xref: /freebsd/sys/net/rtsock.c (revision 7850fa71f55a16f414bb21163d80a03a5ab34522)
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_sctp.h"
33 #include "opt_mpath.h"
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 
37 #include <sys/param.h>
38 #include <sys/domain.h>
39 #include <sys/jail.h>
40 #include <sys/kernel.h>
41 #include <sys/lock.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/priv.h>
45 #include <sys/proc.h>
46 #include <sys/protosw.h>
47 #include <sys/rwlock.h>
48 #include <sys/signalvar.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/sysctl.h>
52 #include <sys/systm.h>
53 #include <sys/vimage.h>
54 
55 #include <net/if.h>
56 #include <net/if_dl.h>
57 #include <net/if_llatbl.h>
58 #include <net/netisr.h>
59 #include <net/raw_cb.h>
60 #include <net/route.h>
61 #include <net/vnet.h>
62 
63 #include <netinet/in.h>
64 #ifdef INET6
65 #include <netinet6/scope6_var.h>
66 #endif
67 
68 #if defined(INET) || defined(INET6)
69 #ifdef SCTP
70 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
71 #endif /* SCTP */
72 #endif
73 
74 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
75 
76 /* NB: these are not modified */
77 static struct	sockaddr route_src = { 2, PF_ROUTE, };
78 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
79 
80 static struct {
81 	int	ip_count;	/* attached w/ AF_INET */
82 	int	ip6_count;	/* attached w/ AF_INET6 */
83 	int	ipx_count;	/* attached w/ AF_IPX */
84 	int	any_count;	/* total attached */
85 } route_cb;
86 
87 struct mtx rtsock_mtx;
88 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
89 
90 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
91 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
92 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
93 
94 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
95 
96 struct walkarg {
97 	int	w_tmemsize;
98 	int	w_op, w_arg;
99 	caddr_t	w_tmem;
100 	struct sysctl_req *w_req;
101 };
102 
103 static void	rts_input(struct mbuf *m);
104 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
105 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
106 			caddr_t cp, struct walkarg *w);
107 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
108 			struct rt_addrinfo *rtinfo);
109 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
110 static int	sysctl_iflist(int af, struct walkarg *w);
111 static int	sysctl_ifmalist(int af, struct walkarg *w);
112 static int	route_output(struct mbuf *m, struct socket *so);
113 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
114 			struct rt_metrics_lite *out);
115 static void	rt_getmetrics(const struct rt_metrics_lite *in,
116 			struct rt_metrics *out);
117 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
118 
119 static struct netisr_handler rtsock_nh = {
120 	.nh_name = "rtsock",
121 	.nh_handler = rts_input,
122 	.nh_proto = NETISR_ROUTE,
123 	.nh_policy = NETISR_POLICY_SOURCE,
124 };
125 
126 static int
127 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
128 {
129 	int error, qlimit;
130 
131 	netisr_getqlimit(&rtsock_nh, &qlimit);
132 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
133         if (error || !req->newptr)
134                 return (error);
135 	if (qlimit < 1)
136 		return (EINVAL);
137 	return (netisr_setqlimit(&rtsock_nh, qlimit));
138 }
139 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
140     0, 0, sysctl_route_netisr_maxqlen, "I",
141     "maximum routing socket dispatch queue length");
142 
143 static void
144 rts_init(void)
145 {
146 	int tmp;
147 
148 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
149 		rtsock_nh.nh_qlimit = tmp;
150 	netisr_register(&rtsock_nh);
151 }
152 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
153 
154 static void
155 rts_input(struct mbuf *m)
156 {
157 	struct sockproto route_proto;
158 	unsigned short *family;
159 	struct m_tag *tag;
160 
161 	route_proto.sp_family = PF_ROUTE;
162 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
163 	if (tag != NULL) {
164 		family = (unsigned short *)(tag + 1);
165 		route_proto.sp_protocol = *family;
166 		m_tag_delete(m, tag);
167 	} else
168 		route_proto.sp_protocol = 0;
169 
170 	raw_input(m, &route_proto, &route_src);
171 }
172 
173 /*
174  * It really doesn't make any sense at all for this code to share much
175  * with raw_usrreq.c, since its functionality is so restricted.  XXX
176  */
177 static void
178 rts_abort(struct socket *so)
179 {
180 
181 	raw_usrreqs.pru_abort(so);
182 }
183 
184 static void
185 rts_close(struct socket *so)
186 {
187 
188 	raw_usrreqs.pru_close(so);
189 }
190 
191 /* pru_accept is EOPNOTSUPP */
192 
193 static int
194 rts_attach(struct socket *so, int proto, struct thread *td)
195 {
196 	struct rawcb *rp;
197 	int s, error;
198 
199 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
200 
201 	/* XXX */
202 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
203 	if (rp == NULL)
204 		return ENOBUFS;
205 
206 	/*
207 	 * The splnet() is necessary to block protocols from sending
208 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
209 	 * this PCB is extant but incompletely initialized.
210 	 * Probably we should try to do more of this work beforehand and
211 	 * eliminate the spl.
212 	 */
213 	s = splnet();
214 	so->so_pcb = (caddr_t)rp;
215 	so->so_fibnum = td->td_proc->p_fibnum;
216 	error = raw_attach(so, proto);
217 	rp = sotorawcb(so);
218 	if (error) {
219 		splx(s);
220 		so->so_pcb = NULL;
221 		free(rp, M_PCB);
222 		return error;
223 	}
224 	RTSOCK_LOCK();
225 	switch(rp->rcb_proto.sp_protocol) {
226 	case AF_INET:
227 		route_cb.ip_count++;
228 		break;
229 	case AF_INET6:
230 		route_cb.ip6_count++;
231 		break;
232 	case AF_IPX:
233 		route_cb.ipx_count++;
234 		break;
235 	}
236 	route_cb.any_count++;
237 	RTSOCK_UNLOCK();
238 	soisconnected(so);
239 	so->so_options |= SO_USELOOPBACK;
240 	splx(s);
241 	return 0;
242 }
243 
244 static int
245 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
246 {
247 
248 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
249 }
250 
251 static int
252 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
253 {
254 
255 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
256 }
257 
258 /* pru_connect2 is EOPNOTSUPP */
259 /* pru_control is EOPNOTSUPP */
260 
261 static void
262 rts_detach(struct socket *so)
263 {
264 	struct rawcb *rp = sotorawcb(so);
265 
266 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
267 
268 	RTSOCK_LOCK();
269 	switch(rp->rcb_proto.sp_protocol) {
270 	case AF_INET:
271 		route_cb.ip_count--;
272 		break;
273 	case AF_INET6:
274 		route_cb.ip6_count--;
275 		break;
276 	case AF_IPX:
277 		route_cb.ipx_count--;
278 		break;
279 	}
280 	route_cb.any_count--;
281 	RTSOCK_UNLOCK();
282 	raw_usrreqs.pru_detach(so);
283 }
284 
285 static int
286 rts_disconnect(struct socket *so)
287 {
288 
289 	return (raw_usrreqs.pru_disconnect(so));
290 }
291 
292 /* pru_listen is EOPNOTSUPP */
293 
294 static int
295 rts_peeraddr(struct socket *so, struct sockaddr **nam)
296 {
297 
298 	return (raw_usrreqs.pru_peeraddr(so, nam));
299 }
300 
301 /* pru_rcvd is EOPNOTSUPP */
302 /* pru_rcvoob is EOPNOTSUPP */
303 
304 static int
305 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
306 	 struct mbuf *control, struct thread *td)
307 {
308 
309 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
310 }
311 
312 /* pru_sense is null */
313 
314 static int
315 rts_shutdown(struct socket *so)
316 {
317 
318 	return (raw_usrreqs.pru_shutdown(so));
319 }
320 
321 static int
322 rts_sockaddr(struct socket *so, struct sockaddr **nam)
323 {
324 
325 	return (raw_usrreqs.pru_sockaddr(so, nam));
326 }
327 
328 static struct pr_usrreqs route_usrreqs = {
329 	.pru_abort =		rts_abort,
330 	.pru_attach =		rts_attach,
331 	.pru_bind =		rts_bind,
332 	.pru_connect =		rts_connect,
333 	.pru_detach =		rts_detach,
334 	.pru_disconnect =	rts_disconnect,
335 	.pru_peeraddr =		rts_peeraddr,
336 	.pru_send =		rts_send,
337 	.pru_shutdown =		rts_shutdown,
338 	.pru_sockaddr =		rts_sockaddr,
339 	.pru_close =		rts_close,
340 };
341 
342 #ifndef _SOCKADDR_UNION_DEFINED
343 #define	_SOCKADDR_UNION_DEFINED
344 /*
345  * The union of all possible address formats we handle.
346  */
347 union sockaddr_union {
348 	struct sockaddr		sa;
349 	struct sockaddr_in	sin;
350 	struct sockaddr_in6	sin6;
351 };
352 #endif /* _SOCKADDR_UNION_DEFINED */
353 
354 static int
355 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
356     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
357 {
358 
359 	/* First, see if the returned address is part of the jail. */
360 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
361 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
362 		return (0);
363 	}
364 
365 	switch (info->rti_info[RTAX_DST]->sa_family) {
366 #ifdef INET
367 	case AF_INET:
368 	{
369 		struct in_addr ia;
370 		struct ifaddr *ifa;
371 		int found;
372 
373 		found = 0;
374 		/*
375 		 * Try to find an address on the given outgoing interface
376 		 * that belongs to the jail.
377 		 */
378 		IF_ADDR_LOCK(ifp);
379 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
380 			struct sockaddr *sa;
381 			sa = ifa->ifa_addr;
382 			if (sa->sa_family != AF_INET)
383 				continue;
384 			ia = ((struct sockaddr_in *)sa)->sin_addr;
385 			if (prison_check_ip4(cred, &ia) == 0) {
386 				found = 1;
387 				break;
388 			}
389 		}
390 		IF_ADDR_UNLOCK(ifp);
391 		if (!found) {
392 			/*
393 			 * As a last resort return the 'default' jail address.
394 			 */
395 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
396 			    sin_addr;
397 			if (prison_get_ip4(cred, &ia) != 0)
398 				return (ESRCH);
399 		}
400 		bzero(&saun->sin, sizeof(struct sockaddr_in));
401 		saun->sin.sin_len = sizeof(struct sockaddr_in);
402 		saun->sin.sin_family = AF_INET;
403 		saun->sin.sin_addr.s_addr = ia.s_addr;
404 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
405 		break;
406 	}
407 #endif
408 #ifdef INET6
409 	case AF_INET6:
410 	{
411 		struct in6_addr ia6;
412 		struct ifaddr *ifa;
413 		int found;
414 
415 		found = 0;
416 		/*
417 		 * Try to find an address on the given outgoing interface
418 		 * that belongs to the jail.
419 		 */
420 		IF_ADDR_LOCK(ifp);
421 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
422 			struct sockaddr *sa;
423 			sa = ifa->ifa_addr;
424 			if (sa->sa_family != AF_INET6)
425 				continue;
426 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
427 			    &ia6, sizeof(struct in6_addr));
428 			if (prison_check_ip6(cred, &ia6) == 0) {
429 				found = 1;
430 				break;
431 			}
432 		}
433 		IF_ADDR_UNLOCK(ifp);
434 		if (!found) {
435 			/*
436 			 * As a last resort return the 'default' jail address.
437 			 */
438 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
439 			    sin6_addr;
440 			if (prison_get_ip6(cred, &ia6) != 0)
441 				return (ESRCH);
442 		}
443 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
444 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
445 		saun->sin6.sin6_family = AF_INET6;
446 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
447 		if (sa6_recoverscope(&saun->sin6) != 0)
448 			return (ESRCH);
449 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
450 		break;
451 	}
452 #endif
453 	default:
454 		return (ESRCH);
455 	}
456 	return (0);
457 }
458 
459 /*ARGSUSED*/
460 static int
461 route_output(struct mbuf *m, struct socket *so)
462 {
463 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
464 	struct rt_msghdr *rtm = NULL;
465 	struct rtentry *rt = NULL;
466 	struct radix_node_head *rnh;
467 	struct rt_addrinfo info;
468 	int len, error = 0;
469 	struct ifnet *ifp = NULL;
470 	union sockaddr_union saun;
471 
472 #define senderr(e) { error = e; goto flush;}
473 	if (m == NULL || ((m->m_len < sizeof(long)) &&
474 		       (m = m_pullup(m, sizeof(long))) == NULL))
475 		return (ENOBUFS);
476 	if ((m->m_flags & M_PKTHDR) == 0)
477 		panic("route_output");
478 	len = m->m_pkthdr.len;
479 	if (len < sizeof(*rtm) ||
480 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
481 		info.rti_info[RTAX_DST] = NULL;
482 		senderr(EINVAL);
483 	}
484 	R_Malloc(rtm, struct rt_msghdr *, len);
485 	if (rtm == NULL) {
486 		info.rti_info[RTAX_DST] = NULL;
487 		senderr(ENOBUFS);
488 	}
489 	m_copydata(m, 0, len, (caddr_t)rtm);
490 	if (rtm->rtm_version != RTM_VERSION) {
491 		info.rti_info[RTAX_DST] = NULL;
492 		senderr(EPROTONOSUPPORT);
493 	}
494 	rtm->rtm_pid = curproc->p_pid;
495 	bzero(&info, sizeof(info));
496 	info.rti_addrs = rtm->rtm_addrs;
497 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
498 		info.rti_info[RTAX_DST] = NULL;
499 		senderr(EINVAL);
500 	}
501 	info.rti_flags = rtm->rtm_flags;
502 	if (info.rti_info[RTAX_DST] == NULL ||
503 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
504 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
505 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
506 		senderr(EINVAL);
507 	/*
508 	 * Verify that the caller has the appropriate privilege; RTM_GET
509 	 * is the only operation the non-superuser is allowed.
510 	 */
511 	if (rtm->rtm_type != RTM_GET) {
512 		error = priv_check(curthread, PRIV_NET_ROUTE);
513 		if (error)
514 			senderr(error);
515 	}
516 
517 	switch (rtm->rtm_type) {
518 		struct rtentry *saved_nrt;
519 
520 	case RTM_ADD:
521 		if (info.rti_info[RTAX_GATEWAY] == NULL)
522 			senderr(EINVAL);
523 		saved_nrt = NULL;
524 
525 		/* support for new ARP code */
526 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
527 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
528 			error = lla_rt_output(rtm, &info);
529 			break;
530 		}
531 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
532 		    so->so_fibnum);
533 		if (error == 0 && saved_nrt) {
534 			RT_LOCK(saved_nrt);
535 			rt_setmetrics(rtm->rtm_inits,
536 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
537 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
538 			RT_REMREF(saved_nrt);
539 			RT_UNLOCK(saved_nrt);
540 		}
541 		break;
542 
543 	case RTM_DELETE:
544 		saved_nrt = NULL;
545 		/* support for new ARP code */
546 		if (info.rti_info[RTAX_GATEWAY] &&
547 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
548 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
549 			error = lla_rt_output(rtm, &info);
550 			break;
551 		}
552 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
553 		    so->so_fibnum);
554 		if (error == 0) {
555 			RT_LOCK(saved_nrt);
556 			rt = saved_nrt;
557 			goto report;
558 		}
559 		break;
560 
561 	case RTM_GET:
562 	case RTM_CHANGE:
563 	case RTM_LOCK:
564 		rnh = rt_tables_get_rnh(so->so_fibnum,
565 		    info.rti_info[RTAX_DST]->sa_family);
566 		if (rnh == NULL)
567 			senderr(EAFNOSUPPORT);
568 		RADIX_NODE_HEAD_RLOCK(rnh);
569 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
570 			info.rti_info[RTAX_NETMASK], rnh);
571 		if (rt == NULL) {	/* XXX looks bogus */
572 			RADIX_NODE_HEAD_RUNLOCK(rnh);
573 			senderr(ESRCH);
574 		}
575 #ifdef RADIX_MPATH
576 		/*
577 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
578 		 * we require users to specify a matching RTAX_GATEWAY.
579 		 *
580 		 * for RTM_GET, gate is optional even with multipath.
581 		 * if gate == NULL the first match is returned.
582 		 * (no need to call rt_mpath_matchgate if gate == NULL)
583 		 */
584 		if (rn_mpath_capable(rnh) &&
585 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
586 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
587 			if (!rt) {
588 				RADIX_NODE_HEAD_RUNLOCK(rnh);
589 				senderr(ESRCH);
590 			}
591 		}
592 #endif
593 		RT_LOCK(rt);
594 		RT_ADDREF(rt);
595 		RADIX_NODE_HEAD_RUNLOCK(rnh);
596 
597 		/*
598 		 * Fix for PR: 82974
599 		 *
600 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
601 		 * returns a perfect match in case a netmask is
602 		 * specified.  For host routes only a longest prefix
603 		 * match is returned so it is necessary to compare the
604 		 * existence of the netmask.  If both have a netmask
605 		 * rnh_lookup() did a perfect match and if none of them
606 		 * have a netmask both are host routes which is also a
607 		 * perfect match.
608 		 */
609 
610 		if (rtm->rtm_type != RTM_GET &&
611 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
612 			RT_UNLOCK(rt);
613 			senderr(ESRCH);
614 		}
615 
616 		switch(rtm->rtm_type) {
617 
618 		case RTM_GET:
619 		report:
620 			RT_LOCK_ASSERT(rt);
621 			if ((rt->rt_flags & RTF_HOST) == 0
622 			    ? jailed(curthread->td_ucred)
623 			    : prison_if(curthread->td_ucred,
624 			    rt_key(rt)) != 0) {
625 				RT_UNLOCK(rt);
626 				senderr(ESRCH);
627 			}
628 			info.rti_info[RTAX_DST] = rt_key(rt);
629 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
630 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
631 			info.rti_info[RTAX_GENMASK] = 0;
632 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
633 				ifp = rt->rt_ifp;
634 				if (ifp) {
635 					info.rti_info[RTAX_IFP] =
636 					    ifp->if_addr->ifa_addr;
637 					error = rtm_get_jailed(&info, ifp, rt,
638 					    &saun, curthread->td_ucred);
639 					if (error != 0) {
640 						RT_UNLOCK(rt);
641 						senderr(error);
642 					}
643 					if (ifp->if_flags & IFF_POINTOPOINT)
644 						info.rti_info[RTAX_BRD] =
645 						    rt->rt_ifa->ifa_dstaddr;
646 					rtm->rtm_index = ifp->if_index;
647 				} else {
648 					info.rti_info[RTAX_IFP] = NULL;
649 					info.rti_info[RTAX_IFA] = NULL;
650 				}
651 			} else if ((ifp = rt->rt_ifp) != NULL) {
652 				rtm->rtm_index = ifp->if_index;
653 			}
654 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
655 			if (len > rtm->rtm_msglen) {
656 				struct rt_msghdr *new_rtm;
657 				R_Malloc(new_rtm, struct rt_msghdr *, len);
658 				if (new_rtm == NULL) {
659 					RT_UNLOCK(rt);
660 					senderr(ENOBUFS);
661 				}
662 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
663 				Free(rtm); rtm = new_rtm;
664 			}
665 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
666 			rtm->rtm_flags = rt->rt_flags;
667 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
668 			rtm->rtm_addrs = info.rti_addrs;
669 			break;
670 
671 		case RTM_CHANGE:
672 			/*
673 			 * New gateway could require new ifaddr, ifp;
674 			 * flags may also be different; ifp may be specified
675 			 * by ll sockaddr when protocol address is ambiguous
676 			 */
677 			if (((rt->rt_flags & RTF_GATEWAY) &&
678 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
679 			    info.rti_info[RTAX_IFP] != NULL ||
680 			    (info.rti_info[RTAX_IFA] != NULL &&
681 			     !sa_equal(info.rti_info[RTAX_IFA],
682 				       rt->rt_ifa->ifa_addr))) {
683 				RT_UNLOCK(rt);
684 				RADIX_NODE_HEAD_LOCK(rnh);
685 				error = rt_getifa_fib(&info, rt->rt_fibnum);
686 				/*
687 				 * XXXRW: Really we should release this
688 				 * reference later, but this maintains
689 				 * historical behavior.
690 				 */
691 				if (info.rti_ifa != NULL)
692 					ifa_free(info.rti_ifa);
693 				RADIX_NODE_HEAD_UNLOCK(rnh);
694 				if (error != 0)
695 					senderr(error);
696 				RT_LOCK(rt);
697 			}
698 			if (info.rti_ifa != NULL &&
699 			    info.rti_ifa != rt->rt_ifa &&
700 			    rt->rt_ifa != NULL &&
701 			    rt->rt_ifa->ifa_rtrequest != NULL) {
702 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
703 				    &info);
704 				ifa_free(rt->rt_ifa);
705 			}
706 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
707 				RT_UNLOCK(rt);
708 				RADIX_NODE_HEAD_LOCK(rnh);
709 				RT_LOCK(rt);
710 
711 				error = rt_setgate(rt, rt_key(rt),
712 				    info.rti_info[RTAX_GATEWAY]);
713 				RADIX_NODE_HEAD_UNLOCK(rnh);
714 				if (error != 0) {
715 					RT_UNLOCK(rt);
716 					senderr(error);
717 				}
718 				rt->rt_flags |= RTF_GATEWAY;
719 			}
720 			if (info.rti_ifa != NULL &&
721 			    info.rti_ifa != rt->rt_ifa) {
722 				ifa_ref(info.rti_ifa);
723 				rt->rt_ifa = info.rti_ifa;
724 				rt->rt_ifp = info.rti_ifp;
725 			}
726 			/* Allow some flags to be toggled on change. */
727 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
728 				    (rtm->rtm_flags & RTF_FMASK);
729 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
730 					&rt->rt_rmx);
731 			rtm->rtm_index = rt->rt_ifp->if_index;
732 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
733 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
734 			/* FALLTHROUGH */
735 		case RTM_LOCK:
736 			/* We don't support locks anymore */
737 			break;
738 		}
739 		RT_UNLOCK(rt);
740 		break;
741 
742 	default:
743 		senderr(EOPNOTSUPP);
744 	}
745 
746 flush:
747 	if (rtm) {
748 		if (error)
749 			rtm->rtm_errno = error;
750 		else
751 			rtm->rtm_flags |= RTF_DONE;
752 	}
753 	if (rt)		/* XXX can this be true? */
754 		RTFREE(rt);
755     {
756 	struct rawcb *rp = NULL;
757 	/*
758 	 * Check to see if we don't want our own messages.
759 	 */
760 	if ((so->so_options & SO_USELOOPBACK) == 0) {
761 		if (route_cb.any_count <= 1) {
762 			if (rtm)
763 				Free(rtm);
764 			m_freem(m);
765 			return (error);
766 		}
767 		/* There is another listener, so construct message */
768 		rp = sotorawcb(so);
769 	}
770 	if (rtm) {
771 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
772 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
773 			m_freem(m);
774 			m = NULL;
775 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
776 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
777 		Free(rtm);
778 	}
779 	if (m) {
780 		if (rp) {
781 			/*
782 			 * XXX insure we don't get a copy by
783 			 * invalidating our protocol
784 			 */
785 			unsigned short family = rp->rcb_proto.sp_family;
786 			rp->rcb_proto.sp_family = 0;
787 			rt_dispatch(m, info.rti_info[RTAX_DST]);
788 			rp->rcb_proto.sp_family = family;
789 		} else
790 			rt_dispatch(m, info.rti_info[RTAX_DST]);
791 	}
792     }
793 	return (error);
794 #undef	sa_equal
795 }
796 
797 static void
798 rt_setmetrics(u_long which, const struct rt_metrics *in,
799 	struct rt_metrics_lite *out)
800 {
801 #define metric(f, e) if (which & (f)) out->e = in->e;
802 	/*
803 	 * Only these are stored in the routing entry since introduction
804 	 * of tcp hostcache. The rest is ignored.
805 	 */
806 	metric(RTV_MTU, rmx_mtu);
807 	metric(RTV_WEIGHT, rmx_weight);
808 	/* Userland -> kernel timebase conversion. */
809 	if (which & RTV_EXPIRE)
810 		out->rmx_expire = in->rmx_expire ?
811 		    in->rmx_expire - time_second + time_uptime : 0;
812 #undef metric
813 }
814 
815 static void
816 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
817 {
818 #define metric(e) out->e = in->e;
819 	bzero(out, sizeof(*out));
820 	metric(rmx_mtu);
821 	metric(rmx_weight);
822 	/* Kernel -> userland timebase conversion. */
823 	out->rmx_expire = in->rmx_expire ?
824 	    in->rmx_expire - time_uptime + time_second : 0;
825 #undef metric
826 }
827 
828 /*
829  * Extract the addresses of the passed sockaddrs.
830  * Do a little sanity checking so as to avoid bad memory references.
831  * This data is derived straight from userland.
832  */
833 static int
834 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
835 {
836 	struct sockaddr *sa;
837 	int i;
838 
839 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
840 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
841 			continue;
842 		sa = (struct sockaddr *)cp;
843 		/*
844 		 * It won't fit.
845 		 */
846 		if (cp + sa->sa_len > cplim)
847 			return (EINVAL);
848 		/*
849 		 * there are no more.. quit now
850 		 * If there are more bits, they are in error.
851 		 * I've seen this. route(1) can evidently generate these.
852 		 * This causes kernel to core dump.
853 		 * for compatibility, If we see this, point to a safe address.
854 		 */
855 		if (sa->sa_len == 0) {
856 			rtinfo->rti_info[i] = &sa_zero;
857 			return (0); /* should be EINVAL but for compat */
858 		}
859 		/* accept it */
860 		rtinfo->rti_info[i] = sa;
861 		cp += SA_SIZE(sa);
862 	}
863 	return (0);
864 }
865 
866 static struct mbuf *
867 rt_msg1(int type, struct rt_addrinfo *rtinfo)
868 {
869 	struct rt_msghdr *rtm;
870 	struct mbuf *m;
871 	int i;
872 	struct sockaddr *sa;
873 	int len, dlen;
874 
875 	switch (type) {
876 
877 	case RTM_DELADDR:
878 	case RTM_NEWADDR:
879 		len = sizeof(struct ifa_msghdr);
880 		break;
881 
882 	case RTM_DELMADDR:
883 	case RTM_NEWMADDR:
884 		len = sizeof(struct ifma_msghdr);
885 		break;
886 
887 	case RTM_IFINFO:
888 		len = sizeof(struct if_msghdr);
889 		break;
890 
891 	case RTM_IFANNOUNCE:
892 	case RTM_IEEE80211:
893 		len = sizeof(struct if_announcemsghdr);
894 		break;
895 
896 	default:
897 		len = sizeof(struct rt_msghdr);
898 	}
899 	if (len > MCLBYTES)
900 		panic("rt_msg1");
901 	m = m_gethdr(M_DONTWAIT, MT_DATA);
902 	if (m && len > MHLEN) {
903 		MCLGET(m, M_DONTWAIT);
904 		if ((m->m_flags & M_EXT) == 0) {
905 			m_free(m);
906 			m = NULL;
907 		}
908 	}
909 	if (m == NULL)
910 		return (m);
911 	m->m_pkthdr.len = m->m_len = len;
912 	m->m_pkthdr.rcvif = NULL;
913 	rtm = mtod(m, struct rt_msghdr *);
914 	bzero((caddr_t)rtm, len);
915 	for (i = 0; i < RTAX_MAX; i++) {
916 		if ((sa = rtinfo->rti_info[i]) == NULL)
917 			continue;
918 		rtinfo->rti_addrs |= (1 << i);
919 		dlen = SA_SIZE(sa);
920 		m_copyback(m, len, dlen, (caddr_t)sa);
921 		len += dlen;
922 	}
923 	if (m->m_pkthdr.len != len) {
924 		m_freem(m);
925 		return (NULL);
926 	}
927 	rtm->rtm_msglen = len;
928 	rtm->rtm_version = RTM_VERSION;
929 	rtm->rtm_type = type;
930 	return (m);
931 }
932 
933 static int
934 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
935 {
936 	int i;
937 	int len, dlen, second_time = 0;
938 	caddr_t cp0;
939 
940 	rtinfo->rti_addrs = 0;
941 again:
942 	switch (type) {
943 
944 	case RTM_DELADDR:
945 	case RTM_NEWADDR:
946 		len = sizeof(struct ifa_msghdr);
947 		break;
948 
949 	case RTM_IFINFO:
950 		len = sizeof(struct if_msghdr);
951 		break;
952 
953 	case RTM_NEWMADDR:
954 		len = sizeof(struct ifma_msghdr);
955 		break;
956 
957 	default:
958 		len = sizeof(struct rt_msghdr);
959 	}
960 	cp0 = cp;
961 	if (cp0)
962 		cp += len;
963 	for (i = 0; i < RTAX_MAX; i++) {
964 		struct sockaddr *sa;
965 
966 		if ((sa = rtinfo->rti_info[i]) == NULL)
967 			continue;
968 		rtinfo->rti_addrs |= (1 << i);
969 		dlen = SA_SIZE(sa);
970 		if (cp) {
971 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
972 			cp += dlen;
973 		}
974 		len += dlen;
975 	}
976 	len = ALIGN(len);
977 	if (cp == NULL && w != NULL && !second_time) {
978 		struct walkarg *rw = w;
979 
980 		if (rw->w_req) {
981 			if (rw->w_tmemsize < len) {
982 				if (rw->w_tmem)
983 					free(rw->w_tmem, M_RTABLE);
984 				rw->w_tmem = (caddr_t)
985 					malloc(len, M_RTABLE, M_NOWAIT);
986 				if (rw->w_tmem)
987 					rw->w_tmemsize = len;
988 			}
989 			if (rw->w_tmem) {
990 				cp = rw->w_tmem;
991 				second_time = 1;
992 				goto again;
993 			}
994 		}
995 	}
996 	if (cp) {
997 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
998 
999 		rtm->rtm_version = RTM_VERSION;
1000 		rtm->rtm_type = type;
1001 		rtm->rtm_msglen = len;
1002 	}
1003 	return (len);
1004 }
1005 
1006 /*
1007  * This routine is called to generate a message from the routing
1008  * socket indicating that a redirect has occured, a routing lookup
1009  * has failed, or that a protocol has detected timeouts to a particular
1010  * destination.
1011  */
1012 void
1013 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1014 {
1015 	struct rt_msghdr *rtm;
1016 	struct mbuf *m;
1017 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1018 
1019 	if (route_cb.any_count == 0)
1020 		return;
1021 	m = rt_msg1(type, rtinfo);
1022 	if (m == NULL)
1023 		return;
1024 	rtm = mtod(m, struct rt_msghdr *);
1025 	rtm->rtm_flags = RTF_DONE | flags;
1026 	rtm->rtm_errno = error;
1027 	rtm->rtm_addrs = rtinfo->rti_addrs;
1028 	rt_dispatch(m, sa);
1029 }
1030 
1031 /*
1032  * This routine is called to generate a message from the routing
1033  * socket indicating that the status of a network interface has changed.
1034  */
1035 void
1036 rt_ifmsg(struct ifnet *ifp)
1037 {
1038 	struct if_msghdr *ifm;
1039 	struct mbuf *m;
1040 	struct rt_addrinfo info;
1041 
1042 	if (route_cb.any_count == 0)
1043 		return;
1044 	bzero((caddr_t)&info, sizeof(info));
1045 	m = rt_msg1(RTM_IFINFO, &info);
1046 	if (m == NULL)
1047 		return;
1048 	ifm = mtod(m, struct if_msghdr *);
1049 	ifm->ifm_index = ifp->if_index;
1050 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1051 	ifm->ifm_data = ifp->if_data;
1052 	ifm->ifm_addrs = 0;
1053 	rt_dispatch(m, NULL);
1054 }
1055 
1056 /*
1057  * This is called to generate messages from the routing socket
1058  * indicating a network interface has had addresses associated with it.
1059  * if we ever reverse the logic and replace messages TO the routing
1060  * socket indicate a request to configure interfaces, then it will
1061  * be unnecessary as the routing socket will automatically generate
1062  * copies of it.
1063  */
1064 void
1065 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1066 {
1067 	struct rt_addrinfo info;
1068 	struct sockaddr *sa = NULL;
1069 	int pass;
1070 	struct mbuf *m = NULL;
1071 	struct ifnet *ifp = ifa->ifa_ifp;
1072 
1073 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1074 		("unexpected cmd %u", cmd));
1075 #if defined(INET) || defined(INET6)
1076 #ifdef SCTP
1077 	/*
1078 	 * notify the SCTP stack
1079 	 * this will only get called when an address is added/deleted
1080 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1081 	 */
1082 	sctp_addr_change(ifa, cmd);
1083 #endif /* SCTP */
1084 #endif
1085 	if (route_cb.any_count == 0)
1086 		return;
1087 	for (pass = 1; pass < 3; pass++) {
1088 		bzero((caddr_t)&info, sizeof(info));
1089 		if ((cmd == RTM_ADD && pass == 1) ||
1090 		    (cmd == RTM_DELETE && pass == 2)) {
1091 			struct ifa_msghdr *ifam;
1092 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1093 
1094 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1095 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1096 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1097 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1098 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1099 				continue;
1100 			ifam = mtod(m, struct ifa_msghdr *);
1101 			ifam->ifam_index = ifp->if_index;
1102 			ifam->ifam_metric = ifa->ifa_metric;
1103 			ifam->ifam_flags = ifa->ifa_flags;
1104 			ifam->ifam_addrs = info.rti_addrs;
1105 		}
1106 		if ((cmd == RTM_ADD && pass == 2) ||
1107 		    (cmd == RTM_DELETE && pass == 1)) {
1108 			struct rt_msghdr *rtm;
1109 
1110 			if (rt == NULL)
1111 				continue;
1112 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1113 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1114 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1115 			if ((m = rt_msg1(cmd, &info)) == NULL)
1116 				continue;
1117 			rtm = mtod(m, struct rt_msghdr *);
1118 			rtm->rtm_index = ifp->if_index;
1119 			rtm->rtm_flags |= rt->rt_flags;
1120 			rtm->rtm_errno = error;
1121 			rtm->rtm_addrs = info.rti_addrs;
1122 		}
1123 		rt_dispatch(m, sa);
1124 	}
1125 }
1126 
1127 /*
1128  * This is the analogue to the rt_newaddrmsg which performs the same
1129  * function but for multicast group memberhips.  This is easier since
1130  * there is no route state to worry about.
1131  */
1132 void
1133 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1134 {
1135 	struct rt_addrinfo info;
1136 	struct mbuf *m = NULL;
1137 	struct ifnet *ifp = ifma->ifma_ifp;
1138 	struct ifma_msghdr *ifmam;
1139 
1140 	if (route_cb.any_count == 0)
1141 		return;
1142 
1143 	bzero((caddr_t)&info, sizeof(info));
1144 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1145 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1146 	/*
1147 	 * If a link-layer address is present, present it as a ``gateway''
1148 	 * (similarly to how ARP entries, e.g., are presented).
1149 	 */
1150 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1151 	m = rt_msg1(cmd, &info);
1152 	if (m == NULL)
1153 		return;
1154 	ifmam = mtod(m, struct ifma_msghdr *);
1155 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1156 	    __func__));
1157 	ifmam->ifmam_index = ifp->if_index;
1158 	ifmam->ifmam_addrs = info.rti_addrs;
1159 	rt_dispatch(m, ifma->ifma_addr);
1160 }
1161 
1162 static struct mbuf *
1163 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1164 	struct rt_addrinfo *info)
1165 {
1166 	struct if_announcemsghdr *ifan;
1167 	struct mbuf *m;
1168 
1169 	if (route_cb.any_count == 0)
1170 		return NULL;
1171 	bzero((caddr_t)info, sizeof(*info));
1172 	m = rt_msg1(type, info);
1173 	if (m != NULL) {
1174 		ifan = mtod(m, struct if_announcemsghdr *);
1175 		ifan->ifan_index = ifp->if_index;
1176 		strlcpy(ifan->ifan_name, ifp->if_xname,
1177 			sizeof(ifan->ifan_name));
1178 		ifan->ifan_what = what;
1179 	}
1180 	return m;
1181 }
1182 
1183 /*
1184  * This is called to generate routing socket messages indicating
1185  * IEEE80211 wireless events.
1186  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1187  */
1188 void
1189 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1190 {
1191 	struct mbuf *m;
1192 	struct rt_addrinfo info;
1193 
1194 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1195 	if (m != NULL) {
1196 		/*
1197 		 * Append the ieee80211 data.  Try to stick it in the
1198 		 * mbuf containing the ifannounce msg; otherwise allocate
1199 		 * a new mbuf and append.
1200 		 *
1201 		 * NB: we assume m is a single mbuf.
1202 		 */
1203 		if (data_len > M_TRAILINGSPACE(m)) {
1204 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1205 			if (n == NULL) {
1206 				m_freem(m);
1207 				return;
1208 			}
1209 			bcopy(data, mtod(n, void *), data_len);
1210 			n->m_len = data_len;
1211 			m->m_next = n;
1212 		} else if (data_len > 0) {
1213 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1214 			m->m_len += data_len;
1215 		}
1216 		if (m->m_flags & M_PKTHDR)
1217 			m->m_pkthdr.len += data_len;
1218 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1219 		rt_dispatch(m, NULL);
1220 	}
1221 }
1222 
1223 /*
1224  * This is called to generate routing socket messages indicating
1225  * network interface arrival and departure.
1226  */
1227 void
1228 rt_ifannouncemsg(struct ifnet *ifp, int what)
1229 {
1230 	struct mbuf *m;
1231 	struct rt_addrinfo info;
1232 
1233 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1234 	if (m != NULL)
1235 		rt_dispatch(m, NULL);
1236 }
1237 
1238 static void
1239 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1240 {
1241 	INIT_VNET_NET(curvnet);
1242 	struct m_tag *tag;
1243 
1244 	/*
1245 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1246 	 * use when injecting the mbuf into the routing socket buffer from
1247 	 * the netisr.
1248 	 */
1249 	if (sa != NULL) {
1250 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1251 		    M_NOWAIT);
1252 		if (tag == NULL) {
1253 			m_freem(m);
1254 			return;
1255 		}
1256 		*(unsigned short *)(tag + 1) = sa->sa_family;
1257 		m_tag_prepend(m, tag);
1258 	}
1259 #ifdef VIMAGE
1260 	if (V_loif)
1261 		m->m_pkthdr.rcvif = V_loif;
1262 	else {
1263 		m_freem(m);
1264 		return;
1265 	}
1266 #endif
1267 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1268 }
1269 
1270 /*
1271  * This is used in dumping the kernel table via sysctl().
1272  */
1273 static int
1274 sysctl_dumpentry(struct radix_node *rn, void *vw)
1275 {
1276 	struct walkarg *w = vw;
1277 	struct rtentry *rt = (struct rtentry *)rn;
1278 	int error = 0, size;
1279 	struct rt_addrinfo info;
1280 
1281 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1282 		return 0;
1283 	if ((rt->rt_flags & RTF_HOST) == 0
1284 	    ? jailed(w->w_req->td->td_ucred)
1285 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1286 		return (0);
1287 	bzero((caddr_t)&info, sizeof(info));
1288 	info.rti_info[RTAX_DST] = rt_key(rt);
1289 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1290 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1291 	info.rti_info[RTAX_GENMASK] = 0;
1292 	if (rt->rt_ifp) {
1293 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1294 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1295 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1296 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1297 	}
1298 	size = rt_msg2(RTM_GET, &info, NULL, w);
1299 	if (w->w_req && w->w_tmem) {
1300 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1301 
1302 		rtm->rtm_flags = rt->rt_flags;
1303 		/*
1304 		 * let's be honest about this being a retarded hack
1305 		 */
1306 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1307 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1308 		rtm->rtm_index = rt->rt_ifp->if_index;
1309 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1310 		rtm->rtm_addrs = info.rti_addrs;
1311 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1312 		return (error);
1313 	}
1314 	return (error);
1315 }
1316 
1317 static int
1318 sysctl_iflist(int af, struct walkarg *w)
1319 {
1320 	INIT_VNET_NET(curvnet);
1321 	struct ifnet *ifp;
1322 	struct ifaddr *ifa;
1323 	struct rt_addrinfo info;
1324 	int len, error = 0;
1325 
1326 	bzero((caddr_t)&info, sizeof(info));
1327 	IFNET_RLOCK();
1328 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1329 		if (w->w_arg && w->w_arg != ifp->if_index)
1330 			continue;
1331 		ifa = ifp->if_addr;
1332 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1333 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1334 		info.rti_info[RTAX_IFP] = NULL;
1335 		if (w->w_req && w->w_tmem) {
1336 			struct if_msghdr *ifm;
1337 
1338 			ifm = (struct if_msghdr *)w->w_tmem;
1339 			ifm->ifm_index = ifp->if_index;
1340 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1341 			ifm->ifm_data = ifp->if_data;
1342 			ifm->ifm_addrs = info.rti_addrs;
1343 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
1344 			if (error)
1345 				goto done;
1346 		}
1347 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1348 			if (af && af != ifa->ifa_addr->sa_family)
1349 				continue;
1350 			if (prison_if(w->w_req->td->td_ucred,
1351 			    ifa->ifa_addr) != 0)
1352 				continue;
1353 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1354 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1355 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1356 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1357 			if (w->w_req && w->w_tmem) {
1358 				struct ifa_msghdr *ifam;
1359 
1360 				ifam = (struct ifa_msghdr *)w->w_tmem;
1361 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1362 				ifam->ifam_flags = ifa->ifa_flags;
1363 				ifam->ifam_metric = ifa->ifa_metric;
1364 				ifam->ifam_addrs = info.rti_addrs;
1365 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1366 				if (error)
1367 					goto done;
1368 			}
1369 		}
1370 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1371 			info.rti_info[RTAX_BRD] = NULL;
1372 	}
1373 done:
1374 	IFNET_RUNLOCK();
1375 	return (error);
1376 }
1377 
1378 static int
1379 sysctl_ifmalist(int af, struct walkarg *w)
1380 {
1381 	INIT_VNET_NET(curvnet);
1382 	struct ifnet *ifp;
1383 	struct ifmultiaddr *ifma;
1384 	struct	rt_addrinfo info;
1385 	int	len, error = 0;
1386 	struct ifaddr *ifa;
1387 
1388 	bzero((caddr_t)&info, sizeof(info));
1389 	IFNET_RLOCK();
1390 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1391 		if (w->w_arg && w->w_arg != ifp->if_index)
1392 			continue;
1393 		ifa = ifp->if_addr;
1394 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1395 		IF_ADDR_LOCK(ifp);
1396 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1397 			if (af && af != ifma->ifma_addr->sa_family)
1398 				continue;
1399 			if (prison_if(w->w_req->td->td_ucred,
1400 			    ifma->ifma_addr) != 0)
1401 				continue;
1402 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1403 			info.rti_info[RTAX_GATEWAY] =
1404 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1405 			    ifma->ifma_lladdr : NULL;
1406 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1407 			if (w->w_req && w->w_tmem) {
1408 				struct ifma_msghdr *ifmam;
1409 
1410 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1411 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1412 				ifmam->ifmam_flags = 0;
1413 				ifmam->ifmam_addrs = info.rti_addrs;
1414 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1415 				if (error) {
1416 					IF_ADDR_UNLOCK(ifp);
1417 					goto done;
1418 				}
1419 			}
1420 		}
1421 		IF_ADDR_UNLOCK(ifp);
1422 	}
1423 done:
1424 	IFNET_RUNLOCK();
1425 	return (error);
1426 }
1427 
1428 static int
1429 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1430 {
1431 	int	*name = (int *)arg1;
1432 	u_int	namelen = arg2;
1433 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1434 	int	i, lim, error = EINVAL;
1435 	u_char	af;
1436 	struct	walkarg w;
1437 
1438 	name ++;
1439 	namelen--;
1440 	if (req->newptr)
1441 		return (EPERM);
1442 	if (namelen != 3)
1443 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1444 	af = name[0];
1445 	if (af > AF_MAX)
1446 		return (EINVAL);
1447 	bzero(&w, sizeof(w));
1448 	w.w_op = name[1];
1449 	w.w_arg = name[2];
1450 	w.w_req = req;
1451 
1452 	error = sysctl_wire_old_buffer(req, 0);
1453 	if (error)
1454 		return (error);
1455 	switch (w.w_op) {
1456 
1457 	case NET_RT_DUMP:
1458 	case NET_RT_FLAGS:
1459 		if (af == 0) {			/* dump all tables */
1460 			i = 1;
1461 			lim = AF_MAX;
1462 		} else				/* dump only one table */
1463 			i = lim = af;
1464 
1465 		/*
1466 		 * take care of llinfo entries, the caller must
1467 		 * specify an AF
1468 		 */
1469 		if (w.w_op == NET_RT_FLAGS &&
1470 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1471 			if (af != 0)
1472 				error = lltable_sysctl_dumparp(af, w.w_req);
1473 			else
1474 				error = EINVAL;
1475 			break;
1476 		}
1477 		/*
1478 		 * take care of routing entries
1479 		 */
1480 		for (error = 0; error == 0 && i <= lim; i++)
1481 			rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1482 			if (rnh != NULL) {
1483 				RADIX_NODE_HEAD_LOCK(rnh);
1484 			    	error = rnh->rnh_walktree(rnh,
1485 				    sysctl_dumpentry, &w);
1486 				RADIX_NODE_HEAD_UNLOCK(rnh);
1487 			} else if (af != 0)
1488 				error = EAFNOSUPPORT;
1489 		break;
1490 
1491 	case NET_RT_IFLIST:
1492 		error = sysctl_iflist(af, &w);
1493 		break;
1494 
1495 	case NET_RT_IFMALIST:
1496 		error = sysctl_ifmalist(af, &w);
1497 		break;
1498 	}
1499 	if (w.w_tmem)
1500 		free(w.w_tmem, M_RTABLE);
1501 	return (error);
1502 }
1503 
1504 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1505 
1506 /*
1507  * Definitions of protocols supported in the ROUTE domain.
1508  */
1509 
1510 static struct domain routedomain;		/* or at least forward */
1511 
1512 static struct protosw routesw[] = {
1513 {
1514 	.pr_type =		SOCK_RAW,
1515 	.pr_domain =		&routedomain,
1516 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1517 	.pr_output =		route_output,
1518 	.pr_ctlinput =		raw_ctlinput,
1519 	.pr_init =		raw_init,
1520 	.pr_usrreqs =		&route_usrreqs
1521 }
1522 };
1523 
1524 static struct domain routedomain = {
1525 	.dom_family =		PF_ROUTE,
1526 	.dom_name =		 "route",
1527 	.dom_protosw =		routesw,
1528 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1529 };
1530 
1531 DOMAIN_SET(route);
1532