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_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 #ifdef IPSEC 122 #include <netipsec/key.h> 123 #include <netipsec/ipsec.h> 124 #include <netinet6/ip6_ipsec.h> 125 #include <netipsec/ipsec6.h> 126 #endif /* IPSEC */ 127 128 #include <netinet6/ip6protosw.h> 129 130 extern struct domain inet6domain; 131 132 u_char ip6_protox[IPPROTO_MAX]; 133 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 134 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); 135 VNET_DEFINE(u_long, in6_ifaddrhmask); 136 137 static struct netisr_handler ip6_nh = { 138 .nh_name = "ip6", 139 .nh_handler = ip6_input, 140 .nh_proto = NETISR_IPV6, 141 #ifdef RSS 142 .nh_m2cpuid = rss_soft_m2cpuid_v6, 143 .nh_policy = NETISR_POLICY_CPU, 144 .nh_dispatch = NETISR_DISPATCH_HYBRID, 145 #else 146 .nh_policy = NETISR_POLICY_FLOW, 147 #endif 148 }; 149 150 static int 151 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 152 { 153 int error, qlimit; 154 155 netisr_getqlimit(&ip6_nh, &qlimit); 156 error = sysctl_handle_int(oidp, &qlimit, 0, req); 157 if (error || !req->newptr) 158 return (error); 159 if (qlimit < 1) 160 return (EINVAL); 161 return (netisr_setqlimit(&ip6_nh, qlimit)); 162 } 163 SYSCTL_DECL(_net_inet6_ip6); 164 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen, 165 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet6_intr_queue_maxlen, "I", 166 "Maximum size of the IPv6 input queue"); 167 168 #ifdef RSS 169 static struct netisr_handler ip6_direct_nh = { 170 .nh_name = "ip6_direct", 171 .nh_handler = ip6_direct_input, 172 .nh_proto = NETISR_IPV6_DIRECT, 173 .nh_m2cpuid = rss_soft_m2cpuid_v6, 174 .nh_policy = NETISR_POLICY_CPU, 175 .nh_dispatch = NETISR_DISPATCH_HYBRID, 176 }; 177 178 static int 179 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 180 { 181 int error, qlimit; 182 183 netisr_getqlimit(&ip6_direct_nh, &qlimit); 184 error = sysctl_handle_int(oidp, &qlimit, 0, req); 185 if (error || !req->newptr) 186 return (error); 187 if (qlimit < 1) 188 return (EINVAL); 189 return (netisr_setqlimit(&ip6_direct_nh, qlimit)); 190 } 191 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 192 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, 193 "I", "Maximum size of the IPv6 direct input queue"); 194 195 #endif 196 197 VNET_DEFINE(struct pfil_head, inet6_pfil_hook); 198 199 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 200 VNET_PCPUSTAT_SYSINIT(ip6stat); 201 #ifdef VIMAGE 202 VNET_PCPUSTAT_SYSUNINIT(ip6stat); 203 #endif /* VIMAGE */ 204 205 struct rmlock in6_ifaddr_lock; 206 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 207 208 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 209 #ifdef PULLDOWN_TEST 210 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int); 211 #endif 212 213 /* 214 * IP6 initialization: fill in IP6 protocol switch table. 215 * All protocols not implemented in kernel go to raw IP6 protocol handler. 216 */ 217 void 218 ip6_init(void) 219 { 220 struct protosw *pr; 221 int i; 222 223 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 224 &V_ip6_auto_linklocal); 225 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 226 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 227 228 TAILQ_INIT(&V_in6_ifaddrhead); 229 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 230 &V_in6_ifaddrhmask); 231 232 /* Initialize packet filter hooks. */ 233 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF; 234 V_inet6_pfil_hook.ph_af = AF_INET6; 235 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0) 236 printf("%s: WARNING: unable to register pfil hook, " 237 "error %d\n", __func__, i); 238 239 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 240 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 241 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 242 printf("%s: WARNING: unable to register input helper hook\n", 243 __func__); 244 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 245 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 246 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 247 printf("%s: WARNING: unable to register output helper hook\n", 248 __func__); 249 250 scope6_init(); 251 addrsel_policy_init(); 252 nd6_init(); 253 frag6_init(); 254 255 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 256 257 /* Skip global initialization stuff for non-default instances. */ 258 #ifdef VIMAGE 259 if (!IS_DEFAULT_VNET(curvnet)) { 260 netisr_register_vnet(&ip6_nh); 261 #ifdef RSS 262 netisr_register_vnet(&ip6_direct_nh); 263 #endif 264 return; 265 } 266 #endif 267 268 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 269 if (pr == NULL) 270 panic("ip6_init"); 271 272 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 273 for (i = 0; i < IPPROTO_MAX; i++) 274 ip6_protox[i] = pr - inet6sw; 275 /* 276 * Cycle through IP protocols and put them into the appropriate place 277 * in ip6_protox[]. 278 */ 279 for (pr = inet6domain.dom_protosw; 280 pr < inet6domain.dom_protoswNPROTOSW; pr++) 281 if (pr->pr_domain->dom_family == PF_INET6 && 282 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 283 /* Be careful to only index valid IP protocols. */ 284 if (pr->pr_protocol < IPPROTO_MAX) 285 ip6_protox[pr->pr_protocol] = pr - inet6sw; 286 } 287 288 netisr_register(&ip6_nh); 289 #ifdef RSS 290 netisr_register(&ip6_direct_nh); 291 #endif 292 } 293 294 /* 295 * The protocol to be inserted into ip6_protox[] must be already registered 296 * in inet6sw[], either statically or through pf_proto_register(). 297 */ 298 int 299 ip6proto_register(short ip6proto) 300 { 301 struct protosw *pr; 302 303 /* Sanity checks. */ 304 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 305 return (EPROTONOSUPPORT); 306 307 /* 308 * The protocol slot must not be occupied by another protocol 309 * already. An index pointing to IPPROTO_RAW is unused. 310 */ 311 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 312 if (pr == NULL) 313 return (EPFNOSUPPORT); 314 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 315 return (EEXIST); 316 317 /* 318 * Find the protocol position in inet6sw[] and set the index. 319 */ 320 for (pr = inet6domain.dom_protosw; 321 pr < inet6domain.dom_protoswNPROTOSW; pr++) { 322 if (pr->pr_domain->dom_family == PF_INET6 && 323 pr->pr_protocol && pr->pr_protocol == ip6proto) { 324 ip6_protox[pr->pr_protocol] = pr - inet6sw; 325 return (0); 326 } 327 } 328 return (EPROTONOSUPPORT); 329 } 330 331 int 332 ip6proto_unregister(short ip6proto) 333 { 334 struct protosw *pr; 335 336 /* Sanity checks. */ 337 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 338 return (EPROTONOSUPPORT); 339 340 /* Check if the protocol was indeed registered. */ 341 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 342 if (pr == NULL) 343 return (EPFNOSUPPORT); 344 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 345 return (ENOENT); 346 347 /* Reset the protocol slot to IPPROTO_RAW. */ 348 ip6_protox[ip6proto] = pr - inet6sw; 349 return (0); 350 } 351 352 #ifdef VIMAGE 353 static void 354 ip6_destroy(void *unused __unused) 355 { 356 struct ifaddr *ifa, *nifa; 357 struct ifnet *ifp; 358 int error; 359 360 #ifdef RSS 361 netisr_unregister_vnet(&ip6_direct_nh); 362 #endif 363 netisr_unregister_vnet(&ip6_nh); 364 365 if ((error = pfil_head_unregister(&V_inet6_pfil_hook)) != 0) 366 printf("%s: WARNING: unable to unregister pfil hook, " 367 "error %d\n", __func__, error); 368 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 369 if (error != 0) { 370 printf("%s: WARNING: unable to deregister input helper hook " 371 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 372 "error %d returned\n", __func__, error); 373 } 374 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 375 if (error != 0) { 376 printf("%s: WARNING: unable to deregister output helper hook " 377 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 378 "error %d returned\n", __func__, error); 379 } 380 381 /* Cleanup addresses. */ 382 IFNET_RLOCK(); 383 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 384 /* Cannot lock here - lock recursion. */ 385 /* IF_ADDR_LOCK(ifp); */ 386 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 387 388 if (ifa->ifa_addr->sa_family != AF_INET6) 389 continue; 390 in6_purgeaddr(ifa); 391 } 392 /* IF_ADDR_UNLOCK(ifp); */ 393 in6_ifdetach_destroy(ifp); 394 mld_domifdetach(ifp); 395 /* Make sure any routes are gone as well. */ 396 rt_flushifroutes_af(ifp, AF_INET6); 397 } 398 IFNET_RUNLOCK(); 399 400 nd6_destroy(); 401 in6_ifattach_destroy(); 402 403 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 404 } 405 406 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 407 #endif 408 409 static int 410 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off, 411 int *nxt, int *ours) 412 { 413 struct ip6_hdr *ip6; 414 struct ip6_hbh *hbh; 415 416 if (ip6_hopopts_input(plen, rtalert, &m, off)) { 417 #if 0 /*touches NULL pointer*/ 418 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 419 #endif 420 goto out; /* m have already been freed */ 421 } 422 423 /* adjust pointer */ 424 ip6 = mtod(m, struct ip6_hdr *); 425 426 /* 427 * if the payload length field is 0 and the next header field 428 * indicates Hop-by-Hop Options header, then a Jumbo Payload 429 * option MUST be included. 430 */ 431 if (ip6->ip6_plen == 0 && *plen == 0) { 432 /* 433 * Note that if a valid jumbo payload option is 434 * contained, ip6_hopopts_input() must set a valid 435 * (non-zero) payload length to the variable plen. 436 */ 437 IP6STAT_INC(ip6s_badoptions); 438 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 439 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 440 icmp6_error(m, ICMP6_PARAM_PROB, 441 ICMP6_PARAMPROB_HEADER, 442 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 443 goto out; 444 } 445 #ifndef PULLDOWN_TEST 446 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 447 hbh = (struct ip6_hbh *)(ip6 + 1); 448 #else 449 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 450 sizeof(struct ip6_hbh)); 451 if (hbh == NULL) { 452 IP6STAT_INC(ip6s_tooshort); 453 goto out; 454 } 455 #endif 456 *nxt = hbh->ip6h_nxt; 457 458 /* 459 * If we are acting as a router and the packet contains a 460 * router alert option, see if we know the option value. 461 * Currently, we only support the option value for MLD, in which 462 * case we should pass the packet to the multicast routing 463 * daemon. 464 */ 465 if (*rtalert != ~0) { 466 switch (*rtalert) { 467 case IP6OPT_RTALERT_MLD: 468 if (V_ip6_forwarding) 469 *ours = 1; 470 break; 471 default: 472 /* 473 * RFC2711 requires unrecognized values must be 474 * silently ignored. 475 */ 476 break; 477 } 478 } 479 480 return (0); 481 482 out: 483 return (1); 484 } 485 486 #ifdef RSS 487 /* 488 * IPv6 direct input routine. 489 * 490 * This is called when reinjecting completed fragments where 491 * all of the previous checking and book-keeping has been done. 492 */ 493 void 494 ip6_direct_input(struct mbuf *m) 495 { 496 int off, nxt; 497 int nest; 498 struct m_tag *mtag; 499 struct ip6_direct_ctx *ip6dc; 500 501 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 502 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 503 504 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 505 nxt = ip6dc->ip6dc_nxt; 506 off = ip6dc->ip6dc_off; 507 508 nest = 0; 509 510 m_tag_delete(m, mtag); 511 512 while (nxt != IPPROTO_DONE) { 513 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 514 IP6STAT_INC(ip6s_toomanyhdr); 515 goto bad; 516 } 517 518 /* 519 * protection against faulty packet - there should be 520 * more sanity checks in header chain processing. 521 */ 522 if (m->m_pkthdr.len < off) { 523 IP6STAT_INC(ip6s_tooshort); 524 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 525 goto bad; 526 } 527 528 #ifdef IPSEC 529 /* 530 * enforce IPsec policy checking if we are seeing last header. 531 * note that we do not visit this with protocols with pcb layer 532 * code - like udp/tcp/raw ip. 533 */ 534 if (ip6_ipsec_input(m, nxt)) 535 goto bad; 536 #endif /* IPSEC */ 537 538 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 539 } 540 return; 541 bad: 542 m_freem(m); 543 } 544 #endif 545 546 void 547 ip6_input(struct mbuf *m) 548 { 549 struct in6_addr odst; 550 struct ip6_hdr *ip6; 551 struct in6_ifaddr *ia; 552 struct ifnet *rcvif; 553 u_int32_t plen; 554 u_int32_t rtalert = ~0; 555 int off = sizeof(struct ip6_hdr), nest; 556 int nxt, ours = 0; 557 int srcrt = 0; 558 559 /* 560 * Drop the packet if IPv6 operation is disabled on the interface. 561 */ 562 rcvif = m->m_pkthdr.rcvif; 563 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 564 goto bad; 565 566 #ifdef IPSEC 567 /* 568 * should the inner packet be considered authentic? 569 * see comment in ah4_input(). 570 * NB: m cannot be NULL when passed to the input routine 571 */ 572 573 m->m_flags &= ~M_AUTHIPHDR; 574 m->m_flags &= ~M_AUTHIPDGM; 575 576 #endif /* IPSEC */ 577 578 if (m->m_flags & M_FASTFWD_OURS) { 579 /* 580 * Firewall changed destination to local. 581 */ 582 m->m_flags &= ~M_FASTFWD_OURS; 583 ours = 1; 584 ip6 = mtod(m, struct ip6_hdr *); 585 goto hbhcheck; 586 } 587 588 /* 589 * mbuf statistics 590 */ 591 if (m->m_flags & M_EXT) { 592 if (m->m_next) 593 IP6STAT_INC(ip6s_mext2m); 594 else 595 IP6STAT_INC(ip6s_mext1); 596 } else { 597 if (m->m_next) { 598 if (m->m_flags & M_LOOP) { 599 IP6STAT_INC(ip6s_m2m[V_loif->if_index]); 600 } else if (rcvif->if_index < IP6S_M2MMAX) 601 IP6STAT_INC(ip6s_m2m[rcvif->if_index]); 602 else 603 IP6STAT_INC(ip6s_m2m[0]); 604 } else 605 IP6STAT_INC(ip6s_m1); 606 } 607 608 in6_ifstat_inc(rcvif, ifs6_in_receive); 609 IP6STAT_INC(ip6s_total); 610 611 #ifndef PULLDOWN_TEST 612 /* 613 * L2 bridge code and some other code can return mbuf chain 614 * that does not conform to KAME requirement. too bad. 615 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 616 */ 617 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 618 struct mbuf *n; 619 620 if (m->m_pkthdr.len > MHLEN) 621 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 622 else 623 n = m_gethdr(M_NOWAIT, MT_DATA); 624 if (n == NULL) 625 goto bad; 626 627 m_move_pkthdr(n, m); 628 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 629 n->m_len = n->m_pkthdr.len; 630 m_freem(m); 631 m = n; 632 } 633 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */); 634 #endif 635 636 if (m->m_len < sizeof(struct ip6_hdr)) { 637 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 638 IP6STAT_INC(ip6s_toosmall); 639 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 640 goto bad; 641 } 642 } 643 644 ip6 = mtod(m, struct ip6_hdr *); 645 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 646 IP6STAT_INC(ip6s_badvers); 647 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 648 goto bad; 649 } 650 651 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 652 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 653 654 /* 655 * Check against address spoofing/corruption. 656 */ 657 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 658 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 659 /* 660 * XXX: "badscope" is not very suitable for a multicast source. 661 */ 662 IP6STAT_INC(ip6s_badscope); 663 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 664 goto bad; 665 } 666 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 667 !(m->m_flags & M_LOOP)) { 668 /* 669 * In this case, the packet should come from the loopback 670 * interface. However, we cannot just check the if_flags, 671 * because ip6_mloopback() passes the "actual" interface 672 * as the outgoing/incoming interface. 673 */ 674 IP6STAT_INC(ip6s_badscope); 675 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 676 goto bad; 677 } 678 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 679 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 680 /* 681 * RFC4291 2.7: 682 * Nodes must not originate a packet to a multicast address 683 * whose scop field contains the reserved value 0; if such 684 * a packet is received, it must be silently dropped. 685 */ 686 IP6STAT_INC(ip6s_badscope); 687 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 688 goto bad; 689 } 690 #ifdef ALTQ 691 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 692 /* packet is dropped by traffic conditioner */ 693 return; 694 } 695 #endif 696 /* 697 * The following check is not documented in specs. A malicious 698 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 699 * and bypass security checks (act as if it was from 127.0.0.1 by using 700 * IPv6 src ::ffff:127.0.0.1). Be cautious. 701 * 702 * This check chokes if we are in an SIIT cloud. As none of BSDs 703 * support IPv4-less kernel compilation, we cannot support SIIT 704 * environment at all. So, it makes more sense for us to reject any 705 * malicious packets for non-SIIT environment, than try to do a 706 * partial support for SIIT environment. 707 */ 708 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 709 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 710 IP6STAT_INC(ip6s_badscope); 711 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 712 goto bad; 713 } 714 #if 0 715 /* 716 * Reject packets with IPv4 compatible addresses (auto tunnel). 717 * 718 * The code forbids auto tunnel relay case in RFC1933 (the check is 719 * stronger than RFC1933). We may want to re-enable it if mech-xx 720 * is revised to forbid relaying case. 721 */ 722 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 723 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 724 IP6STAT_INC(ip6s_badscope); 725 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 726 goto bad; 727 } 728 #endif 729 /* 730 * Try to forward the packet, but if we fail continue. 731 * ip6_tryforward() does inbound and outbound packet firewall 732 * processing. If firewall has decided that destination becomes 733 * our local address, it sets M_FASTFWD_OURS flag. In this 734 * case skip another inbound firewall processing and update 735 * ip6 pointer. 736 */ 737 if (V_ip6_forwarding != 0 738 #ifdef IPSEC 739 && !key_havesp(IPSEC_DIR_INBOUND) 740 && !key_havesp(IPSEC_DIR_OUTBOUND) 741 #endif 742 ) { 743 if ((m = ip6_tryforward(m)) == NULL) 744 return; 745 if (m->m_flags & M_FASTFWD_OURS) { 746 m->m_flags &= ~M_FASTFWD_OURS; 747 ours = 1; 748 ip6 = mtod(m, struct ip6_hdr *); 749 goto hbhcheck; 750 } 751 } 752 #ifdef IPSEC 753 /* 754 * Bypass packet filtering for packets previously handled by IPsec. 755 */ 756 if (ip6_ipsec_filtertunnel(m)) 757 goto passin; 758 #endif 759 /* 760 * Run through list of hooks for input packets. 761 * 762 * NB: Beware of the destination address changing 763 * (e.g. by NAT rewriting). When this happens, 764 * tell ip6_forward to do the right thing. 765 */ 766 767 /* Jump over all PFIL processing if hooks are not active. */ 768 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 769 goto passin; 770 771 odst = ip6->ip6_dst; 772 if (pfil_run_hooks(&V_inet6_pfil_hook, &m, 773 m->m_pkthdr.rcvif, PFIL_IN, NULL)) 774 return; 775 if (m == NULL) /* consumed by filter */ 776 return; 777 ip6 = mtod(m, struct ip6_hdr *); 778 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 779 780 if (m->m_flags & M_FASTFWD_OURS) { 781 m->m_flags &= ~M_FASTFWD_OURS; 782 ours = 1; 783 goto hbhcheck; 784 } 785 if ((m->m_flags & M_IP6_NEXTHOP) && 786 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 787 /* 788 * Directly ship the packet on. This allows forwarding 789 * packets originally destined to us to some other directly 790 * connected host. 791 */ 792 ip6_forward(m, 1); 793 return; 794 } 795 796 passin: 797 /* 798 * Disambiguate address scope zones (if there is ambiguity). 799 * We first make sure that the original source or destination address 800 * is not in our internal form for scoped addresses. Such addresses 801 * are not necessarily invalid spec-wise, but we cannot accept them due 802 * to the usage conflict. 803 * in6_setscope() then also checks and rejects the cases where src or 804 * dst are the loopback address and the receiving interface 805 * is not loopback. 806 */ 807 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 808 IP6STAT_INC(ip6s_badscope); /* XXX */ 809 goto bad; 810 } 811 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 812 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 813 IP6STAT_INC(ip6s_badscope); 814 goto bad; 815 } 816 /* 817 * Multicast check. Assume packet is for us to avoid 818 * prematurely taking locks. 819 */ 820 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 821 ours = 1; 822 in6_ifstat_inc(rcvif, ifs6_in_mcast); 823 goto hbhcheck; 824 } 825 /* 826 * Unicast check 827 * XXX: For now we keep link-local IPv6 addresses with embedded 828 * scope zone id, therefore we use zero zoneid here. 829 */ 830 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */); 831 if (ia != NULL) { 832 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 833 char ip6bufs[INET6_ADDRSTRLEN]; 834 char ip6bufd[INET6_ADDRSTRLEN]; 835 /* address is not ready, so discard the packet. */ 836 nd6log((LOG_INFO, 837 "ip6_input: packet to an unready address %s->%s\n", 838 ip6_sprintf(ip6bufs, &ip6->ip6_src), 839 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 840 ifa_free(&ia->ia_ifa); 841 goto bad; 842 } 843 /* Count the packet in the ip address stats */ 844 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 845 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 846 ifa_free(&ia->ia_ifa); 847 ours = 1; 848 goto hbhcheck; 849 } 850 851 /* 852 * Now there is no reason to process the packet if it's not our own 853 * and we're not a router. 854 */ 855 if (!V_ip6_forwarding) { 856 IP6STAT_INC(ip6s_cantforward); 857 goto bad; 858 } 859 860 hbhcheck: 861 /* 862 * Process Hop-by-Hop options header if it's contained. 863 * m may be modified in ip6_hopopts_input(). 864 * If a JumboPayload option is included, plen will also be modified. 865 */ 866 plen = (u_int32_t)ntohs(ip6->ip6_plen); 867 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 868 if (ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours) != 0) 869 return; 870 } else 871 nxt = ip6->ip6_nxt; 872 873 /* 874 * Use mbuf flags to propagate Router Alert option to 875 * ICMPv6 layer, as hop-by-hop options have been stripped. 876 */ 877 if (rtalert != ~0) 878 m->m_flags |= M_RTALERT_MLD; 879 880 /* 881 * Check that the amount of data in the buffers 882 * is as at least much as the IPv6 header would have us expect. 883 * Trim mbufs if longer than we expect. 884 * Drop packet if shorter than we expect. 885 */ 886 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 887 IP6STAT_INC(ip6s_tooshort); 888 in6_ifstat_inc(rcvif, ifs6_in_truncated); 889 goto bad; 890 } 891 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 892 if (m->m_len == m->m_pkthdr.len) { 893 m->m_len = sizeof(struct ip6_hdr) + plen; 894 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 895 } else 896 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 897 } 898 899 /* 900 * Forward if desirable. 901 */ 902 if (V_ip6_mrouter && 903 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 904 /* 905 * If we are acting as a multicast router, all 906 * incoming multicast packets are passed to the 907 * kernel-level multicast forwarding function. 908 * The packet is returned (relatively) intact; if 909 * ip6_mforward() returns a non-zero value, the packet 910 * must be discarded, else it may be accepted below. 911 * 912 * XXX TODO: Check hlim and multicast scope here to avoid 913 * unnecessarily calling into ip6_mforward(). 914 */ 915 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 916 IP6STAT_INC(ip6s_cantforward); 917 goto bad; 918 } 919 } else if (!ours) { 920 ip6_forward(m, srcrt); 921 return; 922 } 923 924 ip6 = mtod(m, struct ip6_hdr *); 925 926 /* 927 * Malicious party may be able to use IPv4 mapped addr to confuse 928 * tcp/udp stack and bypass security checks (act as if it was from 929 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 930 * 931 * For SIIT end node behavior, you may want to disable the check. 932 * However, you will become vulnerable to attacks using IPv4 mapped 933 * source. 934 */ 935 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 936 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 937 IP6STAT_INC(ip6s_badscope); 938 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 939 goto bad; 940 } 941 942 /* 943 * Tell launch routine the next header 944 */ 945 IP6STAT_INC(ip6s_delivered); 946 in6_ifstat_inc(rcvif, ifs6_in_deliver); 947 nest = 0; 948 949 while (nxt != IPPROTO_DONE) { 950 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 951 IP6STAT_INC(ip6s_toomanyhdr); 952 goto bad; 953 } 954 955 /* 956 * protection against faulty packet - there should be 957 * more sanity checks in header chain processing. 958 */ 959 if (m->m_pkthdr.len < off) { 960 IP6STAT_INC(ip6s_tooshort); 961 in6_ifstat_inc(rcvif, ifs6_in_truncated); 962 goto bad; 963 } 964 965 #ifdef IPSEC 966 /* 967 * enforce IPsec policy checking if we are seeing last header. 968 * note that we do not visit this with protocols with pcb layer 969 * code - like udp/tcp/raw ip. 970 */ 971 if (ip6_ipsec_input(m, nxt)) 972 goto bad; 973 #endif /* IPSEC */ 974 975 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 976 } 977 return; 978 bad: 979 in6_ifstat_inc(rcvif, ifs6_in_discard); 980 if (m != NULL) 981 m_freem(m); 982 } 983 984 /* 985 * Hop-by-Hop options header processing. If a valid jumbo payload option is 986 * included, the real payload length will be stored in plenp. 987 * 988 * rtalertp - XXX: should be stored more smart way 989 */ 990 static int 991 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 992 struct mbuf **mp, int *offp) 993 { 994 struct mbuf *m = *mp; 995 int off = *offp, hbhlen; 996 struct ip6_hbh *hbh; 997 998 /* validation of the length of the header */ 999 #ifndef PULLDOWN_TEST 1000 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); 1001 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 1002 hbhlen = (hbh->ip6h_len + 1) << 3; 1003 1004 IP6_EXTHDR_CHECK(m, off, hbhlen, -1); 1005 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 1006 #else 1007 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, 1008 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); 1009 if (hbh == NULL) { 1010 IP6STAT_INC(ip6s_tooshort); 1011 return -1; 1012 } 1013 hbhlen = (hbh->ip6h_len + 1) << 3; 1014 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 1015 hbhlen); 1016 if (hbh == NULL) { 1017 IP6STAT_INC(ip6s_tooshort); 1018 return -1; 1019 } 1020 #endif 1021 off += hbhlen; 1022 hbhlen -= sizeof(struct ip6_hbh); 1023 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 1024 hbhlen, rtalertp, plenp) < 0) 1025 return (-1); 1026 1027 *offp = off; 1028 *mp = m; 1029 return (0); 1030 } 1031 1032 /* 1033 * Search header for all Hop-by-hop options and process each option. 1034 * This function is separate from ip6_hopopts_input() in order to 1035 * handle a case where the sending node itself process its hop-by-hop 1036 * options header. In such a case, the function is called from ip6_output(). 1037 * 1038 * The function assumes that hbh header is located right after the IPv6 header 1039 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1040 * opthead + hbhlen is located in contiguous memory region. 1041 */ 1042 int 1043 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1044 u_int32_t *rtalertp, u_int32_t *plenp) 1045 { 1046 struct ip6_hdr *ip6; 1047 int optlen = 0; 1048 u_int8_t *opt = opthead; 1049 u_int16_t rtalert_val; 1050 u_int32_t jumboplen; 1051 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1052 1053 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1054 switch (*opt) { 1055 case IP6OPT_PAD1: 1056 optlen = 1; 1057 break; 1058 case IP6OPT_PADN: 1059 if (hbhlen < IP6OPT_MINLEN) { 1060 IP6STAT_INC(ip6s_toosmall); 1061 goto bad; 1062 } 1063 optlen = *(opt + 1) + 2; 1064 break; 1065 case IP6OPT_ROUTER_ALERT: 1066 /* XXX may need check for alignment */ 1067 if (hbhlen < IP6OPT_RTALERT_LEN) { 1068 IP6STAT_INC(ip6s_toosmall); 1069 goto bad; 1070 } 1071 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1072 /* XXX stat */ 1073 icmp6_error(m, ICMP6_PARAM_PROB, 1074 ICMP6_PARAMPROB_HEADER, 1075 erroff + opt + 1 - opthead); 1076 return (-1); 1077 } 1078 optlen = IP6OPT_RTALERT_LEN; 1079 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1080 *rtalertp = ntohs(rtalert_val); 1081 break; 1082 case IP6OPT_JUMBO: 1083 /* XXX may need check for alignment */ 1084 if (hbhlen < IP6OPT_JUMBO_LEN) { 1085 IP6STAT_INC(ip6s_toosmall); 1086 goto bad; 1087 } 1088 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1089 /* XXX stat */ 1090 icmp6_error(m, ICMP6_PARAM_PROB, 1091 ICMP6_PARAMPROB_HEADER, 1092 erroff + opt + 1 - opthead); 1093 return (-1); 1094 } 1095 optlen = IP6OPT_JUMBO_LEN; 1096 1097 /* 1098 * IPv6 packets that have non 0 payload length 1099 * must not contain a jumbo payload option. 1100 */ 1101 ip6 = mtod(m, struct ip6_hdr *); 1102 if (ip6->ip6_plen) { 1103 IP6STAT_INC(ip6s_badoptions); 1104 icmp6_error(m, ICMP6_PARAM_PROB, 1105 ICMP6_PARAMPROB_HEADER, 1106 erroff + opt - opthead); 1107 return (-1); 1108 } 1109 1110 /* 1111 * We may see jumbolen in unaligned location, so 1112 * we'd need to perform bcopy(). 1113 */ 1114 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1115 jumboplen = (u_int32_t)htonl(jumboplen); 1116 1117 #if 1 1118 /* 1119 * if there are multiple jumbo payload options, 1120 * *plenp will be non-zero and the packet will be 1121 * rejected. 1122 * the behavior may need some debate in ipngwg - 1123 * multiple options does not make sense, however, 1124 * there's no explicit mention in specification. 1125 */ 1126 if (*plenp != 0) { 1127 IP6STAT_INC(ip6s_badoptions); 1128 icmp6_error(m, ICMP6_PARAM_PROB, 1129 ICMP6_PARAMPROB_HEADER, 1130 erroff + opt + 2 - opthead); 1131 return (-1); 1132 } 1133 #endif 1134 1135 /* 1136 * jumbo payload length must be larger than 65535. 1137 */ 1138 if (jumboplen <= IPV6_MAXPACKET) { 1139 IP6STAT_INC(ip6s_badoptions); 1140 icmp6_error(m, ICMP6_PARAM_PROB, 1141 ICMP6_PARAMPROB_HEADER, 1142 erroff + opt + 2 - opthead); 1143 return (-1); 1144 } 1145 *plenp = jumboplen; 1146 1147 break; 1148 default: /* unknown option */ 1149 if (hbhlen < IP6OPT_MINLEN) { 1150 IP6STAT_INC(ip6s_toosmall); 1151 goto bad; 1152 } 1153 optlen = ip6_unknown_opt(opt, m, 1154 erroff + opt - opthead); 1155 if (optlen == -1) 1156 return (-1); 1157 optlen += 2; 1158 break; 1159 } 1160 } 1161 1162 return (0); 1163 1164 bad: 1165 m_freem(m); 1166 return (-1); 1167 } 1168 1169 /* 1170 * Unknown option processing. 1171 * The third argument `off' is the offset from the IPv6 header to the option, 1172 * which is necessary if the IPv6 header the and option header and IPv6 header 1173 * is not contiguous in order to return an ICMPv6 error. 1174 */ 1175 int 1176 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1177 { 1178 struct ip6_hdr *ip6; 1179 1180 switch (IP6OPT_TYPE(*optp)) { 1181 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1182 return ((int)*(optp + 1)); 1183 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1184 m_freem(m); 1185 return (-1); 1186 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1187 IP6STAT_INC(ip6s_badoptions); 1188 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1189 return (-1); 1190 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1191 IP6STAT_INC(ip6s_badoptions); 1192 ip6 = mtod(m, struct ip6_hdr *); 1193 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1194 (m->m_flags & (M_BCAST|M_MCAST))) 1195 m_freem(m); 1196 else 1197 icmp6_error(m, ICMP6_PARAM_PROB, 1198 ICMP6_PARAMPROB_OPTION, off); 1199 return (-1); 1200 } 1201 1202 m_freem(m); /* XXX: NOTREACHED */ 1203 return (-1); 1204 } 1205 1206 /* 1207 * Create the "control" list for this pcb. 1208 * These functions will not modify mbuf chain at all. 1209 * 1210 * With KAME mbuf chain restriction: 1211 * The routine will be called from upper layer handlers like tcp6_input(). 1212 * Thus the routine assumes that the caller (tcp6_input) have already 1213 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the 1214 * very first mbuf on the mbuf chain. 1215 * 1216 * ip6_savecontrol_v4 will handle those options that are possible to be 1217 * set on a v4-mapped socket. 1218 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1219 * options and handle the v6-only ones itself. 1220 */ 1221 struct mbuf ** 1222 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1223 int *v4only) 1224 { 1225 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1226 1227 #ifdef SO_TIMESTAMP 1228 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1229 struct timeval tv; 1230 1231 microtime(&tv); 1232 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1233 SCM_TIMESTAMP, SOL_SOCKET); 1234 if (*mp) 1235 mp = &(*mp)->m_next; 1236 } 1237 #endif 1238 1239 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1240 /* RFC 2292 sec. 5 */ 1241 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1242 struct in6_pktinfo pi6; 1243 1244 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1245 #ifdef INET 1246 struct ip *ip; 1247 1248 ip = mtod(m, struct ip *); 1249 pi6.ipi6_addr.s6_addr32[0] = 0; 1250 pi6.ipi6_addr.s6_addr32[1] = 0; 1251 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1252 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1253 #else 1254 /* We won't hit this code */ 1255 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1256 #endif 1257 } else { 1258 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1259 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1260 } 1261 pi6.ipi6_ifindex = 1262 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1263 1264 *mp = sbcreatecontrol((caddr_t) &pi6, 1265 sizeof(struct in6_pktinfo), 1266 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1267 if (*mp) 1268 mp = &(*mp)->m_next; 1269 } 1270 1271 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1272 int hlim; 1273 1274 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1275 #ifdef INET 1276 struct ip *ip; 1277 1278 ip = mtod(m, struct ip *); 1279 hlim = ip->ip_ttl; 1280 #else 1281 /* We won't hit this code */ 1282 hlim = 0; 1283 #endif 1284 } else { 1285 hlim = ip6->ip6_hlim & 0xff; 1286 } 1287 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1288 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1289 IPPROTO_IPV6); 1290 if (*mp) 1291 mp = &(*mp)->m_next; 1292 } 1293 1294 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1295 int tclass; 1296 1297 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1298 #ifdef INET 1299 struct ip *ip; 1300 1301 ip = mtod(m, struct ip *); 1302 tclass = ip->ip_tos; 1303 #else 1304 /* We won't hit this code */ 1305 tclass = 0; 1306 #endif 1307 } else { 1308 u_int32_t flowinfo; 1309 1310 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1311 flowinfo >>= 20; 1312 tclass = flowinfo & 0xff; 1313 } 1314 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int), 1315 IPV6_TCLASS, IPPROTO_IPV6); 1316 if (*mp) 1317 mp = &(*mp)->m_next; 1318 } 1319 1320 if (v4only != NULL) { 1321 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1322 *v4only = 1; 1323 } else { 1324 *v4only = 0; 1325 } 1326 } 1327 1328 return (mp); 1329 } 1330 1331 void 1332 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) 1333 { 1334 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1335 int v4only = 0; 1336 1337 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); 1338 if (v4only) 1339 return; 1340 1341 /* 1342 * IPV6_HOPOPTS socket option. Recall that we required super-user 1343 * privilege for the option (see ip6_ctloutput), but it might be too 1344 * strict, since there might be some hop-by-hop options which can be 1345 * returned to normal user. 1346 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1347 */ 1348 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { 1349 /* 1350 * Check if a hop-by-hop options header is contatined in the 1351 * received packet, and if so, store the options as ancillary 1352 * data. Note that a hop-by-hop options header must be 1353 * just after the IPv6 header, which is assured through the 1354 * IPv6 input processing. 1355 */ 1356 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1357 struct ip6_hbh *hbh; 1358 int hbhlen = 0; 1359 #ifdef PULLDOWN_TEST 1360 struct mbuf *ext; 1361 #endif 1362 1363 #ifndef PULLDOWN_TEST 1364 hbh = (struct ip6_hbh *)(ip6 + 1); 1365 hbhlen = (hbh->ip6h_len + 1) << 3; 1366 #else 1367 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), 1368 ip6->ip6_nxt); 1369 if (ext == NULL) { 1370 IP6STAT_INC(ip6s_tooshort); 1371 return; 1372 } 1373 hbh = mtod(ext, struct ip6_hbh *); 1374 hbhlen = (hbh->ip6h_len + 1) << 3; 1375 if (hbhlen != ext->m_len) { 1376 m_freem(ext); 1377 IP6STAT_INC(ip6s_tooshort); 1378 return; 1379 } 1380 #endif 1381 1382 /* 1383 * XXX: We copy the whole header even if a 1384 * jumbo payload option is included, the option which 1385 * is to be removed before returning according to 1386 * RFC2292. 1387 * Note: this constraint is removed in RFC3542 1388 */ 1389 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1390 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1391 IPPROTO_IPV6); 1392 if (*mp) 1393 mp = &(*mp)->m_next; 1394 #ifdef PULLDOWN_TEST 1395 m_freem(ext); 1396 #endif 1397 } 1398 } 1399 1400 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1401 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1402 1403 /* 1404 * Search for destination options headers or routing 1405 * header(s) through the header chain, and stores each 1406 * header as ancillary data. 1407 * Note that the order of the headers remains in 1408 * the chain of ancillary data. 1409 */ 1410 while (1) { /* is explicit loop prevention necessary? */ 1411 struct ip6_ext *ip6e = NULL; 1412 int elen; 1413 #ifdef PULLDOWN_TEST 1414 struct mbuf *ext = NULL; 1415 #endif 1416 1417 /* 1418 * if it is not an extension header, don't try to 1419 * pull it from the chain. 1420 */ 1421 switch (nxt) { 1422 case IPPROTO_DSTOPTS: 1423 case IPPROTO_ROUTING: 1424 case IPPROTO_HOPOPTS: 1425 case IPPROTO_AH: /* is it possible? */ 1426 break; 1427 default: 1428 goto loopend; 1429 } 1430 1431 #ifndef PULLDOWN_TEST 1432 if (off + sizeof(*ip6e) > m->m_len) 1433 goto loopend; 1434 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1435 if (nxt == IPPROTO_AH) 1436 elen = (ip6e->ip6e_len + 2) << 2; 1437 else 1438 elen = (ip6e->ip6e_len + 1) << 3; 1439 if (off + elen > m->m_len) 1440 goto loopend; 1441 #else 1442 ext = ip6_pullexthdr(m, off, nxt); 1443 if (ext == NULL) { 1444 IP6STAT_INC(ip6s_tooshort); 1445 return; 1446 } 1447 ip6e = mtod(ext, struct ip6_ext *); 1448 if (nxt == IPPROTO_AH) 1449 elen = (ip6e->ip6e_len + 2) << 2; 1450 else 1451 elen = (ip6e->ip6e_len + 1) << 3; 1452 if (elen != ext->m_len) { 1453 m_freem(ext); 1454 IP6STAT_INC(ip6s_tooshort); 1455 return; 1456 } 1457 #endif 1458 1459 switch (nxt) { 1460 case IPPROTO_DSTOPTS: 1461 if (!(in6p->inp_flags & IN6P_DSTOPTS)) 1462 break; 1463 1464 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1465 IS2292(in6p, 1466 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1467 IPPROTO_IPV6); 1468 if (*mp) 1469 mp = &(*mp)->m_next; 1470 break; 1471 case IPPROTO_ROUTING: 1472 if (!(in6p->inp_flags & IN6P_RTHDR)) 1473 break; 1474 1475 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1476 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), 1477 IPPROTO_IPV6); 1478 if (*mp) 1479 mp = &(*mp)->m_next; 1480 break; 1481 case IPPROTO_HOPOPTS: 1482 case IPPROTO_AH: /* is it possible? */ 1483 break; 1484 1485 default: 1486 /* 1487 * other cases have been filtered in the above. 1488 * none will visit this case. here we supply 1489 * the code just in case (nxt overwritten or 1490 * other cases). 1491 */ 1492 #ifdef PULLDOWN_TEST 1493 m_freem(ext); 1494 #endif 1495 goto loopend; 1496 1497 } 1498 1499 /* proceed with the next header. */ 1500 off += elen; 1501 nxt = ip6e->ip6e_nxt; 1502 ip6e = NULL; 1503 #ifdef PULLDOWN_TEST 1504 m_freem(ext); 1505 ext = NULL; 1506 #endif 1507 } 1508 loopend: 1509 ; 1510 } 1511 1512 if (in6p->inp_flags2 & INP_RECVFLOWID) { 1513 uint32_t flowid, flow_type; 1514 1515 flowid = m->m_pkthdr.flowid; 1516 flow_type = M_HASHTYPE_GET(m); 1517 1518 /* 1519 * XXX should handle the failure of one or the 1520 * other - don't populate both? 1521 */ 1522 *mp = sbcreatecontrol((caddr_t) &flowid, 1523 sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6); 1524 if (*mp) 1525 mp = &(*mp)->m_next; 1526 *mp = sbcreatecontrol((caddr_t) &flow_type, 1527 sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6); 1528 if (*mp) 1529 mp = &(*mp)->m_next; 1530 } 1531 1532 #ifdef RSS 1533 if (in6p->inp_flags2 & INP_RECVRSSBUCKETID) { 1534 uint32_t flowid, flow_type; 1535 uint32_t rss_bucketid; 1536 1537 flowid = m->m_pkthdr.flowid; 1538 flow_type = M_HASHTYPE_GET(m); 1539 1540 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1541 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1542 sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6); 1543 if (*mp) 1544 mp = &(*mp)->m_next; 1545 } 1546 } 1547 #endif 1548 1549 } 1550 #undef IS2292 1551 1552 void 1553 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1554 { 1555 struct socket *so; 1556 struct mbuf *m_mtu; 1557 struct ip6_mtuinfo mtuctl; 1558 1559 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1560 /* 1561 * Notify the error by sending IPV6_PATHMTU ancillary data if 1562 * application wanted to know the MTU value. 1563 * NOTE: we notify disconnected sockets, because some udp 1564 * applications keep sending sockets disconnected. 1565 * NOTE: our implementation doesn't notify connected sockets that has 1566 * foreign address that is different than given destination addresses 1567 * (this is permitted by RFC 3542). 1568 */ 1569 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1570 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1571 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1572 return; 1573 1574 mtuctl.ip6m_mtu = mtu; 1575 mtuctl.ip6m_addr = *dst; 1576 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1577 return; 1578 1579 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1580 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1581 return; 1582 1583 so = inp->inp_socket; 1584 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1585 == 0) { 1586 m_freem(m_mtu); 1587 /* XXX: should count statistics */ 1588 } else 1589 sorwakeup(so); 1590 } 1591 1592 #ifdef PULLDOWN_TEST 1593 /* 1594 * pull single extension header from mbuf chain. returns single mbuf that 1595 * contains the result, or NULL on error. 1596 */ 1597 static struct mbuf * 1598 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) 1599 { 1600 struct ip6_ext ip6e; 1601 size_t elen; 1602 struct mbuf *n; 1603 1604 #ifdef DIAGNOSTIC 1605 switch (nxt) { 1606 case IPPROTO_DSTOPTS: 1607 case IPPROTO_ROUTING: 1608 case IPPROTO_HOPOPTS: 1609 case IPPROTO_AH: /* is it possible? */ 1610 break; 1611 default: 1612 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); 1613 } 1614 #endif 1615 1616 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1617 if (nxt == IPPROTO_AH) 1618 elen = (ip6e.ip6e_len + 2) << 2; 1619 else 1620 elen = (ip6e.ip6e_len + 1) << 3; 1621 1622 if (elen > MLEN) 1623 n = m_getcl(M_NOWAIT, MT_DATA, 0); 1624 else 1625 n = m_get(M_NOWAIT, MT_DATA); 1626 if (n == NULL) 1627 return NULL; 1628 1629 m_copydata(m, off, elen, mtod(n, caddr_t)); 1630 n->m_len = elen; 1631 return n; 1632 } 1633 #endif 1634 1635 /* 1636 * Get pointer to the previous header followed by the header 1637 * currently processed. 1638 * XXX: This function supposes that 1639 * M includes all headers, 1640 * the next header field and the header length field of each header 1641 * are valid, and 1642 * the sum of each header length equals to OFF. 1643 * Because of these assumptions, this function must be called very 1644 * carefully. Moreover, it will not be used in the near future when 1645 * we develop `neater' mechanism to process extension headers. 1646 */ 1647 char * 1648 ip6_get_prevhdr(const struct mbuf *m, int off) 1649 { 1650 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1651 1652 if (off == sizeof(struct ip6_hdr)) 1653 return (&ip6->ip6_nxt); 1654 else { 1655 int len, nxt; 1656 struct ip6_ext *ip6e = NULL; 1657 1658 nxt = ip6->ip6_nxt; 1659 len = sizeof(struct ip6_hdr); 1660 while (len < off) { 1661 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len); 1662 1663 switch (nxt) { 1664 case IPPROTO_FRAGMENT: 1665 len += sizeof(struct ip6_frag); 1666 break; 1667 case IPPROTO_AH: 1668 len += (ip6e->ip6e_len + 2) << 2; 1669 break; 1670 default: 1671 len += (ip6e->ip6e_len + 1) << 3; 1672 break; 1673 } 1674 nxt = ip6e->ip6e_nxt; 1675 } 1676 if (ip6e) 1677 return (&ip6e->ip6e_nxt); 1678 else 1679 return NULL; 1680 } 1681 } 1682 1683 /* 1684 * get next header offset. m will be retained. 1685 */ 1686 int 1687 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1688 { 1689 struct ip6_hdr ip6; 1690 struct ip6_ext ip6e; 1691 struct ip6_frag fh; 1692 1693 /* just in case */ 1694 if (m == NULL) 1695 panic("ip6_nexthdr: m == NULL"); 1696 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1697 return -1; 1698 1699 switch (proto) { 1700 case IPPROTO_IPV6: 1701 if (m->m_pkthdr.len < off + sizeof(ip6)) 1702 return -1; 1703 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1704 if (nxtp) 1705 *nxtp = ip6.ip6_nxt; 1706 off += sizeof(ip6); 1707 return off; 1708 1709 case IPPROTO_FRAGMENT: 1710 /* 1711 * terminate parsing if it is not the first fragment, 1712 * it does not make sense to parse through it. 1713 */ 1714 if (m->m_pkthdr.len < off + sizeof(fh)) 1715 return -1; 1716 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1717 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1718 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1719 return -1; 1720 if (nxtp) 1721 *nxtp = fh.ip6f_nxt; 1722 off += sizeof(struct ip6_frag); 1723 return off; 1724 1725 case IPPROTO_AH: 1726 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1727 return -1; 1728 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1729 if (nxtp) 1730 *nxtp = ip6e.ip6e_nxt; 1731 off += (ip6e.ip6e_len + 2) << 2; 1732 return off; 1733 1734 case IPPROTO_HOPOPTS: 1735 case IPPROTO_ROUTING: 1736 case IPPROTO_DSTOPTS: 1737 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1738 return -1; 1739 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1740 if (nxtp) 1741 *nxtp = ip6e.ip6e_nxt; 1742 off += (ip6e.ip6e_len + 1) << 3; 1743 return off; 1744 1745 case IPPROTO_NONE: 1746 case IPPROTO_ESP: 1747 case IPPROTO_IPCOMP: 1748 /* give up */ 1749 return -1; 1750 1751 default: 1752 return -1; 1753 } 1754 1755 /* NOTREACHED */ 1756 } 1757 1758 /* 1759 * get offset for the last header in the chain. m will be kept untainted. 1760 */ 1761 int 1762 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1763 { 1764 int newoff; 1765 int nxt; 1766 1767 if (!nxtp) { 1768 nxt = -1; 1769 nxtp = &nxt; 1770 } 1771 while (1) { 1772 newoff = ip6_nexthdr(m, off, proto, nxtp); 1773 if (newoff < 0) 1774 return off; 1775 else if (newoff < off) 1776 return -1; /* invalid */ 1777 else if (newoff == off) 1778 return newoff; 1779 1780 off = newoff; 1781 proto = *nxtp; 1782 } 1783 } 1784 1785 /* 1786 * System control for IP6 1787 */ 1788 1789 u_char inet6ctlerrmap[PRC_NCMDS] = { 1790 0, 0, 0, 0, 1791 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1792 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1793 EMSGSIZE, EHOSTUNREACH, 0, 0, 1794 0, 0, EHOSTUNREACH, 0, 1795 ENOPROTOOPT, ECONNREFUSED 1796 }; 1797