xref: /freebsd/sys/net/rtsock.c (revision bbd421cdf6d8c6102e6fd3979c5bec21ace3c2e3)
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_route.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_ctl.h>
66 #include <net/route/route_var.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/if_ether.h>
71 #include <netinet/ip_carp.h>
72 #ifdef INET6
73 #include <netinet6/in6_var.h>
74 #include <netinet6/ip6_var.h>
75 #include <netinet6/scope6_var.h>
76 #endif
77 #include <net/route/nhop.h>
78 
79 #ifdef COMPAT_FREEBSD32
80 #include <sys/mount.h>
81 #include <compat/freebsd32/freebsd32.h>
82 
83 struct if_msghdr32 {
84 	uint16_t ifm_msglen;
85 	uint8_t	ifm_version;
86 	uint8_t	ifm_type;
87 	int32_t	ifm_addrs;
88 	int32_t	ifm_flags;
89 	uint16_t ifm_index;
90 	uint16_t _ifm_spare1;
91 	struct	if_data ifm_data;
92 };
93 
94 struct if_msghdrl32 {
95 	uint16_t ifm_msglen;
96 	uint8_t	ifm_version;
97 	uint8_t	ifm_type;
98 	int32_t	ifm_addrs;
99 	int32_t	ifm_flags;
100 	uint16_t ifm_index;
101 	uint16_t _ifm_spare1;
102 	uint16_t ifm_len;
103 	uint16_t ifm_data_off;
104 	uint32_t _ifm_spare2;
105 	struct	if_data ifm_data;
106 };
107 
108 struct ifa_msghdrl32 {
109 	uint16_t ifam_msglen;
110 	uint8_t	ifam_version;
111 	uint8_t	ifam_type;
112 	int32_t	ifam_addrs;
113 	int32_t	ifam_flags;
114 	uint16_t ifam_index;
115 	uint16_t _ifam_spare1;
116 	uint16_t ifam_len;
117 	uint16_t ifam_data_off;
118 	int32_t	ifam_metric;
119 	struct	if_data ifam_data;
120 };
121 
122 #define SA_SIZE32(sa)						\
123     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
124 	sizeof(int)		:				\
125 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
126 
127 #endif /* COMPAT_FREEBSD32 */
128 
129 #define	RTS_PID_PRINTF(_fmt, ...) \
130     printf("rtsock:%s(): PID %d: " _fmt "\n", __func__, curproc->p_pid, ## __VA_ARGS__)
131 
132 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
133 
134 /* NB: these are not modified */
135 static struct	sockaddr route_src = { 2, PF_ROUTE, };
136 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
137 
138 /* These are external hooks for CARP. */
139 int	(*carp_get_vhid_p)(struct ifaddr *);
140 
141 /*
142  * Used by rtsock/raw_input callback code to decide whether to filter the update
143  * notification to a socket bound to a particular FIB.
144  */
145 #define	RTS_FILTER_FIB	M_PROTO8
146 
147 typedef struct {
148 	int	ip_count;	/* attached w/ AF_INET */
149 	int	ip6_count;	/* attached w/ AF_INET6 */
150 	int	any_count;	/* total attached */
151 } route_cb_t;
152 VNET_DEFINE_STATIC(route_cb_t, route_cb);
153 #define	V_route_cb VNET(route_cb)
154 
155 struct mtx rtsock_mtx;
156 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
157 
158 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
159 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
160 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
161 
162 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
163 
164 struct walkarg {
165 	int	family;
166 	int	w_tmemsize;
167 	int	w_op, w_arg;
168 	caddr_t	w_tmem;
169 	struct sysctl_req *w_req;
170 	struct sockaddr *dst;
171 	struct sockaddr *mask;
172 };
173 
174 static void	rts_input(struct mbuf *m);
175 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
176 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
177 			struct walkarg *w, int *plen);
178 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
179 			struct rt_addrinfo *rtinfo);
180 static int	cleanup_xaddrs(struct rt_addrinfo *info);
181 static int	sysctl_dumpentry(struct rtentry *rt, void *vw);
182 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
183 			uint32_t weight, struct walkarg *w);
184 static int	sysctl_iflist(int af, struct walkarg *w);
185 static int	sysctl_ifmalist(int af, struct walkarg *w);
186 static int	route_output(struct mbuf *m, struct socket *so, ...);
187 static void	rt_getmetrics(const struct rtentry *rt,
188 			const struct nhop_object *nh, struct rt_metrics *out);
189 static void	rt_dispatch(struct mbuf *, sa_family_t);
190 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
191 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
192 static int	update_rtm_from_rc(struct rt_addrinfo *info,
193 			struct rt_msghdr **prtm, int alloc_len,
194 			struct rib_cmd_info *rc, struct nhop_object *nh);
195 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
196 			struct mbuf *m, sa_family_t saf, u_int fibnum,
197 			int rtm_errno);
198 static bool	can_export_rte(struct ucred *td_ucred, bool rt_is_host,
199 			const struct sockaddr *rt_dst);
200 
201 static struct netisr_handler rtsock_nh = {
202 	.nh_name = "rtsock",
203 	.nh_handler = rts_input,
204 	.nh_proto = NETISR_ROUTE,
205 	.nh_policy = NETISR_POLICY_SOURCE,
206 };
207 
208 static int
209 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
210 {
211 	int error, qlimit;
212 
213 	netisr_getqlimit(&rtsock_nh, &qlimit);
214 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
215         if (error || !req->newptr)
216                 return (error);
217 	if (qlimit < 1)
218 		return (EINVAL);
219 	return (netisr_setqlimit(&rtsock_nh, qlimit));
220 }
221 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
222     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
223     0, 0, sysctl_route_netisr_maxqlen, "I",
224     "maximum routing socket dispatch queue length");
225 
226 static void
227 vnet_rts_init(void)
228 {
229 	int tmp;
230 
231 	if (IS_DEFAULT_VNET(curvnet)) {
232 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
233 			rtsock_nh.nh_qlimit = tmp;
234 		netisr_register(&rtsock_nh);
235 	}
236 #ifdef VIMAGE
237 	 else
238 		netisr_register_vnet(&rtsock_nh);
239 #endif
240 }
241 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
242     vnet_rts_init, 0);
243 
244 #ifdef VIMAGE
245 static void
246 vnet_rts_uninit(void)
247 {
248 
249 	netisr_unregister_vnet(&rtsock_nh);
250 }
251 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
252     vnet_rts_uninit, 0);
253 #endif
254 
255 static int
256 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
257     struct rawcb *rp)
258 {
259 	int fibnum;
260 
261 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
262 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
263 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
264 
265 	/* No filtering requested. */
266 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
267 		return (0);
268 
269 	/* Check if it is a rts and the fib matches the one of the socket. */
270 	fibnum = M_GETFIB(m);
271 	if (proto->sp_family != PF_ROUTE ||
272 	    rp->rcb_socket == NULL ||
273 	    rp->rcb_socket->so_fibnum == fibnum)
274 		return (0);
275 
276 	/* Filtering requested and no match, the socket shall be skipped. */
277 	return (1);
278 }
279 
280 static void
281 rts_input(struct mbuf *m)
282 {
283 	struct sockproto route_proto;
284 	unsigned short *family;
285 	struct m_tag *tag;
286 
287 	route_proto.sp_family = PF_ROUTE;
288 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
289 	if (tag != NULL) {
290 		family = (unsigned short *)(tag + 1);
291 		route_proto.sp_protocol = *family;
292 		m_tag_delete(m, tag);
293 	} else
294 		route_proto.sp_protocol = 0;
295 
296 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
297 }
298 
299 /*
300  * It really doesn't make any sense at all for this code to share much
301  * with raw_usrreq.c, since its functionality is so restricted.  XXX
302  */
303 static void
304 rts_abort(struct socket *so)
305 {
306 
307 	raw_usrreqs.pru_abort(so);
308 }
309 
310 static void
311 rts_close(struct socket *so)
312 {
313 
314 	raw_usrreqs.pru_close(so);
315 }
316 
317 /* pru_accept is EOPNOTSUPP */
318 
319 static int
320 rts_attach(struct socket *so, int proto, struct thread *td)
321 {
322 	struct rawcb *rp;
323 	int error;
324 
325 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
326 
327 	/* XXX */
328 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
329 
330 	so->so_pcb = (caddr_t)rp;
331 	so->so_fibnum = td->td_proc->p_fibnum;
332 	error = raw_attach(so, proto);
333 	rp = sotorawcb(so);
334 	if (error) {
335 		so->so_pcb = NULL;
336 		free(rp, M_PCB);
337 		return error;
338 	}
339 	RTSOCK_LOCK();
340 	switch(rp->rcb_proto.sp_protocol) {
341 	case AF_INET:
342 		V_route_cb.ip_count++;
343 		break;
344 	case AF_INET6:
345 		V_route_cb.ip6_count++;
346 		break;
347 	}
348 	V_route_cb.any_count++;
349 	RTSOCK_UNLOCK();
350 	soisconnected(so);
351 	so->so_options |= SO_USELOOPBACK;
352 	return 0;
353 }
354 
355 static int
356 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
357 {
358 
359 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
360 }
361 
362 static int
363 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
364 {
365 
366 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
367 }
368 
369 /* pru_connect2 is EOPNOTSUPP */
370 /* pru_control is EOPNOTSUPP */
371 
372 static void
373 rts_detach(struct socket *so)
374 {
375 	struct rawcb *rp = sotorawcb(so);
376 
377 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
378 
379 	RTSOCK_LOCK();
380 	switch(rp->rcb_proto.sp_protocol) {
381 	case AF_INET:
382 		V_route_cb.ip_count--;
383 		break;
384 	case AF_INET6:
385 		V_route_cb.ip6_count--;
386 		break;
387 	}
388 	V_route_cb.any_count--;
389 	RTSOCK_UNLOCK();
390 	raw_usrreqs.pru_detach(so);
391 }
392 
393 static int
394 rts_disconnect(struct socket *so)
395 {
396 
397 	return (raw_usrreqs.pru_disconnect(so));
398 }
399 
400 /* pru_listen is EOPNOTSUPP */
401 
402 static int
403 rts_peeraddr(struct socket *so, struct sockaddr **nam)
404 {
405 
406 	return (raw_usrreqs.pru_peeraddr(so, nam));
407 }
408 
409 /* pru_rcvd is EOPNOTSUPP */
410 /* pru_rcvoob is EOPNOTSUPP */
411 
412 static int
413 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
414 	 struct mbuf *control, struct thread *td)
415 {
416 
417 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
418 }
419 
420 /* pru_sense is null */
421 
422 static int
423 rts_shutdown(struct socket *so)
424 {
425 
426 	return (raw_usrreqs.pru_shutdown(so));
427 }
428 
429 static int
430 rts_sockaddr(struct socket *so, struct sockaddr **nam)
431 {
432 
433 	return (raw_usrreqs.pru_sockaddr(so, nam));
434 }
435 
436 static struct pr_usrreqs route_usrreqs = {
437 	.pru_abort =		rts_abort,
438 	.pru_attach =		rts_attach,
439 	.pru_bind =		rts_bind,
440 	.pru_connect =		rts_connect,
441 	.pru_detach =		rts_detach,
442 	.pru_disconnect =	rts_disconnect,
443 	.pru_peeraddr =		rts_peeraddr,
444 	.pru_send =		rts_send,
445 	.pru_shutdown =		rts_shutdown,
446 	.pru_sockaddr =		rts_sockaddr,
447 	.pru_close =		rts_close,
448 };
449 
450 #ifndef _SOCKADDR_UNION_DEFINED
451 #define	_SOCKADDR_UNION_DEFINED
452 /*
453  * The union of all possible address formats we handle.
454  */
455 union sockaddr_union {
456 	struct sockaddr		sa;
457 	struct sockaddr_in	sin;
458 	struct sockaddr_in6	sin6;
459 };
460 #endif /* _SOCKADDR_UNION_DEFINED */
461 
462 static int
463 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
464     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
465 {
466 #if defined(INET) || defined(INET6)
467 	struct epoch_tracker et;
468 #endif
469 
470 	/* First, see if the returned address is part of the jail. */
471 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
472 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
473 		return (0);
474 	}
475 
476 	switch (info->rti_info[RTAX_DST]->sa_family) {
477 #ifdef INET
478 	case AF_INET:
479 	{
480 		struct in_addr ia;
481 		struct ifaddr *ifa;
482 		int found;
483 
484 		found = 0;
485 		/*
486 		 * Try to find an address on the given outgoing interface
487 		 * that belongs to the jail.
488 		 */
489 		NET_EPOCH_ENTER(et);
490 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
491 			struct sockaddr *sa;
492 			sa = ifa->ifa_addr;
493 			if (sa->sa_family != AF_INET)
494 				continue;
495 			ia = ((struct sockaddr_in *)sa)->sin_addr;
496 			if (prison_check_ip4(cred, &ia) == 0) {
497 				found = 1;
498 				break;
499 			}
500 		}
501 		NET_EPOCH_EXIT(et);
502 		if (!found) {
503 			/*
504 			 * As a last resort return the 'default' jail address.
505 			 */
506 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
507 			    sin_addr;
508 			if (prison_get_ip4(cred, &ia) != 0)
509 				return (ESRCH);
510 		}
511 		bzero(&saun->sin, sizeof(struct sockaddr_in));
512 		saun->sin.sin_len = sizeof(struct sockaddr_in);
513 		saun->sin.sin_family = AF_INET;
514 		saun->sin.sin_addr.s_addr = ia.s_addr;
515 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
516 		break;
517 	}
518 #endif
519 #ifdef INET6
520 	case AF_INET6:
521 	{
522 		struct in6_addr ia6;
523 		struct ifaddr *ifa;
524 		int found;
525 
526 		found = 0;
527 		/*
528 		 * Try to find an address on the given outgoing interface
529 		 * that belongs to the jail.
530 		 */
531 		NET_EPOCH_ENTER(et);
532 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
533 			struct sockaddr *sa;
534 			sa = ifa->ifa_addr;
535 			if (sa->sa_family != AF_INET6)
536 				continue;
537 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
538 			    &ia6, sizeof(struct in6_addr));
539 			if (prison_check_ip6(cred, &ia6) == 0) {
540 				found = 1;
541 				break;
542 			}
543 		}
544 		NET_EPOCH_EXIT(et);
545 		if (!found) {
546 			/*
547 			 * As a last resort return the 'default' jail address.
548 			 */
549 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
550 			    sin6_addr;
551 			if (prison_get_ip6(cred, &ia6) != 0)
552 				return (ESRCH);
553 		}
554 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
555 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
556 		saun->sin6.sin6_family = AF_INET6;
557 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
558 		if (sa6_recoverscope(&saun->sin6) != 0)
559 			return (ESRCH);
560 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
561 		break;
562 	}
563 #endif
564 	default:
565 		return (ESRCH);
566 	}
567 	return (0);
568 }
569 
570 static int
571 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
572 {
573 	struct ifaddr *ifa;
574 	sa_family_t saf;
575 
576 	if (V_loif == NULL) {
577 		RTS_PID_PRINTF("Unable to add blackhole/reject nhop without loopback");
578 		return (ENOTSUP);
579 	}
580 	info->rti_ifp = V_loif;
581 
582 	saf = info->rti_info[RTAX_DST]->sa_family;
583 
584 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
585 		if (ifa->ifa_addr->sa_family == saf) {
586 			info->rti_ifa = ifa;
587 			break;
588 		}
589 	}
590 	if (info->rti_ifa == NULL)
591 		return (ENOTSUP);
592 
593 	bzero(saun, sizeof(union sockaddr_union));
594 	switch (saf) {
595 #ifdef INET
596 	case AF_INET:
597 		saun->sin.sin_family = AF_INET;
598 		saun->sin.sin_len = sizeof(struct sockaddr_in);
599 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
600 		break;
601 #endif
602 #ifdef INET6
603 	case AF_INET6:
604 		saun->sin6.sin6_family = AF_INET6;
605 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
606 		saun->sin6.sin6_addr = in6addr_loopback;
607 		break;
608 #endif
609 	default:
610 		return (ENOTSUP);
611 	}
612 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
613 	info->rti_flags |= RTF_GATEWAY;
614 
615 	return (0);
616 }
617 
618 /*
619  * Fills in @info based on userland-provided @rtm message.
620  *
621  * Returns 0 on success.
622  */
623 static int
624 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
625 {
626 	int error;
627 	sa_family_t saf;
628 
629 	rtm->rtm_pid = curproc->p_pid;
630 	info->rti_addrs = rtm->rtm_addrs;
631 
632 	info->rti_mflags = rtm->rtm_inits;
633 	info->rti_rmx = &rtm->rtm_rmx;
634 
635 	/*
636 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
637 	 * link-local address because rtrequest requires addresses with
638 	 * embedded scope id.
639 	 */
640 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
641 		return (EINVAL);
642 
643 	info->rti_flags = rtm->rtm_flags;
644 	error = cleanup_xaddrs(info);
645 	if (error != 0)
646 		return (error);
647 	saf = info->rti_info[RTAX_DST]->sa_family;
648 	/*
649 	 * Verify that the caller has the appropriate privilege; RTM_GET
650 	 * is the only operation the non-superuser is allowed.
651 	 */
652 	if (rtm->rtm_type != RTM_GET) {
653 		error = priv_check(curthread, PRIV_NET_ROUTE);
654 		if (error != 0)
655 			return (error);
656 	}
657 
658 	/*
659 	 * The given gateway address may be an interface address.
660 	 * For example, issuing a "route change" command on a route
661 	 * entry that was created from a tunnel, and the gateway
662 	 * address given is the local end point. In this case the
663 	 * RTF_GATEWAY flag must be cleared or the destination will
664 	 * not be reachable even though there is no error message.
665 	 */
666 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
667 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
668 		struct rt_addrinfo ginfo;
669 		struct sockaddr *gdst;
670 		struct sockaddr_storage ss;
671 
672 		bzero(&ginfo, sizeof(ginfo));
673 		bzero(&ss, sizeof(ss));
674 		ss.ss_len = sizeof(ss);
675 
676 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
677 		gdst = info->rti_info[RTAX_GATEWAY];
678 
679 		/*
680 		 * A host route through the loopback interface is
681 		 * installed for each interface adddress. In pre 8.0
682 		 * releases the interface address of a PPP link type
683 		 * is not reachable locally. This behavior is fixed as
684 		 * part of the new L2/L3 redesign and rewrite work. The
685 		 * signature of this interface address route is the
686 		 * AF_LINK sa_family type of the gateway, and the
687 		 * rt_ifp has the IFF_LOOPBACK flag set.
688 		 */
689 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
690 			if (ss.ss_family == AF_LINK &&
691 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
692 				info->rti_flags &= ~RTF_GATEWAY;
693 				info->rti_flags |= RTF_GWFLAG_COMPAT;
694 			}
695 			rib_free_info(&ginfo);
696 		}
697 	}
698 
699 	return (0);
700 }
701 
702 static struct nhop_object *
703 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
704 {
705 	if (!NH_IS_NHGRP(nh))
706 		return (nh);
707 #ifdef ROUTE_MPATH
708 	struct weightened_nhop *wn;
709 	uint32_t num_nhops;
710 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
711 	if (gw == NULL)
712 		return (wn[0].nh);
713 	for (int i = 0; i < num_nhops; i++) {
714 		if (match_nhop_gw(wn[i].nh, gw))
715 			return (wn[i].nh);
716 	}
717 #endif
718 	return (NULL);
719 }
720 
721 /*
722  * Handles RTM_GET message from routing socket, returning matching rt.
723  *
724  * Returns:
725  * 0 on success, with locked and referenced matching rt in @rt_nrt
726  * errno of failure
727  */
728 static int
729 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
730     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
731 {
732 	RIB_RLOCK_TRACKER;
733 	struct rib_head *rnh;
734 	struct nhop_object *nh;
735 	sa_family_t saf;
736 
737 	saf = info->rti_info[RTAX_DST]->sa_family;
738 
739 	rnh = rt_tables_get_rnh(fibnum, saf);
740 	if (rnh == NULL)
741 		return (EAFNOSUPPORT);
742 
743 	RIB_RLOCK(rnh);
744 
745 	/*
746 	 * By (implicit) convention host route (one without netmask)
747 	 * means longest-prefix-match request and the route with netmask
748 	 * means exact-match lookup.
749 	 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
750 	 * prefixes, use original data to check for the netmask presence.
751 	 */
752 	if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
753 		/*
754 		 * Provide longest prefix match for
755 		 * address lookup (no mask).
756 		 * 'route -n get addr'
757 		 */
758 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
759 		    info->rti_info[RTAX_DST], &rnh->head);
760 	} else
761 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
762 		    info->rti_info[RTAX_DST],
763 		    info->rti_info[RTAX_NETMASK], &rnh->head);
764 
765 	if (rc->rc_rt == NULL) {
766 		RIB_RUNLOCK(rnh);
767 		return (ESRCH);
768 	}
769 
770 	nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
771 	if (nh == NULL) {
772 		RIB_RUNLOCK(rnh);
773 		return (ESRCH);
774 	}
775 	/*
776 	 * If performing proxied L2 entry insertion, and
777 	 * the actual PPP host entry is found, perform
778 	 * another search to retrieve the prefix route of
779 	 * the local end point of the PPP link.
780 	 * TODO: move this logic to userland.
781 	 */
782 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
783 		struct sockaddr laddr;
784 
785 		if (nh->nh_ifp != NULL &&
786 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
787 			struct ifaddr *ifa;
788 
789 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
790 					RT_ALL_FIBS);
791 			if (ifa != NULL)
792 				rt_maskedcopy(ifa->ifa_addr,
793 					      &laddr,
794 					      ifa->ifa_netmask);
795 		} else
796 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
797 				      &laddr,
798 				      nh->nh_ifa->ifa_netmask);
799 		/*
800 		 * refactor rt and no lock operation necessary
801 		 */
802 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
803 		    &rnh->head);
804 		if (rc->rc_rt == NULL) {
805 			RIB_RUNLOCK(rnh);
806 			return (ESRCH);
807 		}
808 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
809 		if (nh == NULL) {
810 			RIB_RUNLOCK(rnh);
811 			return (ESRCH);
812 		}
813 	}
814 	rc->rc_nh_new = nh;
815 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
816 	RIB_RUNLOCK(rnh);
817 
818 	return (0);
819 }
820 
821 static void
822 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
823 {
824 #ifdef INET
825 	if (family == AF_INET) {
826 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
827 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
828 
829 		bzero(dst4, sizeof(struct sockaddr_in));
830 		bzero(mask4, sizeof(struct sockaddr_in));
831 
832 		dst4->sin_family = AF_INET;
833 		dst4->sin_len = sizeof(struct sockaddr_in);
834 		mask4->sin_family = AF_INET;
835 		mask4->sin_len = sizeof(struct sockaddr_in);
836 	}
837 #endif
838 #ifdef INET6
839 	if (family == AF_INET6) {
840 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
841 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
842 
843 		bzero(dst6, sizeof(struct sockaddr_in6));
844 		bzero(mask6, sizeof(struct sockaddr_in6));
845 
846 		dst6->sin6_family = AF_INET6;
847 		dst6->sin6_len = sizeof(struct sockaddr_in6);
848 		mask6->sin6_family = AF_INET6;
849 		mask6->sin6_len = sizeof(struct sockaddr_in6);
850 	}
851 #endif
852 }
853 
854 static void
855 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
856     struct sockaddr *mask)
857 {
858 #ifdef INET
859 	if (dst->sa_family == AF_INET) {
860 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
861 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
862 		uint32_t scopeid = 0;
863 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
864 		    &scopeid);
865 		return;
866 	}
867 #endif
868 #ifdef INET6
869 	if (dst->sa_family == AF_INET6) {
870 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
871 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
872 		uint32_t scopeid = 0;
873 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
874 		    &mask6->sin6_addr, &scopeid);
875 		dst6->sin6_scope_id = scopeid;
876 		return;
877 	}
878 #endif
879 }
880 
881 
882 /*
883  * Update sockaddrs, flags, etc in @prtm based on @rc data.
884  * rtm can be reallocated.
885  *
886  * Returns 0 on success, along with pointer to (potentially reallocated)
887  *  rtm.
888  *
889  */
890 static int
891 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
892     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
893 {
894 	struct walkarg w;
895 	union sockaddr_union saun;
896 	struct rt_msghdr *rtm, *orig_rtm = NULL;
897 	struct ifnet *ifp;
898 	int error, len;
899 
900 	rtm = *prtm;
901 	union sockaddr_union sa_dst, sa_mask;
902 	int family = info->rti_info[RTAX_DST]->sa_family;
903 	init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
904 	export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
905 
906 	info->rti_info[RTAX_DST] = &sa_dst.sa;
907 	info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
908 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
909 	info->rti_info[RTAX_GENMASK] = 0;
910 	ifp = nh->nh_ifp;
911 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
912 		if (ifp) {
913 			info->rti_info[RTAX_IFP] =
914 			    ifp->if_addr->ifa_addr;
915 			error = rtm_get_jailed(info, ifp, nh,
916 			    &saun, curthread->td_ucred);
917 			if (error != 0)
918 				return (error);
919 			if (ifp->if_flags & IFF_POINTOPOINT)
920 				info->rti_info[RTAX_BRD] =
921 				    nh->nh_ifa->ifa_dstaddr;
922 			rtm->rtm_index = ifp->if_index;
923 		} else {
924 			info->rti_info[RTAX_IFP] = NULL;
925 			info->rti_info[RTAX_IFA] = NULL;
926 		}
927 	} else if (ifp != NULL)
928 		rtm->rtm_index = ifp->if_index;
929 
930 	/* Check if we need to realloc storage */
931 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
932 	if (len > alloc_len) {
933 		struct rt_msghdr *tmp_rtm;
934 
935 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
936 		if (tmp_rtm == NULL)
937 			return (ENOBUFS);
938 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
939 		orig_rtm = rtm;
940 		rtm = tmp_rtm;
941 		alloc_len = len;
942 
943 		/*
944 		 * Delay freeing original rtm as info contains
945 		 * data referencing it.
946 		 */
947 	}
948 
949 	w.w_tmem = (caddr_t)rtm;
950 	w.w_tmemsize = alloc_len;
951 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
952 
953 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
954 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
955 		rtm->rtm_flags = RTF_GATEWAY |
956 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
957 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
958 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
959 	rtm->rtm_addrs = info->rti_addrs;
960 
961 	if (orig_rtm != NULL)
962 		free(orig_rtm, M_TEMP);
963 	*prtm = rtm;
964 
965 	return (0);
966 }
967 
968 #ifdef ROUTE_MPATH
969 static void
970 save_del_notification(struct rib_cmd_info *rc, void *_cbdata)
971 {
972 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
973 
974 	if (rc->rc_cmd == RTM_DELETE)
975 		*rc_new = *rc;
976 }
977 
978 static void
979 save_add_notification(struct rib_cmd_info *rc, void *_cbdata)
980 {
981 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
982 
983 	if (rc->rc_cmd == RTM_ADD)
984 		*rc_new = *rc;
985 }
986 #endif
987 
988 /*ARGSUSED*/
989 static int
990 route_output(struct mbuf *m, struct socket *so, ...)
991 {
992 	struct rt_msghdr *rtm = NULL;
993 	struct rtentry *rt = NULL;
994 	struct rt_addrinfo info;
995 	struct epoch_tracker et;
996 #ifdef INET6
997 	struct sockaddr_storage ss;
998 	struct sockaddr_in6 *sin6;
999 	int i, rti_need_deembed = 0;
1000 #endif
1001 	int alloc_len = 0, len, error = 0, fibnum;
1002 	sa_family_t saf = AF_UNSPEC;
1003 	struct rib_cmd_info rc;
1004 	struct nhop_object *nh;
1005 
1006 	fibnum = so->so_fibnum;
1007 #define senderr(e) { error = e; goto flush;}
1008 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1009 		       (m = m_pullup(m, sizeof(long))) == NULL))
1010 		return (ENOBUFS);
1011 	if ((m->m_flags & M_PKTHDR) == 0)
1012 		panic("route_output");
1013 	NET_EPOCH_ENTER(et);
1014 	len = m->m_pkthdr.len;
1015 	if (len < sizeof(*rtm) ||
1016 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1017 		senderr(EINVAL);
1018 
1019 	/*
1020 	 * Most of current messages are in range 200-240 bytes,
1021 	 * minimize possible re-allocation on reply using larger size
1022 	 * buffer aligned on 1k boundaty.
1023 	 */
1024 	alloc_len = roundup2(len, 1024);
1025 	if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
1026 		senderr(ENOBUFS);
1027 
1028 	m_copydata(m, 0, len, (caddr_t)rtm);
1029 	bzero(&info, sizeof(info));
1030 	nh = NULL;
1031 
1032 	if (rtm->rtm_version != RTM_VERSION) {
1033 		/* Do not touch message since format is unknown */
1034 		free(rtm, M_TEMP);
1035 		rtm = NULL;
1036 		senderr(EPROTONOSUPPORT);
1037 	}
1038 
1039 	/*
1040 	 * Starting from here, it is possible
1041 	 * to alter original message and insert
1042 	 * caller PID and error value.
1043 	 */
1044 
1045 	if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) {
1046 		senderr(error);
1047 	}
1048 
1049 	saf = info.rti_info[RTAX_DST]->sa_family;
1050 
1051 	/* support for new ARP code */
1052 	if (rtm->rtm_flags & RTF_LLDATA) {
1053 		error = lla_rt_output(rtm, &info);
1054 #ifdef INET6
1055 		if (error == 0)
1056 			rti_need_deembed = 1;
1057 #endif
1058 		goto flush;
1059 	}
1060 
1061 	union sockaddr_union gw_saun;
1062 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1063 	if (blackhole_flags != 0) {
1064 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1065 			error = fill_blackholeinfo(&info, &gw_saun);
1066 		else
1067 			error = EINVAL;
1068 		if (error != 0)
1069 			senderr(error);
1070 		/* TODO: rebuild rtm from scratch */
1071 	}
1072 
1073 	switch (rtm->rtm_type) {
1074 	case RTM_ADD:
1075 	case RTM_CHANGE:
1076 		if (rtm->rtm_type == RTM_ADD) {
1077 			if (info.rti_info[RTAX_GATEWAY] == NULL)
1078 				senderr(EINVAL);
1079 		}
1080 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1081 		if (error == 0) {
1082 #ifdef INET6
1083 			rti_need_deembed = 1;
1084 #endif
1085 #ifdef ROUTE_MPATH
1086 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1087 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1088 				struct rib_cmd_info rc_simple = {};
1089 				rib_decompose_notification(&rc,
1090 				    save_add_notification, (void *)&rc_simple);
1091 				rc = rc_simple;
1092 			}
1093 #endif
1094 			nh = rc.rc_nh_new;
1095 			rtm->rtm_index = nh->nh_ifp->if_index;
1096 			rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1097 		}
1098 		break;
1099 
1100 	case RTM_DELETE:
1101 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1102 		if (error == 0) {
1103 #ifdef ROUTE_MPATH
1104 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1105 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1106 				struct rib_cmd_info rc_simple = {};
1107 				rib_decompose_notification(&rc,
1108 				    save_del_notification, (void *)&rc_simple);
1109 				rc = rc_simple;
1110 			}
1111 #endif
1112 			nh = rc.rc_nh_old;
1113 			goto report;
1114 		}
1115 #ifdef INET6
1116 		/* rt_msg2() will not be used when RTM_DELETE fails. */
1117 		rti_need_deembed = 1;
1118 #endif
1119 		break;
1120 
1121 	case RTM_GET:
1122 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1123 		if (error != 0)
1124 			senderr(error);
1125 		nh = rc.rc_nh_new;
1126 
1127 report:
1128 		if (!can_export_rte(curthread->td_ucred,
1129 		    info.rti_info[RTAX_NETMASK] == NULL,
1130 		    info.rti_info[RTAX_DST])) {
1131 			senderr(ESRCH);
1132 		}
1133 
1134 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1135 		/*
1136 		 * Note that some sockaddr pointers may have changed to
1137 		 * point to memory outsize @rtm. Some may be pointing
1138 		 * to the on-stack variables.
1139 		 * Given that, any pointer in @info CANNOT BE USED.
1140 		 */
1141 
1142 		/*
1143 		 * scopeid deembedding has been performed while
1144 		 * writing updated rtm in rtsock_msg_buffer().
1145 		 * With that in mind, skip deembedding procedure below.
1146 		 */
1147 #ifdef INET6
1148 		rti_need_deembed = 0;
1149 #endif
1150 		if (error != 0)
1151 			senderr(error);
1152 		break;
1153 
1154 	default:
1155 		senderr(EOPNOTSUPP);
1156 	}
1157 
1158 flush:
1159 	NET_EPOCH_EXIT(et);
1160 	rt = NULL;
1161 
1162 #ifdef INET6
1163 	if (rtm != NULL) {
1164 		if (rti_need_deembed) {
1165 			/* sin6_scope_id is recovered before sending rtm. */
1166 			sin6 = (struct sockaddr_in6 *)&ss;
1167 			for (i = 0; i < RTAX_MAX; i++) {
1168 				if (info.rti_info[i] == NULL)
1169 					continue;
1170 				if (info.rti_info[i]->sa_family != AF_INET6)
1171 					continue;
1172 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1173 				if (sa6_recoverscope(sin6) == 0)
1174 					bcopy(sin6, info.rti_info[i],
1175 						    sizeof(*sin6));
1176 			}
1177 		}
1178 	}
1179 #endif
1180 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1181 
1182 	return (error);
1183 }
1184 
1185 /*
1186  * Sends the prepared reply message in @rtm to all rtsock clients.
1187  * Frees @m and @rtm.
1188  *
1189  */
1190 static void
1191 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1192     sa_family_t saf, u_int fibnum, int rtm_errno)
1193 {
1194 	struct rawcb *rp = NULL;
1195 
1196 	/*
1197 	 * Check to see if we don't want our own messages.
1198 	 */
1199 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1200 		if (V_route_cb.any_count <= 1) {
1201 			if (rtm != NULL)
1202 				free(rtm, M_TEMP);
1203 			m_freem(m);
1204 			return;
1205 		}
1206 		/* There is another listener, so construct message */
1207 		rp = sotorawcb(so);
1208 	}
1209 
1210 	if (rtm != NULL) {
1211 		if (rtm_errno!= 0)
1212 			rtm->rtm_errno = rtm_errno;
1213 		else
1214 			rtm->rtm_flags |= RTF_DONE;
1215 
1216 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1217 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1218 			m_freem(m);
1219 			m = NULL;
1220 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1221 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1222 
1223 		free(rtm, M_TEMP);
1224 	}
1225 	if (m != NULL) {
1226 		M_SETFIB(m, fibnum);
1227 		m->m_flags |= RTS_FILTER_FIB;
1228 		if (rp) {
1229 			/*
1230 			 * XXX insure we don't get a copy by
1231 			 * invalidating our protocol
1232 			 */
1233 			unsigned short family = rp->rcb_proto.sp_family;
1234 			rp->rcb_proto.sp_family = 0;
1235 			rt_dispatch(m, saf);
1236 			rp->rcb_proto.sp_family = family;
1237 		} else
1238 			rt_dispatch(m, saf);
1239 	}
1240 }
1241 
1242 static void
1243 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1244     struct rt_metrics *out)
1245 {
1246 
1247 	bzero(out, sizeof(*out));
1248 	out->rmx_mtu = nh->nh_mtu;
1249 	out->rmx_weight = rt->rt_weight;
1250 	out->rmx_nhidx = nhop_get_idx(nh);
1251 	/* Kernel -> userland timebase conversion. */
1252 	out->rmx_expire = rt->rt_expire ?
1253 	    rt->rt_expire - time_uptime + time_second : 0;
1254 }
1255 
1256 /*
1257  * Extract the addresses of the passed sockaddrs.
1258  * Do a little sanity checking so as to avoid bad memory references.
1259  * This data is derived straight from userland.
1260  */
1261 static int
1262 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1263 {
1264 	struct sockaddr *sa;
1265 	int i;
1266 
1267 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1268 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1269 			continue;
1270 		sa = (struct sockaddr *)cp;
1271 		/*
1272 		 * It won't fit.
1273 		 */
1274 		if (cp + sa->sa_len > cplim)
1275 			return (EINVAL);
1276 		/*
1277 		 * there are no more.. quit now
1278 		 * If there are more bits, they are in error.
1279 		 * I've seen this. route(1) can evidently generate these.
1280 		 * This causes kernel to core dump.
1281 		 * for compatibility, If we see this, point to a safe address.
1282 		 */
1283 		if (sa->sa_len == 0) {
1284 			rtinfo->rti_info[i] = &sa_zero;
1285 			return (0); /* should be EINVAL but for compat */
1286 		}
1287 		/* accept it */
1288 #ifdef INET6
1289 		if (sa->sa_family == AF_INET6)
1290 			sa6_embedscope((struct sockaddr_in6 *)sa,
1291 			    V_ip6_use_defzone);
1292 #endif
1293 		rtinfo->rti_info[i] = sa;
1294 		cp += SA_SIZE(sa);
1295 	}
1296 	return (0);
1297 }
1298 
1299 #ifdef INET
1300 static inline void
1301 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1302 {
1303 
1304 	const struct sockaddr_in nsin = {
1305 		.sin_family = AF_INET,
1306 		.sin_len = sizeof(struct sockaddr_in),
1307 		.sin_addr = addr,
1308 	};
1309 	*sin = nsin;
1310 }
1311 #endif
1312 
1313 #ifdef INET6
1314 static inline void
1315 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1316     uint32_t scopeid)
1317 {
1318 
1319 	const struct sockaddr_in6 nsin6 = {
1320 		.sin6_family = AF_INET6,
1321 		.sin6_len = sizeof(struct sockaddr_in6),
1322 		.sin6_addr = *addr6,
1323 		.sin6_scope_id = scopeid,
1324 	};
1325 	*sin6 = nsin6;
1326 }
1327 #endif
1328 
1329 /*
1330  * Checks if gateway is suitable for lltable operations.
1331  * Lltable code requires AF_LINK gateway with ifindex
1332  *  and mac address specified.
1333  * Returns 0 on success.
1334  */
1335 static int
1336 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1337 {
1338 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1339 
1340 	if (sdl->sdl_family != AF_LINK)
1341 		return (EINVAL);
1342 
1343 	if (sdl->sdl_index == 0)
1344 		return (EINVAL);
1345 
1346 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len)
1347 		return (EINVAL);
1348 
1349 	return (0);
1350 }
1351 
1352 static int
1353 cleanup_xaddrs_gateway(struct rt_addrinfo *info)
1354 {
1355 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1356 
1357 	if (info->rti_flags & RTF_LLDATA)
1358 		return (cleanup_xaddrs_lladdr(info));
1359 
1360 	switch (gw->sa_family) {
1361 #ifdef INET
1362 	case AF_INET:
1363 		{
1364 			struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1365 			if (gw_sin->sin_len < sizeof(struct sockaddr_in)) {
1366 				RTS_PID_PRINTF("gateway sin_len too small: %d", gw->sa_len);
1367 				return (EINVAL);
1368 			}
1369 			fill_sockaddr_inet(gw_sin, gw_sin->sin_addr);
1370 		}
1371 		break;
1372 #endif
1373 #ifdef INET6
1374 	case AF_INET6:
1375 		{
1376 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1377 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1378 				RTS_PID_PRINTF("gateway sin6_len too small: %d", gw->sa_len);
1379 				return (EINVAL);
1380 			}
1381 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1382 			break;
1383 		}
1384 #endif
1385 	case AF_LINK:
1386 		{
1387 			struct sockaddr_dl *gw_sdl;
1388 
1389 			size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1390 			gw_sdl = (struct sockaddr_dl *)gw;
1391 			if (gw_sdl->sdl_len < sdl_min_len) {
1392 				RTS_PID_PRINTF("gateway sdl_len too small: %d", gw_sdl->sdl_len);
1393 				return (EINVAL);
1394 			}
1395 
1396 			const struct sockaddr_dl_short sdl = {
1397 				.sdl_family = AF_LINK,
1398 				.sdl_len = sdl_min_len,
1399 				.sdl_index = gw_sdl->sdl_index,
1400 			};
1401 			memcpy(gw_sdl, &sdl, sdl_min_len);
1402 			break;
1403 		}
1404 	}
1405 
1406 	return (0);
1407 }
1408 
1409 static void
1410 remove_netmask(struct rt_addrinfo *info)
1411 {
1412 	info->rti_info[RTAX_NETMASK] = NULL;
1413 	info->rti_flags |= RTF_HOST;
1414 	info->rti_addrs &= ~RTA_NETMASK;
1415 }
1416 
1417 #ifdef INET
1418 static int
1419 cleanup_xaddrs_inet(struct rt_addrinfo *info)
1420 {
1421 	struct sockaddr_in *dst_sa, *mask_sa;
1422 
1423 	/* Check & fixup dst/netmask combination first */
1424 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1425 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1426 
1427 	struct in_addr mask = {
1428 		.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST,
1429 	};
1430 	struct in_addr dst = {
1431 		.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr))
1432 	};
1433 
1434 	if (dst_sa->sin_len < sizeof(struct sockaddr_in)) {
1435 		printf("dst sin_len too small\n");
1436 		return (EINVAL);
1437 	}
1438 	if (mask_sa && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1439 		RTS_PID_PRINTF("prefix mask sin_len too small: %d", mask_sa->sin_len);
1440 		return (EINVAL);
1441 	}
1442 	fill_sockaddr_inet(dst_sa, dst);
1443 
1444 	if (mask.s_addr != INADDR_BROADCAST)
1445 		fill_sockaddr_inet(mask_sa, mask);
1446 	else
1447 		remove_netmask(info);
1448 
1449 	/* Check gateway */
1450 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1451 		return (cleanup_xaddrs_gateway(info));
1452 
1453 	return (0);
1454 }
1455 #endif
1456 
1457 #ifdef INET6
1458 static int
1459 cleanup_xaddrs_inet6(struct rt_addrinfo *info)
1460 {
1461 	struct sockaddr_in6 *dst_sa, *mask_sa;
1462 	struct in6_addr mask;
1463 
1464 	/* Check & fixup dst/netmask combination first */
1465 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1466 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1467 
1468 	mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1469 	IN6_MASK_ADDR(&dst_sa->sin6_addr, &mask);
1470 
1471 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1472 		RTS_PID_PRINTF("prefix dst sin6_len too small: %d", dst_sa->sin6_len);
1473 		return (EINVAL);
1474 	}
1475 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1476 		RTS_PID_PRINTF("rtsock: prefix mask sin6_len too small: %d", mask_sa->sin6_len);
1477 		return (EINVAL);
1478 	}
1479 	fill_sockaddr_inet6(dst_sa, &dst_sa->sin6_addr, 0);
1480 
1481 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128))
1482 		fill_sockaddr_inet6(mask_sa, &mask, 0);
1483 	else
1484 		remove_netmask(info);
1485 
1486 	/* Check gateway */
1487 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1488 		return (cleanup_xaddrs_gateway(info));
1489 
1490 	return (0);
1491 }
1492 #endif
1493 
1494 static int
1495 cleanup_xaddrs(struct rt_addrinfo *info)
1496 {
1497 	int error = EAFNOSUPPORT;
1498 
1499 	if (info->rti_info[RTAX_DST] == NULL)
1500 		return (EINVAL);
1501 
1502 	if (info->rti_flags & RTF_LLDATA) {
1503 		/*
1504 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1505 		 * prefix along with the actual address in RTA_DST.
1506 		 * Remove netmask to avoid unnecessary address masking.
1507 		 */
1508 		remove_netmask(info);
1509 	}
1510 
1511 	switch (info->rti_info[RTAX_DST]->sa_family) {
1512 #ifdef INET
1513 	case AF_INET:
1514 		error = cleanup_xaddrs_inet(info);
1515 		break;
1516 #endif
1517 #ifdef INET6
1518 	case AF_INET6:
1519 		error = cleanup_xaddrs_inet6(info);
1520 		break;
1521 #endif
1522 	}
1523 
1524 	return (error);
1525 }
1526 
1527 /*
1528  * Fill in @dmask with valid netmask leaving original @smask
1529  * intact. Mostly used with radix netmasks.
1530  */
1531 struct sockaddr *
1532 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1533     struct sockaddr_storage *dmask)
1534 {
1535 	if (dst == NULL || smask == NULL)
1536 		return (NULL);
1537 
1538 	memset(dmask, 0, dst->sa_len);
1539 	memcpy(dmask, smask, smask->sa_len);
1540 	dmask->ss_len = dst->sa_len;
1541 	dmask->ss_family = dst->sa_family;
1542 
1543 	return ((struct sockaddr *)dmask);
1544 }
1545 
1546 /*
1547  * Writes information related to @rtinfo object to newly-allocated mbuf.
1548  * Assumes MCLBYTES is enough to construct any message.
1549  * Used for OS notifications of vaious events (if/ifa announces,etc)
1550  *
1551  * Returns allocated mbuf or NULL on failure.
1552  */
1553 static struct mbuf *
1554 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1555 {
1556 	struct sockaddr_storage ss;
1557 	struct rt_msghdr *rtm;
1558 	struct mbuf *m;
1559 	int i;
1560 	struct sockaddr *sa;
1561 #ifdef INET6
1562 	struct sockaddr_in6 *sin6;
1563 #endif
1564 	int len, dlen;
1565 
1566 	switch (type) {
1567 	case RTM_DELADDR:
1568 	case RTM_NEWADDR:
1569 		len = sizeof(struct ifa_msghdr);
1570 		break;
1571 
1572 	case RTM_DELMADDR:
1573 	case RTM_NEWMADDR:
1574 		len = sizeof(struct ifma_msghdr);
1575 		break;
1576 
1577 	case RTM_IFINFO:
1578 		len = sizeof(struct if_msghdr);
1579 		break;
1580 
1581 	case RTM_IFANNOUNCE:
1582 	case RTM_IEEE80211:
1583 		len = sizeof(struct if_announcemsghdr);
1584 		break;
1585 
1586 	default:
1587 		len = sizeof(struct rt_msghdr);
1588 	}
1589 
1590 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1591 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1592 	if (len > MHLEN)
1593 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1594 	else
1595 		m = m_gethdr(M_NOWAIT, MT_DATA);
1596 	if (m == NULL)
1597 		return (m);
1598 
1599 	m->m_pkthdr.len = m->m_len = len;
1600 	rtm = mtod(m, struct rt_msghdr *);
1601 	bzero((caddr_t)rtm, len);
1602 	for (i = 0; i < RTAX_MAX; i++) {
1603 		if ((sa = rtinfo->rti_info[i]) == NULL)
1604 			continue;
1605 		rtinfo->rti_addrs |= (1 << i);
1606 
1607 		dlen = SA_SIZE(sa);
1608 		KASSERT(dlen <= sizeof(ss),
1609 		    ("%s: sockaddr size overflow", __func__));
1610 		bzero(&ss, sizeof(ss));
1611 		bcopy(sa, &ss, sa->sa_len);
1612 		sa = (struct sockaddr *)&ss;
1613 #ifdef INET6
1614 		if (sa->sa_family == AF_INET6) {
1615 			sin6 = (struct sockaddr_in6 *)sa;
1616 			(void)sa6_recoverscope(sin6);
1617 		}
1618 #endif
1619 		m_copyback(m, len, dlen, (caddr_t)sa);
1620 		len += dlen;
1621 	}
1622 	if (m->m_pkthdr.len != len) {
1623 		m_freem(m);
1624 		return (NULL);
1625 	}
1626 	rtm->rtm_msglen = len;
1627 	rtm->rtm_version = RTM_VERSION;
1628 	rtm->rtm_type = type;
1629 	return (m);
1630 }
1631 
1632 /*
1633  * Writes information related to @rtinfo object to preallocated buffer.
1634  * Stores needed size in @plen. If @w is NULL, calculates size without
1635  * writing.
1636  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1637  *
1638  * Returns 0 on success.
1639  *
1640  */
1641 static int
1642 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1643 {
1644 	struct sockaddr_storage ss;
1645 	int len, buflen = 0, dlen, i;
1646 	caddr_t cp = NULL;
1647 	struct rt_msghdr *rtm = NULL;
1648 #ifdef INET6
1649 	struct sockaddr_in6 *sin6;
1650 #endif
1651 #ifdef COMPAT_FREEBSD32
1652 	bool compat32 = false;
1653 #endif
1654 
1655 	switch (type) {
1656 	case RTM_DELADDR:
1657 	case RTM_NEWADDR:
1658 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1659 #ifdef COMPAT_FREEBSD32
1660 			if (w->w_req->flags & SCTL_MASK32) {
1661 				len = sizeof(struct ifa_msghdrl32);
1662 				compat32 = true;
1663 			} else
1664 #endif
1665 				len = sizeof(struct ifa_msghdrl);
1666 		} else
1667 			len = sizeof(struct ifa_msghdr);
1668 		break;
1669 
1670 	case RTM_IFINFO:
1671 #ifdef COMPAT_FREEBSD32
1672 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1673 			if (w->w_op == NET_RT_IFLISTL)
1674 				len = sizeof(struct if_msghdrl32);
1675 			else
1676 				len = sizeof(struct if_msghdr32);
1677 			compat32 = true;
1678 			break;
1679 		}
1680 #endif
1681 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1682 			len = sizeof(struct if_msghdrl);
1683 		else
1684 			len = sizeof(struct if_msghdr);
1685 		break;
1686 
1687 	case RTM_NEWMADDR:
1688 		len = sizeof(struct ifma_msghdr);
1689 		break;
1690 
1691 	default:
1692 		len = sizeof(struct rt_msghdr);
1693 	}
1694 
1695 	if (w != NULL) {
1696 		rtm = (struct rt_msghdr *)w->w_tmem;
1697 		buflen = w->w_tmemsize - len;
1698 		cp = (caddr_t)w->w_tmem + len;
1699 	}
1700 
1701 	rtinfo->rti_addrs = 0;
1702 	for (i = 0; i < RTAX_MAX; i++) {
1703 		struct sockaddr *sa;
1704 
1705 		if ((sa = rtinfo->rti_info[i]) == NULL)
1706 			continue;
1707 		rtinfo->rti_addrs |= (1 << i);
1708 #ifdef COMPAT_FREEBSD32
1709 		if (compat32)
1710 			dlen = SA_SIZE32(sa);
1711 		else
1712 #endif
1713 			dlen = SA_SIZE(sa);
1714 		if (cp != NULL && buflen >= dlen) {
1715 			KASSERT(dlen <= sizeof(ss),
1716 			    ("%s: sockaddr size overflow", __func__));
1717 			bzero(&ss, sizeof(ss));
1718 			bcopy(sa, &ss, sa->sa_len);
1719 			sa = (struct sockaddr *)&ss;
1720 #ifdef INET6
1721 			if (sa->sa_family == AF_INET6) {
1722 				sin6 = (struct sockaddr_in6 *)sa;
1723 				(void)sa6_recoverscope(sin6);
1724 			}
1725 #endif
1726 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1727 			cp += dlen;
1728 			buflen -= dlen;
1729 		} else if (cp != NULL) {
1730 			/*
1731 			 * Buffer too small. Count needed size
1732 			 * and return with error.
1733 			 */
1734 			cp = NULL;
1735 		}
1736 
1737 		len += dlen;
1738 	}
1739 
1740 	if (cp != NULL) {
1741 		dlen = ALIGN(len) - len;
1742 		if (buflen < dlen)
1743 			cp = NULL;
1744 		else {
1745 			bzero(cp, dlen);
1746 			cp += dlen;
1747 			buflen -= dlen;
1748 		}
1749 	}
1750 	len = ALIGN(len);
1751 
1752 	if (cp != NULL) {
1753 		/* fill header iff buffer is large enough */
1754 		rtm->rtm_version = RTM_VERSION;
1755 		rtm->rtm_type = type;
1756 		rtm->rtm_msglen = len;
1757 	}
1758 
1759 	*plen = len;
1760 
1761 	if (w != NULL && cp == NULL)
1762 		return (ENOBUFS);
1763 
1764 	return (0);
1765 }
1766 
1767 /*
1768  * This routine is called to generate a message from the routing
1769  * socket indicating that a redirect has occurred, a routing lookup
1770  * has failed, or that a protocol has detected timeouts to a particular
1771  * destination.
1772  */
1773 void
1774 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1775     int fibnum)
1776 {
1777 	struct rt_msghdr *rtm;
1778 	struct mbuf *m;
1779 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1780 
1781 	if (V_route_cb.any_count == 0)
1782 		return;
1783 	m = rtsock_msg_mbuf(type, rtinfo);
1784 	if (m == NULL)
1785 		return;
1786 
1787 	if (fibnum != RT_ALL_FIBS) {
1788 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1789 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1790 		M_SETFIB(m, fibnum);
1791 		m->m_flags |= RTS_FILTER_FIB;
1792 	}
1793 
1794 	rtm = mtod(m, struct rt_msghdr *);
1795 	rtm->rtm_flags = RTF_DONE | flags;
1796 	rtm->rtm_errno = error;
1797 	rtm->rtm_addrs = rtinfo->rti_addrs;
1798 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1799 }
1800 
1801 void
1802 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1803 {
1804 
1805 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1806 }
1807 
1808 /*
1809  * This routine is called to generate a message from the routing
1810  * socket indicating that the status of a network interface has changed.
1811  */
1812 void
1813 rt_ifmsg(struct ifnet *ifp)
1814 {
1815 	struct if_msghdr *ifm;
1816 	struct mbuf *m;
1817 	struct rt_addrinfo info;
1818 
1819 	if (V_route_cb.any_count == 0)
1820 		return;
1821 	bzero((caddr_t)&info, sizeof(info));
1822 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1823 	if (m == NULL)
1824 		return;
1825 	ifm = mtod(m, struct if_msghdr *);
1826 	ifm->ifm_index = ifp->if_index;
1827 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1828 	if_data_copy(ifp, &ifm->ifm_data);
1829 	ifm->ifm_addrs = 0;
1830 	rt_dispatch(m, AF_UNSPEC);
1831 }
1832 
1833 /*
1834  * Announce interface address arrival/withdraw.
1835  * Please do not call directly, use rt_addrmsg().
1836  * Assume input data to be valid.
1837  * Returns 0 on success.
1838  */
1839 int
1840 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1841 {
1842 	struct rt_addrinfo info;
1843 	struct sockaddr *sa;
1844 	int ncmd;
1845 	struct mbuf *m;
1846 	struct ifa_msghdr *ifam;
1847 	struct ifnet *ifp = ifa->ifa_ifp;
1848 	struct sockaddr_storage ss;
1849 
1850 	if (V_route_cb.any_count == 0)
1851 		return (0);
1852 
1853 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1854 
1855 	bzero((caddr_t)&info, sizeof(info));
1856 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1857 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1858 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1859 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1860 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1861 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1862 		return (ENOBUFS);
1863 	ifam = mtod(m, struct ifa_msghdr *);
1864 	ifam->ifam_index = ifp->if_index;
1865 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1866 	ifam->ifam_flags = ifa->ifa_flags;
1867 	ifam->ifam_addrs = info.rti_addrs;
1868 
1869 	if (fibnum != RT_ALL_FIBS) {
1870 		M_SETFIB(m, fibnum);
1871 		m->m_flags |= RTS_FILTER_FIB;
1872 	}
1873 
1874 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1875 
1876 	return (0);
1877 }
1878 
1879 /*
1880  * Announce route addition/removal to rtsock based on @rt data.
1881  * Callers are advives to use rt_routemsg() instead of using this
1882  *  function directly.
1883  * Assume @rt data is consistent.
1884  *
1885  * Returns 0 on success.
1886  */
1887 int
1888 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
1889     int fibnum)
1890 {
1891 	union sockaddr_union dst, mask;
1892 	struct rt_addrinfo info;
1893 
1894 	if (V_route_cb.any_count == 0)
1895 		return (0);
1896 
1897 	int family = rt_get_family(rt);
1898 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
1899 	export_rtaddrs(rt, &dst.sa, &mask.sa);
1900 
1901 	bzero((caddr_t)&info, sizeof(info));
1902 	info.rti_info[RTAX_DST] = &dst.sa;
1903 	info.rti_info[RTAX_NETMASK] = &mask.sa;
1904 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1905 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
1906 	info.rti_ifp = nh->nh_ifp;
1907 
1908 	return (rtsock_routemsg_info(cmd, &info, fibnum));
1909 }
1910 
1911 int
1912 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1913 {
1914 	struct rt_msghdr *rtm;
1915 	struct sockaddr *sa;
1916 	struct mbuf *m;
1917 
1918 	if (V_route_cb.any_count == 0)
1919 		return (0);
1920 
1921 	if (info->rti_flags & RTF_HOST)
1922 		info->rti_info[RTAX_NETMASK] = NULL;
1923 
1924 	m = rtsock_msg_mbuf(cmd, info);
1925 	if (m == NULL)
1926 		return (ENOBUFS);
1927 
1928 	if (fibnum != RT_ALL_FIBS) {
1929 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1930 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1931 		M_SETFIB(m, fibnum);
1932 		m->m_flags |= RTS_FILTER_FIB;
1933 	}
1934 
1935 	rtm = mtod(m, struct rt_msghdr *);
1936 	rtm->rtm_addrs = info->rti_addrs;
1937 	if (info->rti_ifp != NULL)
1938 		rtm->rtm_index = info->rti_ifp->if_index;
1939 	/* Add RTF_DONE to indicate command 'completion' required by API */
1940 	info->rti_flags |= RTF_DONE;
1941 	/* Reported routes has to be up */
1942 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1943 		info->rti_flags |= RTF_UP;
1944 	rtm->rtm_flags = info->rti_flags;
1945 
1946 	sa = info->rti_info[RTAX_DST];
1947 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1948 
1949 	return (0);
1950 }
1951 
1952 /*
1953  * This is the analogue to the rt_newaddrmsg which performs the same
1954  * function but for multicast group memberhips.  This is easier since
1955  * there is no route state to worry about.
1956  */
1957 void
1958 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1959 {
1960 	struct rt_addrinfo info;
1961 	struct mbuf *m = NULL;
1962 	struct ifnet *ifp = ifma->ifma_ifp;
1963 	struct ifma_msghdr *ifmam;
1964 
1965 	if (V_route_cb.any_count == 0)
1966 		return;
1967 
1968 	bzero((caddr_t)&info, sizeof(info));
1969 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1970 	if (ifp && ifp->if_addr)
1971 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1972 	else
1973 		info.rti_info[RTAX_IFP] = NULL;
1974 	/*
1975 	 * If a link-layer address is present, present it as a ``gateway''
1976 	 * (similarly to how ARP entries, e.g., are presented).
1977 	 */
1978 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1979 	m = rtsock_msg_mbuf(cmd, &info);
1980 	if (m == NULL)
1981 		return;
1982 	ifmam = mtod(m, struct ifma_msghdr *);
1983 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1984 	    __func__));
1985 	ifmam->ifmam_index = ifp->if_index;
1986 	ifmam->ifmam_addrs = info.rti_addrs;
1987 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1988 }
1989 
1990 static struct mbuf *
1991 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1992 	struct rt_addrinfo *info)
1993 {
1994 	struct if_announcemsghdr *ifan;
1995 	struct mbuf *m;
1996 
1997 	if (V_route_cb.any_count == 0)
1998 		return NULL;
1999 	bzero((caddr_t)info, sizeof(*info));
2000 	m = rtsock_msg_mbuf(type, info);
2001 	if (m != NULL) {
2002 		ifan = mtod(m, struct if_announcemsghdr *);
2003 		ifan->ifan_index = ifp->if_index;
2004 		strlcpy(ifan->ifan_name, ifp->if_xname,
2005 			sizeof(ifan->ifan_name));
2006 		ifan->ifan_what = what;
2007 	}
2008 	return m;
2009 }
2010 
2011 /*
2012  * This is called to generate routing socket messages indicating
2013  * IEEE80211 wireless events.
2014  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2015  */
2016 void
2017 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2018 {
2019 	struct mbuf *m;
2020 	struct rt_addrinfo info;
2021 
2022 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2023 	if (m != NULL) {
2024 		/*
2025 		 * Append the ieee80211 data.  Try to stick it in the
2026 		 * mbuf containing the ifannounce msg; otherwise allocate
2027 		 * a new mbuf and append.
2028 		 *
2029 		 * NB: we assume m is a single mbuf.
2030 		 */
2031 		if (data_len > M_TRAILINGSPACE(m)) {
2032 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2033 			if (n == NULL) {
2034 				m_freem(m);
2035 				return;
2036 			}
2037 			bcopy(data, mtod(n, void *), data_len);
2038 			n->m_len = data_len;
2039 			m->m_next = n;
2040 		} else if (data_len > 0) {
2041 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2042 			m->m_len += data_len;
2043 		}
2044 		if (m->m_flags & M_PKTHDR)
2045 			m->m_pkthdr.len += data_len;
2046 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2047 		rt_dispatch(m, AF_UNSPEC);
2048 	}
2049 }
2050 
2051 /*
2052  * This is called to generate routing socket messages indicating
2053  * network interface arrival and departure.
2054  */
2055 void
2056 rt_ifannouncemsg(struct ifnet *ifp, int what)
2057 {
2058 	struct mbuf *m;
2059 	struct rt_addrinfo info;
2060 
2061 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2062 	if (m != NULL)
2063 		rt_dispatch(m, AF_UNSPEC);
2064 }
2065 
2066 static void
2067 rt_dispatch(struct mbuf *m, sa_family_t saf)
2068 {
2069 	struct m_tag *tag;
2070 
2071 	/*
2072 	 * Preserve the family from the sockaddr, if any, in an m_tag for
2073 	 * use when injecting the mbuf into the routing socket buffer from
2074 	 * the netisr.
2075 	 */
2076 	if (saf != AF_UNSPEC) {
2077 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
2078 		    M_NOWAIT);
2079 		if (tag == NULL) {
2080 			m_freem(m);
2081 			return;
2082 		}
2083 		*(unsigned short *)(tag + 1) = saf;
2084 		m_tag_prepend(m, tag);
2085 	}
2086 #ifdef VIMAGE
2087 	if (V_loif)
2088 		m->m_pkthdr.rcvif = V_loif;
2089 	else {
2090 		m_freem(m);
2091 		return;
2092 	}
2093 #endif
2094 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2095 }
2096 
2097 /*
2098  * Checks if rte can be exported w.r.t jails/vnets.
2099  *
2100  * Returns true if it can, false otherwise.
2101  */
2102 static bool
2103 can_export_rte(struct ucred *td_ucred, bool rt_is_host,
2104     const struct sockaddr *rt_dst)
2105 {
2106 
2107 	if ((!rt_is_host) ? jailed_without_vnet(td_ucred)
2108 	    : prison_if(td_ucred, rt_dst) != 0)
2109 		return (false);
2110 	return (true);
2111 }
2112 
2113 
2114 /*
2115  * This is used in dumping the kernel table via sysctl().
2116  */
2117 static int
2118 sysctl_dumpentry(struct rtentry *rt, void *vw)
2119 {
2120 	struct walkarg *w = vw;
2121 	struct nhop_object *nh;
2122 	int error = 0;
2123 
2124 	NET_EPOCH_ASSERT();
2125 
2126 	export_rtaddrs(rt, w->dst, w->mask);
2127 	if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst))
2128 		return (0);
2129 	nh = rt_get_raw_nhop(rt);
2130 #ifdef ROUTE_MPATH
2131 	if (NH_IS_NHGRP(nh)) {
2132 		struct weightened_nhop *wn;
2133 		uint32_t num_nhops;
2134 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2135 		for (int i = 0; i < num_nhops; i++) {
2136 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2137 			if (error != 0)
2138 				return (error);
2139 		}
2140 	} else
2141 #endif
2142 		error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2143 
2144 	return (0);
2145 }
2146 
2147 
2148 static int
2149 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2150     struct walkarg *w)
2151 {
2152 	struct rt_addrinfo info;
2153 	int error = 0, size;
2154 	uint32_t rtflags;
2155 
2156 	rtflags = nhop_get_rtflags(nh);
2157 
2158 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2159 		return (0);
2160 
2161 	bzero((caddr_t)&info, sizeof(info));
2162 	info.rti_info[RTAX_DST] = w->dst;
2163 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2164 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2165 	info.rti_info[RTAX_GENMASK] = 0;
2166 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2167 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2168 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2169 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2170 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2171 	}
2172 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2173 		return (error);
2174 	if (w->w_req && w->w_tmem) {
2175 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2176 
2177 		bzero(&rtm->rtm_index,
2178 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2179 
2180 		/*
2181 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2182 		 * and RTF_UP (if entry is linked, which is always true here).
2183 		 * Given that, use nhop rtflags & add RTF_UP.
2184 		 */
2185 		rtm->rtm_flags = rtflags | RTF_UP;
2186 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2187 			rtm->rtm_flags = RTF_GATEWAY |
2188 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2189 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2190 		rtm->rtm_rmx.rmx_weight = weight;
2191 		rtm->rtm_index = nh->nh_ifp->if_index;
2192 		rtm->rtm_addrs = info.rti_addrs;
2193 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2194 		return (error);
2195 	}
2196 	return (error);
2197 }
2198 
2199 static int
2200 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2201     struct rt_addrinfo *info, struct walkarg *w, int len)
2202 {
2203 	struct if_msghdrl *ifm;
2204 	struct if_data *ifd;
2205 
2206 	ifm = (struct if_msghdrl *)w->w_tmem;
2207 
2208 #ifdef COMPAT_FREEBSD32
2209 	if (w->w_req->flags & SCTL_MASK32) {
2210 		struct if_msghdrl32 *ifm32;
2211 
2212 		ifm32 = (struct if_msghdrl32 *)ifm;
2213 		ifm32->ifm_addrs = info->rti_addrs;
2214 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2215 		ifm32->ifm_index = ifp->if_index;
2216 		ifm32->_ifm_spare1 = 0;
2217 		ifm32->ifm_len = sizeof(*ifm32);
2218 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2219 		ifm32->_ifm_spare2 = 0;
2220 		ifd = &ifm32->ifm_data;
2221 	} else
2222 #endif
2223 	{
2224 		ifm->ifm_addrs = info->rti_addrs;
2225 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2226 		ifm->ifm_index = ifp->if_index;
2227 		ifm->_ifm_spare1 = 0;
2228 		ifm->ifm_len = sizeof(*ifm);
2229 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2230 		ifm->_ifm_spare2 = 0;
2231 		ifd = &ifm->ifm_data;
2232 	}
2233 
2234 	memcpy(ifd, src_ifd, sizeof(*ifd));
2235 
2236 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2237 }
2238 
2239 static int
2240 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2241     struct rt_addrinfo *info, struct walkarg *w, int len)
2242 {
2243 	struct if_msghdr *ifm;
2244 	struct if_data *ifd;
2245 
2246 	ifm = (struct if_msghdr *)w->w_tmem;
2247 
2248 #ifdef COMPAT_FREEBSD32
2249 	if (w->w_req->flags & SCTL_MASK32) {
2250 		struct if_msghdr32 *ifm32;
2251 
2252 		ifm32 = (struct if_msghdr32 *)ifm;
2253 		ifm32->ifm_addrs = info->rti_addrs;
2254 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2255 		ifm32->ifm_index = ifp->if_index;
2256 		ifm32->_ifm_spare1 = 0;
2257 		ifd = &ifm32->ifm_data;
2258 	} else
2259 #endif
2260 	{
2261 		ifm->ifm_addrs = info->rti_addrs;
2262 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2263 		ifm->ifm_index = ifp->if_index;
2264 		ifm->_ifm_spare1 = 0;
2265 		ifd = &ifm->ifm_data;
2266 	}
2267 
2268 	memcpy(ifd, src_ifd, sizeof(*ifd));
2269 
2270 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2271 }
2272 
2273 static int
2274 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2275     struct walkarg *w, int len)
2276 {
2277 	struct ifa_msghdrl *ifam;
2278 	struct if_data *ifd;
2279 
2280 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2281 
2282 #ifdef COMPAT_FREEBSD32
2283 	if (w->w_req->flags & SCTL_MASK32) {
2284 		struct ifa_msghdrl32 *ifam32;
2285 
2286 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2287 		ifam32->ifam_addrs = info->rti_addrs;
2288 		ifam32->ifam_flags = ifa->ifa_flags;
2289 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2290 		ifam32->_ifam_spare1 = 0;
2291 		ifam32->ifam_len = sizeof(*ifam32);
2292 		ifam32->ifam_data_off =
2293 		    offsetof(struct ifa_msghdrl32, ifam_data);
2294 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2295 		ifd = &ifam32->ifam_data;
2296 	} else
2297 #endif
2298 	{
2299 		ifam->ifam_addrs = info->rti_addrs;
2300 		ifam->ifam_flags = ifa->ifa_flags;
2301 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2302 		ifam->_ifam_spare1 = 0;
2303 		ifam->ifam_len = sizeof(*ifam);
2304 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2305 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2306 		ifd = &ifam->ifam_data;
2307 	}
2308 
2309 	bzero(ifd, sizeof(*ifd));
2310 	ifd->ifi_datalen = sizeof(struct if_data);
2311 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2312 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2313 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2314 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2315 
2316 	/* Fixup if_data carp(4) vhid. */
2317 	if (carp_get_vhid_p != NULL)
2318 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2319 
2320 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2321 }
2322 
2323 static int
2324 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2325     struct walkarg *w, int len)
2326 {
2327 	struct ifa_msghdr *ifam;
2328 
2329 	ifam = (struct ifa_msghdr *)w->w_tmem;
2330 	ifam->ifam_addrs = info->rti_addrs;
2331 	ifam->ifam_flags = ifa->ifa_flags;
2332 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2333 	ifam->_ifam_spare1 = 0;
2334 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2335 
2336 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2337 }
2338 
2339 static int
2340 sysctl_iflist(int af, struct walkarg *w)
2341 {
2342 	struct ifnet *ifp;
2343 	struct ifaddr *ifa;
2344 	struct if_data ifd;
2345 	struct rt_addrinfo info;
2346 	int len, error = 0;
2347 	struct sockaddr_storage ss;
2348 
2349 	bzero((caddr_t)&info, sizeof(info));
2350 	bzero(&ifd, sizeof(ifd));
2351 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2352 		if (w->w_arg && w->w_arg != ifp->if_index)
2353 			continue;
2354 		if_data_copy(ifp, &ifd);
2355 		ifa = ifp->if_addr;
2356 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2357 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2358 		if (error != 0)
2359 			goto done;
2360 		info.rti_info[RTAX_IFP] = NULL;
2361 		if (w->w_req && w->w_tmem) {
2362 			if (w->w_op == NET_RT_IFLISTL)
2363 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2364 				    len);
2365 			else
2366 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2367 				    len);
2368 			if (error)
2369 				goto done;
2370 		}
2371 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2372 			if (af && af != ifa->ifa_addr->sa_family)
2373 				continue;
2374 			if (prison_if(w->w_req->td->td_ucred,
2375 			    ifa->ifa_addr) != 0)
2376 				continue;
2377 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2378 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2379 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2380 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2381 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2382 			if (error != 0)
2383 				goto done;
2384 			if (w->w_req && w->w_tmem) {
2385 				if (w->w_op == NET_RT_IFLISTL)
2386 					error = sysctl_iflist_ifaml(ifa, &info,
2387 					    w, len);
2388 				else
2389 					error = sysctl_iflist_ifam(ifa, &info,
2390 					    w, len);
2391 				if (error)
2392 					goto done;
2393 			}
2394 		}
2395 		info.rti_info[RTAX_IFA] = NULL;
2396 		info.rti_info[RTAX_NETMASK] = NULL;
2397 		info.rti_info[RTAX_BRD] = NULL;
2398 	}
2399 done:
2400 	return (error);
2401 }
2402 
2403 static int
2404 sysctl_ifmalist(int af, struct walkarg *w)
2405 {
2406 	struct rt_addrinfo info;
2407 	struct ifaddr *ifa;
2408 	struct ifmultiaddr *ifma;
2409 	struct ifnet *ifp;
2410 	int error, len;
2411 
2412 	NET_EPOCH_ASSERT();
2413 
2414 	error = 0;
2415 	bzero((caddr_t)&info, sizeof(info));
2416 
2417 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2418 		if (w->w_arg && w->w_arg != ifp->if_index)
2419 			continue;
2420 		ifa = ifp->if_addr;
2421 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2422 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2423 			if (af && af != ifma->ifma_addr->sa_family)
2424 				continue;
2425 			if (prison_if(w->w_req->td->td_ucred,
2426 			    ifma->ifma_addr) != 0)
2427 				continue;
2428 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2429 			info.rti_info[RTAX_GATEWAY] =
2430 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2431 			    ifma->ifma_lladdr : NULL;
2432 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2433 			if (error != 0)
2434 				break;
2435 			if (w->w_req && w->w_tmem) {
2436 				struct ifma_msghdr *ifmam;
2437 
2438 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2439 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2440 				ifmam->ifmam_flags = 0;
2441 				ifmam->ifmam_addrs = info.rti_addrs;
2442 				ifmam->_ifmam_spare1 = 0;
2443 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2444 				if (error != 0)
2445 					break;
2446 			}
2447 		}
2448 		if (error != 0)
2449 			break;
2450 	}
2451 	return (error);
2452 }
2453 
2454 static void
2455 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2456 {
2457 	union sockaddr_union sa_dst, sa_mask;
2458 
2459 	w->family = family;
2460 	w->dst = (struct sockaddr *)&sa_dst;
2461 	w->mask = (struct sockaddr *)&sa_mask;
2462 
2463 	init_sockaddrs_family(family, w->dst, w->mask);
2464 
2465 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2466 }
2467 
2468 static int
2469 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2470 {
2471 	struct epoch_tracker et;
2472 	int	*name = (int *)arg1;
2473 	u_int	namelen = arg2;
2474 	struct rib_head *rnh = NULL; /* silence compiler. */
2475 	int	i, lim, error = EINVAL;
2476 	int	fib = 0;
2477 	u_char	af;
2478 	struct	walkarg w;
2479 
2480 	name ++;
2481 	namelen--;
2482 	if (req->newptr)
2483 		return (EPERM);
2484 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2485 		if (namelen == 3)
2486 			fib = req->td->td_proc->p_fibnum;
2487 		else if (namelen == 4)
2488 			fib = (name[3] == RT_ALL_FIBS) ?
2489 			    req->td->td_proc->p_fibnum : name[3];
2490 		else
2491 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2492 		if (fib < 0 || fib >= rt_numfibs)
2493 			return (EINVAL);
2494 	} else if (namelen != 3)
2495 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2496 	af = name[0];
2497 	if (af > AF_MAX)
2498 		return (EINVAL);
2499 	bzero(&w, sizeof(w));
2500 	w.w_op = name[1];
2501 	w.w_arg = name[2];
2502 	w.w_req = req;
2503 
2504 	error = sysctl_wire_old_buffer(req, 0);
2505 	if (error)
2506 		return (error);
2507 
2508 	/*
2509 	 * Allocate reply buffer in advance.
2510 	 * All rtsock messages has maximum length of u_short.
2511 	 */
2512 	w.w_tmemsize = 65536;
2513 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2514 
2515 	NET_EPOCH_ENTER(et);
2516 	switch (w.w_op) {
2517 	case NET_RT_DUMP:
2518 	case NET_RT_FLAGS:
2519 		if (af == 0) {			/* dump all tables */
2520 			i = 1;
2521 			lim = AF_MAX;
2522 		} else				/* dump only one table */
2523 			i = lim = af;
2524 
2525 		/*
2526 		 * take care of llinfo entries, the caller must
2527 		 * specify an AF
2528 		 */
2529 		if (w.w_op == NET_RT_FLAGS &&
2530 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2531 			if (af != 0)
2532 				error = lltable_sysctl_dumparp(af, w.w_req);
2533 			else
2534 				error = EINVAL;
2535 			break;
2536 		}
2537 		/*
2538 		 * take care of routing entries
2539 		 */
2540 		for (error = 0; error == 0 && i <= lim; i++) {
2541 			rnh = rt_tables_get_rnh(fib, i);
2542 			if (rnh != NULL) {
2543 				rtable_sysctl_dump(fib, i, &w);
2544 			} else if (af != 0)
2545 				error = EAFNOSUPPORT;
2546 		}
2547 		break;
2548 	case NET_RT_NHOP:
2549 	case NET_RT_NHGRP:
2550 		/* Allow dumping one specific af/fib at a time */
2551 		if (namelen < 4) {
2552 			error = EINVAL;
2553 			break;
2554 		}
2555 		fib = name[3];
2556 		if (fib < 0 || fib > rt_numfibs) {
2557 			error = EINVAL;
2558 			break;
2559 		}
2560 		rnh = rt_tables_get_rnh(fib, af);
2561 		if (rnh == NULL) {
2562 			error = EAFNOSUPPORT;
2563 			break;
2564 		}
2565 		if (w.w_op == NET_RT_NHOP)
2566 			error = nhops_dump_sysctl(rnh, w.w_req);
2567 		else
2568 #ifdef ROUTE_MPATH
2569 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2570 #else
2571 			error = ENOTSUP;
2572 #endif
2573 		break;
2574 	case NET_RT_IFLIST:
2575 	case NET_RT_IFLISTL:
2576 		error = sysctl_iflist(af, &w);
2577 		break;
2578 
2579 	case NET_RT_IFMALIST:
2580 		error = sysctl_ifmalist(af, &w);
2581 		break;
2582 	}
2583 	NET_EPOCH_EXIT(et);
2584 
2585 	free(w.w_tmem, M_TEMP);
2586 	return (error);
2587 }
2588 
2589 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2590     sysctl_rtsock, "Return route tables and interface/address lists");
2591 
2592 /*
2593  * Definitions of protocols supported in the ROUTE domain.
2594  */
2595 
2596 static struct domain routedomain;		/* or at least forward */
2597 
2598 static struct protosw routesw[] = {
2599 {
2600 	.pr_type =		SOCK_RAW,
2601 	.pr_domain =		&routedomain,
2602 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2603 	.pr_output =		route_output,
2604 	.pr_ctlinput =		raw_ctlinput,
2605 	.pr_init =		raw_init,
2606 	.pr_usrreqs =		&route_usrreqs
2607 }
2608 };
2609 
2610 static struct domain routedomain = {
2611 	.dom_family =		PF_ROUTE,
2612 	.dom_name =		"route",
2613 	.dom_protosw =		routesw,
2614 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
2615 };
2616 
2617 VNET_DOMAIN_SET(route);
2618