xref: /dragonfly/sys/net/stf/if_stf.c (revision 7485684f)
1 /*	$FreeBSD: src/sys/net/if_stf.c,v 1.1.2.11 2003/01/23 21:06:44 sam Exp $	*/
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 
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/socket.h>
83 #include <sys/sockio.h>
84 #include <sys/mbuf.h>
85 #include <sys/errno.h>
86 #include <sys/protosw.h>
87 #include <sys/kernel.h>
88 #include <machine/cpu.h>
89 
90 #include <sys/malloc.h>
91 
92 #include <net/if.h>
93 #include <net/route.h>
94 #include <net/netisr.h>
95 #include <net/if_types.h>
96 #include <net/ifq_var.h>
97 #include <net/netisr2.h>
98 #include "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_var.h>
109 #include <netinet6/nd6.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 IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
121 #define GET_V4(x)	((struct in_addr *)(&(x)->s6_addr16[1]))
122 
123 struct stf_softc {
124 	struct ifnet	sc_if;	   /* common area */
125 	struct route	*route_pcpu;
126 	const struct encaptab *encap_cookie;
127 };
128 
129 static struct stf_softc *stf;
130 
131 static MALLOC_DEFINE(M_STF, "stf", "6to4 Tunnel Interface");
132 static int ip_stf_ttl = 40;
133 
134 extern  struct domain inetdomain;
135 struct protosw in_stf_protosw =
136     {
137 	.pr_type = SOCK_RAW,
138 	.pr_domain = &inetdomain,
139 	.pr_protocol = IPPROTO_IPV6,
140 	.pr_flags = PR_ATOMIC|PR_ADDR,
141 
142 	.pr_input = in_stf_input,
143 	.pr_output = rip_output,
144 	.pr_ctlinput = NULL,
145 	.pr_ctloutput = rip_ctloutput,
146 
147 	.pr_usrreqs = &rip_usrreqs
148     };
149 
150 static int stfmodevent (module_t, int, void *);
151 static int stf_encapcheck (const struct mbuf *, int, int, void *);
152 static struct in6_ifaddr *stf_getsrcifa6 (struct ifnet *);
153 static int stf_output (struct ifnet *, struct mbuf *, struct sockaddr *,
154 	struct rtentry *);
155 static int stf_checkaddr4 (struct stf_softc *, struct in_addr *,
156 	struct ifnet *);
157 static int stf_checkaddr6 (struct stf_softc *, struct in6_addr *,
158 	struct ifnet *);
159 static void stf_rtrequest (int, struct rtentry *);
160 static int stf_ioctl (struct ifnet *, u_long, caddr_t, struct ucred *);
161 
162 static int
163 stfmodevent(module_t mod, int type, void *data)
164 {
165 	struct stf_softc *sc;
166 	int err, cpu;
167 	const struct encaptab *p;
168 
169 	switch (type) {
170 	case MOD_LOAD:
171 		stf = kmalloc(sizeof(struct stf_softc), M_STF,
172 		    M_WAITOK | M_ZERO);
173 		sc = stf;
174 
175 		bzero(sc, sizeof(*sc));
176 		if_initname(&(sc->sc_if), "stf", 0);
177 
178 		p = encap_attach_func(AF_INET, IPPROTO_IPV6, stf_encapcheck,
179 		    (void *)&in_stf_protosw, sc);
180 		if (p == NULL) {
181 			kprintf("%s: attach failed\n", if_name(&sc->sc_if));
182 			return (ENOMEM);
183 		}
184 		sc->encap_cookie = p;
185 		sc->route_pcpu = kmalloc(netisr_ncpus * sizeof(struct route),
186 		    M_STF, M_WAITOK | M_ZERO);
187 
188 		sc->sc_if.if_mtu    = IPV6_MMTU;
189 		sc->sc_if.if_flags  = 0;
190 		sc->sc_if.if_ioctl  = stf_ioctl;
191 		sc->sc_if.if_output = stf_output;
192 		sc->sc_if.if_type   = IFT_STF;
193 #if 0
194 		/* turn off ingress filter */
195 		sc->sc_if.if_flags  |= IFF_LINK2;
196 #endif
197 		ifq_set_maxlen(&sc->sc_if.if_snd, IFQ_MAXLEN);
198 		if_attach(&sc->sc_if, NULL);
199 		bpfattach(&sc->sc_if, DLT_NULL, sizeof(uint32_t));
200 		/*
201 		 * 6to4 interface is very special: no multicast, no linklocal.
202 		 * RFC2529 specifies how to make linklocals, but there's no
203 		 * use and it is rather harmful to have one.
204 		 *
205 		 * NOTE: ND_IFINFO() is only available after if_attach().
206 		 */
207 		ND_IFINFO(&sc->sc_if)->flags &= ~ND6_IFF_AUTO_LINKLOCAL;
208 		ND_IFINFO(&sc->sc_if)->flags |= ND6_IFF_NO_DAD;
209 		break;
210 	case MOD_UNLOAD:
211 		sc = stf;
212 		bpfdetach(&sc->sc_if);
213 		if_detach(&sc->sc_if);
214 		err = encap_detach(sc->encap_cookie);
215 		KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
216 		for (cpu = 0; cpu < netisr_ncpus; ++cpu) {
217 			if (sc->route_pcpu[cpu].ro_rt != NULL) {
218 				rtfree_async(sc->route_pcpu[cpu].ro_rt);
219 				sc->route_pcpu[cpu].ro_rt = NULL;
220 			}
221 		}
222 		kfree(sc->route_pcpu, M_STF);
223 		kfree(sc, M_STF);
224 		break;
225 	}
226 
227 	return (0);
228 }
229 
230 static moduledata_t stf_mod = {
231 	"if_stf",
232 	stfmodevent,
233 	0
234 };
235 
236 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
237 
238 static int
239 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
240 {
241 	struct ip ip;
242 	struct in6_ifaddr *ia6;
243 	struct stf_softc *sc;
244 	struct in_addr a, b;
245 
246 	sc = (struct stf_softc *)arg;
247 	if (sc == NULL)
248 		return 0;
249 
250 	if ((sc->sc_if.if_flags & IFF_UP) == 0)
251 		return 0;
252 
253 	/* IFF_LINK0 means "no decapsulation" */
254 	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
255 		return 0;
256 
257 	if (proto != IPPROTO_IPV6)
258 		return 0;
259 
260 	m_copydata(m, 0, sizeof(ip), &ip);
261 
262 	if (ip.ip_v != 4)
263 		return 0;
264 
265 	ia6 = stf_getsrcifa6(&sc->sc_if);
266 	if (ia6 == NULL)
267 		return 0;
268 
269 	/*
270 	 * check if IPv4 dst matches the IPv4 address derived from the
271 	 * local 6to4 address.
272 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
273 	 */
274 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
275 	    sizeof(ip.ip_dst)) != 0)
276 		return 0;
277 
278 	/*
279 	 * check if IPv4 src matches the IPv4 address derived from the
280 	 * local 6to4 address masked by prefixmask.
281 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
282 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
283 	 */
284 	bzero(&a, sizeof(a));
285 	a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
286 	a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
287 	b = ip.ip_src;
288 	b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
289 	if (a.s_addr != b.s_addr)
290 		return 0;
291 
292 	/* stf interface makes single side match only */
293 	return 32;
294 }
295 
296 static struct in6_ifaddr *
297 stf_getsrcifa6(struct ifnet *ifp)
298 {
299 	struct ifaddr_container *ifac;
300 	struct sockaddr_in6 *sin6;
301 	struct in_addr in;
302 
303 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
304 		struct ifaddr *ia = ifac->ifa;
305 		struct in_ifaddr_container *iac;
306 
307 		if (ia->ifa_addr == NULL)
308 			continue;
309 		if (ia->ifa_addr->sa_family != AF_INET6)
310 			continue;
311 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
312 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
313 			continue;
314 
315 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
316 		LIST_FOREACH(iac, INADDR_HASH(in.s_addr), ia_hash) {
317 			if (iac->ia->ia_addr.sin_addr.s_addr == in.s_addr)
318 				break;
319 		}
320 		if (iac == NULL)
321 			continue;
322 
323 		return (struct in6_ifaddr *)ia;
324 	}
325 
326 	return NULL;
327 }
328 
329 static int
330 stf_output_serialized(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
331 		      struct rtentry *rt)
332 {
333 	struct stf_softc *sc;
334 	struct sockaddr_in6 *dst6;
335 	struct in_addr *in4;
336 	struct sockaddr_in *dst4;
337 	u_int8_t tos;
338 	struct ip *ip;
339 	struct ip6_hdr *ip6;
340 	struct in6_ifaddr *ia6;
341 	struct route *ro;
342 	static const uint32_t af = AF_INET6;
343 
344 	ASSERT_NETISR_NCPUS(mycpuid);
345 
346 	sc = (struct stf_softc*)ifp;
347 	dst6 = (struct sockaddr_in6 *)dst;
348 
349 	/* just in case */
350 	if ((ifp->if_flags & IFF_UP) == 0) {
351 		m_freem(m);
352 		return ENETDOWN;
353 	}
354 
355 	/*
356 	 * If we don't have an ip4 address that match my inner ip6 address,
357 	 * we shouldn't generate output.  Without this check, we'll end up
358 	 * using wrong IPv4 source.
359 	 */
360 	ia6 = stf_getsrcifa6(ifp);
361 	if (ia6 == NULL) {
362 		m_freem(m);
363 		return ENETDOWN;
364 	}
365 
366 	if (m->m_len < sizeof(*ip6)) {
367 		m = m_pullup(m, sizeof(*ip6));
368 		if (!m)
369 			return ENOBUFS;
370 	}
371 	ip6 = mtod(m, struct ip6_hdr *);
372 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
373 
374 	/*
375 	 * Pickup the right outer dst addr from the list of candidates.
376 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
377 	 */
378 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
379 		in4 = GET_V4(&ip6->ip6_dst);
380 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
381 		in4 = GET_V4(&dst6->sin6_addr);
382 	else {
383 		m_freem(m);
384 		return ENETUNREACH;
385 	}
386 
387 	if (ifp->if_bpf) {
388 		bpf_gettoken();
389 		if (ifp->if_bpf)
390 			bpf_ptap(ifp->if_bpf, m, &af, sizeof(af));
391 		bpf_reltoken();
392 	}
393 
394 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
395 	if (m && m->m_len < sizeof(struct ip))
396 		m = m_pullup(m, sizeof(struct ip));
397 	if (m == NULL)
398 		return ENOBUFS;
399 	ip = mtod(m, struct ip *);
400 
401 	bzero(ip, sizeof(*ip));
402 
403 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
404 	    &ip->ip_src, sizeof(ip->ip_src));
405 	bcopy(in4, &ip->ip_dst, sizeof(ip->ip_dst));
406 	ip->ip_p = IPPROTO_IPV6;
407 	ip->ip_ttl = ip_stf_ttl;
408 	ip->ip_len = htons(m->m_pkthdr.len);	/* network order */
409 	if (ifp->if_flags & IFF_LINK1)
410 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
411 	else
412 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
413 
414 	ro = &sc->route_pcpu[mycpuid];
415 	dst4 = (struct sockaddr_in *)&ro->ro_dst;
416 	if (dst4->sin_family != AF_INET ||
417 	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
418 		/* cache route doesn't match */
419 		dst4->sin_family = AF_INET;
420 		dst4->sin_len = sizeof(struct sockaddr_in);
421 		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
422 		if (ro->ro_rt) {
423 			RTFREE(ro->ro_rt);
424 			ro->ro_rt = NULL;
425 		}
426 	}
427 	if (ro->ro_rt != NULL && (ro->ro_rt->rt_flags & RTF_UP) == 0) {
428 		RTFREE(ro->ro_rt);
429 		ro->ro_rt = NULL;
430 	}
431 
432 	if (ro->ro_rt == NULL) {
433 		rtalloc(ro);
434 		if (ro->ro_rt == NULL) {
435 			m_freem(m);
436 			return ENETUNREACH;
437 		}
438 	}
439 
440 	return ip_output(m, NULL, ro, IP_DEBUGROUTE, NULL, NULL);
441 }
442 
443 static int
444 stf_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
445 	   struct rtentry *rt)
446 {
447 	struct ifaltq_subque *ifsq = ifq_get_subq_default(&ifp->if_snd);
448 	int error;
449 
450 	ifsq_serialize_hw(ifsq);
451 	error = stf_output_serialized(ifp, m, dst, rt);
452 	ifsq_deserialize_hw(ifsq);
453 
454 	return error;
455 }
456 
457 /*
458  * Parameters:
459  *	inifp:	incoming interface
460  */
461 static int
462 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
463 {
464 	struct in_ifaddr_container *iac;
465 
466 	/*
467 	 * reject packets with the following address:
468 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
469 	 */
470 	if (IN_MULTICAST(ntohl(in->s_addr)))
471 		return -1;
472 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
473 	case 0: case 127: case 255:
474 		return -1;
475 	}
476 
477 	/*
478 	 * reject packets with broadcast
479 	 */
480 	TAILQ_FOREACH(iac, &in_ifaddrheads[mycpuid], ia_link) {
481 		struct in_ifaddr *ia4 = iac->ia;
482 
483 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
484 			continue;
485 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
486 			return -1;
487 	}
488 
489 	/*
490 	 * perform ingress filter
491 	 */
492 	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
493 		struct sockaddr_in sin;
494 		struct rtentry *rt;
495 
496 		bzero(&sin, sizeof(sin));
497 		sin.sin_family = AF_INET;
498 		sin.sin_len = sizeof(struct sockaddr_in);
499 		sin.sin_addr = *in;
500 		rt = rtpurelookup((struct sockaddr *)&sin);
501 		if (!rt || rt->rt_ifp != inifp) {
502 #if 0
503 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
504 			    "due to ingress filter\n", if_name(&sc->sc_if),
505 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
506 #endif
507 			if (rt)
508 				rtfree(rt);
509 			return -1;
510 		}
511 		rtfree(rt);
512 	}
513 
514 	return 0;
515 }
516 
517 /*
518  * Parameters:
519  *	inifp:	incoming interface
520  */
521 static int
522 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
523 {
524 	/*
525 	 * check 6to4 addresses
526 	 */
527 	if (IN6_IS_ADDR_6TO4(in6))
528 		return stf_checkaddr4(sc, GET_V4(in6), inifp);
529 
530 	/*
531 	 * reject anything that look suspicious.  the test is implemented
532 	 * in ip6_input too, but we check here as well to
533 	 * (1) reject bad packets earlier, and
534 	 * (2) to be safe against future ip6_input change.
535 	 */
536 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
537 		return -1;
538 
539 	return 0;
540 }
541 
542 int
543 in_stf_input(struct mbuf **mp, int *offp, int proto)
544 {
545 	struct mbuf *m;
546 	struct stf_softc *sc;
547 	struct ip *ip;
548 	struct ip6_hdr *ip6;
549 	u_int8_t otos, itos;
550 	struct ifnet *ifp;
551 	int off;
552 	static const uint32_t af = AF_INET6;
553 
554 	m = *mp;
555 	off = *offp;
556 
557 	if (proto != IPPROTO_IPV6) {
558 		m_freem(m);
559 		return(IPPROTO_DONE);
560 	}
561 
562 	ip = mtod(m, struct ip *);
563 
564 	sc = (struct stf_softc *)encap_getarg(m);
565 
566 	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
567 		m_freem(m);
568 		return(IPPROTO_DONE);
569 	}
570 
571 	ifp = &sc->sc_if;
572 
573 	/*
574 	 * perform sanity check against outer src/dst.
575 	 * for source, perform ingress filter as well.
576 	 */
577 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
578 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
579 		m_freem(m);
580 		return(IPPROTO_DONE);
581 	}
582 
583 	otos = ip->ip_tos;
584 	m_adj(m, off);
585 
586 	if (m->m_len < sizeof(*ip6)) {
587 		m = m_pullup(m, sizeof(*ip6));
588 		if (!m)
589 			return(IPPROTO_DONE);
590 	}
591 	ip6 = mtod(m, struct ip6_hdr *);
592 
593 	/*
594 	 * perform sanity check against inner src/dst.
595 	 * for source, perform ingress filter as well.
596 	 */
597 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
598 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
599 		m_freem(m);
600 		return(IPPROTO_DONE);
601 	}
602 
603 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
604 	if ((ifp->if_flags & IFF_LINK1) != 0)
605 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
606 	else
607 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
608 	ip6->ip6_flow &= ~htonl(0xff << 20);
609 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
610 
611 	m->m_pkthdr.rcvif = ifp;
612 
613 	if (ifp->if_bpf) {
614 		bpf_gettoken();
615 		if (ifp->if_bpf)
616 			bpf_ptap(ifp->if_bpf, m, &af, sizeof(af));
617 		bpf_reltoken();
618 	}
619 
620 	/*
621 	 * Put the packet to the network layer input queue according to the
622 	 * specified address family.
623 	 * See net/if_gif.c for possible issues with packet processing
624 	 * reorder due to extra queueing.
625 	 */
626 	IFNET_STAT_INC(ifp, ipackets, 1);
627 	IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
628 	m->m_flags &= ~M_HASH;
629 	netisr_queue(NETISR_IPV6, m);
630 	return(IPPROTO_DONE);
631 }
632 
633 /* ARGSUSED */
634 static void
635 stf_rtrequest(int cmd, struct rtentry *rt)
636 {
637 
638 	if (rt)
639 		rt->rt_rmx.rmx_mtu = IPV6_MMTU;
640 }
641 
642 static int
643 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
644 {
645 	struct ifaddr *ifa;
646 	struct ifreq *ifr;
647 	struct sockaddr_in6 *sin6;
648 	int error;
649 
650 	error = 0;
651 	switch (cmd) {
652 	case SIOCSIFADDR:
653 		ifa = (struct ifaddr *)data;
654 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
655 			error = EAFNOSUPPORT;
656 			break;
657 		}
658 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
659 		if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
660 			ifa->ifa_rtrequest = stf_rtrequest;
661 			ifp->if_flags |= IFF_UP;
662 		} else
663 			error = EINVAL;
664 		break;
665 
666 	case SIOCADDMULTI:
667 	case SIOCDELMULTI:
668 		ifr = (struct ifreq *)data;
669 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
670 			;
671 		else
672 			error = EAFNOSUPPORT;
673 		break;
674 
675 	default:
676 		error = EINVAL;
677 		break;
678 	}
679 
680 	return error;
681 }
682