xref: /freebsd/sys/net/if_stf.c (revision ca9ac06c99bfd0150b85d4d83c396ce6237c0e05)
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/mac.h>
86 #include <sys/mbuf.h>
87 #include <sys/errno.h>
88 #include <sys/kernel.h>
89 #include <sys/module.h>
90 #include <sys/protosw.h>
91 #include <sys/queue.h>
92 #include <machine/cpu.h>
93 
94 #include <sys/malloc.h>
95 
96 #include <net/if.h>
97 #include <net/if_clone.h>
98 #include <net/route.h>
99 #include <net/netisr.h>
100 #include <net/if_types.h>
101 #include <net/if_stf.h>
102 
103 #include <netinet/in.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/ip.h>
106 #include <netinet/ip_var.h>
107 #include <netinet/in_var.h>
108 
109 #include <netinet/ip6.h>
110 #include <netinet6/ip6_var.h>
111 #include <netinet6/in6_var.h>
112 #include <netinet/ip_ecn.h>
113 
114 #include <netinet/ip_encap.h>
115 
116 #include <machine/stdarg.h>
117 
118 #include <net/net_osdep.h>
119 
120 #include <net/bpf.h>
121 
122 #define STFNAME		"stf"
123 #define STFUNIT		0
124 
125 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
126 
127 /*
128  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
129  * struct in_addr *; use bcopy() instead.
130  */
131 #define GET_V4(x)	((caddr_t)(&(x)->s6_addr16[1]))
132 
133 struct stf_softc {
134 	struct ifnet	sc_if;	   /* common area */
135 	union {
136 		struct route  __sc_ro4;
137 		struct route_in6 __sc_ro6; /* just for safety */
138 	} __sc_ro46;
139 #define sc_ro	__sc_ro46.__sc_ro4
140 	const struct encaptab *encap_cookie;
141 	LIST_ENTRY(stf_softc) sc_list;	/* all stf's are linked */
142 };
143 
144 /*
145  * All mutable global variables in if_stf.c are protected by stf_mtx.
146  * XXXRW: Note that mutable fields in the softc are not currently locked:
147  * in particular, sc_ro needs to be protected from concurrent entrance
148  * of stf_output().
149  */
150 static struct mtx stf_mtx;
151 static LIST_HEAD(, stf_softc) stf_softc_list;
152 
153 static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
154 static const int ip_stf_ttl = 40;
155 
156 extern  struct domain inetdomain;
157 struct protosw in_stf_protosw =
158 { SOCK_RAW,	&inetdomain,	IPPROTO_IPV6,	PR_ATOMIC|PR_ADDR,
159   in_stf_input,	(pr_output_t*)rip_output, 0,	rip_ctloutput,
160   0,
161   0,            0,              0,              0,
162   &rip_usrreqs
163 };
164 
165 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
166 
167 static int stfmodevent(module_t, int, void *);
168 static int stf_encapcheck(const struct mbuf *, int, int, void *);
169 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
170 static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
171 	struct rtentry *);
172 static int isrfc1918addr(struct in_addr *);
173 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
174 	struct ifnet *);
175 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
176 	struct ifnet *);
177 static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
178 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
179 
180 static int stf_clone_match(struct if_clone *, const char *);
181 static int stf_clone_create(struct if_clone *, char *, size_t);
182 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
183 struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0,
184     NULL, stf_clone_match, stf_clone_create, stf_clone_destroy);
185 
186 static int
187 stf_clone_match(struct if_clone *ifc, const char *name)
188 {
189 	int i;
190 
191 	for(i = 0; stfnames[i] != NULL; i++) {
192 		if (strcmp(stfnames[i], name) == 0)
193 			return (1);
194 	}
195 
196 	return (0);
197 }
198 
199 static int
200 stf_clone_create(struct if_clone *ifc, char *name, size_t len)
201 {
202 	int err, unit;
203 	struct stf_softc *sc;
204 	struct ifnet *ifp;
205 
206 	/*
207 	 * We can only have one unit, but since unit allocation is
208 	 * already locked, we use it to keep from allocating extra
209 	 * interfaces.
210 	 */
211 	unit = STFUNIT;
212 	err = ifc_alloc_unit(ifc, &unit);
213 	if (err != 0)
214 		return (err);
215 
216 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
217 	ifp = &sc->sc_if;
218 	/*
219 	 * Set the name manually rather then using if_initname because
220 	 * we don't conform to the default naming convention for interfaces.
221 	 */
222 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
223 	ifp->if_dname = ifc->ifc_name;
224 	ifp->if_dunit = IF_DUNIT_NONE;
225 
226 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
227 	    stf_encapcheck, &in_stf_protosw, sc);
228 	if (sc->encap_cookie == NULL) {
229 		if_printf(ifp, "attach failed\n");
230 		free(sc, M_STF);
231 		ifc_free_unit(ifc, unit);
232 		return (ENOMEM);
233 	}
234 
235 	ifp->if_mtu    = IPV6_MMTU;
236 	ifp->if_ioctl  = stf_ioctl;
237 	ifp->if_output = stf_output;
238 	ifp->if_type   = IFT_STF;
239 	ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
240 	if_attach(ifp);
241 	bpfattach(ifp, DLT_NULL, sizeof(u_int));
242 	mtx_lock(&stf_mtx);
243 	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
244 	mtx_unlock(&stf_mtx);
245 	return (0);
246 }
247 
248 static void
249 stf_destroy(struct stf_softc *sc)
250 {
251 	int err;
252 
253 	err = encap_detach(sc->encap_cookie);
254 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
255 	bpfdetach(&sc->sc_if);
256 	if_detach(&sc->sc_if);
257 
258 	free(sc, M_STF);
259 }
260 
261 static int
262 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
263 {
264 	struct stf_softc *sc = (void *) ifp;
265 
266 	mtx_lock(&stf_mtx);
267 	LIST_REMOVE(sc, sc_list);
268 	mtx_unlock(&stf_mtx);
269 
270 	stf_destroy(sc);
271 	ifc_free_unit(ifc, STFUNIT);
272 
273 	return (0);
274 }
275 
276 static int
277 stfmodevent(mod, type, data)
278 	module_t mod;
279 	int type;
280 	void *data;
281 {
282 	struct stf_softc *sc;
283 
284 	switch (type) {
285 	case MOD_LOAD:
286 		mtx_init(&stf_mtx, "stf_mtx", NULL, MTX_DEF);
287 		LIST_INIT(&stf_softc_list);
288 		if_clone_attach(&stf_cloner);
289 
290 		break;
291 	case MOD_UNLOAD:
292 		if_clone_detach(&stf_cloner);
293 
294 		mtx_lock(&stf_mtx);
295 		while ((sc = LIST_FIRST(&stf_softc_list)) != NULL) {
296 			LIST_REMOVE(sc, sc_list);
297 			mtx_unlock(&stf_mtx);
298 			stf_destroy(sc);
299 			mtx_lock(&stf_mtx);
300 		}
301 		mtx_unlock(&stf_mtx);
302 		mtx_destroy(&stf_mtx);
303 		break;
304 	default:
305 		return (EOPNOTSUPP);
306 	}
307 
308 	return (0);
309 }
310 
311 static moduledata_t stf_mod = {
312 	"if_stf",
313 	stfmodevent,
314 	0
315 };
316 
317 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
318 
319 static int
320 stf_encapcheck(m, off, proto, arg)
321 	const struct mbuf *m;
322 	int off;
323 	int proto;
324 	void *arg;
325 {
326 	struct ip ip;
327 	struct in6_ifaddr *ia6;
328 	struct stf_softc *sc;
329 	struct in_addr a, b, mask;
330 
331 	sc = (struct stf_softc *)arg;
332 	if (sc == NULL)
333 		return 0;
334 
335 	if ((sc->sc_if.if_flags & IFF_UP) == 0)
336 		return 0;
337 
338 	/* IFF_LINK0 means "no decapsulation" */
339 	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
340 		return 0;
341 
342 	if (proto != IPPROTO_IPV6)
343 		return 0;
344 
345 	/* LINTED const cast */
346 	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
347 
348 	if (ip.ip_v != 4)
349 		return 0;
350 
351 	ia6 = stf_getsrcifa6(&sc->sc_if);
352 	if (ia6 == NULL)
353 		return 0;
354 
355 	/*
356 	 * check if IPv4 dst matches the IPv4 address derived from the
357 	 * local 6to4 address.
358 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
359 	 */
360 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
361 	    sizeof(ip.ip_dst)) != 0)
362 		return 0;
363 
364 	/*
365 	 * check if IPv4 src matches the IPv4 address derived from the
366 	 * local 6to4 address masked by prefixmask.
367 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
368 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
369 	 */
370 	bzero(&a, sizeof(a));
371 	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
372 	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
373 	a.s_addr &= mask.s_addr;
374 	b = ip.ip_src;
375 	b.s_addr &= mask.s_addr;
376 	if (a.s_addr != b.s_addr)
377 		return 0;
378 
379 	/* stf interface makes single side match only */
380 	return 32;
381 }
382 
383 static struct in6_ifaddr *
384 stf_getsrcifa6(ifp)
385 	struct ifnet *ifp;
386 {
387 	struct ifaddr *ia;
388 	struct in_ifaddr *ia4;
389 	struct sockaddr_in6 *sin6;
390 	struct in_addr in;
391 
392 	for (ia = TAILQ_FIRST(&ifp->if_addrlist);
393 	     ia;
394 	     ia = TAILQ_NEXT(ia, ifa_list))
395 	{
396 		if (ia->ifa_addr == NULL)
397 			continue;
398 		if (ia->ifa_addr->sa_family != AF_INET6)
399 			continue;
400 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
401 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
402 			continue;
403 
404 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
405 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
406 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
407 				break;
408 		if (ia4 == NULL)
409 			continue;
410 
411 		return (struct in6_ifaddr *)ia;
412 	}
413 
414 	return NULL;
415 }
416 
417 static int
418 stf_output(ifp, m, dst, rt)
419 	struct ifnet *ifp;
420 	struct mbuf *m;
421 	struct sockaddr *dst;
422 	struct rtentry *rt;
423 {
424 	struct stf_softc *sc;
425 	struct sockaddr_in6 *dst6;
426 	struct in_addr in4;
427 	caddr_t ptr;
428 	struct sockaddr_in *dst4;
429 	u_int8_t tos;
430 	struct ip *ip;
431 	struct ip6_hdr *ip6;
432 	struct in6_ifaddr *ia6;
433 #ifdef MAC
434 	int error;
435 
436 	error = mac_check_ifnet_transmit(ifp, m);
437 	if (error) {
438 		m_freem(m);
439 		return (error);
440 	}
441 #endif
442 
443 	sc = (struct stf_softc*)ifp;
444 	dst6 = (struct sockaddr_in6 *)dst;
445 
446 	/* just in case */
447 	if ((ifp->if_flags & IFF_UP) == 0) {
448 		m_freem(m);
449 		ifp->if_oerrors++;
450 		return ENETDOWN;
451 	}
452 
453 	/*
454 	 * If we don't have an ip4 address that match my inner ip6 address,
455 	 * we shouldn't generate output.  Without this check, we'll end up
456 	 * using wrong IPv4 source.
457 	 */
458 	ia6 = stf_getsrcifa6(ifp);
459 	if (ia6 == NULL) {
460 		m_freem(m);
461 		ifp->if_oerrors++;
462 		return ENETDOWN;
463 	}
464 
465 	if (m->m_len < sizeof(*ip6)) {
466 		m = m_pullup(m, sizeof(*ip6));
467 		if (!m) {
468 			ifp->if_oerrors++;
469 			return ENOBUFS;
470 		}
471 	}
472 	ip6 = mtod(m, struct ip6_hdr *);
473 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
474 
475 	/*
476 	 * Pickup the right outer dst addr from the list of candidates.
477 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
478 	 */
479 	ptr = NULL;
480 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
481 		ptr = GET_V4(&ip6->ip6_dst);
482 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
483 		ptr = GET_V4(&dst6->sin6_addr);
484 	else {
485 		m_freem(m);
486 		ifp->if_oerrors++;
487 		return ENETUNREACH;
488 	}
489 	bcopy(ptr, &in4, sizeof(in4));
490 
491 	if (ifp->if_bpf) {
492 		/*
493 		 * We need to prepend the address family as
494 		 * a four byte field.  Cons up a dummy header
495 		 * to pacify bpf.  This is safe because bpf
496 		 * will only read from the mbuf (i.e., it won't
497 		 * try to free it or keep a pointer a to it).
498 		 */
499 		u_int32_t af = AF_INET6;
500 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
501 	}
502 
503 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
504 	if (m && m->m_len < sizeof(struct ip))
505 		m = m_pullup(m, sizeof(struct ip));
506 	if (m == NULL) {
507 		ifp->if_oerrors++;
508 		return ENOBUFS;
509 	}
510 	ip = mtod(m, struct ip *);
511 
512 	bzero(ip, sizeof(*ip));
513 
514 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
515 	    &ip->ip_src, sizeof(ip->ip_src));
516 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
517 	ip->ip_p = IPPROTO_IPV6;
518 	ip->ip_ttl = ip_stf_ttl;
519 	ip->ip_len = m->m_pkthdr.len;	/*host order*/
520 	if (ifp->if_flags & IFF_LINK1)
521 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
522 	else
523 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
524 
525 	/*
526 	 * XXXRW: Locking of sc_ro required.
527 	 */
528 	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
529 	if (dst4->sin_family != AF_INET ||
530 	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
531 		/* cache route doesn't match */
532 		dst4->sin_family = AF_INET;
533 		dst4->sin_len = sizeof(struct sockaddr_in);
534 		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
535 		if (sc->sc_ro.ro_rt) {
536 			RTFREE(sc->sc_ro.ro_rt);
537 			sc->sc_ro.ro_rt = NULL;
538 		}
539 	}
540 
541 	if (sc->sc_ro.ro_rt == NULL) {
542 		rtalloc(&sc->sc_ro);
543 		if (sc->sc_ro.ro_rt == NULL) {
544 			m_freem(m);
545 			ifp->if_oerrors++;
546 			return ENETUNREACH;
547 		}
548 	}
549 
550 	ifp->if_opackets++;
551 	return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL);
552 }
553 
554 static int
555 isrfc1918addr(in)
556 	struct in_addr *in;
557 {
558 	/*
559 	 * returns 1 if private address range:
560 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
561 	 */
562 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
563 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
564 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
565 		return 1;
566 
567 	return 0;
568 }
569 
570 static int
571 stf_checkaddr4(sc, in, inifp)
572 	struct stf_softc *sc;
573 	struct in_addr *in;
574 	struct ifnet *inifp;	/* incoming interface */
575 {
576 	struct in_ifaddr *ia4;
577 
578 	/*
579 	 * reject packets with the following address:
580 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
581 	 */
582 	if (IN_MULTICAST(ntohl(in->s_addr)))
583 		return -1;
584 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
585 	case 0: case 127: case 255:
586 		return -1;
587 	}
588 
589 	/*
590 	 * reject packets with private address range.
591 	 * (requirement from RFC3056 section 2 1st paragraph)
592 	 */
593 	if (isrfc1918addr(in))
594 		return -1;
595 
596 	/*
597 	 * reject packets with broadcast
598 	 */
599 	for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
600 	     ia4;
601 	     ia4 = TAILQ_NEXT(ia4, ia_link))
602 	{
603 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
604 			continue;
605 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
606 			return -1;
607 	}
608 
609 	/*
610 	 * perform ingress filter
611 	 */
612 	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
613 		struct sockaddr_in sin;
614 		struct rtentry *rt;
615 
616 		bzero(&sin, sizeof(sin));
617 		sin.sin_family = AF_INET;
618 		sin.sin_len = sizeof(struct sockaddr_in);
619 		sin.sin_addr = *in;
620 		rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
621 		if (!rt || rt->rt_ifp != inifp) {
622 #if 0
623 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
624 			    "due to ingress filter\n", if_name(&sc->sc_if),
625 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
626 #endif
627 			if (rt)
628 				rtfree(rt);
629 			return -1;
630 		}
631 		rtfree(rt);
632 	}
633 
634 	return 0;
635 }
636 
637 static int
638 stf_checkaddr6(sc, in6, inifp)
639 	struct stf_softc *sc;
640 	struct in6_addr *in6;
641 	struct ifnet *inifp;	/* incoming interface */
642 {
643 	/*
644 	 * check 6to4 addresses
645 	 */
646 	if (IN6_IS_ADDR_6TO4(in6)) {
647 		struct in_addr in4;
648 		bcopy(GET_V4(in6), &in4, sizeof(in4));
649 		return stf_checkaddr4(sc, &in4, inifp);
650 	}
651 
652 	/*
653 	 * reject anything that look suspicious.  the test is implemented
654 	 * in ip6_input too, but we check here as well to
655 	 * (1) reject bad packets earlier, and
656 	 * (2) to be safe against future ip6_input change.
657 	 */
658 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
659 		return -1;
660 
661 	return 0;
662 }
663 
664 void
665 in_stf_input(m, off)
666 	struct mbuf *m;
667 	int off;
668 {
669 	int proto;
670 	struct stf_softc *sc;
671 	struct ip *ip;
672 	struct ip6_hdr *ip6;
673 	u_int8_t otos, itos;
674 	struct ifnet *ifp;
675 
676 	proto = mtod(m, struct ip *)->ip_p;
677 
678 	if (proto != IPPROTO_IPV6) {
679 		m_freem(m);
680 		return;
681 	}
682 
683 	ip = mtod(m, struct ip *);
684 
685 	sc = (struct stf_softc *)encap_getarg(m);
686 
687 	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
688 		m_freem(m);
689 		return;
690 	}
691 
692 	ifp = &sc->sc_if;
693 
694 #ifdef MAC
695 	mac_create_mbuf_from_ifnet(ifp, m);
696 #endif
697 
698 	/*
699 	 * perform sanity check against outer src/dst.
700 	 * for source, perform ingress filter as well.
701 	 */
702 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
703 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
704 		m_freem(m);
705 		return;
706 	}
707 
708 	otos = ip->ip_tos;
709 	m_adj(m, off);
710 
711 	if (m->m_len < sizeof(*ip6)) {
712 		m = m_pullup(m, sizeof(*ip6));
713 		if (!m)
714 			return;
715 	}
716 	ip6 = mtod(m, struct ip6_hdr *);
717 
718 	/*
719 	 * perform sanity check against inner src/dst.
720 	 * for source, perform ingress filter as well.
721 	 */
722 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
723 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
724 		m_freem(m);
725 		return;
726 	}
727 
728 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
729 	if ((ifp->if_flags & IFF_LINK1) != 0)
730 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
731 	else
732 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
733 	ip6->ip6_flow &= ~htonl(0xff << 20);
734 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
735 
736 	m->m_pkthdr.rcvif = ifp;
737 
738 	if (ifp->if_bpf) {
739 		/*
740 		 * We need to prepend the address family as
741 		 * a four byte field.  Cons up a dummy header
742 		 * to pacify bpf.  This is safe because bpf
743 		 * will only read from the mbuf (i.e., it won't
744 		 * try to free it or keep a pointer a to it).
745 		 */
746 		u_int32_t af = AF_INET6;
747 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
748 	}
749 
750 	/*
751 	 * Put the packet to the network layer input queue according to the
752 	 * specified address family.
753 	 * See net/if_gif.c for possible issues with packet processing
754 	 * reorder due to extra queueing.
755 	 */
756 	ifp->if_ipackets++;
757 	ifp->if_ibytes += m->m_pkthdr.len;
758 	netisr_dispatch(NETISR_IPV6, m);
759 }
760 
761 /* ARGSUSED */
762 static void
763 stf_rtrequest(cmd, rt, info)
764 	int cmd;
765 	struct rtentry *rt;
766 	struct rt_addrinfo *info;
767 {
768 	RT_LOCK_ASSERT(rt);
769 	rt->rt_rmx.rmx_mtu = IPV6_MMTU;
770 }
771 
772 static int
773 stf_ioctl(ifp, cmd, data)
774 	struct ifnet *ifp;
775 	u_long cmd;
776 	caddr_t data;
777 {
778 	struct ifaddr *ifa;
779 	struct ifreq *ifr;
780 	struct sockaddr_in6 *sin6;
781 	struct in_addr addr;
782 	int error;
783 
784 	error = 0;
785 	switch (cmd) {
786 	case SIOCSIFADDR:
787 		ifa = (struct ifaddr *)data;
788 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
789 			error = EAFNOSUPPORT;
790 			break;
791 		}
792 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
793 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
794 			error = EINVAL;
795 			break;
796 		}
797 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
798 		if (isrfc1918addr(&addr)) {
799 			error = EINVAL;
800 			break;
801 		}
802 
803 		ifa->ifa_rtrequest = stf_rtrequest;
804 		ifp->if_flags |= IFF_UP;
805 		break;
806 
807 	case SIOCADDMULTI:
808 	case SIOCDELMULTI:
809 		ifr = (struct ifreq *)data;
810 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
811 			;
812 		else
813 			error = EAFNOSUPPORT;
814 		break;
815 
816 	default:
817 		error = EINVAL;
818 		break;
819 	}
820 
821 	return error;
822 }
823