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 #ifdef RSS 177 static struct netisr_handler ip6_direct_nh = { 178 .nh_name = "ip6_direct", 179 .nh_handler = ip6_direct_input, 180 .nh_proto = NETISR_IPV6_DIRECT, 181 .nh_m2cpuid = rss_soft_m2cpuid_v6, 182 .nh_policy = NETISR_POLICY_CPU, 183 .nh_dispatch = NETISR_DISPATCH_HYBRID, 184 }; 185 186 static int 187 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 188 { 189 int error, qlimit; 190 191 netisr_getqlimit(&ip6_direct_nh, &qlimit); 192 error = sysctl_handle_int(oidp, &qlimit, 0, req); 193 if (error || !req->newptr) 194 return (error); 195 if (qlimit < 1) 196 return (EINVAL); 197 return (netisr_setqlimit(&ip6_direct_nh, qlimit)); 198 } 199 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 200 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 201 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I", 202 "Maximum size of the IPv6 direct input queue"); 203 204 #endif 205 206 VNET_DEFINE(pfil_head_t, inet6_pfil_head); 207 VNET_DEFINE(pfil_head_t, inet6_local_pfil_head); 208 209 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 210 VNET_PCPUSTAT_SYSINIT(ip6stat); 211 #ifdef VIMAGE 212 VNET_PCPUSTAT_SYSUNINIT(ip6stat); 213 #endif /* VIMAGE */ 214 215 struct rmlock in6_ifaddr_lock; 216 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 217 218 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 219 220 /* 221 * IP6 initialization: fill in IP6 protocol switch table. 222 * All protocols not implemented in kernel go to raw IP6 protocol handler. 223 */ 224 static void 225 ip6_vnet_init(void *arg __unused) 226 { 227 struct pfil_head_args args; 228 229 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 230 &V_ip6_auto_linklocal); 231 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 232 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 233 234 CK_STAILQ_INIT(&V_in6_ifaddrhead); 235 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 236 &V_in6_ifaddrhmask); 237 238 /* Initialize packet filter hooks. */ 239 args.pa_version = PFIL_VERSION; 240 args.pa_flags = PFIL_IN | PFIL_OUT; 241 args.pa_type = PFIL_TYPE_IP6; 242 args.pa_headname = PFIL_INET6_NAME; 243 V_inet6_pfil_head = pfil_head_register(&args); 244 245 args.pa_flags = PFIL_OUT; 246 args.pa_headname = PFIL_INET6_LOCAL_NAME; 247 V_inet6_local_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, ip6proto_input_t input, 312 ip6proto_ctlinput_t ctl) 313 { 314 315 MPASS(proto > 0); 316 317 if (ip6_protox[proto] == rip6_input) { 318 ip6_protox[proto] = input; 319 ip6_ctlprotox[proto] = ctl; 320 return (0); 321 } else 322 return (EEXIST); 323 } 324 325 int 326 ip6proto_unregister(uint8_t proto) 327 { 328 329 MPASS(proto > 0); 330 331 if (ip6_protox[proto] != rip6_input) { 332 ip6_protox[proto] = rip6_input; 333 ip6_ctlprotox[proto] = rip6_ctlinput; 334 return (0); 335 } else 336 return (ENOENT); 337 } 338 339 #ifdef VIMAGE 340 static void 341 ip6_destroy(void *unused __unused) 342 { 343 struct ifaddr *ifa, *nifa; 344 struct ifnet *ifp; 345 int error; 346 347 #ifdef RSS 348 netisr_unregister_vnet(&ip6_direct_nh); 349 #endif 350 netisr_unregister_vnet(&ip6_nh); 351 352 pfil_head_unregister(V_inet6_pfil_head); 353 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 354 if (error != 0) { 355 printf("%s: WARNING: unable to deregister input helper hook " 356 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 357 "error %d returned\n", __func__, error); 358 } 359 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 360 if (error != 0) { 361 printf("%s: WARNING: unable to deregister output helper hook " 362 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 363 "error %d returned\n", __func__, error); 364 } 365 366 /* Cleanup addresses. */ 367 IFNET_RLOCK(); 368 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 369 /* Cannot lock here - lock recursion. */ 370 /* IF_ADDR_LOCK(ifp); */ 371 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 372 if (ifa->ifa_addr->sa_family != AF_INET6) 373 continue; 374 in6_purgeaddr(ifa); 375 } 376 /* IF_ADDR_UNLOCK(ifp); */ 377 in6_ifdetach_destroy(ifp); 378 mld_domifdetach(ifp); 379 } 380 IFNET_RUNLOCK(); 381 382 /* Make sure any routes are gone as well. */ 383 rib_flush_routes_family(AF_INET6); 384 385 frag6_destroy(); 386 nd6_destroy(); 387 in6_ifattach_destroy(); 388 389 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 390 } 391 392 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 393 #endif 394 395 static int 396 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off, 397 int *nxt, int *ours) 398 { 399 struct mbuf *m; 400 struct ip6_hdr *ip6; 401 struct ip6_hbh *hbh; 402 403 if (ip6_hopopts_input(plen, rtalert, mp, off)) { 404 #if 0 /*touches NULL pointer*/ 405 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard); 406 #endif 407 goto out; /* m have already been freed */ 408 } 409 410 /* adjust pointer */ 411 m = *mp; 412 ip6 = mtod(m, struct ip6_hdr *); 413 414 /* 415 * if the payload length field is 0 and the next header field 416 * indicates Hop-by-Hop Options header, then a Jumbo Payload 417 * option MUST be included. 418 */ 419 if (ip6->ip6_plen == 0 && *plen == 0) { 420 /* 421 * Note that if a valid jumbo payload option is 422 * contained, ip6_hopopts_input() must set a valid 423 * (non-zero) payload length to the variable plen. 424 */ 425 IP6STAT_INC(ip6s_badoptions); 426 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 427 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 428 icmp6_error(m, ICMP6_PARAM_PROB, 429 ICMP6_PARAMPROB_HEADER, 430 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 431 goto out; 432 } 433 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 434 hbh = (struct ip6_hbh *)(ip6 + 1); 435 *nxt = hbh->ip6h_nxt; 436 437 /* 438 * If we are acting as a router and the packet contains a 439 * router alert option, see if we know the option value. 440 * Currently, we only support the option value for MLD, in which 441 * case we should pass the packet to the multicast routing 442 * daemon. 443 */ 444 if (*rtalert != ~0) { 445 switch (*rtalert) { 446 case IP6OPT_RTALERT_MLD: 447 if (V_ip6_forwarding) 448 *ours = 1; 449 break; 450 default: 451 /* 452 * RFC2711 requires unrecognized values must be 453 * silently ignored. 454 */ 455 break; 456 } 457 } 458 459 return (0); 460 461 out: 462 return (1); 463 } 464 465 #ifdef RSS 466 /* 467 * IPv6 direct input routine. 468 * 469 * This is called when reinjecting completed fragments where 470 * all of the previous checking and book-keeping has been done. 471 */ 472 void 473 ip6_direct_input(struct mbuf *m) 474 { 475 int off, nxt; 476 int nest; 477 struct m_tag *mtag; 478 struct ip6_direct_ctx *ip6dc; 479 480 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 481 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 482 483 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 484 nxt = ip6dc->ip6dc_nxt; 485 off = ip6dc->ip6dc_off; 486 487 nest = 0; 488 489 m_tag_delete(m, mtag); 490 491 while (nxt != IPPROTO_DONE) { 492 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 493 IP6STAT_INC(ip6s_toomanyhdr); 494 goto bad; 495 } 496 497 /* 498 * protection against faulty packet - there should be 499 * more sanity checks in header chain processing. 500 */ 501 if (m->m_pkthdr.len < off) { 502 IP6STAT_INC(ip6s_tooshort); 503 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 504 goto bad; 505 } 506 507 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 508 if (IPSEC_ENABLED(ipv6)) { 509 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 510 return; 511 } 512 #endif /* IPSEC */ 513 514 nxt = ip6_protox[nxt](&m, &off, nxt); 515 } 516 return; 517 bad: 518 m_freem(m); 519 } 520 #endif 521 522 void 523 ip6_input(struct mbuf *m) 524 { 525 struct in6_addr odst; 526 struct ip6_hdr *ip6; 527 struct in6_ifaddr *ia; 528 struct ifnet *rcvif; 529 u_int32_t plen; 530 u_int32_t rtalert = ~0; 531 int off = sizeof(struct ip6_hdr), nest; 532 int nxt, ours = 0; 533 int srcrt = 0; 534 535 /* 536 * Drop the packet if IPv6 operation is disabled on the interface. 537 */ 538 rcvif = m->m_pkthdr.rcvif; 539 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 540 goto bad; 541 542 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 543 /* 544 * should the inner packet be considered authentic? 545 * see comment in ah4_input(). 546 * NB: m cannot be NULL when passed to the input routine 547 */ 548 549 m->m_flags &= ~M_AUTHIPHDR; 550 m->m_flags &= ~M_AUTHIPDGM; 551 552 #endif /* IPSEC */ 553 554 if (m->m_flags & M_FASTFWD_OURS) { 555 /* 556 * Firewall changed destination to local. 557 */ 558 ip6 = mtod(m, struct ip6_hdr *); 559 goto passin; 560 } 561 562 /* 563 * mbuf statistics 564 */ 565 if (m->m_flags & M_EXT) { 566 if (m->m_next) 567 IP6STAT_INC(ip6s_mext2m); 568 else 569 IP6STAT_INC(ip6s_mext1); 570 } else { 571 if (m->m_next) { 572 struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif; 573 int ifindex = ifp->if_index; 574 if (ifindex >= IP6S_M2MMAX) 575 ifindex = 0; 576 IP6STAT_INC(ip6s_m2m[ifindex]); 577 } else 578 IP6STAT_INC(ip6s_m1); 579 } 580 581 in6_ifstat_inc(rcvif, ifs6_in_receive); 582 IP6STAT_INC(ip6s_total); 583 584 /* 585 * L2 bridge code and some other code can return mbuf chain 586 * that does not conform to KAME requirement. too bad. 587 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 588 */ 589 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 590 struct mbuf *n; 591 592 if (m->m_pkthdr.len > MHLEN) 593 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 594 else 595 n = m_gethdr(M_NOWAIT, MT_DATA); 596 if (n == NULL) 597 goto bad; 598 599 m_move_pkthdr(n, m); 600 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 601 n->m_len = n->m_pkthdr.len; 602 m_freem(m); 603 m = n; 604 } 605 if (m->m_len < sizeof(struct ip6_hdr)) { 606 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 607 IP6STAT_INC(ip6s_toosmall); 608 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 609 goto bad; 610 } 611 } 612 613 ip6 = mtod(m, struct ip6_hdr *); 614 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 615 IP6STAT_INC(ip6s_badvers); 616 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 617 goto bad; 618 } 619 620 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 621 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 622 623 /* 624 * Check against address spoofing/corruption. 625 */ 626 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 627 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 628 /* 629 * XXX: "badscope" is not very suitable for a multicast source. 630 */ 631 IP6STAT_INC(ip6s_badscope); 632 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 633 goto bad; 634 } 635 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 636 !(m->m_flags & M_LOOP)) { 637 /* 638 * In this case, the packet should come from the loopback 639 * interface. However, we cannot just check the if_flags, 640 * because ip6_mloopback() passes the "actual" interface 641 * as the outgoing/incoming interface. 642 */ 643 IP6STAT_INC(ip6s_badscope); 644 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 645 goto bad; 646 } 647 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 648 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 649 /* 650 * RFC4291 2.7: 651 * Nodes must not originate a packet to a multicast address 652 * whose scop field contains the reserved value 0; if such 653 * a packet is received, it must be silently dropped. 654 */ 655 IP6STAT_INC(ip6s_badscope); 656 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 657 goto bad; 658 } 659 /* 660 * The following check is not documented in specs. A malicious 661 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 662 * and bypass security checks (act as if it was from 127.0.0.1 by using 663 * IPv6 src ::ffff:127.0.0.1). Be cautious. 664 * 665 * We have supported IPv6-only kernels for a few years and this issue 666 * has not come up. The world seems to move mostly towards not using 667 * v4mapped on the wire, so it makes sense for us to keep rejecting 668 * any such packets. 669 */ 670 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 671 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 672 IP6STAT_INC(ip6s_badscope); 673 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 674 goto bad; 675 } 676 #if 0 677 /* 678 * Reject packets with IPv4 compatible addresses (auto tunnel). 679 * 680 * The code forbids auto tunnel relay case in RFC1933 (the check is 681 * stronger than RFC1933). We may want to re-enable it if mech-xx 682 * is revised to forbid relaying case. 683 */ 684 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 685 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 686 IP6STAT_INC(ip6s_badscope); 687 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 688 goto bad; 689 } 690 #endif 691 /* 692 * Try to forward the packet, but if we fail continue. 693 * ip6_tryforward() does not generate redirects, so fall 694 * through to normal processing if redirects are required. 695 * ip6_tryforward() does inbound and outbound packet firewall 696 * processing. If firewall has decided that destination becomes 697 * our local address, it sets M_FASTFWD_OURS flag. In this 698 * case skip another inbound firewall processing and update 699 * ip6 pointer. 700 */ 701 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0 702 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 703 && (!IPSEC_ENABLED(ipv6) || 704 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0) 705 #endif 706 ) { 707 if ((m = ip6_tryforward(m)) == NULL) 708 return; 709 if (m->m_flags & M_FASTFWD_OURS) { 710 ip6 = mtod(m, struct ip6_hdr *); 711 goto passin; 712 } 713 } 714 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 715 /* 716 * Bypass packet filtering for packets previously handled by IPsec. 717 */ 718 if (IPSEC_ENABLED(ipv6) && 719 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0) 720 goto passin; 721 #endif 722 /* 723 * Run through list of hooks for input packets. 724 * 725 * NB: Beware of the destination address changing 726 * (e.g. by NAT rewriting). When this happens, 727 * tell ip6_forward to do the right thing. 728 */ 729 730 /* Jump over all PFIL processing if hooks are not active. */ 731 if (!PFIL_HOOKED_IN(V_inet6_pfil_head)) 732 goto passin; 733 734 odst = ip6->ip6_dst; 735 if (pfil_mbuf_in(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif, 736 NULL) != PFIL_PASS) 737 return; 738 ip6 = mtod(m, struct ip6_hdr *); 739 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 740 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 741 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 742 /* 743 * Directly ship the packet on. This allows forwarding 744 * packets originally destined to us to some other directly 745 * connected host. 746 */ 747 ip6_forward(m, 1); 748 return; 749 } 750 751 passin: 752 /* 753 * Disambiguate address scope zones (if there is ambiguity). 754 * We first make sure that the original source or destination address 755 * is not in our internal form for scoped addresses. Such addresses 756 * are not necessarily invalid spec-wise, but we cannot accept them due 757 * to the usage conflict. 758 * in6_setscope() then also checks and rejects the cases where src or 759 * dst are the loopback address and the receiving interface 760 * is not loopback. 761 */ 762 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 763 IP6STAT_INC(ip6s_badscope); /* XXX */ 764 goto bad; 765 } 766 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 767 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 768 IP6STAT_INC(ip6s_badscope); 769 goto bad; 770 } 771 if (m->m_flags & M_FASTFWD_OURS) { 772 m->m_flags &= ~M_FASTFWD_OURS; 773 ours = 1; 774 goto hbhcheck; 775 } 776 /* 777 * Multicast check. Assume packet is for us to avoid 778 * prematurely taking locks. 779 */ 780 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 781 ours = 1; 782 in6_ifstat_inc(rcvif, ifs6_in_mcast); 783 goto hbhcheck; 784 } 785 /* 786 * Unicast check 787 * XXX: For now we keep link-local IPv6 addresses with embedded 788 * scope zone id, therefore we use zero zoneid here. 789 */ 790 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false); 791 if (ia != NULL) { 792 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 793 char ip6bufs[INET6_ADDRSTRLEN]; 794 char ip6bufd[INET6_ADDRSTRLEN]; 795 /* address is not ready, so discard the packet. */ 796 nd6log((LOG_INFO, 797 "ip6_input: packet to an unready address %s->%s\n", 798 ip6_sprintf(ip6bufs, &ip6->ip6_src), 799 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 800 goto bad; 801 } 802 if (V_ip6_sav && !(m->m_flags & M_LOOP) && 803 __predict_false(in6_localip_fib(&ip6->ip6_src, 804 rcvif->if_fib))) { 805 IP6STAT_INC(ip6s_badscope); /* XXX */ 806 goto bad; 807 } 808 /* Count the packet in the ip address stats */ 809 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 810 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 811 ours = 1; 812 goto hbhcheck; 813 } 814 815 /* 816 * Now there is no reason to process the packet if it's not our own 817 * and we're not a router. 818 */ 819 if (!V_ip6_forwarding) { 820 IP6STAT_INC(ip6s_cantforward); 821 goto bad; 822 } 823 824 hbhcheck: 825 /* 826 * Process Hop-by-Hop options header if it's contained. 827 * m may be modified in ip6_hopopts_input(). 828 * If a JumboPayload option is included, plen will also be modified. 829 */ 830 plen = (u_int32_t)ntohs(ip6->ip6_plen); 831 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 832 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0) 833 return; 834 } else 835 nxt = ip6->ip6_nxt; 836 837 /* 838 * Use mbuf flags to propagate Router Alert option to 839 * ICMPv6 layer, as hop-by-hop options have been stripped. 840 */ 841 if (rtalert != ~0) 842 m->m_flags |= M_RTALERT_MLD; 843 844 /* 845 * Check that the amount of data in the buffers 846 * is as at least much as the IPv6 header would have us expect. 847 * Trim mbufs if longer than we expect. 848 * Drop packet if shorter than we expect. 849 */ 850 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 851 IP6STAT_INC(ip6s_tooshort); 852 in6_ifstat_inc(rcvif, ifs6_in_truncated); 853 goto bad; 854 } 855 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 856 if (m->m_len == m->m_pkthdr.len) { 857 m->m_len = sizeof(struct ip6_hdr) + plen; 858 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 859 } else 860 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 861 } 862 863 /* 864 * Forward if desirable. 865 */ 866 if (V_ip6_mrouter && 867 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 868 /* 869 * If we are acting as a multicast router, all 870 * incoming multicast packets are passed to the 871 * kernel-level multicast forwarding function. 872 * The packet is returned (relatively) intact; if 873 * ip6_mforward() returns a non-zero value, the packet 874 * must be discarded, else it may be accepted below. 875 * 876 * XXX TODO: Check hlim and multicast scope here to avoid 877 * unnecessarily calling into ip6_mforward(). 878 */ 879 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 880 IP6STAT_INC(ip6s_cantforward); 881 goto bad; 882 } 883 } else if (!ours) { 884 ip6_forward(m, srcrt); 885 return; 886 } 887 888 /* 889 * We are going to ship the packet to the local protocol stack. Call the 890 * filter again for this 'output' action, allowing redirect-like rules 891 * to adjust the source address. 892 */ 893 if (PFIL_HOOKED_OUT(V_inet6_local_pfil_head)) { 894 if (pfil_mbuf_out(V_inet6_local_pfil_head, &m, V_loif, NULL) != 895 PFIL_PASS) 896 return; 897 if (m == NULL) /* consumed by filter */ 898 return; 899 ip6 = mtod(m, struct ip6_hdr *); 900 } 901 902 /* 903 * Tell launch routine the next header 904 */ 905 IP6STAT_INC(ip6s_delivered); 906 in6_ifstat_inc(rcvif, ifs6_in_deliver); 907 nest = 0; 908 909 while (nxt != IPPROTO_DONE) { 910 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 911 IP6STAT_INC(ip6s_toomanyhdr); 912 goto bad; 913 } 914 915 /* 916 * protection against faulty packet - there should be 917 * more sanity checks in header chain processing. 918 */ 919 if (m->m_pkthdr.len < off) { 920 IP6STAT_INC(ip6s_tooshort); 921 in6_ifstat_inc(rcvif, ifs6_in_truncated); 922 goto bad; 923 } 924 925 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 926 if (IPSEC_ENABLED(ipv6)) { 927 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 928 return; 929 } 930 #endif /* IPSEC */ 931 932 nxt = ip6_protox[nxt](&m, &off, nxt); 933 } 934 return; 935 bad: 936 in6_ifstat_inc(rcvif, ifs6_in_discard); 937 if (m != NULL) 938 m_freem(m); 939 } 940 941 /* 942 * Hop-by-Hop options header processing. If a valid jumbo payload option is 943 * included, the real payload length will be stored in plenp. 944 * 945 * rtalertp - XXX: should be stored more smart way 946 */ 947 static int 948 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 949 struct mbuf **mp, int *offp) 950 { 951 struct mbuf *m = *mp; 952 int off = *offp, hbhlen; 953 struct ip6_hbh *hbh; 954 955 /* validation of the length of the header */ 956 if (m->m_len < off + sizeof(*hbh)) { 957 m = m_pullup(m, off + sizeof(*hbh)); 958 if (m == NULL) { 959 IP6STAT_INC(ip6s_exthdrtoolong); 960 *mp = NULL; 961 return (-1); 962 } 963 } 964 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 965 hbhlen = (hbh->ip6h_len + 1) << 3; 966 967 if (m->m_len < off + hbhlen) { 968 m = m_pullup(m, off + hbhlen); 969 if (m == NULL) { 970 IP6STAT_INC(ip6s_exthdrtoolong); 971 *mp = NULL; 972 return (-1); 973 } 974 } 975 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 976 off += hbhlen; 977 hbhlen -= sizeof(struct ip6_hbh); 978 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 979 hbhlen, rtalertp, plenp) < 0) { 980 *mp = NULL; 981 return (-1); 982 } 983 984 *offp = off; 985 *mp = m; 986 return (0); 987 } 988 989 /* 990 * Search header for all Hop-by-hop options and process each option. 991 * This function is separate from ip6_hopopts_input() in order to 992 * handle a case where the sending node itself process its hop-by-hop 993 * options header. In such a case, the function is called from ip6_output(). 994 * 995 * The function assumes that hbh header is located right after the IPv6 header 996 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 997 * opthead + hbhlen is located in contiguous memory region. 998 */ 999 int 1000 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1001 u_int32_t *rtalertp, u_int32_t *plenp) 1002 { 1003 struct ip6_hdr *ip6; 1004 int optlen = 0; 1005 u_int8_t *opt = opthead; 1006 u_int16_t rtalert_val; 1007 u_int32_t jumboplen; 1008 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1009 1010 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1011 switch (*opt) { 1012 case IP6OPT_PAD1: 1013 optlen = 1; 1014 break; 1015 case IP6OPT_PADN: 1016 if (hbhlen < IP6OPT_MINLEN) { 1017 IP6STAT_INC(ip6s_toosmall); 1018 goto bad; 1019 } 1020 optlen = *(opt + 1) + 2; 1021 break; 1022 case IP6OPT_ROUTER_ALERT: 1023 /* XXX may need check for alignment */ 1024 if (hbhlen < IP6OPT_RTALERT_LEN) { 1025 IP6STAT_INC(ip6s_toosmall); 1026 goto bad; 1027 } 1028 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1029 /* XXX stat */ 1030 icmp6_error(m, ICMP6_PARAM_PROB, 1031 ICMP6_PARAMPROB_HEADER, 1032 erroff + opt + 1 - opthead); 1033 return (-1); 1034 } 1035 optlen = IP6OPT_RTALERT_LEN; 1036 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1037 *rtalertp = ntohs(rtalert_val); 1038 break; 1039 case IP6OPT_JUMBO: 1040 /* XXX may need check for alignment */ 1041 if (hbhlen < IP6OPT_JUMBO_LEN) { 1042 IP6STAT_INC(ip6s_toosmall); 1043 goto bad; 1044 } 1045 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1046 /* XXX stat */ 1047 icmp6_error(m, ICMP6_PARAM_PROB, 1048 ICMP6_PARAMPROB_HEADER, 1049 erroff + opt + 1 - opthead); 1050 return (-1); 1051 } 1052 optlen = IP6OPT_JUMBO_LEN; 1053 1054 /* 1055 * IPv6 packets that have non 0 payload length 1056 * must not contain a jumbo payload option. 1057 */ 1058 ip6 = mtod(m, struct ip6_hdr *); 1059 if (ip6->ip6_plen) { 1060 IP6STAT_INC(ip6s_badoptions); 1061 icmp6_error(m, ICMP6_PARAM_PROB, 1062 ICMP6_PARAMPROB_HEADER, 1063 erroff + opt - opthead); 1064 return (-1); 1065 } 1066 1067 /* 1068 * We may see jumbolen in unaligned location, so 1069 * we'd need to perform bcopy(). 1070 */ 1071 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1072 jumboplen = (u_int32_t)htonl(jumboplen); 1073 1074 #if 1 1075 /* 1076 * if there are multiple jumbo payload options, 1077 * *plenp will be non-zero and the packet will be 1078 * rejected. 1079 * the behavior may need some debate in ipngwg - 1080 * multiple options does not make sense, however, 1081 * there's no explicit mention in specification. 1082 */ 1083 if (*plenp != 0) { 1084 IP6STAT_INC(ip6s_badoptions); 1085 icmp6_error(m, ICMP6_PARAM_PROB, 1086 ICMP6_PARAMPROB_HEADER, 1087 erroff + opt + 2 - opthead); 1088 return (-1); 1089 } 1090 #endif 1091 1092 /* 1093 * jumbo payload length must be larger than 65535. 1094 */ 1095 if (jumboplen <= IPV6_MAXPACKET) { 1096 IP6STAT_INC(ip6s_badoptions); 1097 icmp6_error(m, ICMP6_PARAM_PROB, 1098 ICMP6_PARAMPROB_HEADER, 1099 erroff + opt + 2 - opthead); 1100 return (-1); 1101 } 1102 *plenp = jumboplen; 1103 1104 break; 1105 default: /* unknown option */ 1106 if (hbhlen < IP6OPT_MINLEN) { 1107 IP6STAT_INC(ip6s_toosmall); 1108 goto bad; 1109 } 1110 optlen = ip6_unknown_opt(opt, m, 1111 erroff + opt - opthead); 1112 if (optlen == -1) 1113 return (-1); 1114 optlen += 2; 1115 break; 1116 } 1117 } 1118 1119 return (0); 1120 1121 bad: 1122 m_freem(m); 1123 return (-1); 1124 } 1125 1126 /* 1127 * Unknown option processing. 1128 * The third argument `off' is the offset from the IPv6 header to the option, 1129 * which is necessary if the IPv6 header the and option header and IPv6 header 1130 * is not contiguous in order to return an ICMPv6 error. 1131 */ 1132 int 1133 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1134 { 1135 struct ip6_hdr *ip6; 1136 1137 switch (IP6OPT_TYPE(*optp)) { 1138 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1139 return ((int)*(optp + 1)); 1140 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1141 m_freem(m); 1142 return (-1); 1143 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1144 IP6STAT_INC(ip6s_badoptions); 1145 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1146 return (-1); 1147 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1148 IP6STAT_INC(ip6s_badoptions); 1149 ip6 = mtod(m, struct ip6_hdr *); 1150 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1151 (m->m_flags & (M_BCAST|M_MCAST))) 1152 m_freem(m); 1153 else 1154 icmp6_error(m, ICMP6_PARAM_PROB, 1155 ICMP6_PARAMPROB_OPTION, off); 1156 return (-1); 1157 } 1158 1159 m_freem(m); /* XXX: NOTREACHED */ 1160 return (-1); 1161 } 1162 1163 /* 1164 * Create the "control" list for this pcb. 1165 * These functions will not modify mbuf chain at all. 1166 * 1167 * The routine will be called from upper layer handlers like tcp6_input(). 1168 * Thus the routine assumes that the caller (tcp6_input) have already 1169 * called m_pullup() and all the extension headers are located in the 1170 * very first mbuf on the mbuf chain. 1171 * 1172 * ip6_savecontrol_v4 will handle those options that are possible to be 1173 * set on a v4-mapped socket. 1174 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1175 * options and handle the v6-only ones itself. 1176 */ 1177 struct mbuf ** 1178 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1179 int *v4only) 1180 { 1181 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1182 1183 #ifdef SO_TIMESTAMP 1184 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1185 union { 1186 struct timeval tv; 1187 struct bintime bt; 1188 struct timespec ts; 1189 } t; 1190 struct bintime boottimebin, bt1; 1191 struct timespec ts1; 1192 bool stamped; 1193 1194 stamped = false; 1195 switch (inp->inp_socket->so_ts_clock) { 1196 case SO_TS_REALTIME_MICRO: 1197 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1198 M_TSTMP)) { 1199 mbuf_tstmp2timespec(m, &ts1); 1200 timespec2bintime(&ts1, &bt1); 1201 getboottimebin(&boottimebin); 1202 bintime_add(&bt1, &boottimebin); 1203 bintime2timeval(&bt1, &t.tv); 1204 } else { 1205 microtime(&t.tv); 1206 } 1207 *mp = sbcreatecontrol(&t.tv, sizeof(t.tv), 1208 SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT); 1209 if (*mp != NULL) { 1210 mp = &(*mp)->m_next; 1211 stamped = true; 1212 } 1213 break; 1214 1215 case SO_TS_BINTIME: 1216 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1217 M_TSTMP)) { 1218 mbuf_tstmp2timespec(m, &ts1); 1219 timespec2bintime(&ts1, &t.bt); 1220 getboottimebin(&boottimebin); 1221 bintime_add(&t.bt, &boottimebin); 1222 } else { 1223 bintime(&t.bt); 1224 } 1225 *mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME, 1226 SOL_SOCKET, M_NOWAIT); 1227 if (*mp != NULL) { 1228 mp = &(*mp)->m_next; 1229 stamped = true; 1230 } 1231 break; 1232 1233 case SO_TS_REALTIME: 1234 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1235 M_TSTMP)) { 1236 mbuf_tstmp2timespec(m, &t.ts); 1237 getboottimebin(&boottimebin); 1238 bintime2timespec(&boottimebin, &ts1); 1239 timespecadd(&t.ts, &ts1, &t.ts); 1240 } else { 1241 nanotime(&t.ts); 1242 } 1243 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1244 SCM_REALTIME, SOL_SOCKET, M_NOWAIT); 1245 if (*mp != NULL) { 1246 mp = &(*mp)->m_next; 1247 stamped = true; 1248 } 1249 break; 1250 1251 case SO_TS_MONOTONIC: 1252 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1253 M_TSTMP)) 1254 mbuf_tstmp2timespec(m, &t.ts); 1255 else 1256 nanouptime(&t.ts); 1257 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1258 SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT); 1259 if (*mp != NULL) { 1260 mp = &(*mp)->m_next; 1261 stamped = true; 1262 } 1263 break; 1264 1265 default: 1266 panic("unknown (corrupted) so_ts_clock"); 1267 } 1268 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == 1269 (M_PKTHDR | M_TSTMP)) { 1270 struct sock_timestamp_info sti; 1271 1272 bzero(&sti, sizeof(sti)); 1273 sti.st_info_flags = ST_INFO_HW; 1274 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1275 sti.st_info_flags |= ST_INFO_HW_HPREC; 1276 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO, 1277 SOL_SOCKET, M_NOWAIT); 1278 if (*mp != NULL) 1279 mp = &(*mp)->m_next; 1280 } 1281 } 1282 #endif 1283 1284 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1285 /* RFC 2292 sec. 5 */ 1286 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1287 struct in6_pktinfo pi6; 1288 1289 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1290 #ifdef INET 1291 struct ip *ip; 1292 1293 ip = mtod(m, struct ip *); 1294 pi6.ipi6_addr.s6_addr32[0] = 0; 1295 pi6.ipi6_addr.s6_addr32[1] = 0; 1296 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1297 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1298 #else 1299 /* We won't hit this code */ 1300 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1301 #endif 1302 } else { 1303 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1304 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1305 } 1306 pi6.ipi6_ifindex = 1307 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1308 1309 *mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo), 1310 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6, 1311 M_NOWAIT); 1312 if (*mp) 1313 mp = &(*mp)->m_next; 1314 } 1315 1316 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1317 int hlim; 1318 1319 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1320 #ifdef INET 1321 struct ip *ip; 1322 1323 ip = mtod(m, struct ip *); 1324 hlim = ip->ip_ttl; 1325 #else 1326 /* We won't hit this code */ 1327 hlim = 0; 1328 #endif 1329 } else { 1330 hlim = ip6->ip6_hlim & 0xff; 1331 } 1332 *mp = sbcreatecontrol(&hlim, sizeof(int), 1333 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1334 IPPROTO_IPV6, M_NOWAIT); 1335 if (*mp) 1336 mp = &(*mp)->m_next; 1337 } 1338 1339 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1340 int tclass; 1341 1342 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1343 #ifdef INET 1344 struct ip *ip; 1345 1346 ip = mtod(m, struct ip *); 1347 tclass = ip->ip_tos; 1348 #else 1349 /* We won't hit this code */ 1350 tclass = 0; 1351 #endif 1352 } else { 1353 u_int32_t flowinfo; 1354 1355 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1356 flowinfo >>= 20; 1357 tclass = flowinfo & 0xff; 1358 } 1359 *mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS, 1360 IPPROTO_IPV6, M_NOWAIT); 1361 if (*mp) 1362 mp = &(*mp)->m_next; 1363 } 1364 1365 if (v4only != NULL) { 1366 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1367 *v4only = 1; 1368 } else { 1369 *v4only = 0; 1370 } 1371 } 1372 1373 return (mp); 1374 } 1375 1376 void 1377 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp) 1378 { 1379 struct ip6_hdr *ip6; 1380 int v4only = 0; 1381 1382 mp = ip6_savecontrol_v4(inp, m, mp, &v4only); 1383 if (v4only) 1384 return; 1385 1386 ip6 = mtod(m, struct ip6_hdr *); 1387 /* 1388 * IPV6_HOPOPTS socket option. Recall that we required super-user 1389 * privilege for the option (see ip6_ctloutput), but it might be too 1390 * strict, since there might be some hop-by-hop options which can be 1391 * returned to normal user. 1392 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1393 */ 1394 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) { 1395 /* 1396 * Check if a hop-by-hop options header is contatined in the 1397 * received packet, and if so, store the options as ancillary 1398 * data. Note that a hop-by-hop options header must be 1399 * just after the IPv6 header, which is assured through the 1400 * IPv6 input processing. 1401 */ 1402 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1403 struct ip6_hbh *hbh; 1404 u_int hbhlen; 1405 1406 hbh = (struct ip6_hbh *)(ip6 + 1); 1407 hbhlen = (hbh->ip6h_len + 1) << 3; 1408 1409 /* 1410 * XXX: We copy the whole header even if a 1411 * jumbo payload option is included, the option which 1412 * is to be removed before returning according to 1413 * RFC2292. 1414 * Note: this constraint is removed in RFC3542 1415 */ 1416 *mp = sbcreatecontrol(hbh, hbhlen, 1417 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1418 IPPROTO_IPV6, M_NOWAIT); 1419 if (*mp) 1420 mp = &(*mp)->m_next; 1421 } 1422 } 1423 1424 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1425 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1426 1427 /* 1428 * Search for destination options headers or routing 1429 * header(s) through the header chain, and stores each 1430 * header as ancillary data. 1431 * Note that the order of the headers remains in 1432 * the chain of ancillary data. 1433 */ 1434 while (1) { /* is explicit loop prevention necessary? */ 1435 struct ip6_ext *ip6e = NULL; 1436 u_int elen; 1437 1438 /* 1439 * if it is not an extension header, don't try to 1440 * pull it from the chain. 1441 */ 1442 switch (nxt) { 1443 case IPPROTO_DSTOPTS: 1444 case IPPROTO_ROUTING: 1445 case IPPROTO_HOPOPTS: 1446 case IPPROTO_AH: /* is it possible? */ 1447 break; 1448 default: 1449 goto loopend; 1450 } 1451 1452 if (off + sizeof(*ip6e) > m->m_len) 1453 goto loopend; 1454 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1455 if (nxt == IPPROTO_AH) 1456 elen = (ip6e->ip6e_len + 2) << 2; 1457 else 1458 elen = (ip6e->ip6e_len + 1) << 3; 1459 if (off + elen > m->m_len) 1460 goto loopend; 1461 1462 switch (nxt) { 1463 case IPPROTO_DSTOPTS: 1464 if (!(inp->inp_flags & IN6P_DSTOPTS)) 1465 break; 1466 1467 *mp = sbcreatecontrol(ip6e, elen, 1468 IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1469 IPPROTO_IPV6, M_NOWAIT); 1470 if (*mp) 1471 mp = &(*mp)->m_next; 1472 break; 1473 case IPPROTO_ROUTING: 1474 if (!(inp->inp_flags & IN6P_RTHDR)) 1475 break; 1476 1477 *mp = sbcreatecontrol(ip6e, elen, 1478 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR), 1479 IPPROTO_IPV6, M_NOWAIT); 1480 if (*mp) 1481 mp = &(*mp)->m_next; 1482 break; 1483 case IPPROTO_HOPOPTS: 1484 case IPPROTO_AH: /* is it possible? */ 1485 break; 1486 1487 default: 1488 /* 1489 * other cases have been filtered in the above. 1490 * none will visit this case. here we supply 1491 * the code just in case (nxt overwritten or 1492 * other cases). 1493 */ 1494 goto loopend; 1495 } 1496 1497 /* proceed with the next header. */ 1498 off += elen; 1499 nxt = ip6e->ip6e_nxt; 1500 ip6e = NULL; 1501 } 1502 loopend: 1503 ; 1504 } 1505 1506 if (inp->inp_flags2 & INP_RECVFLOWID) { 1507 uint32_t flowid, flow_type; 1508 1509 flowid = m->m_pkthdr.flowid; 1510 flow_type = M_HASHTYPE_GET(m); 1511 1512 /* 1513 * XXX should handle the failure of one or the 1514 * other - don't populate both? 1515 */ 1516 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID, 1517 IPPROTO_IPV6, M_NOWAIT); 1518 if (*mp) 1519 mp = &(*mp)->m_next; 1520 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t), 1521 IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT); 1522 if (*mp) 1523 mp = &(*mp)->m_next; 1524 } 1525 1526 #ifdef RSS 1527 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1528 uint32_t flowid, flow_type; 1529 uint32_t rss_bucketid; 1530 1531 flowid = m->m_pkthdr.flowid; 1532 flow_type = M_HASHTYPE_GET(m); 1533 1534 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1535 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t), 1536 IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT); 1537 if (*mp) 1538 mp = &(*mp)->m_next; 1539 } 1540 } 1541 #endif 1542 1543 } 1544 #undef IS2292 1545 1546 void 1547 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1548 { 1549 struct socket *so; 1550 struct mbuf *m_mtu; 1551 struct ip6_mtuinfo mtuctl; 1552 1553 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1554 /* 1555 * Notify the error by sending IPV6_PATHMTU ancillary data if 1556 * application wanted to know the MTU value. 1557 * NOTE: we notify disconnected sockets, because some udp 1558 * applications keep sending sockets disconnected. 1559 * NOTE: our implementation doesn't notify connected sockets that has 1560 * foreign address that is different than given destination addresses 1561 * (this is permitted by RFC 3542). 1562 */ 1563 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1564 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1565 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1566 return; 1567 1568 mtuctl.ip6m_mtu = mtu; 1569 mtuctl.ip6m_addr = *dst; 1570 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1571 return; 1572 1573 if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU, 1574 IPPROTO_IPV6, M_NOWAIT)) == NULL) 1575 return; 1576 1577 so = inp->inp_socket; 1578 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1579 == 0) { 1580 soroverflow(so); 1581 m_freem(m_mtu); 1582 /* XXX: should count statistics */ 1583 } else 1584 sorwakeup(so); 1585 } 1586 1587 /* 1588 * Get pointer to the previous header followed by the header 1589 * currently processed. 1590 */ 1591 int 1592 ip6_get_prevhdr(const struct mbuf *m, int off) 1593 { 1594 struct ip6_ext ip6e; 1595 struct ip6_hdr *ip6; 1596 int len, nlen, nxt; 1597 1598 if (off == sizeof(struct ip6_hdr)) 1599 return (offsetof(struct ip6_hdr, ip6_nxt)); 1600 if (off < sizeof(struct ip6_hdr)) 1601 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1602 1603 ip6 = mtod(m, struct ip6_hdr *); 1604 nxt = ip6->ip6_nxt; 1605 len = sizeof(struct ip6_hdr); 1606 nlen = 0; 1607 while (len < off) { 1608 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1609 switch (nxt) { 1610 case IPPROTO_FRAGMENT: 1611 nlen = sizeof(struct ip6_frag); 1612 break; 1613 case IPPROTO_AH: 1614 nlen = (ip6e.ip6e_len + 2) << 2; 1615 break; 1616 default: 1617 nlen = (ip6e.ip6e_len + 1) << 3; 1618 } 1619 len += nlen; 1620 nxt = ip6e.ip6e_nxt; 1621 } 1622 return (len - nlen); 1623 } 1624 1625 /* 1626 * get next header offset. m will be retained. 1627 */ 1628 int 1629 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1630 { 1631 struct ip6_hdr ip6; 1632 struct ip6_ext ip6e; 1633 struct ip6_frag fh; 1634 1635 /* just in case */ 1636 if (m == NULL) 1637 panic("ip6_nexthdr: m == NULL"); 1638 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1639 return -1; 1640 1641 switch (proto) { 1642 case IPPROTO_IPV6: 1643 if (m->m_pkthdr.len < off + sizeof(ip6)) 1644 return -1; 1645 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1646 if (nxtp) 1647 *nxtp = ip6.ip6_nxt; 1648 off += sizeof(ip6); 1649 return off; 1650 1651 case IPPROTO_FRAGMENT: 1652 /* 1653 * terminate parsing if it is not the first fragment, 1654 * it does not make sense to parse through it. 1655 */ 1656 if (m->m_pkthdr.len < off + sizeof(fh)) 1657 return -1; 1658 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1659 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1660 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1661 return -1; 1662 if (nxtp) 1663 *nxtp = fh.ip6f_nxt; 1664 off += sizeof(struct ip6_frag); 1665 return off; 1666 1667 case IPPROTO_AH: 1668 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1669 return -1; 1670 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1671 if (nxtp) 1672 *nxtp = ip6e.ip6e_nxt; 1673 off += (ip6e.ip6e_len + 2) << 2; 1674 return off; 1675 1676 case IPPROTO_HOPOPTS: 1677 case IPPROTO_ROUTING: 1678 case IPPROTO_DSTOPTS: 1679 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1680 return -1; 1681 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1682 if (nxtp) 1683 *nxtp = ip6e.ip6e_nxt; 1684 off += (ip6e.ip6e_len + 1) << 3; 1685 return off; 1686 1687 case IPPROTO_NONE: 1688 case IPPROTO_ESP: 1689 case IPPROTO_IPCOMP: 1690 /* give up */ 1691 return -1; 1692 1693 default: 1694 return -1; 1695 } 1696 1697 /* NOTREACHED */ 1698 } 1699 1700 /* 1701 * get offset for the last header in the chain. m will be kept untainted. 1702 */ 1703 int 1704 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1705 { 1706 int newoff; 1707 int nxt; 1708 1709 if (!nxtp) { 1710 nxt = -1; 1711 nxtp = &nxt; 1712 } 1713 while (1) { 1714 newoff = ip6_nexthdr(m, off, proto, nxtp); 1715 if (newoff < 0) 1716 return off; 1717 else if (newoff < off) 1718 return -1; /* invalid */ 1719 else if (newoff == off) 1720 return newoff; 1721 1722 off = newoff; 1723 proto = *nxtp; 1724 } 1725 } 1726