xref: /freebsd/sys/net/rtsock.c (revision 3f0164abf32b9b761e0a2cb4bdca3a8b84f156d4)
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_dl.h>
57 #include <net/if_llatbl.h>
58 #include <net/if_types.h>
59 #include <net/netisr.h>
60 #include <net/raw_cb.h>
61 #include <net/route.h>
62 #include <net/vnet.h>
63 
64 #include <netinet/in.h>
65 #include <netinet/if_ether.h>
66 #include <netinet/ip_carp.h>
67 #ifdef INET6
68 #include <netinet6/ip6_var.h>
69 #include <netinet6/scope6_var.h>
70 #endif
71 
72 #if defined(INET) || defined(INET6)
73 #ifdef SCTP
74 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
75 #endif /* SCTP */
76 #endif
77 
78 #ifdef COMPAT_FREEBSD32
79 #include <sys/mount.h>
80 #include <compat/freebsd32/freebsd32.h>
81 
82 struct if_data32 {
83 	uint8_t	ifi_type;
84 	uint8_t	ifi_physical;
85 	uint8_t	ifi_addrlen;
86 	uint8_t	ifi_hdrlen;
87 	uint8_t	ifi_link_state;
88 	uint8_t	ifi_vhid;
89 	uint8_t	ifi_baudrate_pf;
90 	uint8_t	ifi_datalen;
91 	uint32_t ifi_mtu;
92 	uint32_t ifi_metric;
93 	uint32_t ifi_baudrate;
94 	uint32_t ifi_ipackets;
95 	uint32_t ifi_ierrors;
96 	uint32_t ifi_opackets;
97 	uint32_t ifi_oerrors;
98 	uint32_t ifi_collisions;
99 	uint32_t ifi_ibytes;
100 	uint32_t ifi_obytes;
101 	uint32_t ifi_imcasts;
102 	uint32_t ifi_omcasts;
103 	uint32_t ifi_iqdrops;
104 	uint32_t ifi_noproto;
105 	uint32_t ifi_hwassist;
106 	int32_t	ifi_epoch;
107 	struct	timeval32 ifi_lastchange;
108 };
109 
110 struct if_msghdr32 {
111 	uint16_t ifm_msglen;
112 	uint8_t	ifm_version;
113 	uint8_t	ifm_type;
114 	int32_t	ifm_addrs;
115 	int32_t	ifm_flags;
116 	uint16_t ifm_index;
117 	struct	if_data32 ifm_data;
118 };
119 
120 struct if_msghdrl32 {
121 	uint16_t ifm_msglen;
122 	uint8_t	ifm_version;
123 	uint8_t	ifm_type;
124 	int32_t	ifm_addrs;
125 	int32_t	ifm_flags;
126 	uint16_t ifm_index;
127 	uint16_t _ifm_spare1;
128 	uint16_t ifm_len;
129 	uint16_t ifm_data_off;
130 	struct	if_data32 ifm_data;
131 };
132 
133 struct ifa_msghdrl32 {
134 	uint16_t ifam_msglen;
135 	uint8_t	ifam_version;
136 	uint8_t	ifam_type;
137 	int32_t	ifam_addrs;
138 	int32_t	ifam_flags;
139 	uint16_t ifam_index;
140 	uint16_t _ifam_spare1;
141 	uint16_t ifam_len;
142 	uint16_t ifam_data_off;
143 	int32_t	ifam_metric;
144 	struct	if_data32 ifam_data;
145 };
146 #endif /* COMPAT_FREEBSD32 */
147 
148 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
149 
150 /* NB: these are not modified */
151 static struct	sockaddr route_src = { 2, PF_ROUTE, };
152 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
153 
154 /* These are external hooks for CARP. */
155 int	(*carp_get_vhid_p)(struct ifaddr *);
156 
157 /*
158  * Used by rtsock/raw_input callback code to decide whether to filter the update
159  * notification to a socket bound to a particular FIB.
160  */
161 #define	RTS_FILTER_FIB	M_PROTO8
162 #define	RTS_ALLFIBS	-1
163 
164 static struct {
165 	int	ip_count;	/* attached w/ AF_INET */
166 	int	ip6_count;	/* attached w/ AF_INET6 */
167 	int	ipx_count;	/* attached w/ AF_IPX */
168 	int	any_count;	/* total attached */
169 } route_cb;
170 
171 struct mtx rtsock_mtx;
172 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
173 
174 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
175 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
176 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
177 
178 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
179 
180 struct walkarg {
181 	int	w_tmemsize;
182 	int	w_op, w_arg;
183 	caddr_t	w_tmem;
184 	struct sysctl_req *w_req;
185 };
186 
187 static void	rts_input(struct mbuf *m);
188 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
189 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
190 			caddr_t cp, struct walkarg *w);
191 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
192 			struct rt_addrinfo *rtinfo);
193 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
194 static int	sysctl_iflist(int af, struct walkarg *w);
195 static int	sysctl_ifmalist(int af, struct walkarg *w);
196 static int	route_output(struct mbuf *m, struct socket *so);
197 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
198 			struct rt_metrics_lite *out);
199 static void	rt_getmetrics(const struct rt_metrics_lite *in,
200 			struct rt_metrics *out);
201 static void	rt_dispatch(struct mbuf *, sa_family_t);
202 
203 static struct netisr_handler rtsock_nh = {
204 	.nh_name = "rtsock",
205 	.nh_handler = rts_input,
206 	.nh_proto = NETISR_ROUTE,
207 	.nh_policy = NETISR_POLICY_SOURCE,
208 };
209 
210 static int
211 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
212 {
213 	int error, qlimit;
214 
215 	netisr_getqlimit(&rtsock_nh, &qlimit);
216 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
217         if (error || !req->newptr)
218                 return (error);
219 	if (qlimit < 1)
220 		return (EINVAL);
221 	return (netisr_setqlimit(&rtsock_nh, qlimit));
222 }
223 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
224     0, 0, sysctl_route_netisr_maxqlen, "I",
225     "maximum routing socket dispatch queue length");
226 
227 static void
228 rts_init(void)
229 {
230 	int tmp;
231 
232 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
233 		rtsock_nh.nh_qlimit = tmp;
234 	netisr_register(&rtsock_nh);
235 }
236 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
237 
238 static int
239 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
240     struct rawcb *rp)
241 {
242 	int fibnum;
243 
244 	KASSERT(m != NULL, ("%s: m is NULL", __func__));
245 	KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
246 	KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
247 
248 	/* No filtering requested. */
249 	if ((m->m_flags & RTS_FILTER_FIB) == 0)
250 		return (0);
251 
252 	/* Check if it is a rts and the fib matches the one of the socket. */
253 	fibnum = M_GETFIB(m);
254 	if (proto->sp_family != PF_ROUTE ||
255 	    rp->rcb_socket == NULL ||
256 	    rp->rcb_socket->so_fibnum == fibnum)
257 		return (0);
258 
259 	/* Filtering requested and no match, the socket shall be skipped. */
260 	return (1);
261 }
262 
263 static void
264 rts_input(struct mbuf *m)
265 {
266 	struct sockproto route_proto;
267 	unsigned short *family;
268 	struct m_tag *tag;
269 
270 	route_proto.sp_family = PF_ROUTE;
271 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
272 	if (tag != NULL) {
273 		family = (unsigned short *)(tag + 1);
274 		route_proto.sp_protocol = *family;
275 		m_tag_delete(m, tag);
276 	} else
277 		route_proto.sp_protocol = 0;
278 
279 	raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
280 }
281 
282 /*
283  * It really doesn't make any sense at all for this code to share much
284  * with raw_usrreq.c, since its functionality is so restricted.  XXX
285  */
286 static void
287 rts_abort(struct socket *so)
288 {
289 
290 	raw_usrreqs.pru_abort(so);
291 }
292 
293 static void
294 rts_close(struct socket *so)
295 {
296 
297 	raw_usrreqs.pru_close(so);
298 }
299 
300 /* pru_accept is EOPNOTSUPP */
301 
302 static int
303 rts_attach(struct socket *so, int proto, struct thread *td)
304 {
305 	struct rawcb *rp;
306 	int error;
307 
308 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
309 
310 	/* XXX */
311 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
312 	if (rp == NULL)
313 		return ENOBUFS;
314 
315 	so->so_pcb = (caddr_t)rp;
316 	so->so_fibnum = td->td_proc->p_fibnum;
317 	error = raw_attach(so, proto);
318 	rp = sotorawcb(so);
319 	if (error) {
320 		so->so_pcb = NULL;
321 		free(rp, M_PCB);
322 		return error;
323 	}
324 	RTSOCK_LOCK();
325 	switch(rp->rcb_proto.sp_protocol) {
326 	case AF_INET:
327 		route_cb.ip_count++;
328 		break;
329 	case AF_INET6:
330 		route_cb.ip6_count++;
331 		break;
332 	case AF_IPX:
333 		route_cb.ipx_count++;
334 		break;
335 	}
336 	route_cb.any_count++;
337 	RTSOCK_UNLOCK();
338 	soisconnected(so);
339 	so->so_options |= SO_USELOOPBACK;
340 	return 0;
341 }
342 
343 static int
344 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
345 {
346 
347 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
348 }
349 
350 static int
351 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
352 {
353 
354 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
355 }
356 
357 /* pru_connect2 is EOPNOTSUPP */
358 /* pru_control is EOPNOTSUPP */
359 
360 static void
361 rts_detach(struct socket *so)
362 {
363 	struct rawcb *rp = sotorawcb(so);
364 
365 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
366 
367 	RTSOCK_LOCK();
368 	switch(rp->rcb_proto.sp_protocol) {
369 	case AF_INET:
370 		route_cb.ip_count--;
371 		break;
372 	case AF_INET6:
373 		route_cb.ip6_count--;
374 		break;
375 	case AF_IPX:
376 		route_cb.ipx_count--;
377 		break;
378 	}
379 	route_cb.any_count--;
380 	RTSOCK_UNLOCK();
381 	raw_usrreqs.pru_detach(so);
382 }
383 
384 static int
385 rts_disconnect(struct socket *so)
386 {
387 
388 	return (raw_usrreqs.pru_disconnect(so));
389 }
390 
391 /* pru_listen is EOPNOTSUPP */
392 
393 static int
394 rts_peeraddr(struct socket *so, struct sockaddr **nam)
395 {
396 
397 	return (raw_usrreqs.pru_peeraddr(so, nam));
398 }
399 
400 /* pru_rcvd is EOPNOTSUPP */
401 /* pru_rcvoob is EOPNOTSUPP */
402 
403 static int
404 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
405 	 struct mbuf *control, struct thread *td)
406 {
407 
408 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
409 }
410 
411 /* pru_sense is null */
412 
413 static int
414 rts_shutdown(struct socket *so)
415 {
416 
417 	return (raw_usrreqs.pru_shutdown(so));
418 }
419 
420 static int
421 rts_sockaddr(struct socket *so, struct sockaddr **nam)
422 {
423 
424 	return (raw_usrreqs.pru_sockaddr(so, nam));
425 }
426 
427 static struct pr_usrreqs route_usrreqs = {
428 	.pru_abort =		rts_abort,
429 	.pru_attach =		rts_attach,
430 	.pru_bind =		rts_bind,
431 	.pru_connect =		rts_connect,
432 	.pru_detach =		rts_detach,
433 	.pru_disconnect =	rts_disconnect,
434 	.pru_peeraddr =		rts_peeraddr,
435 	.pru_send =		rts_send,
436 	.pru_shutdown =		rts_shutdown,
437 	.pru_sockaddr =		rts_sockaddr,
438 	.pru_close =		rts_close,
439 };
440 
441 #ifndef _SOCKADDR_UNION_DEFINED
442 #define	_SOCKADDR_UNION_DEFINED
443 /*
444  * The union of all possible address formats we handle.
445  */
446 union sockaddr_union {
447 	struct sockaddr		sa;
448 	struct sockaddr_in	sin;
449 	struct sockaddr_in6	sin6;
450 };
451 #endif /* _SOCKADDR_UNION_DEFINED */
452 
453 static int
454 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
455     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
456 {
457 
458 	/* First, see if the returned address is part of the jail. */
459 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
460 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
461 		return (0);
462 	}
463 
464 	switch (info->rti_info[RTAX_DST]->sa_family) {
465 #ifdef INET
466 	case AF_INET:
467 	{
468 		struct in_addr ia;
469 		struct ifaddr *ifa;
470 		int found;
471 
472 		found = 0;
473 		/*
474 		 * Try to find an address on the given outgoing interface
475 		 * that belongs to the jail.
476 		 */
477 		IF_ADDR_RLOCK(ifp);
478 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
479 			struct sockaddr *sa;
480 			sa = ifa->ifa_addr;
481 			if (sa->sa_family != AF_INET)
482 				continue;
483 			ia = ((struct sockaddr_in *)sa)->sin_addr;
484 			if (prison_check_ip4(cred, &ia) == 0) {
485 				found = 1;
486 				break;
487 			}
488 		}
489 		IF_ADDR_RUNLOCK(ifp);
490 		if (!found) {
491 			/*
492 			 * As a last resort return the 'default' jail address.
493 			 */
494 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
495 			    sin_addr;
496 			if (prison_get_ip4(cred, &ia) != 0)
497 				return (ESRCH);
498 		}
499 		bzero(&saun->sin, sizeof(struct sockaddr_in));
500 		saun->sin.sin_len = sizeof(struct sockaddr_in);
501 		saun->sin.sin_family = AF_INET;
502 		saun->sin.sin_addr.s_addr = ia.s_addr;
503 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
504 		break;
505 	}
506 #endif
507 #ifdef INET6
508 	case AF_INET6:
509 	{
510 		struct in6_addr ia6;
511 		struct ifaddr *ifa;
512 		int found;
513 
514 		found = 0;
515 		/*
516 		 * Try to find an address on the given outgoing interface
517 		 * that belongs to the jail.
518 		 */
519 		IF_ADDR_RLOCK(ifp);
520 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
521 			struct sockaddr *sa;
522 			sa = ifa->ifa_addr;
523 			if (sa->sa_family != AF_INET6)
524 				continue;
525 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
526 			    &ia6, sizeof(struct in6_addr));
527 			if (prison_check_ip6(cred, &ia6) == 0) {
528 				found = 1;
529 				break;
530 			}
531 		}
532 		IF_ADDR_RUNLOCK(ifp);
533 		if (!found) {
534 			/*
535 			 * As a last resort return the 'default' jail address.
536 			 */
537 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
538 			    sin6_addr;
539 			if (prison_get_ip6(cred, &ia6) != 0)
540 				return (ESRCH);
541 		}
542 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
543 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
544 		saun->sin6.sin6_family = AF_INET6;
545 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
546 		if (sa6_recoverscope(&saun->sin6) != 0)
547 			return (ESRCH);
548 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
549 		break;
550 	}
551 #endif
552 	default:
553 		return (ESRCH);
554 	}
555 	return (0);
556 }
557 
558 /*ARGSUSED*/
559 static int
560 route_output(struct mbuf *m, struct socket *so)
561 {
562 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
563 	struct rt_msghdr *rtm = NULL;
564 	struct rtentry *rt = NULL;
565 	struct radix_node_head *rnh;
566 	struct rt_addrinfo info;
567 #ifdef INET6
568 	struct sockaddr_storage ss;
569 	struct sockaddr_in6 *sin6;
570 	int i, rti_need_deembed = 0;
571 #endif
572 	int len, error = 0;
573 	struct ifnet *ifp = NULL;
574 	union sockaddr_union saun;
575 	sa_family_t saf = AF_UNSPEC;
576 
577 #define senderr(e) { error = e; goto flush;}
578 	if (m == NULL || ((m->m_len < sizeof(long)) &&
579 		       (m = m_pullup(m, sizeof(long))) == NULL))
580 		return (ENOBUFS);
581 	if ((m->m_flags & M_PKTHDR) == 0)
582 		panic("route_output");
583 	len = m->m_pkthdr.len;
584 	if (len < sizeof(*rtm) ||
585 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
586 		info.rti_info[RTAX_DST] = NULL;
587 		senderr(EINVAL);
588 	}
589 	R_Malloc(rtm, struct rt_msghdr *, len);
590 	if (rtm == NULL) {
591 		info.rti_info[RTAX_DST] = NULL;
592 		senderr(ENOBUFS);
593 	}
594 	m_copydata(m, 0, len, (caddr_t)rtm);
595 	if (rtm->rtm_version != RTM_VERSION) {
596 		info.rti_info[RTAX_DST] = NULL;
597 		senderr(EPROTONOSUPPORT);
598 	}
599 	rtm->rtm_pid = curproc->p_pid;
600 	bzero(&info, sizeof(info));
601 	info.rti_addrs = rtm->rtm_addrs;
602 	/*
603 	 * rt_xaddrs() performs s6_addr[2] := sin6_scope_id for AF_INET6
604 	 * link-local address because rtrequest requires addresses with
605 	 * embedded scope id.
606 	 */
607 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
608 		info.rti_info[RTAX_DST] = NULL;
609 		senderr(EINVAL);
610 	}
611 	info.rti_flags = rtm->rtm_flags;
612 	if (info.rti_info[RTAX_DST] == NULL ||
613 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
614 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
615 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
616 		senderr(EINVAL);
617 	saf = info.rti_info[RTAX_DST]->sa_family;
618 	/*
619 	 * Verify that the caller has the appropriate privilege; RTM_GET
620 	 * is the only operation the non-superuser is allowed.
621 	 */
622 	if (rtm->rtm_type != RTM_GET) {
623 		error = priv_check(curthread, PRIV_NET_ROUTE);
624 		if (error)
625 			senderr(error);
626 	}
627 
628 	/*
629 	 * The given gateway address may be an interface address.
630 	 * For example, issuing a "route change" command on a route
631 	 * entry that was created from a tunnel, and the gateway
632 	 * address given is the local end point. In this case the
633 	 * RTF_GATEWAY flag must be cleared or the destination will
634 	 * not be reachable even though there is no error message.
635 	 */
636 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
637 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
638 		struct route gw_ro;
639 
640 		bzero(&gw_ro, sizeof(gw_ro));
641 		gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
642 		rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
643 		/*
644 		 * A host route through the loopback interface is
645 		 * installed for each interface adddress. In pre 8.0
646 		 * releases the interface address of a PPP link type
647 		 * is not reachable locally. This behavior is fixed as
648 		 * part of the new L2/L3 redesign and rewrite work. The
649 		 * signature of this interface address route is the
650 		 * AF_LINK sa_family type of the rt_gateway, and the
651 		 * rt_ifp has the IFF_LOOPBACK flag set.
652 		 */
653 		if (gw_ro.ro_rt != NULL &&
654 		    gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
655 		    gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) {
656 			info.rti_flags &= ~RTF_GATEWAY;
657 			info.rti_flags |= RTF_GWFLAG_COMPAT;
658 		}
659 		if (gw_ro.ro_rt != NULL)
660 			RTFREE(gw_ro.ro_rt);
661 	}
662 
663 	switch (rtm->rtm_type) {
664 		struct rtentry *saved_nrt;
665 
666 	case RTM_ADD:
667 		if (info.rti_info[RTAX_GATEWAY] == NULL)
668 			senderr(EINVAL);
669 		saved_nrt = NULL;
670 
671 		/* support for new ARP code */
672 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
673 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
674 			error = lla_rt_output(rtm, &info);
675 #ifdef INET6
676 			if (error == 0)
677 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
678 #endif
679 			break;
680 		}
681 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
682 		    so->so_fibnum);
683 		if (error == 0 && saved_nrt) {
684 #ifdef INET6
685 			rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
686 #endif
687 			RT_LOCK(saved_nrt);
688 			rt_setmetrics(rtm->rtm_inits,
689 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
690 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
691 			RT_REMREF(saved_nrt);
692 			RT_UNLOCK(saved_nrt);
693 		}
694 		break;
695 
696 	case RTM_DELETE:
697 		saved_nrt = NULL;
698 		/* support for new ARP code */
699 		if (info.rti_info[RTAX_GATEWAY] &&
700 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
701 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
702 			error = lla_rt_output(rtm, &info);
703 #ifdef INET6
704 			if (error == 0)
705 				rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
706 #endif
707 			break;
708 		}
709 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
710 		    so->so_fibnum);
711 		if (error == 0) {
712 			RT_LOCK(saved_nrt);
713 			rt = saved_nrt;
714 			goto report;
715 		}
716 #ifdef INET6
717 		/* rt_msg2() will not be used when RTM_DELETE fails. */
718 		rti_need_deembed = (V_deembed_scopeid) ? 1 : 0;
719 #endif
720 		break;
721 
722 	case RTM_GET:
723 	case RTM_CHANGE:
724 	case RTM_LOCK:
725 		rnh = rt_tables_get_rnh(so->so_fibnum,
726 		    info.rti_info[RTAX_DST]->sa_family);
727 		if (rnh == NULL)
728 			senderr(EAFNOSUPPORT);
729 		RADIX_NODE_HEAD_RLOCK(rnh);
730 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
731 			info.rti_info[RTAX_NETMASK], rnh);
732 		if (rt == NULL) {	/* XXX looks bogus */
733 			RADIX_NODE_HEAD_RUNLOCK(rnh);
734 			senderr(ESRCH);
735 		}
736 #ifdef RADIX_MPATH
737 		/*
738 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
739 		 * we require users to specify a matching RTAX_GATEWAY.
740 		 *
741 		 * for RTM_GET, gate is optional even with multipath.
742 		 * if gate == NULL the first match is returned.
743 		 * (no need to call rt_mpath_matchgate if gate == NULL)
744 		 */
745 		if (rn_mpath_capable(rnh) &&
746 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
747 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
748 			if (!rt) {
749 				RADIX_NODE_HEAD_RUNLOCK(rnh);
750 				senderr(ESRCH);
751 			}
752 		}
753 #endif
754 		/*
755 		 * If performing proxied L2 entry insertion, and
756 		 * the actual PPP host entry is found, perform
757 		 * another search to retrieve the prefix route of
758 		 * the local end point of the PPP link.
759 		 */
760 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
761 			struct sockaddr laddr;
762 
763 			if (rt->rt_ifp != NULL &&
764 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
765 				struct ifaddr *ifa;
766 
767 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
768 				if (ifa != NULL)
769 					rt_maskedcopy(ifa->ifa_addr,
770 						      &laddr,
771 						      ifa->ifa_netmask);
772 			} else
773 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
774 					      &laddr,
775 					      rt->rt_ifa->ifa_netmask);
776 			/*
777 			 * refactor rt and no lock operation necessary
778 			 */
779 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
780 			if (rt == NULL) {
781 				RADIX_NODE_HEAD_RUNLOCK(rnh);
782 				senderr(ESRCH);
783 			}
784 		}
785 		RT_LOCK(rt);
786 		RT_ADDREF(rt);
787 		RADIX_NODE_HEAD_RUNLOCK(rnh);
788 
789 		/*
790 		 * Fix for PR: 82974
791 		 *
792 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
793 		 * returns a perfect match in case a netmask is
794 		 * specified.  For host routes only a longest prefix
795 		 * match is returned so it is necessary to compare the
796 		 * existence of the netmask.  If both have a netmask
797 		 * rnh_lookup() did a perfect match and if none of them
798 		 * have a netmask both are host routes which is also a
799 		 * perfect match.
800 		 */
801 
802 		if (rtm->rtm_type != RTM_GET &&
803 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
804 			RT_UNLOCK(rt);
805 			senderr(ESRCH);
806 		}
807 
808 		switch(rtm->rtm_type) {
809 
810 		case RTM_GET:
811 		report:
812 			RT_LOCK_ASSERT(rt);
813 			if ((rt->rt_flags & RTF_HOST) == 0
814 			    ? jailed_without_vnet(curthread->td_ucred)
815 			    : prison_if(curthread->td_ucred,
816 			    rt_key(rt)) != 0) {
817 				RT_UNLOCK(rt);
818 				senderr(ESRCH);
819 			}
820 			info.rti_info[RTAX_DST] = rt_key(rt);
821 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
822 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
823 			info.rti_info[RTAX_GENMASK] = 0;
824 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
825 				ifp = rt->rt_ifp;
826 				if (ifp) {
827 					info.rti_info[RTAX_IFP] =
828 					    ifp->if_addr->ifa_addr;
829 					error = rtm_get_jailed(&info, ifp, rt,
830 					    &saun, curthread->td_ucred);
831 					if (error != 0) {
832 						RT_UNLOCK(rt);
833 						senderr(error);
834 					}
835 					if (ifp->if_flags & IFF_POINTOPOINT)
836 						info.rti_info[RTAX_BRD] =
837 						    rt->rt_ifa->ifa_dstaddr;
838 					rtm->rtm_index = ifp->if_index;
839 				} else {
840 					info.rti_info[RTAX_IFP] = NULL;
841 					info.rti_info[RTAX_IFA] = NULL;
842 				}
843 			} else if ((ifp = rt->rt_ifp) != NULL) {
844 				rtm->rtm_index = ifp->if_index;
845 			}
846 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
847 			if (len > rtm->rtm_msglen) {
848 				struct rt_msghdr *new_rtm;
849 				R_Malloc(new_rtm, struct rt_msghdr *, len);
850 				if (new_rtm == NULL) {
851 					RT_UNLOCK(rt);
852 					senderr(ENOBUFS);
853 				}
854 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
855 				Free(rtm); rtm = new_rtm;
856 			}
857 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
858 			if (rt->rt_flags & RTF_GWFLAG_COMPAT)
859 				rtm->rtm_flags = RTF_GATEWAY |
860 					(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
861 			else
862 				rtm->rtm_flags = rt->rt_flags;
863 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
864 			rtm->rtm_addrs = info.rti_addrs;
865 			break;
866 
867 		case RTM_CHANGE:
868 			/*
869 			 * New gateway could require new ifaddr, ifp;
870 			 * flags may also be different; ifp may be specified
871 			 * by ll sockaddr when protocol address is ambiguous
872 			 */
873 			if (((rt->rt_flags & RTF_GATEWAY) &&
874 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
875 			    info.rti_info[RTAX_IFP] != NULL ||
876 			    (info.rti_info[RTAX_IFA] != NULL &&
877 			     !sa_equal(info.rti_info[RTAX_IFA],
878 				       rt->rt_ifa->ifa_addr))) {
879 				RT_UNLOCK(rt);
880 				RADIX_NODE_HEAD_LOCK(rnh);
881 				error = rt_getifa_fib(&info, rt->rt_fibnum);
882 				/*
883 				 * XXXRW: Really we should release this
884 				 * reference later, but this maintains
885 				 * historical behavior.
886 				 */
887 				if (info.rti_ifa != NULL)
888 					ifa_free(info.rti_ifa);
889 				RADIX_NODE_HEAD_UNLOCK(rnh);
890 				if (error != 0)
891 					senderr(error);
892 				RT_LOCK(rt);
893 			}
894 			if (info.rti_ifa != NULL &&
895 			    info.rti_ifa != rt->rt_ifa &&
896 			    rt->rt_ifa != NULL &&
897 			    rt->rt_ifa->ifa_rtrequest != NULL) {
898 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
899 				    &info);
900 				ifa_free(rt->rt_ifa);
901 			}
902 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
903 				RT_UNLOCK(rt);
904 				RADIX_NODE_HEAD_LOCK(rnh);
905 				RT_LOCK(rt);
906 
907 				error = rt_setgate(rt, rt_key(rt),
908 				    info.rti_info[RTAX_GATEWAY]);
909 				RADIX_NODE_HEAD_UNLOCK(rnh);
910 				if (error != 0) {
911 					RT_UNLOCK(rt);
912 					senderr(error);
913 				}
914 				rt->rt_flags &= ~RTF_GATEWAY;
915 				rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
916 			}
917 			if (info.rti_ifa != NULL &&
918 			    info.rti_ifa != rt->rt_ifa) {
919 				ifa_ref(info.rti_ifa);
920 				rt->rt_ifa = info.rti_ifa;
921 				rt->rt_ifp = info.rti_ifp;
922 			}
923 			/* Allow some flags to be toggled on change. */
924 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
925 				    (rtm->rtm_flags & RTF_FMASK);
926 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
927 					&rt->rt_rmx);
928 			rtm->rtm_index = rt->rt_ifp->if_index;
929 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
930 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
931 			/* FALLTHROUGH */
932 		case RTM_LOCK:
933 			/* We don't support locks anymore */
934 			break;
935 		}
936 		RT_UNLOCK(rt);
937 		break;
938 
939 	default:
940 		senderr(EOPNOTSUPP);
941 	}
942 
943 flush:
944 	if (rtm) {
945 		if (error)
946 			rtm->rtm_errno = error;
947 		else
948 			rtm->rtm_flags |= RTF_DONE;
949 	}
950 	if (rt)		/* XXX can this be true? */
951 		RTFREE(rt);
952     {
953 	struct rawcb *rp = NULL;
954 	/*
955 	 * Check to see if we don't want our own messages.
956 	 */
957 	if ((so->so_options & SO_USELOOPBACK) == 0) {
958 		if (route_cb.any_count <= 1) {
959 			if (rtm)
960 				Free(rtm);
961 			m_freem(m);
962 			return (error);
963 		}
964 		/* There is another listener, so construct message */
965 		rp = sotorawcb(so);
966 	}
967 	if (rtm) {
968 #ifdef INET6
969 		if (rti_need_deembed) {
970 			/* sin6_scope_id is recovered before sending rtm. */
971 			sin6 = (struct sockaddr_in6 *)&ss;
972 			for (i = 0; i < RTAX_MAX; i++) {
973 				if (info.rti_info[i] == NULL)
974 					continue;
975 				if (info.rti_info[i]->sa_family != AF_INET6)
976 					continue;
977 				bcopy(info.rti_info[i], sin6, sizeof(*sin6));
978 				if (sa6_recoverscope(sin6) == 0)
979 					bcopy(sin6, info.rti_info[i],
980 						    sizeof(*sin6));
981 			}
982 		}
983 #endif
984 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
985 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
986 			m_freem(m);
987 			m = NULL;
988 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
989 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
990 	}
991 	if (m) {
992 		M_SETFIB(m, so->so_fibnum);
993 		m->m_flags |= RTS_FILTER_FIB;
994 		if (rp) {
995 			/*
996 			 * XXX insure we don't get a copy by
997 			 * invalidating our protocol
998 			 */
999 			unsigned short family = rp->rcb_proto.sp_family;
1000 			rp->rcb_proto.sp_family = 0;
1001 			rt_dispatch(m, saf);
1002 			rp->rcb_proto.sp_family = family;
1003 		} else
1004 			rt_dispatch(m, saf);
1005 	}
1006 	/* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
1007 	if (rtm)
1008 		Free(rtm);
1009     }
1010 	return (error);
1011 #undef	sa_equal
1012 }
1013 
1014 static void
1015 rt_setmetrics(u_long which, const struct rt_metrics *in,
1016 	struct rt_metrics_lite *out)
1017 {
1018 #define metric(f, e) if (which & (f)) out->e = in->e;
1019 	/*
1020 	 * Only these are stored in the routing entry since introduction
1021 	 * of tcp hostcache. The rest is ignored.
1022 	 */
1023 	metric(RTV_MTU, rmx_mtu);
1024 	metric(RTV_WEIGHT, rmx_weight);
1025 	/* Userland -> kernel timebase conversion. */
1026 	if (which & RTV_EXPIRE)
1027 		out->rmx_expire = in->rmx_expire ?
1028 		    in->rmx_expire - time_second + time_uptime : 0;
1029 #undef metric
1030 }
1031 
1032 static void
1033 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
1034 {
1035 #define metric(e) out->e = in->e;
1036 	bzero(out, sizeof(*out));
1037 	metric(rmx_mtu);
1038 	metric(rmx_weight);
1039 	/* Kernel -> userland timebase conversion. */
1040 	out->rmx_expire = in->rmx_expire ?
1041 	    in->rmx_expire - time_uptime + time_second : 0;
1042 #undef metric
1043 }
1044 
1045 /*
1046  * Extract the addresses of the passed sockaddrs.
1047  * Do a little sanity checking so as to avoid bad memory references.
1048  * This data is derived straight from userland.
1049  */
1050 static int
1051 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
1052 {
1053 	struct sockaddr *sa;
1054 	int i;
1055 
1056 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1057 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1058 			continue;
1059 		sa = (struct sockaddr *)cp;
1060 		/*
1061 		 * It won't fit.
1062 		 */
1063 		if (cp + sa->sa_len > cplim)
1064 			return (EINVAL);
1065 		/*
1066 		 * there are no more.. quit now
1067 		 * If there are more bits, they are in error.
1068 		 * I've seen this. route(1) can evidently generate these.
1069 		 * This causes kernel to core dump.
1070 		 * for compatibility, If we see this, point to a safe address.
1071 		 */
1072 		if (sa->sa_len == 0) {
1073 			rtinfo->rti_info[i] = &sa_zero;
1074 			return (0); /* should be EINVAL but for compat */
1075 		}
1076 		/* accept it */
1077 #ifdef INET6
1078 		if (sa->sa_family == AF_INET6)
1079 			sa6_embedscope((struct sockaddr_in6 *)sa,
1080 			    V_ip6_use_defzone);
1081 #endif
1082 		rtinfo->rti_info[i] = sa;
1083 		cp += SA_SIZE(sa);
1084 	}
1085 	return (0);
1086 }
1087 
1088 /*
1089  * Used by the routing socket.
1090  */
1091 static struct mbuf *
1092 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1093 {
1094 	struct rt_msghdr *rtm;
1095 	struct mbuf *m;
1096 	int i;
1097 	struct sockaddr *sa;
1098 #ifdef INET6
1099 	struct sockaddr_storage ss;
1100 	struct sockaddr_in6 *sin6;
1101 #endif
1102 	int len, dlen;
1103 
1104 	switch (type) {
1105 
1106 	case RTM_DELADDR:
1107 	case RTM_NEWADDR:
1108 		len = sizeof(struct ifa_msghdr);
1109 		break;
1110 
1111 	case RTM_DELMADDR:
1112 	case RTM_NEWMADDR:
1113 		len = sizeof(struct ifma_msghdr);
1114 		break;
1115 
1116 	case RTM_IFINFO:
1117 		len = sizeof(struct if_msghdr);
1118 		break;
1119 
1120 	case RTM_IFANNOUNCE:
1121 	case RTM_IEEE80211:
1122 		len = sizeof(struct if_announcemsghdr);
1123 		break;
1124 
1125 	default:
1126 		len = sizeof(struct rt_msghdr);
1127 	}
1128 
1129 	/* XXXGL: can we use MJUMPAGESIZE cluster here? */
1130 	KASSERT(len <= MCLBYTES, ("%s: message too big", __func__));
1131 	if (len > MHLEN)
1132 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1133 	else
1134 		m = m_gethdr(M_NOWAIT, MT_DATA);
1135 	if (m == NULL)
1136 		return (m);
1137 
1138 	m->m_pkthdr.len = m->m_len = len;
1139 	rtm = mtod(m, struct rt_msghdr *);
1140 	bzero((caddr_t)rtm, len);
1141 	for (i = 0; i < RTAX_MAX; i++) {
1142 		if ((sa = rtinfo->rti_info[i]) == NULL)
1143 			continue;
1144 		rtinfo->rti_addrs |= (1 << i);
1145 		dlen = SA_SIZE(sa);
1146 #ifdef INET6
1147 		if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1148 			sin6 = (struct sockaddr_in6 *)&ss;
1149 			bcopy(sa, sin6, sizeof(*sin6));
1150 			if (sa6_recoverscope(sin6) == 0)
1151 				sa = (struct sockaddr *)sin6;
1152 		}
1153 #endif
1154 		m_copyback(m, len, dlen, (caddr_t)sa);
1155 		len += dlen;
1156 	}
1157 	if (m->m_pkthdr.len != len) {
1158 		m_freem(m);
1159 		return (NULL);
1160 	}
1161 	rtm->rtm_msglen = len;
1162 	rtm->rtm_version = RTM_VERSION;
1163 	rtm->rtm_type = type;
1164 	return (m);
1165 }
1166 
1167 /*
1168  * Used by the sysctl code and routing socket.
1169  */
1170 static int
1171 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1172 {
1173 	int i;
1174 	int len, dlen, second_time = 0;
1175 	caddr_t cp0;
1176 #ifdef INET6
1177 	struct sockaddr_storage ss;
1178 	struct sockaddr_in6 *sin6;
1179 #endif
1180 
1181 	rtinfo->rti_addrs = 0;
1182 again:
1183 	switch (type) {
1184 
1185 	case RTM_DELADDR:
1186 	case RTM_NEWADDR:
1187 		if (w != NULL && w->w_op == NET_RT_IFLISTL) {
1188 #ifdef COMPAT_FREEBSD32
1189 			if (w->w_req->flags & SCTL_MASK32)
1190 				len = sizeof(struct ifa_msghdrl32);
1191 			else
1192 #endif
1193 				len = sizeof(struct ifa_msghdrl);
1194 		} else
1195 			len = sizeof(struct ifa_msghdr);
1196 		break;
1197 
1198 	case RTM_IFINFO:
1199 #ifdef COMPAT_FREEBSD32
1200 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1201 			if (w->w_op == NET_RT_IFLISTL)
1202 				len = sizeof(struct if_msghdrl32);
1203 			else
1204 				len = sizeof(struct if_msghdr32);
1205 			break;
1206 		}
1207 #endif
1208 		if (w != NULL && w->w_op == NET_RT_IFLISTL)
1209 			len = sizeof(struct if_msghdrl);
1210 		else
1211 			len = sizeof(struct if_msghdr);
1212 		break;
1213 
1214 	case RTM_NEWMADDR:
1215 		len = sizeof(struct ifma_msghdr);
1216 		break;
1217 
1218 	default:
1219 		len = sizeof(struct rt_msghdr);
1220 	}
1221 	cp0 = cp;
1222 	if (cp0)
1223 		cp += len;
1224 	for (i = 0; i < RTAX_MAX; i++) {
1225 		struct sockaddr *sa;
1226 
1227 		if ((sa = rtinfo->rti_info[i]) == NULL)
1228 			continue;
1229 		rtinfo->rti_addrs |= (1 << i);
1230 		dlen = SA_SIZE(sa);
1231 		if (cp) {
1232 #ifdef INET6
1233 			if (V_deembed_scopeid && sa->sa_family == AF_INET6) {
1234 				sin6 = (struct sockaddr_in6 *)&ss;
1235 				bcopy(sa, sin6, sizeof(*sin6));
1236 				if (sa6_recoverscope(sin6) == 0)
1237 					sa = (struct sockaddr *)sin6;
1238 			}
1239 #endif
1240 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1241 			cp += dlen;
1242 		}
1243 		len += dlen;
1244 	}
1245 	len = ALIGN(len);
1246 	if (cp == NULL && w != NULL && !second_time) {
1247 		struct walkarg *rw = w;
1248 
1249 		if (rw->w_req) {
1250 			if (rw->w_tmemsize < len) {
1251 				if (rw->w_tmem)
1252 					free(rw->w_tmem, M_RTABLE);
1253 				rw->w_tmem = (caddr_t)
1254 					malloc(len, M_RTABLE, M_NOWAIT);
1255 				if (rw->w_tmem)
1256 					rw->w_tmemsize = len;
1257 			}
1258 			if (rw->w_tmem) {
1259 				cp = rw->w_tmem;
1260 				second_time = 1;
1261 				goto again;
1262 			}
1263 		}
1264 	}
1265 	if (cp) {
1266 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1267 
1268 		rtm->rtm_version = RTM_VERSION;
1269 		rtm->rtm_type = type;
1270 		rtm->rtm_msglen = len;
1271 	}
1272 	return (len);
1273 }
1274 
1275 /*
1276  * This routine is called to generate a message from the routing
1277  * socket indicating that a redirect has occured, a routing lookup
1278  * has failed, or that a protocol has detected timeouts to a particular
1279  * destination.
1280  */
1281 void
1282 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1283     int fibnum)
1284 {
1285 	struct rt_msghdr *rtm;
1286 	struct mbuf *m;
1287 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1288 
1289 	if (route_cb.any_count == 0)
1290 		return;
1291 	m = rt_msg1(type, rtinfo);
1292 	if (m == NULL)
1293 		return;
1294 
1295 	if (fibnum != RTS_ALLFIBS) {
1296 		KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1297 		    "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1298 		M_SETFIB(m, fibnum);
1299 		m->m_flags |= RTS_FILTER_FIB;
1300 	}
1301 
1302 	rtm = mtod(m, struct rt_msghdr *);
1303 	rtm->rtm_flags = RTF_DONE | flags;
1304 	rtm->rtm_errno = error;
1305 	rtm->rtm_addrs = rtinfo->rti_addrs;
1306 	rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1307 }
1308 
1309 void
1310 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1311 {
1312 
1313 	rt_missmsg_fib(type, rtinfo, flags, error, RTS_ALLFIBS);
1314 }
1315 
1316 /*
1317  * This routine is called to generate a message from the routing
1318  * socket indicating that the status of a network interface has changed.
1319  */
1320 void
1321 rt_ifmsg(struct ifnet *ifp)
1322 {
1323 	struct if_msghdr *ifm;
1324 	struct mbuf *m;
1325 	struct rt_addrinfo info;
1326 
1327 	if (route_cb.any_count == 0)
1328 		return;
1329 	bzero((caddr_t)&info, sizeof(info));
1330 	m = rt_msg1(RTM_IFINFO, &info);
1331 	if (m == NULL)
1332 		return;
1333 	ifm = mtod(m, struct if_msghdr *);
1334 	ifm->ifm_index = ifp->if_index;
1335 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1336 	ifm->ifm_data = ifp->if_data;
1337 	ifm->ifm_addrs = 0;
1338 	rt_dispatch(m, AF_UNSPEC);
1339 }
1340 
1341 /*
1342  * This is called to generate messages from the routing socket
1343  * indicating a network interface has had addresses associated with it.
1344  * if we ever reverse the logic and replace messages TO the routing
1345  * socket indicate a request to configure interfaces, then it will
1346  * be unnecessary as the routing socket will automatically generate
1347  * copies of it.
1348  */
1349 void
1350 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
1351     int fibnum)
1352 {
1353 	struct rt_addrinfo info;
1354 	struct sockaddr *sa = NULL;
1355 	int pass;
1356 	struct mbuf *m = NULL;
1357 	struct ifnet *ifp = ifa->ifa_ifp;
1358 
1359 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1360 		("unexpected cmd %u", cmd));
1361 #if defined(INET) || defined(INET6)
1362 #ifdef SCTP
1363 	/*
1364 	 * notify the SCTP stack
1365 	 * this will only get called when an address is added/deleted
1366 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1367 	 */
1368 	sctp_addr_change(ifa, cmd);
1369 #endif /* SCTP */
1370 #endif
1371 	if (route_cb.any_count == 0)
1372 		return;
1373 	for (pass = 1; pass < 3; pass++) {
1374 		bzero((caddr_t)&info, sizeof(info));
1375 		if ((cmd == RTM_ADD && pass == 1) ||
1376 		    (cmd == RTM_DELETE && pass == 2)) {
1377 			struct ifa_msghdr *ifam;
1378 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1379 
1380 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1381 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1382 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1383 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1384 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1385 				continue;
1386 			ifam = mtod(m, struct ifa_msghdr *);
1387 			ifam->ifam_index = ifp->if_index;
1388 			ifam->ifam_metric = ifa->ifa_metric;
1389 			ifam->ifam_flags = ifa->ifa_flags;
1390 			ifam->ifam_addrs = info.rti_addrs;
1391 		}
1392 		if ((cmd == RTM_ADD && pass == 2) ||
1393 		    (cmd == RTM_DELETE && pass == 1)) {
1394 			struct rt_msghdr *rtm;
1395 
1396 			if (rt == NULL)
1397 				continue;
1398 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1399 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1400 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1401 			if ((m = rt_msg1(cmd, &info)) == NULL)
1402 				continue;
1403 			rtm = mtod(m, struct rt_msghdr *);
1404 			rtm->rtm_index = ifp->if_index;
1405 			rtm->rtm_flags |= rt->rt_flags;
1406 			rtm->rtm_errno = error;
1407 			rtm->rtm_addrs = info.rti_addrs;
1408 		}
1409 		if (fibnum != RTS_ALLFIBS) {
1410 			KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: "
1411 			    "fibnum out of range 0 <= %d < %d", __func__,
1412 			     fibnum, rt_numfibs));
1413 			M_SETFIB(m, fibnum);
1414 			m->m_flags |= RTS_FILTER_FIB;
1415 		}
1416 		rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1417 	}
1418 }
1419 
1420 void
1421 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1422 {
1423 
1424 	rt_newaddrmsg_fib(cmd, ifa, error, rt, RTS_ALLFIBS);
1425 }
1426 
1427 /*
1428  * This is the analogue to the rt_newaddrmsg which performs the same
1429  * function but for multicast group memberhips.  This is easier since
1430  * there is no route state to worry about.
1431  */
1432 void
1433 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1434 {
1435 	struct rt_addrinfo info;
1436 	struct mbuf *m = NULL;
1437 	struct ifnet *ifp = ifma->ifma_ifp;
1438 	struct ifma_msghdr *ifmam;
1439 
1440 	if (route_cb.any_count == 0)
1441 		return;
1442 
1443 	bzero((caddr_t)&info, sizeof(info));
1444 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1445 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1446 	/*
1447 	 * If a link-layer address is present, present it as a ``gateway''
1448 	 * (similarly to how ARP entries, e.g., are presented).
1449 	 */
1450 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1451 	m = rt_msg1(cmd, &info);
1452 	if (m == NULL)
1453 		return;
1454 	ifmam = mtod(m, struct ifma_msghdr *);
1455 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1456 	    __func__));
1457 	ifmam->ifmam_index = ifp->if_index;
1458 	ifmam->ifmam_addrs = info.rti_addrs;
1459 	rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1460 }
1461 
1462 static struct mbuf *
1463 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1464 	struct rt_addrinfo *info)
1465 {
1466 	struct if_announcemsghdr *ifan;
1467 	struct mbuf *m;
1468 
1469 	if (route_cb.any_count == 0)
1470 		return NULL;
1471 	bzero((caddr_t)info, sizeof(*info));
1472 	m = rt_msg1(type, info);
1473 	if (m != NULL) {
1474 		ifan = mtod(m, struct if_announcemsghdr *);
1475 		ifan->ifan_index = ifp->if_index;
1476 		strlcpy(ifan->ifan_name, ifp->if_xname,
1477 			sizeof(ifan->ifan_name));
1478 		ifan->ifan_what = what;
1479 	}
1480 	return m;
1481 }
1482 
1483 /*
1484  * This is called to generate routing socket messages indicating
1485  * IEEE80211 wireless events.
1486  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1487  */
1488 void
1489 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1490 {
1491 	struct mbuf *m;
1492 	struct rt_addrinfo info;
1493 
1494 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1495 	if (m != NULL) {
1496 		/*
1497 		 * Append the ieee80211 data.  Try to stick it in the
1498 		 * mbuf containing the ifannounce msg; otherwise allocate
1499 		 * a new mbuf and append.
1500 		 *
1501 		 * NB: we assume m is a single mbuf.
1502 		 */
1503 		if (data_len > M_TRAILINGSPACE(m)) {
1504 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1505 			if (n == NULL) {
1506 				m_freem(m);
1507 				return;
1508 			}
1509 			bcopy(data, mtod(n, void *), data_len);
1510 			n->m_len = data_len;
1511 			m->m_next = n;
1512 		} else if (data_len > 0) {
1513 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1514 			m->m_len += data_len;
1515 		}
1516 		if (m->m_flags & M_PKTHDR)
1517 			m->m_pkthdr.len += data_len;
1518 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1519 		rt_dispatch(m, AF_UNSPEC);
1520 	}
1521 }
1522 
1523 /*
1524  * This is called to generate routing socket messages indicating
1525  * network interface arrival and departure.
1526  */
1527 void
1528 rt_ifannouncemsg(struct ifnet *ifp, int what)
1529 {
1530 	struct mbuf *m;
1531 	struct rt_addrinfo info;
1532 
1533 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1534 	if (m != NULL)
1535 		rt_dispatch(m, AF_UNSPEC);
1536 }
1537 
1538 static void
1539 rt_dispatch(struct mbuf *m, sa_family_t saf)
1540 {
1541 	struct m_tag *tag;
1542 
1543 	/*
1544 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1545 	 * use when injecting the mbuf into the routing socket buffer from
1546 	 * the netisr.
1547 	 */
1548 	if (saf != AF_UNSPEC) {
1549 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1550 		    M_NOWAIT);
1551 		if (tag == NULL) {
1552 			m_freem(m);
1553 			return;
1554 		}
1555 		*(unsigned short *)(tag + 1) = saf;
1556 		m_tag_prepend(m, tag);
1557 	}
1558 #ifdef VIMAGE
1559 	if (V_loif)
1560 		m->m_pkthdr.rcvif = V_loif;
1561 	else {
1562 		m_freem(m);
1563 		return;
1564 	}
1565 #endif
1566 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1567 }
1568 
1569 /*
1570  * This is used in dumping the kernel table via sysctl().
1571  */
1572 static int
1573 sysctl_dumpentry(struct radix_node *rn, void *vw)
1574 {
1575 	struct walkarg *w = vw;
1576 	struct rtentry *rt = (struct rtentry *)rn;
1577 	int error = 0, size;
1578 	struct rt_addrinfo info;
1579 
1580 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1581 		return 0;
1582 	if ((rt->rt_flags & RTF_HOST) == 0
1583 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1584 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1585 		return (0);
1586 	bzero((caddr_t)&info, sizeof(info));
1587 	info.rti_info[RTAX_DST] = rt_key(rt);
1588 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1589 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1590 	info.rti_info[RTAX_GENMASK] = 0;
1591 	if (rt->rt_ifp) {
1592 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1593 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1594 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1595 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1596 	}
1597 	size = rt_msg2(RTM_GET, &info, NULL, w);
1598 	if (w->w_req && w->w_tmem) {
1599 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1600 
1601 		if (rt->rt_flags & RTF_GWFLAG_COMPAT)
1602 			rtm->rtm_flags = RTF_GATEWAY |
1603 				(rt->rt_flags & ~RTF_GWFLAG_COMPAT);
1604 		else
1605 			rtm->rtm_flags = rt->rt_flags;
1606 		/*
1607 		 * let's be honest about this being a retarded hack
1608 		 */
1609 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1610 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1611 		rtm->rtm_index = rt->rt_ifp->if_index;
1612 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1613 		rtm->rtm_addrs = info.rti_addrs;
1614 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1615 		return (error);
1616 	}
1617 	return (error);
1618 }
1619 
1620 #ifdef COMPAT_FREEBSD32
1621 static void
1622 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1623 {
1624 
1625 	bzero(dst, sizeof(*dst));
1626 	CP(*src, *dst, ifi_type);
1627 	CP(*src, *dst, ifi_physical);
1628 	CP(*src, *dst, ifi_addrlen);
1629 	CP(*src, *dst, ifi_hdrlen);
1630 	CP(*src, *dst, ifi_link_state);
1631 	CP(*src, *dst, ifi_vhid);
1632 	CP(*src, *dst, ifi_baudrate_pf);
1633 	dst->ifi_datalen = sizeof(struct if_data32);
1634 	CP(*src, *dst, ifi_mtu);
1635 	CP(*src, *dst, ifi_metric);
1636 	CP(*src, *dst, ifi_baudrate);
1637 	CP(*src, *dst, ifi_ipackets);
1638 	CP(*src, *dst, ifi_ierrors);
1639 	CP(*src, *dst, ifi_opackets);
1640 	CP(*src, *dst, ifi_oerrors);
1641 	CP(*src, *dst, ifi_collisions);
1642 	CP(*src, *dst, ifi_ibytes);
1643 	CP(*src, *dst, ifi_obytes);
1644 	CP(*src, *dst, ifi_imcasts);
1645 	CP(*src, *dst, ifi_omcasts);
1646 	CP(*src, *dst, ifi_iqdrops);
1647 	CP(*src, *dst, ifi_noproto);
1648 	CP(*src, *dst, ifi_hwassist);
1649 	CP(*src, *dst, ifi_epoch);
1650 	TV_CP(*src, *dst, ifi_lastchange);
1651 }
1652 #endif
1653 
1654 static int
1655 sysctl_iflist_ifml(struct ifnet *ifp, struct rt_addrinfo *info,
1656     struct walkarg *w, int len)
1657 {
1658 	struct if_msghdrl *ifm;
1659 
1660 #ifdef COMPAT_FREEBSD32
1661 	if (w->w_req->flags & SCTL_MASK32) {
1662 		struct if_msghdrl32 *ifm32;
1663 
1664 		ifm32 = (struct if_msghdrl32 *)w->w_tmem;
1665 		ifm32->ifm_addrs = info->rti_addrs;
1666 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1667 		ifm32->ifm_index = ifp->if_index;
1668 		ifm32->_ifm_spare1 = 0;
1669 		ifm32->ifm_len = sizeof(*ifm32);
1670 		ifm32->ifm_data_off = offsetof(struct if_msghdrl32, ifm_data);
1671 
1672 		copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1673 		/* Fixup if_data carp(4) vhid. */
1674 		if (carp_get_vhid_p != NULL)
1675 			ifm32->ifm_data.ifi_vhid =
1676 			    (*carp_get_vhid_p)(ifp->if_addr);
1677 
1678 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1679 	}
1680 #endif
1681 	ifm = (struct if_msghdrl *)w->w_tmem;
1682 	ifm->ifm_addrs = info->rti_addrs;
1683 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1684 	ifm->ifm_index = ifp->if_index;
1685 	ifm->_ifm_spare1 = 0;
1686 	ifm->ifm_len = sizeof(*ifm);
1687 	ifm->ifm_data_off = offsetof(struct if_msghdrl, ifm_data);
1688 
1689 	ifm->ifm_data = ifp->if_data;
1690 	/* Fixup if_data carp(4) vhid. */
1691 	if (carp_get_vhid_p != NULL)
1692 		ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1693 
1694 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1695 }
1696 
1697 static int
1698 sysctl_iflist_ifm(struct ifnet *ifp, struct rt_addrinfo *info,
1699     struct walkarg *w, int len)
1700 {
1701 	struct if_msghdr *ifm;
1702 
1703 #ifdef COMPAT_FREEBSD32
1704 	if (w->w_req->flags & SCTL_MASK32) {
1705 		struct if_msghdr32 *ifm32;
1706 
1707 		ifm32 = (struct if_msghdr32 *)w->w_tmem;
1708 		ifm32->ifm_addrs = info->rti_addrs;
1709 		ifm32->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1710 		ifm32->ifm_index = ifp->if_index;
1711 
1712 		copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1713 		/* Fixup if_data carp(4) vhid. */
1714 		if (carp_get_vhid_p != NULL)
1715 			ifm32->ifm_data.ifi_vhid =
1716 			    (*carp_get_vhid_p)(ifp->if_addr);
1717 
1718 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifm32, len));
1719 	}
1720 #endif
1721 	ifm = (struct if_msghdr *)w->w_tmem;
1722 	ifm->ifm_addrs = info->rti_addrs;
1723 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1724 	ifm->ifm_index = ifp->if_index;
1725 
1726 	ifm->ifm_data = ifp->if_data;
1727 	/* Fixup if_data carp(4) vhid. */
1728 	if (carp_get_vhid_p != NULL)
1729 		ifm->ifm_data.ifi_vhid = (*carp_get_vhid_p)(ifp->if_addr);
1730 
1731 	return (SYSCTL_OUT(w->w_req, (caddr_t)ifm, len));
1732 }
1733 
1734 static int
1735 sysctl_iflist_ifaml(struct ifaddr *ifa, struct rt_addrinfo *info,
1736     struct walkarg *w, int len)
1737 {
1738 	struct ifa_msghdrl *ifam;
1739 
1740 #ifdef COMPAT_FREEBSD32
1741 	if (w->w_req->flags & SCTL_MASK32) {
1742 		struct ifa_msghdrl32 *ifam32;
1743 
1744 		ifam32 = (struct ifa_msghdrl32 *)w->w_tmem;
1745 		ifam32->ifam_addrs = info->rti_addrs;
1746 		ifam32->ifam_flags = ifa->ifa_flags;
1747 		ifam32->ifam_index = ifa->ifa_ifp->if_index;
1748 		ifam32->_ifam_spare1 = 0;
1749 		ifam32->ifam_len = sizeof(*ifam32);
1750 		ifam32->ifam_data_off =
1751 		    offsetof(struct ifa_msghdrl32, ifam_data);
1752 		ifam32->ifam_metric = ifa->ifa_metric;
1753 
1754 		copy_ifdata32(&ifa->ifa_ifp->if_data, &ifam32->ifam_data);
1755 		/* Fixup if_data carp(4) vhid. */
1756 		if (carp_get_vhid_p != NULL)
1757 			ifam32->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1758 
1759 		return (SYSCTL_OUT(w->w_req, (caddr_t)ifam32, len));
1760 	}
1761 #endif
1762 
1763 	ifam = (struct ifa_msghdrl *)w->w_tmem;
1764 	ifam->ifam_addrs = info->rti_addrs;
1765 	ifam->ifam_flags = ifa->ifa_flags;
1766 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1767 	ifam->_ifam_spare1 = 0;
1768 	ifam->ifam_len = sizeof(*ifam);
1769 	ifam->ifam_data_off = offsetof(struct ifa_msghdrl, ifam_data);
1770 	ifam->ifam_metric = ifa->ifa_metric;
1771 
1772 	ifam->ifam_data = ifa->if_data;
1773 	/* Fixup if_data carp(4) vhid. */
1774 	if (carp_get_vhid_p != NULL)
1775 		ifam->ifam_data.ifi_vhid = (*carp_get_vhid_p)(ifa);
1776 
1777 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1778 }
1779 
1780 static int
1781 sysctl_iflist_ifam(struct ifaddr *ifa, struct rt_addrinfo *info,
1782     struct walkarg *w, int len)
1783 {
1784 	struct ifa_msghdr *ifam;
1785 
1786 	ifam = (struct ifa_msghdr *)w->w_tmem;
1787 	ifam->ifam_addrs = info->rti_addrs;
1788 	ifam->ifam_flags = ifa->ifa_flags;
1789 	ifam->ifam_index = ifa->ifa_ifp->if_index;
1790 	ifam->ifam_metric = ifa->ifa_metric;
1791 
1792 	return (SYSCTL_OUT(w->w_req, w->w_tmem, len));
1793 }
1794 
1795 static int
1796 sysctl_iflist(int af, struct walkarg *w)
1797 {
1798 	struct ifnet *ifp;
1799 	struct ifaddr *ifa;
1800 	struct rt_addrinfo info;
1801 	int len, error = 0;
1802 
1803 	bzero((caddr_t)&info, sizeof(info));
1804 	IFNET_RLOCK_NOSLEEP();
1805 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1806 		if (w->w_arg && w->w_arg != ifp->if_index)
1807 			continue;
1808 		IF_ADDR_RLOCK(ifp);
1809 		ifa = ifp->if_addr;
1810 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1811 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1812 		info.rti_info[RTAX_IFP] = NULL;
1813 		if (w->w_req && w->w_tmem) {
1814 			if (w->w_op == NET_RT_IFLISTL)
1815 				error = sysctl_iflist_ifml(ifp, &info, w, len);
1816 			else
1817 				error = sysctl_iflist_ifm(ifp, &info, w, len);
1818 			if (error)
1819 				goto done;
1820 		}
1821 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1822 			if (af && af != ifa->ifa_addr->sa_family)
1823 				continue;
1824 			if (prison_if(w->w_req->td->td_ucred,
1825 			    ifa->ifa_addr) != 0)
1826 				continue;
1827 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1828 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1829 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1830 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1831 			if (w->w_req && w->w_tmem) {
1832 				if (w->w_op == NET_RT_IFLISTL)
1833 					error = sysctl_iflist_ifaml(ifa, &info,
1834 					    w, len);
1835 				else
1836 					error = sysctl_iflist_ifam(ifa, &info,
1837 					    w, len);
1838 				if (error)
1839 					goto done;
1840 			}
1841 		}
1842 		IF_ADDR_RUNLOCK(ifp);
1843 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1844 			info.rti_info[RTAX_BRD] = NULL;
1845 	}
1846 done:
1847 	if (ifp != NULL)
1848 		IF_ADDR_RUNLOCK(ifp);
1849 	IFNET_RUNLOCK_NOSLEEP();
1850 	return (error);
1851 }
1852 
1853 static int
1854 sysctl_ifmalist(int af, struct walkarg *w)
1855 {
1856 	struct ifnet *ifp;
1857 	struct ifmultiaddr *ifma;
1858 	struct	rt_addrinfo info;
1859 	int	len, error = 0;
1860 	struct ifaddr *ifa;
1861 
1862 	bzero((caddr_t)&info, sizeof(info));
1863 	IFNET_RLOCK_NOSLEEP();
1864 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1865 		if (w->w_arg && w->w_arg != ifp->if_index)
1866 			continue;
1867 		ifa = ifp->if_addr;
1868 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1869 		IF_ADDR_RLOCK(ifp);
1870 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1871 			if (af && af != ifma->ifma_addr->sa_family)
1872 				continue;
1873 			if (prison_if(w->w_req->td->td_ucred,
1874 			    ifma->ifma_addr) != 0)
1875 				continue;
1876 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1877 			info.rti_info[RTAX_GATEWAY] =
1878 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1879 			    ifma->ifma_lladdr : NULL;
1880 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1881 			if (w->w_req && w->w_tmem) {
1882 				struct ifma_msghdr *ifmam;
1883 
1884 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1885 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1886 				ifmam->ifmam_flags = 0;
1887 				ifmam->ifmam_addrs = info.rti_addrs;
1888 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1889 				if (error) {
1890 					IF_ADDR_RUNLOCK(ifp);
1891 					goto done;
1892 				}
1893 			}
1894 		}
1895 		IF_ADDR_RUNLOCK(ifp);
1896 	}
1897 done:
1898 	IFNET_RUNLOCK_NOSLEEP();
1899 	return (error);
1900 }
1901 
1902 static int
1903 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1904 {
1905 	int	*name = (int *)arg1;
1906 	u_int	namelen = arg2;
1907 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1908 	int	i, lim, error = EINVAL;
1909 	int	fib = 0;
1910 	u_char	af;
1911 	struct	walkarg w;
1912 
1913 	name ++;
1914 	namelen--;
1915 	if (req->newptr)
1916 		return (EPERM);
1917 	if (name[1] == NET_RT_DUMP) {
1918 		if (namelen == 3)
1919 			fib = req->td->td_proc->p_fibnum;
1920 		else if (namelen == 4)
1921 			fib = (name[3] == -1) ?
1922 			    req->td->td_proc->p_fibnum : name[3];
1923 		else
1924 			return ((namelen < 3) ? EISDIR : ENOTDIR);
1925 		if (fib < 0 || fib >= rt_numfibs)
1926 			return (EINVAL);
1927 	} else if (namelen != 3)
1928 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1929 	af = name[0];
1930 	if (af > AF_MAX)
1931 		return (EINVAL);
1932 	bzero(&w, sizeof(w));
1933 	w.w_op = name[1];
1934 	w.w_arg = name[2];
1935 	w.w_req = req;
1936 
1937 	error = sysctl_wire_old_buffer(req, 0);
1938 	if (error)
1939 		return (error);
1940 	switch (w.w_op) {
1941 
1942 	case NET_RT_DUMP:
1943 	case NET_RT_FLAGS:
1944 		if (af == 0) {			/* dump all tables */
1945 			i = 1;
1946 			lim = AF_MAX;
1947 		} else				/* dump only one table */
1948 			i = lim = af;
1949 
1950 		/*
1951 		 * take care of llinfo entries, the caller must
1952 		 * specify an AF
1953 		 */
1954 		if (w.w_op == NET_RT_FLAGS &&
1955 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1956 			if (af != 0)
1957 				error = lltable_sysctl_dumparp(af, w.w_req);
1958 			else
1959 				error = EINVAL;
1960 			break;
1961 		}
1962 		/*
1963 		 * take care of routing entries
1964 		 */
1965 		for (error = 0; error == 0 && i <= lim; i++) {
1966 			rnh = rt_tables_get_rnh(fib, i);
1967 			if (rnh != NULL) {
1968 				RADIX_NODE_HEAD_RLOCK(rnh);
1969 			    	error = rnh->rnh_walktree(rnh,
1970 				    sysctl_dumpentry, &w);
1971 				RADIX_NODE_HEAD_RUNLOCK(rnh);
1972 			} else if (af != 0)
1973 				error = EAFNOSUPPORT;
1974 		}
1975 		break;
1976 
1977 	case NET_RT_IFLIST:
1978 	case NET_RT_IFLISTL:
1979 		error = sysctl_iflist(af, &w);
1980 		break;
1981 
1982 	case NET_RT_IFMALIST:
1983 		error = sysctl_ifmalist(af, &w);
1984 		break;
1985 	}
1986 	if (w.w_tmem)
1987 		free(w.w_tmem, M_RTABLE);
1988 	return (error);
1989 }
1990 
1991 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1992 
1993 /*
1994  * Definitions of protocols supported in the ROUTE domain.
1995  */
1996 
1997 static struct domain routedomain;		/* or at least forward */
1998 
1999 static struct protosw routesw[] = {
2000 {
2001 	.pr_type =		SOCK_RAW,
2002 	.pr_domain =		&routedomain,
2003 	.pr_flags =		PR_ATOMIC|PR_ADDR,
2004 	.pr_output =		route_output,
2005 	.pr_ctlinput =		raw_ctlinput,
2006 	.pr_init =		raw_init,
2007 	.pr_usrreqs =		&route_usrreqs
2008 }
2009 };
2010 
2011 static struct domain routedomain = {
2012 	.dom_family =		PF_ROUTE,
2013 	.dom_name =		 "route",
2014 	.dom_protosw =		routesw,
2015 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
2016 };
2017 
2018 VNET_DOMAIN_SET(route);
2019