xref: /freebsd/sys/net/rtsock.c (revision f15e18a642cb3f7ebc747f8e9cdf11274140107d)
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	family;
162 	int	w_tmemsize;
163 	int	w_op, w_arg;
164 	caddr_t	w_tmem;
165 	struct sysctl_req *w_req;
166 	struct sockaddr *dst;
167 	struct sockaddr *mask;
168 };
169 
170 static void	rts_input(struct mbuf *m);
171 static struct mbuf *rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo);
172 static int	rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo,
173 			struct walkarg *w, int *plen);
174 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
175 			struct rt_addrinfo *rtinfo);
176 static int	sysctl_dumpentry(struct rtentry *rt, void *vw);
177 static int	sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh,
178 			uint32_t weight, struct walkarg *w);
179 static int	sysctl_iflist(int af, struct walkarg *w);
180 static int	sysctl_ifmalist(int af, struct walkarg *w);
181 static int	route_output(struct mbuf *m, struct socket *so, ...);
182 static void	rt_getmetrics(const struct rtentry *rt,
183 			const struct nhop_object *nh, struct rt_metrics *out);
184 static void	rt_dispatch(struct mbuf *, sa_family_t);
185 static int	handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
186 			struct rt_msghdr *rtm, struct rib_cmd_info *rc);
187 static int	update_rtm_from_rc(struct rt_addrinfo *info,
188 			struct rt_msghdr **prtm, int alloc_len,
189 			struct rib_cmd_info *rc, struct nhop_object *nh);
190 static void	send_rtm_reply(struct socket *so, struct rt_msghdr *rtm,
191 			struct mbuf *m, sa_family_t saf, u_int fibnum,
192 			int rtm_errno);
193 static bool	can_export_rte(struct ucred *td_ucred, bool rt_is_host,
194 			const struct sockaddr *rt_dst);
195 
196 static struct netisr_handler rtsock_nh = {
197 	.nh_name = "rtsock",
198 	.nh_handler = rts_input,
199 	.nh_proto = NETISR_ROUTE,
200 	.nh_policy = NETISR_POLICY_SOURCE,
201 };
202 
203 static int
204 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
205 {
206 	int error, qlimit;
207 
208 	netisr_getqlimit(&rtsock_nh, &qlimit);
209 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
210         if (error || !req->newptr)
211                 return (error);
212 	if (qlimit < 1)
213 		return (EINVAL);
214 	return (netisr_setqlimit(&rtsock_nh, qlimit));
215 }
216 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen,
217     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
218     0, 0, sysctl_route_netisr_maxqlen, "I",
219     "maximum routing socket dispatch queue length");
220 
221 static void
222 vnet_rts_init(void)
223 {
224 	int tmp;
225 
226 	if (IS_DEFAULT_VNET(curvnet)) {
227 		if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
228 			rtsock_nh.nh_qlimit = tmp;
229 		netisr_register(&rtsock_nh);
230 	}
231 #ifdef VIMAGE
232 	 else
233 		netisr_register_vnet(&rtsock_nh);
234 #endif
235 }
236 VNET_SYSINIT(vnet_rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
237     vnet_rts_init, 0);
238 
239 #ifdef VIMAGE
240 static void
241 vnet_rts_uninit(void)
242 {
243 
244 	netisr_unregister_vnet(&rtsock_nh);
245 }
246 VNET_SYSUNINIT(vnet_rts_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
247     vnet_rts_uninit, 0);
248 #endif
249 
250 static int
251 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
252     struct rawcb *rp)
253 {
254 	int fibnum;
255 
256 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
257 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
258 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
259 
260 	/* No filtering requested. */
261 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
262 		return (0);
263 
264 	/* Check if it is a rts and the fib matches the one of the socket. */
265 	fibnum = M_GETFIB(m);
266 	if (proto->sp_family != PF_ROUTE ||
267 	    rp->rcb_socket == NULL ||
268 	    rp->rcb_socket->so_fibnum == fibnum)
269 		return (0);
270 
271 	/* Filtering requested and no match, the socket shall be skipped. */
272 	return (1);
273 }
274 
275 static void
276 rts_input(struct mbuf *m)
277 {
278 	struct sockproto route_proto;
279 	unsigned short *family;
280 	struct m_tag *tag;
281 
282 	route_proto.sp_family = PF_ROUTE;
283 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
284 	if (tag != NULL) {
285 		family = (unsigned short *)(tag + 1);
286 		route_proto.sp_protocol = *family;
287 		m_tag_delete(m, tag);
288 	} else
289 		route_proto.sp_protocol = 0;
290 
291 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
292 }
293 
294 /*
295  * It really doesn't make any sense at all for this code to share much
296  * with raw_usrreq.c, since its functionality is so restricted.  XXX
297  */
298 static void
299 rts_abort(struct socket *so)
300 {
301 
302 	raw_usrreqs.pru_abort(so);
303 }
304 
305 static void
306 rts_close(struct socket *so)
307 {
308 
309 	raw_usrreqs.pru_close(so);
310 }
311 
312 /* pru_accept is EOPNOTSUPP */
313 
314 static int
315 rts_attach(struct socket *so, int proto, struct thread *td)
316 {
317 	struct rawcb *rp;
318 	int error;
319 
320 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
321 
322 	/* XXX */
323 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
324 
325 	so->so_pcb = (caddr_t)rp;
326 	so->so_fibnum = td->td_proc->p_fibnum;
327 	error = raw_attach(so, proto);
328 	rp = sotorawcb(so);
329 	if (error) {
330 		so->so_pcb = NULL;
331 		free(rp, M_PCB);
332 		return error;
333 	}
334 	RTSOCK_LOCK();
335 	switch(rp->rcb_proto.sp_protocol) {
336 	case AF_INET:
337 		V_route_cb.ip_count++;
338 		break;
339 	case AF_INET6:
340 		V_route_cb.ip6_count++;
341 		break;
342 	}
343 	V_route_cb.any_count++;
344 	RTSOCK_UNLOCK();
345 	soisconnected(so);
346 	so->so_options |= SO_USELOOPBACK;
347 	return 0;
348 }
349 
350 static int
351 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
352 {
353 
354 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
355 }
356 
357 static int
358 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
359 {
360 
361 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
362 }
363 
364 /* pru_connect2 is EOPNOTSUPP */
365 /* pru_control is EOPNOTSUPP */
366 
367 static void
368 rts_detach(struct socket *so)
369 {
370 	struct rawcb *rp = sotorawcb(so);
371 
372 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
373 
374 	RTSOCK_LOCK();
375 	switch(rp->rcb_proto.sp_protocol) {
376 	case AF_INET:
377 		V_route_cb.ip_count--;
378 		break;
379 	case AF_INET6:
380 		V_route_cb.ip6_count--;
381 		break;
382 	}
383 	V_route_cb.any_count--;
384 	RTSOCK_UNLOCK();
385 	raw_usrreqs.pru_detach(so);
386 }
387 
388 static int
389 rts_disconnect(struct socket *so)
390 {
391 
392 	return (raw_usrreqs.pru_disconnect(so));
393 }
394 
395 /* pru_listen is EOPNOTSUPP */
396 
397 static int
398 rts_peeraddr(struct socket *so, struct sockaddr **nam)
399 {
400 
401 	return (raw_usrreqs.pru_peeraddr(so, nam));
402 }
403 
404 /* pru_rcvd is EOPNOTSUPP */
405 /* pru_rcvoob is EOPNOTSUPP */
406 
407 static int
408 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
409 	 struct mbuf *control, struct thread *td)
410 {
411 
412 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
413 }
414 
415 /* pru_sense is null */
416 
417 static int
418 rts_shutdown(struct socket *so)
419 {
420 
421 	return (raw_usrreqs.pru_shutdown(so));
422 }
423 
424 static int
425 rts_sockaddr(struct socket *so, struct sockaddr **nam)
426 {
427 
428 	return (raw_usrreqs.pru_sockaddr(so, nam));
429 }
430 
431 static struct pr_usrreqs route_usrreqs = {
432 	.pru_abort =		rts_abort,
433 	.pru_attach =		rts_attach,
434 	.pru_bind =		rts_bind,
435 	.pru_connect =		rts_connect,
436 	.pru_detach =		rts_detach,
437 	.pru_disconnect =	rts_disconnect,
438 	.pru_peeraddr =		rts_peeraddr,
439 	.pru_send =		rts_send,
440 	.pru_shutdown =		rts_shutdown,
441 	.pru_sockaddr =		rts_sockaddr,
442 	.pru_close =		rts_close,
443 };
444 
445 #ifndef _SOCKADDR_UNION_DEFINED
446 #define	_SOCKADDR_UNION_DEFINED
447 /*
448  * The union of all possible address formats we handle.
449  */
450 union sockaddr_union {
451 	struct sockaddr		sa;
452 	struct sockaddr_in	sin;
453 	struct sockaddr_in6	sin6;
454 };
455 #endif /* _SOCKADDR_UNION_DEFINED */
456 
457 static int
458 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
459     struct nhop_object *nh, union sockaddr_union *saun, struct ucred *cred)
460 {
461 #if defined(INET) || defined(INET6)
462 	struct epoch_tracker et;
463 #endif
464 
465 	/* First, see if the returned address is part of the jail. */
466 	if (prison_if(cred, nh->nh_ifa->ifa_addr) == 0) {
467 		info->rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
468 		return (0);
469 	}
470 
471 	switch (info->rti_info[RTAX_DST]->sa_family) {
472 #ifdef INET
473 	case AF_INET:
474 	{
475 		struct in_addr ia;
476 		struct ifaddr *ifa;
477 		int found;
478 
479 		found = 0;
480 		/*
481 		 * Try to find an address on the given outgoing interface
482 		 * that belongs to the jail.
483 		 */
484 		NET_EPOCH_ENTER(et);
485 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
486 			struct sockaddr *sa;
487 			sa = ifa->ifa_addr;
488 			if (sa->sa_family != AF_INET)
489 				continue;
490 			ia = ((struct sockaddr_in *)sa)->sin_addr;
491 			if (prison_check_ip4(cred, &ia) == 0) {
492 				found = 1;
493 				break;
494 			}
495 		}
496 		NET_EPOCH_EXIT(et);
497 		if (!found) {
498 			/*
499 			 * As a last resort return the 'default' jail address.
500 			 */
501 			ia = ((struct sockaddr_in *)nh->nh_ifa->ifa_addr)->
502 			    sin_addr;
503 			if (prison_get_ip4(cred, &ia) != 0)
504 				return (ESRCH);
505 		}
506 		bzero(&saun->sin, sizeof(struct sockaddr_in));
507 		saun->sin.sin_len = sizeof(struct sockaddr_in);
508 		saun->sin.sin_family = AF_INET;
509 		saun->sin.sin_addr.s_addr = ia.s_addr;
510 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
511 		break;
512 	}
513 #endif
514 #ifdef INET6
515 	case AF_INET6:
516 	{
517 		struct in6_addr ia6;
518 		struct ifaddr *ifa;
519 		int found;
520 
521 		found = 0;
522 		/*
523 		 * Try to find an address on the given outgoing interface
524 		 * that belongs to the jail.
525 		 */
526 		NET_EPOCH_ENTER(et);
527 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
528 			struct sockaddr *sa;
529 			sa = ifa->ifa_addr;
530 			if (sa->sa_family != AF_INET6)
531 				continue;
532 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
533 			    &ia6, sizeof(struct in6_addr));
534 			if (prison_check_ip6(cred, &ia6) == 0) {
535 				found = 1;
536 				break;
537 			}
538 		}
539 		NET_EPOCH_EXIT(et);
540 		if (!found) {
541 			/*
542 			 * As a last resort return the 'default' jail address.
543 			 */
544 			ia6 = ((struct sockaddr_in6 *)nh->nh_ifa->ifa_addr)->
545 			    sin6_addr;
546 			if (prison_get_ip6(cred, &ia6) != 0)
547 				return (ESRCH);
548 		}
549 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
550 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
551 		saun->sin6.sin6_family = AF_INET6;
552 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
553 		if (sa6_recoverscope(&saun->sin6) != 0)
554 			return (ESRCH);
555 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
556 		break;
557 	}
558 #endif
559 	default:
560 		return (ESRCH);
561 	}
562 	return (0);
563 }
564 
565 static int
566 fill_blackholeinfo(struct rt_addrinfo *info, union sockaddr_union *saun)
567 {
568 	struct ifaddr *ifa;
569 	sa_family_t saf;
570 
571 	if (V_loif == NULL) {
572 		printf("Unable to add blackhole/reject nhop without loopback");
573 		return (ENOTSUP);
574 	}
575 	info->rti_ifp = V_loif;
576 
577 	saf = info->rti_info[RTAX_DST]->sa_family;
578 
579 	CK_STAILQ_FOREACH(ifa, &info->rti_ifp->if_addrhead, ifa_link) {
580 		if (ifa->ifa_addr->sa_family == saf) {
581 			info->rti_ifa = ifa;
582 			break;
583 		}
584 	}
585 	if (info->rti_ifa == NULL)
586 		return (ENOTSUP);
587 
588 	bzero(saun, sizeof(union sockaddr_union));
589 	switch (saf) {
590 #ifdef INET
591 	case AF_INET:
592 		saun->sin.sin_family = AF_INET;
593 		saun->sin.sin_len = sizeof(struct sockaddr_in);
594 		saun->sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
595 		break;
596 #endif
597 #ifdef INET6
598 	case AF_INET6:
599 		saun->sin6.sin6_family = AF_INET6;
600 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
601 		saun->sin6.sin6_addr = in6addr_loopback;
602 		break;
603 #endif
604 	default:
605 		return (ENOTSUP);
606 	}
607 	info->rti_info[RTAX_GATEWAY] = &saun->sa;
608 	info->rti_flags |= RTF_GATEWAY;
609 
610 	return (0);
611 }
612 
613 /*
614  * Fills in @info based on userland-provided @rtm message.
615  *
616  * Returns 0 on success.
617  */
618 static int
619 fill_addrinfo(struct rt_msghdr *rtm, int len, u_int fibnum, struct rt_addrinfo *info)
620 {
621 	int error;
622 	sa_family_t saf;
623 
624 	rtm->rtm_pid = curproc->p_pid;
625 	info->rti_addrs = rtm->rtm_addrs;
626 
627 	info->rti_mflags = rtm->rtm_inits;
628 	info->rti_rmx = &rtm->rtm_rmx;
629 
630 	/*
631 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
632 	 * link-local address because rtrequest requires addresses with
633 	 * embedded scope id.
634 	 */
635 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, info))
636 		return (EINVAL);
637 
638 	if (rtm->rtm_flags & RTF_RNH_LOCKED)
639 		return (EINVAL);
640 	info->rti_flags = rtm->rtm_flags;
641 	if (info->rti_info[RTAX_DST] == NULL ||
642 	    info->rti_info[RTAX_DST]->sa_family >= AF_MAX ||
643 	    (info->rti_info[RTAX_GATEWAY] != NULL &&
644 	     info->rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
645 		return (EINVAL);
646 	saf = info->rti_info[RTAX_DST]->sa_family;
647 	/*
648 	 * Verify that the caller has the appropriate privilege; RTM_GET
649 	 * is the only operation the non-superuser is allowed.
650 	 */
651 	if (rtm->rtm_type != RTM_GET) {
652 		error = priv_check(curthread, PRIV_NET_ROUTE);
653 		if (error != 0)
654 			return (error);
655 	}
656 
657 	/*
658 	 * The given gateway address may be an interface address.
659 	 * For example, issuing a "route change" command on a route
660 	 * entry that was created from a tunnel, and the gateway
661 	 * address given is the local end point. In this case the
662 	 * RTF_GATEWAY flag must be cleared or the destination will
663 	 * not be reachable even though there is no error message.
664 	 */
665 	if (info->rti_info[RTAX_GATEWAY] != NULL &&
666 	    info->rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
667 		struct rt_addrinfo ginfo;
668 		struct sockaddr *gdst;
669 		struct sockaddr_storage ss;
670 
671 		bzero(&ginfo, sizeof(ginfo));
672 		bzero(&ss, sizeof(ss));
673 		ss.ss_len = sizeof(ss);
674 
675 		ginfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ss;
676 		gdst = info->rti_info[RTAX_GATEWAY];
677 
678 		/*
679 		 * A host route through the loopback interface is
680 		 * installed for each interface adddress. In pre 8.0
681 		 * releases the interface address of a PPP link type
682 		 * is not reachable locally. This behavior is fixed as
683 		 * part of the new L2/L3 redesign and rewrite work. The
684 		 * signature of this interface address route is the
685 		 * AF_LINK sa_family type of the gateway, and the
686 		 * rt_ifp has the IFF_LOOPBACK flag set.
687 		 */
688 		if (rib_lookup_info(fibnum, gdst, NHR_REF, 0, &ginfo) == 0) {
689 			if (ss.ss_family == AF_LINK &&
690 			    ginfo.rti_ifp->if_flags & IFF_LOOPBACK) {
691 				info->rti_flags &= ~RTF_GATEWAY;
692 				info->rti_flags |= RTF_GWFLAG_COMPAT;
693 			}
694 			rib_free_info(&ginfo);
695 		}
696 	}
697 
698 	return (0);
699 }
700 
701 static struct nhop_object *
702 select_nhop(struct nhop_object *nh, const struct sockaddr *gw)
703 {
704 	if (!NH_IS_NHGRP(nh))
705 		return (nh);
706 #ifdef ROUTE_MPATH
707 	struct weightened_nhop *wn;
708 	uint32_t num_nhops;
709 	wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
710 	if (gw == NULL)
711 		return (wn[0].nh);
712 	for (int i = 0; i < num_nhops; i++) {
713 		if (match_nhop_gw(wn[i].nh, gw))
714 			return (wn[i].nh);
715 	}
716 #endif
717 	return (NULL);
718 }
719 
720 /*
721  * Handles RTM_GET message from routing socket, returning matching rt.
722  *
723  * Returns:
724  * 0 on success, with locked and referenced matching rt in @rt_nrt
725  * errno of failure
726  */
727 static int
728 handle_rtm_get(struct rt_addrinfo *info, u_int fibnum,
729     struct rt_msghdr *rtm, struct rib_cmd_info *rc)
730 {
731 	RIB_RLOCK_TRACKER;
732 	struct rib_head *rnh;
733 	struct nhop_object *nh;
734 	sa_family_t saf;
735 
736 	saf = info->rti_info[RTAX_DST]->sa_family;
737 
738 	rnh = rt_tables_get_rnh(fibnum, saf);
739 	if (rnh == NULL)
740 		return (EAFNOSUPPORT);
741 
742 	RIB_RLOCK(rnh);
743 
744 	if (info->rti_info[RTAX_NETMASK] == NULL) {
745 		/*
746 		 * Provide longest prefix match for
747 		 * address lookup (no mask).
748 		 * 'route -n get addr'
749 		 */
750 		rc->rc_rt = (struct rtentry *) rnh->rnh_matchaddr(
751 		    info->rti_info[RTAX_DST], &rnh->head);
752 	} else
753 		rc->rc_rt = (struct rtentry *) rnh->rnh_lookup(
754 		    info->rti_info[RTAX_DST],
755 		    info->rti_info[RTAX_NETMASK], &rnh->head);
756 
757 	if (rc->rc_rt == NULL) {
758 		RIB_RUNLOCK(rnh);
759 		return (ESRCH);
760 	}
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 	 * If performing proxied L2 entry insertion, and
769 	 * the actual PPP host entry is found, perform
770 	 * another search to retrieve the prefix route of
771 	 * the local end point of the PPP link.
772 	 * TODO: move this logic to userland.
773 	 */
774 	if (rtm->rtm_flags & RTF_ANNOUNCE) {
775 		struct sockaddr laddr;
776 
777 		if (nh->nh_ifp != NULL &&
778 		    nh->nh_ifp->if_type == IFT_PROPVIRTUAL) {
779 			struct ifaddr *ifa;
780 
781 			ifa = ifa_ifwithnet(info->rti_info[RTAX_DST], 1,
782 					RT_ALL_FIBS);
783 			if (ifa != NULL)
784 				rt_maskedcopy(ifa->ifa_addr,
785 					      &laddr,
786 					      ifa->ifa_netmask);
787 		} else
788 			rt_maskedcopy(nh->nh_ifa->ifa_addr,
789 				      &laddr,
790 				      nh->nh_ifa->ifa_netmask);
791 		/*
792 		 * refactor rt and no lock operation necessary
793 		 */
794 		rc->rc_rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr,
795 		    &rnh->head);
796 		if (rc->rc_rt == NULL) {
797 			RIB_RUNLOCK(rnh);
798 			return (ESRCH);
799 		}
800 		nh = select_nhop(rt_get_raw_nhop(rc->rc_rt), info->rti_info[RTAX_GATEWAY]);
801 		if (nh == NULL) {
802 			RIB_RUNLOCK(rnh);
803 			return (ESRCH);
804 		}
805 	}
806 	rc->rc_nh_new = nh;
807 	rc->rc_nh_weight = rc->rc_rt->rt_weight;
808 	RIB_RUNLOCK(rnh);
809 
810 	return (0);
811 }
812 
813 static void
814 init_sockaddrs_family(int family, struct sockaddr *dst, struct sockaddr *mask)
815 {
816 #ifdef INET
817 	if (family == AF_INET) {
818 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
819 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
820 
821 		bzero(dst4, sizeof(struct sockaddr_in));
822 		bzero(mask4, sizeof(struct sockaddr_in));
823 
824 		dst4->sin_family = AF_INET;
825 		dst4->sin_len = sizeof(struct sockaddr_in);
826 		mask4->sin_family = AF_INET;
827 		mask4->sin_len = sizeof(struct sockaddr_in);
828 	}
829 #endif
830 #ifdef INET6
831 	if (family == AF_INET6) {
832 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
833 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
834 
835 		bzero(dst6, sizeof(struct sockaddr_in6));
836 		bzero(mask6, sizeof(struct sockaddr_in6));
837 
838 		dst6->sin6_family = AF_INET6;
839 		dst6->sin6_len = sizeof(struct sockaddr_in6);
840 		mask6->sin6_family = AF_INET6;
841 		mask6->sin6_len = sizeof(struct sockaddr_in6);
842 	}
843 #endif
844 }
845 
846 static void
847 export_rtaddrs(const struct rtentry *rt, struct sockaddr *dst,
848     struct sockaddr *mask)
849 {
850 #ifdef INET
851 	if (dst->sa_family == AF_INET) {
852 		struct sockaddr_in *dst4 = (struct sockaddr_in *)dst;
853 		struct sockaddr_in *mask4 = (struct sockaddr_in *)mask;
854 		uint32_t scopeid = 0;
855 		rt_get_inet_prefix_pmask(rt, &dst4->sin_addr, &mask4->sin_addr,
856 		    &scopeid);
857 		return;
858 	}
859 #endif
860 #ifdef INET6
861 	if (dst->sa_family == AF_INET6) {
862 		struct sockaddr_in6 *dst6 = (struct sockaddr_in6 *)dst;
863 		struct sockaddr_in6 *mask6 = (struct sockaddr_in6 *)mask;
864 		uint32_t scopeid = 0;
865 		rt_get_inet6_prefix_pmask(rt, &dst6->sin6_addr,
866 		    &mask6->sin6_addr, &scopeid);
867 		dst6->sin6_scope_id = scopeid;
868 		return;
869 	}
870 #endif
871 }
872 
873 
874 /*
875  * Update sockaddrs, flags, etc in @prtm based on @rc data.
876  * rtm can be reallocated.
877  *
878  * Returns 0 on success, along with pointer to (potentially reallocated)
879  *  rtm.
880  *
881  */
882 static int
883 update_rtm_from_rc(struct rt_addrinfo *info, struct rt_msghdr **prtm,
884     int alloc_len, struct rib_cmd_info *rc, struct nhop_object *nh)
885 {
886 	struct walkarg w;
887 	union sockaddr_union saun;
888 	struct rt_msghdr *rtm, *orig_rtm = NULL;
889 	struct ifnet *ifp;
890 	int error, len;
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 	/* Check if we need to realloc storage */
923 	rtsock_msg_buffer(rtm->rtm_type, info, NULL, &len);
924 	if (len > alloc_len) {
925 		struct rt_msghdr *tmp_rtm;
926 
927 		tmp_rtm = malloc(len, M_TEMP, M_NOWAIT);
928 		if (tmp_rtm == NULL)
929 			return (ENOBUFS);
930 		bcopy(rtm, tmp_rtm, rtm->rtm_msglen);
931 		orig_rtm = rtm;
932 		rtm = tmp_rtm;
933 		alloc_len = len;
934 
935 		/*
936 		 * Delay freeing original rtm as info contains
937 		 * data referencing it.
938 		 */
939 	}
940 
941 	w.w_tmem = (caddr_t)rtm;
942 	w.w_tmemsize = alloc_len;
943 	rtsock_msg_buffer(rtm->rtm_type, info, &w, &len);
944 
945 	rtm->rtm_flags = rc->rc_rt->rte_flags | nhop_get_rtflags(nh);
946 	if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
947 		rtm->rtm_flags = RTF_GATEWAY |
948 			(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
949 	rt_getmetrics(rc->rc_rt, nh, &rtm->rtm_rmx);
950 	rtm->rtm_rmx.rmx_weight = rc->rc_nh_weight;
951 	rtm->rtm_addrs = info->rti_addrs;
952 
953 	if (orig_rtm != NULL)
954 		free(orig_rtm, M_TEMP);
955 	*prtm = rtm;
956 
957 	return (0);
958 }
959 
960 #ifdef ROUTE_MPATH
961 static void
962 save_del_notification(struct rib_cmd_info *rc, void *_cbdata)
963 {
964 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
965 
966 	if (rc->rc_cmd == RTM_DELETE)
967 		*rc_new = *rc;
968 }
969 
970 static void
971 save_add_notification(struct rib_cmd_info *rc, void *_cbdata)
972 {
973 	struct rib_cmd_info *rc_new = (struct rib_cmd_info *)_cbdata;
974 
975 	if (rc->rc_cmd == RTM_ADD)
976 		*rc_new = *rc;
977 }
978 #endif
979 
980 /*ARGSUSED*/
981 static int
982 route_output(struct mbuf *m, struct socket *so, ...)
983 {
984 	struct rt_msghdr *rtm = NULL;
985 	struct rtentry *rt = NULL;
986 	struct rt_addrinfo info;
987 	struct epoch_tracker et;
988 #ifdef INET6
989 	struct sockaddr_storage ss;
990 	struct sockaddr_in6 *sin6;
991 	int i, rti_need_deembed = 0;
992 #endif
993 	int alloc_len = 0, len, error = 0, fibnum;
994 	sa_family_t saf = AF_UNSPEC;
995 	struct rib_cmd_info rc;
996 	struct nhop_object *nh;
997 
998 	fibnum = so->so_fibnum;
999 #define senderr(e) { error = e; goto flush;}
1000 	if (m == NULL || ((m->m_len < sizeof(long)) &&
1001 		       (m = m_pullup(m, sizeof(long))) == NULL))
1002 		return (ENOBUFS);
1003 	if ((m->m_flags & M_PKTHDR) == 0)
1004 		panic("route_output");
1005 	NET_EPOCH_ENTER(et);
1006 	len = m->m_pkthdr.len;
1007 	if (len < sizeof(*rtm) ||
1008 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
1009 		senderr(EINVAL);
1010 
1011 	/*
1012 	 * Most of current messages are in range 200-240 bytes,
1013 	 * minimize possible re-allocation on reply using larger size
1014 	 * buffer aligned on 1k boundaty.
1015 	 */
1016 	alloc_len = roundup2(len, 1024);
1017 	if ((rtm = malloc(alloc_len, M_TEMP, M_NOWAIT)) == NULL)
1018 		senderr(ENOBUFS);
1019 
1020 	m_copydata(m, 0, len, (caddr_t)rtm);
1021 	bzero(&info, sizeof(info));
1022 	nh = NULL;
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, fibnum, &info)) != 0) {
1038 		senderr(error);
1039 	}
1040 
1041 	saf = info.rti_info[RTAX_DST]->sa_family;
1042 
1043 	/* support for new ARP code */
1044 	if (rtm->rtm_flags & RTF_LLDATA) {
1045 		error = lla_rt_output(rtm, &info);
1046 #ifdef INET6
1047 		if (error == 0)
1048 			rti_need_deembed = 1;
1049 #endif
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 			error = EINVAL;
1060 		if (error != 0)
1061 			senderr(error);
1062 		/* TODO: rebuild rtm from scratch */
1063 	}
1064 
1065 	switch (rtm->rtm_type) {
1066 	case RTM_ADD:
1067 	case RTM_CHANGE:
1068 		if (rtm->rtm_type == RTM_ADD) {
1069 			if (info.rti_info[RTAX_GATEWAY] == NULL)
1070 				senderr(EINVAL);
1071 		}
1072 		error = rib_action(fibnum, rtm->rtm_type, &info, &rc);
1073 		if (error == 0) {
1074 #ifdef INET6
1075 			rti_need_deembed = 1;
1076 #endif
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 = rc.rc_nh_new;
1087 			rtm->rtm_index = nh->nh_ifp->if_index;
1088 			rtm->rtm_flags = rc.rc_rt->rte_flags | nhop_get_rtflags(nh);
1089 		}
1090 		break;
1091 
1092 	case RTM_DELETE:
1093 		error = rib_action(fibnum, RTM_DELETE, &info, &rc);
1094 		if (error == 0) {
1095 #ifdef ROUTE_MPATH
1096 			if (NH_IS_NHGRP(rc.rc_nh_old) ||
1097 			    (rc.rc_nh_new && NH_IS_NHGRP(rc.rc_nh_new))) {
1098 				struct rib_cmd_info rc_simple = {};
1099 				rib_decompose_notification(&rc,
1100 				    save_del_notification, (void *)&rc_simple);
1101 				rc = rc_simple;
1102 			}
1103 #endif
1104 			nh = rc.rc_nh_old;
1105 			goto report;
1106 		}
1107 #ifdef INET6
1108 		/* rt_msg2() will not be used when RTM_DELETE fails. */
1109 		rti_need_deembed = 1;
1110 #endif
1111 		break;
1112 
1113 	case RTM_GET:
1114 		error = handle_rtm_get(&info, fibnum, rtm, &rc);
1115 		if (error != 0)
1116 			senderr(error);
1117 		nh = rc.rc_nh_new;
1118 
1119 report:
1120 		if (!can_export_rte(curthread->td_ucred,
1121 		    info.rti_info[RTAX_NETMASK] == NULL,
1122 		    info.rti_info[RTAX_DST])) {
1123 			senderr(ESRCH);
1124 		}
1125 
1126 		error = update_rtm_from_rc(&info, &rtm, alloc_len, &rc, nh);
1127 		/*
1128 		 * Note that some sockaddr pointers may have changed to
1129 		 * point to memory outsize @rtm. Some may be pointing
1130 		 * to the on-stack variables.
1131 		 * Given that, any pointer in @info CANNOT BE USED.
1132 		 */
1133 
1134 		/*
1135 		 * scopeid deembedding has been performed while
1136 		 * writing updated rtm in rtsock_msg_buffer().
1137 		 * With that in mind, skip deembedding procedure below.
1138 		 */
1139 #ifdef INET6
1140 		rti_need_deembed = 0;
1141 #endif
1142 		if (error != 0)
1143 			senderr(error);
1144 		break;
1145 
1146 	default:
1147 		senderr(EOPNOTSUPP);
1148 	}
1149 
1150 flush:
1151 	NET_EPOCH_EXIT(et);
1152 	rt = NULL;
1153 
1154 #ifdef INET6
1155 	if (rtm != NULL) {
1156 		if (rti_need_deembed) {
1157 			/* sin6_scope_id is recovered before sending rtm. */
1158 			sin6 = (struct sockaddr_in6 *)&ss;
1159 			for (i = 0; i < RTAX_MAX; i++) {
1160 				if (info.rti_info[i] == NULL)
1161 					continue;
1162 				if (info.rti_info[i]->sa_family != AF_INET6)
1163 					continue;
1164 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
1165 				if (sa6_recoverscope(sin6) == 0)
1166 					bcopy(sin6, info.rti_info[i],
1167 						    sizeof(*sin6));
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 rawcb *rp = 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 		rp = sotorawcb(so);
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 (rp) {
1221 			/*
1222 			 * XXX insure we don't get a copy by
1223 			 * invalidating our protocol
1224 			 */
1225 			unsigned short family = rp->rcb_proto.sp_family;
1226 			rp->rcb_proto.sp_family = 0;
1227 			rt_dispatch(m, saf);
1228 			rp->rcb_proto.sp_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 = rt->rt_expire ?
1245 	    rt->rt_expire - 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 			return (EINVAL);
1268 		/*
1269 		 * there are no more.. quit now
1270 		 * If there are more bits, they are in error.
1271 		 * I've seen this. route(1) can evidently generate these.
1272 		 * This causes kernel to core dump.
1273 		 * for compatibility, If we see this, point to a safe address.
1274 		 */
1275 		if (sa->sa_len == 0) {
1276 			rtinfo->rti_info[i] = &sa_zero;
1277 			return (0); /* should be EINVAL but for compat */
1278 		}
1279 		/* accept it */
1280 #ifdef INET6
1281 		if (sa->sa_family == AF_INET6)
1282 			sa6_embedscope((struct sockaddr_in6 *)sa,
1283 			    V_ip6_use_defzone);
1284 #endif
1285 		rtinfo->rti_info[i] = sa;
1286 		cp += SA_SIZE(sa);
1287 	}
1288 	return (0);
1289 }
1290 
1291 /*
1292  * Fill in @dmask with valid netmask leaving original @smask
1293  * intact. Mostly used with radix netmasks.
1294  */
1295 struct sockaddr *
1296 rtsock_fix_netmask(const struct sockaddr *dst, const struct sockaddr *smask,
1297     struct sockaddr_storage *dmask)
1298 {
1299 	if (dst == NULL || smask == NULL)
1300 		return (NULL);
1301 
1302 	memset(dmask, 0, dst->sa_len);
1303 	memcpy(dmask, smask, smask->sa_len);
1304 	dmask->ss_len = dst->sa_len;
1305 	dmask->ss_family = dst->sa_family;
1306 
1307 	return ((struct sockaddr *)dmask);
1308 }
1309 
1310 /*
1311  * Writes information related to @rtinfo object to newly-allocated mbuf.
1312  * Assumes MCLBYTES is enough to construct any message.
1313  * Used for OS notifications of vaious events (if/ifa announces,etc)
1314  *
1315  * Returns allocated mbuf or NULL on failure.
1316  */
1317 static struct mbuf *
1318 rtsock_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1319 {
1320 	struct sockaddr_storage ss;
1321 	struct rt_msghdr *rtm;
1322 	struct mbuf *m;
1323 	int i;
1324 	struct sockaddr *sa;
1325 #ifdef INET6
1326 	struct sockaddr_in6 *sin6;
1327 #endif
1328 	int len, dlen;
1329 
1330 	switch (type) {
1331 	case RTM_DELADDR:
1332 	case RTM_NEWADDR:
1333 		len = sizeof(struct ifa_msghdr);
1334 		break;
1335 
1336 	case RTM_DELMADDR:
1337 	case RTM_NEWMADDR:
1338 		len = sizeof(struct ifma_msghdr);
1339 		break;
1340 
1341 	case RTM_IFINFO:
1342 		len = sizeof(struct if_msghdr);
1343 		break;
1344 
1345 	case RTM_IFANNOUNCE:
1346 	case RTM_IEEE80211:
1347 		len = sizeof(struct if_announcemsghdr);
1348 		break;
1349 
1350 	default:
1351 		len = sizeof(struct rt_msghdr);
1352 	}
1353 
1354 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1355 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1356 	if (len > MHLEN)
1357 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1358 	else
1359 		m = m_gethdr(M_NOWAIT, MT_DATA);
1360 	if (m == NULL)
1361 		return (m);
1362 
1363 	m->m_pkthdr.len = m->m_len = len;
1364 	rtm = mtod(m, struct rt_msghdr *);
1365 	bzero((caddr_t)rtm, len);
1366 	for (i = 0; i < RTAX_MAX; i++) {
1367 		if ((sa = rtinfo->rti_info[i]) == NULL)
1368 			continue;
1369 		rtinfo->rti_addrs |= (1 << i);
1370 
1371 		dlen = SA_SIZE(sa);
1372 		KASSERT(dlen <= sizeof(ss),
1373 		    ("%s: sockaddr size overflow", __func__));
1374 		bzero(&ss, sizeof(ss));
1375 		bcopy(sa, &ss, sa->sa_len);
1376 		sa = (struct sockaddr *)&ss;
1377 #ifdef INET6
1378 		if (sa->sa_family == AF_INET6) {
1379 			sin6 = (struct sockaddr_in6 *)sa;
1380 			(void)sa6_recoverscope(sin6);
1381 		}
1382 #endif
1383 		m_copyback(m, len, dlen, (caddr_t)sa);
1384 		len += dlen;
1385 	}
1386 	if (m->m_pkthdr.len != len) {
1387 		m_freem(m);
1388 		return (NULL);
1389 	}
1390 	rtm->rtm_msglen = len;
1391 	rtm->rtm_version = RTM_VERSION;
1392 	rtm->rtm_type = type;
1393 	return (m);
1394 }
1395 
1396 /*
1397  * Writes information related to @rtinfo object to preallocated buffer.
1398  * Stores needed size in @plen. If @w is NULL, calculates size without
1399  * writing.
1400  * Used for sysctl dumps and rtsock answers (RTM_DEL/RTM_GET) generation.
1401  *
1402  * Returns 0 on success.
1403  *
1404  */
1405 static int
1406 rtsock_msg_buffer(int type, struct rt_addrinfo *rtinfo, struct walkarg *w, int *plen)
1407 {
1408 	struct sockaddr_storage ss;
1409 	int len, buflen = 0, dlen, i;
1410 	caddr_t cp = NULL;
1411 	struct rt_msghdr *rtm = NULL;
1412 #ifdef INET6
1413 	struct sockaddr_in6 *sin6;
1414 #endif
1415 #ifdef COMPAT_FREEBSD32
1416 	bool compat32 = false;
1417 #endif
1418 
1419 	switch (type) {
1420 	case RTM_DELADDR:
1421 	case RTM_NEWADDR:
1422 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1423 #ifdef COMPAT_FREEBSD32
1424 			if (w->w_req->flags & SCTL_MASK32) {
1425 				len = sizeof(struct ifa_msghdrl32);
1426 				compat32 = true;
1427 			} else
1428 #endif
1429 				len = sizeof(struct ifa_msghdrl);
1430 		} else
1431 			len = sizeof(struct ifa_msghdr);
1432 		break;
1433 
1434 	case RTM_IFINFO:
1435 #ifdef COMPAT_FREEBSD32
1436 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1437 			if (w->w_op == NET_RT_IFLISTL)
1438 				len = sizeof(struct if_msghdrl32);
1439 			else
1440 				len = sizeof(struct if_msghdr32);
1441 			compat32 = true;
1442 			break;
1443 		}
1444 #endif
1445 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1446 			len = sizeof(struct if_msghdrl);
1447 		else
1448 			len = sizeof(struct if_msghdr);
1449 		break;
1450 
1451 	case RTM_NEWMADDR:
1452 		len = sizeof(struct ifma_msghdr);
1453 		break;
1454 
1455 	default:
1456 		len = sizeof(struct rt_msghdr);
1457 	}
1458 
1459 	if (w != NULL) {
1460 		rtm = (struct rt_msghdr *)w->w_tmem;
1461 		buflen = w->w_tmemsize - len;
1462 		cp = (caddr_t)w->w_tmem + len;
1463 	}
1464 
1465 	rtinfo->rti_addrs = 0;
1466 	for (i = 0; i < RTAX_MAX; i++) {
1467 		struct sockaddr *sa;
1468 
1469 		if ((sa = rtinfo->rti_info[i]) == NULL)
1470 			continue;
1471 		rtinfo->rti_addrs |= (1 << i);
1472 #ifdef COMPAT_FREEBSD32
1473 		if (compat32)
1474 			dlen = SA_SIZE32(sa);
1475 		else
1476 #endif
1477 			dlen = SA_SIZE(sa);
1478 		if (cp != NULL && buflen >= dlen) {
1479 			KASSERT(dlen <= sizeof(ss),
1480 			    ("%s: sockaddr size overflow", __func__));
1481 			bzero(&ss, sizeof(ss));
1482 			bcopy(sa, &ss, sa->sa_len);
1483 			sa = (struct sockaddr *)&ss;
1484 #ifdef INET6
1485 			if (sa->sa_family == AF_INET6) {
1486 				sin6 = (struct sockaddr_in6 *)sa;
1487 				(void)sa6_recoverscope(sin6);
1488 			}
1489 #endif
1490 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1491 			cp += dlen;
1492 			buflen -= dlen;
1493 		} else if (cp != NULL) {
1494 			/*
1495 			 * Buffer too small. Count needed size
1496 			 * and return with error.
1497 			 */
1498 			cp = NULL;
1499 		}
1500 
1501 		len += dlen;
1502 	}
1503 
1504 	if (cp != NULL) {
1505 		dlen = ALIGN(len) - len;
1506 		if (buflen < dlen)
1507 			cp = NULL;
1508 		else {
1509 			bzero(cp, dlen);
1510 			cp += dlen;
1511 			buflen -= dlen;
1512 		}
1513 	}
1514 	len = ALIGN(len);
1515 
1516 	if (cp != NULL) {
1517 		/* fill header iff buffer is large enough */
1518 		rtm->rtm_version = RTM_VERSION;
1519 		rtm->rtm_type = type;
1520 		rtm->rtm_msglen = len;
1521 	}
1522 
1523 	*plen = len;
1524 
1525 	if (w != NULL && cp == NULL)
1526 		return (ENOBUFS);
1527 
1528 	return (0);
1529 }
1530 
1531 /*
1532  * This routine is called to generate a message from the routing
1533  * socket indicating that a redirect has occurred, a routing lookup
1534  * has failed, or that a protocol has detected timeouts to a particular
1535  * destination.
1536  */
1537 void
1538 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1539     int fibnum)
1540 {
1541 	struct rt_msghdr *rtm;
1542 	struct mbuf *m;
1543 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1544 
1545 	if (V_route_cb.any_count == 0)
1546 		return;
1547 	m = rtsock_msg_mbuf(type, rtinfo);
1548 	if (m == NULL)
1549 		return;
1550 
1551 	if (fibnum != RT_ALL_FIBS) {
1552 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1553 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1554 		M_SETFIB(m, fibnum);
1555 		m->m_flags |= RTS_FILTER_FIB;
1556 	}
1557 
1558 	rtm = mtod(m, struct rt_msghdr *);
1559 	rtm->rtm_flags = RTF_DONE | flags;
1560 	rtm->rtm_errno = error;
1561 	rtm->rtm_addrs = rtinfo->rti_addrs;
1562 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1563 }
1564 
1565 void
1566 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1567 {
1568 
1569 	rt_missmsg_fib(type, rtinfo, flags, error, RT_ALL_FIBS);
1570 }
1571 
1572 /*
1573  * This routine is called to generate a message from the routing
1574  * socket indicating that the status of a network interface has changed.
1575  */
1576 void
1577 rt_ifmsg(struct ifnet *ifp)
1578 {
1579 	struct if_msghdr *ifm;
1580 	struct mbuf *m;
1581 	struct rt_addrinfo info;
1582 
1583 	if (V_route_cb.any_count == 0)
1584 		return;
1585 	bzero((caddr_t)&info, sizeof(info));
1586 	m = rtsock_msg_mbuf(RTM_IFINFO, &info);
1587 	if (m == NULL)
1588 		return;
1589 	ifm = mtod(m, struct if_msghdr *);
1590 	ifm->ifm_index = ifp->if_index;
1591 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1592 	if_data_copy(ifp, &ifm->ifm_data);
1593 	ifm->ifm_addrs = 0;
1594 	rt_dispatch(m, AF_UNSPEC);
1595 }
1596 
1597 /*
1598  * Announce interface address arrival/withdraw.
1599  * Please do not call directly, use rt_addrmsg().
1600  * Assume input data to be valid.
1601  * Returns 0 on success.
1602  */
1603 int
1604 rtsock_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
1605 {
1606 	struct rt_addrinfo info;
1607 	struct sockaddr *sa;
1608 	int ncmd;
1609 	struct mbuf *m;
1610 	struct ifa_msghdr *ifam;
1611 	struct ifnet *ifp = ifa->ifa_ifp;
1612 	struct sockaddr_storage ss;
1613 
1614 	if (V_route_cb.any_count == 0)
1615 		return (0);
1616 
1617 	ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1618 
1619 	bzero((caddr_t)&info, sizeof(info));
1620 	info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1621 	info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1622 	info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
1623 	    info.rti_info[RTAX_IFA], ifa->ifa_netmask, &ss);
1624 	info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1625 	if ((m = rtsock_msg_mbuf(ncmd, &info)) == NULL)
1626 		return (ENOBUFS);
1627 	ifam = mtod(m, struct ifa_msghdr *);
1628 	ifam->ifam_index = ifp->if_index;
1629 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
1630 	ifam->ifam_flags = ifa->ifa_flags;
1631 	ifam->ifam_addrs = info.rti_addrs;
1632 
1633 	if (fibnum != RT_ALL_FIBS) {
1634 		M_SETFIB(m, fibnum);
1635 		m->m_flags |= RTS_FILTER_FIB;
1636 	}
1637 
1638 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1639 
1640 	return (0);
1641 }
1642 
1643 /*
1644  * Announce route addition/removal to rtsock based on @rt data.
1645  * Callers are advives to use rt_routemsg() instead of using this
1646  *  function directly.
1647  * Assume @rt data is consistent.
1648  *
1649  * Returns 0 on success.
1650  */
1651 int
1652 rtsock_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
1653     int fibnum)
1654 {
1655 	union sockaddr_union dst, mask;
1656 	struct rt_addrinfo info;
1657 
1658 	if (V_route_cb.any_count == 0)
1659 		return (0);
1660 
1661 	int family = rt_get_family(rt);
1662 	init_sockaddrs_family(family, &dst.sa, &mask.sa);
1663 	export_rtaddrs(rt, &dst.sa, &mask.sa);
1664 
1665 	bzero((caddr_t)&info, sizeof(info));
1666 	info.rti_info[RTAX_DST] = &dst.sa;
1667 	info.rti_info[RTAX_NETMASK] = &mask.sa;
1668 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1669 	info.rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
1670 	info.rti_ifp = nh->nh_ifp;
1671 
1672 	return (rtsock_routemsg_info(cmd, &info, fibnum));
1673 }
1674 
1675 int
1676 rtsock_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
1677 {
1678 	struct rt_msghdr *rtm;
1679 	struct sockaddr *sa;
1680 	struct mbuf *m;
1681 
1682 	if (V_route_cb.any_count == 0)
1683 		return (0);
1684 
1685 	if (info->rti_flags & RTF_HOST)
1686 		info->rti_info[RTAX_NETMASK] = NULL;
1687 
1688 	m = rtsock_msg_mbuf(cmd, info);
1689 	if (m == NULL)
1690 		return (ENOBUFS);
1691 
1692 	if (fibnum != RT_ALL_FIBS) {
1693 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1694 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1695 		M_SETFIB(m, fibnum);
1696 		m->m_flags |= RTS_FILTER_FIB;
1697 	}
1698 
1699 	rtm = mtod(m, struct rt_msghdr *);
1700 	rtm->rtm_addrs = info->rti_addrs;
1701 	if (info->rti_ifp != NULL)
1702 		rtm->rtm_index = info->rti_ifp->if_index;
1703 	/* Add RTF_DONE to indicate command 'completion' required by API */
1704 	info->rti_flags |= RTF_DONE;
1705 	/* Reported routes has to be up */
1706 	if (cmd == RTM_ADD || cmd == RTM_CHANGE)
1707 		info->rti_flags |= RTF_UP;
1708 	rtm->rtm_flags = info->rti_flags;
1709 
1710 	sa = info->rti_info[RTAX_DST];
1711 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1712 
1713 	return (0);
1714 }
1715 
1716 /*
1717  * This is the analogue to the rt_newaddrmsg which performs the same
1718  * function but for multicast group memberhips.  This is easier since
1719  * there is no route state to worry about.
1720  */
1721 void
1722 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1723 {
1724 	struct rt_addrinfo info;
1725 	struct mbuf *m = NULL;
1726 	struct ifnet *ifp = ifma->ifma_ifp;
1727 	struct ifma_msghdr *ifmam;
1728 
1729 	if (V_route_cb.any_count == 0)
1730 		return;
1731 
1732 	bzero((caddr_t)&info, sizeof(info));
1733 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1734 	if (ifp && ifp->if_addr)
1735 		info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1736 	else
1737 		info.rti_info[RTAX_IFP] = NULL;
1738 	/*
1739 	 * If a link-layer address is present, present it as a ``gateway''
1740 	 * (similarly to how ARP entries, e.g., are presented).
1741 	 */
1742 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1743 	m = rtsock_msg_mbuf(cmd, &info);
1744 	if (m == NULL)
1745 		return;
1746 	ifmam = mtod(m, struct ifma_msghdr *);
1747 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1748 	    __func__));
1749 	ifmam->ifmam_index = ifp->if_index;
1750 	ifmam->ifmam_addrs = info.rti_addrs;
1751 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1752 }
1753 
1754 static struct mbuf *
1755 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1756 	struct rt_addrinfo *info)
1757 {
1758 	struct if_announcemsghdr *ifan;
1759 	struct mbuf *m;
1760 
1761 	if (V_route_cb.any_count == 0)
1762 		return NULL;
1763 	bzero((caddr_t)info, sizeof(*info));
1764 	m = rtsock_msg_mbuf(type, info);
1765 	if (m != NULL) {
1766 		ifan = mtod(m, struct if_announcemsghdr *);
1767 		ifan->ifan_index = ifp->if_index;
1768 		strlcpy(ifan->ifan_name, ifp->if_xname,
1769 			sizeof(ifan->ifan_name));
1770 		ifan->ifan_what = what;
1771 	}
1772 	return m;
1773 }
1774 
1775 /*
1776  * This is called to generate routing socket messages indicating
1777  * IEEE80211 wireless events.
1778  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1779  */
1780 void
1781 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1782 {
1783 	struct mbuf *m;
1784 	struct rt_addrinfo info;
1785 
1786 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1787 	if (m != NULL) {
1788 		/*
1789 		 * Append the ieee80211 data.  Try to stick it in the
1790 		 * mbuf containing the ifannounce msg; otherwise allocate
1791 		 * a new mbuf and append.
1792 		 *
1793 		 * NB: we assume m is a single mbuf.
1794 		 */
1795 		if (data_len > M_TRAILINGSPACE(m)) {
1796 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1797 			if (n == NULL) {
1798 				m_freem(m);
1799 				return;
1800 			}
1801 			bcopy(data, mtod(n, void *), data_len);
1802 			n->m_len = data_len;
1803 			m->m_next = n;
1804 		} else if (data_len > 0) {
1805 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1806 			m->m_len += data_len;
1807 		}
1808 		if (m->m_flags & M_PKTHDR)
1809 			m->m_pkthdr.len += data_len;
1810 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1811 		rt_dispatch(m, AF_UNSPEC);
1812 	}
1813 }
1814 
1815 /*
1816  * This is called to generate routing socket messages indicating
1817  * network interface arrival and departure.
1818  */
1819 void
1820 rt_ifannouncemsg(struct ifnet *ifp, int what)
1821 {
1822 	struct mbuf *m;
1823 	struct rt_addrinfo info;
1824 
1825 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1826 	if (m != NULL)
1827 		rt_dispatch(m, AF_UNSPEC);
1828 }
1829 
1830 static void
1831 rt_dispatch(struct mbuf *m, sa_family_t saf)
1832 {
1833 	struct m_tag *tag;
1834 
1835 	/*
1836 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1837 	 * use when injecting the mbuf into the routing socket buffer from
1838 	 * the netisr.
1839 	 */
1840 	if (saf != AF_UNSPEC) {
1841 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1842 		    M_NOWAIT);
1843 		if (tag == NULL) {
1844 			m_freem(m);
1845 			return;
1846 		}
1847 		*(unsigned short *)(tag + 1) = saf;
1848 		m_tag_prepend(m, tag);
1849 	}
1850 #ifdef VIMAGE
1851 	if (V_loif)
1852 		m->m_pkthdr.rcvif = V_loif;
1853 	else {
1854 		m_freem(m);
1855 		return;
1856 	}
1857 #endif
1858 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1859 }
1860 
1861 /*
1862  * Checks if rte can be exported v.r.t jails/vnets.
1863  *
1864  * Returns 1 if it can, 0 otherwise.
1865  */
1866 static bool
1867 can_export_rte(struct ucred *td_ucred, bool rt_is_host,
1868     const struct sockaddr *rt_dst)
1869 {
1870 
1871 	if ((!rt_is_host) ? jailed_without_vnet(td_ucred)
1872 	    : prison_if(td_ucred, rt_dst) != 0)
1873 		return (false);
1874 	return (true);
1875 }
1876 
1877 
1878 /*
1879  * This is used in dumping the kernel table via sysctl().
1880  */
1881 static int
1882 sysctl_dumpentry(struct rtentry *rt, void *vw)
1883 {
1884 	struct walkarg *w = vw;
1885 	struct nhop_object *nh;
1886 	int error = 0;
1887 
1888 	NET_EPOCH_ASSERT();
1889 
1890 	export_rtaddrs(rt, w->dst, w->mask);
1891 	if (!can_export_rte(w->w_req->td->td_ucred, rt_is_host(rt), w->dst))
1892 		return (0);
1893 	nh = rt_get_raw_nhop(rt);
1894 #ifdef ROUTE_MPATH
1895 	if (NH_IS_NHGRP(nh)) {
1896 		struct weightened_nhop *wn;
1897 		uint32_t num_nhops;
1898 		wn = nhgrp_get_nhops((struct nhgrp_object *)nh, &num_nhops);
1899 		for (int i = 0; i < num_nhops; i++) {
1900 			error = sysctl_dumpnhop(rt, wn[i].nh, wn[i].weight, w);
1901 			if (error != 0)
1902 				return (error);
1903 		}
1904 	} else
1905 #endif
1906 		error = sysctl_dumpnhop(rt, nh, rt->rt_weight, w);
1907 
1908 	return (0);
1909 }
1910 
1911 
1912 static int
1913 sysctl_dumpnhop(struct rtentry *rt, struct nhop_object *nh, uint32_t weight,
1914     struct walkarg *w)
1915 {
1916 	struct rt_addrinfo info;
1917 	int error = 0, size;
1918 	uint32_t rtflags;
1919 
1920 	rtflags = nhop_get_rtflags(nh);
1921 
1922 	if (w->w_op == NET_RT_FLAGS && !(rtflags & w->w_arg))
1923 		return (0);
1924 
1925 	bzero((caddr_t)&info, sizeof(info));
1926 	info.rti_info[RTAX_DST] = w->dst;
1927 	info.rti_info[RTAX_GATEWAY] = &nh->gw_sa;
1928 	info.rti_info[RTAX_NETMASK] = (rtflags & RTF_HOST) ? NULL : w->mask;
1929 	info.rti_info[RTAX_GENMASK] = 0;
1930 	if (nh->nh_ifp && !(nh->nh_ifp->if_flags & IFF_DYING)) {
1931 		info.rti_info[RTAX_IFP] = nh->nh_ifp->if_addr->ifa_addr;
1932 		info.rti_info[RTAX_IFA] = nh->nh_ifa->ifa_addr;
1933 		if (nh->nh_ifp->if_flags & IFF_POINTOPOINT)
1934 			info.rti_info[RTAX_BRD] = nh->nh_ifa->ifa_dstaddr;
1935 	}
1936 	if ((error = rtsock_msg_buffer(RTM_GET, &info, w, &size)) != 0)
1937 		return (error);
1938 	if (w->w_req && w->w_tmem) {
1939 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1940 
1941 		bzero(&rtm->rtm_index,
1942 		    sizeof(*rtm) - offsetof(struct rt_msghdr, rtm_index));
1943 
1944 		/*
1945 		 * rte flags may consist of RTF_HOST (duplicated in nhop rtflags)
1946 		 * and RTF_UP (if entry is linked, which is always true here).
1947 		 * Given that, use nhop rtflags & add RTF_UP.
1948 		 */
1949 		rtm->rtm_flags = rtflags | RTF_UP;
1950 		if (rtm->rtm_flags & RTF_GWFLAG_COMPAT)
1951 			rtm->rtm_flags = RTF_GATEWAY |
1952 				(rtm->rtm_flags & ~RTF_GWFLAG_COMPAT);
1953 		rt_getmetrics(rt, nh, &rtm->rtm_rmx);
1954 		rtm->rtm_rmx.rmx_weight = weight;
1955 		rtm->rtm_index = nh->nh_ifp->if_index;
1956 		rtm->rtm_addrs = info.rti_addrs;
1957 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1958 		return (error);
1959 	}
1960 	return (error);
1961 }
1962 
1963 static int
1964 sysctl_iflist_ifml(struct ifnet *ifp, const struct if_data *src_ifd,
1965     struct rt_addrinfo *info, struct walkarg *w, int len)
1966 {
1967 	struct if_msghdrl *ifm;
1968 	struct if_data *ifd;
1969 
1970 	ifm = (struct if_msghdrl *)w->w_tmem;
1971 
1972 #ifdef COMPAT_FREEBSD32
1973 	if (w->w_req->flags & SCTL_MASK32) {
1974 		struct if_msghdrl32 *ifm32;
1975 
1976 		ifm32 = (struct if_msghdrl32 *)ifm;
1977 		ifm32->ifm_addrs = info->rti_addrs;
1978 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1979 		ifm32->ifm_index = ifp->if_index;
1980 		ifm32->_ifm_spare1 = 0;
1981 		ifm32->ifm_len = sizeof(*ifm32);
1982 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1983 		ifm32->_ifm_spare2 = 0;
1984 		ifd = &ifm32->ifm_data;
1985 	} else
1986 #endif
1987 	{
1988 		ifm->ifm_addrs = info->rti_addrs;
1989 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1990 		ifm->ifm_index = ifp->if_index;
1991 		ifm->_ifm_spare1 = 0;
1992 		ifm->ifm_len = sizeof(*ifm);
1993 		ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1994 		ifm->_ifm_spare2 = 0;
1995 		ifd = &ifm->ifm_data;
1996 	}
1997 
1998 	memcpy(ifd, src_ifd, sizeof(*ifd));
1999 
2000 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2001 }
2002 
2003 static int
2004 sysctl_iflist_ifm(struct ifnet *ifp, const struct if_data *src_ifd,
2005     struct rt_addrinfo *info, struct walkarg *w, int len)
2006 {
2007 	struct if_msghdr *ifm;
2008 	struct if_data *ifd;
2009 
2010 	ifm = (struct if_msghdr *)w->w_tmem;
2011 
2012 #ifdef COMPAT_FREEBSD32
2013 	if (w->w_req->flags & SCTL_MASK32) {
2014 		struct if_msghdr32 *ifm32;
2015 
2016 		ifm32 = (struct if_msghdr32 *)ifm;
2017 		ifm32->ifm_addrs = info->rti_addrs;
2018 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2019 		ifm32->ifm_index = ifp->if_index;
2020 		ifm32->_ifm_spare1 = 0;
2021 		ifd = &ifm32->ifm_data;
2022 	} else
2023 #endif
2024 	{
2025 		ifm->ifm_addrs = info->rti_addrs;
2026 		ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
2027 		ifm->ifm_index = ifp->if_index;
2028 		ifm->_ifm_spare1 = 0;
2029 		ifd = &ifm->ifm_data;
2030 	}
2031 
2032 	memcpy(ifd, src_ifd, sizeof(*ifd));
2033 
2034 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
2035 }
2036 
2037 static int
2038 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
2039     struct walkarg *w, int len)
2040 {
2041 	struct ifa_msghdrl *ifam;
2042 	struct if_data *ifd;
2043 
2044 	ifam = (struct ifa_msghdrl *)w->w_tmem;
2045 
2046 #ifdef COMPAT_FREEBSD32
2047 	if (w->w_req->flags & SCTL_MASK32) {
2048 		struct ifa_msghdrl32 *ifam32;
2049 
2050 		ifam32 = (struct ifa_msghdrl32 *)ifam;
2051 		ifam32->ifam_addrs = info->rti_addrs;
2052 		ifam32->ifam_flags = ifa->ifa_flags;
2053 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
2054 		ifam32->_ifam_spare1 = 0;
2055 		ifam32->ifam_len = sizeof(*ifam32);
2056 		ifam32->ifam_data_off =
2057 		    offsetof(struct ifa_msghdrl32, ifam_data);
2058 		ifam32->ifam_metric = ifa->ifa_ifp->if_metric;
2059 		ifd = &ifam32->ifam_data;
2060 	} else
2061 #endif
2062 	{
2063 		ifam->ifam_addrs = info->rti_addrs;
2064 		ifam->ifam_flags = ifa->ifa_flags;
2065 		ifam->ifam_index = ifa->ifa_ifp->if_index;
2066 		ifam->_ifam_spare1 = 0;
2067 		ifam->ifam_len = sizeof(*ifam);
2068 		ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
2069 		ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2070 		ifd = &ifam->ifam_data;
2071 	}
2072 
2073 	bzero(ifd, sizeof(*ifd));
2074 	ifd->ifi_datalen = sizeof(struct if_data);
2075 	ifd->ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
2076 	ifd->ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
2077 	ifd->ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
2078 	ifd->ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
2079 
2080 	/* Fixup if_data carp(4) vhid. */
2081 	if (carp_get_vhid_p != NULL)
2082 		ifd->ifi_vhid = (*carp_get_vhid_p)(ifa);
2083 
2084 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2085 }
2086 
2087 static int
2088 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
2089     struct walkarg *w, int len)
2090 {
2091 	struct ifa_msghdr *ifam;
2092 
2093 	ifam = (struct ifa_msghdr *)w->w_tmem;
2094 	ifam->ifam_addrs = info->rti_addrs;
2095 	ifam->ifam_flags = ifa->ifa_flags;
2096 	ifam->ifam_index = ifa->ifa_ifp->if_index;
2097 	ifam->_ifam_spare1 = 0;
2098 	ifam->ifam_metric = ifa->ifa_ifp->if_metric;
2099 
2100 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
2101 }
2102 
2103 static int
2104 sysctl_iflist(int af, struct walkarg *w)
2105 {
2106 	struct ifnet *ifp;
2107 	struct ifaddr *ifa;
2108 	struct if_data ifd;
2109 	struct rt_addrinfo info;
2110 	int len, error = 0;
2111 	struct sockaddr_storage ss;
2112 
2113 	bzero((caddr_t)&info, sizeof(info));
2114 	bzero(&ifd, sizeof(ifd));
2115 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2116 		if (w->w_arg && w->w_arg != ifp->if_index)
2117 			continue;
2118 		if_data_copy(ifp, &ifd);
2119 		ifa = ifp->if_addr;
2120 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
2121 		error = rtsock_msg_buffer(RTM_IFINFO, &info, w, &len);
2122 		if (error != 0)
2123 			goto done;
2124 		info.rti_info[RTAX_IFP] = NULL;
2125 		if (w->w_req && w->w_tmem) {
2126 			if (w->w_op == NET_RT_IFLISTL)
2127 				error = sysctl_iflist_ifml(ifp, &ifd, &info, w,
2128 				    len);
2129 			else
2130 				error = sysctl_iflist_ifm(ifp, &ifd, &info, w,
2131 				    len);
2132 			if (error)
2133 				goto done;
2134 		}
2135 		while ((ifa = CK_STAILQ_NEXT(ifa, ifa_link)) != NULL) {
2136 			if (af && af != ifa->ifa_addr->sa_family)
2137 				continue;
2138 			if (prison_if(w->w_req->td->td_ucred,
2139 			    ifa->ifa_addr) != 0)
2140 				continue;
2141 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
2142 			info.rti_info[RTAX_NETMASK] = rtsock_fix_netmask(
2143 			    ifa->ifa_addr, ifa->ifa_netmask, &ss);
2144 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
2145 			error = rtsock_msg_buffer(RTM_NEWADDR, &info, w, &len);
2146 			if (error != 0)
2147 				goto done;
2148 			if (w->w_req && w->w_tmem) {
2149 				if (w->w_op == NET_RT_IFLISTL)
2150 					error = sysctl_iflist_ifaml(ifa, &info,
2151 					    w, len);
2152 				else
2153 					error = sysctl_iflist_ifam(ifa, &info,
2154 					    w, len);
2155 				if (error)
2156 					goto done;
2157 			}
2158 		}
2159 		info.rti_info[RTAX_IFA] = NULL;
2160 		info.rti_info[RTAX_NETMASK] = NULL;
2161 		info.rti_info[RTAX_BRD] = NULL;
2162 	}
2163 done:
2164 	return (error);
2165 }
2166 
2167 static int
2168 sysctl_ifmalist(int af, struct walkarg *w)
2169 {
2170 	struct rt_addrinfo info;
2171 	struct ifaddr *ifa;
2172 	struct ifmultiaddr *ifma;
2173 	struct ifnet *ifp;
2174 	int error, len;
2175 
2176 	NET_EPOCH_ASSERT();
2177 
2178 	error = 0;
2179 	bzero((caddr_t)&info, sizeof(info));
2180 
2181 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2182 		if (w->w_arg && w->w_arg != ifp->if_index)
2183 			continue;
2184 		ifa = ifp->if_addr;
2185 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
2186 		CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2187 			if (af && af != ifma->ifma_addr->sa_family)
2188 				continue;
2189 			if (prison_if(w->w_req->td->td_ucred,
2190 			    ifma->ifma_addr) != 0)
2191 				continue;
2192 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
2193 			info.rti_info[RTAX_GATEWAY] =
2194 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
2195 			    ifma->ifma_lladdr : NULL;
2196 			error = rtsock_msg_buffer(RTM_NEWMADDR, &info, w, &len);
2197 			if (error != 0)
2198 				break;
2199 			if (w->w_req && w->w_tmem) {
2200 				struct ifma_msghdr *ifmam;
2201 
2202 				ifmam = (struct ifma_msghdr *)w->w_tmem;
2203 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
2204 				ifmam->ifmam_flags = 0;
2205 				ifmam->ifmam_addrs = info.rti_addrs;
2206 				ifmam->_ifmam_spare1 = 0;
2207 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
2208 				if (error != 0)
2209 					break;
2210 			}
2211 		}
2212 		if (error != 0)
2213 			break;
2214 	}
2215 	return (error);
2216 }
2217 
2218 static void
2219 rtable_sysctl_dump(uint32_t fibnum, int family, struct walkarg *w)
2220 {
2221 	union sockaddr_union sa_dst, sa_mask;
2222 
2223 	w->family = family;
2224 	w->dst = (struct sockaddr *)&sa_dst;
2225 	w->mask = (struct sockaddr *)&sa_mask;
2226 
2227 	init_sockaddrs_family(family, w->dst, w->mask);
2228 
2229 	rib_walk(fibnum, family, false, sysctl_dumpentry, w);
2230 }
2231 
2232 static int
2233 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
2234 {
2235 	struct epoch_tracker et;
2236 	int	*name = (int *)arg1;
2237 	u_int	namelen = arg2;
2238 	struct rib_head *rnh = NULL; /* silence compiler. */
2239 	int	i, lim, error = EINVAL;
2240 	int	fib = 0;
2241 	u_char	af;
2242 	struct	walkarg w;
2243 
2244 	name ++;
2245 	namelen--;
2246 	if (req->newptr)
2247 		return (EPERM);
2248 	if (name[1] == NET_RT_DUMP || name[1] == NET_RT_NHOP || name[1] == NET_RT_NHGRP) {
2249 		if (namelen == 3)
2250 			fib = req->td->td_proc->p_fibnum;
2251 		else if (namelen == 4)
2252 			fib = (name[3] == RT_ALL_FIBS) ?
2253 			    req->td->td_proc->p_fibnum : name[3];
2254 		else
2255 			return ((namelen < 3) ? EISDIR : ENOTDIR);
2256 		if (fib < 0 || fib >= rt_numfibs)
2257 			return (EINVAL);
2258 	} else if (namelen != 3)
2259 		return ((namelen < 3) ? EISDIR : ENOTDIR);
2260 	af = name[0];
2261 	if (af > AF_MAX)
2262 		return (EINVAL);
2263 	bzero(&w, sizeof(w));
2264 	w.w_op = name[1];
2265 	w.w_arg = name[2];
2266 	w.w_req = req;
2267 
2268 	error = sysctl_wire_old_buffer(req, 0);
2269 	if (error)
2270 		return (error);
2271 
2272 	/*
2273 	 * Allocate reply buffer in advance.
2274 	 * All rtsock messages has maximum length of u_short.
2275 	 */
2276 	w.w_tmemsize = 65536;
2277 	w.w_tmem = malloc(w.w_tmemsize, M_TEMP, M_WAITOK);
2278 
2279 	NET_EPOCH_ENTER(et);
2280 	switch (w.w_op) {
2281 	case NET_RT_DUMP:
2282 	case NET_RT_FLAGS:
2283 		if (af == 0) {			/* dump all tables */
2284 			i = 1;
2285 			lim = AF_MAX;
2286 		} else				/* dump only one table */
2287 			i = lim = af;
2288 
2289 		/*
2290 		 * take care of llinfo entries, the caller must
2291 		 * specify an AF
2292 		 */
2293 		if (w.w_op == NET_RT_FLAGS &&
2294 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
2295 			if (af != 0)
2296 				error = lltable_sysctl_dumparp(af, w.w_req);
2297 			else
2298 				error = EINVAL;
2299 			break;
2300 		}
2301 		/*
2302 		 * take care of routing entries
2303 		 */
2304 		for (error = 0; error == 0 && i <= lim; i++) {
2305 			rnh = rt_tables_get_rnh(fib, i);
2306 			if (rnh != NULL) {
2307 				rtable_sysctl_dump(fib, i, &w);
2308 			} else if (af != 0)
2309 				error = EAFNOSUPPORT;
2310 		}
2311 		break;
2312 	case NET_RT_NHOP:
2313 	case NET_RT_NHGRP:
2314 		/* Allow dumping one specific af/fib at a time */
2315 		if (namelen < 4) {
2316 			error = EINVAL;
2317 			break;
2318 		}
2319 		fib = name[3];
2320 		if (fib < 0 || fib > rt_numfibs) {
2321 			error = EINVAL;
2322 			break;
2323 		}
2324 		rnh = rt_tables_get_rnh(fib, af);
2325 		if (rnh == NULL) {
2326 			error = EAFNOSUPPORT;
2327 			break;
2328 		}
2329 		if (w.w_op == NET_RT_NHOP)
2330 			error = nhops_dump_sysctl(rnh, w.w_req);
2331 		else
2332 #ifdef ROUTE_MPATH
2333 			error = nhgrp_dump_sysctl(rnh, w.w_req);
2334 #else
2335 			error = ENOTSUP;
2336 #endif
2337 		break;
2338 	case NET_RT_IFLIST:
2339 	case NET_RT_IFLISTL:
2340 		error = sysctl_iflist(af, &w);
2341 		break;
2342 
2343 	case NET_RT_IFMALIST:
2344 		error = sysctl_ifmalist(af, &w);
2345 		break;
2346 	}
2347 	NET_EPOCH_EXIT(et);
2348 
2349 	free(w.w_tmem, M_TEMP);
2350 	return (error);
2351 }
2352 
2353 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD | CTLFLAG_MPSAFE,
2354     sysctl_rtsock, "Return route tables and interface/address lists");
2355 
2356 /*
2357  * Definitions of protocols supported in the ROUTE domain.
2358  */
2359 
2360 static struct domain routedomain;		/* or at least forward */
2361 
2362 static struct protosw routesw[] = {
2363 {
2364 	.pr_type =		SOCK_RAW,
2365 	.pr_domain =		&routedomain,
2366 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2367 	.pr_output =		route_output,
2368 	.pr_ctlinput =		raw_ctlinput,
2369 	.pr_init =		raw_init,
2370 	.pr_usrreqs =		&route_usrreqs
2371 }
2372 };
2373 
2374 static struct domain routedomain = {
2375 	.dom_family =		PF_ROUTE,
2376 	.dom_name =		"route",
2377 	.dom_protosw =		routesw,
2378 	.dom_protoswNPROTOSW =	&routesw[nitems(routesw)]
2379 };
2380 
2381 VNET_DOMAIN_SET(route);
2382