1 /* $FreeBSD: src/sys/net/if_stf.c,v 1.1.2.11 2003/01/23 21:06:44 sam Exp $ */ 2 /* $DragonFly: src/sys/net/stf/if_stf.c,v 1.14 2005/01/26 00:37:39 joerg Exp $ */ 3 /* $KAME: if_stf.c,v 1.73 2001/12/03 11:08:30 keiichi Exp $ */ 4 5 /* 6 * Copyright (C) 2000 WIDE Project. 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. Neither the name of the project nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34 /* 35 * 6to4 interface, based on RFC3056. 36 * 37 * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting. 38 * There is no address mapping defined from IPv6 multicast address to IPv4 39 * address. Therefore, we do not have IFF_MULTICAST on the interface. 40 * 41 * Due to the lack of address mapping for link-local addresses, we cannot 42 * throw packets toward link-local addresses (fe80::x). Also, we cannot throw 43 * packets to link-local multicast addresses (ff02::x). 44 * 45 * Here are interesting symptoms due to the lack of link-local address: 46 * 47 * Unicast routing exchange: 48 * - RIPng: Impossible. Uses link-local multicast packet toward ff02::9, 49 * and link-local addresses as nexthop. 50 * - OSPFv6: Impossible. OSPFv6 assumes that there's link-local address 51 * assigned to the link, and makes use of them. Also, HELLO packets use 52 * link-local multicast addresses (ff02::5 and ff02::6). 53 * - BGP4+: Maybe. You can only use global address as nexthop, and global 54 * address as TCP endpoint address. 55 * 56 * Multicast routing protocols: 57 * - PIM: Hello packet cannot be used to discover adjacent PIM routers. 58 * Adjacent PIM routers must be configured manually (is it really spec-wise 59 * correct thing to do?). 60 * 61 * ICMPv6: 62 * - Redirects cannot be used due to the lack of link-local address. 63 * 64 * stf interface does not have, and will not need, a link-local address. 65 * It seems to have no real benefit and does not help the above symptoms much. 66 * Even if we assign link-locals to interface, we cannot really 67 * use link-local unicast/multicast on top of 6to4 cloud (since there's no 68 * encapsulation defined for link-local address), and the above analysis does 69 * not change. RFC3056 does not mandate the assignment of link-local address 70 * either. 71 * 72 * 6to4 interface has security issues. Refer to 73 * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt 74 * for details. The code tries to filter out some of malicious packets. 75 * Note that there is no way to be 100% secure. 76 */ 77 78 #include "opt_inet.h" 79 #include "opt_inet6.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/mbuf.h> 86 #include <sys/errno.h> 87 #include <sys/protosw.h> 88 #include <sys/kernel.h> 89 #include <machine/cpu.h> 90 91 #include <sys/malloc.h> 92 93 #include <net/if.h> 94 #include <net/route.h> 95 #include <net/netisr.h> 96 #include <net/if_types.h> 97 #include "if_stf.h" 98 99 #include <netinet/in.h> 100 #include <netinet/in_systm.h> 101 #include <netinet/ip.h> 102 #include <netinet/ip_var.h> 103 #include <netinet/in_var.h> 104 105 #include <netinet/ip6.h> 106 #include <netinet6/ip6_var.h> 107 #include <netinet6/in6_var.h> 108 #include <netinet/ip_ecn.h> 109 110 #include <netinet/ip_encap.h> 111 112 #include <machine/stdarg.h> 113 114 #include <net/net_osdep.h> 115 116 #include <net/bpf.h> 117 118 #define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002) 119 #define GET_V4(x) ((struct in_addr *)(&(x)->s6_addr16[1])) 120 121 struct stf_softc { 122 struct ifnet sc_if; /* common area */ 123 union { 124 struct route __sc_ro4; 125 struct route_in6 __sc_ro6; /* just for safety */ 126 } __sc_ro46; 127 #define sc_ro __sc_ro46.__sc_ro4 128 const struct encaptab *encap_cookie; 129 }; 130 131 static struct stf_softc *stf; 132 133 static MALLOC_DEFINE(M_STF, "stf", "6to4 Tunnel Interface"); 134 static int ip_stf_ttl = 40; 135 136 extern struct domain inetdomain; 137 struct protosw in_stf_protosw = 138 { SOCK_RAW, &inetdomain, IPPROTO_IPV6, PR_ATOMIC|PR_ADDR, 139 in_stf_input, rip_output, 0, rip_ctloutput, 140 0, 141 0, 0, 0, 0, 142 &rip_usrreqs 143 }; 144 145 static int stfmodevent (module_t, int, void *); 146 static int stf_encapcheck (const struct mbuf *, int, int, void *); 147 static struct in6_ifaddr *stf_getsrcifa6 (struct ifnet *); 148 static int stf_output (struct ifnet *, struct mbuf *, struct sockaddr *, 149 struct rtentry *); 150 static int stf_checkaddr4 (struct stf_softc *, struct in_addr *, 151 struct ifnet *); 152 static int stf_checkaddr6 (struct stf_softc *, struct in6_addr *, 153 struct ifnet *); 154 static void stf_rtrequest (int, struct rtentry *, struct rt_addrinfo *); 155 static int stf_ioctl (struct ifnet *, u_long, caddr_t, struct ucred *); 156 157 static int 158 stfmodevent(mod, type, data) 159 module_t mod; 160 int type; 161 void *data; 162 { 163 struct stf_softc *sc; 164 int err; 165 const struct encaptab *p; 166 167 switch (type) { 168 case MOD_LOAD: 169 stf = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK); 170 bzero(stf, sizeof(struct stf_softc)); 171 sc = stf; 172 173 bzero(sc, sizeof(*sc)); 174 if_initname(&(sc->sc_if), "stf", 0); 175 176 p = encap_attach_func(AF_INET, IPPROTO_IPV6, stf_encapcheck, 177 (void *)&in_stf_protosw, sc); 178 if (p == NULL) { 179 printf("%s: attach failed\n", if_name(&sc->sc_if)); 180 return (ENOMEM); 181 } 182 sc->encap_cookie = p; 183 184 sc->sc_if.if_mtu = IPV6_MMTU; 185 sc->sc_if.if_flags = 0; 186 sc->sc_if.if_ioctl = stf_ioctl; 187 sc->sc_if.if_output = stf_output; 188 sc->sc_if.if_type = IFT_STF; 189 #if 0 190 /* turn off ingress filter */ 191 sc->sc_if.if_flags |= IFF_LINK2; 192 #endif 193 sc->sc_if.if_snd.ifq_maxlen = IFQ_MAXLEN; 194 if_attach(&sc->sc_if); 195 bpfattach(&sc->sc_if, DLT_NULL, sizeof(u_int)); 196 break; 197 case MOD_UNLOAD: 198 sc = stf; 199 bpfdetach(&sc->sc_if); 200 if_detach(&sc->sc_if); 201 err = encap_detach(sc->encap_cookie); 202 KASSERT(err == 0, ("Unexpected error detaching encap_cookie")); 203 free(sc, M_STF); 204 break; 205 } 206 207 return (0); 208 } 209 210 static moduledata_t stf_mod = { 211 "if_stf", 212 stfmodevent, 213 0 214 }; 215 216 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 217 218 static int 219 stf_encapcheck(m, off, proto, arg) 220 const struct mbuf *m; 221 int off; 222 int proto; 223 void *arg; 224 { 225 struct ip ip; 226 struct in6_ifaddr *ia6; 227 struct stf_softc *sc; 228 struct in_addr a, b; 229 230 sc = (struct stf_softc *)arg; 231 if (sc == NULL) 232 return 0; 233 234 if ((sc->sc_if.if_flags & IFF_UP) == 0) 235 return 0; 236 237 /* IFF_LINK0 means "no decapsulation" */ 238 if ((sc->sc_if.if_flags & IFF_LINK0) != 0) 239 return 0; 240 241 if (proto != IPPROTO_IPV6) 242 return 0; 243 244 m_copydata(m, 0, sizeof(ip), (caddr_t)&ip); 245 246 if (ip.ip_v != 4) 247 return 0; 248 249 ia6 = stf_getsrcifa6(&sc->sc_if); 250 if (ia6 == NULL) 251 return 0; 252 253 /* 254 * check if IPv4 dst matches the IPv4 address derived from the 255 * local 6to4 address. 256 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:... 257 */ 258 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst, 259 sizeof(ip.ip_dst)) != 0) 260 return 0; 261 262 /* 263 * check if IPv4 src matches the IPv4 address derived from the 264 * local 6to4 address masked by prefixmask. 265 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24 266 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24 267 */ 268 bzero(&a, sizeof(a)); 269 a.s_addr = GET_V4(&ia6->ia_addr.sin6_addr)->s_addr; 270 a.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr; 271 b = ip.ip_src; 272 b.s_addr &= GET_V4(&ia6->ia_prefixmask.sin6_addr)->s_addr; 273 if (a.s_addr != b.s_addr) 274 return 0; 275 276 /* stf interface makes single side match only */ 277 return 32; 278 } 279 280 static struct in6_ifaddr * 281 stf_getsrcifa6(ifp) 282 struct ifnet *ifp; 283 { 284 struct ifaddr *ia; 285 struct in_ifaddr *ia4; 286 struct sockaddr_in6 *sin6; 287 struct in_addr in; 288 289 TAILQ_FOREACH(ia, &ifp->if_addrlist, ifa_list) { 290 if (ia->ifa_addr == NULL) 291 continue; 292 if (ia->ifa_addr->sa_family != AF_INET6) 293 continue; 294 sin6 = (struct sockaddr_in6 *)ia->ifa_addr; 295 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) 296 continue; 297 298 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in)); 299 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash) 300 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr) 301 break; 302 if (ia4 == NULL) 303 continue; 304 305 return (struct in6_ifaddr *)ia; 306 } 307 308 return NULL; 309 } 310 311 static int 312 stf_output(ifp, m, dst, rt) 313 struct ifnet *ifp; 314 struct mbuf *m; 315 struct sockaddr *dst; 316 struct rtentry *rt; 317 { 318 struct stf_softc *sc; 319 struct sockaddr_in6 *dst6; 320 struct in_addr *in4; 321 struct sockaddr_in *dst4; 322 u_int8_t tos; 323 struct ip *ip; 324 struct ip6_hdr *ip6; 325 struct in6_ifaddr *ia6; 326 static const uint32_t af = AF_INET6; 327 328 sc = (struct stf_softc*)ifp; 329 dst6 = (struct sockaddr_in6 *)dst; 330 331 /* just in case */ 332 if ((ifp->if_flags & IFF_UP) == 0) { 333 m_freem(m); 334 return ENETDOWN; 335 } 336 337 /* 338 * If we don't have an ip4 address that match my inner ip6 address, 339 * we shouldn't generate output. Without this check, we'll end up 340 * using wrong IPv4 source. 341 */ 342 ia6 = stf_getsrcifa6(ifp); 343 if (ia6 == NULL) { 344 m_freem(m); 345 return ENETDOWN; 346 } 347 348 if (m->m_len < sizeof(*ip6)) { 349 m = m_pullup(m, sizeof(*ip6)); 350 if (!m) 351 return ENOBUFS; 352 } 353 ip6 = mtod(m, struct ip6_hdr *); 354 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 355 356 /* 357 * Pickup the right outer dst addr from the list of candidates. 358 * ip6_dst has priority as it may be able to give us shorter IPv4 hops. 359 */ 360 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst)) 361 in4 = GET_V4(&ip6->ip6_dst); 362 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr)) 363 in4 = GET_V4(&dst6->sin6_addr); 364 else { 365 m_freem(m); 366 return ENETUNREACH; 367 } 368 369 if (ifp->if_bpf) 370 bpf_ptap(ifp->if_bpf, m, &af, sizeof(af)); 371 372 M_PREPEND(m, sizeof(struct ip), MB_DONTWAIT); 373 if (m && m->m_len < sizeof(struct ip)) 374 m = m_pullup(m, sizeof(struct ip)); 375 if (m == NULL) 376 return ENOBUFS; 377 ip = mtod(m, struct ip *); 378 379 bzero(ip, sizeof(*ip)); 380 381 bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr), 382 &ip->ip_src, sizeof(ip->ip_src)); 383 bcopy(in4, &ip->ip_dst, sizeof(ip->ip_dst)); 384 ip->ip_p = IPPROTO_IPV6; 385 ip->ip_ttl = ip_stf_ttl; 386 ip->ip_len = m->m_pkthdr.len; /*host order*/ 387 if (ifp->if_flags & IFF_LINK1) 388 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos); 389 else 390 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos); 391 392 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst; 393 if (dst4->sin_family != AF_INET || 394 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) { 395 /* cache route doesn't match */ 396 dst4->sin_family = AF_INET; 397 dst4->sin_len = sizeof(struct sockaddr_in); 398 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr)); 399 if (sc->sc_ro.ro_rt) { 400 RTFREE(sc->sc_ro.ro_rt); 401 sc->sc_ro.ro_rt = NULL; 402 } 403 } 404 405 if (sc->sc_ro.ro_rt == NULL) { 406 rtalloc(&sc->sc_ro); 407 if (sc->sc_ro.ro_rt == NULL) { 408 m_freem(m); 409 return ENETUNREACH; 410 } 411 } 412 413 return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL); 414 } 415 416 static int 417 stf_checkaddr4(sc, in, inifp) 418 struct stf_softc *sc; 419 struct in_addr *in; 420 struct ifnet *inifp; /* incoming interface */ 421 { 422 struct in_ifaddr *ia4; 423 424 /* 425 * reject packets with the following address: 426 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8 427 */ 428 if (IN_MULTICAST(ntohl(in->s_addr))) 429 return -1; 430 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) { 431 case 0: case 127: case 255: 432 return -1; 433 } 434 435 /* 436 * reject packets with broadcast 437 */ 438 for (ia4 = TAILQ_FIRST(&in_ifaddrhead); 439 ia4; 440 ia4 = TAILQ_NEXT(ia4, ia_link)) 441 { 442 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) 443 continue; 444 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) 445 return -1; 446 } 447 448 /* 449 * perform ingress filter 450 */ 451 if (sc && (sc->sc_if.if_flags & IFF_LINK2) == 0 && inifp) { 452 struct sockaddr_in sin; 453 struct rtentry *rt; 454 455 bzero(&sin, sizeof(sin)); 456 sin.sin_family = AF_INET; 457 sin.sin_len = sizeof(struct sockaddr_in); 458 sin.sin_addr = *in; 459 rt = rtpurelookup((struct sockaddr *)&sin); 460 if (!rt || rt->rt_ifp != inifp) { 461 #if 0 462 log(LOG_WARNING, "%s: packet from 0x%x dropped " 463 "due to ingress filter\n", if_name(&sc->sc_if), 464 (u_int32_t)ntohl(sin.sin_addr.s_addr)); 465 #endif 466 if (rt) 467 rtfree(rt); 468 return -1; 469 } 470 rtfree(rt); 471 } 472 473 return 0; 474 } 475 476 static int 477 stf_checkaddr6(sc, in6, inifp) 478 struct stf_softc *sc; 479 struct in6_addr *in6; 480 struct ifnet *inifp; /* incoming interface */ 481 { 482 /* 483 * check 6to4 addresses 484 */ 485 if (IN6_IS_ADDR_6TO4(in6)) 486 return stf_checkaddr4(sc, GET_V4(in6), inifp); 487 488 /* 489 * reject anything that look suspicious. the test is implemented 490 * in ip6_input too, but we check here as well to 491 * (1) reject bad packets earlier, and 492 * (2) to be safe against future ip6_input change. 493 */ 494 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6)) 495 return -1; 496 497 return 0; 498 } 499 500 void 501 in_stf_input(struct mbuf *m, ...) 502 { 503 struct stf_softc *sc; 504 struct ip *ip; 505 struct ip6_hdr *ip6; 506 u_int8_t otos, itos; 507 struct ifnet *ifp; 508 int off, proto; 509 static const uint32_t af = AF_INET6; 510 __va_list ap; 511 512 __va_start(ap, m); 513 off = __va_arg(ap, int); 514 proto = __va_arg(ap, int); 515 __va_end(ap); 516 517 if (proto != IPPROTO_IPV6) { 518 m_freem(m); 519 return; 520 } 521 522 ip = mtod(m, struct ip *); 523 524 sc = (struct stf_softc *)encap_getarg(m); 525 526 if (sc == NULL || (sc->sc_if.if_flags & IFF_UP) == 0) { 527 m_freem(m); 528 return; 529 } 530 531 ifp = &sc->sc_if; 532 533 /* 534 * perform sanity check against outer src/dst. 535 * for source, perform ingress filter as well. 536 */ 537 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 || 538 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) { 539 m_freem(m); 540 return; 541 } 542 543 otos = ip->ip_tos; 544 m_adj(m, off); 545 546 if (m->m_len < sizeof(*ip6)) { 547 m = m_pullup(m, sizeof(*ip6)); 548 if (!m) 549 return; 550 } 551 ip6 = mtod(m, struct ip6_hdr *); 552 553 /* 554 * perform sanity check against inner src/dst. 555 * for source, perform ingress filter as well. 556 */ 557 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 || 558 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) { 559 m_freem(m); 560 return; 561 } 562 563 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 564 if ((ifp->if_flags & IFF_LINK1) != 0) 565 ip_ecn_egress(ECN_ALLOWED, &otos, &itos); 566 else 567 ip_ecn_egress(ECN_NOCARE, &otos, &itos); 568 ip6->ip6_flow &= ~htonl(0xff << 20); 569 ip6->ip6_flow |= htonl((u_int32_t)itos << 20); 570 571 m->m_pkthdr.rcvif = ifp; 572 573 if (ifp->if_bpf) 574 bpf_ptap(ifp->if_bpf, m, &af, sizeof(af)); 575 576 /* 577 * Put the packet to the network layer input queue according to the 578 * specified address family. 579 * See net/if_gif.c for possible issues with packet processing 580 * reorder due to extra queueing. 581 */ 582 ifp->if_ipackets++; 583 ifp->if_ibytes += m->m_pkthdr.len; 584 netisr_dispatch(NETISR_IPV6, m); 585 } 586 587 /* ARGSUSED */ 588 static void 589 stf_rtrequest(cmd, rt, info) 590 int cmd; 591 struct rtentry *rt; 592 struct rt_addrinfo *info; 593 { 594 595 if (rt) 596 rt->rt_rmx.rmx_mtu = IPV6_MMTU; 597 } 598 599 static int 600 stf_ioctl(ifp, cmd, data, cr) 601 struct ifnet *ifp; 602 u_long cmd; 603 caddr_t data; 604 struct ucred *cr; 605 { 606 struct ifaddr *ifa; 607 struct ifreq *ifr; 608 struct sockaddr_in6 *sin6; 609 int error; 610 611 error = 0; 612 switch (cmd) { 613 case SIOCSIFADDR: 614 ifa = (struct ifaddr *)data; 615 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) { 616 error = EAFNOSUPPORT; 617 break; 618 } 619 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; 620 if (IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) { 621 ifa->ifa_rtrequest = stf_rtrequest; 622 ifp->if_flags |= IFF_UP; 623 } else 624 error = EINVAL; 625 break; 626 627 case SIOCADDMULTI: 628 case SIOCDELMULTI: 629 ifr = (struct ifreq *)data; 630 if (ifr && ifr->ifr_addr.sa_family == AF_INET6) 631 ; 632 else 633 error = EAFNOSUPPORT; 634 break; 635 636 default: 637 error = EINVAL; 638 break; 639 } 640 641 return error; 642 } 643