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 74 #include <sys/param.h> 75 #include <sys/systm.h> 76 #include <sys/hhook.h> 77 #include <sys/malloc.h> 78 #include <sys/mbuf.h> 79 #include <sys/proc.h> 80 #include <sys/domain.h> 81 #include <sys/protosw.h> 82 #include <sys/sdt.h> 83 #include <sys/socket.h> 84 #include <sys/socketvar.h> 85 #include <sys/errno.h> 86 #include <sys/time.h> 87 #include <sys/kernel.h> 88 #include <sys/lock.h> 89 #include <sys/rmlock.h> 90 #include <sys/syslog.h> 91 #include <sys/sysctl.h> 92 93 #include <net/if.h> 94 #include <net/if_var.h> 95 #include <net/if_types.h> 96 #include <net/if_dl.h> 97 #include <net/route.h> 98 #include <net/netisr.h> 99 #include <net/rss_config.h> 100 #include <net/pfil.h> 101 #include <net/vnet.h> 102 103 #include <netinet/in.h> 104 #include <netinet/in_kdtrace.h> 105 #include <netinet/ip_var.h> 106 #include <netinet/in_systm.h> 107 #include <net/if_llatbl.h> 108 #ifdef INET 109 #include <netinet/ip.h> 110 #include <netinet/ip_icmp.h> 111 #endif /* INET */ 112 #include <netinet/ip6.h> 113 #include <netinet6/in6_var.h> 114 #include <netinet6/ip6_var.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> 122 123 #include <netipsec/ipsec_support.h> 124 125 #include <netinet6/ip6protosw.h> 126 127 extern struct domain inet6domain; 128 129 u_char ip6_protox[IPPROTO_MAX]; 130 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 131 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); 132 VNET_DEFINE(u_long, in6_ifaddrhmask); 133 134 static struct netisr_handler ip6_nh = { 135 .nh_name = "ip6", 136 .nh_handler = ip6_input, 137 .nh_proto = NETISR_IPV6, 138 #ifdef RSS 139 .nh_m2cpuid = rss_soft_m2cpuid_v6, 140 .nh_policy = NETISR_POLICY_CPU, 141 .nh_dispatch = NETISR_DISPATCH_HYBRID, 142 #else 143 .nh_policy = NETISR_POLICY_FLOW, 144 #endif 145 }; 146 147 static int 148 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 149 { 150 int error, qlimit; 151 152 netisr_getqlimit(&ip6_nh, &qlimit); 153 error = sysctl_handle_int(oidp, &qlimit, 0, req); 154 if (error || !req->newptr) 155 return (error); 156 if (qlimit < 1) 157 return (EINVAL); 158 return (netisr_setqlimit(&ip6_nh, qlimit)); 159 } 160 SYSCTL_DECL(_net_inet6_ip6); 161 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen, 162 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 163 0, 0, sysctl_netinet6_intr_queue_maxlen, "I", 164 "Maximum size of the IPv6 input queue"); 165 166 VNET_DEFINE_STATIC(bool, ip6_sav) = true; 167 #define V_ip6_sav VNET(ip6_sav) 168 SYSCTL_BOOL(_net_inet6_ip6, OID_AUTO, source_address_validation, 169 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_sav), true, 170 "Drop incoming packets with source address that is a local address"); 171 172 #ifdef RSS 173 static struct netisr_handler ip6_direct_nh = { 174 .nh_name = "ip6_direct", 175 .nh_handler = ip6_direct_input, 176 .nh_proto = NETISR_IPV6_DIRECT, 177 .nh_m2cpuid = rss_soft_m2cpuid_v6, 178 .nh_policy = NETISR_POLICY_CPU, 179 .nh_dispatch = NETISR_DISPATCH_HYBRID, 180 }; 181 182 static int 183 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 184 { 185 int error, qlimit; 186 187 netisr_getqlimit(&ip6_direct_nh, &qlimit); 188 error = sysctl_handle_int(oidp, &qlimit, 0, req); 189 if (error || !req->newptr) 190 return (error); 191 if (qlimit < 1) 192 return (EINVAL); 193 return (netisr_setqlimit(&ip6_direct_nh, qlimit)); 194 } 195 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 196 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 197 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I", 198 "Maximum size of the IPv6 direct input queue"); 199 200 #endif 201 202 VNET_DEFINE(pfil_head_t, inet6_pfil_head); 203 204 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 205 VNET_PCPUSTAT_SYSINIT(ip6stat); 206 #ifdef VIMAGE 207 VNET_PCPUSTAT_SYSUNINIT(ip6stat); 208 #endif /* VIMAGE */ 209 210 struct rmlock in6_ifaddr_lock; 211 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 212 213 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 214 215 /* 216 * IP6 initialization: fill in IP6 protocol switch table. 217 * All protocols not implemented in kernel go to raw IP6 protocol handler. 218 */ 219 void 220 ip6_init(void) 221 { 222 struct pfil_head_args args; 223 struct protosw *pr; 224 int i; 225 226 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 227 &V_ip6_auto_linklocal); 228 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 229 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 230 231 CK_STAILQ_INIT(&V_in6_ifaddrhead); 232 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 233 &V_in6_ifaddrhmask); 234 235 /* Initialize packet filter hooks. */ 236 args.pa_version = PFIL_VERSION; 237 args.pa_flags = PFIL_IN | PFIL_OUT; 238 args.pa_type = PFIL_TYPE_IP6; 239 args.pa_headname = PFIL_INET6_NAME; 240 V_inet6_pfil_head = pfil_head_register(&args); 241 242 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 243 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 244 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 245 printf("%s: WARNING: unable to register input helper hook\n", 246 __func__); 247 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 248 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 249 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 250 printf("%s: WARNING: unable to register output helper hook\n", 251 __func__); 252 253 scope6_init(); 254 addrsel_policy_init(); 255 nd6_init(); 256 frag6_init(); 257 258 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 259 260 /* Skip global initialization stuff for non-default instances. */ 261 #ifdef VIMAGE 262 if (!IS_DEFAULT_VNET(curvnet)) { 263 netisr_register_vnet(&ip6_nh); 264 #ifdef RSS 265 netisr_register_vnet(&ip6_direct_nh); 266 #endif 267 return; 268 } 269 #endif 270 271 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 272 if (pr == NULL) 273 panic("ip6_init"); 274 275 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 276 for (i = 0; i < IPPROTO_MAX; i++) 277 ip6_protox[i] = pr - inet6sw; 278 /* 279 * Cycle through IP protocols and put them into the appropriate place 280 * in ip6_protox[]. 281 */ 282 for (pr = inet6domain.dom_protosw; 283 pr < inet6domain.dom_protoswNPROTOSW; pr++) 284 if (pr->pr_domain->dom_family == PF_INET6 && 285 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 286 /* Be careful to only index valid IP protocols. */ 287 if (pr->pr_protocol < IPPROTO_MAX) 288 ip6_protox[pr->pr_protocol] = pr - inet6sw; 289 } 290 291 netisr_register(&ip6_nh); 292 #ifdef RSS 293 netisr_register(&ip6_direct_nh); 294 #endif 295 } 296 297 /* 298 * The protocol to be inserted into ip6_protox[] must be already registered 299 * in inet6sw[], either statically or through pf_proto_register(). 300 */ 301 int 302 ip6proto_register(short ip6proto) 303 { 304 struct protosw *pr; 305 306 /* Sanity checks. */ 307 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 308 return (EPROTONOSUPPORT); 309 310 /* 311 * The protocol slot must not be occupied by another protocol 312 * already. An index pointing to IPPROTO_RAW is unused. 313 */ 314 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 315 if (pr == NULL) 316 return (EPFNOSUPPORT); 317 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 318 return (EEXIST); 319 320 /* 321 * Find the protocol position in inet6sw[] and set the index. 322 */ 323 for (pr = inet6domain.dom_protosw; 324 pr < inet6domain.dom_protoswNPROTOSW; pr++) { 325 if (pr->pr_domain->dom_family == PF_INET6 && 326 pr->pr_protocol && pr->pr_protocol == ip6proto) { 327 ip6_protox[pr->pr_protocol] = pr - inet6sw; 328 return (0); 329 } 330 } 331 return (EPROTONOSUPPORT); 332 } 333 334 int 335 ip6proto_unregister(short ip6proto) 336 { 337 struct protosw *pr; 338 339 /* Sanity checks. */ 340 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 341 return (EPROTONOSUPPORT); 342 343 /* Check if the protocol was indeed registered. */ 344 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 345 if (pr == NULL) 346 return (EPFNOSUPPORT); 347 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 348 return (ENOENT); 349 350 /* Reset the protocol slot to IPPROTO_RAW. */ 351 ip6_protox[ip6proto] = pr - inet6sw; 352 return (0); 353 } 354 355 #ifdef VIMAGE 356 static void 357 ip6_destroy(void *unused __unused) 358 { 359 struct ifaddr *ifa, *nifa; 360 struct ifnet *ifp; 361 int error; 362 363 #ifdef RSS 364 netisr_unregister_vnet(&ip6_direct_nh); 365 #endif 366 netisr_unregister_vnet(&ip6_nh); 367 368 pfil_head_unregister(V_inet6_pfil_head); 369 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 370 if (error != 0) { 371 printf("%s: WARNING: unable to deregister input helper hook " 372 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 373 "error %d returned\n", __func__, error); 374 } 375 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 376 if (error != 0) { 377 printf("%s: WARNING: unable to deregister output helper hook " 378 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 379 "error %d returned\n", __func__, error); 380 } 381 382 /* Cleanup addresses. */ 383 IFNET_RLOCK(); 384 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 385 /* Cannot lock here - lock recursion. */ 386 /* IF_ADDR_LOCK(ifp); */ 387 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 388 if (ifa->ifa_addr->sa_family != AF_INET6) 389 continue; 390 in6_purgeaddr(ifa); 391 } 392 /* IF_ADDR_UNLOCK(ifp); */ 393 in6_ifdetach_destroy(ifp); 394 mld_domifdetach(ifp); 395 } 396 IFNET_RUNLOCK(); 397 398 /* Make sure any routes are gone as well. */ 399 rib_flush_routes_family(AF_INET6); 400 401 frag6_destroy(); 402 nd6_destroy(); 403 in6_ifattach_destroy(); 404 405 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 406 } 407 408 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 409 #endif 410 411 static int 412 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off, 413 int *nxt, int *ours) 414 { 415 struct mbuf *m; 416 struct ip6_hdr *ip6; 417 struct ip6_hbh *hbh; 418 419 if (ip6_hopopts_input(plen, rtalert, mp, off)) { 420 #if 0 /*touches NULL pointer*/ 421 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard); 422 #endif 423 goto out; /* m have already been freed */ 424 } 425 426 /* adjust pointer */ 427 m = *mp; 428 ip6 = mtod(m, struct ip6_hdr *); 429 430 /* 431 * if the payload length field is 0 and the next header field 432 * indicates Hop-by-Hop Options header, then a Jumbo Payload 433 * option MUST be included. 434 */ 435 if (ip6->ip6_plen == 0 && *plen == 0) { 436 /* 437 * Note that if a valid jumbo payload option is 438 * contained, ip6_hopopts_input() must set a valid 439 * (non-zero) payload length to the variable plen. 440 */ 441 IP6STAT_INC(ip6s_badoptions); 442 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 443 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 444 icmp6_error(m, ICMP6_PARAM_PROB, 445 ICMP6_PARAMPROB_HEADER, 446 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 447 goto out; 448 } 449 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 450 hbh = (struct ip6_hbh *)(ip6 + 1); 451 *nxt = hbh->ip6h_nxt; 452 453 /* 454 * If we are acting as a router and the packet contains a 455 * router alert option, see if we know the option value. 456 * Currently, we only support the option value for MLD, in which 457 * case we should pass the packet to the multicast routing 458 * daemon. 459 */ 460 if (*rtalert != ~0) { 461 switch (*rtalert) { 462 case IP6OPT_RTALERT_MLD: 463 if (V_ip6_forwarding) 464 *ours = 1; 465 break; 466 default: 467 /* 468 * RFC2711 requires unrecognized values must be 469 * silently ignored. 470 */ 471 break; 472 } 473 } 474 475 return (0); 476 477 out: 478 return (1); 479 } 480 481 #ifdef RSS 482 /* 483 * IPv6 direct input routine. 484 * 485 * This is called when reinjecting completed fragments where 486 * all of the previous checking and book-keeping has been done. 487 */ 488 void 489 ip6_direct_input(struct mbuf *m) 490 { 491 int off, nxt; 492 int nest; 493 struct m_tag *mtag; 494 struct ip6_direct_ctx *ip6dc; 495 496 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 497 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 498 499 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 500 nxt = ip6dc->ip6dc_nxt; 501 off = ip6dc->ip6dc_off; 502 503 nest = 0; 504 505 m_tag_delete(m, mtag); 506 507 while (nxt != IPPROTO_DONE) { 508 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 509 IP6STAT_INC(ip6s_toomanyhdr); 510 goto bad; 511 } 512 513 /* 514 * protection against faulty packet - there should be 515 * more sanity checks in header chain processing. 516 */ 517 if (m->m_pkthdr.len < off) { 518 IP6STAT_INC(ip6s_tooshort); 519 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 520 goto bad; 521 } 522 523 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 524 if (IPSEC_ENABLED(ipv6)) { 525 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 526 return; 527 } 528 #endif /* IPSEC */ 529 530 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 531 } 532 return; 533 bad: 534 m_freem(m); 535 } 536 #endif 537 538 void 539 ip6_input(struct mbuf *m) 540 { 541 struct in6_addr odst; 542 struct ip6_hdr *ip6; 543 struct in6_ifaddr *ia; 544 struct ifnet *rcvif; 545 u_int32_t plen; 546 u_int32_t rtalert = ~0; 547 int off = sizeof(struct ip6_hdr), nest; 548 int nxt, ours = 0; 549 int srcrt = 0; 550 551 /* 552 * Drop the packet if IPv6 operation is disabled on the interface. 553 */ 554 rcvif = m->m_pkthdr.rcvif; 555 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 556 goto bad; 557 558 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 559 /* 560 * should the inner packet be considered authentic? 561 * see comment in ah4_input(). 562 * NB: m cannot be NULL when passed to the input routine 563 */ 564 565 m->m_flags &= ~M_AUTHIPHDR; 566 m->m_flags &= ~M_AUTHIPDGM; 567 568 #endif /* IPSEC */ 569 570 if (m->m_flags & M_FASTFWD_OURS) { 571 /* 572 * Firewall changed destination to local. 573 */ 574 ip6 = mtod(m, struct ip6_hdr *); 575 goto passin; 576 } 577 578 /* 579 * mbuf statistics 580 */ 581 if (m->m_flags & M_EXT) { 582 if (m->m_next) 583 IP6STAT_INC(ip6s_mext2m); 584 else 585 IP6STAT_INC(ip6s_mext1); 586 } else { 587 if (m->m_next) { 588 if (m->m_flags & M_LOOP) { 589 IP6STAT_INC(ip6s_m2m[V_loif->if_index]); 590 } else if (rcvif->if_index < IP6S_M2MMAX) 591 IP6STAT_INC(ip6s_m2m[rcvif->if_index]); 592 else 593 IP6STAT_INC(ip6s_m2m[0]); 594 } else 595 IP6STAT_INC(ip6s_m1); 596 } 597 598 in6_ifstat_inc(rcvif, ifs6_in_receive); 599 IP6STAT_INC(ip6s_total); 600 601 /* 602 * L2 bridge code and some other code can return mbuf chain 603 * that does not conform to KAME requirement. too bad. 604 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 605 */ 606 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 607 struct mbuf *n; 608 609 if (m->m_pkthdr.len > MHLEN) 610 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 611 else 612 n = m_gethdr(M_NOWAIT, MT_DATA); 613 if (n == NULL) 614 goto bad; 615 616 m_move_pkthdr(n, m); 617 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 618 n->m_len = n->m_pkthdr.len; 619 m_freem(m); 620 m = n; 621 } 622 if (m->m_len < sizeof(struct ip6_hdr)) { 623 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 624 IP6STAT_INC(ip6s_toosmall); 625 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 626 goto bad; 627 } 628 } 629 630 ip6 = mtod(m, struct ip6_hdr *); 631 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 632 IP6STAT_INC(ip6s_badvers); 633 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 634 goto bad; 635 } 636 637 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 638 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 639 640 /* 641 * Check against address spoofing/corruption. 642 */ 643 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 644 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 645 /* 646 * XXX: "badscope" is not very suitable for a multicast source. 647 */ 648 IP6STAT_INC(ip6s_badscope); 649 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 650 goto bad; 651 } 652 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 653 !(m->m_flags & M_LOOP)) { 654 /* 655 * In this case, the packet should come from the loopback 656 * interface. However, we cannot just check the if_flags, 657 * because ip6_mloopback() passes the "actual" interface 658 * as the outgoing/incoming interface. 659 */ 660 IP6STAT_INC(ip6s_badscope); 661 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 662 goto bad; 663 } 664 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 665 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 666 /* 667 * RFC4291 2.7: 668 * Nodes must not originate a packet to a multicast address 669 * whose scop field contains the reserved value 0; if such 670 * a packet is received, it must be silently dropped. 671 */ 672 IP6STAT_INC(ip6s_badscope); 673 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 674 goto bad; 675 } 676 #ifdef ALTQ 677 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 678 /* packet is dropped by traffic conditioner */ 679 return; 680 } 681 #endif 682 /* 683 * The following check is not documented in specs. A malicious 684 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 685 * and bypass security checks (act as if it was from 127.0.0.1 by using 686 * IPv6 src ::ffff:127.0.0.1). Be cautious. 687 * 688 * We have supported IPv6-only kernels for a few years and this issue 689 * has not come up. The world seems to move mostly towards not using 690 * v4mapped on the wire, so it makes sense for us to keep rejecting 691 * any such packets. 692 */ 693 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 694 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 695 IP6STAT_INC(ip6s_badscope); 696 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 697 goto bad; 698 } 699 #if 0 700 /* 701 * Reject packets with IPv4 compatible addresses (auto tunnel). 702 * 703 * The code forbids auto tunnel relay case in RFC1933 (the check is 704 * stronger than RFC1933). We may want to re-enable it if mech-xx 705 * is revised to forbid relaying case. 706 */ 707 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 708 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 709 IP6STAT_INC(ip6s_badscope); 710 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 711 goto bad; 712 } 713 #endif 714 /* 715 * Try to forward the packet, but if we fail continue. 716 * ip6_tryforward() does not generate redirects, so fall 717 * through to normal processing if redirects are required. 718 * ip6_tryforward() does inbound and outbound packet firewall 719 * processing. If firewall has decided that destination becomes 720 * our local address, it sets M_FASTFWD_OURS flag. In this 721 * case skip another inbound firewall processing and update 722 * ip6 pointer. 723 */ 724 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0 725 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 726 && (!IPSEC_ENABLED(ipv6) || 727 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0) 728 #endif 729 ) { 730 if ((m = ip6_tryforward(m)) == NULL) 731 return; 732 if (m->m_flags & M_FASTFWD_OURS) { 733 ip6 = mtod(m, struct ip6_hdr *); 734 goto passin; 735 } 736 } 737 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 738 /* 739 * Bypass packet filtering for packets previously handled by IPsec. 740 */ 741 if (IPSEC_ENABLED(ipv6) && 742 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0) 743 goto passin; 744 #endif 745 /* 746 * Run through list of hooks for input packets. 747 * 748 * NB: Beware of the destination address changing 749 * (e.g. by NAT rewriting). When this happens, 750 * tell ip6_forward to do the right thing. 751 */ 752 753 /* Jump over all PFIL processing if hooks are not active. */ 754 if (!PFIL_HOOKED_IN(V_inet6_pfil_head)) 755 goto passin; 756 757 odst = ip6->ip6_dst; 758 if (pfil_run_hooks(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif, PFIL_IN, 759 NULL) != PFIL_PASS) 760 return; 761 ip6 = mtod(m, struct ip6_hdr *); 762 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 763 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 764 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 765 /* 766 * Directly ship the packet on. This allows forwarding 767 * packets originally destined to us to some other directly 768 * connected host. 769 */ 770 ip6_forward(m, 1); 771 return; 772 } 773 774 passin: 775 /* 776 * Disambiguate address scope zones (if there is ambiguity). 777 * We first make sure that the original source or destination address 778 * is not in our internal form for scoped addresses. Such addresses 779 * are not necessarily invalid spec-wise, but we cannot accept them due 780 * to the usage conflict. 781 * in6_setscope() then also checks and rejects the cases where src or 782 * dst are the loopback address and the receiving interface 783 * is not loopback. 784 */ 785 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 786 IP6STAT_INC(ip6s_badscope); /* XXX */ 787 goto bad; 788 } 789 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 790 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 791 IP6STAT_INC(ip6s_badscope); 792 goto bad; 793 } 794 if (m->m_flags & M_FASTFWD_OURS) { 795 m->m_flags &= ~M_FASTFWD_OURS; 796 ours = 1; 797 goto hbhcheck; 798 } 799 /* 800 * Multicast check. Assume packet is for us to avoid 801 * prematurely taking locks. 802 */ 803 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 804 ours = 1; 805 in6_ifstat_inc(rcvif, ifs6_in_mcast); 806 goto hbhcheck; 807 } 808 /* 809 * Unicast check 810 * XXX: For now we keep link-local IPv6 addresses with embedded 811 * scope zone id, therefore we use zero zoneid here. 812 */ 813 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false); 814 if (ia != NULL) { 815 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 816 char ip6bufs[INET6_ADDRSTRLEN]; 817 char ip6bufd[INET6_ADDRSTRLEN]; 818 /* address is not ready, so discard the packet. */ 819 nd6log((LOG_INFO, 820 "ip6_input: packet to an unready address %s->%s\n", 821 ip6_sprintf(ip6bufs, &ip6->ip6_src), 822 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 823 goto bad; 824 } 825 if (V_ip6_sav && !(rcvif->if_flags & IFF_LOOPBACK) && 826 __predict_false(in6_localip_fib(&ip6->ip6_src, 827 rcvif->if_fib))) { 828 IP6STAT_INC(ip6s_badscope); /* XXX */ 829 goto bad; 830 } 831 /* Count the packet in the ip address stats */ 832 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 833 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 834 ours = 1; 835 goto hbhcheck; 836 } 837 838 /* 839 * Now there is no reason to process the packet if it's not our own 840 * and we're not a router. 841 */ 842 if (!V_ip6_forwarding) { 843 IP6STAT_INC(ip6s_cantforward); 844 goto bad; 845 } 846 847 hbhcheck: 848 /* 849 * Process Hop-by-Hop options header if it's contained. 850 * m may be modified in ip6_hopopts_input(). 851 * If a JumboPayload option is included, plen will also be modified. 852 */ 853 plen = (u_int32_t)ntohs(ip6->ip6_plen); 854 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 855 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0) 856 return; 857 } else 858 nxt = ip6->ip6_nxt; 859 860 /* 861 * Use mbuf flags to propagate Router Alert option to 862 * ICMPv6 layer, as hop-by-hop options have been stripped. 863 */ 864 if (rtalert != ~0) 865 m->m_flags |= M_RTALERT_MLD; 866 867 /* 868 * Check that the amount of data in the buffers 869 * is as at least much as the IPv6 header would have us expect. 870 * Trim mbufs if longer than we expect. 871 * Drop packet if shorter than we expect. 872 */ 873 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 874 IP6STAT_INC(ip6s_tooshort); 875 in6_ifstat_inc(rcvif, ifs6_in_truncated); 876 goto bad; 877 } 878 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 879 if (m->m_len == m->m_pkthdr.len) { 880 m->m_len = sizeof(struct ip6_hdr) + plen; 881 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 882 } else 883 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 884 } 885 886 /* 887 * Forward if desirable. 888 */ 889 if (V_ip6_mrouter && 890 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 891 /* 892 * If we are acting as a multicast router, all 893 * incoming multicast packets are passed to the 894 * kernel-level multicast forwarding function. 895 * The packet is returned (relatively) intact; if 896 * ip6_mforward() returns a non-zero value, the packet 897 * must be discarded, else it may be accepted below. 898 * 899 * XXX TODO: Check hlim and multicast scope here to avoid 900 * unnecessarily calling into ip6_mforward(). 901 */ 902 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 903 IP6STAT_INC(ip6s_cantforward); 904 goto bad; 905 } 906 } else if (!ours) { 907 ip6_forward(m, srcrt); 908 return; 909 } 910 911 /* 912 * Tell launch routine the next header 913 */ 914 IP6STAT_INC(ip6s_delivered); 915 in6_ifstat_inc(rcvif, ifs6_in_deliver); 916 nest = 0; 917 918 while (nxt != IPPROTO_DONE) { 919 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 920 IP6STAT_INC(ip6s_toomanyhdr); 921 goto bad; 922 } 923 924 /* 925 * protection against faulty packet - there should be 926 * more sanity checks in header chain processing. 927 */ 928 if (m->m_pkthdr.len < off) { 929 IP6STAT_INC(ip6s_tooshort); 930 in6_ifstat_inc(rcvif, ifs6_in_truncated); 931 goto bad; 932 } 933 934 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 935 if (IPSEC_ENABLED(ipv6)) { 936 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 937 return; 938 } 939 #endif /* IPSEC */ 940 941 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 942 } 943 return; 944 bad: 945 in6_ifstat_inc(rcvif, ifs6_in_discard); 946 if (m != NULL) 947 m_freem(m); 948 } 949 950 /* 951 * Hop-by-Hop options header processing. If a valid jumbo payload option is 952 * included, the real payload length will be stored in plenp. 953 * 954 * rtalertp - XXX: should be stored more smart way 955 */ 956 static int 957 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 958 struct mbuf **mp, int *offp) 959 { 960 struct mbuf *m = *mp; 961 int off = *offp, hbhlen; 962 struct ip6_hbh *hbh; 963 964 /* validation of the length of the header */ 965 if (m->m_len < off + sizeof(*hbh)) { 966 m = m_pullup(m, off + sizeof(*hbh)); 967 if (m == NULL) { 968 IP6STAT_INC(ip6s_exthdrtoolong); 969 *mp = NULL; 970 return (-1); 971 } 972 } 973 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 974 hbhlen = (hbh->ip6h_len + 1) << 3; 975 976 if (m->m_len < off + hbhlen) { 977 m = m_pullup(m, off + hbhlen); 978 if (m == NULL) { 979 IP6STAT_INC(ip6s_exthdrtoolong); 980 *mp = NULL; 981 return (-1); 982 } 983 } 984 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 985 off += hbhlen; 986 hbhlen -= sizeof(struct ip6_hbh); 987 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 988 hbhlen, rtalertp, plenp) < 0) { 989 *mp = NULL; 990 return (-1); 991 } 992 993 *offp = off; 994 *mp = m; 995 return (0); 996 } 997 998 /* 999 * Search header for all Hop-by-hop options and process each option. 1000 * This function is separate from ip6_hopopts_input() in order to 1001 * handle a case where the sending node itself process its hop-by-hop 1002 * options header. In such a case, the function is called from ip6_output(). 1003 * 1004 * The function assumes that hbh header is located right after the IPv6 header 1005 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1006 * opthead + hbhlen is located in contiguous memory region. 1007 */ 1008 int 1009 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1010 u_int32_t *rtalertp, u_int32_t *plenp) 1011 { 1012 struct ip6_hdr *ip6; 1013 int optlen = 0; 1014 u_int8_t *opt = opthead; 1015 u_int16_t rtalert_val; 1016 u_int32_t jumboplen; 1017 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1018 1019 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1020 switch (*opt) { 1021 case IP6OPT_PAD1: 1022 optlen = 1; 1023 break; 1024 case IP6OPT_PADN: 1025 if (hbhlen < IP6OPT_MINLEN) { 1026 IP6STAT_INC(ip6s_toosmall); 1027 goto bad; 1028 } 1029 optlen = *(opt + 1) + 2; 1030 break; 1031 case IP6OPT_ROUTER_ALERT: 1032 /* XXX may need check for alignment */ 1033 if (hbhlen < IP6OPT_RTALERT_LEN) { 1034 IP6STAT_INC(ip6s_toosmall); 1035 goto bad; 1036 } 1037 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1038 /* XXX stat */ 1039 icmp6_error(m, ICMP6_PARAM_PROB, 1040 ICMP6_PARAMPROB_HEADER, 1041 erroff + opt + 1 - opthead); 1042 return (-1); 1043 } 1044 optlen = IP6OPT_RTALERT_LEN; 1045 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1046 *rtalertp = ntohs(rtalert_val); 1047 break; 1048 case IP6OPT_JUMBO: 1049 /* XXX may need check for alignment */ 1050 if (hbhlen < IP6OPT_JUMBO_LEN) { 1051 IP6STAT_INC(ip6s_toosmall); 1052 goto bad; 1053 } 1054 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1055 /* XXX stat */ 1056 icmp6_error(m, ICMP6_PARAM_PROB, 1057 ICMP6_PARAMPROB_HEADER, 1058 erroff + opt + 1 - opthead); 1059 return (-1); 1060 } 1061 optlen = IP6OPT_JUMBO_LEN; 1062 1063 /* 1064 * IPv6 packets that have non 0 payload length 1065 * must not contain a jumbo payload option. 1066 */ 1067 ip6 = mtod(m, struct ip6_hdr *); 1068 if (ip6->ip6_plen) { 1069 IP6STAT_INC(ip6s_badoptions); 1070 icmp6_error(m, ICMP6_PARAM_PROB, 1071 ICMP6_PARAMPROB_HEADER, 1072 erroff + opt - opthead); 1073 return (-1); 1074 } 1075 1076 /* 1077 * We may see jumbolen in unaligned location, so 1078 * we'd need to perform bcopy(). 1079 */ 1080 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1081 jumboplen = (u_int32_t)htonl(jumboplen); 1082 1083 #if 1 1084 /* 1085 * if there are multiple jumbo payload options, 1086 * *plenp will be non-zero and the packet will be 1087 * rejected. 1088 * the behavior may need some debate in ipngwg - 1089 * multiple options does not make sense, however, 1090 * there's no explicit mention in specification. 1091 */ 1092 if (*plenp != 0) { 1093 IP6STAT_INC(ip6s_badoptions); 1094 icmp6_error(m, ICMP6_PARAM_PROB, 1095 ICMP6_PARAMPROB_HEADER, 1096 erroff + opt + 2 - opthead); 1097 return (-1); 1098 } 1099 #endif 1100 1101 /* 1102 * jumbo payload length must be larger than 65535. 1103 */ 1104 if (jumboplen <= IPV6_MAXPACKET) { 1105 IP6STAT_INC(ip6s_badoptions); 1106 icmp6_error(m, ICMP6_PARAM_PROB, 1107 ICMP6_PARAMPROB_HEADER, 1108 erroff + opt + 2 - opthead); 1109 return (-1); 1110 } 1111 *plenp = jumboplen; 1112 1113 break; 1114 default: /* unknown option */ 1115 if (hbhlen < IP6OPT_MINLEN) { 1116 IP6STAT_INC(ip6s_toosmall); 1117 goto bad; 1118 } 1119 optlen = ip6_unknown_opt(opt, m, 1120 erroff + opt - opthead); 1121 if (optlen == -1) 1122 return (-1); 1123 optlen += 2; 1124 break; 1125 } 1126 } 1127 1128 return (0); 1129 1130 bad: 1131 m_freem(m); 1132 return (-1); 1133 } 1134 1135 /* 1136 * Unknown option processing. 1137 * The third argument `off' is the offset from the IPv6 header to the option, 1138 * which is necessary if the IPv6 header the and option header and IPv6 header 1139 * is not contiguous in order to return an ICMPv6 error. 1140 */ 1141 int 1142 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1143 { 1144 struct ip6_hdr *ip6; 1145 1146 switch (IP6OPT_TYPE(*optp)) { 1147 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1148 return ((int)*(optp + 1)); 1149 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1150 m_freem(m); 1151 return (-1); 1152 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1153 IP6STAT_INC(ip6s_badoptions); 1154 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1155 return (-1); 1156 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1157 IP6STAT_INC(ip6s_badoptions); 1158 ip6 = mtod(m, struct ip6_hdr *); 1159 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1160 (m->m_flags & (M_BCAST|M_MCAST))) 1161 m_freem(m); 1162 else 1163 icmp6_error(m, ICMP6_PARAM_PROB, 1164 ICMP6_PARAMPROB_OPTION, off); 1165 return (-1); 1166 } 1167 1168 m_freem(m); /* XXX: NOTREACHED */ 1169 return (-1); 1170 } 1171 1172 /* 1173 * Create the "control" list for this pcb. 1174 * These functions will not modify mbuf chain at all. 1175 * 1176 * The routine will be called from upper layer handlers like tcp6_input(). 1177 * Thus the routine assumes that the caller (tcp6_input) have already 1178 * called m_pullup() and all the extension headers are located in the 1179 * very first mbuf on the mbuf chain. 1180 * 1181 * ip6_savecontrol_v4 will handle those options that are possible to be 1182 * set on a v4-mapped socket. 1183 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1184 * options and handle the v6-only ones itself. 1185 */ 1186 struct mbuf ** 1187 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1188 int *v4only) 1189 { 1190 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1191 1192 #ifdef SO_TIMESTAMP 1193 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1194 union { 1195 struct timeval tv; 1196 struct bintime bt; 1197 struct timespec ts; 1198 } t; 1199 struct bintime boottimebin, bt1; 1200 struct timespec ts1; 1201 bool stamped; 1202 1203 stamped = false; 1204 switch (inp->inp_socket->so_ts_clock) { 1205 case SO_TS_REALTIME_MICRO: 1206 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1207 M_TSTMP)) { 1208 mbuf_tstmp2timespec(m, &ts1); 1209 timespec2bintime(&ts1, &bt1); 1210 getboottimebin(&boottimebin); 1211 bintime_add(&bt1, &boottimebin); 1212 bintime2timeval(&bt1, &t.tv); 1213 } else { 1214 microtime(&t.tv); 1215 } 1216 *mp = sbcreatecontrol((caddr_t) &t.tv, sizeof(t.tv), 1217 SCM_TIMESTAMP, SOL_SOCKET); 1218 if (*mp != NULL) { 1219 mp = &(*mp)->m_next; 1220 stamped = true; 1221 } 1222 break; 1223 1224 case SO_TS_BINTIME: 1225 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1226 M_TSTMP)) { 1227 mbuf_tstmp2timespec(m, &ts1); 1228 timespec2bintime(&ts1, &t.bt); 1229 getboottimebin(&boottimebin); 1230 bintime_add(&t.bt, &boottimebin); 1231 } else { 1232 bintime(&t.bt); 1233 } 1234 *mp = sbcreatecontrol((caddr_t)&t.bt, sizeof(t.bt), 1235 SCM_BINTIME, SOL_SOCKET); 1236 if (*mp != NULL) { 1237 mp = &(*mp)->m_next; 1238 stamped = true; 1239 } 1240 break; 1241 1242 case SO_TS_REALTIME: 1243 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1244 M_TSTMP)) { 1245 mbuf_tstmp2timespec(m, &t.ts); 1246 getboottimebin(&boottimebin); 1247 bintime2timespec(&boottimebin, &ts1); 1248 timespecadd(&t.ts, &ts1, &t.ts); 1249 } else { 1250 nanotime(&t.ts); 1251 } 1252 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts), 1253 SCM_REALTIME, SOL_SOCKET); 1254 if (*mp != NULL) { 1255 mp = &(*mp)->m_next; 1256 stamped = true; 1257 } 1258 break; 1259 1260 case SO_TS_MONOTONIC: 1261 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1262 M_TSTMP)) 1263 mbuf_tstmp2timespec(m, &t.ts); 1264 else 1265 nanouptime(&t.ts); 1266 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts), 1267 SCM_MONOTONIC, SOL_SOCKET); 1268 if (*mp != NULL) { 1269 mp = &(*mp)->m_next; 1270 stamped = true; 1271 } 1272 break; 1273 1274 default: 1275 panic("unknown (corrupted) so_ts_clock"); 1276 } 1277 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == 1278 (M_PKTHDR | M_TSTMP)) { 1279 struct sock_timestamp_info sti; 1280 1281 bzero(&sti, sizeof(sti)); 1282 sti.st_info_flags = ST_INFO_HW; 1283 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1284 sti.st_info_flags |= ST_INFO_HW_HPREC; 1285 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), 1286 SCM_TIME_INFO, SOL_SOCKET); 1287 if (*mp != NULL) 1288 mp = &(*mp)->m_next; 1289 } 1290 } 1291 #endif 1292 1293 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1294 /* RFC 2292 sec. 5 */ 1295 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1296 struct in6_pktinfo pi6; 1297 1298 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1299 #ifdef INET 1300 struct ip *ip; 1301 1302 ip = mtod(m, struct ip *); 1303 pi6.ipi6_addr.s6_addr32[0] = 0; 1304 pi6.ipi6_addr.s6_addr32[1] = 0; 1305 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1306 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1307 #else 1308 /* We won't hit this code */ 1309 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1310 #endif 1311 } else { 1312 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1313 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1314 } 1315 pi6.ipi6_ifindex = 1316 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1317 1318 *mp = sbcreatecontrol((caddr_t) &pi6, 1319 sizeof(struct in6_pktinfo), 1320 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1321 if (*mp) 1322 mp = &(*mp)->m_next; 1323 } 1324 1325 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1326 int hlim; 1327 1328 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1329 #ifdef INET 1330 struct ip *ip; 1331 1332 ip = mtod(m, struct ip *); 1333 hlim = ip->ip_ttl; 1334 #else 1335 /* We won't hit this code */ 1336 hlim = 0; 1337 #endif 1338 } else { 1339 hlim = ip6->ip6_hlim & 0xff; 1340 } 1341 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1342 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1343 IPPROTO_IPV6); 1344 if (*mp) 1345 mp = &(*mp)->m_next; 1346 } 1347 1348 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1349 int tclass; 1350 1351 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1352 #ifdef INET 1353 struct ip *ip; 1354 1355 ip = mtod(m, struct ip *); 1356 tclass = ip->ip_tos; 1357 #else 1358 /* We won't hit this code */ 1359 tclass = 0; 1360 #endif 1361 } else { 1362 u_int32_t flowinfo; 1363 1364 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1365 flowinfo >>= 20; 1366 tclass = flowinfo & 0xff; 1367 } 1368 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int), 1369 IPV6_TCLASS, IPPROTO_IPV6); 1370 if (*mp) 1371 mp = &(*mp)->m_next; 1372 } 1373 1374 if (v4only != NULL) { 1375 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1376 *v4only = 1; 1377 } else { 1378 *v4only = 0; 1379 } 1380 } 1381 1382 return (mp); 1383 } 1384 1385 void 1386 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp) 1387 { 1388 struct ip6_hdr *ip6; 1389 int v4only = 0; 1390 1391 mp = ip6_savecontrol_v4(inp, m, mp, &v4only); 1392 if (v4only) 1393 return; 1394 1395 ip6 = mtod(m, struct ip6_hdr *); 1396 /* 1397 * IPV6_HOPOPTS socket option. Recall that we required super-user 1398 * privilege for the option (see ip6_ctloutput), but it might be too 1399 * strict, since there might be some hop-by-hop options which can be 1400 * returned to normal user. 1401 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1402 */ 1403 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) { 1404 /* 1405 * Check if a hop-by-hop options header is contatined in the 1406 * received packet, and if so, store the options as ancillary 1407 * data. Note that a hop-by-hop options header must be 1408 * just after the IPv6 header, which is assured through the 1409 * IPv6 input processing. 1410 */ 1411 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1412 struct ip6_hbh *hbh; 1413 int hbhlen; 1414 1415 hbh = (struct ip6_hbh *)(ip6 + 1); 1416 hbhlen = (hbh->ip6h_len + 1) << 3; 1417 1418 /* 1419 * XXX: We copy the whole header even if a 1420 * jumbo payload option is included, the option which 1421 * is to be removed before returning according to 1422 * RFC2292. 1423 * Note: this constraint is removed in RFC3542 1424 */ 1425 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1426 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1427 IPPROTO_IPV6); 1428 if (*mp) 1429 mp = &(*mp)->m_next; 1430 } 1431 } 1432 1433 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1434 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1435 1436 /* 1437 * Search for destination options headers or routing 1438 * header(s) through the header chain, and stores each 1439 * header as ancillary data. 1440 * Note that the order of the headers remains in 1441 * the chain of ancillary data. 1442 */ 1443 while (1) { /* is explicit loop prevention necessary? */ 1444 struct ip6_ext *ip6e = NULL; 1445 int elen; 1446 1447 /* 1448 * if it is not an extension header, don't try to 1449 * pull it from the chain. 1450 */ 1451 switch (nxt) { 1452 case IPPROTO_DSTOPTS: 1453 case IPPROTO_ROUTING: 1454 case IPPROTO_HOPOPTS: 1455 case IPPROTO_AH: /* is it possible? */ 1456 break; 1457 default: 1458 goto loopend; 1459 } 1460 1461 if (off + sizeof(*ip6e) > m->m_len) 1462 goto loopend; 1463 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1464 if (nxt == IPPROTO_AH) 1465 elen = (ip6e->ip6e_len + 2) << 2; 1466 else 1467 elen = (ip6e->ip6e_len + 1) << 3; 1468 if (off + elen > m->m_len) 1469 goto loopend; 1470 1471 switch (nxt) { 1472 case IPPROTO_DSTOPTS: 1473 if (!(inp->inp_flags & IN6P_DSTOPTS)) 1474 break; 1475 1476 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1477 IS2292(inp, 1478 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1479 IPPROTO_IPV6); 1480 if (*mp) 1481 mp = &(*mp)->m_next; 1482 break; 1483 case IPPROTO_ROUTING: 1484 if (!(inp->inp_flags & IN6P_RTHDR)) 1485 break; 1486 1487 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1488 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR), 1489 IPPROTO_IPV6); 1490 if (*mp) 1491 mp = &(*mp)->m_next; 1492 break; 1493 case IPPROTO_HOPOPTS: 1494 case IPPROTO_AH: /* is it possible? */ 1495 break; 1496 1497 default: 1498 /* 1499 * other cases have been filtered in the above. 1500 * none will visit this case. here we supply 1501 * the code just in case (nxt overwritten or 1502 * other cases). 1503 */ 1504 goto loopend; 1505 } 1506 1507 /* proceed with the next header. */ 1508 off += elen; 1509 nxt = ip6e->ip6e_nxt; 1510 ip6e = NULL; 1511 } 1512 loopend: 1513 ; 1514 } 1515 1516 if (inp->inp_flags2 & INP_RECVFLOWID) { 1517 uint32_t flowid, flow_type; 1518 1519 flowid = m->m_pkthdr.flowid; 1520 flow_type = M_HASHTYPE_GET(m); 1521 1522 /* 1523 * XXX should handle the failure of one or the 1524 * other - don't populate both? 1525 */ 1526 *mp = sbcreatecontrol((caddr_t) &flowid, 1527 sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6); 1528 if (*mp) 1529 mp = &(*mp)->m_next; 1530 *mp = sbcreatecontrol((caddr_t) &flow_type, 1531 sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6); 1532 if (*mp) 1533 mp = &(*mp)->m_next; 1534 } 1535 1536 #ifdef RSS 1537 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1538 uint32_t flowid, flow_type; 1539 uint32_t rss_bucketid; 1540 1541 flowid = m->m_pkthdr.flowid; 1542 flow_type = M_HASHTYPE_GET(m); 1543 1544 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1545 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1546 sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6); 1547 if (*mp) 1548 mp = &(*mp)->m_next; 1549 } 1550 } 1551 #endif 1552 1553 } 1554 #undef IS2292 1555 1556 void 1557 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1558 { 1559 struct socket *so; 1560 struct mbuf *m_mtu; 1561 struct ip6_mtuinfo mtuctl; 1562 1563 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1564 /* 1565 * Notify the error by sending IPV6_PATHMTU ancillary data if 1566 * application wanted to know the MTU value. 1567 * NOTE: we notify disconnected sockets, because some udp 1568 * applications keep sending sockets disconnected. 1569 * NOTE: our implementation doesn't notify connected sockets that has 1570 * foreign address that is different than given destination addresses 1571 * (this is permitted by RFC 3542). 1572 */ 1573 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1574 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1575 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1576 return; 1577 1578 mtuctl.ip6m_mtu = mtu; 1579 mtuctl.ip6m_addr = *dst; 1580 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1581 return; 1582 1583 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1584 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1585 return; 1586 1587 so = inp->inp_socket; 1588 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1589 == 0) { 1590 soroverflow(so); 1591 m_freem(m_mtu); 1592 /* XXX: should count statistics */ 1593 } else 1594 sorwakeup(so); 1595 } 1596 1597 /* 1598 * Get pointer to the previous header followed by the header 1599 * currently processed. 1600 */ 1601 int 1602 ip6_get_prevhdr(const struct mbuf *m, int off) 1603 { 1604 struct ip6_ext ip6e; 1605 struct ip6_hdr *ip6; 1606 int len, nlen, nxt; 1607 1608 if (off == sizeof(struct ip6_hdr)) 1609 return (offsetof(struct ip6_hdr, ip6_nxt)); 1610 if (off < sizeof(struct ip6_hdr)) 1611 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1612 1613 ip6 = mtod(m, struct ip6_hdr *); 1614 nxt = ip6->ip6_nxt; 1615 len = sizeof(struct ip6_hdr); 1616 nlen = 0; 1617 while (len < off) { 1618 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1619 switch (nxt) { 1620 case IPPROTO_FRAGMENT: 1621 nlen = sizeof(struct ip6_frag); 1622 break; 1623 case IPPROTO_AH: 1624 nlen = (ip6e.ip6e_len + 2) << 2; 1625 break; 1626 default: 1627 nlen = (ip6e.ip6e_len + 1) << 3; 1628 } 1629 len += nlen; 1630 nxt = ip6e.ip6e_nxt; 1631 } 1632 return (len - nlen); 1633 } 1634 1635 /* 1636 * get next header offset. m will be retained. 1637 */ 1638 int 1639 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1640 { 1641 struct ip6_hdr ip6; 1642 struct ip6_ext ip6e; 1643 struct ip6_frag fh; 1644 1645 /* just in case */ 1646 if (m == NULL) 1647 panic("ip6_nexthdr: m == NULL"); 1648 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1649 return -1; 1650 1651 switch (proto) { 1652 case IPPROTO_IPV6: 1653 if (m->m_pkthdr.len < off + sizeof(ip6)) 1654 return -1; 1655 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1656 if (nxtp) 1657 *nxtp = ip6.ip6_nxt; 1658 off += sizeof(ip6); 1659 return off; 1660 1661 case IPPROTO_FRAGMENT: 1662 /* 1663 * terminate parsing if it is not the first fragment, 1664 * it does not make sense to parse through it. 1665 */ 1666 if (m->m_pkthdr.len < off + sizeof(fh)) 1667 return -1; 1668 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1669 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1670 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1671 return -1; 1672 if (nxtp) 1673 *nxtp = fh.ip6f_nxt; 1674 off += sizeof(struct ip6_frag); 1675 return off; 1676 1677 case IPPROTO_AH: 1678 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1679 return -1; 1680 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1681 if (nxtp) 1682 *nxtp = ip6e.ip6e_nxt; 1683 off += (ip6e.ip6e_len + 2) << 2; 1684 return off; 1685 1686 case IPPROTO_HOPOPTS: 1687 case IPPROTO_ROUTING: 1688 case IPPROTO_DSTOPTS: 1689 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1690 return -1; 1691 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1692 if (nxtp) 1693 *nxtp = ip6e.ip6e_nxt; 1694 off += (ip6e.ip6e_len + 1) << 3; 1695 return off; 1696 1697 case IPPROTO_NONE: 1698 case IPPROTO_ESP: 1699 case IPPROTO_IPCOMP: 1700 /* give up */ 1701 return -1; 1702 1703 default: 1704 return -1; 1705 } 1706 1707 /* NOTREACHED */ 1708 } 1709 1710 /* 1711 * get offset for the last header in the chain. m will be kept untainted. 1712 */ 1713 int 1714 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1715 { 1716 int newoff; 1717 int nxt; 1718 1719 if (!nxtp) { 1720 nxt = -1; 1721 nxtp = &nxt; 1722 } 1723 while (1) { 1724 newoff = ip6_nexthdr(m, off, proto, nxtp); 1725 if (newoff < 0) 1726 return off; 1727 else if (newoff < off) 1728 return -1; /* invalid */ 1729 else if (newoff == off) 1730 return newoff; 1731 1732 off = newoff; 1733 proto = *nxtp; 1734 } 1735 } 1736 1737 /* 1738 * System control for IP6 1739 */ 1740 1741 u_char inet6ctlerrmap[PRC_NCMDS] = { 1742 0, 0, 0, 0, 1743 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1744 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1745 EMSGSIZE, EHOSTUNREACH, 0, 0, 1746 0, 0, EHOSTUNREACH, 0, 1747 ENOPROTOOPT, ECONNREFUSED 1748 }; 1749