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 struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif; 592 int ifindex = ifp->if_index; 593 if (ifindex >= IP6S_M2MMAX) 594 ifindex = 0; 595 IP6STAT_INC(ip6s_m2m[ifindex]); 596 } else 597 IP6STAT_INC(ip6s_m1); 598 } 599 600 in6_ifstat_inc(rcvif, ifs6_in_receive); 601 IP6STAT_INC(ip6s_total); 602 603 /* 604 * L2 bridge code and some other code can return mbuf chain 605 * that does not conform to KAME requirement. too bad. 606 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 607 */ 608 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 609 struct mbuf *n; 610 611 if (m->m_pkthdr.len > MHLEN) 612 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 613 else 614 n = m_gethdr(M_NOWAIT, MT_DATA); 615 if (n == NULL) 616 goto bad; 617 618 m_move_pkthdr(n, m); 619 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 620 n->m_len = n->m_pkthdr.len; 621 m_freem(m); 622 m = n; 623 } 624 if (m->m_len < sizeof(struct ip6_hdr)) { 625 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 626 IP6STAT_INC(ip6s_toosmall); 627 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 628 goto bad; 629 } 630 } 631 632 ip6 = mtod(m, struct ip6_hdr *); 633 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 634 IP6STAT_INC(ip6s_badvers); 635 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 636 goto bad; 637 } 638 639 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 640 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 641 642 /* 643 * Check against address spoofing/corruption. 644 */ 645 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 646 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 647 /* 648 * XXX: "badscope" is not very suitable for a multicast source. 649 */ 650 IP6STAT_INC(ip6s_badscope); 651 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 652 goto bad; 653 } 654 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 655 !(m->m_flags & M_LOOP)) { 656 /* 657 * In this case, the packet should come from the loopback 658 * interface. However, we cannot just check the if_flags, 659 * because ip6_mloopback() passes the "actual" interface 660 * as the outgoing/incoming interface. 661 */ 662 IP6STAT_INC(ip6s_badscope); 663 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 664 goto bad; 665 } 666 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 667 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 668 /* 669 * RFC4291 2.7: 670 * Nodes must not originate a packet to a multicast address 671 * whose scop field contains the reserved value 0; if such 672 * a packet is received, it must be silently dropped. 673 */ 674 IP6STAT_INC(ip6s_badscope); 675 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 676 goto bad; 677 } 678 #ifdef ALTQ 679 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 680 /* packet is dropped by traffic conditioner */ 681 return; 682 } 683 #endif 684 /* 685 * The following check is not documented in specs. A malicious 686 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 687 * and bypass security checks (act as if it was from 127.0.0.1 by using 688 * IPv6 src ::ffff:127.0.0.1). Be cautious. 689 * 690 * We have supported IPv6-only kernels for a few years and this issue 691 * has not come up. The world seems to move mostly towards not using 692 * v4mapped on the wire, so it makes sense for us to keep rejecting 693 * any such packets. 694 */ 695 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 696 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 697 IP6STAT_INC(ip6s_badscope); 698 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 699 goto bad; 700 } 701 #if 0 702 /* 703 * Reject packets with IPv4 compatible addresses (auto tunnel). 704 * 705 * The code forbids auto tunnel relay case in RFC1933 (the check is 706 * stronger than RFC1933). We may want to re-enable it if mech-xx 707 * is revised to forbid relaying case. 708 */ 709 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 710 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 711 IP6STAT_INC(ip6s_badscope); 712 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 713 goto bad; 714 } 715 #endif 716 /* 717 * Try to forward the packet, but if we fail continue. 718 * ip6_tryforward() does not generate redirects, so fall 719 * through to normal processing if redirects are required. 720 * ip6_tryforward() does inbound and outbound packet firewall 721 * processing. If firewall has decided that destination becomes 722 * our local address, it sets M_FASTFWD_OURS flag. In this 723 * case skip another inbound firewall processing and update 724 * ip6 pointer. 725 */ 726 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0 727 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 728 && (!IPSEC_ENABLED(ipv6) || 729 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0) 730 #endif 731 ) { 732 if ((m = ip6_tryforward(m)) == NULL) 733 return; 734 if (m->m_flags & M_FASTFWD_OURS) { 735 ip6 = mtod(m, struct ip6_hdr *); 736 goto passin; 737 } 738 } 739 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 740 /* 741 * Bypass packet filtering for packets previously handled by IPsec. 742 */ 743 if (IPSEC_ENABLED(ipv6) && 744 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0) 745 goto passin; 746 #endif 747 /* 748 * Run through list of hooks for input packets. 749 * 750 * NB: Beware of the destination address changing 751 * (e.g. by NAT rewriting). When this happens, 752 * tell ip6_forward to do the right thing. 753 */ 754 755 /* Jump over all PFIL processing if hooks are not active. */ 756 if (!PFIL_HOOKED_IN(V_inet6_pfil_head)) 757 goto passin; 758 759 odst = ip6->ip6_dst; 760 if (pfil_run_hooks(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif, PFIL_IN, 761 NULL) != PFIL_PASS) 762 return; 763 ip6 = mtod(m, struct ip6_hdr *); 764 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 765 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 766 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 767 /* 768 * Directly ship the packet on. This allows forwarding 769 * packets originally destined to us to some other directly 770 * connected host. 771 */ 772 ip6_forward(m, 1); 773 return; 774 } 775 776 passin: 777 /* 778 * Disambiguate address scope zones (if there is ambiguity). 779 * We first make sure that the original source or destination address 780 * is not in our internal form for scoped addresses. Such addresses 781 * are not necessarily invalid spec-wise, but we cannot accept them due 782 * to the usage conflict. 783 * in6_setscope() then also checks and rejects the cases where src or 784 * dst are the loopback address and the receiving interface 785 * is not loopback. 786 */ 787 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 788 IP6STAT_INC(ip6s_badscope); /* XXX */ 789 goto bad; 790 } 791 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 792 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 793 IP6STAT_INC(ip6s_badscope); 794 goto bad; 795 } 796 if (m->m_flags & M_FASTFWD_OURS) { 797 m->m_flags &= ~M_FASTFWD_OURS; 798 ours = 1; 799 goto hbhcheck; 800 } 801 /* 802 * Multicast check. Assume packet is for us to avoid 803 * prematurely taking locks. 804 */ 805 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 806 ours = 1; 807 in6_ifstat_inc(rcvif, ifs6_in_mcast); 808 goto hbhcheck; 809 } 810 /* 811 * Unicast check 812 * XXX: For now we keep link-local IPv6 addresses with embedded 813 * scope zone id, therefore we use zero zoneid here. 814 */ 815 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false); 816 if (ia != NULL) { 817 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 818 char ip6bufs[INET6_ADDRSTRLEN]; 819 char ip6bufd[INET6_ADDRSTRLEN]; 820 /* address is not ready, so discard the packet. */ 821 nd6log((LOG_INFO, 822 "ip6_input: packet to an unready address %s->%s\n", 823 ip6_sprintf(ip6bufs, &ip6->ip6_src), 824 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 825 goto bad; 826 } 827 if (V_ip6_sav && !(m->m_flags & M_LOOP) && 828 __predict_false(in6_localip_fib(&ip6->ip6_src, 829 rcvif->if_fib))) { 830 IP6STAT_INC(ip6s_badscope); /* XXX */ 831 goto bad; 832 } 833 /* Count the packet in the ip address stats */ 834 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 835 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 836 ours = 1; 837 goto hbhcheck; 838 } 839 840 /* 841 * Now there is no reason to process the packet if it's not our own 842 * and we're not a router. 843 */ 844 if (!V_ip6_forwarding) { 845 IP6STAT_INC(ip6s_cantforward); 846 goto bad; 847 } 848 849 hbhcheck: 850 /* 851 * Process Hop-by-Hop options header if it's contained. 852 * m may be modified in ip6_hopopts_input(). 853 * If a JumboPayload option is included, plen will also be modified. 854 */ 855 plen = (u_int32_t)ntohs(ip6->ip6_plen); 856 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 857 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0) 858 return; 859 } else 860 nxt = ip6->ip6_nxt; 861 862 /* 863 * Use mbuf flags to propagate Router Alert option to 864 * ICMPv6 layer, as hop-by-hop options have been stripped. 865 */ 866 if (rtalert != ~0) 867 m->m_flags |= M_RTALERT_MLD; 868 869 /* 870 * Check that the amount of data in the buffers 871 * is as at least much as the IPv6 header would have us expect. 872 * Trim mbufs if longer than we expect. 873 * Drop packet if shorter than we expect. 874 */ 875 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 876 IP6STAT_INC(ip6s_tooshort); 877 in6_ifstat_inc(rcvif, ifs6_in_truncated); 878 goto bad; 879 } 880 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 881 if (m->m_len == m->m_pkthdr.len) { 882 m->m_len = sizeof(struct ip6_hdr) + plen; 883 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 884 } else 885 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 886 } 887 888 /* 889 * Forward if desirable. 890 */ 891 if (V_ip6_mrouter && 892 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 893 /* 894 * If we are acting as a multicast router, all 895 * incoming multicast packets are passed to the 896 * kernel-level multicast forwarding function. 897 * The packet is returned (relatively) intact; if 898 * ip6_mforward() returns a non-zero value, the packet 899 * must be discarded, else it may be accepted below. 900 * 901 * XXX TODO: Check hlim and multicast scope here to avoid 902 * unnecessarily calling into ip6_mforward(). 903 */ 904 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 905 IP6STAT_INC(ip6s_cantforward); 906 goto bad; 907 } 908 } else if (!ours) { 909 ip6_forward(m, srcrt); 910 return; 911 } 912 913 /* 914 * Tell launch routine the next header 915 */ 916 IP6STAT_INC(ip6s_delivered); 917 in6_ifstat_inc(rcvif, ifs6_in_deliver); 918 nest = 0; 919 920 while (nxt != IPPROTO_DONE) { 921 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 922 IP6STAT_INC(ip6s_toomanyhdr); 923 goto bad; 924 } 925 926 /* 927 * protection against faulty packet - there should be 928 * more sanity checks in header chain processing. 929 */ 930 if (m->m_pkthdr.len < off) { 931 IP6STAT_INC(ip6s_tooshort); 932 in6_ifstat_inc(rcvif, ifs6_in_truncated); 933 goto bad; 934 } 935 936 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 937 if (IPSEC_ENABLED(ipv6)) { 938 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 939 return; 940 } 941 #endif /* IPSEC */ 942 943 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 944 } 945 return; 946 bad: 947 in6_ifstat_inc(rcvif, ifs6_in_discard); 948 if (m != NULL) 949 m_freem(m); 950 } 951 952 /* 953 * Hop-by-Hop options header processing. If a valid jumbo payload option is 954 * included, the real payload length will be stored in plenp. 955 * 956 * rtalertp - XXX: should be stored more smart way 957 */ 958 static int 959 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 960 struct mbuf **mp, int *offp) 961 { 962 struct mbuf *m = *mp; 963 int off = *offp, hbhlen; 964 struct ip6_hbh *hbh; 965 966 /* validation of the length of the header */ 967 if (m->m_len < off + sizeof(*hbh)) { 968 m = m_pullup(m, off + sizeof(*hbh)); 969 if (m == NULL) { 970 IP6STAT_INC(ip6s_exthdrtoolong); 971 *mp = NULL; 972 return (-1); 973 } 974 } 975 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 976 hbhlen = (hbh->ip6h_len + 1) << 3; 977 978 if (m->m_len < off + hbhlen) { 979 m = m_pullup(m, off + hbhlen); 980 if (m == NULL) { 981 IP6STAT_INC(ip6s_exthdrtoolong); 982 *mp = NULL; 983 return (-1); 984 } 985 } 986 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 987 off += hbhlen; 988 hbhlen -= sizeof(struct ip6_hbh); 989 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 990 hbhlen, rtalertp, plenp) < 0) { 991 *mp = NULL; 992 return (-1); 993 } 994 995 *offp = off; 996 *mp = m; 997 return (0); 998 } 999 1000 /* 1001 * Search header for all Hop-by-hop options and process each option. 1002 * This function is separate from ip6_hopopts_input() in order to 1003 * handle a case where the sending node itself process its hop-by-hop 1004 * options header. In such a case, the function is called from ip6_output(). 1005 * 1006 * The function assumes that hbh header is located right after the IPv6 header 1007 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1008 * opthead + hbhlen is located in contiguous memory region. 1009 */ 1010 int 1011 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1012 u_int32_t *rtalertp, u_int32_t *plenp) 1013 { 1014 struct ip6_hdr *ip6; 1015 int optlen = 0; 1016 u_int8_t *opt = opthead; 1017 u_int16_t rtalert_val; 1018 u_int32_t jumboplen; 1019 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1020 1021 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1022 switch (*opt) { 1023 case IP6OPT_PAD1: 1024 optlen = 1; 1025 break; 1026 case IP6OPT_PADN: 1027 if (hbhlen < IP6OPT_MINLEN) { 1028 IP6STAT_INC(ip6s_toosmall); 1029 goto bad; 1030 } 1031 optlen = *(opt + 1) + 2; 1032 break; 1033 case IP6OPT_ROUTER_ALERT: 1034 /* XXX may need check for alignment */ 1035 if (hbhlen < IP6OPT_RTALERT_LEN) { 1036 IP6STAT_INC(ip6s_toosmall); 1037 goto bad; 1038 } 1039 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1040 /* XXX stat */ 1041 icmp6_error(m, ICMP6_PARAM_PROB, 1042 ICMP6_PARAMPROB_HEADER, 1043 erroff + opt + 1 - opthead); 1044 return (-1); 1045 } 1046 optlen = IP6OPT_RTALERT_LEN; 1047 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1048 *rtalertp = ntohs(rtalert_val); 1049 break; 1050 case IP6OPT_JUMBO: 1051 /* XXX may need check for alignment */ 1052 if (hbhlen < IP6OPT_JUMBO_LEN) { 1053 IP6STAT_INC(ip6s_toosmall); 1054 goto bad; 1055 } 1056 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1057 /* XXX stat */ 1058 icmp6_error(m, ICMP6_PARAM_PROB, 1059 ICMP6_PARAMPROB_HEADER, 1060 erroff + opt + 1 - opthead); 1061 return (-1); 1062 } 1063 optlen = IP6OPT_JUMBO_LEN; 1064 1065 /* 1066 * IPv6 packets that have non 0 payload length 1067 * must not contain a jumbo payload option. 1068 */ 1069 ip6 = mtod(m, struct ip6_hdr *); 1070 if (ip6->ip6_plen) { 1071 IP6STAT_INC(ip6s_badoptions); 1072 icmp6_error(m, ICMP6_PARAM_PROB, 1073 ICMP6_PARAMPROB_HEADER, 1074 erroff + opt - opthead); 1075 return (-1); 1076 } 1077 1078 /* 1079 * We may see jumbolen in unaligned location, so 1080 * we'd need to perform bcopy(). 1081 */ 1082 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1083 jumboplen = (u_int32_t)htonl(jumboplen); 1084 1085 #if 1 1086 /* 1087 * if there are multiple jumbo payload options, 1088 * *plenp will be non-zero and the packet will be 1089 * rejected. 1090 * the behavior may need some debate in ipngwg - 1091 * multiple options does not make sense, however, 1092 * there's no explicit mention in specification. 1093 */ 1094 if (*plenp != 0) { 1095 IP6STAT_INC(ip6s_badoptions); 1096 icmp6_error(m, ICMP6_PARAM_PROB, 1097 ICMP6_PARAMPROB_HEADER, 1098 erroff + opt + 2 - opthead); 1099 return (-1); 1100 } 1101 #endif 1102 1103 /* 1104 * jumbo payload length must be larger than 65535. 1105 */ 1106 if (jumboplen <= IPV6_MAXPACKET) { 1107 IP6STAT_INC(ip6s_badoptions); 1108 icmp6_error(m, ICMP6_PARAM_PROB, 1109 ICMP6_PARAMPROB_HEADER, 1110 erroff + opt + 2 - opthead); 1111 return (-1); 1112 } 1113 *plenp = jumboplen; 1114 1115 break; 1116 default: /* unknown option */ 1117 if (hbhlen < IP6OPT_MINLEN) { 1118 IP6STAT_INC(ip6s_toosmall); 1119 goto bad; 1120 } 1121 optlen = ip6_unknown_opt(opt, m, 1122 erroff + opt - opthead); 1123 if (optlen == -1) 1124 return (-1); 1125 optlen += 2; 1126 break; 1127 } 1128 } 1129 1130 return (0); 1131 1132 bad: 1133 m_freem(m); 1134 return (-1); 1135 } 1136 1137 /* 1138 * Unknown option processing. 1139 * The third argument `off' is the offset from the IPv6 header to the option, 1140 * which is necessary if the IPv6 header the and option header and IPv6 header 1141 * is not contiguous in order to return an ICMPv6 error. 1142 */ 1143 int 1144 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1145 { 1146 struct ip6_hdr *ip6; 1147 1148 switch (IP6OPT_TYPE(*optp)) { 1149 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1150 return ((int)*(optp + 1)); 1151 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1152 m_freem(m); 1153 return (-1); 1154 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1155 IP6STAT_INC(ip6s_badoptions); 1156 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1157 return (-1); 1158 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1159 IP6STAT_INC(ip6s_badoptions); 1160 ip6 = mtod(m, struct ip6_hdr *); 1161 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1162 (m->m_flags & (M_BCAST|M_MCAST))) 1163 m_freem(m); 1164 else 1165 icmp6_error(m, ICMP6_PARAM_PROB, 1166 ICMP6_PARAMPROB_OPTION, off); 1167 return (-1); 1168 } 1169 1170 m_freem(m); /* XXX: NOTREACHED */ 1171 return (-1); 1172 } 1173 1174 /* 1175 * Create the "control" list for this pcb. 1176 * These functions will not modify mbuf chain at all. 1177 * 1178 * The routine will be called from upper layer handlers like tcp6_input(). 1179 * Thus the routine assumes that the caller (tcp6_input) have already 1180 * called m_pullup() and all the extension headers are located in the 1181 * very first mbuf on the mbuf chain. 1182 * 1183 * ip6_savecontrol_v4 will handle those options that are possible to be 1184 * set on a v4-mapped socket. 1185 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1186 * options and handle the v6-only ones itself. 1187 */ 1188 struct mbuf ** 1189 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1190 int *v4only) 1191 { 1192 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1193 1194 #ifdef SO_TIMESTAMP 1195 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1196 union { 1197 struct timeval tv; 1198 struct bintime bt; 1199 struct timespec ts; 1200 } t; 1201 struct bintime boottimebin, bt1; 1202 struct timespec ts1; 1203 bool stamped; 1204 1205 stamped = false; 1206 switch (inp->inp_socket->so_ts_clock) { 1207 case SO_TS_REALTIME_MICRO: 1208 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1209 M_TSTMP)) { 1210 mbuf_tstmp2timespec(m, &ts1); 1211 timespec2bintime(&ts1, &bt1); 1212 getboottimebin(&boottimebin); 1213 bintime_add(&bt1, &boottimebin); 1214 bintime2timeval(&bt1, &t.tv); 1215 } else { 1216 microtime(&t.tv); 1217 } 1218 *mp = sbcreatecontrol(&t.tv, sizeof(t.tv), 1219 SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT); 1220 if (*mp != NULL) { 1221 mp = &(*mp)->m_next; 1222 stamped = true; 1223 } 1224 break; 1225 1226 case SO_TS_BINTIME: 1227 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1228 M_TSTMP)) { 1229 mbuf_tstmp2timespec(m, &ts1); 1230 timespec2bintime(&ts1, &t.bt); 1231 getboottimebin(&boottimebin); 1232 bintime_add(&t.bt, &boottimebin); 1233 } else { 1234 bintime(&t.bt); 1235 } 1236 *mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME, 1237 SOL_SOCKET, M_NOWAIT); 1238 if (*mp != NULL) { 1239 mp = &(*mp)->m_next; 1240 stamped = true; 1241 } 1242 break; 1243 1244 case SO_TS_REALTIME: 1245 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1246 M_TSTMP)) { 1247 mbuf_tstmp2timespec(m, &t.ts); 1248 getboottimebin(&boottimebin); 1249 bintime2timespec(&boottimebin, &ts1); 1250 timespecadd(&t.ts, &ts1, &t.ts); 1251 } else { 1252 nanotime(&t.ts); 1253 } 1254 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1255 SCM_REALTIME, SOL_SOCKET, M_NOWAIT); 1256 if (*mp != NULL) { 1257 mp = &(*mp)->m_next; 1258 stamped = true; 1259 } 1260 break; 1261 1262 case SO_TS_MONOTONIC: 1263 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1264 M_TSTMP)) 1265 mbuf_tstmp2timespec(m, &t.ts); 1266 else 1267 nanouptime(&t.ts); 1268 *mp = sbcreatecontrol(&t.ts, sizeof(t.ts), 1269 SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT); 1270 if (*mp != NULL) { 1271 mp = &(*mp)->m_next; 1272 stamped = true; 1273 } 1274 break; 1275 1276 default: 1277 panic("unknown (corrupted) so_ts_clock"); 1278 } 1279 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == 1280 (M_PKTHDR | M_TSTMP)) { 1281 struct sock_timestamp_info sti; 1282 1283 bzero(&sti, sizeof(sti)); 1284 sti.st_info_flags = ST_INFO_HW; 1285 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1286 sti.st_info_flags |= ST_INFO_HW_HPREC; 1287 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO, 1288 SOL_SOCKET, M_NOWAIT); 1289 if (*mp != NULL) 1290 mp = &(*mp)->m_next; 1291 } 1292 } 1293 #endif 1294 1295 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1296 /* RFC 2292 sec. 5 */ 1297 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1298 struct in6_pktinfo pi6; 1299 1300 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1301 #ifdef INET 1302 struct ip *ip; 1303 1304 ip = mtod(m, struct ip *); 1305 pi6.ipi6_addr.s6_addr32[0] = 0; 1306 pi6.ipi6_addr.s6_addr32[1] = 0; 1307 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1308 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1309 #else 1310 /* We won't hit this code */ 1311 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1312 #endif 1313 } else { 1314 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1315 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1316 } 1317 pi6.ipi6_ifindex = 1318 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1319 1320 *mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo), 1321 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6, 1322 M_NOWAIT); 1323 if (*mp) 1324 mp = &(*mp)->m_next; 1325 } 1326 1327 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1328 int hlim; 1329 1330 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1331 #ifdef INET 1332 struct ip *ip; 1333 1334 ip = mtod(m, struct ip *); 1335 hlim = ip->ip_ttl; 1336 #else 1337 /* We won't hit this code */ 1338 hlim = 0; 1339 #endif 1340 } else { 1341 hlim = ip6->ip6_hlim & 0xff; 1342 } 1343 *mp = sbcreatecontrol(&hlim, sizeof(int), 1344 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1345 IPPROTO_IPV6, M_NOWAIT); 1346 if (*mp) 1347 mp = &(*mp)->m_next; 1348 } 1349 1350 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1351 int tclass; 1352 1353 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1354 #ifdef INET 1355 struct ip *ip; 1356 1357 ip = mtod(m, struct ip *); 1358 tclass = ip->ip_tos; 1359 #else 1360 /* We won't hit this code */ 1361 tclass = 0; 1362 #endif 1363 } else { 1364 u_int32_t flowinfo; 1365 1366 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1367 flowinfo >>= 20; 1368 tclass = flowinfo & 0xff; 1369 } 1370 *mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS, 1371 IPPROTO_IPV6, M_NOWAIT); 1372 if (*mp) 1373 mp = &(*mp)->m_next; 1374 } 1375 1376 if (v4only != NULL) { 1377 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1378 *v4only = 1; 1379 } else { 1380 *v4only = 0; 1381 } 1382 } 1383 1384 return (mp); 1385 } 1386 1387 void 1388 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp) 1389 { 1390 struct ip6_hdr *ip6; 1391 int v4only = 0; 1392 1393 mp = ip6_savecontrol_v4(inp, m, mp, &v4only); 1394 if (v4only) 1395 return; 1396 1397 ip6 = mtod(m, struct ip6_hdr *); 1398 /* 1399 * IPV6_HOPOPTS socket option. Recall that we required super-user 1400 * privilege for the option (see ip6_ctloutput), but it might be too 1401 * strict, since there might be some hop-by-hop options which can be 1402 * returned to normal user. 1403 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1404 */ 1405 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) { 1406 /* 1407 * Check if a hop-by-hop options header is contatined in the 1408 * received packet, and if so, store the options as ancillary 1409 * data. Note that a hop-by-hop options header must be 1410 * just after the IPv6 header, which is assured through the 1411 * IPv6 input processing. 1412 */ 1413 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1414 struct ip6_hbh *hbh; 1415 u_int hbhlen; 1416 1417 hbh = (struct ip6_hbh *)(ip6 + 1); 1418 hbhlen = (hbh->ip6h_len + 1) << 3; 1419 1420 /* 1421 * XXX: We copy the whole header even if a 1422 * jumbo payload option is included, the option which 1423 * is to be removed before returning according to 1424 * RFC2292. 1425 * Note: this constraint is removed in RFC3542 1426 */ 1427 *mp = sbcreatecontrol(hbh, hbhlen, 1428 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1429 IPPROTO_IPV6, M_NOWAIT); 1430 if (*mp) 1431 mp = &(*mp)->m_next; 1432 } 1433 } 1434 1435 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1436 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1437 1438 /* 1439 * Search for destination options headers or routing 1440 * header(s) through the header chain, and stores each 1441 * header as ancillary data. 1442 * Note that the order of the headers remains in 1443 * the chain of ancillary data. 1444 */ 1445 while (1) { /* is explicit loop prevention necessary? */ 1446 struct ip6_ext *ip6e = NULL; 1447 u_int elen; 1448 1449 /* 1450 * if it is not an extension header, don't try to 1451 * pull it from the chain. 1452 */ 1453 switch (nxt) { 1454 case IPPROTO_DSTOPTS: 1455 case IPPROTO_ROUTING: 1456 case IPPROTO_HOPOPTS: 1457 case IPPROTO_AH: /* is it possible? */ 1458 break; 1459 default: 1460 goto loopend; 1461 } 1462 1463 if (off + sizeof(*ip6e) > m->m_len) 1464 goto loopend; 1465 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1466 if (nxt == IPPROTO_AH) 1467 elen = (ip6e->ip6e_len + 2) << 2; 1468 else 1469 elen = (ip6e->ip6e_len + 1) << 3; 1470 if (off + elen > m->m_len) 1471 goto loopend; 1472 1473 switch (nxt) { 1474 case IPPROTO_DSTOPTS: 1475 if (!(inp->inp_flags & IN6P_DSTOPTS)) 1476 break; 1477 1478 *mp = sbcreatecontrol(ip6e, elen, 1479 IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1480 IPPROTO_IPV6, M_NOWAIT); 1481 if (*mp) 1482 mp = &(*mp)->m_next; 1483 break; 1484 case IPPROTO_ROUTING: 1485 if (!(inp->inp_flags & IN6P_RTHDR)) 1486 break; 1487 1488 *mp = sbcreatecontrol(ip6e, elen, 1489 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR), 1490 IPPROTO_IPV6, M_NOWAIT); 1491 if (*mp) 1492 mp = &(*mp)->m_next; 1493 break; 1494 case IPPROTO_HOPOPTS: 1495 case IPPROTO_AH: /* is it possible? */ 1496 break; 1497 1498 default: 1499 /* 1500 * other cases have been filtered in the above. 1501 * none will visit this case. here we supply 1502 * the code just in case (nxt overwritten or 1503 * other cases). 1504 */ 1505 goto loopend; 1506 } 1507 1508 /* proceed with the next header. */ 1509 off += elen; 1510 nxt = ip6e->ip6e_nxt; 1511 ip6e = NULL; 1512 } 1513 loopend: 1514 ; 1515 } 1516 1517 if (inp->inp_flags2 & INP_RECVFLOWID) { 1518 uint32_t flowid, flow_type; 1519 1520 flowid = m->m_pkthdr.flowid; 1521 flow_type = M_HASHTYPE_GET(m); 1522 1523 /* 1524 * XXX should handle the failure of one or the 1525 * other - don't populate both? 1526 */ 1527 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID, 1528 IPPROTO_IPV6, M_NOWAIT); 1529 if (*mp) 1530 mp = &(*mp)->m_next; 1531 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t), 1532 IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT); 1533 if (*mp) 1534 mp = &(*mp)->m_next; 1535 } 1536 1537 #ifdef RSS 1538 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1539 uint32_t flowid, flow_type; 1540 uint32_t rss_bucketid; 1541 1542 flowid = m->m_pkthdr.flowid; 1543 flow_type = M_HASHTYPE_GET(m); 1544 1545 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1546 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t), 1547 IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT); 1548 if (*mp) 1549 mp = &(*mp)->m_next; 1550 } 1551 } 1552 #endif 1553 1554 } 1555 #undef IS2292 1556 1557 void 1558 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1559 { 1560 struct socket *so; 1561 struct mbuf *m_mtu; 1562 struct ip6_mtuinfo mtuctl; 1563 1564 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1565 /* 1566 * Notify the error by sending IPV6_PATHMTU ancillary data if 1567 * application wanted to know the MTU value. 1568 * NOTE: we notify disconnected sockets, because some udp 1569 * applications keep sending sockets disconnected. 1570 * NOTE: our implementation doesn't notify connected sockets that has 1571 * foreign address that is different than given destination addresses 1572 * (this is permitted by RFC 3542). 1573 */ 1574 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1575 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1576 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1577 return; 1578 1579 mtuctl.ip6m_mtu = mtu; 1580 mtuctl.ip6m_addr = *dst; 1581 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1582 return; 1583 1584 if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU, 1585 IPPROTO_IPV6, M_NOWAIT)) == NULL) 1586 return; 1587 1588 so = inp->inp_socket; 1589 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1590 == 0) { 1591 soroverflow(so); 1592 m_freem(m_mtu); 1593 /* XXX: should count statistics */ 1594 } else 1595 sorwakeup(so); 1596 } 1597 1598 /* 1599 * Get pointer to the previous header followed by the header 1600 * currently processed. 1601 */ 1602 int 1603 ip6_get_prevhdr(const struct mbuf *m, int off) 1604 { 1605 struct ip6_ext ip6e; 1606 struct ip6_hdr *ip6; 1607 int len, nlen, nxt; 1608 1609 if (off == sizeof(struct ip6_hdr)) 1610 return (offsetof(struct ip6_hdr, ip6_nxt)); 1611 if (off < sizeof(struct ip6_hdr)) 1612 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1613 1614 ip6 = mtod(m, struct ip6_hdr *); 1615 nxt = ip6->ip6_nxt; 1616 len = sizeof(struct ip6_hdr); 1617 nlen = 0; 1618 while (len < off) { 1619 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1620 switch (nxt) { 1621 case IPPROTO_FRAGMENT: 1622 nlen = sizeof(struct ip6_frag); 1623 break; 1624 case IPPROTO_AH: 1625 nlen = (ip6e.ip6e_len + 2) << 2; 1626 break; 1627 default: 1628 nlen = (ip6e.ip6e_len + 1) << 3; 1629 } 1630 len += nlen; 1631 nxt = ip6e.ip6e_nxt; 1632 } 1633 return (len - nlen); 1634 } 1635 1636 /* 1637 * get next header offset. m will be retained. 1638 */ 1639 int 1640 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1641 { 1642 struct ip6_hdr ip6; 1643 struct ip6_ext ip6e; 1644 struct ip6_frag fh; 1645 1646 /* just in case */ 1647 if (m == NULL) 1648 panic("ip6_nexthdr: m == NULL"); 1649 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1650 return -1; 1651 1652 switch (proto) { 1653 case IPPROTO_IPV6: 1654 if (m->m_pkthdr.len < off + sizeof(ip6)) 1655 return -1; 1656 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1657 if (nxtp) 1658 *nxtp = ip6.ip6_nxt; 1659 off += sizeof(ip6); 1660 return off; 1661 1662 case IPPROTO_FRAGMENT: 1663 /* 1664 * terminate parsing if it is not the first fragment, 1665 * it does not make sense to parse through it. 1666 */ 1667 if (m->m_pkthdr.len < off + sizeof(fh)) 1668 return -1; 1669 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1670 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1671 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1672 return -1; 1673 if (nxtp) 1674 *nxtp = fh.ip6f_nxt; 1675 off += sizeof(struct ip6_frag); 1676 return off; 1677 1678 case IPPROTO_AH: 1679 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1680 return -1; 1681 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1682 if (nxtp) 1683 *nxtp = ip6e.ip6e_nxt; 1684 off += (ip6e.ip6e_len + 2) << 2; 1685 return off; 1686 1687 case IPPROTO_HOPOPTS: 1688 case IPPROTO_ROUTING: 1689 case IPPROTO_DSTOPTS: 1690 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1691 return -1; 1692 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1693 if (nxtp) 1694 *nxtp = ip6e.ip6e_nxt; 1695 off += (ip6e.ip6e_len + 1) << 3; 1696 return off; 1697 1698 case IPPROTO_NONE: 1699 case IPPROTO_ESP: 1700 case IPPROTO_IPCOMP: 1701 /* give up */ 1702 return -1; 1703 1704 default: 1705 return -1; 1706 } 1707 1708 /* NOTREACHED */ 1709 } 1710 1711 /* 1712 * get offset for the last header in the chain. m will be kept untainted. 1713 */ 1714 int 1715 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1716 { 1717 int newoff; 1718 int nxt; 1719 1720 if (!nxtp) { 1721 nxt = -1; 1722 nxtp = &nxt; 1723 } 1724 while (1) { 1725 newoff = ip6_nexthdr(m, off, proto, nxtp); 1726 if (newoff < 0) 1727 return off; 1728 else if (newoff < off) 1729 return -1; /* invalid */ 1730 else if (newoff == off) 1731 return newoff; 1732 1733 off = newoff; 1734 proto = *nxtp; 1735 } 1736 } 1737 1738 /* 1739 * System control for IP6 1740 */ 1741 1742 u_char inet6ctlerrmap[PRC_NCMDS] = { 1743 0, 0, 0, 0, 1744 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1745 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1746 EMSGSIZE, EHOSTUNREACH, 0, 0, 1747 0, 0, EHOSTUNREACH, 0, 1748 ENOPROTOOPT, ECONNREFUSED 1749 }; 1750