xref: /freebsd/sys/net/if_stf.c (revision 675aae732d3dd2ec0f12f39df3a1fbfe693a4ebd)
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/proc.h>
89 #include <sys/queue.h>
90 #include <sys/rmlock.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/route/nhop.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 | CTLFLAG_MPSAFE, 0,
127     "6to4 Interface");
128 
129 static int stf_permit_rfc1918 = 0;
130 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN,
131     &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses");
132 
133 #define STFUNIT		0
134 
135 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
136 
137 /*
138  * XXX: Return a pointer with 16-bit aligned.  Don't cast it to
139  * struct in_addr *; use bcopy() instead.
140  */
141 #define GET_V4(x)	(&(x)->s6_addr16[1])
142 
143 struct stf_softc {
144 	struct ifnet	*sc_ifp;
145 	u_int	sc_fibnum;
146 	const struct encaptab *encap_cookie;
147 };
148 #define STF2IFP(sc)	((sc)->sc_ifp)
149 
150 static const char stfname[] = "stf";
151 
152 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
153 static const int ip_stf_ttl = 40;
154 
155 static int in_stf_input(struct mbuf *, int, int, void *);
156 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
157 
158 static int stfmodevent(module_t, int, void *);
159 static int stf_encapcheck(const struct mbuf *, int, int, void *);
160 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *);
161 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *,
162 	struct route *);
163 static int isrfc1918addr(struct in_addr *);
164 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
165 	struct ifnet *);
166 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
167 	struct ifnet *);
168 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
169 
170 static int stf_clone_match(struct if_clone *, const char *);
171 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
172 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
173 static struct if_clone *stf_cloner;
174 
175 static const struct encap_config ipv4_encap_cfg = {
176 	.proto = IPPROTO_IPV6,
177 	.min_length = sizeof(struct ip),
178 	.exact_match = (sizeof(in_addr_t) << 3) + 8,
179 	.check = stf_encapcheck,
180 	.input = in_stf_input
181 };
182 
183 static int
184 stf_clone_match(struct if_clone *ifc, const char *name)
185 {
186 	int i;
187 
188 	for(i = 0; stfnames[i] != NULL; i++) {
189 		if (strcmp(stfnames[i], name) == 0)
190 			return (1);
191 	}
192 
193 	return (0);
194 }
195 
196 static int
197 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
198 {
199 	char *dp;
200 	int err, unit, wildcard;
201 	struct stf_softc *sc;
202 	struct ifnet *ifp;
203 
204 	err = ifc_name2unit(name, &unit);
205 	if (err != 0)
206 		return (err);
207 	wildcard = (unit < 0);
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 	 * In the wildcard case, we need to update the name.
233 	 */
234 	if (wildcard) {
235 		for (dp = name; *dp != '\0'; dp++);
236 		if (snprintf(dp, len - (dp-name), "%d", unit) >
237 		    len - (dp-name) - 1) {
238 			/*
239 			 * This can only be a programmer error and
240 			 * there's no straightforward way to recover if
241 			 * it happens.
242 			 */
243 			panic("if_clone_create(): interface name too long");
244 		}
245 	}
246 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
247 	ifp->if_dname = stfname;
248 	ifp->if_dunit = IF_DUNIT_NONE;
249 
250 	sc->encap_cookie = ip_encap_attach(&ipv4_encap_cfg, sc, M_WAITOK);
251 	if (sc->encap_cookie == NULL) {
252 		if_printf(ifp, "attach failed\n");
253 		free(sc, M_STF);
254 		ifc_free_unit(ifc, unit);
255 		return (ENOMEM);
256 	}
257 
258 	ifp->if_mtu    = IPV6_MMTU;
259 	ifp->if_ioctl  = stf_ioctl;
260 	ifp->if_output = stf_output;
261 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
262 	if_attach(ifp);
263 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
264 	return (0);
265 }
266 
267 static int
268 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
269 {
270 	struct stf_softc *sc = ifp->if_softc;
271 	int err __unused;
272 
273 	err = ip_encap_detach(sc->encap_cookie);
274 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
275 	bpfdetach(ifp);
276 	if_detach(ifp);
277 	if_free(ifp);
278 
279 	free(sc, M_STF);
280 	ifc_free_unit(ifc, STFUNIT);
281 
282 	return (0);
283 }
284 
285 static int
286 stfmodevent(module_t mod, int type, void *data)
287 {
288 
289 	switch (type) {
290 	case MOD_LOAD:
291 		stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
292 		    stf_clone_create, stf_clone_destroy);
293 		break;
294 	case MOD_UNLOAD:
295 		if_clone_detach(stf_cloner);
296 		break;
297 	default:
298 		return (EOPNOTSUPP);
299 	}
300 
301 	return (0);
302 }
303 
304 static moduledata_t stf_mod = {
305 	"if_stf",
306 	stfmodevent,
307 	0
308 };
309 
310 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
311 
312 static int
313 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
314 {
315 	struct ip ip;
316 	struct stf_softc *sc;
317 	struct in_addr a, b, mask;
318 	struct in6_addr addr6, mask6;
319 
320 	sc = (struct stf_softc *)arg;
321 	if (sc == NULL)
322 		return 0;
323 
324 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
325 		return 0;
326 
327 	/* IFF_LINK0 means "no decapsulation" */
328 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
329 		return 0;
330 
331 	if (proto != IPPROTO_IPV6)
332 		return 0;
333 
334 	m_copydata(m, 0, sizeof(ip), (caddr_t)&ip);
335 
336 	if (ip.ip_v != 4)
337 		return 0;
338 
339 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
340 		return (0);
341 
342 	/*
343 	 * check if IPv4 dst matches the IPv4 address derived from the
344 	 * local 6to4 address.
345 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
346 	 */
347 	if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
348 		return 0;
349 
350 	/*
351 	 * check if IPv4 src matches the IPv4 address derived from the
352 	 * local 6to4 address masked by prefixmask.
353 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
354 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
355 	 */
356 	bzero(&a, sizeof(a));
357 	bcopy(GET_V4(&addr6), &a, sizeof(a));
358 	bcopy(GET_V4(&mask6), &mask, sizeof(mask));
359 	a.s_addr &= mask.s_addr;
360 	b = ip.ip_src;
361 	b.s_addr &= mask.s_addr;
362 	if (a.s_addr != b.s_addr)
363 		return 0;
364 
365 	/* stf interface makes single side match only */
366 	return 32;
367 }
368 
369 static int
370 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
371 {
372 	struct rm_priotracker in_ifa_tracker;
373 	struct ifaddr *ia;
374 	struct in_ifaddr *ia4;
375 	struct in6_ifaddr *ia6;
376 	struct sockaddr_in6 *sin6;
377 	struct in_addr in;
378 
379 	NET_EPOCH_ASSERT();
380 
381 	CK_STAILQ_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 		IN_IFADDR_RLOCK(&in_ifa_tracker);
390 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
391 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
392 				break;
393 		IN_IFADDR_RUNLOCK(&in_ifa_tracker);
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 		return (0);
402 	}
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 	CK_STAILQ_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 nhop_object *nh;
573 
574 		NET_EPOCH_ASSERT();
575 		nh = fib4_lookup(sc->sc_fibnum, *in, 0, 0, 0);
576 		if (nh == NULL)
577 			return (-1);
578 
579 		if (nh->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 *m, int off, int proto, void *arg)
612 {
613 	struct stf_softc *sc = arg;
614 	struct ip *ip;
615 	struct ip6_hdr *ip6;
616 	u_int8_t otos, itos;
617 	struct ifnet *ifp;
618 
619 	NET_EPOCH_ASSERT();
620 
621 	if (proto != IPPROTO_IPV6) {
622 		m_freem(m);
623 		return (IPPROTO_DONE);
624 	}
625 
626 	ip = mtod(m, struct ip *);
627 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
628 		m_freem(m);
629 		return (IPPROTO_DONE);
630 	}
631 
632 	ifp = STF2IFP(sc);
633 
634 #ifdef MAC
635 	mac_ifnet_create_mbuf(ifp, m);
636 #endif
637 
638 	/*
639 	 * perform sanity check against outer src/dst.
640 	 * for source, perform ingress filter as well.
641 	 */
642 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
643 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
644 		m_freem(m);
645 		return (IPPROTO_DONE);
646 	}
647 
648 	otos = ip->ip_tos;
649 	m_adj(m, off);
650 
651 	if (m->m_len < sizeof(*ip6)) {
652 		m = m_pullup(m, sizeof(*ip6));
653 		if (!m)
654 			return (IPPROTO_DONE);
655 	}
656 	ip6 = mtod(m, struct ip6_hdr *);
657 
658 	/*
659 	 * perform sanity check against inner src/dst.
660 	 * for source, perform ingress filter as well.
661 	 */
662 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
663 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
664 		m_freem(m);
665 		return (IPPROTO_DONE);
666 	}
667 
668 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
669 	if ((ifp->if_flags & IFF_LINK1) != 0)
670 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
671 	else
672 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
673 	ip6->ip6_flow &= ~htonl(0xff << 20);
674 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
675 
676 	m->m_pkthdr.rcvif = ifp;
677 
678 	if (bpf_peers_present(ifp->if_bpf)) {
679 		/*
680 		 * We need to prepend the address family as
681 		 * a four byte field.  Cons up a dummy header
682 		 * to pacify bpf.  This is safe because bpf
683 		 * will only read from the mbuf (i.e., it won't
684 		 * try to free it or keep a pointer a to it).
685 		 */
686 		u_int32_t af = AF_INET6;
687 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
688 	}
689 
690 	/*
691 	 * Put the packet to the network layer input queue according to the
692 	 * specified address family.
693 	 * See net/if_gif.c for possible issues with packet processing
694 	 * reorder due to extra queueing.
695 	 */
696 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
697 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
698 	M_SETFIB(m, ifp->if_fib);
699 	netisr_dispatch(NETISR_IPV6, m);
700 	return (IPPROTO_DONE);
701 }
702 
703 static int
704 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
705 {
706 	struct ifaddr *ifa;
707 	struct ifreq *ifr;
708 	struct sockaddr_in6 *sin6;
709 	struct in_addr addr;
710 	int error, mtu;
711 
712 	error = 0;
713 	switch (cmd) {
714 	case SIOCSIFADDR:
715 		ifa = (struct ifaddr *)data;
716 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
717 			error = EAFNOSUPPORT;
718 			break;
719 		}
720 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
721 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
722 			error = EINVAL;
723 			break;
724 		}
725 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
726 		if (isrfc1918addr(&addr)) {
727 			error = EINVAL;
728 			break;
729 		}
730 
731 		ifp->if_flags |= IFF_UP;
732 		ifp->if_drv_flags |= IFF_DRV_RUNNING;
733 		break;
734 
735 	case SIOCADDMULTI:
736 	case SIOCDELMULTI:
737 		ifr = (struct ifreq *)data;
738 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
739 			;
740 		else
741 			error = EAFNOSUPPORT;
742 		break;
743 
744 	case SIOCGIFMTU:
745 		break;
746 
747 	case SIOCSIFMTU:
748 		ifr = (struct ifreq *)data;
749 		mtu = ifr->ifr_mtu;
750 		/* RFC 4213 3.2 ideal world MTU */
751 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
752 			return (EINVAL);
753 		ifp->if_mtu = mtu;
754 		break;
755 
756 	default:
757 		error = EINVAL;
758 		break;
759 	}
760 
761 	return error;
762 }
763