xref: /netbsd-src/sys/net/if_stf.c (revision a4ddc2c8fb9af816efe3b1c375a5530aef0e89e9)
1 /*	$NetBSD: if_stf.c,v 1.77 2011/10/28 20:13:32 dyoung 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 1.77 2011/10/28 20:13:32 dyoung 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/proc.h>
90 #include <sys/protosw.h>
91 #include <sys/queue.h>
92 #include <sys/syslog.h>
93 
94 #include <sys/cpu.h>
95 
96 #include <net/if.h>
97 #include <net/route.h>
98 #include <net/netisr.h>
99 #include <net/if_types.h>
100 #include <net/if_stf.h>
101 
102 #include <netinet/in.h>
103 #include <netinet/in_systm.h>
104 #include <netinet/ip.h>
105 #include <netinet/ip_var.h>
106 #include <netinet/in_var.h>
107 
108 #include <netinet/ip6.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet6/in6_gif.h>
111 #include <netinet6/in6_var.h>
112 #include <netinet/ip_ecn.h>
113 
114 #include <netinet/ip_encap.h>
115 
116 #include <net/net_osdep.h>
117 
118 #include "stf.h"
119 #include "gif.h"	/*XXX*/
120 
121 #include <net/bpf.h>
122 
123 #if NGIF > 0
124 #include <net/if_gif.h>
125 #endif
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 #if NGIF > 0
146 extern int ip_gif_ttl;	/*XXX*/
147 #else
148 static int ip_gif_ttl = 40;	/*XXX*/
149 #endif
150 
151 extern struct domain inetdomain;
152 static const struct protosw in_stf_protosw =
153 { SOCK_RAW,	&inetdomain,	IPPROTO_IPV6,	PR_ATOMIC|PR_ADDR,
154   in_stf_input, rip_output,	0,		rip_ctloutput,
155   rip_usrreq,
156   0,            0,              0,              0
157 };
158 
159 void	stfattach(int);
160 
161 static int stf_encapcheck(struct mbuf *, int, int, void *);
162 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
163 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
164 	struct rtentry *);
165 static int isrfc1918addr(const struct in_addr *);
166 static int stf_checkaddr4(struct stf_softc *, const struct in_addr *,
167 	struct ifnet *);
168 static int stf_checkaddr6(struct stf_softc *, const struct in6_addr *,
169 	struct ifnet *);
170 static void stf_rtrequest(int, struct rtentry *, const struct rt_addrinfo *);
171 static int stf_ioctl(struct ifnet *, u_long, void *);
172 
173 /* ARGSUSED */
174 void
175 stfattach(int count)
176 {
177 
178 	LIST_INIT(&stf_softc_list);
179 	if_clone_attach(&stf_cloner);
180 }
181 
182 static int
183 stf_clone_create(struct if_clone *ifc, int unit)
184 {
185 	struct stf_softc *sc;
186 
187 	if (LIST_FIRST(&stf_softc_list) != NULL) {
188 		/* Only one stf interface is allowed. */
189 		return (EEXIST);
190 	}
191 
192 	sc = malloc(sizeof(struct stf_softc), M_DEVBUF, M_WAIT|M_ZERO);
193 
194 	if_initname(&sc->sc_if, ifc->ifc_name, unit);
195 
196 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
197 	    stf_encapcheck, &in_stf_protosw, sc);
198 	if (sc->encap_cookie == NULL) {
199 		printf("%s: unable to attach encap\n", if_name(&sc->sc_if));
200 		free(sc, M_DEVBUF);
201 		return (EIO);	/* XXX */
202 	}
203 
204 	sc->sc_if.if_mtu    = STF_MTU;
205 	sc->sc_if.if_flags  = 0;
206 	sc->sc_if.if_ioctl  = stf_ioctl;
207 	sc->sc_if.if_output = stf_output;
208 	sc->sc_if.if_type   = IFT_STF;
209 	sc->sc_if.if_dlt    = DLT_NULL;
210 	if_attach(&sc->sc_if);
211 	if_alloc_sadl(&sc->sc_if);
212 	bpf_attach(&sc->sc_if, DLT_NULL, sizeof(u_int));
213 	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
214 	return (0);
215 }
216 
217 static int
218 stf_clone_destroy(struct ifnet *ifp)
219 {
220 	struct stf_softc *sc = (void *) ifp;
221 
222 	LIST_REMOVE(sc, sc_list);
223 	encap_detach(sc->encap_cookie);
224 	bpf_detach(ifp);
225 	if_detach(ifp);
226 	rtcache_free(&sc->sc_ro);
227 	free(sc, M_DEVBUF);
228 
229 	return (0);
230 }
231 
232 static int
233 stf_encapcheck(struct mbuf *m, int off, int proto, void *arg)
234 {
235 	struct ip ip;
236 	struct in6_ifaddr *ia6;
237 	struct stf_softc *sc;
238 	struct in_addr a, b;
239 
240 	sc = (struct stf_softc *)arg;
241 	if (sc == NULL)
242 		return 0;
243 
244 	if ((sc->sc_if.if_flags & IFF_UP) == 0)
245 		return 0;
246 
247 	/* IFF_LINK0 means "no decapsulation" */
248 	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
249 		return 0;
250 
251 	if (proto != IPPROTO_IPV6)
252 		return 0;
253 
254 	m_copydata(m, 0, sizeof(ip), (void *)&ip);
255 
256 	if (ip.ip_v != 4)
257 		return 0;
258 
259 	ia6 = stf_getsrcifa6(&sc->sc_if);
260 	if (ia6 == NULL)
261 		return 0;
262 
263 	/*
264 	 * check if IPv4 dst matches the IPv4 address derived from the
265 	 * local 6to4 address.
266 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
267 	 */
268 	if (memcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
269 	    sizeof(ip.ip_dst)) != 0)
270 		return 0;
271 
272 	/*
273 	 * check if IPv4 src matches the IPv4 address derived from the
274 	 * local 6to4 address masked by prefixmask.
275 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
276 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
277 	 */
278 	memset(&a, 0, sizeof(a));
279 	a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
280 	a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
281 	b = ip.ip_src;
282 	b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
283 	if (a.s_addr != b.s_addr)
284 		return 0;
285 
286 	/* stf interface makes single side match only */
287 	return 32;
288 }
289 
290 static struct in6_ifaddr *
291 stf_getsrcifa6(struct ifnet *ifp)
292 {
293 	struct ifaddr *ifa;
294 	struct in_ifaddr *ia4;
295 	struct sockaddr_in6 *sin6;
296 	struct in_addr in;
297 
298 	IFADDR_FOREACH(ifa, ifp)
299 	{
300 		if (ifa->ifa_addr == NULL)
301 			continue;
302 		if (ifa->ifa_addr->sa_family != AF_INET6)
303 			continue;
304 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
305 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
306 			continue;
307 
308 		memcpy(&in, GET_V4(&sin6->sin6_addr), sizeof(in));
309 		INADDR_TO_IA(in, ia4);
310 		if (ia4 == NULL)
311 			continue;
312 
313 		return (struct in6_ifaddr *)ifa;
314 	}
315 
316 	return NULL;
317 }
318 
319 static int
320 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
321     struct rtentry *rt0)
322 {
323 	struct rtentry *rt;
324 	struct stf_softc *sc;
325 	const struct sockaddr_in6 *dst6;
326 	const struct in_addr *in4;
327 	uint8_t tos;
328 	struct ip *ip;
329 	struct ip6_hdr *ip6;
330 	struct in6_ifaddr *ia6;
331 	union {
332 		struct sockaddr		dst;
333 		struct sockaddr_in	dst4;
334 	} u;
335 
336 	sc = (struct stf_softc*)ifp;
337 	dst6 = (const struct sockaddr_in6 *)dst;
338 
339 	/* just in case */
340 	if ((ifp->if_flags & IFF_UP) == 0) {
341 		m_freem(m);
342 		return ENETDOWN;
343 	}
344 
345 	/*
346 	 * If we don't have an ip4 address that match my inner ip6 address,
347 	 * we shouldn't generate output.  Without this check, we'll end up
348 	 * using wrong IPv4 source.
349 	 */
350 	ia6 = stf_getsrcifa6(ifp);
351 	if (ia6 == NULL) {
352 		m_freem(m);
353 		ifp->if_oerrors++;
354 		return ENETDOWN;
355 	}
356 
357 	if (m->m_len < sizeof(*ip6)) {
358 		m = m_pullup(m, sizeof(*ip6));
359 		if (m == NULL) {
360 			ifp->if_oerrors++;
361 			return ENOBUFS;
362 		}
363 	}
364 	ip6 = mtod(m, struct ip6_hdr *);
365 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
366 
367 	/*
368 	 * Pickup the right outer dst addr from the list of candidates.
369 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
370 	 */
371 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
372 		in4 = GET_V4(&ip6->ip6_dst);
373 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
374 		in4 = GET_V4(&dst6->sin6_addr);
375 	else {
376 		m_freem(m);
377 		ifp->if_oerrors++;
378 		return ENETUNREACH;
379 	}
380 
381 	bpf_mtap_af(ifp, AF_INET6, m);
382 
383 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
384 	if (m && m->m_len < sizeof(struct ip))
385 		m = m_pullup(m, sizeof(struct ip));
386 	if (m == NULL) {
387 		ifp->if_oerrors++;
388 		return ENOBUFS;
389 	}
390 	ip = mtod(m, struct ip *);
391 
392 	memset(ip, 0, sizeof(*ip));
393 
394 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
395 	    &ip->ip_src, sizeof(ip->ip_src));
396 	memcpy(&ip->ip_dst, in4, sizeof(ip->ip_dst));
397 	ip->ip_p = IPPROTO_IPV6;
398 	ip->ip_ttl = ip_gif_ttl;	/*XXX*/
399 	ip->ip_len = htons(m->m_pkthdr.len);
400 	if (ifp->if_flags & IFF_LINK1)
401 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
402 	else
403 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
404 
405 	sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
406 	if ((rt = rtcache_lookup(&sc->sc_ro, &u.dst)) == NULL) {
407 		m_freem(m);
408 		ifp->if_oerrors++;
409 		return ENETUNREACH;
410 	}
411 
412 	/* If the route constitutes infinite encapsulation, punt. */
413 	if (rt->rt_ifp == ifp) {
414 		rtcache_free(&sc->sc_ro);
415 		m_freem(m);
416 		ifp->if_oerrors++;
417 		return ENETUNREACH;
418 	}
419 
420 	ifp->if_opackets++;
421 	ifp->if_obytes += m->m_pkthdr.len - sizeof(struct ip);
422 	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
423 }
424 
425 static int
426 isrfc1918addr(const struct in_addr *in)
427 {
428 	/*
429 	 * returns 1 if private address range:
430 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
431 	 */
432 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
433 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
434 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
435 		return 1;
436 
437 	return 0;
438 }
439 
440 static int
441 stf_checkaddr4(struct stf_softc *sc, const struct in_addr *in,
442     struct ifnet *inifp /*incoming interface*/)
443 {
444 	struct in_ifaddr *ia4;
445 
446 	/*
447 	 * reject packets with the following address:
448 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
449 	 */
450 	if (IN_MULTICAST(in->s_addr))
451 		return -1;
452 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
453 	case 0: case 127: case 255:
454 		return -1;
455 	}
456 
457 	/*
458 	 * reject packets with private address range.
459 	 * (requirement from RFC3056 section 2 1st paragraph)
460 	 */
461 	if (isrfc1918addr(in))
462 		return -1;
463 
464 	/*
465 	 * reject packet with IPv4 link-local (169.254.0.0/16),
466 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
467 	 */
468 	if (((ntohl(in->s_addr) & 0xff000000) >> 24) == 169 &&
469 	    ((ntohl(in->s_addr) & 0x00ff0000) >> 16) == 254)
470 		return -1;
471 
472 	/*
473 	 * reject packets with broadcast
474 	 */
475 	TAILQ_FOREACH(ia4, &in_ifaddrhead, ia_list)
476 	{
477 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
478 			continue;
479 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
480 			return -1;
481 	}
482 
483 	/*
484 	 * perform ingress filter
485 	 */
486 	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
487 		struct sockaddr_in sin;
488 		struct rtentry *rt;
489 
490 		memset(&sin, 0, sizeof(sin));
491 		sin.sin_family = AF_INET;
492 		sin.sin_len = sizeof(struct sockaddr_in);
493 		sin.sin_addr = *in;
494 		rt = rtalloc1((struct sockaddr *)&sin, 0);
495 		if (!rt || rt->rt_ifp != inifp) {
496 #if 0
497 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
498 			    "due to ingress filter\n", if_name(&sc->sc_if),
499 			    (uint32_t)ntohl(sin.sin_addr.s_addr));
500 #endif
501 			if (rt)
502 				rtfree(rt);
503 			return -1;
504 		}
505 		rtfree(rt);
506 	}
507 
508 	return 0;
509 }
510 
511 static int
512 stf_checkaddr6(struct stf_softc *sc, const struct in6_addr *in6,
513     struct ifnet *inifp /*incoming interface*/)
514 {
515 
516 	/*
517 	 * check 6to4 addresses
518 	 */
519 	if (IN6_IS_ADDR_6TO4(in6))
520 		return stf_checkaddr4(sc, GET_V4(in6), inifp);
521 
522 	/*
523 	 * reject anything that look suspicious.  the test is implemented
524 	 * in ip6_input too, but we check here as well to
525 	 * (1) reject bad packets earlier, and
526 	 * (2) to be safe against future ip6_input change.
527 	 */
528 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
529 		return -1;
530 
531 	/*
532 	 * reject link-local and site-local unicast
533 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
534 	 */
535 	if (IN6_IS_ADDR_LINKLOCAL(in6) || IN6_IS_ADDR_SITELOCAL(in6))
536 		return -1;
537 
538 	/*
539 	 * reject node-local and link-local multicast
540 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
541 	 */
542 	if (IN6_IS_ADDR_MC_NODELOCAL(in6) || IN6_IS_ADDR_MC_LINKLOCAL(in6))
543 		return -1;
544 
545 	return 0;
546 }
547 
548 void
549 in_stf_input(struct mbuf *m, ...)
550 {
551 	int off, proto;
552 	struct stf_softc *sc;
553 	struct ip *ip;
554 	struct ip6_hdr *ip6;
555 	uint8_t otos, itos;
556 	int s, isr;
557 	struct ifqueue *ifq = NULL;
558 	struct ifnet *ifp;
559 	va_list ap;
560 
561 	va_start(ap, m);
562 	off = va_arg(ap, int);
563 	proto = va_arg(ap, int);
564 	va_end(ap);
565 
566 	if (proto != IPPROTO_IPV6) {
567 		m_freem(m);
568 		return;
569 	}
570 
571 	ip = mtod(m, struct ip *);
572 
573 	sc = (struct stf_softc *)encap_getarg(m);
574 
575 	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
576 		m_freem(m);
577 		return;
578 	}
579 
580 	ifp = &sc->sc_if;
581 
582 	/*
583 	 * perform sanity check against outer src/dst.
584 	 * for source, perform ingress filter as well.
585 	 */
586 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
587 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
588 		m_freem(m);
589 		return;
590 	}
591 
592 	otos = ip->ip_tos;
593 	m_adj(m, off);
594 
595 	if (m->m_len < sizeof(*ip6)) {
596 		m = m_pullup(m, sizeof(*ip6));
597 		if (!m)
598 			return;
599 	}
600 	ip6 = mtod(m, struct ip6_hdr *);
601 
602 	/*
603 	 * perform sanity check against inner src/dst.
604 	 * for source, perform ingress filter as well.
605 	 */
606 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
607 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
608 		m_freem(m);
609 		return;
610 	}
611 
612 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
613 	if ((ifp->if_flags & IFF_LINK1) != 0)
614 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
615 	else
616 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
617 	ip6->ip6_flow &= ~htonl(0xff << 20);
618 	ip6->ip6_flow |= htonl((uint32_t)itos << 20);
619 
620 	m->m_pkthdr.rcvif = ifp;
621 
622 	bpf_mtap_af(ifp, AF_INET6, m);
623 
624 	/*
625 	 * Put the packet to the network layer input queue according to the
626 	 * specified address family.
627 	 * See net/if_gif.c for possible issues with packet processing
628 	 * reorder due to extra queueing.
629 	 */
630 	ifq = &ip6intrq;
631 	isr = NETISR_IPV6;
632 
633 	s = splnet();
634 	if (IF_QFULL(ifq)) {
635 		IF_DROP(ifq);	/* update statistics */
636 		m_freem(m);
637 		splx(s);
638 		return;
639 	}
640 	IF_ENQUEUE(ifq, m);
641 	schednetisr(isr);
642 	ifp->if_ipackets++;
643 	ifp->if_ibytes += m->m_pkthdr.len;
644 	splx(s);
645 }
646 
647 /* ARGSUSED */
648 static void
649 stf_rtrequest(int cmd, struct rtentry *rt,
650     const struct rt_addrinfo *info)
651 {
652 	if (rt != NULL) {
653 		struct stf_softc *sc;
654 
655 		sc = LIST_FIRST(&stf_softc_list);
656 		rt->rt_rmx.rmx_mtu = (sc != NULL) ? sc->sc_if.if_mtu : STF_MTU;
657 	}
658 }
659 
660 static int
661 stf_ioctl(struct ifnet *ifp, u_long cmd, void *data)
662 {
663 	struct ifaddr		*ifa;
664 	struct ifreq		*ifr = data;
665 	struct sockaddr_in6	*sin6;
666 	int			error;
667 
668 	error = 0;
669 	switch (cmd) {
670 	case SIOCINITIFADDR:
671 		ifa = (struct ifaddr *)data;
672 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
673 			error = EAFNOSUPPORT;
674 			break;
675 		}
676 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
677 		if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr) &&
678 		    !isrfc1918addr(GET_V4(&sin6->sin6_addr))) {
679 			ifa->ifa_rtrequest = stf_rtrequest;
680 			ifp->if_flags |= IFF_UP;
681 		} else
682 			error = EINVAL;
683 		break;
684 
685 	case SIOCADDMULTI:
686 	case SIOCDELMULTI:
687 		if (ifr != NULL &&
688 		    ifreq_getaddr(cmd, ifr)->sa_family == AF_INET6)
689 			;
690 		else
691 			error = EAFNOSUPPORT;
692 		break;
693 
694 	case SIOCSIFMTU:
695 		if (ifr->ifr_mtu < STF_MTU_MIN || ifr->ifr_mtu > STF_MTU_MAX)
696 			return EINVAL;
697 		else if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET)
698 			error = 0;
699 		break;
700 
701 	default:
702 		error = ifioctl_common(ifp, cmd, data);
703 		break;
704 	}
705 
706 	return error;
707 }
708