xref: /freebsd/sys/net/if_stf.c (revision 195ebc7e9e4b129de810833791a19dfb4349d6a9) 
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 <sys/sysctl.h>
93 #include <machine/cpu.h>
94 
95 #include <sys/malloc.h>
96 #include <sys/vimage.h>
97 
98 #include <net/if.h>
99 #include <net/if_clone.h>
100 #include <net/route.h>
101 #include <net/netisr.h>
102 #include <net/if_types.h>
103 #include <net/if_stf.h>
104 
105 #include <netinet/in.h>
106 #include <netinet/in_systm.h>
107 #include <netinet/ip.h>
108 #include <netinet/ip_var.h>
109 #include <netinet/in_var.h>
110 #include <netinet/vinet.h>
111 
112 #include <netinet/ip6.h>
113 #include <netinet6/ip6_var.h>
114 #include <netinet6/in6_var.h>
115 #include <netinet/ip_ecn.h>
116 
117 #include <netinet/ip_encap.h>
118 
119 #include <machine/stdarg.h>
120 
121 #include <net/bpf.h>
122 
123 #include <security/mac/mac_framework.h>
124 
125 SYSCTL_DECL(_net_link);
126 SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
127 
128 static int stf_route_cache = 1;
129 SYSCTL_INT(_net_link_stf, OID_AUTO, route_cache, CTLFLAG_RW,
130     &stf_route_cache, 0, "Caching of IPv4 routes for 6to4 Output");
131 
132 #define STFNAME		"stf"
133 #define STFUNIT		0
134 
135 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
136 
137 /*
138  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
139  * struct in_addr *; use bcopy() instead.
140  */
141 #define GET_V4(x)	((caddr_t)(&(x)->s6_addr16[1]))
142 
143 struct stf_softc {
144 	struct ifnet	*sc_ifp;
145 	union {
146 		struct route  __sc_ro4;
147 		struct route_in6 __sc_ro6; /* just for safety */
148 	} __sc_ro46;
149 #define sc_ro	__sc_ro46.__sc_ro4
150 	struct mtx	sc_ro_mtx;
151 	u_int	sc_fibnum;
152 	const struct encaptab *encap_cookie;
153 };
154 #define STF2IFP(sc)	((sc)->sc_ifp)
155 
156 /*
157  * Note that mutable fields in the softc are not currently locked.
158  * We do lock sc_ro in stf_output though.
159  */
160 static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
161 static const int ip_stf_ttl = 40;
162 
163 extern  struct domain inetdomain;
164 struct protosw in_stf_protosw = {
165 	.pr_type =		SOCK_RAW,
166 	.pr_domain =		&inetdomain,
167 	.pr_protocol =		IPPROTO_IPV6,
168 	.pr_flags =		PR_ATOMIC|PR_ADDR,
169 	.pr_input =		in_stf_input,
170 	.pr_output =		(pr_output_t *)rip_output,
171 	.pr_ctloutput =		rip_ctloutput,
172 	.pr_usrreqs =		&rip_usrreqs
173 };
174 
175 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
176 
177 static int stfmodevent(module_t, int, void *);
178 static int stf_encapcheck(const struct mbuf *, int, int, void *);
179 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
180 static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
181 	struct route *);
182 static int isrfc1918addr(struct in_addr *);
183 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
184 	struct ifnet *);
185 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
186 	struct ifnet *);
187 static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
188 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
189 
190 static int stf_clone_match(struct if_clone *, const char *);
191 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
192 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
193 struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0,
194     NULL, stf_clone_match, stf_clone_create, stf_clone_destroy);
195 
196 static int
197 stf_clone_match(struct if_clone *ifc, const char *name)
198 {
199 	int i;
200 
201 	for(i = 0; stfnames[i] != NULL; i++) {
202 		if (strcmp(stfnames[i], name) == 0)
203 			return (1);
204 	}
205 
206 	return (0);
207 }
208 
209 static int
210 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
211 {
212 	int err, unit;
213 	struct stf_softc *sc;
214 	struct ifnet *ifp;
215 
216 	/*
217 	 * We can only have one unit, but since unit allocation is
218 	 * already locked, we use it to keep from allocating extra
219 	 * interfaces.
220 	 */
221 	unit = STFUNIT;
222 	err = ifc_alloc_unit(ifc, &unit);
223 	if (err != 0)
224 		return (err);
225 
226 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
227 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
228 	if (ifp == NULL) {
229 		free(sc, M_STF);
230 		ifc_free_unit(ifc, unit);
231 		return (ENOSPC);
232 	}
233 	ifp->if_softc = sc;
234 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
235 
236 	/*
237 	 * Set the name manually rather then using if_initname because
238 	 * we don't conform to the default naming convention for interfaces.
239 	 */
240 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
241 	ifp->if_dname = ifc->ifc_name;
242 	ifp->if_dunit = IF_DUNIT_NONE;
243 
244 	mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
245 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
246 	    stf_encapcheck, &in_stf_protosw, sc);
247 	if (sc->encap_cookie == NULL) {
248 		if_printf(ifp, "attach failed\n");
249 		free(sc, M_STF);
250 		ifc_free_unit(ifc, unit);
251 		return (ENOMEM);
252 	}
253 
254 	ifp->if_mtu    = IPV6_MMTU;
255 	ifp->if_ioctl  = stf_ioctl;
256 	ifp->if_output = stf_output;
257 	ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
258 	if_attach(ifp);
259 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
260 	return (0);
261 }
262 
263 static int
264 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
265 {
266 	struct stf_softc *sc = ifp->if_softc;
267 	int err;
268 
269 	err = encap_detach(sc->encap_cookie);
270 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
271 	mtx_destroy(&(sc)->sc_ro_mtx);
272 	bpfdetach(ifp);
273 	if_detach(ifp);
274 	if_free(ifp);
275 
276 	free(sc, M_STF);
277 	ifc_free_unit(ifc, STFUNIT);
278 
279 	return (0);
280 }
281 
282 static int
283 stfmodevent(mod, type, data)
284 	module_t mod;
285 	int type;
286 	void *data;
287 {
288 
289 	switch (type) {
290 	case MOD_LOAD:
291 		if_clone_attach(&stf_cloner);
292 		break;
293 	case MOD_UNLOAD:
294 		if_clone_detach(&stf_cloner);
295 		break;
296 	default:
297 		return (EOPNOTSUPP);
298 	}
299 
300 	return (0);
301 }
302 
303 static moduledata_t stf_mod = {
304 	"if_stf",
305 	stfmodevent,
306 	0
307 };
308 
309 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
310 
311 static int
312 stf_encapcheck(m, off, proto, arg)
313 	const struct mbuf *m;
314 	int off;
315 	int proto;
316 	void *arg;
317 {
318 	struct ip ip;
319 	struct in6_ifaddr *ia6;
320 	struct stf_softc *sc;
321 	struct in_addr a, b, mask;
322 
323 	sc = (struct stf_softc *)arg;
324 	if (sc == NULL)
325 		return 0;
326 
327 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
328 		return 0;
329 
330 	/* IFF_LINK0 means "no decapsulation" */
331 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
332 		return 0;
333 
334 	if (proto != IPPROTO_IPV6)
335 		return 0;
336 
337 	/* LINTED const cast */
338 	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
339 
340 	if (ip.ip_v != 4)
341 		return 0;
342 
343 	ia6 = stf_getsrcifa6(STF2IFP(sc));
344 	if (ia6 == NULL)
345 		return 0;
346 
347 	/*
348 	 * check if IPv4 dst matches the IPv4 address derived from the
349 	 * local 6to4 address.
350 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
351 	 */
352 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
353 	    sizeof(ip.ip_dst)) != 0)
354 		return 0;
355 
356 	/*
357 	 * check if IPv4 src matches the IPv4 address derived from the
358 	 * local 6to4 address masked by prefixmask.
359 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
360 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
361 	 */
362 	bzero(&a, sizeof(a));
363 	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
364 	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
365 	a.s_addr &= mask.s_addr;
366 	b = ip.ip_src;
367 	b.s_addr &= mask.s_addr;
368 	if (a.s_addr != b.s_addr)
369 		return 0;
370 
371 	/* stf interface makes single side match only */
372 	return 32;
373 }
374 
375 static struct in6_ifaddr *
376 stf_getsrcifa6(ifp)
377 	struct ifnet *ifp;
378 {
379 	INIT_VNET_INET(ifp->if_vnet);
380 	struct ifaddr *ia;
381 	struct in_ifaddr *ia4;
382 	struct sockaddr_in6 *sin6;
383 	struct in_addr in;
384 
385 	IF_ADDR_LOCK(ifp);
386 	TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
387 		if (ia->ifa_addr->sa_family != AF_INET6)
388 			continue;
389 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
390 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
391 			continue;
392 
393 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
394 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
395 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
396 				break;
397 		if (ia4 == NULL)
398 			continue;
399 
400 		IF_ADDR_UNLOCK(ifp);
401 		return (struct in6_ifaddr *)ia;
402 	}
403 	IF_ADDR_UNLOCK(ifp);
404 
405 	return NULL;
406 }
407 
408 static int
409 stf_output(ifp, m, dst, ro)
410 	struct ifnet *ifp;
411 	struct mbuf *m;
412 	struct sockaddr *dst;
413 	struct route *ro;
414 {
415 	struct stf_softc *sc;
416 	struct sockaddr_in6 *dst6;
417 	struct route *cached_route;
418 	struct in_addr in4;
419 	caddr_t ptr;
420 	struct sockaddr_in *dst4;
421 	u_int8_t tos;
422 	struct ip *ip;
423 	struct ip6_hdr *ip6;
424 	struct in6_ifaddr *ia6;
425 	u_int32_t af;
426 	int error;
427 
428 #ifdef MAC
429 	error = mac_ifnet_check_transmit(ifp, m);
430 	if (error) {
431 		m_freem(m);
432 		return (error);
433 	}
434 #endif
435 
436 	sc = ifp->if_softc;
437 	dst6 = (struct sockaddr_in6 *)dst;
438 
439 	/* just in case */
440 	if ((ifp->if_flags & IFF_UP) == 0) {
441 		m_freem(m);
442 		ifp->if_oerrors++;
443 		return ENETDOWN;
444 	}
445 
446 	/*
447 	 * If we don't have an ip4 address that match my inner ip6 address,
448 	 * we shouldn't generate output.  Without this check, we'll end up
449 	 * using wrong IPv4 source.
450 	 */
451 	ia6 = stf_getsrcifa6(ifp);
452 	if (ia6 == NULL) {
453 		m_freem(m);
454 		ifp->if_oerrors++;
455 		return ENETDOWN;
456 	}
457 
458 	if (m->m_len < sizeof(*ip6)) {
459 		m = m_pullup(m, sizeof(*ip6));
460 		if (!m) {
461 			ifp->if_oerrors++;
462 			return ENOBUFS;
463 		}
464 	}
465 	ip6 = mtod(m, struct ip6_hdr *);
466 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
467 
468 	/*
469 	 * BPF writes need to be handled specially.
470 	 * This is a null operation, nothing here checks dst->sa_family.
471 	 */
472 	if (dst->sa_family == AF_UNSPEC) {
473 		bcopy(dst->sa_data, &af, sizeof(af));
474 		dst->sa_family = af;
475 	}
476 
477 	/*
478 	 * Pickup the right outer dst addr from the list of candidates.
479 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
480 	 */
481 	ptr = NULL;
482 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
483 		ptr = GET_V4(&ip6->ip6_dst);
484 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
485 		ptr = GET_V4(&dst6->sin6_addr);
486 	else {
487 		m_freem(m);
488 		ifp->if_oerrors++;
489 		return ENETUNREACH;
490 	}
491 	bcopy(ptr, &in4, sizeof(in4));
492 
493 	if (bpf_peers_present(ifp->if_bpf)) {
494 		/*
495 		 * We need to prepend the address family as
496 		 * a four byte field.  Cons up a dummy header
497 		 * to pacify bpf.  This is safe because bpf
498 		 * will only read from the mbuf (i.e., it won't
499 		 * try to free it or keep a pointer a to it).
500 		 */
501 		af = AF_INET6;
502 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
503 	}
504 
505 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
506 	if (m && m->m_len < sizeof(struct ip))
507 		m = m_pullup(m, sizeof(struct ip));
508 	if (m == NULL) {
509 		ifp->if_oerrors++;
510 		return ENOBUFS;
511 	}
512 	ip = mtod(m, struct ip *);
513 
514 	bzero(ip, sizeof(*ip));
515 
516 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
517 	    &ip->ip_src, sizeof(ip->ip_src));
518 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
519 	ip->ip_p = IPPROTO_IPV6;
520 	ip->ip_ttl = ip_stf_ttl;
521 	ip->ip_len = m->m_pkthdr.len;	/*host order*/
522 	if (ifp->if_flags & IFF_LINK1)
523 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
524 	else
525 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
526 
527 	if (!stf_route_cache) {
528 		cached_route = NULL;
529 		goto sendit;
530 	}
531 
532 	/*
533 	 * Do we have a cached route?
534 	 */
535 	mtx_lock(&(sc)->sc_ro_mtx);
536 	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
537 	if (dst4->sin_family != AF_INET ||
538 	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
539 		/* cache route doesn't match */
540 		dst4->sin_family = AF_INET;
541 		dst4->sin_len = sizeof(struct sockaddr_in);
542 		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
543 		if (sc->sc_ro.ro_rt) {
544 			RTFREE(sc->sc_ro.ro_rt);
545 			sc->sc_ro.ro_rt = NULL;
546 		}
547 	}
548 
549 	if (sc->sc_ro.ro_rt == NULL) {
550 		rtalloc_fib(&sc->sc_ro, sc->sc_fibnum);
551 		if (sc->sc_ro.ro_rt == NULL) {
552 			m_freem(m);
553 			mtx_unlock(&(sc)->sc_ro_mtx);
554 			ifp->if_oerrors++;
555 			return ENETUNREACH;
556 		}
557 	}
558 	cached_route = &sc->sc_ro;
559 
560 sendit:
561 	M_SETFIB(m, sc->sc_fibnum);
562 	ifp->if_opackets++;
563 	error = ip_output(m, NULL, cached_route, 0, NULL, NULL);
564 
565 	if (cached_route != NULL)
566 		mtx_unlock(&(sc)->sc_ro_mtx);
567 	return error;
568 }
569 
570 static int
571 isrfc1918addr(in)
572 	struct in_addr *in;
573 {
574 	/*
575 	 * returns 1 if private address range:
576 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
577 	 */
578 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
579 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
580 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
581 		return 1;
582 
583 	return 0;
584 }
585 
586 static int
587 stf_checkaddr4(sc, in, inifp)
588 	struct stf_softc *sc;
589 	struct in_addr *in;
590 	struct ifnet *inifp;	/* incoming interface */
591 {
592 	INIT_VNET_INET(curvnet);
593 	struct in_ifaddr *ia4;
594 
595 	/*
596 	 * reject packets with the following address:
597 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
598 	 */
599 	if (IN_MULTICAST(ntohl(in->s_addr)))
600 		return -1;
601 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
602 	case 0: case 127: case 255:
603 		return -1;
604 	}
605 
606 	/*
607 	 * reject packets with private address range.
608 	 * (requirement from RFC3056 section 2 1st paragraph)
609 	 */
610 	if (isrfc1918addr(in))
611 		return -1;
612 
613 	/*
614 	 * reject packets with broadcast
615 	 */
616 	for (ia4 = TAILQ_FIRST(&V_in_ifaddrhead);
617 	     ia4;
618 	     ia4 = TAILQ_NEXT(ia4, ia_link))
619 	{
620 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
621 			continue;
622 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
623 			return -1;
624 	}
625 
626 	/*
627 	 * perform ingress filter
628 	 */
629 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
630 		struct sockaddr_in sin;
631 		struct rtentry *rt;
632 
633 		bzero(&sin, sizeof(sin));
634 		sin.sin_family = AF_INET;
635 		sin.sin_len = sizeof(struct sockaddr_in);
636 		sin.sin_addr = *in;
637 		rt = rtalloc1_fib((struct sockaddr *)&sin, 0,
638 		    0UL, sc->sc_fibnum);
639 		if (!rt || rt->rt_ifp != inifp) {
640 #if 0
641 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
642 			    "due to ingress filter\n", if_name(STF2IFP(sc)),
643 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
644 #endif
645 			if (rt)
646 				RTFREE_LOCKED(rt);
647 			return -1;
648 		}
649 		RTFREE_LOCKED(rt);
650 	}
651 
652 	return 0;
653 }
654 
655 static int
656 stf_checkaddr6(sc, in6, inifp)
657 	struct stf_softc *sc;
658 	struct in6_addr *in6;
659 	struct ifnet *inifp;	/* incoming interface */
660 {
661 	/*
662 	 * check 6to4 addresses
663 	 */
664 	if (IN6_IS_ADDR_6TO4(in6)) {
665 		struct in_addr in4;
666 		bcopy(GET_V4(in6), &in4, sizeof(in4));
667 		return stf_checkaddr4(sc, &in4, inifp);
668 	}
669 
670 	/*
671 	 * reject anything that look suspicious.  the test is implemented
672 	 * in ip6_input too, but we check here as well to
673 	 * (1) reject bad packets earlier, and
674 	 * (2) to be safe against future ip6_input change.
675 	 */
676 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
677 		return -1;
678 
679 	return 0;
680 }
681 
682 void
683 in_stf_input(m, off)
684 	struct mbuf *m;
685 	int off;
686 {
687 	int proto;
688 	struct stf_softc *sc;
689 	struct ip *ip;
690 	struct ip6_hdr *ip6;
691 	u_int8_t otos, itos;
692 	struct ifnet *ifp;
693 
694 	proto = mtod(m, struct ip *)->ip_p;
695 
696 	if (proto != IPPROTO_IPV6) {
697 		m_freem(m);
698 		return;
699 	}
700 
701 	ip = mtod(m, struct ip *);
702 
703 	sc = (struct stf_softc *)encap_getarg(m);
704 
705 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
706 		m_freem(m);
707 		return;
708 	}
709 
710 	ifp = STF2IFP(sc);
711 
712 #ifdef MAC
713 	mac_ifnet_create_mbuf(ifp, m);
714 #endif
715 
716 	/*
717 	 * perform sanity check against outer src/dst.
718 	 * for source, perform ingress filter as well.
719 	 */
720 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
721 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
722 		m_freem(m);
723 		return;
724 	}
725 
726 	otos = ip->ip_tos;
727 	m_adj(m, off);
728 
729 	if (m->m_len < sizeof(*ip6)) {
730 		m = m_pullup(m, sizeof(*ip6));
731 		if (!m)
732 			return;
733 	}
734 	ip6 = mtod(m, struct ip6_hdr *);
735 
736 	/*
737 	 * perform sanity check against inner src/dst.
738 	 * for source, perform ingress filter as well.
739 	 */
740 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
741 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
742 		m_freem(m);
743 		return;
744 	}
745 
746 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
747 	if ((ifp->if_flags & IFF_LINK1) != 0)
748 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
749 	else
750 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
751 	ip6->ip6_flow &= ~htonl(0xff << 20);
752 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
753 
754 	m->m_pkthdr.rcvif = ifp;
755 
756 	if (bpf_peers_present(ifp->if_bpf)) {
757 		/*
758 		 * We need to prepend the address family as
759 		 * a four byte field.  Cons up a dummy header
760 		 * to pacify bpf.  This is safe because bpf
761 		 * will only read from the mbuf (i.e., it won't
762 		 * try to free it or keep a pointer a to it).
763 		 */
764 		u_int32_t af = AF_INET6;
765 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
766 	}
767 
768 	/*
769 	 * Put the packet to the network layer input queue according to the
770 	 * specified address family.
771 	 * See net/if_gif.c for possible issues with packet processing
772 	 * reorder due to extra queueing.
773 	 */
774 	ifp->if_ipackets++;
775 	ifp->if_ibytes += m->m_pkthdr.len;
776 	netisr_dispatch(NETISR_IPV6, m);
777 }
778 
779 /* ARGSUSED */
780 static void
781 stf_rtrequest(cmd, rt, info)
782 	int cmd;
783 	struct rtentry *rt;
784 	struct rt_addrinfo *info;
785 {
786 	RT_LOCK_ASSERT(rt);
787 	rt->rt_rmx.rmx_mtu = IPV6_MMTU;
788 }
789 
790 static int
791 stf_ioctl(ifp, cmd, data)
792 	struct ifnet *ifp;
793 	u_long cmd;
794 	caddr_t data;
795 {
796 	struct ifaddr *ifa;
797 	struct ifreq *ifr;
798 	struct sockaddr_in6 *sin6;
799 	struct in_addr addr;
800 	int error;
801 
802 	error = 0;
803 	switch (cmd) {
804 	case SIOCSIFADDR:
805 		ifa = (struct ifaddr *)data;
806 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
807 			error = EAFNOSUPPORT;
808 			break;
809 		}
810 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
811 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
812 			error = EINVAL;
813 			break;
814 		}
815 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
816 		if (isrfc1918addr(&addr)) {
817 			error = EINVAL;
818 			break;
819 		}
820 
821 		ifa->ifa_rtrequest = stf_rtrequest;
822 		ifp->if_flags |= IFF_UP;
823 		break;
824 
825 	case SIOCADDMULTI:
826 	case SIOCDELMULTI:
827 		ifr = (struct ifreq *)data;
828 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
829 			;
830 		else
831 			error = EAFNOSUPPORT;
832 		break;
833 
834 	default:
835 		error = EINVAL;
836 		break;
837 	}
838 
839 	return error;
840 }
841