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_ipfw.h" 69 #include "opt_ipsec.h" 70 #include "opt_route.h" 71 72 #include <sys/param.h> 73 #include <sys/systm.h> 74 #include <sys/malloc.h> 75 #include <sys/mbuf.h> 76 #include <sys/proc.h> 77 #include <sys/domain.h> 78 #include <sys/protosw.h> 79 #include <sys/socket.h> 80 #include <sys/socketvar.h> 81 #include <sys/errno.h> 82 #include <sys/time.h> 83 #include <sys/kernel.h> 84 #include <sys/syslog.h> 85 86 #include <net/if.h> 87 #include <net/if_types.h> 88 #include <net/if_dl.h> 89 #include <net/route.h> 90 #include <net/netisr.h> 91 #include <net/pfil.h> 92 #include <net/vnet.h> 93 94 #include <netinet/in.h> 95 #include <netinet/ip_var.h> 96 #include <netinet/in_systm.h> 97 #include <net/if_llatbl.h> 98 #ifdef INET 99 #include <netinet/ip.h> 100 #include <netinet/ip_icmp.h> 101 #endif /* INET */ 102 #include <netinet/ip6.h> 103 #include <netinet6/in6_var.h> 104 #include <netinet6/ip6_var.h> 105 #include <netinet/in_pcb.h> 106 #include <netinet/icmp6.h> 107 #include <netinet6/scope6_var.h> 108 #include <netinet6/in6_ifattach.h> 109 #include <netinet6/nd6.h> 110 111 #ifdef IPSEC 112 #include <netipsec/ipsec.h> 113 #include <netinet6/ip6_ipsec.h> 114 #include <netipsec/ipsec6.h> 115 #endif /* IPSEC */ 116 117 #include <netinet6/ip6protosw.h> 118 119 #ifdef FLOWTABLE 120 #include <net/flowtable.h> 121 VNET_DECLARE(int, ip6_output_flowtable_size); 122 #define V_ip6_output_flowtable_size VNET(ip6_output_flowtable_size) 123 #endif 124 125 extern struct domain inet6domain; 126 127 u_char ip6_protox[IPPROTO_MAX]; 128 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 129 130 static struct netisr_handler ip6_nh = { 131 .nh_name = "ip6", 132 .nh_handler = ip6_input, 133 .nh_proto = NETISR_IPV6, 134 .nh_policy = NETISR_POLICY_FLOW, 135 }; 136 137 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch); 138 #define V_in6_tmpaddrtimer_ch VNET(in6_tmpaddrtimer_ch) 139 140 VNET_DEFINE(struct pfil_head, inet6_pfil_hook); 141 142 VNET_DEFINE(struct ip6stat, ip6stat); 143 144 struct rwlock in6_ifaddr_lock; 145 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 146 147 static void ip6_init2(void *); 148 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *); 149 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 150 #ifdef PULLDOWN_TEST 151 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int); 152 #endif 153 154 /* 155 * IP6 initialization: fill in IP6 protocol switch table. 156 * All protocols not implemented in kernel go to raw IP6 protocol handler. 157 */ 158 void 159 ip6_init(void) 160 { 161 struct ip6protosw *pr; 162 int i; 163 164 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 165 &V_ip6_auto_linklocal); 166 167 TAILQ_INIT(&V_in6_ifaddrhead); 168 169 /* Initialize packet filter hooks. */ 170 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF; 171 V_inet6_pfil_hook.ph_af = AF_INET6; 172 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0) 173 printf("%s: WARNING: unable to register pfil hook, " 174 "error %d\n", __func__, i); 175 176 scope6_init(); 177 addrsel_policy_init(); 178 nd6_init(); 179 frag6_init(); 180 181 #ifdef FLOWTABLE 182 if (TUNABLE_INT_FETCH("net.inet6.ip6.output_flowtable_size", 183 &V_ip6_output_flowtable_size)) { 184 if (V_ip6_output_flowtable_size < 256) 185 V_ip6_output_flowtable_size = 256; 186 if (!powerof2(V_ip6_output_flowtable_size)) { 187 printf("flowtable must be power of 2 size\n"); 188 V_ip6_output_flowtable_size = 2048; 189 } 190 } else { 191 /* 192 * round up to the next power of 2 193 */ 194 V_ip6_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1); 195 } 196 V_ip6_ft = flowtable_alloc("ipv6", V_ip6_output_flowtable_size, FL_IPV6|FL_PCPU); 197 #endif 198 199 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 200 201 /* Skip global initialization stuff for non-default instances. */ 202 if (!IS_DEFAULT_VNET(curvnet)) 203 return; 204 205 #ifdef DIAGNOSTIC 206 if (sizeof(struct protosw) != sizeof(struct ip6protosw)) 207 panic("sizeof(protosw) != sizeof(ip6protosw)"); 208 #endif 209 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 210 if (pr == NULL) 211 panic("ip6_init"); 212 213 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 214 for (i = 0; i < IPPROTO_MAX; i++) 215 ip6_protox[i] = pr - inet6sw; 216 /* 217 * Cycle through IP protocols and put them into the appropriate place 218 * in ip6_protox[]. 219 */ 220 for (pr = (struct ip6protosw *)inet6domain.dom_protosw; 221 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) 222 if (pr->pr_domain->dom_family == PF_INET6 && 223 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 224 /* Be careful to only index valid IP protocols. */ 225 if (pr->pr_protocol < IPPROTO_MAX) 226 ip6_protox[pr->pr_protocol] = pr - inet6sw; 227 } 228 229 netisr_register(&ip6_nh); 230 } 231 232 /* 233 * The protocol to be inserted into ip6_protox[] must be already registered 234 * in inet6sw[], either statically or through pf_proto_register(). 235 */ 236 int 237 ip6proto_register(short ip6proto) 238 { 239 struct ip6protosw *pr; 240 241 /* Sanity checks. */ 242 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 243 return (EPROTONOSUPPORT); 244 245 /* 246 * The protocol slot must not be occupied by another protocol 247 * already. An index pointing to IPPROTO_RAW is unused. 248 */ 249 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 250 if (pr == NULL) 251 return (EPFNOSUPPORT); 252 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 253 return (EEXIST); 254 255 /* 256 * Find the protocol position in inet6sw[] and set the index. 257 */ 258 for (pr = (struct ip6protosw *)inet6domain.dom_protosw; 259 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) { 260 if (pr->pr_domain->dom_family == PF_INET6 && 261 pr->pr_protocol && pr->pr_protocol == ip6proto) { 262 ip6_protox[pr->pr_protocol] = pr - inet6sw; 263 return (0); 264 } 265 } 266 return (EPROTONOSUPPORT); 267 } 268 269 int 270 ip6proto_unregister(short ip6proto) 271 { 272 struct ip6protosw *pr; 273 274 /* Sanity checks. */ 275 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 276 return (EPROTONOSUPPORT); 277 278 /* Check if the protocol was indeed registered. */ 279 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 280 if (pr == NULL) 281 return (EPFNOSUPPORT); 282 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 283 return (ENOENT); 284 285 /* Reset the protocol slot to IPPROTO_RAW. */ 286 ip6_protox[ip6proto] = pr - inet6sw; 287 return (0); 288 } 289 290 #ifdef VIMAGE 291 void 292 ip6_destroy() 293 { 294 295 nd6_destroy(); 296 callout_drain(&V_in6_tmpaddrtimer_ch); 297 } 298 #endif 299 300 static int 301 ip6_init2_vnet(const void *unused __unused) 302 { 303 304 /* nd6_timer_init */ 305 callout_init(&V_nd6_timer_ch, 0); 306 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); 307 308 /* timer for regeneranation of temporary addresses randomize ID */ 309 callout_init(&V_in6_tmpaddrtimer_ch, 0); 310 callout_reset(&V_in6_tmpaddrtimer_ch, 311 (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor - 312 V_ip6_temp_regen_advance) * hz, 313 in6_tmpaddrtimer, curvnet); 314 315 return (0); 316 } 317 318 static void 319 ip6_init2(void *dummy) 320 { 321 322 ip6_init2_vnet(NULL); 323 } 324 325 /* cheat */ 326 /* This must be after route_init(), which is now SI_ORDER_THIRD */ 327 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL); 328 329 void 330 ip6_input(struct mbuf *m) 331 { 332 struct ip6_hdr *ip6; 333 int off = sizeof(struct ip6_hdr), nest; 334 u_int32_t plen; 335 u_int32_t rtalert = ~0; 336 int nxt, ours = 0; 337 struct ifnet *deliverifp = NULL, *ifp = NULL; 338 struct in6_addr odst; 339 struct route_in6 rin6; 340 int srcrt = 0; 341 struct llentry *lle = NULL; 342 struct sockaddr_in6 dst6, *dst; 343 344 bzero(&rin6, sizeof(struct route_in6)); 345 #ifdef IPSEC 346 /* 347 * should the inner packet be considered authentic? 348 * see comment in ah4_input(). 349 * NB: m cannot be NULL when passed to the input routine 350 */ 351 352 m->m_flags &= ~M_AUTHIPHDR; 353 m->m_flags &= ~M_AUTHIPDGM; 354 355 #endif /* IPSEC */ 356 357 /* 358 * make sure we don't have onion peering information into m_tag. 359 */ 360 ip6_delaux(m); 361 362 if (m->m_flags & M_FASTFWD_OURS) { 363 /* 364 * Firewall changed destination to local. 365 */ 366 m->m_flags &= ~M_FASTFWD_OURS; 367 ours = 1; 368 deliverifp = m->m_pkthdr.rcvif; 369 ip6 = mtod(m, struct ip6_hdr *); 370 goto hbhcheck; 371 } 372 373 /* 374 * mbuf statistics 375 */ 376 if (m->m_flags & M_EXT) { 377 if (m->m_next) 378 V_ip6stat.ip6s_mext2m++; 379 else 380 V_ip6stat.ip6s_mext1++; 381 } else { 382 #define M2MMAX (sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0])) 383 if (m->m_next) { 384 if (m->m_flags & M_LOOP) { 385 V_ip6stat.ip6s_m2m[V_loif->if_index]++; 386 } else if (m->m_pkthdr.rcvif->if_index < M2MMAX) 387 V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++; 388 else 389 V_ip6stat.ip6s_m2m[0]++; 390 } else 391 V_ip6stat.ip6s_m1++; 392 #undef M2MMAX 393 } 394 395 /* drop the packet if IPv6 operation is disabled on the IF */ 396 if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) { 397 m_freem(m); 398 return; 399 } 400 401 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive); 402 V_ip6stat.ip6s_total++; 403 404 #ifndef PULLDOWN_TEST 405 /* 406 * L2 bridge code and some other code can return mbuf chain 407 * that does not conform to KAME requirement. too bad. 408 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 409 */ 410 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 411 struct mbuf *n; 412 413 MGETHDR(n, M_DONTWAIT, MT_HEADER); 414 if (n) 415 M_MOVE_PKTHDR(n, m); 416 if (n && n->m_pkthdr.len > MHLEN) { 417 MCLGET(n, M_DONTWAIT); 418 if ((n->m_flags & M_EXT) == 0) { 419 m_freem(n); 420 n = NULL; 421 } 422 } 423 if (n == NULL) { 424 m_freem(m); 425 return; /* ENOBUFS */ 426 } 427 428 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 429 n->m_len = n->m_pkthdr.len; 430 m_freem(m); 431 m = n; 432 } 433 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */); 434 #endif 435 436 if (m->m_len < sizeof(struct ip6_hdr)) { 437 struct ifnet *inifp; 438 inifp = m->m_pkthdr.rcvif; 439 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 440 V_ip6stat.ip6s_toosmall++; 441 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 442 return; 443 } 444 } 445 446 ip6 = mtod(m, struct ip6_hdr *); 447 448 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 449 V_ip6stat.ip6s_badvers++; 450 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 451 goto bad; 452 } 453 454 V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++; 455 456 /* 457 * Check against address spoofing/corruption. 458 */ 459 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 460 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 461 /* 462 * XXX: "badscope" is not very suitable for a multicast source. 463 */ 464 V_ip6stat.ip6s_badscope++; 465 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 466 goto bad; 467 } 468 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 469 !(m->m_flags & M_LOOP)) { 470 /* 471 * In this case, the packet should come from the loopback 472 * interface. However, we cannot just check the if_flags, 473 * because ip6_mloopback() passes the "actual" interface 474 * as the outgoing/incoming interface. 475 */ 476 V_ip6stat.ip6s_badscope++; 477 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 478 goto bad; 479 } 480 481 #ifdef ALTQ 482 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 483 /* packet is dropped by traffic conditioner */ 484 return; 485 } 486 #endif 487 /* 488 * The following check is not documented in specs. A malicious 489 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 490 * and bypass security checks (act as if it was from 127.0.0.1 by using 491 * IPv6 src ::ffff:127.0.0.1). Be cautious. 492 * 493 * This check chokes if we are in an SIIT cloud. As none of BSDs 494 * support IPv4-less kernel compilation, we cannot support SIIT 495 * environment at all. So, it makes more sense for us to reject any 496 * malicious packets for non-SIIT environment, than try to do a 497 * partial support for SIIT environment. 498 */ 499 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 500 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 501 V_ip6stat.ip6s_badscope++; 502 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 503 goto bad; 504 } 505 #if 0 506 /* 507 * Reject packets with IPv4 compatible addresses (auto tunnel). 508 * 509 * The code forbids auto tunnel relay case in RFC1933 (the check is 510 * stronger than RFC1933). We may want to re-enable it if mech-xx 511 * is revised to forbid relaying case. 512 */ 513 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 514 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 515 V_ip6stat.ip6s_badscope++; 516 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 517 goto bad; 518 } 519 #endif 520 #ifdef IPSEC 521 /* 522 * Bypass packet filtering for packets previously handled by IPsec. 523 */ 524 if (ip6_ipsec_filtertunnel(m)) 525 goto passin; 526 #endif /* IPSEC */ 527 528 /* 529 * Run through list of hooks for input packets. 530 * 531 * NB: Beware of the destination address changing 532 * (e.g. by NAT rewriting). When this happens, 533 * tell ip6_forward to do the right thing. 534 */ 535 odst = ip6->ip6_dst; 536 537 /* Jump over all PFIL processing if hooks are not active. */ 538 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 539 goto passin; 540 541 if (pfil_run_hooks(&V_inet6_pfil_hook, &m, 542 m->m_pkthdr.rcvif, PFIL_IN, NULL)) 543 return; 544 if (m == NULL) /* consumed by filter */ 545 return; 546 ip6 = mtod(m, struct ip6_hdr *); 547 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 548 549 #ifdef IPFIREWALL_FORWARD 550 if (m->m_flags & M_FASTFWD_OURS) { 551 m->m_flags &= ~M_FASTFWD_OURS; 552 ours = 1; 553 deliverifp = m->m_pkthdr.rcvif; 554 goto hbhcheck; 555 } 556 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 557 /* 558 * Directly ship the packet on. This allows forwarding 559 * packets originally destined to us to some other directly 560 * connected host. 561 */ 562 ip6_forward(m, 1); 563 goto out; 564 } 565 #endif /* IPFIREWALL_FORWARD */ 566 567 passin: 568 /* 569 * Disambiguate address scope zones (if there is ambiguity). 570 * We first make sure that the original source or destination address 571 * is not in our internal form for scoped addresses. Such addresses 572 * are not necessarily invalid spec-wise, but we cannot accept them due 573 * to the usage conflict. 574 * in6_setscope() then also checks and rejects the cases where src or 575 * dst are the loopback address and the receiving interface 576 * is not loopback. 577 */ 578 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 579 V_ip6stat.ip6s_badscope++; /* XXX */ 580 goto bad; 581 } 582 if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) || 583 in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) { 584 V_ip6stat.ip6s_badscope++; 585 goto bad; 586 } 587 588 /* 589 * Multicast check. Assume packet is for us to avoid 590 * prematurely taking locks. 591 */ 592 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 593 ours = 1; 594 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast); 595 deliverifp = m->m_pkthdr.rcvif; 596 goto hbhcheck; 597 } 598 599 /* 600 * Unicast check 601 */ 602 603 bzero(&dst6, sizeof(dst6)); 604 dst6.sin6_family = AF_INET6; 605 dst6.sin6_len = sizeof(struct sockaddr_in6); 606 dst6.sin6_addr = ip6->ip6_dst; 607 ifp = m->m_pkthdr.rcvif; 608 IF_AFDATA_LOCK(ifp); 609 lle = lla_lookup(LLTABLE6(ifp), 0, 610 (struct sockaddr *)&dst6); 611 IF_AFDATA_UNLOCK(ifp); 612 if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) { 613 struct ifaddr *ifa; 614 struct in6_ifaddr *ia6; 615 int bad; 616 617 bad = 1; 618 #define sa_equal(a1, a2) \ 619 (bcmp((a1), (a2), ((a1))->sin6_len) == 0) 620 IF_ADDR_LOCK(ifp); 621 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 622 if (ifa->ifa_addr->sa_family != dst6.sin6_family) 623 continue; 624 if (sa_equal(&dst6, ifa->ifa_addr)) 625 break; 626 } 627 KASSERT(ifa != NULL, ("%s: ifa not found for lle %p", 628 __func__, lle)); 629 #undef sa_equal 630 631 ia6 = (struct in6_ifaddr *)ifa; 632 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) { 633 /* Count the packet in the ip address stats */ 634 ia6->ia_ifa.if_ipackets++; 635 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len; 636 637 /* 638 * record address information into m_tag. 639 */ 640 (void)ip6_setdstifaddr(m, ia6); 641 642 bad = 0; 643 } else { 644 char ip6bufs[INET6_ADDRSTRLEN]; 645 char ip6bufd[INET6_ADDRSTRLEN]; 646 /* address is not ready, so discard the packet. */ 647 nd6log((LOG_INFO, 648 "ip6_input: packet to an unready address %s->%s\n", 649 ip6_sprintf(ip6bufs, &ip6->ip6_src), 650 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 651 } 652 IF_ADDR_UNLOCK(ifp); 653 LLE_RUNLOCK(lle); 654 if (bad) 655 goto bad; 656 else { 657 ours = 1; 658 deliverifp = ifp; 659 goto hbhcheck; 660 } 661 } 662 if (lle != NULL) 663 LLE_RUNLOCK(lle); 664 665 dst = &rin6.ro_dst; 666 dst->sin6_len = sizeof(struct sockaddr_in6); 667 dst->sin6_family = AF_INET6; 668 dst->sin6_addr = ip6->ip6_dst; 669 rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0); 670 if (rin6.ro_rt) 671 RT_UNLOCK(rin6.ro_rt); 672 673 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key)) 674 675 /* 676 * Accept the packet if the forwarding interface to the destination 677 * according to the routing table is the loopback interface, 678 * unless the associated route has a gateway. 679 * Note that this approach causes to accept a packet if there is a 680 * route to the loopback interface for the destination of the packet. 681 * But we think it's even useful in some situations, e.g. when using 682 * a special daemon which wants to intercept the packet. 683 * 684 * XXX: some OSes automatically make a cloned route for the destination 685 * of an outgoing packet. If the outgoing interface of the packet 686 * is a loopback one, the kernel would consider the packet to be 687 * accepted, even if we have no such address assinged on the interface. 688 * We check the cloned flag of the route entry to reject such cases, 689 * assuming that route entries for our own addresses are not made by 690 * cloning (it should be true because in6_addloop explicitly installs 691 * the host route). However, we might have to do an explicit check 692 * while it would be less efficient. Or, should we rather install a 693 * reject route for such a case? 694 */ 695 if (rin6.ro_rt && 696 (rin6.ro_rt->rt_flags & 697 (RTF_HOST|RTF_GATEWAY)) == RTF_HOST && 698 #ifdef RTF_WASCLONED 699 !(rin6.ro_rt->rt_flags & RTF_WASCLONED) && 700 #endif 701 #ifdef RTF_CLONED 702 !(rin6.ro_rt->rt_flags & RTF_CLONED) && 703 #endif 704 #if 0 705 /* 706 * The check below is redundant since the comparison of 707 * the destination and the key of the rtentry has 708 * already done through looking up the routing table. 709 */ 710 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 711 &rt6_key(rin6.ro_rt)->sin6_addr) 712 #endif 713 rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) { 714 int free_ia6 = 0; 715 struct in6_ifaddr *ia6; 716 717 /* 718 * found the loopback route to the interface address 719 */ 720 if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) { 721 struct sockaddr_in6 dest6; 722 723 bzero(&dest6, sizeof(dest6)); 724 dest6.sin6_family = AF_INET6; 725 dest6.sin6_len = sizeof(dest6); 726 dest6.sin6_addr = ip6->ip6_dst; 727 ia6 = (struct in6_ifaddr *) 728 ifa_ifwithaddr((struct sockaddr *)&dest6); 729 if (ia6 == NULL) 730 goto bad; 731 free_ia6 = 1; 732 } 733 else 734 ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa; 735 736 /* 737 * record address information into m_tag. 738 */ 739 (void)ip6_setdstifaddr(m, ia6); 740 741 /* 742 * packets to a tentative, duplicated, or somehow invalid 743 * address must not be accepted. 744 */ 745 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) { 746 /* this address is ready */ 747 ours = 1; 748 deliverifp = ia6->ia_ifp; /* correct? */ 749 /* Count the packet in the ip address stats */ 750 ia6->ia_ifa.if_ipackets++; 751 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len; 752 if (ia6 != NULL && free_ia6 != 0) 753 ifa_free(&ia6->ia_ifa); 754 goto hbhcheck; 755 } else { 756 char ip6bufs[INET6_ADDRSTRLEN]; 757 char ip6bufd[INET6_ADDRSTRLEN]; 758 /* address is not ready, so discard the packet. */ 759 nd6log((LOG_INFO, 760 "ip6_input: packet to an unready address %s->%s\n", 761 ip6_sprintf(ip6bufs, &ip6->ip6_src), 762 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 763 764 if (ia6 != NULL && free_ia6 != 0) 765 ifa_free(&ia6->ia_ifa); 766 goto bad; 767 } 768 } 769 770 /* 771 * FAITH (Firewall Aided Internet Translator) 772 */ 773 if (V_ip6_keepfaith) { 774 if (rin6.ro_rt && rin6.ro_rt->rt_ifp && 775 rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) { 776 /* XXX do we need more sanity checks? */ 777 ours = 1; 778 deliverifp = rin6.ro_rt->rt_ifp; /* faith */ 779 goto hbhcheck; 780 } 781 } 782 783 /* 784 * Now there is no reason to process the packet if it's not our own 785 * and we're not a router. 786 */ 787 if (!V_ip6_forwarding) { 788 V_ip6stat.ip6s_cantforward++; 789 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 790 goto bad; 791 } 792 793 hbhcheck: 794 /* 795 * record address information into m_tag, if we don't have one yet. 796 * note that we are unable to record it, if the address is not listed 797 * as our interface address (e.g. multicast addresses, addresses 798 * within FAITH prefixes and such). 799 */ 800 if (deliverifp && !ip6_getdstifaddr(m)) { 801 struct in6_ifaddr *ia6; 802 803 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst); 804 if (ia6) { 805 if (!ip6_setdstifaddr(m, ia6)) { 806 /* 807 * XXX maybe we should drop the packet here, 808 * as we could not provide enough information 809 * to the upper layers. 810 */ 811 } 812 ifa_free(&ia6->ia_ifa); 813 } 814 } 815 816 /* 817 * Process Hop-by-Hop options header if it's contained. 818 * m may be modified in ip6_hopopts_input(). 819 * If a JumboPayload option is included, plen will also be modified. 820 */ 821 plen = (u_int32_t)ntohs(ip6->ip6_plen); 822 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 823 struct ip6_hbh *hbh; 824 825 if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) { 826 #if 0 /*touches NULL pointer*/ 827 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 828 #endif 829 goto out; /* m have already been freed */ 830 } 831 832 /* adjust pointer */ 833 ip6 = mtod(m, struct ip6_hdr *); 834 835 /* 836 * if the payload length field is 0 and the next header field 837 * indicates Hop-by-Hop Options header, then a Jumbo Payload 838 * option MUST be included. 839 */ 840 if (ip6->ip6_plen == 0 && plen == 0) { 841 /* 842 * Note that if a valid jumbo payload option is 843 * contained, ip6_hopopts_input() must set a valid 844 * (non-zero) payload length to the variable plen. 845 */ 846 V_ip6stat.ip6s_badoptions++; 847 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 848 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 849 icmp6_error(m, ICMP6_PARAM_PROB, 850 ICMP6_PARAMPROB_HEADER, 851 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 852 goto out; 853 } 854 #ifndef PULLDOWN_TEST 855 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 856 hbh = (struct ip6_hbh *)(ip6 + 1); 857 #else 858 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 859 sizeof(struct ip6_hbh)); 860 if (hbh == NULL) { 861 V_ip6stat.ip6s_tooshort++; 862 goto out; 863 } 864 #endif 865 nxt = hbh->ip6h_nxt; 866 867 /* 868 * If we are acting as a router and the packet contains a 869 * router alert option, see if we know the option value. 870 * Currently, we only support the option value for MLD, in which 871 * case we should pass the packet to the multicast routing 872 * daemon. 873 */ 874 if (rtalert != ~0) { 875 switch (rtalert) { 876 case IP6OPT_RTALERT_MLD: 877 if (V_ip6_forwarding) 878 ours = 1; 879 break; 880 default: 881 /* 882 * RFC2711 requires unrecognized values must be 883 * silently ignored. 884 */ 885 break; 886 } 887 } 888 } else 889 nxt = ip6->ip6_nxt; 890 891 /* 892 * Check that the amount of data in the buffers 893 * is as at least much as the IPv6 header would have us expect. 894 * Trim mbufs if longer than we expect. 895 * Drop packet if shorter than we expect. 896 */ 897 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 898 V_ip6stat.ip6s_tooshort++; 899 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 900 goto bad; 901 } 902 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 903 if (m->m_len == m->m_pkthdr.len) { 904 m->m_len = sizeof(struct ip6_hdr) + plen; 905 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 906 } else 907 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 908 } 909 910 /* 911 * Forward if desirable. 912 */ 913 if (V_ip6_mrouter && 914 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 915 /* 916 * If we are acting as a multicast router, all 917 * incoming multicast packets are passed to the 918 * kernel-level multicast forwarding function. 919 * The packet is returned (relatively) intact; if 920 * ip6_mforward() returns a non-zero value, the packet 921 * must be discarded, else it may be accepted below. 922 * 923 * XXX TODO: Check hlim and multicast scope here to avoid 924 * unnecessarily calling into ip6_mforward(). 925 */ 926 if (ip6_mforward && 927 ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) { 928 IP6STAT_INC(ip6s_cantforward); 929 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 930 goto bad; 931 } 932 } else if (!ours) { 933 ip6_forward(m, srcrt); 934 goto out; 935 } 936 937 ip6 = mtod(m, struct ip6_hdr *); 938 939 /* 940 * Malicious party may be able to use IPv4 mapped addr to confuse 941 * tcp/udp stack and bypass security checks (act as if it was from 942 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 943 * 944 * For SIIT end node behavior, you may want to disable the check. 945 * However, you will become vulnerable to attacks using IPv4 mapped 946 * source. 947 */ 948 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 949 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 950 V_ip6stat.ip6s_badscope++; 951 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 952 goto bad; 953 } 954 955 /* 956 * Tell launch routine the next header 957 */ 958 V_ip6stat.ip6s_delivered++; 959 in6_ifstat_inc(deliverifp, ifs6_in_deliver); 960 nest = 0; 961 962 while (nxt != IPPROTO_DONE) { 963 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 964 V_ip6stat.ip6s_toomanyhdr++; 965 goto bad; 966 } 967 968 /* 969 * protection against faulty packet - there should be 970 * more sanity checks in header chain processing. 971 */ 972 if (m->m_pkthdr.len < off) { 973 V_ip6stat.ip6s_tooshort++; 974 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 975 goto bad; 976 } 977 978 #ifdef IPSEC 979 /* 980 * enforce IPsec policy checking if we are seeing last header. 981 * note that we do not visit this with protocols with pcb layer 982 * code - like udp/tcp/raw ip. 983 */ 984 if (ip6_ipsec_input(m, nxt)) 985 goto bad; 986 #endif /* IPSEC */ 987 988 /* 989 * Use mbuf flags to propagate Router Alert option to 990 * ICMPv6 layer, as hop-by-hop options have been stripped. 991 */ 992 if (nxt == IPPROTO_ICMPV6 && rtalert != ~0) 993 m->m_flags |= M_RTALERT_MLD; 994 995 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 996 } 997 goto out; 998 bad: 999 m_freem(m); 1000 out: 1001 if (rin6.ro_rt) 1002 RTFREE(rin6.ro_rt); 1003 } 1004 1005 /* 1006 * set/grab in6_ifaddr correspond to IPv6 destination address. 1007 * XXX backward compatibility wrapper 1008 * 1009 * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag, 1010 * and then bump it when the tag is copied, and release it when the tag is 1011 * freed. Unfortunately, m_tags don't support deep copies (yet), so instead 1012 * we just bump the ia refcount when we receive it. This should be fixed. 1013 */ 1014 static struct ip6aux * 1015 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6) 1016 { 1017 struct ip6aux *ip6a; 1018 1019 ip6a = ip6_addaux(m); 1020 if (ip6a) 1021 ip6a->ip6a_dstia6 = ia6; 1022 return ip6a; /* NULL if failed to set */ 1023 } 1024 1025 struct in6_ifaddr * 1026 ip6_getdstifaddr(struct mbuf *m) 1027 { 1028 struct ip6aux *ip6a; 1029 struct in6_ifaddr *ia; 1030 1031 ip6a = ip6_findaux(m); 1032 if (ip6a) { 1033 ia = ip6a->ip6a_dstia6; 1034 ifa_ref(&ia->ia_ifa); 1035 return ia; 1036 } else 1037 return NULL; 1038 } 1039 1040 /* 1041 * Hop-by-Hop options header processing. If a valid jumbo payload option is 1042 * included, the real payload length will be stored in plenp. 1043 * 1044 * rtalertp - XXX: should be stored more smart way 1045 */ 1046 static int 1047 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 1048 struct mbuf **mp, int *offp) 1049 { 1050 struct mbuf *m = *mp; 1051 int off = *offp, hbhlen; 1052 struct ip6_hbh *hbh; 1053 u_int8_t *opt; 1054 1055 /* validation of the length of the header */ 1056 #ifndef PULLDOWN_TEST 1057 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); 1058 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 1059 hbhlen = (hbh->ip6h_len + 1) << 3; 1060 1061 IP6_EXTHDR_CHECK(m, off, hbhlen, -1); 1062 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 1063 #else 1064 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, 1065 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); 1066 if (hbh == NULL) { 1067 V_ip6stat.ip6s_tooshort++; 1068 return -1; 1069 } 1070 hbhlen = (hbh->ip6h_len + 1) << 3; 1071 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 1072 hbhlen); 1073 if (hbh == NULL) { 1074 V_ip6stat.ip6s_tooshort++; 1075 return -1; 1076 } 1077 #endif 1078 off += hbhlen; 1079 hbhlen -= sizeof(struct ip6_hbh); 1080 opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh); 1081 1082 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 1083 hbhlen, rtalertp, plenp) < 0) 1084 return (-1); 1085 1086 *offp = off; 1087 *mp = m; 1088 return (0); 1089 } 1090 1091 /* 1092 * Search header for all Hop-by-hop options and process each option. 1093 * This function is separate from ip6_hopopts_input() in order to 1094 * handle a case where the sending node itself process its hop-by-hop 1095 * options header. In such a case, the function is called from ip6_output(). 1096 * 1097 * The function assumes that hbh header is located right after the IPv6 header 1098 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1099 * opthead + hbhlen is located in contiguous memory region. 1100 */ 1101 int 1102 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1103 u_int32_t *rtalertp, u_int32_t *plenp) 1104 { 1105 struct ip6_hdr *ip6; 1106 int optlen = 0; 1107 u_int8_t *opt = opthead; 1108 u_int16_t rtalert_val; 1109 u_int32_t jumboplen; 1110 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1111 1112 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1113 switch (*opt) { 1114 case IP6OPT_PAD1: 1115 optlen = 1; 1116 break; 1117 case IP6OPT_PADN: 1118 if (hbhlen < IP6OPT_MINLEN) { 1119 V_ip6stat.ip6s_toosmall++; 1120 goto bad; 1121 } 1122 optlen = *(opt + 1) + 2; 1123 break; 1124 case IP6OPT_ROUTER_ALERT: 1125 /* XXX may need check for alignment */ 1126 if (hbhlen < IP6OPT_RTALERT_LEN) { 1127 V_ip6stat.ip6s_toosmall++; 1128 goto bad; 1129 } 1130 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1131 /* XXX stat */ 1132 icmp6_error(m, ICMP6_PARAM_PROB, 1133 ICMP6_PARAMPROB_HEADER, 1134 erroff + opt + 1 - opthead); 1135 return (-1); 1136 } 1137 optlen = IP6OPT_RTALERT_LEN; 1138 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1139 *rtalertp = ntohs(rtalert_val); 1140 break; 1141 case IP6OPT_JUMBO: 1142 /* XXX may need check for alignment */ 1143 if (hbhlen < IP6OPT_JUMBO_LEN) { 1144 V_ip6stat.ip6s_toosmall++; 1145 goto bad; 1146 } 1147 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1148 /* XXX stat */ 1149 icmp6_error(m, ICMP6_PARAM_PROB, 1150 ICMP6_PARAMPROB_HEADER, 1151 erroff + opt + 1 - opthead); 1152 return (-1); 1153 } 1154 optlen = IP6OPT_JUMBO_LEN; 1155 1156 /* 1157 * IPv6 packets that have non 0 payload length 1158 * must not contain a jumbo payload option. 1159 */ 1160 ip6 = mtod(m, struct ip6_hdr *); 1161 if (ip6->ip6_plen) { 1162 V_ip6stat.ip6s_badoptions++; 1163 icmp6_error(m, ICMP6_PARAM_PROB, 1164 ICMP6_PARAMPROB_HEADER, 1165 erroff + opt - opthead); 1166 return (-1); 1167 } 1168 1169 /* 1170 * We may see jumbolen in unaligned location, so 1171 * we'd need to perform bcopy(). 1172 */ 1173 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1174 jumboplen = (u_int32_t)htonl(jumboplen); 1175 1176 #if 1 1177 /* 1178 * if there are multiple jumbo payload options, 1179 * *plenp will be non-zero and the packet will be 1180 * rejected. 1181 * the behavior may need some debate in ipngwg - 1182 * multiple options does not make sense, however, 1183 * there's no explicit mention in specification. 1184 */ 1185 if (*plenp != 0) { 1186 V_ip6stat.ip6s_badoptions++; 1187 icmp6_error(m, ICMP6_PARAM_PROB, 1188 ICMP6_PARAMPROB_HEADER, 1189 erroff + opt + 2 - opthead); 1190 return (-1); 1191 } 1192 #endif 1193 1194 /* 1195 * jumbo payload length must be larger than 65535. 1196 */ 1197 if (jumboplen <= IPV6_MAXPACKET) { 1198 V_ip6stat.ip6s_badoptions++; 1199 icmp6_error(m, ICMP6_PARAM_PROB, 1200 ICMP6_PARAMPROB_HEADER, 1201 erroff + opt + 2 - opthead); 1202 return (-1); 1203 } 1204 *plenp = jumboplen; 1205 1206 break; 1207 default: /* unknown option */ 1208 if (hbhlen < IP6OPT_MINLEN) { 1209 V_ip6stat.ip6s_toosmall++; 1210 goto bad; 1211 } 1212 optlen = ip6_unknown_opt(opt, m, 1213 erroff + opt - opthead); 1214 if (optlen == -1) 1215 return (-1); 1216 optlen += 2; 1217 break; 1218 } 1219 } 1220 1221 return (0); 1222 1223 bad: 1224 m_freem(m); 1225 return (-1); 1226 } 1227 1228 /* 1229 * Unknown option processing. 1230 * The third argument `off' is the offset from the IPv6 header to the option, 1231 * which is necessary if the IPv6 header the and option header and IPv6 header 1232 * is not contiguous in order to return an ICMPv6 error. 1233 */ 1234 int 1235 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1236 { 1237 struct ip6_hdr *ip6; 1238 1239 switch (IP6OPT_TYPE(*optp)) { 1240 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1241 return ((int)*(optp + 1)); 1242 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1243 m_freem(m); 1244 return (-1); 1245 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1246 V_ip6stat.ip6s_badoptions++; 1247 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1248 return (-1); 1249 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1250 V_ip6stat.ip6s_badoptions++; 1251 ip6 = mtod(m, struct ip6_hdr *); 1252 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1253 (m->m_flags & (M_BCAST|M_MCAST))) 1254 m_freem(m); 1255 else 1256 icmp6_error(m, ICMP6_PARAM_PROB, 1257 ICMP6_PARAMPROB_OPTION, off); 1258 return (-1); 1259 } 1260 1261 m_freem(m); /* XXX: NOTREACHED */ 1262 return (-1); 1263 } 1264 1265 /* 1266 * Create the "control" list for this pcb. 1267 * These functions will not modify mbuf chain at all. 1268 * 1269 * With KAME mbuf chain restriction: 1270 * The routine will be called from upper layer handlers like tcp6_input(). 1271 * Thus the routine assumes that the caller (tcp6_input) have already 1272 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the 1273 * very first mbuf on the mbuf chain. 1274 * 1275 * ip6_savecontrol_v4 will handle those options that are possible to be 1276 * set on a v4-mapped socket. 1277 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1278 * options and handle the v6-only ones itself. 1279 */ 1280 struct mbuf ** 1281 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1282 int *v4only) 1283 { 1284 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1285 1286 #ifdef SO_TIMESTAMP 1287 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1288 struct timeval tv; 1289 1290 microtime(&tv); 1291 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1292 SCM_TIMESTAMP, SOL_SOCKET); 1293 if (*mp) 1294 mp = &(*mp)->m_next; 1295 } 1296 #endif 1297 1298 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1299 if (v4only != NULL) 1300 *v4only = 1; 1301 return (mp); 1302 } 1303 1304 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1305 /* RFC 2292 sec. 5 */ 1306 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1307 struct in6_pktinfo pi6; 1308 1309 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1310 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1311 pi6.ipi6_ifindex = 1312 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1313 1314 *mp = sbcreatecontrol((caddr_t) &pi6, 1315 sizeof(struct in6_pktinfo), 1316 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1317 if (*mp) 1318 mp = &(*mp)->m_next; 1319 } 1320 1321 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1322 int hlim = ip6->ip6_hlim & 0xff; 1323 1324 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1325 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1326 IPPROTO_IPV6); 1327 if (*mp) 1328 mp = &(*mp)->m_next; 1329 } 1330 1331 if (v4only != NULL) 1332 *v4only = 0; 1333 return (mp); 1334 } 1335 1336 void 1337 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) 1338 { 1339 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1340 int v4only = 0; 1341 1342 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); 1343 if (v4only) 1344 return; 1345 1346 if ((in6p->inp_flags & IN6P_TCLASS) != 0) { 1347 u_int32_t flowinfo; 1348 int tclass; 1349 1350 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1351 flowinfo >>= 20; 1352 1353 tclass = flowinfo & 0xff; 1354 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass), 1355 IPV6_TCLASS, IPPROTO_IPV6); 1356 if (*mp) 1357 mp = &(*mp)->m_next; 1358 } 1359 1360 /* 1361 * IPV6_HOPOPTS socket option. Recall that we required super-user 1362 * privilege for the option (see ip6_ctloutput), but it might be too 1363 * strict, since there might be some hop-by-hop options which can be 1364 * returned to normal user. 1365 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1366 */ 1367 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { 1368 /* 1369 * Check if a hop-by-hop options header is contatined in the 1370 * received packet, and if so, store the options as ancillary 1371 * data. Note that a hop-by-hop options header must be 1372 * just after the IPv6 header, which is assured through the 1373 * IPv6 input processing. 1374 */ 1375 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1376 struct ip6_hbh *hbh; 1377 int hbhlen = 0; 1378 #ifdef PULLDOWN_TEST 1379 struct mbuf *ext; 1380 #endif 1381 1382 #ifndef PULLDOWN_TEST 1383 hbh = (struct ip6_hbh *)(ip6 + 1); 1384 hbhlen = (hbh->ip6h_len + 1) << 3; 1385 #else 1386 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), 1387 ip6->ip6_nxt); 1388 if (ext == NULL) { 1389 V_ip6stat.ip6s_tooshort++; 1390 return; 1391 } 1392 hbh = mtod(ext, struct ip6_hbh *); 1393 hbhlen = (hbh->ip6h_len + 1) << 3; 1394 if (hbhlen != ext->m_len) { 1395 m_freem(ext); 1396 V_ip6stat.ip6s_tooshort++; 1397 return; 1398 } 1399 #endif 1400 1401 /* 1402 * XXX: We copy the whole header even if a 1403 * jumbo payload option is included, the option which 1404 * is to be removed before returning according to 1405 * RFC2292. 1406 * Note: this constraint is removed in RFC3542 1407 */ 1408 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1409 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1410 IPPROTO_IPV6); 1411 if (*mp) 1412 mp = &(*mp)->m_next; 1413 #ifdef PULLDOWN_TEST 1414 m_freem(ext); 1415 #endif 1416 } 1417 } 1418 1419 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1420 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1421 1422 /* 1423 * Search for destination options headers or routing 1424 * header(s) through the header chain, and stores each 1425 * header as ancillary data. 1426 * Note that the order of the headers remains in 1427 * the chain of ancillary data. 1428 */ 1429 while (1) { /* is explicit loop prevention necessary? */ 1430 struct ip6_ext *ip6e = NULL; 1431 int elen; 1432 #ifdef PULLDOWN_TEST 1433 struct mbuf *ext = NULL; 1434 #endif 1435 1436 /* 1437 * if it is not an extension header, don't try to 1438 * pull it from the chain. 1439 */ 1440 switch (nxt) { 1441 case IPPROTO_DSTOPTS: 1442 case IPPROTO_ROUTING: 1443 case IPPROTO_HOPOPTS: 1444 case IPPROTO_AH: /* is it possible? */ 1445 break; 1446 default: 1447 goto loopend; 1448 } 1449 1450 #ifndef PULLDOWN_TEST 1451 if (off + sizeof(*ip6e) > m->m_len) 1452 goto loopend; 1453 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1454 if (nxt == IPPROTO_AH) 1455 elen = (ip6e->ip6e_len + 2) << 2; 1456 else 1457 elen = (ip6e->ip6e_len + 1) << 3; 1458 if (off + elen > m->m_len) 1459 goto loopend; 1460 #else 1461 ext = ip6_pullexthdr(m, off, nxt); 1462 if (ext == NULL) { 1463 V_ip6stat.ip6s_tooshort++; 1464 return; 1465 } 1466 ip6e = mtod(ext, struct ip6_ext *); 1467 if (nxt == IPPROTO_AH) 1468 elen = (ip6e->ip6e_len + 2) << 2; 1469 else 1470 elen = (ip6e->ip6e_len + 1) << 3; 1471 if (elen != ext->m_len) { 1472 m_freem(ext); 1473 V_ip6stat.ip6s_tooshort++; 1474 return; 1475 } 1476 #endif 1477 1478 switch (nxt) { 1479 case IPPROTO_DSTOPTS: 1480 if (!(in6p->inp_flags & IN6P_DSTOPTS)) 1481 break; 1482 1483 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1484 IS2292(in6p, 1485 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1486 IPPROTO_IPV6); 1487 if (*mp) 1488 mp = &(*mp)->m_next; 1489 break; 1490 case IPPROTO_ROUTING: 1491 if (!(in6p->inp_flags & IN6P_RTHDR)) 1492 break; 1493 1494 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1495 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), 1496 IPPROTO_IPV6); 1497 if (*mp) 1498 mp = &(*mp)->m_next; 1499 break; 1500 case IPPROTO_HOPOPTS: 1501 case IPPROTO_AH: /* is it possible? */ 1502 break; 1503 1504 default: 1505 /* 1506 * other cases have been filtered in the above. 1507 * none will visit this case. here we supply 1508 * the code just in case (nxt overwritten or 1509 * other cases). 1510 */ 1511 #ifdef PULLDOWN_TEST 1512 m_freem(ext); 1513 #endif 1514 goto loopend; 1515 1516 } 1517 1518 /* proceed with the next header. */ 1519 off += elen; 1520 nxt = ip6e->ip6e_nxt; 1521 ip6e = NULL; 1522 #ifdef PULLDOWN_TEST 1523 m_freem(ext); 1524 ext = NULL; 1525 #endif 1526 } 1527 loopend: 1528 ; 1529 } 1530 } 1531 #undef IS2292 1532 1533 void 1534 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu) 1535 { 1536 struct socket *so; 1537 struct mbuf *m_mtu; 1538 struct ip6_mtuinfo mtuctl; 1539 1540 so = in6p->inp_socket; 1541 1542 if (mtu == NULL) 1543 return; 1544 1545 #ifdef DIAGNOSTIC 1546 if (so == NULL) /* I believe this is impossible */ 1547 panic("ip6_notify_pmtu: socket is NULL"); 1548 #endif 1549 1550 bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */ 1551 mtuctl.ip6m_mtu = *mtu; 1552 mtuctl.ip6m_addr = *dst; 1553 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1554 return; 1555 1556 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1557 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1558 return; 1559 1560 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1561 == 0) { 1562 m_freem(m_mtu); 1563 /* XXX: should count statistics */ 1564 } else 1565 sorwakeup(so); 1566 1567 return; 1568 } 1569 1570 #ifdef PULLDOWN_TEST 1571 /* 1572 * pull single extension header from mbuf chain. returns single mbuf that 1573 * contains the result, or NULL on error. 1574 */ 1575 static struct mbuf * 1576 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) 1577 { 1578 struct ip6_ext ip6e; 1579 size_t elen; 1580 struct mbuf *n; 1581 1582 #ifdef DIAGNOSTIC 1583 switch (nxt) { 1584 case IPPROTO_DSTOPTS: 1585 case IPPROTO_ROUTING: 1586 case IPPROTO_HOPOPTS: 1587 case IPPROTO_AH: /* is it possible? */ 1588 break; 1589 default: 1590 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); 1591 } 1592 #endif 1593 1594 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1595 if (nxt == IPPROTO_AH) 1596 elen = (ip6e.ip6e_len + 2) << 2; 1597 else 1598 elen = (ip6e.ip6e_len + 1) << 3; 1599 1600 MGET(n, M_DONTWAIT, MT_DATA); 1601 if (n && elen >= MLEN) { 1602 MCLGET(n, M_DONTWAIT); 1603 if ((n->m_flags & M_EXT) == 0) { 1604 m_free(n); 1605 n = NULL; 1606 } 1607 } 1608 if (!n) 1609 return NULL; 1610 1611 n->m_len = 0; 1612 if (elen >= M_TRAILINGSPACE(n)) { 1613 m_free(n); 1614 return NULL; 1615 } 1616 1617 m_copydata(m, off, elen, mtod(n, caddr_t)); 1618 n->m_len = elen; 1619 return n; 1620 } 1621 #endif 1622 1623 /* 1624 * Get pointer to the previous header followed by the header 1625 * currently processed. 1626 * XXX: This function supposes that 1627 * M includes all headers, 1628 * the next header field and the header length field of each header 1629 * are valid, and 1630 * the sum of each header length equals to OFF. 1631 * Because of these assumptions, this function must be called very 1632 * carefully. Moreover, it will not be used in the near future when 1633 * we develop `neater' mechanism to process extension headers. 1634 */ 1635 char * 1636 ip6_get_prevhdr(struct mbuf *m, int off) 1637 { 1638 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1639 1640 if (off == sizeof(struct ip6_hdr)) 1641 return (&ip6->ip6_nxt); 1642 else { 1643 int len, nxt; 1644 struct ip6_ext *ip6e = NULL; 1645 1646 nxt = ip6->ip6_nxt; 1647 len = sizeof(struct ip6_hdr); 1648 while (len < off) { 1649 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len); 1650 1651 switch (nxt) { 1652 case IPPROTO_FRAGMENT: 1653 len += sizeof(struct ip6_frag); 1654 break; 1655 case IPPROTO_AH: 1656 len += (ip6e->ip6e_len + 2) << 2; 1657 break; 1658 default: 1659 len += (ip6e->ip6e_len + 1) << 3; 1660 break; 1661 } 1662 nxt = ip6e->ip6e_nxt; 1663 } 1664 if (ip6e) 1665 return (&ip6e->ip6e_nxt); 1666 else 1667 return NULL; 1668 } 1669 } 1670 1671 /* 1672 * get next header offset. m will be retained. 1673 */ 1674 int 1675 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp) 1676 { 1677 struct ip6_hdr ip6; 1678 struct ip6_ext ip6e; 1679 struct ip6_frag fh; 1680 1681 /* just in case */ 1682 if (m == NULL) 1683 panic("ip6_nexthdr: m == NULL"); 1684 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1685 return -1; 1686 1687 switch (proto) { 1688 case IPPROTO_IPV6: 1689 if (m->m_pkthdr.len < off + sizeof(ip6)) 1690 return -1; 1691 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1692 if (nxtp) 1693 *nxtp = ip6.ip6_nxt; 1694 off += sizeof(ip6); 1695 return off; 1696 1697 case IPPROTO_FRAGMENT: 1698 /* 1699 * terminate parsing if it is not the first fragment, 1700 * it does not make sense to parse through it. 1701 */ 1702 if (m->m_pkthdr.len < off + sizeof(fh)) 1703 return -1; 1704 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1705 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1706 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1707 return -1; 1708 if (nxtp) 1709 *nxtp = fh.ip6f_nxt; 1710 off += sizeof(struct ip6_frag); 1711 return off; 1712 1713 case IPPROTO_AH: 1714 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1715 return -1; 1716 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1717 if (nxtp) 1718 *nxtp = ip6e.ip6e_nxt; 1719 off += (ip6e.ip6e_len + 2) << 2; 1720 return off; 1721 1722 case IPPROTO_HOPOPTS: 1723 case IPPROTO_ROUTING: 1724 case IPPROTO_DSTOPTS: 1725 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1726 return -1; 1727 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1728 if (nxtp) 1729 *nxtp = ip6e.ip6e_nxt; 1730 off += (ip6e.ip6e_len + 1) << 3; 1731 return off; 1732 1733 case IPPROTO_NONE: 1734 case IPPROTO_ESP: 1735 case IPPROTO_IPCOMP: 1736 /* give up */ 1737 return -1; 1738 1739 default: 1740 return -1; 1741 } 1742 1743 return -1; 1744 } 1745 1746 /* 1747 * get offset for the last header in the chain. m will be kept untainted. 1748 */ 1749 int 1750 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp) 1751 { 1752 int newoff; 1753 int nxt; 1754 1755 if (!nxtp) { 1756 nxt = -1; 1757 nxtp = &nxt; 1758 } 1759 while (1) { 1760 newoff = ip6_nexthdr(m, off, proto, nxtp); 1761 if (newoff < 0) 1762 return off; 1763 else if (newoff < off) 1764 return -1; /* invalid */ 1765 else if (newoff == off) 1766 return newoff; 1767 1768 off = newoff; 1769 proto = *nxtp; 1770 } 1771 } 1772 1773 struct ip6aux * 1774 ip6_addaux(struct mbuf *m) 1775 { 1776 struct m_tag *mtag; 1777 1778 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1779 if (!mtag) { 1780 mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux), 1781 M_NOWAIT); 1782 if (mtag) { 1783 m_tag_prepend(m, mtag); 1784 bzero(mtag + 1, sizeof(struct ip6aux)); 1785 } 1786 } 1787 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1788 } 1789 1790 struct ip6aux * 1791 ip6_findaux(struct mbuf *m) 1792 { 1793 struct m_tag *mtag; 1794 1795 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1796 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1797 } 1798 1799 void 1800 ip6_delaux(struct mbuf *m) 1801 { 1802 struct m_tag *mtag; 1803 1804 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1805 if (mtag) 1806 m_tag_delete(m, mtag); 1807 } 1808 1809 /* 1810 * System control for IP6 1811 */ 1812 1813 u_char inet6ctlerrmap[PRC_NCMDS] = { 1814 0, 0, 0, 0, 1815 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1816 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1817 EMSGSIZE, EHOSTUNREACH, 0, 0, 1818 0, 0, 0, 0, 1819 ENOPROTOOPT 1820 }; 1821