xref: /freebsd/sys/net/rtsock.c (revision 30d239bc4c510432e65a84fa1c14ed67a3ab1c92)
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 <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/priv.h>
40 #include <sys/proc.h>
41 #include <sys/protosw.h>
42 #include <sys/signalvar.h>
43 #include <sys/socket.h>
44 #include <sys/socketvar.h>
45 #include <sys/sysctl.h>
46 #include <sys/systm.h>
47 
48 #include <net/if.h>
49 #include <net/netisr.h>
50 #include <net/raw_cb.h>
51 #include <net/route.h>
52 
53 #include <netinet/in.h>
54 
55 #ifdef SCTP
56 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
57 #endif /* SCTP */
58 
59 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
60 
61 /* NB: these are not modified */
62 static struct	sockaddr route_dst = { 2, PF_ROUTE, };
63 static struct	sockaddr route_src = { 2, PF_ROUTE, };
64 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
65 
66 static struct {
67 	int	ip_count;	/* attached w/ AF_INET */
68 	int	ip6_count;	/* attached w/ AF_INET6 */
69 	int	ipx_count;	/* attached w/ AF_IPX */
70 	int	any_count;	/* total attached */
71 } route_cb;
72 
73 struct mtx rtsock_mtx;
74 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
75 
76 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
77 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
78 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
79 
80 static struct	ifqueue rtsintrq;
81 
82 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
83 SYSCTL_INT(_net_route, OID_AUTO, netisr_maxqlen, CTLFLAG_RW,
84     &rtsintrq.ifq_maxlen, 0, "maximum routing socket dispatch queue length");
85 
86 struct walkarg {
87 	int	w_tmemsize;
88 	int	w_op, w_arg;
89 	caddr_t	w_tmem;
90 	struct sysctl_req *w_req;
91 };
92 
93 static void	rts_input(struct mbuf *m);
94 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
95 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
96 			caddr_t cp, struct walkarg *w);
97 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
98 			struct rt_addrinfo *rtinfo);
99 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
100 static int	sysctl_iflist(int af, struct walkarg *w);
101 static int	sysctl_ifmalist(int af, struct walkarg *w);
102 static int	route_output(struct mbuf *m, struct socket *so);
103 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
104 			struct rt_metrics_lite *out);
105 static void	rt_getmetrics(const struct rt_metrics_lite *in,
106 			struct rt_metrics *out);
107 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
108 
109 static void
110 rts_init(void)
111 {
112 	int tmp;
113 
114 	rtsintrq.ifq_maxlen = 256;
115 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
116 		rtsintrq.ifq_maxlen = tmp;
117 	mtx_init(&rtsintrq.ifq_mtx, "rts_inq", NULL, MTX_DEF);
118 	netisr_register(NETISR_ROUTE, rts_input, &rtsintrq, NETISR_MPSAFE);
119 }
120 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0)
121 
122 static void
123 rts_input(struct mbuf *m)
124 {
125 	struct sockproto route_proto;
126 	unsigned short *family;
127 	struct m_tag *tag;
128 
129 	route_proto.sp_family = PF_ROUTE;
130 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
131 	if (tag != NULL) {
132 		family = (unsigned short *)(tag + 1);
133 		route_proto.sp_protocol = *family;
134 		m_tag_delete(m, tag);
135 	} else
136 		route_proto.sp_protocol = 0;
137 
138 	raw_input(m, &route_proto, &route_src, &route_dst);
139 }
140 
141 /*
142  * It really doesn't make any sense at all for this code to share much
143  * with raw_usrreq.c, since its functionality is so restricted.  XXX
144  */
145 static void
146 rts_abort(struct socket *so)
147 {
148 
149 	raw_usrreqs.pru_abort(so);
150 }
151 
152 static void
153 rts_close(struct socket *so)
154 {
155 
156 	raw_usrreqs.pru_close(so);
157 }
158 
159 /* pru_accept is EOPNOTSUPP */
160 
161 static int
162 rts_attach(struct socket *so, int proto, struct thread *td)
163 {
164 	struct rawcb *rp;
165 	int s, error;
166 
167 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
168 
169 	/* XXX */
170 	MALLOC(rp, struct rawcb *, sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
171 	if (rp == NULL)
172 		return ENOBUFS;
173 
174 	/*
175 	 * The splnet() is necessary to block protocols from sending
176 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
177 	 * this PCB is extant but incompletely initialized.
178 	 * Probably we should try to do more of this work beforehand and
179 	 * eliminate the spl.
180 	 */
181 	s = splnet();
182 	so->so_pcb = (caddr_t)rp;
183 	error = raw_attach(so, proto);
184 	rp = sotorawcb(so);
185 	if (error) {
186 		splx(s);
187 		so->so_pcb = NULL;
188 		free(rp, M_PCB);
189 		return error;
190 	}
191 	RTSOCK_LOCK();
192 	switch(rp->rcb_proto.sp_protocol) {
193 	case AF_INET:
194 		route_cb.ip_count++;
195 		break;
196 	case AF_INET6:
197 		route_cb.ip6_count++;
198 		break;
199 	case AF_IPX:
200 		route_cb.ipx_count++;
201 		break;
202 	}
203 	rp->rcb_faddr = &route_src;
204 	route_cb.any_count++;
205 	RTSOCK_UNLOCK();
206 	soisconnected(so);
207 	so->so_options |= SO_USELOOPBACK;
208 	splx(s);
209 	return 0;
210 }
211 
212 static int
213 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
214 {
215 
216 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
217 }
218 
219 static int
220 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
221 {
222 
223 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
224 }
225 
226 /* pru_connect2 is EOPNOTSUPP */
227 /* pru_control is EOPNOTSUPP */
228 
229 static void
230 rts_detach(struct socket *so)
231 {
232 	struct rawcb *rp = sotorawcb(so);
233 
234 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
235 
236 	RTSOCK_LOCK();
237 	switch(rp->rcb_proto.sp_protocol) {
238 	case AF_INET:
239 		route_cb.ip_count--;
240 		break;
241 	case AF_INET6:
242 		route_cb.ip6_count--;
243 		break;
244 	case AF_IPX:
245 		route_cb.ipx_count--;
246 		break;
247 	}
248 	route_cb.any_count--;
249 	RTSOCK_UNLOCK();
250 	raw_usrreqs.pru_detach(so);
251 }
252 
253 static int
254 rts_disconnect(struct socket *so)
255 {
256 
257 	return (raw_usrreqs.pru_disconnect(so));
258 }
259 
260 /* pru_listen is EOPNOTSUPP */
261 
262 static int
263 rts_peeraddr(struct socket *so, struct sockaddr **nam)
264 {
265 
266 	return (raw_usrreqs.pru_peeraddr(so, nam));
267 }
268 
269 /* pru_rcvd is EOPNOTSUPP */
270 /* pru_rcvoob is EOPNOTSUPP */
271 
272 static int
273 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
274 	 struct mbuf *control, struct thread *td)
275 {
276 
277 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
278 }
279 
280 /* pru_sense is null */
281 
282 static int
283 rts_shutdown(struct socket *so)
284 {
285 
286 	return (raw_usrreqs.pru_shutdown(so));
287 }
288 
289 static int
290 rts_sockaddr(struct socket *so, struct sockaddr **nam)
291 {
292 
293 	return (raw_usrreqs.pru_sockaddr(so, nam));
294 }
295 
296 static struct pr_usrreqs route_usrreqs = {
297 	.pru_abort =		rts_abort,
298 	.pru_attach =		rts_attach,
299 	.pru_bind =		rts_bind,
300 	.pru_connect =		rts_connect,
301 	.pru_detach =		rts_detach,
302 	.pru_disconnect =	rts_disconnect,
303 	.pru_peeraddr =		rts_peeraddr,
304 	.pru_send =		rts_send,
305 	.pru_shutdown =		rts_shutdown,
306 	.pru_sockaddr =		rts_sockaddr,
307 	.pru_close =		rts_close,
308 };
309 
310 /*ARGSUSED*/
311 static int
312 route_output(struct mbuf *m, struct socket *so)
313 {
314 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
315 	struct rt_msghdr *rtm = NULL;
316 	struct rtentry *rt = NULL;
317 	struct radix_node_head *rnh;
318 	struct rt_addrinfo info;
319 	int len, error = 0;
320 	struct ifnet *ifp = NULL;
321 	struct sockaddr_in jail;
322 
323 #define senderr(e) { error = e; goto flush;}
324 	if (m == NULL || ((m->m_len < sizeof(long)) &&
325 		       (m = m_pullup(m, sizeof(long))) == NULL))
326 		return (ENOBUFS);
327 	if ((m->m_flags & M_PKTHDR) == 0)
328 		panic("route_output");
329 	len = m->m_pkthdr.len;
330 	if (len < sizeof(*rtm) ||
331 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
332 		info.rti_info[RTAX_DST] = NULL;
333 		senderr(EINVAL);
334 	}
335 	R_Malloc(rtm, struct rt_msghdr *, len);
336 	if (rtm == NULL) {
337 		info.rti_info[RTAX_DST] = NULL;
338 		senderr(ENOBUFS);
339 	}
340 	m_copydata(m, 0, len, (caddr_t)rtm);
341 	if (rtm->rtm_version != RTM_VERSION) {
342 		info.rti_info[RTAX_DST] = NULL;
343 		senderr(EPROTONOSUPPORT);
344 	}
345 	rtm->rtm_pid = curproc->p_pid;
346 	bzero(&info, sizeof(info));
347 	info.rti_addrs = rtm->rtm_addrs;
348 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
349 		info.rti_info[RTAX_DST] = NULL;
350 		senderr(EINVAL);
351 	}
352 	info.rti_flags = rtm->rtm_flags;
353 	if (info.rti_info[RTAX_DST] == NULL ||
354 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
355 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
356 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
357 		senderr(EINVAL);
358 	if (info.rti_info[RTAX_GENMASK]) {
359 		struct radix_node *t;
360 		t = rn_addmask((caddr_t) info.rti_info[RTAX_GENMASK], 0, 1);
361 		if (t != NULL &&
362 		    bcmp((char *)(void *)info.rti_info[RTAX_GENMASK] + 1,
363 		    (char *)(void *)t->rn_key + 1,
364 		    ((struct sockaddr *)t->rn_key)->sa_len - 1) == 0)
365 			info.rti_info[RTAX_GENMASK] =
366 			    (struct sockaddr *)t->rn_key;
367 		else
368 			senderr(ENOBUFS);
369 	}
370 
371 	/*
372 	 * Verify that the caller has the appropriate privilege; RTM_GET
373 	 * is the only operation the non-superuser is allowed.
374 	 */
375 	if (rtm->rtm_type != RTM_GET) {
376 		error = priv_check(curthread, PRIV_NET_ROUTE);
377 		if (error)
378 			senderr(error);
379 	}
380 
381 	switch (rtm->rtm_type) {
382 		struct rtentry *saved_nrt;
383 
384 	case RTM_ADD:
385 		if (info.rti_info[RTAX_GATEWAY] == NULL)
386 			senderr(EINVAL);
387 		saved_nrt = NULL;
388 		error = rtrequest1(RTM_ADD, &info, &saved_nrt);
389 		if (error == 0 && saved_nrt) {
390 			RT_LOCK(saved_nrt);
391 			rt_setmetrics(rtm->rtm_inits,
392 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
393 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
394 			RT_REMREF(saved_nrt);
395 			saved_nrt->rt_genmask = info.rti_info[RTAX_GENMASK];
396 			RT_UNLOCK(saved_nrt);
397 		}
398 		break;
399 
400 	case RTM_DELETE:
401 		saved_nrt = NULL;
402 		error = rtrequest1(RTM_DELETE, &info, &saved_nrt);
403 		if (error == 0) {
404 			RT_LOCK(saved_nrt);
405 			rt = saved_nrt;
406 			goto report;
407 		}
408 		break;
409 
410 	case RTM_GET:
411 	case RTM_CHANGE:
412 	case RTM_LOCK:
413 		rnh = rt_tables[info.rti_info[RTAX_DST]->sa_family];
414 		if (rnh == NULL)
415 			senderr(EAFNOSUPPORT);
416 		RADIX_NODE_HEAD_LOCK(rnh);
417 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
418 			info.rti_info[RTAX_NETMASK], rnh);
419 		if (rt == NULL) {	/* XXX looks bogus */
420 			RADIX_NODE_HEAD_UNLOCK(rnh);
421 			senderr(ESRCH);
422 		}
423 		RT_LOCK(rt);
424 		RT_ADDREF(rt);
425 		RADIX_NODE_HEAD_UNLOCK(rnh);
426 
427 		/*
428 		 * Fix for PR: 82974
429 		 *
430 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
431 		 * returns a perfect match in case a netmask is
432 		 * specified.  For host routes only a longest prefix
433 		 * match is returned so it is necessary to compare the
434 		 * existence of the netmask.  If both have a netmask
435 		 * rnh_lookup() did a perfect match and if none of them
436 		 * have a netmask both are host routes which is also a
437 		 * perfect match.
438 		 */
439 
440 		if (rtm->rtm_type != RTM_GET &&
441 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
442 			RT_UNLOCK(rt);
443 			senderr(ESRCH);
444 		}
445 
446 		switch(rtm->rtm_type) {
447 
448 		case RTM_GET:
449 		report:
450 			RT_LOCK_ASSERT(rt);
451 			info.rti_info[RTAX_DST] = rt_key(rt);
452 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
453 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
454 			info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
455 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
456 				ifp = rt->rt_ifp;
457 				if (ifp) {
458 					info.rti_info[RTAX_IFP] =
459 					    ifp->if_addr->ifa_addr;
460 					if (jailed(so->so_cred)) {
461 						bzero(&jail, sizeof(jail));
462 						jail.sin_family = PF_INET;
463 						jail.sin_len = sizeof(jail);
464 						jail.sin_addr.s_addr =
465 						htonl(prison_getip(so->so_cred));
466 						info.rti_info[RTAX_IFA] =
467 						    (struct sockaddr *)&jail;
468 					} else
469 						info.rti_info[RTAX_IFA] =
470 						    rt->rt_ifa->ifa_addr;
471 					if (ifp->if_flags & IFF_POINTOPOINT)
472 						info.rti_info[RTAX_BRD] =
473 						    rt->rt_ifa->ifa_dstaddr;
474 					rtm->rtm_index = ifp->if_index;
475 				} else {
476 					info.rti_info[RTAX_IFP] = NULL;
477 					info.rti_info[RTAX_IFA] = NULL;
478 				}
479 			} else if ((ifp = rt->rt_ifp) != NULL) {
480 				rtm->rtm_index = ifp->if_index;
481 			}
482 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
483 			if (len > rtm->rtm_msglen) {
484 				struct rt_msghdr *new_rtm;
485 				R_Malloc(new_rtm, struct rt_msghdr *, len);
486 				if (new_rtm == NULL) {
487 					RT_UNLOCK(rt);
488 					senderr(ENOBUFS);
489 				}
490 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
491 				Free(rtm); rtm = new_rtm;
492 			}
493 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
494 			rtm->rtm_flags = rt->rt_flags;
495 			rtm->rtm_use = 0;
496 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
497 			rtm->rtm_addrs = info.rti_addrs;
498 			break;
499 
500 		case RTM_CHANGE:
501 			/*
502 			 * New gateway could require new ifaddr, ifp;
503 			 * flags may also be different; ifp may be specified
504 			 * by ll sockaddr when protocol address is ambiguous
505 			 */
506 			if (((rt->rt_flags & RTF_GATEWAY) &&
507 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
508 			    info.rti_info[RTAX_IFP] != NULL ||
509 			    (info.rti_info[RTAX_IFA] != NULL &&
510 			     !sa_equal(info.rti_info[RTAX_IFA],
511 				       rt->rt_ifa->ifa_addr))) {
512 				RT_UNLOCK(rt);
513 				if ((error = rt_getifa(&info)) != 0)
514 					senderr(error);
515 				RT_LOCK(rt);
516 			}
517 			if (info.rti_ifa != NULL &&
518 			    info.rti_ifa != rt->rt_ifa &&
519 			    rt->rt_ifa != NULL &&
520 			    rt->rt_ifa->ifa_rtrequest != NULL) {
521 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
522 				    &info);
523 				IFAFREE(rt->rt_ifa);
524 			}
525 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
526 				if ((error = rt_setgate(rt, rt_key(rt),
527 					info.rti_info[RTAX_GATEWAY])) != 0) {
528 					RT_UNLOCK(rt);
529 					senderr(error);
530 				}
531 				if (!(rt->rt_flags & RTF_LLINFO))
532 					rt->rt_flags |= RTF_GATEWAY;
533 			}
534 			if (info.rti_ifa != NULL &&
535 			    info.rti_ifa != rt->rt_ifa) {
536 				IFAREF(info.rti_ifa);
537 				rt->rt_ifa = info.rti_ifa;
538 				rt->rt_ifp = info.rti_ifp;
539 			}
540 			/* Allow some flags to be toggled on change. */
541 			if (rtm->rtm_fmask & RTF_FMASK)
542 				rt->rt_flags = (rt->rt_flags &
543 				    ~rtm->rtm_fmask) |
544 				    (rtm->rtm_flags & rtm->rtm_fmask);
545 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
546 					&rt->rt_rmx);
547 			rtm->rtm_index = rt->rt_ifp->if_index;
548 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
549 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
550 			if (info.rti_info[RTAX_GENMASK])
551 				rt->rt_genmask = info.rti_info[RTAX_GENMASK];
552 			/* FALLTHROUGH */
553 		case RTM_LOCK:
554 			/* We don't support locks anymore */
555 			break;
556 		}
557 		RT_UNLOCK(rt);
558 		break;
559 
560 	default:
561 		senderr(EOPNOTSUPP);
562 	}
563 
564 flush:
565 	if (rtm) {
566 		if (error)
567 			rtm->rtm_errno = error;
568 		else
569 			rtm->rtm_flags |= RTF_DONE;
570 	}
571 	if (rt)		/* XXX can this be true? */
572 		RTFREE(rt);
573     {
574 	struct rawcb *rp = NULL;
575 	/*
576 	 * Check to see if we don't want our own messages.
577 	 */
578 	if ((so->so_options & SO_USELOOPBACK) == 0) {
579 		if (route_cb.any_count <= 1) {
580 			if (rtm)
581 				Free(rtm);
582 			m_freem(m);
583 			return (error);
584 		}
585 		/* There is another listener, so construct message */
586 		rp = sotorawcb(so);
587 	}
588 	if (rtm) {
589 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
590 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
591 			m_freem(m);
592 			m = NULL;
593 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
594 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
595 		Free(rtm);
596 	}
597 	if (m) {
598 		if (rp) {
599 			/*
600 			 * XXX insure we don't get a copy by
601 			 * invalidating our protocol
602 			 */
603 			unsigned short family = rp->rcb_proto.sp_family;
604 			rp->rcb_proto.sp_family = 0;
605 			rt_dispatch(m, info.rti_info[RTAX_DST]);
606 			rp->rcb_proto.sp_family = family;
607 		} else
608 			rt_dispatch(m, info.rti_info[RTAX_DST]);
609 	}
610     }
611 	return (error);
612 #undef	sa_equal
613 }
614 
615 static void
616 rt_setmetrics(u_long which, const struct rt_metrics *in,
617 	struct rt_metrics_lite *out)
618 {
619 #define metric(f, e) if (which & (f)) out->e = in->e;
620 	/*
621 	 * Only these are stored in the routing entry since introduction
622 	 * of tcp hostcache. The rest is ignored.
623 	 */
624 	metric(RTV_MTU, rmx_mtu);
625 	/* Userland -> kernel timebase conversion. */
626 	if (which & RTV_EXPIRE)
627 		out->rmx_expire = in->rmx_expire ?
628 		    in->rmx_expire - time_second + time_uptime : 0;
629 #undef metric
630 }
631 
632 static void
633 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
634 {
635 #define metric(e) out->e = in->e;
636 	bzero(out, sizeof(*out));
637 	metric(rmx_mtu);
638 	/* Kernel -> userland timebase conversion. */
639 	out->rmx_expire = in->rmx_expire ?
640 	    in->rmx_expire - time_uptime + time_second : 0;
641 #undef metric
642 }
643 
644 /*
645  * Extract the addresses of the passed sockaddrs.
646  * Do a little sanity checking so as to avoid bad memory references.
647  * This data is derived straight from userland.
648  */
649 static int
650 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
651 {
652 	struct sockaddr *sa;
653 	int i;
654 
655 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
656 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
657 			continue;
658 		sa = (struct sockaddr *)cp;
659 		/*
660 		 * It won't fit.
661 		 */
662 		if (cp + sa->sa_len > cplim)
663 			return (EINVAL);
664 		/*
665 		 * there are no more.. quit now
666 		 * If there are more bits, they are in error.
667 		 * I've seen this. route(1) can evidently generate these.
668 		 * This causes kernel to core dump.
669 		 * for compatibility, If we see this, point to a safe address.
670 		 */
671 		if (sa->sa_len == 0) {
672 			rtinfo->rti_info[i] = &sa_zero;
673 			return (0); /* should be EINVAL but for compat */
674 		}
675 		/* accept it */
676 		rtinfo->rti_info[i] = sa;
677 		cp += SA_SIZE(sa);
678 	}
679 	return (0);
680 }
681 
682 static struct mbuf *
683 rt_msg1(int type, struct rt_addrinfo *rtinfo)
684 {
685 	struct rt_msghdr *rtm;
686 	struct mbuf *m;
687 	int i;
688 	struct sockaddr *sa;
689 	int len, dlen;
690 
691 	switch (type) {
692 
693 	case RTM_DELADDR:
694 	case RTM_NEWADDR:
695 		len = sizeof(struct ifa_msghdr);
696 		break;
697 
698 	case RTM_DELMADDR:
699 	case RTM_NEWMADDR:
700 		len = sizeof(struct ifma_msghdr);
701 		break;
702 
703 	case RTM_IFINFO:
704 		len = sizeof(struct if_msghdr);
705 		break;
706 
707 	case RTM_IFANNOUNCE:
708 	case RTM_IEEE80211:
709 		len = sizeof(struct if_announcemsghdr);
710 		break;
711 
712 	default:
713 		len = sizeof(struct rt_msghdr);
714 	}
715 	if (len > MCLBYTES)
716 		panic("rt_msg1");
717 	m = m_gethdr(M_DONTWAIT, MT_DATA);
718 	if (m && len > MHLEN) {
719 		MCLGET(m, M_DONTWAIT);
720 		if ((m->m_flags & M_EXT) == 0) {
721 			m_free(m);
722 			m = NULL;
723 		}
724 	}
725 	if (m == NULL)
726 		return (m);
727 	m->m_pkthdr.len = m->m_len = len;
728 	m->m_pkthdr.rcvif = NULL;
729 	rtm = mtod(m, struct rt_msghdr *);
730 	bzero((caddr_t)rtm, len);
731 	for (i = 0; i < RTAX_MAX; i++) {
732 		if ((sa = rtinfo->rti_info[i]) == NULL)
733 			continue;
734 		rtinfo->rti_addrs |= (1 << i);
735 		dlen = SA_SIZE(sa);
736 		m_copyback(m, len, dlen, (caddr_t)sa);
737 		len += dlen;
738 	}
739 	if (m->m_pkthdr.len != len) {
740 		m_freem(m);
741 		return (NULL);
742 	}
743 	rtm->rtm_msglen = len;
744 	rtm->rtm_version = RTM_VERSION;
745 	rtm->rtm_type = type;
746 	return (m);
747 }
748 
749 static int
750 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
751 {
752 	int i;
753 	int len, dlen, second_time = 0;
754 	caddr_t cp0;
755 
756 	rtinfo->rti_addrs = 0;
757 again:
758 	switch (type) {
759 
760 	case RTM_DELADDR:
761 	case RTM_NEWADDR:
762 		len = sizeof(struct ifa_msghdr);
763 		break;
764 
765 	case RTM_IFINFO:
766 		len = sizeof(struct if_msghdr);
767 		break;
768 
769 	case RTM_NEWMADDR:
770 		len = sizeof(struct ifma_msghdr);
771 		break;
772 
773 	default:
774 		len = sizeof(struct rt_msghdr);
775 	}
776 	cp0 = cp;
777 	if (cp0)
778 		cp += len;
779 	for (i = 0; i < RTAX_MAX; i++) {
780 		struct sockaddr *sa;
781 
782 		if ((sa = rtinfo->rti_info[i]) == NULL)
783 			continue;
784 		rtinfo->rti_addrs |= (1 << i);
785 		dlen = SA_SIZE(sa);
786 		if (cp) {
787 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
788 			cp += dlen;
789 		}
790 		len += dlen;
791 	}
792 	len = ALIGN(len);
793 	if (cp == NULL && w != NULL && !second_time) {
794 		struct walkarg *rw = w;
795 
796 		if (rw->w_req) {
797 			if (rw->w_tmemsize < len) {
798 				if (rw->w_tmem)
799 					free(rw->w_tmem, M_RTABLE);
800 				rw->w_tmem = (caddr_t)
801 					malloc(len, M_RTABLE, M_NOWAIT);
802 				if (rw->w_tmem)
803 					rw->w_tmemsize = len;
804 			}
805 			if (rw->w_tmem) {
806 				cp = rw->w_tmem;
807 				second_time = 1;
808 				goto again;
809 			}
810 		}
811 	}
812 	if (cp) {
813 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
814 
815 		rtm->rtm_version = RTM_VERSION;
816 		rtm->rtm_type = type;
817 		rtm->rtm_msglen = len;
818 	}
819 	return (len);
820 }
821 
822 /*
823  * This routine is called to generate a message from the routing
824  * socket indicating that a redirect has occured, a routing lookup
825  * has failed, or that a protocol has detected timeouts to a particular
826  * destination.
827  */
828 void
829 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
830 {
831 	struct rt_msghdr *rtm;
832 	struct mbuf *m;
833 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
834 
835 	if (route_cb.any_count == 0)
836 		return;
837 	m = rt_msg1(type, rtinfo);
838 	if (m == NULL)
839 		return;
840 	rtm = mtod(m, struct rt_msghdr *);
841 	rtm->rtm_flags = RTF_DONE | flags;
842 	rtm->rtm_errno = error;
843 	rtm->rtm_addrs = rtinfo->rti_addrs;
844 	rt_dispatch(m, sa);
845 }
846 
847 /*
848  * This routine is called to generate a message from the routing
849  * socket indicating that the status of a network interface has changed.
850  */
851 void
852 rt_ifmsg(struct ifnet *ifp)
853 {
854 	struct if_msghdr *ifm;
855 	struct mbuf *m;
856 	struct rt_addrinfo info;
857 
858 	if (route_cb.any_count == 0)
859 		return;
860 	bzero((caddr_t)&info, sizeof(info));
861 	m = rt_msg1(RTM_IFINFO, &info);
862 	if (m == NULL)
863 		return;
864 	ifm = mtod(m, struct if_msghdr *);
865 	ifm->ifm_index = ifp->if_index;
866 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
867 	ifm->ifm_data = ifp->if_data;
868 	ifm->ifm_addrs = 0;
869 	rt_dispatch(m, NULL);
870 }
871 
872 /*
873  * This is called to generate messages from the routing socket
874  * indicating a network interface has had addresses associated with it.
875  * if we ever reverse the logic and replace messages TO the routing
876  * socket indicate a request to configure interfaces, then it will
877  * be unnecessary as the routing socket will automatically generate
878  * copies of it.
879  */
880 void
881 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
882 {
883 	struct rt_addrinfo info;
884 	struct sockaddr *sa = NULL;
885 	int pass;
886 	struct mbuf *m = NULL;
887 	struct ifnet *ifp = ifa->ifa_ifp;
888 
889 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
890 		("unexpected cmd %u", cmd));
891 #ifdef SCTP
892 	/*
893 	 * notify the SCTP stack
894 	 * this will only get called when an address is added/deleted
895 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
896 	 */
897 	sctp_addr_change(ifa, cmd);
898 #endif /* SCTP */
899 	if (route_cb.any_count == 0)
900 		return;
901 	for (pass = 1; pass < 3; pass++) {
902 		bzero((caddr_t)&info, sizeof(info));
903 		if ((cmd == RTM_ADD && pass == 1) ||
904 		    (cmd == RTM_DELETE && pass == 2)) {
905 			struct ifa_msghdr *ifam;
906 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
907 
908 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
909 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
910 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
911 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
912 			if ((m = rt_msg1(ncmd, &info)) == NULL)
913 				continue;
914 			ifam = mtod(m, struct ifa_msghdr *);
915 			ifam->ifam_index = ifp->if_index;
916 			ifam->ifam_metric = ifa->ifa_metric;
917 			ifam->ifam_flags = ifa->ifa_flags;
918 			ifam->ifam_addrs = info.rti_addrs;
919 		}
920 		if ((cmd == RTM_ADD && pass == 2) ||
921 		    (cmd == RTM_DELETE && pass == 1)) {
922 			struct rt_msghdr *rtm;
923 
924 			if (rt == NULL)
925 				continue;
926 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
927 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
928 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
929 			if ((m = rt_msg1(cmd, &info)) == NULL)
930 				continue;
931 			rtm = mtod(m, struct rt_msghdr *);
932 			rtm->rtm_index = ifp->if_index;
933 			rtm->rtm_flags |= rt->rt_flags;
934 			rtm->rtm_errno = error;
935 			rtm->rtm_addrs = info.rti_addrs;
936 		}
937 		rt_dispatch(m, sa);
938 	}
939 }
940 
941 /*
942  * This is the analogue to the rt_newaddrmsg which performs the same
943  * function but for multicast group memberhips.  This is easier since
944  * there is no route state to worry about.
945  */
946 void
947 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
948 {
949 	struct rt_addrinfo info;
950 	struct mbuf *m = NULL;
951 	struct ifnet *ifp = ifma->ifma_ifp;
952 	struct ifma_msghdr *ifmam;
953 
954 	if (route_cb.any_count == 0)
955 		return;
956 
957 	bzero((caddr_t)&info, sizeof(info));
958 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
959 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
960 	/*
961 	 * If a link-layer address is present, present it as a ``gateway''
962 	 * (similarly to how ARP entries, e.g., are presented).
963 	 */
964 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
965 	m = rt_msg1(cmd, &info);
966 	if (m == NULL)
967 		return;
968 	ifmam = mtod(m, struct ifma_msghdr *);
969 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
970 	    __func__));
971 	ifmam->ifmam_index = ifp->if_index;
972 	ifmam->ifmam_addrs = info.rti_addrs;
973 	rt_dispatch(m, ifma->ifma_addr);
974 }
975 
976 static struct mbuf *
977 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
978 	struct rt_addrinfo *info)
979 {
980 	struct if_announcemsghdr *ifan;
981 	struct mbuf *m;
982 
983 	if (route_cb.any_count == 0)
984 		return NULL;
985 	bzero((caddr_t)info, sizeof(*info));
986 	m = rt_msg1(type, info);
987 	if (m != NULL) {
988 		ifan = mtod(m, struct if_announcemsghdr *);
989 		ifan->ifan_index = ifp->if_index;
990 		strlcpy(ifan->ifan_name, ifp->if_xname,
991 			sizeof(ifan->ifan_name));
992 		ifan->ifan_what = what;
993 	}
994 	return m;
995 }
996 
997 /*
998  * This is called to generate routing socket messages indicating
999  * IEEE80211 wireless events.
1000  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1001  */
1002 void
1003 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1004 {
1005 	struct mbuf *m;
1006 	struct rt_addrinfo info;
1007 
1008 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1009 	if (m != NULL) {
1010 		/*
1011 		 * Append the ieee80211 data.  Try to stick it in the
1012 		 * mbuf containing the ifannounce msg; otherwise allocate
1013 		 * a new mbuf and append.
1014 		 *
1015 		 * NB: we assume m is a single mbuf.
1016 		 */
1017 		if (data_len > M_TRAILINGSPACE(m)) {
1018 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1019 			if (n == NULL) {
1020 				m_freem(m);
1021 				return;
1022 			}
1023 			bcopy(data, mtod(n, void *), data_len);
1024 			n->m_len = data_len;
1025 			m->m_next = n;
1026 		} else if (data_len > 0) {
1027 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1028 			m->m_len += data_len;
1029 		}
1030 		if (m->m_flags & M_PKTHDR)
1031 			m->m_pkthdr.len += data_len;
1032 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1033 		rt_dispatch(m, NULL);
1034 	}
1035 }
1036 
1037 /*
1038  * This is called to generate routing socket messages indicating
1039  * network interface arrival and departure.
1040  */
1041 void
1042 rt_ifannouncemsg(struct ifnet *ifp, int what)
1043 {
1044 	struct mbuf *m;
1045 	struct rt_addrinfo info;
1046 
1047 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1048 	if (m != NULL)
1049 		rt_dispatch(m, NULL);
1050 }
1051 
1052 static void
1053 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1054 {
1055 	struct m_tag *tag;
1056 
1057 	/*
1058 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1059 	 * use when injecting the mbuf into the routing socket buffer from
1060 	 * the netisr.
1061 	 */
1062 	if (sa != NULL) {
1063 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1064 		    M_NOWAIT);
1065 		if (tag == NULL) {
1066 			m_freem(m);
1067 			return;
1068 		}
1069 		*(unsigned short *)(tag + 1) = sa->sa_family;
1070 		m_tag_prepend(m, tag);
1071 	}
1072 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1073 }
1074 
1075 /*
1076  * This is used in dumping the kernel table via sysctl().
1077  */
1078 static int
1079 sysctl_dumpentry(struct radix_node *rn, void *vw)
1080 {
1081 	struct walkarg *w = vw;
1082 	struct rtentry *rt = (struct rtentry *)rn;
1083 	int error = 0, size;
1084 	struct rt_addrinfo info;
1085 
1086 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1087 		return 0;
1088 	bzero((caddr_t)&info, sizeof(info));
1089 	info.rti_info[RTAX_DST] = rt_key(rt);
1090 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1091 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1092 	info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1093 	if (rt->rt_ifp) {
1094 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1095 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1096 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1097 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1098 	}
1099 	size = rt_msg2(RTM_GET, &info, NULL, w);
1100 	if (w->w_req && w->w_tmem) {
1101 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1102 
1103 		rtm->rtm_flags = rt->rt_flags;
1104 		rtm->rtm_use = rt->rt_rmx.rmx_pksent;
1105 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1106 		rtm->rtm_index = rt->rt_ifp->if_index;
1107 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1108 		rtm->rtm_addrs = info.rti_addrs;
1109 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1110 		return (error);
1111 	}
1112 	return (error);
1113 }
1114 
1115 static int
1116 sysctl_iflist(int af, struct walkarg *w)
1117 {
1118 	struct ifnet *ifp;
1119 	struct ifaddr *ifa;
1120 	struct rt_addrinfo info;
1121 	int len, error = 0;
1122 
1123 	bzero((caddr_t)&info, sizeof(info));
1124 	IFNET_RLOCK();
1125 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1126 		if (w->w_arg && w->w_arg != ifp->if_index)
1127 			continue;
1128 		ifa = ifp->if_addr;
1129 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1130 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1131 		info.rti_info[RTAX_IFP] = NULL;
1132 		if (w->w_req && w->w_tmem) {
1133 			struct if_msghdr *ifm;
1134 
1135 			ifm = (struct if_msghdr *)w->w_tmem;
1136 			ifm->ifm_index = ifp->if_index;
1137 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1138 			ifm->ifm_data = ifp->if_data;
1139 			ifm->ifm_addrs = info.rti_addrs;
1140 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
1141 			if (error)
1142 				goto done;
1143 		}
1144 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1145 			if (af && af != ifa->ifa_addr->sa_family)
1146 				continue;
1147 			if (jailed(curthread->td_ucred) &&
1148 			    prison_if(curthread->td_ucred, ifa->ifa_addr))
1149 				continue;
1150 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1151 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1152 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1153 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1154 			if (w->w_req && w->w_tmem) {
1155 				struct ifa_msghdr *ifam;
1156 
1157 				ifam = (struct ifa_msghdr *)w->w_tmem;
1158 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1159 				ifam->ifam_flags = ifa->ifa_flags;
1160 				ifam->ifam_metric = ifa->ifa_metric;
1161 				ifam->ifam_addrs = info.rti_addrs;
1162 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1163 				if (error)
1164 					goto done;
1165 			}
1166 		}
1167 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1168 			info.rti_info[RTAX_BRD] = NULL;
1169 	}
1170 done:
1171 	IFNET_RUNLOCK();
1172 	return (error);
1173 }
1174 
1175 int
1176 sysctl_ifmalist(int af, struct walkarg *w)
1177 {
1178 	struct ifnet *ifp;
1179 	struct ifmultiaddr *ifma;
1180 	struct	rt_addrinfo info;
1181 	int	len, error = 0;
1182 	struct ifaddr *ifa;
1183 
1184 	bzero((caddr_t)&info, sizeof(info));
1185 	IFNET_RLOCK();
1186 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1187 		if (w->w_arg && w->w_arg != ifp->if_index)
1188 			continue;
1189 		ifa = ifp->if_addr;
1190 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1191 		IF_ADDR_LOCK(ifp);
1192 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1193 			if (af && af != ifma->ifma_addr->sa_family)
1194 				continue;
1195 			if (jailed(curproc->p_ucred) &&
1196 			    prison_if(curproc->p_ucred, ifma->ifma_addr))
1197 				continue;
1198 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1199 			info.rti_info[RTAX_GATEWAY] =
1200 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1201 			    ifma->ifma_lladdr : NULL;
1202 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1203 			if (w->w_req && w->w_tmem) {
1204 				struct ifma_msghdr *ifmam;
1205 
1206 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1207 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1208 				ifmam->ifmam_flags = 0;
1209 				ifmam->ifmam_addrs = info.rti_addrs;
1210 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1211 				if (error) {
1212 					IF_ADDR_UNLOCK(ifp);
1213 					goto done;
1214 				}
1215 			}
1216 		}
1217 		IF_ADDR_UNLOCK(ifp);
1218 	}
1219 done:
1220 	IFNET_RUNLOCK();
1221 	return (error);
1222 }
1223 
1224 static int
1225 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1226 {
1227 	int	*name = (int *)arg1;
1228 	u_int	namelen = arg2;
1229 	struct radix_node_head *rnh;
1230 	int	i, lim, error = EINVAL;
1231 	u_char	af;
1232 	struct	walkarg w;
1233 
1234 	name ++;
1235 	namelen--;
1236 	if (req->newptr)
1237 		return (EPERM);
1238 	if (namelen != 3)
1239 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1240 	af = name[0];
1241 	if (af > AF_MAX)
1242 		return (EINVAL);
1243 	bzero(&w, sizeof(w));
1244 	w.w_op = name[1];
1245 	w.w_arg = name[2];
1246 	w.w_req = req;
1247 
1248 	error = sysctl_wire_old_buffer(req, 0);
1249 	if (error)
1250 		return (error);
1251 	switch (w.w_op) {
1252 
1253 	case NET_RT_DUMP:
1254 	case NET_RT_FLAGS:
1255 		if (af == 0) {			/* dump all tables */
1256 			i = 1;
1257 			lim = AF_MAX;
1258 		} else				/* dump only one table */
1259 			i = lim = af;
1260 		for (error = 0; error == 0 && i <= lim; i++)
1261 			if ((rnh = rt_tables[i]) != NULL) {
1262 				RADIX_NODE_HEAD_LOCK(rnh);
1263 			    	error = rnh->rnh_walktree(rnh,
1264 				    sysctl_dumpentry, &w);
1265 				RADIX_NODE_HEAD_UNLOCK(rnh);
1266 			} else if (af != 0)
1267 				error = EAFNOSUPPORT;
1268 		break;
1269 
1270 	case NET_RT_IFLIST:
1271 		error = sysctl_iflist(af, &w);
1272 		break;
1273 
1274 	case NET_RT_IFMALIST:
1275 		error = sysctl_ifmalist(af, &w);
1276 		break;
1277 	}
1278 	if (w.w_tmem)
1279 		free(w.w_tmem, M_RTABLE);
1280 	return (error);
1281 }
1282 
1283 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1284 
1285 /*
1286  * Definitions of protocols supported in the ROUTE domain.
1287  */
1288 
1289 static struct domain routedomain;		/* or at least forward */
1290 
1291 static struct protosw routesw[] = {
1292 {
1293 	.pr_type =		SOCK_RAW,
1294 	.pr_domain =		&routedomain,
1295 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1296 	.pr_output =		route_output,
1297 	.pr_ctlinput =		raw_ctlinput,
1298 	.pr_init =		raw_init,
1299 	.pr_usrreqs =		&route_usrreqs
1300 }
1301 };
1302 
1303 static struct domain routedomain = {
1304 	.dom_family =		PF_ROUTE,
1305 	.dom_name =		 "route",
1306 	.dom_protosw =		routesw,
1307 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1308 };
1309 
1310 DOMAIN_SET(route);
1311