xref: /freebsd/sys/net/rtsock.c (revision dba7640e44c5ec148a84b0d58c6c9a3c9e5147f3)
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/ip6_var.h>
74 #include <netinet6/scope6_var.h>
75 #endif
76 #include <net/route/nhop.h>
77 
78 #ifdef COMPAT_FREEBSD32
79 #include <sys/mount.h>
80 #include <compat/freebsd32/freebsd32.h>
81 
82 struct if_msghdr32 {
83 	uint16_t ifm_msglen;
84 	uint8_t	ifm_version;
85 	uint8_t	ifm_type;
86 	int32_t	ifm_addrs;
87 	int32_t	ifm_flags;
88 	uint16_t ifm_index;
89 	uint16_t _ifm_spare1;
90 	struct	if_data ifm_data;
91 };
92 
93 struct if_msghdrl32 {
94 	uint16_t ifm_msglen;
95 	uint8_t	ifm_version;
96 	uint8_t	ifm_type;
97 	int32_t	ifm_addrs;
98 	int32_t	ifm_flags;
99 	uint16_t ifm_index;
100 	uint16_t _ifm_spare1;
101 	uint16_t ifm_len;
102 	uint16_t ifm_data_off;
103 	uint32_t _ifm_spare2;
104 	struct	if_data ifm_data;
105 };
106 
107 struct ifa_msghdrl32 {
108 	uint16_t ifam_msglen;
109 	uint8_t	ifam_version;
110 	uint8_t	ifam_type;
111 	int32_t	ifam_addrs;
112 	int32_t	ifam_flags;
113 	uint16_t ifam_index;
114 	uint16_t _ifam_spare1;
115 	uint16_t ifam_len;
116 	uint16_t ifam_data_off;
117 	int32_t	ifam_metric;
118 	struct	if_data ifam_data;
119 };
120 
121 #define SA_SIZE32(sa)						\
122     (  (((struct sockaddr *)(sa))->sa_len == 0) ?		\
123 	sizeof(int)		:				\
124 	1 + ( (((struct sockaddr *)(sa))->sa_len - 1) | (sizeof(int) - 1) ) )
125 
126 #endif /* COMPAT_FREEBSD32 */
127 
128 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
129 
130 /* NB: these are not modified */
131 static struct	sockaddr route_src = { 2, PF_ROUTE, };
132 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
133 
134 /* These are external hooks for CARP. */
135 int	(*carp_get_vhid_p)(struct ifaddr *);
136 
137 /*
138  * Used by rtsock/raw_input callback code to decide whether to filter the update
139  * notification to a socket bound to a particular FIB.
140  */
141 #define	RTS_FILTER_FIB	M_PROTO8
142 
143 typedef struct {
144 	int	ip_count;	/* attached w/ AF_INET */
145 	int	ip6_count;	/* attached w/ AF_INET6 */
146 	int	any_count;	/* total attached */
147 } route_cb_t;
148 VNET_DEFINE_STATIC(route_cb_t, route_cb);
149 #define	V_route_cb VNET(route_cb)
150 
151 struct mtx rtsock_mtx;
152 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
153 
154 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
155 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
156 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
157 
158 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "");
159 
160 struct walkarg {
161 	int	w_tmemsize;
162 	int	w_op, w_arg;
163 	caddr_t	w_tmem;
164 	struct sysctl_req *w_req;
165 };
166 
167 static void	rts_input(struct mbuf *m);
168 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
169 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
170 			struct walkarg *w, int *plen);
171 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
172 			struct rt_addrinfo *rtinfo);
173 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
174 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
175 			uint32_t weight, struct walkarg *w);
176 static int	sysctl_iflist(int af, struct walkarg *w);
177 static int	sysctl_ifmalist(int af, struct walkarg *w);
178 static int	route_output(struct mbuf *m, struct socket *so, ...);
179 static void	rt_getmetrics(const struct rtentry *rt,
180 			const struct nhop_object *nh, struct rt_metrics *out);
181 static void	rt_dispatch(struct mbuf *, sa_family_t);
182 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
183 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
184 static int	update_rtm_from_rc(struct rt_addrinfo *info,
185 			struct rt_msghdr **prtm, int alloc_len,
186 			struct rib_cmd_info *rc, struct nhop_object *nh);
187 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
188 			struct mbuf *m, sa_family_t saf, u_int fibnum,
189 			int rtm_errno);
190 static int	can_export_rte(struct ucred *td_ucred, const struct rtentry *rt);
191 
192 static struct netisr_handler rtsock_nh = {
193 	.nh_name = "rtsock",
194 	.nh_handler = rts_input,
195 	.nh_proto = NETISR_ROUTE,
196 	.nh_policy = NETISR_POLICY_SOURCE,
197 };
198 
199 static int
200 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
201 {
202 	int error, qlimit;
203 
204 	netisr_getqlimit(&rtsock_nh, &qlimit);
205 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
206         if (error || !req->newptr)
207                 return (error);
208 	if (qlimit < 1)
209 		return (EINVAL);
210 	return (netisr_setqlimit(&rtsock_nh, qlimit));
211 }
212 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
213     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
214     0, 0, sysctl_route_netisr_maxqlen, "I",
215     "maximum routing socket dispatch queue length");
216 
217 static void
218 vnet_rts_init(void)
219 {
220 	int tmp;
221 
222 	if (IS_DEFAULT_VNET(curvnet)) {
223 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
224 			rtsock_nh.nh_qlimit = tmp;
225 		netisr_register(&rtsock_nh);
226 	}
227 #ifdef VIMAGE
228 	 else
229 		netisr_register_vnet(&rtsock_nh);
230 #endif
231 }
232 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
233     vnet_rts_init, 0);
234 
235 #ifdef VIMAGE
236 static void
237 vnet_rts_uninit(void)
238 {
239 
240 	netisr_unregister_vnet(&rtsock_nh);
241 }
242 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
243     vnet_rts_uninit, 0);
244 #endif
245 
246 static int
247 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
248     struct rawcb *rp)
249 {
250 	int fibnum;
251 
252 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
253 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
254 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
255 
256 	/* No filtering requested. */
257 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
258 		return (0);
259 
260 	/* Check if it is a rts and the fib matches the one of the socket. */
261 	fibnum = M_GETFIB(m);
262 	if (proto->sp_family != PF_ROUTE ||
263 	    rp->rcb_socket == NULL ||
264 	    rp->rcb_socket->so_fibnum == fibnum)
265 		return (0);
266 
267 	/* Filtering requested and no match, the socket shall be skipped. */
268 	return (1);
269 }
270 
271 static void
272 rts_input(struct mbuf *m)
273 {
274 	struct sockproto route_proto;
275 	unsigned short *family;
276 	struct m_tag *tag;
277 
278 	route_proto.sp_family = PF_ROUTE;
279 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
280 	if (tag != NULL) {
281 		family = (unsigned short *)(tag + 1);
282 		route_proto.sp_protocol = *family;
283 		m_tag_delete(m, tag);
284 	} else
285 		route_proto.sp_protocol = 0;
286 
287 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
288 }
289 
290 /*
291  * It really doesn't make any sense at all for this code to share much
292  * with raw_usrreq.c, since its functionality is so restricted.  XXX
293  */
294 static void
295 rts_abort(struct socket *so)
296 {
297 
298 	raw_usrreqs.pru_abort(so);
299 }
300 
301 static void
302 rts_close(struct socket *so)
303 {
304 
305 	raw_usrreqs.pru_close(so);
306 }
307 
308 /* pru_accept is EOPNOTSUPP */
309 
310 static int
311 rts_attach(struct socket *so, int proto, struct thread *td)
312 {
313 	struct rawcb *rp;
314 	int error;
315 
316 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
317 
318 	/* XXX */
319 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
320 
321 	so->so_pcb = (caddr_t)rp;
322 	so->so_fibnum = td->td_proc->p_fibnum;
323 	error = raw_attach(so, proto);
324 	rp = sotorawcb(so);
325 	if (error) {
326 		so->so_pcb = NULL;
327 		free(rp, M_PCB);
328 		return error;
329 	}
330 	RTSOCK_LOCK();
331 	switch(rp->rcb_proto.sp_protocol) {
332 	case AF_INET:
333 		V_route_cb.ip_count++;
334 		break;
335 	case AF_INET6:
336 		V_route_cb.ip6_count++;
337 		break;
338 	}
339 	V_route_cb.any_count++;
340 	RTSOCK_UNLOCK();
341 	soisconnected(so);
342 	so->so_options |= SO_USELOOPBACK;
343 	return 0;
344 }
345 
346 static int
347 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
348 {
349 
350 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
351 }
352 
353 static int
354 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
355 {
356 
357 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
358 }
359 
360 /* pru_connect2 is EOPNOTSUPP */
361 /* pru_control is EOPNOTSUPP */
362 
363 static void
364 rts_detach(struct socket *so)
365 {
366 	struct rawcb *rp = sotorawcb(so);
367 
368 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
369 
370 	RTSOCK_LOCK();
371 	switch(rp->rcb_proto.sp_protocol) {
372 	case AF_INET:
373 		V_route_cb.ip_count--;
374 		break;
375 	case AF_INET6:
376 		V_route_cb.ip6_count--;
377 		break;
378 	}
379 	V_route_cb.any_count--;
380 	RTSOCK_UNLOCK();
381 	raw_usrreqs.pru_detach(so);
382 }
383 
384 static int
385 rts_disconnect(struct socket *so)
386 {
387 
388 	return (raw_usrreqs.pru_disconnect(so));
389 }
390 
391 /* pru_listen is EOPNOTSUPP */
392 
393 static int
394 rts_peeraddr(struct socket *so, struct sockaddr **nam)
395 {
396 
397 	return (raw_usrreqs.pru_peeraddr(so, nam));
398 }
399 
400 /* pru_rcvd is EOPNOTSUPP */
401 /* pru_rcvoob is EOPNOTSUPP */
402 
403 static int
404 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
405 	 struct mbuf *control, struct thread *td)
406 {
407 
408 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
409 }
410 
411 /* pru_sense is null */
412 
413 static int
414 rts_shutdown(struct socket *so)
415 {
416 
417 	return (raw_usrreqs.pru_shutdown(so));
418 }
419 
420 static int
421 rts_sockaddr(struct socket *so, struct sockaddr **nam)
422 {
423 
424 	return (raw_usrreqs.pru_sockaddr(so, nam));
425 }
426 
427 static struct pr_usrreqs route_usrreqs = {
428 	.pru_abort =		rts_abort,
429 	.pru_attach =		rts_attach,
430 	.pru_bind =		rts_bind,
431 	.pru_connect =		rts_connect,
432 	.pru_detach =		rts_detach,
433 	.pru_disconnect =	rts_disconnect,
434 	.pru_peeraddr =		rts_peeraddr,
435 	.pru_send =		rts_send,
436 	.pru_shutdown =		rts_shutdown,
437 	.pru_sockaddr =		rts_sockaddr,
438 	.pru_close =		rts_close,
439 };
440 
441 #ifndef _SOCKADDR_UNION_DEFINED
442 #define	_SOCKADDR_UNION_DEFINED
443 /*
444  * The union of all possible address formats we handle.
445  */
446 union sockaddr_union {
447 	struct sockaddr		sa;
448 	struct sockaddr_in	sin;
449 	struct sockaddr_in6	sin6;
450 };
451 #endif /* _SOCKADDR_UNION_DEFINED */
452 
453 static int
454 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
455     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
456 {
457 #if defined(INET) || defined(INET6)
458 	struct epoch_tracker et;
459 #endif
460 
461 	/* First, see if the returned address is part of the jail. */
462 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
463 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
464 		return (0);
465 	}
466 
467 	switch (info->rti_info[RTAX_DST]->sa_family) {
468 #ifdef INET
469 	case AF_INET:
470 	{
471 		struct in_addr ia;
472 		struct ifaddr *ifa;
473 		int found;
474 
475 		found = 0;
476 		/*
477 		 * Try to find an address on the given outgoing interface
478 		 * that belongs to the jail.
479 		 */
480 		NET_EPOCH_ENTER(et);
481 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
482 			struct sockaddr *sa;
483 			sa = ifa->ifa_addr;
484 			if (sa->sa_family != AF_INET)
485 				continue;
486 			ia = ((struct sockaddr_in *)sa)->sin_addr;
487 			if (prison_check_ip4(cred, &ia) == 0) {
488 				found = 1;
489 				break;
490 			}
491 		}
492 		NET_EPOCH_EXIT(et);
493 		if (!found) {
494 			/*
495 			 * As a last resort return the 'default' jail address.
496 			 */
497 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
498 			    sin_addr;
499 			if (prison_get_ip4(cred, &ia) != 0)
500 				return (ESRCH);
501 		}
502 		bzero(&saun->sin, sizeof(struct sockaddr_in));
503 		saun->sin.sin_len = sizeof(struct sockaddr_in);
504 		saun->sin.sin_family = AF_INET;
505 		saun->sin.sin_addr.s_addr = ia.s_addr;
506 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
507 		break;
508 	}
509 #endif
510 #ifdef INET6
511 	case AF_INET6:
512 	{
513 		struct in6_addr ia6;
514 		struct ifaddr *ifa;
515 		int found;
516 
517 		found = 0;
518 		/*
519 		 * Try to find an address on the given outgoing interface
520 		 * that belongs to the jail.
521 		 */
522 		NET_EPOCH_ENTER(et);
523 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
524 			struct sockaddr *sa;
525 			sa = ifa->ifa_addr;
526 			if (sa->sa_family != AF_INET6)
527 				continue;
528 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
529 			    &ia6, sizeof(struct in6_addr));
530 			if (prison_check_ip6(cred, &ia6) == 0) {
531 				found = 1;
532 				break;
533 			}
534 		}
535 		NET_EPOCH_EXIT(et);
536 		if (!found) {
537 			/*
538 			 * As a last resort return the 'default' jail address.
539 			 */
540 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
541 			    sin6_addr;
542 			if (prison_get_ip6(cred, &ia6) != 0)
543 				return (ESRCH);
544 		}
545 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
546 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
547 		saun->sin6.sin6_family = AF_INET6;
548 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
549 		if (sa6_recoverscope(&saun->sin6) != 0)
550 			return (ESRCH);
551 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
552 		break;
553 	}
554 #endif
555 	default:
556 		return (ESRCH);
557 	}
558 	return (0);
559 }
560 
561 /*
562  * Fills in @info based on userland-provided @rtm message.
563  *
564  * Returns 0 on success.
565  */
566 static int
567 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
568 {
569 	int error;
570 	sa_family_t saf;
571 
572 	rtm->rtm_pid = curproc->p_pid;
573 	info->rti_addrs = rtm->rtm_addrs;
574 
575 	info->rti_mflags = rtm->rtm_inits;
576 	info->rti_rmx = &rtm->rtm_rmx;
577 
578 	/*
579 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
580 	 * link-local address because rtrequest requires addresses with
581 	 * embedded scope id.
582 	 */
583 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
584 		return (EINVAL);
585 
586 	if (rtm->rtm_flags & RTF_RNH_LOCKED)
587 		return (EINVAL);
588 	info->rti_flags = rtm->rtm_flags;
589 	if (info->rti_info[RTAX_DST] == NULL ||
590 	    info->rti_info[RTAX_DST]->sa_family >= AF_MAX ||
591 	    (info->rti_info[RTAX_GATEWAY] != NULL &&
592 	     info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
593 		return (EINVAL);
594 	saf = info->rti_info[RTAX_DST]->sa_family;
595 	/*
596 	 * Verify that the caller has the appropriate privilege; RTM_GET
597 	 * is the only operation the non-superuser is allowed.
598 	 */
599 	if (rtm->rtm_type != RTM_GET) {
600 		error = priv_check(curthread, PRIV_NET_ROUTE);
601 		if (error != 0)
602 			return (error);
603 	}
604 
605 	/*
606 	 * The given gateway address may be an interface address.
607 	 * For example, issuing a "route change" command on a route
608 	 * entry that was created from a tunnel, and the gateway
609 	 * address given is the local end point. In this case the
610 	 * RTF_GATEWAY flag must be cleared or the destination will
611 	 * not be reachable even though there is no error message.
612 	 */
613 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
614 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
615 		struct rt_addrinfo ginfo;
616 		struct sockaddr *gdst;
617 		struct sockaddr_storage ss;
618 
619 		bzero(&ginfo, sizeof(ginfo));
620 		bzero(&ss, sizeof(ss));
621 		ss.ss_len = sizeof(ss);
622 
623 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
624 		gdst = info->rti_info[RTAX_GATEWAY];
625 
626 		/*
627 		 * A host route through the loopback interface is
628 		 * installed for each interface adddress. In pre 8.0
629 		 * releases the interface address of a PPP link type
630 		 * is not reachable locally. This behavior is fixed as
631 		 * part of the new L2/L3 redesign and rewrite work. The
632 		 * signature of this interface address route is the
633 		 * AF_LINK sa_family type of the gateway, and the
634 		 * rt_ifp has the IFF_LOOPBACK flag set.
635 		 */
636 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
637 			if (ss.ss_family == AF_LINK &&
638 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
639 				info->rti_flags &= ~RTF_GATEWAY;
640 				info->rti_flags |= RTF_GWFLAG_COMPAT;
641 			}
642 			rib_free_info(&ginfo);
643 		}
644 	}
645 
646 	return (0);
647 }
648 
649 static struct nhop_object *
650 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
651 {
652 	if (!NH_IS_NHGRP(nh))
653 		return (nh);
654 #ifdef ROUTE_MPATH
655 	struct weightened_nhop *wn;
656 	uint32_t num_nhops;
657 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
658 	if (gw == NULL)
659 		return (wn[0].nh);
660 	for (int i = 0; i < num_nhops; i++) {
661 		if (match_nhop_gw(wn[i].nh, gw))
662 			return (wn[i].nh);
663 	}
664 #endif
665 	return (NULL);
666 }
667 
668 /*
669  * Handles RTM_GET message from routing socket, returning matching rt.
670  *
671  * Returns:
672  * 0 on success, with locked and referenced matching rt in @rt_nrt
673  * errno of failure
674  */
675 static int
676 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
677     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
678 {
679 	RIB_RLOCK_TRACKER;
680 	struct rib_head *rnh;
681 	struct nhop_object *nh;
682 	sa_family_t saf;
683 
684 	saf = info->rti_info[RTAX_DST]->sa_family;
685 
686 	rnh = rt_tables_get_rnh(fibnum, saf);
687 	if (rnh == NULL)
688 		return (EAFNOSUPPORT);
689 
690 	RIB_RLOCK(rnh);
691 
692 	if (info->rti_info[RTAX_NETMASK] == NULL) {
693 		/*
694 		 * Provide longest prefix match for
695 		 * address lookup (no mask).
696 		 * 'route -n get addr'
697 		 */
698 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
699 		    info->rti_info[RTAX_DST], &rnh->head);
700 	} else
701 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
702 		    info->rti_info[RTAX_DST],
703 		    info->rti_info[RTAX_NETMASK], &rnh->head);
704 
705 	if (rc->rc_rt == NULL) {
706 		RIB_RUNLOCK(rnh);
707 		return (ESRCH);
708 	}
709 
710 	nh = select_nhop(rc->rc_rt->rt_nhop, info->rti_info[RTAX_GATEWAY]);
711 	if (nh == NULL) {
712 		RIB_RUNLOCK(rnh);
713 		return (ESRCH);
714 	}
715 	/*
716 	 * If performing proxied L2 entry insertion, and
717 	 * the actual PPP host entry is found, perform
718 	 * another search to retrieve the prefix route of
719 	 * the local end point of the PPP link.
720 	 * TODO: move this logic to userland.
721 	 */
722 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
723 		struct sockaddr laddr;
724 		struct nhop_object *nh;
725 
726 		nh = rc->rc_rt->rt_nhop;
727 		if (nh->nh_ifp != NULL &&
728 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
729 			struct ifaddr *ifa;
730 
731 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
732 					RT_ALL_FIBS);
733 			if (ifa != NULL)
734 				rt_maskedcopy(ifa->ifa_addr,
735 					      &laddr,
736 					      ifa->ifa_netmask);
737 		} else
738 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
739 				      &laddr,
740 				      nh->nh_ifa->ifa_netmask);
741 		/*
742 		 * refactor rt and no lock operation necessary
743 		 */
744 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
745 		    &rnh->head);
746 		if (rc->rc_rt == NULL) {
747 			RIB_RUNLOCK(rnh);
748 			return (ESRCH);
749 		}
750 		nh = select_nhop(rc->rc_rt->rt_nhop, info->rti_info[RTAX_GATEWAY]);
751 		if (nh == NULL) {
752 			RIB_RUNLOCK(rnh);
753 			return (ESRCH);
754 		}
755 	}
756 	rc->rc_nh_new = nh;
757 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
758 	RIB_RUNLOCK(rnh);
759 
760 	return (0);
761 }
762 
763 /*
764  * Update sockaddrs, flags, etc in @prtm based on @rc data.
765  * rtm can be reallocated.
766  *
767  * Returns 0 on success, along with pointer to (potentially reallocated)
768  *  rtm.
769  *
770  */
771 static int
772 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
773     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
774 {
775 	struct sockaddr_storage netmask_ss;
776 	struct walkarg w;
777 	union sockaddr_union saun;
778 	struct rt_msghdr *rtm, *orig_rtm = NULL;
779 	struct ifnet *ifp;
780 	int error, len;
781 
782 	rtm = *prtm;
783 
784 	info->rti_info[RTAX_DST] = rt_key(rc->rc_rt);
785 	info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
786 	info->rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rc->rc_rt),
787 	    rt_mask(rc->rc_rt), &netmask_ss);
788 	info->rti_info[RTAX_GENMASK] = 0;
789 	ifp = nh->nh_ifp;
790 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
791 		if (ifp) {
792 			info->rti_info[RTAX_IFP] =
793 			    ifp->if_addr->ifa_addr;
794 			error = rtm_get_jailed(info, ifp, nh,
795 			    &saun, curthread->td_ucred);
796 			if (error != 0)
797 				return (error);
798 			if (ifp->if_flags & IFF_POINTOPOINT)
799 				info->rti_info[RTAX_BRD] =
800 				    nh->nh_ifa->ifa_dstaddr;
801 			rtm->rtm_index = ifp->if_index;
802 		} else {
803 			info->rti_info[RTAX_IFP] = NULL;
804 			info->rti_info[RTAX_IFA] = NULL;
805 		}
806 	} else if (ifp != NULL)
807 		rtm->rtm_index = ifp->if_index;
808 
809 	/* Check if we need to realloc storage */
810 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
811 	if (len > alloc_len) {
812 		struct rt_msghdr *tmp_rtm;
813 
814 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
815 		if (tmp_rtm == NULL)
816 			return (ENOBUFS);
817 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
818 		orig_rtm = rtm;
819 		rtm = tmp_rtm;
820 		alloc_len = len;
821 
822 		/*
823 		 * Delay freeing original rtm as info contains
824 		 * data referencing it.
825 		 */
826 	}
827 
828 	w.w_tmem = (caddr_t)rtm;
829 	w.w_tmemsize = alloc_len;
830 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
831 
832 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
833 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
834 		rtm->rtm_flags = RTF_GATEWAY |
835 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
836 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
837 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
838 	rtm->rtm_addrs = info->rti_addrs;
839 
840 	if (orig_rtm != NULL)
841 		free(orig_rtm, M_TEMP);
842 	*prtm = rtm;
843 
844 	return (0);
845 }
846 
847 #ifdef ROUTE_MPATH
848 static void
849 save_del_notification(struct rib_cmd_info *rc, void *_cbdata)
850 {
851 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
852 
853 	if (rc->rc_cmd == RTM_DELETE)
854 		*rc_new = *rc;
855 }
856 
857 static void
858 save_add_notification(struct rib_cmd_info *rc, void *_cbdata)
859 {
860 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
861 
862 	if (rc->rc_cmd == RTM_ADD)
863 		*rc_new = *rc;
864 }
865 #endif
866 
867 /*ARGSUSED*/
868 static int
869 route_output(struct mbuf *m, struct socket *so, ...)
870 {
871 	struct rt_msghdr *rtm = NULL;
872 	struct rtentry *rt = NULL;
873 	struct rt_addrinfo info;
874 	struct epoch_tracker et;
875 #ifdef INET6
876 	struct sockaddr_storage ss;
877 	struct sockaddr_in6 *sin6;
878 	int i, rti_need_deembed = 0;
879 #endif
880 	int alloc_len = 0, len, error = 0, fibnum;
881 	sa_family_t saf = AF_UNSPEC;
882 	struct walkarg w;
883 	struct rib_cmd_info rc;
884 	struct nhop_object *nh;
885 
886 	fibnum = so->so_fibnum;
887 #define senderr(e) { error = e; goto flush;}
888 	if (m == NULL || ((m->m_len < sizeof(long)) &&
889 		       (m = m_pullup(m, sizeof(long))) == NULL))
890 		return (ENOBUFS);
891 	if ((m->m_flags & M_PKTHDR) == 0)
892 		panic("route_output");
893 	NET_EPOCH_ENTER(et);
894 	len = m->m_pkthdr.len;
895 	if (len < sizeof(*rtm) ||
896 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
897 		senderr(EINVAL);
898 
899 	/*
900 	 * Most of current messages are in range 200-240 bytes,
901 	 * minimize possible re-allocation on reply using larger size
902 	 * buffer aligned on 1k boundaty.
903 	 */
904 	alloc_len = roundup2(len, 1024);
905 	if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
906 		senderr(ENOBUFS);
907 
908 	m_copydata(m, 0, len, (caddr_t)rtm);
909 	bzero(&info, sizeof(info));
910 	bzero(&w, sizeof(w));
911 	nh = NULL;
912 
913 	if (rtm->rtm_version != RTM_VERSION) {
914 		/* Do not touch message since format is unknown */
915 		free(rtm, M_TEMP);
916 		rtm = NULL;
917 		senderr(EPROTONOSUPPORT);
918 	}
919 
920 	/*
921 	 * Starting from here, it is possible
922 	 * to alter original message and insert
923 	 * caller PID and error value.
924 	 */
925 
926 	if ((error = fill_addrinfo(rtm, len, fibnum, &info)) != 0) {
927 		senderr(error);
928 	}
929 
930 	saf = info.rti_info[RTAX_DST]->sa_family;
931 
932 	/* support for new ARP code */
933 	if (rtm->rtm_flags & RTF_LLDATA) {
934 		error = lla_rt_output(rtm, &info);
935 #ifdef INET6
936 		if (error == 0)
937 			rti_need_deembed = 1;
938 #endif
939 		goto flush;
940 	}
941 
942 	switch (rtm->rtm_type) {
943 	case RTM_ADD:
944 	case RTM_CHANGE:
945 		if (rtm->rtm_type == RTM_ADD) {
946 			if (info.rti_info[RTAX_GATEWAY] == NULL)
947 				senderr(EINVAL);
948 		}
949 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
950 		if (error == 0) {
951 #ifdef INET6
952 			rti_need_deembed = 1;
953 #endif
954 #ifdef ROUTE_MPATH
955 			if (NH_IS_NHGRP(rc.rc_nh_new) ||
956 			    (rc.rc_nh_old && NH_IS_NHGRP(rc.rc_nh_old))) {
957 				struct rib_cmd_info rc_simple = {};
958 				rib_decompose_notification(&rc,
959 				    save_add_notification, (void *)&rc_simple);
960 				rc = rc_simple;
961 			}
962 #endif
963 			nh = rc.rc_nh_new;
964 			rtm->rtm_index = nh->nh_ifp->if_index;
965 			rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
966 		}
967 		break;
968 
969 	case RTM_DELETE:
970 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
971 		if (error == 0) {
972 #ifdef ROUTE_MPATH
973 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
974 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
975 				struct rib_cmd_info rc_simple = {};
976 				rib_decompose_notification(&rc,
977 				    save_del_notification, (void *)&rc_simple);
978 				rc = rc_simple;
979 			}
980 #endif
981 			nh = rc.rc_nh_old;
982 			goto report;
983 		}
984 #ifdef INET6
985 		/* rt_msg2() will not be used when RTM_DELETE fails. */
986 		rti_need_deembed = 1;
987 #endif
988 		break;
989 
990 	case RTM_GET:
991 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
992 		if (error != 0)
993 			senderr(error);
994 		nh = rc.rc_nh_new;
995 
996 report:
997 		if (!can_export_rte(curthread->td_ucred, rc.rc_rt)) {
998 			senderr(ESRCH);
999 		}
1000 
1001 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1002 		/*
1003 		 * Note that some sockaddr pointers may have changed to
1004 		 * point to memory outsize @rtm. Some may be pointing
1005 		 * to the on-stack variables.
1006 		 * Given that, any pointer in @info CANNOT BE USED.
1007 		 */
1008 
1009 		/*
1010 		 * scopeid deembedding has been performed while
1011 		 * writing updated rtm in rtsock_msg_buffer().
1012 		 * With that in mind, skip deembedding procedure below.
1013 		 */
1014 #ifdef INET6
1015 		rti_need_deembed = 0;
1016 #endif
1017 		if (error != 0)
1018 			senderr(error);
1019 		break;
1020 
1021 	default:
1022 		senderr(EOPNOTSUPP);
1023 	}
1024 
1025 flush:
1026 	NET_EPOCH_EXIT(et);
1027 	rt = NULL;
1028 
1029 #ifdef INET6
1030 	if (rtm != NULL) {
1031 		if (rti_need_deembed) {
1032 			/* sin6_scope_id is recovered before sending rtm. */
1033 			sin6 = (struct sockaddr_in6 *)&ss;
1034 			for (i = 0; i < RTAX_MAX; i++) {
1035 				if (info.rti_info[i] == NULL)
1036 					continue;
1037 				if (info.rti_info[i]->sa_family != AF_INET6)
1038 					continue;
1039 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1040 				if (sa6_recoverscope(sin6) == 0)
1041 					bcopy(sin6, info.rti_info[i],
1042 						    sizeof(*sin6));
1043 			}
1044 		}
1045 	}
1046 #endif
1047 	send_rtm_reply(so, rtm, m, saf, fibnum, error);
1048 
1049 	return (error);
1050 }
1051 
1052 /*
1053  * Sends the prepared reply message in @rtm to all rtsock clients.
1054  * Frees @m and @rtm.
1055  *
1056  */
1057 static void
1058 send_rtm_reply(struct socket *so, struct rt_msghdr *rtm, struct mbuf *m,
1059     sa_family_t saf, u_int fibnum, int rtm_errno)
1060 {
1061 	struct rawcb *rp = NULL;
1062 
1063 	/*
1064 	 * Check to see if we don't want our own messages.
1065 	 */
1066 	if ((so->so_options & SO_USELOOPBACK) == 0) {
1067 		if (V_route_cb.any_count <= 1) {
1068 			if (rtm != NULL)
1069 				free(rtm, M_TEMP);
1070 			m_freem(m);
1071 			return;
1072 		}
1073 		/* There is another listener, so construct message */
1074 		rp = sotorawcb(so);
1075 	}
1076 
1077 	if (rtm != NULL) {
1078 		if (rtm_errno!= 0)
1079 			rtm->rtm_errno = rtm_errno;
1080 		else
1081 			rtm->rtm_flags |= RTF_DONE;
1082 
1083 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
1084 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
1085 			m_freem(m);
1086 			m = NULL;
1087 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
1088 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1089 
1090 		free(rtm, M_TEMP);
1091 	}
1092 	if (m != NULL) {
1093 		M_SETFIB(m, fibnum);
1094 		m->m_flags |= RTS_FILTER_FIB;
1095 		if (rp) {
1096 			/*
1097 			 * XXX insure we don't get a copy by
1098 			 * invalidating our protocol
1099 			 */
1100 			unsigned short family = rp->rcb_proto.sp_family;
1101 			rp->rcb_proto.sp_family = 0;
1102 			rt_dispatch(m, saf);
1103 			rp->rcb_proto.sp_family = family;
1104 		} else
1105 			rt_dispatch(m, saf);
1106 	}
1107 }
1108 
1109 static void
1110 rt_getmetrics(const struct rtentry *rt, const struct nhop_object *nh,
1111     struct rt_metrics *out)
1112 {
1113 
1114 	bzero(out, sizeof(*out));
1115 	out->rmx_mtu = nh->nh_mtu;
1116 	out->rmx_weight = rt->rt_weight;
1117 	out->rmx_nhidx = nhop_get_idx(nh);
1118 	/* Kernel -> userland timebase conversion. */
1119 	out->rmx_expire = rt->rt_expire ?
1120 	    rt->rt_expire - time_uptime + time_second : 0;
1121 }
1122 
1123 /*
1124  * Extract the addresses of the passed sockaddrs.
1125  * Do a little sanity checking so as to avoid bad memory references.
1126  * This data is derived straight from userland.
1127  */
1128 static int
1129 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1130 {
1131 	struct sockaddr *sa;
1132 	int i;
1133 
1134 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1135 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1136 			continue;
1137 		sa = (struct sockaddr *)cp;
1138 		/*
1139 		 * It won't fit.
1140 		 */
1141 		if (cp + sa->sa_len > cplim)
1142 			return (EINVAL);
1143 		/*
1144 		 * there are no more.. quit now
1145 		 * If there are more bits, they are in error.
1146 		 * I've seen this. route(1) can evidently generate these.
1147 		 * This causes kernel to core dump.
1148 		 * for compatibility, If we see this, point to a safe address.
1149 		 */
1150 		if (sa->sa_len == 0) {
1151 			rtinfo->rti_info[i] = &sa_zero;
1152 			return (0); /* should be EINVAL but for compat */
1153 		}
1154 		/* accept it */
1155 #ifdef INET6
1156 		if (sa->sa_family == AF_INET6)
1157 			sa6_embedscope((struct sockaddr_in6 *)sa,
1158 			    V_ip6_use_defzone);
1159 #endif
1160 		rtinfo->rti_info[i] = sa;
1161 		cp += SA_SIZE(sa);
1162 	}
1163 	return (0);
1164 }
1165 
1166 /*
1167  * Fill in @dmask with valid netmask leaving original @smask
1168  * intact. Mostly used with radix netmasks.
1169  */
1170 struct sockaddr *
1171 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1172     struct sockaddr_storage *dmask)
1173 {
1174 	if (dst == NULL || smask == NULL)
1175 		return (NULL);
1176 
1177 	memset(dmask, 0, dst->sa_len);
1178 	memcpy(dmask, smask, smask->sa_len);
1179 	dmask->ss_len = dst->sa_len;
1180 	dmask->ss_family = dst->sa_family;
1181 
1182 	return ((struct sockaddr *)dmask);
1183 }
1184 
1185 /*
1186  * Writes information related to @rtinfo object to newly-allocated mbuf.
1187  * Assumes MCLBYTES is enough to construct any message.
1188  * Used for OS notifications of vaious events (if/ifa announces,etc)
1189  *
1190  * Returns allocated mbuf or NULL on failure.
1191  */
1192 static struct mbuf *
1193 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1194 {
1195 	struct rt_msghdr *rtm;
1196 	struct mbuf *m;
1197 	int i;
1198 	struct sockaddr *sa;
1199 #ifdef INET6
1200 	struct sockaddr_storage ss;
1201 	struct sockaddr_in6 *sin6;
1202 #endif
1203 	int len, dlen;
1204 
1205 	switch (type) {
1206 	case RTM_DELADDR:
1207 	case RTM_NEWADDR:
1208 		len = sizeof(struct ifa_msghdr);
1209 		break;
1210 
1211 	case RTM_DELMADDR:
1212 	case RTM_NEWMADDR:
1213 		len = sizeof(struct ifma_msghdr);
1214 		break;
1215 
1216 	case RTM_IFINFO:
1217 		len = sizeof(struct if_msghdr);
1218 		break;
1219 
1220 	case RTM_IFANNOUNCE:
1221 	case RTM_IEEE80211:
1222 		len = sizeof(struct if_announcemsghdr);
1223 		break;
1224 
1225 	default:
1226 		len = sizeof(struct rt_msghdr);
1227 	}
1228 
1229 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1230 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1231 	if (len > MHLEN)
1232 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1233 	else
1234 		m = m_gethdr(M_NOWAIT, MT_DATA);
1235 	if (m == NULL)
1236 		return (m);
1237 
1238 	m->m_pkthdr.len = m->m_len = len;
1239 	rtm = mtod(m, struct rt_msghdr *);
1240 	bzero((caddr_t)rtm, len);
1241 	for (i = 0; i < RTAX_MAX; i++) {
1242 		if ((sa = rtinfo->rti_info[i]) == NULL)
1243 			continue;
1244 		rtinfo->rti_addrs |= (1 << i);
1245 		dlen = SA_SIZE(sa);
1246 #ifdef INET6
1247 		if (sa->sa_family == AF_INET6) {
1248 			sin6 = (struct sockaddr_in6 *)&ss;
1249 			bcopy(sa, sin6, sizeof(*sin6));
1250 			if (sa6_recoverscope(sin6) == 0)
1251 				sa = (struct sockaddr *)sin6;
1252 		}
1253 #endif
1254 		m_copyback(m, len, dlen, (caddr_t)sa);
1255 		len += dlen;
1256 	}
1257 	if (m->m_pkthdr.len != len) {
1258 		m_freem(m);
1259 		return (NULL);
1260 	}
1261 	rtm->rtm_msglen = len;
1262 	rtm->rtm_version = RTM_VERSION;
1263 	rtm->rtm_type = type;
1264 	return (m);
1265 }
1266 
1267 /*
1268  * Writes information related to @rtinfo object to preallocated buffer.
1269  * Stores needed size in @plen. If @w is NULL, calculates size without
1270  * writing.
1271  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1272  *
1273  * Returns 0 on success.
1274  *
1275  */
1276 static int
1277 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1278 {
1279 	int i;
1280 	int len, buflen = 0, dlen;
1281 	caddr_t cp = NULL;
1282 	struct rt_msghdr *rtm = NULL;
1283 #ifdef INET6
1284 	struct sockaddr_storage ss;
1285 	struct sockaddr_in6 *sin6;
1286 #endif
1287 #ifdef COMPAT_FREEBSD32
1288 	bool compat32 = false;
1289 #endif
1290 
1291 	switch (type) {
1292 	case RTM_DELADDR:
1293 	case RTM_NEWADDR:
1294 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1295 #ifdef COMPAT_FREEBSD32
1296 			if (w->w_req->flags & SCTL_MASK32) {
1297 				len = sizeof(struct ifa_msghdrl32);
1298 				compat32 = true;
1299 			} else
1300 #endif
1301 				len = sizeof(struct ifa_msghdrl);
1302 		} else
1303 			len = sizeof(struct ifa_msghdr);
1304 		break;
1305 
1306 	case RTM_IFINFO:
1307 #ifdef COMPAT_FREEBSD32
1308 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1309 			if (w->w_op == NET_RT_IFLISTL)
1310 				len = sizeof(struct if_msghdrl32);
1311 			else
1312 				len = sizeof(struct if_msghdr32);
1313 			compat32 = true;
1314 			break;
1315 		}
1316 #endif
1317 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1318 			len = sizeof(struct if_msghdrl);
1319 		else
1320 			len = sizeof(struct if_msghdr);
1321 		break;
1322 
1323 	case RTM_NEWMADDR:
1324 		len = sizeof(struct ifma_msghdr);
1325 		break;
1326 
1327 	default:
1328 		len = sizeof(struct rt_msghdr);
1329 	}
1330 
1331 	if (w != NULL) {
1332 		rtm = (struct rt_msghdr *)w->w_tmem;
1333 		buflen = w->w_tmemsize - len;
1334 		cp = (caddr_t)w->w_tmem + len;
1335 	}
1336 
1337 	rtinfo->rti_addrs = 0;
1338 	for (i = 0; i < RTAX_MAX; i++) {
1339 		struct sockaddr *sa;
1340 
1341 		if ((sa = rtinfo->rti_info[i]) == NULL)
1342 			continue;
1343 		rtinfo->rti_addrs |= (1 << i);
1344 #ifdef COMPAT_FREEBSD32
1345 		if (compat32)
1346 			dlen = SA_SIZE32(sa);
1347 		else
1348 #endif
1349 			dlen = SA_SIZE(sa);
1350 		if (cp != NULL && buflen >= dlen) {
1351 #ifdef INET6
1352 			if (sa->sa_family == AF_INET6) {
1353 				sin6 = (struct sockaddr_in6 *)&ss;
1354 				bcopy(sa, sin6, sizeof(*sin6));
1355 				if (sa6_recoverscope(sin6) == 0)
1356 					sa = (struct sockaddr *)sin6;
1357 			}
1358 #endif
1359 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1360 			cp += dlen;
1361 			buflen -= dlen;
1362 		} else if (cp != NULL) {
1363 			/*
1364 			 * Buffer too small. Count needed size
1365 			 * and return with error.
1366 			 */
1367 			cp = NULL;
1368 		}
1369 
1370 		len += dlen;
1371 	}
1372 
1373 	if (cp != NULL) {
1374 		dlen = ALIGN(len) - len;
1375 		if (buflen < dlen)
1376 			cp = NULL;
1377 		else {
1378 			bzero(cp, dlen);
1379 			cp += dlen;
1380 			buflen -= dlen;
1381 		}
1382 	}
1383 	len = ALIGN(len);
1384 
1385 	if (cp != NULL) {
1386 		/* fill header iff buffer is large enough */
1387 		rtm->rtm_version = RTM_VERSION;
1388 		rtm->rtm_type = type;
1389 		rtm->rtm_msglen = len;
1390 	}
1391 
1392 	*plen = len;
1393 
1394 	if (w != NULL && cp == NULL)
1395 		return (ENOBUFS);
1396 
1397 	return (0);
1398 }
1399 
1400 /*
1401  * This routine is called to generate a message from the routing
1402  * socket indicating that a redirect has occurred, a routing lookup
1403  * has failed, or that a protocol has detected timeouts to a particular
1404  * destination.
1405  */
1406 void
1407 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1408     int fibnum)
1409 {
1410 	struct rt_msghdr *rtm;
1411 	struct mbuf *m;
1412 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1413 
1414 	if (V_route_cb.any_count == 0)
1415 		return;
1416 	m = rtsock_msg_mbuf(type, rtinfo);
1417 	if (m == NULL)
1418 		return;
1419 
1420 	if (fibnum != RT_ALL_FIBS) {
1421 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1422 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1423 		M_SETFIB(m, fibnum);
1424 		m->m_flags |= RTS_FILTER_FIB;
1425 	}
1426 
1427 	rtm = mtod(m, struct rt_msghdr *);
1428 	rtm->rtm_flags = RTF_DONE | flags;
1429 	rtm->rtm_errno = error;
1430 	rtm->rtm_addrs = rtinfo->rti_addrs;
1431 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1432 }
1433 
1434 void
1435 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1436 {
1437 
1438 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1439 }
1440 
1441 /*
1442  * This routine is called to generate a message from the routing
1443  * socket indicating that the status of a network interface has changed.
1444  */
1445 void
1446 rt_ifmsg(struct ifnet *ifp)
1447 {
1448 	struct if_msghdr *ifm;
1449 	struct mbuf *m;
1450 	struct rt_addrinfo info;
1451 
1452 	if (V_route_cb.any_count == 0)
1453 		return;
1454 	bzero((caddr_t)&info, sizeof(info));
1455 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1456 	if (m == NULL)
1457 		return;
1458 	ifm = mtod(m, struct if_msghdr *);
1459 	ifm->ifm_index = ifp->if_index;
1460 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1461 	if_data_copy(ifp, &ifm->ifm_data);
1462 	ifm->ifm_addrs = 0;
1463 	rt_dispatch(m, AF_UNSPEC);
1464 }
1465 
1466 /*
1467  * Announce interface address arrival/withdraw.
1468  * Please do not call directly, use rt_addrmsg().
1469  * Assume input data to be valid.
1470  * Returns 0 on success.
1471  */
1472 int
1473 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1474 {
1475 	struct rt_addrinfo info;
1476 	struct sockaddr *sa;
1477 	int ncmd;
1478 	struct mbuf *m;
1479 	struct ifa_msghdr *ifam;
1480 	struct ifnet *ifp = ifa->ifa_ifp;
1481 	struct sockaddr_storage ss;
1482 
1483 	if (V_route_cb.any_count == 0)
1484 		return (0);
1485 
1486 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1487 
1488 	bzero((caddr_t)&info, sizeof(info));
1489 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1490 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1491 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1492 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1493 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1494 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1495 		return (ENOBUFS);
1496 	ifam = mtod(m, struct ifa_msghdr *);
1497 	ifam->ifam_index = ifp->if_index;
1498 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1499 	ifam->ifam_flags = ifa->ifa_flags;
1500 	ifam->ifam_addrs = info.rti_addrs;
1501 
1502 	if (fibnum != RT_ALL_FIBS) {
1503 		M_SETFIB(m, fibnum);
1504 		m->m_flags |= RTS_FILTER_FIB;
1505 	}
1506 
1507 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1508 
1509 	return (0);
1510 }
1511 
1512 /*
1513  * Announce route addition/removal to rtsock based on @rt data.
1514  * Callers are advives to use rt_routemsg() instead of using this
1515  *  function directly.
1516  * Assume @rt data is consistent.
1517  *
1518  * Returns 0 on success.
1519  */
1520 int
1521 rtsock_routemsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int rti_addrs,
1522     int fibnum)
1523 {
1524 	struct sockaddr_storage ss;
1525 	struct rt_addrinfo info;
1526 	struct nhop_object *nh;
1527 
1528 	if (V_route_cb.any_count == 0)
1529 		return (0);
1530 
1531 	nh = rt->rt_nhop;
1532 	bzero((caddr_t)&info, sizeof(info));
1533 	info.rti_info[RTAX_DST] = rt_key(rt);
1534 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt), rt_mask(rt), &ss);
1535 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1536 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
1537 	info.rti_ifp = ifp;
1538 
1539 	return (rtsock_routemsg_info(cmd, &info, fibnum));
1540 }
1541 
1542 int
1543 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1544 {
1545 	struct rt_msghdr *rtm;
1546 	struct sockaddr *sa;
1547 	struct mbuf *m;
1548 
1549 	if (V_route_cb.any_count == 0)
1550 		return (0);
1551 
1552 	if (info->rti_flags & RTF_HOST)
1553 		info->rti_info[RTAX_NETMASK] = NULL;
1554 
1555 	m = rtsock_msg_mbuf(cmd, info);
1556 	if (m == NULL)
1557 		return (ENOBUFS);
1558 
1559 	if (fibnum != RT_ALL_FIBS) {
1560 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1561 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1562 		M_SETFIB(m, fibnum);
1563 		m->m_flags |= RTS_FILTER_FIB;
1564 	}
1565 
1566 	rtm = mtod(m, struct rt_msghdr *);
1567 	rtm->rtm_addrs = info->rti_addrs;
1568 	if (info->rti_ifp != NULL)
1569 		rtm->rtm_index = info->rti_ifp->if_index;
1570 	/* Add RTF_DONE to indicate command 'completion' required by API */
1571 	info->rti_flags |= RTF_DONE;
1572 	/* Reported routes has to be up */
1573 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1574 		info->rti_flags |= RTF_UP;
1575 	rtm->rtm_flags = info->rti_flags;
1576 
1577 	sa = info->rti_info[RTAX_DST];
1578 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1579 
1580 	return (0);
1581 }
1582 
1583 /*
1584  * This is the analogue to the rt_newaddrmsg which performs the same
1585  * function but for multicast group memberhips.  This is easier since
1586  * there is no route state to worry about.
1587  */
1588 void
1589 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1590 {
1591 	struct rt_addrinfo info;
1592 	struct mbuf *m = NULL;
1593 	struct ifnet *ifp = ifma->ifma_ifp;
1594 	struct ifma_msghdr *ifmam;
1595 
1596 	if (V_route_cb.any_count == 0)
1597 		return;
1598 
1599 	bzero((caddr_t)&info, sizeof(info));
1600 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1601 	if (ifp && ifp->if_addr)
1602 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1603 	else
1604 		info.rti_info[RTAX_IFP] = NULL;
1605 	/*
1606 	 * If a link-layer address is present, present it as a ``gateway''
1607 	 * (similarly to how ARP entries, e.g., are presented).
1608 	 */
1609 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1610 	m = rtsock_msg_mbuf(cmd, &info);
1611 	if (m == NULL)
1612 		return;
1613 	ifmam = mtod(m, struct ifma_msghdr *);
1614 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1615 	    __func__));
1616 	ifmam->ifmam_index = ifp->if_index;
1617 	ifmam->ifmam_addrs = info.rti_addrs;
1618 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1619 }
1620 
1621 static struct mbuf *
1622 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1623 	struct rt_addrinfo *info)
1624 {
1625 	struct if_announcemsghdr *ifan;
1626 	struct mbuf *m;
1627 
1628 	if (V_route_cb.any_count == 0)
1629 		return NULL;
1630 	bzero((caddr_t)info, sizeof(*info));
1631 	m = rtsock_msg_mbuf(type, info);
1632 	if (m != NULL) {
1633 		ifan = mtod(m, struct if_announcemsghdr *);
1634 		ifan->ifan_index = ifp->if_index;
1635 		strlcpy(ifan->ifan_name, ifp->if_xname,
1636 			sizeof(ifan->ifan_name));
1637 		ifan->ifan_what = what;
1638 	}
1639 	return m;
1640 }
1641 
1642 /*
1643  * This is called to generate routing socket messages indicating
1644  * IEEE80211 wireless events.
1645  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1646  */
1647 void
1648 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1649 {
1650 	struct mbuf *m;
1651 	struct rt_addrinfo info;
1652 
1653 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1654 	if (m != NULL) {
1655 		/*
1656 		 * Append the ieee80211 data.  Try to stick it in the
1657 		 * mbuf containing the ifannounce msg; otherwise allocate
1658 		 * a new mbuf and append.
1659 		 *
1660 		 * NB: we assume m is a single mbuf.
1661 		 */
1662 		if (data_len > M_TRAILINGSPACE(m)) {
1663 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1664 			if (n == NULL) {
1665 				m_freem(m);
1666 				return;
1667 			}
1668 			bcopy(data, mtod(n, void *), data_len);
1669 			n->m_len = data_len;
1670 			m->m_next = n;
1671 		} else if (data_len > 0) {
1672 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1673 			m->m_len += data_len;
1674 		}
1675 		if (m->m_flags & M_PKTHDR)
1676 			m->m_pkthdr.len += data_len;
1677 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1678 		rt_dispatch(m, AF_UNSPEC);
1679 	}
1680 }
1681 
1682 /*
1683  * This is called to generate routing socket messages indicating
1684  * network interface arrival and departure.
1685  */
1686 void
1687 rt_ifannouncemsg(struct ifnet *ifp, int what)
1688 {
1689 	struct mbuf *m;
1690 	struct rt_addrinfo info;
1691 
1692 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1693 	if (m != NULL)
1694 		rt_dispatch(m, AF_UNSPEC);
1695 }
1696 
1697 static void
1698 rt_dispatch(struct mbuf *m, sa_family_t saf)
1699 {
1700 	struct m_tag *tag;
1701 
1702 	/*
1703 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1704 	 * use when injecting the mbuf into the routing socket buffer from
1705 	 * the netisr.
1706 	 */
1707 	if (saf != AF_UNSPEC) {
1708 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1709 		    M_NOWAIT);
1710 		if (tag == NULL) {
1711 			m_freem(m);
1712 			return;
1713 		}
1714 		*(unsigned short *)(tag + 1) = saf;
1715 		m_tag_prepend(m, tag);
1716 	}
1717 #ifdef VIMAGE
1718 	if (V_loif)
1719 		m->m_pkthdr.rcvif = V_loif;
1720 	else {
1721 		m_freem(m);
1722 		return;
1723 	}
1724 #endif
1725 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1726 }
1727 
1728 /*
1729  * Checks if rte can be exported v.r.t jails/vnets.
1730  *
1731  * Returns 1 if it can, 0 otherwise.
1732  */
1733 static int
1734 can_export_rte(struct ucred *td_ucred, const struct rtentry *rt)
1735 {
1736 
1737 	if ((rt->rte_flags & RTF_HOST) == 0
1738 	    ? jailed_without_vnet(td_ucred)
1739 	    : prison_if(td_ucred, rt_key_const(rt)) != 0)
1740 		return (0);
1741 	return (1);
1742 }
1743 
1744 /*
1745  * This is used in dumping the kernel table via sysctl().
1746  */
1747 static int
1748 sysctl_dumpentry(struct radix_node *rn, void *vw)
1749 {
1750 	struct walkarg *w = vw;
1751 	struct rtentry *rt = (struct rtentry *)rn;
1752 	struct nhop_object *nh;
1753 	int error = 0;
1754 
1755 	NET_EPOCH_ASSERT();
1756 
1757 	if (w->w_op == NET_RT_FLAGS && !(rt->rte_flags & w->w_arg))
1758 		return 0;
1759 	if (!can_export_rte(w->w_req->td->td_ucred, rt))
1760 		return (0);
1761 	nh = rt->rt_nhop;
1762 #ifdef ROUTE_MPATH
1763 	if (NH_IS_NHGRP(nh)) {
1764 		struct weightened_nhop *wn;
1765 		uint32_t num_nhops;
1766 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
1767 		for (int i = 0; i < num_nhops; i++) {
1768 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
1769 			if (error != 0)
1770 				return (error);
1771 		}
1772 	} else
1773 #endif
1774 		error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
1775 
1776 	return (0);
1777 }
1778 
1779 
1780 static int
1781 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
1782     struct walkarg *w)
1783 {
1784 	struct rt_addrinfo info;
1785 	int error = 0, size;
1786 	struct sockaddr_storage ss;
1787 
1788 	bzero((caddr_t)&info, sizeof(info));
1789 	info.rti_info[RTAX_DST] = rt_key(rt);
1790 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1791 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(rt_key(rt),
1792 	    rt_mask(rt), &ss);
1793 	info.rti_info[RTAX_GENMASK] = 0;
1794 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
1795 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
1796 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
1797 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
1798 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
1799 	}
1800 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1801 		return (error);
1802 	if (w->w_req && w->w_tmem) {
1803 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1804 
1805 		bzero(&rtm->rtm_index,
1806 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1807 		if (rt->rte_flags & RTF_GWFLAG_COMPAT)
1808 			rtm->rtm_flags = RTF_GATEWAY |
1809 				(rt->rte_flags & ~RTF_GWFLAG_COMPAT);
1810 		else
1811 			rtm->rtm_flags = rt->rte_flags;
1812 		rtm->rtm_flags |= nhop_get_rtflags(nh);
1813 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
1814 		rtm->rtm_rmx.rmx_weight = weight;
1815 		rtm->rtm_index = nh->nh_ifp->if_index;
1816 		rtm->rtm_addrs = info.rti_addrs;
1817 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1818 		return (error);
1819 	}
1820 	return (error);
1821 }
1822 
1823 static int
1824 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1825     struct rt_addrinfo *info, struct walkarg *w, int len)
1826 {
1827 	struct if_msghdrl *ifm;
1828 	struct if_data *ifd;
1829 
1830 	ifm = (struct if_msghdrl *)w->w_tmem;
1831 
1832 #ifdef COMPAT_FREEBSD32
1833 	if (w->w_req->flags & SCTL_MASK32) {
1834 		struct if_msghdrl32 *ifm32;
1835 
1836 		ifm32 = (struct if_msghdrl32 *)ifm;
1837 		ifm32->ifm_addrs = info->rti_addrs;
1838 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1839 		ifm32->ifm_index = ifp->if_index;
1840 		ifm32->_ifm_spare1 = 0;
1841 		ifm32->ifm_len = sizeof(*ifm32);
1842 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1843 		ifm32->_ifm_spare2 = 0;
1844 		ifd = &ifm32->ifm_data;
1845 	} else
1846 #endif
1847 	{
1848 		ifm->ifm_addrs = info->rti_addrs;
1849 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1850 		ifm->ifm_index = ifp->if_index;
1851 		ifm->_ifm_spare1 = 0;
1852 		ifm->ifm_len = sizeof(*ifm);
1853 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1854 		ifm->_ifm_spare2 = 0;
1855 		ifd = &ifm->ifm_data;
1856 	}
1857 
1858 	memcpy(ifd, src_ifd, sizeof(*ifd));
1859 
1860 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1861 }
1862 
1863 static int
1864 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
1865     struct rt_addrinfo *info, struct walkarg *w, int len)
1866 {
1867 	struct if_msghdr *ifm;
1868 	struct if_data *ifd;
1869 
1870 	ifm = (struct if_msghdr *)w->w_tmem;
1871 
1872 #ifdef COMPAT_FREEBSD32
1873 	if (w->w_req->flags & SCTL_MASK32) {
1874 		struct if_msghdr32 *ifm32;
1875 
1876 		ifm32 = (struct if_msghdr32 *)ifm;
1877 		ifm32->ifm_addrs = info->rti_addrs;
1878 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1879 		ifm32->ifm_index = ifp->if_index;
1880 		ifm32->_ifm_spare1 = 0;
1881 		ifd = &ifm32->ifm_data;
1882 	} else
1883 #endif
1884 	{
1885 		ifm->ifm_addrs = info->rti_addrs;
1886 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1887 		ifm->ifm_index = ifp->if_index;
1888 		ifm->_ifm_spare1 = 0;
1889 		ifd = &ifm->ifm_data;
1890 	}
1891 
1892 	memcpy(ifd, src_ifd, sizeof(*ifd));
1893 
1894 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1895 }
1896 
1897 static int
1898 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1899     struct walkarg *w, int len)
1900 {
1901 	struct ifa_msghdrl *ifam;
1902 	struct if_data *ifd;
1903 
1904 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1905 
1906 #ifdef COMPAT_FREEBSD32
1907 	if (w->w_req->flags & SCTL_MASK32) {
1908 		struct ifa_msghdrl32 *ifam32;
1909 
1910 		ifam32 = (struct ifa_msghdrl32 *)ifam;
1911 		ifam32->ifam_addrs = info->rti_addrs;
1912 		ifam32->ifam_flags = ifa->ifa_flags;
1913 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1914 		ifam32->_ifam_spare1 = 0;
1915 		ifam32->ifam_len = sizeof(*ifam32);
1916 		ifam32->ifam_data_off =
1917 		    offsetof(struct ifa_msghdrl32, ifam_data);
1918 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
1919 		ifd = &ifam32->ifam_data;
1920 	} else
1921 #endif
1922 	{
1923 		ifam->ifam_addrs = info->rti_addrs;
1924 		ifam->ifam_flags = ifa->ifa_flags;
1925 		ifam->ifam_index = ifa->ifa_ifp->if_index;
1926 		ifam->_ifam_spare1 = 0;
1927 		ifam->ifam_len = sizeof(*ifam);
1928 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1929 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1930 		ifd = &ifam->ifam_data;
1931 	}
1932 
1933 	bzero(ifd, sizeof(*ifd));
1934 	ifd->ifi_datalen = sizeof(struct if_data);
1935 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1936 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1937 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1938 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1939 
1940 	/* Fixup if_data carp(4) vhid. */
1941 	if (carp_get_vhid_p != NULL)
1942 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
1943 
1944 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1945 }
1946 
1947 static int
1948 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1949     struct walkarg *w, int len)
1950 {
1951 	struct ifa_msghdr *ifam;
1952 
1953 	ifam = (struct ifa_msghdr *)w->w_tmem;
1954 	ifam->ifam_addrs = info->rti_addrs;
1955 	ifam->ifam_flags = ifa->ifa_flags;
1956 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1957 	ifam->_ifam_spare1 = 0;
1958 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1959 
1960 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1961 }
1962 
1963 static int
1964 sysctl_iflist(int af, struct walkarg *w)
1965 {
1966 	struct ifnet *ifp;
1967 	struct ifaddr *ifa;
1968 	struct if_data ifd;
1969 	struct rt_addrinfo info;
1970 	int len, error = 0;
1971 	struct sockaddr_storage ss;
1972 
1973 	bzero((caddr_t)&info, sizeof(info));
1974 	bzero(&ifd, sizeof(ifd));
1975 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1976 		if (w->w_arg && w->w_arg != ifp->if_index)
1977 			continue;
1978 		if_data_copy(ifp, &ifd);
1979 		ifa = ifp->if_addr;
1980 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1981 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
1982 		if (error != 0)
1983 			goto done;
1984 		info.rti_info[RTAX_IFP] = NULL;
1985 		if (w->w_req && w->w_tmem) {
1986 			if (w->w_op == NET_RT_IFLISTL)
1987 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
1988 				    len);
1989 			else
1990 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
1991 				    len);
1992 			if (error)
1993 				goto done;
1994 		}
1995 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
1996 			if (af && af != ifa->ifa_addr->sa_family)
1997 				continue;
1998 			if (prison_if(w->w_req->td->td_ucred,
1999 			    ifa->ifa_addr) != 0)
2000 				continue;
2001 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2002 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2003 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2004 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2005 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2006 			if (error != 0)
2007 				goto done;
2008 			if (w->w_req && w->w_tmem) {
2009 				if (w->w_op == NET_RT_IFLISTL)
2010 					error = sysctl_iflist_ifaml(ifa, &info,
2011 					    w, len);
2012 				else
2013 					error = sysctl_iflist_ifam(ifa, &info,
2014 					    w, len);
2015 				if (error)
2016 					goto done;
2017 			}
2018 		}
2019 		info.rti_info[RTAX_IFA] = NULL;
2020 		info.rti_info[RTAX_NETMASK] = NULL;
2021 		info.rti_info[RTAX_BRD] = NULL;
2022 	}
2023 done:
2024 	return (error);
2025 }
2026 
2027 static int
2028 sysctl_ifmalist(int af, struct walkarg *w)
2029 {
2030 	struct rt_addrinfo info;
2031 	struct ifaddr *ifa;
2032 	struct ifmultiaddr *ifma;
2033 	struct ifnet *ifp;
2034 	int error, len;
2035 
2036 	NET_EPOCH_ASSERT();
2037 
2038 	error = 0;
2039 	bzero((caddr_t)&info, sizeof(info));
2040 
2041 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2042 		if (w->w_arg && w->w_arg != ifp->if_index)
2043 			continue;
2044 		ifa = ifp->if_addr;
2045 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2046 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2047 			if (af && af != ifma->ifma_addr->sa_family)
2048 				continue;
2049 			if (prison_if(w->w_req->td->td_ucred,
2050 			    ifma->ifma_addr) != 0)
2051 				continue;
2052 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2053 			info.rti_info[RTAX_GATEWAY] =
2054 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2055 			    ifma->ifma_lladdr : NULL;
2056 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2057 			if (error != 0)
2058 				break;
2059 			if (w->w_req && w->w_tmem) {
2060 				struct ifma_msghdr *ifmam;
2061 
2062 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2063 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2064 				ifmam->ifmam_flags = 0;
2065 				ifmam->ifmam_addrs = info.rti_addrs;
2066 				ifmam->_ifmam_spare1 = 0;
2067 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2068 				if (error != 0)
2069 					break;
2070 			}
2071 		}
2072 		if (error != 0)
2073 			break;
2074 	}
2075 	return (error);
2076 }
2077 
2078 static int
2079 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2080 {
2081 	RIB_RLOCK_TRACKER;
2082 	struct epoch_tracker et;
2083 	int	*name = (int *)arg1;
2084 	u_int	namelen = arg2;
2085 	struct rib_head *rnh = NULL; /* silence compiler. */
2086 	int	i, lim, error = EINVAL;
2087 	int	fib = 0;
2088 	u_char	af;
2089 	struct	walkarg w;
2090 
2091 	name ++;
2092 	namelen--;
2093 	if (req->newptr)
2094 		return (EPERM);
2095 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2096 		if (namelen == 3)
2097 			fib = req->td->td_proc->p_fibnum;
2098 		else if (namelen == 4)
2099 			fib = (name[3] == RT_ALL_FIBS) ?
2100 			    req->td->td_proc->p_fibnum : name[3];
2101 		else
2102 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2103 		if (fib < 0 || fib >= rt_numfibs)
2104 			return (EINVAL);
2105 	} else if (namelen != 3)
2106 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2107 	af = name[0];
2108 	if (af > AF_MAX)
2109 		return (EINVAL);
2110 	bzero(&w, sizeof(w));
2111 	w.w_op = name[1];
2112 	w.w_arg = name[2];
2113 	w.w_req = req;
2114 
2115 	error = sysctl_wire_old_buffer(req, 0);
2116 	if (error)
2117 		return (error);
2118 
2119 	/*
2120 	 * Allocate reply buffer in advance.
2121 	 * All rtsock messages has maximum length of u_short.
2122 	 */
2123 	w.w_tmemsize = 65536;
2124 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2125 
2126 	NET_EPOCH_ENTER(et);
2127 	switch (w.w_op) {
2128 	case NET_RT_DUMP:
2129 	case NET_RT_FLAGS:
2130 		if (af == 0) {			/* dump all tables */
2131 			i = 1;
2132 			lim = AF_MAX;
2133 		} else				/* dump only one table */
2134 			i = lim = af;
2135 
2136 		/*
2137 		 * take care of llinfo entries, the caller must
2138 		 * specify an AF
2139 		 */
2140 		if (w.w_op == NET_RT_FLAGS &&
2141 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2142 			if (af != 0)
2143 				error = lltable_sysctl_dumparp(af, w.w_req);
2144 			else
2145 				error = EINVAL;
2146 			break;
2147 		}
2148 		/*
2149 		 * take care of routing entries
2150 		 */
2151 		for (error = 0; error == 0 && i <= lim; i++) {
2152 			rnh = rt_tables_get_rnh(fib, i);
2153 			if (rnh != NULL) {
2154 				RIB_RLOCK(rnh);
2155 			    	error = rnh->rnh_walktree(&rnh->head,
2156 				    sysctl_dumpentry, &w);
2157 				RIB_RUNLOCK(rnh);
2158 			} else if (af != 0)
2159 				error = EAFNOSUPPORT;
2160 		}
2161 		break;
2162 	case NET_RT_NHOP:
2163 	case NET_RT_NHGRP:
2164 		/* Allow dumping one specific af/fib at a time */
2165 		if (namelen < 4) {
2166 			error = EINVAL;
2167 			break;
2168 		}
2169 		fib = name[3];
2170 		if (fib < 0 || fib > rt_numfibs) {
2171 			error = EINVAL;
2172 			break;
2173 		}
2174 		rnh = rt_tables_get_rnh(fib, af);
2175 		if (rnh == NULL) {
2176 			error = EAFNOSUPPORT;
2177 			break;
2178 		}
2179 		if (w.w_op == NET_RT_NHOP)
2180 			error = nhops_dump_sysctl(rnh, w.w_req);
2181 		else
2182 #ifdef ROUTE_MPATH
2183 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2184 #else
2185 			error = ENOTSUP;
2186 #endif
2187 		break;
2188 	case NET_RT_IFLIST:
2189 	case NET_RT_IFLISTL:
2190 		error = sysctl_iflist(af, &w);
2191 		break;
2192 
2193 	case NET_RT_IFMALIST:
2194 		error = sysctl_ifmalist(af, &w);
2195 		break;
2196 	}
2197 	NET_EPOCH_EXIT(et);
2198 
2199 	free(w.w_tmem, M_TEMP);
2200 	return (error);
2201 }
2202 
2203 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2204     sysctl_rtsock, "Return route tables and interface/address lists");
2205 
2206 /*
2207  * Definitions of protocols supported in the ROUTE domain.
2208  */
2209 
2210 static struct domain routedomain;		/* or at least forward */
2211 
2212 static struct protosw routesw[] = {
2213 {
2214 	.pr_type =		SOCK_RAW,
2215 	.pr_domain =		&routedomain,
2216 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2217 	.pr_output =		route_output,
2218 	.pr_ctlinput =		raw_ctlinput,
2219 	.pr_init =		raw_init,
2220 	.pr_usrreqs =		&route_usrreqs
2221 }
2222 };
2223 
2224 static struct domain routedomain = {
2225 	.dom_family =		PF_ROUTE,
2226 	.dom_name =		"route",
2227 	.dom_protosw =		routesw,
2228 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
2229 };
2230 
2231 VNET_DOMAIN_SET(route);
2232