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 "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])) 132 133 struct 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 }; 142 #define STF2IFP(sc) ((sc)->sc_ifp) 143 144 /* 145 * XXXRW: Note that mutable fields in the softc are not currently locked: 146 * in particular, sc_ro needs to be protected from concurrent entrance 147 * of stf_output(). 148 */ 149 static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface"); 150 static const int ip_stf_ttl = 40; 151 152 extern struct domain inetdomain; 153 struct protosw in_stf_protosw = { 154 .pr_type = SOCK_RAW, 155 .pr_domain = &inetdomain, 156 .pr_protocol = IPPROTO_IPV6, 157 .pr_flags = PR_ATOMIC|PR_ADDR, 158 .pr_input = in_stf_input, 159 .pr_output = (pr_output_t *)rip_output, 160 .pr_ctloutput = rip_ctloutput, 161 .pr_usrreqs = &rip_usrreqs 162 }; 163 164 static char *stfnames[] = {"stf0", "stf", "6to4", NULL}; 165 166 static int stfmodevent(module_t, int, void *); 167 static int stf_encapcheck(const struct mbuf *, int, int, void *); 168 static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *); 169 static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *, 170 struct rtentry *); 171 static int isrfc1918addr(struct in_addr *); 172 static int stf_checkaddr4(struct stf_softc *, struct in_addr *, 173 struct ifnet *); 174 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *, 175 struct ifnet *); 176 static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 177 static int stf_ioctl(struct ifnet *, u_long, caddr_t); 178 179 static int stf_clone_match(struct if_clone *, const char *); 180 static int stf_clone_create(struct if_clone *, char *, size_t); 181 static int stf_clone_destroy(struct if_clone *, struct ifnet *); 182 struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0, 183 NULL, stf_clone_match, stf_clone_create, stf_clone_destroy); 184 185 static int 186 stf_clone_match(struct if_clone *ifc, const char *name) 187 { 188 int i; 189 190 for(i = 0; stfnames[i] != NULL; i++) { 191 if (strcmp(stfnames[i], name) == 0) 192 return (1); 193 } 194 195 return (0); 196 } 197 198 static int 199 stf_clone_create(struct if_clone *ifc, char *name, size_t len) 200 { 201 int err, unit; 202 struct stf_softc *sc; 203 struct ifnet *ifp; 204 205 /* 206 * We can only have one unit, but since unit allocation is 207 * already locked, we use it to keep from allocating extra 208 * interfaces. 209 */ 210 unit = STFUNIT; 211 err = ifc_alloc_unit(ifc, &unit); 212 if (err != 0) 213 return (err); 214 215 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO); 216 ifp = STF2IFP(sc) = if_alloc(IFT_STF); 217 if (ifp == NULL) { 218 free(sc, M_STF); 219 ifc_free_unit(ifc, unit); 220 return (ENOSPC); 221 } 222 ifp->if_softc = sc; 223 224 /* 225 * Set the name manually rather then using if_initname because 226 * we don't conform to the default naming convention for interfaces. 227 */ 228 strlcpy(ifp->if_xname, name, IFNAMSIZ); 229 ifp->if_dname = ifc->ifc_name; 230 ifp->if_dunit = IF_DUNIT_NONE; 231 232 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6, 233 stf_encapcheck, &in_stf_protosw, sc); 234 if (sc->encap_cookie == NULL) { 235 if_printf(ifp, "attach failed\n"); 236 free(sc, M_STF); 237 ifc_free_unit(ifc, unit); 238 return (ENOMEM); 239 } 240 241 ifp->if_mtu = IPV6_MMTU; 242 ifp->if_ioctl = stf_ioctl; 243 ifp->if_output = stf_output; 244 ifp->if_snd.ifq_maxlen = IFQ_MAXLEN; 245 if_attach(ifp); 246 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 247 return (0); 248 } 249 250 static int 251 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp) 252 { 253 struct stf_softc *sc = ifp->if_softc; 254 int err; 255 256 err = encap_detach(sc->encap_cookie); 257 KASSERT(err == 0, ("Unexpected error detaching encap_cookie")); 258 bpfdetach(ifp); 259 if_detach(ifp); 260 if_free(ifp); 261 262 free(sc, M_STF); 263 ifc_free_unit(ifc, STFUNIT); 264 265 return (0); 266 } 267 268 static int 269 stfmodevent(mod, type, data) 270 module_t mod; 271 int type; 272 void *data; 273 { 274 275 switch (type) { 276 case MOD_LOAD: 277 if_clone_attach(&stf_cloner); 278 break; 279 case MOD_UNLOAD: 280 if_clone_detach(&stf_cloner); 281 break; 282 default: 283 return (EOPNOTSUPP); 284 } 285 286 return (0); 287 } 288 289 static moduledata_t stf_mod = { 290 "if_stf", 291 stfmodevent, 292 0 293 }; 294 295 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 296 297 static int 298 stf_encapcheck(m, off, proto, arg) 299 const struct mbuf *m; 300 int off; 301 int proto; 302 void *arg; 303 { 304 struct ip ip; 305 struct in6_ifaddr *ia6; 306 struct stf_softc *sc; 307 struct in_addr a, b, mask; 308 309 sc = (struct stf_softc *)arg; 310 if (sc == NULL) 311 return 0; 312 313 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0) 314 return 0; 315 316 /* IFF_LINK0 means "no decapsulation" */ 317 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0) 318 return 0; 319 320 if (proto != IPPROTO_IPV6) 321 return 0; 322 323 /* LINTED const cast */ 324 m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip); 325 326 if (ip.ip_v != 4) 327 return 0; 328 329 ia6 = stf_getsrcifa6(STF2IFP(sc)); 330 if (ia6 == NULL) 331 return 0; 332 333 /* 334 * check if IPv4 dst matches the IPv4 address derived from the 335 * local 6to4 address. 336 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:... 337 */ 338 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst, 339 sizeof(ip.ip_dst)) != 0) 340 return 0; 341 342 /* 343 * check if IPv4 src matches the IPv4 address derived from the 344 * local 6to4 address masked by prefixmask. 345 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24 346 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24 347 */ 348 bzero(&a, sizeof(a)); 349 bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a)); 350 bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask)); 351 a.s_addr &= mask.s_addr; 352 b = ip.ip_src; 353 b.s_addr &= mask.s_addr; 354 if (a.s_addr != b.s_addr) 355 return 0; 356 357 /* stf interface makes single side match only */ 358 return 32; 359 } 360 361 static struct in6_ifaddr * 362 stf_getsrcifa6(ifp) 363 struct ifnet *ifp; 364 { 365 struct ifaddr *ia; 366 struct in_ifaddr *ia4; 367 struct sockaddr_in6 *sin6; 368 struct in_addr in; 369 370 for (ia = TAILQ_FIRST(&ifp->if_addrlist); 371 ia; 372 ia = TAILQ_NEXT(ia, ifa_list)) 373 { 374 if (ia->ifa_addr == NULL) 375 continue; 376 if (ia->ifa_addr->sa_family != AF_INET6) 377 continue; 378 sin6 = (struct sockaddr_in6 *)ia->ifa_addr; 379 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) 380 continue; 381 382 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in)); 383 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash) 384 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr) 385 break; 386 if (ia4 == NULL) 387 continue; 388 389 return (struct in6_ifaddr *)ia; 390 } 391 392 return NULL; 393 } 394 395 static int 396 stf_output(ifp, m, dst, rt) 397 struct ifnet *ifp; 398 struct mbuf *m; 399 struct sockaddr *dst; 400 struct rtentry *rt; 401 { 402 struct stf_softc *sc; 403 struct sockaddr_in6 *dst6; 404 struct in_addr in4; 405 caddr_t ptr; 406 struct sockaddr_in *dst4; 407 u_int8_t tos; 408 struct ip *ip; 409 struct ip6_hdr *ip6; 410 struct in6_ifaddr *ia6; 411 u_int32_t af; 412 #ifdef MAC 413 int error; 414 415 error = mac_check_ifnet_transmit(ifp, m); 416 if (error) { 417 m_freem(m); 418 return (error); 419 } 420 #endif 421 422 sc = ifp->if_softc; 423 dst6 = (struct sockaddr_in6 *)dst; 424 425 /* just in case */ 426 if ((ifp->if_flags & IFF_UP) == 0) { 427 m_freem(m); 428 ifp->if_oerrors++; 429 return ENETDOWN; 430 } 431 432 /* 433 * If we don't have an ip4 address that match my inner ip6 address, 434 * we shouldn't generate output. Without this check, we'll end up 435 * using wrong IPv4 source. 436 */ 437 ia6 = stf_getsrcifa6(ifp); 438 if (ia6 == NULL) { 439 m_freem(m); 440 ifp->if_oerrors++; 441 return ENETDOWN; 442 } 443 444 if (m->m_len < sizeof(*ip6)) { 445 m = m_pullup(m, sizeof(*ip6)); 446 if (!m) { 447 ifp->if_oerrors++; 448 return ENOBUFS; 449 } 450 } 451 ip6 = mtod(m, struct ip6_hdr *); 452 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 453 454 /* 455 * BPF writes need to be handled specially. 456 * This is a null operation, nothing here checks dst->sa_family. 457 */ 458 if (dst->sa_family == AF_UNSPEC) { 459 bcopy(dst->sa_data, &af, sizeof(af)); 460 dst->sa_family = af; 461 } 462 463 /* 464 * Pickup the right outer dst addr from the list of candidates. 465 * ip6_dst has priority as it may be able to give us shorter IPv4 hops. 466 */ 467 ptr = NULL; 468 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst)) 469 ptr = GET_V4(&ip6->ip6_dst); 470 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr)) 471 ptr = GET_V4(&dst6->sin6_addr); 472 else { 473 m_freem(m); 474 ifp->if_oerrors++; 475 return ENETUNREACH; 476 } 477 bcopy(ptr, &in4, sizeof(in4)); 478 479 if (ifp->if_bpf) { 480 /* 481 * We need to prepend the address family as 482 * a four byte field. Cons up a dummy header 483 * to pacify bpf. This is safe because bpf 484 * will only read from the mbuf (i.e., it won't 485 * try to free it or keep a pointer a to it). 486 */ 487 af = AF_INET6; 488 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 489 } 490 491 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 492 if (m && m->m_len < sizeof(struct ip)) 493 m = m_pullup(m, sizeof(struct ip)); 494 if (m == NULL) { 495 ifp->if_oerrors++; 496 return ENOBUFS; 497 } 498 ip = mtod(m, struct ip *); 499 500 bzero(ip, sizeof(*ip)); 501 502 bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr), 503 &ip->ip_src, sizeof(ip->ip_src)); 504 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst)); 505 ip->ip_p = IPPROTO_IPV6; 506 ip->ip_ttl = ip_stf_ttl; 507 ip->ip_len = m->m_pkthdr.len; /*host order*/ 508 if (ifp->if_flags & IFF_LINK1) 509 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos); 510 else 511 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos); 512 513 /* 514 * XXXRW: Locking of sc_ro required. 515 */ 516 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst; 517 if (dst4->sin_family != AF_INET || 518 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) { 519 /* cache route doesn't match */ 520 dst4->sin_family = AF_INET; 521 dst4->sin_len = sizeof(struct sockaddr_in); 522 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr)); 523 if (sc->sc_ro.ro_rt) { 524 RTFREE(sc->sc_ro.ro_rt); 525 sc->sc_ro.ro_rt = NULL; 526 } 527 } 528 529 if (sc->sc_ro.ro_rt == NULL) { 530 rtalloc(&sc->sc_ro); 531 if (sc->sc_ro.ro_rt == NULL) { 532 m_freem(m); 533 ifp->if_oerrors++; 534 return ENETUNREACH; 535 } 536 } 537 538 ifp->if_opackets++; 539 return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL); 540 } 541 542 static int 543 isrfc1918addr(in) 544 struct in_addr *in; 545 { 546 /* 547 * returns 1 if private address range: 548 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16 549 */ 550 if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 || 551 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 || 552 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168) 553 return 1; 554 555 return 0; 556 } 557 558 static int 559 stf_checkaddr4(sc, in, inifp) 560 struct stf_softc *sc; 561 struct in_addr *in; 562 struct ifnet *inifp; /* incoming interface */ 563 { 564 struct in_ifaddr *ia4; 565 566 /* 567 * reject packets with the following address: 568 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8 569 */ 570 if (IN_MULTICAST(ntohl(in->s_addr))) 571 return -1; 572 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) { 573 case 0: case 127: case 255: 574 return -1; 575 } 576 577 /* 578 * reject packets with private address range. 579 * (requirement from RFC3056 section 2 1st paragraph) 580 */ 581 if (isrfc1918addr(in)) 582 return -1; 583 584 /* 585 * reject packets with broadcast 586 */ 587 for (ia4 = TAILQ_FIRST(&in_ifaddrhead); 588 ia4; 589 ia4 = TAILQ_NEXT(ia4, ia_link)) 590 { 591 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) 592 continue; 593 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) 594 return -1; 595 } 596 597 /* 598 * perform ingress filter 599 */ 600 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) { 601 struct sockaddr_in sin; 602 struct rtentry *rt; 603 604 bzero(&sin, sizeof(sin)); 605 sin.sin_family = AF_INET; 606 sin.sin_len = sizeof(struct sockaddr_in); 607 sin.sin_addr = *in; 608 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL); 609 if (!rt || rt->rt_ifp != inifp) { 610 #if 0 611 log(LOG_WARNING, "%s: packet from 0x%x dropped " 612 "due to ingress filter\n", if_name(STF2IFP(sc)), 613 (u_int32_t)ntohl(sin.sin_addr.s_addr)); 614 #endif 615 if (rt) 616 rtfree(rt); 617 return -1; 618 } 619 rtfree(rt); 620 } 621 622 return 0; 623 } 624 625 static int 626 stf_checkaddr6(sc, in6, inifp) 627 struct stf_softc *sc; 628 struct in6_addr *in6; 629 struct ifnet *inifp; /* incoming interface */ 630 { 631 /* 632 * check 6to4 addresses 633 */ 634 if (IN6_IS_ADDR_6TO4(in6)) { 635 struct in_addr in4; 636 bcopy(GET_V4(in6), &in4, sizeof(in4)); 637 return stf_checkaddr4(sc, &in4, inifp); 638 } 639 640 /* 641 * reject anything that look suspicious. the test is implemented 642 * in ip6_input too, but we check here as well to 643 * (1) reject bad packets earlier, and 644 * (2) to be safe against future ip6_input change. 645 */ 646 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6)) 647 return -1; 648 649 return 0; 650 } 651 652 void 653 in_stf_input(m, off) 654 struct mbuf *m; 655 int off; 656 { 657 int proto; 658 struct stf_softc *sc; 659 struct ip *ip; 660 struct ip6_hdr *ip6; 661 u_int8_t otos, itos; 662 struct ifnet *ifp; 663 664 proto = mtod(m, struct ip *)->ip_p; 665 666 if (proto != IPPROTO_IPV6) { 667 m_freem(m); 668 return; 669 } 670 671 ip = mtod(m, struct ip *); 672 673 sc = (struct stf_softc *)encap_getarg(m); 674 675 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) { 676 m_freem(m); 677 return; 678 } 679 680 ifp = STF2IFP(sc); 681 682 #ifdef MAC 683 mac_create_mbuf_from_ifnet(ifp, m); 684 #endif 685 686 /* 687 * perform sanity check against outer src/dst. 688 * for source, perform ingress filter as well. 689 */ 690 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 || 691 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) { 692 m_freem(m); 693 return; 694 } 695 696 otos = ip->ip_tos; 697 m_adj(m, off); 698 699 if (m->m_len < sizeof(*ip6)) { 700 m = m_pullup(m, sizeof(*ip6)); 701 if (!m) 702 return; 703 } 704 ip6 = mtod(m, struct ip6_hdr *); 705 706 /* 707 * perform sanity check against inner src/dst. 708 * for source, perform ingress filter as well. 709 */ 710 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 || 711 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) { 712 m_freem(m); 713 return; 714 } 715 716 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 717 if ((ifp->if_flags & IFF_LINK1) != 0) 718 ip_ecn_egress(ECN_ALLOWED, &otos, &itos); 719 else 720 ip_ecn_egress(ECN_NOCARE, &otos, &itos); 721 ip6->ip6_flow &= ~htonl(0xff << 20); 722 ip6->ip6_flow |= htonl((u_int32_t)itos << 20); 723 724 m->m_pkthdr.rcvif = ifp; 725 726 if (ifp->if_bpf) { 727 /* 728 * We need to prepend the address family as 729 * a four byte field. Cons up a dummy header 730 * to pacify bpf. This is safe because bpf 731 * will only read from the mbuf (i.e., it won't 732 * try to free it or keep a pointer a to it). 733 */ 734 u_int32_t af = AF_INET6; 735 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 736 } 737 738 /* 739 * Put the packet to the network layer input queue according to the 740 * specified address family. 741 * See net/if_gif.c for possible issues with packet processing 742 * reorder due to extra queueing. 743 */ 744 ifp->if_ipackets++; 745 ifp->if_ibytes += m->m_pkthdr.len; 746 netisr_dispatch(NETISR_IPV6, m); 747 } 748 749 /* ARGSUSED */ 750 static void 751 stf_rtrequest(cmd, rt, info) 752 int cmd; 753 struct rtentry *rt; 754 struct rt_addrinfo *info; 755 { 756 RT_LOCK_ASSERT(rt); 757 rt->rt_rmx.rmx_mtu = IPV6_MMTU; 758 } 759 760 static int 761 stf_ioctl(ifp, cmd, data) 762 struct ifnet *ifp; 763 u_long cmd; 764 caddr_t data; 765 { 766 struct ifaddr *ifa; 767 struct ifreq *ifr; 768 struct sockaddr_in6 *sin6; 769 struct in_addr addr; 770 int error; 771 772 error = 0; 773 switch (cmd) { 774 case SIOCSIFADDR: 775 ifa = (struct ifaddr *)data; 776 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) { 777 error = EAFNOSUPPORT; 778 break; 779 } 780 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; 781 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) { 782 error = EINVAL; 783 break; 784 } 785 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr)); 786 if (isrfc1918addr(&addr)) { 787 error = EINVAL; 788 break; 789 } 790 791 ifa->ifa_rtrequest = stf_rtrequest; 792 ifp->if_flags |= IFF_UP; 793 break; 794 795 case SIOCADDMULTI: 796 case SIOCDELMULTI: 797 ifr = (struct ifreq *)data; 798 if (ifr && ifr->ifr_addr.sa_family == AF_INET6) 799 ; 800 else 801 error = EAFNOSUPPORT; 802 break; 803 804 default: 805 error = EINVAL; 806 break; 807 } 808 809 return error; 810 } 811