xref: /dflybsd-src/sys/net/rtsock.c (revision 9b5a99654f820c32a9fb8fa9bae3c3b12fe27a0b)
1 /*
2  * Copyright (c) 2004, 2005 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Jeffrey M. Hsu.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of The DragonFly Project nor the names of its
16  *    contributors may be used to endorse or promote products derived
17  *    from this software without specific, prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
22  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
23  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
25  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
27  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
28  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
29  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1988, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
66  * $FreeBSD: src/sys/net/rtsock.c,v 1.44.2.11 2002/12/04 14:05:41 ru Exp $
67  * $DragonFly: src/sys/net/rtsock.c,v 1.39 2007/06/24 20:00:00 dillon Exp $
68  */
69 
70 #include "opt_sctp.h"
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/kernel.h>
75 #include <sys/sysctl.h>
76 #include <sys/proc.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/domain.h>
83 #include <sys/thread2.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 #include <net/raw_cb.h>
88 #include <net/netmsg2.h>
89 
90 #ifdef SCTP
91 extern void sctp_add_ip_address(struct ifaddr *ifa);
92 extern void sctp_delete_ip_address(struct ifaddr *ifa);
93 #endif /* SCTP */
94 
95 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
96 
97 static struct route_cb {
98 	int	ip_count;
99 	int	ip6_count;
100 	int	ipx_count;
101 	int	ns_count;
102 	int	any_count;
103 } route_cb;
104 
105 static const struct sockaddr route_src = { 2, PF_ROUTE, };
106 
107 struct walkarg {
108 	int	w_tmemsize;
109 	int	w_op, w_arg;
110 	void	*w_tmem;
111 	struct sysctl_req *w_req;
112 };
113 
114 static struct mbuf *
115 		rt_msg_mbuf (int, struct rt_addrinfo *);
116 static void	rt_msg_buffer (int, struct rt_addrinfo *, void *buf, int len);
117 static int	rt_msgsize (int type, struct rt_addrinfo *rtinfo);
118 static int	rt_xaddrs (char *, char *, struct rt_addrinfo *);
119 static int	sysctl_dumpentry (struct radix_node *rn, void *vw);
120 static int	sysctl_iflist (int af, struct walkarg *w);
121 static int	route_output(struct mbuf *, struct socket *, ...);
122 static void	rt_setmetrics (u_long, struct rt_metrics *,
123 			       struct rt_metrics *);
124 
125 /*
126  * It really doesn't make any sense at all for this code to share much
127  * with raw_usrreq.c, since its functionality is so restricted.  XXX
128  */
129 static int
130 rts_abort(struct socket *so)
131 {
132 	int error;
133 
134 	crit_enter();
135 	error = raw_usrreqs.pru_abort(so);
136 	crit_exit();
137 	return error;
138 }
139 
140 /* pru_accept is EOPNOTSUPP */
141 
142 static int
143 rts_attach(struct socket *so, int proto, struct pru_attach_info *ai)
144 {
145 	struct rawcb *rp;
146 	int error;
147 
148 	if (sotorawcb(so) != NULL)
149 		return EISCONN;	/* XXX panic? */
150 
151 	rp = kmalloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
152 	if (rp == NULL)
153 		return ENOBUFS;
154 
155 	/*
156 	 * The critical section is necessary to block protocols from sending
157 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
158 	 * this PCB is extant but incompletely initialized.
159 	 * Probably we should try to do more of this work beforehand and
160 	 * eliminate the critical section.
161 	 */
162 	crit_enter();
163 	so->so_pcb = rp;
164 	error = raw_attach(so, proto, ai->sb_rlimit);
165 	rp = sotorawcb(so);
166 	if (error) {
167 		crit_exit();
168 		kfree(rp, M_PCB);
169 		return error;
170 	}
171 	switch(rp->rcb_proto.sp_protocol) {
172 	case AF_INET:
173 		route_cb.ip_count++;
174 		break;
175 	case AF_INET6:
176 		route_cb.ip6_count++;
177 		break;
178 	case AF_IPX:
179 		route_cb.ipx_count++;
180 		break;
181 	case AF_NS:
182 		route_cb.ns_count++;
183 		break;
184 	}
185 	rp->rcb_faddr = &route_src;
186 	route_cb.any_count++;
187 	soisconnected(so);
188 	so->so_options |= SO_USELOOPBACK;
189 	crit_exit();
190 	return 0;
191 }
192 
193 static int
194 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
195 {
196 	int error;
197 
198 	crit_enter();
199 	error = raw_usrreqs.pru_bind(so, nam, td); /* xxx just EINVAL */
200 	crit_exit();
201 	return error;
202 }
203 
204 static int
205 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
206 {
207 	int error;
208 
209 	crit_enter();
210 	error = raw_usrreqs.pru_connect(so, nam, td); /* XXX just EINVAL */
211 	crit_exit();
212 	return error;
213 }
214 
215 /* pru_connect2 is EOPNOTSUPP */
216 /* pru_control is EOPNOTSUPP */
217 
218 static int
219 rts_detach(struct socket *so)
220 {
221 	struct rawcb *rp = sotorawcb(so);
222 	int error;
223 
224 	crit_enter();
225 	if (rp != NULL) {
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 		case AF_NS:
237 			route_cb.ns_count--;
238 			break;
239 		}
240 		route_cb.any_count--;
241 	}
242 	error = raw_usrreqs.pru_detach(so);
243 	crit_exit();
244 	return error;
245 }
246 
247 static int
248 rts_disconnect(struct socket *so)
249 {
250 	int error;
251 
252 	crit_enter();
253 	error = raw_usrreqs.pru_disconnect(so);
254 	crit_exit();
255 	return error;
256 }
257 
258 /* pru_listen is EOPNOTSUPP */
259 
260 static int
261 rts_peeraddr(struct socket *so, struct sockaddr **nam)
262 {
263 	int error;
264 
265 	crit_enter();
266 	error = raw_usrreqs.pru_peeraddr(so, nam);
267 	crit_exit();
268 	return error;
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 	int error;
279 
280 	crit_enter();
281 	error = raw_usrreqs.pru_send(so, flags, m, nam, control, td);
282 	crit_exit();
283 	return error;
284 }
285 
286 /* pru_sense is null */
287 
288 static int
289 rts_shutdown(struct socket *so)
290 {
291 	int error;
292 
293 	crit_enter();
294 	error = raw_usrreqs.pru_shutdown(so);
295 	crit_exit();
296 	return error;
297 }
298 
299 static int
300 rts_sockaddr(struct socket *so, struct sockaddr **nam)
301 {
302 	int error;
303 
304 	crit_enter();
305 	error = raw_usrreqs.pru_sockaddr(so, nam);
306 	crit_exit();
307 	return error;
308 }
309 
310 static struct pr_usrreqs route_usrreqs = {
311 	.pru_abort = rts_abort,
312 	.pru_accept = pru_accept_notsupp,
313 	.pru_attach = rts_attach,
314 	.pru_bind = rts_bind,
315 	.pru_connect = rts_connect,
316 	.pru_connect2 = pru_connect2_notsupp,
317 	.pru_control = pru_control_notsupp,
318 	.pru_detach = rts_detach,
319 	.pru_disconnect = rts_disconnect,
320 	.pru_listen = pru_listen_notsupp,
321 	.pru_peeraddr = rts_peeraddr,
322 	.pru_rcvd = pru_rcvd_notsupp,
323 	.pru_rcvoob = pru_rcvoob_notsupp,
324 	.pru_send = rts_send,
325 	.pru_sense = pru_sense_null,
326 	.pru_shutdown = rts_shutdown,
327 	.pru_sockaddr = rts_sockaddr,
328 	.pru_sosend = sosend,
329 	.pru_soreceive = soreceive,
330 	.pru_sopoll = sopoll
331 };
332 
333 static __inline sa_family_t
334 familyof(struct sockaddr *sa)
335 {
336 	return (sa != NULL ? sa->sa_family : 0);
337 }
338 
339 /*
340  * Routing socket input function.  The packet must be serialized onto cpu 0.
341  * We use the cpu0_soport() netisr processing loop to handle it.
342  *
343  * This looks messy but it means that anyone, including interrupt code,
344  * can send a message to the routing socket.
345  */
346 static void
347 rts_input_handler(struct netmsg *msg)
348 {
349 	static const struct sockaddr route_dst = { 2, PF_ROUTE, };
350 	struct sockproto route_proto;
351 	struct netmsg_packet *pmsg;
352 	struct mbuf *m;
353 	sa_family_t family;
354 
355 	pmsg = (void *)msg;
356 	m = pmsg->nm_packet;
357 	family = pmsg->nm_netmsg.nm_lmsg.u.ms_result;
358 	route_proto.sp_family = PF_ROUTE;
359 	route_proto.sp_protocol = family;
360 
361 	raw_input(m, &route_proto, &route_src, &route_dst);
362 }
363 
364 static void
365 rts_input(struct mbuf *m, sa_family_t family)
366 {
367 	struct netmsg_packet *pmsg;
368 	lwkt_port_t port;
369 
370 	port = cpu0_soport(NULL, NULL, 0);
371 	pmsg = &m->m_hdr.mh_netmsg;
372 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport,
373 		    0, rts_input_handler);
374 	pmsg->nm_packet = m;
375 	pmsg->nm_netmsg.nm_lmsg.u.ms_result = family;
376 	lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
377 }
378 
379 static void *
380 reallocbuf(void *ptr, size_t len, size_t olen)
381 {
382 	void *newptr;
383 
384 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
385 	if (newptr == NULL)
386 		return NULL;
387 	bcopy(ptr, newptr, olen);
388 	kfree(ptr, M_RTABLE);
389 	return (newptr);
390 }
391 
392 /*
393  * Internal helper routine for route_output().
394  */
395 static int
396 fillrtmsg(struct rt_msghdr **prtm, struct rtentry *rt,
397 	  struct rt_addrinfo *rtinfo)
398 {
399 	int msglen;
400 	struct rt_msghdr *rtm = *prtm;
401 
402 	/* Fill in rt_addrinfo for call to rt_msg_buffer(). */
403 	rtinfo->rti_dst = rt_key(rt);
404 	rtinfo->rti_gateway = rt->rt_gateway;
405 	rtinfo->rti_netmask = rt_mask(rt);		/* might be NULL */
406 	rtinfo->rti_genmask = rt->rt_genmask;		/* might be NULL */
407 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
408 		if (rt->rt_ifp != NULL) {
409 			rtinfo->rti_ifpaddr =
410 			    TAILQ_FIRST(&rt->rt_ifp->if_addrhead)->ifa_addr;
411 			rtinfo->rti_ifaaddr = rt->rt_ifa->ifa_addr;
412 			if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
413 				rtinfo->rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
414 			rtm->rtm_index = rt->rt_ifp->if_index;
415 		} else {
416 			rtinfo->rti_ifpaddr = NULL;
417 			rtinfo->rti_ifaaddr = NULL;
418 	    }
419 	}
420 
421 	msglen = rt_msgsize(rtm->rtm_type, rtinfo);
422 	if (rtm->rtm_msglen < msglen) {
423 		rtm = reallocbuf(rtm, msglen, rtm->rtm_msglen);
424 		if (rtm == NULL)
425 			return (ENOBUFS);
426 		*prtm = rtm;
427 	}
428 	rt_msg_buffer(rtm->rtm_type, rtinfo, rtm, msglen);
429 
430 	rtm->rtm_flags = rt->rt_flags;
431 	rtm->rtm_rmx = rt->rt_rmx;
432 	rtm->rtm_addrs = rtinfo->rti_addrs;
433 
434 	return (0);
435 }
436 
437 static void route_output_add_callback(int, int, struct rt_addrinfo *,
438 					struct rtentry *, void *);
439 static void route_output_delete_callback(int, int, struct rt_addrinfo *,
440 					struct rtentry *, void *);
441 static void route_output_change_callback(int, int, struct rt_addrinfo *,
442 					struct rtentry *, void *);
443 static void route_output_lock_callback(int, int, struct rt_addrinfo *,
444 					struct rtentry *, void *);
445 
446 /*ARGSUSED*/
447 static int
448 route_output(struct mbuf *m, struct socket *so, ...)
449 {
450 	struct rt_msghdr *rtm = NULL;
451 	struct rtentry *rt;
452 	struct radix_node_head *rnh;
453 	struct rawcb *rp = NULL;
454 	struct pr_output_info *oi;
455 	struct rt_addrinfo rtinfo;
456 	int len, error = 0;
457 	__va_list ap;
458 
459 	__va_start(ap, so);
460 	oi = __va_arg(ap, struct pr_output_info *);
461 	__va_end(ap);
462 
463 #define gotoerr(e) { error = e; goto flush;}
464 
465 	if (m == NULL ||
466 	    (m->m_len < sizeof(long) &&
467 	     (m = m_pullup(m, sizeof(long))) == NULL))
468 		return (ENOBUFS);
469 	if (!(m->m_flags & M_PKTHDR))
470 		panic("route_output");
471 	len = m->m_pkthdr.len;
472 	if (len < sizeof(struct rt_msghdr) ||
473 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
474 		rtinfo.rti_dst = NULL;
475 		gotoerr(EINVAL);
476 	}
477 	rtm = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
478 	if (rtm == NULL) {
479 		rtinfo.rti_dst = NULL;
480 		gotoerr(ENOBUFS);
481 	}
482 	m_copydata(m, 0, len, (caddr_t)rtm);
483 	if (rtm->rtm_version != RTM_VERSION) {
484 		rtinfo.rti_dst = NULL;
485 		gotoerr(EPROTONOSUPPORT);
486 	}
487 	rtm->rtm_pid = oi->p_pid;
488 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
489 	rtinfo.rti_addrs = rtm->rtm_addrs;
490 	if (rt_xaddrs((char *)(rtm + 1), (char *)rtm + len, &rtinfo) != 0) {
491 		rtinfo.rti_dst = NULL;
492 		gotoerr(EINVAL);
493 	}
494 	rtinfo.rti_flags = rtm->rtm_flags;
495 	if (rtinfo.rti_dst == NULL || rtinfo.rti_dst->sa_family >= AF_MAX ||
496 	    (rtinfo.rti_gateway && rtinfo.rti_gateway->sa_family >= AF_MAX))
497 		gotoerr(EINVAL);
498 
499 	if (rtinfo.rti_genmask != NULL) {
500 		struct radix_node *n;
501 
502 #define	clen(s)	(*(u_char *)(s))
503 		n = rn_addmask((char *)rtinfo.rti_genmask, TRUE, 1);
504 		if (n != NULL &&
505 		    rtinfo.rti_genmask->sa_len >= clen(n->rn_key) &&
506 		    bcmp((char *)rtinfo.rti_genmask + 1,
507 		         (char *)n->rn_key + 1, clen(n->rn_key) - 1) == 0)
508 			rtinfo.rti_genmask = (struct sockaddr *)n->rn_key;
509 		else
510 			gotoerr(ENOBUFS);
511 	}
512 
513 	/*
514 	 * Verify that the caller has the appropriate privilege; RTM_GET
515 	 * is the only operation the non-superuser is allowed.
516 	 */
517 	if (rtm->rtm_type != RTM_GET && suser_cred(so->so_cred, 0) != 0)
518 		gotoerr(EPERM);
519 
520 	switch (rtm->rtm_type) {
521 	case RTM_ADD:
522 		if (rtinfo.rti_gateway == NULL) {
523 			error = EINVAL;
524 		} else {
525 			error = rtrequest1_global(RTM_ADD, &rtinfo,
526 					  route_output_add_callback, rtm);
527 		}
528 		break;
529 	case RTM_DELETE:
530 		/*
531 		 * note: &rtm passed as argument so 'rtm' can be replaced.
532 		 */
533 		error = rtrequest1_global(RTM_DELETE, &rtinfo,
534 					  route_output_delete_callback, &rtm);
535 		break;
536 	case RTM_GET:
537 		rnh = rt_tables[mycpuid][rtinfo.rti_dst->sa_family];
538 		if (rnh == NULL) {
539 			error = EAFNOSUPPORT;
540 			break;
541 		}
542 		rt = (struct rtentry *)
543 		    rnh->rnh_lookup((char *)rtinfo.rti_dst,
544 		    		    (char *)rtinfo.rti_netmask, rnh);
545 		if (rt == NULL) {
546 			error = ESRCH;
547 			break;
548 		}
549 		rt->rt_refcnt++;
550 		if (fillrtmsg(&rtm, rt, &rtinfo) != 0)
551 			gotoerr(ENOBUFS);
552 		--rt->rt_refcnt;
553 		break;
554 	case RTM_CHANGE:
555 		error = rtrequest1_global(RTM_GET, &rtinfo,
556 					  route_output_change_callback, rtm);
557 		break;
558 	case RTM_LOCK:
559 		error = rtrequest1_global(RTM_GET, &rtinfo,
560 					  route_output_lock_callback, rtm);
561 		break;
562 	default:
563 		error = EOPNOTSUPP;
564 		break;
565 	}
566 
567 flush:
568 	if (rtm != NULL) {
569 		if (error != 0)
570 			rtm->rtm_errno = error;
571 		else
572 			rtm->rtm_flags |= RTF_DONE;
573 	}
574 
575 	/*
576 	 * Check to see if we don't want our own messages.
577 	 */
578 	if (!(so->so_options & SO_USELOOPBACK)) {
579 		if (route_cb.any_count <= 1) {
580 			if (rtm != NULL)
581 				kfree(rtm, M_RTABLE);
582 			m_freem(m);
583 			return (error);
584 		}
585 		/* There is another listener, so construct message */
586 		rp = sotorawcb(so);
587 	}
588 	if (rtm != NULL) {
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 		kfree(rtm, M_RTABLE);
596 	}
597 	if (rp != NULL)
598 		rp->rcb_proto.sp_family = 0; /* Avoid us */
599 	if (m != NULL)
600 		rts_input(m, familyof(rtinfo.rti_dst));
601 	if (rp != NULL)
602 		rp->rcb_proto.sp_family = PF_ROUTE;
603 	return (error);
604 }
605 
606 static void
607 route_output_add_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
608 			  struct rtentry *rt, void *arg)
609 {
610 	struct rt_msghdr *rtm = arg;
611 
612 	if (error == 0 && rt != NULL) {
613 		rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
614 		    &rt->rt_rmx);
615 		rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
616 		rt->rt_rmx.rmx_locks |=
617 		    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
618 		rt->rt_genmask = rtinfo->rti_genmask;
619 	}
620 }
621 
622 static void
623 route_output_delete_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
624 			  struct rtentry *rt, void *arg)
625 {
626 	struct rt_msghdr **rtm = arg;
627 
628 	if (error == 0 && rt) {
629 		++rt->rt_refcnt;
630 		if (fillrtmsg(rtm, rt, rtinfo) != 0) {
631 			error = ENOBUFS;
632 			/* XXX no way to return the error */
633 		}
634 		--rt->rt_refcnt;
635 	}
636 }
637 
638 static void
639 route_output_change_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
640 			  struct rtentry *rt, void *arg)
641 {
642 	struct rt_msghdr *rtm = arg;
643 	struct ifaddr *ifa;
644 
645 	if (error)
646 		goto done;
647 
648 	/*
649 	 * new gateway could require new ifaddr, ifp;
650 	 * flags may also be different; ifp may be specified
651 	 * by ll sockaddr when protocol address is ambiguous
652 	 */
653 	if (((rt->rt_flags & RTF_GATEWAY) && rtinfo->rti_gateway != NULL) ||
654 	    rtinfo->rti_ifpaddr != NULL || (rtinfo->rti_ifaaddr != NULL &&
655 	    sa_equal(rtinfo->rti_ifaaddr, rt->rt_ifa->ifa_addr))
656 	) {
657 		error = rt_getifa(rtinfo);
658 		if (error != 0)
659 			goto done;
660 	}
661 	if (rtinfo->rti_gateway != NULL) {
662 		error = rt_setgate(rt, rt_key(rt), rtinfo->rti_gateway);
663 		if (error != 0)
664 			goto done;
665 	}
666 	if ((ifa = rtinfo->rti_ifa) != NULL) {
667 		struct ifaddr *oifa = rt->rt_ifa;
668 
669 		if (oifa != ifa) {
670 			if (oifa && oifa->ifa_rtrequest)
671 				oifa->ifa_rtrequest(RTM_DELETE, rt, rtinfo);
672 			IFAFREE(rt->rt_ifa);
673 			IFAREF(ifa);
674 			rt->rt_ifa = ifa;
675 			rt->rt_ifp = rtinfo->rti_ifp;
676 		}
677 	}
678 	rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, &rt->rt_rmx);
679 	if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
680 	       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, rtinfo);
681 	if (rtinfo->rti_genmask != NULL)
682 		rt->rt_genmask = rtinfo->rti_genmask;
683 done:
684 	/* XXX no way to return error */
685 	;
686 }
687 
688 static void
689 route_output_lock_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
690 			   struct rtentry *rt, void *arg)
691 {
692 	struct rt_msghdr *rtm = arg;
693 
694 	rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
695 	rt->rt_rmx.rmx_locks |=
696 		(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
697 }
698 
699 static void
700 rt_setmetrics(u_long which, struct rt_metrics *in, struct rt_metrics *out)
701 {
702 #define setmetric(flag, elt) if (which & (flag)) out->elt = in->elt;
703 	setmetric(RTV_RPIPE, rmx_recvpipe);
704 	setmetric(RTV_SPIPE, rmx_sendpipe);
705 	setmetric(RTV_SSTHRESH, rmx_ssthresh);
706 	setmetric(RTV_RTT, rmx_rtt);
707 	setmetric(RTV_RTTVAR, rmx_rttvar);
708 	setmetric(RTV_HOPCOUNT, rmx_hopcount);
709 	setmetric(RTV_MTU, rmx_mtu);
710 	setmetric(RTV_EXPIRE, rmx_expire);
711 #undef setmetric
712 }
713 
714 #define ROUNDUP(a) \
715 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
716 
717 /*
718  * Extract the addresses of the passed sockaddrs.
719  * Do a little sanity checking so as to avoid bad memory references.
720  * This data is derived straight from userland.
721  */
722 static int
723 rt_xaddrs(char *cp, char *cplim, struct rt_addrinfo *rtinfo)
724 {
725 	struct sockaddr *sa;
726 	int i;
727 
728 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
729 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
730 			continue;
731 		sa = (struct sockaddr *)cp;
732 		/*
733 		 * It won't fit.
734 		 */
735 		if ((cp + sa->sa_len) > cplim) {
736 			return (EINVAL);
737 		}
738 
739 		/*
740 		 * There are no more...  Quit now.
741 		 * If there are more bits, they are in error.
742 		 * I've seen this.  route(1) can evidently generate these.
743 		 * This causes kernel to core dump.
744 		 * For compatibility, if we see this, point to a safe address.
745 		 */
746 		if (sa->sa_len == 0) {
747 			static struct sockaddr sa_zero = {
748 				sizeof sa_zero, AF_INET,
749 			};
750 
751 			rtinfo->rti_info[i] = &sa_zero;
752 			return (0); /* should be EINVAL but for compat */
753 		}
754 
755 		/* Accept the sockaddr. */
756 		rtinfo->rti_info[i] = sa;
757 		cp += ROUNDUP(sa->sa_len);
758 	}
759 	return (0);
760 }
761 
762 static int
763 rt_msghdrsize(int type)
764 {
765 	switch (type) {
766 	case RTM_DELADDR:
767 	case RTM_NEWADDR:
768 		return sizeof(struct ifa_msghdr);
769 	case RTM_DELMADDR:
770 	case RTM_NEWMADDR:
771 		return sizeof(struct ifma_msghdr);
772 	case RTM_IFINFO:
773 		return sizeof(struct if_msghdr);
774 	case RTM_IFANNOUNCE:
775 	case RTM_IEEE80211:
776 		return sizeof(struct if_announcemsghdr);
777 	default:
778 		return sizeof(struct rt_msghdr);
779 	}
780 }
781 
782 static int
783 rt_msgsize(int type, struct rt_addrinfo *rtinfo)
784 {
785 	int len, i;
786 
787 	len = rt_msghdrsize(type);
788 	for (i = 0; i < RTAX_MAX; i++) {
789 		if (rtinfo->rti_info[i] != NULL)
790 			len += ROUNDUP(rtinfo->rti_info[i]->sa_len);
791 	}
792 	len = ALIGN(len);
793 	return len;
794 }
795 
796 /*
797  * Build a routing message in a buffer.
798  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
799  * to the end of the buffer after the message header.
800  *
801  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
802  * This side-effect can be avoided if we reorder the addrs bitmask field in all
803  * the route messages to line up so we can set it here instead of back in the
804  * calling routine.
805  */
806 static void
807 rt_msg_buffer(int type, struct rt_addrinfo *rtinfo, void *buf, int msglen)
808 {
809 	struct rt_msghdr *rtm;
810 	char *cp;
811 	int dlen, i;
812 
813 	rtm = (struct rt_msghdr *) buf;
814 	rtm->rtm_version = RTM_VERSION;
815 	rtm->rtm_type = type;
816 	rtm->rtm_msglen = msglen;
817 
818 	cp = (char *)buf + rt_msghdrsize(type);
819 	rtinfo->rti_addrs = 0;
820 	for (i = 0; i < RTAX_MAX; i++) {
821 		struct sockaddr *sa;
822 
823 		if ((sa = rtinfo->rti_info[i]) == NULL)
824 			continue;
825 		rtinfo->rti_addrs |= (1 << i);
826 		dlen = ROUNDUP(sa->sa_len);
827 		bcopy(sa, cp, dlen);
828 		cp += dlen;
829 	}
830 }
831 
832 /*
833  * Build a routing message in a mbuf chain.
834  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
835  * to the end of the mbuf after the message header.
836  *
837  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
838  * This side-effect can be avoided if we reorder the addrs bitmask field in all
839  * the route messages to line up so we can set it here instead of back in the
840  * calling routine.
841  */
842 static struct mbuf *
843 rt_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
844 {
845 	struct mbuf *m;
846 	struct rt_msghdr *rtm;
847 	int hlen, len;
848 	int i;
849 
850 	hlen = rt_msghdrsize(type);
851 	KASSERT(hlen <= MCLBYTES, ("rt_msg_mbuf: hlen %d doesn't fit", hlen));
852 
853 	m = m_getl(hlen, MB_DONTWAIT, MT_DATA, M_PKTHDR, NULL);
854 	if (m == NULL)
855 		return (NULL);
856 	m->m_pkthdr.len = m->m_len = hlen;
857 	m->m_pkthdr.rcvif = NULL;
858 	rtinfo->rti_addrs = 0;
859 	len = hlen;
860 	for (i = 0; i < RTAX_MAX; i++) {
861 		struct sockaddr *sa;
862 		int dlen;
863 
864 		if ((sa = rtinfo->rti_info[i]) == NULL)
865 			continue;
866 		rtinfo->rti_addrs |= (1 << i);
867 		dlen = ROUNDUP(sa->sa_len);
868 		m_copyback(m, len, dlen, (caddr_t)sa); /* can grow mbuf chain */
869 		len += dlen;
870 	}
871 	if (m->m_pkthdr.len != len) { /* one of the m_copyback() calls failed */
872 		m_freem(m);
873 		return (NULL);
874 	}
875 	rtm = mtod(m, struct rt_msghdr *);
876 	bzero(rtm, hlen);
877 	rtm->rtm_msglen = len;
878 	rtm->rtm_version = RTM_VERSION;
879 	rtm->rtm_type = type;
880 	return (m);
881 }
882 
883 /*
884  * This routine is called to generate a message from the routing
885  * socket indicating that a redirect has occurred, a routing lookup
886  * has failed, or that a protocol has detected timeouts to a particular
887  * destination.
888  */
889 void
890 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
891 {
892 	struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
893 	struct rt_msghdr *rtm;
894 	struct mbuf *m;
895 
896 	if (route_cb.any_count == 0)
897 		return;
898 	m = rt_msg_mbuf(type, rtinfo);
899 	if (m == NULL)
900 		return;
901 	rtm = mtod(m, struct rt_msghdr *);
902 	rtm->rtm_flags = RTF_DONE | flags;
903 	rtm->rtm_errno = error;
904 	rtm->rtm_addrs = rtinfo->rti_addrs;
905 	rts_input(m, familyof(dst));
906 }
907 
908 void
909 rt_dstmsg(int type, struct sockaddr *dst, int error)
910 {
911 	struct rt_msghdr *rtm;
912 	struct rt_addrinfo addrs;
913 	struct mbuf *m;
914 
915 	if (route_cb.any_count == 0)
916 		return;
917 	bzero(&addrs, sizeof(struct rt_addrinfo));
918 	addrs.rti_info[RTAX_DST] = dst;
919 	m = rt_msg_mbuf(type, &addrs);
920 	if (m == NULL)
921 		return;
922 	rtm = mtod(m, struct rt_msghdr *);
923 	rtm->rtm_flags = RTF_DONE;
924 	rtm->rtm_errno = error;
925 	rtm->rtm_addrs = addrs.rti_addrs;
926 	rts_input(m, familyof(dst));
927 }
928 
929 /*
930  * This routine is called to generate a message from the routing
931  * socket indicating that the status of a network interface has changed.
932  */
933 void
934 rt_ifmsg(struct ifnet *ifp)
935 {
936 	struct if_msghdr *ifm;
937 	struct mbuf *m;
938 	struct rt_addrinfo rtinfo;
939 
940 	if (route_cb.any_count == 0)
941 		return;
942 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
943 	m = rt_msg_mbuf(RTM_IFINFO, &rtinfo);
944 	if (m == NULL)
945 		return;
946 	ifm = mtod(m, struct if_msghdr *);
947 	ifm->ifm_index = ifp->if_index;
948 	ifm->ifm_flags = ifp->if_flags;
949 	ifm->ifm_data = ifp->if_data;
950 	ifm->ifm_addrs = 0;
951 	rts_input(m, 0);
952 }
953 
954 static void
955 rt_ifamsg(int cmd, struct ifaddr *ifa)
956 {
957 	struct ifa_msghdr *ifam;
958 	struct rt_addrinfo rtinfo;
959 	struct mbuf *m;
960 	struct ifnet *ifp = ifa->ifa_ifp;
961 
962 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
963 	rtinfo.rti_ifaaddr = ifa->ifa_addr;
964 	rtinfo.rti_ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
965 	rtinfo.rti_netmask = ifa->ifa_netmask;
966 	rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
967 
968 	m = rt_msg_mbuf(cmd, &rtinfo);
969 	if (m == NULL)
970 		return;
971 
972 	ifam = mtod(m, struct ifa_msghdr *);
973 	ifam->ifam_index = ifp->if_index;
974 	ifam->ifam_metric = ifa->ifa_metric;
975 	ifam->ifam_flags = ifa->ifa_flags;
976 	ifam->ifam_addrs = rtinfo.rti_addrs;
977 
978 	rts_input(m, familyof(ifa->ifa_addr));
979 }
980 
981 void
982 rt_rtmsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int error)
983 {
984 	struct rt_msghdr *rtm;
985 	struct rt_addrinfo rtinfo;
986 	struct mbuf *m;
987 	struct sockaddr *dst;
988 
989 	if (rt == NULL)
990 		return;
991 
992 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
993 	rtinfo.rti_dst = dst = rt_key(rt);
994 	rtinfo.rti_gateway = rt->rt_gateway;
995 	rtinfo.rti_netmask = rt_mask(rt);
996 	if (ifp != NULL)
997 		rtinfo.rti_ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
998 	rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
999 
1000 	m = rt_msg_mbuf(cmd, &rtinfo);
1001 	if (m == NULL)
1002 		return;
1003 
1004 	rtm = mtod(m, struct rt_msghdr *);
1005 	if (ifp != NULL)
1006 		rtm->rtm_index = ifp->if_index;
1007 	rtm->rtm_flags |= rt->rt_flags;
1008 	rtm->rtm_errno = error;
1009 	rtm->rtm_addrs = rtinfo.rti_addrs;
1010 
1011 	rts_input(m, familyof(dst));
1012 }
1013 
1014 /*
1015  * This is called to generate messages from the routing socket
1016  * indicating a network interface has had addresses associated with it.
1017  * if we ever reverse the logic and replace messages TO the routing
1018  * socket indicate a request to configure interfaces, then it will
1019  * be unnecessary as the routing socket will automatically generate
1020  * copies of it.
1021  */
1022 void
1023 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1024 {
1025 #ifdef SCTP
1026 	/*
1027 	 * notify the SCTP stack
1028 	 * this will only get called when an address is added/deleted
1029 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1030 	 */
1031 	if (cmd == RTM_ADD)
1032 		sctp_add_ip_address(ifa);
1033 	else if (cmd == RTM_DELETE)
1034 		sctp_delete_ip_address(ifa);
1035 #endif /* SCTP */
1036 
1037 	if (route_cb.any_count == 0)
1038 		return;
1039 
1040 	if (cmd == RTM_ADD) {
1041 		rt_ifamsg(RTM_NEWADDR, ifa);
1042 		rt_rtmsg(RTM_ADD, rt, ifa->ifa_ifp, error);
1043 	} else {
1044 		KASSERT((cmd == RTM_DELETE), ("unknown cmd %d", cmd));
1045 		rt_rtmsg(RTM_DELETE, rt, ifa->ifa_ifp, error);
1046 		rt_ifamsg(RTM_DELADDR, ifa);
1047 	}
1048 }
1049 
1050 /*
1051  * This is the analogue to the rt_newaddrmsg which performs the same
1052  * function but for multicast group memberhips.  This is easier since
1053  * there is no route state to worry about.
1054  */
1055 void
1056 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1057 {
1058 	struct rt_addrinfo rtinfo;
1059 	struct mbuf *m = NULL;
1060 	struct ifnet *ifp = ifma->ifma_ifp;
1061 	struct ifma_msghdr *ifmam;
1062 
1063 	if (route_cb.any_count == 0)
1064 		return;
1065 
1066 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1067 	rtinfo.rti_ifaaddr = ifma->ifma_addr;
1068 	if (ifp != NULL && !TAILQ_EMPTY(&ifp->if_addrhead))
1069 		rtinfo.rti_ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
1070 	/*
1071 	 * If a link-layer address is present, present it as a ``gateway''
1072 	 * (similarly to how ARP entries, e.g., are presented).
1073 	 */
1074 	rtinfo.rti_gateway = ifma->ifma_lladdr;
1075 
1076 	m = rt_msg_mbuf(cmd, &rtinfo);
1077 	if (m == NULL)
1078 		return;
1079 
1080 	ifmam = mtod(m, struct ifma_msghdr *);
1081 	ifmam->ifmam_index = ifp->if_index;
1082 	ifmam->ifmam_addrs = rtinfo.rti_addrs;
1083 
1084 	rts_input(m, familyof(ifma->ifma_addr));
1085 }
1086 
1087 static struct mbuf *
1088 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1089 		     struct rt_addrinfo *info)
1090 {
1091 	struct if_announcemsghdr *ifan;
1092 	struct mbuf *m;
1093 
1094 	if (route_cb.any_count == 0)
1095 		return NULL;
1096 
1097 	bzero(info, sizeof(*info));
1098 	m = rt_msg_mbuf(type, info);
1099 	if (m == NULL)
1100 		return NULL;
1101 
1102 	ifan = mtod(m, struct if_announcemsghdr *);
1103 	ifan->ifan_index = ifp->if_index;
1104 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof ifan->ifan_name);
1105 	ifan->ifan_what = what;
1106 	return m;
1107 }
1108 
1109 /*
1110  * This is called to generate routing socket messages indicating
1111  * IEEE80211 wireless events.
1112  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1113  */
1114 void
1115 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1116 {
1117 	struct rt_addrinfo info;
1118 	struct mbuf *m;
1119 
1120 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1121 	if (m == NULL)
1122 		return;
1123 
1124 	/*
1125 	 * Append the ieee80211 data.  Try to stick it in the
1126 	 * mbuf containing the ifannounce msg; otherwise allocate
1127 	 * a new mbuf and append.
1128 	 *
1129 	 * NB: we assume m is a single mbuf.
1130 	 */
1131 	if (data_len > M_TRAILINGSPACE(m)) {
1132 		struct mbuf *n = m_get(MB_DONTWAIT, MT_DATA);
1133 		if (n == NULL) {
1134 			m_freem(m);
1135 			return;
1136 		}
1137 		bcopy(data, mtod(n, void *), data_len);
1138 		n->m_len = data_len;
1139 		m->m_next = n;
1140 	} else if (data_len > 0) {
1141 		bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1142 		m->m_len += data_len;
1143 	}
1144 	if (m->m_flags & M_PKTHDR)
1145 		m->m_pkthdr.len += data_len;
1146 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1147 	rts_input(m, 0);
1148 }
1149 
1150 /*
1151  * This is called to generate routing socket messages indicating
1152  * network interface arrival and departure.
1153  */
1154 void
1155 rt_ifannouncemsg(struct ifnet *ifp, int what)
1156 {
1157 	struct rt_addrinfo addrinfo;
1158 	struct mbuf *m;
1159 
1160 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &addrinfo);
1161 	if (m != NULL)
1162 		rts_input(m, 0);
1163 }
1164 
1165 static int
1166 resizewalkarg(struct walkarg *w, int len)
1167 {
1168 	void *newptr;
1169 
1170 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
1171 	if (newptr == NULL)
1172 		return (ENOMEM);
1173 	if (w->w_tmem != NULL)
1174 		kfree(w->w_tmem, M_RTABLE);
1175 	w->w_tmem = newptr;
1176 	w->w_tmemsize = len;
1177 	return (0);
1178 }
1179 
1180 /*
1181  * This is used in dumping the kernel table via sysctl().
1182  */
1183 int
1184 sysctl_dumpentry(struct radix_node *rn, void *vw)
1185 {
1186 	struct walkarg *w = vw;
1187 	struct rtentry *rt = (struct rtentry *)rn;
1188 	struct rt_addrinfo rtinfo;
1189 	int error, msglen;
1190 
1191 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1192 		return 0;
1193 
1194 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1195 	rtinfo.rti_dst = rt_key(rt);
1196 	rtinfo.rti_gateway = rt->rt_gateway;
1197 	rtinfo.rti_netmask = rt_mask(rt);
1198 	rtinfo.rti_genmask = rt->rt_genmask;
1199 	if (rt->rt_ifp != NULL) {
1200 		rtinfo.rti_ifpaddr =
1201 		    TAILQ_FIRST(&rt->rt_ifp->if_addrhead)->ifa_addr;
1202 		rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
1203 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1204 			rtinfo.rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
1205 	}
1206 	msglen = rt_msgsize(RTM_GET, &rtinfo);
1207 	if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1208 		return (ENOMEM);
1209 	rt_msg_buffer(RTM_GET, &rtinfo, w->w_tmem, msglen);
1210 	if (w->w_req != NULL) {
1211 		struct rt_msghdr *rtm = w->w_tmem;
1212 
1213 		rtm->rtm_flags = rt->rt_flags;
1214 		rtm->rtm_use = rt->rt_use;
1215 		rtm->rtm_rmx = rt->rt_rmx;
1216 		rtm->rtm_index = rt->rt_ifp->if_index;
1217 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1218 		rtm->rtm_addrs = rtinfo.rti_addrs;
1219 		error = SYSCTL_OUT(w->w_req, rtm, msglen);
1220 		return (error);
1221 	}
1222 	return (0);
1223 }
1224 
1225 static int
1226 sysctl_iflist(int af, struct walkarg *w)
1227 {
1228 	struct ifnet *ifp;
1229 	struct ifaddr *ifa;
1230 	struct rt_addrinfo rtinfo;
1231 	int msglen, error;
1232 
1233 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1234 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1235 		if (w->w_arg && w->w_arg != ifp->if_index)
1236 			continue;
1237 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1238 		rtinfo.rti_ifpaddr = ifa->ifa_addr;
1239 		msglen = rt_msgsize(RTM_IFINFO, &rtinfo);
1240 		if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1241 			return (ENOMEM);
1242 		rt_msg_buffer(RTM_IFINFO, &rtinfo, w->w_tmem, msglen);
1243 		rtinfo.rti_ifpaddr = NULL;
1244 		if (w->w_req != NULL && w->w_tmem != NULL) {
1245 			struct if_msghdr *ifm = w->w_tmem;
1246 
1247 			ifm->ifm_index = ifp->if_index;
1248 			ifm->ifm_flags = ifp->if_flags;
1249 			ifm->ifm_data = ifp->if_data;
1250 			ifm->ifm_addrs = rtinfo.rti_addrs;
1251 			error = SYSCTL_OUT(w->w_req, ifm, msglen);
1252 			if (error)
1253 				return (error);
1254 		}
1255 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1256 			if (af && af != ifa->ifa_addr->sa_family)
1257 				continue;
1258 			if (curproc->p_ucred->cr_prison &&
1259 			    prison_if(curproc->p_ucred, ifa->ifa_addr))
1260 				continue;
1261 			rtinfo.rti_ifaaddr = ifa->ifa_addr;
1262 			rtinfo.rti_netmask = ifa->ifa_netmask;
1263 			rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
1264 			msglen = rt_msgsize(RTM_NEWADDR, &rtinfo);
1265 			if (w->w_tmemsize < msglen &&
1266 			    resizewalkarg(w, msglen) != 0)
1267 				return (ENOMEM);
1268 			rt_msg_buffer(RTM_NEWADDR, &rtinfo, w->w_tmem, msglen);
1269 			if (w->w_req != NULL) {
1270 				struct ifa_msghdr *ifam = w->w_tmem;
1271 
1272 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1273 				ifam->ifam_flags = ifa->ifa_flags;
1274 				ifam->ifam_metric = ifa->ifa_metric;
1275 				ifam->ifam_addrs = rtinfo.rti_addrs;
1276 				error = SYSCTL_OUT(w->w_req, w->w_tmem, msglen);
1277 				if (error)
1278 					return (error);
1279 			}
1280 		}
1281 		rtinfo.rti_netmask = NULL;
1282 		rtinfo.rti_ifaaddr = NULL;
1283 		rtinfo.rti_bcastaddr = NULL;
1284 	}
1285 	return (0);
1286 }
1287 
1288 static int
1289 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1290 {
1291 	int	*name = (int *)arg1;
1292 	u_int	namelen = arg2;
1293 	struct radix_node_head *rnh;
1294 	int	i, error = EINVAL;
1295 	int	origcpu;
1296 	u_char  af;
1297 	struct	walkarg w;
1298 
1299 	name ++;
1300 	namelen--;
1301 	if (req->newptr)
1302 		return (EPERM);
1303 	if (namelen != 3 && namelen != 4)
1304 		return (EINVAL);
1305 	af = name[0];
1306 	bzero(&w, sizeof w);
1307 	w.w_op = name[1];
1308 	w.w_arg = name[2];
1309 	w.w_req = req;
1310 
1311 	/*
1312 	 * Optional third argument specifies cpu, used primarily for
1313 	 * debugging the route table.
1314 	 */
1315 	if (namelen == 4) {
1316 		if (name[3] < 0 || name[3] >= ncpus)
1317 			return (EINVAL);
1318 		origcpu = mycpuid;
1319 		lwkt_migratecpu(name[3]);
1320 	} else {
1321 		origcpu = -1;
1322 	}
1323 	crit_enter();
1324 	switch (w.w_op) {
1325 	case NET_RT_DUMP:
1326 	case NET_RT_FLAGS:
1327 		for (i = 1; i <= AF_MAX; i++)
1328 			if ((rnh = rt_tables[mycpuid][i]) &&
1329 			    (af == 0 || af == i) &&
1330 			    (error = rnh->rnh_walktree(rnh,
1331 						       sysctl_dumpentry, &w)))
1332 				break;
1333 		break;
1334 
1335 	case NET_RT_IFLIST:
1336 		error = sysctl_iflist(af, &w);
1337 	}
1338 	crit_exit();
1339 	if (w.w_tmem != NULL)
1340 		kfree(w.w_tmem, M_RTABLE);
1341 	if (origcpu >= 0)
1342 		lwkt_migratecpu(origcpu);
1343 	return (error);
1344 }
1345 
1346 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1347 
1348 /*
1349  * Definitions of protocols supported in the ROUTE domain.
1350  */
1351 
1352 static struct domain routedomain;		/* or at least forward */
1353 
1354 static struct protosw routesw[] = {
1355 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
1356   0,		route_output,	raw_ctlinput,	0,
1357   cpu0_soport,
1358   raw_init,	0,		0,		0,
1359   &route_usrreqs
1360 }
1361 };
1362 
1363 static struct domain routedomain = {
1364 	PF_ROUTE, "route", NULL, NULL, NULL,
1365 	routesw, &routesw[(sizeof routesw)/(sizeof routesw[0])],
1366 };
1367 
1368 DOMAIN_SET(route);
1369 
1370