bsd/net/if_stf.c

   1 /*      $FreeBSD: src/sys/net/if_stf.c,v 1.1.2.6 2001/07/24 19:10:18 brooks 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/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/protosw.h>
  84 #include <sys/kernel.h>
  85 #include <sys/syslog.h>
  86
  87 #include <sys/malloc.h>
  88
  89 #include <net/if.h>
  90 #include <net/route.h>
  91 #include <net/if_types.h>
  92
  93 #include <netinet/in.h>
  94 #include <netinet/in_systm.h>
  95 #include <netinet/ip.h>
  96 #include <netinet/ip_var.h>
  97 #include <netinet/in_var.h>
  98
  99 #include <netinet/ip6.h>
 100 #include <netinet6/ip6_var.h>
 101 #include <netinet6/in6_var.h>
 102 #include <netinet/ip_ecn.h>
 103
 104 #include <netinet/ip_encap.h>
 105 #include <net/dlil.h>
 106
 107
 108 #include <net/net_osdep.h>
 109
 110 #include <net/bpf.h>
 111
 112 #define IN6_IS_ADDR_6TO4(x)     (ntohs((x)->s6_addr16[0]) == 0x2002)
 113 #define GET_V4(x)       ((struct in_addr *)(&(x)->s6_addr16[1]))
 114
 115 struct stf_softc {
 116         struct ifnet    sc_if;     /* common area */
 117 #ifdef __APPLE__
 118         u_long sc_protocol_family; /* dlil protocol attached */
 119 #endif
 120         union {
 121                 struct route  __sc_ro4;
 122                 struct route_in6 __sc_ro6; /* just for safety */
 123         } __sc_ro46;
 124 #define sc_ro   __sc_ro46.__sc_ro4
 125         const struct encaptab *encap_cookie;
 126 };
 127
 128 static struct stf_softc *stf;
 129
 130 #ifdef __APPLE__
 131 void stfattach (void);
 132 #endif
 133
 134 #ifndef __APPLE__
 135 static MALLOC_DEFINE(M_STF, "stf", "6to4 Tunnel Interface");
 136 #endif
 137 static int ip_stf_ttl = 40;
 138
 139 static void in_stf_input(struct mbuf *, int);
 140 extern  struct domain inetdomain;
 141 struct protosw in_stf_protosw =
 142 { SOCK_RAW,     &inetdomain,    IPPROTO_IPV6,   PR_ATOMIC|PR_ADDR,
 143   in_stf_input, 0,      0,              rip_ctloutput,
 144   0,
 145   0,
 146   &rip_usrreqs,
 147   0,            rip_unlock,     0
 148 };
 149
 150 static int stf_encapcheck(const struct mbuf *, int, int, void *);
 151 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *);
 152 int stf_pre_output(struct ifnet *, u_long, register struct mbuf **,
 153         const struct sockaddr *, caddr_t, char *, char *);
 154 static int stf_checkaddr4(struct stf_softc *, struct in_addr *,
 155         struct ifnet *);
 156 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *,
 157         struct ifnet *);
 158 static void stf_rtrequest(int, struct rtentry *, struct sockaddr *);
 159 int stf_ioctl(struct ifnet *, u_long, void *);
 160
 161 static
 162 int  stf_add_proto(
 163         struct ifnet *ifp,
 164         u_long protocol_family,
 165         struct ddesc_head_str *desc_head)
 166 {
 167         /* Only one protocol may be attached at a time */
 168         struct stf_softc* stf = (struct stf_softc*)ifp;
 169         if (stf->sc_protocol_family == 0)
 170                 stf->sc_protocol_family = protocol_family;
 171         else {
 172                 printf("stf_add_proto: stf already has a proto\n");
 173                 return EBUSY;
 174         }
 175
 176         return 0;
 177 }
 178
 179 static
 180 int  stf_del_proto(
 181         struct ifnet *ifp,
 182         u_long protocol_family)
 183 {
 184         if (((struct stf_softc*)ifp)->sc_protocol_family == protocol_family)
 185                 ((struct stf_softc*)ifp)->sc_protocol_family = 0;
 186         else
 187                 return ENOENT;
 188
 189         return 0;
 190 }
 191
 192 static int
 193 stf_attach_inet6(struct ifnet *ifp, u_long protocol_family)
 194 {
 195     struct dlil_proto_reg_str   reg;
 196     int   stat, i;
 197
 198         bzero(&reg, sizeof(reg));
 199     TAILQ_INIT(&reg.demux_desc_head);
 200     reg.interface_family = ifp->if_family;
 201     reg.unit_number      = ifp->if_unit;
 202     reg.pre_output       = stf_pre_output;
 203     reg.protocol_family  = PF_INET6;
 204
 205     stat = dlil_attach_protocol(&reg);
 206
 207     return stat;
 208 }
 209
 210 static int
 211 stf_demux(
 212         struct ifnet *ifp,
 213         struct mbuf *m,
 214         char *frame_ptr,
 215         u_long *protocol_family)
 216 {
 217         *protocol_family = PF_INET6;
 218         return 0;
 219 }
 220
 221 void stf_reg_if_mods()
 222 {
 223         int error;
 224
 225         /* Register protocol registration functions */
 226         if ( error = dlil_reg_proto_module(AF_INET6, APPLE_IF_FAM_STF, stf_attach_inet6, NULL) != 0)
 227                 kprintf("dlil_reg_proto_module failed for AF_INET6 error=%d\n", error);
 228 }
 229
 230 void
 231 stfattach(void)
 232 {
 233         struct ifnet *ifp;
 234         struct stf_softc *sc;
 235         int i, error;
 236         int err;
 237         const struct encaptab *p;
 238
 239         stf_reg_if_mods(); /* DLIL modules */
 240
 241         sc = _MALLOC(sizeof(struct stf_softc), M_DEVBUF, M_WAITOK);
 242         if (sc == 0) {
 243                 printf("stf softc attach failed\n" );
 244                 return;
 245         }
 246
 247         bzero(sc, sizeof(*sc));
 248         sc->sc_if.if_name = "stf";
 249         sc->sc_if.if_unit = 0;
 250
 251         p = encap_attach_func(AF_INET, IPPROTO_IPV6, stf_encapcheck,
 252             &in_stf_protosw, sc);
 253         if (p == NULL) {
 254                 printf("%s: attach failed\n", if_name(&sc->sc_if));
 255                 FREE(sc, M_DEVBUF);
 256                 return;
 257         }
 258         sc->encap_cookie = p;
 259         sc->sc_if.if_mtu    = IPV6_MMTU;
 260         sc->sc_if.if_flags  = 0;
 261         sc->sc_if.if_ioctl  = stf_ioctl;
 262         sc->sc_if.if_output = NULL; /* processing done in pre_output */
 263         sc->sc_if.if_type   = IFT_STF;
 264         sc->sc_if.if_family= APPLE_IF_FAM_STF;
 265         sc->sc_if.if_add_proto = stf_add_proto;
 266         sc->sc_if.if_del_proto = stf_del_proto;
 267         sc->sc_if.if_demux = stf_demux;
 268 #if 0
 269         /* turn off ingress filter */
 270         sc->sc_if.if_flags  |= IFF_LINK2;
 271 #endif
 272         sc->sc_if.if_snd.ifq_maxlen = IFQ_MAXLEN;
 273
 274         if (error = dlil_if_attach(&sc->sc_if))
 275                 printf("stfattach: can't dlil_if_attach error=%d\n");
 276         else
 277                 bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int));
 278
 279         return ;
 280 }
 281
 282 static int
 283 stf_encapcheck(m, off, proto, arg)
 284         const struct mbuf *m;
 285         int off;
 286         int proto;
 287         void *arg;
 288 {
 289         struct ip ip;
 290         struct in6_ifaddr *ia6;
 291         struct stf_softc *sc;
 292         struct in_addr a, b;
 293
 294         sc = (struct stf_softc *)arg;
 295         if (sc == NULL)
 296                 return 0;
 297
 298         if ((sc->sc_if.if_flags & IFF_UP) == 0)
 299                 return 0;
 300
 301         /* IFF_LINK0 means "no decapsulation" */
 302         if ((sc->sc_if.if_flags & IFF_LINK0) != 0)
 303                 return 0;
 304
 305         if (proto != IPPROTO_IPV6)
 306                 return 0;
 307
 308         /* LINTED const cast */
 309         m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip);
 310
 311         if (ip.ip_v != 4)
 312                 return 0;
 313
 314         ia6 = stf_getsrcifa6(&sc->sc_if);
 315         if (ia6 == NULL)
 316                 return 0;
 317
 318         /*
 319          * check if IPv4 dst matches the IPv4 address derived from the
 320          * local 6to4 address.
 321          * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:...
 322          */
 323         if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst,
 324             sizeof(ip.ip_dst)) != 0)
 325                 return 0;
 326
 327         /*
 328          * check if IPv4 src matches the IPv4 address derived from the
 329          * local 6to4 address masked by prefixmask.
 330          * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24
 331          * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24
 332          */
 333         bzero(&a, sizeof(a));
 334         a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr;
 335         a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
 336         b = ip.ip_src;
 337         b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr;
 338         if (a.s_addr != b.s_addr)
 339                 return 0;
 340
 341         /* stf interface makes single side match only */
 342         return 32;
 343 }
 344
 345 static struct in6_ifaddr *
 346 stf_getsrcifa6(ifp)
 347         struct ifnet *ifp;
 348 {
 349         struct ifaddr *ia;
 350         struct in_ifaddr *ia4;
 351         struct sockaddr_in6 *sin6;
 352         struct in_addr in;
 353
 354         ifnet_lock_shared(ifp);
 355         for (ia = ifp->if_addrlist.tqh_first;
 356              ia;
 357              ia = ia->ifa_list.tqe_next)
 358         {
 359                 if (ia->ifa_addr == NULL)
 360                         continue;
 361                 if (ia->ifa_addr->sa_family != AF_INET6)
 362                         continue;
 363                 sin6 = (struct sockaddr_in6 *)ia->ifa_addr;
 364                 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr))
 365                         continue;
 366
 367                 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in));
 368                 lck_mtx_lock(rt_mtx);
 369                 for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
 370                      ia4;
 371                      ia4 = TAILQ_NEXT(ia4, ia_link))
 372                 {
 373                         if (ia4->ia_addr.sin_addr.s_addr == in.s_addr)
 374                                 break;
 375                 }
 376                 lck_mtx_unlock(rt_mtx);
 377                 if (ia4 == NULL)
 378                         continue;
 379
 380                 ifnet_lock_done(ifp);
 381                 return (struct in6_ifaddr *)ia;
 382         }
 383         ifnet_lock_done(ifp);
 384
 385         return NULL;
 386 }
 387
 388 int
 389 stf_pre_output(
 390         struct ifnet *ifp,
 391         u_long protocol_family,
 392         register struct mbuf **m0,
 393         const struct sockaddr *dst,
 394         caddr_t                 rt,
 395         char                 *frame_type,
 396         char                 *address)
 397 {
 398         register struct mbuf *m = *m0;
 399         struct stf_softc *sc;
 400         struct sockaddr_in6 *dst6;
 401         struct in_addr *in4;
 402         struct sockaddr_in *dst4;
 403         u_int8_t tos;
 404         struct ip *ip;
 405         struct ip6_hdr *ip6;
 406         struct in6_ifaddr *ia6;
 407         int error = 0 ;
 408
 409         sc = (struct stf_softc*)ifp;
 410         dst6 = (struct sockaddr_in6 *)dst;
 411
 412         /* just in case */
 413         if ((ifp->if_flags & IFF_UP) == 0) {
 414                 printf("stf: IFF_DOWN\n");
 415                 return ENETDOWN;
 416         }
 417
 418         /*
 419          * If we don't have an ip4 address that match my inner ip6 address,
 420          * we shouldn't generate output.  Without this check, we'll end up
 421          * using wrong IPv4 source.
 422          */
 423         ia6 = stf_getsrcifa6(ifp);
 424         if (ia6 == NULL) {
 425                 return ENETDOWN;
 426         }
 427
 428         if (m->m_len < sizeof(*ip6)) {
 429                 m = m_pullup(m, sizeof(*ip6));
 430                 if (!m)
 431                         return ENOBUFS;
 432         }
 433         ip6 = mtod(m, struct ip6_hdr *);
 434         tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
 435
 436         /*
 437          * Pickup the right outer dst addr from the list of candidates.
 438          * ip6_dst has priority as it may be able to give us shorter IPv4 hops.
 439          */
 440         if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst))
 441                 in4 = GET_V4(&ip6->ip6_dst);
 442         else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr))
 443                 in4 = GET_V4(&dst6->sin6_addr);
 444         else {
 445                 return ENETUNREACH;
 446         }
 447
 448         if (ifp->if_bpf) {
 449                 /*
 450                  * We need to prepend the address family as
 451                  * a four byte field.  Cons up a dummy header
 452                  * to pacify bpf.  This is safe because bpf
 453                  * will only read from the mbuf (i.e., it won't
 454                  * try to free it or keep a pointer a to it).
 455                  */
 456                 struct mbuf m1;
 457                 u_int32_t af = AF_INET6;
 458
 459                 m1.m_next = m;
 460                 m1.m_len = 4;
 461                 m1.m_data = (char *)&af;
 462
 463                 bpf_mtap(ifp, &m1);
 464         }
 465
 466         M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
 467         if (m && m->m_len < sizeof(struct ip))
 468                 m = m_pullup(m, sizeof(struct ip));
 469         if (m == NULL)
 470                 return ENOBUFS;
 471         ip = mtod(m, struct ip *);
 472
 473         bzero(ip, sizeof(*ip));
 474
 475         bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr),
 476             &ip->ip_src, sizeof(ip->ip_src));
 477         bcopy(in4, &ip->ip_dst, sizeof(ip->ip_dst));
 478         ip->ip_p = IPPROTO_IPV6;
 479         ip->ip_ttl = ip_stf_ttl;
 480         ip->ip_len = m->m_pkthdr.len;   /*host order*/
 481         if (ifp->if_flags & IFF_LINK1)
 482                 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
 483         else
 484                 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
 485
 486         dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst;
 487         if (dst4->sin_family != AF_INET ||
 488             bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) {
 489                 /* cache route doesn't match */
 490                 dst4->sin_family = AF_INET;
 491                 dst4->sin_len = sizeof(struct sockaddr_in);
 492                 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr));
 493                 if (sc->sc_ro.ro_rt) {
 494                         rtfree(sc->sc_ro.ro_rt);
 495                         sc->sc_ro.ro_rt = NULL;
 496                 }
 497         }
 498
 499         if (sc->sc_ro.ro_rt == NULL) {
 500                 rtalloc(&sc->sc_ro);
 501                 if (sc->sc_ro.ro_rt == NULL) {
 502                         return ENETUNREACH;
 503                 }
 504         }
 505
 506         error = ip_output(m, NULL, &sc->sc_ro, 0, NULL);
 507         if (error == 0)
 508                 return EJUSTRETURN;
 509
 510         return error;
 511 }
 512
 513 static int
 514 stf_checkaddr4(sc, in, inifp)
 515         struct stf_softc *sc;
 516         struct in_addr *in;
 517         struct ifnet *inifp;    /* incoming interface */
 518 {
 519         struct in_ifaddr *ia4;
 520
 521         /*
 522          * reject packets with the following address:
 523          * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8
 524          */
 525         if (IN_MULTICAST(ntohl(in->s_addr)))
 526                 return -1;
 527         switch ((ntohl(in->s_addr) & 0xff000000) >> 24) {
 528         case 0: case 127: case 255:
 529                 return -1;
 530         }
 531
 532         /*
 533          * reject packets with broadcast
 534          */
 535         lck_mtx_lock(rt_mtx);
 536         for (ia4 = TAILQ_FIRST(&in_ifaddrhead);
 537              ia4;
 538              ia4 = TAILQ_NEXT(ia4, ia_link))
 539         {
 540                 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
 541                         continue;
 542                 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
 543                         lck_mtx_unlock(rt_mtx);
 544                         return -1;
 545                 }
 546         }
 547         lck_mtx_unlock(rt_mtx);
 548
 549         /*
 550          * perform ingress filter
 551          */
 552         if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) {
 553                 struct sockaddr_in sin;
 554                 struct rtentry *rt;
 555
 556                 bzero(&sin, sizeof(sin));
 557                 sin.sin_family = AF_INET;
 558                 sin.sin_len = sizeof(struct sockaddr_in);
 559                 sin.sin_addr = *in;
 560                 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL);
 561                 if (!rt || rt->rt_ifp != inifp) {
 562 #if 1
 563                         log(LOG_WARNING, "%s: packet from 0x%x dropped "
 564                             "due to ingress filter\n", if_name(&sc->sc_if),
 565                             (u_int32_t)ntohl(sin.sin_addr.s_addr));
 566 #endif
 567                         if (rt)
 568                                 rtfree(rt);
 569                         return -1;
 570                 }
 571                 rtfree(rt);
 572         }
 573
 574         return 0;
 575 }
 576
 577 static int
 578 stf_checkaddr6(sc, in6, inifp)
 579         struct stf_softc *sc;
 580         struct in6_addr *in6;
 581         struct ifnet *inifp;    /* incoming interface */
 582 {
 583         /*
 584          * check 6to4 addresses
 585          */
 586         if (IN6_IS_ADDR_6TO4(in6))
 587                 return stf_checkaddr4(sc, GET_V4(in6), inifp);
 588
 589         /*
 590          * reject anything that look suspicious.  the test is implemented
 591          * in ip6_input too, but we check here as well to
 592          * (1) reject bad packets earlier, and
 593          * (2) to be safe against future ip6_input change.
 594          */
 595         if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6))
 596                 return -1;
 597
 598         return 0;
 599 }
 600
 601 static void
 602 in_stf_input(m, off)
 603         struct mbuf *m;
 604         int off;
 605 {
 606         struct stf_softc *sc;
 607         struct ip *ip;
 608         struct ip6_hdr *ip6;
 609         u_int8_t otos, itos;
 610         int proto;
 611         struct ifnet *ifp;
 612
 613         ip = mtod(m, struct ip *);
 614         proto = ip->ip_p;
 615
 616
 617         if (proto != IPPROTO_IPV6) {
 618                 m_freem(m);
 619                 return;
 620         }
 621
 622         ip = mtod(m, struct ip *);
 623
 624         sc = (struct stf_softc *)encap_getarg(m);
 625
 626         if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) {
 627                 m_freem(m);
 628                 return;
 629         }
 630
 631         ifp = &sc->sc_if;
 632
 633         /*
 634          * perform sanity check against outer src/dst.
 635          * for source, perform ingress filter as well.
 636          */
 637         if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 ||
 638             stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) {
 639                 m_freem(m);
 640                 return;
 641         }
 642
 643         otos = ip->ip_tos;
 644         m_adj(m, off);
 645
 646         if (m->m_len < sizeof(*ip6)) {
 647                 m = m_pullup(m, sizeof(*ip6));
 648                 if (!m)
 649                         return;
 650         }
 651         ip6 = mtod(m, struct ip6_hdr *);
 652
 653         /*
 654          * perform sanity check against inner src/dst.
 655          * for source, perform ingress filter as well.
 656          */
 657         if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 ||
 658             stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) {
 659                 m_freem(m);
 660                 return;
 661         }
 662
 663         itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
 664         if ((ifp->if_flags & IFF_LINK1) != 0)
 665                 ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
 666         else
 667                 ip_ecn_egress(ECN_NOCARE, &otos, &itos);
 668         ip6->ip6_flow &= ~htonl(0xff << 20);
 669         ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
 670
 671         m->m_pkthdr.rcvif = ifp;
 672
 673         if (ifp->if_bpf) {
 674                 /*
 675                  * We need to prepend the address family as
 676                  * a four byte field.  Cons up a dummy header
 677                  * to pacify bpf.  This is safe because bpf
 678                  * will only read from the mbuf (i.e., it won't
 679                  * try to free it or keep a pointer a to it).
 680                  */
 681                 struct mbuf m0;
 682                 u_int32_t af = AF_INET6;
 683
 684                 m0.m_next = m;
 685                 m0.m_len = 4;
 686                 m0.m_data = (char *)&af;
 687
 688 #ifdef HAVE_OLD_BPF
 689                 bpf_mtap(ifp, &m0);
 690 #else
 691                 bpf_mtap(ifp->if_bpf, &m0);
 692 #endif
 693         }
 694
 695         /*
 696          * Put the packet to the network layer input queue according to the
 697          * specified address family.
 698          * See net/if_gif.c for possible issues with packet processing
 699          * reorder due to extra queueing.
 700          */
 701         proto_input(PF_INET6, m);
 702         ifp->if_ipackets++;
 703         ifp->if_ibytes += m->m_pkthdr.len;
 704
 705         return;
 706 }
 707
 708 /* ARGSUSED */
 709 static void
 710 stf_rtrequest(cmd, rt, sa)
 711         int cmd;
 712         struct rtentry *rt;
 713         struct sockaddr *sa;
 714 {
 715
 716         if (rt)
 717                 rt->rt_rmx.rmx_mtu = IPV6_MMTU;
 718 }
 719
 720 int
 721 stf_ioctl(ifp, cmd, data)
 722         struct ifnet *ifp;
 723         u_long cmd;
 724         void *data;
 725 {
 726         struct ifaddr *ifa;
 727         struct ifreq *ifr;
 728         struct sockaddr_in6 *sin6;
 729         int error;
 730
 731         error = 0;
 732         switch (cmd) {
 733         case SIOCSIFADDR:
 734                 ifa = (struct ifaddr *)data;
 735                 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) {
 736                         error = EAFNOSUPPORT;
 737                         break;
 738                 }
 739                 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
 740                 if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) {
 741                         if ( !(ifnet_flags( ifp ) & IFF_UP) ) {
 742                                 /* do this only if the interface is not already up */
 743                                 ifa->ifa_rtrequest = stf_rtrequest;
 744                                 ifnet_set_flags(ifp, IFF_UP, IFF_UP);
 745                         }
 746                 } else
 747                         error = EINVAL;
 748                 break;
 749
 750         case SIOCADDMULTI:
 751         case SIOCDELMULTI:
 752                 ifr = (struct ifreq *)data;
 753                 if (ifr && ifr->ifr_addr.sa_family == AF_INET6)
 754                         ;
 755                 else
 756                         error = EAFNOSUPPORT;
 757                 break;
 758
 759         default:
 760                 error = EINVAL;
 761                 break;
 762         }
 763
 764         return error;
 765 }