icmp6.c revision 1.129
1 /* $NetBSD: icmp6.c,v 1.129 2007/02/10 09:43:05 degroote Exp $ */ 2 /* $KAME: icmp6.c,v 1.217 2001/06/20 15:03:29 jinmei Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 62 */ 63 64 #include <sys/cdefs.h> 65 __KERNEL_RCSID(0, "$NetBSD: icmp6.c,v 1.129 2007/02/10 09:43:05 degroote Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_ipsec.h" 69 70 #include <sys/param.h> 71 #include <sys/systm.h> 72 #include <sys/malloc.h> 73 #include <sys/mbuf.h> 74 #include <sys/protosw.h> 75 #include <sys/socket.h> 76 #include <sys/socketvar.h> 77 #include <sys/time.h> 78 #include <sys/kernel.h> 79 #include <sys/syslog.h> 80 #include <sys/domain.h> 81 #include <sys/sysctl.h> 82 83 #include <net/if.h> 84 #include <net/route.h> 85 #include <net/if_dl.h> 86 #include <net/if_types.h> 87 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip6.h> 91 #include <netinet6/ip6_var.h> 92 #include <netinet/icmp6.h> 93 #include <netinet6/mld6_var.h> 94 #include <netinet6/in6_pcb.h> 95 #include <netinet6/nd6.h> 96 #include <netinet6/in6_ifattach.h> 97 #include <netinet6/ip6protosw.h> 98 #include <netinet6/scope6_var.h> 99 100 #ifdef IPSEC 101 #include <netinet6/ipsec.h> 102 #include <netkey/key.h> 103 #endif 104 105 #ifdef FAST_IPSEC 106 #include <netipsec/ipsec.h> 107 #include <netipsec/key.h> 108 #endif 109 110 111 #include "faith.h" 112 #if defined(NFAITH) && 0 < NFAITH 113 #include <net/if_faith.h> 114 #endif 115 116 #include <net/net_osdep.h> 117 118 extern struct domain inet6domain; 119 120 struct icmp6stat icmp6stat; 121 122 extern struct inpcbtable raw6cbtable; 123 extern int icmp6errppslim; 124 static int icmp6errpps_count = 0; 125 static struct timeval icmp6errppslim_last; 126 extern int icmp6_nodeinfo; 127 128 /* 129 * List of callbacks to notify when Path MTU changes are made. 130 */ 131 struct icmp6_mtudisc_callback { 132 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 133 void (*mc_func) __P((struct in6_addr *)); 134 }; 135 136 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 137 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 138 139 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 140 extern int pmtu_expire; 141 142 /* XXX do these values make any sense? */ 143 static int icmp6_mtudisc_hiwat = 1280; 144 static int icmp6_mtudisc_lowat = 256; 145 146 /* 147 * keep track of # of redirect routes. 148 */ 149 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 150 151 /* XXX experimental, turned off */ 152 static int icmp6_redirect_hiwat = -1; 153 static int icmp6_redirect_lowat = -1; 154 155 static void icmp6_errcount __P((struct icmp6errstat *, int, int)); 156 static int icmp6_rip6_input __P((struct mbuf **, int)); 157 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int)); 158 static const char *icmp6_redirect_diag __P((struct in6_addr *, 159 struct in6_addr *, struct in6_addr *)); 160 static struct mbuf *ni6_input __P((struct mbuf *, int)); 161 static struct mbuf *ni6_nametodns __P((const char *, int, int)); 162 static int ni6_dnsmatch __P((const char *, int, const char *, int)); 163 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *, 164 struct ifnet **, char *)); 165 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 166 struct ifnet *, int)); 167 static int icmp6_notify_error __P((struct mbuf *, int, int, int)); 168 static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *)); 169 static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *)); 170 static void icmp6_redirect_timeout __P((struct rtentry *, struct rttimer *)); 171 172 173 void 174 icmp6_init() 175 { 176 mld_init(); 177 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 178 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 179 } 180 181 static void 182 icmp6_errcount(stat, type, code) 183 struct icmp6errstat *stat; 184 int type, code; 185 { 186 switch (type) { 187 case ICMP6_DST_UNREACH: 188 switch (code) { 189 case ICMP6_DST_UNREACH_NOROUTE: 190 stat->icp6errs_dst_unreach_noroute++; 191 return; 192 case ICMP6_DST_UNREACH_ADMIN: 193 stat->icp6errs_dst_unreach_admin++; 194 return; 195 case ICMP6_DST_UNREACH_BEYONDSCOPE: 196 stat->icp6errs_dst_unreach_beyondscope++; 197 return; 198 case ICMP6_DST_UNREACH_ADDR: 199 stat->icp6errs_dst_unreach_addr++; 200 return; 201 case ICMP6_DST_UNREACH_NOPORT: 202 stat->icp6errs_dst_unreach_noport++; 203 return; 204 } 205 break; 206 case ICMP6_PACKET_TOO_BIG: 207 stat->icp6errs_packet_too_big++; 208 return; 209 case ICMP6_TIME_EXCEEDED: 210 switch (code) { 211 case ICMP6_TIME_EXCEED_TRANSIT: 212 stat->icp6errs_time_exceed_transit++; 213 return; 214 case ICMP6_TIME_EXCEED_REASSEMBLY: 215 stat->icp6errs_time_exceed_reassembly++; 216 return; 217 } 218 break; 219 case ICMP6_PARAM_PROB: 220 switch (code) { 221 case ICMP6_PARAMPROB_HEADER: 222 stat->icp6errs_paramprob_header++; 223 return; 224 case ICMP6_PARAMPROB_NEXTHEADER: 225 stat->icp6errs_paramprob_nextheader++; 226 return; 227 case ICMP6_PARAMPROB_OPTION: 228 stat->icp6errs_paramprob_option++; 229 return; 230 } 231 break; 232 case ND_REDIRECT: 233 stat->icp6errs_redirect++; 234 return; 235 } 236 stat->icp6errs_unknown++; 237 } 238 239 /* 240 * Register a Path MTU Discovery callback. 241 */ 242 void 243 icmp6_mtudisc_callback_register(func) 244 void (*func) __P((struct in6_addr *)); 245 { 246 struct icmp6_mtudisc_callback *mc; 247 248 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 249 mc = LIST_NEXT(mc, mc_list)) { 250 if (mc->mc_func == func) 251 return; 252 } 253 254 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 255 if (mc == NULL) 256 panic("icmp6_mtudisc_callback_register"); 257 258 mc->mc_func = func; 259 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 260 } 261 262 /* 263 * A wrapper function for icmp6_error() necessary when the erroneous packet 264 * may not contain enough scope zone information. 265 */ 266 void 267 icmp6_error2(m, type, code, param, ifp) 268 struct mbuf *m; 269 int type, code, param; 270 struct ifnet *ifp; 271 { 272 struct ip6_hdr *ip6; 273 274 if (ifp == NULL) 275 return; 276 277 if (m->m_len < sizeof(struct ip6_hdr)) { 278 m = m_pullup(m, sizeof(struct ip6_hdr)); 279 if (m == NULL) 280 return; 281 } 282 283 ip6 = mtod(m, struct ip6_hdr *); 284 285 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 286 return; 287 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 288 return; 289 290 icmp6_error(m, type, code, param); 291 } 292 293 /* 294 * Generate an error packet of type error in response to bad IP6 packet. 295 */ 296 void 297 icmp6_error(m, type, code, param) 298 struct mbuf *m; 299 int type, code, param; 300 { 301 struct ip6_hdr *oip6, *nip6; 302 struct icmp6_hdr *icmp6; 303 u_int preplen; 304 int off; 305 int nxt; 306 307 icmp6stat.icp6s_error++; 308 309 /* count per-type-code statistics */ 310 icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code); 311 312 if (m->m_flags & M_DECRYPTED) { 313 icmp6stat.icp6s_canterror++; 314 goto freeit; 315 } 316 317 if (m->m_len < sizeof(struct ip6_hdr)) { 318 m = m_pullup(m, sizeof(struct ip6_hdr)); 319 if (m == NULL) 320 return; 321 } 322 oip6 = mtod(m, struct ip6_hdr *); 323 324 /* 325 * If the destination address of the erroneous packet is a multicast 326 * address, or the packet was sent using link-layer multicast, 327 * we should basically suppress sending an error (RFC 2463, Section 328 * 2.4). 329 * We have two exceptions (the item e.2 in that section): 330 * - the Pakcet Too Big message can be sent for path MTU discovery. 331 * - the Parameter Problem Message that can be allowed an icmp6 error 332 * in the option type field. This check has been done in 333 * ip6_unknown_opt(), so we can just check the type and code. 334 */ 335 if ((m->m_flags & (M_BCAST|M_MCAST) || 336 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 337 (type != ICMP6_PACKET_TOO_BIG && 338 (type != ICMP6_PARAM_PROB || 339 code != ICMP6_PARAMPROB_OPTION))) 340 goto freeit; 341 342 /* 343 * RFC 2463, 2.4 (e.5): source address check. 344 * XXX: the case of anycast source? 345 */ 346 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 347 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 348 goto freeit; 349 350 /* 351 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 352 * don't do it. 353 */ 354 nxt = -1; 355 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 356 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 357 struct icmp6_hdr *icp; 358 359 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 360 sizeof(*icp)); 361 if (icp == NULL) { 362 icmp6stat.icp6s_tooshort++; 363 return; 364 } 365 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 366 icp->icmp6_type == ND_REDIRECT) { 367 /* 368 * ICMPv6 error 369 * Special case: for redirect (which is 370 * informational) we must not send icmp6 error. 371 */ 372 icmp6stat.icp6s_canterror++; 373 goto freeit; 374 } else { 375 /* ICMPv6 informational - send the error */ 376 } 377 } 378 #if 0 /* controversial */ 379 else if (off >= 0 && nxt == IPPROTO_ESP) { 380 /* 381 * It could be ICMPv6 error inside ESP. Take a safer side, 382 * don't respond. 383 */ 384 icmp6stat.icp6s_canterror++; 385 goto freeit; 386 } 387 #endif 388 else { 389 /* non-ICMPv6 - send the error */ 390 } 391 392 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 393 394 /* Finally, do rate limitation check. */ 395 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 396 icmp6stat.icp6s_toofreq++; 397 goto freeit; 398 } 399 400 /* 401 * OK, ICMP6 can be generated. 402 */ 403 404 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 405 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 406 407 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 408 M_PREPEND(m, preplen, M_DONTWAIT); 409 if (m && m->m_len < preplen) 410 m = m_pullup(m, preplen); 411 if (m == NULL) { 412 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 413 return; 414 } 415 416 nip6 = mtod(m, struct ip6_hdr *); 417 nip6->ip6_src = oip6->ip6_src; 418 nip6->ip6_dst = oip6->ip6_dst; 419 420 in6_clearscope(&oip6->ip6_src); 421 in6_clearscope(&oip6->ip6_dst); 422 423 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 424 icmp6->icmp6_type = type; 425 icmp6->icmp6_code = code; 426 icmp6->icmp6_pptr = htonl((u_int32_t)param); 427 428 /* 429 * icmp6_reflect() is designed to be in the input path. 430 * icmp6_error() can be called from both input and output path, 431 * and if we are in output path rcvif could contain bogus value. 432 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 433 * information in ip header (nip6). 434 */ 435 m->m_pkthdr.rcvif = NULL; 436 437 icmp6stat.icp6s_outhist[type]++; 438 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 439 440 return; 441 442 freeit: 443 /* 444 * If we can't tell whether or not we can generate ICMP6, free it. 445 */ 446 m_freem(m); 447 } 448 449 /* 450 * Process a received ICMP6 message. 451 */ 452 int 453 icmp6_input(struct mbuf **mp, int *offp, int proto) 454 { 455 struct mbuf *m = *mp, *n; 456 struct ip6_hdr *ip6, *nip6; 457 struct icmp6_hdr *icmp6, *nicmp6; 458 int off = *offp; 459 int icmp6len = m->m_pkthdr.len - *offp; 460 int code, sum, noff; 461 462 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4) 463 KASSERT(ICMP6_MAXLEN < MCLBYTES); 464 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 465 466 /* 467 * Locate icmp6 structure in mbuf, and check 468 * that not corrupted and of at least minimum length 469 */ 470 471 ip6 = mtod(m, struct ip6_hdr *); 472 if (icmp6len < sizeof(struct icmp6_hdr)) { 473 icmp6stat.icp6s_tooshort++; 474 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 475 goto freeit; 476 } 477 478 /* 479 * calculate the checksum 480 */ 481 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 482 if (icmp6 == NULL) { 483 icmp6stat.icp6s_tooshort++; 484 /* m is invalid */ 485 /*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/ 486 return IPPROTO_DONE; 487 } 488 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 489 code = icmp6->icmp6_code; 490 491 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 492 nd6log((LOG_ERR, 493 "ICMP6 checksum error(%d|%x) %s\n", 494 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 495 icmp6stat.icp6s_checksum++; 496 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 497 goto freeit; 498 } 499 500 #if defined(NFAITH) && 0 < NFAITH 501 if (faithprefix(&ip6->ip6_dst)) { 502 /* 503 * Deliver very specific ICMP6 type only. 504 * This is important to deliver TOOBIG. Otherwise PMTUD 505 * will not work. 506 */ 507 switch (icmp6->icmp6_type) { 508 case ICMP6_DST_UNREACH: 509 case ICMP6_PACKET_TOO_BIG: 510 case ICMP6_TIME_EXCEEDED: 511 break; 512 default: 513 goto freeit; 514 } 515 } 516 #endif 517 518 icmp6stat.icp6s_inhist[icmp6->icmp6_type]++; 519 520 switch (icmp6->icmp6_type) { 521 case ICMP6_DST_UNREACH: 522 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 523 switch (code) { 524 case ICMP6_DST_UNREACH_NOROUTE: 525 code = PRC_UNREACH_NET; 526 break; 527 case ICMP6_DST_UNREACH_ADMIN: 528 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 529 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 530 break; 531 case ICMP6_DST_UNREACH_ADDR: 532 code = PRC_HOSTDEAD; 533 break; 534 #ifdef COMPAT_RFC1885 535 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 536 code = PRC_UNREACH_SRCFAIL; 537 break; 538 #else 539 case ICMP6_DST_UNREACH_BEYONDSCOPE: 540 /* I mean "source address was incorrect." */ 541 code = PRC_UNREACH_NET; 542 break; 543 #endif 544 case ICMP6_DST_UNREACH_NOPORT: 545 code = PRC_UNREACH_PORT; 546 break; 547 default: 548 goto badcode; 549 } 550 goto deliver; 551 552 case ICMP6_PACKET_TOO_BIG: 553 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 554 555 code = PRC_MSGSIZE; 556 557 /* 558 * Updating the path MTU will be done after examining 559 * intermediate extension headers. 560 */ 561 goto deliver; 562 563 case ICMP6_TIME_EXCEEDED: 564 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 565 switch (code) { 566 case ICMP6_TIME_EXCEED_TRANSIT: 567 code = PRC_TIMXCEED_INTRANS; 568 break; 569 case ICMP6_TIME_EXCEED_REASSEMBLY: 570 code = PRC_TIMXCEED_REASS; 571 break; 572 default: 573 goto badcode; 574 } 575 goto deliver; 576 577 case ICMP6_PARAM_PROB: 578 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 579 switch (code) { 580 case ICMP6_PARAMPROB_NEXTHEADER: 581 code = PRC_UNREACH_PROTOCOL; 582 break; 583 case ICMP6_PARAMPROB_HEADER: 584 case ICMP6_PARAMPROB_OPTION: 585 code = PRC_PARAMPROB; 586 break; 587 default: 588 goto badcode; 589 } 590 goto deliver; 591 592 case ICMP6_ECHO_REQUEST: 593 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 594 if (code != 0) 595 goto badcode; 596 /* 597 * Copy mbuf to send to two data paths: userland socket(s), 598 * and to the querier (echo reply). 599 * m: a copy for socket, n: a copy for querier 600 * 601 * If the first mbuf is shared, or the first mbuf is too short, 602 * copy the first part of the data into a fresh mbuf. 603 * Otherwise, we will wrongly overwrite both copies. 604 */ 605 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 606 /* Give up local */ 607 n = m; 608 m = NULL; 609 } else if (M_READONLY(n) || 610 n->m_len < off + sizeof(struct icmp6_hdr)) { 611 struct mbuf *n0 = n; 612 613 /* 614 * Prepare an internal mbuf. m_pullup() doesn't 615 * always copy the length we specified. 616 */ 617 if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 618 /* Give up local */ 619 n = m; 620 m = NULL; 621 } 622 m_freem(n0); 623 } 624 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, 625 sizeof(*nicmp6)); 626 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 627 nicmp6->icmp6_code = 0; 628 if (n) { 629 icmp6stat.icp6s_reflect++; 630 icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++; 631 icmp6_reflect(n, off); 632 } 633 if (!m) 634 goto freeit; 635 break; 636 637 case ICMP6_ECHO_REPLY: 638 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 639 if (code != 0) 640 goto badcode; 641 break; 642 643 case MLD_LISTENER_QUERY: 644 case MLD_LISTENER_REPORT: 645 if (icmp6len < sizeof(struct mld_hdr)) 646 goto badlen; 647 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */ 648 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 649 else 650 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 651 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 652 /* give up local */ 653 mld_input(m, off); 654 m = NULL; 655 goto freeit; 656 } 657 mld_input(n, off); 658 /* m stays. */ 659 break; 660 661 case MLD_LISTENER_DONE: 662 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 663 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */ 664 goto badlen; 665 break; /* nothing to be done in kernel */ 666 667 case MLD_MTRACE_RESP: 668 case MLD_MTRACE: 669 /* XXX: these two are experimental. not officially defined. */ 670 /* XXX: per-interface statistics? */ 671 break; /* just pass it to applications */ 672 673 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 674 { 675 enum { WRU, FQDN } mode; 676 677 if (!icmp6_nodeinfo) 678 break; 679 680 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 681 mode = WRU; 682 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 683 mode = FQDN; 684 else 685 goto badlen; 686 687 if (mode == FQDN) { 688 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 689 if (n) 690 n = ni6_input(n, off); 691 /* XXX meaningless if n == NULL */ 692 noff = sizeof(struct ip6_hdr); 693 } else { 694 u_char *p; 695 int maxhlen; 696 697 if ((icmp6_nodeinfo & 5) != 5) 698 break; 699 700 if (code != 0) 701 goto badcode; 702 MGETHDR(n, M_DONTWAIT, m->m_type); 703 if (n && ICMP6_MAXLEN > MHLEN) { 704 MCLGET(n, M_DONTWAIT); 705 if ((n->m_flags & M_EXT) == 0) { 706 m_free(n); 707 n = NULL; 708 } 709 } 710 if (n == NULL) { 711 /* Give up remote */ 712 break; 713 } 714 n->m_pkthdr.rcvif = NULL; 715 n->m_len = 0; 716 maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN; 717 if (maxhlen > hostnamelen) 718 maxhlen = hostnamelen; 719 /* 720 * Copy IPv6 and ICMPv6 only. 721 */ 722 nip6 = mtod(n, struct ip6_hdr *); 723 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 724 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 725 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 726 p = (u_char *)(nicmp6 + 1); 727 bzero(p, 4); 728 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 729 noff = sizeof(struct ip6_hdr); 730 M_COPY_PKTHDR(n, m); /* just for rcvif */ 731 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 732 sizeof(struct icmp6_hdr) + 4 + maxhlen; 733 nicmp6->icmp6_type = ICMP6_WRUREPLY; 734 nicmp6->icmp6_code = 0; 735 } 736 #undef hostnamelen 737 if (n) { 738 icmp6stat.icp6s_reflect++; 739 icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++; 740 icmp6_reflect(n, noff); 741 } 742 break; 743 } 744 745 case ICMP6_WRUREPLY: 746 if (code != 0) 747 goto badcode; 748 break; 749 750 case ND_ROUTER_SOLICIT: 751 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 752 if (code != 0) 753 goto badcode; 754 if (icmp6len < sizeof(struct nd_router_solicit)) 755 goto badlen; 756 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 757 /* give up local */ 758 nd6_rs_input(m, off, icmp6len); 759 m = NULL; 760 goto freeit; 761 } 762 nd6_rs_input(n, off, icmp6len); 763 /* m stays. */ 764 break; 765 766 case ND_ROUTER_ADVERT: 767 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 768 if (code != 0) 769 goto badcode; 770 if (icmp6len < sizeof(struct nd_router_advert)) 771 goto badlen; 772 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 773 /* give up local */ 774 nd6_ra_input(m, off, icmp6len); 775 m = NULL; 776 goto freeit; 777 } 778 nd6_ra_input(n, off, icmp6len); 779 /* m stays. */ 780 break; 781 782 case ND_NEIGHBOR_SOLICIT: 783 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 784 if (code != 0) 785 goto badcode; 786 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 787 goto badlen; 788 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 789 /* give up local */ 790 nd6_ns_input(m, off, icmp6len); 791 m = NULL; 792 goto freeit; 793 } 794 nd6_ns_input(n, off, icmp6len); 795 /* m stays. */ 796 break; 797 798 case ND_NEIGHBOR_ADVERT: 799 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 800 if (code != 0) 801 goto badcode; 802 if (icmp6len < sizeof(struct nd_neighbor_advert)) 803 goto badlen; 804 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 805 /* give up local */ 806 nd6_na_input(m, off, icmp6len); 807 m = NULL; 808 goto freeit; 809 } 810 nd6_na_input(n, off, icmp6len); 811 /* m stays. */ 812 break; 813 814 case ND_REDIRECT: 815 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 816 if (code != 0) 817 goto badcode; 818 if (icmp6len < sizeof(struct nd_redirect)) 819 goto badlen; 820 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 821 /* give up local */ 822 icmp6_redirect_input(m, off); 823 m = NULL; 824 goto freeit; 825 } 826 icmp6_redirect_input(n, off); 827 /* m stays. */ 828 break; 829 830 case ICMP6_ROUTER_RENUMBERING: 831 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 832 code != ICMP6_ROUTER_RENUMBERING_RESULT) 833 goto badcode; 834 if (icmp6len < sizeof(struct icmp6_router_renum)) 835 goto badlen; 836 break; 837 838 default: 839 nd6log((LOG_DEBUG, 840 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 841 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 842 ip6_sprintf(&ip6->ip6_dst), 843 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 844 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 845 /* ICMPv6 error: MUST deliver it by spec... */ 846 code = PRC_NCMDS; 847 /* deliver */ 848 } else { 849 /* ICMPv6 informational: MUST not deliver */ 850 break; 851 } 852 deliver: 853 if (icmp6_notify_error(m, off, icmp6len, code)) { 854 /* In this case, m should've been freed. */ 855 return (IPPROTO_DONE); 856 } 857 break; 858 859 badcode: 860 icmp6stat.icp6s_badcode++; 861 break; 862 863 badlen: 864 icmp6stat.icp6s_badlen++; 865 break; 866 } 867 868 /* deliver the packet to appropriate sockets */ 869 icmp6_rip6_input(&m, *offp); 870 871 return IPPROTO_DONE; 872 873 freeit: 874 m_freem(m); 875 return IPPROTO_DONE; 876 } 877 878 static int 879 icmp6_notify_error(m, off, icmp6len, code) 880 struct mbuf *m; 881 int off, icmp6len, code; 882 { 883 struct icmp6_hdr *icmp6; 884 struct ip6_hdr *eip6; 885 u_int32_t notifymtu; 886 struct sockaddr_in6 icmp6src, icmp6dst; 887 888 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 889 icmp6stat.icp6s_tooshort++; 890 goto freeit; 891 } 892 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 893 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 894 if (icmp6 == NULL) { 895 icmp6stat.icp6s_tooshort++; 896 return (-1); 897 } 898 eip6 = (struct ip6_hdr *)(icmp6 + 1); 899 900 /* Detect the upper level protocol */ 901 { 902 void (*ctlfunc) __P((int, struct sockaddr *, void *)); 903 u_int8_t nxt = eip6->ip6_nxt; 904 int eoff = off + sizeof(struct icmp6_hdr) + 905 sizeof(struct ip6_hdr); 906 struct ip6ctlparam ip6cp; 907 struct in6_addr *finaldst = NULL; 908 int icmp6type = icmp6->icmp6_type; 909 struct ip6_frag *fh; 910 struct ip6_rthdr *rth; 911 struct ip6_rthdr0 *rth0; 912 int rthlen; 913 914 while (1) { /* XXX: should avoid infinite loop explicitly? */ 915 struct ip6_ext *eh; 916 917 switch (nxt) { 918 case IPPROTO_HOPOPTS: 919 case IPPROTO_DSTOPTS: 920 case IPPROTO_AH: 921 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 922 eoff, sizeof(*eh)); 923 if (eh == NULL) { 924 icmp6stat.icp6s_tooshort++; 925 return (-1); 926 } 927 928 if (nxt == IPPROTO_AH) 929 eoff += (eh->ip6e_len + 2) << 2; 930 else 931 eoff += (eh->ip6e_len + 1) << 3; 932 nxt = eh->ip6e_nxt; 933 break; 934 case IPPROTO_ROUTING: 935 /* 936 * When the erroneous packet contains a 937 * routing header, we should examine the 938 * header to determine the final destination. 939 * Otherwise, we can't properly update 940 * information that depends on the final 941 * destination (e.g. path MTU). 942 */ 943 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 944 eoff, sizeof(*rth)); 945 if (rth == NULL) { 946 icmp6stat.icp6s_tooshort++; 947 return (-1); 948 } 949 rthlen = (rth->ip6r_len + 1) << 3; 950 /* 951 * XXX: currently there is no 952 * officially defined type other 953 * than type-0. 954 * Note that if the segment left field 955 * is 0, all intermediate hops must 956 * have been passed. 957 */ 958 if (rth->ip6r_segleft && 959 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 960 int hops; 961 962 IP6_EXTHDR_GET(rth0, 963 struct ip6_rthdr0 *, m, 964 eoff, rthlen); 965 if (rth0 == NULL) { 966 icmp6stat.icp6s_tooshort++; 967 return (-1); 968 } 969 /* just ignore a bogus header */ 970 if ((rth0->ip6r0_len % 2) == 0 && 971 (hops = rth0->ip6r0_len/2)) 972 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 973 } 974 eoff += rthlen; 975 nxt = rth->ip6r_nxt; 976 break; 977 case IPPROTO_FRAGMENT: 978 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 979 eoff, sizeof(*fh)); 980 if (fh == NULL) { 981 icmp6stat.icp6s_tooshort++; 982 return (-1); 983 } 984 /* 985 * Data after a fragment header is meaningless 986 * unless it is the first fragment, but 987 * we'll go to the notify label for path MTU 988 * discovery. 989 */ 990 if (fh->ip6f_offlg & IP6F_OFF_MASK) 991 goto notify; 992 993 eoff += sizeof(struct ip6_frag); 994 nxt = fh->ip6f_nxt; 995 break; 996 default: 997 /* 998 * This case includes ESP and the No Next 999 * Header. In such cases going to the notify 1000 * label does not have any meaning 1001 * (i.e. ctlfunc will be NULL), but we go 1002 * anyway since we might have to update 1003 * path MTU information. 1004 */ 1005 goto notify; 1006 } 1007 } 1008 notify: 1009 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1010 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1011 if (icmp6 == NULL) { 1012 icmp6stat.icp6s_tooshort++; 1013 return (-1); 1014 } 1015 1016 /* 1017 * retrieve parameters from the inner IPv6 header, and convert 1018 * them into sockaddr structures. 1019 * XXX: there is no guarantee that the source or destination 1020 * addresses of the inner packet are in the same scope zone as 1021 * the addresses of the icmp packet. But there is no other 1022 * way to determine the zone. 1023 */ 1024 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1025 1026 bzero(&icmp6dst, sizeof(icmp6dst)); 1027 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1028 icmp6dst.sin6_family = AF_INET6; 1029 if (finaldst == NULL) 1030 icmp6dst.sin6_addr = eip6->ip6_dst; 1031 else 1032 icmp6dst.sin6_addr = *finaldst; 1033 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1034 goto freeit; 1035 bzero(&icmp6src, sizeof(icmp6src)); 1036 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1037 icmp6src.sin6_family = AF_INET6; 1038 icmp6src.sin6_addr = eip6->ip6_src; 1039 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1040 goto freeit; 1041 icmp6src.sin6_flowinfo = 1042 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1043 1044 if (finaldst == NULL) 1045 finaldst = &eip6->ip6_dst; 1046 ip6cp.ip6c_m = m; 1047 ip6cp.ip6c_icmp6 = icmp6; 1048 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1049 ip6cp.ip6c_off = eoff; 1050 ip6cp.ip6c_finaldst = finaldst; 1051 ip6cp.ip6c_src = &icmp6src; 1052 ip6cp.ip6c_nxt = nxt; 1053 1054 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1055 notifymtu = ntohl(icmp6->icmp6_mtu); 1056 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1057 } 1058 1059 ctlfunc = (void (*) __P((int, struct sockaddr *, void *))) 1060 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1061 if (ctlfunc) { 1062 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1063 &ip6cp); 1064 } 1065 } 1066 return (0); 1067 1068 freeit: 1069 m_freem(m); 1070 return (-1); 1071 } 1072 1073 void 1074 icmp6_mtudisc_update(ip6cp, validated) 1075 struct ip6ctlparam *ip6cp; 1076 int validated; 1077 { 1078 unsigned long rtcount; 1079 struct icmp6_mtudisc_callback *mc; 1080 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1081 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1082 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1083 u_int mtu = ntohl(icmp6->icmp6_mtu); 1084 struct rtentry *rt = NULL; 1085 struct sockaddr_in6 sin6; 1086 1087 /* 1088 * allow non-validated cases if memory is plenty, to make traffic 1089 * from non-connected pcb happy. 1090 */ 1091 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1092 if (validated) { 1093 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1094 return; 1095 else if (0 <= icmp6_mtudisc_lowat && 1096 rtcount > icmp6_mtudisc_lowat) { 1097 /* 1098 * XXX nuke a victim, install the new one. 1099 */ 1100 } 1101 } else { 1102 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1103 return; 1104 } 1105 1106 bzero(&sin6, sizeof(sin6)); 1107 sin6.sin6_family = PF_INET6; 1108 sin6.sin6_len = sizeof(struct sockaddr_in6); 1109 sin6.sin6_addr = *dst; 1110 if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1111 return; 1112 1113 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1114 1115 if (rt && (rt->rt_flags & RTF_HOST) && 1116 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1117 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1118 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1119 icmp6stat.icp6s_pmtuchg++; 1120 rt->rt_rmx.rmx_mtu = mtu; 1121 } 1122 } 1123 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */ 1124 RTFREE(rt); 1125 } 1126 1127 /* 1128 * Notify protocols that the MTU for this destination 1129 * has changed. 1130 */ 1131 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1132 mc = LIST_NEXT(mc, mc_list)) 1133 (*mc->mc_func)(&sin6.sin6_addr); 1134 } 1135 1136 /* 1137 * Process a Node Information Query packet, based on 1138 * draft-ietf-ipngwg-icmp-name-lookups-07. 1139 * 1140 * Spec incompatibilities: 1141 * - IPv6 Subject address handling 1142 * - IPv4 Subject address handling support missing 1143 * - Proxy reply (answer even if it's not for me) 1144 * - joins NI group address at in6_ifattach() time only, does not cope 1145 * with hostname changes by sethostname(3) 1146 */ 1147 #ifndef offsetof /* XXX */ 1148 #define offsetof(type, member) ((size_t)(&((type *)0)->member)) 1149 #endif 1150 static struct mbuf * 1151 ni6_input(m, off) 1152 struct mbuf *m; 1153 int off; 1154 { 1155 struct icmp6_nodeinfo *ni6, *nni6; 1156 struct mbuf *n = NULL; 1157 u_int16_t qtype; 1158 int subjlen; 1159 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1160 struct ni_reply_fqdn *fqdn; 1161 int addrs; /* for NI_QTYPE_NODEADDR */ 1162 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1163 struct sockaddr_in6 sin6; /* ip6_dst */ 1164 struct in6_addr in6_subj; /* subject address */ 1165 struct ip6_hdr *ip6; 1166 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1167 char *subj = NULL; 1168 1169 ip6 = mtod(m, struct ip6_hdr *); 1170 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1171 if (ni6 == NULL) { 1172 /* m is already reclaimed */ 1173 return NULL; 1174 } 1175 1176 /* 1177 * Validate IPv6 destination address. 1178 * 1179 * The Responder must discard the Query without further processing 1180 * unless it is one of the Responder's unicast or anycast addresses, or 1181 * a link-local scope multicast address which the Responder has joined. 1182 * [icmp-name-lookups-07, Section 4.] 1183 */ 1184 bzero(&sin6, sizeof(sin6)); 1185 sin6.sin6_family = AF_INET6; 1186 sin6.sin6_len = sizeof(struct sockaddr_in6); 1187 bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr)); 1188 /* XXX scopeid */ 1189 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1190 ; /* unicast/anycast, fine */ 1191 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1192 ; /* link-local multicast, fine */ 1193 else 1194 goto bad; 1195 1196 /* validate query Subject field. */ 1197 qtype = ntohs(ni6->ni_qtype); 1198 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1199 switch (qtype) { 1200 case NI_QTYPE_NOOP: 1201 case NI_QTYPE_SUPTYPES: 1202 /* 07 draft */ 1203 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1204 break; 1205 /* FALLTHROUGH */ 1206 case NI_QTYPE_FQDN: 1207 case NI_QTYPE_NODEADDR: 1208 case NI_QTYPE_IPV4ADDR: 1209 switch (ni6->ni_code) { 1210 case ICMP6_NI_SUBJ_IPV6: 1211 #if ICMP6_NI_SUBJ_IPV6 != 0 1212 case 0: 1213 #endif 1214 /* 1215 * backward compatibility - try to accept 03 draft 1216 * format, where no Subject is present. 1217 */ 1218 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1219 subjlen == 0) { 1220 oldfqdn++; 1221 break; 1222 } 1223 #if ICMP6_NI_SUBJ_IPV6 != 0 1224 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1225 goto bad; 1226 #endif 1227 1228 if (subjlen != sizeof(sin6.sin6_addr)) 1229 goto bad; 1230 1231 /* 1232 * Validate Subject address. 1233 * 1234 * Not sure what exactly "address belongs to the node" 1235 * means in the spec, is it just unicast, or what? 1236 * 1237 * At this moment we consider Subject address as 1238 * "belong to the node" if the Subject address equals 1239 * to the IPv6 destination address; validation for 1240 * IPv6 destination address should have done enough 1241 * check for us. 1242 * 1243 * We do not do proxy at this moment. 1244 */ 1245 /* m_pulldown instead of copy? */ 1246 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1247 subjlen, (caddr_t)&in6_subj); 1248 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1249 goto bad; 1250 1251 subj = (char *)&in6_subj; 1252 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1253 break; 1254 1255 /* 1256 * XXX if we are to allow other cases, we should really 1257 * be careful about scope here. 1258 * basically, we should disallow queries toward IPv6 1259 * destination X with subject Y, if scope(X) > scope(Y). 1260 * if we allow scope(X) > scope(Y), it will result in 1261 * information leakage across scope boundary. 1262 */ 1263 goto bad; 1264 1265 case ICMP6_NI_SUBJ_FQDN: 1266 /* 1267 * Validate Subject name with gethostname(3). 1268 * 1269 * The behavior may need some debate, since: 1270 * - we are not sure if the node has FQDN as 1271 * hostname (returned by gethostname(3)). 1272 * - the code does wildcard match for truncated names. 1273 * however, we are not sure if we want to perform 1274 * wildcard match, if gethostname(3) side has 1275 * truncated hostname. 1276 */ 1277 n = ni6_nametodns(hostname, hostnamelen, 0); 1278 if (!n || n->m_next || n->m_len == 0) 1279 goto bad; 1280 IP6_EXTHDR_GET(subj, char *, m, 1281 off + sizeof(struct icmp6_nodeinfo), subjlen); 1282 if (subj == NULL) 1283 goto bad; 1284 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1285 n->m_len)) { 1286 goto bad; 1287 } 1288 m_freem(n); 1289 n = NULL; 1290 break; 1291 1292 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1293 default: 1294 goto bad; 1295 } 1296 break; 1297 } 1298 1299 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1300 switch (qtype) { 1301 case NI_QTYPE_FQDN: 1302 if ((icmp6_nodeinfo & 1) == 0) 1303 goto bad; 1304 break; 1305 case NI_QTYPE_NODEADDR: 1306 case NI_QTYPE_IPV4ADDR: 1307 if ((icmp6_nodeinfo & 2) == 0) 1308 goto bad; 1309 break; 1310 } 1311 1312 /* guess reply length */ 1313 switch (qtype) { 1314 case NI_QTYPE_NOOP: 1315 break; /* no reply data */ 1316 case NI_QTYPE_SUPTYPES: 1317 replylen += sizeof(u_int32_t); 1318 break; 1319 case NI_QTYPE_FQDN: 1320 /* XXX will append an mbuf */ 1321 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1322 break; 1323 case NI_QTYPE_NODEADDR: 1324 addrs = ni6_addrs(ni6, m, &ifp, subj); 1325 if ((replylen += addrs * (sizeof(struct in6_addr) + 1326 sizeof(u_int32_t))) > MCLBYTES) 1327 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1328 break; 1329 case NI_QTYPE_IPV4ADDR: 1330 /* unsupported - should respond with unknown Qtype? */ 1331 goto bad; 1332 default: 1333 /* 1334 * XXX: We must return a reply with the ICMP6 code 1335 * `unknown Qtype' in this case. However we regard the case 1336 * as an FQDN query for backward compatibility. 1337 * Older versions set a random value to this field, 1338 * so it rarely varies in the defined qtypes. 1339 * But the mechanism is not reliable... 1340 * maybe we should obsolete older versions. 1341 */ 1342 qtype = NI_QTYPE_FQDN; 1343 /* XXX will append an mbuf */ 1344 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1345 oldfqdn++; 1346 break; 1347 } 1348 1349 /* allocate an mbuf to reply. */ 1350 MGETHDR(n, M_DONTWAIT, m->m_type); 1351 if (n == NULL) { 1352 m_freem(m); 1353 return (NULL); 1354 } 1355 M_MOVE_PKTHDR(n, m); /* just for rcvif */ 1356 if (replylen > MHLEN) { 1357 if (replylen > MCLBYTES) { 1358 /* 1359 * XXX: should we try to allocate more? But MCLBYTES 1360 * is probably much larger than IPV6_MMTU... 1361 */ 1362 goto bad; 1363 } 1364 MCLGET(n, M_DONTWAIT); 1365 if ((n->m_flags & M_EXT) == 0) { 1366 goto bad; 1367 } 1368 } 1369 n->m_pkthdr.len = n->m_len = replylen; 1370 1371 /* copy mbuf header and IPv6 + Node Information base headers */ 1372 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1373 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1374 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1375 1376 /* qtype dependent procedure */ 1377 switch (qtype) { 1378 case NI_QTYPE_NOOP: 1379 nni6->ni_code = ICMP6_NI_SUCCESS; 1380 nni6->ni_flags = 0; 1381 break; 1382 case NI_QTYPE_SUPTYPES: 1383 { 1384 u_int32_t v; 1385 nni6->ni_code = ICMP6_NI_SUCCESS; 1386 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1387 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1388 v = (u_int32_t)htonl(0x0000000f); 1389 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1390 break; 1391 } 1392 case NI_QTYPE_FQDN: 1393 nni6->ni_code = ICMP6_NI_SUCCESS; 1394 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1395 sizeof(struct ip6_hdr) + 1396 sizeof(struct icmp6_nodeinfo)); 1397 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1398 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1399 /* 1400 * XXX do we really have FQDN in variable "hostname"? 1401 */ 1402 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1403 if (n->m_next == NULL) 1404 goto bad; 1405 /* XXX we assume that n->m_next is not a chain */ 1406 if (n->m_next->m_next != NULL) 1407 goto bad; 1408 n->m_pkthdr.len += n->m_next->m_len; 1409 break; 1410 case NI_QTYPE_NODEADDR: 1411 { 1412 int lenlim, copied; 1413 1414 nni6->ni_code = ICMP6_NI_SUCCESS; 1415 n->m_pkthdr.len = n->m_len = 1416 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1417 lenlim = M_TRAILINGSPACE(n); 1418 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1419 /* XXX: reset mbuf length */ 1420 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1421 sizeof(struct icmp6_nodeinfo) + copied; 1422 break; 1423 } 1424 default: 1425 break; /* XXX impossible! */ 1426 } 1427 1428 nni6->ni_type = ICMP6_NI_REPLY; 1429 m_freem(m); 1430 return (n); 1431 1432 bad: 1433 m_freem(m); 1434 if (n) 1435 m_freem(n); 1436 return (NULL); 1437 } 1438 #undef hostnamelen 1439 1440 #define isupper(x) ('A' <= (x) && (x) <= 'Z') 1441 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1442 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1443 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1444 1445 /* 1446 * make a mbuf with DNS-encoded string. no compression support. 1447 * 1448 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1449 * treated as truncated name (two \0 at the end). this is a wild guess. 1450 */ 1451 static struct mbuf * 1452 ni6_nametodns(name, namelen, old) 1453 const char *name; 1454 int namelen; 1455 int old; /* return pascal string if non-zero */ 1456 { 1457 struct mbuf *m; 1458 char *cp, *ep; 1459 const char *p, *q; 1460 int i, len, nterm; 1461 1462 if (old) 1463 len = namelen + 1; 1464 else 1465 len = MCLBYTES; 1466 1467 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1468 MGET(m, M_DONTWAIT, MT_DATA); 1469 if (m && len > MLEN) { 1470 MCLGET(m, M_DONTWAIT); 1471 if ((m->m_flags & M_EXT) == 0) 1472 goto fail; 1473 } 1474 if (!m) 1475 goto fail; 1476 m->m_next = NULL; 1477 1478 if (old) { 1479 m->m_len = len; 1480 *mtod(m, char *) = namelen; 1481 bcopy(name, mtod(m, char *) + 1, namelen); 1482 return m; 1483 } else { 1484 m->m_len = 0; 1485 cp = mtod(m, char *); 1486 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1487 1488 /* if not certain about my name, return empty buffer */ 1489 if (namelen == 0) 1490 return m; 1491 1492 /* 1493 * guess if it looks like shortened hostname, or FQDN. 1494 * shortened hostname needs two trailing "\0". 1495 */ 1496 i = 0; 1497 for (p = name; p < name + namelen; p++) { 1498 if (*p && *p == '.') 1499 i++; 1500 } 1501 if (i < 2) 1502 nterm = 2; 1503 else 1504 nterm = 1; 1505 1506 p = name; 1507 while (cp < ep && p < name + namelen) { 1508 i = 0; 1509 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1510 i++; 1511 /* result does not fit into mbuf */ 1512 if (cp + i + 1 >= ep) 1513 goto fail; 1514 /* 1515 * DNS label length restriction, RFC1035 page 8. 1516 * "i == 0" case is included here to avoid returning 1517 * 0-length label on "foo..bar". 1518 */ 1519 if (i <= 0 || i >= 64) 1520 goto fail; 1521 *cp++ = i; 1522 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1523 goto fail; 1524 while (i > 0) { 1525 if (!isalnum(*p) && *p != '-') 1526 goto fail; 1527 if (isupper(*p)) { 1528 *cp++ = tolower(*p); 1529 p++; 1530 } else 1531 *cp++ = *p++; 1532 i--; 1533 } 1534 p = q; 1535 if (p < name + namelen && *p == '.') 1536 p++; 1537 } 1538 /* termination */ 1539 if (cp + nterm >= ep) 1540 goto fail; 1541 while (nterm-- > 0) 1542 *cp++ = '\0'; 1543 m->m_len = cp - mtod(m, char *); 1544 return m; 1545 } 1546 1547 panic("should not reach here"); 1548 /* NOTREACHED */ 1549 1550 fail: 1551 if (m) 1552 m_freem(m); 1553 return NULL; 1554 } 1555 1556 /* 1557 * check if two DNS-encoded string matches. takes care of truncated 1558 * form (with \0\0 at the end). no compression support. 1559 * XXX upper/lowercase match (see RFC2065) 1560 */ 1561 static int 1562 ni6_dnsmatch(a, alen, b, blen) 1563 const char *a; 1564 int alen; 1565 const char *b; 1566 int blen; 1567 { 1568 const char *a0, *b0; 1569 int l; 1570 1571 /* simplest case - need validation? */ 1572 if (alen == blen && bcmp(a, b, alen) == 0) 1573 return 1; 1574 1575 a0 = a; 1576 b0 = b; 1577 1578 /* termination is mandatory */ 1579 if (alen < 2 || blen < 2) 1580 return 0; 1581 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1582 return 0; 1583 alen--; 1584 blen--; 1585 1586 while (a - a0 < alen && b - b0 < blen) { 1587 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1588 return 0; 1589 1590 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1591 return 0; 1592 /* we don't support compression yet */ 1593 if (a[0] >= 64 || b[0] >= 64) 1594 return 0; 1595 1596 /* truncated case */ 1597 if (a[0] == 0 && a - a0 == alen - 1) 1598 return 1; 1599 if (b[0] == 0 && b - b0 == blen - 1) 1600 return 1; 1601 if (a[0] == 0 || b[0] == 0) 1602 return 0; 1603 1604 if (a[0] != b[0]) 1605 return 0; 1606 l = a[0]; 1607 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1608 return 0; 1609 if (bcmp(a + 1, b + 1, l) != 0) 1610 return 0; 1611 1612 a += 1 + l; 1613 b += 1 + l; 1614 } 1615 1616 if (a - a0 == alen && b - b0 == blen) 1617 return 1; 1618 else 1619 return 0; 1620 } 1621 1622 /* 1623 * calculate the number of addresses to be returned in the node info reply. 1624 */ 1625 static int 1626 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, 1627 struct ifnet **ifpp, char *subj) 1628 { 1629 struct ifnet *ifp; 1630 struct in6_ifaddr *ifa6; 1631 struct ifaddr *ifa; 1632 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1633 int addrs = 0, addrsofif, iffound = 0; 1634 int niflags = ni6->ni_flags; 1635 1636 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1637 switch (ni6->ni_code) { 1638 case ICMP6_NI_SUBJ_IPV6: 1639 if (subj == NULL) /* must be impossible... */ 1640 return (0); 1641 subj_ip6 = (struct sockaddr_in6 *)subj; 1642 break; 1643 default: 1644 /* 1645 * XXX: we only support IPv6 subject address for 1646 * this Qtype. 1647 */ 1648 return (0); 1649 } 1650 } 1651 1652 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1653 { 1654 addrsofif = 0; 1655 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1656 ifa = ifa->ifa_list.tqe_next) 1657 { 1658 if (ifa->ifa_addr->sa_family != AF_INET6) 1659 continue; 1660 ifa6 = (struct in6_ifaddr *)ifa; 1661 1662 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1663 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1664 &ifa6->ia_addr.sin6_addr)) 1665 iffound = 1; 1666 1667 /* 1668 * IPv4-mapped addresses can only be returned by a 1669 * Node Information proxy, since they represent 1670 * addresses of IPv4-only nodes, which perforce do 1671 * not implement this protocol. 1672 * [icmp-name-lookups-07, Section 5.4] 1673 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1674 * this function at this moment. 1675 */ 1676 1677 /* What do we have to do about ::1? */ 1678 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1679 case IPV6_ADDR_SCOPE_LINKLOCAL: 1680 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1681 continue; 1682 break; 1683 case IPV6_ADDR_SCOPE_SITELOCAL: 1684 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1685 continue; 1686 break; 1687 case IPV6_ADDR_SCOPE_GLOBAL: 1688 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1689 continue; 1690 break; 1691 default: 1692 continue; 1693 } 1694 1695 /* 1696 * check if anycast is okay. 1697 * XXX: just experimental. not in the spec. 1698 */ 1699 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1700 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1701 continue; /* we need only unicast addresses */ 1702 1703 addrsofif++; /* count the address */ 1704 } 1705 if (iffound) { 1706 *ifpp = ifp; 1707 return (addrsofif); 1708 } 1709 1710 addrs += addrsofif; 1711 } 1712 1713 return (addrs); 1714 } 1715 1716 static int 1717 ni6_store_addrs(ni6, nni6, ifp0, resid) 1718 struct icmp6_nodeinfo *ni6, *nni6; 1719 struct ifnet *ifp0; 1720 int resid; 1721 { 1722 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1723 struct in6_ifaddr *ifa6; 1724 struct ifaddr *ifa; 1725 struct ifnet *ifp_dep = NULL; 1726 int copied = 0, allow_deprecated = 0; 1727 u_char *cp = (u_char *)(nni6 + 1); 1728 int niflags = ni6->ni_flags; 1729 u_int32_t ltime; 1730 1731 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1732 return (0); /* needless to copy */ 1733 1734 again: 1735 1736 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1737 { 1738 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1739 ifa = ifa->ifa_list.tqe_next) 1740 { 1741 if (ifa->ifa_addr->sa_family != AF_INET6) 1742 continue; 1743 ifa6 = (struct in6_ifaddr *)ifa; 1744 1745 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1746 allow_deprecated == 0) { 1747 /* 1748 * prefererred address should be put before 1749 * deprecated addresses. 1750 */ 1751 1752 /* record the interface for later search */ 1753 if (ifp_dep == NULL) 1754 ifp_dep = ifp; 1755 1756 continue; 1757 } 1758 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1759 allow_deprecated != 0) 1760 continue; /* we now collect deprecated addrs */ 1761 1762 /* What do we have to do about ::1? */ 1763 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1764 case IPV6_ADDR_SCOPE_LINKLOCAL: 1765 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1766 continue; 1767 break; 1768 case IPV6_ADDR_SCOPE_SITELOCAL: 1769 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1770 continue; 1771 break; 1772 case IPV6_ADDR_SCOPE_GLOBAL: 1773 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1774 continue; 1775 break; 1776 default: 1777 continue; 1778 } 1779 1780 /* 1781 * check if anycast is okay. 1782 * XXX: just experimental. not in the spec. 1783 */ 1784 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1785 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1786 continue; 1787 1788 /* now we can copy the address */ 1789 if (resid < sizeof(struct in6_addr) + 1790 sizeof(u_int32_t)) { 1791 /* 1792 * We give up much more copy. 1793 * Set the truncate flag and return. 1794 */ 1795 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1796 return (copied); 1797 } 1798 1799 /* 1800 * Set the TTL of the address. 1801 * The TTL value should be one of the following 1802 * according to the specification: 1803 * 1804 * 1. The remaining lifetime of a DHCP lease on the 1805 * address, or 1806 * 2. The remaining Valid Lifetime of a prefix from 1807 * which the address was derived through Stateless 1808 * Autoconfiguration. 1809 * 1810 * Note that we currently do not support stateful 1811 * address configuration by DHCPv6, so the former 1812 * case can't happen. 1813 * 1814 * TTL must be 2^31 > TTL >= 0. 1815 */ 1816 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1817 ltime = ND6_INFINITE_LIFETIME; 1818 else { 1819 if (ifa6->ia6_lifetime.ia6t_expire > 1820 time_second) 1821 ltime = ifa6->ia6_lifetime.ia6t_expire - 1822 time_second; 1823 else 1824 ltime = 0; 1825 } 1826 if (ltime > 0x7fffffff) 1827 ltime = 0x7fffffff; 1828 ltime = htonl(ltime); 1829 1830 bcopy(<ime, cp, sizeof(u_int32_t)); 1831 cp += sizeof(u_int32_t); 1832 1833 /* copy the address itself */ 1834 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1835 sizeof(struct in6_addr)); 1836 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1837 cp += sizeof(struct in6_addr); 1838 1839 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1840 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1841 } 1842 if (ifp0) /* we need search only on the specified IF */ 1843 break; 1844 } 1845 1846 if (allow_deprecated == 0 && ifp_dep != NULL) { 1847 ifp = ifp_dep; 1848 allow_deprecated = 1; 1849 1850 goto again; 1851 } 1852 1853 return (copied); 1854 } 1855 1856 /* 1857 * XXX almost dup'ed code with rip6_input. 1858 */ 1859 static int 1860 icmp6_rip6_input(mp, off) 1861 struct mbuf **mp; 1862 int off; 1863 { 1864 struct mbuf *m = *mp; 1865 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1866 struct inpcb_hdr *inph; 1867 struct in6pcb *in6p; 1868 struct in6pcb *last = NULL; 1869 struct sockaddr_in6 rip6src; 1870 struct icmp6_hdr *icmp6; 1871 struct mbuf *opts = NULL; 1872 1873 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1874 if (icmp6 == NULL) { 1875 /* m is already reclaimed */ 1876 return IPPROTO_DONE; 1877 } 1878 1879 /* 1880 * XXX: the address may have embedded scope zone ID, which should be 1881 * hidden from applications. 1882 */ 1883 bzero(&rip6src, sizeof(rip6src)); 1884 rip6src.sin6_len = sizeof(struct sockaddr_in6); 1885 rip6src.sin6_family = AF_INET6; 1886 rip6src.sin6_addr = ip6->ip6_src; 1887 if (sa6_recoverscope(&rip6src)) { 1888 m_freem(m); 1889 return (IPPROTO_DONE); 1890 } 1891 1892 CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) { 1893 in6p = (struct in6pcb *)inph; 1894 if (in6p->in6p_af != AF_INET6) 1895 continue; 1896 if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6) 1897 continue; 1898 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1899 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1900 continue; 1901 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1902 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1903 continue; 1904 if (in6p->in6p_icmp6filt 1905 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1906 in6p->in6p_icmp6filt)) 1907 continue; 1908 if (last) { 1909 struct mbuf *n; 1910 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1911 if (last->in6p_flags & IN6P_CONTROLOPTS) 1912 ip6_savecontrol(last, &opts, ip6, n); 1913 /* strip intermediate headers */ 1914 m_adj(n, off); 1915 if (sbappendaddr(&last->in6p_socket->so_rcv, 1916 (struct sockaddr *)&rip6src, 1917 n, opts) == 0) { 1918 /* should notify about lost packet */ 1919 m_freem(n); 1920 if (opts) 1921 m_freem(opts); 1922 } else 1923 sorwakeup(last->in6p_socket); 1924 opts = NULL; 1925 } 1926 } 1927 last = in6p; 1928 } 1929 if (last) { 1930 if (last->in6p_flags & IN6P_CONTROLOPTS) 1931 ip6_savecontrol(last, &opts, ip6, m); 1932 /* strip intermediate headers */ 1933 m_adj(m, off); 1934 if (sbappendaddr(&last->in6p_socket->so_rcv, 1935 (struct sockaddr *)&rip6src, m, opts) == 0) { 1936 m_freem(m); 1937 if (opts) 1938 m_freem(opts); 1939 } else 1940 sorwakeup(last->in6p_socket); 1941 } else { 1942 m_freem(m); 1943 ip6stat.ip6s_delivered--; 1944 } 1945 return IPPROTO_DONE; 1946 } 1947 1948 /* 1949 * Reflect the ip6 packet back to the source. 1950 * OFF points to the icmp6 header, counted from the top of the mbuf. 1951 * 1952 * Note: RFC 1885 required that an echo reply should be truncated if it 1953 * did not fit in with (return) path MTU, and KAME code supported the 1954 * behavior. However, as a clarification after the RFC, this limitation 1955 * was removed in a revised version of the spec, RFC 2463. We had kept the 1956 * old behavior, with a (non-default) ifdef block, while the new version of 1957 * the spec was an internet-draft status, and even after the new RFC was 1958 * published. But it would rather make sense to clean the obsoleted part 1959 * up, and to make the code simpler at this stage. 1960 */ 1961 void 1962 icmp6_reflect(m, off) 1963 struct mbuf *m; 1964 size_t off; 1965 { 1966 struct ip6_hdr *ip6; 1967 struct icmp6_hdr *icmp6; 1968 struct in6_ifaddr *ia; 1969 int plen; 1970 int type, code; 1971 struct ifnet *outif = NULL; 1972 struct in6_addr origdst, *src = NULL; 1973 1974 /* too short to reflect */ 1975 if (off < sizeof(struct ip6_hdr)) { 1976 nd6log((LOG_DEBUG, 1977 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 1978 (u_long)off, (u_long)sizeof(struct ip6_hdr), 1979 __FILE__, __LINE__)); 1980 goto bad; 1981 } 1982 1983 /* 1984 * If there are extra headers between IPv6 and ICMPv6, strip 1985 * off that header first. 1986 */ 1987 #ifdef DIAGNOSTIC 1988 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 1989 panic("assumption failed in icmp6_reflect"); 1990 #endif 1991 if (off > sizeof(struct ip6_hdr)) { 1992 size_t l; 1993 struct ip6_hdr nip6; 1994 1995 l = off - sizeof(struct ip6_hdr); 1996 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 1997 m_adj(m, l); 1998 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1999 if (m->m_len < l) { 2000 if ((m = m_pullup(m, l)) == NULL) 2001 return; 2002 } 2003 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2004 } else /* off == sizeof(struct ip6_hdr) */ { 2005 size_t l; 2006 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2007 if (m->m_len < l) { 2008 if ((m = m_pullup(m, l)) == NULL) 2009 return; 2010 } 2011 } 2012 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2013 ip6 = mtod(m, struct ip6_hdr *); 2014 ip6->ip6_nxt = IPPROTO_ICMPV6; 2015 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2016 type = icmp6->icmp6_type; /* keep type for statistics */ 2017 code = icmp6->icmp6_code; /* ditto. */ 2018 2019 origdst = ip6->ip6_dst; 2020 /* 2021 * ip6_input() drops a packet if its src is multicast. 2022 * So, the src is never multicast. 2023 */ 2024 ip6->ip6_dst = ip6->ip6_src; 2025 2026 /* 2027 * If the incoming packet was addressed directly to us (i.e. unicast), 2028 * use dst as the src for the reply. 2029 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2030 * (for example) when we encounter an error while forwarding procedure 2031 * destined to a duplicated address of ours. 2032 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2033 * procedure of an outgoing packet of our own, in which case we need 2034 * to search in the ifaddr list. 2035 */ 2036 if (!IN6_IS_ADDR_MULTICAST(&origdst)) { 2037 if ((ia = ip6_getdstifaddr(m))) { 2038 if (!(ia->ia6_flags & 2039 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) 2040 src = &ia->ia_addr.sin6_addr; 2041 } else { 2042 struct sockaddr_in6 d; 2043 2044 bzero(&d, sizeof(d)); 2045 d.sin6_family = AF_INET6; 2046 d.sin6_len = sizeof(d); 2047 d.sin6_addr = origdst; 2048 ia = (struct in6_ifaddr *) 2049 ifa_ifwithaddr((struct sockaddr *)&d); 2050 if (ia && 2051 !(ia->ia6_flags & 2052 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) { 2053 src = &ia->ia_addr.sin6_addr; 2054 } 2055 } 2056 } 2057 2058 if (src == NULL) { 2059 int e; 2060 struct sockaddr_in6 sin6; 2061 struct route_in6 ro; 2062 2063 /* 2064 * This case matches to multicasts, our anycast, or unicasts 2065 * that we do not own. Select a source address based on the 2066 * source address of the erroneous packet. 2067 */ 2068 memset(&sin6, 0, sizeof(sin6)); 2069 sin6.sin6_family = AF_INET6; 2070 sin6.sin6_len = sizeof(sin6); 2071 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */ 2072 2073 memset(&ro, 0, sizeof(ro)); 2074 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2075 rtcache_free((struct route *)&ro); 2076 if (src == NULL) { 2077 nd6log((LOG_DEBUG, 2078 "icmp6_reflect: source can't be determined: " 2079 "dst=%s, error=%d\n", 2080 ip6_sprintf(&sin6.sin6_addr), e)); 2081 goto bad; 2082 } 2083 } 2084 2085 ip6->ip6_src = *src; 2086 ip6->ip6_flow = 0; 2087 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2088 ip6->ip6_vfc |= IPV6_VERSION; 2089 ip6->ip6_nxt = IPPROTO_ICMPV6; 2090 if (m->m_pkthdr.rcvif) { 2091 /* XXX: This may not be the outgoing interface */ 2092 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2093 } else 2094 ip6->ip6_hlim = ip6_defhlim; 2095 2096 m->m_pkthdr.csum_flags = 0; 2097 icmp6->icmp6_cksum = 0; 2098 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2099 sizeof(struct ip6_hdr), plen); 2100 2101 /* 2102 * XXX option handling 2103 */ 2104 2105 m->m_flags &= ~(M_BCAST|M_MCAST); 2106 2107 /* 2108 * To avoid a "too big" situation at an intermediate router 2109 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2110 * Note that only echo and node information replies are affected, 2111 * since the length of ICMP6 errors is limited to the minimum MTU. 2112 */ 2113 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 && 2114 outif) 2115 icmp6_ifstat_inc(outif, ifs6_out_error); 2116 2117 if (outif) 2118 icmp6_ifoutstat_inc(outif, type, code); 2119 2120 return; 2121 2122 bad: 2123 m_freem(m); 2124 return; 2125 } 2126 2127 static const char * 2128 icmp6_redirect_diag(src6, dst6, tgt6) 2129 struct in6_addr *src6; 2130 struct in6_addr *dst6; 2131 struct in6_addr *tgt6; 2132 { 2133 static char buf[1024]; 2134 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2135 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2136 return buf; 2137 } 2138 2139 void 2140 icmp6_redirect_input(m, off) 2141 struct mbuf *m; 2142 int off; 2143 { 2144 struct ifnet *ifp = m->m_pkthdr.rcvif; 2145 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2146 struct nd_redirect *nd_rd; 2147 int icmp6len = ntohs(ip6->ip6_plen); 2148 char *lladdr = NULL; 2149 int lladdrlen = 0; 2150 struct rtentry *rt = NULL; 2151 int is_router; 2152 int is_onlink; 2153 struct in6_addr src6 = ip6->ip6_src; 2154 struct in6_addr redtgt6; 2155 struct in6_addr reddst6; 2156 union nd_opts ndopts; 2157 2158 if (!ifp) 2159 return; 2160 2161 /* XXX if we are router, we don't update route by icmp6 redirect */ 2162 if (ip6_forwarding) 2163 goto freeit; 2164 if (!icmp6_rediraccept) 2165 goto freeit; 2166 2167 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2168 if (nd_rd == NULL) { 2169 icmp6stat.icp6s_tooshort++; 2170 return; 2171 } 2172 redtgt6 = nd_rd->nd_rd_target; 2173 reddst6 = nd_rd->nd_rd_dst; 2174 2175 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2176 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2177 goto freeit; 2178 } 2179 2180 /* validation */ 2181 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2182 nd6log((LOG_ERR, 2183 "ICMP6 redirect sent from %s rejected; " 2184 "must be from linklocal\n", ip6_sprintf(&src6))); 2185 goto bad; 2186 } 2187 if (ip6->ip6_hlim != 255) { 2188 nd6log((LOG_ERR, 2189 "ICMP6 redirect sent from %s rejected; " 2190 "hlim=%d (must be 255)\n", 2191 ip6_sprintf(&src6), ip6->ip6_hlim)); 2192 goto bad; 2193 } 2194 { 2195 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2196 struct sockaddr_in6 sin6; 2197 struct in6_addr *gw6; 2198 2199 bzero(&sin6, sizeof(sin6)); 2200 sin6.sin6_family = AF_INET6; 2201 sin6.sin6_len = sizeof(struct sockaddr_in6); 2202 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2203 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2204 if (rt) { 2205 if (rt->rt_gateway == NULL || 2206 rt->rt_gateway->sa_family != AF_INET6) { 2207 nd6log((LOG_ERR, 2208 "ICMP6 redirect rejected; no route " 2209 "with inet6 gateway found for redirect dst: %s\n", 2210 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2211 RTFREE(rt); 2212 goto bad; 2213 } 2214 2215 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2216 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2217 nd6log((LOG_ERR, 2218 "ICMP6 redirect rejected; " 2219 "not equal to gw-for-src=%s (must be same): " 2220 "%s\n", 2221 ip6_sprintf(gw6), 2222 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2223 RTFREE(rt); 2224 goto bad; 2225 } 2226 } else { 2227 nd6log((LOG_ERR, 2228 "ICMP6 redirect rejected; " 2229 "no route found for redirect dst: %s\n", 2230 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2231 goto bad; 2232 } 2233 RTFREE(rt); 2234 rt = NULL; 2235 } 2236 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2237 nd6log((LOG_ERR, 2238 "ICMP6 redirect rejected; " 2239 "redirect dst must be unicast: %s\n", 2240 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2241 goto bad; 2242 } 2243 2244 is_router = is_onlink = 0; 2245 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2246 is_router = 1; /* router case */ 2247 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2248 is_onlink = 1; /* on-link destination case */ 2249 if (!is_router && !is_onlink) { 2250 nd6log((LOG_ERR, 2251 "ICMP6 redirect rejected; " 2252 "neither router case nor onlink case: %s\n", 2253 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2254 goto bad; 2255 } 2256 /* validation passed */ 2257 2258 icmp6len -= sizeof(*nd_rd); 2259 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2260 if (nd6_options(&ndopts) < 0) { 2261 nd6log((LOG_INFO, "icmp6_redirect_input: " 2262 "invalid ND option, rejected: %s\n", 2263 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2264 /* nd6_options have incremented stats */ 2265 goto freeit; 2266 } 2267 2268 if (ndopts.nd_opts_tgt_lladdr) { 2269 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2270 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2271 } 2272 2273 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2274 nd6log((LOG_INFO, 2275 "icmp6_redirect_input: lladdrlen mismatch for %s " 2276 "(if %d, icmp6 packet %d): %s\n", 2277 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2278 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2279 goto bad; 2280 } 2281 2282 /* RFC 2461 8.3 */ 2283 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2284 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2285 2286 if (!is_onlink) { /* better router case. perform rtredirect. */ 2287 /* perform rtredirect */ 2288 struct sockaddr_in6 sdst; 2289 struct sockaddr_in6 sgw; 2290 struct sockaddr_in6 ssrc; 2291 unsigned long rtcount; 2292 struct rtentry *newrt = NULL; 2293 2294 /* 2295 * do not install redirect route, if the number of entries 2296 * is too much (> hiwat). note that, the node (= host) will 2297 * work just fine even if we do not install redirect route 2298 * (there will be additional hops, though). 2299 */ 2300 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2301 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2302 return; 2303 else if (0 <= icmp6_redirect_lowat && 2304 rtcount > icmp6_redirect_lowat) { 2305 /* 2306 * XXX nuke a victim, install the new one. 2307 */ 2308 } 2309 2310 bzero(&sdst, sizeof(sdst)); 2311 bzero(&sgw, sizeof(sgw)); 2312 bzero(&ssrc, sizeof(ssrc)); 2313 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2314 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2315 sizeof(struct sockaddr_in6); 2316 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2317 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2318 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2319 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2320 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2321 (struct sockaddr *)&ssrc, 2322 &newrt); 2323 2324 if (newrt) { 2325 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2326 icmp6_redirect_timeout_q); 2327 rtfree(newrt); 2328 } 2329 } 2330 /* finally update cached route in each socket via pfctlinput */ 2331 { 2332 struct sockaddr_in6 sdst; 2333 2334 bzero(&sdst, sizeof(sdst)); 2335 sdst.sin6_family = AF_INET6; 2336 sdst.sin6_len = sizeof(struct sockaddr_in6); 2337 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2338 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2339 #if defined(IPSEC) || defined(FAST_IPSEC) 2340 key_sa_routechange((struct sockaddr *)&sdst); 2341 #endif 2342 } 2343 2344 freeit: 2345 m_freem(m); 2346 return; 2347 2348 bad: 2349 icmp6stat.icp6s_badredirect++; 2350 m_freem(m); 2351 } 2352 2353 void 2354 icmp6_redirect_output(m0, rt) 2355 struct mbuf *m0; 2356 struct rtentry *rt; 2357 { 2358 struct ifnet *ifp; /* my outgoing interface */ 2359 struct in6_addr *ifp_ll6; 2360 struct in6_addr *nexthop; 2361 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2362 struct mbuf *m = NULL; /* newly allocated one */ 2363 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2364 struct nd_redirect *nd_rd; 2365 size_t maxlen; 2366 u_char *p; 2367 struct sockaddr_in6 src_sa; 2368 2369 icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2370 2371 /* if we are not router, we don't send icmp6 redirect */ 2372 if (!ip6_forwarding) 2373 goto fail; 2374 2375 /* sanity check */ 2376 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2377 goto fail; 2378 2379 /* 2380 * Address check: 2381 * the source address must identify a neighbor, and 2382 * the destination address must not be a multicast address 2383 * [RFC 2461, sec 8.2] 2384 */ 2385 sip6 = mtod(m0, struct ip6_hdr *); 2386 bzero(&src_sa, sizeof(src_sa)); 2387 src_sa.sin6_family = AF_INET6; 2388 src_sa.sin6_len = sizeof(src_sa); 2389 src_sa.sin6_addr = sip6->ip6_src; 2390 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2391 goto fail; 2392 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2393 goto fail; /* what should we do here? */ 2394 2395 /* rate limit */ 2396 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2397 goto fail; 2398 2399 /* 2400 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2401 * we almost always ask for an mbuf cluster for simplicity. 2402 * (MHLEN < IPV6_MMTU is almost always true) 2403 */ 2404 #if IPV6_MMTU >= MCLBYTES 2405 # error assumption failed about IPV6_MMTU and MCLBYTES 2406 #endif 2407 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2408 if (m && IPV6_MMTU >= MHLEN) 2409 MCLGET(m, M_DONTWAIT); 2410 if (!m) 2411 goto fail; 2412 m->m_pkthdr.rcvif = NULL; 2413 m->m_len = 0; 2414 maxlen = M_TRAILINGSPACE(m); 2415 maxlen = min(IPV6_MMTU, maxlen); 2416 /* just for safety */ 2417 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2418 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2419 goto fail; 2420 } 2421 2422 { 2423 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2424 struct in6_ifaddr *ia; 2425 if ((ia = in6ifa_ifpforlinklocal(ifp, 2426 IN6_IFF_NOTREADY| 2427 IN6_IFF_ANYCAST)) == NULL) 2428 goto fail; 2429 ifp_ll6 = &ia->ia_addr.sin6_addr; 2430 } 2431 2432 /* get ip6 linklocal address for the router. */ 2433 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2434 struct sockaddr_in6 *sin6; 2435 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2436 nexthop = &sin6->sin6_addr; 2437 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2438 nexthop = NULL; 2439 } else 2440 nexthop = NULL; 2441 2442 /* ip6 */ 2443 ip6 = mtod(m, struct ip6_hdr *); 2444 ip6->ip6_flow = 0; 2445 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2446 ip6->ip6_vfc |= IPV6_VERSION; 2447 /* ip6->ip6_plen will be set later */ 2448 ip6->ip6_nxt = IPPROTO_ICMPV6; 2449 ip6->ip6_hlim = 255; 2450 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2451 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2452 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2453 2454 /* ND Redirect */ 2455 nd_rd = (struct nd_redirect *)(ip6 + 1); 2456 nd_rd->nd_rd_type = ND_REDIRECT; 2457 nd_rd->nd_rd_code = 0; 2458 nd_rd->nd_rd_reserved = 0; 2459 if (rt->rt_flags & RTF_GATEWAY) { 2460 /* 2461 * nd_rd->nd_rd_target must be a link-local address in 2462 * better router cases. 2463 */ 2464 if (!nexthop) 2465 goto fail; 2466 bcopy(nexthop, &nd_rd->nd_rd_target, 2467 sizeof(nd_rd->nd_rd_target)); 2468 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2469 sizeof(nd_rd->nd_rd_dst)); 2470 } else { 2471 /* make sure redtgt == reddst */ 2472 nexthop = &sip6->ip6_dst; 2473 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2474 sizeof(nd_rd->nd_rd_target)); 2475 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2476 sizeof(nd_rd->nd_rd_dst)); 2477 } 2478 2479 p = (u_char *)(nd_rd + 1); 2480 2481 { 2482 /* target lladdr option */ 2483 struct rtentry *rt_nexthop = NULL; 2484 int len; 2485 struct sockaddr_dl *sdl; 2486 struct nd_opt_hdr *nd_opt; 2487 char *lladdr; 2488 2489 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2490 if (!rt_nexthop) 2491 goto nolladdropt; 2492 len = sizeof(*nd_opt) + ifp->if_addrlen; 2493 len = (len + 7) & ~7; /* round by 8 */ 2494 /* safety check */ 2495 if (len + (p - (u_char *)ip6) > maxlen) 2496 goto nolladdropt; 2497 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2498 (rt_nexthop->rt_flags & RTF_LLINFO) && 2499 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2500 (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) && 2501 sdl->sdl_alen) { 2502 nd_opt = (struct nd_opt_hdr *)p; 2503 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2504 nd_opt->nd_opt_len = len >> 3; 2505 lladdr = (char *)(nd_opt + 1); 2506 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen); 2507 p += len; 2508 } 2509 } 2510 nolladdropt:; 2511 2512 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2513 2514 /* just to be safe */ 2515 if (m0->m_flags & M_DECRYPTED) 2516 goto noredhdropt; 2517 if (p - (u_char *)ip6 > maxlen) 2518 goto noredhdropt; 2519 2520 { 2521 /* redirected header option */ 2522 int len; 2523 struct nd_opt_rd_hdr *nd_opt_rh; 2524 2525 /* 2526 * compute the maximum size for icmp6 redirect header option. 2527 * XXX room for auth header? 2528 */ 2529 len = maxlen - (p - (u_char *)ip6); 2530 len &= ~7; 2531 2532 /* 2533 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2534 * about padding/truncate rule for the original IP packet. 2535 * From the discussion on IPv6imp in Feb 1999, 2536 * the consensus was: 2537 * - "attach as much as possible" is the goal 2538 * - pad if not aligned (original size can be guessed by 2539 * original ip6 header) 2540 * Following code adds the padding if it is simple enough, 2541 * and truncates if not. 2542 */ 2543 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2544 /* not enough room, truncate */ 2545 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2546 m0->m_pkthdr.len); 2547 } else { 2548 /* 2549 * enough room, truncate if not aligned. 2550 * we don't pad here for simplicity. 2551 */ 2552 size_t extra; 2553 2554 extra = m0->m_pkthdr.len % 8; 2555 if (extra) { 2556 /* truncate */ 2557 m_adj(m0, -extra); 2558 } 2559 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2560 } 2561 2562 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2563 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2564 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2565 nd_opt_rh->nd_opt_rh_len = len >> 3; 2566 p += sizeof(*nd_opt_rh); 2567 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2568 2569 /* connect m0 to m */ 2570 m->m_pkthdr.len += m0->m_pkthdr.len; 2571 m_cat(m, m0); 2572 m0 = NULL; 2573 } 2574 noredhdropt: 2575 if (m0) { 2576 m_freem(m0); 2577 m0 = NULL; 2578 } 2579 2580 /* XXX: clear embedded link IDs in the inner header */ 2581 in6_clearscope(&sip6->ip6_src); 2582 in6_clearscope(&sip6->ip6_dst); 2583 in6_clearscope(&nd_rd->nd_rd_target); 2584 in6_clearscope(&nd_rd->nd_rd_dst); 2585 2586 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2587 2588 nd_rd->nd_rd_cksum = 0; 2589 nd_rd->nd_rd_cksum 2590 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2591 2592 /* send the packet to outside... */ 2593 if (ip6_output(m, NULL, NULL, 0, 2594 (struct ip6_moptions *)NULL, (struct socket *)NULL, NULL) != 0) 2595 icmp6_ifstat_inc(ifp, ifs6_out_error); 2596 2597 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2598 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2599 icmp6stat.icp6s_outhist[ND_REDIRECT]++; 2600 2601 return; 2602 2603 fail: 2604 if (m) 2605 m_freem(m); 2606 if (m0) 2607 m_freem(m0); 2608 } 2609 2610 /* 2611 * ICMPv6 socket option processing. 2612 */ 2613 int 2614 icmp6_ctloutput(op, so, level, optname, mp) 2615 int op; 2616 struct socket *so; 2617 int level, optname; 2618 struct mbuf **mp; 2619 { 2620 int error = 0; 2621 int optlen; 2622 struct in6pcb *in6p = sotoin6pcb(so); 2623 struct mbuf *m = *mp; 2624 2625 optlen = m ? m->m_len : 0; 2626 2627 if (level != IPPROTO_ICMPV6) { 2628 if (op == PRCO_SETOPT && m) 2629 (void)m_free(m); 2630 return EINVAL; 2631 } 2632 2633 switch (op) { 2634 case PRCO_SETOPT: 2635 switch (optname) { 2636 case ICMP6_FILTER: 2637 { 2638 struct icmp6_filter *p; 2639 2640 if (optlen != sizeof(*p)) { 2641 error = EMSGSIZE; 2642 break; 2643 } 2644 p = mtod(m, struct icmp6_filter *); 2645 if (!p || !in6p->in6p_icmp6filt) { 2646 error = EINVAL; 2647 break; 2648 } 2649 bcopy(p, in6p->in6p_icmp6filt, 2650 sizeof(struct icmp6_filter)); 2651 error = 0; 2652 break; 2653 } 2654 2655 default: 2656 error = ENOPROTOOPT; 2657 break; 2658 } 2659 if (m) 2660 (void)m_freem(m); 2661 break; 2662 2663 case PRCO_GETOPT: 2664 switch (optname) { 2665 case ICMP6_FILTER: 2666 { 2667 struct icmp6_filter *p; 2668 2669 if (!in6p->in6p_icmp6filt) { 2670 error = EINVAL; 2671 break; 2672 } 2673 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2674 m->m_len = sizeof(struct icmp6_filter); 2675 p = mtod(m, struct icmp6_filter *); 2676 bcopy(in6p->in6p_icmp6filt, p, 2677 sizeof(struct icmp6_filter)); 2678 error = 0; 2679 break; 2680 } 2681 2682 default: 2683 error = ENOPROTOOPT; 2684 break; 2685 } 2686 break; 2687 } 2688 2689 return (error); 2690 } 2691 2692 /* 2693 * Perform rate limit check. 2694 * Returns 0 if it is okay to send the icmp6 packet. 2695 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2696 * limitation. 2697 * 2698 * XXX per-destination/type check necessary? 2699 */ 2700 static int 2701 icmp6_ratelimit( 2702 const struct in6_addr *dst, /* not used at this moment */ 2703 const int type, /* not used at this moment */ 2704 const int code) /* not used at this moment */ 2705 { 2706 int ret; 2707 2708 ret = 0; /* okay to send */ 2709 2710 /* PPS limit */ 2711 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2712 icmp6errppslim)) { 2713 /* The packet is subject to rate limit */ 2714 ret++; 2715 } 2716 2717 return ret; 2718 } 2719 2720 static struct rtentry * 2721 icmp6_mtudisc_clone(dst) 2722 struct sockaddr *dst; 2723 { 2724 struct rtentry *rt; 2725 int error; 2726 2727 rt = rtalloc1(dst, 1); 2728 if (rt == 0) 2729 return NULL; 2730 2731 /* If we didn't get a host route, allocate one */ 2732 if ((rt->rt_flags & RTF_HOST) == 0) { 2733 struct rtentry *nrt; 2734 2735 error = rtrequest((int) RTM_ADD, dst, 2736 (struct sockaddr *) rt->rt_gateway, 2737 (struct sockaddr *) 0, 2738 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2739 if (error) { 2740 rtfree(rt); 2741 return NULL; 2742 } 2743 nrt->rt_rmx = rt->rt_rmx; 2744 rtfree(rt); 2745 rt = nrt; 2746 } 2747 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2748 icmp6_mtudisc_timeout_q); 2749 if (error) { 2750 rtfree(rt); 2751 return NULL; 2752 } 2753 2754 return rt; /* caller need to call rtfree() */ 2755 } 2756 2757 static void 2758 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 2759 { 2760 if (rt == NULL) 2761 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2762 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2763 (RTF_DYNAMIC | RTF_HOST)) { 2764 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2765 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2766 } else { 2767 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2768 rt->rt_rmx.rmx_mtu = 0; 2769 } 2770 } 2771 2772 static void 2773 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r) 2774 { 2775 if (rt == NULL) 2776 panic("icmp6_redirect_timeout: bad route to timeout"); 2777 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2778 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2779 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2780 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2781 } 2782 } 2783 2784 /* 2785 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes. silly? 2786 */ 2787 static int 2788 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS) 2789 { 2790 (void)&name; 2791 (void)&l; 2792 (void)&oname; 2793 2794 if (namelen != 0) 2795 return (EINVAL); 2796 2797 return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp, 2798 /*XXXUNCONST*/ 2799 __UNCONST(newp), newlen)); 2800 } 2801 2802 SYSCTL_SETUP(sysctl_net_inet6_icmp6_setup, 2803 "sysctl net.inet6.icmp6 subtree setup") 2804 { 2805 extern int nd6_maxqueuelen; /* defined in nd6.c */ 2806 2807 sysctl_createv(clog, 0, NULL, NULL, 2808 CTLFLAG_PERMANENT, 2809 CTLTYPE_NODE, "net", NULL, 2810 NULL, 0, NULL, 0, 2811 CTL_NET, CTL_EOL); 2812 sysctl_createv(clog, 0, NULL, NULL, 2813 CTLFLAG_PERMANENT, 2814 CTLTYPE_NODE, "inet6", NULL, 2815 NULL, 0, NULL, 0, 2816 CTL_NET, PF_INET6, CTL_EOL); 2817 sysctl_createv(clog, 0, NULL, NULL, 2818 CTLFLAG_PERMANENT, 2819 CTLTYPE_NODE, "icmp6", 2820 SYSCTL_DESCR("ICMPv6 related settings"), 2821 NULL, 0, NULL, 0, 2822 CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL); 2823 2824 sysctl_createv(clog, 0, NULL, NULL, 2825 CTLFLAG_PERMANENT, 2826 CTLTYPE_STRUCT, "stats", 2827 SYSCTL_DESCR("ICMPv6 transmission statistics"), 2828 NULL, 0, &icmp6stat, sizeof(icmp6stat), 2829 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2830 ICMPV6CTL_STATS, CTL_EOL); 2831 sysctl_createv(clog, 0, NULL, NULL, 2832 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2833 CTLTYPE_INT, "rediraccept", 2834 SYSCTL_DESCR("Accept and process redirect messages"), 2835 NULL, 0, &icmp6_rediraccept, 0, 2836 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2837 ICMPV6CTL_REDIRACCEPT, CTL_EOL); 2838 sysctl_createv(clog, 0, NULL, NULL, 2839 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2840 CTLTYPE_INT, "redirtimeout", 2841 SYSCTL_DESCR("Redirect generated route lifetime"), 2842 NULL, 0, &icmp6_redirtimeout, 0, 2843 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2844 ICMPV6CTL_REDIRTIMEOUT, CTL_EOL); 2845 #if 0 /* obsoleted */ 2846 sysctl_createv(clog, 0, NULL, NULL, 2847 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2848 CTLTYPE_INT, "errratelimit", NULL, 2849 NULL, 0, &icmp6_errratelimit, 0, 2850 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2851 ICMPV6CTL_ERRRATELIMIT, CTL_EOL); 2852 #endif 2853 sysctl_createv(clog, 0, NULL, NULL, 2854 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2855 CTLTYPE_INT, "nd6_prune", 2856 SYSCTL_DESCR("Neighbor discovery prune interval"), 2857 NULL, 0, &nd6_prune, 0, 2858 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2859 ICMPV6CTL_ND6_PRUNE, CTL_EOL); 2860 sysctl_createv(clog, 0, NULL, NULL, 2861 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2862 CTLTYPE_INT, "nd6_delay", 2863 SYSCTL_DESCR("First probe delay time"), 2864 NULL, 0, &nd6_delay, 0, 2865 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2866 ICMPV6CTL_ND6_DELAY, CTL_EOL); 2867 sysctl_createv(clog, 0, NULL, NULL, 2868 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2869 CTLTYPE_INT, "nd6_umaxtries", 2870 SYSCTL_DESCR("Number of unicast discovery attempts"), 2871 NULL, 0, &nd6_umaxtries, 0, 2872 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2873 ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL); 2874 sysctl_createv(clog, 0, NULL, NULL, 2875 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2876 CTLTYPE_INT, "nd6_mmaxtries", 2877 SYSCTL_DESCR("Number of multicast discovery attempts"), 2878 NULL, 0, &nd6_mmaxtries, 0, 2879 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2880 ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL); 2881 sysctl_createv(clog, 0, NULL, NULL, 2882 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2883 CTLTYPE_INT, "nd6_useloopback", 2884 SYSCTL_DESCR("Use loopback interface for local traffic"), 2885 NULL, 0, &nd6_useloopback, 0, 2886 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2887 ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL); 2888 #if 0 /* obsoleted */ 2889 sysctl_createv(clog, 0, NULL, NULL, 2890 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2891 CTLTYPE_INT, "nd6_proxyall", NULL, 2892 NULL, 0, &nd6_proxyall, 0, 2893 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2894 ICMPV6CTL_ND6_PROXYALL, CTL_EOL); 2895 #endif 2896 sysctl_createv(clog, 0, NULL, NULL, 2897 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2898 CTLTYPE_INT, "nodeinfo", 2899 SYSCTL_DESCR("Respond to node information requests"), 2900 NULL, 0, &icmp6_nodeinfo, 0, 2901 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2902 ICMPV6CTL_NODEINFO, CTL_EOL); 2903 sysctl_createv(clog, 0, NULL, NULL, 2904 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2905 CTLTYPE_INT, "errppslimit", 2906 SYSCTL_DESCR("Maximum ICMP errors sent per second"), 2907 NULL, 0, &icmp6errppslim, 0, 2908 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2909 ICMPV6CTL_ERRPPSLIMIT, CTL_EOL); 2910 sysctl_createv(clog, 0, NULL, NULL, 2911 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2912 CTLTYPE_INT, "nd6_maxnudhint", 2913 SYSCTL_DESCR("Maximum neighbor unreachable hint count"), 2914 NULL, 0, &nd6_maxnudhint, 0, 2915 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2916 ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL); 2917 sysctl_createv(clog, 0, NULL, NULL, 2918 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2919 CTLTYPE_INT, "mtudisc_hiwat", 2920 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2921 NULL, 0, &icmp6_mtudisc_hiwat, 0, 2922 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2923 ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL); 2924 sysctl_createv(clog, 0, NULL, NULL, 2925 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2926 CTLTYPE_INT, "mtudisc_lowat", 2927 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2928 NULL, 0, &icmp6_mtudisc_lowat, 0, 2929 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2930 ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL); 2931 sysctl_createv(clog, 0, NULL, NULL, 2932 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2933 CTLTYPE_INT, "nd6_debug", 2934 SYSCTL_DESCR("Enable neighbor discovery debug output"), 2935 NULL, 0, &nd6_debug, 0, 2936 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2937 ICMPV6CTL_ND6_DEBUG, CTL_EOL); 2938 sysctl_createv(clog, 0, NULL, NULL, 2939 CTLFLAG_PERMANENT, 2940 CTLTYPE_STRUCT, "nd6_drlist", 2941 SYSCTL_DESCR("Default router list"), 2942 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2943 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2944 ICMPV6CTL_ND6_DRLIST, CTL_EOL); 2945 sysctl_createv(clog, 0, NULL, NULL, 2946 CTLFLAG_PERMANENT, 2947 CTLTYPE_STRUCT, "nd6_prlist", 2948 SYSCTL_DESCR("Prefix list"), 2949 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2950 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2951 ICMPV6CTL_ND6_PRLIST, CTL_EOL); 2952 sysctl_createv(clog, 0, NULL, NULL, 2953 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2954 CTLTYPE_INT, "maxqueuelen", 2955 SYSCTL_DESCR("max packet queue len for a unresolved ND"), 2956 NULL, 1, &nd6_maxqueuelen, 0, 2957 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2958 ICMPV6CTL_ND6_MAXQLEN, CTL_EOL); 2959 } 2960
Indexes created Wed Sep 24 05:09:52 GMT 2025