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 507 508 /* 509 * Run through list of hooks for input packets. 510 * 511 * NB: Beware of the destination address changing 512 * (e.g. by NAT rewriting). When this happens, 513 * tell ip6_forward to do the right thing. 514 */ 515 odst = ip6->ip6_dst; 516 517 /* Jump over all PFIL processing if hooks are not active. */ 518 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 519 goto passin; 520 521 if (pfil_run_hooks(&V_inet6_pfil_hook, &m, 522 m->m_pkthdr.rcvif, PFIL_IN, NULL)) 523 return; 524 if (m == NULL) /* consumed by filter */ 525 return; 526 ip6 = mtod(m, struct ip6_hdr *); 527 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 528 529 passin: 530 /* 531 * Disambiguate address scope zones (if there is ambiguity). 532 * We first make sure that the original source or destination address 533 * is not in our internal form for scoped addresses. Such addresses 534 * are not necessarily invalid spec-wise, but we cannot accept them due 535 * to the usage conflict. 536 * in6_setscope() then also checks and rejects the cases where src or 537 * dst are the loopback address and the receiving interface 538 * is not loopback. 539 */ 540 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 541 V_ip6stat.ip6s_badscope++; /* XXX */ 542 goto bad; 543 } 544 if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) || 545 in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) { 546 V_ip6stat.ip6s_badscope++; 547 goto bad; 548 } 549 550 /* 551 * Multicast check. Assume packet is for us to avoid 552 * prematurely taking locks. 553 */ 554 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 555 ours = 1; 556 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast); 557 deliverifp = m->m_pkthdr.rcvif; 558 goto hbhcheck; 559 } 560 561 /* 562 * Unicast check 563 */ 564 565 bzero(&dst6, sizeof(dst6)); 566 dst6.sin6_family = AF_INET6; 567 dst6.sin6_len = sizeof(struct sockaddr_in6); 568 dst6.sin6_addr = ip6->ip6_dst; 569 ifp = m->m_pkthdr.rcvif; 570 IF_AFDATA_LOCK(ifp); 571 lle = lla_lookup(LLTABLE6(ifp), 0, 572 (struct sockaddr *)&dst6); 573 IF_AFDATA_UNLOCK(ifp); 574 if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) { 575 struct ifaddr *ifa; 576 struct in6_ifaddr *ia6; 577 int bad; 578 579 bad = 1; 580 #define sa_equal(a1, a2) \ 581 (bcmp((a1), (a2), ((a1))->sin6_len) == 0) 582 IF_ADDR_LOCK(ifp); 583 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 584 if (ifa->ifa_addr->sa_family != dst6.sin6_family) 585 continue; 586 if (sa_equal(&dst6, ifa->ifa_addr)) 587 break; 588 } 589 KASSERT(ifa != NULL, ("%s: ifa not found for lle %p", 590 __func__, lle)); 591 #undef sa_equal 592 593 ia6 = (struct in6_ifaddr *)ifa; 594 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) { 595 /* Count the packet in the ip address stats */ 596 ia6->ia_ifa.if_ipackets++; 597 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len; 598 599 /* 600 * record address information into m_tag. 601 */ 602 (void)ip6_setdstifaddr(m, ia6); 603 604 bad = 0; 605 } else { 606 char ip6bufs[INET6_ADDRSTRLEN]; 607 char ip6bufd[INET6_ADDRSTRLEN]; 608 /* address is not ready, so discard the packet. */ 609 nd6log((LOG_INFO, 610 "ip6_input: packet to an unready address %s->%s\n", 611 ip6_sprintf(ip6bufs, &ip6->ip6_src), 612 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 613 } 614 IF_ADDR_UNLOCK(ifp); 615 LLE_RUNLOCK(lle); 616 if (bad) 617 goto bad; 618 else { 619 ours = 1; 620 deliverifp = ifp; 621 goto hbhcheck; 622 } 623 } 624 if (lle != NULL) 625 LLE_RUNLOCK(lle); 626 627 dst = &rin6.ro_dst; 628 dst->sin6_len = sizeof(struct sockaddr_in6); 629 dst->sin6_family = AF_INET6; 630 dst->sin6_addr = ip6->ip6_dst; 631 rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0); 632 if (rin6.ro_rt) 633 RT_UNLOCK(rin6.ro_rt); 634 635 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key)) 636 637 /* 638 * Accept the packet if the forwarding interface to the destination 639 * according to the routing table is the loopback interface, 640 * unless the associated route has a gateway. 641 * Note that this approach causes to accept a packet if there is a 642 * route to the loopback interface for the destination of the packet. 643 * But we think it's even useful in some situations, e.g. when using 644 * a special daemon which wants to intercept the packet. 645 * 646 * XXX: some OSes automatically make a cloned route for the destination 647 * of an outgoing packet. If the outgoing interface of the packet 648 * is a loopback one, the kernel would consider the packet to be 649 * accepted, even if we have no such address assinged on the interface. 650 * We check the cloned flag of the route entry to reject such cases, 651 * assuming that route entries for our own addresses are not made by 652 * cloning (it should be true because in6_addloop explicitly installs 653 * the host route). However, we might have to do an explicit check 654 * while it would be less efficient. Or, should we rather install a 655 * reject route for such a case? 656 */ 657 if (rin6.ro_rt && 658 (rin6.ro_rt->rt_flags & 659 (RTF_HOST|RTF_GATEWAY)) == RTF_HOST && 660 #ifdef RTF_WASCLONED 661 !(rin6.ro_rt->rt_flags & RTF_WASCLONED) && 662 #endif 663 #ifdef RTF_CLONED 664 !(rin6.ro_rt->rt_flags & RTF_CLONED) && 665 #endif 666 #if 0 667 /* 668 * The check below is redundant since the comparison of 669 * the destination and the key of the rtentry has 670 * already done through looking up the routing table. 671 */ 672 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 673 &rt6_key(rin6.ro_rt)->sin6_addr) 674 #endif 675 rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) { 676 int free_ia6 = 0; 677 struct in6_ifaddr *ia6; 678 679 /* 680 * found the loopback route to the interface address 681 */ 682 if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) { 683 struct sockaddr_in6 dest6; 684 685 bzero(&dest6, sizeof(dest6)); 686 dest6.sin6_family = AF_INET6; 687 dest6.sin6_len = sizeof(dest6); 688 dest6.sin6_addr = ip6->ip6_dst; 689 ia6 = (struct in6_ifaddr *) 690 ifa_ifwithaddr((struct sockaddr *)&dest6); 691 if (ia6 == NULL) 692 goto bad; 693 free_ia6 = 1; 694 } 695 else 696 ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa; 697 698 /* 699 * record address information into m_tag. 700 */ 701 (void)ip6_setdstifaddr(m, ia6); 702 703 /* 704 * packets to a tentative, duplicated, or somehow invalid 705 * address must not be accepted. 706 */ 707 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) { 708 /* this address is ready */ 709 ours = 1; 710 deliverifp = ia6->ia_ifp; /* correct? */ 711 /* Count the packet in the ip address stats */ 712 ia6->ia_ifa.if_ipackets++; 713 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len; 714 if (ia6 != NULL && free_ia6 != 0) 715 ifa_free(&ia6->ia_ifa); 716 goto hbhcheck; 717 } else { 718 char ip6bufs[INET6_ADDRSTRLEN]; 719 char ip6bufd[INET6_ADDRSTRLEN]; 720 /* address is not ready, so discard the packet. */ 721 nd6log((LOG_INFO, 722 "ip6_input: packet to an unready address %s->%s\n", 723 ip6_sprintf(ip6bufs, &ip6->ip6_src), 724 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 725 726 if (ia6 != NULL && free_ia6 != 0) 727 ifa_free(&ia6->ia_ifa); 728 goto bad; 729 } 730 } 731 732 /* 733 * FAITH (Firewall Aided Internet Translator) 734 */ 735 if (V_ip6_keepfaith) { 736 if (rin6.ro_rt && rin6.ro_rt->rt_ifp && 737 rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) { 738 /* XXX do we need more sanity checks? */ 739 ours = 1; 740 deliverifp = rin6.ro_rt->rt_ifp; /* faith */ 741 goto hbhcheck; 742 } 743 } 744 745 /* 746 * Now there is no reason to process the packet if it's not our own 747 * and we're not a router. 748 */ 749 if (!V_ip6_forwarding) { 750 V_ip6stat.ip6s_cantforward++; 751 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 752 goto bad; 753 } 754 755 hbhcheck: 756 /* 757 * record address information into m_tag, if we don't have one yet. 758 * note that we are unable to record it, if the address is not listed 759 * as our interface address (e.g. multicast addresses, addresses 760 * within FAITH prefixes and such). 761 */ 762 if (deliverifp && !ip6_getdstifaddr(m)) { 763 struct in6_ifaddr *ia6; 764 765 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst); 766 if (ia6) { 767 if (!ip6_setdstifaddr(m, ia6)) { 768 /* 769 * XXX maybe we should drop the packet here, 770 * as we could not provide enough information 771 * to the upper layers. 772 */ 773 } 774 ifa_free(&ia6->ia_ifa); 775 } 776 } 777 778 /* 779 * Process Hop-by-Hop options header if it's contained. 780 * m may be modified in ip6_hopopts_input(). 781 * If a JumboPayload option is included, plen will also be modified. 782 */ 783 plen = (u_int32_t)ntohs(ip6->ip6_plen); 784 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 785 struct ip6_hbh *hbh; 786 787 if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) { 788 #if 0 /*touches NULL pointer*/ 789 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 790 #endif 791 goto out; /* m have already been freed */ 792 } 793 794 /* adjust pointer */ 795 ip6 = mtod(m, struct ip6_hdr *); 796 797 /* 798 * if the payload length field is 0 and the next header field 799 * indicates Hop-by-Hop Options header, then a Jumbo Payload 800 * option MUST be included. 801 */ 802 if (ip6->ip6_plen == 0 && plen == 0) { 803 /* 804 * Note that if a valid jumbo payload option is 805 * contained, ip6_hopopts_input() must set a valid 806 * (non-zero) payload length to the variable plen. 807 */ 808 V_ip6stat.ip6s_badoptions++; 809 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 810 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 811 icmp6_error(m, ICMP6_PARAM_PROB, 812 ICMP6_PARAMPROB_HEADER, 813 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 814 goto out; 815 } 816 #ifndef PULLDOWN_TEST 817 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 818 hbh = (struct ip6_hbh *)(ip6 + 1); 819 #else 820 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 821 sizeof(struct ip6_hbh)); 822 if (hbh == NULL) { 823 V_ip6stat.ip6s_tooshort++; 824 goto out; 825 } 826 #endif 827 nxt = hbh->ip6h_nxt; 828 829 /* 830 * If we are acting as a router and the packet contains a 831 * router alert option, see if we know the option value. 832 * Currently, we only support the option value for MLD, in which 833 * case we should pass the packet to the multicast routing 834 * daemon. 835 */ 836 if (rtalert != ~0) { 837 switch (rtalert) { 838 case IP6OPT_RTALERT_MLD: 839 if (V_ip6_forwarding) 840 ours = 1; 841 break; 842 default: 843 /* 844 * RFC2711 requires unrecognized values must be 845 * silently ignored. 846 */ 847 break; 848 } 849 } 850 } else 851 nxt = ip6->ip6_nxt; 852 853 /* 854 * Check that the amount of data in the buffers 855 * is as at least much as the IPv6 header would have us expect. 856 * Trim mbufs if longer than we expect. 857 * Drop packet if shorter than we expect. 858 */ 859 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 860 V_ip6stat.ip6s_tooshort++; 861 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 862 goto bad; 863 } 864 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 865 if (m->m_len == m->m_pkthdr.len) { 866 m->m_len = sizeof(struct ip6_hdr) + plen; 867 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 868 } else 869 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 870 } 871 872 /* 873 * Forward if desirable. 874 */ 875 if (V_ip6_mrouter && 876 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 877 /* 878 * If we are acting as a multicast router, all 879 * incoming multicast packets are passed to the 880 * kernel-level multicast forwarding function. 881 * The packet is returned (relatively) intact; if 882 * ip6_mforward() returns a non-zero value, the packet 883 * must be discarded, else it may be accepted below. 884 * 885 * XXX TODO: Check hlim and multicast scope here to avoid 886 * unnecessarily calling into ip6_mforward(). 887 */ 888 if (ip6_mforward && 889 ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) { 890 IP6STAT_INC(ip6s_cantforward); 891 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 892 goto bad; 893 } 894 } else if (!ours) { 895 ip6_forward(m, srcrt); 896 goto out; 897 } 898 899 ip6 = mtod(m, struct ip6_hdr *); 900 901 /* 902 * Malicious party may be able to use IPv4 mapped addr to confuse 903 * tcp/udp stack and bypass security checks (act as if it was from 904 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 905 * 906 * For SIIT end node behavior, you may want to disable the check. 907 * However, you will become vulnerable to attacks using IPv4 mapped 908 * source. 909 */ 910 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 911 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 912 V_ip6stat.ip6s_badscope++; 913 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 914 goto bad; 915 } 916 917 /* 918 * Tell launch routine the next header 919 */ 920 V_ip6stat.ip6s_delivered++; 921 in6_ifstat_inc(deliverifp, ifs6_in_deliver); 922 nest = 0; 923 924 while (nxt != IPPROTO_DONE) { 925 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 926 V_ip6stat.ip6s_toomanyhdr++; 927 goto bad; 928 } 929 930 /* 931 * protection against faulty packet - there should be 932 * more sanity checks in header chain processing. 933 */ 934 if (m->m_pkthdr.len < off) { 935 V_ip6stat.ip6s_tooshort++; 936 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 937 goto bad; 938 } 939 940 #ifdef IPSEC 941 /* 942 * enforce IPsec policy checking if we are seeing last header. 943 * note that we do not visit this with protocols with pcb layer 944 * code - like udp/tcp/raw ip. 945 */ 946 if (ip6_ipsec_input(m, nxt)) 947 goto bad; 948 #endif /* IPSEC */ 949 950 /* 951 * Use mbuf flags to propagate Router Alert option to 952 * ICMPv6 layer, as hop-by-hop options have been stripped. 953 */ 954 if (nxt == IPPROTO_ICMPV6 && rtalert != ~0) 955 m->m_flags |= M_RTALERT_MLD; 956 957 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 958 } 959 goto out; 960 bad: 961 m_freem(m); 962 out: 963 if (rin6.ro_rt) 964 RTFREE(rin6.ro_rt); 965 } 966 967 /* 968 * set/grab in6_ifaddr correspond to IPv6 destination address. 969 * XXX backward compatibility wrapper 970 * 971 * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag, 972 * and then bump it when the tag is copied, and release it when the tag is 973 * freed. Unfortunately, m_tags don't support deep copies (yet), so instead 974 * we just bump the ia refcount when we receive it. This should be fixed. 975 */ 976 static struct ip6aux * 977 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6) 978 { 979 struct ip6aux *ip6a; 980 981 ip6a = ip6_addaux(m); 982 if (ip6a) 983 ip6a->ip6a_dstia6 = ia6; 984 return ip6a; /* NULL if failed to set */ 985 } 986 987 struct in6_ifaddr * 988 ip6_getdstifaddr(struct mbuf *m) 989 { 990 struct ip6aux *ip6a; 991 struct in6_ifaddr *ia; 992 993 ip6a = ip6_findaux(m); 994 if (ip6a) { 995 ia = ip6a->ip6a_dstia6; 996 ifa_ref(&ia->ia_ifa); 997 return ia; 998 } else 999 return NULL; 1000 } 1001 1002 /* 1003 * Hop-by-Hop options header processing. If a valid jumbo payload option is 1004 * included, the real payload length will be stored in plenp. 1005 * 1006 * rtalertp - XXX: should be stored more smart way 1007 */ 1008 static int 1009 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 1010 struct mbuf **mp, int *offp) 1011 { 1012 struct mbuf *m = *mp; 1013 int off = *offp, hbhlen; 1014 struct ip6_hbh *hbh; 1015 u_int8_t *opt; 1016 1017 /* validation of the length of the header */ 1018 #ifndef PULLDOWN_TEST 1019 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); 1020 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 1021 hbhlen = (hbh->ip6h_len + 1) << 3; 1022 1023 IP6_EXTHDR_CHECK(m, off, hbhlen, -1); 1024 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 1025 #else 1026 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, 1027 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); 1028 if (hbh == NULL) { 1029 V_ip6stat.ip6s_tooshort++; 1030 return -1; 1031 } 1032 hbhlen = (hbh->ip6h_len + 1) << 3; 1033 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 1034 hbhlen); 1035 if (hbh == NULL) { 1036 V_ip6stat.ip6s_tooshort++; 1037 return -1; 1038 } 1039 #endif 1040 off += hbhlen; 1041 hbhlen -= sizeof(struct ip6_hbh); 1042 opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh); 1043 1044 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 1045 hbhlen, rtalertp, plenp) < 0) 1046 return (-1); 1047 1048 *offp = off; 1049 *mp = m; 1050 return (0); 1051 } 1052 1053 /* 1054 * Search header for all Hop-by-hop options and process each option. 1055 * This function is separate from ip6_hopopts_input() in order to 1056 * handle a case where the sending node itself process its hop-by-hop 1057 * options header. In such a case, the function is called from ip6_output(). 1058 * 1059 * The function assumes that hbh header is located right after the IPv6 header 1060 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1061 * opthead + hbhlen is located in contiguous memory region. 1062 */ 1063 int 1064 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1065 u_int32_t *rtalertp, u_int32_t *plenp) 1066 { 1067 struct ip6_hdr *ip6; 1068 int optlen = 0; 1069 u_int8_t *opt = opthead; 1070 u_int16_t rtalert_val; 1071 u_int32_t jumboplen; 1072 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1073 1074 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1075 switch (*opt) { 1076 case IP6OPT_PAD1: 1077 optlen = 1; 1078 break; 1079 case IP6OPT_PADN: 1080 if (hbhlen < IP6OPT_MINLEN) { 1081 V_ip6stat.ip6s_toosmall++; 1082 goto bad; 1083 } 1084 optlen = *(opt + 1) + 2; 1085 break; 1086 case IP6OPT_ROUTER_ALERT: 1087 /* XXX may need check for alignment */ 1088 if (hbhlen < IP6OPT_RTALERT_LEN) { 1089 V_ip6stat.ip6s_toosmall++; 1090 goto bad; 1091 } 1092 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1093 /* XXX stat */ 1094 icmp6_error(m, ICMP6_PARAM_PROB, 1095 ICMP6_PARAMPROB_HEADER, 1096 erroff + opt + 1 - opthead); 1097 return (-1); 1098 } 1099 optlen = IP6OPT_RTALERT_LEN; 1100 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1101 *rtalertp = ntohs(rtalert_val); 1102 break; 1103 case IP6OPT_JUMBO: 1104 /* XXX may need check for alignment */ 1105 if (hbhlen < IP6OPT_JUMBO_LEN) { 1106 V_ip6stat.ip6s_toosmall++; 1107 goto bad; 1108 } 1109 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1110 /* XXX stat */ 1111 icmp6_error(m, ICMP6_PARAM_PROB, 1112 ICMP6_PARAMPROB_HEADER, 1113 erroff + opt + 1 - opthead); 1114 return (-1); 1115 } 1116 optlen = IP6OPT_JUMBO_LEN; 1117 1118 /* 1119 * IPv6 packets that have non 0 payload length 1120 * must not contain a jumbo payload option. 1121 */ 1122 ip6 = mtod(m, struct ip6_hdr *); 1123 if (ip6->ip6_plen) { 1124 V_ip6stat.ip6s_badoptions++; 1125 icmp6_error(m, ICMP6_PARAM_PROB, 1126 ICMP6_PARAMPROB_HEADER, 1127 erroff + opt - opthead); 1128 return (-1); 1129 } 1130 1131 /* 1132 * We may see jumbolen in unaligned location, so 1133 * we'd need to perform bcopy(). 1134 */ 1135 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1136 jumboplen = (u_int32_t)htonl(jumboplen); 1137 1138 #if 1 1139 /* 1140 * if there are multiple jumbo payload options, 1141 * *plenp will be non-zero and the packet will be 1142 * rejected. 1143 * the behavior may need some debate in ipngwg - 1144 * multiple options does not make sense, however, 1145 * there's no explicit mention in specification. 1146 */ 1147 if (*plenp != 0) { 1148 V_ip6stat.ip6s_badoptions++; 1149 icmp6_error(m, ICMP6_PARAM_PROB, 1150 ICMP6_PARAMPROB_HEADER, 1151 erroff + opt + 2 - opthead); 1152 return (-1); 1153 } 1154 #endif 1155 1156 /* 1157 * jumbo payload length must be larger than 65535. 1158 */ 1159 if (jumboplen <= IPV6_MAXPACKET) { 1160 V_ip6stat.ip6s_badoptions++; 1161 icmp6_error(m, ICMP6_PARAM_PROB, 1162 ICMP6_PARAMPROB_HEADER, 1163 erroff + opt + 2 - opthead); 1164 return (-1); 1165 } 1166 *plenp = jumboplen; 1167 1168 break; 1169 default: /* unknown option */ 1170 if (hbhlen < IP6OPT_MINLEN) { 1171 V_ip6stat.ip6s_toosmall++; 1172 goto bad; 1173 } 1174 optlen = ip6_unknown_opt(opt, m, 1175 erroff + opt - opthead); 1176 if (optlen == -1) 1177 return (-1); 1178 optlen += 2; 1179 break; 1180 } 1181 } 1182 1183 return (0); 1184 1185 bad: 1186 m_freem(m); 1187 return (-1); 1188 } 1189 1190 /* 1191 * Unknown option processing. 1192 * The third argument `off' is the offset from the IPv6 header to the option, 1193 * which is necessary if the IPv6 header the and option header and IPv6 header 1194 * is not contiguous in order to return an ICMPv6 error. 1195 */ 1196 int 1197 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1198 { 1199 struct ip6_hdr *ip6; 1200 1201 switch (IP6OPT_TYPE(*optp)) { 1202 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1203 return ((int)*(optp + 1)); 1204 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1205 m_freem(m); 1206 return (-1); 1207 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1208 V_ip6stat.ip6s_badoptions++; 1209 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1210 return (-1); 1211 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1212 V_ip6stat.ip6s_badoptions++; 1213 ip6 = mtod(m, struct ip6_hdr *); 1214 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1215 (m->m_flags & (M_BCAST|M_MCAST))) 1216 m_freem(m); 1217 else 1218 icmp6_error(m, ICMP6_PARAM_PROB, 1219 ICMP6_PARAMPROB_OPTION, off); 1220 return (-1); 1221 } 1222 1223 m_freem(m); /* XXX: NOTREACHED */ 1224 return (-1); 1225 } 1226 1227 /* 1228 * Create the "control" list for this pcb. 1229 * These functions will not modify mbuf chain at all. 1230 * 1231 * With KAME mbuf chain restriction: 1232 * The routine will be called from upper layer handlers like tcp6_input(). 1233 * Thus the routine assumes that the caller (tcp6_input) have already 1234 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the 1235 * very first mbuf on the mbuf chain. 1236 * 1237 * ip6_savecontrol_v4 will handle those options that are possible to be 1238 * set on a v4-mapped socket. 1239 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1240 * options and handle the v6-only ones itself. 1241 */ 1242 struct mbuf ** 1243 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1244 int *v4only) 1245 { 1246 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1247 1248 #ifdef SO_TIMESTAMP 1249 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1250 struct timeval tv; 1251 1252 microtime(&tv); 1253 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1254 SCM_TIMESTAMP, SOL_SOCKET); 1255 if (*mp) 1256 mp = &(*mp)->m_next; 1257 } 1258 #endif 1259 1260 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1261 if (v4only != NULL) 1262 *v4only = 1; 1263 return (mp); 1264 } 1265 1266 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1267 /* RFC 2292 sec. 5 */ 1268 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1269 struct in6_pktinfo pi6; 1270 1271 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1272 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1273 pi6.ipi6_ifindex = 1274 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1275 1276 *mp = sbcreatecontrol((caddr_t) &pi6, 1277 sizeof(struct in6_pktinfo), 1278 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1279 if (*mp) 1280 mp = &(*mp)->m_next; 1281 } 1282 1283 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1284 int hlim = ip6->ip6_hlim & 0xff; 1285 1286 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1287 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1288 IPPROTO_IPV6); 1289 if (*mp) 1290 mp = &(*mp)->m_next; 1291 } 1292 1293 if (v4only != NULL) 1294 *v4only = 0; 1295 return (mp); 1296 } 1297 1298 void 1299 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) 1300 { 1301 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1302 int v4only = 0; 1303 1304 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); 1305 if (v4only) 1306 return; 1307 1308 if ((in6p->inp_flags & IN6P_TCLASS) != 0) { 1309 u_int32_t flowinfo; 1310 int tclass; 1311 1312 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1313 flowinfo >>= 20; 1314 1315 tclass = flowinfo & 0xff; 1316 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass), 1317 IPV6_TCLASS, IPPROTO_IPV6); 1318 if (*mp) 1319 mp = &(*mp)->m_next; 1320 } 1321 1322 /* 1323 * IPV6_HOPOPTS socket option. Recall that we required super-user 1324 * privilege for the option (see ip6_ctloutput), but it might be too 1325 * strict, since there might be some hop-by-hop options which can be 1326 * returned to normal user. 1327 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1328 */ 1329 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { 1330 /* 1331 * Check if a hop-by-hop options header is contatined in the 1332 * received packet, and if so, store the options as ancillary 1333 * data. Note that a hop-by-hop options header must be 1334 * just after the IPv6 header, which is assured through the 1335 * IPv6 input processing. 1336 */ 1337 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1338 struct ip6_hbh *hbh; 1339 int hbhlen = 0; 1340 #ifdef PULLDOWN_TEST 1341 struct mbuf *ext; 1342 #endif 1343 1344 #ifndef PULLDOWN_TEST 1345 hbh = (struct ip6_hbh *)(ip6 + 1); 1346 hbhlen = (hbh->ip6h_len + 1) << 3; 1347 #else 1348 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), 1349 ip6->ip6_nxt); 1350 if (ext == NULL) { 1351 V_ip6stat.ip6s_tooshort++; 1352 return; 1353 } 1354 hbh = mtod(ext, struct ip6_hbh *); 1355 hbhlen = (hbh->ip6h_len + 1) << 3; 1356 if (hbhlen != ext->m_len) { 1357 m_freem(ext); 1358 V_ip6stat.ip6s_tooshort++; 1359 return; 1360 } 1361 #endif 1362 1363 /* 1364 * XXX: We copy the whole header even if a 1365 * jumbo payload option is included, the option which 1366 * is to be removed before returning according to 1367 * RFC2292. 1368 * Note: this constraint is removed in RFC3542 1369 */ 1370 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1371 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1372 IPPROTO_IPV6); 1373 if (*mp) 1374 mp = &(*mp)->m_next; 1375 #ifdef PULLDOWN_TEST 1376 m_freem(ext); 1377 #endif 1378 } 1379 } 1380 1381 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1382 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1383 1384 /* 1385 * Search for destination options headers or routing 1386 * header(s) through the header chain, and stores each 1387 * header as ancillary data. 1388 * Note that the order of the headers remains in 1389 * the chain of ancillary data. 1390 */ 1391 while (1) { /* is explicit loop prevention necessary? */ 1392 struct ip6_ext *ip6e = NULL; 1393 int elen; 1394 #ifdef PULLDOWN_TEST 1395 struct mbuf *ext = NULL; 1396 #endif 1397 1398 /* 1399 * if it is not an extension header, don't try to 1400 * pull it from the chain. 1401 */ 1402 switch (nxt) { 1403 case IPPROTO_DSTOPTS: 1404 case IPPROTO_ROUTING: 1405 case IPPROTO_HOPOPTS: 1406 case IPPROTO_AH: /* is it possible? */ 1407 break; 1408 default: 1409 goto loopend; 1410 } 1411 1412 #ifndef PULLDOWN_TEST 1413 if (off + sizeof(*ip6e) > m->m_len) 1414 goto loopend; 1415 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1416 if (nxt == IPPROTO_AH) 1417 elen = (ip6e->ip6e_len + 2) << 2; 1418 else 1419 elen = (ip6e->ip6e_len + 1) << 3; 1420 if (off + elen > m->m_len) 1421 goto loopend; 1422 #else 1423 ext = ip6_pullexthdr(m, off, nxt); 1424 if (ext == NULL) { 1425 V_ip6stat.ip6s_tooshort++; 1426 return; 1427 } 1428 ip6e = mtod(ext, struct ip6_ext *); 1429 if (nxt == IPPROTO_AH) 1430 elen = (ip6e->ip6e_len + 2) << 2; 1431 else 1432 elen = (ip6e->ip6e_len + 1) << 3; 1433 if (elen != ext->m_len) { 1434 m_freem(ext); 1435 V_ip6stat.ip6s_tooshort++; 1436 return; 1437 } 1438 #endif 1439 1440 switch (nxt) { 1441 case IPPROTO_DSTOPTS: 1442 if (!(in6p->inp_flags & IN6P_DSTOPTS)) 1443 break; 1444 1445 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1446 IS2292(in6p, 1447 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1448 IPPROTO_IPV6); 1449 if (*mp) 1450 mp = &(*mp)->m_next; 1451 break; 1452 case IPPROTO_ROUTING: 1453 if (!(in6p->inp_flags & IN6P_RTHDR)) 1454 break; 1455 1456 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1457 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), 1458 IPPROTO_IPV6); 1459 if (*mp) 1460 mp = &(*mp)->m_next; 1461 break; 1462 case IPPROTO_HOPOPTS: 1463 case IPPROTO_AH: /* is it possible? */ 1464 break; 1465 1466 default: 1467 /* 1468 * other cases have been filtered in the above. 1469 * none will visit this case. here we supply 1470 * the code just in case (nxt overwritten or 1471 * other cases). 1472 */ 1473 #ifdef PULLDOWN_TEST 1474 m_freem(ext); 1475 #endif 1476 goto loopend; 1477 1478 } 1479 1480 /* proceed with the next header. */ 1481 off += elen; 1482 nxt = ip6e->ip6e_nxt; 1483 ip6e = NULL; 1484 #ifdef PULLDOWN_TEST 1485 m_freem(ext); 1486 ext = NULL; 1487 #endif 1488 } 1489 loopend: 1490 ; 1491 } 1492 } 1493 #undef IS2292 1494 1495 void 1496 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu) 1497 { 1498 struct socket *so; 1499 struct mbuf *m_mtu; 1500 struct ip6_mtuinfo mtuctl; 1501 1502 so = in6p->inp_socket; 1503 1504 if (mtu == NULL) 1505 return; 1506 1507 #ifdef DIAGNOSTIC 1508 if (so == NULL) /* I believe this is impossible */ 1509 panic("ip6_notify_pmtu: socket is NULL"); 1510 #endif 1511 1512 bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */ 1513 mtuctl.ip6m_mtu = *mtu; 1514 mtuctl.ip6m_addr = *dst; 1515 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1516 return; 1517 1518 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1519 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1520 return; 1521 1522 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1523 == 0) { 1524 m_freem(m_mtu); 1525 /* XXX: should count statistics */ 1526 } else 1527 sorwakeup(so); 1528 1529 return; 1530 } 1531 1532 #ifdef PULLDOWN_TEST 1533 /* 1534 * pull single extension header from mbuf chain. returns single mbuf that 1535 * contains the result, or NULL on error. 1536 */ 1537 static struct mbuf * 1538 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) 1539 { 1540 struct ip6_ext ip6e; 1541 size_t elen; 1542 struct mbuf *n; 1543 1544 #ifdef DIAGNOSTIC 1545 switch (nxt) { 1546 case IPPROTO_DSTOPTS: 1547 case IPPROTO_ROUTING: 1548 case IPPROTO_HOPOPTS: 1549 case IPPROTO_AH: /* is it possible? */ 1550 break; 1551 default: 1552 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); 1553 } 1554 #endif 1555 1556 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1557 if (nxt == IPPROTO_AH) 1558 elen = (ip6e.ip6e_len + 2) << 2; 1559 else 1560 elen = (ip6e.ip6e_len + 1) << 3; 1561 1562 MGET(n, M_DONTWAIT, MT_DATA); 1563 if (n && elen >= MLEN) { 1564 MCLGET(n, M_DONTWAIT); 1565 if ((n->m_flags & M_EXT) == 0) { 1566 m_free(n); 1567 n = NULL; 1568 } 1569 } 1570 if (!n) 1571 return NULL; 1572 1573 n->m_len = 0; 1574 if (elen >= M_TRAILINGSPACE(n)) { 1575 m_free(n); 1576 return NULL; 1577 } 1578 1579 m_copydata(m, off, elen, mtod(n, caddr_t)); 1580 n->m_len = elen; 1581 return n; 1582 } 1583 #endif 1584 1585 /* 1586 * Get pointer to the previous header followed by the header 1587 * currently processed. 1588 * XXX: This function supposes that 1589 * M includes all headers, 1590 * the next header field and the header length field of each header 1591 * are valid, and 1592 * the sum of each header length equals to OFF. 1593 * Because of these assumptions, this function must be called very 1594 * carefully. Moreover, it will not be used in the near future when 1595 * we develop `neater' mechanism to process extension headers. 1596 */ 1597 char * 1598 ip6_get_prevhdr(struct mbuf *m, int off) 1599 { 1600 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1601 1602 if (off == sizeof(struct ip6_hdr)) 1603 return (&ip6->ip6_nxt); 1604 else { 1605 int len, nxt; 1606 struct ip6_ext *ip6e = NULL; 1607 1608 nxt = ip6->ip6_nxt; 1609 len = sizeof(struct ip6_hdr); 1610 while (len < off) { 1611 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len); 1612 1613 switch (nxt) { 1614 case IPPROTO_FRAGMENT: 1615 len += sizeof(struct ip6_frag); 1616 break; 1617 case IPPROTO_AH: 1618 len += (ip6e->ip6e_len + 2) << 2; 1619 break; 1620 default: 1621 len += (ip6e->ip6e_len + 1) << 3; 1622 break; 1623 } 1624 nxt = ip6e->ip6e_nxt; 1625 } 1626 if (ip6e) 1627 return (&ip6e->ip6e_nxt); 1628 else 1629 return NULL; 1630 } 1631 } 1632 1633 /* 1634 * get next header offset. m will be retained. 1635 */ 1636 int 1637 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp) 1638 { 1639 struct ip6_hdr ip6; 1640 struct ip6_ext ip6e; 1641 struct ip6_frag fh; 1642 1643 /* just in case */ 1644 if (m == NULL) 1645 panic("ip6_nexthdr: m == NULL"); 1646 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1647 return -1; 1648 1649 switch (proto) { 1650 case IPPROTO_IPV6: 1651 if (m->m_pkthdr.len < off + sizeof(ip6)) 1652 return -1; 1653 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1654 if (nxtp) 1655 *nxtp = ip6.ip6_nxt; 1656 off += sizeof(ip6); 1657 return off; 1658 1659 case IPPROTO_FRAGMENT: 1660 /* 1661 * terminate parsing if it is not the first fragment, 1662 * it does not make sense to parse through it. 1663 */ 1664 if (m->m_pkthdr.len < off + sizeof(fh)) 1665 return -1; 1666 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1667 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1668 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1669 return -1; 1670 if (nxtp) 1671 *nxtp = fh.ip6f_nxt; 1672 off += sizeof(struct ip6_frag); 1673 return off; 1674 1675 case IPPROTO_AH: 1676 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1677 return -1; 1678 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1679 if (nxtp) 1680 *nxtp = ip6e.ip6e_nxt; 1681 off += (ip6e.ip6e_len + 2) << 2; 1682 return off; 1683 1684 case IPPROTO_HOPOPTS: 1685 case IPPROTO_ROUTING: 1686 case IPPROTO_DSTOPTS: 1687 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1688 return -1; 1689 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1690 if (nxtp) 1691 *nxtp = ip6e.ip6e_nxt; 1692 off += (ip6e.ip6e_len + 1) << 3; 1693 return off; 1694 1695 case IPPROTO_NONE: 1696 case IPPROTO_ESP: 1697 case IPPROTO_IPCOMP: 1698 /* give up */ 1699 return -1; 1700 1701 default: 1702 return -1; 1703 } 1704 1705 return -1; 1706 } 1707 1708 /* 1709 * get offset for the last header in the chain. m will be kept untainted. 1710 */ 1711 int 1712 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp) 1713 { 1714 int newoff; 1715 int nxt; 1716 1717 if (!nxtp) { 1718 nxt = -1; 1719 nxtp = &nxt; 1720 } 1721 while (1) { 1722 newoff = ip6_nexthdr(m, off, proto, nxtp); 1723 if (newoff < 0) 1724 return off; 1725 else if (newoff < off) 1726 return -1; /* invalid */ 1727 else if (newoff == off) 1728 return newoff; 1729 1730 off = newoff; 1731 proto = *nxtp; 1732 } 1733 } 1734 1735 struct ip6aux * 1736 ip6_addaux(struct mbuf *m) 1737 { 1738 struct m_tag *mtag; 1739 1740 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1741 if (!mtag) { 1742 mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux), 1743 M_NOWAIT); 1744 if (mtag) { 1745 m_tag_prepend(m, mtag); 1746 bzero(mtag + 1, sizeof(struct ip6aux)); 1747 } 1748 } 1749 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1750 } 1751 1752 struct ip6aux * 1753 ip6_findaux(struct mbuf *m) 1754 { 1755 struct m_tag *mtag; 1756 1757 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1758 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1759 } 1760 1761 void 1762 ip6_delaux(struct mbuf *m) 1763 { 1764 struct m_tag *mtag; 1765 1766 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1767 if (mtag) 1768 m_tag_delete(m, mtag); 1769 } 1770 1771 /* 1772 * System control for IP6 1773 */ 1774 1775 u_char inet6ctlerrmap[PRC_NCMDS] = { 1776 0, 0, 0, 0, 1777 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1778 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1779 EMSGSIZE, EHOSTUNREACH, 0, 0, 1780 0, 0, 0, 0, 1781 ENOPROTOOPT 1782 }; 1783