xref: /freebsd/sys/net/if_stf.c (revision 271c3a9060f2ee55607ebe146523f888e1db2654)
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/proc.h>
91 #include <sys/queue.h>
92 #include <machine/cpu.h>
93 
94 #include <sys/malloc.h>
95 #include <sys/vimage.h>
96 
97 #include <net/if.h>
98 #include <net/if_clone.h>
99 #include <net/route.h>
100 #include <net/netisr.h>
101 #include <net/if_types.h>
102 #include <net/if_stf.h>
103 
104 #include <netinet/in.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip_var.h>
108 #include <netinet/in_var.h>
109 
110 #include <netinet/ip6.h>
111 #include <netinet6/ip6_var.h>
112 #include <netinet6/in6_var.h>
113 #include <netinet/ip_ecn.h>
114 
115 #include <netinet/ip_encap.h>
116 
117 #include <machine/stdarg.h>
118 
119 #include <net/bpf.h>
120 
121 #include <security/mac/mac_framework.h>
122 
123 #define STFNAME		"stf"
124 #define STFUNIT		0
125 
126 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
127 
128 /*
129  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
130  * struct in_addr *; use bcopy() instead.
131  */
132 #define GET_V4(x)	((caddr_t)(&(x)->s6_addr16[1]))
133 
134 struct stf_softc {
135 	struct ifnet	*sc_ifp;
136 	union {
137 		struct route  __sc_ro4;
138 		struct route_in6 __sc_ro6; /* just for safety */
139 	} __sc_ro46;
140 #define sc_ro	__sc_ro46.__sc_ro4
141 	u_int	sc_fibnum;
142 	const struct encaptab *encap_cookie;
143 };
144 #define STF2IFP(sc)	((sc)->sc_ifp)
145 
146 /*
147  * XXXRW: Note that mutable fields in the softc are not currently locked:
148  * in particular, sc_ro needs to be protected from concurrent entrance
149  * of stf_output().
150  */
151 static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
152 static const int ip_stf_ttl = 40;
153 
154 extern  struct domain inetdomain;
155 struct protosw in_stf_protosw = {
156 	.pr_type =		SOCK_RAW,
157 	.pr_domain =		&inetdomain,
158 	.pr_protocol =		IPPROTO_IPV6,
159 	.pr_flags =		PR_ATOMIC|PR_ADDR,
160 	.pr_input =		in_stf_input,
161 	.pr_output =		(pr_output_t *)rip_output,
162 	.pr_ctloutput =		rip_ctloutput,
163 	.pr_usrreqs =		&rip_usrreqs
164 };
165 
166 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
167 
168 static int stfmodevent(module_t, int, void *);
169 static int stf_encapcheck(const struct mbuf *, int, int, void *);
170 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
171 static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
172 	struct rtentry *);
173 static int isrfc1918addr(struct in_addr *);
174 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
175 	struct ifnet *);
176 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
177 	struct ifnet *);
178 static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
179 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
180 
181 static int stf_clone_match(struct if_clone *, const char *);
182 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
183 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
184 struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0,
185     NULL, stf_clone_match, stf_clone_create, stf_clone_destroy);
186 
187 static int
188 stf_clone_match(struct if_clone *ifc, const char *name)
189 {
190 	int i;
191 
192 	for(i = 0; stfnames[i] != NULL; i++) {
193 		if (strcmp(stfnames[i], name) == 0)
194 			return (1);
195 	}
196 
197 	return (0);
198 }
199 
200 static int
201 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
202 {
203 	int err, unit;
204 	struct stf_softc *sc;
205 	struct ifnet *ifp;
206 
207 	/*
208 	 * We can only have one unit, but since unit allocation is
209 	 * already locked, we use it to keep from allocating extra
210 	 * interfaces.
211 	 */
212 	unit = STFUNIT;
213 	err = ifc_alloc_unit(ifc, &unit);
214 	if (err != 0)
215 		return (err);
216 
217 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
218 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
219 	if (ifp == NULL) {
220 		free(sc, M_STF);
221 		ifc_free_unit(ifc, unit);
222 		return (ENOSPC);
223 	}
224 	ifp->if_softc = sc;
225 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
226 
227 	/*
228 	 * Set the name manually rather then using if_initname because
229 	 * we don't conform to the default naming convention for interfaces.
230 	 */
231 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
232 	ifp->if_dname = ifc->ifc_name;
233 	ifp->if_dunit = IF_DUNIT_NONE;
234 
235 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
236 	    stf_encapcheck, &in_stf_protosw, sc);
237 	if (sc->encap_cookie == NULL) {
238 		if_printf(ifp, "attach failed\n");
239 		free(sc, M_STF);
240 		ifc_free_unit(ifc, unit);
241 		return (ENOMEM);
242 	}
243 
244 	ifp->if_mtu    = IPV6_MMTU;
245 	ifp->if_ioctl  = stf_ioctl;
246 	ifp->if_output = stf_output;
247 	ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
248 	if_attach(ifp);
249 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
250 	return (0);
251 }
252 
253 static int
254 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
255 {
256 	struct stf_softc *sc = ifp->if_softc;
257 	int err;
258 
259 	err = encap_detach(sc->encap_cookie);
260 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
261 	bpfdetach(ifp);
262 	if_detach(ifp);
263 	if_free(ifp);
264 
265 	free(sc, M_STF);
266 	ifc_free_unit(ifc, STFUNIT);
267 
268 	return (0);
269 }
270 
271 static int
272 stfmodevent(mod, type, data)
273 	module_t mod;
274 	int type;
275 	void *data;
276 {
277 
278 	switch (type) {
279 	case MOD_LOAD:
280 		if_clone_attach(&stf_cloner);
281 		break;
282 	case MOD_UNLOAD:
283 		if_clone_detach(&stf_cloner);
284 		break;
285 	default:
286 		return (EOPNOTSUPP);
287 	}
288 
289 	return (0);
290 }
291 
292 static moduledata_t stf_mod = {
293 	"if_stf",
294 	stfmodevent,
295 	0
296 };
297 
298 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
299 
300 static int
301 stf_encapcheck(m, off, proto, arg)
302 	const struct mbuf *m;
303 	int off;
304 	int proto;
305 	void *arg;
306 {
307 	struct ip ip;
308 	struct in6_ifaddr *ia6;
309 	struct stf_softc *sc;
310 	struct in_addr a, b, mask;
311 
312 	sc = (struct stf_softc *)arg;
313 	if (sc == NULL)
314 		return 0;
315 
316 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
317 		return 0;
318 
319 	/* IFF_LINK0 means "no decapsulation" */
320 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
321 		return 0;
322 
323 	if (proto != IPPROTO_IPV6)
324 		return 0;
325 
326 	/* LINTED const cast */
327 	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
328 
329 	if (ip.ip_v != 4)
330 		return 0;
331 
332 	ia6 = stf_getsrcifa6(STF2IFP(sc));
333 	if (ia6 == NULL)
334 		return 0;
335 
336 	/*
337 	 * check if IPv4 dst matches the IPv4 address derived from the
338 	 * local 6to4 address.
339 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
340 	 */
341 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
342 	    sizeof(ip.ip_dst)) != 0)
343 		return 0;
344 
345 	/*
346 	 * check if IPv4 src matches the IPv4 address derived from the
347 	 * local 6to4 address masked by prefixmask.
348 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
349 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
350 	 */
351 	bzero(&a, sizeof(a));
352 	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
353 	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
354 	a.s_addr &= mask.s_addr;
355 	b = ip.ip_src;
356 	b.s_addr &= mask.s_addr;
357 	if (a.s_addr != b.s_addr)
358 		return 0;
359 
360 	/* stf interface makes single side match only */
361 	return 32;
362 }
363 
364 static struct in6_ifaddr *
365 stf_getsrcifa6(ifp)
366 	struct ifnet *ifp;
367 {
368 	struct ifaddr *ia;
369 	struct in_ifaddr *ia4;
370 	struct sockaddr_in6 *sin6;
371 	struct in_addr in;
372 
373 	TAILQ_FOREACH(ia, &ifp->if_addrlist, ifa_list) {
374 		if (ia->ifa_addr->sa_family != AF_INET6)
375 			continue;
376 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
377 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
378 			continue;
379 
380 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
381 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
382 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
383 				break;
384 		if (ia4 == NULL)
385 			continue;
386 
387 		return (struct in6_ifaddr *)ia;
388 	}
389 
390 	return NULL;
391 }
392 
393 static int
394 stf_output(ifp, m, dst, rt)
395 	struct ifnet *ifp;
396 	struct mbuf *m;
397 	struct sockaddr *dst;
398 	struct rtentry *rt;
399 {
400 	struct stf_softc *sc;
401 	struct sockaddr_in6 *dst6;
402 	struct in_addr in4;
403 	caddr_t ptr;
404 	struct sockaddr_in *dst4;
405 	u_int8_t tos;
406 	struct ip *ip;
407 	struct ip6_hdr *ip6;
408 	struct in6_ifaddr *ia6;
409 	u_int32_t af;
410 #ifdef MAC
411 	int error;
412 
413 	error = mac_ifnet_check_transmit(ifp, m);
414 	if (error) {
415 		m_freem(m);
416 		return (error);
417 	}
418 #endif
419 
420 	sc = ifp->if_softc;
421 	dst6 = (struct sockaddr_in6 *)dst;
422 
423 	/* just in case */
424 	if ((ifp->if_flags & IFF_UP) == 0) {
425 		m_freem(m);
426 		ifp->if_oerrors++;
427 		return ENETDOWN;
428 	}
429 
430 	/*
431 	 * If we don't have an ip4 address that match my inner ip6 address,
432 	 * we shouldn't generate output.  Without this check, we'll end up
433 	 * using wrong IPv4 source.
434 	 */
435 	ia6 = stf_getsrcifa6(ifp);
436 	if (ia6 == NULL) {
437 		m_freem(m);
438 		ifp->if_oerrors++;
439 		return ENETDOWN;
440 	}
441 
442 	if (m->m_len < sizeof(*ip6)) {
443 		m = m_pullup(m, sizeof(*ip6));
444 		if (!m) {
445 			ifp->if_oerrors++;
446 			return ENOBUFS;
447 		}
448 	}
449 	ip6 = mtod(m, struct ip6_hdr *);
450 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
451 
452 	/*
453 	 * BPF writes need to be handled specially.
454 	 * This is a null operation, nothing here checks dst->sa_family.
455 	 */
456 	if (dst->sa_family == AF_UNSPEC) {
457 		bcopy(dst->sa_data, &af, sizeof(af));
458 		dst->sa_family = af;
459 	}
460 
461 	/*
462 	 * Pickup the right outer dst addr from the list of candidates.
463 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
464 	 */
465 	ptr = NULL;
466 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
467 		ptr = GET_V4(&ip6->ip6_dst);
468 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
469 		ptr = GET_V4(&dst6->sin6_addr);
470 	else {
471 		m_freem(m);
472 		ifp->if_oerrors++;
473 		return ENETUNREACH;
474 	}
475 	bcopy(ptr, &in4, sizeof(in4));
476 
477 	if (bpf_peers_present(ifp->if_bpf)) {
478 		/*
479 		 * We need to prepend the address family as
480 		 * a four byte field.  Cons up a dummy header
481 		 * to pacify bpf.  This is safe because bpf
482 		 * will only read from the mbuf (i.e., it won't
483 		 * try to free it or keep a pointer a to it).
484 		 */
485 		af = AF_INET6;
486 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
487 	}
488 
489 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
490 	if (m && m->m_len < sizeof(struct ip))
491 		m = m_pullup(m, sizeof(struct ip));
492 	if (m == NULL) {
493 		ifp->if_oerrors++;
494 		return ENOBUFS;
495 	}
496 	ip = mtod(m, struct ip *);
497 
498 	bzero(ip, sizeof(*ip));
499 
500 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
501 	    &ip->ip_src, sizeof(ip->ip_src));
502 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
503 	ip->ip_p = IPPROTO_IPV6;
504 	ip->ip_ttl = ip_stf_ttl;
505 	ip->ip_len = m->m_pkthdr.len;	/*host order*/
506 	if (ifp->if_flags & IFF_LINK1)
507 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
508 	else
509 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
510 
511 	/*
512 	 * XXXRW: Locking of sc_ro required.
513 	 */
514 	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
515 	if (dst4->sin_family != AF_INET ||
516 	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
517 		/* cache route doesn't match */
518 		dst4->sin_family = AF_INET;
519 		dst4->sin_len = sizeof(struct sockaddr_in);
520 		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
521 		if (sc->sc_ro.ro_rt) {
522 			RTFREE(sc->sc_ro.ro_rt);
523 			sc->sc_ro.ro_rt = NULL;
524 		}
525 	}
526 
527 	if (sc->sc_ro.ro_rt == NULL) {
528 		rtalloc_fib(&sc->sc_ro, sc->sc_fibnum);
529 		if (sc->sc_ro.ro_rt == NULL) {
530 			m_freem(m);
531 			ifp->if_oerrors++;
532 			return ENETUNREACH;
533 		}
534 	}
535 
536 	M_SETFIB(m, sc->sc_fibnum);
537 	ifp->if_opackets++;
538 	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
539 }
540 
541 static int
542 isrfc1918addr(in)
543 	struct in_addr *in;
544 {
545 	/*
546 	 * returns 1 if private address range:
547 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
548 	 */
549 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
550 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
551 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
552 		return 1;
553 
554 	return 0;
555 }
556 
557 static int
558 stf_checkaddr4(sc, in, inifp)
559 	struct stf_softc *sc;
560 	struct in_addr *in;
561 	struct ifnet *inifp;	/* incoming interface */
562 {
563 	struct in_ifaddr *ia4;
564 
565 	/*
566 	 * reject packets with the following address:
567 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
568 	 */
569 	if (IN_MULTICAST(ntohl(in->s_addr)))
570 		return -1;
571 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
572 	case 0: case 127: case 255:
573 		return -1;
574 	}
575 
576 	/*
577 	 * reject packets with private address range.
578 	 * (requirement from RFC3056 section 2 1st paragraph)
579 	 */
580 	if (isrfc1918addr(in))
581 		return -1;
582 
583 	/*
584 	 * reject packets with broadcast
585 	 */
586 	for (ia4 = TAILQ_FIRST(&V_in_ifaddrhead);
587 	     ia4;
588 	     ia4 = TAILQ_NEXT(ia4, ia_link))
589 	{
590 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
591 			continue;
592 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
593 			return -1;
594 	}
595 
596 	/*
597 	 * perform ingress filter
598 	 */
599 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
600 		struct sockaddr_in sin;
601 		struct rtentry *rt;
602 
603 		bzero(&sin, sizeof(sin));
604 		sin.sin_family = AF_INET;
605 		sin.sin_len = sizeof(struct sockaddr_in);
606 		sin.sin_addr = *in;
607 		rt = rtalloc1_fib((struct sockaddr *)&sin, 0,
608 		    0UL, sc->sc_fibnum);
609 		if (!rt || rt->rt_ifp != inifp) {
610 #if 0
611 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
612 			    "due to ingress filter\n", if_name(STF2IFP(sc)),
613 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
614 #endif
615 			if (rt)
616 				RTFREE_LOCKED(rt);
617 			return -1;
618 		}
619 		RTFREE_LOCKED(rt);
620 	}
621 
622 	return 0;
623 }
624 
625 static int
626 stf_checkaddr6(sc, in6, inifp)
627 	struct stf_softc *sc;
628 	struct in6_addr *in6;
629 	struct ifnet *inifp;	/* incoming interface */
630 {
631 	/*
632 	 * check 6to4 addresses
633 	 */
634 	if (IN6_IS_ADDR_6TO4(in6)) {
635 		struct in_addr in4;
636 		bcopy(GET_V4(in6), &in4, sizeof(in4));
637 		return stf_checkaddr4(sc, &in4, inifp);
638 	}
639 
640 	/*
641 	 * reject anything that look suspicious.  the test is implemented
642 	 * in ip6_input too, but we check here as well to
643 	 * (1) reject bad packets earlier, and
644 	 * (2) to be safe against future ip6_input change.
645 	 */
646 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
647 		return -1;
648 
649 	return 0;
650 }
651 
652 void
653 in_stf_input(m, off)
654 	struct mbuf *m;
655 	int off;
656 {
657 	int proto;
658 	struct stf_softc *sc;
659 	struct ip *ip;
660 	struct ip6_hdr *ip6;
661 	u_int8_t otos, itos;
662 	struct ifnet *ifp;
663 
664 	proto = mtod(m, struct ip *)->ip_p;
665 
666 	if (proto != IPPROTO_IPV6) {
667 		m_freem(m);
668 		return;
669 	}
670 
671 	ip = mtod(m, struct ip *);
672 
673 	sc = (struct stf_softc *)encap_getarg(m);
674 
675 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
676 		m_freem(m);
677 		return;
678 	}
679 
680 	ifp = STF2IFP(sc);
681 
682 #ifdef MAC
683 	mac_ifnet_create_mbuf(ifp, m);
684 #endif
685 
686 	/*
687 	 * perform sanity check against outer src/dst.
688 	 * for source, perform ingress filter as well.
689 	 */
690 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
691 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
692 		m_freem(m);
693 		return;
694 	}
695 
696 	otos = ip->ip_tos;
697 	m_adj(m, off);
698 
699 	if (m->m_len < sizeof(*ip6)) {
700 		m = m_pullup(m, sizeof(*ip6));
701 		if (!m)
702 			return;
703 	}
704 	ip6 = mtod(m, struct ip6_hdr *);
705 
706 	/*
707 	 * perform sanity check against inner src/dst.
708 	 * for source, perform ingress filter as well.
709 	 */
710 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
711 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
712 		m_freem(m);
713 		return;
714 	}
715 
716 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
717 	if ((ifp->if_flags & IFF_LINK1) != 0)
718 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
719 	else
720 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
721 	ip6->ip6_flow &= ~htonl(0xff << 20);
722 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
723 
724 	m->m_pkthdr.rcvif = ifp;
725 
726 	if (bpf_peers_present(ifp->if_bpf)) {
727 		/*
728 		 * We need to prepend the address family as
729 		 * a four byte field.  Cons up a dummy header
730 		 * to pacify bpf.  This is safe because bpf
731 		 * will only read from the mbuf (i.e., it won't
732 		 * try to free it or keep a pointer a to it).
733 		 */
734 		u_int32_t af = AF_INET6;
735 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
736 	}
737 
738 	/*
739 	 * Put the packet to the network layer input queue according to the
740 	 * specified address family.
741 	 * See net/if_gif.c for possible issues with packet processing
742 	 * reorder due to extra queueing.
743 	 */
744 	ifp->if_ipackets++;
745 	ifp->if_ibytes += m->m_pkthdr.len;
746 	netisr_dispatch(NETISR_IPV6, m);
747 }
748 
749 /* ARGSUSED */
750 static void
751 stf_rtrequest(cmd, rt, info)
752 	int cmd;
753 	struct rtentry *rt;
754 	struct rt_addrinfo *info;
755 {
756 	RT_LOCK_ASSERT(rt);
757 	rt->rt_rmx.rmx_mtu = IPV6_MMTU;
758 }
759 
760 static int
761 stf_ioctl(ifp, cmd, data)
762 	struct ifnet *ifp;
763 	u_long cmd;
764 	caddr_t data;
765 {
766 	struct ifaddr *ifa;
767 	struct ifreq *ifr;
768 	struct sockaddr_in6 *sin6;
769 	struct in_addr addr;
770 	int error;
771 
772 	error = 0;
773 	switch (cmd) {
774 	case SIOCSIFADDR:
775 		ifa = (struct ifaddr *)data;
776 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
777 			error = EAFNOSUPPORT;
778 			break;
779 		}
780 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
781 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
782 			error = EINVAL;
783 			break;
784 		}
785 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
786 		if (isrfc1918addr(&addr)) {
787 			error = EINVAL;
788 			break;
789 		}
790 
791 		ifa->ifa_rtrequest = stf_rtrequest;
792 		ifp->if_flags |= IFF_UP;
793 		break;
794 
795 	case SIOCADDMULTI:
796 	case SIOCDELMULTI:
797 		ifr = (struct ifreq *)data;
798 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
799 			;
800 		else
801 			error = EAFNOSUPPORT;
802 		break;
803 
804 	default:
805 		error = EINVAL;
806 		break;
807 	}
808 
809 	return error;
810 }
811