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