if_stf.c revision 126709
1/*	$FreeBSD: head/sys/net/if_stf.c 126709 2004-03-07 05:15:42Z 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
141static LIST_HEAD(, stf_softc) stf_softc_list;
142
143static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
144static const 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	if_initname(&sc->sc_if, ifc->ifc_name, unit);
184
185	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
186	    stf_encapcheck, &in_stf_protosw, sc);
187	if (sc->encap_cookie == NULL) {
188		printf("%s: attach failed\n", if_name(&sc->sc_if));
189		free(sc, M_STF);
190		return (ENOMEM);
191	}
192
193	sc->sc_if.if_mtu    = IPV6_MMTU;
194	sc->sc_if.if_ioctl  = stf_ioctl;
195	sc->sc_if.if_output = stf_output;
196	sc->sc_if.if_type   = IFT_STF;
197	sc->sc_if.if_snd.ifq_maxlen = IFQ_MAXLEN;
198	if_attach(&sc->sc_if);
199	bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int));
200	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
201	return (0);
202}
203
204void
205stf_clone_destroy(ifp)
206	struct ifnet *ifp;
207{
208	int err;
209	struct stf_softc *sc = (void *) ifp;
210
211	LIST_REMOVE(sc, sc_list);
212	err = encap_detach(sc->encap_cookie);
213	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
214	bpfdetach(ifp);
215	if_detach(ifp);
216
217	free(sc, M_STF);
218}
219
220static int
221stfmodevent(mod, type, data)
222	module_t mod;
223	int type;
224	void *data;
225{
226
227	switch (type) {
228	case MOD_LOAD:
229		LIST_INIT(&stf_softc_list);
230		if_clone_attach(&stf_cloner);
231
232		break;
233	case MOD_UNLOAD:
234		if_clone_detach(&stf_cloner);
235
236		while (!LIST_EMPTY(&stf_softc_list))
237			stf_clone_destroy(&LIST_FIRST(&stf_softc_list)->sc_if);
238		break;
239	}
240
241	return (0);
242}
243
244static moduledata_t stf_mod = {
245	"if_stf",
246	stfmodevent,
247	0
248};
249
250DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
251
252static int
253stf_encapcheck(m, off, proto, arg)
254	const struct mbuf *m;
255	int off;
256	int proto;
257	void *arg;
258{
259	struct ip ip;
260	struct in6_ifaddr *ia6;
261	struct stf_softc *sc;
262	struct in_addr a, b, mask;
263
264	sc = (struct stf_softc *)arg;
265	if (sc == NULL)
266		return 0;
267
268	if ((sc->sc_if.if_flags & IFF_UP) == 0)
269		return 0;
270
271	/* IFF_LINK0 means "no decapsulation" */
272	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
273		return 0;
274
275	if (proto != IPPROTO_IPV6)
276		return 0;
277
278	/* LINTED const cast */
279	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
280
281	if (ip.ip_v != 4)
282		return 0;
283
284	ia6 = stf_getsrcifa6(&sc->sc_if);
285	if (ia6 == NULL)
286		return 0;
287
288	/*
289	 * check if IPv4 dst matches the IPv4 address derived from the
290	 * local 6to4 address.
291	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
292	 */
293	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
294	    sizeof(ip.ip_dst)) != 0)
295		return 0;
296
297	/*
298	 * check if IPv4 src matches the IPv4 address derived from the
299	 * local 6to4 address masked by prefixmask.
300	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
301	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
302	 */
303	bzero(&a, sizeof(a));
304	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
305	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
306	a.s_addr &= mask.s_addr;
307	b = ip.ip_src;
308	b.s_addr &= mask.s_addr;
309	if (a.s_addr != b.s_addr)
310		return 0;
311
312	/* stf interface makes single side match only */
313	return 32;
314}
315
316static struct in6_ifaddr *
317stf_getsrcifa6(ifp)
318	struct ifnet *ifp;
319{
320	struct ifaddr *ia;
321	struct in_ifaddr *ia4;
322	struct sockaddr_in6 *sin6;
323	struct in_addr in;
324
325	for (ia = TAILQ_FIRST(&ifp->if_addrlist);
326	     ia;
327	     ia = TAILQ_NEXT(ia, ifa_list))
328	{
329		if (ia->ifa_addr == NULL)
330			continue;
331		if (ia->ifa_addr->sa_family != AF_INET6)
332			continue;
333		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
334		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
335			continue;
336
337		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
338		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
339			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
340				break;
341		if (ia4 == NULL)
342			continue;
343
344		return (struct in6_ifaddr *)ia;
345	}
346
347	return NULL;
348}
349
350static int
351stf_output(ifp, m, dst, rt)
352	struct ifnet *ifp;
353	struct mbuf *m;
354	struct sockaddr *dst;
355	struct rtentry *rt;
356{
357	struct stf_softc *sc;
358	struct sockaddr_in6 *dst6;
359	struct in_addr in4;
360	caddr_t ptr;
361	struct sockaddr_in *dst4;
362	u_int8_t tos;
363	struct ip *ip;
364	struct ip6_hdr *ip6;
365	struct in6_ifaddr *ia6;
366#ifdef MAC
367	int error;
368
369	error = mac_check_ifnet_transmit(ifp, m);
370	if (error) {
371		m_freem(m);
372		return (error);
373	}
374#endif
375
376	sc = (struct stf_softc*)ifp;
377	dst6 = (struct sockaddr_in6 *)dst;
378
379	/* just in case */
380	if ((ifp->if_flags & IFF_UP) == 0) {
381		m_freem(m);
382		ifp->if_oerrors++;
383		return ENETDOWN;
384	}
385
386	/*
387	 * If we don't have an ip4 address that match my inner ip6 address,
388	 * we shouldn't generate output.  Without this check, we'll end up
389	 * using wrong IPv4 source.
390	 */
391	ia6 = stf_getsrcifa6(ifp);
392	if (ia6 == NULL) {
393		m_freem(m);
394		ifp->if_oerrors++;
395		return ENETDOWN;
396	}
397
398	if (m->m_len < sizeof(*ip6)) {
399		m = m_pullup(m, sizeof(*ip6));
400		if (!m) {
401			ifp->if_oerrors++;
402			return ENOBUFS;
403		}
404	}
405	ip6 = mtod(m, struct ip6_hdr *);
406	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
407
408	/*
409	 * Pickup the right outer dst addr from the list of candidates.
410	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
411	 */
412	ptr = NULL;
413	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
414		ptr = GET_V4(&ip6->ip6_dst);
415	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
416		ptr = GET_V4(&dst6->sin6_addr);
417	else {
418		m_freem(m);
419		ifp->if_oerrors++;
420		return ENETUNREACH;
421	}
422	bcopy(ptr, &in4, sizeof(in4));
423
424#if NBPFILTER > 0
425	if (ifp->if_bpf) {
426		/*
427		 * We need to prepend the address family as
428		 * a four byte field.  Cons up a dummy header
429		 * to pacify bpf.  This is safe because bpf
430		 * will only read from the mbuf (i.e., it won't
431		 * try to free it or keep a pointer a to it).
432		 */
433		u_int32_t af = AF_INET6;
434#ifdef HAVE_OLD_BPF
435		struct mbuf m0;
436
437		m0.m_next = m;
438		m0.m_len = 4;
439		m0.m_data = (char *)&af;
440
441		BPF_MTAP(ifp, &m0);
442#else
443		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
444#endif
445	}
446#endif /*NBPFILTER > 0*/
447
448	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
449	if (m && m->m_len < sizeof(struct ip))
450		m = m_pullup(m, sizeof(struct ip));
451	if (m == NULL) {
452		ifp->if_oerrors++;
453		return ENOBUFS;
454	}
455	ip = mtod(m, struct ip *);
456
457	bzero(ip, sizeof(*ip));
458
459	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
460	    &ip->ip_src, sizeof(ip->ip_src));
461	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
462	ip->ip_p = IPPROTO_IPV6;
463	ip->ip_ttl = ip_stf_ttl;
464	ip->ip_len = m->m_pkthdr.len;	/*host order*/
465	if (ifp->if_flags & IFF_LINK1)
466		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
467	else
468		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
469
470	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
471	if (dst4->sin_family != AF_INET ||
472	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
473		/* cache route doesn't match */
474		dst4->sin_family = AF_INET;
475		dst4->sin_len = sizeof(struct sockaddr_in);
476		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
477		if (sc->sc_ro.ro_rt) {
478			RTFREE(sc->sc_ro.ro_rt);
479			sc->sc_ro.ro_rt = NULL;
480		}
481	}
482
483	if (sc->sc_ro.ro_rt == NULL) {
484		rtalloc(&sc->sc_ro);
485		if (sc->sc_ro.ro_rt == NULL) {
486			m_freem(m);
487			ifp->if_oerrors++;
488			return ENETUNREACH;
489		}
490	}
491
492	ifp->if_opackets++;
493	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
494}
495
496static int
497isrfc1918addr(in)
498	struct in_addr *in;
499{
500	/*
501	 * returns 1 if private address range:
502	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
503	 */
504	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
505	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
506	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
507		return 1;
508
509	return 0;
510}
511
512static int
513stf_checkaddr4(sc, in, inifp)
514	struct stf_softc *sc;
515	struct in_addr *in;
516	struct ifnet *inifp;	/* incoming interface */
517{
518	struct in_ifaddr *ia4;
519
520	/*
521	 * reject packets with the following address:
522	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
523	 */
524	if (IN_MULTICAST(ntohl(in->s_addr)))
525		return -1;
526	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
527	case 0: case 127: case 255:
528		return -1;
529	}
530
531	/*
532	 * reject packets with private address range.
533	 * (requirement from RFC3056 section 2 1st paragraph)
534	 */
535	if (isrfc1918addr(in))
536		return -1;
537
538	/*
539	 * reject packets with broadcast
540	 */
541	for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
542	     ia4;
543	     ia4 = TAILQ_NEXT(ia4, ia_link))
544	{
545		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
546			continue;
547		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
548			return -1;
549	}
550
551	/*
552	 * perform ingress filter
553	 */
554	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
555		struct sockaddr_in sin;
556		struct rtentry *rt;
557
558		bzero(&sin, sizeof(sin));
559		sin.sin_family = AF_INET;
560		sin.sin_len = sizeof(struct sockaddr_in);
561		sin.sin_addr = *in;
562		rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
563		if (!rt || rt->rt_ifp != inifp) {
564#if 0
565			log(LOG_WARNING, "%s: packet from 0x%x dropped "
566			    "due to ingress filter\n", if_name(&sc->sc_if),
567			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
568#endif
569			if (rt)
570				rtfree(rt);
571			return -1;
572		}
573		rtfree(rt);
574	}
575
576	return 0;
577}
578
579static int
580stf_checkaddr6(sc, in6, inifp)
581	struct stf_softc *sc;
582	struct in6_addr *in6;
583	struct ifnet *inifp;	/* incoming interface */
584{
585	/*
586	 * check 6to4 addresses
587	 */
588	if (IN6_IS_ADDR_6TO4(in6)) {
589		struct in_addr in4;
590		bcopy(GET_V4(in6), &in4, sizeof(in4));
591		return stf_checkaddr4(sc, &in4, inifp);
592	}
593
594	/*
595	 * reject anything that look suspicious.  the test is implemented
596	 * in ip6_input too, but we check here as well to
597	 * (1) reject bad packets earlier, and
598	 * (2) to be safe against future ip6_input change.
599	 */
600	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
601		return -1;
602
603	return 0;
604}
605
606void
607in_stf_input(m, off)
608	struct mbuf *m;
609	int off;
610{
611	int proto;
612	struct stf_softc *sc;
613	struct ip *ip;
614	struct ip6_hdr *ip6;
615	u_int8_t otos, itos;
616	struct ifnet *ifp;
617
618	proto = mtod(m, struct ip *)->ip_p;
619
620	if (proto != IPPROTO_IPV6) {
621		m_freem(m);
622		return;
623	}
624
625	ip = mtod(m, struct ip *);
626
627	sc = (struct stf_softc *)encap_getarg(m);
628
629	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
630		m_freem(m);
631		return;
632	}
633
634	ifp = &sc->sc_if;
635
636#ifdef MAC
637	mac_create_mbuf_from_ifnet(ifp, m);
638#endif
639
640	/*
641	 * perform sanity check against outer src/dst.
642	 * for source, perform ingress filter as well.
643	 */
644	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
645	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
646		m_freem(m);
647		return;
648	}
649
650	otos = ip->ip_tos;
651	m_adj(m, off);
652
653	if (m->m_len < sizeof(*ip6)) {
654		m = m_pullup(m, sizeof(*ip6));
655		if (!m)
656			return;
657	}
658	ip6 = mtod(m, struct ip6_hdr *);
659
660	/*
661	 * perform sanity check against inner src/dst.
662	 * for source, perform ingress filter as well.
663	 */
664	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
665	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
666		m_freem(m);
667		return;
668	}
669
670	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
671	if ((ifp->if_flags & IFF_LINK1) != 0)
672		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
673	else
674		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
675	ip6->ip6_flow &= ~htonl(0xff << 20);
676	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
677
678	m->m_pkthdr.rcvif = ifp;
679
680	if (ifp->if_bpf) {
681		/*
682		 * We need to prepend the address family as
683		 * a four byte field.  Cons up a dummy header
684		 * to pacify bpf.  This is safe because bpf
685		 * will only read from the mbuf (i.e., it won't
686		 * try to free it or keep a pointer a to it).
687		 */
688		u_int32_t af = AF_INET6;
689#ifdef HAVE_OLD_BPF
690		struct mbuf m0;
691
692		m0.m_next = m;
693		m0.m_len = 4;
694		m0.m_data = (char *)&af;
695
696		BPF_MTAP(ifp, &m0);
697#else
698		bpf_mtap2(ifp->if_bpf, &af, sizeof(ah), m);
699#endif
700	}
701
702	/*
703	 * Put the packet to the network layer input queue according to the
704	 * specified address family.
705	 * See net/if_gif.c for possible issues with packet processing
706	 * reorder due to extra queueing.
707	 */
708	ifp->if_ipackets++;
709	ifp->if_ibytes += m->m_pkthdr.len;
710	netisr_dispatch(NETISR_IPV6, m);
711}
712
713/* ARGSUSED */
714static void
715stf_rtrequest(cmd, rt, info)
716	int cmd;
717	struct rtentry *rt;
718	struct rt_addrinfo *info;
719{
720	RT_LOCK_ASSERT(rt);
721
722	if (rt)
723		rt->rt_rmx.rmx_mtu = IPV6_MMTU;
724}
725
726static int
727stf_ioctl(ifp, cmd, data)
728	struct ifnet *ifp;
729	u_long cmd;
730	caddr_t data;
731{
732	struct ifaddr *ifa;
733	struct ifreq *ifr;
734	struct sockaddr_in6 *sin6;
735	struct in_addr addr;
736	int error;
737
738	error = 0;
739	switch (cmd) {
740	case SIOCSIFADDR:
741		ifa = (struct ifaddr *)data;
742		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
743			error = EAFNOSUPPORT;
744			break;
745		}
746		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
747		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
748			error = EINVAL;
749			break;
750		}
751		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
752		if (isrfc1918addr(&addr)) {
753			error = EINVAL;
754			break;
755		}
756
757		ifa->ifa_rtrequest = stf_rtrequest;
758		ifp->if_flags |= IFF_UP;
759		break;
760
761	case SIOCADDMULTI:
762	case SIOCDELMULTI:
763		ifr = (struct ifreq *)data;
764		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
765			;
766		else
767			error = EAFNOSUPPORT;
768		break;
769
770	default:
771		error = EINVAL;
772		break;
773	}
774
775	return error;
776}
777