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