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, 0, 0, sysctl_netinet6_intr_queue_maxlen, "I", 163 "Maximum size of the IPv6 input queue"); 164 165 #ifdef RSS 166 static 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 }; 174 175 static int 176 sysctl_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 } 188 SYSCTL_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 193 194 VNET_DEFINE(pfil_head_t, inet6_pfil_head); 195 196 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 197 VNET_PCPUSTAT_SYSINIT(ip6stat); 198 #ifdef VIMAGE 199 VNET_PCPUSTAT_SYSUNINIT(ip6stat); 200 #endif /* VIMAGE */ 201 202 struct rmlock in6_ifaddr_lock; 203 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 204 205 static 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 */ 211 void 212 ip6_init(void) 213 { 214 struct pfil_head_args args; 215 struct protosw *pr; 216 int i; 217 218 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 219 &V_ip6_auto_linklocal); 220 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 221 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 222 223 CK_STAILQ_INIT(&V_in6_ifaddrhead); 224 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 225 &V_in6_ifaddrhmask); 226 227 /* Initialize packet filter hooks. */ 228 args.pa_version = PFIL_VERSION; 229 args.pa_flags = PFIL_IN | PFIL_OUT; 230 args.pa_type = PFIL_TYPE_IP6; 231 args.pa_headname = PFIL_INET6_NAME; 232 V_inet6_pfil_head = pfil_head_register(&args); 233 234 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 235 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 236 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 237 printf("%s: WARNING: unable to register input helper hook\n", 238 __func__); 239 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 240 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 241 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 242 printf("%s: WARNING: unable to register output helper hook\n", 243 __func__); 244 245 scope6_init(); 246 addrsel_policy_init(); 247 nd6_init(); 248 frag6_init(); 249 250 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 251 252 /* Skip global initialization stuff for non-default instances. */ 253 #ifdef VIMAGE 254 if (!IS_DEFAULT_VNET(curvnet)) { 255 netisr_register_vnet(&ip6_nh); 256 #ifdef RSS 257 netisr_register_vnet(&ip6_direct_nh); 258 #endif 259 return; 260 } 261 #endif 262 263 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 264 if (pr == NULL) 265 panic("ip6_init"); 266 267 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 268 for (i = 0; i < IPPROTO_MAX; i++) 269 ip6_protox[i] = pr - inet6sw; 270 /* 271 * Cycle through IP protocols and put them into the appropriate place 272 * in ip6_protox[]. 273 */ 274 for (pr = inet6domain.dom_protosw; 275 pr < inet6domain.dom_protoswNPROTOSW; pr++) 276 if (pr->pr_domain->dom_family == PF_INET6 && 277 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 278 /* Be careful to only index valid IP protocols. */ 279 if (pr->pr_protocol < IPPROTO_MAX) 280 ip6_protox[pr->pr_protocol] = pr - inet6sw; 281 } 282 283 netisr_register(&ip6_nh); 284 #ifdef RSS 285 netisr_register(&ip6_direct_nh); 286 #endif 287 } 288 289 /* 290 * The protocol to be inserted into ip6_protox[] must be already registered 291 * in inet6sw[], either statically or through pf_proto_register(). 292 */ 293 int 294 ip6proto_register(short ip6proto) 295 { 296 struct protosw *pr; 297 298 /* Sanity checks. */ 299 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 300 return (EPROTONOSUPPORT); 301 302 /* 303 * The protocol slot must not be occupied by another protocol 304 * already. An index pointing to IPPROTO_RAW is unused. 305 */ 306 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 307 if (pr == NULL) 308 return (EPFNOSUPPORT); 309 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 310 return (EEXIST); 311 312 /* 313 * Find the protocol position in inet6sw[] and set the index. 314 */ 315 for (pr = inet6domain.dom_protosw; 316 pr < inet6domain.dom_protoswNPROTOSW; pr++) { 317 if (pr->pr_domain->dom_family == PF_INET6 && 318 pr->pr_protocol && pr->pr_protocol == ip6proto) { 319 ip6_protox[pr->pr_protocol] = pr - inet6sw; 320 return (0); 321 } 322 } 323 return (EPROTONOSUPPORT); 324 } 325 326 int 327 ip6proto_unregister(short ip6proto) 328 { 329 struct protosw *pr; 330 331 /* Sanity checks. */ 332 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 333 return (EPROTONOSUPPORT); 334 335 /* Check if the protocol was indeed registered. */ 336 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 337 if (pr == NULL) 338 return (EPFNOSUPPORT); 339 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 340 return (ENOENT); 341 342 /* Reset the protocol slot to IPPROTO_RAW. */ 343 ip6_protox[ip6proto] = pr - inet6sw; 344 return (0); 345 } 346 347 #ifdef VIMAGE 348 static void 349 ip6_destroy(void *unused __unused) 350 { 351 struct ifaddr *ifa, *nifa; 352 struct ifnet *ifp; 353 int error; 354 355 #ifdef RSS 356 netisr_unregister_vnet(&ip6_direct_nh); 357 #endif 358 netisr_unregister_vnet(&ip6_nh); 359 360 pfil_head_unregister(V_inet6_pfil_head); 361 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 362 if (error != 0) { 363 printf("%s: WARNING: unable to deregister input helper hook " 364 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 365 "error %d returned\n", __func__, error); 366 } 367 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 368 if (error != 0) { 369 printf("%s: WARNING: unable to deregister output helper hook " 370 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 371 "error %d returned\n", __func__, error); 372 } 373 374 /* Cleanup addresses. */ 375 IFNET_RLOCK(); 376 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 377 /* Cannot lock here - lock recursion. */ 378 /* IF_ADDR_LOCK(ifp); */ 379 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 380 381 if (ifa->ifa_addr->sa_family != AF_INET6) 382 continue; 383 in6_purgeaddr(ifa); 384 } 385 /* IF_ADDR_UNLOCK(ifp); */ 386 in6_ifdetach_destroy(ifp); 387 mld_domifdetach(ifp); 388 /* Make sure any routes are gone as well. */ 389 rt_flushifroutes_af(ifp, AF_INET6); 390 } 391 IFNET_RUNLOCK(); 392 393 frag6_destroy(); 394 nd6_destroy(); 395 in6_ifattach_destroy(); 396 397 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 398 } 399 400 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 401 #endif 402 403 static int 404 ip6_input_hbh(struct mbuf **mp, uint32_t *plen, uint32_t *rtalert, int *off, 405 int *nxt, int *ours) 406 { 407 struct mbuf *m; 408 struct ip6_hdr *ip6; 409 struct ip6_hbh *hbh; 410 411 if (ip6_hopopts_input(plen, rtalert, mp, off)) { 412 #if 0 /*touches NULL pointer*/ 413 in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard); 414 #endif 415 goto out; /* m have already been freed */ 416 } 417 418 /* adjust pointer */ 419 m = *mp; 420 ip6 = mtod(m, struct ip6_hdr *); 421 422 /* 423 * if the payload length field is 0 and the next header field 424 * indicates Hop-by-Hop Options header, then a Jumbo Payload 425 * option MUST be included. 426 */ 427 if (ip6->ip6_plen == 0 && *plen == 0) { 428 /* 429 * Note that if a valid jumbo payload option is 430 * contained, ip6_hopopts_input() must set a valid 431 * (non-zero) payload length to the variable plen. 432 */ 433 IP6STAT_INC(ip6s_badoptions); 434 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 435 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 436 icmp6_error(m, ICMP6_PARAM_PROB, 437 ICMP6_PARAMPROB_HEADER, 438 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 439 goto out; 440 } 441 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 442 hbh = (struct ip6_hbh *)(ip6 + 1); 443 *nxt = hbh->ip6h_nxt; 444 445 /* 446 * If we are acting as a router and the packet contains a 447 * router alert option, see if we know the option value. 448 * Currently, we only support the option value for MLD, in which 449 * case we should pass the packet to the multicast routing 450 * daemon. 451 */ 452 if (*rtalert != ~0) { 453 switch (*rtalert) { 454 case IP6OPT_RTALERT_MLD: 455 if (V_ip6_forwarding) 456 *ours = 1; 457 break; 458 default: 459 /* 460 * RFC2711 requires unrecognized values must be 461 * silently ignored. 462 */ 463 break; 464 } 465 } 466 467 return (0); 468 469 out: 470 return (1); 471 } 472 473 #ifdef RSS 474 /* 475 * IPv6 direct input routine. 476 * 477 * This is called when reinjecting completed fragments where 478 * all of the previous checking and book-keeping has been done. 479 */ 480 void 481 ip6_direct_input(struct mbuf *m) 482 { 483 int off, nxt; 484 int nest; 485 struct m_tag *mtag; 486 struct ip6_direct_ctx *ip6dc; 487 488 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 489 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 490 491 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 492 nxt = ip6dc->ip6dc_nxt; 493 off = ip6dc->ip6dc_off; 494 495 nest = 0; 496 497 m_tag_delete(m, mtag); 498 499 while (nxt != IPPROTO_DONE) { 500 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 501 IP6STAT_INC(ip6s_toomanyhdr); 502 goto bad; 503 } 504 505 /* 506 * protection against faulty packet - there should be 507 * more sanity checks in header chain processing. 508 */ 509 if (m->m_pkthdr.len < off) { 510 IP6STAT_INC(ip6s_tooshort); 511 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 512 goto bad; 513 } 514 515 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 516 if (IPSEC_ENABLED(ipv6)) { 517 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 518 return; 519 } 520 #endif /* IPSEC */ 521 522 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 523 } 524 return; 525 bad: 526 m_freem(m); 527 } 528 #endif 529 530 void 531 ip6_input(struct mbuf *m) 532 { 533 struct in6_addr odst; 534 struct ip6_hdr *ip6; 535 struct in6_ifaddr *ia; 536 struct ifnet *rcvif; 537 u_int32_t plen; 538 u_int32_t rtalert = ~0; 539 int off = sizeof(struct ip6_hdr), nest; 540 int nxt, ours = 0; 541 int srcrt = 0; 542 543 /* 544 * Drop the packet if IPv6 operation is disabled on the interface. 545 */ 546 rcvif = m->m_pkthdr.rcvif; 547 if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED)) 548 goto bad; 549 550 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 551 /* 552 * should the inner packet be considered authentic? 553 * see comment in ah4_input(). 554 * NB: m cannot be NULL when passed to the input routine 555 */ 556 557 m->m_flags &= ~M_AUTHIPHDR; 558 m->m_flags &= ~M_AUTHIPDGM; 559 560 #endif /* IPSEC */ 561 562 if (m->m_flags & M_FASTFWD_OURS) { 563 /* 564 * Firewall changed destination to local. 565 */ 566 ip6 = mtod(m, struct ip6_hdr *); 567 goto passin; 568 } 569 570 /* 571 * mbuf statistics 572 */ 573 if (m->m_flags & M_EXT) { 574 if (m->m_next) 575 IP6STAT_INC(ip6s_mext2m); 576 else 577 IP6STAT_INC(ip6s_mext1); 578 } else { 579 if (m->m_next) { 580 if (m->m_flags & M_LOOP) { 581 IP6STAT_INC(ip6s_m2m[V_loif->if_index]); 582 } else if (rcvif->if_index < IP6S_M2MMAX) 583 IP6STAT_INC(ip6s_m2m[rcvif->if_index]); 584 else 585 IP6STAT_INC(ip6s_m2m[0]); 586 } else 587 IP6STAT_INC(ip6s_m1); 588 } 589 590 in6_ifstat_inc(rcvif, ifs6_in_receive); 591 IP6STAT_INC(ip6s_total); 592 593 /* 594 * L2 bridge code and some other code can return mbuf chain 595 * that does not conform to KAME requirement. too bad. 596 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 597 */ 598 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 599 struct mbuf *n; 600 601 if (m->m_pkthdr.len > MHLEN) 602 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 603 else 604 n = m_gethdr(M_NOWAIT, MT_DATA); 605 if (n == NULL) 606 goto bad; 607 608 m_move_pkthdr(n, m); 609 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 610 n->m_len = n->m_pkthdr.len; 611 m_freem(m); 612 m = n; 613 } 614 if (m->m_len < sizeof(struct ip6_hdr)) { 615 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 616 IP6STAT_INC(ip6s_toosmall); 617 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 618 goto bad; 619 } 620 } 621 622 ip6 = mtod(m, struct ip6_hdr *); 623 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 624 IP6STAT_INC(ip6s_badvers); 625 in6_ifstat_inc(rcvif, ifs6_in_hdrerr); 626 goto bad; 627 } 628 629 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 630 IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6); 631 632 /* 633 * Check against address spoofing/corruption. 634 */ 635 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 636 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 637 /* 638 * XXX: "badscope" is not very suitable for a multicast source. 639 */ 640 IP6STAT_INC(ip6s_badscope); 641 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 642 goto bad; 643 } 644 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 645 !(m->m_flags & M_LOOP)) { 646 /* 647 * In this case, the packet should come from the loopback 648 * interface. However, we cannot just check the if_flags, 649 * because ip6_mloopback() passes the "actual" interface 650 * as the outgoing/incoming interface. 651 */ 652 IP6STAT_INC(ip6s_badscope); 653 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 654 goto bad; 655 } 656 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 657 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 658 /* 659 * RFC4291 2.7: 660 * Nodes must not originate a packet to a multicast address 661 * whose scop field contains the reserved value 0; if such 662 * a packet is received, it must be silently dropped. 663 */ 664 IP6STAT_INC(ip6s_badscope); 665 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 666 goto bad; 667 } 668 #ifdef ALTQ 669 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 670 /* packet is dropped by traffic conditioner */ 671 return; 672 } 673 #endif 674 /* 675 * The following check is not documented in specs. A malicious 676 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 677 * and bypass security checks (act as if it was from 127.0.0.1 by using 678 * IPv6 src ::ffff:127.0.0.1). Be cautious. 679 * 680 * This check chokes if we are in an SIIT cloud. As none of BSDs 681 * support IPv4-less kernel compilation, we cannot support SIIT 682 * environment at all. So, it makes more sense for us to reject any 683 * malicious packets for non-SIIT environment, than try to do a 684 * partial support for SIIT environment. 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_IN(V_inet6_pfil_head)) 748 goto passin; 749 750 odst = ip6->ip6_dst; 751 if (pfil_run_hooks(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif, PFIL_IN, 752 NULL) != PFIL_PASS) 753 return; 754 ip6 = mtod(m, struct ip6_hdr *); 755 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 756 if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP && 757 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 758 /* 759 * Directly ship the packet on. This allows forwarding 760 * packets originally destined to us to some other directly 761 * connected host. 762 */ 763 ip6_forward(m, 1); 764 return; 765 } 766 767 passin: 768 /* 769 * Disambiguate address scope zones (if there is ambiguity). 770 * We first make sure that the original source or destination address 771 * is not in our internal form for scoped addresses. Such addresses 772 * are not necessarily invalid spec-wise, but we cannot accept them due 773 * to the usage conflict. 774 * in6_setscope() then also checks and rejects the cases where src or 775 * dst are the loopback address and the receiving interface 776 * is not loopback. 777 */ 778 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 779 IP6STAT_INC(ip6s_badscope); /* XXX */ 780 goto bad; 781 } 782 if (in6_setscope(&ip6->ip6_src, rcvif, NULL) || 783 in6_setscope(&ip6->ip6_dst, rcvif, NULL)) { 784 IP6STAT_INC(ip6s_badscope); 785 goto bad; 786 } 787 if (m->m_flags & M_FASTFWD_OURS) { 788 m->m_flags &= ~M_FASTFWD_OURS; 789 ours = 1; 790 goto hbhcheck; 791 } 792 /* 793 * Multicast check. Assume packet is for us to avoid 794 * prematurely taking locks. 795 */ 796 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 797 ours = 1; 798 in6_ifstat_inc(rcvif, ifs6_in_mcast); 799 goto hbhcheck; 800 } 801 /* 802 * Unicast check 803 * XXX: For now we keep link-local IPv6 addresses with embedded 804 * scope zone id, therefore we use zero zoneid here. 805 */ 806 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */); 807 if (ia != NULL) { 808 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 809 char ip6bufs[INET6_ADDRSTRLEN]; 810 char ip6bufd[INET6_ADDRSTRLEN]; 811 /* address is not ready, so discard the packet. */ 812 nd6log((LOG_INFO, 813 "ip6_input: packet to an unready address %s->%s\n", 814 ip6_sprintf(ip6bufs, &ip6->ip6_src), 815 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 816 ifa_free(&ia->ia_ifa); 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 ifa_free(&ia->ia_ifa); 823 ours = 1; 824 goto hbhcheck; 825 } 826 827 /* 828 * Now there is no reason to process the packet if it's not our own 829 * and we're not a router. 830 */ 831 if (!V_ip6_forwarding) { 832 IP6STAT_INC(ip6s_cantforward); 833 goto bad; 834 } 835 836 hbhcheck: 837 /* 838 * Process Hop-by-Hop options header if it's contained. 839 * m may be modified in ip6_hopopts_input(). 840 * If a JumboPayload option is included, plen will also be modified. 841 */ 842 plen = (u_int32_t)ntohs(ip6->ip6_plen); 843 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 844 if (ip6_input_hbh(&m, &plen, &rtalert, &off, &nxt, &ours) != 0) 845 return; 846 } else 847 nxt = ip6->ip6_nxt; 848 849 /* 850 * Use mbuf flags to propagate Router Alert option to 851 * ICMPv6 layer, as hop-by-hop options have been stripped. 852 */ 853 if (rtalert != ~0) 854 m->m_flags |= M_RTALERT_MLD; 855 856 /* 857 * Check that the amount of data in the buffers 858 * is as at least much as the IPv6 header would have us expect. 859 * Trim mbufs if longer than we expect. 860 * Drop packet if shorter than we expect. 861 */ 862 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 863 IP6STAT_INC(ip6s_tooshort); 864 in6_ifstat_inc(rcvif, ifs6_in_truncated); 865 goto bad; 866 } 867 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 868 if (m->m_len == m->m_pkthdr.len) { 869 m->m_len = sizeof(struct ip6_hdr) + plen; 870 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 871 } else 872 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 873 } 874 875 /* 876 * Forward if desirable. 877 */ 878 if (V_ip6_mrouter && 879 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 880 /* 881 * If we are acting as a multicast router, all 882 * incoming multicast packets are passed to the 883 * kernel-level multicast forwarding function. 884 * The packet is returned (relatively) intact; if 885 * ip6_mforward() returns a non-zero value, the packet 886 * must be discarded, else it may be accepted below. 887 * 888 * XXX TODO: Check hlim and multicast scope here to avoid 889 * unnecessarily calling into ip6_mforward(). 890 */ 891 if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) { 892 IP6STAT_INC(ip6s_cantforward); 893 goto bad; 894 } 895 } else if (!ours) { 896 ip6_forward(m, srcrt); 897 return; 898 } 899 900 ip6 = mtod(m, struct ip6_hdr *); 901 902 /* 903 * Malicious party may be able to use IPv4 mapped addr to confuse 904 * tcp/udp stack and bypass security checks (act as if it was from 905 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 906 * 907 * For SIIT end node behavior, you may want to disable the check. 908 * However, you will become vulnerable to attacks using IPv4 mapped 909 * source. 910 */ 911 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 912 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 913 IP6STAT_INC(ip6s_badscope); 914 in6_ifstat_inc(rcvif, ifs6_in_addrerr); 915 goto bad; 916 } 917 918 /* 919 * Tell launch routine the next header 920 */ 921 IP6STAT_INC(ip6s_delivered); 922 in6_ifstat_inc(rcvif, ifs6_in_deliver); 923 nest = 0; 924 925 while (nxt != IPPROTO_DONE) { 926 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 927 IP6STAT_INC(ip6s_toomanyhdr); 928 goto bad; 929 } 930 931 /* 932 * protection against faulty packet - there should be 933 * more sanity checks in header chain processing. 934 */ 935 if (m->m_pkthdr.len < off) { 936 IP6STAT_INC(ip6s_tooshort); 937 in6_ifstat_inc(rcvif, ifs6_in_truncated); 938 goto bad; 939 } 940 941 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 942 if (IPSEC_ENABLED(ipv6)) { 943 if (IPSEC_INPUT(ipv6, m, off, nxt) != 0) 944 return; 945 } 946 #endif /* IPSEC */ 947 948 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 949 } 950 return; 951 bad: 952 in6_ifstat_inc(rcvif, ifs6_in_discard); 953 if (m != NULL) 954 m_freem(m); 955 } 956 957 /* 958 * Hop-by-Hop options header processing. If a valid jumbo payload option is 959 * included, the real payload length will be stored in plenp. 960 * 961 * rtalertp - XXX: should be stored more smart way 962 */ 963 static int 964 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 965 struct mbuf **mp, int *offp) 966 { 967 struct mbuf *m = *mp; 968 int off = *offp, hbhlen; 969 struct ip6_hbh *hbh; 970 971 /* validation of the length of the header */ 972 if (m->m_len < off + sizeof(*hbh)) { 973 m = m_pullup(m, off + sizeof(*hbh)); 974 if (m == NULL) { 975 IP6STAT_INC(ip6s_exthdrtoolong); 976 *mp = NULL; 977 return (-1); 978 } 979 } 980 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 981 hbhlen = (hbh->ip6h_len + 1) << 3; 982 983 if (m->m_len < off + hbhlen) { 984 m = m_pullup(m, off + hbhlen); 985 if (m == NULL) { 986 IP6STAT_INC(ip6s_exthdrtoolong); 987 *mp = NULL; 988 return (-1); 989 } 990 } 991 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 992 off += hbhlen; 993 hbhlen -= sizeof(struct ip6_hbh); 994 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 995 hbhlen, rtalertp, plenp) < 0) { 996 *mp = NULL; 997 return (-1); 998 } 999 1000 *offp = off; 1001 *mp = m; 1002 return (0); 1003 } 1004 1005 /* 1006 * Search header for all Hop-by-hop options and process each option. 1007 * This function is separate from ip6_hopopts_input() in order to 1008 * handle a case where the sending node itself process its hop-by-hop 1009 * options header. In such a case, the function is called from ip6_output(). 1010 * 1011 * The function assumes that hbh header is located right after the IPv6 header 1012 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 1013 * opthead + hbhlen is located in contiguous memory region. 1014 */ 1015 int 1016 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 1017 u_int32_t *rtalertp, u_int32_t *plenp) 1018 { 1019 struct ip6_hdr *ip6; 1020 int optlen = 0; 1021 u_int8_t *opt = opthead; 1022 u_int16_t rtalert_val; 1023 u_int32_t jumboplen; 1024 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 1025 1026 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 1027 switch (*opt) { 1028 case IP6OPT_PAD1: 1029 optlen = 1; 1030 break; 1031 case IP6OPT_PADN: 1032 if (hbhlen < IP6OPT_MINLEN) { 1033 IP6STAT_INC(ip6s_toosmall); 1034 goto bad; 1035 } 1036 optlen = *(opt + 1) + 2; 1037 break; 1038 case IP6OPT_ROUTER_ALERT: 1039 /* XXX may need check for alignment */ 1040 if (hbhlen < IP6OPT_RTALERT_LEN) { 1041 IP6STAT_INC(ip6s_toosmall); 1042 goto bad; 1043 } 1044 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1045 /* XXX stat */ 1046 icmp6_error(m, ICMP6_PARAM_PROB, 1047 ICMP6_PARAMPROB_HEADER, 1048 erroff + opt + 1 - opthead); 1049 return (-1); 1050 } 1051 optlen = IP6OPT_RTALERT_LEN; 1052 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1053 *rtalertp = ntohs(rtalert_val); 1054 break; 1055 case IP6OPT_JUMBO: 1056 /* XXX may need check for alignment */ 1057 if (hbhlen < IP6OPT_JUMBO_LEN) { 1058 IP6STAT_INC(ip6s_toosmall); 1059 goto bad; 1060 } 1061 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1062 /* XXX stat */ 1063 icmp6_error(m, ICMP6_PARAM_PROB, 1064 ICMP6_PARAMPROB_HEADER, 1065 erroff + opt + 1 - opthead); 1066 return (-1); 1067 } 1068 optlen = IP6OPT_JUMBO_LEN; 1069 1070 /* 1071 * IPv6 packets that have non 0 payload length 1072 * must not contain a jumbo payload option. 1073 */ 1074 ip6 = mtod(m, struct ip6_hdr *); 1075 if (ip6->ip6_plen) { 1076 IP6STAT_INC(ip6s_badoptions); 1077 icmp6_error(m, ICMP6_PARAM_PROB, 1078 ICMP6_PARAMPROB_HEADER, 1079 erroff + opt - opthead); 1080 return (-1); 1081 } 1082 1083 /* 1084 * We may see jumbolen in unaligned location, so 1085 * we'd need to perform bcopy(). 1086 */ 1087 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1088 jumboplen = (u_int32_t)htonl(jumboplen); 1089 1090 #if 1 1091 /* 1092 * if there are multiple jumbo payload options, 1093 * *plenp will be non-zero and the packet will be 1094 * rejected. 1095 * the behavior may need some debate in ipngwg - 1096 * multiple options does not make sense, however, 1097 * there's no explicit mention in specification. 1098 */ 1099 if (*plenp != 0) { 1100 IP6STAT_INC(ip6s_badoptions); 1101 icmp6_error(m, ICMP6_PARAM_PROB, 1102 ICMP6_PARAMPROB_HEADER, 1103 erroff + opt + 2 - opthead); 1104 return (-1); 1105 } 1106 #endif 1107 1108 /* 1109 * jumbo payload length must be larger than 65535. 1110 */ 1111 if (jumboplen <= IPV6_MAXPACKET) { 1112 IP6STAT_INC(ip6s_badoptions); 1113 icmp6_error(m, ICMP6_PARAM_PROB, 1114 ICMP6_PARAMPROB_HEADER, 1115 erroff + opt + 2 - opthead); 1116 return (-1); 1117 } 1118 *plenp = jumboplen; 1119 1120 break; 1121 default: /* unknown option */ 1122 if (hbhlen < IP6OPT_MINLEN) { 1123 IP6STAT_INC(ip6s_toosmall); 1124 goto bad; 1125 } 1126 optlen = ip6_unknown_opt(opt, m, 1127 erroff + opt - opthead); 1128 if (optlen == -1) 1129 return (-1); 1130 optlen += 2; 1131 break; 1132 } 1133 } 1134 1135 return (0); 1136 1137 bad: 1138 m_freem(m); 1139 return (-1); 1140 } 1141 1142 /* 1143 * Unknown option processing. 1144 * The third argument `off' is the offset from the IPv6 header to the option, 1145 * which is necessary if the IPv6 header the and option header and IPv6 header 1146 * is not contiguous in order to return an ICMPv6 error. 1147 */ 1148 int 1149 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1150 { 1151 struct ip6_hdr *ip6; 1152 1153 switch (IP6OPT_TYPE(*optp)) { 1154 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1155 return ((int)*(optp + 1)); 1156 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1157 m_freem(m); 1158 return (-1); 1159 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1160 IP6STAT_INC(ip6s_badoptions); 1161 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1162 return (-1); 1163 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1164 IP6STAT_INC(ip6s_badoptions); 1165 ip6 = mtod(m, struct ip6_hdr *); 1166 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1167 (m->m_flags & (M_BCAST|M_MCAST))) 1168 m_freem(m); 1169 else 1170 icmp6_error(m, ICMP6_PARAM_PROB, 1171 ICMP6_PARAMPROB_OPTION, off); 1172 return (-1); 1173 } 1174 1175 m_freem(m); /* XXX: NOTREACHED */ 1176 return (-1); 1177 } 1178 1179 /* 1180 * Create the "control" list for this pcb. 1181 * These functions will not modify mbuf chain at all. 1182 * 1183 * The routine will be called from upper layer handlers like tcp6_input(). 1184 * Thus the routine assumes that the caller (tcp6_input) have already 1185 * called m_pullup() and all the extension headers are located in the 1186 * very first mbuf on the mbuf chain. 1187 * 1188 * ip6_savecontrol_v4 will handle those options that are possible to be 1189 * set on a v4-mapped socket. 1190 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1191 * options and handle the v6-only ones itself. 1192 */ 1193 struct mbuf ** 1194 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1195 int *v4only) 1196 { 1197 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1198 1199 #ifdef SO_TIMESTAMP 1200 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1201 union { 1202 struct timeval tv; 1203 struct bintime bt; 1204 struct timespec ts; 1205 } t; 1206 struct bintime boottimebin, bt1; 1207 struct timespec ts1; 1208 bool stamped; 1209 1210 stamped = false; 1211 switch (inp->inp_socket->so_ts_clock) { 1212 case SO_TS_REALTIME_MICRO: 1213 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1214 M_TSTMP)) { 1215 mbuf_tstmp2timespec(m, &ts1); 1216 timespec2bintime(&ts1, &bt1); 1217 getboottimebin(&boottimebin); 1218 bintime_add(&bt1, &boottimebin); 1219 bintime2timeval(&bt1, &t.tv); 1220 } else { 1221 microtime(&t.tv); 1222 } 1223 *mp = sbcreatecontrol((caddr_t) &t.tv, sizeof(t.tv), 1224 SCM_TIMESTAMP, SOL_SOCKET); 1225 if (*mp != NULL) { 1226 mp = &(*mp)->m_next; 1227 stamped = true; 1228 } 1229 break; 1230 1231 case SO_TS_BINTIME: 1232 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1233 M_TSTMP)) { 1234 mbuf_tstmp2timespec(m, &ts1); 1235 timespec2bintime(&ts1, &t.bt); 1236 getboottimebin(&boottimebin); 1237 bintime_add(&t.bt, &boottimebin); 1238 } else { 1239 bintime(&t.bt); 1240 } 1241 *mp = sbcreatecontrol((caddr_t)&t.bt, sizeof(t.bt), 1242 SCM_BINTIME, SOL_SOCKET); 1243 if (*mp != NULL) { 1244 mp = &(*mp)->m_next; 1245 stamped = true; 1246 } 1247 break; 1248 1249 case SO_TS_REALTIME: 1250 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1251 M_TSTMP)) { 1252 mbuf_tstmp2timespec(m, &t.ts); 1253 getboottimebin(&boottimebin); 1254 bintime2timespec(&boottimebin, &ts1); 1255 timespecadd(&t.ts, &ts1, &t.ts); 1256 } else { 1257 nanotime(&t.ts); 1258 } 1259 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts), 1260 SCM_REALTIME, SOL_SOCKET); 1261 if (*mp != NULL) { 1262 mp = &(*mp)->m_next; 1263 stamped = true; 1264 } 1265 break; 1266 1267 case SO_TS_MONOTONIC: 1268 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1269 M_TSTMP)) 1270 mbuf_tstmp2timespec(m, &t.ts); 1271 else 1272 nanouptime(&t.ts); 1273 *mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts), 1274 SCM_MONOTONIC, SOL_SOCKET); 1275 if (*mp != NULL) { 1276 mp = &(*mp)->m_next; 1277 stamped = true; 1278 } 1279 break; 1280 1281 default: 1282 panic("unknown (corrupted) so_ts_clock"); 1283 } 1284 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == 1285 (M_PKTHDR | M_TSTMP)) { 1286 struct sock_timestamp_info sti; 1287 1288 bzero(&sti, sizeof(sti)); 1289 sti.st_info_flags = ST_INFO_HW; 1290 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1291 sti.st_info_flags |= ST_INFO_HW_HPREC; 1292 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), 1293 SCM_TIME_INFO, SOL_SOCKET); 1294 if (*mp != NULL) 1295 mp = &(*mp)->m_next; 1296 } 1297 } 1298 #endif 1299 1300 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1301 /* RFC 2292 sec. 5 */ 1302 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1303 struct in6_pktinfo pi6; 1304 1305 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1306 #ifdef INET 1307 struct ip *ip; 1308 1309 ip = mtod(m, struct ip *); 1310 pi6.ipi6_addr.s6_addr32[0] = 0; 1311 pi6.ipi6_addr.s6_addr32[1] = 0; 1312 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1313 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1314 #else 1315 /* We won't hit this code */ 1316 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1317 #endif 1318 } else { 1319 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1320 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1321 } 1322 pi6.ipi6_ifindex = 1323 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1324 1325 *mp = sbcreatecontrol((caddr_t) &pi6, 1326 sizeof(struct in6_pktinfo), 1327 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1328 if (*mp) 1329 mp = &(*mp)->m_next; 1330 } 1331 1332 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1333 int hlim; 1334 1335 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1336 #ifdef INET 1337 struct ip *ip; 1338 1339 ip = mtod(m, struct ip *); 1340 hlim = ip->ip_ttl; 1341 #else 1342 /* We won't hit this code */ 1343 hlim = 0; 1344 #endif 1345 } else { 1346 hlim = ip6->ip6_hlim & 0xff; 1347 } 1348 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1349 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1350 IPPROTO_IPV6); 1351 if (*mp) 1352 mp = &(*mp)->m_next; 1353 } 1354 1355 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1356 int tclass; 1357 1358 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1359 #ifdef INET 1360 struct ip *ip; 1361 1362 ip = mtod(m, struct ip *); 1363 tclass = ip->ip_tos; 1364 #else 1365 /* We won't hit this code */ 1366 tclass = 0; 1367 #endif 1368 } else { 1369 u_int32_t flowinfo; 1370 1371 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1372 flowinfo >>= 20; 1373 tclass = flowinfo & 0xff; 1374 } 1375 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int), 1376 IPV6_TCLASS, IPPROTO_IPV6); 1377 if (*mp) 1378 mp = &(*mp)->m_next; 1379 } 1380 1381 if (v4only != NULL) { 1382 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1383 *v4only = 1; 1384 } else { 1385 *v4only = 0; 1386 } 1387 } 1388 1389 return (mp); 1390 } 1391 1392 void 1393 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp) 1394 { 1395 struct ip6_hdr *ip6; 1396 int v4only = 0; 1397 1398 mp = ip6_savecontrol_v4(inp, m, mp, &v4only); 1399 if (v4only) 1400 return; 1401 1402 ip6 = mtod(m, struct ip6_hdr *); 1403 /* 1404 * IPV6_HOPOPTS socket option. Recall that we required super-user 1405 * privilege for the option (see ip6_ctloutput), but it might be too 1406 * strict, since there might be some hop-by-hop options which can be 1407 * returned to normal user. 1408 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1409 */ 1410 if ((inp->inp_flags & IN6P_HOPOPTS) != 0) { 1411 /* 1412 * Check if a hop-by-hop options header is contatined in the 1413 * received packet, and if so, store the options as ancillary 1414 * data. Note that a hop-by-hop options header must be 1415 * just after the IPv6 header, which is assured through the 1416 * IPv6 input processing. 1417 */ 1418 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1419 struct ip6_hbh *hbh; 1420 int hbhlen; 1421 1422 hbh = (struct ip6_hbh *)(ip6 + 1); 1423 hbhlen = (hbh->ip6h_len + 1) << 3; 1424 1425 /* 1426 * XXX: We copy the whole header even if a 1427 * jumbo payload option is included, the option which 1428 * is to be removed before returning according to 1429 * RFC2292. 1430 * Note: this constraint is removed in RFC3542 1431 */ 1432 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1433 IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1434 IPPROTO_IPV6); 1435 if (*mp) 1436 mp = &(*mp)->m_next; 1437 } 1438 } 1439 1440 if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1441 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1442 1443 /* 1444 * Search for destination options headers or routing 1445 * header(s) through the header chain, and stores each 1446 * header as ancillary data. 1447 * Note that the order of the headers remains in 1448 * the chain of ancillary data. 1449 */ 1450 while (1) { /* is explicit loop prevention necessary? */ 1451 struct ip6_ext *ip6e = NULL; 1452 int elen; 1453 1454 /* 1455 * if it is not an extension header, don't try to 1456 * pull it from the chain. 1457 */ 1458 switch (nxt) { 1459 case IPPROTO_DSTOPTS: 1460 case IPPROTO_ROUTING: 1461 case IPPROTO_HOPOPTS: 1462 case IPPROTO_AH: /* is it possible? */ 1463 break; 1464 default: 1465 goto loopend; 1466 } 1467 1468 if (off + sizeof(*ip6e) > m->m_len) 1469 goto loopend; 1470 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1471 if (nxt == IPPROTO_AH) 1472 elen = (ip6e->ip6e_len + 2) << 2; 1473 else 1474 elen = (ip6e->ip6e_len + 1) << 3; 1475 if (off + elen > m->m_len) 1476 goto loopend; 1477 1478 switch (nxt) { 1479 case IPPROTO_DSTOPTS: 1480 if (!(inp->inp_flags & IN6P_DSTOPTS)) 1481 break; 1482 1483 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1484 IS2292(inp, 1485 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1486 IPPROTO_IPV6); 1487 if (*mp) 1488 mp = &(*mp)->m_next; 1489 break; 1490 case IPPROTO_ROUTING: 1491 if (!(inp->inp_flags & IN6P_RTHDR)) 1492 break; 1493 1494 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1495 IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR), 1496 IPPROTO_IPV6); 1497 if (*mp) 1498 mp = &(*mp)->m_next; 1499 break; 1500 case IPPROTO_HOPOPTS: 1501 case IPPROTO_AH: /* is it possible? */ 1502 break; 1503 1504 default: 1505 /* 1506 * other cases have been filtered in the above. 1507 * none will visit this case. here we supply 1508 * the code just in case (nxt overwritten or 1509 * other cases). 1510 */ 1511 goto loopend; 1512 1513 } 1514 1515 /* proceed with the next header. */ 1516 off += elen; 1517 nxt = ip6e->ip6e_nxt; 1518 ip6e = NULL; 1519 } 1520 loopend: 1521 ; 1522 } 1523 1524 if (inp->inp_flags2 & INP_RECVFLOWID) { 1525 uint32_t flowid, flow_type; 1526 1527 flowid = m->m_pkthdr.flowid; 1528 flow_type = M_HASHTYPE_GET(m); 1529 1530 /* 1531 * XXX should handle the failure of one or the 1532 * other - don't populate both? 1533 */ 1534 *mp = sbcreatecontrol((caddr_t) &flowid, 1535 sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6); 1536 if (*mp) 1537 mp = &(*mp)->m_next; 1538 *mp = sbcreatecontrol((caddr_t) &flow_type, 1539 sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6); 1540 if (*mp) 1541 mp = &(*mp)->m_next; 1542 } 1543 1544 #ifdef RSS 1545 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1546 uint32_t flowid, flow_type; 1547 uint32_t rss_bucketid; 1548 1549 flowid = m->m_pkthdr.flowid; 1550 flow_type = M_HASHTYPE_GET(m); 1551 1552 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1553 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1554 sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6); 1555 if (*mp) 1556 mp = &(*mp)->m_next; 1557 } 1558 } 1559 #endif 1560 1561 } 1562 #undef IS2292 1563 1564 void 1565 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1566 { 1567 struct socket *so; 1568 struct mbuf *m_mtu; 1569 struct ip6_mtuinfo mtuctl; 1570 1571 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1572 /* 1573 * Notify the error by sending IPV6_PATHMTU ancillary data if 1574 * application wanted to know the MTU value. 1575 * NOTE: we notify disconnected sockets, because some udp 1576 * applications keep sending sockets disconnected. 1577 * NOTE: our implementation doesn't notify connected sockets that has 1578 * foreign address that is different than given destination addresses 1579 * (this is permitted by RFC 3542). 1580 */ 1581 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1582 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1583 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1584 return; 1585 1586 mtuctl.ip6m_mtu = mtu; 1587 mtuctl.ip6m_addr = *dst; 1588 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1589 return; 1590 1591 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1592 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1593 return; 1594 1595 so = inp->inp_socket; 1596 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1597 == 0) { 1598 m_freem(m_mtu); 1599 /* XXX: should count statistics */ 1600 } else 1601 sorwakeup(so); 1602 } 1603 1604 /* 1605 * Get pointer to the previous header followed by the header 1606 * currently processed. 1607 */ 1608 int 1609 ip6_get_prevhdr(const struct mbuf *m, int off) 1610 { 1611 struct ip6_ext ip6e; 1612 struct ip6_hdr *ip6; 1613 int len, nlen, nxt; 1614 1615 if (off == sizeof(struct ip6_hdr)) 1616 return (offsetof(struct ip6_hdr, ip6_nxt)); 1617 if (off < sizeof(struct ip6_hdr)) 1618 panic("%s: off < sizeof(struct ip6_hdr)", __func__); 1619 1620 ip6 = mtod(m, struct ip6_hdr *); 1621 nxt = ip6->ip6_nxt; 1622 len = sizeof(struct ip6_hdr); 1623 nlen = 0; 1624 while (len < off) { 1625 m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e); 1626 switch (nxt) { 1627 case IPPROTO_FRAGMENT: 1628 nlen = sizeof(struct ip6_frag); 1629 break; 1630 case IPPROTO_AH: 1631 nlen = (ip6e.ip6e_len + 2) << 2; 1632 break; 1633 default: 1634 nlen = (ip6e.ip6e_len + 1) << 3; 1635 } 1636 len += nlen; 1637 nxt = ip6e.ip6e_nxt; 1638 } 1639 return (len - nlen); 1640 } 1641 1642 /* 1643 * get next header offset. m will be retained. 1644 */ 1645 int 1646 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1647 { 1648 struct ip6_hdr ip6; 1649 struct ip6_ext ip6e; 1650 struct ip6_frag fh; 1651 1652 /* just in case */ 1653 if (m == NULL) 1654 panic("ip6_nexthdr: m == NULL"); 1655 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1656 return -1; 1657 1658 switch (proto) { 1659 case IPPROTO_IPV6: 1660 if (m->m_pkthdr.len < off + sizeof(ip6)) 1661 return -1; 1662 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1663 if (nxtp) 1664 *nxtp = ip6.ip6_nxt; 1665 off += sizeof(ip6); 1666 return off; 1667 1668 case IPPROTO_FRAGMENT: 1669 /* 1670 * terminate parsing if it is not the first fragment, 1671 * it does not make sense to parse through it. 1672 */ 1673 if (m->m_pkthdr.len < off + sizeof(fh)) 1674 return -1; 1675 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1676 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1677 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1678 return -1; 1679 if (nxtp) 1680 *nxtp = fh.ip6f_nxt; 1681 off += sizeof(struct ip6_frag); 1682 return off; 1683 1684 case IPPROTO_AH: 1685 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1686 return -1; 1687 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1688 if (nxtp) 1689 *nxtp = ip6e.ip6e_nxt; 1690 off += (ip6e.ip6e_len + 2) << 2; 1691 return off; 1692 1693 case IPPROTO_HOPOPTS: 1694 case IPPROTO_ROUTING: 1695 case IPPROTO_DSTOPTS: 1696 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1697 return -1; 1698 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1699 if (nxtp) 1700 *nxtp = ip6e.ip6e_nxt; 1701 off += (ip6e.ip6e_len + 1) << 3; 1702 return off; 1703 1704 case IPPROTO_NONE: 1705 case IPPROTO_ESP: 1706 case IPPROTO_IPCOMP: 1707 /* give up */ 1708 return -1; 1709 1710 default: 1711 return -1; 1712 } 1713 1714 /* NOTREACHED */ 1715 } 1716 1717 /* 1718 * get offset for the last header in the chain. m will be kept untainted. 1719 */ 1720 int 1721 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1722 { 1723 int newoff; 1724 int nxt; 1725 1726 if (!nxtp) { 1727 nxt = -1; 1728 nxtp = &nxt; 1729 } 1730 while (1) { 1731 newoff = ip6_nexthdr(m, off, proto, nxtp); 1732 if (newoff < 0) 1733 return off; 1734 else if (newoff < off) 1735 return -1; /* invalid */ 1736 else if (newoff == off) 1737 return newoff; 1738 1739 off = newoff; 1740 proto = *nxtp; 1741 } 1742 } 1743 1744 /* 1745 * System control for IP6 1746 */ 1747 1748 u_char inet6ctlerrmap[PRC_NCMDS] = { 1749 0, 0, 0, 0, 1750 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1751 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1752 EMSGSIZE, EHOSTUNREACH, 0, 0, 1753 0, 0, EHOSTUNREACH, 0, 1754 ENOPROTOOPT, ECONNREFUSED 1755 }; 1756