xref: /freebsd/sys/net/if_stf.c (revision 1e413cf93298b5b97441a21d9a50fdcd0ee9945e)
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 #include "opt_mac.h"
80 
81 #include <sys/param.h>
82 #include <sys/systm.h>
83 #include <sys/socket.h>
84 #include <sys/sockio.h>
85 #include <sys/mbuf.h>
86 #include <sys/errno.h>
87 #include <sys/kernel.h>
88 #include <sys/module.h>
89 #include <sys/protosw.h>
90 #include <sys/queue.h>
91 #include <machine/cpu.h>
92 
93 #include <sys/malloc.h>
94 
95 #include <net/if.h>
96 #include <net/if_clone.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_var.h>
111 #include <netinet/ip_ecn.h>
112 
113 #include <netinet/ip_encap.h>
114 
115 #include <machine/stdarg.h>
116 
117 #include <net/bpf.h>
118 
119 #include <security/mac/mac_framework.h>
120 
121 #define STFNAME		"stf"
122 #define STFUNIT		0
123 
124 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
125 
126 /*
127  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
128  * struct in_addr *; use bcopy() instead.
129  */
130 #define GET_V4(x)	((caddr_t)(&(x)->s6_addr16[1]))
131 
132 struct stf_softc {
133 	struct ifnet	*sc_ifp;
134 	union {
135 		struct route  __sc_ro4;
136 		struct route_in6 __sc_ro6; /* just for safety */
137 	} __sc_ro46;
138 #define sc_ro	__sc_ro46.__sc_ro4
139 	const struct encaptab *encap_cookie;
140 };
141 #define STF2IFP(sc)	((sc)->sc_ifp)
142 
143 /*
144  * XXXRW: Note that mutable fields in the softc are not currently locked:
145  * in particular, sc_ro needs to be protected from concurrent entrance
146  * of stf_output().
147  */
148 static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
149 static const int ip_stf_ttl = 40;
150 
151 extern  struct domain inetdomain;
152 struct protosw in_stf_protosw = {
153 	.pr_type =		SOCK_RAW,
154 	.pr_domain =		&inetdomain,
155 	.pr_protocol =		IPPROTO_IPV6,
156 	.pr_flags =		PR_ATOMIC|PR_ADDR,
157 	.pr_input =		in_stf_input,
158 	.pr_output =		(pr_output_t *)rip_output,
159 	.pr_ctloutput =		rip_ctloutput,
160 	.pr_usrreqs =		&rip_usrreqs
161 };
162 
163 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
164 
165 static int stfmodevent(module_t, int, void *);
166 static int stf_encapcheck(const struct mbuf *, int, int, void *);
167 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
168 static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
169 	struct rtentry *);
170 static int isrfc1918addr(struct in_addr *);
171 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
172 	struct ifnet *);
173 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
174 	struct ifnet *);
175 static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
176 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
177 
178 static int stf_clone_match(struct if_clone *, const char *);
179 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
180 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
181 struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0,
182     NULL, stf_clone_match, stf_clone_create, stf_clone_destroy);
183 
184 static int
185 stf_clone_match(struct if_clone *ifc, const char *name)
186 {
187 	int i;
188 
189 	for(i = 0; stfnames[i] != NULL; i++) {
190 		if (strcmp(stfnames[i], name) == 0)
191 			return (1);
192 	}
193 
194 	return (0);
195 }
196 
197 static int
198 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
199 {
200 	int err, unit;
201 	struct stf_softc *sc;
202 	struct ifnet *ifp;
203 
204 	/*
205 	 * We can only have one unit, but since unit allocation is
206 	 * already locked, we use it to keep from allocating extra
207 	 * interfaces.
208 	 */
209 	unit = STFUNIT;
210 	err = ifc_alloc_unit(ifc, &unit);
211 	if (err != 0)
212 		return (err);
213 
214 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
215 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
216 	if (ifp == NULL) {
217 		free(sc, M_STF);
218 		ifc_free_unit(ifc, unit);
219 		return (ENOSPC);
220 	}
221 	ifp->if_softc = sc;
222 
223 	/*
224 	 * Set the name manually rather then using if_initname because
225 	 * we don't conform to the default naming convention for interfaces.
226 	 */
227 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
228 	ifp->if_dname = ifc->ifc_name;
229 	ifp->if_dunit = IF_DUNIT_NONE;
230 
231 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
232 	    stf_encapcheck, &in_stf_protosw, sc);
233 	if (sc->encap_cookie == NULL) {
234 		if_printf(ifp, "attach failed\n");
235 		free(sc, M_STF);
236 		ifc_free_unit(ifc, unit);
237 		return (ENOMEM);
238 	}
239 
240 	ifp->if_mtu    = IPV6_MMTU;
241 	ifp->if_ioctl  = stf_ioctl;
242 	ifp->if_output = stf_output;
243 	ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
244 	if_attach(ifp);
245 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
246 	return (0);
247 }
248 
249 static int
250 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
251 {
252 	struct stf_softc *sc = ifp->if_softc;
253 	int err;
254 
255 	err = encap_detach(sc->encap_cookie);
256 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
257 	bpfdetach(ifp);
258 	if_detach(ifp);
259 	if_free(ifp);
260 
261 	free(sc, M_STF);
262 	ifc_free_unit(ifc, STFUNIT);
263 
264 	return (0);
265 }
266 
267 static int
268 stfmodevent(mod, type, data)
269 	module_t mod;
270 	int type;
271 	void *data;
272 {
273 
274 	switch (type) {
275 	case MOD_LOAD:
276 		if_clone_attach(&stf_cloner);
277 		break;
278 	case MOD_UNLOAD:
279 		if_clone_detach(&stf_cloner);
280 		break;
281 	default:
282 		return (EOPNOTSUPP);
283 	}
284 
285 	return (0);
286 }
287 
288 static moduledata_t stf_mod = {
289 	"if_stf",
290 	stfmodevent,
291 	0
292 };
293 
294 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
295 
296 static int
297 stf_encapcheck(m, off, proto, arg)
298 	const struct mbuf *m;
299 	int off;
300 	int proto;
301 	void *arg;
302 {
303 	struct ip ip;
304 	struct in6_ifaddr *ia6;
305 	struct stf_softc *sc;
306 	struct in_addr a, b, mask;
307 
308 	sc = (struct stf_softc *)arg;
309 	if (sc == NULL)
310 		return 0;
311 
312 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
313 		return 0;
314 
315 	/* IFF_LINK0 means "no decapsulation" */
316 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
317 		return 0;
318 
319 	if (proto != IPPROTO_IPV6)
320 		return 0;
321 
322 	/* LINTED const cast */
323 	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
324 
325 	if (ip.ip_v != 4)
326 		return 0;
327 
328 	ia6 = stf_getsrcifa6(STF2IFP(sc));
329 	if (ia6 == NULL)
330 		return 0;
331 
332 	/*
333 	 * check if IPv4 dst matches the IPv4 address derived from the
334 	 * local 6to4 address.
335 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
336 	 */
337 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
338 	    sizeof(ip.ip_dst)) != 0)
339 		return 0;
340 
341 	/*
342 	 * check if IPv4 src matches the IPv4 address derived from the
343 	 * local 6to4 address masked by prefixmask.
344 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
345 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
346 	 */
347 	bzero(&a, sizeof(a));
348 	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
349 	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
350 	a.s_addr &= mask.s_addr;
351 	b = ip.ip_src;
352 	b.s_addr &= mask.s_addr;
353 	if (a.s_addr != b.s_addr)
354 		return 0;
355 
356 	/* stf interface makes single side match only */
357 	return 32;
358 }
359 
360 static struct in6_ifaddr *
361 stf_getsrcifa6(ifp)
362 	struct ifnet *ifp;
363 {
364 	struct ifaddr *ia;
365 	struct in_ifaddr *ia4;
366 	struct sockaddr_in6 *sin6;
367 	struct in_addr in;
368 
369 	TAILQ_FOREACH(ia, &ifp->if_addrlist, ifa_list) {
370 		if (ia->ifa_addr->sa_family != AF_INET6)
371 			continue;
372 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
373 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
374 			continue;
375 
376 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
377 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
378 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
379 				break;
380 		if (ia4 == NULL)
381 			continue;
382 
383 		return (struct in6_ifaddr *)ia;
384 	}
385 
386 	return NULL;
387 }
388 
389 static int
390 stf_output(ifp, m, dst, rt)
391 	struct ifnet *ifp;
392 	struct mbuf *m;
393 	struct sockaddr *dst;
394 	struct rtentry *rt;
395 {
396 	struct stf_softc *sc;
397 	struct sockaddr_in6 *dst6;
398 	struct in_addr in4;
399 	caddr_t ptr;
400 	struct sockaddr_in *dst4;
401 	u_int8_t tos;
402 	struct ip *ip;
403 	struct ip6_hdr *ip6;
404 	struct in6_ifaddr *ia6;
405 	u_int32_t af;
406 #ifdef MAC
407 	int error;
408 
409 	error = mac_ifnet_check_transmit(ifp, m);
410 	if (error) {
411 		m_freem(m);
412 		return (error);
413 	}
414 #endif
415 
416 	sc = ifp->if_softc;
417 	dst6 = (struct sockaddr_in6 *)dst;
418 
419 	/* just in case */
420 	if ((ifp->if_flags & IFF_UP) == 0) {
421 		m_freem(m);
422 		ifp->if_oerrors++;
423 		return ENETDOWN;
424 	}
425 
426 	/*
427 	 * If we don't have an ip4 address that match my inner ip6 address,
428 	 * we shouldn't generate output.  Without this check, we'll end up
429 	 * using wrong IPv4 source.
430 	 */
431 	ia6 = stf_getsrcifa6(ifp);
432 	if (ia6 == NULL) {
433 		m_freem(m);
434 		ifp->if_oerrors++;
435 		return ENETDOWN;
436 	}
437 
438 	if (m->m_len < sizeof(*ip6)) {
439 		m = m_pullup(m, sizeof(*ip6));
440 		if (!m) {
441 			ifp->if_oerrors++;
442 			return ENOBUFS;
443 		}
444 	}
445 	ip6 = mtod(m, struct ip6_hdr *);
446 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
447 
448 	/*
449 	 * BPF writes need to be handled specially.
450 	 * This is a null operation, nothing here checks dst->sa_family.
451 	 */
452 	if (dst->sa_family == AF_UNSPEC) {
453 		bcopy(dst->sa_data, &af, sizeof(af));
454 		dst->sa_family = af;
455 	}
456 
457 	/*
458 	 * Pickup the right outer dst addr from the list of candidates.
459 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
460 	 */
461 	ptr = NULL;
462 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
463 		ptr = GET_V4(&ip6->ip6_dst);
464 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
465 		ptr = GET_V4(&dst6->sin6_addr);
466 	else {
467 		m_freem(m);
468 		ifp->if_oerrors++;
469 		return ENETUNREACH;
470 	}
471 	bcopy(ptr, &in4, sizeof(in4));
472 
473 	if (bpf_peers_present(ifp->if_bpf)) {
474 		/*
475 		 * We need to prepend the address family as
476 		 * a four byte field.  Cons up a dummy header
477 		 * to pacify bpf.  This is safe because bpf
478 		 * will only read from the mbuf (i.e., it won't
479 		 * try to free it or keep a pointer a to it).
480 		 */
481 		af = AF_INET6;
482 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
483 	}
484 
485 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
486 	if (m && m->m_len < sizeof(struct ip))
487 		m = m_pullup(m, sizeof(struct ip));
488 	if (m == NULL) {
489 		ifp->if_oerrors++;
490 		return ENOBUFS;
491 	}
492 	ip = mtod(m, struct ip *);
493 
494 	bzero(ip, sizeof(*ip));
495 
496 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
497 	    &ip->ip_src, sizeof(ip->ip_src));
498 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
499 	ip->ip_p = IPPROTO_IPV6;
500 	ip->ip_ttl = ip_stf_ttl;
501 	ip->ip_len = m->m_pkthdr.len;	/*host order*/
502 	if (ifp->if_flags & IFF_LINK1)
503 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
504 	else
505 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
506 
507 	/*
508 	 * XXXRW: Locking of sc_ro required.
509 	 */
510 	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
511 	if (dst4->sin_family != AF_INET ||
512 	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
513 		/* cache route doesn't match */
514 		dst4->sin_family = AF_INET;
515 		dst4->sin_len = sizeof(struct sockaddr_in);
516 		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
517 		if (sc->sc_ro.ro_rt) {
518 			RTFREE(sc->sc_ro.ro_rt);
519 			sc->sc_ro.ro_rt = NULL;
520 		}
521 	}
522 
523 	if (sc->sc_ro.ro_rt == NULL) {
524 		rtalloc(&sc->sc_ro);
525 		if (sc->sc_ro.ro_rt == NULL) {
526 			m_freem(m);
527 			ifp->if_oerrors++;
528 			return ENETUNREACH;
529 		}
530 	}
531 
532 	ifp->if_opackets++;
533 	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
534 }
535 
536 static int
537 isrfc1918addr(in)
538 	struct in_addr *in;
539 {
540 	/*
541 	 * returns 1 if private address range:
542 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
543 	 */
544 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
545 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
546 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
547 		return 1;
548 
549 	return 0;
550 }
551 
552 static int
553 stf_checkaddr4(sc, in, inifp)
554 	struct stf_softc *sc;
555 	struct in_addr *in;
556 	struct ifnet *inifp;	/* incoming interface */
557 {
558 	struct in_ifaddr *ia4;
559 
560 	/*
561 	 * reject packets with the following address:
562 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
563 	 */
564 	if (IN_MULTICAST(ntohl(in->s_addr)))
565 		return -1;
566 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
567 	case 0: case 127: case 255:
568 		return -1;
569 	}
570 
571 	/*
572 	 * reject packets with private address range.
573 	 * (requirement from RFC3056 section 2 1st paragraph)
574 	 */
575 	if (isrfc1918addr(in))
576 		return -1;
577 
578 	/*
579 	 * reject packets with broadcast
580 	 */
581 	for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
582 	     ia4;
583 	     ia4 = TAILQ_NEXT(ia4, ia_link))
584 	{
585 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
586 			continue;
587 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
588 			return -1;
589 	}
590 
591 	/*
592 	 * perform ingress filter
593 	 */
594 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
595 		struct sockaddr_in sin;
596 		struct rtentry *rt;
597 
598 		bzero(&sin, sizeof(sin));
599 		sin.sin_family = AF_INET;
600 		sin.sin_len = sizeof(struct sockaddr_in);
601 		sin.sin_addr = *in;
602 		rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
603 		if (!rt || rt->rt_ifp != inifp) {
604 #if 0
605 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
606 			    "due to ingress filter\n", if_name(STF2IFP(sc)),
607 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
608 #endif
609 			if (rt)
610 				RTFREE_LOCKED(rt);
611 			return -1;
612 		}
613 		RTFREE_LOCKED(rt);
614 	}
615 
616 	return 0;
617 }
618 
619 static int
620 stf_checkaddr6(sc, in6, inifp)
621 	struct stf_softc *sc;
622 	struct in6_addr *in6;
623 	struct ifnet *inifp;	/* incoming interface */
624 {
625 	/*
626 	 * check 6to4 addresses
627 	 */
628 	if (IN6_IS_ADDR_6TO4(in6)) {
629 		struct in_addr in4;
630 		bcopy(GET_V4(in6), &in4, sizeof(in4));
631 		return stf_checkaddr4(sc, &in4, inifp);
632 	}
633 
634 	/*
635 	 * reject anything that look suspicious.  the test is implemented
636 	 * in ip6_input too, but we check here as well to
637 	 * (1) reject bad packets earlier, and
638 	 * (2) to be safe against future ip6_input change.
639 	 */
640 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
641 		return -1;
642 
643 	return 0;
644 }
645 
646 void
647 in_stf_input(m, off)
648 	struct mbuf *m;
649 	int off;
650 {
651 	int proto;
652 	struct stf_softc *sc;
653 	struct ip *ip;
654 	struct ip6_hdr *ip6;
655 	u_int8_t otos, itos;
656 	struct ifnet *ifp;
657 
658 	proto = mtod(m, struct ip *)->ip_p;
659 
660 	if (proto != IPPROTO_IPV6) {
661 		m_freem(m);
662 		return;
663 	}
664 
665 	ip = mtod(m, struct ip *);
666 
667 	sc = (struct stf_softc *)encap_getarg(m);
668 
669 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
670 		m_freem(m);
671 		return;
672 	}
673 
674 	ifp = STF2IFP(sc);
675 
676 #ifdef MAC
677 	mac_ifnet_create_mbuf(ifp, m);
678 #endif
679 
680 	/*
681 	 * perform sanity check against outer src/dst.
682 	 * for source, perform ingress filter as well.
683 	 */
684 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
685 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
686 		m_freem(m);
687 		return;
688 	}
689 
690 	otos = ip->ip_tos;
691 	m_adj(m, off);
692 
693 	if (m->m_len < sizeof(*ip6)) {
694 		m = m_pullup(m, sizeof(*ip6));
695 		if (!m)
696 			return;
697 	}
698 	ip6 = mtod(m, struct ip6_hdr *);
699 
700 	/*
701 	 * perform sanity check against inner src/dst.
702 	 * for source, perform ingress filter as well.
703 	 */
704 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
705 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
706 		m_freem(m);
707 		return;
708 	}
709 
710 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
711 	if ((ifp->if_flags & IFF_LINK1) != 0)
712 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
713 	else
714 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
715 	ip6->ip6_flow &= ~htonl(0xff << 20);
716 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
717 
718 	m->m_pkthdr.rcvif = ifp;
719 
720 	if (bpf_peers_present(ifp->if_bpf)) {
721 		/*
722 		 * We need to prepend the address family as
723 		 * a four byte field.  Cons up a dummy header
724 		 * to pacify bpf.  This is safe because bpf
725 		 * will only read from the mbuf (i.e., it won't
726 		 * try to free it or keep a pointer a to it).
727 		 */
728 		u_int32_t af = AF_INET6;
729 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
730 	}
731 
732 	/*
733 	 * Put the packet to the network layer input queue according to the
734 	 * specified address family.
735 	 * See net/if_gif.c for possible issues with packet processing
736 	 * reorder due to extra queueing.
737 	 */
738 	ifp->if_ipackets++;
739 	ifp->if_ibytes += m->m_pkthdr.len;
740 	netisr_dispatch(NETISR_IPV6, m);
741 }
742 
743 /* ARGSUSED */
744 static void
745 stf_rtrequest(cmd, rt, info)
746 	int cmd;
747 	struct rtentry *rt;
748 	struct rt_addrinfo *info;
749 {
750 	RT_LOCK_ASSERT(rt);
751 	rt->rt_rmx.rmx_mtu = IPV6_MMTU;
752 }
753 
754 static int
755 stf_ioctl(ifp, cmd, data)
756 	struct ifnet *ifp;
757 	u_long cmd;
758 	caddr_t data;
759 {
760 	struct ifaddr *ifa;
761 	struct ifreq *ifr;
762 	struct sockaddr_in6 *sin6;
763 	struct in_addr addr;
764 	int error;
765 
766 	error = 0;
767 	switch (cmd) {
768 	case SIOCSIFADDR:
769 		ifa = (struct ifaddr *)data;
770 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
771 			error = EAFNOSUPPORT;
772 			break;
773 		}
774 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
775 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
776 			error = EINVAL;
777 			break;
778 		}
779 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
780 		if (isrfc1918addr(&addr)) {
781 			error = EINVAL;
782 			break;
783 		}
784 
785 		ifa->ifa_rtrequest = stf_rtrequest;
786 		ifp->if_flags |= IFF_UP;
787 		break;
788 
789 	case SIOCADDMULTI:
790 	case SIOCDELMULTI:
791 		ifr = (struct ifreq *)data;
792 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
793 			;
794 		else
795 			error = EAFNOSUPPORT;
796 		break;
797 
798 	default:
799 		error = EINVAL;
800 		break;
801 	}
802 
803 	return error;
804 }
805