if_stf.c revision 178888
1/* $FreeBSD: head/sys/net/if_stf.c 178888 2008-05-09 23:03:00Z julian $ */ 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/mbuf.h> 86#include <sys/errno.h> 87#include <sys/kernel.h> 88#include <sys/module.h> 89#include <sys/protosw.h> 90#include <sys/proc.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/bpf.h> 119 120#include <security/mac/mac_framework.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 u_int sc_fibnum; 141 const struct encaptab *encap_cookie; 142}; 143#define STF2IFP(sc) ((sc)->sc_ifp) 144 145/* 146 * XXXRW: Note that mutable fields in the softc are not currently locked: 147 * in particular, sc_ro needs to be protected from concurrent entrance 148 * of stf_output(). 149 */ 150static MALLOC_DEFINE(M_STF, STFNAME, "6to4 Tunnel Interface"); 151static const int ip_stf_ttl = 40; 152 153extern struct domain inetdomain; 154struct protosw in_stf_protosw = { 155 .pr_type = SOCK_RAW, 156 .pr_domain = &inetdomain, 157 .pr_protocol = IPPROTO_IPV6, 158 .pr_flags = PR_ATOMIC|PR_ADDR, 159 .pr_input = in_stf_input, 160 .pr_output = (pr_output_t *)rip_output, 161 .pr_ctloutput = rip_ctloutput, 162 .pr_usrreqs = &rip_usrreqs 163}; 164 165static char *stfnames[] = {"stf0", "stf", "6to4", NULL}; 166 167static int stfmodevent(module_t, int, void *); 168static int stf_encapcheck(const struct mbuf *, int, int, void *); 169static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *); 170static int stf_output(struct ifnet *, struct mbuf *, struct sockaddr *, 171 struct rtentry *); 172static int isrfc1918addr(struct in_addr *); 173static int stf_checkaddr4(struct stf_softc *, struct in_addr *, 174 struct ifnet *); 175static int stf_checkaddr6(struct stf_softc *, struct in6_addr *, 176 struct ifnet *); 177static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 178static int stf_ioctl(struct ifnet *, u_long, caddr_t); 179 180static int stf_clone_match(struct if_clone *, const char *); 181static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t); 182static int stf_clone_destroy(struct if_clone *, struct ifnet *); 183struct if_clone stf_cloner = IFC_CLONE_INITIALIZER(STFNAME, NULL, 0, 184 NULL, stf_clone_match, stf_clone_create, stf_clone_destroy); 185 186static int 187stf_clone_match(struct if_clone *ifc, const char *name) 188{ 189 int i; 190 191 for(i = 0; stfnames[i] != NULL; i++) { 192 if (strcmp(stfnames[i], name) == 0) 193 return (1); 194 } 195 196 return (0); 197} 198 199static int 200stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params) 201{ 202 int err, unit; 203 struct stf_softc *sc; 204 struct ifnet *ifp; 205 206 /* 207 * We can only have one unit, but since unit allocation is 208 * already locked, we use it to keep from allocating extra 209 * interfaces. 210 */ 211 unit = STFUNIT; 212 err = ifc_alloc_unit(ifc, &unit); 213 if (err != 0) 214 return (err); 215 216 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO); 217 ifp = STF2IFP(sc) = if_alloc(IFT_STF); 218 if (ifp == NULL) { 219 free(sc, M_STF); 220 ifc_free_unit(ifc, unit); 221 return (ENOSPC); 222 } 223 ifp->if_softc = sc; 224 sc->sc_fibnum = curthread->td_proc->p_fibnum; 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_int32_t)); 249 return (0); 250} 251 252static int 253stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp) 254{ 255 struct stf_softc *sc = ifp->if_softc; 256 int err; 257 258 err = encap_detach(sc->encap_cookie); 259 KASSERT(err == 0, ("Unexpected error detaching encap_cookie")); 260 bpfdetach(ifp); 261 if_detach(ifp); 262 if_free(ifp); 263 264 free(sc, M_STF); 265 ifc_free_unit(ifc, STFUNIT); 266 267 return (0); 268} 269 270static int 271stfmodevent(mod, type, data) 272 module_t mod; 273 int type; 274 void *data; 275{ 276 277 switch (type) { 278 case MOD_LOAD: 279 if_clone_attach(&stf_cloner); 280 break; 281 case MOD_UNLOAD: 282 if_clone_detach(&stf_cloner); 283 break; 284 default: 285 return (EOPNOTSUPP); 286 } 287 288 return (0); 289} 290 291static moduledata_t stf_mod = { 292 "if_stf", 293 stfmodevent, 294 0 295}; 296 297DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 298 299static int 300stf_encapcheck(m, off, proto, arg) 301 const struct mbuf *m; 302 int off; 303 int proto; 304 void *arg; 305{ 306 struct ip ip; 307 struct in6_ifaddr *ia6; 308 struct stf_softc *sc; 309 struct in_addr a, b, mask; 310 311 sc = (struct stf_softc *)arg; 312 if (sc == NULL) 313 return 0; 314 315 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0) 316 return 0; 317 318 /* IFF_LINK0 means "no decapsulation" */ 319 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0) 320 return 0; 321 322 if (proto != IPPROTO_IPV6) 323 return 0; 324 325 /* LINTED const cast */ 326 m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip); 327 328 if (ip.ip_v != 4) 329 return 0; 330 331 ia6 = stf_getsrcifa6(STF2IFP(sc)); 332 if (ia6 == NULL) 333 return 0; 334 335 /* 336 * check if IPv4 dst matches the IPv4 address derived from the 337 * local 6to4 address. 338 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:... 339 */ 340 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst, 341 sizeof(ip.ip_dst)) != 0) 342 return 0; 343 344 /* 345 * check if IPv4 src matches the IPv4 address derived from the 346 * local 6to4 address masked by prefixmask. 347 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24 348 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24 349 */ 350 bzero(&a, sizeof(a)); 351 bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a)); 352 bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask)); 353 a.s_addr &= mask.s_addr; 354 b = ip.ip_src; 355 b.s_addr &= mask.s_addr; 356 if (a.s_addr != b.s_addr) 357 return 0; 358 359 /* stf interface makes single side match only */ 360 return 32; 361} 362 363static struct in6_ifaddr * 364stf_getsrcifa6(ifp) 365 struct ifnet *ifp; 366{ 367 struct ifaddr *ia; 368 struct in_ifaddr *ia4; 369 struct sockaddr_in6 *sin6; 370 struct in_addr in; 371 372 TAILQ_FOREACH(ia, &ifp->if_addrlist, ifa_list) { 373 if (ia->ifa_addr->sa_family != AF_INET6) 374 continue; 375 sin6 = (struct sockaddr_in6 *)ia->ifa_addr; 376 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) 377 continue; 378 379 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in)); 380 LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash) 381 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr) 382 break; 383 if (ia4 == NULL) 384 continue; 385 386 return (struct in6_ifaddr *)ia; 387 } 388 389 return NULL; 390} 391 392static int 393stf_output(ifp, m, dst, rt) 394 struct ifnet *ifp; 395 struct mbuf *m; 396 struct sockaddr *dst; 397 struct rtentry *rt; 398{ 399 struct stf_softc *sc; 400 struct sockaddr_in6 *dst6; 401 struct in_addr in4; 402 caddr_t ptr; 403 struct sockaddr_in *dst4; 404 u_int8_t tos; 405 struct ip *ip; 406 struct ip6_hdr *ip6; 407 struct in6_ifaddr *ia6; 408 u_int32_t af; 409#ifdef MAC 410 int error; 411 412 error = mac_ifnet_check_transmit(ifp, m); 413 if (error) { 414 m_freem(m); 415 return (error); 416 } 417#endif 418 419 sc = ifp->if_softc; 420 dst6 = (struct sockaddr_in6 *)dst; 421 422 /* just in case */ 423 if ((ifp->if_flags & IFF_UP) == 0) { 424 m_freem(m); 425 ifp->if_oerrors++; 426 return ENETDOWN; 427 } 428 429 /* 430 * If we don't have an ip4 address that match my inner ip6 address, 431 * we shouldn't generate output. Without this check, we'll end up 432 * using wrong IPv4 source. 433 */ 434 ia6 = stf_getsrcifa6(ifp); 435 if (ia6 == NULL) { 436 m_freem(m); 437 ifp->if_oerrors++; 438 return ENETDOWN; 439 } 440 441 if (m->m_len < sizeof(*ip6)) { 442 m = m_pullup(m, sizeof(*ip6)); 443 if (!m) { 444 ifp->if_oerrors++; 445 return ENOBUFS; 446 } 447 } 448 ip6 = mtod(m, struct ip6_hdr *); 449 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 450 451 /* 452 * BPF writes need to be handled specially. 453 * This is a null operation, nothing here checks dst->sa_family. 454 */ 455 if (dst->sa_family == AF_UNSPEC) { 456 bcopy(dst->sa_data, &af, sizeof(af)); 457 dst->sa_family = af; 458 } 459 460 /* 461 * Pickup the right outer dst addr from the list of candidates. 462 * ip6_dst has priority as it may be able to give us shorter IPv4 hops. 463 */ 464 ptr = NULL; 465 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst)) 466 ptr = GET_V4(&ip6->ip6_dst); 467 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr)) 468 ptr = GET_V4(&dst6->sin6_addr); 469 else { 470 m_freem(m); 471 ifp->if_oerrors++; 472 return ENETUNREACH; 473 } 474 bcopy(ptr, &in4, sizeof(in4)); 475 476 if (bpf_peers_present(ifp->if_bpf)) { 477 /* 478 * We need to prepend the address family as 479 * a four byte field. Cons up a dummy header 480 * to pacify bpf. This is safe because bpf 481 * will only read from the mbuf (i.e., it won't 482 * try to free it or keep a pointer a to it). 483 */ 484 af = AF_INET6; 485 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 486 } 487 488 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 489 if (m && m->m_len < sizeof(struct ip)) 490 m = m_pullup(m, sizeof(struct ip)); 491 if (m == NULL) { 492 ifp->if_oerrors++; 493 return ENOBUFS; 494 } 495 ip = mtod(m, struct ip *); 496 497 bzero(ip, sizeof(*ip)); 498 499 bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr), 500 &ip->ip_src, sizeof(ip->ip_src)); 501 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst)); 502 ip->ip_p = IPPROTO_IPV6; 503 ip->ip_ttl = ip_stf_ttl; 504 ip->ip_len = m->m_pkthdr.len; /*host order*/ 505 if (ifp->if_flags & IFF_LINK1) 506 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos); 507 else 508 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos); 509 510 /* 511 * XXXRW: Locking of sc_ro required. 512 */ 513 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst; 514 if (dst4->sin_family != AF_INET || 515 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) { 516 /* cache route doesn't match */ 517 dst4->sin_family = AF_INET; 518 dst4->sin_len = sizeof(struct sockaddr_in); 519 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr)); 520 if (sc->sc_ro.ro_rt) { 521 RTFREE(sc->sc_ro.ro_rt); 522 sc->sc_ro.ro_rt = NULL; 523 } 524 } 525 526 if (sc->sc_ro.ro_rt == NULL) { 527 rtalloc_fib(&sc->sc_ro, sc->sc_fibnum); 528 if (sc->sc_ro.ro_rt == NULL) { 529 m_freem(m); 530 ifp->if_oerrors++; 531 return ENETUNREACH; 532 } 533 } 534 535 M_SETFIB(m, sc->sc_fibnum); 536 ifp->if_opackets++; 537 return ip_output(m, NULL, &sc->sc_ro, 0, NULL, NULL); 538} 539 540static int 541isrfc1918addr(in) 542 struct in_addr *in; 543{ 544 /* 545 * returns 1 if private address range: 546 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16 547 */ 548 if ((ntohl(in->s_addr) & 0xff000000) >> 24 == 10 || 549 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 || 550 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168) 551 return 1; 552 553 return 0; 554} 555 556static int 557stf_checkaddr4(sc, in, inifp) 558 struct stf_softc *sc; 559 struct in_addr *in; 560 struct ifnet *inifp; /* incoming interface */ 561{ 562 struct in_ifaddr *ia4; 563 564 /* 565 * reject packets with the following address: 566 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8 567 */ 568 if (IN_MULTICAST(ntohl(in->s_addr))) 569 return -1; 570 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) { 571 case 0: case 127: case 255: 572 return -1; 573 } 574 575 /* 576 * reject packets with private address range. 577 * (requirement from RFC3056 section 2 1st paragraph) 578 */ 579 if (isrfc1918addr(in)) 580 return -1; 581 582 /* 583 * reject packets with broadcast 584 */ 585 for (ia4 = TAILQ_FIRST(&in_ifaddrhead); 586 ia4; 587 ia4 = TAILQ_NEXT(ia4, ia_link)) 588 { 589 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) 590 continue; 591 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) 592 return -1; 593 } 594 595 /* 596 * perform ingress filter 597 */ 598 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) { 599 struct sockaddr_in sin; 600 struct rtentry *rt; 601 602 bzero(&sin, sizeof(sin)); 603 sin.sin_family = AF_INET; 604 sin.sin_len = sizeof(struct sockaddr_in); 605 sin.sin_addr = *in; 606 rt = rtalloc1_fib((struct sockaddr *)&sin, 0, 607 0UL, sc->sc_fibnum); 608 if (!rt || rt->rt_ifp != inifp) { 609#if 0 610 log(LOG_WARNING, "%s: packet from 0x%x dropped " 611 "due to ingress filter\n", if_name(STF2IFP(sc)), 612 (u_int32_t)ntohl(sin.sin_addr.s_addr)); 613#endif 614 if (rt) 615 RTFREE_LOCKED(rt); 616 return -1; 617 } 618 RTFREE_LOCKED(rt); 619 } 620 621 return 0; 622} 623 624static int 625stf_checkaddr6(sc, in6, inifp) 626 struct stf_softc *sc; 627 struct in6_addr *in6; 628 struct ifnet *inifp; /* incoming interface */ 629{ 630 /* 631 * check 6to4 addresses 632 */ 633 if (IN6_IS_ADDR_6TO4(in6)) { 634 struct in_addr in4; 635 bcopy(GET_V4(in6), &in4, sizeof(in4)); 636 return stf_checkaddr4(sc, &in4, inifp); 637 } 638 639 /* 640 * reject anything that look suspicious. the test is implemented 641 * in ip6_input too, but we check here as well to 642 * (1) reject bad packets earlier, and 643 * (2) to be safe against future ip6_input change. 644 */ 645 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6)) 646 return -1; 647 648 return 0; 649} 650 651void 652in_stf_input(m, off) 653 struct mbuf *m; 654 int off; 655{ 656 int proto; 657 struct stf_softc *sc; 658 struct ip *ip; 659 struct ip6_hdr *ip6; 660 u_int8_t otos, itos; 661 struct ifnet *ifp; 662 663 proto = mtod(m, struct ip *)->ip_p; 664 665 if (proto != IPPROTO_IPV6) { 666 m_freem(m); 667 return; 668 } 669 670 ip = mtod(m, struct ip *); 671 672 sc = (struct stf_softc *)encap_getarg(m); 673 674 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) { 675 m_freem(m); 676 return; 677 } 678 679 ifp = STF2IFP(sc); 680 681#ifdef MAC 682 mac_ifnet_create_mbuf(ifp, m); 683#endif 684 685 /* 686 * perform sanity check against outer src/dst. 687 * for source, perform ingress filter as well. 688 */ 689 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 || 690 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) { 691 m_freem(m); 692 return; 693 } 694 695 otos = ip->ip_tos; 696 m_adj(m, off); 697 698 if (m->m_len < sizeof(*ip6)) { 699 m = m_pullup(m, sizeof(*ip6)); 700 if (!m) 701 return; 702 } 703 ip6 = mtod(m, struct ip6_hdr *); 704 705 /* 706 * perform sanity check against inner src/dst. 707 * for source, perform ingress filter as well. 708 */ 709 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 || 710 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) { 711 m_freem(m); 712 return; 713 } 714 715 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 716 if ((ifp->if_flags & IFF_LINK1) != 0) 717 ip_ecn_egress(ECN_ALLOWED, &otos, &itos); 718 else 719 ip_ecn_egress(ECN_NOCARE, &otos, &itos); 720 ip6->ip6_flow &= ~htonl(0xff << 20); 721 ip6->ip6_flow |= htonl((u_int32_t)itos << 20); 722 723 m->m_pkthdr.rcvif = ifp; 724 725 if (bpf_peers_present(ifp->if_bpf)) { 726 /* 727 * We need to prepend the address family as 728 * a four byte field. Cons up a dummy header 729 * to pacify bpf. This is safe because bpf 730 * will only read from the mbuf (i.e., it won't 731 * try to free it or keep a pointer a to it). 732 */ 733 u_int32_t af = AF_INET6; 734 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 735 } 736 737 /* 738 * Put the packet to the network layer input queue according to the 739 * specified address family. 740 * See net/if_gif.c for possible issues with packet processing 741 * reorder due to extra queueing. 742 */ 743 ifp->if_ipackets++; 744 ifp->if_ibytes += m->m_pkthdr.len; 745 netisr_dispatch(NETISR_IPV6, m); 746} 747 748/* ARGSUSED */ 749static void 750stf_rtrequest(cmd, rt, info) 751 int cmd; 752 struct rtentry *rt; 753 struct rt_addrinfo *info; 754{ 755 RT_LOCK_ASSERT(rt); 756 rt->rt_rmx.rmx_mtu = IPV6_MMTU; 757} 758 759static int 760stf_ioctl(ifp, cmd, data) 761 struct ifnet *ifp; 762 u_long cmd; 763 caddr_t data; 764{ 765 struct ifaddr *ifa; 766 struct ifreq *ifr; 767 struct sockaddr_in6 *sin6; 768 struct in_addr addr; 769 int error; 770 771 error = 0; 772 switch (cmd) { 773 case SIOCSIFADDR: 774 ifa = (struct ifaddr *)data; 775 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) { 776 error = EAFNOSUPPORT; 777 break; 778 } 779 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; 780 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) { 781 error = EINVAL; 782 break; 783 } 784 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr)); 785 if (isrfc1918addr(&addr)) { 786 error = EINVAL; 787 break; 788 } 789 790 ifa->ifa_rtrequest = stf_rtrequest; 791 ifp->if_flags |= IFF_UP; 792 break; 793 794 case SIOCADDMULTI: 795 case SIOCDELMULTI: 796 ifr = (struct ifreq *)data; 797 if (ifr && ifr->ifr_addr.sa_family == AF_INET6) 798 ; 799 else 800 error = EAFNOSUPPORT; 801 break; 802 803 default: 804 error = EINVAL; 805 break; 806 } 807 808 return error; 809} 810