xref: /freebsd/sys/net/if_stf.c (revision d8a0fe102c0cfdfcd5b818f850eff09d8536c9bc)
1 /*	$FreeBSD$	*/
2 /*	$KAME: if_stf.c,v 1.73 2001/12/03 11:08:30 keiichi Exp $	*/
3 
4 /*-
5  * SPDX-License-Identifier: BSD-3-Clause
6  *
7  * Copyright (C) 2000 WIDE Project.
8  * All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the project nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 /*
36  * 6to4 interface, based on RFC3056.
37  *
38  * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting.
39  * There is no address mapping defined from IPv6 multicast address to IPv4
40  * address.  Therefore, we do not have IFF_MULTICAST on the interface.
41  *
42  * Due to the lack of address mapping for link-local addresses, we cannot
43  * throw packets toward link-local addresses (fe80::x).  Also, we cannot throw
44  * packets to link-local multicast addresses (ff02::x).
45  *
46  * Here are interesting symptoms due to the lack of link-local address:
47  *
48  * Unicast routing exchange:
49  * - RIPng: Impossible.  Uses link-local multicast packet toward ff02::9,
50  *   and link-local addresses as nexthop.
51  * - OSPFv6: Impossible.  OSPFv6 assumes that there's link-local address
52  *   assigned to the link, and makes use of them.  Also, HELLO packets use
53  *   link-local multicast addresses (ff02::5 and ff02::6).
54  * - BGP4+: Maybe.  You can only use global address as nexthop, and global
55  *   address as TCP endpoint address.
56  *
57  * Multicast routing protocols:
58  * - PIM: Hello packet cannot be used to discover adjacent PIM routers.
59  *   Adjacent PIM routers must be configured manually (is it really spec-wise
60  *   correct thing to do?).
61  *
62  * ICMPv6:
63  * - Redirects cannot be used due to the lack of link-local address.
64  *
65  * stf interface does not have, and will not need, a link-local address.
66  * It seems to have no real benefit and does not help the above symptoms much.
67  * Even if we assign link-locals to interface, we cannot really
68  * use link-local unicast/multicast on top of 6to4 cloud (since there's no
69  * encapsulation defined for link-local address), and the above analysis does
70  * not change.  RFC3056 does not mandate the assignment of link-local address
71  * either.
72  *
73  * 6to4 interface has security issues.  Refer to
74  * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt
75  * for details.  The code tries to filter out some of malicious packets.
76  * Note that there is no way to be 100% secure.
77  */
78 
79 #include <sys/param.h>
80 #include <sys/systm.h>
81 #include <sys/socket.h>
82 #include <sys/sockio.h>
83 #include <sys/mbuf.h>
84 #include <sys/errno.h>
85 #include <sys/kernel.h>
86 #include <sys/lock.h>
87 #include <sys/module.h>
88 #include <sys/protosw.h>
89 #include <sys/proc.h>
90 #include <sys/queue.h>
91 #include <sys/rmlock.h>
92 #include <sys/sysctl.h>
93 #include <machine/cpu.h>
94 
95 #include <sys/malloc.h>
96 
97 #include <net/if.h>
98 #include <net/if_var.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/vnet.h>
104 
105 #include <netinet/in.h>
106 #include <netinet/in_fib.h>
107 #include <netinet/in_systm.h>
108 #include <netinet/ip.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/in_var.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 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
127 
128 static int stf_permit_rfc1918 = 0;
129 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
130     &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
131 
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)	(&(x)->s6_addr16[1])
141 
142 struct stf_softc {
143 	struct ifnet	*sc_ifp;
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 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
152 static const int ip_stf_ttl = 40;
153 
154 extern  struct domain inetdomain;
155 static int in_stf_input(struct mbuf **, int *, int);
156 static struct protosw in_stf_protosw = {
157 	.pr_type =		SOCK_RAW,
158 	.pr_domain =		&inetdomain,
159 	.pr_protocol =		IPPROTO_IPV6,
160 	.pr_flags =		PR_ATOMIC|PR_ADDR,
161 	.pr_input =		in_stf_input,
162 	.pr_output =		rip_output,
163 	.pr_ctloutput =		rip_ctloutput,
164 	.pr_usrreqs =		&rip_usrreqs
165 };
166 
167 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
168 
169 static int stfmodevent(module_t, int, void *);
170 static int stf_encapcheck(const struct mbuf *, int, int, void *);
171 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
172 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
173 	struct route *);
174 static int isrfc1918addr(struct in_addr *);
175 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
176 	struct ifnet *);
177 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
178 	struct ifnet *);
179 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
180 
181 static int stf_clone_match(struct if_clone *, const char *);
182 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
183 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
184 static struct if_clone *stf_cloner;
185 
186 static int
187 stf_clone_match(struct if_clone *ifc, const char *name)
188 {
189 	int i;
190 
191 	for(i = 0; stfnames[i] != NULL; i++) {
192 		if (strcmp(stfnames[i], name) == 0)
193 			return (1);
194 	}
195 
196 	return (0);
197 }
198 
199 static int
200 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
201 {
202 	char *dp;
203 	int err, unit, wildcard;
204 	struct stf_softc *sc;
205 	struct ifnet *ifp;
206 
207 	err = ifc_name2unit(name, &unit);
208 	if (err != 0)
209 		return (err);
210 	wildcard = (unit < 0);
211 
212 	/*
213 	 * We can only have one unit, but since unit allocation is
214 	 * already locked, we use it to keep from allocating extra
215 	 * interfaces.
216 	 */
217 	unit = STFUNIT;
218 	err = ifc_alloc_unit(ifc, &unit);
219 	if (err != 0)
220 		return (err);
221 
222 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
223 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
224 	if (ifp == NULL) {
225 		free(sc, M_STF);
226 		ifc_free_unit(ifc, unit);
227 		return (ENOSPC);
228 	}
229 	ifp->if_softc = sc;
230 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
231 
232 	/*
233 	 * Set the name manually rather then using if_initname because
234 	 * we don't conform to the default naming convention for interfaces.
235 	 * In the wildcard case, we need to update the name.
236 	 */
237 	if (wildcard) {
238 		for (dp = name; *dp != '\0'; dp++);
239 		if (snprintf(dp, len - (dp-name), "%d", unit) >
240 		    len - (dp-name) - 1) {
241 			/*
242 			 * This can only be a programmer error and
243 			 * there's no straightforward way to recover if
244 			 * it happens.
245 			 */
246 			panic("if_clone_create(): interface name too long");
247 		}
248 	}
249 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
250 	ifp->if_dname = stfname;
251 	ifp->if_dunit = IF_DUNIT_NONE;
252 
253 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
254 	    stf_encapcheck, &in_stf_protosw, sc);
255 	if (sc->encap_cookie == NULL) {
256 		if_printf(ifp, "attach failed\n");
257 		free(sc, M_STF);
258 		ifc_free_unit(ifc, unit);
259 		return (ENOMEM);
260 	}
261 
262 	ifp->if_mtu    = IPV6_MMTU;
263 	ifp->if_ioctl  = stf_ioctl;
264 	ifp->if_output = stf_output;
265 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
266 	if_attach(ifp);
267 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
268 	return (0);
269 }
270 
271 static int
272 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
273 {
274 	struct stf_softc *sc = ifp->if_softc;
275 	int err;
276 
277 	err = encap_detach(sc->encap_cookie);
278 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
279 	bpfdetach(ifp);
280 	if_detach(ifp);
281 	if_free(ifp);
282 
283 	free(sc, M_STF);
284 	ifc_free_unit(ifc, STFUNIT);
285 
286 	return (0);
287 }
288 
289 static int
290 stfmodevent(module_t mod, int type, void *data)
291 {
292 
293 	switch (type) {
294 	case MOD_LOAD:
295 		stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
296 		    stf_clone_create, stf_clone_destroy);
297 		break;
298 	case MOD_UNLOAD:
299 		if_clone_detach(stf_cloner);
300 		break;
301 	default:
302 		return (EOPNOTSUPP);
303 	}
304 
305 	return (0);
306 }
307 
308 static moduledata_t stf_mod = {
309 	"if_stf",
310 	stfmodevent,
311 	0
312 };
313 
314 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
315 
316 static int
317 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
318 {
319 	struct ip ip;
320 	struct stf_softc *sc;
321 	struct in_addr a, b, mask;
322 	struct in6_addr addr6, mask6;
323 
324 	sc = (struct stf_softc *)arg;
325 	if (sc == NULL)
326 		return 0;
327 
328 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
329 		return 0;
330 
331 	/* IFF_LINK0 means "no decapsulation" */
332 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
333 		return 0;
334 
335 	if (proto != IPPROTO_IPV6)
336 		return 0;
337 
338 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
339 
340 	if (ip.ip_v != 4)
341 		return 0;
342 
343 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
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(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
352 		return 0;
353 
354 	/*
355 	 * check if IPv4 src matches the IPv4 address derived from the
356 	 * local 6to4 address masked by prefixmask.
357 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
358 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
359 	 */
360 	bzero(&a, sizeof(a));
361 	bcopy(GET_V4(&addr6), &a, sizeof(a));
362 	bcopy(GET_V4(&mask6), &mask, sizeof(mask));
363 	a.s_addr &= mask.s_addr;
364 	b = ip.ip_src;
365 	b.s_addr &= mask.s_addr;
366 	if (a.s_addr != b.s_addr)
367 		return 0;
368 
369 	/* stf interface makes single side match only */
370 	return 32;
371 }
372 
373 static int
374 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
375 {
376 	struct ifaddr *ia;
377 	struct in_ifaddr *ia4;
378 	struct in6_ifaddr *ia6;
379 	struct sockaddr_in6 *sin6;
380 	struct in_addr in;
381 
382 	if_addr_rlock(ifp);
383 	TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
384 		if (ia->ifa_addr->sa_family != AF_INET6)
385 			continue;
386 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
387 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
388 			continue;
389 
390 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
391 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
392 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
393 				break;
394 		if (ia4 == NULL)
395 			continue;
396 
397 		ia6 = (struct in6_ifaddr *)ia;
398 
399 		*addr = sin6->sin6_addr;
400 		*mask = ia6->ia_prefixmask.sin6_addr;
401 		if_addr_runlock(ifp);
402 		return (0);
403 	}
404 	if_addr_runlock(ifp);
405 
406 	return (ENOENT);
407 }
408 
409 static int
410 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
411     struct route *ro)
412 {
413 	struct stf_softc *sc;
414 	const struct sockaddr_in6 *dst6;
415 	struct in_addr in4;
416 	const void *ptr;
417 	u_int8_t tos;
418 	struct ip *ip;
419 	struct ip6_hdr *ip6;
420 	struct in6_addr addr6, mask6;
421 	int error;
422 
423 #ifdef MAC
424 	error = mac_ifnet_check_transmit(ifp, m);
425 	if (error) {
426 		m_freem(m);
427 		return (error);
428 	}
429 #endif
430 
431 	sc = ifp->if_softc;
432 	dst6 = (const struct sockaddr_in6 *)dst;
433 
434 	/* just in case */
435 	if ((ifp->if_flags & IFF_UP) == 0) {
436 		m_freem(m);
437 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
438 		return ENETDOWN;
439 	}
440 
441 	/*
442 	 * If we don't have an ip4 address that match my inner ip6 address,
443 	 * we shouldn't generate output.  Without this check, we'll end up
444 	 * using wrong IPv4 source.
445 	 */
446 	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
447 		m_freem(m);
448 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
449 		return ENETDOWN;
450 	}
451 
452 	if (m->m_len < sizeof(*ip6)) {
453 		m = m_pullup(m, sizeof(*ip6));
454 		if (!m) {
455 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
456 			return ENOBUFS;
457 		}
458 	}
459 	ip6 = mtod(m, struct ip6_hdr *);
460 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
461 
462 	/*
463 	 * Pickup the right outer dst addr from the list of candidates.
464 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
465 	 */
466 	ptr = NULL;
467 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
468 		ptr = GET_V4(&ip6->ip6_dst);
469 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
470 		ptr = GET_V4(&dst6->sin6_addr);
471 	else {
472 		m_freem(m);
473 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
474 		return ENETUNREACH;
475 	}
476 	bcopy(ptr, &in4, sizeof(in4));
477 
478 	if (bpf_peers_present(ifp->if_bpf)) {
479 		/*
480 		 * We need to prepend the address family as
481 		 * a four byte field.  Cons up a dummy header
482 		 * to pacify bpf.  This is safe because bpf
483 		 * will only read from the mbuf (i.e., it won't
484 		 * try to free it or keep a pointer a to it).
485 		 */
486 		u_int af = AF_INET6;
487 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
488 	}
489 
490 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
491 	if (m == NULL) {
492 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
493 		return ENOBUFS;
494 	}
495 	ip = mtod(m, struct ip *);
496 
497 	bzero(ip, sizeof(*ip));
498 
499 	bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src));
500 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
501 	ip->ip_p = IPPROTO_IPV6;
502 	ip->ip_ttl = ip_stf_ttl;
503 	ip->ip_len = htons(m->m_pkthdr.len);
504 	if (ifp->if_flags & IFF_LINK1)
505 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
506 	else
507 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
508 
509 	M_SETFIB(m, sc->sc_fibnum);
510 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
511 	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
512 
513 	return error;
514 }
515 
516 static int
517 isrfc1918addr(struct in_addr *in)
518 {
519 	/*
520 	 * returns 1 if private address range:
521 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
522 	 */
523 	if (stf_permit_rfc1918 == 0 && (
524 	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
525 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
526 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
527 		return 1;
528 
529 	return 0;
530 }
531 
532 static int
533 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
534 {
535 	struct rm_priotracker in_ifa_tracker;
536 	struct in_ifaddr *ia4;
537 
538 	/*
539 	 * reject packets with the following address:
540 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
541 	 */
542 	if (IN_MULTICAST(ntohl(in->s_addr)))
543 		return -1;
544 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
545 	case 0: case 127: case 255:
546 		return -1;
547 	}
548 
549 	/*
550 	 * reject packets with private address range.
551 	 * (requirement from RFC3056 section 2 1st paragraph)
552 	 */
553 	if (isrfc1918addr(in))
554 		return -1;
555 
556 	/*
557 	 * reject packets with broadcast
558 	 */
559 	IN_IFADDR_RLOCK(&in_ifa_tracker);
560 	TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
561 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
562 			continue;
563 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
564 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
565 			return -1;
566 		}
567 	}
568 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
569 
570 	/*
571 	 * perform ingress filter
572 	 */
573 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
574 		struct nhop4_basic nh4;
575 
576 		if (fib4_lookup_nh_basic(sc->sc_fibnum, *in, 0, 0, &nh4) != 0)
577 			return (-1);
578 
579 		if (nh4.nh_ifp != inifp)
580 			return (-1);
581 	}
582 
583 	return 0;
584 }
585 
586 static int
587 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
588 {
589 	/*
590 	 * check 6to4 addresses
591 	 */
592 	if (IN6_IS_ADDR_6TO4(in6)) {
593 		struct in_addr in4;
594 		bcopy(GET_V4(in6), &in4, sizeof(in4));
595 		return stf_checkaddr4(sc, &in4, inifp);
596 	}
597 
598 	/*
599 	 * reject anything that look suspicious.  the test is implemented
600 	 * in ip6_input too, but we check here as well to
601 	 * (1) reject bad packets earlier, and
602 	 * (2) to be safe against future ip6_input change.
603 	 */
604 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
605 		return -1;
606 
607 	return 0;
608 }
609 
610 static int
611 in_stf_input(struct mbuf **mp, int *offp, int proto)
612 {
613 	struct stf_softc *sc;
614 	struct ip *ip;
615 	struct ip6_hdr *ip6;
616 	struct mbuf *m;
617 	u_int8_t otos, itos;
618 	struct ifnet *ifp;
619 	int off;
620 
621 	m = *mp;
622 	off = *offp;
623 
624 	if (proto != IPPROTO_IPV6) {
625 		m_freem(m);
626 		return (IPPROTO_DONE);
627 	}
628 
629 	ip = mtod(m, struct ip *);
630 
631 	sc = (struct stf_softc *)encap_getarg(m);
632 
633 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
634 		m_freem(m);
635 		return (IPPROTO_DONE);
636 	}
637 
638 	ifp = STF2IFP(sc);
639 
640 #ifdef MAC
641 	mac_ifnet_create_mbuf(ifp, m);
642 #endif
643 
644 	/*
645 	 * perform sanity check against outer src/dst.
646 	 * for source, perform ingress filter as well.
647 	 */
648 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
649 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
650 		m_freem(m);
651 		return (IPPROTO_DONE);
652 	}
653 
654 	otos = ip->ip_tos;
655 	m_adj(m, off);
656 
657 	if (m->m_len < sizeof(*ip6)) {
658 		m = m_pullup(m, sizeof(*ip6));
659 		if (!m)
660 			return (IPPROTO_DONE);
661 	}
662 	ip6 = mtod(m, struct ip6_hdr *);
663 
664 	/*
665 	 * perform sanity check against inner src/dst.
666 	 * for source, perform ingress filter as well.
667 	 */
668 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
669 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
670 		m_freem(m);
671 		return (IPPROTO_DONE);
672 	}
673 
674 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
675 	if ((ifp->if_flags & IFF_LINK1) != 0)
676 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
677 	else
678 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
679 	ip6->ip6_flow &= ~htonl(0xff << 20);
680 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
681 
682 	m->m_pkthdr.rcvif = ifp;
683 
684 	if (bpf_peers_present(ifp->if_bpf)) {
685 		/*
686 		 * We need to prepend the address family as
687 		 * a four byte field.  Cons up a dummy header
688 		 * to pacify bpf.  This is safe because bpf
689 		 * will only read from the mbuf (i.e., it won't
690 		 * try to free it or keep a pointer a to it).
691 		 */
692 		u_int32_t af = AF_INET6;
693 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
694 	}
695 
696 	/*
697 	 * Put the packet to the network layer input queue according to the
698 	 * specified address family.
699 	 * See net/if_gif.c for possible issues with packet processing
700 	 * reorder due to extra queueing.
701 	 */
702 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
703 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
704 	M_SETFIB(m, ifp->if_fib);
705 	netisr_dispatch(NETISR_IPV6, m);
706 	return (IPPROTO_DONE);
707 }
708 
709 static int
710 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
711 {
712 	struct ifaddr *ifa;
713 	struct ifreq *ifr;
714 	struct sockaddr_in6 *sin6;
715 	struct in_addr addr;
716 	int error, mtu;
717 
718 	error = 0;
719 	switch (cmd) {
720 	case SIOCSIFADDR:
721 		ifa = (struct ifaddr *)data;
722 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
723 			error = EAFNOSUPPORT;
724 			break;
725 		}
726 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
727 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
728 			error = EINVAL;
729 			break;
730 		}
731 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
732 		if (isrfc1918addr(&addr)) {
733 			error = EINVAL;
734 			break;
735 		}
736 
737 		ifp->if_flags |= IFF_UP;
738 		break;
739 
740 	case SIOCADDMULTI:
741 	case SIOCDELMULTI:
742 		ifr = (struct ifreq *)data;
743 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
744 			;
745 		else
746 			error = EAFNOSUPPORT;
747 		break;
748 
749 	case SIOCGIFMTU:
750 		break;
751 
752 	case SIOCSIFMTU:
753 		ifr = (struct ifreq *)data;
754 		mtu = ifr->ifr_mtu;
755 		/* RFC 4213 3.2 ideal world MTU */
756 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
757 			return (EINVAL);
758 		ifp->if_mtu = mtu;
759 		break;
760 
761 	default:
762 		error = EINVAL;
763 		break;
764 	}
765 
766 	return error;
767 }
768