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_private.h> 99 #include <net/if_dl.h> 100 #include <net/route.h> 101 #include <net/netisr.h> 102 #include <net/rss_config.h> 103 #include <net/pfil.h> 104 #include <net/vnet.h> 105 106 #include <netinet/in.h> 107 #include <netinet/in_kdtrace.h> 108 #include <netinet/ip_var.h> 109 #include <netinet/in_systm.h> 110 #include <net/if_llatbl.h> 111 #ifdef INET 112 #include <netinet/ip.h> 113 #include <netinet/ip_icmp.h> 114 #endif /* INET */ 115 #include <netinet/ip6.h> 116 #include <netinet6/in6_var.h> 117 #include <netinet6/ip6_var.h> 118 #include <netinet/ip_encap.h> 119 #include <netinet/in_pcb.h> 120 #include <netinet/icmp6.h> 121 #include <netinet6/scope6_var.h> 122 #include <netinet6/in6_ifattach.h> 123 #include <netinet6/mld6_var.h> 124 #include <netinet6/nd6.h> 125 #include <netinet6/in6_rss.h> 126 #ifdef SCTP 127 #include <netinet/sctp_pcb.h> 128 #include <netinet6/sctp6_var.h> 129 #endif 130 131 #include <netipsec/ipsec_support.h> 132 133 ip6proto_input_t *ip6_protox[IPPROTO_MAX] = { 134 [0 ... IPPROTO_MAX - 1] = rip6_input }; 135 ip6proto_ctlinput_t *ip6_ctlprotox[IPPROTO_MAX] = { 136 [0 ... IPPROTO_MAX - 1] = rip6_ctlinput }; 137 138 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 139 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); 140 VNET_DEFINE(u_long, in6_ifaddrhmask); 141 142 static struct netisr_handler ip6_nh = { 143 .nh_name = "ip6", 144 .nh_handler = ip6_input, 145 .nh_proto = NETISR_IPV6, 146 #ifdef RSS 147 .nh_m2cpuid = rss_soft_m2cpuid_v6, 148 .nh_policy = NETISR_POLICY_CPU, 149 .nh_dispatch = NETISR_DISPATCH_HYBRID, 150 #else 151 .nh_policy = NETISR_POLICY_FLOW, 152 #endif 153 }; 154 155 static int 156 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 157 { 158 int error, qlimit; 159 160 netisr_getqlimit(&ip6_nh, &qlimit); 161 error = sysctl_handle_int(oidp, &qlimit, 0, req); 162 if (error || !req->newptr) 163 return (error); 164 if (qlimit < 1) 165 return (EINVAL); 166 return (netisr_setqlimit(&ip6_nh, qlimit)); 167 } 168 SYSCTL_DECL(_net_inet6_ip6); 169 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen, 170 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 171 0, 0, sysctl_netinet6_intr_queue_maxlen, "I", 172 "Maximum size of the IPv6 input queue"); 173 174 VNET_DEFINE_STATIC(bool, ip6_sav) = true; 175 #define V_ip6_sav VNET(ip6_sav) 176 SYSCTL_BOOL(_net_inet6_ip6, OID_AUTO, source_address_validation, 177 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_sav), true, 178 "Drop incoming packets with source address that is a local address"); 179 180 #ifdef RSS 181 static struct netisr_handler ip6_direct_nh = { 182 .nh_name = "ip6_direct", 183 .nh_handler = ip6_direct_input, 184 .nh_proto = NETISR_IPV6_DIRECT, 185 .nh_m2cpuid = rss_soft_m2cpuid_v6, 186 .nh_policy = NETISR_POLICY_CPU, 187 .nh_dispatch = NETISR_DISPATCH_HYBRID, 188 }; 189 190 static int 191 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 192 { 193 int error, qlimit; 194 195 netisr_getqlimit(&ip6_direct_nh, &qlimit); 196 error = sysctl_handle_int(oidp, &qlimit, 0, req); 197 if (error || !req->newptr) 198 return (error); 199 if (qlimit < 1) 200 return (EINVAL); 201 return (netisr_setqlimit(&ip6_direct_nh, qlimit)); 202 } 203 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 204 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 205 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I", 206 "Maximum size of the IPv6 direct input queue"); 207 208 #endif 209 210 VNET_DEFINE(pfil_head_t, inet6_pfil_head); 211 VNET_DEFINE(pfil_head_t, inet6_local_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 args.pa_flags = PFIL_OUT; 250 args.pa_headname = PFIL_INET6_LOCAL_NAME; 251 V_inet6_local_pfil_head = pfil_head_register(&args); 252 253 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 254 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 255 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 256 printf("%s: WARNING: unable to register input helper hook\n", 257 __func__); 258 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 259 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 260 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 261 printf("%s: WARNING: unable to register output helper hook\n", 262 __func__); 263 264 scope6_init(); 265 addrsel_policy_init(); 266 nd6_init(); 267 frag6_init(); 268 269 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 270 271 /* Skip global initialization stuff for non-default instances. */ 272 #ifdef VIMAGE 273 netisr_register_vnet(&ip6_nh); 274 #ifdef RSS 275 netisr_register_vnet(&ip6_direct_nh); 276 #endif 277 #endif 278 } 279 VNET_SYSINIT(ip6_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, 280 ip6_vnet_init, NULL); 281 282 static void 283 ip6_init(void *arg __unused) 284 { 285 286 /* 287 * Register statically those protocols that are unlikely to ever go 288 * dynamic. 289 */ 290 IP6PROTO_REGISTER(IPPROTO_ICMPV6, icmp6_input, rip6_ctlinput); 291 IP6PROTO_REGISTER(IPPROTO_DSTOPTS, dest6_input, NULL); 292 IP6PROTO_REGISTER(IPPROTO_ROUTING, route6_input, NULL); 293 IP6PROTO_REGISTER(IPPROTO_FRAGMENT, frag6_input, NULL); 294 IP6PROTO_REGISTER(IPPROTO_IPV4, encap6_input, NULL); 295 IP6PROTO_REGISTER(IPPROTO_IPV6, encap6_input, NULL); 296 IP6PROTO_REGISTER(IPPROTO_ETHERIP, encap6_input, NULL); 297 IP6PROTO_REGISTER(IPPROTO_GRE, encap6_input, NULL); 298 IP6PROTO_REGISTER(IPPROTO_PIM, encap6_input, NULL); 299 #ifdef SCTP /* XXX: has a loadable & static version */ 300 IP6PROTO_REGISTER(IPPROTO_SCTP, sctp6_input, sctp6_ctlinput); 301 #endif 302 303 EVENTHANDLER_REGISTER(vm_lowmem, frag6_drain, NULL, LOWMEM_PRI_DEFAULT); 304 EVENTHANDLER_REGISTER(mbuf_lowmem, frag6_drain, NULL, 305 LOWMEM_PRI_DEFAULT); 306 307 netisr_register(&ip6_nh); 308 #ifdef RSS 309 netisr_register(&ip6_direct_nh); 310 #endif 311 } 312 SYSINIT(ip6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init, NULL); 313 314 int 315 ip6proto_register(uint8_t proto, ip6proto_input_t input, 316 ip6proto_ctlinput_t ctl) 317 { 318 319 MPASS(proto > 0); 320 321 if (ip6_protox[proto] == rip6_input) { 322 ip6_protox[proto] = input; 323 ip6_ctlprotox[proto] = ctl; 324 return (0); 325 } else 326 return (EEXIST); 327 } 328 329 int 330 ip6proto_unregister(uint8_t proto) 331 { 332 333 MPASS(proto > 0); 334 335 if (ip6_protox[proto] != rip6_input) { 336 ip6_protox[proto] = rip6_input; 337 ip6_ctlprotox[proto] = rip6_ctlinput; 338 return (0); 339 } else 340 return (ENOENT); 341 } 342 343 #ifdef VIMAGE 344 static void 345 ip6_destroy(void *unused __unused) 346 { 347 struct ifaddr *ifa, *nifa; 348 struct ifnet *ifp; 349 int error; 350 351 #ifdef RSS 352 netisr_unregister_vnet(&ip6_direct_nh); 353 #endif 354 netisr_unregister_vnet(&ip6_nh); 355 356 pfil_head_unregister(V_inet6_pfil_head); 357 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 358 if (error != 0) { 359 printf("%s: WARNING: unable to deregister input helper hook " 360 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 361 "error %d returned\n", __func__, error); 362 } 363 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 364 if (error != 0) { 365 printf("%s: WARNING: unable to deregister output helper hook " 366 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 367 "error %d returned\n", __func__, error); 368 } 369 370 /* Cleanup addresses. */ 371 IFNET_RLOCK(); 372 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 373 /* Cannot lock here - lock recursion. */ 374 /* IF_ADDR_LOCK(ifp); */ 375 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 376 if (ifa->ifa_addr->sa_family != AF_INET6) 377 continue; 378 in6_purgeaddr(ifa); 379 } 380 /* IF_ADDR_UNLOCK(ifp); */ 381 in6_ifdetach_destroy(ifp); 382 mld_domifdetach(ifp); 383 } 384 IFNET_RUNLOCK(); 385 386 /* Make sure any routes are gone as well. */ 387 rib_flush_routes_family(AF_INET6); 388 389 frag6_destroy(); 390 nd6_destroy(); 391 in6_ifattach_destroy(); 392 393 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 394 } 395 396 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 397 #endif 398 399 static int 400 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off, 401 int *nxt, int *ours) 402 { 403 struct mbuf *m; 404 struct ip6_hdr *ip6; 405 struct ip6_hbh *hbh; 406 407 if (ip6_hopopts_input(plen, rtalert, mp, off)) { 408 #if 0 /*touches NULL pointer*/ 409 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard); 410 #endif 411 goto out; /* m have already been freed */ 412 } 413 414 /* adjust pointer */ 415 m = *mp; 416 ip6 = mtod(m, struct ip6_hdr *); 417 418 /* 419 * if the payload length field is 0 and the next header field 420 * indicates Hop-by-Hop Options header, then a Jumbo Payload 421 * option MUST be included. 422 */ 423 if (ip6->ip6_plen == 0 && *plen == 0) { 424 /* 425 * Note that if a valid jumbo payload option is 426 * contained, ip6_hopopts_input() must set a valid 427 * (non-zero) payload length to the variable plen. 428 */ 429 IP6STAT_INC(ip6s_badoptions); 430 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 431 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 432 icmp6_error(m, ICMP6_PARAM_PROB, 433 ICMP6_PARAMPROB_HEADER, 434 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 435 goto out; 436 } 437 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 438 hbh = (struct ip6_hbh *)(ip6 + 1); 439 *nxt = hbh->ip6h_nxt; 440 441 /* 442 * If we are acting as a router and the packet contains a 443 * router alert option, see if we know the option value. 444 * Currently, we only support the option value for MLD, in which 445 * case we should pass the packet to the multicast routing 446 * daemon. 447 */ 448 if (*rtalert != ~0) { 449 switch (*rtalert) { 450 case IP6OPT_RTALERT_MLD: 451 if (V_ip6_forwarding) 452 *ours = 1; 453 break; 454 default: 455 /* 456 * RFC2711 requires unrecognized values must be 457 * silently ignored. 458 */ 459 break; 460 } 461 } 462 463 return (0); 464 465 out: 466 return (1); 467 } 468 469 #ifdef RSS 470 /* 471 * IPv6 direct input routine. 472 * 473 * This is called when reinjecting completed fragments where 474 * all of the previous checking and book-keeping has been done. 475 */ 476 void 477 ip6_direct_input(struct mbuf *m) 478 { 479 int off, nxt; 480 int nest; 481 struct m_tag *mtag; 482 struct ip6_direct_ctx *ip6dc; 483 484 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 485 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 486 487 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 488 nxt = ip6dc->ip6dc_nxt; 489 off = ip6dc->ip6dc_off; 490 491 nest = 0; 492 493 m_tag_delete(m, mtag); 494 495 while (nxt != IPPROTO_DONE) { 496 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 497 IP6STAT_INC(ip6s_toomanyhdr); 498 goto bad; 499 } 500 501 /* 502 * protection against faulty packet - there should be 503 * more sanity checks in header chain processing. 504 */ 505 if (m->m_pkthdr.len < off) { 506 IP6STAT_INC(ip6s_tooshort); 507 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 508 goto bad; 509 } 510 511 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 512 if (IPSEC_ENABLED(ipv6)) { 513 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 514 return; 515 } 516 #endif /* IPSEC */ 517 518 nxt = ip6_protox[nxt](&m, &off, nxt); 519 } 520 return; 521 bad: 522 m_freem(m); 523 } 524 #endif 525 526 void 527 ip6_input(struct mbuf *m) 528 { 529 struct in6_addr odst; 530 struct ip6_hdr *ip6; 531 struct in6_ifaddr *ia; 532 struct ifnet *rcvif; 533 u_int32_t plen; 534 u_int32_t rtalert = ~0; 535 int off = sizeof(struct ip6_hdr), nest; 536 int nxt, ours = 0; 537 int srcrt = 0; 538 539 /* 540 * Drop the packet if IPv6 operation is disabled on the interface. 541 */ 542 rcvif = m->m_pkthdr.rcvif; 543 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 544 goto bad; 545 546 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 547 /* 548 * should the inner packet be considered authentic? 549 * see comment in ah4_input(). 550 * NB: m cannot be NULL when passed to the input routine 551 */ 552 553 m->m_flags &= ~M_AUTHIPHDR; 554 m->m_flags &= ~M_AUTHIPDGM; 555 556 #endif /* IPSEC */ 557 558 if (m->m_flags & M_FASTFWD_OURS) { 559 /* 560 * Firewall changed destination to local. 561 */ 562 ip6 = mtod(m, struct ip6_hdr *); 563 goto passin; 564 } 565 566 /* 567 * mbuf statistics 568 */ 569 if (m->m_flags & M_EXT) { 570 if (m->m_next) 571 IP6STAT_INC(ip6s_mext2m); 572 else 573 IP6STAT_INC(ip6s_mext1); 574 } else { 575 if (m->m_next) { 576 struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif; 577 int ifindex = ifp->if_index; 578 if (ifindex >= IP6S_M2MMAX) 579 ifindex = 0; 580 IP6STAT_INC(ip6s_m2m[ifindex]); 581 } else 582 IP6STAT_INC(ip6s_m1); 583 } 584 585 in6_ifstat_inc(rcvif, ifs6_in_receive); 586 IP6STAT_INC(ip6s_total); 587 588 /* 589 * L2 bridge code and some other code can return mbuf chain 590 * that does not conform to KAME requirement. too bad. 591 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 592 */ 593 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 594 struct mbuf *n; 595 596 if (m->m_pkthdr.len > MHLEN) 597 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 598 else 599 n = m_gethdr(M_NOWAIT, MT_DATA); 600 if (n == NULL) 601 goto bad; 602 603 m_move_pkthdr(n, m); 604 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 605 n->m_len = n->m_pkthdr.len; 606 m_freem(m); 607 m = n; 608 } 609 if (m->m_len < sizeof(struct ip6_hdr)) { 610 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 611 IP6STAT_INC(ip6s_toosmall); 612 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 613 goto bad; 614 } 615 } 616 617 ip6 = mtod(m, struct ip6_hdr *); 618 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 619 IP6STAT_INC(ip6s_badvers); 620 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 621 goto bad; 622 } 623 624 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 625 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 626 627 /* 628 * Check against address spoofing/corruption. 629 */ 630 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 631 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 632 /* 633 * XXX: "badscope" is not very suitable for a multicast source. 634 */ 635 IP6STAT_INC(ip6s_badscope); 636 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 637 goto bad; 638 } 639 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 640 !(m->m_flags & M_LOOP)) { 641 /* 642 * In this case, the packet should come from the loopback 643 * interface. However, we cannot just check the if_flags, 644 * because ip6_mloopback() passes the "actual" interface 645 * as the outgoing/incoming interface. 646 */ 647 IP6STAT_INC(ip6s_badscope); 648 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 649 goto bad; 650 } 651 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 652 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 653 /* 654 * RFC4291 2.7: 655 * Nodes must not originate a packet to a multicast address 656 * whose scop field contains the reserved value 0; if such 657 * a packet is received, it must be silently dropped. 658 */ 659 IP6STAT_INC(ip6s_badscope); 660 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 661 goto bad; 662 } 663 /* 664 * The following check is not documented in specs. A malicious 665 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 666 * and bypass security checks (act as if it was from 127.0.0.1 by using 667 * IPv6 src ::ffff:127.0.0.1). Be cautious. 668 * 669 * We have supported IPv6-only kernels for a few years and this issue 670 * has not come up. The world seems to move mostly towards not using 671 * v4mapped on the wire, so it makes sense for us to keep rejecting 672 * any such packets. 673 */ 674 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 675 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 676 IP6STAT_INC(ip6s_badscope); 677 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 678 goto bad; 679 } 680 #if 0 681 /* 682 * Reject packets with IPv4 compatible addresses (auto tunnel). 683 * 684 * The code forbids auto tunnel relay case in RFC1933 (the check is 685 * stronger than RFC1933). We may want to re-enable it if mech-xx 686 * is revised to forbid relaying case. 687 */ 688 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 689 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 690 IP6STAT_INC(ip6s_badscope); 691 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 692 goto bad; 693 } 694 #endif 695 /* 696 * Try to forward the packet, but if we fail continue. 697 * ip6_tryforward() does not generate redirects, so fall 698 * through to normal processing if redirects are required. 699 * ip6_tryforward() does inbound and outbound packet firewall 700 * processing. If firewall has decided that destination becomes 701 * our local address, it sets M_FASTFWD_OURS flag. In this 702 * case skip another inbound firewall processing and update 703 * ip6 pointer. 704 */ 705 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0 706 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 707 && (!IPSEC_ENABLED(ipv6) || 708 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0) 709 #endif 710 ) { 711 if ((m = ip6_tryforward(m)) == NULL) 712 return; 713 if (m->m_flags & M_FASTFWD_OURS) { 714 ip6 = mtod(m, struct ip6_hdr *); 715 goto passin; 716 } 717 } 718 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 719 /* 720 * Bypass packet filtering for packets previously handled by IPsec. 721 */ 722 if (IPSEC_ENABLED(ipv6) && 723 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0) 724 goto passin; 725 #endif 726 /* 727 * Run through list of hooks for input packets. 728 * 729 * NB: Beware of the destination address changing 730 * (e.g. by NAT rewriting). When this happens, 731 * tell ip6_forward to do the right thing. 732 */ 733 734 /* Jump over all PFIL processing if hooks are not active. */ 735 if (!PFIL_HOOKED_IN(V_inet6_pfil_head)) 736 goto passin; 737 738 odst = ip6->ip6_dst; 739 if (pfil_mbuf_in(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif, 740 NULL) != PFIL_PASS) 741 return; 742 ip6 = mtod(m, struct ip6_hdr *); 743 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 744 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 745 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 746 /* 747 * Directly ship the packet on. This allows forwarding 748 * packets originally destined to us to some other directly 749 * connected host. 750 */ 751 ip6_forward(m, 1); 752 return; 753 } 754 755 passin: 756 /* 757 * Disambiguate address scope zones (if there is ambiguity). 758 * We first make sure that the original source or destination address 759 * is not in our internal form for scoped addresses. Such addresses 760 * are not necessarily invalid spec-wise, but we cannot accept them due 761 * to the usage conflict. 762 * in6_setscope() then also checks and rejects the cases where src or 763 * dst are the loopback address and the receiving interface 764 * is not loopback. 765 */ 766 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 767 IP6STAT_INC(ip6s_badscope); /* XXX */ 768 goto bad; 769 } 770 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 771 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 772 IP6STAT_INC(ip6s_badscope); 773 goto bad; 774 } 775 if (m->m_flags & M_FASTFWD_OURS) { 776 m->m_flags &= ~M_FASTFWD_OURS; 777 ours = 1; 778 goto hbhcheck; 779 } 780 /* 781 * Multicast check. Assume packet is for us to avoid 782 * prematurely taking locks. 783 */ 784 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 785 ours = 1; 786 in6_ifstat_inc(rcvif, ifs6_in_mcast); 787 goto hbhcheck; 788 } 789 /* 790 * Unicast check 791 * XXX: For now we keep link-local IPv6 addresses with embedded 792 * scope zone id, therefore we use zero zoneid here. 793 */ 794 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false); 795 if (ia != NULL) { 796 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 797 char ip6bufs[INET6_ADDRSTRLEN]; 798 char ip6bufd[INET6_ADDRSTRLEN]; 799 /* address is not ready, so discard the packet. */ 800 nd6log((LOG_INFO, 801 "ip6_input: packet to an unready address %s->%s\n", 802 ip6_sprintf(ip6bufs, &ip6->ip6_src), 803 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 804 goto bad; 805 } 806 if (V_ip6_sav && !(m->m_flags & M_LOOP) && 807 __predict_false(in6_localip_fib(&ip6->ip6_src, 808 rcvif->if_fib))) { 809 IP6STAT_INC(ip6s_badscope); /* XXX */ 810 goto bad; 811 } 812 /* Count the packet in the ip address stats */ 813 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 814 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 815 ours = 1; 816 goto hbhcheck; 817 } 818 819 /* 820 * Now there is no reason to process the packet if it's not our own 821 * and we're not a router. 822 */ 823 if (!V_ip6_forwarding) { 824 IP6STAT_INC(ip6s_cantforward); 825 goto bad; 826 } 827 828 hbhcheck: 829 /* 830 * Process Hop-by-Hop options header if it's contained. 831 * m may be modified in ip6_hopopts_input(). 832 * If a JumboPayload option is included, plen will also be modified. 833 */ 834 plen = (u_int32_t)ntohs(ip6->ip6_plen); 835 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 836 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0) 837 return; 838 } else 839 nxt = ip6->ip6_nxt; 840 841 /* 842 * Use mbuf flags to propagate Router Alert option to 843 * ICMPv6 layer, as hop-by-hop options have been stripped. 844 */ 845 if (rtalert != ~0) 846 m->m_flags |= M_RTALERT_MLD; 847 848 /* 849 * Check that the amount of data in the buffers 850 * is as at least much as the IPv6 header would have us expect. 851 * Trim mbufs if longer than we expect. 852 * Drop packet if shorter than we expect. 853 */ 854 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 855 IP6STAT_INC(ip6s_tooshort); 856 in6_ifstat_inc(rcvif, ifs6_in_truncated); 857 goto bad; 858 } 859 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 860 if (m->m_len == m->m_pkthdr.len) { 861 m->m_len = sizeof(struct ip6_hdr) + plen; 862 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 863 } else 864 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 865 } 866 867 /* 868 * Forward if desirable. 869 */ 870 if (V_ip6_mrouter && 871 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 872 /* 873 * If we are acting as a multicast router, all 874 * incoming multicast packets are passed to the 875 * kernel-level multicast forwarding function. 876 * The packet is returned (relatively) intact; if 877 * ip6_mforward() returns a non-zero value, the packet 878 * must be discarded, else it may be accepted below. 879 * 880 * XXX TODO: Check hlim and multicast scope here to avoid 881 * unnecessarily calling into ip6_mforward(). 882 */ 883 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 884 IP6STAT_INC(ip6s_cantforward); 885 goto bad; 886 } 887 } else if (!ours) { 888 ip6_forward(m, srcrt); 889 return; 890 } 891 892 /* 893 * We are going to ship the packet to the local protocol stack. Call the 894 * filter again for this 'output' action, allowing redirect-like rules 895 * to adjust the source address. 896 */ 897 if (PFIL_HOOKED_OUT(V_inet6_local_pfil_head)) { 898 if (pfil_mbuf_out(V_inet6_local_pfil_head, &m, V_loif, NULL) != 899 PFIL_PASS) 900 return; 901 if (m == NULL) /* consumed by filter */ 902 return; 903 ip6 = mtod(m, struct ip6_hdr *); 904 } 905 906 /* 907 * Tell launch routine the next header 908 */ 909 IP6STAT_INC(ip6s_delivered); 910 in6_ifstat_inc(rcvif, ifs6_in_deliver); 911 nest = 0; 912 913 while (nxt != IPPROTO_DONE) { 914 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 915 IP6STAT_INC(ip6s_toomanyhdr); 916 goto bad; 917 } 918 919 /* 920 * protection against faulty packet - there should be 921 * more sanity checks in header chain processing. 922 */ 923 if (m->m_pkthdr.len < off) { 924 IP6STAT_INC(ip6s_tooshort); 925 in6_ifstat_inc(rcvif, ifs6_in_truncated); 926 goto bad; 927 } 928 929 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 930 if (IPSEC_ENABLED(ipv6)) { 931 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 932 return; 933 } 934 #endif /* IPSEC */ 935 936 nxt = ip6_protox[nxt](&m, &off, nxt); 937 } 938 return; 939 bad: 940 in6_ifstat_inc(rcvif, ifs6_in_discard); 941 if (m != NULL) 942 m_freem(m); 943 } 944 945 /* 946 * Hop-by-Hop options header processing. If a valid jumbo payload option is 947 * included, the real payload length will be stored in plenp. 948 * 949 * rtalertp - XXX: should be stored more smart way 950 */ 951 static int 952 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 953 struct mbuf **mp, int *offp) 954 { 955 struct mbuf *m = *mp; 956 int off = *offp, hbhlen; 957 struct ip6_hbh *hbh; 958 959 /* validation of the length of the header */ 960 if (m->m_len < off + sizeof(*hbh)) { 961 m = m_pullup(m, off + sizeof(*hbh)); 962 if (m == NULL) { 963 IP6STAT_INC(ip6s_exthdrtoolong); 964 *mp = NULL; 965 return (-1); 966 } 967 } 968 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 969 hbhlen = (hbh->ip6h_len + 1) << 3; 970 971 if (m->m_len < off + hbhlen) { 972 m = m_pullup(m, off + hbhlen); 973 if (m == NULL) { 974 IP6STAT_INC(ip6s_exthdrtoolong); 975 *mp = NULL; 976 return (-1); 977 } 978 } 979 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 980 off += hbhlen; 981 hbhlen -= sizeof(struct ip6_hbh); 982 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 983 hbhlen, rtalertp, plenp) < 0) { 984 *mp = NULL; 985 return (-1); 986 } 987 988 *offp = off; 989 *mp = m; 990 return (0); 991 } 992 993 /* 994 * Search header for all Hop-by-hop options and process each option. 995 * This function is separate from ip6_hopopts_input() in order to 996 * handle a case where the sending node itself process its hop-by-hop 997 * options header. In such a case, the function is called from ip6_output(). 998 * 999 * The function assumes that hbh header is located right after the IPv6 header 1000 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1001 * opthead + hbhlen is located in contiguous memory region. 1002 */ 1003 int 1004 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1005 u_int32_t *rtalertp, u_int32_t *plenp) 1006 { 1007 struct ip6_hdr *ip6; 1008 int optlen = 0; 1009 u_int8_t *opt = opthead; 1010 u_int16_t rtalert_val; 1011 u_int32_t jumboplen; 1012 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1013 1014 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1015 switch (*opt) { 1016 case IP6OPT_PAD1: 1017 optlen = 1; 1018 break; 1019 case IP6OPT_PADN: 1020 if (hbhlen < IP6OPT_MINLEN) { 1021 IP6STAT_INC(ip6s_toosmall); 1022 goto bad; 1023 } 1024 optlen = *(opt + 1) + 2; 1025 break; 1026 case IP6OPT_ROUTER_ALERT: 1027 /* XXX may need check for alignment */ 1028 if (hbhlen < IP6OPT_RTALERT_LEN) { 1029 IP6STAT_INC(ip6s_toosmall); 1030 goto bad; 1031 } 1032 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1033 /* XXX stat */ 1034 icmp6_error(m, ICMP6_PARAM_PROB, 1035 ICMP6_PARAMPROB_HEADER, 1036 erroff + opt + 1 - opthead); 1037 return (-1); 1038 } 1039 optlen = IP6OPT_RTALERT_LEN; 1040 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1041 *rtalertp = ntohs(rtalert_val); 1042 break; 1043 case IP6OPT_JUMBO: 1044 /* XXX may need check for alignment */ 1045 if (hbhlen < IP6OPT_JUMBO_LEN) { 1046 IP6STAT_INC(ip6s_toosmall); 1047 goto bad; 1048 } 1049 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1050 /* XXX stat */ 1051 icmp6_error(m, ICMP6_PARAM_PROB, 1052 ICMP6_PARAMPROB_HEADER, 1053 erroff + opt + 1 - opthead); 1054 return (-1); 1055 } 1056 optlen = IP6OPT_JUMBO_LEN; 1057 1058 /* 1059 * IPv6 packets that have non 0 payload length 1060 * must not contain a jumbo payload option. 1061 */ 1062 ip6 = mtod(m, struct ip6_hdr *); 1063 if (ip6->ip6_plen) { 1064 IP6STAT_INC(ip6s_badoptions); 1065 icmp6_error(m, ICMP6_PARAM_PROB, 1066 ICMP6_PARAMPROB_HEADER, 1067 erroff + opt - opthead); 1068 return (-1); 1069 } 1070 1071 /* 1072 * We may see jumbolen in unaligned location, so 1073 * we'd need to perform bcopy(). 1074 */ 1075 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1076 jumboplen = (u_int32_t)htonl(jumboplen); 1077 1078 #if 1 1079 /* 1080 * if there are multiple jumbo payload options, 1081 * *plenp will be non-zero and the packet will be 1082 * rejected. 1083 * the behavior may need some debate in ipngwg - 1084 * multiple options does not make sense, however, 1085 * there's no explicit mention in specification. 1086 */ 1087 if (*plenp != 0) { 1088 IP6STAT_INC(ip6s_badoptions); 1089 icmp6_error(m, ICMP6_PARAM_PROB, 1090 ICMP6_PARAMPROB_HEADER, 1091 erroff + opt + 2 - opthead); 1092 return (-1); 1093 } 1094 #endif 1095 1096 /* 1097 * jumbo payload length must be larger than 65535. 1098 */ 1099 if (jumboplen <= IPV6_MAXPACKET) { 1100 IP6STAT_INC(ip6s_badoptions); 1101 icmp6_error(m, ICMP6_PARAM_PROB, 1102 ICMP6_PARAMPROB_HEADER, 1103 erroff + opt + 2 - opthead); 1104 return (-1); 1105 } 1106 *plenp = jumboplen; 1107 1108 break; 1109 default: /* unknown option */ 1110 if (hbhlen < IP6OPT_MINLEN) { 1111 IP6STAT_INC(ip6s_toosmall); 1112 goto bad; 1113 } 1114 optlen = ip6_unknown_opt(opt, m, 1115 erroff + opt - opthead); 1116 if (optlen == -1) 1117 return (-1); 1118 optlen += 2; 1119 break; 1120 } 1121 } 1122 1123 return (0); 1124 1125 bad: 1126 m_freem(m); 1127 return (-1); 1128 } 1129 1130 /* 1131 * Unknown option processing. 1132 * The third argument `off' is the offset from the IPv6 header to the option, 1133 * which is necessary if the IPv6 header the and option header and IPv6 header 1134 * is not contiguous in order to return an ICMPv6 error. 1135 */ 1136 int 1137 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1138 { 1139 struct ip6_hdr *ip6; 1140 1141 switch (IP6OPT_TYPE(*optp)) { 1142 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1143 return ((int)*(optp + 1)); 1144 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1145 m_freem(m); 1146 return (-1); 1147 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1148 IP6STAT_INC(ip6s_badoptions); 1149 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1150 return (-1); 1151 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1152 IP6STAT_INC(ip6s_badoptions); 1153 ip6 = mtod(m, struct ip6_hdr *); 1154 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1155 (m->m_flags & (M_BCAST|M_MCAST))) 1156 m_freem(m); 1157 else 1158 icmp6_error(m, ICMP6_PARAM_PROB, 1159 ICMP6_PARAMPROB_OPTION, off); 1160 return (-1); 1161 } 1162 1163 m_freem(m); /* XXX: NOTREACHED */ 1164 return (-1); 1165 } 1166 1167 /* 1168 * Create the "control" list for this pcb. 1169 * These functions will not modify mbuf chain at all. 1170 * 1171 * The routine will be called from upper layer handlers like tcp6_input(). 1172 * Thus the routine assumes that the caller (tcp6_input) have already 1173 * called m_pullup() and all the extension headers are located in the 1174 * very first mbuf on the mbuf chain. 1175 * 1176 * ip6_savecontrol_v4 will handle those options that are possible to be 1177 * set on a v4-mapped socket. 1178 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1179 * options and handle the v6-only ones itself. 1180 */ 1181 struct mbuf ** 1182 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1183 int *v4only) 1184 { 1185 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1186 1187 #ifdef SO_TIMESTAMP 1188 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1189 union { 1190 struct timeval tv; 1191 struct bintime bt; 1192 struct timespec ts; 1193 } t; 1194 struct bintime boottimebin, bt1; 1195 struct timespec ts1; 1196 bool stamped; 1197 1198 stamped = false; 1199 switch (inp->inp_socket->so_ts_clock) { 1200 case SO_TS_REALTIME_MICRO: 1201 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1202 M_TSTMP)) { 1203 mbuf_tstmp2timespec(m, &ts1); 1204 timespec2bintime(&ts1, &bt1); 1205 getboottimebin(&boottimebin); 1206 bintime_add(&bt1, &boottimebin); 1207 bintime2timeval(&bt1, &t.tv); 1208 } else { 1209 microtime(&t.tv); 1210 } 1211 *mp = sbcreatecontrol(&t.tv, sizeof(t.tv), 1212 SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT); 1213 if (*mp != NULL) { 1214 mp = &(*mp)->m_next; 1215 stamped = true; 1216 } 1217 break; 1218 1219 case SO_TS_BINTIME: 1220 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1221 M_TSTMP)) { 1222 mbuf_tstmp2timespec(m, &ts1); 1223 timespec2bintime(&ts1, &t.bt); 1224 getboottimebin(&boottimebin); 1225 bintime_add(&t.bt, &boottimebin); 1226 } else { 1227 bintime(&t.bt); 1228 } 1229 *mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME, 1230 SOL_SOCKET, M_NOWAIT); 1231 if (*mp != NULL) { 1232 mp = &(*mp)->m_next; 1233 stamped = true; 1234 } 1235 break; 1236 1237 case SO_TS_REALTIME: 1238 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1239 M_TSTMP)) { 1240 mbuf_tstmp2timespec(m, &t.ts); 1241 getboottimebin(&boottimebin); 1242 bintime2timespec(&boottimebin, &ts1); 1243 timespecadd(&t.ts, &ts1, &t.ts); 1244 } else { 1245 nanotime(&t.ts); 1246 } 1247 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1248 SCM_REALTIME, SOL_SOCKET, M_NOWAIT); 1249 if (*mp != NULL) { 1250 mp = &(*mp)->m_next; 1251 stamped = true; 1252 } 1253 break; 1254 1255 case SO_TS_MONOTONIC: 1256 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1257 M_TSTMP)) 1258 mbuf_tstmp2timespec(m, &t.ts); 1259 else 1260 nanouptime(&t.ts); 1261 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1262 SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT); 1263 if (*mp != NULL) { 1264 mp = &(*mp)->m_next; 1265 stamped = true; 1266 } 1267 break; 1268 1269 default: 1270 panic("unknown (corrupted) so_ts_clock"); 1271 } 1272 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == 1273 (M_PKTHDR | M_TSTMP)) { 1274 struct sock_timestamp_info sti; 1275 1276 bzero(&sti, sizeof(sti)); 1277 sti.st_info_flags = ST_INFO_HW; 1278 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1279 sti.st_info_flags |= ST_INFO_HW_HPREC; 1280 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO, 1281 SOL_SOCKET, M_NOWAIT); 1282 if (*mp != NULL) 1283 mp = &(*mp)->m_next; 1284 } 1285 } 1286 #endif 1287 1288 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1289 /* RFC 2292 sec. 5 */ 1290 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1291 struct in6_pktinfo pi6; 1292 1293 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1294 #ifdef INET 1295 struct ip *ip; 1296 1297 ip = mtod(m, struct ip *); 1298 pi6.ipi6_addr.s6_addr32[0] = 0; 1299 pi6.ipi6_addr.s6_addr32[1] = 0; 1300 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1301 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1302 #else 1303 /* We won't hit this code */ 1304 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1305 #endif 1306 } else { 1307 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1308 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1309 } 1310 pi6.ipi6_ifindex = 1311 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1312 1313 *mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo), 1314 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6, 1315 M_NOWAIT); 1316 if (*mp) 1317 mp = &(*mp)->m_next; 1318 } 1319 1320 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1321 int hlim; 1322 1323 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1324 #ifdef INET 1325 struct ip *ip; 1326 1327 ip = mtod(m, struct ip *); 1328 hlim = ip->ip_ttl; 1329 #else 1330 /* We won't hit this code */ 1331 hlim = 0; 1332 #endif 1333 } else { 1334 hlim = ip6->ip6_hlim & 0xff; 1335 } 1336 *mp = sbcreatecontrol(&hlim, sizeof(int), 1337 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1338 IPPROTO_IPV6, M_NOWAIT); 1339 if (*mp) 1340 mp = &(*mp)->m_next; 1341 } 1342 1343 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1344 int tclass; 1345 1346 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1347 #ifdef INET 1348 struct ip *ip; 1349 1350 ip = mtod(m, struct ip *); 1351 tclass = ip->ip_tos; 1352 #else 1353 /* We won't hit this code */ 1354 tclass = 0; 1355 #endif 1356 } else { 1357 u_int32_t flowinfo; 1358 1359 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1360 flowinfo >>= 20; 1361 tclass = flowinfo & 0xff; 1362 } 1363 *mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS, 1364 IPPROTO_IPV6, M_NOWAIT); 1365 if (*mp) 1366 mp = &(*mp)->m_next; 1367 } 1368 1369 if (v4only != NULL) { 1370 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1371 *v4only = 1; 1372 } else { 1373 *v4only = 0; 1374 } 1375 } 1376 1377 return (mp); 1378 } 1379 1380 void 1381 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp) 1382 { 1383 struct ip6_hdr *ip6; 1384 int v4only = 0; 1385 1386 mp = ip6_savecontrol_v4(inp, m, mp, &v4only); 1387 if (v4only) 1388 return; 1389 1390 ip6 = mtod(m, struct ip6_hdr *); 1391 /* 1392 * IPV6_HOPOPTS socket option. Recall that we required super-user 1393 * privilege for the option (see ip6_ctloutput), but it might be too 1394 * strict, since there might be some hop-by-hop options which can be 1395 * returned to normal user. 1396 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1397 */ 1398 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) { 1399 /* 1400 * Check if a hop-by-hop options header is contatined in the 1401 * received packet, and if so, store the options as ancillary 1402 * data. Note that a hop-by-hop options header must be 1403 * just after the IPv6 header, which is assured through the 1404 * IPv6 input processing. 1405 */ 1406 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1407 struct ip6_hbh *hbh; 1408 u_int hbhlen; 1409 1410 hbh = (struct ip6_hbh *)(ip6 + 1); 1411 hbhlen = (hbh->ip6h_len + 1) << 3; 1412 1413 /* 1414 * XXX: We copy the whole header even if a 1415 * jumbo payload option is included, the option which 1416 * is to be removed before returning according to 1417 * RFC2292. 1418 * Note: this constraint is removed in RFC3542 1419 */ 1420 *mp = sbcreatecontrol(hbh, hbhlen, 1421 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1422 IPPROTO_IPV6, M_NOWAIT); 1423 if (*mp) 1424 mp = &(*mp)->m_next; 1425 } 1426 } 1427 1428 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1429 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1430 1431 /* 1432 * Search for destination options headers or routing 1433 * header(s) through the header chain, and stores each 1434 * header as ancillary data. 1435 * Note that the order of the headers remains in 1436 * the chain of ancillary data. 1437 */ 1438 while (1) { /* is explicit loop prevention necessary? */ 1439 struct ip6_ext *ip6e = NULL; 1440 u_int elen; 1441 1442 /* 1443 * if it is not an extension header, don't try to 1444 * pull it from the chain. 1445 */ 1446 switch (nxt) { 1447 case IPPROTO_DSTOPTS: 1448 case IPPROTO_ROUTING: 1449 case IPPROTO_HOPOPTS: 1450 case IPPROTO_AH: /* is it possible? */ 1451 break; 1452 default: 1453 goto loopend; 1454 } 1455 1456 if (off + sizeof(*ip6e) > m->m_len) 1457 goto loopend; 1458 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1459 if (nxt == IPPROTO_AH) 1460 elen = (ip6e->ip6e_len + 2) << 2; 1461 else 1462 elen = (ip6e->ip6e_len + 1) << 3; 1463 if (off + elen > m->m_len) 1464 goto loopend; 1465 1466 switch (nxt) { 1467 case IPPROTO_DSTOPTS: 1468 if (!(inp->inp_flags & IN6P_DSTOPTS)) 1469 break; 1470 1471 *mp = sbcreatecontrol(ip6e, elen, 1472 IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1473 IPPROTO_IPV6, M_NOWAIT); 1474 if (*mp) 1475 mp = &(*mp)->m_next; 1476 break; 1477 case IPPROTO_ROUTING: 1478 if (!(inp->inp_flags & IN6P_RTHDR)) 1479 break; 1480 1481 *mp = sbcreatecontrol(ip6e, elen, 1482 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR), 1483 IPPROTO_IPV6, M_NOWAIT); 1484 if (*mp) 1485 mp = &(*mp)->m_next; 1486 break; 1487 case IPPROTO_HOPOPTS: 1488 case IPPROTO_AH: /* is it possible? */ 1489 break; 1490 1491 default: 1492 /* 1493 * other cases have been filtered in the above. 1494 * none will visit this case. here we supply 1495 * the code just in case (nxt overwritten or 1496 * other cases). 1497 */ 1498 goto loopend; 1499 } 1500 1501 /* proceed with the next header. */ 1502 off += elen; 1503 nxt = ip6e->ip6e_nxt; 1504 ip6e = NULL; 1505 } 1506 loopend: 1507 ; 1508 } 1509 1510 if (inp->inp_flags2 & INP_RECVFLOWID) { 1511 uint32_t flowid, flow_type; 1512 1513 flowid = m->m_pkthdr.flowid; 1514 flow_type = M_HASHTYPE_GET(m); 1515 1516 /* 1517 * XXX should handle the failure of one or the 1518 * other - don't populate both? 1519 */ 1520 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID, 1521 IPPROTO_IPV6, M_NOWAIT); 1522 if (*mp) 1523 mp = &(*mp)->m_next; 1524 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t), 1525 IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT); 1526 if (*mp) 1527 mp = &(*mp)->m_next; 1528 } 1529 1530 #ifdef RSS 1531 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1532 uint32_t flowid, flow_type; 1533 uint32_t rss_bucketid; 1534 1535 flowid = m->m_pkthdr.flowid; 1536 flow_type = M_HASHTYPE_GET(m); 1537 1538 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1539 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t), 1540 IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT); 1541 if (*mp) 1542 mp = &(*mp)->m_next; 1543 } 1544 } 1545 #endif 1546 1547 } 1548 #undef IS2292 1549 1550 void 1551 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1552 { 1553 struct socket *so; 1554 struct mbuf *m_mtu; 1555 struct ip6_mtuinfo mtuctl; 1556 1557 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1558 /* 1559 * Notify the error by sending IPV6_PATHMTU ancillary data if 1560 * application wanted to know the MTU value. 1561 * NOTE: we notify disconnected sockets, because some udp 1562 * applications keep sending sockets disconnected. 1563 * NOTE: our implementation doesn't notify connected sockets that has 1564 * foreign address that is different than given destination addresses 1565 * (this is permitted by RFC 3542). 1566 */ 1567 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1568 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1569 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1570 return; 1571 1572 mtuctl.ip6m_mtu = mtu; 1573 mtuctl.ip6m_addr = *dst; 1574 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1575 return; 1576 1577 if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU, 1578 IPPROTO_IPV6, M_NOWAIT)) == NULL) 1579 return; 1580 1581 so = inp->inp_socket; 1582 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1583 == 0) { 1584 soroverflow(so); 1585 m_freem(m_mtu); 1586 /* XXX: should count statistics */ 1587 } else 1588 sorwakeup(so); 1589 } 1590 1591 /* 1592 * Get pointer to the previous header followed by the header 1593 * currently processed. 1594 */ 1595 int 1596 ip6_get_prevhdr(const struct mbuf *m, int off) 1597 { 1598 struct ip6_ext ip6e; 1599 struct ip6_hdr *ip6; 1600 int len, nlen, nxt; 1601 1602 if (off == sizeof(struct ip6_hdr)) 1603 return (offsetof(struct ip6_hdr, ip6_nxt)); 1604 if (off < sizeof(struct ip6_hdr)) 1605 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1606 1607 ip6 = mtod(m, struct ip6_hdr *); 1608 nxt = ip6->ip6_nxt; 1609 len = sizeof(struct ip6_hdr); 1610 nlen = 0; 1611 while (len < off) { 1612 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1613 switch (nxt) { 1614 case IPPROTO_FRAGMENT: 1615 nlen = sizeof(struct ip6_frag); 1616 break; 1617 case IPPROTO_AH: 1618 nlen = (ip6e.ip6e_len + 2) << 2; 1619 break; 1620 default: 1621 nlen = (ip6e.ip6e_len + 1) << 3; 1622 } 1623 len += nlen; 1624 nxt = ip6e.ip6e_nxt; 1625 } 1626 return (len - nlen); 1627 } 1628 1629 /* 1630 * get next header offset. m will be retained. 1631 */ 1632 int 1633 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1634 { 1635 struct ip6_hdr ip6; 1636 struct ip6_ext ip6e; 1637 struct ip6_frag fh; 1638 1639 /* just in case */ 1640 if (m == NULL) 1641 panic("ip6_nexthdr: m == NULL"); 1642 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1643 return -1; 1644 1645 switch (proto) { 1646 case IPPROTO_IPV6: 1647 if (m->m_pkthdr.len < off + sizeof(ip6)) 1648 return -1; 1649 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1650 if (nxtp) 1651 *nxtp = ip6.ip6_nxt; 1652 off += sizeof(ip6); 1653 return off; 1654 1655 case IPPROTO_FRAGMENT: 1656 /* 1657 * terminate parsing if it is not the first fragment, 1658 * it does not make sense to parse through it. 1659 */ 1660 if (m->m_pkthdr.len < off + sizeof(fh)) 1661 return -1; 1662 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1663 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1664 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1665 return -1; 1666 if (nxtp) 1667 *nxtp = fh.ip6f_nxt; 1668 off += sizeof(struct ip6_frag); 1669 return off; 1670 1671 case IPPROTO_AH: 1672 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1673 return -1; 1674 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1675 if (nxtp) 1676 *nxtp = ip6e.ip6e_nxt; 1677 off += (ip6e.ip6e_len + 2) << 2; 1678 return off; 1679 1680 case IPPROTO_HOPOPTS: 1681 case IPPROTO_ROUTING: 1682 case IPPROTO_DSTOPTS: 1683 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1684 return -1; 1685 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1686 if (nxtp) 1687 *nxtp = ip6e.ip6e_nxt; 1688 off += (ip6e.ip6e_len + 1) << 3; 1689 return off; 1690 1691 case IPPROTO_NONE: 1692 case IPPROTO_ESP: 1693 case IPPROTO_IPCOMP: 1694 /* give up */ 1695 return -1; 1696 1697 default: 1698 return -1; 1699 } 1700 1701 /* NOTREACHED */ 1702 } 1703 1704 /* 1705 * get offset for the last header in the chain. m will be kept untainted. 1706 */ 1707 int 1708 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1709 { 1710 int newoff; 1711 int nxt; 1712 1713 if (!nxtp) { 1714 nxt = -1; 1715 nxtp = &nxt; 1716 } 1717 while (1) { 1718 newoff = ip6_nexthdr(m, off, proto, nxtp); 1719 if (newoff < 0) 1720 return off; 1721 else if (newoff < off) 1722 return -1; /* invalid */ 1723 else if (newoff == off) 1724 return newoff; 1725 1726 off = newoff; 1727 proto = *nxtp; 1728 } 1729 } 1730