ip6_input.c revision 194971
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: head/sys/netinet6/ip6_input.c 194971 2009-06-25 16:35:28Z rwatson $"); 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#include <sys/vimage.h> 84 85#include <net/if.h> 86#include <net/if_types.h> 87#include <net/if_dl.h> 88#include <net/route.h> 89#include <net/netisr.h> 90#include <net/pfil.h> 91#include <net/vnet.h> 92 93#include <netinet/in.h> 94#include <netinet/in_systm.h> 95#include <net/if_llatbl.h> 96#ifdef INET 97#include <netinet/ip.h> 98#include <netinet/ip_icmp.h> 99#include <netinet/vinet.h> 100#endif /* INET */ 101#include <netinet/ip6.h> 102#include <netinet6/in6_var.h> 103#include <netinet6/ip6_var.h> 104#include <netinet/in_pcb.h> 105#include <netinet/icmp6.h> 106#include <netinet6/scope6_var.h> 107#include <netinet6/in6_ifattach.h> 108#include <netinet6/nd6.h> 109#include <netinet6/vinet6.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 119extern struct domain inet6domain; 120 121u_char ip6_protox[IPPROTO_MAX]; 122 123static struct netisr_handler ip6_nh = { 124 .nh_name = "ip6", 125 .nh_handler = ip6_input, 126 .nh_proto = NETISR_IPV6, 127 .nh_policy = NETISR_POLICY_FLOW, 128}; 129 130#ifndef VIMAGE 131#ifndef VIMAGE_GLOBALS 132struct vnet_inet6 vnet_inet6_0; 133#endif 134#endif 135 136#ifdef VIMAGE_GLOBALS 137struct in6_ifaddrhead in6_ifaddrhead; 138struct ip6stat ip6stat; 139 140extern struct callout in6_tmpaddrtimer_ch; 141 142extern int dad_init; 143extern int pmtu_expire; 144extern int pmtu_probe; 145extern u_long rip6_sendspace; 146extern u_long rip6_recvspace; 147extern int icmp6errppslim; 148extern int icmp6_nodeinfo; 149extern int udp6_sendspace; 150extern int udp6_recvspace; 151#endif 152 153struct rwlock in6_ifaddr_lock; 154RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 155 156struct pfil_head inet6_pfil_hook; 157 158static void ip6_init2(void *); 159static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *); 160static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 161#ifdef PULLDOWN_TEST 162static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int); 163#endif 164 165#ifndef VIMAGE_GLOBALS 166static void vnet_inet6_register(void); 167 168static const vnet_modinfo_t vnet_inet6_modinfo = { 169 .vmi_id = VNET_MOD_INET6, 170 .vmi_name = "inet6", 171 .vmi_size = sizeof(struct vnet_inet6), 172 .vmi_dependson = VNET_MOD_INET /* XXX revisit - TCP/UDP needs this? */ 173}; 174 175static void 176vnet_inet6_register(void) 177{ 178 179 vnet_mod_register(&vnet_inet6_modinfo); 180} 181 182SYSINIT(inet6, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet6_register, 0); 183#endif 184 185/* 186 * IP6 initialization: fill in IP6 protocol switch table. 187 * All protocols not implemented in kernel go to raw IP6 protocol handler. 188 */ 189void 190ip6_init(void) 191{ 192 INIT_VNET_INET6(curvnet); 193 struct ip6protosw *pr; 194 int i; 195 196 V_in6_maxmtu = 0; 197#ifdef IP6_AUTO_LINKLOCAL 198 V_ip6_auto_linklocal = IP6_AUTO_LINKLOCAL; 199#else 200 V_ip6_auto_linklocal = 1; /* enable by default */ 201#endif 202 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 203 &V_ip6_auto_linklocal); 204 205#ifndef IPV6FORWARDING 206#ifdef GATEWAY6 207#define IPV6FORWARDING 1 /* forward IP6 packets not for us */ 208#else 209#define IPV6FORWARDING 0 /* don't forward IP6 packets not for us */ 210#endif /* GATEWAY6 */ 211#endif /* !IPV6FORWARDING */ 212 213#ifndef IPV6_SENDREDIRECTS 214#define IPV6_SENDREDIRECTS 1 215#endif 216 217 V_ip6_forwarding = IPV6FORWARDING; /* act as router? */ 218 V_ip6_sendredirects = IPV6_SENDREDIRECTS; 219 V_ip6_defhlim = IPV6_DEFHLIM; 220 V_ip6_defmcasthlim = IPV6_DEFAULT_MULTICAST_HOPS; 221 V_ip6_accept_rtadv = 0; /* "IPV6FORWARDING ? 0 : 1" is dangerous */ 222 V_ip6_log_interval = 5; 223 V_ip6_hdrnestlimit = 15; /* How many header options will we process? */ 224 V_ip6_dad_count = 1; /* DupAddrDetectionTransmits */ 225 V_ip6_auto_flowlabel = 1; 226 V_ip6_use_deprecated = 1;/* allow deprecated addr (RFC2462 5.5.4) */ 227 V_ip6_rr_prune = 5; /* router renumbering prefix 228 * walk list every 5 sec. */ 229 V_ip6_mcast_pmtu = 0; /* enable pMTU discovery for multicast? */ 230 V_ip6_v6only = 1; 231 V_ip6_keepfaith = 0; 232 V_ip6_log_time = (time_t)0L; 233#ifdef IPSTEALTH 234 V_ip6stealth = 0; 235#endif 236 V_nd6_onlink_ns_rfc4861 = 0; /* allow 'on-link' nd6 NS (RFC 4861) */ 237 238 V_pmtu_expire = 60*10; 239 V_pmtu_probe = 60*2; 240 241 /* raw IP6 parameters */ 242 /* 243 * Nominal space allocated to a raw ip socket. 244 */ 245#define RIPV6SNDQ 8192 246#define RIPV6RCVQ 8192 247 V_rip6_sendspace = RIPV6SNDQ; 248 V_rip6_recvspace = RIPV6RCVQ; 249 250 /* ICMPV6 parameters */ 251 V_icmp6_rediraccept = 1; /* accept and process redirects */ 252 V_icmp6_redirtimeout = 10 * 60; /* 10 minutes */ 253 V_icmp6errppslim = 100; /* 100pps */ 254 /* control how to respond to NI queries */ 255 V_icmp6_nodeinfo = (ICMP6_NODEINFO_FQDNOK|ICMP6_NODEINFO_NODEADDROK); 256 257 /* UDP on IP6 parameters */ 258 V_udp6_sendspace = 9216; /* really max datagram size */ 259 V_udp6_recvspace = 40 * (1024 + sizeof(struct sockaddr_in6)); 260 /* 40 1K datagrams */ 261 V_dad_init = 0; 262 263 TAILQ_INIT(&V_in6_ifaddrhead); 264 265 scope6_init(); 266 addrsel_policy_init(); 267 nd6_init(); 268 frag6_init(); 269 270 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 271 272 /* Skip global initialization stuff for non-default instances. */ 273 if (!IS_DEFAULT_VNET(curvnet)) 274 return; 275 276#ifdef DIAGNOSTIC 277 if (sizeof(struct protosw) != sizeof(struct ip6protosw)) 278 panic("sizeof(protosw) != sizeof(ip6protosw)"); 279#endif 280 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 281 if (pr == NULL) 282 panic("ip6_init"); 283 284 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 285 for (i = 0; i < IPPROTO_MAX; i++) 286 ip6_protox[i] = pr - inet6sw; 287 /* 288 * Cycle through IP protocols and put them into the appropriate place 289 * in ip6_protox[]. 290 */ 291 for (pr = (struct ip6protosw *)inet6domain.dom_protosw; 292 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) 293 if (pr->pr_domain->dom_family == PF_INET6 && 294 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 295 /* Be careful to only index valid IP protocols. */ 296 if (pr->pr_protocol < IPPROTO_MAX) 297 ip6_protox[pr->pr_protocol] = pr - inet6sw; 298 } 299 300 /* Initialize packet filter hooks. */ 301 inet6_pfil_hook.ph_type = PFIL_TYPE_AF; 302 inet6_pfil_hook.ph_af = AF_INET6; 303 if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) 304 printf("%s: WARNING: unable to register pfil hook, " 305 "error %d\n", __func__, i); 306 307 netisr_register(&ip6_nh); 308} 309 310#ifdef VIMAGE 311void 312ip6_destroy() 313{ 314 INIT_VNET_INET6(curvnet); 315 316 nd6_destroy(); 317 callout_drain(&V_in6_tmpaddrtimer_ch); 318} 319#endif 320 321static int 322ip6_init2_vnet(const void *unused __unused) 323{ 324 INIT_VNET_INET6(curvnet); 325 326 /* nd6_timer_init */ 327 callout_init(&V_nd6_timer_ch, 0); 328 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); 329 330 /* timer for regeneranation of temporary addresses randomize ID */ 331 callout_init(&V_in6_tmpaddrtimer_ch, 0); 332 callout_reset(&V_in6_tmpaddrtimer_ch, 333 (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor - 334 V_ip6_temp_regen_advance) * hz, 335 in6_tmpaddrtimer, curvnet); 336 337 return (0); 338} 339 340static void 341ip6_init2(void *dummy) 342{ 343 344 ip6_init2_vnet(NULL); 345} 346 347/* cheat */ 348/* This must be after route_init(), which is now SI_ORDER_THIRD */ 349SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL); 350 351void 352ip6_input(struct mbuf *m) 353{ 354 INIT_VNET_NET(curvnet); 355 INIT_VNET_INET6(curvnet); 356 struct ip6_hdr *ip6; 357 int off = sizeof(struct ip6_hdr), nest; 358 u_int32_t plen; 359 u_int32_t rtalert = ~0; 360 int nxt, ours = 0; 361 struct ifnet *deliverifp = NULL, *ifp = NULL; 362 struct in6_addr odst; 363 struct route_in6 rin6; 364 int srcrt = 0; 365 struct llentry *lle = NULL; 366 struct sockaddr_in6 dst6, *dst; 367 368 bzero(&rin6, sizeof(struct route_in6)); 369#ifdef IPSEC 370 /* 371 * should the inner packet be considered authentic? 372 * see comment in ah4_input(). 373 * NB: m cannot be NULL when passed to the input routine 374 */ 375 376 m->m_flags &= ~M_AUTHIPHDR; 377 m->m_flags &= ~M_AUTHIPDGM; 378 379#endif /* IPSEC */ 380 381 /* 382 * make sure we don't have onion peering information into m_tag. 383 */ 384 ip6_delaux(m); 385 386 /* 387 * mbuf statistics 388 */ 389 if (m->m_flags & M_EXT) { 390 if (m->m_next) 391 V_ip6stat.ip6s_mext2m++; 392 else 393 V_ip6stat.ip6s_mext1++; 394 } else { 395#define M2MMAX (sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0])) 396 if (m->m_next) { 397 if (m->m_flags & M_LOOP) { 398 V_ip6stat.ip6s_m2m[V_loif->if_index]++; 399 } else if (m->m_pkthdr.rcvif->if_index < M2MMAX) 400 V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++; 401 else 402 V_ip6stat.ip6s_m2m[0]++; 403 } else 404 V_ip6stat.ip6s_m1++; 405#undef M2MMAX 406 } 407 408 /* drop the packet if IPv6 operation is disabled on the IF */ 409 if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) { 410 m_freem(m); 411 return; 412 } 413 414 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive); 415 V_ip6stat.ip6s_total++; 416 417#ifndef PULLDOWN_TEST 418 /* 419 * L2 bridge code and some other code can return mbuf chain 420 * that does not conform to KAME requirement. too bad. 421 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 422 */ 423 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 424 struct mbuf *n; 425 426 MGETHDR(n, M_DONTWAIT, MT_HEADER); 427 if (n) 428 M_MOVE_PKTHDR(n, m); 429 if (n && n->m_pkthdr.len > MHLEN) { 430 MCLGET(n, M_DONTWAIT); 431 if ((n->m_flags & M_EXT) == 0) { 432 m_freem(n); 433 n = NULL; 434 } 435 } 436 if (n == NULL) { 437 m_freem(m); 438 return; /* ENOBUFS */ 439 } 440 441 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 442 n->m_len = n->m_pkthdr.len; 443 m_freem(m); 444 m = n; 445 } 446 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */); 447#endif 448 449 if (m->m_len < sizeof(struct ip6_hdr)) { 450 struct ifnet *inifp; 451 inifp = m->m_pkthdr.rcvif; 452 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 453 V_ip6stat.ip6s_toosmall++; 454 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 455 return; 456 } 457 } 458 459 ip6 = mtod(m, struct ip6_hdr *); 460 461 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 462 V_ip6stat.ip6s_badvers++; 463 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 464 goto bad; 465 } 466 467 V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++; 468 469 /* 470 * Check against address spoofing/corruption. 471 */ 472 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 473 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 474 /* 475 * XXX: "badscope" is not very suitable for a multicast source. 476 */ 477 V_ip6stat.ip6s_badscope++; 478 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 479 goto bad; 480 } 481 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 482 !(m->m_flags & M_LOOP)) { 483 /* 484 * In this case, the packet should come from the loopback 485 * interface. However, we cannot just check the if_flags, 486 * because ip6_mloopback() passes the "actual" interface 487 * as the outgoing/incoming interface. 488 */ 489 V_ip6stat.ip6s_badscope++; 490 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 491 goto bad; 492 } 493 494#ifdef ALTQ 495 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 496 /* packet is dropped by traffic conditioner */ 497 return; 498 } 499#endif 500 /* 501 * The following check is not documented in specs. A malicious 502 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 503 * and bypass security checks (act as if it was from 127.0.0.1 by using 504 * IPv6 src ::ffff:127.0.0.1). Be cautious. 505 * 506 * This check chokes if we are in an SIIT cloud. As none of BSDs 507 * support IPv4-less kernel compilation, we cannot support SIIT 508 * environment at all. So, it makes more sense for us to reject any 509 * malicious packets for non-SIIT environment, than try to do a 510 * partial support for SIIT environment. 511 */ 512 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 513 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 514 V_ip6stat.ip6s_badscope++; 515 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 516 goto bad; 517 } 518#if 0 519 /* 520 * Reject packets with IPv4 compatible addresses (auto tunnel). 521 * 522 * The code forbids auto tunnel relay case in RFC1933 (the check is 523 * stronger than RFC1933). We may want to re-enable it if mech-xx 524 * is revised to forbid relaying case. 525 */ 526 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 527 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 528 V_ip6stat.ip6s_badscope++; 529 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 530 goto bad; 531 } 532#endif 533 534 /* 535 * Run through list of hooks for input packets. 536 * 537 * NB: Beware of the destination address changing 538 * (e.g. by NAT rewriting). When this happens, 539 * tell ip6_forward to do the right thing. 540 */ 541 odst = ip6->ip6_dst; 542 543 /* Jump over all PFIL processing if hooks are not active. */ 544 if (!PFIL_HOOKED(&inet6_pfil_hook)) 545 goto passin; 546 547 if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL)) 548 return; 549 if (m == NULL) /* consumed by filter */ 550 return; 551 ip6 = mtod(m, struct ip6_hdr *); 552 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 553 554passin: 555 /* 556 * Disambiguate address scope zones (if there is ambiguity). 557 * We first make sure that the original source or destination address 558 * is not in our internal form for scoped addresses. Such addresses 559 * are not necessarily invalid spec-wise, but we cannot accept them due 560 * to the usage conflict. 561 * in6_setscope() then also checks and rejects the cases where src or 562 * dst are the loopback address and the receiving interface 563 * is not loopback. 564 */ 565 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 566 V_ip6stat.ip6s_badscope++; /* XXX */ 567 goto bad; 568 } 569 if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) || 570 in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) { 571 V_ip6stat.ip6s_badscope++; 572 goto bad; 573 } 574 575 /* 576 * Multicast check. Assume packet is for us to avoid 577 * prematurely taking locks. 578 */ 579 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 580 ours = 1; 581 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast); 582 deliverifp = m->m_pkthdr.rcvif; 583 goto hbhcheck; 584 } 585 586 /* 587 * Unicast check 588 */ 589 590 bzero(&dst6, sizeof(dst6)); 591 dst6.sin6_family = AF_INET6; 592 dst6.sin6_len = sizeof(struct sockaddr_in6); 593 dst6.sin6_addr = ip6->ip6_dst; 594 ifp = m->m_pkthdr.rcvif; 595 IF_AFDATA_LOCK(ifp); 596 lle = lla_lookup(LLTABLE6(ifp), 0, 597 (struct sockaddr *)&dst6); 598 IF_AFDATA_UNLOCK(ifp); 599 if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) { 600 ours = 1; 601 deliverifp = ifp; 602 LLE_RUNLOCK(lle); 603 goto hbhcheck; 604 } 605 if (lle != NULL) 606 LLE_RUNLOCK(lle); 607 608 dst = &rin6.ro_dst; 609 dst->sin6_len = sizeof(struct sockaddr_in6); 610 dst->sin6_family = AF_INET6; 611 dst->sin6_addr = ip6->ip6_dst; 612 rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0); 613 if (rin6.ro_rt) 614 RT_UNLOCK(rin6.ro_rt); 615 616#define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key)) 617 618 /* 619 * Accept the packet if the forwarding interface to the destination 620 * according to the routing table is the loopback interface, 621 * unless the associated route has a gateway. 622 * Note that this approach causes to accept a packet if there is a 623 * route to the loopback interface for the destination of the packet. 624 * But we think it's even useful in some situations, e.g. when using 625 * a special daemon which wants to intercept the packet. 626 * 627 * XXX: some OSes automatically make a cloned route for the destination 628 * of an outgoing packet. If the outgoing interface of the packet 629 * is a loopback one, the kernel would consider the packet to be 630 * accepted, even if we have no such address assinged on the interface. 631 * We check the cloned flag of the route entry to reject such cases, 632 * assuming that route entries for our own addresses are not made by 633 * cloning (it should be true because in6_addloop explicitly installs 634 * the host route). However, we might have to do an explicit check 635 * while it would be less efficient. Or, should we rather install a 636 * reject route for such a case? 637 */ 638 if (rin6.ro_rt && 639 (rin6.ro_rt->rt_flags & 640 (RTF_HOST|RTF_GATEWAY)) == RTF_HOST && 641#ifdef RTF_WASCLONED 642 !(rin6.ro_rt->rt_flags & RTF_WASCLONED) && 643#endif 644#ifdef RTF_CLONED 645 !(rin6.ro_rt->rt_flags & RTF_CLONED) && 646#endif 647#if 0 648 /* 649 * The check below is redundant since the comparison of 650 * the destination and the key of the rtentry has 651 * already done through looking up the routing table. 652 */ 653 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 654 &rt6_key(rin6.ro_rt)->sin6_addr) 655#endif 656 rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) { 657 struct in6_ifaddr *ia6 = 658 (struct in6_ifaddr *)rin6.ro_rt->rt_ifa; 659 660 /* 661 * record address information into m_tag. 662 */ 663 (void)ip6_setdstifaddr(m, ia6); 664 665 /* 666 * packets to a tentative, duplicated, or somehow invalid 667 * address must not be accepted. 668 */ 669 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) { 670 /* this address is ready */ 671 ours = 1; 672 deliverifp = ia6->ia_ifp; /* correct? */ 673 /* Count the packet in the ip address stats */ 674 ia6->ia_ifa.if_ipackets++; 675 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len; 676 goto hbhcheck; 677 } else { 678 char ip6bufs[INET6_ADDRSTRLEN]; 679 char ip6bufd[INET6_ADDRSTRLEN]; 680 /* address is not ready, so discard the packet. */ 681 nd6log((LOG_INFO, 682 "ip6_input: packet to an unready address %s->%s\n", 683 ip6_sprintf(ip6bufs, &ip6->ip6_src), 684 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 685 686 goto bad; 687 } 688 } 689 690 /* 691 * FAITH (Firewall Aided Internet Translator) 692 */ 693 if (V_ip6_keepfaith) { 694 if (rin6.ro_rt && rin6.ro_rt->rt_ifp && 695 rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) { 696 /* XXX do we need more sanity checks? */ 697 ours = 1; 698 deliverifp = rin6.ro_rt->rt_ifp; /* faith */ 699 goto hbhcheck; 700 } 701 } 702 703 /* 704 * Now there is no reason to process the packet if it's not our own 705 * and we're not a router. 706 */ 707 if (!V_ip6_forwarding) { 708 V_ip6stat.ip6s_cantforward++; 709 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 710 goto bad; 711 } 712 713 hbhcheck: 714 /* 715 * record address information into m_tag, if we don't have one yet. 716 * note that we are unable to record it, if the address is not listed 717 * as our interface address (e.g. multicast addresses, addresses 718 * within FAITH prefixes and such). 719 */ 720 if (deliverifp && !ip6_getdstifaddr(m)) { 721 struct in6_ifaddr *ia6; 722 723 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst); 724 if (ia6) { 725 if (!ip6_setdstifaddr(m, ia6)) { 726 /* 727 * XXX maybe we should drop the packet here, 728 * as we could not provide enough information 729 * to the upper layers. 730 */ 731 } 732 ifa_free(&ia6->ia_ifa); 733 } 734 } 735 736 /* 737 * Process Hop-by-Hop options header if it's contained. 738 * m may be modified in ip6_hopopts_input(). 739 * If a JumboPayload option is included, plen will also be modified. 740 */ 741 plen = (u_int32_t)ntohs(ip6->ip6_plen); 742 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 743 struct ip6_hbh *hbh; 744 745 if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) { 746#if 0 /*touches NULL pointer*/ 747 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 748#endif 749 goto out; /* m have already been freed */ 750 } 751 752 /* adjust pointer */ 753 ip6 = mtod(m, struct ip6_hdr *); 754 755 /* 756 * if the payload length field is 0 and the next header field 757 * indicates Hop-by-Hop Options header, then a Jumbo Payload 758 * option MUST be included. 759 */ 760 if (ip6->ip6_plen == 0 && plen == 0) { 761 /* 762 * Note that if a valid jumbo payload option is 763 * contained, ip6_hopopts_input() must set a valid 764 * (non-zero) payload length to the variable plen. 765 */ 766 V_ip6stat.ip6s_badoptions++; 767 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 768 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 769 icmp6_error(m, ICMP6_PARAM_PROB, 770 ICMP6_PARAMPROB_HEADER, 771 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 772 goto out; 773 } 774#ifndef PULLDOWN_TEST 775 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 776 hbh = (struct ip6_hbh *)(ip6 + 1); 777#else 778 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 779 sizeof(struct ip6_hbh)); 780 if (hbh == NULL) { 781 V_ip6stat.ip6s_tooshort++; 782 goto out; 783 } 784#endif 785 nxt = hbh->ip6h_nxt; 786 787 /* 788 * If we are acting as a router and the packet contains a 789 * router alert option, see if we know the option value. 790 * Currently, we only support the option value for MLD, in which 791 * case we should pass the packet to the multicast routing 792 * daemon. 793 */ 794 if (rtalert != ~0) { 795 switch (rtalert) { 796 case IP6OPT_RTALERT_MLD: 797 if (V_ip6_forwarding) 798 ours = 1; 799 break; 800 default: 801 /* 802 * RFC2711 requires unrecognized values must be 803 * silently ignored. 804 */ 805 break; 806 } 807 } 808 } else 809 nxt = ip6->ip6_nxt; 810 811 /* 812 * Check that the amount of data in the buffers 813 * is as at least much as the IPv6 header would have us expect. 814 * Trim mbufs if longer than we expect. 815 * Drop packet if shorter than we expect. 816 */ 817 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 818 V_ip6stat.ip6s_tooshort++; 819 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 820 goto bad; 821 } 822 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 823 if (m->m_len == m->m_pkthdr.len) { 824 m->m_len = sizeof(struct ip6_hdr) + plen; 825 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 826 } else 827 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 828 } 829 830 /* 831 * Forward if desirable. 832 */ 833 if (V_ip6_mrouter && 834 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 835 /* 836 * If we are acting as a multicast router, all 837 * incoming multicast packets are passed to the 838 * kernel-level multicast forwarding function. 839 * The packet is returned (relatively) intact; if 840 * ip6_mforward() returns a non-zero value, the packet 841 * must be discarded, else it may be accepted below. 842 * 843 * XXX TODO: Check hlim and multicast scope here to avoid 844 * unnecessarily calling into ip6_mforward(). 845 */ 846 if (ip6_mforward && 847 ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) { 848 IP6STAT_INC(ip6s_cantforward); 849 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 850 goto bad; 851 } 852 } else if (!ours) { 853 ip6_forward(m, srcrt); 854 goto out; 855 } 856 857 ip6 = mtod(m, struct ip6_hdr *); 858 859 /* 860 * Malicious party may be able to use IPv4 mapped addr to confuse 861 * tcp/udp stack and bypass security checks (act as if it was from 862 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 863 * 864 * For SIIT end node behavior, you may want to disable the check. 865 * However, you will become vulnerable to attacks using IPv4 mapped 866 * source. 867 */ 868 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 869 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 870 V_ip6stat.ip6s_badscope++; 871 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 872 goto bad; 873 } 874 875 /* 876 * Tell launch routine the next header 877 */ 878 V_ip6stat.ip6s_delivered++; 879 in6_ifstat_inc(deliverifp, ifs6_in_deliver); 880 nest = 0; 881 882 while (nxt != IPPROTO_DONE) { 883 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 884 V_ip6stat.ip6s_toomanyhdr++; 885 goto bad; 886 } 887 888 /* 889 * protection against faulty packet - there should be 890 * more sanity checks in header chain processing. 891 */ 892 if (m->m_pkthdr.len < off) { 893 V_ip6stat.ip6s_tooshort++; 894 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 895 goto bad; 896 } 897 898#ifdef IPSEC 899 /* 900 * enforce IPsec policy checking if we are seeing last header. 901 * note that we do not visit this with protocols with pcb layer 902 * code - like udp/tcp/raw ip. 903 */ 904 if (ip6_ipsec_input(m, nxt)) 905 goto bad; 906#endif /* IPSEC */ 907 908 /* 909 * Use mbuf flags to propagate Router Alert option to 910 * ICMPv6 layer, as hop-by-hop options have been stripped. 911 */ 912 if (nxt == IPPROTO_ICMPV6 && rtalert != ~0) 913 m->m_flags |= M_RTALERT_MLD; 914 915 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 916 } 917 goto out; 918bad: 919 m_freem(m); 920out: 921 if (rin6.ro_rt) 922 RTFREE(rin6.ro_rt); 923} 924 925/* 926 * set/grab in6_ifaddr correspond to IPv6 destination address. 927 * XXX backward compatibility wrapper 928 * 929 * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag, 930 * and then bump it when the tag is copied, and release it when the tag is 931 * freed. Unfortunately, m_tags don't support deep copies (yet), so instead 932 * we just bump the ia refcount when we receive it. This should be fixed. 933 */ 934static struct ip6aux * 935ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6) 936{ 937 struct ip6aux *ip6a; 938 939 ip6a = ip6_addaux(m); 940 if (ip6a) 941 ip6a->ip6a_dstia6 = ia6; 942 return ip6a; /* NULL if failed to set */ 943} 944 945struct in6_ifaddr * 946ip6_getdstifaddr(struct mbuf *m) 947{ 948 struct ip6aux *ip6a; 949 struct in6_ifaddr *ia; 950 951 ip6a = ip6_findaux(m); 952 if (ip6a) { 953 ia = ip6a->ip6a_dstia6; 954 ifa_ref(&ia->ia_ifa); 955 return ia; 956 } else 957 return NULL; 958} 959 960/* 961 * Hop-by-Hop options header processing. If a valid jumbo payload option is 962 * included, the real payload length will be stored in plenp. 963 * 964 * rtalertp - XXX: should be stored more smart way 965 */ 966static int 967ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 968 struct mbuf **mp, int *offp) 969{ 970 INIT_VNET_INET6(curvnet); 971 struct mbuf *m = *mp; 972 int off = *offp, hbhlen; 973 struct ip6_hbh *hbh; 974 u_int8_t *opt; 975 976 /* validation of the length of the header */ 977#ifndef PULLDOWN_TEST 978 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); 979 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 980 hbhlen = (hbh->ip6h_len + 1) << 3; 981 982 IP6_EXTHDR_CHECK(m, off, hbhlen, -1); 983 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 984#else 985 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, 986 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); 987 if (hbh == NULL) { 988 V_ip6stat.ip6s_tooshort++; 989 return -1; 990 } 991 hbhlen = (hbh->ip6h_len + 1) << 3; 992 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 993 hbhlen); 994 if (hbh == NULL) { 995 V_ip6stat.ip6s_tooshort++; 996 return -1; 997 } 998#endif 999 off += hbhlen; 1000 hbhlen -= sizeof(struct ip6_hbh); 1001 opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh); 1002 1003 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 1004 hbhlen, rtalertp, plenp) < 0) 1005 return (-1); 1006 1007 *offp = off; 1008 *mp = m; 1009 return (0); 1010} 1011 1012/* 1013 * Search header for all Hop-by-hop options and process each option. 1014 * This function is separate from ip6_hopopts_input() in order to 1015 * handle a case where the sending node itself process its hop-by-hop 1016 * options header. In such a case, the function is called from ip6_output(). 1017 * 1018 * The function assumes that hbh header is located right after the IPv6 header 1019 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1020 * opthead + hbhlen is located in continuous memory region. 1021 */ 1022int 1023ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1024 u_int32_t *rtalertp, u_int32_t *plenp) 1025{ 1026 INIT_VNET_INET6(curvnet); 1027 struct ip6_hdr *ip6; 1028 int optlen = 0; 1029 u_int8_t *opt = opthead; 1030 u_int16_t rtalert_val; 1031 u_int32_t jumboplen; 1032 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1033 1034 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1035 switch (*opt) { 1036 case IP6OPT_PAD1: 1037 optlen = 1; 1038 break; 1039 case IP6OPT_PADN: 1040 if (hbhlen < IP6OPT_MINLEN) { 1041 V_ip6stat.ip6s_toosmall++; 1042 goto bad; 1043 } 1044 optlen = *(opt + 1) + 2; 1045 break; 1046 case IP6OPT_ROUTER_ALERT: 1047 /* XXX may need check for alignment */ 1048 if (hbhlen < IP6OPT_RTALERT_LEN) { 1049 V_ip6stat.ip6s_toosmall++; 1050 goto bad; 1051 } 1052 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1053 /* XXX stat */ 1054 icmp6_error(m, ICMP6_PARAM_PROB, 1055 ICMP6_PARAMPROB_HEADER, 1056 erroff + opt + 1 - opthead); 1057 return (-1); 1058 } 1059 optlen = IP6OPT_RTALERT_LEN; 1060 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1061 *rtalertp = ntohs(rtalert_val); 1062 break; 1063 case IP6OPT_JUMBO: 1064 /* XXX may need check for alignment */ 1065 if (hbhlen < IP6OPT_JUMBO_LEN) { 1066 V_ip6stat.ip6s_toosmall++; 1067 goto bad; 1068 } 1069 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1070 /* XXX stat */ 1071 icmp6_error(m, ICMP6_PARAM_PROB, 1072 ICMP6_PARAMPROB_HEADER, 1073 erroff + opt + 1 - opthead); 1074 return (-1); 1075 } 1076 optlen = IP6OPT_JUMBO_LEN; 1077 1078 /* 1079 * IPv6 packets that have non 0 payload length 1080 * must not contain a jumbo payload option. 1081 */ 1082 ip6 = mtod(m, struct ip6_hdr *); 1083 if (ip6->ip6_plen) { 1084 V_ip6stat.ip6s_badoptions++; 1085 icmp6_error(m, ICMP6_PARAM_PROB, 1086 ICMP6_PARAMPROB_HEADER, 1087 erroff + opt - opthead); 1088 return (-1); 1089 } 1090 1091 /* 1092 * We may see jumbolen in unaligned location, so 1093 * we'd need to perform bcopy(). 1094 */ 1095 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1096 jumboplen = (u_int32_t)htonl(jumboplen); 1097 1098#if 1 1099 /* 1100 * if there are multiple jumbo payload options, 1101 * *plenp will be non-zero and the packet will be 1102 * rejected. 1103 * the behavior may need some debate in ipngwg - 1104 * multiple options does not make sense, however, 1105 * there's no explicit mention in specification. 1106 */ 1107 if (*plenp != 0) { 1108 V_ip6stat.ip6s_badoptions++; 1109 icmp6_error(m, ICMP6_PARAM_PROB, 1110 ICMP6_PARAMPROB_HEADER, 1111 erroff + opt + 2 - opthead); 1112 return (-1); 1113 } 1114#endif 1115 1116 /* 1117 * jumbo payload length must be larger than 65535. 1118 */ 1119 if (jumboplen <= IPV6_MAXPACKET) { 1120 V_ip6stat.ip6s_badoptions++; 1121 icmp6_error(m, ICMP6_PARAM_PROB, 1122 ICMP6_PARAMPROB_HEADER, 1123 erroff + opt + 2 - opthead); 1124 return (-1); 1125 } 1126 *plenp = jumboplen; 1127 1128 break; 1129 default: /* unknown option */ 1130 if (hbhlen < IP6OPT_MINLEN) { 1131 V_ip6stat.ip6s_toosmall++; 1132 goto bad; 1133 } 1134 optlen = ip6_unknown_opt(opt, m, 1135 erroff + opt - opthead); 1136 if (optlen == -1) 1137 return (-1); 1138 optlen += 2; 1139 break; 1140 } 1141 } 1142 1143 return (0); 1144 1145 bad: 1146 m_freem(m); 1147 return (-1); 1148} 1149 1150/* 1151 * Unknown option processing. 1152 * The third argument `off' is the offset from the IPv6 header to the option, 1153 * which is necessary if the IPv6 header the and option header and IPv6 header 1154 * is not continuous in order to return an ICMPv6 error. 1155 */ 1156int 1157ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1158{ 1159 INIT_VNET_INET6(curvnet); 1160 struct ip6_hdr *ip6; 1161 1162 switch (IP6OPT_TYPE(*optp)) { 1163 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1164 return ((int)*(optp + 1)); 1165 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1166 m_freem(m); 1167 return (-1); 1168 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1169 V_ip6stat.ip6s_badoptions++; 1170 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1171 return (-1); 1172 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1173 V_ip6stat.ip6s_badoptions++; 1174 ip6 = mtod(m, struct ip6_hdr *); 1175 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1176 (m->m_flags & (M_BCAST|M_MCAST))) 1177 m_freem(m); 1178 else 1179 icmp6_error(m, ICMP6_PARAM_PROB, 1180 ICMP6_PARAMPROB_OPTION, off); 1181 return (-1); 1182 } 1183 1184 m_freem(m); /* XXX: NOTREACHED */ 1185 return (-1); 1186} 1187 1188/* 1189 * Create the "control" list for this pcb. 1190 * These functions will not modify mbuf chain at all. 1191 * 1192 * With KAME mbuf chain restriction: 1193 * The routine will be called from upper layer handlers like tcp6_input(). 1194 * Thus the routine assumes that the caller (tcp6_input) have already 1195 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the 1196 * very first mbuf on the mbuf chain. 1197 * 1198 * ip6_savecontrol_v4 will handle those options that are possible to be 1199 * set on a v4-mapped socket. 1200 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1201 * options and handle the v6-only ones itself. 1202 */ 1203struct mbuf ** 1204ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1205 int *v4only) 1206{ 1207 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1208 1209#ifdef SO_TIMESTAMP 1210 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1211 struct timeval tv; 1212 1213 microtime(&tv); 1214 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1215 SCM_TIMESTAMP, SOL_SOCKET); 1216 if (*mp) 1217 mp = &(*mp)->m_next; 1218 } 1219#endif 1220 1221 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1222 if (v4only != NULL) 1223 *v4only = 1; 1224 return (mp); 1225 } 1226 1227#define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1228 /* RFC 2292 sec. 5 */ 1229 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1230 struct in6_pktinfo pi6; 1231 1232 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1233 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1234 pi6.ipi6_ifindex = 1235 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1236 1237 *mp = sbcreatecontrol((caddr_t) &pi6, 1238 sizeof(struct in6_pktinfo), 1239 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1240 if (*mp) 1241 mp = &(*mp)->m_next; 1242 } 1243 1244 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1245 int hlim = ip6->ip6_hlim & 0xff; 1246 1247 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1248 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1249 IPPROTO_IPV6); 1250 if (*mp) 1251 mp = &(*mp)->m_next; 1252 } 1253 1254 if (v4only != NULL) 1255 *v4only = 0; 1256 return (mp); 1257} 1258 1259void 1260ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) 1261{ 1262 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1263 int v4only = 0; 1264 1265 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); 1266 if (v4only) 1267 return; 1268 1269 if ((in6p->inp_flags & IN6P_TCLASS) != 0) { 1270 u_int32_t flowinfo; 1271 int tclass; 1272 1273 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1274 flowinfo >>= 20; 1275 1276 tclass = flowinfo & 0xff; 1277 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass), 1278 IPV6_TCLASS, IPPROTO_IPV6); 1279 if (*mp) 1280 mp = &(*mp)->m_next; 1281 } 1282 1283 /* 1284 * IPV6_HOPOPTS socket option. Recall that we required super-user 1285 * privilege for the option (see ip6_ctloutput), but it might be too 1286 * strict, since there might be some hop-by-hop options which can be 1287 * returned to normal user. 1288 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1289 */ 1290 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { 1291 /* 1292 * Check if a hop-by-hop options header is contatined in the 1293 * received packet, and if so, store the options as ancillary 1294 * data. Note that a hop-by-hop options header must be 1295 * just after the IPv6 header, which is assured through the 1296 * IPv6 input processing. 1297 */ 1298 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1299 struct ip6_hbh *hbh; 1300 int hbhlen = 0; 1301#ifdef PULLDOWN_TEST 1302 struct mbuf *ext; 1303#endif 1304 1305#ifndef PULLDOWN_TEST 1306 hbh = (struct ip6_hbh *)(ip6 + 1); 1307 hbhlen = (hbh->ip6h_len + 1) << 3; 1308#else 1309 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), 1310 ip6->ip6_nxt); 1311 if (ext == NULL) { 1312 V_ip6stat.ip6s_tooshort++; 1313 return; 1314 } 1315 hbh = mtod(ext, struct ip6_hbh *); 1316 hbhlen = (hbh->ip6h_len + 1) << 3; 1317 if (hbhlen != ext->m_len) { 1318 m_freem(ext); 1319 V_ip6stat.ip6s_tooshort++; 1320 return; 1321 } 1322#endif 1323 1324 /* 1325 * XXX: We copy the whole header even if a 1326 * jumbo payload option is included, the option which 1327 * is to be removed before returning according to 1328 * RFC2292. 1329 * Note: this constraint is removed in RFC3542 1330 */ 1331 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1332 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1333 IPPROTO_IPV6); 1334 if (*mp) 1335 mp = &(*mp)->m_next; 1336#ifdef PULLDOWN_TEST 1337 m_freem(ext); 1338#endif 1339 } 1340 } 1341 1342 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1343 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1344 1345 /* 1346 * Search for destination options headers or routing 1347 * header(s) through the header chain, and stores each 1348 * header as ancillary data. 1349 * Note that the order of the headers remains in 1350 * the chain of ancillary data. 1351 */ 1352 while (1) { /* is explicit loop prevention necessary? */ 1353 struct ip6_ext *ip6e = NULL; 1354 int elen; 1355#ifdef PULLDOWN_TEST 1356 struct mbuf *ext = NULL; 1357#endif 1358 1359 /* 1360 * if it is not an extension header, don't try to 1361 * pull it from the chain. 1362 */ 1363 switch (nxt) { 1364 case IPPROTO_DSTOPTS: 1365 case IPPROTO_ROUTING: 1366 case IPPROTO_HOPOPTS: 1367 case IPPROTO_AH: /* is it possible? */ 1368 break; 1369 default: 1370 goto loopend; 1371 } 1372 1373#ifndef PULLDOWN_TEST 1374 if (off + sizeof(*ip6e) > m->m_len) 1375 goto loopend; 1376 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1377 if (nxt == IPPROTO_AH) 1378 elen = (ip6e->ip6e_len + 2) << 2; 1379 else 1380 elen = (ip6e->ip6e_len + 1) << 3; 1381 if (off + elen > m->m_len) 1382 goto loopend; 1383#else 1384 ext = ip6_pullexthdr(m, off, nxt); 1385 if (ext == NULL) { 1386 V_ip6stat.ip6s_tooshort++; 1387 return; 1388 } 1389 ip6e = mtod(ext, struct ip6_ext *); 1390 if (nxt == IPPROTO_AH) 1391 elen = (ip6e->ip6e_len + 2) << 2; 1392 else 1393 elen = (ip6e->ip6e_len + 1) << 3; 1394 if (elen != ext->m_len) { 1395 m_freem(ext); 1396 V_ip6stat.ip6s_tooshort++; 1397 return; 1398 } 1399#endif 1400 1401 switch (nxt) { 1402 case IPPROTO_DSTOPTS: 1403 if (!(in6p->inp_flags & IN6P_DSTOPTS)) 1404 break; 1405 1406 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1407 IS2292(in6p, 1408 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1409 IPPROTO_IPV6); 1410 if (*mp) 1411 mp = &(*mp)->m_next; 1412 break; 1413 case IPPROTO_ROUTING: 1414 if (!in6p->inp_flags & IN6P_RTHDR) 1415 break; 1416 1417 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1418 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), 1419 IPPROTO_IPV6); 1420 if (*mp) 1421 mp = &(*mp)->m_next; 1422 break; 1423 case IPPROTO_HOPOPTS: 1424 case IPPROTO_AH: /* is it possible? */ 1425 break; 1426 1427 default: 1428 /* 1429 * other cases have been filtered in the above. 1430 * none will visit this case. here we supply 1431 * the code just in case (nxt overwritten or 1432 * other cases). 1433 */ 1434#ifdef PULLDOWN_TEST 1435 m_freem(ext); 1436#endif 1437 goto loopend; 1438 1439 } 1440 1441 /* proceed with the next header. */ 1442 off += elen; 1443 nxt = ip6e->ip6e_nxt; 1444 ip6e = NULL; 1445#ifdef PULLDOWN_TEST 1446 m_freem(ext); 1447 ext = NULL; 1448#endif 1449 } 1450 loopend: 1451 ; 1452 } 1453} 1454#undef IS2292 1455 1456void 1457ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu) 1458{ 1459 struct socket *so; 1460 struct mbuf *m_mtu; 1461 struct ip6_mtuinfo mtuctl; 1462 1463 so = in6p->inp_socket; 1464 1465 if (mtu == NULL) 1466 return; 1467 1468#ifdef DIAGNOSTIC 1469 if (so == NULL) /* I believe this is impossible */ 1470 panic("ip6_notify_pmtu: socket is NULL"); 1471#endif 1472 1473 bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */ 1474 mtuctl.ip6m_mtu = *mtu; 1475 mtuctl.ip6m_addr = *dst; 1476 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1477 return; 1478 1479 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1480 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1481 return; 1482 1483 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1484 == 0) { 1485 m_freem(m_mtu); 1486 /* XXX: should count statistics */ 1487 } else 1488 sorwakeup(so); 1489 1490 return; 1491} 1492 1493#ifdef PULLDOWN_TEST 1494/* 1495 * pull single extension header from mbuf chain. returns single mbuf that 1496 * contains the result, or NULL on error. 1497 */ 1498static struct mbuf * 1499ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) 1500{ 1501 struct ip6_ext ip6e; 1502 size_t elen; 1503 struct mbuf *n; 1504 1505#ifdef DIAGNOSTIC 1506 switch (nxt) { 1507 case IPPROTO_DSTOPTS: 1508 case IPPROTO_ROUTING: 1509 case IPPROTO_HOPOPTS: 1510 case IPPROTO_AH: /* is it possible? */ 1511 break; 1512 default: 1513 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); 1514 } 1515#endif 1516 1517 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1518 if (nxt == IPPROTO_AH) 1519 elen = (ip6e.ip6e_len + 2) << 2; 1520 else 1521 elen = (ip6e.ip6e_len + 1) << 3; 1522 1523 MGET(n, M_DONTWAIT, MT_DATA); 1524 if (n && elen >= MLEN) { 1525 MCLGET(n, M_DONTWAIT); 1526 if ((n->m_flags & M_EXT) == 0) { 1527 m_free(n); 1528 n = NULL; 1529 } 1530 } 1531 if (!n) 1532 return NULL; 1533 1534 n->m_len = 0; 1535 if (elen >= M_TRAILINGSPACE(n)) { 1536 m_free(n); 1537 return NULL; 1538 } 1539 1540 m_copydata(m, off, elen, mtod(n, caddr_t)); 1541 n->m_len = elen; 1542 return n; 1543} 1544#endif 1545 1546/* 1547 * Get pointer to the previous header followed by the header 1548 * currently processed. 1549 * XXX: This function supposes that 1550 * M includes all headers, 1551 * the next header field and the header length field of each header 1552 * are valid, and 1553 * the sum of each header length equals to OFF. 1554 * Because of these assumptions, this function must be called very 1555 * carefully. Moreover, it will not be used in the near future when 1556 * we develop `neater' mechanism to process extension headers. 1557 */ 1558char * 1559ip6_get_prevhdr(struct mbuf *m, int off) 1560{ 1561 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1562 1563 if (off == sizeof(struct ip6_hdr)) 1564 return (&ip6->ip6_nxt); 1565 else { 1566 int len, nxt; 1567 struct ip6_ext *ip6e = NULL; 1568 1569 nxt = ip6->ip6_nxt; 1570 len = sizeof(struct ip6_hdr); 1571 while (len < off) { 1572 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len); 1573 1574 switch (nxt) { 1575 case IPPROTO_FRAGMENT: 1576 len += sizeof(struct ip6_frag); 1577 break; 1578 case IPPROTO_AH: 1579 len += (ip6e->ip6e_len + 2) << 2; 1580 break; 1581 default: 1582 len += (ip6e->ip6e_len + 1) << 3; 1583 break; 1584 } 1585 nxt = ip6e->ip6e_nxt; 1586 } 1587 if (ip6e) 1588 return (&ip6e->ip6e_nxt); 1589 else 1590 return NULL; 1591 } 1592} 1593 1594/* 1595 * get next header offset. m will be retained. 1596 */ 1597int 1598ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp) 1599{ 1600 struct ip6_hdr ip6; 1601 struct ip6_ext ip6e; 1602 struct ip6_frag fh; 1603 1604 /* just in case */ 1605 if (m == NULL) 1606 panic("ip6_nexthdr: m == NULL"); 1607 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1608 return -1; 1609 1610 switch (proto) { 1611 case IPPROTO_IPV6: 1612 if (m->m_pkthdr.len < off + sizeof(ip6)) 1613 return -1; 1614 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1615 if (nxtp) 1616 *nxtp = ip6.ip6_nxt; 1617 off += sizeof(ip6); 1618 return off; 1619 1620 case IPPROTO_FRAGMENT: 1621 /* 1622 * terminate parsing if it is not the first fragment, 1623 * it does not make sense to parse through it. 1624 */ 1625 if (m->m_pkthdr.len < off + sizeof(fh)) 1626 return -1; 1627 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1628 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1629 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1630 return -1; 1631 if (nxtp) 1632 *nxtp = fh.ip6f_nxt; 1633 off += sizeof(struct ip6_frag); 1634 return off; 1635 1636 case IPPROTO_AH: 1637 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1638 return -1; 1639 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1640 if (nxtp) 1641 *nxtp = ip6e.ip6e_nxt; 1642 off += (ip6e.ip6e_len + 2) << 2; 1643 return off; 1644 1645 case IPPROTO_HOPOPTS: 1646 case IPPROTO_ROUTING: 1647 case IPPROTO_DSTOPTS: 1648 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1649 return -1; 1650 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1651 if (nxtp) 1652 *nxtp = ip6e.ip6e_nxt; 1653 off += (ip6e.ip6e_len + 1) << 3; 1654 return off; 1655 1656 case IPPROTO_NONE: 1657 case IPPROTO_ESP: 1658 case IPPROTO_IPCOMP: 1659 /* give up */ 1660 return -1; 1661 1662 default: 1663 return -1; 1664 } 1665 1666 return -1; 1667} 1668 1669/* 1670 * get offset for the last header in the chain. m will be kept untainted. 1671 */ 1672int 1673ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp) 1674{ 1675 int newoff; 1676 int nxt; 1677 1678 if (!nxtp) { 1679 nxt = -1; 1680 nxtp = &nxt; 1681 } 1682 while (1) { 1683 newoff = ip6_nexthdr(m, off, proto, nxtp); 1684 if (newoff < 0) 1685 return off; 1686 else if (newoff < off) 1687 return -1; /* invalid */ 1688 else if (newoff == off) 1689 return newoff; 1690 1691 off = newoff; 1692 proto = *nxtp; 1693 } 1694} 1695 1696struct ip6aux * 1697ip6_addaux(struct mbuf *m) 1698{ 1699 struct m_tag *mtag; 1700 1701 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1702 if (!mtag) { 1703 mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux), 1704 M_NOWAIT); 1705 if (mtag) { 1706 m_tag_prepend(m, mtag); 1707 bzero(mtag + 1, sizeof(struct ip6aux)); 1708 } 1709 } 1710 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1711} 1712 1713struct ip6aux * 1714ip6_findaux(struct mbuf *m) 1715{ 1716 struct m_tag *mtag; 1717 1718 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1719 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1720} 1721 1722void 1723ip6_delaux(struct mbuf *m) 1724{ 1725 struct m_tag *mtag; 1726 1727 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1728 if (mtag) 1729 m_tag_delete(m, mtag); 1730} 1731 1732/* 1733 * System control for IP6 1734 */ 1735 1736u_char inet6ctlerrmap[PRC_NCMDS] = { 1737 0, 0, 0, 0, 1738 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1739 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1740 EMSGSIZE, EHOSTUNREACH, 0, 0, 1741 0, 0, 0, 0, 1742 ENOPROTOOPT 1743}; 1744