1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $ 30 */ 31 32 /*- 33 * Copyright (c) 1982, 1986, 1988, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 4. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 61 */ 62 63 #include <sys/cdefs.h> 64 __FBSDID("$FreeBSD$"); 65 66 #include "opt_inet.h" 67 #include "opt_inet6.h" 68 #include "opt_ipfw.h" 69 #include "opt_ipsec.h" 70 #include "opt_route.h" 71 72 #include <sys/param.h> 73 #include <sys/systm.h> 74 #include <sys/malloc.h> 75 #include <sys/mbuf.h> 76 #include <sys/proc.h> 77 #include <sys/domain.h> 78 #include <sys/protosw.h> 79 #include <sys/socket.h> 80 #include <sys/socketvar.h> 81 #include <sys/errno.h> 82 #include <sys/time.h> 83 #include <sys/kernel.h> 84 #include <sys/syslog.h> 85 86 #include <net/if.h> 87 #include <net/if_types.h> 88 #include <net/if_dl.h> 89 #include <net/route.h> 90 #include <net/netisr.h> 91 #include <net/pfil.h> 92 #include <net/vnet.h> 93 94 #include <netinet/in.h> 95 #include <netinet/ip_var.h> 96 #include <netinet/in_systm.h> 97 #include <net/if_llatbl.h> 98 #ifdef INET 99 #include <netinet/ip.h> 100 #include <netinet/ip_icmp.h> 101 #endif /* INET */ 102 #include <netinet/ip6.h> 103 #include <netinet6/in6_var.h> 104 #include <netinet6/ip6_var.h> 105 #include <netinet/in_pcb.h> 106 #include <netinet/icmp6.h> 107 #include <netinet6/scope6_var.h> 108 #include <netinet6/in6_ifattach.h> 109 #include <netinet6/nd6.h> 110 111 #ifdef IPSEC 112 #include <netipsec/ipsec.h> 113 #include <netinet6/ip6_ipsec.h> 114 #include <netipsec/ipsec6.h> 115 #endif /* IPSEC */ 116 117 #include <netinet6/ip6protosw.h> 118 119 #ifdef FLOWTABLE 120 #include <net/flowtable.h> 121 VNET_DECLARE(int, ip6_output_flowtable_size); 122 #define V_ip6_output_flowtable_size VNET(ip6_output_flowtable_size) 123 #endif 124 125 extern struct domain inet6domain; 126 127 u_char ip6_protox[IPPROTO_MAX]; 128 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 129 130 static struct netisr_handler ip6_nh = { 131 .nh_name = "ip6", 132 .nh_handler = ip6_input, 133 .nh_proto = NETISR_IPV6, 134 .nh_policy = NETISR_POLICY_FLOW, 135 }; 136 137 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch); 138 #define V_in6_tmpaddrtimer_ch VNET(in6_tmpaddrtimer_ch) 139 140 VNET_DEFINE(struct pfil_head, inet6_pfil_hook); 141 142 VNET_DEFINE(struct ip6stat, ip6stat); 143 144 struct rwlock in6_ifaddr_lock; 145 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 146 147 static void ip6_init2(void *); 148 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *); 149 static struct ip6aux *ip6_addaux(struct mbuf *); 150 static struct ip6aux *ip6_findaux(struct mbuf *m); 151 static void ip6_delaux (struct mbuf *); 152 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 153 #ifdef PULLDOWN_TEST 154 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int); 155 #endif 156 157 /* 158 * IP6 initialization: fill in IP6 protocol switch table. 159 * All protocols not implemented in kernel go to raw IP6 protocol handler. 160 */ 161 void 162 ip6_init(void) 163 { 164 struct ip6protosw *pr; 165 int i; 166 167 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 168 &V_ip6_auto_linklocal); 169 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 170 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 171 172 TAILQ_INIT(&V_in6_ifaddrhead); 173 174 /* Initialize packet filter hooks. */ 175 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF; 176 V_inet6_pfil_hook.ph_af = AF_INET6; 177 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0) 178 printf("%s: WARNING: unable to register pfil hook, " 179 "error %d\n", __func__, i); 180 181 scope6_init(); 182 addrsel_policy_init(); 183 nd6_init(); 184 frag6_init(); 185 186 #ifdef FLOWTABLE 187 if (TUNABLE_INT_FETCH("net.inet6.ip6.output_flowtable_size", 188 &V_ip6_output_flowtable_size)) { 189 if (V_ip6_output_flowtable_size < 256) 190 V_ip6_output_flowtable_size = 256; 191 if (!powerof2(V_ip6_output_flowtable_size)) { 192 printf("flowtable must be power of 2 size\n"); 193 V_ip6_output_flowtable_size = 2048; 194 } 195 } else { 196 /* 197 * round up to the next power of 2 198 */ 199 V_ip6_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1); 200 } 201 V_ip6_ft = flowtable_alloc("ipv6", V_ip6_output_flowtable_size, FL_IPV6|FL_PCPU); 202 #endif 203 204 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 205 206 /* Skip global initialization stuff for non-default instances. */ 207 if (!IS_DEFAULT_VNET(curvnet)) 208 return; 209 210 #ifdef DIAGNOSTIC 211 if (sizeof(struct protosw) != sizeof(struct ip6protosw)) 212 panic("sizeof(protosw) != sizeof(ip6protosw)"); 213 #endif 214 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 215 if (pr == NULL) 216 panic("ip6_init"); 217 218 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 219 for (i = 0; i < IPPROTO_MAX; i++) 220 ip6_protox[i] = pr - inet6sw; 221 /* 222 * Cycle through IP protocols and put them into the appropriate place 223 * in ip6_protox[]. 224 */ 225 for (pr = (struct ip6protosw *)inet6domain.dom_protosw; 226 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) 227 if (pr->pr_domain->dom_family == PF_INET6 && 228 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 229 /* Be careful to only index valid IP protocols. */ 230 if (pr->pr_protocol < IPPROTO_MAX) 231 ip6_protox[pr->pr_protocol] = pr - inet6sw; 232 } 233 234 netisr_register(&ip6_nh); 235 } 236 237 /* 238 * The protocol to be inserted into ip6_protox[] must be already registered 239 * in inet6sw[], either statically or through pf_proto_register(). 240 */ 241 int 242 ip6proto_register(short ip6proto) 243 { 244 struct ip6protosw *pr; 245 246 /* Sanity checks. */ 247 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 248 return (EPROTONOSUPPORT); 249 250 /* 251 * The protocol slot must not be occupied by another protocol 252 * already. An index pointing to IPPROTO_RAW is unused. 253 */ 254 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 255 if (pr == NULL) 256 return (EPFNOSUPPORT); 257 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 258 return (EEXIST); 259 260 /* 261 * Find the protocol position in inet6sw[] and set the index. 262 */ 263 for (pr = (struct ip6protosw *)inet6domain.dom_protosw; 264 pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) { 265 if (pr->pr_domain->dom_family == PF_INET6 && 266 pr->pr_protocol && pr->pr_protocol == ip6proto) { 267 ip6_protox[pr->pr_protocol] = pr - inet6sw; 268 return (0); 269 } 270 } 271 return (EPROTONOSUPPORT); 272 } 273 274 int 275 ip6proto_unregister(short ip6proto) 276 { 277 struct ip6protosw *pr; 278 279 /* Sanity checks. */ 280 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 281 return (EPROTONOSUPPORT); 282 283 /* Check if the protocol was indeed registered. */ 284 pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 285 if (pr == NULL) 286 return (EPFNOSUPPORT); 287 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 288 return (ENOENT); 289 290 /* Reset the protocol slot to IPPROTO_RAW. */ 291 ip6_protox[ip6proto] = pr - inet6sw; 292 return (0); 293 } 294 295 #ifdef VIMAGE 296 void 297 ip6_destroy() 298 { 299 300 nd6_destroy(); 301 callout_drain(&V_in6_tmpaddrtimer_ch); 302 } 303 #endif 304 305 static int 306 ip6_init2_vnet(const void *unused __unused) 307 { 308 309 /* nd6_timer_init */ 310 callout_init(&V_nd6_timer_ch, 0); 311 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); 312 313 /* timer for regeneranation of temporary addresses randomize ID */ 314 callout_init(&V_in6_tmpaddrtimer_ch, 0); 315 callout_reset(&V_in6_tmpaddrtimer_ch, 316 (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor - 317 V_ip6_temp_regen_advance) * hz, 318 in6_tmpaddrtimer, curvnet); 319 320 return (0); 321 } 322 323 static void 324 ip6_init2(void *dummy) 325 { 326 327 ip6_init2_vnet(NULL); 328 } 329 330 /* cheat */ 331 /* This must be after route_init(), which is now SI_ORDER_THIRD */ 332 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL); 333 334 static int 335 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off, 336 int *nxt, int *ours) 337 { 338 struct ip6_hdr *ip6; 339 struct ip6_hbh *hbh; 340 341 if (ip6_hopopts_input(plen, rtalert, &m, off)) { 342 #if 0 /*touches NULL pointer*/ 343 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 344 #endif 345 goto out; /* m have already been freed */ 346 } 347 348 /* adjust pointer */ 349 ip6 = mtod(m, struct ip6_hdr *); 350 351 /* 352 * if the payload length field is 0 and the next header field 353 * indicates Hop-by-Hop Options header, then a Jumbo Payload 354 * option MUST be included. 355 */ 356 if (ip6->ip6_plen == 0 && *plen == 0) { 357 /* 358 * Note that if a valid jumbo payload option is 359 * contained, ip6_hopopts_input() must set a valid 360 * (non-zero) payload length to the variable plen. 361 */ 362 V_ip6stat.ip6s_badoptions++; 363 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 364 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 365 icmp6_error(m, ICMP6_PARAM_PROB, 366 ICMP6_PARAMPROB_HEADER, 367 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 368 goto out; 369 } 370 #ifndef PULLDOWN_TEST 371 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 372 hbh = (struct ip6_hbh *)(ip6 + 1); 373 #else 374 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 375 sizeof(struct ip6_hbh)); 376 if (hbh == NULL) { 377 V_ip6stat.ip6s_tooshort++; 378 goto out; 379 } 380 #endif 381 *nxt = hbh->ip6h_nxt; 382 383 /* 384 * If we are acting as a router and the packet contains a 385 * router alert option, see if we know the option value. 386 * Currently, we only support the option value for MLD, in which 387 * case we should pass the packet to the multicast routing 388 * daemon. 389 */ 390 if (*rtalert != ~0) { 391 switch (*rtalert) { 392 case IP6OPT_RTALERT_MLD: 393 if (V_ip6_forwarding) 394 *ours = 1; 395 break; 396 default: 397 /* 398 * RFC2711 requires unrecognized values must be 399 * silently ignored. 400 */ 401 break; 402 } 403 } 404 405 return (0); 406 407 out: 408 return (1); 409 } 410 411 void 412 ip6_input(struct mbuf *m) 413 { 414 struct ip6_hdr *ip6; 415 int off = sizeof(struct ip6_hdr), nest; 416 u_int32_t plen; 417 u_int32_t rtalert = ~0; 418 int nxt, ours = 0; 419 struct ifnet *deliverifp = NULL, *ifp = NULL; 420 struct in6_addr odst; 421 struct route_in6 rin6; 422 int srcrt = 0; 423 struct llentry *lle = NULL; 424 struct sockaddr_in6 dst6, *dst; 425 426 bzero(&rin6, sizeof(struct route_in6)); 427 #ifdef IPSEC 428 /* 429 * should the inner packet be considered authentic? 430 * see comment in ah4_input(). 431 * NB: m cannot be NULL when passed to the input routine 432 */ 433 434 m->m_flags &= ~M_AUTHIPHDR; 435 m->m_flags &= ~M_AUTHIPDGM; 436 437 #endif /* IPSEC */ 438 439 /* 440 * make sure we don't have onion peering information into m_tag. 441 */ 442 ip6_delaux(m); 443 444 if (m->m_flags & M_FASTFWD_OURS) { 445 /* 446 * Firewall changed destination to local. 447 */ 448 m->m_flags &= ~M_FASTFWD_OURS; 449 ours = 1; 450 deliverifp = m->m_pkthdr.rcvif; 451 ip6 = mtod(m, struct ip6_hdr *); 452 goto hbhcheck; 453 } 454 455 /* 456 * mbuf statistics 457 */ 458 if (m->m_flags & M_EXT) { 459 if (m->m_next) 460 V_ip6stat.ip6s_mext2m++; 461 else 462 V_ip6stat.ip6s_mext1++; 463 } else { 464 #define M2MMAX (sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0])) 465 if (m->m_next) { 466 if (m->m_flags & M_LOOP) { 467 V_ip6stat.ip6s_m2m[V_loif->if_index]++; 468 } else if (m->m_pkthdr.rcvif->if_index < M2MMAX) 469 V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++; 470 else 471 V_ip6stat.ip6s_m2m[0]++; 472 } else 473 V_ip6stat.ip6s_m1++; 474 #undef M2MMAX 475 } 476 477 /* drop the packet if IPv6 operation is disabled on the IF */ 478 if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) { 479 m_freem(m); 480 return; 481 } 482 483 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive); 484 V_ip6stat.ip6s_total++; 485 486 #ifndef PULLDOWN_TEST 487 /* 488 * L2 bridge code and some other code can return mbuf chain 489 * that does not conform to KAME requirement. too bad. 490 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 491 */ 492 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 493 struct mbuf *n; 494 495 MGETHDR(n, M_DONTWAIT, MT_HEADER); 496 if (n) 497 M_MOVE_PKTHDR(n, m); 498 if (n && n->m_pkthdr.len > MHLEN) { 499 MCLGET(n, M_DONTWAIT); 500 if ((n->m_flags & M_EXT) == 0) { 501 m_freem(n); 502 n = NULL; 503 } 504 } 505 if (n == NULL) { 506 m_freem(m); 507 return; /* ENOBUFS */ 508 } 509 510 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 511 n->m_len = n->m_pkthdr.len; 512 m_freem(m); 513 m = n; 514 } 515 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */); 516 #endif 517 518 if (m->m_len < sizeof(struct ip6_hdr)) { 519 struct ifnet *inifp; 520 inifp = m->m_pkthdr.rcvif; 521 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 522 V_ip6stat.ip6s_toosmall++; 523 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 524 return; 525 } 526 } 527 528 ip6 = mtod(m, struct ip6_hdr *); 529 530 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 531 V_ip6stat.ip6s_badvers++; 532 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 533 goto bad; 534 } 535 536 V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++; 537 538 /* 539 * Check against address spoofing/corruption. 540 */ 541 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 542 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 543 /* 544 * XXX: "badscope" is not very suitable for a multicast source. 545 */ 546 V_ip6stat.ip6s_badscope++; 547 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 548 goto bad; 549 } 550 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 551 !(m->m_flags & M_LOOP)) { 552 /* 553 * In this case, the packet should come from the loopback 554 * interface. However, we cannot just check the if_flags, 555 * because ip6_mloopback() passes the "actual" interface 556 * as the outgoing/incoming interface. 557 */ 558 V_ip6stat.ip6s_badscope++; 559 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 560 goto bad; 561 } 562 563 #ifdef ALTQ 564 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 565 /* packet is dropped by traffic conditioner */ 566 return; 567 } 568 #endif 569 /* 570 * The following check is not documented in specs. A malicious 571 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 572 * and bypass security checks (act as if it was from 127.0.0.1 by using 573 * IPv6 src ::ffff:127.0.0.1). Be cautious. 574 * 575 * This check chokes if we are in an SIIT cloud. As none of BSDs 576 * support IPv4-less kernel compilation, we cannot support SIIT 577 * environment at all. So, it makes more sense for us to reject any 578 * malicious packets for non-SIIT environment, than try to do a 579 * partial support for SIIT environment. 580 */ 581 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 582 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 583 V_ip6stat.ip6s_badscope++; 584 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 585 goto bad; 586 } 587 #if 0 588 /* 589 * Reject packets with IPv4 compatible addresses (auto tunnel). 590 * 591 * The code forbids auto tunnel relay case in RFC1933 (the check is 592 * stronger than RFC1933). We may want to re-enable it if mech-xx 593 * is revised to forbid relaying case. 594 */ 595 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 596 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 597 V_ip6stat.ip6s_badscope++; 598 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 599 goto bad; 600 } 601 #endif 602 #ifdef IPSEC 603 /* 604 * Bypass packet filtering for packets previously handled by IPsec. 605 */ 606 if (ip6_ipsec_filtertunnel(m)) 607 goto passin; 608 #endif /* IPSEC */ 609 610 /* 611 * Run through list of hooks for input packets. 612 * 613 * NB: Beware of the destination address changing 614 * (e.g. by NAT rewriting). When this happens, 615 * tell ip6_forward to do the right thing. 616 */ 617 odst = ip6->ip6_dst; 618 619 /* Jump over all PFIL processing if hooks are not active. */ 620 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 621 goto passin; 622 623 if (pfil_run_hooks(&V_inet6_pfil_hook, &m, 624 m->m_pkthdr.rcvif, PFIL_IN, NULL)) 625 return; 626 if (m == NULL) /* consumed by filter */ 627 return; 628 ip6 = mtod(m, struct ip6_hdr *); 629 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 630 631 #ifdef IPFIREWALL_FORWARD 632 if (m->m_flags & M_FASTFWD_OURS) { 633 m->m_flags &= ~M_FASTFWD_OURS; 634 ours = 1; 635 deliverifp = m->m_pkthdr.rcvif; 636 goto hbhcheck; 637 } 638 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 639 /* 640 * Directly ship the packet on. This allows forwarding 641 * packets originally destined to us to some other directly 642 * connected host. 643 */ 644 ip6_forward(m, 1); 645 goto out; 646 } 647 #endif /* IPFIREWALL_FORWARD */ 648 649 passin: 650 /* 651 * Disambiguate address scope zones (if there is ambiguity). 652 * We first make sure that the original source or destination address 653 * is not in our internal form for scoped addresses. Such addresses 654 * are not necessarily invalid spec-wise, but we cannot accept them due 655 * to the usage conflict. 656 * in6_setscope() then also checks and rejects the cases where src or 657 * dst are the loopback address and the receiving interface 658 * is not loopback. 659 */ 660 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 661 V_ip6stat.ip6s_badscope++; /* XXX */ 662 goto bad; 663 } 664 if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) || 665 in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) { 666 V_ip6stat.ip6s_badscope++; 667 goto bad; 668 } 669 670 /* 671 * Multicast check. Assume packet is for us to avoid 672 * prematurely taking locks. 673 */ 674 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 675 ours = 1; 676 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast); 677 deliverifp = m->m_pkthdr.rcvif; 678 goto hbhcheck; 679 } 680 681 /* 682 * Unicast check 683 */ 684 685 bzero(&dst6, sizeof(dst6)); 686 dst6.sin6_family = AF_INET6; 687 dst6.sin6_len = sizeof(struct sockaddr_in6); 688 dst6.sin6_addr = ip6->ip6_dst; 689 ifp = m->m_pkthdr.rcvif; 690 IF_AFDATA_LOCK(ifp); 691 lle = lla_lookup(LLTABLE6(ifp), 0, 692 (struct sockaddr *)&dst6); 693 IF_AFDATA_UNLOCK(ifp); 694 if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) { 695 struct ifaddr *ifa; 696 struct in6_ifaddr *ia6; 697 int bad; 698 699 bad = 1; 700 #define sa_equal(a1, a2) \ 701 (bcmp((a1), (a2), ((a1))->sin6_len) == 0) 702 IF_ADDR_RLOCK(ifp); 703 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 704 if (ifa->ifa_addr->sa_family != dst6.sin6_family) 705 continue; 706 if (sa_equal(&dst6, ifa->ifa_addr)) 707 break; 708 } 709 KASSERT(ifa != NULL, ("%s: ifa not found for lle %p", 710 __func__, lle)); 711 #undef sa_equal 712 713 ia6 = (struct in6_ifaddr *)ifa; 714 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) { 715 /* Count the packet in the ip address stats */ 716 ia6->ia_ifa.if_ipackets++; 717 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len; 718 719 /* 720 * record address information into m_tag. 721 */ 722 (void)ip6_setdstifaddr(m, ia6); 723 724 bad = 0; 725 } else { 726 char ip6bufs[INET6_ADDRSTRLEN]; 727 char ip6bufd[INET6_ADDRSTRLEN]; 728 /* address is not ready, so discard the packet. */ 729 nd6log((LOG_INFO, 730 "ip6_input: packet to an unready address %s->%s\n", 731 ip6_sprintf(ip6bufs, &ip6->ip6_src), 732 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 733 } 734 IF_ADDR_RUNLOCK(ifp); 735 LLE_RUNLOCK(lle); 736 if (bad) 737 goto bad; 738 else { 739 ours = 1; 740 deliverifp = ifp; 741 goto hbhcheck; 742 } 743 } 744 if (lle != NULL) 745 LLE_RUNLOCK(lle); 746 747 dst = &rin6.ro_dst; 748 dst->sin6_len = sizeof(struct sockaddr_in6); 749 dst->sin6_family = AF_INET6; 750 dst->sin6_addr = ip6->ip6_dst; 751 rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m)); 752 if (rin6.ro_rt) 753 RT_UNLOCK(rin6.ro_rt); 754 755 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key)) 756 757 /* 758 * Accept the packet if the forwarding interface to the destination 759 * according to the routing table is the loopback interface, 760 * unless the associated route has a gateway. 761 * Note that this approach causes to accept a packet if there is a 762 * route to the loopback interface for the destination of the packet. 763 * But we think it's even useful in some situations, e.g. when using 764 * a special daemon which wants to intercept the packet. 765 * 766 * XXX: some OSes automatically make a cloned route for the destination 767 * of an outgoing packet. If the outgoing interface of the packet 768 * is a loopback one, the kernel would consider the packet to be 769 * accepted, even if we have no such address assinged on the interface. 770 * We check the cloned flag of the route entry to reject such cases, 771 * assuming that route entries for our own addresses are not made by 772 * cloning (it should be true because in6_addloop explicitly installs 773 * the host route). However, we might have to do an explicit check 774 * while it would be less efficient. Or, should we rather install a 775 * reject route for such a case? 776 */ 777 if (rin6.ro_rt && 778 (rin6.ro_rt->rt_flags & 779 (RTF_HOST|RTF_GATEWAY)) == RTF_HOST && 780 #ifdef RTF_WASCLONED 781 !(rin6.ro_rt->rt_flags & RTF_WASCLONED) && 782 #endif 783 #ifdef RTF_CLONED 784 !(rin6.ro_rt->rt_flags & RTF_CLONED) && 785 #endif 786 #if 0 787 /* 788 * The check below is redundant since the comparison of 789 * the destination and the key of the rtentry has 790 * already done through looking up the routing table. 791 */ 792 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, 793 &rt6_key(rin6.ro_rt)->sin6_addr) 794 #endif 795 rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) { 796 int free_ia6 = 0; 797 struct in6_ifaddr *ia6; 798 799 /* 800 * found the loopback route to the interface address 801 */ 802 if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) { 803 struct sockaddr_in6 dest6; 804 805 bzero(&dest6, sizeof(dest6)); 806 dest6.sin6_family = AF_INET6; 807 dest6.sin6_len = sizeof(dest6); 808 dest6.sin6_addr = ip6->ip6_dst; 809 ia6 = (struct in6_ifaddr *) 810 ifa_ifwithaddr((struct sockaddr *)&dest6); 811 if (ia6 == NULL) 812 goto bad; 813 free_ia6 = 1; 814 } 815 else 816 ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa; 817 818 /* 819 * record address information into m_tag. 820 */ 821 (void)ip6_setdstifaddr(m, ia6); 822 823 /* 824 * packets to a tentative, duplicated, or somehow invalid 825 * address must not be accepted. 826 */ 827 if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) { 828 /* this address is ready */ 829 ours = 1; 830 deliverifp = ia6->ia_ifp; /* correct? */ 831 /* Count the packet in the ip address stats */ 832 ia6->ia_ifa.if_ipackets++; 833 ia6->ia_ifa.if_ibytes += m->m_pkthdr.len; 834 if (ia6 != NULL && free_ia6 != 0) 835 ifa_free(&ia6->ia_ifa); 836 goto hbhcheck; 837 } else { 838 char ip6bufs[INET6_ADDRSTRLEN]; 839 char ip6bufd[INET6_ADDRSTRLEN]; 840 /* address is not ready, so discard the packet. */ 841 nd6log((LOG_INFO, 842 "ip6_input: packet to an unready address %s->%s\n", 843 ip6_sprintf(ip6bufs, &ip6->ip6_src), 844 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 845 846 if (ia6 != NULL && free_ia6 != 0) 847 ifa_free(&ia6->ia_ifa); 848 goto bad; 849 } 850 } 851 852 /* 853 * FAITH (Firewall Aided Internet Translator) 854 */ 855 if (V_ip6_keepfaith) { 856 if (rin6.ro_rt && rin6.ro_rt->rt_ifp && 857 rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) { 858 /* XXX do we need more sanity checks? */ 859 ours = 1; 860 deliverifp = rin6.ro_rt->rt_ifp; /* faith */ 861 goto hbhcheck; 862 } 863 } 864 865 /* 866 * Now there is no reason to process the packet if it's not our own 867 * and we're not a router. 868 */ 869 if (!V_ip6_forwarding) { 870 V_ip6stat.ip6s_cantforward++; 871 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 872 goto bad; 873 } 874 875 hbhcheck: 876 /* 877 * record address information into m_tag, if we don't have one yet. 878 * note that we are unable to record it, if the address is not listed 879 * as our interface address (e.g. multicast addresses, addresses 880 * within FAITH prefixes and such). 881 */ 882 if (deliverifp) { 883 struct in6_ifaddr *ia6; 884 885 if ((ia6 = ip6_getdstifaddr(m)) != NULL) { 886 ifa_free(&ia6->ia_ifa); 887 } else { 888 ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst); 889 if (ia6) { 890 if (!ip6_setdstifaddr(m, ia6)) { 891 /* 892 * XXX maybe we should drop the packet here, 893 * as we could not provide enough information 894 * to the upper layers. 895 */ 896 } 897 ifa_free(&ia6->ia_ifa); 898 } 899 } 900 } 901 902 /* 903 * Process Hop-by-Hop options header if it's contained. 904 * m may be modified in ip6_hopopts_input(). 905 * If a JumboPayload option is included, plen will also be modified. 906 */ 907 plen = (u_int32_t)ntohs(ip6->ip6_plen); 908 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 909 int error; 910 911 error = ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours); 912 if (error != 0) 913 goto out; 914 } else 915 nxt = ip6->ip6_nxt; 916 917 /* 918 * Check that the amount of data in the buffers 919 * is as at least much as the IPv6 header would have us expect. 920 * Trim mbufs if longer than we expect. 921 * Drop packet if shorter than we expect. 922 */ 923 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 924 V_ip6stat.ip6s_tooshort++; 925 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 926 goto bad; 927 } 928 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 929 if (m->m_len == m->m_pkthdr.len) { 930 m->m_len = sizeof(struct ip6_hdr) + plen; 931 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 932 } else 933 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 934 } 935 936 /* 937 * Forward if desirable. 938 */ 939 if (V_ip6_mrouter && 940 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 941 /* 942 * If we are acting as a multicast router, all 943 * incoming multicast packets are passed to the 944 * kernel-level multicast forwarding function. 945 * The packet is returned (relatively) intact; if 946 * ip6_mforward() returns a non-zero value, the packet 947 * must be discarded, else it may be accepted below. 948 * 949 * XXX TODO: Check hlim and multicast scope here to avoid 950 * unnecessarily calling into ip6_mforward(). 951 */ 952 if (ip6_mforward && 953 ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) { 954 IP6STAT_INC(ip6s_cantforward); 955 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 956 goto bad; 957 } 958 } else if (!ours) { 959 ip6_forward(m, srcrt); 960 goto out; 961 } 962 963 ip6 = mtod(m, struct ip6_hdr *); 964 965 /* 966 * Malicious party may be able to use IPv4 mapped addr to confuse 967 * tcp/udp stack and bypass security checks (act as if it was from 968 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 969 * 970 * For SIIT end node behavior, you may want to disable the check. 971 * However, you will become vulnerable to attacks using IPv4 mapped 972 * source. 973 */ 974 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 975 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 976 V_ip6stat.ip6s_badscope++; 977 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 978 goto bad; 979 } 980 981 /* 982 * Tell launch routine the next header 983 */ 984 V_ip6stat.ip6s_delivered++; 985 in6_ifstat_inc(deliverifp, ifs6_in_deliver); 986 nest = 0; 987 988 while (nxt != IPPROTO_DONE) { 989 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 990 V_ip6stat.ip6s_toomanyhdr++; 991 goto bad; 992 } 993 994 /* 995 * protection against faulty packet - there should be 996 * more sanity checks in header chain processing. 997 */ 998 if (m->m_pkthdr.len < off) { 999 V_ip6stat.ip6s_tooshort++; 1000 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 1001 goto bad; 1002 } 1003 1004 #ifdef IPSEC 1005 /* 1006 * enforce IPsec policy checking if we are seeing last header. 1007 * note that we do not visit this with protocols with pcb layer 1008 * code - like udp/tcp/raw ip. 1009 */ 1010 if (ip6_ipsec_input(m, nxt)) 1011 goto bad; 1012 #endif /* IPSEC */ 1013 1014 /* 1015 * Use mbuf flags to propagate Router Alert option to 1016 * ICMPv6 layer, as hop-by-hop options have been stripped. 1017 */ 1018 if (nxt == IPPROTO_ICMPV6 && rtalert != ~0) 1019 m->m_flags |= M_RTALERT_MLD; 1020 1021 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 1022 } 1023 goto out; 1024 bad: 1025 m_freem(m); 1026 out: 1027 if (rin6.ro_rt) 1028 RTFREE(rin6.ro_rt); 1029 } 1030 1031 /* 1032 * set/grab in6_ifaddr correspond to IPv6 destination address. 1033 * XXX backward compatibility wrapper 1034 * 1035 * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag, 1036 * and then bump it when the tag is copied, and release it when the tag is 1037 * freed. Unfortunately, m_tags don't support deep copies (yet), so instead 1038 * we just bump the ia refcount when we receive it. This should be fixed. 1039 */ 1040 static struct ip6aux * 1041 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6) 1042 { 1043 struct ip6aux *ip6a; 1044 1045 ip6a = ip6_addaux(m); 1046 if (ip6a) 1047 ip6a->ip6a_dstia6 = ia6; 1048 return ip6a; /* NULL if failed to set */ 1049 } 1050 1051 struct in6_ifaddr * 1052 ip6_getdstifaddr(struct mbuf *m) 1053 { 1054 struct ip6aux *ip6a; 1055 struct in6_ifaddr *ia; 1056 1057 ip6a = ip6_findaux(m); 1058 if (ip6a) { 1059 ia = ip6a->ip6a_dstia6; 1060 ifa_ref(&ia->ia_ifa); 1061 return ia; 1062 } else 1063 return NULL; 1064 } 1065 1066 /* 1067 * Hop-by-Hop options header processing. If a valid jumbo payload option is 1068 * included, the real payload length will be stored in plenp. 1069 * 1070 * rtalertp - XXX: should be stored more smart way 1071 */ 1072 static int 1073 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 1074 struct mbuf **mp, int *offp) 1075 { 1076 struct mbuf *m = *mp; 1077 int off = *offp, hbhlen; 1078 struct ip6_hbh *hbh; 1079 u_int8_t *opt; 1080 1081 /* validation of the length of the header */ 1082 #ifndef PULLDOWN_TEST 1083 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); 1084 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 1085 hbhlen = (hbh->ip6h_len + 1) << 3; 1086 1087 IP6_EXTHDR_CHECK(m, off, hbhlen, -1); 1088 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 1089 #else 1090 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, 1091 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); 1092 if (hbh == NULL) { 1093 V_ip6stat.ip6s_tooshort++; 1094 return -1; 1095 } 1096 hbhlen = (hbh->ip6h_len + 1) << 3; 1097 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 1098 hbhlen); 1099 if (hbh == NULL) { 1100 V_ip6stat.ip6s_tooshort++; 1101 return -1; 1102 } 1103 #endif 1104 off += hbhlen; 1105 hbhlen -= sizeof(struct ip6_hbh); 1106 opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh); 1107 1108 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 1109 hbhlen, rtalertp, plenp) < 0) 1110 return (-1); 1111 1112 *offp = off; 1113 *mp = m; 1114 return (0); 1115 } 1116 1117 /* 1118 * Search header for all Hop-by-hop options and process each option. 1119 * This function is separate from ip6_hopopts_input() in order to 1120 * handle a case where the sending node itself process its hop-by-hop 1121 * options header. In such a case, the function is called from ip6_output(). 1122 * 1123 * The function assumes that hbh header is located right after the IPv6 header 1124 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1125 * opthead + hbhlen is located in contiguous memory region. 1126 */ 1127 int 1128 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1129 u_int32_t *rtalertp, u_int32_t *plenp) 1130 { 1131 struct ip6_hdr *ip6; 1132 int optlen = 0; 1133 u_int8_t *opt = opthead; 1134 u_int16_t rtalert_val; 1135 u_int32_t jumboplen; 1136 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1137 1138 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1139 switch (*opt) { 1140 case IP6OPT_PAD1: 1141 optlen = 1; 1142 break; 1143 case IP6OPT_PADN: 1144 if (hbhlen < IP6OPT_MINLEN) { 1145 V_ip6stat.ip6s_toosmall++; 1146 goto bad; 1147 } 1148 optlen = *(opt + 1) + 2; 1149 break; 1150 case IP6OPT_ROUTER_ALERT: 1151 /* XXX may need check for alignment */ 1152 if (hbhlen < IP6OPT_RTALERT_LEN) { 1153 V_ip6stat.ip6s_toosmall++; 1154 goto bad; 1155 } 1156 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1157 /* XXX stat */ 1158 icmp6_error(m, ICMP6_PARAM_PROB, 1159 ICMP6_PARAMPROB_HEADER, 1160 erroff + opt + 1 - opthead); 1161 return (-1); 1162 } 1163 optlen = IP6OPT_RTALERT_LEN; 1164 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1165 *rtalertp = ntohs(rtalert_val); 1166 break; 1167 case IP6OPT_JUMBO: 1168 /* XXX may need check for alignment */ 1169 if (hbhlen < IP6OPT_JUMBO_LEN) { 1170 V_ip6stat.ip6s_toosmall++; 1171 goto bad; 1172 } 1173 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1174 /* XXX stat */ 1175 icmp6_error(m, ICMP6_PARAM_PROB, 1176 ICMP6_PARAMPROB_HEADER, 1177 erroff + opt + 1 - opthead); 1178 return (-1); 1179 } 1180 optlen = IP6OPT_JUMBO_LEN; 1181 1182 /* 1183 * IPv6 packets that have non 0 payload length 1184 * must not contain a jumbo payload option. 1185 */ 1186 ip6 = mtod(m, struct ip6_hdr *); 1187 if (ip6->ip6_plen) { 1188 V_ip6stat.ip6s_badoptions++; 1189 icmp6_error(m, ICMP6_PARAM_PROB, 1190 ICMP6_PARAMPROB_HEADER, 1191 erroff + opt - opthead); 1192 return (-1); 1193 } 1194 1195 /* 1196 * We may see jumbolen in unaligned location, so 1197 * we'd need to perform bcopy(). 1198 */ 1199 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1200 jumboplen = (u_int32_t)htonl(jumboplen); 1201 1202 #if 1 1203 /* 1204 * if there are multiple jumbo payload options, 1205 * *plenp will be non-zero and the packet will be 1206 * rejected. 1207 * the behavior may need some debate in ipngwg - 1208 * multiple options does not make sense, however, 1209 * there's no explicit mention in specification. 1210 */ 1211 if (*plenp != 0) { 1212 V_ip6stat.ip6s_badoptions++; 1213 icmp6_error(m, ICMP6_PARAM_PROB, 1214 ICMP6_PARAMPROB_HEADER, 1215 erroff + opt + 2 - opthead); 1216 return (-1); 1217 } 1218 #endif 1219 1220 /* 1221 * jumbo payload length must be larger than 65535. 1222 */ 1223 if (jumboplen <= IPV6_MAXPACKET) { 1224 V_ip6stat.ip6s_badoptions++; 1225 icmp6_error(m, ICMP6_PARAM_PROB, 1226 ICMP6_PARAMPROB_HEADER, 1227 erroff + opt + 2 - opthead); 1228 return (-1); 1229 } 1230 *plenp = jumboplen; 1231 1232 break; 1233 default: /* unknown option */ 1234 if (hbhlen < IP6OPT_MINLEN) { 1235 V_ip6stat.ip6s_toosmall++; 1236 goto bad; 1237 } 1238 optlen = ip6_unknown_opt(opt, m, 1239 erroff + opt - opthead); 1240 if (optlen == -1) 1241 return (-1); 1242 optlen += 2; 1243 break; 1244 } 1245 } 1246 1247 return (0); 1248 1249 bad: 1250 m_freem(m); 1251 return (-1); 1252 } 1253 1254 /* 1255 * Unknown option processing. 1256 * The third argument `off' is the offset from the IPv6 header to the option, 1257 * which is necessary if the IPv6 header the and option header and IPv6 header 1258 * is not contiguous in order to return an ICMPv6 error. 1259 */ 1260 int 1261 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1262 { 1263 struct ip6_hdr *ip6; 1264 1265 switch (IP6OPT_TYPE(*optp)) { 1266 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1267 return ((int)*(optp + 1)); 1268 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1269 m_freem(m); 1270 return (-1); 1271 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1272 V_ip6stat.ip6s_badoptions++; 1273 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1274 return (-1); 1275 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1276 V_ip6stat.ip6s_badoptions++; 1277 ip6 = mtod(m, struct ip6_hdr *); 1278 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1279 (m->m_flags & (M_BCAST|M_MCAST))) 1280 m_freem(m); 1281 else 1282 icmp6_error(m, ICMP6_PARAM_PROB, 1283 ICMP6_PARAMPROB_OPTION, off); 1284 return (-1); 1285 } 1286 1287 m_freem(m); /* XXX: NOTREACHED */ 1288 return (-1); 1289 } 1290 1291 /* 1292 * Create the "control" list for this pcb. 1293 * These functions will not modify mbuf chain at all. 1294 * 1295 * With KAME mbuf chain restriction: 1296 * The routine will be called from upper layer handlers like tcp6_input(). 1297 * Thus the routine assumes that the caller (tcp6_input) have already 1298 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the 1299 * very first mbuf on the mbuf chain. 1300 * 1301 * ip6_savecontrol_v4 will handle those options that are possible to be 1302 * set on a v4-mapped socket. 1303 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1304 * options and handle the v6-only ones itself. 1305 */ 1306 struct mbuf ** 1307 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1308 int *v4only) 1309 { 1310 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1311 1312 #ifdef SO_TIMESTAMP 1313 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1314 struct timeval tv; 1315 1316 microtime(&tv); 1317 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1318 SCM_TIMESTAMP, SOL_SOCKET); 1319 if (*mp) 1320 mp = &(*mp)->m_next; 1321 } 1322 #endif 1323 1324 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1325 if (v4only != NULL) 1326 *v4only = 1; 1327 return (mp); 1328 } 1329 1330 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1331 /* RFC 2292 sec. 5 */ 1332 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1333 struct in6_pktinfo pi6; 1334 1335 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1336 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1337 pi6.ipi6_ifindex = 1338 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1339 1340 *mp = sbcreatecontrol((caddr_t) &pi6, 1341 sizeof(struct in6_pktinfo), 1342 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1343 if (*mp) 1344 mp = &(*mp)->m_next; 1345 } 1346 1347 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1348 int hlim = ip6->ip6_hlim & 0xff; 1349 1350 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1351 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1352 IPPROTO_IPV6); 1353 if (*mp) 1354 mp = &(*mp)->m_next; 1355 } 1356 1357 if (v4only != NULL) 1358 *v4only = 0; 1359 return (mp); 1360 } 1361 1362 void 1363 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) 1364 { 1365 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1366 int v4only = 0; 1367 1368 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); 1369 if (v4only) 1370 return; 1371 1372 if ((in6p->inp_flags & IN6P_TCLASS) != 0) { 1373 u_int32_t flowinfo; 1374 int tclass; 1375 1376 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1377 flowinfo >>= 20; 1378 1379 tclass = flowinfo & 0xff; 1380 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass), 1381 IPV6_TCLASS, IPPROTO_IPV6); 1382 if (*mp) 1383 mp = &(*mp)->m_next; 1384 } 1385 1386 /* 1387 * IPV6_HOPOPTS socket option. Recall that we required super-user 1388 * privilege for the option (see ip6_ctloutput), but it might be too 1389 * strict, since there might be some hop-by-hop options which can be 1390 * returned to normal user. 1391 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1392 */ 1393 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { 1394 /* 1395 * Check if a hop-by-hop options header is contatined in the 1396 * received packet, and if so, store the options as ancillary 1397 * data. Note that a hop-by-hop options header must be 1398 * just after the IPv6 header, which is assured through the 1399 * IPv6 input processing. 1400 */ 1401 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1402 struct ip6_hbh *hbh; 1403 int hbhlen = 0; 1404 #ifdef PULLDOWN_TEST 1405 struct mbuf *ext; 1406 #endif 1407 1408 #ifndef PULLDOWN_TEST 1409 hbh = (struct ip6_hbh *)(ip6 + 1); 1410 hbhlen = (hbh->ip6h_len + 1) << 3; 1411 #else 1412 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), 1413 ip6->ip6_nxt); 1414 if (ext == NULL) { 1415 V_ip6stat.ip6s_tooshort++; 1416 return; 1417 } 1418 hbh = mtod(ext, struct ip6_hbh *); 1419 hbhlen = (hbh->ip6h_len + 1) << 3; 1420 if (hbhlen != ext->m_len) { 1421 m_freem(ext); 1422 V_ip6stat.ip6s_tooshort++; 1423 return; 1424 } 1425 #endif 1426 1427 /* 1428 * XXX: We copy the whole header even if a 1429 * jumbo payload option is included, the option which 1430 * is to be removed before returning according to 1431 * RFC2292. 1432 * Note: this constraint is removed in RFC3542 1433 */ 1434 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1435 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1436 IPPROTO_IPV6); 1437 if (*mp) 1438 mp = &(*mp)->m_next; 1439 #ifdef PULLDOWN_TEST 1440 m_freem(ext); 1441 #endif 1442 } 1443 } 1444 1445 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1446 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1447 1448 /* 1449 * Search for destination options headers or routing 1450 * header(s) through the header chain, and stores each 1451 * header as ancillary data. 1452 * Note that the order of the headers remains in 1453 * the chain of ancillary data. 1454 */ 1455 while (1) { /* is explicit loop prevention necessary? */ 1456 struct ip6_ext *ip6e = NULL; 1457 int elen; 1458 #ifdef PULLDOWN_TEST 1459 struct mbuf *ext = NULL; 1460 #endif 1461 1462 /* 1463 * if it is not an extension header, don't try to 1464 * pull it from the chain. 1465 */ 1466 switch (nxt) { 1467 case IPPROTO_DSTOPTS: 1468 case IPPROTO_ROUTING: 1469 case IPPROTO_HOPOPTS: 1470 case IPPROTO_AH: /* is it possible? */ 1471 break; 1472 default: 1473 goto loopend; 1474 } 1475 1476 #ifndef PULLDOWN_TEST 1477 if (off + sizeof(*ip6e) > m->m_len) 1478 goto loopend; 1479 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1480 if (nxt == IPPROTO_AH) 1481 elen = (ip6e->ip6e_len + 2) << 2; 1482 else 1483 elen = (ip6e->ip6e_len + 1) << 3; 1484 if (off + elen > m->m_len) 1485 goto loopend; 1486 #else 1487 ext = ip6_pullexthdr(m, off, nxt); 1488 if (ext == NULL) { 1489 V_ip6stat.ip6s_tooshort++; 1490 return; 1491 } 1492 ip6e = mtod(ext, struct ip6_ext *); 1493 if (nxt == IPPROTO_AH) 1494 elen = (ip6e->ip6e_len + 2) << 2; 1495 else 1496 elen = (ip6e->ip6e_len + 1) << 3; 1497 if (elen != ext->m_len) { 1498 m_freem(ext); 1499 V_ip6stat.ip6s_tooshort++; 1500 return; 1501 } 1502 #endif 1503 1504 switch (nxt) { 1505 case IPPROTO_DSTOPTS: 1506 if (!(in6p->inp_flags & IN6P_DSTOPTS)) 1507 break; 1508 1509 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1510 IS2292(in6p, 1511 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1512 IPPROTO_IPV6); 1513 if (*mp) 1514 mp = &(*mp)->m_next; 1515 break; 1516 case IPPROTO_ROUTING: 1517 if (!(in6p->inp_flags & IN6P_RTHDR)) 1518 break; 1519 1520 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1521 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), 1522 IPPROTO_IPV6); 1523 if (*mp) 1524 mp = &(*mp)->m_next; 1525 break; 1526 case IPPROTO_HOPOPTS: 1527 case IPPROTO_AH: /* is it possible? */ 1528 break; 1529 1530 default: 1531 /* 1532 * other cases have been filtered in the above. 1533 * none will visit this case. here we supply 1534 * the code just in case (nxt overwritten or 1535 * other cases). 1536 */ 1537 #ifdef PULLDOWN_TEST 1538 m_freem(ext); 1539 #endif 1540 goto loopend; 1541 1542 } 1543 1544 /* proceed with the next header. */ 1545 off += elen; 1546 nxt = ip6e->ip6e_nxt; 1547 ip6e = NULL; 1548 #ifdef PULLDOWN_TEST 1549 m_freem(ext); 1550 ext = NULL; 1551 #endif 1552 } 1553 loopend: 1554 ; 1555 } 1556 } 1557 #undef IS2292 1558 1559 void 1560 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu) 1561 { 1562 struct socket *so; 1563 struct mbuf *m_mtu; 1564 struct ip6_mtuinfo mtuctl; 1565 1566 so = in6p->inp_socket; 1567 1568 if (mtu == NULL) 1569 return; 1570 1571 #ifdef DIAGNOSTIC 1572 if (so == NULL) /* I believe this is impossible */ 1573 panic("ip6_notify_pmtu: socket is NULL"); 1574 #endif 1575 1576 bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */ 1577 mtuctl.ip6m_mtu = *mtu; 1578 mtuctl.ip6m_addr = *dst; 1579 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1580 return; 1581 1582 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1583 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1584 return; 1585 1586 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1587 == 0) { 1588 m_freem(m_mtu); 1589 /* XXX: should count statistics */ 1590 } else 1591 sorwakeup(so); 1592 1593 return; 1594 } 1595 1596 #ifdef PULLDOWN_TEST 1597 /* 1598 * pull single extension header from mbuf chain. returns single mbuf that 1599 * contains the result, or NULL on error. 1600 */ 1601 static struct mbuf * 1602 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) 1603 { 1604 struct ip6_ext ip6e; 1605 size_t elen; 1606 struct mbuf *n; 1607 1608 #ifdef DIAGNOSTIC 1609 switch (nxt) { 1610 case IPPROTO_DSTOPTS: 1611 case IPPROTO_ROUTING: 1612 case IPPROTO_HOPOPTS: 1613 case IPPROTO_AH: /* is it possible? */ 1614 break; 1615 default: 1616 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); 1617 } 1618 #endif 1619 1620 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1621 if (nxt == IPPROTO_AH) 1622 elen = (ip6e.ip6e_len + 2) << 2; 1623 else 1624 elen = (ip6e.ip6e_len + 1) << 3; 1625 1626 MGET(n, M_DONTWAIT, MT_DATA); 1627 if (n && elen >= MLEN) { 1628 MCLGET(n, M_DONTWAIT); 1629 if ((n->m_flags & M_EXT) == 0) { 1630 m_free(n); 1631 n = NULL; 1632 } 1633 } 1634 if (!n) 1635 return NULL; 1636 1637 n->m_len = 0; 1638 if (elen >= M_TRAILINGSPACE(n)) { 1639 m_free(n); 1640 return NULL; 1641 } 1642 1643 m_copydata(m, off, elen, mtod(n, caddr_t)); 1644 n->m_len = elen; 1645 return n; 1646 } 1647 #endif 1648 1649 /* 1650 * Get pointer to the previous header followed by the header 1651 * currently processed. 1652 * XXX: This function supposes that 1653 * M includes all headers, 1654 * the next header field and the header length field of each header 1655 * are valid, and 1656 * the sum of each header length equals to OFF. 1657 * Because of these assumptions, this function must be called very 1658 * carefully. Moreover, it will not be used in the near future when 1659 * we develop `neater' mechanism to process extension headers. 1660 */ 1661 char * 1662 ip6_get_prevhdr(struct mbuf *m, int off) 1663 { 1664 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1665 1666 if (off == sizeof(struct ip6_hdr)) 1667 return (&ip6->ip6_nxt); 1668 else { 1669 int len, nxt; 1670 struct ip6_ext *ip6e = NULL; 1671 1672 nxt = ip6->ip6_nxt; 1673 len = sizeof(struct ip6_hdr); 1674 while (len < off) { 1675 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len); 1676 1677 switch (nxt) { 1678 case IPPROTO_FRAGMENT: 1679 len += sizeof(struct ip6_frag); 1680 break; 1681 case IPPROTO_AH: 1682 len += (ip6e->ip6e_len + 2) << 2; 1683 break; 1684 default: 1685 len += (ip6e->ip6e_len + 1) << 3; 1686 break; 1687 } 1688 nxt = ip6e->ip6e_nxt; 1689 } 1690 if (ip6e) 1691 return (&ip6e->ip6e_nxt); 1692 else 1693 return NULL; 1694 } 1695 } 1696 1697 /* 1698 * get next header offset. m will be retained. 1699 */ 1700 int 1701 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp) 1702 { 1703 struct ip6_hdr ip6; 1704 struct ip6_ext ip6e; 1705 struct ip6_frag fh; 1706 1707 /* just in case */ 1708 if (m == NULL) 1709 panic("ip6_nexthdr: m == NULL"); 1710 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1711 return -1; 1712 1713 switch (proto) { 1714 case IPPROTO_IPV6: 1715 if (m->m_pkthdr.len < off + sizeof(ip6)) 1716 return -1; 1717 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1718 if (nxtp) 1719 *nxtp = ip6.ip6_nxt; 1720 off += sizeof(ip6); 1721 return off; 1722 1723 case IPPROTO_FRAGMENT: 1724 /* 1725 * terminate parsing if it is not the first fragment, 1726 * it does not make sense to parse through it. 1727 */ 1728 if (m->m_pkthdr.len < off + sizeof(fh)) 1729 return -1; 1730 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1731 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1732 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1733 return -1; 1734 if (nxtp) 1735 *nxtp = fh.ip6f_nxt; 1736 off += sizeof(struct ip6_frag); 1737 return off; 1738 1739 case IPPROTO_AH: 1740 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1741 return -1; 1742 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1743 if (nxtp) 1744 *nxtp = ip6e.ip6e_nxt; 1745 off += (ip6e.ip6e_len + 2) << 2; 1746 return off; 1747 1748 case IPPROTO_HOPOPTS: 1749 case IPPROTO_ROUTING: 1750 case IPPROTO_DSTOPTS: 1751 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1752 return -1; 1753 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1754 if (nxtp) 1755 *nxtp = ip6e.ip6e_nxt; 1756 off += (ip6e.ip6e_len + 1) << 3; 1757 return off; 1758 1759 case IPPROTO_NONE: 1760 case IPPROTO_ESP: 1761 case IPPROTO_IPCOMP: 1762 /* give up */ 1763 return -1; 1764 1765 default: 1766 return -1; 1767 } 1768 1769 return -1; 1770 } 1771 1772 /* 1773 * get offset for the last header in the chain. m will be kept untainted. 1774 */ 1775 int 1776 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp) 1777 { 1778 int newoff; 1779 int nxt; 1780 1781 if (!nxtp) { 1782 nxt = -1; 1783 nxtp = &nxt; 1784 } 1785 while (1) { 1786 newoff = ip6_nexthdr(m, off, proto, nxtp); 1787 if (newoff < 0) 1788 return off; 1789 else if (newoff < off) 1790 return -1; /* invalid */ 1791 else if (newoff == off) 1792 return newoff; 1793 1794 off = newoff; 1795 proto = *nxtp; 1796 } 1797 } 1798 1799 static struct ip6aux * 1800 ip6_addaux(struct mbuf *m) 1801 { 1802 struct m_tag *mtag; 1803 1804 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1805 if (!mtag) { 1806 mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux), 1807 M_NOWAIT); 1808 if (mtag) { 1809 m_tag_prepend(m, mtag); 1810 bzero(mtag + 1, sizeof(struct ip6aux)); 1811 } 1812 } 1813 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1814 } 1815 1816 static struct ip6aux * 1817 ip6_findaux(struct mbuf *m) 1818 { 1819 struct m_tag *mtag; 1820 1821 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1822 return mtag ? (struct ip6aux *)(mtag + 1) : NULL; 1823 } 1824 1825 static void 1826 ip6_delaux(struct mbuf *m) 1827 { 1828 struct m_tag *mtag; 1829 1830 mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL); 1831 if (mtag) 1832 m_tag_delete(m, mtag); 1833 } 1834 1835 /* 1836 * System control for IP6 1837 */ 1838 1839 u_char inet6ctlerrmap[PRC_NCMDS] = { 1840 0, 0, 0, 0, 1841 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1842 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1843 EMSGSIZE, EHOSTUNREACH, 0, 0, 1844 0, 0, 0, 0, 1845 ENOPROTOOPT 1846 }; 1847