xref: /netbsd-src/sys/net/if_stf.c (revision 37b34d511dea595d3ba03a661cf3b775038ea5f8)
1 /*	$NetBSD: if_stf.c,v 1.31 2002/09/17 06:32:49 itojun Exp $	*/
2 /*	$KAME: if_stf.c,v 1.62 2001/06/07 22:32:16 itojun 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 <sys/cdefs.h>
78 __KERNEL_RCSID(0, "$NetBSD: if_stf.c,v 1.31 2002/09/17 06:32:49 itojun Exp $");
79 
80 #include "opt_inet.h"
81 
82 #include <sys/param.h>
83 #include <sys/systm.h>
84 #include <sys/socket.h>
85 #include <sys/sockio.h>
86 #include <sys/mbuf.h>
87 #include <sys/errno.h>
88 #include <sys/ioctl.h>
89 #include <sys/protosw.h>
90 #include <sys/queue.h>
91 #include <sys/syslog.h>
92 #include <machine/cpu.h>
93 
94 #include <net/if.h>
95 #include <net/route.h>
96 #include <net/netisr.h>
97 #include <net/if_types.h>
98 #include <net/if_stf.h>
99 
100 #include <netinet/in.h>
101 #include <netinet/in_systm.h>
102 #include <netinet/ip.h>
103 #include <netinet/ip_var.h>
104 #include <netinet/in_var.h>
105 
106 #include <netinet/ip6.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/in6_gif.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 "bpfilter.h"
119 #include "stf.h"
120 #include "gif.h"	/*XXX*/
121 
122 #if NBPFILTER > 0
123 #include <net/bpf.h>
124 #endif
125 
126 #if NGIF > 0
127 #include <net/if_gif.h>
128 #endif
129 
130 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
131 #define GET_V4(x)	((struct in_addr *)(&(x)->s6_addr16[1]))
132 
133 struct stf_softc {
134 	struct ifnet	sc_if;	   /* common area */
135 	union {
136 		struct route  __sc_ro4;
137 		struct route_in6 __sc_ro6; /* just for safety */
138 	} __sc_ro46;
139 #define sc_ro	__sc_ro46.__sc_ro4
140 	const struct encaptab *encap_cookie;
141 	LIST_ENTRY(stf_softc) sc_list;
142 };
143 
144 LIST_HEAD(, stf_softc) stf_softc_list;
145 
146 int	stf_clone_create __P((struct if_clone *, int));
147 void	stf_clone_destroy __P((struct ifnet *));
148 
149 struct if_clone stf_cloner =
150     IF_CLONE_INITIALIZER("stf", stf_clone_create, stf_clone_destroy);
151 
152 #if NGIF > 0
153 extern int ip_gif_ttl;	/*XXX*/
154 #else
155 static int ip_gif_ttl = 40;	/*XXX*/
156 #endif
157 
158 extern struct domain inetdomain;
159 struct protosw in_stf_protosw =
160 { SOCK_RAW,	&inetdomain,	IPPROTO_IPV6,	PR_ATOMIC|PR_ADDR,
161   in_stf_input, rip_output,	0,		rip_ctloutput,
162   rip_usrreq,
163   0,            0,              0,              0
164 };
165 
166 void stfattach __P((int));
167 static int stf_encapcheck __P((const struct mbuf *, int, int, void *));
168 static struct in6_ifaddr *stf_getsrcifa6 __P((struct ifnet *));
169 static int stf_output __P((struct ifnet *, struct mbuf *, struct sockaddr *,
170 	struct rtentry *));
171 static int isrfc1918addr __P((struct in_addr *));
172 static int stf_checkaddr4 __P((struct stf_softc *, struct in_addr *,
173 	struct ifnet *));
174 static int stf_checkaddr6 __P((struct stf_softc *, struct in6_addr *,
175 	struct ifnet *));
176 static void stf_rtrequest __P((int, struct rtentry *, struct rt_addrinfo *));
177 static int stf_ioctl __P((struct ifnet *, u_long, caddr_t));
178 
179 /* ARGSUSED */
180 void
181 stfattach(count)
182 	int count;
183 {
184 
185 	LIST_INIT(&stf_softc_list);
186 	if_clone_attach(&stf_cloner);
187 }
188 
189 int
190 stf_clone_create(ifc, unit)
191 	struct if_clone *ifc;
192 	int unit;
193 {
194 	struct stf_softc *sc;
195 
196 	if (LIST_FIRST(&stf_softc_list) != NULL) {
197 		/* Only one stf interface is allowed. */
198 		return (EEXIST);
199 	}
200 
201 	sc = malloc(sizeof(struct stf_softc), M_DEVBUF, M_WAIT);
202 	memset(sc, 0, sizeof(struct stf_softc));
203 
204 	sprintf(sc->sc_if.if_xname, "%s%d", ifc->ifc_name, unit);
205 
206 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
207 	    stf_encapcheck, &in_stf_protosw, sc);
208 	if (sc->encap_cookie == NULL) {
209 		printf("%s: unable to attach encap\n", if_name(&sc->sc_if));
210 		free(sc, M_DEVBUF);
211 		return (EIO);	/* XXX */
212 	}
213 
214 	sc->sc_if.if_mtu    = IPV6_MMTU;
215 	sc->sc_if.if_flags  = 0;
216 	sc->sc_if.if_ioctl  = stf_ioctl;
217 	sc->sc_if.if_output = stf_output;
218 	sc->sc_if.if_type   = IFT_STF;
219 	sc->sc_if.if_dlt    = DLT_NULL;
220 	if_attach(&sc->sc_if);
221 	if_alloc_sadl(&sc->sc_if);
222 #if NBPFILTER > 0
223 	bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int));
224 #endif
225 	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
226 	return (0);
227 }
228 
229 void
230 stf_clone_destroy(ifp)
231 	struct ifnet *ifp;
232 {
233 	struct stf_softc *sc = (void *) ifp;
234 
235 	LIST_REMOVE(sc, sc_list);
236 	encap_detach(sc->encap_cookie);
237 #if NBPFILTER > 0
238 	bpfdetach(ifp);
239 #endif
240 	if_detach(ifp);
241 	free(sc, M_DEVBUF);
242 }
243 
244 static int
245 stf_encapcheck(m, off, proto, arg)
246 	const struct mbuf *m;
247 	int off;
248 	int proto;
249 	void *arg;
250 {
251 	struct ip ip;
252 	struct in6_ifaddr *ia6;
253 	struct stf_softc *sc;
254 	struct in_addr a, b;
255 
256 	sc = (struct stf_softc *)arg;
257 	if (sc == NULL)
258 		return 0;
259 
260 	if ((sc->sc_if.if_flags & IFF_UP) == 0)
261 		return 0;
262 
263 	/* IFF_LINK0 means "no decapsulation" */
264 	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
265 		return 0;
266 
267 	if (proto != IPPROTO_IPV6)
268 		return 0;
269 
270 	/* LINTED const cast */
271 	m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip);
272 
273 	if (ip.ip_v != 4)
274 		return 0;
275 
276 	ia6 = stf_getsrcifa6(&sc->sc_if);
277 	if (ia6 == NULL)
278 		return 0;
279 
280 	/*
281 	 * check if IPv4 dst matches the IPv4 address derived from the
282 	 * local 6to4 address.
283 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
284 	 */
285 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
286 	    sizeof(ip.ip_dst)) != 0)
287 		return 0;
288 
289 	/*
290 	 * check if IPv4 src matches the IPv4 address derived from the
291 	 * local 6to4 address masked by prefixmask.
292 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
293 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
294 	 */
295 	memset(&a, 0, sizeof(a));
296 	a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
297 	a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
298 	b = ip.ip_src;
299 	b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
300 	if (a.s_addr != b.s_addr)
301 		return 0;
302 
303 	/* stf interface makes single side match only */
304 	return 32;
305 }
306 
307 static struct in6_ifaddr *
308 stf_getsrcifa6(ifp)
309 	struct ifnet *ifp;
310 {
311 	struct ifaddr *ia;
312 	struct in_ifaddr *ia4;
313 	struct sockaddr_in6 *sin6;
314 	struct in_addr in;
315 
316 	TAILQ_FOREACH(ia, &ifp->if_addrlist, ifa_list)
317 	{
318 		if (ia->ifa_addr == NULL)
319 			continue;
320 		if (ia->ifa_addr->sa_family != AF_INET6)
321 			continue;
322 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
323 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
324 			continue;
325 
326 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
327 		INADDR_TO_IA(in, ia4);
328 		if (ia4 == NULL)
329 			continue;
330 
331 		return (struct in6_ifaddr *)ia;
332 	}
333 
334 	return NULL;
335 }
336 
337 static int
338 stf_output(ifp, m, dst, rt)
339 	struct ifnet *ifp;
340 	struct mbuf *m;
341 	struct sockaddr *dst;
342 	struct rtentry *rt;
343 {
344 	struct stf_softc *sc;
345 	struct sockaddr_in6 *dst6;
346 	struct in_addr *in4;
347 	struct sockaddr_in *dst4;
348 	u_int8_t tos;
349 	struct ip *ip;
350 	struct ip6_hdr *ip6;
351 	struct in6_ifaddr *ia6;
352 
353 	sc = (struct stf_softc*)ifp;
354 	dst6 = (struct sockaddr_in6 *)dst;
355 
356 	/* just in case */
357 	if ((ifp->if_flags & IFF_UP) == 0) {
358 		m_freem(m);
359 		return ENETDOWN;
360 	}
361 
362 	/*
363 	 * If we don't have an ip4 address that match my inner ip6 address,
364 	 * we shouldn't generate output.  Without this check, we'll end up
365 	 * using wrong IPv4 source.
366 	 */
367 	ia6 = stf_getsrcifa6(ifp);
368 	if (ia6 == NULL) {
369 		m_freem(m);
370 		ifp->if_oerrors++;
371 		return ENETDOWN;
372 	}
373 
374 	if (m->m_len < sizeof(*ip6)) {
375 		m = m_pullup(m, sizeof(*ip6));
376 		if (m == NULL) {
377 			ifp->if_oerrors++;
378 			return ENOBUFS;
379 		}
380 	}
381 	ip6 = mtod(m, struct ip6_hdr *);
382 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
383 
384 	/*
385 	 * Pickup the right outer dst addr from the list of candidates.
386 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
387 	 */
388 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
389 		in4 = GET_V4(&ip6->ip6_dst);
390 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
391 		in4 = GET_V4(&dst6->sin6_addr);
392 	else {
393 		m_freem(m);
394 		ifp->if_oerrors++;
395 		return ENETUNREACH;
396 	}
397 
398 #if NBPFILTER > 0
399 	if (ifp->if_bpf) {
400 		/*
401 		 * We need to prepend the address family as
402 		 * a four byte field.  Cons up a dummy header
403 		 * to pacify bpf.  This is safe because bpf
404 		 * will only read from the mbuf (i.e., it won't
405 		 * try to free it or keep a pointer a to it).
406 		 */
407 		struct mbuf m0;
408 		u_int32_t af = AF_INET6;
409 
410 		m0.m_next = m;
411 		m0.m_len = 4;
412 		m0.m_data = (char *)&af;
413 
414 #ifdef HAVE_OLD_BPF
415 		bpf_mtap(ifp, &m0);
416 #else
417 		bpf_mtap(ifp->if_bpf, &m0);
418 #endif
419 	}
420 #endif /*NBPFILTER > 0*/
421 
422 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
423 	if (m && m->m_len < sizeof(struct ip))
424 		m = m_pullup(m, sizeof(struct ip));
425 	if (m == NULL) {
426 		ifp->if_oerrors++;
427 		return ENOBUFS;
428 	}
429 	ip = mtod(m, struct ip *);
430 
431 	memset(ip, 0, sizeof(*ip));
432 
433 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
434 	    &ip->ip_src, sizeof(ip->ip_src));
435 	bcopy(in4, &ip->ip_dst, sizeof(ip->ip_dst));
436 	ip->ip_p = IPPROTO_IPV6;
437 	ip->ip_ttl = ip_gif_ttl;	/*XXX*/
438 	ip->ip_len = htons(m->m_pkthdr.len);
439 	if (ifp->if_flags & IFF_LINK1)
440 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
441 	else
442 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
443 
444 	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
445 	if (dst4->sin_family != AF_INET ||
446 	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
447 		/* cache route doesn't match */
448 		dst4->sin_family = AF_INET;
449 		dst4->sin_len = sizeof(struct sockaddr_in);
450 		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
451 		if (sc->sc_ro.ro_rt) {
452 			RTFREE(sc->sc_ro.ro_rt);
453 			sc->sc_ro.ro_rt = NULL;
454 		}
455 	}
456 
457 	if (sc->sc_ro.ro_rt == NULL) {
458 		rtalloc(&sc->sc_ro);
459 		if (sc->sc_ro.ro_rt == NULL) {
460 			m_freem(m);
461 			ifp->if_oerrors++;
462 			return ENETUNREACH;
463 		}
464 	}
465 
466 	ifp->if_opackets++;
467 	return ip_output(m, NULL, &sc->sc_ro, 0, NULL);
468 }
469 
470 static int
471 isrfc1918addr(in)
472 	struct in_addr *in;
473 {
474 	/*
475 	 * returns 1 if private address range:
476 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
477 	 */
478 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
479 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
480 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
481 		return 1;
482 
483 	return 0;
484 }
485 
486 static int
487 stf_checkaddr4(sc, in, inifp)
488 	struct stf_softc *sc;
489 	struct in_addr *in;
490 	struct ifnet *inifp;	/* incoming interface */
491 {
492 	struct in_ifaddr *ia4;
493 
494 	/*
495 	 * reject packets with the following address:
496 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
497 	 */
498 	if (IN_MULTICAST(in->s_addr))
499 		return -1;
500 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
501 	case 0: case 127: case 255:
502 		return -1;
503 	}
504 
505 	/*
506 	 * reject packets with private address range.
507 	 * (requirement from RFC3056 section 2 1st paragraph)
508 	 */
509 	if (isrfc1918addr(in))
510 		return -1;
511 
512 	/*
513 	 * reject packets with broadcast
514 	 */
515 	TAILQ_FOREACH(ia4, &in_ifaddr, ia_list)
516 	{
517 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
518 			continue;
519 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
520 			return -1;
521 	}
522 
523 	/*
524 	 * perform ingress filter
525 	 */
526 	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
527 		struct sockaddr_in sin;
528 		struct rtentry *rt;
529 
530 		memset(&sin, 0, sizeof(sin));
531 		sin.sin_family = AF_INET;
532 		sin.sin_len = sizeof(struct sockaddr_in);
533 		sin.sin_addr = *in;
534 		rt = rtalloc1((struct sockaddr *)&sin, 0);
535 		if (!rt || rt->rt_ifp != inifp) {
536 #if 0
537 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
538 			    "due to ingress filter\n", if_name(&sc->sc_if),
539 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
540 #endif
541 			if (rt)
542 				rtfree(rt);
543 			return -1;
544 		}
545 		rtfree(rt);
546 	}
547 
548 	return 0;
549 }
550 
551 static int
552 stf_checkaddr6(sc, in6, inifp)
553 	struct stf_softc *sc;
554 	struct in6_addr *in6;
555 	struct ifnet *inifp;	/* incoming interface */
556 {
557 	/*
558 	 * check 6to4 addresses
559 	 */
560 	if (IN6_IS_ADDR_6TO4(in6))
561 		return stf_checkaddr4(sc, GET_V4(in6), inifp);
562 
563 	/*
564 	 * reject anything that look suspicious.  the test is implemented
565 	 * in ip6_input too, but we check here as well to
566 	 * (1) reject bad packets earlier, and
567 	 * (2) to be safe against future ip6_input change.
568 	 */
569 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
570 		return -1;
571 
572 	return 0;
573 }
574 
575 void
576 #if __STDC__
577 in_stf_input(struct mbuf *m, ...)
578 #else
579 in_stf_input(m, va_alist)
580 	struct mbuf *m;
581 #endif
582 {
583 	int off, proto;
584 	struct stf_softc *sc;
585 	struct ip *ip;
586 	struct ip6_hdr *ip6;
587 	u_int8_t otos, itos;
588 	int s, isr;
589 	struct ifqueue *ifq = NULL;
590 	struct ifnet *ifp;
591 	va_list ap;
592 
593 	va_start(ap, m);
594 	off = va_arg(ap, int);
595 	proto = va_arg(ap, int);
596 	va_end(ap);
597 
598 	if (proto != IPPROTO_IPV6) {
599 		m_freem(m);
600 		return;
601 	}
602 
603 	ip = mtod(m, struct ip *);
604 
605 	sc = (struct stf_softc *)encap_getarg(m);
606 
607 	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
608 		m_freem(m);
609 		return;
610 	}
611 
612 	ifp = &sc->sc_if;
613 
614 	/*
615 	 * perform sanity check against outer src/dst.
616 	 * for source, perform ingress filter as well.
617 	 */
618 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
619 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
620 		m_freem(m);
621 		return;
622 	}
623 
624 	otos = ip->ip_tos;
625 	m_adj(m, off);
626 
627 	if (m->m_len < sizeof(*ip6)) {
628 		m = m_pullup(m, sizeof(*ip6));
629 		if (!m)
630 			return;
631 	}
632 	ip6 = mtod(m, struct ip6_hdr *);
633 
634 	/*
635 	 * perform sanity check against inner src/dst.
636 	 * for source, perform ingress filter as well.
637 	 */
638 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
639 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
640 		m_freem(m);
641 		return;
642 	}
643 
644 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
645 	if ((ifp->if_flags & IFF_LINK1) != 0)
646 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
647 	else
648 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
649 	ip6->ip6_flow &= ~htonl(0xff << 20);
650 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
651 
652 	m->m_pkthdr.rcvif = ifp;
653 
654 #if NBPFILTER > 0
655 	if (ifp->if_bpf) {
656 		/*
657 		 * We need to prepend the address family as
658 		 * a four byte field.  Cons up a dummy header
659 		 * to pacify bpf.  This is safe because bpf
660 		 * will only read from the mbuf (i.e., it won't
661 		 * try to free it or keep a pointer a to it).
662 		 */
663 		struct mbuf m0;
664 		u_int32_t af = AF_INET6;
665 
666 		m0.m_next = m;
667 		m0.m_len = 4;
668 		m0.m_data = (char *)&af;
669 
670 #ifdef HAVE_OLD_BPF
671 		bpf_mtap(ifp, &m0);
672 #else
673 		bpf_mtap(ifp->if_bpf, &m0);
674 #endif
675 	}
676 #endif /*NBPFILTER > 0*/
677 
678 	/*
679 	 * Put the packet to the network layer input queue according to the
680 	 * specified address family.
681 	 * See net/if_gif.c for possible issues with packet processing
682 	 * reorder due to extra queueing.
683 	 */
684 	ifq = &ip6intrq;
685 	isr = NETISR_IPV6;
686 
687 	s = splnet();
688 	if (IF_QFULL(ifq)) {
689 		IF_DROP(ifq);	/* update statistics */
690 		m_freem(m);
691 		splx(s);
692 		return;
693 	}
694 	IF_ENQUEUE(ifq, m);
695 	schednetisr(isr);
696 	ifp->if_ipackets++;
697 	ifp->if_ibytes += m->m_pkthdr.len;
698 	splx(s);
699 }
700 
701 /* ARGSUSED */
702 static void
703 stf_rtrequest(cmd, rt, info)
704 	int cmd;
705 	struct rtentry *rt;
706 	struct rt_addrinfo *info;
707 {
708 
709 	if (rt)
710 		rt->rt_rmx.rmx_mtu = IPV6_MMTU;
711 }
712 
713 static int
714 stf_ioctl(ifp, cmd, data)
715 	struct ifnet *ifp;
716 	u_long cmd;
717 	caddr_t data;
718 {
719 	struct ifaddr *ifa;
720 	struct ifreq *ifr;
721 	struct sockaddr_in6 *sin6;
722 	int error;
723 
724 	error = 0;
725 	switch (cmd) {
726 	case SIOCSIFADDR:
727 		ifa = (struct ifaddr *)data;
728 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
729 			error = EAFNOSUPPORT;
730 			break;
731 		}
732 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
733 		if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr) &&
734 		    !isrfc1918addr(GET_V4(&sin6->sin6_addr))) {
735 			ifa->ifa_rtrequest = stf_rtrequest;
736 			ifp->if_flags |= IFF_UP;
737 		} else
738 			error = EINVAL;
739 		break;
740 
741 	case SIOCADDMULTI:
742 	case SIOCDELMULTI:
743 		ifr = (struct ifreq *)data;
744 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
745 			;
746 		else
747 			error = EAFNOSUPPORT;
748 		break;
749 
750 	default:
751 		error = EINVAL;
752 		break;
753 	}
754 
755 	return error;
756 }
757