icmp6.c revision 1.180
1/* $NetBSD: icmp6.c,v 1.180 2016/04/01 08:12:00 ozaki-r 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 1.180 2016/04/01 08:12:00 ozaki-r Exp $"); 66 67#ifdef _KERNEL_OPT 68#include "opt_inet.h" 69#include "opt_ipsec.h" 70#endif 71 72#include <sys/param.h> 73#include <sys/systm.h> 74#include <sys/kmem.h> 75#include <sys/mbuf.h> 76#include <sys/protosw.h> 77#include <sys/socket.h> 78#include <sys/socketvar.h> 79#include <sys/time.h> 80#include <sys/kernel.h> 81#include <sys/syslog.h> 82#include <sys/domain.h> 83#include <sys/sysctl.h> 84 85#include <net/if.h> 86#include <net/route.h> 87#include <net/if_dl.h> 88#include <net/if_types.h> 89 90#include <netinet/in.h> 91#include <netinet/in_var.h> 92#include <netinet/ip6.h> 93#include <netinet6/ip6_var.h> 94#include <netinet6/ip6_private.h> 95#include <netinet/icmp6.h> 96#include <netinet6/icmp6_private.h> 97#include <netinet6/mld6_var.h> 98#include <netinet6/in6_pcb.h> 99#include <netinet6/nd6.h> 100#include <netinet6/in6_ifattach.h> 101#include <netinet6/ip6protosw.h> 102#include <netinet6/scope6_var.h> 103 104#ifdef IPSEC 105#include <netipsec/ipsec.h> 106#include <netipsec/key.h> 107#endif 108 109 110#include "faith.h" 111#if defined(NFAITH) && 0 < NFAITH 112#include <net/if_faith.h> 113#endif 114 115#include <net/net_osdep.h> 116 117extern struct domain inet6domain; 118 119percpu_t *icmp6stat_percpu; 120 121extern struct inpcbtable raw6cbtable; 122extern int icmp6errppslim; 123static int icmp6errpps_count = 0; 124static struct timeval icmp6errppslim_last; 125extern int icmp6_nodeinfo; 126 127/* 128 * List of callbacks to notify when Path MTU changes are made. 129 */ 130struct icmp6_mtudisc_callback { 131 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 132 void (*mc_func)(struct in6_addr *); 133}; 134 135LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 136 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 137 138static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 139extern int pmtu_expire; 140 141/* XXX do these values make any sense? */ 142static int icmp6_mtudisc_hiwat = 1280; 143static int icmp6_mtudisc_lowat = 256; 144 145/* 146 * keep track of # of redirect routes. 147 */ 148static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 149 150/* XXX experimental, turned off */ 151static int icmp6_redirect_hiwat = -1; 152static int icmp6_redirect_lowat = -1; 153 154static void icmp6_errcount(u_int, int, int); 155static int icmp6_rip6_input(struct mbuf **, int); 156static int icmp6_ratelimit(const struct in6_addr *, const int, const int); 157static const char *icmp6_redirect_diag(struct in6_addr *, 158 struct in6_addr *, struct in6_addr *); 159static struct mbuf *ni6_input(struct mbuf *, int); 160static struct mbuf *ni6_nametodns(const char *, int, int); 161static int ni6_dnsmatch(const char *, int, const char *, int); 162static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *, 163 struct ifnet **, char *); 164static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 165 struct ifnet *, int); 166static int icmp6_notify_error(struct mbuf *, int, int, int); 167static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *); 168static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *); 169static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *); 170static void sysctl_net_inet6_icmp6_setup(struct sysctllog **); 171 172 173void 174icmp6_init(void) 175{ 176 177 sysctl_net_inet6_icmp6_setup(NULL); 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 = kmem_alloc(sizeof(*mc), KM_SLEEP); 259 mc->mc_func = func; 260 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 261} 262 263/* 264 * A wrapper function for icmp6_error() necessary when the erroneous packet 265 * may not contain enough scope zone information. 266 */ 267void 268icmp6_error2(struct mbuf *m, int type, int code, int param, 269 struct ifnet *ifp) 270{ 271 struct ip6_hdr *ip6; 272 273 if (ifp == NULL) 274 return; 275 276 if (m->m_len < sizeof(struct ip6_hdr)) { 277 m = m_pullup(m, sizeof(struct ip6_hdr)); 278 if (m == NULL) 279 return; 280 } 281 282 ip6 = mtod(m, struct ip6_hdr *); 283 284 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 285 return; 286 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 287 return; 288 289 icmp6_error(m, type, code, param); 290} 291 292/* 293 * Generate an error packet of type error in response to bad IP6 packet. 294 */ 295void 296icmp6_error(struct mbuf *m, int type, int code, int param) 297{ 298 struct ip6_hdr *oip6, *nip6; 299 struct icmp6_hdr *icmp6; 300 u_int preplen; 301 int off; 302 int nxt; 303 304 ICMP6_STATINC(ICMP6_STAT_ERROR); 305 306 /* count per-type-code statistics */ 307 icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code); 308 309 if (m->m_flags & M_DECRYPTED) { 310 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 311 goto freeit; 312 } 313 314 if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) && 315 (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) 316 return; 317 oip6 = mtod(m, struct ip6_hdr *); 318 319 /* 320 * If the destination address of the erroneous packet is a multicast 321 * address, or the packet was sent using link-layer multicast, 322 * we should basically suppress sending an error (RFC 2463, Section 323 * 2.4). 324 * We have two exceptions (the item e.2 in that section): 325 * - the Pakcet Too Big message can be sent for path MTU discovery. 326 * - the Parameter Problem Message that can be allowed an icmp6 error 327 * in the option type field. This check has been done in 328 * ip6_unknown_opt(), so we can just check the type and code. 329 */ 330 if ((m->m_flags & (M_BCAST|M_MCAST) || 331 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 332 (type != ICMP6_PACKET_TOO_BIG && 333 (type != ICMP6_PARAM_PROB || 334 code != ICMP6_PARAMPROB_OPTION))) 335 goto freeit; 336 337 /* 338 * RFC 2463, 2.4 (e.5): source address check. 339 * XXX: the case of anycast source? 340 */ 341 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 342 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 343 goto freeit; 344 345 /* 346 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 347 * don't do it. 348 */ 349 nxt = -1; 350 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 351 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 352 struct icmp6_hdr *icp; 353 354 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 355 sizeof(*icp)); 356 if (icp == NULL) { 357 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 358 return; 359 } 360 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 361 icp->icmp6_type == ND_REDIRECT) { 362 /* 363 * ICMPv6 error 364 * Special case: for redirect (which is 365 * informational) we must not send icmp6 error. 366 */ 367 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 368 goto freeit; 369 } else { 370 /* ICMPv6 informational - send the error */ 371 } 372 } 373#if 0 /* controversial */ 374 else if (off >= 0 && nxt == IPPROTO_ESP) { 375 /* 376 * It could be ICMPv6 error inside ESP. Take a safer side, 377 * don't respond. 378 */ 379 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 380 goto freeit; 381 } 382#endif 383 else { 384 /* non-ICMPv6 - send the error */ 385 } 386 387 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 388 389 /* Finally, do rate limitation check. */ 390 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 391 ICMP6_STATINC(ICMP6_STAT_TOOFREQ); 392 goto freeit; 393 } 394 395 /* 396 * OK, ICMP6 can be generated. 397 */ 398 399 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 400 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 401 402 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 403 M_PREPEND(m, preplen, M_DONTWAIT); 404 if (m && M_UNWRITABLE(m, preplen)) 405 m = m_pullup(m, preplen); 406 if (m == NULL) { 407 nd6log(LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__); 408 return; 409 } 410 411 nip6 = mtod(m, struct ip6_hdr *); 412 nip6->ip6_src = oip6->ip6_src; 413 nip6->ip6_dst = oip6->ip6_dst; 414 415 in6_clearscope(&oip6->ip6_src); 416 in6_clearscope(&oip6->ip6_dst); 417 418 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 419 icmp6->icmp6_type = type; 420 icmp6->icmp6_code = code; 421 icmp6->icmp6_pptr = htonl((u_int32_t)param); 422 423 /* 424 * icmp6_reflect() is designed to be in the input path. 425 * icmp6_error() can be called from both input and output path, 426 * and if we are in output path rcvif could contain bogus value. 427 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 428 * information in ip header (nip6). 429 */ 430 m->m_pkthdr.rcvif = NULL; 431 432 ICMP6_STATINC(ICMP6_STAT_OUTHIST + type); 433 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 434 435 return; 436 437 freeit: 438 /* 439 * If we can't tell whether or not we can generate ICMP6, free it. 440 */ 441 m_freem(m); 442} 443 444/* 445 * Process a received ICMP6 message. 446 */ 447int 448icmp6_input(struct mbuf **mp, int *offp, int proto) 449{ 450 struct mbuf *m = *mp, *n; 451 struct ip6_hdr *ip6, *nip6; 452 struct icmp6_hdr *icmp6, *nicmp6; 453 int off = *offp; 454 int icmp6len = m->m_pkthdr.len - *offp; 455 int code, sum, noff, i; 456 457#define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4) 458 KASSERT(ICMP6_MAXLEN < MCLBYTES); 459 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 460 461 /* 462 * Locate icmp6 structure in mbuf, and check 463 * that not corrupted and of at least minimum length 464 */ 465 466 if (icmp6len < sizeof(struct icmp6_hdr)) { 467 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 468 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 469 goto freeit; 470 } 471 472 i = off + sizeof(*icmp6); 473 if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) { 474 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 475#if 0 /* m is 0 here */ 476 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 477#endif 478 goto freeit; 479 } 480 ip6 = mtod(m, struct ip6_hdr *); 481 /* 482 * calculate the checksum 483 */ 484 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 485 if (icmp6 == NULL) { 486 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 487 /* m is invalid */ 488 /*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/ 489 return IPPROTO_DONE; 490 } 491 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 492 code = icmp6->icmp6_code; 493 494 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 495 nd6log(LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n", 496 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)); 497 ICMP6_STATINC(ICMP6_STAT_CHECKSUM); 498 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 499 goto freeit; 500 } 501 502#if defined(NFAITH) && 0 < NFAITH 503 if (faithprefix(&ip6->ip6_dst)) { 504 /* 505 * Deliver very specific ICMP6 type only. 506 * This is important to deliver TOOBIG. Otherwise PMTUD 507 * will not work. 508 */ 509 switch (icmp6->icmp6_type) { 510 case ICMP6_DST_UNREACH: 511 case ICMP6_PACKET_TOO_BIG: 512 case ICMP6_TIME_EXCEEDED: 513 break; 514 default: 515 goto freeit; 516 } 517 } 518#endif 519 520 ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type); 521 522 switch (icmp6->icmp6_type) { 523 case ICMP6_DST_UNREACH: 524 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 525 switch (code) { 526 case ICMP6_DST_UNREACH_NOROUTE: 527 code = PRC_UNREACH_NET; 528 break; 529 case ICMP6_DST_UNREACH_ADMIN: 530 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 531 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 532 break; 533 case ICMP6_DST_UNREACH_ADDR: 534 code = PRC_HOSTDEAD; 535 break; 536#ifdef COMPAT_RFC1885 537 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 538 code = PRC_UNREACH_SRCFAIL; 539 break; 540#else 541 case ICMP6_DST_UNREACH_BEYONDSCOPE: 542 /* I mean "source address was incorrect." */ 543 code = PRC_UNREACH_NET; 544 break; 545#endif 546 case ICMP6_DST_UNREACH_NOPORT: 547 code = PRC_UNREACH_PORT; 548 break; 549 default: 550 goto badcode; 551 } 552 goto deliver; 553 554 case ICMP6_PACKET_TOO_BIG: 555 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 556 557 /* 558 * MTU is checked in icmp6_mtudisc. 559 */ 560 code = PRC_MSGSIZE; 561 562 /* 563 * Updating the path MTU will be done after examining 564 * intermediate extension headers. 565 */ 566 goto deliver; 567 568 case ICMP6_TIME_EXCEEDED: 569 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 570 switch (code) { 571 case ICMP6_TIME_EXCEED_TRANSIT: 572 code = PRC_TIMXCEED_INTRANS; 573 break; 574 case ICMP6_TIME_EXCEED_REASSEMBLY: 575 code = PRC_TIMXCEED_REASS; 576 break; 577 default: 578 goto badcode; 579 } 580 goto deliver; 581 582 case ICMP6_PARAM_PROB: 583 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 584 switch (code) { 585 case ICMP6_PARAMPROB_NEXTHEADER: 586 code = PRC_UNREACH_PROTOCOL; 587 break; 588 case ICMP6_PARAMPROB_HEADER: 589 case ICMP6_PARAMPROB_OPTION: 590 code = PRC_PARAMPROB; 591 break; 592 default: 593 goto badcode; 594 } 595 goto deliver; 596 597 case ICMP6_ECHO_REQUEST: 598 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 599 if (code != 0) 600 goto badcode; 601 /* 602 * Copy mbuf to send to two data paths: userland socket(s), 603 * and to the querier (echo reply). 604 * m: a copy for socket, n: a copy for querier 605 * 606 * If the first mbuf is shared, or the first mbuf is too short, 607 * copy the first part of the data into a fresh mbuf. 608 * Otherwise, we will wrongly overwrite both copies. 609 */ 610 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 611 /* Give up local */ 612 n = m; 613 m = NULL; 614 } else if (M_READONLY(n) || 615 n->m_len < off + sizeof(struct icmp6_hdr)) { 616 struct mbuf *n0 = n; 617 618 /* 619 * Prepare an internal mbuf. m_pullup() doesn't 620 * always copy the length we specified. 621 */ 622 if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 623 /* Give up local */ 624 n = m; 625 m = NULL; 626 } 627 m_freem(n0); 628 } 629 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, 630 sizeof(*nicmp6)); 631 if (nicmp6 == NULL) 632 goto freeit; 633 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 634 nicmp6->icmp6_code = 0; 635 if (n) { 636 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 637 icmp6s[ICMP6_STAT_REFLECT]++; 638 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++; 639 ICMP6_STAT_PUTREF(); 640 icmp6_reflect(n, off); 641 } 642 if (!m) 643 goto freeit; 644 break; 645 646 case ICMP6_ECHO_REPLY: 647 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 648 if (code != 0) 649 goto badcode; 650 break; 651 652 case MLD_LISTENER_QUERY: 653 case MLD_LISTENER_REPORT: 654 if (icmp6len < sizeof(struct mld_hdr)) 655 goto badlen; 656 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */ 657 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 658 else 659 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 660 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 661 /* give up local */ 662 mld_input(m, off); 663 m = NULL; 664 goto freeit; 665 } 666 mld_input(n, off); 667 /* m stays. */ 668 break; 669 670 case MLD_LISTENER_DONE: 671 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 672 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */ 673 goto badlen; 674 break; /* nothing to be done in kernel */ 675 676 case MLD_MTRACE_RESP: 677 case MLD_MTRACE: 678 /* XXX: these two are experimental. not officially defined. */ 679 /* XXX: per-interface statistics? */ 680 break; /* just pass it to applications */ 681 682 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 683 { 684 enum { WRU, FQDN } mode; 685 686 if (!icmp6_nodeinfo) 687 break; 688 689 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 690 mode = WRU; 691 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 692 mode = FQDN; 693 else 694 goto badlen; 695 696 if (mode == FQDN) { 697 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 698 if (n) 699 n = ni6_input(n, off); 700 /* XXX meaningless if n == NULL */ 701 noff = sizeof(struct ip6_hdr); 702 } else { 703 u_char *p; 704 int maxhlen; 705 706 if ((icmp6_nodeinfo & 5) != 5) 707 break; 708 709 if (code != 0) 710 goto badcode; 711 MGETHDR(n, M_DONTWAIT, m->m_type); 712 if (n && ICMP6_MAXLEN > MHLEN) { 713 MCLGET(n, M_DONTWAIT); 714 if ((n->m_flags & M_EXT) == 0) { 715 m_free(n); 716 n = NULL; 717 } 718 } 719 if (n == NULL) { 720 /* Give up remote */ 721 break; 722 } 723 n->m_pkthdr.rcvif = NULL; 724 n->m_len = 0; 725 maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN; 726 if (maxhlen < 0) 727 break; 728 if (maxhlen > hostnamelen) 729 maxhlen = hostnamelen; 730 /* 731 * Copy IPv6 and ICMPv6 only. 732 */ 733 nip6 = mtod(n, struct ip6_hdr *); 734 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 735 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 736 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 737 p = (u_char *)(nicmp6 + 1); 738 memset(p, 0, 4); 739 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 740 noff = sizeof(struct ip6_hdr); 741 M_COPY_PKTHDR(n, m); /* just for rcvif */ 742 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 743 sizeof(struct icmp6_hdr) + 4 + maxhlen; 744 nicmp6->icmp6_type = ICMP6_WRUREPLY; 745 nicmp6->icmp6_code = 0; 746 } 747#undef hostnamelen 748 if (n) { 749 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 750 icmp6s[ICMP6_STAT_REFLECT]++; 751 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++; 752 ICMP6_STAT_PUTREF(); 753 icmp6_reflect(n, noff); 754 } 755 break; 756 } 757 758 case ICMP6_WRUREPLY: 759 if (code != 0) 760 goto badcode; 761 break; 762 763 case ND_ROUTER_SOLICIT: 764 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 765 if (code != 0) 766 goto badcode; 767 if (icmp6len < sizeof(struct nd_router_solicit)) 768 goto badlen; 769 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 770 /* give up local */ 771 nd6_rs_input(m, off, icmp6len); 772 m = NULL; 773 goto freeit; 774 } 775 nd6_rs_input(n, off, icmp6len); 776 /* m stays. */ 777 break; 778 779 case ND_ROUTER_ADVERT: 780 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 781 if (code != 0) 782 goto badcode; 783 if (icmp6len < sizeof(struct nd_router_advert)) 784 goto badlen; 785 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 786 /* give up local */ 787 nd6_ra_input(m, off, icmp6len); 788 m = NULL; 789 goto freeit; 790 } 791 nd6_ra_input(n, off, icmp6len); 792 /* m stays. */ 793 break; 794 795 case ND_NEIGHBOR_SOLICIT: 796 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 797 if (code != 0) 798 goto badcode; 799 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 800 goto badlen; 801 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 802 /* give up local */ 803 nd6_ns_input(m, off, icmp6len); 804 m = NULL; 805 goto freeit; 806 } 807 nd6_ns_input(n, off, icmp6len); 808 /* m stays. */ 809 break; 810 811 case ND_NEIGHBOR_ADVERT: 812 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 813 if (code != 0) 814 goto badcode; 815 if (icmp6len < sizeof(struct nd_neighbor_advert)) 816 goto badlen; 817 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 818 /* give up local */ 819 nd6_na_input(m, off, icmp6len); 820 m = NULL; 821 goto freeit; 822 } 823 nd6_na_input(n, off, icmp6len); 824 /* m stays. */ 825 break; 826 827 case ND_REDIRECT: 828 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 829 if (code != 0) 830 goto badcode; 831 if (icmp6len < sizeof(struct nd_redirect)) 832 goto badlen; 833 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 834 /* give up local */ 835 icmp6_redirect_input(m, off); 836 m = NULL; 837 goto freeit; 838 } 839 icmp6_redirect_input(n, off); 840 /* m stays. */ 841 break; 842 843 case ICMP6_ROUTER_RENUMBERING: 844 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 845 code != ICMP6_ROUTER_RENUMBERING_RESULT) 846 goto badcode; 847 if (icmp6len < sizeof(struct icmp6_router_renum)) 848 goto badlen; 849 break; 850 851 default: 852 nd6log(LOG_DEBUG, "unknown type %d(src=%s, dst=%s, ifid=%d)\n", 853 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 854 ip6_sprintf(&ip6->ip6_dst), 855 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0); 856 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 857 /* ICMPv6 error: MUST deliver it by spec... */ 858 code = PRC_NCMDS; 859 /* deliver */ 860 } else { 861 /* ICMPv6 informational: MUST not deliver */ 862 break; 863 } 864 deliver: 865 if (icmp6_notify_error(m, off, icmp6len, code)) { 866 /* In this case, m should've been freed. */ 867 return (IPPROTO_DONE); 868 } 869 break; 870 871 badcode: 872 ICMP6_STATINC(ICMP6_STAT_BADCODE); 873 break; 874 875 badlen: 876 ICMP6_STATINC(ICMP6_STAT_BADLEN); 877 break; 878 } 879 880 /* deliver the packet to appropriate sockets */ 881 icmp6_rip6_input(&m, *offp); 882 883 return IPPROTO_DONE; 884 885 freeit: 886 m_freem(m); 887 return IPPROTO_DONE; 888} 889 890static int 891icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code) 892{ 893 struct icmp6_hdr *icmp6; 894 struct ip6_hdr *eip6; 895 u_int32_t notifymtu; 896 struct sockaddr_in6 icmp6src, icmp6dst; 897 898 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 899 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 900 goto freeit; 901 } 902 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 903 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 904 if (icmp6 == NULL) { 905 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 906 return (-1); 907 } 908 eip6 = (struct ip6_hdr *)(icmp6 + 1); 909 910 /* Detect the upper level protocol */ 911 { 912 void (*ctlfunc)(int, struct sockaddr *, void *); 913 u_int8_t nxt = eip6->ip6_nxt; 914 int eoff = off + sizeof(struct icmp6_hdr) + 915 sizeof(struct ip6_hdr); 916 struct ip6ctlparam ip6cp; 917 struct in6_addr *finaldst = NULL; 918 int icmp6type = icmp6->icmp6_type; 919 struct ip6_frag *fh; 920 struct ip6_rthdr *rth; 921 struct ip6_rthdr0 *rth0; 922 int rthlen; 923 924 while (1) { /* XXX: should avoid infinite loop explicitly? */ 925 struct ip6_ext *eh; 926 927 switch (nxt) { 928 case IPPROTO_HOPOPTS: 929 case IPPROTO_DSTOPTS: 930 case IPPROTO_AH: 931 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 932 eoff, sizeof(*eh)); 933 if (eh == NULL) { 934 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 935 return (-1); 936 } 937 938 if (nxt == IPPROTO_AH) 939 eoff += (eh->ip6e_len + 2) << 2; 940 else 941 eoff += (eh->ip6e_len + 1) << 3; 942 nxt = eh->ip6e_nxt; 943 break; 944 case IPPROTO_ROUTING: 945 /* 946 * When the erroneous packet contains a 947 * routing header, we should examine the 948 * header to determine the final destination. 949 * Otherwise, we can't properly update 950 * information that depends on the final 951 * destination (e.g. path MTU). 952 */ 953 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 954 eoff, sizeof(*rth)); 955 if (rth == NULL) { 956 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 957 return (-1); 958 } 959 rthlen = (rth->ip6r_len + 1) << 3; 960 /* 961 * XXX: currently there is no 962 * officially defined type other 963 * than type-0. 964 * Note that if the segment left field 965 * is 0, all intermediate hops must 966 * have been passed. 967 */ 968 if (rth->ip6r_segleft && 969 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 970 int hops; 971 972 IP6_EXTHDR_GET(rth0, 973 struct ip6_rthdr0 *, m, 974 eoff, rthlen); 975 if (rth0 == NULL) { 976 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 977 return (-1); 978 } 979 /* just ignore a bogus header */ 980 if ((rth0->ip6r0_len % 2) == 0 && 981 (hops = rth0->ip6r0_len/2)) 982 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 983 } 984 eoff += rthlen; 985 nxt = rth->ip6r_nxt; 986 break; 987 case IPPROTO_FRAGMENT: 988 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 989 eoff, sizeof(*fh)); 990 if (fh == NULL) { 991 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 992 return (-1); 993 } 994 /* 995 * Data after a fragment header is meaningless 996 * unless it is the first fragment, but 997 * we'll go to the notify label for path MTU 998 * discovery. 999 */ 1000 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1001 goto notify; 1002 1003 eoff += sizeof(struct ip6_frag); 1004 nxt = fh->ip6f_nxt; 1005 break; 1006 default: 1007 /* 1008 * This case includes ESP and the No Next 1009 * Header. In such cases going to the notify 1010 * label does not have any meaning 1011 * (i.e. ctlfunc will be NULL), but we go 1012 * anyway since we might have to update 1013 * path MTU information. 1014 */ 1015 goto notify; 1016 } 1017 } 1018 notify: 1019 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1020 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1021 if (icmp6 == NULL) { 1022 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 1023 return (-1); 1024 } 1025 1026 /* 1027 * retrieve parameters from the inner IPv6 header, and convert 1028 * them into sockaddr structures. 1029 * XXX: there is no guarantee that the source or destination 1030 * addresses of the inner packet are in the same scope zone as 1031 * the addresses of the icmp packet. But there is no other 1032 * way to determine the zone. 1033 */ 1034 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1035 1036 sockaddr_in6_init(&icmp6dst, 1037 (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0); 1038 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1039 goto freeit; 1040 sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0); 1041 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1042 goto freeit; 1043 icmp6src.sin6_flowinfo = 1044 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1045 1046 if (finaldst == NULL) 1047 finaldst = &eip6->ip6_dst; 1048 ip6cp.ip6c_m = m; 1049 ip6cp.ip6c_icmp6 = icmp6; 1050 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1051 ip6cp.ip6c_off = eoff; 1052 ip6cp.ip6c_finaldst = finaldst; 1053 ip6cp.ip6c_src = &icmp6src; 1054 ip6cp.ip6c_nxt = nxt; 1055 1056 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1057 notifymtu = ntohl(icmp6->icmp6_mtu); 1058 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1059 } 1060 1061 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1062 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1063 if (ctlfunc) { 1064 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1065 &ip6cp); 1066 } 1067 } 1068 return (0); 1069 1070 freeit: 1071 m_freem(m); 1072 return (-1); 1073} 1074 1075void 1076icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1077{ 1078 unsigned long rtcount; 1079 struct icmp6_mtudisc_callback *mc; 1080 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1081 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1082 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1083 u_int mtu = ntohl(icmp6->icmp6_mtu); 1084 struct rtentry *rt = NULL; 1085 struct sockaddr_in6 sin6; 1086 1087 /* 1088 * The MTU should not be less than the minimal IPv6 MTU except for the 1089 * hack in ip6_output/ip6_setpmtu where we always include a frag header. 1090 * In that one case, the MTU might be less than 1280. 1091 */ 1092 if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) { 1093 /* is the mtu even sane? */ 1094 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) 1095 return; 1096 if (!validated) 1097 return; 1098 mtu = IPV6_MMTU - sizeof(struct ip6_frag); 1099 } 1100 1101 /* 1102 * allow non-validated cases if memory is plenty, to make traffic 1103 * from non-connected pcb happy. 1104 */ 1105 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1106 if (validated) { 1107 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1108 return; 1109 else if (0 <= icmp6_mtudisc_lowat && 1110 rtcount > icmp6_mtudisc_lowat) { 1111 /* 1112 * XXX nuke a victim, install the new one. 1113 */ 1114 } 1115 } else { 1116 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1117 return; 1118 } 1119 1120 memset(&sin6, 0, sizeof(sin6)); 1121 sin6.sin6_family = PF_INET6; 1122 sin6.sin6_len = sizeof(struct sockaddr_in6); 1123 sin6.sin6_addr = *dst; 1124 if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1125 return; 1126 1127 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1128 1129 if (rt && (rt->rt_flags & RTF_HOST) && 1130 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1131 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1132 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1133 ICMP6_STATINC(ICMP6_STAT_PMTUCHG); 1134 rt->rt_rmx.rmx_mtu = mtu; 1135 } 1136 } 1137 if (rt) { 1138 rtfree(rt); 1139 } 1140 1141 /* 1142 * Notify protocols that the MTU for this destination 1143 * has changed. 1144 */ 1145 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1146 mc = LIST_NEXT(mc, mc_list)) 1147 (*mc->mc_func)(&sin6.sin6_addr); 1148} 1149 1150/* 1151 * Process a Node Information Query packet, based on 1152 * draft-ietf-ipngwg-icmp-name-lookups-07. 1153 * 1154 * Spec incompatibilities: 1155 * - IPv6 Subject address handling 1156 * - IPv4 Subject address handling support missing 1157 * - Proxy reply (answer even if it's not for me) 1158 * - joins NI group address at in6_ifattach() time only, does not cope 1159 * with hostname changes by sethostname(3) 1160 */ 1161static struct mbuf * 1162ni6_input(struct mbuf *m, int off) 1163{ 1164 struct icmp6_nodeinfo *ni6, *nni6; 1165 struct mbuf *n = NULL; 1166 u_int16_t qtype; 1167 int subjlen; 1168 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1169 struct ni_reply_fqdn *fqdn; 1170 int addrs; /* for NI_QTYPE_NODEADDR */ 1171 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1172 struct sockaddr_in6 sin6; /* ip6_dst */ 1173 struct in6_addr in6_subj; /* subject address */ 1174 struct ip6_hdr *ip6; 1175 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1176 char *subj = NULL; 1177 1178 ip6 = mtod(m, struct ip6_hdr *); 1179 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1180 if (ni6 == NULL) { 1181 /* m is already reclaimed */ 1182 return NULL; 1183 } 1184 1185 /* 1186 * Validate IPv6 destination address. 1187 * 1188 * The Responder must discard the Query without further processing 1189 * unless it is one of the Responder's unicast or anycast addresses, or 1190 * a link-local scope multicast address which the Responder has joined. 1191 * [icmp-name-lookups-07, Section 4.] 1192 */ 1193 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 1194 /* XXX scopeid */ 1195 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1196 ; /* unicast/anycast, fine */ 1197 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1198 ; /* link-local multicast, fine */ 1199 else 1200 goto bad; 1201 1202 /* validate query Subject field. */ 1203 qtype = ntohs(ni6->ni_qtype); 1204 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1205 switch (qtype) { 1206 case NI_QTYPE_NOOP: 1207 case NI_QTYPE_SUPTYPES: 1208 /* 07 draft */ 1209 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1210 break; 1211 /* FALLTHROUGH */ 1212 case NI_QTYPE_FQDN: 1213 case NI_QTYPE_NODEADDR: 1214 case NI_QTYPE_IPV4ADDR: 1215 switch (ni6->ni_code) { 1216 case ICMP6_NI_SUBJ_IPV6: 1217#if ICMP6_NI_SUBJ_IPV6 != 0 1218 case 0: 1219#endif 1220 /* 1221 * backward compatibility - try to accept 03 draft 1222 * format, where no Subject is present. 1223 */ 1224 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1225 subjlen == 0) { 1226 oldfqdn++; 1227 break; 1228 } 1229#if ICMP6_NI_SUBJ_IPV6 != 0 1230 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1231 goto bad; 1232#endif 1233 1234 if (subjlen != sizeof(sin6.sin6_addr)) 1235 goto bad; 1236 1237 /* 1238 * Validate Subject address. 1239 * 1240 * Not sure what exactly "address belongs to the node" 1241 * means in the spec, is it just unicast, or what? 1242 * 1243 * At this moment we consider Subject address as 1244 * "belong to the node" if the Subject address equals 1245 * to the IPv6 destination address; validation for 1246 * IPv6 destination address should have done enough 1247 * check for us. 1248 * 1249 * We do not do proxy at this moment. 1250 */ 1251 /* m_pulldown instead of copy? */ 1252 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1253 subjlen, (void *)&in6_subj); 1254 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1255 goto bad; 1256 1257 subj = (char *)&in6_subj; 1258 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1259 break; 1260 1261 /* 1262 * XXX if we are to allow other cases, we should really 1263 * be careful about scope here. 1264 * basically, we should disallow queries toward IPv6 1265 * destination X with subject Y, if scope(X) > scope(Y). 1266 * if we allow scope(X) > scope(Y), it will result in 1267 * information leakage across scope boundary. 1268 */ 1269 goto bad; 1270 1271 case ICMP6_NI_SUBJ_FQDN: 1272 /* 1273 * Validate Subject name with gethostname(3). 1274 * 1275 * The behavior may need some debate, since: 1276 * - we are not sure if the node has FQDN as 1277 * hostname (returned by gethostname(3)). 1278 * - the code does wildcard match for truncated names. 1279 * however, we are not sure if we want to perform 1280 * wildcard match, if gethostname(3) side has 1281 * truncated hostname. 1282 */ 1283 n = ni6_nametodns(hostname, hostnamelen, 0); 1284 if (!n || n->m_next || n->m_len == 0) 1285 goto bad; 1286 IP6_EXTHDR_GET(subj, char *, m, 1287 off + sizeof(struct icmp6_nodeinfo), subjlen); 1288 if (subj == NULL) 1289 goto bad; 1290 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1291 n->m_len)) { 1292 goto bad; 1293 } 1294 m_freem(n); 1295 n = NULL; 1296 break; 1297 1298 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1299 default: 1300 goto bad; 1301 } 1302 break; 1303 } 1304 1305 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1306 switch (qtype) { 1307 case NI_QTYPE_FQDN: 1308 if ((icmp6_nodeinfo & 1) == 0) 1309 goto bad; 1310 break; 1311 case NI_QTYPE_NODEADDR: 1312 case NI_QTYPE_IPV4ADDR: 1313 if ((icmp6_nodeinfo & 2) == 0) 1314 goto bad; 1315 break; 1316 } 1317 1318 /* guess reply length */ 1319 switch (qtype) { 1320 case NI_QTYPE_NOOP: 1321 break; /* no reply data */ 1322 case NI_QTYPE_SUPTYPES: 1323 replylen += sizeof(u_int32_t); 1324 break; 1325 case NI_QTYPE_FQDN: 1326 /* XXX will append an mbuf */ 1327 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1328 break; 1329 case NI_QTYPE_NODEADDR: 1330 addrs = ni6_addrs(ni6, m, &ifp, subj); 1331 if ((replylen += addrs * (sizeof(struct in6_addr) + 1332 sizeof(u_int32_t))) > MCLBYTES) 1333 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1334 break; 1335 case NI_QTYPE_IPV4ADDR: 1336 /* unsupported - should respond with unknown Qtype? */ 1337 goto bad; 1338 default: 1339 /* 1340 * XXX: We must return a reply with the ICMP6 code 1341 * `unknown Qtype' in this case. However we regard the case 1342 * as an FQDN query for backward compatibility. 1343 * Older versions set a random value to this field, 1344 * so it rarely varies in the defined qtypes. 1345 * But the mechanism is not reliable... 1346 * maybe we should obsolete older versions. 1347 */ 1348 qtype = NI_QTYPE_FQDN; 1349 /* XXX will append an mbuf */ 1350 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1351 oldfqdn++; 1352 break; 1353 } 1354 1355 /* allocate an mbuf to reply. */ 1356 MGETHDR(n, M_DONTWAIT, m->m_type); 1357 if (n == NULL) { 1358 m_freem(m); 1359 return (NULL); 1360 } 1361 M_MOVE_PKTHDR(n, m); /* just for rcvif */ 1362 if (replylen > MHLEN) { 1363 if (replylen > MCLBYTES) { 1364 /* 1365 * XXX: should we try to allocate more? But MCLBYTES 1366 * is probably much larger than IPV6_MMTU... 1367 */ 1368 goto bad; 1369 } 1370 MCLGET(n, M_DONTWAIT); 1371 if ((n->m_flags & M_EXT) == 0) { 1372 goto bad; 1373 } 1374 } 1375 n->m_pkthdr.len = n->m_len = replylen; 1376 1377 /* copy mbuf header and IPv6 + Node Information base headers */ 1378 bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr)); 1379 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1380 bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo)); 1381 1382 /* qtype dependent procedure */ 1383 switch (qtype) { 1384 case NI_QTYPE_NOOP: 1385 nni6->ni_code = ICMP6_NI_SUCCESS; 1386 nni6->ni_flags = 0; 1387 break; 1388 case NI_QTYPE_SUPTYPES: 1389 { 1390 u_int32_t v; 1391 nni6->ni_code = ICMP6_NI_SUCCESS; 1392 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1393 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1394 v = (u_int32_t)htonl(0x0000000f); 1395 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1396 break; 1397 } 1398 case NI_QTYPE_FQDN: 1399 nni6->ni_code = ICMP6_NI_SUCCESS; 1400 fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) + 1401 sizeof(struct ip6_hdr) + 1402 sizeof(struct icmp6_nodeinfo)); 1403 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1404 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1405 /* 1406 * XXX do we really have FQDN in variable "hostname"? 1407 */ 1408 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1409 if (n->m_next == NULL) 1410 goto bad; 1411 /* XXX we assume that n->m_next is not a chain */ 1412 if (n->m_next->m_next != NULL) 1413 goto bad; 1414 n->m_pkthdr.len += n->m_next->m_len; 1415 break; 1416 case NI_QTYPE_NODEADDR: 1417 { 1418 int lenlim, copied; 1419 1420 nni6->ni_code = ICMP6_NI_SUCCESS; 1421 n->m_pkthdr.len = n->m_len = 1422 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1423 lenlim = M_TRAILINGSPACE(n); 1424 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1425 /* XXX: reset mbuf length */ 1426 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1427 sizeof(struct icmp6_nodeinfo) + copied; 1428 break; 1429 } 1430 default: 1431 break; /* XXX impossible! */ 1432 } 1433 1434 nni6->ni_type = ICMP6_NI_REPLY; 1435 m_freem(m); 1436 return (n); 1437 1438 bad: 1439 m_freem(m); 1440 if (n) 1441 m_freem(n); 1442 return (NULL); 1443} 1444#undef hostnamelen 1445 1446#define isupper(x) ('A' <= (x) && (x) <= 'Z') 1447#define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1448#define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1449#define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1450 1451/* 1452 * make a mbuf with DNS-encoded string. no compression support. 1453 * 1454 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1455 * treated as truncated name (two \0 at the end). this is a wild guess. 1456 * 1457 * old - return pascal string if non-zero 1458 */ 1459static struct mbuf * 1460ni6_nametodns(const char *name, int namelen, int old) 1461{ 1462 struct mbuf *m; 1463 char *cp, *ep; 1464 const char *p, *q; 1465 int i, len, nterm; 1466 1467 if (old) 1468 len = namelen + 1; 1469 else 1470 len = MCLBYTES; 1471 1472 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1473 MGET(m, M_DONTWAIT, MT_DATA); 1474 if (m && len > MLEN) { 1475 MCLGET(m, M_DONTWAIT); 1476 if ((m->m_flags & M_EXT) == 0) 1477 goto fail; 1478 } 1479 if (!m) 1480 goto fail; 1481 m->m_next = NULL; 1482 1483 if (old) { 1484 m->m_len = len; 1485 *mtod(m, char *) = namelen; 1486 bcopy(name, mtod(m, char *) + 1, namelen); 1487 return m; 1488 } else { 1489 m->m_len = 0; 1490 cp = mtod(m, char *); 1491 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1492 1493 /* if not certain about my name, return empty buffer */ 1494 if (namelen == 0) 1495 return m; 1496 1497 /* 1498 * guess if it looks like shortened hostname, or FQDN. 1499 * shortened hostname needs two trailing "\0". 1500 */ 1501 i = 0; 1502 for (p = name; p < name + namelen; p++) { 1503 if (*p && *p == '.') 1504 i++; 1505 } 1506 if (i < 2) 1507 nterm = 2; 1508 else 1509 nterm = 1; 1510 1511 p = name; 1512 while (cp < ep && p < name + namelen) { 1513 i = 0; 1514 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1515 i++; 1516 /* result does not fit into mbuf */ 1517 if (cp + i + 1 >= ep) 1518 goto fail; 1519 /* 1520 * DNS label length restriction, RFC1035 page 8. 1521 * "i == 0" case is included here to avoid returning 1522 * 0-length label on "foo..bar". 1523 */ 1524 if (i <= 0 || i >= 64) 1525 goto fail; 1526 *cp++ = i; 1527 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1528 goto fail; 1529 while (i > 0) { 1530 if (!isalnum(*p) && *p != '-') 1531 goto fail; 1532 if (isupper(*p)) { 1533 *cp++ = tolower(*p); 1534 p++; 1535 } else 1536 *cp++ = *p++; 1537 i--; 1538 } 1539 p = q; 1540 if (p < name + namelen && *p == '.') 1541 p++; 1542 } 1543 /* termination */ 1544 if (cp + nterm >= ep) 1545 goto fail; 1546 while (nterm-- > 0) 1547 *cp++ = '\0'; 1548 m->m_len = cp - mtod(m, char *); 1549 return m; 1550 } 1551 1552 panic("should not reach here"); 1553 /* NOTREACHED */ 1554 1555 fail: 1556 if (m) 1557 m_freem(m); 1558 return NULL; 1559} 1560 1561/* 1562 * check if two DNS-encoded string matches. takes care of truncated 1563 * form (with \0\0 at the end). no compression support. 1564 * XXX upper/lowercase match (see RFC2065) 1565 */ 1566static int 1567ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1568{ 1569 const char *a0, *b0; 1570 int l; 1571 1572 /* simplest case - need validation? */ 1573 if (alen == blen && memcmp(a, b, alen) == 0) 1574 return 1; 1575 1576 a0 = a; 1577 b0 = b; 1578 1579 /* termination is mandatory */ 1580 if (alen < 2 || blen < 2) 1581 return 0; 1582 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1583 return 0; 1584 alen--; 1585 blen--; 1586 1587 while (a - a0 < alen && b - b0 < blen) { 1588 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1589 return 0; 1590 1591 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1592 return 0; 1593 /* we don't support compression yet */ 1594 if (a[0] >= 64 || b[0] >= 64) 1595 return 0; 1596 1597 /* truncated case */ 1598 if (a[0] == 0 && a - a0 == alen - 1) 1599 return 1; 1600 if (b[0] == 0 && b - b0 == blen - 1) 1601 return 1; 1602 if (a[0] == 0 || b[0] == 0) 1603 return 0; 1604 1605 if (a[0] != b[0]) 1606 return 0; 1607 l = a[0]; 1608 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1609 return 0; 1610 if (memcmp(a + 1, b + 1, l) != 0) 1611 return 0; 1612 1613 a += 1 + l; 1614 b += 1 + l; 1615 } 1616 1617 if (a - a0 == alen && b - b0 == blen) 1618 return 1; 1619 else 1620 return 0; 1621} 1622 1623/* 1624 * calculate the number of addresses to be returned in the node info reply. 1625 */ 1626static int 1627ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, 1628 struct ifnet **ifpp, char *subj) 1629{ 1630 struct ifnet *ifp; 1631 struct in6_ifaddr *ifa6; 1632 struct ifaddr *ifa; 1633 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1634 int addrs = 0, addrsofif, iffound = 0; 1635 int niflags = ni6->ni_flags; 1636 1637 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1638 switch (ni6->ni_code) { 1639 case ICMP6_NI_SUBJ_IPV6: 1640 if (subj == NULL) /* must be impossible... */ 1641 return (0); 1642 subj_ip6 = (struct sockaddr_in6 *)subj; 1643 break; 1644 default: 1645 /* 1646 * XXX: we only support IPv6 subject address for 1647 * this Qtype. 1648 */ 1649 return (0); 1650 } 1651 } 1652 1653 IFNET_FOREACH(ifp) { 1654 addrsofif = 0; 1655 IFADDR_FOREACH(ifa, ifp) { 1656 if (ifa->ifa_addr->sa_family != AF_INET6) 1657 continue; 1658 ifa6 = (struct in6_ifaddr *)ifa; 1659 1660 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1661 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1662 &ifa6->ia_addr.sin6_addr)) 1663 iffound = 1; 1664 1665 /* 1666 * IPv4-mapped addresses can only be returned by a 1667 * Node Information proxy, since they represent 1668 * addresses of IPv4-only nodes, which perforce do 1669 * not implement this protocol. 1670 * [icmp-name-lookups-07, Section 5.4] 1671 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1672 * this function at this moment. 1673 */ 1674 1675 /* What do we have to do about ::1? */ 1676 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1677 case IPV6_ADDR_SCOPE_LINKLOCAL: 1678 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1679 continue; 1680 break; 1681 case IPV6_ADDR_SCOPE_SITELOCAL: 1682 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1683 continue; 1684 break; 1685 case IPV6_ADDR_SCOPE_GLOBAL: 1686 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1687 continue; 1688 break; 1689 default: 1690 continue; 1691 } 1692 1693 /* 1694 * check if anycast is okay. 1695 * XXX: just experimental. not in the spec. 1696 */ 1697 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1698 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1699 continue; /* we need only unicast addresses */ 1700 1701 addrsofif++; /* count the address */ 1702 } 1703 if (iffound) { 1704 *ifpp = ifp; 1705 return (addrsofif); 1706 } 1707 1708 addrs += addrsofif; 1709 } 1710 1711 return (addrs); 1712} 1713 1714static int 1715ni6_store_addrs(struct icmp6_nodeinfo *ni6, 1716 struct icmp6_nodeinfo *nni6, struct ifnet *ifp0, 1717 int resid) 1718{ 1719 struct ifnet *ifp = ifp0 ? ifp0 : IFNET_FIRST(); 1720 struct in6_ifaddr *ifa6; 1721 struct ifaddr *ifa; 1722 struct ifnet *ifp_dep = NULL; 1723 int copied = 0, allow_deprecated = 0; 1724 u_char *cp = (u_char *)(nni6 + 1); 1725 int niflags = ni6->ni_flags; 1726 u_int32_t ltime; 1727 1728 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1729 return (0); /* needless to copy */ 1730 1731 again: 1732 1733 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1734 { 1735 IFADDR_FOREACH(ifa, ifp) { 1736 if (ifa->ifa_addr->sa_family != AF_INET6) 1737 continue; 1738 ifa6 = (struct in6_ifaddr *)ifa; 1739 1740 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1741 allow_deprecated == 0) { 1742 /* 1743 * prefererred address should be put before 1744 * deprecated addresses. 1745 */ 1746 1747 /* record the interface for later search */ 1748 if (ifp_dep == NULL) 1749 ifp_dep = ifp; 1750 1751 continue; 1752 } 1753 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1754 allow_deprecated != 0) 1755 continue; /* we now collect deprecated addrs */ 1756 1757 /* What do we have to do about ::1? */ 1758 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1759 case IPV6_ADDR_SCOPE_LINKLOCAL: 1760 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1761 continue; 1762 break; 1763 case IPV6_ADDR_SCOPE_SITELOCAL: 1764 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1765 continue; 1766 break; 1767 case IPV6_ADDR_SCOPE_GLOBAL: 1768 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1769 continue; 1770 break; 1771 default: 1772 continue; 1773 } 1774 1775 /* 1776 * check if anycast is okay. 1777 * XXX: just experimental. not in the spec. 1778 */ 1779 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1780 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1781 continue; 1782 1783 /* now we can copy the address */ 1784 if (resid < sizeof(struct in6_addr) + 1785 sizeof(u_int32_t)) { 1786 /* 1787 * We give up much more copy. 1788 * Set the truncate flag and return. 1789 */ 1790 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1791 return (copied); 1792 } 1793 1794 /* 1795 * Set the TTL of the address. 1796 * The TTL value should be one of the following 1797 * according to the specification: 1798 * 1799 * 1. The remaining lifetime of a DHCP lease on the 1800 * address, or 1801 * 2. The remaining Valid Lifetime of a prefix from 1802 * which the address was derived through Stateless 1803 * Autoconfiguration. 1804 * 1805 * Note that we currently do not support stateful 1806 * address configuration by DHCPv6, so the former 1807 * case can't happen. 1808 * 1809 * TTL must be 2^31 > TTL >= 0. 1810 */ 1811 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1812 ltime = ND6_INFINITE_LIFETIME; 1813 else { 1814 if (ifa6->ia6_lifetime.ia6t_expire > 1815 time_uptime) 1816 ltime = ifa6->ia6_lifetime.ia6t_expire - 1817 time_uptime; 1818 else 1819 ltime = 0; 1820 } 1821 if (ltime > 0x7fffffff) 1822 ltime = 0x7fffffff; 1823 ltime = htonl(ltime); 1824 1825 bcopy(<ime, cp, sizeof(u_int32_t)); 1826 cp += sizeof(u_int32_t); 1827 1828 /* copy the address itself */ 1829 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1830 sizeof(struct in6_addr)); 1831 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1832 cp += sizeof(struct in6_addr); 1833 1834 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1835 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1836 } 1837 if (ifp0) /* we need search only on the specified IF */ 1838 break; 1839 } 1840 1841 if (allow_deprecated == 0 && ifp_dep != NULL) { 1842 ifp = ifp_dep; 1843 allow_deprecated = 1; 1844 1845 goto again; 1846 } 1847 1848 return (copied); 1849} 1850 1851/* 1852 * XXX almost dup'ed code with rip6_input. 1853 */ 1854static int 1855icmp6_rip6_input(struct mbuf **mp, int off) 1856{ 1857 struct mbuf *m = *mp; 1858 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1859 struct inpcb_hdr *inph; 1860 struct in6pcb *in6p; 1861 struct in6pcb *last = NULL; 1862 struct sockaddr_in6 rip6src; 1863 struct icmp6_hdr *icmp6; 1864 struct mbuf *opts = NULL; 1865 1866 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1867 if (icmp6 == NULL) { 1868 /* m is already reclaimed */ 1869 return IPPROTO_DONE; 1870 } 1871 1872 /* 1873 * XXX: the address may have embedded scope zone ID, which should be 1874 * hidden from applications. 1875 */ 1876 sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0); 1877 if (sa6_recoverscope(&rip6src)) { 1878 m_freem(m); 1879 return (IPPROTO_DONE); 1880 } 1881 1882 TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) { 1883 in6p = (struct in6pcb *)inph; 1884 if (in6p->in6p_af != AF_INET6) 1885 continue; 1886 if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6) 1887 continue; 1888 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1889 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1890 continue; 1891 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1892 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1893 continue; 1894 if (in6p->in6p_icmp6filt 1895 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1896 in6p->in6p_icmp6filt)) 1897 continue; 1898 if (last) { 1899 struct mbuf *n; 1900 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1901 if (last->in6p_flags & IN6P_CONTROLOPTS) 1902 ip6_savecontrol(last, &opts, ip6, n); 1903 /* strip intermediate headers */ 1904 m_adj(n, off); 1905 if (sbappendaddr(&last->in6p_socket->so_rcv, 1906 (struct sockaddr *)&rip6src, 1907 n, opts) == 0) { 1908 /* should notify about lost packet */ 1909 m_freem(n); 1910 if (opts) 1911 m_freem(opts); 1912 } else 1913 sorwakeup(last->in6p_socket); 1914 opts = NULL; 1915 } 1916 } 1917 last = in6p; 1918 } 1919 if (last) { 1920 if (last->in6p_flags & IN6P_CONTROLOPTS) 1921 ip6_savecontrol(last, &opts, ip6, m); 1922 /* strip intermediate headers */ 1923 m_adj(m, off); 1924 if (sbappendaddr(&last->in6p_socket->so_rcv, 1925 (struct sockaddr *)&rip6src, m, opts) == 0) { 1926 m_freem(m); 1927 if (opts) 1928 m_freem(opts); 1929 } else 1930 sorwakeup(last->in6p_socket); 1931 } else { 1932 m_freem(m); 1933 IP6_STATDEC(IP6_STAT_DELIVERED); 1934 } 1935 return IPPROTO_DONE; 1936} 1937 1938/* 1939 * Reflect the ip6 packet back to the source. 1940 * OFF points to the icmp6 header, counted from the top of the mbuf. 1941 * 1942 * Note: RFC 1885 required that an echo reply should be truncated if it 1943 * did not fit in with (return) path MTU, and KAME code supported the 1944 * behavior. However, as a clarification after the RFC, this limitation 1945 * was removed in a revised version of the spec, RFC 2463. We had kept the 1946 * old behavior, with a (non-default) ifdef block, while the new version of 1947 * the spec was an internet-draft status, and even after the new RFC was 1948 * published. But it would rather make sense to clean the obsoleted part 1949 * up, and to make the code simpler at this stage. 1950 */ 1951void 1952icmp6_reflect(struct mbuf *m, size_t off) 1953{ 1954 struct ip6_hdr *ip6; 1955 struct icmp6_hdr *icmp6; 1956 const struct in6_ifaddr *ia; 1957 const struct ip6aux *ip6a; 1958 int plen; 1959 int type, code; 1960 struct ifnet *outif = NULL; 1961 struct in6_addr origdst; 1962 const struct in6_addr *src = NULL; 1963 1964 /* too short to reflect */ 1965 if (off < sizeof(struct ip6_hdr)) { 1966 nd6log(LOG_DEBUG, 1967 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 1968 (u_long)off, (u_long)sizeof(struct ip6_hdr), 1969 __FILE__, __LINE__); 1970 goto bad; 1971 } 1972 1973 /* 1974 * If there are extra headers between IPv6 and ICMPv6, strip 1975 * off that header first. 1976 */ 1977#ifdef DIAGNOSTIC 1978 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 1979 panic("assumption failed in icmp6_reflect"); 1980#endif 1981 if (off > sizeof(struct ip6_hdr)) { 1982 size_t l; 1983 struct ip6_hdr nip6; 1984 1985 l = off - sizeof(struct ip6_hdr); 1986 m_copydata(m, 0, sizeof(nip6), (void *)&nip6); 1987 m_adj(m, l); 1988 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1989 if (m->m_len < l) { 1990 if ((m = m_pullup(m, l)) == NULL) 1991 return; 1992 } 1993 bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6)); 1994 } else /* off == sizeof(struct ip6_hdr) */ { 1995 size_t l; 1996 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1997 if (m->m_len < l) { 1998 if ((m = m_pullup(m, l)) == NULL) 1999 return; 2000 } 2001 } 2002 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2003 ip6 = mtod(m, struct ip6_hdr *); 2004 ip6->ip6_nxt = IPPROTO_ICMPV6; 2005 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2006 type = icmp6->icmp6_type; /* keep type for statistics */ 2007 code = icmp6->icmp6_code; /* ditto. */ 2008 2009 origdst = ip6->ip6_dst; 2010 /* 2011 * ip6_input() drops a packet if its src is multicast. 2012 * So, the src is never multicast. 2013 */ 2014 ip6->ip6_dst = ip6->ip6_src; 2015 2016 /* 2017 * If the incoming packet was addressed directly to us (i.e. unicast), 2018 * use dst as the src for the reply. 2019 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2020 * (for example) when we encounter an error while forwarding procedure 2021 * destined to a duplicated address of ours. 2022 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2023 * procedure of an outgoing packet of our own, in which case we need 2024 * to search in the ifaddr list. 2025 */ 2026 if (IN6_IS_ADDR_MULTICAST(&origdst)) 2027 ; 2028 else if ((ip6a = ip6_getdstifaddr(m)) != NULL) { 2029 if ((ip6a->ip6a_flags & 2030 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2031 src = &ip6a->ip6a_src; 2032 } else { 2033 union { 2034 struct sockaddr_in6 sin6; 2035 struct sockaddr sa; 2036 } u; 2037 2038 sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0); 2039 2040 ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa); 2041 2042 if (ia == NULL) 2043 ; 2044 else if ((ia->ia6_flags & 2045 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2046 src = &ia->ia_addr.sin6_addr; 2047 } 2048 2049 if (src == NULL) { 2050 int e; 2051 struct sockaddr_in6 sin6; 2052 struct route ro; 2053 2054 /* 2055 * This case matches to multicasts, our anycast, or unicasts 2056 * that we do not own. Select a source address based on the 2057 * source address of the erroneous packet. 2058 */ 2059 /* zone ID should be embedded */ 2060 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 2061 2062 memset(&ro, 0, sizeof(ro)); 2063 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2064 rtcache_free(&ro); 2065 if (src == NULL) { 2066 nd6log(LOG_DEBUG, 2067 "source can't be determined: " 2068 "dst=%s, error=%d\n", 2069 ip6_sprintf(&sin6.sin6_addr), e); 2070 goto bad; 2071 } 2072 } 2073 2074 ip6->ip6_src = *src; 2075 ip6->ip6_flow = 0; 2076 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2077 ip6->ip6_vfc |= IPV6_VERSION; 2078 ip6->ip6_nxt = IPPROTO_ICMPV6; 2079 if (m->m_pkthdr.rcvif) { 2080 /* XXX: This may not be the outgoing interface */ 2081 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2082 } else 2083 ip6->ip6_hlim = ip6_defhlim; 2084 2085 m->m_pkthdr.csum_flags = 0; 2086 icmp6->icmp6_cksum = 0; 2087 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2088 sizeof(struct ip6_hdr), plen); 2089 2090 /* 2091 * XXX option handling 2092 */ 2093 2094 m->m_flags &= ~(M_BCAST|M_MCAST); 2095 2096 /* 2097 * To avoid a "too big" situation at an intermediate router 2098 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2099 * Note that only echo and node information replies are affected, 2100 * since the length of ICMP6 errors is limited to the minimum MTU. 2101 */ 2102 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 && 2103 outif) 2104 icmp6_ifstat_inc(outif, ifs6_out_error); 2105 2106 if (outif) 2107 icmp6_ifoutstat_inc(outif, type, code); 2108 2109 return; 2110 2111 bad: 2112 m_freem(m); 2113 return; 2114} 2115 2116static const char * 2117icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2118 struct in6_addr *tgt6) 2119{ 2120 static char buf[1024]; 2121 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2122 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2123 return buf; 2124} 2125 2126void 2127icmp6_redirect_input(struct mbuf *m, int off) 2128{ 2129 struct ifnet *ifp = m->m_pkthdr.rcvif; 2130 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2131 struct nd_redirect *nd_rd; 2132 int icmp6len = ntohs(ip6->ip6_plen); 2133 char *lladdr = NULL; 2134 int lladdrlen = 0; 2135 struct rtentry *rt = NULL; 2136 int is_router; 2137 int is_onlink; 2138 struct in6_addr src6 = ip6->ip6_src; 2139 struct in6_addr redtgt6; 2140 struct in6_addr reddst6; 2141 union nd_opts ndopts; 2142 2143 if (!ifp) 2144 return; 2145 2146 /* XXX if we are router, we don't update route by icmp6 redirect */ 2147 if (ip6_forwarding) 2148 goto freeit; 2149 if (!icmp6_rediraccept) 2150 goto freeit; 2151 2152 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2153 if (nd_rd == NULL) { 2154 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 2155 return; 2156 } 2157 redtgt6 = nd_rd->nd_rd_target; 2158 reddst6 = nd_rd->nd_rd_dst; 2159 2160 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2161 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2162 goto freeit; 2163 } 2164 2165 /* validation */ 2166 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2167 nd6log(LOG_ERR, 2168 "ICMP6 redirect sent from %s rejected; " 2169 "must be from linklocal\n", ip6_sprintf(&src6)); 2170 goto bad; 2171 } 2172 if (ip6->ip6_hlim != 255) { 2173 nd6log(LOG_ERR, 2174 "ICMP6 redirect sent from %s rejected; " 2175 "hlim=%d (must be 255)\n", 2176 ip6_sprintf(&src6), ip6->ip6_hlim); 2177 goto bad; 2178 } 2179 { 2180 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2181 struct sockaddr_in6 sin6; 2182 struct in6_addr *gw6; 2183 2184 sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0); 2185 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2186 if (rt) { 2187 if (rt->rt_gateway == NULL || 2188 rt->rt_gateway->sa_family != AF_INET6) { 2189 nd6log(LOG_ERR, 2190 "ICMP6 redirect rejected; no route " 2191 "with inet6 gateway found for redirect dst: %s\n", 2192 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2193 rtfree(rt); 2194 goto bad; 2195 } 2196 2197 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2198 if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2199 nd6log(LOG_ERR, 2200 "ICMP6 redirect rejected; " 2201 "not equal to gw-for-src=%s (must be same): %s\n", 2202 ip6_sprintf(gw6), 2203 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2204 rtfree(rt); 2205 goto bad; 2206 } 2207 } else { 2208 nd6log(LOG_ERR, "ICMP6 redirect rejected; " 2209 "no route found for redirect dst: %s\n", 2210 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2211 goto bad; 2212 } 2213 rtfree(rt); 2214 rt = NULL; 2215 } 2216 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2217 nd6log(LOG_ERR, "ICMP6 redirect rejected; " 2218 "redirect dst must be unicast: %s\n", 2219 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2220 goto bad; 2221 } 2222 2223 is_router = is_onlink = 0; 2224 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2225 is_router = 1; /* router case */ 2226 if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2227 is_onlink = 1; /* on-link destination case */ 2228 if (!is_router && !is_onlink) { 2229 nd6log(LOG_ERR, "ICMP6 redirect rejected; " 2230 "neither router case nor onlink case: %s\n", 2231 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2232 goto bad; 2233 } 2234 /* validation passed */ 2235 2236 icmp6len -= sizeof(*nd_rd); 2237 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2238 if (nd6_options(&ndopts) < 0) { 2239 nd6log(LOG_INFO, "invalid ND option, rejected: %s\n", 2240 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2241 /* nd6_options have incremented stats */ 2242 goto freeit; 2243 } 2244 2245 if (ndopts.nd_opts_tgt_lladdr) { 2246 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2247 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2248 } 2249 2250 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2251 nd6log(LOG_INFO, "lladdrlen mismatch for %s " 2252 "(if %d, icmp6 packet %d): %s\n", 2253 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2254 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2255 goto bad; 2256 } 2257 2258 /* RFC 2461 8.3 */ 2259 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2260 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2261 2262 if (!is_onlink) { /* better router case. perform rtredirect. */ 2263 /* perform rtredirect */ 2264 struct sockaddr_in6 sdst; 2265 struct sockaddr_in6 sgw; 2266 struct sockaddr_in6 ssrc; 2267 unsigned long rtcount; 2268 struct rtentry *newrt = NULL; 2269 2270 /* 2271 * do not install redirect route, if the number of entries 2272 * is too much (> hiwat). note that, the node (= host) will 2273 * work just fine even if we do not install redirect route 2274 * (there will be additional hops, though). 2275 */ 2276 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2277 if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes) 2278 goto freeit; 2279 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2280 return; 2281 else if (0 <= icmp6_redirect_lowat && 2282 rtcount > icmp6_redirect_lowat) { 2283 /* 2284 * XXX nuke a victim, install the new one. 2285 */ 2286 } 2287 2288 memset(&sdst, 0, sizeof(sdst)); 2289 memset(&sgw, 0, sizeof(sgw)); 2290 memset(&ssrc, 0, sizeof(ssrc)); 2291 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2292 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2293 sizeof(struct sockaddr_in6); 2294 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2295 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2296 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2297 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2298 NULL, RTF_GATEWAY | RTF_HOST, 2299 (struct sockaddr *)&ssrc, 2300 &newrt); 2301 2302 if (newrt) { 2303 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2304 icmp6_redirect_timeout_q); 2305 rtfree(newrt); 2306 } 2307 } 2308 /* finally update cached route in each socket via pfctlinput */ 2309 { 2310 struct sockaddr_in6 sdst; 2311 2312 sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0); 2313 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2314#if defined(IPSEC) 2315 if (ipsec_used) 2316 key_sa_routechange((struct sockaddr *)&sdst); 2317#endif 2318 } 2319 2320 freeit: 2321 m_freem(m); 2322 return; 2323 2324 bad: 2325 ICMP6_STATINC(ICMP6_STAT_BADREDIRECT); 2326 m_freem(m); 2327} 2328 2329void 2330icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2331{ 2332 struct ifnet *ifp; /* my outgoing interface */ 2333 struct in6_addr *ifp_ll6; 2334 struct in6_addr *nexthop; 2335 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2336 struct mbuf *m = NULL; /* newly allocated one */ 2337 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2338 struct nd_redirect *nd_rd; 2339 size_t maxlen; 2340 u_char *p; 2341 struct sockaddr_in6 src_sa; 2342 2343 icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0); 2344 2345 /* if we are not router, we don't send icmp6 redirect */ 2346 if (!ip6_forwarding) 2347 goto fail; 2348 2349 /* sanity check */ 2350 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2351 goto fail; 2352 2353 /* 2354 * Address check: 2355 * the source address must identify a neighbor, and 2356 * the destination address must not be a multicast address 2357 * [RFC 2461, sec 8.2] 2358 */ 2359 sip6 = mtod(m0, struct ip6_hdr *); 2360 sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0); 2361 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2362 goto fail; 2363 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2364 goto fail; /* what should we do here? */ 2365 2366 /* rate limit */ 2367 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2368 goto fail; 2369 2370 /* 2371 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2372 * we almost always ask for an mbuf cluster for simplicity. 2373 * (MHLEN < IPV6_MMTU is almost always true) 2374 */ 2375 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2376 if (m && IPV6_MMTU >= MHLEN) { 2377#if IPV6_MMTU >= MCLBYTES 2378 _MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT); 2379#else 2380 MCLGET(m, M_DONTWAIT); 2381#endif 2382 } 2383 2384 if (!m) 2385 goto fail; 2386 m->m_pkthdr.rcvif = NULL; 2387 m->m_len = 0; 2388 maxlen = M_TRAILINGSPACE(m); 2389 maxlen = min(IPV6_MMTU, maxlen); 2390 /* just for safety */ 2391 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2392 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2393 goto fail; 2394 } 2395 2396 { 2397 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2398 struct in6_ifaddr *ia; 2399 if ((ia = in6ifa_ifpforlinklocal(ifp, 2400 IN6_IFF_NOTREADY| 2401 IN6_IFF_ANYCAST)) == NULL) 2402 goto fail; 2403 ifp_ll6 = &ia->ia_addr.sin6_addr; 2404 } 2405 2406 /* get ip6 linklocal address for the router. */ 2407 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2408 struct sockaddr_in6 *sin6; 2409 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2410 nexthop = &sin6->sin6_addr; 2411 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2412 nexthop = NULL; 2413 } else 2414 nexthop = NULL; 2415 2416 /* ip6 */ 2417 ip6 = mtod(m, struct ip6_hdr *); 2418 ip6->ip6_flow = 0; 2419 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2420 ip6->ip6_vfc |= IPV6_VERSION; 2421 /* ip6->ip6_plen will be set later */ 2422 ip6->ip6_nxt = IPPROTO_ICMPV6; 2423 ip6->ip6_hlim = 255; 2424 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2425 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2426 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2427 2428 /* ND Redirect */ 2429 nd_rd = (struct nd_redirect *)(ip6 + 1); 2430 nd_rd->nd_rd_type = ND_REDIRECT; 2431 nd_rd->nd_rd_code = 0; 2432 nd_rd->nd_rd_reserved = 0; 2433 if (rt->rt_flags & RTF_GATEWAY) { 2434 /* 2435 * nd_rd->nd_rd_target must be a link-local address in 2436 * better router cases. 2437 */ 2438 if (!nexthop) 2439 goto fail; 2440 bcopy(nexthop, &nd_rd->nd_rd_target, 2441 sizeof(nd_rd->nd_rd_target)); 2442 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2443 sizeof(nd_rd->nd_rd_dst)); 2444 } else { 2445 /* make sure redtgt == reddst */ 2446 nexthop = &sip6->ip6_dst; 2447 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2448 sizeof(nd_rd->nd_rd_target)); 2449 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2450 sizeof(nd_rd->nd_rd_dst)); 2451 } 2452 2453 p = (u_char *)(nd_rd + 1); 2454 2455 { 2456 /* target lladdr option */ 2457 struct rtentry *rt_nexthop = NULL; 2458 int len; 2459 const struct sockaddr_dl *sdl; 2460 struct nd_opt_hdr *nd_opt; 2461 char *lladdr; 2462 2463 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2464 if (!rt_nexthop) 2465 goto nolladdropt; 2466 len = sizeof(*nd_opt) + ifp->if_addrlen; 2467 len = (len + 7) & ~7; /* round by 8 */ 2468 /* safety check */ 2469 if (len + (p - (u_char *)ip6) > maxlen) { 2470 rtfree(rt_nexthop); 2471 goto nolladdropt; 2472 } 2473 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2474 (rt_nexthop->rt_flags & RTF_LLINFO) && 2475 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2476 (sdl = satocsdl(rt_nexthop->rt_gateway)) && 2477 sdl->sdl_alen) { 2478 nd_opt = (struct nd_opt_hdr *)p; 2479 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2480 nd_opt->nd_opt_len = len >> 3; 2481 lladdr = (char *)(nd_opt + 1); 2482 memcpy(lladdr, CLLADDR(sdl), ifp->if_addrlen); 2483 p += len; 2484 } 2485 rtfree(rt_nexthop); 2486 } 2487 nolladdropt:; 2488 2489 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2490 2491 /* just to be safe */ 2492 if (m0->m_flags & M_DECRYPTED) 2493 goto noredhdropt; 2494 if (p - (u_char *)ip6 > maxlen) 2495 goto noredhdropt; 2496 2497 { 2498 /* redirected header option */ 2499 int len; 2500 struct nd_opt_rd_hdr *nd_opt_rh; 2501 2502 /* 2503 * compute the maximum size for icmp6 redirect header option. 2504 * XXX room for auth header? 2505 */ 2506 len = maxlen - (p - (u_char *)ip6); 2507 len &= ~7; 2508 2509 /* 2510 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2511 * about padding/truncate rule for the original IP packet. 2512 * From the discussion on IPv6imp in Feb 1999, 2513 * the consensus was: 2514 * - "attach as much as possible" is the goal 2515 * - pad if not aligned (original size can be guessed by 2516 * original ip6 header) 2517 * Following code adds the padding if it is simple enough, 2518 * and truncates if not. 2519 */ 2520 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2521 /* not enough room, truncate */ 2522 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2523 m0->m_pkthdr.len); 2524 } else { 2525 /* 2526 * enough room, truncate if not aligned. 2527 * we don't pad here for simplicity. 2528 */ 2529 size_t extra; 2530 2531 extra = m0->m_pkthdr.len % 8; 2532 if (extra) { 2533 /* truncate */ 2534 m_adj(m0, -extra); 2535 } 2536 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2537 } 2538 2539 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2540 memset(nd_opt_rh, 0, sizeof(*nd_opt_rh)); 2541 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2542 nd_opt_rh->nd_opt_rh_len = len >> 3; 2543 p += sizeof(*nd_opt_rh); 2544 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2545 2546 /* connect m0 to m */ 2547 m->m_pkthdr.len += m0->m_pkthdr.len; 2548 m_cat(m, m0); 2549 m0 = NULL; 2550 } 2551noredhdropt: 2552 if (m0) { 2553 m_freem(m0); 2554 m0 = NULL; 2555 } 2556 2557 /* XXX: clear embedded link IDs in the inner header */ 2558 in6_clearscope(&sip6->ip6_src); 2559 in6_clearscope(&sip6->ip6_dst); 2560 in6_clearscope(&nd_rd->nd_rd_target); 2561 in6_clearscope(&nd_rd->nd_rd_dst); 2562 2563 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2564 2565 nd_rd->nd_rd_cksum = 0; 2566 nd_rd->nd_rd_cksum 2567 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2568 2569 /* send the packet to outside... */ 2570 if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0) 2571 icmp6_ifstat_inc(ifp, ifs6_out_error); 2572 2573 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2574 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2575 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT); 2576 2577 return; 2578 2579fail: 2580 if (m) 2581 m_freem(m); 2582 if (m0) 2583 m_freem(m0); 2584} 2585 2586/* 2587 * ICMPv6 socket option processing. 2588 */ 2589int 2590icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 2591{ 2592 int error = 0; 2593 struct in6pcb *in6p = sotoin6pcb(so); 2594 2595 if (sopt->sopt_level != IPPROTO_ICMPV6) 2596 return rip6_ctloutput(op, so, sopt); 2597 2598 switch (op) { 2599 case PRCO_SETOPT: 2600 switch (sopt->sopt_name) { 2601 case ICMP6_FILTER: 2602 { 2603 struct icmp6_filter fil; 2604 2605 error = sockopt_get(sopt, &fil, sizeof(fil)); 2606 if (error) 2607 break; 2608 memcpy(in6p->in6p_icmp6filt, &fil, 2609 sizeof(struct icmp6_filter)); 2610 error = 0; 2611 break; 2612 } 2613 2614 default: 2615 error = ENOPROTOOPT; 2616 break; 2617 } 2618 break; 2619 2620 case PRCO_GETOPT: 2621 switch (sopt->sopt_name) { 2622 case ICMP6_FILTER: 2623 { 2624 if (in6p->in6p_icmp6filt == NULL) { 2625 error = EINVAL; 2626 break; 2627 } 2628 error = sockopt_set(sopt, in6p->in6p_icmp6filt, 2629 sizeof(struct icmp6_filter)); 2630 break; 2631 } 2632 2633 default: 2634 error = ENOPROTOOPT; 2635 break; 2636 } 2637 break; 2638 } 2639 2640 return (error); 2641} 2642 2643/* 2644 * Perform rate limit check. 2645 * Returns 0 if it is okay to send the icmp6 packet. 2646 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2647 * limitation. 2648 * 2649 * XXX per-destination/type check necessary? 2650 */ 2651static int 2652icmp6_ratelimit( 2653 const struct in6_addr *dst, /* not used at this moment */ 2654 const int type, /* not used at this moment */ 2655 const int code) /* not used at this moment */ 2656{ 2657 int ret; 2658 2659 ret = 0; /* okay to send */ 2660 2661 /* PPS limit */ 2662 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2663 icmp6errppslim)) { 2664 /* The packet is subject to rate limit */ 2665 ret++; 2666 } 2667 2668 return ret; 2669} 2670 2671static struct rtentry * 2672icmp6_mtudisc_clone(struct sockaddr *dst) 2673{ 2674 struct rtentry *rt; 2675 int error; 2676 2677 rt = rtalloc1(dst, 1); 2678 if (rt == 0) 2679 return NULL; 2680 2681 /* If we didn't get a host route, allocate one */ 2682 if ((rt->rt_flags & RTF_HOST) == 0) { 2683 struct rtentry *nrt; 2684 2685 error = rtrequest((int) RTM_ADD, dst, 2686 (struct sockaddr *) rt->rt_gateway, NULL, 2687 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2688 if (error) { 2689 rtfree(rt); 2690 return NULL; 2691 } 2692 nrt->rt_rmx = rt->rt_rmx; 2693 rtfree(rt); 2694 rt = nrt; 2695 } 2696 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2697 icmp6_mtudisc_timeout_q); 2698 if (error) { 2699 rtfree(rt); 2700 return NULL; 2701 } 2702 2703 return rt; /* caller need to call rtfree() */ 2704} 2705 2706static void 2707icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 2708{ 2709 2710 KASSERT(rt != NULL); 2711 rt_assert_referenced(rt); 2712 2713 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2714 (RTF_DYNAMIC | RTF_HOST)) { 2715 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2716 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL); 2717 } else { 2718 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2719 rt->rt_rmx.rmx_mtu = 0; 2720 } 2721} 2722 2723static void 2724icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r) 2725{ 2726 2727 KASSERT(rt != NULL); 2728 rt_assert_referenced(rt); 2729 2730 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2731 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2732 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2733 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL); 2734 } 2735} 2736 2737/* 2738 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes. silly? 2739 */ 2740static int 2741sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS) 2742{ 2743 (void)&name; 2744 (void)&l; 2745 (void)&oname; 2746 2747 if (namelen != 0) 2748 return (EINVAL); 2749 2750 return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp, 2751 /*XXXUNCONST*/ 2752 __UNCONST(newp), newlen)); 2753} 2754 2755static int 2756sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS) 2757{ 2758 2759 return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS)); 2760} 2761 2762static int 2763sysctl_net_inet6_icmp6_redirtimeout(SYSCTLFN_ARGS) 2764{ 2765 int error, tmp; 2766 struct sysctlnode node; 2767 2768 node = *rnode; 2769 node.sysctl_data = &tmp; 2770 tmp = icmp6_redirtimeout; 2771 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2772 if (error || newp == NULL) 2773 return error; 2774 if (tmp < 0) 2775 return EINVAL; 2776 icmp6_redirtimeout = tmp; 2777 2778 if (icmp6_redirect_timeout_q != NULL) { 2779 if (icmp6_redirtimeout == 0) { 2780 rt_timer_queue_destroy(icmp6_redirect_timeout_q, 2781 true); 2782 } else { 2783 rt_timer_queue_change(icmp6_redirect_timeout_q, 2784 icmp6_redirtimeout); 2785 } 2786 } else if (icmp6_redirtimeout > 0) { 2787 icmp6_redirect_timeout_q = 2788 rt_timer_queue_create(icmp6_redirtimeout); 2789 } 2790 2791 return 0; 2792} 2793 2794static void 2795sysctl_net_inet6_icmp6_setup(struct sysctllog **clog) 2796{ 2797 extern int nd6_maxqueuelen; /* defined in nd6.c */ 2798 2799 sysctl_createv(clog, 0, NULL, NULL, 2800 CTLFLAG_PERMANENT, 2801 CTLTYPE_NODE, "inet6", NULL, 2802 NULL, 0, NULL, 0, 2803 CTL_NET, PF_INET6, CTL_EOL); 2804 sysctl_createv(clog, 0, NULL, NULL, 2805 CTLFLAG_PERMANENT, 2806 CTLTYPE_NODE, "icmp6", 2807 SYSCTL_DESCR("ICMPv6 related settings"), 2808 NULL, 0, NULL, 0, 2809 CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL); 2810 2811 sysctl_createv(clog, 0, NULL, NULL, 2812 CTLFLAG_PERMANENT, 2813 CTLTYPE_STRUCT, "stats", 2814 SYSCTL_DESCR("ICMPv6 transmission statistics"), 2815 sysctl_net_inet6_icmp6_stats, 0, NULL, 0, 2816 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2817 ICMPV6CTL_STATS, CTL_EOL); 2818 sysctl_createv(clog, 0, NULL, NULL, 2819 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2820 CTLTYPE_INT, "rediraccept", 2821 SYSCTL_DESCR("Accept and process redirect messages"), 2822 NULL, 0, &icmp6_rediraccept, 0, 2823 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2824 ICMPV6CTL_REDIRACCEPT, CTL_EOL); 2825 sysctl_createv(clog, 0, NULL, NULL, 2826 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2827 CTLTYPE_INT, "redirtimeout", 2828 SYSCTL_DESCR("Redirect generated route lifetime"), 2829 sysctl_net_inet6_icmp6_redirtimeout, 0, 2830 &icmp6_redirtimeout, 0, 2831 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2832 ICMPV6CTL_REDIRTIMEOUT, CTL_EOL); 2833#if 0 /* obsoleted */ 2834 sysctl_createv(clog, 0, NULL, NULL, 2835 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2836 CTLTYPE_INT, "errratelimit", NULL, 2837 NULL, 0, &icmp6_errratelimit, 0, 2838 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2839 ICMPV6CTL_ERRRATELIMIT, CTL_EOL); 2840#endif 2841 sysctl_createv(clog, 0, NULL, NULL, 2842 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2843 CTLTYPE_INT, "nd6_prune", 2844 SYSCTL_DESCR("Neighbor discovery prune interval"), 2845 NULL, 0, &nd6_prune, 0, 2846 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2847 ICMPV6CTL_ND6_PRUNE, CTL_EOL); 2848 sysctl_createv(clog, 0, NULL, NULL, 2849 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2850 CTLTYPE_INT, "nd6_delay", 2851 SYSCTL_DESCR("First probe delay time"), 2852 NULL, 0, &nd6_delay, 0, 2853 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2854 ICMPV6CTL_ND6_DELAY, CTL_EOL); 2855 sysctl_createv(clog, 0, NULL, NULL, 2856 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2857 CTLTYPE_INT, "nd6_umaxtries", 2858 SYSCTL_DESCR("Number of unicast discovery attempts"), 2859 NULL, 0, &nd6_umaxtries, 0, 2860 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2861 ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL); 2862 sysctl_createv(clog, 0, NULL, NULL, 2863 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2864 CTLTYPE_INT, "nd6_mmaxtries", 2865 SYSCTL_DESCR("Number of multicast discovery attempts"), 2866 NULL, 0, &nd6_mmaxtries, 0, 2867 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2868 ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL); 2869 sysctl_createv(clog, 0, NULL, NULL, 2870 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2871 CTLTYPE_INT, "nd6_useloopback", 2872 SYSCTL_DESCR("Use loopback interface for local traffic"), 2873 NULL, 0, &nd6_useloopback, 0, 2874 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2875 ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL); 2876#if 0 /* obsoleted */ 2877 sysctl_createv(clog, 0, NULL, NULL, 2878 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2879 CTLTYPE_INT, "nd6_proxyall", NULL, 2880 NULL, 0, &nd6_proxyall, 0, 2881 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2882 ICMPV6CTL_ND6_PROXYALL, CTL_EOL); 2883#endif 2884 sysctl_createv(clog, 0, NULL, NULL, 2885 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2886 CTLTYPE_INT, "nodeinfo", 2887 SYSCTL_DESCR("Respond to node information requests"), 2888 NULL, 0, &icmp6_nodeinfo, 0, 2889 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2890 ICMPV6CTL_NODEINFO, CTL_EOL); 2891 sysctl_createv(clog, 0, NULL, NULL, 2892 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2893 CTLTYPE_INT, "errppslimit", 2894 SYSCTL_DESCR("Maximum ICMP errors sent per second"), 2895 NULL, 0, &icmp6errppslim, 0, 2896 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2897 ICMPV6CTL_ERRPPSLIMIT, CTL_EOL); 2898 sysctl_createv(clog, 0, NULL, NULL, 2899 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2900 CTLTYPE_INT, "nd6_maxnudhint", 2901 SYSCTL_DESCR("Maximum neighbor unreachable hint count"), 2902 NULL, 0, &nd6_maxnudhint, 0, 2903 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2904 ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL); 2905 sysctl_createv(clog, 0, NULL, NULL, 2906 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2907 CTLTYPE_INT, "mtudisc_hiwat", 2908 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2909 NULL, 0, &icmp6_mtudisc_hiwat, 0, 2910 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2911 ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL); 2912 sysctl_createv(clog, 0, NULL, NULL, 2913 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2914 CTLTYPE_INT, "mtudisc_lowat", 2915 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2916 NULL, 0, &icmp6_mtudisc_lowat, 0, 2917 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2918 ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL); 2919 sysctl_createv(clog, 0, NULL, NULL, 2920 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2921 CTLTYPE_INT, "nd6_debug", 2922 SYSCTL_DESCR("Enable neighbor discovery debug output"), 2923 NULL, 0, &nd6_debug, 0, 2924 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2925 ICMPV6CTL_ND6_DEBUG, CTL_EOL); 2926 sysctl_createv(clog, 0, NULL, NULL, 2927 CTLFLAG_PERMANENT, 2928 CTLTYPE_STRUCT, "nd6_drlist", 2929 SYSCTL_DESCR("Default router list"), 2930 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2931 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2932 ICMPV6CTL_ND6_DRLIST, CTL_EOL); 2933 sysctl_createv(clog, 0, NULL, NULL, 2934 CTLFLAG_PERMANENT, 2935 CTLTYPE_STRUCT, "nd6_prlist", 2936 SYSCTL_DESCR("Prefix list"), 2937 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2938 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2939 ICMPV6CTL_ND6_PRLIST, CTL_EOL); 2940 sysctl_createv(clog, 0, NULL, NULL, 2941 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2942 CTLTYPE_INT, "maxqueuelen", 2943 SYSCTL_DESCR("max packet queue len for a unresolved ND"), 2944 NULL, 1, &nd6_maxqueuelen, 0, 2945 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2946 ICMPV6CTL_ND6_MAXQLEN, CTL_EOL); 2947} 2948 2949void 2950icmp6_statinc(u_int stat) 2951{ 2952 2953 KASSERT(stat < ICMP6_NSTATS); 2954 ICMP6_STATINC(stat); 2955} 2956