if_stf.c revision 195022
1/* $FreeBSD: head/sys/net/if_stf.c 195022 2009-06-26 00:45:20Z rwatson $ */ 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 80#include <sys/param.h> 81#include <sys/systm.h> 82#include <sys/socket.h> 83#include <sys/sockio.h> 84#include <sys/mbuf.h> 85#include <sys/errno.h> 86#include <sys/kernel.h> 87#include <sys/module.h> 88#include <sys/protosw.h> 89#include <sys/proc.h> 90#include <sys/queue.h> 91#include <sys/sysctl.h> 92#include <machine/cpu.h> 93 94#include <sys/malloc.h> 95#include <sys/vimage.h> 96 97#include <net/if.h> 98#include <net/if_clone.h> 99#include <net/route.h> 100#include <net/netisr.h> 101#include <net/if_types.h> 102#include <net/if_stf.h> 103 104#include <netinet/in.h> 105#include <netinet/in_systm.h> 106#include <netinet/ip.h> 107#include <netinet/ip_var.h> 108#include <netinet/in_var.h> 109#include <netinet/vinet.h> 110 111#include <netinet/ip6.h> 112#include <netinet6/ip6_var.h> 113#include <netinet6/in6_var.h> 114#include <netinet/ip_ecn.h> 115 116#include <netinet/ip_encap.h> 117 118#include <machine/stdarg.h> 119 120#include <net/bpf.h> 121 122#include <security/mac/mac_framework.h> 123 124SYSCTL_DECL(_net_link); 125SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface"); 126 127static int stf_route_cache = 1; 128SYSCTL_INT(_net_link_stf, OID_AUTO, route_cache, CTLFLAG_RW, 129 &stf_route_cache, 0, "Caching of IPv4 routes for 6to4 Output"); 130 131#define STFNAME "stf" 132#define STFUNIT 0 133 134#define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002) 135 136/* 137 * XXX: Return a pointer with 16-bit aligned. Don't cast it to 138 * struct in_addr *; use bcopy() instead. 139 */ 140#define GET_V4(x) ((caddr_t)(&(x)->s6_addr16[1])) 141 142struct stf_softc { 143 struct ifnet *sc_ifp; 144 union { 145 struct route __sc_ro4; 146 struct route_in6 __sc_ro6; /* just for safety */ 147 } __sc_ro46; 148#define sc_ro __sc_ro46.__sc_ro4 149 struct mtx sc_ro_mtx; 150 u_int sc_fibnum; 151 const struct encaptab *encap_cookie; 152}; 153#define STF2IFP(sc) ((sc)->sc_ifp) 154 155/* 156 * Note that mutable fields in the softc are not currently locked. 157 * We do lock sc_ro in stf_output though. 158 */ 159static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface"); 160static const int ip_stf_ttl = 40; 161 162extern struct domain inetdomain; 163struct protosw in_stf_protosw = { 164 .pr_type = SOCK_RAW, 165 .pr_domain = &inetdomain, 166 .pr_protocol = IPPROTO_IPV6, 167 .pr_flags = PR_ATOMIC|PR_ADDR, 168 .pr_input = in_stf_input, 169 .pr_output = (pr_output_t *)rip_output, 170 .pr_ctloutput = rip_ctloutput, 171 .pr_usrreqs = &rip_usrreqs 172}; 173 174static char *stfnames[] = {"stf0", "stf", "6to4", NULL}; 175 176static int stfmodevent(module_t, int, void *); 177static int stf_encapcheck(const struct mbuf *, int, int, void *); 178static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *); 179static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *, 180 struct route *); 181static int isrfc1918addr(struct in_addr *); 182static int stf_checkaddr4(struct stf_softc *, struct in_addr *, 183 struct ifnet *); 184static int stf_checkaddr6(struct stf_softc *, struct in6_addr *, 185 struct ifnet *); 186static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 187static int stf_ioctl(struct ifnet *, u_long, caddr_t); 188 189static int stf_clone_match(struct if_clone *, const char *); 190static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t); 191static int stf_clone_destroy(struct if_clone *, struct ifnet *); 192struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0, 193 NULL, stf_clone_match, stf_clone_create, stf_clone_destroy); 194 195static int 196stf_clone_match(struct if_clone *ifc, const char *name) 197{ 198 int i; 199 200 for(i = 0; stfnames[i] != NULL; i++) { 201 if (strcmp(stfnames[i], name) == 0) 202 return (1); 203 } 204 205 return (0); 206} 207 208static int 209stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params) 210{ 211 int err, unit; 212 struct stf_softc *sc; 213 struct ifnet *ifp; 214 215 /* 216 * We can only have one unit, but since unit allocation is 217 * already locked, we use it to keep from allocating extra 218 * interfaces. 219 */ 220 unit = STFUNIT; 221 err = ifc_alloc_unit(ifc, &unit); 222 if (err != 0) 223 return (err); 224 225 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO); 226 ifp = STF2IFP(sc) = if_alloc(IFT_STF); 227 if (ifp == NULL) { 228 free(sc, M_STF); 229 ifc_free_unit(ifc, unit); 230 return (ENOSPC); 231 } 232 ifp->if_softc = sc; 233 sc->sc_fibnum = curthread->td_proc->p_fibnum; 234 235 /* 236 * Set the name manually rather then using if_initname because 237 * we don't conform to the default naming convention for interfaces. 238 */ 239 strlcpy(ifp->if_xname, name, IFNAMSIZ); 240 ifp->if_dname = ifc->ifc_name; 241 ifp->if_dunit = IF_DUNIT_NONE; 242 243 mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF); 244 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6, 245 stf_encapcheck, &in_stf_protosw, sc); 246 if (sc->encap_cookie == NULL) { 247 if_printf(ifp, "attach failed\n"); 248 free(sc, M_STF); 249 ifc_free_unit(ifc, unit); 250 return (ENOMEM); 251 } 252 253 ifp->if_mtu = IPV6_MMTU; 254 ifp->if_ioctl = stf_ioctl; 255 ifp->if_output = stf_output; 256 ifp->if_snd.ifq_maxlen = IFQ_MAXLEN; 257 if_attach(ifp); 258 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 259 return (0); 260} 261 262static int 263stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp) 264{ 265 struct stf_softc *sc = ifp->if_softc; 266 int err; 267 268 err = encap_detach(sc->encap_cookie); 269 KASSERT(err == 0, ("Unexpected error detaching encap_cookie")); 270 mtx_destroy(&(sc)->sc_ro_mtx); 271 bpfdetach(ifp); 272 if_detach(ifp); 273 if_free(ifp); 274 275 free(sc, M_STF); 276 ifc_free_unit(ifc, STFUNIT); 277 278 return (0); 279} 280 281static int 282stfmodevent(mod, type, data) 283 module_t mod; 284 int type; 285 void *data; 286{ 287 288 switch (type) { 289 case MOD_LOAD: 290 if_clone_attach(&stf_cloner); 291 break; 292 case MOD_UNLOAD: 293 if_clone_detach(&stf_cloner); 294 break; 295 default: 296 return (EOPNOTSUPP); 297 } 298 299 return (0); 300} 301 302static moduledata_t stf_mod = { 303 "if_stf", 304 stfmodevent, 305 0 306}; 307 308DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 309 310static int 311stf_encapcheck(m, off, proto, arg) 312 const struct mbuf *m; 313 int off; 314 int proto; 315 void *arg; 316{ 317 struct ip ip; 318 struct in6_ifaddr *ia6; 319 struct stf_softc *sc; 320 struct in_addr a, b, mask; 321 322 sc = (struct stf_softc *)arg; 323 if (sc == NULL) 324 return 0; 325 326 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0) 327 return 0; 328 329 /* IFF_LINK0 means "no decapsulation" */ 330 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0) 331 return 0; 332 333 if (proto != IPPROTO_IPV6) 334 return 0; 335 336 /* LINTED const cast */ 337 m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip); 338 339 if (ip.ip_v != 4) 340 return 0; 341 342 ia6 = stf_getsrcifa6(STF2IFP(sc)); 343 if (ia6 == NULL) 344 return 0; 345 346 /* 347 * check if IPv4 dst matches the IPv4 address derived from the 348 * local 6to4 address. 349 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:... 350 */ 351 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst, 352 sizeof(ip.ip_dst)) != 0) { 353 ifa_free(&ia6->ia_ifa); 354 return 0; 355 } 356 357 /* 358 * check if IPv4 src matches the IPv4 address derived from the 359 * local 6to4 address masked by prefixmask. 360 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24 361 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24 362 */ 363 bzero(&a, sizeof(a)); 364 bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a)); 365 bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask)); 366 ifa_free(&ia6->ia_ifa); 367 a.s_addr &= mask.s_addr; 368 b = ip.ip_src; 369 b.s_addr &= mask.s_addr; 370 if (a.s_addr != b.s_addr) 371 return 0; 372 373 /* stf interface makes single side match only */ 374 return 32; 375} 376 377static struct in6_ifaddr * 378stf_getsrcifa6(ifp) 379 struct ifnet *ifp; 380{ 381 INIT_VNET_INET(ifp->if_vnet); 382 struct ifaddr *ia; 383 struct in_ifaddr *ia4; 384 struct sockaddr_in6 *sin6; 385 struct in_addr in; 386 387 if_addr_rlock(ifp); 388 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 389 if (ia->ifa_addr->sa_family != AF_INET6) 390 continue; 391 sin6 = (struct sockaddr_in6 *)ia->ifa_addr; 392 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) 393 continue; 394 395 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in)); 396 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash) 397 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr) 398 break; 399 if (ia4 == NULL) 400 continue; 401 402 ifa_ref(ia); 403 if_addr_runlock(ifp); 404 return (struct in6_ifaddr *)ia; 405 } 406 if_addr_runlock(ifp); 407 408 return NULL; 409} 410 411static int 412stf_output(ifp, m, dst, ro) 413 struct ifnet *ifp; 414 struct mbuf *m; 415 struct sockaddr *dst; 416 struct route *ro; 417{ 418 struct stf_softc *sc; 419 struct sockaddr_in6 *dst6; 420 struct route *cached_route; 421 struct in_addr in4; 422 caddr_t ptr; 423 struct sockaddr_in *dst4; 424 u_int8_t tos; 425 struct ip *ip; 426 struct ip6_hdr *ip6; 427 struct in6_ifaddr *ia6; 428 u_int32_t af; 429 int error; 430 431#ifdef MAC 432 error = mac_ifnet_check_transmit(ifp, m); 433 if (error) { 434 m_freem(m); 435 return (error); 436 } 437#endif 438 439 sc = ifp->if_softc; 440 dst6 = (struct sockaddr_in6 *)dst; 441 442 /* just in case */ 443 if ((ifp->if_flags & IFF_UP) == 0) { 444 m_freem(m); 445 ifp->if_oerrors++; 446 return ENETDOWN; 447 } 448 449 /* 450 * If we don't have an ip4 address that match my inner ip6 address, 451 * we shouldn't generate output. Without this check, we'll end up 452 * using wrong IPv4 source. 453 */ 454 ia6 = stf_getsrcifa6(ifp); 455 if (ia6 == NULL) { 456 m_freem(m); 457 ifp->if_oerrors++; 458 return ENETDOWN; 459 } 460 461 if (m->m_len < sizeof(*ip6)) { 462 m = m_pullup(m, sizeof(*ip6)); 463 if (!m) { 464 ifa_free(&ia6->ia_ifa); 465 ifp->if_oerrors++; 466 return ENOBUFS; 467 } 468 } 469 ip6 = mtod(m, struct ip6_hdr *); 470 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 471 472 /* 473 * BPF writes need to be handled specially. 474 * This is a null operation, nothing here checks dst->sa_family. 475 */ 476 if (dst->sa_family == AF_UNSPEC) { 477 bcopy(dst->sa_data, &af, sizeof(af)); 478 dst->sa_family = af; 479 } 480 481 /* 482 * Pickup the right outer dst addr from the list of candidates. 483 * ip6_dst has priority as it may be able to give us shorter IPv4 hops. 484 */ 485 ptr = NULL; 486 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst)) 487 ptr = GET_V4(&ip6->ip6_dst); 488 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr)) 489 ptr = GET_V4(&dst6->sin6_addr); 490 else { 491 ifa_free(&ia6->ia_ifa); 492 m_freem(m); 493 ifp->if_oerrors++; 494 return ENETUNREACH; 495 } 496 bcopy(ptr, &in4, sizeof(in4)); 497 498 if (bpf_peers_present(ifp->if_bpf)) { 499 /* 500 * We need to prepend the address family as 501 * a four byte field. Cons up a dummy header 502 * to pacify bpf. This is safe because bpf 503 * will only read from the mbuf (i.e., it won't 504 * try to free it or keep a pointer a to it). 505 */ 506 af = AF_INET6; 507 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 508 } 509 510 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 511 if (m && m->m_len < sizeof(struct ip)) 512 m = m_pullup(m, sizeof(struct ip)); 513 if (m == NULL) { 514 ifa_free(&ia6->ia_ifa); 515 ifp->if_oerrors++; 516 return ENOBUFS; 517 } 518 ip = mtod(m, struct ip *); 519 520 bzero(ip, sizeof(*ip)); 521 522 bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr), 523 &ip->ip_src, sizeof(ip->ip_src)); 524 ifa_free(&ia6->ia_ifa); 525 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst)); 526 ip->ip_p = IPPROTO_IPV6; 527 ip->ip_ttl = ip_stf_ttl; 528 ip->ip_len = m->m_pkthdr.len; /*host order*/ 529 if (ifp->if_flags & IFF_LINK1) 530 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos); 531 else 532 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos); 533 534 if (!stf_route_cache) { 535 cached_route = NULL; 536 goto sendit; 537 } 538 539 /* 540 * Do we have a cached route? 541 */ 542 mtx_lock(&(sc)->sc_ro_mtx); 543 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst; 544 if (dst4->sin_family != AF_INET || 545 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) { 546 /* cache route doesn't match */ 547 dst4->sin_family = AF_INET; 548 dst4->sin_len = sizeof(struct sockaddr_in); 549 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr)); 550 if (sc->sc_ro.ro_rt) { 551 RTFREE(sc->sc_ro.ro_rt); 552 sc->sc_ro.ro_rt = NULL; 553 } 554 } 555 556 if (sc->sc_ro.ro_rt == NULL) { 557 rtalloc_fib(&sc->sc_ro, sc->sc_fibnum); 558 if (sc->sc_ro.ro_rt == NULL) { 559 m_freem(m); 560 mtx_unlock(&(sc)->sc_ro_mtx); 561 ifp->if_oerrors++; 562 return ENETUNREACH; 563 } 564 } 565 cached_route = &sc->sc_ro; 566 567sendit: 568 M_SETFIB(m, sc->sc_fibnum); 569 ifp->if_opackets++; 570 error = ip_output(m, NULL, cached_route, 0, NULL, NULL); 571 572 if (cached_route != NULL) 573 mtx_unlock(&(sc)->sc_ro_mtx); 574 return error; 575} 576 577static int 578isrfc1918addr(in) 579 struct in_addr *in; 580{ 581 /* 582 * returns 1 if private address range: 583 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16 584 */ 585 if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 || 586 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 || 587 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168) 588 return 1; 589 590 return 0; 591} 592 593static int 594stf_checkaddr4(sc, in, inifp) 595 struct stf_softc *sc; 596 struct in_addr *in; 597 struct ifnet *inifp; /* incoming interface */ 598{ 599 INIT_VNET_INET(curvnet); 600 struct in_ifaddr *ia4; 601 602 /* 603 * reject packets with the following address: 604 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8 605 */ 606 if (IN_MULTICAST(ntohl(in->s_addr))) 607 return -1; 608 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) { 609 case 0: case 127: case 255: 610 return -1; 611 } 612 613 /* 614 * reject packets with private address range. 615 * (requirement from RFC3056 section 2 1st paragraph) 616 */ 617 if (isrfc1918addr(in)) 618 return -1; 619 620 /* 621 * reject packets with broadcast 622 */ 623 IN_IFADDR_RLOCK(); 624 for (ia4 = TAILQ_FIRST(&V_in_ifaddrhead); 625 ia4; 626 ia4 = TAILQ_NEXT(ia4, ia_link)) 627 { 628 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) 629 continue; 630 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) { 631 IN_IFADDR_RUNLOCK(); 632 return -1; 633 } 634 } 635 IN_IFADDR_RUNLOCK(); 636 637 /* 638 * perform ingress filter 639 */ 640 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) { 641 struct sockaddr_in sin; 642 struct rtentry *rt; 643 644 bzero(&sin, sizeof(sin)); 645 sin.sin_family = AF_INET; 646 sin.sin_len = sizeof(struct sockaddr_in); 647 sin.sin_addr = *in; 648 rt = rtalloc1_fib((struct sockaddr *)&sin, 0, 649 0UL, sc->sc_fibnum); 650 if (!rt || rt->rt_ifp != inifp) { 651#if 0 652 log(LOG_WARNING, "%s: packet from 0x%x dropped " 653 "due to ingress filter\n", if_name(STF2IFP(sc)), 654 (u_int32_t)ntohl(sin.sin_addr.s_addr)); 655#endif 656 if (rt) 657 RTFREE_LOCKED(rt); 658 return -1; 659 } 660 RTFREE_LOCKED(rt); 661 } 662 663 return 0; 664} 665 666static int 667stf_checkaddr6(sc, in6, inifp) 668 struct stf_softc *sc; 669 struct in6_addr *in6; 670 struct ifnet *inifp; /* incoming interface */ 671{ 672 /* 673 * check 6to4 addresses 674 */ 675 if (IN6_IS_ADDR_6TO4(in6)) { 676 struct in_addr in4; 677 bcopy(GET_V4(in6), &in4, sizeof(in4)); 678 return stf_checkaddr4(sc, &in4, inifp); 679 } 680 681 /* 682 * reject anything that look suspicious. the test is implemented 683 * in ip6_input too, but we check here as well to 684 * (1) reject bad packets earlier, and 685 * (2) to be safe against future ip6_input change. 686 */ 687 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6)) 688 return -1; 689 690 return 0; 691} 692 693void 694in_stf_input(m, off) 695 struct mbuf *m; 696 int off; 697{ 698 int proto; 699 struct stf_softc *sc; 700 struct ip *ip; 701 struct ip6_hdr *ip6; 702 u_int8_t otos, itos; 703 struct ifnet *ifp; 704 705 proto = mtod(m, struct ip *)->ip_p; 706 707 if (proto != IPPROTO_IPV6) { 708 m_freem(m); 709 return; 710 } 711 712 ip = mtod(m, struct ip *); 713 714 sc = (struct stf_softc *)encap_getarg(m); 715 716 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) { 717 m_freem(m); 718 return; 719 } 720 721 ifp = STF2IFP(sc); 722 723#ifdef MAC 724 mac_ifnet_create_mbuf(ifp, m); 725#endif 726 727 /* 728 * perform sanity check against outer src/dst. 729 * for source, perform ingress filter as well. 730 */ 731 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 || 732 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) { 733 m_freem(m); 734 return; 735 } 736 737 otos = ip->ip_tos; 738 m_adj(m, off); 739 740 if (m->m_len < sizeof(*ip6)) { 741 m = m_pullup(m, sizeof(*ip6)); 742 if (!m) 743 return; 744 } 745 ip6 = mtod(m, struct ip6_hdr *); 746 747 /* 748 * perform sanity check against inner src/dst. 749 * for source, perform ingress filter as well. 750 */ 751 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 || 752 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) { 753 m_freem(m); 754 return; 755 } 756 757 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 758 if ((ifp->if_flags & IFF_LINK1) != 0) 759 ip_ecn_egress(ECN_ALLOWED, &otos, &itos); 760 else 761 ip_ecn_egress(ECN_NOCARE, &otos, &itos); 762 ip6->ip6_flow &= ~htonl(0xff << 20); 763 ip6->ip6_flow |= htonl((u_int32_t)itos << 20); 764 765 m->m_pkthdr.rcvif = ifp; 766 767 if (bpf_peers_present(ifp->if_bpf)) { 768 /* 769 * We need to prepend the address family as 770 * a four byte field. Cons up a dummy header 771 * to pacify bpf. This is safe because bpf 772 * will only read from the mbuf (i.e., it won't 773 * try to free it or keep a pointer a to it). 774 */ 775 u_int32_t af = AF_INET6; 776 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 777 } 778 779 /* 780 * Put the packet to the network layer input queue according to the 781 * specified address family. 782 * See net/if_gif.c for possible issues with packet processing 783 * reorder due to extra queueing. 784 */ 785 ifp->if_ipackets++; 786 ifp->if_ibytes += m->m_pkthdr.len; 787 netisr_dispatch(NETISR_IPV6, m); 788} 789 790/* ARGSUSED */ 791static void 792stf_rtrequest(cmd, rt, info) 793 int cmd; 794 struct rtentry *rt; 795 struct rt_addrinfo *info; 796{ 797 RT_LOCK_ASSERT(rt); 798 rt->rt_rmx.rmx_mtu = IPV6_MMTU; 799} 800 801static int 802stf_ioctl(ifp, cmd, data) 803 struct ifnet *ifp; 804 u_long cmd; 805 caddr_t data; 806{ 807 struct ifaddr *ifa; 808 struct ifreq *ifr; 809 struct sockaddr_in6 *sin6; 810 struct in_addr addr; 811 int error; 812 813 error = 0; 814 switch (cmd) { 815 case SIOCSIFADDR: 816 ifa = (struct ifaddr *)data; 817 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) { 818 error = EAFNOSUPPORT; 819 break; 820 } 821 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; 822 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) { 823 error = EINVAL; 824 break; 825 } 826 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr)); 827 if (isrfc1918addr(&addr)) { 828 error = EINVAL; 829 break; 830 } 831 832 ifa->ifa_rtrequest = stf_rtrequest; 833 ifp->if_flags |= IFF_UP; 834 break; 835 836 case SIOCADDMULTI: 837 case SIOCDELMULTI: 838 ifr = (struct ifreq *)data; 839 if (ifr && ifr->ifr_addr.sa_family == AF_INET6) 840 ; 841 else 842 error = EAFNOSUPPORT; 843 break; 844 845 default: 846 error = EINVAL; 847 break; 848 } 849 850 return error; 851} 852