1 /*- 2 * Copyright (c) 2014, 2018 Andrey V. Elsukov <ae@FreeBSD.org> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include "opt_inet.h" 31 #include "opt_inet6.h" 32 33 #include <sys/param.h> 34 #include <sys/jail.h> 35 #include <sys/systm.h> 36 #include <sys/socket.h> 37 #include <sys/socketvar.h> 38 #include <sys/sockio.h> 39 #include <sys/mbuf.h> 40 #include <sys/errno.h> 41 #include <sys/kernel.h> 42 #include <sys/sysctl.h> 43 #include <sys/malloc.h> 44 #include <sys/proc.h> 45 46 #include <net/if.h> 47 #include <net/if_var.h> 48 #include <net/vnet.h> 49 50 #include <netinet/in.h> 51 #ifdef INET 52 #include <net/ethernet.h> 53 #include <netinet/ip.h> 54 #endif 55 #include <netinet/in_pcb.h> 56 #include <netinet/ip_encap.h> 57 #include <netinet/ip_var.h> 58 #include <netinet/ip6.h> 59 #include <netinet/udp.h> 60 #include <netinet/udp_var.h> 61 #include <netinet6/ip6_var.h> 62 #include <netinet6/in6_var.h> 63 #include <netinet6/scope6_var.h> 64 #include <net/if_gre.h> 65 66 VNET_DEFINE(int, ip6_gre_hlim) = IPV6_DEFHLIM; 67 #define V_ip6_gre_hlim VNET(ip6_gre_hlim) 68 69 SYSCTL_DECL(_net_inet6_ip6); 70 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, grehlim, CTLFLAG_VNET | CTLFLAG_RW, 71 &VNET_NAME(ip6_gre_hlim), 0, "Default hop limit for encapsulated packets"); 72 73 struct in6_gre_socket { 74 struct gre_socket base; 75 struct in6_addr addr; /* scope zone id is embedded */ 76 }; 77 VNET_DEFINE_STATIC(struct gre_sockets *, ipv6_sockets) = NULL; 78 VNET_DEFINE_STATIC(struct gre_list *, ipv6_hashtbl) = NULL; 79 VNET_DEFINE_STATIC(struct gre_list *, ipv6_srchashtbl) = NULL; 80 #define V_ipv6_sockets VNET(ipv6_sockets) 81 #define V_ipv6_hashtbl VNET(ipv6_hashtbl) 82 #define V_ipv6_srchashtbl VNET(ipv6_srchashtbl) 83 #define GRE_HASH(src, dst) (V_ipv6_hashtbl[\ 84 in6_gre_hashval((src), (dst)) & (GRE_HASH_SIZE - 1)]) 85 #define GRE_SRCHASH(src) (V_ipv6_srchashtbl[\ 86 fnv_32_buf((src), sizeof(*src), FNV1_32_INIT) & (GRE_HASH_SIZE - 1)]) 87 #define GRE_SOCKHASH(src) (V_ipv6_sockets[\ 88 fnv_32_buf((src), sizeof(*src), FNV1_32_INIT) & (GRE_HASH_SIZE - 1)]) 89 #define GRE_HASH_SC(sc) GRE_HASH(&(sc)->gre_oip6.ip6_src,\ 90 &(sc)->gre_oip6.ip6_dst) 91 92 static uint32_t 93 in6_gre_hashval(const struct in6_addr *src, const struct in6_addr *dst) 94 { 95 uint32_t ret; 96 97 ret = fnv_32_buf(src, sizeof(*src), FNV1_32_INIT); 98 return (fnv_32_buf(dst, sizeof(*dst), ret)); 99 } 100 101 static struct gre_socket* 102 in6_gre_lookup_socket(const struct in6_addr *addr) 103 { 104 struct gre_socket *gs; 105 struct in6_gre_socket *s; 106 107 CK_LIST_FOREACH(gs, &GRE_SOCKHASH(addr), chain) { 108 s = __containerof(gs, struct in6_gre_socket, base); 109 if (IN6_ARE_ADDR_EQUAL(&s->addr, addr)) 110 break; 111 } 112 return (gs); 113 } 114 115 static int 116 in6_gre_checkdup(const struct gre_softc *sc, const struct in6_addr *src, 117 const struct in6_addr *dst, uint32_t opts) 118 { 119 struct gre_list *head; 120 struct gre_softc *tmp; 121 struct gre_socket *gs; 122 123 if (sc->gre_family == AF_INET6 && 124 IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_src, src) && 125 IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_dst, dst) && 126 (sc->gre_options & GRE_UDPENCAP) == (opts & GRE_UDPENCAP)) 127 return (EEXIST); 128 129 if (opts & GRE_UDPENCAP) { 130 gs = in6_gre_lookup_socket(src); 131 if (gs == NULL) 132 return (0); 133 head = &gs->list; 134 } else 135 head = &GRE_HASH(src, dst); 136 137 CK_LIST_FOREACH(tmp, head, chain) { 138 if (tmp == sc) 139 continue; 140 if (IN6_ARE_ADDR_EQUAL(&tmp->gre_oip6.ip6_src, src) && 141 IN6_ARE_ADDR_EQUAL(&tmp->gre_oip6.ip6_dst, dst)) 142 return (EADDRNOTAVAIL); 143 } 144 return (0); 145 } 146 147 static int 148 in6_gre_lookup(const struct mbuf *m, int off, int proto, void **arg) 149 { 150 const struct ip6_hdr *ip6; 151 struct gre_softc *sc; 152 153 if (V_ipv6_hashtbl == NULL) 154 return (0); 155 156 NET_EPOCH_ASSERT(); 157 ip6 = mtod(m, const struct ip6_hdr *); 158 CK_LIST_FOREACH(sc, &GRE_HASH(&ip6->ip6_dst, &ip6->ip6_src), chain) { 159 /* 160 * This is an inbound packet, its ip6_dst is source address 161 * in softc. 162 */ 163 if (IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_src, 164 &ip6->ip6_dst) && 165 IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_dst, 166 &ip6->ip6_src)) { 167 if ((GRE2IFP(sc)->if_flags & IFF_UP) == 0) 168 return (0); 169 *arg = sc; 170 return (ENCAP_DRV_LOOKUP); 171 } 172 } 173 return (0); 174 } 175 176 /* 177 * Check that ingress address belongs to local host. 178 */ 179 static void 180 in6_gre_set_running(struct gre_softc *sc) 181 { 182 183 if (in6_localip(&sc->gre_oip6.ip6_src)) 184 GRE2IFP(sc)->if_drv_flags |= IFF_DRV_RUNNING; 185 else 186 GRE2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 187 } 188 189 /* 190 * ifaddr_event handler. 191 * Clear IFF_DRV_RUNNING flag when ingress address disappears to prevent 192 * source address spoofing. 193 */ 194 static void 195 in6_gre_srcaddr(void *arg __unused, const struct sockaddr *sa, 196 int event __unused) 197 { 198 const struct sockaddr_in6 *sin; 199 struct gre_softc *sc; 200 201 /* Check that VNET is ready */ 202 if (V_ipv6_hashtbl == NULL) 203 return; 204 205 NET_EPOCH_ASSERT(); 206 sin = (const struct sockaddr_in6 *)sa; 207 CK_LIST_FOREACH(sc, &GRE_SRCHASH(&sin->sin6_addr), srchash) { 208 if (IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_src, 209 &sin->sin6_addr) == 0) 210 continue; 211 in6_gre_set_running(sc); 212 } 213 } 214 215 static void 216 in6_gre_udp_input(struct mbuf *m, int off, struct inpcb *inp, 217 const struct sockaddr *sa, void *ctx) 218 { 219 struct epoch_tracker et; 220 struct gre_socket *gs; 221 struct gre_softc *sc; 222 struct sockaddr_in6 dst; 223 224 NET_EPOCH_ENTER(et); 225 /* 226 * udp_append() holds reference to inp, it is safe to check 227 * inp_flags2 without INP_RLOCK(). 228 * If socket was closed before we have entered NET_EPOCH section, 229 * INP_FREED flag should be set. Otherwise it should be safe to 230 * make access to ctx data, because gre_so will be freed by 231 * gre_sofree() via NET_EPOCH_CALL(). 232 */ 233 if (__predict_false(inp->inp_flags2 & INP_FREED)) { 234 NET_EPOCH_EXIT(et); 235 m_freem(m); 236 return; 237 } 238 239 gs = (struct gre_socket *)ctx; 240 dst = *(const struct sockaddr_in6 *)sa; 241 if (sa6_embedscope(&dst, 0)) { 242 NET_EPOCH_EXIT(et); 243 m_freem(m); 244 return; 245 } 246 CK_LIST_FOREACH(sc, &gs->list, chain) { 247 if (IN6_ARE_ADDR_EQUAL(&sc->gre_oip6.ip6_dst, &dst.sin6_addr)) 248 break; 249 } 250 if (sc != NULL && (GRE2IFP(sc)->if_flags & IFF_UP) != 0){ 251 gre_input(m, off + sizeof(struct udphdr), IPPROTO_UDP, sc); 252 NET_EPOCH_EXIT(et); 253 return; 254 } 255 m_freem(m); 256 NET_EPOCH_EXIT(et); 257 } 258 259 static int 260 in6_gre_setup_socket(struct gre_softc *sc) 261 { 262 struct sockopt sopt; 263 struct sockaddr_in6 sin6; 264 struct in6_gre_socket *s; 265 struct gre_socket *gs; 266 int error, value; 267 268 /* 269 * NOTE: we are protected with gre_ioctl_sx lock. 270 * 271 * First check that socket is already configured. 272 * If so, check that source addres was not changed. 273 * If address is different, check that there are no other tunnels 274 * and close socket. 275 */ 276 gs = sc->gre_so; 277 if (gs != NULL) { 278 s = __containerof(gs, struct in6_gre_socket, base); 279 if (!IN6_ARE_ADDR_EQUAL(&s->addr, &sc->gre_oip6.ip6_src)) { 280 if (CK_LIST_EMPTY(&gs->list)) { 281 CK_LIST_REMOVE(gs, chain); 282 soclose(gs->so); 283 NET_EPOCH_CALL(gre_sofree, &gs->epoch_ctx); 284 } 285 gs = sc->gre_so = NULL; 286 } 287 } 288 289 if (gs == NULL) { 290 /* 291 * Check that socket for given address is already 292 * configured. 293 */ 294 gs = in6_gre_lookup_socket(&sc->gre_oip6.ip6_src); 295 if (gs == NULL) { 296 s = malloc(sizeof(*s), M_GRE, M_WAITOK | M_ZERO); 297 s->addr = sc->gre_oip6.ip6_src; 298 gs = &s->base; 299 300 error = socreate(sc->gre_family, &gs->so, 301 SOCK_DGRAM, IPPROTO_UDP, curthread->td_ucred, 302 curthread); 303 if (error != 0) { 304 if_printf(GRE2IFP(sc), 305 "cannot create socket: %d\n", error); 306 free(s, M_GRE); 307 return (error); 308 } 309 310 error = udp_set_kernel_tunneling(gs->so, 311 in6_gre_udp_input, NULL, gs); 312 if (error != 0) { 313 if_printf(GRE2IFP(sc), 314 "cannot set UDP tunneling: %d\n", error); 315 goto fail; 316 } 317 318 memset(&sopt, 0, sizeof(sopt)); 319 sopt.sopt_dir = SOPT_SET; 320 sopt.sopt_level = IPPROTO_IPV6; 321 sopt.sopt_name = IPV6_BINDANY; 322 sopt.sopt_val = &value; 323 sopt.sopt_valsize = sizeof(value); 324 value = 1; 325 error = sosetopt(gs->so, &sopt); 326 if (error != 0) { 327 if_printf(GRE2IFP(sc), 328 "cannot set IPV6_BINDANY opt: %d\n", 329 error); 330 goto fail; 331 } 332 333 memset(&sin6, 0, sizeof(sin6)); 334 sin6.sin6_family = AF_INET6; 335 sin6.sin6_len = sizeof(sin6); 336 sin6.sin6_addr = sc->gre_oip6.ip6_src; 337 sin6.sin6_port = htons(GRE_UDPPORT); 338 error = sa6_recoverscope(&sin6); 339 if (error != 0) { 340 if_printf(GRE2IFP(sc), 341 "cannot determine scope zone id: %d\n", 342 error); 343 goto fail; 344 } 345 error = sobind(gs->so, (struct sockaddr *)&sin6, 346 curthread); 347 if (error != 0) { 348 if_printf(GRE2IFP(sc), 349 "cannot bind socket: %d\n", error); 350 goto fail; 351 } 352 /* Add socket to the chain */ 353 CK_LIST_INSERT_HEAD( 354 &GRE_SOCKHASH(&sc->gre_oip6.ip6_src), gs, chain); 355 } 356 } 357 358 /* Add softc to the socket's list */ 359 CK_LIST_INSERT_HEAD(&gs->list, sc, chain); 360 sc->gre_so = gs; 361 return (0); 362 fail: 363 soclose(gs->so); 364 free(s, M_GRE); 365 return (error); 366 } 367 368 static int 369 in6_gre_attach(struct gre_softc *sc) 370 { 371 struct grehdr *gh; 372 int error; 373 374 if (sc->gre_options & GRE_UDPENCAP) { 375 sc->gre_csumflags = CSUM_UDP_IPV6; 376 sc->gre_hlen = sizeof(struct greudp6); 377 sc->gre_oip6.ip6_nxt = IPPROTO_UDP; 378 gh = &sc->gre_udp6hdr->gi6_gre; 379 gre_update_udphdr(sc, &sc->gre_udp6, 380 in6_cksum_pseudo(&sc->gre_oip6, 0, 0, 0)); 381 } else { 382 sc->gre_hlen = sizeof(struct greip6); 383 sc->gre_oip6.ip6_nxt = IPPROTO_GRE; 384 gh = &sc->gre_ip6hdr->gi6_gre; 385 } 386 sc->gre_oip6.ip6_vfc = IPV6_VERSION; 387 gre_update_hdr(sc, gh); 388 389 /* 390 * If we return error, this means that sc is not linked, 391 * and caller should reset gre_family and free(sc->gre_hdr). 392 */ 393 if (sc->gre_options & GRE_UDPENCAP) { 394 error = in6_gre_setup_socket(sc); 395 if (error != 0) 396 return (error); 397 } else 398 CK_LIST_INSERT_HEAD(&GRE_HASH_SC(sc), sc, chain); 399 CK_LIST_INSERT_HEAD(&GRE_SRCHASH(&sc->gre_oip6.ip6_src), sc, srchash); 400 401 /* Set IFF_DRV_RUNNING if interface is ready */ 402 in6_gre_set_running(sc); 403 return (0); 404 } 405 406 int 407 in6_gre_setopts(struct gre_softc *sc, u_long cmd, uint32_t value) 408 { 409 int error; 410 411 /* NOTE: we are protected with gre_ioctl_sx lock */ 412 MPASS(cmd == GRESKEY || cmd == GRESOPTS || cmd == GRESPORT); 413 MPASS(sc->gre_family == AF_INET6); 414 415 /* 416 * If we are going to change encapsulation protocol, do check 417 * for duplicate tunnels. Return EEXIST here to do not confuse 418 * user. 419 */ 420 if (cmd == GRESOPTS && 421 (sc->gre_options & GRE_UDPENCAP) != (value & GRE_UDPENCAP) && 422 in6_gre_checkdup(sc, &sc->gre_oip6.ip6_src, 423 &sc->gre_oip6.ip6_dst, value) == EADDRNOTAVAIL) 424 return (EEXIST); 425 426 CK_LIST_REMOVE(sc, chain); 427 CK_LIST_REMOVE(sc, srchash); 428 GRE_WAIT(); 429 switch (cmd) { 430 case GRESKEY: 431 sc->gre_key = value; 432 break; 433 case GRESOPTS: 434 sc->gre_options = value; 435 break; 436 case GRESPORT: 437 sc->gre_port = value; 438 break; 439 } 440 error = in6_gre_attach(sc); 441 if (error != 0) { 442 sc->gre_family = 0; 443 free(sc->gre_hdr, M_GRE); 444 } 445 return (error); 446 } 447 448 int 449 in6_gre_ioctl(struct gre_softc *sc, u_long cmd, caddr_t data) 450 { 451 struct in6_ifreq *ifr = (struct in6_ifreq *)data; 452 struct sockaddr_in6 *dst, *src; 453 struct ip6_hdr *ip6; 454 int error; 455 456 /* NOTE: we are protected with gre_ioctl_sx lock */ 457 error = EINVAL; 458 switch (cmd) { 459 case SIOCSIFPHYADDR_IN6: 460 src = &((struct in6_aliasreq *)data)->ifra_addr; 461 dst = &((struct in6_aliasreq *)data)->ifra_dstaddr; 462 463 /* sanity checks */ 464 if (src->sin6_family != dst->sin6_family || 465 src->sin6_family != AF_INET6 || 466 src->sin6_len != dst->sin6_len || 467 src->sin6_len != sizeof(*src)) 468 break; 469 if (IN6_IS_ADDR_UNSPECIFIED(&src->sin6_addr) || 470 IN6_IS_ADDR_UNSPECIFIED(&dst->sin6_addr)) { 471 error = EADDRNOTAVAIL; 472 break; 473 } 474 /* 475 * Check validity of the scope zone ID of the 476 * addresses, and convert it into the kernel 477 * internal form if necessary. 478 */ 479 if ((error = sa6_embedscope(src, 0)) != 0 || 480 (error = sa6_embedscope(dst, 0)) != 0) 481 break; 482 483 if (V_ipv6_hashtbl == NULL) { 484 V_ipv6_hashtbl = gre_hashinit(); 485 V_ipv6_srchashtbl = gre_hashinit(); 486 V_ipv6_sockets = (struct gre_sockets *)gre_hashinit(); 487 } 488 error = in6_gre_checkdup(sc, &src->sin6_addr, 489 &dst->sin6_addr, sc->gre_options); 490 if (error == EADDRNOTAVAIL) 491 break; 492 if (error == EEXIST) { 493 /* Addresses are the same. Just return. */ 494 error = 0; 495 break; 496 } 497 ip6 = malloc(sizeof(struct greudp6) + 3 * sizeof(uint32_t), 498 M_GRE, M_WAITOK | M_ZERO); 499 ip6->ip6_src = src->sin6_addr; 500 ip6->ip6_dst = dst->sin6_addr; 501 if (sc->gre_family != 0) { 502 /* Detach existing tunnel first */ 503 CK_LIST_REMOVE(sc, chain); 504 CK_LIST_REMOVE(sc, srchash); 505 GRE_WAIT(); 506 free(sc->gre_hdr, M_GRE); 507 /* XXX: should we notify about link state change? */ 508 } 509 sc->gre_family = AF_INET6; 510 sc->gre_hdr = ip6; 511 sc->gre_oseq = 0; 512 sc->gre_iseq = UINT32_MAX; 513 error = in6_gre_attach(sc); 514 if (error != 0) { 515 sc->gre_family = 0; 516 free(sc->gre_hdr, M_GRE); 517 } 518 break; 519 case SIOCGIFPSRCADDR_IN6: 520 case SIOCGIFPDSTADDR_IN6: 521 if (sc->gre_family != AF_INET6) { 522 error = EADDRNOTAVAIL; 523 break; 524 } 525 src = (struct sockaddr_in6 *)&ifr->ifr_addr; 526 memset(src, 0, sizeof(*src)); 527 src->sin6_family = AF_INET6; 528 src->sin6_len = sizeof(*src); 529 src->sin6_addr = (cmd == SIOCGIFPSRCADDR_IN6) ? 530 sc->gre_oip6.ip6_src: sc->gre_oip6.ip6_dst; 531 error = prison_if(curthread->td_ucred, (struct sockaddr *)src); 532 if (error == 0) 533 error = sa6_recoverscope(src); 534 if (error != 0) 535 memset(src, 0, sizeof(*src)); 536 break; 537 } 538 return (error); 539 } 540 541 int 542 in6_gre_output(struct mbuf *m, int af __unused, int hlen __unused, 543 uint32_t flowid) 544 { 545 struct greip6 *gi6; 546 547 gi6 = mtod(m, struct greip6 *); 548 gi6->gi6_ip6.ip6_hlim = V_ip6_gre_hlim; 549 gi6->gi6_ip6.ip6_flow |= flowid & IPV6_FLOWLABEL_MASK; 550 return (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, NULL)); 551 } 552 553 static const struct srcaddrtab *ipv6_srcaddrtab = NULL; 554 static const struct encaptab *ecookie = NULL; 555 static const struct encap_config ipv6_encap_cfg = { 556 .proto = IPPROTO_GRE, 557 .min_length = sizeof(struct greip6) + 558 #ifdef INET 559 sizeof(struct ip), 560 #else 561 sizeof(struct ip6_hdr), 562 #endif 563 .exact_match = ENCAP_DRV_LOOKUP, 564 .lookup = in6_gre_lookup, 565 .input = gre_input 566 }; 567 568 void 569 in6_gre_init(void) 570 { 571 572 if (!IS_DEFAULT_VNET(curvnet)) 573 return; 574 ipv6_srcaddrtab = ip6_encap_register_srcaddr(in6_gre_srcaddr, 575 NULL, M_WAITOK); 576 ecookie = ip6_encap_attach(&ipv6_encap_cfg, NULL, M_WAITOK); 577 } 578 579 void 580 in6_gre_uninit(void) 581 { 582 583 if (IS_DEFAULT_VNET(curvnet)) { 584 ip6_encap_detach(ecookie); 585 ip6_encap_unregister_srcaddr(ipv6_srcaddrtab); 586 } 587 if (V_ipv6_hashtbl != NULL) { 588 gre_hashdestroy(V_ipv6_hashtbl); 589 V_ipv6_hashtbl = NULL; 590 GRE_WAIT(); 591 gre_hashdestroy(V_ipv6_srchashtbl); 592 gre_hashdestroy((struct gre_list *)V_ipv6_sockets); 593 } 594 } 595