ip6_input.c revision 197138
1258945Sroberto/*- 2258945Sroberto * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3258945Sroberto * All rights reserved. 4258945Sroberto * 5258945Sroberto * Redistribution and use in source and binary forms, with or without 6258945Sroberto * modification, are permitted provided that the following conditions 7258945Sroberto * are met: 8258945Sroberto * 1. Redistributions of source code must retain the above copyright 9258945Sroberto * notice, this list of conditions and the following disclaimer. 10258945Sroberto * 2. Redistributions in binary form must reproduce the above copyright 11258945Sroberto * notice, this list of conditions and the following disclaimer in the 12258945Sroberto * documentation and/or other materials provided with the distribution. 13258945Sroberto * 3. Neither the name of the project nor the names of its contributors 14258945Sroberto * may be used to endorse or promote products derived from this software 15258945Sroberto * without specific prior written permission. 16258945Sroberto * 17258945Sroberto * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18258945Sroberto * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19258945Sroberto * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20258945Sroberto * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21258945Sroberto * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22258945Sroberto * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23258945Sroberto * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24258945Sroberto * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25258945Sroberto * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26258945Sroberto * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27258945Sroberto * SUCH DAMAGE. 28258945Sroberto * 29258945Sroberto * $KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $ 30258945Sroberto */ 31258945Sroberto 32258945Sroberto/*- 33258945Sroberto * Copyright (c) 1982, 1986, 1988, 1993 34258945Sroberto * The Regents of the University of California. All rights reserved. 35258945Sroberto * 36258945Sroberto * Redistribution and use in source and binary forms, with or without 37258945Sroberto * modification, are permitted provided that the following conditions 38258945Sroberto * are met: 39258945Sroberto * 1. Redistributions of source code must retain the above copyright 40258945Sroberto * notice, this list of conditions and the following disclaimer. 41258945Sroberto * 2. Redistributions in binary form must reproduce the above copyright 42258945Sroberto * notice, this list of conditions and the following disclaimer in the 43258945Sroberto * documentation and/or other materials provided with the distribution. 44258945Sroberto * 4. Neither the name of the University nor the names of its contributors 45258945Sroberto * may be used to endorse or promote products derived from this software 46258945Sroberto * without specific prior written permission. 47258945Sroberto * 48258945Sroberto * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49258945Sroberto * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50258945Sroberto * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51258945Sroberto * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52258945Sroberto * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53258945Sroberto * 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: head/sys/netinet6/ip6_input.c 197138 2009-09-12 22:08:20Z hrs $"); 65 66#include "opt_inet.h" 67#include "opt_inet6.h" 68#include "opt_ipsec.h" 69 70#include <sys/param.h> 71#include <sys/systm.h> 72#include <sys/malloc.h> 73#include <sys/mbuf.h> 74#include <sys/proc.h> 75#include <sys/domain.h> 76#include <sys/protosw.h> 77#include <sys/socket.h> 78#include <sys/socketvar.h> 79#include <sys/errno.h> 80#include <sys/time.h> 81#include <sys/kernel.h> 82#include <sys/syslog.h> 83 84#include <net/if.h> 85#include <net/if_types.h> 86#include <net/if_dl.h> 87#include <net/route.h> 88#include <net/netisr.h> 89#include <net/pfil.h> 90#include <net/vnet.h> 91 92#include <netinet/in.h> 93#include <netinet/in_systm.h> 94#include <net/if_llatbl.h> 95#ifdef INET 96#include <netinet/ip.h> 97#include <netinet/ip_icmp.h> 98#endif /* INET */ 99#include <netinet/ip6.h> 100#include <netinet6/in6_var.h> 101#include <netinet6/ip6_var.h> 102#include <netinet/in_pcb.h> 103#include <netinet/icmp6.h> 104#include <netinet6/scope6_var.h> 105#include <netinet6/in6_ifattach.h> 106#include <netinet6/nd6.h> 107 108#ifdef IPSEC 109#include <netipsec/ipsec.h> 110#include <netinet6/ip6_ipsec.h> 111#include <netipsec/ipsec6.h> 112#endif /* IPSEC */ 113 114#include <netinet6/ip6protosw.h> 115 116extern struct domain inet6domain; 117 118u_char ip6_protox[IPPROTO_MAX]; 119 120static struct netisr_handler ip6_nh = { 121 .nh_name = "ip6", 122 .nh_handler = ip6_input, 123 .nh_proto = NETISR_IPV6, 124 .nh_policy = NETISR_POLICY_FLOW, 125}; 126 127VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 128VNET_DEFINE(struct ip6stat, ip6stat); 129 130VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch); 131VNET_DECLARE(int, dad_init); 132VNET_DECLARE(int, pmtu_expire); 133VNET_DECLARE(int, pmtu_probe); 134VNET_DECLARE(u_long, rip6_sendspace); 135VNET_DECLARE(u_long, rip6_recvspace); 136VNET_DECLARE(int, icmp6errppslim); 137VNET_DECLARE(int, icmp6_nodeinfo); 138VNET_DECLARE(int, udp6_sendspace); 139VNET_DECLARE(int, udp6_recvspace); 140 141#define V_in6_tmpaddrtimer_ch VNET(in6_tmpaddrtimer_ch) 142#define V_dad_init VNET(dad_init) 143#define V_pmtu_expire VNET(pmtu_expire) 144#define V_pmtu_probe VNET(pmtu_probe) 145#define V_rip6_sendspace VNET(rip6_sendspace) 146#define V_rip6_recvspace VNET(rip6_recvspace) 147#define V_icmp6errppslim VNET(icmp6errppslim) 148#define V_icmp6_nodeinfo VNET(icmp6_nodeinfo) 149#define V_udp6_sendspace VNET(udp6_sendspace) 150#define V_udp6_recvspace VNET(udp6_recvspace) 151 152struct rwlock in6_ifaddr_lock; 153RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 154 155struct pfil_head inet6_pfil_hook; 156 157static void ip6_init2(void *); 158static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *); 159static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 160#ifdef PULLDOWN_TEST 161static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int); 162#endif 163 164/* 165 * IP6 initialization: fill in IP6 protocol switch table. 166 * All protocols not implemented in kernel go to raw IP6 protocol handler. 167 */ 168void 169ip6_init(void) 170{ 171 struct ip6protosw *pr; 172 int i; 173 174 V_in6_maxmtu = 0; 175#ifdef IP6_AUTO_LINKLOCAL 176 V_ip6_auto_linklocal = IP6_AUTO_LINKLOCAL; 177#else 178 V_ip6_auto_linklocal = 1; /* enabled by default */ 179#endif 180 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 181 &V_ip6_auto_linklocal); 182 183#ifndef IPV6FORWARDING 184#ifdef GATEWAY6 185#define IPV6FORWARDING 1 /* forward IP6 packets not for us */ 186#else 187#define IPV6FORWARDING 0 /* don't forward IP6 packets not for us */ 188#endif /* GATEWAY6 */ 189#endif /* !IPV6FORWARDING */ 190 191#ifndef IPV6_SENDREDIRECTS 192#define IPV6_SENDREDIRECTS 1 193#endif 194 195 V_ip6_forwarding = IPV6FORWARDING; /* act as router? */ 196 V_ip6_sendredirects = IPV6_SENDREDIRECTS; 197 V_ip6_defhlim = IPV6_DEFHLIM; 198 V_ip6_defmcasthlim = IPV6_DEFAULT_MULTICAST_HOPS; 199 V_ip6_accept_rtadv = 0; 200 V_ip6_log_interval = 5; 201 V_ip6_hdrnestlimit = 15; /* How many header options will we process? */ 202 V_ip6_dad_count = 1; /* DupAddrDetectionTransmits */ 203 V_ip6_auto_flowlabel = 1; 204 V_ip6_use_deprecated = 1;/* allow deprecated addr (RFC2462 5.5.4) */ 205 V_ip6_rr_prune = 5; /* router renumbering prefix 206 * walk list every 5 sec. */ 207 V_ip6_mcast_pmtu = 0; /* enable pMTU discovery for multicast? */ 208 V_ip6_v6only = 1; 209 V_ip6_keepfaith = 0; 210 V_ip6_log_time = (time_t)0L; 211#ifdef IPSTEALTH 212 V_ip6stealth = 0; 213#endif 214 V_nd6_onlink_ns_rfc4861 = 0; /* allow 'on-link' nd6 NS (RFC 4861) */ 215 216 V_pmtu_expire = 60*10; 217 V_pmtu_probe = 60*2; 218 219 /* raw IP6 parameters */ 220 /* 221 * Nominal space allocated to a raw ip socket. 222 */ 223#define RIPV6SNDQ 8192 224#define RIPV6RCVQ 8192 225 V_rip6_sendspace = RIPV6SNDQ; 226 V_rip6_recvspace = RIPV6RCVQ; 227 228 /* ICMPV6 parameters */ 229 V_icmp6_rediraccept = 1; /* accept and process redirects */ 230 V_icmp6_redirtimeout = 10 * 60; /* 10 minutes */ 231 V_icmp6errppslim = 100; /* 100pps */ 232 /* control how to respond to NI queries */ 233 V_icmp6_nodeinfo = (ICMP6_NODEINFO_FQDNOK|ICMP6_NODEINFO_NODEADDROK); 234 235 /* UDP on IP6 parameters */ 236 V_udp6_sendspace = 9216; /* really max datagram size */ 237 V_udp6_recvspace = 40 * (1024 + sizeof(struct sockaddr_in6)); 238 /* 40 1K datagrams */ 239 V_dad_init = 0; 240 241 TAILQ_INIT(&V_in6_ifaddrhead); 242 243 scope6_init(); 244 addrsel_policy_init(); 245 nd6_init(); 246 frag6_init(); 247 248 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 249 250 /* Skip global initialization stuff for non-default instances. */ 251 if (!IS_DEFAULT_VNET(curvnet)) 252 return; 253 254#ifdef DIAGNOSTIC 255 if (sizeof(struct protosw) != sizeof(struct ip6protosw)) 256 panic("sizeof(protosw) != sizeof(ip6protosw)"); 257#endif 258 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 259 if (pr == NULL) 260 panic("ip6_init"); 261 262 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 263 for (i = 0; i < IPPROTO_MAX; i++) 264 ip6_protox[i] = pr - inet6sw; 265 /* 266 * Cycle through IP protocols and put them into the appropriate place 267 * in ip6_protox[]. 268 */ 269 for (pr = (struct ip6protosw *)inet6domain.dom_protosw; 270 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) 271 if (pr->pr_domain->dom_family == PF_INET6 && 272 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 273 /* Be careful to only index valid IP protocols. */ 274 if (pr->pr_protocol < IPPROTO_MAX) 275 ip6_protox[pr->pr_protocol] = pr - inet6sw; 276 } 277 278 /* Initialize packet filter hooks. */ 279 inet6_pfil_hook.ph_type = PFIL_TYPE_AF; 280 inet6_pfil_hook.ph_af = AF_INET6; 281 if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) 282 printf("%s: WARNING: unable to register pfil hook, " 283 "error %d\n", __func__, i); 284 285 netisr_register(&ip6_nh); 286} 287 288#ifdef VIMAGE 289void 290ip6_destroy() 291{ 292 293 nd6_destroy(); 294 callout_drain(&V_in6_tmpaddrtimer_ch); 295} 296#endif 297 298static int 299ip6_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 316static void 317ip6_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 */ 325SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL); 326 327void 328ip6_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(&inet6_pfil_hook)) 519 goto passin; 520 521 if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL)) 522 return; 523 if (m == NULL) /* consumed by filter */ 524 return; 525 ip6 = mtod(m, struct ip6_hdr *); 526 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 527 528passin: 529 /* 530 * Disambiguate address scope zones (if there is ambiguity). 531 * We first make sure that the original source or destination address 532 * is not in our internal form for scoped addresses. Such addresses 533 * are not necessarily invalid spec-wise, but we cannot accept them due 534 * to the usage conflict. 535 * in6_setscope() then also checks and rejects the cases where src or 536 * dst are the loopback address and the receiving interface 537 * is not loopback. 538 */ 539 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 540 V_ip6stat.ip6s_badscope++; /* XXX */ 541 goto bad; 542 } 543 if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) || 544 in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) { 545 V_ip6stat.ip6s_badscope++; 546 goto bad; 547 } 548 549 /* 550 * Multicast check. Assume packet is for us to avoid 551 * prematurely taking locks. 552 */ 553 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 554 ours = 1; 555 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast); 556 deliverifp = m->m_pkthdr.rcvif; 557 goto hbhcheck; 558 } 559 560 /* 561 * Unicast check 562 */ 563 564 bzero(&dst6, sizeof(dst6)); 565 dst6.sin6_family = AF_INET6; 566 dst6.sin6_len = sizeof(struct sockaddr_in6); 567 dst6.sin6_addr = ip6->ip6_dst; 568 ifp = m->m_pkthdr.rcvif; 569 IF_AFDATA_LOCK(ifp); 570 lle = lla_lookup(LLTABLE6(ifp), 0, 571 (struct sockaddr *)&dst6); 572 IF_AFDATA_UNLOCK(ifp); 573 if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) { 574 ours = 1; 575 deliverifp = ifp; 576 LLE_RUNLOCK(lle); 577 goto hbhcheck; 578 } 579 if (lle != NULL) 580 LLE_RUNLOCK(lle); 581 582 dst = &rin6.ro_dst; 583 dst->sin6_len = sizeof(struct sockaddr_in6); 584 dst->sin6_family = AF_INET6; 585 dst->sin6_addr = ip6->ip6_dst; 586 rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0); 587 if (rin6.ro_rt) 588 RT_UNLOCK(rin6.ro_rt); 589 590#define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key)) 591 592 /* 593 * Accept the packet if the forwarding interface to the destination 594 * according to the routing table is the loopback interface, 595 * unless the associated route has a gateway. 596 * Note that this approach causes to accept a packet if there is a 597 * route to the loopback interface for the destination of the packet. 598 * But we think it's even useful in some situations, e.g. when using 599 * a special daemon which wants to intercept the packet. 600 * 601 * XXX: some OSes automatically make a cloned route for the destination 602 * of an outgoing packet. If the outgoing interface of the packet 603 * is a loopback one, the kernel would consider the packet to be 604 * accepted, even if we have no such address assinged on the interface. 605 * We check the cloned flag of the route entry to reject such cases, 606 * assuming that route entries for our own addresses are not made by 607 * cloning (it should be true because in6_addloop explicitly installs 608 * the host route). However, we might have to do an explicit check 609 * while it would be less efficient. Or, should we rather install a 610 * reject route for such a case? 611 */ 612 if (rin6.ro_rt && 613 (rin6.ro_rt->rt_flags & 614 (RTF_HOST|RTF_GATEWAY)) == RTF_HOST && 615#ifdef RTF_WASCLONED 616 !(rin6.ro_rt->rt_flags & RTF_WASCLONED) && 617#endif 618#ifdef RTF_CLONED 619 !(rin6.ro_rt->rt_flags & RTF_CLONED) && 620#endif 621#if 0 622 /* 623 * The check below is redundant since the comparison of 624 * the destination and the key of the rtentry has 625 * already done through looking up the routing table. 626 */ 627 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 628 &rt6_key(rin6.ro_rt)->sin6_addr) 629#endif 630 rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) { 631 int free_ia6 = 0; 632 struct in6_ifaddr *ia6; 633 634 /* 635 * found the loopback route to the interface address 636 */ 637 if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) { 638 struct sockaddr_in6 dest6; 639 640 bzero(&dest6, sizeof(dest6)); 641 dest6.sin6_family = AF_INET6; 642 dest6.sin6_len = sizeof(dest6); 643 dest6.sin6_addr = ip6->ip6_dst; 644 ia6 = (struct in6_ifaddr *) 645 ifa_ifwithaddr((struct sockaddr *)&dest6); 646 if (ia6 == NULL) 647 goto bad; 648 free_ia6 = 1; 649 } 650 else 651 ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa; 652 653 /* 654 * record address information into m_tag. 655 */ 656 (void)ip6_setdstifaddr(m, ia6); 657 658 /* 659 * packets to a tentative, duplicated, or somehow invalid 660 * address must not be accepted. 661 */ 662 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) { 663 /* this address is ready */ 664 ours = 1; 665 deliverifp = ia6->ia_ifp; /* correct? */ 666 /* Count the packet in the ip address stats */ 667 ia6->ia_ifa.if_ipackets++; 668 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len; 669 if (ia6 != NULL && free_ia6 != 0) 670 ifa_free(&ia6->ia_ifa); 671 goto hbhcheck; 672 } else { 673 char ip6bufs[INET6_ADDRSTRLEN]; 674 char ip6bufd[INET6_ADDRSTRLEN]; 675 /* address is not ready, so discard the packet. */ 676 nd6log((LOG_INFO, 677 "ip6_input: packet to an unready address %s->%s\n", 678 ip6_sprintf(ip6bufs, &ip6->ip6_src), 679 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 680 681 if (ia6 != NULL && free_ia6 != 0) 682 ifa_free(&ia6->ia_ifa); 683 goto bad; 684 } 685 } 686 687 /* 688 * FAITH (Firewall Aided Internet Translator) 689 */ 690 if (V_ip6_keepfaith) { 691 if (rin6.ro_rt && rin6.ro_rt->rt_ifp && 692 rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) { 693 /* XXX do we need more sanity checks? */ 694 ours = 1; 695 deliverifp = rin6.ro_rt->rt_ifp; /* faith */ 696 goto hbhcheck; 697 } 698 } 699 700 /* 701 * Now there is no reason to process the packet if it's not our own 702 * and we're not a router. 703 */ 704 if (!V_ip6_forwarding) { 705 V_ip6stat.ip6s_cantforward++; 706 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 707 goto bad; 708 } 709 710 hbhcheck: 711 /* 712 * record address information into m_tag, if we don't have one yet. 713 * note that we are unable to record it, if the address is not listed 714 * as our interface address (e.g. multicast addresses, addresses 715 * within FAITH prefixes and such). 716 */ 717 if (deliverifp && !ip6_getdstifaddr(m)) { 718 struct in6_ifaddr *ia6; 719 720 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst); 721 if (ia6) { 722 if (!ip6_setdstifaddr(m, ia6)) { 723 /* 724 * XXX maybe we should drop the packet here, 725 * as we could not provide enough information 726 * to the upper layers. 727 */ 728 } 729 ifa_free(&ia6->ia_ifa); 730 } 731 } 732 733 /* 734 * Process Hop-by-Hop options header if it's contained. 735 * m may be modified in ip6_hopopts_input(). 736 * If a JumboPayload option is included, plen will also be modified. 737 */ 738 plen = (u_int32_t)ntohs(ip6->ip6_plen); 739 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 740 struct ip6_hbh *hbh; 741 742 if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) { 743#if 0 /*touches NULL pointer*/ 744 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 745#endif 746 goto out; /* m have already been freed */ 747 } 748 749 /* adjust pointer */ 750 ip6 = mtod(m, struct ip6_hdr *); 751 752 /* 753 * if the payload length field is 0 and the next header field 754 * indicates Hop-by-Hop Options header, then a Jumbo Payload 755 * option MUST be included. 756 */ 757 if (ip6->ip6_plen == 0 && plen == 0) { 758 /* 759 * Note that if a valid jumbo payload option is 760 * contained, ip6_hopopts_input() must set a valid 761 * (non-zero) payload length to the variable plen. 762 */ 763 V_ip6stat.ip6s_badoptions++; 764 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 765 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 766 icmp6_error(m, ICMP6_PARAM_PROB, 767 ICMP6_PARAMPROB_HEADER, 768 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 769 goto out; 770 } 771#ifndef PULLDOWN_TEST 772 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 773 hbh = (struct ip6_hbh *)(ip6 + 1); 774#else 775 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 776 sizeof(struct ip6_hbh)); 777 if (hbh == NULL) { 778 V_ip6stat.ip6s_tooshort++; 779 goto out; 780 } 781#endif 782 nxt = hbh->ip6h_nxt; 783 784 /* 785 * If we are acting as a router and the packet contains a 786 * router alert option, see if we know the option value. 787 * Currently, we only support the option value for MLD, in which 788 * case we should pass the packet to the multicast routing 789 * daemon. 790 */ 791 if (rtalert != ~0) { 792 switch (rtalert) { 793 case IP6OPT_RTALERT_MLD: 794 if (V_ip6_forwarding) 795 ours = 1; 796 break; 797 default: 798 /* 799 * RFC2711 requires unrecognized values must be 800 * silently ignored. 801 */ 802 break; 803 } 804 } 805 } else 806 nxt = ip6->ip6_nxt; 807 808 /* 809 * Check that the amount of data in the buffers 810 * is as at least much as the IPv6 header would have us expect. 811 * Trim mbufs if longer than we expect. 812 * Drop packet if shorter than we expect. 813 */ 814 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 815 V_ip6stat.ip6s_tooshort++; 816 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 817 goto bad; 818 } 819 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 820 if (m->m_len == m->m_pkthdr.len) { 821 m->m_len = sizeof(struct ip6_hdr) + plen; 822 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 823 } else 824 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 825 } 826 827 /* 828 * Forward if desirable. 829 */ 830 if (V_ip6_mrouter && 831 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 832 /* 833 * If we are acting as a multicast router, all 834 * incoming multicast packets are passed to the 835 * kernel-level multicast forwarding function. 836 * The packet is returned (relatively) intact; if 837 * ip6_mforward() returns a non-zero value, the packet 838 * must be discarded, else it may be accepted below. 839 * 840 * XXX TODO: Check hlim and multicast scope here to avoid 841 * unnecessarily calling into ip6_mforward(). 842 */ 843 if (ip6_mforward && 844 ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) { 845 IP6STAT_INC(ip6s_cantforward); 846 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 847 goto bad; 848 } 849 } else if (!ours) { 850 ip6_forward(m, srcrt); 851 goto out; 852 } 853 854 ip6 = mtod(m, struct ip6_hdr *); 855 856 /* 857 * Malicious party may be able to use IPv4 mapped addr to confuse 858 * tcp/udp stack and bypass security checks (act as if it was from 859 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 860 * 861 * For SIIT end node behavior, you may want to disable the check. 862 * However, you will become vulnerable to attacks using IPv4 mapped 863 * source. 864 */ 865 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 866 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 867 V_ip6stat.ip6s_badscope++; 868 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 869 goto bad; 870 } 871 872 /* 873 * Tell launch routine the next header 874 */ 875 V_ip6stat.ip6s_delivered++; 876 in6_ifstat_inc(deliverifp, ifs6_in_deliver); 877 nest = 0; 878 879 while (nxt != IPPROTO_DONE) { 880 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 881 V_ip6stat.ip6s_toomanyhdr++; 882 goto bad; 883 } 884 885 /* 886 * protection against faulty packet - there should be 887 * more sanity checks in header chain processing. 888 */ 889 if (m->m_pkthdr.len < off) { 890 V_ip6stat.ip6s_tooshort++; 891 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 892 goto bad; 893 } 894 895#ifdef IPSEC 896 /* 897 * enforce IPsec policy checking if we are seeing last header. 898 * note that we do not visit this with protocols with pcb layer 899 * code - like udp/tcp/raw ip. 900 */ 901 if (ip6_ipsec_input(m, nxt)) 902 goto bad; 903#endif /* IPSEC */ 904 905 /* 906 * Use mbuf flags to propagate Router Alert option to 907 * ICMPv6 layer, as hop-by-hop options have been stripped. 908 */ 909 if (nxt == IPPROTO_ICMPV6 && rtalert != ~0) 910 m->m_flags |= M_RTALERT_MLD; 911 912 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 913 } 914 goto out; 915bad: 916 m_freem(m); 917out: 918 if (rin6.ro_rt) 919 RTFREE(rin6.ro_rt); 920} 921 922/* 923 * set/grab in6_ifaddr correspond to IPv6 destination address. 924 * XXX backward compatibility wrapper 925 * 926 * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag, 927 * and then bump it when the tag is copied, and release it when the tag is 928 * freed. Unfortunately, m_tags don't support deep copies (yet), so instead 929 * we just bump the ia refcount when we receive it. This should be fixed. 930 */ 931static struct ip6aux * 932ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6) 933{ 934 struct ip6aux *ip6a; 935 936 ip6a = ip6_addaux(m); 937 if (ip6a) 938 ip6a->ip6a_dstia6 = ia6; 939 return ip6a; /* NULL if failed to set */ 940} 941 942struct in6_ifaddr * 943ip6_getdstifaddr(struct mbuf *m) 944{ 945 struct ip6aux *ip6a; 946 struct in6_ifaddr *ia; 947 948 ip6a = ip6_findaux(m); 949 if (ip6a) { 950 ia = ip6a->ip6a_dstia6; 951 ifa_ref(&ia->ia_ifa); 952 return ia; 953 } else 954 return NULL; 955} 956 957/* 958 * Hop-by-Hop options header processing. If a valid jumbo payload option is 959 * included, the real payload length will be stored in plenp. 960 * 961 * rtalertp - XXX: should be stored more smart way 962 */ 963static int 964ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 965 struct mbuf **mp, int *offp) 966{ 967 struct mbuf *m = *mp; 968 int off = *offp, hbhlen; 969 struct ip6_hbh *hbh; 970 u_int8_t *opt; 971 972 /* validation of the length of the header */ 973#ifndef PULLDOWN_TEST 974 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); 975 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 976 hbhlen = (hbh->ip6h_len + 1) << 3; 977 978 IP6_EXTHDR_CHECK(m, off, hbhlen, -1); 979 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 980#else 981 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, 982 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); 983 if (hbh == NULL) { 984 V_ip6stat.ip6s_tooshort++; 985 return -1; 986 } 987 hbhlen = (hbh->ip6h_len + 1) << 3; 988 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 989 hbhlen); 990 if (hbh == NULL) { 991 V_ip6stat.ip6s_tooshort++; 992 return -1; 993 } 994#endif 995 off += hbhlen; 996 hbhlen -= sizeof(struct ip6_hbh); 997 opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh); 998 999 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 1000 hbhlen, rtalertp, plenp) < 0) 1001 return (-1); 1002 1003 *offp = off; 1004 *mp = m; 1005 return (0); 1006} 1007 1008/* 1009 * Search header for all Hop-by-hop options and process each option. 1010 * This function is separate from ip6_hopopts_input() in order to 1011 * handle a case where the sending node itself process its hop-by-hop 1012 * options header. In such a case, the function is called from ip6_output(). 1013 * 1014 * The function assumes that hbh header is located right after the IPv6 header 1015 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1016 * opthead + hbhlen is located in continuous memory region. 1017 */ 1018int 1019ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1020 u_int32_t *rtalertp, u_int32_t *plenp) 1021{ 1022 struct ip6_hdr *ip6; 1023 int optlen = 0; 1024 u_int8_t *opt = opthead; 1025 u_int16_t rtalert_val; 1026 u_int32_t jumboplen; 1027 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1028 1029 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1030 switch (*opt) { 1031 case IP6OPT_PAD1: 1032 optlen = 1; 1033 break; 1034 case IP6OPT_PADN: 1035 if (hbhlen < IP6OPT_MINLEN) { 1036 V_ip6stat.ip6s_toosmall++; 1037 goto bad; 1038 } 1039 optlen = *(opt + 1) + 2; 1040 break; 1041 case IP6OPT_ROUTER_ALERT: 1042 /* XXX may need check for alignment */ 1043 if (hbhlen < IP6OPT_RTALERT_LEN) { 1044 V_ip6stat.ip6s_toosmall++; 1045 goto bad; 1046 } 1047 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1048 /* XXX stat */ 1049 icmp6_error(m, ICMP6_PARAM_PROB, 1050 ICMP6_PARAMPROB_HEADER, 1051 erroff + opt + 1 - opthead); 1052 return (-1); 1053 } 1054 optlen = IP6OPT_RTALERT_LEN; 1055 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1056 *rtalertp = ntohs(rtalert_val); 1057 break; 1058 case IP6OPT_JUMBO: 1059 /* XXX may need check for alignment */ 1060 if (hbhlen < IP6OPT_JUMBO_LEN) { 1061 V_ip6stat.ip6s_toosmall++; 1062 goto bad; 1063 } 1064 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1065 /* XXX stat */ 1066 icmp6_error(m, ICMP6_PARAM_PROB, 1067 ICMP6_PARAMPROB_HEADER, 1068 erroff + opt + 1 - opthead); 1069 return (-1); 1070 } 1071 optlen = IP6OPT_JUMBO_LEN; 1072 1073 /* 1074 * IPv6 packets that have non 0 payload length 1075 * must not contain a jumbo payload option. 1076 */ 1077 ip6 = mtod(m, struct ip6_hdr *); 1078 if (ip6->ip6_plen) { 1079 V_ip6stat.ip6s_badoptions++; 1080 icmp6_error(m, ICMP6_PARAM_PROB, 1081 ICMP6_PARAMPROB_HEADER, 1082 erroff + opt - opthead); 1083 return (-1); 1084 } 1085 1086 /* 1087 * We may see jumbolen in unaligned location, so 1088 * we'd need to perform bcopy(). 1089 */ 1090 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1091 jumboplen = (u_int32_t)htonl(jumboplen); 1092 1093#if 1 1094 /* 1095 * if there are multiple jumbo payload options, 1096 * *plenp will be non-zero and the packet will be 1097 * rejected. 1098 * the behavior may need some debate in ipngwg - 1099 * multiple options does not make sense, however, 1100 * there's no explicit mention in specification. 1101 */ 1102 if (*plenp != 0) { 1103 V_ip6stat.ip6s_badoptions++; 1104 icmp6_error(m, ICMP6_PARAM_PROB, 1105 ICMP6_PARAMPROB_HEADER, 1106 erroff + opt + 2 - opthead); 1107 return (-1); 1108 } 1109#endif 1110 1111 /* 1112 * jumbo payload length must be larger than 65535. 1113 */ 1114 if (jumboplen <= IPV6_MAXPACKET) { 1115 V_ip6stat.ip6s_badoptions++; 1116 icmp6_error(m, ICMP6_PARAM_PROB, 1117 ICMP6_PARAMPROB_HEADER, 1118 erroff + opt + 2 - opthead); 1119 return (-1); 1120 } 1121 *plenp = jumboplen; 1122 1123 break; 1124 default: /* unknown option */ 1125 if (hbhlen < IP6OPT_MINLEN) { 1126 V_ip6stat.ip6s_toosmall++; 1127 goto bad; 1128 } 1129 optlen = ip6_unknown_opt(opt, m, 1130 erroff + opt - opthead); 1131 if (optlen == -1) 1132 return (-1); 1133 optlen += 2; 1134 break; 1135 } 1136 } 1137 1138 return (0); 1139 1140 bad: 1141 m_freem(m); 1142 return (-1); 1143} 1144 1145/* 1146 * Unknown option processing. 1147 * The third argument `off' is the offset from the IPv6 header to the option, 1148 * which is necessary if the IPv6 header the and option header and IPv6 header 1149 * is not continuous in order to return an ICMPv6 error. 1150 */ 1151int 1152ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1153{ 1154 struct ip6_hdr *ip6; 1155 1156 switch (IP6OPT_TYPE(*optp)) { 1157 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1158 return ((int)*(optp + 1)); 1159 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1160 m_freem(m); 1161 return (-1); 1162 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1163 V_ip6stat.ip6s_badoptions++; 1164 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1165 return (-1); 1166 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1167 V_ip6stat.ip6s_badoptions++; 1168 ip6 = mtod(m, struct ip6_hdr *); 1169 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1170 (m->m_flags & (M_BCAST|M_MCAST))) 1171 m_freem(m); 1172 else 1173 icmp6_error(m, ICMP6_PARAM_PROB, 1174 ICMP6_PARAMPROB_OPTION, off); 1175 return (-1); 1176 } 1177 1178 m_freem(m); /* XXX: NOTREACHED */ 1179 return (-1); 1180} 1181 1182/* 1183 * Create the "control" list for this pcb. 1184 * These functions will not modify mbuf chain at all. 1185 * 1186 * With KAME mbuf chain restriction: 1187 * The routine will be called from upper layer handlers like tcp6_input(). 1188 * Thus the routine assumes that the caller (tcp6_input) have already 1189 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the 1190 * very first mbuf on the mbuf chain. 1191 * 1192 * ip6_savecontrol_v4 will handle those options that are possible to be 1193 * set on a v4-mapped socket. 1194 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1195 * options and handle the v6-only ones itself. 1196 */ 1197struct mbuf ** 1198ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1199 int *v4only) 1200{ 1201 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1202 1203#ifdef SO_TIMESTAMP 1204 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1205 struct timeval tv; 1206 1207 microtime(&tv); 1208 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1209 SCM_TIMESTAMP, SOL_SOCKET); 1210 if (*mp) 1211 mp = &(*mp)->m_next; 1212 } 1213#endif 1214 1215 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1216 if (v4only != NULL) 1217 *v4only = 1; 1218 return (mp); 1219 } 1220 1221#define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1222 /* RFC 2292 sec. 5 */ 1223 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1224 struct in6_pktinfo pi6; 1225 1226 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1227 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1228 pi6.ipi6_ifindex = 1229 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1230 1231 *mp = sbcreatecontrol((caddr_t) &pi6, 1232 sizeof(struct in6_pktinfo), 1233 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1234 if (*mp) 1235 mp = &(*mp)->m_next; 1236 } 1237 1238 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1239 int hlim = ip6->ip6_hlim & 0xff; 1240 1241 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1242 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1243 IPPROTO_IPV6); 1244 if (*mp) 1245 mp = &(*mp)->m_next; 1246 } 1247 1248 if (v4only != NULL) 1249 *v4only = 0; 1250 return (mp); 1251} 1252 1253void 1254ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) 1255{ 1256 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1257 int v4only = 0; 1258 1259 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); 1260 if (v4only) 1261 return; 1262 1263 if ((in6p->inp_flags & IN6P_TCLASS) != 0) { 1264 u_int32_t flowinfo; 1265 int tclass; 1266 1267 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1268 flowinfo >>= 20; 1269 1270 tclass = flowinfo & 0xff; 1271 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass), 1272 IPV6_TCLASS, IPPROTO_IPV6); 1273 if (*mp) 1274 mp = &(*mp)->m_next; 1275 } 1276 1277 /* 1278 * IPV6_HOPOPTS socket option. Recall that we required super-user 1279 * privilege for the option (see ip6_ctloutput), but it might be too 1280 * strict, since there might be some hop-by-hop options which can be 1281 * returned to normal user. 1282 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1283 */ 1284 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { 1285 /* 1286 * Check if a hop-by-hop options header is contatined in the 1287 * received packet, and if so, store the options as ancillary 1288 * data. Note that a hop-by-hop options header must be 1289 * just after the IPv6 header, which is assured through the 1290 * IPv6 input processing. 1291 */ 1292 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1293 struct ip6_hbh *hbh; 1294 int hbhlen = 0; 1295#ifdef PULLDOWN_TEST 1296 struct mbuf *ext; 1297#endif 1298 1299#ifndef PULLDOWN_TEST 1300 hbh = (struct ip6_hbh *)(ip6 + 1); 1301 hbhlen = (hbh->ip6h_len + 1) << 3; 1302#else 1303 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), 1304 ip6->ip6_nxt); 1305 if (ext == NULL) { 1306 V_ip6stat.ip6s_tooshort++; 1307 return; 1308 } 1309 hbh = mtod(ext, struct ip6_hbh *); 1310 hbhlen = (hbh->ip6h_len + 1) << 3; 1311 if (hbhlen != ext->m_len) { 1312 m_freem(ext); 1313 V_ip6stat.ip6s_tooshort++; 1314 return; 1315 } 1316#endif 1317 1318 /* 1319 * XXX: We copy the whole header even if a 1320 * jumbo payload option is included, the option which 1321 * is to be removed before returning according to 1322 * RFC2292. 1323 * Note: this constraint is removed in RFC3542 1324 */ 1325 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1326 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1327 IPPROTO_IPV6); 1328 if (*mp) 1329 mp = &(*mp)->m_next; 1330#ifdef PULLDOWN_TEST 1331 m_freem(ext); 1332#endif 1333 } 1334 } 1335 1336 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1337 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1338 1339 /* 1340 * Search for destination options headers or routing 1341 * header(s) through the header chain, and stores each 1342 * header as ancillary data. 1343 * Note that the order of the headers remains in 1344 * the chain of ancillary data. 1345 */ 1346 while (1) { /* is explicit loop prevention necessary? */ 1347 struct ip6_ext *ip6e = NULL; 1348 int elen; 1349#ifdef PULLDOWN_TEST 1350 struct mbuf *ext = NULL; 1351#endif 1352 1353 /* 1354 * if it is not an extension header, don't try to 1355 * pull it from the chain. 1356 */ 1357 switch (nxt) { 1358 case IPPROTO_DSTOPTS: 1359 case IPPROTO_ROUTING: 1360 case IPPROTO_HOPOPTS: 1361 case IPPROTO_AH: /* is it possible? */ 1362 break; 1363 default: 1364 goto loopend; 1365 } 1366 1367#ifndef PULLDOWN_TEST 1368 if (off + sizeof(*ip6e) > m->m_len) 1369 goto loopend; 1370 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1371 if (nxt == IPPROTO_AH) 1372 elen = (ip6e->ip6e_len + 2) << 2; 1373 else 1374 elen = (ip6e->ip6e_len + 1) << 3; 1375 if (off + elen > m->m_len) 1376 goto loopend; 1377#else 1378 ext = ip6_pullexthdr(m, off, nxt); 1379 if (ext == NULL) { 1380 V_ip6stat.ip6s_tooshort++; 1381 return; 1382 } 1383 ip6e = mtod(ext, struct ip6_ext *); 1384 if (nxt == IPPROTO_AH) 1385 elen = (ip6e->ip6e_len + 2) << 2; 1386 else 1387 elen = (ip6e->ip6e_len + 1) << 3; 1388 if (elen != ext->m_len) { 1389 m_freem(ext); 1390 V_ip6stat.ip6s_tooshort++; 1391 return; 1392 } 1393#endif 1394 1395 switch (nxt) { 1396 case IPPROTO_DSTOPTS: 1397 if (!(in6p->inp_flags & IN6P_DSTOPTS)) 1398 break; 1399 1400 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1401 IS2292(in6p, 1402 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1403 IPPROTO_IPV6); 1404 if (*mp) 1405 mp = &(*mp)->m_next; 1406 break; 1407 case IPPROTO_ROUTING: 1408 if (!in6p->inp_flags & IN6P_RTHDR) 1409 break; 1410 1411 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1412 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), 1413 IPPROTO_IPV6); 1414 if (*mp) 1415 mp = &(*mp)->m_next; 1416 break; 1417 case IPPROTO_HOPOPTS: 1418 case IPPROTO_AH: /* is it possible? */ 1419 break; 1420 1421 default: 1422 /* 1423 * other cases have been filtered in the above. 1424 * none will visit this case. here we supply 1425 * the code just in case (nxt overwritten or 1426 * other cases). 1427 */ 1428#ifdef PULLDOWN_TEST 1429 m_freem(ext); 1430#endif 1431 goto loopend; 1432 1433 } 1434 1435 /* proceed with the next header. */ 1436 off += elen; 1437 nxt = ip6e->ip6e_nxt; 1438 ip6e = NULL; 1439#ifdef PULLDOWN_TEST 1440 m_freem(ext); 1441 ext = NULL; 1442#endif 1443 } 1444 loopend: 1445 ; 1446 } 1447} 1448#undef IS2292 1449 1450void 1451ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu) 1452{ 1453 struct socket *so; 1454 struct mbuf *m_mtu; 1455 struct ip6_mtuinfo mtuctl; 1456 1457 so = in6p->inp_socket; 1458 1459 if (mtu == NULL) 1460 return; 1461 1462#ifdef DIAGNOSTIC 1463 if (so == NULL) /* I believe this is impossible */ 1464 panic("ip6_notify_pmtu: socket is NULL"); 1465#endif 1466 1467 bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */ 1468 mtuctl.ip6m_mtu = *mtu; 1469 mtuctl.ip6m_addr = *dst; 1470 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1471 return; 1472 1473 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1474 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1475 return; 1476 1477 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1478 == 0) { 1479 m_freem(m_mtu); 1480 /* XXX: should count statistics */ 1481 } else 1482 sorwakeup(so); 1483 1484 return; 1485} 1486 1487#ifdef PULLDOWN_TEST 1488/* 1489 * pull single extension header from mbuf chain. returns single mbuf that 1490 * contains the result, or NULL on error. 1491 */ 1492static struct mbuf * 1493ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) 1494{ 1495 struct ip6_ext ip6e; 1496 size_t elen; 1497 struct mbuf *n; 1498 1499#ifdef DIAGNOSTIC 1500 switch (nxt) { 1501 case IPPROTO_DSTOPTS: 1502 case IPPROTO_ROUTING: 1503 case IPPROTO_HOPOPTS: 1504 case IPPROTO_AH: /* is it possible? */ 1505 break; 1506 default: 1507 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); 1508 } 1509#endif 1510 1511 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1512 if (nxt == IPPROTO_AH) 1513 elen = (ip6e.ip6e_len + 2) << 2; 1514 else 1515 elen = (ip6e.ip6e_len + 1) << 3; 1516 1517 MGET(n, M_DONTWAIT, MT_DATA); 1518 if (n && elen >= MLEN) { 1519 MCLGET(n, M_DONTWAIT); 1520 if ((n->m_flags & M_EXT) == 0) { 1521 m_free(n); 1522 n = NULL; 1523 } 1524 } 1525 if (!n) 1526 return NULL; 1527 1528 n->m_len = 0; 1529 if (elen >= M_TRAILINGSPACE(n)) { 1530 m_free(n); 1531 return NULL; 1532 } 1533 1534 m_copydata(m, off, elen, mtod(n, caddr_t)); 1535 n->m_len = elen; 1536 return n; 1537} 1538#endif 1539 1540/* 1541 * Get pointer to the previous header followed by the header 1542 * currently processed. 1543 * XXX: This function supposes that 1544 * M includes all headers, 1545 * the next header field and the header length field of each header 1546 * are valid, and 1547 * the sum of each header length equals to OFF. 1548 * Because of these assumptions, this function must be called very 1549 * carefully. Moreover, it will not be used in the near future when 1550 * we develop `neater' mechanism to process extension headers. 1551 */ 1552char * 1553ip6_get_prevhdr(struct mbuf *m, int off) 1554{ 1555 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1556 1557 if (off == sizeof(struct ip6_hdr)) 1558 return (&ip6->ip6_nxt); 1559 else { 1560 int len, nxt; 1561 struct ip6_ext *ip6e = NULL; 1562 1563 nxt = ip6->ip6_nxt; 1564 len = sizeof(struct ip6_hdr); 1565 while (len < off) { 1566 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len); 1567 1568 switch (nxt) { 1569 case IPPROTO_FRAGMENT: 1570 len += sizeof(struct ip6_frag); 1571 break; 1572 case IPPROTO_AH: 1573 len += (ip6e->ip6e_len + 2) << 2; 1574 break; 1575 default: 1576 len += (ip6e->ip6e_len + 1) << 3; 1577 break; 1578 } 1579 nxt = ip6e->ip6e_nxt; 1580 } 1581 if (ip6e) 1582 return (&ip6e->ip6e_nxt); 1583 else 1584 return NULL; 1585 } 1586} 1587 1588/* 1589 * get next header offset. m will be retained. 1590 */ 1591int 1592ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp) 1593{ 1594 struct ip6_hdr ip6; 1595 struct ip6_ext ip6e; 1596 struct ip6_frag fh; 1597 1598 /* just in case */ 1599 if (m == NULL) 1600 panic("ip6_nexthdr: m == NULL"); 1601 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1602 return -1; 1603 1604 switch (proto) { 1605 case IPPROTO_IPV6: 1606 if (m->m_pkthdr.len < off + sizeof(ip6)) 1607 return -1; 1608 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1609 if (nxtp) 1610 *nxtp = ip6.ip6_nxt; 1611 off += sizeof(ip6); 1612 return off; 1613 1614 case IPPROTO_FRAGMENT: 1615 /* 1616 * terminate parsing if it is not the first fragment, 1617 * it does not make sense to parse through it. 1618 */ 1619 if (m->m_pkthdr.len < off + sizeof(fh)) 1620 return -1; 1621 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1622 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1623 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1624 return -1; 1625 if (nxtp) 1626 *nxtp = fh.ip6f_nxt; 1627 off += sizeof(struct ip6_frag); 1628 return off; 1629 1630 case IPPROTO_AH: 1631 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1632 return -1; 1633 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1634 if (nxtp) 1635 *nxtp = ip6e.ip6e_nxt; 1636 off += (ip6e.ip6e_len + 2) << 2; 1637 return off; 1638 1639 case IPPROTO_HOPOPTS: 1640 case IPPROTO_ROUTING: 1641 case IPPROTO_DSTOPTS: 1642 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1643 return -1; 1644 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1645 if (nxtp) 1646 *nxtp = ip6e.ip6e_nxt; 1647 off += (ip6e.ip6e_len + 1) << 3; 1648 return off; 1649 1650 case IPPROTO_NONE: 1651 case IPPROTO_ESP: 1652 case IPPROTO_IPCOMP: 1653 /* give up */ 1654 return -1; 1655 1656 default: 1657 return -1; 1658 } 1659 1660 return -1; 1661} 1662 1663/* 1664 * get offset for the last header in the chain. m will be kept untainted. 1665 */ 1666int 1667ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp) 1668{ 1669 int newoff; 1670 int nxt; 1671 1672 if (!nxtp) { 1673 nxt = -1; 1674 nxtp = &nxt; 1675 } 1676 while (1) { 1677 newoff = ip6_nexthdr(m, off, proto, nxtp); 1678 if (newoff < 0) 1679 return off; 1680 else if (newoff < off) 1681 return -1; /* invalid */ 1682 else if (newoff == off) 1683 return newoff; 1684 1685 off = newoff; 1686 proto = *nxtp; 1687 } 1688} 1689 1690struct ip6aux * 1691ip6_addaux(struct mbuf *m) 1692{ 1693 struct m_tag *mtag; 1694 1695 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1696 if (!mtag) { 1697 mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux), 1698 M_NOWAIT); 1699 if (mtag) { 1700 m_tag_prepend(m, mtag); 1701 bzero(mtag + 1, sizeof(struct ip6aux)); 1702 } 1703 } 1704 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1705} 1706 1707struct ip6aux * 1708ip6_findaux(struct mbuf *m) 1709{ 1710 struct m_tag *mtag; 1711 1712 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1713 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1714} 1715 1716void 1717ip6_delaux(struct mbuf *m) 1718{ 1719 struct m_tag *mtag; 1720 1721 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1722 if (mtag) 1723 m_tag_delete(m, mtag); 1724} 1725 1726/* 1727 * System control for IP6 1728 */ 1729 1730u_char inet6ctlerrmap[PRC_NCMDS] = { 1731 0, 0, 0, 0, 1732 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1733 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1734 EMSGSIZE, EHOSTUNREACH, 0, 0, 1735 0, 0, 0, 0, 1736 ENOPROTOOPT 1737}; 1738