if_stf.c revision 111119
1/*	$FreeBSD: head/sys/net/if_stf.c 111119 2003-02-19 05:47:46Z imp $	*/
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
141static LIST_HEAD(, stf_softc) stf_softc_list;
142
143static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
144static int ip_stf_ttl = 40;
145
146extern  struct domain inetdomain;
147struct protosw in_stf_protosw =
148{ SOCK_RAW,	&inetdomain,	IPPROTO_IPV6,	PR_ATOMIC|PR_ADDR,
149  in_stf_input,	(pr_output_t*)rip_output, 0,	rip_ctloutput,
150  0,
151  0,            0,              0,              0,
152  &rip_usrreqs
153};
154
155static int stfmodevent(module_t, int, void *);
156static int stf_encapcheck(const struct mbuf *, int, int, void *);
157static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
158static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
159	struct rtentry *);
160static int isrfc1918addr(struct in_addr *);
161static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
162	struct ifnet *);
163static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
164	struct ifnet *);
165static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
166static int stf_ioctl(struct ifnet *, u_long, caddr_t);
167
168int	stf_clone_create(struct if_clone *, int);
169void	stf_clone_destroy(struct ifnet *);
170
171/* only one clone is currently allowed */
172struct if_clone stf_cloner =
173    IF_CLONE_INITIALIZER(STFNAME, stf_clone_create, stf_clone_destroy, 0, 0);
174
175int
176stf_clone_create(ifc, unit)
177	struct if_clone *ifc;
178	int unit;
179{
180	struct stf_softc *sc;
181
182	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
183	sc->sc_if.if_name = STFNAME;
184	sc->sc_if.if_unit = unit;
185
186	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
187	    stf_encapcheck, &in_stf_protosw, sc);
188	if (sc->encap_cookie == NULL) {
189		printf("%s: attach failed\n", if_name(&sc->sc_if));
190		free(sc, M_STF);
191		return (ENOMEM);
192	}
193
194	sc->sc_if.if_mtu    = IPV6_MMTU;
195	sc->sc_if.if_ioctl  = stf_ioctl;
196	sc->sc_if.if_output = stf_output;
197	sc->sc_if.if_type   = IFT_STF;
198	sc->sc_if.if_snd.ifq_maxlen = IFQ_MAXLEN;
199	if_attach(&sc->sc_if);
200	bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int));
201	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
202	return (0);
203}
204
205void
206stf_clone_destroy(ifp)
207	struct ifnet *ifp;
208{
209	int err;
210	struct stf_softc *sc = (void *) ifp;
211
212	LIST_REMOVE(sc, sc_list);
213	err = encap_detach(sc->encap_cookie);
214	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
215	bpfdetach(ifp);
216	if_detach(ifp);
217
218	free(sc, M_STF);
219}
220
221static int
222stfmodevent(mod, type, data)
223	module_t mod;
224	int type;
225	void *data;
226{
227
228	switch (type) {
229	case MOD_LOAD:
230		LIST_INIT(&stf_softc_list);
231		if_clone_attach(&stf_cloner);
232
233		break;
234	case MOD_UNLOAD:
235		if_clone_detach(&stf_cloner);
236
237		while (!LIST_EMPTY(&stf_softc_list))
238			stf_clone_destroy(&LIST_FIRST(&stf_softc_list)->sc_if);
239		break;
240	}
241
242	return (0);
243}
244
245static moduledata_t stf_mod = {
246	"if_stf",
247	stfmodevent,
248	0
249};
250
251DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
252
253static int
254stf_encapcheck(m, off, proto, arg)
255	const struct mbuf *m;
256	int off;
257	int proto;
258	void *arg;
259{
260	struct ip ip;
261	struct in6_ifaddr *ia6;
262	struct stf_softc *sc;
263	struct in_addr a, b, mask;
264
265	sc = (struct stf_softc *)arg;
266	if (sc == NULL)
267		return 0;
268
269	if ((sc->sc_if.if_flags & IFF_UP) == 0)
270		return 0;
271
272	/* IFF_LINK0 means "no decapsulation" */
273	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
274		return 0;
275
276	if (proto != IPPROTO_IPV6)
277		return 0;
278
279	/* LINTED const cast */
280	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
281
282	if (ip.ip_v != 4)
283		return 0;
284
285	ia6 = stf_getsrcifa6(&sc->sc_if);
286	if (ia6 == NULL)
287		return 0;
288
289	/*
290	 * check if IPv4 dst matches the IPv4 address derived from the
291	 * local 6to4 address.
292	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
293	 */
294	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
295	    sizeof(ip.ip_dst)) != 0)
296		return 0;
297
298	/*
299	 * check if IPv4 src matches the IPv4 address derived from the
300	 * local 6to4 address masked by prefixmask.
301	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
302	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
303	 */
304	bzero(&a, sizeof(a));
305	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
306	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
307	a.s_addr &= mask.s_addr;
308	b = ip.ip_src;
309	b.s_addr &= mask.s_addr;
310	if (a.s_addr != b.s_addr)
311		return 0;
312
313	/* stf interface makes single side match only */
314	return 32;
315}
316
317static struct in6_ifaddr *
318stf_getsrcifa6(ifp)
319	struct ifnet *ifp;
320{
321	struct ifaddr *ia;
322	struct in_ifaddr *ia4;
323	struct sockaddr_in6 *sin6;
324	struct in_addr in;
325
326	for (ia = TAILQ_FIRST(&ifp->if_addrlist);
327	     ia;
328	     ia = TAILQ_NEXT(ia, ifa_list))
329	{
330		if (ia->ifa_addr == NULL)
331			continue;
332		if (ia->ifa_addr->sa_family != AF_INET6)
333			continue;
334		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
335		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
336			continue;
337
338		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
339		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
340			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
341				break;
342		if (ia4 == NULL)
343			continue;
344
345		return (struct in6_ifaddr *)ia;
346	}
347
348	return NULL;
349}
350
351static int
352stf_output(ifp, m, dst, rt)
353	struct ifnet *ifp;
354	struct mbuf *m;
355	struct sockaddr *dst;
356	struct rtentry *rt;
357{
358	struct stf_softc *sc;
359	struct sockaddr_in6 *dst6;
360	struct in_addr in4;
361	caddr_t ptr;
362	struct sockaddr_in *dst4;
363	u_int8_t tos;
364	struct ip *ip;
365	struct ip6_hdr *ip6;
366	struct in6_ifaddr *ia6;
367#ifdef MAC
368	int error;
369
370	error = mac_check_ifnet_transmit(ifp, m);
371	if (error) {
372		m_freem(m);
373		return (error);
374	}
375#endif
376
377	sc = (struct stf_softc*)ifp;
378	dst6 = (struct sockaddr_in6 *)dst;
379
380	/* just in case */
381	if ((ifp->if_flags & IFF_UP) == 0) {
382		m_freem(m);
383		ifp->if_oerrors++;
384		return ENETDOWN;
385	}
386
387	/*
388	 * If we don't have an ip4 address that match my inner ip6 address,
389	 * we shouldn't generate output.  Without this check, we'll end up
390	 * using wrong IPv4 source.
391	 */
392	ia6 = stf_getsrcifa6(ifp);
393	if (ia6 == NULL) {
394		m_freem(m);
395		ifp->if_oerrors++;
396		return ENETDOWN;
397	}
398
399	if (m->m_len < sizeof(*ip6)) {
400		m = m_pullup(m, sizeof(*ip6));
401		if (!m) {
402			ifp->if_oerrors++;
403			return ENOBUFS;
404		}
405	}
406	ip6 = mtod(m, struct ip6_hdr *);
407	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
408
409	/*
410	 * Pickup the right outer dst addr from the list of candidates.
411	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
412	 */
413	ptr = NULL;
414	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
415		ptr = GET_V4(&ip6->ip6_dst);
416	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
417		ptr = GET_V4(&dst6->sin6_addr);
418	else {
419		m_freem(m);
420		ifp->if_oerrors++;
421		return ENETUNREACH;
422	}
423	bcopy(ptr, &in4, sizeof(in4));
424
425#if NBPFILTER > 0
426	if (ifp->if_bpf) {
427		/*
428		 * We need to prepend the address family as
429		 * a four byte field.  Cons up a dummy header
430		 * to pacify bpf.  This is safe because bpf
431		 * will only read from the mbuf (i.e., it won't
432		 * try to free it or keep a pointer a to it).
433		 */
434		struct mbuf m0;
435		u_int32_t af = AF_INET6;
436
437		m0.m_next = m;
438		m0.m_len = 4;
439		m0.m_data = (char *)&af;
440
441#ifdef HAVE_OLD_BPF
442		BPF_MTAP(ifp, &m0);
443#else
444		bpf_mtap(ifp->if_bpf, &m0);
445#endif
446	}
447#endif /*NBPFILTER > 0*/
448
449	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
450	if (m && m->m_len < sizeof(struct ip))
451		m = m_pullup(m, sizeof(struct ip));
452	if (m == NULL) {
453		ifp->if_oerrors++;
454		return ENOBUFS;
455	}
456	ip = mtod(m, struct ip *);
457
458	bzero(ip, sizeof(*ip));
459
460	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
461	    &ip->ip_src, sizeof(ip->ip_src));
462	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
463	ip->ip_p = IPPROTO_IPV6;
464	ip->ip_ttl = ip_stf_ttl;
465	ip->ip_len = m->m_pkthdr.len;	/*host order*/
466	if (ifp->if_flags & IFF_LINK1)
467		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
468	else
469		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
470
471	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
472	if (dst4->sin_family != AF_INET ||
473	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
474		/* cache route doesn't match */
475		dst4->sin_family = AF_INET;
476		dst4->sin_len = sizeof(struct sockaddr_in);
477		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
478		if (sc->sc_ro.ro_rt) {
479			RTFREE(sc->sc_ro.ro_rt);
480			sc->sc_ro.ro_rt = NULL;
481		}
482	}
483
484	if (sc->sc_ro.ro_rt == NULL) {
485		rtalloc(&sc->sc_ro);
486		if (sc->sc_ro.ro_rt == NULL) {
487			m_freem(m);
488			ifp->if_oerrors++;
489			return ENETUNREACH;
490		}
491	}
492
493	ifp->if_opackets++;
494	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
495}
496
497static int
498isrfc1918addr(in)
499	struct in_addr *in;
500{
501	/*
502	 * returns 1 if private address range:
503	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
504	 */
505	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
506	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
507	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
508		return 1;
509
510	return 0;
511}
512
513static int
514stf_checkaddr4(sc, in, inifp)
515	struct stf_softc *sc;
516	struct in_addr *in;
517	struct ifnet *inifp;	/* incoming interface */
518{
519	struct in_ifaddr *ia4;
520
521	/*
522	 * reject packets with the following address:
523	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
524	 */
525	if (IN_MULTICAST(ntohl(in->s_addr)))
526		return -1;
527	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
528	case 0: case 127: case 255:
529		return -1;
530	}
531
532	/*
533	 * reject packets with private address range.
534	 * (requirement from RFC3056 section 2 1st paragraph)
535	 */
536	if (isrfc1918addr(in))
537		return -1;
538
539	/*
540	 * reject packets with broadcast
541	 */
542	for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
543	     ia4;
544	     ia4 = TAILQ_NEXT(ia4, ia_link))
545	{
546		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
547			continue;
548		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
549			return -1;
550	}
551
552	/*
553	 * perform ingress filter
554	 */
555	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
556		struct sockaddr_in sin;
557		struct rtentry *rt;
558
559		bzero(&sin, sizeof(sin));
560		sin.sin_family = AF_INET;
561		sin.sin_len = sizeof(struct sockaddr_in);
562		sin.sin_addr = *in;
563		rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
564		if (!rt || rt->rt_ifp != inifp) {
565#if 0
566			log(LOG_WARNING, "%s: packet from 0x%x dropped "
567			    "due to ingress filter\n", if_name(&sc->sc_if),
568			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
569#endif
570			if (rt)
571				rtfree(rt);
572			return -1;
573		}
574		rtfree(rt);
575	}
576
577	return 0;
578}
579
580static int
581stf_checkaddr6(sc, in6, inifp)
582	struct stf_softc *sc;
583	struct in6_addr *in6;
584	struct ifnet *inifp;	/* incoming interface */
585{
586	/*
587	 * check 6to4 addresses
588	 */
589	if (IN6_IS_ADDR_6TO4(in6)) {
590		struct in_addr in4;
591		bcopy(GET_V4(in6), &in4, sizeof(in4));
592		return stf_checkaddr4(sc, &in4, inifp);
593	}
594
595	/*
596	 * reject anything that look suspicious.  the test is implemented
597	 * in ip6_input too, but we check here as well to
598	 * (1) reject bad packets earlier, and
599	 * (2) to be safe against future ip6_input change.
600	 */
601	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
602		return -1;
603
604	return 0;
605}
606
607void
608in_stf_input(m, off)
609	struct mbuf *m;
610	int off;
611{
612	int proto;
613	struct stf_softc *sc;
614	struct ip *ip;
615	struct ip6_hdr *ip6;
616	u_int8_t otos, itos;
617	int len, isr;
618	struct ifqueue *ifq = NULL;
619	struct ifnet *ifp;
620
621	proto = mtod(m, struct ip *)->ip_p;
622
623	if (proto != IPPROTO_IPV6) {
624		m_freem(m);
625		return;
626	}
627
628	ip = mtod(m, struct ip *);
629
630	sc = (struct stf_softc *)encap_getarg(m);
631
632	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
633		m_freem(m);
634		return;
635	}
636
637	ifp = &sc->sc_if;
638
639#ifdef MAC
640	mac_create_mbuf_from_ifnet(ifp, m);
641#endif
642
643	/*
644	 * perform sanity check against outer src/dst.
645	 * for source, perform ingress filter as well.
646	 */
647	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
648	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
649		m_freem(m);
650		return;
651	}
652
653	otos = ip->ip_tos;
654	m_adj(m, off);
655
656	if (m->m_len < sizeof(*ip6)) {
657		m = m_pullup(m, sizeof(*ip6));
658		if (!m)
659			return;
660	}
661	ip6 = mtod(m, struct ip6_hdr *);
662
663	/*
664	 * perform sanity check against inner src/dst.
665	 * for source, perform ingress filter as well.
666	 */
667	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
668	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
669		m_freem(m);
670		return;
671	}
672
673	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
674	if ((ifp->if_flags & IFF_LINK1) != 0)
675		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
676	else
677		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
678	ip6->ip6_flow &= ~htonl(0xff << 20);
679	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
680
681	m->m_pkthdr.rcvif = ifp;
682
683	if (ifp->if_bpf) {
684		/*
685		 * We need to prepend the address family as
686		 * a four byte field.  Cons up a dummy header
687		 * to pacify bpf.  This is safe because bpf
688		 * will only read from the mbuf (i.e., it won't
689		 * try to free it or keep a pointer a to it).
690		 */
691		struct mbuf m0;
692		u_int32_t af = AF_INET6;
693
694		m0.m_next = m;
695		m0.m_len = 4;
696		m0.m_data = (char *)&af;
697
698#ifdef HAVE_OLD_BPF
699		BPF_MTAP(ifp, &m0);
700#else
701		bpf_mtap(ifp->if_bpf, &m0);
702#endif
703	}
704
705	/*
706	 * Put the packet to the network layer input queue according to the
707	 * specified address family.
708	 * See net/if_gif.c for possible issues with packet processing
709	 * reorder due to extra queueing.
710	 */
711	ifq = &ip6intrq;
712	isr = NETISR_IPV6;
713
714	len = m->m_pkthdr.len;
715	if (! IF_HANDOFF(ifq, m, NULL))
716		return;
717	schednetisr(isr);
718	ifp->if_ipackets++;
719	ifp->if_ibytes += len;
720}
721
722/* ARGSUSED */
723static void
724stf_rtrequest(cmd, rt, info)
725	int cmd;
726	struct rtentry *rt;
727	struct rt_addrinfo *info;
728{
729
730	if (rt)
731		rt->rt_rmx.rmx_mtu = IPV6_MMTU;
732}
733
734static int
735stf_ioctl(ifp, cmd, data)
736	struct ifnet *ifp;
737	u_long cmd;
738	caddr_t data;
739{
740	struct ifaddr *ifa;
741	struct ifreq *ifr;
742	struct sockaddr_in6 *sin6;
743	struct in_addr addr;
744	int error;
745
746	error = 0;
747	switch (cmd) {
748	case SIOCSIFADDR:
749		ifa = (struct ifaddr *)data;
750		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
751			error = EAFNOSUPPORT;
752			break;
753		}
754		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
755		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
756			error = EINVAL;
757			break;
758		}
759		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
760		if (isrfc1918addr(&addr)) {
761			error = EINVAL;
762			break;
763		}
764
765		ifa->ifa_rtrequest = stf_rtrequest;
766		ifp->if_flags |= IFF_UP;
767		break;
768
769	case SIOCADDMULTI:
770	case SIOCDELMULTI:
771		ifr = (struct ifreq *)data;
772		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
773			;
774		else
775			error = EAFNOSUPPORT;
776		break;
777
778	default:
779		error = EINVAL;
780		break;
781	}
782
783	return error;
784}
785