xref: /freebsd/sys/net/rtsock.c (revision 8d20be1e22095c27faf8fe8b2f0d089739cc742e)
1 /*-
2  * Copyright (c) 1988, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
30  * $FreeBSD$
31  */
32 #include "opt_compat.h"
33 #include "opt_sctp.h"
34 #include "opt_mpath.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/jail.h>
40 #include <sys/kernel.h>
41 #include <sys/domain.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/rwlock.h>
49 #include <sys/signalvar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/systm.h>
54 
55 #include <net/if.h>
56 #include <net/if_var.h>
57 #include <net/if_dl.h>
58 #include <net/if_llatbl.h>
59 #include <net/if_types.h>
60 #include <net/netisr.h>
61 #include <net/raw_cb.h>
62 #include <net/route.h>
63 #include <net/vnet.h>
64 
65 #include <netinet/in.h>
66 #include <netinet/if_ether.h>
67 #include <netinet/ip_carp.h>
68 #ifdef INET6
69 #include <netinet6/ip6_var.h>
70 #include <netinet6/scope6_var.h>
71 #endif
72 
73 #if defined(INET) || defined(INET6)
74 #ifdef SCTP
75 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
76 #endif /* SCTP */
77 #endif
78 
79 #ifdef COMPAT_FREEBSD32
80 #include <sys/mount.h>
81 #include <compat/freebsd32/freebsd32.h>
82 
83 struct if_data32 {
84 	uint8_t	ifi_type;
85 	uint8_t	ifi_physical;
86 	uint8_t	ifi_addrlen;
87 	uint8_t	ifi_hdrlen;
88 	uint8_t	ifi_link_state;
89 	uint8_t	ifi_vhid;
90 	uint8_t	ifi_baudrate_pf;
91 	uint8_t	ifi_datalen;
92 	uint32_t ifi_mtu;
93 	uint32_t ifi_metric;
94 	uint32_t ifi_baudrate;
95 	uint32_t ifi_ipackets;
96 	uint32_t ifi_ierrors;
97 	uint32_t ifi_opackets;
98 	uint32_t ifi_oerrors;
99 	uint32_t ifi_collisions;
100 	uint32_t ifi_ibytes;
101 	uint32_t ifi_obytes;
102 	uint32_t ifi_imcasts;
103 	uint32_t ifi_omcasts;
104 	uint32_t ifi_iqdrops;
105 	uint32_t ifi_noproto;
106 	uint32_t ifi_hwassist;
107 	int32_t	ifi_epoch;
108 	struct	timeval32 ifi_lastchange;
109 };
110 
111 struct if_msghdr32 {
112 	uint16_t ifm_msglen;
113 	uint8_t	ifm_version;
114 	uint8_t	ifm_type;
115 	int32_t	ifm_addrs;
116 	int32_t	ifm_flags;
117 	uint16_t ifm_index;
118 	struct	if_data32 ifm_data;
119 };
120 
121 struct if_msghdrl32 {
122 	uint16_t ifm_msglen;
123 	uint8_t	ifm_version;
124 	uint8_t	ifm_type;
125 	int32_t	ifm_addrs;
126 	int32_t	ifm_flags;
127 	uint16_t ifm_index;
128 	uint16_t _ifm_spare1;
129 	uint16_t ifm_len;
130 	uint16_t ifm_data_off;
131 	struct	if_data32 ifm_data;
132 };
133 
134 struct ifa_msghdrl32 {
135 	uint16_t ifam_msglen;
136 	uint8_t	ifam_version;
137 	uint8_t	ifam_type;
138 	int32_t	ifam_addrs;
139 	int32_t	ifam_flags;
140 	uint16_t ifam_index;
141 	uint16_t _ifam_spare1;
142 	uint16_t ifam_len;
143 	uint16_t ifam_data_off;
144 	int32_t	ifam_metric;
145 	struct	if_data32 ifam_data;
146 };
147 #endif /* COMPAT_FREEBSD32 */
148 
149 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
150 
151 /* NB: these are not modified */
152 static struct	sockaddr route_src = { 2, PF_ROUTE, };
153 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
154 
155 /* These are external hooks for CARP. */
156 int	(*carp_get_vhid_p)(struct ifaddr *);
157 
158 /*
159  * Used by rtsock/raw_input callback code to decide whether to filter the update
160  * notification to a socket bound to a particular FIB.
161  */
162 #define	RTS_FILTER_FIB	M_PROTO8
163 #define	RTS_ALLFIBS	-1
164 
165 static struct {
166 	int	ip_count;	/* attached w/ AF_INET */
167 	int	ip6_count;	/* attached w/ AF_INET6 */
168 	int	ipx_count;	/* attached w/ AF_IPX */
169 	int	any_count;	/* total attached */
170 } route_cb;
171 
172 struct mtx rtsock_mtx;
173 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
174 
175 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
176 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
177 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
178 
179 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
180 
181 struct walkarg {
182 	int	w_tmemsize;
183 	int	w_op, w_arg;
184 	caddr_t	w_tmem;
185 	struct sysctl_req *w_req;
186 };
187 
188 static void	rts_input(struct mbuf *m);
189 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
190 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
191 			caddr_t cp, struct walkarg *w);
192 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
193 			struct rt_addrinfo *rtinfo);
194 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
195 static int	sysctl_iflist(int af, struct walkarg *w);
196 static int	sysctl_ifmalist(int af, struct walkarg *w);
197 static int	route_output(struct mbuf *m, struct socket *so);
198 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
199 			struct rt_metrics_lite *out);
200 static void	rt_getmetrics(const struct rt_metrics_lite *in,
201 			struct rt_metrics *out);
202 static void	rt_dispatch(struct mbuf *, sa_family_t);
203 
204 static struct netisr_handler rtsock_nh = {
205 	.nh_name = "rtsock",
206 	.nh_handler = rts_input,
207 	.nh_proto = NETISR_ROUTE,
208 	.nh_policy = NETISR_POLICY_SOURCE,
209 };
210 
211 static int
212 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
213 {
214 	int error, qlimit;
215 
216 	netisr_getqlimit(&rtsock_nh, &qlimit);
217 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
218         if (error || !req->newptr)
219                 return (error);
220 	if (qlimit < 1)
221 		return (EINVAL);
222 	return (netisr_setqlimit(&rtsock_nh, qlimit));
223 }
224 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
225     0, 0, sysctl_route_netisr_maxqlen, "I",
226     "maximum routing socket dispatch queue length");
227 
228 static void
229 rts_init(void)
230 {
231 	int tmp;
232 
233 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
234 		rtsock_nh.nh_qlimit = tmp;
235 	netisr_register(&rtsock_nh);
236 }
237 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
238 
239 static int
240 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
241     struct rawcb *rp)
242 {
243 	int fibnum;
244 
245 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
246 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
247 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
248 
249 	/* No filtering requested. */
250 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
251 		return (0);
252 
253 	/* Check if it is a rts and the fib matches the one of the socket. */
254 	fibnum = M_GETFIB(m);
255 	if (proto->sp_family != PF_ROUTE ||
256 	    rp->rcb_socket == NULL ||
257 	    rp->rcb_socket->so_fibnum == fibnum)
258 		return (0);
259 
260 	/* Filtering requested and no match, the socket shall be skipped. */
261 	return (1);
262 }
263 
264 static void
265 rts_input(struct mbuf *m)
266 {
267 	struct sockproto route_proto;
268 	unsigned short *family;
269 	struct m_tag *tag;
270 
271 	route_proto.sp_family = PF_ROUTE;
272 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
273 	if (tag != NULL) {
274 		family = (unsigned short *)(tag + 1);
275 		route_proto.sp_protocol = *family;
276 		m_tag_delete(m, tag);
277 	} else
278 		route_proto.sp_protocol = 0;
279 
280 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
281 }
282 
283 /*
284  * It really doesn't make any sense at all for this code to share much
285  * with raw_usrreq.c, since its functionality is so restricted.  XXX
286  */
287 static void
288 rts_abort(struct socket *so)
289 {
290 
291 	raw_usrreqs.pru_abort(so);
292 }
293 
294 static void
295 rts_close(struct socket *so)
296 {
297 
298 	raw_usrreqs.pru_close(so);
299 }
300 
301 /* pru_accept is EOPNOTSUPP */
302 
303 static int
304 rts_attach(struct socket *so, int proto, struct thread *td)
305 {
306 	struct rawcb *rp;
307 	int error;
308 
309 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
310 
311 	/* XXX */
312 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
313 	if (rp == NULL)
314 		return ENOBUFS;
315 
316 	so->so_pcb = (caddr_t)rp;
317 	so->so_fibnum = td->td_proc->p_fibnum;
318 	error = raw_attach(so, proto);
319 	rp = sotorawcb(so);
320 	if (error) {
321 		so->so_pcb = NULL;
322 		free(rp, M_PCB);
323 		return error;
324 	}
325 	RTSOCK_LOCK();
326 	switch(rp->rcb_proto.sp_protocol) {
327 	case AF_INET:
328 		route_cb.ip_count++;
329 		break;
330 	case AF_INET6:
331 		route_cb.ip6_count++;
332 		break;
333 	case AF_IPX:
334 		route_cb.ipx_count++;
335 		break;
336 	}
337 	route_cb.any_count++;
338 	RTSOCK_UNLOCK();
339 	soisconnected(so);
340 	so->so_options |= SO_USELOOPBACK;
341 	return 0;
342 }
343 
344 static int
345 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
346 {
347 
348 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
349 }
350 
351 static int
352 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
353 {
354 
355 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
356 }
357 
358 /* pru_connect2 is EOPNOTSUPP */
359 /* pru_control is EOPNOTSUPP */
360 
361 static void
362 rts_detach(struct socket *so)
363 {
364 	struct rawcb *rp = sotorawcb(so);
365 
366 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
367 
368 	RTSOCK_LOCK();
369 	switch(rp->rcb_proto.sp_protocol) {
370 	case AF_INET:
371 		route_cb.ip_count--;
372 		break;
373 	case AF_INET6:
374 		route_cb.ip6_count--;
375 		break;
376 	case AF_IPX:
377 		route_cb.ipx_count--;
378 		break;
379 	}
380 	route_cb.any_count--;
381 	RTSOCK_UNLOCK();
382 	raw_usrreqs.pru_detach(so);
383 }
384 
385 static int
386 rts_disconnect(struct socket *so)
387 {
388 
389 	return (raw_usrreqs.pru_disconnect(so));
390 }
391 
392 /* pru_listen is EOPNOTSUPP */
393 
394 static int
395 rts_peeraddr(struct socket *so, struct sockaddr **nam)
396 {
397 
398 	return (raw_usrreqs.pru_peeraddr(so, nam));
399 }
400 
401 /* pru_rcvd is EOPNOTSUPP */
402 /* pru_rcvoob is EOPNOTSUPP */
403 
404 static int
405 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
406 	 struct mbuf *control, struct thread *td)
407 {
408 
409 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
410 }
411 
412 /* pru_sense is null */
413 
414 static int
415 rts_shutdown(struct socket *so)
416 {
417 
418 	return (raw_usrreqs.pru_shutdown(so));
419 }
420 
421 static int
422 rts_sockaddr(struct socket *so, struct sockaddr **nam)
423 {
424 
425 	return (raw_usrreqs.pru_sockaddr(so, nam));
426 }
427 
428 static struct pr_usrreqs route_usrreqs = {
429 	.pru_abort =		rts_abort,
430 	.pru_attach =		rts_attach,
431 	.pru_bind =		rts_bind,
432 	.pru_connect =		rts_connect,
433 	.pru_detach =		rts_detach,
434 	.pru_disconnect =	rts_disconnect,
435 	.pru_peeraddr =		rts_peeraddr,
436 	.pru_send =		rts_send,
437 	.pru_shutdown =		rts_shutdown,
438 	.pru_sockaddr =		rts_sockaddr,
439 	.pru_close =		rts_close,
440 };
441 
442 #ifndef _SOCKADDR_UNION_DEFINED
443 #define	_SOCKADDR_UNION_DEFINED
444 /*
445  * The union of all possible address formats we handle.
446  */
447 union sockaddr_union {
448 	struct sockaddr		sa;
449 	struct sockaddr_in	sin;
450 	struct sockaddr_in6	sin6;
451 };
452 #endif /* _SOCKADDR_UNION_DEFINED */
453 
454 static int
455 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
456     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
457 {
458 
459 	/* First, see if the returned address is part of the jail. */
460 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
461 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
462 		return (0);
463 	}
464 
465 	switch (info->rti_info[RTAX_DST]->sa_family) {
466 #ifdef INET
467 	case AF_INET:
468 	{
469 		struct in_addr ia;
470 		struct ifaddr *ifa;
471 		int found;
472 
473 		found = 0;
474 		/*
475 		 * Try to find an address on the given outgoing interface
476 		 * that belongs to the jail.
477 		 */
478 		IF_ADDR_RLOCK(ifp);
479 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
480 			struct sockaddr *sa;
481 			sa = ifa->ifa_addr;
482 			if (sa->sa_family != AF_INET)
483 				continue;
484 			ia = ((struct sockaddr_in *)sa)->sin_addr;
485 			if (prison_check_ip4(cred, &ia) == 0) {
486 				found = 1;
487 				break;
488 			}
489 		}
490 		IF_ADDR_RUNLOCK(ifp);
491 		if (!found) {
492 			/*
493 			 * As a last resort return the 'default' jail address.
494 			 */
495 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
496 			    sin_addr;
497 			if (prison_get_ip4(cred, &ia) != 0)
498 				return (ESRCH);
499 		}
500 		bzero(&saun->sin, sizeof(struct sockaddr_in));
501 		saun->sin.sin_len = sizeof(struct sockaddr_in);
502 		saun->sin.sin_family = AF_INET;
503 		saun->sin.sin_addr.s_addr = ia.s_addr;
504 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
505 		break;
506 	}
507 #endif
508 #ifdef INET6
509 	case AF_INET6:
510 	{
511 		struct in6_addr ia6;
512 		struct ifaddr *ifa;
513 		int found;
514 
515 		found = 0;
516 		/*
517 		 * Try to find an address on the given outgoing interface
518 		 * that belongs to the jail.
519 		 */
520 		IF_ADDR_RLOCK(ifp);
521 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
522 			struct sockaddr *sa;
523 			sa = ifa->ifa_addr;
524 			if (sa->sa_family != AF_INET6)
525 				continue;
526 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
527 			    &ia6, sizeof(struct in6_addr));
528 			if (prison_check_ip6(cred, &ia6) == 0) {
529 				found = 1;
530 				break;
531 			}
532 		}
533 		IF_ADDR_RUNLOCK(ifp);
534 		if (!found) {
535 			/*
536 			 * As a last resort return the 'default' jail address.
537 			 */
538 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
539 			    sin6_addr;
540 			if (prison_get_ip6(cred, &ia6) != 0)
541 				return (ESRCH);
542 		}
543 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
544 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
545 		saun->sin6.sin6_family = AF_INET6;
546 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
547 		if (sa6_recoverscope(&saun->sin6) != 0)
548 			return (ESRCH);
549 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
550 		break;
551 	}
552 #endif
553 	default:
554 		return (ESRCH);
555 	}
556 	return (0);
557 }
558 
559 /*ARGSUSED*/
560 static int
561 route_output(struct mbuf *m, struct socket *so)
562 {
563 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
564 	struct rt_msghdr *rtm = NULL;
565 	struct rtentry *rt = NULL;
566 	struct radix_node_head *rnh;
567 	struct rt_addrinfo info;
568 #ifdef INET6
569 	struct sockaddr_storage ss;
570 	struct sockaddr_in6 *sin6;
571 	int i, rti_need_deembed = 0;
572 #endif
573 	int len, error = 0;
574 	struct ifnet *ifp = NULL;
575 	union sockaddr_union saun;
576 	sa_family_t saf = AF_UNSPEC;
577 
578 #define senderr(e) { error = e; goto flush;}
579 	if (m == NULL || ((m->m_len < sizeof(long)) &&
580 		       (m = m_pullup(m, sizeof(long))) == NULL))
581 		return (ENOBUFS);
582 	if ((m->m_flags & M_PKTHDR) == 0)
583 		panic("route_output");
584 	len = m->m_pkthdr.len;
585 	if (len < sizeof(*rtm) ||
586 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
587 		info.rti_info[RTAX_DST] = NULL;
588 		senderr(EINVAL);
589 	}
590 	R_Malloc(rtm, struct rt_msghdr *, len);
591 	if (rtm == NULL) {
592 		info.rti_info[RTAX_DST] = NULL;
593 		senderr(ENOBUFS);
594 	}
595 	m_copydata(m, 0, len, (caddr_t)rtm);
596 	if (rtm->rtm_version != RTM_VERSION) {
597 		info.rti_info[RTAX_DST] = NULL;
598 		senderr(EPROTONOSUPPORT);
599 	}
600 	rtm->rtm_pid = curproc->p_pid;
601 	bzero(&info, sizeof(info));
602 	info.rti_addrs = rtm->rtm_addrs;
603 	/*
604 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
605 	 * link-local address because rtrequest requires addresses with
606 	 * embedded scope id.
607 	 */
608 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
609 		info.rti_info[RTAX_DST] = NULL;
610 		senderr(EINVAL);
611 	}
612 	info.rti_flags = rtm->rtm_flags;
613 	if (info.rti_info[RTAX_DST] == NULL ||
614 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
615 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
616 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
617 		senderr(EINVAL);
618 	saf = info.rti_info[RTAX_DST]->sa_family;
619 	/*
620 	 * Verify that the caller has the appropriate privilege; RTM_GET
621 	 * is the only operation the non-superuser is allowed.
622 	 */
623 	if (rtm->rtm_type != RTM_GET) {
624 		error = priv_check(curthread, PRIV_NET_ROUTE);
625 		if (error)
626 			senderr(error);
627 	}
628 
629 	/*
630 	 * The given gateway address may be an interface address.
631 	 * For example, issuing a "route change" command on a route
632 	 * entry that was created from a tunnel, and the gateway
633 	 * address given is the local end point. In this case the
634 	 * RTF_GATEWAY flag must be cleared or the destination will
635 	 * not be reachable even though there is no error message.
636 	 */
637 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
638 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
639 		struct route gw_ro;
640 
641 		bzero(&gw_ro, sizeof(gw_ro));
642 		gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
643 		rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
644 		/*
645 		 * A host route through the loopback interface is
646 		 * installed for each interface adddress. In pre 8.0
647 		 * releases the interface address of a PPP link type
648 		 * is not reachable locally. This behavior is fixed as
649 		 * part of the new L2/L3 redesign and rewrite work. The
650 		 * signature of this interface address route is the
651 		 * AF_LINK sa_family type of the rt_gateway, and the
652 		 * rt_ifp has the IFF_LOOPBACK flag set.
653 		 */
654 		if (gw_ro.ro_rt != NULL &&
655 		    gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
656 		    gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) {
657 			info.rti_flags &= ~RTF_GATEWAY;
658 			info.rti_flags |= RTF_GWFLAG_COMPAT;
659 		}
660 		if (gw_ro.ro_rt != NULL)
661 			RTFREE(gw_ro.ro_rt);
662 	}
663 
664 	switch (rtm->rtm_type) {
665 		struct rtentry *saved_nrt;
666 
667 	case RTM_ADD:
668 		if (info.rti_info[RTAX_GATEWAY] == NULL)
669 			senderr(EINVAL);
670 		saved_nrt = NULL;
671 
672 		/* support for new ARP code */
673 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
674 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
675 			error = lla_rt_output(rtm, &info);
676 #ifdef INET6
677 			if (error == 0)
678 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
679 #endif
680 			break;
681 		}
682 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
683 		    so->so_fibnum);
684 		if (error == 0 && saved_nrt) {
685 #ifdef INET6
686 			rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
687 #endif
688 			RT_LOCK(saved_nrt);
689 			rt_setmetrics(rtm->rtm_inits,
690 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
691 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
692 			RT_REMREF(saved_nrt);
693 			RT_UNLOCK(saved_nrt);
694 		}
695 		break;
696 
697 	case RTM_DELETE:
698 		saved_nrt = NULL;
699 		/* support for new ARP code */
700 		if (info.rti_info[RTAX_GATEWAY] &&
701 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
702 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
703 			error = lla_rt_output(rtm, &info);
704 #ifdef INET6
705 			if (error == 0)
706 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
707 #endif
708 			break;
709 		}
710 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
711 		    so->so_fibnum);
712 		if (error == 0) {
713 			RT_LOCK(saved_nrt);
714 			rt = saved_nrt;
715 			goto report;
716 		}
717 #ifdef INET6
718 		/* rt_msg2() will not be used when RTM_DELETE fails. */
719 		rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
720 #endif
721 		break;
722 
723 	case RTM_GET:
724 	case RTM_CHANGE:
725 	case RTM_LOCK:
726 		rnh = rt_tables_get_rnh(so->so_fibnum,
727 		    info.rti_info[RTAX_DST]->sa_family);
728 		if (rnh == NULL)
729 			senderr(EAFNOSUPPORT);
730 		RADIX_NODE_HEAD_RLOCK(rnh);
731 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
732 			info.rti_info[RTAX_NETMASK], rnh);
733 		if (rt == NULL) {	/* XXX looks bogus */
734 			RADIX_NODE_HEAD_RUNLOCK(rnh);
735 			senderr(ESRCH);
736 		}
737 #ifdef RADIX_MPATH
738 		/*
739 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
740 		 * we require users to specify a matching RTAX_GATEWAY.
741 		 *
742 		 * for RTM_GET, gate is optional even with multipath.
743 		 * if gate == NULL the first match is returned.
744 		 * (no need to call rt_mpath_matchgate if gate == NULL)
745 		 */
746 		if (rn_mpath_capable(rnh) &&
747 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
748 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
749 			if (!rt) {
750 				RADIX_NODE_HEAD_RUNLOCK(rnh);
751 				senderr(ESRCH);
752 			}
753 		}
754 #endif
755 		/*
756 		 * If performing proxied L2 entry insertion, and
757 		 * the actual PPP host entry is found, perform
758 		 * another search to retrieve the prefix route of
759 		 * the local end point of the PPP link.
760 		 */
761 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
762 			struct sockaddr laddr;
763 
764 			if (rt->rt_ifp != NULL &&
765 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
766 				struct ifaddr *ifa;
767 
768 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
769 				if (ifa != NULL)
770 					rt_maskedcopy(ifa->ifa_addr,
771 						      &laddr,
772 						      ifa->ifa_netmask);
773 			} else
774 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
775 					      &laddr,
776 					      rt->rt_ifa->ifa_netmask);
777 			/*
778 			 * refactor rt and no lock operation necessary
779 			 */
780 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
781 			if (rt == NULL) {
782 				RADIX_NODE_HEAD_RUNLOCK(rnh);
783 				senderr(ESRCH);
784 			}
785 		}
786 		RT_LOCK(rt);
787 		RT_ADDREF(rt);
788 		RADIX_NODE_HEAD_RUNLOCK(rnh);
789 
790 		/*
791 		 * Fix for PR: 82974
792 		 *
793 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
794 		 * returns a perfect match in case a netmask is
795 		 * specified.  For host routes only a longest prefix
796 		 * match is returned so it is necessary to compare the
797 		 * existence of the netmask.  If both have a netmask
798 		 * rnh_lookup() did a perfect match and if none of them
799 		 * have a netmask both are host routes which is also a
800 		 * perfect match.
801 		 */
802 
803 		if (rtm->rtm_type != RTM_GET &&
804 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
805 			RT_UNLOCK(rt);
806 			senderr(ESRCH);
807 		}
808 
809 		switch(rtm->rtm_type) {
810 
811 		case RTM_GET:
812 		report:
813 			RT_LOCK_ASSERT(rt);
814 			if ((rt->rt_flags & RTF_HOST) == 0
815 			    ? jailed_without_vnet(curthread->td_ucred)
816 			    : prison_if(curthread->td_ucred,
817 			    rt_key(rt)) != 0) {
818 				RT_UNLOCK(rt);
819 				senderr(ESRCH);
820 			}
821 			info.rti_info[RTAX_DST] = rt_key(rt);
822 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
823 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
824 			info.rti_info[RTAX_GENMASK] = 0;
825 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
826 				ifp = rt->rt_ifp;
827 				if (ifp) {
828 					info.rti_info[RTAX_IFP] =
829 					    ifp->if_addr->ifa_addr;
830 					error = rtm_get_jailed(&info, ifp, rt,
831 					    &saun, curthread->td_ucred);
832 					if (error != 0) {
833 						RT_UNLOCK(rt);
834 						senderr(error);
835 					}
836 					if (ifp->if_flags & IFF_POINTOPOINT)
837 						info.rti_info[RTAX_BRD] =
838 						    rt->rt_ifa->ifa_dstaddr;
839 					rtm->rtm_index = ifp->if_index;
840 				} else {
841 					info.rti_info[RTAX_IFP] = NULL;
842 					info.rti_info[RTAX_IFA] = NULL;
843 				}
844 			} else if ((ifp = rt->rt_ifp) != NULL) {
845 				rtm->rtm_index = ifp->if_index;
846 			}
847 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
848 			if (len > rtm->rtm_msglen) {
849 				struct rt_msghdr *new_rtm;
850 				R_Malloc(new_rtm, struct rt_msghdr *, len);
851 				if (new_rtm == NULL) {
852 					RT_UNLOCK(rt);
853 					senderr(ENOBUFS);
854 				}
855 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
856 				Free(rtm); rtm = new_rtm;
857 			}
858 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
859 			if (rt->rt_flags & RTF_GWFLAG_COMPAT)
860 				rtm->rtm_flags = RTF_GATEWAY |
861 					(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
862 			else
863 				rtm->rtm_flags = rt->rt_flags;
864 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
865 			rtm->rtm_addrs = info.rti_addrs;
866 			break;
867 
868 		case RTM_CHANGE:
869 			/*
870 			 * New gateway could require new ifaddr, ifp;
871 			 * flags may also be different; ifp may be specified
872 			 * by ll sockaddr when protocol address is ambiguous
873 			 */
874 			if (((rt->rt_flags & RTF_GATEWAY) &&
875 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
876 			    info.rti_info[RTAX_IFP] != NULL ||
877 			    (info.rti_info[RTAX_IFA] != NULL &&
878 			     !sa_equal(info.rti_info[RTAX_IFA],
879 				       rt->rt_ifa->ifa_addr))) {
880 				RT_UNLOCK(rt);
881 				RADIX_NODE_HEAD_LOCK(rnh);
882 				error = rt_getifa_fib(&info, rt->rt_fibnum);
883 				/*
884 				 * XXXRW: Really we should release this
885 				 * reference later, but this maintains
886 				 * historical behavior.
887 				 */
888 				if (info.rti_ifa != NULL)
889 					ifa_free(info.rti_ifa);
890 				RADIX_NODE_HEAD_UNLOCK(rnh);
891 				if (error != 0)
892 					senderr(error);
893 				RT_LOCK(rt);
894 			}
895 			if (info.rti_ifa != NULL &&
896 			    info.rti_ifa != rt->rt_ifa &&
897 			    rt->rt_ifa != NULL &&
898 			    rt->rt_ifa->ifa_rtrequest != NULL) {
899 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
900 				    &info);
901 				ifa_free(rt->rt_ifa);
902 			}
903 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
904 				RT_UNLOCK(rt);
905 				RADIX_NODE_HEAD_LOCK(rnh);
906 				RT_LOCK(rt);
907 
908 				error = rt_setgate(rt, rt_key(rt),
909 				    info.rti_info[RTAX_GATEWAY]);
910 				RADIX_NODE_HEAD_UNLOCK(rnh);
911 				if (error != 0) {
912 					RT_UNLOCK(rt);
913 					senderr(error);
914 				}
915 				rt->rt_flags &= ~RTF_GATEWAY;
916 				rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
917 			}
918 			if (info.rti_ifa != NULL &&
919 			    info.rti_ifa != rt->rt_ifa) {
920 				ifa_ref(info.rti_ifa);
921 				rt->rt_ifa = info.rti_ifa;
922 				rt->rt_ifp = info.rti_ifp;
923 			}
924 			/* Allow some flags to be toggled on change. */
925 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
926 				    (rtm->rtm_flags & RTF_FMASK);
927 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
928 					&rt->rt_rmx);
929 			rtm->rtm_index = rt->rt_ifp->if_index;
930 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
931 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
932 			/* FALLTHROUGH */
933 		case RTM_LOCK:
934 			/* We don't support locks anymore */
935 			break;
936 		}
937 		RT_UNLOCK(rt);
938 		break;
939 
940 	default:
941 		senderr(EOPNOTSUPP);
942 	}
943 
944 flush:
945 	if (rtm) {
946 		if (error)
947 			rtm->rtm_errno = error;
948 		else
949 			rtm->rtm_flags |= RTF_DONE;
950 	}
951 	if (rt)		/* XXX can this be true? */
952 		RTFREE(rt);
953     {
954 	struct rawcb *rp = NULL;
955 	/*
956 	 * Check to see if we don't want our own messages.
957 	 */
958 	if ((so->so_options & SO_USELOOPBACK) == 0) {
959 		if (route_cb.any_count <= 1) {
960 			if (rtm)
961 				Free(rtm);
962 			m_freem(m);
963 			return (error);
964 		}
965 		/* There is another listener, so construct message */
966 		rp = sotorawcb(so);
967 	}
968 	if (rtm) {
969 #ifdef INET6
970 		if (rti_need_deembed) {
971 			/* sin6_scope_id is recovered before sending rtm. */
972 			sin6 = (struct sockaddr_in6 *)&ss;
973 			for (i = 0; i < RTAX_MAX; i++) {
974 				if (info.rti_info[i] == NULL)
975 					continue;
976 				if (info.rti_info[i]->sa_family != AF_INET6)
977 					continue;
978 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
979 				if (sa6_recoverscope(sin6) == 0)
980 					bcopy(sin6, info.rti_info[i],
981 						    sizeof(*sin6));
982 			}
983 		}
984 #endif
985 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
986 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
987 			m_freem(m);
988 			m = NULL;
989 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
990 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
991 	}
992 	if (m) {
993 		M_SETFIB(m, so->so_fibnum);
994 		m->m_flags |= RTS_FILTER_FIB;
995 		if (rp) {
996 			/*
997 			 * XXX insure we don't get a copy by
998 			 * invalidating our protocol
999 			 */
1000 			unsigned short family = rp->rcb_proto.sp_family;
1001 			rp->rcb_proto.sp_family = 0;
1002 			rt_dispatch(m, saf);
1003 			rp->rcb_proto.sp_family = family;
1004 		} else
1005 			rt_dispatch(m, saf);
1006 	}
1007 	/* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
1008 	if (rtm)
1009 		Free(rtm);
1010     }
1011 	return (error);
1012 #undef	sa_equal
1013 }
1014 
1015 static void
1016 rt_setmetrics(u_long which, const struct rt_metrics *in,
1017 	struct rt_metrics_lite *out)
1018 {
1019 #define metric(f, e) if (which & (f)) out->e = in->e;
1020 	/*
1021 	 * Only these are stored in the routing entry since introduction
1022 	 * of tcp hostcache. The rest is ignored.
1023 	 */
1024 	metric(RTV_MTU, rmx_mtu);
1025 	metric(RTV_WEIGHT, rmx_weight);
1026 	/* Userland -> kernel timebase conversion. */
1027 	if (which & RTV_EXPIRE)
1028 		out->rmx_expire = in->rmx_expire ?
1029 		    in->rmx_expire - time_second + time_uptime : 0;
1030 #undef metric
1031 }
1032 
1033 static void
1034 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
1035 {
1036 #define metric(e) out->e = in->e;
1037 	bzero(out, sizeof(*out));
1038 	metric(rmx_mtu);
1039 	metric(rmx_weight);
1040 	/* Kernel -> userland timebase conversion. */
1041 	out->rmx_expire = in->rmx_expire ?
1042 	    in->rmx_expire - time_uptime + time_second : 0;
1043 #undef metric
1044 }
1045 
1046 /*
1047  * Extract the addresses of the passed sockaddrs.
1048  * Do a little sanity checking so as to avoid bad memory references.
1049  * This data is derived straight from userland.
1050  */
1051 static int
1052 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1053 {
1054 	struct sockaddr *sa;
1055 	int i;
1056 
1057 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1058 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1059 			continue;
1060 		sa = (struct sockaddr *)cp;
1061 		/*
1062 		 * It won't fit.
1063 		 */
1064 		if (cp + sa->sa_len > cplim)
1065 			return (EINVAL);
1066 		/*
1067 		 * there are no more.. quit now
1068 		 * If there are more bits, they are in error.
1069 		 * I've seen this. route(1) can evidently generate these.
1070 		 * This causes kernel to core dump.
1071 		 * for compatibility, If we see this, point to a safe address.
1072 		 */
1073 		if (sa->sa_len == 0) {
1074 			rtinfo->rti_info[i] = &sa_zero;
1075 			return (0); /* should be EINVAL but for compat */
1076 		}
1077 		/* accept it */
1078 #ifdef INET6
1079 		if (sa->sa_family == AF_INET6)
1080 			sa6_embedscope((struct sockaddr_in6 *)sa,
1081 			    V_ip6_use_defzone);
1082 #endif
1083 		rtinfo->rti_info[i] = sa;
1084 		cp += SA_SIZE(sa);
1085 	}
1086 	return (0);
1087 }
1088 
1089 /*
1090  * Used by the routing socket.
1091  */
1092 static struct mbuf *
1093 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1094 {
1095 	struct rt_msghdr *rtm;
1096 	struct mbuf *m;
1097 	int i;
1098 	struct sockaddr *sa;
1099 #ifdef INET6
1100 	struct sockaddr_storage ss;
1101 	struct sockaddr_in6 *sin6;
1102 #endif
1103 	int len, dlen;
1104 
1105 	switch (type) {
1106 
1107 	case RTM_DELADDR:
1108 	case RTM_NEWADDR:
1109 		len = sizeof(struct ifa_msghdr);
1110 		break;
1111 
1112 	case RTM_DELMADDR:
1113 	case RTM_NEWMADDR:
1114 		len = sizeof(struct ifma_msghdr);
1115 		break;
1116 
1117 	case RTM_IFINFO:
1118 		len = sizeof(struct if_msghdr);
1119 		break;
1120 
1121 	case RTM_IFANNOUNCE:
1122 	case RTM_IEEE80211:
1123 		len = sizeof(struct if_announcemsghdr);
1124 		break;
1125 
1126 	default:
1127 		len = sizeof(struct rt_msghdr);
1128 	}
1129 
1130 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1131 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1132 	if (len > MHLEN)
1133 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1134 	else
1135 		m = m_gethdr(M_NOWAIT, MT_DATA);
1136 	if (m == NULL)
1137 		return (m);
1138 
1139 	m->m_pkthdr.len = m->m_len = len;
1140 	rtm = mtod(m, struct rt_msghdr *);
1141 	bzero((caddr_t)rtm, len);
1142 	for (i = 0; i < RTAX_MAX; i++) {
1143 		if ((sa = rtinfo->rti_info[i]) == NULL)
1144 			continue;
1145 		rtinfo->rti_addrs |= (1 << i);
1146 		dlen = SA_SIZE(sa);
1147 #ifdef INET6
1148 		if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1149 			sin6 = (struct sockaddr_in6 *)&ss;
1150 			bcopy(sa, sin6, sizeof(*sin6));
1151 			if (sa6_recoverscope(sin6) == 0)
1152 				sa = (struct sockaddr *)sin6;
1153 		}
1154 #endif
1155 		m_copyback(m, len, dlen, (caddr_t)sa);
1156 		len += dlen;
1157 	}
1158 	if (m->m_pkthdr.len != len) {
1159 		m_freem(m);
1160 		return (NULL);
1161 	}
1162 	rtm->rtm_msglen = len;
1163 	rtm->rtm_version = RTM_VERSION;
1164 	rtm->rtm_type = type;
1165 	return (m);
1166 }
1167 
1168 /*
1169  * Used by the sysctl code and routing socket.
1170  */
1171 static int
1172 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1173 {
1174 	int i;
1175 	int len, dlen, second_time = 0;
1176 	caddr_t cp0;
1177 #ifdef INET6
1178 	struct sockaddr_storage ss;
1179 	struct sockaddr_in6 *sin6;
1180 #endif
1181 
1182 	rtinfo->rti_addrs = 0;
1183 again:
1184 	switch (type) {
1185 
1186 	case RTM_DELADDR:
1187 	case RTM_NEWADDR:
1188 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1189 #ifdef COMPAT_FREEBSD32
1190 			if (w->w_req->flags & SCTL_MASK32)
1191 				len = sizeof(struct ifa_msghdrl32);
1192 			else
1193 #endif
1194 				len = sizeof(struct ifa_msghdrl);
1195 		} else
1196 			len = sizeof(struct ifa_msghdr);
1197 		break;
1198 
1199 	case RTM_IFINFO:
1200 #ifdef COMPAT_FREEBSD32
1201 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1202 			if (w->w_op == NET_RT_IFLISTL)
1203 				len = sizeof(struct if_msghdrl32);
1204 			else
1205 				len = sizeof(struct if_msghdr32);
1206 			break;
1207 		}
1208 #endif
1209 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1210 			len = sizeof(struct if_msghdrl);
1211 		else
1212 			len = sizeof(struct if_msghdr);
1213 		break;
1214 
1215 	case RTM_NEWMADDR:
1216 		len = sizeof(struct ifma_msghdr);
1217 		break;
1218 
1219 	default:
1220 		len = sizeof(struct rt_msghdr);
1221 	}
1222 	cp0 = cp;
1223 	if (cp0)
1224 		cp += len;
1225 	for (i = 0; i < RTAX_MAX; i++) {
1226 		struct sockaddr *sa;
1227 
1228 		if ((sa = rtinfo->rti_info[i]) == NULL)
1229 			continue;
1230 		rtinfo->rti_addrs |= (1 << i);
1231 		dlen = SA_SIZE(sa);
1232 		if (cp) {
1233 #ifdef INET6
1234 			if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1235 				sin6 = (struct sockaddr_in6 *)&ss;
1236 				bcopy(sa, sin6, sizeof(*sin6));
1237 				if (sa6_recoverscope(sin6) == 0)
1238 					sa = (struct sockaddr *)sin6;
1239 			}
1240 #endif
1241 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1242 			cp += dlen;
1243 		}
1244 		len += dlen;
1245 	}
1246 	len = ALIGN(len);
1247 	if (cp == NULL && w != NULL && !second_time) {
1248 		struct walkarg *rw = w;
1249 
1250 		if (rw->w_req) {
1251 			if (rw->w_tmemsize < len) {
1252 				if (rw->w_tmem)
1253 					free(rw->w_tmem, M_RTABLE);
1254 				rw->w_tmem = (caddr_t)
1255 					malloc(len, M_RTABLE, M_NOWAIT);
1256 				if (rw->w_tmem)
1257 					rw->w_tmemsize = len;
1258 			}
1259 			if (rw->w_tmem) {
1260 				cp = rw->w_tmem;
1261 				second_time = 1;
1262 				goto again;
1263 			}
1264 		}
1265 	}
1266 	if (cp) {
1267 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1268 
1269 		rtm->rtm_version = RTM_VERSION;
1270 		rtm->rtm_type = type;
1271 		rtm->rtm_msglen = len;
1272 	}
1273 	return (len);
1274 }
1275 
1276 /*
1277  * This routine is called to generate a message from the routing
1278  * socket indicating that a redirect has occured, a routing lookup
1279  * has failed, or that a protocol has detected timeouts to a particular
1280  * destination.
1281  */
1282 void
1283 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1284     int fibnum)
1285 {
1286 	struct rt_msghdr *rtm;
1287 	struct mbuf *m;
1288 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1289 
1290 	if (route_cb.any_count == 0)
1291 		return;
1292 	m = rt_msg1(type, rtinfo);
1293 	if (m == NULL)
1294 		return;
1295 
1296 	if (fibnum != RTS_ALLFIBS) {
1297 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1298 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1299 		M_SETFIB(m, fibnum);
1300 		m->m_flags |= RTS_FILTER_FIB;
1301 	}
1302 
1303 	rtm = mtod(m, struct rt_msghdr *);
1304 	rtm->rtm_flags = RTF_DONE | flags;
1305 	rtm->rtm_errno = error;
1306 	rtm->rtm_addrs = rtinfo->rti_addrs;
1307 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1308 }
1309 
1310 void
1311 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1312 {
1313 
1314 	rt_missmsg_fib(type, rtinfo, flags, error, RTS_ALLFIBS);
1315 }
1316 
1317 /*
1318  * This routine is called to generate a message from the routing
1319  * socket indicating that the status of a network interface has changed.
1320  */
1321 void
1322 rt_ifmsg(struct ifnet *ifp)
1323 {
1324 	struct if_msghdr *ifm;
1325 	struct mbuf *m;
1326 	struct rt_addrinfo info;
1327 
1328 	if (route_cb.any_count == 0)
1329 		return;
1330 	bzero((caddr_t)&info, sizeof(info));
1331 	m = rt_msg1(RTM_IFINFO, &info);
1332 	if (m == NULL)
1333 		return;
1334 	ifm = mtod(m, struct if_msghdr *);
1335 	ifm->ifm_index = ifp->if_index;
1336 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1337 	ifm->ifm_data = ifp->if_data;
1338 	ifm->ifm_addrs = 0;
1339 	rt_dispatch(m, AF_UNSPEC);
1340 }
1341 
1342 /*
1343  * This is called to generate messages from the routing socket
1344  * indicating a network interface has had addresses associated with it.
1345  * if we ever reverse the logic and replace messages TO the routing
1346  * socket indicate a request to configure interfaces, then it will
1347  * be unnecessary as the routing socket will automatically generate
1348  * copies of it.
1349  */
1350 void
1351 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
1352     int fibnum)
1353 {
1354 	struct rt_addrinfo info;
1355 	struct sockaddr *sa = NULL;
1356 	int pass;
1357 	struct mbuf *m = NULL;
1358 	struct ifnet *ifp = ifa->ifa_ifp;
1359 
1360 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1361 		("unexpected cmd %u", cmd));
1362 #if defined(INET) || defined(INET6)
1363 #ifdef SCTP
1364 	/*
1365 	 * notify the SCTP stack
1366 	 * this will only get called when an address is added/deleted
1367 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1368 	 */
1369 	sctp_addr_change(ifa, cmd);
1370 #endif /* SCTP */
1371 #endif
1372 	if (route_cb.any_count == 0)
1373 		return;
1374 	for (pass = 1; pass < 3; pass++) {
1375 		bzero((caddr_t)&info, sizeof(info));
1376 		if ((cmd == RTM_ADD && pass == 1) ||
1377 		    (cmd == RTM_DELETE && pass == 2)) {
1378 			struct ifa_msghdr *ifam;
1379 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1380 
1381 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1382 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1383 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1384 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1385 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1386 				continue;
1387 			ifam = mtod(m, struct ifa_msghdr *);
1388 			ifam->ifam_index = ifp->if_index;
1389 			ifam->ifam_metric = ifa->ifa_metric;
1390 			ifam->ifam_flags = ifa->ifa_flags;
1391 			ifam->ifam_addrs = info.rti_addrs;
1392 		}
1393 		if ((cmd == RTM_ADD && pass == 2) ||
1394 		    (cmd == RTM_DELETE && pass == 1)) {
1395 			struct rt_msghdr *rtm;
1396 
1397 			if (rt == NULL)
1398 				continue;
1399 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1400 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1401 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1402 			if ((m = rt_msg1(cmd, &info)) == NULL)
1403 				continue;
1404 			rtm = mtod(m, struct rt_msghdr *);
1405 			rtm->rtm_index = ifp->if_index;
1406 			rtm->rtm_flags |= rt->rt_flags;
1407 			rtm->rtm_errno = error;
1408 			rtm->rtm_addrs = info.rti_addrs;
1409 		}
1410 		if (fibnum != RTS_ALLFIBS) {
1411 			KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: "
1412 			    "fibnum out of range 0 <= %d < %d", __func__,
1413 			     fibnum, rt_numfibs));
1414 			M_SETFIB(m, fibnum);
1415 			m->m_flags |= RTS_FILTER_FIB;
1416 		}
1417 		rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1418 	}
1419 }
1420 
1421 void
1422 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1423 {
1424 
1425 	rt_newaddrmsg_fib(cmd, ifa, error, rt, RTS_ALLFIBS);
1426 }
1427 
1428 /*
1429  * This is the analogue to the rt_newaddrmsg which performs the same
1430  * function but for multicast group memberhips.  This is easier since
1431  * there is no route state to worry about.
1432  */
1433 void
1434 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1435 {
1436 	struct rt_addrinfo info;
1437 	struct mbuf *m = NULL;
1438 	struct ifnet *ifp = ifma->ifma_ifp;
1439 	struct ifma_msghdr *ifmam;
1440 
1441 	if (route_cb.any_count == 0)
1442 		return;
1443 
1444 	bzero((caddr_t)&info, sizeof(info));
1445 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1446 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1447 	/*
1448 	 * If a link-layer address is present, present it as a ``gateway''
1449 	 * (similarly to how ARP entries, e.g., are presented).
1450 	 */
1451 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1452 	m = rt_msg1(cmd, &info);
1453 	if (m == NULL)
1454 		return;
1455 	ifmam = mtod(m, struct ifma_msghdr *);
1456 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1457 	    __func__));
1458 	ifmam->ifmam_index = ifp->if_index;
1459 	ifmam->ifmam_addrs = info.rti_addrs;
1460 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1461 }
1462 
1463 static struct mbuf *
1464 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1465 	struct rt_addrinfo *info)
1466 {
1467 	struct if_announcemsghdr *ifan;
1468 	struct mbuf *m;
1469 
1470 	if (route_cb.any_count == 0)
1471 		return NULL;
1472 	bzero((caddr_t)info, sizeof(*info));
1473 	m = rt_msg1(type, info);
1474 	if (m != NULL) {
1475 		ifan = mtod(m, struct if_announcemsghdr *);
1476 		ifan->ifan_index = ifp->if_index;
1477 		strlcpy(ifan->ifan_name, ifp->if_xname,
1478 			sizeof(ifan->ifan_name));
1479 		ifan->ifan_what = what;
1480 	}
1481 	return m;
1482 }
1483 
1484 /*
1485  * This is called to generate routing socket messages indicating
1486  * IEEE80211 wireless events.
1487  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1488  */
1489 void
1490 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1491 {
1492 	struct mbuf *m;
1493 	struct rt_addrinfo info;
1494 
1495 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1496 	if (m != NULL) {
1497 		/*
1498 		 * Append the ieee80211 data.  Try to stick it in the
1499 		 * mbuf containing the ifannounce msg; otherwise allocate
1500 		 * a new mbuf and append.
1501 		 *
1502 		 * NB: we assume m is a single mbuf.
1503 		 */
1504 		if (data_len > M_TRAILINGSPACE(m)) {
1505 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1506 			if (n == NULL) {
1507 				m_freem(m);
1508 				return;
1509 			}
1510 			bcopy(data, mtod(n, void *), data_len);
1511 			n->m_len = data_len;
1512 			m->m_next = n;
1513 		} else if (data_len > 0) {
1514 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1515 			m->m_len += data_len;
1516 		}
1517 		if (m->m_flags & M_PKTHDR)
1518 			m->m_pkthdr.len += data_len;
1519 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1520 		rt_dispatch(m, AF_UNSPEC);
1521 	}
1522 }
1523 
1524 /*
1525  * This is called to generate routing socket messages indicating
1526  * network interface arrival and departure.
1527  */
1528 void
1529 rt_ifannouncemsg(struct ifnet *ifp, int what)
1530 {
1531 	struct mbuf *m;
1532 	struct rt_addrinfo info;
1533 
1534 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1535 	if (m != NULL)
1536 		rt_dispatch(m, AF_UNSPEC);
1537 }
1538 
1539 static void
1540 rt_dispatch(struct mbuf *m, sa_family_t saf)
1541 {
1542 	struct m_tag *tag;
1543 
1544 	/*
1545 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1546 	 * use when injecting the mbuf into the routing socket buffer from
1547 	 * the netisr.
1548 	 */
1549 	if (saf != AF_UNSPEC) {
1550 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1551 		    M_NOWAIT);
1552 		if (tag == NULL) {
1553 			m_freem(m);
1554 			return;
1555 		}
1556 		*(unsigned short *)(tag + 1) = saf;
1557 		m_tag_prepend(m, tag);
1558 	}
1559 #ifdef VIMAGE
1560 	if (V_loif)
1561 		m->m_pkthdr.rcvif = V_loif;
1562 	else {
1563 		m_freem(m);
1564 		return;
1565 	}
1566 #endif
1567 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1568 }
1569 
1570 /*
1571  * This is used in dumping the kernel table via sysctl().
1572  */
1573 static int
1574 sysctl_dumpentry(struct radix_node *rn, void *vw)
1575 {
1576 	struct walkarg *w = vw;
1577 	struct rtentry *rt = (struct rtentry *)rn;
1578 	int error = 0, size;
1579 	struct rt_addrinfo info;
1580 
1581 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1582 		return 0;
1583 	if ((rt->rt_flags & RTF_HOST) == 0
1584 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1585 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1586 		return (0);
1587 	bzero((caddr_t)&info, sizeof(info));
1588 	info.rti_info[RTAX_DST] = rt_key(rt);
1589 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1590 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1591 	info.rti_info[RTAX_GENMASK] = 0;
1592 	if (rt->rt_ifp) {
1593 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1594 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1595 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1596 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1597 	}
1598 	size = rt_msg2(RTM_GET, &info, NULL, w);
1599 	if (w->w_req && w->w_tmem) {
1600 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1601 
1602 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1603 			rtm->rtm_flags = RTF_GATEWAY |
1604 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1605 		else
1606 			rtm->rtm_flags = rt->rt_flags;
1607 		/*
1608 		 * let's be honest about this being a retarded hack
1609 		 */
1610 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1611 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1612 		rtm->rtm_index = rt->rt_ifp->if_index;
1613 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1614 		rtm->rtm_addrs = info.rti_addrs;
1615 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1616 		return (error);
1617 	}
1618 	return (error);
1619 }
1620 
1621 #ifdef COMPAT_FREEBSD32
1622 static void
1623 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1624 {
1625 
1626 	bzero(dst, sizeof(*dst));
1627 	CP(*src, *dst, ifi_type);
1628 	CP(*src, *dst, ifi_physical);
1629 	CP(*src, *dst, ifi_addrlen);
1630 	CP(*src, *dst, ifi_hdrlen);
1631 	CP(*src, *dst, ifi_link_state);
1632 	CP(*src, *dst, ifi_vhid);
1633 	CP(*src, *dst, ifi_baudrate_pf);
1634 	dst->ifi_datalen = sizeof(struct if_data32);
1635 	CP(*src, *dst, ifi_mtu);
1636 	CP(*src, *dst, ifi_metric);
1637 	CP(*src, *dst, ifi_baudrate);
1638 	CP(*src, *dst, ifi_ipackets);
1639 	CP(*src, *dst, ifi_ierrors);
1640 	CP(*src, *dst, ifi_opackets);
1641 	CP(*src, *dst, ifi_oerrors);
1642 	CP(*src, *dst, ifi_collisions);
1643 	CP(*src, *dst, ifi_ibytes);
1644 	CP(*src, *dst, ifi_obytes);
1645 	CP(*src, *dst, ifi_imcasts);
1646 	CP(*src, *dst, ifi_omcasts);
1647 	CP(*src, *dst, ifi_iqdrops);
1648 	CP(*src, *dst, ifi_noproto);
1649 	CP(*src, *dst, ifi_hwassist);
1650 	CP(*src, *dst, ifi_epoch);
1651 	TV_CP(*src, *dst, ifi_lastchange);
1652 }
1653 #endif
1654 
1655 static int
1656 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info,
1657     struct walkarg *w, int len)
1658 {
1659 	struct if_msghdrl *ifm;
1660 
1661 #ifdef COMPAT_FREEBSD32
1662 	if (w->w_req->flags & SCTL_MASK32) {
1663 		struct if_msghdrl32 *ifm32;
1664 
1665 		ifm32 = (struct if_msghdrl32 *)w->w_tmem;
1666 		ifm32->ifm_addrs = info->rti_addrs;
1667 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1668 		ifm32->ifm_index = ifp->if_index;
1669 		ifm32->_ifm_spare1 = 0;
1670 		ifm32->ifm_len = sizeof(*ifm32);
1671 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1672 
1673 		copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1674 		/* Fixup if_data carp(4) vhid. */
1675 		if (carp_get_vhid_p != NULL)
1676 			ifm32->ifm_data.ifi_vhid =
1677 			    (*carp_get_vhid_p)(ifp->if_addr);
1678 
1679 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1680 	}
1681 #endif
1682 	ifm = (struct if_msghdrl *)w->w_tmem;
1683 	ifm->ifm_addrs = info->rti_addrs;
1684 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1685 	ifm->ifm_index = ifp->if_index;
1686 	ifm->_ifm_spare1 = 0;
1687 	ifm->ifm_len = sizeof(*ifm);
1688 	ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1689 
1690 	ifm->ifm_data = ifp->if_data;
1691 	/* Fixup if_data carp(4) vhid. */
1692 	if (carp_get_vhid_p != NULL)
1693 		ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1694 
1695 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1696 }
1697 
1698 static int
1699 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info,
1700     struct walkarg *w, int len)
1701 {
1702 	struct if_msghdr *ifm;
1703 
1704 #ifdef COMPAT_FREEBSD32
1705 	if (w->w_req->flags & SCTL_MASK32) {
1706 		struct if_msghdr32 *ifm32;
1707 
1708 		ifm32 = (struct if_msghdr32 *)w->w_tmem;
1709 		ifm32->ifm_addrs = info->rti_addrs;
1710 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1711 		ifm32->ifm_index = ifp->if_index;
1712 
1713 		copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1714 		/* Fixup if_data carp(4) vhid. */
1715 		if (carp_get_vhid_p != NULL)
1716 			ifm32->ifm_data.ifi_vhid =
1717 			    (*carp_get_vhid_p)(ifp->if_addr);
1718 
1719 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1720 	}
1721 #endif
1722 	ifm = (struct if_msghdr *)w->w_tmem;
1723 	ifm->ifm_addrs = info->rti_addrs;
1724 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1725 	ifm->ifm_index = ifp->if_index;
1726 
1727 	ifm->ifm_data = ifp->if_data;
1728 	/* Fixup if_data carp(4) vhid. */
1729 	if (carp_get_vhid_p != NULL)
1730 		ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1731 
1732 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1733 }
1734 
1735 static int
1736 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1737     struct walkarg *w, int len)
1738 {
1739 	struct ifa_msghdrl *ifam;
1740 
1741 #ifdef COMPAT_FREEBSD32
1742 	if (w->w_req->flags & SCTL_MASK32) {
1743 		struct ifa_msghdrl32 *ifam32;
1744 
1745 		ifam32 = (struct ifa_msghdrl32 *)w->w_tmem;
1746 		ifam32->ifam_addrs = info->rti_addrs;
1747 		ifam32->ifam_flags = ifa->ifa_flags;
1748 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1749 		ifam32->_ifam_spare1 = 0;
1750 		ifam32->ifam_len = sizeof(*ifam32);
1751 		ifam32->ifam_data_off =
1752 		    offsetof(struct ifa_msghdrl32, ifam_data);
1753 		ifam32->ifam_metric = ifa->ifa_metric;
1754 
1755 		bzero(&ifam32->ifam_data, sizeof(ifam32->ifam_data));
1756 		ifam32->ifam_data.ifi_datalen = sizeof(struct if_data32);
1757 		ifam32->ifam_data.ifi_ipackets =
1758 		    counter_u64_fetch(ifa->ifa_ipackets);
1759 		ifam32->ifam_data.ifi_opackets =
1760 		    counter_u64_fetch(ifa->ifa_opackets);
1761 		ifam32->ifam_data.ifi_ibytes =
1762 		    counter_u64_fetch(ifa->ifa_ibytes);
1763 		ifam32->ifam_data.ifi_obytes =
1764 		    counter_u64_fetch(ifa->ifa_obytes);
1765 
1766 		/* Fixup if_data carp(4) vhid. */
1767 		if (carp_get_vhid_p != NULL)
1768 			ifam32->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1769 
1770 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifam32, len));
1771 	}
1772 #endif
1773 
1774 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1775 	ifam->ifam_addrs = info->rti_addrs;
1776 	ifam->ifam_flags = ifa->ifa_flags;
1777 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1778 	ifam->_ifam_spare1 = 0;
1779 	ifam->ifam_len = sizeof(*ifam);
1780 	ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1781 	ifam->ifam_metric = ifa->ifa_metric;
1782 
1783 	bzero(&ifam->ifam_data, sizeof(ifam->ifam_data));
1784 	ifam->ifam_data.ifi_datalen = sizeof(struct if_data);
1785 	ifam->ifam_data.ifi_ipackets = counter_u64_fetch(ifa->ifa_ipackets);
1786 	ifam->ifam_data.ifi_opackets = counter_u64_fetch(ifa->ifa_opackets);
1787 	ifam->ifam_data.ifi_ibytes = counter_u64_fetch(ifa->ifa_ibytes);
1788 	ifam->ifam_data.ifi_obytes = counter_u64_fetch(ifa->ifa_obytes);
1789 
1790 	/* Fixup if_data carp(4) vhid. */
1791 	if (carp_get_vhid_p != NULL)
1792 		ifam->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1793 
1794 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1795 }
1796 
1797 static int
1798 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1799     struct walkarg *w, int len)
1800 {
1801 	struct ifa_msghdr *ifam;
1802 
1803 	ifam = (struct ifa_msghdr *)w->w_tmem;
1804 	ifam->ifam_addrs = info->rti_addrs;
1805 	ifam->ifam_flags = ifa->ifa_flags;
1806 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1807 	ifam->ifam_metric = ifa->ifa_metric;
1808 
1809 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1810 }
1811 
1812 static int
1813 sysctl_iflist(int af, struct walkarg *w)
1814 {
1815 	struct ifnet *ifp;
1816 	struct ifaddr *ifa;
1817 	struct rt_addrinfo info;
1818 	int len, error = 0;
1819 
1820 	bzero((caddr_t)&info, sizeof(info));
1821 	IFNET_RLOCK_NOSLEEP();
1822 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1823 		if (w->w_arg && w->w_arg != ifp->if_index)
1824 			continue;
1825 		IF_ADDR_RLOCK(ifp);
1826 		ifa = ifp->if_addr;
1827 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1828 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1829 		info.rti_info[RTAX_IFP] = NULL;
1830 		if (w->w_req && w->w_tmem) {
1831 			if (w->w_op == NET_RT_IFLISTL)
1832 				error = sysctl_iflist_ifml(ifp, &info, w, len);
1833 			else
1834 				error = sysctl_iflist_ifm(ifp, &info, w, len);
1835 			if (error)
1836 				goto done;
1837 		}
1838 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1839 			if (af && af != ifa->ifa_addr->sa_family)
1840 				continue;
1841 			if (prison_if(w->w_req->td->td_ucred,
1842 			    ifa->ifa_addr) != 0)
1843 				continue;
1844 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1845 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1846 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1847 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1848 			if (w->w_req && w->w_tmem) {
1849 				if (w->w_op == NET_RT_IFLISTL)
1850 					error = sysctl_iflist_ifaml(ifa, &info,
1851 					    w, len);
1852 				else
1853 					error = sysctl_iflist_ifam(ifa, &info,
1854 					    w, len);
1855 				if (error)
1856 					goto done;
1857 			}
1858 		}
1859 		IF_ADDR_RUNLOCK(ifp);
1860 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1861 			info.rti_info[RTAX_BRD] = NULL;
1862 	}
1863 done:
1864 	if (ifp != NULL)
1865 		IF_ADDR_RUNLOCK(ifp);
1866 	IFNET_RUNLOCK_NOSLEEP();
1867 	return (error);
1868 }
1869 
1870 static int
1871 sysctl_ifmalist(int af, struct walkarg *w)
1872 {
1873 	struct ifnet *ifp;
1874 	struct ifmultiaddr *ifma;
1875 	struct	rt_addrinfo info;
1876 	int	len, error = 0;
1877 	struct ifaddr *ifa;
1878 
1879 	bzero((caddr_t)&info, sizeof(info));
1880 	IFNET_RLOCK_NOSLEEP();
1881 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1882 		if (w->w_arg && w->w_arg != ifp->if_index)
1883 			continue;
1884 		ifa = ifp->if_addr;
1885 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1886 		IF_ADDR_RLOCK(ifp);
1887 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1888 			if (af && af != ifma->ifma_addr->sa_family)
1889 				continue;
1890 			if (prison_if(w->w_req->td->td_ucred,
1891 			    ifma->ifma_addr) != 0)
1892 				continue;
1893 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1894 			info.rti_info[RTAX_GATEWAY] =
1895 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1896 			    ifma->ifma_lladdr : NULL;
1897 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1898 			if (w->w_req && w->w_tmem) {
1899 				struct ifma_msghdr *ifmam;
1900 
1901 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1902 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1903 				ifmam->ifmam_flags = 0;
1904 				ifmam->ifmam_addrs = info.rti_addrs;
1905 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1906 				if (error) {
1907 					IF_ADDR_RUNLOCK(ifp);
1908 					goto done;
1909 				}
1910 			}
1911 		}
1912 		IF_ADDR_RUNLOCK(ifp);
1913 	}
1914 done:
1915 	IFNET_RUNLOCK_NOSLEEP();
1916 	return (error);
1917 }
1918 
1919 static int
1920 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1921 {
1922 	int	*name = (int *)arg1;
1923 	u_int	namelen = arg2;
1924 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1925 	int	i, lim, error = EINVAL;
1926 	int	fib = 0;
1927 	u_char	af;
1928 	struct	walkarg w;
1929 
1930 	name ++;
1931 	namelen--;
1932 	if (req->newptr)
1933 		return (EPERM);
1934 	if (name[1] == NET_RT_DUMP) {
1935 		if (namelen == 3)
1936 			fib = req->td->td_proc->p_fibnum;
1937 		else if (namelen == 4)
1938 			fib = (name[3] == -1) ?
1939 			    req->td->td_proc->p_fibnum : name[3];
1940 		else
1941 			return ((namelen < 3) ? EISDIR : ENOTDIR);
1942 		if (fib < 0 || fib >= rt_numfibs)
1943 			return (EINVAL);
1944 	} else if (namelen != 3)
1945 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1946 	af = name[0];
1947 	if (af > AF_MAX)
1948 		return (EINVAL);
1949 	bzero(&w, sizeof(w));
1950 	w.w_op = name[1];
1951 	w.w_arg = name[2];
1952 	w.w_req = req;
1953 
1954 	error = sysctl_wire_old_buffer(req, 0);
1955 	if (error)
1956 		return (error);
1957 	switch (w.w_op) {
1958 
1959 	case NET_RT_DUMP:
1960 	case NET_RT_FLAGS:
1961 		if (af == 0) {			/* dump all tables */
1962 			i = 1;
1963 			lim = AF_MAX;
1964 		} else				/* dump only one table */
1965 			i = lim = af;
1966 
1967 		/*
1968 		 * take care of llinfo entries, the caller must
1969 		 * specify an AF
1970 		 */
1971 		if (w.w_op == NET_RT_FLAGS &&
1972 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1973 			if (af != 0)
1974 				error = lltable_sysctl_dumparp(af, w.w_req);
1975 			else
1976 				error = EINVAL;
1977 			break;
1978 		}
1979 		/*
1980 		 * take care of routing entries
1981 		 */
1982 		for (error = 0; error == 0 && i <= lim; i++) {
1983 			rnh = rt_tables_get_rnh(fib, i);
1984 			if (rnh != NULL) {
1985 				RADIX_NODE_HEAD_RLOCK(rnh);
1986 			    	error = rnh->rnh_walktree(rnh,
1987 				    sysctl_dumpentry, &w);
1988 				RADIX_NODE_HEAD_RUNLOCK(rnh);
1989 			} else if (af != 0)
1990 				error = EAFNOSUPPORT;
1991 		}
1992 		break;
1993 
1994 	case NET_RT_IFLIST:
1995 	case NET_RT_IFLISTL:
1996 		error = sysctl_iflist(af, &w);
1997 		break;
1998 
1999 	case NET_RT_IFMALIST:
2000 		error = sysctl_ifmalist(af, &w);
2001 		break;
2002 	}
2003 	if (w.w_tmem)
2004 		free(w.w_tmem, M_RTABLE);
2005 	return (error);
2006 }
2007 
2008 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
2009 
2010 /*
2011  * Definitions of protocols supported in the ROUTE domain.
2012  */
2013 
2014 static struct domain routedomain;		/* or at least forward */
2015 
2016 static struct protosw routesw[] = {
2017 {
2018 	.pr_type =		SOCK_RAW,
2019 	.pr_domain =		&routedomain,
2020 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2021 	.pr_output =		route_output,
2022 	.pr_ctlinput =		raw_ctlinput,
2023 	.pr_init =		raw_init,
2024 	.pr_usrreqs =		&route_usrreqs
2025 }
2026 };
2027 
2028 static struct domain routedomain = {
2029 	.dom_family =		PF_ROUTE,
2030 	.dom_name =		 "route",
2031 	.dom_protosw =		routesw,
2032 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
2033 };
2034 
2035 VNET_DOMAIN_SET(route);
2036