1 /* $FreeBSD$ */ 2 /* $KAME: if_stf.c,v 1.73 2001/12/03 11:08:30 keiichi Exp $ */ 3 4 /*- 5 * SPDX-License-Identifier: BSD-3-Clause 6 * 7 * Copyright (C) 2000 WIDE Project. 8 * Copyright (c) 2010 Hiroki Sato <hrs@FreeBSD.org> 9 * Copyright (c) 2013 Ermal Luci <eri@FreeBSD.org> 10 * Copyright (c) 2017-2021 Rubicon Communications, LLC (Netgate) 11 * All rights reserved. 12 * 13 * Redistribution and use in source and binary forms, with or without 14 * modification, are permitted provided that the following conditions 15 * are met: 16 * 1. Redistributions of source code must retain the above copyright 17 * notice, this list of conditions and the following disclaimer. 18 * 2. Redistributions in binary form must reproduce the above copyright 19 * notice, this list of conditions and the following disclaimer in the 20 * documentation and/or other materials provided with the distribution. 21 * 3. Neither the name of the project nor the names of its contributors 22 * may be used to endorse or promote products derived from this software 23 * without specific prior written permission. 24 * 25 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 28 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 35 * SUCH DAMAGE. 36 */ 37 38 /* 39 * 6to4 interface, based on RFC3056. 40 * 41 * 6to4 interface is NOT capable of link-layer (I mean, IPv4) multicasting. 42 * There is no address mapping defined from IPv6 multicast address to IPv4 43 * address. Therefore, we do not have IFF_MULTICAST on the interface. 44 * 45 * Due to the lack of address mapping for link-local addresses, we cannot 46 * throw packets toward link-local addresses (fe80::x). Also, we cannot throw 47 * packets to link-local multicast addresses (ff02::x). 48 * 49 * Here are interesting symptoms due to the lack of link-local address: 50 * 51 * Unicast routing exchange: 52 * - RIPng: Impossible. Uses link-local multicast packet toward ff02::9, 53 * and link-local addresses as nexthop. 54 * - OSPFv6: Impossible. OSPFv6 assumes that there's link-local address 55 * assigned to the link, and makes use of them. Also, HELLO packets use 56 * link-local multicast addresses (ff02::5 and ff02::6). 57 * - BGP4+: Maybe. You can only use global address as nexthop, and global 58 * address as TCP endpoint address. 59 * 60 * Multicast routing protocols: 61 * - PIM: Hello packet cannot be used to discover adjacent PIM routers. 62 * Adjacent PIM routers must be configured manually (is it really spec-wise 63 * correct thing to do?). 64 * 65 * ICMPv6: 66 * - Redirects cannot be used due to the lack of link-local address. 67 * 68 * stf interface does not have, and will not need, a link-local address. 69 * It seems to have no real benefit and does not help the above symptoms much. 70 * Even if we assign link-locals to interface, we cannot really 71 * use link-local unicast/multicast on top of 6to4 cloud (since there's no 72 * encapsulation defined for link-local address), and the above analysis does 73 * not change. RFC3056 does not mandate the assignment of link-local address 74 * either. 75 * 76 * 6to4 interface has security issues. Refer to 77 * http://playground.iijlab.net/i-d/draft-itojun-ipv6-transition-abuse-00.txt 78 * for details. The code tries to filter out some of malicious packets. 79 * Note that there is no way to be 100% secure. 80 */ 81 82 #include <sys/param.h> 83 #include <sys/systm.h> 84 #include <sys/socket.h> 85 #include <sys/sockio.h> 86 #include <sys/mbuf.h> 87 #include <sys/endian.h> 88 #include <sys/errno.h> 89 #include <sys/kernel.h> 90 #include <sys/lock.h> 91 #include <sys/module.h> 92 #include <sys/priv.h> 93 #include <sys/proc.h> 94 #include <sys/queue.h> 95 #include <sys/sdt.h> 96 #include <sys/sysctl.h> 97 #include <machine/cpu.h> 98 99 #include <sys/malloc.h> 100 101 #include <net/if.h> 102 #include <net/if_var.h> 103 #include <net/if_private.h> 104 #include <net/if_clone.h> 105 #include <net/route.h> 106 #include <net/route/nhop.h> 107 #include <net/netisr.h> 108 #include <net/if_stf.h> 109 #include <net/if_types.h> 110 #include <net/vnet.h> 111 112 #include <netinet/in.h> 113 #include <netinet/in_fib.h> 114 #include <netinet/in_systm.h> 115 #include <netinet/ip.h> 116 #include <netinet/ip_var.h> 117 #include <netinet/in_var.h> 118 119 #include <netinet/ip6.h> 120 #include <netinet6/in6_fib.h> 121 #include <netinet6/ip6_var.h> 122 #include <netinet6/in6_var.h> 123 #include <netinet/ip_ecn.h> 124 125 #include <netinet/ip_encap.h> 126 127 #include <machine/stdarg.h> 128 129 #include <net/bpf.h> 130 131 #include <security/mac/mac_framework.h> 132 133 SDT_PROVIDER_DEFINE(if_stf); 134 SDT_PROBE_DEFINE3(if_stf, , encapcheck, in, "struct mbuf *", "int", "int"); 135 SDT_PROBE_DEFINE0(if_stf, , encapcheck, accept); 136 SDT_PROBE_DEFINE3(if_stf, , getsrcifa6, in, "struct ifnet *", 137 "struct in6_addr *", "struct in6_addr *"); 138 SDT_PROBE_DEFINE2(if_stf, , getsrcifa6, found, "struct in6_addr *", 139 "struct in6_addr *"); 140 SDT_PROBE_DEFINE0(if_stf, , getsrcifa6, notfound); 141 142 SDT_PROBE_DEFINE4(if_stf, , stf_output, in, "struct ifnet *", "struct mbuf *", 143 "struct sockaddr *", "struct route *"); 144 SDT_PROBE_DEFINE2(if_stf, , stf_output, error, "int", "int"); 145 SDT_PROBE_DEFINE1(if_stf, , stf_output, out, "int"); 146 147 SDT_PROBE_DEFINE3(if_stf, , checkaddr6, in, "struct stf_softc *", 148 "struct in6_addr *", "struct ifnet *"); 149 SDT_PROBE_DEFINE2(if_stf, , checkaddr6, out, "int", "int"); 150 151 SDT_PROBE_DEFINE3(if_stf, , stf_input, in, "struct mbuf *", "int", "int"); 152 SDT_PROBE_DEFINE2(if_stf, , stf_input, out, "int", "int"); 153 154 SDT_PROBE_DEFINE3(if_stf, , ioctl, sv4net, "struct in_addr *", 155 "struct in_addr *", "int"); 156 SDT_PROBE_DEFINE1(if_stf, , ioctl, sdstv4, "struct in_addr *"); 157 SDT_PROBE_DEFINE1(if_stf, , ioctl, ifaddr, "struct ifaddr *"); 158 159 SDT_PROBE_DEFINE4(if_stf, , getin4addr_in6, out, "struct in6_addr *", 160 "struct in6_addr *", "struct in6_addr *", "struct sockaddr_in *"); 161 162 SDT_PROBE_DEFINE2(if_stf, , getin4addr, in, "struct in6_addr *", "struct in6_addr *"); 163 SDT_PROBE_DEFINE1(if_stf, , getin4addr, out, "struct sockaddr_in *"); 164 165 SYSCTL_DECL(_net_link); 166 static SYSCTL_NODE(_net_link, IFT_STF, stf, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 167 "6to4 Interface"); 168 169 static int stf_permit_rfc1918 = 0; 170 SYSCTL_INT(_net_link_stf, OID_AUTO, permit_rfc1918, CTLFLAG_RWTUN, 171 &stf_permit_rfc1918, 0, "Permit the use of private IPv4 addresses"); 172 173 #define STFUNIT 0 174 175 #define IN6_IS_ADDR_6TO4(x) (ntohs((x)->s6_addr16[0]) == 0x2002) 176 177 /* 178 * XXX: Return a pointer with 16-bit aligned. Don't cast it to 179 * struct in_addr *; use bcopy() instead. 180 */ 181 #define GET_V4(x) (&(x)->s6_addr16[1]) 182 183 struct stf_softc { 184 struct ifnet *sc_ifp; 185 in_addr_t braddr; /* Border relay IPv4 address */ 186 in_addr_t srcv4_addr; /* Our IPv4 WAN address */ 187 u_int v4prefixlen; /* How much of the v4 address to include in our address. */ 188 u_int sc_fibnum; 189 const struct encaptab *encap_cookie; 190 }; 191 #define STF2IFP(sc) ((sc)->sc_ifp) 192 193 static const char stfname[] = "stf"; 194 195 static MALLOC_DEFINE(M_STF, stfname, "6to4 Tunnel Interface"); 196 static const int ip_stf_ttl = 40; 197 198 static int in_stf_input(struct mbuf *, int, int, void *); 199 static char *stfnames[] = {"stf0", "stf", "6to4", NULL}; 200 201 static int stfmodevent(module_t, int, void *); 202 static int stf_encapcheck(const struct mbuf *, int, int, void *); 203 static int stf_getsrcifa6(struct ifnet *, struct in6_addr *, struct in6_addr *); 204 static int stf_output(struct ifnet *, struct mbuf *, const struct sockaddr *, 205 struct route *); 206 static int isrfc1918addr(struct in_addr *); 207 static int stf_checkaddr4(struct stf_softc *, struct in_addr *, 208 struct ifnet *); 209 static int stf_checkaddr6(struct stf_softc *, struct in6_addr *, 210 struct ifnet *); 211 static struct sockaddr_in *stf_getin4addr_in6(struct stf_softc *, 212 struct sockaddr_in *, struct in6_addr, struct in6_addr, 213 struct in6_addr); 214 static struct sockaddr_in *stf_getin4addr(struct stf_softc *, 215 struct sockaddr_in *, struct in6_addr, struct in6_addr); 216 static int stf_ioctl(struct ifnet *, u_long, caddr_t); 217 218 VNET_DEFINE_STATIC(struct if_clone *, stf_cloner); 219 #define V_stf_cloner VNET(stf_cloner) 220 221 static const struct encap_config ipv4_encap_cfg = { 222 .proto = IPPROTO_IPV6, 223 .min_length = sizeof(struct ip), 224 .exact_match = (sizeof(in_addr_t) << 3) + 8, 225 .check = stf_encapcheck, 226 .input = in_stf_input 227 }; 228 229 static int 230 stf_clone_match(struct if_clone *ifc, const char *name) 231 { 232 int i; 233 234 for(i = 0; stfnames[i] != NULL; i++) { 235 if (strcmp(stfnames[i], name) == 0) 236 return (1); 237 } 238 239 return (0); 240 } 241 242 static int 243 stf_clone_create(struct if_clone *ifc, char *name, size_t len, 244 struct ifc_data *ifd, struct ifnet **ifpp) 245 { 246 char *dp; 247 int err, unit, wildcard; 248 struct stf_softc *sc; 249 struct ifnet *ifp; 250 251 err = ifc_name2unit(name, &unit); 252 if (err != 0) 253 return (err); 254 wildcard = (unit < 0); 255 256 /* 257 * We can only have one unit, but since unit allocation is 258 * already locked, we use it to keep from allocating extra 259 * interfaces. 260 */ 261 unit = STFUNIT; 262 err = ifc_alloc_unit(ifc, &unit); 263 if (err != 0) 264 return (err); 265 266 sc = malloc(sizeof(struct stf_softc), M_STF, M_WAITOK | M_ZERO); 267 ifp = STF2IFP(sc) = if_alloc(IFT_STF); 268 if (ifp == NULL) { 269 free(sc, M_STF); 270 ifc_free_unit(ifc, unit); 271 return (ENOSPC); 272 } 273 ifp->if_softc = sc; 274 sc->sc_fibnum = curthread->td_proc->p_fibnum; 275 276 /* 277 * Set the name manually rather then using if_initname because 278 * we don't conform to the default naming convention for interfaces. 279 * In the wildcard case, we need to update the name. 280 */ 281 if (wildcard) { 282 for (dp = name; *dp != '\0'; dp++); 283 if (snprintf(dp, len - (dp-name), "%d", unit) > 284 len - (dp-name) - 1) { 285 /* 286 * This can only be a programmer error and 287 * there's no straightforward way to recover if 288 * it happens. 289 */ 290 panic("if_clone_create(): interface name too long"); 291 } 292 } 293 strlcpy(ifp->if_xname, name, IFNAMSIZ); 294 ifp->if_dname = stfname; 295 ifp->if_dunit = IF_DUNIT_NONE; 296 297 sc->encap_cookie = ip_encap_attach(&ipv4_encap_cfg, sc, M_WAITOK); 298 if (sc->encap_cookie == NULL) { 299 if_printf(ifp, "attach failed\n"); 300 free(sc, M_STF); 301 ifc_free_unit(ifc, unit); 302 return (ENOMEM); 303 } 304 305 ifp->if_mtu = IPV6_MMTU; 306 ifp->if_ioctl = stf_ioctl; 307 ifp->if_output = stf_output; 308 ifp->if_snd.ifq_maxlen = ifqmaxlen; 309 if_attach(ifp); 310 bpfattach(ifp, DLT_NULL, sizeof(u_int32_t)); 311 *ifpp = ifp; 312 313 return (0); 314 } 315 316 static int 317 stf_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags) 318 { 319 struct stf_softc *sc = ifp->if_softc; 320 int err __unused; 321 322 err = ip_encap_detach(sc->encap_cookie); 323 KASSERT(err == 0, ("Unexpected error detaching encap_cookie")); 324 bpfdetach(ifp); 325 if_detach(ifp); 326 if_free(ifp); 327 328 free(sc, M_STF); 329 ifc_free_unit(ifc, STFUNIT); 330 331 return (0); 332 } 333 334 static void 335 vnet_stf_init(const void *unused __unused) 336 { 337 struct if_clone_addreq req = { 338 .match_f = stf_clone_match, 339 .create_f = stf_clone_create, 340 .destroy_f = stf_clone_destroy, 341 }; 342 V_stf_cloner = ifc_attach_cloner(stfname, &req); 343 } 344 VNET_SYSINIT(vnet_stf_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_init, NULL); 345 346 static void 347 vnet_stf_uninit(const void *unused __unused) 348 { 349 if_clone_detach(V_stf_cloner); 350 V_stf_cloner = NULL; 351 } 352 VNET_SYSUNINIT(vnet_stf_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_stf_uninit, 353 NULL); 354 355 static int 356 stfmodevent(module_t mod, int type, void *data) 357 { 358 359 switch (type) { 360 case MOD_LOAD: 361 /* Done in vnet_stf_init() */ 362 break; 363 case MOD_UNLOAD: 364 /* Done in vnet_stf_uninit() */ 365 break; 366 default: 367 return (EOPNOTSUPP); 368 } 369 370 return (0); 371 } 372 373 static moduledata_t stf_mod = { 374 "if_stf", 375 stfmodevent, 376 0 377 }; 378 379 DECLARE_MODULE(if_stf, stf_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 380 MODULE_VERSION(if_stf, 2); 381 382 static int 383 stf_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 384 { 385 struct ip ip; 386 struct stf_softc *sc; 387 struct in6_addr addr6, mask6; 388 struct sockaddr_in sin4addr, sin4mask; 389 390 SDT_PROBE3(if_stf, , encapcheck, in, m, off, proto); 391 392 sc = (struct stf_softc *)arg; 393 if (sc == NULL) 394 return (0); 395 396 if ((STF2IFP(sc)->if_flags & IFF_UP) == 0) 397 return (0); 398 399 /* IFF_LINK0 means "no decapsulation" */ 400 if ((STF2IFP(sc)->if_flags & IFF_LINK0) != 0) 401 return (0); 402 403 if (proto != IPPROTO_IPV6) 404 return (0); 405 406 m_copydata(m, 0, sizeof(ip), (caddr_t)&ip); 407 408 if (ip.ip_v != 4) 409 return (0); 410 411 if (stf_getsrcifa6(STF2IFP(sc), &addr6, &mask6) != 0) 412 return (0); 413 414 if (sc->srcv4_addr != INADDR_ANY) { 415 sin4addr.sin_addr.s_addr = sc->srcv4_addr; 416 sin4addr.sin_family = AF_INET; 417 } else 418 if (stf_getin4addr(sc, &sin4addr, addr6, mask6) == NULL) 419 return (0); 420 421 if (sin4addr.sin_addr.s_addr != ip.ip_dst.s_addr) 422 return (0); 423 424 if (IN6_IS_ADDR_6TO4(&addr6)) { 425 /* 426 * 6to4 (RFC 3056). 427 * Check if IPv4 src matches the IPv4 address derived 428 * from the local 6to4 address masked by prefixmask. 429 * success on: src = 10.1.1.1, ia6->ia_addr = 2002:0a00:.../24 430 * fail on: src = 10.1.1.1, ia6->ia_addr = 2002:0b00:.../24 431 */ 432 memcpy(&sin4mask.sin_addr, GET_V4(&mask6), 433 sizeof(sin4mask.sin_addr)); 434 if ((sin4addr.sin_addr.s_addr & sin4mask.sin_addr.s_addr) != 435 (ip.ip_src.s_addr & sin4mask.sin_addr.s_addr)) 436 return (0); 437 } else { 438 /* 6rd (RFC 5569) */ 439 /* 440 * No restriction on the src address in the case of 441 * 6rd because the stf(4) interface always has a 442 * prefix which covers whole of IPv4 src address 443 * range. So, stf_output() will catch all of 444 * 6rd-capsuled IPv4 traffic with suspicious inner dst 445 * IPv4 address (i.e. the IPv6 destination address is 446 * one the admin does not like to route to outside), 447 * and then it discard them silently. 448 */ 449 } 450 451 SDT_PROBE0(if_stf, , encapcheck, accept); 452 453 /* stf interface makes single side match only */ 454 return (32); 455 } 456 457 static int 458 stf_getsrcifa6(struct ifnet *ifp, struct in6_addr *addr, struct in6_addr *mask) 459 { 460 struct ifaddr *ia; 461 struct in_ifaddr *ia4; 462 struct in6_addr addr6, mask6; 463 struct sockaddr_in sin4; 464 struct stf_softc *sc; 465 struct in_addr in; 466 467 NET_EPOCH_ASSERT(); 468 469 sc = ifp->if_softc; 470 471 SDT_PROBE3(if_stf, , getsrcifa6, in, ifp, addr, mask); 472 473 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 474 if (ia->ifa_addr->sa_family != AF_INET6) 475 continue; 476 477 addr6 = *IFA_IN6(ia); 478 mask6 = *IFA_MASKIN6(ia); 479 if (sc->srcv4_addr != INADDR_ANY) 480 bcopy(&sc->srcv4_addr, &in, sizeof(in)); 481 else { 482 if (stf_getin4addr(sc, &sin4, addr6, mask6) == NULL) 483 continue; 484 bcopy(&sin4.sin_addr, &in, sizeof(in)); 485 } 486 487 CK_LIST_FOREACH(ia4, INADDR_HASH(in.s_addr), ia_hash) 488 if (ia4->ia_addr.sin_addr.s_addr == in.s_addr) 489 break; 490 if (ia4 == NULL) 491 continue; 492 493 *addr = addr6; 494 *mask = mask6; 495 496 SDT_PROBE2(if_stf, , getsrcifa6, found, addr, mask); 497 498 return (0); 499 } 500 501 SDT_PROBE0(if_stf, , getsrcifa6, notfound); 502 503 return (ENOENT); 504 } 505 506 static int 507 stf_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, 508 struct route *ro) 509 { 510 struct stf_softc *sc; 511 const struct sockaddr_in6 *dst6; 512 struct sockaddr_in dst4, src4; 513 u_int8_t tos; 514 struct ip *ip; 515 struct ip6_hdr *ip6; 516 struct in6_addr addr6, mask6; 517 int error; 518 519 SDT_PROBE4(if_stf, , stf_output, in, ifp, m, dst, ro); 520 521 #ifdef MAC 522 error = mac_ifnet_check_transmit(ifp, m); 523 if (error) { 524 m_freem(m); 525 SDT_PROBE2(if_stf, , stf_output, error, error, __LINE__); 526 return (error); 527 } 528 #endif 529 530 sc = ifp->if_softc; 531 dst6 = (const struct sockaddr_in6 *)dst; 532 533 /* just in case */ 534 if ((ifp->if_flags & IFF_UP) == 0) { 535 m_freem(m); 536 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 537 SDT_PROBE2(if_stf, , stf_output, error, ENETDOWN, __LINE__); 538 return (ENETDOWN); 539 } 540 541 /* 542 * If we don't have an ip4 address that match my inner ip6 address, 543 * we shouldn't generate output. Without this check, we'll end up 544 * using wrong IPv4 source. 545 */ 546 if (stf_getsrcifa6(ifp, &addr6, &mask6) != 0) { 547 m_freem(m); 548 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 549 SDT_PROBE2(if_stf, , stf_output, error, ENETDOWN, __LINE__); 550 return (ENETDOWN); 551 } 552 553 if (m->m_len < sizeof(*ip6)) { 554 m = m_pullup(m, sizeof(*ip6)); 555 if (!m) { 556 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 557 SDT_PROBE2(if_stf, , stf_output, error, ENOBUFS, 558 __LINE__); 559 return (ENOBUFS); 560 } 561 } 562 ip6 = mtod(m, struct ip6_hdr *); 563 tos = IPV6_TRAFFIC_CLASS(ip6); 564 565 /* 566 * Pickup the right outer dst addr from the list of candidates. 567 * ip6_dst has priority as it may be able to give us shorter IPv4 hops. 568 */ 569 if (stf_getin4addr_in6(sc, &dst4, addr6, mask6, 570 ip6->ip6_dst) == NULL) { 571 if (sc->braddr != INADDR_ANY) 572 dst4.sin_addr.s_addr = sc->braddr; 573 else if (stf_getin4addr_in6(sc, &dst4, addr6, mask6, 574 dst6->sin6_addr) == NULL) { 575 m_freem(m); 576 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 577 SDT_PROBE2(if_stf, , stf_output, error, ENETUNREACH, 578 __LINE__); 579 return (ENETUNREACH); 580 } 581 } 582 583 if (bpf_peers_present(ifp->if_bpf)) { 584 /* 585 * We need to prepend the address family as 586 * a four byte field. Cons up a dummy header 587 * to pacify bpf. This is safe because bpf 588 * will only read from the mbuf (i.e., it won't 589 * try to free it or keep a pointer a to it). 590 */ 591 u_int af = AF_INET6; 592 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 593 } 594 595 M_PREPEND(m, sizeof(struct ip), M_NOWAIT); 596 if (m == NULL) { 597 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 598 SDT_PROBE2(if_stf, , stf_output, error, ENOBUFS, __LINE__); 599 return (ENOBUFS); 600 } 601 ip = mtod(m, struct ip *); 602 603 bzero(ip, sizeof(*ip)); 604 605 if (sc->srcv4_addr != INADDR_ANY) 606 src4.sin_addr.s_addr = sc->srcv4_addr; 607 else if (stf_getin4addr(sc, &src4, addr6, mask6) == NULL) { 608 m_freem(m); 609 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); 610 SDT_PROBE2(if_stf, , stf_output, error, ENETUNREACH, __LINE__); 611 return (ENETUNREACH); 612 } 613 bcopy(&src4.sin_addr, &ip->ip_src, sizeof(ip->ip_src)); 614 bcopy(&dst4.sin_addr, &ip->ip_dst, sizeof(ip->ip_dst)); 615 616 ip->ip_p = IPPROTO_IPV6; 617 ip->ip_ttl = ip_stf_ttl; 618 ip->ip_len = htons(m->m_pkthdr.len); 619 if (ifp->if_flags & IFF_LINK1) 620 ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos); 621 else 622 ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos); 623 624 M_SETFIB(m, sc->sc_fibnum); 625 if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); 626 error = ip_output(m, NULL, NULL, 0, NULL, NULL); 627 628 SDT_PROBE1(if_stf, , stf_output, out, error); 629 return (error); 630 } 631 632 static int 633 isrfc1918addr(struct in_addr *in) 634 { 635 /* 636 * returns 1 if private address range: 637 * 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16 638 */ 639 if (stf_permit_rfc1918 == 0 && ( 640 (ntohl(in->s_addr) & 0xff000000) >> 24 == 10 || 641 (ntohl(in->s_addr) & 0xfff00000) >> 16 == 172 * 256 + 16 || 642 (ntohl(in->s_addr) & 0xffff0000) >> 16 == 192 * 256 + 168)) 643 return (1); 644 645 return (0); 646 } 647 648 static int 649 stf_checkaddr4(struct stf_softc *sc, struct in_addr *in, struct ifnet *inifp) 650 { 651 struct in_ifaddr *ia4; 652 653 /* 654 * reject packets with the following address: 655 * 224.0.0.0/4 0.0.0.0/8 127.0.0.0/8 255.0.0.0/8 656 */ 657 if (IN_MULTICAST(ntohl(in->s_addr))) 658 return (-1); 659 switch ((ntohl(in->s_addr) & 0xff000000) >> 24) { 660 case 0: case 127: case 255: 661 return (-1); 662 } 663 664 /* 665 * reject packets with broadcast 666 */ 667 CK_STAILQ_FOREACH(ia4, &V_in_ifaddrhead, ia_link) { 668 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) 669 continue; 670 if (in->s_addr == ia4->ia_broadaddr.sin_addr.s_addr) { 671 return (-1); 672 } 673 } 674 675 /* 676 * perform ingress filter 677 */ 678 if (sc && (STF2IFP(sc)->if_flags & IFF_LINK2) == 0 && inifp) { 679 struct nhop_object *nh; 680 681 NET_EPOCH_ASSERT(); 682 nh = fib4_lookup(sc->sc_fibnum, *in, 0, 0, 0); 683 if (nh == NULL) 684 return (-1); 685 686 if (nh->nh_ifp != inifp) 687 return (-1); 688 } 689 690 return (0); 691 } 692 693 static int 694 stf_checkaddr6(struct stf_softc *sc, struct in6_addr *in6, struct ifnet *inifp) 695 { 696 SDT_PROBE3(if_stf, , checkaddr6, in, sc, in6, inifp); 697 698 /* 699 * check 6to4 addresses 700 */ 701 if (IN6_IS_ADDR_6TO4(in6)) { 702 struct in_addr in4; 703 int ret; 704 705 bcopy(GET_V4(in6), &in4, sizeof(in4)); 706 ret = stf_checkaddr4(sc, &in4, inifp); 707 SDT_PROBE2(if_stf, , checkaddr6, out, ret, __LINE__); 708 return (ret); 709 } 710 711 /* 712 * reject anything that look suspicious. the test is implemented 713 * in ip6_input too, but we check here as well to 714 * (1) reject bad packets earlier, and 715 * (2) to be safe against future ip6_input change. 716 */ 717 if (IN6_IS_ADDR_V4COMPAT(in6)) { 718 SDT_PROBE2(if_stf, , checkaddr6, out, -1, __LINE__); 719 return (-1); 720 } 721 722 if (IN6_IS_ADDR_V4MAPPED(in6)) { 723 SDT_PROBE2(if_stf, , checkaddr6, out, -1, __LINE__); 724 return (-1); 725 } 726 727 SDT_PROBE2(if_stf, , checkaddr6, out, 0, __LINE__); 728 return (0); 729 } 730 731 static int 732 in_stf_input(struct mbuf *m, int off, int proto, void *arg) 733 { 734 struct stf_softc *sc = arg; 735 struct ip ip; 736 struct ip6_hdr *ip6; 737 u_int8_t otos, itos; 738 struct ifnet *ifp; 739 struct nhop_object *nh; 740 741 NET_EPOCH_ASSERT(); 742 743 SDT_PROBE3(if_stf, , stf_input, in, m, off, proto); 744 745 if (proto != IPPROTO_IPV6) { 746 m_freem(m); 747 SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__); 748 return (IPPROTO_DONE); 749 } 750 751 m_copydata(m, 0, sizeof(struct ip), (caddr_t)&ip); 752 if (sc == NULL || (STF2IFP(sc)->if_flags & IFF_UP) == 0) { 753 m_freem(m); 754 SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__); 755 return (IPPROTO_DONE); 756 } 757 758 ifp = STF2IFP(sc); 759 760 #ifdef MAC 761 mac_ifnet_create_mbuf(ifp, m); 762 #endif 763 764 /* 765 * perform sanity check against outer src/dst. 766 * for source, perform ingress filter as well. 767 */ 768 if (stf_checkaddr4(sc, &ip.ip_dst, NULL) < 0 || 769 stf_checkaddr4(sc, &ip.ip_src, m->m_pkthdr.rcvif) < 0) { 770 m_freem(m); 771 SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__); 772 return (IPPROTO_DONE); 773 } 774 775 otos = ip.ip_tos; 776 m_adj(m, off); 777 778 if (m->m_len < sizeof(*ip6)) { 779 m = m_pullup(m, sizeof(*ip6)); 780 if (!m) { 781 SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, 782 __LINE__); 783 return (IPPROTO_DONE); 784 } 785 } 786 ip6 = mtod(m, struct ip6_hdr *); 787 788 /* 789 * perform sanity check against inner src/dst. 790 * for source, perform ingress filter as well. 791 */ 792 if (stf_checkaddr6(sc, &ip6->ip6_dst, NULL) < 0 || 793 stf_checkaddr6(sc, &ip6->ip6_src, m->m_pkthdr.rcvif) < 0) { 794 m_freem(m); 795 SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__); 796 return (IPPROTO_DONE); 797 } 798 799 /* 800 * reject packets with private address range. 801 * (requirement from RFC3056 section 2 1st paragraph) 802 */ 803 if ((IN6_IS_ADDR_6TO4(&ip6->ip6_src) && isrfc1918addr(&ip.ip_src)) || 804 (IN6_IS_ADDR_6TO4(&ip6->ip6_dst) && isrfc1918addr(&ip.ip_dst))) { 805 m_freem(m); 806 SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__); 807 return (IPPROTO_DONE); 808 } 809 810 /* 811 * Ignore if the destination is the same stf interface because 812 * all of valid IPv6 outgoing traffic should go interfaces 813 * except for it. 814 */ 815 nh = fib6_lookup(sc->sc_fibnum, &ip6->ip6_dst, 0, 0, 0); 816 if (nh == NULL) { 817 m_free(m); 818 SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__); 819 return (IPPROTO_DONE); 820 } 821 if ((nh->nh_ifp == ifp) && 822 (!IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &nh->gw6_sa.sin6_addr))) { 823 m_free(m); 824 SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__); 825 return (IPPROTO_DONE); 826 } 827 828 itos = IPV6_TRAFFIC_CLASS(ip6); 829 if ((ifp->if_flags & IFF_LINK1) != 0) 830 ip_ecn_egress(ECN_ALLOWED, &otos, &itos); 831 else 832 ip_ecn_egress(ECN_NOCARE, &otos, &itos); 833 ip6->ip6_flow &= ~htonl(0xff << 20); 834 ip6->ip6_flow |= htonl((u_int32_t)itos << 20); 835 836 m->m_pkthdr.rcvif = ifp; 837 838 if (bpf_peers_present(ifp->if_bpf)) { 839 /* 840 * We need to prepend the address family as 841 * a four byte field. Cons up a dummy header 842 * to pacify bpf. This is safe because bpf 843 * will only read from the mbuf (i.e., it won't 844 * try to free it or keep a pointer a to it). 845 */ 846 u_int32_t af = AF_INET6; 847 bpf_mtap2(ifp->if_bpf, &af, sizeof(af), m); 848 } 849 850 /* 851 * Put the packet to the network layer input queue according to the 852 * specified address family. 853 * See net/if_gif.c for possible issues with packet processing 854 * reorder due to extra queueing. 855 */ 856 if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1); 857 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); 858 M_SETFIB(m, ifp->if_fib); 859 netisr_dispatch(NETISR_IPV6, m); 860 SDT_PROBE2(if_stf, , stf_input, out, IPPROTO_DONE, __LINE__); 861 return (IPPROTO_DONE); 862 } 863 864 static struct sockaddr_in * 865 stf_getin4addr_in6(struct stf_softc *sc, struct sockaddr_in *sin, 866 struct in6_addr addr6, struct in6_addr mask6, struct in6_addr in6) 867 { 868 int i; 869 struct sockaddr_in *out; 870 871 /* 872 * When (src addr & src mask) != (in6 & src mask), 873 * the dst is not in the 6rd domain. The IPv4 address must 874 * not be used. 875 */ 876 for (i = 0; i < sizeof(addr6); i++) { 877 if ((((u_char *)&addr6)[i] & ((u_char *)&mask6)[i]) != 878 (((u_char *)&in6)[i] & ((u_char *)&mask6)[i])) { 879 SDT_PROBE4(if_stf, , getin4addr_in6, out, &addr6, 880 &mask6, &in6, NULL); 881 return (NULL); 882 } 883 } 884 885 /* After the mask check, use in6 instead of addr6. */ 886 out = stf_getin4addr(sc, sin, in6, mask6); 887 SDT_PROBE4(if_stf, , getin4addr_in6, out, &addr6, &mask6, &in6, out); 888 return (out); 889 } 890 891 static struct sockaddr_in * 892 stf_getin4addr(struct stf_softc *sc, struct sockaddr_in *sin, 893 struct in6_addr addr6, struct in6_addr mask6) 894 { 895 struct in_addr *in; 896 897 SDT_PROBE2(if_stf, , getin4addr, in, &addr6, &mask6); 898 899 memset(sin, 0, sizeof(*sin)); 900 in = &sin->sin_addr; 901 if (IN6_IS_ADDR_6TO4(&addr6)) { 902 /* 6to4 (RFC 3056) */ 903 bcopy(GET_V4(&addr6), in, sizeof(*in)); 904 if (isrfc1918addr(in)) 905 return (NULL); 906 } else { 907 /* 6rd (RFC 5569) */ 908 in_addr_t v4prefix; 909 uint8_t *v6 = (uint8_t*)&addr6; 910 uint64_t v6prefix; 911 u_int plen; 912 u_int v4suffixlen; 913 914 v4prefix = 0; 915 if (sc->v4prefixlen < 32) { 916 v4suffixlen = 32 - sc->v4prefixlen; 917 v4prefix = ntohl(sc->srcv4_addr) & 918 (0xffffffffU << v4suffixlen); 919 } else { 920 MPASS(sc->v4prefixlen == 32); 921 v4suffixlen = 32; 922 } 923 924 plen = in6_mask2len(&mask6, NULL); 925 if (plen > 64) 926 return (NULL); 927 928 /* To make this simple we do not support prefixes longer than 929 * 64 bits. RFC5969 says "a 6rd delegated prefix SHOULD be /64 930 * or shorter." so this is a moderately safe assumption. */ 931 v6prefix = be64toh(*(uint64_t *)v6); 932 933 /* Shift away the v6 prefix itself. */ 934 v6prefix <<= plen; 935 v6prefix >>= plen; 936 937 /* Now shift away everything after the v4 address. */ 938 v6prefix >>= 64 - plen - v4suffixlen; 939 940 sin->sin_addr.s_addr = htonl(v4prefix | (uint32_t)v6prefix); 941 } 942 943 SDT_PROBE1(if_stf, , getin4addr, out, sin); 944 945 return (sin); 946 } 947 948 static int 949 stf_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) 950 { 951 struct ifaddr *ifa; 952 struct ifdrv *ifd; 953 struct ifreq *ifr; 954 struct sockaddr_in sin4; 955 struct stf_softc *sc_cur; 956 struct stfv4args args; 957 int error, mtu; 958 959 error = 0; 960 sc_cur = ifp->if_softc; 961 962 switch (cmd) { 963 case SIOCSDRVSPEC: 964 ifd = (struct ifdrv *)data; 965 error = priv_check(curthread, PRIV_NET_ADDIFADDR); 966 if (error) 967 break; 968 if (ifd->ifd_cmd == STF6RD_SV4NET) { 969 if (ifd->ifd_len != sizeof(args)) { 970 error = EINVAL; 971 break; 972 } 973 bzero(&args, sizeof(args)); 974 error = copyin(ifd->ifd_data, &args, ifd->ifd_len); 975 if (error) 976 break; 977 978 if (args.v4_prefixlen < 1 || args.v4_prefixlen > 32) { 979 error = EINVAL; 980 break; 981 } 982 983 bcopy(&args.srcv4_addr, &sc_cur->srcv4_addr, 984 sizeof(sc_cur->srcv4_addr)); 985 sc_cur->v4prefixlen = args.v4_prefixlen; 986 SDT_PROBE3(if_stf, , ioctl, sv4net, sc_cur->srcv4_addr, 987 sc_cur->srcv4_addr, sc_cur->v4prefixlen); 988 } else if (ifd->ifd_cmd == STF6RD_SBR) { 989 if (ifd->ifd_len != sizeof(args)) { 990 error = EINVAL; 991 break; 992 } 993 bzero(&args, sizeof(args)); 994 error = copyin(ifd->ifd_data, &args, ifd->ifd_len); 995 if (error) 996 break; 997 sc_cur->braddr = args.braddr.s_addr; 998 SDT_PROBE1(if_stf, , ioctl, sdstv4, 999 sc_cur->braddr); 1000 } else 1001 error = EINVAL; 1002 break; 1003 case SIOCGDRVSPEC: 1004 ifd = (struct ifdrv *)data; 1005 if (ifd->ifd_cmd != STF6RD_GV4NET) { 1006 error = EINVAL; 1007 break; 1008 } 1009 if (ifd->ifd_len != sizeof(args)) { 1010 error = EINVAL; 1011 break; 1012 } 1013 bzero(&args, sizeof(args)); 1014 args.srcv4_addr.s_addr = sc_cur->srcv4_addr; 1015 args.braddr.s_addr = sc_cur->braddr; 1016 args.v4_prefixlen = sc_cur->v4prefixlen; 1017 error = copyout(&args, ifd->ifd_data, ifd->ifd_len); 1018 break; 1019 case SIOCSIFADDR: 1020 ifa = (struct ifaddr *)data; 1021 SDT_PROBE1(if_stf, , ioctl, ifaddr, ifa); 1022 if (ifa == NULL || ifa->ifa_addr->sa_family != AF_INET6) { 1023 error = EAFNOSUPPORT; 1024 break; 1025 } 1026 if (stf_getin4addr(sc_cur, &sin4, 1027 satosin6(ifa->ifa_addr)->sin6_addr, 1028 satosin6(ifa->ifa_netmask)->sin6_addr) == NULL) { 1029 error = EINVAL; 1030 break; 1031 } 1032 ifp->if_flags |= IFF_UP; 1033 ifp->if_drv_flags |= IFF_DRV_RUNNING; 1034 break; 1035 1036 case SIOCADDMULTI: 1037 case SIOCDELMULTI: 1038 ifr = (struct ifreq *)data; 1039 if (ifr && ifr->ifr_addr.sa_family == AF_INET6) 1040 ; 1041 else 1042 error = EAFNOSUPPORT; 1043 break; 1044 1045 case SIOCGIFMTU: 1046 break; 1047 1048 case SIOCSIFMTU: 1049 ifr = (struct ifreq *)data; 1050 mtu = ifr->ifr_mtu; 1051 /* RFC 4213 3.2 ideal world MTU */ 1052 if (mtu < IPV6_MINMTU || mtu > IF_MAXMTU - 20) 1053 return (EINVAL); 1054 ifp->if_mtu = mtu; 1055 break; 1056 1057 default: 1058 error = EINVAL; 1059 break; 1060 } 1061 1062 return (error); 1063 } 1064