1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * $KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $ 32 */ 33 34 /*- 35 * Copyright (c) 1982, 1986, 1988, 1993 36 * The Regents of the University of California. All rights reserved. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. Neither the name of the University nor the names of its contributors 47 * may be used to endorse or promote products derived from this software 48 * without specific prior written permission. 49 * 50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 60 * SUCH DAMAGE. 61 * 62 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 63 */ 64 65 #include <sys/cdefs.h> 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 #include "opt_sctp.h" 72 73 #include <sys/param.h> 74 #include <sys/systm.h> 75 #include <sys/hhook.h> 76 #include <sys/malloc.h> 77 #include <sys/mbuf.h> 78 #include <sys/proc.h> 79 #include <sys/domain.h> 80 #include <sys/protosw.h> 81 #include <sys/sdt.h> 82 #include <sys/socket.h> 83 #include <sys/socketvar.h> 84 #include <sys/errno.h> 85 #include <sys/time.h> 86 #include <sys/kernel.h> 87 #include <sys/lock.h> 88 #include <sys/rmlock.h> 89 #include <sys/syslog.h> 90 #include <sys/sysctl.h> 91 #include <sys/eventhandler.h> 92 93 #include <net/if.h> 94 #include <net/if_var.h> 95 #include <net/if_types.h> 96 #include <net/if_private.h> 97 #include <net/if_dl.h> 98 #include <net/route.h> 99 #include <net/netisr.h> 100 #include <net/rss_config.h> 101 #include <net/pfil.h> 102 #include <net/vnet.h> 103 104 #include <netinet/in.h> 105 #include <netinet/in_kdtrace.h> 106 #include <netinet/ip_var.h> 107 #include <netinet/in_systm.h> 108 #include <net/if_llatbl.h> 109 #ifdef INET 110 #include <netinet/ip.h> 111 #include <netinet/ip_icmp.h> 112 #endif /* INET */ 113 #include <netinet/ip6.h> 114 #include <netinet6/in6_var.h> 115 #include <netinet6/ip6_var.h> 116 #include <netinet/ip_encap.h> 117 #include <netinet/in_pcb.h> 118 #include <netinet/icmp6.h> 119 #include <netinet6/scope6_var.h> 120 #include <netinet6/in6_ifattach.h> 121 #include <netinet6/mld6_var.h> 122 #include <netinet6/nd6.h> 123 #include <netinet6/in6_rss.h> 124 #ifdef SCTP 125 #include <netinet/sctp_pcb.h> 126 #include <netinet6/sctp6_var.h> 127 #endif 128 129 #include <netipsec/ipsec_support.h> 130 131 ip6proto_input_t *ip6_protox[IPPROTO_MAX] = { 132 [0 ... IPPROTO_MAX - 1] = rip6_input }; 133 ip6proto_ctlinput_t *ip6_ctlprotox[IPPROTO_MAX] = { 134 [0 ... IPPROTO_MAX - 1] = rip6_ctlinput }; 135 136 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 137 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); 138 VNET_DEFINE(u_long, in6_ifaddrhmask); 139 140 static struct netisr_handler ip6_nh = { 141 .nh_name = "ip6", 142 .nh_handler = ip6_input, 143 .nh_proto = NETISR_IPV6, 144 #ifdef RSS 145 .nh_m2cpuid = rss_soft_m2cpuid_v6, 146 .nh_policy = NETISR_POLICY_CPU, 147 .nh_dispatch = NETISR_DISPATCH_HYBRID, 148 #else 149 .nh_policy = NETISR_POLICY_FLOW, 150 #endif 151 }; 152 153 static int 154 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 155 { 156 int error, qlimit; 157 158 netisr_getqlimit(&ip6_nh, &qlimit); 159 error = sysctl_handle_int(oidp, &qlimit, 0, req); 160 if (error || !req->newptr) 161 return (error); 162 if (qlimit < 1) 163 return (EINVAL); 164 return (netisr_setqlimit(&ip6_nh, qlimit)); 165 } 166 SYSCTL_DECL(_net_inet6_ip6); 167 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen, 168 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 169 0, 0, sysctl_netinet6_intr_queue_maxlen, "I", 170 "Maximum size of the IPv6 input queue"); 171 172 VNET_DEFINE_STATIC(bool, ip6_sav) = true; 173 #define V_ip6_sav VNET(ip6_sav) 174 SYSCTL_BOOL(_net_inet6_ip6, OID_AUTO, source_address_validation, 175 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_sav), true, 176 "Drop incoming packets with source address that is a local address"); 177 178 #ifdef RSS 179 static struct netisr_handler ip6_direct_nh = { 180 .nh_name = "ip6_direct", 181 .nh_handler = ip6_direct_input, 182 .nh_proto = NETISR_IPV6_DIRECT, 183 .nh_m2cpuid = rss_soft_m2cpuid_v6, 184 .nh_policy = NETISR_POLICY_CPU, 185 .nh_dispatch = NETISR_DISPATCH_HYBRID, 186 }; 187 188 static int 189 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 190 { 191 int error, qlimit; 192 193 netisr_getqlimit(&ip6_direct_nh, &qlimit); 194 error = sysctl_handle_int(oidp, &qlimit, 0, req); 195 if (error || !req->newptr) 196 return (error); 197 if (qlimit < 1) 198 return (EINVAL); 199 return (netisr_setqlimit(&ip6_direct_nh, qlimit)); 200 } 201 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 202 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 203 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I", 204 "Maximum size of the IPv6 direct input queue"); 205 206 #endif 207 208 VNET_DEFINE(pfil_head_t, inet6_pfil_head); 209 VNET_DEFINE(pfil_head_t, inet6_local_pfil_head); 210 211 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 212 VNET_PCPUSTAT_SYSINIT(ip6stat); 213 #ifdef VIMAGE 214 VNET_PCPUSTAT_SYSUNINIT(ip6stat); 215 #endif /* VIMAGE */ 216 217 struct rmlock in6_ifaddr_lock; 218 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 219 220 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 221 222 /* 223 * IP6 initialization: fill in IP6 protocol switch table. 224 * All protocols not implemented in kernel go to raw IP6 protocol handler. 225 */ 226 static void 227 ip6_vnet_init(void *arg __unused) 228 { 229 struct pfil_head_args args; 230 231 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 232 &V_ip6_auto_linklocal); 233 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 234 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 235 236 CK_STAILQ_INIT(&V_in6_ifaddrhead); 237 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 238 &V_in6_ifaddrhmask); 239 240 /* Initialize packet filter hooks. */ 241 args.pa_version = PFIL_VERSION; 242 args.pa_flags = PFIL_IN | PFIL_OUT; 243 args.pa_type = PFIL_TYPE_IP6; 244 args.pa_headname = PFIL_INET6_NAME; 245 V_inet6_pfil_head = pfil_head_register(&args); 246 247 args.pa_flags = PFIL_OUT; 248 args.pa_headname = PFIL_INET6_LOCAL_NAME; 249 V_inet6_local_pfil_head = pfil_head_register(&args); 250 251 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 252 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 253 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 254 printf("%s: WARNING: unable to register input helper hook\n", 255 __func__); 256 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 257 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 258 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 259 printf("%s: WARNING: unable to register output helper hook\n", 260 __func__); 261 262 scope6_init(); 263 addrsel_policy_init(); 264 nd6_init(); 265 frag6_init(); 266 267 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 268 269 /* Skip global initialization stuff for non-default instances. */ 270 #ifdef VIMAGE 271 netisr_register_vnet(&ip6_nh); 272 #ifdef RSS 273 netisr_register_vnet(&ip6_direct_nh); 274 #endif 275 #endif 276 } 277 VNET_SYSINIT(ip6_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, 278 ip6_vnet_init, NULL); 279 280 static void 281 ip6_init(void *arg __unused) 282 { 283 284 /* 285 * Register statically those protocols that are unlikely to ever go 286 * dynamic. 287 */ 288 IP6PROTO_REGISTER(IPPROTO_ICMPV6, icmp6_input, rip6_ctlinput); 289 IP6PROTO_REGISTER(IPPROTO_DSTOPTS, dest6_input, NULL); 290 IP6PROTO_REGISTER(IPPROTO_ROUTING, route6_input, NULL); 291 IP6PROTO_REGISTER(IPPROTO_FRAGMENT, frag6_input, NULL); 292 IP6PROTO_REGISTER(IPPROTO_IPV4, encap6_input, NULL); 293 IP6PROTO_REGISTER(IPPROTO_IPV6, encap6_input, NULL); 294 IP6PROTO_REGISTER(IPPROTO_ETHERIP, encap6_input, NULL); 295 IP6PROTO_REGISTER(IPPROTO_GRE, encap6_input, NULL); 296 IP6PROTO_REGISTER(IPPROTO_PIM, encap6_input, NULL); 297 #ifdef SCTP /* XXX: has a loadable & static version */ 298 IP6PROTO_REGISTER(IPPROTO_SCTP, sctp6_input, sctp6_ctlinput); 299 #endif 300 301 EVENTHANDLER_REGISTER(vm_lowmem, frag6_drain, NULL, LOWMEM_PRI_DEFAULT); 302 EVENTHANDLER_REGISTER(mbuf_lowmem, frag6_drain, NULL, 303 LOWMEM_PRI_DEFAULT); 304 305 netisr_register(&ip6_nh); 306 #ifdef RSS 307 netisr_register(&ip6_direct_nh); 308 #endif 309 } 310 SYSINIT(ip6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init, NULL); 311 312 int 313 ip6proto_register(uint8_t proto, ip6proto_input_t input, 314 ip6proto_ctlinput_t ctl) 315 { 316 317 MPASS(proto > 0); 318 319 if (ip6_protox[proto] == rip6_input) { 320 ip6_protox[proto] = input; 321 ip6_ctlprotox[proto] = ctl; 322 return (0); 323 } else 324 return (EEXIST); 325 } 326 327 int 328 ip6proto_unregister(uint8_t proto) 329 { 330 331 MPASS(proto > 0); 332 333 if (ip6_protox[proto] != rip6_input) { 334 ip6_protox[proto] = rip6_input; 335 ip6_ctlprotox[proto] = rip6_ctlinput; 336 return (0); 337 } else 338 return (ENOENT); 339 } 340 341 #ifdef VIMAGE 342 static void 343 ip6_destroy(void *unused __unused) 344 { 345 struct ifaddr *ifa, *nifa; 346 struct ifnet *ifp; 347 int error; 348 349 #ifdef RSS 350 netisr_unregister_vnet(&ip6_direct_nh); 351 #endif 352 netisr_unregister_vnet(&ip6_nh); 353 354 pfil_head_unregister(V_inet6_pfil_head); 355 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 356 if (error != 0) { 357 printf("%s: WARNING: unable to deregister input helper hook " 358 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 359 "error %d returned\n", __func__, error); 360 } 361 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 362 if (error != 0) { 363 printf("%s: WARNING: unable to deregister output helper hook " 364 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 365 "error %d returned\n", __func__, error); 366 } 367 368 /* Cleanup addresses. */ 369 IFNET_RLOCK(); 370 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 371 /* Cannot lock here - lock recursion. */ 372 /* IF_ADDR_LOCK(ifp); */ 373 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 374 if (ifa->ifa_addr->sa_family != AF_INET6) 375 continue; 376 in6_purgeaddr(ifa); 377 } 378 /* IF_ADDR_UNLOCK(ifp); */ 379 in6_ifdetach_destroy(ifp); 380 mld_domifdetach(ifp); 381 } 382 IFNET_RUNLOCK(); 383 384 /* Make sure any routes are gone as well. */ 385 rib_flush_routes_family(AF_INET6); 386 387 frag6_destroy(); 388 nd6_destroy(); 389 in6_ifattach_destroy(); 390 391 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 392 } 393 394 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 395 #endif 396 397 static int 398 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off, 399 int *nxt, int *ours) 400 { 401 struct mbuf *m; 402 struct ip6_hdr *ip6; 403 struct ip6_hbh *hbh; 404 405 if (ip6_hopopts_input(plen, rtalert, mp, off)) { 406 #if 0 /*touches NULL pointer*/ 407 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard); 408 #endif 409 goto out; /* m have already been freed */ 410 } 411 412 /* adjust pointer */ 413 m = *mp; 414 ip6 = mtod(m, struct ip6_hdr *); 415 416 /* 417 * if the payload length field is 0 and the next header field 418 * indicates Hop-by-Hop Options header, then a Jumbo Payload 419 * option MUST be included. 420 */ 421 if (ip6->ip6_plen == 0 && *plen == 0) { 422 /* 423 * Note that if a valid jumbo payload option is 424 * contained, ip6_hopopts_input() must set a valid 425 * (non-zero) payload length to the variable plen. 426 */ 427 IP6STAT_INC(ip6s_badoptions); 428 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 429 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 430 icmp6_error(m, ICMP6_PARAM_PROB, 431 ICMP6_PARAMPROB_HEADER, 432 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 433 goto out; 434 } 435 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 436 hbh = (struct ip6_hbh *)(ip6 + 1); 437 *nxt = hbh->ip6h_nxt; 438 439 /* 440 * If we are acting as a router and the packet contains a 441 * router alert option, see if we know the option value. 442 * Currently, we only support the option value for MLD, in which 443 * case we should pass the packet to the multicast routing 444 * daemon. 445 */ 446 if (*rtalert != ~0) { 447 switch (*rtalert) { 448 case IP6OPT_RTALERT_MLD: 449 if (V_ip6_forwarding) 450 *ours = 1; 451 break; 452 default: 453 /* 454 * RFC2711 requires unrecognized values must be 455 * silently ignored. 456 */ 457 break; 458 } 459 } 460 461 return (0); 462 463 out: 464 return (1); 465 } 466 467 #ifdef RSS 468 /* 469 * IPv6 direct input routine. 470 * 471 * This is called when reinjecting completed fragments where 472 * all of the previous checking and book-keeping has been done. 473 */ 474 void 475 ip6_direct_input(struct mbuf *m) 476 { 477 int off, nxt; 478 int nest; 479 struct m_tag *mtag; 480 struct ip6_direct_ctx *ip6dc; 481 482 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 483 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 484 485 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 486 nxt = ip6dc->ip6dc_nxt; 487 off = ip6dc->ip6dc_off; 488 489 nest = 0; 490 491 m_tag_delete(m, mtag); 492 493 while (nxt != IPPROTO_DONE) { 494 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 495 IP6STAT_INC(ip6s_toomanyhdr); 496 goto bad; 497 } 498 499 /* 500 * protection against faulty packet - there should be 501 * more sanity checks in header chain processing. 502 */ 503 if (m->m_pkthdr.len < off) { 504 IP6STAT_INC(ip6s_tooshort); 505 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 506 goto bad; 507 } 508 509 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 510 if (IPSEC_ENABLED(ipv6)) { 511 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 512 return; 513 } 514 #endif /* IPSEC */ 515 516 nxt = ip6_protox[nxt](&m, &off, nxt); 517 } 518 return; 519 bad: 520 m_freem(m); 521 } 522 #endif 523 524 void 525 ip6_input(struct mbuf *m) 526 { 527 struct in6_addr odst; 528 struct ip6_hdr *ip6; 529 struct in6_ifaddr *ia; 530 struct ifnet *rcvif; 531 u_int32_t plen; 532 u_int32_t rtalert = ~0; 533 int off = sizeof(struct ip6_hdr), nest; 534 int nxt, ours = 0; 535 int srcrt = 0; 536 537 /* 538 * Drop the packet if IPv6 operation is disabled on the interface. 539 */ 540 rcvif = m->m_pkthdr.rcvif; 541 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 542 goto bad; 543 544 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 545 /* 546 * should the inner packet be considered authentic? 547 * see comment in ah4_input(). 548 * NB: m cannot be NULL when passed to the input routine 549 */ 550 551 m->m_flags &= ~M_AUTHIPHDR; 552 m->m_flags &= ~M_AUTHIPDGM; 553 554 #endif /* IPSEC */ 555 556 if (m->m_flags & M_FASTFWD_OURS) { 557 /* 558 * Firewall changed destination to local. 559 */ 560 ip6 = mtod(m, struct ip6_hdr *); 561 goto passin; 562 } 563 564 /* 565 * mbuf statistics 566 */ 567 if (m->m_flags & M_EXT) { 568 if (m->m_next) 569 IP6STAT_INC(ip6s_mext2m); 570 else 571 IP6STAT_INC(ip6s_mext1); 572 } else { 573 if (m->m_next) { 574 struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif; 575 int ifindex = ifp->if_index; 576 if (ifindex >= IP6S_M2MMAX) 577 ifindex = 0; 578 IP6STAT_INC(ip6s_m2m[ifindex]); 579 } else 580 IP6STAT_INC(ip6s_m1); 581 } 582 583 in6_ifstat_inc(rcvif, ifs6_in_receive); 584 IP6STAT_INC(ip6s_total); 585 586 /* 587 * L2 bridge code and some other code can return mbuf chain 588 * that does not conform to KAME requirement. too bad. 589 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 590 */ 591 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 592 struct mbuf *n; 593 594 if (m->m_pkthdr.len > MHLEN) 595 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 596 else 597 n = m_gethdr(M_NOWAIT, MT_DATA); 598 if (n == NULL) 599 goto bad; 600 601 m_move_pkthdr(n, m); 602 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 603 n->m_len = n->m_pkthdr.len; 604 m_freem(m); 605 m = n; 606 } 607 if (m->m_len < sizeof(struct ip6_hdr)) { 608 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 609 IP6STAT_INC(ip6s_toosmall); 610 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 611 goto bad; 612 } 613 } 614 615 ip6 = mtod(m, struct ip6_hdr *); 616 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 617 IP6STAT_INC(ip6s_badvers); 618 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 619 goto bad; 620 } 621 622 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 623 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 624 625 /* 626 * Check against address spoofing/corruption. 627 */ 628 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 629 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 630 /* 631 * XXX: "badscope" is not very suitable for a multicast source. 632 */ 633 IP6STAT_INC(ip6s_badscope); 634 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 635 goto bad; 636 } 637 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 638 !(m->m_flags & M_LOOP)) { 639 /* 640 * In this case, the packet should come from the loopback 641 * interface. However, we cannot just check the if_flags, 642 * because ip6_mloopback() passes the "actual" interface 643 * as the outgoing/incoming interface. 644 */ 645 IP6STAT_INC(ip6s_badscope); 646 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 647 goto bad; 648 } 649 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 650 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 651 /* 652 * RFC4291 2.7: 653 * Nodes must not originate a packet to a multicast address 654 * whose scop field contains the reserved value 0; if such 655 * a packet is received, it must be silently dropped. 656 */ 657 IP6STAT_INC(ip6s_badscope); 658 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 659 goto bad; 660 } 661 /* 662 * The following check is not documented in specs. A malicious 663 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 664 * and bypass security checks (act as if it was from 127.0.0.1 by using 665 * IPv6 src ::ffff:127.0.0.1). Be cautious. 666 * 667 * We have supported IPv6-only kernels for a few years and this issue 668 * has not come up. The world seems to move mostly towards not using 669 * v4mapped on the wire, so it makes sense for us to keep rejecting 670 * any such packets. 671 */ 672 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 673 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 674 IP6STAT_INC(ip6s_badscope); 675 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 676 goto bad; 677 } 678 #if 0 679 /* 680 * Reject packets with IPv4 compatible addresses (auto tunnel). 681 * 682 * The code forbids auto tunnel relay case in RFC1933 (the check is 683 * stronger than RFC1933). We may want to re-enable it if mech-xx 684 * is revised to forbid relaying case. 685 */ 686 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 687 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 688 IP6STAT_INC(ip6s_badscope); 689 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 690 goto bad; 691 } 692 #endif 693 /* 694 * Try to forward the packet, but if we fail continue. 695 * ip6_tryforward() does not generate redirects, so fall 696 * through to normal processing if redirects are required. 697 * ip6_tryforward() does inbound and outbound packet firewall 698 * processing. If firewall has decided that destination becomes 699 * our local address, it sets M_FASTFWD_OURS flag. In this 700 * case skip another inbound firewall processing and update 701 * ip6 pointer. 702 */ 703 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0 704 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 705 && (!IPSEC_ENABLED(ipv6) || 706 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0) 707 #endif 708 ) { 709 if ((m = ip6_tryforward(m)) == NULL) 710 return; 711 if (m->m_flags & M_FASTFWD_OURS) { 712 ip6 = mtod(m, struct ip6_hdr *); 713 goto passin; 714 } 715 } 716 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 717 /* 718 * Bypass packet filtering for packets previously handled by IPsec. 719 */ 720 if (IPSEC_ENABLED(ipv6) && 721 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0) 722 goto passin; 723 #endif 724 /* 725 * Run through list of hooks for input packets. 726 * 727 * NB: Beware of the destination address changing 728 * (e.g. by NAT rewriting). When this happens, 729 * tell ip6_forward to do the right thing. 730 */ 731 732 /* Jump over all PFIL processing if hooks are not active. */ 733 if (!PFIL_HOOKED_IN(V_inet6_pfil_head)) 734 goto passin; 735 736 odst = ip6->ip6_dst; 737 if (pfil_mbuf_in(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif, 738 NULL) != PFIL_PASS) 739 return; 740 ip6 = mtod(m, struct ip6_hdr *); 741 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 742 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 743 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 744 /* 745 * Directly ship the packet on. This allows forwarding 746 * packets originally destined to us to some other directly 747 * connected host. 748 */ 749 ip6_forward(m, 1); 750 return; 751 } 752 753 passin: 754 /* 755 * Disambiguate address scope zones (if there is ambiguity). 756 * We first make sure that the original source or destination address 757 * is not in our internal form for scoped addresses. Such addresses 758 * are not necessarily invalid spec-wise, but we cannot accept them due 759 * to the usage conflict. 760 * in6_setscope() then also checks and rejects the cases where src or 761 * dst are the loopback address and the receiving interface 762 * is not loopback. 763 */ 764 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 765 IP6STAT_INC(ip6s_badscope); /* XXX */ 766 goto bad; 767 } 768 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 769 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 770 IP6STAT_INC(ip6s_badscope); 771 goto bad; 772 } 773 if (m->m_flags & M_FASTFWD_OURS) { 774 m->m_flags &= ~M_FASTFWD_OURS; 775 ours = 1; 776 goto hbhcheck; 777 } 778 /* 779 * Multicast check. Assume packet is for us to avoid 780 * prematurely taking locks. 781 */ 782 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 783 ours = 1; 784 in6_ifstat_inc(rcvif, ifs6_in_mcast); 785 goto hbhcheck; 786 } 787 /* 788 * Unicast check 789 * XXX: For now we keep link-local IPv6 addresses with embedded 790 * scope zone id, therefore we use zero zoneid here. 791 */ 792 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false); 793 if (ia != NULL) { 794 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 795 char ip6bufs[INET6_ADDRSTRLEN]; 796 char ip6bufd[INET6_ADDRSTRLEN]; 797 /* address is not ready, so discard the packet. */ 798 nd6log((LOG_INFO, 799 "ip6_input: packet to an unready address %s->%s\n", 800 ip6_sprintf(ip6bufs, &ip6->ip6_src), 801 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 802 goto bad; 803 } 804 if (V_ip6_sav && !(m->m_flags & M_LOOP) && 805 __predict_false(in6_localip_fib(&ip6->ip6_src, 806 rcvif->if_fib))) { 807 IP6STAT_INC(ip6s_badscope); /* XXX */ 808 goto bad; 809 } 810 /* Count the packet in the ip address stats */ 811 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 812 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 813 ours = 1; 814 goto hbhcheck; 815 } 816 817 /* 818 * Now there is no reason to process the packet if it's not our own 819 * and we're not a router. 820 */ 821 if (!V_ip6_forwarding) { 822 IP6STAT_INC(ip6s_cantforward); 823 goto bad; 824 } 825 826 hbhcheck: 827 /* 828 * Process Hop-by-Hop options header if it's contained. 829 * m may be modified in ip6_hopopts_input(). 830 * If a JumboPayload option is included, plen will also be modified. 831 */ 832 plen = (u_int32_t)ntohs(ip6->ip6_plen); 833 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 834 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0) 835 return; 836 } else 837 nxt = ip6->ip6_nxt; 838 839 /* 840 * Use mbuf flags to propagate Router Alert option to 841 * ICMPv6 layer, as hop-by-hop options have been stripped. 842 */ 843 if (rtalert != ~0) 844 m->m_flags |= M_RTALERT_MLD; 845 846 /* 847 * Check that the amount of data in the buffers 848 * is as at least much as the IPv6 header would have us expect. 849 * Trim mbufs if longer than we expect. 850 * Drop packet if shorter than we expect. 851 */ 852 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 853 IP6STAT_INC(ip6s_tooshort); 854 in6_ifstat_inc(rcvif, ifs6_in_truncated); 855 goto bad; 856 } 857 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 858 if (m->m_len == m->m_pkthdr.len) { 859 m->m_len = sizeof(struct ip6_hdr) + plen; 860 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 861 } else 862 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 863 } 864 865 /* 866 * Forward if desirable. 867 */ 868 if (V_ip6_mrouter && 869 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 870 /* 871 * If we are acting as a multicast router, all 872 * incoming multicast packets are passed to the 873 * kernel-level multicast forwarding function. 874 * The packet is returned (relatively) intact; if 875 * ip6_mforward() returns a non-zero value, the packet 876 * must be discarded, else it may be accepted below. 877 * 878 * XXX TODO: Check hlim and multicast scope here to avoid 879 * unnecessarily calling into ip6_mforward(). 880 */ 881 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 882 IP6STAT_INC(ip6s_cantforward); 883 goto bad; 884 } 885 } else if (!ours) { 886 ip6_forward(m, srcrt); 887 return; 888 } 889 890 /* 891 * We are going to ship the packet to the local protocol stack. Call the 892 * filter again for this 'output' action, allowing redirect-like rules 893 * to adjust the source address. 894 */ 895 if (PFIL_HOOKED_OUT(V_inet6_local_pfil_head)) { 896 if (pfil_mbuf_out(V_inet6_local_pfil_head, &m, V_loif, NULL) != 897 PFIL_PASS) 898 return; 899 if (m == NULL) /* consumed by filter */ 900 return; 901 ip6 = mtod(m, struct ip6_hdr *); 902 } 903 904 /* 905 * Tell launch routine the next header 906 */ 907 IP6STAT_INC(ip6s_delivered); 908 in6_ifstat_inc(rcvif, ifs6_in_deliver); 909 nest = 0; 910 911 while (nxt != IPPROTO_DONE) { 912 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 913 IP6STAT_INC(ip6s_toomanyhdr); 914 goto bad; 915 } 916 917 /* 918 * protection against faulty packet - there should be 919 * more sanity checks in header chain processing. 920 */ 921 if (m->m_pkthdr.len < off) { 922 IP6STAT_INC(ip6s_tooshort); 923 in6_ifstat_inc(rcvif, ifs6_in_truncated); 924 goto bad; 925 } 926 927 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 928 if (IPSEC_ENABLED(ipv6)) { 929 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 930 return; 931 } 932 #endif /* IPSEC */ 933 934 nxt = ip6_protox[nxt](&m, &off, nxt); 935 } 936 return; 937 bad: 938 in6_ifstat_inc(rcvif, ifs6_in_discard); 939 if (m != NULL) 940 m_freem(m); 941 } 942 943 /* 944 * Hop-by-Hop options header processing. If a valid jumbo payload option is 945 * included, the real payload length will be stored in plenp. 946 * 947 * rtalertp - XXX: should be stored more smart way 948 */ 949 static int 950 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 951 struct mbuf **mp, int *offp) 952 { 953 struct mbuf *m = *mp; 954 int off = *offp, hbhlen; 955 struct ip6_hbh *hbh; 956 957 /* validation of the length of the header */ 958 if (m->m_len < off + sizeof(*hbh)) { 959 m = m_pullup(m, off + sizeof(*hbh)); 960 if (m == NULL) { 961 IP6STAT_INC(ip6s_exthdrtoolong); 962 *mp = NULL; 963 return (-1); 964 } 965 } 966 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 967 hbhlen = (hbh->ip6h_len + 1) << 3; 968 969 if (m->m_len < off + hbhlen) { 970 m = m_pullup(m, off + hbhlen); 971 if (m == NULL) { 972 IP6STAT_INC(ip6s_exthdrtoolong); 973 *mp = NULL; 974 return (-1); 975 } 976 } 977 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 978 off += hbhlen; 979 hbhlen -= sizeof(struct ip6_hbh); 980 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 981 hbhlen, rtalertp, plenp) < 0) { 982 *mp = NULL; 983 return (-1); 984 } 985 986 *offp = off; 987 *mp = m; 988 return (0); 989 } 990 991 /* 992 * Search header for all Hop-by-hop options and process each option. 993 * This function is separate from ip6_hopopts_input() in order to 994 * handle a case where the sending node itself process its hop-by-hop 995 * options header. In such a case, the function is called from ip6_output(). 996 * 997 * The function assumes that hbh header is located right after the IPv6 header 998 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 999 * opthead + hbhlen is located in contiguous memory region. 1000 */ 1001 int 1002 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1003 u_int32_t *rtalertp, u_int32_t *plenp) 1004 { 1005 struct ip6_hdr *ip6; 1006 int optlen = 0; 1007 u_int8_t *opt = opthead; 1008 u_int16_t rtalert_val; 1009 u_int32_t jumboplen; 1010 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1011 1012 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1013 switch (*opt) { 1014 case IP6OPT_PAD1: 1015 optlen = 1; 1016 break; 1017 case IP6OPT_PADN: 1018 if (hbhlen < IP6OPT_MINLEN) { 1019 IP6STAT_INC(ip6s_toosmall); 1020 goto bad; 1021 } 1022 optlen = *(opt + 1) + 2; 1023 break; 1024 case IP6OPT_ROUTER_ALERT: 1025 /* XXX may need check for alignment */ 1026 if (hbhlen < IP6OPT_RTALERT_LEN) { 1027 IP6STAT_INC(ip6s_toosmall); 1028 goto bad; 1029 } 1030 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1031 /* XXX stat */ 1032 icmp6_error(m, ICMP6_PARAM_PROB, 1033 ICMP6_PARAMPROB_HEADER, 1034 erroff + opt + 1 - opthead); 1035 return (-1); 1036 } 1037 optlen = IP6OPT_RTALERT_LEN; 1038 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1039 *rtalertp = ntohs(rtalert_val); 1040 break; 1041 case IP6OPT_JUMBO: 1042 /* XXX may need check for alignment */ 1043 if (hbhlen < IP6OPT_JUMBO_LEN) { 1044 IP6STAT_INC(ip6s_toosmall); 1045 goto bad; 1046 } 1047 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1048 /* XXX stat */ 1049 icmp6_error(m, ICMP6_PARAM_PROB, 1050 ICMP6_PARAMPROB_HEADER, 1051 erroff + opt + 1 - opthead); 1052 return (-1); 1053 } 1054 optlen = IP6OPT_JUMBO_LEN; 1055 1056 /* 1057 * IPv6 packets that have non 0 payload length 1058 * must not contain a jumbo payload option. 1059 */ 1060 ip6 = mtod(m, struct ip6_hdr *); 1061 if (ip6->ip6_plen) { 1062 IP6STAT_INC(ip6s_badoptions); 1063 icmp6_error(m, ICMP6_PARAM_PROB, 1064 ICMP6_PARAMPROB_HEADER, 1065 erroff + opt - opthead); 1066 return (-1); 1067 } 1068 1069 /* 1070 * We may see jumbolen in unaligned location, so 1071 * we'd need to perform bcopy(). 1072 */ 1073 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1074 jumboplen = (u_int32_t)htonl(jumboplen); 1075 1076 #if 1 1077 /* 1078 * if there are multiple jumbo payload options, 1079 * *plenp will be non-zero and the packet will be 1080 * rejected. 1081 * the behavior may need some debate in ipngwg - 1082 * multiple options does not make sense, however, 1083 * there's no explicit mention in specification. 1084 */ 1085 if (*plenp != 0) { 1086 IP6STAT_INC(ip6s_badoptions); 1087 icmp6_error(m, ICMP6_PARAM_PROB, 1088 ICMP6_PARAMPROB_HEADER, 1089 erroff + opt + 2 - opthead); 1090 return (-1); 1091 } 1092 #endif 1093 1094 /* 1095 * jumbo payload length must be larger than 65535. 1096 */ 1097 if (jumboplen <= IPV6_MAXPACKET) { 1098 IP6STAT_INC(ip6s_badoptions); 1099 icmp6_error(m, ICMP6_PARAM_PROB, 1100 ICMP6_PARAMPROB_HEADER, 1101 erroff + opt + 2 - opthead); 1102 return (-1); 1103 } 1104 *plenp = jumboplen; 1105 1106 break; 1107 default: /* unknown option */ 1108 if (hbhlen < IP6OPT_MINLEN) { 1109 IP6STAT_INC(ip6s_toosmall); 1110 goto bad; 1111 } 1112 optlen = ip6_unknown_opt(opt, m, 1113 erroff + opt - opthead); 1114 if (optlen == -1) 1115 return (-1); 1116 optlen += 2; 1117 break; 1118 } 1119 } 1120 1121 return (0); 1122 1123 bad: 1124 m_freem(m); 1125 return (-1); 1126 } 1127 1128 /* 1129 * Unknown option processing. 1130 * The third argument `off' is the offset from the IPv6 header to the option, 1131 * which is necessary if the IPv6 header the and option header and IPv6 header 1132 * is not contiguous in order to return an ICMPv6 error. 1133 */ 1134 int 1135 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1136 { 1137 struct ip6_hdr *ip6; 1138 1139 switch (IP6OPT_TYPE(*optp)) { 1140 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1141 return ((int)*(optp + 1)); 1142 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1143 m_freem(m); 1144 return (-1); 1145 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1146 IP6STAT_INC(ip6s_badoptions); 1147 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1148 return (-1); 1149 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1150 IP6STAT_INC(ip6s_badoptions); 1151 ip6 = mtod(m, struct ip6_hdr *); 1152 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1153 (m->m_flags & (M_BCAST|M_MCAST))) 1154 m_freem(m); 1155 else 1156 icmp6_error(m, ICMP6_PARAM_PROB, 1157 ICMP6_PARAMPROB_OPTION, off); 1158 return (-1); 1159 } 1160 1161 m_freem(m); /* XXX: NOTREACHED */ 1162 return (-1); 1163 } 1164 1165 /* 1166 * Create the "control" list for this pcb. 1167 * These functions will not modify mbuf chain at all. 1168 * 1169 * The routine will be called from upper layer handlers like tcp6_input(). 1170 * Thus the routine assumes that the caller (tcp6_input) have already 1171 * called m_pullup() and all the extension headers are located in the 1172 * very first mbuf on the mbuf chain. 1173 * 1174 * ip6_savecontrol_v4 will handle those options that are possible to be 1175 * set on a v4-mapped socket. 1176 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1177 * options and handle the v6-only ones itself. 1178 */ 1179 struct mbuf ** 1180 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1181 int *v4only) 1182 { 1183 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1184 1185 #ifdef SO_TIMESTAMP 1186 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1187 union { 1188 struct timeval tv; 1189 struct bintime bt; 1190 struct timespec ts; 1191 } t; 1192 struct bintime boottimebin, bt1; 1193 struct timespec ts1; 1194 bool stamped; 1195 1196 stamped = false; 1197 switch (inp->inp_socket->so_ts_clock) { 1198 case SO_TS_REALTIME_MICRO: 1199 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1200 M_TSTMP)) { 1201 mbuf_tstmp2timespec(m, &ts1); 1202 timespec2bintime(&ts1, &bt1); 1203 getboottimebin(&boottimebin); 1204 bintime_add(&bt1, &boottimebin); 1205 bintime2timeval(&bt1, &t.tv); 1206 } else { 1207 microtime(&t.tv); 1208 } 1209 *mp = sbcreatecontrol(&t.tv, sizeof(t.tv), 1210 SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT); 1211 if (*mp != NULL) { 1212 mp = &(*mp)->m_next; 1213 stamped = true; 1214 } 1215 break; 1216 1217 case SO_TS_BINTIME: 1218 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1219 M_TSTMP)) { 1220 mbuf_tstmp2timespec(m, &ts1); 1221 timespec2bintime(&ts1, &t.bt); 1222 getboottimebin(&boottimebin); 1223 bintime_add(&t.bt, &boottimebin); 1224 } else { 1225 bintime(&t.bt); 1226 } 1227 *mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME, 1228 SOL_SOCKET, M_NOWAIT); 1229 if (*mp != NULL) { 1230 mp = &(*mp)->m_next; 1231 stamped = true; 1232 } 1233 break; 1234 1235 case SO_TS_REALTIME: 1236 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1237 M_TSTMP)) { 1238 mbuf_tstmp2timespec(m, &t.ts); 1239 getboottimebin(&boottimebin); 1240 bintime2timespec(&boottimebin, &ts1); 1241 timespecadd(&t.ts, &ts1, &t.ts); 1242 } else { 1243 nanotime(&t.ts); 1244 } 1245 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1246 SCM_REALTIME, SOL_SOCKET, M_NOWAIT); 1247 if (*mp != NULL) { 1248 mp = &(*mp)->m_next; 1249 stamped = true; 1250 } 1251 break; 1252 1253 case SO_TS_MONOTONIC: 1254 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1255 M_TSTMP)) 1256 mbuf_tstmp2timespec(m, &t.ts); 1257 else 1258 nanouptime(&t.ts); 1259 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1260 SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT); 1261 if (*mp != NULL) { 1262 mp = &(*mp)->m_next; 1263 stamped = true; 1264 } 1265 break; 1266 1267 default: 1268 panic("unknown (corrupted) so_ts_clock"); 1269 } 1270 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == 1271 (M_PKTHDR | M_TSTMP)) { 1272 struct sock_timestamp_info sti; 1273 1274 bzero(&sti, sizeof(sti)); 1275 sti.st_info_flags = ST_INFO_HW; 1276 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1277 sti.st_info_flags |= ST_INFO_HW_HPREC; 1278 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO, 1279 SOL_SOCKET, M_NOWAIT); 1280 if (*mp != NULL) 1281 mp = &(*mp)->m_next; 1282 } 1283 } 1284 #endif 1285 1286 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1287 /* RFC 2292 sec. 5 */ 1288 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1289 struct in6_pktinfo pi6; 1290 1291 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1292 #ifdef INET 1293 struct ip *ip; 1294 1295 ip = mtod(m, struct ip *); 1296 pi6.ipi6_addr.s6_addr32[0] = 0; 1297 pi6.ipi6_addr.s6_addr32[1] = 0; 1298 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1299 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1300 #else 1301 /* We won't hit this code */ 1302 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1303 #endif 1304 } else { 1305 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1306 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1307 } 1308 pi6.ipi6_ifindex = 1309 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1310 1311 *mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo), 1312 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6, 1313 M_NOWAIT); 1314 if (*mp) 1315 mp = &(*mp)->m_next; 1316 } 1317 1318 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1319 int hlim; 1320 1321 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1322 #ifdef INET 1323 struct ip *ip; 1324 1325 ip = mtod(m, struct ip *); 1326 hlim = ip->ip_ttl; 1327 #else 1328 /* We won't hit this code */ 1329 hlim = 0; 1330 #endif 1331 } else { 1332 hlim = ip6->ip6_hlim & 0xff; 1333 } 1334 *mp = sbcreatecontrol(&hlim, sizeof(int), 1335 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1336 IPPROTO_IPV6, M_NOWAIT); 1337 if (*mp) 1338 mp = &(*mp)->m_next; 1339 } 1340 1341 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1342 int tclass; 1343 1344 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1345 #ifdef INET 1346 struct ip *ip; 1347 1348 ip = mtod(m, struct ip *); 1349 tclass = ip->ip_tos; 1350 #else 1351 /* We won't hit this code */ 1352 tclass = 0; 1353 #endif 1354 } else { 1355 u_int32_t flowinfo; 1356 1357 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1358 flowinfo >>= 20; 1359 tclass = flowinfo & 0xff; 1360 } 1361 *mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS, 1362 IPPROTO_IPV6, M_NOWAIT); 1363 if (*mp) 1364 mp = &(*mp)->m_next; 1365 } 1366 1367 if (v4only != NULL) { 1368 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1369 *v4only = 1; 1370 } else { 1371 *v4only = 0; 1372 } 1373 } 1374 1375 return (mp); 1376 } 1377 1378 void 1379 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp) 1380 { 1381 struct ip6_hdr *ip6; 1382 int v4only = 0; 1383 1384 mp = ip6_savecontrol_v4(inp, m, mp, &v4only); 1385 if (v4only) 1386 return; 1387 1388 ip6 = mtod(m, struct ip6_hdr *); 1389 /* 1390 * IPV6_HOPOPTS socket option. Recall that we required super-user 1391 * privilege for the option (see ip6_ctloutput), but it might be too 1392 * strict, since there might be some hop-by-hop options which can be 1393 * returned to normal user. 1394 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1395 */ 1396 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) { 1397 /* 1398 * Check if a hop-by-hop options header is contatined in the 1399 * received packet, and if so, store the options as ancillary 1400 * data. Note that a hop-by-hop options header must be 1401 * just after the IPv6 header, which is assured through the 1402 * IPv6 input processing. 1403 */ 1404 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1405 struct ip6_hbh *hbh; 1406 u_int hbhlen; 1407 1408 hbh = (struct ip6_hbh *)(ip6 + 1); 1409 hbhlen = (hbh->ip6h_len + 1) << 3; 1410 1411 /* 1412 * XXX: We copy the whole header even if a 1413 * jumbo payload option is included, the option which 1414 * is to be removed before returning according to 1415 * RFC2292. 1416 * Note: this constraint is removed in RFC3542 1417 */ 1418 *mp = sbcreatecontrol(hbh, hbhlen, 1419 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1420 IPPROTO_IPV6, M_NOWAIT); 1421 if (*mp) 1422 mp = &(*mp)->m_next; 1423 } 1424 } 1425 1426 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1427 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1428 1429 /* 1430 * Search for destination options headers or routing 1431 * header(s) through the header chain, and stores each 1432 * header as ancillary data. 1433 * Note that the order of the headers remains in 1434 * the chain of ancillary data. 1435 */ 1436 while (1) { /* is explicit loop prevention necessary? */ 1437 struct ip6_ext *ip6e = NULL; 1438 u_int elen; 1439 1440 /* 1441 * if it is not an extension header, don't try to 1442 * pull it from the chain. 1443 */ 1444 switch (nxt) { 1445 case IPPROTO_DSTOPTS: 1446 case IPPROTO_ROUTING: 1447 case IPPROTO_HOPOPTS: 1448 case IPPROTO_AH: /* is it possible? */ 1449 break; 1450 default: 1451 goto loopend; 1452 } 1453 1454 if (off + sizeof(*ip6e) > m->m_len) 1455 goto loopend; 1456 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1457 if (nxt == IPPROTO_AH) 1458 elen = (ip6e->ip6e_len + 2) << 2; 1459 else 1460 elen = (ip6e->ip6e_len + 1) << 3; 1461 if (off + elen > m->m_len) 1462 goto loopend; 1463 1464 switch (nxt) { 1465 case IPPROTO_DSTOPTS: 1466 if (!(inp->inp_flags & IN6P_DSTOPTS)) 1467 break; 1468 1469 *mp = sbcreatecontrol(ip6e, elen, 1470 IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1471 IPPROTO_IPV6, M_NOWAIT); 1472 if (*mp) 1473 mp = &(*mp)->m_next; 1474 break; 1475 case IPPROTO_ROUTING: 1476 if (!(inp->inp_flags & IN6P_RTHDR)) 1477 break; 1478 1479 *mp = sbcreatecontrol(ip6e, elen, 1480 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR), 1481 IPPROTO_IPV6, M_NOWAIT); 1482 if (*mp) 1483 mp = &(*mp)->m_next; 1484 break; 1485 case IPPROTO_HOPOPTS: 1486 case IPPROTO_AH: /* is it possible? */ 1487 break; 1488 1489 default: 1490 /* 1491 * other cases have been filtered in the above. 1492 * none will visit this case. here we supply 1493 * the code just in case (nxt overwritten or 1494 * other cases). 1495 */ 1496 goto loopend; 1497 } 1498 1499 /* proceed with the next header. */ 1500 off += elen; 1501 nxt = ip6e->ip6e_nxt; 1502 ip6e = NULL; 1503 } 1504 loopend: 1505 ; 1506 } 1507 1508 if (inp->inp_flags2 & INP_RECVFLOWID) { 1509 uint32_t flowid, flow_type; 1510 1511 flowid = m->m_pkthdr.flowid; 1512 flow_type = M_HASHTYPE_GET(m); 1513 1514 /* 1515 * XXX should handle the failure of one or the 1516 * other - don't populate both? 1517 */ 1518 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID, 1519 IPPROTO_IPV6, M_NOWAIT); 1520 if (*mp) 1521 mp = &(*mp)->m_next; 1522 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t), 1523 IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT); 1524 if (*mp) 1525 mp = &(*mp)->m_next; 1526 } 1527 1528 #ifdef RSS 1529 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1530 uint32_t flowid, flow_type; 1531 uint32_t rss_bucketid; 1532 1533 flowid = m->m_pkthdr.flowid; 1534 flow_type = M_HASHTYPE_GET(m); 1535 1536 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1537 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t), 1538 IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT); 1539 if (*mp) 1540 mp = &(*mp)->m_next; 1541 } 1542 } 1543 #endif 1544 1545 } 1546 #undef IS2292 1547 1548 void 1549 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1550 { 1551 struct socket *so; 1552 struct mbuf *m_mtu; 1553 struct ip6_mtuinfo mtuctl; 1554 1555 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1556 /* 1557 * Notify the error by sending IPV6_PATHMTU ancillary data if 1558 * application wanted to know the MTU value. 1559 * NOTE: we notify disconnected sockets, because some udp 1560 * applications keep sending sockets disconnected. 1561 * NOTE: our implementation doesn't notify connected sockets that has 1562 * foreign address that is different than given destination addresses 1563 * (this is permitted by RFC 3542). 1564 */ 1565 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1566 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1567 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1568 return; 1569 1570 mtuctl.ip6m_mtu = mtu; 1571 mtuctl.ip6m_addr = *dst; 1572 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1573 return; 1574 1575 if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU, 1576 IPPROTO_IPV6, M_NOWAIT)) == NULL) 1577 return; 1578 1579 so = inp->inp_socket; 1580 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1581 == 0) { 1582 soroverflow(so); 1583 m_freem(m_mtu); 1584 /* XXX: should count statistics */ 1585 } else 1586 sorwakeup(so); 1587 } 1588 1589 /* 1590 * Get pointer to the previous header followed by the header 1591 * currently processed. 1592 */ 1593 int 1594 ip6_get_prevhdr(const struct mbuf *m, int off) 1595 { 1596 struct ip6_ext ip6e; 1597 struct ip6_hdr *ip6; 1598 int len, nlen, nxt; 1599 1600 if (off == sizeof(struct ip6_hdr)) 1601 return (offsetof(struct ip6_hdr, ip6_nxt)); 1602 if (off < sizeof(struct ip6_hdr)) 1603 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1604 1605 ip6 = mtod(m, struct ip6_hdr *); 1606 nxt = ip6->ip6_nxt; 1607 len = sizeof(struct ip6_hdr); 1608 nlen = 0; 1609 while (len < off) { 1610 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1611 switch (nxt) { 1612 case IPPROTO_FRAGMENT: 1613 nlen = sizeof(struct ip6_frag); 1614 break; 1615 case IPPROTO_AH: 1616 nlen = (ip6e.ip6e_len + 2) << 2; 1617 break; 1618 default: 1619 nlen = (ip6e.ip6e_len + 1) << 3; 1620 } 1621 len += nlen; 1622 nxt = ip6e.ip6e_nxt; 1623 } 1624 return (len - nlen); 1625 } 1626 1627 /* 1628 * get next header offset. m will be retained. 1629 */ 1630 int 1631 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1632 { 1633 struct ip6_hdr ip6; 1634 struct ip6_ext ip6e; 1635 struct ip6_frag fh; 1636 1637 /* just in case */ 1638 if (m == NULL) 1639 panic("ip6_nexthdr: m == NULL"); 1640 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1641 return -1; 1642 1643 switch (proto) { 1644 case IPPROTO_IPV6: 1645 if (m->m_pkthdr.len < off + sizeof(ip6)) 1646 return -1; 1647 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1648 if (nxtp) 1649 *nxtp = ip6.ip6_nxt; 1650 off += sizeof(ip6); 1651 return off; 1652 1653 case IPPROTO_FRAGMENT: 1654 /* 1655 * terminate parsing if it is not the first fragment, 1656 * it does not make sense to parse through it. 1657 */ 1658 if (m->m_pkthdr.len < off + sizeof(fh)) 1659 return -1; 1660 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1661 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1662 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1663 return -1; 1664 if (nxtp) 1665 *nxtp = fh.ip6f_nxt; 1666 off += sizeof(struct ip6_frag); 1667 return off; 1668 1669 case IPPROTO_AH: 1670 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1671 return -1; 1672 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1673 if (nxtp) 1674 *nxtp = ip6e.ip6e_nxt; 1675 off += (ip6e.ip6e_len + 2) << 2; 1676 return off; 1677 1678 case IPPROTO_HOPOPTS: 1679 case IPPROTO_ROUTING: 1680 case IPPROTO_DSTOPTS: 1681 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1682 return -1; 1683 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1684 if (nxtp) 1685 *nxtp = ip6e.ip6e_nxt; 1686 off += (ip6e.ip6e_len + 1) << 3; 1687 return off; 1688 1689 case IPPROTO_NONE: 1690 case IPPROTO_ESP: 1691 case IPPROTO_IPCOMP: 1692 /* give up */ 1693 return -1; 1694 1695 default: 1696 return -1; 1697 } 1698 1699 /* NOTREACHED */ 1700 } 1701 1702 /* 1703 * get offset for the last header in the chain. m will be kept untainted. 1704 */ 1705 int 1706 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1707 { 1708 int newoff; 1709 int nxt; 1710 1711 if (!nxtp) { 1712 nxt = -1; 1713 nxtp = &nxt; 1714 } 1715 while (1) { 1716 newoff = ip6_nexthdr(m, off, proto, nxtp); 1717 if (newoff < 0) 1718 return off; 1719 else if (newoff < off) 1720 return -1; /* invalid */ 1721 else if (newoff == off) 1722 return newoff; 1723 1724 off = newoff; 1725 proto = *nxtp; 1726 } 1727 } 1728