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_systm.h> 105 #include <netinet/ip.h> 106 #include <netinet/ip_var.h> 107 #include <netinet/in_var.h> 108 109 #include <netinet/ip6.h> 110 #include <netinet6/ip6_var.h> 111 #include <netinet6/in6_var.h> 112 #include <netinet/ip_ecn.h> 113 114 #include <netinet/ip_encap.h> 115 116 #include <machine/stdarg.h> 117 118 #include <net/bpf.h> 119 120 #include <security/mac/mac_framework.h> 121 122 SYSCTL_DECL(_net_link); 123 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface"); 124 125 static int stf_permit_rfc1918 = 0; 126 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN, 127 &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses"); 128 129 #define STFUNIT 0 130 131 #define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002) 132 133 /* 134 * XXX: Return a pointer with 16-bit aligned. Don't cast it to 135 * struct in_addr *; use bcopy() instead. 136 */ 137 #define GET_V4(x) (&(x)->s6_addr16[1]) 138 139 struct stf_softc { 140 struct ifnet *sc_ifp; 141 struct mtx sc_ro_mtx; 142 u_int sc_fibnum; 143 const struct encaptab *encap_cookie; 144 }; 145 #define STF2IFP(sc) ((sc)->sc_ifp) 146 147 static const char stfname[] = "stf"; 148 149 /* 150 * Note that mutable fields in the softc are not currently locked. 151 * We do lock sc_ro in stf_output though. 152 */ 153 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface"); 154 static const int ip_stf_ttl = 40; 155 156 extern struct domain inetdomain; 157 static int in_stf_input(struct mbuf **, int *, int); 158 static struct protosw in_stf_protosw = { 159 .pr_type = SOCK_RAW, 160 .pr_domain = &inetdomain, 161 .pr_protocol = IPPROTO_IPV6, 162 .pr_flags = PR_ATOMIC|PR_ADDR, 163 .pr_input = in_stf_input, 164 .pr_output = rip_output, 165 .pr_ctloutput = rip_ctloutput, 166 .pr_usrreqs = &rip_usrreqs 167 }; 168 169 static char *stfnames[] = {"stf0", "stf", "6to4", NULL}; 170 171 static int stfmodevent(module_t, int, void *); 172 static int stf_encapcheck(const struct mbuf *, int, int, void *); 173 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *); 174 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *, 175 struct route *); 176 static int isrfc1918addr(struct in_addr *); 177 static int stf_checkaddr4(struct stf_softc *, struct in_addr *, 178 struct ifnet *); 179 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *, 180 struct ifnet *); 181 static int stf_ioctl(struct ifnet *, u_long, caddr_t); 182 183 static int stf_clone_match(struct if_clone *, const char *); 184 static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t); 185 static int stf_clone_destroy(struct if_clone *, struct ifnet *); 186 static struct if_clone *stf_cloner; 187 188 static int 189 stf_clone_match(struct if_clone *ifc, const char *name) 190 { 191 int i; 192 193 for(i = 0; stfnames[i] != NULL; i++) { 194 if (strcmp(stfnames[i], name) == 0) 195 return (1); 196 } 197 198 return (0); 199 } 200 201 static int 202 stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params) 203 { 204 int err, unit; 205 struct stf_softc *sc; 206 struct ifnet *ifp; 207 208 /* 209 * We can only have one unit, but since unit allocation is 210 * already locked, we use it to keep from allocating extra 211 * interfaces. 212 */ 213 unit = STFUNIT; 214 err = ifc_alloc_unit(ifc, &unit); 215 if (err != 0) 216 return (err); 217 218 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO); 219 ifp = STF2IFP(sc) = if_alloc(IFT_STF); 220 if (ifp == NULL) { 221 free(sc, M_STF); 222 ifc_free_unit(ifc, unit); 223 return (ENOSPC); 224 } 225 ifp->if_softc = sc; 226 sc->sc_fibnum = curthread->td_proc->p_fibnum; 227 228 /* 229 * Set the name manually rather then using if_initname because 230 * we don't conform to the default naming convention for interfaces. 231 */ 232 strlcpy(ifp->if_xname, name, IFNAMSIZ); 233 ifp->if_dname = stfname; 234 ifp->if_dunit = IF_DUNIT_NONE; 235 236 mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF); 237 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6, 238 stf_encapcheck, &in_stf_protosw, sc); 239 if (sc->encap_cookie == NULL) { 240 if_printf(ifp, "attach failed\n"); 241 free(sc, M_STF); 242 ifc_free_unit(ifc, unit); 243 return (ENOMEM); 244 } 245 246 ifp->if_mtu = IPV6_MMTU; 247 ifp->if_ioctl = stf_ioctl; 248 ifp->if_output = stf_output; 249 ifp->if_snd.ifq_maxlen = ifqmaxlen; 250 if_attach(ifp); 251 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 252 return (0); 253 } 254 255 static int 256 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp) 257 { 258 struct stf_softc *sc = ifp->if_softc; 259 int err; 260 261 err = encap_detach(sc->encap_cookie); 262 KASSERT(err == 0, ("Unexpected error detaching encap_cookie")); 263 mtx_destroy(&(sc)->sc_ro_mtx); 264 bpfdetach(ifp); 265 if_detach(ifp); 266 if_free(ifp); 267 268 free(sc, M_STF); 269 ifc_free_unit(ifc, STFUNIT); 270 271 return (0); 272 } 273 274 static int 275 stfmodevent(mod, type, data) 276 module_t mod; 277 int type; 278 void *data; 279 { 280 281 switch (type) { 282 case MOD_LOAD: 283 stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match, 284 stf_clone_create, stf_clone_destroy); 285 break; 286 case MOD_UNLOAD: 287 if_clone_detach(stf_cloner); 288 break; 289 default: 290 return (EOPNOTSUPP); 291 } 292 293 return (0); 294 } 295 296 static moduledata_t stf_mod = { 297 "if_stf", 298 stfmodevent, 299 0 300 }; 301 302 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 303 304 static int 305 stf_encapcheck(m, off, proto, arg) 306 const struct mbuf *m; 307 int off; 308 int proto; 309 void *arg; 310 { 311 struct ip ip; 312 struct stf_softc *sc; 313 struct in_addr a, b, mask; 314 struct in6_addr addr6, mask6; 315 316 sc = (struct stf_softc *)arg; 317 if (sc == NULL) 318 return 0; 319 320 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0) 321 return 0; 322 323 /* IFF_LINK0 means "no decapsulation" */ 324 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0) 325 return 0; 326 327 if (proto != IPPROTO_IPV6) 328 return 0; 329 330 /* LINTED const cast */ 331 m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip); 332 333 if (ip.ip_v != 4) 334 return 0; 335 336 if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0) 337 return (0); 338 339 /* 340 * check if IPv4 dst matches the IPv4 address derived from the 341 * local 6to4 address. 342 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:... 343 */ 344 if (bcmp(GET_V4(&addr6), &ip.ip_dst, sizeof(ip.ip_dst)) != 0) 345 return 0; 346 347 /* 348 * check if IPv4 src matches the IPv4 address derived from the 349 * local 6to4 address masked by prefixmask. 350 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24 351 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24 352 */ 353 bzero(&a, sizeof(a)); 354 bcopy(GET_V4(&addr6), &a, sizeof(a)); 355 bcopy(GET_V4(&mask6), &mask, sizeof(mask)); 356 a.s_addr &= mask.s_addr; 357 b = ip.ip_src; 358 b.s_addr &= mask.s_addr; 359 if (a.s_addr != b.s_addr) 360 return 0; 361 362 /* stf interface makes single side match only */ 363 return 32; 364 } 365 366 static int 367 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask) 368 { 369 struct ifaddr *ia; 370 struct in_ifaddr *ia4; 371 struct in6_ifaddr *ia6; 372 struct sockaddr_in6 *sin6; 373 struct in_addr in; 374 375 if_addr_rlock(ifp); 376 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 377 if (ia->ifa_addr->sa_family != AF_INET6) 378 continue; 379 sin6 = (struct sockaddr_in6 *)ia->ifa_addr; 380 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) 381 continue; 382 383 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in)); 384 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash) 385 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr) 386 break; 387 if (ia4 == NULL) 388 continue; 389 390 ia6 = (struct in6_ifaddr *)ia; 391 392 *addr = sin6->sin6_addr; 393 *mask = ia6->ia_prefixmask.sin6_addr; 394 if_addr_runlock(ifp); 395 return (0); 396 } 397 if_addr_runlock(ifp); 398 399 return (ENOENT); 400 } 401 402 static int 403 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, 404 struct route *ro) 405 { 406 struct stf_softc *sc; 407 const struct sockaddr_in6 *dst6; 408 struct in_addr in4; 409 const void *ptr; 410 u_int8_t tos; 411 struct ip *ip; 412 struct ip6_hdr *ip6; 413 struct in6_addr addr6, mask6; 414 int error; 415 416 #ifdef MAC 417 error = mac_ifnet_check_transmit(ifp, m); 418 if (error) { 419 m_freem(m); 420 return (error); 421 } 422 #endif 423 424 sc = ifp->if_softc; 425 dst6 = (const struct sockaddr_in6 *)dst; 426 427 /* just in case */ 428 if ((ifp->if_flags & IFF_UP) == 0) { 429 m_freem(m); 430 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 431 return ENETDOWN; 432 } 433 434 /* 435 * If we don't have an ip4 address that match my inner ip6 address, 436 * we shouldn't generate output. Without this check, we'll end up 437 * using wrong IPv4 source. 438 */ 439 if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) { 440 m_freem(m); 441 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 442 return ENETDOWN; 443 } 444 445 if (m->m_len < sizeof(*ip6)) { 446 m = m_pullup(m, sizeof(*ip6)); 447 if (!m) { 448 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 449 return ENOBUFS; 450 } 451 } 452 ip6 = mtod(m, struct ip6_hdr *); 453 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 454 455 /* 456 * Pickup the right outer dst addr from the list of candidates. 457 * ip6_dst has priority as it may be able to give us shorter IPv4 hops. 458 */ 459 ptr = NULL; 460 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst)) 461 ptr = GET_V4(&ip6->ip6_dst); 462 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr)) 463 ptr = GET_V4(&dst6->sin6_addr); 464 else { 465 m_freem(m); 466 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 467 return ENETUNREACH; 468 } 469 bcopy(ptr, &in4, sizeof(in4)); 470 471 if (bpf_peers_present(ifp->if_bpf)) { 472 /* 473 * We need to prepend the address family as 474 * a four byte field. Cons up a dummy header 475 * to pacify bpf. This is safe because bpf 476 * will only read from the mbuf (i.e., it won't 477 * try to free it or keep a pointer a to it). 478 */ 479 u_int af = AF_INET6; 480 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 481 } 482 483 M_PREPEND(m, sizeof(struct ip), M_NOWAIT); 484 if (m == NULL) { 485 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 486 return ENOBUFS; 487 } 488 ip = mtod(m, struct ip *); 489 490 bzero(ip, sizeof(*ip)); 491 492 bcopy(GET_V4(&addr6), &ip->ip_src, sizeof(ip->ip_src)); 493 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst)); 494 ip->ip_p = IPPROTO_IPV6; 495 ip->ip_ttl = ip_stf_ttl; 496 ip->ip_len = htons(m->m_pkthdr.len); 497 if (ifp->if_flags & IFF_LINK1) 498 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos); 499 else 500 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos); 501 502 M_SETFIB(m, sc->sc_fibnum); 503 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); 504 error = ip_output(m, NULL, NULL, 0, NULL, NULL); 505 506 return error; 507 } 508 509 static int 510 isrfc1918addr(in) 511 struct in_addr *in; 512 { 513 /* 514 * returns 1 if private address range: 515 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16 516 */ 517 if (stf_permit_rfc1918 == 0 && ( 518 (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 || 519 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 || 520 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)) 521 return 1; 522 523 return 0; 524 } 525 526 static int 527 stf_checkaddr4(sc, in, inifp) 528 struct stf_softc *sc; 529 struct in_addr *in; 530 struct ifnet *inifp; /* incoming interface */ 531 { 532 struct rm_priotracker in_ifa_tracker; 533 struct in_ifaddr *ia4; 534 535 /* 536 * reject packets with the following address: 537 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8 538 */ 539 if (IN_MULTICAST(ntohl(in->s_addr))) 540 return -1; 541 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) { 542 case 0: case 127: case 255: 543 return -1; 544 } 545 546 /* 547 * reject packets with private address range. 548 * (requirement from RFC3056 section 2 1st paragraph) 549 */ 550 if (isrfc1918addr(in)) 551 return -1; 552 553 /* 554 * reject packets with broadcast 555 */ 556 IN_IFADDR_RLOCK(&in_ifa_tracker); 557 TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) { 558 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) 559 continue; 560 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) { 561 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 562 return -1; 563 } 564 } 565 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 566 567 /* 568 * perform ingress filter 569 */ 570 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) { 571 struct sockaddr_in sin; 572 struct rtentry *rt; 573 574 bzero(&sin, sizeof(sin)); 575 sin.sin_family = AF_INET; 576 sin.sin_len = sizeof(struct sockaddr_in); 577 sin.sin_addr = *in; 578 rt = rtalloc1_fib((struct sockaddr *)&sin, 0, 579 0UL, sc->sc_fibnum); 580 if (!rt || rt->rt_ifp != inifp) { 581 #if 0 582 log(LOG_WARNING, "%s: packet from 0x%x dropped " 583 "due to ingress filter\n", if_name(STF2IFP(sc)), 584 (u_int32_t)ntohl(sin.sin_addr.s_addr)); 585 #endif 586 if (rt) 587 RTFREE_LOCKED(rt); 588 return -1; 589 } 590 RTFREE_LOCKED(rt); 591 } 592 593 return 0; 594 } 595 596 static int 597 stf_checkaddr6(sc, in6, inifp) 598 struct stf_softc *sc; 599 struct in6_addr *in6; 600 struct ifnet *inifp; /* incoming interface */ 601 { 602 /* 603 * check 6to4 addresses 604 */ 605 if (IN6_IS_ADDR_6TO4(in6)) { 606 struct in_addr in4; 607 bcopy(GET_V4(in6), &in4, sizeof(in4)); 608 return stf_checkaddr4(sc, &in4, inifp); 609 } 610 611 /* 612 * reject anything that look suspicious. the test is implemented 613 * in ip6_input too, but we check here as well to 614 * (1) reject bad packets earlier, and 615 * (2) to be safe against future ip6_input change. 616 */ 617 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6)) 618 return -1; 619 620 return 0; 621 } 622 623 static int 624 in_stf_input(struct mbuf **mp, int *offp, int proto) 625 { 626 struct stf_softc *sc; 627 struct ip *ip; 628 struct ip6_hdr *ip6; 629 struct mbuf *m; 630 u_int8_t otos, itos; 631 struct ifnet *ifp; 632 int off; 633 634 m = *mp; 635 off = *offp; 636 637 if (proto != IPPROTO_IPV6) { 638 m_freem(m); 639 return (IPPROTO_DONE); 640 } 641 642 ip = mtod(m, struct ip *); 643 644 sc = (struct stf_softc *)encap_getarg(m); 645 646 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) { 647 m_freem(m); 648 return (IPPROTO_DONE); 649 } 650 651 ifp = STF2IFP(sc); 652 653 #ifdef MAC 654 mac_ifnet_create_mbuf(ifp, m); 655 #endif 656 657 /* 658 * perform sanity check against outer src/dst. 659 * for source, perform ingress filter as well. 660 */ 661 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 || 662 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) { 663 m_freem(m); 664 return (IPPROTO_DONE); 665 } 666 667 otos = ip->ip_tos; 668 m_adj(m, off); 669 670 if (m->m_len < sizeof(*ip6)) { 671 m = m_pullup(m, sizeof(*ip6)); 672 if (!m) 673 return (IPPROTO_DONE); 674 } 675 ip6 = mtod(m, struct ip6_hdr *); 676 677 /* 678 * perform sanity check against inner src/dst. 679 * for source, perform ingress filter as well. 680 */ 681 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 || 682 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) { 683 m_freem(m); 684 return (IPPROTO_DONE); 685 } 686 687 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 688 if ((ifp->if_flags & IFF_LINK1) != 0) 689 ip_ecn_egress(ECN_ALLOWED, &otos, &itos); 690 else 691 ip_ecn_egress(ECN_NOCARE, &otos, &itos); 692 ip6->ip6_flow &= ~htonl(0xff << 20); 693 ip6->ip6_flow |= htonl((u_int32_t)itos << 20); 694 695 m->m_pkthdr.rcvif = ifp; 696 697 if (bpf_peers_present(ifp->if_bpf)) { 698 /* 699 * We need to prepend the address family as 700 * a four byte field. Cons up a dummy header 701 * to pacify bpf. This is safe because bpf 702 * will only read from the mbuf (i.e., it won't 703 * try to free it or keep a pointer a to it). 704 */ 705 u_int32_t af = AF_INET6; 706 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 707 } 708 709 /* 710 * Put the packet to the network layer input queue according to the 711 * specified address family. 712 * See net/if_gif.c for possible issues with packet processing 713 * reorder due to extra queueing. 714 */ 715 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1); 716 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); 717 M_SETFIB(m, ifp->if_fib); 718 netisr_dispatch(NETISR_IPV6, m); 719 return (IPPROTO_DONE); 720 } 721 722 static int 723 stf_ioctl(ifp, cmd, data) 724 struct ifnet *ifp; 725 u_long cmd; 726 caddr_t data; 727 { 728 struct ifaddr *ifa; 729 struct ifreq *ifr; 730 struct sockaddr_in6 *sin6; 731 struct in_addr addr; 732 int error, mtu; 733 734 error = 0; 735 switch (cmd) { 736 case SIOCSIFADDR: 737 ifa = (struct ifaddr *)data; 738 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) { 739 error = EAFNOSUPPORT; 740 break; 741 } 742 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; 743 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) { 744 error = EINVAL; 745 break; 746 } 747 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr)); 748 if (isrfc1918addr(&addr)) { 749 error = EINVAL; 750 break; 751 } 752 753 ifp->if_flags |= IFF_UP; 754 break; 755 756 case SIOCADDMULTI: 757 case SIOCDELMULTI: 758 ifr = (struct ifreq *)data; 759 if (ifr && ifr->ifr_addr.sa_family == AF_INET6) 760 ; 761 else 762 error = EAFNOSUPPORT; 763 break; 764 765 case SIOCGIFMTU: 766 break; 767 768 case SIOCSIFMTU: 769 ifr = (struct ifreq *)data; 770 mtu = ifr->ifr_mtu; 771 /* RFC 4213 3.2 ideal world MTU */ 772 if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20) 773 return (EINVAL); 774 ifp->if_mtu = mtu; 775 break; 776 777 default: 778 error = EINVAL; 779 break; 780 } 781 782 return error; 783 } 784