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 static void 220 ip6_vnet_init(void *arg __unused) 221 { 222 struct pfil_head_args args; 223 224 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 225 &V_ip6_auto_linklocal); 226 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 227 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 228 229 CK_STAILQ_INIT(&V_in6_ifaddrhead); 230 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 231 &V_in6_ifaddrhmask); 232 233 /* Initialize packet filter hooks. */ 234 args.pa_version = PFIL_VERSION; 235 args.pa_flags = PFIL_IN | PFIL_OUT; 236 args.pa_type = PFIL_TYPE_IP6; 237 args.pa_headname = PFIL_INET6_NAME; 238 V_inet6_pfil_head = pfil_head_register(&args); 239 240 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 241 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 242 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 243 printf("%s: WARNING: unable to register input helper hook\n", 244 __func__); 245 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 246 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 247 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 248 printf("%s: WARNING: unable to register output helper hook\n", 249 __func__); 250 251 scope6_init(); 252 addrsel_policy_init(); 253 nd6_init(); 254 frag6_init(); 255 256 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 257 258 /* Skip global initialization stuff for non-default instances. */ 259 #ifdef VIMAGE 260 netisr_register_vnet(&ip6_nh); 261 #ifdef RSS 262 netisr_register_vnet(&ip6_direct_nh); 263 #endif 264 #endif 265 } 266 VNET_SYSINIT(ip6_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, 267 ip6_vnet_init, NULL); 268 269 static void 270 ip6_init(void *arg __unused) 271 { 272 struct protosw *pr; 273 274 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 275 KASSERT(pr, ("%s: PF_INET6 not found", __func__)); 276 277 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 278 for (int i = 0; i < IPPROTO_MAX; i++) 279 ip6_protox[i] = pr - inet6sw; 280 /* 281 * Cycle through IP protocols and put them into the appropriate place 282 * in ip6_protox[]. 283 */ 284 for (pr = inet6domain.dom_protosw; 285 pr < inet6domain.dom_protoswNPROTOSW; pr++) 286 if (pr->pr_domain->dom_family == PF_INET6 && 287 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 288 /* Be careful to only index valid IP protocols. */ 289 if (pr->pr_protocol < IPPROTO_MAX) 290 ip6_protox[pr->pr_protocol] = pr - inet6sw; 291 } 292 293 netisr_register(&ip6_nh); 294 #ifdef RSS 295 netisr_register(&ip6_direct_nh); 296 #endif 297 } 298 SYSINIT(ip6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init, NULL); 299 300 /* 301 * The protocol to be inserted into ip6_protox[] must be already registered 302 * in inet6sw[], either statically or through pf_proto_register(). 303 */ 304 int 305 ip6proto_register(short ip6proto) 306 { 307 struct protosw *pr; 308 309 /* Sanity checks. */ 310 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 311 return (EPROTONOSUPPORT); 312 313 /* 314 * The protocol slot must not be occupied by another protocol 315 * already. An index pointing to IPPROTO_RAW is unused. 316 */ 317 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 318 if (pr == NULL) 319 return (EPFNOSUPPORT); 320 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 321 return (EEXIST); 322 323 /* 324 * Find the protocol position in inet6sw[] and set the index. 325 */ 326 for (pr = inet6domain.dom_protosw; 327 pr < inet6domain.dom_protoswNPROTOSW; pr++) { 328 if (pr->pr_domain->dom_family == PF_INET6 && 329 pr->pr_protocol && pr->pr_protocol == ip6proto) { 330 ip6_protox[pr->pr_protocol] = pr - inet6sw; 331 return (0); 332 } 333 } 334 return (EPROTONOSUPPORT); 335 } 336 337 int 338 ip6proto_unregister(short ip6proto) 339 { 340 struct protosw *pr; 341 342 /* Sanity checks. */ 343 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 344 return (EPROTONOSUPPORT); 345 346 /* Check if the protocol was indeed registered. */ 347 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 348 if (pr == NULL) 349 return (EPFNOSUPPORT); 350 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 351 return (ENOENT); 352 353 /* Reset the protocol slot to IPPROTO_RAW. */ 354 ip6_protox[ip6proto] = pr - inet6sw; 355 return (0); 356 } 357 358 #ifdef VIMAGE 359 static void 360 ip6_destroy(void *unused __unused) 361 { 362 struct ifaddr *ifa, *nifa; 363 struct ifnet *ifp; 364 int error; 365 366 #ifdef RSS 367 netisr_unregister_vnet(&ip6_direct_nh); 368 #endif 369 netisr_unregister_vnet(&ip6_nh); 370 371 pfil_head_unregister(V_inet6_pfil_head); 372 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 373 if (error != 0) { 374 printf("%s: WARNING: unable to deregister input helper hook " 375 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 376 "error %d returned\n", __func__, error); 377 } 378 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 379 if (error != 0) { 380 printf("%s: WARNING: unable to deregister output helper hook " 381 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 382 "error %d returned\n", __func__, error); 383 } 384 385 /* Cleanup addresses. */ 386 IFNET_RLOCK(); 387 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 388 /* Cannot lock here - lock recursion. */ 389 /* IF_ADDR_LOCK(ifp); */ 390 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 391 if (ifa->ifa_addr->sa_family != AF_INET6) 392 continue; 393 in6_purgeaddr(ifa); 394 } 395 /* IF_ADDR_UNLOCK(ifp); */ 396 in6_ifdetach_destroy(ifp); 397 mld_domifdetach(ifp); 398 } 399 IFNET_RUNLOCK(); 400 401 /* Make sure any routes are gone as well. */ 402 rib_flush_routes_family(AF_INET6); 403 404 frag6_destroy(); 405 nd6_destroy(); 406 in6_ifattach_destroy(); 407 408 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 409 } 410 411 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 412 #endif 413 414 static int 415 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off, 416 int *nxt, int *ours) 417 { 418 struct mbuf *m; 419 struct ip6_hdr *ip6; 420 struct ip6_hbh *hbh; 421 422 if (ip6_hopopts_input(plen, rtalert, mp, off)) { 423 #if 0 /*touches NULL pointer*/ 424 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard); 425 #endif 426 goto out; /* m have already been freed */ 427 } 428 429 /* adjust pointer */ 430 m = *mp; 431 ip6 = mtod(m, struct ip6_hdr *); 432 433 /* 434 * if the payload length field is 0 and the next header field 435 * indicates Hop-by-Hop Options header, then a Jumbo Payload 436 * option MUST be included. 437 */ 438 if (ip6->ip6_plen == 0 && *plen == 0) { 439 /* 440 * Note that if a valid jumbo payload option is 441 * contained, ip6_hopopts_input() must set a valid 442 * (non-zero) payload length to the variable plen. 443 */ 444 IP6STAT_INC(ip6s_badoptions); 445 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 446 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 447 icmp6_error(m, ICMP6_PARAM_PROB, 448 ICMP6_PARAMPROB_HEADER, 449 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 450 goto out; 451 } 452 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 453 hbh = (struct ip6_hbh *)(ip6 + 1); 454 *nxt = hbh->ip6h_nxt; 455 456 /* 457 * If we are acting as a router and the packet contains a 458 * router alert option, see if we know the option value. 459 * Currently, we only support the option value for MLD, in which 460 * case we should pass the packet to the multicast routing 461 * daemon. 462 */ 463 if (*rtalert != ~0) { 464 switch (*rtalert) { 465 case IP6OPT_RTALERT_MLD: 466 if (V_ip6_forwarding) 467 *ours = 1; 468 break; 469 default: 470 /* 471 * RFC2711 requires unrecognized values must be 472 * silently ignored. 473 */ 474 break; 475 } 476 } 477 478 return (0); 479 480 out: 481 return (1); 482 } 483 484 #ifdef RSS 485 /* 486 * IPv6 direct input routine. 487 * 488 * This is called when reinjecting completed fragments where 489 * all of the previous checking and book-keeping has been done. 490 */ 491 void 492 ip6_direct_input(struct mbuf *m) 493 { 494 int off, nxt; 495 int nest; 496 struct m_tag *mtag; 497 struct ip6_direct_ctx *ip6dc; 498 499 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 500 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 501 502 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 503 nxt = ip6dc->ip6dc_nxt; 504 off = ip6dc->ip6dc_off; 505 506 nest = 0; 507 508 m_tag_delete(m, mtag); 509 510 while (nxt != IPPROTO_DONE) { 511 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 512 IP6STAT_INC(ip6s_toomanyhdr); 513 goto bad; 514 } 515 516 /* 517 * protection against faulty packet - there should be 518 * more sanity checks in header chain processing. 519 */ 520 if (m->m_pkthdr.len < off) { 521 IP6STAT_INC(ip6s_tooshort); 522 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 523 goto bad; 524 } 525 526 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 527 if (IPSEC_ENABLED(ipv6)) { 528 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 529 return; 530 } 531 #endif /* IPSEC */ 532 533 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 534 } 535 return; 536 bad: 537 m_freem(m); 538 } 539 #endif 540 541 void 542 ip6_input(struct mbuf *m) 543 { 544 struct in6_addr odst; 545 struct ip6_hdr *ip6; 546 struct in6_ifaddr *ia; 547 struct ifnet *rcvif; 548 u_int32_t plen; 549 u_int32_t rtalert = ~0; 550 int off = sizeof(struct ip6_hdr), nest; 551 int nxt, ours = 0; 552 int srcrt = 0; 553 554 /* 555 * Drop the packet if IPv6 operation is disabled on the interface. 556 */ 557 rcvif = m->m_pkthdr.rcvif; 558 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 559 goto bad; 560 561 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 562 /* 563 * should the inner packet be considered authentic? 564 * see comment in ah4_input(). 565 * NB: m cannot be NULL when passed to the input routine 566 */ 567 568 m->m_flags &= ~M_AUTHIPHDR; 569 m->m_flags &= ~M_AUTHIPDGM; 570 571 #endif /* IPSEC */ 572 573 if (m->m_flags & M_FASTFWD_OURS) { 574 /* 575 * Firewall changed destination to local. 576 */ 577 ip6 = mtod(m, struct ip6_hdr *); 578 goto passin; 579 } 580 581 /* 582 * mbuf statistics 583 */ 584 if (m->m_flags & M_EXT) { 585 if (m->m_next) 586 IP6STAT_INC(ip6s_mext2m); 587 else 588 IP6STAT_INC(ip6s_mext1); 589 } else { 590 if (m->m_next) { 591 if (m->m_flags & M_LOOP) { 592 IP6STAT_INC(ip6s_m2m[V_loif->if_index]); 593 } else if (rcvif->if_index < IP6S_M2MMAX) 594 IP6STAT_INC(ip6s_m2m[rcvif->if_index]); 595 else 596 IP6STAT_INC(ip6s_m2m[0]); 597 } else 598 IP6STAT_INC(ip6s_m1); 599 } 600 601 in6_ifstat_inc(rcvif, ifs6_in_receive); 602 IP6STAT_INC(ip6s_total); 603 604 /* 605 * L2 bridge code and some other code can return mbuf chain 606 * that does not conform to KAME requirement. too bad. 607 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 608 */ 609 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 610 struct mbuf *n; 611 612 if (m->m_pkthdr.len > MHLEN) 613 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 614 else 615 n = m_gethdr(M_NOWAIT, MT_DATA); 616 if (n == NULL) 617 goto bad; 618 619 m_move_pkthdr(n, m); 620 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 621 n->m_len = n->m_pkthdr.len; 622 m_freem(m); 623 m = n; 624 } 625 if (m->m_len < sizeof(struct ip6_hdr)) { 626 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 627 IP6STAT_INC(ip6s_toosmall); 628 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 629 goto bad; 630 } 631 } 632 633 ip6 = mtod(m, struct ip6_hdr *); 634 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 635 IP6STAT_INC(ip6s_badvers); 636 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 637 goto bad; 638 } 639 640 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 641 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 642 643 /* 644 * Check against address spoofing/corruption. 645 */ 646 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 647 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 648 /* 649 * XXX: "badscope" is not very suitable for a multicast source. 650 */ 651 IP6STAT_INC(ip6s_badscope); 652 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 653 goto bad; 654 } 655 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 656 !(m->m_flags & M_LOOP)) { 657 /* 658 * In this case, the packet should come from the loopback 659 * interface. However, we cannot just check the if_flags, 660 * because ip6_mloopback() passes the "actual" interface 661 * as the outgoing/incoming interface. 662 */ 663 IP6STAT_INC(ip6s_badscope); 664 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 665 goto bad; 666 } 667 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 668 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 669 /* 670 * RFC4291 2.7: 671 * Nodes must not originate a packet to a multicast address 672 * whose scop field contains the reserved value 0; if such 673 * a packet is received, it must be silently dropped. 674 */ 675 IP6STAT_INC(ip6s_badscope); 676 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 677 goto bad; 678 } 679 #ifdef ALTQ 680 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 681 /* packet is dropped by traffic conditioner */ 682 return; 683 } 684 #endif 685 /* 686 * The following check is not documented in specs. A malicious 687 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 688 * and bypass security checks (act as if it was from 127.0.0.1 by using 689 * IPv6 src ::ffff:127.0.0.1). Be cautious. 690 * 691 * We have supported IPv6-only kernels for a few years and this issue 692 * has not come up. The world seems to move mostly towards not using 693 * v4mapped on the wire, so it makes sense for us to keep rejecting 694 * any such packets. 695 */ 696 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 697 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 698 IP6STAT_INC(ip6s_badscope); 699 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 700 goto bad; 701 } 702 #if 0 703 /* 704 * Reject packets with IPv4 compatible addresses (auto tunnel). 705 * 706 * The code forbids auto tunnel relay case in RFC1933 (the check is 707 * stronger than RFC1933). We may want to re-enable it if mech-xx 708 * is revised to forbid relaying case. 709 */ 710 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 711 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 712 IP6STAT_INC(ip6s_badscope); 713 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 714 goto bad; 715 } 716 #endif 717 /* 718 * Try to forward the packet, but if we fail continue. 719 * ip6_tryforward() does not generate redirects, so fall 720 * through to normal processing if redirects are required. 721 * ip6_tryforward() does inbound and outbound packet firewall 722 * processing. If firewall has decided that destination becomes 723 * our local address, it sets M_FASTFWD_OURS flag. In this 724 * case skip another inbound firewall processing and update 725 * ip6 pointer. 726 */ 727 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0 728 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 729 && (!IPSEC_ENABLED(ipv6) || 730 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0) 731 #endif 732 ) { 733 if ((m = ip6_tryforward(m)) == NULL) 734 return; 735 if (m->m_flags & M_FASTFWD_OURS) { 736 ip6 = mtod(m, struct ip6_hdr *); 737 goto passin; 738 } 739 } 740 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 741 /* 742 * Bypass packet filtering for packets previously handled by IPsec. 743 */ 744 if (IPSEC_ENABLED(ipv6) && 745 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0) 746 goto passin; 747 #endif 748 /* 749 * Run through list of hooks for input packets. 750 * 751 * NB: Beware of the destination address changing 752 * (e.g. by NAT rewriting). When this happens, 753 * tell ip6_forward to do the right thing. 754 */ 755 756 /* Jump over all PFIL processing if hooks are not active. */ 757 if (!PFIL_HOOKED_IN(V_inet6_pfil_head)) 758 goto passin; 759 760 odst = ip6->ip6_dst; 761 if (pfil_run_hooks(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif, PFIL_IN, 762 NULL) != PFIL_PASS) 763 return; 764 ip6 = mtod(m, struct ip6_hdr *); 765 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 766 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 767 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 768 /* 769 * Directly ship the packet on. This allows forwarding 770 * packets originally destined to us to some other directly 771 * connected host. 772 */ 773 ip6_forward(m, 1); 774 return; 775 } 776 777 passin: 778 /* 779 * Disambiguate address scope zones (if there is ambiguity). 780 * We first make sure that the original source or destination address 781 * is not in our internal form for scoped addresses. Such addresses 782 * are not necessarily invalid spec-wise, but we cannot accept them due 783 * to the usage conflict. 784 * in6_setscope() then also checks and rejects the cases where src or 785 * dst are the loopback address and the receiving interface 786 * is not loopback. 787 */ 788 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 789 IP6STAT_INC(ip6s_badscope); /* XXX */ 790 goto bad; 791 } 792 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 793 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 794 IP6STAT_INC(ip6s_badscope); 795 goto bad; 796 } 797 if (m->m_flags & M_FASTFWD_OURS) { 798 m->m_flags &= ~M_FASTFWD_OURS; 799 ours = 1; 800 goto hbhcheck; 801 } 802 /* 803 * Multicast check. Assume packet is for us to avoid 804 * prematurely taking locks. 805 */ 806 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 807 ours = 1; 808 in6_ifstat_inc(rcvif, ifs6_in_mcast); 809 goto hbhcheck; 810 } 811 /* 812 * Unicast check 813 * XXX: For now we keep link-local IPv6 addresses with embedded 814 * scope zone id, therefore we use zero zoneid here. 815 */ 816 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false); 817 if (ia != NULL) { 818 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 819 char ip6bufs[INET6_ADDRSTRLEN]; 820 char ip6bufd[INET6_ADDRSTRLEN]; 821 /* address is not ready, so discard the packet. */ 822 nd6log((LOG_INFO, 823 "ip6_input: packet to an unready address %s->%s\n", 824 ip6_sprintf(ip6bufs, &ip6->ip6_src), 825 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 826 goto bad; 827 } 828 if (V_ip6_sav && !(rcvif->if_flags & IFF_LOOPBACK) && 829 __predict_false(in6_localip_fib(&ip6->ip6_src, 830 rcvif->if_fib))) { 831 IP6STAT_INC(ip6s_badscope); /* XXX */ 832 goto bad; 833 } 834 /* Count the packet in the ip address stats */ 835 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 836 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 837 ours = 1; 838 goto hbhcheck; 839 } 840 841 /* 842 * Now there is no reason to process the packet if it's not our own 843 * and we're not a router. 844 */ 845 if (!V_ip6_forwarding) { 846 IP6STAT_INC(ip6s_cantforward); 847 goto bad; 848 } 849 850 hbhcheck: 851 /* 852 * Process Hop-by-Hop options header if it's contained. 853 * m may be modified in ip6_hopopts_input(). 854 * If a JumboPayload option is included, plen will also be modified. 855 */ 856 plen = (u_int32_t)ntohs(ip6->ip6_plen); 857 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 858 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0) 859 return; 860 } else 861 nxt = ip6->ip6_nxt; 862 863 /* 864 * Use mbuf flags to propagate Router Alert option to 865 * ICMPv6 layer, as hop-by-hop options have been stripped. 866 */ 867 if (rtalert != ~0) 868 m->m_flags |= M_RTALERT_MLD; 869 870 /* 871 * Check that the amount of data in the buffers 872 * is as at least much as the IPv6 header would have us expect. 873 * Trim mbufs if longer than we expect. 874 * Drop packet if shorter than we expect. 875 */ 876 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 877 IP6STAT_INC(ip6s_tooshort); 878 in6_ifstat_inc(rcvif, ifs6_in_truncated); 879 goto bad; 880 } 881 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 882 if (m->m_len == m->m_pkthdr.len) { 883 m->m_len = sizeof(struct ip6_hdr) + plen; 884 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 885 } else 886 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 887 } 888 889 /* 890 * Forward if desirable. 891 */ 892 if (V_ip6_mrouter && 893 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 894 /* 895 * If we are acting as a multicast router, all 896 * incoming multicast packets are passed to the 897 * kernel-level multicast forwarding function. 898 * The packet is returned (relatively) intact; if 899 * ip6_mforward() returns a non-zero value, the packet 900 * must be discarded, else it may be accepted below. 901 * 902 * XXX TODO: Check hlim and multicast scope here to avoid 903 * unnecessarily calling into ip6_mforward(). 904 */ 905 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 906 IP6STAT_INC(ip6s_cantforward); 907 goto bad; 908 } 909 } else if (!ours) { 910 ip6_forward(m, srcrt); 911 return; 912 } 913 914 /* 915 * Tell launch routine the next header 916 */ 917 IP6STAT_INC(ip6s_delivered); 918 in6_ifstat_inc(rcvif, ifs6_in_deliver); 919 nest = 0; 920 921 while (nxt != IPPROTO_DONE) { 922 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 923 IP6STAT_INC(ip6s_toomanyhdr); 924 goto bad; 925 } 926 927 /* 928 * protection against faulty packet - there should be 929 * more sanity checks in header chain processing. 930 */ 931 if (m->m_pkthdr.len < off) { 932 IP6STAT_INC(ip6s_tooshort); 933 in6_ifstat_inc(rcvif, ifs6_in_truncated); 934 goto bad; 935 } 936 937 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 938 if (IPSEC_ENABLED(ipv6)) { 939 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 940 return; 941 } 942 #endif /* IPSEC */ 943 944 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 945 } 946 return; 947 bad: 948 in6_ifstat_inc(rcvif, ifs6_in_discard); 949 if (m != NULL) 950 m_freem(m); 951 } 952 953 /* 954 * Hop-by-Hop options header processing. If a valid jumbo payload option is 955 * included, the real payload length will be stored in plenp. 956 * 957 * rtalertp - XXX: should be stored more smart way 958 */ 959 static int 960 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 961 struct mbuf **mp, int *offp) 962 { 963 struct mbuf *m = *mp; 964 int off = *offp, hbhlen; 965 struct ip6_hbh *hbh; 966 967 /* validation of the length of the header */ 968 if (m->m_len < off + sizeof(*hbh)) { 969 m = m_pullup(m, off + sizeof(*hbh)); 970 if (m == NULL) { 971 IP6STAT_INC(ip6s_exthdrtoolong); 972 *mp = NULL; 973 return (-1); 974 } 975 } 976 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 977 hbhlen = (hbh->ip6h_len + 1) << 3; 978 979 if (m->m_len < off + hbhlen) { 980 m = m_pullup(m, off + hbhlen); 981 if (m == NULL) { 982 IP6STAT_INC(ip6s_exthdrtoolong); 983 *mp = NULL; 984 return (-1); 985 } 986 } 987 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 988 off += hbhlen; 989 hbhlen -= sizeof(struct ip6_hbh); 990 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 991 hbhlen, rtalertp, plenp) < 0) { 992 *mp = NULL; 993 return (-1); 994 } 995 996 *offp = off; 997 *mp = m; 998 return (0); 999 } 1000 1001 /* 1002 * Search header for all Hop-by-hop options and process each option. 1003 * This function is separate from ip6_hopopts_input() in order to 1004 * handle a case where the sending node itself process its hop-by-hop 1005 * options header. In such a case, the function is called from ip6_output(). 1006 * 1007 * The function assumes that hbh header is located right after the IPv6 header 1008 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1009 * opthead + hbhlen is located in contiguous memory region. 1010 */ 1011 int 1012 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1013 u_int32_t *rtalertp, u_int32_t *plenp) 1014 { 1015 struct ip6_hdr *ip6; 1016 int optlen = 0; 1017 u_int8_t *opt = opthead; 1018 u_int16_t rtalert_val; 1019 u_int32_t jumboplen; 1020 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1021 1022 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1023 switch (*opt) { 1024 case IP6OPT_PAD1: 1025 optlen = 1; 1026 break; 1027 case IP6OPT_PADN: 1028 if (hbhlen < IP6OPT_MINLEN) { 1029 IP6STAT_INC(ip6s_toosmall); 1030 goto bad; 1031 } 1032 optlen = *(opt + 1) + 2; 1033 break; 1034 case IP6OPT_ROUTER_ALERT: 1035 /* XXX may need check for alignment */ 1036 if (hbhlen < IP6OPT_RTALERT_LEN) { 1037 IP6STAT_INC(ip6s_toosmall); 1038 goto bad; 1039 } 1040 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1041 /* XXX stat */ 1042 icmp6_error(m, ICMP6_PARAM_PROB, 1043 ICMP6_PARAMPROB_HEADER, 1044 erroff + opt + 1 - opthead); 1045 return (-1); 1046 } 1047 optlen = IP6OPT_RTALERT_LEN; 1048 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1049 *rtalertp = ntohs(rtalert_val); 1050 break; 1051 case IP6OPT_JUMBO: 1052 /* XXX may need check for alignment */ 1053 if (hbhlen < IP6OPT_JUMBO_LEN) { 1054 IP6STAT_INC(ip6s_toosmall); 1055 goto bad; 1056 } 1057 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1058 /* XXX stat */ 1059 icmp6_error(m, ICMP6_PARAM_PROB, 1060 ICMP6_PARAMPROB_HEADER, 1061 erroff + opt + 1 - opthead); 1062 return (-1); 1063 } 1064 optlen = IP6OPT_JUMBO_LEN; 1065 1066 /* 1067 * IPv6 packets that have non 0 payload length 1068 * must not contain a jumbo payload option. 1069 */ 1070 ip6 = mtod(m, struct ip6_hdr *); 1071 if (ip6->ip6_plen) { 1072 IP6STAT_INC(ip6s_badoptions); 1073 icmp6_error(m, ICMP6_PARAM_PROB, 1074 ICMP6_PARAMPROB_HEADER, 1075 erroff + opt - opthead); 1076 return (-1); 1077 } 1078 1079 /* 1080 * We may see jumbolen in unaligned location, so 1081 * we'd need to perform bcopy(). 1082 */ 1083 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1084 jumboplen = (u_int32_t)htonl(jumboplen); 1085 1086 #if 1 1087 /* 1088 * if there are multiple jumbo payload options, 1089 * *plenp will be non-zero and the packet will be 1090 * rejected. 1091 * the behavior may need some debate in ipngwg - 1092 * multiple options does not make sense, however, 1093 * there's no explicit mention in specification. 1094 */ 1095 if (*plenp != 0) { 1096 IP6STAT_INC(ip6s_badoptions); 1097 icmp6_error(m, ICMP6_PARAM_PROB, 1098 ICMP6_PARAMPROB_HEADER, 1099 erroff + opt + 2 - opthead); 1100 return (-1); 1101 } 1102 #endif 1103 1104 /* 1105 * jumbo payload length must be larger than 65535. 1106 */ 1107 if (jumboplen <= IPV6_MAXPACKET) { 1108 IP6STAT_INC(ip6s_badoptions); 1109 icmp6_error(m, ICMP6_PARAM_PROB, 1110 ICMP6_PARAMPROB_HEADER, 1111 erroff + opt + 2 - opthead); 1112 return (-1); 1113 } 1114 *plenp = jumboplen; 1115 1116 break; 1117 default: /* unknown option */ 1118 if (hbhlen < IP6OPT_MINLEN) { 1119 IP6STAT_INC(ip6s_toosmall); 1120 goto bad; 1121 } 1122 optlen = ip6_unknown_opt(opt, m, 1123 erroff + opt - opthead); 1124 if (optlen == -1) 1125 return (-1); 1126 optlen += 2; 1127 break; 1128 } 1129 } 1130 1131 return (0); 1132 1133 bad: 1134 m_freem(m); 1135 return (-1); 1136 } 1137 1138 /* 1139 * Unknown option processing. 1140 * The third argument `off' is the offset from the IPv6 header to the option, 1141 * which is necessary if the IPv6 header the and option header and IPv6 header 1142 * is not contiguous in order to return an ICMPv6 error. 1143 */ 1144 int 1145 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1146 { 1147 struct ip6_hdr *ip6; 1148 1149 switch (IP6OPT_TYPE(*optp)) { 1150 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1151 return ((int)*(optp + 1)); 1152 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1153 m_freem(m); 1154 return (-1); 1155 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1156 IP6STAT_INC(ip6s_badoptions); 1157 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1158 return (-1); 1159 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1160 IP6STAT_INC(ip6s_badoptions); 1161 ip6 = mtod(m, struct ip6_hdr *); 1162 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1163 (m->m_flags & (M_BCAST|M_MCAST))) 1164 m_freem(m); 1165 else 1166 icmp6_error(m, ICMP6_PARAM_PROB, 1167 ICMP6_PARAMPROB_OPTION, off); 1168 return (-1); 1169 } 1170 1171 m_freem(m); /* XXX: NOTREACHED */ 1172 return (-1); 1173 } 1174 1175 /* 1176 * Create the "control" list for this pcb. 1177 * These functions will not modify mbuf chain at all. 1178 * 1179 * The routine will be called from upper layer handlers like tcp6_input(). 1180 * Thus the routine assumes that the caller (tcp6_input) have already 1181 * called m_pullup() and all the extension headers are located in the 1182 * very first mbuf on the mbuf chain. 1183 * 1184 * ip6_savecontrol_v4 will handle those options that are possible to be 1185 * set on a v4-mapped socket. 1186 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1187 * options and handle the v6-only ones itself. 1188 */ 1189 struct mbuf ** 1190 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1191 int *v4only) 1192 { 1193 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1194 1195 #ifdef SO_TIMESTAMP 1196 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1197 union { 1198 struct timeval tv; 1199 struct bintime bt; 1200 struct timespec ts; 1201 } t; 1202 struct bintime boottimebin, bt1; 1203 struct timespec ts1; 1204 bool stamped; 1205 1206 stamped = false; 1207 switch (inp->inp_socket->so_ts_clock) { 1208 case SO_TS_REALTIME_MICRO: 1209 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1210 M_TSTMP)) { 1211 mbuf_tstmp2timespec(m, &ts1); 1212 timespec2bintime(&ts1, &bt1); 1213 getboottimebin(&boottimebin); 1214 bintime_add(&bt1, &boottimebin); 1215 bintime2timeval(&bt1, &t.tv); 1216 } else { 1217 microtime(&t.tv); 1218 } 1219 *mp = sbcreatecontrol((caddr_t) &t.tv, sizeof(t.tv), 1220 SCM_TIMESTAMP, SOL_SOCKET); 1221 if (*mp != NULL) { 1222 mp = &(*mp)->m_next; 1223 stamped = true; 1224 } 1225 break; 1226 1227 case SO_TS_BINTIME: 1228 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1229 M_TSTMP)) { 1230 mbuf_tstmp2timespec(m, &ts1); 1231 timespec2bintime(&ts1, &t.bt); 1232 getboottimebin(&boottimebin); 1233 bintime_add(&t.bt, &boottimebin); 1234 } else { 1235 bintime(&t.bt); 1236 } 1237 *mp = sbcreatecontrol((caddr_t)&t.bt, sizeof(t.bt), 1238 SCM_BINTIME, SOL_SOCKET); 1239 if (*mp != NULL) { 1240 mp = &(*mp)->m_next; 1241 stamped = true; 1242 } 1243 break; 1244 1245 case SO_TS_REALTIME: 1246 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1247 M_TSTMP)) { 1248 mbuf_tstmp2timespec(m, &t.ts); 1249 getboottimebin(&boottimebin); 1250 bintime2timespec(&boottimebin, &ts1); 1251 timespecadd(&t.ts, &ts1, &t.ts); 1252 } else { 1253 nanotime(&t.ts); 1254 } 1255 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts), 1256 SCM_REALTIME, SOL_SOCKET); 1257 if (*mp != NULL) { 1258 mp = &(*mp)->m_next; 1259 stamped = true; 1260 } 1261 break; 1262 1263 case SO_TS_MONOTONIC: 1264 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1265 M_TSTMP)) 1266 mbuf_tstmp2timespec(m, &t.ts); 1267 else 1268 nanouptime(&t.ts); 1269 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts), 1270 SCM_MONOTONIC, SOL_SOCKET); 1271 if (*mp != NULL) { 1272 mp = &(*mp)->m_next; 1273 stamped = true; 1274 } 1275 break; 1276 1277 default: 1278 panic("unknown (corrupted) so_ts_clock"); 1279 } 1280 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == 1281 (M_PKTHDR | M_TSTMP)) { 1282 struct sock_timestamp_info sti; 1283 1284 bzero(&sti, sizeof(sti)); 1285 sti.st_info_flags = ST_INFO_HW; 1286 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1287 sti.st_info_flags |= ST_INFO_HW_HPREC; 1288 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), 1289 SCM_TIME_INFO, SOL_SOCKET); 1290 if (*mp != NULL) 1291 mp = &(*mp)->m_next; 1292 } 1293 } 1294 #endif 1295 1296 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1297 /* RFC 2292 sec. 5 */ 1298 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1299 struct in6_pktinfo pi6; 1300 1301 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1302 #ifdef INET 1303 struct ip *ip; 1304 1305 ip = mtod(m, struct ip *); 1306 pi6.ipi6_addr.s6_addr32[0] = 0; 1307 pi6.ipi6_addr.s6_addr32[1] = 0; 1308 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1309 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1310 #else 1311 /* We won't hit this code */ 1312 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1313 #endif 1314 } else { 1315 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1316 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1317 } 1318 pi6.ipi6_ifindex = 1319 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1320 1321 *mp = sbcreatecontrol((caddr_t) &pi6, 1322 sizeof(struct in6_pktinfo), 1323 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1324 if (*mp) 1325 mp = &(*mp)->m_next; 1326 } 1327 1328 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1329 int hlim; 1330 1331 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1332 #ifdef INET 1333 struct ip *ip; 1334 1335 ip = mtod(m, struct ip *); 1336 hlim = ip->ip_ttl; 1337 #else 1338 /* We won't hit this code */ 1339 hlim = 0; 1340 #endif 1341 } else { 1342 hlim = ip6->ip6_hlim & 0xff; 1343 } 1344 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1345 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1346 IPPROTO_IPV6); 1347 if (*mp) 1348 mp = &(*mp)->m_next; 1349 } 1350 1351 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1352 int tclass; 1353 1354 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1355 #ifdef INET 1356 struct ip *ip; 1357 1358 ip = mtod(m, struct ip *); 1359 tclass = ip->ip_tos; 1360 #else 1361 /* We won't hit this code */ 1362 tclass = 0; 1363 #endif 1364 } else { 1365 u_int32_t flowinfo; 1366 1367 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1368 flowinfo >>= 20; 1369 tclass = flowinfo & 0xff; 1370 } 1371 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int), 1372 IPV6_TCLASS, IPPROTO_IPV6); 1373 if (*mp) 1374 mp = &(*mp)->m_next; 1375 } 1376 1377 if (v4only != NULL) { 1378 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1379 *v4only = 1; 1380 } else { 1381 *v4only = 0; 1382 } 1383 } 1384 1385 return (mp); 1386 } 1387 1388 void 1389 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp) 1390 { 1391 struct ip6_hdr *ip6; 1392 int v4only = 0; 1393 1394 mp = ip6_savecontrol_v4(inp, m, mp, &v4only); 1395 if (v4only) 1396 return; 1397 1398 ip6 = mtod(m, struct ip6_hdr *); 1399 /* 1400 * IPV6_HOPOPTS socket option. Recall that we required super-user 1401 * privilege for the option (see ip6_ctloutput), but it might be too 1402 * strict, since there might be some hop-by-hop options which can be 1403 * returned to normal user. 1404 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1405 */ 1406 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) { 1407 /* 1408 * Check if a hop-by-hop options header is contatined in the 1409 * received packet, and if so, store the options as ancillary 1410 * data. Note that a hop-by-hop options header must be 1411 * just after the IPv6 header, which is assured through the 1412 * IPv6 input processing. 1413 */ 1414 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1415 struct ip6_hbh *hbh; 1416 int hbhlen; 1417 1418 hbh = (struct ip6_hbh *)(ip6 + 1); 1419 hbhlen = (hbh->ip6h_len + 1) << 3; 1420 1421 /* 1422 * XXX: We copy the whole header even if a 1423 * jumbo payload option is included, the option which 1424 * is to be removed before returning according to 1425 * RFC2292. 1426 * Note: this constraint is removed in RFC3542 1427 */ 1428 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1429 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1430 IPPROTO_IPV6); 1431 if (*mp) 1432 mp = &(*mp)->m_next; 1433 } 1434 } 1435 1436 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1437 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1438 1439 /* 1440 * Search for destination options headers or routing 1441 * header(s) through the header chain, and stores each 1442 * header as ancillary data. 1443 * Note that the order of the headers remains in 1444 * the chain of ancillary data. 1445 */ 1446 while (1) { /* is explicit loop prevention necessary? */ 1447 struct ip6_ext *ip6e = NULL; 1448 int elen; 1449 1450 /* 1451 * if it is not an extension header, don't try to 1452 * pull it from the chain. 1453 */ 1454 switch (nxt) { 1455 case IPPROTO_DSTOPTS: 1456 case IPPROTO_ROUTING: 1457 case IPPROTO_HOPOPTS: 1458 case IPPROTO_AH: /* is it possible? */ 1459 break; 1460 default: 1461 goto loopend; 1462 } 1463 1464 if (off + sizeof(*ip6e) > m->m_len) 1465 goto loopend; 1466 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1467 if (nxt == IPPROTO_AH) 1468 elen = (ip6e->ip6e_len + 2) << 2; 1469 else 1470 elen = (ip6e->ip6e_len + 1) << 3; 1471 if (off + elen > m->m_len) 1472 goto loopend; 1473 1474 switch (nxt) { 1475 case IPPROTO_DSTOPTS: 1476 if (!(inp->inp_flags & IN6P_DSTOPTS)) 1477 break; 1478 1479 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1480 IS2292(inp, 1481 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1482 IPPROTO_IPV6); 1483 if (*mp) 1484 mp = &(*mp)->m_next; 1485 break; 1486 case IPPROTO_ROUTING: 1487 if (!(inp->inp_flags & IN6P_RTHDR)) 1488 break; 1489 1490 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1491 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR), 1492 IPPROTO_IPV6); 1493 if (*mp) 1494 mp = &(*mp)->m_next; 1495 break; 1496 case IPPROTO_HOPOPTS: 1497 case IPPROTO_AH: /* is it possible? */ 1498 break; 1499 1500 default: 1501 /* 1502 * other cases have been filtered in the above. 1503 * none will visit this case. here we supply 1504 * the code just in case (nxt overwritten or 1505 * other cases). 1506 */ 1507 goto loopend; 1508 } 1509 1510 /* proceed with the next header. */ 1511 off += elen; 1512 nxt = ip6e->ip6e_nxt; 1513 ip6e = NULL; 1514 } 1515 loopend: 1516 ; 1517 } 1518 1519 if (inp->inp_flags2 & INP_RECVFLOWID) { 1520 uint32_t flowid, flow_type; 1521 1522 flowid = m->m_pkthdr.flowid; 1523 flow_type = M_HASHTYPE_GET(m); 1524 1525 /* 1526 * XXX should handle the failure of one or the 1527 * other - don't populate both? 1528 */ 1529 *mp = sbcreatecontrol((caddr_t) &flowid, 1530 sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6); 1531 if (*mp) 1532 mp = &(*mp)->m_next; 1533 *mp = sbcreatecontrol((caddr_t) &flow_type, 1534 sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6); 1535 if (*mp) 1536 mp = &(*mp)->m_next; 1537 } 1538 1539 #ifdef RSS 1540 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1541 uint32_t flowid, flow_type; 1542 uint32_t rss_bucketid; 1543 1544 flowid = m->m_pkthdr.flowid; 1545 flow_type = M_HASHTYPE_GET(m); 1546 1547 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1548 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1549 sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6); 1550 if (*mp) 1551 mp = &(*mp)->m_next; 1552 } 1553 } 1554 #endif 1555 1556 } 1557 #undef IS2292 1558 1559 void 1560 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1561 { 1562 struct socket *so; 1563 struct mbuf *m_mtu; 1564 struct ip6_mtuinfo mtuctl; 1565 1566 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1567 /* 1568 * Notify the error by sending IPV6_PATHMTU ancillary data if 1569 * application wanted to know the MTU value. 1570 * NOTE: we notify disconnected sockets, because some udp 1571 * applications keep sending sockets disconnected. 1572 * NOTE: our implementation doesn't notify connected sockets that has 1573 * foreign address that is different than given destination addresses 1574 * (this is permitted by RFC 3542). 1575 */ 1576 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1577 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1578 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1579 return; 1580 1581 mtuctl.ip6m_mtu = mtu; 1582 mtuctl.ip6m_addr = *dst; 1583 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1584 return; 1585 1586 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1587 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1588 return; 1589 1590 so = inp->inp_socket; 1591 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1592 == 0) { 1593 soroverflow(so); 1594 m_freem(m_mtu); 1595 /* XXX: should count statistics */ 1596 } else 1597 sorwakeup(so); 1598 } 1599 1600 /* 1601 * Get pointer to the previous header followed by the header 1602 * currently processed. 1603 */ 1604 int 1605 ip6_get_prevhdr(const struct mbuf *m, int off) 1606 { 1607 struct ip6_ext ip6e; 1608 struct ip6_hdr *ip6; 1609 int len, nlen, nxt; 1610 1611 if (off == sizeof(struct ip6_hdr)) 1612 return (offsetof(struct ip6_hdr, ip6_nxt)); 1613 if (off < sizeof(struct ip6_hdr)) 1614 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1615 1616 ip6 = mtod(m, struct ip6_hdr *); 1617 nxt = ip6->ip6_nxt; 1618 len = sizeof(struct ip6_hdr); 1619 nlen = 0; 1620 while (len < off) { 1621 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1622 switch (nxt) { 1623 case IPPROTO_FRAGMENT: 1624 nlen = sizeof(struct ip6_frag); 1625 break; 1626 case IPPROTO_AH: 1627 nlen = (ip6e.ip6e_len + 2) << 2; 1628 break; 1629 default: 1630 nlen = (ip6e.ip6e_len + 1) << 3; 1631 } 1632 len += nlen; 1633 nxt = ip6e.ip6e_nxt; 1634 } 1635 return (len - nlen); 1636 } 1637 1638 /* 1639 * get next header offset. m will be retained. 1640 */ 1641 int 1642 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1643 { 1644 struct ip6_hdr ip6; 1645 struct ip6_ext ip6e; 1646 struct ip6_frag fh; 1647 1648 /* just in case */ 1649 if (m == NULL) 1650 panic("ip6_nexthdr: m == NULL"); 1651 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1652 return -1; 1653 1654 switch (proto) { 1655 case IPPROTO_IPV6: 1656 if (m->m_pkthdr.len < off + sizeof(ip6)) 1657 return -1; 1658 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1659 if (nxtp) 1660 *nxtp = ip6.ip6_nxt; 1661 off += sizeof(ip6); 1662 return off; 1663 1664 case IPPROTO_FRAGMENT: 1665 /* 1666 * terminate parsing if it is not the first fragment, 1667 * it does not make sense to parse through it. 1668 */ 1669 if (m->m_pkthdr.len < off + sizeof(fh)) 1670 return -1; 1671 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1672 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1673 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1674 return -1; 1675 if (nxtp) 1676 *nxtp = fh.ip6f_nxt; 1677 off += sizeof(struct ip6_frag); 1678 return off; 1679 1680 case IPPROTO_AH: 1681 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1682 return -1; 1683 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1684 if (nxtp) 1685 *nxtp = ip6e.ip6e_nxt; 1686 off += (ip6e.ip6e_len + 2) << 2; 1687 return off; 1688 1689 case IPPROTO_HOPOPTS: 1690 case IPPROTO_ROUTING: 1691 case IPPROTO_DSTOPTS: 1692 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1693 return -1; 1694 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1695 if (nxtp) 1696 *nxtp = ip6e.ip6e_nxt; 1697 off += (ip6e.ip6e_len + 1) << 3; 1698 return off; 1699 1700 case IPPROTO_NONE: 1701 case IPPROTO_ESP: 1702 case IPPROTO_IPCOMP: 1703 /* give up */ 1704 return -1; 1705 1706 default: 1707 return -1; 1708 } 1709 1710 /* NOTREACHED */ 1711 } 1712 1713 /* 1714 * get offset for the last header in the chain. m will be kept untainted. 1715 */ 1716 int 1717 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1718 { 1719 int newoff; 1720 int nxt; 1721 1722 if (!nxtp) { 1723 nxt = -1; 1724 nxtp = &nxt; 1725 } 1726 while (1) { 1727 newoff = ip6_nexthdr(m, off, proto, nxtp); 1728 if (newoff < 0) 1729 return off; 1730 else if (newoff < off) 1731 return -1; /* invalid */ 1732 else if (newoff == off) 1733 return newoff; 1734 1735 off = newoff; 1736 proto = *nxtp; 1737 } 1738 } 1739 1740 /* 1741 * System control for IP6 1742 */ 1743 1744 u_char inet6ctlerrmap[PRC_NCMDS] = { 1745 0, 0, 0, 0, 1746 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1747 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1748 EMSGSIZE, EHOSTUNREACH, 0, 0, 1749 0, 0, EHOSTUNREACH, 0, 1750 ENOPROTOOPT, ECONNREFUSED 1751 }; 1752