xref: /freebsd/sys/net/rtsock.c (revision 410556f1f10fd35b350102725fd8504c3cb0afc8)
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 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
130 
131 /* NB: these are not modified */
132 static struct	sockaddr route_src = { 2, PF_ROUTE, };
133 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
134 
135 /* These are external hooks for CARP. */
136 int	(*carp_get_vhid_p)(struct ifaddr *);
137 
138 /*
139  * Used by rtsock/raw_input callback code to decide whether to filter the update
140  * notification to a socket bound to a particular FIB.
141  */
142 #define	RTS_FILTER_FIB	M_PROTO8
143 
144 typedef struct {
145 	int	ip_count;	/* attached w/ AF_INET */
146 	int	ip6_count;	/* attached w/ AF_INET6 */
147 	int	any_count;	/* total attached */
148 } route_cb_t;
149 VNET_DEFINE_STATIC(route_cb_t, route_cb);
150 #define	V_route_cb VNET(route_cb)
151 
152 struct mtx rtsock_mtx;
153 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
154 
155 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
156 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
157 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
158 
159 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
160 
161 struct walkarg {
162 	int	family;
163 	int	w_tmemsize;
164 	int	w_op, w_arg;
165 	caddr_t	w_tmem;
166 	struct sysctl_req *w_req;
167 	struct sockaddr *dst;
168 	struct sockaddr *mask;
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	cleanup_xaddrs(struct rt_addrinfo *info);
178 static int	sysctl_dumpentry(struct rtentry *rt, void *vw);
179 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
180 			uint32_t weight, struct walkarg *w);
181 static int	sysctl_iflist(int af, struct walkarg *w);
182 static int	sysctl_ifmalist(int af, struct walkarg *w);
183 static int	route_output(struct mbuf *m, struct socket *so, ...);
184 static void	rt_getmetrics(const struct rtentry *rt,
185 			const struct nhop_object *nh, struct rt_metrics *out);
186 static void	rt_dispatch(struct mbuf *, sa_family_t);
187 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
188 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
189 static int	update_rtm_from_rc(struct rt_addrinfo *info,
190 			struct rt_msghdr **prtm, int alloc_len,
191 			struct rib_cmd_info *rc, struct nhop_object *nh);
192 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
193 			struct mbuf *m, sa_family_t saf, u_int fibnum,
194 			int rtm_errno);
195 static bool	can_export_rte(struct ucred *td_ucred, bool rt_is_host,
196 			const struct sockaddr *rt_dst);
197 
198 static struct netisr_handler rtsock_nh = {
199 	.nh_name = "rtsock",
200 	.nh_handler = rts_input,
201 	.nh_proto = NETISR_ROUTE,
202 	.nh_policy = NETISR_POLICY_SOURCE,
203 };
204 
205 static int
206 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
207 {
208 	int error, qlimit;
209 
210 	netisr_getqlimit(&rtsock_nh, &qlimit);
211 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
212         if (error || !req->newptr)
213                 return (error);
214 	if (qlimit < 1)
215 		return (EINVAL);
216 	return (netisr_setqlimit(&rtsock_nh, qlimit));
217 }
218 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
219     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
220     0, 0, sysctl_route_netisr_maxqlen, "I",
221     "maximum routing socket dispatch queue length");
222 
223 static void
224 vnet_rts_init(void)
225 {
226 	int tmp;
227 
228 	if (IS_DEFAULT_VNET(curvnet)) {
229 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
230 			rtsock_nh.nh_qlimit = tmp;
231 		netisr_register(&rtsock_nh);
232 	}
233 #ifdef VIMAGE
234 	 else
235 		netisr_register_vnet(&rtsock_nh);
236 #endif
237 }
238 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
239     vnet_rts_init, 0);
240 
241 #ifdef VIMAGE
242 static void
243 vnet_rts_uninit(void)
244 {
245 
246 	netisr_unregister_vnet(&rtsock_nh);
247 }
248 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
249     vnet_rts_uninit, 0);
250 #endif
251 
252 static int
253 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
254     struct rawcb *rp)
255 {
256 	int fibnum;
257 
258 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
259 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
260 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
261 
262 	/* No filtering requested. */
263 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
264 		return (0);
265 
266 	/* Check if it is a rts and the fib matches the one of the socket. */
267 	fibnum = M_GETFIB(m);
268 	if (proto->sp_family != PF_ROUTE ||
269 	    rp->rcb_socket == NULL ||
270 	    rp->rcb_socket->so_fibnum == fibnum)
271 		return (0);
272 
273 	/* Filtering requested and no match, the socket shall be skipped. */
274 	return (1);
275 }
276 
277 static void
278 rts_input(struct mbuf *m)
279 {
280 	struct sockproto route_proto;
281 	unsigned short *family;
282 	struct m_tag *tag;
283 
284 	route_proto.sp_family = PF_ROUTE;
285 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
286 	if (tag != NULL) {
287 		family = (unsigned short *)(tag + 1);
288 		route_proto.sp_protocol = *family;
289 		m_tag_delete(m, tag);
290 	} else
291 		route_proto.sp_protocol = 0;
292 
293 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
294 }
295 
296 /*
297  * It really doesn't make any sense at all for this code to share much
298  * with raw_usrreq.c, since its functionality is so restricted.  XXX
299  */
300 static void
301 rts_abort(struct socket *so)
302 {
303 
304 	raw_usrreqs.pru_abort(so);
305 }
306 
307 static void
308 rts_close(struct socket *so)
309 {
310 
311 	raw_usrreqs.pru_close(so);
312 }
313 
314 /* pru_accept is EOPNOTSUPP */
315 
316 static int
317 rts_attach(struct socket *so, int proto, struct thread *td)
318 {
319 	struct rawcb *rp;
320 	int error;
321 
322 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
323 
324 	/* XXX */
325 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
326 
327 	so->so_pcb = (caddr_t)rp;
328 	so->so_fibnum = td->td_proc->p_fibnum;
329 	error = raw_attach(so, proto);
330 	rp = sotorawcb(so);
331 	if (error) {
332 		so->so_pcb = NULL;
333 		free(rp, M_PCB);
334 		return error;
335 	}
336 	RTSOCK_LOCK();
337 	switch(rp->rcb_proto.sp_protocol) {
338 	case AF_INET:
339 		V_route_cb.ip_count++;
340 		break;
341 	case AF_INET6:
342 		V_route_cb.ip6_count++;
343 		break;
344 	}
345 	V_route_cb.any_count++;
346 	RTSOCK_UNLOCK();
347 	soisconnected(so);
348 	so->so_options |= SO_USELOOPBACK;
349 	return 0;
350 }
351 
352 static int
353 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
354 {
355 
356 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
357 }
358 
359 static int
360 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
361 {
362 
363 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
364 }
365 
366 /* pru_connect2 is EOPNOTSUPP */
367 /* pru_control is EOPNOTSUPP */
368 
369 static void
370 rts_detach(struct socket *so)
371 {
372 	struct rawcb *rp = sotorawcb(so);
373 
374 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
375 
376 	RTSOCK_LOCK();
377 	switch(rp->rcb_proto.sp_protocol) {
378 	case AF_INET:
379 		V_route_cb.ip_count--;
380 		break;
381 	case AF_INET6:
382 		V_route_cb.ip6_count--;
383 		break;
384 	}
385 	V_route_cb.any_count--;
386 	RTSOCK_UNLOCK();
387 	raw_usrreqs.pru_detach(so);
388 }
389 
390 static int
391 rts_disconnect(struct socket *so)
392 {
393 
394 	return (raw_usrreqs.pru_disconnect(so));
395 }
396 
397 /* pru_listen is EOPNOTSUPP */
398 
399 static int
400 rts_peeraddr(struct socket *so, struct sockaddr **nam)
401 {
402 
403 	return (raw_usrreqs.pru_peeraddr(so, nam));
404 }
405 
406 /* pru_rcvd is EOPNOTSUPP */
407 /* pru_rcvoob is EOPNOTSUPP */
408 
409 static int
410 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
411 	 struct mbuf *control, struct thread *td)
412 {
413 
414 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
415 }
416 
417 /* pru_sense is null */
418 
419 static int
420 rts_shutdown(struct socket *so)
421 {
422 
423 	return (raw_usrreqs.pru_shutdown(so));
424 }
425 
426 static int
427 rts_sockaddr(struct socket *so, struct sockaddr **nam)
428 {
429 
430 	return (raw_usrreqs.pru_sockaddr(so, nam));
431 }
432 
433 static struct pr_usrreqs route_usrreqs = {
434 	.pru_abort =		rts_abort,
435 	.pru_attach =		rts_attach,
436 	.pru_bind =		rts_bind,
437 	.pru_connect =		rts_connect,
438 	.pru_detach =		rts_detach,
439 	.pru_disconnect =	rts_disconnect,
440 	.pru_peeraddr =		rts_peeraddr,
441 	.pru_send =		rts_send,
442 	.pru_shutdown =		rts_shutdown,
443 	.pru_sockaddr =		rts_sockaddr,
444 	.pru_close =		rts_close,
445 };
446 
447 #ifndef _SOCKADDR_UNION_DEFINED
448 #define	_SOCKADDR_UNION_DEFINED
449 /*
450  * The union of all possible address formats we handle.
451  */
452 union sockaddr_union {
453 	struct sockaddr		sa;
454 	struct sockaddr_in	sin;
455 	struct sockaddr_in6	sin6;
456 };
457 #endif /* _SOCKADDR_UNION_DEFINED */
458 
459 static int
460 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
461     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
462 {
463 #if defined(INET) || defined(INET6)
464 	struct epoch_tracker et;
465 #endif
466 
467 	/* First, see if the returned address is part of the jail. */
468 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
469 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
470 		return (0);
471 	}
472 
473 	switch (info->rti_info[RTAX_DST]->sa_family) {
474 #ifdef INET
475 	case AF_INET:
476 	{
477 		struct in_addr ia;
478 		struct ifaddr *ifa;
479 		int found;
480 
481 		found = 0;
482 		/*
483 		 * Try to find an address on the given outgoing interface
484 		 * that belongs to the jail.
485 		 */
486 		NET_EPOCH_ENTER(et);
487 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
488 			struct sockaddr *sa;
489 			sa = ifa->ifa_addr;
490 			if (sa->sa_family != AF_INET)
491 				continue;
492 			ia = ((struct sockaddr_in *)sa)->sin_addr;
493 			if (prison_check_ip4(cred, &ia) == 0) {
494 				found = 1;
495 				break;
496 			}
497 		}
498 		NET_EPOCH_EXIT(et);
499 		if (!found) {
500 			/*
501 			 * As a last resort return the 'default' jail address.
502 			 */
503 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
504 			    sin_addr;
505 			if (prison_get_ip4(cred, &ia) != 0)
506 				return (ESRCH);
507 		}
508 		bzero(&saun->sin, sizeof(struct sockaddr_in));
509 		saun->sin.sin_len = sizeof(struct sockaddr_in);
510 		saun->sin.sin_family = AF_INET;
511 		saun->sin.sin_addr.s_addr = ia.s_addr;
512 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
513 		break;
514 	}
515 #endif
516 #ifdef INET6
517 	case AF_INET6:
518 	{
519 		struct in6_addr ia6;
520 		struct ifaddr *ifa;
521 		int found;
522 
523 		found = 0;
524 		/*
525 		 * Try to find an address on the given outgoing interface
526 		 * that belongs to the jail.
527 		 */
528 		NET_EPOCH_ENTER(et);
529 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
530 			struct sockaddr *sa;
531 			sa = ifa->ifa_addr;
532 			if (sa->sa_family != AF_INET6)
533 				continue;
534 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
535 			    &ia6, sizeof(struct in6_addr));
536 			if (prison_check_ip6(cred, &ia6) == 0) {
537 				found = 1;
538 				break;
539 			}
540 		}
541 		NET_EPOCH_EXIT(et);
542 		if (!found) {
543 			/*
544 			 * As a last resort return the 'default' jail address.
545 			 */
546 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
547 			    sin6_addr;
548 			if (prison_get_ip6(cred, &ia6) != 0)
549 				return (ESRCH);
550 		}
551 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
552 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
553 		saun->sin6.sin6_family = AF_INET6;
554 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
555 		if (sa6_recoverscope(&saun->sin6) != 0)
556 			return (ESRCH);
557 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
558 		break;
559 	}
560 #endif
561 	default:
562 		return (ESRCH);
563 	}
564 	return (0);
565 }
566 
567 static int
568 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
569 {
570 	struct ifaddr *ifa;
571 	sa_family_t saf;
572 
573 	if (V_loif == NULL) {
574 		printf("Unable to add blackhole/reject nhop without loopback");
575 		return (ENOTSUP);
576 	}
577 	info->rti_ifp = V_loif;
578 
579 	saf = info->rti_info[RTAX_DST]->sa_family;
580 
581 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
582 		if (ifa->ifa_addr->sa_family == saf) {
583 			info->rti_ifa = ifa;
584 			break;
585 		}
586 	}
587 	if (info->rti_ifa == NULL)
588 		return (ENOTSUP);
589 
590 	bzero(saun, sizeof(union sockaddr_union));
591 	switch (saf) {
592 #ifdef INET
593 	case AF_INET:
594 		saun->sin.sin_family = AF_INET;
595 		saun->sin.sin_len = sizeof(struct sockaddr_in);
596 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
597 		break;
598 #endif
599 #ifdef INET6
600 	case AF_INET6:
601 		saun->sin6.sin6_family = AF_INET6;
602 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
603 		saun->sin6.sin6_addr = in6addr_loopback;
604 		break;
605 #endif
606 	default:
607 		return (ENOTSUP);
608 	}
609 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
610 	info->rti_flags |= RTF_GATEWAY;
611 
612 	return (0);
613 }
614 
615 /*
616  * Fills in @info based on userland-provided @rtm message.
617  *
618  * Returns 0 on success.
619  */
620 static int
621 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
622 {
623 	int error;
624 	sa_family_t saf;
625 
626 	rtm->rtm_pid = curproc->p_pid;
627 	info->rti_addrs = rtm->rtm_addrs;
628 
629 	info->rti_mflags = rtm->rtm_inits;
630 	info->rti_rmx = &rtm->rtm_rmx;
631 
632 	/*
633 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
634 	 * link-local address because rtrequest requires addresses with
635 	 * embedded scope id.
636 	 */
637 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
638 		return (EINVAL);
639 
640 	info->rti_flags = rtm->rtm_flags;
641 	error = cleanup_xaddrs(info);
642 	if (error != 0)
643 		return (error);
644 	saf = info->rti_info[RTAX_DST]->sa_family;
645 	/*
646 	 * Verify that the caller has the appropriate privilege; RTM_GET
647 	 * is the only operation the non-superuser is allowed.
648 	 */
649 	if (rtm->rtm_type != RTM_GET) {
650 		error = priv_check(curthread, PRIV_NET_ROUTE);
651 		if (error != 0)
652 			return (error);
653 	}
654 
655 	/*
656 	 * The given gateway address may be an interface address.
657 	 * For example, issuing a "route change" command on a route
658 	 * entry that was created from a tunnel, and the gateway
659 	 * address given is the local end point. In this case the
660 	 * RTF_GATEWAY flag must be cleared or the destination will
661 	 * not be reachable even though there is no error message.
662 	 */
663 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
664 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
665 		struct rt_addrinfo ginfo;
666 		struct sockaddr *gdst;
667 		struct sockaddr_storage ss;
668 
669 		bzero(&ginfo, sizeof(ginfo));
670 		bzero(&ss, sizeof(ss));
671 		ss.ss_len = sizeof(ss);
672 
673 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
674 		gdst = info->rti_info[RTAX_GATEWAY];
675 
676 		/*
677 		 * A host route through the loopback interface is
678 		 * installed for each interface adddress. In pre 8.0
679 		 * releases the interface address of a PPP link type
680 		 * is not reachable locally. This behavior is fixed as
681 		 * part of the new L2/L3 redesign and rewrite work. The
682 		 * signature of this interface address route is the
683 		 * AF_LINK sa_family type of the gateway, and the
684 		 * rt_ifp has the IFF_LOOPBACK flag set.
685 		 */
686 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
687 			if (ss.ss_family == AF_LINK &&
688 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
689 				info->rti_flags &= ~RTF_GATEWAY;
690 				info->rti_flags |= RTF_GWFLAG_COMPAT;
691 			}
692 			rib_free_info(&ginfo);
693 		}
694 	}
695 
696 	return (0);
697 }
698 
699 static struct nhop_object *
700 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
701 {
702 	if (!NH_IS_NHGRP(nh))
703 		return (nh);
704 #ifdef ROUTE_MPATH
705 	struct weightened_nhop *wn;
706 	uint32_t num_nhops;
707 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
708 	if (gw == NULL)
709 		return (wn[0].nh);
710 	for (int i = 0; i < num_nhops; i++) {
711 		if (match_nhop_gw(wn[i].nh, gw))
712 			return (wn[i].nh);
713 	}
714 #endif
715 	return (NULL);
716 }
717 
718 /*
719  * Handles RTM_GET message from routing socket, returning matching rt.
720  *
721  * Returns:
722  * 0 on success, with locked and referenced matching rt in @rt_nrt
723  * errno of failure
724  */
725 static int
726 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
727     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
728 {
729 	RIB_RLOCK_TRACKER;
730 	struct rib_head *rnh;
731 	struct nhop_object *nh;
732 	sa_family_t saf;
733 
734 	saf = info->rti_info[RTAX_DST]->sa_family;
735 
736 	rnh = rt_tables_get_rnh(fibnum, saf);
737 	if (rnh == NULL)
738 		return (EAFNOSUPPORT);
739 
740 	RIB_RLOCK(rnh);
741 
742 	/*
743 	 * By (implicit) convention host route (one without netmask)
744 	 * means longest-prefix-match request and the route with netmask
745 	 * means exact-match lookup.
746 	 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
747 	 * prefixes, use original data to check for the netmask presence.
748 	 */
749 	if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
750 		/*
751 		 * Provide longest prefix match for
752 		 * address lookup (no mask).
753 		 * 'route -n get addr'
754 		 */
755 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
756 		    info->rti_info[RTAX_DST], &rnh->head);
757 	} else
758 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
759 		    info->rti_info[RTAX_DST],
760 		    info->rti_info[RTAX_NETMASK], &rnh->head);
761 
762 	if (rc->rc_rt == NULL) {
763 		RIB_RUNLOCK(rnh);
764 		return (ESRCH);
765 	}
766 
767 	nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
768 	if (nh == NULL) {
769 		RIB_RUNLOCK(rnh);
770 		return (ESRCH);
771 	}
772 	/*
773 	 * If performing proxied L2 entry insertion, and
774 	 * the actual PPP host entry is found, perform
775 	 * another search to retrieve the prefix route of
776 	 * the local end point of the PPP link.
777 	 * TODO: move this logic to userland.
778 	 */
779 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
780 		struct sockaddr laddr;
781 
782 		if (nh->nh_ifp != NULL &&
783 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
784 			struct ifaddr *ifa;
785 
786 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
787 					RT_ALL_FIBS);
788 			if (ifa != NULL)
789 				rt_maskedcopy(ifa->ifa_addr,
790 					      &laddr,
791 					      ifa->ifa_netmask);
792 		} else
793 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
794 				      &laddr,
795 				      nh->nh_ifa->ifa_netmask);
796 		/*
797 		 * refactor rt and no lock operation necessary
798 		 */
799 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
800 		    &rnh->head);
801 		if (rc->rc_rt == NULL) {
802 			RIB_RUNLOCK(rnh);
803 			return (ESRCH);
804 		}
805 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
806 		if (nh == NULL) {
807 			RIB_RUNLOCK(rnh);
808 			return (ESRCH);
809 		}
810 	}
811 	rc->rc_nh_new = nh;
812 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
813 	RIB_RUNLOCK(rnh);
814 
815 	return (0);
816 }
817 
818 static void
819 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
820 {
821 #ifdef INET
822 	if (family == AF_INET) {
823 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
824 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
825 
826 		bzero(dst4, sizeof(struct sockaddr_in));
827 		bzero(mask4, sizeof(struct sockaddr_in));
828 
829 		dst4->sin_family = AF_INET;
830 		dst4->sin_len = sizeof(struct sockaddr_in);
831 		mask4->sin_family = AF_INET;
832 		mask4->sin_len = sizeof(struct sockaddr_in);
833 	}
834 #endif
835 #ifdef INET6
836 	if (family == AF_INET6) {
837 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
838 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
839 
840 		bzero(dst6, sizeof(struct sockaddr_in6));
841 		bzero(mask6, sizeof(struct sockaddr_in6));
842 
843 		dst6->sin6_family = AF_INET6;
844 		dst6->sin6_len = sizeof(struct sockaddr_in6);
845 		mask6->sin6_family = AF_INET6;
846 		mask6->sin6_len = sizeof(struct sockaddr_in6);
847 	}
848 #endif
849 }
850 
851 static void
852 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
853     struct sockaddr *mask)
854 {
855 #ifdef INET
856 	if (dst->sa_family == AF_INET) {
857 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
858 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
859 		uint32_t scopeid = 0;
860 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
861 		    &scopeid);
862 		return;
863 	}
864 #endif
865 #ifdef INET6
866 	if (dst->sa_family == AF_INET6) {
867 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
868 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
869 		uint32_t scopeid = 0;
870 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
871 		    &mask6->sin6_addr, &scopeid);
872 		dst6->sin6_scope_id = scopeid;
873 		return;
874 	}
875 #endif
876 }
877 
878 
879 /*
880  * Update sockaddrs, flags, etc in @prtm based on @rc data.
881  * rtm can be reallocated.
882  *
883  * Returns 0 on success, along with pointer to (potentially reallocated)
884  *  rtm.
885  *
886  */
887 static int
888 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
889     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
890 {
891 	struct walkarg w;
892 	union sockaddr_union saun;
893 	struct rt_msghdr *rtm, *orig_rtm = NULL;
894 	struct ifnet *ifp;
895 	int error, len;
896 
897 	rtm = *prtm;
898 	union sockaddr_union sa_dst, sa_mask;
899 	int family = info->rti_info[RTAX_DST]->sa_family;
900 	init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
901 	export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
902 
903 	info->rti_info[RTAX_DST] = &sa_dst.sa;
904 	info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
905 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
906 	info->rti_info[RTAX_GENMASK] = 0;
907 	ifp = nh->nh_ifp;
908 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
909 		if (ifp) {
910 			info->rti_info[RTAX_IFP] =
911 			    ifp->if_addr->ifa_addr;
912 			error = rtm_get_jailed(info, ifp, nh,
913 			    &saun, curthread->td_ucred);
914 			if (error != 0)
915 				return (error);
916 			if (ifp->if_flags & IFF_POINTOPOINT)
917 				info->rti_info[RTAX_BRD] =
918 				    nh->nh_ifa->ifa_dstaddr;
919 			rtm->rtm_index = ifp->if_index;
920 		} else {
921 			info->rti_info[RTAX_IFP] = NULL;
922 			info->rti_info[RTAX_IFA] = NULL;
923 		}
924 	} else if (ifp != NULL)
925 		rtm->rtm_index = ifp->if_index;
926 
927 	/* Check if we need to realloc storage */
928 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
929 	if (len > alloc_len) {
930 		struct rt_msghdr *tmp_rtm;
931 
932 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
933 		if (tmp_rtm == NULL)
934 			return (ENOBUFS);
935 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
936 		orig_rtm = rtm;
937 		rtm = tmp_rtm;
938 		alloc_len = len;
939 
940 		/*
941 		 * Delay freeing original rtm as info contains
942 		 * data referencing it.
943 		 */
944 	}
945 
946 	w.w_tmem = (caddr_t)rtm;
947 	w.w_tmemsize = alloc_len;
948 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
949 
950 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
951 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
952 		rtm->rtm_flags = RTF_GATEWAY |
953 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
954 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
955 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
956 	rtm->rtm_addrs = info->rti_addrs;
957 
958 	if (orig_rtm != NULL)
959 		free(orig_rtm, M_TEMP);
960 	*prtm = rtm;
961 
962 	return (0);
963 }
964 
965 #ifdef ROUTE_MPATH
966 static void
967 save_del_notification(struct rib_cmd_info *rc, void *_cbdata)
968 {
969 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
970 
971 	if (rc->rc_cmd == RTM_DELETE)
972 		*rc_new = *rc;
973 }
974 
975 static void
976 save_add_notification(struct rib_cmd_info *rc, void *_cbdata)
977 {
978 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
979 
980 	if (rc->rc_cmd == RTM_ADD)
981 		*rc_new = *rc;
982 }
983 #endif
984 
985 /*ARGSUSED*/
986 static int
987 route_output(struct mbuf *m, struct socket *so, ...)
988 {
989 	struct rt_msghdr *rtm = NULL;
990 	struct rtentry *rt = NULL;
991 	struct rt_addrinfo info;
992 	struct epoch_tracker et;
993 #ifdef INET6
994 	struct sockaddr_storage ss;
995 	struct sockaddr_in6 *sin6;
996 	int i, rti_need_deembed = 0;
997 #endif
998 	int alloc_len = 0, len, error = 0, fibnum;
999 	sa_family_t saf = AF_UNSPEC;
1000 	struct rib_cmd_info rc;
1001 	struct nhop_object *nh;
1002 
1003 	fibnum = so->so_fibnum;
1004 #define senderr(e) { error = e; goto flush;}
1005 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1006 		       (m = m_pullup(m, sizeof(long))) == NULL))
1007 		return (ENOBUFS);
1008 	if ((m->m_flags & M_PKTHDR) == 0)
1009 		panic("route_output");
1010 	NET_EPOCH_ENTER(et);
1011 	len = m->m_pkthdr.len;
1012 	if (len < sizeof(*rtm) ||
1013 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1014 		senderr(EINVAL);
1015 
1016 	/*
1017 	 * Most of current messages are in range 200-240 bytes,
1018 	 * minimize possible re-allocation on reply using larger size
1019 	 * buffer aligned on 1k boundaty.
1020 	 */
1021 	alloc_len = roundup2(len, 1024);
1022 	if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
1023 		senderr(ENOBUFS);
1024 
1025 	m_copydata(m, 0, len, (caddr_t)rtm);
1026 	bzero(&info, sizeof(info));
1027 	nh = NULL;
1028 
1029 	if (rtm->rtm_version != RTM_VERSION) {
1030 		/* Do not touch message since format is unknown */
1031 		free(rtm, M_TEMP);
1032 		rtm = NULL;
1033 		senderr(EPROTONOSUPPORT);
1034 	}
1035 
1036 	/*
1037 	 * Starting from here, it is possible
1038 	 * to alter original message and insert
1039 	 * caller PID and error value.
1040 	 */
1041 
1042 	if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) {
1043 		senderr(error);
1044 	}
1045 
1046 	saf = info.rti_info[RTAX_DST]->sa_family;
1047 
1048 	/* support for new ARP code */
1049 	if (rtm->rtm_flags & RTF_LLDATA) {
1050 		error = lla_rt_output(rtm, &info);
1051 #ifdef INET6
1052 		if (error == 0)
1053 			rti_need_deembed = 1;
1054 #endif
1055 		goto flush;
1056 	}
1057 
1058 	union sockaddr_union gw_saun;
1059 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1060 	if (blackhole_flags != 0) {
1061 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1062 			error = fill_blackholeinfo(&info, &gw_saun);
1063 		else
1064 			error = EINVAL;
1065 		if (error != 0)
1066 			senderr(error);
1067 		/* TODO: rebuild rtm from scratch */
1068 	}
1069 
1070 	switch (rtm->rtm_type) {
1071 	case RTM_ADD:
1072 	case RTM_CHANGE:
1073 		if (rtm->rtm_type == RTM_ADD) {
1074 			if (info.rti_info[RTAX_GATEWAY] == NULL)
1075 				senderr(EINVAL);
1076 		}
1077 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1078 		if (error == 0) {
1079 #ifdef INET6
1080 			rti_need_deembed = 1;
1081 #endif
1082 #ifdef ROUTE_MPATH
1083 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1084 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1085 				struct rib_cmd_info rc_simple = {};
1086 				rib_decompose_notification(&rc,
1087 				    save_add_notification, (void *)&rc_simple);
1088 				rc = rc_simple;
1089 			}
1090 #endif
1091 			nh = rc.rc_nh_new;
1092 			rtm->rtm_index = nh->nh_ifp->if_index;
1093 			rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1094 		}
1095 		break;
1096 
1097 	case RTM_DELETE:
1098 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1099 		if (error == 0) {
1100 #ifdef ROUTE_MPATH
1101 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1102 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1103 				struct rib_cmd_info rc_simple = {};
1104 				rib_decompose_notification(&rc,
1105 				    save_del_notification, (void *)&rc_simple);
1106 				rc = rc_simple;
1107 			}
1108 #endif
1109 			nh = rc.rc_nh_old;
1110 			goto report;
1111 		}
1112 #ifdef INET6
1113 		/* rt_msg2() will not be used when RTM_DELETE fails. */
1114 		rti_need_deembed = 1;
1115 #endif
1116 		break;
1117 
1118 	case RTM_GET:
1119 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1120 		if (error != 0)
1121 			senderr(error);
1122 		nh = rc.rc_nh_new;
1123 
1124 report:
1125 		if (!can_export_rte(curthread->td_ucred,
1126 		    info.rti_info[RTAX_NETMASK] == NULL,
1127 		    info.rti_info[RTAX_DST])) {
1128 			senderr(ESRCH);
1129 		}
1130 
1131 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1132 		/*
1133 		 * Note that some sockaddr pointers may have changed to
1134 		 * point to memory outsize @rtm. Some may be pointing
1135 		 * to the on-stack variables.
1136 		 * Given that, any pointer in @info CANNOT BE USED.
1137 		 */
1138 
1139 		/*
1140 		 * scopeid deembedding has been performed while
1141 		 * writing updated rtm in rtsock_msg_buffer().
1142 		 * With that in mind, skip deembedding procedure below.
1143 		 */
1144 #ifdef INET6
1145 		rti_need_deembed = 0;
1146 #endif
1147 		if (error != 0)
1148 			senderr(error);
1149 		break;
1150 
1151 	default:
1152 		senderr(EOPNOTSUPP);
1153 	}
1154 
1155 flush:
1156 	NET_EPOCH_EXIT(et);
1157 	rt = NULL;
1158 
1159 #ifdef INET6
1160 	if (rtm != NULL) {
1161 		if (rti_need_deembed) {
1162 			/* sin6_scope_id is recovered before sending rtm. */
1163 			sin6 = (struct sockaddr_in6 *)&ss;
1164 			for (i = 0; i < RTAX_MAX; i++) {
1165 				if (info.rti_info[i] == NULL)
1166 					continue;
1167 				if (info.rti_info[i]->sa_family != AF_INET6)
1168 					continue;
1169 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1170 				if (sa6_recoverscope(sin6) == 0)
1171 					bcopy(sin6, info.rti_info[i],
1172 						    sizeof(*sin6));
1173 			}
1174 		}
1175 	}
1176 #endif
1177 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1178 
1179 	return (error);
1180 }
1181 
1182 /*
1183  * Sends the prepared reply message in @rtm to all rtsock clients.
1184  * Frees @m and @rtm.
1185  *
1186  */
1187 static void
1188 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1189     sa_family_t saf, u_int fibnum, int rtm_errno)
1190 {
1191 	struct rawcb *rp = NULL;
1192 
1193 	/*
1194 	 * Check to see if we don't want our own messages.
1195 	 */
1196 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1197 		if (V_route_cb.any_count <= 1) {
1198 			if (rtm != NULL)
1199 				free(rtm, M_TEMP);
1200 			m_freem(m);
1201 			return;
1202 		}
1203 		/* There is another listener, so construct message */
1204 		rp = sotorawcb(so);
1205 	}
1206 
1207 	if (rtm != NULL) {
1208 		if (rtm_errno!= 0)
1209 			rtm->rtm_errno = rtm_errno;
1210 		else
1211 			rtm->rtm_flags |= RTF_DONE;
1212 
1213 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1214 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1215 			m_freem(m);
1216 			m = NULL;
1217 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1218 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1219 
1220 		free(rtm, M_TEMP);
1221 	}
1222 	if (m != NULL) {
1223 		M_SETFIB(m, fibnum);
1224 		m->m_flags |= RTS_FILTER_FIB;
1225 		if (rp) {
1226 			/*
1227 			 * XXX insure we don't get a copy by
1228 			 * invalidating our protocol
1229 			 */
1230 			unsigned short family = rp->rcb_proto.sp_family;
1231 			rp->rcb_proto.sp_family = 0;
1232 			rt_dispatch(m, saf);
1233 			rp->rcb_proto.sp_family = family;
1234 		} else
1235 			rt_dispatch(m, saf);
1236 	}
1237 }
1238 
1239 static void
1240 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1241     struct rt_metrics *out)
1242 {
1243 
1244 	bzero(out, sizeof(*out));
1245 	out->rmx_mtu = nh->nh_mtu;
1246 	out->rmx_weight = rt->rt_weight;
1247 	out->rmx_nhidx = nhop_get_idx(nh);
1248 	/* Kernel -> userland timebase conversion. */
1249 	out->rmx_expire = rt->rt_expire ?
1250 	    rt->rt_expire - time_uptime + time_second : 0;
1251 }
1252 
1253 /*
1254  * Extract the addresses of the passed sockaddrs.
1255  * Do a little sanity checking so as to avoid bad memory references.
1256  * This data is derived straight from userland.
1257  */
1258 static int
1259 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1260 {
1261 	struct sockaddr *sa;
1262 	int i;
1263 
1264 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1265 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1266 			continue;
1267 		sa = (struct sockaddr *)cp;
1268 		/*
1269 		 * It won't fit.
1270 		 */
1271 		if (cp + sa->sa_len > cplim)
1272 			return (EINVAL);
1273 		/*
1274 		 * there are no more.. quit now
1275 		 * If there are more bits, they are in error.
1276 		 * I've seen this. route(1) can evidently generate these.
1277 		 * This causes kernel to core dump.
1278 		 * for compatibility, If we see this, point to a safe address.
1279 		 */
1280 		if (sa->sa_len == 0) {
1281 			rtinfo->rti_info[i] = &sa_zero;
1282 			return (0); /* should be EINVAL but for compat */
1283 		}
1284 		/* accept it */
1285 #ifdef INET6
1286 		if (sa->sa_family == AF_INET6)
1287 			sa6_embedscope((struct sockaddr_in6 *)sa,
1288 			    V_ip6_use_defzone);
1289 #endif
1290 		rtinfo->rti_info[i] = sa;
1291 		cp += SA_SIZE(sa);
1292 	}
1293 	return (0);
1294 }
1295 
1296 #ifdef INET
1297 static inline void
1298 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1299 {
1300 
1301 	const struct sockaddr_in nsin = {
1302 		.sin_family = AF_INET,
1303 		.sin_len = sizeof(struct sockaddr_in),
1304 		.sin_addr = addr,
1305 	};
1306 	*sin = nsin;
1307 }
1308 #endif
1309 
1310 #ifdef INET6
1311 static inline void
1312 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1313     uint32_t scopeid)
1314 {
1315 
1316 	const struct sockaddr_in6 nsin6 = {
1317 		.sin6_family = AF_INET6,
1318 		.sin6_len = sizeof(struct sockaddr_in6),
1319 		.sin6_addr = *addr6,
1320 		.sin6_scope_id = scopeid,
1321 	};
1322 	*sin6 = nsin6;
1323 }
1324 #endif
1325 
1326 /*
1327  * Checks if gateway is suitable for lltable operations.
1328  * Lltable code requires AF_LINK gateway with ifindex
1329  *  and mac address specified.
1330  * Returns 0 on success.
1331  */
1332 static int
1333 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1334 {
1335 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1336 
1337 	if (sdl->sdl_family != AF_LINK)
1338 		return (EINVAL);
1339 
1340 	if (sdl->sdl_index == 0)
1341 		return (EINVAL);
1342 
1343 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len)
1344 		return (EINVAL);
1345 
1346 	return (0);
1347 }
1348 
1349 static int
1350 cleanup_xaddrs_gateway(struct rt_addrinfo *info)
1351 {
1352 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1353 
1354 	if (info->rti_flags & RTF_LLDATA)
1355 		return (cleanup_xaddrs_lladdr(info));
1356 
1357 	switch (gw->sa_family) {
1358 #ifdef INET
1359 	case AF_INET:
1360 		{
1361 			struct sockaddr_in *gw_sin = (struct sockaddr_in *)gw;
1362 			if (gw_sin->sin_len < sizeof(struct sockaddr_in)) {
1363 				printf("gw sin_len too small\n");
1364 				return (EINVAL);
1365 			}
1366 			fill_sockaddr_inet(gw_sin, gw_sin->sin_addr);
1367 		}
1368 		break;
1369 #endif
1370 #ifdef INET6
1371 	case AF_INET6:
1372 		{
1373 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1374 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1375 				printf("gw sin6_len too small\n");
1376 				return (EINVAL);
1377 			}
1378 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1379 			break;
1380 		}
1381 #endif
1382 	case AF_LINK:
1383 		{
1384 			struct sockaddr_dl_short *gw_sdl;
1385 
1386 			gw_sdl = (struct sockaddr_dl_short *)gw;
1387 			if (gw_sdl->sdl_len < sizeof(struct sockaddr_dl_short)) {
1388 				printf("gw sdl_len too small\n");
1389 				return (EINVAL);
1390 			}
1391 
1392 			const struct sockaddr_dl_short sdl = {
1393 				.sdl_family = AF_LINK,
1394 				.sdl_len = sizeof(struct sockaddr_dl_short),
1395 				.sdl_index = gw_sdl->sdl_index,
1396 			};
1397 			*gw_sdl = sdl;
1398 			break;
1399 		}
1400 	}
1401 
1402 	return (0);
1403 }
1404 
1405 static void
1406 remove_netmask(struct rt_addrinfo *info)
1407 {
1408 	info->rti_info[RTAX_NETMASK] = NULL;
1409 	info->rti_flags |= RTF_HOST;
1410 	info->rti_addrs &= ~RTA_NETMASK;
1411 }
1412 
1413 #ifdef INET
1414 static int
1415 cleanup_xaddrs_inet(struct rt_addrinfo *info)
1416 {
1417 	struct sockaddr_in *dst_sa, *mask_sa;
1418 
1419 	/* Check & fixup dst/netmask combination first */
1420 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1421 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1422 
1423 	struct in_addr mask = {
1424 		.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST,
1425 	};
1426 	struct in_addr dst = {
1427 		.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr))
1428 	};
1429 
1430 	if (dst_sa->sin_len < sizeof(struct sockaddr_in)) {
1431 		printf("dst sin_len too small\n");
1432 		return (EINVAL);
1433 	}
1434 	if (mask_sa && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1435 		printf("mask sin_len too small\n");
1436 		return (EINVAL);
1437 	}
1438 	fill_sockaddr_inet(dst_sa, dst);
1439 
1440 	if (mask.s_addr != INADDR_BROADCAST)
1441 		fill_sockaddr_inet(mask_sa, mask);
1442 	else
1443 		remove_netmask(info);
1444 
1445 	/* Check gateway */
1446 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1447 		return (cleanup_xaddrs_gateway(info));
1448 
1449 	return (0);
1450 }
1451 #endif
1452 
1453 #ifdef INET6
1454 static int
1455 cleanup_xaddrs_inet6(struct rt_addrinfo *info)
1456 {
1457 	struct sockaddr_in6 *dst_sa, *mask_sa;
1458 	struct in6_addr mask;
1459 
1460 	/* Check & fixup dst/netmask combination first */
1461 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1462 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1463 
1464 	mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1465 	IN6_MASK_ADDR(&dst_sa->sin6_addr, &mask);
1466 
1467 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1468 		printf("dst sin6_len too small\n");
1469 		return (EINVAL);
1470 	}
1471 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1472 		printf("mask sin6_len too small\n");
1473 		return (EINVAL);
1474 	}
1475 	fill_sockaddr_inet6(dst_sa, &dst_sa->sin6_addr, 0);
1476 
1477 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128))
1478 		fill_sockaddr_inet6(mask_sa, &mask, 0);
1479 	else
1480 		remove_netmask(info);
1481 
1482 	/* Check gateway */
1483 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1484 		return (cleanup_xaddrs_gateway(info));
1485 
1486 	return (0);
1487 }
1488 #endif
1489 
1490 static int
1491 cleanup_xaddrs(struct rt_addrinfo *info)
1492 {
1493 	int error = EAFNOSUPPORT;
1494 
1495 	if (info->rti_info[RTAX_DST] == NULL)
1496 		return (EINVAL);
1497 
1498 	if (info->rti_flags & RTF_LLDATA) {
1499 		/*
1500 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1501 		 * prefix along with the actual address in RTA_DST.
1502 		 * Remove netmask to avoid unnecessary address masking.
1503 		 */
1504 		remove_netmask(info);
1505 	}
1506 
1507 	switch (info->rti_info[RTAX_DST]->sa_family) {
1508 #ifdef INET
1509 	case AF_INET:
1510 		error = cleanup_xaddrs_inet(info);
1511 		break;
1512 #endif
1513 #ifdef INET6
1514 	case AF_INET6:
1515 		error = cleanup_xaddrs_inet6(info);
1516 		break;
1517 #endif
1518 	}
1519 
1520 	return (error);
1521 }
1522 
1523 /*
1524  * Fill in @dmask with valid netmask leaving original @smask
1525  * intact. Mostly used with radix netmasks.
1526  */
1527 struct sockaddr *
1528 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1529     struct sockaddr_storage *dmask)
1530 {
1531 	if (dst == NULL || smask == NULL)
1532 		return (NULL);
1533 
1534 	memset(dmask, 0, dst->sa_len);
1535 	memcpy(dmask, smask, smask->sa_len);
1536 	dmask->ss_len = dst->sa_len;
1537 	dmask->ss_family = dst->sa_family;
1538 
1539 	return ((struct sockaddr *)dmask);
1540 }
1541 
1542 /*
1543  * Writes information related to @rtinfo object to newly-allocated mbuf.
1544  * Assumes MCLBYTES is enough to construct any message.
1545  * Used for OS notifications of vaious events (if/ifa announces,etc)
1546  *
1547  * Returns allocated mbuf or NULL on failure.
1548  */
1549 static struct mbuf *
1550 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1551 {
1552 	struct sockaddr_storage ss;
1553 	struct rt_msghdr *rtm;
1554 	struct mbuf *m;
1555 	int i;
1556 	struct sockaddr *sa;
1557 #ifdef INET6
1558 	struct sockaddr_in6 *sin6;
1559 #endif
1560 	int len, dlen;
1561 
1562 	switch (type) {
1563 	case RTM_DELADDR:
1564 	case RTM_NEWADDR:
1565 		len = sizeof(struct ifa_msghdr);
1566 		break;
1567 
1568 	case RTM_DELMADDR:
1569 	case RTM_NEWMADDR:
1570 		len = sizeof(struct ifma_msghdr);
1571 		break;
1572 
1573 	case RTM_IFINFO:
1574 		len = sizeof(struct if_msghdr);
1575 		break;
1576 
1577 	case RTM_IFANNOUNCE:
1578 	case RTM_IEEE80211:
1579 		len = sizeof(struct if_announcemsghdr);
1580 		break;
1581 
1582 	default:
1583 		len = sizeof(struct rt_msghdr);
1584 	}
1585 
1586 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1587 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1588 	if (len > MHLEN)
1589 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1590 	else
1591 		m = m_gethdr(M_NOWAIT, MT_DATA);
1592 	if (m == NULL)
1593 		return (m);
1594 
1595 	m->m_pkthdr.len = m->m_len = len;
1596 	rtm = mtod(m, struct rt_msghdr *);
1597 	bzero((caddr_t)rtm, len);
1598 	for (i = 0; i < RTAX_MAX; i++) {
1599 		if ((sa = rtinfo->rti_info[i]) == NULL)
1600 			continue;
1601 		rtinfo->rti_addrs |= (1 << i);
1602 
1603 		dlen = SA_SIZE(sa);
1604 		KASSERT(dlen <= sizeof(ss),
1605 		    ("%s: sockaddr size overflow", __func__));
1606 		bzero(&ss, sizeof(ss));
1607 		bcopy(sa, &ss, sa->sa_len);
1608 		sa = (struct sockaddr *)&ss;
1609 #ifdef INET6
1610 		if (sa->sa_family == AF_INET6) {
1611 			sin6 = (struct sockaddr_in6 *)sa;
1612 			(void)sa6_recoverscope(sin6);
1613 		}
1614 #endif
1615 		m_copyback(m, len, dlen, (caddr_t)sa);
1616 		len += dlen;
1617 	}
1618 	if (m->m_pkthdr.len != len) {
1619 		m_freem(m);
1620 		return (NULL);
1621 	}
1622 	rtm->rtm_msglen = len;
1623 	rtm->rtm_version = RTM_VERSION;
1624 	rtm->rtm_type = type;
1625 	return (m);
1626 }
1627 
1628 /*
1629  * Writes information related to @rtinfo object to preallocated buffer.
1630  * Stores needed size in @plen. If @w is NULL, calculates size without
1631  * writing.
1632  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1633  *
1634  * Returns 0 on success.
1635  *
1636  */
1637 static int
1638 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1639 {
1640 	struct sockaddr_storage ss;
1641 	int len, buflen = 0, dlen, i;
1642 	caddr_t cp = NULL;
1643 	struct rt_msghdr *rtm = NULL;
1644 #ifdef INET6
1645 	struct sockaddr_in6 *sin6;
1646 #endif
1647 #ifdef COMPAT_FREEBSD32
1648 	bool compat32 = false;
1649 #endif
1650 
1651 	switch (type) {
1652 	case RTM_DELADDR:
1653 	case RTM_NEWADDR:
1654 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1655 #ifdef COMPAT_FREEBSD32
1656 			if (w->w_req->flags & SCTL_MASK32) {
1657 				len = sizeof(struct ifa_msghdrl32);
1658 				compat32 = true;
1659 			} else
1660 #endif
1661 				len = sizeof(struct ifa_msghdrl);
1662 		} else
1663 			len = sizeof(struct ifa_msghdr);
1664 		break;
1665 
1666 	case RTM_IFINFO:
1667 #ifdef COMPAT_FREEBSD32
1668 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1669 			if (w->w_op == NET_RT_IFLISTL)
1670 				len = sizeof(struct if_msghdrl32);
1671 			else
1672 				len = sizeof(struct if_msghdr32);
1673 			compat32 = true;
1674 			break;
1675 		}
1676 #endif
1677 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1678 			len = sizeof(struct if_msghdrl);
1679 		else
1680 			len = sizeof(struct if_msghdr);
1681 		break;
1682 
1683 	case RTM_NEWMADDR:
1684 		len = sizeof(struct ifma_msghdr);
1685 		break;
1686 
1687 	default:
1688 		len = sizeof(struct rt_msghdr);
1689 	}
1690 
1691 	if (w != NULL) {
1692 		rtm = (struct rt_msghdr *)w->w_tmem;
1693 		buflen = w->w_tmemsize - len;
1694 		cp = (caddr_t)w->w_tmem + len;
1695 	}
1696 
1697 	rtinfo->rti_addrs = 0;
1698 	for (i = 0; i < RTAX_MAX; i++) {
1699 		struct sockaddr *sa;
1700 
1701 		if ((sa = rtinfo->rti_info[i]) == NULL)
1702 			continue;
1703 		rtinfo->rti_addrs |= (1 << i);
1704 #ifdef COMPAT_FREEBSD32
1705 		if (compat32)
1706 			dlen = SA_SIZE32(sa);
1707 		else
1708 #endif
1709 			dlen = SA_SIZE(sa);
1710 		if (cp != NULL && buflen >= dlen) {
1711 			KASSERT(dlen <= sizeof(ss),
1712 			    ("%s: sockaddr size overflow", __func__));
1713 			bzero(&ss, sizeof(ss));
1714 			bcopy(sa, &ss, sa->sa_len);
1715 			sa = (struct sockaddr *)&ss;
1716 #ifdef INET6
1717 			if (sa->sa_family == AF_INET6) {
1718 				sin6 = (struct sockaddr_in6 *)sa;
1719 				(void)sa6_recoverscope(sin6);
1720 			}
1721 #endif
1722 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1723 			cp += dlen;
1724 			buflen -= dlen;
1725 		} else if (cp != NULL) {
1726 			/*
1727 			 * Buffer too small. Count needed size
1728 			 * and return with error.
1729 			 */
1730 			cp = NULL;
1731 		}
1732 
1733 		len += dlen;
1734 	}
1735 
1736 	if (cp != NULL) {
1737 		dlen = ALIGN(len) - len;
1738 		if (buflen < dlen)
1739 			cp = NULL;
1740 		else {
1741 			bzero(cp, dlen);
1742 			cp += dlen;
1743 			buflen -= dlen;
1744 		}
1745 	}
1746 	len = ALIGN(len);
1747 
1748 	if (cp != NULL) {
1749 		/* fill header iff buffer is large enough */
1750 		rtm->rtm_version = RTM_VERSION;
1751 		rtm->rtm_type = type;
1752 		rtm->rtm_msglen = len;
1753 	}
1754 
1755 	*plen = len;
1756 
1757 	if (w != NULL && cp == NULL)
1758 		return (ENOBUFS);
1759 
1760 	return (0);
1761 }
1762 
1763 /*
1764  * This routine is called to generate a message from the routing
1765  * socket indicating that a redirect has occurred, a routing lookup
1766  * has failed, or that a protocol has detected timeouts to a particular
1767  * destination.
1768  */
1769 void
1770 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1771     int fibnum)
1772 {
1773 	struct rt_msghdr *rtm;
1774 	struct mbuf *m;
1775 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1776 
1777 	if (V_route_cb.any_count == 0)
1778 		return;
1779 	m = rtsock_msg_mbuf(type, rtinfo);
1780 	if (m == NULL)
1781 		return;
1782 
1783 	if (fibnum != RT_ALL_FIBS) {
1784 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1785 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1786 		M_SETFIB(m, fibnum);
1787 		m->m_flags |= RTS_FILTER_FIB;
1788 	}
1789 
1790 	rtm = mtod(m, struct rt_msghdr *);
1791 	rtm->rtm_flags = RTF_DONE | flags;
1792 	rtm->rtm_errno = error;
1793 	rtm->rtm_addrs = rtinfo->rti_addrs;
1794 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1795 }
1796 
1797 void
1798 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1799 {
1800 
1801 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1802 }
1803 
1804 /*
1805  * This routine is called to generate a message from the routing
1806  * socket indicating that the status of a network interface has changed.
1807  */
1808 void
1809 rt_ifmsg(struct ifnet *ifp)
1810 {
1811 	struct if_msghdr *ifm;
1812 	struct mbuf *m;
1813 	struct rt_addrinfo info;
1814 
1815 	if (V_route_cb.any_count == 0)
1816 		return;
1817 	bzero((caddr_t)&info, sizeof(info));
1818 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1819 	if (m == NULL)
1820 		return;
1821 	ifm = mtod(m, struct if_msghdr *);
1822 	ifm->ifm_index = ifp->if_index;
1823 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1824 	if_data_copy(ifp, &ifm->ifm_data);
1825 	ifm->ifm_addrs = 0;
1826 	rt_dispatch(m, AF_UNSPEC);
1827 }
1828 
1829 /*
1830  * Announce interface address arrival/withdraw.
1831  * Please do not call directly, use rt_addrmsg().
1832  * Assume input data to be valid.
1833  * Returns 0 on success.
1834  */
1835 int
1836 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1837 {
1838 	struct rt_addrinfo info;
1839 	struct sockaddr *sa;
1840 	int ncmd;
1841 	struct mbuf *m;
1842 	struct ifa_msghdr *ifam;
1843 	struct ifnet *ifp = ifa->ifa_ifp;
1844 	struct sockaddr_storage ss;
1845 
1846 	if (V_route_cb.any_count == 0)
1847 		return (0);
1848 
1849 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1850 
1851 	bzero((caddr_t)&info, sizeof(info));
1852 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1853 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1854 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1855 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1856 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1857 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1858 		return (ENOBUFS);
1859 	ifam = mtod(m, struct ifa_msghdr *);
1860 	ifam->ifam_index = ifp->if_index;
1861 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1862 	ifam->ifam_flags = ifa->ifa_flags;
1863 	ifam->ifam_addrs = info.rti_addrs;
1864 
1865 	if (fibnum != RT_ALL_FIBS) {
1866 		M_SETFIB(m, fibnum);
1867 		m->m_flags |= RTS_FILTER_FIB;
1868 	}
1869 
1870 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1871 
1872 	return (0);
1873 }
1874 
1875 /*
1876  * Announce route addition/removal to rtsock based on @rt data.
1877  * Callers are advives to use rt_routemsg() instead of using this
1878  *  function directly.
1879  * Assume @rt data is consistent.
1880  *
1881  * Returns 0 on success.
1882  */
1883 int
1884 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
1885     int fibnum)
1886 {
1887 	union sockaddr_union dst, mask;
1888 	struct rt_addrinfo info;
1889 
1890 	if (V_route_cb.any_count == 0)
1891 		return (0);
1892 
1893 	int family = rt_get_family(rt);
1894 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
1895 	export_rtaddrs(rt, &dst.sa, &mask.sa);
1896 
1897 	bzero((caddr_t)&info, sizeof(info));
1898 	info.rti_info[RTAX_DST] = &dst.sa;
1899 	info.rti_info[RTAX_NETMASK] = &mask.sa;
1900 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1901 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
1902 	info.rti_ifp = nh->nh_ifp;
1903 
1904 	return (rtsock_routemsg_info(cmd, &info, fibnum));
1905 }
1906 
1907 int
1908 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1909 {
1910 	struct rt_msghdr *rtm;
1911 	struct sockaddr *sa;
1912 	struct mbuf *m;
1913 
1914 	if (V_route_cb.any_count == 0)
1915 		return (0);
1916 
1917 	if (info->rti_flags & RTF_HOST)
1918 		info->rti_info[RTAX_NETMASK] = NULL;
1919 
1920 	m = rtsock_msg_mbuf(cmd, info);
1921 	if (m == NULL)
1922 		return (ENOBUFS);
1923 
1924 	if (fibnum != RT_ALL_FIBS) {
1925 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1926 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1927 		M_SETFIB(m, fibnum);
1928 		m->m_flags |= RTS_FILTER_FIB;
1929 	}
1930 
1931 	rtm = mtod(m, struct rt_msghdr *);
1932 	rtm->rtm_addrs = info->rti_addrs;
1933 	if (info->rti_ifp != NULL)
1934 		rtm->rtm_index = info->rti_ifp->if_index;
1935 	/* Add RTF_DONE to indicate command 'completion' required by API */
1936 	info->rti_flags |= RTF_DONE;
1937 	/* Reported routes has to be up */
1938 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1939 		info->rti_flags |= RTF_UP;
1940 	rtm->rtm_flags = info->rti_flags;
1941 
1942 	sa = info->rti_info[RTAX_DST];
1943 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1944 
1945 	return (0);
1946 }
1947 
1948 /*
1949  * This is the analogue to the rt_newaddrmsg which performs the same
1950  * function but for multicast group memberhips.  This is easier since
1951  * there is no route state to worry about.
1952  */
1953 void
1954 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1955 {
1956 	struct rt_addrinfo info;
1957 	struct mbuf *m = NULL;
1958 	struct ifnet *ifp = ifma->ifma_ifp;
1959 	struct ifma_msghdr *ifmam;
1960 
1961 	if (V_route_cb.any_count == 0)
1962 		return;
1963 
1964 	bzero((caddr_t)&info, sizeof(info));
1965 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1966 	if (ifp && ifp->if_addr)
1967 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1968 	else
1969 		info.rti_info[RTAX_IFP] = NULL;
1970 	/*
1971 	 * If a link-layer address is present, present it as a ``gateway''
1972 	 * (similarly to how ARP entries, e.g., are presented).
1973 	 */
1974 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1975 	m = rtsock_msg_mbuf(cmd, &info);
1976 	if (m == NULL)
1977 		return;
1978 	ifmam = mtod(m, struct ifma_msghdr *);
1979 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1980 	    __func__));
1981 	ifmam->ifmam_index = ifp->if_index;
1982 	ifmam->ifmam_addrs = info.rti_addrs;
1983 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1984 }
1985 
1986 static struct mbuf *
1987 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1988 	struct rt_addrinfo *info)
1989 {
1990 	struct if_announcemsghdr *ifan;
1991 	struct mbuf *m;
1992 
1993 	if (V_route_cb.any_count == 0)
1994 		return NULL;
1995 	bzero((caddr_t)info, sizeof(*info));
1996 	m = rtsock_msg_mbuf(type, info);
1997 	if (m != NULL) {
1998 		ifan = mtod(m, struct if_announcemsghdr *);
1999 		ifan->ifan_index = ifp->if_index;
2000 		strlcpy(ifan->ifan_name, ifp->if_xname,
2001 			sizeof(ifan->ifan_name));
2002 		ifan->ifan_what = what;
2003 	}
2004 	return m;
2005 }
2006 
2007 /*
2008  * This is called to generate routing socket messages indicating
2009  * IEEE80211 wireless events.
2010  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2011  */
2012 void
2013 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2014 {
2015 	struct mbuf *m;
2016 	struct rt_addrinfo info;
2017 
2018 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2019 	if (m != NULL) {
2020 		/*
2021 		 * Append the ieee80211 data.  Try to stick it in the
2022 		 * mbuf containing the ifannounce msg; otherwise allocate
2023 		 * a new mbuf and append.
2024 		 *
2025 		 * NB: we assume m is a single mbuf.
2026 		 */
2027 		if (data_len > M_TRAILINGSPACE(m)) {
2028 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2029 			if (n == NULL) {
2030 				m_freem(m);
2031 				return;
2032 			}
2033 			bcopy(data, mtod(n, void *), data_len);
2034 			n->m_len = data_len;
2035 			m->m_next = n;
2036 		} else if (data_len > 0) {
2037 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2038 			m->m_len += data_len;
2039 		}
2040 		if (m->m_flags & M_PKTHDR)
2041 			m->m_pkthdr.len += data_len;
2042 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2043 		rt_dispatch(m, AF_UNSPEC);
2044 	}
2045 }
2046 
2047 /*
2048  * This is called to generate routing socket messages indicating
2049  * network interface arrival and departure.
2050  */
2051 void
2052 rt_ifannouncemsg(struct ifnet *ifp, int what)
2053 {
2054 	struct mbuf *m;
2055 	struct rt_addrinfo info;
2056 
2057 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2058 	if (m != NULL)
2059 		rt_dispatch(m, AF_UNSPEC);
2060 }
2061 
2062 static void
2063 rt_dispatch(struct mbuf *m, sa_family_t saf)
2064 {
2065 	struct m_tag *tag;
2066 
2067 	/*
2068 	 * Preserve the family from the sockaddr, if any, in an m_tag for
2069 	 * use when injecting the mbuf into the routing socket buffer from
2070 	 * the netisr.
2071 	 */
2072 	if (saf != AF_UNSPEC) {
2073 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
2074 		    M_NOWAIT);
2075 		if (tag == NULL) {
2076 			m_freem(m);
2077 			return;
2078 		}
2079 		*(unsigned short *)(tag + 1) = saf;
2080 		m_tag_prepend(m, tag);
2081 	}
2082 #ifdef VIMAGE
2083 	if (V_loif)
2084 		m->m_pkthdr.rcvif = V_loif;
2085 	else {
2086 		m_freem(m);
2087 		return;
2088 	}
2089 #endif
2090 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2091 }
2092 
2093 /*
2094  * Checks if rte can be exported v.r.t jails/vnets.
2095  *
2096  * Returns 1 if it can, 0 otherwise.
2097  */
2098 static bool
2099 can_export_rte(struct ucred *td_ucred, bool rt_is_host,
2100     const struct sockaddr *rt_dst)
2101 {
2102 
2103 	if ((!rt_is_host) ? jailed_without_vnet(td_ucred)
2104 	    : prison_if(td_ucred, rt_dst) != 0)
2105 		return (false);
2106 	return (true);
2107 }
2108 
2109 
2110 /*
2111  * This is used in dumping the kernel table via sysctl().
2112  */
2113 static int
2114 sysctl_dumpentry(struct rtentry *rt, void *vw)
2115 {
2116 	struct walkarg *w = vw;
2117 	struct nhop_object *nh;
2118 	int error = 0;
2119 
2120 	NET_EPOCH_ASSERT();
2121 
2122 	export_rtaddrs(rt, w->dst, w->mask);
2123 	if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst))
2124 		return (0);
2125 	nh = rt_get_raw_nhop(rt);
2126 #ifdef ROUTE_MPATH
2127 	if (NH_IS_NHGRP(nh)) {
2128 		struct weightened_nhop *wn;
2129 		uint32_t num_nhops;
2130 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2131 		for (int i = 0; i < num_nhops; i++) {
2132 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2133 			if (error != 0)
2134 				return (error);
2135 		}
2136 	} else
2137 #endif
2138 		error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2139 
2140 	return (0);
2141 }
2142 
2143 
2144 static int
2145 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2146     struct walkarg *w)
2147 {
2148 	struct rt_addrinfo info;
2149 	int error = 0, size;
2150 	uint32_t rtflags;
2151 
2152 	rtflags = nhop_get_rtflags(nh);
2153 
2154 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2155 		return (0);
2156 
2157 	bzero((caddr_t)&info, sizeof(info));
2158 	info.rti_info[RTAX_DST] = w->dst;
2159 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2160 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2161 	info.rti_info[RTAX_GENMASK] = 0;
2162 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2163 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2164 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2165 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2166 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2167 	}
2168 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2169 		return (error);
2170 	if (w->w_req && w->w_tmem) {
2171 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2172 
2173 		bzero(&rtm->rtm_index,
2174 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2175 
2176 		/*
2177 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2178 		 * and RTF_UP (if entry is linked, which is always true here).
2179 		 * Given that, use nhop rtflags & add RTF_UP.
2180 		 */
2181 		rtm->rtm_flags = rtflags | RTF_UP;
2182 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2183 			rtm->rtm_flags = RTF_GATEWAY |
2184 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2185 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2186 		rtm->rtm_rmx.rmx_weight = weight;
2187 		rtm->rtm_index = nh->nh_ifp->if_index;
2188 		rtm->rtm_addrs = info.rti_addrs;
2189 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2190 		return (error);
2191 	}
2192 	return (error);
2193 }
2194 
2195 static int
2196 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2197     struct rt_addrinfo *info, struct walkarg *w, int len)
2198 {
2199 	struct if_msghdrl *ifm;
2200 	struct if_data *ifd;
2201 
2202 	ifm = (struct if_msghdrl *)w->w_tmem;
2203 
2204 #ifdef COMPAT_FREEBSD32
2205 	if (w->w_req->flags & SCTL_MASK32) {
2206 		struct if_msghdrl32 *ifm32;
2207 
2208 		ifm32 = (struct if_msghdrl32 *)ifm;
2209 		ifm32->ifm_addrs = info->rti_addrs;
2210 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2211 		ifm32->ifm_index = ifp->if_index;
2212 		ifm32->_ifm_spare1 = 0;
2213 		ifm32->ifm_len = sizeof(*ifm32);
2214 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2215 		ifm32->_ifm_spare2 = 0;
2216 		ifd = &ifm32->ifm_data;
2217 	} else
2218 #endif
2219 	{
2220 		ifm->ifm_addrs = info->rti_addrs;
2221 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2222 		ifm->ifm_index = ifp->if_index;
2223 		ifm->_ifm_spare1 = 0;
2224 		ifm->ifm_len = sizeof(*ifm);
2225 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2226 		ifm->_ifm_spare2 = 0;
2227 		ifd = &ifm->ifm_data;
2228 	}
2229 
2230 	memcpy(ifd, src_ifd, sizeof(*ifd));
2231 
2232 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2233 }
2234 
2235 static int
2236 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2237     struct rt_addrinfo *info, struct walkarg *w, int len)
2238 {
2239 	struct if_msghdr *ifm;
2240 	struct if_data *ifd;
2241 
2242 	ifm = (struct if_msghdr *)w->w_tmem;
2243 
2244 #ifdef COMPAT_FREEBSD32
2245 	if (w->w_req->flags & SCTL_MASK32) {
2246 		struct if_msghdr32 *ifm32;
2247 
2248 		ifm32 = (struct if_msghdr32 *)ifm;
2249 		ifm32->ifm_addrs = info->rti_addrs;
2250 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2251 		ifm32->ifm_index = ifp->if_index;
2252 		ifm32->_ifm_spare1 = 0;
2253 		ifd = &ifm32->ifm_data;
2254 	} else
2255 #endif
2256 	{
2257 		ifm->ifm_addrs = info->rti_addrs;
2258 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2259 		ifm->ifm_index = ifp->if_index;
2260 		ifm->_ifm_spare1 = 0;
2261 		ifd = &ifm->ifm_data;
2262 	}
2263 
2264 	memcpy(ifd, src_ifd, sizeof(*ifd));
2265 
2266 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2267 }
2268 
2269 static int
2270 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2271     struct walkarg *w, int len)
2272 {
2273 	struct ifa_msghdrl *ifam;
2274 	struct if_data *ifd;
2275 
2276 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2277 
2278 #ifdef COMPAT_FREEBSD32
2279 	if (w->w_req->flags & SCTL_MASK32) {
2280 		struct ifa_msghdrl32 *ifam32;
2281 
2282 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2283 		ifam32->ifam_addrs = info->rti_addrs;
2284 		ifam32->ifam_flags = ifa->ifa_flags;
2285 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2286 		ifam32->_ifam_spare1 = 0;
2287 		ifam32->ifam_len = sizeof(*ifam32);
2288 		ifam32->ifam_data_off =
2289 		    offsetof(struct ifa_msghdrl32, ifam_data);
2290 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2291 		ifd = &ifam32->ifam_data;
2292 	} else
2293 #endif
2294 	{
2295 		ifam->ifam_addrs = info->rti_addrs;
2296 		ifam->ifam_flags = ifa->ifa_flags;
2297 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2298 		ifam->_ifam_spare1 = 0;
2299 		ifam->ifam_len = sizeof(*ifam);
2300 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2301 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2302 		ifd = &ifam->ifam_data;
2303 	}
2304 
2305 	bzero(ifd, sizeof(*ifd));
2306 	ifd->ifi_datalen = sizeof(struct if_data);
2307 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2308 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2309 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2310 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2311 
2312 	/* Fixup if_data carp(4) vhid. */
2313 	if (carp_get_vhid_p != NULL)
2314 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2315 
2316 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2317 }
2318 
2319 static int
2320 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2321     struct walkarg *w, int len)
2322 {
2323 	struct ifa_msghdr *ifam;
2324 
2325 	ifam = (struct ifa_msghdr *)w->w_tmem;
2326 	ifam->ifam_addrs = info->rti_addrs;
2327 	ifam->ifam_flags = ifa->ifa_flags;
2328 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2329 	ifam->_ifam_spare1 = 0;
2330 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2331 
2332 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2333 }
2334 
2335 static int
2336 sysctl_iflist(int af, struct walkarg *w)
2337 {
2338 	struct ifnet *ifp;
2339 	struct ifaddr *ifa;
2340 	struct if_data ifd;
2341 	struct rt_addrinfo info;
2342 	int len, error = 0;
2343 	struct sockaddr_storage ss;
2344 
2345 	bzero((caddr_t)&info, sizeof(info));
2346 	bzero(&ifd, sizeof(ifd));
2347 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2348 		if (w->w_arg && w->w_arg != ifp->if_index)
2349 			continue;
2350 		if_data_copy(ifp, &ifd);
2351 		ifa = ifp->if_addr;
2352 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2353 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2354 		if (error != 0)
2355 			goto done;
2356 		info.rti_info[RTAX_IFP] = NULL;
2357 		if (w->w_req && w->w_tmem) {
2358 			if (w->w_op == NET_RT_IFLISTL)
2359 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2360 				    len);
2361 			else
2362 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2363 				    len);
2364 			if (error)
2365 				goto done;
2366 		}
2367 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2368 			if (af && af != ifa->ifa_addr->sa_family)
2369 				continue;
2370 			if (prison_if(w->w_req->td->td_ucred,
2371 			    ifa->ifa_addr) != 0)
2372 				continue;
2373 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2374 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2375 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2376 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2377 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2378 			if (error != 0)
2379 				goto done;
2380 			if (w->w_req && w->w_tmem) {
2381 				if (w->w_op == NET_RT_IFLISTL)
2382 					error = sysctl_iflist_ifaml(ifa, &info,
2383 					    w, len);
2384 				else
2385 					error = sysctl_iflist_ifam(ifa, &info,
2386 					    w, len);
2387 				if (error)
2388 					goto done;
2389 			}
2390 		}
2391 		info.rti_info[RTAX_IFA] = NULL;
2392 		info.rti_info[RTAX_NETMASK] = NULL;
2393 		info.rti_info[RTAX_BRD] = NULL;
2394 	}
2395 done:
2396 	return (error);
2397 }
2398 
2399 static int
2400 sysctl_ifmalist(int af, struct walkarg *w)
2401 {
2402 	struct rt_addrinfo info;
2403 	struct ifaddr *ifa;
2404 	struct ifmultiaddr *ifma;
2405 	struct ifnet *ifp;
2406 	int error, len;
2407 
2408 	NET_EPOCH_ASSERT();
2409 
2410 	error = 0;
2411 	bzero((caddr_t)&info, sizeof(info));
2412 
2413 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2414 		if (w->w_arg && w->w_arg != ifp->if_index)
2415 			continue;
2416 		ifa = ifp->if_addr;
2417 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2418 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2419 			if (af && af != ifma->ifma_addr->sa_family)
2420 				continue;
2421 			if (prison_if(w->w_req->td->td_ucred,
2422 			    ifma->ifma_addr) != 0)
2423 				continue;
2424 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2425 			info.rti_info[RTAX_GATEWAY] =
2426 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2427 			    ifma->ifma_lladdr : NULL;
2428 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2429 			if (error != 0)
2430 				break;
2431 			if (w->w_req && w->w_tmem) {
2432 				struct ifma_msghdr *ifmam;
2433 
2434 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2435 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2436 				ifmam->ifmam_flags = 0;
2437 				ifmam->ifmam_addrs = info.rti_addrs;
2438 				ifmam->_ifmam_spare1 = 0;
2439 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2440 				if (error != 0)
2441 					break;
2442 			}
2443 		}
2444 		if (error != 0)
2445 			break;
2446 	}
2447 	return (error);
2448 }
2449 
2450 static void
2451 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2452 {
2453 	union sockaddr_union sa_dst, sa_mask;
2454 
2455 	w->family = family;
2456 	w->dst = (struct sockaddr *)&sa_dst;
2457 	w->mask = (struct sockaddr *)&sa_mask;
2458 
2459 	init_sockaddrs_family(family, w->dst, w->mask);
2460 
2461 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2462 }
2463 
2464 static int
2465 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2466 {
2467 	struct epoch_tracker et;
2468 	int	*name = (int *)arg1;
2469 	u_int	namelen = arg2;
2470 	struct rib_head *rnh = NULL; /* silence compiler. */
2471 	int	i, lim, error = EINVAL;
2472 	int	fib = 0;
2473 	u_char	af;
2474 	struct	walkarg w;
2475 
2476 	name ++;
2477 	namelen--;
2478 	if (req->newptr)
2479 		return (EPERM);
2480 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2481 		if (namelen == 3)
2482 			fib = req->td->td_proc->p_fibnum;
2483 		else if (namelen == 4)
2484 			fib = (name[3] == RT_ALL_FIBS) ?
2485 			    req->td->td_proc->p_fibnum : name[3];
2486 		else
2487 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2488 		if (fib < 0 || fib >= rt_numfibs)
2489 			return (EINVAL);
2490 	} else if (namelen != 3)
2491 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2492 	af = name[0];
2493 	if (af > AF_MAX)
2494 		return (EINVAL);
2495 	bzero(&w, sizeof(w));
2496 	w.w_op = name[1];
2497 	w.w_arg = name[2];
2498 	w.w_req = req;
2499 
2500 	error = sysctl_wire_old_buffer(req, 0);
2501 	if (error)
2502 		return (error);
2503 
2504 	/*
2505 	 * Allocate reply buffer in advance.
2506 	 * All rtsock messages has maximum length of u_short.
2507 	 */
2508 	w.w_tmemsize = 65536;
2509 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2510 
2511 	NET_EPOCH_ENTER(et);
2512 	switch (w.w_op) {
2513 	case NET_RT_DUMP:
2514 	case NET_RT_FLAGS:
2515 		if (af == 0) {			/* dump all tables */
2516 			i = 1;
2517 			lim = AF_MAX;
2518 		} else				/* dump only one table */
2519 			i = lim = af;
2520 
2521 		/*
2522 		 * take care of llinfo entries, the caller must
2523 		 * specify an AF
2524 		 */
2525 		if (w.w_op == NET_RT_FLAGS &&
2526 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2527 			if (af != 0)
2528 				error = lltable_sysctl_dumparp(af, w.w_req);
2529 			else
2530 				error = EINVAL;
2531 			break;
2532 		}
2533 		/*
2534 		 * take care of routing entries
2535 		 */
2536 		for (error = 0; error == 0 && i <= lim; i++) {
2537 			rnh = rt_tables_get_rnh(fib, i);
2538 			if (rnh != NULL) {
2539 				rtable_sysctl_dump(fib, i, &w);
2540 			} else if (af != 0)
2541 				error = EAFNOSUPPORT;
2542 		}
2543 		break;
2544 	case NET_RT_NHOP:
2545 	case NET_RT_NHGRP:
2546 		/* Allow dumping one specific af/fib at a time */
2547 		if (namelen < 4) {
2548 			error = EINVAL;
2549 			break;
2550 		}
2551 		fib = name[3];
2552 		if (fib < 0 || fib > rt_numfibs) {
2553 			error = EINVAL;
2554 			break;
2555 		}
2556 		rnh = rt_tables_get_rnh(fib, af);
2557 		if (rnh == NULL) {
2558 			error = EAFNOSUPPORT;
2559 			break;
2560 		}
2561 		if (w.w_op == NET_RT_NHOP)
2562 			error = nhops_dump_sysctl(rnh, w.w_req);
2563 		else
2564 #ifdef ROUTE_MPATH
2565 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2566 #else
2567 			error = ENOTSUP;
2568 #endif
2569 		break;
2570 	case NET_RT_IFLIST:
2571 	case NET_RT_IFLISTL:
2572 		error = sysctl_iflist(af, &w);
2573 		break;
2574 
2575 	case NET_RT_IFMALIST:
2576 		error = sysctl_ifmalist(af, &w);
2577 		break;
2578 	}
2579 	NET_EPOCH_EXIT(et);
2580 
2581 	free(w.w_tmem, M_TEMP);
2582 	return (error);
2583 }
2584 
2585 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2586     sysctl_rtsock, "Return route tables and interface/address lists");
2587 
2588 /*
2589  * Definitions of protocols supported in the ROUTE domain.
2590  */
2591 
2592 static struct domain routedomain;		/* or at least forward */
2593 
2594 static struct protosw routesw[] = {
2595 {
2596 	.pr_type =		SOCK_RAW,
2597 	.pr_domain =		&routedomain,
2598 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2599 	.pr_output =		route_output,
2600 	.pr_ctlinput =		raw_ctlinput,
2601 	.pr_init =		raw_init,
2602 	.pr_usrreqs =		&route_usrreqs
2603 }
2604 };
2605 
2606 static struct domain routedomain = {
2607 	.dom_family =		PF_ROUTE,
2608 	.dom_name =		"route",
2609 	.dom_protosw =		routesw,
2610 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
2611 };
2612 
2613 VNET_DOMAIN_SET(route);
2614