if_stf.c revision 340958
1/* $FreeBSD: stable/10/sys/net/if_stf.c 340958 2018-11-26 12:19:30Z eugen $ */ 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 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#include <net/vnet.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> 109 110#include <netinet/ip6.h> 111#include <netinet6/ip6_var.h> 112#include <netinet6/in6_var.h> 113#include <netinet/ip_ecn.h> 114 115#include <netinet/ip_encap.h> 116 117#include <machine/stdarg.h> 118 119#include <net/bpf.h> 120 121#include <security/mac/mac_framework.h> 122 123SYSCTL_DECL(_net_link); 124static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW, 0, "6to4 Interface"); 125 126static int stf_route_cache = 1; 127SYSCTL_INT(_net_link_stf, OID_AUTO, route_cache, CTLFLAG_RW, 128 &stf_route_cache, 0, "Caching of IPv4 routes for 6to4 Output"); 129 130static int stf_permit_rfc1918 = 0; 131TUNABLE_INT("net.link.stf.permit_rfc1918", &stf_permit_rfc1918); 132SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RW | CTLFLAG_TUN, 133 &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses"); 134 135#define STFUNIT 0 136 137#define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002) 138 139/* 140 * XXX: Return a pointer with 16-bit aligned. Don't cast it to 141 * struct in_addr *; use bcopy() instead. 142 */ 143#define GET_V4(x) (&(x)->s6_addr16[1]) 144 145struct stf_softc { 146 struct ifnet *sc_ifp; 147 union { 148 struct route __sc_ro4; 149 struct route_in6 __sc_ro6; /* just for safety */ 150 } __sc_ro46; 151#define sc_ro __sc_ro46.__sc_ro4 152 struct mtx sc_ro_mtx; 153 u_int sc_fibnum; 154 const struct encaptab *encap_cookie; 155}; 156#define STF2IFP(sc) ((sc)->sc_ifp) 157 158static const char stfname[] = "stf"; 159 160/* 161 * Note that mutable fields in the softc are not currently locked. 162 * We do lock sc_ro in stf_output though. 163 */ 164static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface"); 165static const int ip_stf_ttl = 40; 166 167extern struct domain inetdomain; 168struct protosw in_stf_protosw = { 169 .pr_type = SOCK_RAW, 170 .pr_domain = &inetdomain, 171 .pr_protocol = IPPROTO_IPV6, 172 .pr_flags = PR_ATOMIC|PR_ADDR, 173 .pr_input = in_stf_input, 174 .pr_output = (pr_output_t *)rip_output, 175 .pr_ctloutput = rip_ctloutput, 176 .pr_usrreqs = &rip_usrreqs 177}; 178 179static char *stfnames[] = {"stf0", "stf", "6to4", NULL}; 180 181static int stfmodevent(module_t, int, void *); 182static int stf_encapcheck(const struct mbuf *, int, int, void *); 183static struct in6_ifaddr *stf_getsrcifa6(struct ifnet *); 184static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *, 185 struct route *); 186static int isrfc1918addr(struct in_addr *); 187static int stf_checkaddr4(struct stf_softc *, struct in_addr *, 188 struct ifnet *); 189static int stf_checkaddr6(struct stf_softc *, struct in6_addr *, 190 struct ifnet *); 191static void stf_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 192static int stf_ioctl(struct ifnet *, u_long, caddr_t); 193 194static int stf_clone_match(struct if_clone *, const char *); 195static int stf_clone_create(struct if_clone *, char *, size_t, caddr_t); 196static int stf_clone_destroy(struct if_clone *, struct ifnet *); 197static struct if_clone *stf_cloner; 198 199static int 200stf_clone_match(struct if_clone *ifc, const char *name) 201{ 202 int i; 203 204 for(i = 0; stfnames[i] != NULL; i++) { 205 if (strcmp(stfnames[i], name) == 0) 206 return (1); 207 } 208 209 return (0); 210} 211 212static int 213stf_clone_create(struct if_clone *ifc, char *name, size_t len, caddr_t params) 214{ 215 int err, unit; 216 struct stf_softc *sc; 217 struct ifnet *ifp; 218 219 /* 220 * We can only have one unit, but since unit allocation is 221 * already locked, we use it to keep from allocating extra 222 * interfaces. 223 */ 224 unit = STFUNIT; 225 err = ifc_alloc_unit(ifc, &unit); 226 if (err != 0) 227 return (err); 228 229 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO); 230 ifp = STF2IFP(sc) = if_alloc(IFT_STF); 231 if (ifp == NULL) { 232 free(sc, M_STF); 233 ifc_free_unit(ifc, unit); 234 return (ENOSPC); 235 } 236 ifp->if_softc = sc; 237 sc->sc_fibnum = curthread->td_proc->p_fibnum; 238 239 /* 240 * Set the name manually rather then using if_initname because 241 * we don't conform to the default naming convention for interfaces. 242 */ 243 strlcpy(ifp->if_xname, name, IFNAMSIZ); 244 ifp->if_dname = stfname; 245 ifp->if_dunit = IF_DUNIT_NONE; 246 247 mtx_init(&(sc)->sc_ro_mtx, "stf ro", NULL, MTX_DEF); 248 sc->encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV6, 249 stf_encapcheck, &in_stf_protosw, sc); 250 if (sc->encap_cookie == NULL) { 251 if_printf(ifp, "attach failed\n"); 252 free(sc, M_STF); 253 ifc_free_unit(ifc, unit); 254 return (ENOMEM); 255 } 256 257 ifp->if_mtu = IPV6_MMTU; 258 ifp->if_ioctl = stf_ioctl; 259 ifp->if_output = stf_output; 260 ifp->if_snd.ifq_maxlen = ifqmaxlen; 261 if_attach(ifp); 262 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 263 return (0); 264} 265 266static int 267stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp) 268{ 269 struct stf_softc *sc = ifp->if_softc; 270 int err; 271 272 err = encap_detach(sc->encap_cookie); 273 KASSERT(err == 0, ("Unexpected error detaching encap_cookie")); 274 mtx_destroy(&(sc)->sc_ro_mtx); 275 bpfdetach(ifp); 276 if_detach(ifp); 277 if_free(ifp); 278 279 free(sc, M_STF); 280 ifc_free_unit(ifc, STFUNIT); 281 282 return (0); 283} 284 285static int 286stfmodevent(mod, type, data) 287 module_t mod; 288 int type; 289 void *data; 290{ 291 292 switch (type) { 293 case MOD_LOAD: 294 stf_cloner = if_clone_advanced(stfname, 0, stf_clone_match, 295 stf_clone_create, stf_clone_destroy); 296 break; 297 case MOD_UNLOAD: 298 if_clone_detach(stf_cloner); 299 break; 300 default: 301 return (EOPNOTSUPP); 302 } 303 304 return (0); 305} 306 307static moduledata_t stf_mod = { 308 "if_stf", 309 stfmodevent, 310 0 311}; 312 313DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 314 315static int 316stf_encapcheck(m, off, proto, arg) 317 const struct mbuf *m; 318 int off; 319 int proto; 320 void *arg; 321{ 322 struct ip ip; 323 struct in6_ifaddr *ia6; 324 struct stf_softc *sc; 325 struct in_addr a, b, mask; 326 327 sc = (struct stf_softc *)arg; 328 if (sc == NULL) 329 return 0; 330 331 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0) 332 return 0; 333 334 /* IFF_LINK0 means "no decapsulation" */ 335 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0) 336 return 0; 337 338 if (proto != IPPROTO_IPV6) 339 return 0; 340 341 /* LINTED const cast */ 342 m_copydata((struct mbuf *)(uintptr_t)m, 0, sizeof(ip), (caddr_t)&ip); 343 344 if (ip.ip_v != 4) 345 return 0; 346 347 ia6 = stf_getsrcifa6(STF2IFP(sc)); 348 if (ia6 == NULL) 349 return 0; 350 351 /* 352 * check if IPv4 dst matches the IPv4 address derived from the 353 * local 6to4 address. 354 * success on: dst = 10.1.1.1, ia6->ia_addr = 2002:0a01:0101:... 355 */ 356 if (bcmp(GET_V4(&ia6->ia_addr.sin6_addr), &ip.ip_dst, 357 sizeof(ip.ip_dst)) != 0) { 358 ifa_free(&ia6->ia_ifa); 359 return 0; 360 } 361 362 /* 363 * check if IPv4 src matches the IPv4 address derived from the 364 * local 6to4 address masked by prefixmask. 365 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24 366 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24 367 */ 368 bzero(&a, sizeof(a)); 369 bcopy(GET_V4(&ia6->ia_addr.sin6_addr), &a, sizeof(a)); 370 bcopy(GET_V4(&ia6->ia_prefixmask.sin6_addr), &mask, sizeof(mask)); 371 ifa_free(&ia6->ia_ifa); 372 a.s_addr &= mask.s_addr; 373 b = ip.ip_src; 374 b.s_addr &= mask.s_addr; 375 if (a.s_addr != b.s_addr) 376 return 0; 377 378 /* stf interface makes single side match only */ 379 return 32; 380} 381 382static struct in6_ifaddr * 383stf_getsrcifa6(ifp) 384 struct ifnet *ifp; 385{ 386 struct ifaddr *ia; 387 struct sockaddr_in6 *sin6; 388 struct in_addr in; 389 390 if_addr_rlock(ifp); 391 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 392 if (ia->ifa_addr->sa_family != AF_INET6) 393 continue; 394 sin6 = (struct sockaddr_in6 *)ia->ifa_addr; 395 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) 396 continue; 397 398 bcopy(GET_V4(&sin6->sin6_addr), &in, sizeof(in)); 399 if (!in_localip(in)) 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(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, 413 struct route *ro) 414{ 415 struct stf_softc *sc; 416 const struct sockaddr_in6 *dst6; 417 struct route *cached_route; 418 struct in_addr in4; 419 const void *ptr; 420 struct sockaddr_in *dst4; 421 u_int8_t tos; 422 struct ip *ip; 423 struct ip6_hdr *ip6; 424 struct in6_ifaddr *ia6; 425 int error; 426 427#ifdef MAC 428 error = mac_ifnet_check_transmit(ifp, m); 429 if (error) { 430 m_freem(m); 431 return (error); 432 } 433#endif 434 435 sc = ifp->if_softc; 436 dst6 = (const struct sockaddr_in6 *)dst; 437 438 /* just in case */ 439 if ((ifp->if_flags & IFF_UP) == 0) { 440 m_freem(m); 441 ifp->if_oerrors++; 442 return ENETDOWN; 443 } 444 445 /* 446 * If we don't have an ip4 address that match my inner ip6 address, 447 * we shouldn't generate output. Without this check, we'll end up 448 * using wrong IPv4 source. 449 */ 450 ia6 = stf_getsrcifa6(ifp); 451 if (ia6 == NULL) { 452 m_freem(m); 453 ifp->if_oerrors++; 454 return ENETDOWN; 455 } 456 457 if (m->m_len < sizeof(*ip6)) { 458 m = m_pullup(m, sizeof(*ip6)); 459 if (!m) { 460 ifa_free(&ia6->ia_ifa); 461 ifp->if_oerrors++; 462 return ENOBUFS; 463 } 464 } 465 ip6 = mtod(m, struct ip6_hdr *); 466 tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 467 468 /* 469 * Pickup the right outer dst addr from the list of candidates. 470 * ip6_dst has priority as it may be able to give us shorter IPv4 hops. 471 */ 472 ptr = NULL; 473 if (IN6_IS_ADDR_6TO4(&ip6->ip6_dst)) 474 ptr = GET_V4(&ip6->ip6_dst); 475 else if (IN6_IS_ADDR_6TO4(&dst6->sin6_addr)) 476 ptr = GET_V4(&dst6->sin6_addr); 477 else { 478 ifa_free(&ia6->ia_ifa); 479 m_freem(m); 480 ifp->if_oerrors++; 481 return ENETUNREACH; 482 } 483 bcopy(ptr, &in4, sizeof(in4)); 484 485 if (bpf_peers_present(ifp->if_bpf)) { 486 /* 487 * We need to prepend the address family as 488 * a four byte field. Cons up a dummy header 489 * to pacify bpf. This is safe because bpf 490 * will only read from the mbuf (i.e., it won't 491 * try to free it or keep a pointer a to it). 492 */ 493 u_int af = AF_INET6; 494 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 495 } 496 497 M_PREPEND(m, sizeof(struct ip), M_NOWAIT); 498 if (m == NULL) { 499 ifa_free(&ia6->ia_ifa); 500 ifp->if_oerrors++; 501 return ENOBUFS; 502 } 503 ip = mtod(m, struct ip *); 504 505 bzero(ip, sizeof(*ip)); 506 507 bcopy(GET_V4(&((struct sockaddr_in6 *)&ia6->ia_addr)->sin6_addr), 508 &ip->ip_src, sizeof(ip->ip_src)); 509 ifa_free(&ia6->ia_ifa); 510 bcopy(&in4, &ip->ip_dst, sizeof(ip->ip_dst)); 511 ip->ip_p = IPPROTO_IPV6; 512 ip->ip_ttl = ip_stf_ttl; 513 ip->ip_len = htons(m->m_pkthdr.len); 514 if (ifp->if_flags & IFF_LINK1) 515 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos); 516 else 517 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos); 518 519 if (!stf_route_cache) { 520 cached_route = NULL; 521 goto sendit; 522 } 523 524 /* 525 * Do we have a cached route? 526 */ 527 mtx_lock(&(sc)->sc_ro_mtx); 528 dst4 = (struct sockaddr_in *)&sc->sc_ro.ro_dst; 529 if (dst4->sin_family != AF_INET || 530 bcmp(&dst4->sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)) != 0) { 531 /* cache route doesn't match */ 532 dst4->sin_family = AF_INET; 533 dst4->sin_len = sizeof(struct sockaddr_in); 534 bcopy(&ip->ip_dst, &dst4->sin_addr, sizeof(dst4->sin_addr)); 535 if (sc->sc_ro.ro_rt) { 536 RTFREE(sc->sc_ro.ro_rt); 537 sc->sc_ro.ro_rt = NULL; 538 } 539 } 540 541 if (sc->sc_ro.ro_rt == NULL) { 542 rtalloc_fib(&sc->sc_ro, sc->sc_fibnum); 543 if (sc->sc_ro.ro_rt == NULL) { 544 m_freem(m); 545 mtx_unlock(&(sc)->sc_ro_mtx); 546 ifp->if_oerrors++; 547 return ENETUNREACH; 548 } 549 } 550 cached_route = &sc->sc_ro; 551 552sendit: 553 M_SETFIB(m, sc->sc_fibnum); 554 ifp->if_opackets++; 555 error = ip_output(m, NULL, cached_route, 0, NULL, NULL); 556 557 if (cached_route != NULL) 558 mtx_unlock(&(sc)->sc_ro_mtx); 559 return error; 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 (stf_permit_rfc1918 == 0 && ( 571 (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 || 572 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 || 573 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)) 574 return 1; 575 576 return 0; 577} 578 579static int 580stf_checkaddr4(sc, in, inifp) 581 struct stf_softc *sc; 582 struct in_addr *in; 583 struct ifnet *inifp; /* incoming interface */ 584{ 585 struct in_ifaddr *ia4; 586 587 /* 588 * reject packets with the following address: 589 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8 590 */ 591 if (IN_MULTICAST(ntohl(in->s_addr))) 592 return -1; 593 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) { 594 case 0: case 127: case 255: 595 return -1; 596 } 597 598 /* 599 * reject packets with private address range. 600 * (requirement from RFC3056 section 2 1st paragraph) 601 */ 602 if (isrfc1918addr(in)) 603 return -1; 604 605 /* 606 * reject packets with broadcast 607 */ 608 IN_IFADDR_RLOCK(); 609 TAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) { 610 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) 611 continue; 612 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) { 613 IN_IFADDR_RUNLOCK(); 614 return -1; 615 } 616 } 617 IN_IFADDR_RUNLOCK(); 618 619 /* 620 * perform ingress filter 621 */ 622 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) { 623 struct sockaddr_in sin; 624 struct rtentry *rt; 625 626 bzero(&sin, sizeof(sin)); 627 sin.sin_family = AF_INET; 628 sin.sin_len = sizeof(struct sockaddr_in); 629 sin.sin_addr = *in; 630 rt = rtalloc1_fib((struct sockaddr *)&sin, 0, 631 0UL, sc->sc_fibnum); 632 if (!rt || rt->rt_ifp != inifp) { 633#if 0 634 log(LOG_WARNING, "%s: packet from 0x%x dropped " 635 "due to ingress filter\n", if_name(STF2IFP(sc)), 636 (u_int32_t)ntohl(sin.sin_addr.s_addr)); 637#endif 638 if (rt) 639 RTFREE_LOCKED(rt); 640 return -1; 641 } 642 RTFREE_LOCKED(rt); 643 } 644 645 return 0; 646} 647 648static int 649stf_checkaddr6(sc, in6, inifp) 650 struct stf_softc *sc; 651 struct in6_addr *in6; 652 struct ifnet *inifp; /* incoming interface */ 653{ 654 /* 655 * check 6to4 addresses 656 */ 657 if (IN6_IS_ADDR_6TO4(in6)) { 658 struct in_addr in4; 659 bcopy(GET_V4(in6), &in4, sizeof(in4)); 660 return stf_checkaddr4(sc, &in4, inifp); 661 } 662 663 /* 664 * reject anything that look suspicious. the test is implemented 665 * in ip6_input too, but we check here as well to 666 * (1) reject bad packets earlier, and 667 * (2) to be safe against future ip6_input change. 668 */ 669 if (IN6_IS_ADDR_V4COMPAT(in6) || IN6_IS_ADDR_V4MAPPED(in6)) 670 return -1; 671 672 return 0; 673} 674 675void 676in_stf_input(m, off) 677 struct mbuf *m; 678 int off; 679{ 680 int proto; 681 struct stf_softc *sc; 682 struct ip *ip; 683 struct ip6_hdr *ip6; 684 u_int8_t otos, itos; 685 struct ifnet *ifp; 686 687 proto = mtod(m, struct ip *)->ip_p; 688 689 if (proto != IPPROTO_IPV6) { 690 m_freem(m); 691 return; 692 } 693 694 ip = mtod(m, struct ip *); 695 696 sc = (struct stf_softc *)encap_getarg(m); 697 698 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) { 699 m_freem(m); 700 return; 701 } 702 703 ifp = STF2IFP(sc); 704 705#ifdef MAC 706 mac_ifnet_create_mbuf(ifp, m); 707#endif 708 709 /* 710 * perform sanity check against outer src/dst. 711 * for source, perform ingress filter as well. 712 */ 713 if (stf_checkaddr4(sc, &ip->ip_dst, NULL) < 0 || 714 stf_checkaddr4(sc, &ip->ip_src, m->m_pkthdr.rcvif) < 0) { 715 m_freem(m); 716 return; 717 } 718 719 otos = ip->ip_tos; 720 m_adj(m, off); 721 722 if (m->m_len < sizeof(*ip6)) { 723 m = m_pullup(m, sizeof(*ip6)); 724 if (!m) 725 return; 726 } 727 ip6 = mtod(m, struct ip6_hdr *); 728 729 /* 730 * perform sanity check against inner src/dst. 731 * for source, perform ingress filter as well. 732 */ 733 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 || 734 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) { 735 m_freem(m); 736 return; 737 } 738 739 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 740 if ((ifp->if_flags & IFF_LINK1) != 0) 741 ip_ecn_egress(ECN_ALLOWED, &otos, &itos); 742 else 743 ip_ecn_egress(ECN_NOCARE, &otos, &itos); 744 ip6->ip6_flow &= ~htonl(0xff << 20); 745 ip6->ip6_flow |= htonl((u_int32_t)itos << 20); 746 747 m->m_pkthdr.rcvif = ifp; 748 749 if (bpf_peers_present(ifp->if_bpf)) { 750 /* 751 * We need to prepend the address family as 752 * a four byte field. Cons up a dummy header 753 * to pacify bpf. This is safe because bpf 754 * will only read from the mbuf (i.e., it won't 755 * try to free it or keep a pointer a to it). 756 */ 757 u_int32_t af = AF_INET6; 758 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 759 } 760 761 /* 762 * Put the packet to the network layer input queue according to the 763 * specified address family. 764 * See net/if_gif.c for possible issues with packet processing 765 * reorder due to extra queueing. 766 */ 767 ifp->if_ipackets++; 768 ifp->if_ibytes += m->m_pkthdr.len; 769 M_SETFIB(m, ifp->if_fib); 770 netisr_dispatch(NETISR_IPV6, m); 771} 772 773/* ARGSUSED */ 774static void 775stf_rtrequest(cmd, rt, info) 776 int cmd; 777 struct rtentry *rt; 778 struct rt_addrinfo *info; 779{ 780 RT_LOCK_ASSERT(rt); 781 rt->rt_mtu = rt->rt_ifp->if_mtu; 782} 783 784static int 785stf_ioctl(ifp, cmd, data) 786 struct ifnet *ifp; 787 u_long cmd; 788 caddr_t data; 789{ 790 struct ifaddr *ifa; 791 struct ifreq *ifr; 792 struct sockaddr_in6 *sin6; 793 struct in_addr addr; 794 int error, mtu; 795 796 error = 0; 797 switch (cmd) { 798 case SIOCSIFADDR: 799 ifa = (struct ifaddr *)data; 800 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) { 801 error = EAFNOSUPPORT; 802 break; 803 } 804 sin6 = (struct sockaddr_in6 *)ifa->ifa_addr; 805 if (!IN6_IS_ADDR_6TO4(&sin6->sin6_addr)) { 806 error = EINVAL; 807 break; 808 } 809 bcopy(GET_V4(&sin6->sin6_addr), &addr, sizeof(addr)); 810 if (isrfc1918addr(&addr)) { 811 error = EINVAL; 812 break; 813 } 814 815 ifa->ifa_rtrequest = stf_rtrequest; 816 ifp->if_flags |= IFF_UP; 817 break; 818 819 case SIOCADDMULTI: 820 case SIOCDELMULTI: 821 ifr = (struct ifreq *)data; 822 if (ifr && ifr->ifr_addr.sa_family == AF_INET6) 823 ; 824 else 825 error = EAFNOSUPPORT; 826 break; 827 828 case SIOCGIFMTU: 829 break; 830 831 case SIOCSIFMTU: 832 ifr = (struct ifreq *)data; 833 mtu = ifr->ifr_mtu; 834 /* RFC 4213 3.2 ideal world MTU */ 835 if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20) 836 return (EINVAL); 837 ifp->if_mtu = mtu; 838 break; 839 840 default: 841 error = EINVAL; 842 break; 843 } 844 845 return error; 846} 847