xref: /dflybsd-src/sys/net/rtsock.c (revision 978400d3b04daf8f91ba8bb2dcc382a37ef632f4)
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 1.42 2008/01/06 16:55:51 swildner 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_addrhead)->ifa_addr;
409 			rtinfo->rti_ifaaddr = rt->rt_ifa->ifa_addr;
410 			if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
411 				rtinfo->rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
412 			rtm->rtm_index = rt->rt_ifp->if_index;
413 		} else {
414 			rtinfo->rti_ifpaddr = NULL;
415 			rtinfo->rti_ifaaddr = NULL;
416 	    }
417 	}
418 
419 	msglen = rt_msgsize(rtm->rtm_type, rtinfo);
420 	if (rtm->rtm_msglen < msglen) {
421 		rtm = reallocbuf(rtm, msglen, rtm->rtm_msglen);
422 		if (rtm == NULL)
423 			return (ENOBUFS);
424 		*prtm = rtm;
425 	}
426 	rt_msg_buffer(rtm->rtm_type, rtinfo, rtm, msglen);
427 
428 	rtm->rtm_flags = rt->rt_flags;
429 	rtm->rtm_rmx = rt->rt_rmx;
430 	rtm->rtm_addrs = rtinfo->rti_addrs;
431 
432 	return (0);
433 }
434 
435 static void route_output_add_callback(int, int, struct rt_addrinfo *,
436 					struct rtentry *, void *);
437 static void route_output_delete_callback(int, int, struct rt_addrinfo *,
438 					struct rtentry *, void *);
439 static void route_output_change_callback(int, int, struct rt_addrinfo *,
440 					struct rtentry *, void *);
441 static void route_output_lock_callback(int, int, struct rt_addrinfo *,
442 					struct rtentry *, void *);
443 
444 /*ARGSUSED*/
445 static int
446 route_output(struct mbuf *m, struct socket *so, ...)
447 {
448 	struct rt_msghdr *rtm = NULL;
449 	struct rtentry *rt;
450 	struct radix_node_head *rnh;
451 	struct rawcb *rp = NULL;
452 	struct pr_output_info *oi;
453 	struct rt_addrinfo rtinfo;
454 	int len, error = 0;
455 	__va_list ap;
456 
457 	__va_start(ap, so);
458 	oi = __va_arg(ap, struct pr_output_info *);
459 	__va_end(ap);
460 
461 #define gotoerr(e) { error = e; goto flush;}
462 
463 	if (m == NULL ||
464 	    (m->m_len < sizeof(long) &&
465 	     (m = m_pullup(m, sizeof(long))) == NULL))
466 		return (ENOBUFS);
467 	if (!(m->m_flags & M_PKTHDR))
468 		panic("route_output");
469 	len = m->m_pkthdr.len;
470 	if (len < sizeof(struct rt_msghdr) ||
471 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
472 		rtinfo.rti_dst = NULL;
473 		gotoerr(EINVAL);
474 	}
475 	rtm = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
476 	if (rtm == NULL) {
477 		rtinfo.rti_dst = NULL;
478 		gotoerr(ENOBUFS);
479 	}
480 	m_copydata(m, 0, len, (caddr_t)rtm);
481 	if (rtm->rtm_version != RTM_VERSION) {
482 		rtinfo.rti_dst = NULL;
483 		gotoerr(EPROTONOSUPPORT);
484 	}
485 	rtm->rtm_pid = oi->p_pid;
486 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
487 	rtinfo.rti_addrs = rtm->rtm_addrs;
488 	if (rt_xaddrs((char *)(rtm + 1), (char *)rtm + len, &rtinfo) != 0) {
489 		rtinfo.rti_dst = NULL;
490 		gotoerr(EINVAL);
491 	}
492 	rtinfo.rti_flags = rtm->rtm_flags;
493 	if (rtinfo.rti_dst == NULL || rtinfo.rti_dst->sa_family >= AF_MAX ||
494 	    (rtinfo.rti_gateway && rtinfo.rti_gateway->sa_family >= AF_MAX))
495 		gotoerr(EINVAL);
496 
497 	if (rtinfo.rti_genmask != NULL) {
498 		struct radix_node *n;
499 
500 #define	clen(s)	(*(u_char *)(s))
501 		n = rn_addmask((char *)rtinfo.rti_genmask, TRUE, 1);
502 		if (n != NULL &&
503 		    rtinfo.rti_genmask->sa_len >= clen(n->rn_key) &&
504 		    bcmp((char *)rtinfo.rti_genmask + 1,
505 		         (char *)n->rn_key + 1, clen(n->rn_key) - 1) == 0)
506 			rtinfo.rti_genmask = (struct sockaddr *)n->rn_key;
507 		else
508 			gotoerr(ENOBUFS);
509 	}
510 
511 	/*
512 	 * Verify that the caller has the appropriate privilege; RTM_GET
513 	 * is the only operation the non-superuser is allowed.
514 	 */
515 	if (rtm->rtm_type != RTM_GET && suser_cred(so->so_cred, 0) != 0)
516 		gotoerr(EPERM);
517 
518 	switch (rtm->rtm_type) {
519 	case RTM_ADD:
520 		if (rtinfo.rti_gateway == NULL) {
521 			error = EINVAL;
522 		} else {
523 			error = rtrequest1_global(RTM_ADD, &rtinfo,
524 					  route_output_add_callback, rtm);
525 		}
526 		break;
527 	case RTM_DELETE:
528 		/*
529 		 * note: &rtm passed as argument so 'rtm' can be replaced.
530 		 */
531 		error = rtrequest1_global(RTM_DELETE, &rtinfo,
532 					  route_output_delete_callback, &rtm);
533 		break;
534 	case RTM_GET:
535 		rnh = rt_tables[mycpuid][rtinfo.rti_dst->sa_family];
536 		if (rnh == NULL) {
537 			error = EAFNOSUPPORT;
538 			break;
539 		}
540 		rt = (struct rtentry *)
541 		    rnh->rnh_lookup((char *)rtinfo.rti_dst,
542 		    		    (char *)rtinfo.rti_netmask, rnh);
543 		if (rt == NULL) {
544 			error = ESRCH;
545 			break;
546 		}
547 		rt->rt_refcnt++;
548 		if (fillrtmsg(&rtm, rt, &rtinfo) != 0)
549 			gotoerr(ENOBUFS);
550 		--rt->rt_refcnt;
551 		break;
552 	case RTM_CHANGE:
553 		error = rtrequest1_global(RTM_GET, &rtinfo,
554 					  route_output_change_callback, rtm);
555 		break;
556 	case RTM_LOCK:
557 		error = rtrequest1_global(RTM_GET, &rtinfo,
558 					  route_output_lock_callback, rtm);
559 		break;
560 	default:
561 		error = EOPNOTSUPP;
562 		break;
563 	}
564 
565 flush:
566 	if (rtm != NULL) {
567 		if (error != 0)
568 			rtm->rtm_errno = error;
569 		else
570 			rtm->rtm_flags |= RTF_DONE;
571 	}
572 
573 	/*
574 	 * Check to see if we don't want our own messages.
575 	 */
576 	if (!(so->so_options & SO_USELOOPBACK)) {
577 		if (route_cb.any_count <= 1) {
578 			if (rtm != NULL)
579 				kfree(rtm, M_RTABLE);
580 			m_freem(m);
581 			return (error);
582 		}
583 		/* There is another listener, so construct message */
584 		rp = sotorawcb(so);
585 	}
586 	if (rtm != NULL) {
587 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
588 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
589 			m_freem(m);
590 			m = NULL;
591 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
592 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
593 		kfree(rtm, M_RTABLE);
594 	}
595 	if (rp != NULL)
596 		rp->rcb_proto.sp_family = 0; /* Avoid us */
597 	if (m != NULL)
598 		rts_input(m, familyof(rtinfo.rti_dst));
599 	if (rp != NULL)
600 		rp->rcb_proto.sp_family = PF_ROUTE;
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 		rt->rt_genmask = rtinfo->rti_genmask;
617 	}
618 }
619 
620 static void
621 route_output_delete_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
622 			  struct rtentry *rt, void *arg)
623 {
624 	struct rt_msghdr **rtm = arg;
625 
626 	if (error == 0 && rt) {
627 		++rt->rt_refcnt;
628 		if (fillrtmsg(rtm, rt, rtinfo) != 0) {
629 			error = ENOBUFS;
630 			/* XXX no way to return the error */
631 		}
632 		--rt->rt_refcnt;
633 	}
634 }
635 
636 static void
637 route_output_change_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
638 			  struct rtentry *rt, void *arg)
639 {
640 	struct rt_msghdr *rtm = arg;
641 	struct ifaddr *ifa;
642 
643 	if (error)
644 		goto done;
645 
646 	/*
647 	 * new gateway could require new ifaddr, ifp;
648 	 * flags may also be different; ifp may be specified
649 	 * by ll sockaddr when protocol address is ambiguous
650 	 */
651 	if (((rt->rt_flags & RTF_GATEWAY) && rtinfo->rti_gateway != NULL) ||
652 	    rtinfo->rti_ifpaddr != NULL || (rtinfo->rti_ifaaddr != NULL &&
653 	    sa_equal(rtinfo->rti_ifaaddr, rt->rt_ifa->ifa_addr))
654 	) {
655 		error = rt_getifa(rtinfo);
656 		if (error != 0)
657 			goto done;
658 	}
659 	if (rtinfo->rti_gateway != NULL) {
660 		error = rt_setgate(rt, rt_key(rt), rtinfo->rti_gateway);
661 		if (error != 0)
662 			goto done;
663 	}
664 	if ((ifa = rtinfo->rti_ifa) != NULL) {
665 		struct ifaddr *oifa = rt->rt_ifa;
666 
667 		if (oifa != ifa) {
668 			if (oifa && oifa->ifa_rtrequest)
669 				oifa->ifa_rtrequest(RTM_DELETE, rt, rtinfo);
670 			IFAFREE(rt->rt_ifa);
671 			IFAREF(ifa);
672 			rt->rt_ifa = ifa;
673 			rt->rt_ifp = rtinfo->rti_ifp;
674 		}
675 	}
676 	rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, &rt->rt_rmx);
677 	if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
678 	       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, rtinfo);
679 	if (rtinfo->rti_genmask != NULL)
680 		rt->rt_genmask = rtinfo->rti_genmask;
681 done:
682 	/* XXX no way to return error */
683 	;
684 }
685 
686 static void
687 route_output_lock_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
688 			   struct rtentry *rt, void *arg)
689 {
690 	struct rt_msghdr *rtm = arg;
691 
692 	rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
693 	rt->rt_rmx.rmx_locks |=
694 		(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
695 }
696 
697 static void
698 rt_setmetrics(u_long which, struct rt_metrics *in, struct rt_metrics *out)
699 {
700 #define setmetric(flag, elt) if (which & (flag)) out->elt = in->elt;
701 	setmetric(RTV_RPIPE, rmx_recvpipe);
702 	setmetric(RTV_SPIPE, rmx_sendpipe);
703 	setmetric(RTV_SSTHRESH, rmx_ssthresh);
704 	setmetric(RTV_RTT, rmx_rtt);
705 	setmetric(RTV_RTTVAR, rmx_rttvar);
706 	setmetric(RTV_HOPCOUNT, rmx_hopcount);
707 	setmetric(RTV_MTU, rmx_mtu);
708 	setmetric(RTV_EXPIRE, rmx_expire);
709 #undef setmetric
710 }
711 
712 #define ROUNDUP(a) \
713 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
714 
715 /*
716  * Extract the addresses of the passed sockaddrs.
717  * Do a little sanity checking so as to avoid bad memory references.
718  * This data is derived straight from userland.
719  */
720 static int
721 rt_xaddrs(char *cp, char *cplim, struct rt_addrinfo *rtinfo)
722 {
723 	struct sockaddr *sa;
724 	int i;
725 
726 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
727 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
728 			continue;
729 		sa = (struct sockaddr *)cp;
730 		/*
731 		 * It won't fit.
732 		 */
733 		if ((cp + sa->sa_len) > cplim) {
734 			return (EINVAL);
735 		}
736 
737 		/*
738 		 * There are no more...  Quit now.
739 		 * If there are more bits, they are in error.
740 		 * I've seen this.  route(1) can evidently generate these.
741 		 * This causes kernel to core dump.
742 		 * For compatibility, if we see this, point to a safe address.
743 		 */
744 		if (sa->sa_len == 0) {
745 			static struct sockaddr sa_zero = {
746 				sizeof sa_zero, AF_INET,
747 			};
748 
749 			rtinfo->rti_info[i] = &sa_zero;
750 			return (0); /* should be EINVAL but for compat */
751 		}
752 
753 		/* Accept the sockaddr. */
754 		rtinfo->rti_info[i] = sa;
755 		cp += ROUNDUP(sa->sa_len);
756 	}
757 	return (0);
758 }
759 
760 static int
761 rt_msghdrsize(int type)
762 {
763 	switch (type) {
764 	case RTM_DELADDR:
765 	case RTM_NEWADDR:
766 		return sizeof(struct ifa_msghdr);
767 	case RTM_DELMADDR:
768 	case RTM_NEWMADDR:
769 		return sizeof(struct ifma_msghdr);
770 	case RTM_IFINFO:
771 		return sizeof(struct if_msghdr);
772 	case RTM_IFANNOUNCE:
773 	case RTM_IEEE80211:
774 		return sizeof(struct if_announcemsghdr);
775 	default:
776 		return sizeof(struct rt_msghdr);
777 	}
778 }
779 
780 static int
781 rt_msgsize(int type, struct rt_addrinfo *rtinfo)
782 {
783 	int len, i;
784 
785 	len = rt_msghdrsize(type);
786 	for (i = 0; i < RTAX_MAX; i++) {
787 		if (rtinfo->rti_info[i] != NULL)
788 			len += ROUNDUP(rtinfo->rti_info[i]->sa_len);
789 	}
790 	len = ALIGN(len);
791 	return len;
792 }
793 
794 /*
795  * Build a routing message in a buffer.
796  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
797  * to the end of the buffer after the message header.
798  *
799  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
800  * This side-effect can be avoided if we reorder the addrs bitmask field in all
801  * the route messages to line up so we can set it here instead of back in the
802  * calling routine.
803  */
804 static void
805 rt_msg_buffer(int type, struct rt_addrinfo *rtinfo, void *buf, int msglen)
806 {
807 	struct rt_msghdr *rtm;
808 	char *cp;
809 	int dlen, i;
810 
811 	rtm = (struct rt_msghdr *) buf;
812 	rtm->rtm_version = RTM_VERSION;
813 	rtm->rtm_type = type;
814 	rtm->rtm_msglen = msglen;
815 
816 	cp = (char *)buf + rt_msghdrsize(type);
817 	rtinfo->rti_addrs = 0;
818 	for (i = 0; i < RTAX_MAX; i++) {
819 		struct sockaddr *sa;
820 
821 		if ((sa = rtinfo->rti_info[i]) == NULL)
822 			continue;
823 		rtinfo->rti_addrs |= (1 << i);
824 		dlen = ROUNDUP(sa->sa_len);
825 		bcopy(sa, cp, dlen);
826 		cp += dlen;
827 	}
828 }
829 
830 /*
831  * Build a routing message in a mbuf chain.
832  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
833  * to the end of the mbuf after the message header.
834  *
835  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
836  * This side-effect can be avoided if we reorder the addrs bitmask field in all
837  * the route messages to line up so we can set it here instead of back in the
838  * calling routine.
839  */
840 static struct mbuf *
841 rt_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
842 {
843 	struct mbuf *m;
844 	struct rt_msghdr *rtm;
845 	int hlen, len;
846 	int i;
847 
848 	hlen = rt_msghdrsize(type);
849 	KASSERT(hlen <= MCLBYTES, ("rt_msg_mbuf: hlen %d doesn't fit", hlen));
850 
851 	m = m_getl(hlen, MB_DONTWAIT, MT_DATA, M_PKTHDR, NULL);
852 	if (m == NULL)
853 		return (NULL);
854 	mbuftrackid(m, 32);
855 	m->m_pkthdr.len = m->m_len = hlen;
856 	m->m_pkthdr.rcvif = NULL;
857 	rtinfo->rti_addrs = 0;
858 	len = hlen;
859 	for (i = 0; i < RTAX_MAX; i++) {
860 		struct sockaddr *sa;
861 		int dlen;
862 
863 		if ((sa = rtinfo->rti_info[i]) == NULL)
864 			continue;
865 		rtinfo->rti_addrs |= (1 << i);
866 		dlen = ROUNDUP(sa->sa_len);
867 		m_copyback(m, len, dlen, (caddr_t)sa); /* can grow mbuf chain */
868 		len += dlen;
869 	}
870 	if (m->m_pkthdr.len != len) { /* one of the m_copyback() calls failed */
871 		m_freem(m);
872 		return (NULL);
873 	}
874 	rtm = mtod(m, struct rt_msghdr *);
875 	bzero(rtm, hlen);
876 	rtm->rtm_msglen = len;
877 	rtm->rtm_version = RTM_VERSION;
878 	rtm->rtm_type = type;
879 	return (m);
880 }
881 
882 /*
883  * This routine is called to generate a message from the routing
884  * socket indicating that a redirect has occurred, a routing lookup
885  * has failed, or that a protocol has detected timeouts to a particular
886  * destination.
887  */
888 void
889 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
890 {
891 	struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
892 	struct rt_msghdr *rtm;
893 	struct mbuf *m;
894 
895 	if (route_cb.any_count == 0)
896 		return;
897 	m = rt_msg_mbuf(type, rtinfo);
898 	if (m == NULL)
899 		return;
900 	rtm = mtod(m, struct rt_msghdr *);
901 	rtm->rtm_flags = RTF_DONE | flags;
902 	rtm->rtm_errno = error;
903 	rtm->rtm_addrs = rtinfo->rti_addrs;
904 	rts_input(m, familyof(dst));
905 }
906 
907 void
908 rt_dstmsg(int type, struct sockaddr *dst, int error)
909 {
910 	struct rt_msghdr *rtm;
911 	struct rt_addrinfo addrs;
912 	struct mbuf *m;
913 
914 	if (route_cb.any_count == 0)
915 		return;
916 	bzero(&addrs, sizeof(struct rt_addrinfo));
917 	addrs.rti_info[RTAX_DST] = dst;
918 	m = rt_msg_mbuf(type, &addrs);
919 	if (m == NULL)
920 		return;
921 	rtm = mtod(m, struct rt_msghdr *);
922 	rtm->rtm_flags = RTF_DONE;
923 	rtm->rtm_errno = error;
924 	rtm->rtm_addrs = addrs.rti_addrs;
925 	rts_input(m, familyof(dst));
926 }
927 
928 /*
929  * This routine is called to generate a message from the routing
930  * socket indicating that the status of a network interface has changed.
931  */
932 void
933 rt_ifmsg(struct ifnet *ifp)
934 {
935 	struct if_msghdr *ifm;
936 	struct mbuf *m;
937 	struct rt_addrinfo rtinfo;
938 
939 	if (route_cb.any_count == 0)
940 		return;
941 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
942 	m = rt_msg_mbuf(RTM_IFINFO, &rtinfo);
943 	if (m == NULL)
944 		return;
945 	ifm = mtod(m, struct if_msghdr *);
946 	ifm->ifm_index = ifp->if_index;
947 	ifm->ifm_flags = ifp->if_flags;
948 	ifm->ifm_data = ifp->if_data;
949 	ifm->ifm_addrs = 0;
950 	rts_input(m, 0);
951 }
952 
953 static void
954 rt_ifamsg(int cmd, struct ifaddr *ifa)
955 {
956 	struct ifa_msghdr *ifam;
957 	struct rt_addrinfo rtinfo;
958 	struct mbuf *m;
959 	struct ifnet *ifp = ifa->ifa_ifp;
960 
961 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
962 	rtinfo.rti_ifaaddr = ifa->ifa_addr;
963 	rtinfo.rti_ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
964 	rtinfo.rti_netmask = ifa->ifa_netmask;
965 	rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
966 
967 	m = rt_msg_mbuf(cmd, &rtinfo);
968 	if (m == NULL)
969 		return;
970 
971 	ifam = mtod(m, struct ifa_msghdr *);
972 	ifam->ifam_index = ifp->if_index;
973 	ifam->ifam_metric = ifa->ifa_metric;
974 	ifam->ifam_flags = ifa->ifa_flags;
975 	ifam->ifam_addrs = rtinfo.rti_addrs;
976 
977 	rts_input(m, familyof(ifa->ifa_addr));
978 }
979 
980 void
981 rt_rtmsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int error)
982 {
983 	struct rt_msghdr *rtm;
984 	struct rt_addrinfo rtinfo;
985 	struct mbuf *m;
986 	struct sockaddr *dst;
987 
988 	if (rt == NULL)
989 		return;
990 
991 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
992 	rtinfo.rti_dst = dst = rt_key(rt);
993 	rtinfo.rti_gateway = rt->rt_gateway;
994 	rtinfo.rti_netmask = rt_mask(rt);
995 	if (ifp != NULL)
996 		rtinfo.rti_ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
997 	rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
998 
999 	m = rt_msg_mbuf(cmd, &rtinfo);
1000 	if (m == NULL)
1001 		return;
1002 
1003 	rtm = mtod(m, struct rt_msghdr *);
1004 	if (ifp != NULL)
1005 		rtm->rtm_index = ifp->if_index;
1006 	rtm->rtm_flags |= rt->rt_flags;
1007 	rtm->rtm_errno = error;
1008 	rtm->rtm_addrs = rtinfo.rti_addrs;
1009 
1010 	rts_input(m, familyof(dst));
1011 }
1012 
1013 /*
1014  * This is called to generate messages from the routing socket
1015  * indicating a network interface has had addresses associated with it.
1016  * if we ever reverse the logic and replace messages TO the routing
1017  * socket indicate a request to configure interfaces, then it will
1018  * be unnecessary as the routing socket will automatically generate
1019  * copies of it.
1020  */
1021 void
1022 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1023 {
1024 #ifdef SCTP
1025 	/*
1026 	 * notify the SCTP stack
1027 	 * this will only get called when an address is added/deleted
1028 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1029 	 */
1030 	if (cmd == RTM_ADD)
1031 		sctp_add_ip_address(ifa);
1032 	else if (cmd == RTM_DELETE)
1033 		sctp_delete_ip_address(ifa);
1034 #endif /* SCTP */
1035 
1036 	if (route_cb.any_count == 0)
1037 		return;
1038 
1039 	if (cmd == RTM_ADD) {
1040 		rt_ifamsg(RTM_NEWADDR, ifa);
1041 		rt_rtmsg(RTM_ADD, rt, ifa->ifa_ifp, error);
1042 	} else {
1043 		KASSERT((cmd == RTM_DELETE), ("unknown cmd %d", cmd));
1044 		rt_rtmsg(RTM_DELETE, rt, ifa->ifa_ifp, error);
1045 		rt_ifamsg(RTM_DELADDR, ifa);
1046 	}
1047 }
1048 
1049 /*
1050  * This is the analogue to the rt_newaddrmsg which performs the same
1051  * function but for multicast group memberhips.  This is easier since
1052  * there is no route state to worry about.
1053  */
1054 void
1055 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1056 {
1057 	struct rt_addrinfo rtinfo;
1058 	struct mbuf *m = NULL;
1059 	struct ifnet *ifp = ifma->ifma_ifp;
1060 	struct ifma_msghdr *ifmam;
1061 
1062 	if (route_cb.any_count == 0)
1063 		return;
1064 
1065 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1066 	rtinfo.rti_ifaaddr = ifma->ifma_addr;
1067 	if (ifp != NULL && !TAILQ_EMPTY(&ifp->if_addrhead))
1068 		rtinfo.rti_ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
1069 	/*
1070 	 * If a link-layer address is present, present it as a ``gateway''
1071 	 * (similarly to how ARP entries, e.g., are presented).
1072 	 */
1073 	rtinfo.rti_gateway = ifma->ifma_lladdr;
1074 
1075 	m = rt_msg_mbuf(cmd, &rtinfo);
1076 	if (m == NULL)
1077 		return;
1078 
1079 	ifmam = mtod(m, struct ifma_msghdr *);
1080 	ifmam->ifmam_index = ifp->if_index;
1081 	ifmam->ifmam_addrs = rtinfo.rti_addrs;
1082 
1083 	rts_input(m, familyof(ifma->ifma_addr));
1084 }
1085 
1086 static struct mbuf *
1087 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1088 		     struct rt_addrinfo *info)
1089 {
1090 	struct if_announcemsghdr *ifan;
1091 	struct mbuf *m;
1092 
1093 	if (route_cb.any_count == 0)
1094 		return NULL;
1095 
1096 	bzero(info, sizeof(*info));
1097 	m = rt_msg_mbuf(type, info);
1098 	if (m == NULL)
1099 		return NULL;
1100 
1101 	ifan = mtod(m, struct if_announcemsghdr *);
1102 	ifan->ifan_index = ifp->if_index;
1103 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof ifan->ifan_name);
1104 	ifan->ifan_what = what;
1105 	return m;
1106 }
1107 
1108 /*
1109  * This is called to generate routing socket messages indicating
1110  * IEEE80211 wireless events.
1111  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1112  */
1113 void
1114 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1115 {
1116 	struct rt_addrinfo info;
1117 	struct mbuf *m;
1118 
1119 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1120 	if (m == NULL)
1121 		return;
1122 
1123 	/*
1124 	 * Append the ieee80211 data.  Try to stick it in the
1125 	 * mbuf containing the ifannounce msg; otherwise allocate
1126 	 * a new mbuf and append.
1127 	 *
1128 	 * NB: we assume m is a single mbuf.
1129 	 */
1130 	if (data_len > M_TRAILINGSPACE(m)) {
1131 		struct mbuf *n = m_get(MB_DONTWAIT, MT_DATA);
1132 		if (n == NULL) {
1133 			m_freem(m);
1134 			return;
1135 		}
1136 		bcopy(data, mtod(n, void *), data_len);
1137 		n->m_len = data_len;
1138 		m->m_next = n;
1139 	} else if (data_len > 0) {
1140 		bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1141 		m->m_len += data_len;
1142 	}
1143 	mbuftrackid(m, 33);
1144 	if (m->m_flags & M_PKTHDR)
1145 		m->m_pkthdr.len += data_len;
1146 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1147 	rts_input(m, 0);
1148 }
1149 
1150 /*
1151  * This is called to generate routing socket messages indicating
1152  * network interface arrival and departure.
1153  */
1154 void
1155 rt_ifannouncemsg(struct ifnet *ifp, int what)
1156 {
1157 	struct rt_addrinfo addrinfo;
1158 	struct mbuf *m;
1159 
1160 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &addrinfo);
1161 	if (m != NULL)
1162 		rts_input(m, 0);
1163 }
1164 
1165 static int
1166 resizewalkarg(struct walkarg *w, int len)
1167 {
1168 	void *newptr;
1169 
1170 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
1171 	if (newptr == NULL)
1172 		return (ENOMEM);
1173 	if (w->w_tmem != NULL)
1174 		kfree(w->w_tmem, M_RTABLE);
1175 	w->w_tmem = newptr;
1176 	w->w_tmemsize = len;
1177 	return (0);
1178 }
1179 
1180 /*
1181  * This is used in dumping the kernel table via sysctl().
1182  */
1183 int
1184 sysctl_dumpentry(struct radix_node *rn, void *vw)
1185 {
1186 	struct walkarg *w = vw;
1187 	struct rtentry *rt = (struct rtentry *)rn;
1188 	struct rt_addrinfo rtinfo;
1189 	int error, msglen;
1190 
1191 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1192 		return 0;
1193 
1194 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1195 	rtinfo.rti_dst = rt_key(rt);
1196 	rtinfo.rti_gateway = rt->rt_gateway;
1197 	rtinfo.rti_netmask = rt_mask(rt);
1198 	rtinfo.rti_genmask = rt->rt_genmask;
1199 	if (rt->rt_ifp != NULL) {
1200 		rtinfo.rti_ifpaddr =
1201 		    TAILQ_FIRST(&rt->rt_ifp->if_addrhead)->ifa_addr;
1202 		rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
1203 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1204 			rtinfo.rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
1205 	}
1206 	msglen = rt_msgsize(RTM_GET, &rtinfo);
1207 	if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1208 		return (ENOMEM);
1209 	rt_msg_buffer(RTM_GET, &rtinfo, w->w_tmem, msglen);
1210 	if (w->w_req != NULL) {
1211 		struct rt_msghdr *rtm = w->w_tmem;
1212 
1213 		rtm->rtm_flags = rt->rt_flags;
1214 		rtm->rtm_use = rt->rt_use;
1215 		rtm->rtm_rmx = rt->rt_rmx;
1216 		rtm->rtm_index = rt->rt_ifp->if_index;
1217 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1218 		rtm->rtm_addrs = rtinfo.rti_addrs;
1219 		error = SYSCTL_OUT(w->w_req, rtm, msglen);
1220 		return (error);
1221 	}
1222 	return (0);
1223 }
1224 
1225 static int
1226 sysctl_iflist(int af, struct walkarg *w)
1227 {
1228 	struct ifnet *ifp;
1229 	struct ifaddr *ifa;
1230 	struct rt_addrinfo rtinfo;
1231 	int msglen, error;
1232 
1233 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1234 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1235 		if (w->w_arg && w->w_arg != ifp->if_index)
1236 			continue;
1237 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1238 		rtinfo.rti_ifpaddr = ifa->ifa_addr;
1239 		msglen = rt_msgsize(RTM_IFINFO, &rtinfo);
1240 		if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1241 			return (ENOMEM);
1242 		rt_msg_buffer(RTM_IFINFO, &rtinfo, w->w_tmem, msglen);
1243 		rtinfo.rti_ifpaddr = NULL;
1244 		if (w->w_req != NULL && w->w_tmem != NULL) {
1245 			struct if_msghdr *ifm = w->w_tmem;
1246 
1247 			ifm->ifm_index = ifp->if_index;
1248 			ifm->ifm_flags = ifp->if_flags;
1249 			ifm->ifm_data = ifp->if_data;
1250 			ifm->ifm_addrs = rtinfo.rti_addrs;
1251 			error = SYSCTL_OUT(w->w_req, ifm, msglen);
1252 			if (error)
1253 				return (error);
1254 		}
1255 		while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1256 			if (af && af != ifa->ifa_addr->sa_family)
1257 				continue;
1258 			if (curproc->p_ucred->cr_prison &&
1259 			    prison_if(curproc->p_ucred, ifa->ifa_addr))
1260 				continue;
1261 			rtinfo.rti_ifaaddr = ifa->ifa_addr;
1262 			rtinfo.rti_netmask = ifa->ifa_netmask;
1263 			rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
1264 			msglen = rt_msgsize(RTM_NEWADDR, &rtinfo);
1265 			if (w->w_tmemsize < msglen &&
1266 			    resizewalkarg(w, msglen) != 0)
1267 				return (ENOMEM);
1268 			rt_msg_buffer(RTM_NEWADDR, &rtinfo, w->w_tmem, msglen);
1269 			if (w->w_req != NULL) {
1270 				struct ifa_msghdr *ifam = w->w_tmem;
1271 
1272 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1273 				ifam->ifam_flags = ifa->ifa_flags;
1274 				ifam->ifam_metric = ifa->ifa_metric;
1275 				ifam->ifam_addrs = rtinfo.rti_addrs;
1276 				error = SYSCTL_OUT(w->w_req, w->w_tmem, msglen);
1277 				if (error)
1278 					return (error);
1279 			}
1280 		}
1281 		rtinfo.rti_netmask = NULL;
1282 		rtinfo.rti_ifaaddr = NULL;
1283 		rtinfo.rti_bcastaddr = NULL;
1284 	}
1285 	return (0);
1286 }
1287 
1288 static int
1289 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1290 {
1291 	int	*name = (int *)arg1;
1292 	u_int	namelen = arg2;
1293 	struct radix_node_head *rnh;
1294 	int	i, error = EINVAL;
1295 	int	origcpu;
1296 	u_char  af;
1297 	struct	walkarg w;
1298 
1299 	name ++;
1300 	namelen--;
1301 	if (req->newptr)
1302 		return (EPERM);
1303 	if (namelen != 3 && namelen != 4)
1304 		return (EINVAL);
1305 	af = name[0];
1306 	bzero(&w, sizeof w);
1307 	w.w_op = name[1];
1308 	w.w_arg = name[2];
1309 	w.w_req = req;
1310 
1311 	/*
1312 	 * Optional third argument specifies cpu, used primarily for
1313 	 * debugging the route table.
1314 	 */
1315 	if (namelen == 4) {
1316 		if (name[3] < 0 || name[3] >= ncpus)
1317 			return (EINVAL);
1318 		origcpu = mycpuid;
1319 		lwkt_migratecpu(name[3]);
1320 	} else {
1321 		origcpu = -1;
1322 	}
1323 	crit_enter();
1324 	switch (w.w_op) {
1325 	case NET_RT_DUMP:
1326 	case NET_RT_FLAGS:
1327 		for (i = 1; i <= AF_MAX; i++)
1328 			if ((rnh = rt_tables[mycpuid][i]) &&
1329 			    (af == 0 || af == i) &&
1330 			    (error = rnh->rnh_walktree(rnh,
1331 						       sysctl_dumpentry, &w)))
1332 				break;
1333 		break;
1334 
1335 	case NET_RT_IFLIST:
1336 		error = sysctl_iflist(af, &w);
1337 	}
1338 	crit_exit();
1339 	if (w.w_tmem != NULL)
1340 		kfree(w.w_tmem, M_RTABLE);
1341 	if (origcpu >= 0)
1342 		lwkt_migratecpu(origcpu);
1343 	return (error);
1344 }
1345 
1346 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1347 
1348 /*
1349  * Definitions of protocols supported in the ROUTE domain.
1350  */
1351 
1352 static struct domain routedomain;		/* or at least forward */
1353 
1354 static struct protosw routesw[] = {
1355 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
1356   0,		route_output,	raw_ctlinput,	0,
1357   cpu0_soport,
1358   raw_init,	0,		0,		0,
1359   &route_usrreqs
1360 }
1361 };
1362 
1363 static struct domain routedomain = {
1364 	PF_ROUTE, "route", NULL, NULL, NULL,
1365 	routesw, &routesw[(sizeof routesw)/(sizeof routesw[0])],
1366 };
1367 
1368 DOMAIN_SET(route);
1369 
1370