xref: /dragonfly/sys/net/rtsock.c (revision 1847e88f)
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) 2004, 2005 Jeffrey M. Hsu.  All rights reserved.
35  *
36  * License terms: all terms for the DragonFly license above plus the following:
37  *
38  * 4. All advertising materials mentioning features or use of this software
39  *    must display the following acknowledgement:
40  *
41  *	This product includes software developed by Jeffrey M. Hsu
42  *	for the DragonFly Project.
43  *
44  *    This requirement may be waived with permission from Jeffrey Hsu.
45  *    Permission will be granted to any DragonFly user for free.
46  *    This requirement will sunset and may be removed on Jan 31, 2006,
47  *    after which the standard DragonFly license (as shown above) will
48  *    apply.
49  */
50 
51 /*
52  * Copyright (c) 1988, 1991, 1993
53  *	The Regents of the University of California.  All rights reserved.
54  *
55  * Redistribution and use in source and binary forms, with or without
56  * modification, are permitted provided that the following conditions
57  * are met:
58  * 1. Redistributions of source code must retain the above copyright
59  *    notice, this list of conditions and the following disclaimer.
60  * 2. Redistributions in binary form must reproduce the above copyright
61  *    notice, this list of conditions and the following disclaimer in the
62  *    documentation and/or other materials provided with the distribution.
63  * 3. All advertising materials mentioning features or use of this software
64  *    must display the following acknowledgement:
65  *	This product includes software developed by the University of
66  *	California, Berkeley and its contributors.
67  * 4. Neither the name of the University nor the names of its contributors
68  *    may be used to endorse or promote products derived from this software
69  *    without specific prior written permission.
70  *
71  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
72  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
73  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
74  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
75  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
76  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
77  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
78  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
79  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
80  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
81  * SUCH DAMAGE.
82  *
83  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
84  * $FreeBSD: src/sys/net/rtsock.c,v 1.44.2.11 2002/12/04 14:05:41 ru Exp $
85  * $DragonFly: src/sys/net/rtsock.c,v 1.30 2006/01/31 19:05:35 dillon Exp $
86  */
87 
88 #include "opt_sctp.h"
89 
90 #include <sys/param.h>
91 #include <sys/systm.h>
92 #include <sys/kernel.h>
93 #include <sys/sysctl.h>
94 #include <sys/proc.h>
95 #include <sys/malloc.h>
96 #include <sys/mbuf.h>
97 #include <sys/protosw.h>
98 #include <sys/socket.h>
99 #include <sys/socketvar.h>
100 #include <sys/domain.h>
101 #include <sys/thread2.h>
102 
103 #include <net/if.h>
104 #include <net/route.h>
105 #include <net/raw_cb.h>
106 
107 #ifdef SCTP
108 extern void sctp_add_ip_address(struct ifaddr *ifa);
109 extern void sctp_delete_ip_address(struct ifaddr *ifa);
110 #endif /* SCTP */
111 
112 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
113 
114 static struct route_cb {
115 	int	ip_count;
116 	int	ip6_count;
117 	int	ipx_count;
118 	int	ns_count;
119 	int	any_count;
120 } route_cb;
121 
122 static const struct sockaddr route_src = { 2, PF_ROUTE, };
123 
124 struct walkarg {
125 	int	w_tmemsize;
126 	int	w_op, w_arg;
127 	void	*w_tmem;
128 	struct sysctl_req *w_req;
129 };
130 
131 static struct mbuf *
132 		rt_msg_mbuf (int, struct rt_addrinfo *);
133 static void	rt_msg_buffer (int, struct rt_addrinfo *, void *buf, int len);
134 static int	rt_msgsize (int type, struct rt_addrinfo *rtinfo);
135 static int	rt_xaddrs (char *, char *, struct rt_addrinfo *);
136 static int	sysctl_dumpentry (struct radix_node *rn, void *vw);
137 static int	sysctl_iflist (int af, struct walkarg *w);
138 static int	route_output(struct mbuf *, struct socket *, ...);
139 static void	rt_setmetrics (u_long, struct rt_metrics *,
140 			       struct rt_metrics *);
141 
142 /*
143  * It really doesn't make any sense at all for this code to share much
144  * with raw_usrreq.c, since its functionality is so restricted.  XXX
145  */
146 static int
147 rts_abort(struct socket *so)
148 {
149 	int error;
150 
151 	crit_enter();
152 	error = raw_usrreqs.pru_abort(so);
153 	crit_exit();
154 	return error;
155 }
156 
157 /* pru_accept is EOPNOTSUPP */
158 
159 static int
160 rts_attach(struct socket *so, int proto, struct pru_attach_info *ai)
161 {
162 	struct rawcb *rp;
163 	int error;
164 
165 	if (sotorawcb(so) != NULL)
166 		return EISCONN;	/* XXX panic? */
167 
168 	rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
169 	if (rp == NULL)
170 		return ENOBUFS;
171 
172 	/*
173 	 * The critical section is necessary to block protocols from sending
174 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
175 	 * this PCB is extant but incompletely initialized.
176 	 * Probably we should try to do more of this work beforehand and
177 	 * eliminate the critical section.
178 	 */
179 	crit_enter();
180 	so->so_pcb = rp;
181 	error = raw_attach(so, proto, ai->sb_rlimit);
182 	rp = sotorawcb(so);
183 	if (error) {
184 		crit_exit();
185 		free(rp, M_PCB);
186 		return error;
187 	}
188 	switch(rp->rcb_proto.sp_protocol) {
189 	case AF_INET:
190 		route_cb.ip_count++;
191 		break;
192 	case AF_INET6:
193 		route_cb.ip6_count++;
194 		break;
195 	case AF_IPX:
196 		route_cb.ipx_count++;
197 		break;
198 	case AF_NS:
199 		route_cb.ns_count++;
200 		break;
201 	}
202 	rp->rcb_faddr = &route_src;
203 	route_cb.any_count++;
204 	soisconnected(so);
205 	so->so_options |= SO_USELOOPBACK;
206 	crit_exit();
207 	return 0;
208 }
209 
210 static int
211 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
212 {
213 	int error;
214 
215 	crit_enter();
216 	error = raw_usrreqs.pru_bind(so, nam, td); /* xxx just EINVAL */
217 	crit_exit();
218 	return error;
219 }
220 
221 static int
222 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
223 {
224 	int error;
225 
226 	crit_enter();
227 	error = raw_usrreqs.pru_connect(so, nam, td); /* XXX just EINVAL */
228 	crit_exit();
229 	return error;
230 }
231 
232 /* pru_connect2 is EOPNOTSUPP */
233 /* pru_control is EOPNOTSUPP */
234 
235 static int
236 rts_detach(struct socket *so)
237 {
238 	struct rawcb *rp = sotorawcb(so);
239 	int error;
240 
241 	crit_enter();
242 	if (rp != NULL) {
243 		switch(rp->rcb_proto.sp_protocol) {
244 		case AF_INET:
245 			route_cb.ip_count--;
246 			break;
247 		case AF_INET6:
248 			route_cb.ip6_count--;
249 			break;
250 		case AF_IPX:
251 			route_cb.ipx_count--;
252 			break;
253 		case AF_NS:
254 			route_cb.ns_count--;
255 			break;
256 		}
257 		route_cb.any_count--;
258 	}
259 	error = raw_usrreqs.pru_detach(so);
260 	crit_exit();
261 	return error;
262 }
263 
264 static int
265 rts_disconnect(struct socket *so)
266 {
267 	int error;
268 
269 	crit_enter();
270 	error = raw_usrreqs.pru_disconnect(so);
271 	crit_exit();
272 	return error;
273 }
274 
275 /* pru_listen is EOPNOTSUPP */
276 
277 static int
278 rts_peeraddr(struct socket *so, struct sockaddr **nam)
279 {
280 	int error;
281 
282 	crit_enter();
283 	error = raw_usrreqs.pru_peeraddr(so, nam);
284 	crit_exit();
285 	return error;
286 }
287 
288 /* pru_rcvd is EOPNOTSUPP */
289 /* pru_rcvoob is EOPNOTSUPP */
290 
291 static int
292 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
293 	 struct mbuf *control, struct thread *td)
294 {
295 	int error;
296 
297 	crit_enter();
298 	error = raw_usrreqs.pru_send(so, flags, m, nam, control, td);
299 	crit_exit();
300 	return error;
301 }
302 
303 /* pru_sense is null */
304 
305 static int
306 rts_shutdown(struct socket *so)
307 {
308 	int error;
309 
310 	crit_enter();
311 	error = raw_usrreqs.pru_shutdown(so);
312 	crit_exit();
313 	return error;
314 }
315 
316 static int
317 rts_sockaddr(struct socket *so, struct sockaddr **nam)
318 {
319 	int error;
320 
321 	crit_enter();
322 	error = raw_usrreqs.pru_sockaddr(so, nam);
323 	crit_exit();
324 	return error;
325 }
326 
327 static struct pr_usrreqs route_usrreqs = {
328 	rts_abort, pru_accept_notsupp, rts_attach, rts_bind, rts_connect,
329 	pru_connect2_notsupp, pru_control_notsupp, rts_detach, rts_disconnect,
330 	pru_listen_notsupp, rts_peeraddr, pru_rcvd_notsupp, pru_rcvoob_notsupp,
331 	rts_send, pru_sense_null, rts_shutdown, rts_sockaddr,
332 	sosend, soreceive, sopoll
333 };
334 
335 static __inline sa_family_t
336 familyof(struct sockaddr *sa)
337 {
338 	return (sa != NULL ? sa->sa_family : 0);
339 }
340 
341 static void
342 rts_input(struct mbuf *m, sa_family_t family)
343 {
344 	static const struct sockaddr route_dst = { 2, PF_ROUTE, };
345 	struct sockproto route_proto = { PF_ROUTE, family };
346 
347 	raw_input(m, &route_proto, &route_src, &route_dst);
348 }
349 
350 static void *
351 reallocbuf(void *ptr, size_t len, size_t olen)
352 {
353 	void *newptr;
354 
355 	newptr = malloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
356 	if (newptr == NULL)
357 		return NULL;
358 	bcopy(ptr, newptr, olen);
359 	free(ptr, M_RTABLE);
360 	return (newptr);
361 }
362 
363 /*
364  * Internal helper routine for route_output().
365  */
366 static int
367 fillrtmsg(struct rt_msghdr **prtm, struct rtentry *rt,
368 	  struct rt_addrinfo *rtinfo)
369 {
370 	int msglen;
371 	struct rt_msghdr *rtm = *prtm;
372 
373 	/* Fill in rt_addrinfo for call to rt_msg_buffer(). */
374 	rtinfo->rti_dst = rt_key(rt);
375 	rtinfo->rti_gateway = rt->rt_gateway;
376 	rtinfo->rti_netmask = rt_mask(rt);		/* might be NULL */
377 	rtinfo->rti_genmask = rt->rt_genmask;		/* might be NULL */
378 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
379 		if (rt->rt_ifp != NULL) {
380 			rtinfo->rti_ifpaddr =
381 			    TAILQ_FIRST(&rt->rt_ifp->if_addrhead)->ifa_addr;
382 			rtinfo->rti_ifaaddr = rt->rt_ifa->ifa_addr;
383 			if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
384 				rtinfo->rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
385 			rtm->rtm_index = rt->rt_ifp->if_index;
386 		} else {
387 			rtinfo->rti_ifpaddr = NULL;
388 			rtinfo->rti_ifaaddr = NULL;
389 	    }
390 	}
391 
392 	msglen = rt_msgsize(rtm->rtm_type, rtinfo);
393 	if (rtm->rtm_msglen < msglen) {
394 		rtm = reallocbuf(rtm, msglen, rtm->rtm_msglen);
395 		if (rtm == NULL)
396 			return (ENOBUFS);
397 		*prtm = rtm;
398 	}
399 	rt_msg_buffer(rtm->rtm_type, rtinfo, rtm, msglen);
400 
401 	rtm->rtm_flags = rt->rt_flags;
402 	rtm->rtm_rmx = rt->rt_rmx;
403 	rtm->rtm_addrs = rtinfo->rti_addrs;
404 
405 	return (0);
406 }
407 
408 static void route_output_add_callback(int, int, struct rt_addrinfo *,
409 					struct rtentry *, void *);
410 static void route_output_delete_callback(int, int, struct rt_addrinfo *,
411 					struct rtentry *, void *);
412 static void route_output_change_callback(int, int, struct rt_addrinfo *,
413 					struct rtentry *, void *);
414 static void route_output_lock_callback(int, int, struct rt_addrinfo *,
415 					struct rtentry *, void *);
416 
417 /*ARGSUSED*/
418 static int
419 route_output(struct mbuf *m, struct socket *so, ...)
420 {
421 	struct rt_msghdr *rtm = NULL;
422 	struct rtentry *rt;
423 	struct radix_node_head *rnh;
424 	struct rawcb *rp = NULL;
425 	struct pr_output_info *oi;
426 	struct rt_addrinfo rtinfo;
427 	int len, error = 0;
428 	__va_list ap;
429 
430 	__va_start(ap, so);
431 	oi = __va_arg(ap, struct pr_output_info *);
432 	__va_end(ap);
433 
434 #define gotoerr(e) { error = e; goto flush;}
435 
436 	if (m == NULL ||
437 	    (m->m_len < sizeof(long) &&
438 	     (m = m_pullup(m, sizeof(long))) == NULL))
439 		return (ENOBUFS);
440 	if (!(m->m_flags & M_PKTHDR))
441 		panic("route_output");
442 	len = m->m_pkthdr.len;
443 	if (len < sizeof(struct rt_msghdr) ||
444 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
445 		rtinfo.rti_dst = NULL;
446 		gotoerr(EINVAL);
447 	}
448 	rtm = malloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
449 	if (rtm == NULL) {
450 		rtinfo.rti_dst = NULL;
451 		gotoerr(ENOBUFS);
452 	}
453 	m_copydata(m, 0, len, (caddr_t)rtm);
454 	if (rtm->rtm_version != RTM_VERSION) {
455 		rtinfo.rti_dst = NULL;
456 		gotoerr(EPROTONOSUPPORT);
457 	}
458 	rtm->rtm_pid = oi->p_pid;
459 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
460 	rtinfo.rti_addrs = rtm->rtm_addrs;
461 	if (rt_xaddrs((char *)(rtm + 1), (char *)rtm + len, &rtinfo) != 0) {
462 		rtinfo.rti_dst = NULL;
463 		gotoerr(EINVAL);
464 	}
465 	rtinfo.rti_flags = rtm->rtm_flags;
466 	if (rtinfo.rti_dst == NULL || rtinfo.rti_dst->sa_family >= AF_MAX ||
467 	    (rtinfo.rti_gateway && rtinfo.rti_gateway->sa_family >= AF_MAX))
468 		gotoerr(EINVAL);
469 
470 	if (rtinfo.rti_genmask != NULL) {
471 		struct radix_node *n;
472 
473 #define	clen(s)	(*(u_char *)(s))
474 		n = rn_addmask((char *)rtinfo.rti_genmask, TRUE, 1);
475 		if (n != NULL &&
476 		    rtinfo.rti_genmask->sa_len >= clen(n->rn_key) &&
477 		    bcmp((char *)rtinfo.rti_genmask + 1,
478 		         (char *)n->rn_key + 1, clen(n->rn_key) - 1) == 0)
479 			rtinfo.rti_genmask = (struct sockaddr *)n->rn_key;
480 		else
481 			gotoerr(ENOBUFS);
482 	}
483 
484 	/*
485 	 * Verify that the caller has the appropriate privilege; RTM_GET
486 	 * is the only operation the non-superuser is allowed.
487 	 */
488 	if (rtm->rtm_type != RTM_GET && suser_cred(so->so_cred, 0) != 0)
489 		gotoerr(EPERM);
490 
491 	switch (rtm->rtm_type) {
492 	case RTM_ADD:
493 		if (rtinfo.rti_gateway == NULL) {
494 			error = EINVAL;
495 		} else {
496 			error = rtrequest1_global(RTM_ADD, &rtinfo,
497 					  route_output_add_callback, rtm);
498 		}
499 		break;
500 	case RTM_DELETE:
501 		/*
502 		 * note: &rtm passed as argument so 'rtm' can be replaced.
503 		 */
504 		error = rtrequest1_global(RTM_DELETE, &rtinfo,
505 					  route_output_delete_callback, &rtm);
506 		break;
507 	case RTM_GET:
508 		rnh = rt_tables[mycpuid][rtinfo.rti_dst->sa_family];
509 		if (rnh == NULL) {
510 			error = EAFNOSUPPORT;
511 			break;
512 		}
513 		rt = (struct rtentry *)
514 		    rnh->rnh_lookup((char *)rtinfo.rti_dst,
515 		    		    (char *)rtinfo.rti_netmask, rnh);
516 		if (rt == NULL) {
517 			error = ESRCH;
518 			break;
519 		}
520 		rt->rt_refcnt++;
521 		if (fillrtmsg(&rtm, rt, &rtinfo) != 0)
522 			gotoerr(ENOBUFS);
523 		--rt->rt_refcnt;
524 		break;
525 	case RTM_CHANGE:
526 		error = rtrequest1_global(RTM_GET, &rtinfo,
527 					  route_output_change_callback, rtm);
528 		break;
529 	case RTM_LOCK:
530 		error = rtrequest1_global(RTM_GET, &rtinfo,
531 					  route_output_lock_callback, rtm);
532 		break;
533 	default:
534 		error = EOPNOTSUPP;
535 		break;
536 	}
537 
538 flush:
539 	if (rtm != NULL) {
540 		if (error != 0)
541 			rtm->rtm_errno = error;
542 		else
543 			rtm->rtm_flags |= RTF_DONE;
544 	}
545 
546 	/*
547 	 * Check to see if we don't want our own messages.
548 	 */
549 	if (!(so->so_options & SO_USELOOPBACK)) {
550 		if (route_cb.any_count <= 1) {
551 			if (rtm != NULL)
552 				free(rtm, M_RTABLE);
553 			m_freem(m);
554 			return (error);
555 		}
556 		/* There is another listener, so construct message */
557 		rp = sotorawcb(so);
558 	}
559 	if (rtm != NULL) {
560 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
561 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
562 			m_freem(m);
563 			m = NULL;
564 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
565 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
566 		free(rtm, M_RTABLE);
567 	}
568 	if (rp != NULL)
569 		rp->rcb_proto.sp_family = 0; /* Avoid us */
570 	if (m != NULL)
571 		rts_input(m, familyof(rtinfo.rti_dst));
572 	if (rp != NULL)
573 		rp->rcb_proto.sp_family = PF_ROUTE;
574 	return (error);
575 }
576 
577 static void
578 route_output_add_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
579 			  struct rtentry *rt, void *arg)
580 {
581 	struct rt_msghdr *rtm = arg;
582 
583 	if (error == 0 && rt != NULL) {
584 		rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
585 		    &rt->rt_rmx);
586 		rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
587 		rt->rt_rmx.rmx_locks |=
588 		    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
589 		rt->rt_genmask = rtinfo->rti_genmask;
590 	}
591 }
592 
593 static void
594 route_output_delete_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
595 			  struct rtentry *rt, void *arg)
596 {
597 	struct rt_msghdr **rtm = arg;
598 
599 	if (error == 0 && rt) {
600 		++rt->rt_refcnt;
601 		if (fillrtmsg(rtm, rt, rtinfo) != 0) {
602 			error = ENOBUFS;
603 			/* XXX no way to return the error */
604 		}
605 		--rt->rt_refcnt;
606 	}
607 }
608 
609 static void
610 route_output_change_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
611 			  struct rtentry *rt, void *arg)
612 {
613 	struct rt_msghdr *rtm = arg;
614 	struct ifaddr *ifa;
615 
616 	if (error)
617 		goto done;
618 
619 	/*
620 	 * new gateway could require new ifaddr, ifp;
621 	 * flags may also be different; ifp may be specified
622 	 * by ll sockaddr when protocol address is ambiguous
623 	 */
624 	if (((rt->rt_flags & RTF_GATEWAY) && rtinfo->rti_gateway != NULL) ||
625 	    rtinfo->rti_ifpaddr != NULL || (rtinfo->rti_ifaaddr != NULL &&
626 	    sa_equal(rtinfo->rti_ifaaddr, rt->rt_ifa->ifa_addr))
627 	) {
628 		error = rt_getifa(rtinfo);
629 		if (error != 0)
630 			goto done;
631 	}
632 	if (rtinfo->rti_gateway != NULL) {
633 		error = rt_setgate(rt, rt_key(rt), rtinfo->rti_gateway);
634 		if (error != 0)
635 			goto done;
636 	}
637 	if ((ifa = rtinfo->rti_ifa) != NULL) {
638 		struct ifaddr *oifa = rt->rt_ifa;
639 
640 		if (oifa != ifa) {
641 			if (oifa && oifa->ifa_rtrequest)
642 				oifa->ifa_rtrequest(RTM_DELETE, rt, rtinfo);
643 			IFAFREE(rt->rt_ifa);
644 			IFAREF(ifa);
645 			rt->rt_ifa = ifa;
646 			rt->rt_ifp = rtinfo->rti_ifp;
647 		}
648 	}
649 	rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, &rt->rt_rmx);
650 	if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
651 	       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, rtinfo);
652 	if (rtinfo->rti_genmask != NULL)
653 		rt->rt_genmask = rtinfo->rti_genmask;
654 done:
655 	/* XXX no way to return error */
656 	;
657 }
658 
659 static void
660 route_output_lock_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
661 			   struct rtentry *rt, void *arg)
662 {
663 	struct rt_msghdr *rtm = arg;
664 
665 	rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
666 	rt->rt_rmx.rmx_locks |=
667 		(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
668 }
669 
670 static void
671 rt_setmetrics(u_long which, struct rt_metrics *in, struct rt_metrics *out)
672 {
673 #define setmetric(flag, elt) if (which & (flag)) out->elt = in->elt;
674 	setmetric(RTV_RPIPE, rmx_recvpipe);
675 	setmetric(RTV_SPIPE, rmx_sendpipe);
676 	setmetric(RTV_SSTHRESH, rmx_ssthresh);
677 	setmetric(RTV_RTT, rmx_rtt);
678 	setmetric(RTV_RTTVAR, rmx_rttvar);
679 	setmetric(RTV_HOPCOUNT, rmx_hopcount);
680 	setmetric(RTV_MTU, rmx_mtu);
681 	setmetric(RTV_EXPIRE, rmx_expire);
682 #undef setmetric
683 }
684 
685 #define ROUNDUP(a) \
686 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
687 
688 /*
689  * Extract the addresses of the passed sockaddrs.
690  * Do a little sanity checking so as to avoid bad memory references.
691  * This data is derived straight from userland.
692  */
693 static int
694 rt_xaddrs(char *cp, char *cplim, struct rt_addrinfo *rtinfo)
695 {
696 	struct sockaddr *sa;
697 	int i;
698 
699 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
700 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
701 			continue;
702 		sa = (struct sockaddr *)cp;
703 		/*
704 		 * It won't fit.
705 		 */
706 		if ((cp + sa->sa_len) > cplim) {
707 			return (EINVAL);
708 		}
709 
710 		/*
711 		 * There are no more...  Quit now.
712 		 * If there are more bits, they are in error.
713 		 * I've seen this.  route(1) can evidently generate these.
714 		 * This causes kernel to core dump.
715 		 * For compatibility, if we see this, point to a safe address.
716 		 */
717 		if (sa->sa_len == 0) {
718 			static struct sockaddr sa_zero = {
719 				sizeof sa_zero, AF_INET,
720 			};
721 
722 			rtinfo->rti_info[i] = &sa_zero;
723 			return (0); /* should be EINVAL but for compat */
724 		}
725 
726 		/* Accept the sockaddr. */
727 		rtinfo->rti_info[i] = sa;
728 		cp += ROUNDUP(sa->sa_len);
729 	}
730 	return (0);
731 }
732 
733 static int
734 rt_msghdrsize(int type)
735 {
736 	switch (type) {
737 	case RTM_DELADDR:
738 	case RTM_NEWADDR:
739 		return sizeof(struct ifa_msghdr);
740 	case RTM_DELMADDR:
741 	case RTM_NEWMADDR:
742 		return sizeof(struct ifma_msghdr);
743 	case RTM_IFINFO:
744 		return sizeof(struct if_msghdr);
745 	case RTM_IFANNOUNCE:
746 		return sizeof(struct if_announcemsghdr);
747 	default:
748 		return sizeof(struct rt_msghdr);
749 	}
750 }
751 
752 static int
753 rt_msgsize(int type, struct rt_addrinfo *rtinfo)
754 {
755 	int len, i;
756 
757 	len = rt_msghdrsize(type);
758 	for (i = 0; i < RTAX_MAX; i++) {
759 		if (rtinfo->rti_info[i] != NULL)
760 			len += ROUNDUP(rtinfo->rti_info[i]->sa_len);
761 	}
762 	len = ALIGN(len);
763 	return len;
764 }
765 
766 /*
767  * Build a routing message in a buffer.
768  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
769  * to the end of the buffer after the message header.
770  *
771  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
772  * This side-effect can be avoided if we reorder the addrs bitmask field in all
773  * the route messages to line up so we can set it here instead of back in the
774  * calling routine.
775  */
776 static void
777 rt_msg_buffer(int type, struct rt_addrinfo *rtinfo, void *buf, int msglen)
778 {
779 	struct rt_msghdr *rtm;
780 	char *cp;
781 	int dlen, i;
782 
783 	rtm = (struct rt_msghdr *) buf;
784 	rtm->rtm_version = RTM_VERSION;
785 	rtm->rtm_type = type;
786 	rtm->rtm_msglen = msglen;
787 
788 	cp = (char *)buf + rt_msghdrsize(type);
789 	rtinfo->rti_addrs = 0;
790 	for (i = 0; i < RTAX_MAX; i++) {
791 		struct sockaddr *sa;
792 
793 		if ((sa = rtinfo->rti_info[i]) == NULL)
794 			continue;
795 		rtinfo->rti_addrs |= (1 << i);
796 		dlen = ROUNDUP(sa->sa_len);
797 		bcopy(sa, cp, dlen);
798 		cp += dlen;
799 	}
800 }
801 
802 /*
803  * Build a routing message in a mbuf chain.
804  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
805  * to the end of the mbuf after the message header.
806  *
807  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
808  * This side-effect can be avoided if we reorder the addrs bitmask field in all
809  * the route messages to line up so we can set it here instead of back in the
810  * calling routine.
811  */
812 static struct mbuf *
813 rt_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
814 {
815 	struct mbuf *m;
816 	struct rt_msghdr *rtm;
817 	int hlen, len;
818 	int i;
819 
820 	hlen = rt_msghdrsize(type);
821 	KASSERT(hlen <= MCLBYTES, ("rt_msg_mbuf: hlen %d doesn't fit", hlen));
822 
823 	m = m_getl(hlen, MB_DONTWAIT, MT_DATA, M_PKTHDR, NULL);
824 	if (m == NULL)
825 		return (NULL);
826 	m->m_pkthdr.len = m->m_len = hlen;
827 	m->m_pkthdr.rcvif = NULL;
828 	rtinfo->rti_addrs = 0;
829 	len = hlen;
830 	for (i = 0; i < RTAX_MAX; i++) {
831 		struct sockaddr *sa;
832 		int dlen;
833 
834 		if ((sa = rtinfo->rti_info[i]) == NULL)
835 			continue;
836 		rtinfo->rti_addrs |= (1 << i);
837 		dlen = ROUNDUP(sa->sa_len);
838 		m_copyback(m, len, dlen, (caddr_t)sa); /* can grow mbuf chain */
839 		len += dlen;
840 	}
841 	if (m->m_pkthdr.len != len) { /* one of the m_copyback() calls failed */
842 		m_freem(m);
843 		return (NULL);
844 	}
845 	rtm = mtod(m, struct rt_msghdr *);
846 	bzero(rtm, hlen);
847 	rtm->rtm_msglen = len;
848 	rtm->rtm_version = RTM_VERSION;
849 	rtm->rtm_type = type;
850 	return (m);
851 }
852 
853 /*
854  * This routine is called to generate a message from the routing
855  * socket indicating that a redirect has occurred, a routing lookup
856  * has failed, or that a protocol has detected timeouts to a particular
857  * destination.
858  */
859 void
860 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
861 {
862 	struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
863 	struct rt_msghdr *rtm;
864 	struct mbuf *m;
865 
866 	if (route_cb.any_count == 0)
867 		return;
868 	m = rt_msg_mbuf(type, rtinfo);
869 	if (m == NULL)
870 		return;
871 	rtm = mtod(m, struct rt_msghdr *);
872 	rtm->rtm_flags = RTF_DONE | flags;
873 	rtm->rtm_errno = error;
874 	rtm->rtm_addrs = rtinfo->rti_addrs;
875 	rts_input(m, familyof(dst));
876 }
877 
878 void
879 rt_dstmsg(int type, struct sockaddr *dst, int error)
880 {
881 	struct rt_msghdr *rtm;
882 	struct rt_addrinfo addrs;
883 	struct mbuf *m;
884 
885 	if (route_cb.any_count == 0)
886 		return;
887 	bzero(&addrs, sizeof(struct rt_addrinfo));
888 	addrs.rti_info[RTAX_DST] = dst;
889 	m = rt_msg_mbuf(type, &addrs);
890 	if (m == NULL)
891 		return;
892 	rtm = mtod(m, struct rt_msghdr *);
893 	rtm->rtm_flags = RTF_DONE;
894 	rtm->rtm_errno = error;
895 	rtm->rtm_addrs = addrs.rti_addrs;
896 	rts_input(m, familyof(dst));
897 }
898 
899 /*
900  * This routine is called to generate a message from the routing
901  * socket indicating that the status of a network interface has changed.
902  */
903 void
904 rt_ifmsg(struct ifnet *ifp)
905 {
906 	struct if_msghdr *ifm;
907 	struct mbuf *m;
908 	struct rt_addrinfo rtinfo;
909 
910 	if (route_cb.any_count == 0)
911 		return;
912 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
913 	m = rt_msg_mbuf(RTM_IFINFO, &rtinfo);
914 	if (m == NULL)
915 		return;
916 	ifm = mtod(m, struct if_msghdr *);
917 	ifm->ifm_index = ifp->if_index;
918 	ifm->ifm_flags = ifp->if_flags;
919 	ifm->ifm_data = ifp->if_data;
920 	ifm->ifm_addrs = 0;
921 	rts_input(m, 0);
922 }
923 
924 static void
925 rt_ifamsg(int cmd, struct ifaddr *ifa)
926 {
927 	struct ifa_msghdr *ifam;
928 	struct rt_addrinfo rtinfo;
929 	struct mbuf *m;
930 	struct ifnet *ifp = ifa->ifa_ifp;
931 
932 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
933 	rtinfo.rti_ifaaddr = ifa->ifa_addr;
934 	rtinfo.rti_ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
935 	rtinfo.rti_netmask = ifa->ifa_netmask;
936 	rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
937 
938 	m = rt_msg_mbuf(cmd, &rtinfo);
939 	if (m == NULL)
940 		return;
941 
942 	ifam = mtod(m, struct ifa_msghdr *);
943 	ifam->ifam_index = ifp->if_index;
944 	ifam->ifam_metric = ifa->ifa_metric;
945 	ifam->ifam_flags = ifa->ifa_flags;
946 	ifam->ifam_addrs = rtinfo.rti_addrs;
947 
948 	rts_input(m, familyof(ifa->ifa_addr));
949 }
950 
951 void
952 rt_rtmsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int error)
953 {
954 	struct rt_msghdr *rtm;
955 	struct rt_addrinfo rtinfo;
956 	struct mbuf *m;
957 	struct sockaddr *dst;
958 
959 	if (rt == NULL)
960 		return;
961 
962 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
963 	rtinfo.rti_dst = dst = rt_key(rt);
964 	rtinfo.rti_gateway = rt->rt_gateway;
965 	rtinfo.rti_netmask = rt_mask(rt);
966 	if (ifp != NULL)
967 		rtinfo.rti_ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
968 	rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
969 
970 	m = rt_msg_mbuf(cmd, &rtinfo);
971 	if (m == NULL)
972 		return;
973 
974 	rtm = mtod(m, struct rt_msghdr *);
975 	if (ifp != NULL)
976 		rtm->rtm_index = ifp->if_index;
977 	rtm->rtm_flags |= rt->rt_flags;
978 	rtm->rtm_errno = error;
979 	rtm->rtm_addrs = rtinfo.rti_addrs;
980 
981 	rts_input(m, familyof(dst));
982 }
983 
984 /*
985  * This is called to generate messages from the routing socket
986  * indicating a network interface has had addresses associated with it.
987  * if we ever reverse the logic and replace messages TO the routing
988  * socket indicate a request to configure interfaces, then it will
989  * be unnecessary as the routing socket will automatically generate
990  * copies of it.
991  */
992 void
993 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
994 {
995 #ifdef SCTP
996 	/*
997 	 * notify the SCTP stack
998 	 * this will only get called when an address is added/deleted
999 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1000 	 */
1001 	if (cmd == RTM_ADD)
1002 		sctp_add_ip_address(ifa);
1003 	else if (cmd == RTM_DELETE)
1004 		sctp_delete_ip_address(ifa);
1005 #endif /* SCTP */
1006 
1007 	if (route_cb.any_count == 0)
1008 		return;
1009 
1010 	if (cmd == RTM_ADD) {
1011 		rt_ifamsg(RTM_NEWADDR, ifa);
1012 		rt_rtmsg(RTM_ADD, rt, ifa->ifa_ifp, error);
1013 	} else {
1014 		KASSERT((cmd == RTM_DELETE), ("unknown cmd %d", cmd));
1015 		rt_rtmsg(RTM_DELETE, rt, ifa->ifa_ifp, error);
1016 		rt_ifamsg(RTM_DELADDR, ifa);
1017 	}
1018 }
1019 
1020 /*
1021  * This is the analogue to the rt_newaddrmsg which performs the same
1022  * function but for multicast group memberhips.  This is easier since
1023  * there is no route state to worry about.
1024  */
1025 void
1026 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1027 {
1028 	struct rt_addrinfo rtinfo;
1029 	struct mbuf *m = NULL;
1030 	struct ifnet *ifp = ifma->ifma_ifp;
1031 	struct ifma_msghdr *ifmam;
1032 
1033 	if (route_cb.any_count == 0)
1034 		return;
1035 
1036 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1037 	rtinfo.rti_ifaaddr = ifma->ifma_addr;
1038 	if (ifp != NULL && !TAILQ_EMPTY(&ifp->if_addrhead))
1039 		rtinfo.rti_ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
1040 	/*
1041 	 * If a link-layer address is present, present it as a ``gateway''
1042 	 * (similarly to how ARP entries, e.g., are presented).
1043 	 */
1044 	rtinfo.rti_gateway = ifma->ifma_lladdr;
1045 
1046 	m = rt_msg_mbuf(cmd, &rtinfo);
1047 	if (m == NULL)
1048 		return;
1049 
1050 	ifmam = mtod(m, struct ifma_msghdr *);
1051 	ifmam->ifmam_index = ifp->if_index;
1052 	ifmam->ifmam_addrs = rtinfo.rti_addrs;
1053 
1054 	rts_input(m, familyof(ifma->ifma_addr));
1055 }
1056 
1057 /*
1058  * This is called to generate routing socket messages indicating
1059  * network interface arrival and departure.
1060  */
1061 void
1062 rt_ifannouncemsg(struct ifnet *ifp, int what)
1063 {
1064 	struct rt_addrinfo addrinfo;
1065 	struct mbuf *m;
1066 	struct if_announcemsghdr *ifan;
1067 
1068 	if (route_cb.any_count == 0)
1069 		return;
1070 
1071 	bzero(&addrinfo, sizeof addrinfo);
1072 	m = rt_msg_mbuf(RTM_IFANNOUNCE, &addrinfo);
1073 	if (m == NULL)
1074 		return;
1075 
1076 	ifan = mtod(m, struct if_announcemsghdr *);
1077 	ifan->ifan_index = ifp->if_index;
1078 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof ifan->ifan_name);
1079 	ifan->ifan_what = what;
1080 
1081 	rts_input(m, 0);
1082 }
1083 
1084 static int
1085 resizewalkarg(struct walkarg *w, int len)
1086 {
1087 	void *newptr;
1088 
1089 	newptr = malloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
1090 	if (newptr == NULL)
1091 		return (ENOMEM);
1092 	if (w->w_tmem != NULL)
1093 		free(w->w_tmem, M_RTABLE);
1094 	w->w_tmem = newptr;
1095 	w->w_tmemsize = len;
1096 	return (0);
1097 }
1098 
1099 /*
1100  * This is used in dumping the kernel table via sysctl().
1101  */
1102 int
1103 sysctl_dumpentry(struct radix_node *rn, void *vw)
1104 {
1105 	struct walkarg *w = vw;
1106 	struct rtentry *rt = (struct rtentry *)rn;
1107 	struct rt_addrinfo rtinfo;
1108 	int error, msglen;
1109 
1110 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1111 		return 0;
1112 
1113 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1114 	rtinfo.rti_dst = rt_key(rt);
1115 	rtinfo.rti_gateway = rt->rt_gateway;
1116 	rtinfo.rti_netmask = rt_mask(rt);
1117 	rtinfo.rti_genmask = rt->rt_genmask;
1118 	if (rt->rt_ifp != NULL) {
1119 		rtinfo.rti_ifpaddr =
1120 		    TAILQ_FIRST(&rt->rt_ifp->if_addrhead)->ifa_addr;
1121 		rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
1122 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1123 			rtinfo.rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
1124 	}
1125 	msglen = rt_msgsize(RTM_GET, &rtinfo);
1126 	if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1127 		return (ENOMEM);
1128 	rt_msg_buffer(RTM_GET, &rtinfo, w->w_tmem, msglen);
1129 	if (w->w_req != NULL) {
1130 		struct rt_msghdr *rtm = w->w_tmem;
1131 
1132 		rtm->rtm_flags = rt->rt_flags;
1133 		rtm->rtm_use = rt->rt_use;
1134 		rtm->rtm_rmx = rt->rt_rmx;
1135 		rtm->rtm_index = rt->rt_ifp->if_index;
1136 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1137 		rtm->rtm_addrs = rtinfo.rti_addrs;
1138 		error = SYSCTL_OUT(w->w_req, rtm, msglen);
1139 		return (error);
1140 	}
1141 	return (0);
1142 }
1143 
1144 static int
1145 sysctl_iflist(int af, struct walkarg *w)
1146 {
1147 	struct ifnet *ifp;
1148 	struct ifaddr *ifa;
1149 	struct rt_addrinfo rtinfo;
1150 	int msglen, error;
1151 
1152 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1153 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1154 		if (w->w_arg && w->w_arg != ifp->if_index)
1155 			continue;
1156 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1157 		rtinfo.rti_ifpaddr = ifa->ifa_addr;
1158 		msglen = rt_msgsize(RTM_IFINFO, &rtinfo);
1159 		if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1160 			return (ENOMEM);
1161 		rt_msg_buffer(RTM_IFINFO, &rtinfo, w->w_tmem, msglen);
1162 		rtinfo.rti_ifpaddr = NULL;
1163 		if (w->w_req != NULL && w->w_tmem != NULL) {
1164 			struct if_msghdr *ifm = w->w_tmem;
1165 
1166 			ifm->ifm_index = ifp->if_index;
1167 			ifm->ifm_flags = ifp->if_flags;
1168 			ifm->ifm_data = ifp->if_data;
1169 			ifm->ifm_addrs = rtinfo.rti_addrs;
1170 			error = SYSCTL_OUT(w->w_req, ifm, msglen);
1171 			if (error)
1172 				return (error);
1173 		}
1174 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1175 			if (af && af != ifa->ifa_addr->sa_family)
1176 				continue;
1177 			if (curproc->p_ucred->cr_prison &&
1178 			    prison_if(curthread, ifa->ifa_addr))
1179 				continue;
1180 			rtinfo.rti_ifaaddr = ifa->ifa_addr;
1181 			rtinfo.rti_netmask = ifa->ifa_netmask;
1182 			rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
1183 			msglen = rt_msgsize(RTM_NEWADDR, &rtinfo);
1184 			if (w->w_tmemsize < msglen &&
1185 			    resizewalkarg(w, msglen) != 0)
1186 				return (ENOMEM);
1187 			rt_msg_buffer(RTM_NEWADDR, &rtinfo, w->w_tmem, msglen);
1188 			if (w->w_req != NULL) {
1189 				struct ifa_msghdr *ifam = w->w_tmem;
1190 
1191 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1192 				ifam->ifam_flags = ifa->ifa_flags;
1193 				ifam->ifam_metric = ifa->ifa_metric;
1194 				ifam->ifam_addrs = rtinfo.rti_addrs;
1195 				error = SYSCTL_OUT(w->w_req, w->w_tmem, msglen);
1196 				if (error)
1197 					return (error);
1198 			}
1199 		}
1200 		rtinfo.rti_netmask = NULL;
1201 		rtinfo.rti_ifaaddr = NULL;
1202 		rtinfo.rti_bcastaddr = NULL;
1203 	}
1204 	return (0);
1205 }
1206 
1207 static int
1208 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1209 {
1210 	int	*name = (int *)arg1;
1211 	u_int	namelen = arg2;
1212 	struct radix_node_head *rnh;
1213 	int	i, error = EINVAL;
1214 	int	origcpu;
1215 	u_char  af;
1216 	struct	walkarg w;
1217 
1218 	name ++;
1219 	namelen--;
1220 	if (req->newptr)
1221 		return (EPERM);
1222 	if (namelen != 3 && namelen != 4)
1223 		return (EINVAL);
1224 	af = name[0];
1225 	bzero(&w, sizeof w);
1226 	w.w_op = name[1];
1227 	w.w_arg = name[2];
1228 	w.w_req = req;
1229 
1230 	/*
1231 	 * Optional third argument specifies cpu, used primarily for
1232 	 * debugging the route table.
1233 	 */
1234 	if (namelen == 4) {
1235 		if (name[3] < 0 || name[3] >= ncpus)
1236 			return (EINVAL);
1237 		origcpu = mycpuid;
1238 		lwkt_migratecpu(name[3]);
1239 	} else {
1240 		origcpu = -1;
1241 	}
1242 	crit_enter();
1243 	switch (w.w_op) {
1244 	case NET_RT_DUMP:
1245 	case NET_RT_FLAGS:
1246 		for (i = 1; i <= AF_MAX; i++)
1247 			if ((rnh = rt_tables[mycpuid][i]) &&
1248 			    (af == 0 || af == i) &&
1249 			    (error = rnh->rnh_walktree(rnh,
1250 						       sysctl_dumpentry, &w)))
1251 				break;
1252 		break;
1253 
1254 	case NET_RT_IFLIST:
1255 		error = sysctl_iflist(af, &w);
1256 	}
1257 	crit_exit();
1258 	if (w.w_tmem != NULL)
1259 		free(w.w_tmem, M_RTABLE);
1260 	if (origcpu >= 0)
1261 		lwkt_migratecpu(origcpu);
1262 	return (error);
1263 }
1264 
1265 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1266 
1267 /*
1268  * Definitions of protocols supported in the ROUTE domain.
1269  */
1270 
1271 extern struct domain routedomain;		/* or at least forward */
1272 
1273 static struct protosw routesw[] = {
1274 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
1275   0,		route_output,	raw_ctlinput,	0,
1276   cpu0_soport,
1277   raw_init,	0,		0,		0,
1278   &route_usrreqs
1279 }
1280 };
1281 
1282 static struct domain routedomain = {
1283 	PF_ROUTE, "route", NULL, NULL, NULL,
1284 	routesw, &routesw[(sizeof routesw)/(sizeof routesw[0])],
1285 };
1286 
1287 DOMAIN_SET(route);
1288