if_stf.c revision 128417
1/*	$FreeBSD: head/sys/net/if_stf.c 128417 2004-04-19 05:06:27Z brooks $	*/
2/*	$KAME: if_stf.c,v 1.73 2001/12/03 11:08:30 keiichi Exp $	*/
3
4/*
5 * Copyright (C) 2000 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33/*
34 * 6to4 interface, based on RFC3056.
35 *
36 * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting.
37 * There is no address mapping defined from IPv6 multicast address to IPv4
38 * address.  Therefore, we do not have IFF_MULTICAST on the interface.
39 *
40 * Due to the lack of address mapping for link-local addresses, we cannot
41 * throw packets toward link-local addresses (fe80::x).  Also, we cannot throw
42 * packets to link-local multicast addresses (ff02::x).
43 *
44 * Here are interesting symptoms due to the lack of link-local address:
45 *
46 * Unicast routing exchange:
47 * - RIPng: Impossible.  Uses link-local multicast packet toward ff02::9,
48 *   and link-local addresses as nexthop.
49 * - OSPFv6: Impossible.  OSPFv6 assumes that there's link-local address
50 *   assigned to the link, and makes use of them.  Also, HELLO packets use
51 *   link-local multicast addresses (ff02::5 and ff02::6).
52 * - BGP4+: Maybe.  You can only use global address as nexthop, and global
53 *   address as TCP endpoint address.
54 *
55 * Multicast routing protocols:
56 * - PIM: Hello packet cannot be used to discover adjacent PIM routers.
57 *   Adjacent PIM routers must be configured manually (is it really spec-wise
58 *   correct thing to do?).
59 *
60 * ICMPv6:
61 * - Redirects cannot be used due to the lack of link-local address.
62 *
63 * stf interface does not have, and will not need, a link-local address.
64 * It seems to have no real benefit and does not help the above symptoms much.
65 * Even if we assign link-locals to interface, we cannot really
66 * use link-local unicast/multicast on top of 6to4 cloud (since there's no
67 * encapsulation defined for link-local address), and the above analysis does
68 * not change.  RFC3056 does not mandate the assignment of link-local address
69 * either.
70 *
71 * 6to4 interface has security issues.  Refer to
72 * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt
73 * for details.  The code tries to filter out some of malicious packets.
74 * Note that there is no way to be 100% secure.
75 */
76
77#include "opt_inet.h"
78#include "opt_inet6.h"
79#include "opt_mac.h"
80
81#include <sys/param.h>
82#include <sys/systm.h>
83#include <sys/socket.h>
84#include <sys/sockio.h>
85#include <sys/mac.h>
86#include <sys/mbuf.h>
87#include <sys/errno.h>
88#include <sys/kernel.h>
89#include <sys/protosw.h>
90#include <sys/queue.h>
91#include <machine/cpu.h>
92
93#include <sys/malloc.h>
94
95#include <net/if.h>
96#include <net/route.h>
97#include <net/netisr.h>
98#include <net/if_types.h>
99#include <net/if_stf.h>
100
101#include <netinet/in.h>
102#include <netinet/in_systm.h>
103#include <netinet/ip.h>
104#include <netinet/ip_var.h>
105#include <netinet/in_var.h>
106
107#include <netinet/ip6.h>
108#include <netinet6/ip6_var.h>
109#include <netinet6/in6_var.h>
110#include <netinet/ip_ecn.h>
111
112#include <netinet/ip_encap.h>
113
114#include <machine/stdarg.h>
115
116#include <net/net_osdep.h>
117
118#include <net/bpf.h>
119
120#define STFNAME		"stf"
121
122#define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
123
124/*
125 * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
126 * struct in_addr *; use bcopy() instead.
127 */
128#define GET_V4(x)	((caddr_t)(&(x)->s6_addr16[1]))
129
130struct stf_softc {
131	struct ifnet	sc_if;	   /* common area */
132	union {
133		struct route  __sc_ro4;
134		struct route_in6 __sc_ro6; /* just for safety */
135	} __sc_ro46;
136#define sc_ro	__sc_ro46.__sc_ro4
137	const struct encaptab *encap_cookie;
138	LIST_ENTRY(stf_softc) sc_list;	/* all stf's are linked */
139};
140
141/*
142 * All mutable global variables in if_stf.c are protected by stf_mtx.
143 * XXXRW: Note that mutable fields in the softc are not currently locked:
144 * in particular, sc_ro needs to be protected from concurrent entrance
145 * of stf_output().
146 */
147static struct mtx stf_mtx;
148static LIST_HEAD(, stf_softc) stf_softc_list;
149
150static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
151static const int ip_stf_ttl = 40;
152
153extern  struct domain inetdomain;
154struct protosw in_stf_protosw =
155{ SOCK_RAW,	&inetdomain,	IPPROTO_IPV6,	PR_ATOMIC|PR_ADDR,
156  in_stf_input,	(pr_output_t*)rip_output, 0,	rip_ctloutput,
157  0,
158  0,            0,              0,              0,
159  &rip_usrreqs
160};
161
162static int stfmodevent(module_t, int, void *);
163static int stf_encapcheck(const struct mbuf *, int, int, void *);
164static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
165static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
166	struct rtentry *);
167static int isrfc1918addr(struct in_addr *);
168static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
169	struct ifnet *);
170static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
171	struct ifnet *);
172static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
173static int stf_ioctl(struct ifnet *, u_long, caddr_t);
174
175static int stf_clone_create(struct if_clone *, int);
176static void stf_clone_destroy(struct ifnet *);
177
178/* only one clone is currently allowed */
179struct if_clone stf_cloner =
180    IF_CLONE_INITIALIZER(STFNAME, stf_clone_create, stf_clone_destroy, 0, 0);
181
182static int
183stf_clone_create(ifc, unit)
184	struct if_clone *ifc;
185	int unit;
186{
187	struct stf_softc *sc;
188	struct ifnet *ifp;
189
190	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
191	ifp = &sc->sc_if;
192	if_initname(ifp, ifc->ifc_name, unit);
193
194	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
195	    stf_encapcheck, &in_stf_protosw, sc);
196	if (sc->encap_cookie == NULL) {
197		if_printf(ifp, "attach failed\n");
198		free(sc, M_STF);
199		return (ENOMEM);
200	}
201
202	ifp->if_mtu    = IPV6_MMTU;
203	ifp->if_ioctl  = stf_ioctl;
204	ifp->if_output = stf_output;
205	ifp->if_type   = IFT_STF;
206	ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
207	if_attach(ifp);
208	bpfattach(ifp, DLT_NULL, sizeof(u_int));
209	mtx_lock(&stf_mtx);
210	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
211	mtx_unlock(&stf_mtx);
212	return (0);
213}
214
215static void
216stf_destroy(struct stf_softc *sc)
217{
218	int err;
219
220	err = encap_detach(sc->encap_cookie);
221	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
222	bpfdetach(&sc->sc_if);
223	if_detach(&sc->sc_if);
224
225	free(sc, M_STF);
226}
227
228static void
229stf_clone_destroy(ifp)
230	struct ifnet *ifp;
231{
232	struct stf_softc *sc = (void *) ifp;
233
234	mtx_lock(&stf_mtx);
235	LIST_REMOVE(sc, sc_list);
236	mtx_unlock(&stf_mtx);
237
238	stf_destroy(sc);
239}
240
241static int
242stfmodevent(mod, type, data)
243	module_t mod;
244	int type;
245	void *data;
246{
247	struct stf_softc *sc;
248
249	switch (type) {
250	case MOD_LOAD:
251		mtx_init(&stf_mtx, "stf_mtx", NULL, MTX_DEF);
252		LIST_INIT(&stf_softc_list);
253		if_clone_attach(&stf_cloner);
254
255		break;
256	case MOD_UNLOAD:
257		if_clone_detach(&stf_cloner);
258
259		mtx_lock(&stf_mtx);
260		while ((sc = LIST_FIRST(&stf_softc_list)) != NULL) {
261			LIST_REMOVE(sc, sc_list);
262			mtx_unlock(&stf_mtx);
263			stf_destroy(sc);
264			mtx_lock(&stf_mtx);
265		}
266		mtx_unlock(&stf_mtx);
267		mtx_destroy(&stf_mtx);
268		break;
269	}
270
271	return (0);
272}
273
274static moduledata_t stf_mod = {
275	"if_stf",
276	stfmodevent,
277	0
278};
279
280DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
281
282static int
283stf_encapcheck(m, off, proto, arg)
284	const struct mbuf *m;
285	int off;
286	int proto;
287	void *arg;
288{
289	struct ip ip;
290	struct in6_ifaddr *ia6;
291	struct stf_softc *sc;
292	struct in_addr a, b, mask;
293
294	sc = (struct stf_softc *)arg;
295	if (sc == NULL)
296		return 0;
297
298	if ((sc->sc_if.if_flags & IFF_UP) == 0)
299		return 0;
300
301	/* IFF_LINK0 means "no decapsulation" */
302	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
303		return 0;
304
305	if (proto != IPPROTO_IPV6)
306		return 0;
307
308	/* LINTED const cast */
309	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
310
311	if (ip.ip_v != 4)
312		return 0;
313
314	ia6 = stf_getsrcifa6(&sc->sc_if);
315	if (ia6 == NULL)
316		return 0;
317
318	/*
319	 * check if IPv4 dst matches the IPv4 address derived from the
320	 * local 6to4 address.
321	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
322	 */
323	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
324	    sizeof(ip.ip_dst)) != 0)
325		return 0;
326
327	/*
328	 * check if IPv4 src matches the IPv4 address derived from the
329	 * local 6to4 address masked by prefixmask.
330	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
331	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
332	 */
333	bzero(&a, sizeof(a));
334	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
335	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
336	a.s_addr &= mask.s_addr;
337	b = ip.ip_src;
338	b.s_addr &= mask.s_addr;
339	if (a.s_addr != b.s_addr)
340		return 0;
341
342	/* stf interface makes single side match only */
343	return 32;
344}
345
346static struct in6_ifaddr *
347stf_getsrcifa6(ifp)
348	struct ifnet *ifp;
349{
350	struct ifaddr *ia;
351	struct in_ifaddr *ia4;
352	struct sockaddr_in6 *sin6;
353	struct in_addr in;
354
355	for (ia = TAILQ_FIRST(&ifp->if_addrlist);
356	     ia;
357	     ia = TAILQ_NEXT(ia, ifa_list))
358	{
359		if (ia->ifa_addr == NULL)
360			continue;
361		if (ia->ifa_addr->sa_family != AF_INET6)
362			continue;
363		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
364		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
365			continue;
366
367		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
368		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
369			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
370				break;
371		if (ia4 == NULL)
372			continue;
373
374		return (struct in6_ifaddr *)ia;
375	}
376
377	return NULL;
378}
379
380static int
381stf_output(ifp, m, dst, rt)
382	struct ifnet *ifp;
383	struct mbuf *m;
384	struct sockaddr *dst;
385	struct rtentry *rt;
386{
387	struct stf_softc *sc;
388	struct sockaddr_in6 *dst6;
389	struct in_addr in4;
390	caddr_t ptr;
391	struct sockaddr_in *dst4;
392	u_int8_t tos;
393	struct ip *ip;
394	struct ip6_hdr *ip6;
395	struct in6_ifaddr *ia6;
396#ifdef MAC
397	int error;
398
399	error = mac_check_ifnet_transmit(ifp, m);
400	if (error) {
401		m_freem(m);
402		return (error);
403	}
404#endif
405
406	sc = (struct stf_softc*)ifp;
407	dst6 = (struct sockaddr_in6 *)dst;
408
409	/* just in case */
410	if ((ifp->if_flags & IFF_UP) == 0) {
411		m_freem(m);
412		ifp->if_oerrors++;
413		return ENETDOWN;
414	}
415
416	/*
417	 * If we don't have an ip4 address that match my inner ip6 address,
418	 * we shouldn't generate output.  Without this check, we'll end up
419	 * using wrong IPv4 source.
420	 */
421	ia6 = stf_getsrcifa6(ifp);
422	if (ia6 == NULL) {
423		m_freem(m);
424		ifp->if_oerrors++;
425		return ENETDOWN;
426	}
427
428	if (m->m_len < sizeof(*ip6)) {
429		m = m_pullup(m, sizeof(*ip6));
430		if (!m) {
431			ifp->if_oerrors++;
432			return ENOBUFS;
433		}
434	}
435	ip6 = mtod(m, struct ip6_hdr *);
436	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
437
438	/*
439	 * Pickup the right outer dst addr from the list of candidates.
440	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
441	 */
442	ptr = NULL;
443	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
444		ptr = GET_V4(&ip6->ip6_dst);
445	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
446		ptr = GET_V4(&dst6->sin6_addr);
447	else {
448		m_freem(m);
449		ifp->if_oerrors++;
450		return ENETUNREACH;
451	}
452	bcopy(ptr, &in4, sizeof(in4));
453
454#if NBPFILTER > 0
455	if (ifp->if_bpf) {
456		/*
457		 * We need to prepend the address family as
458		 * a four byte field.  Cons up a dummy header
459		 * to pacify bpf.  This is safe because bpf
460		 * will only read from the mbuf (i.e., it won't
461		 * try to free it or keep a pointer a to it).
462		 */
463		u_int32_t af = AF_INET6;
464#ifdef HAVE_OLD_BPF
465		struct mbuf m0;
466
467		m0.m_next = m;
468		m0.m_len = 4;
469		m0.m_data = (char *)&af;
470
471		BPF_MTAP(ifp, &m0);
472#else
473		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
474#endif
475	}
476#endif /*NBPFILTER > 0*/
477
478	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
479	if (m && m->m_len < sizeof(struct ip))
480		m = m_pullup(m, sizeof(struct ip));
481	if (m == NULL) {
482		ifp->if_oerrors++;
483		return ENOBUFS;
484	}
485	ip = mtod(m, struct ip *);
486
487	bzero(ip, sizeof(*ip));
488
489	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
490	    &ip->ip_src, sizeof(ip->ip_src));
491	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
492	ip->ip_p = IPPROTO_IPV6;
493	ip->ip_ttl = ip_stf_ttl;
494	ip->ip_len = m->m_pkthdr.len;	/*host order*/
495	if (ifp->if_flags & IFF_LINK1)
496		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
497	else
498		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
499
500	/*
501	 * XXXRW: Locking of sc_ro required.
502	 */
503	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
504	if (dst4->sin_family != AF_INET ||
505	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
506		/* cache route doesn't match */
507		dst4->sin_family = AF_INET;
508		dst4->sin_len = sizeof(struct sockaddr_in);
509		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
510		if (sc->sc_ro.ro_rt) {
511			RTFREE(sc->sc_ro.ro_rt);
512			sc->sc_ro.ro_rt = NULL;
513		}
514	}
515
516	if (sc->sc_ro.ro_rt == NULL) {
517		rtalloc(&sc->sc_ro);
518		if (sc->sc_ro.ro_rt == NULL) {
519			m_freem(m);
520			ifp->if_oerrors++;
521			return ENETUNREACH;
522		}
523	}
524
525	ifp->if_opackets++;
526	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
527}
528
529static int
530isrfc1918addr(in)
531	struct in_addr *in;
532{
533	/*
534	 * returns 1 if private address range:
535	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
536	 */
537	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
538	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
539	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
540		return 1;
541
542	return 0;
543}
544
545static int
546stf_checkaddr4(sc, in, inifp)
547	struct stf_softc *sc;
548	struct in_addr *in;
549	struct ifnet *inifp;	/* incoming interface */
550{
551	struct in_ifaddr *ia4;
552
553	/*
554	 * reject packets with the following address:
555	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
556	 */
557	if (IN_MULTICAST(ntohl(in->s_addr)))
558		return -1;
559	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
560	case 0: case 127: case 255:
561		return -1;
562	}
563
564	/*
565	 * reject packets with private address range.
566	 * (requirement from RFC3056 section 2 1st paragraph)
567	 */
568	if (isrfc1918addr(in))
569		return -1;
570
571	/*
572	 * reject packets with broadcast
573	 */
574	for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
575	     ia4;
576	     ia4 = TAILQ_NEXT(ia4, ia_link))
577	{
578		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
579			continue;
580		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
581			return -1;
582	}
583
584	/*
585	 * perform ingress filter
586	 */
587	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
588		struct sockaddr_in sin;
589		struct rtentry *rt;
590
591		bzero(&sin, sizeof(sin));
592		sin.sin_family = AF_INET;
593		sin.sin_len = sizeof(struct sockaddr_in);
594		sin.sin_addr = *in;
595		rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
596		if (!rt || rt->rt_ifp != inifp) {
597#if 0
598			log(LOG_WARNING, "%s: packet from 0x%x dropped "
599			    "due to ingress filter\n", if_name(&sc->sc_if),
600			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
601#endif
602			if (rt)
603				rtfree(rt);
604			return -1;
605		}
606		rtfree(rt);
607	}
608
609	return 0;
610}
611
612static int
613stf_checkaddr6(sc, in6, inifp)
614	struct stf_softc *sc;
615	struct in6_addr *in6;
616	struct ifnet *inifp;	/* incoming interface */
617{
618	/*
619	 * check 6to4 addresses
620	 */
621	if (IN6_IS_ADDR_6TO4(in6)) {
622		struct in_addr in4;
623		bcopy(GET_V4(in6), &in4, sizeof(in4));
624		return stf_checkaddr4(sc, &in4, inifp);
625	}
626
627	/*
628	 * reject anything that look suspicious.  the test is implemented
629	 * in ip6_input too, but we check here as well to
630	 * (1) reject bad packets earlier, and
631	 * (2) to be safe against future ip6_input change.
632	 */
633	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
634		return -1;
635
636	return 0;
637}
638
639void
640in_stf_input(m, off)
641	struct mbuf *m;
642	int off;
643{
644	int proto;
645	struct stf_softc *sc;
646	struct ip *ip;
647	struct ip6_hdr *ip6;
648	u_int8_t otos, itos;
649	struct ifnet *ifp;
650
651	proto = mtod(m, struct ip *)->ip_p;
652
653	if (proto != IPPROTO_IPV6) {
654		m_freem(m);
655		return;
656	}
657
658	ip = mtod(m, struct ip *);
659
660	sc = (struct stf_softc *)encap_getarg(m);
661
662	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
663		m_freem(m);
664		return;
665	}
666
667	ifp = &sc->sc_if;
668
669#ifdef MAC
670	mac_create_mbuf_from_ifnet(ifp, m);
671#endif
672
673	/*
674	 * perform sanity check against outer src/dst.
675	 * for source, perform ingress filter as well.
676	 */
677	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
678	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
679		m_freem(m);
680		return;
681	}
682
683	otos = ip->ip_tos;
684	m_adj(m, off);
685
686	if (m->m_len < sizeof(*ip6)) {
687		m = m_pullup(m, sizeof(*ip6));
688		if (!m)
689			return;
690	}
691	ip6 = mtod(m, struct ip6_hdr *);
692
693	/*
694	 * perform sanity check against inner src/dst.
695	 * for source, perform ingress filter as well.
696	 */
697	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
698	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
699		m_freem(m);
700		return;
701	}
702
703	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
704	if ((ifp->if_flags & IFF_LINK1) != 0)
705		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
706	else
707		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
708	ip6->ip6_flow &= ~htonl(0xff << 20);
709	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
710
711	m->m_pkthdr.rcvif = ifp;
712
713	if (ifp->if_bpf) {
714		/*
715		 * We need to prepend the address family as
716		 * a four byte field.  Cons up a dummy header
717		 * to pacify bpf.  This is safe because bpf
718		 * will only read from the mbuf (i.e., it won't
719		 * try to free it or keep a pointer a to it).
720		 */
721		u_int32_t af = AF_INET6;
722#ifdef HAVE_OLD_BPF
723		struct mbuf m0;
724
725		m0.m_next = m;
726		m0.m_len = 4;
727		m0.m_data = (char *)&af;
728
729		BPF_MTAP(ifp, &m0);
730#else
731		bpf_mtap2(ifp->if_bpf, &af, sizeof(ah), m);
732#endif
733	}
734
735	/*
736	 * Put the packet to the network layer input queue according to the
737	 * specified address family.
738	 * See net/if_gif.c for possible issues with packet processing
739	 * reorder due to extra queueing.
740	 */
741	ifp->if_ipackets++;
742	ifp->if_ibytes += m->m_pkthdr.len;
743	netisr_dispatch(NETISR_IPV6, m);
744}
745
746/* ARGSUSED */
747static void
748stf_rtrequest(cmd, rt, info)
749	int cmd;
750	struct rtentry *rt;
751	struct rt_addrinfo *info;
752{
753	RT_LOCK_ASSERT(rt);
754
755	if (rt)
756		rt->rt_rmx.rmx_mtu = IPV6_MMTU;
757}
758
759static int
760stf_ioctl(ifp, cmd, data)
761	struct ifnet *ifp;
762	u_long cmd;
763	caddr_t data;
764{
765	struct ifaddr *ifa;
766	struct ifreq *ifr;
767	struct sockaddr_in6 *sin6;
768	struct in_addr addr;
769	int error;
770
771	error = 0;
772	switch (cmd) {
773	case SIOCSIFADDR:
774		ifa = (struct ifaddr *)data;
775		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
776			error = EAFNOSUPPORT;
777			break;
778		}
779		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
780		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
781			error = EINVAL;
782			break;
783		}
784		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
785		if (isrfc1918addr(&addr)) {
786			error = EINVAL;
787			break;
788		}
789
790		ifa->ifa_rtrequest = stf_rtrequest;
791		ifp->if_flags |= IFF_UP;
792		break;
793
794	case SIOCADDMULTI:
795	case SIOCDELMULTI:
796		ifr = (struct ifreq *)data;
797		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
798			;
799		else
800			error = EAFNOSUPPORT;
801		break;
802
803	default:
804		error = EINVAL;
805		break;
806	}
807
808	return error;
809}
810