xref: /freebsd/sys/net/rtsock.c (revision 66e576525d35c68fcb86f142ebaa5a448555c0c7)
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 #ifdef INET6
67 #include <netinet6/scope6_var.h>
68 #endif
69 
70 #if defined(INET) || defined(INET6)
71 #ifdef SCTP
72 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
73 #endif /* SCTP */
74 #endif
75 
76 #ifdef COMPAT_FREEBSD32
77 #include <sys/mount.h>
78 #include <compat/freebsd32/freebsd32.h>
79 
80 struct if_data32 {
81 	uint8_t	ifi_type;
82 	uint8_t	ifi_physical;
83 	uint8_t	ifi_addrlen;
84 	uint8_t	ifi_hdrlen;
85 	uint8_t	ifi_link_state;
86 	uint8_t	ifi_spare_char1;
87 	uint8_t	ifi_spare_char2;
88 	uint8_t	ifi_datalen;
89 	uint32_t ifi_mtu;
90 	uint32_t ifi_metric;
91 	uint32_t ifi_baudrate;
92 	uint32_t ifi_ipackets;
93 	uint32_t ifi_ierrors;
94 	uint32_t ifi_opackets;
95 	uint32_t ifi_oerrors;
96 	uint32_t ifi_collisions;
97 	uint32_t ifi_ibytes;
98 	uint32_t ifi_obytes;
99 	uint32_t ifi_imcasts;
100 	uint32_t ifi_omcasts;
101 	uint32_t ifi_iqdrops;
102 	uint32_t ifi_noproto;
103 	uint32_t ifi_hwassist;
104 	int32_t	ifi_epoch;
105 	struct	timeval32 ifi_lastchange;
106 };
107 
108 struct if_msghdr32 {
109 	uint16_t ifm_msglen;
110 	uint8_t	ifm_version;
111 	uint8_t	ifm_type;
112 	int32_t	ifm_addrs;
113 	int32_t	ifm_flags;
114 	uint16_t ifm_index;
115 	struct	if_data32 ifm_data;
116 };
117 #endif
118 
119 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
120 
121 /* NB: these are not modified */
122 static struct	sockaddr route_src = { 2, PF_ROUTE, };
123 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
124 
125 static struct {
126 	int	ip_count;	/* attached w/ AF_INET */
127 	int	ip6_count;	/* attached w/ AF_INET6 */
128 	int	ipx_count;	/* attached w/ AF_IPX */
129 	int	any_count;	/* total attached */
130 } route_cb;
131 
132 struct mtx rtsock_mtx;
133 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
134 
135 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
136 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
137 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
138 
139 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
140 
141 struct walkarg {
142 	int	w_tmemsize;
143 	int	w_op, w_arg;
144 	caddr_t	w_tmem;
145 	struct sysctl_req *w_req;
146 };
147 
148 static void	rts_input(struct mbuf *m);
149 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
150 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
151 			caddr_t cp, struct walkarg *w);
152 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
153 			struct rt_addrinfo *rtinfo);
154 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
155 static int	sysctl_iflist(int af, struct walkarg *w);
156 static int	sysctl_ifmalist(int af, struct walkarg *w);
157 static int	route_output(struct mbuf *m, struct socket *so);
158 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
159 			struct rt_metrics_lite *out);
160 static void	rt_getmetrics(const struct rt_metrics_lite *in,
161 			struct rt_metrics *out);
162 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
163 
164 static struct netisr_handler rtsock_nh = {
165 	.nh_name = "rtsock",
166 	.nh_handler = rts_input,
167 	.nh_proto = NETISR_ROUTE,
168 	.nh_policy = NETISR_POLICY_SOURCE,
169 };
170 
171 static int
172 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
173 {
174 	int error, qlimit;
175 
176 	netisr_getqlimit(&rtsock_nh, &qlimit);
177 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
178         if (error || !req->newptr)
179                 return (error);
180 	if (qlimit < 1)
181 		return (EINVAL);
182 	return (netisr_setqlimit(&rtsock_nh, qlimit));
183 }
184 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
185     0, 0, sysctl_route_netisr_maxqlen, "I",
186     "maximum routing socket dispatch queue length");
187 
188 static void
189 rts_init(void)
190 {
191 	int tmp;
192 
193 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
194 		rtsock_nh.nh_qlimit = tmp;
195 	netisr_register(&rtsock_nh);
196 }
197 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
198 
199 static void
200 rts_input(struct mbuf *m)
201 {
202 	struct sockproto route_proto;
203 	unsigned short *family;
204 	struct m_tag *tag;
205 
206 	route_proto.sp_family = PF_ROUTE;
207 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
208 	if (tag != NULL) {
209 		family = (unsigned short *)(tag + 1);
210 		route_proto.sp_protocol = *family;
211 		m_tag_delete(m, tag);
212 	} else
213 		route_proto.sp_protocol = 0;
214 
215 	raw_input(m, &route_proto, &route_src);
216 }
217 
218 /*
219  * It really doesn't make any sense at all for this code to share much
220  * with raw_usrreq.c, since its functionality is so restricted.  XXX
221  */
222 static void
223 rts_abort(struct socket *so)
224 {
225 
226 	raw_usrreqs.pru_abort(so);
227 }
228 
229 static void
230 rts_close(struct socket *so)
231 {
232 
233 	raw_usrreqs.pru_close(so);
234 }
235 
236 /* pru_accept is EOPNOTSUPP */
237 
238 static int
239 rts_attach(struct socket *so, int proto, struct thread *td)
240 {
241 	struct rawcb *rp;
242 	int s, error;
243 
244 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
245 
246 	/* XXX */
247 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
248 	if (rp == NULL)
249 		return ENOBUFS;
250 
251 	/*
252 	 * The splnet() is necessary to block protocols from sending
253 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
254 	 * this PCB is extant but incompletely initialized.
255 	 * Probably we should try to do more of this work beforehand and
256 	 * eliminate the spl.
257 	 */
258 	s = splnet();
259 	so->so_pcb = (caddr_t)rp;
260 	so->so_fibnum = td->td_proc->p_fibnum;
261 	error = raw_attach(so, proto);
262 	rp = sotorawcb(so);
263 	if (error) {
264 		splx(s);
265 		so->so_pcb = NULL;
266 		free(rp, M_PCB);
267 		return error;
268 	}
269 	RTSOCK_LOCK();
270 	switch(rp->rcb_proto.sp_protocol) {
271 	case AF_INET:
272 		route_cb.ip_count++;
273 		break;
274 	case AF_INET6:
275 		route_cb.ip6_count++;
276 		break;
277 	case AF_IPX:
278 		route_cb.ipx_count++;
279 		break;
280 	}
281 	route_cb.any_count++;
282 	RTSOCK_UNLOCK();
283 	soisconnected(so);
284 	so->so_options |= SO_USELOOPBACK;
285 	splx(s);
286 	return 0;
287 }
288 
289 static int
290 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
291 {
292 
293 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
294 }
295 
296 static int
297 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
298 {
299 
300 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
301 }
302 
303 /* pru_connect2 is EOPNOTSUPP */
304 /* pru_control is EOPNOTSUPP */
305 
306 static void
307 rts_detach(struct socket *so)
308 {
309 	struct rawcb *rp = sotorawcb(so);
310 
311 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
312 
313 	RTSOCK_LOCK();
314 	switch(rp->rcb_proto.sp_protocol) {
315 	case AF_INET:
316 		route_cb.ip_count--;
317 		break;
318 	case AF_INET6:
319 		route_cb.ip6_count--;
320 		break;
321 	case AF_IPX:
322 		route_cb.ipx_count--;
323 		break;
324 	}
325 	route_cb.any_count--;
326 	RTSOCK_UNLOCK();
327 	raw_usrreqs.pru_detach(so);
328 }
329 
330 static int
331 rts_disconnect(struct socket *so)
332 {
333 
334 	return (raw_usrreqs.pru_disconnect(so));
335 }
336 
337 /* pru_listen is EOPNOTSUPP */
338 
339 static int
340 rts_peeraddr(struct socket *so, struct sockaddr **nam)
341 {
342 
343 	return (raw_usrreqs.pru_peeraddr(so, nam));
344 }
345 
346 /* pru_rcvd is EOPNOTSUPP */
347 /* pru_rcvoob is EOPNOTSUPP */
348 
349 static int
350 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
351 	 struct mbuf *control, struct thread *td)
352 {
353 
354 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
355 }
356 
357 /* pru_sense is null */
358 
359 static int
360 rts_shutdown(struct socket *so)
361 {
362 
363 	return (raw_usrreqs.pru_shutdown(so));
364 }
365 
366 static int
367 rts_sockaddr(struct socket *so, struct sockaddr **nam)
368 {
369 
370 	return (raw_usrreqs.pru_sockaddr(so, nam));
371 }
372 
373 static struct pr_usrreqs route_usrreqs = {
374 	.pru_abort =		rts_abort,
375 	.pru_attach =		rts_attach,
376 	.pru_bind =		rts_bind,
377 	.pru_connect =		rts_connect,
378 	.pru_detach =		rts_detach,
379 	.pru_disconnect =	rts_disconnect,
380 	.pru_peeraddr =		rts_peeraddr,
381 	.pru_send =		rts_send,
382 	.pru_shutdown =		rts_shutdown,
383 	.pru_sockaddr =		rts_sockaddr,
384 	.pru_close =		rts_close,
385 };
386 
387 #ifndef _SOCKADDR_UNION_DEFINED
388 #define	_SOCKADDR_UNION_DEFINED
389 /*
390  * The union of all possible address formats we handle.
391  */
392 union sockaddr_union {
393 	struct sockaddr		sa;
394 	struct sockaddr_in	sin;
395 	struct sockaddr_in6	sin6;
396 };
397 #endif /* _SOCKADDR_UNION_DEFINED */
398 
399 static int
400 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
401     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
402 {
403 
404 	/* First, see if the returned address is part of the jail. */
405 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
406 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
407 		return (0);
408 	}
409 
410 	switch (info->rti_info[RTAX_DST]->sa_family) {
411 #ifdef INET
412 	case AF_INET:
413 	{
414 		struct in_addr ia;
415 		struct ifaddr *ifa;
416 		int found;
417 
418 		found = 0;
419 		/*
420 		 * Try to find an address on the given outgoing interface
421 		 * that belongs to the jail.
422 		 */
423 		IF_ADDR_LOCK(ifp);
424 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
425 			struct sockaddr *sa;
426 			sa = ifa->ifa_addr;
427 			if (sa->sa_family != AF_INET)
428 				continue;
429 			ia = ((struct sockaddr_in *)sa)->sin_addr;
430 			if (prison_check_ip4(cred, &ia) == 0) {
431 				found = 1;
432 				break;
433 			}
434 		}
435 		IF_ADDR_UNLOCK(ifp);
436 		if (!found) {
437 			/*
438 			 * As a last resort return the 'default' jail address.
439 			 */
440 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
441 			    sin_addr;
442 			if (prison_get_ip4(cred, &ia) != 0)
443 				return (ESRCH);
444 		}
445 		bzero(&saun->sin, sizeof(struct sockaddr_in));
446 		saun->sin.sin_len = sizeof(struct sockaddr_in);
447 		saun->sin.sin_family = AF_INET;
448 		saun->sin.sin_addr.s_addr = ia.s_addr;
449 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
450 		break;
451 	}
452 #endif
453 #ifdef INET6
454 	case AF_INET6:
455 	{
456 		struct in6_addr ia6;
457 		struct ifaddr *ifa;
458 		int found;
459 
460 		found = 0;
461 		/*
462 		 * Try to find an address on the given outgoing interface
463 		 * that belongs to the jail.
464 		 */
465 		IF_ADDR_LOCK(ifp);
466 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
467 			struct sockaddr *sa;
468 			sa = ifa->ifa_addr;
469 			if (sa->sa_family != AF_INET6)
470 				continue;
471 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
472 			    &ia6, sizeof(struct in6_addr));
473 			if (prison_check_ip6(cred, &ia6) == 0) {
474 				found = 1;
475 				break;
476 			}
477 		}
478 		IF_ADDR_UNLOCK(ifp);
479 		if (!found) {
480 			/*
481 			 * As a last resort return the 'default' jail address.
482 			 */
483 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
484 			    sin6_addr;
485 			if (prison_get_ip6(cred, &ia6) != 0)
486 				return (ESRCH);
487 		}
488 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
489 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
490 		saun->sin6.sin6_family = AF_INET6;
491 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
492 		if (sa6_recoverscope(&saun->sin6) != 0)
493 			return (ESRCH);
494 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
495 		break;
496 	}
497 #endif
498 	default:
499 		return (ESRCH);
500 	}
501 	return (0);
502 }
503 
504 /*ARGSUSED*/
505 static int
506 route_output(struct mbuf *m, struct socket *so)
507 {
508 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
509 	struct rt_msghdr *rtm = NULL;
510 	struct rtentry *rt = NULL;
511 	struct radix_node_head *rnh;
512 	struct rt_addrinfo info;
513 	int len, error = 0;
514 	struct ifnet *ifp = NULL;
515 	union sockaddr_union saun;
516 
517 #define senderr(e) { error = e; goto flush;}
518 	if (m == NULL || ((m->m_len < sizeof(long)) &&
519 		       (m = m_pullup(m, sizeof(long))) == NULL))
520 		return (ENOBUFS);
521 	if ((m->m_flags & M_PKTHDR) == 0)
522 		panic("route_output");
523 	len = m->m_pkthdr.len;
524 	if (len < sizeof(*rtm) ||
525 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
526 		info.rti_info[RTAX_DST] = NULL;
527 		senderr(EINVAL);
528 	}
529 	R_Malloc(rtm, struct rt_msghdr *, len);
530 	if (rtm == NULL) {
531 		info.rti_info[RTAX_DST] = NULL;
532 		senderr(ENOBUFS);
533 	}
534 	m_copydata(m, 0, len, (caddr_t)rtm);
535 	if (rtm->rtm_version != RTM_VERSION) {
536 		info.rti_info[RTAX_DST] = NULL;
537 		senderr(EPROTONOSUPPORT);
538 	}
539 	rtm->rtm_pid = curproc->p_pid;
540 	bzero(&info, sizeof(info));
541 	info.rti_addrs = rtm->rtm_addrs;
542 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
543 		info.rti_info[RTAX_DST] = NULL;
544 		senderr(EINVAL);
545 	}
546 	info.rti_flags = rtm->rtm_flags;
547 	if (info.rti_info[RTAX_DST] == NULL ||
548 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
549 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
550 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
551 		senderr(EINVAL);
552 	/*
553 	 * Verify that the caller has the appropriate privilege; RTM_GET
554 	 * is the only operation the non-superuser is allowed.
555 	 */
556 	if (rtm->rtm_type != RTM_GET) {
557 		error = priv_check(curthread, PRIV_NET_ROUTE);
558 		if (error)
559 			senderr(error);
560 	}
561 
562 	/*
563 	 * The given gateway address may be an interface address.
564 	 * For example, issuing a "route change" command on a route
565 	 * entry that was created from a tunnel, and the gateway
566 	 * address given is the local end point. In this case the
567 	 * RTF_GATEWAY flag must be cleared or the destination will
568 	 * not be reachable even though there is no error message.
569 	 */
570 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
571 	    info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
572 		struct route gw_ro;
573 
574 		bzero(&gw_ro, sizeof(gw_ro));
575 		gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
576 		rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
577 		/*
578 		 * A host route through the loopback interface is
579 		 * installed for each interface adddress. In pre 8.0
580 		 * releases the interface address of a PPP link type
581 		 * is not reachable locally. This behavior is fixed as
582 		 * part of the new L2/L3 redesign and rewrite work. The
583 		 * signature of this interface address route is the
584 		 * AF_LINK sa_family type of the rt_gateway, and the
585 		 * rt_ifp has the IFF_LOOPBACK flag set.
586 		 */
587 		if (gw_ro.ro_rt != NULL &&
588 		    gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
589 		    gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)
590 			info.rti_flags &= ~RTF_GATEWAY;
591 		if (gw_ro.ro_rt != NULL)
592 			RTFREE(gw_ro.ro_rt);
593 	}
594 
595 	switch (rtm->rtm_type) {
596 		struct rtentry *saved_nrt;
597 
598 	case RTM_ADD:
599 		if (info.rti_info[RTAX_GATEWAY] == NULL)
600 			senderr(EINVAL);
601 		saved_nrt = NULL;
602 
603 		/* support for new ARP code */
604 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
605 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
606 			error = lla_rt_output(rtm, &info);
607 			break;
608 		}
609 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
610 		    so->so_fibnum);
611 		if (error == 0 && saved_nrt) {
612 			RT_LOCK(saved_nrt);
613 			rt_setmetrics(rtm->rtm_inits,
614 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
615 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
616 			RT_REMREF(saved_nrt);
617 			RT_UNLOCK(saved_nrt);
618 		}
619 		break;
620 
621 	case RTM_DELETE:
622 		saved_nrt = NULL;
623 		/* support for new ARP code */
624 		if (info.rti_info[RTAX_GATEWAY] &&
625 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
626 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
627 			error = lla_rt_output(rtm, &info);
628 			break;
629 		}
630 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
631 		    so->so_fibnum);
632 		if (error == 0) {
633 			RT_LOCK(saved_nrt);
634 			rt = saved_nrt;
635 			goto report;
636 		}
637 		break;
638 
639 	case RTM_GET:
640 	case RTM_CHANGE:
641 	case RTM_LOCK:
642 		rnh = rt_tables_get_rnh(so->so_fibnum,
643 		    info.rti_info[RTAX_DST]->sa_family);
644 		if (rnh == NULL)
645 			senderr(EAFNOSUPPORT);
646 		RADIX_NODE_HEAD_RLOCK(rnh);
647 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
648 			info.rti_info[RTAX_NETMASK], rnh);
649 		if (rt == NULL) {	/* XXX looks bogus */
650 			RADIX_NODE_HEAD_RUNLOCK(rnh);
651 			senderr(ESRCH);
652 		}
653 #ifdef RADIX_MPATH
654 		/*
655 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
656 		 * we require users to specify a matching RTAX_GATEWAY.
657 		 *
658 		 * for RTM_GET, gate is optional even with multipath.
659 		 * if gate == NULL the first match is returned.
660 		 * (no need to call rt_mpath_matchgate if gate == NULL)
661 		 */
662 		if (rn_mpath_capable(rnh) &&
663 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
664 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
665 			if (!rt) {
666 				RADIX_NODE_HEAD_RUNLOCK(rnh);
667 				senderr(ESRCH);
668 			}
669 		}
670 #endif
671 		/*
672 		 * If performing proxied L2 entry insertion, and
673 		 * the actual PPP host entry is found, perform
674 		 * another search to retrieve the prefix route of
675 		 * the local end point of the PPP link.
676 		 */
677 		if (rtm->rtm_flags & RTF_ANNOUNCE) {
678 			struct sockaddr laddr;
679 
680 			if (rt->rt_ifp != NULL &&
681 			    rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
682 				struct ifaddr *ifa;
683 
684 				ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
685 				if (ifa != NULL)
686 					rt_maskedcopy(ifa->ifa_addr,
687 						      &laddr,
688 						      ifa->ifa_netmask);
689 			} else
690 				rt_maskedcopy(rt->rt_ifa->ifa_addr,
691 					      &laddr,
692 					      rt->rt_ifa->ifa_netmask);
693 			/*
694 			 * refactor rt and no lock operation necessary
695 			 */
696 			rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
697 			if (rt == NULL) {
698 				RADIX_NODE_HEAD_RUNLOCK(rnh);
699 				senderr(ESRCH);
700 			}
701 		}
702 		RT_LOCK(rt);
703 		RT_ADDREF(rt);
704 		RADIX_NODE_HEAD_RUNLOCK(rnh);
705 
706 		/*
707 		 * Fix for PR: 82974
708 		 *
709 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
710 		 * returns a perfect match in case a netmask is
711 		 * specified.  For host routes only a longest prefix
712 		 * match is returned so it is necessary to compare the
713 		 * existence of the netmask.  If both have a netmask
714 		 * rnh_lookup() did a perfect match and if none of them
715 		 * have a netmask both are host routes which is also a
716 		 * perfect match.
717 		 */
718 
719 		if (rtm->rtm_type != RTM_GET &&
720 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
721 			RT_UNLOCK(rt);
722 			senderr(ESRCH);
723 		}
724 
725 		switch(rtm->rtm_type) {
726 
727 		case RTM_GET:
728 		report:
729 			RT_LOCK_ASSERT(rt);
730 			if ((rt->rt_flags & RTF_HOST) == 0
731 			    ? jailed_without_vnet(curthread->td_ucred)
732 			    : prison_if(curthread->td_ucred,
733 			    rt_key(rt)) != 0) {
734 				RT_UNLOCK(rt);
735 				senderr(ESRCH);
736 			}
737 			info.rti_info[RTAX_DST] = rt_key(rt);
738 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
739 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
740 			info.rti_info[RTAX_GENMASK] = 0;
741 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
742 				ifp = rt->rt_ifp;
743 				if (ifp) {
744 					info.rti_info[RTAX_IFP] =
745 					    ifp->if_addr->ifa_addr;
746 					error = rtm_get_jailed(&info, ifp, rt,
747 					    &saun, curthread->td_ucred);
748 					if (error != 0) {
749 						RT_UNLOCK(rt);
750 						senderr(error);
751 					}
752 					if (ifp->if_flags & IFF_POINTOPOINT)
753 						info.rti_info[RTAX_BRD] =
754 						    rt->rt_ifa->ifa_dstaddr;
755 					rtm->rtm_index = ifp->if_index;
756 				} else {
757 					info.rti_info[RTAX_IFP] = NULL;
758 					info.rti_info[RTAX_IFA] = NULL;
759 				}
760 			} else if ((ifp = rt->rt_ifp) != NULL) {
761 				rtm->rtm_index = ifp->if_index;
762 			}
763 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
764 			if (len > rtm->rtm_msglen) {
765 				struct rt_msghdr *new_rtm;
766 				R_Malloc(new_rtm, struct rt_msghdr *, len);
767 				if (new_rtm == NULL) {
768 					RT_UNLOCK(rt);
769 					senderr(ENOBUFS);
770 				}
771 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
772 				Free(rtm); rtm = new_rtm;
773 			}
774 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
775 			rtm->rtm_flags = rt->rt_flags;
776 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
777 			rtm->rtm_addrs = info.rti_addrs;
778 			break;
779 
780 		case RTM_CHANGE:
781 			/*
782 			 * New gateway could require new ifaddr, ifp;
783 			 * flags may also be different; ifp may be specified
784 			 * by ll sockaddr when protocol address is ambiguous
785 			 */
786 			if (((rt->rt_flags & RTF_GATEWAY) &&
787 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
788 			    info.rti_info[RTAX_IFP] != NULL ||
789 			    (info.rti_info[RTAX_IFA] != NULL &&
790 			     !sa_equal(info.rti_info[RTAX_IFA],
791 				       rt->rt_ifa->ifa_addr))) {
792 				RT_UNLOCK(rt);
793 				RADIX_NODE_HEAD_LOCK(rnh);
794 				error = rt_getifa_fib(&info, rt->rt_fibnum);
795 				/*
796 				 * XXXRW: Really we should release this
797 				 * reference later, but this maintains
798 				 * historical behavior.
799 				 */
800 				if (info.rti_ifa != NULL)
801 					ifa_free(info.rti_ifa);
802 				RADIX_NODE_HEAD_UNLOCK(rnh);
803 				if (error != 0)
804 					senderr(error);
805 				RT_LOCK(rt);
806 			}
807 			if (info.rti_ifa != NULL &&
808 			    info.rti_ifa != rt->rt_ifa &&
809 			    rt->rt_ifa != NULL &&
810 			    rt->rt_ifa->ifa_rtrequest != NULL) {
811 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
812 				    &info);
813 				ifa_free(rt->rt_ifa);
814 			}
815 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
816 				RT_UNLOCK(rt);
817 				RADIX_NODE_HEAD_LOCK(rnh);
818 				RT_LOCK(rt);
819 
820 				error = rt_setgate(rt, rt_key(rt),
821 				    info.rti_info[RTAX_GATEWAY]);
822 				RADIX_NODE_HEAD_UNLOCK(rnh);
823 				if (error != 0) {
824 					RT_UNLOCK(rt);
825 					senderr(error);
826 				}
827 				rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
828 			}
829 			if (info.rti_ifa != NULL &&
830 			    info.rti_ifa != rt->rt_ifa) {
831 				ifa_ref(info.rti_ifa);
832 				rt->rt_ifa = info.rti_ifa;
833 				rt->rt_ifp = info.rti_ifp;
834 			}
835 			/* Allow some flags to be toggled on change. */
836 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
837 				    (rtm->rtm_flags & RTF_FMASK);
838 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
839 					&rt->rt_rmx);
840 			rtm->rtm_index = rt->rt_ifp->if_index;
841 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
842 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
843 			/* FALLTHROUGH */
844 		case RTM_LOCK:
845 			/* We don't support locks anymore */
846 			break;
847 		}
848 		RT_UNLOCK(rt);
849 		break;
850 
851 	default:
852 		senderr(EOPNOTSUPP);
853 	}
854 
855 flush:
856 	if (rtm) {
857 		if (error)
858 			rtm->rtm_errno = error;
859 		else
860 			rtm->rtm_flags |= RTF_DONE;
861 	}
862 	if (rt)		/* XXX can this be true? */
863 		RTFREE(rt);
864     {
865 	struct rawcb *rp = NULL;
866 	/*
867 	 * Check to see if we don't want our own messages.
868 	 */
869 	if ((so->so_options & SO_USELOOPBACK) == 0) {
870 		if (route_cb.any_count <= 1) {
871 			if (rtm)
872 				Free(rtm);
873 			m_freem(m);
874 			return (error);
875 		}
876 		/* There is another listener, so construct message */
877 		rp = sotorawcb(so);
878 	}
879 	if (rtm) {
880 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
881 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
882 			m_freem(m);
883 			m = NULL;
884 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
885 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
886 		Free(rtm);
887 	}
888 	if (m) {
889 		if (rp) {
890 			/*
891 			 * XXX insure we don't get a copy by
892 			 * invalidating our protocol
893 			 */
894 			unsigned short family = rp->rcb_proto.sp_family;
895 			rp->rcb_proto.sp_family = 0;
896 			rt_dispatch(m, info.rti_info[RTAX_DST]);
897 			rp->rcb_proto.sp_family = family;
898 		} else
899 			rt_dispatch(m, info.rti_info[RTAX_DST]);
900 	}
901     }
902 	return (error);
903 #undef	sa_equal
904 }
905 
906 static void
907 rt_setmetrics(u_long which, const struct rt_metrics *in,
908 	struct rt_metrics_lite *out)
909 {
910 #define metric(f, e) if (which & (f)) out->e = in->e;
911 	/*
912 	 * Only these are stored in the routing entry since introduction
913 	 * of tcp hostcache. The rest is ignored.
914 	 */
915 	metric(RTV_MTU, rmx_mtu);
916 	metric(RTV_WEIGHT, rmx_weight);
917 	/* Userland -> kernel timebase conversion. */
918 	if (which & RTV_EXPIRE)
919 		out->rmx_expire = in->rmx_expire ?
920 		    in->rmx_expire - time_second + time_uptime : 0;
921 #undef metric
922 }
923 
924 static void
925 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
926 {
927 #define metric(e) out->e = in->e;
928 	bzero(out, sizeof(*out));
929 	metric(rmx_mtu);
930 	metric(rmx_weight);
931 	/* Kernel -> userland timebase conversion. */
932 	out->rmx_expire = in->rmx_expire ?
933 	    in->rmx_expire - time_uptime + time_second : 0;
934 #undef metric
935 }
936 
937 /*
938  * Extract the addresses of the passed sockaddrs.
939  * Do a little sanity checking so as to avoid bad memory references.
940  * This data is derived straight from userland.
941  */
942 static int
943 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
944 {
945 	struct sockaddr *sa;
946 	int i;
947 
948 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
949 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
950 			continue;
951 		sa = (struct sockaddr *)cp;
952 		/*
953 		 * It won't fit.
954 		 */
955 		if (cp + sa->sa_len > cplim)
956 			return (EINVAL);
957 		/*
958 		 * there are no more.. quit now
959 		 * If there are more bits, they are in error.
960 		 * I've seen this. route(1) can evidently generate these.
961 		 * This causes kernel to core dump.
962 		 * for compatibility, If we see this, point to a safe address.
963 		 */
964 		if (sa->sa_len == 0) {
965 			rtinfo->rti_info[i] = &sa_zero;
966 			return (0); /* should be EINVAL but for compat */
967 		}
968 		/* accept it */
969 		rtinfo->rti_info[i] = sa;
970 		cp += SA_SIZE(sa);
971 	}
972 	return (0);
973 }
974 
975 static struct mbuf *
976 rt_msg1(int type, struct rt_addrinfo *rtinfo)
977 {
978 	struct rt_msghdr *rtm;
979 	struct mbuf *m;
980 	int i;
981 	struct sockaddr *sa;
982 	int len, dlen;
983 
984 	switch (type) {
985 
986 	case RTM_DELADDR:
987 	case RTM_NEWADDR:
988 		len = sizeof(struct ifa_msghdr);
989 		break;
990 
991 	case RTM_DELMADDR:
992 	case RTM_NEWMADDR:
993 		len = sizeof(struct ifma_msghdr);
994 		break;
995 
996 	case RTM_IFINFO:
997 		len = sizeof(struct if_msghdr);
998 		break;
999 
1000 	case RTM_IFANNOUNCE:
1001 	case RTM_IEEE80211:
1002 		len = sizeof(struct if_announcemsghdr);
1003 		break;
1004 
1005 	default:
1006 		len = sizeof(struct rt_msghdr);
1007 	}
1008 	if (len > MCLBYTES)
1009 		panic("rt_msg1");
1010 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1011 	if (m && len > MHLEN) {
1012 		MCLGET(m, M_DONTWAIT);
1013 		if ((m->m_flags & M_EXT) == 0) {
1014 			m_free(m);
1015 			m = NULL;
1016 		}
1017 	}
1018 	if (m == NULL)
1019 		return (m);
1020 	m->m_pkthdr.len = m->m_len = len;
1021 	m->m_pkthdr.rcvif = NULL;
1022 	rtm = mtod(m, struct rt_msghdr *);
1023 	bzero((caddr_t)rtm, len);
1024 	for (i = 0; i < RTAX_MAX; i++) {
1025 		if ((sa = rtinfo->rti_info[i]) == NULL)
1026 			continue;
1027 		rtinfo->rti_addrs |= (1 << i);
1028 		dlen = SA_SIZE(sa);
1029 		m_copyback(m, len, dlen, (caddr_t)sa);
1030 		len += dlen;
1031 	}
1032 	if (m->m_pkthdr.len != len) {
1033 		m_freem(m);
1034 		return (NULL);
1035 	}
1036 	rtm->rtm_msglen = len;
1037 	rtm->rtm_version = RTM_VERSION;
1038 	rtm->rtm_type = type;
1039 	return (m);
1040 }
1041 
1042 static int
1043 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1044 {
1045 	int i;
1046 	int len, dlen, second_time = 0;
1047 	caddr_t cp0;
1048 
1049 	rtinfo->rti_addrs = 0;
1050 again:
1051 	switch (type) {
1052 
1053 	case RTM_DELADDR:
1054 	case RTM_NEWADDR:
1055 		len = sizeof(struct ifa_msghdr);
1056 		break;
1057 
1058 	case RTM_IFINFO:
1059 #ifdef COMPAT_FREEBSD32
1060 		if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1061 			len = sizeof(struct if_msghdr32);
1062 			break;
1063 		}
1064 #endif
1065 		len = sizeof(struct if_msghdr);
1066 		break;
1067 
1068 	case RTM_NEWMADDR:
1069 		len = sizeof(struct ifma_msghdr);
1070 		break;
1071 
1072 	default:
1073 		len = sizeof(struct rt_msghdr);
1074 	}
1075 	cp0 = cp;
1076 	if (cp0)
1077 		cp += len;
1078 	for (i = 0; i < RTAX_MAX; i++) {
1079 		struct sockaddr *sa;
1080 
1081 		if ((sa = rtinfo->rti_info[i]) == NULL)
1082 			continue;
1083 		rtinfo->rti_addrs |= (1 << i);
1084 		dlen = SA_SIZE(sa);
1085 		if (cp) {
1086 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
1087 			cp += dlen;
1088 		}
1089 		len += dlen;
1090 	}
1091 	len = ALIGN(len);
1092 	if (cp == NULL && w != NULL && !second_time) {
1093 		struct walkarg *rw = w;
1094 
1095 		if (rw->w_req) {
1096 			if (rw->w_tmemsize < len) {
1097 				if (rw->w_tmem)
1098 					free(rw->w_tmem, M_RTABLE);
1099 				rw->w_tmem = (caddr_t)
1100 					malloc(len, M_RTABLE, M_NOWAIT);
1101 				if (rw->w_tmem)
1102 					rw->w_tmemsize = len;
1103 			}
1104 			if (rw->w_tmem) {
1105 				cp = rw->w_tmem;
1106 				second_time = 1;
1107 				goto again;
1108 			}
1109 		}
1110 	}
1111 	if (cp) {
1112 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1113 
1114 		rtm->rtm_version = RTM_VERSION;
1115 		rtm->rtm_type = type;
1116 		rtm->rtm_msglen = len;
1117 	}
1118 	return (len);
1119 }
1120 
1121 /*
1122  * This routine is called to generate a message from the routing
1123  * socket indicating that a redirect has occured, a routing lookup
1124  * has failed, or that a protocol has detected timeouts to a particular
1125  * destination.
1126  */
1127 void
1128 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1129 {
1130 	struct rt_msghdr *rtm;
1131 	struct mbuf *m;
1132 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1133 
1134 	if (route_cb.any_count == 0)
1135 		return;
1136 	m = rt_msg1(type, rtinfo);
1137 	if (m == NULL)
1138 		return;
1139 	rtm = mtod(m, struct rt_msghdr *);
1140 	rtm->rtm_flags = RTF_DONE | flags;
1141 	rtm->rtm_errno = error;
1142 	rtm->rtm_addrs = rtinfo->rti_addrs;
1143 	rt_dispatch(m, sa);
1144 }
1145 
1146 /*
1147  * This routine is called to generate a message from the routing
1148  * socket indicating that the status of a network interface has changed.
1149  */
1150 void
1151 rt_ifmsg(struct ifnet *ifp)
1152 {
1153 	struct if_msghdr *ifm;
1154 	struct mbuf *m;
1155 	struct rt_addrinfo info;
1156 
1157 	if (route_cb.any_count == 0)
1158 		return;
1159 	bzero((caddr_t)&info, sizeof(info));
1160 	m = rt_msg1(RTM_IFINFO, &info);
1161 	if (m == NULL)
1162 		return;
1163 	ifm = mtod(m, struct if_msghdr *);
1164 	ifm->ifm_index = ifp->if_index;
1165 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1166 	ifm->ifm_data = ifp->if_data;
1167 	ifm->ifm_addrs = 0;
1168 	rt_dispatch(m, NULL);
1169 }
1170 
1171 /*
1172  * This is called to generate messages from the routing socket
1173  * indicating a network interface has had addresses associated with it.
1174  * if we ever reverse the logic and replace messages TO the routing
1175  * socket indicate a request to configure interfaces, then it will
1176  * be unnecessary as the routing socket will automatically generate
1177  * copies of it.
1178  */
1179 void
1180 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1181 {
1182 	struct rt_addrinfo info;
1183 	struct sockaddr *sa = NULL;
1184 	int pass;
1185 	struct mbuf *m = NULL;
1186 	struct ifnet *ifp = ifa->ifa_ifp;
1187 
1188 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1189 		("unexpected cmd %u", cmd));
1190 #if defined(INET) || defined(INET6)
1191 #ifdef SCTP
1192 	/*
1193 	 * notify the SCTP stack
1194 	 * this will only get called when an address is added/deleted
1195 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1196 	 */
1197 	sctp_addr_change(ifa, cmd);
1198 #endif /* SCTP */
1199 #endif
1200 	if (route_cb.any_count == 0)
1201 		return;
1202 	for (pass = 1; pass < 3; pass++) {
1203 		bzero((caddr_t)&info, sizeof(info));
1204 		if ((cmd == RTM_ADD && pass == 1) ||
1205 		    (cmd == RTM_DELETE && pass == 2)) {
1206 			struct ifa_msghdr *ifam;
1207 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1208 
1209 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1210 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1211 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1212 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1213 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1214 				continue;
1215 			ifam = mtod(m, struct ifa_msghdr *);
1216 			ifam->ifam_index = ifp->if_index;
1217 			ifam->ifam_metric = ifa->ifa_metric;
1218 			ifam->ifam_flags = ifa->ifa_flags;
1219 			ifam->ifam_addrs = info.rti_addrs;
1220 		}
1221 		if ((cmd == RTM_ADD && pass == 2) ||
1222 		    (cmd == RTM_DELETE && pass == 1)) {
1223 			struct rt_msghdr *rtm;
1224 
1225 			if (rt == NULL)
1226 				continue;
1227 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1228 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1229 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1230 			if ((m = rt_msg1(cmd, &info)) == NULL)
1231 				continue;
1232 			rtm = mtod(m, struct rt_msghdr *);
1233 			rtm->rtm_index = ifp->if_index;
1234 			rtm->rtm_flags |= rt->rt_flags;
1235 			rtm->rtm_errno = error;
1236 			rtm->rtm_addrs = info.rti_addrs;
1237 		}
1238 		rt_dispatch(m, sa);
1239 	}
1240 }
1241 
1242 /*
1243  * This is the analogue to the rt_newaddrmsg which performs the same
1244  * function but for multicast group memberhips.  This is easier since
1245  * there is no route state to worry about.
1246  */
1247 void
1248 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1249 {
1250 	struct rt_addrinfo info;
1251 	struct mbuf *m = NULL;
1252 	struct ifnet *ifp = ifma->ifma_ifp;
1253 	struct ifma_msghdr *ifmam;
1254 
1255 	if (route_cb.any_count == 0)
1256 		return;
1257 
1258 	bzero((caddr_t)&info, sizeof(info));
1259 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1260 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1261 	/*
1262 	 * If a link-layer address is present, present it as a ``gateway''
1263 	 * (similarly to how ARP entries, e.g., are presented).
1264 	 */
1265 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1266 	m = rt_msg1(cmd, &info);
1267 	if (m == NULL)
1268 		return;
1269 	ifmam = mtod(m, struct ifma_msghdr *);
1270 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1271 	    __func__));
1272 	ifmam->ifmam_index = ifp->if_index;
1273 	ifmam->ifmam_addrs = info.rti_addrs;
1274 	rt_dispatch(m, ifma->ifma_addr);
1275 }
1276 
1277 static struct mbuf *
1278 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1279 	struct rt_addrinfo *info)
1280 {
1281 	struct if_announcemsghdr *ifan;
1282 	struct mbuf *m;
1283 
1284 	if (route_cb.any_count == 0)
1285 		return NULL;
1286 	bzero((caddr_t)info, sizeof(*info));
1287 	m = rt_msg1(type, info);
1288 	if (m != NULL) {
1289 		ifan = mtod(m, struct if_announcemsghdr *);
1290 		ifan->ifan_index = ifp->if_index;
1291 		strlcpy(ifan->ifan_name, ifp->if_xname,
1292 			sizeof(ifan->ifan_name));
1293 		ifan->ifan_what = what;
1294 	}
1295 	return m;
1296 }
1297 
1298 /*
1299  * This is called to generate routing socket messages indicating
1300  * IEEE80211 wireless events.
1301  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1302  */
1303 void
1304 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1305 {
1306 	struct mbuf *m;
1307 	struct rt_addrinfo info;
1308 
1309 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1310 	if (m != NULL) {
1311 		/*
1312 		 * Append the ieee80211 data.  Try to stick it in the
1313 		 * mbuf containing the ifannounce msg; otherwise allocate
1314 		 * a new mbuf and append.
1315 		 *
1316 		 * NB: we assume m is a single mbuf.
1317 		 */
1318 		if (data_len > M_TRAILINGSPACE(m)) {
1319 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1320 			if (n == NULL) {
1321 				m_freem(m);
1322 				return;
1323 			}
1324 			bcopy(data, mtod(n, void *), data_len);
1325 			n->m_len = data_len;
1326 			m->m_next = n;
1327 		} else if (data_len > 0) {
1328 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1329 			m->m_len += data_len;
1330 		}
1331 		if (m->m_flags & M_PKTHDR)
1332 			m->m_pkthdr.len += data_len;
1333 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1334 		rt_dispatch(m, NULL);
1335 	}
1336 }
1337 
1338 /*
1339  * This is called to generate routing socket messages indicating
1340  * network interface arrival and departure.
1341  */
1342 void
1343 rt_ifannouncemsg(struct ifnet *ifp, int what)
1344 {
1345 	struct mbuf *m;
1346 	struct rt_addrinfo info;
1347 
1348 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1349 	if (m != NULL)
1350 		rt_dispatch(m, NULL);
1351 }
1352 
1353 static void
1354 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1355 {
1356 	struct m_tag *tag;
1357 
1358 	/*
1359 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1360 	 * use when injecting the mbuf into the routing socket buffer from
1361 	 * the netisr.
1362 	 */
1363 	if (sa != NULL) {
1364 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1365 		    M_NOWAIT);
1366 		if (tag == NULL) {
1367 			m_freem(m);
1368 			return;
1369 		}
1370 		*(unsigned short *)(tag + 1) = sa->sa_family;
1371 		m_tag_prepend(m, tag);
1372 	}
1373 #ifdef VIMAGE
1374 	if (V_loif)
1375 		m->m_pkthdr.rcvif = V_loif;
1376 	else {
1377 		m_freem(m);
1378 		return;
1379 	}
1380 #endif
1381 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1382 }
1383 
1384 /*
1385  * This is used in dumping the kernel table via sysctl().
1386  */
1387 static int
1388 sysctl_dumpentry(struct radix_node *rn, void *vw)
1389 {
1390 	struct walkarg *w = vw;
1391 	struct rtentry *rt = (struct rtentry *)rn;
1392 	int error = 0, size;
1393 	struct rt_addrinfo info;
1394 
1395 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1396 		return 0;
1397 	if ((rt->rt_flags & RTF_HOST) == 0
1398 	    ? jailed_without_vnet(w->w_req->td->td_ucred)
1399 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1400 		return (0);
1401 	bzero((caddr_t)&info, sizeof(info));
1402 	info.rti_info[RTAX_DST] = rt_key(rt);
1403 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1404 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1405 	info.rti_info[RTAX_GENMASK] = 0;
1406 	if (rt->rt_ifp) {
1407 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1408 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1409 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1410 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1411 	}
1412 	size = rt_msg2(RTM_GET, &info, NULL, w);
1413 	if (w->w_req && w->w_tmem) {
1414 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1415 
1416 		rtm->rtm_flags = rt->rt_flags;
1417 		/*
1418 		 * let's be honest about this being a retarded hack
1419 		 */
1420 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1421 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1422 		rtm->rtm_index = rt->rt_ifp->if_index;
1423 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1424 		rtm->rtm_addrs = info.rti_addrs;
1425 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1426 		return (error);
1427 	}
1428 	return (error);
1429 }
1430 
1431 #ifdef COMPAT_FREEBSD32
1432 static void
1433 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1434 {
1435 
1436 	bzero(dst, sizeof(*dst));
1437 	CP(*src, *dst, ifi_type);
1438 	CP(*src, *dst, ifi_physical);
1439 	CP(*src, *dst, ifi_addrlen);
1440 	CP(*src, *dst, ifi_hdrlen);
1441 	CP(*src, *dst, ifi_link_state);
1442 	dst->ifi_datalen = sizeof(struct if_data32);
1443 	CP(*src, *dst, ifi_mtu);
1444 	CP(*src, *dst, ifi_metric);
1445 	CP(*src, *dst, ifi_baudrate);
1446 	CP(*src, *dst, ifi_ipackets);
1447 	CP(*src, *dst, ifi_ierrors);
1448 	CP(*src, *dst, ifi_opackets);
1449 	CP(*src, *dst, ifi_oerrors);
1450 	CP(*src, *dst, ifi_collisions);
1451 	CP(*src, *dst, ifi_ibytes);
1452 	CP(*src, *dst, ifi_obytes);
1453 	CP(*src, *dst, ifi_imcasts);
1454 	CP(*src, *dst, ifi_omcasts);
1455 	CP(*src, *dst, ifi_iqdrops);
1456 	CP(*src, *dst, ifi_noproto);
1457 	CP(*src, *dst, ifi_hwassist);
1458 	CP(*src, *dst, ifi_epoch);
1459 	TV_CP(*src, *dst, ifi_lastchange);
1460 }
1461 #endif
1462 
1463 static int
1464 sysctl_iflist(int af, struct walkarg *w)
1465 {
1466 	struct ifnet *ifp;
1467 	struct ifaddr *ifa;
1468 	struct rt_addrinfo info;
1469 	int len, error = 0;
1470 
1471 	bzero((caddr_t)&info, sizeof(info));
1472 	IFNET_RLOCK();
1473 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1474 		if (w->w_arg && w->w_arg != ifp->if_index)
1475 			continue;
1476 		ifa = ifp->if_addr;
1477 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1478 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1479 		info.rti_info[RTAX_IFP] = NULL;
1480 		if (w->w_req && w->w_tmem) {
1481 			struct if_msghdr *ifm;
1482 
1483 #ifdef COMPAT_FREEBSD32
1484 			if (w->w_req->flags & SCTL_MASK32) {
1485 				struct if_msghdr32 *ifm32;
1486 
1487 				ifm32 = (struct if_msghdr32 *)w->w_tmem;
1488 				ifm32->ifm_index = ifp->if_index;
1489 				ifm32->ifm_flags = ifp->if_flags |
1490 				    ifp->if_drv_flags;
1491 				copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1492 				ifm32->ifm_addrs = info.rti_addrs;
1493 				error = SYSCTL_OUT(w->w_req, (caddr_t)ifm32,
1494 				    len);
1495 				goto sysctl_out;
1496 			}
1497 #endif
1498 			ifm = (struct if_msghdr *)w->w_tmem;
1499 			ifm->ifm_index = ifp->if_index;
1500 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1501 			ifm->ifm_data = ifp->if_data;
1502 			ifm->ifm_addrs = info.rti_addrs;
1503 			error = SYSCTL_OUT(w->w_req, (caddr_t)ifm, len);
1504 #ifdef COMPAT_FREEBSD32
1505 		sysctl_out:
1506 #endif
1507 			if (error)
1508 				goto done;
1509 		}
1510 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1511 			if (af && af != ifa->ifa_addr->sa_family)
1512 				continue;
1513 			if (prison_if(w->w_req->td->td_ucred,
1514 			    ifa->ifa_addr) != 0)
1515 				continue;
1516 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1517 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1518 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1519 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1520 			if (w->w_req && w->w_tmem) {
1521 				struct ifa_msghdr *ifam;
1522 
1523 				ifam = (struct ifa_msghdr *)w->w_tmem;
1524 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1525 				ifam->ifam_flags = ifa->ifa_flags;
1526 				ifam->ifam_metric = ifa->ifa_metric;
1527 				ifam->ifam_addrs = info.rti_addrs;
1528 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1529 				if (error)
1530 					goto done;
1531 			}
1532 		}
1533 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1534 			info.rti_info[RTAX_BRD] = NULL;
1535 	}
1536 done:
1537 	IFNET_RUNLOCK();
1538 	return (error);
1539 }
1540 
1541 static int
1542 sysctl_ifmalist(int af, struct walkarg *w)
1543 {
1544 	struct ifnet *ifp;
1545 	struct ifmultiaddr *ifma;
1546 	struct	rt_addrinfo info;
1547 	int	len, error = 0;
1548 	struct ifaddr *ifa;
1549 
1550 	bzero((caddr_t)&info, sizeof(info));
1551 	IFNET_RLOCK();
1552 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1553 		if (w->w_arg && w->w_arg != ifp->if_index)
1554 			continue;
1555 		ifa = ifp->if_addr;
1556 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1557 		IF_ADDR_LOCK(ifp);
1558 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1559 			if (af && af != ifma->ifma_addr->sa_family)
1560 				continue;
1561 			if (prison_if(w->w_req->td->td_ucred,
1562 			    ifma->ifma_addr) != 0)
1563 				continue;
1564 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1565 			info.rti_info[RTAX_GATEWAY] =
1566 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1567 			    ifma->ifma_lladdr : NULL;
1568 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1569 			if (w->w_req && w->w_tmem) {
1570 				struct ifma_msghdr *ifmam;
1571 
1572 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1573 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1574 				ifmam->ifmam_flags = 0;
1575 				ifmam->ifmam_addrs = info.rti_addrs;
1576 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1577 				if (error) {
1578 					IF_ADDR_UNLOCK(ifp);
1579 					goto done;
1580 				}
1581 			}
1582 		}
1583 		IF_ADDR_UNLOCK(ifp);
1584 	}
1585 done:
1586 	IFNET_RUNLOCK();
1587 	return (error);
1588 }
1589 
1590 static int
1591 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1592 {
1593 	int	*name = (int *)arg1;
1594 	u_int	namelen = arg2;
1595 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1596 	int	i, lim, error = EINVAL;
1597 	u_char	af;
1598 	struct	walkarg w;
1599 
1600 	name ++;
1601 	namelen--;
1602 	if (req->newptr)
1603 		return (EPERM);
1604 	if (namelen != 3)
1605 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1606 	af = name[0];
1607 	if (af > AF_MAX)
1608 		return (EINVAL);
1609 	bzero(&w, sizeof(w));
1610 	w.w_op = name[1];
1611 	w.w_arg = name[2];
1612 	w.w_req = req;
1613 
1614 	error = sysctl_wire_old_buffer(req, 0);
1615 	if (error)
1616 		return (error);
1617 	switch (w.w_op) {
1618 
1619 	case NET_RT_DUMP:
1620 	case NET_RT_FLAGS:
1621 		if (af == 0) {			/* dump all tables */
1622 			i = 1;
1623 			lim = AF_MAX;
1624 		} else				/* dump only one table */
1625 			i = lim = af;
1626 
1627 		/*
1628 		 * take care of llinfo entries, the caller must
1629 		 * specify an AF
1630 		 */
1631 		if (w.w_op == NET_RT_FLAGS &&
1632 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1633 			if (af != 0)
1634 				error = lltable_sysctl_dumparp(af, w.w_req);
1635 			else
1636 				error = EINVAL;
1637 			break;
1638 		}
1639 		/*
1640 		 * take care of routing entries
1641 		 */
1642 		for (error = 0; error == 0 && i <= lim; i++) {
1643 			rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1644 			if (rnh != NULL) {
1645 				RADIX_NODE_HEAD_LOCK(rnh);
1646 			    	error = rnh->rnh_walktree(rnh,
1647 				    sysctl_dumpentry, &w);
1648 				RADIX_NODE_HEAD_UNLOCK(rnh);
1649 			} else if (af != 0)
1650 				error = EAFNOSUPPORT;
1651 		}
1652 		break;
1653 
1654 	case NET_RT_IFLIST:
1655 		error = sysctl_iflist(af, &w);
1656 		break;
1657 
1658 	case NET_RT_IFMALIST:
1659 		error = sysctl_ifmalist(af, &w);
1660 		break;
1661 	}
1662 	if (w.w_tmem)
1663 		free(w.w_tmem, M_RTABLE);
1664 	return (error);
1665 }
1666 
1667 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1668 
1669 /*
1670  * Definitions of protocols supported in the ROUTE domain.
1671  */
1672 
1673 static struct domain routedomain;		/* or at least forward */
1674 
1675 static struct protosw routesw[] = {
1676 {
1677 	.pr_type =		SOCK_RAW,
1678 	.pr_domain =		&routedomain,
1679 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1680 	.pr_output =		route_output,
1681 	.pr_ctlinput =		raw_ctlinput,
1682 	.pr_init =		raw_init,
1683 	.pr_usrreqs =		&route_usrreqs
1684 }
1685 };
1686 
1687 static struct domain routedomain = {
1688 	.dom_family =		PF_ROUTE,
1689 	.dom_name =		 "route",
1690 	.dom_protosw =		routesw,
1691 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1692 };
1693 
1694 VNET_DOMAIN_SET(route);
1695