1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $ 30 */ 31 32 /*- 33 * Copyright (c) 1982, 1986, 1988, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 4. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 61 */ 62 63 #include <sys/cdefs.h> 64 __FBSDID("$FreeBSD$"); 65 66 #include "opt_inet.h" 67 #include "opt_inet6.h" 68 #include "opt_ipsec.h" 69 #include "opt_route.h" 70 #include "opt_rss.h" 71 72 #include <sys/param.h> 73 #include <sys/systm.h> 74 #include <sys/hhook.h> 75 #include <sys/malloc.h> 76 #include <sys/mbuf.h> 77 #include <sys/proc.h> 78 #include <sys/domain.h> 79 #include <sys/protosw.h> 80 #include <sys/sdt.h> 81 #include <sys/socket.h> 82 #include <sys/socketvar.h> 83 #include <sys/errno.h> 84 #include <sys/time.h> 85 #include <sys/kernel.h> 86 #include <sys/lock.h> 87 #include <sys/rmlock.h> 88 #include <sys/syslog.h> 89 90 #include <net/if.h> 91 #include <net/if_var.h> 92 #include <net/if_types.h> 93 #include <net/if_dl.h> 94 #include <net/route.h> 95 #include <net/netisr.h> 96 #include <net/rss_config.h> 97 #include <net/pfil.h> 98 #include <net/vnet.h> 99 100 #include <netinet/in.h> 101 #include <netinet/in_kdtrace.h> 102 #include <netinet/ip_var.h> 103 #include <netinet/in_systm.h> 104 #include <net/if_llatbl.h> 105 #ifdef INET 106 #include <netinet/ip.h> 107 #include <netinet/ip_icmp.h> 108 #endif /* INET */ 109 #include <netinet/ip6.h> 110 #include <netinet6/in6_var.h> 111 #include <netinet6/ip6_var.h> 112 #include <netinet/in_pcb.h> 113 #include <netinet/icmp6.h> 114 #include <netinet6/scope6_var.h> 115 #include <netinet6/in6_ifattach.h> 116 #include <netinet6/mld6_var.h> 117 #include <netinet6/nd6.h> 118 #include <netinet6/in6_rss.h> 119 120 #ifdef IPSEC 121 #include <netipsec/ipsec.h> 122 #include <netinet6/ip6_ipsec.h> 123 #include <netipsec/ipsec6.h> 124 #endif /* IPSEC */ 125 126 #include <netinet6/ip6protosw.h> 127 128 extern struct domain inet6domain; 129 130 u_char ip6_protox[IPPROTO_MAX]; 131 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead); 132 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl); 133 VNET_DEFINE(u_long, in6_ifaddrhmask); 134 135 static struct netisr_handler ip6_nh = { 136 .nh_name = "ip6", 137 .nh_handler = ip6_input, 138 .nh_proto = NETISR_IPV6, 139 #ifdef RSS 140 .nh_m2cpuid = rss_soft_m2cpuid_v6, 141 .nh_policy = NETISR_POLICY_CPU, 142 .nh_dispatch = NETISR_DISPATCH_HYBRID, 143 #else 144 .nh_policy = NETISR_POLICY_FLOW, 145 #endif 146 }; 147 148 #ifdef RSS 149 static struct netisr_handler ip6_direct_nh = { 150 .nh_name = "ip6_direct", 151 .nh_handler = ip6_direct_input, 152 .nh_proto = NETISR_IPV6_DIRECT, 153 .nh_m2cpuid = rss_soft_m2cpuid_v6, 154 .nh_policy = NETISR_POLICY_CPU, 155 .nh_dispatch = NETISR_DISPATCH_HYBRID, 156 }; 157 #endif 158 159 VNET_DEFINE(struct pfil_head, inet6_pfil_hook); 160 161 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat); 162 VNET_PCPUSTAT_SYSINIT(ip6stat); 163 #ifdef VIMAGE 164 VNET_PCPUSTAT_SYSUNINIT(ip6stat); 165 #endif /* VIMAGE */ 166 167 struct rmlock in6_ifaddr_lock; 168 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock"); 169 170 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *); 171 #ifdef PULLDOWN_TEST 172 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int); 173 #endif 174 175 /* 176 * IP6 initialization: fill in IP6 protocol switch table. 177 * All protocols not implemented in kernel go to raw IP6 protocol handler. 178 */ 179 void 180 ip6_init(void) 181 { 182 struct protosw *pr; 183 int i; 184 185 TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal", 186 &V_ip6_auto_linklocal); 187 TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv); 188 TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr); 189 190 TAILQ_INIT(&V_in6_ifaddrhead); 191 V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR, 192 &V_in6_ifaddrhmask); 193 194 /* Initialize packet filter hooks. */ 195 V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF; 196 V_inet6_pfil_hook.ph_af = AF_INET6; 197 if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0) 198 printf("%s: WARNING: unable to register pfil hook, " 199 "error %d\n", __func__, i); 200 201 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6, 202 &V_ipsec_hhh_in[HHOOK_IPSEC_INET6], 203 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 204 printf("%s: WARNING: unable to register input helper hook\n", 205 __func__); 206 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6, 207 &V_ipsec_hhh_out[HHOOK_IPSEC_INET6], 208 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 209 printf("%s: WARNING: unable to register output helper hook\n", 210 __func__); 211 212 scope6_init(); 213 addrsel_policy_init(); 214 nd6_init(); 215 frag6_init(); 216 217 V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR; 218 219 /* Skip global initialization stuff for non-default instances. */ 220 #ifdef VIMAGE 221 if (!IS_DEFAULT_VNET(curvnet)) { 222 netisr_register_vnet(&ip6_nh); 223 #ifdef RSS 224 netisr_register_vnet(&ip6_direct_nh); 225 #endif 226 return; 227 } 228 #endif 229 230 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 231 if (pr == NULL) 232 panic("ip6_init"); 233 234 /* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */ 235 for (i = 0; i < IPPROTO_MAX; i++) 236 ip6_protox[i] = pr - inet6sw; 237 /* 238 * Cycle through IP protocols and put them into the appropriate place 239 * in ip6_protox[]. 240 */ 241 for (pr = inet6domain.dom_protosw; 242 pr < inet6domain.dom_protoswNPROTOSW; pr++) 243 if (pr->pr_domain->dom_family == PF_INET6 && 244 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 245 /* Be careful to only index valid IP protocols. */ 246 if (pr->pr_protocol < IPPROTO_MAX) 247 ip6_protox[pr->pr_protocol] = pr - inet6sw; 248 } 249 250 netisr_register(&ip6_nh); 251 #ifdef RSS 252 netisr_register(&ip6_direct_nh); 253 #endif 254 } 255 256 /* 257 * The protocol to be inserted into ip6_protox[] must be already registered 258 * in inet6sw[], either statically or through pf_proto_register(). 259 */ 260 int 261 ip6proto_register(short ip6proto) 262 { 263 struct protosw *pr; 264 265 /* Sanity checks. */ 266 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 267 return (EPROTONOSUPPORT); 268 269 /* 270 * The protocol slot must not be occupied by another protocol 271 * already. An index pointing to IPPROTO_RAW is unused. 272 */ 273 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 274 if (pr == NULL) 275 return (EPFNOSUPPORT); 276 if (ip6_protox[ip6proto] != pr - inet6sw) /* IPPROTO_RAW */ 277 return (EEXIST); 278 279 /* 280 * Find the protocol position in inet6sw[] and set the index. 281 */ 282 for (pr = inet6domain.dom_protosw; 283 pr < inet6domain.dom_protoswNPROTOSW; pr++) { 284 if (pr->pr_domain->dom_family == PF_INET6 && 285 pr->pr_protocol && pr->pr_protocol == ip6proto) { 286 ip6_protox[pr->pr_protocol] = pr - inet6sw; 287 return (0); 288 } 289 } 290 return (EPROTONOSUPPORT); 291 } 292 293 int 294 ip6proto_unregister(short ip6proto) 295 { 296 struct protosw *pr; 297 298 /* Sanity checks. */ 299 if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX) 300 return (EPROTONOSUPPORT); 301 302 /* Check if the protocol was indeed registered. */ 303 pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW); 304 if (pr == NULL) 305 return (EPFNOSUPPORT); 306 if (ip6_protox[ip6proto] == pr - inet6sw) /* IPPROTO_RAW */ 307 return (ENOENT); 308 309 /* Reset the protocol slot to IPPROTO_RAW. */ 310 ip6_protox[ip6proto] = pr - inet6sw; 311 return (0); 312 } 313 314 #ifdef VIMAGE 315 static void 316 ip6_destroy(void *unused __unused) 317 { 318 struct ifaddr *ifa, *nifa; 319 struct ifnet *ifp; 320 int error; 321 322 #ifdef RSS 323 netisr_unregister_vnet(&ip6_direct_nh); 324 #endif 325 netisr_unregister_vnet(&ip6_nh); 326 327 if ((error = pfil_head_unregister(&V_inet6_pfil_hook)) != 0) 328 printf("%s: WARNING: unable to unregister pfil hook, " 329 "error %d\n", __func__, error); 330 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]); 331 if (error != 0) { 332 printf("%s: WARNING: unable to deregister input helper hook " 333 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: " 334 "error %d returned\n", __func__, error); 335 } 336 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]); 337 if (error != 0) { 338 printf("%s: WARNING: unable to deregister output helper hook " 339 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: " 340 "error %d returned\n", __func__, error); 341 } 342 343 /* Cleanup addresses. */ 344 IFNET_RLOCK(); 345 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 346 /* Cannot lock here - lock recursion. */ 347 /* IF_ADDR_LOCK(ifp); */ 348 TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 349 350 if (ifa->ifa_addr->sa_family != AF_INET6) 351 continue; 352 in6_purgeaddr(ifa); 353 } 354 /* IF_ADDR_UNLOCK(ifp); */ 355 in6_ifdetach_destroy(ifp); 356 mld_domifdetach(ifp); 357 /* Make sure any routes are gone as well. */ 358 rt_flushifroutes_af(ifp, AF_INET6); 359 } 360 IFNET_RUNLOCK(); 361 362 nd6_destroy(); 363 in6_ifattach_destroy(); 364 365 hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask); 366 } 367 368 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL); 369 #endif 370 371 static int 372 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off, 373 int *nxt, int *ours) 374 { 375 struct ip6_hdr *ip6; 376 struct ip6_hbh *hbh; 377 378 if (ip6_hopopts_input(plen, rtalert, &m, off)) { 379 #if 0 /*touches NULL pointer*/ 380 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 381 #endif 382 goto out; /* m have already been freed */ 383 } 384 385 /* adjust pointer */ 386 ip6 = mtod(m, struct ip6_hdr *); 387 388 /* 389 * if the payload length field is 0 and the next header field 390 * indicates Hop-by-Hop Options header, then a Jumbo Payload 391 * option MUST be included. 392 */ 393 if (ip6->ip6_plen == 0 && *plen == 0) { 394 /* 395 * Note that if a valid jumbo payload option is 396 * contained, ip6_hopopts_input() must set a valid 397 * (non-zero) payload length to the variable plen. 398 */ 399 IP6STAT_INC(ip6s_badoptions); 400 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 401 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 402 icmp6_error(m, ICMP6_PARAM_PROB, 403 ICMP6_PARAMPROB_HEADER, 404 (caddr_t)&ip6->ip6_plen - (caddr_t)ip6); 405 goto out; 406 } 407 #ifndef PULLDOWN_TEST 408 /* ip6_hopopts_input() ensures that mbuf is contiguous */ 409 hbh = (struct ip6_hbh *)(ip6 + 1); 410 #else 411 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 412 sizeof(struct ip6_hbh)); 413 if (hbh == NULL) { 414 IP6STAT_INC(ip6s_tooshort); 415 goto out; 416 } 417 #endif 418 *nxt = hbh->ip6h_nxt; 419 420 /* 421 * If we are acting as a router and the packet contains a 422 * router alert option, see if we know the option value. 423 * Currently, we only support the option value for MLD, in which 424 * case we should pass the packet to the multicast routing 425 * daemon. 426 */ 427 if (*rtalert != ~0) { 428 switch (*rtalert) { 429 case IP6OPT_RTALERT_MLD: 430 if (V_ip6_forwarding) 431 *ours = 1; 432 break; 433 default: 434 /* 435 * RFC2711 requires unrecognized values must be 436 * silently ignored. 437 */ 438 break; 439 } 440 } 441 442 return (0); 443 444 out: 445 return (1); 446 } 447 448 #ifdef RSS 449 /* 450 * IPv6 direct input routine. 451 * 452 * This is called when reinjecting completed fragments where 453 * all of the previous checking and book-keeping has been done. 454 */ 455 void 456 ip6_direct_input(struct mbuf *m) 457 { 458 int off, nxt; 459 int nest; 460 struct m_tag *mtag; 461 struct ip6_direct_ctx *ip6dc; 462 463 mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL); 464 KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!")); 465 466 ip6dc = (struct ip6_direct_ctx *)(mtag + 1); 467 nxt = ip6dc->ip6dc_nxt; 468 off = ip6dc->ip6dc_off; 469 470 nest = 0; 471 472 m_tag_delete(m, mtag); 473 474 while (nxt != IPPROTO_DONE) { 475 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 476 IP6STAT_INC(ip6s_toomanyhdr); 477 goto bad; 478 } 479 480 /* 481 * protection against faulty packet - there should be 482 * more sanity checks in header chain processing. 483 */ 484 if (m->m_pkthdr.len < off) { 485 IP6STAT_INC(ip6s_tooshort); 486 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 487 goto bad; 488 } 489 490 #ifdef IPSEC 491 /* 492 * enforce IPsec policy checking if we are seeing last header. 493 * note that we do not visit this with protocols with pcb layer 494 * code - like udp/tcp/raw ip. 495 */ 496 if (ip6_ipsec_input(m, nxt)) 497 goto bad; 498 #endif /* IPSEC */ 499 500 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 501 } 502 return; 503 bad: 504 m_freem(m); 505 } 506 #endif 507 508 void 509 ip6_input(struct mbuf *m) 510 { 511 struct in6_addr odst; 512 struct ip6_hdr *ip6; 513 struct in6_ifaddr *ia; 514 u_int32_t plen; 515 u_int32_t rtalert = ~0; 516 int off = sizeof(struct ip6_hdr), nest; 517 int nxt, ours = 0; 518 int srcrt = 0; 519 520 #ifdef IPSEC 521 /* 522 * should the inner packet be considered authentic? 523 * see comment in ah4_input(). 524 * NB: m cannot be NULL when passed to the input routine 525 */ 526 527 m->m_flags &= ~M_AUTHIPHDR; 528 m->m_flags &= ~M_AUTHIPDGM; 529 530 #endif /* IPSEC */ 531 532 if (m->m_flags & M_FASTFWD_OURS) { 533 /* 534 * Firewall changed destination to local. 535 */ 536 m->m_flags &= ~M_FASTFWD_OURS; 537 ours = 1; 538 ip6 = mtod(m, struct ip6_hdr *); 539 goto hbhcheck; 540 } 541 542 /* 543 * mbuf statistics 544 */ 545 if (m->m_flags & M_EXT) { 546 if (m->m_next) 547 IP6STAT_INC(ip6s_mext2m); 548 else 549 IP6STAT_INC(ip6s_mext1); 550 } else { 551 if (m->m_next) { 552 if (m->m_flags & M_LOOP) { 553 IP6STAT_INC(ip6s_m2m[V_loif->if_index]); 554 } else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX) 555 IP6STAT_INC( 556 ip6s_m2m[m->m_pkthdr.rcvif->if_index]); 557 else 558 IP6STAT_INC(ip6s_m2m[0]); 559 } else 560 IP6STAT_INC(ip6s_m1); 561 } 562 563 /* drop the packet if IPv6 operation is disabled on the IF */ 564 if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) 565 goto bad; 566 567 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive); 568 IP6STAT_INC(ip6s_total); 569 570 #ifndef PULLDOWN_TEST 571 /* 572 * L2 bridge code and some other code can return mbuf chain 573 * that does not conform to KAME requirement. too bad. 574 * XXX: fails to join if interface MTU > MCLBYTES. jumbogram? 575 */ 576 if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) { 577 struct mbuf *n; 578 579 if (m->m_pkthdr.len > MHLEN) 580 n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 581 else 582 n = m_gethdr(M_NOWAIT, MT_DATA); 583 if (n == NULL) { 584 m_freem(m); 585 return; /* ENOBUFS */ 586 } 587 588 m_move_pkthdr(n, m); 589 m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t)); 590 n->m_len = n->m_pkthdr.len; 591 m_freem(m); 592 m = n; 593 } 594 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */); 595 #endif 596 597 if (m->m_len < sizeof(struct ip6_hdr)) { 598 struct ifnet *inifp; 599 inifp = m->m_pkthdr.rcvif; 600 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { 601 IP6STAT_INC(ip6s_toosmall); 602 in6_ifstat_inc(inifp, ifs6_in_hdrerr); 603 return; 604 } 605 } 606 607 ip6 = mtod(m, struct ip6_hdr *); 608 609 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 610 IP6STAT_INC(ip6s_badvers); 611 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr); 612 goto bad; 613 } 614 615 IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]); 616 617 IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6); 618 619 /* 620 * Check against address spoofing/corruption. 621 */ 622 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) || 623 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) { 624 /* 625 * XXX: "badscope" is not very suitable for a multicast source. 626 */ 627 IP6STAT_INC(ip6s_badscope); 628 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 629 goto bad; 630 } 631 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) && 632 !(m->m_flags & M_LOOP)) { 633 /* 634 * In this case, the packet should come from the loopback 635 * interface. However, we cannot just check the if_flags, 636 * because ip6_mloopback() passes the "actual" interface 637 * as the outgoing/incoming interface. 638 */ 639 IP6STAT_INC(ip6s_badscope); 640 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 641 goto bad; 642 } 643 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && 644 IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) { 645 /* 646 * RFC4291 2.7: 647 * Nodes must not originate a packet to a multicast address 648 * whose scop field contains the reserved value 0; if such 649 * a packet is received, it must be silently dropped. 650 */ 651 IP6STAT_INC(ip6s_badscope); 652 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 653 goto bad; 654 } 655 #ifdef ALTQ 656 if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) { 657 /* packet is dropped by traffic conditioner */ 658 return; 659 } 660 #endif 661 /* 662 * The following check is not documented in specs. A malicious 663 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack 664 * and bypass security checks (act as if it was from 127.0.0.1 by using 665 * IPv6 src ::ffff:127.0.0.1). Be cautious. 666 * 667 * This check chokes if we are in an SIIT cloud. As none of BSDs 668 * support IPv4-less kernel compilation, we cannot support SIIT 669 * environment at all. So, it makes more sense for us to reject any 670 * malicious packets for non-SIIT environment, than try to do a 671 * partial support for SIIT environment. 672 */ 673 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 674 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 675 IP6STAT_INC(ip6s_badscope); 676 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 677 goto bad; 678 } 679 #if 0 680 /* 681 * Reject packets with IPv4 compatible addresses (auto tunnel). 682 * 683 * The code forbids auto tunnel relay case in RFC1933 (the check is 684 * stronger than RFC1933). We may want to re-enable it if mech-xx 685 * is revised to forbid relaying case. 686 */ 687 if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) || 688 IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) { 689 IP6STAT_INC(ip6s_badscope); 690 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 691 goto bad; 692 } 693 #endif 694 #ifdef IPSEC 695 /* 696 * Bypass packet filtering for packets previously handled by IPsec. 697 */ 698 if (ip6_ipsec_filtertunnel(m)) 699 goto passin; 700 #endif /* IPSEC */ 701 702 /* 703 * Run through list of hooks for input packets. 704 * 705 * NB: Beware of the destination address changing 706 * (e.g. by NAT rewriting). When this happens, 707 * tell ip6_forward to do the right thing. 708 */ 709 odst = ip6->ip6_dst; 710 711 /* Jump over all PFIL processing if hooks are not active. */ 712 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 713 goto passin; 714 715 if (pfil_run_hooks(&V_inet6_pfil_hook, &m, 716 m->m_pkthdr.rcvif, PFIL_IN, NULL)) 717 return; 718 if (m == NULL) /* consumed by filter */ 719 return; 720 ip6 = mtod(m, struct ip6_hdr *); 721 srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst); 722 723 if (m->m_flags & M_FASTFWD_OURS) { 724 m->m_flags &= ~M_FASTFWD_OURS; 725 ours = 1; 726 goto hbhcheck; 727 } 728 if ((m->m_flags & M_IP6_NEXTHOP) && 729 m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 730 /* 731 * Directly ship the packet on. This allows forwarding 732 * packets originally destined to us to some other directly 733 * connected host. 734 */ 735 ip6_forward(m, 1); 736 return; 737 } 738 739 passin: 740 /* 741 * Disambiguate address scope zones (if there is ambiguity). 742 * We first make sure that the original source or destination address 743 * is not in our internal form for scoped addresses. Such addresses 744 * are not necessarily invalid spec-wise, but we cannot accept them due 745 * to the usage conflict. 746 * in6_setscope() then also checks and rejects the cases where src or 747 * dst are the loopback address and the receiving interface 748 * is not loopback. 749 */ 750 if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) { 751 IP6STAT_INC(ip6s_badscope); /* XXX */ 752 goto bad; 753 } 754 if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) || 755 in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) { 756 IP6STAT_INC(ip6s_badscope); 757 goto bad; 758 } 759 /* 760 * Multicast check. Assume packet is for us to avoid 761 * prematurely taking locks. 762 */ 763 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 764 ours = 1; 765 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast); 766 goto hbhcheck; 767 } 768 /* 769 * Unicast check 770 * XXX: For now we keep link-local IPv6 addresses with embedded 771 * scope zone id, therefore we use zero zoneid here. 772 */ 773 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */); 774 if (ia != NULL) { 775 if (ia->ia6_flags & IN6_IFF_NOTREADY) { 776 char ip6bufs[INET6_ADDRSTRLEN]; 777 char ip6bufd[INET6_ADDRSTRLEN]; 778 /* address is not ready, so discard the packet. */ 779 nd6log((LOG_INFO, 780 "ip6_input: packet to an unready address %s->%s\n", 781 ip6_sprintf(ip6bufs, &ip6->ip6_src), 782 ip6_sprintf(ip6bufd, &ip6->ip6_dst))); 783 ifa_free(&ia->ia_ifa); 784 goto bad; 785 } 786 /* Count the packet in the ip address stats */ 787 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 788 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 789 ifa_free(&ia->ia_ifa); 790 ours = 1; 791 goto hbhcheck; 792 } 793 794 /* 795 * Now there is no reason to process the packet if it's not our own 796 * and we're not a router. 797 */ 798 if (!V_ip6_forwarding) { 799 IP6STAT_INC(ip6s_cantforward); 800 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 801 goto bad; 802 } 803 804 hbhcheck: 805 /* 806 * Process Hop-by-Hop options header if it's contained. 807 * m may be modified in ip6_hopopts_input(). 808 * If a JumboPayload option is included, plen will also be modified. 809 */ 810 plen = (u_int32_t)ntohs(ip6->ip6_plen); 811 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 812 if (ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours) != 0) 813 return; 814 } else 815 nxt = ip6->ip6_nxt; 816 817 /* 818 * Use mbuf flags to propagate Router Alert option to 819 * ICMPv6 layer, as hop-by-hop options have been stripped. 820 */ 821 if (rtalert != ~0) 822 m->m_flags |= M_RTALERT_MLD; 823 824 /* 825 * Check that the amount of data in the buffers 826 * is as at least much as the IPv6 header would have us expect. 827 * Trim mbufs if longer than we expect. 828 * Drop packet if shorter than we expect. 829 */ 830 if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) { 831 IP6STAT_INC(ip6s_tooshort); 832 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 833 goto bad; 834 } 835 if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) { 836 if (m->m_len == m->m_pkthdr.len) { 837 m->m_len = sizeof(struct ip6_hdr) + plen; 838 m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen; 839 } else 840 m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len); 841 } 842 843 /* 844 * Forward if desirable. 845 */ 846 if (V_ip6_mrouter && 847 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 848 /* 849 * If we are acting as a multicast router, all 850 * incoming multicast packets are passed to the 851 * kernel-level multicast forwarding function. 852 * The packet is returned (relatively) intact; if 853 * ip6_mforward() returns a non-zero value, the packet 854 * must be discarded, else it may be accepted below. 855 * 856 * XXX TODO: Check hlim and multicast scope here to avoid 857 * unnecessarily calling into ip6_mforward(). 858 */ 859 if (ip6_mforward && 860 ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) { 861 IP6STAT_INC(ip6s_cantforward); 862 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 863 goto bad; 864 } 865 } else if (!ours) { 866 ip6_forward(m, srcrt); 867 return; 868 } 869 870 ip6 = mtod(m, struct ip6_hdr *); 871 872 /* 873 * Malicious party may be able to use IPv4 mapped addr to confuse 874 * tcp/udp stack and bypass security checks (act as if it was from 875 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1). Be cautious. 876 * 877 * For SIIT end node behavior, you may want to disable the check. 878 * However, you will become vulnerable to attacks using IPv4 mapped 879 * source. 880 */ 881 if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || 882 IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { 883 IP6STAT_INC(ip6s_badscope); 884 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr); 885 goto bad; 886 } 887 888 /* 889 * Tell launch routine the next header 890 */ 891 IP6STAT_INC(ip6s_delivered); 892 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_deliver); 893 nest = 0; 894 895 while (nxt != IPPROTO_DONE) { 896 if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) { 897 IP6STAT_INC(ip6s_toomanyhdr); 898 goto bad; 899 } 900 901 /* 902 * protection against faulty packet - there should be 903 * more sanity checks in header chain processing. 904 */ 905 if (m->m_pkthdr.len < off) { 906 IP6STAT_INC(ip6s_tooshort); 907 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated); 908 goto bad; 909 } 910 911 #ifdef IPSEC 912 /* 913 * enforce IPsec policy checking if we are seeing last header. 914 * note that we do not visit this with protocols with pcb layer 915 * code - like udp/tcp/raw ip. 916 */ 917 if (ip6_ipsec_input(m, nxt)) 918 goto bad; 919 #endif /* IPSEC */ 920 921 nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt); 922 } 923 return; 924 bad: 925 m_freem(m); 926 } 927 928 /* 929 * Hop-by-Hop options header processing. If a valid jumbo payload option is 930 * included, the real payload length will be stored in plenp. 931 * 932 * rtalertp - XXX: should be stored more smart way 933 */ 934 static int 935 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, 936 struct mbuf **mp, int *offp) 937 { 938 struct mbuf *m = *mp; 939 int off = *offp, hbhlen; 940 struct ip6_hbh *hbh; 941 942 /* validation of the length of the header */ 943 #ifndef PULLDOWN_TEST 944 IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1); 945 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 946 hbhlen = (hbh->ip6h_len + 1) << 3; 947 948 IP6_EXTHDR_CHECK(m, off, hbhlen, -1); 949 hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off); 950 #else 951 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, 952 sizeof(struct ip6_hdr), sizeof(struct ip6_hbh)); 953 if (hbh == NULL) { 954 IP6STAT_INC(ip6s_tooshort); 955 return -1; 956 } 957 hbhlen = (hbh->ip6h_len + 1) << 3; 958 IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr), 959 hbhlen); 960 if (hbh == NULL) { 961 IP6STAT_INC(ip6s_tooshort); 962 return -1; 963 } 964 #endif 965 off += hbhlen; 966 hbhlen -= sizeof(struct ip6_hbh); 967 if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh), 968 hbhlen, rtalertp, plenp) < 0) 969 return (-1); 970 971 *offp = off; 972 *mp = m; 973 return (0); 974 } 975 976 /* 977 * Search header for all Hop-by-hop options and process each option. 978 * This function is separate from ip6_hopopts_input() in order to 979 * handle a case where the sending node itself process its hop-by-hop 980 * options header. In such a case, the function is called from ip6_output(). 981 * 982 * The function assumes that hbh header is located right after the IPv6 header 983 * (RFC2460 p7), opthead is pointer into data content in m, and opthead to 984 * opthead + hbhlen is located in contiguous memory region. 985 */ 986 int 987 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen, 988 u_int32_t *rtalertp, u_int32_t *plenp) 989 { 990 struct ip6_hdr *ip6; 991 int optlen = 0; 992 u_int8_t *opt = opthead; 993 u_int16_t rtalert_val; 994 u_int32_t jumboplen; 995 const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh); 996 997 for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) { 998 switch (*opt) { 999 case IP6OPT_PAD1: 1000 optlen = 1; 1001 break; 1002 case IP6OPT_PADN: 1003 if (hbhlen < IP6OPT_MINLEN) { 1004 IP6STAT_INC(ip6s_toosmall); 1005 goto bad; 1006 } 1007 optlen = *(opt + 1) + 2; 1008 break; 1009 case IP6OPT_ROUTER_ALERT: 1010 /* XXX may need check for alignment */ 1011 if (hbhlen < IP6OPT_RTALERT_LEN) { 1012 IP6STAT_INC(ip6s_toosmall); 1013 goto bad; 1014 } 1015 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) { 1016 /* XXX stat */ 1017 icmp6_error(m, ICMP6_PARAM_PROB, 1018 ICMP6_PARAMPROB_HEADER, 1019 erroff + opt + 1 - opthead); 1020 return (-1); 1021 } 1022 optlen = IP6OPT_RTALERT_LEN; 1023 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2); 1024 *rtalertp = ntohs(rtalert_val); 1025 break; 1026 case IP6OPT_JUMBO: 1027 /* XXX may need check for alignment */ 1028 if (hbhlen < IP6OPT_JUMBO_LEN) { 1029 IP6STAT_INC(ip6s_toosmall); 1030 goto bad; 1031 } 1032 if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) { 1033 /* XXX stat */ 1034 icmp6_error(m, ICMP6_PARAM_PROB, 1035 ICMP6_PARAMPROB_HEADER, 1036 erroff + opt + 1 - opthead); 1037 return (-1); 1038 } 1039 optlen = IP6OPT_JUMBO_LEN; 1040 1041 /* 1042 * IPv6 packets that have non 0 payload length 1043 * must not contain a jumbo payload option. 1044 */ 1045 ip6 = mtod(m, struct ip6_hdr *); 1046 if (ip6->ip6_plen) { 1047 IP6STAT_INC(ip6s_badoptions); 1048 icmp6_error(m, ICMP6_PARAM_PROB, 1049 ICMP6_PARAMPROB_HEADER, 1050 erroff + opt - opthead); 1051 return (-1); 1052 } 1053 1054 /* 1055 * We may see jumbolen in unaligned location, so 1056 * we'd need to perform bcopy(). 1057 */ 1058 bcopy(opt + 2, &jumboplen, sizeof(jumboplen)); 1059 jumboplen = (u_int32_t)htonl(jumboplen); 1060 1061 #if 1 1062 /* 1063 * if there are multiple jumbo payload options, 1064 * *plenp will be non-zero and the packet will be 1065 * rejected. 1066 * the behavior may need some debate in ipngwg - 1067 * multiple options does not make sense, however, 1068 * there's no explicit mention in specification. 1069 */ 1070 if (*plenp != 0) { 1071 IP6STAT_INC(ip6s_badoptions); 1072 icmp6_error(m, ICMP6_PARAM_PROB, 1073 ICMP6_PARAMPROB_HEADER, 1074 erroff + opt + 2 - opthead); 1075 return (-1); 1076 } 1077 #endif 1078 1079 /* 1080 * jumbo payload length must be larger than 65535. 1081 */ 1082 if (jumboplen <= IPV6_MAXPACKET) { 1083 IP6STAT_INC(ip6s_badoptions); 1084 icmp6_error(m, ICMP6_PARAM_PROB, 1085 ICMP6_PARAMPROB_HEADER, 1086 erroff + opt + 2 - opthead); 1087 return (-1); 1088 } 1089 *plenp = jumboplen; 1090 1091 break; 1092 default: /* unknown option */ 1093 if (hbhlen < IP6OPT_MINLEN) { 1094 IP6STAT_INC(ip6s_toosmall); 1095 goto bad; 1096 } 1097 optlen = ip6_unknown_opt(opt, m, 1098 erroff + opt - opthead); 1099 if (optlen == -1) 1100 return (-1); 1101 optlen += 2; 1102 break; 1103 } 1104 } 1105 1106 return (0); 1107 1108 bad: 1109 m_freem(m); 1110 return (-1); 1111 } 1112 1113 /* 1114 * Unknown option processing. 1115 * The third argument `off' is the offset from the IPv6 header to the option, 1116 * which is necessary if the IPv6 header the and option header and IPv6 header 1117 * is not contiguous in order to return an ICMPv6 error. 1118 */ 1119 int 1120 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off) 1121 { 1122 struct ip6_hdr *ip6; 1123 1124 switch (IP6OPT_TYPE(*optp)) { 1125 case IP6OPT_TYPE_SKIP: /* ignore the option */ 1126 return ((int)*(optp + 1)); 1127 case IP6OPT_TYPE_DISCARD: /* silently discard */ 1128 m_freem(m); 1129 return (-1); 1130 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */ 1131 IP6STAT_INC(ip6s_badoptions); 1132 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off); 1133 return (-1); 1134 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */ 1135 IP6STAT_INC(ip6s_badoptions); 1136 ip6 = mtod(m, struct ip6_hdr *); 1137 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 1138 (m->m_flags & (M_BCAST|M_MCAST))) 1139 m_freem(m); 1140 else 1141 icmp6_error(m, ICMP6_PARAM_PROB, 1142 ICMP6_PARAMPROB_OPTION, off); 1143 return (-1); 1144 } 1145 1146 m_freem(m); /* XXX: NOTREACHED */ 1147 return (-1); 1148 } 1149 1150 /* 1151 * Create the "control" list for this pcb. 1152 * These functions will not modify mbuf chain at all. 1153 * 1154 * With KAME mbuf chain restriction: 1155 * The routine will be called from upper layer handlers like tcp6_input(). 1156 * Thus the routine assumes that the caller (tcp6_input) have already 1157 * called IP6_EXTHDR_CHECK() and all the extension headers are located in the 1158 * very first mbuf on the mbuf chain. 1159 * 1160 * ip6_savecontrol_v4 will handle those options that are possible to be 1161 * set on a v4-mapped socket. 1162 * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those 1163 * options and handle the v6-only ones itself. 1164 */ 1165 struct mbuf ** 1166 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp, 1167 int *v4only) 1168 { 1169 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1170 1171 #ifdef SO_TIMESTAMP 1172 if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) { 1173 struct timeval tv; 1174 1175 microtime(&tv); 1176 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1177 SCM_TIMESTAMP, SOL_SOCKET); 1178 if (*mp) 1179 mp = &(*mp)->m_next; 1180 } 1181 #endif 1182 1183 #define IS2292(inp, x, y) (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y)) 1184 /* RFC 2292 sec. 5 */ 1185 if ((inp->inp_flags & IN6P_PKTINFO) != 0) { 1186 struct in6_pktinfo pi6; 1187 1188 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1189 #ifdef INET 1190 struct ip *ip; 1191 1192 ip = mtod(m, struct ip *); 1193 pi6.ipi6_addr.s6_addr32[0] = 0; 1194 pi6.ipi6_addr.s6_addr32[1] = 0; 1195 pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP; 1196 pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr; 1197 #else 1198 /* We won't hit this code */ 1199 bzero(&pi6.ipi6_addr, sizeof(struct in6_addr)); 1200 #endif 1201 } else { 1202 bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr)); 1203 in6_clearscope(&pi6.ipi6_addr); /* XXX */ 1204 } 1205 pi6.ipi6_ifindex = 1206 (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0; 1207 1208 *mp = sbcreatecontrol((caddr_t) &pi6, 1209 sizeof(struct in6_pktinfo), 1210 IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6); 1211 if (*mp) 1212 mp = &(*mp)->m_next; 1213 } 1214 1215 if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) { 1216 int hlim; 1217 1218 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1219 #ifdef INET 1220 struct ip *ip; 1221 1222 ip = mtod(m, struct ip *); 1223 hlim = ip->ip_ttl; 1224 #else 1225 /* We won't hit this code */ 1226 hlim = 0; 1227 #endif 1228 } else { 1229 hlim = ip6->ip6_hlim & 0xff; 1230 } 1231 *mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int), 1232 IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), 1233 IPPROTO_IPV6); 1234 if (*mp) 1235 mp = &(*mp)->m_next; 1236 } 1237 1238 if ((inp->inp_flags & IN6P_TCLASS) != 0) { 1239 int tclass; 1240 1241 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1242 #ifdef INET 1243 struct ip *ip; 1244 1245 ip = mtod(m, struct ip *); 1246 tclass = ip->ip_tos; 1247 #else 1248 /* We won't hit this code */ 1249 tclass = 0; 1250 #endif 1251 } else { 1252 u_int32_t flowinfo; 1253 1254 flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK); 1255 flowinfo >>= 20; 1256 tclass = flowinfo & 0xff; 1257 } 1258 *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int), 1259 IPV6_TCLASS, IPPROTO_IPV6); 1260 if (*mp) 1261 mp = &(*mp)->m_next; 1262 } 1263 1264 if (v4only != NULL) { 1265 if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) { 1266 *v4only = 1; 1267 } else { 1268 *v4only = 0; 1269 } 1270 } 1271 1272 return (mp); 1273 } 1274 1275 void 1276 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp) 1277 { 1278 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1279 int v4only = 0; 1280 1281 mp = ip6_savecontrol_v4(in6p, m, mp, &v4only); 1282 if (v4only) 1283 return; 1284 1285 /* 1286 * IPV6_HOPOPTS socket option. Recall that we required super-user 1287 * privilege for the option (see ip6_ctloutput), but it might be too 1288 * strict, since there might be some hop-by-hop options which can be 1289 * returned to normal user. 1290 * See also RFC 2292 section 6 (or RFC 3542 section 8). 1291 */ 1292 if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) { 1293 /* 1294 * Check if a hop-by-hop options header is contatined in the 1295 * received packet, and if so, store the options as ancillary 1296 * data. Note that a hop-by-hop options header must be 1297 * just after the IPv6 header, which is assured through the 1298 * IPv6 input processing. 1299 */ 1300 if (ip6->ip6_nxt == IPPROTO_HOPOPTS) { 1301 struct ip6_hbh *hbh; 1302 int hbhlen = 0; 1303 #ifdef PULLDOWN_TEST 1304 struct mbuf *ext; 1305 #endif 1306 1307 #ifndef PULLDOWN_TEST 1308 hbh = (struct ip6_hbh *)(ip6 + 1); 1309 hbhlen = (hbh->ip6h_len + 1) << 3; 1310 #else 1311 ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr), 1312 ip6->ip6_nxt); 1313 if (ext == NULL) { 1314 IP6STAT_INC(ip6s_tooshort); 1315 return; 1316 } 1317 hbh = mtod(ext, struct ip6_hbh *); 1318 hbhlen = (hbh->ip6h_len + 1) << 3; 1319 if (hbhlen != ext->m_len) { 1320 m_freem(ext); 1321 IP6STAT_INC(ip6s_tooshort); 1322 return; 1323 } 1324 #endif 1325 1326 /* 1327 * XXX: We copy the whole header even if a 1328 * jumbo payload option is included, the option which 1329 * is to be removed before returning according to 1330 * RFC2292. 1331 * Note: this constraint is removed in RFC3542 1332 */ 1333 *mp = sbcreatecontrol((caddr_t)hbh, hbhlen, 1334 IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS), 1335 IPPROTO_IPV6); 1336 if (*mp) 1337 mp = &(*mp)->m_next; 1338 #ifdef PULLDOWN_TEST 1339 m_freem(ext); 1340 #endif 1341 } 1342 } 1343 1344 if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) { 1345 int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr); 1346 1347 /* 1348 * Search for destination options headers or routing 1349 * header(s) through the header chain, and stores each 1350 * header as ancillary data. 1351 * Note that the order of the headers remains in 1352 * the chain of ancillary data. 1353 */ 1354 while (1) { /* is explicit loop prevention necessary? */ 1355 struct ip6_ext *ip6e = NULL; 1356 int elen; 1357 #ifdef PULLDOWN_TEST 1358 struct mbuf *ext = NULL; 1359 #endif 1360 1361 /* 1362 * if it is not an extension header, don't try to 1363 * pull it from the chain. 1364 */ 1365 switch (nxt) { 1366 case IPPROTO_DSTOPTS: 1367 case IPPROTO_ROUTING: 1368 case IPPROTO_HOPOPTS: 1369 case IPPROTO_AH: /* is it possible? */ 1370 break; 1371 default: 1372 goto loopend; 1373 } 1374 1375 #ifndef PULLDOWN_TEST 1376 if (off + sizeof(*ip6e) > m->m_len) 1377 goto loopend; 1378 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off); 1379 if (nxt == IPPROTO_AH) 1380 elen = (ip6e->ip6e_len + 2) << 2; 1381 else 1382 elen = (ip6e->ip6e_len + 1) << 3; 1383 if (off + elen > m->m_len) 1384 goto loopend; 1385 #else 1386 ext = ip6_pullexthdr(m, off, nxt); 1387 if (ext == NULL) { 1388 IP6STAT_INC(ip6s_tooshort); 1389 return; 1390 } 1391 ip6e = mtod(ext, struct ip6_ext *); 1392 if (nxt == IPPROTO_AH) 1393 elen = (ip6e->ip6e_len + 2) << 2; 1394 else 1395 elen = (ip6e->ip6e_len + 1) << 3; 1396 if (elen != ext->m_len) { 1397 m_freem(ext); 1398 IP6STAT_INC(ip6s_tooshort); 1399 return; 1400 } 1401 #endif 1402 1403 switch (nxt) { 1404 case IPPROTO_DSTOPTS: 1405 if (!(in6p->inp_flags & IN6P_DSTOPTS)) 1406 break; 1407 1408 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1409 IS2292(in6p, 1410 IPV6_2292DSTOPTS, IPV6_DSTOPTS), 1411 IPPROTO_IPV6); 1412 if (*mp) 1413 mp = &(*mp)->m_next; 1414 break; 1415 case IPPROTO_ROUTING: 1416 if (!(in6p->inp_flags & IN6P_RTHDR)) 1417 break; 1418 1419 *mp = sbcreatecontrol((caddr_t)ip6e, elen, 1420 IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR), 1421 IPPROTO_IPV6); 1422 if (*mp) 1423 mp = &(*mp)->m_next; 1424 break; 1425 case IPPROTO_HOPOPTS: 1426 case IPPROTO_AH: /* is it possible? */ 1427 break; 1428 1429 default: 1430 /* 1431 * other cases have been filtered in the above. 1432 * none will visit this case. here we supply 1433 * the code just in case (nxt overwritten or 1434 * other cases). 1435 */ 1436 #ifdef PULLDOWN_TEST 1437 m_freem(ext); 1438 #endif 1439 goto loopend; 1440 1441 } 1442 1443 /* proceed with the next header. */ 1444 off += elen; 1445 nxt = ip6e->ip6e_nxt; 1446 ip6e = NULL; 1447 #ifdef PULLDOWN_TEST 1448 m_freem(ext); 1449 ext = NULL; 1450 #endif 1451 } 1452 loopend: 1453 ; 1454 } 1455 1456 if (in6p->inp_flags2 & INP_RECVFLOWID) { 1457 uint32_t flowid, flow_type; 1458 1459 flowid = m->m_pkthdr.flowid; 1460 flow_type = M_HASHTYPE_GET(m); 1461 1462 /* 1463 * XXX should handle the failure of one or the 1464 * other - don't populate both? 1465 */ 1466 *mp = sbcreatecontrol((caddr_t) &flowid, 1467 sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6); 1468 if (*mp) 1469 mp = &(*mp)->m_next; 1470 *mp = sbcreatecontrol((caddr_t) &flow_type, 1471 sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6); 1472 if (*mp) 1473 mp = &(*mp)->m_next; 1474 } 1475 1476 #ifdef RSS 1477 if (in6p->inp_flags2 & INP_RECVRSSBUCKETID) { 1478 uint32_t flowid, flow_type; 1479 uint32_t rss_bucketid; 1480 1481 flowid = m->m_pkthdr.flowid; 1482 flow_type = M_HASHTYPE_GET(m); 1483 1484 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1485 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1486 sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6); 1487 if (*mp) 1488 mp = &(*mp)->m_next; 1489 } 1490 } 1491 #endif 1492 1493 } 1494 #undef IS2292 1495 1496 void 1497 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu) 1498 { 1499 struct socket *so; 1500 struct mbuf *m_mtu; 1501 struct ip6_mtuinfo mtuctl; 1502 1503 KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); 1504 /* 1505 * Notify the error by sending IPV6_PATHMTU ancillary data if 1506 * application wanted to know the MTU value. 1507 * NOTE: we notify disconnected sockets, because some udp 1508 * applications keep sending sockets disconnected. 1509 * NOTE: our implementation doesn't notify connected sockets that has 1510 * foreign address that is different than given destination addresses 1511 * (this is permitted by RFC 3542). 1512 */ 1513 if ((inp->inp_flags & IN6P_MTU) == 0 || ( 1514 !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1515 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr))) 1516 return; 1517 1518 mtuctl.ip6m_mtu = mtu; 1519 mtuctl.ip6m_addr = *dst; 1520 if (sa6_recoverscope(&mtuctl.ip6m_addr)) 1521 return; 1522 1523 if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl), 1524 IPV6_PATHMTU, IPPROTO_IPV6)) == NULL) 1525 return; 1526 1527 so = inp->inp_socket; 1528 if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu) 1529 == 0) { 1530 m_freem(m_mtu); 1531 /* XXX: should count statistics */ 1532 } else 1533 sorwakeup(so); 1534 } 1535 1536 #ifdef PULLDOWN_TEST 1537 /* 1538 * pull single extension header from mbuf chain. returns single mbuf that 1539 * contains the result, or NULL on error. 1540 */ 1541 static struct mbuf * 1542 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt) 1543 { 1544 struct ip6_ext ip6e; 1545 size_t elen; 1546 struct mbuf *n; 1547 1548 #ifdef DIAGNOSTIC 1549 switch (nxt) { 1550 case IPPROTO_DSTOPTS: 1551 case IPPROTO_ROUTING: 1552 case IPPROTO_HOPOPTS: 1553 case IPPROTO_AH: /* is it possible? */ 1554 break; 1555 default: 1556 printf("ip6_pullexthdr: invalid nxt=%d\n", nxt); 1557 } 1558 #endif 1559 1560 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1561 if (nxt == IPPROTO_AH) 1562 elen = (ip6e.ip6e_len + 2) << 2; 1563 else 1564 elen = (ip6e.ip6e_len + 1) << 3; 1565 1566 if (elen > MLEN) 1567 n = m_getcl(M_NOWAIT, MT_DATA, 0); 1568 else 1569 n = m_get(M_NOWAIT, MT_DATA); 1570 if (n == NULL) 1571 return NULL; 1572 1573 m_copydata(m, off, elen, mtod(n, caddr_t)); 1574 n->m_len = elen; 1575 return n; 1576 } 1577 #endif 1578 1579 /* 1580 * Get pointer to the previous header followed by the header 1581 * currently processed. 1582 * XXX: This function supposes that 1583 * M includes all headers, 1584 * the next header field and the header length field of each header 1585 * are valid, and 1586 * the sum of each header length equals to OFF. 1587 * Because of these assumptions, this function must be called very 1588 * carefully. Moreover, it will not be used in the near future when 1589 * we develop `neater' mechanism to process extension headers. 1590 */ 1591 char * 1592 ip6_get_prevhdr(const struct mbuf *m, int off) 1593 { 1594 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1595 1596 if (off == sizeof(struct ip6_hdr)) 1597 return (&ip6->ip6_nxt); 1598 else { 1599 int len, nxt; 1600 struct ip6_ext *ip6e = NULL; 1601 1602 nxt = ip6->ip6_nxt; 1603 len = sizeof(struct ip6_hdr); 1604 while (len < off) { 1605 ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len); 1606 1607 switch (nxt) { 1608 case IPPROTO_FRAGMENT: 1609 len += sizeof(struct ip6_frag); 1610 break; 1611 case IPPROTO_AH: 1612 len += (ip6e->ip6e_len + 2) << 2; 1613 break; 1614 default: 1615 len += (ip6e->ip6e_len + 1) << 3; 1616 break; 1617 } 1618 nxt = ip6e->ip6e_nxt; 1619 } 1620 if (ip6e) 1621 return (&ip6e->ip6e_nxt); 1622 else 1623 return NULL; 1624 } 1625 } 1626 1627 /* 1628 * get next header offset. m will be retained. 1629 */ 1630 int 1631 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1632 { 1633 struct ip6_hdr ip6; 1634 struct ip6_ext ip6e; 1635 struct ip6_frag fh; 1636 1637 /* just in case */ 1638 if (m == NULL) 1639 panic("ip6_nexthdr: m == NULL"); 1640 if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off) 1641 return -1; 1642 1643 switch (proto) { 1644 case IPPROTO_IPV6: 1645 if (m->m_pkthdr.len < off + sizeof(ip6)) 1646 return -1; 1647 m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6); 1648 if (nxtp) 1649 *nxtp = ip6.ip6_nxt; 1650 off += sizeof(ip6); 1651 return off; 1652 1653 case IPPROTO_FRAGMENT: 1654 /* 1655 * terminate parsing if it is not the first fragment, 1656 * it does not make sense to parse through it. 1657 */ 1658 if (m->m_pkthdr.len < off + sizeof(fh)) 1659 return -1; 1660 m_copydata(m, off, sizeof(fh), (caddr_t)&fh); 1661 /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */ 1662 if (fh.ip6f_offlg & IP6F_OFF_MASK) 1663 return -1; 1664 if (nxtp) 1665 *nxtp = fh.ip6f_nxt; 1666 off += sizeof(struct ip6_frag); 1667 return off; 1668 1669 case IPPROTO_AH: 1670 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1671 return -1; 1672 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1673 if (nxtp) 1674 *nxtp = ip6e.ip6e_nxt; 1675 off += (ip6e.ip6e_len + 2) << 2; 1676 return off; 1677 1678 case IPPROTO_HOPOPTS: 1679 case IPPROTO_ROUTING: 1680 case IPPROTO_DSTOPTS: 1681 if (m->m_pkthdr.len < off + sizeof(ip6e)) 1682 return -1; 1683 m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e); 1684 if (nxtp) 1685 *nxtp = ip6e.ip6e_nxt; 1686 off += (ip6e.ip6e_len + 1) << 3; 1687 return off; 1688 1689 case IPPROTO_NONE: 1690 case IPPROTO_ESP: 1691 case IPPROTO_IPCOMP: 1692 /* give up */ 1693 return -1; 1694 1695 default: 1696 return -1; 1697 } 1698 1699 /* NOTREACHED */ 1700 } 1701 1702 /* 1703 * get offset for the last header in the chain. m will be kept untainted. 1704 */ 1705 int 1706 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp) 1707 { 1708 int newoff; 1709 int nxt; 1710 1711 if (!nxtp) { 1712 nxt = -1; 1713 nxtp = &nxt; 1714 } 1715 while (1) { 1716 newoff = ip6_nexthdr(m, off, proto, nxtp); 1717 if (newoff < 0) 1718 return off; 1719 else if (newoff < off) 1720 return -1; /* invalid */ 1721 else if (newoff == off) 1722 return newoff; 1723 1724 off = newoff; 1725 proto = *nxtp; 1726 } 1727 } 1728 1729 /* 1730 * System control for IP6 1731 */ 1732 1733 u_char inet6ctlerrmap[PRC_NCMDS] = { 1734 0, 0, 0, 0, 1735 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1736 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1737 EMSGSIZE, EHOSTUNREACH, 0, 0, 1738 0, 0, 0, 0, 1739 ENOPROTOOPT 1740 }; 1741