xref: /netbsd-src/sys/net/rtsock.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: rtsock.c,v 1.241 2018/04/25 03:49:57 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 1.241 2018/04/25 03:49:57 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 #include "opt_net_mpsafe.h"
72 #endif
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/proc.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/domain.h>
80 #include <sys/protosw.h>
81 #include <sys/sysctl.h>
82 #include <sys/kauth.h>
83 #include <sys/kmem.h>
84 #include <sys/intr.h>
85 #include <sys/condvar.h>
86 
87 #include <net/if.h>
88 #include <net/if_llatbl.h>
89 #include <net/if_types.h>
90 #include <net/route.h>
91 #include <net/raw_cb.h>
92 
93 #include <netinet/in_var.h>
94 #include <netinet/if_inarp.h>
95 
96 #include <netmpls/mpls.h>
97 
98 #ifdef SCTP
99 extern void sctp_add_ip_address(struct ifaddr *);
100 extern void sctp_delete_ip_address(struct ifaddr *);
101 #endif
102 
103 #if defined(COMPAT_14) || defined(COMPAT_50) || defined(COMPAT_70)
104 #include <compat/net/if.h>
105 #include <compat/net/route.h>
106 #endif
107 #ifdef COMPAT_RTSOCK
108 #define	RTM_XVERSION	RTM_OVERSION
109 #define	RTM_XNEWADDR	RTM_ONEWADDR
110 #define	RTM_XDELADDR	RTM_ODELADDR
111 #define	RTM_XCHGADDR	RTM_OCHGADDR
112 #define	RT_XADVANCE(a,b) RT_OADVANCE(a,b)
113 #define	RT_XROUNDUP(n)	RT_OROUNDUP(n)
114 #define	PF_XROUTE	PF_OROUTE
115 #define	rt_xmsghdr	rt_msghdr50
116 #define	if_xmsghdr	if_msghdr	/* if_msghdr50 is for RTM_OIFINFO */
117 #define	ifa_xmsghdr	ifa_msghdr50
118 #define	if_xannouncemsghdr	if_announcemsghdr50
119 #define	COMPATNAME(x)	compat_50_ ## x
120 #define	DOMAINNAME	"oroute"
121 CTASSERT(sizeof(struct ifa_xmsghdr) == 20);
122 DOMAIN_DEFINE(compat_50_routedomain); /* forward declare and add to link set */
123 #undef COMPAT_70
124 #else /* COMPAT_RTSOCK */
125 #define	RTM_XVERSION	RTM_VERSION
126 #define	RTM_XNEWADDR	RTM_NEWADDR
127 #define	RTM_XDELADDR	RTM_DELADDR
128 #define	RTM_XCHGADDR	RTM_CHGADDR
129 #define	RT_XADVANCE(a,b) RT_ADVANCE(a,b)
130 #define	RT_XROUNDUP(n)	RT_ROUNDUP(n)
131 #define	PF_XROUTE	PF_ROUTE
132 #define	rt_xmsghdr	rt_msghdr
133 #define	if_xmsghdr	if_msghdr
134 #define	ifa_xmsghdr	ifa_msghdr
135 #define	if_xannouncemsghdr	if_announcemsghdr
136 #define	COMPATNAME(x)	x
137 #define	DOMAINNAME	"route"
138 CTASSERT(sizeof(struct ifa_xmsghdr) == 32);
139 #ifdef COMPAT_50
140 #define	COMPATCALL(name, args)	compat_50_ ## name args
141 #endif
142 DOMAIN_DEFINE(routedomain); /* forward declare and add to link set */
143 #undef COMPAT_50
144 #undef COMPAT_14
145 #endif /* COMPAT_RTSOCK */
146 
147 #ifndef COMPATCALL
148 #define	COMPATCALL(name, args)	do { } while (/*CONSTCOND*/ 0)
149 #endif
150 
151 #ifdef RTSOCK_DEBUG
152 #define RT_IN_PRINT(info, b, a) (in_print((b), sizeof(b), \
153     &((const struct sockaddr_in *)(info)->rti_info[(a)])->sin_addr), (b))
154 #endif /* RTSOCK_DEBUG */
155 
156 struct route_info COMPATNAME(route_info) = {
157 	.ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, },
158 	.ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, },
159 	.ri_maxqlen = IFQ_MAXLEN,
160 };
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 struct routecb {
180 	struct rawcb	rocb_rcb;
181 	unsigned int	rocb_msgfilter;
182 #define	RTMSGFILTER(m)	(1U << (m))
183 };
184 #define sotoroutecb(so)	((struct routecb *)(so)->so_pcb)
185 
186 static struct rawcbhead rt_rawcb;
187 #ifdef NET_MPSAFE
188 static kmutex_t *rt_so_mtx;
189 
190 static bool rt_updating = false;
191 static kcondvar_t rt_update_cv;
192 #endif
193 
194 static void
195 rt_adjustcount(int af, int cnt)
196 {
197 	struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
198 
199 	cb->any_count += cnt;
200 
201 	switch (af) {
202 	case AF_INET:
203 		cb->ip_count += cnt;
204 		return;
205 #ifdef INET6
206 	case AF_INET6:
207 		cb->ip6_count += cnt;
208 		return;
209 #endif
210 	case AF_MPLS:
211 		cb->mpls_count += cnt;
212 		return;
213 	}
214 }
215 
216 static int
217 COMPATNAME(route_filter)(struct mbuf *m, struct sockproto *proto,
218     struct rawcb *rp)
219 {
220 	struct routecb *rop = (struct routecb *)rp;
221 	struct rt_xmsghdr *rtm;
222 
223 	KASSERT(m != NULL);
224 	KASSERT(proto != NULL);
225 	KASSERT(rp != NULL);
226 
227 	/* Wrong family for this socket. */
228 	if (proto->sp_family != PF_ROUTE)
229 		return ENOPROTOOPT;
230 
231 	/* If no filter set, just return. */
232 	if (rop->rocb_msgfilter == 0)
233 		return 0;
234 
235 	/* Ensure we can access rtm_type */
236 	if (m->m_len <
237 	    offsetof(struct rt_xmsghdr, rtm_type) + sizeof(rtm->rtm_type))
238 		return EINVAL;
239 
240 	rtm = mtod(m, struct rt_xmsghdr *);
241 	/* If the rtm type is filtered out, return a positive. */
242 	if (!(rop->rocb_msgfilter & RTMSGFILTER(rtm->rtm_type)))
243 		return EEXIST;
244 
245 	/* Passed the filter. */
246 	return 0;
247 }
248 
249 static void
250 rt_pr_init(void)
251 {
252 
253 	LIST_INIT(&rt_rawcb);
254 }
255 
256 static int
257 COMPATNAME(route_attach)(struct socket *so, int proto)
258 {
259 	struct rawcb *rp;
260 	struct routecb *rop;
261 	int s, error;
262 
263 	KASSERT(sotorawcb(so) == NULL);
264 	rop = kmem_zalloc(sizeof(*rop), KM_SLEEP);
265 	rp = &rop->rocb_rcb;
266 	rp->rcb_len = sizeof(*rop);
267 	so->so_pcb = rp;
268 
269 	s = splsoftnet();
270 
271 #ifdef NET_MPSAFE
272 	KASSERT(so->so_lock == NULL);
273 	mutex_obj_hold(rt_so_mtx);
274 	so->so_lock = rt_so_mtx;
275 	solock(so);
276 #endif
277 
278 	if ((error = raw_attach(so, proto, &rt_rawcb)) == 0) {
279 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
280 		rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
281 		rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
282 		rp->rcb_filter = COMPATNAME(route_filter);
283 	}
284 	splx(s);
285 
286 	if (error) {
287 		kmem_free(rop, sizeof(*rop));
288 		so->so_pcb = NULL;
289 		return error;
290 	}
291 
292 	soisconnected(so);
293 	so->so_options |= SO_USELOOPBACK;
294 	KASSERT(solocked(so));
295 
296 	return error;
297 }
298 
299 static void
300 COMPATNAME(route_detach)(struct socket *so)
301 {
302 	struct rawcb *rp = sotorawcb(so);
303 	int s;
304 
305 	KASSERT(rp != NULL);
306 	KASSERT(solocked(so));
307 
308 	s = splsoftnet();
309 	rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
310 	raw_detach(so);
311 	splx(s);
312 }
313 
314 static int
315 COMPATNAME(route_accept)(struct socket *so, struct sockaddr *nam)
316 {
317 	KASSERT(solocked(so));
318 
319 	panic("route_accept");
320 
321 	return EOPNOTSUPP;
322 }
323 
324 static int
325 COMPATNAME(route_bind)(struct socket *so, struct sockaddr *nam, struct lwp *l)
326 {
327 	KASSERT(solocked(so));
328 
329 	return EOPNOTSUPP;
330 }
331 
332 static int
333 COMPATNAME(route_listen)(struct socket *so, struct lwp *l)
334 {
335 	KASSERT(solocked(so));
336 
337 	return EOPNOTSUPP;
338 }
339 
340 static int
341 COMPATNAME(route_connect)(struct socket *so, struct sockaddr *nam, struct lwp *l)
342 {
343 	KASSERT(solocked(so));
344 
345 	return EOPNOTSUPP;
346 }
347 
348 static int
349 COMPATNAME(route_connect2)(struct socket *so, struct socket *so2)
350 {
351 	KASSERT(solocked(so));
352 
353 	return EOPNOTSUPP;
354 }
355 
356 static int
357 COMPATNAME(route_disconnect)(struct socket *so)
358 {
359 	struct rawcb *rp = sotorawcb(so);
360 	int s;
361 
362 	KASSERT(solocked(so));
363 	KASSERT(rp != NULL);
364 
365 	s = splsoftnet();
366 	soisdisconnected(so);
367 	raw_disconnect(rp);
368 	splx(s);
369 
370 	return 0;
371 }
372 
373 static int
374 COMPATNAME(route_shutdown)(struct socket *so)
375 {
376 	int s;
377 
378 	KASSERT(solocked(so));
379 
380 	/*
381 	 * Mark the connection as being incapable of further input.
382 	 */
383 	s = splsoftnet();
384 	socantsendmore(so);
385 	splx(s);
386 	return 0;
387 }
388 
389 static int
390 COMPATNAME(route_abort)(struct socket *so)
391 {
392 	KASSERT(solocked(so));
393 
394 	panic("route_abort");
395 
396 	return EOPNOTSUPP;
397 }
398 
399 static int
400 COMPATNAME(route_ioctl)(struct socket *so, u_long cmd, void *nam,
401     struct ifnet * ifp)
402 {
403 	return EOPNOTSUPP;
404 }
405 
406 static int
407 COMPATNAME(route_stat)(struct socket *so, struct stat *ub)
408 {
409 	KASSERT(solocked(so));
410 
411 	return 0;
412 }
413 
414 static int
415 COMPATNAME(route_peeraddr)(struct socket *so, struct sockaddr *nam)
416 {
417 	struct rawcb *rp = sotorawcb(so);
418 
419 	KASSERT(solocked(so));
420 	KASSERT(rp != NULL);
421 	KASSERT(nam != NULL);
422 
423 	if (rp->rcb_faddr == NULL)
424 		return ENOTCONN;
425 
426 	raw_setpeeraddr(rp, nam);
427 	return 0;
428 }
429 
430 static int
431 COMPATNAME(route_sockaddr)(struct socket *so, struct sockaddr *nam)
432 {
433 	struct rawcb *rp = sotorawcb(so);
434 
435 	KASSERT(solocked(so));
436 	KASSERT(rp != NULL);
437 	KASSERT(nam != NULL);
438 
439 	if (rp->rcb_faddr == NULL)
440 		return ENOTCONN;
441 
442 	raw_setsockaddr(rp, nam);
443 	return 0;
444 }
445 
446 static int
447 COMPATNAME(route_rcvd)(struct socket *so, int flags, struct lwp *l)
448 {
449 	KASSERT(solocked(so));
450 
451 	return EOPNOTSUPP;
452 }
453 
454 static int
455 COMPATNAME(route_recvoob)(struct socket *so, struct mbuf *m, int flags)
456 {
457 	KASSERT(solocked(so));
458 
459 	return EOPNOTSUPP;
460 }
461 
462 static int
463 COMPATNAME(route_send)(struct socket *so, struct mbuf *m,
464     struct sockaddr *nam, struct mbuf *control, struct lwp *l)
465 {
466 	int error = 0;
467 	int s;
468 
469 	KASSERT(solocked(so));
470 	KASSERT(so->so_proto == &COMPATNAME(route_protosw)[0]);
471 
472 	s = splsoftnet();
473 	error = raw_send(so, m, nam, control, l, &COMPATNAME(route_output));
474 	splx(s);
475 
476 	return error;
477 }
478 
479 static int
480 COMPATNAME(route_sendoob)(struct socket *so, struct mbuf *m,
481     struct mbuf *control)
482 {
483 	KASSERT(solocked(so));
484 
485 	m_freem(m);
486 	m_freem(control);
487 
488 	return EOPNOTSUPP;
489 }
490 static int
491 COMPATNAME(route_purgeif)(struct socket *so, struct ifnet *ifp)
492 {
493 
494 	panic("route_purgeif");
495 
496 	return EOPNOTSUPP;
497 }
498 
499 #if defined(INET) || defined(INET6)
500 static int
501 route_get_sdl_index(struct rt_addrinfo *info, int *sdl_index)
502 {
503 	struct rtentry *nrt;
504 	int error;
505 
506 	error = rtrequest1(RTM_GET, info, &nrt);
507 	if (error != 0)
508 		return error;
509 	/*
510 	 * nrt->rt_ifp->if_index may not be correct
511 	 * due to changing to ifplo0.
512 	 */
513 	*sdl_index = satosdl(nrt->rt_gateway)->sdl_index;
514 	rt_unref(nrt);
515 
516 	return 0;
517 }
518 #endif
519 
520 static void
521 route_get_sdl(const struct ifnet *ifp, const struct sockaddr *dst,
522     struct sockaddr_dl *sdl, int *flags)
523 {
524 	struct llentry *la;
525 
526 	KASSERT(ifp != NULL);
527 
528 	IF_AFDATA_RLOCK(ifp);
529 	switch (dst->sa_family) {
530 	case AF_INET:
531 		la = lla_lookup(LLTABLE(ifp), 0, dst);
532 		break;
533 	case AF_INET6:
534 		la = lla_lookup(LLTABLE6(ifp), 0, dst);
535 		break;
536 	default:
537 		la = NULL;
538 		KASSERTMSG(0, "Invalid AF=%d\n", dst->sa_family);
539 		break;
540 	}
541 	IF_AFDATA_RUNLOCK(ifp);
542 
543 	void *a = (LLE_IS_VALID(la) && (la->la_flags & LLE_VALID) == LLE_VALID)
544 	    ? &la->ll_addr : NULL;
545 
546 	a = sockaddr_dl_init(sdl, sizeof(*sdl), ifp->if_index, ifp->if_type,
547 	    NULL, 0, a, ifp->if_addrlen);
548 	KASSERT(a != NULL);
549 
550 	if (la != NULL) {
551 		*flags = la->la_flags;
552 		LLE_RUNLOCK(la);
553 	}
554 }
555 
556 static int
557 route_output_report(struct rtentry *rt, struct rt_addrinfo *info,
558     struct rt_xmsghdr *rtm, struct rt_xmsghdr **new_rtm)
559 {
560 	int len;
561 
562 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
563 		const struct ifaddr *rtifa;
564 		const struct ifnet *ifp = rt->rt_ifp;
565 
566 		info->rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
567 		/* rtifa used to be simply rt->rt_ifa.
568 		 * If rt->rt_ifa != NULL, then
569 		 * rt_get_ifa() != NULL.  So this
570 		 * ought to still be safe. --dyoung
571 		 */
572 		rtifa = rt_get_ifa(rt);
573 		info->rti_info[RTAX_IFA] = rtifa->ifa_addr;
574 #ifdef RTSOCK_DEBUG
575 		if (info->rti_info[RTAX_IFA]->sa_family == AF_INET) {
576 			char ibuf[INET_ADDRSTRLEN];
577 			char abuf[INET_ADDRSTRLEN];
578 			printf("%s: copying out RTAX_IFA %s "
579 			    "for info->rti_info[RTAX_DST] %s "
580 			    "ifa_getifa %p ifa_seqno %p\n",
581 			    __func__,
582 			    RT_IN_PRINT(info, ibuf, RTAX_IFA),
583 			    RT_IN_PRINT(info, abuf, RTAX_DST),
584 			    (void *)rtifa->ifa_getifa,
585 			    rtifa->ifa_seqno);
586 		}
587 #endif /* RTSOCK_DEBUG */
588 		if (ifp->if_flags & IFF_POINTOPOINT)
589 			info->rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
590 		else
591 			info->rti_info[RTAX_BRD] = NULL;
592 		rtm->rtm_index = ifp->if_index;
593 	}
594 	(void)rt_msg2(rtm->rtm_type, info, NULL, NULL, &len);
595 	if (len > rtm->rtm_msglen) {
596 		struct rt_xmsghdr *old_rtm = rtm;
597 		R_Malloc(*new_rtm, struct rt_xmsghdr *, len);
598 		if (*new_rtm == NULL)
599 			return ENOBUFS;
600 		(void)memcpy(*new_rtm, old_rtm, old_rtm->rtm_msglen);
601 		rtm = *new_rtm;
602 	}
603 	(void)rt_msg2(rtm->rtm_type, info, rtm, NULL, 0);
604 	rtm->rtm_flags = rt->rt_flags;
605 	rtm_setmetrics(rt, rtm);
606 	rtm->rtm_addrs = info->rti_addrs;
607 
608 	return 0;
609 }
610 
611 /*ARGSUSED*/
612 int
613 COMPATNAME(route_output)(struct mbuf *m, struct socket *so)
614 {
615 	struct sockproto proto = { .sp_family = PF_XROUTE, };
616 	struct rt_xmsghdr *rtm = NULL;
617 	struct rt_xmsghdr *old_rtm = NULL, *new_rtm = NULL;
618 	struct rtentry *rt = NULL;
619 	struct rtentry *saved_nrt = NULL;
620 	struct rt_addrinfo info;
621 	int len, error = 0;
622 	sa_family_t family;
623 	struct sockaddr_dl sdl;
624 	int bound = curlwp_bind();
625 	bool do_rt_free = false;
626 	struct sockaddr_storage netmask;
627 
628 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
629 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
630 	   (m = m_pullup(m, sizeof(int32_t))) == NULL)) {
631 		error = ENOBUFS;
632 		goto out;
633 	}
634 	if ((m->m_flags & M_PKTHDR) == 0)
635 		panic("%s", __func__);
636 	len = m->m_pkthdr.len;
637 	if (len < sizeof(*rtm) ||
638 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
639 		info.rti_info[RTAX_DST] = NULL;
640 		senderr(EINVAL);
641 	}
642 	R_Malloc(rtm, struct rt_xmsghdr *, len);
643 	if (rtm == NULL) {
644 		info.rti_info[RTAX_DST] = NULL;
645 		senderr(ENOBUFS);
646 	}
647 	m_copydata(m, 0, len, rtm);
648 	if (rtm->rtm_version != RTM_XVERSION) {
649 		info.rti_info[RTAX_DST] = NULL;
650 		senderr(EPROTONOSUPPORT);
651 	}
652 	rtm->rtm_pid = curproc->p_pid;
653 	memset(&info, 0, sizeof(info));
654 	info.rti_addrs = rtm->rtm_addrs;
655 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
656 	    &info)) {
657 		senderr(EINVAL);
658 	}
659 	info.rti_flags = rtm->rtm_flags;
660 #ifdef RTSOCK_DEBUG
661 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
662 		char abuf[INET_ADDRSTRLEN];
663 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
664 		    RT_IN_PRINT(&info, abuf, RTAX_DST));
665 	}
666 #endif /* RTSOCK_DEBUG */
667 	if (info.rti_info[RTAX_DST] == NULL ||
668 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
669 		senderr(EINVAL);
670 	}
671 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
672 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
673 		senderr(EINVAL);
674 	}
675 
676 	/*
677 	 * Verify that the caller has the appropriate privilege; RTM_GET
678 	 * is the only operation the non-superuser is allowed.
679 	 */
680 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
681 	    0, rtm, NULL, NULL) != 0)
682 		senderr(EACCES);
683 
684 	/*
685 	 * route(8) passes a sockaddr truncated with prefixlen.
686 	 * The kernel doesn't expect such sockaddr and need to
687 	 * use a buffer that is big enough for the sockaddr expected
688 	 * (padded with 0's). We keep the original length of the sockaddr.
689 	 */
690 	if (info.rti_info[RTAX_NETMASK]) {
691 		/*
692 		 * Use the family of RTAX_DST, because RTAX_NETMASK
693 		 * can have a zero family if it comes from the radix
694 		 * tree via rt_mask().
695 		 */
696 		socklen_t sa_len = sockaddr_getsize_by_family(
697 		    info.rti_info[RTAX_DST]->sa_family);
698 		socklen_t masklen = sockaddr_getlen(
699 		    info.rti_info[RTAX_NETMASK]);
700 		if (sa_len != 0 && sa_len > masklen) {
701 			KASSERT(sa_len <= sizeof(netmask));
702 			memcpy(&netmask, info.rti_info[RTAX_NETMASK], masklen);
703 			memset((char *)&netmask + masklen, 0, sa_len - masklen);
704 			info.rti_info[RTAX_NETMASK] = sstocsa(&netmask);
705 		}
706 	}
707 
708 	switch (rtm->rtm_type) {
709 
710 	case RTM_ADD:
711 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
712 			senderr(EINVAL);
713 		}
714 #if defined(INET) || defined(INET6)
715 		/* support for new ARP/NDP code with keeping backcompat */
716 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) {
717 			const struct sockaddr_dl *sdlp =
718 			    satocsdl(info.rti_info[RTAX_GATEWAY]);
719 
720 			/* Allow routing requests by interface index */
721 			if (sdlp->sdl_nlen == 0 && sdlp->sdl_alen == 0
722 			    && sdlp->sdl_slen == 0)
723 				goto fallback;
724 			/*
725 			 * Old arp binaries don't set the sdl_index
726 			 * so we have to complement it.
727 			 */
728 			int sdl_index = sdlp->sdl_index;
729 			if (sdl_index == 0) {
730 				error = route_get_sdl_index(&info, &sdl_index);
731 				if (error != 0)
732 					goto fallback;
733 			} else if (
734 			    info.rti_info[RTAX_DST]->sa_family == AF_INET) {
735 				/*
736 				 * XXX workaround for SIN_PROXY case; proxy arp
737 				 * entry should be in an interface that has
738 				 * a network route including the destination,
739 				 * not a local (link) route that may not be a
740 				 * desired place, for example a tap.
741 				 */
742 				const struct sockaddr_inarp *sina =
743 				    (const struct sockaddr_inarp *)
744 				    info.rti_info[RTAX_DST];
745 				if (sina->sin_other & SIN_PROXY) {
746 					error = route_get_sdl_index(&info,
747 					    &sdl_index);
748 					if (error != 0)
749 						goto fallback;
750 				}
751 			}
752 			error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
753 			    rtm->rtm_rmx.rmx_expire, &info, sdl_index);
754 			break;
755 		}
756 	fallback:
757 #endif /* defined(INET) || defined(INET6) */
758 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
759 		if (error == 0) {
760 			_rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
761 			rt_unref(saved_nrt);
762 		}
763 		break;
764 
765 	case RTM_DELETE:
766 #if defined(INET) || defined(INET6)
767 		/* support for new ARP/NDP code */
768 		if (info.rti_info[RTAX_GATEWAY] &&
769 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
770 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
771 			const struct sockaddr_dl *sdlp =
772 			    satocsdl(info.rti_info[RTAX_GATEWAY]);
773 			error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
774 			    rtm->rtm_rmx.rmx_expire, &info, sdlp->sdl_index);
775 			rtm->rtm_flags &= ~RTF_UP;
776 			break;
777 		}
778 #endif
779 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
780 		if (error != 0)
781 			break;
782 
783 		rt = saved_nrt;
784 		do_rt_free = true;
785 		info.rti_info[RTAX_DST] = rt_getkey(rt);
786 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
787 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
788 		info.rti_info[RTAX_TAG] = rt_gettag(rt);
789 		error = route_output_report(rt, &info, rtm, &new_rtm);
790 		if (error)
791 			senderr(error);
792 		if (new_rtm != NULL) {
793 			old_rtm = rtm;
794 			rtm = new_rtm;
795 		}
796 		break;
797 
798 	case RTM_GET:
799 	case RTM_CHANGE:
800 	case RTM_LOCK:
801                 /* XXX This will mask info.rti_info[RTAX_DST] with
802 		 * info.rti_info[RTAX_NETMASK] before
803                  * searching.  It did not used to do that.  --dyoung
804 		 */
805 		rt = NULL;
806 		error = rtrequest1(RTM_GET, &info, &rt);
807 		if (error != 0)
808 			senderr(error);
809 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
810 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
811 			    info.rti_info[RTAX_DST]->sa_len) != 0)
812 				senderr(ESRCH);
813 			if (info.rti_info[RTAX_NETMASK] == NULL &&
814 			    rt_mask(rt) != NULL)
815 				senderr(ETOOMANYREFS);
816 		}
817 
818 		/*
819 		 * XXX if arp/ndp requests an L2 entry, we have to obtain
820 		 * it from lltable while for the route command we have to
821 		 * return a route as it is. How to distinguish them?
822 		 * For newer arp/ndp, RTF_LLDATA flag set by arp/ndp
823 		 * indicates an L2 entry is requested. For old arp/ndp
824 		 * binaries, we check RTF_UP flag is NOT set; it works
825 		 * by the fact that arp/ndp don't set it while the route
826 		 * command sets it.
827 		 */
828 		if (((rtm->rtm_flags & RTF_LLDATA) != 0 ||
829 		     (rtm->rtm_flags & RTF_UP) == 0) &&
830 		    rtm->rtm_type == RTM_GET &&
831 		    sockaddr_cmp(rt_getkey(rt), info.rti_info[RTAX_DST]) != 0) {
832 			int ll_flags = 0;
833 			route_get_sdl(rt->rt_ifp, info.rti_info[RTAX_DST], &sdl,
834 			    &ll_flags);
835 			info.rti_info[RTAX_GATEWAY] = sstocsa(&sdl);
836 			error = route_output_report(rt, &info, rtm, &new_rtm);
837 			if (error)
838 				senderr(error);
839 			if (new_rtm != NULL) {
840 				old_rtm = rtm;
841 				rtm = new_rtm;
842 			}
843 			rtm->rtm_flags |= RTF_LLDATA;
844 			rtm->rtm_flags &= ~RTF_CONNECTED;
845 			rtm->rtm_flags |= (ll_flags & LLE_STATIC) ? RTF_STATIC : 0;
846 			break;
847 		}
848 
849 		switch (rtm->rtm_type) {
850 		case RTM_GET:
851 			info.rti_info[RTAX_DST] = rt_getkey(rt);
852 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
853 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
854 			info.rti_info[RTAX_TAG] = rt_gettag(rt);
855 			error = route_output_report(rt, &info, rtm, &new_rtm);
856 			if (error)
857 				senderr(error);
858 			if (new_rtm != NULL) {
859 				old_rtm = rtm;
860 				rtm = new_rtm;
861 			}
862 			break;
863 
864 		case RTM_CHANGE:
865 #ifdef NET_MPSAFE
866 			/*
867 			 * Release rt_so_mtx to avoid a deadlock with route_intr
868 			 * and also serialize updating routes to avoid another.
869 			 */
870 			if (rt_updating) {
871 				/* Release to allow the updater to proceed */
872 				rt_unref(rt);
873 				rt = NULL;
874 			}
875 			while (rt_updating) {
876 				error = cv_wait_sig(&rt_update_cv, rt_so_mtx);
877 				if (error != 0)
878 					goto flush;
879 			}
880 			if (rt == NULL) {
881 				error = rtrequest1(RTM_GET, &info, &rt);
882 				if (error != 0)
883 					goto flush;
884 			}
885 			rt_updating = true;
886 			mutex_exit(rt_so_mtx);
887 
888 			error = rt_update_prepare(rt);
889 			if (error == 0) {
890 				error = rt_update(rt, &info, rtm);
891 				rt_update_finish(rt);
892 			}
893 
894 			mutex_enter(rt_so_mtx);
895 			rt_updating = false;
896 			cv_broadcast(&rt_update_cv);
897 #else
898 			error = rt_update(rt, &info, rtm);
899 #endif
900 			if (error != 0)
901 				goto flush;
902 			/*FALLTHROUGH*/
903 		case RTM_LOCK:
904 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
905 			rt->rt_rmx.rmx_locks |=
906 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
907 			break;
908 		}
909 		break;
910 
911 	default:
912 		senderr(EOPNOTSUPP);
913 	}
914 
915 flush:
916 	if (rtm) {
917 		if (error)
918 			rtm->rtm_errno = error;
919 		else
920 			rtm->rtm_flags |= RTF_DONE;
921 	}
922 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
923 	    0;
924 	/* We cannot free old_rtm until we have stopped using the
925 	 * pointers in info, some of which may point to sockaddrs
926 	 * in old_rtm.
927 	 */
928 	if (old_rtm != NULL)
929 		Free(old_rtm);
930 	if (rt) {
931 		if (do_rt_free) {
932 #ifdef NET_MPSAFE
933 			/*
934 			 * Release rt_so_mtx to avoid a deadlock with
935 			 * route_intr.
936 			 */
937 			mutex_exit(rt_so_mtx);
938 			rt_free(rt);
939 			mutex_enter(rt_so_mtx);
940 #else
941 			rt_free(rt);
942 #endif
943 		} else
944 			rt_unref(rt);
945 	}
946     {
947 	struct rawcb *rp = NULL;
948 	/*
949 	 * Check to see if we don't want our own messages.
950 	 */
951 	if ((so->so_options & SO_USELOOPBACK) == 0) {
952 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
953 			if (rtm)
954 				Free(rtm);
955 			m_freem(m);
956 			goto out;
957 		}
958 		/* There is another listener, so construct message */
959 		rp = sotorawcb(so);
960 	}
961 	if (rtm) {
962 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
963 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
964 			m_freem(m);
965 			m = NULL;
966 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
967 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
968 		Free(rtm);
969 	}
970 	if (rp)
971 		rp->rcb_proto.sp_family = 0; /* Avoid us */
972 	if (family)
973 		proto.sp_protocol = family;
974 	if (m)
975 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
976 		    &COMPATNAME(route_info).ri_dst, &rt_rawcb);
977 	if (rp)
978 		rp->rcb_proto.sp_family = PF_XROUTE;
979     }
980 out:
981 	curlwp_bindx(bound);
982 	return error;
983 }
984 
985 static int
986 route_ctloutput(int op, struct socket *so, struct sockopt *sopt)
987 {
988 	struct routecb *rop = sotoroutecb(so);
989 	int error = 0;
990 	unsigned char *rtm_type;
991 	size_t len;
992 	unsigned int msgfilter;
993 
994 	KASSERT(solocked(so));
995 
996 	if (sopt->sopt_level != AF_ROUTE) {
997 		error = ENOPROTOOPT;
998 	} else switch (op) {
999 	case PRCO_SETOPT:
1000 		switch (sopt->sopt_name) {
1001 		case RO_MSGFILTER:
1002 			msgfilter = 0;
1003 			for (rtm_type = sopt->sopt_data, len = sopt->sopt_size;
1004 			     len != 0;
1005 			     rtm_type++, len -= sizeof(*rtm_type))
1006 			{
1007 				/* Guard against overflowing our storage. */
1008 				if (*rtm_type >= sizeof(msgfilter) * CHAR_BIT) {
1009 					error = EOVERFLOW;
1010 					break;
1011 				}
1012 				msgfilter |= RTMSGFILTER(*rtm_type);
1013 			}
1014 			if (error == 0)
1015 				rop->rocb_msgfilter = msgfilter;
1016 			break;
1017 		default:
1018 			error = ENOPROTOOPT;
1019 			break;
1020 		}
1021 		break;
1022 	case PRCO_GETOPT:
1023 		switch (sopt->sopt_name) {
1024 		case RO_MSGFILTER:
1025 			error = ENOTSUP;
1026 			break;
1027 		default:
1028 			error = ENOPROTOOPT;
1029 			break;
1030 		}
1031 	}
1032 	return error;
1033 }
1034 
1035 static void
1036 _rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
1037 {
1038 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
1039 	metric(RTV_RPIPE, rmx_recvpipe);
1040 	metric(RTV_SPIPE, rmx_sendpipe);
1041 	metric(RTV_SSTHRESH, rmx_ssthresh);
1042 	metric(RTV_RTT, rmx_rtt);
1043 	metric(RTV_RTTVAR, rmx_rttvar);
1044 	metric(RTV_HOPCOUNT, rmx_hopcount);
1045 	metric(RTV_MTU, rmx_mtu);
1046 #undef metric
1047 	if (which & RTV_EXPIRE) {
1048 		out->rt_rmx.rmx_expire = in->rtm_rmx.rmx_expire ?
1049 		    time_wall_to_mono(in->rtm_rmx.rmx_expire) : 0;
1050 	}
1051 }
1052 
1053 #ifndef COMPAT_RTSOCK
1054 /*
1055  * XXX avoid using void * once msghdr compat disappears.
1056  */
1057 void
1058 rt_setmetrics(void *in, struct rtentry *out)
1059 {
1060 	const struct rt_xmsghdr *rtm = in;
1061 
1062 	_rt_setmetrics(rtm->rtm_inits, rtm, out);
1063 }
1064 #endif
1065 
1066 static void
1067 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
1068 {
1069 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
1070 	metric(rmx_recvpipe);
1071 	metric(rmx_sendpipe);
1072 	metric(rmx_ssthresh);
1073 	metric(rmx_rtt);
1074 	metric(rmx_rttvar);
1075 	metric(rmx_hopcount);
1076 	metric(rmx_mtu);
1077 	metric(rmx_locks);
1078 #undef metric
1079 	out->rtm_rmx.rmx_expire = in->rt_rmx.rmx_expire ?
1080 	    time_mono_to_wall(in->rt_rmx.rmx_expire) : 0;
1081 }
1082 
1083 static int
1084 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
1085     struct rt_addrinfo *rtinfo)
1086 {
1087 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
1088 	int i;
1089 
1090 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1091 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1092 			continue;
1093 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
1094 		RT_XADVANCE(cp, sa);
1095 	}
1096 
1097 	/*
1098 	 * Check for extra addresses specified, except RTM_GET asking
1099 	 * for interface info.
1100 	 */
1101 	if (rtmtype == RTM_GET) {
1102 		if (((rtinfo->rti_addrs &
1103 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0U << i)) != 0)
1104 			return 1;
1105 	} else if ((rtinfo->rti_addrs & (~0U << i)) != 0)
1106 		return 1;
1107 	/* Check for bad data length.  */
1108 	if (cp != cplim) {
1109 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
1110 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
1111 			/*
1112 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
1113 			 * We accept this for now for the sake of old
1114 			 * binaries or third party softwares.
1115 			 */
1116 			;
1117 		else
1118 			return 1;
1119 	}
1120 	return 0;
1121 }
1122 
1123 static int
1124 rt_getlen(int type)
1125 {
1126 #ifndef COMPAT_RTSOCK
1127 	CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
1128 	CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
1129 	CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
1130 	CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
1131 #endif
1132 
1133 	switch (type) {
1134 	case RTM_ODELADDR:
1135 	case RTM_ONEWADDR:
1136 	case RTM_OCHGADDR:
1137 #ifdef COMPAT_70
1138 		return sizeof(struct ifa_msghdr70);
1139 #else
1140 #ifdef RTSOCK_DEBUG
1141 		printf("%s: unsupported RTM type %d\n", __func__, type);
1142 #endif
1143 		return -1;
1144 #endif
1145 	case RTM_DELADDR:
1146 	case RTM_NEWADDR:
1147 	case RTM_CHGADDR:
1148 		return sizeof(struct ifa_xmsghdr);
1149 
1150 	case RTM_OOIFINFO:
1151 #ifdef COMPAT_14
1152 		return sizeof(struct if_msghdr14);
1153 #else
1154 #ifdef RTSOCK_DEBUG
1155 		printf("%s: unsupported RTM type RTM_OOIFINFO\n", __func__);
1156 #endif
1157 		return -1;
1158 #endif
1159 	case RTM_OIFINFO:
1160 #ifdef COMPAT_50
1161 		return sizeof(struct if_msghdr50);
1162 #else
1163 #ifdef RTSOCK_DEBUG
1164 		printf("%s: unsupported RTM type RTM_OIFINFO\n", __func__);
1165 #endif
1166 		return -1;
1167 #endif
1168 
1169 	case RTM_IFINFO:
1170 		return sizeof(struct if_xmsghdr);
1171 
1172 	case RTM_IFANNOUNCE:
1173 	case RTM_IEEE80211:
1174 		return sizeof(struct if_xannouncemsghdr);
1175 
1176 	default:
1177 		return sizeof(struct rt_xmsghdr);
1178 	}
1179 }
1180 
1181 
1182 struct mbuf *
1183 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
1184 {
1185 	struct rt_xmsghdr *rtm;
1186 	struct mbuf *m;
1187 	int i;
1188 	const struct sockaddr *sa;
1189 	int len, dlen;
1190 
1191 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1192 	if (m == NULL)
1193 		return m;
1194 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
1195 
1196 	if ((len = rt_getlen(type)) == -1)
1197 		goto out;
1198 	if (len > MHLEN + MLEN)
1199 		panic("%s: message too long", __func__);
1200 	else if (len > MHLEN) {
1201 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
1202 		if (m->m_next == NULL)
1203 			goto out;
1204 		MCLAIM(m->m_next, m->m_owner);
1205 		m->m_pkthdr.len = len;
1206 		m->m_len = MHLEN;
1207 		m->m_next->m_len = len - MHLEN;
1208 	} else {
1209 		m->m_pkthdr.len = m->m_len = len;
1210 	}
1211 	m_reset_rcvif(m);
1212 	m_copyback(m, 0, datalen, data);
1213 	if (len > datalen)
1214 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
1215 	rtm = mtod(m, struct rt_xmsghdr *);
1216 	for (i = 0; i < RTAX_MAX; i++) {
1217 		if ((sa = rtinfo->rti_info[i]) == NULL)
1218 			continue;
1219 		rtinfo->rti_addrs |= (1 << i);
1220 		dlen = RT_XROUNDUP(sa->sa_len);
1221 		m_copyback(m, len, sa->sa_len, sa);
1222 		if (dlen != sa->sa_len) {
1223 			/*
1224 			 * Up to 6 + 1 nul's since roundup is to
1225 			 * sizeof(uint64_t) (8 bytes)
1226 			 */
1227 			m_copyback(m, len + sa->sa_len,
1228 			    dlen - sa->sa_len, "\0\0\0\0\0\0");
1229 		}
1230 		len += dlen;
1231 	}
1232 	if (m->m_pkthdr.len != len)
1233 		goto out;
1234 	rtm->rtm_msglen = len;
1235 	rtm->rtm_version = RTM_XVERSION;
1236 	rtm->rtm_type = type;
1237 	return m;
1238 out:
1239 	m_freem(m);
1240 	return NULL;
1241 }
1242 
1243 /*
1244  * rt_msg2
1245  *
1246  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
1247  *		returns the length of the message in 'lenp'.
1248  *
1249  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
1250  *	the message
1251  * otherwise walkarg's w_needed is updated and if the user buffer is
1252  *	specified and w_needed indicates space exists the information is copied
1253  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
1254  *	if the allocation fails ENOBUFS is returned.
1255  */
1256 static int
1257 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1258 	int *lenp)
1259 {
1260 	int i;
1261 	int len, dlen, second_time = 0;
1262 	char *cp0, *cp = cpv;
1263 
1264 	rtinfo->rti_addrs = 0;
1265 again:
1266 	if ((len = rt_getlen(type)) == -1)
1267 		return EINVAL;
1268 
1269 	if ((cp0 = cp) != NULL)
1270 		cp += len;
1271 	for (i = 0; i < RTAX_MAX; i++) {
1272 		const struct sockaddr *sa;
1273 
1274 		if ((sa = rtinfo->rti_info[i]) == NULL)
1275 			continue;
1276 		rtinfo->rti_addrs |= (1 << i);
1277 		dlen = RT_XROUNDUP(sa->sa_len);
1278 		if (cp) {
1279 			int diff = dlen - sa->sa_len;
1280 			(void)memcpy(cp, sa, (size_t)sa->sa_len);
1281 			cp += sa->sa_len;
1282 			if (diff > 0) {
1283 				(void)memset(cp, 0, (size_t)diff);
1284 				cp += diff;
1285 			}
1286 		}
1287 		len += dlen;
1288 	}
1289 	if (cp == NULL && w != NULL && !second_time) {
1290 		struct rt_walkarg *rw = w;
1291 
1292 		rw->w_needed += len;
1293 		if (rw->w_needed <= 0 && rw->w_where) {
1294 			if (rw->w_tmemsize < len) {
1295 				if (rw->w_tmem)
1296 					kmem_free(rw->w_tmem, rw->w_tmemsize);
1297 				rw->w_tmem = kmem_alloc(len, KM_SLEEP);
1298 				rw->w_tmemsize = len;
1299 			}
1300 			if (rw->w_tmem) {
1301 				cp = rw->w_tmem;
1302 				second_time = 1;
1303 				goto again;
1304 			} else {
1305 				rw->w_tmemneeded = len;
1306 				return ENOBUFS;
1307 			}
1308 		}
1309 	}
1310 	if (cp) {
1311 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
1312 
1313 		rtm->rtm_version = RTM_XVERSION;
1314 		rtm->rtm_type = type;
1315 		rtm->rtm_msglen = len;
1316 	}
1317 	if (lenp)
1318 		*lenp = len;
1319 	return 0;
1320 }
1321 
1322 #ifndef COMPAT_RTSOCK
1323 int
1324 rt_msg3(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1325 	int *lenp)
1326 {
1327 	return rt_msg2(type, rtinfo, cpv, w, lenp);
1328 }
1329 #endif
1330 
1331 /*
1332  * This routine is called to generate a message from the routing
1333  * socket indicating that a redirect has occurred, a routing lookup
1334  * has failed, or that a protocol has detected timeouts to a particular
1335  * destination.
1336  */
1337 void
1338 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
1339     int error)
1340 {
1341 	struct rt_xmsghdr rtm;
1342 	struct mbuf *m;
1343 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1344 	struct rt_addrinfo info = *rtinfo;
1345 
1346 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
1347 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1348 		return;
1349 	memset(&rtm, 0, sizeof(rtm));
1350 	rtm.rtm_pid = curproc->p_pid;
1351 	rtm.rtm_flags = RTF_DONE | flags;
1352 	rtm.rtm_errno = error;
1353 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
1354 	if (m == NULL)
1355 		return;
1356 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1357 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1358 }
1359 
1360 /*
1361  * This routine is called to generate a message from the routing
1362  * socket indicating that the status of a network interface has changed.
1363  */
1364 void
1365 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
1366 {
1367 	struct if_xmsghdr ifm;
1368 	struct mbuf *m;
1369 	struct rt_addrinfo info;
1370 
1371 	COMPATCALL(rt_ifmsg, (ifp));
1372 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1373 		return;
1374 	(void)memset(&info, 0, sizeof(info));
1375 	(void)memset(&ifm, 0, sizeof(ifm));
1376 	ifm.ifm_index = ifp->if_index;
1377 	ifm.ifm_flags = ifp->if_flags;
1378 	ifm.ifm_data = ifp->if_data;
1379 	ifm.ifm_addrs = 0;
1380 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
1381 	if (m == NULL)
1382 		return;
1383 	COMPATNAME(route_enqueue)(m, 0);
1384 #ifdef COMPAT_14
1385 	compat_14_rt_oifmsg(ifp);
1386 #endif
1387 #ifdef COMPAT_50
1388 	compat_50_rt_oifmsg(ifp);
1389 #endif
1390 }
1391 
1392 #ifndef COMPAT_RTSOCK
1393 static int
1394 if_addrflags(struct ifaddr *ifa)
1395 {
1396 
1397 	switch (ifa->ifa_addr->sa_family) {
1398 #ifdef INET
1399 	case AF_INET:
1400 		return ((struct in_ifaddr *)ifa)->ia4_flags;
1401 #endif
1402 #ifdef INET6
1403 	case AF_INET6:
1404 		return ((struct in6_ifaddr *)ifa)->ia6_flags;
1405 #endif
1406 	default:
1407 		return 0;
1408 	}
1409 }
1410 #endif
1411 
1412 /*
1413  * This is called to generate messages from the routing socket
1414  * indicating a network interface has had addresses associated with it.
1415  * if we ever reverse the logic and replace messages TO the routing
1416  * socket indicate a request to configure interfaces, then it will
1417  * be unnecessary as the routing socket will automatically generate
1418  * copies of it.
1419  */
1420 void
1421 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
1422     struct rtentry *rt)
1423 {
1424 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
1425 	struct rt_addrinfo info;
1426 	const struct sockaddr *sa;
1427 	int pass;
1428 	struct mbuf *m;
1429 	struct ifnet *ifp;
1430 	struct rt_xmsghdr rtm;
1431 	struct ifa_xmsghdr ifam;
1432 	int ncmd;
1433 
1434 	KASSERT(ifa != NULL);
1435 	KASSERT(ifa->ifa_addr != NULL);
1436 	ifp = ifa->ifa_ifp;
1437 #ifdef SCTP
1438 	if (cmd == RTM_ADD) {
1439 		sctp_add_ip_address(ifa);
1440 	} else if (cmd == RTM_DELETE) {
1441 		sctp_delete_ip_address(ifa);
1442 	}
1443 #endif
1444 
1445 	COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
1446 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1447 		return;
1448 	for (pass = 1; pass < 3; pass++) {
1449 		memset(&info, 0, sizeof(info));
1450 		switch (cmdpass(cmd, pass)) {
1451 		case cmdpass(RTM_ADD, 1):
1452 		case cmdpass(RTM_CHANGE, 1):
1453 		case cmdpass(RTM_DELETE, 2):
1454 		case cmdpass(RTM_NEWADDR, 1):
1455 		case cmdpass(RTM_DELADDR, 1):
1456 		case cmdpass(RTM_CHGADDR, 1):
1457 			switch (cmd) {
1458 			case RTM_ADD:
1459 				ncmd = RTM_XNEWADDR;
1460 				break;
1461 			case RTM_DELETE:
1462 				ncmd = RTM_XDELADDR;
1463 				break;
1464 			case RTM_CHANGE:
1465 				ncmd = RTM_XCHGADDR;
1466 				break;
1467 			case RTM_NEWADDR:
1468 				ncmd = RTM_XNEWADDR;
1469 				break;
1470 			case RTM_DELADDR:
1471 				ncmd = RTM_XDELADDR;
1472 				break;
1473 			case RTM_CHGADDR:
1474 				ncmd = RTM_XCHGADDR;
1475 				break;
1476 			default:
1477 				panic("%s: unknown command %d", __func__, cmd);
1478 			}
1479 #ifdef COMPAT_70
1480 			compat_70_rt_newaddrmsg1(ncmd, ifa);
1481 #endif
1482 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1483 			KASSERT(ifp->if_dl != NULL);
1484 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1485 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1486 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1487 			memset(&ifam, 0, sizeof(ifam));
1488 			ifam.ifam_index = ifp->if_index;
1489 			ifam.ifam_metric = ifa->ifa_metric;
1490 			ifam.ifam_flags = ifa->ifa_flags;
1491 #ifndef COMPAT_RTSOCK
1492 			ifam.ifam_pid = curproc->p_pid;
1493 			ifam.ifam_addrflags = if_addrflags(ifa);
1494 #endif
1495 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
1496 			if (m == NULL)
1497 				continue;
1498 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
1499 			    info.rti_addrs;
1500 			break;
1501 		case cmdpass(RTM_ADD, 2):
1502 		case cmdpass(RTM_CHANGE, 2):
1503 		case cmdpass(RTM_DELETE, 1):
1504 			if (rt == NULL)
1505 				continue;
1506 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1507 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
1508 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1509 			memset(&rtm, 0, sizeof(rtm));
1510 			rtm.rtm_pid = curproc->p_pid;
1511 			rtm.rtm_index = ifp->if_index;
1512 			rtm.rtm_flags |= rt->rt_flags;
1513 			rtm.rtm_errno = error;
1514 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
1515 			if (m == NULL)
1516 				continue;
1517 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1518 			break;
1519 		default:
1520 			continue;
1521 		}
1522 		KASSERTMSG(m != NULL, "called with wrong command");
1523 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1524 	}
1525 #undef cmdpass
1526 
1527 }
1528 
1529 static struct mbuf *
1530 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1531     struct rt_addrinfo *info)
1532 {
1533 	struct if_xannouncemsghdr ifan;
1534 
1535 	memset(info, 0, sizeof(*info));
1536 	memset(&ifan, 0, sizeof(ifan));
1537 	ifan.ifan_index = ifp->if_index;
1538 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
1539 	ifan.ifan_what = what;
1540 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
1541 }
1542 
1543 /*
1544  * This is called to generate routing socket messages indicating
1545  * network interface arrival and departure.
1546  */
1547 void
1548 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
1549 {
1550 	struct mbuf *m;
1551 	struct rt_addrinfo info;
1552 
1553 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
1554 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1555 		return;
1556 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1557 	if (m == NULL)
1558 		return;
1559 	COMPATNAME(route_enqueue)(m, 0);
1560 }
1561 
1562 /*
1563  * This is called to generate routing socket messages indicating
1564  * IEEE80211 wireless events.
1565  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1566  */
1567 void
1568 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
1569 	size_t data_len)
1570 {
1571 	struct mbuf *m;
1572 	struct rt_addrinfo info;
1573 
1574 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
1575 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1576 		return;
1577 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1578 	if (m == NULL)
1579 		return;
1580 	/*
1581 	 * Append the ieee80211 data.  Try to stick it in the
1582 	 * mbuf containing the ifannounce msg; otherwise allocate
1583 	 * a new mbuf and append.
1584 	 *
1585 	 * NB: we assume m is a single mbuf.
1586 	 */
1587 	if (data_len > M_TRAILINGSPACE(m)) {
1588 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1589 		if (n == NULL) {
1590 			m_freem(m);
1591 			return;
1592 		}
1593 		(void)memcpy(mtod(n, void *), data, data_len);
1594 		n->m_len = data_len;
1595 		m->m_next = n;
1596 	} else if (data_len > 0) {
1597 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
1598 		m->m_len += data_len;
1599 	}
1600 	if (m->m_flags & M_PKTHDR)
1601 		m->m_pkthdr.len += data_len;
1602 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
1603 	COMPATNAME(route_enqueue)(m, 0);
1604 }
1605 
1606 #ifndef COMPAT_RTSOCK
1607 /*
1608  * Send a routing message as mimicing that a cloned route is added.
1609  */
1610 void
1611 rt_clonedmsg(const struct sockaddr *dst, const struct ifnet *ifp,
1612     const struct rtentry *rt)
1613 {
1614 	struct rt_addrinfo info;
1615 	/* Mimic flags exactly */
1616 #define RTF_LLINFO	0x400
1617 #define RTF_CLONED	0x2000
1618 	int flags = RTF_UP | RTF_HOST | RTF_DONE | RTF_LLINFO | RTF_CLONED;
1619 	union {
1620 		struct sockaddr sa;
1621 		struct sockaddr_storage ss;
1622 		struct sockaddr_dl sdl;
1623 	} u;
1624 	uint8_t namelen = strlen(ifp->if_xname);
1625 	uint8_t addrlen = ifp->if_addrlen;
1626 
1627 	if (rt == NULL)
1628 		return; /* XXX */
1629 
1630 	memset(&info, 0, sizeof(info));
1631 	info.rti_info[RTAX_DST] = dst;
1632 	sockaddr_dl_init(&u.sdl, sizeof(u.ss), ifp->if_index, ifp->if_type,
1633 	    NULL, namelen, NULL, addrlen);
1634 	info.rti_info[RTAX_GATEWAY] = &u.sa;
1635 
1636 	rt_missmsg(RTM_ADD, &info, flags, 0);
1637 #undef RTF_LLINFO
1638 #undef RTF_CLONED
1639 }
1640 #endif /* COMPAT_RTSOCK */
1641 
1642 /*
1643  * This is used in dumping the kernel table via sysctl().
1644  */
1645 static int
1646 sysctl_dumpentry(struct rtentry *rt, void *v)
1647 {
1648 	struct rt_walkarg *w = v;
1649 	int error = 0, size;
1650 	struct rt_addrinfo info;
1651 
1652 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1653 		return 0;
1654 	memset(&info, 0, sizeof(info));
1655 	info.rti_info[RTAX_DST] = rt_getkey(rt);
1656 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1657 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1658 	info.rti_info[RTAX_TAG] = rt_gettag(rt);
1659 	if (rt->rt_ifp) {
1660 		const struct ifaddr *rtifa;
1661 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
1662 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
1663 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
1664 		 * --dyoung
1665 		 */
1666 		rtifa = rt_get_ifa(rt);
1667 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
1668 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1669 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
1670 	}
1671 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
1672 		return error;
1673 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1674 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
1675 
1676 		rtm->rtm_flags = rt->rt_flags;
1677 		rtm->rtm_use = rt->rt_use;
1678 		rtm_setmetrics(rt, rtm);
1679 		KASSERT(rt->rt_ifp != NULL);
1680 		rtm->rtm_index = rt->rt_ifp->if_index;
1681 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1682 		rtm->rtm_addrs = info.rti_addrs;
1683 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1684 			w->w_where = NULL;
1685 		else
1686 			w->w_where = (char *)w->w_where + size;
1687 	}
1688 	return error;
1689 }
1690 
1691 static int
1692 sysctl_iflist_if(struct ifnet *ifp, struct rt_walkarg *w,
1693     struct rt_addrinfo *info, size_t len)
1694 {
1695 	struct if_xmsghdr *ifm;
1696 	int error;
1697 
1698 	ifm = (struct if_xmsghdr *)w->w_tmem;
1699 	ifm->ifm_index = ifp->if_index;
1700 	ifm->ifm_flags = ifp->if_flags;
1701 	ifm->ifm_data = ifp->if_data;
1702 	ifm->ifm_addrs = info->rti_addrs;
1703 	if ((error = copyout(ifm, w->w_where, len)) == 0)
1704 		w->w_where = (char *)w->w_where + len;
1705 	return error;
1706 }
1707 
1708 static int
1709 sysctl_iflist_addr(struct rt_walkarg *w, struct ifaddr *ifa,
1710      struct rt_addrinfo *info)
1711 {
1712 	int len, error;
1713 
1714 	if ((error = rt_msg2(RTM_XNEWADDR, info, 0, w, &len)))
1715 		return error;
1716 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1717 		struct ifa_xmsghdr *ifam;
1718 
1719 		ifam = (struct ifa_xmsghdr *)w->w_tmem;
1720 		ifam->ifam_index = ifa->ifa_ifp->if_index;
1721 		ifam->ifam_flags = ifa->ifa_flags;
1722 		ifam->ifam_metric = ifa->ifa_metric;
1723 		ifam->ifam_addrs = info->rti_addrs;
1724 #ifndef COMPAT_RTSOCK
1725 		ifam->ifam_pid = 0;
1726 		ifam->ifam_addrflags = if_addrflags(ifa);
1727 #endif
1728 		if ((error = copyout(w->w_tmem, w->w_where, len)) == 0)
1729 			w->w_where = (char *)w->w_where + len;
1730 	}
1731 	return error;
1732 }
1733 
1734 static int
1735 sysctl_iflist(int af, struct rt_walkarg *w, int type)
1736 {
1737 	struct ifnet *ifp;
1738 	struct ifaddr *ifa;
1739 	struct	rt_addrinfo info;
1740 	int	cmd, len, error = 0;
1741 	int	(*iflist_if)(struct ifnet *, struct rt_walkarg *,
1742 			     struct rt_addrinfo *, size_t);
1743 	int	(*iflist_addr)(struct rt_walkarg *, struct ifaddr *,
1744 			       struct rt_addrinfo *);
1745 	int s;
1746 	struct psref psref;
1747 	int bound;
1748 
1749 	switch (type) {
1750 	case NET_RT_IFLIST:
1751 		cmd = RTM_IFINFO;
1752 		iflist_if = sysctl_iflist_if;
1753 		iflist_addr = sysctl_iflist_addr;
1754 		break;
1755 #ifdef COMPAT_14
1756 	case NET_RT_OOOIFLIST:
1757 		cmd = RTM_OOIFINFO;
1758 		iflist_if = compat_14_iflist;
1759 		iflist_addr = compat_70_iflist_addr;
1760 		break;
1761 #endif
1762 #ifdef COMPAT_50
1763 	case NET_RT_OOIFLIST:
1764 		cmd = RTM_OIFINFO;
1765 		iflist_if = compat_50_iflist;
1766 		iflist_addr = compat_70_iflist_addr;
1767 		break;
1768 #endif
1769 #ifdef COMPAT_70
1770 	case NET_RT_OIFLIST:
1771 		cmd = RTM_IFINFO;
1772 		iflist_if = sysctl_iflist_if;
1773 		iflist_addr = compat_70_iflist_addr;
1774 		break;
1775 #endif
1776 	default:
1777 #ifdef RTSOCK_DEBUG
1778 		printf("%s: unsupported IFLIST type %d\n", __func__, type);
1779 #endif
1780 		return EINVAL;
1781 	}
1782 
1783 	memset(&info, 0, sizeof(info));
1784 
1785 	bound = curlwp_bind();
1786 	s = pserialize_read_enter();
1787 	IFNET_READER_FOREACH(ifp) {
1788 		int _s;
1789 		if (w->w_arg && w->w_arg != ifp->if_index)
1790 			continue;
1791 		if (IFADDR_READER_EMPTY(ifp))
1792 			continue;
1793 
1794 		if_acquire(ifp, &psref);
1795 		pserialize_read_exit(s);
1796 
1797 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1798 		if ((error = rt_msg2(cmd, &info, NULL, w, &len)) != 0)
1799 			goto release_exit;
1800 		info.rti_info[RTAX_IFP] = NULL;
1801 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1802 			if ((error = iflist_if(ifp, w, &info, len)) != 0)
1803 				goto release_exit;
1804 		}
1805 		_s = pserialize_read_enter();
1806 		IFADDR_READER_FOREACH(ifa, ifp) {
1807 			struct psref _psref;
1808 			if (af && af != ifa->ifa_addr->sa_family)
1809 				continue;
1810 			ifa_acquire(ifa, &_psref);
1811 			pserialize_read_exit(_s);
1812 
1813 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1814 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1815 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1816 			error = iflist_addr(w, ifa, &info);
1817 
1818 			_s = pserialize_read_enter();
1819 			ifa_release(ifa, &_psref);
1820 			if (error != 0) {
1821 				pserialize_read_exit(_s);
1822 				goto release_exit;
1823 			}
1824 		}
1825 		pserialize_read_exit(_s);
1826 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1827 		    info.rti_info[RTAX_BRD] = NULL;
1828 
1829 		s = pserialize_read_enter();
1830 		if_release(ifp, &psref);
1831 	}
1832 	pserialize_read_exit(s);
1833 	curlwp_bindx(bound);
1834 
1835 	return 0;
1836 
1837 release_exit:
1838 	if_release(ifp, &psref);
1839 	curlwp_bindx(bound);
1840 	return error;
1841 }
1842 
1843 static int
1844 sysctl_rtable(SYSCTLFN_ARGS)
1845 {
1846 	void 	*where = oldp;
1847 	size_t	*given = oldlenp;
1848 	int	i, s, error = EINVAL;
1849 	u_char  af;
1850 	struct	rt_walkarg w;
1851 
1852 	if (namelen == 1 && name[0] == CTL_QUERY)
1853 		return sysctl_query(SYSCTLFN_CALL(rnode));
1854 
1855 	if (newp)
1856 		return EPERM;
1857 	if (namelen != 3)
1858 		return EINVAL;
1859 	af = name[0];
1860 	w.w_tmemneeded = 0;
1861 	w.w_tmemsize = 0;
1862 	w.w_tmem = NULL;
1863 again:
1864 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1865 	if (w.w_tmemneeded) {
1866 		w.w_tmem = kmem_alloc(w.w_tmemneeded, KM_SLEEP);
1867 		w.w_tmemsize = w.w_tmemneeded;
1868 		w.w_tmemneeded = 0;
1869 	}
1870 	w.w_op = name[1];
1871 	w.w_arg = name[2];
1872 	w.w_given = *given;
1873 	w.w_needed = 0 - w.w_given;
1874 	w.w_where = where;
1875 
1876 	s = splsoftnet();
1877 	switch (w.w_op) {
1878 
1879 	case NET_RT_DUMP:
1880 	case NET_RT_FLAGS:
1881 #if defined(INET) || defined(INET6)
1882 		/*
1883 		 * take care of llinfo entries, the caller must
1884 		 * specify an AF
1885 		 */
1886 		if (w.w_op == NET_RT_FLAGS &&
1887 		    (w.w_arg == 0 || w.w_arg & RTF_LLDATA)) {
1888 			if (af != 0)
1889 				error = lltable_sysctl_dump(af, &w);
1890 			else
1891 				error = EINVAL;
1892 			break;
1893 		}
1894 #endif
1895 
1896 		for (i = 1; i <= AF_MAX; i++) {
1897 			if (af == 0 || af == i) {
1898 				error = rt_walktree(i, sysctl_dumpentry, &w);
1899 				if (error != 0)
1900 					break;
1901 #if defined(INET) || defined(INET6)
1902 				/*
1903 				 * Return ARP/NDP entries too for
1904 				 * backward compatibility.
1905 				 */
1906 				error = lltable_sysctl_dump(i, &w);
1907 				if (error != 0)
1908 					break;
1909 #endif
1910 			}
1911 		}
1912 		break;
1913 
1914 #ifdef COMPAT_14
1915 	case NET_RT_OOOIFLIST:
1916 		error = sysctl_iflist(af, &w, w.w_op);
1917 		break;
1918 #endif
1919 #ifdef COMPAT_50
1920 	case NET_RT_OOIFLIST:
1921 		error = sysctl_iflist(af, &w, w.w_op);
1922 		break;
1923 #endif
1924 #ifdef COMPAT_70
1925 	case NET_RT_OIFLIST:
1926 		error = sysctl_iflist(af, &w, w.w_op);
1927 		break;
1928 #endif
1929 	case NET_RT_IFLIST:
1930 		error = sysctl_iflist(af, &w, w.w_op);
1931 		break;
1932 	}
1933 	splx(s);
1934 
1935 	/* check to see if we couldn't allocate memory with NOWAIT */
1936 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1937 		goto again;
1938 
1939 	if (w.w_tmem)
1940 		kmem_free(w.w_tmem, w.w_tmemsize);
1941 	w.w_needed += w.w_given;
1942 	if (where) {
1943 		*given = (char *)w.w_where - (char *)where;
1944 		if (*given < w.w_needed)
1945 			return ENOMEM;
1946 	} else {
1947 		*given = (11 * w.w_needed) / 10;
1948 	}
1949 	return error;
1950 }
1951 
1952 /*
1953  * Routing message software interrupt routine
1954  */
1955 static void
1956 COMPATNAME(route_intr)(void *cookie)
1957 {
1958 	struct sockproto proto = { .sp_family = PF_XROUTE, };
1959 	struct route_info * const ri = &COMPATNAME(route_info);
1960 	struct mbuf *m;
1961 
1962 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
1963 	for (;;) {
1964 		IFQ_LOCK(&ri->ri_intrq);
1965 		IF_DEQUEUE(&ri->ri_intrq, m);
1966 		IFQ_UNLOCK(&ri->ri_intrq);
1967 		if (m == NULL)
1968 			break;
1969 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
1970 #ifdef NET_MPSAFE
1971 		mutex_enter(rt_so_mtx);
1972 #endif
1973 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst, &rt_rawcb);
1974 #ifdef NET_MPSAFE
1975 		mutex_exit(rt_so_mtx);
1976 #endif
1977 	}
1978 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1979 }
1980 
1981 /*
1982  * Enqueue a message to the software interrupt routine.
1983  */
1984 void
1985 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
1986 {
1987 	struct route_info * const ri = &COMPATNAME(route_info);
1988 	int wasempty;
1989 
1990 	IFQ_LOCK(&ri->ri_intrq);
1991 	if (IF_QFULL(&ri->ri_intrq)) {
1992 		printf("%s: queue full, dropped message\n", __func__);
1993 		IF_DROP(&ri->ri_intrq);
1994 		IFQ_UNLOCK(&ri->ri_intrq);
1995 		m_freem(m);
1996 	} else {
1997 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
1998 		M_SETCTX(m, (uintptr_t)family);
1999 		IF_ENQUEUE(&ri->ri_intrq, m);
2000 		IFQ_UNLOCK(&ri->ri_intrq);
2001 		if (wasempty) {
2002 			kpreempt_disable();
2003 			softint_schedule(ri->ri_sih);
2004 			kpreempt_enable();
2005 		}
2006 	}
2007 }
2008 
2009 static void
2010 COMPATNAME(route_init)(void)
2011 {
2012 	struct route_info * const ri = &COMPATNAME(route_info);
2013 
2014 #ifndef COMPAT_RTSOCK
2015 	rt_init();
2016 #endif
2017 #ifdef NET_MPSAFE
2018 	rt_so_mtx = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
2019 
2020 	cv_init(&rt_update_cv, "rtsock_cv");
2021 #endif
2022 
2023 	sysctl_net_route_setup(NULL);
2024 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
2025 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
2026 	    COMPATNAME(route_intr), NULL);
2027 	IFQ_LOCK_INIT(&ri->ri_intrq);
2028 }
2029 
2030 /*
2031  * Definitions of protocols supported in the ROUTE domain.
2032  */
2033 #ifndef COMPAT_RTSOCK
2034 PR_WRAP_USRREQS(route);
2035 #else
2036 PR_WRAP_USRREQS(compat_50_route);
2037 #endif
2038 
2039 static const struct pr_usrreqs route_usrreqs = {
2040 	.pr_attach	= COMPATNAME(route_attach_wrapper),
2041 	.pr_detach	= COMPATNAME(route_detach_wrapper),
2042 	.pr_accept	= COMPATNAME(route_accept_wrapper),
2043 	.pr_bind	= COMPATNAME(route_bind_wrapper),
2044 	.pr_listen	= COMPATNAME(route_listen_wrapper),
2045 	.pr_connect	= COMPATNAME(route_connect_wrapper),
2046 	.pr_connect2	= COMPATNAME(route_connect2_wrapper),
2047 	.pr_disconnect	= COMPATNAME(route_disconnect_wrapper),
2048 	.pr_shutdown	= COMPATNAME(route_shutdown_wrapper),
2049 	.pr_abort	= COMPATNAME(route_abort_wrapper),
2050 	.pr_ioctl	= COMPATNAME(route_ioctl_wrapper),
2051 	.pr_stat	= COMPATNAME(route_stat_wrapper),
2052 	.pr_peeraddr	= COMPATNAME(route_peeraddr_wrapper),
2053 	.pr_sockaddr	= COMPATNAME(route_sockaddr_wrapper),
2054 	.pr_rcvd	= COMPATNAME(route_rcvd_wrapper),
2055 	.pr_recvoob	= COMPATNAME(route_recvoob_wrapper),
2056 	.pr_send	= COMPATNAME(route_send_wrapper),
2057 	.pr_sendoob	= COMPATNAME(route_sendoob_wrapper),
2058 	.pr_purgeif	= COMPATNAME(route_purgeif_wrapper),
2059 };
2060 
2061 static const struct protosw COMPATNAME(route_protosw)[] = {
2062 	{
2063 		.pr_type = SOCK_RAW,
2064 		.pr_domain = &COMPATNAME(routedomain),
2065 		.pr_flags = PR_ATOMIC|PR_ADDR,
2066 		.pr_input = raw_input,
2067 		.pr_ctlinput = raw_ctlinput,
2068 		.pr_ctloutput = route_ctloutput,
2069 		.pr_usrreqs = &route_usrreqs,
2070 		.pr_init = rt_pr_init,
2071 	},
2072 };
2073 
2074 struct domain COMPATNAME(routedomain) = {
2075 	.dom_family = PF_XROUTE,
2076 	.dom_name = DOMAINNAME,
2077 	.dom_init = COMPATNAME(route_init),
2078 	.dom_protosw = COMPATNAME(route_protosw),
2079 	.dom_protoswNPROTOSW =
2080 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
2081 };
2082 
2083 static void
2084 sysctl_net_route_setup(struct sysctllog **clog)
2085 {
2086 	const struct sysctlnode *rnode = NULL;
2087 
2088 	sysctl_createv(clog, 0, NULL, &rnode,
2089 		       CTLFLAG_PERMANENT,
2090 		       CTLTYPE_NODE, DOMAINNAME,
2091 		       SYSCTL_DESCR("PF_ROUTE information"),
2092 		       NULL, 0, NULL, 0,
2093 		       CTL_NET, PF_XROUTE, CTL_EOL);
2094 
2095 	sysctl_createv(clog, 0, NULL, NULL,
2096 		       CTLFLAG_PERMANENT,
2097 		       CTLTYPE_NODE, "rtable",
2098 		       SYSCTL_DESCR("Routing table information"),
2099 		       sysctl_rtable, 0, NULL, 0,
2100 		       CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
2101 
2102 	sysctl_createv(clog, 0, &rnode, NULL,
2103 		       CTLFLAG_PERMANENT,
2104 		       CTLTYPE_STRUCT, "stats",
2105 		       SYSCTL_DESCR("Routing statistics"),
2106 		       NULL, 0, &rtstat, sizeof(rtstat),
2107 		       CTL_CREATE, CTL_EOL);
2108 }
2109