if_stf.c revision 147346
1/* $FreeBSD: head/sys/net/if_stf.c 147346 2005-06-13 17:17:07Z brooks $ */ 2/* $KAME: if_stf.c,v 1.73 2001/12/03 11:08:30 keiichi Exp $ */ 3 4/*- 5 * Copyright (C) 2000 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33/* 34 * 6to4 interface, based on RFC3056. 35 * 36 * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting. 37 * There is no address mapping defined from IPv6 multicast address to IPv4 38 * address. Therefore, we do not have IFF_MULTICAST on the interface. 39 * 40 * Due to the lack of address mapping for link-local addresses, we cannot 41 * throw packets toward link-local addresses (fe80::x). Also, we cannot throw 42 * packets to link-local multicast addresses (ff02::x). 43 * 44 * Here are interesting symptoms due to the lack of link-local address: 45 * 46 * Unicast routing exchange: 47 * - RIPng: Impossible. Uses link-local multicast packet toward ff02::9, 48 * and link-local addresses as nexthop. 49 * - OSPFv6: Impossible. OSPFv6 assumes that there's link-local address 50 * assigned to the link, and makes use of them. Also, HELLO packets use 51 * link-local multicast addresses (ff02::5 and ff02::6). 52 * - BGP4+: Maybe. You can only use global address as nexthop, and global 53 * address as TCP endpoint address. 54 * 55 * Multicast routing protocols: 56 * - PIM: Hello packet cannot be used to discover adjacent PIM routers. 57 * Adjacent PIM routers must be configured manually (is it really spec-wise 58 * correct thing to do?). 59 * 60 * ICMPv6: 61 * - Redirects cannot be used due to the lack of link-local address. 62 * 63 * stf interface does not have, and will not need, a link-local address. 64 * It seems to have no real benefit and does not help the above symptoms much. 65 * Even if we assign link-locals to interface, we cannot really 66 * use link-local unicast/multicast on top of 6to4 cloud (since there's no 67 * encapsulation defined for link-local address), and the above analysis does 68 * not change. RFC3056 does not mandate the assignment of link-local address 69 * either. 70 * 71 * 6to4 interface has security issues. Refer to 72 * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt 73 * for details. The code tries to filter out some of malicious packets. 74 * Note that there is no way to be 100% secure. 75 */ 76 77#include "opt_inet.h" 78#include "opt_inet6.h" 79#include "opt_mac.h" 80 81#include <sys/param.h> 82#include <sys/systm.h> 83#include <sys/socket.h> 84#include <sys/sockio.h> 85#include <sys/mac.h> 86#include <sys/mbuf.h> 87#include <sys/errno.h> 88#include <sys/kernel.h> 89#include <sys/module.h> 90#include <sys/protosw.h> 91#include <sys/queue.h> 92#include <machine/cpu.h> 93 94#include <sys/malloc.h> 95 96#include <net/if.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/if_stf.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/net_osdep.h> 119 120#include <net/bpf.h> 121 122#define STFNAME "stf" 123#define STFUNIT 0 124 125#define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002) 126 127/* 128 * XXX: Return a pointer with 16-bit aligned. Don't cast it to 129 * struct in_addr *; use bcopy() instead. 130 */ 131#define GET_V4(x) ((caddr_t)(&(x)->s6_addr16[1])) 132 133struct stf_softc { 134 struct ifnet *sc_ifp; 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; /* all stf's are linked */ 142}; 143#define STF2IFP(sc) ((sc)->sc_ifp) 144 145/* 146 * All mutable global variables in if_stf.c are protected by stf_mtx. 147 * XXXRW: Note that mutable fields in the softc are not currently locked: 148 * in particular, sc_ro needs to be protected from concurrent entrance 149 * of stf_output(). 150 */ 151static struct mtx stf_mtx; 152static LIST_HEAD(, stf_softc) stf_softc_list; 153 154static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface"); 155static const int ip_stf_ttl = 40; 156 157extern struct domain inetdomain; 158struct protosw in_stf_protosw = 159{ SOCK_RAW, &inetdomain, IPPROTO_IPV6, PR_ATOMIC|PR_ADDR, 160 in_stf_input, (pr_output_t*)rip_output, 0, rip_ctloutput, 161 0, 162 0, 0, 0, 0, 163 &rip_usrreqs 164}; 165 166static char *stfnames[] = {"stf0", "stf", "6to4", NULL}; 167 168static int stfmodevent(module_t, int, void *); 169static int stf_encapcheck(const struct mbuf *, int, int, void *); 170static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *); 171static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *, 172 struct rtentry *); 173static int isrfc1918addr(struct in_addr *); 174static int stf_checkaddr4(struct stf_softc *, struct in_addr *, 175 struct ifnet *); 176static int stf_checkaddr6(struct stf_softc *, struct in6_addr *, 177 struct ifnet *); 178static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 179static int stf_ioctl(struct ifnet *, u_long, caddr_t); 180 181static int stf_clone_match(struct if_clone *, const char *); 182static int stf_clone_create(struct if_clone *, char *, size_t); 183static int stf_clone_destroy(struct if_clone *, struct ifnet *); 184struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0, 185 NULL, stf_clone_match, stf_clone_create, stf_clone_destroy); 186 187static int 188stf_clone_match(struct if_clone *ifc, const char *name) 189{ 190 int i; 191 192 for(i = 0; stfnames[i] != NULL; i++) { 193 if (strcmp(stfnames[i], name) == 0) 194 return (1); 195 } 196 197 return (0); 198} 199 200static int 201stf_clone_create(struct if_clone *ifc, char *name, size_t len) 202{ 203 int err, unit; 204 struct stf_softc *sc; 205 struct ifnet *ifp; 206 207 /* 208 * We can only have one unit, but since unit allocation is 209 * already locked, we use it to keep from allocating extra 210 * interfaces. 211 */ 212 unit = STFUNIT; 213 err = ifc_alloc_unit(ifc, &unit); 214 if (err != 0) 215 return (err); 216 217 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO); 218 ifp = STF2IFP(sc) = if_alloc(IFT_STF); 219 if (ifp == NULL) { 220 free(sc, M_STF); 221 ifc_free_unit(ifc, unit); 222 return (ENOSPC); 223 } 224 ifp->if_softc = sc; 225 226 /* 227 * Set the name manually rather then using if_initname because 228 * we don't conform to the default naming convention for interfaces. 229 */ 230 strlcpy(ifp->if_xname, name, IFNAMSIZ); 231 ifp->if_dname = ifc->ifc_name; 232 ifp->if_dunit = IF_DUNIT_NONE; 233 234 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6, 235 stf_encapcheck, &in_stf_protosw, sc); 236 if (sc->encap_cookie == NULL) { 237 if_printf(ifp, "attach failed\n"); 238 free(sc, M_STF); 239 ifc_free_unit(ifc, unit); 240 return (ENOMEM); 241 } 242 243 ifp->if_mtu = IPV6_MMTU; 244 ifp->if_ioctl = stf_ioctl; 245 ifp->if_output = stf_output; 246 ifp->if_snd.ifq_maxlen = IFQ_MAXLEN; 247 if_attach(ifp); 248 bpfattach(ifp, DLT_NULL, sizeof(u_int)); 249 mtx_lock(&stf_mtx); 250 LIST_INSERT_HEAD(&stf_softc_list, sc, sc_list); 251 mtx_unlock(&stf_mtx); 252 return (0); 253} 254 255static void 256stf_destroy(struct stf_softc *sc) 257{ 258 int err; 259 260 err = encap_detach(sc->encap_cookie); 261 KASSERT(err == 0, ("Unexpected error detaching encap_cookie")); 262 bpfdetach(STF2IFP(sc)); 263 if_detach(STF2IFP(sc)); 264 if_free(STF2IFP(sc)); 265 266 free(sc, M_STF); 267} 268 269static int 270stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp) 271{ 272 struct stf_softc *sc = ifp->if_softc; 273 274 mtx_lock(&stf_mtx); 275 LIST_REMOVE(sc, sc_list); 276 mtx_unlock(&stf_mtx); 277 278 stf_destroy(sc); 279 ifc_free_unit(ifc, STFUNIT); 280 281 return (0); 282} 283 284static int 285stfmodevent(mod, type, data) 286 module_t mod; 287 int type; 288 void *data; 289{ 290 struct stf_softc *sc; 291 292 switch (type) { 293 case MOD_LOAD: 294 mtx_init(&stf_mtx, "stf_mtx", NULL, MTX_DEF); 295 LIST_INIT(&stf_softc_list); 296 if_clone_attach(&stf_cloner); 297 298 break; 299 case MOD_UNLOAD: 300 if_clone_detach(&stf_cloner); 301 302 mtx_lock(&stf_mtx); 303 while ((sc = LIST_FIRST(&stf_softc_list)) != NULL) { 304 LIST_REMOVE(sc, sc_list); 305 mtx_unlock(&stf_mtx); 306 stf_destroy(sc); 307 mtx_lock(&stf_mtx); 308 } 309 mtx_unlock(&stf_mtx); 310 mtx_destroy(&stf_mtx); 311 break; 312 default: 313 return (EOPNOTSUPP); 314 } 315 316 return (0); 317} 318 319static moduledata_t stf_mod = { 320 "if_stf", 321 stfmodevent, 322 0 323}; 324 325DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 326 327static int 328stf_encapcheck(m, off, proto, arg) 329 const struct mbuf *m; 330 int off; 331 int proto; 332 void *arg; 333{ 334 struct ip ip; 335 struct in6_ifaddr *ia6; 336 struct stf_softc *sc; 337 struct in_addr a, b, mask; 338 339 sc = (struct stf_softc *)arg; 340 if (sc == NULL) 341 return 0; 342 343 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0) 344 return 0; 345 346 /* IFF_LINK0 means "no decapsulation" */ 347 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0) 348 return 0; 349 350 if (proto != IPPROTO_IPV6) 351 return 0; 352 353 /* LINTED const cast */ 354 m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip); 355 356 if (ip.ip_v != 4) 357 return 0; 358 359 ia6 = stf_getsrcifa6(STF2IFP(sc)); 360 if (ia6 == NULL) 361 return 0; 362 363 /* 364 * check if IPv4 dst matches the IPv4 address derived from the 365 * local 6to4 address. 366 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:... 367 */ 368 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst, 369 sizeof(ip.ip_dst)) != 0) 370 return 0; 371 372 /* 373 * check if IPv4 src matches the IPv4 address derived from the 374 * local 6to4 address masked by prefixmask. 375 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24 376 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24 377 */ 378 bzero(&a, sizeof(a)); 379 bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a)); 380 bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask)); 381 a.s_addr &= mask.s_addr; 382 b = ip.ip_src; 383 b.s_addr &= mask.s_addr; 384 if (a.s_addr != b.s_addr) 385 return 0; 386 387 /* stf interface makes single side match only */ 388 return 32; 389} 390 391static struct in6_ifaddr * 392stf_getsrcifa6(ifp) 393 struct ifnet *ifp; 394{ 395 struct ifaddr *ia; 396 struct in_ifaddr *ia4; 397 struct sockaddr_in6 *sin6; 398 struct in_addr in; 399 400 for (ia = TAILQ_FIRST(&ifp->if_addrlist); 401 ia; 402 ia = TAILQ_NEXT(ia, ifa_list)) 403 { 404 if (ia->ifa_addr == NULL) 405 continue; 406 if (ia->ifa_addr->sa_family != AF_INET6) 407 continue; 408 sin6 = (struct sockaddr_in6 *)ia->ifa_addr; 409 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) 410 continue; 411 412 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in)); 413 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash) 414 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr) 415 break; 416 if (ia4 == NULL) 417 continue; 418 419 return (struct in6_ifaddr *)ia; 420 } 421 422 return NULL; 423} 424 425static int 426stf_output(ifp, m, dst, rt) 427 struct ifnet *ifp; 428 struct mbuf *m; 429 struct sockaddr *dst; 430 struct rtentry *rt; 431{ 432 struct stf_softc *sc; 433 struct sockaddr_in6 *dst6; 434 struct in_addr in4; 435 caddr_t ptr; 436 struct sockaddr_in *dst4; 437 u_int8_t tos; 438 struct ip *ip; 439 struct ip6_hdr *ip6; 440 struct in6_ifaddr *ia6; 441#ifdef MAC 442 int error; 443 444 error = mac_check_ifnet_transmit(ifp, m); 445 if (error) { 446 m_freem(m); 447 return (error); 448 } 449#endif 450 451 sc = ifp->if_softc; 452 dst6 = (struct sockaddr_in6 *)dst; 453 454 /* just in case */ 455 if ((ifp->if_flags & IFF_UP) == 0) { 456 m_freem(m); 457 ifp->if_oerrors++; 458 return ENETDOWN; 459 } 460 461 /* 462 * If we don't have an ip4 address that match my inner ip6 address, 463 * we shouldn't generate output. Without this check, we'll end up 464 * using wrong IPv4 source. 465 */ 466 ia6 = stf_getsrcifa6(ifp); 467 if (ia6 == NULL) { 468 m_freem(m); 469 ifp->if_oerrors++; 470 return ENETDOWN; 471 } 472 473 if (m->m_len < sizeof(*ip6)) { 474 m = m_pullup(m, sizeof(*ip6)); 475 if (!m) { 476 ifp->if_oerrors++; 477 return ENOBUFS; 478 } 479 } 480 ip6 = mtod(m, struct ip6_hdr *); 481 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 482 483 /* 484 * Pickup the right outer dst addr from the list of candidates. 485 * ip6_dst has priority as it may be able to give us shorter IPv4 hops. 486 */ 487 ptr = NULL; 488 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst)) 489 ptr = GET_V4(&ip6->ip6_dst); 490 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr)) 491 ptr = GET_V4(&dst6->sin6_addr); 492 else { 493 m_freem(m); 494 ifp->if_oerrors++; 495 return ENETUNREACH; 496 } 497 bcopy(ptr, &in4, sizeof(in4)); 498 499 if (ifp->if_bpf) { 500 /* 501 * We need to prepend the address family as 502 * a four byte field. Cons up a dummy header 503 * to pacify bpf. This is safe because bpf 504 * will only read from the mbuf (i.e., it won't 505 * try to free it or keep a pointer a to it). 506 */ 507 u_int32_t af = AF_INET6; 508 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 509 } 510 511 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 512 if (m && m->m_len < sizeof(struct ip)) 513 m = m_pullup(m, sizeof(struct ip)); 514 if (m == NULL) { 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 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst)); 525 ip->ip_p = IPPROTO_IPV6; 526 ip->ip_ttl = ip_stf_ttl; 527 ip->ip_len = m->m_pkthdr.len; /*host order*/ 528 if (ifp->if_flags & IFF_LINK1) 529 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos); 530 else 531 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos); 532 533 /* 534 * XXXRW: Locking of sc_ro required. 535 */ 536 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst; 537 if (dst4->sin_family != AF_INET || 538 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) { 539 /* cache route doesn't match */ 540 dst4->sin_family = AF_INET; 541 dst4->sin_len = sizeof(struct sockaddr_in); 542 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr)); 543 if (sc->sc_ro.ro_rt) { 544 RTFREE(sc->sc_ro.ro_rt); 545 sc->sc_ro.ro_rt = NULL; 546 } 547 } 548 549 if (sc->sc_ro.ro_rt == NULL) { 550 rtalloc(&sc->sc_ro); 551 if (sc->sc_ro.ro_rt == NULL) { 552 m_freem(m); 553 ifp->if_oerrors++; 554 return ENETUNREACH; 555 } 556 } 557 558 ifp->if_opackets++; 559 return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL); 560} 561 562static int 563isrfc1918addr(in) 564 struct in_addr *in; 565{ 566 /* 567 * returns 1 if private address range: 568 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16 569 */ 570 if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 || 571 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 || 572 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168) 573 return 1; 574 575 return 0; 576} 577 578static int 579stf_checkaddr4(sc, in, inifp) 580 struct stf_softc *sc; 581 struct in_addr *in; 582 struct ifnet *inifp; /* incoming interface */ 583{ 584 struct in_ifaddr *ia4; 585 586 /* 587 * reject packets with the following address: 588 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8 589 */ 590 if (IN_MULTICAST(ntohl(in->s_addr))) 591 return -1; 592 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) { 593 case 0: case 127: case 255: 594 return -1; 595 } 596 597 /* 598 * reject packets with private address range. 599 * (requirement from RFC3056 section 2 1st paragraph) 600 */ 601 if (isrfc1918addr(in)) 602 return -1; 603 604 /* 605 * reject packets with broadcast 606 */ 607 for (ia4 = TAILQ_FIRST(&in_ifaddrhead); 608 ia4; 609 ia4 = TAILQ_NEXT(ia4, ia_link)) 610 { 611 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) 612 continue; 613 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) 614 return -1; 615 } 616 617 /* 618 * perform ingress filter 619 */ 620 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) { 621 struct sockaddr_in sin; 622 struct rtentry *rt; 623 624 bzero(&sin, sizeof(sin)); 625 sin.sin_family = AF_INET; 626 sin.sin_len = sizeof(struct sockaddr_in); 627 sin.sin_addr = *in; 628 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL); 629 if (!rt || rt->rt_ifp != inifp) { 630#if 0 631 log(LOG_WARNING, "%s: packet from 0x%x dropped " 632 "due to ingress filter\n", if_name(STF2IFP(sc)), 633 (u_int32_t)ntohl(sin.sin_addr.s_addr)); 634#endif 635 if (rt) 636 rtfree(rt); 637 return -1; 638 } 639 rtfree(rt); 640 } 641 642 return 0; 643} 644 645static int 646stf_checkaddr6(sc, in6, inifp) 647 struct stf_softc *sc; 648 struct in6_addr *in6; 649 struct ifnet *inifp; /* incoming interface */ 650{ 651 /* 652 * check 6to4 addresses 653 */ 654 if (IN6_IS_ADDR_6TO4(in6)) { 655 struct in_addr in4; 656 bcopy(GET_V4(in6), &in4, sizeof(in4)); 657 return stf_checkaddr4(sc, &in4, inifp); 658 } 659 660 /* 661 * reject anything that look suspicious. the test is implemented 662 * in ip6_input too, but we check here as well to 663 * (1) reject bad packets earlier, and 664 * (2) to be safe against future ip6_input change. 665 */ 666 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6)) 667 return -1; 668 669 return 0; 670} 671 672void 673in_stf_input(m, off) 674 struct mbuf *m; 675 int off; 676{ 677 int proto; 678 struct stf_softc *sc; 679 struct ip *ip; 680 struct ip6_hdr *ip6; 681 u_int8_t otos, itos; 682 struct ifnet *ifp; 683 684 proto = mtod(m, struct ip *)->ip_p; 685 686 if (proto != IPPROTO_IPV6) { 687 m_freem(m); 688 return; 689 } 690 691 ip = mtod(m, struct ip *); 692 693 sc = (struct stf_softc *)encap_getarg(m); 694 695 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) { 696 m_freem(m); 697 return; 698 } 699 700 ifp = STF2IFP(sc); 701 702#ifdef MAC 703 mac_create_mbuf_from_ifnet(ifp, m); 704#endif 705 706 /* 707 * perform sanity check against outer src/dst. 708 * for source, perform ingress filter as well. 709 */ 710 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 || 711 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) { 712 m_freem(m); 713 return; 714 } 715 716 otos = ip->ip_tos; 717 m_adj(m, off); 718 719 if (m->m_len < sizeof(*ip6)) { 720 m = m_pullup(m, sizeof(*ip6)); 721 if (!m) 722 return; 723 } 724 ip6 = mtod(m, struct ip6_hdr *); 725 726 /* 727 * perform sanity check against inner src/dst. 728 * for source, perform ingress filter as well. 729 */ 730 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 || 731 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) { 732 m_freem(m); 733 return; 734 } 735 736 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 737 if ((ifp->if_flags & IFF_LINK1) != 0) 738 ip_ecn_egress(ECN_ALLOWED, &otos, &itos); 739 else 740 ip_ecn_egress(ECN_NOCARE, &otos, &itos); 741 ip6->ip6_flow &= ~htonl(0xff << 20); 742 ip6->ip6_flow |= htonl((u_int32_t)itos << 20); 743 744 m->m_pkthdr.rcvif = ifp; 745 746 if (ifp->if_bpf) { 747 /* 748 * We need to prepend the address family as 749 * a four byte field. Cons up a dummy header 750 * to pacify bpf. This is safe because bpf 751 * will only read from the mbuf (i.e., it won't 752 * try to free it or keep a pointer a to it). 753 */ 754 u_int32_t af = AF_INET6; 755 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 756 } 757 758 /* 759 * Put the packet to the network layer input queue according to the 760 * specified address family. 761 * See net/if_gif.c for possible issues with packet processing 762 * reorder due to extra queueing. 763 */ 764 ifp->if_ipackets++; 765 ifp->if_ibytes += m->m_pkthdr.len; 766 netisr_dispatch(NETISR_IPV6, m); 767} 768 769/* ARGSUSED */ 770static void 771stf_rtrequest(cmd, rt, info) 772 int cmd; 773 struct rtentry *rt; 774 struct rt_addrinfo *info; 775{ 776 RT_LOCK_ASSERT(rt); 777 rt->rt_rmx.rmx_mtu = IPV6_MMTU; 778} 779 780static int 781stf_ioctl(ifp, cmd, data) 782 struct ifnet *ifp; 783 u_long cmd; 784 caddr_t data; 785{ 786 struct ifaddr *ifa; 787 struct ifreq *ifr; 788 struct sockaddr_in6 *sin6; 789 struct in_addr addr; 790 int error; 791 792 error = 0; 793 switch (cmd) { 794 case SIOCSIFADDR: 795 ifa = (struct ifaddr *)data; 796 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) { 797 error = EAFNOSUPPORT; 798 break; 799 } 800 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; 801 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) { 802 error = EINVAL; 803 break; 804 } 805 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr)); 806 if (isrfc1918addr(&addr)) { 807 error = EINVAL; 808 break; 809 } 810 811 ifa->ifa_rtrequest = stf_rtrequest; 812 ifp->if_flags |= IFF_UP; 813 break; 814 815 case SIOCADDMULTI: 816 case SIOCDELMULTI: 817 ifr = (struct ifreq *)data; 818 if (ifr && ifr->ifr_addr.sa_family == AF_INET6) 819 ; 820 else 821 error = EAFNOSUPPORT; 822 break; 823 824 default: 825 error = EINVAL; 826 break; 827 } 828 829 return error; 830} 831