xref: /netbsd-src/sys/net/rtsock_shared.c (revision f3cfa6f6ce31685c6c4a758bc430e69eb99f50a4)
1 /*	$NetBSD: rtsock_shared.c,v 1.9 2019/05/03 02:10:58 pgoyette 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_shared.c,v 1.9 2019/05/03 02:10:58 pgoyette Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_net_mpsafe.h"
69 #endif
70 
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/proc.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/domain.h>
77 #include <sys/protosw.h>
78 #include <sys/sysctl.h>
79 #include <sys/kauth.h>
80 #include <sys/kmem.h>
81 #include <sys/intr.h>
82 #include <sys/condvar.h>
83 #include <sys/compat_stub.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 #include <compat/net/if.h>
97 #include <compat/net/route.h>
98 
99 #ifdef COMPAT_RTSOCK
100 /*
101  * These are used when #include-d from compat/common/rtsock_50.c
102  */
103 #define	RTM_XVERSION	RTM_OVERSION
104 #define	RTM_XNEWADDR	RTM_ONEWADDR
105 #define	RTM_XDELADDR	RTM_ODELADDR
106 #define	RTM_XCHGADDR	RTM_OCHGADDR
107 #define	RT_XADVANCE(a,b) RT_OADVANCE(a,b)
108 #define	RT_XROUNDUP(n)	RT_OROUNDUP(n)
109 #define	PF_XROUTE	PF_OROUTE
110 #define	rt_xmsghdr	rt_msghdr50
111 #define	if_xmsghdr	if_msghdr	/* if_msghdr50 is for RTM_OIFINFO */
112 #define	ifa_xmsghdr	ifa_msghdr50
113 #define	if_xannouncemsghdr	if_announcemsghdr50
114 #define	COMPATNAME(x)	compat_50_ ## x
115 #define	DOMAINNAME	"oroute"
116 #define	COMPATCALL(name, args)		\
117 	MODULE_HOOK_CALL_VOID(rtsock_ ## name ## _50_hook, args, __nothing);
118 #define	RTS_CTASSERT(x)	__nothing
119 CTASSERT(sizeof(struct ifa_xmsghdr) == 20);
120 DOMAIN_DEFINE(compat_50_routedomain); /* forward declare and add to link set */
121 #else /* COMPAT_RTSOCK */
122 /*
123  * These are used when #include-d from compat/common/rtsock_50.c
124  */
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 #define	COMPATCALL(name, args)	__nothing;
139 #define	RTS_CTASSERT(x)	CTASSERT(x)
140 CTASSERT(sizeof(struct ifa_xmsghdr) == 32);
141 DOMAIN_DEFINE(routedomain); /* forward declare and add to link set */
142 #endif /* COMPAT_RTSOCK */
143 
144 #ifdef RTSOCK_DEBUG
145 #define RT_IN_PRINT(info, b, a) (in_print((b), sizeof(b), \
146     &((const struct sockaddr_in *)(info)->rti_info[(a)])->sin_addr), (b))
147 #endif /* RTSOCK_DEBUG */
148 
149 struct route_info COMPATNAME(route_info) = {
150 	.ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, },
151 	.ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, },
152 	.ri_maxqlen = IFQ_MAXLEN,
153 };
154 
155 static void COMPATNAME(route_init)(void);
156 static int COMPATNAME(route_output)(struct mbuf *, struct socket *);
157 
158 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
159 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
160     struct rt_addrinfo *);
161 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
162 static void _rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *);
163 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *);
164 static void rt_adjustcount(int, int);
165 
166 static const struct protosw COMPATNAME(route_protosw)[];
167 
168 struct routecb {
169 	struct rawcb	rocb_rcb;
170 	unsigned int	rocb_msgfilter;
171 #define	RTMSGFILTER(m)	(1U << (m))
172 };
173 #define sotoroutecb(so)	((struct routecb *)(so)->so_pcb)
174 
175 static struct rawcbhead rt_rawcb;
176 #ifdef NET_MPSAFE
177 static kmutex_t *rt_so_mtx;
178 
179 static bool rt_updating = false;
180 static kcondvar_t rt_update_cv;
181 #endif
182 
183 static void
184 rt_adjustcount(int af, int cnt)
185 {
186 	struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
187 
188 	cb->any_count += cnt;
189 
190 	switch (af) {
191 	case AF_INET:
192 		cb->ip_count += cnt;
193 		return;
194 #ifdef INET6
195 	case AF_INET6:
196 		cb->ip6_count += cnt;
197 		return;
198 #endif
199 	case AF_MPLS:
200 		cb->mpls_count += cnt;
201 		return;
202 	}
203 }
204 
205 static int
206 COMPATNAME(route_filter)(struct mbuf *m, struct sockproto *proto,
207     struct rawcb *rp)
208 {
209 	struct routecb *rop = (struct routecb *)rp;
210 	struct rt_xmsghdr *rtm;
211 
212 	KASSERT(m != NULL);
213 	KASSERT(proto != NULL);
214 	KASSERT(rp != NULL);
215 
216 	/* Wrong family for this socket. */
217 	if (proto->sp_family != PF_ROUTE)
218 		return ENOPROTOOPT;
219 
220 	/* If no filter set, just return. */
221 	if (rop->rocb_msgfilter == 0)
222 		return 0;
223 
224 	/* Ensure we can access rtm_type */
225 	if (m->m_len <
226 	    offsetof(struct rt_xmsghdr, rtm_type) + sizeof(rtm->rtm_type))
227 		return EINVAL;
228 
229 	rtm = mtod(m, struct rt_xmsghdr *);
230 	/* If the rtm type is filtered out, return a positive. */
231 	if (!(rop->rocb_msgfilter & RTMSGFILTER(rtm->rtm_type)))
232 		return EEXIST;
233 
234 	/* Passed the filter. */
235 	return 0;
236 }
237 
238 static void
239 rt_pr_init(void)
240 {
241 
242 	LIST_INIT(&rt_rawcb);
243 }
244 
245 static int
246 COMPATNAME(route_attach)(struct socket *so, int proto)
247 {
248 	struct rawcb *rp;
249 	struct routecb *rop;
250 	int s, error;
251 
252 	KASSERT(sotorawcb(so) == NULL);
253 	rop = kmem_zalloc(sizeof(*rop), KM_SLEEP);
254 	rp = &rop->rocb_rcb;
255 	rp->rcb_len = sizeof(*rop);
256 	so->so_pcb = rp;
257 
258 	s = splsoftnet();
259 
260 #ifdef NET_MPSAFE
261 	KASSERT(so->so_lock == NULL);
262 	mutex_obj_hold(rt_so_mtx);
263 	so->so_lock = rt_so_mtx;
264 	solock(so);
265 #endif
266 
267 	if ((error = raw_attach(so, proto, &rt_rawcb)) == 0) {
268 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
269 		rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
270 		rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
271 		rp->rcb_filter = COMPATNAME(route_filter);
272 	}
273 	splx(s);
274 
275 	if (error) {
276 		kmem_free(rop, sizeof(*rop));
277 		so->so_pcb = NULL;
278 		return error;
279 	}
280 
281 	soisconnected(so);
282 	so->so_options |= SO_USELOOPBACK;
283 	KASSERT(solocked(so));
284 
285 	return error;
286 }
287 
288 static void
289 COMPATNAME(route_detach)(struct socket *so)
290 {
291 	struct rawcb *rp = sotorawcb(so);
292 	int s;
293 
294 	KASSERT(rp != NULL);
295 	KASSERT(solocked(so));
296 
297 	s = splsoftnet();
298 	rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
299 	raw_detach(so);
300 	splx(s);
301 }
302 
303 static int
304 COMPATNAME(route_accept)(struct socket *so, struct sockaddr *nam)
305 {
306 	KASSERT(solocked(so));
307 
308 	panic("route_accept");
309 
310 	return EOPNOTSUPP;
311 }
312 
313 static int
314 COMPATNAME(route_bind)(struct socket *so, struct sockaddr *nam, struct lwp *l)
315 {
316 	KASSERT(solocked(so));
317 
318 	return EOPNOTSUPP;
319 }
320 
321 static int
322 COMPATNAME(route_listen)(struct socket *so, struct lwp *l)
323 {
324 	KASSERT(solocked(so));
325 
326 	return EOPNOTSUPP;
327 }
328 
329 static int
330 COMPATNAME(route_connect)(struct socket *so, struct sockaddr *nam, struct lwp *l)
331 {
332 	KASSERT(solocked(so));
333 
334 	return EOPNOTSUPP;
335 }
336 
337 static int
338 COMPATNAME(route_connect2)(struct socket *so, struct socket *so2)
339 {
340 	KASSERT(solocked(so));
341 
342 	return EOPNOTSUPP;
343 }
344 
345 static int
346 COMPATNAME(route_disconnect)(struct socket *so)
347 {
348 	struct rawcb *rp = sotorawcb(so);
349 	int s;
350 
351 	KASSERT(solocked(so));
352 	KASSERT(rp != NULL);
353 
354 	s = splsoftnet();
355 	soisdisconnected(so);
356 	raw_disconnect(rp);
357 	splx(s);
358 
359 	return 0;
360 }
361 
362 static int
363 COMPATNAME(route_shutdown)(struct socket *so)
364 {
365 	int s;
366 
367 	KASSERT(solocked(so));
368 
369 	/*
370 	 * Mark the connection as being incapable of further input.
371 	 */
372 	s = splsoftnet();
373 	socantsendmore(so);
374 	splx(s);
375 	return 0;
376 }
377 
378 static int
379 COMPATNAME(route_abort)(struct socket *so)
380 {
381 	KASSERT(solocked(so));
382 
383 	panic("route_abort");
384 
385 	return EOPNOTSUPP;
386 }
387 
388 static int
389 COMPATNAME(route_ioctl)(struct socket *so, u_long cmd, void *nam,
390     struct ifnet * ifp)
391 {
392 	return EOPNOTSUPP;
393 }
394 
395 static int
396 COMPATNAME(route_stat)(struct socket *so, struct stat *ub)
397 {
398 	KASSERT(solocked(so));
399 
400 	return 0;
401 }
402 
403 static int
404 COMPATNAME(route_peeraddr)(struct socket *so, struct sockaddr *nam)
405 {
406 	struct rawcb *rp = sotorawcb(so);
407 
408 	KASSERT(solocked(so));
409 	KASSERT(rp != NULL);
410 	KASSERT(nam != NULL);
411 
412 	if (rp->rcb_faddr == NULL)
413 		return ENOTCONN;
414 
415 	raw_setpeeraddr(rp, nam);
416 	return 0;
417 }
418 
419 static int
420 COMPATNAME(route_sockaddr)(struct socket *so, struct sockaddr *nam)
421 {
422 	struct rawcb *rp = sotorawcb(so);
423 
424 	KASSERT(solocked(so));
425 	KASSERT(rp != NULL);
426 	KASSERT(nam != NULL);
427 
428 	if (rp->rcb_faddr == NULL)
429 		return ENOTCONN;
430 
431 	raw_setsockaddr(rp, nam);
432 	return 0;
433 }
434 
435 static int
436 COMPATNAME(route_rcvd)(struct socket *so, int flags, struct lwp *l)
437 {
438 	KASSERT(solocked(so));
439 
440 	return EOPNOTSUPP;
441 }
442 
443 static int
444 COMPATNAME(route_recvoob)(struct socket *so, struct mbuf *m, int flags)
445 {
446 	KASSERT(solocked(so));
447 
448 	return EOPNOTSUPP;
449 }
450 
451 static int
452 COMPATNAME(route_send)(struct socket *so, struct mbuf *m,
453     struct sockaddr *nam, struct mbuf *control, struct lwp *l)
454 {
455 	int error = 0;
456 	int s;
457 
458 	KASSERT(solocked(so));
459 	KASSERT(so->so_proto == &COMPATNAME(route_protosw)[0]);
460 
461 	s = splsoftnet();
462 	error = raw_send(so, m, nam, control, l, &COMPATNAME(route_output));
463 	splx(s);
464 
465 	return error;
466 }
467 
468 static int
469 COMPATNAME(route_sendoob)(struct socket *so, struct mbuf *m,
470     struct mbuf *control)
471 {
472 	KASSERT(solocked(so));
473 
474 	m_freem(m);
475 	m_freem(control);
476 
477 	return EOPNOTSUPP;
478 }
479 static int
480 COMPATNAME(route_purgeif)(struct socket *so, struct ifnet *ifp)
481 {
482 
483 	panic("route_purgeif");
484 
485 	return EOPNOTSUPP;
486 }
487 
488 #if defined(INET) || defined(INET6)
489 static int
490 route_get_sdl_index(struct rt_addrinfo *info, int *sdl_index)
491 {
492 	struct rtentry *nrt;
493 	int error;
494 
495 	error = rtrequest1(RTM_GET, info, &nrt);
496 	if (error != 0)
497 		return error;
498 	/*
499 	 * nrt->rt_ifp->if_index may not be correct
500 	 * due to changing to ifplo0.
501 	 */
502 	*sdl_index = satosdl(nrt->rt_gateway)->sdl_index;
503 	rt_unref(nrt);
504 
505 	return 0;
506 }
507 #endif
508 
509 static void
510 route_get_sdl(const struct ifnet *ifp, const struct sockaddr *dst,
511     struct sockaddr_dl *sdl, int *flags)
512 {
513 	struct llentry *la;
514 
515 	KASSERT(ifp != NULL);
516 
517 	IF_AFDATA_RLOCK(ifp);
518 	switch (dst->sa_family) {
519 	case AF_INET:
520 		la = lla_lookup(LLTABLE(ifp), 0, dst);
521 		break;
522 	case AF_INET6:
523 		la = lla_lookup(LLTABLE6(ifp), 0, dst);
524 		break;
525 	default:
526 		la = NULL;
527 		KASSERTMSG(0, "Invalid AF=%d\n", dst->sa_family);
528 		break;
529 	}
530 	IF_AFDATA_RUNLOCK(ifp);
531 
532 	void *a = (LLE_IS_VALID(la) && (la->la_flags & LLE_VALID) == LLE_VALID)
533 	    ? &la->ll_addr : NULL;
534 
535 	a = sockaddr_dl_init(sdl, sizeof(*sdl), ifp->if_index, ifp->if_type,
536 	    NULL, 0, a, ifp->if_addrlen);
537 	KASSERT(a != NULL);
538 
539 	if (la != NULL) {
540 		*flags = la->la_flags;
541 		LLE_RUNLOCK(la);
542 	}
543 }
544 
545 static int
546 route_output_report(struct rtentry *rt, struct rt_addrinfo *info,
547     struct rt_xmsghdr *rtm, struct rt_xmsghdr **new_rtm)
548 {
549 	int len, error;
550 
551 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
552 		const struct ifaddr *rtifa;
553 		const struct ifnet *ifp = rt->rt_ifp;
554 
555 		info->rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
556 		/* rtifa used to be simply rt->rt_ifa.
557 		 * If rt->rt_ifa != NULL, then
558 		 * rt_get_ifa() != NULL.  So this
559 		 * ought to still be safe. --dyoung
560 		 */
561 		rtifa = rt_get_ifa(rt);
562 		info->rti_info[RTAX_IFA] = rtifa->ifa_addr;
563 #ifdef RTSOCK_DEBUG
564 		if (info->rti_info[RTAX_IFA]->sa_family == AF_INET) {
565 			char ibuf[INET_ADDRSTRLEN];
566 			char abuf[INET_ADDRSTRLEN];
567 			printf("%s: copying out RTAX_IFA %s "
568 			    "for info->rti_info[RTAX_DST] %s "
569 			    "ifa_getifa %p ifa_seqno %p\n",
570 			    __func__,
571 			    RT_IN_PRINT(info, ibuf, RTAX_IFA),
572 			    RT_IN_PRINT(info, abuf, RTAX_DST),
573 			    (void *)rtifa->ifa_getifa,
574 			    rtifa->ifa_seqno);
575 		}
576 #endif /* RTSOCK_DEBUG */
577 		if (ifp->if_flags & IFF_POINTOPOINT)
578 			info->rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
579 		else
580 			info->rti_info[RTAX_BRD] = NULL;
581 		rtm->rtm_index = ifp->if_index;
582 	}
583 	error = rt_msg2(rtm->rtm_type, info, NULL, NULL, &len);
584 	if (error)
585 		return error;
586 	if (len > rtm->rtm_msglen) {
587 		struct rt_xmsghdr *old_rtm = rtm;
588 		R_Malloc(*new_rtm, struct rt_xmsghdr *, len);
589 		if (*new_rtm == NULL)
590 			return ENOBUFS;
591 		(void)memcpy(*new_rtm, old_rtm, old_rtm->rtm_msglen);
592 		rtm = *new_rtm;
593 	}
594 	(void)rt_msg2(rtm->rtm_type, info, rtm, NULL, 0);
595 	rtm->rtm_flags = rt->rt_flags;
596 	rtm_setmetrics(rt, rtm);
597 	rtm->rtm_addrs = info->rti_addrs;
598 
599 	return 0;
600 }
601 
602 /*ARGSUSED*/
603 int
604 COMPATNAME(route_output)(struct mbuf *m, struct socket *so)
605 {
606 	struct sockproto proto = { .sp_family = PF_XROUTE, };
607 	struct rt_xmsghdr *rtm = NULL;
608 	struct rt_xmsghdr *old_rtm = NULL, *new_rtm = NULL;
609 	struct rtentry *rt = NULL;
610 	struct rtentry *saved_nrt = NULL;
611 	struct rt_addrinfo info;
612 	int len, error = 0;
613 	sa_family_t family;
614 	struct sockaddr_dl sdl;
615 	int bound = curlwp_bind();
616 	bool do_rt_free = false;
617 	struct sockaddr_storage netmask;
618 
619 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
620 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
621 	   (m = m_pullup(m, sizeof(int32_t))) == NULL)) {
622 		error = ENOBUFS;
623 		goto out;
624 	}
625 	if ((m->m_flags & M_PKTHDR) == 0)
626 		panic("%s", __func__);
627 	len = m->m_pkthdr.len;
628 	if (len < sizeof(*rtm) ||
629 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
630 		info.rti_info[RTAX_DST] = NULL;
631 		senderr(EINVAL);
632 	}
633 	R_Malloc(rtm, struct rt_xmsghdr *, len);
634 	if (rtm == NULL) {
635 		info.rti_info[RTAX_DST] = NULL;
636 		senderr(ENOBUFS);
637 	}
638 	m_copydata(m, 0, len, rtm);
639 	if (rtm->rtm_version != RTM_XVERSION) {
640 		info.rti_info[RTAX_DST] = NULL;
641 		senderr(EPROTONOSUPPORT);
642 	}
643 	rtm->rtm_pid = curproc->p_pid;
644 	memset(&info, 0, sizeof(info));
645 	info.rti_addrs = rtm->rtm_addrs;
646 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
647 	    &info)) {
648 		senderr(EINVAL);
649 	}
650 	info.rti_flags = rtm->rtm_flags;
651 #ifdef RTSOCK_DEBUG
652 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
653 		char abuf[INET_ADDRSTRLEN];
654 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
655 		    RT_IN_PRINT(&info, abuf, RTAX_DST));
656 	}
657 #endif /* RTSOCK_DEBUG */
658 	if (info.rti_info[RTAX_DST] == NULL ||
659 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
660 		senderr(EINVAL);
661 	}
662 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
663 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
664 		senderr(EINVAL);
665 	}
666 
667 	/*
668 	 * Verify that the caller has the appropriate privilege; RTM_GET
669 	 * is the only operation the non-superuser is allowed.
670 	 */
671 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
672 	    0, rtm, NULL, NULL) != 0)
673 		senderr(EACCES);
674 
675 	/*
676 	 * route(8) passes a sockaddr truncated with prefixlen.
677 	 * The kernel doesn't expect such sockaddr and need to
678 	 * use a buffer that is big enough for the sockaddr expected
679 	 * (padded with 0's). We keep the original length of the sockaddr.
680 	 */
681 	if (info.rti_info[RTAX_NETMASK]) {
682 		/*
683 		 * Use the family of RTAX_DST, because RTAX_NETMASK
684 		 * can have a zero family if it comes from the radix
685 		 * tree via rt_mask().
686 		 */
687 		socklen_t sa_len = sockaddr_getsize_by_family(
688 		    info.rti_info[RTAX_DST]->sa_family);
689 		socklen_t masklen = sockaddr_getlen(
690 		    info.rti_info[RTAX_NETMASK]);
691 		if (sa_len != 0 && sa_len > masklen) {
692 			KASSERT(sa_len <= sizeof(netmask));
693 			memcpy(&netmask, info.rti_info[RTAX_NETMASK], masklen);
694 			memset((char *)&netmask + masklen, 0, sa_len - masklen);
695 			info.rti_info[RTAX_NETMASK] = sstocsa(&netmask);
696 		}
697 	}
698 
699 	switch (rtm->rtm_type) {
700 
701 	case RTM_ADD:
702 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
703 			senderr(EINVAL);
704 		}
705 #if defined(INET) || defined(INET6)
706 		/* support for new ARP/NDP code with keeping backcompat */
707 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) {
708 			const struct sockaddr_dl *sdlp =
709 			    satocsdl(info.rti_info[RTAX_GATEWAY]);
710 
711 			/* Allow routing requests by interface index */
712 			if (sdlp->sdl_nlen == 0 && sdlp->sdl_alen == 0
713 			    && sdlp->sdl_slen == 0)
714 				goto fallback;
715 			/*
716 			 * Old arp binaries don't set the sdl_index
717 			 * so we have to complement it.
718 			 */
719 			int sdl_index = sdlp->sdl_index;
720 			if (sdl_index == 0) {
721 				error = route_get_sdl_index(&info, &sdl_index);
722 				if (error != 0)
723 					goto fallback;
724 			} else if (
725 			    info.rti_info[RTAX_DST]->sa_family == AF_INET) {
726 				/*
727 				 * XXX workaround for SIN_PROXY case; proxy arp
728 				 * entry should be in an interface that has
729 				 * a network route including the destination,
730 				 * not a local (link) route that may not be a
731 				 * desired place, for example a tap.
732 				 */
733 				const struct sockaddr_inarp *sina =
734 				    (const struct sockaddr_inarp *)
735 				    info.rti_info[RTAX_DST];
736 				if (sina->sin_other & SIN_PROXY) {
737 					error = route_get_sdl_index(&info,
738 					    &sdl_index);
739 					if (error != 0)
740 						goto fallback;
741 				}
742 			}
743 			error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
744 			    rtm->rtm_rmx.rmx_expire, &info, sdl_index);
745 			break;
746 		}
747 	fallback:
748 #endif /* defined(INET) || defined(INET6) */
749 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
750 		if (error == 0) {
751 			_rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
752 			rt_unref(saved_nrt);
753 		}
754 		break;
755 
756 	case RTM_DELETE:
757 #if defined(INET) || defined(INET6)
758 		/* support for new ARP/NDP code */
759 		if (info.rti_info[RTAX_GATEWAY] &&
760 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
761 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
762 			const struct sockaddr_dl *sdlp =
763 			    satocsdl(info.rti_info[RTAX_GATEWAY]);
764 			error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
765 			    rtm->rtm_rmx.rmx_expire, &info, sdlp->sdl_index);
766 			rtm->rtm_flags &= ~RTF_UP;
767 			break;
768 		}
769 #endif
770 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
771 		if (error != 0)
772 			break;
773 
774 		rt = saved_nrt;
775 		do_rt_free = true;
776 		info.rti_info[RTAX_DST] = rt_getkey(rt);
777 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
778 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
779 		info.rti_info[RTAX_TAG] = rt_gettag(rt);
780 		error = route_output_report(rt, &info, rtm, &new_rtm);
781 		if (error)
782 			senderr(error);
783 		if (new_rtm != NULL) {
784 			old_rtm = rtm;
785 			rtm = new_rtm;
786 		}
787 		break;
788 
789 	case RTM_GET:
790 	case RTM_CHANGE:
791 	case RTM_LOCK:
792                 /* XXX This will mask info.rti_info[RTAX_DST] with
793 		 * info.rti_info[RTAX_NETMASK] before
794                  * searching.  It did not used to do that.  --dyoung
795 		 */
796 		rt = NULL;
797 		error = rtrequest1(RTM_GET, &info, &rt);
798 		if (error != 0)
799 			senderr(error);
800 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
801 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
802 			    info.rti_info[RTAX_DST]->sa_len) != 0)
803 				senderr(ESRCH);
804 			if (info.rti_info[RTAX_NETMASK] == NULL &&
805 			    rt_mask(rt) != NULL)
806 				senderr(ETOOMANYREFS);
807 		}
808 
809 		/*
810 		 * XXX if arp/ndp requests an L2 entry, we have to obtain
811 		 * it from lltable while for the route command we have to
812 		 * return a route as it is. How to distinguish them?
813 		 * For newer arp/ndp, RTF_LLDATA flag set by arp/ndp
814 		 * indicates an L2 entry is requested. For old arp/ndp
815 		 * binaries, we check RTF_UP flag is NOT set; it works
816 		 * by the fact that arp/ndp don't set it while the route
817 		 * command sets it.
818 		 */
819 		if (((rtm->rtm_flags & RTF_LLDATA) != 0 ||
820 		     (rtm->rtm_flags & RTF_UP) == 0) &&
821 		    rtm->rtm_type == RTM_GET &&
822 		    sockaddr_cmp(rt_getkey(rt), info.rti_info[RTAX_DST]) != 0) {
823 			int ll_flags = 0;
824 			route_get_sdl(rt->rt_ifp, info.rti_info[RTAX_DST], &sdl,
825 			    &ll_flags);
826 			info.rti_info[RTAX_GATEWAY] = sstocsa(&sdl);
827 			error = route_output_report(rt, &info, rtm, &new_rtm);
828 			if (error)
829 				senderr(error);
830 			if (new_rtm != NULL) {
831 				old_rtm = rtm;
832 				rtm = new_rtm;
833 			}
834 			rtm->rtm_flags |= RTF_LLDATA;
835 			rtm->rtm_flags &= ~RTF_CONNECTED;
836 			rtm->rtm_flags |= (ll_flags & LLE_STATIC) ? RTF_STATIC : 0;
837 			break;
838 		}
839 
840 		switch (rtm->rtm_type) {
841 		case RTM_GET:
842 			info.rti_info[RTAX_DST] = rt_getkey(rt);
843 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
844 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
845 			info.rti_info[RTAX_TAG] = rt_gettag(rt);
846 			error = route_output_report(rt, &info, rtm, &new_rtm);
847 			if (error)
848 				senderr(error);
849 			if (new_rtm != NULL) {
850 				old_rtm = rtm;
851 				rtm = new_rtm;
852 			}
853 			break;
854 
855 		case RTM_CHANGE:
856 #ifdef NET_MPSAFE
857 			/*
858 			 * Release rt_so_mtx to avoid a deadlock with route_intr
859 			 * and also serialize updating routes to avoid another.
860 			 */
861 			if (rt_updating) {
862 				/* Release to allow the updater to proceed */
863 				rt_unref(rt);
864 				rt = NULL;
865 			}
866 			while (rt_updating) {
867 				error = cv_wait_sig(&rt_update_cv, rt_so_mtx);
868 				if (error != 0)
869 					goto flush;
870 			}
871 			if (rt == NULL) {
872 				error = rtrequest1(RTM_GET, &info, &rt);
873 				if (error != 0)
874 					goto flush;
875 			}
876 			rt_updating = true;
877 			mutex_exit(rt_so_mtx);
878 
879 			error = rt_update_prepare(rt);
880 			if (error == 0) {
881 				error = rt_update(rt, &info, rtm);
882 				rt_update_finish(rt);
883 			}
884 
885 			mutex_enter(rt_so_mtx);
886 			rt_updating = false;
887 			cv_broadcast(&rt_update_cv);
888 #else
889 			error = rt_update(rt, &info, rtm);
890 #endif
891 			if (error != 0)
892 				goto flush;
893 			/*FALLTHROUGH*/
894 		case RTM_LOCK:
895 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
896 			rt->rt_rmx.rmx_locks |=
897 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
898 			break;
899 		}
900 		break;
901 
902 	default:
903 		senderr(EOPNOTSUPP);
904 	}
905 
906 flush:
907 	if (rtm) {
908 		if (error)
909 			rtm->rtm_errno = error;
910 		else
911 			rtm->rtm_flags |= RTF_DONE;
912 	}
913 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
914 	    0;
915 	/* We cannot free old_rtm until we have stopped using the
916 	 * pointers in info, some of which may point to sockaddrs
917 	 * in old_rtm.
918 	 */
919 	if (old_rtm != NULL)
920 		Free(old_rtm);
921 	if (rt) {
922 		if (do_rt_free) {
923 #ifdef NET_MPSAFE
924 			/*
925 			 * Release rt_so_mtx to avoid a deadlock with
926 			 * route_intr.
927 			 */
928 			mutex_exit(rt_so_mtx);
929 			rt_free(rt);
930 			mutex_enter(rt_so_mtx);
931 #else
932 			rt_free(rt);
933 #endif
934 		} else
935 			rt_unref(rt);
936 	}
937     {
938 	struct rawcb *rp = NULL;
939 	/*
940 	 * Check to see if we don't want our own messages.
941 	 */
942 	if ((so->so_options & SO_USELOOPBACK) == 0) {
943 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
944 			if (rtm)
945 				Free(rtm);
946 			m_freem(m);
947 			goto out;
948 		}
949 		/* There is another listener, so construct message */
950 		rp = sotorawcb(so);
951 	}
952 	if (rtm) {
953 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
954 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
955 			m_freem(m);
956 			m = NULL;
957 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
958 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
959 		Free(rtm);
960 	}
961 	if (rp)
962 		rp->rcb_proto.sp_family = 0; /* Avoid us */
963 	if (family)
964 		proto.sp_protocol = family;
965 	if (m)
966 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
967 		    &COMPATNAME(route_info).ri_dst, &rt_rawcb);
968 	if (rp)
969 		rp->rcb_proto.sp_family = PF_XROUTE;
970     }
971 out:
972 	curlwp_bindx(bound);
973 	return error;
974 }
975 
976 static int
977 route_ctloutput(int op, struct socket *so, struct sockopt *sopt)
978 {
979 	struct routecb *rop = sotoroutecb(so);
980 	int error = 0;
981 	unsigned char *rtm_type;
982 	size_t len;
983 	unsigned int msgfilter;
984 
985 	KASSERT(solocked(so));
986 
987 	if (sopt->sopt_level != AF_ROUTE) {
988 		error = ENOPROTOOPT;
989 	} else switch (op) {
990 	case PRCO_SETOPT:
991 		switch (sopt->sopt_name) {
992 		case RO_MSGFILTER:
993 			msgfilter = 0;
994 			for (rtm_type = sopt->sopt_data, len = sopt->sopt_size;
995 			     len != 0;
996 			     rtm_type++, len -= sizeof(*rtm_type))
997 			{
998 				/* Guard against overflowing our storage. */
999 				if (*rtm_type >= sizeof(msgfilter) * CHAR_BIT) {
1000 					error = EOVERFLOW;
1001 					break;
1002 				}
1003 				msgfilter |= RTMSGFILTER(*rtm_type);
1004 			}
1005 			if (error == 0)
1006 				rop->rocb_msgfilter = msgfilter;
1007 			break;
1008 		default:
1009 			error = ENOPROTOOPT;
1010 			break;
1011 		}
1012 		break;
1013 	case PRCO_GETOPT:
1014 		switch (sopt->sopt_name) {
1015 		case RO_MSGFILTER:
1016 			error = ENOTSUP;
1017 			break;
1018 		default:
1019 			error = ENOPROTOOPT;
1020 			break;
1021 		}
1022 	}
1023 	return error;
1024 }
1025 
1026 static void
1027 _rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
1028 {
1029 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
1030 	metric(RTV_RPIPE, rmx_recvpipe);
1031 	metric(RTV_SPIPE, rmx_sendpipe);
1032 	metric(RTV_SSTHRESH, rmx_ssthresh);
1033 	metric(RTV_RTT, rmx_rtt);
1034 	metric(RTV_RTTVAR, rmx_rttvar);
1035 	metric(RTV_HOPCOUNT, rmx_hopcount);
1036 	metric(RTV_MTU, rmx_mtu);
1037 #undef metric
1038 	if (which & RTV_EXPIRE) {
1039 		out->rt_rmx.rmx_expire = in->rtm_rmx.rmx_expire ?
1040 		    time_wall_to_mono(in->rtm_rmx.rmx_expire) : 0;
1041 	}
1042 }
1043 
1044 static void
1045 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
1046 {
1047 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
1048 	metric(rmx_recvpipe);
1049 	metric(rmx_sendpipe);
1050 	metric(rmx_ssthresh);
1051 	metric(rmx_rtt);
1052 	metric(rmx_rttvar);
1053 	metric(rmx_hopcount);
1054 	metric(rmx_mtu);
1055 	metric(rmx_locks);
1056 #undef metric
1057 	out->rtm_rmx.rmx_expire = in->rt_rmx.rmx_expire ?
1058 	    time_mono_to_wall(in->rt_rmx.rmx_expire) : 0;
1059 }
1060 
1061 static int
1062 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
1063     struct rt_addrinfo *rtinfo)
1064 {
1065 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
1066 	int i;
1067 
1068 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1069 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1070 			continue;
1071 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
1072 		RT_XADVANCE(cp, sa);
1073 	}
1074 
1075 	/*
1076 	 * Check for extra addresses specified, except RTM_GET asking
1077 	 * for interface info.
1078 	 */
1079 	if (rtmtype == RTM_GET) {
1080 		if (((rtinfo->rti_addrs &
1081 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0U << i)) != 0)
1082 			return 1;
1083 	} else if ((rtinfo->rti_addrs & (~0U << i)) != 0)
1084 		return 1;
1085 	/* Check for bad data length.  */
1086 	if (cp != cplim) {
1087 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
1088 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
1089 			/*
1090 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
1091 			 * We accept this for now for the sake of old
1092 			 * binaries or third party softwares.
1093 			 */
1094 			;
1095 		else
1096 			return 1;
1097 	}
1098 	return 0;
1099 }
1100 
1101 static int
1102 rt_getlen(int type)
1103 {
1104 	RTS_CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
1105 	RTS_CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
1106 	RTS_CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
1107 	RTS_CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
1108 
1109 	switch (type) {
1110 	case RTM_ODELADDR:
1111 	case RTM_ONEWADDR:
1112 	case RTM_OCHGADDR:
1113 		if (rtsock_iflist_70_hook.hooked)
1114 			return sizeof(struct ifa_msghdr70);
1115 		else {
1116 #ifdef RTSOCK_DEBUG
1117 			printf("%s: unsupported RTM type %d\n", __func__, type);
1118 #endif
1119 			return -1;
1120 		}
1121 
1122 	case RTM_DELADDR:
1123 	case RTM_NEWADDR:
1124 	case RTM_CHGADDR:
1125 		return sizeof(struct ifa_xmsghdr);
1126 
1127 	case RTM_OOIFINFO:
1128 		if (rtsock_iflist_14_hook.hooked)
1129 			return sizeof(struct if_msghdr14);
1130 		else {
1131 #ifdef RTSOCK_DEBUG
1132 			printf("%s: unsupported RTM type RTM_OOIFINFO\n",
1133 			    __func__);
1134 #endif
1135 			return -1;
1136 		}
1137 
1138 	case RTM_OIFINFO:
1139 		if (rtsock_iflist_50_hook.hooked)
1140 			return sizeof(struct if_msghdr50);
1141 		else {
1142 #ifdef RTSOCK_DEBUG
1143 			printf("%s: unsupported RTM type RTM_OIFINFO\n",
1144 			    __func__);
1145 #endif
1146 			return -1;
1147 		}
1148 
1149 	case RTM_IFINFO:
1150 		return sizeof(struct if_xmsghdr);
1151 
1152 	case RTM_IFANNOUNCE:
1153 	case RTM_IEEE80211:
1154 		return sizeof(struct if_xannouncemsghdr);
1155 
1156 	default:
1157 		return sizeof(struct rt_xmsghdr);
1158 	}
1159 }
1160 
1161 
1162 struct mbuf *
1163 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
1164 {
1165 	struct rt_xmsghdr *rtm;
1166 	struct mbuf *m;
1167 	int i;
1168 	const struct sockaddr *sa;
1169 	int len, dlen;
1170 
1171 	m = m_gethdr(M_DONTWAIT, MT_DATA);
1172 	if (m == NULL)
1173 		return m;
1174 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
1175 
1176 	if ((len = rt_getlen(type)) == -1)
1177 		goto out;
1178 	if (len > MHLEN + MLEN)
1179 		panic("%s: message too long", __func__);
1180 	else if (len > MHLEN) {
1181 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
1182 		if (m->m_next == NULL)
1183 			goto out;
1184 		MCLAIM(m->m_next, m->m_owner);
1185 		m->m_pkthdr.len = len;
1186 		m->m_len = MHLEN;
1187 		m->m_next->m_len = len - MHLEN;
1188 	} else {
1189 		m->m_pkthdr.len = m->m_len = len;
1190 	}
1191 	m_reset_rcvif(m);
1192 	m_copyback(m, 0, datalen, data);
1193 	if (len > datalen)
1194 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
1195 	rtm = mtod(m, struct rt_xmsghdr *);
1196 	for (i = 0; i < RTAX_MAX; i++) {
1197 		if ((sa = rtinfo->rti_info[i]) == NULL)
1198 			continue;
1199 		rtinfo->rti_addrs |= (1 << i);
1200 		dlen = RT_XROUNDUP(sa->sa_len);
1201 		m_copyback(m, len, sa->sa_len, sa);
1202 		if (dlen != sa->sa_len) {
1203 			/*
1204 			 * Up to 7 + 1 nul's since roundup is to
1205 			 * sizeof(uint64_t) (8 bytes)
1206 			 */
1207 			m_copyback(m, len + sa->sa_len,
1208 			    dlen - sa->sa_len, "\0\0\0\0\0\0\0");
1209 		}
1210 		len += dlen;
1211 	}
1212 	if (m->m_pkthdr.len != len)
1213 		goto out;
1214 	rtm->rtm_msglen = len;
1215 	rtm->rtm_version = RTM_XVERSION;
1216 	rtm->rtm_type = type;
1217 	return m;
1218 out:
1219 	m_freem(m);
1220 	return NULL;
1221 }
1222 
1223 /*
1224  * rt_msg2
1225  *
1226  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
1227  *		returns the length of the message in 'lenp'.
1228  *
1229  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
1230  *	the message
1231  * otherwise walkarg's w_needed is updated and if the user buffer is
1232  *	specified and w_needed indicates space exists the information is copied
1233  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
1234  *	if the allocation fails ENOBUFS is returned.
1235  */
1236 static int
1237 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1238 	int *lenp)
1239 {
1240 	int i;
1241 	int len, dlen, second_time = 0;
1242 	char *cp0, *cp = cpv;
1243 
1244 	rtinfo->rti_addrs = 0;
1245 again:
1246 	if ((len = rt_getlen(type)) == -1)
1247 		return EINVAL;
1248 
1249 	if ((cp0 = cp) != NULL)
1250 		cp += len;
1251 	for (i = 0; i < RTAX_MAX; i++) {
1252 		const struct sockaddr *sa;
1253 
1254 		if ((sa = rtinfo->rti_info[i]) == NULL)
1255 			continue;
1256 		rtinfo->rti_addrs |= (1 << i);
1257 		dlen = RT_XROUNDUP(sa->sa_len);
1258 		if (cp) {
1259 			int diff = dlen - sa->sa_len;
1260 			(void)memcpy(cp, sa, (size_t)sa->sa_len);
1261 			cp += sa->sa_len;
1262 			if (diff > 0) {
1263 				(void)memset(cp, 0, (size_t)diff);
1264 				cp += diff;
1265 			}
1266 		}
1267 		len += dlen;
1268 	}
1269 	if (cp == NULL && w != NULL && !second_time) {
1270 		struct rt_walkarg *rw = w;
1271 
1272 		rw->w_needed += len;
1273 		if (rw->w_needed <= 0 && rw->w_where) {
1274 			if (rw->w_tmemsize < len) {
1275 				if (rw->w_tmem)
1276 					kmem_free(rw->w_tmem, rw->w_tmemsize);
1277 				rw->w_tmem = kmem_zalloc(len, KM_SLEEP);
1278 				rw->w_tmemsize = len;
1279 			}
1280 			if (rw->w_tmem) {
1281 				cp = rw->w_tmem;
1282 				second_time = 1;
1283 				goto again;
1284 			} else {
1285 				rw->w_tmemneeded = len;
1286 				return ENOBUFS;
1287 			}
1288 		}
1289 	}
1290 	if (cp) {
1291 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
1292 
1293 		rtm->rtm_version = RTM_XVERSION;
1294 		rtm->rtm_type = type;
1295 		rtm->rtm_msglen = len;
1296 	}
1297 	if (lenp)
1298 		*lenp = len;
1299 	return 0;
1300 }
1301 
1302 /*
1303  * This routine is called to generate a message from the routing
1304  * socket indicating that a redirect has occurred, a routing lookup
1305  * has failed, or that a protocol has detected timeouts to a particular
1306  * destination.
1307  */
1308 void
1309 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
1310     int error)
1311 {
1312 	struct rt_xmsghdr rtm;
1313 	struct mbuf *m;
1314 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1315 	struct rt_addrinfo info = *rtinfo;
1316 
1317 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
1318 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1319 		return;
1320 	memset(&rtm, 0, sizeof(rtm));
1321 	rtm.rtm_pid = curproc->p_pid;
1322 	rtm.rtm_flags = RTF_DONE | flags;
1323 	rtm.rtm_errno = error;
1324 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
1325 	if (m == NULL)
1326 		return;
1327 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1328 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1329 }
1330 
1331 /*
1332  * This routine is called to generate a message from the routing
1333  * socket indicating that the status of a network interface has changed.
1334  */
1335 void
1336 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
1337 {
1338 	struct if_xmsghdr ifm;
1339 	struct mbuf *m;
1340 	struct rt_addrinfo info;
1341 
1342 	COMPATCALL(rt_ifmsg, (ifp));
1343 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1344 		return;
1345 	(void)memset(&info, 0, sizeof(info));
1346 	(void)memset(&ifm, 0, sizeof(ifm));
1347 	ifm.ifm_index = ifp->if_index;
1348 	ifm.ifm_flags = ifp->if_flags;
1349 	ifm.ifm_data = ifp->if_data;
1350 	ifm.ifm_addrs = 0;
1351 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
1352 	if (m == NULL)
1353 		return;
1354 	COMPATNAME(route_enqueue)(m, 0);
1355 	MODULE_HOOK_CALL_VOID(rtsock_oifmsg_14_hook, (ifp), __nothing);
1356 	MODULE_HOOK_CALL_VOID(rtsock_oifmsg_50_hook, (ifp), __nothing);
1357 }
1358 
1359 /*
1360  * This is called to generate messages from the routing socket
1361  * indicating a network interface has had addresses associated with it.
1362  * if we ever reverse the logic and replace messages TO the routing
1363  * socket indicate a request to configure interfaces, then it will
1364  * be unnecessary as the routing socket will automatically generate
1365  * copies of it.
1366  */
1367 static void
1368 COMPATNAME(rt_addrmsg0)(int cmd, struct ifaddr *ifa, int error,
1369     struct rtentry *rt, const struct sockaddr *src)
1370 {
1371 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
1372 	struct rt_addrinfo info;
1373 	const struct sockaddr *sa;
1374 	int pass;
1375 	struct mbuf *m;
1376 	struct ifnet *ifp;
1377 	struct rt_xmsghdr rtm;
1378 	struct ifa_xmsghdr ifam;
1379 	int ncmd;
1380 
1381 	KASSERT(ifa != NULL);
1382 	KASSERT(ifa->ifa_addr != NULL);
1383 	ifp = ifa->ifa_ifp;
1384 	if (cmd == RTM_ADD && vec_sctp_add_ip_address != NULL) {
1385 		(*vec_sctp_add_ip_address)(ifa);
1386 	} else if (cmd == RTM_DELETE && vec_sctp_delete_ip_address != NULL) {
1387 		(*vec_sctp_delete_ip_address)(ifa);
1388 	}
1389 
1390 	COMPATCALL(rt_addrmsg_rt, (cmd, ifa, error, rt));
1391 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1392 		return;
1393 	for (pass = 1; pass < 3; pass++) {
1394 		memset(&info, 0, sizeof(info));
1395 		switch (cmdpass(cmd, pass)) {
1396 		case cmdpass(RTM_ADD, 1):
1397 		case cmdpass(RTM_CHANGE, 1):
1398 		case cmdpass(RTM_DELETE, 2):
1399 		case cmdpass(RTM_NEWADDR, 1):
1400 		case cmdpass(RTM_DELADDR, 1):
1401 		case cmdpass(RTM_CHGADDR, 1):
1402 			switch (cmd) {
1403 			case RTM_ADD:
1404 				ncmd = RTM_XNEWADDR;
1405 				break;
1406 			case RTM_DELETE:
1407 				ncmd = RTM_XDELADDR;
1408 				break;
1409 			case RTM_CHANGE:
1410 				ncmd = RTM_XCHGADDR;
1411 				break;
1412 			case RTM_NEWADDR:
1413 				ncmd = RTM_XNEWADDR;
1414 				break;
1415 			case RTM_DELADDR:
1416 				ncmd = RTM_XDELADDR;
1417 				break;
1418 			case RTM_CHGADDR:
1419 				ncmd = RTM_XCHGADDR;
1420 				break;
1421 			default:
1422 				panic("%s: unknown command %d", __func__, cmd);
1423 			}
1424 			MODULE_HOOK_CALL_VOID(rtsock_newaddr_70_hook,
1425 			    (ncmd, ifa), __nothing);
1426 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1427 			KASSERT(ifp->if_dl != NULL);
1428 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1429 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1430 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1431 			info.rti_info[RTAX_AUTHOR] = src;
1432 			memset(&ifam, 0, sizeof(ifam));
1433 			ifam.ifam_index = ifp->if_index;
1434 			ifam.ifam_metric = ifa->ifa_metric;
1435 			ifam.ifam_flags = ifa->ifa_flags;
1436 #ifndef COMPAT_RTSOCK
1437 			ifam.ifam_pid = curproc->p_pid;
1438 			ifam.ifam_addrflags = if_addrflags(ifa);
1439 #endif
1440 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
1441 			if (m == NULL)
1442 				continue;
1443 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
1444 			    info.rti_addrs;
1445 			break;
1446 		case cmdpass(RTM_ADD, 2):
1447 		case cmdpass(RTM_CHANGE, 2):
1448 		case cmdpass(RTM_DELETE, 1):
1449 			if (rt == NULL)
1450 				continue;
1451 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1452 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
1453 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1454 			memset(&rtm, 0, sizeof(rtm));
1455 			rtm.rtm_pid = curproc->p_pid;
1456 			rtm.rtm_index = ifp->if_index;
1457 			rtm.rtm_flags |= rt->rt_flags;
1458 			rtm.rtm_errno = error;
1459 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
1460 			if (m == NULL)
1461 				continue;
1462 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1463 			break;
1464 		default:
1465 			continue;
1466 		}
1467 		KASSERTMSG(m != NULL, "called with wrong command");
1468 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1469 	}
1470 #undef cmdpass
1471 }
1472 
1473 void
1474 COMPATNAME(rt_addrmsg)(int cmd, struct ifaddr *ifa)
1475 {
1476 
1477 	COMPATNAME(rt_addrmsg0)(cmd, ifa, 0, NULL, NULL);
1478 }
1479 
1480 void
1481 COMPATNAME(rt_addrmsg_rt)(int cmd, struct ifaddr *ifa, int error,
1482     struct rtentry *rt)
1483 {
1484 
1485 	COMPATNAME(rt_addrmsg0)(cmd, ifa, error, rt, NULL);
1486 }
1487 
1488 void
1489 COMPATNAME(rt_addrmsg_src)(int cmd, struct ifaddr *ifa,
1490     const struct sockaddr *src)
1491 {
1492 
1493 	COMPATNAME(rt_addrmsg0)(cmd, ifa, 0, NULL, src);
1494 }
1495 
1496 static struct mbuf *
1497 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1498     struct rt_addrinfo *info)
1499 {
1500 	struct if_xannouncemsghdr ifan;
1501 
1502 	memset(info, 0, sizeof(*info));
1503 	memset(&ifan, 0, sizeof(ifan));
1504 	ifan.ifan_index = ifp->if_index;
1505 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
1506 	ifan.ifan_what = what;
1507 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
1508 }
1509 
1510 /*
1511  * This is called to generate routing socket messages indicating
1512  * network interface arrival and departure.
1513  */
1514 void
1515 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
1516 {
1517 	struct mbuf *m;
1518 	struct rt_addrinfo info;
1519 
1520 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
1521 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1522 		return;
1523 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1524 	if (m == NULL)
1525 		return;
1526 	COMPATNAME(route_enqueue)(m, 0);
1527 }
1528 
1529 /*
1530  * This is called to generate routing socket messages indicating
1531  * IEEE80211 wireless events.
1532  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1533  */
1534 void
1535 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
1536 	size_t data_len)
1537 {
1538 	struct mbuf *m;
1539 	struct rt_addrinfo info;
1540 
1541 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
1542 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
1543 		return;
1544 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1545 	if (m == NULL)
1546 		return;
1547 	/*
1548 	 * Append the ieee80211 data.  Try to stick it in the
1549 	 * mbuf containing the ifannounce msg; otherwise allocate
1550 	 * a new mbuf and append.
1551 	 *
1552 	 * NB: we assume m is a single mbuf.
1553 	 */
1554 	if (data_len > M_TRAILINGSPACE(m)) {
1555 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1556 		if (n == NULL) {
1557 			m_freem(m);
1558 			return;
1559 		}
1560 		(void)memcpy(mtod(n, void *), data, data_len);
1561 		n->m_len = data_len;
1562 		m->m_next = n;
1563 	} else if (data_len > 0) {
1564 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
1565 		m->m_len += data_len;
1566 	}
1567 	if (m->m_flags & M_PKTHDR)
1568 		m->m_pkthdr.len += data_len;
1569 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
1570 	COMPATNAME(route_enqueue)(m, 0);
1571 }
1572 
1573 /*
1574  * Routing message software interrupt routine
1575  */
1576 static void
1577 COMPATNAME(route_intr)(void *cookie)
1578 {
1579 	struct sockproto proto = { .sp_family = PF_XROUTE, };
1580 	struct route_info * const ri = &COMPATNAME(route_info);
1581 	struct mbuf *m;
1582 
1583 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
1584 	for (;;) {
1585 		IFQ_LOCK(&ri->ri_intrq);
1586 		IF_DEQUEUE(&ri->ri_intrq, m);
1587 		IFQ_UNLOCK(&ri->ri_intrq);
1588 		if (m == NULL)
1589 			break;
1590 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
1591 #ifdef NET_MPSAFE
1592 		mutex_enter(rt_so_mtx);
1593 #endif
1594 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst, &rt_rawcb);
1595 #ifdef NET_MPSAFE
1596 		mutex_exit(rt_so_mtx);
1597 #endif
1598 	}
1599 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1600 }
1601 
1602 /*
1603  * Enqueue a message to the software interrupt routine.
1604  */
1605 void
1606 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
1607 {
1608 	struct route_info * const ri = &COMPATNAME(route_info);
1609 	int wasempty;
1610 
1611 	IFQ_LOCK(&ri->ri_intrq);
1612 	if (IF_QFULL(&ri->ri_intrq)) {
1613 		printf("%s: queue full, dropped message\n", __func__);
1614 		IF_DROP(&ri->ri_intrq);
1615 		IFQ_UNLOCK(&ri->ri_intrq);
1616 		m_freem(m);
1617 	} else {
1618 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
1619 		M_SETCTX(m, (uintptr_t)family);
1620 		IF_ENQUEUE(&ri->ri_intrq, m);
1621 		IFQ_UNLOCK(&ri->ri_intrq);
1622 		if (wasempty) {
1623 			kpreempt_disable();
1624 			softint_schedule(ri->ri_sih);
1625 			kpreempt_enable();
1626 		}
1627 	}
1628 }
1629 
1630 static void
1631 COMPATNAME(route_init)(void)
1632 {
1633 	struct route_info * const ri = &COMPATNAME(route_info);
1634 
1635 #ifndef COMPAT_RTSOCK
1636 	rt_init();
1637 #ifdef NET_MPSAFE
1638 	rt_so_mtx = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
1639 
1640 	cv_init(&rt_update_cv, "rtsock_cv");
1641 #endif
1642 
1643 	sysctl_net_route_setup(NULL, PF_ROUTE, "rtable");
1644 #endif
1645 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
1646 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
1647 	    COMPATNAME(route_intr), NULL);
1648 	IFQ_LOCK_INIT(&ri->ri_intrq);
1649 }
1650 
1651 /*
1652  * Definitions of protocols supported in the ROUTE domain.
1653  */
1654 #ifndef COMPAT_RTSOCK
1655 PR_WRAP_USRREQS(route);
1656 #else
1657 PR_WRAP_USRREQS(compat_50_route);
1658 #endif
1659 
1660 static const struct pr_usrreqs route_usrreqs = {
1661 	.pr_attach	= COMPATNAME(route_attach_wrapper),
1662 	.pr_detach	= COMPATNAME(route_detach_wrapper),
1663 	.pr_accept	= COMPATNAME(route_accept_wrapper),
1664 	.pr_bind	= COMPATNAME(route_bind_wrapper),
1665 	.pr_listen	= COMPATNAME(route_listen_wrapper),
1666 	.pr_connect	= COMPATNAME(route_connect_wrapper),
1667 	.pr_connect2	= COMPATNAME(route_connect2_wrapper),
1668 	.pr_disconnect	= COMPATNAME(route_disconnect_wrapper),
1669 	.pr_shutdown	= COMPATNAME(route_shutdown_wrapper),
1670 	.pr_abort	= COMPATNAME(route_abort_wrapper),
1671 	.pr_ioctl	= COMPATNAME(route_ioctl_wrapper),
1672 	.pr_stat	= COMPATNAME(route_stat_wrapper),
1673 	.pr_peeraddr	= COMPATNAME(route_peeraddr_wrapper),
1674 	.pr_sockaddr	= COMPATNAME(route_sockaddr_wrapper),
1675 	.pr_rcvd	= COMPATNAME(route_rcvd_wrapper),
1676 	.pr_recvoob	= COMPATNAME(route_recvoob_wrapper),
1677 	.pr_send	= COMPATNAME(route_send_wrapper),
1678 	.pr_sendoob	= COMPATNAME(route_sendoob_wrapper),
1679 	.pr_purgeif	= COMPATNAME(route_purgeif_wrapper),
1680 };
1681 
1682 static const struct protosw COMPATNAME(route_protosw)[] = {
1683 	{
1684 		.pr_type = SOCK_RAW,
1685 		.pr_domain = &COMPATNAME(routedomain),
1686 		.pr_flags = PR_ATOMIC|PR_ADDR,
1687 		.pr_ctlinput = raw_ctlinput,
1688 		.pr_ctloutput = route_ctloutput,
1689 		.pr_usrreqs = &route_usrreqs,
1690 		.pr_init = rt_pr_init,
1691 	},
1692 };
1693 
1694 struct domain COMPATNAME(routedomain) = {
1695 	.dom_family = PF_XROUTE,
1696 	.dom_name = DOMAINNAME,
1697 	.dom_init = COMPATNAME(route_init),
1698 	.dom_protosw = COMPATNAME(route_protosw),
1699 	.dom_protoswNPROTOSW =
1700 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
1701 };
1702