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: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun 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_icmp.c 8.2 (Berkeley) 1/4/94 63 */ 64 65 #include <sys/cdefs.h> 66 __FBSDID("$FreeBSD$"); 67 68 #define MBUF_PRIVATE /* XXXRW: Optimisation tries to avoid M_EXT mbufs */ 69 70 #include "opt_inet.h" 71 #include "opt_inet6.h" 72 73 #include <sys/param.h> 74 #include <sys/domain.h> 75 #include <sys/jail.h> 76 #include <sys/kernel.h> 77 #include <sys/lock.h> 78 #include <sys/malloc.h> 79 #include <sys/mbuf.h> 80 #include <sys/proc.h> 81 #include <sys/protosw.h> 82 #include <sys/signalvar.h> 83 #include <sys/socket.h> 84 #include <sys/socketvar.h> 85 #include <sys/sx.h> 86 #include <sys/syslog.h> 87 #include <sys/systm.h> 88 #include <sys/time.h> 89 90 #include <net/if.h> 91 #include <net/if_var.h> 92 #include <net/if_dl.h> 93 #include <net/if_llatbl.h> 94 #include <net/if_types.h> 95 #include <net/route.h> 96 #include <net/route/route_ctl.h> 97 #include <net/route/nhop.h> 98 #include <net/vnet.h> 99 100 #include <netinet/in.h> 101 #include <netinet/in_pcb.h> 102 #include <netinet/in_var.h> 103 #include <netinet/ip6.h> 104 #include <netinet/icmp6.h> 105 #include <netinet/tcp_var.h> 106 107 #include <netinet6/in6_fib.h> 108 #include <netinet6/in6_ifattach.h> 109 #include <netinet6/in6_pcb.h> 110 #include <netinet6/ip6protosw.h> 111 #include <netinet6/ip6_var.h> 112 #include <netinet6/scope6_var.h> 113 #include <netinet6/mld6_var.h> 114 #include <netinet6/nd6.h> 115 #include <netinet6/send.h> 116 117 extern struct domain inet6domain; 118 119 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat); 120 VNET_PCPUSTAT_SYSINIT(icmp6stat); 121 122 #ifdef VIMAGE 123 VNET_PCPUSTAT_SYSUNINIT(icmp6stat); 124 #endif /* VIMAGE */ 125 126 VNET_DECLARE(struct inpcbinfo, ripcbinfo); 127 VNET_DECLARE(struct inpcbhead, ripcb); 128 VNET_DECLARE(int, icmp6errppslim); 129 VNET_DEFINE_STATIC(int, icmp6errpps_count) = 0; 130 VNET_DEFINE_STATIC(struct timeval, icmp6errppslim_last); 131 VNET_DECLARE(int, icmp6_nodeinfo); 132 133 #define V_ripcbinfo VNET(ripcbinfo) 134 #define V_ripcb VNET(ripcb) 135 #define V_icmp6errppslim VNET(icmp6errppslim) 136 #define V_icmp6errpps_count VNET(icmp6errpps_count) 137 #define V_icmp6errppslim_last VNET(icmp6errppslim_last) 138 #define V_icmp6_nodeinfo VNET(icmp6_nodeinfo) 139 140 static void icmp6_errcount(int, int); 141 static int icmp6_rip6_input(struct mbuf **, int); 142 static void icmp6_reflect(struct mbuf *, size_t); 143 static int icmp6_ratelimit(const struct in6_addr *, const int, const int); 144 static const char *icmp6_redirect_diag(struct in6_addr *, 145 struct in6_addr *, struct in6_addr *); 146 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *); 147 static struct mbuf *ni6_nametodns(const char *, int, int); 148 static int ni6_dnsmatch(const char *, int, const char *, int); 149 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *, 150 struct ifnet **, struct in6_addr *); 151 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 152 struct ifnet *, int); 153 static int icmp6_notify_error(struct mbuf **, int, int, int); 154 155 /* 156 * Kernel module interface for updating icmp6stat. The argument is an index 157 * into icmp6stat treated as an array of u_quad_t. While this encodes the 158 * general layout of icmp6stat into the caller, it doesn't encode its 159 * location, so that future changes to add, for example, per-CPU stats 160 * support won't cause binary compatibility problems for kernel modules. 161 */ 162 void 163 kmod_icmp6stat_inc(int statnum) 164 { 165 166 counter_u64_add(VNET(icmp6stat)[statnum], 1); 167 } 168 169 static void 170 icmp6_errcount(int type, int code) 171 { 172 switch (type) { 173 case ICMP6_DST_UNREACH: 174 switch (code) { 175 case ICMP6_DST_UNREACH_NOROUTE: 176 ICMP6STAT_INC(icp6s_odst_unreach_noroute); 177 return; 178 case ICMP6_DST_UNREACH_ADMIN: 179 ICMP6STAT_INC(icp6s_odst_unreach_admin); 180 return; 181 case ICMP6_DST_UNREACH_BEYONDSCOPE: 182 ICMP6STAT_INC(icp6s_odst_unreach_beyondscope); 183 return; 184 case ICMP6_DST_UNREACH_ADDR: 185 ICMP6STAT_INC(icp6s_odst_unreach_addr); 186 return; 187 case ICMP6_DST_UNREACH_NOPORT: 188 ICMP6STAT_INC(icp6s_odst_unreach_noport); 189 return; 190 } 191 break; 192 case ICMP6_PACKET_TOO_BIG: 193 ICMP6STAT_INC(icp6s_opacket_too_big); 194 return; 195 case ICMP6_TIME_EXCEEDED: 196 switch (code) { 197 case ICMP6_TIME_EXCEED_TRANSIT: 198 ICMP6STAT_INC(icp6s_otime_exceed_transit); 199 return; 200 case ICMP6_TIME_EXCEED_REASSEMBLY: 201 ICMP6STAT_INC(icp6s_otime_exceed_reassembly); 202 return; 203 } 204 break; 205 case ICMP6_PARAM_PROB: 206 switch (code) { 207 case ICMP6_PARAMPROB_HEADER: 208 ICMP6STAT_INC(icp6s_oparamprob_header); 209 return; 210 case ICMP6_PARAMPROB_NEXTHEADER: 211 ICMP6STAT_INC(icp6s_oparamprob_nextheader); 212 return; 213 case ICMP6_PARAMPROB_OPTION: 214 ICMP6STAT_INC(icp6s_oparamprob_option); 215 return; 216 } 217 break; 218 case ND_REDIRECT: 219 ICMP6STAT_INC(icp6s_oredirect); 220 return; 221 } 222 ICMP6STAT_INC(icp6s_ounknown); 223 } 224 225 /* 226 * A wrapper function for icmp6_error() necessary when the erroneous packet 227 * may not contain enough scope zone information. 228 */ 229 void 230 icmp6_error2(struct mbuf *m, int type, int code, int param, 231 struct ifnet *ifp) 232 { 233 struct ip6_hdr *ip6; 234 235 if (ifp == NULL) 236 return; 237 238 if (m->m_len < sizeof(struct ip6_hdr)) { 239 m = m_pullup(m, sizeof(struct ip6_hdr)); 240 if (m == NULL) { 241 IP6STAT_INC(ip6s_exthdrtoolong); 242 return; 243 } 244 } 245 ip6 = mtod(m, struct ip6_hdr *); 246 247 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 248 return; 249 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 250 return; 251 252 icmp6_error(m, type, code, param); 253 } 254 255 /* 256 * Generate an error packet of type error in response to bad IP6 packet. 257 */ 258 void 259 icmp6_error(struct mbuf *m, int type, int code, int param) 260 { 261 struct ip6_hdr *oip6, *nip6; 262 struct icmp6_hdr *icmp6; 263 u_int preplen; 264 int off; 265 int nxt; 266 267 ICMP6STAT_INC(icp6s_error); 268 269 /* count per-type-code statistics */ 270 icmp6_errcount(type, code); 271 272 #ifdef M_DECRYPTED /*not openbsd*/ 273 if (m->m_flags & M_DECRYPTED) { 274 ICMP6STAT_INC(icp6s_canterror); 275 goto freeit; 276 } 277 #endif 278 279 if (m->m_len < sizeof(struct ip6_hdr)) { 280 m = m_pullup(m, sizeof(struct ip6_hdr)); 281 if (m == NULL) { 282 IP6STAT_INC(ip6s_exthdrtoolong); 283 return; 284 } 285 } 286 oip6 = mtod(m, struct ip6_hdr *); 287 288 /* 289 * If the destination address of the erroneous packet is a multicast 290 * address, or the packet was sent using link-layer multicast, 291 * we should basically suppress sending an error (RFC 2463, Section 292 * 2.4). 293 * We have two exceptions (the item e.2 in that section): 294 * - the Packet Too Big message can be sent for path MTU discovery. 295 * - the Parameter Problem Message that can be allowed an icmp6 error 296 * in the option type field. This check has been done in 297 * ip6_unknown_opt(), so we can just check the type and code. 298 */ 299 if ((m->m_flags & (M_BCAST|M_MCAST) || 300 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 301 (type != ICMP6_PACKET_TOO_BIG && 302 (type != ICMP6_PARAM_PROB || 303 code != ICMP6_PARAMPROB_OPTION))) 304 goto freeit; 305 306 /* 307 * RFC 2463, 2.4 (e.5): source address check. 308 * XXX: the case of anycast source? 309 */ 310 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 311 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 312 goto freeit; 313 314 /* 315 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 316 * don't do it. 317 */ 318 nxt = -1; 319 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 320 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 321 struct icmp6_hdr *icp; 322 323 if (m->m_len < off + sizeof(struct icmp6_hdr)) { 324 m = m_pullup(m, off + sizeof(struct icmp6_hdr)); 325 if (m == NULL) { 326 IP6STAT_INC(ip6s_exthdrtoolong); 327 return; 328 } 329 } 330 oip6 = mtod(m, struct ip6_hdr *); 331 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 332 333 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 334 icp->icmp6_type == ND_REDIRECT) { 335 /* 336 * ICMPv6 error 337 * Special case: for redirect (which is 338 * informational) we must not send icmp6 error. 339 */ 340 ICMP6STAT_INC(icp6s_canterror); 341 goto freeit; 342 } else { 343 /* ICMPv6 informational - send the error */ 344 } 345 } else { 346 /* non-ICMPv6 - send the error */ 347 } 348 349 /* Finally, do rate limitation check. */ 350 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 351 ICMP6STAT_INC(icp6s_toofreq); 352 goto freeit; 353 } 354 355 /* 356 * OK, ICMP6 can be generated. 357 */ 358 359 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 360 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 361 362 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 363 M_PREPEND(m, preplen, M_NOWAIT); /* FIB is also copied over. */ 364 if (m == NULL) { 365 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 366 return; 367 } 368 369 nip6 = mtod(m, struct ip6_hdr *); 370 nip6->ip6_src = oip6->ip6_src; 371 nip6->ip6_dst = oip6->ip6_dst; 372 373 in6_clearscope(&oip6->ip6_src); 374 in6_clearscope(&oip6->ip6_dst); 375 376 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 377 icmp6->icmp6_type = type; 378 icmp6->icmp6_code = code; 379 icmp6->icmp6_pptr = htonl((u_int32_t)param); 380 381 ICMP6STAT_INC(icp6s_outhist[type]); 382 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 383 384 return; 385 386 freeit: 387 /* 388 * If we can't tell whether or not we can generate ICMP6, free it. 389 */ 390 m_freem(m); 391 } 392 393 /* 394 * Process a received ICMP6 message. 395 */ 396 int 397 icmp6_input(struct mbuf **mp, int *offp, int proto) 398 { 399 struct mbuf *m, *n; 400 struct ifnet *ifp; 401 struct ip6_hdr *ip6, *nip6; 402 struct icmp6_hdr *icmp6, *nicmp6; 403 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 404 int code, error, icmp6len, ip6len, noff, off, sum; 405 406 NET_EPOCH_ASSERT(); 407 408 m = *mp; 409 off = *offp; 410 411 if (m->m_len < off + sizeof(struct icmp6_hdr)) { 412 m = m_pullup(m, off + sizeof(struct icmp6_hdr)); 413 if (m == NULL) { 414 IP6STAT_INC(ip6s_exthdrtoolong); 415 *mp = m; 416 return (IPPROTO_DONE); 417 } 418 } 419 420 /* 421 * Locate icmp6 structure in mbuf, and check 422 * that not corrupted and of at least minimum length 423 */ 424 425 icmp6len = m->m_pkthdr.len - off; 426 if (icmp6len < sizeof(struct icmp6_hdr)) { 427 ICMP6STAT_INC(icp6s_tooshort); 428 goto freeit; 429 } 430 431 ip6 = mtod(m, struct ip6_hdr *); 432 ifp = m->m_pkthdr.rcvif; 433 /* 434 * Check multicast group membership. 435 * Note: SSM filters are not applied for ICMPv6 traffic. 436 */ 437 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 438 struct in6_multi *inm; 439 440 inm = in6m_lookup(ifp, &ip6->ip6_dst); 441 if (inm == NULL) { 442 IP6STAT_INC(ip6s_notmember); 443 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 444 goto freeit; 445 } 446 } 447 448 /* Calculate the checksum. */ 449 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 450 code = icmp6->icmp6_code; 451 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 452 nd6log((LOG_ERR, 453 "ICMP6 checksum error(%d|%x) %s\n", 454 icmp6->icmp6_type, sum, 455 ip6_sprintf(ip6bufs, &ip6->ip6_src))); 456 ICMP6STAT_INC(icp6s_checksum); 457 goto freeit; 458 } 459 460 ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]); 461 icmp6_ifstat_inc(ifp, ifs6_in_msg); 462 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK) 463 icmp6_ifstat_inc(ifp, ifs6_in_error); 464 465 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 466 switch (icmp6->icmp6_type) { 467 case ICMP6_DST_UNREACH: 468 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach); 469 switch (code) { 470 case ICMP6_DST_UNREACH_NOROUTE: 471 case ICMP6_DST_UNREACH_ADDR: /* PRC_HOSTDEAD is a DOS */ 472 code = PRC_UNREACH_NET; 473 break; 474 case ICMP6_DST_UNREACH_ADMIN: 475 icmp6_ifstat_inc(ifp, ifs6_in_adminprohib); 476 code = PRC_UNREACH_ADMIN_PROHIB; 477 break; 478 case ICMP6_DST_UNREACH_BEYONDSCOPE: 479 /* I mean "source address was incorrect." */ 480 code = PRC_PARAMPROB; 481 break; 482 case ICMP6_DST_UNREACH_NOPORT: 483 code = PRC_UNREACH_PORT; 484 break; 485 default: 486 goto badcode; 487 } 488 goto deliver; 489 break; 490 491 case ICMP6_PACKET_TOO_BIG: 492 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig); 493 494 /* validation is made in icmp6_mtudisc_update */ 495 496 code = PRC_MSGSIZE; 497 498 /* 499 * Updating the path MTU will be done after examining 500 * intermediate extension headers. 501 */ 502 goto deliver; 503 break; 504 505 case ICMP6_TIME_EXCEEDED: 506 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed); 507 switch (code) { 508 case ICMP6_TIME_EXCEED_TRANSIT: 509 code = PRC_TIMXCEED_INTRANS; 510 break; 511 case ICMP6_TIME_EXCEED_REASSEMBLY: 512 code = PRC_TIMXCEED_REASS; 513 break; 514 default: 515 goto badcode; 516 } 517 goto deliver; 518 break; 519 520 case ICMP6_PARAM_PROB: 521 icmp6_ifstat_inc(ifp, ifs6_in_paramprob); 522 switch (code) { 523 case ICMP6_PARAMPROB_NEXTHEADER: 524 code = PRC_UNREACH_PROTOCOL; 525 break; 526 case ICMP6_PARAMPROB_HEADER: 527 case ICMP6_PARAMPROB_OPTION: 528 code = PRC_PARAMPROB; 529 break; 530 default: 531 goto badcode; 532 } 533 goto deliver; 534 break; 535 536 case ICMP6_ECHO_REQUEST: 537 icmp6_ifstat_inc(ifp, ifs6_in_echo); 538 if (code != 0) 539 goto badcode; 540 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) { 541 /* Give up remote */ 542 break; 543 } 544 if (!M_WRITABLE(n) 545 || n->m_len < off + sizeof(struct icmp6_hdr)) { 546 struct mbuf *n0 = n; 547 int n0len; 548 549 CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN); 550 n = m_gethdr(M_NOWAIT, n0->m_type); 551 if (n == NULL) { 552 /* Give up remote */ 553 m_freem(n0); 554 break; 555 } 556 557 m_move_pkthdr(n, n0); /* FIB copied. */ 558 n0len = n0->m_pkthdr.len; /* save for use below */ 559 /* 560 * Copy IPv6 and ICMPv6 only. 561 */ 562 nip6 = mtod(n, struct ip6_hdr *); 563 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 564 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 565 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 566 noff = sizeof(struct ip6_hdr); 567 /* new mbuf contains only ipv6+icmpv6 headers */ 568 n->m_len = noff + sizeof(struct icmp6_hdr); 569 /* 570 * Adjust mbuf. ip6_plen will be adjusted in 571 * ip6_output(). 572 */ 573 m_adj(n0, off + sizeof(struct icmp6_hdr)); 574 /* recalculate complete packet size */ 575 n->m_pkthdr.len = n0len + (noff - off); 576 n->m_next = n0; 577 } else { 578 if (n->m_len < off + sizeof(*nicmp6)) { 579 n = m_pullup(n, off + sizeof(*nicmp6)); 580 if (n == NULL) { 581 IP6STAT_INC(ip6s_exthdrtoolong); 582 break; 583 } 584 } 585 nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off); 586 noff = off; 587 } 588 if (n) { 589 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 590 nicmp6->icmp6_code = 0; 591 ICMP6STAT_INC(icp6s_reflect); 592 ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]); 593 icmp6_reflect(n, noff); 594 } 595 break; 596 597 case ICMP6_ECHO_REPLY: 598 icmp6_ifstat_inc(ifp, ifs6_in_echoreply); 599 if (code != 0) 600 goto badcode; 601 break; 602 603 case MLD_LISTENER_QUERY: 604 case MLD_LISTENER_REPORT: 605 case MLD_LISTENER_DONE: 606 case MLDV2_LISTENER_REPORT: 607 /* 608 * Drop MLD traffic which is not link-local, has a hop limit 609 * of greater than 1 hop, or which does not have the 610 * IPv6 HBH Router Alert option. 611 * As IPv6 HBH options are stripped in ip6_input() we must 612 * check an mbuf header flag. 613 * XXX Should we also sanity check that these messages 614 * were directed to a link-local multicast prefix? 615 */ 616 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0) 617 goto freeit; 618 if (mld_input(&m, off, icmp6len) != 0) { 619 *mp = NULL; 620 return (IPPROTO_DONE); 621 } 622 /* m stays. */ 623 break; 624 625 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 626 { 627 enum { WRU, FQDN } mode; 628 struct prison *pr; 629 630 if (!V_icmp6_nodeinfo) 631 break; 632 633 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 634 mode = WRU; 635 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 636 mode = FQDN; 637 else 638 goto badlen; 639 640 pr = NULL; 641 sx_slock(&allprison_lock); 642 TAILQ_FOREACH(pr, &allprison, pr_list) 643 if (pr->pr_vnet == ifp->if_vnet) 644 break; 645 sx_sunlock(&allprison_lock); 646 if (pr == NULL) 647 pr = curthread->td_ucred->cr_prison; 648 if (mode == FQDN) { 649 if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) { 650 m = m_pullup(m, off + 651 sizeof(struct icmp6_nodeinfo)); 652 if (m == NULL) { 653 IP6STAT_INC(ip6s_exthdrtoolong); 654 *mp = m; 655 return (IPPROTO_DONE); 656 } 657 } 658 n = m_copym(m, 0, M_COPYALL, M_NOWAIT); 659 if (n) 660 n = ni6_input(n, off, pr); 661 /* XXX meaningless if n == NULL */ 662 noff = sizeof(struct ip6_hdr); 663 } else { 664 u_char *p; 665 int maxhlen, hlen; 666 667 /* 668 * XXX: this combination of flags is pointless, 669 * but should we keep this for compatibility? 670 */ 671 if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK | 672 ICMP6_NODEINFO_TMPADDROK)) != 673 (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK)) 674 break; 675 676 if (code != 0) 677 goto badcode; 678 679 CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN); 680 n = m_gethdr(M_NOWAIT, m->m_type); 681 if (n == NULL) { 682 /* Give up remote */ 683 break; 684 } 685 if (!m_dup_pkthdr(n, m, M_NOWAIT)) { 686 /* 687 * Previous code did a blind M_COPY_PKTHDR 688 * and said "just for rcvif". If true, then 689 * we could tolerate the dup failing (due to 690 * the deep copy of the tag chain). For now 691 * be conservative and just fail. 692 */ 693 m_free(n); 694 n = NULL; 695 break; 696 } 697 /* 698 * Copy IPv6 and ICMPv6 only. 699 */ 700 nip6 = mtod(n, struct ip6_hdr *); 701 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 702 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 703 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 704 p = (u_char *)(nicmp6 + 1); 705 bzero(p, 4); 706 707 maxhlen = M_TRAILINGSPACE(n) - 708 (sizeof(*nip6) + sizeof(*nicmp6) + 4); 709 mtx_lock(&pr->pr_mtx); 710 hlen = strlen(pr->pr_hostname); 711 if (maxhlen > hlen) 712 maxhlen = hlen; 713 /* meaningless TTL */ 714 bcopy(pr->pr_hostname, p + 4, maxhlen); 715 mtx_unlock(&pr->pr_mtx); 716 noff = sizeof(struct ip6_hdr); 717 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 718 sizeof(struct icmp6_hdr) + 4 + maxhlen; 719 nicmp6->icmp6_type = ICMP6_WRUREPLY; 720 nicmp6->icmp6_code = 0; 721 } 722 if (n) { 723 ICMP6STAT_INC(icp6s_reflect); 724 ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]); 725 icmp6_reflect(n, noff); 726 } 727 break; 728 } 729 730 case ICMP6_WRUREPLY: 731 if (code != 0) 732 goto badcode; 733 break; 734 735 case ND_ROUTER_SOLICIT: 736 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit); 737 if (code != 0) 738 goto badcode; 739 if (icmp6len < sizeof(struct nd_router_solicit)) 740 goto badlen; 741 if (send_sendso_input_hook != NULL) { 742 if (m->m_len < off + icmp6len) { 743 m = m_pullup(m, off + icmp6len); 744 if (m == NULL) { 745 IP6STAT_INC(ip6s_exthdrtoolong); 746 *mp = NULL; 747 return (IPPROTO_DONE); 748 } 749 } 750 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len); 751 if (error == 0) { 752 m = NULL; 753 goto freeit; 754 } 755 } 756 n = m_copym(m, 0, M_COPYALL, M_NOWAIT); 757 nd6_rs_input(m, off, icmp6len); 758 m = n; 759 if (m == NULL) 760 goto freeit; 761 break; 762 763 case ND_ROUTER_ADVERT: 764 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert); 765 if (code != 0) 766 goto badcode; 767 if (icmp6len < sizeof(struct nd_router_advert)) 768 goto badlen; 769 if (send_sendso_input_hook != NULL) { 770 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len); 771 if (error == 0) { 772 m = NULL; 773 goto freeit; 774 } 775 } 776 n = m_copym(m, 0, M_COPYALL, M_NOWAIT); 777 nd6_ra_input(m, off, icmp6len); 778 m = n; 779 if (m == NULL) 780 goto freeit; 781 break; 782 783 case ND_NEIGHBOR_SOLICIT: 784 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit); 785 if (code != 0) 786 goto badcode; 787 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 788 goto badlen; 789 if (send_sendso_input_hook != NULL) { 790 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len); 791 if (error == 0) { 792 m = NULL; 793 goto freeit; 794 } 795 } 796 n = m_copym(m, 0, M_COPYALL, M_NOWAIT); 797 nd6_ns_input(m, off, icmp6len); 798 m = n; 799 if (m == NULL) 800 goto freeit; 801 break; 802 803 case ND_NEIGHBOR_ADVERT: 804 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert); 805 if (code != 0) 806 goto badcode; 807 if (icmp6len < sizeof(struct nd_neighbor_advert)) 808 goto badlen; 809 if (send_sendso_input_hook != NULL) { 810 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len); 811 if (error == 0) { 812 m = NULL; 813 goto freeit; 814 } 815 } 816 n = m_copym(m, 0, M_COPYALL, M_NOWAIT); 817 nd6_na_input(m, off, icmp6len); 818 m = n; 819 if (m == NULL) 820 goto freeit; 821 break; 822 823 case ND_REDIRECT: 824 icmp6_ifstat_inc(ifp, ifs6_in_redirect); 825 if (code != 0) 826 goto badcode; 827 if (icmp6len < sizeof(struct nd_redirect)) 828 goto badlen; 829 if (send_sendso_input_hook != NULL) { 830 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len); 831 if (error == 0) { 832 m = NULL; 833 goto freeit; 834 } 835 } 836 n = m_copym(m, 0, M_COPYALL, M_NOWAIT); 837 icmp6_redirect_input(m, off); 838 m = n; 839 if (m == NULL) 840 goto freeit; 841 break; 842 843 case ICMP6_ROUTER_RENUMBERING: 844 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 845 code != ICMP6_ROUTER_RENUMBERING_RESULT) 846 goto badcode; 847 if (icmp6len < sizeof(struct icmp6_router_renum)) 848 goto badlen; 849 break; 850 851 default: 852 nd6log((LOG_DEBUG, 853 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 854 icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src), 855 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 856 ifp ? ifp->if_index : 0)); 857 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 858 /* ICMPv6 error: MUST deliver it by spec... */ 859 code = PRC_NCMDS; 860 /* deliver */ 861 } else { 862 /* ICMPv6 informational: MUST not deliver */ 863 break; 864 } 865 deliver: 866 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) { 867 /* In this case, m should've been freed. */ 868 *mp = NULL; 869 return (IPPROTO_DONE); 870 } 871 break; 872 873 badcode: 874 ICMP6STAT_INC(icp6s_badcode); 875 break; 876 877 badlen: 878 ICMP6STAT_INC(icp6s_badlen); 879 break; 880 } 881 882 /* deliver the packet to appropriate sockets */ 883 icmp6_rip6_input(&m, *offp); 884 885 *mp = m; 886 return (IPPROTO_DONE); 887 888 freeit: 889 m_freem(m); 890 *mp = NULL; 891 return (IPPROTO_DONE); 892 } 893 894 static int 895 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code) 896 { 897 struct mbuf *m; 898 struct icmp6_hdr *icmp6; 899 struct ip6_hdr *eip6; 900 u_int32_t notifymtu; 901 struct sockaddr_in6 icmp6src, icmp6dst; 902 903 m = *mp; 904 905 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 906 ICMP6STAT_INC(icp6s_tooshort); 907 goto freeit; 908 } 909 910 if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) { 911 m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr)); 912 if (m == NULL) { 913 IP6STAT_INC(ip6s_exthdrtoolong); 914 *mp = m; 915 return (-1); 916 } 917 } 918 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 919 eip6 = (struct ip6_hdr *)(icmp6 + 1); 920 921 /* Detect the upper level protocol */ 922 { 923 void (*ctlfunc)(int, struct sockaddr *, void *); 924 u_int8_t nxt = eip6->ip6_nxt; 925 int eoff = off + sizeof(struct icmp6_hdr) + 926 sizeof(struct ip6_hdr); 927 struct ip6ctlparam ip6cp; 928 struct in6_addr *finaldst = NULL; 929 int icmp6type = icmp6->icmp6_type; 930 struct ip6_frag *fh; 931 struct ip6_rthdr *rth; 932 struct ip6_rthdr0 *rth0; 933 int rthlen; 934 935 while (1) { /* XXX: should avoid infinite loop explicitly? */ 936 struct ip6_ext *eh; 937 938 switch (nxt) { 939 case IPPROTO_HOPOPTS: 940 case IPPROTO_DSTOPTS: 941 case IPPROTO_AH: 942 if (m->m_len < eoff + sizeof(struct ip6_ext)) { 943 m = m_pullup(m, eoff + 944 sizeof(struct ip6_ext)); 945 if (m == NULL) { 946 IP6STAT_INC(ip6s_exthdrtoolong); 947 *mp = m; 948 return (-1); 949 } 950 } 951 eh = (struct ip6_ext *) 952 (mtod(m, caddr_t) + eoff); 953 if (nxt == IPPROTO_AH) 954 eoff += (eh->ip6e_len + 2) << 2; 955 else 956 eoff += (eh->ip6e_len + 1) << 3; 957 nxt = eh->ip6e_nxt; 958 break; 959 case IPPROTO_ROUTING: 960 /* 961 * When the erroneous packet contains a 962 * routing header, we should examine the 963 * header to determine the final destination. 964 * Otherwise, we can't properly update 965 * information that depends on the final 966 * destination (e.g. path MTU). 967 */ 968 if (m->m_len < eoff + sizeof(*rth)) { 969 m = m_pullup(m, eoff + sizeof(*rth)); 970 if (m == NULL) { 971 IP6STAT_INC(ip6s_exthdrtoolong); 972 *mp = m; 973 return (-1); 974 } 975 } 976 rth = (struct ip6_rthdr *) 977 (mtod(m, caddr_t) + eoff); 978 rthlen = (rth->ip6r_len + 1) << 3; 979 /* 980 * XXX: currently there is no 981 * officially defined type other 982 * than type-0. 983 * Note that if the segment left field 984 * is 0, all intermediate hops must 985 * have been passed. 986 */ 987 if (rth->ip6r_segleft && 988 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 989 int hops; 990 991 if (m->m_len < eoff + rthlen) { 992 m = m_pullup(m, eoff + rthlen); 993 if (m == NULL) { 994 IP6STAT_INC( 995 ip6s_exthdrtoolong); 996 *mp = m; 997 return (-1); 998 } 999 } 1000 rth0 = (struct ip6_rthdr0 *) 1001 (mtod(m, caddr_t) + eoff); 1002 /* just ignore a bogus header */ 1003 if ((rth0->ip6r0_len % 2) == 0 && 1004 (hops = rth0->ip6r0_len/2)) 1005 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 1006 } 1007 eoff += rthlen; 1008 nxt = rth->ip6r_nxt; 1009 break; 1010 case IPPROTO_FRAGMENT: 1011 if (m->m_len < eoff + sizeof(struct ip6_frag)) { 1012 m = m_pullup(m, eoff + 1013 sizeof(struct ip6_frag)); 1014 if (m == NULL) { 1015 IP6STAT_INC(ip6s_exthdrtoolong); 1016 *mp = m; 1017 return (-1); 1018 } 1019 } 1020 fh = (struct ip6_frag *)(mtod(m, caddr_t) + 1021 eoff); 1022 /* 1023 * Data after a fragment header is meaningless 1024 * unless it is the first fragment, but 1025 * we'll go to the notify label for path MTU 1026 * discovery. 1027 */ 1028 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1029 goto notify; 1030 1031 eoff += sizeof(struct ip6_frag); 1032 nxt = fh->ip6f_nxt; 1033 break; 1034 default: 1035 /* 1036 * This case includes ESP and the No Next 1037 * Header. In such cases going to the notify 1038 * label does not have any meaning 1039 * (i.e. ctlfunc will be NULL), but we go 1040 * anyway since we might have to update 1041 * path MTU information. 1042 */ 1043 goto notify; 1044 } 1045 } 1046 notify: 1047 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 1048 1049 /* 1050 * retrieve parameters from the inner IPv6 header, and convert 1051 * them into sockaddr structures. 1052 * XXX: there is no guarantee that the source or destination 1053 * addresses of the inner packet are in the same scope as 1054 * the addresses of the icmp packet. But there is no other 1055 * way to determine the zone. 1056 */ 1057 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1058 1059 bzero(&icmp6dst, sizeof(icmp6dst)); 1060 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1061 icmp6dst.sin6_family = AF_INET6; 1062 if (finaldst == NULL) 1063 icmp6dst.sin6_addr = eip6->ip6_dst; 1064 else 1065 icmp6dst.sin6_addr = *finaldst; 1066 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1067 goto freeit; 1068 bzero(&icmp6src, sizeof(icmp6src)); 1069 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1070 icmp6src.sin6_family = AF_INET6; 1071 icmp6src.sin6_addr = eip6->ip6_src; 1072 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1073 goto freeit; 1074 icmp6src.sin6_flowinfo = 1075 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1076 1077 if (finaldst == NULL) 1078 finaldst = &eip6->ip6_dst; 1079 ip6cp.ip6c_m = m; 1080 ip6cp.ip6c_icmp6 = icmp6; 1081 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1082 ip6cp.ip6c_off = eoff; 1083 ip6cp.ip6c_finaldst = finaldst; 1084 ip6cp.ip6c_src = &icmp6src; 1085 ip6cp.ip6c_nxt = nxt; 1086 1087 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1088 notifymtu = ntohl(icmp6->icmp6_mtu); 1089 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1090 icmp6_mtudisc_update(&ip6cp, 1); /*XXX*/ 1091 } 1092 1093 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1094 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1095 if (ctlfunc) { 1096 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1097 &ip6cp); 1098 } 1099 } 1100 *mp = m; 1101 return (0); 1102 1103 freeit: 1104 m_freem(m); 1105 *mp = NULL; 1106 return (-1); 1107 } 1108 1109 void 1110 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1111 { 1112 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1113 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1114 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1115 u_int mtu = ntohl(icmp6->icmp6_mtu); 1116 struct in_conninfo inc; 1117 1118 #if 0 1119 /* 1120 * RFC2460 section 5, last paragraph. 1121 * even though minimum link MTU for IPv6 is IPV6_MMTU, 1122 * we may see ICMPv6 too big with mtu < IPV6_MMTU 1123 * due to packet translator in the middle. 1124 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for 1125 * special handling. 1126 */ 1127 if (mtu < IPV6_MMTU) 1128 return; 1129 #endif 1130 1131 /* 1132 * we reject ICMPv6 too big with abnormally small value. 1133 * XXX what is the good definition of "abnormally small"? 1134 */ 1135 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) 1136 return; 1137 1138 if (!validated) 1139 return; 1140 1141 /* 1142 * In case the suggested mtu is less than IPV6_MMTU, we 1143 * only need to remember that it was for above mentioned 1144 * "alwaysfrag" case. 1145 * Try to be as close to the spec as possible. 1146 */ 1147 if (mtu < IPV6_MMTU) 1148 mtu = IPV6_MMTU - 8; 1149 1150 bzero(&inc, sizeof(inc)); 1151 inc.inc_fibnum = M_GETFIB(m); 1152 inc.inc_flags |= INC_ISIPV6; 1153 inc.inc6_faddr = *dst; 1154 if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL)) 1155 return; 1156 1157 if (mtu < tcp_maxmtu6(&inc, NULL)) { 1158 tcp_hc_updatemtu(&inc, mtu); 1159 ICMP6STAT_INC(icp6s_pmtuchg); 1160 } 1161 } 1162 1163 /* 1164 * Process a Node Information Query packet, based on 1165 * draft-ietf-ipngwg-icmp-name-lookups-07. 1166 * 1167 * Spec incompatibilities: 1168 * - IPv6 Subject address handling 1169 * - IPv4 Subject address handling support missing 1170 * - Proxy reply (answer even if it's not for me) 1171 * - joins NI group address at in6_ifattach() time only, does not cope 1172 * with hostname changes by sethostname(3) 1173 */ 1174 static struct mbuf * 1175 ni6_input(struct mbuf *m, int off, struct prison *pr) 1176 { 1177 struct icmp6_nodeinfo *ni6, *nni6; 1178 struct mbuf *n = NULL; 1179 u_int16_t qtype; 1180 int subjlen; 1181 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1182 struct ni_reply_fqdn *fqdn; 1183 int addrs; /* for NI_QTYPE_NODEADDR */ 1184 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1185 struct in6_addr in6_subj; /* subject address */ 1186 struct ip6_hdr *ip6; 1187 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1188 char *subj = NULL; 1189 struct in6_ifaddr *ia6 = NULL; 1190 1191 ip6 = mtod(m, struct ip6_hdr *); 1192 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off); 1193 1194 /* 1195 * Validate IPv6 source address. 1196 * The default configuration MUST be to refuse answering queries from 1197 * global-scope addresses according to RFC4602. 1198 * Notes: 1199 * - it's not very clear what "refuse" means; this implementation 1200 * simply drops it. 1201 * - it's not very easy to identify global-scope (unicast) addresses 1202 * since there are many prefixes for them. It should be safer 1203 * and in practice sufficient to check "all" but loopback and 1204 * link-local (note that site-local unicast was deprecated and 1205 * ULA is defined as global scope-wise) 1206 */ 1207 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 && 1208 !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) && 1209 !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) 1210 goto bad; 1211 1212 /* 1213 * Validate IPv6 destination address. 1214 * 1215 * The Responder must discard the Query without further processing 1216 * unless it is one of the Responder's unicast or anycast addresses, or 1217 * a link-local scope multicast address which the Responder has joined. 1218 * [RFC4602, Section 5.] 1219 */ 1220 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 1221 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst)) 1222 goto bad; 1223 /* else it's a link-local multicast, fine */ 1224 } else { /* unicast or anycast */ 1225 ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */); 1226 if (ia6 == NULL) 1227 goto bad; /* XXX impossible */ 1228 1229 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) && 1230 !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) { 1231 ifa_free(&ia6->ia_ifa); 1232 nd6log((LOG_DEBUG, "ni6_input: ignore node info to " 1233 "a temporary address in %s:%d", 1234 __FILE__, __LINE__)); 1235 goto bad; 1236 } 1237 ifa_free(&ia6->ia_ifa); 1238 } 1239 1240 /* validate query Subject field. */ 1241 qtype = ntohs(ni6->ni_qtype); 1242 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1243 switch (qtype) { 1244 case NI_QTYPE_NOOP: 1245 case NI_QTYPE_SUPTYPES: 1246 /* 07 draft */ 1247 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1248 break; 1249 /* FALLTHROUGH */ 1250 case NI_QTYPE_FQDN: 1251 case NI_QTYPE_NODEADDR: 1252 case NI_QTYPE_IPV4ADDR: 1253 switch (ni6->ni_code) { 1254 case ICMP6_NI_SUBJ_IPV6: 1255 #if ICMP6_NI_SUBJ_IPV6 != 0 1256 case 0: 1257 #endif 1258 /* 1259 * backward compatibility - try to accept 03 draft 1260 * format, where no Subject is present. 1261 */ 1262 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1263 subjlen == 0) { 1264 oldfqdn++; 1265 break; 1266 } 1267 #if ICMP6_NI_SUBJ_IPV6 != 0 1268 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1269 goto bad; 1270 #endif 1271 1272 if (subjlen != sizeof(struct in6_addr)) 1273 goto bad; 1274 1275 /* 1276 * Validate Subject address. 1277 * 1278 * Not sure what exactly "address belongs to the node" 1279 * means in the spec, is it just unicast, or what? 1280 * 1281 * At this moment we consider Subject address as 1282 * "belong to the node" if the Subject address equals 1283 * to the IPv6 destination address; validation for 1284 * IPv6 destination address should have done enough 1285 * check for us. 1286 * 1287 * We do not do proxy at this moment. 1288 */ 1289 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1290 subjlen, (caddr_t)&in6_subj); 1291 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1292 goto bad; 1293 1294 subj = (char *)&in6_subj; 1295 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1296 break; 1297 1298 /* 1299 * XXX if we are to allow other cases, we should really 1300 * be careful about scope here. 1301 * basically, we should disallow queries toward IPv6 1302 * destination X with subject Y, 1303 * if scope(X) > scope(Y). 1304 * if we allow scope(X) > scope(Y), it will result in 1305 * information leakage across scope boundary. 1306 */ 1307 goto bad; 1308 1309 case ICMP6_NI_SUBJ_FQDN: 1310 /* 1311 * Validate Subject name with gethostname(3). 1312 * 1313 * The behavior may need some debate, since: 1314 * - we are not sure if the node has FQDN as 1315 * hostname (returned by gethostname(3)). 1316 * - the code does wildcard match for truncated names. 1317 * however, we are not sure if we want to perform 1318 * wildcard match, if gethostname(3) side has 1319 * truncated hostname. 1320 */ 1321 mtx_lock(&pr->pr_mtx); 1322 n = ni6_nametodns(pr->pr_hostname, 1323 strlen(pr->pr_hostname), 0); 1324 mtx_unlock(&pr->pr_mtx); 1325 if (!n || n->m_next || n->m_len == 0) 1326 goto bad; 1327 if (m->m_len < off + sizeof(struct icmp6_nodeinfo) + 1328 subjlen) { 1329 m = m_pullup(m, off + 1330 sizeof(struct icmp6_nodeinfo) + subjlen); 1331 if (m == NULL) { 1332 IP6STAT_INC(ip6s_exthdrtoolong); 1333 goto bad; 1334 } 1335 } 1336 /* ip6 possibly invalid but not used after. */ 1337 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off); 1338 subj = (char *)(mtod(m, caddr_t) + off + 1339 sizeof(struct icmp6_nodeinfo)); 1340 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1341 n->m_len)) { 1342 goto bad; 1343 } 1344 m_freem(n); 1345 n = NULL; 1346 break; 1347 1348 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1349 default: 1350 goto bad; 1351 } 1352 break; 1353 } 1354 1355 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1356 switch (qtype) { 1357 case NI_QTYPE_FQDN: 1358 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0) 1359 goto bad; 1360 break; 1361 case NI_QTYPE_NODEADDR: 1362 case NI_QTYPE_IPV4ADDR: 1363 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0) 1364 goto bad; 1365 break; 1366 } 1367 1368 /* guess reply length */ 1369 switch (qtype) { 1370 case NI_QTYPE_NOOP: 1371 break; /* no reply data */ 1372 case NI_QTYPE_SUPTYPES: 1373 replylen += sizeof(u_int32_t); 1374 break; 1375 case NI_QTYPE_FQDN: 1376 /* XXX will append an mbuf */ 1377 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1378 break; 1379 case NI_QTYPE_NODEADDR: 1380 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj); 1381 if ((replylen += addrs * (sizeof(struct in6_addr) + 1382 sizeof(u_int32_t))) > MCLBYTES) 1383 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1384 break; 1385 case NI_QTYPE_IPV4ADDR: 1386 /* unsupported - should respond with unknown Qtype? */ 1387 break; 1388 default: 1389 /* 1390 * XXX: We must return a reply with the ICMP6 code 1391 * `unknown Qtype' in this case. However we regard the case 1392 * as an FQDN query for backward compatibility. 1393 * Older versions set a random value to this field, 1394 * so it rarely varies in the defined qtypes. 1395 * But the mechanism is not reliable... 1396 * maybe we should obsolete older versions. 1397 */ 1398 qtype = NI_QTYPE_FQDN; 1399 /* XXX will append an mbuf */ 1400 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1401 oldfqdn++; 1402 break; 1403 } 1404 1405 /* Allocate an mbuf to reply. */ 1406 if (replylen > MCLBYTES) { 1407 /* 1408 * XXX: should we try to allocate more? But MCLBYTES 1409 * is probably much larger than IPV6_MMTU... 1410 */ 1411 goto bad; 1412 } 1413 if (replylen > MHLEN) 1414 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR); 1415 else 1416 n = m_gethdr(M_NOWAIT, m->m_type); 1417 if (n == NULL) { 1418 m_freem(m); 1419 return (NULL); 1420 } 1421 m_move_pkthdr(n, m); /* just for recvif and FIB */ 1422 n->m_pkthdr.len = n->m_len = replylen; 1423 1424 /* copy mbuf header and IPv6 + Node Information base headers */ 1425 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1426 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1427 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1428 1429 /* qtype dependent procedure */ 1430 switch (qtype) { 1431 case NI_QTYPE_NOOP: 1432 nni6->ni_code = ICMP6_NI_SUCCESS; 1433 nni6->ni_flags = 0; 1434 break; 1435 case NI_QTYPE_SUPTYPES: 1436 { 1437 u_int32_t v; 1438 nni6->ni_code = ICMP6_NI_SUCCESS; 1439 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1440 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1441 v = (u_int32_t)htonl(0x0000000f); 1442 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1443 break; 1444 } 1445 case NI_QTYPE_FQDN: 1446 nni6->ni_code = ICMP6_NI_SUCCESS; 1447 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1448 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo)); 1449 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1450 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1451 /* 1452 * XXX do we really have FQDN in hostname? 1453 */ 1454 mtx_lock(&pr->pr_mtx); 1455 n->m_next = ni6_nametodns(pr->pr_hostname, 1456 strlen(pr->pr_hostname), oldfqdn); 1457 mtx_unlock(&pr->pr_mtx); 1458 if (n->m_next == NULL) 1459 goto bad; 1460 /* XXX we assume that n->m_next is not a chain */ 1461 if (n->m_next->m_next != NULL) 1462 goto bad; 1463 n->m_pkthdr.len += n->m_next->m_len; 1464 break; 1465 case NI_QTYPE_NODEADDR: 1466 { 1467 int lenlim, copied; 1468 1469 nni6->ni_code = ICMP6_NI_SUCCESS; 1470 n->m_pkthdr.len = n->m_len = 1471 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1472 lenlim = M_TRAILINGSPACE(n); 1473 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1474 /* XXX: reset mbuf length */ 1475 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1476 sizeof(struct icmp6_nodeinfo) + copied; 1477 break; 1478 } 1479 default: 1480 break; /* XXX impossible! */ 1481 } 1482 1483 nni6->ni_type = ICMP6_NI_REPLY; 1484 m_freem(m); 1485 return (n); 1486 1487 bad: 1488 m_freem(m); 1489 if (n) 1490 m_freem(n); 1491 return (NULL); 1492 } 1493 1494 /* 1495 * make a mbuf with DNS-encoded string. no compression support. 1496 * 1497 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1498 * treated as truncated name (two \0 at the end). this is a wild guess. 1499 * 1500 * old - return pascal string if non-zero 1501 */ 1502 static struct mbuf * 1503 ni6_nametodns(const char *name, int namelen, int old) 1504 { 1505 struct mbuf *m; 1506 char *cp, *ep; 1507 const char *p, *q; 1508 int i, len, nterm; 1509 1510 if (old) 1511 len = namelen + 1; 1512 else 1513 len = MCLBYTES; 1514 1515 /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */ 1516 if (len > MLEN) 1517 m = m_getcl(M_NOWAIT, MT_DATA, 0); 1518 else 1519 m = m_get(M_NOWAIT, MT_DATA); 1520 if (m == NULL) 1521 goto fail; 1522 1523 if (old) { 1524 m->m_len = len; 1525 *mtod(m, char *) = namelen; 1526 bcopy(name, mtod(m, char *) + 1, namelen); 1527 return m; 1528 } else { 1529 m->m_len = 0; 1530 cp = mtod(m, char *); 1531 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1532 1533 /* if not certain about my name, return empty buffer */ 1534 if (namelen == 0) 1535 return m; 1536 1537 /* 1538 * guess if it looks like shortened hostname, or FQDN. 1539 * shortened hostname needs two trailing "\0". 1540 */ 1541 i = 0; 1542 for (p = name; p < name + namelen; p++) { 1543 if (*p && *p == '.') 1544 i++; 1545 } 1546 if (i < 2) 1547 nterm = 2; 1548 else 1549 nterm = 1; 1550 1551 p = name; 1552 while (cp < ep && p < name + namelen) { 1553 i = 0; 1554 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1555 i++; 1556 /* result does not fit into mbuf */ 1557 if (cp + i + 1 >= ep) 1558 goto fail; 1559 /* 1560 * DNS label length restriction, RFC1035 page 8. 1561 * "i == 0" case is included here to avoid returning 1562 * 0-length label on "foo..bar". 1563 */ 1564 if (i <= 0 || i >= 64) 1565 goto fail; 1566 *cp++ = i; 1567 bcopy(p, cp, i); 1568 cp += i; 1569 p = q; 1570 if (p < name + namelen && *p == '.') 1571 p++; 1572 } 1573 /* termination */ 1574 if (cp + nterm >= ep) 1575 goto fail; 1576 while (nterm-- > 0) 1577 *cp++ = '\0'; 1578 m->m_len = cp - mtod(m, char *); 1579 return m; 1580 } 1581 1582 panic("should not reach here"); 1583 /* NOTREACHED */ 1584 1585 fail: 1586 if (m) 1587 m_freem(m); 1588 return NULL; 1589 } 1590 1591 /* 1592 * check if two DNS-encoded string matches. takes care of truncated 1593 * form (with \0\0 at the end). no compression support. 1594 * XXX upper/lowercase match (see RFC2065) 1595 */ 1596 static int 1597 ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1598 { 1599 const char *a0, *b0; 1600 int l; 1601 1602 /* simplest case - need validation? */ 1603 if (alen == blen && bcmp(a, b, alen) == 0) 1604 return 1; 1605 1606 a0 = a; 1607 b0 = b; 1608 1609 /* termination is mandatory */ 1610 if (alen < 2 || blen < 2) 1611 return 0; 1612 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1613 return 0; 1614 alen--; 1615 blen--; 1616 1617 while (a - a0 < alen && b - b0 < blen) { 1618 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1619 return 0; 1620 1621 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1622 return 0; 1623 /* we don't support compression yet */ 1624 if (a[0] >= 64 || b[0] >= 64) 1625 return 0; 1626 1627 /* truncated case */ 1628 if (a[0] == 0 && a - a0 == alen - 1) 1629 return 1; 1630 if (b[0] == 0 && b - b0 == blen - 1) 1631 return 1; 1632 if (a[0] == 0 || b[0] == 0) 1633 return 0; 1634 1635 if (a[0] != b[0]) 1636 return 0; 1637 l = a[0]; 1638 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1639 return 0; 1640 if (bcmp(a + 1, b + 1, l) != 0) 1641 return 0; 1642 1643 a += 1 + l; 1644 b += 1 + l; 1645 } 1646 1647 if (a - a0 == alen && b - b0 == blen) 1648 return 1; 1649 else 1650 return 0; 1651 } 1652 1653 /* 1654 * calculate the number of addresses to be returned in the node info reply. 1655 */ 1656 static int 1657 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp, 1658 struct in6_addr *subj) 1659 { 1660 struct ifnet *ifp; 1661 struct in6_ifaddr *ifa6; 1662 struct ifaddr *ifa; 1663 int addrs = 0, addrsofif, iffound = 0; 1664 int niflags = ni6->ni_flags; 1665 1666 NET_EPOCH_ASSERT(); 1667 1668 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1669 switch (ni6->ni_code) { 1670 case ICMP6_NI_SUBJ_IPV6: 1671 if (subj == NULL) /* must be impossible... */ 1672 return (0); 1673 break; 1674 default: 1675 /* 1676 * XXX: we only support IPv6 subject address for 1677 * this Qtype. 1678 */ 1679 return (0); 1680 } 1681 } 1682 1683 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1684 addrsofif = 0; 1685 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1686 if (ifa->ifa_addr->sa_family != AF_INET6) 1687 continue; 1688 ifa6 = (struct in6_ifaddr *)ifa; 1689 1690 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1691 IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr)) 1692 iffound = 1; 1693 1694 /* 1695 * IPv4-mapped addresses can only be returned by a 1696 * Node Information proxy, since they represent 1697 * addresses of IPv4-only nodes, which perforce do 1698 * not implement this protocol. 1699 * [icmp-name-lookups-07, Section 5.4] 1700 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1701 * this function at this moment. 1702 */ 1703 1704 /* What do we have to do about ::1? */ 1705 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1706 case IPV6_ADDR_SCOPE_LINKLOCAL: 1707 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1708 continue; 1709 break; 1710 case IPV6_ADDR_SCOPE_SITELOCAL: 1711 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1712 continue; 1713 break; 1714 case IPV6_ADDR_SCOPE_GLOBAL: 1715 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1716 continue; 1717 break; 1718 default: 1719 continue; 1720 } 1721 1722 /* 1723 * check if anycast is okay. 1724 * XXX: just experimental. not in the spec. 1725 */ 1726 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1727 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1728 continue; /* we need only unicast addresses */ 1729 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1730 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) { 1731 continue; 1732 } 1733 addrsofif++; /* count the address */ 1734 } 1735 if (iffound) { 1736 *ifpp = ifp; 1737 return (addrsofif); 1738 } 1739 1740 addrs += addrsofif; 1741 } 1742 1743 return (addrs); 1744 } 1745 1746 static int 1747 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6, 1748 struct ifnet *ifp0, int resid) 1749 { 1750 struct ifnet *ifp; 1751 struct in6_ifaddr *ifa6; 1752 struct ifaddr *ifa; 1753 struct ifnet *ifp_dep = NULL; 1754 int copied = 0, allow_deprecated = 0; 1755 u_char *cp = (u_char *)(nni6 + 1); 1756 int niflags = ni6->ni_flags; 1757 u_int32_t ltime; 1758 1759 NET_EPOCH_ASSERT(); 1760 1761 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1762 return (0); /* needless to copy */ 1763 1764 ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet); 1765 again: 1766 1767 for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) { 1768 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1769 if (ifa->ifa_addr->sa_family != AF_INET6) 1770 continue; 1771 ifa6 = (struct in6_ifaddr *)ifa; 1772 1773 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1774 allow_deprecated == 0) { 1775 /* 1776 * prefererred address should be put before 1777 * deprecated addresses. 1778 */ 1779 1780 /* record the interface for later search */ 1781 if (ifp_dep == NULL) 1782 ifp_dep = ifp; 1783 1784 continue; 1785 } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1786 allow_deprecated != 0) 1787 continue; /* we now collect deprecated addrs */ 1788 1789 /* What do we have to do about ::1? */ 1790 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1791 case IPV6_ADDR_SCOPE_LINKLOCAL: 1792 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1793 continue; 1794 break; 1795 case IPV6_ADDR_SCOPE_SITELOCAL: 1796 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1797 continue; 1798 break; 1799 case IPV6_ADDR_SCOPE_GLOBAL: 1800 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1801 continue; 1802 break; 1803 default: 1804 continue; 1805 } 1806 1807 /* 1808 * check if anycast is okay. 1809 * XXX: just experimental. not in the spec. 1810 */ 1811 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1812 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1813 continue; 1814 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1815 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) { 1816 continue; 1817 } 1818 1819 /* now we can copy the address */ 1820 if (resid < sizeof(struct in6_addr) + 1821 sizeof(u_int32_t)) { 1822 /* 1823 * We give up much more copy. 1824 * Set the truncate flag and return. 1825 */ 1826 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1827 return (copied); 1828 } 1829 1830 /* 1831 * Set the TTL of the address. 1832 * The TTL value should be one of the following 1833 * according to the specification: 1834 * 1835 * 1. The remaining lifetime of a DHCP lease on the 1836 * address, or 1837 * 2. The remaining Valid Lifetime of a prefix from 1838 * which the address was derived through Stateless 1839 * Autoconfiguration. 1840 * 1841 * Note that we currently do not support stateful 1842 * address configuration by DHCPv6, so the former 1843 * case can't happen. 1844 */ 1845 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1846 ltime = ND6_INFINITE_LIFETIME; 1847 else { 1848 if (ifa6->ia6_lifetime.ia6t_expire > 1849 time_uptime) 1850 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime); 1851 else 1852 ltime = 0; 1853 } 1854 1855 bcopy(<ime, cp, sizeof(u_int32_t)); 1856 cp += sizeof(u_int32_t); 1857 1858 /* copy the address itself */ 1859 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1860 sizeof(struct in6_addr)); 1861 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1862 cp += sizeof(struct in6_addr); 1863 1864 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1865 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1866 } 1867 if (ifp0) /* we need search only on the specified IF */ 1868 break; 1869 } 1870 1871 if (allow_deprecated == 0 && ifp_dep != NULL) { 1872 ifp = ifp_dep; 1873 allow_deprecated = 1; 1874 1875 goto again; 1876 } 1877 1878 return (copied); 1879 } 1880 1881 /* 1882 * XXX almost dup'ed code with rip6_input. 1883 */ 1884 static int 1885 icmp6_rip6_input(struct mbuf **mp, int off) 1886 { 1887 struct mbuf *m = *mp; 1888 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1889 struct inpcb *inp; 1890 struct inpcb *last = NULL; 1891 struct sockaddr_in6 fromsa; 1892 struct icmp6_hdr *icmp6; 1893 struct mbuf *opts = NULL; 1894 1895 NET_EPOCH_ASSERT(); 1896 1897 /* This is assumed to be safe; icmp6_input() does a pullup. */ 1898 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 1899 1900 /* 1901 * XXX: the address may have embedded scope zone ID, which should be 1902 * hidden from applications. 1903 */ 1904 bzero(&fromsa, sizeof(fromsa)); 1905 fromsa.sin6_family = AF_INET6; 1906 fromsa.sin6_len = sizeof(struct sockaddr_in6); 1907 fromsa.sin6_addr = ip6->ip6_src; 1908 if (sa6_recoverscope(&fromsa)) { 1909 m_freem(m); 1910 *mp = NULL; 1911 return (IPPROTO_DONE); 1912 } 1913 1914 CK_LIST_FOREACH(inp, &V_ripcb, inp_list) { 1915 if ((inp->inp_vflag & INP_IPV6) == 0) 1916 continue; 1917 if (inp->inp_ip_p != IPPROTO_ICMPV6) 1918 continue; 1919 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) && 1920 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst)) 1921 continue; 1922 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) && 1923 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src)) 1924 continue; 1925 INP_RLOCK(inp); 1926 if (__predict_false(inp->inp_flags2 & INP_FREED)) { 1927 INP_RUNLOCK(inp); 1928 continue; 1929 } 1930 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1931 inp->in6p_icmp6filt)) { 1932 INP_RUNLOCK(inp); 1933 continue; 1934 } 1935 if (last != NULL) { 1936 struct mbuf *n = NULL; 1937 1938 /* 1939 * Recent network drivers tend to allocate a single 1940 * mbuf cluster, rather than to make a couple of 1941 * mbufs without clusters. Also, since the IPv6 code 1942 * path tries to avoid m_pullup(), it is highly 1943 * probable that we still have an mbuf cluster here 1944 * even though the necessary length can be stored in an 1945 * mbuf's internal buffer. 1946 * Meanwhile, the default size of the receive socket 1947 * buffer for raw sockets is not so large. This means 1948 * the possibility of packet loss is relatively higher 1949 * than before. To avoid this scenario, we copy the 1950 * received data to a separate mbuf that does not use 1951 * a cluster, if possible. 1952 * XXX: it is better to copy the data after stripping 1953 * intermediate headers. 1954 */ 1955 if ((m->m_flags & M_EXT) && m->m_next == NULL && 1956 m->m_len <= MHLEN) { 1957 n = m_get(M_NOWAIT, m->m_type); 1958 if (n != NULL) { 1959 if (m_dup_pkthdr(n, m, M_NOWAIT)) { 1960 bcopy(m->m_data, n->m_data, 1961 m->m_len); 1962 n->m_len = m->m_len; 1963 } else { 1964 m_free(n); 1965 n = NULL; 1966 } 1967 } 1968 } 1969 if (n != NULL || 1970 (n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) { 1971 if (last->inp_flags & INP_CONTROLOPTS) 1972 ip6_savecontrol(last, n, &opts); 1973 /* strip intermediate headers */ 1974 m_adj(n, off); 1975 SOCKBUF_LOCK(&last->inp_socket->so_rcv); 1976 if (sbappendaddr_locked( 1977 &last->inp_socket->so_rcv, 1978 (struct sockaddr *)&fromsa, n, opts) 1979 == 0) { 1980 /* should notify about lost packet */ 1981 m_freem(n); 1982 if (opts) { 1983 m_freem(opts); 1984 } 1985 SOCKBUF_UNLOCK( 1986 &last->inp_socket->so_rcv); 1987 } else 1988 sorwakeup_locked(last->inp_socket); 1989 opts = NULL; 1990 } 1991 INP_RUNLOCK(last); 1992 } 1993 last = inp; 1994 } 1995 if (last != NULL) { 1996 if (last->inp_flags & INP_CONTROLOPTS) 1997 ip6_savecontrol(last, m, &opts); 1998 /* strip intermediate headers */ 1999 m_adj(m, off); 2000 2001 /* avoid using mbuf clusters if possible (see above) */ 2002 if ((m->m_flags & M_EXT) && m->m_next == NULL && 2003 m->m_len <= MHLEN) { 2004 struct mbuf *n; 2005 2006 n = m_get(M_NOWAIT, m->m_type); 2007 if (n != NULL) { 2008 if (m_dup_pkthdr(n, m, M_NOWAIT)) { 2009 bcopy(m->m_data, n->m_data, m->m_len); 2010 n->m_len = m->m_len; 2011 2012 m_freem(m); 2013 m = n; 2014 } else { 2015 m_freem(n); 2016 n = NULL; 2017 } 2018 } 2019 } 2020 SOCKBUF_LOCK(&last->inp_socket->so_rcv); 2021 if (sbappendaddr_locked(&last->inp_socket->so_rcv, 2022 (struct sockaddr *)&fromsa, m, opts) == 0) { 2023 m_freem(m); 2024 if (opts) 2025 m_freem(opts); 2026 SOCKBUF_UNLOCK(&last->inp_socket->so_rcv); 2027 } else 2028 sorwakeup_locked(last->inp_socket); 2029 INP_RUNLOCK(last); 2030 } else { 2031 m_freem(m); 2032 IP6STAT_DEC(ip6s_delivered); 2033 } 2034 *mp = NULL; 2035 return (IPPROTO_DONE); 2036 } 2037 2038 /* 2039 * Reflect the ip6 packet back to the source. 2040 * OFF points to the icmp6 header, counted from the top of the mbuf. 2041 */ 2042 static void 2043 icmp6_reflect(struct mbuf *m, size_t off) 2044 { 2045 struct in6_addr src6, *srcp; 2046 struct ip6_hdr *ip6; 2047 struct icmp6_hdr *icmp6; 2048 struct in6_ifaddr *ia = NULL; 2049 struct ifnet *outif = NULL; 2050 int plen; 2051 int type, code, hlim; 2052 2053 /* too short to reflect */ 2054 if (off < sizeof(struct ip6_hdr)) { 2055 nd6log((LOG_DEBUG, 2056 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 2057 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2058 __FILE__, __LINE__)); 2059 goto bad; 2060 } 2061 2062 /* 2063 * If there are extra headers between IPv6 and ICMPv6, strip 2064 * off that header first. 2065 */ 2066 #ifdef DIAGNOSTIC 2067 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2068 panic("assumption failed in icmp6_reflect"); 2069 #endif 2070 if (off > sizeof(struct ip6_hdr)) { 2071 size_t l; 2072 struct ip6_hdr nip6; 2073 2074 l = off - sizeof(struct ip6_hdr); 2075 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2076 m_adj(m, l); 2077 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2078 if (m->m_len < l) { 2079 if ((m = m_pullup(m, l)) == NULL) 2080 return; 2081 } 2082 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2083 } else /* off == sizeof(struct ip6_hdr) */ { 2084 size_t l; 2085 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2086 if (m->m_len < l) { 2087 if ((m = m_pullup(m, l)) == NULL) 2088 return; 2089 } 2090 } 2091 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2092 ip6 = mtod(m, struct ip6_hdr *); 2093 ip6->ip6_nxt = IPPROTO_ICMPV6; 2094 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2095 type = icmp6->icmp6_type; /* keep type for statistics */ 2096 code = icmp6->icmp6_code; /* ditto. */ 2097 hlim = 0; 2098 srcp = NULL; 2099 2100 /* 2101 * If the incoming packet was addressed directly to us (i.e. unicast), 2102 * use dst as the src for the reply. 2103 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2104 * (for example) when we encounter an error while forwarding procedure 2105 * destined to a duplicated address of ours. 2106 */ 2107 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 2108 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */); 2109 if (ia != NULL && !(ia->ia6_flags & 2110 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) { 2111 src6 = ia->ia_addr.sin6_addr; 2112 srcp = &src6; 2113 2114 if (m->m_pkthdr.rcvif != NULL) { 2115 /* XXX: This may not be the outgoing interface */ 2116 hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2117 } else 2118 hlim = V_ip6_defhlim; 2119 } 2120 if (ia != NULL) 2121 ifa_free(&ia->ia_ifa); 2122 } 2123 2124 if (srcp == NULL) { 2125 int error; 2126 struct in6_addr dst6; 2127 uint32_t scopeid; 2128 2129 /* 2130 * This case matches to multicasts, our anycast, or unicasts 2131 * that we do not own. Select a source address based on the 2132 * source address of the erroneous packet. 2133 */ 2134 in6_splitscope(&ip6->ip6_src, &dst6, &scopeid); 2135 error = in6_selectsrc_addr(M_GETFIB(m), &dst6, 2136 scopeid, NULL, &src6, &hlim); 2137 2138 if (error) { 2139 char ip6buf[INET6_ADDRSTRLEN]; 2140 nd6log((LOG_DEBUG, 2141 "icmp6_reflect: source can't be determined: " 2142 "dst=%s, error=%d\n", 2143 ip6_sprintf(ip6buf, &ip6->ip6_dst), error)); 2144 goto bad; 2145 } 2146 srcp = &src6; 2147 } 2148 /* 2149 * ip6_input() drops a packet if its src is multicast. 2150 * So, the src is never multicast. 2151 */ 2152 ip6->ip6_dst = ip6->ip6_src; 2153 ip6->ip6_src = *srcp; 2154 ip6->ip6_flow = 0; 2155 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2156 ip6->ip6_vfc |= IPV6_VERSION; 2157 ip6->ip6_nxt = IPPROTO_ICMPV6; 2158 ip6->ip6_hlim = hlim; 2159 2160 icmp6->icmp6_cksum = 0; 2161 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2162 sizeof(struct ip6_hdr), plen); 2163 2164 /* 2165 * XXX option handling 2166 */ 2167 2168 m->m_flags &= ~(M_BCAST|M_MCAST); 2169 m->m_pkthdr.rcvif = NULL; 2170 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL); 2171 if (outif) 2172 icmp6_ifoutstat_inc(outif, type, code); 2173 2174 return; 2175 2176 bad: 2177 m_freem(m); 2178 return; 2179 } 2180 2181 void 2182 icmp6_fasttimo(void) 2183 { 2184 2185 mld_fasttimo(); 2186 } 2187 2188 void 2189 icmp6_slowtimo(void) 2190 { 2191 2192 mld_slowtimo(); 2193 } 2194 2195 static const char * 2196 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2197 struct in6_addr *tgt6) 2198 { 2199 static char buf[1024]; 2200 char ip6bufs[INET6_ADDRSTRLEN]; 2201 char ip6bufd[INET6_ADDRSTRLEN]; 2202 char ip6buft[INET6_ADDRSTRLEN]; 2203 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2204 ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6), 2205 ip6_sprintf(ip6buft, tgt6)); 2206 return buf; 2207 } 2208 2209 void 2210 icmp6_redirect_input(struct mbuf *m, int off) 2211 { 2212 struct ifnet *ifp; 2213 struct ip6_hdr *ip6; 2214 struct nd_redirect *nd_rd; 2215 struct in6_addr src6, redtgt6, reddst6; 2216 union nd_opts ndopts; 2217 char ip6buf[INET6_ADDRSTRLEN]; 2218 char *lladdr; 2219 int icmp6len, is_onlink, is_router, lladdrlen; 2220 2221 M_ASSERTPKTHDR(m); 2222 KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__)); 2223 2224 /* XXX if we are router, we don't update route by icmp6 redirect */ 2225 if (V_ip6_forwarding) 2226 goto freeit; 2227 if (!V_icmp6_rediraccept) 2228 goto freeit; 2229 2230 /* RFC 6980: Nodes MUST silently ignore fragments */ 2231 if(m->m_flags & M_FRAGMENTED) 2232 goto freeit; 2233 2234 ip6 = mtod(m, struct ip6_hdr *); 2235 icmp6len = ntohs(ip6->ip6_plen); 2236 if (m->m_len < off + icmp6len) { 2237 m = m_pullup(m, off + icmp6len); 2238 if (m == NULL) { 2239 IP6STAT_INC(ip6s_exthdrtoolong); 2240 return; 2241 } 2242 } 2243 ip6 = mtod(m, struct ip6_hdr *); 2244 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off); 2245 2246 ifp = m->m_pkthdr.rcvif; 2247 redtgt6 = nd_rd->nd_rd_target; 2248 reddst6 = nd_rd->nd_rd_dst; 2249 2250 if (in6_setscope(&redtgt6, ifp, NULL) || 2251 in6_setscope(&reddst6, ifp, NULL)) { 2252 goto freeit; 2253 } 2254 2255 /* validation */ 2256 src6 = ip6->ip6_src; 2257 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2258 nd6log((LOG_ERR, 2259 "ICMP6 redirect sent from %s rejected; " 2260 "must be from linklocal\n", 2261 ip6_sprintf(ip6buf, &src6))); 2262 goto bad; 2263 } 2264 if (ip6->ip6_hlim != 255) { 2265 nd6log((LOG_ERR, 2266 "ICMP6 redirect sent from %s rejected; " 2267 "hlim=%d (must be 255)\n", 2268 ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim)); 2269 goto bad; 2270 } 2271 { 2272 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2273 struct nhop6_basic nh6; 2274 struct in6_addr kdst; 2275 uint32_t scopeid; 2276 2277 in6_splitscope(&reddst6, &kdst, &scopeid); 2278 if (fib6_lookup_nh_basic(ifp->if_fib, &kdst, scopeid, 0, 0,&nh6)==0){ 2279 if ((nh6.nh_flags & NHF_GATEWAY) == 0) { 2280 nd6log((LOG_ERR, 2281 "ICMP6 redirect rejected; no route " 2282 "with inet6 gateway found for redirect dst: %s\n", 2283 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2284 goto bad; 2285 } 2286 2287 /* 2288 * Embed scope zone id into next hop address, since 2289 * fib6_lookup_nh_basic() returns address without embedded 2290 * scope zone id. 2291 */ 2292 if (in6_setscope(&nh6.nh_addr, m->m_pkthdr.rcvif, NULL)) 2293 goto freeit; 2294 2295 if (IN6_ARE_ADDR_EQUAL(&src6, &nh6.nh_addr) == 0) { 2296 nd6log((LOG_ERR, 2297 "ICMP6 redirect rejected; " 2298 "not equal to gw-for-src=%s (must be same): " 2299 "%s\n", 2300 ip6_sprintf(ip6buf, &nh6.nh_addr), 2301 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2302 goto bad; 2303 } 2304 } else { 2305 nd6log((LOG_ERR, 2306 "ICMP6 redirect rejected; " 2307 "no route found for redirect dst: %s\n", 2308 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2309 goto bad; 2310 } 2311 } 2312 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2313 nd6log((LOG_ERR, 2314 "ICMP6 redirect rejected; " 2315 "redirect dst must be unicast: %s\n", 2316 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2317 goto bad; 2318 } 2319 2320 is_router = is_onlink = 0; 2321 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2322 is_router = 1; /* router case */ 2323 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2324 is_onlink = 1; /* on-link destination case */ 2325 if (!is_router && !is_onlink) { 2326 nd6log((LOG_ERR, 2327 "ICMP6 redirect rejected; " 2328 "neither router case nor onlink case: %s\n", 2329 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2330 goto bad; 2331 } 2332 2333 icmp6len -= sizeof(*nd_rd); 2334 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2335 if (nd6_options(&ndopts) < 0) { 2336 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n", 2337 __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2338 /* nd6_options have incremented stats */ 2339 goto freeit; 2340 } 2341 2342 lladdr = NULL; 2343 lladdrlen = 0; 2344 if (ndopts.nd_opts_tgt_lladdr) { 2345 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2346 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2347 } 2348 2349 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2350 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s " 2351 "(if %d, icmp6 packet %d): %s\n", 2352 __func__, ip6_sprintf(ip6buf, &redtgt6), 2353 ifp->if_addrlen, lladdrlen - 2, 2354 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2355 goto bad; 2356 } 2357 2358 /* Validation passed. */ 2359 2360 /* RFC 2461 8.3 */ 2361 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2362 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2363 2364 /* 2365 * Install a gateway route in the better-router case or an interface 2366 * route in the on-link-destination case. 2367 */ 2368 { 2369 struct sockaddr_in6 sdst; 2370 struct sockaddr_in6 sgw; 2371 struct sockaddr_in6 ssrc; 2372 struct sockaddr *gw; 2373 int rt_flags; 2374 u_int fibnum; 2375 2376 bzero(&sdst, sizeof(sdst)); 2377 bzero(&ssrc, sizeof(ssrc)); 2378 sdst.sin6_family = ssrc.sin6_family = AF_INET6; 2379 sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6); 2380 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2381 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2382 rt_flags = 0; 2383 if (is_router) { 2384 bzero(&sgw, sizeof(sgw)); 2385 sgw.sin6_family = AF_INET6; 2386 sgw.sin6_len = sizeof(struct sockaddr_in6); 2387 bcopy(&redtgt6, &sgw.sin6_addr, 2388 sizeof(struct in6_addr)); 2389 gw = (struct sockaddr *)&sgw; 2390 rt_flags |= RTF_GATEWAY; 2391 } else 2392 gw = ifp->if_addr->ifa_addr; 2393 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) 2394 rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw, 2395 (struct sockaddr *)&ssrc, ifp, rt_flags, 2396 V_icmp6_redirtimeout); 2397 } 2398 /* finally update cached route in each socket via pfctlinput */ 2399 { 2400 struct sockaddr_in6 sdst; 2401 2402 bzero(&sdst, sizeof(sdst)); 2403 sdst.sin6_family = AF_INET6; 2404 sdst.sin6_len = sizeof(struct sockaddr_in6); 2405 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2406 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2407 } 2408 2409 freeit: 2410 m_freem(m); 2411 return; 2412 2413 bad: 2414 ICMP6STAT_INC(icp6s_badredirect); 2415 m_freem(m); 2416 } 2417 2418 void 2419 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh) 2420 { 2421 struct ifnet *ifp; /* my outgoing interface */ 2422 struct in6_addr *ifp_ll6; 2423 struct in6_addr *router_ll6; 2424 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2425 struct mbuf *m = NULL; /* newly allocated one */ 2426 struct m_tag *mtag; 2427 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2428 struct nd_redirect *nd_rd; 2429 struct llentry *ln = NULL; 2430 size_t maxlen; 2431 u_char *p; 2432 struct ifnet *outif = NULL; 2433 struct sockaddr_in6 src_sa; 2434 2435 icmp6_errcount(ND_REDIRECT, 0); 2436 2437 /* if we are not router, we don't send icmp6 redirect */ 2438 if (!V_ip6_forwarding) 2439 goto fail; 2440 2441 /* sanity check */ 2442 if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp)) 2443 goto fail; 2444 2445 /* 2446 * Address check: 2447 * the source address must identify a neighbor, and 2448 * the destination address must not be a multicast address 2449 * [RFC 2461, sec 8.2] 2450 */ 2451 sip6 = mtod(m0, struct ip6_hdr *); 2452 bzero(&src_sa, sizeof(src_sa)); 2453 src_sa.sin6_family = AF_INET6; 2454 src_sa.sin6_len = sizeof(src_sa); 2455 src_sa.sin6_addr = sip6->ip6_src; 2456 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2457 goto fail; 2458 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2459 goto fail; /* what should we do here? */ 2460 2461 /* rate limit */ 2462 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2463 goto fail; 2464 2465 /* 2466 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2467 * we almost always ask for an mbuf cluster for simplicity. 2468 * (MHLEN < IPV6_MMTU is almost always true) 2469 */ 2470 #if IPV6_MMTU >= MCLBYTES 2471 # error assumption failed about IPV6_MMTU and MCLBYTES 2472 #endif 2473 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 2474 if (m == NULL) 2475 goto fail; 2476 M_SETFIB(m, M_GETFIB(m0)); 2477 maxlen = M_TRAILINGSPACE(m); 2478 maxlen = min(IPV6_MMTU, maxlen); 2479 /* just for safety */ 2480 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2481 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2482 goto fail; 2483 } 2484 2485 { 2486 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2487 struct in6_ifaddr *ia; 2488 if ((ia = in6ifa_ifpforlinklocal(ifp, 2489 IN6_IFF_NOTREADY| 2490 IN6_IFF_ANYCAST)) == NULL) 2491 goto fail; 2492 ifp_ll6 = &ia->ia_addr.sin6_addr; 2493 /* XXXRW: reference released prematurely. */ 2494 ifa_free(&ia->ia_ifa); 2495 } 2496 2497 /* get ip6 linklocal address for the router. */ 2498 if (nh->nh_flags & NHF_GATEWAY) { 2499 struct sockaddr_in6 *sin6; 2500 sin6 = &nh->gw6_sa; 2501 router_ll6 = &sin6->sin6_addr; 2502 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6)) 2503 router_ll6 = (struct in6_addr *)NULL; 2504 } else 2505 router_ll6 = (struct in6_addr *)NULL; 2506 2507 /* ip6 */ 2508 ip6 = mtod(m, struct ip6_hdr *); 2509 ip6->ip6_flow = 0; 2510 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2511 ip6->ip6_vfc |= IPV6_VERSION; 2512 /* ip6->ip6_plen will be set later */ 2513 ip6->ip6_nxt = IPPROTO_ICMPV6; 2514 ip6->ip6_hlim = 255; 2515 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2516 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2517 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2518 2519 /* ND Redirect */ 2520 nd_rd = (struct nd_redirect *)(ip6 + 1); 2521 nd_rd->nd_rd_type = ND_REDIRECT; 2522 nd_rd->nd_rd_code = 0; 2523 nd_rd->nd_rd_reserved = 0; 2524 if (nh->nh_flags & NHF_GATEWAY) { 2525 /* 2526 * nd_rd->nd_rd_target must be a link-local address in 2527 * better router cases. 2528 */ 2529 if (!router_ll6) 2530 goto fail; 2531 bcopy(router_ll6, &nd_rd->nd_rd_target, 2532 sizeof(nd_rd->nd_rd_target)); 2533 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2534 sizeof(nd_rd->nd_rd_dst)); 2535 } else { 2536 /* make sure redtgt == reddst */ 2537 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2538 sizeof(nd_rd->nd_rd_target)); 2539 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2540 sizeof(nd_rd->nd_rd_dst)); 2541 } 2542 2543 p = (u_char *)(nd_rd + 1); 2544 2545 if (!router_ll6) 2546 goto nolladdropt; 2547 2548 { 2549 /* target lladdr option */ 2550 int len; 2551 struct nd_opt_hdr *nd_opt; 2552 char *lladdr; 2553 2554 ln = nd6_lookup(router_ll6, 0, ifp); 2555 if (ln == NULL) 2556 goto nolladdropt; 2557 2558 len = sizeof(*nd_opt) + ifp->if_addrlen; 2559 len = (len + 7) & ~7; /* round by 8 */ 2560 /* safety check */ 2561 if (len + (p - (u_char *)ip6) > maxlen) 2562 goto nolladdropt; 2563 2564 if (ln->la_flags & LLE_VALID) { 2565 nd_opt = (struct nd_opt_hdr *)p; 2566 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2567 nd_opt->nd_opt_len = len >> 3; 2568 lladdr = (char *)(nd_opt + 1); 2569 bcopy(ln->ll_addr, lladdr, ifp->if_addrlen); 2570 p += len; 2571 } 2572 } 2573 nolladdropt: 2574 if (ln != NULL) 2575 LLE_RUNLOCK(ln); 2576 2577 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2578 2579 /* just to be safe */ 2580 #ifdef M_DECRYPTED /*not openbsd*/ 2581 if (m0->m_flags & M_DECRYPTED) 2582 goto noredhdropt; 2583 #endif 2584 if (p - (u_char *)ip6 > maxlen) 2585 goto noredhdropt; 2586 2587 { 2588 /* redirected header option */ 2589 int len; 2590 struct nd_opt_rd_hdr *nd_opt_rh; 2591 2592 /* 2593 * compute the maximum size for icmp6 redirect header option. 2594 * XXX room for auth header? 2595 */ 2596 len = maxlen - (p - (u_char *)ip6); 2597 len &= ~7; 2598 2599 /* This is just for simplicity. */ 2600 if (m0->m_pkthdr.len != m0->m_len) { 2601 if (m0->m_next) { 2602 m_freem(m0->m_next); 2603 m0->m_next = NULL; 2604 } 2605 m0->m_pkthdr.len = m0->m_len; 2606 } 2607 2608 /* 2609 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2610 * about padding/truncate rule for the original IP packet. 2611 * From the discussion on IPv6imp in Feb 1999, 2612 * the consensus was: 2613 * - "attach as much as possible" is the goal 2614 * - pad if not aligned (original size can be guessed by 2615 * original ip6 header) 2616 * Following code adds the padding if it is simple enough, 2617 * and truncates if not. 2618 */ 2619 if (m0->m_next || m0->m_pkthdr.len != m0->m_len) 2620 panic("assumption failed in %s:%d", __FILE__, 2621 __LINE__); 2622 2623 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2624 /* not enough room, truncate */ 2625 m0->m_pkthdr.len = m0->m_len = len - 2626 sizeof(*nd_opt_rh); 2627 } else { 2628 /* enough room, pad or truncate */ 2629 size_t extra; 2630 2631 extra = m0->m_pkthdr.len % 8; 2632 if (extra) { 2633 /* pad if easy enough, truncate if not */ 2634 if (8 - extra <= M_TRAILINGSPACE(m0)) { 2635 /* pad */ 2636 m0->m_len += (8 - extra); 2637 m0->m_pkthdr.len += (8 - extra); 2638 } else { 2639 /* truncate */ 2640 m0->m_pkthdr.len -= extra; 2641 m0->m_len -= extra; 2642 } 2643 } 2644 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2645 m0->m_pkthdr.len = m0->m_len = len - 2646 sizeof(*nd_opt_rh); 2647 } 2648 2649 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2650 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2651 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2652 nd_opt_rh->nd_opt_rh_len = len >> 3; 2653 p += sizeof(*nd_opt_rh); 2654 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2655 2656 /* connect m0 to m */ 2657 m_tag_delete_chain(m0, NULL); 2658 m0->m_flags &= ~M_PKTHDR; 2659 m->m_next = m0; 2660 m->m_pkthdr.len = m->m_len + m0->m_len; 2661 m0 = NULL; 2662 } 2663 noredhdropt:; 2664 if (m0) { 2665 m_freem(m0); 2666 m0 = NULL; 2667 } 2668 2669 /* XXX: clear embedded link IDs in the inner header */ 2670 in6_clearscope(&sip6->ip6_src); 2671 in6_clearscope(&sip6->ip6_dst); 2672 in6_clearscope(&nd_rd->nd_rd_target); 2673 in6_clearscope(&nd_rd->nd_rd_dst); 2674 2675 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2676 2677 nd_rd->nd_rd_cksum = 0; 2678 nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2679 sizeof(*ip6), ntohs(ip6->ip6_plen)); 2680 2681 if (send_sendso_input_hook != NULL) { 2682 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short), 2683 M_NOWAIT); 2684 if (mtag == NULL) 2685 goto fail; 2686 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type; 2687 m_tag_prepend(m, mtag); 2688 } 2689 2690 /* send the packet to outside... */ 2691 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL); 2692 if (outif) { 2693 icmp6_ifstat_inc(outif, ifs6_out_msg); 2694 icmp6_ifstat_inc(outif, ifs6_out_redirect); 2695 } 2696 ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]); 2697 2698 return; 2699 2700 fail: 2701 if (m) 2702 m_freem(m); 2703 if (m0) 2704 m_freem(m0); 2705 } 2706 2707 /* 2708 * ICMPv6 socket option processing. 2709 */ 2710 int 2711 icmp6_ctloutput(struct socket *so, struct sockopt *sopt) 2712 { 2713 int error = 0; 2714 int optlen; 2715 struct inpcb *inp = sotoinpcb(so); 2716 int level, op, optname; 2717 2718 if (sopt) { 2719 level = sopt->sopt_level; 2720 op = sopt->sopt_dir; 2721 optname = sopt->sopt_name; 2722 optlen = sopt->sopt_valsize; 2723 } else 2724 level = op = optname = optlen = 0; 2725 2726 if (level != IPPROTO_ICMPV6) { 2727 return EINVAL; 2728 } 2729 2730 switch (op) { 2731 case PRCO_SETOPT: 2732 switch (optname) { 2733 case ICMP6_FILTER: 2734 { 2735 struct icmp6_filter ic6f; 2736 2737 if (optlen != sizeof(ic6f)) { 2738 error = EMSGSIZE; 2739 break; 2740 } 2741 error = sooptcopyin(sopt, &ic6f, optlen, optlen); 2742 if (error == 0) { 2743 INP_WLOCK(inp); 2744 *inp->in6p_icmp6filt = ic6f; 2745 INP_WUNLOCK(inp); 2746 } 2747 break; 2748 } 2749 2750 default: 2751 error = ENOPROTOOPT; 2752 break; 2753 } 2754 break; 2755 2756 case PRCO_GETOPT: 2757 switch (optname) { 2758 case ICMP6_FILTER: 2759 { 2760 struct icmp6_filter ic6f; 2761 2762 INP_RLOCK(inp); 2763 ic6f = *inp->in6p_icmp6filt; 2764 INP_RUNLOCK(inp); 2765 error = sooptcopyout(sopt, &ic6f, sizeof(ic6f)); 2766 break; 2767 } 2768 2769 default: 2770 error = ENOPROTOOPT; 2771 break; 2772 } 2773 break; 2774 } 2775 2776 return (error); 2777 } 2778 2779 /* 2780 * Perform rate limit check. 2781 * Returns 0 if it is okay to send the icmp6 packet. 2782 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2783 * limitation. 2784 * 2785 * XXX per-destination/type check necessary? 2786 * 2787 * dst - not used at this moment 2788 * type - not used at this moment 2789 * code - not used at this moment 2790 */ 2791 static int 2792 icmp6_ratelimit(const struct in6_addr *dst, const int type, 2793 const int code) 2794 { 2795 int ret; 2796 2797 ret = 0; /* okay to send */ 2798 2799 /* PPS limit */ 2800 if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count, 2801 V_icmp6errppslim)) { 2802 /* The packet is subject to rate limit */ 2803 ret++; 2804 } 2805 2806 return ret; 2807 } 2808