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 127extern struct domain inet6domain; 128 129u_char ip6_protox[IPPROTO_MAX]; 130VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 131VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); 132VNET_DEFINE(u_long, in6_ifaddrhmask); 133 134static 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 147static int 148sysctl_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} 160SYSCTL_DECL(_net_inet6_ip6); 161SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen, 162 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet6_intr_queue_maxlen, "I", 163 "Maximum size of the IPv6 input queue"); 164 165#ifdef RSS 166static struct netisr_handler ip6_direct_nh = { 167 .nh_name = "ip6_direct", 168 .nh_handler = ip6_direct_input, 169 .nh_proto = NETISR_IPV6_DIRECT, 170 .nh_m2cpuid = rss_soft_m2cpuid_v6, 171 .nh_policy = NETISR_POLICY_CPU, 172 .nh_dispatch = NETISR_DISPATCH_HYBRID, 173}; 174 175static int 176sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 177{ 178 int error, qlimit; 179 180 netisr_getqlimit(&ip6_direct_nh, &qlimit); 181 error = sysctl_handle_int(oidp, &qlimit, 0, req); 182 if (error || !req->newptr) 183 return (error); 184 if (qlimit < 1) 185 return (EINVAL); 186 return (netisr_setqlimit(&ip6_direct_nh, qlimit)); 187} 188SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 189 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet6_intr_direct_queue_maxlen, 190 "I", "Maximum size of the IPv6 direct input queue"); 191 192#endif 193 194VNET_DEFINE(struct pfil_head, inet6_pfil_hook); 195 196VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 197VNET_PCPUSTAT_SYSINIT(ip6stat); 198#ifdef VIMAGE 199VNET_PCPUSTAT_SYSUNINIT(ip6stat); 200#endif /* VIMAGE */ 201 202struct rmlock in6_ifaddr_lock; 203RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 204 205static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 206 207/* 208 * IP6 initialization: fill in IP6 protocol switch table. 209 * All protocols not implemented in kernel go to raw IP6 protocol handler. 210 */ 211void 212ip6_init(void) 213{ 214 struct protosw *pr; 215 int i; 216 217 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 218 &V_ip6_auto_linklocal); 219 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 220 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 221 222 CK_STAILQ_INIT(&V_in6_ifaddrhead); 223 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 224 &V_in6_ifaddrhmask); 225 226 /* Initialize packet filter hooks. */ 227 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF; 228 V_inet6_pfil_hook.ph_af = AF_INET6; 229 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0) 230 printf("%s: WARNING: unable to register pfil hook, " 231 "error %d\n", __func__, i); 232 233 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 234 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 235 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 236 printf("%s: WARNING: unable to register input helper hook\n", 237 __func__); 238 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 239 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 240 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 241 printf("%s: WARNING: unable to register output helper hook\n", 242 __func__); 243 244 scope6_init(); 245 addrsel_policy_init(); 246 nd6_init(); 247 frag6_init(); 248 249 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 250 251 /* Skip global initialization stuff for non-default instances. */ 252#ifdef VIMAGE 253 if (!IS_DEFAULT_VNET(curvnet)) { 254 netisr_register_vnet(&ip6_nh); 255#ifdef RSS 256 netisr_register_vnet(&ip6_direct_nh); 257#endif 258 return; 259 } 260#endif 261 262 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 263 if (pr == NULL) 264 panic("ip6_init"); 265 266 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 267 for (i = 0; i < IPPROTO_MAX; i++) 268 ip6_protox[i] = pr - inet6sw; 269 /* 270 * Cycle through IP protocols and put them into the appropriate place 271 * in ip6_protox[]. 272 */ 273 for (pr = inet6domain.dom_protosw; 274 pr < inet6domain.dom_protoswNPROTOSW; pr++) 275 if (pr->pr_domain->dom_family == PF_INET6 && 276 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 277 /* Be careful to only index valid IP protocols. */ 278 if (pr->pr_protocol < IPPROTO_MAX) 279 ip6_protox[pr->pr_protocol] = pr - inet6sw; 280 } 281 282 netisr_register(&ip6_nh); 283#ifdef RSS 284 netisr_register(&ip6_direct_nh); 285#endif 286} 287 288/* 289 * The protocol to be inserted into ip6_protox[] must be already registered 290 * in inet6sw[], either statically or through pf_proto_register(). 291 */ 292int 293ip6proto_register(short ip6proto) 294{ 295 struct protosw *pr; 296 297 /* Sanity checks. */ 298 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 299 return (EPROTONOSUPPORT); 300 301 /* 302 * The protocol slot must not be occupied by another protocol 303 * already. An index pointing to IPPROTO_RAW is unused. 304 */ 305 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 306 if (pr == NULL) 307 return (EPFNOSUPPORT); 308 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 309 return (EEXIST); 310 311 /* 312 * Find the protocol position in inet6sw[] and set the index. 313 */ 314 for (pr = inet6domain.dom_protosw; 315 pr < inet6domain.dom_protoswNPROTOSW; pr++) { 316 if (pr->pr_domain->dom_family == PF_INET6 && 317 pr->pr_protocol && pr->pr_protocol == ip6proto) { 318 ip6_protox[pr->pr_protocol] = pr - inet6sw; 319 return (0); 320 } 321 } 322 return (EPROTONOSUPPORT); 323} 324 325int 326ip6proto_unregister(short ip6proto) 327{ 328 struct protosw *pr; 329 330 /* Sanity checks. */ 331 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 332 return (EPROTONOSUPPORT); 333 334 /* Check if the protocol was indeed registered. */ 335 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 336 if (pr == NULL) 337 return (EPFNOSUPPORT); 338 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 339 return (ENOENT); 340 341 /* Reset the protocol slot to IPPROTO_RAW. */ 342 ip6_protox[ip6proto] = pr - inet6sw; 343 return (0); 344} 345 346#ifdef VIMAGE 347static void 348ip6_destroy(void *unused __unused) 349{ 350 struct ifaddr *ifa, *nifa; 351 struct ifnet *ifp; 352 int error; 353 354#ifdef RSS 355 netisr_unregister_vnet(&ip6_direct_nh); 356#endif 357 netisr_unregister_vnet(&ip6_nh); 358 359 if ((error = pfil_head_unregister(&V_inet6_pfil_hook)) != 0) 360 printf("%s: WARNING: unable to unregister pfil hook, " 361 "error %d\n", __func__, error); 362 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 363 if (error != 0) { 364 printf("%s: WARNING: unable to deregister input helper hook " 365 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 366 "error %d returned\n", __func__, error); 367 } 368 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 369 if (error != 0) { 370 printf("%s: WARNING: unable to deregister output helper hook " 371 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 372 "error %d returned\n", __func__, error); 373 } 374 375 /* Cleanup addresses. */ 376 IFNET_RLOCK(); 377 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 378 /* Cannot lock here - lock recursion. */ 379 /* IF_ADDR_LOCK(ifp); */ 380 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 381 382 if (ifa->ifa_addr->sa_family != AF_INET6) 383 continue; 384 in6_purgeaddr(ifa); 385 } 386 /* IF_ADDR_UNLOCK(ifp); */ 387 in6_ifdetach_destroy(ifp); 388 mld_domifdetach(ifp); 389 /* Make sure any routes are gone as well. */ 390 rt_flushifroutes_af(ifp, AF_INET6); 391 } 392 IFNET_RUNLOCK(); 393 394 frag6_destroy(); 395 nd6_destroy(); 396 in6_ifattach_destroy(); 397 398 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 399} 400 401VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 402#endif 403 404static int 405ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off, 406 int *nxt, int *ours) 407{ 408 struct mbuf *m; 409 struct ip6_hdr *ip6; 410 struct ip6_hbh *hbh; 411 412 if (ip6_hopopts_input(plen, rtalert, mp, off)) { 413#if 0 /*touches NULL pointer*/ 414 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard); 415#endif 416 goto out; /* m have already been freed */ 417 } 418 419 /* adjust pointer */ 420 m = *mp; 421 ip6 = mtod(m, struct ip6_hdr *); 422 423 /* 424 * if the payload length field is 0 and the next header field 425 * indicates Hop-by-Hop Options header, then a Jumbo Payload 426 * option MUST be included. 427 */ 428 if (ip6->ip6_plen == 0 && *plen == 0) { 429 /* 430 * Note that if a valid jumbo payload option is 431 * contained, ip6_hopopts_input() must set a valid 432 * (non-zero) payload length to the variable plen. 433 */ 434 IP6STAT_INC(ip6s_badoptions); 435 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 436 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 437 icmp6_error(m, ICMP6_PARAM_PROB, 438 ICMP6_PARAMPROB_HEADER, 439 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 440 goto out; 441 } 442 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 443 hbh = (struct ip6_hbh *)(ip6 + 1); 444 *nxt = hbh->ip6h_nxt; 445 446 /* 447 * If we are acting as a router and the packet contains a 448 * router alert option, see if we know the option value. 449 * Currently, we only support the option value for MLD, in which 450 * case we should pass the packet to the multicast routing 451 * daemon. 452 */ 453 if (*rtalert != ~0) { 454 switch (*rtalert) { 455 case IP6OPT_RTALERT_MLD: 456 if (V_ip6_forwarding) 457 *ours = 1; 458 break; 459 default: 460 /* 461 * RFC2711 requires unrecognized values must be 462 * silently ignored. 463 */ 464 break; 465 } 466 } 467 468 return (0); 469 470out: 471 return (1); 472} 473 474#ifdef RSS 475/* 476 * IPv6 direct input routine. 477 * 478 * This is called when reinjecting completed fragments where 479 * all of the previous checking and book-keeping has been done. 480 */ 481void 482ip6_direct_input(struct mbuf *m) 483{ 484 int off, nxt; 485 int nest; 486 struct m_tag *mtag; 487 struct ip6_direct_ctx *ip6dc; 488 489 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 490 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 491 492 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 493 nxt = ip6dc->ip6dc_nxt; 494 off = ip6dc->ip6dc_off; 495 496 nest = 0; 497 498 m_tag_delete(m, mtag); 499 500 while (nxt != IPPROTO_DONE) { 501 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 502 IP6STAT_INC(ip6s_toomanyhdr); 503 goto bad; 504 } 505 506 /* 507 * protection against faulty packet - there should be 508 * more sanity checks in header chain processing. 509 */ 510 if (m->m_pkthdr.len < off) { 511 IP6STAT_INC(ip6s_tooshort); 512 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 513 goto bad; 514 } 515 516#if defined(IPSEC) || defined(IPSEC_SUPPORT) 517 if (IPSEC_ENABLED(ipv6)) { 518 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 519 return; 520 } 521#endif /* IPSEC */ 522 523 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 524 } 525 return; 526bad: 527 m_freem(m); 528} 529#endif 530 531void 532ip6_input(struct mbuf *m) 533{ 534 struct in6_addr odst; 535 struct ip6_hdr *ip6; 536 struct in6_ifaddr *ia; 537 struct ifnet *rcvif; 538 u_int32_t plen; 539 u_int32_t rtalert = ~0; 540 int off = sizeof(struct ip6_hdr), nest; 541 int nxt, ours = 0; 542 int srcrt = 0; 543 544 /* 545 * Drop the packet if IPv6 operation is disabled on the interface. 546 */ 547 rcvif = m->m_pkthdr.rcvif; 548 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 549 goto bad; 550 551#if defined(IPSEC) || defined(IPSEC_SUPPORT) 552 /* 553 * should the inner packet be considered authentic? 554 * see comment in ah4_input(). 555 * NB: m cannot be NULL when passed to the input routine 556 */ 557 558 m->m_flags &= ~M_AUTHIPHDR; 559 m->m_flags &= ~M_AUTHIPDGM; 560 561#endif /* IPSEC */ 562 563 if (m->m_flags & M_FASTFWD_OURS) { 564 /* 565 * Firewall changed destination to local. 566 */ 567 ip6 = mtod(m, struct ip6_hdr *); 568 goto passin; 569 } 570 571 /* 572 * mbuf statistics 573 */ 574 if (m->m_flags & M_EXT) { 575 if (m->m_next) 576 IP6STAT_INC(ip6s_mext2m); 577 else 578 IP6STAT_INC(ip6s_mext1); 579 } else { 580 if (m->m_next) { 581 if (m->m_flags & M_LOOP) { 582 IP6STAT_INC(ip6s_m2m[V_loif->if_index]); 583 } else if (rcvif->if_index < IP6S_M2MMAX) 584 IP6STAT_INC(ip6s_m2m[rcvif->if_index]); 585 else 586 IP6STAT_INC(ip6s_m2m[0]); 587 } else 588 IP6STAT_INC(ip6s_m1); 589 } 590 591 in6_ifstat_inc(rcvif, ifs6_in_receive); 592 IP6STAT_INC(ip6s_total); 593 594 /* 595 * L2 bridge code and some other code can return mbuf chain 596 * that does not conform to KAME requirement. too bad. 597 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 598 */ 599 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 600 struct mbuf *n; 601 602 if (m->m_pkthdr.len > MHLEN) 603 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 604 else 605 n = m_gethdr(M_NOWAIT, MT_DATA); 606 if (n == NULL) 607 goto bad; 608 609 m_move_pkthdr(n, m); 610 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 611 n->m_len = n->m_pkthdr.len; 612 m_freem(m); 613 m = n; 614 } 615 if (m->m_len < sizeof(struct ip6_hdr)) { 616 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 617 IP6STAT_INC(ip6s_toosmall); 618 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 619 goto bad; 620 } 621 } 622 623 ip6 = mtod(m, struct ip6_hdr *); 624 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 625 IP6STAT_INC(ip6s_badvers); 626 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 627 goto bad; 628 } 629 630 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 631 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 632 633 /* 634 * Check against address spoofing/corruption. 635 */ 636 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 637 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 638 /* 639 * XXX: "badscope" is not very suitable for a multicast source. 640 */ 641 IP6STAT_INC(ip6s_badscope); 642 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 643 goto bad; 644 } 645 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 646 !(m->m_flags & M_LOOP)) { 647 /* 648 * In this case, the packet should come from the loopback 649 * interface. However, we cannot just check the if_flags, 650 * because ip6_mloopback() passes the "actual" interface 651 * as the outgoing/incoming interface. 652 */ 653 IP6STAT_INC(ip6s_badscope); 654 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 655 goto bad; 656 } 657 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 658 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 659 /* 660 * RFC4291 2.7: 661 * Nodes must not originate a packet to a multicast address 662 * whose scop field contains the reserved value 0; if such 663 * a packet is received, it must be silently dropped. 664 */ 665 IP6STAT_INC(ip6s_badscope); 666 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 667 goto bad; 668 } 669#ifdef ALTQ 670 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 671 /* packet is dropped by traffic conditioner */ 672 return; 673 } 674#endif 675 /* 676 * The following check is not documented in specs. A malicious 677 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 678 * and bypass security checks (act as if it was from 127.0.0.1 by using 679 * IPv6 src ::ffff:127.0.0.1). Be cautious. 680 * 681 * We have supported IPv6-only kernels for a few years and this issue 682 * has not come up. The world seems to move mostly towards not using 683 * v4mapped on the wire, so it makes sense for us to keep rejecting 684 * any such packets. 685 */ 686 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 687 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 688 IP6STAT_INC(ip6s_badscope); 689 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 690 goto bad; 691 } 692#if 0 693 /* 694 * Reject packets with IPv4 compatible addresses (auto tunnel). 695 * 696 * The code forbids auto tunnel relay case in RFC1933 (the check is 697 * stronger than RFC1933). We may want to re-enable it if mech-xx 698 * is revised to forbid relaying case. 699 */ 700 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 701 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 702 IP6STAT_INC(ip6s_badscope); 703 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 704 goto bad; 705 } 706#endif 707 /* 708 * Try to forward the packet, but if we fail continue. 709 * ip6_tryforward() does not generate redirects, so fall 710 * through to normal processing if redirects are required. 711 * ip6_tryforward() does inbound and outbound packet firewall 712 * processing. If firewall has decided that destination becomes 713 * our local address, it sets M_FASTFWD_OURS flag. In this 714 * case skip another inbound firewall processing and update 715 * ip6 pointer. 716 */ 717 if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0 718#if defined(IPSEC) || defined(IPSEC_SUPPORT) 719 && (!IPSEC_ENABLED(ipv6) || 720 IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0) 721#endif 722 ) { 723 if ((m = ip6_tryforward(m)) == NULL) 724 return; 725 if (m->m_flags & M_FASTFWD_OURS) { 726 ip6 = mtod(m, struct ip6_hdr *); 727 goto passin; 728 } 729 } 730#if defined(IPSEC) || defined(IPSEC_SUPPORT) 731 /* 732 * Bypass packet filtering for packets previously handled by IPsec. 733 */ 734 if (IPSEC_ENABLED(ipv6) && 735 IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0) 736 goto passin; 737#endif 738 /* 739 * Run through list of hooks for input packets. 740 * 741 * NB: Beware of the destination address changing 742 * (e.g. by NAT rewriting). When this happens, 743 * tell ip6_forward to do the right thing. 744 */ 745 746 /* Jump over all PFIL processing if hooks are not active. */ 747 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 748 goto passin; 749 750 odst = ip6->ip6_dst; 751 if (pfil_run_hooks(&V_inet6_pfil_hook, &m, 752 m->m_pkthdr.rcvif, PFIL_IN, 0, NULL)) 753 return; 754 if (m == NULL) /* consumed by filter */ 755 return; 756 ip6 = mtod(m, struct ip6_hdr *); 757 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 758 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 759 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 760 /* 761 * Directly ship the packet on. This allows forwarding 762 * packets originally destined to us to some other directly 763 * connected host. 764 */ 765 ip6_forward(m, 1); 766 return; 767 } 768 769passin: 770 /* 771 * Disambiguate address scope zones (if there is ambiguity). 772 * We first make sure that the original source or destination address 773 * is not in our internal form for scoped addresses. Such addresses 774 * are not necessarily invalid spec-wise, but we cannot accept them due 775 * to the usage conflict. 776 * in6_setscope() then also checks and rejects the cases where src or 777 * dst are the loopback address and the receiving interface 778 * is not loopback. 779 */ 780 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 781 IP6STAT_INC(ip6s_badscope); /* XXX */ 782 goto bad; 783 } 784 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 785 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 786 IP6STAT_INC(ip6s_badscope); 787 goto bad; 788 } 789 if (m->m_flags & M_FASTFWD_OURS) { 790 m->m_flags &= ~M_FASTFWD_OURS; 791 ours = 1; 792 goto hbhcheck; 793 } 794 /* 795 * Multicast check. Assume packet is for us to avoid 796 * prematurely taking locks. 797 */ 798 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 799 ours = 1; 800 in6_ifstat_inc(rcvif, ifs6_in_mcast); 801 goto hbhcheck; 802 } 803 /* 804 * Unicast check 805 * XXX: For now we keep link-local IPv6 addresses with embedded 806 * scope zone id, therefore we use zero zoneid here. 807 */ 808 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */); 809 if (ia != NULL) { 810 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 811 char ip6bufs[INET6_ADDRSTRLEN]; 812 char ip6bufd[INET6_ADDRSTRLEN]; 813 /* address is not ready, so discard the packet. */ 814 nd6log((LOG_INFO, 815 "ip6_input: packet to an unready address %s->%s\n", 816 ip6_sprintf(ip6bufs, &ip6->ip6_src), 817 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 818 ifa_free(&ia->ia_ifa); 819 goto bad; 820 } 821 /* Count the packet in the ip address stats */ 822 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 823 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 824 ifa_free(&ia->ia_ifa); 825 ours = 1; 826 goto hbhcheck; 827 } 828 829 /* 830 * Now there is no reason to process the packet if it's not our own 831 * and we're not a router. 832 */ 833 if (!V_ip6_forwarding) { 834 IP6STAT_INC(ip6s_cantforward); 835 goto bad; 836 } 837 838 hbhcheck: 839 /* 840 * Process Hop-by-Hop options header if it's contained. 841 * m may be modified in ip6_hopopts_input(). 842 * If a JumboPayload option is included, plen will also be modified. 843 */ 844 plen = (u_int32_t)ntohs(ip6->ip6_plen); 845 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 846 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0) 847 return; 848 } else 849 nxt = ip6->ip6_nxt; 850 851 /* 852 * Use mbuf flags to propagate Router Alert option to 853 * ICMPv6 layer, as hop-by-hop options have been stripped. 854 */ 855 if (rtalert != ~0) 856 m->m_flags |= M_RTALERT_MLD; 857 858 /* 859 * Check that the amount of data in the buffers 860 * is as at least much as the IPv6 header would have us expect. 861 * Trim mbufs if longer than we expect. 862 * Drop packet if shorter than we expect. 863 */ 864 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 865 IP6STAT_INC(ip6s_tooshort); 866 in6_ifstat_inc(rcvif, ifs6_in_truncated); 867 goto bad; 868 } 869 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 870 if (m->m_len == m->m_pkthdr.len) { 871 m->m_len = sizeof(struct ip6_hdr) + plen; 872 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 873 } else 874 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 875 } 876 877 /* 878 * Forward if desirable. 879 */ 880 if (V_ip6_mrouter && 881 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 882 /* 883 * If we are acting as a multicast router, all 884 * incoming multicast packets are passed to the 885 * kernel-level multicast forwarding function. 886 * The packet is returned (relatively) intact; if 887 * ip6_mforward() returns a non-zero value, the packet 888 * must be discarded, else it may be accepted below. 889 * 890 * XXX TODO: Check hlim and multicast scope here to avoid 891 * unnecessarily calling into ip6_mforward(). 892 */ 893 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 894 IP6STAT_INC(ip6s_cantforward); 895 goto bad; 896 } 897 } else if (!ours) { 898 ip6_forward(m, srcrt); 899 return; 900 } 901 902 /* 903 * Tell launch routine the next header 904 */ 905 IP6STAT_INC(ip6s_delivered); 906 in6_ifstat_inc(rcvif, ifs6_in_deliver); 907 nest = 0; 908 909 while (nxt != IPPROTO_DONE) { 910 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 911 IP6STAT_INC(ip6s_toomanyhdr); 912 goto bad; 913 } 914 915 /* 916 * protection against faulty packet - there should be 917 * more sanity checks in header chain processing. 918 */ 919 if (m->m_pkthdr.len < off) { 920 IP6STAT_INC(ip6s_tooshort); 921 in6_ifstat_inc(rcvif, ifs6_in_truncated); 922 goto bad; 923 } 924 925#if defined(IPSEC) || defined(IPSEC_SUPPORT) 926 if (IPSEC_ENABLED(ipv6)) { 927 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 928 return; 929 } 930#endif /* IPSEC */ 931 932 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 933 } 934 return; 935bad: 936 in6_ifstat_inc(rcvif, ifs6_in_discard); 937 if (m != NULL) 938 m_freem(m); 939} 940 941/* 942 * Hop-by-Hop options header processing. If a valid jumbo payload option is 943 * included, the real payload length will be stored in plenp. 944 * 945 * rtalertp - XXX: should be stored more smart way 946 */ 947static int 948ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 949 struct mbuf **mp, int *offp) 950{ 951 struct mbuf *m = *mp; 952 int off = *offp, hbhlen; 953 struct ip6_hbh *hbh; 954 955 /* validation of the length of the header */ 956 if (m->m_len < off + sizeof(*hbh)) { 957 m = m_pullup(m, off + sizeof(*hbh)); 958 if (m == NULL) { 959 IP6STAT_INC(ip6s_exthdrtoolong); 960 *mp = NULL; 961 return (-1); 962 } 963 } 964 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 965 hbhlen = (hbh->ip6h_len + 1) << 3; 966 967 if (m->m_len < off + hbhlen) { 968 m = m_pullup(m, off + hbhlen); 969 if (m == NULL) { 970 IP6STAT_INC(ip6s_exthdrtoolong); 971 *mp = NULL; 972 return (-1); 973 } 974 } 975 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 976 off += hbhlen; 977 hbhlen -= sizeof(struct ip6_hbh); 978 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 979 hbhlen, rtalertp, plenp) < 0) { 980 *mp = NULL; 981 return (-1); 982 } 983 984 *offp = off; 985 *mp = m; 986 return (0); 987} 988 989/* 990 * Search header for all Hop-by-hop options and process each option. 991 * This function is separate from ip6_hopopts_input() in order to 992 * handle a case where the sending node itself process its hop-by-hop 993 * options header. In such a case, the function is called from ip6_output(). 994 * 995 * The function assumes that hbh header is located right after the IPv6 header 996 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 997 * opthead + hbhlen is located in contiguous memory region. 998 */ 999int 1000ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1001 u_int32_t *rtalertp, u_int32_t *plenp) 1002{ 1003 struct ip6_hdr *ip6; 1004 int optlen = 0; 1005 u_int8_t *opt = opthead; 1006 u_int16_t rtalert_val; 1007 u_int32_t jumboplen; 1008 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1009 1010 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1011 switch (*opt) { 1012 case IP6OPT_PAD1: 1013 optlen = 1; 1014 break; 1015 case IP6OPT_PADN: 1016 if (hbhlen < IP6OPT_MINLEN) { 1017 IP6STAT_INC(ip6s_toosmall); 1018 goto bad; 1019 } 1020 optlen = *(opt + 1) + 2; 1021 break; 1022 case IP6OPT_ROUTER_ALERT: 1023 /* XXX may need check for alignment */ 1024 if (hbhlen < IP6OPT_RTALERT_LEN) { 1025 IP6STAT_INC(ip6s_toosmall); 1026 goto bad; 1027 } 1028 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1029 /* XXX stat */ 1030 icmp6_error(m, ICMP6_PARAM_PROB, 1031 ICMP6_PARAMPROB_HEADER, 1032 erroff + opt + 1 - opthead); 1033 return (-1); 1034 } 1035 optlen = IP6OPT_RTALERT_LEN; 1036 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1037 *rtalertp = ntohs(rtalert_val); 1038 break; 1039 case IP6OPT_JUMBO: 1040 /* XXX may need check for alignment */ 1041 if (hbhlen < IP6OPT_JUMBO_LEN) { 1042 IP6STAT_INC(ip6s_toosmall); 1043 goto bad; 1044 } 1045 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1046 /* XXX stat */ 1047 icmp6_error(m, ICMP6_PARAM_PROB, 1048 ICMP6_PARAMPROB_HEADER, 1049 erroff + opt + 1 - opthead); 1050 return (-1); 1051 } 1052 optlen = IP6OPT_JUMBO_LEN; 1053 1054 /* 1055 * IPv6 packets that have non 0 payload length 1056 * must not contain a jumbo payload option. 1057 */ 1058 ip6 = mtod(m, struct ip6_hdr *); 1059 if (ip6->ip6_plen) { 1060 IP6STAT_INC(ip6s_badoptions); 1061 icmp6_error(m, ICMP6_PARAM_PROB, 1062 ICMP6_PARAMPROB_HEADER, 1063 erroff + opt - opthead); 1064 return (-1); 1065 } 1066 1067 /* 1068 * We may see jumbolen in unaligned location, so 1069 * we'd need to perform bcopy(). 1070 */ 1071 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1072 jumboplen = (u_int32_t)htonl(jumboplen); 1073 1074#if 1 1075 /* 1076 * if there are multiple jumbo payload options, 1077 * *plenp will be non-zero and the packet will be 1078 * rejected. 1079 * the behavior may need some debate in ipngwg - 1080 * multiple options does not make sense, however, 1081 * there's no explicit mention in specification. 1082 */ 1083 if (*plenp != 0) { 1084 IP6STAT_INC(ip6s_badoptions); 1085 icmp6_error(m, ICMP6_PARAM_PROB, 1086 ICMP6_PARAMPROB_HEADER, 1087 erroff + opt + 2 - opthead); 1088 return (-1); 1089 } 1090#endif 1091 1092 /* 1093 * jumbo payload length must be larger than 65535. 1094 */ 1095 if (jumboplen <= IPV6_MAXPACKET) { 1096 IP6STAT_INC(ip6s_badoptions); 1097 icmp6_error(m, ICMP6_PARAM_PROB, 1098 ICMP6_PARAMPROB_HEADER, 1099 erroff + opt + 2 - opthead); 1100 return (-1); 1101 } 1102 *plenp = jumboplen; 1103 1104 break; 1105 default: /* unknown option */ 1106 if (hbhlen < IP6OPT_MINLEN) { 1107 IP6STAT_INC(ip6s_toosmall); 1108 goto bad; 1109 } 1110 optlen = ip6_unknown_opt(opt, m, 1111 erroff + opt - opthead); 1112 if (optlen == -1) 1113 return (-1); 1114 optlen += 2; 1115 break; 1116 } 1117 } 1118 1119 return (0); 1120 1121 bad: 1122 m_freem(m); 1123 return (-1); 1124} 1125 1126/* 1127 * Unknown option processing. 1128 * The third argument `off' is the offset from the IPv6 header to the option, 1129 * which is necessary if the IPv6 header the and option header and IPv6 header 1130 * is not contiguous in order to return an ICMPv6 error. 1131 */ 1132int 1133ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1134{ 1135 struct ip6_hdr *ip6; 1136 1137 switch (IP6OPT_TYPE(*optp)) { 1138 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1139 return ((int)*(optp + 1)); 1140 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1141 m_freem(m); 1142 return (-1); 1143 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1144 IP6STAT_INC(ip6s_badoptions); 1145 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1146 return (-1); 1147 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1148 IP6STAT_INC(ip6s_badoptions); 1149 ip6 = mtod(m, struct ip6_hdr *); 1150 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1151 (m->m_flags & (M_BCAST|M_MCAST))) 1152 m_freem(m); 1153 else 1154 icmp6_error(m, ICMP6_PARAM_PROB, 1155 ICMP6_PARAMPROB_OPTION, off); 1156 return (-1); 1157 } 1158 1159 m_freem(m); /* XXX: NOTREACHED */ 1160 return (-1); 1161} 1162 1163/* 1164 * Create the "control" list for this pcb. 1165 * These functions will not modify mbuf chain at all. 1166 * 1167 * The routine will be called from upper layer handlers like tcp6_input(). 1168 * Thus the routine assumes that the caller (tcp6_input) have already 1169 * called m_pullup() and all the extension headers are located in the 1170 * very first mbuf on the mbuf chain. 1171 * 1172 * ip6_savecontrol_v4 will handle those options that are possible to be 1173 * set on a v4-mapped socket. 1174 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1175 * options and handle the v6-only ones itself. 1176 */ 1177struct mbuf ** 1178ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1179 int *v4only) 1180{ 1181 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1182 1183#ifdef SO_TIMESTAMP 1184 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1185 union { 1186 struct timeval tv; 1187 struct bintime bt; 1188 struct timespec ts; 1189 } t; 1190 struct bintime boottimebin, bt1; 1191 struct timespec ts1; 1192 bool stamped; 1193 1194 stamped = false; 1195 switch (inp->inp_socket->so_ts_clock) { 1196 case SO_TS_REALTIME_MICRO: 1197 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1198 M_TSTMP)) { 1199 mbuf_tstmp2timespec(m, &ts1); 1200 timespec2bintime(&ts1, &bt1); 1201 getboottimebin(&boottimebin); 1202 bintime_add(&bt1, &boottimebin); 1203 bintime2timeval(&bt1, &t.tv); 1204 } else { 1205 microtime(&t.tv); 1206 } 1207 *mp = sbcreatecontrol((caddr_t) &t.tv, sizeof(t.tv), 1208 SCM_TIMESTAMP, SOL_SOCKET); 1209 if (*mp != NULL) { 1210 mp = &(*mp)->m_next; 1211 stamped = true; 1212 } 1213 break; 1214 1215 case SO_TS_BINTIME: 1216 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1217 M_TSTMP)) { 1218 mbuf_tstmp2timespec(m, &ts1); 1219 timespec2bintime(&ts1, &t.bt); 1220 getboottimebin(&boottimebin); 1221 bintime_add(&t.bt, &boottimebin); 1222 } else { 1223 bintime(&t.bt); 1224 } 1225 *mp = sbcreatecontrol((caddr_t)&t.bt, sizeof(t.bt), 1226 SCM_BINTIME, SOL_SOCKET); 1227 if (*mp != NULL) { 1228 mp = &(*mp)->m_next; 1229 stamped = true; 1230 } 1231 break; 1232 1233 case SO_TS_REALTIME: 1234 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1235 M_TSTMP)) { 1236 mbuf_tstmp2timespec(m, &t.ts); 1237 getboottimebin(&boottimebin); 1238 bintime2timespec(&boottimebin, &ts1); 1239 timespecadd(&t.ts, &ts1, &t.ts); 1240 } else { 1241 nanotime(&t.ts); 1242 } 1243 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts), 1244 SCM_REALTIME, SOL_SOCKET); 1245 if (*mp != NULL) { 1246 mp = &(*mp)->m_next; 1247 stamped = true; 1248 } 1249 break; 1250 1251 case SO_TS_MONOTONIC: 1252 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1253 M_TSTMP)) 1254 mbuf_tstmp2timespec(m, &t.ts); 1255 else 1256 nanouptime(&t.ts); 1257 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts), 1258 SCM_MONOTONIC, SOL_SOCKET); 1259 if (*mp != NULL) { 1260 mp = &(*mp)->m_next; 1261 stamped = true; 1262 } 1263 break; 1264 1265 default: 1266 panic("unknown (corrupted) so_ts_clock"); 1267 } 1268 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == 1269 (M_PKTHDR | M_TSTMP)) { 1270 struct sock_timestamp_info sti; 1271 1272 bzero(&sti, sizeof(sti)); 1273 sti.st_info_flags = ST_INFO_HW; 1274 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1275 sti.st_info_flags |= ST_INFO_HW_HPREC; 1276 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), 1277 SCM_TIME_INFO, SOL_SOCKET); 1278 if (*mp != NULL) 1279 mp = &(*mp)->m_next; 1280 } 1281 } 1282#endif 1283 1284#define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1285 /* RFC 2292 sec. 5 */ 1286 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1287 struct in6_pktinfo pi6; 1288 1289 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1290#ifdef INET 1291 struct ip *ip; 1292 1293 ip = mtod(m, struct ip *); 1294 pi6.ipi6_addr.s6_addr32[0] = 0; 1295 pi6.ipi6_addr.s6_addr32[1] = 0; 1296 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1297 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1298#else 1299 /* We won't hit this code */ 1300 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1301#endif 1302 } else { 1303 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1304 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1305 } 1306 pi6.ipi6_ifindex = 1307 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1308 1309 *mp = sbcreatecontrol((caddr_t) &pi6, 1310 sizeof(struct in6_pktinfo), 1311 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1312 if (*mp) 1313 mp = &(*mp)->m_next; 1314 } 1315 1316 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1317 int hlim; 1318 1319 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1320#ifdef INET 1321 struct ip *ip; 1322 1323 ip = mtod(m, struct ip *); 1324 hlim = ip->ip_ttl; 1325#else 1326 /* We won't hit this code */ 1327 hlim = 0; 1328#endif 1329 } else { 1330 hlim = ip6->ip6_hlim & 0xff; 1331 } 1332 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1333 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1334 IPPROTO_IPV6); 1335 if (*mp) 1336 mp = &(*mp)->m_next; 1337 } 1338 1339 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1340 int tclass; 1341 1342 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1343#ifdef INET 1344 struct ip *ip; 1345 1346 ip = mtod(m, struct ip *); 1347 tclass = ip->ip_tos; 1348#else 1349 /* We won't hit this code */ 1350 tclass = 0; 1351#endif 1352 } else { 1353 u_int32_t flowinfo; 1354 1355 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1356 flowinfo >>= 20; 1357 tclass = flowinfo & 0xff; 1358 } 1359 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int), 1360 IPV6_TCLASS, IPPROTO_IPV6); 1361 if (*mp) 1362 mp = &(*mp)->m_next; 1363 } 1364 1365 if (v4only != NULL) { 1366 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1367 *v4only = 1; 1368 } else { 1369 *v4only = 0; 1370 } 1371 } 1372 1373 return (mp); 1374} 1375 1376void 1377ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp) 1378{ 1379 struct ip6_hdr *ip6; 1380 int v4only = 0; 1381 1382 mp = ip6_savecontrol_v4(inp, m, mp, &v4only); 1383 if (v4only) 1384 return; 1385 1386 ip6 = mtod(m, struct ip6_hdr *); 1387 /* 1388 * IPV6_HOPOPTS socket option. Recall that we required super-user 1389 * privilege for the option (see ip6_ctloutput), but it might be too 1390 * strict, since there might be some hop-by-hop options which can be 1391 * returned to normal user. 1392 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1393 */ 1394 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) { 1395 /* 1396 * Check if a hop-by-hop options header is contatined in the 1397 * received packet, and if so, store the options as ancillary 1398 * data. Note that a hop-by-hop options header must be 1399 * just after the IPv6 header, which is assured through the 1400 * IPv6 input processing. 1401 */ 1402 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1403 struct ip6_hbh *hbh; 1404 int hbhlen; 1405 1406 hbh = (struct ip6_hbh *)(ip6 + 1); 1407 hbhlen = (hbh->ip6h_len + 1) << 3; 1408 1409 /* 1410 * XXX: We copy the whole header even if a 1411 * jumbo payload option is included, the option which 1412 * is to be removed before returning according to 1413 * RFC2292. 1414 * Note: this constraint is removed in RFC3542 1415 */ 1416 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1417 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1418 IPPROTO_IPV6); 1419 if (*mp) 1420 mp = &(*mp)->m_next; 1421 } 1422 } 1423 1424 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1425 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1426 1427 /* 1428 * Search for destination options headers or routing 1429 * header(s) through the header chain, and stores each 1430 * header as ancillary data. 1431 * Note that the order of the headers remains in 1432 * the chain of ancillary data. 1433 */ 1434 while (1) { /* is explicit loop prevention necessary? */ 1435 struct ip6_ext *ip6e = NULL; 1436 int elen; 1437 1438 /* 1439 * if it is not an extension header, don't try to 1440 * pull it from the chain. 1441 */ 1442 switch (nxt) { 1443 case IPPROTO_DSTOPTS: 1444 case IPPROTO_ROUTING: 1445 case IPPROTO_HOPOPTS: 1446 case IPPROTO_AH: /* is it possible? */ 1447 break; 1448 default: 1449 goto loopend; 1450 } 1451 1452 if (off + sizeof(*ip6e) > m->m_len) 1453 goto loopend; 1454 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1455 if (nxt == IPPROTO_AH) 1456 elen = (ip6e->ip6e_len + 2) << 2; 1457 else 1458 elen = (ip6e->ip6e_len + 1) << 3; 1459 if (off + elen > m->m_len) 1460 goto loopend; 1461 1462 switch (nxt) { 1463 case IPPROTO_DSTOPTS: 1464 if (!(inp->inp_flags & IN6P_DSTOPTS)) 1465 break; 1466 1467 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1468 IS2292(inp, 1469 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1470 IPPROTO_IPV6); 1471 if (*mp) 1472 mp = &(*mp)->m_next; 1473 break; 1474 case IPPROTO_ROUTING: 1475 if (!(inp->inp_flags & IN6P_RTHDR)) 1476 break; 1477 1478 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1479 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR), 1480 IPPROTO_IPV6); 1481 if (*mp) 1482 mp = &(*mp)->m_next; 1483 break; 1484 case IPPROTO_HOPOPTS: 1485 case IPPROTO_AH: /* is it possible? */ 1486 break; 1487 1488 default: 1489 /* 1490 * other cases have been filtered in the above. 1491 * none will visit this case. here we supply 1492 * the code just in case (nxt overwritten or 1493 * other cases). 1494 */ 1495 goto loopend; 1496 1497 } 1498 1499 /* proceed with the next header. */ 1500 off += elen; 1501 nxt = ip6e->ip6e_nxt; 1502 ip6e = NULL; 1503 } 1504 loopend: 1505 ; 1506 } 1507 1508 if (inp->inp_flags2 & INP_RECVFLOWID) { 1509 uint32_t flowid, flow_type; 1510 1511 flowid = m->m_pkthdr.flowid; 1512 flow_type = M_HASHTYPE_GET(m); 1513 1514 /* 1515 * XXX should handle the failure of one or the 1516 * other - don't populate both? 1517 */ 1518 *mp = sbcreatecontrol((caddr_t) &flowid, 1519 sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6); 1520 if (*mp) 1521 mp = &(*mp)->m_next; 1522 *mp = sbcreatecontrol((caddr_t) &flow_type, 1523 sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6); 1524 if (*mp) 1525 mp = &(*mp)->m_next; 1526 } 1527 1528#ifdef RSS 1529 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1530 uint32_t flowid, flow_type; 1531 uint32_t rss_bucketid; 1532 1533 flowid = m->m_pkthdr.flowid; 1534 flow_type = M_HASHTYPE_GET(m); 1535 1536 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1537 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1538 sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6); 1539 if (*mp) 1540 mp = &(*mp)->m_next; 1541 } 1542 } 1543#endif 1544 1545} 1546#undef IS2292 1547 1548void 1549ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1550{ 1551 struct socket *so; 1552 struct mbuf *m_mtu; 1553 struct ip6_mtuinfo mtuctl; 1554 1555 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1556 /* 1557 * Notify the error by sending IPV6_PATHMTU ancillary data if 1558 * application wanted to know the MTU value. 1559 * NOTE: we notify disconnected sockets, because some udp 1560 * applications keep sending sockets disconnected. 1561 * NOTE: our implementation doesn't notify connected sockets that has 1562 * foreign address that is different than given destination addresses 1563 * (this is permitted by RFC 3542). 1564 */ 1565 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1566 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1567 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1568 return; 1569 1570 mtuctl.ip6m_mtu = mtu; 1571 mtuctl.ip6m_addr = *dst; 1572 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1573 return; 1574 1575 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1576 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1577 return; 1578 1579 so = inp->inp_socket; 1580 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1581 == 0) { 1582 m_freem(m_mtu); 1583 /* XXX: should count statistics */ 1584 } else 1585 sorwakeup(so); 1586} 1587 1588/* 1589 * Get pointer to the previous header followed by the header 1590 * currently processed. 1591 */ 1592int 1593ip6_get_prevhdr(const struct mbuf *m, int off) 1594{ 1595 struct ip6_ext ip6e; 1596 struct ip6_hdr *ip6; 1597 int len, nlen, nxt; 1598 1599 if (off == sizeof(struct ip6_hdr)) 1600 return (offsetof(struct ip6_hdr, ip6_nxt)); 1601 if (off < sizeof(struct ip6_hdr)) 1602 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1603 1604 ip6 = mtod(m, struct ip6_hdr *); 1605 nxt = ip6->ip6_nxt; 1606 len = sizeof(struct ip6_hdr); 1607 nlen = 0; 1608 while (len < off) { 1609 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1610 switch (nxt) { 1611 case IPPROTO_FRAGMENT: 1612 nlen = sizeof(struct ip6_frag); 1613 break; 1614 case IPPROTO_AH: 1615 nlen = (ip6e.ip6e_len + 2) << 2; 1616 break; 1617 default: 1618 nlen = (ip6e.ip6e_len + 1) << 3; 1619 } 1620 len += nlen; 1621 nxt = ip6e.ip6e_nxt; 1622 } 1623 return (len - nlen); 1624} 1625 1626/* 1627 * get next header offset. m will be retained. 1628 */ 1629int 1630ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1631{ 1632 struct ip6_hdr ip6; 1633 struct ip6_ext ip6e; 1634 struct ip6_frag fh; 1635 1636 /* just in case */ 1637 if (m == NULL) 1638 panic("ip6_nexthdr: m == NULL"); 1639 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1640 return -1; 1641 1642 switch (proto) { 1643 case IPPROTO_IPV6: 1644 if (m->m_pkthdr.len < off + sizeof(ip6)) 1645 return -1; 1646 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1647 if (nxtp) 1648 *nxtp = ip6.ip6_nxt; 1649 off += sizeof(ip6); 1650 return off; 1651 1652 case IPPROTO_FRAGMENT: 1653 /* 1654 * terminate parsing if it is not the first fragment, 1655 * it does not make sense to parse through it. 1656 */ 1657 if (m->m_pkthdr.len < off + sizeof(fh)) 1658 return -1; 1659 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1660 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1661 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1662 return -1; 1663 if (nxtp) 1664 *nxtp = fh.ip6f_nxt; 1665 off += sizeof(struct ip6_frag); 1666 return off; 1667 1668 case IPPROTO_AH: 1669 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1670 return -1; 1671 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1672 if (nxtp) 1673 *nxtp = ip6e.ip6e_nxt; 1674 off += (ip6e.ip6e_len + 2) << 2; 1675 return off; 1676 1677 case IPPROTO_HOPOPTS: 1678 case IPPROTO_ROUTING: 1679 case IPPROTO_DSTOPTS: 1680 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1681 return -1; 1682 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1683 if (nxtp) 1684 *nxtp = ip6e.ip6e_nxt; 1685 off += (ip6e.ip6e_len + 1) << 3; 1686 return off; 1687 1688 case IPPROTO_NONE: 1689 case IPPROTO_ESP: 1690 case IPPROTO_IPCOMP: 1691 /* give up */ 1692 return -1; 1693 1694 default: 1695 return -1; 1696 } 1697 1698 /* NOTREACHED */ 1699} 1700 1701/* 1702 * get offset for the last header in the chain. m will be kept untainted. 1703 */ 1704int 1705ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1706{ 1707 int newoff; 1708 int nxt; 1709 1710 if (!nxtp) { 1711 nxt = -1; 1712 nxtp = &nxt; 1713 } 1714 while (1) { 1715 newoff = ip6_nexthdr(m, off, proto, nxtp); 1716 if (newoff < 0) 1717 return off; 1718 else if (newoff < off) 1719 return -1; /* invalid */ 1720 else if (newoff == off) 1721 return newoff; 1722 1723 off = newoff; 1724 proto = *nxtp; 1725 } 1726} 1727 1728/* 1729 * System control for IP6 1730 */ 1731 1732u_char inet6ctlerrmap[PRC_NCMDS] = { 1733 0, 0, 0, 0, 1734 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1735 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1736 EMSGSIZE, EHOSTUNREACH, 0, 0, 1737 0, 0, EHOSTUNREACH, 0, 1738 ENOPROTOOPT, ECONNREFUSED 1739}; 1740