti_ifp = ifp;
1499 
1500 	return (rtsock_routemsg_info(cmd, &info, fibnum));
1501 }
1502 
1503 int
1504 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1505 {
1506 	struct rt_msghdr *rtm;
1507 	struct sockaddr *sa;
1508 	struct mbuf *m;
1509 
1510 	if (V_route_cb.any_count == 0)
1511 		return (0);
1512 
1513 	if (info->rti_flags & RTF_HOST)
1514 		info->rti_info[RTAX_NETMASK] = NULL;
1515 
1516 	m = rtsock_msg_mbuf(cmd, info);
1517 	if (m == NULL)
1518 		return (ENOBUFS);
1519 
1520 	if (fibnum != RT_ALL_FIBS) {
1521 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1522 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1523 		M_SETFIB(m, fibnum);
1524 		m->m_flags |= RTS_FILTER_FIB;
1525 	}
1526 
1527 	rtm = mtod(m, struct rt_msghdr *);
1528 	rtm->rtm_addrs = info->rti_addrs;
1529 	if (info->rti_ifp != NULL)
1530 		rtm->rtm_index = info->rti_ifp->if_index;
1531 	/* Add RTF_DONE to indicate command 'completion' required by API */
1532 	info->rti_flags |= RTF_DONE;
1533 	/* Reported routes has to be up */
1534 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1535 		info->rti_flags |= RTF_UP;
1536 	rtm->rtm_flags = info->rti_flags;
1537 
1538 	sa = info->rti_info[RTAX_DST];
1539 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1540 
1541 	return (0);
1542 }
1543 
1544 /*
1545  * This is the analogue to the rt_newaddrmsg which performs the same
1546  * function but for multicast group memberhips.  This is easier since
1547  * there is no route state to worry about.
1548  */
1549 void
1550 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1551 {
1552 	struct rt_addrinfo info;
1553 	struct mbuf *m = NULL;
1554 	struct ifnet *ifp = ifma->ifma_ifp;
1555 	struct ifma_msghdr *ifmam;
1556 
1557 	if (V_route_cb.any_count == 0)
1558 		return;
1559 
1560 	bzero((caddr_t)&info, sizeof(info));
1561 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1562 	if (ifp && ifp->if_addr)
1563 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1564 	else
1565 		info.rti_info[RTAX_IFP] = NULL;
1566 	/*
1567 	 * If a link-layer address is present, present it as a ``gateway''
1568 	 * (similarly to how ARP entries, e.g., are presented).
1569 	 */
1570 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1571 	m = rtsock_msg_mbuf(cmd, &info);
1572 	if (m == NULL)
1573 		return;
1574 	ifmam = mtod(m, struct ifma_msghdr *);
1575 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1576 	    __func__));
1577 	ifmam->ifmam_index = ifp->if_index;
1578 	ifmam->ifmam_addrs = info.rti_addrs;
1579 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1580 }
1581 
1582 static struct mbuf *
1583 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1584 	struct rt_addrinfo *info)
1585 {
1586 	struct if_announcemsghdr *ifan;
1587 	struct mbuf *m;
1588 
1589 	if (V_route_cb.any_count == 0)
1590 		return NULL;
1591 	bzero((caddr_t)info, sizeof(*info));
1592 	m = rtsock_msg_mbuf(type, info);
1593 	if (m != NULL) {
1594 		ifan = mtod(m, struct if_announcemsghdr *);
1595 		ifan->ifan_index = ifp->if_index;
1596 		strlcpy(ifan->ifan_name, ifp->if_xname,
1597 			sizeof(ifan->ifan_name));
1598 		ifan->ifan_what = what;
1599 	}
1600 	return m;
1601 }
1602 
1603 /*
1604  * This is called to generate routing socket messages indicating
1605  * IEEE80211 wireless events.
1606  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1607  */
1608 void
1609 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1610 {
1611 	struct mbuf *m;
1612 	struct rt_addrinfo info;
1613 
1614 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1615 	if (m != NULL) {
1616 		/*
1617 		 * Append the ieee80211 data.  Try to stick it in the
1618 		 * mbuf containing the ifannounce msg; otherwise allocate
1619 		 * a new mbuf and append.
1620 		 *
1621 		 * NB: we assume m is a single mbuf.
1622 		 */
1623 		if (data_len > M_TRAILINGSPACE(m)) {
1624 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1625 			if (n == NULL) {
1626 				m_freem(m);
1627 				return;
1628 			}
1629 			bcopy(data, mtod(n, void *), data_len);
1630 			n->m_len = data_len;
1631 			m->m_next = n;
1632 		} else if (data_len > 0) {
1633 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1634 			m->m_len += data_len;
1635 		}
1636 		if (m->m_flags & M_PKTHDR)
1637 			m->m_pkthdr.len += data_len;
1638 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1639 		rt_dispatch(m, AF_UNSPEC);
1640 	}
1641 }
1642 
1643 /*
1644  * This is called to generate routing socket messages indicating
1645  * network interface arrival and departure.
1646  */
1647 void
1648 rt_ifannouncemsg(struct ifnet *ifp, int what)
1649 {
1650 	struct mbuf *m;
1651 	struct rt_addrinfo info;
1652 
1653 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1654 	if (m != NULL)
1655 		rt_dispatch(m, AF_UNSPEC);
1656 }
1657 
1658 static void
1659 rt_dispatch(struct mbuf *m, sa_family_t saf)
1660 {
1661 	struct m_tag *tag;
1662 
1663 	/*
1664 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1665 	 * use when injecting the mbuf into the routing socket buffer from
1666 	 * the netisr.
1667 	 */
1668 	if (saf != AF_UNSPEC) {
1669 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1670 		    M_NOWAIT);
1671 		if (tag == NULL) {
1672 			m_freem(m);
1673 			return;
1674 		}
1675 		*(unsigned short *)(tag + 1) = saf;
1676 		m_tag_prepend(m, tag);
1677 	}
1678 #ifdef VIMAGE
1679 	if (V_loif)
1680 		m->m_pkthdr.rcvif = V_loif;
1681 	else {
1682 		m_freem(m);
1683 		return;
1684 	}
1685 #endif
1686 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1687 }
1688 
1689 /*
1690  * Checks if rte can be exported v.r.t jails/vnets.
1691  *
1692  * Returns 1 if it can, 0 otherwise.
1693  */
1694 static int
1695 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt)
1696 {
1697 
1698 	if ((rt->rt_flags & RTF_HOST) == 0
1699 	    ? jailed_without_vnet(td_ucred)
1700 	    : prison_if(td_ucred, rt_key_const(rt)) != 0)
1701 		return (0);
1702 	return (1);
1703 }
1704 
1705 /*
1706  * This is used in dumping the kernel table via sysctl().
1707  */
1708 static int
1709 sysctl_dumpentry(struct radix_node *rn, void *vw)
1710 {
1711 	struct walkarg *w = vw;
1712 	struct rtentry *rt = (struct rtentry *)rn;
1713 	struct nhop_object *nh;
1714 	int error = 0, size;
1715 	struct rt_addrinfo info;
1716 	struct sockaddr_storage ss;
1717 
1718 	NET_EPOCH_ASSERT();
1719 
1720 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1721 		return 0;
1722 	if (!can_export_rte(w->w_req->td->td_ucred, rt))
1723 		return (0);
1724 	bzero((caddr_t)&info, sizeof(info));
1725 	info.rti_info[RTAX_DST] = rt_key(rt);
1726 	info.rti_info[RTAX_GATEWAY] = &rt->rt_nhop->gw_sa;
1727 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1728 	    rt_mask(rt), &ss);
1729 	info.rti_info[RTAX_GENMASK] = 0;
1730 	nh = rt->rt_nhop;
1731 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
1732 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
1733 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
1734 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
1735 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
1736 	}
1737 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1738 		return (error);
1739 	if (w->w_req && w->w_tmem) {
1740 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1741 
1742 		bzero(&rtm->rtm_index,
1743 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1744 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1745 			rtm->rtm_flags = RTF_GATEWAY |
1746 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1747 		else
1748 			rtm->rtm_flags = rt->rt_flags;
1749 		rtm->rtm_flags |= nhop_get_rtflags(nh);
1750 		rt_getmetrics(rt, &rtm->rtm_rmx);
1751 		rtm->rtm_index = nh->nh_ifp->if_index;
1752 		rtm->rtm_addrs = info.rti_addrs;
1753 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1754 		return (error);
1755 	}
1756 	return (error);
1757 }
1758 
1759 static int
1760 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1761     struct rt_addrinfo *info, struct walkarg *w, int len)
1762 {
1763 	struct if_msghdrl *ifm;
1764 	struct if_data *ifd;
1765 
1766 	ifm = (struct if_msghdrl *)w->w_tmem;
1767 
1768 #ifdef COMPAT_FREEBSD32
1769 	if (w->w_req->flags & SCTL_MASK32) {
1770 		struct if_msghdrl32 *ifm32;
1771 
1772 		ifm32 = (struct if_msghdrl32 *)ifm;
1773 		ifm32->ifm_addrs = info->rti_addrs;
1774 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1775 		ifm32->ifm_index = ifp->if_index;
1776 		ifm32->_ifm_spare1 = 0;
1777 		ifm32->ifm_len = sizeof(*ifm32);
1778 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1779 		ifm32->_ifm_spare2 = 0;
1780 		ifd = &ifm32->ifm_data;
1781 	} else
1782 #endif
1783 	{
1784 		ifm->ifm_addrs = info->rti_addrs;
1785 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1786 		ifm->ifm_index = ifp->if_index;
1787 		ifm->_ifm_spare1 = 0;
1788 		ifm->ifm_len = sizeof(*ifm);
1789 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1790 		ifm->_ifm_spare2 = 0;
1791 		ifd = &ifm->ifm_data;
1792 	}
1793 
1794 	memcpy(ifd, src_ifd, sizeof(*ifd));
1795 
1796 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1797 }
1798 
1799 static int
1800 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1801     struct rt_addrinfo *info, struct walkarg *w, int len)
1802 {
1803 	struct if_msghdr *ifm;
1804 	struct if_data *ifd;
1805 
1806 	ifm = (struct if_msghdr *)w->w_tmem;
1807 
1808 #ifdef COMPAT_FREEBSD32
1809 	if (w->w_req->flags & SCTL_MASK32) {
1810 		struct if_msghdr32 *ifm32;
1811 
1812 		ifm32 = (struct if_msghdr32 *)ifm;
1813 		ifm32->ifm_addrs = info->rti_addrs;
1814 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1815 		ifm32->ifm_index = ifp->if_index;
1816 		ifm32->_ifm_spare1 = 0;
1817 		ifd = &ifm32->ifm_data;
1818 	} else
1819 #endif
1820 	{
1821 		ifm->ifm_addrs = info->rti_addrs;
1822 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1823 		ifm->ifm_index = ifp->if_index;
1824 		ifm->_ifm_spare1 = 0;
1825 		ifd = &ifm->ifm_data;
1826 	}
1827 
1828 	memcpy(ifd, src_ifd, sizeof(*ifd));
1829 
1830 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1831 }
1832 
1833 static int
1834 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1835     struct walkarg *w, int len)
1836 {
1837 	struct ifa_msghdrl *ifam;
1838 	struct if_data *ifd;
1839 
1840 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1841 
1842 #ifdef COMPAT_FREEBSD32
1843 	if (w->w_req->flags & SCTL_MASK32) {
1844 		struct ifa_msghdrl32 *ifam32;
1845 
1846 		ifam32 = (struct ifa_msghdrl32 *)ifam;
1847 		ifam32->ifam_addrs = info->rti_addrs;
1848 		ifam32->ifam_flags = ifa->ifa_flags;
1849 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1850 		ifam32->_ifam_spare1 = 0;
1851 		ifam32->ifam_len = sizeof(*ifam32);
1852 		ifam32->ifam_data_off =
1853 		    offsetof(struct ifa_msghdrl32, ifam_data);
1854 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1855 		ifd = &ifam32->ifam_data;
1856 	} else
1857 #endif
1858 	{
1859 		ifam->ifam_addrs = info->rti_addrs;
1860 		ifam->ifam_flags = ifa->ifa_flags;
1861 		ifam->ifam_index = ifa->ifa_ifp->if_index;
1862 		ifam->_ifam_spare1 = 0;
1863 		ifam->ifam_len = sizeof(*ifam);
1864 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1865 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1866 		ifd = &ifam->ifam_data;
1867 	}
1868 
1869 	bzero(ifd, sizeof(*ifd));
1870 	ifd->ifi_datalen = sizeof(struct if_data);
1871 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1872 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1873 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1874 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1875 
1876 	/* Fixup if_data carp(4) vhid. */
1877 	if (carp_get_vhid_p != NULL)
1878 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1879 
1880 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1881 }
1882 
1883 static int
1884 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1885     struct walkarg *w, int len)
1886 {
1887 	struct ifa_msghdr *ifam;
1888 
1889 	ifam = (struct ifa_msghdr *)w->w_tmem;
1890 	ifam->ifam_addrs = info->rti_addrs;
1891 	ifam->ifam_flags = ifa->ifa_flags;
1892 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1893 	ifam->_ifam_spare1 = 0;
1894 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1895 
1896 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1897 }
1898 
1899 static int
1900 sysctl_iflist(int af, struct walkarg *w)
1901 {
1902 	struct ifnet *ifp;
1903 	struct ifaddr *ifa;
1904 	struct if_data ifd;
1905 	struct rt_addrinfo info;
1906 	int len, error = 0;
1907 	struct sockaddr_storage ss;
1908 
1909 	bzero((caddr_t)&info, sizeof(info));
1910 	bzero(&ifd, sizeof(ifd));
1911 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1912 		if (w->w_arg && w->w_arg != ifp->if_index)
1913 			continue;
1914 		if_data_copy(ifp, &ifd);
1915 		ifa = ifp->if_addr;
1916 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1917 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1918 		if (error != 0)
1919 			goto done;
1920 		info.rti_info[RTAX_IFP] = NULL;
1921 		if (w->w_req && w->w_tmem) {
1922 			if (w->w_op == NET_RT_IFLISTL)
1923 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1924 				    len);
1925 			else
1926 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1927 				    len);
1928 			if (error)
1929 				goto done;
1930 		}
1931 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1932 			if (af && af != ifa->ifa_addr->sa_family)
1933 				continue;
1934 			if (prison_if(w->w_req->td->td_ucred,
1935 			    ifa->ifa_addr) != 0)
1936 				continue;
1937 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1938 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1939 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
1940 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1941 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
1942 			if (error != 0)
1943 				goto done;
1944 			if (w->w_req && w->w_tmem) {
1945 				if (w->w_op == NET_RT_IFLISTL)
1946 					error = sysctl_iflist_ifaml(ifa, &info,
1947 					    w, len);
1948 				else
1949 					error = sysctl_iflist_ifam(ifa, &info,
1950 					    w, len);
1951 				if (error)
1952 					goto done;
1953 			}
1954 		}
1955 		info.rti_info[RTAX_IFA] = NULL;
1956 		info.rti_info[RTAX_NETMASK] = NULL;
1957 		info.rti_info[RTAX_BRD] = NULL;
1958 	}
1959 done:
1960 	return (error);
1961 }
1962 
1963 static int
1964 sysctl_ifmalist(int af, struct walkarg *w)
1965 {
1966 	struct rt_addrinfo info;
1967 	struct ifaddr *ifa;
1968 	struct ifmultiaddr *ifma;
1969 	struct ifnet *ifp;
1970 	int error, len;
1971 
1972 	NET_EPOCH_ASSERT();
1973 
1974 	error = 0;
1975 	bzero((caddr_t)&info, sizeof(info));
1976 
1977 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1978 		if (w->w_arg && w->w_arg != ifp->if_index)
1979 			continue;
1980 		ifa = ifp->if_addr;
1981 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1982 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1983 			if (af && af != ifma->ifma_addr->sa_family)
1984 				continue;
1985 			if (prison_if(w->w_req->td->td_ucred,
1986 			    ifma->ifma_addr) != 0)
1987 				continue;
1988 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1989 			info.rti_info[RTAX_GATEWAY] =
1990 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1991 			    ifma->ifma_lladdr : NULL;
1992 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
1993 			if (error != 0)
1994 				break;
1995 			if (w->w_req && w->w_tmem) {
1996 				struct ifma_msghdr *ifmam;
1997 
1998 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1999 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2000 				ifmam->ifmam_flags = 0;
2001 				ifmam->ifmam_addrs = info.rti_addrs;
2002 				ifmam->_ifmam_spare1 = 0;
2003 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2004 				if (error != 0)
2005 					break;
2006 			}
2007 		}
2008 		if (error != 0)
2009 			break;
2010 	}
2011 	return (error);
2012 }
2013 
2014 static int
2015 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2016 {
2017 	RIB_RLOCK_TRACKER;
2018 	struct epoch_tracker et;
2019 	int	*name = (int *)arg1;
2020 	u_int	namelen = arg2;
2021 	struct rib_head *rnh = NULL; /* silence compiler. */
2022 	int	i, lim, error = EINVAL;
2023 	int	fib = 0;
2024 	u_char	af;
2025 	struct	walkarg w;
2026 
2027 	name ++;
2028 	namelen--;
2029 	if (req->newptr)
2030 		return (EPERM);
2031 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP) {
2032 		if (namelen == 3)
2033 			fib = req->td->td_proc->p_fibnum;
2034 		else if (namelen == 4)
2035 			fib = (name[3] == RT_ALL_FIBS) ?
2036 			    req->td->td_proc->p_fibnum : name[3];
2037 		else
2038 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2039 		if (fib < 0 || fib >= rt_numfibs)
2040 			return (EINVAL);
2041 	} else if (namelen != 3)
2042 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2043 	af = name[0];
2044 	if (af > AF_MAX)
2045 		return (EINVAL);
2046 	bzero(&w, sizeof(w));
2047 	w.w_op = name[1];
2048 	w.w_arg = name[2];
2049 	w.w_req = req;
2050 
2051 	error = sysctl_wire_old_buffer(req, 0);
2052 	if (error)
2053 		return (error);
2054 
2055 	/*
2056 	 * Allocate reply buffer in advance.
2057 	 * All rtsock messages has maximum length of u_short.
2058 	 */
2059 	w.w_tmemsize = 65536;
2060 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2061 
2062 	NET_EPOCH_ENTER(et);
2063 	switch (w.w_op) {
2064 	case NET_RT_DUMP:
2065 	case NET_RT_FLAGS:
2066 		if (af == 0) {			/* dump all tables */
2067 			i = 1;
2068 			lim = AF_MAX;
2069 		} else				/* dump only one table */
2070 			i = lim = af;
2071 
2072 		/*
2073 		 * take care of llinfo entries, the caller must
2074 		 * specify an AF
2075 		 */
2076 		if (w.w_op == NET_RT_FLAGS &&
2077 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2078 			if (af != 0)
2079 				error = lltable_sysctl_dumparp(af, w.w_req);
2080 			else
2081 				error = EINVAL;
2082 			break;
2083 		}
2084 		/*
2085 		 * take care of routing entries
2086 		 */
2087 		for (error = 0; error == 0 && i <= lim; i++) {
2088 			rnh = rt_tables_get_rnh(fib, i);
2089 			if (rnh != NULL) {
2090 				RIB_RLOCK(rnh);
2091 			    	error = rnh->rnh_walktree(&rnh->head,
2092 				    sysctl_dumpentry, &w);
2093 				RIB_RUNLOCK(rnh);
2094 			} else if (af != 0)
2095 				error = EAFNOSUPPORT;
2096 		}
2097 		break;
2098 	case NET_RT_NHOP:
2099 		/* Allow dumping one specific af/fib at a time */
2100 		if (namelen < 4) {
2101 			error = EINVAL;
2102 			break;
2103 		}
2104 		fib = name[3];
2105 		if (fib < 0 || fib > rt_numfibs) {
2106 			error = EINVAL;
2107 			break;
2108 		}
2109 		rnh = rt_tables_get_rnh(fib, af);
2110 		if (rnh == NULL) {
2111 			error = EAFNOSUPPORT;
2112 			break;
2113 		}
2114 		if (w.w_op == NET_RT_NHOP)
2115 			error = nhops_dump_sysctl(rnh, w.w_req);
2116 		break;
2117 	case NET_RT_IFLIST:
2118 	case NET_RT_IFLISTL:
2119 		error = sysctl_iflist(af, &w);
2120 		break;
2121 
2122 	case NET_RT_IFMALIST:
2123 		error = sysctl_ifmalist(af, &w);
2124 		break;
2125 	}
2126 	NET_EPOCH_EXIT(et);
2127 
2128 	free(w.w_tmem, M_TEMP);
2129 	return (error);
2130 }
2131 
2132 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2133     sysctl_rtsock, "Return route tables and interface/address lists");
2134 
2135 /*
2136  * Definitions of protocols supported in the ROUTE domain.
2137  */
2138 
2139 static struct domain routedomain;		/* or at least forward */
2140 
2141 static struct protosw routesw[] = {
2142 {
2143 	.pr_type =		SOCK_RAW,
2144 	.pr_domain =		&routedomain,
2145 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2146 	.pr_output =		route_output,
2147 	.pr_ctlinput =		raw_ctlinput,
2148 	.pr_init =		raw_init,
2149 	.pr_usrreqs =		&route_usrreqs
2150 }
2151 };
2152 
2153 static struct domain routedomain = {
2154 	.dom_family =		PF_ROUTE,
2155 	.dom_name =		 "route",
2156 	.dom_protosw =		routesw,
2157 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
2158 };
2159 
2160 VNET_DOMAIN_SET(route);
2161 
2162