1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 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 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $ 30 */ 31 32 /*- 33 * Copyright (c) 1982, 1986, 1988, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 4. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 61 */ 62 63 #include <sys/cdefs.h> 64 __FBSDID("$FreeBSD$"); 65 66 #include "opt_inet.h" 67 #include "opt_inet6.h" 68 #include "opt_ipsec.h" 69 #include "opt_route.h" 70 71 #include <sys/param.h> 72 #include <sys/systm.h> 73 #include <sys/malloc.h> 74 #include <sys/mbuf.h> 75 #include <sys/proc.h> 76 #include <sys/domain.h> 77 #include <sys/protosw.h> 78 #include <sys/socket.h> 79 #include <sys/socketvar.h> 80 #include <sys/errno.h> 81 #include <sys/time.h> 82 #include <sys/kernel.h> 83 #include <sys/syslog.h> 84 85 #include <net/if.h> 86 #include <net/if_types.h> 87 #include <net/if_dl.h> 88 #include <net/route.h> 89 #include <net/netisr.h> 90 #include <net/pfil.h> 91 #include <net/vnet.h> 92 93 #include <netinet/in.h> 94 #include <netinet/in_systm.h> 95 #include <net/if_llatbl.h> 96 #ifdef INET 97 #include <netinet/ip.h> 98 #include <netinet/ip_icmp.h> 99 #endif /* INET */ 100 #include <netinet/ip6.h> 101 #include <netinet6/in6_var.h> 102 #include <netinet6/ip6_var.h> 103 #include <netinet/in_pcb.h> 104 #include <netinet/icmp6.h> 105 #include <netinet6/scope6_var.h> 106 #include <netinet6/in6_ifattach.h> 107 #include <netinet6/nd6.h> 108 109 #ifdef IPSEC 110 #include <netipsec/ipsec.h> 111 #include <netinet6/ip6_ipsec.h> 112 #include <netipsec/ipsec6.h> 113 #endif /* IPSEC */ 114 115 #include <netinet6/ip6protosw.h> 116 117 #ifdef FLOWTABLE 118 #include <net/flowtable.h> 119 VNET_DECLARE(int, ip6_output_flowtable_size); 120 #define V_ip6_output_flowtable_size VNET(ip6_output_flowtable_size) 121 #endif 122 123 extern struct domain inet6domain; 124 125 u_char ip6_protox[IPPROTO_MAX]; 126 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 127 128 static struct netisr_handler ip6_nh = { 129 .nh_name = "ip6", 130 .nh_handler = ip6_input, 131 .nh_proto = NETISR_IPV6, 132 .nh_policy = NETISR_POLICY_FLOW, 133 }; 134 135 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch); 136 #define V_in6_tmpaddrtimer_ch VNET(in6_tmpaddrtimer_ch) 137 138 VNET_DEFINE(struct pfil_head, inet6_pfil_hook); 139 140 VNET_DEFINE(struct ip6stat, ip6stat); 141 142 struct rwlock in6_ifaddr_lock; 143 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 144 145 static void ip6_init2(void *); 146 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *); 147 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 148 #ifdef PULLDOWN_TEST 149 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int); 150 #endif 151 152 /* 153 * IP6 initialization: fill in IP6 protocol switch table. 154 * All protocols not implemented in kernel go to raw IP6 protocol handler. 155 */ 156 void 157 ip6_init(void) 158 { 159 struct ip6protosw *pr; 160 int i; 161 162 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 163 &V_ip6_auto_linklocal); 164 165 TAILQ_INIT(&V_in6_ifaddrhead); 166 167 /* Initialize packet filter hooks. */ 168 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF; 169 V_inet6_pfil_hook.ph_af = AF_INET6; 170 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0) 171 printf("%s: WARNING: unable to register pfil hook, " 172 "error %d\n", __func__, i); 173 174 scope6_init(); 175 addrsel_policy_init(); 176 nd6_init(); 177 frag6_init(); 178 179 #ifdef FLOWTABLE 180 if (TUNABLE_INT_FETCH("net.inet6.ip6.output_flowtable_size", 181 &V_ip6_output_flowtable_size)) { 182 if (V_ip6_output_flowtable_size < 256) 183 V_ip6_output_flowtable_size = 256; 184 if (!powerof2(V_ip6_output_flowtable_size)) { 185 printf("flowtable must be power of 2 size\n"); 186 V_ip6_output_flowtable_size = 2048; 187 } 188 } else { 189 /* 190 * round up to the next power of 2 191 */ 192 V_ip6_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1); 193 } 194 V_ip6_ft = flowtable_alloc("ipv6", V_ip6_output_flowtable_size, FL_IPV6|FL_PCPU); 195 #endif 196 197 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 198 199 /* Skip global initialization stuff for non-default instances. */ 200 if (!IS_DEFAULT_VNET(curvnet)) 201 return; 202 203 #ifdef DIAGNOSTIC 204 if (sizeof(struct protosw) != sizeof(struct ip6protosw)) 205 panic("sizeof(protosw) != sizeof(ip6protosw)"); 206 #endif 207 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 208 if (pr == NULL) 209 panic("ip6_init"); 210 211 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 212 for (i = 0; i < IPPROTO_MAX; i++) 213 ip6_protox[i] = pr - inet6sw; 214 /* 215 * Cycle through IP protocols and put them into the appropriate place 216 * in ip6_protox[]. 217 */ 218 for (pr = (struct ip6protosw *)inet6domain.dom_protosw; 219 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) 220 if (pr->pr_domain->dom_family == PF_INET6 && 221 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 222 /* Be careful to only index valid IP protocols. */ 223 if (pr->pr_protocol < IPPROTO_MAX) 224 ip6_protox[pr->pr_protocol] = pr - inet6sw; 225 } 226 227 netisr_register(&ip6_nh); 228 } 229 230 /* 231 * The protocol to be inserted into ip6_protox[] must be already registered 232 * in inet6sw[], either statically or through pf_proto_register(). 233 */ 234 int 235 ip6proto_register(short ip6proto) 236 { 237 struct ip6protosw *pr; 238 239 /* Sanity checks. */ 240 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 241 return (EPROTONOSUPPORT); 242 243 /* 244 * The protocol slot must not be occupied by another protocol 245 * already. An index pointing to IPPROTO_RAW is unused. 246 */ 247 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 248 if (pr == NULL) 249 return (EPFNOSUPPORT); 250 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 251 return (EEXIST); 252 253 /* 254 * Find the protocol position in inet6sw[] and set the index. 255 */ 256 for (pr = (struct ip6protosw *)inet6domain.dom_protosw; 257 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) { 258 if (pr->pr_domain->dom_family == PF_INET6 && 259 pr->pr_protocol && pr->pr_protocol == ip6proto) { 260 ip6_protox[pr->pr_protocol] = pr - inet6sw; 261 return (0); 262 } 263 } 264 return (EPROTONOSUPPORT); 265 } 266 267 int 268 ip6proto_unregister(short ip6proto) 269 { 270 struct ip6protosw *pr; 271 272 /* Sanity checks. */ 273 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 274 return (EPROTONOSUPPORT); 275 276 /* Check if the protocol was indeed registered. */ 277 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 278 if (pr == NULL) 279 return (EPFNOSUPPORT); 280 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 281 return (ENOENT); 282 283 /* Reset the protocol slot to IPPROTO_RAW. */ 284 ip6_protox[ip6proto] = pr - inet6sw; 285 return (0); 286 } 287 288 #ifdef VIMAGE 289 void 290 ip6_destroy() 291 { 292 293 nd6_destroy(); 294 callout_drain(&V_in6_tmpaddrtimer_ch); 295 } 296 #endif 297 298 static int 299 ip6_init2_vnet(const void *unused __unused) 300 { 301 302 /* nd6_timer_init */ 303 callout_init(&V_nd6_timer_ch, 0); 304 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); 305 306 /* timer for regeneranation of temporary addresses randomize ID */ 307 callout_init(&V_in6_tmpaddrtimer_ch, 0); 308 callout_reset(&V_in6_tmpaddrtimer_ch, 309 (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor - 310 V_ip6_temp_regen_advance) * hz, 311 in6_tmpaddrtimer, curvnet); 312 313 return (0); 314 } 315 316 static void 317 ip6_init2(void *dummy) 318 { 319 320 ip6_init2_vnet(NULL); 321 } 322 323 /* cheat */ 324 /* This must be after route_init(), which is now SI_ORDER_THIRD */ 325 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL); 326 327 void 328 ip6_input(struct mbuf *m) 329 { 330 struct ip6_hdr *ip6; 331 int off = sizeof(struct ip6_hdr), nest; 332 u_int32_t plen; 333 u_int32_t rtalert = ~0; 334 int nxt, ours = 0; 335 struct ifnet *deliverifp = NULL, *ifp = NULL; 336 struct in6_addr odst; 337 struct route_in6 rin6; 338 int srcrt = 0; 339 struct llentry *lle = NULL; 340 struct sockaddr_in6 dst6, *dst; 341 342 bzero(&rin6, sizeof(struct route_in6)); 343 #ifdef IPSEC 344 /* 345 * should the inner packet be considered authentic? 346 * see comment in ah4_input(). 347 * NB: m cannot be NULL when passed to the input routine 348 */ 349 350 m->m_flags &= ~M_AUTHIPHDR; 351 m->m_flags &= ~M_AUTHIPDGM; 352 353 #endif /* IPSEC */ 354 355 /* 356 * make sure we don't have onion peering information into m_tag. 357 */ 358 ip6_delaux(m); 359 360 /* 361 * mbuf statistics 362 */ 363 if (m->m_flags & M_EXT) { 364 if (m->m_next) 365 V_ip6stat.ip6s_mext2m++; 366 else 367 V_ip6stat.ip6s_mext1++; 368 } else { 369 #define M2MMAX (sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0])) 370 if (m->m_next) { 371 if (m->m_flags & M_LOOP) { 372 V_ip6stat.ip6s_m2m[V_loif->if_index]++; 373 } else if (m->m_pkthdr.rcvif->if_index < M2MMAX) 374 V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++; 375 else 376 V_ip6stat.ip6s_m2m[0]++; 377 } else 378 V_ip6stat.ip6s_m1++; 379 #undef M2MMAX 380 } 381 382 /* drop the packet if IPv6 operation is disabled on the IF */ 383 if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) { 384 m_freem(m); 385 return; 386 } 387 388 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive); 389 V_ip6stat.ip6s_total++; 390 391 #ifndef PULLDOWN_TEST 392 /* 393 * L2 bridge code and some other code can return mbuf chain 394 * that does not conform to KAME requirement. too bad. 395 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 396 */ 397 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 398 struct mbuf *n; 399 400 MGETHDR(n, M_DONTWAIT, MT_HEADER); 401 if (n) 402 M_MOVE_PKTHDR(n, m); 403 if (n && n->m_pkthdr.len > MHLEN) { 404 MCLGET(n, M_DONTWAIT); 405 if ((n->m_flags & M_EXT) == 0) { 406 m_freem(n); 407 n = NULL; 408 } 409 } 410 if (n == NULL) { 411 m_freem(m); 412 return; /* ENOBUFS */ 413 } 414 415 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 416 n->m_len = n->m_pkthdr.len; 417 m_freem(m); 418 m = n; 419 } 420 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */); 421 #endif 422 423 if (m->m_len < sizeof(struct ip6_hdr)) { 424 struct ifnet *inifp; 425 inifp = m->m_pkthdr.rcvif; 426 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 427 V_ip6stat.ip6s_toosmall++; 428 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 429 return; 430 } 431 } 432 433 ip6 = mtod(m, struct ip6_hdr *); 434 435 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 436 V_ip6stat.ip6s_badvers++; 437 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 438 goto bad; 439 } 440 441 V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++; 442 443 /* 444 * Check against address spoofing/corruption. 445 */ 446 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 447 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 448 /* 449 * XXX: "badscope" is not very suitable for a multicast source. 450 */ 451 V_ip6stat.ip6s_badscope++; 452 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 453 goto bad; 454 } 455 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 456 !(m->m_flags & M_LOOP)) { 457 /* 458 * In this case, the packet should come from the loopback 459 * interface. However, we cannot just check the if_flags, 460 * because ip6_mloopback() passes the "actual" interface 461 * as the outgoing/incoming interface. 462 */ 463 V_ip6stat.ip6s_badscope++; 464 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 465 goto bad; 466 } 467 468 #ifdef ALTQ 469 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 470 /* packet is dropped by traffic conditioner */ 471 return; 472 } 473 #endif 474 /* 475 * The following check is not documented in specs. A malicious 476 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 477 * and bypass security checks (act as if it was from 127.0.0.1 by using 478 * IPv6 src ::ffff:127.0.0.1). Be cautious. 479 * 480 * This check chokes if we are in an SIIT cloud. As none of BSDs 481 * support IPv4-less kernel compilation, we cannot support SIIT 482 * environment at all. So, it makes more sense for us to reject any 483 * malicious packets for non-SIIT environment, than try to do a 484 * partial support for SIIT environment. 485 */ 486 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 487 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 488 V_ip6stat.ip6s_badscope++; 489 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 490 goto bad; 491 } 492 #if 0 493 /* 494 * Reject packets with IPv4 compatible addresses (auto tunnel). 495 * 496 * The code forbids auto tunnel relay case in RFC1933 (the check is 497 * stronger than RFC1933). We may want to re-enable it if mech-xx 498 * is revised to forbid relaying case. 499 */ 500 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 501 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 502 V_ip6stat.ip6s_badscope++; 503 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 504 goto bad; 505 } 506 #endif 507 #ifdef IPSEC 508 /* 509 * Bypass packet filtering for packets previously handled by IPsec. 510 */ 511 if (ip6_ipsec_filtertunnel(m)) 512 goto passin; 513 #endif /* IPSEC */ 514 515 /* 516 * Run through list of hooks for input packets. 517 * 518 * NB: Beware of the destination address changing 519 * (e.g. by NAT rewriting). When this happens, 520 * tell ip6_forward to do the right thing. 521 */ 522 odst = ip6->ip6_dst; 523 524 /* Jump over all PFIL processing if hooks are not active. */ 525 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 526 goto passin; 527 528 if (pfil_run_hooks(&V_inet6_pfil_hook, &m, 529 m->m_pkthdr.rcvif, PFIL_IN, NULL)) 530 return; 531 if (m == NULL) /* consumed by filter */ 532 return; 533 ip6 = mtod(m, struct ip6_hdr *); 534 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 535 536 passin: 537 /* 538 * Disambiguate address scope zones (if there is ambiguity). 539 * We first make sure that the original source or destination address 540 * is not in our internal form for scoped addresses. Such addresses 541 * are not necessarily invalid spec-wise, but we cannot accept them due 542 * to the usage conflict. 543 * in6_setscope() then also checks and rejects the cases where src or 544 * dst are the loopback address and the receiving interface 545 * is not loopback. 546 */ 547 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 548 V_ip6stat.ip6s_badscope++; /* XXX */ 549 goto bad; 550 } 551 if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) || 552 in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) { 553 V_ip6stat.ip6s_badscope++; 554 goto bad; 555 } 556 557 /* 558 * Multicast check. Assume packet is for us to avoid 559 * prematurely taking locks. 560 */ 561 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 562 ours = 1; 563 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast); 564 deliverifp = m->m_pkthdr.rcvif; 565 goto hbhcheck; 566 } 567 568 /* 569 * Unicast check 570 */ 571 572 bzero(&dst6, sizeof(dst6)); 573 dst6.sin6_family = AF_INET6; 574 dst6.sin6_len = sizeof(struct sockaddr_in6); 575 dst6.sin6_addr = ip6->ip6_dst; 576 ifp = m->m_pkthdr.rcvif; 577 IF_AFDATA_LOCK(ifp); 578 lle = lla_lookup(LLTABLE6(ifp), 0, 579 (struct sockaddr *)&dst6); 580 IF_AFDATA_UNLOCK(ifp); 581 if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) { 582 struct ifaddr *ifa; 583 struct in6_ifaddr *ia6; 584 int bad; 585 586 bad = 1; 587 #define sa_equal(a1, a2) \ 588 (bcmp((a1), (a2), ((a1))->sin6_len) == 0) 589 IF_ADDR_LOCK(ifp); 590 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 591 if (ifa->ifa_addr->sa_family != dst6.sin6_family) 592 continue; 593 if (sa_equal(&dst6, ifa->ifa_addr)) 594 break; 595 } 596 KASSERT(ifa != NULL, ("%s: ifa not found for lle %p", 597 __func__, lle)); 598 #undef sa_equal 599 600 ia6 = (struct in6_ifaddr *)ifa; 601 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) { 602 /* Count the packet in the ip address stats */ 603 ia6->ia_ifa.if_ipackets++; 604 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len; 605 606 /* 607 * record address information into m_tag. 608 */ 609 (void)ip6_setdstifaddr(m, ia6); 610 611 bad = 0; 612 } else { 613 char ip6bufs[INET6_ADDRSTRLEN]; 614 char ip6bufd[INET6_ADDRSTRLEN]; 615 /* address is not ready, so discard the packet. */ 616 nd6log((LOG_INFO, 617 "ip6_input: packet to an unready address %s->%s\n", 618 ip6_sprintf(ip6bufs, &ip6->ip6_src), 619 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 620 } 621 IF_ADDR_UNLOCK(ifp); 622 LLE_RUNLOCK(lle); 623 if (bad) 624 goto bad; 625 else { 626 ours = 1; 627 deliverifp = ifp; 628 goto hbhcheck; 629 } 630 } 631 if (lle != NULL) 632 LLE_RUNLOCK(lle); 633 634 dst = &rin6.ro_dst; 635 dst->sin6_len = sizeof(struct sockaddr_in6); 636 dst->sin6_family = AF_INET6; 637 dst->sin6_addr = ip6->ip6_dst; 638 rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0); 639 if (rin6.ro_rt) 640 RT_UNLOCK(rin6.ro_rt); 641 642 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key)) 643 644 /* 645 * Accept the packet if the forwarding interface to the destination 646 * according to the routing table is the loopback interface, 647 * unless the associated route has a gateway. 648 * Note that this approach causes to accept a packet if there is a 649 * route to the loopback interface for the destination of the packet. 650 * But we think it's even useful in some situations, e.g. when using 651 * a special daemon which wants to intercept the packet. 652 * 653 * XXX: some OSes automatically make a cloned route for the destination 654 * of an outgoing packet. If the outgoing interface of the packet 655 * is a loopback one, the kernel would consider the packet to be 656 * accepted, even if we have no such address assinged on the interface. 657 * We check the cloned flag of the route entry to reject such cases, 658 * assuming that route entries for our own addresses are not made by 659 * cloning (it should be true because in6_addloop explicitly installs 660 * the host route). However, we might have to do an explicit check 661 * while it would be less efficient. Or, should we rather install a 662 * reject route for such a case? 663 */ 664 if (rin6.ro_rt && 665 (rin6.ro_rt->rt_flags & 666 (RTF_HOST|RTF_GATEWAY)) == RTF_HOST && 667 #ifdef RTF_WASCLONED 668 !(rin6.ro_rt->rt_flags & RTF_WASCLONED) && 669 #endif 670 #ifdef RTF_CLONED 671 !(rin6.ro_rt->rt_flags & RTF_CLONED) && 672 #endif 673 #if 0 674 /* 675 * The check below is redundant since the comparison of 676 * the destination and the key of the rtentry has 677 * already done through looking up the routing table. 678 */ 679 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 680 &rt6_key(rin6.ro_rt)->sin6_addr) 681 #endif 682 rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) { 683 int free_ia6 = 0; 684 struct in6_ifaddr *ia6; 685 686 /* 687 * found the loopback route to the interface address 688 */ 689 if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) { 690 struct sockaddr_in6 dest6; 691 692 bzero(&dest6, sizeof(dest6)); 693 dest6.sin6_family = AF_INET6; 694 dest6.sin6_len = sizeof(dest6); 695 dest6.sin6_addr = ip6->ip6_dst; 696 ia6 = (struct in6_ifaddr *) 697 ifa_ifwithaddr((struct sockaddr *)&dest6); 698 if (ia6 == NULL) 699 goto bad; 700 free_ia6 = 1; 701 } 702 else 703 ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa; 704 705 /* 706 * record address information into m_tag. 707 */ 708 (void)ip6_setdstifaddr(m, ia6); 709 710 /* 711 * packets to a tentative, duplicated, or somehow invalid 712 * address must not be accepted. 713 */ 714 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) { 715 /* this address is ready */ 716 ours = 1; 717 deliverifp = ia6->ia_ifp; /* correct? */ 718 /* Count the packet in the ip address stats */ 719 ia6->ia_ifa.if_ipackets++; 720 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len; 721 if (ia6 != NULL && free_ia6 != 0) 722 ifa_free(&ia6->ia_ifa); 723 goto hbhcheck; 724 } else { 725 char ip6bufs[INET6_ADDRSTRLEN]; 726 char ip6bufd[INET6_ADDRSTRLEN]; 727 /* address is not ready, so discard the packet. */ 728 nd6log((LOG_INFO, 729 "ip6_input: packet to an unready address %s->%s\n", 730 ip6_sprintf(ip6bufs, &ip6->ip6_src), 731 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 732 733 if (ia6 != NULL && free_ia6 != 0) 734 ifa_free(&ia6->ia_ifa); 735 goto bad; 736 } 737 } 738 739 /* 740 * FAITH (Firewall Aided Internet Translator) 741 */ 742 if (V_ip6_keepfaith) { 743 if (rin6.ro_rt && rin6.ro_rt->rt_ifp && 744 rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) { 745 /* XXX do we need more sanity checks? */ 746 ours = 1; 747 deliverifp = rin6.ro_rt->rt_ifp; /* faith */ 748 goto hbhcheck; 749 } 750 } 751 752 /* 753 * Now there is no reason to process the packet if it's not our own 754 * and we're not a router. 755 */ 756 if (!V_ip6_forwarding) { 757 V_ip6stat.ip6s_cantforward++; 758 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 759 goto bad; 760 } 761 762 hbhcheck: 763 /* 764 * record address information into m_tag, if we don't have one yet. 765 * note that we are unable to record it, if the address is not listed 766 * as our interface address (e.g. multicast addresses, addresses 767 * within FAITH prefixes and such). 768 */ 769 if (deliverifp && !ip6_getdstifaddr(m)) { 770 struct in6_ifaddr *ia6; 771 772 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst); 773 if (ia6) { 774 if (!ip6_setdstifaddr(m, ia6)) { 775 /* 776 * XXX maybe we should drop the packet here, 777 * as we could not provide enough information 778 * to the upper layers. 779 */ 780 } 781 ifa_free(&ia6->ia_ifa); 782 } 783 } 784 785 /* 786 * Process Hop-by-Hop options header if it's contained. 787 * m may be modified in ip6_hopopts_input(). 788 * If a JumboPayload option is included, plen will also be modified. 789 */ 790 plen = (u_int32_t)ntohs(ip6->ip6_plen); 791 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 792 struct ip6_hbh *hbh; 793 794 if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) { 795 #if 0 /*touches NULL pointer*/ 796 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 797 #endif 798 goto out; /* m have already been freed */ 799 } 800 801 /* adjust pointer */ 802 ip6 = mtod(m, struct ip6_hdr *); 803 804 /* 805 * if the payload length field is 0 and the next header field 806 * indicates Hop-by-Hop Options header, then a Jumbo Payload 807 * option MUST be included. 808 */ 809 if (ip6->ip6_plen == 0 && plen == 0) { 810 /* 811 * Note that if a valid jumbo payload option is 812 * contained, ip6_hopopts_input() must set a valid 813 * (non-zero) payload length to the variable plen. 814 */ 815 V_ip6stat.ip6s_badoptions++; 816 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 817 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 818 icmp6_error(m, ICMP6_PARAM_PROB, 819 ICMP6_PARAMPROB_HEADER, 820 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 821 goto out; 822 } 823 #ifndef PULLDOWN_TEST 824 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 825 hbh = (struct ip6_hbh *)(ip6 + 1); 826 #else 827 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 828 sizeof(struct ip6_hbh)); 829 if (hbh == NULL) { 830 V_ip6stat.ip6s_tooshort++; 831 goto out; 832 } 833 #endif 834 nxt = hbh->ip6h_nxt; 835 836 /* 837 * If we are acting as a router and the packet contains a 838 * router alert option, see if we know the option value. 839 * Currently, we only support the option value for MLD, in which 840 * case we should pass the packet to the multicast routing 841 * daemon. 842 */ 843 if (rtalert != ~0) { 844 switch (rtalert) { 845 case IP6OPT_RTALERT_MLD: 846 if (V_ip6_forwarding) 847 ours = 1; 848 break; 849 default: 850 /* 851 * RFC2711 requires unrecognized values must be 852 * silently ignored. 853 */ 854 break; 855 } 856 } 857 } else 858 nxt = ip6->ip6_nxt; 859 860 /* 861 * Check that the amount of data in the buffers 862 * is as at least much as the IPv6 header would have us expect. 863 * Trim mbufs if longer than we expect. 864 * Drop packet if shorter than we expect. 865 */ 866 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 867 V_ip6stat.ip6s_tooshort++; 868 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 869 goto bad; 870 } 871 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 872 if (m->m_len == m->m_pkthdr.len) { 873 m->m_len = sizeof(struct ip6_hdr) + plen; 874 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 875 } else 876 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 877 } 878 879 /* 880 * Forward if desirable. 881 */ 882 if (V_ip6_mrouter && 883 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 884 /* 885 * If we are acting as a multicast router, all 886 * incoming multicast packets are passed to the 887 * kernel-level multicast forwarding function. 888 * The packet is returned (relatively) intact; if 889 * ip6_mforward() returns a non-zero value, the packet 890 * must be discarded, else it may be accepted below. 891 * 892 * XXX TODO: Check hlim and multicast scope here to avoid 893 * unnecessarily calling into ip6_mforward(). 894 */ 895 if (ip6_mforward && 896 ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) { 897 IP6STAT_INC(ip6s_cantforward); 898 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 899 goto bad; 900 } 901 } else if (!ours) { 902 ip6_forward(m, srcrt); 903 goto out; 904 } 905 906 ip6 = mtod(m, struct ip6_hdr *); 907 908 /* 909 * Malicious party may be able to use IPv4 mapped addr to confuse 910 * tcp/udp stack and bypass security checks (act as if it was from 911 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 912 * 913 * For SIIT end node behavior, you may want to disable the check. 914 * However, you will become vulnerable to attacks using IPv4 mapped 915 * source. 916 */ 917 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 918 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 919 V_ip6stat.ip6s_badscope++; 920 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 921 goto bad; 922 } 923 924 /* 925 * Tell launch routine the next header 926 */ 927 V_ip6stat.ip6s_delivered++; 928 in6_ifstat_inc(deliverifp, ifs6_in_deliver); 929 nest = 0; 930 931 while (nxt != IPPROTO_DONE) { 932 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 933 V_ip6stat.ip6s_toomanyhdr++; 934 goto bad; 935 } 936 937 /* 938 * protection against faulty packet - there should be 939 * more sanity checks in header chain processing. 940 */ 941 if (m->m_pkthdr.len < off) { 942 V_ip6stat.ip6s_tooshort++; 943 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 944 goto bad; 945 } 946 947 #ifdef IPSEC 948 /* 949 * enforce IPsec policy checking if we are seeing last header. 950 * note that we do not visit this with protocols with pcb layer 951 * code - like udp/tcp/raw ip. 952 */ 953 if (ip6_ipsec_input(m, nxt)) 954 goto bad; 955 #endif /* IPSEC */ 956 957 /* 958 * Use mbuf flags to propagate Router Alert option to 959 * ICMPv6 layer, as hop-by-hop options have been stripped. 960 */ 961 if (nxt == IPPROTO_ICMPV6 && rtalert != ~0) 962 m->m_flags |= M_RTALERT_MLD; 963 964 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 965 } 966 goto out; 967 bad: 968 m_freem(m); 969 out: 970 if (rin6.ro_rt) 971 RTFREE(rin6.ro_rt); 972 } 973 974 /* 975 * set/grab in6_ifaddr correspond to IPv6 destination address. 976 * XXX backward compatibility wrapper 977 * 978 * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag, 979 * and then bump it when the tag is copied, and release it when the tag is 980 * freed. Unfortunately, m_tags don't support deep copies (yet), so instead 981 * we just bump the ia refcount when we receive it. This should be fixed. 982 */ 983 static struct ip6aux * 984 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6) 985 { 986 struct ip6aux *ip6a; 987 988 ip6a = ip6_addaux(m); 989 if (ip6a) 990 ip6a->ip6a_dstia6 = ia6; 991 return ip6a; /* NULL if failed to set */ 992 } 993 994 struct in6_ifaddr * 995 ip6_getdstifaddr(struct mbuf *m) 996 { 997 struct ip6aux *ip6a; 998 struct in6_ifaddr *ia; 999 1000 ip6a = ip6_findaux(m); 1001 if (ip6a) { 1002 ia = ip6a->ip6a_dstia6; 1003 ifa_ref(&ia->ia_ifa); 1004 return ia; 1005 } else 1006 return NULL; 1007 } 1008 1009 /* 1010 * Hop-by-Hop options header processing. If a valid jumbo payload option is 1011 * included, the real payload length will be stored in plenp. 1012 * 1013 * rtalertp - XXX: should be stored more smart way 1014 */ 1015 static int 1016 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 1017 struct mbuf **mp, int *offp) 1018 { 1019 struct mbuf *m = *mp; 1020 int off = *offp, hbhlen; 1021 struct ip6_hbh *hbh; 1022 u_int8_t *opt; 1023 1024 /* validation of the length of the header */ 1025 #ifndef PULLDOWN_TEST 1026 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); 1027 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 1028 hbhlen = (hbh->ip6h_len + 1) << 3; 1029 1030 IP6_EXTHDR_CHECK(m, off, hbhlen, -1); 1031 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 1032 #else 1033 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, 1034 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); 1035 if (hbh == NULL) { 1036 V_ip6stat.ip6s_tooshort++; 1037 return -1; 1038 } 1039 hbhlen = (hbh->ip6h_len + 1) << 3; 1040 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 1041 hbhlen); 1042 if (hbh == NULL) { 1043 V_ip6stat.ip6s_tooshort++; 1044 return -1; 1045 } 1046 #endif 1047 off += hbhlen; 1048 hbhlen -= sizeof(struct ip6_hbh); 1049 opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh); 1050 1051 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 1052 hbhlen, rtalertp, plenp) < 0) 1053 return (-1); 1054 1055 *offp = off; 1056 *mp = m; 1057 return (0); 1058 } 1059 1060 /* 1061 * Search header for all Hop-by-hop options and process each option. 1062 * This function is separate from ip6_hopopts_input() in order to 1063 * handle a case where the sending node itself process its hop-by-hop 1064 * options header. In such a case, the function is called from ip6_output(). 1065 * 1066 * The function assumes that hbh header is located right after the IPv6 header 1067 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1068 * opthead + hbhlen is located in contiguous memory region. 1069 */ 1070 int 1071 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1072 u_int32_t *rtalertp, u_int32_t *plenp) 1073 { 1074 struct ip6_hdr *ip6; 1075 int optlen = 0; 1076 u_int8_t *opt = opthead; 1077 u_int16_t rtalert_val; 1078 u_int32_t jumboplen; 1079 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1080 1081 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1082 switch (*opt) { 1083 case IP6OPT_PAD1: 1084 optlen = 1; 1085 break; 1086 case IP6OPT_PADN: 1087 if (hbhlen < IP6OPT_MINLEN) { 1088 V_ip6stat.ip6s_toosmall++; 1089 goto bad; 1090 } 1091 optlen = *(opt + 1) + 2; 1092 break; 1093 case IP6OPT_ROUTER_ALERT: 1094 /* XXX may need check for alignment */ 1095 if (hbhlen < IP6OPT_RTALERT_LEN) { 1096 V_ip6stat.ip6s_toosmall++; 1097 goto bad; 1098 } 1099 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1100 /* XXX stat */ 1101 icmp6_error(m, ICMP6_PARAM_PROB, 1102 ICMP6_PARAMPROB_HEADER, 1103 erroff + opt + 1 - opthead); 1104 return (-1); 1105 } 1106 optlen = IP6OPT_RTALERT_LEN; 1107 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1108 *rtalertp = ntohs(rtalert_val); 1109 break; 1110 case IP6OPT_JUMBO: 1111 /* XXX may need check for alignment */ 1112 if (hbhlen < IP6OPT_JUMBO_LEN) { 1113 V_ip6stat.ip6s_toosmall++; 1114 goto bad; 1115 } 1116 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1117 /* XXX stat */ 1118 icmp6_error(m, ICMP6_PARAM_PROB, 1119 ICMP6_PARAMPROB_HEADER, 1120 erroff + opt + 1 - opthead); 1121 return (-1); 1122 } 1123 optlen = IP6OPT_JUMBO_LEN; 1124 1125 /* 1126 * IPv6 packets that have non 0 payload length 1127 * must not contain a jumbo payload option. 1128 */ 1129 ip6 = mtod(m, struct ip6_hdr *); 1130 if (ip6->ip6_plen) { 1131 V_ip6stat.ip6s_badoptions++; 1132 icmp6_error(m, ICMP6_PARAM_PROB, 1133 ICMP6_PARAMPROB_HEADER, 1134 erroff + opt - opthead); 1135 return (-1); 1136 } 1137 1138 /* 1139 * We may see jumbolen in unaligned location, so 1140 * we'd need to perform bcopy(). 1141 */ 1142 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1143 jumboplen = (u_int32_t)htonl(jumboplen); 1144 1145 #if 1 1146 /* 1147 * if there are multiple jumbo payload options, 1148 * *plenp will be non-zero and the packet will be 1149 * rejected. 1150 * the behavior may need some debate in ipngwg - 1151 * multiple options does not make sense, however, 1152 * there's no explicit mention in specification. 1153 */ 1154 if (*plenp != 0) { 1155 V_ip6stat.ip6s_badoptions++; 1156 icmp6_error(m, ICMP6_PARAM_PROB, 1157 ICMP6_PARAMPROB_HEADER, 1158 erroff + opt + 2 - opthead); 1159 return (-1); 1160 } 1161 #endif 1162 1163 /* 1164 * jumbo payload length must be larger than 65535. 1165 */ 1166 if (jumboplen <= IPV6_MAXPACKET) { 1167 V_ip6stat.ip6s_badoptions++; 1168 icmp6_error(m, ICMP6_PARAM_PROB, 1169 ICMP6_PARAMPROB_HEADER, 1170 erroff + opt + 2 - opthead); 1171 return (-1); 1172 } 1173 *plenp = jumboplen; 1174 1175 break; 1176 default: /* unknown option */ 1177 if (hbhlen < IP6OPT_MINLEN) { 1178 V_ip6stat.ip6s_toosmall++; 1179 goto bad; 1180 } 1181 optlen = ip6_unknown_opt(opt, m, 1182 erroff + opt - opthead); 1183 if (optlen == -1) 1184 return (-1); 1185 optlen += 2; 1186 break; 1187 } 1188 } 1189 1190 return (0); 1191 1192 bad: 1193 m_freem(m); 1194 return (-1); 1195 } 1196 1197 /* 1198 * Unknown option processing. 1199 * The third argument `off' is the offset from the IPv6 header to the option, 1200 * which is necessary if the IPv6 header the and option header and IPv6 header 1201 * is not contiguous in order to return an ICMPv6 error. 1202 */ 1203 int 1204 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1205 { 1206 struct ip6_hdr *ip6; 1207 1208 switch (IP6OPT_TYPE(*optp)) { 1209 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1210 return ((int)*(optp + 1)); 1211 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1212 m_freem(m); 1213 return (-1); 1214 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1215 V_ip6stat.ip6s_badoptions++; 1216 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1217 return (-1); 1218 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1219 V_ip6stat.ip6s_badoptions++; 1220 ip6 = mtod(m, struct ip6_hdr *); 1221 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1222 (m->m_flags & (M_BCAST|M_MCAST))) 1223 m_freem(m); 1224 else 1225 icmp6_error(m, ICMP6_PARAM_PROB, 1226 ICMP6_PARAMPROB_OPTION, off); 1227 return (-1); 1228 } 1229 1230 m_freem(m); /* XXX: NOTREACHED */ 1231 return (-1); 1232 } 1233 1234 /* 1235 * Create the "control" list for this pcb. 1236 * These functions will not modify mbuf chain at all. 1237 * 1238 * With KAME mbuf chain restriction: 1239 * The routine will be called from upper layer handlers like tcp6_input(). 1240 * Thus the routine assumes that the caller (tcp6_input) have already 1241 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the 1242 * very first mbuf on the mbuf chain. 1243 * 1244 * ip6_savecontrol_v4 will handle those options that are possible to be 1245 * set on a v4-mapped socket. 1246 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1247 * options and handle the v6-only ones itself. 1248 */ 1249 struct mbuf ** 1250 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1251 int *v4only) 1252 { 1253 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1254 1255 #ifdef SO_TIMESTAMP 1256 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1257 struct timeval tv; 1258 1259 microtime(&tv); 1260 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1261 SCM_TIMESTAMP, SOL_SOCKET); 1262 if (*mp) 1263 mp = &(*mp)->m_next; 1264 } 1265 #endif 1266 1267 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1268 if (v4only != NULL) 1269 *v4only = 1; 1270 return (mp); 1271 } 1272 1273 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1274 /* RFC 2292 sec. 5 */ 1275 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1276 struct in6_pktinfo pi6; 1277 1278 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1279 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1280 pi6.ipi6_ifindex = 1281 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1282 1283 *mp = sbcreatecontrol((caddr_t) &pi6, 1284 sizeof(struct in6_pktinfo), 1285 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1286 if (*mp) 1287 mp = &(*mp)->m_next; 1288 } 1289 1290 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1291 int hlim = ip6->ip6_hlim & 0xff; 1292 1293 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1294 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1295 IPPROTO_IPV6); 1296 if (*mp) 1297 mp = &(*mp)->m_next; 1298 } 1299 1300 if (v4only != NULL) 1301 *v4only = 0; 1302 return (mp); 1303 } 1304 1305 void 1306 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) 1307 { 1308 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1309 int v4only = 0; 1310 1311 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); 1312 if (v4only) 1313 return; 1314 1315 if ((in6p->inp_flags & IN6P_TCLASS) != 0) { 1316 u_int32_t flowinfo; 1317 int tclass; 1318 1319 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1320 flowinfo >>= 20; 1321 1322 tclass = flowinfo & 0xff; 1323 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass), 1324 IPV6_TCLASS, IPPROTO_IPV6); 1325 if (*mp) 1326 mp = &(*mp)->m_next; 1327 } 1328 1329 /* 1330 * IPV6_HOPOPTS socket option. Recall that we required super-user 1331 * privilege for the option (see ip6_ctloutput), but it might be too 1332 * strict, since there might be some hop-by-hop options which can be 1333 * returned to normal user. 1334 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1335 */ 1336 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { 1337 /* 1338 * Check if a hop-by-hop options header is contatined in the 1339 * received packet, and if so, store the options as ancillary 1340 * data. Note that a hop-by-hop options header must be 1341 * just after the IPv6 header, which is assured through the 1342 * IPv6 input processing. 1343 */ 1344 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1345 struct ip6_hbh *hbh; 1346 int hbhlen = 0; 1347 #ifdef PULLDOWN_TEST 1348 struct mbuf *ext; 1349 #endif 1350 1351 #ifndef PULLDOWN_TEST 1352 hbh = (struct ip6_hbh *)(ip6 + 1); 1353 hbhlen = (hbh->ip6h_len + 1) << 3; 1354 #else 1355 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), 1356 ip6->ip6_nxt); 1357 if (ext == NULL) { 1358 V_ip6stat.ip6s_tooshort++; 1359 return; 1360 } 1361 hbh = mtod(ext, struct ip6_hbh *); 1362 hbhlen = (hbh->ip6h_len + 1) << 3; 1363 if (hbhlen != ext->m_len) { 1364 m_freem(ext); 1365 V_ip6stat.ip6s_tooshort++; 1366 return; 1367 } 1368 #endif 1369 1370 /* 1371 * XXX: We copy the whole header even if a 1372 * jumbo payload option is included, the option which 1373 * is to be removed before returning according to 1374 * RFC2292. 1375 * Note: this constraint is removed in RFC3542 1376 */ 1377 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1378 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1379 IPPROTO_IPV6); 1380 if (*mp) 1381 mp = &(*mp)->m_next; 1382 #ifdef PULLDOWN_TEST 1383 m_freem(ext); 1384 #endif 1385 } 1386 } 1387 1388 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1389 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1390 1391 /* 1392 * Search for destination options headers or routing 1393 * header(s) through the header chain, and stores each 1394 * header as ancillary data. 1395 * Note that the order of the headers remains in 1396 * the chain of ancillary data. 1397 */ 1398 while (1) { /* is explicit loop prevention necessary? */ 1399 struct ip6_ext *ip6e = NULL; 1400 int elen; 1401 #ifdef PULLDOWN_TEST 1402 struct mbuf *ext = NULL; 1403 #endif 1404 1405 /* 1406 * if it is not an extension header, don't try to 1407 * pull it from the chain. 1408 */ 1409 switch (nxt) { 1410 case IPPROTO_DSTOPTS: 1411 case IPPROTO_ROUTING: 1412 case IPPROTO_HOPOPTS: 1413 case IPPROTO_AH: /* is it possible? */ 1414 break; 1415 default: 1416 goto loopend; 1417 } 1418 1419 #ifndef PULLDOWN_TEST 1420 if (off + sizeof(*ip6e) > m->m_len) 1421 goto loopend; 1422 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1423 if (nxt == IPPROTO_AH) 1424 elen = (ip6e->ip6e_len + 2) << 2; 1425 else 1426 elen = (ip6e->ip6e_len + 1) << 3; 1427 if (off + elen > m->m_len) 1428 goto loopend; 1429 #else 1430 ext = ip6_pullexthdr(m, off, nxt); 1431 if (ext == NULL) { 1432 V_ip6stat.ip6s_tooshort++; 1433 return; 1434 } 1435 ip6e = mtod(ext, struct ip6_ext *); 1436 if (nxt == IPPROTO_AH) 1437 elen = (ip6e->ip6e_len + 2) << 2; 1438 else 1439 elen = (ip6e->ip6e_len + 1) << 3; 1440 if (elen != ext->m_len) { 1441 m_freem(ext); 1442 V_ip6stat.ip6s_tooshort++; 1443 return; 1444 } 1445 #endif 1446 1447 switch (nxt) { 1448 case IPPROTO_DSTOPTS: 1449 if (!(in6p->inp_flags & IN6P_DSTOPTS)) 1450 break; 1451 1452 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1453 IS2292(in6p, 1454 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1455 IPPROTO_IPV6); 1456 if (*mp) 1457 mp = &(*mp)->m_next; 1458 break; 1459 case IPPROTO_ROUTING: 1460 if (!(in6p->inp_flags & IN6P_RTHDR)) 1461 break; 1462 1463 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1464 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), 1465 IPPROTO_IPV6); 1466 if (*mp) 1467 mp = &(*mp)->m_next; 1468 break; 1469 case IPPROTO_HOPOPTS: 1470 case IPPROTO_AH: /* is it possible? */ 1471 break; 1472 1473 default: 1474 /* 1475 * other cases have been filtered in the above. 1476 * none will visit this case. here we supply 1477 * the code just in case (nxt overwritten or 1478 * other cases). 1479 */ 1480 #ifdef PULLDOWN_TEST 1481 m_freem(ext); 1482 #endif 1483 goto loopend; 1484 1485 } 1486 1487 /* proceed with the next header. */ 1488 off += elen; 1489 nxt = ip6e->ip6e_nxt; 1490 ip6e = NULL; 1491 #ifdef PULLDOWN_TEST 1492 m_freem(ext); 1493 ext = NULL; 1494 #endif 1495 } 1496 loopend: 1497 ; 1498 } 1499 } 1500 #undef IS2292 1501 1502 void 1503 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu) 1504 { 1505 struct socket *so; 1506 struct mbuf *m_mtu; 1507 struct ip6_mtuinfo mtuctl; 1508 1509 so = in6p->inp_socket; 1510 1511 if (mtu == NULL) 1512 return; 1513 1514 #ifdef DIAGNOSTIC 1515 if (so == NULL) /* I believe this is impossible */ 1516 panic("ip6_notify_pmtu: socket is NULL"); 1517 #endif 1518 1519 bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */ 1520 mtuctl.ip6m_mtu = *mtu; 1521 mtuctl.ip6m_addr = *dst; 1522 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1523 return; 1524 1525 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1526 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1527 return; 1528 1529 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1530 == 0) { 1531 m_freem(m_mtu); 1532 /* XXX: should count statistics */ 1533 } else 1534 sorwakeup(so); 1535 1536 return; 1537 } 1538 1539 #ifdef PULLDOWN_TEST 1540 /* 1541 * pull single extension header from mbuf chain. returns single mbuf that 1542 * contains the result, or NULL on error. 1543 */ 1544 static struct mbuf * 1545 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) 1546 { 1547 struct ip6_ext ip6e; 1548 size_t elen; 1549 struct mbuf *n; 1550 1551 #ifdef DIAGNOSTIC 1552 switch (nxt) { 1553 case IPPROTO_DSTOPTS: 1554 case IPPROTO_ROUTING: 1555 case IPPROTO_HOPOPTS: 1556 case IPPROTO_AH: /* is it possible? */ 1557 break; 1558 default: 1559 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); 1560 } 1561 #endif 1562 1563 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1564 if (nxt == IPPROTO_AH) 1565 elen = (ip6e.ip6e_len + 2) << 2; 1566 else 1567 elen = (ip6e.ip6e_len + 1) << 3; 1568 1569 MGET(n, M_DONTWAIT, MT_DATA); 1570 if (n && elen >= MLEN) { 1571 MCLGET(n, M_DONTWAIT); 1572 if ((n->m_flags & M_EXT) == 0) { 1573 m_free(n); 1574 n = NULL; 1575 } 1576 } 1577 if (!n) 1578 return NULL; 1579 1580 n->m_len = 0; 1581 if (elen >= M_TRAILINGSPACE(n)) { 1582 m_free(n); 1583 return NULL; 1584 } 1585 1586 m_copydata(m, off, elen, mtod(n, caddr_t)); 1587 n->m_len = elen; 1588 return n; 1589 } 1590 #endif 1591 1592 /* 1593 * Get pointer to the previous header followed by the header 1594 * currently processed. 1595 * XXX: This function supposes that 1596 * M includes all headers, 1597 * the next header field and the header length field of each header 1598 * are valid, and 1599 * the sum of each header length equals to OFF. 1600 * Because of these assumptions, this function must be called very 1601 * carefully. Moreover, it will not be used in the near future when 1602 * we develop `neater' mechanism to process extension headers. 1603 */ 1604 char * 1605 ip6_get_prevhdr(struct mbuf *m, int off) 1606 { 1607 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1608 1609 if (off == sizeof(struct ip6_hdr)) 1610 return (&ip6->ip6_nxt); 1611 else { 1612 int len, nxt; 1613 struct ip6_ext *ip6e = NULL; 1614 1615 nxt = ip6->ip6_nxt; 1616 len = sizeof(struct ip6_hdr); 1617 while (len < off) { 1618 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len); 1619 1620 switch (nxt) { 1621 case IPPROTO_FRAGMENT: 1622 len += sizeof(struct ip6_frag); 1623 break; 1624 case IPPROTO_AH: 1625 len += (ip6e->ip6e_len + 2) << 2; 1626 break; 1627 default: 1628 len += (ip6e->ip6e_len + 1) << 3; 1629 break; 1630 } 1631 nxt = ip6e->ip6e_nxt; 1632 } 1633 if (ip6e) 1634 return (&ip6e->ip6e_nxt); 1635 else 1636 return NULL; 1637 } 1638 } 1639 1640 /* 1641 * get next header offset. m will be retained. 1642 */ 1643 int 1644 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp) 1645 { 1646 struct ip6_hdr ip6; 1647 struct ip6_ext ip6e; 1648 struct ip6_frag fh; 1649 1650 /* just in case */ 1651 if (m == NULL) 1652 panic("ip6_nexthdr: m == NULL"); 1653 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1654 return -1; 1655 1656 switch (proto) { 1657 case IPPROTO_IPV6: 1658 if (m->m_pkthdr.len < off + sizeof(ip6)) 1659 return -1; 1660 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1661 if (nxtp) 1662 *nxtp = ip6.ip6_nxt; 1663 off += sizeof(ip6); 1664 return off; 1665 1666 case IPPROTO_FRAGMENT: 1667 /* 1668 * terminate parsing if it is not the first fragment, 1669 * it does not make sense to parse through it. 1670 */ 1671 if (m->m_pkthdr.len < off + sizeof(fh)) 1672 return -1; 1673 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1674 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1675 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1676 return -1; 1677 if (nxtp) 1678 *nxtp = fh.ip6f_nxt; 1679 off += sizeof(struct ip6_frag); 1680 return off; 1681 1682 case IPPROTO_AH: 1683 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1684 return -1; 1685 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1686 if (nxtp) 1687 *nxtp = ip6e.ip6e_nxt; 1688 off += (ip6e.ip6e_len + 2) << 2; 1689 return off; 1690 1691 case IPPROTO_HOPOPTS: 1692 case IPPROTO_ROUTING: 1693 case IPPROTO_DSTOPTS: 1694 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1695 return -1; 1696 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1697 if (nxtp) 1698 *nxtp = ip6e.ip6e_nxt; 1699 off += (ip6e.ip6e_len + 1) << 3; 1700 return off; 1701 1702 case IPPROTO_NONE: 1703 case IPPROTO_ESP: 1704 case IPPROTO_IPCOMP: 1705 /* give up */ 1706 return -1; 1707 1708 default: 1709 return -1; 1710 } 1711 1712 return -1; 1713 } 1714 1715 /* 1716 * get offset for the last header in the chain. m will be kept untainted. 1717 */ 1718 int 1719 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp) 1720 { 1721 int newoff; 1722 int nxt; 1723 1724 if (!nxtp) { 1725 nxt = -1; 1726 nxtp = &nxt; 1727 } 1728 while (1) { 1729 newoff = ip6_nexthdr(m, off, proto, nxtp); 1730 if (newoff < 0) 1731 return off; 1732 else if (newoff < off) 1733 return -1; /* invalid */ 1734 else if (newoff == off) 1735 return newoff; 1736 1737 off = newoff; 1738 proto = *nxtp; 1739 } 1740 } 1741 1742 struct ip6aux * 1743 ip6_addaux(struct mbuf *m) 1744 { 1745 struct m_tag *mtag; 1746 1747 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1748 if (!mtag) { 1749 mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux), 1750 M_NOWAIT); 1751 if (mtag) { 1752 m_tag_prepend(m, mtag); 1753 bzero(mtag + 1, sizeof(struct ip6aux)); 1754 } 1755 } 1756 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1757 } 1758 1759 struct ip6aux * 1760 ip6_findaux(struct mbuf *m) 1761 { 1762 struct m_tag *mtag; 1763 1764 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1765 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1766 } 1767 1768 void 1769 ip6_delaux(struct mbuf *m) 1770 { 1771 struct m_tag *mtag; 1772 1773 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1774 if (mtag) 1775 m_tag_delete(m, mtag); 1776 } 1777 1778 /* 1779 * System control for IP6 1780 */ 1781 1782 u_char inet6ctlerrmap[PRC_NCMDS] = { 1783 0, 0, 0, 0, 1784 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1785 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1786 EMSGSIZE, EHOSTUNREACH, 0, 0, 1787 0, 0, 0, 0, 1788 ENOPROTOOPT 1789 }; 1790