if_stf.c revision 126783
1/*	$FreeBSD: head/sys/net/if_stf.c 126783 2004-03-09 20:29:19Z rwatson $	*/
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
175int	stf_clone_create(struct if_clone *, int);
176void	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
182int
183stf_clone_create(ifc, unit)
184	struct if_clone *ifc;
185	int unit;
186{
187	struct stf_softc *sc;
188
189	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
190	if_initname(&sc->sc_if, ifc->ifc_name, unit);
191
192	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
193	    stf_encapcheck, &in_stf_protosw, sc);
194	if (sc->encap_cookie == NULL) {
195		printf("%s: attach failed\n", if_name(&sc->sc_if));
196		free(sc, M_STF);
197		return (ENOMEM);
198	}
199
200	sc->sc_if.if_mtu    = IPV6_MMTU;
201	sc->sc_if.if_ioctl  = stf_ioctl;
202	sc->sc_if.if_output = stf_output;
203	sc->sc_if.if_type   = IFT_STF;
204	sc->sc_if.if_snd.ifq_maxlen = IFQ_MAXLEN;
205	if_attach(&sc->sc_if);
206	bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int));
207	mtx_lock(&stf_mtx);
208	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
209	mtx_unlock(&stf_mtx);
210	return (0);
211}
212
213static void
214stf_destroy(struct stf_softc *sc)
215{
216	int err;
217
218	err = encap_detach(sc->encap_cookie);
219	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
220	bpfdetach(&sc->sc_if);
221	if_detach(&sc->sc_if);
222
223	free(sc, M_STF);
224}
225
226void
227stf_clone_destroy(ifp)
228	struct ifnet *ifp;
229{
230	struct stf_softc *sc = (void *) ifp;
231
232	mtx_lock(&stf_mtx);
233	LIST_REMOVE(sc, sc_list);
234	mtx_unlock(&stf_mtx);
235
236	stf_destroy(sc);
237}
238
239static int
240stfmodevent(mod, type, data)
241	module_t mod;
242	int type;
243	void *data;
244{
245	struct stf_softc *sc;
246
247	switch (type) {
248	case MOD_LOAD:
249		mtx_init(&stf_mtx, "stf_mtx", NULL, MTX_DEF);
250		LIST_INIT(&stf_softc_list);
251		if_clone_attach(&stf_cloner);
252
253		break;
254	case MOD_UNLOAD:
255		if_clone_detach(&stf_cloner);
256
257		mtx_lock(&stf_mtx);
258		while ((sc = LIST_FIRST(&stf_softc_list)) != NULL) {
259			LIST_REMOVE(sc, sc_list);
260			mtx_unlock(&stf_mtx);
261			stf_destroy(sc);
262			mtx_lock(&stf_mtx);
263		}
264		mtx_unlock(&stf_mtx);
265		mtx_destroy(&stf_mtx);
266		break;
267	}
268
269	return (0);
270}
271
272static moduledata_t stf_mod = {
273	"if_stf",
274	stfmodevent,
275	0
276};
277
278DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
279
280static int
281stf_encapcheck(m, off, proto, arg)
282	const struct mbuf *m;
283	int off;
284	int proto;
285	void *arg;
286{
287	struct ip ip;
288	struct in6_ifaddr *ia6;
289	struct stf_softc *sc;
290	struct in_addr a, b, mask;
291
292	sc = (struct stf_softc *)arg;
293	if (sc == NULL)
294		return 0;
295
296	if ((sc->sc_if.if_flags & IFF_UP) == 0)
297		return 0;
298
299	/* IFF_LINK0 means "no decapsulation" */
300	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
301		return 0;
302
303	if (proto != IPPROTO_IPV6)
304		return 0;
305
306	/* LINTED const cast */
307	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
308
309	if (ip.ip_v != 4)
310		return 0;
311
312	ia6 = stf_getsrcifa6(&sc->sc_if);
313	if (ia6 == NULL)
314		return 0;
315
316	/*
317	 * check if IPv4 dst matches the IPv4 address derived from the
318	 * local 6to4 address.
319	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
320	 */
321	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
322	    sizeof(ip.ip_dst)) != 0)
323		return 0;
324
325	/*
326	 * check if IPv4 src matches the IPv4 address derived from the
327	 * local 6to4 address masked by prefixmask.
328	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
329	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
330	 */
331	bzero(&a, sizeof(a));
332	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
333	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
334	a.s_addr &= mask.s_addr;
335	b = ip.ip_src;
336	b.s_addr &= mask.s_addr;
337	if (a.s_addr != b.s_addr)
338		return 0;
339
340	/* stf interface makes single side match only */
341	return 32;
342}
343
344static struct in6_ifaddr *
345stf_getsrcifa6(ifp)
346	struct ifnet *ifp;
347{
348	struct ifaddr *ia;
349	struct in_ifaddr *ia4;
350	struct sockaddr_in6 *sin6;
351	struct in_addr in;
352
353	for (ia = TAILQ_FIRST(&ifp->if_addrlist);
354	     ia;
355	     ia = TAILQ_NEXT(ia, ifa_list))
356	{
357		if (ia->ifa_addr == NULL)
358			continue;
359		if (ia->ifa_addr->sa_family != AF_INET6)
360			continue;
361		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
362		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
363			continue;
364
365		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
366		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
367			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
368				break;
369		if (ia4 == NULL)
370			continue;
371
372		return (struct in6_ifaddr *)ia;
373	}
374
375	return NULL;
376}
377
378static int
379stf_output(ifp, m, dst, rt)
380	struct ifnet *ifp;
381	struct mbuf *m;
382	struct sockaddr *dst;
383	struct rtentry *rt;
384{
385	struct stf_softc *sc;
386	struct sockaddr_in6 *dst6;
387	struct in_addr in4;
388	caddr_t ptr;
389	struct sockaddr_in *dst4;
390	u_int8_t tos;
391	struct ip *ip;
392	struct ip6_hdr *ip6;
393	struct in6_ifaddr *ia6;
394#ifdef MAC
395	int error;
396
397	error = mac_check_ifnet_transmit(ifp, m);
398	if (error) {
399		m_freem(m);
400		return (error);
401	}
402#endif
403
404	sc = (struct stf_softc*)ifp;
405	dst6 = (struct sockaddr_in6 *)dst;
406
407	/* just in case */
408	if ((ifp->if_flags & IFF_UP) == 0) {
409		m_freem(m);
410		ifp->if_oerrors++;
411		return ENETDOWN;
412	}
413
414	/*
415	 * If we don't have an ip4 address that match my inner ip6 address,
416	 * we shouldn't generate output.  Without this check, we'll end up
417	 * using wrong IPv4 source.
418	 */
419	ia6 = stf_getsrcifa6(ifp);
420	if (ia6 == NULL) {
421		m_freem(m);
422		ifp->if_oerrors++;
423		return ENETDOWN;
424	}
425
426	if (m->m_len < sizeof(*ip6)) {
427		m = m_pullup(m, sizeof(*ip6));
428		if (!m) {
429			ifp->if_oerrors++;
430			return ENOBUFS;
431		}
432	}
433	ip6 = mtod(m, struct ip6_hdr *);
434	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
435
436	/*
437	 * Pickup the right outer dst addr from the list of candidates.
438	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
439	 */
440	ptr = NULL;
441	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
442		ptr = GET_V4(&ip6->ip6_dst);
443	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
444		ptr = GET_V4(&dst6->sin6_addr);
445	else {
446		m_freem(m);
447		ifp->if_oerrors++;
448		return ENETUNREACH;
449	}
450	bcopy(ptr, &in4, sizeof(in4));
451
452#if NBPFILTER > 0
453	if (ifp->if_bpf) {
454		/*
455		 * We need to prepend the address family as
456		 * a four byte field.  Cons up a dummy header
457		 * to pacify bpf.  This is safe because bpf
458		 * will only read from the mbuf (i.e., it won't
459		 * try to free it or keep a pointer a to it).
460		 */
461		u_int32_t af = AF_INET6;
462#ifdef HAVE_OLD_BPF
463		struct mbuf m0;
464
465		m0.m_next = m;
466		m0.m_len = 4;
467		m0.m_data = (char *)&af;
468
469		BPF_MTAP(ifp, &m0);
470#else
471		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
472#endif
473	}
474#endif /*NBPFILTER > 0*/
475
476	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
477	if (m && m->m_len < sizeof(struct ip))
478		m = m_pullup(m, sizeof(struct ip));
479	if (m == NULL) {
480		ifp->if_oerrors++;
481		return ENOBUFS;
482	}
483	ip = mtod(m, struct ip *);
484
485	bzero(ip, sizeof(*ip));
486
487	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
488	    &ip->ip_src, sizeof(ip->ip_src));
489	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
490	ip->ip_p = IPPROTO_IPV6;
491	ip->ip_ttl = ip_stf_ttl;
492	ip->ip_len = m->m_pkthdr.len;	/*host order*/
493	if (ifp->if_flags & IFF_LINK1)
494		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
495	else
496		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
497
498	/*
499	 * XXXRW: Locking of sc_ro required.
500	 */
501	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
502	if (dst4->sin_family != AF_INET ||
503	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
504		/* cache route doesn't match */
505		dst4->sin_family = AF_INET;
506		dst4->sin_len = sizeof(struct sockaddr_in);
507		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
508		if (sc->sc_ro.ro_rt) {
509			RTFREE(sc->sc_ro.ro_rt);
510			sc->sc_ro.ro_rt = NULL;
511		}
512	}
513
514	if (sc->sc_ro.ro_rt == NULL) {
515		rtalloc(&sc->sc_ro);
516		if (sc->sc_ro.ro_rt == NULL) {
517			m_freem(m);
518			ifp->if_oerrors++;
519			return ENETUNREACH;
520		}
521	}
522
523	ifp->if_opackets++;
524	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
525}
526
527static int
528isrfc1918addr(in)
529	struct in_addr *in;
530{
531	/*
532	 * returns 1 if private address range:
533	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
534	 */
535	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
536	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
537	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
538		return 1;
539
540	return 0;
541}
542
543static int
544stf_checkaddr4(sc, in, inifp)
545	struct stf_softc *sc;
546	struct in_addr *in;
547	struct ifnet *inifp;	/* incoming interface */
548{
549	struct in_ifaddr *ia4;
550
551	/*
552	 * reject packets with the following address:
553	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
554	 */
555	if (IN_MULTICAST(ntohl(in->s_addr)))
556		return -1;
557	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
558	case 0: case 127: case 255:
559		return -1;
560	}
561
562	/*
563	 * reject packets with private address range.
564	 * (requirement from RFC3056 section 2 1st paragraph)
565	 */
566	if (isrfc1918addr(in))
567		return -1;
568
569	/*
570	 * reject packets with broadcast
571	 */
572	for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
573	     ia4;
574	     ia4 = TAILQ_NEXT(ia4, ia_link))
575	{
576		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
577			continue;
578		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
579			return -1;
580	}
581
582	/*
583	 * perform ingress filter
584	 */
585	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
586		struct sockaddr_in sin;
587		struct rtentry *rt;
588
589		bzero(&sin, sizeof(sin));
590		sin.sin_family = AF_INET;
591		sin.sin_len = sizeof(struct sockaddr_in);
592		sin.sin_addr = *in;
593		rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
594		if (!rt || rt->rt_ifp != inifp) {
595#if 0
596			log(LOG_WARNING, "%s: packet from 0x%x dropped "
597			    "due to ingress filter\n", if_name(&sc->sc_if),
598			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
599#endif
600			if (rt)
601				rtfree(rt);
602			return -1;
603		}
604		rtfree(rt);
605	}
606
607	return 0;
608}
609
610static int
611stf_checkaddr6(sc, in6, inifp)
612	struct stf_softc *sc;
613	struct in6_addr *in6;
614	struct ifnet *inifp;	/* incoming interface */
615{
616	/*
617	 * check 6to4 addresses
618	 */
619	if (IN6_IS_ADDR_6TO4(in6)) {
620		struct in_addr in4;
621		bcopy(GET_V4(in6), &in4, sizeof(in4));
622		return stf_checkaddr4(sc, &in4, inifp);
623	}
624
625	/*
626	 * reject anything that look suspicious.  the test is implemented
627	 * in ip6_input too, but we check here as well to
628	 * (1) reject bad packets earlier, and
629	 * (2) to be safe against future ip6_input change.
630	 */
631	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
632		return -1;
633
634	return 0;
635}
636
637void
638in_stf_input(m, off)
639	struct mbuf *m;
640	int off;
641{
642	int proto;
643	struct stf_softc *sc;
644	struct ip *ip;
645	struct ip6_hdr *ip6;
646	u_int8_t otos, itos;
647	struct ifnet *ifp;
648
649	proto = mtod(m, struct ip *)->ip_p;
650
651	if (proto != IPPROTO_IPV6) {
652		m_freem(m);
653		return;
654	}
655
656	ip = mtod(m, struct ip *);
657
658	sc = (struct stf_softc *)encap_getarg(m);
659
660	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
661		m_freem(m);
662		return;
663	}
664
665	ifp = &sc->sc_if;
666
667#ifdef MAC
668	mac_create_mbuf_from_ifnet(ifp, m);
669#endif
670
671	/*
672	 * perform sanity check against outer src/dst.
673	 * for source, perform ingress filter as well.
674	 */
675	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
676	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
677		m_freem(m);
678		return;
679	}
680
681	otos = ip->ip_tos;
682	m_adj(m, off);
683
684	if (m->m_len < sizeof(*ip6)) {
685		m = m_pullup(m, sizeof(*ip6));
686		if (!m)
687			return;
688	}
689	ip6 = mtod(m, struct ip6_hdr *);
690
691	/*
692	 * perform sanity check against inner src/dst.
693	 * for source, perform ingress filter as well.
694	 */
695	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
696	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
697		m_freem(m);
698		return;
699	}
700
701	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
702	if ((ifp->if_flags & IFF_LINK1) != 0)
703		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
704	else
705		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
706	ip6->ip6_flow &= ~htonl(0xff << 20);
707	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
708
709	m->m_pkthdr.rcvif = ifp;
710
711	if (ifp->if_bpf) {
712		/*
713		 * We need to prepend the address family as
714		 * a four byte field.  Cons up a dummy header
715		 * to pacify bpf.  This is safe because bpf
716		 * will only read from the mbuf (i.e., it won't
717		 * try to free it or keep a pointer a to it).
718		 */
719		u_int32_t af = AF_INET6;
720#ifdef HAVE_OLD_BPF
721		struct mbuf m0;
722
723		m0.m_next = m;
724		m0.m_len = 4;
725		m0.m_data = (char *)&af;
726
727		BPF_MTAP(ifp, &m0);
728#else
729		bpf_mtap2(ifp->if_bpf, &af, sizeof(ah), m);
730#endif
731	}
732
733	/*
734	 * Put the packet to the network layer input queue according to the
735	 * specified address family.
736	 * See net/if_gif.c for possible issues with packet processing
737	 * reorder due to extra queueing.
738	 */
739	ifp->if_ipackets++;
740	ifp->if_ibytes += m->m_pkthdr.len;
741	netisr_dispatch(NETISR_IPV6, m);
742}
743
744/* ARGSUSED */
745static void
746stf_rtrequest(cmd, rt, info)
747	int cmd;
748	struct rtentry *rt;
749	struct rt_addrinfo *info;
750{
751	RT_LOCK_ASSERT(rt);
752
753	if (rt)
754		rt->rt_rmx.rmx_mtu = IPV6_MMTU;
755}
756
757static int
758stf_ioctl(ifp, cmd, data)
759	struct ifnet *ifp;
760	u_long cmd;
761	caddr_t data;
762{
763	struct ifaddr *ifa;
764	struct ifreq *ifr;
765	struct sockaddr_in6 *sin6;
766	struct in_addr addr;
767	int error;
768
769	error = 0;
770	switch (cmd) {
771	case SIOCSIFADDR:
772		ifa = (struct ifaddr *)data;
773		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
774			error = EAFNOSUPPORT;
775			break;
776		}
777		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
778		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
779			error = EINVAL;
780			break;
781		}
782		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
783		if (isrfc1918addr(&addr)) {
784			error = EINVAL;
785			break;
786		}
787
788		ifa->ifa_rtrequest = stf_rtrequest;
789		ifp->if_flags |= IFF_UP;
790		break;
791
792	case SIOCADDMULTI:
793	case SIOCDELMULTI:
794		ifr = (struct ifreq *)data;
795		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
796			;
797		else
798			error = EAFNOSUPPORT;
799		break;
800
801	default:
802		error = EINVAL;
803		break;
804	}
805
806	return error;
807}
808