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 __FBSDID("$FreeBSD$"); 67 68 #include "opt_inet.h" 69 #include "opt_inet6.h" 70 #include "opt_ipsec.h" 71 #include "opt_route.h" 72 #include "opt_rss.h" 73 #include "opt_sctp.h" 74 75 #include <sys/param.h> 76 #include <sys/systm.h> 77 #include <sys/hhook.h> 78 #include <sys/malloc.h> 79 #include <sys/mbuf.h> 80 #include <sys/proc.h> 81 #include <sys/domain.h> 82 #include <sys/protosw.h> 83 #include <sys/sdt.h> 84 #include <sys/socket.h> 85 #include <sys/socketvar.h> 86 #include <sys/errno.h> 87 #include <sys/time.h> 88 #include <sys/kernel.h> 89 #include <sys/lock.h> 90 #include <sys/rmlock.h> 91 #include <sys/syslog.h> 92 #include <sys/sysctl.h> 93 #include <sys/eventhandler.h> 94 95 #include <net/if.h> 96 #include <net/if_var.h> 97 #include <net/if_types.h> 98 #include <net/if_dl.h> 99 #include <net/route.h> 100 #include <net/netisr.h> 101 #include <net/rss_config.h> 102 #include <net/pfil.h> 103 #include <net/vnet.h> 104 105 #include <netinet/in.h> 106 #include <netinet/in_kdtrace.h> 107 #include <netinet/ip_var.h> 108 #include <netinet/in_systm.h> 109 #include <net/if_llatbl.h> 110 #ifdef INET 111 #include <netinet/ip.h> 112 #include <netinet/ip_icmp.h> 113 #endif /* INET */ 114 #include <netinet/ip6.h> 115 #include <netinet6/in6_var.h> 116 #include <netinet6/ip6_var.h> 117 #include <netinet/ip_encap.h> 118 #include <netinet/in_pcb.h> 119 #include <netinet/icmp6.h> 120 #include <netinet6/scope6_var.h> 121 #include <netinet6/in6_ifattach.h> 122 #include <netinet6/mld6_var.h> 123 #include <netinet6/nd6.h> 124 #include <netinet6/in6_rss.h> 125 #ifdef SCTP 126 #include <netinet/sctp_pcb.h> 127 #include <netinet6/sctp6_var.h> 128 #endif 129 130 #include <netipsec/ipsec_support.h> 131 132 #include <netinet6/ip6protosw.h> 133 134 ipproto_input_t *ip6_protox[IPPROTO_MAX] = { 135 [0 ... IPPROTO_MAX - 1] = rip6_input }; 136 ipproto_ctlinput_t *ip6_ctlprotox[IPPROTO_MAX] = { 137 [0 ... IPPROTO_MAX - 1] = rip6_ctlinput }; 138 139 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 140 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); 141 VNET_DEFINE(u_long, in6_ifaddrhmask); 142 143 static struct netisr_handler ip6_nh = { 144 .nh_name = "ip6", 145 .nh_handler = ip6_input, 146 .nh_proto = NETISR_IPV6, 147 #ifdef RSS 148 .nh_m2cpuid = rss_soft_m2cpuid_v6, 149 .nh_policy = NETISR_POLICY_CPU, 150 .nh_dispatch = NETISR_DISPATCH_HYBRID, 151 #else 152 .nh_policy = NETISR_POLICY_FLOW, 153 #endif 154 }; 155 156 static int 157 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 158 { 159 int error, qlimit; 160 161 netisr_getqlimit(&ip6_nh, &qlimit); 162 error = sysctl_handle_int(oidp, &qlimit, 0, req); 163 if (error || !req->newptr) 164 return (error); 165 if (qlimit < 1) 166 return (EINVAL); 167 return (netisr_setqlimit(&ip6_nh, qlimit)); 168 } 169 SYSCTL_DECL(_net_inet6_ip6); 170 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen, 171 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 172 0, 0, sysctl_netinet6_intr_queue_maxlen, "I", 173 "Maximum size of the IPv6 input queue"); 174 175 VNET_DEFINE_STATIC(bool, ip6_sav) = true; 176 #define V_ip6_sav VNET(ip6_sav) 177 SYSCTL_BOOL(_net_inet6_ip6, OID_AUTO, source_address_validation, 178 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_sav), true, 179 "Drop incoming packets with source address that is a local address"); 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); 212 213 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 214 VNET_PCPUSTAT_SYSINIT(ip6stat); 215 #ifdef VIMAGE 216 VNET_PCPUSTAT_SYSUNINIT(ip6stat); 217 #endif /* VIMAGE */ 218 219 struct rmlock in6_ifaddr_lock; 220 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 221 222 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 223 224 /* 225 * IP6 initialization: fill in IP6 protocol switch table. 226 * All protocols not implemented in kernel go to raw IP6 protocol handler. 227 */ 228 static void 229 ip6_vnet_init(void *arg __unused) 230 { 231 struct pfil_head_args args; 232 233 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 234 &V_ip6_auto_linklocal); 235 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 236 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 237 238 CK_STAILQ_INIT(&V_in6_ifaddrhead); 239 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 240 &V_in6_ifaddrhmask); 241 242 /* Initialize packet filter hooks. */ 243 args.pa_version = PFIL_VERSION; 244 args.pa_flags = PFIL_IN | PFIL_OUT; 245 args.pa_type = PFIL_TYPE_IP6; 246 args.pa_headname = PFIL_INET6_NAME; 247 V_inet6_pfil_head = pfil_head_register(&args); 248 249 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 250 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 251 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 252 printf("%s: WARNING: unable to register input helper hook\n", 253 __func__); 254 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 255 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 256 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 257 printf("%s: WARNING: unable to register output helper hook\n", 258 __func__); 259 260 scope6_init(); 261 addrsel_policy_init(); 262 nd6_init(); 263 frag6_init(); 264 265 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 266 267 /* Skip global initialization stuff for non-default instances. */ 268 #ifdef VIMAGE 269 netisr_register_vnet(&ip6_nh); 270 #ifdef RSS 271 netisr_register_vnet(&ip6_direct_nh); 272 #endif 273 #endif 274 } 275 VNET_SYSINIT(ip6_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, 276 ip6_vnet_init, NULL); 277 278 static void 279 ip6_init(void *arg __unused) 280 { 281 282 /* 283 * Register statically those protocols that are unlikely to ever go 284 * dynamic. 285 */ 286 IP6PROTO_REGISTER(IPPROTO_ICMPV6, icmp6_input, rip6_ctlinput); 287 IP6PROTO_REGISTER(IPPROTO_DSTOPTS, dest6_input, NULL); 288 IP6PROTO_REGISTER(IPPROTO_ROUTING, route6_input, NULL); 289 IP6PROTO_REGISTER(IPPROTO_FRAGMENT, frag6_input, NULL); 290 IP6PROTO_REGISTER(IPPROTO_IPV4, encap6_input, NULL); 291 IP6PROTO_REGISTER(IPPROTO_IPV6, encap6_input, NULL); 292 IP6PROTO_REGISTER(IPPROTO_ETHERIP, encap6_input, NULL); 293 IP6PROTO_REGISTER(IPPROTO_GRE, encap6_input, NULL); 294 IP6PROTO_REGISTER(IPPROTO_PIM, encap6_input, NULL); 295 #ifdef SCTP /* XXX: has a loadable & static version */ 296 IP6PROTO_REGISTER(IPPROTO_SCTP, sctp6_input, sctp6_ctlinput); 297 #endif 298 299 EVENTHANDLER_REGISTER(vm_lowmem, frag6_drain, NULL, LOWMEM_PRI_DEFAULT); 300 EVENTHANDLER_REGISTER(mbuf_lowmem, frag6_drain, NULL, 301 LOWMEM_PRI_DEFAULT); 302 303 netisr_register(&ip6_nh); 304 #ifdef RSS 305 netisr_register(&ip6_direct_nh); 306 #endif 307 } 308 SYSINIT(ip6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init, NULL); 309 310 int 311 ip6proto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl) 312 { 313 314 MPASS(proto > 0); 315 316 if (ip6_protox[proto] == rip6_input) { 317 ip6_protox[proto] = input; 318 ip6_ctlprotox[proto] = ctl; 319 return (0); 320 } else 321 return (EEXIST); 322 } 323 324 int 325 ip6proto_unregister(uint8_t proto) 326 { 327 328 MPASS(proto > 0); 329 330 if (ip6_protox[proto] != rip6_input) { 331 ip6_protox[proto] = rip6_input; 332 ip6_ctlprotox[proto] = rip6_ctlinput; 333 return (0); 334 } else 335 return (ENOENT); 336 } 337 338 #ifdef VIMAGE 339 static void 340 ip6_destroy(void *unused __unused) 341 { 342 struct ifaddr *ifa, *nifa; 343 struct ifnet *ifp; 344 int error; 345 346 #ifdef RSS 347 netisr_unregister_vnet(&ip6_direct_nh); 348 #endif 349 netisr_unregister_vnet(&ip6_nh); 350 351 pfil_head_unregister(V_inet6_pfil_head); 352 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 353 if (error != 0) { 354 printf("%s: WARNING: unable to deregister input helper hook " 355 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 356 "error %d returned\n", __func__, error); 357 } 358 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 359 if (error != 0) { 360 printf("%s: WARNING: unable to deregister output helper hook " 361 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 362 "error %d returned\n", __func__, error); 363 } 364 365 /* Cleanup addresses. */ 366 IFNET_RLOCK(); 367 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 368 /* Cannot lock here - lock recursion. */ 369 /* IF_ADDR_LOCK(ifp); */ 370 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 371 if (ifa->ifa_addr->sa_family != AF_INET6) 372 continue; 373 in6_purgeaddr(ifa); 374 } 375 /* IF_ADDR_UNLOCK(ifp); */ 376 in6_ifdetach_destroy(ifp); 377 mld_domifdetach(ifp); 378 } 379 IFNET_RUNLOCK(); 380 381 /* Make sure any routes are gone as well. */ 382 rib_flush_routes_family(AF_INET6); 383 384 frag6_destroy(); 385 nd6_destroy(); 386 in6_ifattach_destroy(); 387 388 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 389 } 390 391 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 392 #endif 393 394 static int 395 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off, 396 int *nxt, int *ours) 397 { 398 struct mbuf *m; 399 struct ip6_hdr *ip6; 400 struct ip6_hbh *hbh; 401 402 if (ip6_hopopts_input(plen, rtalert, mp, off)) { 403 #if 0 /*touches NULL pointer*/ 404 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard); 405 #endif 406 goto out; /* m have already been freed */ 407 } 408 409 /* adjust pointer */ 410 m = *mp; 411 ip6 = mtod(m, struct ip6_hdr *); 412 413 /* 414 * if the payload length field is 0 and the next header field 415 * indicates Hop-by-Hop Options header, then a Jumbo Payload 416 * option MUST be included. 417 */ 418 if (ip6->ip6_plen == 0 && *plen == 0) { 419 /* 420 * Note that if a valid jumbo payload option is 421 * contained, ip6_hopopts_input() must set a valid 422 * (non-zero) payload length to the variable plen. 423 */ 424 IP6STAT_INC(ip6s_badoptions); 425 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 426 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 427 icmp6_error(m, ICMP6_PARAM_PROB, 428 ICMP6_PARAMPROB_HEADER, 429 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 430 goto out; 431 } 432 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 433 hbh = (struct ip6_hbh *)(ip6 + 1); 434 *nxt = hbh->ip6h_nxt; 435 436 /* 437 * If we are acting as a router and the packet contains a 438 * router alert option, see if we know the option value. 439 * Currently, we only support the option value for MLD, in which 440 * case we should pass the packet to the multicast routing 441 * daemon. 442 */ 443 if (*rtalert != ~0) { 444 switch (*rtalert) { 445 case IP6OPT_RTALERT_MLD: 446 if (V_ip6_forwarding) 447 *ours = 1; 448 break; 449 default: 450 /* 451 * RFC2711 requires unrecognized values must be 452 * silently ignored. 453 */ 454 break; 455 } 456 } 457 458 return (0); 459 460 out: 461 return (1); 462 } 463 464 #ifdef RSS 465 /* 466 * IPv6 direct input routine. 467 * 468 * This is called when reinjecting completed fragments where 469 * all of the previous checking and book-keeping has been done. 470 */ 471 void 472 ip6_direct_input(struct mbuf *m) 473 { 474 int off, nxt; 475 int nest; 476 struct m_tag *mtag; 477 struct ip6_direct_ctx *ip6dc; 478 479 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 480 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 481 482 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 483 nxt = ip6dc->ip6dc_nxt; 484 off = ip6dc->ip6dc_off; 485 486 nest = 0; 487 488 m_tag_delete(m, mtag); 489 490 while (nxt != IPPROTO_DONE) { 491 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 492 IP6STAT_INC(ip6s_toomanyhdr); 493 goto bad; 494 } 495 496 /* 497 * protection against faulty packet - there should be 498 * more sanity checks in header chain processing. 499 */ 500 if (m->m_pkthdr.len < off) { 501 IP6STAT_INC(ip6s_tooshort); 502 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 503 goto bad; 504 } 505 506 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 507 if (IPSEC_ENABLED(ipv6)) { 508 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 509 return; 510 } 511 #endif /* IPSEC */ 512 513 nxt = ip6_protox[nxt](&m, &off, nxt); 514 } 515 return; 516 bad: 517 m_freem(m); 518 } 519 #endif 520 521 void 522 ip6_input(struct mbuf *m) 523 { 524 struct in6_addr odst; 525 struct ip6_hdr *ip6; 526 struct in6_ifaddr *ia; 527 struct ifnet *rcvif; 528 u_int32_t plen; 529 u_int32_t rtalert = ~0; 530 int off = sizeof(struct ip6_hdr), nest; 531 int nxt, ours = 0; 532 int srcrt = 0; 533 534 /* 535 * Drop the packet if IPv6 operation is disabled on the interface. 536 */ 537 rcvif = m->m_pkthdr.rcvif; 538 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 539 goto bad; 540 541 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 542 /* 543 * should the inner packet be considered authentic? 544 * see comment in ah4_input(). 545 * NB: m cannot be NULL when passed to the input routine 546 */ 547 548 m->m_flags &= ~M_AUTHIPHDR; 549 m->m_flags &= ~M_AUTHIPDGM; 550 551 #endif /* IPSEC */ 552 553 if (m->m_flags & M_FASTFWD_OURS) { 554 /* 555 * Firewall changed destination to local. 556 */ 557 ip6 = mtod(m, struct ip6_hdr *); 558 goto passin; 559 } 560 561 /* 562 * mbuf statistics 563 */ 564 if (m->m_flags & M_EXT) { 565 if (m->m_next) 566 IP6STAT_INC(ip6s_mext2m); 567 else 568 IP6STAT_INC(ip6s_mext1); 569 } else { 570 if (m->m_next) { 571 struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif; 572 int ifindex = ifp->if_index; 573 if (ifindex >= IP6S_M2MMAX) 574 ifindex = 0; 575 IP6STAT_INC(ip6s_m2m[ifindex]); 576 } else 577 IP6STAT_INC(ip6s_m1); 578 } 579 580 in6_ifstat_inc(rcvif, ifs6_in_receive); 581 IP6STAT_INC(ip6s_total); 582 583 /* 584 * L2 bridge code and some other code can return mbuf chain 585 * that does not conform to KAME requirement. too bad. 586 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 587 */ 588 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 589 struct mbuf *n; 590 591 if (m->m_pkthdr.len > MHLEN) 592 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 593 else 594 n = m_gethdr(M_NOWAIT, MT_DATA); 595 if (n == NULL) 596 goto bad; 597 598 m_move_pkthdr(n, m); 599 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 600 n->m_len = n->m_pkthdr.len; 601 m_freem(m); 602 m = n; 603 } 604 if (m->m_len < sizeof(struct ip6_hdr)) { 605 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 606 IP6STAT_INC(ip6s_toosmall); 607 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 608 goto bad; 609 } 610 } 611 612 ip6 = mtod(m, struct ip6_hdr *); 613 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 614 IP6STAT_INC(ip6s_badvers); 615 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 616 goto bad; 617 } 618 619 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 620 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 621 622 /* 623 * Check against address spoofing/corruption. 624 */ 625 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 626 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 627 /* 628 * XXX: "badscope" is not very suitable for a multicast source. 629 */ 630 IP6STAT_INC(ip6s_badscope); 631 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 632 goto bad; 633 } 634 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 635 !(m->m_flags & M_LOOP)) { 636 /* 637 * In this case, the packet should come from the loopback 638 * interface. However, we cannot just check the if_flags, 639 * because ip6_mloopback() passes the "actual" interface 640 * as the outgoing/incoming interface. 641 */ 642 IP6STAT_INC(ip6s_badscope); 643 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 644 goto bad; 645 } 646 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 647 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 648 /* 649 * RFC4291 2.7: 650 * Nodes must not originate a packet to a multicast address 651 * whose scop field contains the reserved value 0; if such 652 * a packet is received, it must be silently dropped. 653 */ 654 IP6STAT_INC(ip6s_badscope); 655 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 656 goto bad; 657 } 658 #ifdef ALTQ 659 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 660 /* packet is dropped by traffic conditioner */ 661 return; 662 } 663 #endif 664 /* 665 * The following check is not documented in specs. A malicious 666 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 667 * and bypass security checks (act as if it was from 127.0.0.1 by using 668 * IPv6 src ::ffff:127.0.0.1). Be cautious. 669 * 670 * We have supported IPv6-only kernels for a few years and this issue 671 * has not come up. The world seems to move mostly towards not using 672 * v4mapped on the wire, so it makes sense for us to keep rejecting 673 * any such packets. 674 */ 675 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 676 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 677 IP6STAT_INC(ip6s_badscope); 678 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 679 goto bad; 680 } 681 #if 0 682 /* 683 * Reject packets with IPv4 compatible addresses (auto tunnel). 684 * 685 * The code forbids auto tunnel relay case in RFC1933 (the check is 686 * stronger than RFC1933). We may want to re-enable it if mech-xx 687 * is revised to forbid relaying case. 688 */ 689 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 690 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 691 IP6STAT_INC(ip6s_badscope); 692 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 693 goto bad; 694 } 695 #endif 696 /* 697 * Try to forward the packet, but if we fail continue. 698 * ip6_tryforward() does not generate redirects, so fall 699 * through to normal processing if redirects are required. 700 * ip6_tryforward() does inbound and outbound packet firewall 701 * processing. If firewall has decided that destination becomes 702 * our local address, it sets M_FASTFWD_OURS flag. In this 703 * case skip another inbound firewall processing and update 704 * ip6 pointer. 705 */ 706 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0 707 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 708 && (!IPSEC_ENABLED(ipv6) || 709 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0) 710 #endif 711 ) { 712 if ((m = ip6_tryforward(m)) == NULL) 713 return; 714 if (m->m_flags & M_FASTFWD_OURS) { 715 ip6 = mtod(m, struct ip6_hdr *); 716 goto passin; 717 } 718 } 719 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 720 /* 721 * Bypass packet filtering for packets previously handled by IPsec. 722 */ 723 if (IPSEC_ENABLED(ipv6) && 724 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0) 725 goto passin; 726 #endif 727 /* 728 * Run through list of hooks for input packets. 729 * 730 * NB: Beware of the destination address changing 731 * (e.g. by NAT rewriting). When this happens, 732 * tell ip6_forward to do the right thing. 733 */ 734 735 /* Jump over all PFIL processing if hooks are not active. */ 736 if (!PFIL_HOOKED_IN(V_inet6_pfil_head)) 737 goto passin; 738 739 odst = ip6->ip6_dst; 740 if (pfil_run_hooks(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif, PFIL_IN, 741 NULL) != PFIL_PASS) 742 return; 743 ip6 = mtod(m, struct ip6_hdr *); 744 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 745 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 746 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 747 /* 748 * Directly ship the packet on. This allows forwarding 749 * packets originally destined to us to some other directly 750 * connected host. 751 */ 752 ip6_forward(m, 1); 753 return; 754 } 755 756 passin: 757 /* 758 * Disambiguate address scope zones (if there is ambiguity). 759 * We first make sure that the original source or destination address 760 * is not in our internal form for scoped addresses. Such addresses 761 * are not necessarily invalid spec-wise, but we cannot accept them due 762 * to the usage conflict. 763 * in6_setscope() then also checks and rejects the cases where src or 764 * dst are the loopback address and the receiving interface 765 * is not loopback. 766 */ 767 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 768 IP6STAT_INC(ip6s_badscope); /* XXX */ 769 goto bad; 770 } 771 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 772 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 773 IP6STAT_INC(ip6s_badscope); 774 goto bad; 775 } 776 if (m->m_flags & M_FASTFWD_OURS) { 777 m->m_flags &= ~M_FASTFWD_OURS; 778 ours = 1; 779 goto hbhcheck; 780 } 781 /* 782 * Multicast check. Assume packet is for us to avoid 783 * prematurely taking locks. 784 */ 785 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 786 ours = 1; 787 in6_ifstat_inc(rcvif, ifs6_in_mcast); 788 goto hbhcheck; 789 } 790 /* 791 * Unicast check 792 * XXX: For now we keep link-local IPv6 addresses with embedded 793 * scope zone id, therefore we use zero zoneid here. 794 */ 795 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false); 796 if (ia != NULL) { 797 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 798 char ip6bufs[INET6_ADDRSTRLEN]; 799 char ip6bufd[INET6_ADDRSTRLEN]; 800 /* address is not ready, so discard the packet. */ 801 nd6log((LOG_INFO, 802 "ip6_input: packet to an unready address %s->%s\n", 803 ip6_sprintf(ip6bufs, &ip6->ip6_src), 804 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 805 goto bad; 806 } 807 if (V_ip6_sav && !(m->m_flags & M_LOOP) && 808 __predict_false(in6_localip_fib(&ip6->ip6_src, 809 rcvif->if_fib))) { 810 IP6STAT_INC(ip6s_badscope); /* XXX */ 811 goto bad; 812 } 813 /* Count the packet in the ip address stats */ 814 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 815 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 816 ours = 1; 817 goto hbhcheck; 818 } 819 820 /* 821 * Now there is no reason to process the packet if it's not our own 822 * and we're not a router. 823 */ 824 if (!V_ip6_forwarding) { 825 IP6STAT_INC(ip6s_cantforward); 826 goto bad; 827 } 828 829 hbhcheck: 830 /* 831 * Process Hop-by-Hop options header if it's contained. 832 * m may be modified in ip6_hopopts_input(). 833 * If a JumboPayload option is included, plen will also be modified. 834 */ 835 plen = (u_int32_t)ntohs(ip6->ip6_plen); 836 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 837 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0) 838 return; 839 } else 840 nxt = ip6->ip6_nxt; 841 842 /* 843 * Use mbuf flags to propagate Router Alert option to 844 * ICMPv6 layer, as hop-by-hop options have been stripped. 845 */ 846 if (rtalert != ~0) 847 m->m_flags |= M_RTALERT_MLD; 848 849 /* 850 * Check that the amount of data in the buffers 851 * is as at least much as the IPv6 header would have us expect. 852 * Trim mbufs if longer than we expect. 853 * Drop packet if shorter than we expect. 854 */ 855 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 856 IP6STAT_INC(ip6s_tooshort); 857 in6_ifstat_inc(rcvif, ifs6_in_truncated); 858 goto bad; 859 } 860 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 861 if (m->m_len == m->m_pkthdr.len) { 862 m->m_len = sizeof(struct ip6_hdr) + plen; 863 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 864 } else 865 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 866 } 867 868 /* 869 * Forward if desirable. 870 */ 871 if (V_ip6_mrouter && 872 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 873 /* 874 * If we are acting as a multicast router, all 875 * incoming multicast packets are passed to the 876 * kernel-level multicast forwarding function. 877 * The packet is returned (relatively) intact; if 878 * ip6_mforward() returns a non-zero value, the packet 879 * must be discarded, else it may be accepted below. 880 * 881 * XXX TODO: Check hlim and multicast scope here to avoid 882 * unnecessarily calling into ip6_mforward(). 883 */ 884 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 885 IP6STAT_INC(ip6s_cantforward); 886 goto bad; 887 } 888 } else if (!ours) { 889 ip6_forward(m, srcrt); 890 return; 891 } 892 893 /* 894 * Tell launch routine the next header 895 */ 896 IP6STAT_INC(ip6s_delivered); 897 in6_ifstat_inc(rcvif, ifs6_in_deliver); 898 nest = 0; 899 900 while (nxt != IPPROTO_DONE) { 901 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 902 IP6STAT_INC(ip6s_toomanyhdr); 903 goto bad; 904 } 905 906 /* 907 * protection against faulty packet - there should be 908 * more sanity checks in header chain processing. 909 */ 910 if (m->m_pkthdr.len < off) { 911 IP6STAT_INC(ip6s_tooshort); 912 in6_ifstat_inc(rcvif, ifs6_in_truncated); 913 goto bad; 914 } 915 916 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 917 if (IPSEC_ENABLED(ipv6)) { 918 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 919 return; 920 } 921 #endif /* IPSEC */ 922 923 nxt = ip6_protox[nxt](&m, &off, nxt); 924 } 925 return; 926 bad: 927 in6_ifstat_inc(rcvif, ifs6_in_discard); 928 if (m != NULL) 929 m_freem(m); 930 } 931 932 /* 933 * Hop-by-Hop options header processing. If a valid jumbo payload option is 934 * included, the real payload length will be stored in plenp. 935 * 936 * rtalertp - XXX: should be stored more smart way 937 */ 938 static int 939 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 940 struct mbuf **mp, int *offp) 941 { 942 struct mbuf *m = *mp; 943 int off = *offp, hbhlen; 944 struct ip6_hbh *hbh; 945 946 /* validation of the length of the header */ 947 if (m->m_len < off + sizeof(*hbh)) { 948 m = m_pullup(m, off + sizeof(*hbh)); 949 if (m == NULL) { 950 IP6STAT_INC(ip6s_exthdrtoolong); 951 *mp = NULL; 952 return (-1); 953 } 954 } 955 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 956 hbhlen = (hbh->ip6h_len + 1) << 3; 957 958 if (m->m_len < off + hbhlen) { 959 m = m_pullup(m, off + hbhlen); 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 off += hbhlen; 968 hbhlen -= sizeof(struct ip6_hbh); 969 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 970 hbhlen, rtalertp, plenp) < 0) { 971 *mp = NULL; 972 return (-1); 973 } 974 975 *offp = off; 976 *mp = m; 977 return (0); 978 } 979 980 /* 981 * Search header for all Hop-by-hop options and process each option. 982 * This function is separate from ip6_hopopts_input() in order to 983 * handle a case where the sending node itself process its hop-by-hop 984 * options header. In such a case, the function is called from ip6_output(). 985 * 986 * The function assumes that hbh header is located right after the IPv6 header 987 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 988 * opthead + hbhlen is located in contiguous memory region. 989 */ 990 int 991 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 992 u_int32_t *rtalertp, u_int32_t *plenp) 993 { 994 struct ip6_hdr *ip6; 995 int optlen = 0; 996 u_int8_t *opt = opthead; 997 u_int16_t rtalert_val; 998 u_int32_t jumboplen; 999 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1000 1001 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1002 switch (*opt) { 1003 case IP6OPT_PAD1: 1004 optlen = 1; 1005 break; 1006 case IP6OPT_PADN: 1007 if (hbhlen < IP6OPT_MINLEN) { 1008 IP6STAT_INC(ip6s_toosmall); 1009 goto bad; 1010 } 1011 optlen = *(opt + 1) + 2; 1012 break; 1013 case IP6OPT_ROUTER_ALERT: 1014 /* XXX may need check for alignment */ 1015 if (hbhlen < IP6OPT_RTALERT_LEN) { 1016 IP6STAT_INC(ip6s_toosmall); 1017 goto bad; 1018 } 1019 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1020 /* XXX stat */ 1021 icmp6_error(m, ICMP6_PARAM_PROB, 1022 ICMP6_PARAMPROB_HEADER, 1023 erroff + opt + 1 - opthead); 1024 return (-1); 1025 } 1026 optlen = IP6OPT_RTALERT_LEN; 1027 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1028 *rtalertp = ntohs(rtalert_val); 1029 break; 1030 case IP6OPT_JUMBO: 1031 /* XXX may need check for alignment */ 1032 if (hbhlen < IP6OPT_JUMBO_LEN) { 1033 IP6STAT_INC(ip6s_toosmall); 1034 goto bad; 1035 } 1036 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1037 /* XXX stat */ 1038 icmp6_error(m, ICMP6_PARAM_PROB, 1039 ICMP6_PARAMPROB_HEADER, 1040 erroff + opt + 1 - opthead); 1041 return (-1); 1042 } 1043 optlen = IP6OPT_JUMBO_LEN; 1044 1045 /* 1046 * IPv6 packets that have non 0 payload length 1047 * must not contain a jumbo payload option. 1048 */ 1049 ip6 = mtod(m, struct ip6_hdr *); 1050 if (ip6->ip6_plen) { 1051 IP6STAT_INC(ip6s_badoptions); 1052 icmp6_error(m, ICMP6_PARAM_PROB, 1053 ICMP6_PARAMPROB_HEADER, 1054 erroff + opt - opthead); 1055 return (-1); 1056 } 1057 1058 /* 1059 * We may see jumbolen in unaligned location, so 1060 * we'd need to perform bcopy(). 1061 */ 1062 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1063 jumboplen = (u_int32_t)htonl(jumboplen); 1064 1065 #if 1 1066 /* 1067 * if there are multiple jumbo payload options, 1068 * *plenp will be non-zero and the packet will be 1069 * rejected. 1070 * the behavior may need some debate in ipngwg - 1071 * multiple options does not make sense, however, 1072 * there's no explicit mention in specification. 1073 */ 1074 if (*plenp != 0) { 1075 IP6STAT_INC(ip6s_badoptions); 1076 icmp6_error(m, ICMP6_PARAM_PROB, 1077 ICMP6_PARAMPROB_HEADER, 1078 erroff + opt + 2 - opthead); 1079 return (-1); 1080 } 1081 #endif 1082 1083 /* 1084 * jumbo payload length must be larger than 65535. 1085 */ 1086 if (jumboplen <= IPV6_MAXPACKET) { 1087 IP6STAT_INC(ip6s_badoptions); 1088 icmp6_error(m, ICMP6_PARAM_PROB, 1089 ICMP6_PARAMPROB_HEADER, 1090 erroff + opt + 2 - opthead); 1091 return (-1); 1092 } 1093 *plenp = jumboplen; 1094 1095 break; 1096 default: /* unknown option */ 1097 if (hbhlen < IP6OPT_MINLEN) { 1098 IP6STAT_INC(ip6s_toosmall); 1099 goto bad; 1100 } 1101 optlen = ip6_unknown_opt(opt, m, 1102 erroff + opt - opthead); 1103 if (optlen == -1) 1104 return (-1); 1105 optlen += 2; 1106 break; 1107 } 1108 } 1109 1110 return (0); 1111 1112 bad: 1113 m_freem(m); 1114 return (-1); 1115 } 1116 1117 /* 1118 * Unknown option processing. 1119 * The third argument `off' is the offset from the IPv6 header to the option, 1120 * which is necessary if the IPv6 header the and option header and IPv6 header 1121 * is not contiguous in order to return an ICMPv6 error. 1122 */ 1123 int 1124 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1125 { 1126 struct ip6_hdr *ip6; 1127 1128 switch (IP6OPT_TYPE(*optp)) { 1129 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1130 return ((int)*(optp + 1)); 1131 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1132 m_freem(m); 1133 return (-1); 1134 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1135 IP6STAT_INC(ip6s_badoptions); 1136 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1137 return (-1); 1138 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1139 IP6STAT_INC(ip6s_badoptions); 1140 ip6 = mtod(m, struct ip6_hdr *); 1141 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1142 (m->m_flags & (M_BCAST|M_MCAST))) 1143 m_freem(m); 1144 else 1145 icmp6_error(m, ICMP6_PARAM_PROB, 1146 ICMP6_PARAMPROB_OPTION, off); 1147 return (-1); 1148 } 1149 1150 m_freem(m); /* XXX: NOTREACHED */ 1151 return (-1); 1152 } 1153 1154 /* 1155 * Create the "control" list for this pcb. 1156 * These functions will not modify mbuf chain at all. 1157 * 1158 * The routine will be called from upper layer handlers like tcp6_input(). 1159 * Thus the routine assumes that the caller (tcp6_input) have already 1160 * called m_pullup() and all the extension headers are located in the 1161 * very first mbuf on the mbuf chain. 1162 * 1163 * ip6_savecontrol_v4 will handle those options that are possible to be 1164 * set on a v4-mapped socket. 1165 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1166 * options and handle the v6-only ones itself. 1167 */ 1168 struct mbuf ** 1169 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1170 int *v4only) 1171 { 1172 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1173 1174 #ifdef SO_TIMESTAMP 1175 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1176 union { 1177 struct timeval tv; 1178 struct bintime bt; 1179 struct timespec ts; 1180 } t; 1181 struct bintime boottimebin, bt1; 1182 struct timespec ts1; 1183 bool stamped; 1184 1185 stamped = false; 1186 switch (inp->inp_socket->so_ts_clock) { 1187 case SO_TS_REALTIME_MICRO: 1188 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1189 M_TSTMP)) { 1190 mbuf_tstmp2timespec(m, &ts1); 1191 timespec2bintime(&ts1, &bt1); 1192 getboottimebin(&boottimebin); 1193 bintime_add(&bt1, &boottimebin); 1194 bintime2timeval(&bt1, &t.tv); 1195 } else { 1196 microtime(&t.tv); 1197 } 1198 *mp = sbcreatecontrol(&t.tv, sizeof(t.tv), 1199 SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT); 1200 if (*mp != NULL) { 1201 mp = &(*mp)->m_next; 1202 stamped = true; 1203 } 1204 break; 1205 1206 case SO_TS_BINTIME: 1207 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1208 M_TSTMP)) { 1209 mbuf_tstmp2timespec(m, &ts1); 1210 timespec2bintime(&ts1, &t.bt); 1211 getboottimebin(&boottimebin); 1212 bintime_add(&t.bt, &boottimebin); 1213 } else { 1214 bintime(&t.bt); 1215 } 1216 *mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME, 1217 SOL_SOCKET, M_NOWAIT); 1218 if (*mp != NULL) { 1219 mp = &(*mp)->m_next; 1220 stamped = true; 1221 } 1222 break; 1223 1224 case SO_TS_REALTIME: 1225 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1226 M_TSTMP)) { 1227 mbuf_tstmp2timespec(m, &t.ts); 1228 getboottimebin(&boottimebin); 1229 bintime2timespec(&boottimebin, &ts1); 1230 timespecadd(&t.ts, &ts1, &t.ts); 1231 } else { 1232 nanotime(&t.ts); 1233 } 1234 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1235 SCM_REALTIME, SOL_SOCKET, M_NOWAIT); 1236 if (*mp != NULL) { 1237 mp = &(*mp)->m_next; 1238 stamped = true; 1239 } 1240 break; 1241 1242 case SO_TS_MONOTONIC: 1243 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1244 M_TSTMP)) 1245 mbuf_tstmp2timespec(m, &t.ts); 1246 else 1247 nanouptime(&t.ts); 1248 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1249 SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT); 1250 if (*mp != NULL) { 1251 mp = &(*mp)->m_next; 1252 stamped = true; 1253 } 1254 break; 1255 1256 default: 1257 panic("unknown (corrupted) so_ts_clock"); 1258 } 1259 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == 1260 (M_PKTHDR | M_TSTMP)) { 1261 struct sock_timestamp_info sti; 1262 1263 bzero(&sti, sizeof(sti)); 1264 sti.st_info_flags = ST_INFO_HW; 1265 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1266 sti.st_info_flags |= ST_INFO_HW_HPREC; 1267 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO, 1268 SOL_SOCKET, M_NOWAIT); 1269 if (*mp != NULL) 1270 mp = &(*mp)->m_next; 1271 } 1272 } 1273 #endif 1274 1275 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1276 /* RFC 2292 sec. 5 */ 1277 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1278 struct in6_pktinfo pi6; 1279 1280 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1281 #ifdef INET 1282 struct ip *ip; 1283 1284 ip = mtod(m, struct ip *); 1285 pi6.ipi6_addr.s6_addr32[0] = 0; 1286 pi6.ipi6_addr.s6_addr32[1] = 0; 1287 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1288 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1289 #else 1290 /* We won't hit this code */ 1291 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1292 #endif 1293 } else { 1294 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1295 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1296 } 1297 pi6.ipi6_ifindex = 1298 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1299 1300 *mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo), 1301 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6, 1302 M_NOWAIT); 1303 if (*mp) 1304 mp = &(*mp)->m_next; 1305 } 1306 1307 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1308 int hlim; 1309 1310 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1311 #ifdef INET 1312 struct ip *ip; 1313 1314 ip = mtod(m, struct ip *); 1315 hlim = ip->ip_ttl; 1316 #else 1317 /* We won't hit this code */ 1318 hlim = 0; 1319 #endif 1320 } else { 1321 hlim = ip6->ip6_hlim & 0xff; 1322 } 1323 *mp = sbcreatecontrol(&hlim, sizeof(int), 1324 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1325 IPPROTO_IPV6, M_NOWAIT); 1326 if (*mp) 1327 mp = &(*mp)->m_next; 1328 } 1329 1330 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1331 int tclass; 1332 1333 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1334 #ifdef INET 1335 struct ip *ip; 1336 1337 ip = mtod(m, struct ip *); 1338 tclass = ip->ip_tos; 1339 #else 1340 /* We won't hit this code */ 1341 tclass = 0; 1342 #endif 1343 } else { 1344 u_int32_t flowinfo; 1345 1346 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1347 flowinfo >>= 20; 1348 tclass = flowinfo & 0xff; 1349 } 1350 *mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS, 1351 IPPROTO_IPV6, M_NOWAIT); 1352 if (*mp) 1353 mp = &(*mp)->m_next; 1354 } 1355 1356 if (v4only != NULL) { 1357 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1358 *v4only = 1; 1359 } else { 1360 *v4only = 0; 1361 } 1362 } 1363 1364 return (mp); 1365 } 1366 1367 void 1368 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp) 1369 { 1370 struct ip6_hdr *ip6; 1371 int v4only = 0; 1372 1373 mp = ip6_savecontrol_v4(inp, m, mp, &v4only); 1374 if (v4only) 1375 return; 1376 1377 ip6 = mtod(m, struct ip6_hdr *); 1378 /* 1379 * IPV6_HOPOPTS socket option. Recall that we required super-user 1380 * privilege for the option (see ip6_ctloutput), but it might be too 1381 * strict, since there might be some hop-by-hop options which can be 1382 * returned to normal user. 1383 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1384 */ 1385 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) { 1386 /* 1387 * Check if a hop-by-hop options header is contatined in the 1388 * received packet, and if so, store the options as ancillary 1389 * data. Note that a hop-by-hop options header must be 1390 * just after the IPv6 header, which is assured through the 1391 * IPv6 input processing. 1392 */ 1393 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1394 struct ip6_hbh *hbh; 1395 u_int hbhlen; 1396 1397 hbh = (struct ip6_hbh *)(ip6 + 1); 1398 hbhlen = (hbh->ip6h_len + 1) << 3; 1399 1400 /* 1401 * XXX: We copy the whole header even if a 1402 * jumbo payload option is included, the option which 1403 * is to be removed before returning according to 1404 * RFC2292. 1405 * Note: this constraint is removed in RFC3542 1406 */ 1407 *mp = sbcreatecontrol(hbh, hbhlen, 1408 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1409 IPPROTO_IPV6, M_NOWAIT); 1410 if (*mp) 1411 mp = &(*mp)->m_next; 1412 } 1413 } 1414 1415 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1416 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1417 1418 /* 1419 * Search for destination options headers or routing 1420 * header(s) through the header chain, and stores each 1421 * header as ancillary data. 1422 * Note that the order of the headers remains in 1423 * the chain of ancillary data. 1424 */ 1425 while (1) { /* is explicit loop prevention necessary? */ 1426 struct ip6_ext *ip6e = NULL; 1427 u_int elen; 1428 1429 /* 1430 * if it is not an extension header, don't try to 1431 * pull it from the chain. 1432 */ 1433 switch (nxt) { 1434 case IPPROTO_DSTOPTS: 1435 case IPPROTO_ROUTING: 1436 case IPPROTO_HOPOPTS: 1437 case IPPROTO_AH: /* is it possible? */ 1438 break; 1439 default: 1440 goto loopend; 1441 } 1442 1443 if (off + sizeof(*ip6e) > m->m_len) 1444 goto loopend; 1445 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1446 if (nxt == IPPROTO_AH) 1447 elen = (ip6e->ip6e_len + 2) << 2; 1448 else 1449 elen = (ip6e->ip6e_len + 1) << 3; 1450 if (off + elen > m->m_len) 1451 goto loopend; 1452 1453 switch (nxt) { 1454 case IPPROTO_DSTOPTS: 1455 if (!(inp->inp_flags & IN6P_DSTOPTS)) 1456 break; 1457 1458 *mp = sbcreatecontrol(ip6e, elen, 1459 IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1460 IPPROTO_IPV6, M_NOWAIT); 1461 if (*mp) 1462 mp = &(*mp)->m_next; 1463 break; 1464 case IPPROTO_ROUTING: 1465 if (!(inp->inp_flags & IN6P_RTHDR)) 1466 break; 1467 1468 *mp = sbcreatecontrol(ip6e, elen, 1469 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR), 1470 IPPROTO_IPV6, M_NOWAIT); 1471 if (*mp) 1472 mp = &(*mp)->m_next; 1473 break; 1474 case IPPROTO_HOPOPTS: 1475 case IPPROTO_AH: /* is it possible? */ 1476 break; 1477 1478 default: 1479 /* 1480 * other cases have been filtered in the above. 1481 * none will visit this case. here we supply 1482 * the code just in case (nxt overwritten or 1483 * other cases). 1484 */ 1485 goto loopend; 1486 } 1487 1488 /* proceed with the next header. */ 1489 off += elen; 1490 nxt = ip6e->ip6e_nxt; 1491 ip6e = NULL; 1492 } 1493 loopend: 1494 ; 1495 } 1496 1497 if (inp->inp_flags2 & INP_RECVFLOWID) { 1498 uint32_t flowid, flow_type; 1499 1500 flowid = m->m_pkthdr.flowid; 1501 flow_type = M_HASHTYPE_GET(m); 1502 1503 /* 1504 * XXX should handle the failure of one or the 1505 * other - don't populate both? 1506 */ 1507 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID, 1508 IPPROTO_IPV6, M_NOWAIT); 1509 if (*mp) 1510 mp = &(*mp)->m_next; 1511 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t), 1512 IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT); 1513 if (*mp) 1514 mp = &(*mp)->m_next; 1515 } 1516 1517 #ifdef RSS 1518 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1519 uint32_t flowid, flow_type; 1520 uint32_t rss_bucketid; 1521 1522 flowid = m->m_pkthdr.flowid; 1523 flow_type = M_HASHTYPE_GET(m); 1524 1525 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1526 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t), 1527 IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT); 1528 if (*mp) 1529 mp = &(*mp)->m_next; 1530 } 1531 } 1532 #endif 1533 1534 } 1535 #undef IS2292 1536 1537 void 1538 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1539 { 1540 struct socket *so; 1541 struct mbuf *m_mtu; 1542 struct ip6_mtuinfo mtuctl; 1543 1544 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1545 /* 1546 * Notify the error by sending IPV6_PATHMTU ancillary data if 1547 * application wanted to know the MTU value. 1548 * NOTE: we notify disconnected sockets, because some udp 1549 * applications keep sending sockets disconnected. 1550 * NOTE: our implementation doesn't notify connected sockets that has 1551 * foreign address that is different than given destination addresses 1552 * (this is permitted by RFC 3542). 1553 */ 1554 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1555 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1556 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1557 return; 1558 1559 mtuctl.ip6m_mtu = mtu; 1560 mtuctl.ip6m_addr = *dst; 1561 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1562 return; 1563 1564 if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU, 1565 IPPROTO_IPV6, M_NOWAIT)) == NULL) 1566 return; 1567 1568 so = inp->inp_socket; 1569 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1570 == 0) { 1571 soroverflow(so); 1572 m_freem(m_mtu); 1573 /* XXX: should count statistics */ 1574 } else 1575 sorwakeup(so); 1576 } 1577 1578 /* 1579 * Get pointer to the previous header followed by the header 1580 * currently processed. 1581 */ 1582 int 1583 ip6_get_prevhdr(const struct mbuf *m, int off) 1584 { 1585 struct ip6_ext ip6e; 1586 struct ip6_hdr *ip6; 1587 int len, nlen, nxt; 1588 1589 if (off == sizeof(struct ip6_hdr)) 1590 return (offsetof(struct ip6_hdr, ip6_nxt)); 1591 if (off < sizeof(struct ip6_hdr)) 1592 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1593 1594 ip6 = mtod(m, struct ip6_hdr *); 1595 nxt = ip6->ip6_nxt; 1596 len = sizeof(struct ip6_hdr); 1597 nlen = 0; 1598 while (len < off) { 1599 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1600 switch (nxt) { 1601 case IPPROTO_FRAGMENT: 1602 nlen = sizeof(struct ip6_frag); 1603 break; 1604 case IPPROTO_AH: 1605 nlen = (ip6e.ip6e_len + 2) << 2; 1606 break; 1607 default: 1608 nlen = (ip6e.ip6e_len + 1) << 3; 1609 } 1610 len += nlen; 1611 nxt = ip6e.ip6e_nxt; 1612 } 1613 return (len - nlen); 1614 } 1615 1616 /* 1617 * get next header offset. m will be retained. 1618 */ 1619 int 1620 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1621 { 1622 struct ip6_hdr ip6; 1623 struct ip6_ext ip6e; 1624 struct ip6_frag fh; 1625 1626 /* just in case */ 1627 if (m == NULL) 1628 panic("ip6_nexthdr: m == NULL"); 1629 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1630 return -1; 1631 1632 switch (proto) { 1633 case IPPROTO_IPV6: 1634 if (m->m_pkthdr.len < off + sizeof(ip6)) 1635 return -1; 1636 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1637 if (nxtp) 1638 *nxtp = ip6.ip6_nxt; 1639 off += sizeof(ip6); 1640 return off; 1641 1642 case IPPROTO_FRAGMENT: 1643 /* 1644 * terminate parsing if it is not the first fragment, 1645 * it does not make sense to parse through it. 1646 */ 1647 if (m->m_pkthdr.len < off + sizeof(fh)) 1648 return -1; 1649 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1650 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1651 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1652 return -1; 1653 if (nxtp) 1654 *nxtp = fh.ip6f_nxt; 1655 off += sizeof(struct ip6_frag); 1656 return off; 1657 1658 case IPPROTO_AH: 1659 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1660 return -1; 1661 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1662 if (nxtp) 1663 *nxtp = ip6e.ip6e_nxt; 1664 off += (ip6e.ip6e_len + 2) << 2; 1665 return off; 1666 1667 case IPPROTO_HOPOPTS: 1668 case IPPROTO_ROUTING: 1669 case IPPROTO_DSTOPTS: 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 + 1) << 3; 1676 return off; 1677 1678 case IPPROTO_NONE: 1679 case IPPROTO_ESP: 1680 case IPPROTO_IPCOMP: 1681 /* give up */ 1682 return -1; 1683 1684 default: 1685 return -1; 1686 } 1687 1688 /* NOTREACHED */ 1689 } 1690 1691 /* 1692 * get offset for the last header in the chain. m will be kept untainted. 1693 */ 1694 int 1695 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1696 { 1697 int newoff; 1698 int nxt; 1699 1700 if (!nxtp) { 1701 nxt = -1; 1702 nxtp = &nxt; 1703 } 1704 while (1) { 1705 newoff = ip6_nexthdr(m, off, proto, nxtp); 1706 if (newoff < 0) 1707 return off; 1708 else if (newoff < off) 1709 return -1; /* invalid */ 1710 else if (newoff == off) 1711 return newoff; 1712 1713 off = newoff; 1714 proto = *nxtp; 1715 } 1716 } 1717 1718 /* 1719 * System control for IP6 1720 */ 1721 1722 u_char inet6ctlerrmap[PRC_NCMDS] = { 1723 0, 0, 0, 0, 1724 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1725 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1726 EMSGSIZE, EHOSTUNREACH, 0, 0, 1727 0, 0, EHOSTUNREACH, 0, 1728 ENOPROTOOPT, ECONNREFUSED 1729 }; 1730