xref: /netbsd-src/sys/net/rtsock.c (revision f21b7d7f2cbdd5c14b3882c4e8a3d43580d460a6)
1 /*	$NetBSD: rtsock.c,v 1.197 2016/10/03 11:06:06 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 1.197 2016/10/03 11:06:06 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 /*ARGSUSED*/
597 int
598 COMPATNAME(route_output)(struct mbuf *m, struct socket *so)
599 {
600 	struct sockproto proto = { .sp_family = PF_XROUTE, };
601 	struct rt_xmsghdr *rtm = NULL;
602 	struct rt_xmsghdr *old_rtm = NULL, *new_rtm = NULL;
603 	struct rtentry *rt = NULL;
604 	struct rtentry *saved_nrt = NULL;
605 	struct rt_addrinfo info;
606 	int len, error = 0;
607 	struct ifnet *ifp = NULL;
608 	struct ifaddr *ifa = NULL;
609 	sa_family_t family;
610 	struct sockaddr_dl sdl;
611 	struct psref psref;
612 	int bound = curlwp_bind();
613 
614 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
615 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
616 	   (m = m_pullup(m, sizeof(int32_t))) == NULL)) {
617 		error = ENOBUFS;
618 		goto out;
619 	}
620 	if ((m->m_flags & M_PKTHDR) == 0)
621 		panic("%s", __func__);
622 	len = m->m_pkthdr.len;
623 	if (len < sizeof(*rtm) ||
624 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
625 		info.rti_info[RTAX_DST] = NULL;
626 		senderr(EINVAL);
627 	}
628 	R_Malloc(rtm, struct rt_xmsghdr *, len);
629 	if (rtm == NULL) {
630 		info.rti_info[RTAX_DST] = NULL;
631 		senderr(ENOBUFS);
632 	}
633 	m_copydata(m, 0, len, rtm);
634 	if (rtm->rtm_version != RTM_XVERSION) {
635 		info.rti_info[RTAX_DST] = NULL;
636 		senderr(EPROTONOSUPPORT);
637 	}
638 	rtm->rtm_pid = curproc->p_pid;
639 	memset(&info, 0, sizeof(info));
640 	info.rti_addrs = rtm->rtm_addrs;
641 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
642 	    &info)) {
643 		senderr(EINVAL);
644 	}
645 	info.rti_flags = rtm->rtm_flags;
646 #ifdef RTSOCK_DEBUG
647 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
648 		char abuf[INET_ADDRSTRLEN];
649 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
650 		    RT_IN_PRINT(&info, abuf, RTAX_DST));
651 	}
652 #endif /* RTSOCK_DEBUG */
653 	if (info.rti_info[RTAX_DST] == NULL ||
654 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
655 		senderr(EINVAL);
656 	}
657 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
658 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
659 		senderr(EINVAL);
660 	}
661 
662 	/*
663 	 * Verify that the caller has the appropriate privilege; RTM_GET
664 	 * is the only operation the non-superuser is allowed.
665 	 */
666 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
667 	    0, rtm, NULL, NULL) != 0)
668 		senderr(EACCES);
669 
670 	switch (rtm->rtm_type) {
671 
672 	case RTM_ADD:
673 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
674 			senderr(EINVAL);
675 		}
676 #ifdef INET
677 		/* support for new ARP code with keeping backcompat */
678 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) {
679 			const struct sockaddr_dl *sdlp =
680 			    satocsdl(info.rti_info[RTAX_GATEWAY]);
681 
682 			/* Allow routing requests by interface index */
683 			if (sdlp->sdl_nlen == 0 && sdlp->sdl_alen == 0
684 			    && sdlp->sdl_slen == 0)
685 				goto fallback;
686 			/*
687 			 * Old arp binaries don't set the sdl_index
688 			 * so we have to complement it.
689 			 */
690 			int sdl_index = sdlp->sdl_index;
691 			if (sdl_index == 0) {
692 				error = route_get_sdl_index(&info, &sdl_index);
693 				if (error != 0)
694 					goto fallback;
695 			} else if (
696 			    info.rti_info[RTAX_DST]->sa_family == AF_INET) {
697 				/*
698 				 * XXX workaround for SIN_PROXY case; proxy arp
699 				 * entry should be in an interface that has
700 				 * a network route including the destination,
701 				 * not a local (link) route that may not be a
702 				 * desired place, for example a tap.
703 				 */
704 				const struct sockaddr_inarp *sina =
705 				    (const struct sockaddr_inarp *)
706 				    info.rti_info[RTAX_DST];
707 				if (sina->sin_other & SIN_PROXY) {
708 					error = route_get_sdl_index(&info,
709 					    &sdl_index);
710 					if (error != 0)
711 						goto fallback;
712 				}
713 			}
714 			error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
715 			    rtm->rtm_rmx.rmx_expire, &info, sdl_index);
716 			break;
717 		}
718 	fallback:
719 #endif /* INET */
720 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
721 		if (error == 0) {
722 			rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
723 			rtfree(saved_nrt);
724 		}
725 		break;
726 
727 	case RTM_DELETE:
728 #ifdef INET
729 		/* support for new ARP code */
730 		if (info.rti_info[RTAX_GATEWAY] &&
731 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
732 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
733 			error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
734 			    rtm->rtm_rmx.rmx_expire, &info, 0);
735 			break;
736 		}
737 #endif /* INET */
738 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
739 		if (error != 0)
740 			break;
741 
742 		rt = saved_nrt;
743 		info.rti_info[RTAX_DST] = rt_getkey(rt);
744 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
745 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
746 		info.rti_info[RTAX_TAG] = rt_gettag(rt);
747 		error = route_output_report(rt, &info, rtm, &new_rtm);
748 		if (error)
749 			senderr(error);
750 		if (new_rtm != NULL) {
751 			old_rtm = rtm;
752 			rtm = new_rtm;
753 		}
754 		break;
755 
756 	case RTM_GET:
757 	case RTM_CHANGE:
758 	case RTM_LOCK:
759                 /* XXX This will mask info.rti_info[RTAX_DST] with
760 		 * info.rti_info[RTAX_NETMASK] before
761                  * searching.  It did not used to do that.  --dyoung
762 		 */
763 		rt = NULL;
764 		error = rtrequest1(RTM_GET, &info, &rt);
765 		if (error != 0)
766 			senderr(error);
767 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
768 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
769 			    info.rti_info[RTAX_DST]->sa_len) != 0)
770 				senderr(ESRCH);
771 			if (info.rti_info[RTAX_NETMASK] == NULL &&
772 			    rt_mask(rt) != NULL)
773 				senderr(ETOOMANYREFS);
774 		}
775 
776 		/*
777 		 * XXX if arp/ndp requests an L2 entry, we have to obtain
778 		 * it from lltable while for the route command we have to
779 		 * return a route as it is. How to distinguish them?
780 		 * For newer arp/ndp, RTF_LLDATA flag set by arp/ndp
781 		 * indicates an L2 entry is requested. For old arp/ndp
782 		 * binaries, we check RTF_UP flag is NOT set; it works
783 		 * by the fact that arp/ndp don't set it while the route
784 		 * command sets it.
785 		 */
786 		if (((rtm->rtm_flags & RTF_LLDATA) != 0 ||
787 		     (rtm->rtm_flags & RTF_UP) == 0) &&
788 		    rtm->rtm_type == RTM_GET &&
789 		    sockaddr_cmp(rt_getkey(rt), info.rti_info[RTAX_DST]) != 0) {
790 			int ll_flags = 0;
791 			route_get_sdl(rt->rt_ifp, info.rti_info[RTAX_DST], &sdl,
792 			    &ll_flags);
793 			info.rti_info[RTAX_GATEWAY] = sstocsa(&sdl);
794 			error = route_output_report(rt, &info, rtm, &new_rtm);
795 			if (error)
796 				senderr(error);
797 			if (new_rtm != NULL) {
798 				old_rtm = rtm;
799 				rtm = new_rtm;
800 			}
801 			rtm->rtm_flags |= RTF_LLDATA;
802 			rtm->rtm_flags |= (ll_flags & LLE_STATIC) ? RTF_STATIC : 0;
803 			break;
804 		}
805 
806 		switch (rtm->rtm_type) {
807 		case RTM_GET:
808 			info.rti_info[RTAX_DST] = rt_getkey(rt);
809 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
810 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
811 			info.rti_info[RTAX_TAG] = rt_gettag(rt);
812 			error = route_output_report(rt, &info, rtm, &new_rtm);
813 			if (error)
814 				senderr(error);
815 			if (new_rtm != NULL) {
816 				old_rtm = rtm;
817 				rtm = new_rtm;
818 			}
819 			break;
820 
821 		case RTM_CHANGE: {
822 			struct ifnet *_ifp;
823 			struct ifaddr *_ifa;
824 			struct psref _psref, psref_ifp;
825 			/*
826 			 * new gateway could require new ifaddr, ifp;
827 			 * flags may also be different; ifp may be specified
828 			 * by ll sockaddr when protocol address is ambiguous
829 			 */
830 			_ifp = rt_getifp(&info, &psref_ifp);
831 			ifa = rt_getifa(&info, &psref);
832 			if (ifa == NULL) {
833 				if_put(_ifp, &psref_ifp);
834 				senderr(ENETUNREACH);
835 			}
836 			if (info.rti_info[RTAX_GATEWAY]) {
837 				error = rt_setgate(rt,
838 				    info.rti_info[RTAX_GATEWAY]);
839 				if (error != 0) {
840 					if_put(_ifp, &psref_ifp);
841 					senderr(error);
842 				}
843 			}
844 			if (info.rti_info[RTAX_TAG]) {
845 				const struct sockaddr *tag;
846 				tag = rt_settag(rt, info.rti_info[RTAX_TAG]);
847 				if (tag == NULL) {
848 					if_put(_ifp, &psref_ifp);
849 					senderr(ENOBUFS);
850 				}
851 			}
852 			/* new gateway could require new ifaddr, ifp;
853 			   flags may also be different; ifp may be specified
854 			   by ll sockaddr when protocol address is ambiguous */
855 			_ifa = route_output_get_ifa(info, rt, &ifp, &_psref);
856 			if (_ifa != NULL) {
857 				ifa_release(ifa, &psref);
858 				ifa = _ifa;
859 			}
860 			if (ifa) {
861 				struct ifaddr *oifa = rt->rt_ifa;
862 				if (oifa != ifa) {
863 					if (oifa && oifa->ifa_rtrequest) {
864 						oifa->ifa_rtrequest(RTM_DELETE,
865 						    rt, &info);
866 					}
867 					rt_replace_ifa(rt, ifa);
868 					rt->rt_ifp = ifp;
869 				}
870 				if (_ifa == NULL)
871 					ifa_release(ifa, &psref);
872 			}
873 			ifa_release(_ifa, &_psref);
874 			if (ifp && rt->rt_ifp != ifp)
875 				rt->rt_ifp = ifp;
876 			rt_setmetrics(rtm->rtm_inits, rtm, rt);
877 			if (rt->rt_flags != info.rti_flags)
878 				rt->rt_flags = (info.rti_flags & ~PRESERVED_RTF)
879 				    | (rt->rt_flags & PRESERVED_RTF);
880 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
881 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
882 			if_put(_ifp, &psref_ifp);
883 			/*FALLTHROUGH*/
884 		    }
885 		case RTM_LOCK:
886 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
887 			rt->rt_rmx.rmx_locks |=
888 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
889 			break;
890 		}
891 		break;
892 
893 	default:
894 		senderr(EOPNOTSUPP);
895 	}
896 
897 flush:
898 	if (rtm) {
899 		if (error)
900 			rtm->rtm_errno = error;
901 		else
902 			rtm->rtm_flags |= RTF_DONE;
903 	}
904 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
905 	    0;
906 	/* We cannot free old_rtm until we have stopped using the
907 	 * pointers in info, some of which may point to sockaddrs
908 	 * in old_rtm.
909 	 */
910 	if (old_rtm != NULL)
911 		Free(old_rtm);
912 	if (rt)
913 		rtfree(rt);
914     {
915 	struct rawcb *rp = NULL;
916 	/*
917 	 * Check to see if we don't want our own messages.
918 	 */
919 	if ((so->so_options & SO_USELOOPBACK) == 0) {
920 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
921 			if (rtm)
922 				Free(rtm);
923 			m_freem(m);
924 			goto out;
925 		}
926 		/* There is another listener, so construct message */
927 		rp = sotorawcb(so);
928 	}
929 	if (rtm) {
930 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
931 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
932 			m_freem(m);
933 			m = NULL;
934 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
935 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
936 		Free(rtm);
937 	}
938 	if (rp)
939 		rp->rcb_proto.sp_family = 0; /* Avoid us */
940 	if (family)
941 		proto.sp_protocol = family;
942 	if (m)
943 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
944 		    &COMPATNAME(route_info).ri_dst);
945 	if (rp)
946 		rp->rcb_proto.sp_family = PF_XROUTE;
947     }
948 out:
949 	curlwp_bindx(bound);
950 	return error;
951 }
952 
953 static void
954 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
955 {
956 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
957 	metric(RTV_RPIPE, rmx_recvpipe);
958 	metric(RTV_SPIPE, rmx_sendpipe);
959 	metric(RTV_SSTHRESH, rmx_ssthresh);
960 	metric(RTV_RTT, rmx_rtt);
961 	metric(RTV_RTTVAR, rmx_rttvar);
962 	metric(RTV_HOPCOUNT, rmx_hopcount);
963 	metric(RTV_MTU, rmx_mtu);
964 #undef metric
965 	if (which & RTV_EXPIRE) {
966 		out->rt_rmx.rmx_expire = in->rtm_rmx.rmx_expire ?
967 		    time_wall_to_mono(in->rtm_rmx.rmx_expire) : 0;
968 	}
969 }
970 
971 static void
972 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
973 {
974 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
975 	metric(rmx_recvpipe);
976 	metric(rmx_sendpipe);
977 	metric(rmx_ssthresh);
978 	metric(rmx_rtt);
979 	metric(rmx_rttvar);
980 	metric(rmx_hopcount);
981 	metric(rmx_mtu);
982 #undef metric
983 	out->rtm_rmx.rmx_expire = in->rt_rmx.rmx_expire ?
984 	    time_mono_to_wall(in->rt_rmx.rmx_expire) : 0;
985 }
986 
987 static int
988 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
989     struct rt_addrinfo *rtinfo)
990 {
991 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
992 	int i;
993 
994 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
995 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
996 			continue;
997 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
998 		RT_XADVANCE(cp, sa);
999 	}
1000 
1001 	/*
1002 	 * Check for extra addresses specified, except RTM_GET asking
1003 	 * for interface info.
1004 	 */
1005 	if (rtmtype == RTM_GET) {
1006 		if (((rtinfo->rti_addrs &
1007 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0U << i)) != 0)
1008 			return 1;
1009 	} else if ((rtinfo->rti_addrs & (~0U << i)) != 0)
1010 		return 1;
1011 	/* Check for bad data length.  */
1012 	if (cp != cplim) {
1013 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
1014 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
1015 			/*
1016 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
1017 			 * We accept this for now for the sake of old
1018 			 * binaries or third party softwares.
1019 			 */
1020 			;
1021 		else
1022 			return 1;
1023 	}
1024 	return 0;
1025 }
1026 
1027 static int
1028 rt_getlen(int type)
1029 {
1030 #ifndef COMPAT_RTSOCK
1031 	CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
1032 	CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
1033 	CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
1034 	CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
1035 #endif
1036 
1037 	switch (type) {
1038 	case RTM_ODELADDR:
1039 	case RTM_ONEWADDR:
1040 	case RTM_OCHGADDR:
1041 #ifdef COMPAT_70
1042 		return sizeof(struct ifa_msghdr70);
1043 #else
1044 #ifdef DIAGNOSTIC
1045 		printf("RTM_ONEWADDR\n");
1046 #endif
1047 		return -1;
1048 #endif
1049 	case RTM_DELADDR:
1050 	case RTM_NEWADDR:
1051 	case RTM_CHGADDR:
1052 		return sizeof(struct ifa_xmsghdr);
1053 
1054 	case RTM_OOIFINFO:
1055 #ifdef COMPAT_14
1056 		return sizeof(struct if_msghdr14);
1057 #else
1058 #ifdef DIAGNOSTIC
1059 		printf("RTM_OOIFINFO\n");
1060 #endif
1061 		return -1;
1062 #endif
1063 	case RTM_OIFINFO:
1064 #ifdef COMPAT_50
1065 		return sizeof(struct if_msghdr50);
1066 #else
1067 #ifdef DIAGNOSTIC
1068 		printf("RTM_OIFINFO\n");
1069 #endif
1070 		return -1;
1071 #endif
1072 
1073 	case RTM_IFINFO:
1074 		return sizeof(struct if_xmsghdr);
1075 
1076 	case RTM_IFANNOUNCE:
1077 	case RTM_IEEE80211:
1078 		return sizeof(struct if_xannouncemsghdr);
1079 
1080 	default:
1081 		return sizeof(struct rt_xmsghdr);
1082 	}
1083 }
1084 
1085 
1086 struct mbuf *
1087 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
1088 {
1089 	struct rt_xmsghdr *rtm;
1090 	struct mbuf *m;
1091 	int i;
1092 	const struct sockaddr *sa;
1093 	int len, dlen;
1094 
1095 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1096 	if (m == NULL)
1097 		return m;
1098 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
1099 
1100 	if ((len = rt_getlen(type)) == -1)
1101 		goto out;
1102 	if (len > MHLEN + MLEN)
1103 		panic("%s: message too long", __func__);
1104 	else if (len > MHLEN) {
1105 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
1106 		if (m->m_next == NULL)
1107 			goto out;
1108 		MCLAIM(m->m_next, m->m_owner);
1109 		m->m_pkthdr.len = len;
1110 		m->m_len = MHLEN;
1111 		m->m_next->m_len = len - MHLEN;
1112 	} else {
1113 		m->m_pkthdr.len = m->m_len = len;
1114 	}
1115 	m_reset_rcvif(m);
1116 	m_copyback(m, 0, datalen, data);
1117 	if (len > datalen)
1118 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
1119 	rtm = mtod(m, struct rt_xmsghdr *);
1120 	for (i = 0; i < RTAX_MAX; i++) {
1121 		if ((sa = rtinfo->rti_info[i]) == NULL)
1122 			continue;
1123 		rtinfo->rti_addrs |= (1 << i);
1124 		dlen = RT_XROUNDUP(sa->sa_len);
1125 		m_copyback(m, len, sa->sa_len, sa);
1126 		if (dlen != sa->sa_len) {
1127 			/*
1128 			 * Up to 6 + 1 nul's since roundup is to
1129 			 * sizeof(uint64_t) (8 bytes)
1130 			 */
1131 			m_copyback(m, len + sa->sa_len,
1132 			    dlen - sa->sa_len, "\0\0\0\0\0\0");
1133 		}
1134 		len += dlen;
1135 	}
1136 	if (m->m_pkthdr.len != len)
1137 		goto out;
1138 	rtm->rtm_msglen = len;
1139 	rtm->rtm_version = RTM_XVERSION;
1140 	rtm->rtm_type = type;
1141 	return m;
1142 out:
1143 	m_freem(m);
1144 	return NULL;
1145 }
1146 
1147 /*
1148  * rt_msg2
1149  *
1150  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
1151  *		returns the length of the message in 'lenp'.
1152  *
1153  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
1154  *	the message
1155  * otherwise walkarg's w_needed is updated and if the user buffer is
1156  *	specified and w_needed indicates space exists the information is copied
1157  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
1158  *	if the allocation fails ENOBUFS is returned.
1159  */
1160 static int
1161 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1162 	int *lenp)
1163 {
1164 	int i;
1165 	int len, dlen, second_time = 0;
1166 	char *cp0, *cp = cpv;
1167 
1168 	rtinfo->rti_addrs = 0;
1169 again:
1170 	if ((len = rt_getlen(type)) == -1)
1171 		return EINVAL;
1172 
1173 	if ((cp0 = cp) != NULL)
1174 		cp += len;
1175 	for (i = 0; i < RTAX_MAX; i++) {
1176 		const struct sockaddr *sa;
1177 
1178 		if ((sa = rtinfo->rti_info[i]) == NULL)
1179 			continue;
1180 		rtinfo->rti_addrs |= (1 << i);
1181 		dlen = RT_XROUNDUP(sa->sa_len);
1182 		if (cp) {
1183 			int diff = dlen - sa->sa_len;
1184 			(void)memcpy(cp, sa, (size_t)sa->sa_len);
1185 			cp += sa->sa_len;
1186 			if (diff > 0) {
1187 				(void)memset(cp, 0, (size_t)diff);
1188 				cp += diff;
1189 			}
1190 		}
1191 		len += dlen;
1192 	}
1193 	if (cp == NULL && w != NULL && !second_time) {
1194 		struct rt_walkarg *rw = w;
1195 
1196 		rw->w_needed += len;
1197 		if (rw->w_needed <= 0 && rw->w_where) {
1198 			if (rw->w_tmemsize < len) {
1199 				if (rw->w_tmem)
1200 					free(rw->w_tmem, M_RTABLE);
1201 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
1202 				if (rw->w_tmem)
1203 					rw->w_tmemsize = len;
1204 				else
1205 					rw->w_tmemsize = 0;
1206 			}
1207 			if (rw->w_tmem) {
1208 				cp = rw->w_tmem;
1209 				second_time = 1;
1210 				goto again;
1211 			} else {
1212 				rw->w_tmemneeded = len;
1213 				return ENOBUFS;
1214 			}
1215 		}
1216 	}
1217 	if (cp) {
1218 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
1219 
1220 		rtm->rtm_version = RTM_XVERSION;
1221 		rtm->rtm_type = type;
1222 		rtm->rtm_msglen = len;
1223 	}
1224 	if (lenp)
1225 		*lenp = len;
1226 	return 0;
1227 }
1228 
1229 #ifndef COMPAT_RTSOCK
1230 int
1231 rt_msg3(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1232 	int *lenp)
1233 {
1234 	return rt_msg2(type, rtinfo, cpv, w, lenp);
1235 }
1236 #endif
1237 
1238 /*
1239  * This routine is called to generate a message from the routing
1240  * socket indicating that a redirect has occurred, a routing lookup
1241  * has failed, or that a protocol has detected timeouts to a particular
1242  * destination.
1243  */
1244 void
1245 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
1246     int error)
1247 {
1248 	struct rt_xmsghdr rtm;
1249 	struct mbuf *m;
1250 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1251 	struct rt_addrinfo info = *rtinfo;
1252 
1253 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
1254 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1255 		return;
1256 	memset(&rtm, 0, sizeof(rtm));
1257 	rtm.rtm_pid = curproc->p_pid;
1258 	rtm.rtm_flags = RTF_DONE | flags;
1259 	rtm.rtm_errno = error;
1260 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
1261 	if (m == NULL)
1262 		return;
1263 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1264 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1265 }
1266 
1267 /*
1268  * This routine is called to generate a message from the routing
1269  * socket indicating that the status of a network interface has changed.
1270  */
1271 void
1272 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
1273 {
1274 	struct if_xmsghdr ifm;
1275 	struct mbuf *m;
1276 	struct rt_addrinfo info;
1277 
1278 	COMPATCALL(rt_ifmsg, (ifp));
1279 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1280 		return;
1281 	(void)memset(&info, 0, sizeof(info));
1282 	(void)memset(&ifm, 0, sizeof(ifm));
1283 	ifm.ifm_index = ifp->if_index;
1284 	ifm.ifm_flags = ifp->if_flags;
1285 	ifm.ifm_data = ifp->if_data;
1286 	ifm.ifm_addrs = 0;
1287 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
1288 	if (m == NULL)
1289 		return;
1290 	COMPATNAME(route_enqueue)(m, 0);
1291 #ifdef COMPAT_14
1292 	compat_14_rt_oifmsg(ifp);
1293 #endif
1294 #ifdef COMPAT_50
1295 	compat_50_rt_oifmsg(ifp);
1296 #endif
1297 }
1298 
1299 #ifndef COMPAT_RTSOCK
1300 static int
1301 if_addrflags(struct ifaddr *ifa)
1302 {
1303 
1304 	switch (ifa->ifa_addr->sa_family) {
1305 #ifdef INET
1306 	case AF_INET:
1307 		return ((struct in_ifaddr *)ifa)->ia4_flags;
1308 #endif
1309 #ifdef INET6
1310 	case AF_INET6:
1311 		return ((struct in6_ifaddr *)ifa)->ia6_flags;
1312 #endif
1313 	default:
1314 		return 0;
1315 	}
1316 }
1317 #endif
1318 
1319 /*
1320  * This is called to generate messages from the routing socket
1321  * indicating a network interface has had addresses associated with it.
1322  * if we ever reverse the logic and replace messages TO the routing
1323  * socket indicate a request to configure interfaces, then it will
1324  * be unnecessary as the routing socket will automatically generate
1325  * copies of it.
1326  */
1327 void
1328 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
1329     struct rtentry *rt)
1330 {
1331 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
1332 	struct rt_addrinfo info;
1333 	const struct sockaddr *sa;
1334 	int pass;
1335 	struct mbuf *m;
1336 	struct ifnet *ifp;
1337 	struct rt_xmsghdr rtm;
1338 	struct ifa_xmsghdr ifam;
1339 	int ncmd;
1340 
1341 	KASSERT(ifa != NULL);
1342 	KASSERT(ifa->ifa_addr != NULL);
1343 	ifp = ifa->ifa_ifp;
1344 #ifdef SCTP
1345 	if (cmd == RTM_ADD) {
1346 		sctp_add_ip_address(ifa);
1347 	} else if (cmd == RTM_DELETE) {
1348 		sctp_delete_ip_address(ifa);
1349 	}
1350 #endif
1351 
1352 	COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
1353 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1354 		return;
1355 	for (pass = 1; pass < 3; pass++) {
1356 		memset(&info, 0, sizeof(info));
1357 		switch (cmdpass(cmd, pass)) {
1358 		case cmdpass(RTM_ADD, 1):
1359 		case cmdpass(RTM_CHANGE, 1):
1360 		case cmdpass(RTM_DELETE, 2):
1361 		case cmdpass(RTM_NEWADDR, 1):
1362 		case cmdpass(RTM_DELADDR, 1):
1363 		case cmdpass(RTM_CHGADDR, 1):
1364 			switch (cmd) {
1365 			case RTM_ADD:
1366 				ncmd = RTM_XNEWADDR;
1367 				break;
1368 			case RTM_DELETE:
1369 				ncmd = RTM_XDELADDR;
1370 				break;
1371 			case RTM_CHANGE:
1372 				ncmd = RTM_XCHGADDR;
1373 				break;
1374 			case RTM_NEWADDR:
1375 				ncmd = RTM_XNEWADDR;
1376 				break;
1377 			case RTM_DELADDR:
1378 				ncmd = RTM_XDELADDR;
1379 				break;
1380 			case RTM_CHGADDR:
1381 				ncmd = RTM_XCHGADDR;
1382 				break;
1383 			default:
1384 				panic("%s: unknown command %d", __func__, cmd);
1385 			}
1386 #ifdef COMPAT_70
1387 			compat_70_rt_newaddrmsg1(ncmd, ifa);
1388 #endif
1389 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1390 			KASSERT(ifp->if_dl != NULL);
1391 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1392 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1393 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1394 			memset(&ifam, 0, sizeof(ifam));
1395 			ifam.ifam_index = ifp->if_index;
1396 			ifam.ifam_metric = ifa->ifa_metric;
1397 			ifam.ifam_flags = ifa->ifa_flags;
1398 #ifndef COMPAT_RTSOCK
1399 			ifam.ifam_pid = curproc->p_pid;
1400 			ifam.ifam_addrflags = if_addrflags(ifa);
1401 #endif
1402 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
1403 			if (m == NULL)
1404 				continue;
1405 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
1406 			    info.rti_addrs;
1407 			break;
1408 		case cmdpass(RTM_ADD, 2):
1409 		case cmdpass(RTM_CHANGE, 2):
1410 		case cmdpass(RTM_DELETE, 1):
1411 			if (rt == NULL)
1412 				continue;
1413 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1414 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
1415 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1416 			memset(&rtm, 0, sizeof(rtm));
1417 			rtm.rtm_pid = curproc->p_pid;
1418 			rtm.rtm_index = ifp->if_index;
1419 			rtm.rtm_flags |= rt->rt_flags;
1420 			rtm.rtm_errno = error;
1421 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
1422 			if (m == NULL)
1423 				continue;
1424 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1425 			break;
1426 		default:
1427 			continue;
1428 		}
1429 #ifdef DIAGNOSTIC
1430 		if (m == NULL)
1431 			panic("%s: called with wrong command", __func__);
1432 #endif
1433 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1434 	}
1435 #undef cmdpass
1436 
1437 }
1438 
1439 static struct mbuf *
1440 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1441     struct rt_addrinfo *info)
1442 {
1443 	struct if_xannouncemsghdr ifan;
1444 
1445 	memset(info, 0, sizeof(*info));
1446 	memset(&ifan, 0, sizeof(ifan));
1447 	ifan.ifan_index = ifp->if_index;
1448 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
1449 	ifan.ifan_what = what;
1450 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
1451 }
1452 
1453 /*
1454  * This is called to generate routing socket messages indicating
1455  * network interface arrival and departure.
1456  */
1457 void
1458 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
1459 {
1460 	struct mbuf *m;
1461 	struct rt_addrinfo info;
1462 
1463 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
1464 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1465 		return;
1466 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1467 	if (m == NULL)
1468 		return;
1469 	COMPATNAME(route_enqueue)(m, 0);
1470 }
1471 
1472 /*
1473  * This is called to generate routing socket messages indicating
1474  * IEEE80211 wireless events.
1475  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1476  */
1477 void
1478 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
1479 	size_t data_len)
1480 {
1481 	struct mbuf *m;
1482 	struct rt_addrinfo info;
1483 
1484 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
1485 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1486 		return;
1487 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1488 	if (m == NULL)
1489 		return;
1490 	/*
1491 	 * Append the ieee80211 data.  Try to stick it in the
1492 	 * mbuf containing the ifannounce msg; otherwise allocate
1493 	 * a new mbuf and append.
1494 	 *
1495 	 * NB: we assume m is a single mbuf.
1496 	 */
1497 	if (data_len > M_TRAILINGSPACE(m)) {
1498 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1499 		if (n == NULL) {
1500 			m_freem(m);
1501 			return;
1502 		}
1503 		(void)memcpy(mtod(n, void *), data, data_len);
1504 		n->m_len = data_len;
1505 		m->m_next = n;
1506 	} else if (data_len > 0) {
1507 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
1508 		m->m_len += data_len;
1509 	}
1510 	if (m->m_flags & M_PKTHDR)
1511 		m->m_pkthdr.len += data_len;
1512 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
1513 	COMPATNAME(route_enqueue)(m, 0);
1514 }
1515 
1516 /*
1517  * This is used in dumping the kernel table via sysctl().
1518  */
1519 static int
1520 sysctl_dumpentry(struct rtentry *rt, void *v)
1521 {
1522 	struct rt_walkarg *w = v;
1523 	int error = 0, size;
1524 	struct rt_addrinfo info;
1525 
1526 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1527 		return 0;
1528 	memset(&info, 0, sizeof(info));
1529 	info.rti_info[RTAX_DST] = rt_getkey(rt);
1530 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1531 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1532 	info.rti_info[RTAX_TAG] = rt_gettag(rt);
1533 	if (rt->rt_ifp) {
1534 		const struct ifaddr *rtifa;
1535 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
1536 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
1537 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
1538 		 * --dyoung
1539 		 */
1540 		rtifa = rt_get_ifa(rt);
1541 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
1542 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1543 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
1544 	}
1545 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
1546 		return error;
1547 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1548 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
1549 
1550 		rtm->rtm_flags = rt->rt_flags;
1551 		rtm->rtm_use = rt->rt_use;
1552 		rtm_setmetrics(rt, rtm);
1553 		KASSERT(rt->rt_ifp != NULL);
1554 		rtm->rtm_index = rt->rt_ifp->if_index;
1555 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1556 		rtm->rtm_addrs = info.rti_addrs;
1557 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1558 			w->w_where = NULL;
1559 		else
1560 			w->w_where = (char *)w->w_where + size;
1561 	}
1562 	return error;
1563 }
1564 
1565 static int
1566 sysctl_iflist_if(struct ifnet *ifp, struct rt_walkarg *w,
1567     struct rt_addrinfo *info, size_t len)
1568 {
1569 	struct if_xmsghdr *ifm;
1570 	int error;
1571 
1572 	ifm = (struct if_xmsghdr *)w->w_tmem;
1573 	ifm->ifm_index = ifp->if_index;
1574 	ifm->ifm_flags = ifp->if_flags;
1575 	ifm->ifm_data = ifp->if_data;
1576 	ifm->ifm_addrs = info->rti_addrs;
1577 	if ((error = copyout(ifm, w->w_where, len)) == 0)
1578 		w->w_where = (char *)w->w_where + len;
1579 	return error;
1580 }
1581 
1582 static int
1583 sysctl_iflist_addr(struct rt_walkarg *w, struct ifaddr *ifa,
1584      struct rt_addrinfo *info)
1585 {
1586 	int len, error;
1587 
1588 	if ((error = rt_msg2(RTM_XNEWADDR, info, 0, w, &len)))
1589 		return error;
1590 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1591 		struct ifa_xmsghdr *ifam;
1592 
1593 		ifam = (struct ifa_xmsghdr *)w->w_tmem;
1594 		ifam->ifam_index = ifa->ifa_ifp->if_index;
1595 		ifam->ifam_flags = ifa->ifa_flags;
1596 		ifam->ifam_metric = ifa->ifa_metric;
1597 		ifam->ifam_addrs = info->rti_addrs;
1598 #ifndef COMPAT_RTSOCK
1599 		ifam->ifam_pid = 0;
1600 		ifam->ifam_addrflags = if_addrflags(ifa);
1601 #endif
1602 		if ((error = copyout(w->w_tmem, w->w_where, len)) == 0)
1603 			w->w_where = (char *)w->w_where + len;
1604 	}
1605 	return error;
1606 }
1607 
1608 static int
1609 sysctl_iflist(int af, struct rt_walkarg *w, int type)
1610 {
1611 	struct ifnet *ifp;
1612 	struct ifaddr *ifa;
1613 	struct	rt_addrinfo info;
1614 	int	cmd, len, error = 0;
1615 	int	(*iflist_if)(struct ifnet *, struct rt_walkarg *,
1616 			     struct rt_addrinfo *, size_t);
1617 	int	(*iflist_addr)(struct rt_walkarg *, struct ifaddr *,
1618 			       struct rt_addrinfo *);
1619 	int s;
1620 	struct psref psref;
1621 	int bound = curlwp_bind();
1622 
1623 	switch (type) {
1624 	case NET_RT_IFLIST:
1625 		cmd = RTM_IFINFO;
1626 		iflist_if = sysctl_iflist_if;
1627 		iflist_addr = sysctl_iflist_addr;
1628 		break;
1629 #ifdef COMPAT_14
1630 	case NET_RT_OOOIFLIST:
1631 		cmd = RTM_OOIFINFO;
1632 		iflist_if = compat_14_iflist;
1633 		iflist_addr = compat_70_iflist_addr;
1634 		break;
1635 #endif
1636 #ifdef COMPAT_50
1637 	case NET_RT_OOIFLIST:
1638 		cmd = RTM_OIFINFO;
1639 		iflist_if = compat_50_iflist;
1640 		iflist_addr = compat_70_iflist_addr;
1641 		break;
1642 #endif
1643 #ifdef COMPAT_70
1644 	case NET_RT_OIFLIST:
1645 		cmd = RTM_IFINFO;
1646 		iflist_if = sysctl_iflist_if;
1647 		iflist_addr = compat_70_iflist_addr;
1648 		break;
1649 #endif
1650 	default:
1651 #ifdef DIAGNOSTIC
1652 		printf("sysctl_iflist\n");
1653 #endif
1654 		return EINVAL;
1655 	}
1656 
1657 	memset(&info, 0, sizeof(info));
1658 
1659 	s = pserialize_read_enter();
1660 	IFNET_READER_FOREACH(ifp) {
1661 		if (w->w_arg && w->w_arg != ifp->if_index)
1662 			continue;
1663 		if (IFADDR_READER_EMPTY(ifp))
1664 			continue;
1665 
1666 		psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class);
1667 		pserialize_read_exit(s);
1668 
1669 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1670 		if ((error = rt_msg2(cmd, &info, NULL, w, &len)) != 0)
1671 			goto release_exit;
1672 		info.rti_info[RTAX_IFP] = NULL;
1673 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1674 			if ((error = iflist_if(ifp, w, &info, len)) != 0)
1675 				goto release_exit;
1676 		}
1677 		IFADDR_READER_FOREACH(ifa, ifp) {
1678 			if (af && af != ifa->ifa_addr->sa_family)
1679 				continue;
1680 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1681 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1682 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1683 			if ((error = iflist_addr(w, ifa, &info)) != 0)
1684 				goto release_exit;
1685 		}
1686 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1687 		    info.rti_info[RTAX_BRD] = NULL;
1688 
1689 		s = pserialize_read_enter();
1690 		psref_release(&psref, &ifp->if_psref, ifnet_psref_class);
1691 	}
1692 	pserialize_read_exit(s);
1693 	curlwp_bindx(bound);
1694 
1695 	return 0;
1696 
1697 release_exit:
1698 	psref_release(&psref, &ifp->if_psref, ifnet_psref_class);
1699 	curlwp_bindx(bound);
1700 	return error;
1701 }
1702 
1703 static int
1704 sysctl_rtable(SYSCTLFN_ARGS)
1705 {
1706 	void 	*where = oldp;
1707 	size_t	*given = oldlenp;
1708 	int	i, s, error = EINVAL;
1709 	u_char  af;
1710 	struct	rt_walkarg w;
1711 
1712 	if (namelen == 1 && name[0] == CTL_QUERY)
1713 		return sysctl_query(SYSCTLFN_CALL(rnode));
1714 
1715 	if (newp)
1716 		return EPERM;
1717 	if (namelen != 3)
1718 		return EINVAL;
1719 	af = name[0];
1720 	w.w_tmemneeded = 0;
1721 	w.w_tmemsize = 0;
1722 	w.w_tmem = NULL;
1723 again:
1724 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1725 	if (w.w_tmemneeded) {
1726 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
1727 		w.w_tmemsize = w.w_tmemneeded;
1728 		w.w_tmemneeded = 0;
1729 	}
1730 	w.w_op = name[1];
1731 	w.w_arg = name[2];
1732 	w.w_given = *given;
1733 	w.w_needed = 0 - w.w_given;
1734 	w.w_where = where;
1735 
1736 	s = splsoftnet();
1737 	switch (w.w_op) {
1738 
1739 	case NET_RT_DUMP:
1740 	case NET_RT_FLAGS:
1741 #ifdef INET
1742 		/*
1743 		 * take care of llinfo entries, the caller must
1744 		 * specify an AF
1745 		 */
1746 		if (w.w_op == NET_RT_FLAGS &&
1747 		    (w.w_arg == 0 || w.w_arg & RTF_LLDATA)) {
1748 			if (af != 0)
1749 				error = lltable_sysctl_dumparp(af, &w);
1750 			else
1751 				error = EINVAL;
1752 			break;
1753 		}
1754 #endif /* INET */
1755 
1756 		for (i = 1; i <= AF_MAX; i++)
1757 			if ((af == 0 || af == i) &&
1758 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
1759 				break;
1760 		break;
1761 
1762 #ifdef COMPAT_14
1763 	case NET_RT_OOOIFLIST:
1764 		error = sysctl_iflist(af, &w, w.w_op);
1765 		break;
1766 #endif
1767 #ifdef COMPAT_50
1768 	case NET_RT_OOIFLIST:
1769 		error = sysctl_iflist(af, &w, w.w_op);
1770 		break;
1771 #endif
1772 #ifdef COMPAT_70
1773 	case NET_RT_OIFLIST:
1774 		error = sysctl_iflist(af, &w, w.w_op);
1775 		break;
1776 #endif
1777 	case NET_RT_IFLIST:
1778 		error = sysctl_iflist(af, &w, w.w_op);
1779 		break;
1780 	}
1781 	splx(s);
1782 
1783 	/* check to see if we couldn't allocate memory with NOWAIT */
1784 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1785 		goto again;
1786 
1787 	if (w.w_tmem)
1788 		free(w.w_tmem, M_RTABLE);
1789 	w.w_needed += w.w_given;
1790 	if (where) {
1791 		*given = (char *)w.w_where - (char *)where;
1792 		if (*given < w.w_needed)
1793 			return ENOMEM;
1794 	} else {
1795 		*given = (11 * w.w_needed) / 10;
1796 	}
1797 	return error;
1798 }
1799 
1800 /*
1801  * Routing message software interrupt routine
1802  */
1803 static void
1804 COMPATNAME(route_intr)(void *cookie)
1805 {
1806 	struct sockproto proto = { .sp_family = PF_XROUTE, };
1807 	struct route_info * const ri = &COMPATNAME(route_info);
1808 	struct mbuf *m;
1809 
1810 	mutex_enter(softnet_lock);
1811 	KERNEL_LOCK(1, NULL);
1812 	for (;;) {
1813 		IFQ_LOCK(&ri->ri_intrq);
1814 		IF_DEQUEUE(&ri->ri_intrq, m);
1815 		IFQ_UNLOCK(&ri->ri_intrq);
1816 		if (m == NULL)
1817 			break;
1818 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
1819 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst);
1820 	}
1821 	KERNEL_UNLOCK_ONE(NULL);
1822 	mutex_exit(softnet_lock);
1823 }
1824 
1825 /*
1826  * Enqueue a message to the software interrupt routine.
1827  */
1828 void
1829 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
1830 {
1831 	struct route_info * const ri = &COMPATNAME(route_info);
1832 	int wasempty;
1833 
1834 	IFQ_LOCK(&ri->ri_intrq);
1835 	if (IF_QFULL(&ri->ri_intrq)) {
1836 		IF_DROP(&ri->ri_intrq);
1837 		IFQ_UNLOCK(&ri->ri_intrq);
1838 		m_freem(m);
1839 	} else {
1840 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
1841 		M_SETCTX(m, (uintptr_t)family);
1842 		IF_ENQUEUE(&ri->ri_intrq, m);
1843 		IFQ_UNLOCK(&ri->ri_intrq);
1844 		if (wasempty) {
1845 			kpreempt_disable();
1846 			softint_schedule(ri->ri_sih);
1847 			kpreempt_enable();
1848 		}
1849 	}
1850 }
1851 
1852 static void
1853 COMPATNAME(route_init)(void)
1854 {
1855 	struct route_info * const ri = &COMPATNAME(route_info);
1856 
1857 #ifndef COMPAT_RTSOCK
1858 	rt_init();
1859 #endif
1860 
1861 	sysctl_net_route_setup(NULL);
1862 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
1863 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
1864 	    COMPATNAME(route_intr), NULL);
1865 	IFQ_LOCK_INIT(&ri->ri_intrq);
1866 }
1867 
1868 /*
1869  * Definitions of protocols supported in the ROUTE domain.
1870  */
1871 #ifndef COMPAT_RTSOCK
1872 PR_WRAP_USRREQS(route);
1873 #else
1874 PR_WRAP_USRREQS(compat_50_route);
1875 #endif
1876 
1877 static const struct pr_usrreqs route_usrreqs = {
1878 	.pr_attach	= COMPATNAME(route_attach_wrapper),
1879 	.pr_detach	= COMPATNAME(route_detach_wrapper),
1880 	.pr_accept	= COMPATNAME(route_accept_wrapper),
1881 	.pr_bind	= COMPATNAME(route_bind_wrapper),
1882 	.pr_listen	= COMPATNAME(route_listen_wrapper),
1883 	.pr_connect	= COMPATNAME(route_connect_wrapper),
1884 	.pr_connect2	= COMPATNAME(route_connect2_wrapper),
1885 	.pr_disconnect	= COMPATNAME(route_disconnect_wrapper),
1886 	.pr_shutdown	= COMPATNAME(route_shutdown_wrapper),
1887 	.pr_abort	= COMPATNAME(route_abort_wrapper),
1888 	.pr_ioctl	= COMPATNAME(route_ioctl_wrapper),
1889 	.pr_stat	= COMPATNAME(route_stat_wrapper),
1890 	.pr_peeraddr	= COMPATNAME(route_peeraddr_wrapper),
1891 	.pr_sockaddr	= COMPATNAME(route_sockaddr_wrapper),
1892 	.pr_rcvd	= COMPATNAME(route_rcvd_wrapper),
1893 	.pr_recvoob	= COMPATNAME(route_recvoob_wrapper),
1894 	.pr_send	= COMPATNAME(route_send_wrapper),
1895 	.pr_sendoob	= COMPATNAME(route_sendoob_wrapper),
1896 	.pr_purgeif	= COMPATNAME(route_purgeif_wrapper),
1897 };
1898 
1899 static const struct protosw COMPATNAME(route_protosw)[] = {
1900 	{
1901 		.pr_type = SOCK_RAW,
1902 		.pr_domain = &COMPATNAME(routedomain),
1903 		.pr_flags = PR_ATOMIC|PR_ADDR,
1904 		.pr_input = raw_input,
1905 		.pr_ctlinput = raw_ctlinput,
1906 		.pr_usrreqs = &route_usrreqs,
1907 		.pr_init = raw_init,
1908 	},
1909 };
1910 
1911 struct domain COMPATNAME(routedomain) = {
1912 	.dom_family = PF_XROUTE,
1913 	.dom_name = DOMAINNAME,
1914 	.dom_init = COMPATNAME(route_init),
1915 	.dom_protosw = COMPATNAME(route_protosw),
1916 	.dom_protoswNPROTOSW =
1917 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
1918 };
1919 
1920 static void
1921 sysctl_net_route_setup(struct sysctllog **clog)
1922 {
1923 	const struct sysctlnode *rnode = NULL;
1924 
1925 	sysctl_createv(clog, 0, NULL, &rnode,
1926 		       CTLFLAG_PERMANENT,
1927 		       CTLTYPE_NODE, DOMAINNAME,
1928 		       SYSCTL_DESCR("PF_ROUTE information"),
1929 		       NULL, 0, NULL, 0,
1930 		       CTL_NET, PF_XROUTE, CTL_EOL);
1931 
1932 	sysctl_createv(clog, 0, NULL, NULL,
1933 		       CTLFLAG_PERMANENT,
1934 		       CTLTYPE_NODE, "rtable",
1935 		       SYSCTL_DESCR("Routing table information"),
1936 		       sysctl_rtable, 0, NULL, 0,
1937 		       CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
1938 
1939 	sysctl_createv(clog, 0, &rnode, NULL,
1940 		       CTLFLAG_PERMANENT,
1941 		       CTLTYPE_STRUCT, "stats",
1942 		       SYSCTL_DESCR("Routing statistics"),
1943 		       NULL, 0, &rtstat, sizeof(rtstat),
1944 		       CTL_CREATE, CTL_EOL);
1945 }
1946