xref: /freebsd/sys/net/if_stf.c (revision ff0ba87247820afbdfdc1b307c803f7923d0e4d3)
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 
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/socket.h>
83 #include <sys/sockio.h>
84 #include <sys/mbuf.h>
85 #include <sys/errno.h>
86 #include <sys/kernel.h>
87 #include <sys/module.h>
88 #include <sys/protosw.h>
89 #include <sys/proc.h>
90 #include <sys/queue.h>
91 #include <sys/sysctl.h>
92 #include <machine/cpu.h>
93 
94 #include <sys/malloc.h>
95 
96 #include <net/if.h>
97 #include <net/if_var.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 #include <net/vnet.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 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
126 
127 static int stf_permit_rfc1918 = 0;
128 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
129     &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
130 
131 #define STFUNIT		0
132 
133 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
134 
135 /*
136  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
137  * struct in_addr *; use bcopy() instead.
138  */
139 #define GET_V4(x)	(&(x)->s6_addr16[1])
140 
141 struct stf_softc {
142 	struct ifnet	*sc_ifp;
143 	struct mtx	sc_ro_mtx;
144 	u_int	sc_fibnum;
145 	const struct encaptab *encap_cookie;
146 };
147 #define STF2IFP(sc)	((sc)->sc_ifp)
148 
149 static const char stfname[] = "stf";
150 
151 /*
152  * Note that mutable fields in the softc are not currently locked.
153  * We do lock sc_ro in stf_output though.
154  */
155 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
156 static const int ip_stf_ttl = 40;
157 
158 extern  struct domain inetdomain;
159 struct protosw in_stf_protosw = {
160 	.pr_type =		SOCK_RAW,
161 	.pr_domain =		&inetdomain,
162 	.pr_protocol =		IPPROTO_IPV6,
163 	.pr_flags =		PR_ATOMIC|PR_ADDR,
164 	.pr_input =		in_stf_input,
165 	.pr_output =		rip_output,
166 	.pr_ctloutput =		rip_ctloutput,
167 	.pr_usrreqs =		&rip_usrreqs
168 };
169 
170 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
171 
172 static int stfmodevent(module_t, int, void *);
173 static int stf_encapcheck(const struct mbuf *, int, int, void *);
174 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
175 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
176 	struct route *);
177 static int isrfc1918addr(struct in_addr *);
178 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
179 	struct ifnet *);
180 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
181 	struct ifnet *);
182 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
183 
184 static int stf_clone_match(struct if_clone *, const char *);
185 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
186 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
187 static struct if_clone *stf_cloner;
188 
189 static int
190 stf_clone_match(struct if_clone *ifc, const char *name)
191 {
192 	int i;
193 
194 	for(i = 0; stfnames[i] != NULL; i++) {
195 		if (strcmp(stfnames[i], name) == 0)
196 			return (1);
197 	}
198 
199 	return (0);
200 }
201 
202 static int
203 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
204 {
205 	int err, unit;
206 	struct stf_softc *sc;
207 	struct ifnet *ifp;
208 
209 	/*
210 	 * We can only have one unit, but since unit allocation is
211 	 * already locked, we use it to keep from allocating extra
212 	 * interfaces.
213 	 */
214 	unit = STFUNIT;
215 	err = ifc_alloc_unit(ifc, &unit);
216 	if (err != 0)
217 		return (err);
218 
219 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
220 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
221 	if (ifp == NULL) {
222 		free(sc, M_STF);
223 		ifc_free_unit(ifc, unit);
224 		return (ENOSPC);
225 	}
226 	ifp->if_softc = sc;
227 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
228 
229 	/*
230 	 * Set the name manually rather then using if_initname because
231 	 * we don't conform to the default naming convention for interfaces.
232 	 */
233 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
234 	ifp->if_dname = stfname;
235 	ifp->if_dunit = IF_DUNIT_NONE;
236 
237 	mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
238 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
239 	    stf_encapcheck, &in_stf_protosw, sc);
240 	if (sc->encap_cookie == NULL) {
241 		if_printf(ifp, "attach failed\n");
242 		free(sc, M_STF);
243 		ifc_free_unit(ifc, unit);
244 		return (ENOMEM);
245 	}
246 
247 	ifp->if_mtu    = IPV6_MMTU;
248 	ifp->if_ioctl  = stf_ioctl;
249 	ifp->if_output = stf_output;
250 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
251 	if_attach(ifp);
252 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
253 	return (0);
254 }
255 
256 static int
257 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
258 {
259 	struct stf_softc *sc = ifp->if_softc;
260 	int err;
261 
262 	err = encap_detach(sc->encap_cookie);
263 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
264 	mtx_destroy(&(sc)->sc_ro_mtx);
265 	bpfdetach(ifp);
266 	if_detach(ifp);
267 	if_free(ifp);
268 
269 	free(sc, M_STF);
270 	ifc_free_unit(ifc, STFUNIT);
271 
272 	return (0);
273 }
274 
275 static int
276 stfmodevent(mod, type, data)
277 	module_t mod;
278 	int type;
279 	void *data;
280 {
281 
282 	switch (type) {
283 	case MOD_LOAD:
284 		stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
285 		    stf_clone_create, stf_clone_destroy);
286 		break;
287 	case MOD_UNLOAD:
288 		if_clone_detach(stf_cloner);
289 		break;
290 	default:
291 		return (EOPNOTSUPP);
292 	}
293 
294 	return (0);
295 }
296 
297 static moduledata_t stf_mod = {
298 	"if_stf",
299 	stfmodevent,
300 	0
301 };
302 
303 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
304 
305 static int
306 stf_encapcheck(m, off, proto, arg)
307 	const struct mbuf *m;
308 	int off;
309 	int proto;
310 	void *arg;
311 {
312 	struct ip ip;
313 	struct stf_softc *sc;
314 	struct in_addr a, b, mask;
315 	struct in6_addr addr6, mask6;
316 
317 	sc = (struct stf_softc *)arg;
318 	if (sc == NULL)
319 		return 0;
320 
321 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
322 		return 0;
323 
324 	/* IFF_LINK0 means "no decapsulation" */
325 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
326 		return 0;
327 
328 	if (proto != IPPROTO_IPV6)
329 		return 0;
330 
331 	/* LINTED const cast */
332 	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
333 
334 	if (ip.ip_v != 4)
335 		return 0;
336 
337 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
338 		return (0);
339 
340 	/*
341 	 * check if IPv4 dst matches the IPv4 address derived from the
342 	 * local 6to4 address.
343 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
344 	 */
345 	if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
346 		return 0;
347 
348 	/*
349 	 * check if IPv4 src matches the IPv4 address derived from the
350 	 * local 6to4 address masked by prefixmask.
351 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
352 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
353 	 */
354 	bzero(&a, sizeof(a));
355 	bcopy(GET_V4(&addr6), &a, sizeof(a));
356 	bcopy(GET_V4(&mask6), &mask, sizeof(mask));
357 	a.s_addr &= mask.s_addr;
358 	b = ip.ip_src;
359 	b.s_addr &= mask.s_addr;
360 	if (a.s_addr != b.s_addr)
361 		return 0;
362 
363 	/* stf interface makes single side match only */
364 	return 32;
365 }
366 
367 static int
368 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
369 {
370 	struct ifaddr *ia;
371 	struct in_ifaddr *ia4;
372 	struct in6_ifaddr *ia6;
373 	struct sockaddr_in6 *sin6;
374 	struct in_addr in;
375 
376 	if_addr_rlock(ifp);
377 	TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
378 		if (ia->ifa_addr->sa_family != AF_INET6)
379 			continue;
380 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
381 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
382 			continue;
383 
384 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
385 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
386 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
387 				break;
388 		if (ia4 == NULL)
389 			continue;
390 
391 		ia6 = (struct in6_ifaddr *)ia;
392 
393 		*addr = sin6->sin6_addr;
394 		*mask = ia6->ia_prefixmask.sin6_addr;
395 		if_addr_runlock(ifp);
396 		return (0);
397 	}
398 	if_addr_runlock(ifp);
399 
400 	return (ENOENT);
401 }
402 
403 static int
404 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
405 	struct route *ro)
406 {
407 	struct stf_softc *sc;
408 	const struct sockaddr_in6 *dst6;
409 	struct in_addr in4;
410 	const void *ptr;
411 	u_int8_t tos;
412 	struct ip *ip;
413 	struct ip6_hdr *ip6;
414 	struct in6_addr addr6, mask6;
415 	int error;
416 
417 #ifdef MAC
418 	error = mac_ifnet_check_transmit(ifp, m);
419 	if (error) {
420 		m_freem(m);
421 		return (error);
422 	}
423 #endif
424 
425 	sc = ifp->if_softc;
426 	dst6 = (const struct sockaddr_in6 *)dst;
427 
428 	/* just in case */
429 	if ((ifp->if_flags & IFF_UP) == 0) {
430 		m_freem(m);
431 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
432 		return ENETDOWN;
433 	}
434 
435 	/*
436 	 * If we don't have an ip4 address that match my inner ip6 address,
437 	 * we shouldn't generate output.  Without this check, we'll end up
438 	 * using wrong IPv4 source.
439 	 */
440 	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
441 		m_freem(m);
442 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
443 		return ENETDOWN;
444 	}
445 
446 	if (m->m_len < sizeof(*ip6)) {
447 		m = m_pullup(m, sizeof(*ip6));
448 		if (!m) {
449 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
450 			return ENOBUFS;
451 		}
452 	}
453 	ip6 = mtod(m, struct ip6_hdr *);
454 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
455 
456 	/*
457 	 * Pickup the right outer dst addr from the list of candidates.
458 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
459 	 */
460 	ptr = NULL;
461 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
462 		ptr = GET_V4(&ip6->ip6_dst);
463 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
464 		ptr = GET_V4(&dst6->sin6_addr);
465 	else {
466 		m_freem(m);
467 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
468 		return ENETUNREACH;
469 	}
470 	bcopy(ptr, &in4, sizeof(in4));
471 
472 	if (bpf_peers_present(ifp->if_bpf)) {
473 		/*
474 		 * We need to prepend the address family as
475 		 * a four byte field.  Cons up a dummy header
476 		 * to pacify bpf.  This is safe because bpf
477 		 * will only read from the mbuf (i.e., it won't
478 		 * try to free it or keep a pointer a to it).
479 		 */
480 		u_int af = AF_INET6;
481 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
482 	}
483 
484 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
485 	if (m && m->m_len < sizeof(struct ip))
486 		m = m_pullup(m, sizeof(struct ip));
487 	if (m == NULL) {
488 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
489 		return ENOBUFS;
490 	}
491 	ip = mtod(m, struct ip *);
492 
493 	bzero(ip, sizeof(*ip));
494 
495 	bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src));
496 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
497 	ip->ip_p = IPPROTO_IPV6;
498 	ip->ip_ttl = ip_stf_ttl;
499 	ip->ip_len = htons(m->m_pkthdr.len);
500 	if (ifp->if_flags & IFF_LINK1)
501 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
502 	else
503 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
504 
505 	M_SETFIB(m, sc->sc_fibnum);
506 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
507 	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
508 
509 	return error;
510 }
511 
512 static int
513 isrfc1918addr(in)
514 	struct in_addr *in;
515 {
516 	/*
517 	 * returns 1 if private address range:
518 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
519 	 */
520 	if (stf_permit_rfc1918 == 0 && (
521 	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
522 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
523 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
524 		return 1;
525 
526 	return 0;
527 }
528 
529 static int
530 stf_checkaddr4(sc, in, inifp)
531 	struct stf_softc *sc;
532 	struct in_addr *in;
533 	struct ifnet *inifp;	/* incoming interface */
534 {
535 	struct in_ifaddr *ia4;
536 
537 	/*
538 	 * reject packets with the following address:
539 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
540 	 */
541 	if (IN_MULTICAST(ntohl(in->s_addr)))
542 		return -1;
543 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
544 	case 0: case 127: case 255:
545 		return -1;
546 	}
547 
548 	/*
549 	 * reject packets with private address range.
550 	 * (requirement from RFC3056 section 2 1st paragraph)
551 	 */
552 	if (isrfc1918addr(in))
553 		return -1;
554 
555 	/*
556 	 * reject packets with broadcast
557 	 */
558 	IN_IFADDR_RLOCK();
559 	TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
560 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
561 			continue;
562 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
563 			IN_IFADDR_RUNLOCK();
564 			return -1;
565 		}
566 	}
567 	IN_IFADDR_RUNLOCK();
568 
569 	/*
570 	 * perform ingress filter
571 	 */
572 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
573 		struct sockaddr_in sin;
574 		struct rtentry *rt;
575 
576 		bzero(&sin, sizeof(sin));
577 		sin.sin_family = AF_INET;
578 		sin.sin_len = sizeof(struct sockaddr_in);
579 		sin.sin_addr = *in;
580 		rt = rtalloc1_fib((struct sockaddr *)&sin, 0,
581 		    0UL, sc->sc_fibnum);
582 		if (!rt || rt->rt_ifp != inifp) {
583 #if 0
584 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
585 			    "due to ingress filter\n", if_name(STF2IFP(sc)),
586 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
587 #endif
588 			if (rt)
589 				RTFREE_LOCKED(rt);
590 			return -1;
591 		}
592 		RTFREE_LOCKED(rt);
593 	}
594 
595 	return 0;
596 }
597 
598 static int
599 stf_checkaddr6(sc, in6, inifp)
600 	struct stf_softc *sc;
601 	struct in6_addr *in6;
602 	struct ifnet *inifp;	/* incoming interface */
603 {
604 	/*
605 	 * check 6to4 addresses
606 	 */
607 	if (IN6_IS_ADDR_6TO4(in6)) {
608 		struct in_addr in4;
609 		bcopy(GET_V4(in6), &in4, sizeof(in4));
610 		return stf_checkaddr4(sc, &in4, inifp);
611 	}
612 
613 	/*
614 	 * reject anything that look suspicious.  the test is implemented
615 	 * in ip6_input too, but we check here as well to
616 	 * (1) reject bad packets earlier, and
617 	 * (2) to be safe against future ip6_input change.
618 	 */
619 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
620 		return -1;
621 
622 	return 0;
623 }
624 
625 int
626 in_stf_input(struct mbuf **mp, int *offp, int proto)
627 {
628 	struct stf_softc *sc;
629 	struct ip *ip;
630 	struct ip6_hdr *ip6;
631 	struct mbuf *m;
632 	u_int8_t otos, itos;
633 	struct ifnet *ifp;
634 	int off;
635 
636 	m = *mp;
637 	off = *offp;
638 
639 	if (proto != IPPROTO_IPV6) {
640 		m_freem(m);
641 		return (IPPROTO_DONE);
642 	}
643 
644 	ip = mtod(m, struct ip *);
645 
646 	sc = (struct stf_softc *)encap_getarg(m);
647 
648 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
649 		m_freem(m);
650 		return (IPPROTO_DONE);
651 	}
652 
653 	ifp = STF2IFP(sc);
654 
655 #ifdef MAC
656 	mac_ifnet_create_mbuf(ifp, m);
657 #endif
658 
659 	/*
660 	 * perform sanity check against outer src/dst.
661 	 * for source, perform ingress filter as well.
662 	 */
663 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
664 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
665 		m_freem(m);
666 		return (IPPROTO_DONE);
667 	}
668 
669 	otos = ip->ip_tos;
670 	m_adj(m, off);
671 
672 	if (m->m_len < sizeof(*ip6)) {
673 		m = m_pullup(m, sizeof(*ip6));
674 		if (!m)
675 			return (IPPROTO_DONE);
676 	}
677 	ip6 = mtod(m, struct ip6_hdr *);
678 
679 	/*
680 	 * perform sanity check against inner src/dst.
681 	 * for source, perform ingress filter as well.
682 	 */
683 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
684 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
685 		m_freem(m);
686 		return (IPPROTO_DONE);
687 	}
688 
689 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
690 	if ((ifp->if_flags & IFF_LINK1) != 0)
691 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
692 	else
693 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
694 	ip6->ip6_flow &= ~htonl(0xff << 20);
695 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
696 
697 	m->m_pkthdr.rcvif = ifp;
698 
699 	if (bpf_peers_present(ifp->if_bpf)) {
700 		/*
701 		 * We need to prepend the address family as
702 		 * a four byte field.  Cons up a dummy header
703 		 * to pacify bpf.  This is safe because bpf
704 		 * will only read from the mbuf (i.e., it won't
705 		 * try to free it or keep a pointer a to it).
706 		 */
707 		u_int32_t af = AF_INET6;
708 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
709 	}
710 
711 	/*
712 	 * Put the packet to the network layer input queue according to the
713 	 * specified address family.
714 	 * See net/if_gif.c for possible issues with packet processing
715 	 * reorder due to extra queueing.
716 	 */
717 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
718 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
719 	M_SETFIB(m, ifp->if_fib);
720 	netisr_dispatch(NETISR_IPV6, m);
721 	return (IPPROTO_DONE);
722 }
723 
724 static int
725 stf_ioctl(ifp, cmd, data)
726 	struct ifnet *ifp;
727 	u_long cmd;
728 	caddr_t data;
729 {
730 	struct ifaddr *ifa;
731 	struct ifreq *ifr;
732 	struct sockaddr_in6 *sin6;
733 	struct in_addr addr;
734 	int error, mtu;
735 
736 	error = 0;
737 	switch (cmd) {
738 	case SIOCSIFADDR:
739 		ifa = (struct ifaddr *)data;
740 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
741 			error = EAFNOSUPPORT;
742 			break;
743 		}
744 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
745 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
746 			error = EINVAL;
747 			break;
748 		}
749 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
750 		if (isrfc1918addr(&addr)) {
751 			error = EINVAL;
752 			break;
753 		}
754 
755 		ifp->if_flags |= IFF_UP;
756 		break;
757 
758 	case SIOCADDMULTI:
759 	case SIOCDELMULTI:
760 		ifr = (struct ifreq *)data;
761 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
762 			;
763 		else
764 			error = EAFNOSUPPORT;
765 		break;
766 
767 	case SIOCGIFMTU:
768 		break;
769 
770 	case SIOCSIFMTU:
771 		ifr = (struct ifreq *)data;
772 		mtu = ifr->ifr_mtu;
773 		/* RFC 4213 3.2 ideal world MTU */
774 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
775 			return (EINVAL);
776 		ifp->if_mtu = mtu;
777 		break;
778 
779 	default:
780 		error = EINVAL;
781 		break;
782 	}
783 
784 	return error;
785 }
786