xref: /freebsd/sys/net/rtsock.c (revision 5b5b7e2ca2fa9a2418dd51749f4ef6f881ae7179)
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/eventhandler.h>
43 #include <sys/domain.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/protosw.h>
50 #include <sys/rmlock.h>
51 #include <sys/rwlock.h>
52 #include <sys/signalvar.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <sys/systm.h>
57 
58 #include <net/if.h>
59 #include <net/if_var.h>
60 #include <net/if_dl.h>
61 #include <net/if_llatbl.h>
62 #include <net/if_types.h>
63 #include <net/netisr.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 #define	DEBUG_MOD_NAME	rtsock
80 #define	DEBUG_MAX_LEVEL	LOG_DEBUG
81 #include <net/route/route_debug.h>
82 _DECLARE_DEBUG(LOG_INFO);
83 
84 #ifdef COMPAT_FREEBSD32
85 #include <sys/mount.h>
86 #include <compat/freebsd32/freebsd32.h>
87 
88 struct if_msghdr32 {
89 	uint16_t ifm_msglen;
90 	uint8_t	ifm_version;
91 	uint8_t	ifm_type;
92 	int32_t	ifm_addrs;
93 	int32_t	ifm_flags;
94 	uint16_t ifm_index;
95 	uint16_t _ifm_spare1;
96 	struct	if_data ifm_data;
97 };
98 
99 struct if_msghdrl32 {
100 	uint16_t ifm_msglen;
101 	uint8_t	ifm_version;
102 	uint8_t	ifm_type;
103 	int32_t	ifm_addrs;
104 	int32_t	ifm_flags;
105 	uint16_t ifm_index;
106 	uint16_t _ifm_spare1;
107 	uint16_t ifm_len;
108 	uint16_t ifm_data_off;
109 	uint32_t _ifm_spare2;
110 	struct	if_data ifm_data;
111 };
112 
113 struct ifa_msghdrl32 {
114 	uint16_t ifam_msglen;
115 	uint8_t	ifam_version;
116 	uint8_t	ifam_type;
117 	int32_t	ifam_addrs;
118 	int32_t	ifam_flags;
119 	uint16_t ifam_index;
120 	uint16_t _ifam_spare1;
121 	uint16_t ifam_len;
122 	uint16_t ifam_data_off;
123 	int32_t	ifam_metric;
124 	struct	if_data ifam_data;
125 };
126 
127 #define SA_SIZE32(sa)						\
128     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
129 	sizeof(int)		:				\
130 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
131 
132 #endif /* COMPAT_FREEBSD32 */
133 
134 struct linear_buffer {
135 	char		*base;	/* Base allocated memory pointer */
136 	uint32_t	offset;	/* Currently used offset */
137 	uint32_t	size;	/* Total buffer size */
138 };
139 #define	SCRATCH_BUFFER_SIZE	1024
140 
141 #define	RTS_PID_LOG(_l, _fmt, ...)	RT_LOG_##_l(_l, "PID %d: " _fmt, curproc ? curproc->p_pid : 0, ## __VA_ARGS__)
142 
143 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
144 
145 /* NB: these are not modified */
146 static struct	sockaddr route_src = { 2, PF_ROUTE, };
147 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
148 
149 /* These are external hooks for CARP. */
150 int	(*carp_get_vhid_p)(struct ifaddr *);
151 
152 /*
153  * Used by rtsock callback code to decide whether to filter the update
154  * notification to a socket bound to a particular FIB.
155  */
156 #define	RTS_FILTER_FIB	M_PROTO8
157 /*
158  * Used to store address family of the notification.
159  */
160 #define	m_rtsock_family	m_pkthdr.PH_loc.eight[0]
161 
162 struct rcb {
163 	LIST_ENTRY(rcb) list;
164 	struct socket	*rcb_socket;
165 	sa_family_t	rcb_family;
166 };
167 
168 typedef struct {
169 	LIST_HEAD(, rcb)	cblist;
170 	int	ip_count;	/* attached w/ AF_INET */
171 	int	ip6_count;	/* attached w/ AF_INET6 */
172 	int	any_count;	/* total attached */
173 } route_cb_t;
174 VNET_DEFINE_STATIC(route_cb_t, route_cb);
175 #define	V_route_cb VNET(route_cb)
176 
177 struct mtx rtsock_mtx;
178 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
179 
180 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
181 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
182 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
183 
184 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
185 
186 struct walkarg {
187 	int	family;
188 	int	w_tmemsize;
189 	int	w_op, w_arg;
190 	caddr_t	w_tmem;
191 	struct sysctl_req *w_req;
192 	struct sockaddr *dst;
193 	struct sockaddr *mask;
194 };
195 
196 static void	rts_input(struct mbuf *m);
197 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
198 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
199 			struct walkarg *w, int *plen);
200 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
201 			struct rt_addrinfo *rtinfo);
202 static int	cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb);
203 static int	sysctl_dumpentry(struct rtentry *rt, void *vw);
204 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
205 			uint32_t weight, struct walkarg *w);
206 static int	sysctl_iflist(int af, struct walkarg *w);
207 static int	sysctl_ifmalist(int af, struct walkarg *w);
208 static void	rt_getmetrics(const struct rtentry *rt,
209 			const struct nhop_object *nh, struct rt_metrics *out);
210 static void	rt_dispatch(struct mbuf *, sa_family_t);
211 static void	rt_ifannouncemsg(struct ifnet *ifp, int what);
212 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
213 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
214 static int	update_rtm_from_rc(struct rt_addrinfo *info,
215 			struct rt_msghdr **prtm, int alloc_len,
216 			struct rib_cmd_info *rc, struct nhop_object *nh);
217 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
218 			struct mbuf *m, sa_family_t saf, u_int fibnum,
219 			int rtm_errno);
220 static bool	can_export_rte(struct ucred *td_ucred, bool rt_is_host,
221 			const struct sockaddr *rt_dst);
222 
223 static struct netisr_handler rtsock_nh = {
224 	.nh_name = "rtsock",
225 	.nh_handler = rts_input,
226 	.nh_proto = NETISR_ROUTE,
227 	.nh_policy = NETISR_POLICY_SOURCE,
228 };
229 
230 static int
231 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
232 {
233 	int error, qlimit;
234 
235 	netisr_getqlimit(&rtsock_nh, &qlimit);
236 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
237         if (error || !req->newptr)
238                 return (error);
239 	if (qlimit < 1)
240 		return (EINVAL);
241 	return (netisr_setqlimit(&rtsock_nh, qlimit));
242 }
243 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
244     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
245     0, 0, sysctl_route_netisr_maxqlen, "I",
246     "maximum routing socket dispatch queue length");
247 
248 static void
249 vnet_rts_init(void)
250 {
251 	int tmp;
252 
253 	if (IS_DEFAULT_VNET(curvnet)) {
254 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
255 			rtsock_nh.nh_qlimit = tmp;
256 		netisr_register(&rtsock_nh);
257 	}
258 #ifdef VIMAGE
259 	 else
260 		netisr_register_vnet(&rtsock_nh);
261 #endif
262 }
263 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
264     vnet_rts_init, 0);
265 
266 #ifdef VIMAGE
267 static void
268 vnet_rts_uninit(void)
269 {
270 
271 	netisr_unregister_vnet(&rtsock_nh);
272 }
273 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
274     vnet_rts_uninit, 0);
275 #endif
276 
277 static void
278 rts_handle_ifnet_arrival(void *arg __unused, struct ifnet *ifp)
279 {
280 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
281 }
282 EVENTHANDLER_DEFINE(ifnet_arrival_event, rts_handle_ifnet_arrival, NULL, 0);
283 
284 static void
285 rts_handle_ifnet_departure(void *arg __unused, struct ifnet *ifp)
286 {
287 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
288 }
289 EVENTHANDLER_DEFINE(ifnet_departure_event, rts_handle_ifnet_departure, NULL, 0);
290 
291 static void
292 rts_append_data(struct socket *so, struct mbuf *m)
293 {
294 
295 	if (sbappendaddr(&so->so_rcv, &route_src, m, NULL) == 0) {
296 		soroverflow(so);
297 		m_freem(m);
298 	} else
299 		sorwakeup(so);
300 }
301 
302 static void
303 rts_input(struct mbuf *m)
304 {
305 	struct rcb *rcb;
306 	struct socket *last;
307 
308 	last = NULL;
309 	RTSOCK_LOCK();
310 	LIST_FOREACH(rcb, &V_route_cb.cblist, list) {
311 		if (rcb->rcb_family != AF_UNSPEC &&
312 		    rcb->rcb_family != m->m_rtsock_family)
313 			continue;
314 		if ((m->m_flags & RTS_FILTER_FIB) &&
315 		    M_GETFIB(m) != rcb->rcb_socket->so_fibnum)
316 			continue;
317 		if (last != NULL) {
318 			struct mbuf *n;
319 
320 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
321 			if (n != NULL)
322 				rts_append_data(last, n);
323 		}
324 		last = rcb->rcb_socket;
325 	}
326 	if (last != NULL)
327 		rts_append_data(last, m);
328 	else
329 		m_freem(m);
330 	RTSOCK_UNLOCK();
331 }
332 
333 static void
334 rts_close(struct socket *so)
335 {
336 
337 	soisdisconnected(so);
338 }
339 
340 static SYSCTL_NODE(_net, OID_AUTO, rtsock, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
341     "Routing socket infrastructure");
342 static u_long rts_sendspace = 8192;
343 SYSCTL_ULONG(_net_rtsock, OID_AUTO, sendspace, CTLFLAG_RW, &rts_sendspace, 0,
344     "Default routing socket send space");
345 static u_long rts_recvspace = 8192;
346 SYSCTL_ULONG(_net_rtsock, OID_AUTO, recvspace, CTLFLAG_RW, &rts_recvspace, 0,
347     "Default routing socket receive space");
348 
349 static int
350 rts_attach(struct socket *so, int proto, struct thread *td)
351 {
352 	struct rcb *rcb;
353 	int error;
354 
355 	error = soreserve(so, rts_sendspace, rts_recvspace);
356 	if (error)
357 		return (error);
358 
359 	rcb = malloc(sizeof(*rcb), M_PCB, M_WAITOK);
360 	rcb->rcb_socket = so;
361 	rcb->rcb_family = proto;
362 
363 	so->so_pcb = rcb;
364 	so->so_fibnum = td->td_proc->p_fibnum;
365 	so->so_options |= SO_USELOOPBACK;
366 
367 	RTSOCK_LOCK();
368 	LIST_INSERT_HEAD(&V_route_cb.cblist, rcb, list);
369 	switch (proto) {
370 	case AF_INET:
371 		V_route_cb.ip_count++;
372 		break;
373 	case AF_INET6:
374 		V_route_cb.ip6_count++;
375 		break;
376 	}
377 	V_route_cb.any_count++;
378 	RTSOCK_UNLOCK();
379 	soisconnected(so);
380 
381 	return (0);
382 }
383 
384 static void
385 rts_detach(struct socket *so)
386 {
387 	struct rcb *rcb = so->so_pcb;
388 
389 	RTSOCK_LOCK();
390 	LIST_REMOVE(rcb, list);
391 	switch(rcb->rcb_family) {
392 	case AF_INET:
393 		V_route_cb.ip_count--;
394 		break;
395 	case AF_INET6:
396 		V_route_cb.ip6_count--;
397 		break;
398 	}
399 	V_route_cb.any_count--;
400 	RTSOCK_UNLOCK();
401 	free(rcb, M_PCB);
402 	so->so_pcb = NULL;
403 }
404 
405 static int
406 rts_shutdown(struct socket *so)
407 {
408 
409 	socantsendmore(so);
410 	return (0);
411 }
412 
413 #ifndef _SOCKADDR_UNION_DEFINED
414 #define	_SOCKADDR_UNION_DEFINED
415 /*
416  * The union of all possible address formats we handle.
417  */
418 union sockaddr_union {
419 	struct sockaddr		sa;
420 	struct sockaddr_in	sin;
421 	struct sockaddr_in6	sin6;
422 };
423 #endif /* _SOCKADDR_UNION_DEFINED */
424 
425 static int
426 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
427     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
428 {
429 #if defined(INET) || defined(INET6)
430 	struct epoch_tracker et;
431 #endif
432 
433 	/* First, see if the returned address is part of the jail. */
434 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
435 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
436 		return (0);
437 	}
438 
439 	switch (info->rti_info[RTAX_DST]->sa_family) {
440 #ifdef INET
441 	case AF_INET:
442 	{
443 		struct in_addr ia;
444 		struct ifaddr *ifa;
445 		int found;
446 
447 		found = 0;
448 		/*
449 		 * Try to find an address on the given outgoing interface
450 		 * that belongs to the jail.
451 		 */
452 		NET_EPOCH_ENTER(et);
453 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
454 			struct sockaddr *sa;
455 			sa = ifa->ifa_addr;
456 			if (sa->sa_family != AF_INET)
457 				continue;
458 			ia = ((struct sockaddr_in *)sa)->sin_addr;
459 			if (prison_check_ip4(cred, &ia) == 0) {
460 				found = 1;
461 				break;
462 			}
463 		}
464 		NET_EPOCH_EXIT(et);
465 		if (!found) {
466 			/*
467 			 * As a last resort return the 'default' jail address.
468 			 */
469 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
470 			    sin_addr;
471 			if (prison_get_ip4(cred, &ia) != 0)
472 				return (ESRCH);
473 		}
474 		bzero(&saun->sin, sizeof(struct sockaddr_in));
475 		saun->sin.sin_len = sizeof(struct sockaddr_in);
476 		saun->sin.sin_family = AF_INET;
477 		saun->sin.sin_addr.s_addr = ia.s_addr;
478 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
479 		break;
480 	}
481 #endif
482 #ifdef INET6
483 	case AF_INET6:
484 	{
485 		struct in6_addr ia6;
486 		struct ifaddr *ifa;
487 		int found;
488 
489 		found = 0;
490 		/*
491 		 * Try to find an address on the given outgoing interface
492 		 * that belongs to the jail.
493 		 */
494 		NET_EPOCH_ENTER(et);
495 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
496 			struct sockaddr *sa;
497 			sa = ifa->ifa_addr;
498 			if (sa->sa_family != AF_INET6)
499 				continue;
500 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
501 			    &ia6, sizeof(struct in6_addr));
502 			if (prison_check_ip6(cred, &ia6) == 0) {
503 				found = 1;
504 				break;
505 			}
506 		}
507 		NET_EPOCH_EXIT(et);
508 		if (!found) {
509 			/*
510 			 * As a last resort return the 'default' jail address.
511 			 */
512 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
513 			    sin6_addr;
514 			if (prison_get_ip6(cred, &ia6) != 0)
515 				return (ESRCH);
516 		}
517 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
518 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
519 		saun->sin6.sin6_family = AF_INET6;
520 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
521 		if (sa6_recoverscope(&saun->sin6) != 0)
522 			return (ESRCH);
523 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
524 		break;
525 	}
526 #endif
527 	default:
528 		return (ESRCH);
529 	}
530 	return (0);
531 }
532 
533 static int
534 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
535 {
536 	struct ifaddr *ifa;
537 	sa_family_t saf;
538 
539 	if (V_loif == NULL) {
540 		RTS_PID_LOG(LOG_INFO, "Unable to add blackhole/reject nhop without loopback");
541 		return (ENOTSUP);
542 	}
543 	info->rti_ifp = V_loif;
544 
545 	saf = info->rti_info[RTAX_DST]->sa_family;
546 
547 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
548 		if (ifa->ifa_addr->sa_family == saf) {
549 			info->rti_ifa = ifa;
550 			break;
551 		}
552 	}
553 	if (info->rti_ifa == NULL) {
554 		RTS_PID_LOG(LOG_INFO, "Unable to find ifa for blackhole/reject nhop");
555 		return (ENOTSUP);
556 	}
557 
558 	bzero(saun, sizeof(union sockaddr_union));
559 	switch (saf) {
560 #ifdef INET
561 	case AF_INET:
562 		saun->sin.sin_family = AF_INET;
563 		saun->sin.sin_len = sizeof(struct sockaddr_in);
564 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
565 		break;
566 #endif
567 #ifdef INET6
568 	case AF_INET6:
569 		saun->sin6.sin6_family = AF_INET6;
570 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
571 		saun->sin6.sin6_addr = in6addr_loopback;
572 		break;
573 #endif
574 	default:
575 		RTS_PID_LOG(LOG_INFO, "unsupported family: %d", saf);
576 		return (ENOTSUP);
577 	}
578 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
579 	info->rti_flags |= RTF_GATEWAY;
580 
581 	return (0);
582 }
583 
584 /*
585  * Fills in @info based on userland-provided @rtm message.
586  *
587  * Returns 0 on success.
588  */
589 static int
590 fill_addrinfo(struct rt_msghdr *rtm, int len, struct linear_buffer *lb, u_int fibnum,
591     struct rt_addrinfo *info)
592 {
593 	int error;
594 
595 	rtm->rtm_pid = curproc->p_pid;
596 	info->rti_addrs = rtm->rtm_addrs;
597 
598 	info->rti_mflags = rtm->rtm_inits;
599 	info->rti_rmx = &rtm->rtm_rmx;
600 
601 	/*
602 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
603 	 * link-local address because rtrequest requires addresses with
604 	 * embedded scope id.
605 	 */
606 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
607 		return (EINVAL);
608 
609 	info->rti_flags = rtm->rtm_flags;
610 	error = cleanup_xaddrs(info, lb);
611 	if (error != 0)
612 		return (error);
613 	/*
614 	 * Verify that the caller has the appropriate privilege; RTM_GET
615 	 * is the only operation the non-superuser is allowed.
616 	 */
617 	if (rtm->rtm_type != RTM_GET) {
618 		error = priv_check(curthread, PRIV_NET_ROUTE);
619 		if (error != 0)
620 			return (error);
621 	}
622 
623 	/*
624 	 * The given gateway address may be an interface address.
625 	 * For example, issuing a "route change" command on a route
626 	 * entry that was created from a tunnel, and the gateway
627 	 * address given is the local end point. In this case the
628 	 * RTF_GATEWAY flag must be cleared or the destination will
629 	 * not be reachable even though there is no error message.
630 	 */
631 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
632 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
633 		struct nhop_object *nh;
634 
635 		/*
636 		 * A host route through the loopback interface is
637 		 * installed for each interface adddress. In pre 8.0
638 		 * releases the interface address of a PPP link type
639 		 * is not reachable locally. This behavior is fixed as
640 		 * part of the new L2/L3 redesign and rewrite work. The
641 		 * signature of this interface address route is the
642 		 * AF_LINK sa_family type of the gateway, and the
643 		 * rt_ifp has the IFF_LOOPBACK flag set.
644 		 */
645 		nh = rib_lookup(fibnum, info->rti_info[RTAX_GATEWAY], NHR_NONE, 0);
646 		if (nh != NULL && nh->gw_sa.sa_family == AF_LINK &&
647 		    nh->nh_ifp->if_flags & IFF_LOOPBACK) {
648 				info->rti_flags &= ~RTF_GATEWAY;
649 				info->rti_flags |= RTF_GWFLAG_COMPAT;
650 		}
651 	}
652 
653 	return (0);
654 }
655 
656 static struct nhop_object *
657 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
658 {
659 	if (!NH_IS_NHGRP(nh))
660 		return (nh);
661 #ifdef ROUTE_MPATH
662 	const struct weightened_nhop *wn;
663 	uint32_t num_nhops;
664 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
665 	if (gw == NULL)
666 		return (wn[0].nh);
667 	for (int i = 0; i < num_nhops; i++) {
668 		if (match_nhop_gw(wn[i].nh, gw))
669 			return (wn[i].nh);
670 	}
671 #endif
672 	return (NULL);
673 }
674 
675 /*
676  * Handles RTM_GET message from routing socket, returning matching rt.
677  *
678  * Returns:
679  * 0 on success, with locked and referenced matching rt in @rt_nrt
680  * errno of failure
681  */
682 static int
683 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
684     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
685 {
686 	RIB_RLOCK_TRACKER;
687 	struct rib_head *rnh;
688 	struct nhop_object *nh;
689 	sa_family_t saf;
690 
691 	saf = info->rti_info[RTAX_DST]->sa_family;
692 
693 	rnh = rt_tables_get_rnh(fibnum, saf);
694 	if (rnh == NULL)
695 		return (EAFNOSUPPORT);
696 
697 	RIB_RLOCK(rnh);
698 
699 	/*
700 	 * By (implicit) convention host route (one without netmask)
701 	 * means longest-prefix-match request and the route with netmask
702 	 * means exact-match lookup.
703 	 * As cleanup_xaddrs() cleans up info flags&addrs for the /32,/128
704 	 * prefixes, use original data to check for the netmask presence.
705 	 */
706 	if ((rtm->rtm_addrs & RTA_NETMASK) == 0) {
707 		/*
708 		 * Provide longest prefix match for
709 		 * address lookup (no mask).
710 		 * 'route -n get addr'
711 		 */
712 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
713 		    info->rti_info[RTAX_DST], &rnh->head);
714 	} else
715 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
716 		    info->rti_info[RTAX_DST],
717 		    info->rti_info[RTAX_NETMASK], &rnh->head);
718 
719 	if (rc->rc_rt == NULL) {
720 		RIB_RUNLOCK(rnh);
721 		return (ESRCH);
722 	}
723 
724 	nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
725 	if (nh == NULL) {
726 		RIB_RUNLOCK(rnh);
727 		return (ESRCH);
728 	}
729 	/*
730 	 * If performing proxied L2 entry insertion, and
731 	 * the actual PPP host entry is found, perform
732 	 * another search to retrieve the prefix route of
733 	 * the local end point of the PPP link.
734 	 * TODO: move this logic to userland.
735 	 */
736 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
737 		struct sockaddr_storage laddr;
738 
739 		if (nh->nh_ifp != NULL &&
740 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
741 			struct ifaddr *ifa;
742 
743 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
744 					RT_ALL_FIBS);
745 			if (ifa != NULL)
746 				rt_maskedcopy(ifa->ifa_addr,
747 					      (struct sockaddr *)&laddr,
748 					      ifa->ifa_netmask);
749 		} else
750 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
751 				      (struct sockaddr *)&laddr,
752 				      nh->nh_ifa->ifa_netmask);
753 		/*
754 		 * refactor rt and no lock operation necessary
755 		 */
756 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(
757 		    (struct sockaddr *)&laddr, &rnh->head);
758 		if (rc->rc_rt == NULL) {
759 			RIB_RUNLOCK(rnh);
760 			return (ESRCH);
761 		}
762 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
763 		if (nh == NULL) {
764 			RIB_RUNLOCK(rnh);
765 			return (ESRCH);
766 		}
767 	}
768 	rc->rc_nh_new = nh;
769 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
770 	RIB_RUNLOCK(rnh);
771 
772 	return (0);
773 }
774 
775 static void
776 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
777 {
778 #ifdef INET
779 	if (family == AF_INET) {
780 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
781 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
782 
783 		bzero(dst4, sizeof(struct sockaddr_in));
784 		bzero(mask4, sizeof(struct sockaddr_in));
785 
786 		dst4->sin_family = AF_INET;
787 		dst4->sin_len = sizeof(struct sockaddr_in);
788 		mask4->sin_family = AF_INET;
789 		mask4->sin_len = sizeof(struct sockaddr_in);
790 	}
791 #endif
792 #ifdef INET6
793 	if (family == AF_INET6) {
794 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
795 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
796 
797 		bzero(dst6, sizeof(struct sockaddr_in6));
798 		bzero(mask6, sizeof(struct sockaddr_in6));
799 
800 		dst6->sin6_family = AF_INET6;
801 		dst6->sin6_len = sizeof(struct sockaddr_in6);
802 		mask6->sin6_family = AF_INET6;
803 		mask6->sin6_len = sizeof(struct sockaddr_in6);
804 	}
805 #endif
806 }
807 
808 static void
809 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
810     struct sockaddr *mask)
811 {
812 #ifdef INET
813 	if (dst->sa_family == AF_INET) {
814 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
815 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
816 		uint32_t scopeid = 0;
817 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
818 		    &scopeid);
819 		return;
820 	}
821 #endif
822 #ifdef INET6
823 	if (dst->sa_family == AF_INET6) {
824 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
825 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
826 		uint32_t scopeid = 0;
827 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
828 		    &mask6->sin6_addr, &scopeid);
829 		dst6->sin6_scope_id = scopeid;
830 		return;
831 	}
832 #endif
833 }
834 
835 static int
836 update_rtm_from_info(struct rt_addrinfo *info, struct rt_msghdr **prtm,
837     int alloc_len)
838 {
839 	struct rt_msghdr *rtm, *orig_rtm = NULL;
840 	struct walkarg w;
841 	int len;
842 
843 	rtm = *prtm;
844 	/* Check if we need to realloc storage */
845 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
846 	if (len > alloc_len) {
847 		struct rt_msghdr *tmp_rtm;
848 
849 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
850 		if (tmp_rtm == NULL)
851 			return (ENOBUFS);
852 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
853 		orig_rtm = rtm;
854 		rtm = tmp_rtm;
855 		alloc_len = len;
856 
857 		/*
858 		 * Delay freeing original rtm as info contains
859 		 * data referencing it.
860 		 */
861 	}
862 
863 	w.w_tmem = (caddr_t)rtm;
864 	w.w_tmemsize = alloc_len;
865 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
866 	rtm->rtm_addrs = info->rti_addrs;
867 
868 	if (orig_rtm != NULL)
869 		free(orig_rtm, M_TEMP);
870 	*prtm = rtm;
871 	return (0);
872 }
873 
874 
875 /*
876  * Update sockaddrs, flags, etc in @prtm based on @rc data.
877  * rtm can be reallocated.
878  *
879  * Returns 0 on success, along with pointer to (potentially reallocated)
880  *  rtm.
881  *
882  */
883 static int
884 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
885     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
886 {
887 	union sockaddr_union saun;
888 	struct rt_msghdr *rtm;
889 	struct ifnet *ifp;
890 	int error;
891 
892 	rtm = *prtm;
893 	union sockaddr_union sa_dst, sa_mask;
894 	int family = info->rti_info[RTAX_DST]->sa_family;
895 	init_sockaddrs_family(family, &sa_dst.sa, &sa_mask.sa);
896 	export_rtaddrs(rc->rc_rt, &sa_dst.sa, &sa_mask.sa);
897 
898 	info->rti_info[RTAX_DST] = &sa_dst.sa;
899 	info->rti_info[RTAX_NETMASK] = rt_is_host(rc->rc_rt) ? NULL : &sa_mask.sa;
900 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
901 	info->rti_info[RTAX_GENMASK] = 0;
902 	ifp = nh->nh_ifp;
903 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
904 		if (ifp) {
905 			info->rti_info[RTAX_IFP] =
906 			    ifp->if_addr->ifa_addr;
907 			error = rtm_get_jailed(info, ifp, nh,
908 			    &saun, curthread->td_ucred);
909 			if (error != 0)
910 				return (error);
911 			if (ifp->if_flags & IFF_POINTOPOINT)
912 				info->rti_info[RTAX_BRD] =
913 				    nh->nh_ifa->ifa_dstaddr;
914 			rtm->rtm_index = ifp->if_index;
915 		} else {
916 			info->rti_info[RTAX_IFP] = NULL;
917 			info->rti_info[RTAX_IFA] = NULL;
918 		}
919 	} else if (ifp != NULL)
920 		rtm->rtm_index = ifp->if_index;
921 
922 	if ((error = update_rtm_from_info(info, prtm, alloc_len)) != 0)
923 		return (error);
924 
925 	rtm = *prtm;
926 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
927 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
928 		rtm->rtm_flags = RTF_GATEWAY |
929 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
930 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
931 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
932 
933 	return (0);
934 }
935 
936 #ifdef ROUTE_MPATH
937 static void
938 save_del_notification(const struct rib_cmd_info *rc, void *_cbdata)
939 {
940 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
941 
942 	if (rc->rc_cmd == RTM_DELETE)
943 		*rc_new = *rc;
944 }
945 
946 static void
947 save_add_notification(const struct rib_cmd_info *rc, void *_cbdata)
948 {
949 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
950 
951 	if (rc->rc_cmd == RTM_ADD)
952 		*rc_new = *rc;
953 }
954 #endif
955 
956 #if defined(INET6) || defined(INET)
957 static struct sockaddr *
958 alloc_sockaddr_aligned(struct linear_buffer *lb, int len)
959 {
960 	len = roundup2(len, sizeof(uint64_t));
961 	if (lb->offset + len > lb->size)
962 		return (NULL);
963 	struct sockaddr *sa = (struct sockaddr *)(lb->base + lb->offset);
964 	lb->offset += len;
965 	return (sa);
966 }
967 #endif
968 
969 static int
970 rts_send(struct socket *so, int flags, struct mbuf *m,
971     struct sockaddr *nam, struct mbuf *control, struct thread *td)
972 {
973 	struct rt_msghdr *rtm = NULL;
974 	struct rt_addrinfo info;
975 	struct epoch_tracker et;
976 #ifdef INET6
977 	struct sockaddr_storage ss;
978 	struct sockaddr_in6 *sin6;
979 	int i, rti_need_deembed = 0;
980 #endif
981 	int alloc_len = 0, len, error = 0, fibnum;
982 	sa_family_t saf = AF_UNSPEC;
983 	struct rib_cmd_info rc;
984 	struct nhop_object *nh;
985 
986 	if ((flags & PRUS_OOB) || control != NULL) {
987 		m_freem(m);
988 		if (control != NULL)
989 			m_freem(control);
990 		return (EOPNOTSUPP);
991 	}
992 
993 	fibnum = so->so_fibnum;
994 #define senderr(e) { error = e; goto flush;}
995 	if (m == NULL || ((m->m_len < sizeof(long)) &&
996 		       (m = m_pullup(m, sizeof(long))) == NULL))
997 		return (ENOBUFS);
998 	if ((m->m_flags & M_PKTHDR) == 0)
999 		panic("route_output");
1000 	NET_EPOCH_ENTER(et);
1001 	len = m->m_pkthdr.len;
1002 	if (len < sizeof(*rtm) ||
1003 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1004 		senderr(EINVAL);
1005 
1006 	/*
1007 	 * Most of current messages are in range 200-240 bytes,
1008 	 * minimize possible re-allocation on reply using larger size
1009 	 * buffer aligned on 1k boundaty.
1010 	 */
1011 	alloc_len = roundup2(len, 1024);
1012 	int total_len = alloc_len + SCRATCH_BUFFER_SIZE;
1013 	if ((rtm = malloc(total_len, M_TEMP, M_NOWAIT)) == NULL)
1014 		senderr(ENOBUFS);
1015 
1016 	m_copydata(m, 0, len, (caddr_t)rtm);
1017 	bzero(&info, sizeof(info));
1018 	nh = NULL;
1019 	struct linear_buffer lb = {
1020 		.base = (char *)rtm + alloc_len,
1021 		.size = SCRATCH_BUFFER_SIZE,
1022 	};
1023 
1024 	if (rtm->rtm_version != RTM_VERSION) {
1025 		/* Do not touch message since format is unknown */
1026 		free(rtm, M_TEMP);
1027 		rtm = NULL;
1028 		senderr(EPROTONOSUPPORT);
1029 	}
1030 
1031 	/*
1032 	 * Starting from here, it is possible
1033 	 * to alter original message and insert
1034 	 * caller PID and error value.
1035 	 */
1036 
1037 	if ((error = fill_addrinfo(rtm, len, &lb, fibnum, &info)) != 0) {
1038 		senderr(error);
1039 	}
1040 	/* fill_addringo() embeds scope into IPv6 addresses */
1041 #ifdef INET6
1042 	rti_need_deembed = 1;
1043 #endif
1044 
1045 	saf = info.rti_info[RTAX_DST]->sa_family;
1046 
1047 	/* support for new ARP code */
1048 	if (rtm->rtm_flags & RTF_LLDATA) {
1049 		error = lla_rt_output(rtm, &info);
1050 		goto flush;
1051 	}
1052 
1053 	union sockaddr_union gw_saun;
1054 	int blackhole_flags = rtm->rtm_flags & (RTF_BLACKHOLE|RTF_REJECT);
1055 	if (blackhole_flags != 0) {
1056 		if (blackhole_flags != (RTF_BLACKHOLE | RTF_REJECT))
1057 			error = fill_blackholeinfo(&info, &gw_saun);
1058 		else {
1059 			RTS_PID_LOG(LOG_DEBUG, "both BLACKHOLE and REJECT flags specifiied");
1060 			error = EINVAL;
1061 		}
1062 		if (error != 0)
1063 			senderr(error);
1064 	}
1065 
1066 	switch (rtm->rtm_type) {
1067 	case RTM_ADD:
1068 	case RTM_CHANGE:
1069 		if (rtm->rtm_type == RTM_ADD) {
1070 			if (info.rti_info[RTAX_GATEWAY] == NULL) {
1071 				RTS_PID_LOG(LOG_DEBUG, "RTM_ADD w/o gateway");
1072 				senderr(EINVAL);
1073 			}
1074 		}
1075 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1076 		if (error == 0) {
1077 #ifdef ROUTE_MPATH
1078 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
1079 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
1080 				struct rib_cmd_info rc_simple = {};
1081 				rib_decompose_notification(&rc,
1082 				    save_add_notification, (void *)&rc_simple);
1083 				rc = rc_simple;
1084 			}
1085 #endif
1086 			/* nh MAY be empty if RTM_CHANGE request is no-op */
1087 			nh = rc.rc_nh_new;
1088 			if (nh != NULL) {
1089 				rtm->rtm_index = nh->nh_ifp->if_index;
1090 				rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1091 			}
1092 		}
1093 		break;
1094 
1095 	case RTM_DELETE:
1096 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1097 		if (error == 0) {
1098 #ifdef ROUTE_MPATH
1099 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1100 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1101 				struct rib_cmd_info rc_simple = {};
1102 				rib_decompose_notification(&rc,
1103 				    save_del_notification, (void *)&rc_simple);
1104 				rc = rc_simple;
1105 			}
1106 #endif
1107 			nh = rc.rc_nh_old;
1108 		}
1109 		break;
1110 
1111 	case RTM_GET:
1112 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1113 		if (error != 0)
1114 			senderr(error);
1115 		nh = rc.rc_nh_new;
1116 
1117 		if (!can_export_rte(curthread->td_ucred,
1118 		    info.rti_info[RTAX_NETMASK] == NULL,
1119 		    info.rti_info[RTAX_DST])) {
1120 			senderr(ESRCH);
1121 		}
1122 		break;
1123 
1124 	default:
1125 		senderr(EOPNOTSUPP);
1126 	}
1127 
1128 	if (error == 0 && nh != NULL) {
1129 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1130 		/*
1131 		 * Note that some sockaddr pointers may have changed to
1132 		 * point to memory outsize @rtm. Some may be pointing
1133 		 * to the on-stack variables.
1134 		 * Given that, any pointer in @info CANNOT BE USED.
1135 		 */
1136 
1137 		/*
1138 		 * scopeid deembedding has been performed while
1139 		 * writing updated rtm in rtsock_msg_buffer().
1140 		 * With that in mind, skip deembedding procedure below.
1141 		 */
1142 #ifdef INET6
1143 		rti_need_deembed = 0;
1144 #endif
1145 	}
1146 
1147 flush:
1148 	NET_EPOCH_EXIT(et);
1149 
1150 #ifdef INET6
1151 	if (rtm != NULL) {
1152 		if (rti_need_deembed) {
1153 			/* sin6_scope_id is recovered before sending rtm. */
1154 			sin6 = (struct sockaddr_in6 *)&ss;
1155 			for (i = 0; i < RTAX_MAX; i++) {
1156 				if (info.rti_info[i] == NULL)
1157 					continue;
1158 				if (info.rti_info[i]->sa_family != AF_INET6)
1159 					continue;
1160 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1161 				if (sa6_recoverscope(sin6) == 0)
1162 					bcopy(sin6, info.rti_info[i],
1163 						    sizeof(*sin6));
1164 			}
1165 			if (update_rtm_from_info(&info, &rtm, alloc_len) != 0) {
1166 				if (error != 0)
1167 					error = ENOBUFS;
1168 			}
1169 		}
1170 	}
1171 #endif
1172 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1173 
1174 	return (error);
1175 }
1176 
1177 /*
1178  * Sends the prepared reply message in @rtm to all rtsock clients.
1179  * Frees @m and @rtm.
1180  *
1181  */
1182 static void
1183 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1184     sa_family_t saf, u_int fibnum, int rtm_errno)
1185 {
1186 	struct rcb *rcb = NULL;
1187 
1188 	/*
1189 	 * Check to see if we don't want our own messages.
1190 	 */
1191 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1192 		if (V_route_cb.any_count <= 1) {
1193 			if (rtm != NULL)
1194 				free(rtm, M_TEMP);
1195 			m_freem(m);
1196 			return;
1197 		}
1198 		/* There is another listener, so construct message */
1199 		rcb = so->so_pcb;
1200 	}
1201 
1202 	if (rtm != NULL) {
1203 		if (rtm_errno!= 0)
1204 			rtm->rtm_errno = rtm_errno;
1205 		else
1206 			rtm->rtm_flags |= RTF_DONE;
1207 
1208 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1209 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1210 			m_freem(m);
1211 			m = NULL;
1212 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1213 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1214 
1215 		free(rtm, M_TEMP);
1216 	}
1217 	if (m != NULL) {
1218 		M_SETFIB(m, fibnum);
1219 		m->m_flags |= RTS_FILTER_FIB;
1220 		if (rcb) {
1221 			/*
1222 			 * XXX insure we don't get a copy by
1223 			 * invalidating our protocol
1224 			 */
1225 			sa_family_t family = rcb->rcb_family;
1226 			rcb->rcb_family = AF_UNSPEC;
1227 			rt_dispatch(m, saf);
1228 			rcb->rcb_family = family;
1229 		} else
1230 			rt_dispatch(m, saf);
1231 	}
1232 }
1233 
1234 static void
1235 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1236     struct rt_metrics *out)
1237 {
1238 
1239 	bzero(out, sizeof(*out));
1240 	out->rmx_mtu = nh->nh_mtu;
1241 	out->rmx_weight = rt->rt_weight;
1242 	out->rmx_nhidx = nhop_get_idx(nh);
1243 	/* Kernel -> userland timebase conversion. */
1244 	out->rmx_expire = nhop_get_expire(nh) ?
1245 	    nhop_get_expire(nh) - time_uptime + time_second : 0;
1246 }
1247 
1248 /*
1249  * Extract the addresses of the passed sockaddrs.
1250  * Do a little sanity checking so as to avoid bad memory references.
1251  * This data is derived straight from userland.
1252  */
1253 static int
1254 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1255 {
1256 	struct sockaddr *sa;
1257 	int i;
1258 
1259 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1260 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1261 			continue;
1262 		sa = (struct sockaddr *)cp;
1263 		/*
1264 		 * It won't fit.
1265 		 */
1266 		if (cp + sa->sa_len > cplim) {
1267 			RTS_PID_LOG(LOG_DEBUG, "sa_len too big for sa type %d", i);
1268 			return (EINVAL);
1269 		}
1270 		/*
1271 		 * there are no more.. quit now
1272 		 * If there are more bits, they are in error.
1273 		 * I've seen this. route(1) can evidently generate these.
1274 		 * This causes kernel to core dump.
1275 		 * for compatibility, If we see this, point to a safe address.
1276 		 */
1277 		if (sa->sa_len == 0) {
1278 			rtinfo->rti_info[i] = &sa_zero;
1279 			return (0); /* should be EINVAL but for compat */
1280 		}
1281 		/* accept it */
1282 #ifdef INET6
1283 		if (sa->sa_family == AF_INET6)
1284 			sa6_embedscope((struct sockaddr_in6 *)sa,
1285 			    V_ip6_use_defzone);
1286 #endif
1287 		rtinfo->rti_info[i] = sa;
1288 		cp += SA_SIZE(sa);
1289 	}
1290 	return (0);
1291 }
1292 
1293 #ifdef INET
1294 static inline void
1295 fill_sockaddr_inet(struct sockaddr_in *sin, struct in_addr addr)
1296 {
1297 
1298 	const struct sockaddr_in nsin = {
1299 		.sin_family = AF_INET,
1300 		.sin_len = sizeof(struct sockaddr_in),
1301 		.sin_addr = addr,
1302 	};
1303 	*sin = nsin;
1304 }
1305 #endif
1306 
1307 #ifdef INET6
1308 static inline void
1309 fill_sockaddr_inet6(struct sockaddr_in6 *sin6, const struct in6_addr *addr6,
1310     uint32_t scopeid)
1311 {
1312 
1313 	const struct sockaddr_in6 nsin6 = {
1314 		.sin6_family = AF_INET6,
1315 		.sin6_len = sizeof(struct sockaddr_in6),
1316 		.sin6_addr = *addr6,
1317 		.sin6_scope_id = scopeid,
1318 	};
1319 	*sin6 = nsin6;
1320 }
1321 #endif
1322 
1323 #if defined(INET6) || defined(INET)
1324 /*
1325  * Checks if gateway is suitable for lltable operations.
1326  * Lltable code requires AF_LINK gateway with ifindex
1327  *  and mac address specified.
1328  * Returns 0 on success.
1329  */
1330 static int
1331 cleanup_xaddrs_lladdr(struct rt_addrinfo *info)
1332 {
1333 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)info->rti_info[RTAX_GATEWAY];
1334 
1335 	if (sdl->sdl_family != AF_LINK)
1336 		return (EINVAL);
1337 
1338 	if (sdl->sdl_index == 0) {
1339 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gateway w/o ifindex");
1340 		return (EINVAL);
1341 	}
1342 
1343 	if (offsetof(struct sockaddr_dl, sdl_data) + sdl->sdl_nlen + sdl->sdl_alen > sdl->sdl_len) {
1344 		RTS_PID_LOG(LOG_DEBUG, "AF_LINK gw: sdl_nlen/sdl_alen too large");
1345 		return (EINVAL);
1346 	}
1347 
1348 	return (0);
1349 }
1350 
1351 static int
1352 cleanup_xaddrs_gateway(struct rt_addrinfo *info, struct linear_buffer *lb)
1353 {
1354 	struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
1355 	struct sockaddr *sa;
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 
1366 			/* Ensure reads do not go beyoud SA boundary */
1367 			if (SA_SIZE(gw) < offsetof(struct sockaddr_in, sin_zero)) {
1368 				RTS_PID_LOG(LOG_DEBUG, "gateway sin_len too small: %d",
1369 				    gw->sa_len);
1370 				return (EINVAL);
1371 			}
1372 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_in));
1373 			if (sa == NULL)
1374 				return (ENOBUFS);
1375 			fill_sockaddr_inet((struct sockaddr_in *)sa, gw_sin->sin_addr);
1376 			info->rti_info[RTAX_GATEWAY] = sa;
1377 		}
1378 		break;
1379 #endif
1380 #ifdef INET6
1381 	case AF_INET6:
1382 		{
1383 			struct sockaddr_in6 *gw_sin6 = (struct sockaddr_in6 *)gw;
1384 			if (gw_sin6->sin6_len < sizeof(struct sockaddr_in6)) {
1385 				RTS_PID_LOG(LOG_DEBUG, "gateway sin6_len too small: %d",
1386 				    gw->sa_len);
1387 				return (EINVAL);
1388 			}
1389 			fill_sockaddr_inet6(gw_sin6, &gw_sin6->sin6_addr, 0);
1390 			break;
1391 		}
1392 #endif
1393 	case AF_LINK:
1394 		{
1395 			struct sockaddr_dl *gw_sdl;
1396 
1397 			size_t sdl_min_len = offsetof(struct sockaddr_dl, sdl_data);
1398 			gw_sdl = (struct sockaddr_dl *)gw;
1399 			if (gw_sdl->sdl_len < sdl_min_len) {
1400 				RTS_PID_LOG(LOG_DEBUG, "gateway sdl_len too small: %d",
1401 				    gw_sdl->sdl_len);
1402 				return (EINVAL);
1403 			}
1404 			sa = alloc_sockaddr_aligned(lb, sizeof(struct sockaddr_dl_short));
1405 			if (sa == NULL)
1406 				return (ENOBUFS);
1407 
1408 			const struct sockaddr_dl_short sdl = {
1409 				.sdl_family = AF_LINK,
1410 				.sdl_len = sizeof(struct sockaddr_dl_short),
1411 				.sdl_index = gw_sdl->sdl_index,
1412 			};
1413 			*((struct sockaddr_dl_short *)sa) = sdl;
1414 			info->rti_info[RTAX_GATEWAY] = sa;
1415 			break;
1416 		}
1417 	}
1418 
1419 	return (0);
1420 }
1421 #endif
1422 
1423 static void
1424 remove_netmask(struct rt_addrinfo *info)
1425 {
1426 	info->rti_info[RTAX_NETMASK] = NULL;
1427 	info->rti_flags |= RTF_HOST;
1428 	info->rti_addrs &= ~RTA_NETMASK;
1429 }
1430 
1431 #ifdef INET
1432 static int
1433 cleanup_xaddrs_inet(struct rt_addrinfo *info, struct linear_buffer *lb)
1434 {
1435 	struct sockaddr_in *dst_sa, *mask_sa;
1436 	const int sa_len = sizeof(struct sockaddr_in);
1437 	struct in_addr dst, mask;
1438 
1439 	/* Check & fixup dst/netmask combination first */
1440 	dst_sa = (struct sockaddr_in *)info->rti_info[RTAX_DST];
1441 	mask_sa = (struct sockaddr_in *)info->rti_info[RTAX_NETMASK];
1442 
1443 	/* Ensure reads do not go beyound the buffer size */
1444 	if (SA_SIZE(dst_sa) < offsetof(struct sockaddr_in, sin_zero)) {
1445 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin_len too small: %d",
1446 		    dst_sa->sin_len);
1447 		return (EINVAL);
1448 	}
1449 
1450 	if ((mask_sa != NULL) && mask_sa->sin_len < sizeof(struct sockaddr_in)) {
1451 		/*
1452 		 * Some older routing software encode mask length into the
1453 		 * sin_len, thus resulting in "truncated" sockaddr.
1454 		 */
1455 		int len = mask_sa->sin_len - offsetof(struct sockaddr_in, sin_addr);
1456 		if (len >= 0) {
1457 			mask.s_addr = 0;
1458 			if (len > sizeof(struct in_addr))
1459 				len = sizeof(struct in_addr);
1460 			memcpy(&mask, &mask_sa->sin_addr, len);
1461 		} else {
1462 			RTS_PID_LOG(LOG_DEBUG, "prefix mask sin_len too small: %d",
1463 			    mask_sa->sin_len);
1464 			return (EINVAL);
1465 		}
1466 	} else
1467 		mask.s_addr = mask_sa ? mask_sa->sin_addr.s_addr : INADDR_BROADCAST;
1468 
1469 	dst.s_addr = htonl(ntohl(dst_sa->sin_addr.s_addr) & ntohl(mask.s_addr));
1470 
1471 	/* Construct new "clean" dst/mask sockaddresses */
1472 	if ((dst_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1473 		return (ENOBUFS);
1474 	fill_sockaddr_inet(dst_sa, dst);
1475 	info->rti_info[RTAX_DST] = (struct sockaddr *)dst_sa;
1476 
1477 	if (mask.s_addr != INADDR_BROADCAST) {
1478 		if ((mask_sa = (struct sockaddr_in *)alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1479 			return (ENOBUFS);
1480 		fill_sockaddr_inet(mask_sa, mask);
1481 		info->rti_info[RTAX_NETMASK] = (struct sockaddr *)mask_sa;
1482 		info->rti_flags &= ~RTF_HOST;
1483 	} else
1484 		remove_netmask(info);
1485 
1486 	/* Check gateway */
1487 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1488 		return (cleanup_xaddrs_gateway(info, lb));
1489 
1490 	return (0);
1491 }
1492 #endif
1493 
1494 #ifdef INET6
1495 static int
1496 cleanup_xaddrs_inet6(struct rt_addrinfo *info, struct linear_buffer *lb)
1497 {
1498 	struct sockaddr *sa;
1499 	struct sockaddr_in6 *dst_sa, *mask_sa;
1500 	struct in6_addr mask, *dst;
1501 	const int sa_len = sizeof(struct sockaddr_in6);
1502 
1503 	/* Check & fixup dst/netmask combination first */
1504 	dst_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_DST];
1505 	mask_sa = (struct sockaddr_in6 *)info->rti_info[RTAX_NETMASK];
1506 
1507 	if (dst_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1508 		RTS_PID_LOG(LOG_DEBUG, "prefix dst sin6_len too small: %d",
1509 		    dst_sa->sin6_len);
1510 		return (EINVAL);
1511 	}
1512 
1513 	if (mask_sa && mask_sa->sin6_len < sizeof(struct sockaddr_in6)) {
1514 		/*
1515 		 * Some older routing software encode mask length into the
1516 		 * sin6_len, thus resulting in "truncated" sockaddr.
1517 		 */
1518 		int len = mask_sa->sin6_len - offsetof(struct sockaddr_in6, sin6_addr);
1519 		if (len >= 0) {
1520 			bzero(&mask, sizeof(mask));
1521 			if (len > sizeof(struct in6_addr))
1522 				len = sizeof(struct in6_addr);
1523 			memcpy(&mask, &mask_sa->sin6_addr, len);
1524 		} else {
1525 			RTS_PID_LOG(LOG_DEBUG, "rtsock: prefix mask sin6_len too small: %d",
1526 			    mask_sa->sin6_len);
1527 			return (EINVAL);
1528 		}
1529 	} else
1530 		mask = mask_sa ? mask_sa->sin6_addr : in6mask128;
1531 
1532 	dst = &dst_sa->sin6_addr;
1533 	IN6_MASK_ADDR(dst, &mask);
1534 
1535 	if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1536 		return (ENOBUFS);
1537 	fill_sockaddr_inet6((struct sockaddr_in6 *)sa, dst, 0);
1538 	info->rti_info[RTAX_DST] = sa;
1539 
1540 	if (!IN6_ARE_ADDR_EQUAL(&mask, &in6mask128)) {
1541 		if ((sa = alloc_sockaddr_aligned(lb, sa_len)) == NULL)
1542 			return (ENOBUFS);
1543 		fill_sockaddr_inet6((struct sockaddr_in6 *)sa, &mask, 0);
1544 		info->rti_info[RTAX_NETMASK] = sa;
1545 		info->rti_flags &= ~RTF_HOST;
1546 	} else
1547 		remove_netmask(info);
1548 
1549 	/* Check gateway */
1550 	if (info->rti_info[RTAX_GATEWAY] != NULL)
1551 		return (cleanup_xaddrs_gateway(info, lb));
1552 
1553 	return (0);
1554 }
1555 #endif
1556 
1557 static int
1558 cleanup_xaddrs(struct rt_addrinfo *info, struct linear_buffer *lb)
1559 {
1560 	int error = EAFNOSUPPORT;
1561 
1562 	if (info->rti_info[RTAX_DST] == NULL) {
1563 		RTS_PID_LOG(LOG_DEBUG, "prefix dst is not set");
1564 		return (EINVAL);
1565 	}
1566 
1567 	if (info->rti_flags & RTF_LLDATA) {
1568 		/*
1569 		 * arp(8)/ndp(8) sends RTA_NETMASK for the associated
1570 		 * prefix along with the actual address in RTA_DST.
1571 		 * Remove netmask to avoid unnecessary address masking.
1572 		 */
1573 		remove_netmask(info);
1574 	}
1575 
1576 	switch (info->rti_info[RTAX_DST]->sa_family) {
1577 #ifdef INET
1578 	case AF_INET:
1579 		error = cleanup_xaddrs_inet(info, lb);
1580 		break;
1581 #endif
1582 #ifdef INET6
1583 	case AF_INET6:
1584 		error = cleanup_xaddrs_inet6(info, lb);
1585 		break;
1586 #endif
1587 	}
1588 
1589 	return (error);
1590 }
1591 
1592 /*
1593  * Fill in @dmask with valid netmask leaving original @smask
1594  * intact. Mostly used with radix netmasks.
1595  */
1596 struct sockaddr *
1597 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1598     struct sockaddr_storage *dmask)
1599 {
1600 	if (dst == NULL || smask == NULL)
1601 		return (NULL);
1602 
1603 	memset(dmask, 0, dst->sa_len);
1604 	memcpy(dmask, smask, smask->sa_len);
1605 	dmask->ss_len = dst->sa_len;
1606 	dmask->ss_family = dst->sa_family;
1607 
1608 	return ((struct sockaddr *)dmask);
1609 }
1610 
1611 /*
1612  * Writes information related to @rtinfo object to newly-allocated mbuf.
1613  * Assumes MCLBYTES is enough to construct any message.
1614  * Used for OS notifications of vaious events (if/ifa announces,etc)
1615  *
1616  * Returns allocated mbuf or NULL on failure.
1617  */
1618 static struct mbuf *
1619 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1620 {
1621 	struct sockaddr_storage ss;
1622 	struct rt_msghdr *rtm;
1623 	struct mbuf *m;
1624 	int i;
1625 	struct sockaddr *sa;
1626 #ifdef INET6
1627 	struct sockaddr_in6 *sin6;
1628 #endif
1629 	int len, dlen;
1630 
1631 	switch (type) {
1632 	case RTM_DELADDR:
1633 	case RTM_NEWADDR:
1634 		len = sizeof(struct ifa_msghdr);
1635 		break;
1636 
1637 	case RTM_DELMADDR:
1638 	case RTM_NEWMADDR:
1639 		len = sizeof(struct ifma_msghdr);
1640 		break;
1641 
1642 	case RTM_IFINFO:
1643 		len = sizeof(struct if_msghdr);
1644 		break;
1645 
1646 	case RTM_IFANNOUNCE:
1647 	case RTM_IEEE80211:
1648 		len = sizeof(struct if_announcemsghdr);
1649 		break;
1650 
1651 	default:
1652 		len = sizeof(struct rt_msghdr);
1653 	}
1654 
1655 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1656 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1657 	if (len > MHLEN)
1658 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1659 	else
1660 		m = m_gethdr(M_NOWAIT, MT_DATA);
1661 	if (m == NULL)
1662 		return (m);
1663 
1664 	m->m_pkthdr.len = m->m_len = len;
1665 	rtm = mtod(m, struct rt_msghdr *);
1666 	bzero((caddr_t)rtm, len);
1667 	for (i = 0; i < RTAX_MAX; i++) {
1668 		if ((sa = rtinfo->rti_info[i]) == NULL)
1669 			continue;
1670 		rtinfo->rti_addrs |= (1 << i);
1671 
1672 		dlen = SA_SIZE(sa);
1673 		KASSERT(dlen <= sizeof(ss),
1674 		    ("%s: sockaddr size overflow", __func__));
1675 		bzero(&ss, sizeof(ss));
1676 		bcopy(sa, &ss, sa->sa_len);
1677 		sa = (struct sockaddr *)&ss;
1678 #ifdef INET6
1679 		if (sa->sa_family == AF_INET6) {
1680 			sin6 = (struct sockaddr_in6 *)sa;
1681 			(void)sa6_recoverscope(sin6);
1682 		}
1683 #endif
1684 		m_copyback(m, len, dlen, (caddr_t)sa);
1685 		len += dlen;
1686 	}
1687 	if (m->m_pkthdr.len != len) {
1688 		m_freem(m);
1689 		return (NULL);
1690 	}
1691 	rtm->rtm_msglen = len;
1692 	rtm->rtm_version = RTM_VERSION;
1693 	rtm->rtm_type = type;
1694 	return (m);
1695 }
1696 
1697 /*
1698  * Writes information related to @rtinfo object to preallocated buffer.
1699  * Stores needed size in @plen. If @w is NULL, calculates size without
1700  * writing.
1701  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1702  *
1703  * Returns 0 on success.
1704  *
1705  */
1706 static int
1707 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1708 {
1709 	struct sockaddr_storage ss;
1710 	int len, buflen = 0, dlen, i;
1711 	caddr_t cp = NULL;
1712 	struct rt_msghdr *rtm = NULL;
1713 #ifdef INET6
1714 	struct sockaddr_in6 *sin6;
1715 #endif
1716 #ifdef COMPAT_FREEBSD32
1717 	bool compat32 = false;
1718 #endif
1719 
1720 	switch (type) {
1721 	case RTM_DELADDR:
1722 	case RTM_NEWADDR:
1723 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1724 #ifdef COMPAT_FREEBSD32
1725 			if (w->w_req->flags & SCTL_MASK32) {
1726 				len = sizeof(struct ifa_msghdrl32);
1727 				compat32 = true;
1728 			} else
1729 #endif
1730 				len = sizeof(struct ifa_msghdrl);
1731 		} else
1732 			len = sizeof(struct ifa_msghdr);
1733 		break;
1734 
1735 	case RTM_IFINFO:
1736 #ifdef COMPAT_FREEBSD32
1737 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1738 			if (w->w_op == NET_RT_IFLISTL)
1739 				len = sizeof(struct if_msghdrl32);
1740 			else
1741 				len = sizeof(struct if_msghdr32);
1742 			compat32 = true;
1743 			break;
1744 		}
1745 #endif
1746 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1747 			len = sizeof(struct if_msghdrl);
1748 		else
1749 			len = sizeof(struct if_msghdr);
1750 		break;
1751 
1752 	case RTM_NEWMADDR:
1753 		len = sizeof(struct ifma_msghdr);
1754 		break;
1755 
1756 	default:
1757 		len = sizeof(struct rt_msghdr);
1758 	}
1759 
1760 	if (w != NULL) {
1761 		rtm = (struct rt_msghdr *)w->w_tmem;
1762 		buflen = w->w_tmemsize - len;
1763 		cp = (caddr_t)w->w_tmem + len;
1764 	}
1765 
1766 	rtinfo->rti_addrs = 0;
1767 	for (i = 0; i < RTAX_MAX; i++) {
1768 		struct sockaddr *sa;
1769 
1770 		if ((sa = rtinfo->rti_info[i]) == NULL)
1771 			continue;
1772 		rtinfo->rti_addrs |= (1 << i);
1773 #ifdef COMPAT_FREEBSD32
1774 		if (compat32)
1775 			dlen = SA_SIZE32(sa);
1776 		else
1777 #endif
1778 			dlen = SA_SIZE(sa);
1779 		if (cp != NULL && buflen >= dlen) {
1780 			KASSERT(dlen <= sizeof(ss),
1781 			    ("%s: sockaddr size overflow", __func__));
1782 			bzero(&ss, sizeof(ss));
1783 			bcopy(sa, &ss, sa->sa_len);
1784 			sa = (struct sockaddr *)&ss;
1785 #ifdef INET6
1786 			if (sa->sa_family == AF_INET6) {
1787 				sin6 = (struct sockaddr_in6 *)sa;
1788 				(void)sa6_recoverscope(sin6);
1789 			}
1790 #endif
1791 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1792 			cp += dlen;
1793 			buflen -= dlen;
1794 		} else if (cp != NULL) {
1795 			/*
1796 			 * Buffer too small. Count needed size
1797 			 * and return with error.
1798 			 */
1799 			cp = NULL;
1800 		}
1801 
1802 		len += dlen;
1803 	}
1804 
1805 	if (cp != NULL) {
1806 		dlen = ALIGN(len) - len;
1807 		if (buflen < dlen)
1808 			cp = NULL;
1809 		else {
1810 			bzero(cp, dlen);
1811 			cp += dlen;
1812 			buflen -= dlen;
1813 		}
1814 	}
1815 	len = ALIGN(len);
1816 
1817 	if (cp != NULL) {
1818 		/* fill header iff buffer is large enough */
1819 		rtm->rtm_version = RTM_VERSION;
1820 		rtm->rtm_type = type;
1821 		rtm->rtm_msglen = len;
1822 	}
1823 
1824 	*plen = len;
1825 
1826 	if (w != NULL && cp == NULL)
1827 		return (ENOBUFS);
1828 
1829 	return (0);
1830 }
1831 
1832 /*
1833  * This routine is called to generate a message from the routing
1834  * socket indicating that a redirect has occurred, a routing lookup
1835  * has failed, or that a protocol has detected timeouts to a particular
1836  * destination.
1837  */
1838 void
1839 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1840     int fibnum)
1841 {
1842 	struct rt_msghdr *rtm;
1843 	struct mbuf *m;
1844 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1845 
1846 	if (V_route_cb.any_count == 0)
1847 		return;
1848 	m = rtsock_msg_mbuf(type, rtinfo);
1849 	if (m == NULL)
1850 		return;
1851 
1852 	if (fibnum != RT_ALL_FIBS) {
1853 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1854 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1855 		M_SETFIB(m, fibnum);
1856 		m->m_flags |= RTS_FILTER_FIB;
1857 	}
1858 
1859 	rtm = mtod(m, struct rt_msghdr *);
1860 	rtm->rtm_flags = RTF_DONE | flags;
1861 	rtm->rtm_errno = error;
1862 	rtm->rtm_addrs = rtinfo->rti_addrs;
1863 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1864 }
1865 
1866 void
1867 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1868 {
1869 
1870 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1871 }
1872 
1873 /*
1874  * This routine is called to generate a message from the routing
1875  * socket indicating that the status of a network interface has changed.
1876  */
1877 void
1878 rt_ifmsg(struct ifnet *ifp)
1879 {
1880 	struct if_msghdr *ifm;
1881 	struct mbuf *m;
1882 	struct rt_addrinfo info;
1883 
1884 	if (V_route_cb.any_count == 0)
1885 		return;
1886 	bzero((caddr_t)&info, sizeof(info));
1887 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1888 	if (m == NULL)
1889 		return;
1890 	ifm = mtod(m, struct if_msghdr *);
1891 	ifm->ifm_index = ifp->if_index;
1892 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1893 	if_data_copy(ifp, &ifm->ifm_data);
1894 	ifm->ifm_addrs = 0;
1895 	rt_dispatch(m, AF_UNSPEC);
1896 }
1897 
1898 /*
1899  * Announce interface address arrival/withdraw.
1900  * Please do not call directly, use rt_addrmsg().
1901  * Assume input data to be valid.
1902  * Returns 0 on success.
1903  */
1904 int
1905 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1906 {
1907 	struct rt_addrinfo info;
1908 	struct sockaddr *sa;
1909 	int ncmd;
1910 	struct mbuf *m;
1911 	struct ifa_msghdr *ifam;
1912 	struct ifnet *ifp = ifa->ifa_ifp;
1913 	struct sockaddr_storage ss;
1914 
1915 	if (V_route_cb.any_count == 0)
1916 		return (0);
1917 
1918 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1919 
1920 	bzero((caddr_t)&info, sizeof(info));
1921 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1922 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1923 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1924 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1925 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1926 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1927 		return (ENOBUFS);
1928 	ifam = mtod(m, struct ifa_msghdr *);
1929 	ifam->ifam_index = ifp->if_index;
1930 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1931 	ifam->ifam_flags = ifa->ifa_flags;
1932 	ifam->ifam_addrs = info.rti_addrs;
1933 
1934 	if (fibnum != RT_ALL_FIBS) {
1935 		M_SETFIB(m, fibnum);
1936 		m->m_flags |= RTS_FILTER_FIB;
1937 	}
1938 
1939 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1940 
1941 	return (0);
1942 }
1943 
1944 /*
1945  * Announce route addition/removal to rtsock based on @rt data.
1946  * Callers are advives to use rt_routemsg() instead of using this
1947  *  function directly.
1948  * Assume @rt data is consistent.
1949  *
1950  * Returns 0 on success.
1951  */
1952 int
1953 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
1954     int fibnum)
1955 {
1956 	union sockaddr_union dst, mask;
1957 	struct rt_addrinfo info;
1958 
1959 	if (V_route_cb.any_count == 0)
1960 		return (0);
1961 
1962 	int family = rt_get_family(rt);
1963 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
1964 	export_rtaddrs(rt, &dst.sa, &mask.sa);
1965 
1966 	bzero((caddr_t)&info, sizeof(info));
1967 	info.rti_info[RTAX_DST] = &dst.sa;
1968 	info.rti_info[RTAX_NETMASK] = &mask.sa;
1969 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1970 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
1971 	info.rti_ifp = nh->nh_ifp;
1972 
1973 	return (rtsock_routemsg_info(cmd, &info, fibnum));
1974 }
1975 
1976 int
1977 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1978 {
1979 	struct rt_msghdr *rtm;
1980 	struct sockaddr *sa;
1981 	struct mbuf *m;
1982 
1983 	if (V_route_cb.any_count == 0)
1984 		return (0);
1985 
1986 	if (info->rti_flags & RTF_HOST)
1987 		info->rti_info[RTAX_NETMASK] = NULL;
1988 
1989 	m = rtsock_msg_mbuf(cmd, info);
1990 	if (m == NULL)
1991 		return (ENOBUFS);
1992 
1993 	if (fibnum != RT_ALL_FIBS) {
1994 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1995 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1996 		M_SETFIB(m, fibnum);
1997 		m->m_flags |= RTS_FILTER_FIB;
1998 	}
1999 
2000 	rtm = mtod(m, struct rt_msghdr *);
2001 	rtm->rtm_addrs = info->rti_addrs;
2002 	if (info->rti_ifp != NULL)
2003 		rtm->rtm_index = info->rti_ifp->if_index;
2004 	/* Add RTF_DONE to indicate command 'completion' required by API */
2005 	info->rti_flags |= RTF_DONE;
2006 	/* Reported routes has to be up */
2007 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
2008 		info->rti_flags |= RTF_UP;
2009 	rtm->rtm_flags = info->rti_flags;
2010 
2011 	sa = info->rti_info[RTAX_DST];
2012 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
2013 
2014 	return (0);
2015 }
2016 
2017 /*
2018  * This is the analogue to the rt_newaddrmsg which performs the same
2019  * function but for multicast group memberhips.  This is easier since
2020  * there is no route state to worry about.
2021  */
2022 void
2023 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
2024 {
2025 	struct rt_addrinfo info;
2026 	struct mbuf *m = NULL;
2027 	struct ifnet *ifp = ifma->ifma_ifp;
2028 	struct ifma_msghdr *ifmam;
2029 
2030 	if (V_route_cb.any_count == 0)
2031 		return;
2032 
2033 	bzero((caddr_t)&info, sizeof(info));
2034 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2035 	if (ifp && ifp->if_addr)
2036 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
2037 	else
2038 		info.rti_info[RTAX_IFP] = NULL;
2039 	/*
2040 	 * If a link-layer address is present, present it as a ``gateway''
2041 	 * (similarly to how ARP entries, e.g., are presented).
2042 	 */
2043 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
2044 	m = rtsock_msg_mbuf(cmd, &info);
2045 	if (m == NULL)
2046 		return;
2047 	ifmam = mtod(m, struct ifma_msghdr *);
2048 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
2049 	    __func__));
2050 	ifmam->ifmam_index = ifp->if_index;
2051 	ifmam->ifmam_addrs = info.rti_addrs;
2052 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
2053 }
2054 
2055 static struct mbuf *
2056 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
2057 	struct rt_addrinfo *info)
2058 {
2059 	struct if_announcemsghdr *ifan;
2060 	struct mbuf *m;
2061 
2062 	if (V_route_cb.any_count == 0)
2063 		return NULL;
2064 	bzero((caddr_t)info, sizeof(*info));
2065 	m = rtsock_msg_mbuf(type, info);
2066 	if (m != NULL) {
2067 		ifan = mtod(m, struct if_announcemsghdr *);
2068 		ifan->ifan_index = ifp->if_index;
2069 		strlcpy(ifan->ifan_name, ifp->if_xname,
2070 			sizeof(ifan->ifan_name));
2071 		ifan->ifan_what = what;
2072 	}
2073 	return m;
2074 }
2075 
2076 /*
2077  * This is called to generate routing socket messages indicating
2078  * IEEE80211 wireless events.
2079  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
2080  */
2081 void
2082 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
2083 {
2084 	struct mbuf *m;
2085 	struct rt_addrinfo info;
2086 
2087 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
2088 	if (m != NULL) {
2089 		/*
2090 		 * Append the ieee80211 data.  Try to stick it in the
2091 		 * mbuf containing the ifannounce msg; otherwise allocate
2092 		 * a new mbuf and append.
2093 		 *
2094 		 * NB: we assume m is a single mbuf.
2095 		 */
2096 		if (data_len > M_TRAILINGSPACE(m)) {
2097 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
2098 			if (n == NULL) {
2099 				m_freem(m);
2100 				return;
2101 			}
2102 			bcopy(data, mtod(n, void *), data_len);
2103 			n->m_len = data_len;
2104 			m->m_next = n;
2105 		} else if (data_len > 0) {
2106 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
2107 			m->m_len += data_len;
2108 		}
2109 		if (m->m_flags & M_PKTHDR)
2110 			m->m_pkthdr.len += data_len;
2111 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
2112 		rt_dispatch(m, AF_UNSPEC);
2113 	}
2114 }
2115 
2116 /*
2117  * This is called to generate routing socket messages indicating
2118  * network interface arrival and departure.
2119  */
2120 static void
2121 rt_ifannouncemsg(struct ifnet *ifp, int what)
2122 {
2123 	struct mbuf *m;
2124 	struct rt_addrinfo info;
2125 
2126 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
2127 	if (m != NULL)
2128 		rt_dispatch(m, AF_UNSPEC);
2129 }
2130 
2131 static void
2132 rt_dispatch(struct mbuf *m, sa_family_t saf)
2133 {
2134 
2135 	M_ASSERTPKTHDR(m);
2136 
2137 	m->m_rtsock_family = saf;
2138 	if (V_loif)
2139 		m->m_pkthdr.rcvif = V_loif;
2140 	else {
2141 		m_freem(m);
2142 		return;
2143 	}
2144 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
2145 }
2146 
2147 /*
2148  * Checks if rte can be exported w.r.t jails/vnets.
2149  *
2150  * Returns true if it can, false otherwise.
2151  */
2152 static bool
2153 can_export_rte(struct ucred *td_ucred, bool rt_is_host,
2154     const struct sockaddr *rt_dst)
2155 {
2156 
2157 	if ((!rt_is_host) ? jailed_without_vnet(td_ucred)
2158 	    : prison_if(td_ucred, rt_dst) != 0)
2159 		return (false);
2160 	return (true);
2161 }
2162 
2163 
2164 /*
2165  * This is used in dumping the kernel table via sysctl().
2166  */
2167 static int
2168 sysctl_dumpentry(struct rtentry *rt, void *vw)
2169 {
2170 	struct walkarg *w = vw;
2171 	struct nhop_object *nh;
2172 
2173 	NET_EPOCH_ASSERT();
2174 
2175 	export_rtaddrs(rt, w->dst, w->mask);
2176 	if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst))
2177 		return (0);
2178 	nh = rt_get_raw_nhop(rt);
2179 #ifdef ROUTE_MPATH
2180 	if (NH_IS_NHGRP(nh)) {
2181 		const struct weightened_nhop *wn;
2182 		uint32_t num_nhops;
2183 		int error;
2184 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
2185 		for (int i = 0; i < num_nhops; i++) {
2186 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
2187 			if (error != 0)
2188 				return (error);
2189 		}
2190 	} else
2191 #endif
2192 		sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
2193 
2194 	return (0);
2195 }
2196 
2197 
2198 static int
2199 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
2200     struct walkarg *w)
2201 {
2202 	struct rt_addrinfo info;
2203 	int error = 0, size;
2204 	uint32_t rtflags;
2205 
2206 	rtflags = nhop_get_rtflags(nh);
2207 
2208 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
2209 		return (0);
2210 
2211 	bzero((caddr_t)&info, sizeof(info));
2212 	info.rti_info[RTAX_DST] = w->dst;
2213 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
2214 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
2215 	info.rti_info[RTAX_GENMASK] = 0;
2216 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
2217 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
2218 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
2219 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
2220 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
2221 	}
2222 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
2223 		return (error);
2224 	if (w->w_req && w->w_tmem) {
2225 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
2226 
2227 		bzero(&rtm->rtm_index,
2228 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
2229 
2230 		/*
2231 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
2232 		 * and RTF_UP (if entry is linked, which is always true here).
2233 		 * Given that, use nhop rtflags & add RTF_UP.
2234 		 */
2235 		rtm->rtm_flags = rtflags | RTF_UP;
2236 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
2237 			rtm->rtm_flags = RTF_GATEWAY |
2238 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
2239 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
2240 		rtm->rtm_rmx.rmx_weight = weight;
2241 		rtm->rtm_index = nh->nh_ifp->if_index;
2242 		rtm->rtm_addrs = info.rti_addrs;
2243 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
2244 		return (error);
2245 	}
2246 	return (error);
2247 }
2248 
2249 static int
2250 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
2251     struct rt_addrinfo *info, struct walkarg *w, int len)
2252 {
2253 	struct if_msghdrl *ifm;
2254 	struct if_data *ifd;
2255 
2256 	ifm = (struct if_msghdrl *)w->w_tmem;
2257 
2258 #ifdef COMPAT_FREEBSD32
2259 	if (w->w_req->flags & SCTL_MASK32) {
2260 		struct if_msghdrl32 *ifm32;
2261 
2262 		ifm32 = (struct if_msghdrl32 *)ifm;
2263 		ifm32->ifm_addrs = info->rti_addrs;
2264 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2265 		ifm32->ifm_index = ifp->if_index;
2266 		ifm32->_ifm_spare1 = 0;
2267 		ifm32->ifm_len = sizeof(*ifm32);
2268 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
2269 		ifm32->_ifm_spare2 = 0;
2270 		ifd = &ifm32->ifm_data;
2271 	} else
2272 #endif
2273 	{
2274 		ifm->ifm_addrs = info->rti_addrs;
2275 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2276 		ifm->ifm_index = ifp->if_index;
2277 		ifm->_ifm_spare1 = 0;
2278 		ifm->ifm_len = sizeof(*ifm);
2279 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
2280 		ifm->_ifm_spare2 = 0;
2281 		ifd = &ifm->ifm_data;
2282 	}
2283 
2284 	memcpy(ifd, src_ifd, sizeof(*ifd));
2285 
2286 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2287 }
2288 
2289 static int
2290 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2291     struct rt_addrinfo *info, struct walkarg *w, int len)
2292 {
2293 	struct if_msghdr *ifm;
2294 	struct if_data *ifd;
2295 
2296 	ifm = (struct if_msghdr *)w->w_tmem;
2297 
2298 #ifdef COMPAT_FREEBSD32
2299 	if (w->w_req->flags & SCTL_MASK32) {
2300 		struct if_msghdr32 *ifm32;
2301 
2302 		ifm32 = (struct if_msghdr32 *)ifm;
2303 		ifm32->ifm_addrs = info->rti_addrs;
2304 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2305 		ifm32->ifm_index = ifp->if_index;
2306 		ifm32->_ifm_spare1 = 0;
2307 		ifd = &ifm32->ifm_data;
2308 	} else
2309 #endif
2310 	{
2311 		ifm->ifm_addrs = info->rti_addrs;
2312 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2313 		ifm->ifm_index = ifp->if_index;
2314 		ifm->_ifm_spare1 = 0;
2315 		ifd = &ifm->ifm_data;
2316 	}
2317 
2318 	memcpy(ifd, src_ifd, sizeof(*ifd));
2319 
2320 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2321 }
2322 
2323 static int
2324 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2325     struct walkarg *w, int len)
2326 {
2327 	struct ifa_msghdrl *ifam;
2328 	struct if_data *ifd;
2329 
2330 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2331 
2332 #ifdef COMPAT_FREEBSD32
2333 	if (w->w_req->flags & SCTL_MASK32) {
2334 		struct ifa_msghdrl32 *ifam32;
2335 
2336 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2337 		ifam32->ifam_addrs = info->rti_addrs;
2338 		ifam32->ifam_flags = ifa->ifa_flags;
2339 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2340 		ifam32->_ifam_spare1 = 0;
2341 		ifam32->ifam_len = sizeof(*ifam32);
2342 		ifam32->ifam_data_off =
2343 		    offsetof(struct ifa_msghdrl32, ifam_data);
2344 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2345 		ifd = &ifam32->ifam_data;
2346 	} else
2347 #endif
2348 	{
2349 		ifam->ifam_addrs = info->rti_addrs;
2350 		ifam->ifam_flags = ifa->ifa_flags;
2351 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2352 		ifam->_ifam_spare1 = 0;
2353 		ifam->ifam_len = sizeof(*ifam);
2354 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2355 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2356 		ifd = &ifam->ifam_data;
2357 	}
2358 
2359 	bzero(ifd, sizeof(*ifd));
2360 	ifd->ifi_datalen = sizeof(struct if_data);
2361 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2362 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2363 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2364 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2365 
2366 	/* Fixup if_data carp(4) vhid. */
2367 	if (carp_get_vhid_p != NULL)
2368 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2369 
2370 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2371 }
2372 
2373 static int
2374 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2375     struct walkarg *w, int len)
2376 {
2377 	struct ifa_msghdr *ifam;
2378 
2379 	ifam = (struct ifa_msghdr *)w->w_tmem;
2380 	ifam->ifam_addrs = info->rti_addrs;
2381 	ifam->ifam_flags = ifa->ifa_flags;
2382 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2383 	ifam->_ifam_spare1 = 0;
2384 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2385 
2386 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2387 }
2388 
2389 static int
2390 sysctl_iflist(int af, struct walkarg *w)
2391 {
2392 	struct ifnet *ifp;
2393 	struct ifaddr *ifa;
2394 	struct if_data ifd;
2395 	struct rt_addrinfo info;
2396 	int len, error = 0;
2397 	struct sockaddr_storage ss;
2398 
2399 	bzero((caddr_t)&info, sizeof(info));
2400 	bzero(&ifd, sizeof(ifd));
2401 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2402 		if (w->w_arg && w->w_arg != ifp->if_index)
2403 			continue;
2404 		if_data_copy(ifp, &ifd);
2405 		ifa = ifp->if_addr;
2406 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2407 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2408 		if (error != 0)
2409 			goto done;
2410 		info.rti_info[RTAX_IFP] = NULL;
2411 		if (w->w_req && w->w_tmem) {
2412 			if (w->w_op == NET_RT_IFLISTL)
2413 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2414 				    len);
2415 			else
2416 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2417 				    len);
2418 			if (error)
2419 				goto done;
2420 		}
2421 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2422 			if (af && af != ifa->ifa_addr->sa_family)
2423 				continue;
2424 			if (prison_if(w->w_req->td->td_ucred,
2425 			    ifa->ifa_addr) != 0)
2426 				continue;
2427 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2428 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2429 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2430 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2431 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2432 			if (error != 0)
2433 				goto done;
2434 			if (w->w_req && w->w_tmem) {
2435 				if (w->w_op == NET_RT_IFLISTL)
2436 					error = sysctl_iflist_ifaml(ifa, &info,
2437 					    w, len);
2438 				else
2439 					error = sysctl_iflist_ifam(ifa, &info,
2440 					    w, len);
2441 				if (error)
2442 					goto done;
2443 			}
2444 		}
2445 		info.rti_info[RTAX_IFA] = NULL;
2446 		info.rti_info[RTAX_NETMASK] = NULL;
2447 		info.rti_info[RTAX_BRD] = NULL;
2448 	}
2449 done:
2450 	return (error);
2451 }
2452 
2453 static int
2454 sysctl_ifmalist(int af, struct walkarg *w)
2455 {
2456 	struct rt_addrinfo info;
2457 	struct ifaddr *ifa;
2458 	struct ifmultiaddr *ifma;
2459 	struct ifnet *ifp;
2460 	int error, len;
2461 
2462 	NET_EPOCH_ASSERT();
2463 
2464 	error = 0;
2465 	bzero((caddr_t)&info, sizeof(info));
2466 
2467 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2468 		if (w->w_arg && w->w_arg != ifp->if_index)
2469 			continue;
2470 		ifa = ifp->if_addr;
2471 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2472 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2473 			if (af && af != ifma->ifma_addr->sa_family)
2474 				continue;
2475 			if (prison_if(w->w_req->td->td_ucred,
2476 			    ifma->ifma_addr) != 0)
2477 				continue;
2478 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2479 			info.rti_info[RTAX_GATEWAY] =
2480 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2481 			    ifma->ifma_lladdr : NULL;
2482 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2483 			if (error != 0)
2484 				break;
2485 			if (w->w_req && w->w_tmem) {
2486 				struct ifma_msghdr *ifmam;
2487 
2488 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2489 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2490 				ifmam->ifmam_flags = 0;
2491 				ifmam->ifmam_addrs = info.rti_addrs;
2492 				ifmam->_ifmam_spare1 = 0;
2493 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2494 				if (error != 0)
2495 					break;
2496 			}
2497 		}
2498 		if (error != 0)
2499 			break;
2500 	}
2501 	return (error);
2502 }
2503 
2504 static void
2505 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2506 {
2507 	union sockaddr_union sa_dst, sa_mask;
2508 
2509 	w->family = family;
2510 	w->dst = (struct sockaddr *)&sa_dst;
2511 	w->mask = (struct sockaddr *)&sa_mask;
2512 
2513 	init_sockaddrs_family(family, w->dst, w->mask);
2514 
2515 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2516 }
2517 
2518 static int
2519 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2520 {
2521 	struct epoch_tracker et;
2522 	int	*name = (int *)arg1;
2523 	u_int	namelen = arg2;
2524 	struct rib_head *rnh = NULL; /* silence compiler. */
2525 	int	i, lim, error = EINVAL;
2526 	int	fib = 0;
2527 	u_char	af;
2528 	struct	walkarg w;
2529 
2530 	if (namelen < 3)
2531 		return (EINVAL);
2532 
2533 	name++;
2534 	namelen--;
2535 	if (req->newptr)
2536 		return (EPERM);
2537 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2538 		if (namelen == 3)
2539 			fib = req->td->td_proc->p_fibnum;
2540 		else if (namelen == 4)
2541 			fib = (name[3] == RT_ALL_FIBS) ?
2542 			    req->td->td_proc->p_fibnum : name[3];
2543 		else
2544 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2545 		if (fib < 0 || fib >= rt_numfibs)
2546 			return (EINVAL);
2547 	} else if (namelen != 3)
2548 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2549 	af = name[0];
2550 	if (af > AF_MAX)
2551 		return (EINVAL);
2552 	bzero(&w, sizeof(w));
2553 	w.w_op = name[1];
2554 	w.w_arg = name[2];
2555 	w.w_req = req;
2556 
2557 	error = sysctl_wire_old_buffer(req, 0);
2558 	if (error)
2559 		return (error);
2560 
2561 	/*
2562 	 * Allocate reply buffer in advance.
2563 	 * All rtsock messages has maximum length of u_short.
2564 	 */
2565 	w.w_tmemsize = 65536;
2566 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2567 
2568 	NET_EPOCH_ENTER(et);
2569 	switch (w.w_op) {
2570 	case NET_RT_DUMP:
2571 	case NET_RT_FLAGS:
2572 		if (af == 0) {			/* dump all tables */
2573 			i = 1;
2574 			lim = AF_MAX;
2575 		} else				/* dump only one table */
2576 			i = lim = af;
2577 
2578 		/*
2579 		 * take care of llinfo entries, the caller must
2580 		 * specify an AF
2581 		 */
2582 		if (w.w_op == NET_RT_FLAGS &&
2583 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2584 			if (af != 0)
2585 				error = lltable_sysctl_dumparp(af, w.w_req);
2586 			else
2587 				error = EINVAL;
2588 			break;
2589 		}
2590 		/*
2591 		 * take care of routing entries
2592 		 */
2593 		for (error = 0; error == 0 && i <= lim; i++) {
2594 			rnh = rt_tables_get_rnh(fib, i);
2595 			if (rnh != NULL) {
2596 				rtable_sysctl_dump(fib, i, &w);
2597 			} else if (af != 0)
2598 				error = EAFNOSUPPORT;
2599 		}
2600 		break;
2601 	case NET_RT_NHOP:
2602 	case NET_RT_NHGRP:
2603 		/* Allow dumping one specific af/fib at a time */
2604 		if (namelen < 4) {
2605 			error = EINVAL;
2606 			break;
2607 		}
2608 		fib = name[3];
2609 		if (fib < 0 || fib > rt_numfibs) {
2610 			error = EINVAL;
2611 			break;
2612 		}
2613 		rnh = rt_tables_get_rnh(fib, af);
2614 		if (rnh == NULL) {
2615 			error = EAFNOSUPPORT;
2616 			break;
2617 		}
2618 		if (w.w_op == NET_RT_NHOP)
2619 			error = nhops_dump_sysctl(rnh, w.w_req);
2620 		else
2621 #ifdef ROUTE_MPATH
2622 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2623 #else
2624 			error = ENOTSUP;
2625 #endif
2626 		break;
2627 	case NET_RT_IFLIST:
2628 	case NET_RT_IFLISTL:
2629 		error = sysctl_iflist(af, &w);
2630 		break;
2631 
2632 	case NET_RT_IFMALIST:
2633 		error = sysctl_ifmalist(af, &w);
2634 		break;
2635 	}
2636 	NET_EPOCH_EXIT(et);
2637 
2638 	free(w.w_tmem, M_TEMP);
2639 	return (error);
2640 }
2641 
2642 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2643     sysctl_rtsock, "Return route tables and interface/address lists");
2644 
2645 /*
2646  * Definitions of protocols supported in the ROUTE domain.
2647  */
2648 
2649 static struct domain routedomain;		/* or at least forward */
2650 
2651 static struct protosw routesw = {
2652 	.pr_type =		SOCK_RAW,
2653 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2654 	.pr_abort =		rts_close,
2655 	.pr_attach =		rts_attach,
2656 	.pr_detach =		rts_detach,
2657 	.pr_send =		rts_send,
2658 	.pr_shutdown =		rts_shutdown,
2659 	.pr_close =		rts_close,
2660 };
2661 
2662 static struct domain routedomain = {
2663 	.dom_family =		PF_ROUTE,
2664 	.dom_name =		"route",
2665 	.dom_nprotosw =		1,
2666 	.dom_protosw =		{ &routesw },
2667 };
2668 
2669 DOMAIN_SET(route);
2670