xref: /freebsd/sys/net/if_stf.c (revision 076ad2f836d5f49dc1375f1677335a48fe0d4b82)
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 	struct mtx	sc_ro_mtx;
143 	u_int	sc_fibnum;
144 	const struct encaptab *encap_cookie;
145 };
146 #define STF2IFP(sc)	((sc)->sc_ifp)
147 
148 static const char stfname[] = "stf";
149 
150 /*
151  * Note that mutable fields in the softc are not currently locked.
152  * We do lock sc_ro in stf_output though.
153  */
154 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface");
155 static const int ip_stf_ttl = 40;
156 
157 extern  struct domain inetdomain;
158 static int in_stf_input(struct mbuf **, int *, int);
159 static 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 	char *dp;
206 	int err, unit, wildcard;
207 	struct stf_softc *sc;
208 	struct ifnet *ifp;
209 
210 	err = ifc_name2unit(name, &unit);
211 	if (err != 0)
212 		return (err);
213 	wildcard = (unit < 0);
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 	 * In the wildcard case, we need to update the name.
239 	 */
240 	if (wildcard) {
241 		for (dp = name; *dp != '\0'; dp++);
242 		if (snprintf(dp, len - (dp-name), "%d", unit) >
243 		    len - (dp-name) - 1) {
244 			/*
245 			 * This can only be a programmer error and
246 			 * there's no straightforward way to recover if
247 			 * it happens.
248 			 */
249 			panic("if_clone_create(): interface name too long");
250 		}
251 	}
252 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
253 	ifp->if_dname = stfname;
254 	ifp->if_dunit = IF_DUNIT_NONE;
255 
256 	mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
257 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
258 	    stf_encapcheck, &in_stf_protosw, sc);
259 	if (sc->encap_cookie == NULL) {
260 		if_printf(ifp, "attach failed\n");
261 		free(sc, M_STF);
262 		ifc_free_unit(ifc, unit);
263 		return (ENOMEM);
264 	}
265 
266 	ifp->if_mtu    = IPV6_MMTU;
267 	ifp->if_ioctl  = stf_ioctl;
268 	ifp->if_output = stf_output;
269 	ifp->if_snd.ifq_maxlen = ifqmaxlen;
270 	if_attach(ifp);
271 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
272 	return (0);
273 }
274 
275 static int
276 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
277 {
278 	struct stf_softc *sc = ifp->if_softc;
279 	int err;
280 
281 	err = encap_detach(sc->encap_cookie);
282 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
283 	mtx_destroy(&(sc)->sc_ro_mtx);
284 	bpfdetach(ifp);
285 	if_detach(ifp);
286 	if_free(ifp);
287 
288 	free(sc, M_STF);
289 	ifc_free_unit(ifc, STFUNIT);
290 
291 	return (0);
292 }
293 
294 static int
295 stfmodevent(module_t mod, int type, void *data)
296 {
297 
298 	switch (type) {
299 	case MOD_LOAD:
300 		stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match,
301 		    stf_clone_create, stf_clone_destroy);
302 		break;
303 	case MOD_UNLOAD:
304 		if_clone_detach(stf_cloner);
305 		break;
306 	default:
307 		return (EOPNOTSUPP);
308 	}
309 
310 	return (0);
311 }
312 
313 static moduledata_t stf_mod = {
314 	"if_stf",
315 	stfmodevent,
316 	0
317 };
318 
319 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
320 
321 static int
322 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
323 {
324 	struct ip ip;
325 	struct stf_softc *sc;
326 	struct in_addr a, b, mask;
327 	struct in6_addr addr6, mask6;
328 
329 	sc = (struct stf_softc *)arg;
330 	if (sc == NULL)
331 		return 0;
332 
333 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
334 		return 0;
335 
336 	/* IFF_LINK0 means "no decapsulation" */
337 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
338 		return 0;
339 
340 	if (proto != IPPROTO_IPV6)
341 		return 0;
342 
343 	/* LINTED const cast */
344 	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
345 
346 	if (ip.ip_v != 4)
347 		return 0;
348 
349 	if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0)
350 		return (0);
351 
352 	/*
353 	 * check if IPv4 dst matches the IPv4 address derived from the
354 	 * local 6to4 address.
355 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
356 	 */
357 	if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0)
358 		return 0;
359 
360 	/*
361 	 * check if IPv4 src matches the IPv4 address derived from the
362 	 * local 6to4 address masked by prefixmask.
363 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
364 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
365 	 */
366 	bzero(&a, sizeof(a));
367 	bcopy(GET_V4(&addr6), &a, sizeof(a));
368 	bcopy(GET_V4(&mask6), &mask, sizeof(mask));
369 	a.s_addr &= mask.s_addr;
370 	b = ip.ip_src;
371 	b.s_addr &= mask.s_addr;
372 	if (a.s_addr != b.s_addr)
373 		return 0;
374 
375 	/* stf interface makes single side match only */
376 	return 32;
377 }
378 
379 static int
380 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask)
381 {
382 	struct ifaddr *ia;
383 	struct in_ifaddr *ia4;
384 	struct in6_ifaddr *ia6;
385 	struct sockaddr_in6 *sin6;
386 	struct in_addr in;
387 
388 	if_addr_rlock(ifp);
389 	TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
390 		if (ia->ifa_addr->sa_family != AF_INET6)
391 			continue;
392 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
393 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
394 			continue;
395 
396 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
397 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
398 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
399 				break;
400 		if (ia4 == NULL)
401 			continue;
402 
403 		ia6 = (struct in6_ifaddr *)ia;
404 
405 		*addr = sin6->sin6_addr;
406 		*mask = ia6->ia_prefixmask.sin6_addr;
407 		if_addr_runlock(ifp);
408 		return (0);
409 	}
410 	if_addr_runlock(ifp);
411 
412 	return (ENOENT);
413 }
414 
415 static int
416 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst,
417     struct route *ro)
418 {
419 	struct stf_softc *sc;
420 	const struct sockaddr_in6 *dst6;
421 	struct in_addr in4;
422 	const void *ptr;
423 	u_int8_t tos;
424 	struct ip *ip;
425 	struct ip6_hdr *ip6;
426 	struct in6_addr addr6, mask6;
427 	int error;
428 
429 #ifdef MAC
430 	error = mac_ifnet_check_transmit(ifp, m);
431 	if (error) {
432 		m_freem(m);
433 		return (error);
434 	}
435 #endif
436 
437 	sc = ifp->if_softc;
438 	dst6 = (const struct sockaddr_in6 *)dst;
439 
440 	/* just in case */
441 	if ((ifp->if_flags & IFF_UP) == 0) {
442 		m_freem(m);
443 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
444 		return ENETDOWN;
445 	}
446 
447 	/*
448 	 * If we don't have an ip4 address that match my inner ip6 address,
449 	 * we shouldn't generate output.  Without this check, we'll end up
450 	 * using wrong IPv4 source.
451 	 */
452 	if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) {
453 		m_freem(m);
454 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
455 		return ENETDOWN;
456 	}
457 
458 	if (m->m_len < sizeof(*ip6)) {
459 		m = m_pullup(m, sizeof(*ip6));
460 		if (!m) {
461 			if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
462 			return ENOBUFS;
463 		}
464 	}
465 	ip6 = mtod(m, struct ip6_hdr *);
466 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
467 
468 	/*
469 	 * Pickup the right outer dst addr from the list of candidates.
470 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
471 	 */
472 	ptr = NULL;
473 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
474 		ptr = GET_V4(&ip6->ip6_dst);
475 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
476 		ptr = GET_V4(&dst6->sin6_addr);
477 	else {
478 		m_freem(m);
479 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
480 		return ENETUNREACH;
481 	}
482 	bcopy(ptr, &in4, sizeof(in4));
483 
484 	if (bpf_peers_present(ifp->if_bpf)) {
485 		/*
486 		 * We need to prepend the address family as
487 		 * a four byte field.  Cons up a dummy header
488 		 * to pacify bpf.  This is safe because bpf
489 		 * will only read from the mbuf (i.e., it won't
490 		 * try to free it or keep a pointer a to it).
491 		 */
492 		u_int af = AF_INET6;
493 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
494 	}
495 
496 	M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
497 	if (m == NULL) {
498 		if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
499 		return ENOBUFS;
500 	}
501 	ip = mtod(m, struct ip *);
502 
503 	bzero(ip, sizeof(*ip));
504 
505 	bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src));
506 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
507 	ip->ip_p = IPPROTO_IPV6;
508 	ip->ip_ttl = ip_stf_ttl;
509 	ip->ip_len = htons(m->m_pkthdr.len);
510 	if (ifp->if_flags & IFF_LINK1)
511 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
512 	else
513 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
514 
515 	M_SETFIB(m, sc->sc_fibnum);
516 	if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
517 	error = ip_output(m, NULL, NULL, 0, NULL, NULL);
518 
519 	return error;
520 }
521 
522 static int
523 isrfc1918addr(struct in_addr *in)
524 {
525 	/*
526 	 * returns 1 if private address range:
527 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
528 	 */
529 	if (stf_permit_rfc1918 == 0 && (
530 	    (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
531 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
532 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168))
533 		return 1;
534 
535 	return 0;
536 }
537 
538 static int
539 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp)
540 {
541 	struct rm_priotracker in_ifa_tracker;
542 	struct in_ifaddr *ia4;
543 
544 	/*
545 	 * reject packets with the following address:
546 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
547 	 */
548 	if (IN_MULTICAST(ntohl(in->s_addr)))
549 		return -1;
550 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
551 	case 0: case 127: case 255:
552 		return -1;
553 	}
554 
555 	/*
556 	 * reject packets with private address range.
557 	 * (requirement from RFC3056 section 2 1st paragraph)
558 	 */
559 	if (isrfc1918addr(in))
560 		return -1;
561 
562 	/*
563 	 * reject packets with broadcast
564 	 */
565 	IN_IFADDR_RLOCK(&in_ifa_tracker);
566 	TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) {
567 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
568 			continue;
569 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
570 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
571 			return -1;
572 		}
573 	}
574 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
575 
576 	/*
577 	 * perform ingress filter
578 	 */
579 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
580 		struct nhop4_basic nh4;
581 
582 		if (fib4_lookup_nh_basic(sc->sc_fibnum, *in, 0, 0, &nh4) != 0)
583 			return (-1);
584 
585 		if (nh4.nh_ifp != inifp)
586 			return (-1);
587 	}
588 
589 	return 0;
590 }
591 
592 static int
593 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp)
594 {
595 	/*
596 	 * check 6to4 addresses
597 	 */
598 	if (IN6_IS_ADDR_6TO4(in6)) {
599 		struct in_addr in4;
600 		bcopy(GET_V4(in6), &in4, sizeof(in4));
601 		return stf_checkaddr4(sc, &in4, inifp);
602 	}
603 
604 	/*
605 	 * reject anything that look suspicious.  the test is implemented
606 	 * in ip6_input too, but we check here as well to
607 	 * (1) reject bad packets earlier, and
608 	 * (2) to be safe against future ip6_input change.
609 	 */
610 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
611 		return -1;
612 
613 	return 0;
614 }
615 
616 static int
617 in_stf_input(struct mbuf **mp, int *offp, int proto)
618 {
619 	struct stf_softc *sc;
620 	struct ip *ip;
621 	struct ip6_hdr *ip6;
622 	struct mbuf *m;
623 	u_int8_t otos, itos;
624 	struct ifnet *ifp;
625 	int off;
626 
627 	m = *mp;
628 	off = *offp;
629 
630 	if (proto != IPPROTO_IPV6) {
631 		m_freem(m);
632 		return (IPPROTO_DONE);
633 	}
634 
635 	ip = mtod(m, struct ip *);
636 
637 	sc = (struct stf_softc *)encap_getarg(m);
638 
639 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
640 		m_freem(m);
641 		return (IPPROTO_DONE);
642 	}
643 
644 	ifp = STF2IFP(sc);
645 
646 #ifdef MAC
647 	mac_ifnet_create_mbuf(ifp, m);
648 #endif
649 
650 	/*
651 	 * perform sanity check against outer src/dst.
652 	 * for source, perform ingress filter as well.
653 	 */
654 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
655 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
656 		m_freem(m);
657 		return (IPPROTO_DONE);
658 	}
659 
660 	otos = ip->ip_tos;
661 	m_adj(m, off);
662 
663 	if (m->m_len < sizeof(*ip6)) {
664 		m = m_pullup(m, sizeof(*ip6));
665 		if (!m)
666 			return (IPPROTO_DONE);
667 	}
668 	ip6 = mtod(m, struct ip6_hdr *);
669 
670 	/*
671 	 * perform sanity check against inner src/dst.
672 	 * for source, perform ingress filter as well.
673 	 */
674 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
675 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
676 		m_freem(m);
677 		return (IPPROTO_DONE);
678 	}
679 
680 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
681 	if ((ifp->if_flags & IFF_LINK1) != 0)
682 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
683 	else
684 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
685 	ip6->ip6_flow &= ~htonl(0xff << 20);
686 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
687 
688 	m->m_pkthdr.rcvif = ifp;
689 
690 	if (bpf_peers_present(ifp->if_bpf)) {
691 		/*
692 		 * We need to prepend the address family as
693 		 * a four byte field.  Cons up a dummy header
694 		 * to pacify bpf.  This is safe because bpf
695 		 * will only read from the mbuf (i.e., it won't
696 		 * try to free it or keep a pointer a to it).
697 		 */
698 		u_int32_t af = AF_INET6;
699 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
700 	}
701 
702 	/*
703 	 * Put the packet to the network layer input queue according to the
704 	 * specified address family.
705 	 * See net/if_gif.c for possible issues with packet processing
706 	 * reorder due to extra queueing.
707 	 */
708 	if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
709 	if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
710 	M_SETFIB(m, ifp->if_fib);
711 	netisr_dispatch(NETISR_IPV6, m);
712 	return (IPPROTO_DONE);
713 }
714 
715 static int
716 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
717 {
718 	struct ifaddr *ifa;
719 	struct ifreq *ifr;
720 	struct sockaddr_in6 *sin6;
721 	struct in_addr addr;
722 	int error, mtu;
723 
724 	error = 0;
725 	switch (cmd) {
726 	case SIOCSIFADDR:
727 		ifa = (struct ifaddr *)data;
728 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
729 			error = EAFNOSUPPORT;
730 			break;
731 		}
732 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
733 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
734 			error = EINVAL;
735 			break;
736 		}
737 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
738 		if (isrfc1918addr(&addr)) {
739 			error = EINVAL;
740 			break;
741 		}
742 
743 		ifp->if_flags |= IFF_UP;
744 		break;
745 
746 	case SIOCADDMULTI:
747 	case SIOCDELMULTI:
748 		ifr = (struct ifreq *)data;
749 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
750 			;
751 		else
752 			error = EAFNOSUPPORT;
753 		break;
754 
755 	case SIOCGIFMTU:
756 		break;
757 
758 	case SIOCSIFMTU:
759 		ifr = (struct ifreq *)data;
760 		mtu = ifr->ifr_mtu;
761 		/* RFC 4213 3.2 ideal world MTU */
762 		if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20)
763 			return (EINVAL);
764 		ifp->if_mtu = mtu;
765 		break;
766 
767 	default:
768 		error = EINVAL;
769 		break;
770 	}
771 
772 	return error;
773 }
774