xref: /freebsd/sys/net/rtsock.c (revision 271c3a9060f2ee55607ebe146523f888e1db2654)
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 
35 #include <sys/param.h>
36 #include <sys/domain.h>
37 #include <sys/kernel.h>
38 #include <sys/jail.h>
39 #include <sys/malloc.h>
40 #include <sys/mbuf.h>
41 #include <sys/priv.h>
42 #include <sys/proc.h>
43 #include <sys/protosw.h>
44 #include <sys/signalvar.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <sys/sysctl.h>
48 #include <sys/systm.h>
49 #include <sys/vimage.h>
50 
51 #include <net/if.h>
52 #include <net/netisr.h>
53 #include <net/raw_cb.h>
54 #include <net/route.h>
55 
56 #include <netinet/in.h>
57 
58 #ifdef SCTP
59 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
60 #endif /* SCTP */
61 
62 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
63 
64 /* NB: these are not modified */
65 static struct	sockaddr route_src = { 2, PF_ROUTE, };
66 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
67 
68 static struct {
69 	int	ip_count;	/* attached w/ AF_INET */
70 	int	ip6_count;	/* attached w/ AF_INET6 */
71 	int	ipx_count;	/* attached w/ AF_IPX */
72 	int	any_count;	/* total attached */
73 } route_cb;
74 
75 struct mtx rtsock_mtx;
76 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
77 
78 #define	RTSOCK_LOCK()	mtx_lock(&rtsock_mtx)
79 #define	RTSOCK_UNLOCK()	mtx_unlock(&rtsock_mtx)
80 #define	RTSOCK_LOCK_ASSERT()	mtx_assert(&rtsock_mtx, MA_OWNED)
81 
82 static struct	ifqueue rtsintrq;
83 
84 SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
85 SYSCTL_INT(_net_route, OID_AUTO, netisr_maxqlen, CTLFLAG_RW,
86     &rtsintrq.ifq_maxlen, 0, "maximum routing socket dispatch queue length");
87 
88 struct walkarg {
89 	int	w_tmemsize;
90 	int	w_op, w_arg;
91 	caddr_t	w_tmem;
92 	struct sysctl_req *w_req;
93 };
94 
95 static void	rts_input(struct mbuf *m);
96 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
97 static int	rt_msg2(int type, struct rt_addrinfo *rtinfo,
98 			caddr_t cp, struct walkarg *w);
99 static int	rt_xaddrs(caddr_t cp, caddr_t cplim,
100 			struct rt_addrinfo *rtinfo);
101 static int	sysctl_dumpentry(struct radix_node *rn, void *vw);
102 static int	sysctl_iflist(int af, struct walkarg *w);
103 static int	sysctl_ifmalist(int af, struct walkarg *w);
104 static int	route_output(struct mbuf *m, struct socket *so);
105 static void	rt_setmetrics(u_long which, const struct rt_metrics *in,
106 			struct rt_metrics_lite *out);
107 static void	rt_getmetrics(const struct rt_metrics_lite *in,
108 			struct rt_metrics *out);
109 static void	rt_dispatch(struct mbuf *, const struct sockaddr *);
110 
111 static void
112 rts_init(void)
113 {
114 	int tmp;
115 
116 	rtsintrq.ifq_maxlen = 256;
117 	if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
118 		rtsintrq.ifq_maxlen = tmp;
119 	mtx_init(&rtsintrq.ifq_mtx, "rts_inq", NULL, MTX_DEF);
120 	netisr_register(NETISR_ROUTE, rts_input, &rtsintrq, 0);
121 }
122 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
123 
124 static void
125 rts_input(struct mbuf *m)
126 {
127 	struct sockproto route_proto;
128 	unsigned short *family;
129 	struct m_tag *tag;
130 
131 	route_proto.sp_family = PF_ROUTE;
132 	tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
133 	if (tag != NULL) {
134 		family = (unsigned short *)(tag + 1);
135 		route_proto.sp_protocol = *family;
136 		m_tag_delete(m, tag);
137 	} else
138 		route_proto.sp_protocol = 0;
139 
140 	raw_input(m, &route_proto, &route_src);
141 }
142 
143 /*
144  * It really doesn't make any sense at all for this code to share much
145  * with raw_usrreq.c, since its functionality is so restricted.  XXX
146  */
147 static void
148 rts_abort(struct socket *so)
149 {
150 
151 	raw_usrreqs.pru_abort(so);
152 }
153 
154 static void
155 rts_close(struct socket *so)
156 {
157 
158 	raw_usrreqs.pru_close(so);
159 }
160 
161 /* pru_accept is EOPNOTSUPP */
162 
163 static int
164 rts_attach(struct socket *so, int proto, struct thread *td)
165 {
166 	struct rawcb *rp;
167 	int s, error;
168 
169 	KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
170 
171 	/* XXX */
172 	MALLOC(rp, struct rawcb *, sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
173 	if (rp == NULL)
174 		return ENOBUFS;
175 
176 	/*
177 	 * The splnet() is necessary to block protocols from sending
178 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
179 	 * this PCB is extant but incompletely initialized.
180 	 * Probably we should try to do more of this work beforehand and
181 	 * eliminate the spl.
182 	 */
183 	s = splnet();
184 	so->so_pcb = (caddr_t)rp;
185 	so->so_fibnum = td->td_proc->p_fibnum;
186 	error = raw_attach(so, proto);
187 	rp = sotorawcb(so);
188 	if (error) {
189 		splx(s);
190 		so->so_pcb = NULL;
191 		free(rp, M_PCB);
192 		return error;
193 	}
194 	RTSOCK_LOCK();
195 	switch(rp->rcb_proto.sp_protocol) {
196 	case AF_INET:
197 		route_cb.ip_count++;
198 		break;
199 	case AF_INET6:
200 		route_cb.ip6_count++;
201 		break;
202 	case AF_IPX:
203 		route_cb.ipx_count++;
204 		break;
205 	}
206 	route_cb.any_count++;
207 	RTSOCK_UNLOCK();
208 	soisconnected(so);
209 	so->so_options |= SO_USELOOPBACK;
210 	splx(s);
211 	return 0;
212 }
213 
214 static int
215 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
216 {
217 
218 	return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
219 }
220 
221 static int
222 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
223 {
224 
225 	return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
226 }
227 
228 /* pru_connect2 is EOPNOTSUPP */
229 /* pru_control is EOPNOTSUPP */
230 
231 static void
232 rts_detach(struct socket *so)
233 {
234 	struct rawcb *rp = sotorawcb(so);
235 
236 	KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
237 
238 	RTSOCK_LOCK();
239 	switch(rp->rcb_proto.sp_protocol) {
240 	case AF_INET:
241 		route_cb.ip_count--;
242 		break;
243 	case AF_INET6:
244 		route_cb.ip6_count--;
245 		break;
246 	case AF_IPX:
247 		route_cb.ipx_count--;
248 		break;
249 	}
250 	route_cb.any_count--;
251 	RTSOCK_UNLOCK();
252 	raw_usrreqs.pru_detach(so);
253 }
254 
255 static int
256 rts_disconnect(struct socket *so)
257 {
258 
259 	return (raw_usrreqs.pru_disconnect(so));
260 }
261 
262 /* pru_listen is EOPNOTSUPP */
263 
264 static int
265 rts_peeraddr(struct socket *so, struct sockaddr **nam)
266 {
267 
268 	return (raw_usrreqs.pru_peeraddr(so, nam));
269 }
270 
271 /* pru_rcvd is EOPNOTSUPP */
272 /* pru_rcvoob is EOPNOTSUPP */
273 
274 static int
275 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
276 	 struct mbuf *control, struct thread *td)
277 {
278 
279 	return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
280 }
281 
282 /* pru_sense is null */
283 
284 static int
285 rts_shutdown(struct socket *so)
286 {
287 
288 	return (raw_usrreqs.pru_shutdown(so));
289 }
290 
291 static int
292 rts_sockaddr(struct socket *so, struct sockaddr **nam)
293 {
294 
295 	return (raw_usrreqs.pru_sockaddr(so, nam));
296 }
297 
298 static struct pr_usrreqs route_usrreqs = {
299 	.pru_abort =		rts_abort,
300 	.pru_attach =		rts_attach,
301 	.pru_bind =		rts_bind,
302 	.pru_connect =		rts_connect,
303 	.pru_detach =		rts_detach,
304 	.pru_disconnect =	rts_disconnect,
305 	.pru_peeraddr =		rts_peeraddr,
306 	.pru_send =		rts_send,
307 	.pru_shutdown =		rts_shutdown,
308 	.pru_sockaddr =		rts_sockaddr,
309 	.pru_close =		rts_close,
310 };
311 
312 /*ARGSUSED*/
313 static int
314 route_output(struct mbuf *m, struct socket *so)
315 {
316 #define	sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
317 	struct rt_msghdr *rtm = NULL;
318 	struct rtentry *rt = NULL;
319 	struct radix_node_head *rnh;
320 	struct rt_addrinfo info;
321 	int len, error = 0;
322 	struct ifnet *ifp = NULL;
323 	struct sockaddr_in jail;
324 
325 #define senderr(e) { error = e; goto flush;}
326 	if (m == NULL || ((m->m_len < sizeof(long)) &&
327 		       (m = m_pullup(m, sizeof(long))) == NULL))
328 		return (ENOBUFS);
329 	if ((m->m_flags & M_PKTHDR) == 0)
330 		panic("route_output");
331 	len = m->m_pkthdr.len;
332 	if (len < sizeof(*rtm) ||
333 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
334 		info.rti_info[RTAX_DST] = NULL;
335 		senderr(EINVAL);
336 	}
337 	R_Malloc(rtm, struct rt_msghdr *, len);
338 	if (rtm == NULL) {
339 		info.rti_info[RTAX_DST] = NULL;
340 		senderr(ENOBUFS);
341 	}
342 	m_copydata(m, 0, len, (caddr_t)rtm);
343 	if (rtm->rtm_version != RTM_VERSION) {
344 		info.rti_info[RTAX_DST] = NULL;
345 		senderr(EPROTONOSUPPORT);
346 	}
347 	rtm->rtm_pid = curproc->p_pid;
348 	bzero(&info, sizeof(info));
349 	info.rti_addrs = rtm->rtm_addrs;
350 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
351 		info.rti_info[RTAX_DST] = NULL;
352 		senderr(EINVAL);
353 	}
354 	info.rti_flags = rtm->rtm_flags;
355 	if (info.rti_info[RTAX_DST] == NULL ||
356 	    info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
357 	    (info.rti_info[RTAX_GATEWAY] != NULL &&
358 	     info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
359 		senderr(EINVAL);
360 	if (info.rti_info[RTAX_GENMASK]) {
361 		struct radix_node *t;
362 		t = rn_addmask((caddr_t) info.rti_info[RTAX_GENMASK], 0, 1);
363 		if (t != NULL &&
364 		    bcmp((char *)(void *)info.rti_info[RTAX_GENMASK] + 1,
365 		    (char *)(void *)t->rn_key + 1,
366 		    ((struct sockaddr *)t->rn_key)->sa_len - 1) == 0)
367 			info.rti_info[RTAX_GENMASK] =
368 			    (struct sockaddr *)t->rn_key;
369 		else
370 			senderr(ENOBUFS);
371 	}
372 
373 	/*
374 	 * Verify that the caller has the appropriate privilege; RTM_GET
375 	 * is the only operation the non-superuser is allowed.
376 	 */
377 	if (rtm->rtm_type != RTM_GET) {
378 		error = priv_check(curthread, PRIV_NET_ROUTE);
379 		if (error)
380 			senderr(error);
381 	}
382 
383 	switch (rtm->rtm_type) {
384 		struct rtentry *saved_nrt;
385 
386 	case RTM_ADD:
387 		if (info.rti_info[RTAX_GATEWAY] == NULL)
388 			senderr(EINVAL);
389 		saved_nrt = NULL;
390 		error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
391 		    so->so_fibnum);
392 		if (error == 0 && saved_nrt) {
393 			RT_LOCK(saved_nrt);
394 			rt_setmetrics(rtm->rtm_inits,
395 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
396 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
397 			RT_REMREF(saved_nrt);
398 			saved_nrt->rt_genmask = info.rti_info[RTAX_GENMASK];
399 			RT_UNLOCK(saved_nrt);
400 		}
401 		break;
402 
403 	case RTM_DELETE:
404 		saved_nrt = NULL;
405 		error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
406 		    so->so_fibnum);
407 		if (error == 0) {
408 			RT_LOCK(saved_nrt);
409 			rt = saved_nrt;
410 			goto report;
411 		}
412 		break;
413 
414 	case RTM_GET:
415 	case RTM_CHANGE:
416 	case RTM_LOCK:
417 		rnh = V_rt_tables[so->so_fibnum][info.rti_info[RTAX_DST]->sa_family];
418 		if (rnh == NULL)
419 			senderr(EAFNOSUPPORT);
420 		RADIX_NODE_HEAD_LOCK(rnh);
421 		rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
422 			info.rti_info[RTAX_NETMASK], rnh);
423 		if (rt == NULL) {	/* XXX looks bogus */
424 			RADIX_NODE_HEAD_UNLOCK(rnh);
425 			senderr(ESRCH);
426 		}
427 #ifdef RADIX_MPATH
428 		/*
429 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
430 		 * we require users to specify a matching RTAX_GATEWAY.
431 		 *
432 		 * for RTM_GET, gate is optional even with multipath.
433 		 * if gate == NULL the first match is returned.
434 		 * (no need to call rt_mpath_matchgate if gate == NULL)
435 		 */
436 		if (rn_mpath_capable(rnh) &&
437 		    (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
438 			rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
439 			if (!rt) {
440 				RADIX_NODE_HEAD_UNLOCK(rnh);
441 				senderr(ESRCH);
442 			}
443 		}
444 #endif
445 		RT_LOCK(rt);
446 		RT_ADDREF(rt);
447 		RADIX_NODE_HEAD_UNLOCK(rnh);
448 
449 		/*
450 		 * Fix for PR: 82974
451 		 *
452 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
453 		 * returns a perfect match in case a netmask is
454 		 * specified.  For host routes only a longest prefix
455 		 * match is returned so it is necessary to compare the
456 		 * existence of the netmask.  If both have a netmask
457 		 * rnh_lookup() did a perfect match and if none of them
458 		 * have a netmask both are host routes which is also a
459 		 * perfect match.
460 		 */
461 
462 		if (rtm->rtm_type != RTM_GET &&
463 		    (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
464 			RT_UNLOCK(rt);
465 			senderr(ESRCH);
466 		}
467 
468 		switch(rtm->rtm_type) {
469 
470 		case RTM_GET:
471 		report:
472 			RT_LOCK_ASSERT(rt);
473 			info.rti_info[RTAX_DST] = rt_key(rt);
474 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
475 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
476 			info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
477 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
478 				ifp = rt->rt_ifp;
479 				if (ifp) {
480 					info.rti_info[RTAX_IFP] =
481 					    ifp->if_addr->ifa_addr;
482 					if (jailed(so->so_cred)) {
483 						bzero(&jail, sizeof(jail));
484 						jail.sin_family = PF_INET;
485 						jail.sin_len = sizeof(jail);
486 						jail.sin_addr.s_addr =
487 						htonl(prison_getip(so->so_cred));
488 						info.rti_info[RTAX_IFA] =
489 						    (struct sockaddr *)&jail;
490 					} else
491 						info.rti_info[RTAX_IFA] =
492 						    rt->rt_ifa->ifa_addr;
493 					if (ifp->if_flags & IFF_POINTOPOINT)
494 						info.rti_info[RTAX_BRD] =
495 						    rt->rt_ifa->ifa_dstaddr;
496 					rtm->rtm_index = ifp->if_index;
497 				} else {
498 					info.rti_info[RTAX_IFP] = NULL;
499 					info.rti_info[RTAX_IFA] = NULL;
500 				}
501 			} else if ((ifp = rt->rt_ifp) != NULL) {
502 				rtm->rtm_index = ifp->if_index;
503 			}
504 			len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
505 			if (len > rtm->rtm_msglen) {
506 				struct rt_msghdr *new_rtm;
507 				R_Malloc(new_rtm, struct rt_msghdr *, len);
508 				if (new_rtm == NULL) {
509 					RT_UNLOCK(rt);
510 					senderr(ENOBUFS);
511 				}
512 				bcopy(rtm, new_rtm, rtm->rtm_msglen);
513 				Free(rtm); rtm = new_rtm;
514 			}
515 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
516 			rtm->rtm_flags = rt->rt_flags;
517 			rtm->rtm_use = 0;
518 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
519 			rtm->rtm_addrs = info.rti_addrs;
520 			break;
521 
522 		case RTM_CHANGE:
523 			/*
524 			 * New gateway could require new ifaddr, ifp;
525 			 * flags may also be different; ifp may be specified
526 			 * by ll sockaddr when protocol address is ambiguous
527 			 */
528 			if (((rt->rt_flags & RTF_GATEWAY) &&
529 			     info.rti_info[RTAX_GATEWAY] != NULL) ||
530 			    info.rti_info[RTAX_IFP] != NULL ||
531 			    (info.rti_info[RTAX_IFA] != NULL &&
532 			     !sa_equal(info.rti_info[RTAX_IFA],
533 				       rt->rt_ifa->ifa_addr))) {
534 				RT_UNLOCK(rt);
535 				if ((error = rt_getifa_fib(&info,
536 				    rt->rt_fibnum)) != 0)
537 					senderr(error);
538 				RT_LOCK(rt);
539 			}
540 			if (info.rti_ifa != NULL &&
541 			    info.rti_ifa != rt->rt_ifa &&
542 			    rt->rt_ifa != NULL &&
543 			    rt->rt_ifa->ifa_rtrequest != NULL) {
544 				rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
545 				    &info);
546 				IFAFREE(rt->rt_ifa);
547 			}
548 			if (info.rti_info[RTAX_GATEWAY] != NULL) {
549 				if ((error = rt_setgate(rt, rt_key(rt),
550 					info.rti_info[RTAX_GATEWAY])) != 0) {
551 					RT_UNLOCK(rt);
552 					senderr(error);
553 				}
554 				if (!(rt->rt_flags & RTF_LLINFO))
555 					rt->rt_flags |= RTF_GATEWAY;
556 			}
557 			if (info.rti_ifa != NULL &&
558 			    info.rti_ifa != rt->rt_ifa) {
559 				IFAREF(info.rti_ifa);
560 				rt->rt_ifa = info.rti_ifa;
561 				rt->rt_ifp = info.rti_ifp;
562 			}
563 			/* Allow some flags to be toggled on change. */
564 			if (rtm->rtm_fmask & RTF_FMASK)
565 				rt->rt_flags = (rt->rt_flags &
566 				    ~rtm->rtm_fmask) |
567 				    (rtm->rtm_flags & rtm->rtm_fmask);
568 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
569 					&rt->rt_rmx);
570 			rtm->rtm_index = rt->rt_ifp->if_index;
571 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
572 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
573 			if (info.rti_info[RTAX_GENMASK])
574 				rt->rt_genmask = info.rti_info[RTAX_GENMASK];
575 			/* FALLTHROUGH */
576 		case RTM_LOCK:
577 			/* We don't support locks anymore */
578 			break;
579 		}
580 		RT_UNLOCK(rt);
581 		break;
582 
583 	default:
584 		senderr(EOPNOTSUPP);
585 	}
586 
587 flush:
588 	if (rtm) {
589 		if (error)
590 			rtm->rtm_errno = error;
591 		else
592 			rtm->rtm_flags |= RTF_DONE;
593 	}
594 	if (rt)		/* XXX can this be true? */
595 		RTFREE(rt);
596     {
597 	struct rawcb *rp = NULL;
598 	/*
599 	 * Check to see if we don't want our own messages.
600 	 */
601 	if ((so->so_options & SO_USELOOPBACK) == 0) {
602 		if (route_cb.any_count <= 1) {
603 			if (rtm)
604 				Free(rtm);
605 			m_freem(m);
606 			return (error);
607 		}
608 		/* There is another listener, so construct message */
609 		rp = sotorawcb(so);
610 	}
611 	if (rtm) {
612 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
613 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
614 			m_freem(m);
615 			m = NULL;
616 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
617 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
618 		Free(rtm);
619 	}
620 	if (m) {
621 		if (rp) {
622 			/*
623 			 * XXX insure we don't get a copy by
624 			 * invalidating our protocol
625 			 */
626 			unsigned short family = rp->rcb_proto.sp_family;
627 			rp->rcb_proto.sp_family = 0;
628 			rt_dispatch(m, info.rti_info[RTAX_DST]);
629 			rp->rcb_proto.sp_family = family;
630 		} else
631 			rt_dispatch(m, info.rti_info[RTAX_DST]);
632 	}
633     }
634 	return (error);
635 #undef	sa_equal
636 }
637 
638 static void
639 rt_setmetrics(u_long which, const struct rt_metrics *in,
640 	struct rt_metrics_lite *out)
641 {
642 #define metric(f, e) if (which & (f)) out->e = in->e;
643 	/*
644 	 * Only these are stored in the routing entry since introduction
645 	 * of tcp hostcache. The rest is ignored.
646 	 */
647 	metric(RTV_MTU, rmx_mtu);
648 	/* Userland -> kernel timebase conversion. */
649 	if (which & RTV_EXPIRE)
650 		out->rmx_expire = in->rmx_expire ?
651 		    in->rmx_expire - time_second + time_uptime : 0;
652 #undef metric
653 }
654 
655 static void
656 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
657 {
658 #define metric(e) out->e = in->e;
659 	bzero(out, sizeof(*out));
660 	metric(rmx_mtu);
661 	/* Kernel -> userland timebase conversion. */
662 	out->rmx_expire = in->rmx_expire ?
663 	    in->rmx_expire - time_uptime + time_second : 0;
664 #undef metric
665 }
666 
667 /*
668  * Extract the addresses of the passed sockaddrs.
669  * Do a little sanity checking so as to avoid bad memory references.
670  * This data is derived straight from userland.
671  */
672 static int
673 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
674 {
675 	struct sockaddr *sa;
676 	int i;
677 
678 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
679 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
680 			continue;
681 		sa = (struct sockaddr *)cp;
682 		/*
683 		 * It won't fit.
684 		 */
685 		if (cp + sa->sa_len > cplim)
686 			return (EINVAL);
687 		/*
688 		 * there are no more.. quit now
689 		 * If there are more bits, they are in error.
690 		 * I've seen this. route(1) can evidently generate these.
691 		 * This causes kernel to core dump.
692 		 * for compatibility, If we see this, point to a safe address.
693 		 */
694 		if (sa->sa_len == 0) {
695 			rtinfo->rti_info[i] = &sa_zero;
696 			return (0); /* should be EINVAL but for compat */
697 		}
698 		/* accept it */
699 		rtinfo->rti_info[i] = sa;
700 		cp += SA_SIZE(sa);
701 	}
702 	return (0);
703 }
704 
705 static struct mbuf *
706 rt_msg1(int type, struct rt_addrinfo *rtinfo)
707 {
708 	struct rt_msghdr *rtm;
709 	struct mbuf *m;
710 	int i;
711 	struct sockaddr *sa;
712 	int len, dlen;
713 
714 	switch (type) {
715 
716 	case RTM_DELADDR:
717 	case RTM_NEWADDR:
718 		len = sizeof(struct ifa_msghdr);
719 		break;
720 
721 	case RTM_DELMADDR:
722 	case RTM_NEWMADDR:
723 		len = sizeof(struct ifma_msghdr);
724 		break;
725 
726 	case RTM_IFINFO:
727 		len = sizeof(struct if_msghdr);
728 		break;
729 
730 	case RTM_IFANNOUNCE:
731 	case RTM_IEEE80211:
732 		len = sizeof(struct if_announcemsghdr);
733 		break;
734 
735 	default:
736 		len = sizeof(struct rt_msghdr);
737 	}
738 	if (len > MCLBYTES)
739 		panic("rt_msg1");
740 	m = m_gethdr(M_DONTWAIT, MT_DATA);
741 	if (m && len > MHLEN) {
742 		MCLGET(m, M_DONTWAIT);
743 		if ((m->m_flags & M_EXT) == 0) {
744 			m_free(m);
745 			m = NULL;
746 		}
747 	}
748 	if (m == NULL)
749 		return (m);
750 	m->m_pkthdr.len = m->m_len = len;
751 	m->m_pkthdr.rcvif = NULL;
752 	rtm = mtod(m, struct rt_msghdr *);
753 	bzero((caddr_t)rtm, len);
754 	for (i = 0; i < RTAX_MAX; i++) {
755 		if ((sa = rtinfo->rti_info[i]) == NULL)
756 			continue;
757 		rtinfo->rti_addrs |= (1 << i);
758 		dlen = SA_SIZE(sa);
759 		m_copyback(m, len, dlen, (caddr_t)sa);
760 		len += dlen;
761 	}
762 	if (m->m_pkthdr.len != len) {
763 		m_freem(m);
764 		return (NULL);
765 	}
766 	rtm->rtm_msglen = len;
767 	rtm->rtm_version = RTM_VERSION;
768 	rtm->rtm_type = type;
769 	return (m);
770 }
771 
772 static int
773 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
774 {
775 	int i;
776 	int len, dlen, second_time = 0;
777 	caddr_t cp0;
778 
779 	rtinfo->rti_addrs = 0;
780 again:
781 	switch (type) {
782 
783 	case RTM_DELADDR:
784 	case RTM_NEWADDR:
785 		len = sizeof(struct ifa_msghdr);
786 		break;
787 
788 	case RTM_IFINFO:
789 		len = sizeof(struct if_msghdr);
790 		break;
791 
792 	case RTM_NEWMADDR:
793 		len = sizeof(struct ifma_msghdr);
794 		break;
795 
796 	default:
797 		len = sizeof(struct rt_msghdr);
798 	}
799 	cp0 = cp;
800 	if (cp0)
801 		cp += len;
802 	for (i = 0; i < RTAX_MAX; i++) {
803 		struct sockaddr *sa;
804 
805 		if ((sa = rtinfo->rti_info[i]) == NULL)
806 			continue;
807 		rtinfo->rti_addrs |= (1 << i);
808 		dlen = SA_SIZE(sa);
809 		if (cp) {
810 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
811 			cp += dlen;
812 		}
813 		len += dlen;
814 	}
815 	len = ALIGN(len);
816 	if (cp == NULL && w != NULL && !second_time) {
817 		struct walkarg *rw = w;
818 
819 		if (rw->w_req) {
820 			if (rw->w_tmemsize < len) {
821 				if (rw->w_tmem)
822 					free(rw->w_tmem, M_RTABLE);
823 				rw->w_tmem = (caddr_t)
824 					malloc(len, M_RTABLE, M_NOWAIT);
825 				if (rw->w_tmem)
826 					rw->w_tmemsize = len;
827 			}
828 			if (rw->w_tmem) {
829 				cp = rw->w_tmem;
830 				second_time = 1;
831 				goto again;
832 			}
833 		}
834 	}
835 	if (cp) {
836 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
837 
838 		rtm->rtm_version = RTM_VERSION;
839 		rtm->rtm_type = type;
840 		rtm->rtm_msglen = len;
841 	}
842 	return (len);
843 }
844 
845 /*
846  * This routine is called to generate a message from the routing
847  * socket indicating that a redirect has occured, a routing lookup
848  * has failed, or that a protocol has detected timeouts to a particular
849  * destination.
850  */
851 void
852 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
853 {
854 	struct rt_msghdr *rtm;
855 	struct mbuf *m;
856 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
857 
858 	if (route_cb.any_count == 0)
859 		return;
860 	m = rt_msg1(type, rtinfo);
861 	if (m == NULL)
862 		return;
863 	rtm = mtod(m, struct rt_msghdr *);
864 	rtm->rtm_flags = RTF_DONE | flags;
865 	rtm->rtm_errno = error;
866 	rtm->rtm_addrs = rtinfo->rti_addrs;
867 	rt_dispatch(m, sa);
868 }
869 
870 /*
871  * This routine is called to generate a message from the routing
872  * socket indicating that the status of a network interface has changed.
873  */
874 void
875 rt_ifmsg(struct ifnet *ifp)
876 {
877 	struct if_msghdr *ifm;
878 	struct mbuf *m;
879 	struct rt_addrinfo info;
880 
881 	if (route_cb.any_count == 0)
882 		return;
883 	bzero((caddr_t)&info, sizeof(info));
884 	m = rt_msg1(RTM_IFINFO, &info);
885 	if (m == NULL)
886 		return;
887 	ifm = mtod(m, struct if_msghdr *);
888 	ifm->ifm_index = ifp->if_index;
889 	ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
890 	ifm->ifm_data = ifp->if_data;
891 	ifm->ifm_addrs = 0;
892 	rt_dispatch(m, NULL);
893 }
894 
895 /*
896  * This is called to generate messages from the routing socket
897  * indicating a network interface has had addresses associated with it.
898  * if we ever reverse the logic and replace messages TO the routing
899  * socket indicate a request to configure interfaces, then it will
900  * be unnecessary as the routing socket will automatically generate
901  * copies of it.
902  */
903 void
904 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
905 {
906 	struct rt_addrinfo info;
907 	struct sockaddr *sa = NULL;
908 	int pass;
909 	struct mbuf *m = NULL;
910 	struct ifnet *ifp = ifa->ifa_ifp;
911 
912 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
913 		("unexpected cmd %u", cmd));
914 #ifdef SCTP
915 	/*
916 	 * notify the SCTP stack
917 	 * this will only get called when an address is added/deleted
918 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
919 	 */
920 	sctp_addr_change(ifa, cmd);
921 #endif /* SCTP */
922 	if (route_cb.any_count == 0)
923 		return;
924 	for (pass = 1; pass < 3; pass++) {
925 		bzero((caddr_t)&info, sizeof(info));
926 		if ((cmd == RTM_ADD && pass == 1) ||
927 		    (cmd == RTM_DELETE && pass == 2)) {
928 			struct ifa_msghdr *ifam;
929 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
930 
931 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
932 			info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
933 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
934 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
935 			if ((m = rt_msg1(ncmd, &info)) == NULL)
936 				continue;
937 			ifam = mtod(m, struct ifa_msghdr *);
938 			ifam->ifam_index = ifp->if_index;
939 			ifam->ifam_metric = ifa->ifa_metric;
940 			ifam->ifam_flags = ifa->ifa_flags;
941 			ifam->ifam_addrs = info.rti_addrs;
942 		}
943 		if ((cmd == RTM_ADD && pass == 2) ||
944 		    (cmd == RTM_DELETE && pass == 1)) {
945 			struct rt_msghdr *rtm;
946 
947 			if (rt == NULL)
948 				continue;
949 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
950 			info.rti_info[RTAX_DST] = sa = rt_key(rt);
951 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
952 			if ((m = rt_msg1(cmd, &info)) == NULL)
953 				continue;
954 			rtm = mtod(m, struct rt_msghdr *);
955 			rtm->rtm_index = ifp->if_index;
956 			rtm->rtm_flags |= rt->rt_flags;
957 			rtm->rtm_errno = error;
958 			rtm->rtm_addrs = info.rti_addrs;
959 		}
960 		rt_dispatch(m, sa);
961 	}
962 }
963 
964 /*
965  * This is the analogue to the rt_newaddrmsg which performs the same
966  * function but for multicast group memberhips.  This is easier since
967  * there is no route state to worry about.
968  */
969 void
970 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
971 {
972 	struct rt_addrinfo info;
973 	struct mbuf *m = NULL;
974 	struct ifnet *ifp = ifma->ifma_ifp;
975 	struct ifma_msghdr *ifmam;
976 
977 	if (route_cb.any_count == 0)
978 		return;
979 
980 	bzero((caddr_t)&info, sizeof(info));
981 	info.rti_info[RTAX_IFA] = ifma->ifma_addr;
982 	info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
983 	/*
984 	 * If a link-layer address is present, present it as a ``gateway''
985 	 * (similarly to how ARP entries, e.g., are presented).
986 	 */
987 	info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
988 	m = rt_msg1(cmd, &info);
989 	if (m == NULL)
990 		return;
991 	ifmam = mtod(m, struct ifma_msghdr *);
992 	KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
993 	    __func__));
994 	ifmam->ifmam_index = ifp->if_index;
995 	ifmam->ifmam_addrs = info.rti_addrs;
996 	rt_dispatch(m, ifma->ifma_addr);
997 }
998 
999 static struct mbuf *
1000 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1001 	struct rt_addrinfo *info)
1002 {
1003 	struct if_announcemsghdr *ifan;
1004 	struct mbuf *m;
1005 
1006 	if (route_cb.any_count == 0)
1007 		return NULL;
1008 	bzero((caddr_t)info, sizeof(*info));
1009 	m = rt_msg1(type, info);
1010 	if (m != NULL) {
1011 		ifan = mtod(m, struct if_announcemsghdr *);
1012 		ifan->ifan_index = ifp->if_index;
1013 		strlcpy(ifan->ifan_name, ifp->if_xname,
1014 			sizeof(ifan->ifan_name));
1015 		ifan->ifan_what = what;
1016 	}
1017 	return m;
1018 }
1019 
1020 /*
1021  * This is called to generate routing socket messages indicating
1022  * IEEE80211 wireless events.
1023  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1024  */
1025 void
1026 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1027 {
1028 	struct mbuf *m;
1029 	struct rt_addrinfo info;
1030 
1031 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1032 	if (m != NULL) {
1033 		/*
1034 		 * Append the ieee80211 data.  Try to stick it in the
1035 		 * mbuf containing the ifannounce msg; otherwise allocate
1036 		 * a new mbuf and append.
1037 		 *
1038 		 * NB: we assume m is a single mbuf.
1039 		 */
1040 		if (data_len > M_TRAILINGSPACE(m)) {
1041 			struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1042 			if (n == NULL) {
1043 				m_freem(m);
1044 				return;
1045 			}
1046 			bcopy(data, mtod(n, void *), data_len);
1047 			n->m_len = data_len;
1048 			m->m_next = n;
1049 		} else if (data_len > 0) {
1050 			bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1051 			m->m_len += data_len;
1052 		}
1053 		if (m->m_flags & M_PKTHDR)
1054 			m->m_pkthdr.len += data_len;
1055 		mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1056 		rt_dispatch(m, NULL);
1057 	}
1058 }
1059 
1060 /*
1061  * This is called to generate routing socket messages indicating
1062  * network interface arrival and departure.
1063  */
1064 void
1065 rt_ifannouncemsg(struct ifnet *ifp, int what)
1066 {
1067 	struct mbuf *m;
1068 	struct rt_addrinfo info;
1069 
1070 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1071 	if (m != NULL)
1072 		rt_dispatch(m, NULL);
1073 }
1074 
1075 static void
1076 rt_dispatch(struct mbuf *m, const struct sockaddr *sa)
1077 {
1078 	struct m_tag *tag;
1079 
1080 	/*
1081 	 * Preserve the family from the sockaddr, if any, in an m_tag for
1082 	 * use when injecting the mbuf into the routing socket buffer from
1083 	 * the netisr.
1084 	 */
1085 	if (sa != NULL) {
1086 		tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1087 		    M_NOWAIT);
1088 		if (tag == NULL) {
1089 			m_freem(m);
1090 			return;
1091 		}
1092 		*(unsigned short *)(tag + 1) = sa->sa_family;
1093 		m_tag_prepend(m, tag);
1094 	}
1095 	netisr_queue(NETISR_ROUTE, m);	/* mbuf is free'd on failure. */
1096 }
1097 
1098 /*
1099  * This is used in dumping the kernel table via sysctl().
1100  */
1101 static int
1102 sysctl_dumpentry(struct radix_node *rn, void *vw)
1103 {
1104 	struct walkarg *w = vw;
1105 	struct rtentry *rt = (struct rtentry *)rn;
1106 	int error = 0, size;
1107 	struct rt_addrinfo info;
1108 
1109 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1110 		return 0;
1111 	bzero((caddr_t)&info, sizeof(info));
1112 	info.rti_info[RTAX_DST] = rt_key(rt);
1113 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1114 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1115 	info.rti_info[RTAX_GENMASK] = rt->rt_genmask;
1116 	if (rt->rt_ifp) {
1117 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1118 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1119 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1120 			info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1121 	}
1122 	size = rt_msg2(RTM_GET, &info, NULL, w);
1123 	if (w->w_req && w->w_tmem) {
1124 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1125 
1126 		rtm->rtm_flags = rt->rt_flags;
1127 		rtm->rtm_use = rt->rt_rmx.rmx_pksent;
1128 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1129 		rtm->rtm_index = rt->rt_ifp->if_index;
1130 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1131 		rtm->rtm_addrs = info.rti_addrs;
1132 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1133 		return (error);
1134 	}
1135 	return (error);
1136 }
1137 
1138 static int
1139 sysctl_iflist(int af, struct walkarg *w)
1140 {
1141 	struct ifnet *ifp;
1142 	struct ifaddr *ifa;
1143 	struct rt_addrinfo info;
1144 	int len, error = 0;
1145 
1146 	bzero((caddr_t)&info, sizeof(info));
1147 	IFNET_RLOCK();
1148 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1149 		if (w->w_arg && w->w_arg != ifp->if_index)
1150 			continue;
1151 		ifa = ifp->if_addr;
1152 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1153 		len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1154 		info.rti_info[RTAX_IFP] = NULL;
1155 		if (w->w_req && w->w_tmem) {
1156 			struct if_msghdr *ifm;
1157 
1158 			ifm = (struct if_msghdr *)w->w_tmem;
1159 			ifm->ifm_index = ifp->if_index;
1160 			ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1161 			ifm->ifm_data = ifp->if_data;
1162 			ifm->ifm_addrs = info.rti_addrs;
1163 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
1164 			if (error)
1165 				goto done;
1166 		}
1167 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1168 			if (af && af != ifa->ifa_addr->sa_family)
1169 				continue;
1170 			if (jailed(curthread->td_ucred) &&
1171 			    prison_if(curthread->td_ucred, ifa->ifa_addr))
1172 				continue;
1173 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1174 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1175 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1176 			len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1177 			if (w->w_req && w->w_tmem) {
1178 				struct ifa_msghdr *ifam;
1179 
1180 				ifam = (struct ifa_msghdr *)w->w_tmem;
1181 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1182 				ifam->ifam_flags = ifa->ifa_flags;
1183 				ifam->ifam_metric = ifa->ifa_metric;
1184 				ifam->ifam_addrs = info.rti_addrs;
1185 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1186 				if (error)
1187 					goto done;
1188 			}
1189 		}
1190 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1191 			info.rti_info[RTAX_BRD] = NULL;
1192 	}
1193 done:
1194 	IFNET_RUNLOCK();
1195 	return (error);
1196 }
1197 
1198 int
1199 sysctl_ifmalist(int af, struct walkarg *w)
1200 {
1201 	struct ifnet *ifp;
1202 	struct ifmultiaddr *ifma;
1203 	struct	rt_addrinfo info;
1204 	int	len, error = 0;
1205 	struct ifaddr *ifa;
1206 
1207 	bzero((caddr_t)&info, sizeof(info));
1208 	IFNET_RLOCK();
1209 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1210 		if (w->w_arg && w->w_arg != ifp->if_index)
1211 			continue;
1212 		ifa = ifp->if_addr;
1213 		info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1214 		IF_ADDR_LOCK(ifp);
1215 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1216 			if (af && af != ifma->ifma_addr->sa_family)
1217 				continue;
1218 			if (jailed(curproc->p_ucred) &&
1219 			    prison_if(curproc->p_ucred, ifma->ifma_addr))
1220 				continue;
1221 			info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1222 			info.rti_info[RTAX_GATEWAY] =
1223 			    (ifma->ifma_addr->sa_family != AF_LINK) ?
1224 			    ifma->ifma_lladdr : NULL;
1225 			len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1226 			if (w->w_req && w->w_tmem) {
1227 				struct ifma_msghdr *ifmam;
1228 
1229 				ifmam = (struct ifma_msghdr *)w->w_tmem;
1230 				ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1231 				ifmam->ifmam_flags = 0;
1232 				ifmam->ifmam_addrs = info.rti_addrs;
1233 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1234 				if (error) {
1235 					IF_ADDR_UNLOCK(ifp);
1236 					goto done;
1237 				}
1238 			}
1239 		}
1240 		IF_ADDR_UNLOCK(ifp);
1241 	}
1242 done:
1243 	IFNET_RUNLOCK();
1244 	return (error);
1245 }
1246 
1247 static int
1248 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1249 {
1250 	int	*name = (int *)arg1;
1251 	u_int	namelen = arg2;
1252 	struct radix_node_head *rnh;
1253 	int	i, lim, error = EINVAL;
1254 	u_char	af;
1255 	struct	walkarg w;
1256 
1257 	name ++;
1258 	namelen--;
1259 	if (req->newptr)
1260 		return (EPERM);
1261 	if (namelen != 3)
1262 		return ((namelen < 3) ? EISDIR : ENOTDIR);
1263 	af = name[0];
1264 	if (af > AF_MAX)
1265 		return (EINVAL);
1266 	bzero(&w, sizeof(w));
1267 	w.w_op = name[1];
1268 	w.w_arg = name[2];
1269 	w.w_req = req;
1270 
1271 	error = sysctl_wire_old_buffer(req, 0);
1272 	if (error)
1273 		return (error);
1274 	switch (w.w_op) {
1275 
1276 	case NET_RT_DUMP:
1277 	case NET_RT_FLAGS:
1278 		if (af == 0) {			/* dump all tables */
1279 			i = 1;
1280 			lim = AF_MAX;
1281 		} else				/* dump only one table */
1282 			i = lim = af;
1283 		for (error = 0; error == 0 && i <= lim; i++)
1284 			if ((rnh = V_rt_tables[curthread->td_proc->p_fibnum][i]) != NULL) {
1285 				RADIX_NODE_HEAD_LOCK(rnh);
1286 			    	error = rnh->rnh_walktree(rnh,
1287 				    sysctl_dumpentry, &w);
1288 				RADIX_NODE_HEAD_UNLOCK(rnh);
1289 			} else if (af != 0)
1290 				error = EAFNOSUPPORT;
1291 		break;
1292 
1293 	case NET_RT_IFLIST:
1294 		error = sysctl_iflist(af, &w);
1295 		break;
1296 
1297 	case NET_RT_IFMALIST:
1298 		error = sysctl_ifmalist(af, &w);
1299 		break;
1300 	}
1301 	if (w.w_tmem)
1302 		free(w.w_tmem, M_RTABLE);
1303 	return (error);
1304 }
1305 
1306 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1307 
1308 /*
1309  * Definitions of protocols supported in the ROUTE domain.
1310  */
1311 
1312 static struct domain routedomain;		/* or at least forward */
1313 
1314 static struct protosw routesw[] = {
1315 {
1316 	.pr_type =		SOCK_RAW,
1317 	.pr_domain =		&routedomain,
1318 	.pr_flags =		PR_ATOMIC|PR_ADDR,
1319 	.pr_output =		route_output,
1320 	.pr_ctlinput =		raw_ctlinput,
1321 	.pr_init =		raw_init,
1322 	.pr_usrreqs =		&route_usrreqs
1323 }
1324 };
1325 
1326 static struct domain routedomain = {
1327 	.dom_family =		PF_ROUTE,
1328 	.dom_name =		 "route",
1329 	.dom_protosw =		routesw,
1330 	.dom_protoswNPROTOSW =	&routesw[sizeof(routesw)/sizeof(routesw[0])]
1331 };
1332 
1333 DOMAIN_SET(route);
1334