xref: /netbsd-src/sys/net/rtsock.c (revision 80d9064ac03cbb6a4174695f0d5b237c8766d3d0)
1 /*	$NetBSD: rtsock.c,v 1.164 2014/09/05 06:00:05 matt Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
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 project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1988, 1991, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
61  */
62 
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: rtsock.c,v 1.164 2014/09/05 06:00:05 matt Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_mpls.h"
69 #include "opt_compat_netbsd.h"
70 #endif
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/proc.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/sysctl.h>
80 #include <sys/kauth.h>
81 #include <sys/kmem.h>
82 #include <sys/intr.h>
83 #ifdef RTSOCK_DEBUG
84 #include <netinet/in.h>
85 #endif /* RTSOCK_DEBUG */
86 
87 #include <net/if.h>
88 #include <net/route.h>
89 #include <net/raw_cb.h>
90 
91 #include <netmpls/mpls.h>
92 
93 #if defined(COMPAT_14) || defined(COMPAT_50)
94 #include <compat/net/if.h>
95 #include <compat/net/route.h>
96 #endif
97 #ifdef COMPAT_RTSOCK
98 #define	RTM_XVERSION	RTM_OVERSION
99 #define	RT_XADVANCE(a,b) RT_OADVANCE(a,b)
100 #define	RT_XROUNDUP(n)	RT_OROUNDUP(n)
101 #define	PF_XROUTE	PF_OROUTE
102 #define	rt_xmsghdr	rt_msghdr50
103 #define	if_xmsghdr	if_msghdr	/* if_msghdr50 is for RTM_OIFINFO */
104 #define	ifa_xmsghdr	ifa_msghdr50
105 #define	if_xannouncemsghdr	if_announcemsghdr50
106 #define	COMPATNAME(x)	compat_50_ ## x
107 #define	DOMAINNAME	"oroute"
108 CTASSERT(sizeof(struct ifa_xmsghdr) == 20);
109 DOMAIN_DEFINE(compat_50_routedomain); /* forward declare and add to link set */
110 #else
111 #define	RTM_XVERSION	RTM_VERSION
112 #define	RT_XADVANCE(a,b) RT_ADVANCE(a,b)
113 #define	RT_XROUNDUP(n)	RT_ROUNDUP(n)
114 #define	PF_XROUTE	PF_ROUTE
115 #define	rt_xmsghdr	rt_msghdr
116 #define	if_xmsghdr	if_msghdr
117 #define	ifa_xmsghdr	ifa_msghdr
118 #define	if_xannouncemsghdr	if_announcemsghdr
119 #define	COMPATNAME(x)	x
120 #define	DOMAINNAME	"route"
121 CTASSERT(sizeof(struct ifa_xmsghdr) == 24);
122 #ifdef COMPAT_50
123 #define	COMPATCALL(name, args)	compat_50_ ## name args
124 #endif
125 DOMAIN_DEFINE(routedomain); /* forward declare and add to link set */
126 #undef COMPAT_50
127 #undef COMPAT_14
128 #endif
129 
130 #ifndef COMPATCALL
131 #define	COMPATCALL(name, args)	do { } while (/*CONSTCOND*/ 0)
132 #endif
133 
134 struct route_info COMPATNAME(route_info) = {
135 	.ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, },
136 	.ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, },
137 	.ri_maxqlen = IFQ_MAXLEN,
138 };
139 
140 #define	PRESERVED_RTF	(RTF_UP | RTF_GATEWAY | RTF_HOST | RTF_DONE | RTF_MASK)
141 
142 static void COMPATNAME(route_init)(void);
143 static int COMPATNAME(route_output)(struct mbuf *, ...);
144 
145 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
146 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
147 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
148     struct rt_addrinfo *);
149 static void rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *);
150 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *);
151 static void sysctl_net_route_setup(struct sysctllog **);
152 static int sysctl_dumpentry(struct rtentry *, void *);
153 static int sysctl_iflist(int, struct rt_walkarg *, int);
154 static int sysctl_rtable(SYSCTLFN_PROTO);
155 static void rt_adjustcount(int, int);
156 
157 static void
158 rt_adjustcount(int af, int cnt)
159 {
160 	struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
161 
162 	cb->any_count += cnt;
163 
164 	switch (af) {
165 	case AF_INET:
166 		cb->ip_count += cnt;
167 		return;
168 #ifdef INET6
169 	case AF_INET6:
170 		cb->ip6_count += cnt;
171 		return;
172 #endif
173 	case AF_MPLS:
174 		cb->mpls_count += cnt;
175 		return;
176 	}
177 }
178 
179 static int
180 COMPATNAME(route_attach)(struct socket *so, int proto)
181 {
182 	struct rawcb *rp;
183 	int s, error;
184 
185 	KASSERT(sotorawcb(so) == NULL);
186 	rp = kmem_zalloc(sizeof(*rp), KM_SLEEP);
187 	rp->rcb_len = sizeof(*rp);
188 	so->so_pcb = rp;
189 
190 	s = splsoftnet();
191 	if ((error = raw_attach(so, proto)) == 0) {
192 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
193 		rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
194 		rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
195 	}
196 	splx(s);
197 
198 	if (error) {
199 		kmem_free(rp, sizeof(*rp));
200 		so->so_pcb = NULL;
201 		return error;
202 	}
203 
204 	soisconnected(so);
205 	so->so_options |= SO_USELOOPBACK;
206 	KASSERT(solocked(so));
207 
208 	return error;
209 }
210 
211 static void
212 COMPATNAME(route_detach)(struct socket *so)
213 {
214 	struct rawcb *rp = sotorawcb(so);
215 	int s;
216 
217 	KASSERT(rp != NULL);
218 	KASSERT(solocked(so));
219 
220 	s = splsoftnet();
221 	rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
222 	raw_detach(so);
223 	splx(s);
224 }
225 
226 static int
227 COMPATNAME(route_accept)(struct socket *so, struct mbuf *nam)
228 {
229 	KASSERT(solocked(so));
230 
231 	panic("route_accept");
232 
233 	return EOPNOTSUPP;
234 }
235 
236 static int
237 COMPATNAME(route_bind)(struct socket *so, struct mbuf *nam, struct lwp *l)
238 {
239 	KASSERT(solocked(so));
240 
241 	return EOPNOTSUPP;
242 }
243 
244 static int
245 COMPATNAME(route_listen)(struct socket *so, struct lwp *l)
246 {
247 	KASSERT(solocked(so));
248 
249 	return EOPNOTSUPP;
250 }
251 
252 static int
253 COMPATNAME(route_connect)(struct socket *so, struct mbuf *nam, struct lwp *l)
254 {
255 	KASSERT(solocked(so));
256 
257 	return EOPNOTSUPP;
258 }
259 
260 static int
261 COMPATNAME(route_connect2)(struct socket *so, struct socket *so2)
262 {
263 	KASSERT(solocked(so));
264 
265 	return EOPNOTSUPP;
266 }
267 
268 static int
269 COMPATNAME(route_disconnect)(struct socket *so)
270 {
271 	struct rawcb *rp = sotorawcb(so);
272 	int s;
273 
274 	KASSERT(solocked(so));
275 	KASSERT(rp != NULL);
276 
277 	s = splsoftnet();
278 	soisdisconnected(so);
279 	raw_disconnect(rp);
280 	splx(s);
281 
282 	return 0;
283 }
284 
285 static int
286 COMPATNAME(route_shutdown)(struct socket *so)
287 {
288 	int s;
289 
290 	KASSERT(solocked(so));
291 
292 	/*
293 	 * Mark the connection as being incapable of further input.
294 	 */
295 	s = splsoftnet();
296 	socantsendmore(so);
297 	splx(s);
298 	return 0;
299 }
300 
301 static int
302 COMPATNAME(route_abort)(struct socket *so)
303 {
304 	KASSERT(solocked(so));
305 
306 	panic("route_abort");
307 
308 	return EOPNOTSUPP;
309 }
310 
311 static int
312 COMPATNAME(route_ioctl)(struct socket *so, u_long cmd, void *nam,
313     struct ifnet * ifp)
314 {
315 	return EOPNOTSUPP;
316 }
317 
318 static int
319 COMPATNAME(route_stat)(struct socket *so, struct stat *ub)
320 {
321 	KASSERT(solocked(so));
322 
323 	return 0;
324 }
325 
326 static int
327 COMPATNAME(route_peeraddr)(struct socket *so, struct mbuf *nam)
328 {
329 	struct rawcb *rp = sotorawcb(so);
330 
331 	KASSERT(solocked(so));
332 	KASSERT(rp != NULL);
333 	KASSERT(nam != NULL);
334 
335 	if (rp->rcb_faddr == NULL)
336 		return ENOTCONN;
337 
338 	raw_setpeeraddr(rp, nam);
339 	return 0;
340 }
341 
342 static int
343 COMPATNAME(route_sockaddr)(struct socket *so, struct mbuf *nam)
344 {
345 	struct rawcb *rp = sotorawcb(so);
346 
347 	KASSERT(solocked(so));
348 	KASSERT(rp != NULL);
349 	KASSERT(nam != NULL);
350 
351 	if (rp->rcb_faddr == NULL)
352 		return ENOTCONN;
353 
354 	raw_setsockaddr(rp, nam);
355 	return 0;
356 }
357 
358 static int
359 COMPATNAME(route_rcvd)(struct socket *so, int flags, struct lwp *l)
360 {
361 	KASSERT(solocked(so));
362 
363 	return EOPNOTSUPP;
364 }
365 
366 static int
367 COMPATNAME(route_recvoob)(struct socket *so, struct mbuf *m, int flags)
368 {
369 	KASSERT(solocked(so));
370 
371 	return EOPNOTSUPP;
372 }
373 
374 static int
375 COMPATNAME(route_send)(struct socket *so, struct mbuf *m,
376     struct mbuf *nam, struct mbuf *control, struct lwp *l)
377 {
378 	int error = 0;
379 	int s;
380 
381 	KASSERT(solocked(so));
382 
383 	s = splsoftnet();
384 	error = raw_send(so, m, nam, control, l);
385 	splx(s);
386 
387 	return error;
388 }
389 
390 static int
391 COMPATNAME(route_sendoob)(struct socket *so, struct mbuf *m,
392     struct mbuf *control)
393 {
394 	KASSERT(solocked(so));
395 
396 	m_freem(m);
397 	m_freem(control);
398 
399 	return EOPNOTSUPP;
400 }
401 static int
402 COMPATNAME(route_purgeif)(struct socket *so, struct ifnet *ifp)
403 {
404 
405 	panic("route_purgeif");
406 
407 	return EOPNOTSUPP;
408 }
409 
410 static int
411 COMPATNAME(route_usrreq)(struct socket *so, int req, struct mbuf *m,
412     struct mbuf *nam, struct mbuf *control, struct lwp *l)
413 {
414 	int s, error = 0;
415 
416 	KASSERT(req != PRU_ATTACH);
417 	KASSERT(req != PRU_DETACH);
418 	KASSERT(req != PRU_ACCEPT);
419 	KASSERT(req != PRU_BIND);
420 	KASSERT(req != PRU_LISTEN);
421 	KASSERT(req != PRU_CONNECT);
422 	KASSERT(req != PRU_CONNECT2);
423 	KASSERT(req != PRU_DISCONNECT);
424 	KASSERT(req != PRU_SHUTDOWN);
425 	KASSERT(req != PRU_ABORT);
426 	KASSERT(req != PRU_CONTROL);
427 	KASSERT(req != PRU_SENSE);
428 	KASSERT(req != PRU_PEERADDR);
429 	KASSERT(req != PRU_SOCKADDR);
430 	KASSERT(req != PRU_RCVD);
431 	KASSERT(req != PRU_RCVOOB);
432 	KASSERT(req != PRU_SEND);
433 	KASSERT(req != PRU_SENDOOB);
434 	KASSERT(req != PRU_PURGEIF);
435 
436 	s = splsoftnet();
437 	error = raw_usrreq(so, req, m, nam, control, l);
438 	splx(s);
439 
440 	return error;
441 }
442 
443 /*ARGSUSED*/
444 int
445 COMPATNAME(route_output)(struct mbuf *m, ...)
446 {
447 	struct sockproto proto = { .sp_family = PF_XROUTE, };
448 	struct rt_xmsghdr *rtm = NULL;
449 	struct rt_xmsghdr *old_rtm = NULL;
450 	struct rtentry *rt = NULL;
451 	struct rtentry *saved_nrt = NULL;
452 	struct rt_addrinfo info;
453 	int len, error = 0;
454 	struct ifnet *ifp = NULL;
455 	struct ifaddr *ifa = NULL;
456 	struct socket *so;
457 	va_list ap;
458 	sa_family_t family;
459 
460 	va_start(ap, m);
461 	so = va_arg(ap, struct socket *);
462 	va_end(ap);
463 
464 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
465 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
466 	   (m = m_pullup(m, sizeof(int32_t))) == NULL))
467 		return ENOBUFS;
468 	if ((m->m_flags & M_PKTHDR) == 0)
469 		panic("%s", __func__);
470 	len = m->m_pkthdr.len;
471 	if (len < sizeof(*rtm) ||
472 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
473 		info.rti_info[RTAX_DST] = NULL;
474 		senderr(EINVAL);
475 	}
476 	R_Malloc(rtm, struct rt_xmsghdr *, len);
477 	if (rtm == NULL) {
478 		info.rti_info[RTAX_DST] = NULL;
479 		senderr(ENOBUFS);
480 	}
481 	m_copydata(m, 0, len, rtm);
482 	if (rtm->rtm_version != RTM_XVERSION) {
483 		info.rti_info[RTAX_DST] = NULL;
484 		senderr(EPROTONOSUPPORT);
485 	}
486 	rtm->rtm_pid = curproc->p_pid;
487 	memset(&info, 0, sizeof(info));
488 	info.rti_addrs = rtm->rtm_addrs;
489 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
490 	    &info)) {
491 		senderr(EINVAL);
492 	}
493 	info.rti_flags = rtm->rtm_flags;
494 #ifdef RTSOCK_DEBUG
495 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
496 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
497 		    inet_ntoa(((const struct sockaddr_in *)
498 		    info.rti_info[RTAX_DST])->sin_addr));
499 	}
500 #endif /* RTSOCK_DEBUG */
501 	if (info.rti_info[RTAX_DST] == NULL ||
502 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
503 		senderr(EINVAL);
504 	}
505 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
506 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
507 		senderr(EINVAL);
508 	}
509 
510 	/*
511 	 * Verify that the caller has the appropriate privilege; RTM_GET
512 	 * is the only operation the non-superuser is allowed.
513 	 */
514 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
515 	    0, rtm, NULL, NULL) != 0)
516 		senderr(EACCES);
517 
518 	switch (rtm->rtm_type) {
519 
520 	case RTM_ADD:
521 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
522 			senderr(EINVAL);
523 		}
524 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
525 		if (error == 0 && saved_nrt) {
526 			rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
527 			saved_nrt->rt_refcnt--;
528 		}
529 		break;
530 
531 	case RTM_DELETE:
532 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
533 		if (error == 0) {
534 			(rt = saved_nrt)->rt_refcnt++;
535 			goto report;
536 		}
537 		break;
538 
539 	case RTM_GET:
540 	case RTM_CHANGE:
541 	case RTM_LOCK:
542                 /* XXX This will mask info.rti_info[RTAX_DST] with
543 		 * info.rti_info[RTAX_NETMASK] before
544                  * searching.  It did not used to do that.  --dyoung
545 		 */
546 		error = rtrequest1(RTM_GET, &info, &rt);
547 		if (error != 0)
548 			senderr(error);
549 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
550 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
551 			    info.rti_info[RTAX_DST]->sa_len) != 0)
552 				senderr(ESRCH);
553 			if (info.rti_info[RTAX_NETMASK] == NULL &&
554 			    rt_mask(rt) != NULL)
555 				senderr(ETOOMANYREFS);
556 		}
557 
558 		switch (rtm->rtm_type) {
559 		case RTM_GET:
560 		report:
561 			info.rti_info[RTAX_DST] = rt_getkey(rt);
562 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
563 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
564 			info.rti_info[RTAX_TAG] = rt_gettag(rt);
565 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
566 				;
567 			else if ((ifp = rt->rt_ifp) != NULL) {
568 				const struct ifaddr *rtifa;
569 				info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
570                                 /* rtifa used to be simply rt->rt_ifa.
571                                  * If rt->rt_ifa != NULL, then
572                                  * rt_get_ifa() != NULL.  So this
573                                  * ought to still be safe. --dyoung
574 				 */
575 				rtifa = rt_get_ifa(rt);
576 				info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
577 #ifdef RTSOCK_DEBUG
578 				if (info.rti_info[RTAX_IFA]->sa_family ==
579 				    AF_INET) {
580 					printf("%s: copying out RTAX_IFA %s ",
581 					    __func__, inet_ntoa(
582 					    ((const struct sockaddr_in *)
583 					    info.rti_info[RTAX_IFA])->sin_addr)
584 					    );
585 					printf("for info.rti_info[RTAX_DST] %s "
586 					    "ifa_getifa %p ifa_seqno %p\n",
587 					    inet_ntoa(
588 					    ((const struct sockaddr_in *)
589 					    info.rti_info[RTAX_DST])->sin_addr),
590 					    (void *)rtifa->ifa_getifa,
591 					    rtifa->ifa_seqno);
592 				}
593 #endif /* RTSOCK_DEBUG */
594 				if (ifp->if_flags & IFF_POINTOPOINT) {
595 					info.rti_info[RTAX_BRD] =
596 					    rtifa->ifa_dstaddr;
597 				} else
598 					info.rti_info[RTAX_BRD] = NULL;
599 				rtm->rtm_index = ifp->if_index;
600 			} else {
601 				info.rti_info[RTAX_IFP] = NULL;
602 				info.rti_info[RTAX_IFA] = NULL;
603 			}
604 			(void)rt_msg2(rtm->rtm_type, &info, NULL, NULL, &len);
605 			if (len > rtm->rtm_msglen) {
606 				old_rtm = rtm;
607 				R_Malloc(rtm, struct rt_xmsghdr *, len);
608 				if (rtm == NULL)
609 					senderr(ENOBUFS);
610 				(void)memcpy(rtm, old_rtm, old_rtm->rtm_msglen);
611 			}
612 			(void)rt_msg2(rtm->rtm_type, &info, rtm, NULL, 0);
613 			rtm->rtm_flags = rt->rt_flags;
614 			rtm_setmetrics(rt, rtm);
615 			rtm->rtm_addrs = info.rti_addrs;
616 			break;
617 
618 		case RTM_CHANGE:
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 ((error = rt_getifa(&info)) != 0)
625 				senderr(error);
626 			if (info.rti_info[RTAX_GATEWAY] &&
627 			    rt_setgate(rt, info.rti_info[RTAX_GATEWAY]))
628 				senderr(EDQUOT);
629 			if (info.rti_info[RTAX_TAG])
630 				rt_settag(rt, info.rti_info[RTAX_TAG]);
631 			/* new gateway could require new ifaddr, ifp;
632 			   flags may also be different; ifp may be specified
633 			   by ll sockaddr when protocol address is ambiguous */
634 			if (info.rti_info[RTAX_IFP] &&
635 			    (ifa = ifa_ifwithnet(info.rti_info[RTAX_IFP])) &&
636 			    (ifp = ifa->ifa_ifp) && (info.rti_info[RTAX_IFA] ||
637 			    info.rti_info[RTAX_GATEWAY])) {
638 				if (info.rti_info[RTAX_IFA] == NULL ||
639 				    (ifa = ifa_ifwithaddr(
640 				    info.rti_info[RTAX_IFA])) == NULL)
641 					ifa = ifaof_ifpforaddr(
642 					    info.rti_info[RTAX_IFA] ?
643 					    info.rti_info[RTAX_IFA] :
644 					    info.rti_info[RTAX_GATEWAY], ifp);
645 			} else if ((info.rti_info[RTAX_IFA] &&
646 			    (ifa = ifa_ifwithaddr(info.rti_info[RTAX_IFA]))) ||
647 			    (info.rti_info[RTAX_GATEWAY] &&
648 			    (ifa = ifa_ifwithroute(rt->rt_flags,
649 			    rt_getkey(rt), info.rti_info[RTAX_GATEWAY])))) {
650 				ifp = ifa->ifa_ifp;
651 			}
652 			if (ifa) {
653 				struct ifaddr *oifa = rt->rt_ifa;
654 				if (oifa != ifa) {
655 					if (oifa && oifa->ifa_rtrequest) {
656 						oifa->ifa_rtrequest(RTM_DELETE,
657 						    rt, &info);
658 					}
659 					rt_replace_ifa(rt, ifa);
660 					rt->rt_ifp = ifp;
661 				}
662 			}
663 			if (ifp && rt->rt_ifp != ifp)
664 				rt->rt_ifp = ifp;
665 			rt_setmetrics(rtm->rtm_inits, rtm, rt);
666 			if (rt->rt_flags != info.rti_flags)
667 				rt->rt_flags = (info.rti_flags & ~PRESERVED_RTF)
668 				    | (rt->rt_flags & PRESERVED_RTF);
669 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
670 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
671 			/*FALLTHROUGH*/
672 		case RTM_LOCK:
673 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
674 			rt->rt_rmx.rmx_locks |=
675 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
676 			break;
677 		}
678 		break;
679 
680 	default:
681 		senderr(EOPNOTSUPP);
682 	}
683 
684 flush:
685 	if (rtm) {
686 		if (error)
687 			rtm->rtm_errno = error;
688 		else
689 			rtm->rtm_flags |= RTF_DONE;
690 	}
691 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
692 	    0;
693 	/* We cannot free old_rtm until we have stopped using the
694 	 * pointers in info, some of which may point to sockaddrs
695 	 * in old_rtm.
696 	 */
697 	if (old_rtm != NULL)
698 		Free(old_rtm);
699 	if (rt)
700 		rtfree(rt);
701     {
702 	struct rawcb *rp = NULL;
703 	/*
704 	 * Check to see if we don't want our own messages.
705 	 */
706 	if ((so->so_options & SO_USELOOPBACK) == 0) {
707 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
708 			if (rtm)
709 				Free(rtm);
710 			m_freem(m);
711 			return error;
712 		}
713 		/* There is another listener, so construct message */
714 		rp = sotorawcb(so);
715 	}
716 	if (rtm) {
717 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
718 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
719 			m_freem(m);
720 			m = NULL;
721 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
722 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
723 		Free(rtm);
724 	}
725 	if (rp)
726 		rp->rcb_proto.sp_family = 0; /* Avoid us */
727 	if (family)
728 		proto.sp_protocol = family;
729 	if (m)
730 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
731 		    &COMPATNAME(route_info).ri_dst);
732 	if (rp)
733 		rp->rcb_proto.sp_family = PF_XROUTE;
734     }
735 	return error;
736 }
737 
738 static void
739 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
740 {
741 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
742 	metric(RTV_RPIPE, rmx_recvpipe);
743 	metric(RTV_SPIPE, rmx_sendpipe);
744 	metric(RTV_SSTHRESH, rmx_ssthresh);
745 	metric(RTV_RTT, rmx_rtt);
746 	metric(RTV_RTTVAR, rmx_rttvar);
747 	metric(RTV_HOPCOUNT, rmx_hopcount);
748 	metric(RTV_MTU, rmx_mtu);
749 	metric(RTV_EXPIRE, rmx_expire);
750 #undef metric
751 }
752 
753 static void
754 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
755 {
756 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
757 	metric(rmx_recvpipe);
758 	metric(rmx_sendpipe);
759 	metric(rmx_ssthresh);
760 	metric(rmx_rtt);
761 	metric(rmx_rttvar);
762 	metric(rmx_hopcount);
763 	metric(rmx_mtu);
764 	metric(rmx_expire);
765 #undef metric
766 }
767 
768 static int
769 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
770     struct rt_addrinfo *rtinfo)
771 {
772 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
773 	int i;
774 
775 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
776 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
777 			continue;
778 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
779 		RT_XADVANCE(cp, sa);
780 	}
781 
782 	/*
783 	 * Check for extra addresses specified, except RTM_GET asking
784 	 * for interface info.
785 	 */
786 	if (rtmtype == RTM_GET) {
787 		if (((rtinfo->rti_addrs &
788 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
789 			return 1;
790 	} else if ((rtinfo->rti_addrs & (~0 << i)) != 0)
791 		return 1;
792 	/* Check for bad data length.  */
793 	if (cp != cplim) {
794 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
795 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
796 			/*
797 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
798 			 * We accept this for now for the sake of old
799 			 * binaries or third party softwares.
800 			 */
801 			;
802 		else
803 			return 1;
804 	}
805 	return 0;
806 }
807 
808 static int
809 rt_getlen(int type)
810 {
811 #ifndef COMPAT_RTSOCK
812 	CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
813 	CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
814 	CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
815 	CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
816 #endif
817 
818 	switch (type) {
819 	case RTM_DELADDR:
820 	case RTM_NEWADDR:
821 	case RTM_CHGADDR:
822 		return sizeof(struct ifa_xmsghdr);
823 
824 	case RTM_OOIFINFO:
825 #ifdef COMPAT_14
826 		return sizeof(struct if_msghdr14);
827 #else
828 #ifdef DIAGNOSTIC
829 		printf("RTM_OOIFINFO\n");
830 #endif
831 		return -1;
832 #endif
833 	case RTM_OIFINFO:
834 #ifdef COMPAT_50
835 		return sizeof(struct if_msghdr50);
836 #else
837 #ifdef DIAGNOSTIC
838 		printf("RTM_OIFINFO\n");
839 #endif
840 		return -1;
841 #endif
842 
843 	case RTM_IFINFO:
844 		return sizeof(struct if_xmsghdr);
845 
846 	case RTM_IFANNOUNCE:
847 	case RTM_IEEE80211:
848 		return sizeof(struct if_xannouncemsghdr);
849 
850 	default:
851 		return sizeof(struct rt_xmsghdr);
852 	}
853 }
854 
855 
856 struct mbuf *
857 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
858 {
859 	struct rt_xmsghdr *rtm;
860 	struct mbuf *m;
861 	int i;
862 	const struct sockaddr *sa;
863 	int len, dlen;
864 
865 	m = m_gethdr(M_DONTWAIT, MT_DATA);
866 	if (m == NULL)
867 		return m;
868 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
869 
870 	if ((len = rt_getlen(type)) == -1)
871 		goto out;
872 	if (len > MHLEN + MLEN)
873 		panic("%s: message too long", __func__);
874 	else if (len > MHLEN) {
875 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
876 		if (m->m_next == NULL)
877 			goto out;
878 		MCLAIM(m->m_next, m->m_owner);
879 		m->m_pkthdr.len = len;
880 		m->m_len = MHLEN;
881 		m->m_next->m_len = len - MHLEN;
882 	} else {
883 		m->m_pkthdr.len = m->m_len = len;
884 	}
885 	m->m_pkthdr.rcvif = NULL;
886 	m_copyback(m, 0, datalen, data);
887 	if (len > datalen)
888 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
889 	rtm = mtod(m, struct rt_xmsghdr *);
890 	for (i = 0; i < RTAX_MAX; i++) {
891 		if ((sa = rtinfo->rti_info[i]) == NULL)
892 			continue;
893 		rtinfo->rti_addrs |= (1 << i);
894 		dlen = RT_XROUNDUP(sa->sa_len);
895 		m_copyback(m, len, sa->sa_len, sa);
896 		if (dlen != sa->sa_len) {
897 			/*
898 			 * Up to 6 + 1 nul's since roundup is to
899 			 * sizeof(uint64_t) (8 bytes)
900 			 */
901 			m_copyback(m, len + sa->sa_len,
902 			    dlen - sa->sa_len, "\0\0\0\0\0\0");
903 		}
904 		len += dlen;
905 	}
906 	if (m->m_pkthdr.len != len)
907 		goto out;
908 	rtm->rtm_msglen = len;
909 	rtm->rtm_version = RTM_XVERSION;
910 	rtm->rtm_type = type;
911 	return m;
912 out:
913 	m_freem(m);
914 	return NULL;
915 }
916 
917 /*
918  * rt_msg2
919  *
920  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
921  *		returns the length of the message in 'lenp'.
922  *
923  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
924  *	the message
925  * otherwise walkarg's w_needed is updated and if the user buffer is
926  *	specified and w_needed indicates space exists the information is copied
927  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
928  *	if the allocation fails ENOBUFS is returned.
929  */
930 static int
931 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
932 	int *lenp)
933 {
934 	int i;
935 	int len, dlen, second_time = 0;
936 	char *cp0, *cp = cpv;
937 
938 	rtinfo->rti_addrs = 0;
939 again:
940 	if ((len = rt_getlen(type)) == -1)
941 		return EINVAL;
942 
943 	if ((cp0 = cp) != NULL)
944 		cp += len;
945 	for (i = 0; i < RTAX_MAX; i++) {
946 		const struct sockaddr *sa;
947 
948 		if ((sa = rtinfo->rti_info[i]) == NULL)
949 			continue;
950 		rtinfo->rti_addrs |= (1 << i);
951 		dlen = RT_XROUNDUP(sa->sa_len);
952 		if (cp) {
953 			int diff = dlen - sa->sa_len;
954 			(void)memcpy(cp, sa, (size_t)sa->sa_len);
955 			cp += sa->sa_len;
956 			if (diff > 0) {
957 				(void)memset(cp, 0, (size_t)diff);
958 				cp += diff;
959 			}
960 		}
961 		len += dlen;
962 	}
963 	if (cp == NULL && w != NULL && !second_time) {
964 		struct rt_walkarg *rw = w;
965 
966 		rw->w_needed += len;
967 		if (rw->w_needed <= 0 && rw->w_where) {
968 			if (rw->w_tmemsize < len) {
969 				if (rw->w_tmem)
970 					free(rw->w_tmem, M_RTABLE);
971 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
972 				if (rw->w_tmem)
973 					rw->w_tmemsize = len;
974 				else
975 					rw->w_tmemsize = 0;
976 			}
977 			if (rw->w_tmem) {
978 				cp = rw->w_tmem;
979 				second_time = 1;
980 				goto again;
981 			} else {
982 				rw->w_tmemneeded = len;
983 				return ENOBUFS;
984 			}
985 		}
986 	}
987 	if (cp) {
988 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
989 
990 		rtm->rtm_version = RTM_XVERSION;
991 		rtm->rtm_type = type;
992 		rtm->rtm_msglen = len;
993 	}
994 	if (lenp)
995 		*lenp = len;
996 	return 0;
997 }
998 
999 /*
1000  * This routine is called to generate a message from the routing
1001  * socket indicating that a redirect has occurred, a routing lookup
1002  * has failed, or that a protocol has detected timeouts to a particular
1003  * destination.
1004  */
1005 void
1006 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
1007     int error)
1008 {
1009 	struct rt_xmsghdr rtm;
1010 	struct mbuf *m;
1011 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1012 	struct rt_addrinfo info = *rtinfo;
1013 
1014 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
1015 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1016 		return;
1017 	memset(&rtm, 0, sizeof(rtm));
1018 	rtm.rtm_flags = RTF_DONE | flags;
1019 	rtm.rtm_errno = error;
1020 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
1021 	if (m == NULL)
1022 		return;
1023 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1024 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1025 }
1026 
1027 /*
1028  * This routine is called to generate a message from the routing
1029  * socket indicating that the status of a network interface has changed.
1030  */
1031 void
1032 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
1033 {
1034 	struct if_xmsghdr ifm;
1035 	struct mbuf *m;
1036 	struct rt_addrinfo info;
1037 
1038 	COMPATCALL(rt_ifmsg, (ifp));
1039 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1040 		return;
1041 	(void)memset(&info, 0, sizeof(info));
1042 	(void)memset(&ifm, 0, sizeof(ifm));
1043 	ifm.ifm_index = ifp->if_index;
1044 	ifm.ifm_flags = ifp->if_flags;
1045 	ifm.ifm_data = ifp->if_data;
1046 	ifm.ifm_addrs = 0;
1047 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
1048 	if (m == NULL)
1049 		return;
1050 	COMPATNAME(route_enqueue)(m, 0);
1051 #ifdef COMPAT_14
1052 	compat_14_rt_oifmsg(ifp);
1053 #endif
1054 #ifdef COMPAT_50
1055 	compat_50_rt_oifmsg(ifp);
1056 #endif
1057 }
1058 
1059 
1060 /*
1061  * This is called to generate messages from the routing socket
1062  * indicating a network interface has had addresses associated with it.
1063  * if we ever reverse the logic and replace messages TO the routing
1064  * socket indicate a request to configure interfaces, then it will
1065  * be unnecessary as the routing socket will automatically generate
1066  * copies of it.
1067  */
1068 void
1069 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
1070     struct rtentry *rt)
1071 {
1072 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
1073 	struct rt_addrinfo info;
1074 	const struct sockaddr *sa;
1075 	int pass;
1076 	struct mbuf *m;
1077 	struct ifnet *ifp;
1078 	struct rt_xmsghdr rtm;
1079 	struct ifa_xmsghdr ifam;
1080 	int ncmd;
1081 
1082 	KASSERT(ifa != NULL);
1083 	ifp = ifa->ifa_ifp;
1084 	COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
1085 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1086 		return;
1087 	for (pass = 1; pass < 3; pass++) {
1088 		memset(&info, 0, sizeof(info));
1089 		switch (cmdpass(cmd, pass)) {
1090 		case cmdpass(RTM_ADD, 1):
1091 		case cmdpass(RTM_CHANGE, 1):
1092 		case cmdpass(RTM_DELETE, 2):
1093 		case cmdpass(RTM_NEWADDR, 1):
1094 		case cmdpass(RTM_DELADDR, 1):
1095 		case cmdpass(RTM_CHGADDR, 1):
1096 			switch (cmd) {
1097 			case RTM_ADD:
1098 				ncmd = RTM_NEWADDR;
1099 				break;
1100 			case RTM_DELETE:
1101 				ncmd = RTM_DELADDR;
1102 				break;
1103 			case RTM_CHANGE:
1104 				ncmd = RTM_CHGADDR;
1105 				break;
1106 			default:
1107 				ncmd = cmd;
1108 			}
1109 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1110 			KASSERT(ifp->if_dl != NULL);
1111 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1112 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1113 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1114 			memset(&ifam, 0, sizeof(ifam));
1115 			ifam.ifam_index = ifp->if_index;
1116 			ifam.ifam_metric = ifa->ifa_metric;
1117 			ifam.ifam_flags = ifa->ifa_flags;
1118 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
1119 			if (m == NULL)
1120 				continue;
1121 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
1122 			    info.rti_addrs;
1123 			break;
1124 		case cmdpass(RTM_ADD, 2):
1125 		case cmdpass(RTM_CHANGE, 2):
1126 		case cmdpass(RTM_DELETE, 1):
1127 			if (rt == NULL)
1128 				continue;
1129 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1130 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
1131 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1132 			memset(&rtm, 0, sizeof(rtm));
1133 			rtm.rtm_index = ifp->if_index;
1134 			rtm.rtm_flags |= rt->rt_flags;
1135 			rtm.rtm_errno = error;
1136 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
1137 			if (m == NULL)
1138 				continue;
1139 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1140 			break;
1141 		default:
1142 			continue;
1143 		}
1144 #ifdef DIAGNOSTIC
1145 		if (m == NULL)
1146 			panic("%s: called with wrong command", __func__);
1147 #endif
1148 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1149 	}
1150 #undef cmdpass
1151 }
1152 
1153 static struct mbuf *
1154 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1155     struct rt_addrinfo *info)
1156 {
1157 	struct if_xannouncemsghdr ifan;
1158 
1159 	memset(info, 0, sizeof(*info));
1160 	memset(&ifan, 0, sizeof(ifan));
1161 	ifan.ifan_index = ifp->if_index;
1162 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
1163 	ifan.ifan_what = what;
1164 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
1165 }
1166 
1167 /*
1168  * This is called to generate routing socket messages indicating
1169  * network interface arrival and departure.
1170  */
1171 void
1172 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
1173 {
1174 	struct mbuf *m;
1175 	struct rt_addrinfo info;
1176 
1177 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
1178 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1179 		return;
1180 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1181 	if (m == NULL)
1182 		return;
1183 	COMPATNAME(route_enqueue)(m, 0);
1184 }
1185 
1186 /*
1187  * This is called to generate routing socket messages indicating
1188  * IEEE80211 wireless events.
1189  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1190  */
1191 void
1192 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
1193 	size_t data_len)
1194 {
1195 	struct mbuf *m;
1196 	struct rt_addrinfo info;
1197 
1198 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
1199 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1200 		return;
1201 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1202 	if (m == NULL)
1203 		return;
1204 	/*
1205 	 * Append the ieee80211 data.  Try to stick it in the
1206 	 * mbuf containing the ifannounce msg; otherwise allocate
1207 	 * a new mbuf and append.
1208 	 *
1209 	 * NB: we assume m is a single mbuf.
1210 	 */
1211 	if (data_len > M_TRAILINGSPACE(m)) {
1212 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1213 		if (n == NULL) {
1214 			m_freem(m);
1215 			return;
1216 		}
1217 		(void)memcpy(mtod(n, void *), data, data_len);
1218 		n->m_len = data_len;
1219 		m->m_next = n;
1220 	} else if (data_len > 0) {
1221 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
1222 		m->m_len += data_len;
1223 	}
1224 	if (m->m_flags & M_PKTHDR)
1225 		m->m_pkthdr.len += data_len;
1226 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
1227 	COMPATNAME(route_enqueue)(m, 0);
1228 }
1229 
1230 /*
1231  * This is used in dumping the kernel table via sysctl().
1232  */
1233 static int
1234 sysctl_dumpentry(struct rtentry *rt, void *v)
1235 {
1236 	struct rt_walkarg *w = v;
1237 	int error = 0, size;
1238 	struct rt_addrinfo info;
1239 
1240 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1241 		return 0;
1242 	memset(&info, 0, sizeof(info));
1243 	info.rti_info[RTAX_DST] = rt_getkey(rt);
1244 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1245 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1246 	info.rti_info[RTAX_TAG] = rt_gettag(rt);
1247 	if (rt->rt_ifp) {
1248 		const struct ifaddr *rtifa;
1249 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
1250 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
1251 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
1252 		 * --dyoung
1253 		 */
1254 		rtifa = rt_get_ifa(rt);
1255 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
1256 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1257 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
1258 	}
1259 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
1260 		return error;
1261 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1262 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
1263 
1264 		rtm->rtm_flags = rt->rt_flags;
1265 		rtm->rtm_use = rt->rt_use;
1266 		rtm_setmetrics(rt, rtm);
1267 		KASSERT(rt->rt_ifp != NULL);
1268 		rtm->rtm_index = rt->rt_ifp->if_index;
1269 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1270 		rtm->rtm_addrs = info.rti_addrs;
1271 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1272 			w->w_where = NULL;
1273 		else
1274 			w->w_where = (char *)w->w_where + size;
1275 	}
1276 	return error;
1277 }
1278 
1279 static int
1280 sysctl_iflist(int af, struct rt_walkarg *w, int type)
1281 {
1282 	struct ifnet *ifp;
1283 	struct ifaddr *ifa;
1284 	struct	rt_addrinfo info;
1285 	int	len, error = 0;
1286 
1287 	memset(&info, 0, sizeof(info));
1288 	IFNET_FOREACH(ifp) {
1289 		if (w->w_arg && w->w_arg != ifp->if_index)
1290 			continue;
1291 		if (IFADDR_EMPTY(ifp))
1292 			continue;
1293 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1294 		switch (type) {
1295 		case NET_RT_IFLIST:
1296 			error = rt_msg2(RTM_IFINFO, &info, NULL, w, &len);
1297 			break;
1298 #ifdef COMPAT_14
1299 		case NET_RT_OOIFLIST:
1300 			error = rt_msg2(RTM_OOIFINFO, &info, NULL, w, &len);
1301 			break;
1302 #endif
1303 #ifdef COMPAT_50
1304 		case NET_RT_OIFLIST:
1305 			error = rt_msg2(RTM_OIFINFO, &info, NULL, w, &len);
1306 			break;
1307 #endif
1308 		default:
1309 			panic("sysctl_iflist(1)");
1310 		}
1311 		if (error)
1312 			return error;
1313 		info.rti_info[RTAX_IFP] = NULL;
1314 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1315 			switch (type) {
1316 			case NET_RT_IFLIST: {
1317 				struct if_xmsghdr *ifm;
1318 
1319 				ifm = (struct if_xmsghdr *)w->w_tmem;
1320 				ifm->ifm_index = ifp->if_index;
1321 				ifm->ifm_flags = ifp->if_flags;
1322 				ifm->ifm_data = ifp->if_data;
1323 				ifm->ifm_addrs = info.rti_addrs;
1324 				error = copyout(ifm, w->w_where, len);
1325 				if (error)
1326 					return error;
1327 				w->w_where = (char *)w->w_where + len;
1328 				break;
1329 			}
1330 
1331 #ifdef COMPAT_14
1332 			case NET_RT_OOIFLIST:
1333 				error = compat_14_iflist(ifp, w, &info, len);
1334 				if (error)
1335 					return error;
1336 				break;
1337 #endif
1338 #ifdef COMPAT_50
1339 			case NET_RT_OIFLIST:
1340 				error = compat_50_iflist(ifp, w, &info, len);
1341 				if (error)
1342 					return error;
1343 				break;
1344 #endif
1345 			default:
1346 				panic("sysctl_iflist(2)");
1347 			}
1348 		}
1349 		IFADDR_FOREACH(ifa, ifp) {
1350 			if (af && af != ifa->ifa_addr->sa_family)
1351 				continue;
1352 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1353 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1354 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1355 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
1356 				return error;
1357 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1358 				struct ifa_xmsghdr *ifam;
1359 
1360 				ifam = (struct ifa_xmsghdr *)w->w_tmem;
1361 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1362 				ifam->ifam_flags = ifa->ifa_flags;
1363 				ifam->ifam_metric = ifa->ifa_metric;
1364 				ifam->ifam_addrs = info.rti_addrs;
1365 				error = copyout(w->w_tmem, w->w_where, len);
1366 				if (error)
1367 					return error;
1368 				w->w_where = (char *)w->w_where + len;
1369 			}
1370 		}
1371 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1372 		    info.rti_info[RTAX_BRD] = NULL;
1373 	}
1374 	return 0;
1375 }
1376 
1377 static int
1378 sysctl_rtable(SYSCTLFN_ARGS)
1379 {
1380 	void 	*where = oldp;
1381 	size_t	*given = oldlenp;
1382 	int	i, s, error = EINVAL;
1383 	u_char  af;
1384 	struct	rt_walkarg w;
1385 
1386 	if (namelen == 1 && name[0] == CTL_QUERY)
1387 		return sysctl_query(SYSCTLFN_CALL(rnode));
1388 
1389 	if (newp)
1390 		return EPERM;
1391 	if (namelen != 3)
1392 		return EINVAL;
1393 	af = name[0];
1394 	w.w_tmemneeded = 0;
1395 	w.w_tmemsize = 0;
1396 	w.w_tmem = NULL;
1397 again:
1398 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1399 	if (w.w_tmemneeded) {
1400 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
1401 		w.w_tmemsize = w.w_tmemneeded;
1402 		w.w_tmemneeded = 0;
1403 	}
1404 	w.w_op = name[1];
1405 	w.w_arg = name[2];
1406 	w.w_given = *given;
1407 	w.w_needed = 0 - w.w_given;
1408 	w.w_where = where;
1409 
1410 	s = splsoftnet();
1411 	switch (w.w_op) {
1412 
1413 	case NET_RT_DUMP:
1414 	case NET_RT_FLAGS:
1415 		for (i = 1; i <= AF_MAX; i++)
1416 			if ((af == 0 || af == i) &&
1417 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
1418 				break;
1419 		break;
1420 
1421 #ifdef COMPAT_14
1422 	case NET_RT_OOIFLIST:
1423 		error = sysctl_iflist(af, &w, w.w_op);
1424 		break;
1425 #endif
1426 #ifdef COMPAT_50
1427 	case NET_RT_OIFLIST:
1428 		error = sysctl_iflist(af, &w, w.w_op);
1429 		break;
1430 #endif
1431 	case NET_RT_IFLIST:
1432 		error = sysctl_iflist(af, &w, w.w_op);
1433 		break;
1434 	}
1435 	splx(s);
1436 
1437 	/* check to see if we couldn't allocate memory with NOWAIT */
1438 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1439 		goto again;
1440 
1441 	if (w.w_tmem)
1442 		free(w.w_tmem, M_RTABLE);
1443 	w.w_needed += w.w_given;
1444 	if (where) {
1445 		*given = (char *)w.w_where - (char *)where;
1446 		if (*given < w.w_needed)
1447 			return ENOMEM;
1448 	} else {
1449 		*given = (11 * w.w_needed) / 10;
1450 	}
1451 	return error;
1452 }
1453 
1454 /*
1455  * Routing message software interrupt routine
1456  */
1457 static void
1458 COMPATNAME(route_intr)(void *cookie)
1459 {
1460 	struct sockproto proto = { .sp_family = PF_XROUTE, };
1461 	struct route_info * const ri = &COMPATNAME(route_info);
1462 	struct mbuf *m;
1463 	int s;
1464 
1465 	mutex_enter(softnet_lock);
1466 	KERNEL_LOCK(1, NULL);
1467 	while (!IF_IS_EMPTY(&ri->ri_intrq)) {
1468 		s = splnet();
1469 		IF_DEQUEUE(&ri->ri_intrq, m);
1470 		splx(s);
1471 		if (m == NULL)
1472 			break;
1473 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
1474 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst);
1475 	}
1476 	KERNEL_UNLOCK_ONE(NULL);
1477 	mutex_exit(softnet_lock);
1478 }
1479 
1480 /*
1481  * Enqueue a message to the software interrupt routine.
1482  */
1483 void
1484 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
1485 {
1486 	struct route_info * const ri = &COMPATNAME(route_info);
1487 	int s, wasempty;
1488 
1489 	s = splnet();
1490 	if (IF_QFULL(&ri->ri_intrq)) {
1491 		IF_DROP(&ri->ri_intrq);
1492 		m_freem(m);
1493 	} else {
1494 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
1495 		M_SETCTX(m, (uintptr_t)family);
1496 		IF_ENQUEUE(&ri->ri_intrq, m);
1497 		if (wasempty)
1498 			softint_schedule(ri->ri_sih);
1499 	}
1500 	splx(s);
1501 }
1502 
1503 static void
1504 COMPATNAME(route_init)(void)
1505 {
1506 	struct route_info * const ri = &COMPATNAME(route_info);
1507 
1508 #ifndef COMPAT_RTSOCK
1509 	rt_init();
1510 #endif
1511 
1512 	sysctl_net_route_setup(NULL);
1513 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
1514 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
1515 	    COMPATNAME(route_intr), NULL);
1516 }
1517 
1518 /*
1519  * Definitions of protocols supported in the ROUTE domain.
1520  */
1521 #ifndef COMPAT_RTSOCK
1522 PR_WRAP_USRREQS(route);
1523 #else
1524 PR_WRAP_USRREQS(compat_50_route);
1525 #endif
1526 
1527 static const struct pr_usrreqs route_usrreqs = {
1528 	.pr_attach	= COMPATNAME(route_attach_wrapper),
1529 	.pr_detach	= COMPATNAME(route_detach_wrapper),
1530 	.pr_accept	= COMPATNAME(route_accept_wrapper),
1531 	.pr_bind	= COMPATNAME(route_bind_wrapper),
1532 	.pr_listen	= COMPATNAME(route_listen_wrapper),
1533 	.pr_connect	= COMPATNAME(route_connect_wrapper),
1534 	.pr_connect2	= COMPATNAME(route_connect2_wrapper),
1535 	.pr_disconnect	= COMPATNAME(route_disconnect_wrapper),
1536 	.pr_shutdown	= COMPATNAME(route_shutdown_wrapper),
1537 	.pr_abort	= COMPATNAME(route_abort_wrapper),
1538 	.pr_ioctl	= COMPATNAME(route_ioctl_wrapper),
1539 	.pr_stat	= COMPATNAME(route_stat_wrapper),
1540 	.pr_peeraddr	= COMPATNAME(route_peeraddr_wrapper),
1541 	.pr_sockaddr	= COMPATNAME(route_sockaddr_wrapper),
1542 	.pr_rcvd	= COMPATNAME(route_rcvd_wrapper),
1543 	.pr_recvoob	= COMPATNAME(route_recvoob_wrapper),
1544 	.pr_send	= COMPATNAME(route_send_wrapper),
1545 	.pr_sendoob	= COMPATNAME(route_sendoob_wrapper),
1546 	.pr_purgeif	= COMPATNAME(route_purgeif_wrapper),
1547 	.pr_generic	= COMPATNAME(route_usrreq_wrapper),
1548 };
1549 
1550 static const struct protosw COMPATNAME(route_protosw)[] = {
1551 	{
1552 		.pr_type = SOCK_RAW,
1553 		.pr_domain = &COMPATNAME(routedomain),
1554 		.pr_flags = PR_ATOMIC|PR_ADDR,
1555 		.pr_input = raw_input,
1556 		.pr_output = COMPATNAME(route_output),
1557 		.pr_ctlinput = raw_ctlinput,
1558 		.pr_usrreqs = &route_usrreqs,
1559 		.pr_init = raw_init,
1560 	},
1561 };
1562 
1563 struct domain COMPATNAME(routedomain) = {
1564 	.dom_family = PF_XROUTE,
1565 	.dom_name = DOMAINNAME,
1566 	.dom_init = COMPATNAME(route_init),
1567 	.dom_protosw = COMPATNAME(route_protosw),
1568 	.dom_protoswNPROTOSW =
1569 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
1570 };
1571 
1572 static void
1573 sysctl_net_route_setup(struct sysctllog **clog)
1574 {
1575 	const struct sysctlnode *rnode = NULL;
1576 
1577 	sysctl_createv(clog, 0, NULL, &rnode,
1578 		       CTLFLAG_PERMANENT,
1579 		       CTLTYPE_NODE, DOMAINNAME,
1580 		       SYSCTL_DESCR("PF_ROUTE information"),
1581 		       NULL, 0, NULL, 0,
1582 		       CTL_NET, PF_XROUTE, CTL_EOL);
1583 
1584 	sysctl_createv(clog, 0, NULL, NULL,
1585 		       CTLFLAG_PERMANENT,
1586 		       CTLTYPE_NODE, "rtable",
1587 		       SYSCTL_DESCR("Routing table information"),
1588 		       sysctl_rtable, 0, NULL, 0,
1589 		       CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
1590 
1591 	sysctl_createv(clog, 0, &rnode, NULL,
1592 		       CTLFLAG_PERMANENT,
1593 		       CTLTYPE_STRUCT, "stats",
1594 		       SYSCTL_DESCR("Routing statistics"),
1595 		       NULL, 0, &rtstat, sizeof(rtstat),
1596 		       CTL_CREATE, CTL_EOL);
1597 }
1598