xref: /freebsd/sys/net/if_stf.c (revision d93a896e)
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 <sys/param.h>
78 #include <sys/systm.h>
79 #include <sys/socket.h>
80 #include <sys/sockio.h>
81 #include <sys/mbuf.h>
82 #include <sys/errno.h>
83 #include <sys/kernel.h>
84 #include <sys/lock.h>
85 #include <sys/module.h>
86 #include <sys/protosw.h>
87 #include <sys/proc.h>
88 #include <sys/queue.h>
89 #include <sys/rmlock.h>
90 #include <sys/sysctl.h>
91 #include <machine/cpu.h>
92 
93 #include <sys/malloc.h>
94 
95 #include <net/if.h>
96 #include <net/if_var.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/vnet.h>
102 
103 #include <netinet/in.h>
104 #include <netinet/in_fib.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip_var.h>
108 #include <netinet/in_var.h>
109 
110 #include <netinet/ip6.h>
111 #include <netinet6/ip6_var.h>
112 #include <netinet6/in6_var.h>
113 #include <netinet/ip_ecn.h>
114 
115 #include <netinet/ip_encap.h>
116 
117 #include <machine/stdarg.h>
118 
119 #include <net/bpf.h>
120 
121 #include <security/mac/mac_framework.h>
122 
123 SYSCTL_DECL(_net_link);
124 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
125 
126 static int stf_permit_rfc1918 = 0;
127 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
128     &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
129 
130 #define STFUNIT		0
131 
132 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
133 
134 /*
135  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
136  * struct in_addr *; use bcopy() instead.
137  */
138 #define GET_V4(x)	(&(x)->s6_addr16[1])
139 
140 struct stf_softc {
141 	struct ifnet	*sc_ifp;
142 	u_int	sc_fibnum;
143 	const struct encaptab *encap_cookie;
144 };
145 #define STF2IFP(sc)	((sc)->sc_ifp)
146 
147 static const char stfname[] = "stf";
148 
149 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
150 static const int ip_stf_ttl = 40;
151 
152 extern  struct domain inetdomain;
153 static int in_stf_input(struct mbuf **, int *, int);
154 static struct protosw in_stf_protosw = {
155 	.pr_type =		SOCK_RAW,
156 	.pr_domain =		&inetdomain,
157 	.pr_protocol =		IPPROTO_IPV6,
158 	.pr_flags =		PR_ATOMIC|PR_ADDR,
159 	.pr_input =		in_stf_input,
160 	.pr_output =		rip_output,
161 	.pr_ctloutput =		rip_ctloutput,
162 	.pr_usrreqs =		&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 int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
170 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
171 	struct route *);
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 int stf_ioctl(struct ifnet *, u_long, caddr_t);
178 
179 static int stf_clone_match(struct if_clone *, const char *);
180 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
181 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
182 static struct if_clone *stf_cloner;
183 
184 static int
185 stf_clone_match(struct if_clone *ifc, const char *name)
186 {
187 	int i;
188 
189 	for(i = 0; stfnames[i] != NULL; i++) {
190 		if (strcmp(stfnames[i], name) == 0)
191 			return (1);
192 	}
193 
194 	return (0);
195 }
196 
197 static int
198 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
199 {
200 	char *dp;
201 	int err, unit, wildcard;
202 	struct stf_softc *sc;
203 	struct ifnet *ifp;
204 
205 	err = ifc_name2unit(name, &unit);
206 	if (err != 0)
207 		return (err);
208 	wildcard = (unit < 0);
209 
210 	/*
211 	 * We can only have one unit, but since unit allocation is
212 	 * already locked, we use it to keep from allocating extra
213 	 * interfaces.
214 	 */
215 	unit = STFUNIT;
216 	err = ifc_alloc_unit(ifc, &unit);
217 	if (err != 0)
218 		return (err);
219 
220 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
221 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
222 	if (ifp == NULL) {
223 		free(sc, M_STF);
224 		ifc_free_unit(ifc, unit);
225 		return (ENOSPC);
226 	}
227 	ifp->if_softc = sc;
228 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
229 
230 	/*
231 	 * Set the name manually rather then using if_initname because
232 	 * we don't conform to the default naming convention for interfaces.
233 	 * In the wildcard case, we need to update the name.
234 	 */
235 	if (wildcard) {
236 		for (dp = name; *dp != '\0'; dp++);
237 		if (snprintf(dp, len - (dp-name), "%d", unit) >
238 		    len - (dp-name) - 1) {
239 			/*
240 			 * This can only be a programmer error and
241 			 * there's no straightforward way to recover if
242 			 * it happens.
243 			 */
244 			panic("if_clone_create(): interface name too long");
245 		}
246 	}
247 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
248 	ifp->if_dname = stfname;
249 	ifp->if_dunit = IF_DUNIT_NONE;
250 
251 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
252 	    stf_encapcheck, &in_stf_protosw, sc);
253 	if (sc->encap_cookie == NULL) {
254 		if_printf(ifp, "attach failed\n");
255 		free(sc, M_STF);
256 		ifc_free_unit(ifc, unit);
257 		return (ENOMEM);
258 	}
259 
260 	ifp->if_mtu    = IPV6_MMTU;
261 	ifp->if_ioctl  = stf_ioctl;
262 	ifp->if_output = stf_output;
263 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
264 	if_attach(ifp);
265 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
266 	return (0);
267 }
268 
269 static int
270 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
271 {
272 	struct stf_softc *sc = ifp->if_softc;
273 	int err;
274 
275 	err = encap_detach(sc->encap_cookie);
276 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
277 	bpfdetach(ifp);
278 	if_detach(ifp);
279 	if_free(ifp);
280 
281 	free(sc, M_STF);
282 	ifc_free_unit(ifc, STFUNIT);
283 
284 	return (0);
285 }
286 
287 static int
288 stfmodevent(module_t mod, int type, void *data)
289 {
290 
291 	switch (type) {
292 	case MOD_LOAD:
293 		stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
294 		    stf_clone_create, stf_clone_destroy);
295 		break;
296 	case MOD_UNLOAD:
297 		if_clone_detach(stf_cloner);
298 		break;
299 	default:
300 		return (EOPNOTSUPP);
301 	}
302 
303 	return (0);
304 }
305 
306 static moduledata_t stf_mod = {
307 	"if_stf",
308 	stfmodevent,
309 	0
310 };
311 
312 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
313 
314 static int
315 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
316 {
317 	struct ip ip;
318 	struct stf_softc *sc;
319 	struct in_addr a, b, mask;
320 	struct in6_addr addr6, mask6;
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 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
337 
338 	if (ip.ip_v != 4)
339 		return 0;
340 
341 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
342 		return (0);
343 
344 	/*
345 	 * check if IPv4 dst matches the IPv4 address derived from the
346 	 * local 6to4 address.
347 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
348 	 */
349 	if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
350 		return 0;
351 
352 	/*
353 	 * check if IPv4 src matches the IPv4 address derived from the
354 	 * local 6to4 address masked by prefixmask.
355 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
356 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
357 	 */
358 	bzero(&a, sizeof(a));
359 	bcopy(GET_V4(&addr6), &a, sizeof(a));
360 	bcopy(GET_V4(&mask6), &mask, sizeof(mask));
361 	a.s_addr &= mask.s_addr;
362 	b = ip.ip_src;
363 	b.s_addr &= mask.s_addr;
364 	if (a.s_addr != b.s_addr)
365 		return 0;
366 
367 	/* stf interface makes single side match only */
368 	return 32;
369 }
370 
371 static int
372 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
373 {
374 	struct ifaddr *ia;
375 	struct in_ifaddr *ia4;
376 	struct in6_ifaddr *ia6;
377 	struct sockaddr_in6 *sin6;
378 	struct in_addr in;
379 
380 	if_addr_rlock(ifp);
381 	TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
382 		if (ia->ifa_addr->sa_family != AF_INET6)
383 			continue;
384 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
385 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
386 			continue;
387 
388 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
389 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
390 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
391 				break;
392 		if (ia4 == NULL)
393 			continue;
394 
395 		ia6 = (struct in6_ifaddr *)ia;
396 
397 		*addr = sin6->sin6_addr;
398 		*mask = ia6->ia_prefixmask.sin6_addr;
399 		if_addr_runlock(ifp);
400 		return (0);
401 	}
402 	if_addr_runlock(ifp);
403 
404 	return (ENOENT);
405 }
406 
407 static int
408 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
409     struct route *ro)
410 {
411 	struct stf_softc *sc;
412 	const struct sockaddr_in6 *dst6;
413 	struct in_addr in4;
414 	const void *ptr;
415 	u_int8_t tos;
416 	struct ip *ip;
417 	struct ip6_hdr *ip6;
418 	struct in6_addr addr6, mask6;
419 	int error;
420 
421 #ifdef MAC
422 	error = mac_ifnet_check_transmit(ifp, m);
423 	if (error) {
424 		m_freem(m);
425 		return (error);
426 	}
427 #endif
428 
429 	sc = ifp->if_softc;
430 	dst6 = (const struct sockaddr_in6 *)dst;
431 
432 	/* just in case */
433 	if ((ifp->if_flags & IFF_UP) == 0) {
434 		m_freem(m);
435 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
436 		return ENETDOWN;
437 	}
438 
439 	/*
440 	 * If we don't have an ip4 address that match my inner ip6 address,
441 	 * we shouldn't generate output.  Without this check, we'll end up
442 	 * using wrong IPv4 source.
443 	 */
444 	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
445 		m_freem(m);
446 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
447 		return ENETDOWN;
448 	}
449 
450 	if (m->m_len < sizeof(*ip6)) {
451 		m = m_pullup(m, sizeof(*ip6));
452 		if (!m) {
453 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
454 			return ENOBUFS;
455 		}
456 	}
457 	ip6 = mtod(m, struct ip6_hdr *);
458 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
459 
460 	/*
461 	 * Pickup the right outer dst addr from the list of candidates.
462 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
463 	 */
464 	ptr = NULL;
465 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
466 		ptr = GET_V4(&ip6->ip6_dst);
467 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
468 		ptr = GET_V4(&dst6->sin6_addr);
469 	else {
470 		m_freem(m);
471 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
472 		return ENETUNREACH;
473 	}
474 	bcopy(ptr, &in4, sizeof(in4));
475 
476 	if (bpf_peers_present(ifp->if_bpf)) {
477 		/*
478 		 * We need to prepend the address family as
479 		 * a four byte field.  Cons up a dummy header
480 		 * to pacify bpf.  This is safe because bpf
481 		 * will only read from the mbuf (i.e., it won't
482 		 * try to free it or keep a pointer a to it).
483 		 */
484 		u_int af = AF_INET6;
485 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
486 	}
487 
488 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
489 	if (m == NULL) {
490 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
491 		return ENOBUFS;
492 	}
493 	ip = mtod(m, struct ip *);
494 
495 	bzero(ip, sizeof(*ip));
496 
497 	bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src));
498 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
499 	ip->ip_p = IPPROTO_IPV6;
500 	ip->ip_ttl = ip_stf_ttl;
501 	ip->ip_len = htons(m->m_pkthdr.len);
502 	if (ifp->if_flags & IFF_LINK1)
503 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
504 	else
505 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
506 
507 	M_SETFIB(m, sc->sc_fibnum);
508 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
509 	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
510 
511 	return error;
512 }
513 
514 static int
515 isrfc1918addr(struct in_addr *in)
516 {
517 	/*
518 	 * returns 1 if private address range:
519 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
520 	 */
521 	if (stf_permit_rfc1918 == 0 && (
522 	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
523 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
524 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
525 		return 1;
526 
527 	return 0;
528 }
529 
530 static int
531 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
532 {
533 	struct rm_priotracker in_ifa_tracker;
534 	struct in_ifaddr *ia4;
535 
536 	/*
537 	 * reject packets with the following address:
538 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
539 	 */
540 	if (IN_MULTICAST(ntohl(in->s_addr)))
541 		return -1;
542 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
543 	case 0: case 127: case 255:
544 		return -1;
545 	}
546 
547 	/*
548 	 * reject packets with private address range.
549 	 * (requirement from RFC3056 section 2 1st paragraph)
550 	 */
551 	if (isrfc1918addr(in))
552 		return -1;
553 
554 	/*
555 	 * reject packets with broadcast
556 	 */
557 	IN_IFADDR_RLOCK(&in_ifa_tracker);
558 	TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
559 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
560 			continue;
561 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
562 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
563 			return -1;
564 		}
565 	}
566 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
567 
568 	/*
569 	 * perform ingress filter
570 	 */
571 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
572 		struct nhop4_basic nh4;
573 
574 		if (fib4_lookup_nh_basic(sc->sc_fibnum, *in, 0, 0, &nh4) != 0)
575 			return (-1);
576 
577 		if (nh4.nh_ifp != inifp)
578 			return (-1);
579 	}
580 
581 	return 0;
582 }
583 
584 static int
585 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
586 {
587 	/*
588 	 * check 6to4 addresses
589 	 */
590 	if (IN6_IS_ADDR_6TO4(in6)) {
591 		struct in_addr in4;
592 		bcopy(GET_V4(in6), &in4, sizeof(in4));
593 		return stf_checkaddr4(sc, &in4, inifp);
594 	}
595 
596 	/*
597 	 * reject anything that look suspicious.  the test is implemented
598 	 * in ip6_input too, but we check here as well to
599 	 * (1) reject bad packets earlier, and
600 	 * (2) to be safe against future ip6_input change.
601 	 */
602 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
603 		return -1;
604 
605 	return 0;
606 }
607 
608 static int
609 in_stf_input(struct mbuf **mp, int *offp, int proto)
610 {
611 	struct stf_softc *sc;
612 	struct ip *ip;
613 	struct ip6_hdr *ip6;
614 	struct mbuf *m;
615 	u_int8_t otos, itos;
616 	struct ifnet *ifp;
617 	int off;
618 
619 	m = *mp;
620 	off = *offp;
621 
622 	if (proto != IPPROTO_IPV6) {
623 		m_freem(m);
624 		return (IPPROTO_DONE);
625 	}
626 
627 	ip = mtod(m, struct ip *);
628 
629 	sc = (struct stf_softc *)encap_getarg(m);
630 
631 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
632 		m_freem(m);
633 		return (IPPROTO_DONE);
634 	}
635 
636 	ifp = STF2IFP(sc);
637 
638 #ifdef MAC
639 	mac_ifnet_create_mbuf(ifp, m);
640 #endif
641 
642 	/*
643 	 * perform sanity check against outer src/dst.
644 	 * for source, perform ingress filter as well.
645 	 */
646 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
647 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
648 		m_freem(m);
649 		return (IPPROTO_DONE);
650 	}
651 
652 	otos = ip->ip_tos;
653 	m_adj(m, off);
654 
655 	if (m->m_len < sizeof(*ip6)) {
656 		m = m_pullup(m, sizeof(*ip6));
657 		if (!m)
658 			return (IPPROTO_DONE);
659 	}
660 	ip6 = mtod(m, struct ip6_hdr *);
661 
662 	/*
663 	 * perform sanity check against inner src/dst.
664 	 * for source, perform ingress filter as well.
665 	 */
666 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
667 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
668 		m_freem(m);
669 		return (IPPROTO_DONE);
670 	}
671 
672 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
673 	if ((ifp->if_flags & IFF_LINK1) != 0)
674 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
675 	else
676 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
677 	ip6->ip6_flow &= ~htonl(0xff << 20);
678 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
679 
680 	m->m_pkthdr.rcvif = ifp;
681 
682 	if (bpf_peers_present(ifp->if_bpf)) {
683 		/*
684 		 * We need to prepend the address family as
685 		 * a four byte field.  Cons up a dummy header
686 		 * to pacify bpf.  This is safe because bpf
687 		 * will only read from the mbuf (i.e., it won't
688 		 * try to free it or keep a pointer a to it).
689 		 */
690 		u_int32_t af = AF_INET6;
691 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
692 	}
693 
694 	/*
695 	 * Put the packet to the network layer input queue according to the
696 	 * specified address family.
697 	 * See net/if_gif.c for possible issues with packet processing
698 	 * reorder due to extra queueing.
699 	 */
700 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
701 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
702 	M_SETFIB(m, ifp->if_fib);
703 	netisr_dispatch(NETISR_IPV6, m);
704 	return (IPPROTO_DONE);
705 }
706 
707 static int
708 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
709 {
710 	struct ifaddr *ifa;
711 	struct ifreq *ifr;
712 	struct sockaddr_in6 *sin6;
713 	struct in_addr addr;
714 	int error, mtu;
715 
716 	error = 0;
717 	switch (cmd) {
718 	case SIOCSIFADDR:
719 		ifa = (struct ifaddr *)data;
720 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
721 			error = EAFNOSUPPORT;
722 			break;
723 		}
724 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
725 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
726 			error = EINVAL;
727 			break;
728 		}
729 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
730 		if (isrfc1918addr(&addr)) {
731 			error = EINVAL;
732 			break;
733 		}
734 
735 		ifp->if_flags |= IFF_UP;
736 		break;
737 
738 	case SIOCADDMULTI:
739 	case SIOCDELMULTI:
740 		ifr = (struct ifreq *)data;
741 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
742 			;
743 		else
744 			error = EAFNOSUPPORT;
745 		break;
746 
747 	case SIOCGIFMTU:
748 		break;
749 
750 	case SIOCSIFMTU:
751 		ifr = (struct ifreq *)data;
752 		mtu = ifr->ifr_mtu;
753 		/* RFC 4213 3.2 ideal world MTU */
754 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
755 			return (EINVAL);
756 		ifp->if_mtu = mtu;
757 		break;
758 
759 	default:
760 		error = EINVAL;
761 		break;
762 	}
763 
764 	return error;
765 }
766