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 #include "opt_rss.h" 72 73 #include <sys/param.h> 74 #include <sys/systm.h> 75 #include <sys/malloc.h> 76 #include <sys/mbuf.h> 77 #include <sys/proc.h> 78 #include <sys/domain.h> 79 #include <sys/protosw.h> 80 #include <sys/sdt.h> 81 #include <sys/socket.h> 82 #include <sys/socketvar.h> 83 #include <sys/errno.h> 84 #include <sys/time.h> 85 #include <sys/kernel.h> 86 #include <sys/lock.h> 87 #include <sys/rmlock.h> 88 #include <sys/syslog.h> 89 90 #include <net/if.h> 91 #include <net/if_var.h> 92 #include <net/if_types.h> 93 #include <net/if_dl.h> 94 #include <net/route.h> 95 #include <net/netisr.h> 96 #include <net/rss_config.h> 97 #include <net/pfil.h> 98 #include <net/vnet.h> 99 100 #include <netinet/in.h> 101 #include <netinet/in_kdtrace.h> 102 #include <netinet/ip_var.h> 103 #include <netinet/in_systm.h> 104 #include <net/if_llatbl.h> 105 #ifdef INET 106 #include <netinet/ip.h> 107 #include <netinet/ip_icmp.h> 108 #endif /* INET */ 109 #include <netinet/ip6.h> 110 #include <netinet6/in6_var.h> 111 #include <netinet6/ip6_var.h> 112 #include <netinet/in_pcb.h> 113 #include <netinet/icmp6.h> 114 #include <netinet6/scope6_var.h> 115 #include <netinet6/in6_ifattach.h> 116 #include <netinet6/nd6.h> 117 #include <netinet6/in6_rss.h> 118 119 #ifdef IPSEC 120 #include <netipsec/ipsec.h> 121 #include <netinet6/ip6_ipsec.h> 122 #include <netipsec/ipsec6.h> 123 #endif /* IPSEC */ 124 125 #include <netinet6/ip6protosw.h> 126 127 extern struct domain inet6domain; 128 129 u_char ip6_protox[IPPROTO_MAX]; 130 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 131 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); 132 VNET_DEFINE(u_long, in6_ifaddrhmask); 133 134 static struct netisr_handler ip6_nh = { 135 .nh_name = "ip6", 136 .nh_handler = ip6_input, 137 .nh_proto = NETISR_IPV6, 138 #ifdef RSS 139 .nh_m2cpuid = rss_soft_m2cpuid_v6, 140 .nh_policy = NETISR_POLICY_CPU, 141 .nh_dispatch = NETISR_DISPATCH_HYBRID, 142 #else 143 .nh_policy = NETISR_POLICY_FLOW, 144 #endif 145 }; 146 147 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch); 148 #define V_in6_tmpaddrtimer_ch VNET(in6_tmpaddrtimer_ch) 149 150 VNET_DEFINE(struct pfil_head, inet6_pfil_hook); 151 152 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 153 VNET_PCPUSTAT_SYSINIT(ip6stat); 154 #ifdef VIMAGE 155 VNET_PCPUSTAT_SYSUNINIT(ip6stat); 156 #endif /* VIMAGE */ 157 158 struct rmlock in6_ifaddr_lock; 159 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 160 161 static void ip6_init2(void *); 162 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 163 #ifdef PULLDOWN_TEST 164 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int); 165 #endif 166 167 /* 168 * IP6 initialization: fill in IP6 protocol switch table. 169 * All protocols not implemented in kernel go to raw IP6 protocol handler. 170 */ 171 void 172 ip6_init(void) 173 { 174 struct protosw *pr; 175 int i; 176 177 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 178 &V_ip6_auto_linklocal); 179 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 180 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 181 182 TAILQ_INIT(&V_in6_ifaddrhead); 183 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 184 &V_in6_ifaddrhmask); 185 186 /* Initialize packet filter hooks. */ 187 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF; 188 V_inet6_pfil_hook.ph_af = AF_INET6; 189 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0) 190 printf("%s: WARNING: unable to register pfil hook, " 191 "error %d\n", __func__, i); 192 193 scope6_init(); 194 addrsel_policy_init(); 195 nd6_init(); 196 frag6_init(); 197 198 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 199 200 /* Skip global initialization stuff for non-default instances. */ 201 if (!IS_DEFAULT_VNET(curvnet)) 202 return; 203 204 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 205 if (pr == NULL) 206 panic("ip6_init"); 207 208 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 209 for (i = 0; i < IPPROTO_MAX; i++) 210 ip6_protox[i] = pr - inet6sw; 211 /* 212 * Cycle through IP protocols and put them into the appropriate place 213 * in ip6_protox[]. 214 */ 215 for (pr = inet6domain.dom_protosw; 216 pr < inet6domain.dom_protoswNPROTOSW; pr++) 217 if (pr->pr_domain->dom_family == PF_INET6 && 218 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 219 /* Be careful to only index valid IP protocols. */ 220 if (pr->pr_protocol < IPPROTO_MAX) 221 ip6_protox[pr->pr_protocol] = pr - inet6sw; 222 } 223 224 netisr_register(&ip6_nh); 225 } 226 227 /* 228 * The protocol to be inserted into ip6_protox[] must be already registered 229 * in inet6sw[], either statically or through pf_proto_register(). 230 */ 231 int 232 ip6proto_register(short ip6proto) 233 { 234 struct protosw *pr; 235 236 /* Sanity checks. */ 237 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 238 return (EPROTONOSUPPORT); 239 240 /* 241 * The protocol slot must not be occupied by another protocol 242 * already. An index pointing to IPPROTO_RAW is unused. 243 */ 244 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 245 if (pr == NULL) 246 return (EPFNOSUPPORT); 247 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 248 return (EEXIST); 249 250 /* 251 * Find the protocol position in inet6sw[] and set the index. 252 */ 253 for (pr = inet6domain.dom_protosw; 254 pr < inet6domain.dom_protoswNPROTOSW; pr++) { 255 if (pr->pr_domain->dom_family == PF_INET6 && 256 pr->pr_protocol && pr->pr_protocol == ip6proto) { 257 ip6_protox[pr->pr_protocol] = pr - inet6sw; 258 return (0); 259 } 260 } 261 return (EPROTONOSUPPORT); 262 } 263 264 int 265 ip6proto_unregister(short ip6proto) 266 { 267 struct protosw *pr; 268 269 /* Sanity checks. */ 270 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 271 return (EPROTONOSUPPORT); 272 273 /* Check if the protocol was indeed registered. */ 274 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 275 if (pr == NULL) 276 return (EPFNOSUPPORT); 277 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 278 return (ENOENT); 279 280 /* Reset the protocol slot to IPPROTO_RAW. */ 281 ip6_protox[ip6proto] = pr - inet6sw; 282 return (0); 283 } 284 285 #ifdef VIMAGE 286 void 287 ip6_destroy() 288 { 289 int i; 290 291 if ((i = pfil_head_unregister(&V_inet6_pfil_hook)) != 0) 292 printf("%s: WARNING: unable to unregister pfil hook, " 293 "error %d\n", __func__, i); 294 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 295 nd6_destroy(); 296 callout_drain(&V_in6_tmpaddrtimer_ch); 297 } 298 #endif 299 300 static int 301 ip6_init2_vnet(const void *unused __unused) 302 { 303 304 /* nd6_timer_init */ 305 callout_init(&V_nd6_timer_ch, 0); 306 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); 307 308 /* timer for regeneranation of temporary addresses randomize ID */ 309 callout_init(&V_in6_tmpaddrtimer_ch, 0); 310 callout_reset(&V_in6_tmpaddrtimer_ch, 311 (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor - 312 V_ip6_temp_regen_advance) * hz, 313 in6_tmpaddrtimer, curvnet); 314 315 return (0); 316 } 317 318 static void 319 ip6_init2(void *dummy) 320 { 321 322 ip6_init2_vnet(NULL); 323 } 324 325 /* cheat */ 326 /* This must be after route_init(), which is now SI_ORDER_THIRD */ 327 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL); 328 329 static int 330 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off, 331 int *nxt, int *ours) 332 { 333 struct ip6_hdr *ip6; 334 struct ip6_hbh *hbh; 335 336 if (ip6_hopopts_input(plen, rtalert, &m, off)) { 337 #if 0 /*touches NULL pointer*/ 338 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 339 #endif 340 goto out; /* m have already been freed */ 341 } 342 343 /* adjust pointer */ 344 ip6 = mtod(m, struct ip6_hdr *); 345 346 /* 347 * if the payload length field is 0 and the next header field 348 * indicates Hop-by-Hop Options header, then a Jumbo Payload 349 * option MUST be included. 350 */ 351 if (ip6->ip6_plen == 0 && *plen == 0) { 352 /* 353 * Note that if a valid jumbo payload option is 354 * contained, ip6_hopopts_input() must set a valid 355 * (non-zero) payload length to the variable plen. 356 */ 357 IP6STAT_INC(ip6s_badoptions); 358 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 359 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 360 icmp6_error(m, ICMP6_PARAM_PROB, 361 ICMP6_PARAMPROB_HEADER, 362 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 363 goto out; 364 } 365 #ifndef PULLDOWN_TEST 366 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 367 hbh = (struct ip6_hbh *)(ip6 + 1); 368 #else 369 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 370 sizeof(struct ip6_hbh)); 371 if (hbh == NULL) { 372 IP6STAT_INC(ip6s_tooshort); 373 goto out; 374 } 375 #endif 376 *nxt = hbh->ip6h_nxt; 377 378 /* 379 * If we are acting as a router and the packet contains a 380 * router alert option, see if we know the option value. 381 * Currently, we only support the option value for MLD, in which 382 * case we should pass the packet to the multicast routing 383 * daemon. 384 */ 385 if (*rtalert != ~0) { 386 switch (*rtalert) { 387 case IP6OPT_RTALERT_MLD: 388 if (V_ip6_forwarding) 389 *ours = 1; 390 break; 391 default: 392 /* 393 * RFC2711 requires unrecognized values must be 394 * silently ignored. 395 */ 396 break; 397 } 398 } 399 400 return (0); 401 402 out: 403 return (1); 404 } 405 406 void 407 ip6_input(struct mbuf *m) 408 { 409 struct in6_addr odst; 410 struct ip6_hdr *ip6; 411 struct in6_ifaddr *ia; 412 u_int32_t plen; 413 u_int32_t rtalert = ~0; 414 int off = sizeof(struct ip6_hdr), nest; 415 int nxt, ours = 0; 416 int srcrt = 0; 417 418 #ifdef IPSEC 419 /* 420 * should the inner packet be considered authentic? 421 * see comment in ah4_input(). 422 * NB: m cannot be NULL when passed to the input routine 423 */ 424 425 m->m_flags &= ~M_AUTHIPHDR; 426 m->m_flags &= ~M_AUTHIPDGM; 427 428 #endif /* IPSEC */ 429 430 if (m->m_flags & M_FASTFWD_OURS) { 431 /* 432 * Firewall changed destination to local. 433 */ 434 m->m_flags &= ~M_FASTFWD_OURS; 435 ours = 1; 436 ip6 = mtod(m, struct ip6_hdr *); 437 goto hbhcheck; 438 } 439 440 /* 441 * mbuf statistics 442 */ 443 if (m->m_flags & M_EXT) { 444 if (m->m_next) 445 IP6STAT_INC(ip6s_mext2m); 446 else 447 IP6STAT_INC(ip6s_mext1); 448 } else { 449 if (m->m_next) { 450 if (m->m_flags & M_LOOP) { 451 IP6STAT_INC(ip6s_m2m[V_loif->if_index]); 452 } else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX) 453 IP6STAT_INC( 454 ip6s_m2m[m->m_pkthdr.rcvif->if_index]); 455 else 456 IP6STAT_INC(ip6s_m2m[0]); 457 } else 458 IP6STAT_INC(ip6s_m1); 459 } 460 461 /* drop the packet if IPv6 operation is disabled on the IF */ 462 if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) 463 goto bad; 464 465 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive); 466 IP6STAT_INC(ip6s_total); 467 468 #ifndef PULLDOWN_TEST 469 /* 470 * L2 bridge code and some other code can return mbuf chain 471 * that does not conform to KAME requirement. too bad. 472 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 473 */ 474 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 475 struct mbuf *n; 476 477 if (m->m_pkthdr.len > MHLEN) 478 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 479 else 480 n = m_gethdr(M_NOWAIT, MT_DATA); 481 if (n == NULL) { 482 m_freem(m); 483 return; /* ENOBUFS */ 484 } 485 486 m_move_pkthdr(n, m); 487 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 488 n->m_len = n->m_pkthdr.len; 489 m_freem(m); 490 m = n; 491 } 492 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */); 493 #endif 494 495 if (m->m_len < sizeof(struct ip6_hdr)) { 496 struct ifnet *inifp; 497 inifp = m->m_pkthdr.rcvif; 498 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 499 IP6STAT_INC(ip6s_toosmall); 500 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 501 return; 502 } 503 } 504 505 ip6 = mtod(m, struct ip6_hdr *); 506 507 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 508 IP6STAT_INC(ip6s_badvers); 509 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 510 goto bad; 511 } 512 513 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 514 515 IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6); 516 517 /* 518 * Check against address spoofing/corruption. 519 */ 520 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 521 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 522 /* 523 * XXX: "badscope" is not very suitable for a multicast source. 524 */ 525 IP6STAT_INC(ip6s_badscope); 526 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 527 goto bad; 528 } 529 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 530 !(m->m_flags & M_LOOP)) { 531 /* 532 * In this case, the packet should come from the loopback 533 * interface. However, we cannot just check the if_flags, 534 * because ip6_mloopback() passes the "actual" interface 535 * as the outgoing/incoming interface. 536 */ 537 IP6STAT_INC(ip6s_badscope); 538 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 539 goto bad; 540 } 541 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 542 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 543 /* 544 * RFC4291 2.7: 545 * Nodes must not originate a packet to a multicast address 546 * whose scop field contains the reserved value 0; if such 547 * a packet is received, it must be silently dropped. 548 */ 549 IP6STAT_INC(ip6s_badscope); 550 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 551 goto bad; 552 } 553 #ifdef ALTQ 554 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 555 /* packet is dropped by traffic conditioner */ 556 return; 557 } 558 #endif 559 /* 560 * The following check is not documented in specs. A malicious 561 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 562 * and bypass security checks (act as if it was from 127.0.0.1 by using 563 * IPv6 src ::ffff:127.0.0.1). Be cautious. 564 * 565 * This check chokes if we are in an SIIT cloud. As none of BSDs 566 * support IPv4-less kernel compilation, we cannot support SIIT 567 * environment at all. So, it makes more sense for us to reject any 568 * malicious packets for non-SIIT environment, than try to do a 569 * partial support for SIIT environment. 570 */ 571 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 572 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 573 IP6STAT_INC(ip6s_badscope); 574 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 575 goto bad; 576 } 577 #if 0 578 /* 579 * Reject packets with IPv4 compatible addresses (auto tunnel). 580 * 581 * The code forbids auto tunnel relay case in RFC1933 (the check is 582 * stronger than RFC1933). We may want to re-enable it if mech-xx 583 * is revised to forbid relaying case. 584 */ 585 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 586 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 587 IP6STAT_INC(ip6s_badscope); 588 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 589 goto bad; 590 } 591 #endif 592 #ifdef IPSEC 593 /* 594 * Bypass packet filtering for packets previously handled by IPsec. 595 */ 596 if (ip6_ipsec_filtertunnel(m)) 597 goto passin; 598 #endif /* IPSEC */ 599 600 /* 601 * Run through list of hooks for input packets. 602 * 603 * NB: Beware of the destination address changing 604 * (e.g. by NAT rewriting). When this happens, 605 * tell ip6_forward to do the right thing. 606 */ 607 odst = ip6->ip6_dst; 608 609 /* Jump over all PFIL processing if hooks are not active. */ 610 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 611 goto passin; 612 613 if (pfil_run_hooks(&V_inet6_pfil_hook, &m, 614 m->m_pkthdr.rcvif, PFIL_IN, NULL)) 615 return; 616 if (m == NULL) /* consumed by filter */ 617 return; 618 ip6 = mtod(m, struct ip6_hdr *); 619 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 620 621 if (m->m_flags & M_FASTFWD_OURS) { 622 m->m_flags &= ~M_FASTFWD_OURS; 623 ours = 1; 624 goto hbhcheck; 625 } 626 if ((m->m_flags & M_IP6_NEXTHOP) && 627 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 628 /* 629 * Directly ship the packet on. This allows forwarding 630 * packets originally destined to us to some other directly 631 * connected host. 632 */ 633 ip6_forward(m, 1); 634 return; 635 } 636 637 passin: 638 /* 639 * Disambiguate address scope zones (if there is ambiguity). 640 * We first make sure that the original source or destination address 641 * is not in our internal form for scoped addresses. Such addresses 642 * are not necessarily invalid spec-wise, but we cannot accept them due 643 * to the usage conflict. 644 * in6_setscope() then also checks and rejects the cases where src or 645 * dst are the loopback address and the receiving interface 646 * is not loopback. 647 */ 648 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 649 IP6STAT_INC(ip6s_badscope); /* XXX */ 650 goto bad; 651 } 652 if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) || 653 in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) { 654 IP6STAT_INC(ip6s_badscope); 655 goto bad; 656 } 657 /* 658 * Multicast check. Assume packet is for us to avoid 659 * prematurely taking locks. 660 */ 661 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 662 ours = 1; 663 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast); 664 goto hbhcheck; 665 } 666 /* 667 * Unicast check 668 * XXX: For now we keep link-local IPv6 addresses with embedded 669 * scope zone id, therefore we use zero zoneid here. 670 */ 671 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */); 672 if (ia != NULL) { 673 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 674 char ip6bufs[INET6_ADDRSTRLEN]; 675 char ip6bufd[INET6_ADDRSTRLEN]; 676 /* address is not ready, so discard the packet. */ 677 nd6log((LOG_INFO, 678 "ip6_input: packet to an unready address %s->%s\n", 679 ip6_sprintf(ip6bufs, &ip6->ip6_src), 680 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 681 ifa_free(&ia->ia_ifa); 682 goto bad; 683 } 684 /* Count the packet in the ip address stats */ 685 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 686 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 687 ifa_free(&ia->ia_ifa); 688 ours = 1; 689 goto hbhcheck; 690 } 691 692 /* 693 * Now there is no reason to process the packet if it's not our own 694 * and we're not a router. 695 */ 696 if (!V_ip6_forwarding) { 697 IP6STAT_INC(ip6s_cantforward); 698 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 699 goto bad; 700 } 701 702 hbhcheck: 703 /* 704 * Process Hop-by-Hop options header if it's contained. 705 * m may be modified in ip6_hopopts_input(). 706 * If a JumboPayload option is included, plen will also be modified. 707 */ 708 plen = (u_int32_t)ntohs(ip6->ip6_plen); 709 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 710 if (ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours) != 0) 711 return; 712 } else 713 nxt = ip6->ip6_nxt; 714 715 /* 716 * Check that the amount of data in the buffers 717 * is as at least much as the IPv6 header would have us expect. 718 * Trim mbufs if longer than we expect. 719 * Drop packet if shorter than we expect. 720 */ 721 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 722 IP6STAT_INC(ip6s_tooshort); 723 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 724 goto bad; 725 } 726 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 727 if (m->m_len == m->m_pkthdr.len) { 728 m->m_len = sizeof(struct ip6_hdr) + plen; 729 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 730 } else 731 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 732 } 733 734 /* 735 * Forward if desirable. 736 */ 737 if (V_ip6_mrouter && 738 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 739 /* 740 * If we are acting as a multicast router, all 741 * incoming multicast packets are passed to the 742 * kernel-level multicast forwarding function. 743 * The packet is returned (relatively) intact; if 744 * ip6_mforward() returns a non-zero value, the packet 745 * must be discarded, else it may be accepted below. 746 * 747 * XXX TODO: Check hlim and multicast scope here to avoid 748 * unnecessarily calling into ip6_mforward(). 749 */ 750 if (ip6_mforward && 751 ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) { 752 IP6STAT_INC(ip6s_cantforward); 753 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 754 goto bad; 755 } 756 } else if (!ours) { 757 ip6_forward(m, srcrt); 758 return; 759 } 760 761 ip6 = mtod(m, struct ip6_hdr *); 762 763 /* 764 * Malicious party may be able to use IPv4 mapped addr to confuse 765 * tcp/udp stack and bypass security checks (act as if it was from 766 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 767 * 768 * For SIIT end node behavior, you may want to disable the check. 769 * However, you will become vulnerable to attacks using IPv4 mapped 770 * source. 771 */ 772 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 773 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 774 IP6STAT_INC(ip6s_badscope); 775 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 776 goto bad; 777 } 778 779 /* 780 * Tell launch routine the next header 781 */ 782 IP6STAT_INC(ip6s_delivered); 783 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_deliver); 784 nest = 0; 785 786 while (nxt != IPPROTO_DONE) { 787 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 788 IP6STAT_INC(ip6s_toomanyhdr); 789 goto bad; 790 } 791 792 /* 793 * protection against faulty packet - there should be 794 * more sanity checks in header chain processing. 795 */ 796 if (m->m_pkthdr.len < off) { 797 IP6STAT_INC(ip6s_tooshort); 798 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 799 goto bad; 800 } 801 802 #ifdef IPSEC 803 /* 804 * enforce IPsec policy checking if we are seeing last header. 805 * note that we do not visit this with protocols with pcb layer 806 * code - like udp/tcp/raw ip. 807 */ 808 if (ip6_ipsec_input(m, nxt)) 809 goto bad; 810 #endif /* IPSEC */ 811 812 /* 813 * Use mbuf flags to propagate Router Alert option to 814 * ICMPv6 layer, as hop-by-hop options have been stripped. 815 */ 816 if (nxt == IPPROTO_ICMPV6 && rtalert != ~0) 817 m->m_flags |= M_RTALERT_MLD; 818 819 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 820 } 821 return; 822 bad: 823 m_freem(m); 824 } 825 826 /* 827 * Hop-by-Hop options header processing. If a valid jumbo payload option is 828 * included, the real payload length will be stored in plenp. 829 * 830 * rtalertp - XXX: should be stored more smart way 831 */ 832 static int 833 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 834 struct mbuf **mp, int *offp) 835 { 836 struct mbuf *m = *mp; 837 int off = *offp, hbhlen; 838 struct ip6_hbh *hbh; 839 840 /* validation of the length of the header */ 841 #ifndef PULLDOWN_TEST 842 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); 843 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 844 hbhlen = (hbh->ip6h_len + 1) << 3; 845 846 IP6_EXTHDR_CHECK(m, off, hbhlen, -1); 847 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 848 #else 849 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, 850 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); 851 if (hbh == NULL) { 852 IP6STAT_INC(ip6s_tooshort); 853 return -1; 854 } 855 hbhlen = (hbh->ip6h_len + 1) << 3; 856 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 857 hbhlen); 858 if (hbh == NULL) { 859 IP6STAT_INC(ip6s_tooshort); 860 return -1; 861 } 862 #endif 863 off += hbhlen; 864 hbhlen -= sizeof(struct ip6_hbh); 865 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 866 hbhlen, rtalertp, plenp) < 0) 867 return (-1); 868 869 *offp = off; 870 *mp = m; 871 return (0); 872 } 873 874 /* 875 * Search header for all Hop-by-hop options and process each option. 876 * This function is separate from ip6_hopopts_input() in order to 877 * handle a case where the sending node itself process its hop-by-hop 878 * options header. In such a case, the function is called from ip6_output(). 879 * 880 * The function assumes that hbh header is located right after the IPv6 header 881 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 882 * opthead + hbhlen is located in contiguous memory region. 883 */ 884 int 885 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 886 u_int32_t *rtalertp, u_int32_t *plenp) 887 { 888 struct ip6_hdr *ip6; 889 int optlen = 0; 890 u_int8_t *opt = opthead; 891 u_int16_t rtalert_val; 892 u_int32_t jumboplen; 893 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 894 895 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 896 switch (*opt) { 897 case IP6OPT_PAD1: 898 optlen = 1; 899 break; 900 case IP6OPT_PADN: 901 if (hbhlen < IP6OPT_MINLEN) { 902 IP6STAT_INC(ip6s_toosmall); 903 goto bad; 904 } 905 optlen = *(opt + 1) + 2; 906 break; 907 case IP6OPT_ROUTER_ALERT: 908 /* XXX may need check for alignment */ 909 if (hbhlen < IP6OPT_RTALERT_LEN) { 910 IP6STAT_INC(ip6s_toosmall); 911 goto bad; 912 } 913 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 914 /* XXX stat */ 915 icmp6_error(m, ICMP6_PARAM_PROB, 916 ICMP6_PARAMPROB_HEADER, 917 erroff + opt + 1 - opthead); 918 return (-1); 919 } 920 optlen = IP6OPT_RTALERT_LEN; 921 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 922 *rtalertp = ntohs(rtalert_val); 923 break; 924 case IP6OPT_JUMBO: 925 /* XXX may need check for alignment */ 926 if (hbhlen < IP6OPT_JUMBO_LEN) { 927 IP6STAT_INC(ip6s_toosmall); 928 goto bad; 929 } 930 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 931 /* XXX stat */ 932 icmp6_error(m, ICMP6_PARAM_PROB, 933 ICMP6_PARAMPROB_HEADER, 934 erroff + opt + 1 - opthead); 935 return (-1); 936 } 937 optlen = IP6OPT_JUMBO_LEN; 938 939 /* 940 * IPv6 packets that have non 0 payload length 941 * must not contain a jumbo payload option. 942 */ 943 ip6 = mtod(m, struct ip6_hdr *); 944 if (ip6->ip6_plen) { 945 IP6STAT_INC(ip6s_badoptions); 946 icmp6_error(m, ICMP6_PARAM_PROB, 947 ICMP6_PARAMPROB_HEADER, 948 erroff + opt - opthead); 949 return (-1); 950 } 951 952 /* 953 * We may see jumbolen in unaligned location, so 954 * we'd need to perform bcopy(). 955 */ 956 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 957 jumboplen = (u_int32_t)htonl(jumboplen); 958 959 #if 1 960 /* 961 * if there are multiple jumbo payload options, 962 * *plenp will be non-zero and the packet will be 963 * rejected. 964 * the behavior may need some debate in ipngwg - 965 * multiple options does not make sense, however, 966 * there's no explicit mention in specification. 967 */ 968 if (*plenp != 0) { 969 IP6STAT_INC(ip6s_badoptions); 970 icmp6_error(m, ICMP6_PARAM_PROB, 971 ICMP6_PARAMPROB_HEADER, 972 erroff + opt + 2 - opthead); 973 return (-1); 974 } 975 #endif 976 977 /* 978 * jumbo payload length must be larger than 65535. 979 */ 980 if (jumboplen <= IPV6_MAXPACKET) { 981 IP6STAT_INC(ip6s_badoptions); 982 icmp6_error(m, ICMP6_PARAM_PROB, 983 ICMP6_PARAMPROB_HEADER, 984 erroff + opt + 2 - opthead); 985 return (-1); 986 } 987 *plenp = jumboplen; 988 989 break; 990 default: /* unknown option */ 991 if (hbhlen < IP6OPT_MINLEN) { 992 IP6STAT_INC(ip6s_toosmall); 993 goto bad; 994 } 995 optlen = ip6_unknown_opt(opt, m, 996 erroff + opt - opthead); 997 if (optlen == -1) 998 return (-1); 999 optlen += 2; 1000 break; 1001 } 1002 } 1003 1004 return (0); 1005 1006 bad: 1007 m_freem(m); 1008 return (-1); 1009 } 1010 1011 /* 1012 * Unknown option processing. 1013 * The third argument `off' is the offset from the IPv6 header to the option, 1014 * which is necessary if the IPv6 header the and option header and IPv6 header 1015 * is not contiguous in order to return an ICMPv6 error. 1016 */ 1017 int 1018 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1019 { 1020 struct ip6_hdr *ip6; 1021 1022 switch (IP6OPT_TYPE(*optp)) { 1023 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1024 return ((int)*(optp + 1)); 1025 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1026 m_freem(m); 1027 return (-1); 1028 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1029 IP6STAT_INC(ip6s_badoptions); 1030 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1031 return (-1); 1032 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1033 IP6STAT_INC(ip6s_badoptions); 1034 ip6 = mtod(m, struct ip6_hdr *); 1035 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1036 (m->m_flags & (M_BCAST|M_MCAST))) 1037 m_freem(m); 1038 else 1039 icmp6_error(m, ICMP6_PARAM_PROB, 1040 ICMP6_PARAMPROB_OPTION, off); 1041 return (-1); 1042 } 1043 1044 m_freem(m); /* XXX: NOTREACHED */ 1045 return (-1); 1046 } 1047 1048 /* 1049 * Create the "control" list for this pcb. 1050 * These functions will not modify mbuf chain at all. 1051 * 1052 * With KAME mbuf chain restriction: 1053 * The routine will be called from upper layer handlers like tcp6_input(). 1054 * Thus the routine assumes that the caller (tcp6_input) have already 1055 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the 1056 * very first mbuf on the mbuf chain. 1057 * 1058 * ip6_savecontrol_v4 will handle those options that are possible to be 1059 * set on a v4-mapped socket. 1060 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1061 * options and handle the v6-only ones itself. 1062 */ 1063 struct mbuf ** 1064 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1065 int *v4only) 1066 { 1067 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1068 1069 #ifdef SO_TIMESTAMP 1070 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1071 struct timeval tv; 1072 1073 microtime(&tv); 1074 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1075 SCM_TIMESTAMP, SOL_SOCKET); 1076 if (*mp) 1077 mp = &(*mp)->m_next; 1078 } 1079 #endif 1080 1081 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1082 /* RFC 2292 sec. 5 */ 1083 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1084 struct in6_pktinfo pi6; 1085 1086 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1087 #ifdef INET 1088 struct ip *ip; 1089 1090 ip = mtod(m, struct ip *); 1091 pi6.ipi6_addr.s6_addr32[0] = 0; 1092 pi6.ipi6_addr.s6_addr32[1] = 0; 1093 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1094 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1095 #else 1096 /* We won't hit this code */ 1097 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1098 #endif 1099 } else { 1100 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1101 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1102 } 1103 pi6.ipi6_ifindex = 1104 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1105 1106 *mp = sbcreatecontrol((caddr_t) &pi6, 1107 sizeof(struct in6_pktinfo), 1108 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1109 if (*mp) 1110 mp = &(*mp)->m_next; 1111 } 1112 1113 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1114 int hlim; 1115 1116 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1117 #ifdef INET 1118 struct ip *ip; 1119 1120 ip = mtod(m, struct ip *); 1121 hlim = ip->ip_ttl; 1122 #else 1123 /* We won't hit this code */ 1124 hlim = 0; 1125 #endif 1126 } else { 1127 hlim = ip6->ip6_hlim & 0xff; 1128 } 1129 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1130 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1131 IPPROTO_IPV6); 1132 if (*mp) 1133 mp = &(*mp)->m_next; 1134 } 1135 1136 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1137 int tclass; 1138 1139 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1140 #ifdef INET 1141 struct ip *ip; 1142 1143 ip = mtod(m, struct ip *); 1144 tclass = ip->ip_tos; 1145 #else 1146 /* We won't hit this code */ 1147 tclass = 0; 1148 #endif 1149 } else { 1150 u_int32_t flowinfo; 1151 1152 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1153 flowinfo >>= 20; 1154 tclass = flowinfo & 0xff; 1155 } 1156 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int), 1157 IPV6_TCLASS, IPPROTO_IPV6); 1158 if (*mp) 1159 mp = &(*mp)->m_next; 1160 } 1161 1162 if (v4only != NULL) { 1163 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1164 *v4only = 1; 1165 } else { 1166 *v4only = 0; 1167 } 1168 } 1169 1170 return (mp); 1171 } 1172 1173 void 1174 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) 1175 { 1176 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1177 int v4only = 0; 1178 1179 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); 1180 if (v4only) 1181 return; 1182 1183 /* 1184 * IPV6_HOPOPTS socket option. Recall that we required super-user 1185 * privilege for the option (see ip6_ctloutput), but it might be too 1186 * strict, since there might be some hop-by-hop options which can be 1187 * returned to normal user. 1188 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1189 */ 1190 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { 1191 /* 1192 * Check if a hop-by-hop options header is contatined in the 1193 * received packet, and if so, store the options as ancillary 1194 * data. Note that a hop-by-hop options header must be 1195 * just after the IPv6 header, which is assured through the 1196 * IPv6 input processing. 1197 */ 1198 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1199 struct ip6_hbh *hbh; 1200 int hbhlen = 0; 1201 #ifdef PULLDOWN_TEST 1202 struct mbuf *ext; 1203 #endif 1204 1205 #ifndef PULLDOWN_TEST 1206 hbh = (struct ip6_hbh *)(ip6 + 1); 1207 hbhlen = (hbh->ip6h_len + 1) << 3; 1208 #else 1209 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), 1210 ip6->ip6_nxt); 1211 if (ext == NULL) { 1212 IP6STAT_INC(ip6s_tooshort); 1213 return; 1214 } 1215 hbh = mtod(ext, struct ip6_hbh *); 1216 hbhlen = (hbh->ip6h_len + 1) << 3; 1217 if (hbhlen != ext->m_len) { 1218 m_freem(ext); 1219 IP6STAT_INC(ip6s_tooshort); 1220 return; 1221 } 1222 #endif 1223 1224 /* 1225 * XXX: We copy the whole header even if a 1226 * jumbo payload option is included, the option which 1227 * is to be removed before returning according to 1228 * RFC2292. 1229 * Note: this constraint is removed in RFC3542 1230 */ 1231 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1232 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1233 IPPROTO_IPV6); 1234 if (*mp) 1235 mp = &(*mp)->m_next; 1236 #ifdef PULLDOWN_TEST 1237 m_freem(ext); 1238 #endif 1239 } 1240 } 1241 1242 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1243 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1244 1245 /* 1246 * Search for destination options headers or routing 1247 * header(s) through the header chain, and stores each 1248 * header as ancillary data. 1249 * Note that the order of the headers remains in 1250 * the chain of ancillary data. 1251 */ 1252 while (1) { /* is explicit loop prevention necessary? */ 1253 struct ip6_ext *ip6e = NULL; 1254 int elen; 1255 #ifdef PULLDOWN_TEST 1256 struct mbuf *ext = NULL; 1257 #endif 1258 1259 /* 1260 * if it is not an extension header, don't try to 1261 * pull it from the chain. 1262 */ 1263 switch (nxt) { 1264 case IPPROTO_DSTOPTS: 1265 case IPPROTO_ROUTING: 1266 case IPPROTO_HOPOPTS: 1267 case IPPROTO_AH: /* is it possible? */ 1268 break; 1269 default: 1270 goto loopend; 1271 } 1272 1273 #ifndef PULLDOWN_TEST 1274 if (off + sizeof(*ip6e) > m->m_len) 1275 goto loopend; 1276 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1277 if (nxt == IPPROTO_AH) 1278 elen = (ip6e->ip6e_len + 2) << 2; 1279 else 1280 elen = (ip6e->ip6e_len + 1) << 3; 1281 if (off + elen > m->m_len) 1282 goto loopend; 1283 #else 1284 ext = ip6_pullexthdr(m, off, nxt); 1285 if (ext == NULL) { 1286 IP6STAT_INC(ip6s_tooshort); 1287 return; 1288 } 1289 ip6e = mtod(ext, struct ip6_ext *); 1290 if (nxt == IPPROTO_AH) 1291 elen = (ip6e->ip6e_len + 2) << 2; 1292 else 1293 elen = (ip6e->ip6e_len + 1) << 3; 1294 if (elen != ext->m_len) { 1295 m_freem(ext); 1296 IP6STAT_INC(ip6s_tooshort); 1297 return; 1298 } 1299 #endif 1300 1301 switch (nxt) { 1302 case IPPROTO_DSTOPTS: 1303 if (!(in6p->inp_flags & IN6P_DSTOPTS)) 1304 break; 1305 1306 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1307 IS2292(in6p, 1308 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1309 IPPROTO_IPV6); 1310 if (*mp) 1311 mp = &(*mp)->m_next; 1312 break; 1313 case IPPROTO_ROUTING: 1314 if (!(in6p->inp_flags & IN6P_RTHDR)) 1315 break; 1316 1317 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1318 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), 1319 IPPROTO_IPV6); 1320 if (*mp) 1321 mp = &(*mp)->m_next; 1322 break; 1323 case IPPROTO_HOPOPTS: 1324 case IPPROTO_AH: /* is it possible? */ 1325 break; 1326 1327 default: 1328 /* 1329 * other cases have been filtered in the above. 1330 * none will visit this case. here we supply 1331 * the code just in case (nxt overwritten or 1332 * other cases). 1333 */ 1334 #ifdef PULLDOWN_TEST 1335 m_freem(ext); 1336 #endif 1337 goto loopend; 1338 1339 } 1340 1341 /* proceed with the next header. */ 1342 off += elen; 1343 nxt = ip6e->ip6e_nxt; 1344 ip6e = NULL; 1345 #ifdef PULLDOWN_TEST 1346 m_freem(ext); 1347 ext = NULL; 1348 #endif 1349 } 1350 loopend: 1351 ; 1352 } 1353 1354 if (in6p->inp_flags2 & INP_RECVFLOWID) { 1355 uint32_t flowid, flow_type; 1356 1357 flowid = m->m_pkthdr.flowid; 1358 flow_type = M_HASHTYPE_GET(m); 1359 1360 /* 1361 * XXX should handle the failure of one or the 1362 * other - don't populate both? 1363 */ 1364 *mp = sbcreatecontrol((caddr_t) &flowid, 1365 sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6); 1366 if (*mp) 1367 mp = &(*mp)->m_next; 1368 *mp = sbcreatecontrol((caddr_t) &flow_type, 1369 sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6); 1370 if (*mp) 1371 mp = &(*mp)->m_next; 1372 } 1373 1374 #ifdef RSS 1375 if (in6p->inp_flags2 & INP_RECVRSSBUCKETID) { 1376 uint32_t flowid, flow_type; 1377 uint32_t rss_bucketid; 1378 1379 flowid = m->m_pkthdr.flowid; 1380 flow_type = M_HASHTYPE_GET(m); 1381 1382 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1383 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1384 sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6); 1385 if (*mp) 1386 mp = &(*mp)->m_next; 1387 } 1388 } 1389 #endif 1390 1391 } 1392 #undef IS2292 1393 1394 void 1395 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1396 { 1397 struct socket *so; 1398 struct mbuf *m_mtu; 1399 struct ip6_mtuinfo mtuctl; 1400 1401 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1402 /* 1403 * Notify the error by sending IPV6_PATHMTU ancillary data if 1404 * application wanted to know the MTU value. 1405 * NOTE: we notify disconnected sockets, because some udp 1406 * applications keep sending sockets disconnected. 1407 * NOTE: our implementation doesn't notify connected sockets that has 1408 * foreign address that is different than given destination addresses 1409 * (this is permitted by RFC 3542). 1410 */ 1411 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1412 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1413 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1414 return; 1415 1416 mtuctl.ip6m_mtu = mtu; 1417 mtuctl.ip6m_addr = *dst; 1418 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1419 return; 1420 1421 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1422 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1423 return; 1424 1425 so = inp->inp_socket; 1426 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1427 == 0) { 1428 m_freem(m_mtu); 1429 /* XXX: should count statistics */ 1430 } else 1431 sorwakeup(so); 1432 } 1433 1434 #ifdef PULLDOWN_TEST 1435 /* 1436 * pull single extension header from mbuf chain. returns single mbuf that 1437 * contains the result, or NULL on error. 1438 */ 1439 static struct mbuf * 1440 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) 1441 { 1442 struct ip6_ext ip6e; 1443 size_t elen; 1444 struct mbuf *n; 1445 1446 #ifdef DIAGNOSTIC 1447 switch (nxt) { 1448 case IPPROTO_DSTOPTS: 1449 case IPPROTO_ROUTING: 1450 case IPPROTO_HOPOPTS: 1451 case IPPROTO_AH: /* is it possible? */ 1452 break; 1453 default: 1454 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); 1455 } 1456 #endif 1457 1458 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1459 if (nxt == IPPROTO_AH) 1460 elen = (ip6e.ip6e_len + 2) << 2; 1461 else 1462 elen = (ip6e.ip6e_len + 1) << 3; 1463 1464 if (elen > MLEN) 1465 n = m_getcl(M_NOWAIT, MT_DATA, 0); 1466 else 1467 n = m_get(M_NOWAIT, MT_DATA); 1468 if (n == NULL) 1469 return NULL; 1470 1471 m_copydata(m, off, elen, mtod(n, caddr_t)); 1472 n->m_len = elen; 1473 return n; 1474 } 1475 #endif 1476 1477 /* 1478 * Get pointer to the previous header followed by the header 1479 * currently processed. 1480 * XXX: This function supposes that 1481 * M includes all headers, 1482 * the next header field and the header length field of each header 1483 * are valid, and 1484 * the sum of each header length equals to OFF. 1485 * Because of these assumptions, this function must be called very 1486 * carefully. Moreover, it will not be used in the near future when 1487 * we develop `neater' mechanism to process extension headers. 1488 */ 1489 char * 1490 ip6_get_prevhdr(const struct mbuf *m, int off) 1491 { 1492 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1493 1494 if (off == sizeof(struct ip6_hdr)) 1495 return (&ip6->ip6_nxt); 1496 else { 1497 int len, nxt; 1498 struct ip6_ext *ip6e = NULL; 1499 1500 nxt = ip6->ip6_nxt; 1501 len = sizeof(struct ip6_hdr); 1502 while (len < off) { 1503 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len); 1504 1505 switch (nxt) { 1506 case IPPROTO_FRAGMENT: 1507 len += sizeof(struct ip6_frag); 1508 break; 1509 case IPPROTO_AH: 1510 len += (ip6e->ip6e_len + 2) << 2; 1511 break; 1512 default: 1513 len += (ip6e->ip6e_len + 1) << 3; 1514 break; 1515 } 1516 nxt = ip6e->ip6e_nxt; 1517 } 1518 if (ip6e) 1519 return (&ip6e->ip6e_nxt); 1520 else 1521 return NULL; 1522 } 1523 } 1524 1525 /* 1526 * get next header offset. m will be retained. 1527 */ 1528 int 1529 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1530 { 1531 struct ip6_hdr ip6; 1532 struct ip6_ext ip6e; 1533 struct ip6_frag fh; 1534 1535 /* just in case */ 1536 if (m == NULL) 1537 panic("ip6_nexthdr: m == NULL"); 1538 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1539 return -1; 1540 1541 switch (proto) { 1542 case IPPROTO_IPV6: 1543 if (m->m_pkthdr.len < off + sizeof(ip6)) 1544 return -1; 1545 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1546 if (nxtp) 1547 *nxtp = ip6.ip6_nxt; 1548 off += sizeof(ip6); 1549 return off; 1550 1551 case IPPROTO_FRAGMENT: 1552 /* 1553 * terminate parsing if it is not the first fragment, 1554 * it does not make sense to parse through it. 1555 */ 1556 if (m->m_pkthdr.len < off + sizeof(fh)) 1557 return -1; 1558 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1559 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1560 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1561 return -1; 1562 if (nxtp) 1563 *nxtp = fh.ip6f_nxt; 1564 off += sizeof(struct ip6_frag); 1565 return off; 1566 1567 case IPPROTO_AH: 1568 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1569 return -1; 1570 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1571 if (nxtp) 1572 *nxtp = ip6e.ip6e_nxt; 1573 off += (ip6e.ip6e_len + 2) << 2; 1574 return off; 1575 1576 case IPPROTO_HOPOPTS: 1577 case IPPROTO_ROUTING: 1578 case IPPROTO_DSTOPTS: 1579 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1580 return -1; 1581 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1582 if (nxtp) 1583 *nxtp = ip6e.ip6e_nxt; 1584 off += (ip6e.ip6e_len + 1) << 3; 1585 return off; 1586 1587 case IPPROTO_NONE: 1588 case IPPROTO_ESP: 1589 case IPPROTO_IPCOMP: 1590 /* give up */ 1591 return -1; 1592 1593 default: 1594 return -1; 1595 } 1596 1597 /* NOTREACHED */ 1598 } 1599 1600 /* 1601 * get offset for the last header in the chain. m will be kept untainted. 1602 */ 1603 int 1604 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1605 { 1606 int newoff; 1607 int nxt; 1608 1609 if (!nxtp) { 1610 nxt = -1; 1611 nxtp = &nxt; 1612 } 1613 while (1) { 1614 newoff = ip6_nexthdr(m, off, proto, nxtp); 1615 if (newoff < 0) 1616 return off; 1617 else if (newoff < off) 1618 return -1; /* invalid */ 1619 else if (newoff == off) 1620 return newoff; 1621 1622 off = newoff; 1623 proto = *nxtp; 1624 } 1625 } 1626 1627 /* 1628 * System control for IP6 1629 */ 1630 1631 u_char inet6ctlerrmap[PRC_NCMDS] = { 1632 0, 0, 0, 0, 1633 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1634 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1635 EMSGSIZE, EHOSTUNREACH, 0, 0, 1636 0, 0, 0, 0, 1637 ENOPROTOOPT 1638 }; 1639