xref: /dragonfly/sys/net/rtsock.c (revision ab709bfb)
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 1.45 2008/10/27 02:56:30 sephe 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/priv.h>
78 #include <sys/malloc.h>
79 #include <sys/mbuf.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/domain.h>
84 #include <sys/thread2.h>
85 
86 #include <net/if.h>
87 #include <net/route.h>
88 #include <net/raw_cb.h>
89 #include <net/netmsg2.h>
90 
91 #ifdef SCTP
92 extern void sctp_add_ip_address(struct ifaddr *ifa);
93 extern void sctp_delete_ip_address(struct ifaddr *ifa);
94 #endif /* SCTP */
95 
96 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
97 
98 static struct route_cb {
99 	int	ip_count;
100 	int	ip6_count;
101 	int	ipx_count;
102 	int	ns_count;
103 	int	any_count;
104 } route_cb;
105 
106 static const struct sockaddr route_src = { 2, PF_ROUTE, };
107 
108 struct walkarg {
109 	int	w_tmemsize;
110 	int	w_op, w_arg;
111 	void	*w_tmem;
112 	struct sysctl_req *w_req;
113 };
114 
115 static struct mbuf *
116 		rt_msg_mbuf (int, struct rt_addrinfo *);
117 static void	rt_msg_buffer (int, struct rt_addrinfo *, void *buf, int len);
118 static int	rt_msgsize (int type, struct rt_addrinfo *rtinfo);
119 static int	rt_xaddrs (char *, char *, struct rt_addrinfo *);
120 static int	sysctl_dumpentry (struct radix_node *rn, void *vw);
121 static int	sysctl_iflist (int af, struct walkarg *w);
122 static int	route_output(struct mbuf *, struct socket *, ...);
123 static void	rt_setmetrics (u_long, struct rt_metrics *,
124 			       struct rt_metrics *);
125 
126 /*
127  * It really doesn't make any sense at all for this code to share much
128  * with raw_usrreq.c, since its functionality is so restricted.  XXX
129  */
130 static int
131 rts_abort(struct socket *so)
132 {
133 	int error;
134 
135 	crit_enter();
136 	error = raw_usrreqs.pru_abort(so);
137 	crit_exit();
138 	return error;
139 }
140 
141 /* pru_accept is EOPNOTSUPP */
142 
143 static int
144 rts_attach(struct socket *so, int proto, struct pru_attach_info *ai)
145 {
146 	struct rawcb *rp;
147 	int error;
148 
149 	if (sotorawcb(so) != NULL)
150 		return EISCONN;	/* XXX panic? */
151 
152 	rp = kmalloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
153 
154 	/*
155 	 * The critical section is necessary to block protocols from sending
156 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
157 	 * this PCB is extant but incompletely initialized.
158 	 * Probably we should try to do more of this work beforehand and
159 	 * eliminate the critical section.
160 	 */
161 	crit_enter();
162 	so->so_pcb = rp;
163 	error = raw_attach(so, proto, ai->sb_rlimit);
164 	rp = sotorawcb(so);
165 	if (error) {
166 		crit_exit();
167 		kfree(rp, M_PCB);
168 		return error;
169 	}
170 	switch(rp->rcb_proto.sp_protocol) {
171 	case AF_INET:
172 		route_cb.ip_count++;
173 		break;
174 	case AF_INET6:
175 		route_cb.ip6_count++;
176 		break;
177 	case AF_IPX:
178 		route_cb.ipx_count++;
179 		break;
180 	case AF_NS:
181 		route_cb.ns_count++;
182 		break;
183 	}
184 	rp->rcb_faddr = &route_src;
185 	route_cb.any_count++;
186 	soisconnected(so);
187 	so->so_options |= SO_USELOOPBACK;
188 	crit_exit();
189 	return 0;
190 }
191 
192 static int
193 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
194 {
195 	int error;
196 
197 	crit_enter();
198 	error = raw_usrreqs.pru_bind(so, nam, td); /* xxx just EINVAL */
199 	crit_exit();
200 	return error;
201 }
202 
203 static int
204 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
205 {
206 	int error;
207 
208 	crit_enter();
209 	error = raw_usrreqs.pru_connect(so, nam, td); /* XXX just EINVAL */
210 	crit_exit();
211 	return error;
212 }
213 
214 /* pru_connect2 is EOPNOTSUPP */
215 /* pru_control is EOPNOTSUPP */
216 
217 static int
218 rts_detach(struct socket *so)
219 {
220 	struct rawcb *rp = sotorawcb(so);
221 	int error;
222 
223 	crit_enter();
224 	if (rp != NULL) {
225 		switch(rp->rcb_proto.sp_protocol) {
226 		case AF_INET:
227 			route_cb.ip_count--;
228 			break;
229 		case AF_INET6:
230 			route_cb.ip6_count--;
231 			break;
232 		case AF_IPX:
233 			route_cb.ipx_count--;
234 			break;
235 		case AF_NS:
236 			route_cb.ns_count--;
237 			break;
238 		}
239 		route_cb.any_count--;
240 	}
241 	error = raw_usrreqs.pru_detach(so);
242 	crit_exit();
243 	return error;
244 }
245 
246 static int
247 rts_disconnect(struct socket *so)
248 {
249 	int error;
250 
251 	crit_enter();
252 	error = raw_usrreqs.pru_disconnect(so);
253 	crit_exit();
254 	return error;
255 }
256 
257 /* pru_listen is EOPNOTSUPP */
258 
259 static int
260 rts_peeraddr(struct socket *so, struct sockaddr **nam)
261 {
262 	int error;
263 
264 	crit_enter();
265 	error = raw_usrreqs.pru_peeraddr(so, nam);
266 	crit_exit();
267 	return error;
268 }
269 
270 /* pru_rcvd is EOPNOTSUPP */
271 /* pru_rcvoob is EOPNOTSUPP */
272 
273 static int
274 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
275 	 struct mbuf *control, struct thread *td)
276 {
277 	int error;
278 
279 	crit_enter();
280 	error = raw_usrreqs.pru_send(so, flags, m, nam, control, td);
281 	crit_exit();
282 	return error;
283 }
284 
285 /* pru_sense is null */
286 
287 static int
288 rts_shutdown(struct socket *so)
289 {
290 	int error;
291 
292 	crit_enter();
293 	error = raw_usrreqs.pru_shutdown(so);
294 	crit_exit();
295 	return error;
296 }
297 
298 static int
299 rts_sockaddr(struct socket *so, struct sockaddr **nam)
300 {
301 	int error;
302 
303 	crit_enter();
304 	error = raw_usrreqs.pru_sockaddr(so, nam);
305 	crit_exit();
306 	return error;
307 }
308 
309 static struct pr_usrreqs route_usrreqs = {
310 	.pru_abort = rts_abort,
311 	.pru_accept = pru_accept_notsupp,
312 	.pru_attach = rts_attach,
313 	.pru_bind = rts_bind,
314 	.pru_connect = rts_connect,
315 	.pru_connect2 = pru_connect2_notsupp,
316 	.pru_control = pru_control_notsupp,
317 	.pru_detach = rts_detach,
318 	.pru_disconnect = rts_disconnect,
319 	.pru_listen = pru_listen_notsupp,
320 	.pru_peeraddr = rts_peeraddr,
321 	.pru_rcvd = pru_rcvd_notsupp,
322 	.pru_rcvoob = pru_rcvoob_notsupp,
323 	.pru_send = rts_send,
324 	.pru_sense = pru_sense_null,
325 	.pru_shutdown = rts_shutdown,
326 	.pru_sockaddr = rts_sockaddr,
327 	.pru_sosend = sosend,
328 	.pru_soreceive = soreceive,
329 	.pru_sopoll = sopoll
330 };
331 
332 static __inline sa_family_t
333 familyof(struct sockaddr *sa)
334 {
335 	return (sa != NULL ? sa->sa_family : 0);
336 }
337 
338 /*
339  * Routing socket input function.  The packet must be serialized onto cpu 0.
340  * We use the cpu0_soport() netisr processing loop to handle it.
341  *
342  * This looks messy but it means that anyone, including interrupt code,
343  * can send a message to the routing socket.
344  */
345 static void
346 rts_input_handler(struct netmsg *msg)
347 {
348 	static const struct sockaddr route_dst = { 2, PF_ROUTE, };
349 	struct sockproto route_proto;
350 	struct netmsg_packet *pmsg;
351 	struct mbuf *m;
352 	sa_family_t family;
353 	struct rawcb *skip;
354 
355 	pmsg = (void *)msg;
356 	family = pmsg->nm_netmsg.nm_lmsg.u.ms_result;
357 	route_proto.sp_family = PF_ROUTE;
358 	route_proto.sp_protocol = family;
359 
360 	m = pmsg->nm_packet;
361 	M_ASSERTPKTHDR(m);
362 
363 	skip = m->m_pkthdr.header;
364 	m->m_pkthdr.header = NULL;
365 
366 	raw_input(m, &route_proto, &route_src, &route_dst, skip);
367 }
368 
369 static void
370 rts_input_skip(struct mbuf *m, sa_family_t family, struct rawcb *skip)
371 {
372 	struct netmsg_packet *pmsg;
373 	lwkt_port_t port;
374 
375 	M_ASSERTPKTHDR(m);
376 
377 	port = cpu0_soport(NULL, NULL, NULL, 0);
378 	pmsg = &m->m_hdr.mh_netmsg;
379 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport,
380 		    0, rts_input_handler);
381 	pmsg->nm_packet = m;
382 	pmsg->nm_netmsg.nm_lmsg.u.ms_result = family;
383 	m->m_pkthdr.header = skip; /* XXX steal field in pkthdr */
384 	lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
385 }
386 
387 static __inline void
388 rts_input(struct mbuf *m, sa_family_t family)
389 {
390 	rts_input_skip(m, family, NULL);
391 }
392 
393 static void *
394 reallocbuf(void *ptr, size_t len, size_t olen)
395 {
396 	void *newptr;
397 
398 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
399 	if (newptr == NULL)
400 		return NULL;
401 	bcopy(ptr, newptr, olen);
402 	kfree(ptr, M_RTABLE);
403 	return (newptr);
404 }
405 
406 /*
407  * Internal helper routine for route_output().
408  */
409 static int
410 fillrtmsg(struct rt_msghdr **prtm, struct rtentry *rt,
411 	  struct rt_addrinfo *rtinfo)
412 {
413 	int msglen;
414 	struct rt_msghdr *rtm = *prtm;
415 
416 	/* Fill in rt_addrinfo for call to rt_msg_buffer(). */
417 	rtinfo->rti_dst = rt_key(rt);
418 	rtinfo->rti_gateway = rt->rt_gateway;
419 	rtinfo->rti_netmask = rt_mask(rt);		/* might be NULL */
420 	rtinfo->rti_genmask = rt->rt_genmask;		/* might be NULL */
421 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
422 		if (rt->rt_ifp != NULL) {
423 			rtinfo->rti_ifpaddr =
424 			    TAILQ_FIRST(&rt->rt_ifp->if_addrheads[mycpuid])
425 			    ->ifa->ifa_addr;
426 			rtinfo->rti_ifaaddr = rt->rt_ifa->ifa_addr;
427 			if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
428 				rtinfo->rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
429 			rtm->rtm_index = rt->rt_ifp->if_index;
430 		} else {
431 			rtinfo->rti_ifpaddr = NULL;
432 			rtinfo->rti_ifaaddr = NULL;
433 		}
434 	} else if (rt->rt_ifp != NULL) {
435 		rtm->rtm_index = rt->rt_ifp->if_index;
436 	}
437 
438 	msglen = rt_msgsize(rtm->rtm_type, rtinfo);
439 	if (rtm->rtm_msglen < msglen) {
440 		rtm = reallocbuf(rtm, msglen, rtm->rtm_msglen);
441 		if (rtm == NULL)
442 			return (ENOBUFS);
443 		*prtm = rtm;
444 	}
445 	rt_msg_buffer(rtm->rtm_type, rtinfo, rtm, msglen);
446 
447 	rtm->rtm_flags = rt->rt_flags;
448 	rtm->rtm_rmx = rt->rt_rmx;
449 	rtm->rtm_addrs = rtinfo->rti_addrs;
450 
451 	return (0);
452 }
453 
454 static void route_output_add_callback(int, int, struct rt_addrinfo *,
455 					struct rtentry *, void *);
456 static void route_output_delete_callback(int, int, struct rt_addrinfo *,
457 					struct rtentry *, void *);
458 static int route_output_get_callback(int, struct rt_addrinfo *,
459 				     struct rtentry *, void *, int);
460 static int route_output_change_callback(int, struct rt_addrinfo *,
461 					struct rtentry *, void *, int);
462 static int route_output_lock_callback(int, struct rt_addrinfo *,
463 				      struct rtentry *, void *, int);
464 
465 /*ARGSUSED*/
466 static int
467 route_output(struct mbuf *m, struct socket *so, ...)
468 {
469 	struct rt_msghdr *rtm = NULL;
470 	struct rawcb *rp = NULL;
471 	struct pr_output_info *oi;
472 	struct rt_addrinfo rtinfo;
473 	int len, error = 0;
474 	__va_list ap;
475 
476 	M_ASSERTPKTHDR(m);
477 
478 	__va_start(ap, so);
479 	oi = __va_arg(ap, struct pr_output_info *);
480 	__va_end(ap);
481 
482 #define gotoerr(e) { error = e; goto flush;}
483 
484 	if (m == NULL ||
485 	    (m->m_len < sizeof(long) &&
486 	     (m = m_pullup(m, sizeof(long))) == NULL))
487 		return (ENOBUFS);
488 	len = m->m_pkthdr.len;
489 	if (len < sizeof(struct rt_msghdr) ||
490 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
491 		rtinfo.rti_dst = NULL;
492 		gotoerr(EINVAL);
493 	}
494 	rtm = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
495 	if (rtm == NULL) {
496 		rtinfo.rti_dst = NULL;
497 		gotoerr(ENOBUFS);
498 	}
499 	m_copydata(m, 0, len, (caddr_t)rtm);
500 	if (rtm->rtm_version != RTM_VERSION) {
501 		rtinfo.rti_dst = NULL;
502 		gotoerr(EPROTONOSUPPORT);
503 	}
504 	rtm->rtm_pid = oi->p_pid;
505 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
506 	rtinfo.rti_addrs = rtm->rtm_addrs;
507 	if (rt_xaddrs((char *)(rtm + 1), (char *)rtm + len, &rtinfo) != 0) {
508 		rtinfo.rti_dst = NULL;
509 		gotoerr(EINVAL);
510 	}
511 	rtinfo.rti_flags = rtm->rtm_flags;
512 	if (rtinfo.rti_dst == NULL || rtinfo.rti_dst->sa_family >= AF_MAX ||
513 	    (rtinfo.rti_gateway && rtinfo.rti_gateway->sa_family >= AF_MAX))
514 		gotoerr(EINVAL);
515 
516 	if (rtinfo.rti_genmask != NULL) {
517 		error = rtmask_add_global(rtinfo.rti_genmask);
518 		if (error)
519 			goto flush;
520 	}
521 
522 	/*
523 	 * Verify that the caller has the appropriate privilege; RTM_GET
524 	 * is the only operation the non-superuser is allowed.
525 	 */
526 	if (rtm->rtm_type != RTM_GET &&
527 	    priv_check_cred(so->so_cred, PRIV_ROOT, 0) != 0)
528 		gotoerr(EPERM);
529 
530 	switch (rtm->rtm_type) {
531 	case RTM_ADD:
532 		if (rtinfo.rti_gateway == NULL) {
533 			error = EINVAL;
534 		} else {
535 			error = rtrequest1_global(RTM_ADD, &rtinfo,
536 					  route_output_add_callback, rtm);
537 		}
538 		break;
539 	case RTM_DELETE:
540 		/*
541 		 * note: &rtm passed as argument so 'rtm' can be replaced.
542 		 */
543 		error = rtrequest1_global(RTM_DELETE, &rtinfo,
544 					  route_output_delete_callback, &rtm);
545 		break;
546 	case RTM_GET:
547 		/*
548 		 * note: &rtm passed as argument so 'rtm' can be replaced.
549 		 */
550 		error = rtsearch_global(RTM_GET, &rtinfo,
551 					route_output_get_callback, &rtm,
552 					RTS_NOEXACTMATCH);
553 		break;
554 	case RTM_CHANGE:
555 		error = rtsearch_global(RTM_CHANGE, &rtinfo,
556 					route_output_change_callback, rtm,
557 					RTS_EXACTMATCH);
558 		break;
559 	case RTM_LOCK:
560 		error = rtsearch_global(RTM_LOCK, &rtinfo,
561 					route_output_lock_callback, rtm,
562 					RTS_EXACTMATCH);
563 		break;
564 	default:
565 		error = EOPNOTSUPP;
566 		break;
567 	}
568 
569 flush:
570 	if (rtm != NULL) {
571 		if (error != 0)
572 			rtm->rtm_errno = error;
573 		else
574 			rtm->rtm_flags |= RTF_DONE;
575 	}
576 
577 	/*
578 	 * Check to see if we don't want our own messages.
579 	 */
580 	if (!(so->so_options & SO_USELOOPBACK)) {
581 		if (route_cb.any_count <= 1) {
582 			if (rtm != NULL)
583 				kfree(rtm, M_RTABLE);
584 			m_freem(m);
585 			return (error);
586 		}
587 		/* There is another listener, so construct message */
588 		rp = sotorawcb(so);
589 	}
590 	if (rtm != NULL) {
591 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
592 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
593 			m_freem(m);
594 			m = NULL;
595 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
596 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
597 		kfree(rtm, M_RTABLE);
598 	}
599 	if (m != NULL)
600 		rts_input_skip(m, familyof(rtinfo.rti_dst), rp);
601 	return (error);
602 }
603 
604 static void
605 route_output_add_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
606 			  struct rtentry *rt, void *arg)
607 {
608 	struct rt_msghdr *rtm = arg;
609 
610 	if (error == 0 && rt != NULL) {
611 		rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
612 		    &rt->rt_rmx);
613 		rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
614 		rt->rt_rmx.rmx_locks |=
615 		    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
616 		if (rtinfo->rti_genmask != NULL) {
617 			rt->rt_genmask = rtmask_purelookup(rtinfo->rti_genmask);
618 			if (rt->rt_genmask == NULL) {
619 				/*
620 				 * This should not happen, since we
621 				 * have already installed genmask
622 				 * on each CPU before we reach here.
623 				 */
624 				panic("genmask is gone!?");
625 			}
626 		} else {
627 			rt->rt_genmask = NULL;
628 		}
629 		rtm->rtm_index = rt->rt_ifp->if_index;
630 	}
631 }
632 
633 static void
634 route_output_delete_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
635 			  struct rtentry *rt, void *arg)
636 {
637 	struct rt_msghdr **rtm = arg;
638 
639 	if (error == 0 && rt) {
640 		++rt->rt_refcnt;
641 		if (fillrtmsg(rtm, rt, rtinfo) != 0) {
642 			error = ENOBUFS;
643 			/* XXX no way to return the error */
644 		}
645 		--rt->rt_refcnt;
646 	}
647 }
648 
649 static int
650 route_output_get_callback(int cmd, struct rt_addrinfo *rtinfo,
651 			  struct rtentry *rt, void *arg, int found_cnt)
652 {
653 	struct rt_msghdr **rtm = arg;
654 	int error, found = 0;
655 
656 	if (((rtinfo->rti_flags ^ rt->rt_flags) & RTF_HOST) == 0)
657 		found = 1;
658 
659 	error = fillrtmsg(rtm, rt, rtinfo);
660 	if (!error && found) {
661 		/* Got the exact match, we could return now! */
662 		error = EJUSTRETURN;
663 	}
664 	return error;
665 }
666 
667 static int
668 route_output_change_callback(int cmd, struct rt_addrinfo *rtinfo,
669 			     struct rtentry *rt, void *arg, int found_cnt)
670 {
671 	struct rt_msghdr *rtm = arg;
672 	struct ifaddr *ifa;
673 	int error = 0;
674 
675 	/*
676 	 * new gateway could require new ifaddr, ifp;
677 	 * flags may also be different; ifp may be specified
678 	 * by ll sockaddr when protocol address is ambiguous
679 	 */
680 	if (((rt->rt_flags & RTF_GATEWAY) && rtinfo->rti_gateway != NULL) ||
681 	    rtinfo->rti_ifpaddr != NULL ||
682 	    (rtinfo->rti_ifaaddr != NULL &&
683 	     !sa_equal(rtinfo->rti_ifaaddr, rt->rt_ifa->ifa_addr))) {
684 		error = rt_getifa(rtinfo);
685 		if (error != 0)
686 			goto done;
687 	}
688 	if (rtinfo->rti_gateway != NULL) {
689 		/*
690 		 * We only need to generate rtmsg upon the
691 		 * first route to be changed.
692 		 */
693 		error = rt_setgate(rt, rt_key(rt), rtinfo->rti_gateway,
694 			found_cnt == 1 ? RTL_REPORTMSG : RTL_DONTREPORT);
695 		if (error != 0)
696 			goto done;
697 	}
698 	if ((ifa = rtinfo->rti_ifa) != NULL) {
699 		struct ifaddr *oifa = rt->rt_ifa;
700 
701 		if (oifa != ifa) {
702 			if (oifa && oifa->ifa_rtrequest)
703 				oifa->ifa_rtrequest(RTM_DELETE, rt, rtinfo);
704 			IFAFREE(rt->rt_ifa);
705 			IFAREF(ifa);
706 			rt->rt_ifa = ifa;
707 			rt->rt_ifp = rtinfo->rti_ifp;
708 		}
709 	}
710 	rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, &rt->rt_rmx);
711 	if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
712 		rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, rtinfo);
713 	if (rtinfo->rti_genmask != NULL) {
714 		rt->rt_genmask = rtmask_purelookup(rtinfo->rti_genmask);
715 		if (rt->rt_genmask == NULL) {
716 			/*
717 			 * This should not happen, since we
718 			 * have already installed genmask
719 			 * on each CPU before we reach here.
720 			 */
721 			panic("genmask is gone!?\n");
722 		}
723 	}
724 	rtm->rtm_index = rt->rt_ifp->if_index;
725 done:
726 	return error;
727 }
728 
729 static int
730 route_output_lock_callback(int cmd, struct rt_addrinfo *rtinfo,
731 			   struct rtentry *rt, void *arg,
732 			   int found_cnt __unused)
733 {
734 	struct rt_msghdr *rtm = arg;
735 
736 	rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
737 	rt->rt_rmx.rmx_locks |=
738 		(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
739 	return 0;
740 }
741 
742 static void
743 rt_setmetrics(u_long which, struct rt_metrics *in, struct rt_metrics *out)
744 {
745 #define setmetric(flag, elt) if (which & (flag)) out->elt = in->elt;
746 	setmetric(RTV_RPIPE, rmx_recvpipe);
747 	setmetric(RTV_SPIPE, rmx_sendpipe);
748 	setmetric(RTV_SSTHRESH, rmx_ssthresh);
749 	setmetric(RTV_RTT, rmx_rtt);
750 	setmetric(RTV_RTTVAR, rmx_rttvar);
751 	setmetric(RTV_HOPCOUNT, rmx_hopcount);
752 	setmetric(RTV_MTU, rmx_mtu);
753 	setmetric(RTV_EXPIRE, rmx_expire);
754 #undef setmetric
755 }
756 
757 #define ROUNDUP(a) \
758 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
759 
760 /*
761  * Extract the addresses of the passed sockaddrs.
762  * Do a little sanity checking so as to avoid bad memory references.
763  * This data is derived straight from userland.
764  */
765 static int
766 rt_xaddrs(char *cp, char *cplim, struct rt_addrinfo *rtinfo)
767 {
768 	struct sockaddr *sa;
769 	int i;
770 
771 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
772 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
773 			continue;
774 		sa = (struct sockaddr *)cp;
775 		/*
776 		 * It won't fit.
777 		 */
778 		if ((cp + sa->sa_len) > cplim) {
779 			return (EINVAL);
780 		}
781 
782 		/*
783 		 * There are no more...  Quit now.
784 		 * If there are more bits, they are in error.
785 		 * I've seen this.  route(1) can evidently generate these.
786 		 * This causes kernel to core dump.
787 		 * For compatibility, if we see this, point to a safe address.
788 		 */
789 		if (sa->sa_len == 0) {
790 			static struct sockaddr sa_zero = {
791 				sizeof sa_zero, AF_INET,
792 			};
793 
794 			rtinfo->rti_info[i] = &sa_zero;
795 			kprintf("rtsock: received more addr bits than sockaddrs.\n");
796 			return (0); /* should be EINVAL but for compat */
797 		}
798 
799 		/* Accept the sockaddr. */
800 		rtinfo->rti_info[i] = sa;
801 		cp += ROUNDUP(sa->sa_len);
802 	}
803 	return (0);
804 }
805 
806 static int
807 rt_msghdrsize(int type)
808 {
809 	switch (type) {
810 	case RTM_DELADDR:
811 	case RTM_NEWADDR:
812 		return sizeof(struct ifa_msghdr);
813 	case RTM_DELMADDR:
814 	case RTM_NEWMADDR:
815 		return sizeof(struct ifma_msghdr);
816 	case RTM_IFINFO:
817 		return sizeof(struct if_msghdr);
818 	case RTM_IFANNOUNCE:
819 	case RTM_IEEE80211:
820 		return sizeof(struct if_announcemsghdr);
821 	default:
822 		return sizeof(struct rt_msghdr);
823 	}
824 }
825 
826 static int
827 rt_msgsize(int type, struct rt_addrinfo *rtinfo)
828 {
829 	int len, i;
830 
831 	len = rt_msghdrsize(type);
832 	for (i = 0; i < RTAX_MAX; i++) {
833 		if (rtinfo->rti_info[i] != NULL)
834 			len += ROUNDUP(rtinfo->rti_info[i]->sa_len);
835 	}
836 	len = ALIGN(len);
837 	return len;
838 }
839 
840 /*
841  * Build a routing message in a buffer.
842  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
843  * to the end of the buffer after the message header.
844  *
845  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
846  * This side-effect can be avoided if we reorder the addrs bitmask field in all
847  * the route messages to line up so we can set it here instead of back in the
848  * calling routine.
849  */
850 static void
851 rt_msg_buffer(int type, struct rt_addrinfo *rtinfo, void *buf, int msglen)
852 {
853 	struct rt_msghdr *rtm;
854 	char *cp;
855 	int dlen, i;
856 
857 	rtm = (struct rt_msghdr *) buf;
858 	rtm->rtm_version = RTM_VERSION;
859 	rtm->rtm_type = type;
860 	rtm->rtm_msglen = msglen;
861 
862 	cp = (char *)buf + rt_msghdrsize(type);
863 	rtinfo->rti_addrs = 0;
864 	for (i = 0; i < RTAX_MAX; i++) {
865 		struct sockaddr *sa;
866 
867 		if ((sa = rtinfo->rti_info[i]) == NULL)
868 			continue;
869 		rtinfo->rti_addrs |= (1 << i);
870 		dlen = ROUNDUP(sa->sa_len);
871 		bcopy(sa, cp, dlen);
872 		cp += dlen;
873 	}
874 }
875 
876 /*
877  * Build a routing message in a mbuf chain.
878  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
879  * to the end of the mbuf after the message header.
880  *
881  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
882  * This side-effect can be avoided if we reorder the addrs bitmask field in all
883  * the route messages to line up so we can set it here instead of back in the
884  * calling routine.
885  */
886 static struct mbuf *
887 rt_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
888 {
889 	struct mbuf *m;
890 	struct rt_msghdr *rtm;
891 	int hlen, len;
892 	int i;
893 
894 	hlen = rt_msghdrsize(type);
895 	KASSERT(hlen <= MCLBYTES, ("rt_msg_mbuf: hlen %d doesn't fit", hlen));
896 
897 	m = m_getl(hlen, MB_DONTWAIT, MT_DATA, M_PKTHDR, NULL);
898 	if (m == NULL)
899 		return (NULL);
900 	mbuftrackid(m, 32);
901 	m->m_pkthdr.len = m->m_len = hlen;
902 	m->m_pkthdr.rcvif = NULL;
903 	rtinfo->rti_addrs = 0;
904 	len = hlen;
905 	for (i = 0; i < RTAX_MAX; i++) {
906 		struct sockaddr *sa;
907 		int dlen;
908 
909 		if ((sa = rtinfo->rti_info[i]) == NULL)
910 			continue;
911 		rtinfo->rti_addrs |= (1 << i);
912 		dlen = ROUNDUP(sa->sa_len);
913 		m_copyback(m, len, dlen, (caddr_t)sa); /* can grow mbuf chain */
914 		len += dlen;
915 	}
916 	if (m->m_pkthdr.len != len) { /* one of the m_copyback() calls failed */
917 		m_freem(m);
918 		return (NULL);
919 	}
920 	rtm = mtod(m, struct rt_msghdr *);
921 	bzero(rtm, hlen);
922 	rtm->rtm_msglen = len;
923 	rtm->rtm_version = RTM_VERSION;
924 	rtm->rtm_type = type;
925 	return (m);
926 }
927 
928 /*
929  * This routine is called to generate a message from the routing
930  * socket indicating that a redirect has occurred, a routing lookup
931  * has failed, or that a protocol has detected timeouts to a particular
932  * destination.
933  */
934 void
935 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
936 {
937 	struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
938 	struct rt_msghdr *rtm;
939 	struct mbuf *m;
940 
941 	if (route_cb.any_count == 0)
942 		return;
943 	m = rt_msg_mbuf(type, rtinfo);
944 	if (m == NULL)
945 		return;
946 	rtm = mtod(m, struct rt_msghdr *);
947 	rtm->rtm_flags = RTF_DONE | flags;
948 	rtm->rtm_errno = error;
949 	rtm->rtm_addrs = rtinfo->rti_addrs;
950 	rts_input(m, familyof(dst));
951 }
952 
953 void
954 rt_dstmsg(int type, struct sockaddr *dst, int error)
955 {
956 	struct rt_msghdr *rtm;
957 	struct rt_addrinfo addrs;
958 	struct mbuf *m;
959 
960 	if (route_cb.any_count == 0)
961 		return;
962 	bzero(&addrs, sizeof(struct rt_addrinfo));
963 	addrs.rti_info[RTAX_DST] = dst;
964 	m = rt_msg_mbuf(type, &addrs);
965 	if (m == NULL)
966 		return;
967 	rtm = mtod(m, struct rt_msghdr *);
968 	rtm->rtm_flags = RTF_DONE;
969 	rtm->rtm_errno = error;
970 	rtm->rtm_addrs = addrs.rti_addrs;
971 	rts_input(m, familyof(dst));
972 }
973 
974 /*
975  * This routine is called to generate a message from the routing
976  * socket indicating that the status of a network interface has changed.
977  */
978 void
979 rt_ifmsg(struct ifnet *ifp)
980 {
981 	struct if_msghdr *ifm;
982 	struct mbuf *m;
983 	struct rt_addrinfo rtinfo;
984 
985 	if (route_cb.any_count == 0)
986 		return;
987 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
988 	m = rt_msg_mbuf(RTM_IFINFO, &rtinfo);
989 	if (m == NULL)
990 		return;
991 	ifm = mtod(m, struct if_msghdr *);
992 	ifm->ifm_index = ifp->if_index;
993 	ifm->ifm_flags = ifp->if_flags;
994 	ifm->ifm_data = ifp->if_data;
995 	ifm->ifm_addrs = 0;
996 	rts_input(m, 0);
997 }
998 
999 static void
1000 rt_ifamsg(int cmd, struct ifaddr *ifa)
1001 {
1002 	struct ifa_msghdr *ifam;
1003 	struct rt_addrinfo rtinfo;
1004 	struct mbuf *m;
1005 	struct ifnet *ifp = ifa->ifa_ifp;
1006 
1007 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1008 	rtinfo.rti_ifaaddr = ifa->ifa_addr;
1009 	rtinfo.rti_ifpaddr =
1010 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1011 	rtinfo.rti_netmask = ifa->ifa_netmask;
1012 	rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
1013 
1014 	m = rt_msg_mbuf(cmd, &rtinfo);
1015 	if (m == NULL)
1016 		return;
1017 
1018 	ifam = mtod(m, struct ifa_msghdr *);
1019 	ifam->ifam_index = ifp->if_index;
1020 	ifam->ifam_metric = ifa->ifa_metric;
1021 	ifam->ifam_flags = ifa->ifa_flags;
1022 	ifam->ifam_addrs = rtinfo.rti_addrs;
1023 
1024 	rts_input(m, familyof(ifa->ifa_addr));
1025 }
1026 
1027 void
1028 rt_rtmsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int error)
1029 {
1030 	struct rt_msghdr *rtm;
1031 	struct rt_addrinfo rtinfo;
1032 	struct mbuf *m;
1033 	struct sockaddr *dst;
1034 
1035 	if (rt == NULL)
1036 		return;
1037 
1038 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1039 	rtinfo.rti_dst = dst = rt_key(rt);
1040 	rtinfo.rti_gateway = rt->rt_gateway;
1041 	rtinfo.rti_netmask = rt_mask(rt);
1042 	if (ifp != NULL) {
1043 		rtinfo.rti_ifpaddr =
1044 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1045 	}
1046 	rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
1047 
1048 	m = rt_msg_mbuf(cmd, &rtinfo);
1049 	if (m == NULL)
1050 		return;
1051 
1052 	rtm = mtod(m, struct rt_msghdr *);
1053 	if (ifp != NULL)
1054 		rtm->rtm_index = ifp->if_index;
1055 	rtm->rtm_flags |= rt->rt_flags;
1056 	rtm->rtm_errno = error;
1057 	rtm->rtm_addrs = rtinfo.rti_addrs;
1058 
1059 	rts_input(m, familyof(dst));
1060 }
1061 
1062 /*
1063  * This is called to generate messages from the routing socket
1064  * indicating a network interface has had addresses associated with it.
1065  * if we ever reverse the logic and replace messages TO the routing
1066  * socket indicate a request to configure interfaces, then it will
1067  * be unnecessary as the routing socket will automatically generate
1068  * copies of it.
1069  */
1070 void
1071 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1072 {
1073 #ifdef SCTP
1074 	/*
1075 	 * notify the SCTP stack
1076 	 * this will only get called when an address is added/deleted
1077 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1078 	 */
1079 	if (cmd == RTM_ADD)
1080 		sctp_add_ip_address(ifa);
1081 	else if (cmd == RTM_DELETE)
1082 		sctp_delete_ip_address(ifa);
1083 #endif /* SCTP */
1084 
1085 	if (route_cb.any_count == 0)
1086 		return;
1087 
1088 	if (cmd == RTM_ADD) {
1089 		rt_ifamsg(RTM_NEWADDR, ifa);
1090 		rt_rtmsg(RTM_ADD, rt, ifa->ifa_ifp, error);
1091 	} else {
1092 		KASSERT((cmd == RTM_DELETE), ("unknown cmd %d", cmd));
1093 		rt_rtmsg(RTM_DELETE, rt, ifa->ifa_ifp, error);
1094 		rt_ifamsg(RTM_DELADDR, ifa);
1095 	}
1096 }
1097 
1098 /*
1099  * This is the analogue to the rt_newaddrmsg which performs the same
1100  * function but for multicast group memberhips.  This is easier since
1101  * there is no route state to worry about.
1102  */
1103 void
1104 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1105 {
1106 	struct rt_addrinfo rtinfo;
1107 	struct mbuf *m = NULL;
1108 	struct ifnet *ifp = ifma->ifma_ifp;
1109 	struct ifma_msghdr *ifmam;
1110 
1111 	if (route_cb.any_count == 0)
1112 		return;
1113 
1114 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1115 	rtinfo.rti_ifaaddr = ifma->ifma_addr;
1116 	if (ifp != NULL && !TAILQ_EMPTY(&ifp->if_addrheads[mycpuid])) {
1117 		rtinfo.rti_ifpaddr =
1118 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1119 	}
1120 	/*
1121 	 * If a link-layer address is present, present it as a ``gateway''
1122 	 * (similarly to how ARP entries, e.g., are presented).
1123 	 */
1124 	rtinfo.rti_gateway = ifma->ifma_lladdr;
1125 
1126 	m = rt_msg_mbuf(cmd, &rtinfo);
1127 	if (m == NULL)
1128 		return;
1129 
1130 	ifmam = mtod(m, struct ifma_msghdr *);
1131 	ifmam->ifmam_index = ifp->if_index;
1132 	ifmam->ifmam_addrs = rtinfo.rti_addrs;
1133 
1134 	rts_input(m, familyof(ifma->ifma_addr));
1135 }
1136 
1137 static struct mbuf *
1138 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1139 		     struct rt_addrinfo *info)
1140 {
1141 	struct if_announcemsghdr *ifan;
1142 	struct mbuf *m;
1143 
1144 	if (route_cb.any_count == 0)
1145 		return NULL;
1146 
1147 	bzero(info, sizeof(*info));
1148 	m = rt_msg_mbuf(type, info);
1149 	if (m == NULL)
1150 		return NULL;
1151 
1152 	ifan = mtod(m, struct if_announcemsghdr *);
1153 	ifan->ifan_index = ifp->if_index;
1154 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof ifan->ifan_name);
1155 	ifan->ifan_what = what;
1156 	return m;
1157 }
1158 
1159 /*
1160  * This is called to generate routing socket messages indicating
1161  * IEEE80211 wireless events.
1162  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1163  */
1164 void
1165 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1166 {
1167 	struct rt_addrinfo info;
1168 	struct mbuf *m;
1169 
1170 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1171 	if (m == NULL)
1172 		return;
1173 
1174 	/*
1175 	 * Append the ieee80211 data.  Try to stick it in the
1176 	 * mbuf containing the ifannounce msg; otherwise allocate
1177 	 * a new mbuf and append.
1178 	 *
1179 	 * NB: we assume m is a single mbuf.
1180 	 */
1181 	if (data_len > M_TRAILINGSPACE(m)) {
1182 		struct mbuf *n = m_get(MB_DONTWAIT, MT_DATA);
1183 		if (n == NULL) {
1184 			m_freem(m);
1185 			return;
1186 		}
1187 		bcopy(data, mtod(n, void *), data_len);
1188 		n->m_len = data_len;
1189 		m->m_next = n;
1190 	} else if (data_len > 0) {
1191 		bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1192 		m->m_len += data_len;
1193 	}
1194 	mbuftrackid(m, 33);
1195 	if (m->m_flags & M_PKTHDR)
1196 		m->m_pkthdr.len += data_len;
1197 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1198 	rts_input(m, 0);
1199 }
1200 
1201 /*
1202  * This is called to generate routing socket messages indicating
1203  * network interface arrival and departure.
1204  */
1205 void
1206 rt_ifannouncemsg(struct ifnet *ifp, int what)
1207 {
1208 	struct rt_addrinfo addrinfo;
1209 	struct mbuf *m;
1210 
1211 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &addrinfo);
1212 	if (m != NULL)
1213 		rts_input(m, 0);
1214 }
1215 
1216 static int
1217 resizewalkarg(struct walkarg *w, int len)
1218 {
1219 	void *newptr;
1220 
1221 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
1222 	if (newptr == NULL)
1223 		return (ENOMEM);
1224 	if (w->w_tmem != NULL)
1225 		kfree(w->w_tmem, M_RTABLE);
1226 	w->w_tmem = newptr;
1227 	w->w_tmemsize = len;
1228 	return (0);
1229 }
1230 
1231 /*
1232  * This is used in dumping the kernel table via sysctl().
1233  */
1234 int
1235 sysctl_dumpentry(struct radix_node *rn, void *vw)
1236 {
1237 	struct walkarg *w = vw;
1238 	struct rtentry *rt = (struct rtentry *)rn;
1239 	struct rt_addrinfo rtinfo;
1240 	int error, msglen;
1241 
1242 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1243 		return 0;
1244 
1245 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1246 	rtinfo.rti_dst = rt_key(rt);
1247 	rtinfo.rti_gateway = rt->rt_gateway;
1248 	rtinfo.rti_netmask = rt_mask(rt);
1249 	rtinfo.rti_genmask = rt->rt_genmask;
1250 	if (rt->rt_ifp != NULL) {
1251 		rtinfo.rti_ifpaddr =
1252 		TAILQ_FIRST(&rt->rt_ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1253 		rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
1254 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1255 			rtinfo.rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
1256 	}
1257 	msglen = rt_msgsize(RTM_GET, &rtinfo);
1258 	if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1259 		return (ENOMEM);
1260 	rt_msg_buffer(RTM_GET, &rtinfo, w->w_tmem, msglen);
1261 	if (w->w_req != NULL) {
1262 		struct rt_msghdr *rtm = w->w_tmem;
1263 
1264 		rtm->rtm_flags = rt->rt_flags;
1265 		rtm->rtm_use = rt->rt_use;
1266 		rtm->rtm_rmx = rt->rt_rmx;
1267 		rtm->rtm_index = rt->rt_ifp->if_index;
1268 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1269 		rtm->rtm_addrs = rtinfo.rti_addrs;
1270 		error = SYSCTL_OUT(w->w_req, rtm, msglen);
1271 		return (error);
1272 	}
1273 	return (0);
1274 }
1275 
1276 static int
1277 sysctl_iflist(int af, struct walkarg *w)
1278 {
1279 	struct ifnet *ifp;
1280 	struct rt_addrinfo rtinfo;
1281 	int msglen, error;
1282 
1283 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1284 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1285 		struct ifaddr_container *ifac;
1286 		struct ifaddr *ifa;
1287 
1288 		if (w->w_arg && w->w_arg != ifp->if_index)
1289 			continue;
1290 		ifac = TAILQ_FIRST(&ifp->if_addrheads[mycpuid]);
1291 		ifa = ifac->ifa;
1292 		rtinfo.rti_ifpaddr = ifa->ifa_addr;
1293 		msglen = rt_msgsize(RTM_IFINFO, &rtinfo);
1294 		if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1295 			return (ENOMEM);
1296 		rt_msg_buffer(RTM_IFINFO, &rtinfo, w->w_tmem, msglen);
1297 		rtinfo.rti_ifpaddr = NULL;
1298 		if (w->w_req != NULL && w->w_tmem != NULL) {
1299 			struct if_msghdr *ifm = w->w_tmem;
1300 
1301 			ifm->ifm_index = ifp->if_index;
1302 			ifm->ifm_flags = ifp->if_flags;
1303 			ifm->ifm_data = ifp->if_data;
1304 			ifm->ifm_addrs = rtinfo.rti_addrs;
1305 			error = SYSCTL_OUT(w->w_req, ifm, msglen);
1306 			if (error)
1307 				return (error);
1308 		}
1309 		while ((ifac = TAILQ_NEXT(ifac, ifa_link)) != NULL) {
1310 			ifa = ifac->ifa;
1311 
1312 			if (af && af != ifa->ifa_addr->sa_family)
1313 				continue;
1314 			if (curproc->p_ucred->cr_prison &&
1315 			    prison_if(curproc->p_ucred, ifa->ifa_addr))
1316 				continue;
1317 			rtinfo.rti_ifaaddr = ifa->ifa_addr;
1318 			rtinfo.rti_netmask = ifa->ifa_netmask;
1319 			rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
1320 			msglen = rt_msgsize(RTM_NEWADDR, &rtinfo);
1321 			if (w->w_tmemsize < msglen &&
1322 			    resizewalkarg(w, msglen) != 0)
1323 				return (ENOMEM);
1324 			rt_msg_buffer(RTM_NEWADDR, &rtinfo, w->w_tmem, msglen);
1325 			if (w->w_req != NULL) {
1326 				struct ifa_msghdr *ifam = w->w_tmem;
1327 
1328 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1329 				ifam->ifam_flags = ifa->ifa_flags;
1330 				ifam->ifam_metric = ifa->ifa_metric;
1331 				ifam->ifam_addrs = rtinfo.rti_addrs;
1332 				error = SYSCTL_OUT(w->w_req, w->w_tmem, msglen);
1333 				if (error)
1334 					return (error);
1335 			}
1336 		}
1337 		rtinfo.rti_netmask = NULL;
1338 		rtinfo.rti_ifaaddr = NULL;
1339 		rtinfo.rti_bcastaddr = NULL;
1340 	}
1341 	return (0);
1342 }
1343 
1344 static int
1345 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1346 {
1347 	int	*name = (int *)arg1;
1348 	u_int	namelen = arg2;
1349 	struct radix_node_head *rnh;
1350 	int	i, error = EINVAL;
1351 	int	origcpu;
1352 	u_char  af;
1353 	struct	walkarg w;
1354 
1355 	name ++;
1356 	namelen--;
1357 	if (req->newptr)
1358 		return (EPERM);
1359 	if (namelen != 3 && namelen != 4)
1360 		return (EINVAL);
1361 	af = name[0];
1362 	bzero(&w, sizeof w);
1363 	w.w_op = name[1];
1364 	w.w_arg = name[2];
1365 	w.w_req = req;
1366 
1367 	/*
1368 	 * Optional third argument specifies cpu, used primarily for
1369 	 * debugging the route table.
1370 	 */
1371 	if (namelen == 4) {
1372 		if (name[3] < 0 || name[3] >= ncpus)
1373 			return (EINVAL);
1374 		origcpu = mycpuid;
1375 		lwkt_migratecpu(name[3]);
1376 	} else {
1377 		origcpu = -1;
1378 	}
1379 	crit_enter();
1380 	switch (w.w_op) {
1381 	case NET_RT_DUMP:
1382 	case NET_RT_FLAGS:
1383 		for (i = 1; i <= AF_MAX; i++)
1384 			if ((rnh = rt_tables[mycpuid][i]) &&
1385 			    (af == 0 || af == i) &&
1386 			    (error = rnh->rnh_walktree(rnh,
1387 						       sysctl_dumpentry, &w)))
1388 				break;
1389 		break;
1390 
1391 	case NET_RT_IFLIST:
1392 		error = sysctl_iflist(af, &w);
1393 	}
1394 	crit_exit();
1395 	if (w.w_tmem != NULL)
1396 		kfree(w.w_tmem, M_RTABLE);
1397 	if (origcpu >= 0)
1398 		lwkt_migratecpu(origcpu);
1399 	return (error);
1400 }
1401 
1402 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1403 
1404 /*
1405  * Definitions of protocols supported in the ROUTE domain.
1406  */
1407 
1408 static struct domain routedomain;		/* or at least forward */
1409 
1410 static struct protosw routesw[] = {
1411 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
1412   0,		route_output,	raw_ctlinput,	0,
1413   cpu0_soport,	cpu0_ctlport,
1414   raw_init,	0,		0,		0,
1415   &route_usrreqs
1416 }
1417 };
1418 
1419 static struct domain routedomain = {
1420 	PF_ROUTE, "route", NULL, NULL, NULL,
1421 	routesw, &routesw[(sizeof routesw)/(sizeof routesw[0])],
1422 };
1423 
1424 DOMAIN_SET(route);
1425 
1426