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