xref: /freebsd/sys/net/if_stf.c (revision b3aaa0cc21c63d388230c7ef2a80abd631ff20d5) 
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 #include "opt_mac.h"
80 
81 #include <sys/param.h>
82 #include <sys/systm.h>
83 #include <sys/socket.h>
84 #include <sys/sockio.h>
85 #include <sys/mbuf.h>
86 #include <sys/errno.h>
87 #include <sys/kernel.h>
88 #include <sys/module.h>
89 #include <sys/protosw.h>
90 #include <sys/proc.h>
91 #include <sys/queue.h>
92 #include <sys/sysctl.h>
93 #include <machine/cpu.h>
94 
95 #include <sys/malloc.h>
96 #include <sys/vimage.h>
97 
98 #include <net/if.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/if_stf.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 #include <netinet/vinet.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 SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface");
127 
128 static int stf_route_cache = 1;
129 SYSCTL_INT(_net_link_stf, OID_AUTO, route_cache, CTLFLAG_RW,
130     &stf_route_cache, 0, "Caching of IPv4 routes for 6to4 Output");
131 
132 #define STFNAME		"stf"
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)	((caddr_t)(&(x)->s6_addr16[1]))
142 
143 struct stf_softc {
144 	struct ifnet	*sc_ifp;
145 	union {
146 		struct route  __sc_ro4;
147 		struct route_in6 __sc_ro6; /* just for safety */
148 	} __sc_ro46;
149 #define sc_ro	__sc_ro46.__sc_ro4
150 	struct mtx	sc_ro_mtx;
151 	u_int	sc_fibnum;
152 	const struct encaptab *encap_cookie;
153 };
154 #define STF2IFP(sc)	((sc)->sc_ifp)
155 
156 /*
157  * Note that mutable fields in the softc are not currently locked.
158  * We do lock sc_ro in stf_output though.
159  */
160 static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface");
161 static const int ip_stf_ttl = 40;
162 
163 extern  struct domain inetdomain;
164 struct protosw in_stf_protosw = {
165 	.pr_type =		SOCK_RAW,
166 	.pr_domain =		&inetdomain,
167 	.pr_protocol =		IPPROTO_IPV6,
168 	.pr_flags =		PR_ATOMIC|PR_ADDR,
169 	.pr_input =		in_stf_input,
170 	.pr_output =		(pr_output_t *)rip_output,
171 	.pr_ctloutput =		rip_ctloutput,
172 	.pr_usrreqs =		&rip_usrreqs
173 };
174 
175 static char *stfnames[] = {"stf0", "stf", "6to4", NULL};
176 
177 static int stfmodevent(module_t, int, void *);
178 static int stf_encapcheck(const struct mbuf *, int, int, void *);
179 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
180 static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *,
181 	struct rtentry *);
182 static int isrfc1918addr(struct in_addr *);
183 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
184 	struct ifnet *);
185 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
186 	struct ifnet *);
187 static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
188 static int stf_ioctl(struct ifnet *, u_long, caddr_t);
189 
190 static int stf_clone_match(struct if_clone *, const char *);
191 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t);
192 static int stf_clone_destroy(struct if_clone *, struct ifnet *);
193 struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0,
194     NULL, stf_clone_match, stf_clone_create, stf_clone_destroy);
195 
196 static int
197 stf_clone_match(struct if_clone *ifc, const char *name)
198 {
199 	int i;
200 
201 	for(i = 0; stfnames[i] != NULL; i++) {
202 		if (strcmp(stfnames[i], name) == 0)
203 			return (1);
204 	}
205 
206 	return (0);
207 }
208 
209 static int
210 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params)
211 {
212 	int err, unit;
213 	struct stf_softc *sc;
214 	struct ifnet *ifp;
215 
216 	/*
217 	 * We can only have one unit, but since unit allocation is
218 	 * already locked, we use it to keep from allocating extra
219 	 * interfaces.
220 	 */
221 	unit = STFUNIT;
222 	err = ifc_alloc_unit(ifc, &unit);
223 	if (err != 0)
224 		return (err);
225 
226 	sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO);
227 	ifp = STF2IFP(sc) = if_alloc(IFT_STF);
228 	if (ifp == NULL) {
229 		free(sc, M_STF);
230 		ifc_free_unit(ifc, unit);
231 		return (ENOSPC);
232 	}
233 	ifp->if_softc = sc;
234 	sc->sc_fibnum = curthread->td_proc->p_fibnum;
235 
236 	/*
237 	 * Set the name manually rather then using if_initname because
238 	 * we don't conform to the default naming convention for interfaces.
239 	 */
240 	strlcpy(ifp->if_xname, name, IFNAMSIZ);
241 	ifp->if_dname = ifc->ifc_name;
242 	ifp->if_dunit = IF_DUNIT_NONE;
243 
244 	mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF);
245 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
246 	    stf_encapcheck, &in_stf_protosw, sc);
247 	if (sc->encap_cookie == NULL) {
248 		if_printf(ifp, "attach failed\n");
249 		free(sc, M_STF);
250 		ifc_free_unit(ifc, unit);
251 		return (ENOMEM);
252 	}
253 
254 	ifp->if_mtu    = IPV6_MMTU;
255 	ifp->if_ioctl  = stf_ioctl;
256 	ifp->if_output = stf_output;
257 	ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
258 	if_attach(ifp);
259 	bpfattach(ifp, DLT_NULL, sizeof(u_int32_t));
260 	return (0);
261 }
262 
263 static int
264 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp)
265 {
266 	struct stf_softc *sc = ifp->if_softc;
267 	int err;
268 
269 	err = encap_detach(sc->encap_cookie);
270 	KASSERT(err == 0, ("Unexpected error detaching encap_cookie"));
271 	mtx_destroy(&(sc)->sc_ro_mtx);
272 	bpfdetach(ifp);
273 	if_detach(ifp);
274 	if_free(ifp);
275 
276 	free(sc, M_STF);
277 	ifc_free_unit(ifc, STFUNIT);
278 
279 	return (0);
280 }
281 
282 static int
283 stfmodevent(mod, type, data)
284 	module_t mod;
285 	int type;
286 	void *data;
287 {
288 
289 	switch (type) {
290 	case MOD_LOAD:
291 		if_clone_attach(&stf_cloner);
292 		break;
293 	case MOD_UNLOAD:
294 		if_clone_detach(&stf_cloner);
295 		break;
296 	default:
297 		return (EOPNOTSUPP);
298 	}
299 
300 	return (0);
301 }
302 
303 static moduledata_t stf_mod = {
304 	"if_stf",
305 	stfmodevent,
306 	0
307 };
308 
309 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
310 
311 static int
312 stf_encapcheck(m, off, proto, arg)
313 	const struct mbuf *m;
314 	int off;
315 	int proto;
316 	void *arg;
317 {
318 	struct ip ip;
319 	struct in6_ifaddr *ia6;
320 	struct stf_softc *sc;
321 	struct in_addr a, b, mask;
322 
323 	sc = (struct stf_softc *)arg;
324 	if (sc == NULL)
325 		return 0;
326 
327 	if ((STF2IFP(sc)->if_flags & IFF_UP) == 0)
328 		return 0;
329 
330 	/* IFF_LINK0 means "no decapsulation" */
331 	if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0)
332 		return 0;
333 
334 	if (proto != IPPROTO_IPV6)
335 		return 0;
336 
337 	/* LINTED const cast */
338 	m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip);
339 
340 	if (ip.ip_v != 4)
341 		return 0;
342 
343 	ia6 = stf_getsrcifa6(STF2IFP(sc));
344 	if (ia6 == NULL)
345 		return 0;
346 
347 	/*
348 	 * check if IPv4 dst matches the IPv4 address derived from the
349 	 * local 6to4 address.
350 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
351 	 */
352 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
353 	    sizeof(ip.ip_dst)) != 0)
354 		return 0;
355 
356 	/*
357 	 * check if IPv4 src matches the IPv4 address derived from the
358 	 * local 6to4 address masked by prefixmask.
359 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
360 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
361 	 */
362 	bzero(&a, sizeof(a));
363 	bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a));
364 	bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask));
365 	a.s_addr &= mask.s_addr;
366 	b = ip.ip_src;
367 	b.s_addr &= mask.s_addr;
368 	if (a.s_addr != b.s_addr)
369 		return 0;
370 
371 	/* stf interface makes single side match only */
372 	return 32;
373 }
374 
375 static struct in6_ifaddr *
376 stf_getsrcifa6(ifp)
377 	struct ifnet *ifp;
378 {
379 	INIT_VNET_INET(ifp->if_vnet);
380 	struct ifaddr *ia;
381 	struct in_ifaddr *ia4;
382 	struct sockaddr_in6 *sin6;
383 	struct in_addr in;
384 
385 	TAILQ_FOREACH(ia, &ifp->if_addrlist, ifa_list) {
386 		if (ia->ifa_addr->sa_family != AF_INET6)
387 			continue;
388 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
389 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
390 			continue;
391 
392 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
393 		LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash)
394 			if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
395 				break;
396 		if (ia4 == NULL)
397 			continue;
398 
399 		return (struct in6_ifaddr *)ia;
400 	}
401 
402 	return NULL;
403 }
404 
405 static int
406 stf_output(ifp, m, dst, rt)
407 	struct ifnet *ifp;
408 	struct mbuf *m;
409 	struct sockaddr *dst;
410 	struct rtentry *rt;
411 {
412 	struct stf_softc *sc;
413 	struct sockaddr_in6 *dst6;
414 	struct route *cached_route;
415 	struct in_addr in4;
416 	caddr_t ptr;
417 	struct sockaddr_in *dst4;
418 	u_int8_t tos;
419 	struct ip *ip;
420 	struct ip6_hdr *ip6;
421 	struct in6_ifaddr *ia6;
422 	u_int32_t af;
423 	int error;
424 
425 #ifdef MAC
426 	error = mac_ifnet_check_transmit(ifp, m);
427 	if (error) {
428 		m_freem(m);
429 		return (error);
430 	}
431 #endif
432 
433 	sc = ifp->if_softc;
434 	dst6 = (struct sockaddr_in6 *)dst;
435 
436 	/* just in case */
437 	if ((ifp->if_flags & IFF_UP) == 0) {
438 		m_freem(m);
439 		ifp->if_oerrors++;
440 		return ENETDOWN;
441 	}
442 
443 	/*
444 	 * If we don't have an ip4 address that match my inner ip6 address,
445 	 * we shouldn't generate output.  Without this check, we'll end up
446 	 * using wrong IPv4 source.
447 	 */
448 	ia6 = stf_getsrcifa6(ifp);
449 	if (ia6 == NULL) {
450 		m_freem(m);
451 		ifp->if_oerrors++;
452 		return ENETDOWN;
453 	}
454 
455 	if (m->m_len < sizeof(*ip6)) {
456 		m = m_pullup(m, sizeof(*ip6));
457 		if (!m) {
458 			ifp->if_oerrors++;
459 			return ENOBUFS;
460 		}
461 	}
462 	ip6 = mtod(m, struct ip6_hdr *);
463 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
464 
465 	/*
466 	 * BPF writes need to be handled specially.
467 	 * This is a null operation, nothing here checks dst->sa_family.
468 	 */
469 	if (dst->sa_family == AF_UNSPEC) {
470 		bcopy(dst->sa_data, &af, sizeof(af));
471 		dst->sa_family = af;
472 	}
473 
474 	/*
475 	 * Pickup the right outer dst addr from the list of candidates.
476 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
477 	 */
478 	ptr = NULL;
479 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
480 		ptr = GET_V4(&ip6->ip6_dst);
481 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
482 		ptr = GET_V4(&dst6->sin6_addr);
483 	else {
484 		m_freem(m);
485 		ifp->if_oerrors++;
486 		return ENETUNREACH;
487 	}
488 	bcopy(ptr, &in4, sizeof(in4));
489 
490 	if (bpf_peers_present(ifp->if_bpf)) {
491 		/*
492 		 * We need to prepend the address family as
493 		 * a four byte field.  Cons up a dummy header
494 		 * to pacify bpf.  This is safe because bpf
495 		 * will only read from the mbuf (i.e., it won't
496 		 * try to free it or keep a pointer a to it).
497 		 */
498 		af = AF_INET6;
499 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
500 	}
501 
502 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
503 	if (m && m->m_len < sizeof(struct ip))
504 		m = m_pullup(m, sizeof(struct ip));
505 	if (m == NULL) {
506 		ifp->if_oerrors++;
507 		return ENOBUFS;
508 	}
509 	ip = mtod(m, struct ip *);
510 
511 	bzero(ip, sizeof(*ip));
512 
513 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
514 	    &ip->ip_src, sizeof(ip->ip_src));
515 	bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst));
516 	ip->ip_p = IPPROTO_IPV6;
517 	ip->ip_ttl = ip_stf_ttl;
518 	ip->ip_len = m->m_pkthdr.len;	/*host order*/
519 	if (ifp->if_flags & IFF_LINK1)
520 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
521 	else
522 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
523 
524 	if (!stf_route_cache) {
525 		cached_route = NULL;
526 		goto sendit;
527 	}
528 
529 	/*
530 	 * Do we have a cached route?
531 	 */
532 	mtx_lock(&(sc)->sc_ro_mtx);
533 	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
534 	if (dst4->sin_family != AF_INET ||
535 	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
536 		/* cache route doesn't match */
537 		dst4->sin_family = AF_INET;
538 		dst4->sin_len = sizeof(struct sockaddr_in);
539 		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
540 		if (sc->sc_ro.ro_rt) {
541 			RTFREE(sc->sc_ro.ro_rt);
542 			sc->sc_ro.ro_rt = NULL;
543 		}
544 	}
545 
546 	if (sc->sc_ro.ro_rt == NULL) {
547 		rtalloc_fib(&sc->sc_ro, sc->sc_fibnum);
548 		if (sc->sc_ro.ro_rt == NULL) {
549 			m_freem(m);
550 			mtx_unlock(&(sc)->sc_ro_mtx);
551 			ifp->if_oerrors++;
552 			return ENETUNREACH;
553 		}
554 	}
555 	cached_route = &sc->sc_ro;
556 
557 sendit:
558 	M_SETFIB(m, sc->sc_fibnum);
559 	ifp->if_opackets++;
560 	error = ip_output(m, NULL, cached_route, 0, NULL, NULL);
561 
562 	if (cached_route != NULL)
563 		mtx_unlock(&(sc)->sc_ro_mtx);
564 	return error;
565 }
566 
567 static int
568 isrfc1918addr(in)
569 	struct in_addr *in;
570 {
571 	/*
572 	 * returns 1 if private address range:
573 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
574 	 */
575 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
576 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
577 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
578 		return 1;
579 
580 	return 0;
581 }
582 
583 static int
584 stf_checkaddr4(sc, in, inifp)
585 	struct stf_softc *sc;
586 	struct in_addr *in;
587 	struct ifnet *inifp;	/* incoming interface */
588 {
589 	INIT_VNET_INET(curvnet);
590 	struct in_ifaddr *ia4;
591 
592 	/*
593 	 * reject packets with the following address:
594 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
595 	 */
596 	if (IN_MULTICAST(ntohl(in->s_addr)))
597 		return -1;
598 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
599 	case 0: case 127: case 255:
600 		return -1;
601 	}
602 
603 	/*
604 	 * reject packets with private address range.
605 	 * (requirement from RFC3056 section 2 1st paragraph)
606 	 */
607 	if (isrfc1918addr(in))
608 		return -1;
609 
610 	/*
611 	 * reject packets with broadcast
612 	 */
613 	for (ia4 = TAILQ_FIRST(&V_in_ifaddrhead);
614 	     ia4;
615 	     ia4 = TAILQ_NEXT(ia4, ia_link))
616 	{
617 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
618 			continue;
619 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
620 			return -1;
621 	}
622 
623 	/*
624 	 * perform ingress filter
625 	 */
626 	if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) {
627 		struct sockaddr_in sin;
628 		struct rtentry *rt;
629 
630 		bzero(&sin, sizeof(sin));
631 		sin.sin_family = AF_INET;
632 		sin.sin_len = sizeof(struct sockaddr_in);
633 		sin.sin_addr = *in;
634 		rt = rtalloc1_fib((struct sockaddr *)&sin, 0,
635 		    0UL, sc->sc_fibnum);
636 		if (!rt || rt->rt_ifp != inifp) {
637 #if 0
638 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
639 			    "due to ingress filter\n", if_name(STF2IFP(sc)),
640 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
641 #endif
642 			if (rt)
643 				RTFREE_LOCKED(rt);
644 			return -1;
645 		}
646 		RTFREE_LOCKED(rt);
647 	}
648 
649 	return 0;
650 }
651 
652 static int
653 stf_checkaddr6(sc, in6, inifp)
654 	struct stf_softc *sc;
655 	struct in6_addr *in6;
656 	struct ifnet *inifp;	/* incoming interface */
657 {
658 	/*
659 	 * check 6to4 addresses
660 	 */
661 	if (IN6_IS_ADDR_6TO4(in6)) {
662 		struct in_addr in4;
663 		bcopy(GET_V4(in6), &in4, sizeof(in4));
664 		return stf_checkaddr4(sc, &in4, inifp);
665 	}
666 
667 	/*
668 	 * reject anything that look suspicious.  the test is implemented
669 	 * in ip6_input too, but we check here as well to
670 	 * (1) reject bad packets earlier, and
671 	 * (2) to be safe against future ip6_input change.
672 	 */
673 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
674 		return -1;
675 
676 	return 0;
677 }
678 
679 void
680 in_stf_input(m, off)
681 	struct mbuf *m;
682 	int off;
683 {
684 	int proto;
685 	struct stf_softc *sc;
686 	struct ip *ip;
687 	struct ip6_hdr *ip6;
688 	u_int8_t otos, itos;
689 	struct ifnet *ifp;
690 
691 	proto = mtod(m, struct ip *)->ip_p;
692 
693 	if (proto != IPPROTO_IPV6) {
694 		m_freem(m);
695 		return;
696 	}
697 
698 	ip = mtod(m, struct ip *);
699 
700 	sc = (struct stf_softc *)encap_getarg(m);
701 
702 	if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) {
703 		m_freem(m);
704 		return;
705 	}
706 
707 	ifp = STF2IFP(sc);
708 
709 #ifdef MAC
710 	mac_ifnet_create_mbuf(ifp, m);
711 #endif
712 
713 	/*
714 	 * perform sanity check against outer src/dst.
715 	 * for source, perform ingress filter as well.
716 	 */
717 	if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
718 	    stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
719 		m_freem(m);
720 		return;
721 	}
722 
723 	otos = ip->ip_tos;
724 	m_adj(m, off);
725 
726 	if (m->m_len < sizeof(*ip6)) {
727 		m = m_pullup(m, sizeof(*ip6));
728 		if (!m)
729 			return;
730 	}
731 	ip6 = mtod(m, struct ip6_hdr *);
732 
733 	/*
734 	 * perform sanity check against inner src/dst.
735 	 * for source, perform ingress filter as well.
736 	 */
737 	if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
738 	    stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
739 		m_freem(m);
740 		return;
741 	}
742 
743 	itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
744 	if ((ifp->if_flags & IFF_LINK1) != 0)
745 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
746 	else
747 		ip_ecn_egress(ECN_NOCARE, &otos, &itos);
748 	ip6->ip6_flow &= ~htonl(0xff << 20);
749 	ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
750 
751 	m->m_pkthdr.rcvif = ifp;
752 
753 	if (bpf_peers_present(ifp->if_bpf)) {
754 		/*
755 		 * We need to prepend the address family as
756 		 * a four byte field.  Cons up a dummy header
757 		 * to pacify bpf.  This is safe because bpf
758 		 * will only read from the mbuf (i.e., it won't
759 		 * try to free it or keep a pointer a to it).
760 		 */
761 		u_int32_t af = AF_INET6;
762 		bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m);
763 	}
764 
765 	/*
766 	 * Put the packet to the network layer input queue according to the
767 	 * specified address family.
768 	 * See net/if_gif.c for possible issues with packet processing
769 	 * reorder due to extra queueing.
770 	 */
771 	ifp->if_ipackets++;
772 	ifp->if_ibytes += m->m_pkthdr.len;
773 	netisr_dispatch(NETISR_IPV6, m);
774 }
775 
776 /* ARGSUSED */
777 static void
778 stf_rtrequest(cmd, rt, info)
779 	int cmd;
780 	struct rtentry *rt;
781 	struct rt_addrinfo *info;
782 {
783 	RT_LOCK_ASSERT(rt);
784 	rt->rt_rmx.rmx_mtu = IPV6_MMTU;
785 }
786 
787 static int
788 stf_ioctl(ifp, cmd, data)
789 	struct ifnet *ifp;
790 	u_long cmd;
791 	caddr_t data;
792 {
793 	struct ifaddr *ifa;
794 	struct ifreq *ifr;
795 	struct sockaddr_in6 *sin6;
796 	struct in_addr addr;
797 	int error;
798 
799 	error = 0;
800 	switch (cmd) {
801 	case SIOCSIFADDR:
802 		ifa = (struct ifaddr *)data;
803 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
804 			error = EAFNOSUPPORT;
805 			break;
806 		}
807 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
808 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
809 			error = EINVAL;
810 			break;
811 		}
812 		bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr));
813 		if (isrfc1918addr(&addr)) {
814 			error = EINVAL;
815 			break;
816 		}
817 
818 		ifa->ifa_rtrequest = stf_rtrequest;
819 		ifp->if_flags |= IFF_UP;
820 		break;
821 
822 	case SIOCADDMULTI:
823 	case SIOCDELMULTI:
824 		ifr = (struct ifreq *)data;
825 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
826 			;
827 		else
828 			error = EAFNOSUPPORT;
829 		break;
830 
831 	default:
832 		error = EINVAL;
833 		break;
834 	}
835 
836 	return error;
837 }
838