ip6_input.c revision 338343
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: stable/11/sys/netinet6/ip6_input.c 338343 2018-08-28 07:24:09Z ae $"); 65 66#include "opt_inet.h" 67#include "opt_inet6.h" 68#include "opt_ipsec.h" 69#include "opt_route.h" 70#include "opt_rss.h" 71 72#include <sys/param.h> 73#include <sys/systm.h> 74#include <sys/hhook.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#include <sys/sysctl.h> 90 91#include <net/if.h> 92#include <net/if_var.h> 93#include <net/if_types.h> 94#include <net/if_dl.h> 95#include <net/route.h> 96#include <net/netisr.h> 97#include <net/rss_config.h> 98#include <net/pfil.h> 99#include <net/vnet.h> 100 101#include <netinet/in.h> 102#include <netinet/in_kdtrace.h> 103#include <netinet/ip_var.h> 104#include <netinet/in_systm.h> 105#include <net/if_llatbl.h> 106#ifdef INET 107#include <netinet/ip.h> 108#include <netinet/ip_icmp.h> 109#endif /* INET */ 110#include <netinet/ip6.h> 111#include <netinet6/in6_var.h> 112#include <netinet6/ip6_var.h> 113#include <netinet/in_pcb.h> 114#include <netinet/icmp6.h> 115#include <netinet6/scope6_var.h> 116#include <netinet6/in6_ifattach.h> 117#include <netinet6/mld6_var.h> 118#include <netinet6/nd6.h> 119#include <netinet6/in6_rss.h> 120 121#include <netipsec/ipsec_support.h> 122 123#include <netinet6/ip6protosw.h> 124 125extern struct domain inet6domain; 126 127u_char ip6_protox[IPPROTO_MAX]; 128VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 129VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); 130VNET_DEFINE(u_long, in6_ifaddrhmask); 131 132static struct netisr_handler ip6_nh = { 133 .nh_name = "ip6", 134 .nh_handler = ip6_input, 135 .nh_proto = NETISR_IPV6, 136#ifdef RSS 137 .nh_m2cpuid = rss_soft_m2cpuid_v6, 138 .nh_policy = NETISR_POLICY_CPU, 139 .nh_dispatch = NETISR_DISPATCH_HYBRID, 140#else 141 .nh_policy = NETISR_POLICY_FLOW, 142#endif 143}; 144 145static int 146sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 147{ 148 int error, qlimit; 149 150 netisr_getqlimit(&ip6_nh, &qlimit); 151 error = sysctl_handle_int(oidp, &qlimit, 0, req); 152 if (error || !req->newptr) 153 return (error); 154 if (qlimit < 1) 155 return (EINVAL); 156 return (netisr_setqlimit(&ip6_nh, qlimit)); 157} 158SYSCTL_DECL(_net_inet6_ip6); 159SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen, 160 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet6_intr_queue_maxlen, "I", 161 "Maximum size of the IPv6 input queue"); 162 163#ifdef RSS 164static struct netisr_handler ip6_direct_nh = { 165 .nh_name = "ip6_direct", 166 .nh_handler = ip6_direct_input, 167 .nh_proto = NETISR_IPV6_DIRECT, 168 .nh_m2cpuid = rss_soft_m2cpuid_v6, 169 .nh_policy = NETISR_POLICY_CPU, 170 .nh_dispatch = NETISR_DISPATCH_HYBRID, 171}; 172 173static int 174sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 175{ 176 int error, qlimit; 177 178 netisr_getqlimit(&ip6_direct_nh, &qlimit); 179 error = sysctl_handle_int(oidp, &qlimit, 0, req); 180 if (error || !req->newptr) 181 return (error); 182 if (qlimit < 1) 183 return (EINVAL); 184 return (netisr_setqlimit(&ip6_direct_nh, qlimit)); 185} 186SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 187 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, 188 "I", "Maximum size of the IPv6 direct input queue"); 189 190#endif 191 192VNET_DEFINE(struct pfil_head, inet6_pfil_hook); 193 194VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 195VNET_PCPUSTAT_SYSINIT(ip6stat); 196#ifdef VIMAGE 197VNET_PCPUSTAT_SYSUNINIT(ip6stat); 198#endif /* VIMAGE */ 199 200struct rmlock in6_ifaddr_lock; 201RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 202 203static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 204#ifdef PULLDOWN_TEST 205static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int); 206#endif 207 208/* 209 * IP6 initialization: fill in IP6 protocol switch table. 210 * All protocols not implemented in kernel go to raw IP6 protocol handler. 211 */ 212void 213ip6_init(void) 214{ 215 struct protosw *pr; 216 int i; 217 218 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 219 &V_ip6_auto_linklocal); 220 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 221 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 222 223 TAILQ_INIT(&V_in6_ifaddrhead); 224 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 225 &V_in6_ifaddrhmask); 226 227 /* Initialize packet filter hooks. */ 228 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF; 229 V_inet6_pfil_hook.ph_af = AF_INET6; 230 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0) 231 printf("%s: WARNING: unable to register pfil hook, " 232 "error %d\n", __func__, i); 233 234 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 235 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 236 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 237 printf("%s: WARNING: unable to register input helper hook\n", 238 __func__); 239 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 240 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 241 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 242 printf("%s: WARNING: unable to register output helper hook\n", 243 __func__); 244 245 scope6_init(); 246 addrsel_policy_init(); 247 nd6_init(); 248 frag6_init(); 249 250 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 251 252 /* Skip global initialization stuff for non-default instances. */ 253#ifdef VIMAGE 254 if (!IS_DEFAULT_VNET(curvnet)) { 255 netisr_register_vnet(&ip6_nh); 256#ifdef RSS 257 netisr_register_vnet(&ip6_direct_nh); 258#endif 259 return; 260 } 261#endif 262 263 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 264 if (pr == NULL) 265 panic("ip6_init"); 266 267 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 268 for (i = 0; i < IPPROTO_MAX; i++) 269 ip6_protox[i] = pr - inet6sw; 270 /* 271 * Cycle through IP protocols and put them into the appropriate place 272 * in ip6_protox[]. 273 */ 274 for (pr = inet6domain.dom_protosw; 275 pr < inet6domain.dom_protoswNPROTOSW; pr++) 276 if (pr->pr_domain->dom_family == PF_INET6 && 277 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 278 /* Be careful to only index valid IP protocols. */ 279 if (pr->pr_protocol < IPPROTO_MAX) 280 ip6_protox[pr->pr_protocol] = pr - inet6sw; 281 } 282 283 netisr_register(&ip6_nh); 284#ifdef RSS 285 netisr_register(&ip6_direct_nh); 286#endif 287} 288 289/* 290 * The protocol to be inserted into ip6_protox[] must be already registered 291 * in inet6sw[], either statically or through pf_proto_register(). 292 */ 293int 294ip6proto_register(short ip6proto) 295{ 296 struct protosw *pr; 297 298 /* Sanity checks. */ 299 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 300 return (EPROTONOSUPPORT); 301 302 /* 303 * The protocol slot must not be occupied by another protocol 304 * already. An index pointing to IPPROTO_RAW is unused. 305 */ 306 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 307 if (pr == NULL) 308 return (EPFNOSUPPORT); 309 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 310 return (EEXIST); 311 312 /* 313 * Find the protocol position in inet6sw[] and set the index. 314 */ 315 for (pr = inet6domain.dom_protosw; 316 pr < inet6domain.dom_protoswNPROTOSW; pr++) { 317 if (pr->pr_domain->dom_family == PF_INET6 && 318 pr->pr_protocol && pr->pr_protocol == ip6proto) { 319 ip6_protox[pr->pr_protocol] = pr - inet6sw; 320 return (0); 321 } 322 } 323 return (EPROTONOSUPPORT); 324} 325 326int 327ip6proto_unregister(short ip6proto) 328{ 329 struct protosw *pr; 330 331 /* Sanity checks. */ 332 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 333 return (EPROTONOSUPPORT); 334 335 /* Check if the protocol was indeed registered. */ 336 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 337 if (pr == NULL) 338 return (EPFNOSUPPORT); 339 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 340 return (ENOENT); 341 342 /* Reset the protocol slot to IPPROTO_RAW. */ 343 ip6_protox[ip6proto] = pr - inet6sw; 344 return (0); 345} 346 347#ifdef VIMAGE 348static void 349ip6_destroy(void *unused __unused) 350{ 351 struct ifaddr *ifa, *nifa; 352 struct ifnet *ifp; 353 int error; 354 355#ifdef RSS 356 netisr_unregister_vnet(&ip6_direct_nh); 357#endif 358 netisr_unregister_vnet(&ip6_nh); 359 360 if ((error = pfil_head_unregister(&V_inet6_pfil_hook)) != 0) 361 printf("%s: WARNING: unable to unregister pfil hook, " 362 "error %d\n", __func__, error); 363 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 364 if (error != 0) { 365 printf("%s: WARNING: unable to deregister input helper hook " 366 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 367 "error %d returned\n", __func__, error); 368 } 369 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 370 if (error != 0) { 371 printf("%s: WARNING: unable to deregister output helper hook " 372 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 373 "error %d returned\n", __func__, error); 374 } 375 376 /* Cleanup addresses. */ 377 IFNET_RLOCK(); 378 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 379 /* Cannot lock here - lock recursion. */ 380 /* IF_ADDR_LOCK(ifp); */ 381 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 382 383 if (ifa->ifa_addr->sa_family != AF_INET6) 384 continue; 385 in6_purgeaddr(ifa); 386 } 387 /* IF_ADDR_UNLOCK(ifp); */ 388 in6_ifdetach_destroy(ifp); 389 mld_domifdetach(ifp); 390 /* Make sure any routes are gone as well. */ 391 rt_flushifroutes_af(ifp, AF_INET6); 392 } 393 IFNET_RUNLOCK(); 394 395 nd6_destroy(); 396 in6_ifattach_destroy(); 397 398 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 399} 400 401VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 402#endif 403 404static int 405ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off, 406 int *nxt, int *ours) 407{ 408 struct ip6_hdr *ip6; 409 struct ip6_hbh *hbh; 410 411 if (ip6_hopopts_input(plen, rtalert, &m, off)) { 412#if 0 /*touches NULL pointer*/ 413 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 414#endif 415 goto out; /* m have already been freed */ 416 } 417 418 /* adjust pointer */ 419 ip6 = mtod(m, struct ip6_hdr *); 420 421 /* 422 * if the payload length field is 0 and the next header field 423 * indicates Hop-by-Hop Options header, then a Jumbo Payload 424 * option MUST be included. 425 */ 426 if (ip6->ip6_plen == 0 && *plen == 0) { 427 /* 428 * Note that if a valid jumbo payload option is 429 * contained, ip6_hopopts_input() must set a valid 430 * (non-zero) payload length to the variable plen. 431 */ 432 IP6STAT_INC(ip6s_badoptions); 433 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 434 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 435 icmp6_error(m, ICMP6_PARAM_PROB, 436 ICMP6_PARAMPROB_HEADER, 437 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 438 goto out; 439 } 440#ifndef PULLDOWN_TEST 441 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 442 hbh = (struct ip6_hbh *)(ip6 + 1); 443#else 444 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 445 sizeof(struct ip6_hbh)); 446 if (hbh == NULL) { 447 IP6STAT_INC(ip6s_tooshort); 448 goto out; 449 } 450#endif 451 *nxt = hbh->ip6h_nxt; 452 453 /* 454 * If we are acting as a router and the packet contains a 455 * router alert option, see if we know the option value. 456 * Currently, we only support the option value for MLD, in which 457 * case we should pass the packet to the multicast routing 458 * daemon. 459 */ 460 if (*rtalert != ~0) { 461 switch (*rtalert) { 462 case IP6OPT_RTALERT_MLD: 463 if (V_ip6_forwarding) 464 *ours = 1; 465 break; 466 default: 467 /* 468 * RFC2711 requires unrecognized values must be 469 * silently ignored. 470 */ 471 break; 472 } 473 } 474 475 return (0); 476 477out: 478 return (1); 479} 480 481#ifdef RSS 482/* 483 * IPv6 direct input routine. 484 * 485 * This is called when reinjecting completed fragments where 486 * all of the previous checking and book-keeping has been done. 487 */ 488void 489ip6_direct_input(struct mbuf *m) 490{ 491 int off, nxt; 492 int nest; 493 struct m_tag *mtag; 494 struct ip6_direct_ctx *ip6dc; 495 496 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 497 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 498 499 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 500 nxt = ip6dc->ip6dc_nxt; 501 off = ip6dc->ip6dc_off; 502 503 nest = 0; 504 505 m_tag_delete(m, mtag); 506 507 while (nxt != IPPROTO_DONE) { 508 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 509 IP6STAT_INC(ip6s_toomanyhdr); 510 goto bad; 511 } 512 513 /* 514 * protection against faulty packet - there should be 515 * more sanity checks in header chain processing. 516 */ 517 if (m->m_pkthdr.len < off) { 518 IP6STAT_INC(ip6s_tooshort); 519 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 520 goto bad; 521 } 522 523#if defined(IPSEC) || defined(IPSEC_SUPPORT) 524 if (IPSEC_ENABLED(ipv6)) { 525 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 526 return; 527 } 528#endif /* IPSEC */ 529 530 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 531 } 532 return; 533bad: 534 m_freem(m); 535} 536#endif 537 538void 539ip6_input(struct mbuf *m) 540{ 541 struct in6_addr odst; 542 struct ip6_hdr *ip6; 543 struct in6_ifaddr *ia; 544 struct ifnet *rcvif; 545 u_int32_t plen; 546 u_int32_t rtalert = ~0; 547 int off = sizeof(struct ip6_hdr), nest; 548 int nxt, ours = 0; 549 int srcrt = 0; 550 551 /* 552 * Drop the packet if IPv6 operation is disabled on the interface. 553 */ 554 rcvif = m->m_pkthdr.rcvif; 555 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 556 goto bad; 557 558#if defined(IPSEC) || defined(IPSEC_SUPPORT) 559 /* 560 * should the inner packet be considered authentic? 561 * see comment in ah4_input(). 562 * NB: m cannot be NULL when passed to the input routine 563 */ 564 565 m->m_flags &= ~M_AUTHIPHDR; 566 m->m_flags &= ~M_AUTHIPDGM; 567 568#endif /* IPSEC */ 569 570 if (m->m_flags & M_FASTFWD_OURS) { 571 /* 572 * Firewall changed destination to local. 573 */ 574 ip6 = mtod(m, struct ip6_hdr *); 575 goto passin; 576 } 577 578 /* 579 * mbuf statistics 580 */ 581 if (m->m_flags & M_EXT) { 582 if (m->m_next) 583 IP6STAT_INC(ip6s_mext2m); 584 else 585 IP6STAT_INC(ip6s_mext1); 586 } else { 587 if (m->m_next) { 588 if (m->m_flags & M_LOOP) { 589 IP6STAT_INC(ip6s_m2m[V_loif->if_index]); 590 } else if (rcvif->if_index < IP6S_M2MMAX) 591 IP6STAT_INC(ip6s_m2m[rcvif->if_index]); 592 else 593 IP6STAT_INC(ip6s_m2m[0]); 594 } else 595 IP6STAT_INC(ip6s_m1); 596 } 597 598 in6_ifstat_inc(rcvif, ifs6_in_receive); 599 IP6STAT_INC(ip6s_total); 600 601#ifndef PULLDOWN_TEST 602 /* 603 * L2 bridge code and some other code can return mbuf chain 604 * that does not conform to KAME requirement. too bad. 605 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 606 */ 607 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 608 struct mbuf *n; 609 610 if (m->m_pkthdr.len > MHLEN) 611 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 612 else 613 n = m_gethdr(M_NOWAIT, MT_DATA); 614 if (n == NULL) 615 goto bad; 616 617 m_move_pkthdr(n, m); 618 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 619 n->m_len = n->m_pkthdr.len; 620 m_freem(m); 621 m = n; 622 } 623 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */); 624#endif 625 626 if (m->m_len < sizeof(struct ip6_hdr)) { 627 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 628 IP6STAT_INC(ip6s_toosmall); 629 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 630 goto bad; 631 } 632 } 633 634 ip6 = mtod(m, struct ip6_hdr *); 635 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 636 IP6STAT_INC(ip6s_badvers); 637 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 638 goto bad; 639 } 640 641 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 642 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 643 644 /* 645 * Check against address spoofing/corruption. 646 */ 647 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 648 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 649 /* 650 * XXX: "badscope" is not very suitable for a multicast source. 651 */ 652 IP6STAT_INC(ip6s_badscope); 653 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 654 goto bad; 655 } 656 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 657 !(m->m_flags & M_LOOP)) { 658 /* 659 * In this case, the packet should come from the loopback 660 * interface. However, we cannot just check the if_flags, 661 * because ip6_mloopback() passes the "actual" interface 662 * as the outgoing/incoming interface. 663 */ 664 IP6STAT_INC(ip6s_badscope); 665 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 666 goto bad; 667 } 668 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 669 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 670 /* 671 * RFC4291 2.7: 672 * Nodes must not originate a packet to a multicast address 673 * whose scop field contains the reserved value 0; if such 674 * a packet is received, it must be silently dropped. 675 */ 676 IP6STAT_INC(ip6s_badscope); 677 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 678 goto bad; 679 } 680#ifdef ALTQ 681 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 682 /* packet is dropped by traffic conditioner */ 683 return; 684 } 685#endif 686 /* 687 * The following check is not documented in specs. A malicious 688 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 689 * and bypass security checks (act as if it was from 127.0.0.1 by using 690 * IPv6 src ::ffff:127.0.0.1). Be cautious. 691 * 692 * This check chokes if we are in an SIIT cloud. As none of BSDs 693 * support IPv4-less kernel compilation, we cannot support SIIT 694 * environment at all. So, it makes more sense for us to reject any 695 * malicious packets for non-SIIT environment, than try to do a 696 * partial support for SIIT environment. 697 */ 698 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 699 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 700 IP6STAT_INC(ip6s_badscope); 701 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 702 goto bad; 703 } 704#if 0 705 /* 706 * Reject packets with IPv4 compatible addresses (auto tunnel). 707 * 708 * The code forbids auto tunnel relay case in RFC1933 (the check is 709 * stronger than RFC1933). We may want to re-enable it if mech-xx 710 * is revised to forbid relaying case. 711 */ 712 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 713 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 714 IP6STAT_INC(ip6s_badscope); 715 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 716 goto bad; 717 } 718#endif 719 /* 720 * Try to forward the packet, but if we fail continue. 721 * ip6_tryforward() does not generate redirects, so fall 722 * through to normal processing if redirects are required. 723 * ip6_tryforward() does inbound and outbound packet firewall 724 * processing. If firewall has decided that destination becomes 725 * our local address, it sets M_FASTFWD_OURS flag. In this 726 * case skip another inbound firewall processing and update 727 * ip6 pointer. 728 */ 729 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0 730#if defined(IPSEC) || defined(IPSEC_SUPPORT) 731 && (!IPSEC_ENABLED(ipv6) || 732 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0) 733#endif 734 ) { 735 if ((m = ip6_tryforward(m)) == NULL) 736 return; 737 if (m->m_flags & M_FASTFWD_OURS) { 738 ip6 = mtod(m, struct ip6_hdr *); 739 goto passin; 740 } 741 } 742#if defined(IPSEC) || defined(IPSEC_SUPPORT) 743 /* 744 * Bypass packet filtering for packets previously handled by IPsec. 745 */ 746 if (IPSEC_ENABLED(ipv6) && 747 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0) 748 goto passin; 749#endif 750 /* 751 * Run through list of hooks for input packets. 752 * 753 * NB: Beware of the destination address changing 754 * (e.g. by NAT rewriting). When this happens, 755 * tell ip6_forward to do the right thing. 756 */ 757 758 /* Jump over all PFIL processing if hooks are not active. */ 759 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 760 goto passin; 761 762 odst = ip6->ip6_dst; 763 if (pfil_run_hooks(&V_inet6_pfil_hook, &m, 764 m->m_pkthdr.rcvif, PFIL_IN, 0, NULL)) 765 return; 766 if (m == NULL) /* consumed by filter */ 767 return; 768 ip6 = mtod(m, struct ip6_hdr *); 769 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 770 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 771 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 772 /* 773 * Directly ship the packet on. This allows forwarding 774 * packets originally destined to us to some other directly 775 * connected host. 776 */ 777 ip6_forward(m, 1); 778 return; 779 } 780 781passin: 782 /* 783 * Disambiguate address scope zones (if there is ambiguity). 784 * We first make sure that the original source or destination address 785 * is not in our internal form for scoped addresses. Such addresses 786 * are not necessarily invalid spec-wise, but we cannot accept them due 787 * to the usage conflict. 788 * in6_setscope() then also checks and rejects the cases where src or 789 * dst are the loopback address and the receiving interface 790 * is not loopback. 791 */ 792 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 793 IP6STAT_INC(ip6s_badscope); /* XXX */ 794 goto bad; 795 } 796 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 797 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 798 IP6STAT_INC(ip6s_badscope); 799 goto bad; 800 } 801 if (m->m_flags & M_FASTFWD_OURS) { 802 m->m_flags &= ~M_FASTFWD_OURS; 803 ours = 1; 804 goto hbhcheck; 805 } 806 /* 807 * Multicast check. Assume packet is for us to avoid 808 * prematurely taking locks. 809 */ 810 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 811 ours = 1; 812 in6_ifstat_inc(rcvif, ifs6_in_mcast); 813 goto hbhcheck; 814 } 815 /* 816 * Unicast check 817 * XXX: For now we keep link-local IPv6 addresses with embedded 818 * scope zone id, therefore we use zero zoneid here. 819 */ 820 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */); 821 if (ia != NULL) { 822 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 823 char ip6bufs[INET6_ADDRSTRLEN]; 824 char ip6bufd[INET6_ADDRSTRLEN]; 825 /* address is not ready, so discard the packet. */ 826 nd6log((LOG_INFO, 827 "ip6_input: packet to an unready address %s->%s\n", 828 ip6_sprintf(ip6bufs, &ip6->ip6_src), 829 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 830 ifa_free(&ia->ia_ifa); 831 goto bad; 832 } 833 /* Count the packet in the ip address stats */ 834 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 835 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 836 ifa_free(&ia->ia_ifa); 837 ours = 1; 838 goto hbhcheck; 839 } 840 841 /* 842 * Now there is no reason to process the packet if it's not our own 843 * and we're not a router. 844 */ 845 if (!V_ip6_forwarding) { 846 IP6STAT_INC(ip6s_cantforward); 847 goto bad; 848 } 849 850 hbhcheck: 851 /* 852 * Process Hop-by-Hop options header if it's contained. 853 * m may be modified in ip6_hopopts_input(). 854 * If a JumboPayload option is included, plen will also be modified. 855 */ 856 plen = (u_int32_t)ntohs(ip6->ip6_plen); 857 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 858 if (ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours) != 0) 859 return; 860 } else 861 nxt = ip6->ip6_nxt; 862 863 /* 864 * Use mbuf flags to propagate Router Alert option to 865 * ICMPv6 layer, as hop-by-hop options have been stripped. 866 */ 867 if (rtalert != ~0) 868 m->m_flags |= M_RTALERT_MLD; 869 870 /* 871 * Check that the amount of data in the buffers 872 * is as at least much as the IPv6 header would have us expect. 873 * Trim mbufs if longer than we expect. 874 * Drop packet if shorter than we expect. 875 */ 876 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 877 IP6STAT_INC(ip6s_tooshort); 878 in6_ifstat_inc(rcvif, ifs6_in_truncated); 879 goto bad; 880 } 881 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 882 if (m->m_len == m->m_pkthdr.len) { 883 m->m_len = sizeof(struct ip6_hdr) + plen; 884 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 885 } else 886 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 887 } 888 889 /* 890 * Forward if desirable. 891 */ 892 if (V_ip6_mrouter && 893 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 894 /* 895 * If we are acting as a multicast router, all 896 * incoming multicast packets are passed to the 897 * kernel-level multicast forwarding function. 898 * The packet is returned (relatively) intact; if 899 * ip6_mforward() returns a non-zero value, the packet 900 * must be discarded, else it may be accepted below. 901 * 902 * XXX TODO: Check hlim and multicast scope here to avoid 903 * unnecessarily calling into ip6_mforward(). 904 */ 905 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 906 IP6STAT_INC(ip6s_cantforward); 907 goto bad; 908 } 909 } else if (!ours) { 910 ip6_forward(m, srcrt); 911 return; 912 } 913 914 ip6 = mtod(m, struct ip6_hdr *); 915 916 /* 917 * Malicious party may be able to use IPv4 mapped addr to confuse 918 * tcp/udp stack and bypass security checks (act as if it was from 919 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 920 * 921 * For SIIT end node behavior, you may want to disable the check. 922 * However, you will become vulnerable to attacks using IPv4 mapped 923 * source. 924 */ 925 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 926 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 927 IP6STAT_INC(ip6s_badscope); 928 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 929 goto bad; 930 } 931 932 /* 933 * Tell launch routine the next header 934 */ 935 IP6STAT_INC(ip6s_delivered); 936 in6_ifstat_inc(rcvif, ifs6_in_deliver); 937 nest = 0; 938 939 while (nxt != IPPROTO_DONE) { 940 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 941 IP6STAT_INC(ip6s_toomanyhdr); 942 goto bad; 943 } 944 945 /* 946 * protection against faulty packet - there should be 947 * more sanity checks in header chain processing. 948 */ 949 if (m->m_pkthdr.len < off) { 950 IP6STAT_INC(ip6s_tooshort); 951 in6_ifstat_inc(rcvif, ifs6_in_truncated); 952 goto bad; 953 } 954 955#if defined(IPSEC) || defined(IPSEC_SUPPORT) 956 if (IPSEC_ENABLED(ipv6)) { 957 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 958 return; 959 } 960#endif /* IPSEC */ 961 962 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 963 } 964 return; 965bad: 966 in6_ifstat_inc(rcvif, ifs6_in_discard); 967 if (m != NULL) 968 m_freem(m); 969} 970 971/* 972 * Hop-by-Hop options header processing. If a valid jumbo payload option is 973 * included, the real payload length will be stored in plenp. 974 * 975 * rtalertp - XXX: should be stored more smart way 976 */ 977static int 978ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 979 struct mbuf **mp, int *offp) 980{ 981 struct mbuf *m = *mp; 982 int off = *offp, hbhlen; 983 struct ip6_hbh *hbh; 984 985 /* validation of the length of the header */ 986#ifndef PULLDOWN_TEST 987 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); 988 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 989 hbhlen = (hbh->ip6h_len + 1) << 3; 990 991 IP6_EXTHDR_CHECK(m, off, hbhlen, -1); 992 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 993#else 994 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, 995 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); 996 if (hbh == NULL) { 997 IP6STAT_INC(ip6s_tooshort); 998 return -1; 999 } 1000 hbhlen = (hbh->ip6h_len + 1) << 3; 1001 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 1002 hbhlen); 1003 if (hbh == NULL) { 1004 IP6STAT_INC(ip6s_tooshort); 1005 return -1; 1006 } 1007#endif 1008 off += hbhlen; 1009 hbhlen -= sizeof(struct ip6_hbh); 1010 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 1011 hbhlen, rtalertp, plenp) < 0) 1012 return (-1); 1013 1014 *offp = off; 1015 *mp = m; 1016 return (0); 1017} 1018 1019/* 1020 * Search header for all Hop-by-hop options and process each option. 1021 * This function is separate from ip6_hopopts_input() in order to 1022 * handle a case where the sending node itself process its hop-by-hop 1023 * options header. In such a case, the function is called from ip6_output(). 1024 * 1025 * The function assumes that hbh header is located right after the IPv6 header 1026 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1027 * opthead + hbhlen is located in contiguous memory region. 1028 */ 1029int 1030ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1031 u_int32_t *rtalertp, u_int32_t *plenp) 1032{ 1033 struct ip6_hdr *ip6; 1034 int optlen = 0; 1035 u_int8_t *opt = opthead; 1036 u_int16_t rtalert_val; 1037 u_int32_t jumboplen; 1038 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1039 1040 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1041 switch (*opt) { 1042 case IP6OPT_PAD1: 1043 optlen = 1; 1044 break; 1045 case IP6OPT_PADN: 1046 if (hbhlen < IP6OPT_MINLEN) { 1047 IP6STAT_INC(ip6s_toosmall); 1048 goto bad; 1049 } 1050 optlen = *(opt + 1) + 2; 1051 break; 1052 case IP6OPT_ROUTER_ALERT: 1053 /* XXX may need check for alignment */ 1054 if (hbhlen < IP6OPT_RTALERT_LEN) { 1055 IP6STAT_INC(ip6s_toosmall); 1056 goto bad; 1057 } 1058 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1059 /* XXX stat */ 1060 icmp6_error(m, ICMP6_PARAM_PROB, 1061 ICMP6_PARAMPROB_HEADER, 1062 erroff + opt + 1 - opthead); 1063 return (-1); 1064 } 1065 optlen = IP6OPT_RTALERT_LEN; 1066 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1067 *rtalertp = ntohs(rtalert_val); 1068 break; 1069 case IP6OPT_JUMBO: 1070 /* XXX may need check for alignment */ 1071 if (hbhlen < IP6OPT_JUMBO_LEN) { 1072 IP6STAT_INC(ip6s_toosmall); 1073 goto bad; 1074 } 1075 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1076 /* XXX stat */ 1077 icmp6_error(m, ICMP6_PARAM_PROB, 1078 ICMP6_PARAMPROB_HEADER, 1079 erroff + opt + 1 - opthead); 1080 return (-1); 1081 } 1082 optlen = IP6OPT_JUMBO_LEN; 1083 1084 /* 1085 * IPv6 packets that have non 0 payload length 1086 * must not contain a jumbo payload option. 1087 */ 1088 ip6 = mtod(m, struct ip6_hdr *); 1089 if (ip6->ip6_plen) { 1090 IP6STAT_INC(ip6s_badoptions); 1091 icmp6_error(m, ICMP6_PARAM_PROB, 1092 ICMP6_PARAMPROB_HEADER, 1093 erroff + opt - opthead); 1094 return (-1); 1095 } 1096 1097 /* 1098 * We may see jumbolen in unaligned location, so 1099 * we'd need to perform bcopy(). 1100 */ 1101 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1102 jumboplen = (u_int32_t)htonl(jumboplen); 1103 1104#if 1 1105 /* 1106 * if there are multiple jumbo payload options, 1107 * *plenp will be non-zero and the packet will be 1108 * rejected. 1109 * the behavior may need some debate in ipngwg - 1110 * multiple options does not make sense, however, 1111 * there's no explicit mention in specification. 1112 */ 1113 if (*plenp != 0) { 1114 IP6STAT_INC(ip6s_badoptions); 1115 icmp6_error(m, ICMP6_PARAM_PROB, 1116 ICMP6_PARAMPROB_HEADER, 1117 erroff + opt + 2 - opthead); 1118 return (-1); 1119 } 1120#endif 1121 1122 /* 1123 * jumbo payload length must be larger than 65535. 1124 */ 1125 if (jumboplen <= IPV6_MAXPACKET) { 1126 IP6STAT_INC(ip6s_badoptions); 1127 icmp6_error(m, ICMP6_PARAM_PROB, 1128 ICMP6_PARAMPROB_HEADER, 1129 erroff + opt + 2 - opthead); 1130 return (-1); 1131 } 1132 *plenp = jumboplen; 1133 1134 break; 1135 default: /* unknown option */ 1136 if (hbhlen < IP6OPT_MINLEN) { 1137 IP6STAT_INC(ip6s_toosmall); 1138 goto bad; 1139 } 1140 optlen = ip6_unknown_opt(opt, m, 1141 erroff + opt - opthead); 1142 if (optlen == -1) 1143 return (-1); 1144 optlen += 2; 1145 break; 1146 } 1147 } 1148 1149 return (0); 1150 1151 bad: 1152 m_freem(m); 1153 return (-1); 1154} 1155 1156/* 1157 * Unknown option processing. 1158 * The third argument `off' is the offset from the IPv6 header to the option, 1159 * which is necessary if the IPv6 header the and option header and IPv6 header 1160 * is not contiguous in order to return an ICMPv6 error. 1161 */ 1162int 1163ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1164{ 1165 struct ip6_hdr *ip6; 1166 1167 switch (IP6OPT_TYPE(*optp)) { 1168 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1169 return ((int)*(optp + 1)); 1170 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1171 m_freem(m); 1172 return (-1); 1173 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1174 IP6STAT_INC(ip6s_badoptions); 1175 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1176 return (-1); 1177 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1178 IP6STAT_INC(ip6s_badoptions); 1179 ip6 = mtod(m, struct ip6_hdr *); 1180 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1181 (m->m_flags & (M_BCAST|M_MCAST))) 1182 m_freem(m); 1183 else 1184 icmp6_error(m, ICMP6_PARAM_PROB, 1185 ICMP6_PARAMPROB_OPTION, off); 1186 return (-1); 1187 } 1188 1189 m_freem(m); /* XXX: NOTREACHED */ 1190 return (-1); 1191} 1192 1193/* 1194 * Create the "control" list for this pcb. 1195 * These functions will not modify mbuf chain at all. 1196 * 1197 * With KAME mbuf chain restriction: 1198 * The routine will be called from upper layer handlers like tcp6_input(). 1199 * Thus the routine assumes that the caller (tcp6_input) have already 1200 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the 1201 * very first mbuf on the mbuf chain. 1202 * 1203 * ip6_savecontrol_v4 will handle those options that are possible to be 1204 * set on a v4-mapped socket. 1205 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1206 * options and handle the v6-only ones itself. 1207 */ 1208struct mbuf ** 1209ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1210 int *v4only) 1211{ 1212 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1213 1214#ifdef SO_TIMESTAMP 1215 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1216 struct timeval tv; 1217 1218 microtime(&tv); 1219 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1220 SCM_TIMESTAMP, SOL_SOCKET); 1221 if (*mp) 1222 mp = &(*mp)->m_next; 1223 } 1224#endif 1225 1226#define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1227 /* RFC 2292 sec. 5 */ 1228 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1229 struct in6_pktinfo pi6; 1230 1231 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1232#ifdef INET 1233 struct ip *ip; 1234 1235 ip = mtod(m, struct ip *); 1236 pi6.ipi6_addr.s6_addr32[0] = 0; 1237 pi6.ipi6_addr.s6_addr32[1] = 0; 1238 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1239 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1240#else 1241 /* We won't hit this code */ 1242 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1243#endif 1244 } else { 1245 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1246 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1247 } 1248 pi6.ipi6_ifindex = 1249 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1250 1251 *mp = sbcreatecontrol((caddr_t) &pi6, 1252 sizeof(struct in6_pktinfo), 1253 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1254 if (*mp) 1255 mp = &(*mp)->m_next; 1256 } 1257 1258 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1259 int hlim; 1260 1261 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1262#ifdef INET 1263 struct ip *ip; 1264 1265 ip = mtod(m, struct ip *); 1266 hlim = ip->ip_ttl; 1267#else 1268 /* We won't hit this code */ 1269 hlim = 0; 1270#endif 1271 } else { 1272 hlim = ip6->ip6_hlim & 0xff; 1273 } 1274 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1275 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1276 IPPROTO_IPV6); 1277 if (*mp) 1278 mp = &(*mp)->m_next; 1279 } 1280 1281 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1282 int tclass; 1283 1284 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1285#ifdef INET 1286 struct ip *ip; 1287 1288 ip = mtod(m, struct ip *); 1289 tclass = ip->ip_tos; 1290#else 1291 /* We won't hit this code */ 1292 tclass = 0; 1293#endif 1294 } else { 1295 u_int32_t flowinfo; 1296 1297 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1298 flowinfo >>= 20; 1299 tclass = flowinfo & 0xff; 1300 } 1301 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int), 1302 IPV6_TCLASS, IPPROTO_IPV6); 1303 if (*mp) 1304 mp = &(*mp)->m_next; 1305 } 1306 1307 if (v4only != NULL) { 1308 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1309 *v4only = 1; 1310 } else { 1311 *v4only = 0; 1312 } 1313 } 1314 1315 return (mp); 1316} 1317 1318void 1319ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) 1320{ 1321 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1322 int v4only = 0; 1323 1324 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); 1325 if (v4only) 1326 return; 1327 1328 /* 1329 * IPV6_HOPOPTS socket option. Recall that we required super-user 1330 * privilege for the option (see ip6_ctloutput), but it might be too 1331 * strict, since there might be some hop-by-hop options which can be 1332 * returned to normal user. 1333 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1334 */ 1335 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { 1336 /* 1337 * Check if a hop-by-hop options header is contatined in the 1338 * received packet, and if so, store the options as ancillary 1339 * data. Note that a hop-by-hop options header must be 1340 * just after the IPv6 header, which is assured through the 1341 * IPv6 input processing. 1342 */ 1343 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1344 struct ip6_hbh *hbh; 1345 int hbhlen = 0; 1346#ifdef PULLDOWN_TEST 1347 struct mbuf *ext; 1348#endif 1349 1350#ifndef PULLDOWN_TEST 1351 hbh = (struct ip6_hbh *)(ip6 + 1); 1352 hbhlen = (hbh->ip6h_len + 1) << 3; 1353#else 1354 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), 1355 ip6->ip6_nxt); 1356 if (ext == NULL) { 1357 IP6STAT_INC(ip6s_tooshort); 1358 return; 1359 } 1360 hbh = mtod(ext, struct ip6_hbh *); 1361 hbhlen = (hbh->ip6h_len + 1) << 3; 1362 if (hbhlen != ext->m_len) { 1363 m_freem(ext); 1364 IP6STAT_INC(ip6s_tooshort); 1365 return; 1366 } 1367#endif 1368 1369 /* 1370 * XXX: We copy the whole header even if a 1371 * jumbo payload option is included, the option which 1372 * is to be removed before returning according to 1373 * RFC2292. 1374 * Note: this constraint is removed in RFC3542 1375 */ 1376 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1377 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1378 IPPROTO_IPV6); 1379 if (*mp) 1380 mp = &(*mp)->m_next; 1381#ifdef PULLDOWN_TEST 1382 m_freem(ext); 1383#endif 1384 } 1385 } 1386 1387 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1388 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1389 1390 /* 1391 * Search for destination options headers or routing 1392 * header(s) through the header chain, and stores each 1393 * header as ancillary data. 1394 * Note that the order of the headers remains in 1395 * the chain of ancillary data. 1396 */ 1397 while (1) { /* is explicit loop prevention necessary? */ 1398 struct ip6_ext *ip6e = NULL; 1399 int elen; 1400#ifdef PULLDOWN_TEST 1401 struct mbuf *ext = NULL; 1402#endif 1403 1404 /* 1405 * if it is not an extension header, don't try to 1406 * pull it from the chain. 1407 */ 1408 switch (nxt) { 1409 case IPPROTO_DSTOPTS: 1410 case IPPROTO_ROUTING: 1411 case IPPROTO_HOPOPTS: 1412 case IPPROTO_AH: /* is it possible? */ 1413 break; 1414 default: 1415 goto loopend; 1416 } 1417 1418#ifndef PULLDOWN_TEST 1419 if (off + sizeof(*ip6e) > m->m_len) 1420 goto loopend; 1421 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1422 if (nxt == IPPROTO_AH) 1423 elen = (ip6e->ip6e_len + 2) << 2; 1424 else 1425 elen = (ip6e->ip6e_len + 1) << 3; 1426 if (off + elen > m->m_len) 1427 goto loopend; 1428#else 1429 ext = ip6_pullexthdr(m, off, nxt); 1430 if (ext == NULL) { 1431 IP6STAT_INC(ip6s_tooshort); 1432 return; 1433 } 1434 ip6e = mtod(ext, struct ip6_ext *); 1435 if (nxt == IPPROTO_AH) 1436 elen = (ip6e->ip6e_len + 2) << 2; 1437 else 1438 elen = (ip6e->ip6e_len + 1) << 3; 1439 if (elen != ext->m_len) { 1440 m_freem(ext); 1441 IP6STAT_INC(ip6s_tooshort); 1442 return; 1443 } 1444#endif 1445 1446 switch (nxt) { 1447 case IPPROTO_DSTOPTS: 1448 if (!(in6p->inp_flags & IN6P_DSTOPTS)) 1449 break; 1450 1451 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1452 IS2292(in6p, 1453 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1454 IPPROTO_IPV6); 1455 if (*mp) 1456 mp = &(*mp)->m_next; 1457 break; 1458 case IPPROTO_ROUTING: 1459 if (!(in6p->inp_flags & IN6P_RTHDR)) 1460 break; 1461 1462 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1463 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), 1464 IPPROTO_IPV6); 1465 if (*mp) 1466 mp = &(*mp)->m_next; 1467 break; 1468 case IPPROTO_HOPOPTS: 1469 case IPPROTO_AH: /* is it possible? */ 1470 break; 1471 1472 default: 1473 /* 1474 * other cases have been filtered in the above. 1475 * none will visit this case. here we supply 1476 * the code just in case (nxt overwritten or 1477 * other cases). 1478 */ 1479#ifdef PULLDOWN_TEST 1480 m_freem(ext); 1481#endif 1482 goto loopend; 1483 1484 } 1485 1486 /* proceed with the next header. */ 1487 off += elen; 1488 nxt = ip6e->ip6e_nxt; 1489 ip6e = NULL; 1490#ifdef PULLDOWN_TEST 1491 m_freem(ext); 1492 ext = NULL; 1493#endif 1494 } 1495 loopend: 1496 ; 1497 } 1498 1499 if (in6p->inp_flags2 & INP_RECVFLOWID) { 1500 uint32_t flowid, flow_type; 1501 1502 flowid = m->m_pkthdr.flowid; 1503 flow_type = M_HASHTYPE_GET(m); 1504 1505 /* 1506 * XXX should handle the failure of one or the 1507 * other - don't populate both? 1508 */ 1509 *mp = sbcreatecontrol((caddr_t) &flowid, 1510 sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6); 1511 if (*mp) 1512 mp = &(*mp)->m_next; 1513 *mp = sbcreatecontrol((caddr_t) &flow_type, 1514 sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6); 1515 if (*mp) 1516 mp = &(*mp)->m_next; 1517 } 1518 1519#ifdef RSS 1520 if (in6p->inp_flags2 & INP_RECVRSSBUCKETID) { 1521 uint32_t flowid, flow_type; 1522 uint32_t rss_bucketid; 1523 1524 flowid = m->m_pkthdr.flowid; 1525 flow_type = M_HASHTYPE_GET(m); 1526 1527 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1528 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1529 sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6); 1530 if (*mp) 1531 mp = &(*mp)->m_next; 1532 } 1533 } 1534#endif 1535 1536} 1537#undef IS2292 1538 1539void 1540ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1541{ 1542 struct socket *so; 1543 struct mbuf *m_mtu; 1544 struct ip6_mtuinfo mtuctl; 1545 1546 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1547 /* 1548 * Notify the error by sending IPV6_PATHMTU ancillary data if 1549 * application wanted to know the MTU value. 1550 * NOTE: we notify disconnected sockets, because some udp 1551 * applications keep sending sockets disconnected. 1552 * NOTE: our implementation doesn't notify connected sockets that has 1553 * foreign address that is different than given destination addresses 1554 * (this is permitted by RFC 3542). 1555 */ 1556 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1557 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1558 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1559 return; 1560 1561 mtuctl.ip6m_mtu = mtu; 1562 mtuctl.ip6m_addr = *dst; 1563 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1564 return; 1565 1566 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1567 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1568 return; 1569 1570 so = inp->inp_socket; 1571 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1572 == 0) { 1573 m_freem(m_mtu); 1574 /* XXX: should count statistics */ 1575 } else 1576 sorwakeup(so); 1577} 1578 1579#ifdef PULLDOWN_TEST 1580/* 1581 * pull single extension header from mbuf chain. returns single mbuf that 1582 * contains the result, or NULL on error. 1583 */ 1584static struct mbuf * 1585ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) 1586{ 1587 struct ip6_ext ip6e; 1588 size_t elen; 1589 struct mbuf *n; 1590 1591#ifdef DIAGNOSTIC 1592 switch (nxt) { 1593 case IPPROTO_DSTOPTS: 1594 case IPPROTO_ROUTING: 1595 case IPPROTO_HOPOPTS: 1596 case IPPROTO_AH: /* is it possible? */ 1597 break; 1598 default: 1599 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); 1600 } 1601#endif 1602 1603 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1604 if (nxt == IPPROTO_AH) 1605 elen = (ip6e.ip6e_len + 2) << 2; 1606 else 1607 elen = (ip6e.ip6e_len + 1) << 3; 1608 1609 if (elen > MLEN) 1610 n = m_getcl(M_NOWAIT, MT_DATA, 0); 1611 else 1612 n = m_get(M_NOWAIT, MT_DATA); 1613 if (n == NULL) 1614 return NULL; 1615 1616 m_copydata(m, off, elen, mtod(n, caddr_t)); 1617 n->m_len = elen; 1618 return n; 1619} 1620#endif 1621 1622/* 1623 * Get pointer to the previous header followed by the header 1624 * currently processed. 1625 */ 1626int 1627ip6_get_prevhdr(const struct mbuf *m, int off) 1628{ 1629 struct ip6_ext ip6e; 1630 struct ip6_hdr *ip6; 1631 int len, nlen, nxt; 1632 1633 if (off == sizeof(struct ip6_hdr)) 1634 return (offsetof(struct ip6_hdr, ip6_nxt)); 1635 if (off < sizeof(struct ip6_hdr)) 1636 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1637 1638 ip6 = mtod(m, struct ip6_hdr *); 1639 nxt = ip6->ip6_nxt; 1640 len = sizeof(struct ip6_hdr); 1641 nlen = 0; 1642 while (len < off) { 1643 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1644 switch (nxt) { 1645 case IPPROTO_FRAGMENT: 1646 nlen = sizeof(struct ip6_frag); 1647 break; 1648 case IPPROTO_AH: 1649 nlen = (ip6e.ip6e_len + 2) << 2; 1650 break; 1651 default: 1652 nlen = (ip6e.ip6e_len + 1) << 3; 1653 } 1654 len += nlen; 1655 nxt = ip6e.ip6e_nxt; 1656 } 1657 return (len - nlen); 1658} 1659 1660/* 1661 * get next header offset. m will be retained. 1662 */ 1663int 1664ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1665{ 1666 struct ip6_hdr ip6; 1667 struct ip6_ext ip6e; 1668 struct ip6_frag fh; 1669 1670 /* just in case */ 1671 if (m == NULL) 1672 panic("ip6_nexthdr: m == NULL"); 1673 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1674 return -1; 1675 1676 switch (proto) { 1677 case IPPROTO_IPV6: 1678 if (m->m_pkthdr.len < off + sizeof(ip6)) 1679 return -1; 1680 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1681 if (nxtp) 1682 *nxtp = ip6.ip6_nxt; 1683 off += sizeof(ip6); 1684 return off; 1685 1686 case IPPROTO_FRAGMENT: 1687 /* 1688 * terminate parsing if it is not the first fragment, 1689 * it does not make sense to parse through it. 1690 */ 1691 if (m->m_pkthdr.len < off + sizeof(fh)) 1692 return -1; 1693 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1694 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1695 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1696 return -1; 1697 if (nxtp) 1698 *nxtp = fh.ip6f_nxt; 1699 off += sizeof(struct ip6_frag); 1700 return off; 1701 1702 case IPPROTO_AH: 1703 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1704 return -1; 1705 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1706 if (nxtp) 1707 *nxtp = ip6e.ip6e_nxt; 1708 off += (ip6e.ip6e_len + 2) << 2; 1709 return off; 1710 1711 case IPPROTO_HOPOPTS: 1712 case IPPROTO_ROUTING: 1713 case IPPROTO_DSTOPTS: 1714 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1715 return -1; 1716 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1717 if (nxtp) 1718 *nxtp = ip6e.ip6e_nxt; 1719 off += (ip6e.ip6e_len + 1) << 3; 1720 return off; 1721 1722 case IPPROTO_NONE: 1723 case IPPROTO_ESP: 1724 case IPPROTO_IPCOMP: 1725 /* give up */ 1726 return -1; 1727 1728 default: 1729 return -1; 1730 } 1731 1732 /* NOTREACHED */ 1733} 1734 1735/* 1736 * get offset for the last header in the chain. m will be kept untainted. 1737 */ 1738int 1739ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1740{ 1741 int newoff; 1742 int nxt; 1743 1744 if (!nxtp) { 1745 nxt = -1; 1746 nxtp = &nxt; 1747 } 1748 while (1) { 1749 newoff = ip6_nexthdr(m, off, proto, nxtp); 1750 if (newoff < 0) 1751 return off; 1752 else if (newoff < off) 1753 return -1; /* invalid */ 1754 else if (newoff == off) 1755 return newoff; 1756 1757 off = newoff; 1758 proto = *nxtp; 1759 } 1760} 1761 1762/* 1763 * System control for IP6 1764 */ 1765 1766u_char inet6ctlerrmap[PRC_NCMDS] = { 1767 0, 0, 0, 0, 1768 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1769 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1770 EMSGSIZE, EHOSTUNREACH, 0, 0, 1771 0, 0, EHOSTUNREACH, 0, 1772 ENOPROTOOPT, ECONNREFUSED 1773}; 1774