xref: /dragonfly/sys/net/rtsock.c (revision 51f35c5c)
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 1.44 2008/07/07 22:02:10 nant 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 
153 	/*
154 	 * The critical section is necessary to block protocols from sending
155 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
156 	 * this PCB is extant but incompletely initialized.
157 	 * Probably we should try to do more of this work beforehand and
158 	 * eliminate the critical section.
159 	 */
160 	crit_enter();
161 	so->so_pcb = rp;
162 	error = raw_attach(so, proto, ai->sb_rlimit);
163 	rp = sotorawcb(so);
164 	if (error) {
165 		crit_exit();
166 		kfree(rp, M_PCB);
167 		return error;
168 	}
169 	switch(rp->rcb_proto.sp_protocol) {
170 	case AF_INET:
171 		route_cb.ip_count++;
172 		break;
173 	case AF_INET6:
174 		route_cb.ip6_count++;
175 		break;
176 	case AF_IPX:
177 		route_cb.ipx_count++;
178 		break;
179 	case AF_NS:
180 		route_cb.ns_count++;
181 		break;
182 	}
183 	rp->rcb_faddr = &route_src;
184 	route_cb.any_count++;
185 	soisconnected(so);
186 	so->so_options |= SO_USELOOPBACK;
187 	crit_exit();
188 	return 0;
189 }
190 
191 static int
192 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
193 {
194 	int error;
195 
196 	crit_enter();
197 	error = raw_usrreqs.pru_bind(so, nam, td); /* xxx just EINVAL */
198 	crit_exit();
199 	return error;
200 }
201 
202 static int
203 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
204 {
205 	int error;
206 
207 	crit_enter();
208 	error = raw_usrreqs.pru_connect(so, nam, td); /* XXX just EINVAL */
209 	crit_exit();
210 	return error;
211 }
212 
213 /* pru_connect2 is EOPNOTSUPP */
214 /* pru_control is EOPNOTSUPP */
215 
216 static int
217 rts_detach(struct socket *so)
218 {
219 	struct rawcb *rp = sotorawcb(so);
220 	int error;
221 
222 	crit_enter();
223 	if (rp != NULL) {
224 		switch(rp->rcb_proto.sp_protocol) {
225 		case AF_INET:
226 			route_cb.ip_count--;
227 			break;
228 		case AF_INET6:
229 			route_cb.ip6_count--;
230 			break;
231 		case AF_IPX:
232 			route_cb.ipx_count--;
233 			break;
234 		case AF_NS:
235 			route_cb.ns_count--;
236 			break;
237 		}
238 		route_cb.any_count--;
239 	}
240 	error = raw_usrreqs.pru_detach(so);
241 	crit_exit();
242 	return error;
243 }
244 
245 static int
246 rts_disconnect(struct socket *so)
247 {
248 	int error;
249 
250 	crit_enter();
251 	error = raw_usrreqs.pru_disconnect(so);
252 	crit_exit();
253 	return error;
254 }
255 
256 /* pru_listen is EOPNOTSUPP */
257 
258 static int
259 rts_peeraddr(struct socket *so, struct sockaddr **nam)
260 {
261 	int error;
262 
263 	crit_enter();
264 	error = raw_usrreqs.pru_peeraddr(so, nam);
265 	crit_exit();
266 	return error;
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 	int error;
277 
278 	crit_enter();
279 	error = raw_usrreqs.pru_send(so, flags, m, nam, control, td);
280 	crit_exit();
281 	return error;
282 }
283 
284 /* pru_sense is null */
285 
286 static int
287 rts_shutdown(struct socket *so)
288 {
289 	int error;
290 
291 	crit_enter();
292 	error = raw_usrreqs.pru_shutdown(so);
293 	crit_exit();
294 	return error;
295 }
296 
297 static int
298 rts_sockaddr(struct socket *so, struct sockaddr **nam)
299 {
300 	int error;
301 
302 	crit_enter();
303 	error = raw_usrreqs.pru_sockaddr(so, nam);
304 	crit_exit();
305 	return error;
306 }
307 
308 static struct pr_usrreqs route_usrreqs = {
309 	.pru_abort = rts_abort,
310 	.pru_accept = pru_accept_notsupp,
311 	.pru_attach = rts_attach,
312 	.pru_bind = rts_bind,
313 	.pru_connect = rts_connect,
314 	.pru_connect2 = pru_connect2_notsupp,
315 	.pru_control = pru_control_notsupp,
316 	.pru_detach = rts_detach,
317 	.pru_disconnect = rts_disconnect,
318 	.pru_listen = pru_listen_notsupp,
319 	.pru_peeraddr = rts_peeraddr,
320 	.pru_rcvd = pru_rcvd_notsupp,
321 	.pru_rcvoob = pru_rcvoob_notsupp,
322 	.pru_send = rts_send,
323 	.pru_sense = pru_sense_null,
324 	.pru_shutdown = rts_shutdown,
325 	.pru_sockaddr = rts_sockaddr,
326 	.pru_sosend = sosend,
327 	.pru_soreceive = soreceive,
328 	.pru_sopoll = sopoll
329 };
330 
331 static __inline sa_family_t
332 familyof(struct sockaddr *sa)
333 {
334 	return (sa != NULL ? sa->sa_family : 0);
335 }
336 
337 /*
338  * Routing socket input function.  The packet must be serialized onto cpu 0.
339  * We use the cpu0_soport() netisr processing loop to handle it.
340  *
341  * This looks messy but it means that anyone, including interrupt code,
342  * can send a message to the routing socket.
343  */
344 static void
345 rts_input_handler(struct netmsg *msg)
346 {
347 	static const struct sockaddr route_dst = { 2, PF_ROUTE, };
348 	struct sockproto route_proto;
349 	struct netmsg_packet *pmsg;
350 	struct mbuf *m;
351 	sa_family_t family;
352 
353 	pmsg = (void *)msg;
354 	m = pmsg->nm_packet;
355 	family = pmsg->nm_netmsg.nm_lmsg.u.ms_result;
356 	route_proto.sp_family = PF_ROUTE;
357 	route_proto.sp_protocol = family;
358 
359 	raw_input(m, &route_proto, &route_src, &route_dst);
360 }
361 
362 static void
363 rts_input(struct mbuf *m, sa_family_t family)
364 {
365 	struct netmsg_packet *pmsg;
366 	lwkt_port_t port;
367 
368 	port = cpu0_soport(NULL, NULL, NULL, 0);
369 	pmsg = &m->m_hdr.mh_netmsg;
370 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport,
371 		    0, rts_input_handler);
372 	pmsg->nm_packet = m;
373 	pmsg->nm_netmsg.nm_lmsg.u.ms_result = family;
374 	lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
375 }
376 
377 static void *
378 reallocbuf(void *ptr, size_t len, size_t olen)
379 {
380 	void *newptr;
381 
382 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
383 	if (newptr == NULL)
384 		return NULL;
385 	bcopy(ptr, newptr, olen);
386 	kfree(ptr, M_RTABLE);
387 	return (newptr);
388 }
389 
390 /*
391  * Internal helper routine for route_output().
392  */
393 static int
394 fillrtmsg(struct rt_msghdr **prtm, struct rtentry *rt,
395 	  struct rt_addrinfo *rtinfo)
396 {
397 	int msglen;
398 	struct rt_msghdr *rtm = *prtm;
399 
400 	/* Fill in rt_addrinfo for call to rt_msg_buffer(). */
401 	rtinfo->rti_dst = rt_key(rt);
402 	rtinfo->rti_gateway = rt->rt_gateway;
403 	rtinfo->rti_netmask = rt_mask(rt);		/* might be NULL */
404 	rtinfo->rti_genmask = rt->rt_genmask;		/* might be NULL */
405 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
406 		if (rt->rt_ifp != NULL) {
407 			rtinfo->rti_ifpaddr =
408 			    TAILQ_FIRST(&rt->rt_ifp->if_addrheads[mycpuid])
409 			    ->ifa->ifa_addr;
410 			rtinfo->rti_ifaaddr = rt->rt_ifa->ifa_addr;
411 			if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
412 				rtinfo->rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
413 			rtm->rtm_index = rt->rt_ifp->if_index;
414 		} else {
415 			rtinfo->rti_ifpaddr = NULL;
416 			rtinfo->rti_ifaaddr = NULL;
417 	    }
418 	}
419 
420 	msglen = rt_msgsize(rtm->rtm_type, rtinfo);
421 	if (rtm->rtm_msglen < msglen) {
422 		rtm = reallocbuf(rtm, msglen, rtm->rtm_msglen);
423 		if (rtm == NULL)
424 			return (ENOBUFS);
425 		*prtm = rtm;
426 	}
427 	rt_msg_buffer(rtm->rtm_type, rtinfo, rtm, msglen);
428 
429 	rtm->rtm_flags = rt->rt_flags;
430 	rtm->rtm_rmx = rt->rt_rmx;
431 	rtm->rtm_addrs = rtinfo->rti_addrs;
432 
433 	return (0);
434 }
435 
436 static void route_output_add_callback(int, int, struct rt_addrinfo *,
437 					struct rtentry *, void *);
438 static void route_output_delete_callback(int, int, struct rt_addrinfo *,
439 					struct rtentry *, void *);
440 static void route_output_change_callback(int, int, struct rt_addrinfo *,
441 					struct rtentry *, void *);
442 static void route_output_lock_callback(int, int, struct rt_addrinfo *,
443 					struct rtentry *, void *);
444 
445 /*ARGSUSED*/
446 static int
447 route_output(struct mbuf *m, struct socket *so, ...)
448 {
449 	struct rt_msghdr *rtm = NULL;
450 	struct rtentry *rt;
451 	struct radix_node_head *rnh;
452 	struct rawcb *rp = NULL;
453 	struct pr_output_info *oi;
454 	struct rt_addrinfo rtinfo;
455 	int len, error = 0;
456 	__va_list ap;
457 
458 	__va_start(ap, so);
459 	oi = __va_arg(ap, struct pr_output_info *);
460 	__va_end(ap);
461 
462 #define gotoerr(e) { error = e; goto flush;}
463 
464 	if (m == NULL ||
465 	    (m->m_len < sizeof(long) &&
466 	     (m = m_pullup(m, sizeof(long))) == NULL))
467 		return (ENOBUFS);
468 	if (!(m->m_flags & M_PKTHDR))
469 		panic("route_output");
470 	len = m->m_pkthdr.len;
471 	if (len < sizeof(struct rt_msghdr) ||
472 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
473 		rtinfo.rti_dst = NULL;
474 		gotoerr(EINVAL);
475 	}
476 	rtm = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
477 	if (rtm == NULL) {
478 		rtinfo.rti_dst = NULL;
479 		gotoerr(ENOBUFS);
480 	}
481 	m_copydata(m, 0, len, (caddr_t)rtm);
482 	if (rtm->rtm_version != RTM_VERSION) {
483 		rtinfo.rti_dst = NULL;
484 		gotoerr(EPROTONOSUPPORT);
485 	}
486 	rtm->rtm_pid = oi->p_pid;
487 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
488 	rtinfo.rti_addrs = rtm->rtm_addrs;
489 	if (rt_xaddrs((char *)(rtm + 1), (char *)rtm + len, &rtinfo) != 0) {
490 		rtinfo.rti_dst = NULL;
491 		gotoerr(EINVAL);
492 	}
493 	rtinfo.rti_flags = rtm->rtm_flags;
494 	if (rtinfo.rti_dst == NULL || rtinfo.rti_dst->sa_family >= AF_MAX ||
495 	    (rtinfo.rti_gateway && rtinfo.rti_gateway->sa_family >= AF_MAX))
496 		gotoerr(EINVAL);
497 
498 	if (rtinfo.rti_genmask != NULL) {
499 		struct radix_node *n;
500 
501 #define	clen(s)	(*(u_char *)(s))
502 		n = rn_addmask((char *)rtinfo.rti_genmask, TRUE, 1);
503 		if (n != NULL &&
504 		    rtinfo.rti_genmask->sa_len >= clen(n->rn_key) &&
505 		    bcmp((char *)rtinfo.rti_genmask + 1,
506 		         (char *)n->rn_key + 1, clen(n->rn_key) - 1) == 0)
507 			rtinfo.rti_genmask = (struct sockaddr *)n->rn_key;
508 		else
509 			gotoerr(ENOBUFS);
510 	}
511 
512 	/*
513 	 * Verify that the caller has the appropriate privilege; RTM_GET
514 	 * is the only operation the non-superuser is allowed.
515 	 */
516 	if (rtm->rtm_type != RTM_GET && suser_cred(so->so_cred, 0) != 0)
517 		gotoerr(EPERM);
518 
519 	switch (rtm->rtm_type) {
520 	case RTM_ADD:
521 		if (rtinfo.rti_gateway == NULL) {
522 			error = EINVAL;
523 		} else {
524 			error = rtrequest1_global(RTM_ADD, &rtinfo,
525 					  route_output_add_callback, rtm);
526 		}
527 		break;
528 	case RTM_DELETE:
529 		/*
530 		 * note: &rtm passed as argument so 'rtm' can be replaced.
531 		 */
532 		error = rtrequest1_global(RTM_DELETE, &rtinfo,
533 					  route_output_delete_callback, &rtm);
534 		break;
535 	case RTM_GET:
536 		rnh = rt_tables[mycpuid][rtinfo.rti_dst->sa_family];
537 		if (rnh == NULL) {
538 			error = EAFNOSUPPORT;
539 			break;
540 		}
541 		rt = (struct rtentry *)
542 		    rnh->rnh_lookup((char *)rtinfo.rti_dst,
543 		    		    (char *)rtinfo.rti_netmask, rnh);
544 		if (rt == NULL) {
545 			error = ESRCH;
546 			break;
547 		}
548 		rt->rt_refcnt++;
549 		if (fillrtmsg(&rtm, rt, &rtinfo) != 0)
550 			gotoerr(ENOBUFS);
551 		--rt->rt_refcnt;
552 		break;
553 	case RTM_CHANGE:
554 		error = rtrequest1_global(RTM_GET, &rtinfo,
555 					  route_output_change_callback, rtm);
556 		break;
557 	case RTM_LOCK:
558 		error = rtrequest1_global(RTM_GET, &rtinfo,
559 					  route_output_lock_callback, rtm);
560 		break;
561 	default:
562 		error = EOPNOTSUPP;
563 		break;
564 	}
565 
566 flush:
567 	if (rtm != NULL) {
568 		if (error != 0)
569 			rtm->rtm_errno = error;
570 		else
571 			rtm->rtm_flags |= RTF_DONE;
572 	}
573 
574 	/*
575 	 * Check to see if we don't want our own messages.
576 	 */
577 	if (!(so->so_options & SO_USELOOPBACK)) {
578 		if (route_cb.any_count <= 1) {
579 			if (rtm != NULL)
580 				kfree(rtm, M_RTABLE);
581 			m_freem(m);
582 			return (error);
583 		}
584 		/* There is another listener, so construct message */
585 		rp = sotorawcb(so);
586 	}
587 	if (rtm != NULL) {
588 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
589 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
590 			m_freem(m);
591 			m = NULL;
592 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
593 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
594 		kfree(rtm, M_RTABLE);
595 	}
596 	if (rp != NULL)
597 		rp->rcb_proto.sp_family = 0; /* Avoid us */
598 	if (m != NULL)
599 		rts_input(m, familyof(rtinfo.rti_dst));
600 	if (rp != NULL)
601 		rp->rcb_proto.sp_family = PF_ROUTE;
602 	return (error);
603 }
604 
605 static void
606 route_output_add_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
607 			  struct rtentry *rt, void *arg)
608 {
609 	struct rt_msghdr *rtm = arg;
610 
611 	if (error == 0 && rt != NULL) {
612 		rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
613 		    &rt->rt_rmx);
614 		rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
615 		rt->rt_rmx.rmx_locks |=
616 		    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
617 		rt->rt_genmask = rtinfo->rti_genmask;
618 	}
619 }
620 
621 static void
622 route_output_delete_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
623 			  struct rtentry *rt, void *arg)
624 {
625 	struct rt_msghdr **rtm = arg;
626 
627 	if (error == 0 && rt) {
628 		++rt->rt_refcnt;
629 		if (fillrtmsg(rtm, rt, rtinfo) != 0) {
630 			error = ENOBUFS;
631 			/* XXX no way to return the error */
632 		}
633 		--rt->rt_refcnt;
634 	}
635 }
636 
637 static void
638 route_output_change_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
639 			  struct rtentry *rt, void *arg)
640 {
641 	struct rt_msghdr *rtm = arg;
642 	struct ifaddr *ifa;
643 
644 	if (error)
645 		goto done;
646 
647 	/*
648 	 * new gateway could require new ifaddr, ifp;
649 	 * flags may also be different; ifp may be specified
650 	 * by ll sockaddr when protocol address is ambiguous
651 	 */
652 	if (((rt->rt_flags & RTF_GATEWAY) && rtinfo->rti_gateway != NULL) ||
653 	    rtinfo->rti_ifpaddr != NULL || (rtinfo->rti_ifaaddr != NULL &&
654 	    sa_equal(rtinfo->rti_ifaaddr, rt->rt_ifa->ifa_addr))
655 	) {
656 		error = rt_getifa(rtinfo);
657 		if (error != 0)
658 			goto done;
659 	}
660 	if (rtinfo->rti_gateway != NULL) {
661 		error = rt_setgate(rt, rt_key(rt), rtinfo->rti_gateway);
662 		if (error != 0)
663 			goto done;
664 	}
665 	if ((ifa = rtinfo->rti_ifa) != NULL) {
666 		struct ifaddr *oifa = rt->rt_ifa;
667 
668 		if (oifa != ifa) {
669 			if (oifa && oifa->ifa_rtrequest)
670 				oifa->ifa_rtrequest(RTM_DELETE, rt, rtinfo);
671 			IFAFREE(rt->rt_ifa);
672 			IFAREF(ifa);
673 			rt->rt_ifa = ifa;
674 			rt->rt_ifp = rtinfo->rti_ifp;
675 		}
676 	}
677 	rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, &rt->rt_rmx);
678 	if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
679 	       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, rtinfo);
680 	if (rtinfo->rti_genmask != NULL)
681 		rt->rt_genmask = rtinfo->rti_genmask;
682 done:
683 	/* XXX no way to return error */
684 	;
685 }
686 
687 static void
688 route_output_lock_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
689 			   struct rtentry *rt, void *arg)
690 {
691 	struct rt_msghdr *rtm = arg;
692 
693 	rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
694 	rt->rt_rmx.rmx_locks |=
695 		(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
696 }
697 
698 static void
699 rt_setmetrics(u_long which, struct rt_metrics *in, struct rt_metrics *out)
700 {
701 #define setmetric(flag, elt) if (which & (flag)) out->elt = in->elt;
702 	setmetric(RTV_RPIPE, rmx_recvpipe);
703 	setmetric(RTV_SPIPE, rmx_sendpipe);
704 	setmetric(RTV_SSTHRESH, rmx_ssthresh);
705 	setmetric(RTV_RTT, rmx_rtt);
706 	setmetric(RTV_RTTVAR, rmx_rttvar);
707 	setmetric(RTV_HOPCOUNT, rmx_hopcount);
708 	setmetric(RTV_MTU, rmx_mtu);
709 	setmetric(RTV_EXPIRE, rmx_expire);
710 #undef setmetric
711 }
712 
713 #define ROUNDUP(a) \
714 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
715 
716 /*
717  * Extract the addresses of the passed sockaddrs.
718  * Do a little sanity checking so as to avoid bad memory references.
719  * This data is derived straight from userland.
720  */
721 static int
722 rt_xaddrs(char *cp, char *cplim, struct rt_addrinfo *rtinfo)
723 {
724 	struct sockaddr *sa;
725 	int i;
726 
727 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
728 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
729 			continue;
730 		sa = (struct sockaddr *)cp;
731 		/*
732 		 * It won't fit.
733 		 */
734 		if ((cp + sa->sa_len) > cplim) {
735 			return (EINVAL);
736 		}
737 
738 		/*
739 		 * There are no more...  Quit now.
740 		 * If there are more bits, they are in error.
741 		 * I've seen this.  route(1) can evidently generate these.
742 		 * This causes kernel to core dump.
743 		 * For compatibility, if we see this, point to a safe address.
744 		 */
745 		if (sa->sa_len == 0) {
746 			static struct sockaddr sa_zero = {
747 				sizeof sa_zero, AF_INET,
748 			};
749 
750 			rtinfo->rti_info[i] = &sa_zero;
751 			kprintf("rtsock: received more addr bits than sockaddrs.\n");
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 	mbuftrackid(m, 32);
857 	m->m_pkthdr.len = m->m_len = hlen;
858 	m->m_pkthdr.rcvif = NULL;
859 	rtinfo->rti_addrs = 0;
860 	len = hlen;
861 	for (i = 0; i < RTAX_MAX; i++) {
862 		struct sockaddr *sa;
863 		int dlen;
864 
865 		if ((sa = rtinfo->rti_info[i]) == NULL)
866 			continue;
867 		rtinfo->rti_addrs |= (1 << i);
868 		dlen = ROUNDUP(sa->sa_len);
869 		m_copyback(m, len, dlen, (caddr_t)sa); /* can grow mbuf chain */
870 		len += dlen;
871 	}
872 	if (m->m_pkthdr.len != len) { /* one of the m_copyback() calls failed */
873 		m_freem(m);
874 		return (NULL);
875 	}
876 	rtm = mtod(m, struct rt_msghdr *);
877 	bzero(rtm, hlen);
878 	rtm->rtm_msglen = len;
879 	rtm->rtm_version = RTM_VERSION;
880 	rtm->rtm_type = type;
881 	return (m);
882 }
883 
884 /*
885  * This routine is called to generate a message from the routing
886  * socket indicating that a redirect has occurred, a routing lookup
887  * has failed, or that a protocol has detected timeouts to a particular
888  * destination.
889  */
890 void
891 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
892 {
893 	struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
894 	struct rt_msghdr *rtm;
895 	struct mbuf *m;
896 
897 	if (route_cb.any_count == 0)
898 		return;
899 	m = rt_msg_mbuf(type, rtinfo);
900 	if (m == NULL)
901 		return;
902 	rtm = mtod(m, struct rt_msghdr *);
903 	rtm->rtm_flags = RTF_DONE | flags;
904 	rtm->rtm_errno = error;
905 	rtm->rtm_addrs = rtinfo->rti_addrs;
906 	rts_input(m, familyof(dst));
907 }
908 
909 void
910 rt_dstmsg(int type, struct sockaddr *dst, int error)
911 {
912 	struct rt_msghdr *rtm;
913 	struct rt_addrinfo addrs;
914 	struct mbuf *m;
915 
916 	if (route_cb.any_count == 0)
917 		return;
918 	bzero(&addrs, sizeof(struct rt_addrinfo));
919 	addrs.rti_info[RTAX_DST] = dst;
920 	m = rt_msg_mbuf(type, &addrs);
921 	if (m == NULL)
922 		return;
923 	rtm = mtod(m, struct rt_msghdr *);
924 	rtm->rtm_flags = RTF_DONE;
925 	rtm->rtm_errno = error;
926 	rtm->rtm_addrs = addrs.rti_addrs;
927 	rts_input(m, familyof(dst));
928 }
929 
930 /*
931  * This routine is called to generate a message from the routing
932  * socket indicating that the status of a network interface has changed.
933  */
934 void
935 rt_ifmsg(struct ifnet *ifp)
936 {
937 	struct if_msghdr *ifm;
938 	struct mbuf *m;
939 	struct rt_addrinfo rtinfo;
940 
941 	if (route_cb.any_count == 0)
942 		return;
943 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
944 	m = rt_msg_mbuf(RTM_IFINFO, &rtinfo);
945 	if (m == NULL)
946 		return;
947 	ifm = mtod(m, struct if_msghdr *);
948 	ifm->ifm_index = ifp->if_index;
949 	ifm->ifm_flags = ifp->if_flags;
950 	ifm->ifm_data = ifp->if_data;
951 	ifm->ifm_addrs = 0;
952 	rts_input(m, 0);
953 }
954 
955 static void
956 rt_ifamsg(int cmd, struct ifaddr *ifa)
957 {
958 	struct ifa_msghdr *ifam;
959 	struct rt_addrinfo rtinfo;
960 	struct mbuf *m;
961 	struct ifnet *ifp = ifa->ifa_ifp;
962 
963 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
964 	rtinfo.rti_ifaaddr = ifa->ifa_addr;
965 	rtinfo.rti_ifpaddr =
966 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
967 	rtinfo.rti_netmask = ifa->ifa_netmask;
968 	rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
969 
970 	m = rt_msg_mbuf(cmd, &rtinfo);
971 	if (m == NULL)
972 		return;
973 
974 	ifam = mtod(m, struct ifa_msghdr *);
975 	ifam->ifam_index = ifp->if_index;
976 	ifam->ifam_metric = ifa->ifa_metric;
977 	ifam->ifam_flags = ifa->ifa_flags;
978 	ifam->ifam_addrs = rtinfo.rti_addrs;
979 
980 	rts_input(m, familyof(ifa->ifa_addr));
981 }
982 
983 void
984 rt_rtmsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int error)
985 {
986 	struct rt_msghdr *rtm;
987 	struct rt_addrinfo rtinfo;
988 	struct mbuf *m;
989 	struct sockaddr *dst;
990 
991 	if (rt == NULL)
992 		return;
993 
994 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
995 	rtinfo.rti_dst = dst = rt_key(rt);
996 	rtinfo.rti_gateway = rt->rt_gateway;
997 	rtinfo.rti_netmask = rt_mask(rt);
998 	if (ifp != NULL) {
999 		rtinfo.rti_ifpaddr =
1000 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1001 	}
1002 	rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
1003 
1004 	m = rt_msg_mbuf(cmd, &rtinfo);
1005 	if (m == NULL)
1006 		return;
1007 
1008 	rtm = mtod(m, struct rt_msghdr *);
1009 	if (ifp != NULL)
1010 		rtm->rtm_index = ifp->if_index;
1011 	rtm->rtm_flags |= rt->rt_flags;
1012 	rtm->rtm_errno = error;
1013 	rtm->rtm_addrs = rtinfo.rti_addrs;
1014 
1015 	rts_input(m, familyof(dst));
1016 }
1017 
1018 /*
1019  * This is called to generate messages from the routing socket
1020  * indicating a network interface has had addresses associated with it.
1021  * if we ever reverse the logic and replace messages TO the routing
1022  * socket indicate a request to configure interfaces, then it will
1023  * be unnecessary as the routing socket will automatically generate
1024  * copies of it.
1025  */
1026 void
1027 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1028 {
1029 #ifdef SCTP
1030 	/*
1031 	 * notify the SCTP stack
1032 	 * this will only get called when an address is added/deleted
1033 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1034 	 */
1035 	if (cmd == RTM_ADD)
1036 		sctp_add_ip_address(ifa);
1037 	else if (cmd == RTM_DELETE)
1038 		sctp_delete_ip_address(ifa);
1039 #endif /* SCTP */
1040 
1041 	if (route_cb.any_count == 0)
1042 		return;
1043 
1044 	if (cmd == RTM_ADD) {
1045 		rt_ifamsg(RTM_NEWADDR, ifa);
1046 		rt_rtmsg(RTM_ADD, rt, ifa->ifa_ifp, error);
1047 	} else {
1048 		KASSERT((cmd == RTM_DELETE), ("unknown cmd %d", cmd));
1049 		rt_rtmsg(RTM_DELETE, rt, ifa->ifa_ifp, error);
1050 		rt_ifamsg(RTM_DELADDR, ifa);
1051 	}
1052 }
1053 
1054 /*
1055  * This is the analogue to the rt_newaddrmsg which performs the same
1056  * function but for multicast group memberhips.  This is easier since
1057  * there is no route state to worry about.
1058  */
1059 void
1060 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1061 {
1062 	struct rt_addrinfo rtinfo;
1063 	struct mbuf *m = NULL;
1064 	struct ifnet *ifp = ifma->ifma_ifp;
1065 	struct ifma_msghdr *ifmam;
1066 
1067 	if (route_cb.any_count == 0)
1068 		return;
1069 
1070 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1071 	rtinfo.rti_ifaaddr = ifma->ifma_addr;
1072 	if (ifp != NULL && !TAILQ_EMPTY(&ifp->if_addrheads[mycpuid])) {
1073 		rtinfo.rti_ifpaddr =
1074 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1075 	}
1076 	/*
1077 	 * If a link-layer address is present, present it as a ``gateway''
1078 	 * (similarly to how ARP entries, e.g., are presented).
1079 	 */
1080 	rtinfo.rti_gateway = ifma->ifma_lladdr;
1081 
1082 	m = rt_msg_mbuf(cmd, &rtinfo);
1083 	if (m == NULL)
1084 		return;
1085 
1086 	ifmam = mtod(m, struct ifma_msghdr *);
1087 	ifmam->ifmam_index = ifp->if_index;
1088 	ifmam->ifmam_addrs = rtinfo.rti_addrs;
1089 
1090 	rts_input(m, familyof(ifma->ifma_addr));
1091 }
1092 
1093 static struct mbuf *
1094 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1095 		     struct rt_addrinfo *info)
1096 {
1097 	struct if_announcemsghdr *ifan;
1098 	struct mbuf *m;
1099 
1100 	if (route_cb.any_count == 0)
1101 		return NULL;
1102 
1103 	bzero(info, sizeof(*info));
1104 	m = rt_msg_mbuf(type, info);
1105 	if (m == NULL)
1106 		return NULL;
1107 
1108 	ifan = mtod(m, struct if_announcemsghdr *);
1109 	ifan->ifan_index = ifp->if_index;
1110 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof ifan->ifan_name);
1111 	ifan->ifan_what = what;
1112 	return m;
1113 }
1114 
1115 /*
1116  * This is called to generate routing socket messages indicating
1117  * IEEE80211 wireless events.
1118  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1119  */
1120 void
1121 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1122 {
1123 	struct rt_addrinfo info;
1124 	struct mbuf *m;
1125 
1126 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1127 	if (m == NULL)
1128 		return;
1129 
1130 	/*
1131 	 * Append the ieee80211 data.  Try to stick it in the
1132 	 * mbuf containing the ifannounce msg; otherwise allocate
1133 	 * a new mbuf and append.
1134 	 *
1135 	 * NB: we assume m is a single mbuf.
1136 	 */
1137 	if (data_len > M_TRAILINGSPACE(m)) {
1138 		struct mbuf *n = m_get(MB_DONTWAIT, MT_DATA);
1139 		if (n == NULL) {
1140 			m_freem(m);
1141 			return;
1142 		}
1143 		bcopy(data, mtod(n, void *), data_len);
1144 		n->m_len = data_len;
1145 		m->m_next = n;
1146 	} else if (data_len > 0) {
1147 		bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1148 		m->m_len += data_len;
1149 	}
1150 	mbuftrackid(m, 33);
1151 	if (m->m_flags & M_PKTHDR)
1152 		m->m_pkthdr.len += data_len;
1153 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1154 	rts_input(m, 0);
1155 }
1156 
1157 /*
1158  * This is called to generate routing socket messages indicating
1159  * network interface arrival and departure.
1160  */
1161 void
1162 rt_ifannouncemsg(struct ifnet *ifp, int what)
1163 {
1164 	struct rt_addrinfo addrinfo;
1165 	struct mbuf *m;
1166 
1167 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &addrinfo);
1168 	if (m != NULL)
1169 		rts_input(m, 0);
1170 }
1171 
1172 static int
1173 resizewalkarg(struct walkarg *w, int len)
1174 {
1175 	void *newptr;
1176 
1177 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
1178 	if (newptr == NULL)
1179 		return (ENOMEM);
1180 	if (w->w_tmem != NULL)
1181 		kfree(w->w_tmem, M_RTABLE);
1182 	w->w_tmem = newptr;
1183 	w->w_tmemsize = len;
1184 	return (0);
1185 }
1186 
1187 /*
1188  * This is used in dumping the kernel table via sysctl().
1189  */
1190 int
1191 sysctl_dumpentry(struct radix_node *rn, void *vw)
1192 {
1193 	struct walkarg *w = vw;
1194 	struct rtentry *rt = (struct rtentry *)rn;
1195 	struct rt_addrinfo rtinfo;
1196 	int error, msglen;
1197 
1198 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1199 		return 0;
1200 
1201 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1202 	rtinfo.rti_dst = rt_key(rt);
1203 	rtinfo.rti_gateway = rt->rt_gateway;
1204 	rtinfo.rti_netmask = rt_mask(rt);
1205 	rtinfo.rti_genmask = rt->rt_genmask;
1206 	if (rt->rt_ifp != NULL) {
1207 		rtinfo.rti_ifpaddr =
1208 		TAILQ_FIRST(&rt->rt_ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1209 		rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
1210 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1211 			rtinfo.rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
1212 	}
1213 	msglen = rt_msgsize(RTM_GET, &rtinfo);
1214 	if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1215 		return (ENOMEM);
1216 	rt_msg_buffer(RTM_GET, &rtinfo, w->w_tmem, msglen);
1217 	if (w->w_req != NULL) {
1218 		struct rt_msghdr *rtm = w->w_tmem;
1219 
1220 		rtm->rtm_flags = rt->rt_flags;
1221 		rtm->rtm_use = rt->rt_use;
1222 		rtm->rtm_rmx = rt->rt_rmx;
1223 		rtm->rtm_index = rt->rt_ifp->if_index;
1224 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1225 		rtm->rtm_addrs = rtinfo.rti_addrs;
1226 		error = SYSCTL_OUT(w->w_req, rtm, msglen);
1227 		return (error);
1228 	}
1229 	return (0);
1230 }
1231 
1232 static int
1233 sysctl_iflist(int af, struct walkarg *w)
1234 {
1235 	struct ifnet *ifp;
1236 	struct rt_addrinfo rtinfo;
1237 	int msglen, error;
1238 
1239 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1240 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1241 		struct ifaddr_container *ifac;
1242 		struct ifaddr *ifa;
1243 
1244 		if (w->w_arg && w->w_arg != ifp->if_index)
1245 			continue;
1246 		ifac = TAILQ_FIRST(&ifp->if_addrheads[mycpuid]);
1247 		ifa = ifac->ifa;
1248 		rtinfo.rti_ifpaddr = ifa->ifa_addr;
1249 		msglen = rt_msgsize(RTM_IFINFO, &rtinfo);
1250 		if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1251 			return (ENOMEM);
1252 		rt_msg_buffer(RTM_IFINFO, &rtinfo, w->w_tmem, msglen);
1253 		rtinfo.rti_ifpaddr = NULL;
1254 		if (w->w_req != NULL && w->w_tmem != NULL) {
1255 			struct if_msghdr *ifm = w->w_tmem;
1256 
1257 			ifm->ifm_index = ifp->if_index;
1258 			ifm->ifm_flags = ifp->if_flags;
1259 			ifm->ifm_data = ifp->if_data;
1260 			ifm->ifm_addrs = rtinfo.rti_addrs;
1261 			error = SYSCTL_OUT(w->w_req, ifm, msglen);
1262 			if (error)
1263 				return (error);
1264 		}
1265 		while ((ifac = TAILQ_NEXT(ifac, ifa_link)) != NULL) {
1266 			ifa = ifac->ifa;
1267 
1268 			if (af && af != ifa->ifa_addr->sa_family)
1269 				continue;
1270 			if (curproc->p_ucred->cr_prison &&
1271 			    prison_if(curproc->p_ucred, ifa->ifa_addr))
1272 				continue;
1273 			rtinfo.rti_ifaaddr = ifa->ifa_addr;
1274 			rtinfo.rti_netmask = ifa->ifa_netmask;
1275 			rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
1276 			msglen = rt_msgsize(RTM_NEWADDR, &rtinfo);
1277 			if (w->w_tmemsize < msglen &&
1278 			    resizewalkarg(w, msglen) != 0)
1279 				return (ENOMEM);
1280 			rt_msg_buffer(RTM_NEWADDR, &rtinfo, w->w_tmem, msglen);
1281 			if (w->w_req != NULL) {
1282 				struct ifa_msghdr *ifam = w->w_tmem;
1283 
1284 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1285 				ifam->ifam_flags = ifa->ifa_flags;
1286 				ifam->ifam_metric = ifa->ifa_metric;
1287 				ifam->ifam_addrs = rtinfo.rti_addrs;
1288 				error = SYSCTL_OUT(w->w_req, w->w_tmem, msglen);
1289 				if (error)
1290 					return (error);
1291 			}
1292 		}
1293 		rtinfo.rti_netmask = NULL;
1294 		rtinfo.rti_ifaaddr = NULL;
1295 		rtinfo.rti_bcastaddr = NULL;
1296 	}
1297 	return (0);
1298 }
1299 
1300 static int
1301 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1302 {
1303 	int	*name = (int *)arg1;
1304 	u_int	namelen = arg2;
1305 	struct radix_node_head *rnh;
1306 	int	i, error = EINVAL;
1307 	int	origcpu;
1308 	u_char  af;
1309 	struct	walkarg w;
1310 
1311 	name ++;
1312 	namelen--;
1313 	if (req->newptr)
1314 		return (EPERM);
1315 	if (namelen != 3 && namelen != 4)
1316 		return (EINVAL);
1317 	af = name[0];
1318 	bzero(&w, sizeof w);
1319 	w.w_op = name[1];
1320 	w.w_arg = name[2];
1321 	w.w_req = req;
1322 
1323 	/*
1324 	 * Optional third argument specifies cpu, used primarily for
1325 	 * debugging the route table.
1326 	 */
1327 	if (namelen == 4) {
1328 		if (name[3] < 0 || name[3] >= ncpus)
1329 			return (EINVAL);
1330 		origcpu = mycpuid;
1331 		lwkt_migratecpu(name[3]);
1332 	} else {
1333 		origcpu = -1;
1334 	}
1335 	crit_enter();
1336 	switch (w.w_op) {
1337 	case NET_RT_DUMP:
1338 	case NET_RT_FLAGS:
1339 		for (i = 1; i <= AF_MAX; i++)
1340 			if ((rnh = rt_tables[mycpuid][i]) &&
1341 			    (af == 0 || af == i) &&
1342 			    (error = rnh->rnh_walktree(rnh,
1343 						       sysctl_dumpentry, &w)))
1344 				break;
1345 		break;
1346 
1347 	case NET_RT_IFLIST:
1348 		error = sysctl_iflist(af, &w);
1349 	}
1350 	crit_exit();
1351 	if (w.w_tmem != NULL)
1352 		kfree(w.w_tmem, M_RTABLE);
1353 	if (origcpu >= 0)
1354 		lwkt_migratecpu(origcpu);
1355 	return (error);
1356 }
1357 
1358 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1359 
1360 /*
1361  * Definitions of protocols supported in the ROUTE domain.
1362  */
1363 
1364 static struct domain routedomain;		/* or at least forward */
1365 
1366 static struct protosw routesw[] = {
1367 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
1368   0,		route_output,	raw_ctlinput,	0,
1369   cpu0_soport,
1370   raw_init,	0,		0,		0,
1371   &route_usrreqs
1372 }
1373 };
1374 
1375 static struct domain routedomain = {
1376 	PF_ROUTE, "route", NULL, NULL, NULL,
1377 	routesw, &routesw[(sizeof routesw)/(sizeof routesw[0])],
1378 };
1379 
1380 DOMAIN_SET(route);
1381 
1382