xref: /freebsd/sys/net/rtsock.c (revision e1fe3dba5ce2826061f6489765be9b4a341736a9)
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 
33 #include <sys/param.h>
34 #include <sys/domain.h>
35 #include <sys/kernel.h>
36 #include <sys/jail.h>
37 #include <sys/malloc.h>
38 #include <sys/mbuf.h>
39 #include <sys/proc.h>
40 #include <sys/protosw.h>
41 #include <sys/signalvar.h>
42 #include <sys/socket.h>
43 #include <sys/socketvar.h>
44 #include <sys/sysctl.h>
45 #include <sys/systm.h>
46 
47 #include <net/if.h>
48 #include <net/netisr.h>
49 #include <net/raw_cb.h>
50 #include <net/route.h>
51 
52 #include <netinet/in.h>
53 
54 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
55 
56 /* NB: these are not modified */
57 static struct	sockaddr route_dst = { 2, PF_ROUTE, };
58 static struct	sockaddr route_src = { 2, PF_ROUTE, };
59 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
60 
61 static struct {
62 	int	ip_count;	/* attached w/ AF_INET */
63 	int	ip6_count;	/* attached w/ AF_INET6 */
64 	int	ipx_count;	/* attached w/ AF_IPX */
65 	int	any_count;	/* total attached */
66 } route_cb;
67 
68 struct mtx rtsock_mtx;
69 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
70 
71 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
72 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
73 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
74 
75 static struct	ifqueue rtsintrq;
76 
77 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
78 SYSCTL_INT(_net_route, OID_AUTO, netisr_maxqlen, CTLFLAG_RW,
79     &rtsintrq.ifq_maxlen, 0, "maximum routing socket dispatch queue length");
80 
81 struct walkarg {
82 	int	w_tmemsize;
83 	int	w_op, w_arg;
84 	caddr_t	w_tmem;
85 	struct sysctl_req *w_req;
86 };
87 
88 static void	rts_input(struct mbuf *m);
89 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
90 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
91 			caddr_t cp, struct walkarg *w);
92 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
93 			struct rt_addrinfo *rtinfo);
94 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
95 static int	sysctl_iflist(int af, struct walkarg *w);
96 static int	sysctl_ifmalist(int af, struct walkarg *w);
97 static int	route_output(struct mbuf *m, struct socket *so);
98 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
99 			struct rt_metrics_lite *out);
100 static void	rt_getmetrics(const struct rt_metrics_lite *in,
101 			struct rt_metrics *out);
102 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
103 
104 static void
105 rts_init(void)
106 {
107 	int tmp;
108 
109 	rtsintrq.ifq_maxlen = 256;
110 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
111 		rtsintrq.ifq_maxlen = tmp;
112 	mtx_init(&rtsintrq.ifq_mtx, "rts_inq", NULL, MTX_DEF);
113 	netisr_register(NETISR_ROUTE, rts_input, &rtsintrq, NETISR_MPSAFE);
114 }
115 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0)
116 
117 static void
118 rts_input(struct mbuf *m)
119 {
120 	struct sockproto route_proto;
121 	unsigned short *family;
122 	struct m_tag *tag;
123 
124 	route_proto.sp_family = PF_ROUTE;
125 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
126 	if (tag != NULL) {
127 		family = (unsigned short *)(tag + 1);
128 		route_proto.sp_protocol = *family;
129 		m_tag_delete(m, tag);
130 	} else
131 		route_proto.sp_protocol = 0;
132 
133 	raw_input(m, &route_proto, &route_src, &route_dst);
134 }
135 
136 /*
137  * It really doesn't make any sense at all for this code to share much
138  * with raw_usrreq.c, since its functionality is so restricted.  XXX
139  */
140 static int
141 rts_abort(struct socket *so)
142 {
143 
144 	return (raw_usrreqs.pru_abort(so));
145 }
146 
147 /* pru_accept is EOPNOTSUPP */
148 
149 static int
150 rts_attach(struct socket *so, int proto, struct thread *td)
151 {
152 	struct rawcb *rp;
153 	int s, error;
154 
155 	if (sotorawcb(so) != NULL)
156 		return EISCONN;	/* XXX panic? */
157 	/* XXX */
158 	MALLOC(rp, struct rawcb *, sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
159 	if (rp == NULL)
160 		return ENOBUFS;
161 
162 	/*
163 	 * The splnet() is necessary to block protocols from sending
164 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
165 	 * this PCB is extant but incompletely initialized.
166 	 * Probably we should try to do more of this work beforehand and
167 	 * eliminate the spl.
168 	 */
169 	s = splnet();
170 	so->so_pcb = (caddr_t)rp;
171 	error = raw_attach(so, proto);
172 	rp = sotorawcb(so);
173 	if (error) {
174 		splx(s);
175 		so->so_pcb = NULL;
176 		free(rp, M_PCB);
177 		return error;
178 	}
179 	RTSOCK_LOCK();
180 	switch(rp->rcb_proto.sp_protocol) {
181 	case AF_INET:
182 		route_cb.ip_count++;
183 		break;
184 	case AF_INET6:
185 		route_cb.ip6_count++;
186 		break;
187 	case AF_IPX:
188 		route_cb.ipx_count++;
189 		break;
190 	}
191 	rp->rcb_faddr = &route_src;
192 	route_cb.any_count++;
193 	RTSOCK_UNLOCK();
194 	soisconnected(so);
195 	so->so_options |= SO_USELOOPBACK;
196 	splx(s);
197 	return 0;
198 }
199 
200 static int
201 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
202 {
203 
204 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
205 }
206 
207 static int
208 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
209 {
210 
211 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
212 }
213 
214 /* pru_connect2 is EOPNOTSUPP */
215 /* pru_control is EOPNOTSUPP */
216 
217 static int
218 rts_detach(struct socket *so)
219 {
220 	struct rawcb *rp = sotorawcb(so);
221 	int s, error;
222 
223 	s = splnet();
224 	if (rp != NULL) {
225 		RTSOCK_LOCK();
226 		switch(rp->rcb_proto.sp_protocol) {
227 		case AF_INET:
228 			route_cb.ip_count--;
229 			break;
230 		case AF_INET6:
231 			route_cb.ip6_count--;
232 			break;
233 		case AF_IPX:
234 			route_cb.ipx_count--;
235 			break;
236 		}
237 		route_cb.any_count--;
238 		RTSOCK_UNLOCK();
239 	}
240 	error = raw_usrreqs.pru_detach(so);
241 	splx(s);
242 	return error;
243 }
244 
245 static int
246 rts_disconnect(struct socket *so)
247 {
248 
249 	return (raw_usrreqs.pru_disconnect(so));
250 }
251 
252 /* pru_listen is EOPNOTSUPP */
253 
254 static int
255 rts_peeraddr(struct socket *so, struct sockaddr **nam)
256 {
257 
258 	return (raw_usrreqs.pru_peeraddr(so, nam));
259 }
260 
261 /* pru_rcvd is EOPNOTSUPP */
262 /* pru_rcvoob is EOPNOTSUPP */
263 
264 static int
265 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
266 	 struct mbuf *control, struct thread *td)
267 {
268 
269 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
270 }
271 
272 /* pru_sense is null */
273 
274 static int
275 rts_shutdown(struct socket *so)
276 {
277 
278 	return (raw_usrreqs.pru_shutdown(so));
279 }
280 
281 static int
282 rts_sockaddr(struct socket *so, struct sockaddr **nam)
283 {
284 
285 	return (raw_usrreqs.pru_sockaddr(so, nam));
286 }
287 
288 static struct pr_usrreqs route_usrreqs = {
289 	.pru_abort =		rts_abort,
290 	.pru_attach =		rts_attach,
291 	.pru_bind =		rts_bind,
292 	.pru_connect =		rts_connect,
293 	.pru_detach =		rts_detach,
294 	.pru_disconnect =	rts_disconnect,
295 	.pru_peeraddr =		rts_peeraddr,
296 	.pru_send =		rts_send,
297 	.pru_shutdown =		rts_shutdown,
298 	.pru_sockaddr =		rts_sockaddr,
299 };
300 
301 /*ARGSUSED*/
302 static int
303 route_output(struct mbuf *m, struct socket *so)
304 {
305 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
306 	struct rt_msghdr *rtm = NULL;
307 	struct rtentry *rt = NULL;
308 	struct radix_node_head *rnh;
309 	struct rt_addrinfo info;
310 	int len, error = 0;
311 	struct ifnet *ifp = NULL;
312 	struct ifaddr *ifa = NULL;
313 	struct sockaddr_in jail;
314 
315 #define senderr(e) { error = e; goto flush;}
316 	if (m == NULL || ((m->m_len < sizeof(long)) &&
317 		       (m = m_pullup(m, sizeof(long))) == NULL))
318 		return (ENOBUFS);
319 	if ((m->m_flags & M_PKTHDR) == 0)
320 		panic("route_output");
321 	len = m->m_pkthdr.len;
322 	if (len < sizeof(*rtm) ||
323 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
324 		info.rti_info[RTAX_DST] = NULL;
325 		senderr(EINVAL);
326 	}
327 	R_Malloc(rtm, struct rt_msghdr *, len);
328 	if (rtm == NULL) {
329 		info.rti_info[RTAX_DST] = NULL;
330 		senderr(ENOBUFS);
331 	}
332 	m_copydata(m, 0, len, (caddr_t)rtm);
333 	if (rtm->rtm_version != RTM_VERSION) {
334 		info.rti_info[RTAX_DST] = NULL;
335 		senderr(EPROTONOSUPPORT);
336 	}
337 	rtm->rtm_pid = curproc->p_pid;
338 	bzero(&info, sizeof(info));
339 	info.rti_addrs = rtm->rtm_addrs;
340 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
341 		info.rti_info[RTAX_DST] = NULL;
342 		senderr(EINVAL);
343 	}
344 	info.rti_flags = rtm->rtm_flags;
345 	if (info.rti_info[RTAX_DST] == NULL ||
346 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
347 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
348 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
349 		senderr(EINVAL);
350 	if (info.rti_info[RTAX_GENMASK]) {
351 		struct radix_node *t;
352 		t = rn_addmask((caddr_t) info.rti_info[RTAX_GENMASK], 0, 1);
353 		if (t != NULL &&
354 		    bcmp((char *)(void *)info.rti_info[RTAX_GENMASK] + 1,
355 		    (char *)(void *)t->rn_key + 1,
356 		    ((struct sockaddr *)t->rn_key)->sa_len - 1) == 0)
357 			info.rti_info[RTAX_GENMASK] =
358 			    (struct sockaddr *)t->rn_key;
359 		else
360 			senderr(ENOBUFS);
361 	}
362 
363 	/*
364 	 * Verify that the caller has the appropriate privilege; RTM_GET
365 	 * is the only operation the non-superuser is allowed.
366 	 */
367 	if (rtm->rtm_type != RTM_GET && (error = suser(curthread)) != 0)
368 		senderr(error);
369 
370 	switch (rtm->rtm_type) {
371 		struct rtentry *saved_nrt;
372 
373 	case RTM_ADD:
374 		if (info.rti_info[RTAX_GATEWAY] == NULL)
375 			senderr(EINVAL);
376 		saved_nrt = NULL;
377 		error = rtrequest1(RTM_ADD, &info, &saved_nrt);
378 		if (error == 0 && saved_nrt) {
379 			RT_LOCK(saved_nrt);
380 			rt_setmetrics(rtm->rtm_inits,
381 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
382 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
383 			RT_REMREF(saved_nrt);
384 			saved_nrt->rt_genmask = info.rti_info[RTAX_GENMASK];
385 			RT_UNLOCK(saved_nrt);
386 		}
387 		break;
388 
389 	case RTM_DELETE:
390 		saved_nrt = NULL;
391 		error = rtrequest1(RTM_DELETE, &info, &saved_nrt);
392 		if (error == 0) {
393 			RT_LOCK(saved_nrt);
394 			rt = saved_nrt;
395 			goto report;
396 		}
397 		break;
398 
399 	case RTM_GET:
400 	case RTM_CHANGE:
401 	case RTM_LOCK:
402 		rnh = rt_tables[info.rti_info[RTAX_DST]->sa_family];
403 		if (rnh == NULL)
404 			senderr(EAFNOSUPPORT);
405 		RADIX_NODE_HEAD_LOCK(rnh);
406 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
407 			info.rti_info[RTAX_NETMASK], rnh);
408 		if (rt == NULL) {	/* XXX looks bogus */
409 			RADIX_NODE_HEAD_UNLOCK(rnh);
410 			senderr(ESRCH);
411 		}
412 		RT_LOCK(rt);
413 		RT_ADDREF(rt);
414 		RADIX_NODE_HEAD_UNLOCK(rnh);
415 
416 		/*
417 		 * Fix for PR: 82974
418 		 *
419 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
420 		 * returns a perfect match in case a netmask is
421 		 * specified.  For host routes only a longest prefix
422 		 * match is returned so it is necessary to compare the
423 		 * existence of the netmask.  If both have a netmask
424 		 * rnh_lookup() did a perfect match and if none of them
425 		 * have a netmask both are host routes which is also a
426 		 * perfect match.
427 		 */
428 
429 		if (rtm->rtm_type != RTM_GET &&
430 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
431 			RT_UNLOCK(rt);
432 			senderr(ESRCH);
433 		}
434 
435 		switch(rtm->rtm_type) {
436 
437 		case RTM_GET:
438 		report:
439 			RT_LOCK_ASSERT(rt);
440 			info.rti_info[RTAX_DST] = rt_key(rt);
441 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
442 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
443 			info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
444 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
445 				ifp = rt->rt_ifp;
446 				if (ifp) {
447 					info.rti_info[RTAX_IFP] =
448 					    ifp->if_addr->ifa_addr;
449 					if (jailed(so->so_cred)) {
450 						bzero(&jail, sizeof(jail));
451 						jail.sin_family = PF_INET;
452 						jail.sin_len = sizeof(jail);
453 						jail.sin_addr.s_addr =
454 						htonl(prison_getip(so->so_cred));
455 						info.rti_info[RTAX_IFA] =
456 						    (struct sockaddr *)&jail;
457 					} else
458 						info.rti_info[RTAX_IFA] =
459 						    rt->rt_ifa->ifa_addr;
460 					if (ifp->if_flags & IFF_POINTOPOINT)
461 						info.rti_info[RTAX_BRD] =
462 						    rt->rt_ifa->ifa_dstaddr;
463 					rtm->rtm_index = ifp->if_index;
464 				} else {
465 					info.rti_info[RTAX_IFP] = NULL;
466 					info.rti_info[RTAX_IFA] = NULL;
467 				}
468 			} else if ((ifp = rt->rt_ifp) != NULL) {
469 				rtm->rtm_index = ifp->if_index;
470 			}
471 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
472 			if (len > rtm->rtm_msglen) {
473 				struct rt_msghdr *new_rtm;
474 				R_Malloc(new_rtm, struct rt_msghdr *, len);
475 				if (new_rtm == NULL) {
476 					RT_UNLOCK(rt);
477 					senderr(ENOBUFS);
478 				}
479 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
480 				Free(rtm); rtm = new_rtm;
481 			}
482 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
483 			rtm->rtm_flags = rt->rt_flags;
484 			rtm->rtm_use = 0;
485 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
486 			rtm->rtm_addrs = info.rti_addrs;
487 			break;
488 
489 		case RTM_CHANGE:
490 			/*
491 			 * New gateway could require new ifaddr, ifp;
492 			 * flags may also be different; ifp may be specified
493 			 * by ll sockaddr when protocol address is ambiguous
494 			 */
495 			if (((rt->rt_flags & RTF_GATEWAY) &&
496 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
497 			    info.rti_info[RTAX_IFP] != NULL ||
498 			    (info.rti_info[RTAX_IFA] != NULL &&
499 			     !sa_equal(info.rti_info[RTAX_IFA],
500 				       rt->rt_ifa->ifa_addr))) {
501 				RT_UNLOCK(rt);
502 				if ((error = rt_getifa(&info)) != 0)
503 					senderr(error);
504 				RT_LOCK(rt);
505 			}
506 			if (info.rti_info[RTAX_GATEWAY] != NULL &&
507 			    (error = rt_setgate(rt, rt_key(rt),
508 					info.rti_info[RTAX_GATEWAY])) != 0) {
509 				RT_UNLOCK(rt);
510 				senderr(error);
511 			}
512 			if ((ifa = info.rti_ifa) != NULL) {
513 				struct ifaddr *oifa = rt->rt_ifa;
514 				if (oifa != ifa) {
515 					if (oifa) {
516 						if (oifa->ifa_rtrequest)
517 							oifa->ifa_rtrequest(
518 								RTM_DELETE, rt,
519 								&info);
520 						IFAFREE(oifa);
521 					}
522 				        IFAREF(ifa);
523 				        rt->rt_ifa = ifa;
524 				        rt->rt_ifp = info.rti_ifp;
525 				}
526 			}
527 			/* Allow some flags to be toggled on change. */
528 			if (rtm->rtm_fmask & RTF_FMASK)
529 				rt->rt_flags = (rt->rt_flags &
530 				    ~rtm->rtm_fmask) |
531 				    (rtm->rtm_flags & rtm->rtm_fmask);
532 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
533 					&rt->rt_rmx);
534 			rtm->rtm_index = rt->rt_ifp->if_index;
535 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
536 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
537 			if (info.rti_info[RTAX_GENMASK])
538 				rt->rt_genmask = info.rti_info[RTAX_GENMASK];
539 			/* FALLTHROUGH */
540 		case RTM_LOCK:
541 			/* We don't support locks anymore */
542 			break;
543 		}
544 		RT_UNLOCK(rt);
545 		break;
546 
547 	default:
548 		senderr(EOPNOTSUPP);
549 	}
550 
551 flush:
552 	if (rtm) {
553 		if (error)
554 			rtm->rtm_errno = error;
555 		else
556 			rtm->rtm_flags |= RTF_DONE;
557 	}
558 	if (rt)		/* XXX can this be true? */
559 		RTFREE(rt);
560     {
561 	struct rawcb *rp = NULL;
562 	/*
563 	 * Check to see if we don't want our own messages.
564 	 */
565 	if ((so->so_options & SO_USELOOPBACK) == 0) {
566 		if (route_cb.any_count <= 1) {
567 			if (rtm)
568 				Free(rtm);
569 			m_freem(m);
570 			return (error);
571 		}
572 		/* There is another listener, so construct message */
573 		rp = sotorawcb(so);
574 	}
575 	if (rtm) {
576 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
577 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
578 			m_freem(m);
579 			m = NULL;
580 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
581 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
582 		Free(rtm);
583 	}
584 	if (m) {
585 		if (rp) {
586 			/*
587 			 * XXX insure we don't get a copy by
588 			 * invalidating our protocol
589 			 */
590 			unsigned short family = rp->rcb_proto.sp_family;
591 			rp->rcb_proto.sp_family = 0;
592 			rt_dispatch(m, info.rti_info[RTAX_DST]);
593 			rp->rcb_proto.sp_family = family;
594 		} else
595 			rt_dispatch(m, info.rti_info[RTAX_DST]);
596 	}
597     }
598 	return (error);
599 #undef	sa_equal
600 }
601 
602 static void
603 rt_setmetrics(u_long which, const struct rt_metrics *in,
604 	struct rt_metrics_lite *out)
605 {
606 #define metric(f, e) if (which & (f)) out->e = in->e;
607 	/*
608 	 * Only these are stored in the routing entry since introduction
609 	 * of tcp hostcache. The rest is ignored.
610 	 */
611 	metric(RTV_MTU, rmx_mtu);
612 	metric(RTV_EXPIRE, rmx_expire);
613 #undef metric
614 }
615 
616 static void
617 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
618 {
619 #define metric(e) out->e = in->e;
620 	bzero(out, sizeof(*out));
621 	metric(rmx_mtu);
622 	metric(rmx_expire);
623 #undef metric
624 }
625 
626 /*
627  * Extract the addresses of the passed sockaddrs.
628  * Do a little sanity checking so as to avoid bad memory references.
629  * This data is derived straight from userland.
630  */
631 static int
632 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
633 {
634 	struct sockaddr *sa;
635 	int i;
636 
637 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
638 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
639 			continue;
640 		sa = (struct sockaddr *)cp;
641 		/*
642 		 * It won't fit.
643 		 */
644 		if (cp + sa->sa_len > cplim)
645 			return (EINVAL);
646 		/*
647 		 * there are no more.. quit now
648 		 * If there are more bits, they are in error.
649 		 * I've seen this. route(1) can evidently generate these.
650 		 * This causes kernel to core dump.
651 		 * for compatibility, If we see this, point to a safe address.
652 		 */
653 		if (sa->sa_len == 0) {
654 			rtinfo->rti_info[i] = &sa_zero;
655 			return (0); /* should be EINVAL but for compat */
656 		}
657 		/* accept it */
658 		rtinfo->rti_info[i] = sa;
659 		cp += SA_SIZE(sa);
660 	}
661 	return (0);
662 }
663 
664 static struct mbuf *
665 rt_msg1(int type, struct rt_addrinfo *rtinfo)
666 {
667 	struct rt_msghdr *rtm;
668 	struct mbuf *m;
669 	int i;
670 	struct sockaddr *sa;
671 	int len, dlen;
672 
673 	switch (type) {
674 
675 	case RTM_DELADDR:
676 	case RTM_NEWADDR:
677 		len = sizeof(struct ifa_msghdr);
678 		break;
679 
680 	case RTM_DELMADDR:
681 	case RTM_NEWMADDR:
682 		len = sizeof(struct ifma_msghdr);
683 		break;
684 
685 	case RTM_IFINFO:
686 		len = sizeof(struct if_msghdr);
687 		break;
688 
689 	case RTM_IFANNOUNCE:
690 	case RTM_IEEE80211:
691 		len = sizeof(struct if_announcemsghdr);
692 		break;
693 
694 	default:
695 		len = sizeof(struct rt_msghdr);
696 	}
697 	if (len > MCLBYTES)
698 		panic("rt_msg1");
699 	m = m_gethdr(M_DONTWAIT, MT_DATA);
700 	if (m && len > MHLEN) {
701 		MCLGET(m, M_DONTWAIT);
702 		if ((m->m_flags & M_EXT) == 0) {
703 			m_free(m);
704 			m = NULL;
705 		}
706 	}
707 	if (m == NULL)
708 		return (m);
709 	m->m_pkthdr.len = m->m_len = len;
710 	m->m_pkthdr.rcvif = NULL;
711 	rtm = mtod(m, struct rt_msghdr *);
712 	bzero((caddr_t)rtm, len);
713 	for (i = 0; i < RTAX_MAX; i++) {
714 		if ((sa = rtinfo->rti_info[i]) == NULL)
715 			continue;
716 		rtinfo->rti_addrs |= (1 << i);
717 		dlen = SA_SIZE(sa);
718 		m_copyback(m, len, dlen, (caddr_t)sa);
719 		len += dlen;
720 	}
721 	if (m->m_pkthdr.len != len) {
722 		m_freem(m);
723 		return (NULL);
724 	}
725 	rtm->rtm_msglen = len;
726 	rtm->rtm_version = RTM_VERSION;
727 	rtm->rtm_type = type;
728 	return (m);
729 }
730 
731 static int
732 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
733 {
734 	int i;
735 	int len, dlen, second_time = 0;
736 	caddr_t cp0;
737 
738 	rtinfo->rti_addrs = 0;
739 again:
740 	switch (type) {
741 
742 	case RTM_DELADDR:
743 	case RTM_NEWADDR:
744 		len = sizeof(struct ifa_msghdr);
745 		break;
746 
747 	case RTM_IFINFO:
748 		len = sizeof(struct if_msghdr);
749 		break;
750 
751 	case RTM_NEWMADDR:
752 		len = sizeof(struct ifma_msghdr);
753 		break;
754 
755 	default:
756 		len = sizeof(struct rt_msghdr);
757 	}
758 	cp0 = cp;
759 	if (cp0)
760 		cp += len;
761 	for (i = 0; i < RTAX_MAX; i++) {
762 		struct sockaddr *sa;
763 
764 		if ((sa = rtinfo->rti_info[i]) == NULL)
765 			continue;
766 		rtinfo->rti_addrs |= (1 << i);
767 		dlen = SA_SIZE(sa);
768 		if (cp) {
769 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
770 			cp += dlen;
771 		}
772 		len += dlen;
773 	}
774 	len = ALIGN(len);
775 	if (cp == NULL && w != NULL && !second_time) {
776 		struct walkarg *rw = w;
777 
778 		if (rw->w_req) {
779 			if (rw->w_tmemsize < len) {
780 				if (rw->w_tmem)
781 					free(rw->w_tmem, M_RTABLE);
782 				rw->w_tmem = (caddr_t)
783 					malloc(len, M_RTABLE, M_NOWAIT);
784 				if (rw->w_tmem)
785 					rw->w_tmemsize = len;
786 			}
787 			if (rw->w_tmem) {
788 				cp = rw->w_tmem;
789 				second_time = 1;
790 				goto again;
791 			}
792 		}
793 	}
794 	if (cp) {
795 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
796 
797 		rtm->rtm_version = RTM_VERSION;
798 		rtm->rtm_type = type;
799 		rtm->rtm_msglen = len;
800 	}
801 	return (len);
802 }
803 
804 /*
805  * This routine is called to generate a message from the routing
806  * socket indicating that a redirect has occured, a routing lookup
807  * has failed, or that a protocol has detected timeouts to a particular
808  * destination.
809  */
810 void
811 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
812 {
813 	struct rt_msghdr *rtm;
814 	struct mbuf *m;
815 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
816 
817 	if (route_cb.any_count == 0)
818 		return;
819 	m = rt_msg1(type, rtinfo);
820 	if (m == NULL)
821 		return;
822 	rtm = mtod(m, struct rt_msghdr *);
823 	rtm->rtm_flags = RTF_DONE | flags;
824 	rtm->rtm_errno = error;
825 	rtm->rtm_addrs = rtinfo->rti_addrs;
826 	rt_dispatch(m, sa);
827 }
828 
829 /*
830  * This routine is called to generate a message from the routing
831  * socket indicating that the status of a network interface has changed.
832  */
833 void
834 rt_ifmsg(struct ifnet *ifp)
835 {
836 	struct if_msghdr *ifm;
837 	struct mbuf *m;
838 	struct rt_addrinfo info;
839 
840 	if (route_cb.any_count == 0)
841 		return;
842 	bzero((caddr_t)&info, sizeof(info));
843 	m = rt_msg1(RTM_IFINFO, &info);
844 	if (m == NULL)
845 		return;
846 	ifm = mtod(m, struct if_msghdr *);
847 	ifm->ifm_index = ifp->if_index;
848 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
849 	ifm->ifm_data = ifp->if_data;
850 	ifm->ifm_addrs = 0;
851 	rt_dispatch(m, NULL);
852 }
853 
854 /*
855  * This is called to generate messages from the routing socket
856  * indicating a network interface has had addresses associated with it.
857  * if we ever reverse the logic and replace messages TO the routing
858  * socket indicate a request to configure interfaces, then it will
859  * be unnecessary as the routing socket will automatically generate
860  * copies of it.
861  */
862 void
863 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
864 {
865 	struct rt_addrinfo info;
866 	struct sockaddr *sa = NULL;
867 	int pass;
868 	struct mbuf *m = NULL;
869 	struct ifnet *ifp = ifa->ifa_ifp;
870 
871 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
872 		("unexpected cmd %u", cmd));
873 
874 	if (route_cb.any_count == 0)
875 		return;
876 	for (pass = 1; pass < 3; pass++) {
877 		bzero((caddr_t)&info, sizeof(info));
878 		if ((cmd == RTM_ADD && pass == 1) ||
879 		    (cmd == RTM_DELETE && pass == 2)) {
880 			struct ifa_msghdr *ifam;
881 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
882 
883 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
884 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
885 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
886 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
887 			if ((m = rt_msg1(ncmd, &info)) == NULL)
888 				continue;
889 			ifam = mtod(m, struct ifa_msghdr *);
890 			ifam->ifam_index = ifp->if_index;
891 			ifam->ifam_metric = ifa->ifa_metric;
892 			ifam->ifam_flags = ifa->ifa_flags;
893 			ifam->ifam_addrs = info.rti_addrs;
894 		}
895 		if ((cmd == RTM_ADD && pass == 2) ||
896 		    (cmd == RTM_DELETE && pass == 1)) {
897 			struct rt_msghdr *rtm;
898 
899 			if (rt == NULL)
900 				continue;
901 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
902 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
903 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
904 			if ((m = rt_msg1(cmd, &info)) == NULL)
905 				continue;
906 			rtm = mtod(m, struct rt_msghdr *);
907 			rtm->rtm_index = ifp->if_index;
908 			rtm->rtm_flags |= rt->rt_flags;
909 			rtm->rtm_errno = error;
910 			rtm->rtm_addrs = info.rti_addrs;
911 		}
912 		rt_dispatch(m, sa);
913 	}
914 }
915 
916 /*
917  * This is the analogue to the rt_newaddrmsg which performs the same
918  * function but for multicast group memberhips.  This is easier since
919  * there is no route state to worry about.
920  */
921 void
922 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
923 {
924 	struct rt_addrinfo info;
925 	struct mbuf *m = NULL;
926 	struct ifnet *ifp = ifma->ifma_ifp;
927 	struct ifma_msghdr *ifmam;
928 
929 	if (route_cb.any_count == 0)
930 		return;
931 
932 	bzero((caddr_t)&info, sizeof(info));
933 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
934 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
935 	/*
936 	 * If a link-layer address is present, present it as a ``gateway''
937 	 * (similarly to how ARP entries, e.g., are presented).
938 	 */
939 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
940 	m = rt_msg1(cmd, &info);
941 	if (m == NULL)
942 		return;
943 	ifmam = mtod(m, struct ifma_msghdr *);
944 	ifmam->ifmam_index = ifp->if_index;
945 	ifmam->ifmam_addrs = info.rti_addrs;
946 	rt_dispatch(m, ifma->ifma_addr);
947 }
948 
949 static struct mbuf *
950 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
951 	struct rt_addrinfo *info)
952 {
953 	struct if_announcemsghdr *ifan;
954 	struct mbuf *m;
955 
956 	if (route_cb.any_count == 0)
957 		return NULL;
958 	bzero((caddr_t)info, sizeof(*info));
959 	m = rt_msg1(type, info);
960 	if (m != NULL) {
961 		ifan = mtod(m, struct if_announcemsghdr *);
962 		ifan->ifan_index = ifp->if_index;
963 		strlcpy(ifan->ifan_name, ifp->if_xname,
964 			sizeof(ifan->ifan_name));
965 		ifan->ifan_what = what;
966 	}
967 	return m;
968 }
969 
970 /*
971  * This is called to generate routing socket messages indicating
972  * IEEE80211 wireless events.
973  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
974  */
975 void
976 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
977 {
978 	struct mbuf *m;
979 	struct rt_addrinfo info;
980 
981 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
982 	if (m != NULL) {
983 		/*
984 		 * Append the ieee80211 data.  Try to stick it in the
985 		 * mbuf containing the ifannounce msg; otherwise allocate
986 		 * a new mbuf and append.
987 		 *
988 		 * NB: we assume m is a single mbuf.
989 		 */
990 		if (data_len > M_TRAILINGSPACE(m)) {
991 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
992 			if (n == NULL) {
993 				m_freem(m);
994 				return;
995 			}
996 			bcopy(data, mtod(n, void *), data_len);
997 			n->m_len = data_len;
998 			m->m_next = n;
999 		} else if (data_len > 0) {
1000 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1001 			m->m_len += data_len;
1002 		}
1003 		if (m->m_flags & M_PKTHDR)
1004 			m->m_pkthdr.len += data_len;
1005 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1006 		rt_dispatch(m, NULL);
1007 	}
1008 }
1009 
1010 /*
1011  * This is called to generate routing socket messages indicating
1012  * network interface arrival and departure.
1013  */
1014 void
1015 rt_ifannouncemsg(struct ifnet *ifp, int what)
1016 {
1017 	struct mbuf *m;
1018 	struct rt_addrinfo info;
1019 
1020 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1021 	if (m != NULL)
1022 		rt_dispatch(m, NULL);
1023 }
1024 
1025 static void
1026 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1027 {
1028 	struct m_tag *tag;
1029 
1030 	/*
1031 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1032 	 * use when injecting the mbuf into the routing socket buffer from
1033 	 * the netisr.
1034 	 */
1035 	if (sa != NULL) {
1036 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1037 		    M_NOWAIT);
1038 		if (tag == NULL) {
1039 			m_freem(m);
1040 			return;
1041 		}
1042 		*(unsigned short *)(tag + 1) = sa->sa_family;
1043 		m_tag_prepend(m, tag);
1044 	}
1045 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1046 }
1047 
1048 /*
1049  * This is used in dumping the kernel table via sysctl().
1050  */
1051 static int
1052 sysctl_dumpentry(struct radix_node *rn, void *vw)
1053 {
1054 	struct walkarg *w = vw;
1055 	struct rtentry *rt = (struct rtentry *)rn;
1056 	int error = 0, size;
1057 	struct rt_addrinfo info;
1058 
1059 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1060 		return 0;
1061 	bzero((caddr_t)&info, sizeof(info));
1062 	info.rti_info[RTAX_DST] = rt_key(rt);
1063 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1064 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1065 	info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1066 	if (rt->rt_ifp) {
1067 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1068 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1069 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1070 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1071 	}
1072 	size = rt_msg2(RTM_GET, &info, NULL, w);
1073 	if (w->w_req && w->w_tmem) {
1074 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1075 
1076 		rtm->rtm_flags = rt->rt_flags;
1077 		rtm->rtm_use = rt->rt_rmx.rmx_pksent;
1078 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1079 		rtm->rtm_index = rt->rt_ifp->if_index;
1080 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1081 		rtm->rtm_addrs = info.rti_addrs;
1082 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1083 		return (error);
1084 	}
1085 	return (error);
1086 }
1087 
1088 static int
1089 sysctl_iflist(int af, struct walkarg *w)
1090 {
1091 	struct ifnet *ifp;
1092 	struct ifaddr *ifa;
1093 	struct rt_addrinfo info;
1094 	int len, error = 0;
1095 
1096 	bzero((caddr_t)&info, sizeof(info));
1097 	IFNET_RLOCK();
1098 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1099 		if (w->w_arg && w->w_arg != ifp->if_index)
1100 			continue;
1101 		ifa = ifp->if_addr;
1102 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1103 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1104 		info.rti_info[RTAX_IFP] = NULL;
1105 		if (w->w_req && w->w_tmem) {
1106 			struct if_msghdr *ifm;
1107 
1108 			ifm = (struct if_msghdr *)w->w_tmem;
1109 			ifm->ifm_index = ifp->if_index;
1110 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1111 			ifm->ifm_data = ifp->if_data;
1112 			ifm->ifm_addrs = info.rti_addrs;
1113 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
1114 			if (error)
1115 				goto done;
1116 		}
1117 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1118 			if (af && af != ifa->ifa_addr->sa_family)
1119 				continue;
1120 			if (jailed(curthread->td_ucred) &&
1121 			    prison_if(curthread->td_ucred, ifa->ifa_addr))
1122 				continue;
1123 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1124 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1125 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1126 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1127 			if (w->w_req && w->w_tmem) {
1128 				struct ifa_msghdr *ifam;
1129 
1130 				ifam = (struct ifa_msghdr *)w->w_tmem;
1131 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1132 				ifam->ifam_flags = ifa->ifa_flags;
1133 				ifam->ifam_metric = ifa->ifa_metric;
1134 				ifam->ifam_addrs = info.rti_addrs;
1135 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1136 				if (error)
1137 					goto done;
1138 			}
1139 		}
1140 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1141 			info.rti_info[RTAX_BRD] = NULL;
1142 	}
1143 done:
1144 	IFNET_RUNLOCK();
1145 	return (error);
1146 }
1147 
1148 int
1149 sysctl_ifmalist(int af, struct walkarg *w)
1150 {
1151 	struct ifnet *ifp;
1152 	struct ifmultiaddr *ifma;
1153 	struct	rt_addrinfo info;
1154 	int	len, error = 0;
1155 	struct ifaddr *ifa;
1156 
1157 	bzero((caddr_t)&info, sizeof(info));
1158 	IFNET_RLOCK();
1159 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1160 		if (w->w_arg && w->w_arg != ifp->if_index)
1161 			continue;
1162 		ifa = ifp->if_addr;
1163 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1164 		IF_ADDR_LOCK(ifp);
1165 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1166 			if (af && af != ifma->ifma_addr->sa_family)
1167 				continue;
1168 			if (jailed(curproc->p_ucred) &&
1169 			    prison_if(curproc->p_ucred, ifma->ifma_addr))
1170 				continue;
1171 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1172 			info.rti_info[RTAX_GATEWAY] =
1173 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1174 			    ifma->ifma_lladdr : NULL;
1175 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1176 			if (w->w_req && w->w_tmem) {
1177 				struct ifma_msghdr *ifmam;
1178 
1179 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1180 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1181 				ifmam->ifmam_flags = 0;
1182 				ifmam->ifmam_addrs = info.rti_addrs;
1183 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1184 				if (error) {
1185 					IF_ADDR_UNLOCK(ifp);
1186 					goto done;
1187 				}
1188 			}
1189 		}
1190 		IF_ADDR_UNLOCK(ifp);
1191 	}
1192 done:
1193 	IFNET_RUNLOCK();
1194 	return (error);
1195 }
1196 
1197 static int
1198 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1199 {
1200 	int	*name = (int *)arg1;
1201 	u_int	namelen = arg2;
1202 	struct radix_node_head *rnh;
1203 	int	i, lim, error = EINVAL;
1204 	u_char	af;
1205 	struct	walkarg w;
1206 
1207 	name ++;
1208 	namelen--;
1209 	if (req->newptr)
1210 		return (EPERM);
1211 	if (namelen != 3)
1212 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1213 	af = name[0];
1214 	if (af > AF_MAX)
1215 		return (EINVAL);
1216 	bzero(&w, sizeof(w));
1217 	w.w_op = name[1];
1218 	w.w_arg = name[2];
1219 	w.w_req = req;
1220 
1221 	error = sysctl_wire_old_buffer(req, 0);
1222 	if (error)
1223 		return (error);
1224 	switch (w.w_op) {
1225 
1226 	case NET_RT_DUMP:
1227 	case NET_RT_FLAGS:
1228 		if (af == 0) {			/* dump all tables */
1229 			i = 1;
1230 			lim = AF_MAX;
1231 		} else				/* dump only one table */
1232 			i = lim = af;
1233 		for (error = 0; error == 0 && i <= lim; i++)
1234 			if ((rnh = rt_tables[i]) != NULL) {
1235 				RADIX_NODE_HEAD_LOCK(rnh);
1236 			    	error = rnh->rnh_walktree(rnh,
1237 				    sysctl_dumpentry, &w);
1238 				RADIX_NODE_HEAD_UNLOCK(rnh);
1239 			} else if (af != 0)
1240 				error = EAFNOSUPPORT;
1241 		break;
1242 
1243 	case NET_RT_IFLIST:
1244 		error = sysctl_iflist(af, &w);
1245 		break;
1246 
1247 	case NET_RT_IFMALIST:
1248 		error = sysctl_ifmalist(af, &w);
1249 		break;
1250 	}
1251 	if (w.w_tmem)
1252 		free(w.w_tmem, M_RTABLE);
1253 	return (error);
1254 }
1255 
1256 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1257 
1258 /*
1259  * Definitions of protocols supported in the ROUTE domain.
1260  */
1261 
1262 static struct domain routedomain;		/* or at least forward */
1263 
1264 static struct protosw routesw[] = {
1265 {
1266 	.pr_type =		SOCK_RAW,
1267 	.pr_domain =		&routedomain,
1268 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1269 	.pr_output =		route_output,
1270 	.pr_ctlinput =		raw_ctlinput,
1271 	.pr_init =		raw_init,
1272 	.pr_usrreqs =		&route_usrreqs
1273 }
1274 };
1275 
1276 static struct domain routedomain = {
1277 	.dom_family =		PF_ROUTE,
1278 	.dom_name =		 "route",
1279 	.dom_protosw =		routesw,
1280 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1281 };
1282 
1283 DOMAIN_SET(route);
1284