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 63 #include <sys/cdefs.h> 64 #include "opt_inet.h" 65 #include "opt_inet6.h" 66 #include "opt_ipsec.h" 67 #include "opt_route.h" 68 #include "opt_rss.h" 69 #include "opt_sctp.h" 70 71 #include <sys/param.h> 72 #include <sys/systm.h> 73 #include <sys/hhook.h> 74 #include <sys/malloc.h> 75 #include <sys/mbuf.h> 76 #include <sys/proc.h> 77 #include <sys/domain.h> 78 #include <sys/protosw.h> 79 #include <sys/sdt.h> 80 #include <sys/socket.h> 81 #include <sys/socketvar.h> 82 #include <sys/errno.h> 83 #include <sys/time.h> 84 #include <sys/kernel.h> 85 #include <sys/lock.h> 86 #include <sys/rmlock.h> 87 #include <sys/syslog.h> 88 #include <sys/sysctl.h> 89 #include <sys/eventhandler.h> 90 91 #include <net/if.h> 92 #include <net/if_var.h> 93 #include <net/if_types.h> 94 #include <net/if_private.h> 95 #include <net/if_dl.h> 96 #include <net/route.h> 97 #include <net/netisr.h> 98 #include <net/rss_config.h> 99 #include <net/pfil.h> 100 #include <net/vnet.h> 101 102 #include <netinet/in.h> 103 #include <netinet/in_kdtrace.h> 104 #include <netinet/ip_var.h> 105 #include <netinet/in_systm.h> 106 #include <net/if_llatbl.h> 107 #ifdef INET 108 #include <netinet/ip.h> 109 #include <netinet/ip_icmp.h> 110 #endif /* INET */ 111 #include <netinet/ip6.h> 112 #include <netinet6/in6_var.h> 113 #include <netinet6/ip6_var.h> 114 #include <netinet/ip_encap.h> 115 #include <netinet/in_pcb.h> 116 #include <netinet/icmp6.h> 117 #include <netinet6/scope6_var.h> 118 #include <netinet6/in6_ifattach.h> 119 #include <netinet6/mld6_var.h> 120 #include <netinet6/nd6.h> 121 #include <netinet6/in6_rss.h> 122 #ifdef SCTP 123 #include <netinet/sctp_pcb.h> 124 #include <netinet6/sctp6_var.h> 125 #endif 126 127 #include <netipsec/ipsec_support.h> 128 129 ip6proto_input_t *ip6_protox[IPPROTO_MAX] = { 130 [0 ... IPPROTO_MAX - 1] = rip6_input }; 131 ip6proto_ctlinput_t *ip6_ctlprotox[IPPROTO_MAX] = { 132 [0 ... IPPROTO_MAX - 1] = rip6_ctlinput }; 133 134 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 135 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); 136 VNET_DEFINE(u_long, in6_ifaddrhmask); 137 138 static struct netisr_handler ip6_nh = { 139 .nh_name = "ip6", 140 .nh_handler = ip6_input, 141 .nh_proto = NETISR_IPV6, 142 #ifdef RSS 143 .nh_m2cpuid = rss_soft_m2cpuid_v6, 144 .nh_policy = NETISR_POLICY_CPU, 145 .nh_dispatch = NETISR_DISPATCH_HYBRID, 146 #else 147 .nh_policy = NETISR_POLICY_FLOW, 148 #endif 149 }; 150 151 static int 152 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 153 { 154 int error, qlimit; 155 156 netisr_getqlimit(&ip6_nh, &qlimit); 157 error = sysctl_handle_int(oidp, &qlimit, 0, req); 158 if (error || !req->newptr) 159 return (error); 160 if (qlimit < 1) 161 return (EINVAL); 162 return (netisr_setqlimit(&ip6_nh, qlimit)); 163 } 164 SYSCTL_DECL(_net_inet6_ip6); 165 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen, 166 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 167 0, 0, sysctl_netinet6_intr_queue_maxlen, "I", 168 "Maximum size of the IPv6 input queue"); 169 170 VNET_DEFINE_STATIC(bool, ip6_sav) = true; 171 #define V_ip6_sav VNET(ip6_sav) 172 SYSCTL_BOOL(_net_inet6_ip6, OID_AUTO, source_address_validation, 173 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_sav), true, 174 "Drop incoming packets with source address that is a local address"); 175 176 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, temp_max_desync_factor, 177 CTLFLAG_RD | CTLFLAG_VNET, 178 &VNET_NAME(ip6_temp_max_desync_factor), 0, 179 "RFC 8981 max desync factor"); 180 181 #ifdef RSS 182 static struct netisr_handler ip6_direct_nh = { 183 .nh_name = "ip6_direct", 184 .nh_handler = ip6_direct_input, 185 .nh_proto = NETISR_IPV6_DIRECT, 186 .nh_m2cpuid = rss_soft_m2cpuid_v6, 187 .nh_policy = NETISR_POLICY_CPU, 188 .nh_dispatch = NETISR_DISPATCH_HYBRID, 189 }; 190 191 static int 192 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 193 { 194 int error, qlimit; 195 196 netisr_getqlimit(&ip6_direct_nh, &qlimit); 197 error = sysctl_handle_int(oidp, &qlimit, 0, req); 198 if (error || !req->newptr) 199 return (error); 200 if (qlimit < 1) 201 return (EINVAL); 202 return (netisr_setqlimit(&ip6_direct_nh, qlimit)); 203 } 204 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 205 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 206 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I", 207 "Maximum size of the IPv6 direct input queue"); 208 209 #endif 210 211 VNET_DEFINE(pfil_head_t, inet6_pfil_head); 212 VNET_DEFINE(pfil_head_t, inet6_local_pfil_head); 213 214 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 215 VNET_PCPUSTAT_SYSINIT(ip6stat); 216 #ifdef VIMAGE 217 VNET_PCPUSTAT_SYSUNINIT(ip6stat); 218 #endif /* VIMAGE */ 219 220 struct rmlock in6_ifaddr_lock; 221 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 222 223 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 224 225 /* 226 * IP6 initialization: fill in IP6 protocol switch table. 227 * All protocols not implemented in kernel go to raw IP6 protocol handler. 228 */ 229 static void 230 ip6_vnet_init(void *arg __unused) 231 { 232 struct pfil_head_args args; 233 234 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 235 &V_ip6_auto_linklocal); 236 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 237 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 238 TUNABLE_BOOL_FETCH("net.inet6.ip6.use_stableaddr", &V_ip6_use_stableaddr); 239 240 CK_STAILQ_INIT(&V_in6_ifaddrhead); 241 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 242 &V_in6_ifaddrhmask); 243 244 /* Initialize packet filter hooks. */ 245 args.pa_version = PFIL_VERSION; 246 args.pa_flags = PFIL_IN | PFIL_OUT; 247 args.pa_type = PFIL_TYPE_IP6; 248 args.pa_headname = PFIL_INET6_NAME; 249 V_inet6_pfil_head = pfil_head_register(&args); 250 251 args.pa_flags = PFIL_OUT; 252 args.pa_headname = PFIL_INET6_LOCAL_NAME; 253 V_inet6_local_pfil_head = pfil_head_register(&args); 254 255 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 256 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 257 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 258 printf("%s: WARNING: unable to register input helper hook\n", 259 __func__); 260 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 261 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 262 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 263 printf("%s: WARNING: unable to register output helper hook\n", 264 __func__); 265 266 scope6_init(); 267 addrsel_policy_init(); 268 nd6_init(); 269 frag6_init(); 270 271 V_ip6_temp_max_desync_factor = TEMP_MAX_DESYNC_FACTOR_BASE + 272 (V_ip6_temp_preferred_lifetime >> 2) + 273 (V_ip6_temp_preferred_lifetime >> 3); 274 V_ip6_desync_factor = arc4random() % V_ip6_temp_max_desync_factor; 275 276 /* Skip global initialization stuff for non-default instances. */ 277 #ifdef VIMAGE 278 netisr_register_vnet(&ip6_nh); 279 #ifdef RSS 280 netisr_register_vnet(&ip6_direct_nh); 281 #endif 282 #endif 283 } 284 VNET_SYSINIT(ip6_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, 285 ip6_vnet_init, NULL); 286 287 static void 288 ip6_init(void *arg __unused) 289 { 290 291 /* 292 * Register statically those protocols that are unlikely to ever go 293 * dynamic. 294 */ 295 IP6PROTO_REGISTER(IPPROTO_ICMPV6, icmp6_input, rip6_ctlinput); 296 IP6PROTO_REGISTER(IPPROTO_DSTOPTS, dest6_input, NULL); 297 IP6PROTO_REGISTER(IPPROTO_ROUTING, route6_input, NULL); 298 IP6PROTO_REGISTER(IPPROTO_FRAGMENT, frag6_input, NULL); 299 IP6PROTO_REGISTER(IPPROTO_IPV4, encap6_input, NULL); 300 IP6PROTO_REGISTER(IPPROTO_IPV6, encap6_input, NULL); 301 IP6PROTO_REGISTER(IPPROTO_ETHERIP, encap6_input, NULL); 302 IP6PROTO_REGISTER(IPPROTO_GRE, encap6_input, NULL); 303 IP6PROTO_REGISTER(IPPROTO_PIM, encap6_input, NULL); 304 #ifdef SCTP /* XXX: has a loadable & static version */ 305 IP6PROTO_REGISTER(IPPROTO_SCTP, sctp6_input, sctp6_ctlinput); 306 #endif 307 308 EVENTHANDLER_REGISTER(vm_lowmem, frag6_drain, NULL, LOWMEM_PRI_DEFAULT); 309 EVENTHANDLER_REGISTER(mbuf_lowmem, frag6_drain, NULL, 310 LOWMEM_PRI_DEFAULT); 311 312 netisr_register(&ip6_nh); 313 #ifdef RSS 314 netisr_register(&ip6_direct_nh); 315 #endif 316 } 317 SYSINIT(ip6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init, NULL); 318 319 int 320 ip6proto_register(uint8_t proto, ip6proto_input_t input, 321 ip6proto_ctlinput_t ctl) 322 { 323 324 MPASS(proto > 0); 325 326 if (ip6_protox[proto] == rip6_input) { 327 ip6_protox[proto] = input; 328 ip6_ctlprotox[proto] = ctl; 329 return (0); 330 } else 331 return (EEXIST); 332 } 333 334 int 335 ip6proto_unregister(uint8_t proto) 336 { 337 338 MPASS(proto > 0); 339 340 if (ip6_protox[proto] != rip6_input) { 341 ip6_protox[proto] = rip6_input; 342 ip6_ctlprotox[proto] = rip6_ctlinput; 343 return (0); 344 } else 345 return (ENOENT); 346 } 347 348 #ifdef VIMAGE 349 static void 350 ip6_destroy(void *unused __unused) 351 { 352 struct ifaddr *ifa, *nifa; 353 struct ifnet *ifp; 354 int error; 355 356 #ifdef RSS 357 netisr_unregister_vnet(&ip6_direct_nh); 358 #endif 359 netisr_unregister_vnet(&ip6_nh); 360 361 pfil_head_unregister(V_inet6_pfil_head); 362 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 363 if (error != 0) { 364 printf("%s: WARNING: unable to deregister input helper hook " 365 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 366 "error %d returned\n", __func__, error); 367 } 368 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 369 if (error != 0) { 370 printf("%s: WARNING: unable to deregister output helper hook " 371 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 372 "error %d returned\n", __func__, error); 373 } 374 375 /* Cleanup addresses. */ 376 IFNET_RLOCK(); 377 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 378 /* Cannot lock here - lock recursion. */ 379 /* IF_ADDR_LOCK(ifp); */ 380 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 381 if (ifa->ifa_addr->sa_family != AF_INET6) 382 continue; 383 in6_purgeaddr(ifa); 384 } 385 /* IF_ADDR_UNLOCK(ifp); */ 386 in6_ifdetach_destroy(ifp); 387 mld_domifdetach(ifp); 388 } 389 IFNET_RUNLOCK(); 390 391 /* Make sure any routes are gone as well. */ 392 rib_flush_routes_family(AF_INET6); 393 394 frag6_destroy(); 395 nd6_destroy(); 396 in6_ifattach_destroy(); 397 398 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 399 } 400 401 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 402 #endif 403 404 static int 405 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off, 406 int *nxt, int *ours) 407 { 408 struct mbuf *m; 409 struct ip6_hdr *ip6; 410 struct ip6_hbh *hbh; 411 412 if (ip6_hopopts_input(plen, rtalert, mp, off)) { 413 #if 0 /*touches NULL pointer*/ 414 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard); 415 #endif 416 goto out; /* m have already been freed */ 417 } 418 419 /* adjust pointer */ 420 m = *mp; 421 ip6 = mtod(m, struct ip6_hdr *); 422 423 /* 424 * if the payload length field is 0 and the next header field 425 * indicates Hop-by-Hop Options header, then a Jumbo Payload 426 * option MUST be included. 427 */ 428 if (ip6->ip6_plen == 0 && *plen == 0) { 429 /* 430 * Note that if a valid jumbo payload option is 431 * contained, ip6_hopopts_input() must set a valid 432 * (non-zero) payload length to the variable plen. 433 */ 434 IP6STAT_INC(ip6s_badoptions); 435 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 436 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 437 icmp6_error(m, ICMP6_PARAM_PROB, 438 ICMP6_PARAMPROB_HEADER, 439 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 440 goto out; 441 } 442 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 443 hbh = (struct ip6_hbh *)(ip6 + 1); 444 *nxt = hbh->ip6h_nxt; 445 446 /* 447 * If we are acting as a router and the packet contains a 448 * router alert option, see if we know the option value. 449 * Currently, we only support the option value for MLD, in which 450 * case we should pass the packet to the multicast routing 451 * daemon. 452 */ 453 if (*rtalert != ~0) { 454 switch (*rtalert) { 455 case IP6OPT_RTALERT_MLD: 456 if (V_ip6_forwarding) 457 *ours = 1; 458 break; 459 default: 460 /* 461 * RFC2711 requires unrecognized values must be 462 * silently ignored. 463 */ 464 break; 465 } 466 } 467 468 return (0); 469 470 out: 471 return (1); 472 } 473 474 #ifdef RSS 475 /* 476 * IPv6 direct input routine. 477 * 478 * This is called when reinjecting completed fragments where 479 * all of the previous checking and book-keeping has been done. 480 */ 481 void 482 ip6_direct_input(struct mbuf *m) 483 { 484 int off, nxt; 485 int nest; 486 struct m_tag *mtag; 487 struct ip6_direct_ctx *ip6dc; 488 489 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 490 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 491 492 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 493 nxt = ip6dc->ip6dc_nxt; 494 off = ip6dc->ip6dc_off; 495 496 nest = 0; 497 498 m_tag_delete(m, mtag); 499 500 while (nxt != IPPROTO_DONE) { 501 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 502 IP6STAT_INC(ip6s_toomanyhdr); 503 goto bad; 504 } 505 506 /* 507 * protection against faulty packet - there should be 508 * more sanity checks in header chain processing. 509 */ 510 if (m->m_pkthdr.len < off) { 511 IP6STAT_INC(ip6s_tooshort); 512 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 513 goto bad; 514 } 515 516 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 517 if (IPSEC_ENABLED(ipv6)) { 518 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 519 return; 520 } 521 #endif /* IPSEC */ 522 523 nxt = ip6_protox[nxt](&m, &off, nxt); 524 } 525 return; 526 bad: 527 m_freem(m); 528 } 529 #endif 530 531 void 532 ip6_input(struct mbuf *m) 533 { 534 struct in6_addr odst; 535 struct ip6_hdr *ip6; 536 struct in6_ifaddr *ia; 537 struct ifnet *rcvif; 538 u_int32_t plen; 539 u_int32_t rtalert = ~0; 540 int off = sizeof(struct ip6_hdr), nest; 541 int nxt, ours = 0; 542 int srcrt = 0; 543 544 /* 545 * Drop the packet if IPv6 operation is disabled on the interface. 546 */ 547 rcvif = m->m_pkthdr.rcvif; 548 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 549 goto bad; 550 551 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 552 /* 553 * should the inner packet be considered authentic? 554 * see comment in ah4_input(). 555 * NB: m cannot be NULL when passed to the input routine 556 */ 557 558 m->m_flags &= ~M_AUTHIPHDR; 559 m->m_flags &= ~M_AUTHIPDGM; 560 561 #endif /* IPSEC */ 562 563 if (m->m_flags & M_FASTFWD_OURS) { 564 /* 565 * Firewall changed destination to local. 566 */ 567 ip6 = mtod(m, struct ip6_hdr *); 568 goto passin; 569 } 570 571 /* 572 * mbuf statistics 573 */ 574 if (m->m_flags & M_EXT) { 575 if (m->m_next) 576 IP6STAT_INC(ip6s_mext2m); 577 else 578 IP6STAT_INC(ip6s_mext1); 579 } else { 580 if (m->m_next) { 581 struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif; 582 int ifindex = ifp->if_index; 583 if (ifindex >= IP6S_M2MMAX) 584 ifindex = 0; 585 IP6STAT_INC2(ip6s_m2m, ifindex); 586 } else 587 IP6STAT_INC(ip6s_m1); 588 } 589 590 in6_ifstat_inc(rcvif, ifs6_in_receive); 591 IP6STAT_INC(ip6s_total); 592 593 /* 594 * L2 bridge code and some other code can return mbuf chain 595 * that does not conform to KAME requirement. too bad. 596 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 597 */ 598 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 599 struct mbuf *n; 600 601 if (m->m_pkthdr.len > MHLEN) 602 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 603 else 604 n = m_gethdr(M_NOWAIT, MT_DATA); 605 if (n == NULL) 606 goto bad; 607 608 m_move_pkthdr(n, m); 609 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 610 n->m_len = n->m_pkthdr.len; 611 m_freem(m); 612 m = n; 613 } 614 if (m->m_len < sizeof(struct ip6_hdr)) { 615 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 616 IP6STAT_INC(ip6s_toosmall); 617 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 618 goto bad; 619 } 620 } 621 622 ip6 = mtod(m, struct ip6_hdr *); 623 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 624 IP6STAT_INC(ip6s_badvers); 625 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 626 goto bad; 627 } 628 629 IP6STAT_INC2(ip6s_nxthist, ip6->ip6_nxt); 630 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 631 632 /* 633 * Check against address spoofing/corruption. The unspecified address 634 * is checked further below. 635 */ 636 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) { 637 /* 638 * XXX: "badscope" is not very suitable for a multicast source. 639 */ 640 IP6STAT_INC(ip6s_badscope); 641 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 642 goto bad; 643 } 644 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 645 !(m->m_flags & M_LOOP)) { 646 /* 647 * In this case, the packet should come from the loopback 648 * interface. However, we cannot just check the if_flags, 649 * because ip6_mloopback() passes the "actual" interface 650 * as the outgoing/incoming interface. 651 */ 652 IP6STAT_INC(ip6s_badscope); 653 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 654 goto bad; 655 } 656 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 657 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 658 /* 659 * RFC4291 2.7: 660 * Nodes must not originate a packet to a multicast address 661 * whose scop field contains the reserved value 0; if such 662 * a packet is received, it must be silently dropped. 663 */ 664 IP6STAT_INC(ip6s_badscope); 665 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 666 goto bad; 667 } 668 /* 669 * The following check is not documented in specs. A malicious 670 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 671 * and bypass security checks (act as if it was from 127.0.0.1 by using 672 * IPv6 src ::ffff:127.0.0.1). Be cautious. 673 * 674 * We have supported IPv6-only kernels for a few years and this issue 675 * has not come up. The world seems to move mostly towards not using 676 * v4mapped on the wire, so it makes sense for us to keep rejecting 677 * any such packets. 678 */ 679 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 680 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 681 IP6STAT_INC(ip6s_badscope); 682 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 683 goto bad; 684 } 685 #if 0 686 /* 687 * Reject packets with IPv4 compatible addresses (auto tunnel). 688 * 689 * The code forbids auto tunnel relay case in RFC1933 (the check is 690 * stronger than RFC1933). We may want to re-enable it if mech-xx 691 * is revised to forbid relaying case. 692 */ 693 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 694 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 695 IP6STAT_INC(ip6s_badscope); 696 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 697 goto bad; 698 } 699 #endif 700 /* 701 * Try to forward the packet, but if we fail continue. 702 * ip6_tryforward() does not generate redirects, so fall 703 * through to normal processing if redirects are required. 704 * ip6_tryforward() does inbound and outbound packet firewall 705 * processing. If firewall has decided that destination becomes 706 * our local address, it sets M_FASTFWD_OURS flag. In this 707 * case skip another inbound firewall processing and update 708 * ip6 pointer. 709 */ 710 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0 711 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 712 && (!IPSEC_ENABLED(ipv6) || 713 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0) 714 #endif 715 ) { 716 if ((m = ip6_tryforward(m)) == NULL) 717 return; 718 if (m->m_flags & M_FASTFWD_OURS) { 719 ip6 = mtod(m, struct ip6_hdr *); 720 goto passin; 721 } 722 } 723 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 724 /* 725 * Bypass packet filtering for packets previously handled by IPsec. 726 */ 727 if (IPSEC_ENABLED(ipv6) && 728 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0) 729 goto passin; 730 #endif 731 /* 732 * Run through list of hooks for input packets. 733 * 734 * NB: Beware of the destination address changing 735 * (e.g. by NAT rewriting). When this happens, 736 * tell ip6_forward to do the right thing. 737 */ 738 739 /* Jump over all PFIL processing if hooks are not active. */ 740 if (!PFIL_HOOKED_IN(V_inet6_pfil_head)) 741 goto passin; 742 743 odst = ip6->ip6_dst; 744 if (pfil_mbuf_in(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif, 745 NULL) != PFIL_PASS) 746 return; 747 ip6 = mtod(m, struct ip6_hdr *); 748 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 749 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 750 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 751 /* 752 * Directly ship the packet on. This allows forwarding 753 * packets originally destined to us to some other directly 754 * connected host. 755 */ 756 ip6_forward(m, 1); 757 return; 758 } 759 760 passin: 761 /* 762 * The check is deferred to here to give firewalls a chance to block 763 * (and log) such packets. ip6_tryforward() will not process such 764 * packets. 765 */ 766 if (__predict_false(IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst))) { 767 IP6STAT_INC(ip6s_badscope); 768 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 769 goto bad; 770 } 771 772 /* 773 * Disambiguate address scope zones (if there is ambiguity). 774 * We first make sure that the original source or destination address 775 * is not in our internal form for scoped addresses. Such addresses 776 * are not necessarily invalid spec-wise, but we cannot accept them due 777 * to the usage conflict. 778 * in6_setscope() then also checks and rejects the cases where src or 779 * dst are the loopback address and the receiving interface 780 * is not loopback. 781 */ 782 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 783 IP6STAT_INC(ip6s_badscope); /* XXX */ 784 goto bad; 785 } 786 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 787 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 788 IP6STAT_INC(ip6s_badscope); 789 goto bad; 790 } 791 if (m->m_flags & M_FASTFWD_OURS) { 792 m->m_flags &= ~M_FASTFWD_OURS; 793 ours = 1; 794 goto hbhcheck; 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(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 */, false); 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 goto bad; 821 } 822 if (V_ip6_sav && !(m->m_flags & M_LOOP) && 823 __predict_false(in6_localip_fib(&ip6->ip6_src, 824 rcvif->if_fib))) { 825 IP6STAT_INC(ip6s_badscope); /* XXX */ 826 goto bad; 827 } 828 /* Count the packet in the ip address stats */ 829 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 830 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 831 ours = 1; 832 goto hbhcheck; 833 } 834 835 /* 836 * Now there is no reason to process the packet if it's not our own 837 * and we're not a router. 838 */ 839 if (!V_ip6_forwarding) { 840 IP6STAT_INC(ip6s_cantforward); 841 goto bad; 842 } 843 844 hbhcheck: 845 /* 846 * Process Hop-by-Hop options header if it's contained. 847 * m may be modified in ip6_hopopts_input(). 848 * If a JumboPayload option is included, plen will also be modified. 849 */ 850 plen = (u_int32_t)ntohs(ip6->ip6_plen); 851 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 852 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0) 853 return; 854 } else 855 nxt = ip6->ip6_nxt; 856 857 /* 858 * Use mbuf flags to propagate Router Alert option to 859 * ICMPv6 layer, as hop-by-hop options have been stripped. 860 */ 861 if (rtalert != ~0) 862 m->m_flags |= M_RTALERT_MLD; 863 864 /* 865 * Check that the amount of data in the buffers 866 * is as at least much as the IPv6 header would have us expect. 867 * Trim mbufs if longer than we expect. 868 * Drop packet if shorter than we expect. 869 */ 870 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 871 IP6STAT_INC(ip6s_tooshort); 872 in6_ifstat_inc(rcvif, ifs6_in_truncated); 873 goto bad; 874 } 875 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 876 if (m->m_len == m->m_pkthdr.len) { 877 m->m_len = sizeof(struct ip6_hdr) + plen; 878 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 879 } else 880 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 881 } 882 883 /* 884 * Forward if desirable. 885 */ 886 if (V_ip6_mrouter && 887 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 888 /* 889 * If we are acting as a multicast router, all 890 * incoming multicast packets are passed to the 891 * kernel-level multicast forwarding function. 892 * The packet is returned (relatively) intact; if 893 * ip6_mforward() returns a non-zero value, the packet 894 * must be discarded, else it may be accepted below. 895 * 896 * XXX TODO: Check hlim and multicast scope here to avoid 897 * unnecessarily calling into ip6_mforward(). 898 */ 899 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 900 IP6STAT_INC(ip6s_cantforward); 901 goto bad; 902 } 903 } else if (!ours) { 904 ip6_forward(m, srcrt); 905 return; 906 } 907 908 /* 909 * We are going to ship the packet to the local protocol stack. Call the 910 * filter again for this 'output' action, allowing redirect-like rules 911 * to adjust the source address. 912 */ 913 if (PFIL_HOOKED_OUT(V_inet6_local_pfil_head)) { 914 if (pfil_mbuf_out(V_inet6_local_pfil_head, &m, V_loif, NULL) != 915 PFIL_PASS) 916 return; 917 ip6 = mtod(m, struct ip6_hdr *); 918 } 919 920 /* 921 * Tell launch routine the next header 922 */ 923 IP6STAT_INC(ip6s_delivered); 924 in6_ifstat_inc(rcvif, ifs6_in_deliver); 925 nest = 0; 926 927 while (nxt != IPPROTO_DONE) { 928 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 929 IP6STAT_INC(ip6s_toomanyhdr); 930 goto bad; 931 } 932 933 /* 934 * protection against faulty packet - there should be 935 * more sanity checks in header chain processing. 936 */ 937 if (m->m_pkthdr.len < off) { 938 IP6STAT_INC(ip6s_tooshort); 939 in6_ifstat_inc(rcvif, ifs6_in_truncated); 940 goto bad; 941 } 942 943 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 944 if (IPSEC_ENABLED(ipv6)) { 945 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 946 return; 947 } 948 #endif /* IPSEC */ 949 950 nxt = ip6_protox[nxt](&m, &off, nxt); 951 } 952 return; 953 bad: 954 in6_ifstat_inc(rcvif, ifs6_in_discard); 955 if (m != NULL) 956 m_freem(m); 957 } 958 959 /* 960 * Hop-by-Hop options header processing. If a valid jumbo payload option is 961 * included, the real payload length will be stored in plenp. 962 * 963 * rtalertp - XXX: should be stored more smart way 964 */ 965 static int 966 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 967 struct mbuf **mp, int *offp) 968 { 969 struct mbuf *m = *mp; 970 int off = *offp, hbhlen; 971 struct ip6_hbh *hbh; 972 973 /* validation of the length of the header */ 974 if (m->m_len < off + sizeof(*hbh)) { 975 m = m_pullup(m, off + sizeof(*hbh)); 976 if (m == NULL) { 977 IP6STAT_INC(ip6s_exthdrtoolong); 978 *mp = NULL; 979 return (-1); 980 } 981 } 982 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 983 hbhlen = (hbh->ip6h_len + 1) << 3; 984 985 if (m->m_len < off + hbhlen) { 986 m = m_pullup(m, off + hbhlen); 987 if (m == NULL) { 988 IP6STAT_INC(ip6s_exthdrtoolong); 989 *mp = NULL; 990 return (-1); 991 } 992 } 993 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 994 off += hbhlen; 995 hbhlen -= sizeof(struct ip6_hbh); 996 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 997 hbhlen, rtalertp, plenp) < 0) { 998 *mp = NULL; 999 return (-1); 1000 } 1001 1002 *offp = off; 1003 *mp = m; 1004 return (0); 1005 } 1006 1007 /* 1008 * Search header for all Hop-by-hop options and process each option. 1009 * This function is separate from ip6_hopopts_input() in order to 1010 * handle a case where the sending node itself process its hop-by-hop 1011 * options header. In such a case, the function is called from ip6_output(). 1012 * 1013 * The function assumes that hbh header is located right after the IPv6 header 1014 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1015 * opthead + hbhlen is located in contiguous memory region. 1016 */ 1017 int 1018 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1019 u_int32_t *rtalertp, u_int32_t *plenp) 1020 { 1021 struct ip6_hdr *ip6; 1022 int optlen = 0; 1023 u_int8_t *opt = opthead; 1024 u_int16_t rtalert_val; 1025 u_int32_t jumboplen; 1026 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1027 1028 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1029 switch (*opt) { 1030 case IP6OPT_PAD1: 1031 optlen = 1; 1032 break; 1033 case IP6OPT_PADN: 1034 if (hbhlen < IP6OPT_MINLEN) { 1035 IP6STAT_INC(ip6s_toosmall); 1036 goto bad; 1037 } 1038 optlen = *(opt + 1) + 2; 1039 break; 1040 case IP6OPT_ROUTER_ALERT: 1041 /* XXX may need check for alignment */ 1042 if (hbhlen < IP6OPT_RTALERT_LEN) { 1043 IP6STAT_INC(ip6s_toosmall); 1044 goto bad; 1045 } 1046 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1047 /* XXX stat */ 1048 icmp6_error(m, ICMP6_PARAM_PROB, 1049 ICMP6_PARAMPROB_HEADER, 1050 erroff + opt + 1 - opthead); 1051 return (-1); 1052 } 1053 optlen = IP6OPT_RTALERT_LEN; 1054 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1055 *rtalertp = ntohs(rtalert_val); 1056 break; 1057 case IP6OPT_JUMBO: 1058 /* XXX may need check for alignment */ 1059 if (hbhlen < IP6OPT_JUMBO_LEN) { 1060 IP6STAT_INC(ip6s_toosmall); 1061 goto bad; 1062 } 1063 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1064 /* XXX stat */ 1065 icmp6_error(m, ICMP6_PARAM_PROB, 1066 ICMP6_PARAMPROB_HEADER, 1067 erroff + opt + 1 - opthead); 1068 return (-1); 1069 } 1070 optlen = IP6OPT_JUMBO_LEN; 1071 1072 /* 1073 * IPv6 packets that have non 0 payload length 1074 * must not contain a jumbo payload option. 1075 */ 1076 ip6 = mtod(m, struct ip6_hdr *); 1077 if (ip6->ip6_plen) { 1078 IP6STAT_INC(ip6s_badoptions); 1079 icmp6_error(m, ICMP6_PARAM_PROB, 1080 ICMP6_PARAMPROB_HEADER, 1081 erroff + opt - opthead); 1082 return (-1); 1083 } 1084 1085 /* 1086 * We may see jumbolen in unaligned location, so 1087 * we'd need to perform bcopy(). 1088 */ 1089 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1090 jumboplen = (u_int32_t)htonl(jumboplen); 1091 1092 #if 1 1093 /* 1094 * if there are multiple jumbo payload options, 1095 * *plenp will be non-zero and the packet will be 1096 * rejected. 1097 * the behavior may need some debate in ipngwg - 1098 * multiple options does not make sense, however, 1099 * there's no explicit mention in specification. 1100 */ 1101 if (*plenp != 0) { 1102 IP6STAT_INC(ip6s_badoptions); 1103 icmp6_error(m, ICMP6_PARAM_PROB, 1104 ICMP6_PARAMPROB_HEADER, 1105 erroff + opt + 2 - opthead); 1106 return (-1); 1107 } 1108 #endif 1109 1110 /* 1111 * jumbo payload length must be larger than 65535. 1112 */ 1113 if (jumboplen <= IPV6_MAXPACKET) { 1114 IP6STAT_INC(ip6s_badoptions); 1115 icmp6_error(m, ICMP6_PARAM_PROB, 1116 ICMP6_PARAMPROB_HEADER, 1117 erroff + opt + 2 - opthead); 1118 return (-1); 1119 } 1120 *plenp = jumboplen; 1121 1122 break; 1123 default: /* unknown option */ 1124 if (hbhlen < IP6OPT_MINLEN) { 1125 IP6STAT_INC(ip6s_toosmall); 1126 goto bad; 1127 } 1128 optlen = ip6_unknown_opt(opt, m, 1129 erroff + opt - opthead); 1130 if (optlen == -1) 1131 return (-1); 1132 optlen += 2; 1133 break; 1134 } 1135 } 1136 1137 return (0); 1138 1139 bad: 1140 m_freem(m); 1141 return (-1); 1142 } 1143 1144 /* 1145 * Unknown option processing. 1146 * The third argument `off' is the offset from the IPv6 header to the option, 1147 * which is necessary if the IPv6 header the and option header and IPv6 header 1148 * is not contiguous in order to return an ICMPv6 error. 1149 */ 1150 int 1151 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1152 { 1153 struct ip6_hdr *ip6; 1154 1155 switch (IP6OPT_TYPE(*optp)) { 1156 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1157 return ((int)*(optp + 1)); 1158 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1159 m_freem(m); 1160 return (-1); 1161 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1162 IP6STAT_INC(ip6s_badoptions); 1163 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1164 return (-1); 1165 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1166 IP6STAT_INC(ip6s_badoptions); 1167 ip6 = mtod(m, struct ip6_hdr *); 1168 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1169 (m->m_flags & (M_BCAST|M_MCAST))) 1170 m_freem(m); 1171 else 1172 icmp6_error(m, ICMP6_PARAM_PROB, 1173 ICMP6_PARAMPROB_OPTION, off); 1174 return (-1); 1175 } 1176 1177 m_freem(m); /* XXX: NOTREACHED */ 1178 return (-1); 1179 } 1180 1181 /* 1182 * Create the "control" list for this pcb. 1183 * These functions will not modify mbuf chain at all. 1184 * 1185 * The routine will be called from upper layer handlers like tcp6_input(). 1186 * Thus the routine assumes that the caller (tcp6_input) have already 1187 * called m_pullup() and all the extension headers are located in the 1188 * very first mbuf on the mbuf chain. 1189 * 1190 * ip6_savecontrol_v4 will handle those options that are possible to be 1191 * set on a v4-mapped socket. 1192 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1193 * options and handle the v6-only ones itself. 1194 */ 1195 struct mbuf ** 1196 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1197 int *v4only) 1198 { 1199 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1200 1201 #if defined(SO_TIMESTAMP) && defined(SO_BINTIME) 1202 if ((inp->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) != 0) { 1203 union { 1204 struct timeval tv; 1205 struct bintime bt; 1206 struct timespec ts; 1207 } t; 1208 struct bintime boottimebin, bt1; 1209 struct timespec ts1; 1210 int ts_clock; 1211 bool stamped; 1212 1213 ts_clock = inp->inp_socket->so_ts_clock; 1214 stamped = false; 1215 1216 /* 1217 * Handle BINTIME first. We create the same output options 1218 * for both SO_BINTIME and the case where SO_TIMESTAMP is 1219 * set with the timestamp clock set to SO_TS_BINTIME. 1220 */ 1221 if ((inp->inp_socket->so_options & SO_BINTIME) != 0 || 1222 ts_clock == SO_TS_BINTIME) { 1223 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1224 M_TSTMP)) { 1225 mbuf_tstmp2timespec(m, &ts1); 1226 timespec2bintime(&ts1, &t.bt); 1227 getboottimebin(&boottimebin); 1228 bintime_add(&t.bt, &boottimebin); 1229 } else { 1230 bintime(&t.bt); 1231 } 1232 *mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME, 1233 SOL_SOCKET, M_NOWAIT); 1234 if (*mp != NULL) { 1235 mp = &(*mp)->m_next; 1236 stamped = true; 1237 } 1238 1239 /* 1240 * Suppress other timestamps if SO_TIMESTAMP is not 1241 * set. 1242 */ 1243 if ((inp->inp_socket->so_options & SO_TIMESTAMP) == 0) 1244 ts_clock = SO_TS_BINTIME; 1245 } 1246 1247 switch (ts_clock) { 1248 case SO_TS_REALTIME_MICRO: 1249 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1250 M_TSTMP)) { 1251 mbuf_tstmp2timespec(m, &ts1); 1252 timespec2bintime(&ts1, &bt1); 1253 getboottimebin(&boottimebin); 1254 bintime_add(&bt1, &boottimebin); 1255 bintime2timeval(&bt1, &t.tv); 1256 } else { 1257 microtime(&t.tv); 1258 } 1259 *mp = sbcreatecontrol(&t.tv, sizeof(t.tv), 1260 SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT); 1261 if (*mp != NULL) { 1262 mp = &(*mp)->m_next; 1263 stamped = true; 1264 } 1265 break; 1266 1267 case SO_TS_BINTIME: 1268 break; 1269 1270 case SO_TS_REALTIME: 1271 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1272 M_TSTMP)) { 1273 mbuf_tstmp2timespec(m, &t.ts); 1274 getboottimebin(&boottimebin); 1275 bintime2timespec(&boottimebin, &ts1); 1276 timespecadd(&t.ts, &ts1, &t.ts); 1277 } else { 1278 nanotime(&t.ts); 1279 } 1280 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1281 SCM_REALTIME, SOL_SOCKET, M_NOWAIT); 1282 if (*mp != NULL) { 1283 mp = &(*mp)->m_next; 1284 stamped = true; 1285 } 1286 break; 1287 1288 case SO_TS_MONOTONIC: 1289 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1290 M_TSTMP)) 1291 mbuf_tstmp2timespec(m, &t.ts); 1292 else 1293 nanouptime(&t.ts); 1294 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1295 SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT); 1296 if (*mp != NULL) { 1297 mp = &(*mp)->m_next; 1298 stamped = true; 1299 } 1300 break; 1301 1302 default: 1303 panic("unknown (corrupted) so_ts_clock"); 1304 } 1305 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == 1306 (M_PKTHDR | M_TSTMP)) { 1307 struct sock_timestamp_info sti; 1308 1309 bzero(&sti, sizeof(sti)); 1310 sti.st_info_flags = ST_INFO_HW; 1311 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1312 sti.st_info_flags |= ST_INFO_HW_HPREC; 1313 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO, 1314 SOL_SOCKET, M_NOWAIT); 1315 if (*mp != NULL) 1316 mp = &(*mp)->m_next; 1317 } 1318 } 1319 #endif 1320 1321 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1322 /* RFC 2292 sec. 5 */ 1323 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1324 struct in6_pktinfo pi6; 1325 1326 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1327 #ifdef INET 1328 struct ip *ip; 1329 1330 ip = mtod(m, struct ip *); 1331 pi6.ipi6_addr.s6_addr32[0] = 0; 1332 pi6.ipi6_addr.s6_addr32[1] = 0; 1333 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1334 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1335 #else 1336 /* We won't hit this code */ 1337 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1338 #endif 1339 } else { 1340 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1341 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1342 } 1343 pi6.ipi6_ifindex = 1344 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1345 1346 *mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo), 1347 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6, 1348 M_NOWAIT); 1349 if (*mp) 1350 mp = &(*mp)->m_next; 1351 } 1352 1353 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1354 int hlim; 1355 1356 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1357 #ifdef INET 1358 struct ip *ip; 1359 1360 ip = mtod(m, struct ip *); 1361 hlim = ip->ip_ttl; 1362 #else 1363 /* We won't hit this code */ 1364 hlim = 0; 1365 #endif 1366 } else { 1367 hlim = ip6->ip6_hlim & 0xff; 1368 } 1369 *mp = sbcreatecontrol(&hlim, sizeof(int), 1370 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1371 IPPROTO_IPV6, M_NOWAIT); 1372 if (*mp) 1373 mp = &(*mp)->m_next; 1374 } 1375 1376 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1377 int tclass; 1378 1379 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1380 #ifdef INET 1381 struct ip *ip; 1382 1383 ip = mtod(m, struct ip *); 1384 tclass = ip->ip_tos; 1385 #else 1386 /* We won't hit this code */ 1387 tclass = 0; 1388 #endif 1389 } else { 1390 u_int32_t flowinfo; 1391 1392 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1393 flowinfo >>= 20; 1394 tclass = flowinfo & 0xff; 1395 } 1396 *mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS, 1397 IPPROTO_IPV6, M_NOWAIT); 1398 if (*mp) 1399 mp = &(*mp)->m_next; 1400 } 1401 1402 if (v4only != NULL) { 1403 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1404 *v4only = 1; 1405 } else { 1406 *v4only = 0; 1407 } 1408 } 1409 1410 return (mp); 1411 } 1412 1413 void 1414 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp) 1415 { 1416 struct ip6_hdr *ip6; 1417 int v4only = 0; 1418 1419 mp = ip6_savecontrol_v4(inp, m, mp, &v4only); 1420 if (v4only) 1421 return; 1422 1423 ip6 = mtod(m, struct ip6_hdr *); 1424 /* 1425 * IPV6_HOPOPTS socket option. Recall that we required super-user 1426 * privilege for the option (see ip6_ctloutput), but it might be too 1427 * strict, since there might be some hop-by-hop options which can be 1428 * returned to normal user. 1429 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1430 */ 1431 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) { 1432 /* 1433 * Check if a hop-by-hop options header is contatined in the 1434 * received packet, and if so, store the options as ancillary 1435 * data. Note that a hop-by-hop options header must be 1436 * just after the IPv6 header, which is assured through the 1437 * IPv6 input processing. 1438 */ 1439 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1440 struct ip6_hbh *hbh; 1441 u_int hbhlen; 1442 1443 hbh = (struct ip6_hbh *)(ip6 + 1); 1444 hbhlen = (hbh->ip6h_len + 1) << 3; 1445 1446 /* 1447 * XXX: We copy the whole header even if a 1448 * jumbo payload option is included, the option which 1449 * is to be removed before returning according to 1450 * RFC2292. 1451 * Note: this constraint is removed in RFC3542 1452 */ 1453 *mp = sbcreatecontrol(hbh, hbhlen, 1454 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1455 IPPROTO_IPV6, M_NOWAIT); 1456 if (*mp) 1457 mp = &(*mp)->m_next; 1458 } 1459 } 1460 1461 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1462 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1463 1464 /* 1465 * Search for destination options headers or routing 1466 * header(s) through the header chain, and stores each 1467 * header as ancillary data. 1468 * Note that the order of the headers remains in 1469 * the chain of ancillary data. 1470 */ 1471 while (1) { /* is explicit loop prevention necessary? */ 1472 struct ip6_ext *ip6e = NULL; 1473 u_int elen; 1474 1475 /* 1476 * if it is not an extension header, don't try to 1477 * pull it from the chain. 1478 */ 1479 switch (nxt) { 1480 case IPPROTO_DSTOPTS: 1481 case IPPROTO_ROUTING: 1482 case IPPROTO_HOPOPTS: 1483 case IPPROTO_AH: /* is it possible? */ 1484 break; 1485 default: 1486 goto loopend; 1487 } 1488 1489 if (off + sizeof(*ip6e) > m->m_len) 1490 goto loopend; 1491 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1492 if (nxt == IPPROTO_AH) 1493 elen = (ip6e->ip6e_len + 2) << 2; 1494 else 1495 elen = (ip6e->ip6e_len + 1) << 3; 1496 if (off + elen > m->m_len) 1497 goto loopend; 1498 1499 switch (nxt) { 1500 case IPPROTO_DSTOPTS: 1501 if (!(inp->inp_flags & IN6P_DSTOPTS)) 1502 break; 1503 1504 *mp = sbcreatecontrol(ip6e, elen, 1505 IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1506 IPPROTO_IPV6, M_NOWAIT); 1507 if (*mp) 1508 mp = &(*mp)->m_next; 1509 break; 1510 case IPPROTO_ROUTING: 1511 if (!(inp->inp_flags & IN6P_RTHDR)) 1512 break; 1513 1514 *mp = sbcreatecontrol(ip6e, elen, 1515 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR), 1516 IPPROTO_IPV6, M_NOWAIT); 1517 if (*mp) 1518 mp = &(*mp)->m_next; 1519 break; 1520 case IPPROTO_HOPOPTS: 1521 case IPPROTO_AH: /* is it possible? */ 1522 break; 1523 1524 default: 1525 /* 1526 * other cases have been filtered in the above. 1527 * none will visit this case. here we supply 1528 * the code just in case (nxt overwritten or 1529 * other cases). 1530 */ 1531 goto loopend; 1532 } 1533 1534 /* proceed with the next header. */ 1535 off += elen; 1536 nxt = ip6e->ip6e_nxt; 1537 ip6e = NULL; 1538 } 1539 loopend: 1540 ; 1541 } 1542 1543 if (inp->inp_flags2 & INP_RECVFLOWID) { 1544 uint32_t flowid, flow_type; 1545 1546 flowid = m->m_pkthdr.flowid; 1547 flow_type = M_HASHTYPE_GET(m); 1548 1549 /* 1550 * XXX should handle the failure of one or the 1551 * other - don't populate both? 1552 */ 1553 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID, 1554 IPPROTO_IPV6, M_NOWAIT); 1555 if (*mp) 1556 mp = &(*mp)->m_next; 1557 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t), 1558 IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT); 1559 if (*mp) 1560 mp = &(*mp)->m_next; 1561 } 1562 1563 #ifdef RSS 1564 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1565 uint32_t flowid, flow_type; 1566 uint32_t rss_bucketid; 1567 1568 flowid = m->m_pkthdr.flowid; 1569 flow_type = M_HASHTYPE_GET(m); 1570 1571 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1572 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t), 1573 IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT); 1574 if (*mp) 1575 mp = &(*mp)->m_next; 1576 } 1577 } 1578 #endif 1579 1580 } 1581 #undef IS2292 1582 1583 void 1584 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1585 { 1586 struct socket *so; 1587 struct mbuf *m_mtu; 1588 struct ip6_mtuinfo mtuctl; 1589 1590 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1591 /* 1592 * Notify the error by sending IPV6_PATHMTU ancillary data if 1593 * application wanted to know the MTU value. 1594 * NOTE: we notify disconnected sockets, because some udp 1595 * applications keep sending sockets disconnected. 1596 * NOTE: our implementation doesn't notify connected sockets that has 1597 * foreign address that is different than given destination addresses 1598 * (this is permitted by RFC 3542). 1599 */ 1600 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1601 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1602 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1603 return; 1604 1605 mtuctl.ip6m_mtu = mtu; 1606 mtuctl.ip6m_addr = *dst; 1607 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1608 return; 1609 1610 if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU, 1611 IPPROTO_IPV6, M_NOWAIT)) == NULL) 1612 return; 1613 1614 so = inp->inp_socket; 1615 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1616 == 0) { 1617 soroverflow(so); 1618 m_freem(m_mtu); 1619 /* XXX: should count statistics */ 1620 } else 1621 sorwakeup(so); 1622 } 1623 1624 /* 1625 * Get pointer to the previous header followed by the header 1626 * currently processed. 1627 */ 1628 int 1629 ip6_get_prevhdr(const struct mbuf *m, int off) 1630 { 1631 struct ip6_ext ip6e; 1632 struct ip6_hdr *ip6; 1633 int len, nlen, nxt; 1634 1635 if (off == sizeof(struct ip6_hdr)) 1636 return (offsetof(struct ip6_hdr, ip6_nxt)); 1637 if (off < sizeof(struct ip6_hdr)) 1638 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1639 1640 ip6 = mtod(m, struct ip6_hdr *); 1641 nxt = ip6->ip6_nxt; 1642 len = sizeof(struct ip6_hdr); 1643 nlen = 0; 1644 while (len < off) { 1645 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1646 switch (nxt) { 1647 case IPPROTO_FRAGMENT: 1648 nlen = sizeof(struct ip6_frag); 1649 break; 1650 case IPPROTO_AH: 1651 nlen = (ip6e.ip6e_len + 2) << 2; 1652 break; 1653 default: 1654 nlen = (ip6e.ip6e_len + 1) << 3; 1655 } 1656 len += nlen; 1657 nxt = ip6e.ip6e_nxt; 1658 } 1659 return (len - nlen); 1660 } 1661 1662 /* 1663 * get next header offset. m will be retained. 1664 */ 1665 int 1666 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1667 { 1668 struct ip6_hdr ip6; 1669 struct ip6_ext ip6e; 1670 struct ip6_frag fh; 1671 1672 /* just in case */ 1673 if (m == NULL) 1674 panic("ip6_nexthdr: m == NULL"); 1675 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1676 return -1; 1677 1678 switch (proto) { 1679 case IPPROTO_IPV6: 1680 if (m->m_pkthdr.len < off + sizeof(ip6)) 1681 return -1; 1682 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1683 if (nxtp) 1684 *nxtp = ip6.ip6_nxt; 1685 off += sizeof(ip6); 1686 return off; 1687 1688 case IPPROTO_FRAGMENT: 1689 /* 1690 * terminate parsing if it is not the first fragment, 1691 * it does not make sense to parse through it. 1692 */ 1693 if (m->m_pkthdr.len < off + sizeof(fh)) 1694 return -1; 1695 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1696 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1697 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1698 return -1; 1699 if (nxtp) 1700 *nxtp = fh.ip6f_nxt; 1701 off += sizeof(struct ip6_frag); 1702 return off; 1703 1704 case IPPROTO_AH: 1705 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1706 return -1; 1707 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1708 if (nxtp) 1709 *nxtp = ip6e.ip6e_nxt; 1710 off += (ip6e.ip6e_len + 2) << 2; 1711 return off; 1712 1713 case IPPROTO_HOPOPTS: 1714 case IPPROTO_ROUTING: 1715 case IPPROTO_DSTOPTS: 1716 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1717 return -1; 1718 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1719 if (nxtp) 1720 *nxtp = ip6e.ip6e_nxt; 1721 off += (ip6e.ip6e_len + 1) << 3; 1722 return off; 1723 1724 case IPPROTO_NONE: 1725 case IPPROTO_ESP: 1726 case IPPROTO_IPCOMP: 1727 /* give up */ 1728 return -1; 1729 1730 default: 1731 return -1; 1732 } 1733 1734 /* NOTREACHED */ 1735 } 1736 1737 /* 1738 * get offset for the last header in the chain. m will be kept untainted. 1739 */ 1740 int 1741 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1742 { 1743 int newoff; 1744 int nxt; 1745 1746 if (!nxtp) { 1747 nxt = -1; 1748 nxtp = &nxt; 1749 } 1750 while (1) { 1751 newoff = ip6_nexthdr(m, off, proto, nxtp); 1752 if (newoff < 0) 1753 return off; 1754 else if (newoff < off) 1755 return -1; /* invalid */ 1756 else if (newoff == off) 1757 return newoff; 1758 1759 off = newoff; 1760 proto = *nxtp; 1761 } 1762 } 1763