xref: /freebsd/sys/net/rtsock.c (revision 195ebc7e9e4b129de810833791a19dfb4349d6a9)
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_route.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/domain.h>
40 #include <sys/jail.h>
41 #include <sys/kernel.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 #include <sys/vimage.h>
55 
56 #include <net/if.h>
57 #include <net/if_dl.h>
58 #include <net/if_llatbl.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 #ifdef INET6
66 #include <netinet6/scope6_var.h>
67 #endif
68 
69 #ifdef SCTP
70 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
71 #endif /* SCTP */
72 
73 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
74 
75 /* NB: these are not modified */
76 static struct	sockaddr route_src = { 2, PF_ROUTE, };
77 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
78 
79 static struct {
80 	int	ip_count;	/* attached w/ AF_INET */
81 	int	ip6_count;	/* attached w/ AF_INET6 */
82 	int	ipx_count;	/* attached w/ AF_IPX */
83 	int	any_count;	/* total attached */
84 } route_cb;
85 
86 struct mtx rtsock_mtx;
87 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
88 
89 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
90 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
91 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
92 
93 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
94 
95 struct walkarg {
96 	int	w_tmemsize;
97 	int	w_op, w_arg;
98 	caddr_t	w_tmem;
99 	struct sysctl_req *w_req;
100 };
101 
102 static void	rts_input(struct mbuf *m);
103 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
104 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
105 			caddr_t cp, struct walkarg *w);
106 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
107 			struct rt_addrinfo *rtinfo);
108 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
109 static int	sysctl_iflist(int af, struct walkarg *w);
110 static int	sysctl_ifmalist(int af, struct walkarg *w);
111 static int	route_output(struct mbuf *m, struct socket *so);
112 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
113 			struct rt_metrics_lite *out);
114 static void	rt_getmetrics(const struct rt_metrics_lite *in,
115 			struct rt_metrics *out);
116 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
117 
118 static struct netisr_handler rtsock_nh = {
119 	.nh_name = "rtsock",
120 	.nh_handler = rts_input,
121 	.nh_proto = NETISR_ROUTE,
122 	.nh_policy = NETISR_POLICY_SOURCE,
123 };
124 
125 static int
126 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
127 {
128 	int error, qlimit;
129 
130 	netisr_getqlimit(&rtsock_nh, &qlimit);
131 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
132         if (error || !req->newptr)
133                 return (error);
134 	if (qlimit < 1)
135 		return (EINVAL);
136 	return (netisr_setqlimit(&rtsock_nh, qlimit));
137 }
138 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
139     0, 0, sysctl_route_netisr_maxqlen, "I",
140     "maximum routing socket dispatch queue length");
141 
142 static void
143 rts_init(void)
144 {
145 	int tmp;
146 
147 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
148 		rtsock_nh.nh_qlimit = tmp;
149 	netisr_register(&rtsock_nh);
150 }
151 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
152 
153 static void
154 rts_input(struct mbuf *m)
155 {
156 	struct sockproto route_proto;
157 	unsigned short *family;
158 	struct m_tag *tag;
159 
160 	route_proto.sp_family = PF_ROUTE;
161 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
162 	if (tag != NULL) {
163 		family = (unsigned short *)(tag + 1);
164 		route_proto.sp_protocol = *family;
165 		m_tag_delete(m, tag);
166 	} else
167 		route_proto.sp_protocol = 0;
168 
169 	raw_input(m, &route_proto, &route_src);
170 }
171 
172 /*
173  * It really doesn't make any sense at all for this code to share much
174  * with raw_usrreq.c, since its functionality is so restricted.  XXX
175  */
176 static void
177 rts_abort(struct socket *so)
178 {
179 
180 	raw_usrreqs.pru_abort(so);
181 }
182 
183 static void
184 rts_close(struct socket *so)
185 {
186 
187 	raw_usrreqs.pru_close(so);
188 }
189 
190 /* pru_accept is EOPNOTSUPP */
191 
192 static int
193 rts_attach(struct socket *so, int proto, struct thread *td)
194 {
195 	struct rawcb *rp;
196 	int s, error;
197 
198 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
199 
200 	/* XXX */
201 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
202 	if (rp == NULL)
203 		return ENOBUFS;
204 
205 	/*
206 	 * The splnet() is necessary to block protocols from sending
207 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
208 	 * this PCB is extant but incompletely initialized.
209 	 * Probably we should try to do more of this work beforehand and
210 	 * eliminate the spl.
211 	 */
212 	s = splnet();
213 	so->so_pcb = (caddr_t)rp;
214 	so->so_fibnum = td->td_proc->p_fibnum;
215 	error = raw_attach(so, proto);
216 	rp = sotorawcb(so);
217 	if (error) {
218 		splx(s);
219 		so->so_pcb = NULL;
220 		free(rp, M_PCB);
221 		return error;
222 	}
223 	RTSOCK_LOCK();
224 	switch(rp->rcb_proto.sp_protocol) {
225 	case AF_INET:
226 		route_cb.ip_count++;
227 		break;
228 	case AF_INET6:
229 		route_cb.ip6_count++;
230 		break;
231 	case AF_IPX:
232 		route_cb.ipx_count++;
233 		break;
234 	}
235 	route_cb.any_count++;
236 	RTSOCK_UNLOCK();
237 	soisconnected(so);
238 	so->so_options |= SO_USELOOPBACK;
239 	splx(s);
240 	return 0;
241 }
242 
243 static int
244 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
245 {
246 
247 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
248 }
249 
250 static int
251 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
252 {
253 
254 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
255 }
256 
257 /* pru_connect2 is EOPNOTSUPP */
258 /* pru_control is EOPNOTSUPP */
259 
260 static void
261 rts_detach(struct socket *so)
262 {
263 	struct rawcb *rp = sotorawcb(so);
264 
265 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
266 
267 	RTSOCK_LOCK();
268 	switch(rp->rcb_proto.sp_protocol) {
269 	case AF_INET:
270 		route_cb.ip_count--;
271 		break;
272 	case AF_INET6:
273 		route_cb.ip6_count--;
274 		break;
275 	case AF_IPX:
276 		route_cb.ipx_count--;
277 		break;
278 	}
279 	route_cb.any_count--;
280 	RTSOCK_UNLOCK();
281 	raw_usrreqs.pru_detach(so);
282 }
283 
284 static int
285 rts_disconnect(struct socket *so)
286 {
287 
288 	return (raw_usrreqs.pru_disconnect(so));
289 }
290 
291 /* pru_listen is EOPNOTSUPP */
292 
293 static int
294 rts_peeraddr(struct socket *so, struct sockaddr **nam)
295 {
296 
297 	return (raw_usrreqs.pru_peeraddr(so, nam));
298 }
299 
300 /* pru_rcvd is EOPNOTSUPP */
301 /* pru_rcvoob is EOPNOTSUPP */
302 
303 static int
304 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
305 	 struct mbuf *control, struct thread *td)
306 {
307 
308 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
309 }
310 
311 /* pru_sense is null */
312 
313 static int
314 rts_shutdown(struct socket *so)
315 {
316 
317 	return (raw_usrreqs.pru_shutdown(so));
318 }
319 
320 static int
321 rts_sockaddr(struct socket *so, struct sockaddr **nam)
322 {
323 
324 	return (raw_usrreqs.pru_sockaddr(so, nam));
325 }
326 
327 static struct pr_usrreqs route_usrreqs = {
328 	.pru_abort =		rts_abort,
329 	.pru_attach =		rts_attach,
330 	.pru_bind =		rts_bind,
331 	.pru_connect =		rts_connect,
332 	.pru_detach =		rts_detach,
333 	.pru_disconnect =	rts_disconnect,
334 	.pru_peeraddr =		rts_peeraddr,
335 	.pru_send =		rts_send,
336 	.pru_shutdown =		rts_shutdown,
337 	.pru_sockaddr =		rts_sockaddr,
338 	.pru_close =		rts_close,
339 };
340 
341 #ifndef _SOCKADDR_UNION_DEFINED
342 #define	_SOCKADDR_UNION_DEFINED
343 /*
344  * The union of all possible address formats we handle.
345  */
346 union sockaddr_union {
347 	struct sockaddr		sa;
348 	struct sockaddr_in	sin;
349 	struct sockaddr_in6	sin6;
350 };
351 #endif /* _SOCKADDR_UNION_DEFINED */
352 
353 static int
354 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
355     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
356 {
357 
358 	/* First, see if the returned address is part of the jail. */
359 	if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
360 		info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
361 		return (0);
362 	}
363 
364 	switch (info->rti_info[RTAX_DST]->sa_family) {
365 #ifdef INET
366 	case AF_INET:
367 	{
368 		struct in_addr ia;
369 		struct ifaddr *ifa;
370 		int found;
371 
372 		found = 0;
373 		/*
374 		 * Try to find an address on the given outgoing interface
375 		 * that belongs to the jail.
376 		 */
377 		IF_ADDR_LOCK(ifp);
378 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
379 			struct sockaddr *sa;
380 			sa = ifa->ifa_addr;
381 			if (sa->sa_family != AF_INET)
382 				continue;
383 			ia = ((struct sockaddr_in *)sa)->sin_addr;
384 			if (prison_check_ip4(cred, &ia) == 0) {
385 				found = 1;
386 				break;
387 			}
388 		}
389 		IF_ADDR_UNLOCK(ifp);
390 		if (!found) {
391 			/*
392 			 * As a last resort return the 'default' jail address.
393 			 */
394 			ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
395 			    sin_addr;
396 			if (prison_get_ip4(cred, &ia) != 0)
397 				return (ESRCH);
398 		}
399 		bzero(&saun->sin, sizeof(struct sockaddr_in));
400 		saun->sin.sin_len = sizeof(struct sockaddr_in);
401 		saun->sin.sin_family = AF_INET;
402 		saun->sin.sin_addr.s_addr = ia.s_addr;
403 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
404 		break;
405 	}
406 #endif
407 #ifdef INET6
408 	case AF_INET6:
409 	{
410 		struct in6_addr ia6;
411 		struct ifaddr *ifa;
412 		int found;
413 
414 		found = 0;
415 		/*
416 		 * Try to find an address on the given outgoing interface
417 		 * that belongs to the jail.
418 		 */
419 		IF_ADDR_LOCK(ifp);
420 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
421 			struct sockaddr *sa;
422 			sa = ifa->ifa_addr;
423 			if (sa->sa_family != AF_INET6)
424 				continue;
425 			bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
426 			    &ia6, sizeof(struct in6_addr));
427 			if (prison_check_ip6(cred, &ia6) == 0) {
428 				found = 1;
429 				break;
430 			}
431 		}
432 		IF_ADDR_UNLOCK(ifp);
433 		if (!found) {
434 			/*
435 			 * As a last resort return the 'default' jail address.
436 			 */
437 			ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
438 			    sin6_addr;
439 			if (prison_get_ip6(cred, &ia6) != 0)
440 				return (ESRCH);
441 		}
442 		bzero(&saun->sin6, sizeof(struct sockaddr_in6));
443 		saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
444 		saun->sin6.sin6_family = AF_INET6;
445 		bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
446 		if (sa6_recoverscope(&saun->sin6) != 0)
447 			return (ESRCH);
448 		info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
449 		break;
450 	}
451 #endif
452 	default:
453 		return (ESRCH);
454 	}
455 	return (0);
456 }
457 
458 /*ARGSUSED*/
459 static int
460 route_output(struct mbuf *m, struct socket *so)
461 {
462 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
463 	struct rt_msghdr *rtm = NULL;
464 	struct rtentry *rt = NULL;
465 	struct radix_node_head *rnh;
466 	struct rt_addrinfo info;
467 	int len, error = 0;
468 	struct ifnet *ifp = NULL;
469 	union sockaddr_union saun;
470 
471 #define senderr(e) { error = e; goto flush;}
472 	if (m == NULL || ((m->m_len < sizeof(long)) &&
473 		       (m = m_pullup(m, sizeof(long))) == NULL))
474 		return (ENOBUFS);
475 	if ((m->m_flags & M_PKTHDR) == 0)
476 		panic("route_output");
477 	len = m->m_pkthdr.len;
478 	if (len < sizeof(*rtm) ||
479 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
480 		info.rti_info[RTAX_DST] = NULL;
481 		senderr(EINVAL);
482 	}
483 	R_Malloc(rtm, struct rt_msghdr *, len);
484 	if (rtm == NULL) {
485 		info.rti_info[RTAX_DST] = NULL;
486 		senderr(ENOBUFS);
487 	}
488 	m_copydata(m, 0, len, (caddr_t)rtm);
489 	if (rtm->rtm_version != RTM_VERSION) {
490 		info.rti_info[RTAX_DST] = NULL;
491 		senderr(EPROTONOSUPPORT);
492 	}
493 	rtm->rtm_pid = curproc->p_pid;
494 	bzero(&info, sizeof(info));
495 	info.rti_addrs = rtm->rtm_addrs;
496 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
497 		info.rti_info[RTAX_DST] = NULL;
498 		senderr(EINVAL);
499 	}
500 	info.rti_flags = rtm->rtm_flags;
501 	if (info.rti_info[RTAX_DST] == NULL ||
502 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
503 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
504 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
505 		senderr(EINVAL);
506 	/*
507 	 * Verify that the caller has the appropriate privilege; RTM_GET
508 	 * is the only operation the non-superuser is allowed.
509 	 */
510 	if (rtm->rtm_type != RTM_GET) {
511 		error = priv_check(curthread, PRIV_NET_ROUTE);
512 		if (error)
513 			senderr(error);
514 	}
515 
516 	switch (rtm->rtm_type) {
517 		struct rtentry *saved_nrt;
518 
519 	case RTM_ADD:
520 		if (info.rti_info[RTAX_GATEWAY] == NULL)
521 			senderr(EINVAL);
522 		saved_nrt = NULL;
523 
524 		/* support for new ARP code */
525 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
526 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
527 			error = lla_rt_output(rtm, &info);
528 			break;
529 		}
530 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
531 		    so->so_fibnum);
532 		if (error == 0 && saved_nrt) {
533 			RT_LOCK(saved_nrt);
534 			rt_setmetrics(rtm->rtm_inits,
535 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
536 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
537 			RT_REMREF(saved_nrt);
538 			RT_UNLOCK(saved_nrt);
539 		}
540 		break;
541 
542 	case RTM_DELETE:
543 		saved_nrt = NULL;
544 		/* support for new ARP code */
545 		if (info.rti_info[RTAX_GATEWAY] &&
546 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
547 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
548 			error = lla_rt_output(rtm, &info);
549 			break;
550 		}
551 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
552 		    so->so_fibnum);
553 		if (error == 0) {
554 			RT_LOCK(saved_nrt);
555 			rt = saved_nrt;
556 			goto report;
557 		}
558 		break;
559 
560 	case RTM_GET:
561 	case RTM_CHANGE:
562 	case RTM_LOCK:
563 		rnh = rt_tables_get_rnh(so->so_fibnum,
564 		    info.rti_info[RTAX_DST]->sa_family);
565 		if (rnh == NULL)
566 			senderr(EAFNOSUPPORT);
567 		RADIX_NODE_HEAD_RLOCK(rnh);
568 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
569 			info.rti_info[RTAX_NETMASK], rnh);
570 		if (rt == NULL) {	/* XXX looks bogus */
571 			RADIX_NODE_HEAD_RUNLOCK(rnh);
572 			senderr(ESRCH);
573 		}
574 #ifdef RADIX_MPATH
575 		/*
576 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
577 		 * we require users to specify a matching RTAX_GATEWAY.
578 		 *
579 		 * for RTM_GET, gate is optional even with multipath.
580 		 * if gate == NULL the first match is returned.
581 		 * (no need to call rt_mpath_matchgate if gate == NULL)
582 		 */
583 		if (rn_mpath_capable(rnh) &&
584 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
585 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
586 			if (!rt) {
587 				RADIX_NODE_HEAD_RUNLOCK(rnh);
588 				senderr(ESRCH);
589 			}
590 		}
591 #endif
592 		RT_LOCK(rt);
593 		RT_ADDREF(rt);
594 		RADIX_NODE_HEAD_RUNLOCK(rnh);
595 
596 		/*
597 		 * Fix for PR: 82974
598 		 *
599 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
600 		 * returns a perfect match in case a netmask is
601 		 * specified.  For host routes only a longest prefix
602 		 * match is returned so it is necessary to compare the
603 		 * existence of the netmask.  If both have a netmask
604 		 * rnh_lookup() did a perfect match and if none of them
605 		 * have a netmask both are host routes which is also a
606 		 * perfect match.
607 		 */
608 
609 		if (rtm->rtm_type != RTM_GET &&
610 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
611 			RT_UNLOCK(rt);
612 			senderr(ESRCH);
613 		}
614 
615 		switch(rtm->rtm_type) {
616 
617 		case RTM_GET:
618 		report:
619 			RT_LOCK_ASSERT(rt);
620 			if ((rt->rt_flags & RTF_HOST) == 0
621 			    ? jailed(curthread->td_ucred)
622 			    : prison_if(curthread->td_ucred,
623 			    rt_key(rt)) != 0) {
624 				RT_UNLOCK(rt);
625 				senderr(ESRCH);
626 			}
627 			info.rti_info[RTAX_DST] = rt_key(rt);
628 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
629 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
630 			info.rti_info[RTAX_GENMASK] = 0;
631 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
632 				ifp = rt->rt_ifp;
633 				if (ifp) {
634 					info.rti_info[RTAX_IFP] =
635 					    ifp->if_addr->ifa_addr;
636 					error = rtm_get_jailed(&info, ifp, rt,
637 					    &saun, curthread->td_ucred);
638 					if (error != 0) {
639 						RT_UNLOCK(rt);
640 						senderr(error);
641 					}
642 					if (ifp->if_flags & IFF_POINTOPOINT)
643 						info.rti_info[RTAX_BRD] =
644 						    rt->rt_ifa->ifa_dstaddr;
645 					rtm->rtm_index = ifp->if_index;
646 				} else {
647 					info.rti_info[RTAX_IFP] = NULL;
648 					info.rti_info[RTAX_IFA] = NULL;
649 				}
650 			} else if ((ifp = rt->rt_ifp) != NULL) {
651 				rtm->rtm_index = ifp->if_index;
652 			}
653 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
654 			if (len > rtm->rtm_msglen) {
655 				struct rt_msghdr *new_rtm;
656 				R_Malloc(new_rtm, struct rt_msghdr *, len);
657 				if (new_rtm == NULL) {
658 					RT_UNLOCK(rt);
659 					senderr(ENOBUFS);
660 				}
661 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
662 				Free(rtm); rtm = new_rtm;
663 			}
664 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
665 			rtm->rtm_flags = rt->rt_flags;
666 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
667 			rtm->rtm_addrs = info.rti_addrs;
668 			break;
669 
670 		case RTM_CHANGE:
671 			/*
672 			 * New gateway could require new ifaddr, ifp;
673 			 * flags may also be different; ifp may be specified
674 			 * by ll sockaddr when protocol address is ambiguous
675 			 */
676 			if (((rt->rt_flags & RTF_GATEWAY) &&
677 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
678 			    info.rti_info[RTAX_IFP] != NULL ||
679 			    (info.rti_info[RTAX_IFA] != NULL &&
680 			     !sa_equal(info.rti_info[RTAX_IFA],
681 				       rt->rt_ifa->ifa_addr))) {
682 				RT_UNLOCK(rt);
683 				RADIX_NODE_HEAD_LOCK(rnh);
684 				error = rt_getifa_fib(&info, rt->rt_fibnum);
685 				RADIX_NODE_HEAD_UNLOCK(rnh);
686 				if (error != 0)
687 					senderr(error);
688 				RT_LOCK(rt);
689 			}
690 			if (info.rti_ifa != NULL &&
691 			    info.rti_ifa != rt->rt_ifa &&
692 			    rt->rt_ifa != NULL &&
693 			    rt->rt_ifa->ifa_rtrequest != NULL) {
694 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
695 				    &info);
696 				IFAFREE(rt->rt_ifa);
697 			}
698 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
699 				RT_UNLOCK(rt);
700 				RADIX_NODE_HEAD_LOCK(rnh);
701 				RT_LOCK(rt);
702 
703 				error = rt_setgate(rt, rt_key(rt),
704 				    info.rti_info[RTAX_GATEWAY]);
705 				RADIX_NODE_HEAD_UNLOCK(rnh);
706 				if (error != 0) {
707 					RT_UNLOCK(rt);
708 					senderr(error);
709 				}
710 				rt->rt_flags |= RTF_GATEWAY;
711 			}
712 			if (info.rti_ifa != NULL &&
713 			    info.rti_ifa != rt->rt_ifa) {
714 				IFAREF(info.rti_ifa);
715 				rt->rt_ifa = info.rti_ifa;
716 				rt->rt_ifp = info.rti_ifp;
717 			}
718 			/* Allow some flags to be toggled on change. */
719 			rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
720 				    (rtm->rtm_flags & RTF_FMASK);
721 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
722 					&rt->rt_rmx);
723 			rtm->rtm_index = rt->rt_ifp->if_index;
724 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
725 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
726 			/* FALLTHROUGH */
727 		case RTM_LOCK:
728 			/* We don't support locks anymore */
729 			break;
730 		}
731 		RT_UNLOCK(rt);
732 		break;
733 
734 	default:
735 		senderr(EOPNOTSUPP);
736 	}
737 
738 flush:
739 	if (rtm) {
740 		if (error)
741 			rtm->rtm_errno = error;
742 		else
743 			rtm->rtm_flags |= RTF_DONE;
744 	}
745 	if (rt)		/* XXX can this be true? */
746 		RTFREE(rt);
747     {
748 	struct rawcb *rp = NULL;
749 	/*
750 	 * Check to see if we don't want our own messages.
751 	 */
752 	if ((so->so_options & SO_USELOOPBACK) == 0) {
753 		if (route_cb.any_count <= 1) {
754 			if (rtm)
755 				Free(rtm);
756 			m_freem(m);
757 			return (error);
758 		}
759 		/* There is another listener, so construct message */
760 		rp = sotorawcb(so);
761 	}
762 	if (rtm) {
763 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
764 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
765 			m_freem(m);
766 			m = NULL;
767 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
768 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
769 		Free(rtm);
770 	}
771 	if (m) {
772 		if (rp) {
773 			/*
774 			 * XXX insure we don't get a copy by
775 			 * invalidating our protocol
776 			 */
777 			unsigned short family = rp->rcb_proto.sp_family;
778 			rp->rcb_proto.sp_family = 0;
779 			rt_dispatch(m, info.rti_info[RTAX_DST]);
780 			rp->rcb_proto.sp_family = family;
781 		} else
782 			rt_dispatch(m, info.rti_info[RTAX_DST]);
783 	}
784     }
785 	return (error);
786 #undef	sa_equal
787 }
788 
789 static void
790 rt_setmetrics(u_long which, const struct rt_metrics *in,
791 	struct rt_metrics_lite *out)
792 {
793 #define metric(f, e) if (which & (f)) out->e = in->e;
794 	/*
795 	 * Only these are stored in the routing entry since introduction
796 	 * of tcp hostcache. The rest is ignored.
797 	 */
798 	metric(RTV_MTU, rmx_mtu);
799 	metric(RTV_WEIGHT, rmx_weight);
800 	/* Userland -> kernel timebase conversion. */
801 	if (which & RTV_EXPIRE)
802 		out->rmx_expire = in->rmx_expire ?
803 		    in->rmx_expire - time_second + time_uptime : 0;
804 #undef metric
805 }
806 
807 static void
808 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
809 {
810 #define metric(e) out->e = in->e;
811 	bzero(out, sizeof(*out));
812 	metric(rmx_mtu);
813 	metric(rmx_weight);
814 	/* Kernel -> userland timebase conversion. */
815 	out->rmx_expire = in->rmx_expire ?
816 	    in->rmx_expire - time_uptime + time_second : 0;
817 #undef metric
818 }
819 
820 /*
821  * Extract the addresses of the passed sockaddrs.
822  * Do a little sanity checking so as to avoid bad memory references.
823  * This data is derived straight from userland.
824  */
825 static int
826 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
827 {
828 	struct sockaddr *sa;
829 	int i;
830 
831 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
832 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
833 			continue;
834 		sa = (struct sockaddr *)cp;
835 		/*
836 		 * It won't fit.
837 		 */
838 		if (cp + sa->sa_len > cplim)
839 			return (EINVAL);
840 		/*
841 		 * there are no more.. quit now
842 		 * If there are more bits, they are in error.
843 		 * I've seen this. route(1) can evidently generate these.
844 		 * This causes kernel to core dump.
845 		 * for compatibility, If we see this, point to a safe address.
846 		 */
847 		if (sa->sa_len == 0) {
848 			rtinfo->rti_info[i] = &sa_zero;
849 			return (0); /* should be EINVAL but for compat */
850 		}
851 		/* accept it */
852 		rtinfo->rti_info[i] = sa;
853 		cp += SA_SIZE(sa);
854 	}
855 	return (0);
856 }
857 
858 static struct mbuf *
859 rt_msg1(int type, struct rt_addrinfo *rtinfo)
860 {
861 	struct rt_msghdr *rtm;
862 	struct mbuf *m;
863 	int i;
864 	struct sockaddr *sa;
865 	int len, dlen;
866 
867 	switch (type) {
868 
869 	case RTM_DELADDR:
870 	case RTM_NEWADDR:
871 		len = sizeof(struct ifa_msghdr);
872 		break;
873 
874 	case RTM_DELMADDR:
875 	case RTM_NEWMADDR:
876 		len = sizeof(struct ifma_msghdr);
877 		break;
878 
879 	case RTM_IFINFO:
880 		len = sizeof(struct if_msghdr);
881 		break;
882 
883 	case RTM_IFANNOUNCE:
884 	case RTM_IEEE80211:
885 		len = sizeof(struct if_announcemsghdr);
886 		break;
887 
888 	default:
889 		len = sizeof(struct rt_msghdr);
890 	}
891 	if (len > MCLBYTES)
892 		panic("rt_msg1");
893 	m = m_gethdr(M_DONTWAIT, MT_DATA);
894 	if (m && len > MHLEN) {
895 		MCLGET(m, M_DONTWAIT);
896 		if ((m->m_flags & M_EXT) == 0) {
897 			m_free(m);
898 			m = NULL;
899 		}
900 	}
901 	if (m == NULL)
902 		return (m);
903 	m->m_pkthdr.len = m->m_len = len;
904 	m->m_pkthdr.rcvif = NULL;
905 	rtm = mtod(m, struct rt_msghdr *);
906 	bzero((caddr_t)rtm, len);
907 	for (i = 0; i < RTAX_MAX; i++) {
908 		if ((sa = rtinfo->rti_info[i]) == NULL)
909 			continue;
910 		rtinfo->rti_addrs |= (1 << i);
911 		dlen = SA_SIZE(sa);
912 		m_copyback(m, len, dlen, (caddr_t)sa);
913 		len += dlen;
914 	}
915 	if (m->m_pkthdr.len != len) {
916 		m_freem(m);
917 		return (NULL);
918 	}
919 	rtm->rtm_msglen = len;
920 	rtm->rtm_version = RTM_VERSION;
921 	rtm->rtm_type = type;
922 	return (m);
923 }
924 
925 static int
926 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
927 {
928 	int i;
929 	int len, dlen, second_time = 0;
930 	caddr_t cp0;
931 
932 	rtinfo->rti_addrs = 0;
933 again:
934 	switch (type) {
935 
936 	case RTM_DELADDR:
937 	case RTM_NEWADDR:
938 		len = sizeof(struct ifa_msghdr);
939 		break;
940 
941 	case RTM_IFINFO:
942 		len = sizeof(struct if_msghdr);
943 		break;
944 
945 	case RTM_NEWMADDR:
946 		len = sizeof(struct ifma_msghdr);
947 		break;
948 
949 	default:
950 		len = sizeof(struct rt_msghdr);
951 	}
952 	cp0 = cp;
953 	if (cp0)
954 		cp += len;
955 	for (i = 0; i < RTAX_MAX; i++) {
956 		struct sockaddr *sa;
957 
958 		if ((sa = rtinfo->rti_info[i]) == NULL)
959 			continue;
960 		rtinfo->rti_addrs |= (1 << i);
961 		dlen = SA_SIZE(sa);
962 		if (cp) {
963 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
964 			cp += dlen;
965 		}
966 		len += dlen;
967 	}
968 	len = ALIGN(len);
969 	if (cp == NULL && w != NULL && !second_time) {
970 		struct walkarg *rw = w;
971 
972 		if (rw->w_req) {
973 			if (rw->w_tmemsize < len) {
974 				if (rw->w_tmem)
975 					free(rw->w_tmem, M_RTABLE);
976 				rw->w_tmem = (caddr_t)
977 					malloc(len, M_RTABLE, M_NOWAIT);
978 				if (rw->w_tmem)
979 					rw->w_tmemsize = len;
980 			}
981 			if (rw->w_tmem) {
982 				cp = rw->w_tmem;
983 				second_time = 1;
984 				goto again;
985 			}
986 		}
987 	}
988 	if (cp) {
989 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
990 
991 		rtm->rtm_version = RTM_VERSION;
992 		rtm->rtm_type = type;
993 		rtm->rtm_msglen = len;
994 	}
995 	return (len);
996 }
997 
998 /*
999  * This routine is called to generate a message from the routing
1000  * socket indicating that a redirect has occured, a routing lookup
1001  * has failed, or that a protocol has detected timeouts to a particular
1002  * destination.
1003  */
1004 void
1005 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1006 {
1007 	struct rt_msghdr *rtm;
1008 	struct mbuf *m;
1009 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1010 
1011 	if (route_cb.any_count == 0)
1012 		return;
1013 	m = rt_msg1(type, rtinfo);
1014 	if (m == NULL)
1015 		return;
1016 	rtm = mtod(m, struct rt_msghdr *);
1017 	rtm->rtm_flags = RTF_DONE | flags;
1018 	rtm->rtm_errno = error;
1019 	rtm->rtm_addrs = rtinfo->rti_addrs;
1020 	rt_dispatch(m, sa);
1021 }
1022 
1023 /*
1024  * This routine is called to generate a message from the routing
1025  * socket indicating that the status of a network interface has changed.
1026  */
1027 void
1028 rt_ifmsg(struct ifnet *ifp)
1029 {
1030 	struct if_msghdr *ifm;
1031 	struct mbuf *m;
1032 	struct rt_addrinfo info;
1033 
1034 	if (route_cb.any_count == 0)
1035 		return;
1036 	bzero((caddr_t)&info, sizeof(info));
1037 	m = rt_msg1(RTM_IFINFO, &info);
1038 	if (m == NULL)
1039 		return;
1040 	ifm = mtod(m, struct if_msghdr *);
1041 	ifm->ifm_index = ifp->if_index;
1042 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1043 	ifm->ifm_data = ifp->if_data;
1044 	ifm->ifm_addrs = 0;
1045 	rt_dispatch(m, NULL);
1046 }
1047 
1048 /*
1049  * This is called to generate messages from the routing socket
1050  * indicating a network interface has had addresses associated with it.
1051  * if we ever reverse the logic and replace messages TO the routing
1052  * socket indicate a request to configure interfaces, then it will
1053  * be unnecessary as the routing socket will automatically generate
1054  * copies of it.
1055  */
1056 void
1057 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1058 {
1059 	struct rt_addrinfo info;
1060 	struct sockaddr *sa = NULL;
1061 	int pass;
1062 	struct mbuf *m = NULL;
1063 	struct ifnet *ifp = ifa->ifa_ifp;
1064 
1065 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1066 		("unexpected cmd %u", cmd));
1067 #ifdef SCTP
1068 	/*
1069 	 * notify the SCTP stack
1070 	 * this will only get called when an address is added/deleted
1071 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1072 	 */
1073 	sctp_addr_change(ifa, cmd);
1074 #endif /* SCTP */
1075 	if (route_cb.any_count == 0)
1076 		return;
1077 	for (pass = 1; pass < 3; pass++) {
1078 		bzero((caddr_t)&info, sizeof(info));
1079 		if ((cmd == RTM_ADD && pass == 1) ||
1080 		    (cmd == RTM_DELETE && pass == 2)) {
1081 			struct ifa_msghdr *ifam;
1082 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1083 
1084 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1085 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1086 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1087 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1088 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1089 				continue;
1090 			ifam = mtod(m, struct ifa_msghdr *);
1091 			ifam->ifam_index = ifp->if_index;
1092 			ifam->ifam_metric = ifa->ifa_metric;
1093 			ifam->ifam_flags = ifa->ifa_flags;
1094 			ifam->ifam_addrs = info.rti_addrs;
1095 		}
1096 		if ((cmd == RTM_ADD && pass == 2) ||
1097 		    (cmd == RTM_DELETE && pass == 1)) {
1098 			struct rt_msghdr *rtm;
1099 
1100 			if (rt == NULL)
1101 				continue;
1102 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1103 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
1104 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1105 			if ((m = rt_msg1(cmd, &info)) == NULL)
1106 				continue;
1107 			rtm = mtod(m, struct rt_msghdr *);
1108 			rtm->rtm_index = ifp->if_index;
1109 			rtm->rtm_flags |= rt->rt_flags;
1110 			rtm->rtm_errno = error;
1111 			rtm->rtm_addrs = info.rti_addrs;
1112 		}
1113 		rt_dispatch(m, sa);
1114 	}
1115 }
1116 
1117 /*
1118  * This is the analogue to the rt_newaddrmsg which performs the same
1119  * function but for multicast group memberhips.  This is easier since
1120  * there is no route state to worry about.
1121  */
1122 void
1123 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1124 {
1125 	struct rt_addrinfo info;
1126 	struct mbuf *m = NULL;
1127 	struct ifnet *ifp = ifma->ifma_ifp;
1128 	struct ifma_msghdr *ifmam;
1129 
1130 	if (route_cb.any_count == 0)
1131 		return;
1132 
1133 	bzero((caddr_t)&info, sizeof(info));
1134 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1135 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1136 	/*
1137 	 * If a link-layer address is present, present it as a ``gateway''
1138 	 * (similarly to how ARP entries, e.g., are presented).
1139 	 */
1140 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1141 	m = rt_msg1(cmd, &info);
1142 	if (m == NULL)
1143 		return;
1144 	ifmam = mtod(m, struct ifma_msghdr *);
1145 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1146 	    __func__));
1147 	ifmam->ifmam_index = ifp->if_index;
1148 	ifmam->ifmam_addrs = info.rti_addrs;
1149 	rt_dispatch(m, ifma->ifma_addr);
1150 }
1151 
1152 static struct mbuf *
1153 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1154 	struct rt_addrinfo *info)
1155 {
1156 	struct if_announcemsghdr *ifan;
1157 	struct mbuf *m;
1158 
1159 	if (route_cb.any_count == 0)
1160 		return NULL;
1161 	bzero((caddr_t)info, sizeof(*info));
1162 	m = rt_msg1(type, info);
1163 	if (m != NULL) {
1164 		ifan = mtod(m, struct if_announcemsghdr *);
1165 		ifan->ifan_index = ifp->if_index;
1166 		strlcpy(ifan->ifan_name, ifp->if_xname,
1167 			sizeof(ifan->ifan_name));
1168 		ifan->ifan_what = what;
1169 	}
1170 	return m;
1171 }
1172 
1173 /*
1174  * This is called to generate routing socket messages indicating
1175  * IEEE80211 wireless events.
1176  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1177  */
1178 void
1179 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1180 {
1181 	struct mbuf *m;
1182 	struct rt_addrinfo info;
1183 
1184 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1185 	if (m != NULL) {
1186 		/*
1187 		 * Append the ieee80211 data.  Try to stick it in the
1188 		 * mbuf containing the ifannounce msg; otherwise allocate
1189 		 * a new mbuf and append.
1190 		 *
1191 		 * NB: we assume m is a single mbuf.
1192 		 */
1193 		if (data_len > M_TRAILINGSPACE(m)) {
1194 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1195 			if (n == NULL) {
1196 				m_freem(m);
1197 				return;
1198 			}
1199 			bcopy(data, mtod(n, void *), data_len);
1200 			n->m_len = data_len;
1201 			m->m_next = n;
1202 		} else if (data_len > 0) {
1203 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1204 			m->m_len += data_len;
1205 		}
1206 		if (m->m_flags & M_PKTHDR)
1207 			m->m_pkthdr.len += data_len;
1208 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1209 		rt_dispatch(m, NULL);
1210 	}
1211 }
1212 
1213 /*
1214  * This is called to generate routing socket messages indicating
1215  * network interface arrival and departure.
1216  */
1217 void
1218 rt_ifannouncemsg(struct ifnet *ifp, int what)
1219 {
1220 	struct mbuf *m;
1221 	struct rt_addrinfo info;
1222 
1223 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1224 	if (m != NULL)
1225 		rt_dispatch(m, NULL);
1226 }
1227 
1228 static void
1229 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1230 {
1231 	INIT_VNET_NET(curvnet);
1232 	struct m_tag *tag;
1233 
1234 	/*
1235 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1236 	 * use when injecting the mbuf into the routing socket buffer from
1237 	 * the netisr.
1238 	 */
1239 	if (sa != NULL) {
1240 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1241 		    M_NOWAIT);
1242 		if (tag == NULL) {
1243 			m_freem(m);
1244 			return;
1245 		}
1246 		*(unsigned short *)(tag + 1) = sa->sa_family;
1247 		m_tag_prepend(m, tag);
1248 	}
1249 #ifdef VIMAGE
1250 	if (V_loif)
1251 		m->m_pkthdr.rcvif = V_loif;
1252 	else {
1253 		m_freem(m);
1254 		return;
1255 	}
1256 #endif
1257 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1258 }
1259 
1260 /*
1261  * This is used in dumping the kernel table via sysctl().
1262  */
1263 static int
1264 sysctl_dumpentry(struct radix_node *rn, void *vw)
1265 {
1266 	struct walkarg *w = vw;
1267 	struct rtentry *rt = (struct rtentry *)rn;
1268 	int error = 0, size;
1269 	struct rt_addrinfo info;
1270 
1271 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1272 		return 0;
1273 	if ((rt->rt_flags & RTF_HOST) == 0
1274 	    ? jailed(w->w_req->td->td_ucred)
1275 	    : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1276 		return (0);
1277 	bzero((caddr_t)&info, sizeof(info));
1278 	info.rti_info[RTAX_DST] = rt_key(rt);
1279 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1280 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1281 	info.rti_info[RTAX_GENMASK] = 0;
1282 	if (rt->rt_ifp) {
1283 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1284 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1285 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1286 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1287 	}
1288 	size = rt_msg2(RTM_GET, &info, NULL, w);
1289 	if (w->w_req && w->w_tmem) {
1290 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1291 
1292 		rtm->rtm_flags = rt->rt_flags;
1293 		/*
1294 		 * let's be honest about this being a retarded hack
1295 		 */
1296 		rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1297 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1298 		rtm->rtm_index = rt->rt_ifp->if_index;
1299 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1300 		rtm->rtm_addrs = info.rti_addrs;
1301 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1302 		return (error);
1303 	}
1304 	return (error);
1305 }
1306 
1307 static int
1308 sysctl_iflist(int af, struct walkarg *w)
1309 {
1310 	INIT_VNET_NET(curvnet);
1311 	struct ifnet *ifp;
1312 	struct ifaddr *ifa;
1313 	struct rt_addrinfo info;
1314 	int len, error = 0;
1315 
1316 	bzero((caddr_t)&info, sizeof(info));
1317 	IFNET_RLOCK();
1318 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1319 		if (w->w_arg && w->w_arg != ifp->if_index)
1320 			continue;
1321 		ifa = ifp->if_addr;
1322 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1323 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1324 		info.rti_info[RTAX_IFP] = NULL;
1325 		if (w->w_req && w->w_tmem) {
1326 			struct if_msghdr *ifm;
1327 
1328 			ifm = (struct if_msghdr *)w->w_tmem;
1329 			ifm->ifm_index = ifp->if_index;
1330 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1331 			ifm->ifm_data = ifp->if_data;
1332 			ifm->ifm_addrs = info.rti_addrs;
1333 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
1334 			if (error)
1335 				goto done;
1336 		}
1337 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1338 			if (af && af != ifa->ifa_addr->sa_family)
1339 				continue;
1340 			if (prison_if(w->w_req->td->td_ucred,
1341 			    ifa->ifa_addr) != 0)
1342 				continue;
1343 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1344 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1345 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1346 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1347 			if (w->w_req && w->w_tmem) {
1348 				struct ifa_msghdr *ifam;
1349 
1350 				ifam = (struct ifa_msghdr *)w->w_tmem;
1351 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1352 				ifam->ifam_flags = ifa->ifa_flags;
1353 				ifam->ifam_metric = ifa->ifa_metric;
1354 				ifam->ifam_addrs = info.rti_addrs;
1355 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1356 				if (error)
1357 					goto done;
1358 			}
1359 		}
1360 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1361 			info.rti_info[RTAX_BRD] = NULL;
1362 	}
1363 done:
1364 	IFNET_RUNLOCK();
1365 	return (error);
1366 }
1367 
1368 static int
1369 sysctl_ifmalist(int af, struct walkarg *w)
1370 {
1371 	INIT_VNET_NET(curvnet);
1372 	struct ifnet *ifp;
1373 	struct ifmultiaddr *ifma;
1374 	struct	rt_addrinfo info;
1375 	int	len, error = 0;
1376 	struct ifaddr *ifa;
1377 
1378 	bzero((caddr_t)&info, sizeof(info));
1379 	IFNET_RLOCK();
1380 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1381 		if (w->w_arg && w->w_arg != ifp->if_index)
1382 			continue;
1383 		ifa = ifp->if_addr;
1384 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1385 		IF_ADDR_LOCK(ifp);
1386 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1387 			if (af && af != ifma->ifma_addr->sa_family)
1388 				continue;
1389 			if (prison_if(w->w_req->td->td_ucred,
1390 			    ifma->ifma_addr) != 0)
1391 				continue;
1392 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1393 			info.rti_info[RTAX_GATEWAY] =
1394 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1395 			    ifma->ifma_lladdr : NULL;
1396 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1397 			if (w->w_req && w->w_tmem) {
1398 				struct ifma_msghdr *ifmam;
1399 
1400 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1401 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1402 				ifmam->ifmam_flags = 0;
1403 				ifmam->ifmam_addrs = info.rti_addrs;
1404 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1405 				if (error) {
1406 					IF_ADDR_UNLOCK(ifp);
1407 					goto done;
1408 				}
1409 			}
1410 		}
1411 		IF_ADDR_UNLOCK(ifp);
1412 	}
1413 done:
1414 	IFNET_RUNLOCK();
1415 	return (error);
1416 }
1417 
1418 static int
1419 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1420 {
1421 	int	*name = (int *)arg1;
1422 	u_int	namelen = arg2;
1423 	struct radix_node_head *rnh = NULL; /* silence compiler. */
1424 	int	i, lim, error = EINVAL;
1425 	u_char	af;
1426 	struct	walkarg w;
1427 
1428 	name ++;
1429 	namelen--;
1430 	if (req->newptr)
1431 		return (EPERM);
1432 	if (namelen != 3)
1433 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1434 	af = name[0];
1435 	if (af > AF_MAX)
1436 		return (EINVAL);
1437 	bzero(&w, sizeof(w));
1438 	w.w_op = name[1];
1439 	w.w_arg = name[2];
1440 	w.w_req = req;
1441 
1442 	error = sysctl_wire_old_buffer(req, 0);
1443 	if (error)
1444 		return (error);
1445 	switch (w.w_op) {
1446 
1447 	case NET_RT_DUMP:
1448 	case NET_RT_FLAGS:
1449 		if (af == 0) {			/* dump all tables */
1450 			i = 1;
1451 			lim = AF_MAX;
1452 		} else				/* dump only one table */
1453 			i = lim = af;
1454 
1455 		/*
1456 		 * take care of llinfo entries, the caller must
1457 		 * specify an AF
1458 		 */
1459 		if (w.w_op == NET_RT_FLAGS &&
1460 		    (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1461 			if (af != 0)
1462 				error = lltable_sysctl_dumparp(af, w.w_req);
1463 			else
1464 				error = EINVAL;
1465 			break;
1466 		}
1467 		/*
1468 		 * take care of routing entries
1469 		 */
1470 		for (error = 0; error == 0 && i <= lim; i++)
1471 			rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1472 			if (rnh != NULL) {
1473 				RADIX_NODE_HEAD_LOCK(rnh);
1474 			    	error = rnh->rnh_walktree(rnh,
1475 				    sysctl_dumpentry, &w);
1476 				RADIX_NODE_HEAD_UNLOCK(rnh);
1477 			} else if (af != 0)
1478 				error = EAFNOSUPPORT;
1479 		break;
1480 
1481 	case NET_RT_IFLIST:
1482 		error = sysctl_iflist(af, &w);
1483 		break;
1484 
1485 	case NET_RT_IFMALIST:
1486 		error = sysctl_ifmalist(af, &w);
1487 		break;
1488 	}
1489 	if (w.w_tmem)
1490 		free(w.w_tmem, M_RTABLE);
1491 	return (error);
1492 }
1493 
1494 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1495 
1496 /*
1497  * Definitions of protocols supported in the ROUTE domain.
1498  */
1499 
1500 static struct domain routedomain;		/* or at least forward */
1501 
1502 static struct protosw routesw[] = {
1503 {
1504 	.pr_type =		SOCK_RAW,
1505 	.pr_domain =		&routedomain,
1506 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1507 	.pr_output =		route_output,
1508 	.pr_ctlinput =		raw_ctlinput,
1509 	.pr_init =		raw_init,
1510 	.pr_usrreqs =		&route_usrreqs
1511 }
1512 };
1513 
1514 static struct domain routedomain = {
1515 	.dom_family =		PF_ROUTE,
1516 	.dom_name =		 "route",
1517 	.dom_protosw =		routesw,
1518 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1519 };
1520 
1521 DOMAIN_SET(route);
1522