1/* $NetBSD: icmp6.c,v 1.159 2011/12/31 20:41:59 christos 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 1.159 2011/12/31 20:41:59 christos 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 <netinet6/ip6_private.h> 93#include <netinet/icmp6.h> 94#include <netinet6/icmp6_private.h> 95#include <netinet6/mld6_var.h> 96#include <netinet6/in6_pcb.h> 97#include <netinet6/nd6.h> 98#include <netinet6/in6_ifattach.h> 99#include <netinet6/ip6protosw.h> 100#include <netinet6/scope6_var.h> 101 102#ifdef KAME_IPSEC 103#include <netinet6/ipsec.h> 104#include <netkey/key.h> 105#endif 106 107#ifdef FAST_IPSEC 108#include <netipsec/ipsec.h> 109#include <netipsec/key.h> 110#endif 111 112 113#include "faith.h" 114#if defined(NFAITH) && 0 < NFAITH 115#include <net/if_faith.h> 116#endif 117 118#include <net/net_osdep.h> 119 120extern struct domain inet6domain; 121 122percpu_t *icmp6stat_percpu; 123 124extern struct inpcbtable raw6cbtable; 125extern int icmp6errppslim; 126static int icmp6errpps_count = 0; 127static struct timeval icmp6errppslim_last; 128extern int icmp6_nodeinfo; 129 130/* 131 * List of callbacks to notify when Path MTU changes are made. 132 */ 133struct icmp6_mtudisc_callback { 134 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 135 void (*mc_func)(struct in6_addr *); 136}; 137 138LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 139 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 140 141static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 142extern int pmtu_expire; 143 144/* XXX do these values make any sense? */ 145static int icmp6_mtudisc_hiwat = 1280; 146static int icmp6_mtudisc_lowat = 256; 147 148/* 149 * keep track of # of redirect routes. 150 */ 151static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 152 153/* XXX experimental, turned off */ 154static int icmp6_redirect_hiwat = -1; 155static int icmp6_redirect_lowat = -1; 156 157static void icmp6_errcount(u_int, int, int); 158static int icmp6_rip6_input(struct mbuf **, int); 159static int icmp6_ratelimit(const struct in6_addr *, const int, const int); 160static const char *icmp6_redirect_diag(struct in6_addr *, 161 struct in6_addr *, struct in6_addr *); 162static struct mbuf *ni6_input(struct mbuf *, int); 163static struct mbuf *ni6_nametodns(const char *, int, int); 164static int ni6_dnsmatch(const char *, int, const char *, int); 165static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *, 166 struct ifnet **, char *); 167static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 168 struct ifnet *, int); 169static int icmp6_notify_error(struct mbuf *, int, int, int); 170static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *); 171static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *); 172static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *); 173static void sysctl_net_inet6_icmp6_setup(struct sysctllog **); 174 175 176void 177icmp6_init(void) 178{ 179 180 sysctl_net_inet6_icmp6_setup(NULL); 181 mld_init(); 182 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 183 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 184 185 icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS); 186} 187 188static void 189icmp6_errcount(u_int base, int type, int code) 190{ 191 switch (type) { 192 case ICMP6_DST_UNREACH: 193 switch (code) { 194 case ICMP6_DST_UNREACH_NOROUTE: 195 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE); 196 return; 197 case ICMP6_DST_UNREACH_ADMIN: 198 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN); 199 return; 200 case ICMP6_DST_UNREACH_BEYONDSCOPE: 201 ICMP6_STATINC(base + 202 ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE); 203 return; 204 case ICMP6_DST_UNREACH_ADDR: 205 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR); 206 return; 207 case ICMP6_DST_UNREACH_NOPORT: 208 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT); 209 return; 210 } 211 break; 212 case ICMP6_PACKET_TOO_BIG: 213 ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG); 214 return; 215 case ICMP6_TIME_EXCEEDED: 216 switch (code) { 217 case ICMP6_TIME_EXCEED_TRANSIT: 218 ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT); 219 return; 220 case ICMP6_TIME_EXCEED_REASSEMBLY: 221 ICMP6_STATINC(base + 222 ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY); 223 return; 224 } 225 break; 226 case ICMP6_PARAM_PROB: 227 switch (code) { 228 case ICMP6_PARAMPROB_HEADER: 229 ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER); 230 return; 231 case ICMP6_PARAMPROB_NEXTHEADER: 232 ICMP6_STATINC(base + 233 ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER); 234 return; 235 case ICMP6_PARAMPROB_OPTION: 236 ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION); 237 return; 238 } 239 break; 240 case ND_REDIRECT: 241 ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT); 242 return; 243 } 244 ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN); 245} 246 247/* 248 * Register a Path MTU Discovery callback. 249 */ 250void 251icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *)) 252{ 253 struct icmp6_mtudisc_callback *mc; 254 255 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 256 mc = LIST_NEXT(mc, mc_list)) { 257 if (mc->mc_func == func) 258 return; 259 } 260 261 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 262 if (mc == NULL) 263 panic("icmp6_mtudisc_callback_register"); 264 265 mc->mc_func = func; 266 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 267} 268 269/* 270 * A wrapper function for icmp6_error() necessary when the erroneous packet 271 * may not contain enough scope zone information. 272 */ 273void 274icmp6_error2(struct mbuf *m, int type, int code, int param, 275 struct ifnet *ifp) 276{ 277 struct ip6_hdr *ip6; 278 279 if (ifp == NULL) 280 return; 281 282 if (m->m_len < sizeof(struct ip6_hdr)) { 283 m = m_pullup(m, sizeof(struct ip6_hdr)); 284 if (m == NULL) 285 return; 286 } 287 288 ip6 = mtod(m, struct ip6_hdr *); 289 290 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 291 return; 292 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 293 return; 294 295 icmp6_error(m, type, code, param); 296} 297 298/* 299 * Generate an error packet of type error in response to bad IP6 packet. 300 */ 301void 302icmp6_error(struct mbuf *m, int type, int code, int param) 303{ 304 struct ip6_hdr *oip6, *nip6; 305 struct icmp6_hdr *icmp6; 306 u_int preplen; 307 int off; 308 int nxt; 309 310 ICMP6_STATINC(ICMP6_STAT_ERROR); 311 312 /* count per-type-code statistics */ 313 icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code); 314 315 if (m->m_flags & M_DECRYPTED) { 316 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 317 goto freeit; 318 } 319 320 if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) && 321 (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) 322 return; 323 oip6 = mtod(m, struct ip6_hdr *); 324 325 /* 326 * If the destination address of the erroneous packet is a multicast 327 * address, or the packet was sent using link-layer multicast, 328 * we should basically suppress sending an error (RFC 2463, Section 329 * 2.4). 330 * We have two exceptions (the item e.2 in that section): 331 * - the Pakcet Too Big message can be sent for path MTU discovery. 332 * - the Parameter Problem Message that can be allowed an icmp6 error 333 * in the option type field. This check has been done in 334 * ip6_unknown_opt(), so we can just check the type and code. 335 */ 336 if ((m->m_flags & (M_BCAST|M_MCAST) || 337 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 338 (type != ICMP6_PACKET_TOO_BIG && 339 (type != ICMP6_PARAM_PROB || 340 code != ICMP6_PARAMPROB_OPTION))) 341 goto freeit; 342 343 /* 344 * RFC 2463, 2.4 (e.5): source address check. 345 * XXX: the case of anycast source? 346 */ 347 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 348 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 349 goto freeit; 350 351 /* 352 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 353 * don't do it. 354 */ 355 nxt = -1; 356 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 357 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 358 struct icmp6_hdr *icp; 359 360 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 361 sizeof(*icp)); 362 if (icp == NULL) { 363 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 364 return; 365 } 366 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 367 icp->icmp6_type == ND_REDIRECT) { 368 /* 369 * ICMPv6 error 370 * Special case: for redirect (which is 371 * informational) we must not send icmp6 error. 372 */ 373 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 374 goto freeit; 375 } else { 376 /* ICMPv6 informational - send the error */ 377 } 378 } 379#if 0 /* controversial */ 380 else if (off >= 0 && nxt == IPPROTO_ESP) { 381 /* 382 * It could be ICMPv6 error inside ESP. Take a safer side, 383 * don't respond. 384 */ 385 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 386 goto freeit; 387 } 388#endif 389 else { 390 /* non-ICMPv6 - send the error */ 391 } 392 393 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 394 395 /* Finally, do rate limitation check. */ 396 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 397 ICMP6_STATINC(ICMP6_STAT_TOOFREQ); 398 goto freeit; 399 } 400 401 /* 402 * OK, ICMP6 can be generated. 403 */ 404 405 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 406 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 407 408 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 409 M_PREPEND(m, preplen, M_DONTWAIT); 410 if (m && M_UNWRITABLE(m, preplen)) 411 m = m_pullup(m, preplen); 412 if (m == NULL) { 413 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 414 return; 415 } 416 417 nip6 = mtod(m, struct ip6_hdr *); 418 nip6->ip6_src = oip6->ip6_src; 419 nip6->ip6_dst = oip6->ip6_dst; 420 421 in6_clearscope(&oip6->ip6_src); 422 in6_clearscope(&oip6->ip6_dst); 423 424 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 425 icmp6->icmp6_type = type; 426 icmp6->icmp6_code = code; 427 icmp6->icmp6_pptr = htonl((u_int32_t)param); 428 429 /* 430 * icmp6_reflect() is designed to be in the input path. 431 * icmp6_error() can be called from both input and output path, 432 * and if we are in output path rcvif could contain bogus value. 433 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 434 * information in ip header (nip6). 435 */ 436 m->m_pkthdr.rcvif = NULL; 437 438 ICMP6_STATINC(ICMP6_STAT_OUTHIST + type); 439 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 440 441 return; 442 443 freeit: 444 /* 445 * If we can't tell whether or not we can generate ICMP6, free it. 446 */ 447 m_freem(m); 448} 449 450/* 451 * Process a received ICMP6 message. 452 */ 453int 454icmp6_input(struct mbuf **mp, int *offp, int proto) 455{ 456 struct mbuf *m = *mp, *n; 457 struct ip6_hdr *ip6, *nip6; 458 struct icmp6_hdr *icmp6, *nicmp6; 459 int off = *offp; 460 int icmp6len = m->m_pkthdr.len - *offp; 461 int code, sum, noff, i; 462 463#define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4) 464 KASSERT(ICMP6_MAXLEN < MCLBYTES); 465 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 466 467 /* 468 * Locate icmp6 structure in mbuf, and check 469 * that not corrupted and of at least minimum length 470 */ 471 472 if (icmp6len < sizeof(struct icmp6_hdr)) { 473 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 474 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 475 goto freeit; 476 } 477 478 i = off + sizeof(*icmp6); 479 if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) { 480 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 481#if 0 /* m is 0 here */ 482 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 483#endif 484 goto freeit; 485 } 486 ip6 = mtod(m, struct ip6_hdr *); 487 /* 488 * calculate the checksum 489 */ 490 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 491 if (icmp6 == NULL) { 492 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 493 /* m is invalid */ 494 /*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/ 495 return IPPROTO_DONE; 496 } 497 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 498 code = icmp6->icmp6_code; 499 500 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 501 nd6log((LOG_ERR, 502 "ICMP6 checksum error(%d|%x) %s\n", 503 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 504 ICMP6_STATINC(ICMP6_STAT_CHECKSUM); 505 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 506 goto freeit; 507 } 508 509#if defined(NFAITH) && 0 < NFAITH 510 if (faithprefix(&ip6->ip6_dst)) { 511 /* 512 * Deliver very specific ICMP6 type only. 513 * This is important to deliver TOOBIG. Otherwise PMTUD 514 * will not work. 515 */ 516 switch (icmp6->icmp6_type) { 517 case ICMP6_DST_UNREACH: 518 case ICMP6_PACKET_TOO_BIG: 519 case ICMP6_TIME_EXCEEDED: 520 break; 521 default: 522 goto freeit; 523 } 524 } 525#endif 526 527 ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type); 528 529 switch (icmp6->icmp6_type) { 530 case ICMP6_DST_UNREACH: 531 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 532 switch (code) { 533 case ICMP6_DST_UNREACH_NOROUTE: 534 code = PRC_UNREACH_NET; 535 break; 536 case ICMP6_DST_UNREACH_ADMIN: 537 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 538 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 539 break; 540 case ICMP6_DST_UNREACH_ADDR: 541 code = PRC_HOSTDEAD; 542 break; 543#ifdef COMPAT_RFC1885 544 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 545 code = PRC_UNREACH_SRCFAIL; 546 break; 547#else 548 case ICMP6_DST_UNREACH_BEYONDSCOPE: 549 /* I mean "source address was incorrect." */ 550 code = PRC_UNREACH_NET; 551 break; 552#endif 553 case ICMP6_DST_UNREACH_NOPORT: 554 code = PRC_UNREACH_PORT; 555 break; 556 default: 557 goto badcode; 558 } 559 goto deliver; 560 561 case ICMP6_PACKET_TOO_BIG: 562 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 563 564 /* 565 * MTU is checked in icmp6_mtudisc. 566 */ 567 code = PRC_MSGSIZE; 568 569 /* 570 * Updating the path MTU will be done after examining 571 * intermediate extension headers. 572 */ 573 goto deliver; 574 575 case ICMP6_TIME_EXCEEDED: 576 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 577 switch (code) { 578 case ICMP6_TIME_EXCEED_TRANSIT: 579 code = PRC_TIMXCEED_INTRANS; 580 break; 581 case ICMP6_TIME_EXCEED_REASSEMBLY: 582 code = PRC_TIMXCEED_REASS; 583 break; 584 default: 585 goto badcode; 586 } 587 goto deliver; 588 589 case ICMP6_PARAM_PROB: 590 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 591 switch (code) { 592 case ICMP6_PARAMPROB_NEXTHEADER: 593 code = PRC_UNREACH_PROTOCOL; 594 break; 595 case ICMP6_PARAMPROB_HEADER: 596 case ICMP6_PARAMPROB_OPTION: 597 code = PRC_PARAMPROB; 598 break; 599 default: 600 goto badcode; 601 } 602 goto deliver; 603 604 case ICMP6_ECHO_REQUEST: 605 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 606 if (code != 0) 607 goto badcode; 608 /* 609 * Copy mbuf to send to two data paths: userland socket(s), 610 * and to the querier (echo reply). 611 * m: a copy for socket, n: a copy for querier 612 * 613 * If the first mbuf is shared, or the first mbuf is too short, 614 * copy the first part of the data into a fresh mbuf. 615 * Otherwise, we will wrongly overwrite both copies. 616 */ 617 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 618 /* Give up local */ 619 n = m; 620 m = NULL; 621 } else if (M_READONLY(n) || 622 n->m_len < off + sizeof(struct icmp6_hdr)) { 623 struct mbuf *n0 = n; 624 625 /* 626 * Prepare an internal mbuf. m_pullup() doesn't 627 * always copy the length we specified. 628 */ 629 if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 630 /* Give up local */ 631 n = m; 632 m = NULL; 633 } 634 m_freem(n0); 635 } 636 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, 637 sizeof(*nicmp6)); 638 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 639 nicmp6->icmp6_code = 0; 640 if (n) { 641 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 642 icmp6s[ICMP6_STAT_REFLECT]++; 643 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++; 644 ICMP6_STAT_PUTREF(); 645 icmp6_reflect(n, off); 646 } 647 if (!m) 648 goto freeit; 649 break; 650 651 case ICMP6_ECHO_REPLY: 652 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 653 if (code != 0) 654 goto badcode; 655 break; 656 657 case MLD_LISTENER_QUERY: 658 case MLD_LISTENER_REPORT: 659 if (icmp6len < sizeof(struct mld_hdr)) 660 goto badlen; 661 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */ 662 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 663 else 664 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 665 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 666 /* give up local */ 667 mld_input(m, off); 668 m = NULL; 669 goto freeit; 670 } 671 mld_input(n, off); 672 /* m stays. */ 673 break; 674 675 case MLD_LISTENER_DONE: 676 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 677 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */ 678 goto badlen; 679 break; /* nothing to be done in kernel */ 680 681 case MLD_MTRACE_RESP: 682 case MLD_MTRACE: 683 /* XXX: these two are experimental. not officially defined. */ 684 /* XXX: per-interface statistics? */ 685 break; /* just pass it to applications */ 686 687 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 688 { 689 enum { WRU, FQDN } mode; 690 691 if (!icmp6_nodeinfo) 692 break; 693 694 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 695 mode = WRU; 696 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 697 mode = FQDN; 698 else 699 goto badlen; 700 701 if (mode == FQDN) { 702 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 703 if (n) 704 n = ni6_input(n, off); 705 /* XXX meaningless if n == NULL */ 706 noff = sizeof(struct ip6_hdr); 707 } else { 708 u_char *p; 709 int maxhlen; 710 711 if ((icmp6_nodeinfo & 5) != 5) 712 break; 713 714 if (code != 0) 715 goto badcode; 716 MGETHDR(n, M_DONTWAIT, m->m_type); 717 if (n && ICMP6_MAXLEN > MHLEN) { 718 MCLGET(n, M_DONTWAIT); 719 if ((n->m_flags & M_EXT) == 0) { 720 m_free(n); 721 n = NULL; 722 } 723 } 724 if (n == NULL) { 725 /* Give up remote */ 726 break; 727 } 728 n->m_pkthdr.rcvif = NULL; 729 n->m_len = 0; 730 maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN; 731 if (maxhlen > hostnamelen) 732 maxhlen = hostnamelen; 733 /* 734 * Copy IPv6 and ICMPv6 only. 735 */ 736 nip6 = mtod(n, struct ip6_hdr *); 737 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 738 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 739 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 740 p = (u_char *)(nicmp6 + 1); 741 memset(p, 0, 4); 742 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 743 noff = sizeof(struct ip6_hdr); 744 M_COPY_PKTHDR(n, m); /* just for rcvif */ 745 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 746 sizeof(struct icmp6_hdr) + 4 + maxhlen; 747 nicmp6->icmp6_type = ICMP6_WRUREPLY; 748 nicmp6->icmp6_code = 0; 749 } 750#undef hostnamelen 751 if (n) { 752 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 753 icmp6s[ICMP6_STAT_REFLECT]++; 754 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++; 755 ICMP6_STAT_PUTREF(); 756 icmp6_reflect(n, noff); 757 } 758 break; 759 } 760 761 case ICMP6_WRUREPLY: 762 if (code != 0) 763 goto badcode; 764 break; 765 766 case ND_ROUTER_SOLICIT: 767 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 768 if (code != 0) 769 goto badcode; 770 if (icmp6len < sizeof(struct nd_router_solicit)) 771 goto badlen; 772 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 773 /* give up local */ 774 nd6_rs_input(m, off, icmp6len); 775 m = NULL; 776 goto freeit; 777 } 778 nd6_rs_input(n, off, icmp6len); 779 /* m stays. */ 780 break; 781 782 case ND_ROUTER_ADVERT: 783 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 784 if (code != 0) 785 goto badcode; 786 if (icmp6len < sizeof(struct nd_router_advert)) 787 goto badlen; 788 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 789 /* give up local */ 790 nd6_ra_input(m, off, icmp6len); 791 m = NULL; 792 goto freeit; 793 } 794 nd6_ra_input(n, off, icmp6len); 795 /* m stays. */ 796 break; 797 798 case ND_NEIGHBOR_SOLICIT: 799 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 800 if (code != 0) 801 goto badcode; 802 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 803 goto badlen; 804 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 805 /* give up local */ 806 nd6_ns_input(m, off, icmp6len); 807 m = NULL; 808 goto freeit; 809 } 810 nd6_ns_input(n, off, icmp6len); 811 /* m stays. */ 812 break; 813 814 case ND_NEIGHBOR_ADVERT: 815 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 816 if (code != 0) 817 goto badcode; 818 if (icmp6len < sizeof(struct nd_neighbor_advert)) 819 goto badlen; 820 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 821 /* give up local */ 822 nd6_na_input(m, off, icmp6len); 823 m = NULL; 824 goto freeit; 825 } 826 nd6_na_input(n, off, icmp6len); 827 /* m stays. */ 828 break; 829 830 case ND_REDIRECT: 831 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 832 if (code != 0) 833 goto badcode; 834 if (icmp6len < sizeof(struct nd_redirect)) 835 goto badlen; 836 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 837 /* give up local */ 838 icmp6_redirect_input(m, off); 839 m = NULL; 840 goto freeit; 841 } 842 icmp6_redirect_input(n, off); 843 /* m stays. */ 844 break; 845 846 case ICMP6_ROUTER_RENUMBERING: 847 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 848 code != ICMP6_ROUTER_RENUMBERING_RESULT) 849 goto badcode; 850 if (icmp6len < sizeof(struct icmp6_router_renum)) 851 goto badlen; 852 break; 853 854 default: 855 nd6log((LOG_DEBUG, 856 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 857 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 858 ip6_sprintf(&ip6->ip6_dst), 859 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 860 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 861 /* ICMPv6 error: MUST deliver it by spec... */ 862 code = PRC_NCMDS; 863 /* deliver */ 864 } else { 865 /* ICMPv6 informational: MUST not deliver */ 866 break; 867 } 868 deliver: 869 if (icmp6_notify_error(m, off, icmp6len, code)) { 870 /* In this case, m should've been freed. */ 871 return (IPPROTO_DONE); 872 } 873 break; 874 875 badcode: 876 ICMP6_STATINC(ICMP6_STAT_BADCODE); 877 break; 878 879 badlen: 880 ICMP6_STATINC(ICMP6_STAT_BADLEN); 881 break; 882 } 883 884 /* deliver the packet to appropriate sockets */ 885 icmp6_rip6_input(&m, *offp); 886 887 return IPPROTO_DONE; 888 889 freeit: 890 m_freem(m); 891 return IPPROTO_DONE; 892} 893 894static int 895icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code) 896{ 897 struct icmp6_hdr *icmp6; 898 struct ip6_hdr *eip6; 899 u_int32_t notifymtu; 900 struct sockaddr_in6 icmp6src, icmp6dst; 901 902 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 903 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 904 goto freeit; 905 } 906 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 907 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 908 if (icmp6 == NULL) { 909 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 910 return (-1); 911 } 912 eip6 = (struct ip6_hdr *)(icmp6 + 1); 913 914 /* Detect the upper level protocol */ 915 { 916 void (*ctlfunc)(int, struct sockaddr *, void *); 917 u_int8_t nxt = eip6->ip6_nxt; 918 int eoff = off + sizeof(struct icmp6_hdr) + 919 sizeof(struct ip6_hdr); 920 struct ip6ctlparam ip6cp; 921 struct in6_addr *finaldst = NULL; 922 int icmp6type = icmp6->icmp6_type; 923 struct ip6_frag *fh; 924 struct ip6_rthdr *rth; 925 struct ip6_rthdr0 *rth0; 926 int rthlen; 927 928 while (1) { /* XXX: should avoid infinite loop explicitly? */ 929 struct ip6_ext *eh; 930 931 switch (nxt) { 932 case IPPROTO_HOPOPTS: 933 case IPPROTO_DSTOPTS: 934 case IPPROTO_AH: 935 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 936 eoff, sizeof(*eh)); 937 if (eh == NULL) { 938 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 939 return (-1); 940 } 941 942 if (nxt == IPPROTO_AH) 943 eoff += (eh->ip6e_len + 2) << 2; 944 else 945 eoff += (eh->ip6e_len + 1) << 3; 946 nxt = eh->ip6e_nxt; 947 break; 948 case IPPROTO_ROUTING: 949 /* 950 * When the erroneous packet contains a 951 * routing header, we should examine the 952 * header to determine the final destination. 953 * Otherwise, we can't properly update 954 * information that depends on the final 955 * destination (e.g. path MTU). 956 */ 957 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 958 eoff, sizeof(*rth)); 959 if (rth == NULL) { 960 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 961 return (-1); 962 } 963 rthlen = (rth->ip6r_len + 1) << 3; 964 /* 965 * XXX: currently there is no 966 * officially defined type other 967 * than type-0. 968 * Note that if the segment left field 969 * is 0, all intermediate hops must 970 * have been passed. 971 */ 972 if (rth->ip6r_segleft && 973 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 974 int hops; 975 976 IP6_EXTHDR_GET(rth0, 977 struct ip6_rthdr0 *, m, 978 eoff, rthlen); 979 if (rth0 == NULL) { 980 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 981 return (-1); 982 } 983 /* just ignore a bogus header */ 984 if ((rth0->ip6r0_len % 2) == 0 && 985 (hops = rth0->ip6r0_len/2)) 986 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 987 } 988 eoff += rthlen; 989 nxt = rth->ip6r_nxt; 990 break; 991 case IPPROTO_FRAGMENT: 992 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 993 eoff, sizeof(*fh)); 994 if (fh == NULL) { 995 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 996 return (-1); 997 } 998 /* 999 * Data after a fragment header is meaningless 1000 * unless it is the first fragment, but 1001 * we'll go to the notify label for path MTU 1002 * discovery. 1003 */ 1004 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1005 goto notify; 1006 1007 eoff += sizeof(struct ip6_frag); 1008 nxt = fh->ip6f_nxt; 1009 break; 1010 default: 1011 /* 1012 * This case includes ESP and the No Next 1013 * Header. In such cases going to the notify 1014 * label does not have any meaning 1015 * (i.e. ctlfunc will be NULL), but we go 1016 * anyway since we might have to update 1017 * path MTU information. 1018 */ 1019 goto notify; 1020 } 1021 } 1022 notify: 1023 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1024 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1025 if (icmp6 == NULL) { 1026 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 1027 return (-1); 1028 } 1029 1030 /* 1031 * retrieve parameters from the inner IPv6 header, and convert 1032 * them into sockaddr structures. 1033 * XXX: there is no guarantee that the source or destination 1034 * addresses of the inner packet are in the same scope zone as 1035 * the addresses of the icmp packet. But there is no other 1036 * way to determine the zone. 1037 */ 1038 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1039 1040 sockaddr_in6_init(&icmp6dst, 1041 (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0); 1042 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1043 goto freeit; 1044 sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0); 1045 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1046 goto freeit; 1047 icmp6src.sin6_flowinfo = 1048 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1049 1050 if (finaldst == NULL) 1051 finaldst = &eip6->ip6_dst; 1052 ip6cp.ip6c_m = m; 1053 ip6cp.ip6c_icmp6 = icmp6; 1054 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1055 ip6cp.ip6c_off = eoff; 1056 ip6cp.ip6c_finaldst = finaldst; 1057 ip6cp.ip6c_src = &icmp6src; 1058 ip6cp.ip6c_nxt = nxt; 1059 1060 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1061 notifymtu = ntohl(icmp6->icmp6_mtu); 1062 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1063 } 1064 1065 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1066 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1067 if (ctlfunc) { 1068 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1069 &ip6cp); 1070 } 1071 } 1072 return (0); 1073 1074 freeit: 1075 m_freem(m); 1076 return (-1); 1077} 1078 1079void 1080icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1081{ 1082 unsigned long rtcount; 1083 struct icmp6_mtudisc_callback *mc; 1084 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1085 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1086 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1087 u_int mtu = ntohl(icmp6->icmp6_mtu); 1088 struct rtentry *rt = NULL; 1089 struct sockaddr_in6 sin6; 1090 1091 /* 1092 * The MTU should not be less than the minimal IPv6 MTU except for the 1093 * hack in ip6_output/ip6_setpmtu where we always include a frag header. 1094 * In that one case, the MTU might be less than 1280. 1095 */ 1096 if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) { 1097 /* is the mtu even sane? */ 1098 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) 1099 return; 1100 if (!validated) 1101 return; 1102 mtu = IPV6_MMTU - sizeof(struct ip6_frag); 1103 } 1104 1105 /* 1106 * allow non-validated cases if memory is plenty, to make traffic 1107 * from non-connected pcb happy. 1108 */ 1109 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1110 if (validated) { 1111 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1112 return; 1113 else if (0 <= icmp6_mtudisc_lowat && 1114 rtcount > icmp6_mtudisc_lowat) { 1115 /* 1116 * XXX nuke a victim, install the new one. 1117 */ 1118 } 1119 } else { 1120 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1121 return; 1122 } 1123 1124 memset(&sin6, 0, sizeof(sin6)); 1125 sin6.sin6_family = PF_INET6; 1126 sin6.sin6_len = sizeof(struct sockaddr_in6); 1127 sin6.sin6_addr = *dst; 1128 if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1129 return; 1130 1131 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1132 1133 if (rt && (rt->rt_flags & RTF_HOST) && 1134 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1135 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1136 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1137 ICMP6_STATINC(ICMP6_STAT_PMTUCHG); 1138 rt->rt_rmx.rmx_mtu = mtu; 1139 } 1140 } 1141 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */ 1142 RTFREE(rt); 1143 } 1144 1145 /* 1146 * Notify protocols that the MTU for this destination 1147 * has changed. 1148 */ 1149 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1150 mc = LIST_NEXT(mc, mc_list)) 1151 (*mc->mc_func)(&sin6.sin6_addr); 1152} 1153 1154/* 1155 * Process a Node Information Query packet, based on 1156 * draft-ietf-ipngwg-icmp-name-lookups-07. 1157 * 1158 * Spec incompatibilities: 1159 * - IPv6 Subject address handling 1160 * - IPv4 Subject address handling support missing 1161 * - Proxy reply (answer even if it's not for me) 1162 * - joins NI group address at in6_ifattach() time only, does not cope 1163 * with hostname changes by sethostname(3) 1164 */ 1165static struct mbuf * 1166ni6_input(struct mbuf *m, int off) 1167{ 1168 struct icmp6_nodeinfo *ni6, *nni6; 1169 struct mbuf *n = NULL; 1170 u_int16_t qtype; 1171 int subjlen; 1172 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1173 struct ni_reply_fqdn *fqdn; 1174 int addrs; /* for NI_QTYPE_NODEADDR */ 1175 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1176 struct sockaddr_in6 sin6; /* ip6_dst */ 1177 struct in6_addr in6_subj; /* subject address */ 1178 struct ip6_hdr *ip6; 1179 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1180 char *subj = NULL; 1181 1182 ip6 = mtod(m, struct ip6_hdr *); 1183 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1184 if (ni6 == NULL) { 1185 /* m is already reclaimed */ 1186 return NULL; 1187 } 1188 1189 /* 1190 * Validate IPv6 destination address. 1191 * 1192 * The Responder must discard the Query without further processing 1193 * unless it is one of the Responder's unicast or anycast addresses, or 1194 * a link-local scope multicast address which the Responder has joined. 1195 * [icmp-name-lookups-07, Section 4.] 1196 */ 1197 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 1198 /* XXX scopeid */ 1199 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1200 ; /* unicast/anycast, fine */ 1201 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1202 ; /* link-local multicast, fine */ 1203 else 1204 goto bad; 1205 1206 /* validate query Subject field. */ 1207 qtype = ntohs(ni6->ni_qtype); 1208 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1209 switch (qtype) { 1210 case NI_QTYPE_NOOP: 1211 case NI_QTYPE_SUPTYPES: 1212 /* 07 draft */ 1213 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1214 break; 1215 /* FALLTHROUGH */ 1216 case NI_QTYPE_FQDN: 1217 case NI_QTYPE_NODEADDR: 1218 case NI_QTYPE_IPV4ADDR: 1219 switch (ni6->ni_code) { 1220 case ICMP6_NI_SUBJ_IPV6: 1221#if ICMP6_NI_SUBJ_IPV6 != 0 1222 case 0: 1223#endif 1224 /* 1225 * backward compatibility - try to accept 03 draft 1226 * format, where no Subject is present. 1227 */ 1228 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1229 subjlen == 0) { 1230 oldfqdn++; 1231 break; 1232 } 1233#if ICMP6_NI_SUBJ_IPV6 != 0 1234 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1235 goto bad; 1236#endif 1237 1238 if (subjlen != sizeof(sin6.sin6_addr)) 1239 goto bad; 1240 1241 /* 1242 * Validate Subject address. 1243 * 1244 * Not sure what exactly "address belongs to the node" 1245 * means in the spec, is it just unicast, or what? 1246 * 1247 * At this moment we consider Subject address as 1248 * "belong to the node" if the Subject address equals 1249 * to the IPv6 destination address; validation for 1250 * IPv6 destination address should have done enough 1251 * check for us. 1252 * 1253 * We do not do proxy at this moment. 1254 */ 1255 /* m_pulldown instead of copy? */ 1256 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1257 subjlen, (void *)&in6_subj); 1258 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1259 goto bad; 1260 1261 subj = (char *)&in6_subj; 1262 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1263 break; 1264 1265 /* 1266 * XXX if we are to allow other cases, we should really 1267 * be careful about scope here. 1268 * basically, we should disallow queries toward IPv6 1269 * destination X with subject Y, if scope(X) > scope(Y). 1270 * if we allow scope(X) > scope(Y), it will result in 1271 * information leakage across scope boundary. 1272 */ 1273 goto bad; 1274 1275 case ICMP6_NI_SUBJ_FQDN: 1276 /* 1277 * Validate Subject name with gethostname(3). 1278 * 1279 * The behavior may need some debate, since: 1280 * - we are not sure if the node has FQDN as 1281 * hostname (returned by gethostname(3)). 1282 * - the code does wildcard match for truncated names. 1283 * however, we are not sure if we want to perform 1284 * wildcard match, if gethostname(3) side has 1285 * truncated hostname. 1286 */ 1287 n = ni6_nametodns(hostname, hostnamelen, 0); 1288 if (!n || n->m_next || n->m_len == 0) 1289 goto bad; 1290 IP6_EXTHDR_GET(subj, char *, m, 1291 off + sizeof(struct icmp6_nodeinfo), subjlen); 1292 if (subj == NULL) 1293 goto bad; 1294 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1295 n->m_len)) { 1296 goto bad; 1297 } 1298 m_freem(n); 1299 n = NULL; 1300 break; 1301 1302 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1303 default: 1304 goto bad; 1305 } 1306 break; 1307 } 1308 1309 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1310 switch (qtype) { 1311 case NI_QTYPE_FQDN: 1312 if ((icmp6_nodeinfo & 1) == 0) 1313 goto bad; 1314 break; 1315 case NI_QTYPE_NODEADDR: 1316 case NI_QTYPE_IPV4ADDR: 1317 if ((icmp6_nodeinfo & 2) == 0) 1318 goto bad; 1319 break; 1320 } 1321 1322 /* guess reply length */ 1323 switch (qtype) { 1324 case NI_QTYPE_NOOP: 1325 break; /* no reply data */ 1326 case NI_QTYPE_SUPTYPES: 1327 replylen += sizeof(u_int32_t); 1328 break; 1329 case NI_QTYPE_FQDN: 1330 /* XXX will append an mbuf */ 1331 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1332 break; 1333 case NI_QTYPE_NODEADDR: 1334 addrs = ni6_addrs(ni6, m, &ifp, subj); 1335 if ((replylen += addrs * (sizeof(struct in6_addr) + 1336 sizeof(u_int32_t))) > MCLBYTES) 1337 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1338 break; 1339 case NI_QTYPE_IPV4ADDR: 1340 /* unsupported - should respond with unknown Qtype? */ 1341 goto bad; 1342 default: 1343 /* 1344 * XXX: We must return a reply with the ICMP6 code 1345 * `unknown Qtype' in this case. However we regard the case 1346 * as an FQDN query for backward compatibility. 1347 * Older versions set a random value to this field, 1348 * so it rarely varies in the defined qtypes. 1349 * But the mechanism is not reliable... 1350 * maybe we should obsolete older versions. 1351 */ 1352 qtype = NI_QTYPE_FQDN; 1353 /* XXX will append an mbuf */ 1354 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1355 oldfqdn++; 1356 break; 1357 } 1358 1359 /* allocate an mbuf to reply. */ 1360 MGETHDR(n, M_DONTWAIT, m->m_type); 1361 if (n == NULL) { 1362 m_freem(m); 1363 return (NULL); 1364 } 1365 M_MOVE_PKTHDR(n, m); /* just for rcvif */ 1366 if (replylen > MHLEN) { 1367 if (replylen > MCLBYTES) { 1368 /* 1369 * XXX: should we try to allocate more? But MCLBYTES 1370 * is probably much larger than IPV6_MMTU... 1371 */ 1372 goto bad; 1373 } 1374 MCLGET(n, M_DONTWAIT); 1375 if ((n->m_flags & M_EXT) == 0) { 1376 goto bad; 1377 } 1378 } 1379 n->m_pkthdr.len = n->m_len = replylen; 1380 1381 /* copy mbuf header and IPv6 + Node Information base headers */ 1382 bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr)); 1383 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1384 bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo)); 1385 1386 /* qtype dependent procedure */ 1387 switch (qtype) { 1388 case NI_QTYPE_NOOP: 1389 nni6->ni_code = ICMP6_NI_SUCCESS; 1390 nni6->ni_flags = 0; 1391 break; 1392 case NI_QTYPE_SUPTYPES: 1393 { 1394 u_int32_t v; 1395 nni6->ni_code = ICMP6_NI_SUCCESS; 1396 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1397 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1398 v = (u_int32_t)htonl(0x0000000f); 1399 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1400 break; 1401 } 1402 case NI_QTYPE_FQDN: 1403 nni6->ni_code = ICMP6_NI_SUCCESS; 1404 fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) + 1405 sizeof(struct ip6_hdr) + 1406 sizeof(struct icmp6_nodeinfo)); 1407 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1408 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1409 /* 1410 * XXX do we really have FQDN in variable "hostname"? 1411 */ 1412 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1413 if (n->m_next == NULL) 1414 goto bad; 1415 /* XXX we assume that n->m_next is not a chain */ 1416 if (n->m_next->m_next != NULL) 1417 goto bad; 1418 n->m_pkthdr.len += n->m_next->m_len; 1419 break; 1420 case NI_QTYPE_NODEADDR: 1421 { 1422 int lenlim, copied; 1423 1424 nni6->ni_code = ICMP6_NI_SUCCESS; 1425 n->m_pkthdr.len = n->m_len = 1426 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1427 lenlim = M_TRAILINGSPACE(n); 1428 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1429 /* XXX: reset mbuf length */ 1430 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1431 sizeof(struct icmp6_nodeinfo) + copied; 1432 break; 1433 } 1434 default: 1435 break; /* XXX impossible! */ 1436 } 1437 1438 nni6->ni_type = ICMP6_NI_REPLY; 1439 m_freem(m); 1440 return (n); 1441 1442 bad: 1443 m_freem(m); 1444 if (n) 1445 m_freem(n); 1446 return (NULL); 1447} 1448#undef hostnamelen 1449 1450#define isupper(x) ('A' <= (x) && (x) <= 'Z') 1451#define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1452#define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1453#define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1454 1455/* 1456 * make a mbuf with DNS-encoded string. no compression support. 1457 * 1458 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1459 * treated as truncated name (two \0 at the end). this is a wild guess. 1460 * 1461 * old - return pascal string if non-zero 1462 */ 1463static struct mbuf * 1464ni6_nametodns(const char *name, int namelen, int old) 1465{ 1466 struct mbuf *m; 1467 char *cp, *ep; 1468 const char *p, *q; 1469 int i, len, nterm; 1470 1471 if (old) 1472 len = namelen + 1; 1473 else 1474 len = MCLBYTES; 1475 1476 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1477 MGET(m, M_DONTWAIT, MT_DATA); 1478 if (m && len > MLEN) { 1479 MCLGET(m, M_DONTWAIT); 1480 if ((m->m_flags & M_EXT) == 0) 1481 goto fail; 1482 } 1483 if (!m) 1484 goto fail; 1485 m->m_next = NULL; 1486 1487 if (old) { 1488 m->m_len = len; 1489 *mtod(m, char *) = namelen; 1490 bcopy(name, mtod(m, char *) + 1, namelen); 1491 return m; 1492 } else { 1493 m->m_len = 0; 1494 cp = mtod(m, char *); 1495 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1496 1497 /* if not certain about my name, return empty buffer */ 1498 if (namelen == 0) 1499 return m; 1500 1501 /* 1502 * guess if it looks like shortened hostname, or FQDN. 1503 * shortened hostname needs two trailing "\0". 1504 */ 1505 i = 0; 1506 for (p = name; p < name + namelen; p++) { 1507 if (*p && *p == '.') 1508 i++; 1509 } 1510 if (i < 2) 1511 nterm = 2; 1512 else 1513 nterm = 1; 1514 1515 p = name; 1516 while (cp < ep && p < name + namelen) { 1517 i = 0; 1518 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1519 i++; 1520 /* result does not fit into mbuf */ 1521 if (cp + i + 1 >= ep) 1522 goto fail; 1523 /* 1524 * DNS label length restriction, RFC1035 page 8. 1525 * "i == 0" case is included here to avoid returning 1526 * 0-length label on "foo..bar". 1527 */ 1528 if (i <= 0 || i >= 64) 1529 goto fail; 1530 *cp++ = i; 1531 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1532 goto fail; 1533 while (i > 0) { 1534 if (!isalnum(*p) && *p != '-') 1535 goto fail; 1536 if (isupper(*p)) { 1537 *cp++ = tolower(*p); 1538 p++; 1539 } else 1540 *cp++ = *p++; 1541 i--; 1542 } 1543 p = q; 1544 if (p < name + namelen && *p == '.') 1545 p++; 1546 } 1547 /* termination */ 1548 if (cp + nterm >= ep) 1549 goto fail; 1550 while (nterm-- > 0) 1551 *cp++ = '\0'; 1552 m->m_len = cp - mtod(m, char *); 1553 return m; 1554 } 1555 1556 panic("should not reach here"); 1557 /* NOTREACHED */ 1558 1559 fail: 1560 if (m) 1561 m_freem(m); 1562 return NULL; 1563} 1564 1565/* 1566 * check if two DNS-encoded string matches. takes care of truncated 1567 * form (with \0\0 at the end). no compression support. 1568 * XXX upper/lowercase match (see RFC2065) 1569 */ 1570static int 1571ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1572{ 1573 const char *a0, *b0; 1574 int l; 1575 1576 /* simplest case - need validation? */ 1577 if (alen == blen && memcmp(a, b, alen) == 0) 1578 return 1; 1579 1580 a0 = a; 1581 b0 = b; 1582 1583 /* termination is mandatory */ 1584 if (alen < 2 || blen < 2) 1585 return 0; 1586 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1587 return 0; 1588 alen--; 1589 blen--; 1590 1591 while (a - a0 < alen && b - b0 < blen) { 1592 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1593 return 0; 1594 1595 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1596 return 0; 1597 /* we don't support compression yet */ 1598 if (a[0] >= 64 || b[0] >= 64) 1599 return 0; 1600 1601 /* truncated case */ 1602 if (a[0] == 0 && a - a0 == alen - 1) 1603 return 1; 1604 if (b[0] == 0 && b - b0 == blen - 1) 1605 return 1; 1606 if (a[0] == 0 || b[0] == 0) 1607 return 0; 1608 1609 if (a[0] != b[0]) 1610 return 0; 1611 l = a[0]; 1612 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1613 return 0; 1614 if (memcmp(a + 1, b + 1, l) != 0) 1615 return 0; 1616 1617 a += 1 + l; 1618 b += 1 + l; 1619 } 1620 1621 if (a - a0 == alen && b - b0 == blen) 1622 return 1; 1623 else 1624 return 0; 1625} 1626 1627/* 1628 * calculate the number of addresses to be returned in the node info reply. 1629 */ 1630static int 1631ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, 1632 struct ifnet **ifpp, char *subj) 1633{ 1634 struct ifnet *ifp; 1635 struct in6_ifaddr *ifa6; 1636 struct ifaddr *ifa; 1637 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1638 int addrs = 0, addrsofif, iffound = 0; 1639 int niflags = ni6->ni_flags; 1640 1641 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1642 switch (ni6->ni_code) { 1643 case ICMP6_NI_SUBJ_IPV6: 1644 if (subj == NULL) /* must be impossible... */ 1645 return (0); 1646 subj_ip6 = (struct sockaddr_in6 *)subj; 1647 break; 1648 default: 1649 /* 1650 * XXX: we only support IPv6 subject address for 1651 * this Qtype. 1652 */ 1653 return (0); 1654 } 1655 } 1656 1657 IFNET_FOREACH(ifp) { 1658 addrsofif = 0; 1659 IFADDR_FOREACH(ifa, ifp) { 1660 if (ifa->ifa_addr->sa_family != AF_INET6) 1661 continue; 1662 ifa6 = (struct in6_ifaddr *)ifa; 1663 1664 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1665 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1666 &ifa6->ia_addr.sin6_addr)) 1667 iffound = 1; 1668 1669 /* 1670 * IPv4-mapped addresses can only be returned by a 1671 * Node Information proxy, since they represent 1672 * addresses of IPv4-only nodes, which perforce do 1673 * not implement this protocol. 1674 * [icmp-name-lookups-07, Section 5.4] 1675 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1676 * this function at this moment. 1677 */ 1678 1679 /* What do we have to do about ::1? */ 1680 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1681 case IPV6_ADDR_SCOPE_LINKLOCAL: 1682 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1683 continue; 1684 break; 1685 case IPV6_ADDR_SCOPE_SITELOCAL: 1686 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1687 continue; 1688 break; 1689 case IPV6_ADDR_SCOPE_GLOBAL: 1690 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1691 continue; 1692 break; 1693 default: 1694 continue; 1695 } 1696 1697 /* 1698 * check if anycast is okay. 1699 * XXX: just experimental. not in the spec. 1700 */ 1701 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1702 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1703 continue; /* we need only unicast addresses */ 1704 1705 addrsofif++; /* count the address */ 1706 } 1707 if (iffound) { 1708 *ifpp = ifp; 1709 return (addrsofif); 1710 } 1711 1712 addrs += addrsofif; 1713 } 1714 1715 return (addrs); 1716} 1717 1718static int 1719ni6_store_addrs(struct icmp6_nodeinfo *ni6, 1720 struct icmp6_nodeinfo *nni6, struct ifnet *ifp0, 1721 int resid) 1722{ 1723 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1724 struct in6_ifaddr *ifa6; 1725 struct ifaddr *ifa; 1726 struct ifnet *ifp_dep = NULL; 1727 int copied = 0, allow_deprecated = 0; 1728 u_char *cp = (u_char *)(nni6 + 1); 1729 int niflags = ni6->ni_flags; 1730 u_int32_t ltime; 1731 1732 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1733 return (0); /* needless to copy */ 1734 1735 again: 1736 1737 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1738 { 1739 IFADDR_FOREACH(ifa, ifp) { 1740 if (ifa->ifa_addr->sa_family != AF_INET6) 1741 continue; 1742 ifa6 = (struct in6_ifaddr *)ifa; 1743 1744 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1745 allow_deprecated == 0) { 1746 /* 1747 * prefererred address should be put before 1748 * deprecated addresses. 1749 */ 1750 1751 /* record the interface for later search */ 1752 if (ifp_dep == NULL) 1753 ifp_dep = ifp; 1754 1755 continue; 1756 } 1757 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1758 allow_deprecated != 0) 1759 continue; /* we now collect deprecated addrs */ 1760 1761 /* What do we have to do about ::1? */ 1762 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1763 case IPV6_ADDR_SCOPE_LINKLOCAL: 1764 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1765 continue; 1766 break; 1767 case IPV6_ADDR_SCOPE_SITELOCAL: 1768 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1769 continue; 1770 break; 1771 case IPV6_ADDR_SCOPE_GLOBAL: 1772 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1773 continue; 1774 break; 1775 default: 1776 continue; 1777 } 1778 1779 /* 1780 * check if anycast is okay. 1781 * XXX: just experimental. not in the spec. 1782 */ 1783 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1784 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1785 continue; 1786 1787 /* now we can copy the address */ 1788 if (resid < sizeof(struct in6_addr) + 1789 sizeof(u_int32_t)) { 1790 /* 1791 * We give up much more copy. 1792 * Set the truncate flag and return. 1793 */ 1794 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1795 return (copied); 1796 } 1797 1798 /* 1799 * Set the TTL of the address. 1800 * The TTL value should be one of the following 1801 * according to the specification: 1802 * 1803 * 1. The remaining lifetime of a DHCP lease on the 1804 * address, or 1805 * 2. The remaining Valid Lifetime of a prefix from 1806 * which the address was derived through Stateless 1807 * Autoconfiguration. 1808 * 1809 * Note that we currently do not support stateful 1810 * address configuration by DHCPv6, so the former 1811 * case can't happen. 1812 * 1813 * TTL must be 2^31 > TTL >= 0. 1814 */ 1815 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1816 ltime = ND6_INFINITE_LIFETIME; 1817 else { 1818 if (ifa6->ia6_lifetime.ia6t_expire > 1819 time_second) 1820 ltime = ifa6->ia6_lifetime.ia6t_expire - 1821 time_second; 1822 else 1823 ltime = 0; 1824 } 1825 if (ltime > 0x7fffffff) 1826 ltime = 0x7fffffff; 1827 ltime = htonl(ltime); 1828 1829 bcopy(<ime, cp, sizeof(u_int32_t)); 1830 cp += sizeof(u_int32_t); 1831 1832 /* copy the address itself */ 1833 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1834 sizeof(struct in6_addr)); 1835 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1836 cp += sizeof(struct in6_addr); 1837 1838 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1839 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1840 } 1841 if (ifp0) /* we need search only on the specified IF */ 1842 break; 1843 } 1844 1845 if (allow_deprecated == 0 && ifp_dep != NULL) { 1846 ifp = ifp_dep; 1847 allow_deprecated = 1; 1848 1849 goto again; 1850 } 1851 1852 return (copied); 1853} 1854 1855/* 1856 * XXX almost dup'ed code with rip6_input. 1857 */ 1858static int 1859icmp6_rip6_input(struct mbuf **mp, int off) 1860{ 1861 struct mbuf *m = *mp; 1862 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1863 struct inpcb_hdr *inph; 1864 struct in6pcb *in6p; 1865 struct in6pcb *last = NULL; 1866 struct sockaddr_in6 rip6src; 1867 struct icmp6_hdr *icmp6; 1868 struct mbuf *opts = NULL; 1869 1870 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1871 if (icmp6 == NULL) { 1872 /* m is already reclaimed */ 1873 return IPPROTO_DONE; 1874 } 1875 1876 /* 1877 * XXX: the address may have embedded scope zone ID, which should be 1878 * hidden from applications. 1879 */ 1880 sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0); 1881 if (sa6_recoverscope(&rip6src)) { 1882 m_freem(m); 1883 return (IPPROTO_DONE); 1884 } 1885 1886 CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) { 1887 in6p = (struct in6pcb *)inph; 1888 if (in6p->in6p_af != AF_INET6) 1889 continue; 1890 if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6) 1891 continue; 1892 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1893 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1894 continue; 1895 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1896 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1897 continue; 1898 if (in6p->in6p_icmp6filt 1899 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1900 in6p->in6p_icmp6filt)) 1901 continue; 1902 if (last) { 1903 struct mbuf *n; 1904 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1905 if (last->in6p_flags & IN6P_CONTROLOPTS) 1906 ip6_savecontrol(last, &opts, ip6, n); 1907 /* strip intermediate headers */ 1908 m_adj(n, off); 1909 if (sbappendaddr(&last->in6p_socket->so_rcv, 1910 (struct sockaddr *)&rip6src, 1911 n, opts) == 0) { 1912 /* should notify about lost packet */ 1913 m_freem(n); 1914 if (opts) 1915 m_freem(opts); 1916 } else 1917 sorwakeup(last->in6p_socket); 1918 opts = NULL; 1919 } 1920 } 1921 last = in6p; 1922 } 1923 if (last) { 1924 if (last->in6p_flags & IN6P_CONTROLOPTS) 1925 ip6_savecontrol(last, &opts, ip6, m); 1926 /* strip intermediate headers */ 1927 m_adj(m, off); 1928 if (sbappendaddr(&last->in6p_socket->so_rcv, 1929 (struct sockaddr *)&rip6src, m, opts) == 0) { 1930 m_freem(m); 1931 if (opts) 1932 m_freem(opts); 1933 } else 1934 sorwakeup(last->in6p_socket); 1935 } else { 1936 m_freem(m); 1937 IP6_STATDEC(IP6_STAT_DELIVERED); 1938 } 1939 return IPPROTO_DONE; 1940} 1941 1942/* 1943 * Reflect the ip6 packet back to the source. 1944 * OFF points to the icmp6 header, counted from the top of the mbuf. 1945 * 1946 * Note: RFC 1885 required that an echo reply should be truncated if it 1947 * did not fit in with (return) path MTU, and KAME code supported the 1948 * behavior. However, as a clarification after the RFC, this limitation 1949 * was removed in a revised version of the spec, RFC 2463. We had kept the 1950 * old behavior, with a (non-default) ifdef block, while the new version of 1951 * the spec was an internet-draft status, and even after the new RFC was 1952 * published. But it would rather make sense to clean the obsoleted part 1953 * up, and to make the code simpler at this stage. 1954 */ 1955void 1956icmp6_reflect(struct mbuf *m, size_t off) 1957{ 1958 struct ip6_hdr *ip6; 1959 struct icmp6_hdr *icmp6; 1960 const struct in6_ifaddr *ia; 1961 const struct ip6aux *ip6a; 1962 int plen; 1963 int type, code; 1964 struct ifnet *outif = NULL; 1965 struct in6_addr origdst; 1966 const struct in6_addr *src = NULL; 1967 1968 /* too short to reflect */ 1969 if (off < sizeof(struct ip6_hdr)) { 1970 nd6log((LOG_DEBUG, 1971 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 1972 (u_long)off, (u_long)sizeof(struct ip6_hdr), 1973 __FILE__, __LINE__)); 1974 goto bad; 1975 } 1976 1977 /* 1978 * If there are extra headers between IPv6 and ICMPv6, strip 1979 * off that header first. 1980 */ 1981#ifdef DIAGNOSTIC 1982 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 1983 panic("assumption failed in icmp6_reflect"); 1984#endif 1985 if (off > sizeof(struct ip6_hdr)) { 1986 size_t l; 1987 struct ip6_hdr nip6; 1988 1989 l = off - sizeof(struct ip6_hdr); 1990 m_copydata(m, 0, sizeof(nip6), (void *)&nip6); 1991 m_adj(m, l); 1992 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1993 if (m->m_len < l) { 1994 if ((m = m_pullup(m, l)) == NULL) 1995 return; 1996 } 1997 bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6)); 1998 } else /* off == sizeof(struct ip6_hdr) */ { 1999 size_t l; 2000 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2001 if (m->m_len < l) { 2002 if ((m = m_pullup(m, l)) == NULL) 2003 return; 2004 } 2005 } 2006 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2007 ip6 = mtod(m, struct ip6_hdr *); 2008 ip6->ip6_nxt = IPPROTO_ICMPV6; 2009 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2010 type = icmp6->icmp6_type; /* keep type for statistics */ 2011 code = icmp6->icmp6_code; /* ditto. */ 2012 2013 origdst = ip6->ip6_dst; 2014 /* 2015 * ip6_input() drops a packet if its src is multicast. 2016 * So, the src is never multicast. 2017 */ 2018 ip6->ip6_dst = ip6->ip6_src; 2019 2020 /* 2021 * If the incoming packet was addressed directly to us (i.e. unicast), 2022 * use dst as the src for the reply. 2023 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2024 * (for example) when we encounter an error while forwarding procedure 2025 * destined to a duplicated address of ours. 2026 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2027 * procedure of an outgoing packet of our own, in which case we need 2028 * to search in the ifaddr list. 2029 */ 2030 if (IN6_IS_ADDR_MULTICAST(&origdst)) 2031 ; 2032 else if ((ip6a = ip6_getdstifaddr(m)) != NULL) { 2033 if ((ip6a->ip6a_flags & 2034 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2035 src = &ip6a->ip6a_src; 2036 } else { 2037 union { 2038 struct sockaddr_in6 sin6; 2039 struct sockaddr sa; 2040 } u; 2041 2042 sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0); 2043 2044 ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa); 2045 2046 if (ia == NULL) 2047 ; 2048 else if ((ia->ia6_flags & 2049 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2050 src = &ia->ia_addr.sin6_addr; 2051 } 2052 2053 if (src == NULL) { 2054 int e; 2055 struct sockaddr_in6 sin6; 2056 struct route ro; 2057 2058 /* 2059 * This case matches to multicasts, our anycast, or unicasts 2060 * that we do not own. Select a source address based on the 2061 * source address of the erroneous packet. 2062 */ 2063 /* zone ID should be embedded */ 2064 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 2065 2066 memset(&ro, 0, sizeof(ro)); 2067 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2068 rtcache_free(&ro); 2069 if (src == NULL) { 2070 nd6log((LOG_DEBUG, 2071 "icmp6_reflect: source can't be determined: " 2072 "dst=%s, error=%d\n", 2073 ip6_sprintf(&sin6.sin6_addr), e)); 2074 goto bad; 2075 } 2076 } 2077 2078 ip6->ip6_src = *src; 2079 ip6->ip6_flow = 0; 2080 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2081 ip6->ip6_vfc |= IPV6_VERSION; 2082 ip6->ip6_nxt = IPPROTO_ICMPV6; 2083 if (m->m_pkthdr.rcvif) { 2084 /* XXX: This may not be the outgoing interface */ 2085 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2086 } else 2087 ip6->ip6_hlim = ip6_defhlim; 2088 2089 m->m_pkthdr.csum_flags = 0; 2090 icmp6->icmp6_cksum = 0; 2091 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2092 sizeof(struct ip6_hdr), plen); 2093 2094 /* 2095 * XXX option handling 2096 */ 2097 2098 m->m_flags &= ~(M_BCAST|M_MCAST); 2099 2100 /* 2101 * To avoid a "too big" situation at an intermediate router 2102 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2103 * Note that only echo and node information replies are affected, 2104 * since the length of ICMP6 errors is limited to the minimum MTU. 2105 */ 2106 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 && 2107 outif) 2108 icmp6_ifstat_inc(outif, ifs6_out_error); 2109 2110 if (outif) 2111 icmp6_ifoutstat_inc(outif, type, code); 2112 2113 return; 2114 2115 bad: 2116 m_freem(m); 2117 return; 2118} 2119 2120static const char * 2121icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2122 struct in6_addr *tgt6) 2123{ 2124 static char buf[1024]; 2125 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2126 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2127 return buf; 2128} 2129 2130void 2131icmp6_redirect_input(struct mbuf *m, int off) 2132{ 2133 struct ifnet *ifp = m->m_pkthdr.rcvif; 2134 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2135 struct nd_redirect *nd_rd; 2136 int icmp6len = ntohs(ip6->ip6_plen); 2137 char *lladdr = NULL; 2138 int lladdrlen = 0; 2139 struct rtentry *rt = NULL; 2140 int is_router; 2141 int is_onlink; 2142 struct in6_addr src6 = ip6->ip6_src; 2143 struct in6_addr redtgt6; 2144 struct in6_addr reddst6; 2145 union nd_opts ndopts; 2146 2147 if (!ifp) 2148 return; 2149 2150 /* XXX if we are router, we don't update route by icmp6 redirect */ 2151 if (ip6_forwarding) 2152 goto freeit; 2153 if (!icmp6_rediraccept) 2154 goto freeit; 2155 2156 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2157 if (nd_rd == NULL) { 2158 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 2159 return; 2160 } 2161 redtgt6 = nd_rd->nd_rd_target; 2162 reddst6 = nd_rd->nd_rd_dst; 2163 2164 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2165 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2166 goto freeit; 2167 } 2168 2169 /* validation */ 2170 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2171 nd6log((LOG_ERR, 2172 "ICMP6 redirect sent from %s rejected; " 2173 "must be from linklocal\n", ip6_sprintf(&src6))); 2174 goto bad; 2175 } 2176 if (ip6->ip6_hlim != 255) { 2177 nd6log((LOG_ERR, 2178 "ICMP6 redirect sent from %s rejected; " 2179 "hlim=%d (must be 255)\n", 2180 ip6_sprintf(&src6), ip6->ip6_hlim)); 2181 goto bad; 2182 } 2183 { 2184 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2185 struct sockaddr_in6 sin6; 2186 struct in6_addr *gw6; 2187 2188 sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0); 2189 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2190 if (rt) { 2191 if (rt->rt_gateway == NULL || 2192 rt->rt_gateway->sa_family != AF_INET6) { 2193 nd6log((LOG_ERR, 2194 "ICMP6 redirect rejected; no route " 2195 "with inet6 gateway found for redirect dst: %s\n", 2196 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2197 RTFREE(rt); 2198 goto bad; 2199 } 2200 2201 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2202 if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2203 nd6log((LOG_ERR, 2204 "ICMP6 redirect rejected; " 2205 "not equal to gw-for-src=%s (must be same): " 2206 "%s\n", 2207 ip6_sprintf(gw6), 2208 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2209 RTFREE(rt); 2210 goto bad; 2211 } 2212 } else { 2213 nd6log((LOG_ERR, 2214 "ICMP6 redirect rejected; " 2215 "no route found for redirect dst: %s\n", 2216 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2217 goto bad; 2218 } 2219 RTFREE(rt); 2220 rt = NULL; 2221 } 2222 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2223 nd6log((LOG_ERR, 2224 "ICMP6 redirect rejected; " 2225 "redirect dst must be unicast: %s\n", 2226 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2227 goto bad; 2228 } 2229 2230 is_router = is_onlink = 0; 2231 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2232 is_router = 1; /* router case */ 2233 if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2234 is_onlink = 1; /* on-link destination case */ 2235 if (!is_router && !is_onlink) { 2236 nd6log((LOG_ERR, 2237 "ICMP6 redirect rejected; " 2238 "neither router case nor onlink case: %s\n", 2239 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2240 goto bad; 2241 } 2242 /* validation passed */ 2243 2244 icmp6len -= sizeof(*nd_rd); 2245 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2246 if (nd6_options(&ndopts) < 0) { 2247 nd6log((LOG_INFO, "icmp6_redirect_input: " 2248 "invalid ND option, rejected: %s\n", 2249 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2250 /* nd6_options have incremented stats */ 2251 goto freeit; 2252 } 2253 2254 if (ndopts.nd_opts_tgt_lladdr) { 2255 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2256 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2257 } 2258 2259 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2260 nd6log((LOG_INFO, 2261 "icmp6_redirect_input: lladdrlen mismatch for %s " 2262 "(if %d, icmp6 packet %d): %s\n", 2263 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2264 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2265 goto bad; 2266 } 2267 2268 /* RFC 2461 8.3 */ 2269 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2270 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2271 2272 if (!is_onlink) { /* better router case. perform rtredirect. */ 2273 /* perform rtredirect */ 2274 struct sockaddr_in6 sdst; 2275 struct sockaddr_in6 sgw; 2276 struct sockaddr_in6 ssrc; 2277 unsigned long rtcount; 2278 struct rtentry *newrt = NULL; 2279 2280 /* 2281 * do not install redirect route, if the number of entries 2282 * is too much (> hiwat). note that, the node (= host) will 2283 * work just fine even if we do not install redirect route 2284 * (there will be additional hops, though). 2285 */ 2286 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2287 if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes) 2288 goto freeit; 2289 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2290 return; 2291 else if (0 <= icmp6_redirect_lowat && 2292 rtcount > icmp6_redirect_lowat) { 2293 /* 2294 * XXX nuke a victim, install the new one. 2295 */ 2296 } 2297 2298 memset(&sdst, 0, sizeof(sdst)); 2299 memset(&sgw, 0, sizeof(sgw)); 2300 memset(&ssrc, 0, sizeof(ssrc)); 2301 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2302 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2303 sizeof(struct sockaddr_in6); 2304 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2305 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2306 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2307 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2308 NULL, RTF_GATEWAY | RTF_HOST, 2309 (struct sockaddr *)&ssrc, 2310 &newrt); 2311 2312 if (newrt) { 2313 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2314 icmp6_redirect_timeout_q); 2315 rtfree(newrt); 2316 } 2317 } 2318 /* finally update cached route in each socket via pfctlinput */ 2319 { 2320 struct sockaddr_in6 sdst; 2321 2322 sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0); 2323 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2324#if defined(KAME_IPSEC) || defined(FAST_IPSEC) 2325 key_sa_routechange((struct sockaddr *)&sdst); 2326#endif 2327 } 2328 2329 freeit: 2330 m_freem(m); 2331 return; 2332 2333 bad: 2334 ICMP6_STATINC(ICMP6_STAT_BADREDIRECT); 2335 m_freem(m); 2336} 2337 2338void 2339icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2340{ 2341 struct ifnet *ifp; /* my outgoing interface */ 2342 struct in6_addr *ifp_ll6; 2343 struct in6_addr *nexthop; 2344 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2345 struct mbuf *m = NULL; /* newly allocated one */ 2346 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2347 struct nd_redirect *nd_rd; 2348 size_t maxlen; 2349 u_char *p; 2350 struct sockaddr_in6 src_sa; 2351 2352 icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0); 2353 2354 /* if we are not router, we don't send icmp6 redirect */ 2355 if (!ip6_forwarding) 2356 goto fail; 2357 2358 /* sanity check */ 2359 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2360 goto fail; 2361 2362 /* 2363 * Address check: 2364 * the source address must identify a neighbor, and 2365 * the destination address must not be a multicast address 2366 * [RFC 2461, sec 8.2] 2367 */ 2368 sip6 = mtod(m0, struct ip6_hdr *); 2369 sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0); 2370 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2371 goto fail; 2372 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2373 goto fail; /* what should we do here? */ 2374 2375 /* rate limit */ 2376 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2377 goto fail; 2378 2379 /* 2380 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2381 * we almost always ask for an mbuf cluster for simplicity. 2382 * (MHLEN < IPV6_MMTU is almost always true) 2383 */ 2384 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2385 if (m && IPV6_MMTU >= MHLEN) { 2386#if IPV6_MMTU >= MCLBYTES 2387 _MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT); 2388#else 2389 MCLGET(m, M_DONTWAIT); 2390#endif 2391 } 2392 2393 if (!m) 2394 goto fail; 2395 m->m_pkthdr.rcvif = NULL; 2396 m->m_len = 0; 2397 maxlen = M_TRAILINGSPACE(m); 2398 maxlen = min(IPV6_MMTU, maxlen); 2399 /* just for safety */ 2400 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2401 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2402 goto fail; 2403 } 2404 2405 { 2406 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2407 struct in6_ifaddr *ia; 2408 if ((ia = in6ifa_ifpforlinklocal(ifp, 2409 IN6_IFF_NOTREADY| 2410 IN6_IFF_ANYCAST)) == NULL) 2411 goto fail; 2412 ifp_ll6 = &ia->ia_addr.sin6_addr; 2413 } 2414 2415 /* get ip6 linklocal address for the router. */ 2416 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2417 struct sockaddr_in6 *sin6; 2418 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2419 nexthop = &sin6->sin6_addr; 2420 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2421 nexthop = NULL; 2422 } else 2423 nexthop = NULL; 2424 2425 /* ip6 */ 2426 ip6 = mtod(m, struct ip6_hdr *); 2427 ip6->ip6_flow = 0; 2428 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2429 ip6->ip6_vfc |= IPV6_VERSION; 2430 /* ip6->ip6_plen will be set later */ 2431 ip6->ip6_nxt = IPPROTO_ICMPV6; 2432 ip6->ip6_hlim = 255; 2433 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2434 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2435 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2436 2437 /* ND Redirect */ 2438 nd_rd = (struct nd_redirect *)(ip6 + 1); 2439 nd_rd->nd_rd_type = ND_REDIRECT; 2440 nd_rd->nd_rd_code = 0; 2441 nd_rd->nd_rd_reserved = 0; 2442 if (rt->rt_flags & RTF_GATEWAY) { 2443 /* 2444 * nd_rd->nd_rd_target must be a link-local address in 2445 * better router cases. 2446 */ 2447 if (!nexthop) 2448 goto fail; 2449 bcopy(nexthop, &nd_rd->nd_rd_target, 2450 sizeof(nd_rd->nd_rd_target)); 2451 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2452 sizeof(nd_rd->nd_rd_dst)); 2453 } else { 2454 /* make sure redtgt == reddst */ 2455 nexthop = &sip6->ip6_dst; 2456 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2457 sizeof(nd_rd->nd_rd_target)); 2458 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2459 sizeof(nd_rd->nd_rd_dst)); 2460 } 2461 2462 p = (u_char *)(nd_rd + 1); 2463 2464 { 2465 /* target lladdr option */ 2466 struct rtentry *rt_nexthop = NULL; 2467 int len; 2468 const struct sockaddr_dl *sdl; 2469 struct nd_opt_hdr *nd_opt; 2470 char *lladdr; 2471 2472 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2473 if (!rt_nexthop) 2474 goto nolladdropt; 2475 len = sizeof(*nd_opt) + ifp->if_addrlen; 2476 len = (len + 7) & ~7; /* round by 8 */ 2477 /* safety check */ 2478 if (len + (p - (u_char *)ip6) > maxlen) 2479 goto nolladdropt; 2480 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2481 (rt_nexthop->rt_flags & RTF_LLINFO) && 2482 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2483 (sdl = satocsdl(rt_nexthop->rt_gateway)) && 2484 sdl->sdl_alen) { 2485 nd_opt = (struct nd_opt_hdr *)p; 2486 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2487 nd_opt->nd_opt_len = len >> 3; 2488 lladdr = (char *)(nd_opt + 1); 2489 memcpy(lladdr, CLLADDR(sdl), ifp->if_addrlen); 2490 p += len; 2491 } 2492 } 2493 nolladdropt:; 2494 2495 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2496 2497 /* just to be safe */ 2498 if (m0->m_flags & M_DECRYPTED) 2499 goto noredhdropt; 2500 if (p - (u_char *)ip6 > maxlen) 2501 goto noredhdropt; 2502 2503 { 2504 /* redirected header option */ 2505 int len; 2506 struct nd_opt_rd_hdr *nd_opt_rh; 2507 2508 /* 2509 * compute the maximum size for icmp6 redirect header option. 2510 * XXX room for auth header? 2511 */ 2512 len = maxlen - (p - (u_char *)ip6); 2513 len &= ~7; 2514 2515 /* 2516 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2517 * about padding/truncate rule for the original IP packet. 2518 * From the discussion on IPv6imp in Feb 1999, 2519 * the consensus was: 2520 * - "attach as much as possible" is the goal 2521 * - pad if not aligned (original size can be guessed by 2522 * original ip6 header) 2523 * Following code adds the padding if it is simple enough, 2524 * and truncates if not. 2525 */ 2526 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2527 /* not enough room, truncate */ 2528 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2529 m0->m_pkthdr.len); 2530 } else { 2531 /* 2532 * enough room, truncate if not aligned. 2533 * we don't pad here for simplicity. 2534 */ 2535 size_t extra; 2536 2537 extra = m0->m_pkthdr.len % 8; 2538 if (extra) { 2539 /* truncate */ 2540 m_adj(m0, -extra); 2541 } 2542 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2543 } 2544 2545 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2546 memset(nd_opt_rh, 0, sizeof(*nd_opt_rh)); 2547 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2548 nd_opt_rh->nd_opt_rh_len = len >> 3; 2549 p += sizeof(*nd_opt_rh); 2550 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2551 2552 /* connect m0 to m */ 2553 m->m_pkthdr.len += m0->m_pkthdr.len; 2554 m_cat(m, m0); 2555 m0 = NULL; 2556 } 2557noredhdropt: 2558 if (m0) { 2559 m_freem(m0); 2560 m0 = NULL; 2561 } 2562 2563 /* XXX: clear embedded link IDs in the inner header */ 2564 in6_clearscope(&sip6->ip6_src); 2565 in6_clearscope(&sip6->ip6_dst); 2566 in6_clearscope(&nd_rd->nd_rd_target); 2567 in6_clearscope(&nd_rd->nd_rd_dst); 2568 2569 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2570 2571 nd_rd->nd_rd_cksum = 0; 2572 nd_rd->nd_rd_cksum 2573 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2574 2575 /* send the packet to outside... */ 2576 if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0) 2577 icmp6_ifstat_inc(ifp, ifs6_out_error); 2578 2579 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2580 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2581 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT); 2582 2583 return; 2584 2585fail: 2586 if (m) 2587 m_freem(m); 2588 if (m0) 2589 m_freem(m0); 2590} 2591 2592/* 2593 * ICMPv6 socket option processing. 2594 */ 2595int 2596icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 2597{ 2598 int error = 0; 2599 struct in6pcb *in6p = sotoin6pcb(so); 2600 2601 if (sopt->sopt_level != IPPROTO_ICMPV6) 2602 return rip6_ctloutput(op, so, sopt); 2603 2604 switch (op) { 2605 case PRCO_SETOPT: 2606 switch (sopt->sopt_name) { 2607 case ICMP6_FILTER: 2608 { 2609 struct icmp6_filter fil; 2610 2611 error = sockopt_get(sopt, &fil, sizeof(fil)); 2612 if (error) 2613 break; 2614 memcpy(in6p->in6p_icmp6filt, &fil, 2615 sizeof(struct icmp6_filter)); 2616 error = 0; 2617 break; 2618 } 2619 2620 default: 2621 error = ENOPROTOOPT; 2622 break; 2623 } 2624 break; 2625 2626 case PRCO_GETOPT: 2627 switch (sopt->sopt_name) { 2628 case ICMP6_FILTER: 2629 { 2630 if (in6p->in6p_icmp6filt == NULL) { 2631 error = EINVAL; 2632 break; 2633 } 2634 error = sockopt_set(sopt, in6p->in6p_icmp6filt, 2635 sizeof(struct icmp6_filter)); 2636 break; 2637 } 2638 2639 default: 2640 error = ENOPROTOOPT; 2641 break; 2642 } 2643 break; 2644 } 2645 2646 return (error); 2647} 2648 2649/* 2650 * Perform rate limit check. 2651 * Returns 0 if it is okay to send the icmp6 packet. 2652 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2653 * limitation. 2654 * 2655 * XXX per-destination/type check necessary? 2656 */ 2657static int 2658icmp6_ratelimit( 2659 const struct in6_addr *dst, /* not used at this moment */ 2660 const int type, /* not used at this moment */ 2661 const int code) /* not used at this moment */ 2662{ 2663 int ret; 2664 2665 ret = 0; /* okay to send */ 2666 2667 /* PPS limit */ 2668 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2669 icmp6errppslim)) { 2670 /* The packet is subject to rate limit */ 2671 ret++; 2672 } 2673 2674 return ret; 2675} 2676 2677static struct rtentry * 2678icmp6_mtudisc_clone(struct sockaddr *dst) 2679{ 2680 struct rtentry *rt; 2681 int error; 2682 2683 rt = rtalloc1(dst, 1); 2684 if (rt == 0) 2685 return NULL; 2686 2687 /* If we didn't get a host route, allocate one */ 2688 if ((rt->rt_flags & RTF_HOST) == 0) { 2689 struct rtentry *nrt; 2690 2691 error = rtrequest((int) RTM_ADD, dst, 2692 (struct sockaddr *) rt->rt_gateway, NULL, 2693 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2694 if (error) { 2695 rtfree(rt); 2696 return NULL; 2697 } 2698 nrt->rt_rmx = rt->rt_rmx; 2699 rtfree(rt); 2700 rt = nrt; 2701 } 2702 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2703 icmp6_mtudisc_timeout_q); 2704 if (error) { 2705 rtfree(rt); 2706 return NULL; 2707 } 2708 2709 return rt; /* caller need to call rtfree() */ 2710} 2711 2712static void 2713icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 2714{ 2715 if (rt == NULL) 2716 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2717 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2718 (RTF_DYNAMIC | RTF_HOST)) { 2719 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2720 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2721 } else { 2722 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2723 rt->rt_rmx.rmx_mtu = 0; 2724 } 2725} 2726 2727static void 2728icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r) 2729{ 2730 if (rt == NULL) 2731 panic("icmp6_redirect_timeout: bad route to timeout"); 2732 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2733 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2734 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2735 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2736 } 2737} 2738 2739/* 2740 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes. silly? 2741 */ 2742static int 2743sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS) 2744{ 2745 (void)&name; 2746 (void)&l; 2747 (void)&oname; 2748 2749 if (namelen != 0) 2750 return (EINVAL); 2751 2752 return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp, 2753 /*XXXUNCONST*/ 2754 __UNCONST(newp), newlen)); 2755} 2756 2757static int 2758sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS) 2759{ 2760 2761 return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS)); 2762} 2763 2764static void 2765sysctl_net_inet6_icmp6_setup(struct sysctllog **clog) 2766{ 2767 extern int nd6_maxqueuelen; /* defined in nd6.c */ 2768 2769 sysctl_createv(clog, 0, NULL, NULL, 2770 CTLFLAG_PERMANENT, 2771 CTLTYPE_NODE, "net", NULL, 2772 NULL, 0, NULL, 0, 2773 CTL_NET, CTL_EOL); 2774 sysctl_createv(clog, 0, NULL, NULL, 2775 CTLFLAG_PERMANENT, 2776 CTLTYPE_NODE, "inet6", NULL, 2777 NULL, 0, NULL, 0, 2778 CTL_NET, PF_INET6, CTL_EOL); 2779 sysctl_createv(clog, 0, NULL, NULL, 2780 CTLFLAG_PERMANENT, 2781 CTLTYPE_NODE, "icmp6", 2782 SYSCTL_DESCR("ICMPv6 related settings"), 2783 NULL, 0, NULL, 0, 2784 CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL); 2785 2786 sysctl_createv(clog, 0, NULL, NULL, 2787 CTLFLAG_PERMANENT, 2788 CTLTYPE_STRUCT, "stats", 2789 SYSCTL_DESCR("ICMPv6 transmission statistics"), 2790 sysctl_net_inet6_icmp6_stats, 0, NULL, 0, 2791 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2792 ICMPV6CTL_STATS, CTL_EOL); 2793 sysctl_createv(clog, 0, NULL, NULL, 2794 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2795 CTLTYPE_INT, "rediraccept", 2796 SYSCTL_DESCR("Accept and process redirect messages"), 2797 NULL, 0, &icmp6_rediraccept, 0, 2798 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2799 ICMPV6CTL_REDIRACCEPT, CTL_EOL); 2800 sysctl_createv(clog, 0, NULL, NULL, 2801 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2802 CTLTYPE_INT, "redirtimeout", 2803 SYSCTL_DESCR("Redirect generated route lifetime"), 2804 NULL, 0, &icmp6_redirtimeout, 0, 2805 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2806 ICMPV6CTL_REDIRTIMEOUT, CTL_EOL); 2807#if 0 /* obsoleted */ 2808 sysctl_createv(clog, 0, NULL, NULL, 2809 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2810 CTLTYPE_INT, "errratelimit", NULL, 2811 NULL, 0, &icmp6_errratelimit, 0, 2812 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2813 ICMPV6CTL_ERRRATELIMIT, CTL_EOL); 2814#endif 2815 sysctl_createv(clog, 0, NULL, NULL, 2816 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2817 CTLTYPE_INT, "nd6_prune", 2818 SYSCTL_DESCR("Neighbor discovery prune interval"), 2819 NULL, 0, &nd6_prune, 0, 2820 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2821 ICMPV6CTL_ND6_PRUNE, CTL_EOL); 2822 sysctl_createv(clog, 0, NULL, NULL, 2823 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2824 CTLTYPE_INT, "nd6_delay", 2825 SYSCTL_DESCR("First probe delay time"), 2826 NULL, 0, &nd6_delay, 0, 2827 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2828 ICMPV6CTL_ND6_DELAY, CTL_EOL); 2829 sysctl_createv(clog, 0, NULL, NULL, 2830 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2831 CTLTYPE_INT, "nd6_umaxtries", 2832 SYSCTL_DESCR("Number of unicast discovery attempts"), 2833 NULL, 0, &nd6_umaxtries, 0, 2834 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2835 ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL); 2836 sysctl_createv(clog, 0, NULL, NULL, 2837 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2838 CTLTYPE_INT, "nd6_mmaxtries", 2839 SYSCTL_DESCR("Number of multicast discovery attempts"), 2840 NULL, 0, &nd6_mmaxtries, 0, 2841 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2842 ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL); 2843 sysctl_createv(clog, 0, NULL, NULL, 2844 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2845 CTLTYPE_INT, "nd6_useloopback", 2846 SYSCTL_DESCR("Use loopback interface for local traffic"), 2847 NULL, 0, &nd6_useloopback, 0, 2848 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2849 ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL); 2850#if 0 /* obsoleted */ 2851 sysctl_createv(clog, 0, NULL, NULL, 2852 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2853 CTLTYPE_INT, "nd6_proxyall", NULL, 2854 NULL, 0, &nd6_proxyall, 0, 2855 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2856 ICMPV6CTL_ND6_PROXYALL, CTL_EOL); 2857#endif 2858 sysctl_createv(clog, 0, NULL, NULL, 2859 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2860 CTLTYPE_INT, "nodeinfo", 2861 SYSCTL_DESCR("Respond to node information requests"), 2862 NULL, 0, &icmp6_nodeinfo, 0, 2863 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2864 ICMPV6CTL_NODEINFO, CTL_EOL); 2865 sysctl_createv(clog, 0, NULL, NULL, 2866 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2867 CTLTYPE_INT, "errppslimit", 2868 SYSCTL_DESCR("Maximum ICMP errors sent per second"), 2869 NULL, 0, &icmp6errppslim, 0, 2870 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2871 ICMPV6CTL_ERRPPSLIMIT, CTL_EOL); 2872 sysctl_createv(clog, 0, NULL, NULL, 2873 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2874 CTLTYPE_INT, "nd6_maxnudhint", 2875 SYSCTL_DESCR("Maximum neighbor unreachable hint count"), 2876 NULL, 0, &nd6_maxnudhint, 0, 2877 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2878 ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL); 2879 sysctl_createv(clog, 0, NULL, NULL, 2880 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2881 CTLTYPE_INT, "mtudisc_hiwat", 2882 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2883 NULL, 0, &icmp6_mtudisc_hiwat, 0, 2884 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2885 ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL); 2886 sysctl_createv(clog, 0, NULL, NULL, 2887 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2888 CTLTYPE_INT, "mtudisc_lowat", 2889 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2890 NULL, 0, &icmp6_mtudisc_lowat, 0, 2891 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2892 ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL); 2893 sysctl_createv(clog, 0, NULL, NULL, 2894 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2895 CTLTYPE_INT, "nd6_debug", 2896 SYSCTL_DESCR("Enable neighbor discovery debug output"), 2897 NULL, 0, &nd6_debug, 0, 2898 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2899 ICMPV6CTL_ND6_DEBUG, CTL_EOL); 2900 sysctl_createv(clog, 0, NULL, NULL, 2901 CTLFLAG_PERMANENT, 2902 CTLTYPE_STRUCT, "nd6_drlist", 2903 SYSCTL_DESCR("Default router list"), 2904 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2905 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2906 ICMPV6CTL_ND6_DRLIST, CTL_EOL); 2907 sysctl_createv(clog, 0, NULL, NULL, 2908 CTLFLAG_PERMANENT, 2909 CTLTYPE_STRUCT, "nd6_prlist", 2910 SYSCTL_DESCR("Prefix list"), 2911 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2912 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2913 ICMPV6CTL_ND6_PRLIST, CTL_EOL); 2914 sysctl_createv(clog, 0, NULL, NULL, 2915 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2916 CTLTYPE_INT, "maxqueuelen", 2917 SYSCTL_DESCR("max packet queue len for a unresolved ND"), 2918 NULL, 1, &nd6_maxqueuelen, 0, 2919 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2920 ICMPV6CTL_ND6_MAXQLEN, CTL_EOL); 2921} 2922 2923void 2924icmp6_statinc(u_int stat) 2925{ 2926 2927 KASSERT(stat < ICMP6_NSTATS); 2928 ICMP6_STATINC(stat); 2929} 2930