Home | History | Annotate | Line # | Download | only in net
if_stf.c revision 1.33
      1 /*	$NetBSD: if_stf.c,v 1.33 2003/05/01 07:52:59 itojun Exp $	*/
      2 /*	$KAME: if_stf.c,v 1.62 2001/06/07 22:32:16 itojun 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/cdefs.h>
     78 __KERNEL_RCSID(0, "$NetBSD: if_stf.c,v 1.33 2003/05/01 07:52:59 itojun Exp $");
     79 
     80 #include "opt_inet.h"
     81 
     82 #include <sys/param.h>
     83 #include <sys/systm.h>
     84 #include <sys/socket.h>
     85 #include <sys/sockio.h>
     86 #include <sys/mbuf.h>
     87 #include <sys/errno.h>
     88 #include <sys/ioctl.h>
     89 #include <sys/protosw.h>
     90 #include <sys/queue.h>
     91 #include <sys/syslog.h>
     92 #include <machine/cpu.h>
     93 
     94 #include <net/if.h>
     95 #include <net/route.h>
     96 #include <net/netisr.h>
     97 #include <net/if_types.h>
     98 #include <net/if_stf.h>
     99 
    100 #include <netinet/in.h>
    101 #include <netinet/in_systm.h>
    102 #include <netinet/ip.h>
    103 #include <netinet/ip_var.h>
    104 #include <netinet/in_var.h>
    105 
    106 #include <netinet/ip6.h>
    107 #include <netinet6/ip6_var.h>
    108 #include <netinet6/in6_gif.h>
    109 #include <netinet6/in6_var.h>
    110 #include <netinet/ip_ecn.h>
    111 
    112 #include <netinet/ip_encap.h>
    113 
    114 #include <machine/stdarg.h>
    115 
    116 #include <net/net_osdep.h>
    117 
    118 #include "bpfilter.h"
    119 #include "stf.h"
    120 #include "gif.h"	/*XXX*/
    121 
    122 #if NBPFILTER > 0
    123 #include <net/bpf.h>
    124 #endif
    125 
    126 #if NGIF > 0
    127 #include <net/if_gif.h>
    128 #endif
    129 
    130 #define IN6_IS_ADDR_6TO4(x)	(ntohs((x)->s6_addr16[0]) == 0x2002)
    131 #define GET_V4(x)	((struct in_addr *)(&(x)->s6_addr16[1]))
    132 
    133 struct stf_softc {
    134 	struct ifnet	sc_if;	   /* common area */
    135 	union {
    136 		struct route  __sc_ro4;
    137 		struct route_in6 __sc_ro6; /* just for safety */
    138 	} __sc_ro46;
    139 #define sc_ro	__sc_ro46.__sc_ro4
    140 	const struct encaptab *encap_cookie;
    141 	LIST_ENTRY(stf_softc) sc_list;
    142 };
    143 
    144 LIST_HEAD(, stf_softc) stf_softc_list;
    145 
    146 int	stf_clone_create __P((struct if_clone *, int));
    147 void	stf_clone_destroy __P((struct ifnet *));
    148 
    149 struct if_clone stf_cloner =
    150     IF_CLONE_INITIALIZER("stf", stf_clone_create, stf_clone_destroy);
    151 
    152 #if NGIF > 0
    153 extern int ip_gif_ttl;	/*XXX*/
    154 #else
    155 static int ip_gif_ttl = 40;	/*XXX*/
    156 #endif
    157 
    158 extern struct domain inetdomain;
    159 struct protosw in_stf_protosw =
    160 { SOCK_RAW,	&inetdomain,	IPPROTO_IPV6,	PR_ATOMIC|PR_ADDR,
    161   in_stf_input, rip_output,	0,		rip_ctloutput,
    162   rip_usrreq,
    163   0,            0,              0,              0
    164 };
    165 
    166 void stfattach __P((int));
    167 static int stf_encapcheck __P((const struct mbuf *, int, int, void *));
    168 static struct in6_ifaddr *stf_getsrcifa6 __P((struct ifnet *));
    169 static int stf_output __P((struct ifnet *, struct mbuf *, struct sockaddr *,
    170 	struct rtentry *));
    171 static int isrfc1918addr __P((struct in_addr *));
    172 static int stf_checkaddr4 __P((struct stf_softc *, struct in_addr *,
    173 	struct ifnet *));
    174 static int stf_checkaddr6 __P((struct stf_softc *, struct in6_addr *,
    175 	struct ifnet *));
    176 static void stf_rtrequest __P((int, struct rtentry *, struct rt_addrinfo *));
    177 static int stf_ioctl __P((struct ifnet *, u_long, caddr_t));
    178 
    179 /* ARGSUSED */
    180 void
    181 stfattach(count)
    182 	int count;
    183 {
    184 
    185 	LIST_INIT(&stf_softc_list);
    186 	if_clone_attach(&stf_cloner);
    187 }
    188 
    189 int
    190 stf_clone_create(ifc, unit)
    191 	struct if_clone *ifc;
    192 	int unit;
    193 {
    194 	struct stf_softc *sc;
    195 
    196 	if (LIST_FIRST(&stf_softc_list) != NULL) {
    197 		/* Only one stf interface is allowed. */
    198 		return (EEXIST);
    199 	}
    200 
    201 	sc = malloc(sizeof(struct stf_softc), M_DEVBUF, M_WAIT);
    202 	memset(sc, 0, sizeof(struct stf_softc));
    203 
    204 	sprintf(sc->sc_if.if_xname, "%s%d", ifc->ifc_name, unit);
    205 
    206 	sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6,
    207 	    stf_encapcheck, &in_stf_protosw, sc);
    208 	if (sc->encap_cookie == NULL) {
    209 		printf("%s: unable to attach encap\n", if_name(&sc->sc_if));
    210 		free(sc, M_DEVBUF);
    211 		return (EIO);	/* XXX */
    212 	}
    213 
    214 	sc->sc_if.if_mtu    = IPV6_MMTU;
    215 	sc->sc_if.if_flags  = 0;
    216 	sc->sc_if.if_ioctl  = stf_ioctl;
    217 	sc->sc_if.if_output = stf_output;
    218 	sc->sc_if.if_type   = IFT_STF;
    219 	sc->sc_if.if_dlt    = DLT_NULL;
    220 	if_attach(&sc->sc_if);
    221 	if_alloc_sadl(&sc->sc_if);
    222 #if NBPFILTER > 0
    223 	bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int));
    224 #endif
    225 	LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list);
    226 	return (0);
    227 }
    228 
    229 void
    230 stf_clone_destroy(ifp)
    231 	struct ifnet *ifp;
    232 {
    233 	struct stf_softc *sc = (void *) ifp;
    234 
    235 	LIST_REMOVE(sc, sc_list);
    236 	encap_detach(sc->encap_cookie);
    237 #if NBPFILTER > 0
    238 	bpfdetach(ifp);
    239 #endif
    240 	if_detach(ifp);
    241 	free(sc, M_DEVBUF);
    242 }
    243 
    244 static int
    245 stf_encapcheck(m, off, proto, arg)
    246 	const struct mbuf *m;
    247 	int off;
    248 	int proto;
    249 	void *arg;
    250 {
    251 	struct ip ip;
    252 	struct in6_ifaddr *ia6;
    253 	struct stf_softc *sc;
    254 	struct in_addr a, b;
    255 
    256 	sc = (struct stf_softc *)arg;
    257 	if (sc == NULL)
    258 		return 0;
    259 
    260 	if ((sc->sc_if.if_flags & IFF_UP) == 0)
    261 		return 0;
    262 
    263 	/* IFF_LINK0 means "no decapsulation" */
    264 	if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
    265 		return 0;
    266 
    267 	if (proto != IPPROTO_IPV6)
    268 		return 0;
    269 
    270 	/* LINTED const cast */
    271 	m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip);
    272 
    273 	if (ip.ip_v != 4)
    274 		return 0;
    275 
    276 	ia6 = stf_getsrcifa6(&sc->sc_if);
    277 	if (ia6 == NULL)
    278 		return 0;
    279 
    280 	/*
    281 	 * check if IPv4 dst matches the IPv4 address derived from the
    282 	 * local 6to4 address.
    283 	 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
    284 	 */
    285 	if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
    286 	    sizeof(ip.ip_dst)) != 0)
    287 		return 0;
    288 
    289 	/*
    290 	 * check if IPv4 src matches the IPv4 address derived from the
    291 	 * local 6to4 address masked by prefixmask.
    292 	 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
    293 	 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
    294 	 */
    295 	memset(&a, 0, sizeof(a));
    296 	a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
    297 	a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
    298 	b = ip.ip_src;
    299 	b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
    300 	if (a.s_addr != b.s_addr)
    301 		return 0;
    302 
    303 	/* stf interface makes single side match only */
    304 	return 32;
    305 }
    306 
    307 static struct in6_ifaddr *
    308 stf_getsrcifa6(ifp)
    309 	struct ifnet *ifp;
    310 {
    311 	struct ifaddr *ia;
    312 	struct in_ifaddr *ia4;
    313 	struct sockaddr_in6 *sin6;
    314 	struct in_addr in;
    315 
    316 	TAILQ_FOREACH(ia, &ifp->if_addrlist, ifa_list)
    317 	{
    318 		if (ia->ifa_addr == NULL)
    319 			continue;
    320 		if (ia->ifa_addr->sa_family != AF_INET6)
    321 			continue;
    322 		sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
    323 		if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
    324 			continue;
    325 
    326 		bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
    327 		INADDR_TO_IA(in, ia4);
    328 		if (ia4 == NULL)
    329 			continue;
    330 
    331 		return (struct in6_ifaddr *)ia;
    332 	}
    333 
    334 	return NULL;
    335 }
    336 
    337 static int
    338 stf_output(ifp, m, dst, rt)
    339 	struct ifnet *ifp;
    340 	struct mbuf *m;
    341 	struct sockaddr *dst;
    342 	struct rtentry *rt;
    343 {
    344 	struct stf_softc *sc;
    345 	struct sockaddr_in6 *dst6;
    346 	struct in_addr *in4;
    347 	struct sockaddr_in *dst4;
    348 	u_int8_t tos;
    349 	struct ip *ip;
    350 	struct ip6_hdr *ip6;
    351 	struct in6_ifaddr *ia6;
    352 
    353 	sc = (struct stf_softc*)ifp;
    354 	dst6 = (struct sockaddr_in6 *)dst;
    355 
    356 	/* just in case */
    357 	if ((ifp->if_flags & IFF_UP) == 0) {
    358 		m_freem(m);
    359 		return ENETDOWN;
    360 	}
    361 
    362 	/*
    363 	 * If we don't have an ip4 address that match my inner ip6 address,
    364 	 * we shouldn't generate output.  Without this check, we'll end up
    365 	 * using wrong IPv4 source.
    366 	 */
    367 	ia6 = stf_getsrcifa6(ifp);
    368 	if (ia6 == NULL) {
    369 		m_freem(m);
    370 		ifp->if_oerrors++;
    371 		return ENETDOWN;
    372 	}
    373 
    374 	if (m->m_len < sizeof(*ip6)) {
    375 		m = m_pullup(m, sizeof(*ip6));
    376 		if (m == NULL) {
    377 			ifp->if_oerrors++;
    378 			return ENOBUFS;
    379 		}
    380 	}
    381 	ip6 = mtod(m, struct ip6_hdr *);
    382 	tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
    383 
    384 	/*
    385 	 * Pickup the right outer dst addr from the list of candidates.
    386 	 * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
    387 	 */
    388 	if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
    389 		in4 = GET_V4(&ip6->ip6_dst);
    390 	else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
    391 		in4 = GET_V4(&dst6->sin6_addr);
    392 	else {
    393 		m_freem(m);
    394 		ifp->if_oerrors++;
    395 		return ENETUNREACH;
    396 	}
    397 
    398 #if NBPFILTER > 0
    399 	if (ifp->if_bpf) {
    400 		/*
    401 		 * We need to prepend the address family as
    402 		 * a four byte field.  Cons up a dummy header
    403 		 * to pacify bpf.  This is safe because bpf
    404 		 * will only read from the mbuf (i.e., it won't
    405 		 * try to free it or keep a pointer a to it).
    406 		 */
    407 		struct mbuf m0;
    408 		u_int32_t af = AF_INET6;
    409 
    410 		m0.m_flags = 0;
    411 		m0.m_next = m;
    412 		m0.m_len = 4;
    413 		m0.m_data = (char *)&af;
    414 
    415 #ifdef HAVE_OLD_BPF
    416 		bpf_mtap(ifp, &m0);
    417 #else
    418 		bpf_mtap(ifp->if_bpf, &m0);
    419 #endif
    420 	}
    421 #endif /*NBPFILTER > 0*/
    422 
    423 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
    424 	if (m && m->m_len < sizeof(struct ip))
    425 		m = m_pullup(m, sizeof(struct ip));
    426 	if (m == NULL) {
    427 		ifp->if_oerrors++;
    428 		return ENOBUFS;
    429 	}
    430 	ip = mtod(m, struct ip *);
    431 
    432 	memset(ip, 0, sizeof(*ip));
    433 
    434 	bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
    435 	    &ip->ip_src, sizeof(ip->ip_src));
    436 	bcopy(in4, &ip->ip_dst, sizeof(ip->ip_dst));
    437 	ip->ip_p = IPPROTO_IPV6;
    438 	ip->ip_ttl = ip_gif_ttl;	/*XXX*/
    439 	ip->ip_len = htons(m->m_pkthdr.len);
    440 	if (ifp->if_flags & IFF_LINK1)
    441 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
    442 	else
    443 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
    444 
    445 	dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
    446 	if (dst4->sin_family != AF_INET ||
    447 	    bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
    448 		/* cache route doesn't match */
    449 		dst4->sin_family = AF_INET;
    450 		dst4->sin_len = sizeof(struct sockaddr_in);
    451 		bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
    452 		if (sc->sc_ro.ro_rt) {
    453 			RTFREE(sc->sc_ro.ro_rt);
    454 			sc->sc_ro.ro_rt = NULL;
    455 		}
    456 	}
    457 
    458 	if (sc->sc_ro.ro_rt == NULL) {
    459 		rtalloc(&sc->sc_ro);
    460 		if (sc->sc_ro.ro_rt == NULL) {
    461 			m_freem(m);
    462 			ifp->if_oerrors++;
    463 			return ENETUNREACH;
    464 		}
    465 	}
    466 
    467 	ifp->if_opackets++;
    468 	return ip_output(m, NULL, &sc->sc_ro, 0, NULL);
    469 }
    470 
    471 static int
    472 isrfc1918addr(in)
    473 	struct in_addr *in;
    474 {
    475 	/*
    476 	 * returns 1 if private address range:
    477 	 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16
    478 	 */
    479 	if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 ||
    480 	    (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 ||
    481 	    (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)
    482 		return 1;
    483 
    484 	return 0;
    485 }
    486 
    487 static int
    488 stf_checkaddr4(sc, in, inifp)
    489 	struct stf_softc *sc;
    490 	struct in_addr *in;
    491 	struct ifnet *inifp;	/* incoming interface */
    492 {
    493 	struct in_ifaddr *ia4;
    494 
    495 	/*
    496 	 * reject packets with the following address:
    497 	 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
    498 	 */
    499 	if (IN_MULTICAST(in->s_addr))
    500 		return -1;
    501 	switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
    502 	case 0: case 127: case 255:
    503 		return -1;
    504 	}
    505 
    506 	/*
    507 	 * reject packets with private address range.
    508 	 * (requirement from RFC3056 section 2 1st paragraph)
    509 	 */
    510 	if (isrfc1918addr(in))
    511 		return -1;
    512 
    513 	/*
    514 	 * reject packet with IPv4 link-local (169.254.0.0/16),
    515 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
    516 	 */
    517 	if (((ntohl(in->s_addr) & 0xff000000) >> 24) == 169 &&
    518 	    ((ntohl(in->s_addr) & 0x00ff0000) >> 16) == 254)
    519 		return -1;
    520 
    521 	/*
    522 	 * reject packets with broadcast
    523 	 */
    524 	TAILQ_FOREACH(ia4, &in_ifaddr, ia_list)
    525 	{
    526 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
    527 			continue;
    528 		if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr)
    529 			return -1;
    530 	}
    531 
    532 	/*
    533 	 * perform ingress filter
    534 	 */
    535 	if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
    536 		struct sockaddr_in sin;
    537 		struct rtentry *rt;
    538 
    539 		memset(&sin, 0, sizeof(sin));
    540 		sin.sin_family = AF_INET;
    541 		sin.sin_len = sizeof(struct sockaddr_in);
    542 		sin.sin_addr = *in;
    543 		rt = rtalloc1((struct sockaddr *)&sin, 0);
    544 		if (!rt || rt->rt_ifp != inifp) {
    545 #if 0
    546 			log(LOG_WARNING, "%s: packet from 0x%x dropped "
    547 			    "due to ingress filter\n", if_name(&sc->sc_if),
    548 			    (u_int32_t)ntohl(sin.sin_addr.s_addr));
    549 #endif
    550 			if (rt)
    551 				rtfree(rt);
    552 			return -1;
    553 		}
    554 		rtfree(rt);
    555 	}
    556 
    557 	return 0;
    558 }
    559 
    560 static int
    561 stf_checkaddr6(sc, in6, inifp)
    562 	struct stf_softc *sc;
    563 	struct in6_addr *in6;
    564 	struct ifnet *inifp;	/* incoming interface */
    565 {
    566 
    567 	/*
    568 	 * check 6to4 addresses
    569 	 */
    570 	if (IN6_IS_ADDR_6TO4(in6))
    571 		return stf_checkaddr4(sc, GET_V4(in6), inifp);
    572 
    573 	/*
    574 	 * reject anything that look suspicious.  the test is implemented
    575 	 * in ip6_input too, but we check here as well to
    576 	 * (1) reject bad packets earlier, and
    577 	 * (2) to be safe against future ip6_input change.
    578 	 */
    579 	if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
    580 		return -1;
    581 
    582 	/*
    583 	 * reject link-local and site-local unicast
    584 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
    585 	 */
    586 	if (IN6_IS_ADDR_LINKLOCAL(in6) || IN6_IS_ADDR_SITELOCAL(in6))
    587 		return -1;
    588 
    589 	/*
    590 	 * reject node-local and link-local multicast
    591 	 * as suggested in draft-savola-v6ops-6to4-security-00.txt
    592 	 */
    593 	if (IN6_IS_ADDR_MC_NODELOCAL(in6) || IN6_IS_ADDR_MC_LINKLOCAL(in6))
    594 		return -1;
    595 
    596 	return 0;
    597 }
    598 
    599 void
    600 #if __STDC__
    601 in_stf_input(struct mbuf *m, ...)
    602 #else
    603 in_stf_input(m, va_alist)
    604 	struct mbuf *m;
    605 #endif
    606 {
    607 	int off, proto;
    608 	struct stf_softc *sc;
    609 	struct ip *ip;
    610 	struct ip6_hdr *ip6;
    611 	u_int8_t otos, itos;
    612 	int s, isr;
    613 	struct ifqueue *ifq = NULL;
    614 	struct ifnet *ifp;
    615 	va_list ap;
    616 
    617 	va_start(ap, m);
    618 	off = va_arg(ap, int);
    619 	proto = va_arg(ap, int);
    620 	va_end(ap);
    621 
    622 	if (proto != IPPROTO_IPV6) {
    623 		m_freem(m);
    624 		return;
    625 	}
    626 
    627 	ip = mtod(m, struct ip *);
    628 
    629 	sc = (struct stf_softc *)encap_getarg(m);
    630 
    631 	if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
    632 		m_freem(m);
    633 		return;
    634 	}
    635 
    636 	ifp = &sc->sc_if;
    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;
    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;
    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;
    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 NBPFILTER > 0
    679 	if (ifp->if_bpf) {
    680 		/*
    681 		 * We need to prepend the address family as
    682 		 * a four byte field.  Cons up a dummy header
    683 		 * to pacify bpf.  This is safe because bpf
    684 		 * will only read from the mbuf (i.e., it won't
    685 		 * try to free it or keep a pointer a to it).
    686 		 */
    687 		struct mbuf m0;
    688 		u_int32_t af = AF_INET6;
    689 
    690 		m0.m_flags = 0;
    691 		m0.m_next = m;
    692 		m0.m_len = 4;
    693 		m0.m_data = (char *)&af;
    694 
    695 #ifdef HAVE_OLD_BPF
    696 		bpf_mtap(ifp, &m0);
    697 #else
    698 		bpf_mtap(ifp->if_bpf, &m0);
    699 #endif
    700 	}
    701 #endif /*NBPFILTER > 0*/
    702 
    703 	/*
    704 	 * Put the packet to the network layer input queue according to the
    705 	 * specified address family.
    706 	 * See net/if_gif.c for possible issues with packet processing
    707 	 * reorder due to extra queueing.
    708 	 */
    709 	ifq = &ip6intrq;
    710 	isr = NETISR_IPV6;
    711 
    712 	s = splnet();
    713 	if (IF_QFULL(ifq)) {
    714 		IF_DROP(ifq);	/* update statistics */
    715 		m_freem(m);
    716 		splx(s);
    717 		return;
    718 	}
    719 	IF_ENQUEUE(ifq, m);
    720 	schednetisr(isr);
    721 	ifp->if_ipackets++;
    722 	ifp->if_ibytes += m->m_pkthdr.len;
    723 	splx(s);
    724 }
    725 
    726 /* ARGSUSED */
    727 static void
    728 stf_rtrequest(cmd, rt, info)
    729 	int cmd;
    730 	struct rtentry *rt;
    731 	struct rt_addrinfo *info;
    732 {
    733 
    734 	if (rt)
    735 		rt->rt_rmx.rmx_mtu = IPV6_MMTU;
    736 }
    737 
    738 static int
    739 stf_ioctl(ifp, cmd, data)
    740 	struct ifnet *ifp;
    741 	u_long cmd;
    742 	caddr_t data;
    743 {
    744 	struct ifaddr *ifa;
    745 	struct ifreq *ifr;
    746 	struct sockaddr_in6 *sin6;
    747 	int error;
    748 
    749 	error = 0;
    750 	switch (cmd) {
    751 	case SIOCSIFADDR:
    752 		ifa = (struct ifaddr *)data;
    753 		if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
    754 			error = EAFNOSUPPORT;
    755 			break;
    756 		}
    757 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
    758 		if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr) &&
    759 		    !isrfc1918addr(GET_V4(&sin6->sin6_addr))) {
    760 			ifa->ifa_rtrequest = stf_rtrequest;
    761 			ifp->if_flags |= IFF_UP;
    762 		} else
    763 			error = EINVAL;
    764 		break;
    765 
    766 	case SIOCADDMULTI:
    767 	case SIOCDELMULTI:
    768 		ifr = (struct ifreq *)data;
    769 		if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
    770 			;
    771 		else
    772 			error = EAFNOSUPPORT;
    773 		break;
    774 
    775 	default:
    776 		error = EINVAL;
    777 		break;
    778 	}
    779 
    780 	return error;
    781 }
    782