xref: /netbsd-src/sys/net/if_stf.c (revision 181254a7b1bdde6873432bffef2d2decc4b5c22f)
1 /*	$NetBSD: if_stf.c,v 1.107 2020/01/29 04:28:27 thorpej 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 1.107 2020/01/29 04:28:27 thorpej Exp $");
79 
80 #ifdef _KERNEL_OPT
81 #include "opt_inet.h"
82 #include "stf.h"
83 #endif
84 
85 #ifndef INET6
86 	#error "pseudo-device stf requires options INET6"
87 #endif
88 
89 #include <sys/param.h>
90 #include <sys/systm.h>
91 #include <sys/socket.h>
92 #include <sys/sockio.h>
93 #include <sys/mbuf.h>
94 #include <sys/errno.h>
95 #include <sys/ioctl.h>
96 #include <sys/proc.h>
97 #include <sys/queue.h>
98 #include <sys/syslog.h>
99 #include <sys/device.h>
100 #include <sys/module.h>
101 
102 #include <sys/cpu.h>
103 
104 #include <net/if.h>
105 #include <net/route.h>
106 #include <net/netisr.h>
107 #include <net/if_types.h>
108 #include <net/if_stf.h>
109 
110 #include <netinet/in.h>
111 #include <netinet/in_systm.h>
112 #include <netinet/ip.h>
113 #include <netinet/ip_var.h>
114 #include <netinet/in_var.h>
115 
116 #include <netinet/ip6.h>
117 #include <netinet6/ip6_var.h>
118 #include <netinet6/in6_var.h>
119 #include <netinet/ip_ecn.h>
120 
121 #include <netinet/ip_encap.h>
122 
123 #include <net/bpf.h>
124 
125 #include "ioconf.h"
126 
127 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
128 #define GET_V4(x)	((const struct in_addr *)(&(x)->s6_addr16[1]))
129 
130 struct stf_softc {
131 	struct ifnet	sc_if;	   /* common area */
132 	struct route	sc_ro;
133 	const struct encaptab *encap_cookie;
134 	LIST_ENTRY(stf_softc) sc_list;
135 };
136 
137 static LIST_HEAD(, stf_softc) stf_softc_list;
138 
139 static int	stf_clone_create(struct if_clone *, int);
140 static int	stf_clone_destroy(struct ifnet *);
141 
142 struct if_clone stf_cloner =
143     IF_CLONE_INITIALIZER("stf", stf_clone_create, stf_clone_destroy);
144 
145 static int ip_stf_ttl = STF_TTL;
146 
147 extern struct domain inetdomain;
148 
149 static const struct encapsw in_stf_encapsw =
150 {
151 	.encapsw4 = {
152 		.pr_input	= in_stf_input,
153 		.pr_ctlinput	= NULL,
154 	}
155 };
156 
157 static int stf_encapcheck(struct mbuf *, int, int, void *);
158 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
159 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
160 	const struct rtentry *);
161 static int isrfc1918addr(const struct in_addr *);
162 static int stf_checkaddr4(struct stf_softc *, const struct in_addr *,
163 	struct ifnet *);
164 static int stf_checkaddr6(struct stf_softc *, const struct in6_addr *,
165 	struct ifnet *);
166 static void stf_rtrequest(int, struct rtentry *, const struct rt_addrinfo *);
167 static int stf_ioctl(struct ifnet *, u_long, void *);
168 
169 /* ARGSUSED */
170 void
171 stfattach(int count)
172 {
173 
174 	/*
175 	 * Nothing to do here, initialization is handled by the
176 	 * module initialization code in stfinit() below).
177 	 */
178 }
179 
180 static void
181 stfinit(void)
182 {
183 
184 	LIST_INIT(&stf_softc_list);
185 	if_clone_attach(&stf_cloner);
186 }
187 
188 static int
189 stfdetach(void)
190 {
191 	int error = 0;
192 
193 	if (!LIST_EMPTY(&stf_softc_list))
194 		error = EBUSY;
195 
196 	if (error == 0)
197 		if_clone_detach(&stf_cloner);
198 
199 	return error;
200 }
201 
202 static int
203 stf_clone_create(struct if_clone *ifc, int unit)
204 {
205 	struct stf_softc *sc;
206 	int error;
207 
208 	sc = malloc(sizeof(struct stf_softc), M_DEVBUF, M_WAIT|M_ZERO);
209 	if_initname(&sc->sc_if, ifc->ifc_name, unit);
210 
211 	error = encap_lock_enter();
212 	if (error) {
213 		free(sc, M_DEVBUF);
214 		return error;
215 	}
216 
217 	if (LIST_FIRST(&stf_softc_list) != NULL) {
218 		/* Only one stf interface is allowed. */
219 		encap_lock_exit();
220 		free(sc, M_DEVBUF);
221 		return EEXIST;
222 	}
223 
224 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
225 	    stf_encapcheck, &in_stf_encapsw, sc);
226 	encap_lock_exit();
227 	if (sc->encap_cookie == NULL) {
228 		printf("%s: unable to attach encap\n", if_name(&sc->sc_if));
229 		free(sc, M_DEVBUF);
230 		return EIO;	/* XXX */
231 	}
232 
233 	sc->sc_if.if_mtu    = STF_MTU;
234 	sc->sc_if.if_flags  = 0;
235 	sc->sc_if.if_ioctl  = stf_ioctl;
236 	sc->sc_if.if_output = stf_output;
237 	sc->sc_if.if_type   = IFT_STF;
238 	sc->sc_if.if_dlt    = DLT_NULL;
239 	error = if_attach(&sc->sc_if);
240 	if (error != 0) {
241 		aprint_error("%s: if_initialize failed(%d)\n",
242 		    if_name(&sc->sc_if), error);
243 		encap_lock_enter();
244 		encap_detach(sc->encap_cookie);
245 		encap_lock_exit();
246 		free(sc, M_DEVBUF);
247 		return error;
248 	}
249 	if_alloc_sadl(&sc->sc_if);
250 	bpf_attach(&sc->sc_if, DLT_NULL, sizeof(u_int));
251 	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
252 	return 0;
253 }
254 
255 static int
256 stf_clone_destroy(struct ifnet *ifp)
257 {
258 	struct stf_softc *sc = (void *) ifp;
259 
260 	encap_lock_enter();
261 	LIST_REMOVE(sc, sc_list);
262 	encap_detach(sc->encap_cookie);
263 	encap_lock_exit();
264 	bpf_detach(ifp);
265 	if_detach(ifp);
266 	rtcache_free(&sc->sc_ro);
267 	free(sc, M_DEVBUF);
268 
269 	return 0;
270 }
271 
272 static int
273 stf_encapcheck(struct mbuf *m, int off, int proto, void *arg)
274 {
275 	struct ip ip;
276 	struct in6_ifaddr *ia6;
277 	struct stf_softc *sc;
278 	struct in_addr a, b;
279 
280 	sc = (struct stf_softc *)arg;
281 	if (sc == NULL)
282 		return 0;
283 
284 	if ((sc->sc_if.if_flags & IFF_UP) == 0)
285 		return 0;
286 
287 	/* IFF_LINK0 means "no decapsulation" */
288 	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
289 		return 0;
290 
291 	if (proto != IPPROTO_IPV6)
292 		return 0;
293 
294 	m_copydata(m, 0, sizeof(ip), (void *)&ip);
295 
296 	if (ip.ip_v != 4)
297 		return 0;
298 
299 	ia6 = stf_getsrcifa6(&sc->sc_if);
300 	if (ia6 == NULL)
301 		return 0;
302 
303 	/*
304 	 * check if IPv4 dst matches the IPv4 address derived from the
305 	 * local 6to4 address.
306 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
307 	 */
308 	if (memcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
309 	    sizeof(ip.ip_dst)) != 0)
310 		return 0;
311 
312 	/*
313 	 * check if IPv4 src matches the IPv4 address derived from the
314 	 * local 6to4 address masked by prefixmask.
315 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
316 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
317 	 */
318 	memset(&a, 0, sizeof(a));
319 	a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
320 	a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
321 	b = ip.ip_src;
322 	b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
323 	if (a.s_addr != b.s_addr)
324 		return 0;
325 
326 	/* stf interface makes single side match only */
327 	return 32;
328 }
329 
330 static struct in6_ifaddr *
331 stf_getsrcifa6(struct ifnet *ifp)
332 {
333 	struct ifaddr *ifa;
334 	struct in_ifaddr *ia4;
335 	struct sockaddr_in6 *sin6;
336 	struct in_addr in;
337 	int s;
338 
339 	s = pserialize_read_enter();
340 	IFADDR_READER_FOREACH(ifa, ifp) {
341 		if (ifa->ifa_addr->sa_family != AF_INET6)
342 			continue;
343 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
344 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
345 			continue;
346 
347 		memcpy(&in, GET_V4(&sin6->sin6_addr), sizeof(in));
348 		ia4 = in_get_ia(in);
349 		if (ia4 == NULL)
350 			continue;
351 
352 		pserialize_read_exit(s);
353 		/* TODO NOMPSAFE */
354 		return (struct in6_ifaddr *)ifa;
355 	}
356 	pserialize_read_exit(s);
357 
358 	return NULL;
359 }
360 
361 static int
362 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
363     const struct rtentry *rt0)
364 {
365 	struct rtentry *rt;
366 	struct stf_softc *sc;
367 	const struct sockaddr_in6 *dst6;
368 	const struct in_addr *in4;
369 	uint8_t tos;
370 	struct ip *ip;
371 	struct ip6_hdr *ip6;
372 	struct in6_ifaddr *ia6;
373 	union {
374 		struct sockaddr		dst;
375 		struct sockaddr_in	dst4;
376 	} u;
377 
378 	sc = (struct stf_softc*)ifp;
379 	dst6 = (const struct sockaddr_in6 *)dst;
380 
381 	/* just in case */
382 	if ((ifp->if_flags & IFF_UP) == 0) {
383 		m_freem(m);
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 		if_statinc(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 == NULL) {
402 			if_statinc(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 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
414 		in4 = GET_V4(&ip6->ip6_dst);
415 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
416 		in4 = GET_V4(&dst6->sin6_addr);
417 	else {
418 		m_freem(m);
419 		if_statinc(ifp, if_oerrors);
420 		return ENETUNREACH;
421 	}
422 
423 	bpf_mtap_af(ifp, AF_INET6, m, BPF_D_OUT);
424 
425 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
426 	if (m && m->m_len < sizeof(struct ip))
427 		m = m_pullup(m, sizeof(struct ip));
428 	if (m == NULL) {
429 		if_statinc(ifp, if_oerrors);
430 		return ENOBUFS;
431 	}
432 	ip = mtod(m, struct ip *);
433 
434 	memset(ip, 0, sizeof(*ip));
435 
436 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
437 	    &ip->ip_src, sizeof(ip->ip_src));
438 	memcpy(&ip->ip_dst, in4, sizeof(ip->ip_dst));
439 	ip->ip_p = IPPROTO_IPV6;
440 	ip->ip_ttl = ip_stf_ttl;
441 	ip->ip_len = htons(m->m_pkthdr.len);
442 	if (ifp->if_flags & IFF_LINK1)
443 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
444 	else
445 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
446 
447 	sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
448 	if ((rt = rtcache_lookup(&sc->sc_ro, &u.dst)) == NULL) {
449 		m_freem(m);
450 		if_statinc(ifp, if_oerrors);
451 		return ENETUNREACH;
452 	}
453 
454 	/* If the route constitutes infinite encapsulation, punt. */
455 	if (rt->rt_ifp == ifp) {
456 		rtcache_unref(rt, &sc->sc_ro);
457 		rtcache_free(&sc->sc_ro);
458 		m_freem(m);
459 		if_statinc(ifp, if_oerrors);
460 		return ENETUNREACH;
461 	}
462 	rtcache_unref(rt, &sc->sc_ro);
463 
464 	if_statadd2(ifp, if_opackets, 1,
465 	    if_obytes, m->m_pkthdr.len - sizeof(struct ip));
466 	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
467 }
468 
469 static int
470 isrfc1918addr(const struct in_addr *in)
471 {
472 	/*
473 	 * returns 1 if private address range:
474 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
475 	 */
476 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
477 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
478 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
479 		return 1;
480 
481 	return 0;
482 }
483 
484 static int
485 stf_checkaddr4(struct stf_softc *sc, const struct in_addr *in,
486     struct ifnet *inifp /*incoming interface*/)
487 {
488 	struct in_ifaddr *ia4;
489 
490 	/*
491 	 * reject packets with the following address:
492 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
493 	 */
494 	if (IN_MULTICAST(in->s_addr))
495 		return -1;
496 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
497 	case 0: case 127: case 255:
498 		return -1;
499 	}
500 
501 	/*
502 	 * reject packets with private address range.
503 	 * (requirement from RFC3056 section 2 1st paragraph)
504 	 */
505 	if (isrfc1918addr(in))
506 		return -1;
507 
508 	/*
509 	 * reject packet with IPv4 link-local (169.254.0.0/16),
510 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
511 	 */
512 	if (((ntohl(in->s_addr) & 0xff000000) >> 24) == 169 &&
513 	    ((ntohl(in->s_addr) & 0x00ff0000) >> 16) == 254)
514 		return -1;
515 
516 	/*
517 	 * reject packets with broadcast
518 	 */
519 	IN_ADDRLIST_READER_FOREACH(ia4) {
520 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
521 			continue;
522 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
523 			return -1;
524 	}
525 
526 	/*
527 	 * perform ingress filter
528 	 */
529 	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
530 		struct sockaddr_in sin;
531 		struct rtentry *rt;
532 
533 		memset(&sin, 0, sizeof(sin));
534 		sin.sin_family = AF_INET;
535 		sin.sin_len = sizeof(struct sockaddr_in);
536 		sin.sin_addr = *in;
537 		rt = rtalloc1((struct sockaddr *)&sin, 0);
538 		if (!rt || rt->rt_ifp != inifp) {
539 #if 0
540 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
541 			    "due to ingress filter\n", if_name(&sc->sc_if),
542 			    (uint32_t)ntohl(sin.sin_addr.s_addr));
543 #endif
544 			if (rt)
545 				rt_unref(rt);
546 			return -1;
547 		}
548 		rt_unref(rt);
549 	}
550 
551 	return 0;
552 }
553 
554 static int
555 stf_checkaddr6(struct stf_softc *sc, const struct in6_addr *in6,
556     struct ifnet *inifp /*incoming interface*/)
557 {
558 
559 	/*
560 	 * check 6to4 addresses
561 	 */
562 	if (IN6_IS_ADDR_6TO4(in6))
563 		return stf_checkaddr4(sc, GET_V4(in6), inifp);
564 
565 	/*
566 	 * reject anything that look suspicious.  the test is implemented
567 	 * in ip6_input too, but we check here as well to
568 	 * (1) reject bad packets earlier, and
569 	 * (2) to be safe against future ip6_input change.
570 	 */
571 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
572 		return -1;
573 
574 	/*
575 	 * reject link-local and site-local unicast
576 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
577 	 */
578 	if (IN6_IS_ADDR_LINKLOCAL(in6) || IN6_IS_ADDR_SITELOCAL(in6))
579 		return -1;
580 
581 	/*
582 	 * reject node-local and link-local multicast
583 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
584 	 */
585 	if (IN6_IS_ADDR_MC_NODELOCAL(in6) || IN6_IS_ADDR_MC_LINKLOCAL(in6))
586 		return -1;
587 
588 	return 0;
589 }
590 
591 void
592 in_stf_input(struct mbuf *m, int off, int proto, void *eparg)
593 {
594 	int s;
595 	struct stf_softc *sc = eparg;
596 	struct ip *ip;
597 	struct ip6_hdr *ip6;
598 	uint8_t otos, itos;
599 	struct ifnet *ifp;
600 	size_t pktlen;
601 
602 	KASSERT(sc != NULL);
603 
604 	if (proto != IPPROTO_IPV6) {
605 		m_freem(m);
606 		return;
607 	}
608 
609 	ip = mtod(m, struct ip *);
610 
611 	if ((sc->sc_if.if_flags & IFF_UP) == 0) {
612 		m_freem(m);
613 		return;
614 	}
615 
616 	ifp = &sc->sc_if;
617 
618 	/*
619 	 * perform sanity check against outer src/dst.
620 	 * for source, perform ingress filter as well.
621 	 */
622 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
623 	    stf_checkaddr4(sc, &ip->ip_src, m_get_rcvif_NOMPSAFE(m)) < 0) {
624 		m_freem(m);
625 		return;
626 	}
627 
628 	otos = ip->ip_tos;
629 	m_adj(m, off);
630 
631 	if (m->m_len < sizeof(*ip6)) {
632 		m = m_pullup(m, sizeof(*ip6));
633 		if (!m)
634 			return;
635 	}
636 	ip6 = mtod(m, struct ip6_hdr *);
637 
638 	/*
639 	 * perform sanity check against inner src/dst.
640 	 * for source, perform ingress filter as well.
641 	 */
642 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
643 	    stf_checkaddr6(sc, &ip6->ip6_src, m_get_rcvif_NOMPSAFE(m)) < 0) {
644 		m_freem(m);
645 		return;
646 	}
647 
648 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
649 	if ((ifp->if_flags & IFF_LINK1) != 0)
650 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
651 	else
652 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
653 	ip6->ip6_flow &= ~htonl(0xff << 20);
654 	ip6->ip6_flow |= htonl((uint32_t)itos << 20);
655 
656 	pktlen = m->m_pkthdr.len;
657 	m_set_rcvif(m, ifp);
658 
659 	bpf_mtap_af(ifp, AF_INET6, m, BPF_D_IN);
660 
661 	/*
662 	 * Put the packet to the network layer input queue according to the
663 	 * specified address family.
664 	 * See net/if_gif.c for possible issues with packet processing
665 	 * reorder due to extra queueing.
666 	 */
667 
668 	s = splnet();
669 	if (__predict_true(pktq_enqueue(ip6_pktq, m, 0))) {
670 		if_statadd2(ifp, if_ipackets, 1, if_ibytes, pktlen);
671 	} else {
672 		m_freem(m);
673 	}
674 	splx(s);
675 
676 	return;
677 }
678 
679 /* ARGSUSED */
680 static void
681 stf_rtrequest(int cmd, struct rtentry *rt,
682     const struct rt_addrinfo *info)
683 {
684 	if (rt != NULL) {
685 		struct stf_softc *sc;
686 
687 		sc = LIST_FIRST(&stf_softc_list);
688 		rt->rt_rmx.rmx_mtu = (sc != NULL) ? sc->sc_if.if_mtu : STF_MTU;
689 	}
690 }
691 
692 static int
693 stf_ioctl(struct ifnet *ifp, u_long cmd, void *data)
694 {
695 	struct ifaddr		*ifa;
696 	struct ifreq		*ifr = data;
697 	struct sockaddr_in6	*sin6;
698 	int			error;
699 
700 	error = 0;
701 	switch (cmd) {
702 	case SIOCINITIFADDR:
703 		ifa = (struct ifaddr *)data;
704 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
705 			error = EAFNOSUPPORT;
706 			break;
707 		}
708 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
709 		if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr) &&
710 		    !isrfc1918addr(GET_V4(&sin6->sin6_addr))) {
711 			ifa->ifa_rtrequest = stf_rtrequest;
712 			ifp->if_flags |= IFF_UP;
713 		} else
714 			error = EINVAL;
715 		break;
716 
717 	case SIOCADDMULTI:
718 	case SIOCDELMULTI:
719 		if (ifr != NULL &&
720 		    ifreq_getaddr(cmd, ifr)->sa_family == AF_INET6)
721 			;
722 		else
723 			error = EAFNOSUPPORT;
724 		break;
725 
726 	case SIOCSIFMTU:
727 		if (ifr->ifr_mtu < STF_MTU_MIN || ifr->ifr_mtu > STF_MTU_MAX)
728 			return EINVAL;
729 		else if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
730 			error = 0;
731 		break;
732 
733 	default:
734 		error = ifioctl_common(ifp, cmd, data);
735 		break;
736 	}
737 
738 	return error;
739 }
740 
741 /*
742  * Module infrastructure
743  */
744 #include "if_module.h"
745 
746 IF_MODULE(MODULE_CLASS_DRIVER, stf, NULL)
747