xref: /freebsd/sys/net/if_stf.c (revision 13014ca04aad1931d41958b56f71a2c65b9a7a2c) 
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 
111 #include <netinet/ip6.h>
112 #include <netinet6/ip6_var.h>
113 #include <netinet6/in6_var.h>
114 #include <netinet/ip_ecn.h>
115 
116 #include <netinet/ip_encap.h>
117 
118 #include <machine/stdarg.h>
119 
120 #include <net/bpf.h>
121 
122 #include <security/mac/mac_framework.h>
123 
124 SYSCTL_DECL(_net_link);
125 SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
126 
127 static int stf_route_cache = 1;
128 SYSCTL_INT(_net_link_stf, OID_AUTO, route_cache, CTLFLAG_RW,
129     &stf_route_cache, 0, "Caching of IPv4 routes for 6to4 Output");
130 
131 #define STFNAME		"stf"
132 #define STFUNIT		0
133 
134 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
135 
136 /*
137  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
138  * struct in_addr *; use bcopy() instead.
139  */
140 #define GET_V4(x)	((caddr_t)(&(x)->s6_addr16[1]))
141 
142 struct stf_softc {
143 	struct ifnet	*sc_ifp;
144 	union {
145 		struct route  __sc_ro4;
146 		struct route_in6 __sc_ro6; /* just for safety */
147 	} __sc_ro46;
148 #define sc_ro	__sc_ro46.__sc_ro4
149 	struct mtx	sc_ro_mtx;
150 	u_int	sc_fibnum;
151 	const struct encaptab *encap_cookie;
152 };
153 #define STF2IFP(sc)	((sc)->sc_ifp)
154 
155 /*
156  * Note that mutable fields in the softc are not currently locked.
157  * We do lock sc_ro in stf_output though.
158  */
159 static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
160 static const int ip_stf_ttl = 40;
161 
162 extern  struct domain inetdomain;
163 struct protosw in_stf_protosw = {
164 	.pr_type =		SOCK_RAW,
165 	.pr_domain =		&inetdomain,
166 	.pr_protocol =		IPPROTO_IPV6,
167 	.pr_flags =		PR_ATOMIC|PR_ADDR,
168 	.pr_input =		in_stf_input,
169 	.pr_output =		(pr_output_t *)rip_output,
170 	.pr_ctloutput =		rip_ctloutput,
171 	.pr_usrreqs =		&rip_usrreqs
172 };
173 
174 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
175 
176 static int stfmodevent(module_t, int, void *);
177 static int stf_encapcheck(const struct mbuf *, int, int, void *);
178 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
179 static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
180 	struct rtentry *);
181 static int isrfc1918addr(struct in_addr *);
182 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
183 	struct ifnet *);
184 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
185 	struct ifnet *);
186 static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
187 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
188 
189 static int stf_clone_match(struct if_clone *, const char *);
190 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
191 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
192 struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0,
193     NULL, stf_clone_match, stf_clone_create, stf_clone_destroy);
194 
195 static int
196 stf_clone_match(struct if_clone *ifc, const char *name)
197 {
198 	int i;
199 
200 	for(i = 0; stfnames[i] != NULL; i++) {
201 		if (strcmp(stfnames[i], name) == 0)
202 			return (1);
203 	}
204 
205 	return (0);
206 }
207 
208 static int
209 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
210 {
211 	int err, unit;
212 	struct stf_softc *sc;
213 	struct ifnet *ifp;
214 
215 	/*
216 	 * We can only have one unit, but since unit allocation is
217 	 * already locked, we use it to keep from allocating extra
218 	 * interfaces.
219 	 */
220 	unit = STFUNIT;
221 	err = ifc_alloc_unit(ifc, &unit);
222 	if (err != 0)
223 		return (err);
224 
225 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
226 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
227 	if (ifp == NULL) {
228 		free(sc, M_STF);
229 		ifc_free_unit(ifc, unit);
230 		return (ENOSPC);
231 	}
232 	ifp->if_softc = sc;
233 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
234 
235 	/*
236 	 * Set the name manually rather then using if_initname because
237 	 * we don't conform to the default naming convention for interfaces.
238 	 */
239 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
240 	ifp->if_dname = ifc->ifc_name;
241 	ifp->if_dunit = IF_DUNIT_NONE;
242 
243 	mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
244 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
245 	    stf_encapcheck, &in_stf_protosw, sc);
246 	if (sc->encap_cookie == NULL) {
247 		if_printf(ifp, "attach failed\n");
248 		free(sc, M_STF);
249 		ifc_free_unit(ifc, unit);
250 		return (ENOMEM);
251 	}
252 
253 	ifp->if_mtu    = IPV6_MMTU;
254 	ifp->if_ioctl  = stf_ioctl;
255 	ifp->if_output = stf_output;
256 	ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
257 	if_attach(ifp);
258 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
259 	return (0);
260 }
261 
262 static int
263 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
264 {
265 	struct stf_softc *sc = ifp->if_softc;
266 	int err;
267 
268 	err = encap_detach(sc->encap_cookie);
269 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
270 	mtx_destroy(&(sc)->sc_ro_mtx);
271 	bpfdetach(ifp);
272 	if_detach(ifp);
273 	if_free(ifp);
274 
275 	free(sc, M_STF);
276 	ifc_free_unit(ifc, STFUNIT);
277 
278 	return (0);
279 }
280 
281 static int
282 stfmodevent(mod, type, data)
283 	module_t mod;
284 	int type;
285 	void *data;
286 {
287 
288 	switch (type) {
289 	case MOD_LOAD:
290 		if_clone_attach(&stf_cloner);
291 		break;
292 	case MOD_UNLOAD:
293 		if_clone_detach(&stf_cloner);
294 		break;
295 	default:
296 		return (EOPNOTSUPP);
297 	}
298 
299 	return (0);
300 }
301 
302 static moduledata_t stf_mod = {
303 	"if_stf",
304 	stfmodevent,
305 	0
306 };
307 
308 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
309 
310 static int
311 stf_encapcheck(m, off, proto, arg)
312 	const struct mbuf *m;
313 	int off;
314 	int proto;
315 	void *arg;
316 {
317 	struct ip ip;
318 	struct in6_ifaddr *ia6;
319 	struct stf_softc *sc;
320 	struct in_addr a, b, mask;
321 
322 	sc = (struct stf_softc *)arg;
323 	if (sc == NULL)
324 		return 0;
325 
326 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
327 		return 0;
328 
329 	/* IFF_LINK0 means "no decapsulation" */
330 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
331 		return 0;
332 
333 	if (proto != IPPROTO_IPV6)
334 		return 0;
335 
336 	/* LINTED const cast */
337 	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
338 
339 	if (ip.ip_v != 4)
340 		return 0;
341 
342 	ia6 = stf_getsrcifa6(STF2IFP(sc));
343 	if (ia6 == NULL)
344 		return 0;
345 
346 	/*
347 	 * check if IPv4 dst matches the IPv4 address derived from the
348 	 * local 6to4 address.
349 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
350 	 */
351 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
352 	    sizeof(ip.ip_dst)) != 0)
353 		return 0;
354 
355 	/*
356 	 * check if IPv4 src matches the IPv4 address derived from the
357 	 * local 6to4 address masked by prefixmask.
358 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
359 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
360 	 */
361 	bzero(&a, sizeof(a));
362 	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
363 	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
364 	a.s_addr &= mask.s_addr;
365 	b = ip.ip_src;
366 	b.s_addr &= mask.s_addr;
367 	if (a.s_addr != b.s_addr)
368 		return 0;
369 
370 	/* stf interface makes single side match only */
371 	return 32;
372 }
373 
374 static struct in6_ifaddr *
375 stf_getsrcifa6(ifp)
376 	struct ifnet *ifp;
377 {
378 	INIT_VNET_INET(ifp->if_vnet);
379 	struct ifaddr *ia;
380 	struct in_ifaddr *ia4;
381 	struct sockaddr_in6 *sin6;
382 	struct in_addr in;
383 
384 	TAILQ_FOREACH(ia, &ifp->if_addrlist, ifa_list) {
385 		if (ia->ifa_addr->sa_family != AF_INET6)
386 			continue;
387 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
388 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
389 			continue;
390 
391 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
392 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
393 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
394 				break;
395 		if (ia4 == NULL)
396 			continue;
397 
398 		return (struct in6_ifaddr *)ia;
399 	}
400 
401 	return NULL;
402 }
403 
404 static int
405 stf_output(ifp, m, dst, rt)
406 	struct ifnet *ifp;
407 	struct mbuf *m;
408 	struct sockaddr *dst;
409 	struct rtentry *rt;
410 {
411 	struct stf_softc *sc;
412 	struct sockaddr_in6 *dst6;
413 	struct route *cached_route;
414 	struct in_addr in4;
415 	caddr_t ptr;
416 	struct sockaddr_in *dst4;
417 	u_int8_t tos;
418 	struct ip *ip;
419 	struct ip6_hdr *ip6;
420 	struct in6_ifaddr *ia6;
421 	u_int32_t af;
422 	int error;
423 
424 #ifdef MAC
425 	error = mac_ifnet_check_transmit(ifp, m);
426 	if (error) {
427 		m_freem(m);
428 		return (error);
429 	}
430 #endif
431 
432 	sc = ifp->if_softc;
433 	dst6 = (struct sockaddr_in6 *)dst;
434 
435 	/* just in case */
436 	if ((ifp->if_flags & IFF_UP) == 0) {
437 		m_freem(m);
438 		ifp->if_oerrors++;
439 		return ENETDOWN;
440 	}
441 
442 	/*
443 	 * If we don't have an ip4 address that match my inner ip6 address,
444 	 * we shouldn't generate output.  Without this check, we'll end up
445 	 * using wrong IPv4 source.
446 	 */
447 	ia6 = stf_getsrcifa6(ifp);
448 	if (ia6 == NULL) {
449 		m_freem(m);
450 		ifp->if_oerrors++;
451 		return ENETDOWN;
452 	}
453 
454 	if (m->m_len < sizeof(*ip6)) {
455 		m = m_pullup(m, sizeof(*ip6));
456 		if (!m) {
457 			ifp->if_oerrors++;
458 			return ENOBUFS;
459 		}
460 	}
461 	ip6 = mtod(m, struct ip6_hdr *);
462 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
463 
464 	/*
465 	 * BPF writes need to be handled specially.
466 	 * This is a null operation, nothing here checks dst->sa_family.
467 	 */
468 	if (dst->sa_family == AF_UNSPEC) {
469 		bcopy(dst->sa_data, &af, sizeof(af));
470 		dst->sa_family = af;
471 	}
472 
473 	/*
474 	 * Pickup the right outer dst addr from the list of candidates.
475 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
476 	 */
477 	ptr = NULL;
478 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
479 		ptr = GET_V4(&ip6->ip6_dst);
480 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
481 		ptr = GET_V4(&dst6->sin6_addr);
482 	else {
483 		m_freem(m);
484 		ifp->if_oerrors++;
485 		return ENETUNREACH;
486 	}
487 	bcopy(ptr, &in4, sizeof(in4));
488 
489 	if (bpf_peers_present(ifp->if_bpf)) {
490 		/*
491 		 * We need to prepend the address family as
492 		 * a four byte field.  Cons up a dummy header
493 		 * to pacify bpf.  This is safe because bpf
494 		 * will only read from the mbuf (i.e., it won't
495 		 * try to free it or keep a pointer a to it).
496 		 */
497 		af = AF_INET6;
498 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
499 	}
500 
501 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
502 	if (m && m->m_len < sizeof(struct ip))
503 		m = m_pullup(m, sizeof(struct ip));
504 	if (m == NULL) {
505 		ifp->if_oerrors++;
506 		return ENOBUFS;
507 	}
508 	ip = mtod(m, struct ip *);
509 
510 	bzero(ip, sizeof(*ip));
511 
512 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
513 	    &ip->ip_src, sizeof(ip->ip_src));
514 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
515 	ip->ip_p = IPPROTO_IPV6;
516 	ip->ip_ttl = ip_stf_ttl;
517 	ip->ip_len = m->m_pkthdr.len;	/*host order*/
518 	if (ifp->if_flags & IFF_LINK1)
519 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
520 	else
521 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
522 
523 	if (!stf_route_cache) {
524 		cached_route = NULL;
525 		goto sendit;
526 	}
527 
528 	/*
529 	 * Do we have a cached route?
530 	 */
531 	mtx_lock(&(sc)->sc_ro_mtx);
532 	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
533 	if (dst4->sin_family != AF_INET ||
534 	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
535 		/* cache route doesn't match */
536 		dst4->sin_family = AF_INET;
537 		dst4->sin_len = sizeof(struct sockaddr_in);
538 		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
539 		if (sc->sc_ro.ro_rt) {
540 			RTFREE(sc->sc_ro.ro_rt);
541 			sc->sc_ro.ro_rt = NULL;
542 		}
543 	}
544 
545 	if (sc->sc_ro.ro_rt == NULL) {
546 		rtalloc_fib(&sc->sc_ro, sc->sc_fibnum);
547 		if (sc->sc_ro.ro_rt == NULL) {
548 			m_freem(m);
549 			mtx_unlock(&(sc)->sc_ro_mtx);
550 			ifp->if_oerrors++;
551 			return ENETUNREACH;
552 		}
553 	}
554 	cached_route = &sc->sc_ro;
555 
556 sendit:
557 	M_SETFIB(m, sc->sc_fibnum);
558 	ifp->if_opackets++;
559 	error = ip_output(m, NULL, cached_route, 0, NULL, NULL);
560 
561 	if (cached_route != NULL)
562 		mtx_unlock(&(sc)->sc_ro_mtx);
563 	return error;
564 }
565 
566 static int
567 isrfc1918addr(in)
568 	struct in_addr *in;
569 {
570 	/*
571 	 * returns 1 if private address range:
572 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
573 	 */
574 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
575 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
576 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
577 		return 1;
578 
579 	return 0;
580 }
581 
582 static int
583 stf_checkaddr4(sc, in, inifp)
584 	struct stf_softc *sc;
585 	struct in_addr *in;
586 	struct ifnet *inifp;	/* incoming interface */
587 {
588 	INIT_VNET_INET(curvnet);
589 	struct in_ifaddr *ia4;
590 
591 	/*
592 	 * reject packets with the following address:
593 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
594 	 */
595 	if (IN_MULTICAST(ntohl(in->s_addr)))
596 		return -1;
597 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
598 	case 0: case 127: case 255:
599 		return -1;
600 	}
601 
602 	/*
603 	 * reject packets with private address range.
604 	 * (requirement from RFC3056 section 2 1st paragraph)
605 	 */
606 	if (isrfc1918addr(in))
607 		return -1;
608 
609 	/*
610 	 * reject packets with broadcast
611 	 */
612 	for (ia4 = TAILQ_FIRST(&V_in_ifaddrhead);
613 	     ia4;
614 	     ia4 = TAILQ_NEXT(ia4, ia_link))
615 	{
616 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
617 			continue;
618 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
619 			return -1;
620 	}
621 
622 	/*
623 	 * perform ingress filter
624 	 */
625 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
626 		struct sockaddr_in sin;
627 		struct rtentry *rt;
628 
629 		bzero(&sin, sizeof(sin));
630 		sin.sin_family = AF_INET;
631 		sin.sin_len = sizeof(struct sockaddr_in);
632 		sin.sin_addr = *in;
633 		rt = rtalloc1_fib((struct sockaddr *)&sin, 0,
634 		    0UL, sc->sc_fibnum);
635 		if (!rt || rt->rt_ifp != inifp) {
636 #if 0
637 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
638 			    "due to ingress filter\n", if_name(STF2IFP(sc)),
639 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
640 #endif
641 			if (rt)
642 				RTFREE_LOCKED(rt);
643 			return -1;
644 		}
645 		RTFREE_LOCKED(rt);
646 	}
647 
648 	return 0;
649 }
650 
651 static int
652 stf_checkaddr6(sc, in6, inifp)
653 	struct stf_softc *sc;
654 	struct in6_addr *in6;
655 	struct ifnet *inifp;	/* incoming interface */
656 {
657 	/*
658 	 * check 6to4 addresses
659 	 */
660 	if (IN6_IS_ADDR_6TO4(in6)) {
661 		struct in_addr in4;
662 		bcopy(GET_V4(in6), &in4, sizeof(in4));
663 		return stf_checkaddr4(sc, &in4, inifp);
664 	}
665 
666 	/*
667 	 * reject anything that look suspicious.  the test is implemented
668 	 * in ip6_input too, but we check here as well to
669 	 * (1) reject bad packets earlier, and
670 	 * (2) to be safe against future ip6_input change.
671 	 */
672 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
673 		return -1;
674 
675 	return 0;
676 }
677 
678 void
679 in_stf_input(m, off)
680 	struct mbuf *m;
681 	int off;
682 {
683 	int proto;
684 	struct stf_softc *sc;
685 	struct ip *ip;
686 	struct ip6_hdr *ip6;
687 	u_int8_t otos, itos;
688 	struct ifnet *ifp;
689 
690 	proto = mtod(m, struct ip *)->ip_p;
691 
692 	if (proto != IPPROTO_IPV6) {
693 		m_freem(m);
694 		return;
695 	}
696 
697 	ip = mtod(m, struct ip *);
698 
699 	sc = (struct stf_softc *)encap_getarg(m);
700 
701 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
702 		m_freem(m);
703 		return;
704 	}
705 
706 	ifp = STF2IFP(sc);
707 
708 #ifdef MAC
709 	mac_ifnet_create_mbuf(ifp, m);
710 #endif
711 
712 	/*
713 	 * perform sanity check against outer src/dst.
714 	 * for source, perform ingress filter as well.
715 	 */
716 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
717 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
718 		m_freem(m);
719 		return;
720 	}
721 
722 	otos = ip->ip_tos;
723 	m_adj(m, off);
724 
725 	if (m->m_len < sizeof(*ip6)) {
726 		m = m_pullup(m, sizeof(*ip6));
727 		if (!m)
728 			return;
729 	}
730 	ip6 = mtod(m, struct ip6_hdr *);
731 
732 	/*
733 	 * perform sanity check against inner src/dst.
734 	 * for source, perform ingress filter as well.
735 	 */
736 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
737 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
738 		m_freem(m);
739 		return;
740 	}
741 
742 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
743 	if ((ifp->if_flags & IFF_LINK1) != 0)
744 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
745 	else
746 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
747 	ip6->ip6_flow &= ~htonl(0xff << 20);
748 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
749 
750 	m->m_pkthdr.rcvif = ifp;
751 
752 	if (bpf_peers_present(ifp->if_bpf)) {
753 		/*
754 		 * We need to prepend the address family as
755 		 * a four byte field.  Cons up a dummy header
756 		 * to pacify bpf.  This is safe because bpf
757 		 * will only read from the mbuf (i.e., it won't
758 		 * try to free it or keep a pointer a to it).
759 		 */
760 		u_int32_t af = AF_INET6;
761 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
762 	}
763 
764 	/*
765 	 * Put the packet to the network layer input queue according to the
766 	 * specified address family.
767 	 * See net/if_gif.c for possible issues with packet processing
768 	 * reorder due to extra queueing.
769 	 */
770 	ifp->if_ipackets++;
771 	ifp->if_ibytes += m->m_pkthdr.len;
772 	netisr_dispatch(NETISR_IPV6, m);
773 }
774 
775 /* ARGSUSED */
776 static void
777 stf_rtrequest(cmd, rt, info)
778 	int cmd;
779 	struct rtentry *rt;
780 	struct rt_addrinfo *info;
781 {
782 	RT_LOCK_ASSERT(rt);
783 	rt->rt_rmx.rmx_mtu = IPV6_MMTU;
784 }
785 
786 static int
787 stf_ioctl(ifp, cmd, data)
788 	struct ifnet *ifp;
789 	u_long cmd;
790 	caddr_t data;
791 {
792 	struct ifaddr *ifa;
793 	struct ifreq *ifr;
794 	struct sockaddr_in6 *sin6;
795 	struct in_addr addr;
796 	int error;
797 
798 	error = 0;
799 	switch (cmd) {
800 	case SIOCSIFADDR:
801 		ifa = (struct ifaddr *)data;
802 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
803 			error = EAFNOSUPPORT;
804 			break;
805 		}
806 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
807 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
808 			error = EINVAL;
809 			break;
810 		}
811 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
812 		if (isrfc1918addr(&addr)) {
813 			error = EINVAL;
814 			break;
815 		}
816 
817 		ifa->ifa_rtrequest = stf_rtrequest;
818 		ifp->if_flags |= IFF_UP;
819 		break;
820 
821 	case SIOCADDMULTI:
822 	case SIOCDELMULTI:
823 		ifr = (struct ifreq *)data;
824 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
825 			;
826 		else
827 			error = EAFNOSUPPORT;
828 		break;
829 
830 	default:
831 		error = EINVAL;
832 		break;
833 	}
834 
835 	return error;
836 }
837