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