xref: /netbsd-src/sys/net/rtsock.c (revision 8450a7c42673d65e3b1f6560d3b6ecd317a6cbe8)
1 /*	$NetBSD: rtsock.c,v 1.198 2016/10/26 06:49:10 ozaki-r 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 1.198 2016/10/26 06:49:10 ozaki-r Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_mpls.h"
69 #include "opt_compat_netbsd.h"
70 #include "opt_sctp.h"
71 #endif
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/proc.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/domain.h>
79 #include <sys/protosw.h>
80 #include <sys/sysctl.h>
81 #include <sys/kauth.h>
82 #include <sys/kmem.h>
83 #include <sys/intr.h>
84 
85 #include <net/if.h>
86 #include <net/if_llatbl.h>
87 #include <net/if_types.h>
88 #include <net/route.h>
89 #include <net/raw_cb.h>
90 
91 #include <netinet/in_var.h>
92 #include <netinet/if_inarp.h>
93 
94 #include <netmpls/mpls.h>
95 
96 #ifdef SCTP
97 extern void sctp_add_ip_address(struct ifaddr *);
98 extern void sctp_delete_ip_address(struct ifaddr *);
99 #endif
100 
101 #if defined(COMPAT_14) || defined(COMPAT_50) || defined(COMPAT_70)
102 #include <compat/net/if.h>
103 #include <compat/net/route.h>
104 #endif
105 #ifdef COMPAT_RTSOCK
106 #define	RTM_XVERSION	RTM_OVERSION
107 #define	RTM_XNEWADDR	RTM_ONEWADDR
108 #define	RTM_XDELADDR	RTM_ODELADDR
109 #define	RTM_XCHGADDR	RTM_OCHGADDR
110 #define	RT_XADVANCE(a,b) RT_OADVANCE(a,b)
111 #define	RT_XROUNDUP(n)	RT_OROUNDUP(n)
112 #define	PF_XROUTE	PF_OROUTE
113 #define	rt_xmsghdr	rt_msghdr50
114 #define	if_xmsghdr	if_msghdr	/* if_msghdr50 is for RTM_OIFINFO */
115 #define	ifa_xmsghdr	ifa_msghdr50
116 #define	if_xannouncemsghdr	if_announcemsghdr50
117 #define	COMPATNAME(x)	compat_50_ ## x
118 #define	DOMAINNAME	"oroute"
119 CTASSERT(sizeof(struct ifa_xmsghdr) == 20);
120 DOMAIN_DEFINE(compat_50_routedomain); /* forward declare and add to link set */
121 #undef COMPAT_70
122 #else /* COMPAT_RTSOCK */
123 #define	RTM_XVERSION	RTM_VERSION
124 #define	RTM_XNEWADDR	RTM_NEWADDR
125 #define	RTM_XDELADDR	RTM_DELADDR
126 #define	RTM_XCHGADDR	RTM_CHGADDR
127 #define	RT_XADVANCE(a,b) RT_ADVANCE(a,b)
128 #define	RT_XROUNDUP(n)	RT_ROUNDUP(n)
129 #define	PF_XROUTE	PF_ROUTE
130 #define	rt_xmsghdr	rt_msghdr
131 #define	if_xmsghdr	if_msghdr
132 #define	ifa_xmsghdr	ifa_msghdr
133 #define	if_xannouncemsghdr	if_announcemsghdr
134 #define	COMPATNAME(x)	x
135 #define	DOMAINNAME	"route"
136 CTASSERT(sizeof(struct ifa_xmsghdr) == 32);
137 #ifdef COMPAT_50
138 #define	COMPATCALL(name, args)	compat_50_ ## name args
139 #endif
140 DOMAIN_DEFINE(routedomain); /* forward declare and add to link set */
141 #undef COMPAT_50
142 #undef COMPAT_14
143 #endif /* COMPAT_RTSOCK */
144 
145 #ifndef COMPATCALL
146 #define	COMPATCALL(name, args)	do { } while (/*CONSTCOND*/ 0)
147 #endif
148 
149 #ifdef RTSOCK_DEBUG
150 #define RT_IN_PRINT(info, b, a) (in_print((b), sizeof(b), \
151     &((const struct sockaddr_in *)(info)->rti_info[(a)])->sin_addr), (b))
152 #endif /* RTSOCK_DEBUG */
153 
154 struct route_info COMPATNAME(route_info) = {
155 	.ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, },
156 	.ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, },
157 	.ri_maxqlen = IFQ_MAXLEN,
158 };
159 
160 #define	PRESERVED_RTF	(RTF_UP | RTF_GATEWAY | RTF_HOST | RTF_DONE | RTF_MASK)
161 
162 static void COMPATNAME(route_init)(void);
163 static int COMPATNAME(route_output)(struct mbuf *, struct socket *);
164 
165 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
166 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
167     struct rt_addrinfo *);
168 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
169 static void rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *);
170 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *);
171 static void sysctl_net_route_setup(struct sysctllog **);
172 static int sysctl_dumpentry(struct rtentry *, void *);
173 static int sysctl_iflist(int, struct rt_walkarg *, int);
174 static int sysctl_rtable(SYSCTLFN_PROTO);
175 static void rt_adjustcount(int, int);
176 
177 static const struct protosw COMPATNAME(route_protosw)[];
178 
179 static void
180 rt_adjustcount(int af, int cnt)
181 {
182 	struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
183 
184 	cb->any_count += cnt;
185 
186 	switch (af) {
187 	case AF_INET:
188 		cb->ip_count += cnt;
189 		return;
190 #ifdef INET6
191 	case AF_INET6:
192 		cb->ip6_count += cnt;
193 		return;
194 #endif
195 	case AF_MPLS:
196 		cb->mpls_count += cnt;
197 		return;
198 	}
199 }
200 
201 static int
202 COMPATNAME(route_attach)(struct socket *so, int proto)
203 {
204 	struct rawcb *rp;
205 	int s, error;
206 
207 	KASSERT(sotorawcb(so) == NULL);
208 	rp = kmem_zalloc(sizeof(*rp), KM_SLEEP);
209 	rp->rcb_len = sizeof(*rp);
210 	so->so_pcb = rp;
211 
212 	s = splsoftnet();
213 	if ((error = raw_attach(so, proto)) == 0) {
214 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
215 		rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
216 		rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
217 	}
218 	splx(s);
219 
220 	if (error) {
221 		kmem_free(rp, sizeof(*rp));
222 		so->so_pcb = NULL;
223 		return error;
224 	}
225 
226 	soisconnected(so);
227 	so->so_options |= SO_USELOOPBACK;
228 	KASSERT(solocked(so));
229 
230 	return error;
231 }
232 
233 static void
234 COMPATNAME(route_detach)(struct socket *so)
235 {
236 	struct rawcb *rp = sotorawcb(so);
237 	int s;
238 
239 	KASSERT(rp != NULL);
240 	KASSERT(solocked(so));
241 
242 	s = splsoftnet();
243 	rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
244 	raw_detach(so);
245 	splx(s);
246 }
247 
248 static int
249 COMPATNAME(route_accept)(struct socket *so, struct sockaddr *nam)
250 {
251 	KASSERT(solocked(so));
252 
253 	panic("route_accept");
254 
255 	return EOPNOTSUPP;
256 }
257 
258 static int
259 COMPATNAME(route_bind)(struct socket *so, struct sockaddr *nam, struct lwp *l)
260 {
261 	KASSERT(solocked(so));
262 
263 	return EOPNOTSUPP;
264 }
265 
266 static int
267 COMPATNAME(route_listen)(struct socket *so, struct lwp *l)
268 {
269 	KASSERT(solocked(so));
270 
271 	return EOPNOTSUPP;
272 }
273 
274 static int
275 COMPATNAME(route_connect)(struct socket *so, struct sockaddr *nam, struct lwp *l)
276 {
277 	KASSERT(solocked(so));
278 
279 	return EOPNOTSUPP;
280 }
281 
282 static int
283 COMPATNAME(route_connect2)(struct socket *so, struct socket *so2)
284 {
285 	KASSERT(solocked(so));
286 
287 	return EOPNOTSUPP;
288 }
289 
290 static int
291 COMPATNAME(route_disconnect)(struct socket *so)
292 {
293 	struct rawcb *rp = sotorawcb(so);
294 	int s;
295 
296 	KASSERT(solocked(so));
297 	KASSERT(rp != NULL);
298 
299 	s = splsoftnet();
300 	soisdisconnected(so);
301 	raw_disconnect(rp);
302 	splx(s);
303 
304 	return 0;
305 }
306 
307 static int
308 COMPATNAME(route_shutdown)(struct socket *so)
309 {
310 	int s;
311 
312 	KASSERT(solocked(so));
313 
314 	/*
315 	 * Mark the connection as being incapable of further input.
316 	 */
317 	s = splsoftnet();
318 	socantsendmore(so);
319 	splx(s);
320 	return 0;
321 }
322 
323 static int
324 COMPATNAME(route_abort)(struct socket *so)
325 {
326 	KASSERT(solocked(so));
327 
328 	panic("route_abort");
329 
330 	return EOPNOTSUPP;
331 }
332 
333 static int
334 COMPATNAME(route_ioctl)(struct socket *so, u_long cmd, void *nam,
335     struct ifnet * ifp)
336 {
337 	return EOPNOTSUPP;
338 }
339 
340 static int
341 COMPATNAME(route_stat)(struct socket *so, struct stat *ub)
342 {
343 	KASSERT(solocked(so));
344 
345 	return 0;
346 }
347 
348 static int
349 COMPATNAME(route_peeraddr)(struct socket *so, struct sockaddr *nam)
350 {
351 	struct rawcb *rp = sotorawcb(so);
352 
353 	KASSERT(solocked(so));
354 	KASSERT(rp != NULL);
355 	KASSERT(nam != NULL);
356 
357 	if (rp->rcb_faddr == NULL)
358 		return ENOTCONN;
359 
360 	raw_setpeeraddr(rp, nam);
361 	return 0;
362 }
363 
364 static int
365 COMPATNAME(route_sockaddr)(struct socket *so, struct sockaddr *nam)
366 {
367 	struct rawcb *rp = sotorawcb(so);
368 
369 	KASSERT(solocked(so));
370 	KASSERT(rp != NULL);
371 	KASSERT(nam != NULL);
372 
373 	if (rp->rcb_faddr == NULL)
374 		return ENOTCONN;
375 
376 	raw_setsockaddr(rp, nam);
377 	return 0;
378 }
379 
380 static int
381 COMPATNAME(route_rcvd)(struct socket *so, int flags, struct lwp *l)
382 {
383 	KASSERT(solocked(so));
384 
385 	return EOPNOTSUPP;
386 }
387 
388 static int
389 COMPATNAME(route_recvoob)(struct socket *so, struct mbuf *m, int flags)
390 {
391 	KASSERT(solocked(so));
392 
393 	return EOPNOTSUPP;
394 }
395 
396 static int
397 COMPATNAME(route_send)(struct socket *so, struct mbuf *m,
398     struct sockaddr *nam, struct mbuf *control, struct lwp *l)
399 {
400 	int error = 0;
401 	int s;
402 
403 	KASSERT(solocked(so));
404 	KASSERT(so->so_proto == &COMPATNAME(route_protosw)[0]);
405 
406 	s = splsoftnet();
407 	error = raw_send(so, m, nam, control, l, &COMPATNAME(route_output));
408 	splx(s);
409 
410 	return error;
411 }
412 
413 static int
414 COMPATNAME(route_sendoob)(struct socket *so, struct mbuf *m,
415     struct mbuf *control)
416 {
417 	KASSERT(solocked(so));
418 
419 	m_freem(m);
420 	m_freem(control);
421 
422 	return EOPNOTSUPP;
423 }
424 static int
425 COMPATNAME(route_purgeif)(struct socket *so, struct ifnet *ifp)
426 {
427 
428 	panic("route_purgeif");
429 
430 	return EOPNOTSUPP;
431 }
432 
433 #ifdef INET
434 static int
435 route_get_sdl_index(struct rt_addrinfo *info, int *sdl_index)
436 {
437 	struct rtentry *nrt;
438 	int error;
439 
440 	error = rtrequest1(RTM_GET, info, &nrt);
441 	if (error != 0)
442 		return error;
443 	/*
444 	 * nrt->rt_ifp->if_index may not be correct
445 	 * due to changing to ifplo0.
446 	 */
447 	*sdl_index = satosdl(nrt->rt_gateway)->sdl_index;
448 	rtfree(nrt);
449 
450 	return 0;
451 }
452 #endif /* INET */
453 
454 static void
455 route_get_sdl(const struct ifnet *ifp, const struct sockaddr *dst,
456     struct sockaddr_dl *sdl, int *flags)
457 {
458 	struct llentry *la;
459 
460 	KASSERT(ifp != NULL);
461 
462 	IF_AFDATA_RLOCK(ifp);
463 	switch (dst->sa_family) {
464 	case AF_INET:
465 		la = lla_lookup(LLTABLE(ifp), 0, dst);
466 		break;
467 	case AF_INET6:
468 		la = lla_lookup(LLTABLE6(ifp), 0, dst);
469 		break;
470 	default:
471 		la = NULL;
472 		KASSERTMSG(0, "Invalid AF=%d\n", dst->sa_family);
473 		break;
474 	}
475 	IF_AFDATA_RUNLOCK(ifp);
476 
477 	void *a = (LLE_IS_VALID(la) && (la->la_flags & LLE_VALID) == LLE_VALID)
478 	    ? &la->ll_addr : NULL;
479 
480 	a = sockaddr_dl_init(sdl, sizeof(*sdl), ifp->if_index, ifp->if_type,
481 	    NULL, 0, a, ifp->if_addrlen);
482 	KASSERT(a != NULL);
483 
484 	if (la != NULL) {
485 		*flags = la->la_flags;
486 		LLE_RUNLOCK(la);
487 	}
488 }
489 
490 static int
491 route_output_report(struct rtentry *rt, struct rt_addrinfo *info,
492     struct rt_xmsghdr *rtm, struct rt_xmsghdr **new_rtm)
493 {
494 	int len;
495 	struct ifnet *ifp;
496 
497 	if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
498 		;
499 	else if ((ifp = rt->rt_ifp) != NULL) {
500 		const struct ifaddr *rtifa;
501 		info->rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
502 		/* rtifa used to be simply rt->rt_ifa.
503 		 * If rt->rt_ifa != NULL, then
504 		 * rt_get_ifa() != NULL.  So this
505 		 * ought to still be safe. --dyoung
506 		 */
507 		rtifa = rt_get_ifa(rt);
508 		info->rti_info[RTAX_IFA] = rtifa->ifa_addr;
509 #ifdef RTSOCK_DEBUG
510 		if (info->rti_info[RTAX_IFA]->sa_family == AF_INET) {
511 			char ibuf[INET_ADDRSTRLEN];
512 			char abuf[INET_ADDRSTRLEN];
513 			printf("%s: copying out RTAX_IFA %s "
514 			    "for info->rti_info[RTAX_DST] %s "
515 			    "ifa_getifa %p ifa_seqno %p\n",
516 			    __func__,
517 			    RT_IN_PRINT(info, ibuf, RTAX_IFA),
518 			    RT_IN_PRINT(info, abuf, RTAX_DST),
519 			    (void *)rtifa->ifa_getifa,
520 			    rtifa->ifa_seqno);
521 		}
522 #endif /* RTSOCK_DEBUG */
523 		if (ifp->if_flags & IFF_POINTOPOINT)
524 			info->rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
525 		else
526 			info->rti_info[RTAX_BRD] = NULL;
527 		rtm->rtm_index = ifp->if_index;
528 	} else {
529 		info->rti_info[RTAX_IFP] = NULL;
530 		info->rti_info[RTAX_IFA] = NULL;
531 	}
532 	(void)rt_msg2(rtm->rtm_type, info, NULL, NULL, &len);
533 	if (len > rtm->rtm_msglen) {
534 		struct rt_xmsghdr *old_rtm = rtm;
535 		R_Malloc(*new_rtm, struct rt_xmsghdr *, len);
536 		if (*new_rtm == NULL)
537 			return ENOBUFS;
538 		(void)memcpy(*new_rtm, old_rtm, old_rtm->rtm_msglen);
539 		rtm = *new_rtm;
540 	}
541 	(void)rt_msg2(rtm->rtm_type, info, rtm, NULL, 0);
542 	rtm->rtm_flags = rt->rt_flags;
543 	rtm_setmetrics(rt, rtm);
544 	rtm->rtm_addrs = info->rti_addrs;
545 
546 	return 0;
547 }
548 
549 static struct ifaddr *
550 route_output_get_ifa(const struct rt_addrinfo info, const struct rtentry *rt,
551     struct ifnet **ifp, struct psref *psref)
552 {
553 	struct ifaddr *ifa = NULL;
554 
555 	*ifp = NULL;
556 	if (info.rti_info[RTAX_IFP] != NULL) {
557 		ifa = ifa_ifwithnet_psref(info.rti_info[RTAX_IFP], psref);
558 		if (ifa == NULL)
559 			goto next;
560 		*ifp = ifa->ifa_ifp;
561 		if (info.rti_info[RTAX_IFA] == NULL &&
562 		    info.rti_info[RTAX_GATEWAY] == NULL)
563 			goto next;
564 		if (info.rti_info[RTAX_IFA] == NULL) {
565 			/* route change <dst> <gw> -ifp <if> */
566 			ifa = ifaof_ifpforaddr_psref(info.rti_info[RTAX_GATEWAY],
567 			    *ifp, psref);
568 		} else {
569 			/* route change <dst> -ifp <if> -ifa <addr> */
570 			ifa = ifa_ifwithaddr_psref(info.rti_info[RTAX_IFA], psref);
571 			if (ifa != NULL)
572 				goto out;
573 			ifa = ifaof_ifpforaddr_psref(info.rti_info[RTAX_IFA],
574 			    *ifp, psref);
575 		}
576 		goto out;
577 	}
578 next:
579 	if (info.rti_info[RTAX_IFA] != NULL) {
580 		/* route change <dst> <gw> -ifa <addr> */
581 		ifa = ifa_ifwithaddr_psref(info.rti_info[RTAX_IFA], psref);
582 		if (ifa != NULL)
583 			goto out;
584 	}
585 	if (info.rti_info[RTAX_GATEWAY] != NULL) {
586 		/* route change <dst> <gw> */
587 		ifa = ifa_ifwithroute_psref(rt->rt_flags, rt_getkey(rt),
588 		    info.rti_info[RTAX_GATEWAY], psref);
589 	}
590 out:
591 	if (ifa != NULL && *ifp == NULL)
592 		*ifp = ifa->ifa_ifp;
593 	return ifa;
594 }
595 
596 static int
597 route_output_change(struct rtentry *rt, struct rt_addrinfo *info,
598     struct rt_xmsghdr *rtm)
599 {
600 	int error = 0;
601 	struct ifnet *ifp, *new_ifp;
602 	struct ifaddr *ifa, *new_ifa;
603 	struct psref psref_ifa, psref_new_ifa, psref_ifp;
604 
605 	/*
606 	 * new gateway could require new ifaddr, ifp;
607 	 * flags may also be different; ifp may be specified
608 	 * by ll sockaddr when protocol address is ambiguous
609 	 */
610 	ifp = rt_getifp(info, &psref_ifp);
611 	ifa = rt_getifa(info, &psref_ifa);
612 	if (ifa == NULL) {
613 		error = ENETUNREACH;
614 		goto out;
615 	}
616 	if (info->rti_info[RTAX_GATEWAY]) {
617 		error = rt_setgate(rt, info->rti_info[RTAX_GATEWAY]);
618 		if (error != 0)
619 			goto out;
620 	}
621 	if (info->rti_info[RTAX_TAG]) {
622 		const struct sockaddr *tag;
623 		tag = rt_settag(rt, info->rti_info[RTAX_TAG]);
624 		if (tag == NULL) {
625 			error = ENOBUFS;
626 			goto out;
627 		}
628 	}
629 	/* new gateway could require new ifaddr, ifp;
630 	   flags may also be different; ifp may be specified
631 	   by ll sockaddr when protocol address is ambiguous */
632 	new_ifa = route_output_get_ifa(*info, rt, &new_ifp, &psref_new_ifa);
633 	if (new_ifa != NULL) {
634 		ifa_release(ifa, &psref_ifa);
635 		ifa = new_ifa;
636 	}
637 	if (ifa) {
638 		struct ifaddr *oifa = rt->rt_ifa;
639 		if (oifa != ifa) {
640 			if (oifa && oifa->ifa_rtrequest)
641 				oifa->ifa_rtrequest(RTM_DELETE, rt, info);
642 			rt_replace_ifa(rt, ifa);
643 			rt->rt_ifp = new_ifp;
644 		}
645 		if (new_ifa == NULL)
646 			ifa_release(ifa, &psref_ifa);
647 	}
648 	ifa_release(new_ifa, &psref_new_ifa);
649 	if (new_ifp && rt->rt_ifp != new_ifp)
650 		rt->rt_ifp = new_ifp;
651 	rt_setmetrics(rtm->rtm_inits, rtm, rt);
652 	if (rt->rt_flags != info->rti_flags)
653 		rt->rt_flags = (info->rti_flags & ~PRESERVED_RTF)
654 		    | (rt->rt_flags & PRESERVED_RTF);
655 	if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
656 		rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
657 out:
658 	if_put(ifp, &psref_ifp);
659 
660 	return error;
661 }
662 
663 /*ARGSUSED*/
664 int
665 COMPATNAME(route_output)(struct mbuf *m, struct socket *so)
666 {
667 	struct sockproto proto = { .sp_family = PF_XROUTE, };
668 	struct rt_xmsghdr *rtm = NULL;
669 	struct rt_xmsghdr *old_rtm = NULL, *new_rtm = NULL;
670 	struct rtentry *rt = NULL;
671 	struct rtentry *saved_nrt = NULL;
672 	struct rt_addrinfo info;
673 	int len, error = 0;
674 	sa_family_t family;
675 	struct sockaddr_dl sdl;
676 	int bound = curlwp_bind();
677 
678 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
679 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
680 	   (m = m_pullup(m, sizeof(int32_t))) == NULL)) {
681 		error = ENOBUFS;
682 		goto out;
683 	}
684 	if ((m->m_flags & M_PKTHDR) == 0)
685 		panic("%s", __func__);
686 	len = m->m_pkthdr.len;
687 	if (len < sizeof(*rtm) ||
688 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
689 		info.rti_info[RTAX_DST] = NULL;
690 		senderr(EINVAL);
691 	}
692 	R_Malloc(rtm, struct rt_xmsghdr *, len);
693 	if (rtm == NULL) {
694 		info.rti_info[RTAX_DST] = NULL;
695 		senderr(ENOBUFS);
696 	}
697 	m_copydata(m, 0, len, rtm);
698 	if (rtm->rtm_version != RTM_XVERSION) {
699 		info.rti_info[RTAX_DST] = NULL;
700 		senderr(EPROTONOSUPPORT);
701 	}
702 	rtm->rtm_pid = curproc->p_pid;
703 	memset(&info, 0, sizeof(info));
704 	info.rti_addrs = rtm->rtm_addrs;
705 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
706 	    &info)) {
707 		senderr(EINVAL);
708 	}
709 	info.rti_flags = rtm->rtm_flags;
710 #ifdef RTSOCK_DEBUG
711 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
712 		char abuf[INET_ADDRSTRLEN];
713 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
714 		    RT_IN_PRINT(&info, abuf, RTAX_DST));
715 	}
716 #endif /* RTSOCK_DEBUG */
717 	if (info.rti_info[RTAX_DST] == NULL ||
718 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
719 		senderr(EINVAL);
720 	}
721 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
722 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
723 		senderr(EINVAL);
724 	}
725 
726 	/*
727 	 * Verify that the caller has the appropriate privilege; RTM_GET
728 	 * is the only operation the non-superuser is allowed.
729 	 */
730 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
731 	    0, rtm, NULL, NULL) != 0)
732 		senderr(EACCES);
733 
734 	switch (rtm->rtm_type) {
735 
736 	case RTM_ADD:
737 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
738 			senderr(EINVAL);
739 		}
740 #ifdef INET
741 		/* support for new ARP code with keeping backcompat */
742 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) {
743 			const struct sockaddr_dl *sdlp =
744 			    satocsdl(info.rti_info[RTAX_GATEWAY]);
745 
746 			/* Allow routing requests by interface index */
747 			if (sdlp->sdl_nlen == 0 && sdlp->sdl_alen == 0
748 			    && sdlp->sdl_slen == 0)
749 				goto fallback;
750 			/*
751 			 * Old arp binaries don't set the sdl_index
752 			 * so we have to complement it.
753 			 */
754 			int sdl_index = sdlp->sdl_index;
755 			if (sdl_index == 0) {
756 				error = route_get_sdl_index(&info, &sdl_index);
757 				if (error != 0)
758 					goto fallback;
759 			} else if (
760 			    info.rti_info[RTAX_DST]->sa_family == AF_INET) {
761 				/*
762 				 * XXX workaround for SIN_PROXY case; proxy arp
763 				 * entry should be in an interface that has
764 				 * a network route including the destination,
765 				 * not a local (link) route that may not be a
766 				 * desired place, for example a tap.
767 				 */
768 				const struct sockaddr_inarp *sina =
769 				    (const struct sockaddr_inarp *)
770 				    info.rti_info[RTAX_DST];
771 				if (sina->sin_other & SIN_PROXY) {
772 					error = route_get_sdl_index(&info,
773 					    &sdl_index);
774 					if (error != 0)
775 						goto fallback;
776 				}
777 			}
778 			error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
779 			    rtm->rtm_rmx.rmx_expire, &info, sdl_index);
780 			break;
781 		}
782 	fallback:
783 #endif /* INET */
784 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
785 		if (error == 0) {
786 			rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
787 			rtfree(saved_nrt);
788 		}
789 		break;
790 
791 	case RTM_DELETE:
792 #ifdef INET
793 		/* support for new ARP code */
794 		if (info.rti_info[RTAX_GATEWAY] &&
795 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
796 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
797 			error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
798 			    rtm->rtm_rmx.rmx_expire, &info, 0);
799 			break;
800 		}
801 #endif /* INET */
802 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
803 		if (error != 0)
804 			break;
805 
806 		rt = saved_nrt;
807 		info.rti_info[RTAX_DST] = rt_getkey(rt);
808 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
809 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
810 		info.rti_info[RTAX_TAG] = rt_gettag(rt);
811 		error = route_output_report(rt, &info, rtm, &new_rtm);
812 		if (error)
813 			senderr(error);
814 		if (new_rtm != NULL) {
815 			old_rtm = rtm;
816 			rtm = new_rtm;
817 		}
818 		break;
819 
820 	case RTM_GET:
821 	case RTM_CHANGE:
822 	case RTM_LOCK:
823                 /* XXX This will mask info.rti_info[RTAX_DST] with
824 		 * info.rti_info[RTAX_NETMASK] before
825                  * searching.  It did not used to do that.  --dyoung
826 		 */
827 		rt = NULL;
828 		error = rtrequest1(RTM_GET, &info, &rt);
829 		if (error != 0)
830 			senderr(error);
831 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
832 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
833 			    info.rti_info[RTAX_DST]->sa_len) != 0)
834 				senderr(ESRCH);
835 			if (info.rti_info[RTAX_NETMASK] == NULL &&
836 			    rt_mask(rt) != NULL)
837 				senderr(ETOOMANYREFS);
838 		}
839 
840 		/*
841 		 * XXX if arp/ndp requests an L2 entry, we have to obtain
842 		 * it from lltable while for the route command we have to
843 		 * return a route as it is. How to distinguish them?
844 		 * For newer arp/ndp, RTF_LLDATA flag set by arp/ndp
845 		 * indicates an L2 entry is requested. For old arp/ndp
846 		 * binaries, we check RTF_UP flag is NOT set; it works
847 		 * by the fact that arp/ndp don't set it while the route
848 		 * command sets it.
849 		 */
850 		if (((rtm->rtm_flags & RTF_LLDATA) != 0 ||
851 		     (rtm->rtm_flags & RTF_UP) == 0) &&
852 		    rtm->rtm_type == RTM_GET &&
853 		    sockaddr_cmp(rt_getkey(rt), info.rti_info[RTAX_DST]) != 0) {
854 			int ll_flags = 0;
855 			route_get_sdl(rt->rt_ifp, info.rti_info[RTAX_DST], &sdl,
856 			    &ll_flags);
857 			info.rti_info[RTAX_GATEWAY] = sstocsa(&sdl);
858 			error = route_output_report(rt, &info, rtm, &new_rtm);
859 			if (error)
860 				senderr(error);
861 			if (new_rtm != NULL) {
862 				old_rtm = rtm;
863 				rtm = new_rtm;
864 			}
865 			rtm->rtm_flags |= RTF_LLDATA;
866 			rtm->rtm_flags |= (ll_flags & LLE_STATIC) ? RTF_STATIC : 0;
867 			break;
868 		}
869 
870 		switch (rtm->rtm_type) {
871 		case RTM_GET:
872 			info.rti_info[RTAX_DST] = rt_getkey(rt);
873 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
874 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
875 			info.rti_info[RTAX_TAG] = rt_gettag(rt);
876 			error = route_output_report(rt, &info, rtm, &new_rtm);
877 			if (error)
878 				senderr(error);
879 			if (new_rtm != NULL) {
880 				old_rtm = rtm;
881 				rtm = new_rtm;
882 			}
883 			break;
884 
885 		case RTM_CHANGE:
886 			error = route_output_change(rt, &info, rtm);
887 			if (error != 0)
888 				goto flush;
889 			/*FALLTHROUGH*/
890 		case RTM_LOCK:
891 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
892 			rt->rt_rmx.rmx_locks |=
893 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
894 			break;
895 		}
896 		break;
897 
898 	default:
899 		senderr(EOPNOTSUPP);
900 	}
901 
902 flush:
903 	if (rtm) {
904 		if (error)
905 			rtm->rtm_errno = error;
906 		else
907 			rtm->rtm_flags |= RTF_DONE;
908 	}
909 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
910 	    0;
911 	/* We cannot free old_rtm until we have stopped using the
912 	 * pointers in info, some of which may point to sockaddrs
913 	 * in old_rtm.
914 	 */
915 	if (old_rtm != NULL)
916 		Free(old_rtm);
917 	if (rt)
918 		rtfree(rt);
919     {
920 	struct rawcb *rp = NULL;
921 	/*
922 	 * Check to see if we don't want our own messages.
923 	 */
924 	if ((so->so_options & SO_USELOOPBACK) == 0) {
925 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
926 			if (rtm)
927 				Free(rtm);
928 			m_freem(m);
929 			goto out;
930 		}
931 		/* There is another listener, so construct message */
932 		rp = sotorawcb(so);
933 	}
934 	if (rtm) {
935 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
936 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
937 			m_freem(m);
938 			m = NULL;
939 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
940 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
941 		Free(rtm);
942 	}
943 	if (rp)
944 		rp->rcb_proto.sp_family = 0; /* Avoid us */
945 	if (family)
946 		proto.sp_protocol = family;
947 	if (m)
948 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
949 		    &COMPATNAME(route_info).ri_dst);
950 	if (rp)
951 		rp->rcb_proto.sp_family = PF_XROUTE;
952     }
953 out:
954 	curlwp_bindx(bound);
955 	return error;
956 }
957 
958 static void
959 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
960 {
961 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
962 	metric(RTV_RPIPE, rmx_recvpipe);
963 	metric(RTV_SPIPE, rmx_sendpipe);
964 	metric(RTV_SSTHRESH, rmx_ssthresh);
965 	metric(RTV_RTT, rmx_rtt);
966 	metric(RTV_RTTVAR, rmx_rttvar);
967 	metric(RTV_HOPCOUNT, rmx_hopcount);
968 	metric(RTV_MTU, rmx_mtu);
969 #undef metric
970 	if (which & RTV_EXPIRE) {
971 		out->rt_rmx.rmx_expire = in->rtm_rmx.rmx_expire ?
972 		    time_wall_to_mono(in->rtm_rmx.rmx_expire) : 0;
973 	}
974 }
975 
976 static void
977 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
978 {
979 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
980 	metric(rmx_recvpipe);
981 	metric(rmx_sendpipe);
982 	metric(rmx_ssthresh);
983 	metric(rmx_rtt);
984 	metric(rmx_rttvar);
985 	metric(rmx_hopcount);
986 	metric(rmx_mtu);
987 #undef metric
988 	out->rtm_rmx.rmx_expire = in->rt_rmx.rmx_expire ?
989 	    time_mono_to_wall(in->rt_rmx.rmx_expire) : 0;
990 }
991 
992 static int
993 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
994     struct rt_addrinfo *rtinfo)
995 {
996 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
997 	int i;
998 
999 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1000 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1001 			continue;
1002 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
1003 		RT_XADVANCE(cp, sa);
1004 	}
1005 
1006 	/*
1007 	 * Check for extra addresses specified, except RTM_GET asking
1008 	 * for interface info.
1009 	 */
1010 	if (rtmtype == RTM_GET) {
1011 		if (((rtinfo->rti_addrs &
1012 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0U << i)) != 0)
1013 			return 1;
1014 	} else if ((rtinfo->rti_addrs & (~0U << i)) != 0)
1015 		return 1;
1016 	/* Check for bad data length.  */
1017 	if (cp != cplim) {
1018 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
1019 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
1020 			/*
1021 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
1022 			 * We accept this for now for the sake of old
1023 			 * binaries or third party softwares.
1024 			 */
1025 			;
1026 		else
1027 			return 1;
1028 	}
1029 	return 0;
1030 }
1031 
1032 static int
1033 rt_getlen(int type)
1034 {
1035 #ifndef COMPAT_RTSOCK
1036 	CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
1037 	CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
1038 	CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
1039 	CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
1040 #endif
1041 
1042 	switch (type) {
1043 	case RTM_ODELADDR:
1044 	case RTM_ONEWADDR:
1045 	case RTM_OCHGADDR:
1046 #ifdef COMPAT_70
1047 		return sizeof(struct ifa_msghdr70);
1048 #else
1049 #ifdef DIAGNOSTIC
1050 		printf("RTM_ONEWADDR\n");
1051 #endif
1052 		return -1;
1053 #endif
1054 	case RTM_DELADDR:
1055 	case RTM_NEWADDR:
1056 	case RTM_CHGADDR:
1057 		return sizeof(struct ifa_xmsghdr);
1058 
1059 	case RTM_OOIFINFO:
1060 #ifdef COMPAT_14
1061 		return sizeof(struct if_msghdr14);
1062 #else
1063 #ifdef DIAGNOSTIC
1064 		printf("RTM_OOIFINFO\n");
1065 #endif
1066 		return -1;
1067 #endif
1068 	case RTM_OIFINFO:
1069 #ifdef COMPAT_50
1070 		return sizeof(struct if_msghdr50);
1071 #else
1072 #ifdef DIAGNOSTIC
1073 		printf("RTM_OIFINFO\n");
1074 #endif
1075 		return -1;
1076 #endif
1077 
1078 	case RTM_IFINFO:
1079 		return sizeof(struct if_xmsghdr);
1080 
1081 	case RTM_IFANNOUNCE:
1082 	case RTM_IEEE80211:
1083 		return sizeof(struct if_xannouncemsghdr);
1084 
1085 	default:
1086 		return sizeof(struct rt_xmsghdr);
1087 	}
1088 }
1089 
1090 
1091 struct mbuf *
1092 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
1093 {
1094 	struct rt_xmsghdr *rtm;
1095 	struct mbuf *m;
1096 	int i;
1097 	const struct sockaddr *sa;
1098 	int len, dlen;
1099 
1100 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1101 	if (m == NULL)
1102 		return m;
1103 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
1104 
1105 	if ((len = rt_getlen(type)) == -1)
1106 		goto out;
1107 	if (len > MHLEN + MLEN)
1108 		panic("%s: message too long", __func__);
1109 	else if (len > MHLEN) {
1110 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
1111 		if (m->m_next == NULL)
1112 			goto out;
1113 		MCLAIM(m->m_next, m->m_owner);
1114 		m->m_pkthdr.len = len;
1115 		m->m_len = MHLEN;
1116 		m->m_next->m_len = len - MHLEN;
1117 	} else {
1118 		m->m_pkthdr.len = m->m_len = len;
1119 	}
1120 	m_reset_rcvif(m);
1121 	m_copyback(m, 0, datalen, data);
1122 	if (len > datalen)
1123 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
1124 	rtm = mtod(m, struct rt_xmsghdr *);
1125 	for (i = 0; i < RTAX_MAX; i++) {
1126 		if ((sa = rtinfo->rti_info[i]) == NULL)
1127 			continue;
1128 		rtinfo->rti_addrs |= (1 << i);
1129 		dlen = RT_XROUNDUP(sa->sa_len);
1130 		m_copyback(m, len, sa->sa_len, sa);
1131 		if (dlen != sa->sa_len) {
1132 			/*
1133 			 * Up to 6 + 1 nul's since roundup is to
1134 			 * sizeof(uint64_t) (8 bytes)
1135 			 */
1136 			m_copyback(m, len + sa->sa_len,
1137 			    dlen - sa->sa_len, "\0\0\0\0\0\0");
1138 		}
1139 		len += dlen;
1140 	}
1141 	if (m->m_pkthdr.len != len)
1142 		goto out;
1143 	rtm->rtm_msglen = len;
1144 	rtm->rtm_version = RTM_XVERSION;
1145 	rtm->rtm_type = type;
1146 	return m;
1147 out:
1148 	m_freem(m);
1149 	return NULL;
1150 }
1151 
1152 /*
1153  * rt_msg2
1154  *
1155  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
1156  *		returns the length of the message in 'lenp'.
1157  *
1158  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
1159  *	the message
1160  * otherwise walkarg's w_needed is updated and if the user buffer is
1161  *	specified and w_needed indicates space exists the information is copied
1162  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
1163  *	if the allocation fails ENOBUFS is returned.
1164  */
1165 static int
1166 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1167 	int *lenp)
1168 {
1169 	int i;
1170 	int len, dlen, second_time = 0;
1171 	char *cp0, *cp = cpv;
1172 
1173 	rtinfo->rti_addrs = 0;
1174 again:
1175 	if ((len = rt_getlen(type)) == -1)
1176 		return EINVAL;
1177 
1178 	if ((cp0 = cp) != NULL)
1179 		cp += len;
1180 	for (i = 0; i < RTAX_MAX; i++) {
1181 		const struct sockaddr *sa;
1182 
1183 		if ((sa = rtinfo->rti_info[i]) == NULL)
1184 			continue;
1185 		rtinfo->rti_addrs |= (1 << i);
1186 		dlen = RT_XROUNDUP(sa->sa_len);
1187 		if (cp) {
1188 			int diff = dlen - sa->sa_len;
1189 			(void)memcpy(cp, sa, (size_t)sa->sa_len);
1190 			cp += sa->sa_len;
1191 			if (diff > 0) {
1192 				(void)memset(cp, 0, (size_t)diff);
1193 				cp += diff;
1194 			}
1195 		}
1196 		len += dlen;
1197 	}
1198 	if (cp == NULL && w != NULL && !second_time) {
1199 		struct rt_walkarg *rw = w;
1200 
1201 		rw->w_needed += len;
1202 		if (rw->w_needed <= 0 && rw->w_where) {
1203 			if (rw->w_tmemsize < len) {
1204 				if (rw->w_tmem)
1205 					free(rw->w_tmem, M_RTABLE);
1206 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
1207 				if (rw->w_tmem)
1208 					rw->w_tmemsize = len;
1209 				else
1210 					rw->w_tmemsize = 0;
1211 			}
1212 			if (rw->w_tmem) {
1213 				cp = rw->w_tmem;
1214 				second_time = 1;
1215 				goto again;
1216 			} else {
1217 				rw->w_tmemneeded = len;
1218 				return ENOBUFS;
1219 			}
1220 		}
1221 	}
1222 	if (cp) {
1223 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
1224 
1225 		rtm->rtm_version = RTM_XVERSION;
1226 		rtm->rtm_type = type;
1227 		rtm->rtm_msglen = len;
1228 	}
1229 	if (lenp)
1230 		*lenp = len;
1231 	return 0;
1232 }
1233 
1234 #ifndef COMPAT_RTSOCK
1235 int
1236 rt_msg3(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1237 	int *lenp)
1238 {
1239 	return rt_msg2(type, rtinfo, cpv, w, lenp);
1240 }
1241 #endif
1242 
1243 /*
1244  * This routine is called to generate a message from the routing
1245  * socket indicating that a redirect has occurred, a routing lookup
1246  * has failed, or that a protocol has detected timeouts to a particular
1247  * destination.
1248  */
1249 void
1250 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
1251     int error)
1252 {
1253 	struct rt_xmsghdr rtm;
1254 	struct mbuf *m;
1255 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1256 	struct rt_addrinfo info = *rtinfo;
1257 
1258 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
1259 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1260 		return;
1261 	memset(&rtm, 0, sizeof(rtm));
1262 	rtm.rtm_pid = curproc->p_pid;
1263 	rtm.rtm_flags = RTF_DONE | flags;
1264 	rtm.rtm_errno = error;
1265 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
1266 	if (m == NULL)
1267 		return;
1268 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1269 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1270 }
1271 
1272 /*
1273  * This routine is called to generate a message from the routing
1274  * socket indicating that the status of a network interface has changed.
1275  */
1276 void
1277 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
1278 {
1279 	struct if_xmsghdr ifm;
1280 	struct mbuf *m;
1281 	struct rt_addrinfo info;
1282 
1283 	COMPATCALL(rt_ifmsg, (ifp));
1284 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1285 		return;
1286 	(void)memset(&info, 0, sizeof(info));
1287 	(void)memset(&ifm, 0, sizeof(ifm));
1288 	ifm.ifm_index = ifp->if_index;
1289 	ifm.ifm_flags = ifp->if_flags;
1290 	ifm.ifm_data = ifp->if_data;
1291 	ifm.ifm_addrs = 0;
1292 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
1293 	if (m == NULL)
1294 		return;
1295 	COMPATNAME(route_enqueue)(m, 0);
1296 #ifdef COMPAT_14
1297 	compat_14_rt_oifmsg(ifp);
1298 #endif
1299 #ifdef COMPAT_50
1300 	compat_50_rt_oifmsg(ifp);
1301 #endif
1302 }
1303 
1304 #ifndef COMPAT_RTSOCK
1305 static int
1306 if_addrflags(struct ifaddr *ifa)
1307 {
1308 
1309 	switch (ifa->ifa_addr->sa_family) {
1310 #ifdef INET
1311 	case AF_INET:
1312 		return ((struct in_ifaddr *)ifa)->ia4_flags;
1313 #endif
1314 #ifdef INET6
1315 	case AF_INET6:
1316 		return ((struct in6_ifaddr *)ifa)->ia6_flags;
1317 #endif
1318 	default:
1319 		return 0;
1320 	}
1321 }
1322 #endif
1323 
1324 /*
1325  * This is called to generate messages from the routing socket
1326  * indicating a network interface has had addresses associated with it.
1327  * if we ever reverse the logic and replace messages TO the routing
1328  * socket indicate a request to configure interfaces, then it will
1329  * be unnecessary as the routing socket will automatically generate
1330  * copies of it.
1331  */
1332 void
1333 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
1334     struct rtentry *rt)
1335 {
1336 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
1337 	struct rt_addrinfo info;
1338 	const struct sockaddr *sa;
1339 	int pass;
1340 	struct mbuf *m;
1341 	struct ifnet *ifp;
1342 	struct rt_xmsghdr rtm;
1343 	struct ifa_xmsghdr ifam;
1344 	int ncmd;
1345 
1346 	KASSERT(ifa != NULL);
1347 	KASSERT(ifa->ifa_addr != NULL);
1348 	ifp = ifa->ifa_ifp;
1349 #ifdef SCTP
1350 	if (cmd == RTM_ADD) {
1351 		sctp_add_ip_address(ifa);
1352 	} else if (cmd == RTM_DELETE) {
1353 		sctp_delete_ip_address(ifa);
1354 	}
1355 #endif
1356 
1357 	COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
1358 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1359 		return;
1360 	for (pass = 1; pass < 3; pass++) {
1361 		memset(&info, 0, sizeof(info));
1362 		switch (cmdpass(cmd, pass)) {
1363 		case cmdpass(RTM_ADD, 1):
1364 		case cmdpass(RTM_CHANGE, 1):
1365 		case cmdpass(RTM_DELETE, 2):
1366 		case cmdpass(RTM_NEWADDR, 1):
1367 		case cmdpass(RTM_DELADDR, 1):
1368 		case cmdpass(RTM_CHGADDR, 1):
1369 			switch (cmd) {
1370 			case RTM_ADD:
1371 				ncmd = RTM_XNEWADDR;
1372 				break;
1373 			case RTM_DELETE:
1374 				ncmd = RTM_XDELADDR;
1375 				break;
1376 			case RTM_CHANGE:
1377 				ncmd = RTM_XCHGADDR;
1378 				break;
1379 			case RTM_NEWADDR:
1380 				ncmd = RTM_XNEWADDR;
1381 				break;
1382 			case RTM_DELADDR:
1383 				ncmd = RTM_XDELADDR;
1384 				break;
1385 			case RTM_CHGADDR:
1386 				ncmd = RTM_XCHGADDR;
1387 				break;
1388 			default:
1389 				panic("%s: unknown command %d", __func__, cmd);
1390 			}
1391 #ifdef COMPAT_70
1392 			compat_70_rt_newaddrmsg1(ncmd, ifa);
1393 #endif
1394 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1395 			KASSERT(ifp->if_dl != NULL);
1396 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1397 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1398 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1399 			memset(&ifam, 0, sizeof(ifam));
1400 			ifam.ifam_index = ifp->if_index;
1401 			ifam.ifam_metric = ifa->ifa_metric;
1402 			ifam.ifam_flags = ifa->ifa_flags;
1403 #ifndef COMPAT_RTSOCK
1404 			ifam.ifam_pid = curproc->p_pid;
1405 			ifam.ifam_addrflags = if_addrflags(ifa);
1406 #endif
1407 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
1408 			if (m == NULL)
1409 				continue;
1410 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
1411 			    info.rti_addrs;
1412 			break;
1413 		case cmdpass(RTM_ADD, 2):
1414 		case cmdpass(RTM_CHANGE, 2):
1415 		case cmdpass(RTM_DELETE, 1):
1416 			if (rt == NULL)
1417 				continue;
1418 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1419 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
1420 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1421 			memset(&rtm, 0, sizeof(rtm));
1422 			rtm.rtm_pid = curproc->p_pid;
1423 			rtm.rtm_index = ifp->if_index;
1424 			rtm.rtm_flags |= rt->rt_flags;
1425 			rtm.rtm_errno = error;
1426 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
1427 			if (m == NULL)
1428 				continue;
1429 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1430 			break;
1431 		default:
1432 			continue;
1433 		}
1434 #ifdef DIAGNOSTIC
1435 		if (m == NULL)
1436 			panic("%s: called with wrong command", __func__);
1437 #endif
1438 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1439 	}
1440 #undef cmdpass
1441 
1442 }
1443 
1444 static struct mbuf *
1445 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1446     struct rt_addrinfo *info)
1447 {
1448 	struct if_xannouncemsghdr ifan;
1449 
1450 	memset(info, 0, sizeof(*info));
1451 	memset(&ifan, 0, sizeof(ifan));
1452 	ifan.ifan_index = ifp->if_index;
1453 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
1454 	ifan.ifan_what = what;
1455 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
1456 }
1457 
1458 /*
1459  * This is called to generate routing socket messages indicating
1460  * network interface arrival and departure.
1461  */
1462 void
1463 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
1464 {
1465 	struct mbuf *m;
1466 	struct rt_addrinfo info;
1467 
1468 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
1469 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1470 		return;
1471 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1472 	if (m == NULL)
1473 		return;
1474 	COMPATNAME(route_enqueue)(m, 0);
1475 }
1476 
1477 /*
1478  * This is called to generate routing socket messages indicating
1479  * IEEE80211 wireless events.
1480  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1481  */
1482 void
1483 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
1484 	size_t data_len)
1485 {
1486 	struct mbuf *m;
1487 	struct rt_addrinfo info;
1488 
1489 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
1490 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1491 		return;
1492 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1493 	if (m == NULL)
1494 		return;
1495 	/*
1496 	 * Append the ieee80211 data.  Try to stick it in the
1497 	 * mbuf containing the ifannounce msg; otherwise allocate
1498 	 * a new mbuf and append.
1499 	 *
1500 	 * NB: we assume m is a single mbuf.
1501 	 */
1502 	if (data_len > M_TRAILINGSPACE(m)) {
1503 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1504 		if (n == NULL) {
1505 			m_freem(m);
1506 			return;
1507 		}
1508 		(void)memcpy(mtod(n, void *), data, data_len);
1509 		n->m_len = data_len;
1510 		m->m_next = n;
1511 	} else if (data_len > 0) {
1512 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
1513 		m->m_len += data_len;
1514 	}
1515 	if (m->m_flags & M_PKTHDR)
1516 		m->m_pkthdr.len += data_len;
1517 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
1518 	COMPATNAME(route_enqueue)(m, 0);
1519 }
1520 
1521 /*
1522  * This is used in dumping the kernel table via sysctl().
1523  */
1524 static int
1525 sysctl_dumpentry(struct rtentry *rt, void *v)
1526 {
1527 	struct rt_walkarg *w = v;
1528 	int error = 0, size;
1529 	struct rt_addrinfo info;
1530 
1531 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1532 		return 0;
1533 	memset(&info, 0, sizeof(info));
1534 	info.rti_info[RTAX_DST] = rt_getkey(rt);
1535 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1536 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1537 	info.rti_info[RTAX_TAG] = rt_gettag(rt);
1538 	if (rt->rt_ifp) {
1539 		const struct ifaddr *rtifa;
1540 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
1541 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
1542 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
1543 		 * --dyoung
1544 		 */
1545 		rtifa = rt_get_ifa(rt);
1546 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
1547 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1548 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
1549 	}
1550 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
1551 		return error;
1552 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1553 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
1554 
1555 		rtm->rtm_flags = rt->rt_flags;
1556 		rtm->rtm_use = rt->rt_use;
1557 		rtm_setmetrics(rt, rtm);
1558 		KASSERT(rt->rt_ifp != NULL);
1559 		rtm->rtm_index = rt->rt_ifp->if_index;
1560 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1561 		rtm->rtm_addrs = info.rti_addrs;
1562 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1563 			w->w_where = NULL;
1564 		else
1565 			w->w_where = (char *)w->w_where + size;
1566 	}
1567 	return error;
1568 }
1569 
1570 static int
1571 sysctl_iflist_if(struct ifnet *ifp, struct rt_walkarg *w,
1572     struct rt_addrinfo *info, size_t len)
1573 {
1574 	struct if_xmsghdr *ifm;
1575 	int error;
1576 
1577 	ifm = (struct if_xmsghdr *)w->w_tmem;
1578 	ifm->ifm_index = ifp->if_index;
1579 	ifm->ifm_flags = ifp->if_flags;
1580 	ifm->ifm_data = ifp->if_data;
1581 	ifm->ifm_addrs = info->rti_addrs;
1582 	if ((error = copyout(ifm, w->w_where, len)) == 0)
1583 		w->w_where = (char *)w->w_where + len;
1584 	return error;
1585 }
1586 
1587 static int
1588 sysctl_iflist_addr(struct rt_walkarg *w, struct ifaddr *ifa,
1589      struct rt_addrinfo *info)
1590 {
1591 	int len, error;
1592 
1593 	if ((error = rt_msg2(RTM_XNEWADDR, info, 0, w, &len)))
1594 		return error;
1595 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1596 		struct ifa_xmsghdr *ifam;
1597 
1598 		ifam = (struct ifa_xmsghdr *)w->w_tmem;
1599 		ifam->ifam_index = ifa->ifa_ifp->if_index;
1600 		ifam->ifam_flags = ifa->ifa_flags;
1601 		ifam->ifam_metric = ifa->ifa_metric;
1602 		ifam->ifam_addrs = info->rti_addrs;
1603 #ifndef COMPAT_RTSOCK
1604 		ifam->ifam_pid = 0;
1605 		ifam->ifam_addrflags = if_addrflags(ifa);
1606 #endif
1607 		if ((error = copyout(w->w_tmem, w->w_where, len)) == 0)
1608 			w->w_where = (char *)w->w_where + len;
1609 	}
1610 	return error;
1611 }
1612 
1613 static int
1614 sysctl_iflist(int af, struct rt_walkarg *w, int type)
1615 {
1616 	struct ifnet *ifp;
1617 	struct ifaddr *ifa;
1618 	struct	rt_addrinfo info;
1619 	int	cmd, len, error = 0;
1620 	int	(*iflist_if)(struct ifnet *, struct rt_walkarg *,
1621 			     struct rt_addrinfo *, size_t);
1622 	int	(*iflist_addr)(struct rt_walkarg *, struct ifaddr *,
1623 			       struct rt_addrinfo *);
1624 	int s;
1625 	struct psref psref;
1626 	int bound = curlwp_bind();
1627 
1628 	switch (type) {
1629 	case NET_RT_IFLIST:
1630 		cmd = RTM_IFINFO;
1631 		iflist_if = sysctl_iflist_if;
1632 		iflist_addr = sysctl_iflist_addr;
1633 		break;
1634 #ifdef COMPAT_14
1635 	case NET_RT_OOOIFLIST:
1636 		cmd = RTM_OOIFINFO;
1637 		iflist_if = compat_14_iflist;
1638 		iflist_addr = compat_70_iflist_addr;
1639 		break;
1640 #endif
1641 #ifdef COMPAT_50
1642 	case NET_RT_OOIFLIST:
1643 		cmd = RTM_OIFINFO;
1644 		iflist_if = compat_50_iflist;
1645 		iflist_addr = compat_70_iflist_addr;
1646 		break;
1647 #endif
1648 #ifdef COMPAT_70
1649 	case NET_RT_OIFLIST:
1650 		cmd = RTM_IFINFO;
1651 		iflist_if = sysctl_iflist_if;
1652 		iflist_addr = compat_70_iflist_addr;
1653 		break;
1654 #endif
1655 	default:
1656 #ifdef DIAGNOSTIC
1657 		printf("sysctl_iflist\n");
1658 #endif
1659 		return EINVAL;
1660 	}
1661 
1662 	memset(&info, 0, sizeof(info));
1663 
1664 	s = pserialize_read_enter();
1665 	IFNET_READER_FOREACH(ifp) {
1666 		if (w->w_arg && w->w_arg != ifp->if_index)
1667 			continue;
1668 		if (IFADDR_READER_EMPTY(ifp))
1669 			continue;
1670 
1671 		psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class);
1672 		pserialize_read_exit(s);
1673 
1674 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1675 		if ((error = rt_msg2(cmd, &info, NULL, w, &len)) != 0)
1676 			goto release_exit;
1677 		info.rti_info[RTAX_IFP] = NULL;
1678 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1679 			if ((error = iflist_if(ifp, w, &info, len)) != 0)
1680 				goto release_exit;
1681 		}
1682 		IFADDR_READER_FOREACH(ifa, ifp) {
1683 			if (af && af != ifa->ifa_addr->sa_family)
1684 				continue;
1685 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1686 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1687 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1688 			if ((error = iflist_addr(w, ifa, &info)) != 0)
1689 				goto release_exit;
1690 		}
1691 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1692 		    info.rti_info[RTAX_BRD] = NULL;
1693 
1694 		s = pserialize_read_enter();
1695 		psref_release(&psref, &ifp->if_psref, ifnet_psref_class);
1696 	}
1697 	pserialize_read_exit(s);
1698 	curlwp_bindx(bound);
1699 
1700 	return 0;
1701 
1702 release_exit:
1703 	psref_release(&psref, &ifp->if_psref, ifnet_psref_class);
1704 	curlwp_bindx(bound);
1705 	return error;
1706 }
1707 
1708 static int
1709 sysctl_rtable(SYSCTLFN_ARGS)
1710 {
1711 	void 	*where = oldp;
1712 	size_t	*given = oldlenp;
1713 	int	i, s, error = EINVAL;
1714 	u_char  af;
1715 	struct	rt_walkarg w;
1716 
1717 	if (namelen == 1 && name[0] == CTL_QUERY)
1718 		return sysctl_query(SYSCTLFN_CALL(rnode));
1719 
1720 	if (newp)
1721 		return EPERM;
1722 	if (namelen != 3)
1723 		return EINVAL;
1724 	af = name[0];
1725 	w.w_tmemneeded = 0;
1726 	w.w_tmemsize = 0;
1727 	w.w_tmem = NULL;
1728 again:
1729 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1730 	if (w.w_tmemneeded) {
1731 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
1732 		w.w_tmemsize = w.w_tmemneeded;
1733 		w.w_tmemneeded = 0;
1734 	}
1735 	w.w_op = name[1];
1736 	w.w_arg = name[2];
1737 	w.w_given = *given;
1738 	w.w_needed = 0 - w.w_given;
1739 	w.w_where = where;
1740 
1741 	s = splsoftnet();
1742 	switch (w.w_op) {
1743 
1744 	case NET_RT_DUMP:
1745 	case NET_RT_FLAGS:
1746 #ifdef INET
1747 		/*
1748 		 * take care of llinfo entries, the caller must
1749 		 * specify an AF
1750 		 */
1751 		if (w.w_op == NET_RT_FLAGS &&
1752 		    (w.w_arg == 0 || w.w_arg & RTF_LLDATA)) {
1753 			if (af != 0)
1754 				error = lltable_sysctl_dumparp(af, &w);
1755 			else
1756 				error = EINVAL;
1757 			break;
1758 		}
1759 #endif /* INET */
1760 
1761 		for (i = 1; i <= AF_MAX; i++)
1762 			if ((af == 0 || af == i) &&
1763 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
1764 				break;
1765 		break;
1766 
1767 #ifdef COMPAT_14
1768 	case NET_RT_OOOIFLIST:
1769 		error = sysctl_iflist(af, &w, w.w_op);
1770 		break;
1771 #endif
1772 #ifdef COMPAT_50
1773 	case NET_RT_OOIFLIST:
1774 		error = sysctl_iflist(af, &w, w.w_op);
1775 		break;
1776 #endif
1777 #ifdef COMPAT_70
1778 	case NET_RT_OIFLIST:
1779 		error = sysctl_iflist(af, &w, w.w_op);
1780 		break;
1781 #endif
1782 	case NET_RT_IFLIST:
1783 		error = sysctl_iflist(af, &w, w.w_op);
1784 		break;
1785 	}
1786 	splx(s);
1787 
1788 	/* check to see if we couldn't allocate memory with NOWAIT */
1789 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1790 		goto again;
1791 
1792 	if (w.w_tmem)
1793 		free(w.w_tmem, M_RTABLE);
1794 	w.w_needed += w.w_given;
1795 	if (where) {
1796 		*given = (char *)w.w_where - (char *)where;
1797 		if (*given < w.w_needed)
1798 			return ENOMEM;
1799 	} else {
1800 		*given = (11 * w.w_needed) / 10;
1801 	}
1802 	return error;
1803 }
1804 
1805 /*
1806  * Routing message software interrupt routine
1807  */
1808 static void
1809 COMPATNAME(route_intr)(void *cookie)
1810 {
1811 	struct sockproto proto = { .sp_family = PF_XROUTE, };
1812 	struct route_info * const ri = &COMPATNAME(route_info);
1813 	struct mbuf *m;
1814 
1815 	mutex_enter(softnet_lock);
1816 	KERNEL_LOCK(1, NULL);
1817 	for (;;) {
1818 		IFQ_LOCK(&ri->ri_intrq);
1819 		IF_DEQUEUE(&ri->ri_intrq, m);
1820 		IFQ_UNLOCK(&ri->ri_intrq);
1821 		if (m == NULL)
1822 			break;
1823 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
1824 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst);
1825 	}
1826 	KERNEL_UNLOCK_ONE(NULL);
1827 	mutex_exit(softnet_lock);
1828 }
1829 
1830 /*
1831  * Enqueue a message to the software interrupt routine.
1832  */
1833 void
1834 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
1835 {
1836 	struct route_info * const ri = &COMPATNAME(route_info);
1837 	int wasempty;
1838 
1839 	IFQ_LOCK(&ri->ri_intrq);
1840 	if (IF_QFULL(&ri->ri_intrq)) {
1841 		IF_DROP(&ri->ri_intrq);
1842 		IFQ_UNLOCK(&ri->ri_intrq);
1843 		m_freem(m);
1844 	} else {
1845 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
1846 		M_SETCTX(m, (uintptr_t)family);
1847 		IF_ENQUEUE(&ri->ri_intrq, m);
1848 		IFQ_UNLOCK(&ri->ri_intrq);
1849 		if (wasempty) {
1850 			kpreempt_disable();
1851 			softint_schedule(ri->ri_sih);
1852 			kpreempt_enable();
1853 		}
1854 	}
1855 }
1856 
1857 static void
1858 COMPATNAME(route_init)(void)
1859 {
1860 	struct route_info * const ri = &COMPATNAME(route_info);
1861 
1862 #ifndef COMPAT_RTSOCK
1863 	rt_init();
1864 #endif
1865 
1866 	sysctl_net_route_setup(NULL);
1867 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
1868 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
1869 	    COMPATNAME(route_intr), NULL);
1870 	IFQ_LOCK_INIT(&ri->ri_intrq);
1871 }
1872 
1873 /*
1874  * Definitions of protocols supported in the ROUTE domain.
1875  */
1876 #ifndef COMPAT_RTSOCK
1877 PR_WRAP_USRREQS(route);
1878 #else
1879 PR_WRAP_USRREQS(compat_50_route);
1880 #endif
1881 
1882 static const struct pr_usrreqs route_usrreqs = {
1883 	.pr_attach	= COMPATNAME(route_attach_wrapper),
1884 	.pr_detach	= COMPATNAME(route_detach_wrapper),
1885 	.pr_accept	= COMPATNAME(route_accept_wrapper),
1886 	.pr_bind	= COMPATNAME(route_bind_wrapper),
1887 	.pr_listen	= COMPATNAME(route_listen_wrapper),
1888 	.pr_connect	= COMPATNAME(route_connect_wrapper),
1889 	.pr_connect2	= COMPATNAME(route_connect2_wrapper),
1890 	.pr_disconnect	= COMPATNAME(route_disconnect_wrapper),
1891 	.pr_shutdown	= COMPATNAME(route_shutdown_wrapper),
1892 	.pr_abort	= COMPATNAME(route_abort_wrapper),
1893 	.pr_ioctl	= COMPATNAME(route_ioctl_wrapper),
1894 	.pr_stat	= COMPATNAME(route_stat_wrapper),
1895 	.pr_peeraddr	= COMPATNAME(route_peeraddr_wrapper),
1896 	.pr_sockaddr	= COMPATNAME(route_sockaddr_wrapper),
1897 	.pr_rcvd	= COMPATNAME(route_rcvd_wrapper),
1898 	.pr_recvoob	= COMPATNAME(route_recvoob_wrapper),
1899 	.pr_send	= COMPATNAME(route_send_wrapper),
1900 	.pr_sendoob	= COMPATNAME(route_sendoob_wrapper),
1901 	.pr_purgeif	= COMPATNAME(route_purgeif_wrapper),
1902 };
1903 
1904 static const struct protosw COMPATNAME(route_protosw)[] = {
1905 	{
1906 		.pr_type = SOCK_RAW,
1907 		.pr_domain = &COMPATNAME(routedomain),
1908 		.pr_flags = PR_ATOMIC|PR_ADDR,
1909 		.pr_input = raw_input,
1910 		.pr_ctlinput = raw_ctlinput,
1911 		.pr_usrreqs = &route_usrreqs,
1912 		.pr_init = raw_init,
1913 	},
1914 };
1915 
1916 struct domain COMPATNAME(routedomain) = {
1917 	.dom_family = PF_XROUTE,
1918 	.dom_name = DOMAINNAME,
1919 	.dom_init = COMPATNAME(route_init),
1920 	.dom_protosw = COMPATNAME(route_protosw),
1921 	.dom_protoswNPROTOSW =
1922 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
1923 };
1924 
1925 static void
1926 sysctl_net_route_setup(struct sysctllog **clog)
1927 {
1928 	const struct sysctlnode *rnode = NULL;
1929 
1930 	sysctl_createv(clog, 0, NULL, &rnode,
1931 		       CTLFLAG_PERMANENT,
1932 		       CTLTYPE_NODE, DOMAINNAME,
1933 		       SYSCTL_DESCR("PF_ROUTE information"),
1934 		       NULL, 0, NULL, 0,
1935 		       CTL_NET, PF_XROUTE, CTL_EOL);
1936 
1937 	sysctl_createv(clog, 0, NULL, NULL,
1938 		       CTLFLAG_PERMANENT,
1939 		       CTLTYPE_NODE, "rtable",
1940 		       SYSCTL_DESCR("Routing table information"),
1941 		       sysctl_rtable, 0, NULL, 0,
1942 		       CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
1943 
1944 	sysctl_createv(clog, 0, &rnode, NULL,
1945 		       CTLFLAG_PERMANENT,
1946 		       CTLTYPE_STRUCT, "stats",
1947 		       SYSCTL_DESCR("Routing statistics"),
1948 		       NULL, 0, &rtstat, sizeof(rtstat),
1949 		       CTL_CREATE, CTL_EOL);
1950 }
1951