icmp6.c revision 1.145
1/* $NetBSD: icmp6.c,v 1.145 2008/04/15 03:57:04 thorpej 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 1.145 2008/04/15 03:57:04 thorpej 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 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 *); 173 174 175void 176icmp6_init(void) 177{ 178 mld_init(); 179 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 180 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 181 182 icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS); 183} 184 185static void 186icmp6_errcount(u_int base, int type, int code) 187{ 188 switch (type) { 189 case ICMP6_DST_UNREACH: 190 switch (code) { 191 case ICMP6_DST_UNREACH_NOROUTE: 192 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE); 193 return; 194 case ICMP6_DST_UNREACH_ADMIN: 195 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN); 196 return; 197 case ICMP6_DST_UNREACH_BEYONDSCOPE: 198 ICMP6_STATINC(base + 199 ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE); 200 return; 201 case ICMP6_DST_UNREACH_ADDR: 202 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR); 203 return; 204 case ICMP6_DST_UNREACH_NOPORT: 205 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT); 206 return; 207 } 208 break; 209 case ICMP6_PACKET_TOO_BIG: 210 ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG); 211 return; 212 case ICMP6_TIME_EXCEEDED: 213 switch (code) { 214 case ICMP6_TIME_EXCEED_TRANSIT: 215 ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT); 216 return; 217 case ICMP6_TIME_EXCEED_REASSEMBLY: 218 ICMP6_STATINC(base + 219 ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY); 220 return; 221 } 222 break; 223 case ICMP6_PARAM_PROB: 224 switch (code) { 225 case ICMP6_PARAMPROB_HEADER: 226 ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER); 227 return; 228 case ICMP6_PARAMPROB_NEXTHEADER: 229 ICMP6_STATINC(base + 230 ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER); 231 return; 232 case ICMP6_PARAMPROB_OPTION: 233 ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION); 234 return; 235 } 236 break; 237 case ND_REDIRECT: 238 ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT); 239 return; 240 } 241 ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN); 242} 243 244/* 245 * Register a Path MTU Discovery callback. 246 */ 247void 248icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *)) 249{ 250 struct icmp6_mtudisc_callback *mc; 251 252 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 253 mc = LIST_NEXT(mc, mc_list)) { 254 if (mc->mc_func == func) 255 return; 256 } 257 258 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 259 if (mc == NULL) 260 panic("icmp6_mtudisc_callback_register"); 261 262 mc->mc_func = func; 263 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 264} 265 266/* 267 * A wrapper function for icmp6_error() necessary when the erroneous packet 268 * may not contain enough scope zone information. 269 */ 270void 271icmp6_error2(struct mbuf *m, int type, int code, int param, 272 struct ifnet *ifp) 273{ 274 struct ip6_hdr *ip6; 275 276 if (ifp == NULL) 277 return; 278 279 if (m->m_len < sizeof(struct ip6_hdr)) { 280 m = m_pullup(m, sizeof(struct ip6_hdr)); 281 if (m == NULL) 282 return; 283 } 284 285 ip6 = mtod(m, struct ip6_hdr *); 286 287 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 288 return; 289 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 290 return; 291 292 icmp6_error(m, type, code, param); 293} 294 295/* 296 * Generate an error packet of type error in response to bad IP6 packet. 297 */ 298void 299icmp6_error(struct mbuf *m, int type, int code, int 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 ICMP6_STATINC(ICMP6_STAT_ERROR); 308 309 /* count per-type-code statistics */ 310 icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code); 311 312 if (m->m_flags & M_DECRYPTED) { 313 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 314 goto freeit; 315 } 316 317 if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) && 318 (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) 319 return; 320 oip6 = mtod(m, struct ip6_hdr *); 321 322 /* 323 * If the destination address of the erroneous packet is a multicast 324 * address, or the packet was sent using link-layer multicast, 325 * we should basically suppress sending an error (RFC 2463, Section 326 * 2.4). 327 * We have two exceptions (the item e.2 in that section): 328 * - the Pakcet Too Big message can be sent for path MTU discovery. 329 * - the Parameter Problem Message that can be allowed an icmp6 error 330 * in the option type field. This check has been done in 331 * ip6_unknown_opt(), so we can just check the type and code. 332 */ 333 if ((m->m_flags & (M_BCAST|M_MCAST) || 334 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 335 (type != ICMP6_PACKET_TOO_BIG && 336 (type != ICMP6_PARAM_PROB || 337 code != ICMP6_PARAMPROB_OPTION))) 338 goto freeit; 339 340 /* 341 * RFC 2463, 2.4 (e.5): source address check. 342 * XXX: the case of anycast source? 343 */ 344 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 345 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 346 goto freeit; 347 348 /* 349 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 350 * don't do it. 351 */ 352 nxt = -1; 353 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 354 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 355 struct icmp6_hdr *icp; 356 357 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 358 sizeof(*icp)); 359 if (icp == NULL) { 360 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 361 return; 362 } 363 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 364 icp->icmp6_type == ND_REDIRECT) { 365 /* 366 * ICMPv6 error 367 * Special case: for redirect (which is 368 * informational) we must not send icmp6 error. 369 */ 370 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 371 goto freeit; 372 } else { 373 /* ICMPv6 informational - send the error */ 374 } 375 } 376#if 0 /* controversial */ 377 else if (off >= 0 && nxt == IPPROTO_ESP) { 378 /* 379 * It could be ICMPv6 error inside ESP. Take a safer side, 380 * don't respond. 381 */ 382 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 383 goto freeit; 384 } 385#endif 386 else { 387 /* non-ICMPv6 - send the error */ 388 } 389 390 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 391 392 /* Finally, do rate limitation check. */ 393 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 394 ICMP6_STATINC(ICMP6_STAT_TOOFREQ); 395 goto freeit; 396 } 397 398 /* 399 * OK, ICMP6 can be generated. 400 */ 401 402 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 403 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 404 405 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 406 M_PREPEND(m, preplen, M_DONTWAIT); 407 if (m && M_UNWRITABLE(m, preplen)) 408 m = m_pullup(m, preplen); 409 if (m == NULL) { 410 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 411 return; 412 } 413 414 nip6 = mtod(m, struct ip6_hdr *); 415 nip6->ip6_src = oip6->ip6_src; 416 nip6->ip6_dst = oip6->ip6_dst; 417 418 in6_clearscope(&oip6->ip6_src); 419 in6_clearscope(&oip6->ip6_dst); 420 421 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 422 icmp6->icmp6_type = type; 423 icmp6->icmp6_code = code; 424 icmp6->icmp6_pptr = htonl((u_int32_t)param); 425 426 /* 427 * icmp6_reflect() is designed to be in the input path. 428 * icmp6_error() can be called from both input and output path, 429 * and if we are in output path rcvif could contain bogus value. 430 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 431 * information in ip header (nip6). 432 */ 433 m->m_pkthdr.rcvif = NULL; 434 435 ICMP6_STATINC(ICMP6_STAT_OUTHIST + type); 436 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 437 438 return; 439 440 freeit: 441 /* 442 * If we can't tell whether or not we can generate ICMP6, free it. 443 */ 444 m_freem(m); 445} 446 447/* 448 * Process a received ICMP6 message. 449 */ 450int 451icmp6_input(struct mbuf **mp, int *offp, int proto) 452{ 453 struct mbuf *m = *mp, *n; 454 struct ip6_hdr *ip6, *nip6; 455 struct icmp6_hdr *icmp6, *nicmp6; 456 int off = *offp; 457 int icmp6len = m->m_pkthdr.len - *offp; 458 int code, sum, noff; 459 460#define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4) 461 KASSERT(ICMP6_MAXLEN < MCLBYTES); 462 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 463 464 /* 465 * Locate icmp6 structure in mbuf, and check 466 * that not corrupted and of at least minimum length 467 */ 468 469 ip6 = mtod(m, struct ip6_hdr *); 470 if (icmp6len < sizeof(struct icmp6_hdr)) { 471 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 472 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 473 goto freeit; 474 } 475 476 /* 477 * calculate the checksum 478 */ 479 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 480 if (icmp6 == NULL) { 481 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 482 /* m is invalid */ 483 /*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/ 484 return IPPROTO_DONE; 485 } 486 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 487 code = icmp6->icmp6_code; 488 489 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 490 nd6log((LOG_ERR, 491 "ICMP6 checksum error(%d|%x) %s\n", 492 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 493 ICMP6_STATINC(ICMP6_STAT_CHECKSUM); 494 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 495 goto freeit; 496 } 497 498#if defined(NFAITH) && 0 < NFAITH 499 if (faithprefix(&ip6->ip6_dst)) { 500 /* 501 * Deliver very specific ICMP6 type only. 502 * This is important to deliver TOOBIG. Otherwise PMTUD 503 * will not work. 504 */ 505 switch (icmp6->icmp6_type) { 506 case ICMP6_DST_UNREACH: 507 case ICMP6_PACKET_TOO_BIG: 508 case ICMP6_TIME_EXCEEDED: 509 break; 510 default: 511 goto freeit; 512 } 513 } 514#endif 515 516 ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type); 517 518 switch (icmp6->icmp6_type) { 519 case ICMP6_DST_UNREACH: 520 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 521 switch (code) { 522 case ICMP6_DST_UNREACH_NOROUTE: 523 code = PRC_UNREACH_NET; 524 break; 525 case ICMP6_DST_UNREACH_ADMIN: 526 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 527 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 528 break; 529 case ICMP6_DST_UNREACH_ADDR: 530 code = PRC_HOSTDEAD; 531 break; 532#ifdef COMPAT_RFC1885 533 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 534 code = PRC_UNREACH_SRCFAIL; 535 break; 536#else 537 case ICMP6_DST_UNREACH_BEYONDSCOPE: 538 /* I mean "source address was incorrect." */ 539 code = PRC_UNREACH_NET; 540 break; 541#endif 542 case ICMP6_DST_UNREACH_NOPORT: 543 code = PRC_UNREACH_PORT; 544 break; 545 default: 546 goto badcode; 547 } 548 goto deliver; 549 550 case ICMP6_PACKET_TOO_BIG: 551 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 552 553 code = PRC_MSGSIZE; 554 555 /* 556 * Updating the path MTU will be done after examining 557 * intermediate extension headers. 558 */ 559 goto deliver; 560 561 case ICMP6_TIME_EXCEEDED: 562 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 563 switch (code) { 564 case ICMP6_TIME_EXCEED_TRANSIT: 565 code = PRC_TIMXCEED_INTRANS; 566 break; 567 case ICMP6_TIME_EXCEED_REASSEMBLY: 568 code = PRC_TIMXCEED_REASS; 569 break; 570 default: 571 goto badcode; 572 } 573 goto deliver; 574 575 case ICMP6_PARAM_PROB: 576 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 577 switch (code) { 578 case ICMP6_PARAMPROB_NEXTHEADER: 579 code = PRC_UNREACH_PROTOCOL; 580 break; 581 case ICMP6_PARAMPROB_HEADER: 582 case ICMP6_PARAMPROB_OPTION: 583 code = PRC_PARAMPROB; 584 break; 585 default: 586 goto badcode; 587 } 588 goto deliver; 589 590 case ICMP6_ECHO_REQUEST: 591 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 592 if (code != 0) 593 goto badcode; 594 /* 595 * Copy mbuf to send to two data paths: userland socket(s), 596 * and to the querier (echo reply). 597 * m: a copy for socket, n: a copy for querier 598 * 599 * If the first mbuf is shared, or the first mbuf is too short, 600 * copy the first part of the data into a fresh mbuf. 601 * Otherwise, we will wrongly overwrite both copies. 602 */ 603 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 604 /* Give up local */ 605 n = m; 606 m = NULL; 607 } else if (M_READONLY(n) || 608 n->m_len < off + sizeof(struct icmp6_hdr)) { 609 struct mbuf *n0 = n; 610 611 /* 612 * Prepare an internal mbuf. m_pullup() doesn't 613 * always copy the length we specified. 614 */ 615 if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 616 /* Give up local */ 617 n = m; 618 m = NULL; 619 } 620 m_freem(n0); 621 } 622 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, 623 sizeof(*nicmp6)); 624 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 625 nicmp6->icmp6_code = 0; 626 if (n) { 627 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 628 icmp6s[ICMP6_STAT_REFLECT]++; 629 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++; 630 ICMP6_STAT_PUTREF(); 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 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 739 icmp6s[ICMP6_STAT_REFLECT]++; 740 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++; 741 ICMP6_STAT_PUTREF(); 742 icmp6_reflect(n, noff); 743 } 744 break; 745 } 746 747 case ICMP6_WRUREPLY: 748 if (code != 0) 749 goto badcode; 750 break; 751 752 case ND_ROUTER_SOLICIT: 753 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 754 if (code != 0) 755 goto badcode; 756 if (icmp6len < sizeof(struct nd_router_solicit)) 757 goto badlen; 758 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 759 /* give up local */ 760 nd6_rs_input(m, off, icmp6len); 761 m = NULL; 762 goto freeit; 763 } 764 nd6_rs_input(n, off, icmp6len); 765 /* m stays. */ 766 break; 767 768 case ND_ROUTER_ADVERT: 769 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 770 if (code != 0) 771 goto badcode; 772 if (icmp6len < sizeof(struct nd_router_advert)) 773 goto badlen; 774 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 775 /* give up local */ 776 nd6_ra_input(m, off, icmp6len); 777 m = NULL; 778 goto freeit; 779 } 780 nd6_ra_input(n, off, icmp6len); 781 /* m stays. */ 782 break; 783 784 case ND_NEIGHBOR_SOLICIT: 785 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 786 if (code != 0) 787 goto badcode; 788 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 789 goto badlen; 790 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 791 /* give up local */ 792 nd6_ns_input(m, off, icmp6len); 793 m = NULL; 794 goto freeit; 795 } 796 nd6_ns_input(n, off, icmp6len); 797 /* m stays. */ 798 break; 799 800 case ND_NEIGHBOR_ADVERT: 801 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 802 if (code != 0) 803 goto badcode; 804 if (icmp6len < sizeof(struct nd_neighbor_advert)) 805 goto badlen; 806 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 807 /* give up local */ 808 nd6_na_input(m, off, icmp6len); 809 m = NULL; 810 goto freeit; 811 } 812 nd6_na_input(n, off, icmp6len); 813 /* m stays. */ 814 break; 815 816 case ND_REDIRECT: 817 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 818 if (code != 0) 819 goto badcode; 820 if (icmp6len < sizeof(struct nd_redirect)) 821 goto badlen; 822 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 823 /* give up local */ 824 icmp6_redirect_input(m, off); 825 m = NULL; 826 goto freeit; 827 } 828 icmp6_redirect_input(n, off); 829 /* m stays. */ 830 break; 831 832 case ICMP6_ROUTER_RENUMBERING: 833 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 834 code != ICMP6_ROUTER_RENUMBERING_RESULT) 835 goto badcode; 836 if (icmp6len < sizeof(struct icmp6_router_renum)) 837 goto badlen; 838 break; 839 840 default: 841 nd6log((LOG_DEBUG, 842 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 843 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 844 ip6_sprintf(&ip6->ip6_dst), 845 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 846 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 847 /* ICMPv6 error: MUST deliver it by spec... */ 848 code = PRC_NCMDS; 849 /* deliver */ 850 } else { 851 /* ICMPv6 informational: MUST not deliver */ 852 break; 853 } 854 deliver: 855 if (icmp6_notify_error(m, off, icmp6len, code)) { 856 /* In this case, m should've been freed. */ 857 return (IPPROTO_DONE); 858 } 859 break; 860 861 badcode: 862 ICMP6_STATINC(ICMP6_STAT_BADCODE); 863 break; 864 865 badlen: 866 ICMP6_STATINC(ICMP6_STAT_BADLEN); 867 break; 868 } 869 870 /* deliver the packet to appropriate sockets */ 871 icmp6_rip6_input(&m, *offp); 872 873 return IPPROTO_DONE; 874 875 freeit: 876 m_freem(m); 877 return IPPROTO_DONE; 878} 879 880static int 881icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int 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 ICMP6_STATINC(ICMP6_STAT_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 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 896 return (-1); 897 } 898 eip6 = (struct ip6_hdr *)(icmp6 + 1); 899 900 /* Detect the upper level protocol */ 901 { 902 void (*ctlfunc)(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 ICMP6_STATINC(ICMP6_STAT_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 ICMP6_STATINC(ICMP6_STAT_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 ICMP6_STATINC(ICMP6_STAT_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 ICMP6_STATINC(ICMP6_STAT_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 ICMP6_STATINC(ICMP6_STAT_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 sockaddr_in6_init(&icmp6dst, 1027 (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0); 1028 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1029 goto freeit; 1030 sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0); 1031 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1032 goto freeit; 1033 icmp6src.sin6_flowinfo = 1034 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1035 1036 if (finaldst == NULL) 1037 finaldst = &eip6->ip6_dst; 1038 ip6cp.ip6c_m = m; 1039 ip6cp.ip6c_icmp6 = icmp6; 1040 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1041 ip6cp.ip6c_off = eoff; 1042 ip6cp.ip6c_finaldst = finaldst; 1043 ip6cp.ip6c_src = &icmp6src; 1044 ip6cp.ip6c_nxt = nxt; 1045 1046 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1047 notifymtu = ntohl(icmp6->icmp6_mtu); 1048 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1049 } 1050 1051 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1052 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1053 if (ctlfunc) { 1054 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1055 &ip6cp); 1056 } 1057 } 1058 return (0); 1059 1060 freeit: 1061 m_freem(m); 1062 return (-1); 1063} 1064 1065void 1066icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1067{ 1068 unsigned long rtcount; 1069 struct icmp6_mtudisc_callback *mc; 1070 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1071 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1072 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1073 u_int mtu = ntohl(icmp6->icmp6_mtu); 1074 struct rtentry *rt = NULL; 1075 struct sockaddr_in6 sin6; 1076 1077 /* 1078 * allow non-validated cases if memory is plenty, to make traffic 1079 * from non-connected pcb happy. 1080 */ 1081 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1082 if (validated) { 1083 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1084 return; 1085 else if (0 <= icmp6_mtudisc_lowat && 1086 rtcount > icmp6_mtudisc_lowat) { 1087 /* 1088 * XXX nuke a victim, install the new one. 1089 */ 1090 } 1091 } else { 1092 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1093 return; 1094 } 1095 1096 bzero(&sin6, sizeof(sin6)); 1097 sin6.sin6_family = PF_INET6; 1098 sin6.sin6_len = sizeof(struct sockaddr_in6); 1099 sin6.sin6_addr = *dst; 1100 if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1101 return; 1102 1103 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1104 1105 if (rt && (rt->rt_flags & RTF_HOST) && 1106 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1107 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1108 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1109 ICMP6_STATINC(ICMP6_STAT_PMTUCHG); 1110 rt->rt_rmx.rmx_mtu = mtu; 1111 } 1112 } 1113 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */ 1114 RTFREE(rt); 1115 } 1116 1117 /* 1118 * Notify protocols that the MTU for this destination 1119 * has changed. 1120 */ 1121 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1122 mc = LIST_NEXT(mc, mc_list)) 1123 (*mc->mc_func)(&sin6.sin6_addr); 1124} 1125 1126/* 1127 * Process a Node Information Query packet, based on 1128 * draft-ietf-ipngwg-icmp-name-lookups-07. 1129 * 1130 * Spec incompatibilities: 1131 * - IPv6 Subject address handling 1132 * - IPv4 Subject address handling support missing 1133 * - Proxy reply (answer even if it's not for me) 1134 * - joins NI group address at in6_ifattach() time only, does not cope 1135 * with hostname changes by sethostname(3) 1136 */ 1137#ifndef offsetof /* XXX */ 1138#define offsetof(type, member) ((size_t)(&((type *)0)->member)) 1139#endif 1140static struct mbuf * 1141ni6_input(struct mbuf *m, int off) 1142{ 1143 struct icmp6_nodeinfo *ni6, *nni6; 1144 struct mbuf *n = NULL; 1145 u_int16_t qtype; 1146 int subjlen; 1147 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1148 struct ni_reply_fqdn *fqdn; 1149 int addrs; /* for NI_QTYPE_NODEADDR */ 1150 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1151 struct sockaddr_in6 sin6; /* ip6_dst */ 1152 struct in6_addr in6_subj; /* subject address */ 1153 struct ip6_hdr *ip6; 1154 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1155 char *subj = NULL; 1156 1157 ip6 = mtod(m, struct ip6_hdr *); 1158 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1159 if (ni6 == NULL) { 1160 /* m is already reclaimed */ 1161 return NULL; 1162 } 1163 1164 /* 1165 * Validate IPv6 destination address. 1166 * 1167 * The Responder must discard the Query without further processing 1168 * unless it is one of the Responder's unicast or anycast addresses, or 1169 * a link-local scope multicast address which the Responder has joined. 1170 * [icmp-name-lookups-07, Section 4.] 1171 */ 1172 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 1173 /* XXX scopeid */ 1174 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1175 ; /* unicast/anycast, fine */ 1176 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1177 ; /* link-local multicast, fine */ 1178 else 1179 goto bad; 1180 1181 /* validate query Subject field. */ 1182 qtype = ntohs(ni6->ni_qtype); 1183 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1184 switch (qtype) { 1185 case NI_QTYPE_NOOP: 1186 case NI_QTYPE_SUPTYPES: 1187 /* 07 draft */ 1188 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1189 break; 1190 /* FALLTHROUGH */ 1191 case NI_QTYPE_FQDN: 1192 case NI_QTYPE_NODEADDR: 1193 case NI_QTYPE_IPV4ADDR: 1194 switch (ni6->ni_code) { 1195 case ICMP6_NI_SUBJ_IPV6: 1196#if ICMP6_NI_SUBJ_IPV6 != 0 1197 case 0: 1198#endif 1199 /* 1200 * backward compatibility - try to accept 03 draft 1201 * format, where no Subject is present. 1202 */ 1203 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1204 subjlen == 0) { 1205 oldfqdn++; 1206 break; 1207 } 1208#if ICMP6_NI_SUBJ_IPV6 != 0 1209 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1210 goto bad; 1211#endif 1212 1213 if (subjlen != sizeof(sin6.sin6_addr)) 1214 goto bad; 1215 1216 /* 1217 * Validate Subject address. 1218 * 1219 * Not sure what exactly "address belongs to the node" 1220 * means in the spec, is it just unicast, or what? 1221 * 1222 * At this moment we consider Subject address as 1223 * "belong to the node" if the Subject address equals 1224 * to the IPv6 destination address; validation for 1225 * IPv6 destination address should have done enough 1226 * check for us. 1227 * 1228 * We do not do proxy at this moment. 1229 */ 1230 /* m_pulldown instead of copy? */ 1231 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1232 subjlen, (void *)&in6_subj); 1233 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1234 goto bad; 1235 1236 subj = (char *)&in6_subj; 1237 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1238 break; 1239 1240 /* 1241 * XXX if we are to allow other cases, we should really 1242 * be careful about scope here. 1243 * basically, we should disallow queries toward IPv6 1244 * destination X with subject Y, if scope(X) > scope(Y). 1245 * if we allow scope(X) > scope(Y), it will result in 1246 * information leakage across scope boundary. 1247 */ 1248 goto bad; 1249 1250 case ICMP6_NI_SUBJ_FQDN: 1251 /* 1252 * Validate Subject name with gethostname(3). 1253 * 1254 * The behavior may need some debate, since: 1255 * - we are not sure if the node has FQDN as 1256 * hostname (returned by gethostname(3)). 1257 * - the code does wildcard match for truncated names. 1258 * however, we are not sure if we want to perform 1259 * wildcard match, if gethostname(3) side has 1260 * truncated hostname. 1261 */ 1262 n = ni6_nametodns(hostname, hostnamelen, 0); 1263 if (!n || n->m_next || n->m_len == 0) 1264 goto bad; 1265 IP6_EXTHDR_GET(subj, char *, m, 1266 off + sizeof(struct icmp6_nodeinfo), subjlen); 1267 if (subj == NULL) 1268 goto bad; 1269 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1270 n->m_len)) { 1271 goto bad; 1272 } 1273 m_freem(n); 1274 n = NULL; 1275 break; 1276 1277 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1278 default: 1279 goto bad; 1280 } 1281 break; 1282 } 1283 1284 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1285 switch (qtype) { 1286 case NI_QTYPE_FQDN: 1287 if ((icmp6_nodeinfo & 1) == 0) 1288 goto bad; 1289 break; 1290 case NI_QTYPE_NODEADDR: 1291 case NI_QTYPE_IPV4ADDR: 1292 if ((icmp6_nodeinfo & 2) == 0) 1293 goto bad; 1294 break; 1295 } 1296 1297 /* guess reply length */ 1298 switch (qtype) { 1299 case NI_QTYPE_NOOP: 1300 break; /* no reply data */ 1301 case NI_QTYPE_SUPTYPES: 1302 replylen += sizeof(u_int32_t); 1303 break; 1304 case NI_QTYPE_FQDN: 1305 /* XXX will append an mbuf */ 1306 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1307 break; 1308 case NI_QTYPE_NODEADDR: 1309 addrs = ni6_addrs(ni6, m, &ifp, subj); 1310 if ((replylen += addrs * (sizeof(struct in6_addr) + 1311 sizeof(u_int32_t))) > MCLBYTES) 1312 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1313 break; 1314 case NI_QTYPE_IPV4ADDR: 1315 /* unsupported - should respond with unknown Qtype? */ 1316 goto bad; 1317 default: 1318 /* 1319 * XXX: We must return a reply with the ICMP6 code 1320 * `unknown Qtype' in this case. However we regard the case 1321 * as an FQDN query for backward compatibility. 1322 * Older versions set a random value to this field, 1323 * so it rarely varies in the defined qtypes. 1324 * But the mechanism is not reliable... 1325 * maybe we should obsolete older versions. 1326 */ 1327 qtype = NI_QTYPE_FQDN; 1328 /* XXX will append an mbuf */ 1329 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1330 oldfqdn++; 1331 break; 1332 } 1333 1334 /* allocate an mbuf to reply. */ 1335 MGETHDR(n, M_DONTWAIT, m->m_type); 1336 if (n == NULL) { 1337 m_freem(m); 1338 return (NULL); 1339 } 1340 M_MOVE_PKTHDR(n, m); /* just for rcvif */ 1341 if (replylen > MHLEN) { 1342 if (replylen > MCLBYTES) { 1343 /* 1344 * XXX: should we try to allocate more? But MCLBYTES 1345 * is probably much larger than IPV6_MMTU... 1346 */ 1347 goto bad; 1348 } 1349 MCLGET(n, M_DONTWAIT); 1350 if ((n->m_flags & M_EXT) == 0) { 1351 goto bad; 1352 } 1353 } 1354 n->m_pkthdr.len = n->m_len = replylen; 1355 1356 /* copy mbuf header and IPv6 + Node Information base headers */ 1357 bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr)); 1358 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1359 bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo)); 1360 1361 /* qtype dependent procedure */ 1362 switch (qtype) { 1363 case NI_QTYPE_NOOP: 1364 nni6->ni_code = ICMP6_NI_SUCCESS; 1365 nni6->ni_flags = 0; 1366 break; 1367 case NI_QTYPE_SUPTYPES: 1368 { 1369 u_int32_t v; 1370 nni6->ni_code = ICMP6_NI_SUCCESS; 1371 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1372 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1373 v = (u_int32_t)htonl(0x0000000f); 1374 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1375 break; 1376 } 1377 case NI_QTYPE_FQDN: 1378 nni6->ni_code = ICMP6_NI_SUCCESS; 1379 fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) + 1380 sizeof(struct ip6_hdr) + 1381 sizeof(struct icmp6_nodeinfo)); 1382 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1383 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1384 /* 1385 * XXX do we really have FQDN in variable "hostname"? 1386 */ 1387 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1388 if (n->m_next == NULL) 1389 goto bad; 1390 /* XXX we assume that n->m_next is not a chain */ 1391 if (n->m_next->m_next != NULL) 1392 goto bad; 1393 n->m_pkthdr.len += n->m_next->m_len; 1394 break; 1395 case NI_QTYPE_NODEADDR: 1396 { 1397 int lenlim, copied; 1398 1399 nni6->ni_code = ICMP6_NI_SUCCESS; 1400 n->m_pkthdr.len = n->m_len = 1401 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1402 lenlim = M_TRAILINGSPACE(n); 1403 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1404 /* XXX: reset mbuf length */ 1405 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1406 sizeof(struct icmp6_nodeinfo) + copied; 1407 break; 1408 } 1409 default: 1410 break; /* XXX impossible! */ 1411 } 1412 1413 nni6->ni_type = ICMP6_NI_REPLY; 1414 m_freem(m); 1415 return (n); 1416 1417 bad: 1418 m_freem(m); 1419 if (n) 1420 m_freem(n); 1421 return (NULL); 1422} 1423#undef hostnamelen 1424 1425#define isupper(x) ('A' <= (x) && (x) <= 'Z') 1426#define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1427#define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1428#define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1429 1430/* 1431 * make a mbuf with DNS-encoded string. no compression support. 1432 * 1433 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1434 * treated as truncated name (two \0 at the end). this is a wild guess. 1435 * 1436 * old - return pascal string if non-zero 1437 */ 1438static struct mbuf * 1439ni6_nametodns(const char *name, int namelen, int old) 1440{ 1441 struct mbuf *m; 1442 char *cp, *ep; 1443 const char *p, *q; 1444 int i, len, nterm; 1445 1446 if (old) 1447 len = namelen + 1; 1448 else 1449 len = MCLBYTES; 1450 1451 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1452 MGET(m, M_DONTWAIT, MT_DATA); 1453 if (m && len > MLEN) { 1454 MCLGET(m, M_DONTWAIT); 1455 if ((m->m_flags & M_EXT) == 0) 1456 goto fail; 1457 } 1458 if (!m) 1459 goto fail; 1460 m->m_next = NULL; 1461 1462 if (old) { 1463 m->m_len = len; 1464 *mtod(m, char *) = namelen; 1465 bcopy(name, mtod(m, char *) + 1, namelen); 1466 return m; 1467 } else { 1468 m->m_len = 0; 1469 cp = mtod(m, char *); 1470 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1471 1472 /* if not certain about my name, return empty buffer */ 1473 if (namelen == 0) 1474 return m; 1475 1476 /* 1477 * guess if it looks like shortened hostname, or FQDN. 1478 * shortened hostname needs two trailing "\0". 1479 */ 1480 i = 0; 1481 for (p = name; p < name + namelen; p++) { 1482 if (*p && *p == '.') 1483 i++; 1484 } 1485 if (i < 2) 1486 nterm = 2; 1487 else 1488 nterm = 1; 1489 1490 p = name; 1491 while (cp < ep && p < name + namelen) { 1492 i = 0; 1493 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1494 i++; 1495 /* result does not fit into mbuf */ 1496 if (cp + i + 1 >= ep) 1497 goto fail; 1498 /* 1499 * DNS label length restriction, RFC1035 page 8. 1500 * "i == 0" case is included here to avoid returning 1501 * 0-length label on "foo..bar". 1502 */ 1503 if (i <= 0 || i >= 64) 1504 goto fail; 1505 *cp++ = i; 1506 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1507 goto fail; 1508 while (i > 0) { 1509 if (!isalnum(*p) && *p != '-') 1510 goto fail; 1511 if (isupper(*p)) { 1512 *cp++ = tolower(*p); 1513 p++; 1514 } else 1515 *cp++ = *p++; 1516 i--; 1517 } 1518 p = q; 1519 if (p < name + namelen && *p == '.') 1520 p++; 1521 } 1522 /* termination */ 1523 if (cp + nterm >= ep) 1524 goto fail; 1525 while (nterm-- > 0) 1526 *cp++ = '\0'; 1527 m->m_len = cp - mtod(m, char *); 1528 return m; 1529 } 1530 1531 panic("should not reach here"); 1532 /* NOTREACHED */ 1533 1534 fail: 1535 if (m) 1536 m_freem(m); 1537 return NULL; 1538} 1539 1540/* 1541 * check if two DNS-encoded string matches. takes care of truncated 1542 * form (with \0\0 at the end). no compression support. 1543 * XXX upper/lowercase match (see RFC2065) 1544 */ 1545static int 1546ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1547{ 1548 const char *a0, *b0; 1549 int l; 1550 1551 /* simplest case - need validation? */ 1552 if (alen == blen && bcmp(a, b, alen) == 0) 1553 return 1; 1554 1555 a0 = a; 1556 b0 = b; 1557 1558 /* termination is mandatory */ 1559 if (alen < 2 || blen < 2) 1560 return 0; 1561 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1562 return 0; 1563 alen--; 1564 blen--; 1565 1566 while (a - a0 < alen && b - b0 < blen) { 1567 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1568 return 0; 1569 1570 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1571 return 0; 1572 /* we don't support compression yet */ 1573 if (a[0] >= 64 || b[0] >= 64) 1574 return 0; 1575 1576 /* truncated case */ 1577 if (a[0] == 0 && a - a0 == alen - 1) 1578 return 1; 1579 if (b[0] == 0 && b - b0 == blen - 1) 1580 return 1; 1581 if (a[0] == 0 || b[0] == 0) 1582 return 0; 1583 1584 if (a[0] != b[0]) 1585 return 0; 1586 l = a[0]; 1587 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1588 return 0; 1589 if (bcmp(a + 1, b + 1, l) != 0) 1590 return 0; 1591 1592 a += 1 + l; 1593 b += 1 + l; 1594 } 1595 1596 if (a - a0 == alen && b - b0 == blen) 1597 return 1; 1598 else 1599 return 0; 1600} 1601 1602/* 1603 * calculate the number of addresses to be returned in the node info reply. 1604 */ 1605static int 1606ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, 1607 struct ifnet **ifpp, char *subj) 1608{ 1609 struct ifnet *ifp; 1610 struct in6_ifaddr *ifa6; 1611 struct ifaddr *ifa; 1612 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1613 int addrs = 0, addrsofif, iffound = 0; 1614 int niflags = ni6->ni_flags; 1615 1616 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1617 switch (ni6->ni_code) { 1618 case ICMP6_NI_SUBJ_IPV6: 1619 if (subj == NULL) /* must be impossible... */ 1620 return (0); 1621 subj_ip6 = (struct sockaddr_in6 *)subj; 1622 break; 1623 default: 1624 /* 1625 * XXX: we only support IPv6 subject address for 1626 * this Qtype. 1627 */ 1628 return (0); 1629 } 1630 } 1631 1632 IFNET_FOREACH(ifp) { 1633 addrsofif = 0; 1634 IFADDR_FOREACH(ifa, ifp) { 1635 if (ifa->ifa_addr->sa_family != AF_INET6) 1636 continue; 1637 ifa6 = (struct in6_ifaddr *)ifa; 1638 1639 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1640 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1641 &ifa6->ia_addr.sin6_addr)) 1642 iffound = 1; 1643 1644 /* 1645 * IPv4-mapped addresses can only be returned by a 1646 * Node Information proxy, since they represent 1647 * addresses of IPv4-only nodes, which perforce do 1648 * not implement this protocol. 1649 * [icmp-name-lookups-07, Section 5.4] 1650 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1651 * this function at this moment. 1652 */ 1653 1654 /* What do we have to do about ::1? */ 1655 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1656 case IPV6_ADDR_SCOPE_LINKLOCAL: 1657 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1658 continue; 1659 break; 1660 case IPV6_ADDR_SCOPE_SITELOCAL: 1661 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1662 continue; 1663 break; 1664 case IPV6_ADDR_SCOPE_GLOBAL: 1665 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1666 continue; 1667 break; 1668 default: 1669 continue; 1670 } 1671 1672 /* 1673 * check if anycast is okay. 1674 * XXX: just experimental. not in the spec. 1675 */ 1676 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1677 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1678 continue; /* we need only unicast addresses */ 1679 1680 addrsofif++; /* count the address */ 1681 } 1682 if (iffound) { 1683 *ifpp = ifp; 1684 return (addrsofif); 1685 } 1686 1687 addrs += addrsofif; 1688 } 1689 1690 return (addrs); 1691} 1692 1693static int 1694ni6_store_addrs(struct icmp6_nodeinfo *ni6, 1695 struct icmp6_nodeinfo *nni6, struct ifnet *ifp0, 1696 int resid) 1697{ 1698 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1699 struct in6_ifaddr *ifa6; 1700 struct ifaddr *ifa; 1701 struct ifnet *ifp_dep = NULL; 1702 int copied = 0, allow_deprecated = 0; 1703 u_char *cp = (u_char *)(nni6 + 1); 1704 int niflags = ni6->ni_flags; 1705 u_int32_t ltime; 1706 1707 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1708 return (0); /* needless to copy */ 1709 1710 again: 1711 1712 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1713 { 1714 IFADDR_FOREACH(ifa, ifp) { 1715 if (ifa->ifa_addr->sa_family != AF_INET6) 1716 continue; 1717 ifa6 = (struct in6_ifaddr *)ifa; 1718 1719 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1720 allow_deprecated == 0) { 1721 /* 1722 * prefererred address should be put before 1723 * deprecated addresses. 1724 */ 1725 1726 /* record the interface for later search */ 1727 if (ifp_dep == NULL) 1728 ifp_dep = ifp; 1729 1730 continue; 1731 } 1732 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1733 allow_deprecated != 0) 1734 continue; /* we now collect deprecated addrs */ 1735 1736 /* What do we have to do about ::1? */ 1737 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1738 case IPV6_ADDR_SCOPE_LINKLOCAL: 1739 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1740 continue; 1741 break; 1742 case IPV6_ADDR_SCOPE_SITELOCAL: 1743 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1744 continue; 1745 break; 1746 case IPV6_ADDR_SCOPE_GLOBAL: 1747 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1748 continue; 1749 break; 1750 default: 1751 continue; 1752 } 1753 1754 /* 1755 * check if anycast is okay. 1756 * XXX: just experimental. not in the spec. 1757 */ 1758 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1759 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1760 continue; 1761 1762 /* now we can copy the address */ 1763 if (resid < sizeof(struct in6_addr) + 1764 sizeof(u_int32_t)) { 1765 /* 1766 * We give up much more copy. 1767 * Set the truncate flag and return. 1768 */ 1769 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1770 return (copied); 1771 } 1772 1773 /* 1774 * Set the TTL of the address. 1775 * The TTL value should be one of the following 1776 * according to the specification: 1777 * 1778 * 1. The remaining lifetime of a DHCP lease on the 1779 * address, or 1780 * 2. The remaining Valid Lifetime of a prefix from 1781 * which the address was derived through Stateless 1782 * Autoconfiguration. 1783 * 1784 * Note that we currently do not support stateful 1785 * address configuration by DHCPv6, so the former 1786 * case can't happen. 1787 * 1788 * TTL must be 2^31 > TTL >= 0. 1789 */ 1790 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1791 ltime = ND6_INFINITE_LIFETIME; 1792 else { 1793 if (ifa6->ia6_lifetime.ia6t_expire > 1794 time_second) 1795 ltime = ifa6->ia6_lifetime.ia6t_expire - 1796 time_second; 1797 else 1798 ltime = 0; 1799 } 1800 if (ltime > 0x7fffffff) 1801 ltime = 0x7fffffff; 1802 ltime = htonl(ltime); 1803 1804 bcopy(<ime, cp, sizeof(u_int32_t)); 1805 cp += sizeof(u_int32_t); 1806 1807 /* copy the address itself */ 1808 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1809 sizeof(struct in6_addr)); 1810 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1811 cp += sizeof(struct in6_addr); 1812 1813 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1814 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1815 } 1816 if (ifp0) /* we need search only on the specified IF */ 1817 break; 1818 } 1819 1820 if (allow_deprecated == 0 && ifp_dep != NULL) { 1821 ifp = ifp_dep; 1822 allow_deprecated = 1; 1823 1824 goto again; 1825 } 1826 1827 return (copied); 1828} 1829 1830/* 1831 * XXX almost dup'ed code with rip6_input. 1832 */ 1833static int 1834icmp6_rip6_input(struct mbuf **mp, int off) 1835{ 1836 struct mbuf *m = *mp; 1837 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1838 struct inpcb_hdr *inph; 1839 struct in6pcb *in6p; 1840 struct in6pcb *last = NULL; 1841 struct sockaddr_in6 rip6src; 1842 struct icmp6_hdr *icmp6; 1843 struct mbuf *opts = NULL; 1844 1845 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1846 if (icmp6 == NULL) { 1847 /* m is already reclaimed */ 1848 return IPPROTO_DONE; 1849 } 1850 1851 /* 1852 * XXX: the address may have embedded scope zone ID, which should be 1853 * hidden from applications. 1854 */ 1855 sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0); 1856 if (sa6_recoverscope(&rip6src)) { 1857 m_freem(m); 1858 return (IPPROTO_DONE); 1859 } 1860 1861 CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) { 1862 in6p = (struct in6pcb *)inph; 1863 if (in6p->in6p_af != AF_INET6) 1864 continue; 1865 if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6) 1866 continue; 1867 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1868 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1869 continue; 1870 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1871 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1872 continue; 1873 if (in6p->in6p_icmp6filt 1874 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1875 in6p->in6p_icmp6filt)) 1876 continue; 1877 if (last) { 1878 struct mbuf *n; 1879 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1880 if (last->in6p_flags & IN6P_CONTROLOPTS) 1881 ip6_savecontrol(last, &opts, ip6, n); 1882 /* strip intermediate headers */ 1883 m_adj(n, off); 1884 if (sbappendaddr(&last->in6p_socket->so_rcv, 1885 (struct sockaddr *)&rip6src, 1886 n, opts) == 0) { 1887 /* should notify about lost packet */ 1888 m_freem(n); 1889 if (opts) 1890 m_freem(opts); 1891 } else 1892 sorwakeup(last->in6p_socket); 1893 opts = NULL; 1894 } 1895 } 1896 last = in6p; 1897 } 1898 if (last) { 1899 if (last->in6p_flags & IN6P_CONTROLOPTS) 1900 ip6_savecontrol(last, &opts, ip6, m); 1901 /* strip intermediate headers */ 1902 m_adj(m, off); 1903 if (sbappendaddr(&last->in6p_socket->so_rcv, 1904 (struct sockaddr *)&rip6src, m, opts) == 0) { 1905 m_freem(m); 1906 if (opts) 1907 m_freem(opts); 1908 } else 1909 sorwakeup(last->in6p_socket); 1910 } else { 1911 m_freem(m); 1912 IP6_STATDEC(IP6_STAT_DELIVERED); 1913 } 1914 return IPPROTO_DONE; 1915} 1916 1917/* 1918 * Reflect the ip6 packet back to the source. 1919 * OFF points to the icmp6 header, counted from the top of the mbuf. 1920 * 1921 * Note: RFC 1885 required that an echo reply should be truncated if it 1922 * did not fit in with (return) path MTU, and KAME code supported the 1923 * behavior. However, as a clarification after the RFC, this limitation 1924 * was removed in a revised version of the spec, RFC 2463. We had kept the 1925 * old behavior, with a (non-default) ifdef block, while the new version of 1926 * the spec was an internet-draft status, and even after the new RFC was 1927 * published. But it would rather make sense to clean the obsoleted part 1928 * up, and to make the code simpler at this stage. 1929 */ 1930void 1931icmp6_reflect(struct mbuf *m, size_t off) 1932{ 1933 struct ip6_hdr *ip6; 1934 struct icmp6_hdr *icmp6; 1935 const struct in6_ifaddr *ia; 1936 const struct ip6aux *ip6a; 1937 int plen; 1938 int type, code; 1939 struct ifnet *outif = NULL; 1940 struct in6_addr origdst; 1941 const struct in6_addr *src = NULL; 1942 1943 /* too short to reflect */ 1944 if (off < sizeof(struct ip6_hdr)) { 1945 nd6log((LOG_DEBUG, 1946 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 1947 (u_long)off, (u_long)sizeof(struct ip6_hdr), 1948 __FILE__, __LINE__)); 1949 goto bad; 1950 } 1951 1952 /* 1953 * If there are extra headers between IPv6 and ICMPv6, strip 1954 * off that header first. 1955 */ 1956#ifdef DIAGNOSTIC 1957 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 1958 panic("assumption failed in icmp6_reflect"); 1959#endif 1960 if (off > sizeof(struct ip6_hdr)) { 1961 size_t l; 1962 struct ip6_hdr nip6; 1963 1964 l = off - sizeof(struct ip6_hdr); 1965 m_copydata(m, 0, sizeof(nip6), (void *)&nip6); 1966 m_adj(m, l); 1967 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1968 if (m->m_len < l) { 1969 if ((m = m_pullup(m, l)) == NULL) 1970 return; 1971 } 1972 bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6)); 1973 } else /* off == sizeof(struct ip6_hdr) */ { 1974 size_t l; 1975 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1976 if (m->m_len < l) { 1977 if ((m = m_pullup(m, l)) == NULL) 1978 return; 1979 } 1980 } 1981 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 1982 ip6 = mtod(m, struct ip6_hdr *); 1983 ip6->ip6_nxt = IPPROTO_ICMPV6; 1984 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 1985 type = icmp6->icmp6_type; /* keep type for statistics */ 1986 code = icmp6->icmp6_code; /* ditto. */ 1987 1988 origdst = ip6->ip6_dst; 1989 /* 1990 * ip6_input() drops a packet if its src is multicast. 1991 * So, the src is never multicast. 1992 */ 1993 ip6->ip6_dst = ip6->ip6_src; 1994 1995 /* 1996 * If the incoming packet was addressed directly to us (i.e. unicast), 1997 * use dst as the src for the reply. 1998 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 1999 * (for example) when we encounter an error while forwarding procedure 2000 * destined to a duplicated address of ours. 2001 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2002 * procedure of an outgoing packet of our own, in which case we need 2003 * to search in the ifaddr list. 2004 */ 2005 if (IN6_IS_ADDR_MULTICAST(&origdst)) 2006 ; 2007 else if ((ip6a = ip6_getdstifaddr(m)) != NULL) { 2008 if ((ip6a->ip6a_flags & 2009 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2010 src = &ip6a->ip6a_src; 2011 } else { 2012 union { 2013 struct sockaddr_in6 sin6; 2014 struct sockaddr sa; 2015 } u; 2016 2017 sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0); 2018 2019 ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa); 2020 2021 if (ia == NULL) 2022 ; 2023 else if ((ia->ia6_flags & 2024 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2025 src = &ia->ia_addr.sin6_addr; 2026 } 2027 2028 if (src == NULL) { 2029 int e; 2030 struct sockaddr_in6 sin6; 2031 struct route ro; 2032 2033 /* 2034 * This case matches to multicasts, our anycast, or unicasts 2035 * that we do not own. Select a source address based on the 2036 * source address of the erroneous packet. 2037 */ 2038 /* zone ID should be embedded */ 2039 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 2040 2041 memset(&ro, 0, sizeof(ro)); 2042 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2043 rtcache_free(&ro); 2044 if (src == NULL) { 2045 nd6log((LOG_DEBUG, 2046 "icmp6_reflect: source can't be determined: " 2047 "dst=%s, error=%d\n", 2048 ip6_sprintf(&sin6.sin6_addr), e)); 2049 goto bad; 2050 } 2051 } 2052 2053 ip6->ip6_src = *src; 2054 ip6->ip6_flow = 0; 2055 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2056 ip6->ip6_vfc |= IPV6_VERSION; 2057 ip6->ip6_nxt = IPPROTO_ICMPV6; 2058 if (m->m_pkthdr.rcvif) { 2059 /* XXX: This may not be the outgoing interface */ 2060 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2061 } else 2062 ip6->ip6_hlim = ip6_defhlim; 2063 2064 m->m_pkthdr.csum_flags = 0; 2065 icmp6->icmp6_cksum = 0; 2066 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2067 sizeof(struct ip6_hdr), plen); 2068 2069 /* 2070 * XXX option handling 2071 */ 2072 2073 m->m_flags &= ~(M_BCAST|M_MCAST); 2074 2075 /* 2076 * To avoid a "too big" situation at an intermediate router 2077 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2078 * Note that only echo and node information replies are affected, 2079 * since the length of ICMP6 errors is limited to the minimum MTU. 2080 */ 2081 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 && 2082 outif) 2083 icmp6_ifstat_inc(outif, ifs6_out_error); 2084 2085 if (outif) 2086 icmp6_ifoutstat_inc(outif, type, code); 2087 2088 return; 2089 2090 bad: 2091 m_freem(m); 2092 return; 2093} 2094 2095static const char * 2096icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2097 struct in6_addr *tgt6) 2098{ 2099 static char buf[1024]; 2100 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2101 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2102 return buf; 2103} 2104 2105void 2106icmp6_redirect_input(struct mbuf *m, int off) 2107{ 2108 struct ifnet *ifp = m->m_pkthdr.rcvif; 2109 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2110 struct nd_redirect *nd_rd; 2111 int icmp6len = ntohs(ip6->ip6_plen); 2112 char *lladdr = NULL; 2113 int lladdrlen = 0; 2114 struct rtentry *rt = NULL; 2115 int is_router; 2116 int is_onlink; 2117 struct in6_addr src6 = ip6->ip6_src; 2118 struct in6_addr redtgt6; 2119 struct in6_addr reddst6; 2120 union nd_opts ndopts; 2121 2122 if (!ifp) 2123 return; 2124 2125 /* XXX if we are router, we don't update route by icmp6 redirect */ 2126 if (ip6_forwarding) 2127 goto freeit; 2128 if (!icmp6_rediraccept) 2129 goto freeit; 2130 2131 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2132 if (nd_rd == NULL) { 2133 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 2134 return; 2135 } 2136 redtgt6 = nd_rd->nd_rd_target; 2137 reddst6 = nd_rd->nd_rd_dst; 2138 2139 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2140 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2141 goto freeit; 2142 } 2143 2144 /* validation */ 2145 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2146 nd6log((LOG_ERR, 2147 "ICMP6 redirect sent from %s rejected; " 2148 "must be from linklocal\n", ip6_sprintf(&src6))); 2149 goto bad; 2150 } 2151 if (ip6->ip6_hlim != 255) { 2152 nd6log((LOG_ERR, 2153 "ICMP6 redirect sent from %s rejected; " 2154 "hlim=%d (must be 255)\n", 2155 ip6_sprintf(&src6), ip6->ip6_hlim)); 2156 goto bad; 2157 } 2158 { 2159 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2160 struct sockaddr_in6 sin6; 2161 struct in6_addr *gw6; 2162 2163 sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0); 2164 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2165 if (rt) { 2166 if (rt->rt_gateway == NULL || 2167 rt->rt_gateway->sa_family != AF_INET6) { 2168 nd6log((LOG_ERR, 2169 "ICMP6 redirect rejected; no route " 2170 "with inet6 gateway found for redirect dst: %s\n", 2171 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2172 RTFREE(rt); 2173 goto bad; 2174 } 2175 2176 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2177 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2178 nd6log((LOG_ERR, 2179 "ICMP6 redirect rejected; " 2180 "not equal to gw-for-src=%s (must be same): " 2181 "%s\n", 2182 ip6_sprintf(gw6), 2183 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2184 RTFREE(rt); 2185 goto bad; 2186 } 2187 } else { 2188 nd6log((LOG_ERR, 2189 "ICMP6 redirect rejected; " 2190 "no route found for redirect dst: %s\n", 2191 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2192 goto bad; 2193 } 2194 RTFREE(rt); 2195 rt = NULL; 2196 } 2197 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2198 nd6log((LOG_ERR, 2199 "ICMP6 redirect rejected; " 2200 "redirect dst must be unicast: %s\n", 2201 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2202 goto bad; 2203 } 2204 2205 is_router = is_onlink = 0; 2206 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2207 is_router = 1; /* router case */ 2208 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2209 is_onlink = 1; /* on-link destination case */ 2210 if (!is_router && !is_onlink) { 2211 nd6log((LOG_ERR, 2212 "ICMP6 redirect rejected; " 2213 "neither router case nor onlink case: %s\n", 2214 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2215 goto bad; 2216 } 2217 /* validation passed */ 2218 2219 icmp6len -= sizeof(*nd_rd); 2220 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2221 if (nd6_options(&ndopts) < 0) { 2222 nd6log((LOG_INFO, "icmp6_redirect_input: " 2223 "invalid ND option, rejected: %s\n", 2224 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2225 /* nd6_options have incremented stats */ 2226 goto freeit; 2227 } 2228 2229 if (ndopts.nd_opts_tgt_lladdr) { 2230 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2231 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2232 } 2233 2234 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2235 nd6log((LOG_INFO, 2236 "icmp6_redirect_input: lladdrlen mismatch for %s " 2237 "(if %d, icmp6 packet %d): %s\n", 2238 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2239 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2240 goto bad; 2241 } 2242 2243 /* RFC 2461 8.3 */ 2244 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2245 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2246 2247 if (!is_onlink) { /* better router case. perform rtredirect. */ 2248 /* perform rtredirect */ 2249 struct sockaddr_in6 sdst; 2250 struct sockaddr_in6 sgw; 2251 struct sockaddr_in6 ssrc; 2252 unsigned long rtcount; 2253 struct rtentry *newrt = NULL; 2254 2255 /* 2256 * do not install redirect route, if the number of entries 2257 * is too much (> hiwat). note that, the node (= host) will 2258 * work just fine even if we do not install redirect route 2259 * (there will be additional hops, though). 2260 */ 2261 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2262 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2263 return; 2264 else if (0 <= icmp6_redirect_lowat && 2265 rtcount > icmp6_redirect_lowat) { 2266 /* 2267 * XXX nuke a victim, install the new one. 2268 */ 2269 } 2270 2271 bzero(&sdst, sizeof(sdst)); 2272 bzero(&sgw, sizeof(sgw)); 2273 bzero(&ssrc, sizeof(ssrc)); 2274 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2275 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2276 sizeof(struct sockaddr_in6); 2277 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2278 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2279 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2280 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2281 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2282 (struct sockaddr *)&ssrc, 2283 &newrt); 2284 2285 if (newrt) { 2286 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2287 icmp6_redirect_timeout_q); 2288 rtfree(newrt); 2289 } 2290 } 2291 /* finally update cached route in each socket via pfctlinput */ 2292 { 2293 struct sockaddr_in6 sdst; 2294 2295 sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0); 2296 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2297#if defined(IPSEC) || defined(FAST_IPSEC) 2298 key_sa_routechange((struct sockaddr *)&sdst); 2299#endif 2300 } 2301 2302 freeit: 2303 m_freem(m); 2304 return; 2305 2306 bad: 2307 ICMP6_STATINC(ICMP6_STAT_BADREDIRECT); 2308 m_freem(m); 2309} 2310 2311void 2312icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2313{ 2314 struct ifnet *ifp; /* my outgoing interface */ 2315 struct in6_addr *ifp_ll6; 2316 struct in6_addr *nexthop; 2317 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2318 struct mbuf *m = NULL; /* newly allocated one */ 2319 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2320 struct nd_redirect *nd_rd; 2321 size_t maxlen; 2322 u_char *p; 2323 struct sockaddr_in6 src_sa; 2324 2325 icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0); 2326 2327 /* if we are not router, we don't send icmp6 redirect */ 2328 if (!ip6_forwarding) 2329 goto fail; 2330 2331 /* sanity check */ 2332 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2333 goto fail; 2334 2335 /* 2336 * Address check: 2337 * the source address must identify a neighbor, and 2338 * the destination address must not be a multicast address 2339 * [RFC 2461, sec 8.2] 2340 */ 2341 sip6 = mtod(m0, struct ip6_hdr *); 2342 sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0); 2343 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2344 goto fail; 2345 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2346 goto fail; /* what should we do here? */ 2347 2348 /* rate limit */ 2349 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2350 goto fail; 2351 2352 /* 2353 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2354 * we almost always ask for an mbuf cluster for simplicity. 2355 * (MHLEN < IPV6_MMTU is almost always true) 2356 */ 2357#if IPV6_MMTU >= MCLBYTES 2358# error assumption failed about IPV6_MMTU and MCLBYTES 2359#endif 2360 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2361 if (m && IPV6_MMTU >= MHLEN) 2362 MCLGET(m, M_DONTWAIT); 2363 if (!m) 2364 goto fail; 2365 m->m_pkthdr.rcvif = NULL; 2366 m->m_len = 0; 2367 maxlen = M_TRAILINGSPACE(m); 2368 maxlen = min(IPV6_MMTU, maxlen); 2369 /* just for safety */ 2370 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2371 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2372 goto fail; 2373 } 2374 2375 { 2376 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2377 struct in6_ifaddr *ia; 2378 if ((ia = in6ifa_ifpforlinklocal(ifp, 2379 IN6_IFF_NOTREADY| 2380 IN6_IFF_ANYCAST)) == NULL) 2381 goto fail; 2382 ifp_ll6 = &ia->ia_addr.sin6_addr; 2383 } 2384 2385 /* get ip6 linklocal address for the router. */ 2386 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2387 struct sockaddr_in6 *sin6; 2388 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2389 nexthop = &sin6->sin6_addr; 2390 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2391 nexthop = NULL; 2392 } else 2393 nexthop = NULL; 2394 2395 /* ip6 */ 2396 ip6 = mtod(m, struct ip6_hdr *); 2397 ip6->ip6_flow = 0; 2398 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2399 ip6->ip6_vfc |= IPV6_VERSION; 2400 /* ip6->ip6_plen will be set later */ 2401 ip6->ip6_nxt = IPPROTO_ICMPV6; 2402 ip6->ip6_hlim = 255; 2403 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2404 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2405 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2406 2407 /* ND Redirect */ 2408 nd_rd = (struct nd_redirect *)(ip6 + 1); 2409 nd_rd->nd_rd_type = ND_REDIRECT; 2410 nd_rd->nd_rd_code = 0; 2411 nd_rd->nd_rd_reserved = 0; 2412 if (rt->rt_flags & RTF_GATEWAY) { 2413 /* 2414 * nd_rd->nd_rd_target must be a link-local address in 2415 * better router cases. 2416 */ 2417 if (!nexthop) 2418 goto fail; 2419 bcopy(nexthop, &nd_rd->nd_rd_target, 2420 sizeof(nd_rd->nd_rd_target)); 2421 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2422 sizeof(nd_rd->nd_rd_dst)); 2423 } else { 2424 /* make sure redtgt == reddst */ 2425 nexthop = &sip6->ip6_dst; 2426 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2427 sizeof(nd_rd->nd_rd_target)); 2428 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2429 sizeof(nd_rd->nd_rd_dst)); 2430 } 2431 2432 p = (u_char *)(nd_rd + 1); 2433 2434 { 2435 /* target lladdr option */ 2436 struct rtentry *rt_nexthop = NULL; 2437 int len; 2438 const struct sockaddr_dl *sdl; 2439 struct nd_opt_hdr *nd_opt; 2440 char *lladdr; 2441 2442 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2443 if (!rt_nexthop) 2444 goto nolladdropt; 2445 len = sizeof(*nd_opt) + ifp->if_addrlen; 2446 len = (len + 7) & ~7; /* round by 8 */ 2447 /* safety check */ 2448 if (len + (p - (u_char *)ip6) > maxlen) 2449 goto nolladdropt; 2450 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2451 (rt_nexthop->rt_flags & RTF_LLINFO) && 2452 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2453 (sdl = satocsdl(rt_nexthop->rt_gateway)) && 2454 sdl->sdl_alen) { 2455 nd_opt = (struct nd_opt_hdr *)p; 2456 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2457 nd_opt->nd_opt_len = len >> 3; 2458 lladdr = (char *)(nd_opt + 1); 2459 memcpy(lladdr, CLLADDR(sdl), ifp->if_addrlen); 2460 p += len; 2461 } 2462 } 2463 nolladdropt:; 2464 2465 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2466 2467 /* just to be safe */ 2468 if (m0->m_flags & M_DECRYPTED) 2469 goto noredhdropt; 2470 if (p - (u_char *)ip6 > maxlen) 2471 goto noredhdropt; 2472 2473 { 2474 /* redirected header option */ 2475 int len; 2476 struct nd_opt_rd_hdr *nd_opt_rh; 2477 2478 /* 2479 * compute the maximum size for icmp6 redirect header option. 2480 * XXX room for auth header? 2481 */ 2482 len = maxlen - (p - (u_char *)ip6); 2483 len &= ~7; 2484 2485 /* 2486 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2487 * about padding/truncate rule for the original IP packet. 2488 * From the discussion on IPv6imp in Feb 1999, 2489 * the consensus was: 2490 * - "attach as much as possible" is the goal 2491 * - pad if not aligned (original size can be guessed by 2492 * original ip6 header) 2493 * Following code adds the padding if it is simple enough, 2494 * and truncates if not. 2495 */ 2496 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2497 /* not enough room, truncate */ 2498 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2499 m0->m_pkthdr.len); 2500 } else { 2501 /* 2502 * enough room, truncate if not aligned. 2503 * we don't pad here for simplicity. 2504 */ 2505 size_t extra; 2506 2507 extra = m0->m_pkthdr.len % 8; 2508 if (extra) { 2509 /* truncate */ 2510 m_adj(m0, -extra); 2511 } 2512 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2513 } 2514 2515 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2516 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2517 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2518 nd_opt_rh->nd_opt_rh_len = len >> 3; 2519 p += sizeof(*nd_opt_rh); 2520 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2521 2522 /* connect m0 to m */ 2523 m->m_pkthdr.len += m0->m_pkthdr.len; 2524 m_cat(m, m0); 2525 m0 = NULL; 2526 } 2527noredhdropt: 2528 if (m0) { 2529 m_freem(m0); 2530 m0 = NULL; 2531 } 2532 2533 /* XXX: clear embedded link IDs in the inner header */ 2534 in6_clearscope(&sip6->ip6_src); 2535 in6_clearscope(&sip6->ip6_dst); 2536 in6_clearscope(&nd_rd->nd_rd_target); 2537 in6_clearscope(&nd_rd->nd_rd_dst); 2538 2539 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2540 2541 nd_rd->nd_rd_cksum = 0; 2542 nd_rd->nd_rd_cksum 2543 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2544 2545 /* send the packet to outside... */ 2546 if (ip6_output(m, NULL, NULL, 0, 2547 (struct ip6_moptions *)NULL, (struct socket *)NULL, NULL) != 0) 2548 icmp6_ifstat_inc(ifp, ifs6_out_error); 2549 2550 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2551 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2552 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT); 2553 2554 return; 2555 2556fail: 2557 if (m) 2558 m_freem(m); 2559 if (m0) 2560 m_freem(m0); 2561} 2562 2563/* 2564 * ICMPv6 socket option processing. 2565 */ 2566int 2567icmp6_ctloutput(int op, struct socket *so, int level, 2568 int optname, struct mbuf **mp) 2569{ 2570 int error = 0; 2571 int optlen; 2572 struct in6pcb *in6p = sotoin6pcb(so); 2573 struct mbuf *m = *mp; 2574 2575 optlen = m ? m->m_len : 0; 2576 2577 if (level != IPPROTO_ICMPV6) 2578 return rip6_ctloutput(op, so, level, optname, mp); 2579 2580 switch (op) { 2581 case PRCO_SETOPT: 2582 switch (optname) { 2583 case ICMP6_FILTER: 2584 { 2585 struct icmp6_filter *p; 2586 2587 if (optlen != sizeof(*p)) { 2588 error = EMSGSIZE; 2589 break; 2590 } 2591 p = mtod(m, struct icmp6_filter *); 2592 if (!p || !in6p->in6p_icmp6filt) { 2593 error = EINVAL; 2594 break; 2595 } 2596 bcopy(p, in6p->in6p_icmp6filt, 2597 sizeof(struct icmp6_filter)); 2598 error = 0; 2599 break; 2600 } 2601 2602 default: 2603 error = ENOPROTOOPT; 2604 break; 2605 } 2606 if (m) 2607 (void)m_freem(m); 2608 break; 2609 2610 case PRCO_GETOPT: 2611 switch (optname) { 2612 case ICMP6_FILTER: 2613 { 2614 struct icmp6_filter *p; 2615 2616 if (!in6p->in6p_icmp6filt) { 2617 error = EINVAL; 2618 break; 2619 } 2620 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2621 m->m_len = sizeof(struct icmp6_filter); 2622 p = mtod(m, struct icmp6_filter *); 2623 bcopy(in6p->in6p_icmp6filt, p, 2624 sizeof(struct icmp6_filter)); 2625 error = 0; 2626 break; 2627 } 2628 2629 default: 2630 error = ENOPROTOOPT; 2631 break; 2632 } 2633 break; 2634 } 2635 2636 return (error); 2637} 2638 2639/* 2640 * Perform rate limit check. 2641 * Returns 0 if it is okay to send the icmp6 packet. 2642 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2643 * limitation. 2644 * 2645 * XXX per-destination/type check necessary? 2646 */ 2647static int 2648icmp6_ratelimit( 2649 const struct in6_addr *dst, /* not used at this moment */ 2650 const int type, /* not used at this moment */ 2651 const int code) /* not used at this moment */ 2652{ 2653 int ret; 2654 2655 ret = 0; /* okay to send */ 2656 2657 /* PPS limit */ 2658 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2659 icmp6errppslim)) { 2660 /* The packet is subject to rate limit */ 2661 ret++; 2662 } 2663 2664 return ret; 2665} 2666 2667static struct rtentry * 2668icmp6_mtudisc_clone(struct sockaddr *dst) 2669{ 2670 struct rtentry *rt; 2671 int error; 2672 2673 rt = rtalloc1(dst, 1); 2674 if (rt == 0) 2675 return NULL; 2676 2677 /* If we didn't get a host route, allocate one */ 2678 if ((rt->rt_flags & RTF_HOST) == 0) { 2679 struct rtentry *nrt; 2680 2681 error = rtrequest((int) RTM_ADD, dst, 2682 (struct sockaddr *) rt->rt_gateway, 2683 (struct sockaddr *) 0, 2684 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2685 if (error) { 2686 rtfree(rt); 2687 return NULL; 2688 } 2689 nrt->rt_rmx = rt->rt_rmx; 2690 rtfree(rt); 2691 rt = nrt; 2692 } 2693 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2694 icmp6_mtudisc_timeout_q); 2695 if (error) { 2696 rtfree(rt); 2697 return NULL; 2698 } 2699 2700 return rt; /* caller need to call rtfree() */ 2701} 2702 2703static void 2704icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 2705{ 2706 if (rt == NULL) 2707 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2708 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2709 (RTF_DYNAMIC | RTF_HOST)) { 2710 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2711 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2712 } else { 2713 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2714 rt->rt_rmx.rmx_mtu = 0; 2715 } 2716} 2717 2718static void 2719icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r) 2720{ 2721 if (rt == NULL) 2722 panic("icmp6_redirect_timeout: bad route to timeout"); 2723 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2724 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2725 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2726 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2727 } 2728} 2729 2730/* 2731 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes. silly? 2732 */ 2733static int 2734sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS) 2735{ 2736 (void)&name; 2737 (void)&l; 2738 (void)&oname; 2739 2740 if (namelen != 0) 2741 return (EINVAL); 2742 2743 return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp, 2744 /*XXXUNCONST*/ 2745 __UNCONST(newp), newlen)); 2746} 2747 2748static void 2749icmp6stat_convert_to_user_cb(void *v1, void *v2, struct cpu_info *ci) 2750{ 2751 uint64_t *icmp6sc = v1; 2752 uint64_t *icmp6s = v2; 2753 u_int i; 2754 2755 for (i = 0; i < ICMP6_NSTATS; i++) 2756 icmp6s[i] += icmp6sc[i]; 2757} 2758 2759static void 2760icmp6stat_convert_to_user(uint64_t *icmp6s) 2761{ 2762 2763 memset(icmp6s, 0, sizeof(uint64_t) * ICMP6_NSTATS); 2764 percpu_foreach(icmp6stat_percpu, icmp6stat_convert_to_user_cb, icmp6s); 2765} 2766 2767static int 2768sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS) 2769{ 2770 struct sysctlnode node; 2771 uint64_t icmp6s[ICMP6_NSTATS]; 2772 2773 icmp6stat_convert_to_user(icmp6s); 2774 node = *rnode; 2775 node.sysctl_data = icmp6s; 2776 node.sysctl_size = sizeof(icmp6s); 2777 return (sysctl_lookup(SYSCTLFN_CALL(&node))); 2778} 2779 2780SYSCTL_SETUP(sysctl_net_inet6_icmp6_setup, 2781 "sysctl net.inet6.icmp6 subtree setup") 2782{ 2783 extern int nd6_maxqueuelen; /* defined in nd6.c */ 2784 2785 sysctl_createv(clog, 0, NULL, NULL, 2786 CTLFLAG_PERMANENT, 2787 CTLTYPE_NODE, "net", NULL, 2788 NULL, 0, NULL, 0, 2789 CTL_NET, CTL_EOL); 2790 sysctl_createv(clog, 0, NULL, NULL, 2791 CTLFLAG_PERMANENT, 2792 CTLTYPE_NODE, "inet6", NULL, 2793 NULL, 0, NULL, 0, 2794 CTL_NET, PF_INET6, CTL_EOL); 2795 sysctl_createv(clog, 0, NULL, NULL, 2796 CTLFLAG_PERMANENT, 2797 CTLTYPE_NODE, "icmp6", 2798 SYSCTL_DESCR("ICMPv6 related settings"), 2799 NULL, 0, NULL, 0, 2800 CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL); 2801 2802 sysctl_createv(clog, 0, NULL, NULL, 2803 CTLFLAG_PERMANENT, 2804 CTLTYPE_STRUCT, "stats", 2805 SYSCTL_DESCR("ICMPv6 transmission statistics"), 2806 sysctl_net_inet6_icmp6_stats, 0, NULL, 0, 2807 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2808 ICMPV6CTL_STATS, CTL_EOL); 2809 sysctl_createv(clog, 0, NULL, NULL, 2810 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2811 CTLTYPE_INT, "rediraccept", 2812 SYSCTL_DESCR("Accept and process redirect messages"), 2813 NULL, 0, &icmp6_rediraccept, 0, 2814 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2815 ICMPV6CTL_REDIRACCEPT, CTL_EOL); 2816 sysctl_createv(clog, 0, NULL, NULL, 2817 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2818 CTLTYPE_INT, "redirtimeout", 2819 SYSCTL_DESCR("Redirect generated route lifetime"), 2820 NULL, 0, &icmp6_redirtimeout, 0, 2821 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2822 ICMPV6CTL_REDIRTIMEOUT, CTL_EOL); 2823#if 0 /* obsoleted */ 2824 sysctl_createv(clog, 0, NULL, NULL, 2825 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2826 CTLTYPE_INT, "errratelimit", NULL, 2827 NULL, 0, &icmp6_errratelimit, 0, 2828 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2829 ICMPV6CTL_ERRRATELIMIT, CTL_EOL); 2830#endif 2831 sysctl_createv(clog, 0, NULL, NULL, 2832 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2833 CTLTYPE_INT, "nd6_prune", 2834 SYSCTL_DESCR("Neighbor discovery prune interval"), 2835 NULL, 0, &nd6_prune, 0, 2836 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2837 ICMPV6CTL_ND6_PRUNE, CTL_EOL); 2838 sysctl_createv(clog, 0, NULL, NULL, 2839 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2840 CTLTYPE_INT, "nd6_delay", 2841 SYSCTL_DESCR("First probe delay time"), 2842 NULL, 0, &nd6_delay, 0, 2843 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2844 ICMPV6CTL_ND6_DELAY, CTL_EOL); 2845 sysctl_createv(clog, 0, NULL, NULL, 2846 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2847 CTLTYPE_INT, "nd6_umaxtries", 2848 SYSCTL_DESCR("Number of unicast discovery attempts"), 2849 NULL, 0, &nd6_umaxtries, 0, 2850 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2851 ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL); 2852 sysctl_createv(clog, 0, NULL, NULL, 2853 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2854 CTLTYPE_INT, "nd6_mmaxtries", 2855 SYSCTL_DESCR("Number of multicast discovery attempts"), 2856 NULL, 0, &nd6_mmaxtries, 0, 2857 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2858 ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL); 2859 sysctl_createv(clog, 0, NULL, NULL, 2860 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2861 CTLTYPE_INT, "nd6_useloopback", 2862 SYSCTL_DESCR("Use loopback interface for local traffic"), 2863 NULL, 0, &nd6_useloopback, 0, 2864 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2865 ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL); 2866#if 0 /* obsoleted */ 2867 sysctl_createv(clog, 0, NULL, NULL, 2868 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2869 CTLTYPE_INT, "nd6_proxyall", NULL, 2870 NULL, 0, &nd6_proxyall, 0, 2871 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2872 ICMPV6CTL_ND6_PROXYALL, CTL_EOL); 2873#endif 2874 sysctl_createv(clog, 0, NULL, NULL, 2875 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2876 CTLTYPE_INT, "nodeinfo", 2877 SYSCTL_DESCR("Respond to node information requests"), 2878 NULL, 0, &icmp6_nodeinfo, 0, 2879 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2880 ICMPV6CTL_NODEINFO, CTL_EOL); 2881 sysctl_createv(clog, 0, NULL, NULL, 2882 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2883 CTLTYPE_INT, "errppslimit", 2884 SYSCTL_DESCR("Maximum ICMP errors sent per second"), 2885 NULL, 0, &icmp6errppslim, 0, 2886 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2887 ICMPV6CTL_ERRPPSLIMIT, CTL_EOL); 2888 sysctl_createv(clog, 0, NULL, NULL, 2889 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2890 CTLTYPE_INT, "nd6_maxnudhint", 2891 SYSCTL_DESCR("Maximum neighbor unreachable hint count"), 2892 NULL, 0, &nd6_maxnudhint, 0, 2893 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2894 ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL); 2895 sysctl_createv(clog, 0, NULL, NULL, 2896 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2897 CTLTYPE_INT, "mtudisc_hiwat", 2898 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2899 NULL, 0, &icmp6_mtudisc_hiwat, 0, 2900 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2901 ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL); 2902 sysctl_createv(clog, 0, NULL, NULL, 2903 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2904 CTLTYPE_INT, "mtudisc_lowat", 2905 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2906 NULL, 0, &icmp6_mtudisc_lowat, 0, 2907 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2908 ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL); 2909 sysctl_createv(clog, 0, NULL, NULL, 2910 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2911 CTLTYPE_INT, "nd6_debug", 2912 SYSCTL_DESCR("Enable neighbor discovery debug output"), 2913 NULL, 0, &nd6_debug, 0, 2914 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2915 ICMPV6CTL_ND6_DEBUG, CTL_EOL); 2916 sysctl_createv(clog, 0, NULL, NULL, 2917 CTLFLAG_PERMANENT, 2918 CTLTYPE_STRUCT, "nd6_drlist", 2919 SYSCTL_DESCR("Default router list"), 2920 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2921 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2922 ICMPV6CTL_ND6_DRLIST, CTL_EOL); 2923 sysctl_createv(clog, 0, NULL, NULL, 2924 CTLFLAG_PERMANENT, 2925 CTLTYPE_STRUCT, "nd6_prlist", 2926 SYSCTL_DESCR("Prefix list"), 2927 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2928 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2929 ICMPV6CTL_ND6_PRLIST, CTL_EOL); 2930 sysctl_createv(clog, 0, NULL, NULL, 2931 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2932 CTLTYPE_INT, "maxqueuelen", 2933 SYSCTL_DESCR("max packet queue len for a unresolved ND"), 2934 NULL, 1, &nd6_maxqueuelen, 0, 2935 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2936 ICMPV6CTL_ND6_MAXQLEN, CTL_EOL); 2937} 2938 2939void 2940icmp6_statinc(u_int stat) 2941{ 2942 2943 KASSERT(stat < ICMP6_NSTATS); 2944 ICMP6_STATINC(stat); 2945} 2946