xref: /freebsd/sys/net/if_stf.c (revision c11e094d96120a2e0e726ed9705ae0ec08db49b6)
1 /*	$FreeBSD$	*/
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/kernel.h>
87 #include <sys/protosw.h>
88 #include <sys/queue.h>
89 #include <machine/cpu.h>
90 
91 #include <sys/malloc.h>
92 
93 #include <net/if.h>
94 #include <net/route.h>
95 #include <net/netisr.h>
96 #include <net/if_types.h>
97 #include <net/if_stf.h>
98 
99 #include <netinet/in.h>
100 #include <netinet/in_systm.h>
101 #include <netinet/ip.h>
102 #include <netinet/ip_var.h>
103 #include <netinet/in_var.h>
104 
105 #include <netinet/ip6.h>
106 #include <netinet6/ip6_var.h>
107 #include <netinet6/in6_var.h>
108 #include <netinet/ip_ecn.h>
109 
110 #include <netinet/ip_encap.h>
111 
112 #include <machine/stdarg.h>
113 
114 #include <net/net_osdep.h>
115 
116 #include <net/bpf.h>
117 
118 #define STFNAME		"stf"
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 	union {
126 		struct route  __sc_ro4;
127 		struct route_in6 __sc_ro6; /* just for safety */
128 	} __sc_ro46;
129 #define sc_ro	__sc_ro46.__sc_ro4
130 	const struct encaptab *encap_cookie;
131 	LIST_ENTRY(stf_softc) sc_list;	/* all stf's are linked */
132 };
133 
134 static LIST_HEAD(, stf_softc) stf_softc_list;
135 
136 static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
137 static int ip_stf_ttl = 40;
138 
139 extern  struct domain inetdomain;
140 struct protosw in_stf_protosw =
141 { SOCK_RAW,	&inetdomain,	IPPROTO_IPV6,	PR_ATOMIC|PR_ADDR,
142   in_stf_input,	(pr_output_t*)rip_output, 0,	rip_ctloutput,
143   0,
144   0,            0,              0,              0,
145   &rip_usrreqs
146 };
147 
148 static int stfmodevent(module_t, int, void *);
149 static int stf_encapcheck(const struct mbuf *, int, int, void *);
150 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
151 static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
152 	struct rtentry *);
153 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
154 	struct ifnet *);
155 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
156 	struct ifnet *);
157 static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
158 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
159 
160 int	stf_clone_create(struct if_clone *, int);
161 int	stf_clone_destroy(struct ifnet *);
162 
163 /* only one clone is currently allowed */
164 struct if_clone stf_cloner =
165     IF_CLONE_INITIALIZER(STFNAME, stf_clone_create, stf_clone_destroy, 0);
166 
167 int
168 stf_clone_create(ifc, unit)
169 	struct if_clone *ifc;
170 	int unit;
171 {
172 	struct stf_softc *sc;
173 
174 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
175 	sc->sc_if.if_name = STFNAME;
176 	sc->sc_if.if_unit = unit;
177 
178 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
179 	    stf_encapcheck, &in_stf_protosw, sc);
180 	if (sc->encap_cookie == NULL) {
181 		printf("%s: attach failed\n", if_name(&sc->sc_if));
182 		free(sc, M_STF);
183 		return (ENOMEM);
184 	}
185 
186 	sc->sc_if.if_mtu    = IPV6_MMTU;
187 	sc->sc_if.if_ioctl  = stf_ioctl;
188 	sc->sc_if.if_output = stf_output;
189 	sc->sc_if.if_type   = IFT_STF;
190 	sc->sc_if.if_snd.ifq_maxlen = IFQ_MAXLEN;
191 	if_attach(&sc->sc_if);
192 	bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int));
193 	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
194 	return (0);
195 }
196 
197 int
198 stf_clone_destroy(ifp)
199 	struct ifnet *ifp;
200 {
201 	int err;
202 	struct stf_softc *sc = (void *) ifp;
203 
204 	LIST_REMOVE(sc, sc_list);
205 	err = encap_detach(sc->encap_cookie);
206 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
207 	bpfdetach(ifp);
208 	if_detach(ifp);
209 
210 	free(sc, M_STF);
211 	return (0);
212 }
213 
214 static int
215 stfmodevent(mod, type, data)
216 	module_t mod;
217 	int type;
218 	void *data;
219 {
220 
221 	switch (type) {
222 	case MOD_LOAD:
223 		LIST_INIT(&stf_softc_list);
224 		if_clone_attach(&stf_cloner);
225 
226 		break;
227 	case MOD_UNLOAD:
228 		if_clone_detach(&stf_cloner);
229 
230 		while (!LIST_EMPTY(&stf_softc_list))
231 			stf_clone_destroy(&LIST_FIRST(&stf_softc_list)->sc_if);
232 		break;
233 	}
234 
235 	return (0);
236 }
237 
238 static moduledata_t stf_mod = {
239 	"if_stf",
240 	stfmodevent,
241 	0
242 };
243 
244 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
245 
246 static int
247 stf_encapcheck(m, off, proto, arg)
248 	const struct mbuf *m;
249 	int off;
250 	int proto;
251 	void *arg;
252 {
253 	struct ip ip;
254 	struct in6_ifaddr *ia6;
255 	struct stf_softc *sc;
256 	struct in_addr a, b;
257 
258 	sc = (struct stf_softc *)arg;
259 	if (sc == NULL)
260 		return 0;
261 
262 	if ((sc->sc_if.if_flags & IFF_UP) == 0)
263 		return 0;
264 
265 	/* IFF_LINK0 means "no decapsulation" */
266 	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
267 		return 0;
268 
269 	if (proto != IPPROTO_IPV6)
270 		return 0;
271 
272 	/* LINTED const cast */
273 	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
274 
275 	if (ip.ip_v != 4)
276 		return 0;
277 
278 	ia6 = stf_getsrcifa6(&sc->sc_if);
279 	if (ia6 == NULL)
280 		return 0;
281 
282 	/*
283 	 * check if IPv4 dst matches the IPv4 address derived from the
284 	 * local 6to4 address.
285 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
286 	 */
287 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
288 	    sizeof(ip.ip_dst)) != 0)
289 		return 0;
290 
291 	/*
292 	 * check if IPv4 src matches the IPv4 address derived from the
293 	 * local 6to4 address masked by prefixmask.
294 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
295 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
296 	 */
297 	bzero(&a, sizeof(a));
298 	a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
299 	a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
300 	b = ip.ip_src;
301 	b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
302 	if (a.s_addr != b.s_addr)
303 		return 0;
304 
305 	/* stf interface makes single side match only */
306 	return 32;
307 }
308 
309 static struct in6_ifaddr *
310 stf_getsrcifa6(ifp)
311 	struct ifnet *ifp;
312 {
313 	struct ifaddr *ia;
314 	struct in_ifaddr *ia4;
315 	struct sockaddr_in6 *sin6;
316 	struct in_addr in;
317 
318 	for (ia = TAILQ_FIRST(&ifp->if_addrlist);
319 	     ia;
320 	     ia = TAILQ_NEXT(ia, ifa_list))
321 	{
322 		if (ia->ifa_addr == NULL)
323 			continue;
324 		if (ia->ifa_addr->sa_family != AF_INET6)
325 			continue;
326 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
327 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
328 			continue;
329 
330 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
331 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
332 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
333 				break;
334 		if (ia4 == NULL)
335 			continue;
336 
337 		return (struct in6_ifaddr *)ia;
338 	}
339 
340 	return NULL;
341 }
342 
343 static int
344 stf_output(ifp, m, dst, rt)
345 	struct ifnet *ifp;
346 	struct mbuf *m;
347 	struct sockaddr *dst;
348 	struct rtentry *rt;
349 {
350 	struct stf_softc *sc;
351 	struct sockaddr_in6 *dst6;
352 	struct in_addr *in4;
353 	struct sockaddr_in *dst4;
354 	u_int8_t tos;
355 	struct ip *ip;
356 	struct ip6_hdr *ip6;
357 	struct in6_ifaddr *ia6;
358 
359 	sc = (struct stf_softc*)ifp;
360 	dst6 = (struct sockaddr_in6 *)dst;
361 
362 	/* just in case */
363 	if ((ifp->if_flags & IFF_UP) == 0) {
364 		m_freem(m);
365 		return ENETDOWN;
366 	}
367 
368 	/*
369 	 * If we don't have an ip4 address that match my inner ip6 address,
370 	 * we shouldn't generate output.  Without this check, we'll end up
371 	 * using wrong IPv4 source.
372 	 */
373 	ia6 = stf_getsrcifa6(ifp);
374 	if (ia6 == NULL) {
375 		m_freem(m);
376 		return ENETDOWN;
377 	}
378 
379 	if (m->m_len < sizeof(*ip6)) {
380 		m = m_pullup(m, sizeof(*ip6));
381 		if (!m)
382 			return ENOBUFS;
383 	}
384 	ip6 = mtod(m, struct ip6_hdr *);
385 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
386 
387 	/*
388 	 * Pickup the right outer dst addr from the list of candidates.
389 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
390 	 */
391 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
392 		in4 = GET_V4(&ip6->ip6_dst);
393 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
394 		in4 = GET_V4(&dst6->sin6_addr);
395 	else {
396 		m_freem(m);
397 		return ENETUNREACH;
398 	}
399 
400 #if NBPFILTER > 0
401 	if (ifp->if_bpf) {
402 		/*
403 		 * We need to prepend the address family as
404 		 * a four byte field.  Cons up a dummy header
405 		 * to pacify bpf.  This is safe because bpf
406 		 * will only read from the mbuf (i.e., it won't
407 		 * try to free it or keep a pointer a to it).
408 		 */
409 		struct mbuf m0;
410 		u_int32_t af = AF_INET6;
411 
412 		m0.m_next = m;
413 		m0.m_len = 4;
414 		m0.m_data = (char *)&af;
415 
416 #ifdef HAVE_OLD_BPF
417 		bpf_mtap(ifp, &m0);
418 #else
419 		bpf_mtap(ifp->if_bpf, &m0);
420 #endif
421 	}
422 #endif /*NBPFILTER > 0*/
423 
424 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
425 	if (m && m->m_len < sizeof(struct ip))
426 		m = m_pullup(m, sizeof(struct ip));
427 	if (m == NULL)
428 		return ENOBUFS;
429 	ip = mtod(m, struct ip *);
430 
431 	bzero(ip, 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_stf_ttl;
438 	ip->ip_len = m->m_pkthdr.len;	/*host order*/
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 			return ENETUNREACH;
462 		}
463 	}
464 
465 	return ip_output(m, NULL, &sc->sc_ro, 0, NULL);
466 }
467 
468 static int
469 stf_checkaddr4(sc, in, inifp)
470 	struct stf_softc *sc;
471 	struct in_addr *in;
472 	struct ifnet *inifp;	/* incoming interface */
473 {
474 	struct in_ifaddr *ia4;
475 
476 	/*
477 	 * reject packets with the following address:
478 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
479 	 */
480 	if (IN_MULTICAST(ntohl(in->s_addr)))
481 		return -1;
482 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
483 	case 0: case 127: case 255:
484 		return -1;
485 	}
486 
487 	/*
488 	 * reject packets with broadcast
489 	 */
490 	for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
491 	     ia4;
492 	     ia4 = TAILQ_NEXT(ia4, ia_link))
493 	{
494 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
495 			continue;
496 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
497 			return -1;
498 	}
499 
500 	/*
501 	 * perform ingress filter
502 	 */
503 	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
504 		struct sockaddr_in sin;
505 		struct rtentry *rt;
506 
507 		bzero(&sin, sizeof(sin));
508 		sin.sin_family = AF_INET;
509 		sin.sin_len = sizeof(struct sockaddr_in);
510 		sin.sin_addr = *in;
511 		rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
512 		if (!rt || rt->rt_ifp != inifp) {
513 #if 0
514 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
515 			    "due to ingress filter\n", if_name(&sc->sc_if),
516 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
517 #endif
518 			if (rt)
519 				rtfree(rt);
520 			return -1;
521 		}
522 		rtfree(rt);
523 	}
524 
525 	return 0;
526 }
527 
528 static int
529 stf_checkaddr6(sc, in6, inifp)
530 	struct stf_softc *sc;
531 	struct in6_addr *in6;
532 	struct ifnet *inifp;	/* incoming interface */
533 {
534 	/*
535 	 * check 6to4 addresses
536 	 */
537 	if (IN6_IS_ADDR_6TO4(in6))
538 		return stf_checkaddr4(sc, GET_V4(in6), inifp);
539 
540 	/*
541 	 * reject anything that look suspicious.  the test is implemented
542 	 * in ip6_input too, but we check here as well to
543 	 * (1) reject bad packets earlier, and
544 	 * (2) to be safe against future ip6_input change.
545 	 */
546 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
547 		return -1;
548 
549 	return 0;
550 }
551 
552 void
553 in_stf_input(m, off)
554 	struct mbuf *m;
555 	int off;
556 {
557 	int proto;
558 	struct stf_softc *sc;
559 	struct ip *ip;
560 	struct ip6_hdr *ip6;
561 	u_int8_t otos, itos;
562 	int len, isr;
563 	struct ifqueue *ifq = NULL;
564 	struct ifnet *ifp;
565 
566 	proto = mtod(m, struct ip *)->ip_p;
567 
568 	if (proto != IPPROTO_IPV6) {
569 		m_freem(m);
570 		return;
571 	}
572 
573 	ip = mtod(m, struct ip *);
574 
575 	sc = (struct stf_softc *)encap_getarg(m);
576 
577 	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
578 		m_freem(m);
579 		return;
580 	}
581 
582 	ifp = &sc->sc_if;
583 
584 	/*
585 	 * perform sanity check against outer src/dst.
586 	 * for source, perform ingress filter as well.
587 	 */
588 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
589 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
590 		m_freem(m);
591 		return;
592 	}
593 
594 	otos = ip->ip_tos;
595 	m_adj(m, off);
596 
597 	if (m->m_len < sizeof(*ip6)) {
598 		m = m_pullup(m, sizeof(*ip6));
599 		if (!m)
600 			return;
601 	}
602 	ip6 = mtod(m, struct ip6_hdr *);
603 
604 	/*
605 	 * perform sanity check against inner src/dst.
606 	 * for source, perform ingress filter as well.
607 	 */
608 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
609 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
610 		m_freem(m);
611 		return;
612 	}
613 
614 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
615 	if ((ifp->if_flags & IFF_LINK1) != 0)
616 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
617 	else
618 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
619 	ip6->ip6_flow &= ~htonl(0xff << 20);
620 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
621 
622 	m->m_pkthdr.rcvif = ifp;
623 
624 	if (ifp->if_bpf) {
625 		/*
626 		 * We need to prepend the address family as
627 		 * a four byte field.  Cons up a dummy header
628 		 * to pacify bpf.  This is safe because bpf
629 		 * will only read from the mbuf (i.e., it won't
630 		 * try to free it or keep a pointer a to it).
631 		 */
632 		struct mbuf m0;
633 		u_int32_t af = AF_INET6;
634 
635 		m0.m_next = m;
636 		m0.m_len = 4;
637 		m0.m_data = (char *)&af;
638 
639 #ifdef HAVE_OLD_BPF
640 		bpf_mtap(ifp, &m0);
641 #else
642 		bpf_mtap(ifp->if_bpf, &m0);
643 #endif
644 	}
645 
646 	/*
647 	 * Put the packet to the network layer input queue according to the
648 	 * specified address family.
649 	 * See net/if_gif.c for possible issues with packet processing
650 	 * reorder due to extra queueing.
651 	 */
652 	ifq = &ip6intrq;
653 	isr = NETISR_IPV6;
654 
655 	len = m->m_pkthdr.len;
656 	if (! IF_HANDOFF(ifq, m, NULL))
657 		return;
658 	schednetisr(isr);
659 	ifp->if_ipackets++;
660 	ifp->if_ibytes += len;
661 }
662 
663 /* ARGSUSED */
664 static void
665 stf_rtrequest(cmd, rt, info)
666 	int cmd;
667 	struct rtentry *rt;
668 	struct rt_addrinfo *info;
669 {
670 
671 	if (rt)
672 		rt->rt_rmx.rmx_mtu = IPV6_MMTU;
673 }
674 
675 static int
676 stf_ioctl(ifp, cmd, data)
677 	struct ifnet *ifp;
678 	u_long cmd;
679 	caddr_t data;
680 {
681 	struct ifaddr *ifa;
682 	struct ifreq *ifr;
683 	struct sockaddr_in6 *sin6;
684 	int error;
685 
686 	error = 0;
687 	switch (cmd) {
688 	case SIOCSIFADDR:
689 		ifa = (struct ifaddr *)data;
690 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
691 			error = EAFNOSUPPORT;
692 			break;
693 		}
694 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
695 		if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
696 			ifa->ifa_rtrequest = stf_rtrequest;
697 			ifp->if_flags |= IFF_UP;
698 		} else
699 			error = EINVAL;
700 		break;
701 
702 	case SIOCADDMULTI:
703 	case SIOCDELMULTI:
704 		ifr = (struct ifreq *)data;
705 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
706 			;
707 		else
708 			error = EAFNOSUPPORT;
709 		break;
710 
711 	default:
712 		error = EINVAL;
713 		break;
714 	}
715 
716 	return error;
717 }
718