icmp6.c revision 1.89
1/* $NetBSD: icmp6.c,v 1.89 2003/03/31 23:55:46 itojun 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. All advertising materials mentioning features or use of this software 46 * must display the following acknowledgement: 47 * This product includes software developed by the University of 48 * California, Berkeley and its contributors. 49 * 4. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 66 */ 67 68#include <sys/cdefs.h> 69__KERNEL_RCSID(0, "$NetBSD: icmp6.c,v 1.89 2003/03/31 23:55:46 itojun Exp $"); 70 71#include "opt_inet.h" 72#include "opt_ipsec.h" 73 74#include <sys/param.h> 75#include <sys/systm.h> 76#include <sys/malloc.h> 77#include <sys/mbuf.h> 78#include <sys/protosw.h> 79#include <sys/socket.h> 80#include <sys/socketvar.h> 81#include <sys/time.h> 82#include <sys/kernel.h> 83#include <sys/syslog.h> 84#include <sys/domain.h> 85#include <sys/sysctl.h> 86 87#include <net/if.h> 88#include <net/route.h> 89#include <net/if_dl.h> 90#include <net/if_types.h> 91 92#include <netinet/in.h> 93#include <netinet/in_var.h> 94#include <netinet/ip6.h> 95#include <netinet6/ip6_var.h> 96#include <netinet/icmp6.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> 102 103#ifdef IPSEC 104#include <netinet6/ipsec.h> 105#include <netkey/key.h> 106#endif 107 108#include "faith.h" 109#if defined(NFAITH) && 0 < NFAITH 110#include <net/if_faith.h> 111#endif 112 113#include <net/net_osdep.h> 114 115extern struct domain inet6domain; 116 117struct icmp6stat icmp6stat; 118 119extern struct in6pcb rawin6pcb; 120extern int icmp6errppslim; 121static int icmp6errpps_count = 0; 122static struct timeval icmp6errppslim_last; 123extern int icmp6_nodeinfo; 124 125/* 126 * List of callbacks to notify when Path MTU changes are made. 127 */ 128struct icmp6_mtudisc_callback { 129 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 130 void (*mc_func) __P((struct in6_addr *)); 131}; 132 133LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 134 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 135 136static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 137extern int pmtu_expire; 138 139/* XXX do these values make any sense? */ 140static int icmp6_mtudisc_hiwat = 1280; 141static int icmp6_mtudisc_lowat = 256; 142 143/* 144 * keep track of # of redirect routes. 145 */ 146static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 147 148/* XXX experimental, turned off */ 149static int icmp6_redirect_hiwat = -1; 150static int icmp6_redirect_lowat = -1; 151 152static void icmp6_errcount __P((struct icmp6errstat *, int, int)); 153static int icmp6_rip6_input __P((struct mbuf **, int)); 154static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int)); 155static const char *icmp6_redirect_diag __P((struct in6_addr *, 156 struct in6_addr *, struct in6_addr *)); 157static struct mbuf *ni6_input __P((struct mbuf *, int)); 158static struct mbuf *ni6_nametodns __P((const char *, int, int)); 159static int ni6_dnsmatch __P((const char *, int, const char *, int)); 160static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *, 161 struct ifnet **, char *)); 162static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 163 struct ifnet *, int)); 164static int icmp6_notify_error __P((struct mbuf *, int, int, int)); 165static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *)); 166static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *)); 167static void icmp6_redirect_timeout __P((struct rtentry *, struct rttimer *)); 168 169void 170icmp6_init() 171{ 172 mld6_init(); 173 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 174 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 175} 176 177static void 178icmp6_errcount(stat, type, code) 179 struct icmp6errstat *stat; 180 int type, code; 181{ 182 switch (type) { 183 case ICMP6_DST_UNREACH: 184 switch (code) { 185 case ICMP6_DST_UNREACH_NOROUTE: 186 stat->icp6errs_dst_unreach_noroute++; 187 return; 188 case ICMP6_DST_UNREACH_ADMIN: 189 stat->icp6errs_dst_unreach_admin++; 190 return; 191 case ICMP6_DST_UNREACH_BEYONDSCOPE: 192 stat->icp6errs_dst_unreach_beyondscope++; 193 return; 194 case ICMP6_DST_UNREACH_ADDR: 195 stat->icp6errs_dst_unreach_addr++; 196 return; 197 case ICMP6_DST_UNREACH_NOPORT: 198 stat->icp6errs_dst_unreach_noport++; 199 return; 200 } 201 break; 202 case ICMP6_PACKET_TOO_BIG: 203 stat->icp6errs_packet_too_big++; 204 return; 205 case ICMP6_TIME_EXCEEDED: 206 switch (code) { 207 case ICMP6_TIME_EXCEED_TRANSIT: 208 stat->icp6errs_time_exceed_transit++; 209 return; 210 case ICMP6_TIME_EXCEED_REASSEMBLY: 211 stat->icp6errs_time_exceed_reassembly++; 212 return; 213 } 214 break; 215 case ICMP6_PARAM_PROB: 216 switch (code) { 217 case ICMP6_PARAMPROB_HEADER: 218 stat->icp6errs_paramprob_header++; 219 return; 220 case ICMP6_PARAMPROB_NEXTHEADER: 221 stat->icp6errs_paramprob_nextheader++; 222 return; 223 case ICMP6_PARAMPROB_OPTION: 224 stat->icp6errs_paramprob_option++; 225 return; 226 } 227 break; 228 case ND_REDIRECT: 229 stat->icp6errs_redirect++; 230 return; 231 } 232 stat->icp6errs_unknown++; 233} 234 235/* 236 * Register a Path MTU Discovery callback. 237 */ 238void 239icmp6_mtudisc_callback_register(func) 240 void (*func) __P((struct in6_addr *)); 241{ 242 struct icmp6_mtudisc_callback *mc; 243 244 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 245 mc = LIST_NEXT(mc, mc_list)) { 246 if (mc->mc_func == func) 247 return; 248 } 249 250 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 251 if (mc == NULL) 252 panic("icmp6_mtudisc_callback_register"); 253 254 mc->mc_func = func; 255 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 256} 257 258/* 259 * Generate an error packet of type error in response to bad IP6 packet. 260 */ 261void 262icmp6_error(m, type, code, param) 263 struct mbuf *m; 264 int type, code, param; 265{ 266 struct ip6_hdr *oip6, *nip6; 267 struct icmp6_hdr *icmp6; 268 u_int preplen; 269 int off; 270 int nxt; 271 272 icmp6stat.icp6s_error++; 273 274 /* count per-type-code statistics */ 275 icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code); 276 277 if (m->m_flags & M_DECRYPTED) { 278 icmp6stat.icp6s_canterror++; 279 goto freeit; 280 } 281 282#ifndef PULLDOWN_TEST 283 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), ); 284#else 285 if (m->m_len < sizeof(struct ip6_hdr)) { 286 m = m_pullup(m, sizeof(struct ip6_hdr)); 287 if (m == NULL) 288 return; 289 } 290#endif 291 oip6 = mtod(m, struct ip6_hdr *); 292 293 /* 294 * If the destination address of the erroneous packet is a multicast 295 * address, or the packet was sent using link-layer multicast, 296 * we should basically suppress sending an error (RFC 2463, Section 297 * 2.4). 298 * We have two exceptions (the item e.2 in that section): 299 * - the Pakcet Too Big message can be sent for path MTU discovery. 300 * - the Parameter Problem Message that can be allowed an icmp6 error 301 * in the option type field. This check has been done in 302 * ip6_unknown_opt(), so we can just check the type and code. 303 */ 304 if ((m->m_flags & (M_BCAST|M_MCAST) || 305 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 306 (type != ICMP6_PACKET_TOO_BIG && 307 (type != ICMP6_PARAM_PROB || 308 code != ICMP6_PARAMPROB_OPTION))) 309 goto freeit; 310 311 /* 312 * RFC 2463, 2.4 (e.5): source address check. 313 * XXX: the case of anycast source? 314 */ 315 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 316 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 317 goto freeit; 318 319 /* 320 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 321 * don't do it. 322 */ 323 nxt = -1; 324 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 325 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 326 struct icmp6_hdr *icp; 327 328#ifndef PULLDOWN_TEST 329 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), ); 330 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 331#else 332 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 333 sizeof(*icp)); 334 if (icp == NULL) { 335 icmp6stat.icp6s_tooshort++; 336 return; 337 } 338#endif 339 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 340 icp->icmp6_type == ND_REDIRECT) { 341 /* 342 * ICMPv6 error 343 * Special case: for redirect (which is 344 * informational) we must not send icmp6 error. 345 */ 346 icmp6stat.icp6s_canterror++; 347 goto freeit; 348 } else { 349 /* ICMPv6 informational - send the error */ 350 } 351 } 352#if 0 /* controversial */ 353 else if (off >= 0 && nxt == IPPROTO_ESP) { 354 /* 355 * It could be ICMPv6 error inside ESP. Take a safer side, 356 * don't respond. 357 */ 358 icmp6stat.icp6s_canterror++; 359 goto freeit; 360 } 361#endif 362 else { 363 /* non-ICMPv6 - send the error */ 364 } 365 366 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 367 368 /* Finally, do rate limitation check. */ 369 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 370 icmp6stat.icp6s_toofreq++; 371 goto freeit; 372 } 373 374 /* 375 * OK, ICMP6 can be generated. 376 */ 377 378 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 379 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 380 381 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 382 M_PREPEND(m, preplen, M_DONTWAIT); 383 if (m && m->m_len < preplen) 384 m = m_pullup(m, preplen); 385 if (m == NULL) { 386 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 387 return; 388 } 389 390 nip6 = mtod(m, struct ip6_hdr *); 391 nip6->ip6_src = oip6->ip6_src; 392 nip6->ip6_dst = oip6->ip6_dst; 393 394 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src)) 395 oip6->ip6_src.s6_addr16[1] = 0; 396 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst)) 397 oip6->ip6_dst.s6_addr16[1] = 0; 398 399 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 400 icmp6->icmp6_type = type; 401 icmp6->icmp6_code = code; 402 icmp6->icmp6_pptr = htonl((u_int32_t)param); 403 404 /* 405 * icmp6_reflect() is designed to be in the input path. 406 * icmp6_error() can be called from both input and outut path, 407 * and if we are in output path rcvif could contain bogus value. 408 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 409 * information in ip header (nip6). 410 */ 411 m->m_pkthdr.rcvif = NULL; 412 413 icmp6stat.icp6s_outhist[type]++; 414 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 415 416 return; 417 418 freeit: 419 /* 420 * If we can't tell wheter or not we can generate ICMP6, free it. 421 */ 422 m_freem(m); 423} 424 425/* 426 * Process a received ICMP6 message. 427 */ 428int 429icmp6_input(mp, offp, proto) 430 struct mbuf **mp; 431 int *offp, proto; 432{ 433 struct mbuf *m = *mp, *n; 434 struct ip6_hdr *ip6, *nip6; 435 struct icmp6_hdr *icmp6, *nicmp6; 436 int off = *offp; 437 int icmp6len = m->m_pkthdr.len - *offp; 438 int code, sum, noff; 439 440 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 441 442#ifndef PULLDOWN_TEST 443 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE); 444 /* m might change if M_LOOP. So, call mtod after this */ 445#endif 446 447 /* 448 * Locate icmp6 structure in mbuf, and check 449 * that not corrupted and of at least minimum length 450 */ 451 452 ip6 = mtod(m, struct ip6_hdr *); 453 if (icmp6len < sizeof(struct icmp6_hdr)) { 454 icmp6stat.icp6s_tooshort++; 455 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 456 goto freeit; 457 } 458 459 /* 460 * calculate the checksum 461 */ 462#ifndef PULLDOWN_TEST 463 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 464#else 465 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 466 if (icmp6 == NULL) { 467 icmp6stat.icp6s_tooshort++; 468 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 469 return IPPROTO_DONE; 470 } 471#endif 472 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 473 code = icmp6->icmp6_code; 474 475 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 476 nd6log((LOG_ERR, 477 "ICMP6 checksum error(%d|%x) %s\n", 478 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 479 icmp6stat.icp6s_checksum++; 480 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 481 goto freeit; 482 } 483 484#if defined(NFAITH) && 0 < NFAITH 485 if (faithprefix(&ip6->ip6_dst)) { 486 /* 487 * Deliver very specific ICMP6 type only. 488 * This is important to deilver TOOBIG. Otherwise PMTUD 489 * will not work. 490 */ 491 switch (icmp6->icmp6_type) { 492 case ICMP6_DST_UNREACH: 493 case ICMP6_PACKET_TOO_BIG: 494 case ICMP6_TIME_EXCEEDED: 495 break; 496 default: 497 goto freeit; 498 } 499 } 500#endif 501 502 icmp6stat.icp6s_inhist[icmp6->icmp6_type]++; 503 504 switch (icmp6->icmp6_type) { 505 case ICMP6_DST_UNREACH: 506 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 507 switch (code) { 508 case ICMP6_DST_UNREACH_NOROUTE: 509 code = PRC_UNREACH_NET; 510 break; 511 case ICMP6_DST_UNREACH_ADMIN: 512 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 513 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 514 break; 515 case ICMP6_DST_UNREACH_ADDR: 516 code = PRC_HOSTDEAD; 517 break; 518#ifdef COMPAT_RFC1885 519 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 520 code = PRC_UNREACH_SRCFAIL; 521 break; 522#else 523 case ICMP6_DST_UNREACH_BEYONDSCOPE: 524 /* I mean "source address was incorrect." */ 525 code = PRC_UNREACH_NET; 526 break; 527#endif 528 case ICMP6_DST_UNREACH_NOPORT: 529 code = PRC_UNREACH_PORT; 530 break; 531 default: 532 goto badcode; 533 } 534 goto deliver; 535 536 case ICMP6_PACKET_TOO_BIG: 537 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 538 if (code != 0) 539 goto badcode; 540 541 code = PRC_MSGSIZE; 542 543 /* 544 * Updating the path MTU will be done after examining 545 * intermediate extension headers. 546 */ 547 goto deliver; 548 549 case ICMP6_TIME_EXCEEDED: 550 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 551 switch (code) { 552 case ICMP6_TIME_EXCEED_TRANSIT: 553 code = PRC_TIMXCEED_INTRANS; 554 break; 555 case ICMP6_TIME_EXCEED_REASSEMBLY: 556 code = PRC_TIMXCEED_REASS; 557 break; 558 default: 559 goto badcode; 560 } 561 goto deliver; 562 563 case ICMP6_PARAM_PROB: 564 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 565 switch (code) { 566 case ICMP6_PARAMPROB_NEXTHEADER: 567 code = PRC_UNREACH_PROTOCOL; 568 break; 569 case ICMP6_PARAMPROB_HEADER: 570 case ICMP6_PARAMPROB_OPTION: 571 code = PRC_PARAMPROB; 572 break; 573 default: 574 goto badcode; 575 } 576 goto deliver; 577 578 case ICMP6_ECHO_REQUEST: 579 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 580 if (code != 0) 581 goto badcode; 582 /* 583 * Copy mbuf to send to two data paths: userland socket(s), 584 * and to the querier (echo reply). 585 * m: a copy for socket, n: a copy for querier 586 */ 587 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 588 /* Give up local */ 589 n = m; 590 m = NULL; 591 goto deliverecho; 592 } 593 /* 594 * If the first mbuf is shared, or the first mbuf is too short, 595 * copy the first part of the data into a fresh mbuf. 596 * Otherwise, we will wrongly overwrite both copies. 597 */ 598 if ((n->m_flags & M_EXT) != 0 || 599 n->m_len < off + sizeof(struct icmp6_hdr)) { 600 struct mbuf *n0 = n; 601 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6); 602 603 /* 604 * Prepare an internal mbuf. m_pullup() doesn't 605 * always copy the length we specified. 606 */ 607 if (maxlen >= MCLBYTES) { 608 /* Give up remote */ 609 m_freem(n0); 610 break; 611 } 612 MGETHDR(n, M_DONTWAIT, n0->m_type); 613 if (n && maxlen >= MHLEN) { 614 MCLGET(n, M_DONTWAIT); 615 if ((n->m_flags & M_EXT) == 0) { 616 m_free(n); 617 n = NULL; 618 } 619 } 620 if (n == NULL) { 621 /* Give up local */ 622 m_freem(n0); 623 n = m; 624 m = NULL; 625 goto deliverecho; 626 } 627 M_COPY_PKTHDR(n, n0); 628 /* 629 * Copy IPv6 and ICMPv6 only. 630 */ 631 nip6 = mtod(n, struct ip6_hdr *); 632 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 633 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 634 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 635 noff = sizeof(struct ip6_hdr); 636 n->m_len = noff + sizeof(struct icmp6_hdr); 637 /* 638 * Adjust mbuf. ip6_plen will be adjusted in 639 * ip6_output(). 640 * n->m_pkthdr.len == n0->m_pkthdr.len at this point. 641 */ 642 n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr); 643 n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr)); 644 m_adj(n0, off + sizeof(struct icmp6_hdr)); 645 n->m_next = n0; 646 n0->m_flags &= ~M_PKTHDR; 647 } else { 648 deliverecho: 649 nip6 = mtod(n, struct ip6_hdr *); 650 nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off); 651 noff = off; 652 } 653 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 654 nicmp6->icmp6_code = 0; 655 if (n) { 656 icmp6stat.icp6s_reflect++; 657 icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++; 658 icmp6_reflect(n, noff); 659 } 660 if (!m) 661 goto freeit; 662 break; 663 664 case ICMP6_ECHO_REPLY: 665 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 666 if (code != 0) 667 goto badcode; 668 break; 669 670 case MLD6_LISTENER_QUERY: 671 case MLD6_LISTENER_REPORT: 672 if (icmp6len < sizeof(struct mld6_hdr)) 673 goto badlen; 674 if (icmp6->icmp6_type == MLD6_LISTENER_QUERY) /* XXX: ugly... */ 675 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 676 else 677 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 678 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 679 /* give up local */ 680 mld6_input(m, off); 681 m = NULL; 682 goto freeit; 683 } 684 mld6_input(n, off); 685 /* m stays. */ 686 break; 687 688 case MLD6_LISTENER_DONE: 689 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 690 if (icmp6len < sizeof(struct mld6_hdr)) /* necessary? */ 691 goto badlen; 692 break; /* nothing to be done in kernel */ 693 694 case MLD6_MTRACE_RESP: 695 case MLD6_MTRACE: 696 /* XXX: these two are experimental. not officially defined. */ 697 /* XXX: per-interface statistics? */ 698 break; /* just pass it to applications */ 699 700 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 701 { 702 enum { WRU, FQDN } mode; 703 704 if (!icmp6_nodeinfo) 705 break; 706 707 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 708 mode = WRU; 709 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 710 mode = FQDN; 711 else 712 goto badlen; 713 714 if (mode == FQDN) { 715#ifndef PULLDOWN_TEST 716 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo), 717 IPPROTO_DONE); 718#endif 719 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 720 if (n) 721 n = ni6_input(n, off); 722 /* XXX meaningless if n == NULL */ 723 noff = sizeof(struct ip6_hdr); 724 } else { 725 u_char *p; 726 int maxlen, maxhlen; 727 728 if ((icmp6_nodeinfo & 5) != 5) 729 break; 730 731 if (code != 0) 732 goto badcode; 733 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; 734 if (maxlen >= MCLBYTES) { 735 /* Give up remote */ 736 break; 737 } 738 MGETHDR(n, M_DONTWAIT, m->m_type); 739 if (n && maxlen > MHLEN) { 740 MCLGET(n, M_DONTWAIT); 741 if ((n->m_flags & M_EXT) == 0) { 742 m_free(n); 743 n = NULL; 744 } 745 } 746 if (n == NULL) { 747 /* Give up remote */ 748 break; 749 } 750 n->m_pkthdr.rcvif = NULL; 751 n->m_len = 0; 752 maxhlen = M_TRAILINGSPACE(n) - maxlen; 753 if (maxhlen > hostnamelen) 754 maxhlen = hostnamelen; 755 /* 756 * Copy IPv6 and ICMPv6 only. 757 */ 758 nip6 = mtod(n, struct ip6_hdr *); 759 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 760 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 761 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 762 p = (u_char *)(nicmp6 + 1); 763 bzero(p, 4); 764 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 765 noff = sizeof(struct ip6_hdr); 766 M_COPY_PKTHDR(n, m); /* just for rcvif */ 767 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 768 sizeof(struct icmp6_hdr) + 4 + maxhlen; 769 nicmp6->icmp6_type = ICMP6_WRUREPLY; 770 nicmp6->icmp6_code = 0; 771 } 772#undef hostnamelen 773 if (n) { 774 icmp6stat.icp6s_reflect++; 775 icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++; 776 icmp6_reflect(n, noff); 777 } 778 break; 779 } 780 781 case ICMP6_WRUREPLY: 782 if (code != 0) 783 goto badcode; 784 break; 785 786 case ND_ROUTER_SOLICIT: 787 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 788 if (code != 0) 789 goto badcode; 790 if (icmp6len < sizeof(struct nd_router_solicit)) 791 goto badlen; 792 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 793 /* give up local */ 794 nd6_rs_input(m, off, icmp6len); 795 m = NULL; 796 goto freeit; 797 } 798 nd6_rs_input(n, off, icmp6len); 799 /* m stays. */ 800 break; 801 802 case ND_ROUTER_ADVERT: 803 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 804 if (code != 0) 805 goto badcode; 806 if (icmp6len < sizeof(struct nd_router_advert)) 807 goto badlen; 808 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 809 /* give up local */ 810 nd6_ra_input(m, off, icmp6len); 811 m = NULL; 812 goto freeit; 813 } 814 nd6_ra_input(n, off, icmp6len); 815 /* m stays. */ 816 break; 817 818 case ND_NEIGHBOR_SOLICIT: 819 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 820 if (code != 0) 821 goto badcode; 822 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 823 goto badlen; 824 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 825 /* give up local */ 826 nd6_ns_input(m, off, icmp6len); 827 m = NULL; 828 goto freeit; 829 } 830 nd6_ns_input(n, off, icmp6len); 831 /* m stays. */ 832 break; 833 834 case ND_NEIGHBOR_ADVERT: 835 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 836 if (code != 0) 837 goto badcode; 838 if (icmp6len < sizeof(struct nd_neighbor_advert)) 839 goto badlen; 840 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 841 /* give up local */ 842 nd6_na_input(m, off, icmp6len); 843 m = NULL; 844 goto freeit; 845 } 846 nd6_na_input(n, off, icmp6len); 847 /* m stays. */ 848 break; 849 850 case ND_REDIRECT: 851 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 852 if (code != 0) 853 goto badcode; 854 if (icmp6len < sizeof(struct nd_redirect)) 855 goto badlen; 856 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 857 /* give up local */ 858 icmp6_redirect_input(m, off); 859 m = NULL; 860 goto freeit; 861 } 862 icmp6_redirect_input(n, off); 863 /* m stays. */ 864 break; 865 866 case ICMP6_ROUTER_RENUMBERING: 867 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 868 code != ICMP6_ROUTER_RENUMBERING_RESULT) 869 goto badcode; 870 if (icmp6len < sizeof(struct icmp6_router_renum)) 871 goto badlen; 872 break; 873 874 default: 875 nd6log((LOG_DEBUG, 876 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 877 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 878 ip6_sprintf(&ip6->ip6_dst), 879 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 880 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 881 /* ICMPv6 error: MUST deliver it by spec... */ 882 code = PRC_NCMDS; 883 /* deliver */ 884 } else { 885 /* ICMPv6 informational: MUST not deliver */ 886 break; 887 } 888 deliver: 889 if (icmp6_notify_error(m, off, icmp6len, code)) { 890 /* In this case, m should've been freed. */ 891 return (IPPROTO_DONE); 892 } 893 break; 894 895 badcode: 896 icmp6stat.icp6s_badcode++; 897 break; 898 899 badlen: 900 icmp6stat.icp6s_badlen++; 901 break; 902 } 903 904 /* deliver the packet to appropriate sockets */ 905 icmp6_rip6_input(&m, *offp); 906 907 return IPPROTO_DONE; 908 909 freeit: 910 m_freem(m); 911 return IPPROTO_DONE; 912} 913 914static int 915icmp6_notify_error(m, off, icmp6len, code) 916 struct mbuf *m; 917 int off, icmp6len; 918{ 919 struct icmp6_hdr *icmp6; 920 struct ip6_hdr *eip6; 921 u_int32_t notifymtu; 922 struct sockaddr_in6 icmp6src, icmp6dst; 923 924 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 925 icmp6stat.icp6s_tooshort++; 926 goto freeit; 927 } 928#ifndef PULLDOWN_TEST 929 IP6_EXTHDR_CHECK(m, off, 930 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), 931 -1); 932 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 933#else 934 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 935 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 936 if (icmp6 == NULL) { 937 icmp6stat.icp6s_tooshort++; 938 return (-1); 939 } 940#endif 941 eip6 = (struct ip6_hdr *)(icmp6 + 1); 942 943 /* Detect the upper level protocol */ 944 { 945 void (*ctlfunc) __P((int, struct sockaddr *, void *)); 946 u_int8_t nxt = eip6->ip6_nxt; 947 int eoff = off + sizeof(struct icmp6_hdr) + 948 sizeof(struct ip6_hdr); 949 struct ip6ctlparam ip6cp; 950 struct in6_addr *finaldst = NULL; 951 int icmp6type = icmp6->icmp6_type; 952 struct ip6_frag *fh; 953 struct ip6_rthdr *rth; 954 struct ip6_rthdr0 *rth0; 955 int rthlen; 956 957 while (1) { /* XXX: should avoid infinite loop explicitly? */ 958 struct ip6_ext *eh; 959 960 switch (nxt) { 961 case IPPROTO_HOPOPTS: 962 case IPPROTO_DSTOPTS: 963 case IPPROTO_AH: 964#ifndef PULLDOWN_TEST 965 IP6_EXTHDR_CHECK(m, 0, eoff + 966 sizeof(struct ip6_ext), 967 -1); 968 eh = (struct ip6_ext *)(mtod(m, caddr_t) 969 + eoff); 970#else 971 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 972 eoff, sizeof(*eh)); 973 if (eh == NULL) { 974 icmp6stat.icp6s_tooshort++; 975 return (-1); 976 } 977#endif 978 979 if (nxt == IPPROTO_AH) 980 eoff += (eh->ip6e_len + 2) << 2; 981 else 982 eoff += (eh->ip6e_len + 1) << 3; 983 nxt = eh->ip6e_nxt; 984 break; 985 case IPPROTO_ROUTING: 986 /* 987 * When the erroneous packet contains a 988 * routing header, we should examine the 989 * header to determine the final destination. 990 * Otherwise, we can't properly update 991 * information that depends on the final 992 * destination (e.g. path MTU). 993 */ 994#ifndef PULLDOWN_TEST 995 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), 996 -1); 997 rth = (struct ip6_rthdr *)(mtod(m, caddr_t) 998 + eoff); 999#else 1000 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 1001 eoff, sizeof(*rth)); 1002 if (rth == NULL) { 1003 icmp6stat.icp6s_tooshort++; 1004 return (-1); 1005 } 1006#endif 1007 rthlen = (rth->ip6r_len + 1) << 3; 1008 /* 1009 * XXX: currently there is no 1010 * officially defined type other 1011 * than type-0. 1012 * Note that if the segment left field 1013 * is 0, all intermediate hops must 1014 * have been passed. 1015 */ 1016 if (rth->ip6r_segleft && 1017 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 1018 int hops; 1019 1020#ifndef PULLDOWN_TEST 1021 IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, 1022 -1); 1023 rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff); 1024#else 1025 IP6_EXTHDR_GET(rth0, 1026 struct ip6_rthdr0 *, m, 1027 eoff, rthlen); 1028 if (rth0 == NULL) { 1029 icmp6stat.icp6s_tooshort++; 1030 return (-1); 1031 } 1032#endif 1033 /* just ignore a bogus header */ 1034 if ((rth0->ip6r0_len % 2) == 0 && 1035 (hops = rth0->ip6r0_len/2)) 1036 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 1037 } 1038 eoff += rthlen; 1039 nxt = rth->ip6r_nxt; 1040 break; 1041 case IPPROTO_FRAGMENT: 1042#ifndef PULLDOWN_TEST 1043 IP6_EXTHDR_CHECK(m, 0, eoff + 1044 sizeof(struct ip6_frag), 1045 -1); 1046 fh = (struct ip6_frag *)(mtod(m, caddr_t) 1047 + eoff); 1048#else 1049 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 1050 eoff, sizeof(*fh)); 1051 if (fh == NULL) { 1052 icmp6stat.icp6s_tooshort++; 1053 return (-1); 1054 } 1055#endif 1056 /* 1057 * Data after a fragment header is meaningless 1058 * unless it is the first fragment, but 1059 * we'll go to the notify label for path MTU 1060 * discovery. 1061 */ 1062 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1063 goto notify; 1064 1065 eoff += sizeof(struct ip6_frag); 1066 nxt = fh->ip6f_nxt; 1067 break; 1068 default: 1069 /* 1070 * This case includes ESP and the No Next 1071 * Header. In such cases going to the notify 1072 * label does not have any meaning 1073 * (i.e. ctlfunc will be NULL), but we go 1074 * anyway since we might have to update 1075 * path MTU information. 1076 */ 1077 goto notify; 1078 } 1079 } 1080 notify: 1081#ifndef PULLDOWN_TEST 1082 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 1083#else 1084 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1085 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1086 if (icmp6 == NULL) { 1087 icmp6stat.icp6s_tooshort++; 1088 return (-1); 1089 } 1090#endif 1091 1092 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1093 bzero(&icmp6dst, sizeof(icmp6dst)); 1094 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1095 icmp6dst.sin6_family = AF_INET6; 1096 if (finaldst == NULL) 1097 icmp6dst.sin6_addr = eip6->ip6_dst; 1098 else 1099 icmp6dst.sin6_addr = *finaldst; 1100 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1101 &icmp6dst.sin6_addr); 1102#ifndef SCOPEDROUTING 1103 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst, 1104 NULL, NULL)) { 1105 /* should be impossbile */ 1106 nd6log((LOG_DEBUG, 1107 "icmp6_notify_error: in6_embedscope failed\n")); 1108 goto freeit; 1109 } 1110#endif 1111 1112 /* 1113 * retrieve parameters from the inner IPv6 header, and convert 1114 * them into sockaddr structures. 1115 */ 1116 bzero(&icmp6src, sizeof(icmp6src)); 1117 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1118 icmp6src.sin6_family = AF_INET6; 1119 icmp6src.sin6_addr = eip6->ip6_src; 1120 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1121 &icmp6src.sin6_addr); 1122#ifndef SCOPEDROUTING 1123 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src, 1124 NULL, NULL)) { 1125 /* should be impossbile */ 1126 nd6log((LOG_DEBUG, 1127 "icmp6_notify_error: in6_embedscope failed\n")); 1128 goto freeit; 1129 } 1130#endif 1131 icmp6src.sin6_flowinfo = 1132 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1133 1134 if (finaldst == NULL) 1135 finaldst = &eip6->ip6_dst; 1136 ip6cp.ip6c_m = m; 1137 ip6cp.ip6c_icmp6 = icmp6; 1138 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1139 ip6cp.ip6c_off = eoff; 1140 ip6cp.ip6c_finaldst = finaldst; 1141 ip6cp.ip6c_src = &icmp6src; 1142 ip6cp.ip6c_nxt = nxt; 1143 1144 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1145 notifymtu = ntohl(icmp6->icmp6_mtu); 1146 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1147 } 1148 1149 ctlfunc = (void (*) __P((int, struct sockaddr *, void *))) 1150 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1151 if (ctlfunc) { 1152 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1153 &ip6cp); 1154 } 1155 } 1156 return (0); 1157 1158 freeit: 1159 m_freem(m); 1160 return (-1); 1161} 1162 1163void 1164icmp6_mtudisc_update(ip6cp, validated) 1165 struct ip6ctlparam *ip6cp; 1166 int validated; 1167{ 1168 unsigned long rtcount; 1169 struct icmp6_mtudisc_callback *mc; 1170 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1171 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1172 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1173 u_int mtu = ntohl(icmp6->icmp6_mtu); 1174 struct rtentry *rt = NULL; 1175 struct sockaddr_in6 sin6; 1176 1177 /* 1178 * allow non-validated cases if memory is plenty, to make traffic 1179 * from non-connected pcb happy. 1180 */ 1181 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1182 if (validated) { 1183 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1184 return; 1185 else if (0 <= icmp6_mtudisc_lowat && 1186 rtcount > icmp6_mtudisc_lowat) { 1187 /* 1188 * XXX nuke a victim, install the new one. 1189 */ 1190 } 1191 } else { 1192 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1193 return; 1194 } 1195 1196 bzero(&sin6, sizeof(sin6)); 1197 sin6.sin6_family = PF_INET6; 1198 sin6.sin6_len = sizeof(struct sockaddr_in6); 1199 sin6.sin6_addr = *dst; 1200 /* XXX normally, this won't happen */ 1201 if (IN6_IS_ADDR_LINKLOCAL(dst)) { 1202 sin6.sin6_addr.s6_addr16[1] = 1203 htons(m->m_pkthdr.rcvif->if_index); 1204 } 1205 /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */ 1206 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1207 1208 if (rt && (rt->rt_flags & RTF_HOST) && 1209 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1210 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1211 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1212 icmp6stat.icp6s_pmtuchg++; 1213 rt->rt_rmx.rmx_mtu = mtu; 1214 } 1215 } 1216 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */ 1217 RTFREE(rt); 1218 } 1219 1220 /* 1221 * Notify protocols that the MTU for this destination 1222 * has changed. 1223 */ 1224 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1225 mc = LIST_NEXT(mc, mc_list)) 1226 (*mc->mc_func)(&sin6.sin6_addr); 1227} 1228 1229/* 1230 * Process a Node Information Query packet, based on 1231 * draft-ietf-ipngwg-icmp-name-lookups-07. 1232 * 1233 * Spec incompatibilities: 1234 * - IPv6 Subject address handling 1235 * - IPv4 Subject address handling support missing 1236 * - Proxy reply (answer even if it's not for me) 1237 * - joins NI group address at in6_ifattach() time only, does not cope 1238 * with hostname changes by sethostname(3) 1239 */ 1240#ifndef offsetof /* XXX */ 1241#define offsetof(type, member) ((size_t)(&((type *)0)->member)) 1242#endif 1243static struct mbuf * 1244ni6_input(m, off) 1245 struct mbuf *m; 1246 int off; 1247{ 1248 struct icmp6_nodeinfo *ni6, *nni6; 1249 struct mbuf *n = NULL; 1250 u_int16_t qtype; 1251 int subjlen; 1252 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1253 struct ni_reply_fqdn *fqdn; 1254 int addrs; /* for NI_QTYPE_NODEADDR */ 1255 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1256 struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */ 1257 struct ip6_hdr *ip6; 1258 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1259 char *subj = NULL; 1260 1261 ip6 = mtod(m, struct ip6_hdr *); 1262#ifndef PULLDOWN_TEST 1263 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off); 1264#else 1265 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1266 if (ni6 == NULL) { 1267 /* m is already reclaimed */ 1268 return NULL; 1269 } 1270#endif 1271 1272 /* 1273 * Validate IPv6 destination address. 1274 * 1275 * The Responder must discard the Query without further processing 1276 * unless it is one of the Responder's unicast or anycast addresses, or 1277 * a link-local scope multicast address which the Responder has joined. 1278 * [icmp-name-lookups-07, Section 4.] 1279 */ 1280 bzero(&sin6, sizeof(sin6)); 1281 sin6.sin6_family = AF_INET6; 1282 sin6.sin6_len = sizeof(struct sockaddr_in6); 1283 bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr)); 1284 /* XXX scopeid */ 1285 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1286 ; /* unicast/anycast, fine */ 1287 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1288 ; /* link-local multicast, fine */ 1289 else 1290 goto bad; 1291 1292 /* validate query Subject field. */ 1293 qtype = ntohs(ni6->ni_qtype); 1294 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1295 switch (qtype) { 1296 case NI_QTYPE_NOOP: 1297 case NI_QTYPE_SUPTYPES: 1298 /* 07 draft */ 1299 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1300 break; 1301 /* FALLTHROUGH */ 1302 case NI_QTYPE_FQDN: 1303 case NI_QTYPE_NODEADDR: 1304 switch (ni6->ni_code) { 1305 case ICMP6_NI_SUBJ_IPV6: 1306#if ICMP6_NI_SUBJ_IPV6 != 0 1307 case 0: 1308#endif 1309 /* 1310 * backward compatibility - try to accept 03 draft 1311 * format, where no Subject is present. 1312 */ 1313 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1314 subjlen == 0) { 1315 oldfqdn++; 1316 break; 1317 } 1318#if ICMP6_NI_SUBJ_IPV6 != 0 1319 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1320 goto bad; 1321#endif 1322 1323 if (subjlen != sizeof(sin6.sin6_addr)) 1324 goto bad; 1325 1326 /* 1327 * Validate Subject address. 1328 * 1329 * Not sure what exactly "address belongs to the node" 1330 * means in the spec, is it just unicast, or what? 1331 * 1332 * At this moment we consider Subject address as 1333 * "belong to the node" if the Subject address equals 1334 * to the IPv6 destination address; validation for 1335 * IPv6 destination address should have done enough 1336 * check for us. 1337 * 1338 * We do not do proxy at this moment. 1339 */ 1340 /* m_pulldown instead of copy? */ 1341 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1342 subjlen, (caddr_t)&sin6.sin6_addr); 1343 /* XXX kame scope hack */ 1344 if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) { 1345 if ((m->m_flags & M_PKTHDR) != 0 && 1346 m->m_pkthdr.rcvif) { 1347 sin6.sin6_addr.s6_addr16[1] = 1348 htons(m->m_pkthdr.rcvif->if_index); 1349 } 1350 } 1351 subj = (char *)&sin6; 1352 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr)) 1353 break; 1354 1355 /* 1356 * XXX if we are to allow other cases, we should really 1357 * be careful about scope here. 1358 * basically, we should disallow queries toward IPv6 1359 * destination X with subject Y, if scope(X) > scope(Y). 1360 * if we allow scope(X) > scope(Y), it will result in 1361 * information leakage across scope boundary. 1362 */ 1363 goto bad; 1364 1365 case ICMP6_NI_SUBJ_FQDN: 1366 /* 1367 * Validate Subject name with gethostname(3). 1368 * 1369 * The behavior may need some debate, since: 1370 * - we are not sure if the node has FQDN as 1371 * hostname (returned by gethostname(3)). 1372 * - the code does wildcard match for truncated names. 1373 * however, we are not sure if we want to perform 1374 * wildcard match, if gethostname(3) side has 1375 * truncated hostname. 1376 */ 1377 n = ni6_nametodns(hostname, hostnamelen, 0); 1378 if (!n || n->m_next || n->m_len == 0) 1379 goto bad; 1380 IP6_EXTHDR_GET(subj, char *, m, 1381 off + sizeof(struct icmp6_nodeinfo), subjlen); 1382 if (subj == NULL) 1383 goto bad; 1384 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1385 n->m_len)) { 1386 goto bad; 1387 } 1388 m_freem(n); 1389 n = NULL; 1390 break; 1391 1392 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1393 default: 1394 goto bad; 1395 } 1396 break; 1397 } 1398 1399 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1400 switch (qtype) { 1401 case NI_QTYPE_FQDN: 1402 if ((icmp6_nodeinfo & 1) == 0) 1403 goto bad; 1404 break; 1405 case NI_QTYPE_NODEADDR: 1406 if ((icmp6_nodeinfo & 2) == 0) 1407 goto bad; 1408 break; 1409 } 1410 1411 /* guess reply length */ 1412 switch (qtype) { 1413 case NI_QTYPE_NOOP: 1414 break; /* no reply data */ 1415 case NI_QTYPE_SUPTYPES: 1416 replylen += sizeof(u_int32_t); 1417 break; 1418 case NI_QTYPE_FQDN: 1419 /* XXX will append an mbuf */ 1420 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1421 break; 1422 case NI_QTYPE_NODEADDR: 1423 addrs = ni6_addrs(ni6, m, &ifp, subj); 1424 if ((replylen += addrs * (sizeof(struct in6_addr) + 1425 sizeof(u_int32_t))) > MCLBYTES) 1426 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1427 break; 1428 default: 1429 /* 1430 * XXX: We must return a reply with the ICMP6 code 1431 * `unknown Qtype' in this case. However we regard the case 1432 * as an FQDN query for backward compatibility. 1433 * Older versions set a random value to this field, 1434 * so it rarely varies in the defined qtypes. 1435 * But the mechanism is not reliable... 1436 * maybe we should obsolete older versions. 1437 */ 1438 qtype = NI_QTYPE_FQDN; 1439 /* XXX will append an mbuf */ 1440 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1441 oldfqdn++; 1442 break; 1443 } 1444 1445 /* allocate an mbuf to reply. */ 1446 MGETHDR(n, M_DONTWAIT, m->m_type); 1447 if (n == NULL) { 1448 m_freem(m); 1449 return (NULL); 1450 } 1451 M_COPY_PKTHDR(n, m); /* just for rcvif */ 1452 if (replylen > MHLEN) { 1453 if (replylen > MCLBYTES) { 1454 /* 1455 * XXX: should we try to allocate more? But MCLBYTES 1456 * is probably much larger than IPV6_MMTU... 1457 */ 1458 goto bad; 1459 } 1460 MCLGET(n, M_DONTWAIT); 1461 if ((n->m_flags & M_EXT) == 0) { 1462 goto bad; 1463 } 1464 } 1465 n->m_pkthdr.len = n->m_len = replylen; 1466 1467 /* copy mbuf header and IPv6 + Node Information base headers */ 1468 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1469 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1470 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1471 1472 /* qtype dependent procedure */ 1473 switch (qtype) { 1474 case NI_QTYPE_NOOP: 1475 nni6->ni_code = ICMP6_NI_SUCCESS; 1476 nni6->ni_flags = 0; 1477 break; 1478 case NI_QTYPE_SUPTYPES: 1479 { 1480 u_int32_t v; 1481 nni6->ni_code = ICMP6_NI_SUCCESS; 1482 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1483 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1484 v = (u_int32_t)htonl(0x0000000f); 1485 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1486 break; 1487 } 1488 case NI_QTYPE_FQDN: 1489 nni6->ni_code = ICMP6_NI_SUCCESS; 1490 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1491 sizeof(struct ip6_hdr) + 1492 sizeof(struct icmp6_nodeinfo)); 1493 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1494 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1495 /* 1496 * XXX do we really have FQDN in variable "hostname"? 1497 */ 1498 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1499 if (n->m_next == NULL) 1500 goto bad; 1501 /* XXX we assume that n->m_next is not a chain */ 1502 if (n->m_next->m_next != NULL) 1503 goto bad; 1504 n->m_pkthdr.len += n->m_next->m_len; 1505 break; 1506 case NI_QTYPE_NODEADDR: 1507 { 1508 int lenlim, copied; 1509 1510 nni6->ni_code = ICMP6_NI_SUCCESS; 1511 n->m_pkthdr.len = n->m_len = 1512 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1513 lenlim = M_TRAILINGSPACE(n); 1514 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1515 /* XXX: reset mbuf length */ 1516 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1517 sizeof(struct icmp6_nodeinfo) + copied; 1518 break; 1519 } 1520 default: 1521 break; /* XXX impossible! */ 1522 } 1523 1524 nni6->ni_type = ICMP6_NI_REPLY; 1525 m_freem(m); 1526 return (n); 1527 1528 bad: 1529 m_freem(m); 1530 if (n) 1531 m_freem(n); 1532 return (NULL); 1533} 1534#undef hostnamelen 1535 1536#define isupper(x) ('A' <= (x) && (x) <= 'Z') 1537#define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1538#define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1539#define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1540 1541/* 1542 * make a mbuf with DNS-encoded string. no compression support. 1543 * 1544 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1545 * treated as truncated name (two \0 at the end). this is a wild guess. 1546 */ 1547static struct mbuf * 1548ni6_nametodns(name, namelen, old) 1549 const char *name; 1550 int namelen; 1551 int old; /* return pascal string if non-zero */ 1552{ 1553 struct mbuf *m; 1554 char *cp, *ep; 1555 const char *p, *q; 1556 int i, len, nterm; 1557 1558 if (old) 1559 len = namelen + 1; 1560 else 1561 len = MCLBYTES; 1562 1563 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1564 MGET(m, M_DONTWAIT, MT_DATA); 1565 if (m && len > MLEN) { 1566 MCLGET(m, M_DONTWAIT); 1567 if ((m->m_flags & M_EXT) == 0) 1568 goto fail; 1569 } 1570 if (!m) 1571 goto fail; 1572 m->m_next = NULL; 1573 1574 if (old) { 1575 m->m_len = len; 1576 *mtod(m, char *) = namelen; 1577 bcopy(name, mtod(m, char *) + 1, namelen); 1578 return m; 1579 } else { 1580 m->m_len = 0; 1581 cp = mtod(m, char *); 1582 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1583 1584 /* if not certain about my name, return empty buffer */ 1585 if (namelen == 0) 1586 return m; 1587 1588 /* 1589 * guess if it looks like shortened hostname, or FQDN. 1590 * shortened hostname needs two trailing "\0". 1591 */ 1592 i = 0; 1593 for (p = name; p < name + namelen; p++) { 1594 if (*p && *p == '.') 1595 i++; 1596 } 1597 if (i < 2) 1598 nterm = 2; 1599 else 1600 nterm = 1; 1601 1602 p = name; 1603 while (cp < ep && p < name + namelen) { 1604 i = 0; 1605 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1606 i++; 1607 /* result does not fit into mbuf */ 1608 if (cp + i + 1 >= ep) 1609 goto fail; 1610 /* 1611 * DNS label length restriction, RFC1035 page 8. 1612 * "i == 0" case is included here to avoid returning 1613 * 0-length label on "foo..bar". 1614 */ 1615 if (i <= 0 || i >= 64) 1616 goto fail; 1617 *cp++ = i; 1618 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1619 goto fail; 1620 while (i > 0) { 1621 if (!isalnum(*p) && *p != '-') 1622 goto fail; 1623 if (isupper(*p)) { 1624 *cp++ = tolower(*p); 1625 p++; 1626 } else 1627 *cp++ = *p++; 1628 i--; 1629 } 1630 p = q; 1631 if (p < name + namelen && *p == '.') 1632 p++; 1633 } 1634 /* termination */ 1635 if (cp + nterm >= ep) 1636 goto fail; 1637 while (nterm-- > 0) 1638 *cp++ = '\0'; 1639 m->m_len = cp - mtod(m, char *); 1640 return m; 1641 } 1642 1643 panic("should not reach here"); 1644 /* NOTREACHED */ 1645 1646 fail: 1647 if (m) 1648 m_freem(m); 1649 return NULL; 1650} 1651 1652/* 1653 * check if two DNS-encoded string matches. takes care of truncated 1654 * form (with \0\0 at the end). no compression support. 1655 * XXX upper/lowercase match (see RFC2065) 1656 */ 1657static int 1658ni6_dnsmatch(a, alen, b, blen) 1659 const char *a; 1660 int alen; 1661 const char *b; 1662 int blen; 1663{ 1664 const char *a0, *b0; 1665 int l; 1666 1667 /* simplest case - need validation? */ 1668 if (alen == blen && bcmp(a, b, alen) == 0) 1669 return 1; 1670 1671 a0 = a; 1672 b0 = b; 1673 1674 /* termination is mandatory */ 1675 if (alen < 2 || blen < 2) 1676 return 0; 1677 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1678 return 0; 1679 alen--; 1680 blen--; 1681 1682 while (a - a0 < alen && b - b0 < blen) { 1683 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1684 return 0; 1685 1686 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1687 return 0; 1688 /* we don't support compression yet */ 1689 if (a[0] >= 64 || b[0] >= 64) 1690 return 0; 1691 1692 /* truncated case */ 1693 if (a[0] == 0 && a - a0 == alen - 1) 1694 return 1; 1695 if (b[0] == 0 && b - b0 == blen - 1) 1696 return 1; 1697 if (a[0] == 0 || b[0] == 0) 1698 return 0; 1699 1700 if (a[0] != b[0]) 1701 return 0; 1702 l = a[0]; 1703 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1704 return 0; 1705 if (bcmp(a + 1, b + 1, l) != 0) 1706 return 0; 1707 1708 a += 1 + l; 1709 b += 1 + l; 1710 } 1711 1712 if (a - a0 == alen && b - b0 == blen) 1713 return 1; 1714 else 1715 return 0; 1716} 1717 1718/* 1719 * calculate the number of addresses to be returned in the node info reply. 1720 */ 1721static int 1722ni6_addrs(ni6, m, ifpp, subj) 1723 struct icmp6_nodeinfo *ni6; 1724 struct mbuf *m; 1725 struct ifnet **ifpp; 1726 char *subj; 1727{ 1728 struct ifnet *ifp; 1729 struct in6_ifaddr *ifa6; 1730 struct ifaddr *ifa; 1731 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1732 int addrs = 0, addrsofif, iffound = 0; 1733 int niflags = ni6->ni_flags; 1734 1735 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1736 switch (ni6->ni_code) { 1737 case ICMP6_NI_SUBJ_IPV6: 1738 if (subj == NULL) /* must be impossible... */ 1739 return (0); 1740 subj_ip6 = (struct sockaddr_in6 *)subj; 1741 break; 1742 default: 1743 /* 1744 * XXX: we only support IPv6 subject address for 1745 * this Qtype. 1746 */ 1747 return (0); 1748 } 1749 } 1750 1751 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1752 { 1753 addrsofif = 0; 1754 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1755 ifa = ifa->ifa_list.tqe_next) 1756 { 1757 if (ifa->ifa_addr->sa_family != AF_INET6) 1758 continue; 1759 ifa6 = (struct in6_ifaddr *)ifa; 1760 1761 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1762 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1763 &ifa6->ia_addr.sin6_addr)) 1764 iffound = 1; 1765 1766 /* 1767 * IPv4-mapped addresses can only be returned by a 1768 * Node Information proxy, since they represent 1769 * addresses of IPv4-only nodes, which perforce do 1770 * not implement this protocol. 1771 * [icmp-name-lookups-07, Section 5.4] 1772 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1773 * this function at this moment. 1774 */ 1775 1776 /* What do we have to do about ::1? */ 1777 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1778 case IPV6_ADDR_SCOPE_LINKLOCAL: 1779 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1780 continue; 1781 break; 1782 case IPV6_ADDR_SCOPE_SITELOCAL: 1783 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1784 continue; 1785 break; 1786 case IPV6_ADDR_SCOPE_GLOBAL: 1787 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1788 continue; 1789 break; 1790 default: 1791 continue; 1792 } 1793 1794 /* 1795 * check if anycast is okay. 1796 * XXX: just experimental. not in the spec. 1797 */ 1798 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1799 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1800 continue; /* we need only unicast addresses */ 1801 1802 addrsofif++; /* count the address */ 1803 } 1804 if (iffound) { 1805 *ifpp = ifp; 1806 return (addrsofif); 1807 } 1808 1809 addrs += addrsofif; 1810 } 1811 1812 return (addrs); 1813} 1814 1815static int 1816ni6_store_addrs(ni6, nni6, ifp0, resid) 1817 struct icmp6_nodeinfo *ni6, *nni6; 1818 struct ifnet *ifp0; 1819 int resid; 1820{ 1821 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1822 struct in6_ifaddr *ifa6; 1823 struct ifaddr *ifa; 1824 struct ifnet *ifp_dep = NULL; 1825 int copied = 0, allow_deprecated = 0; 1826 u_char *cp = (u_char *)(nni6 + 1); 1827 int niflags = ni6->ni_flags; 1828 u_int32_t ltime; 1829 long time_second = time.tv_sec; 1830 1831 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1832 return (0); /* needless to copy */ 1833 1834 again: 1835 1836 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1837 { 1838 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1839 ifa = ifa->ifa_list.tqe_next) 1840 { 1841 if (ifa->ifa_addr->sa_family != AF_INET6) 1842 continue; 1843 ifa6 = (struct in6_ifaddr *)ifa; 1844 1845 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1846 allow_deprecated == 0) { 1847 /* 1848 * prefererred address should be put before 1849 * deprecated addresses. 1850 */ 1851 1852 /* record the interface for later search */ 1853 if (ifp_dep == NULL) 1854 ifp_dep = ifp; 1855 1856 continue; 1857 } 1858 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1859 allow_deprecated != 0) 1860 continue; /* we now collect deprecated addrs */ 1861 1862 /* What do we have to do about ::1? */ 1863 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1864 case IPV6_ADDR_SCOPE_LINKLOCAL: 1865 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1866 continue; 1867 break; 1868 case IPV6_ADDR_SCOPE_SITELOCAL: 1869 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1870 continue; 1871 break; 1872 case IPV6_ADDR_SCOPE_GLOBAL: 1873 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1874 continue; 1875 break; 1876 default: 1877 continue; 1878 } 1879 1880 /* 1881 * check if anycast is okay. 1882 * XXX: just experimental. not in the spec. 1883 */ 1884 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1885 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1886 continue; 1887 1888 /* now we can copy the address */ 1889 if (resid < sizeof(struct in6_addr) + 1890 sizeof(u_int32_t)) { 1891 /* 1892 * We give up much more copy. 1893 * Set the truncate flag and return. 1894 */ 1895 nni6->ni_flags |= 1896 NI_NODEADDR_FLAG_TRUNCATE; 1897 return (copied); 1898 } 1899 1900 /* 1901 * Set the TTL of the address. 1902 * The TTL value should be one of the following 1903 * according to the specification: 1904 * 1905 * 1. The remaining lifetime of a DHCP lease on the 1906 * address, or 1907 * 2. The remaining Valid Lifetime of a prefix from 1908 * which the address was derived through Stateless 1909 * Autoconfiguration. 1910 * 1911 * Note that we currently do not support stateful 1912 * address configuration by DHCPv6, so the former 1913 * case can't happen. 1914 * 1915 * TTL must be 2^31 > TTL >= 0. 1916 */ 1917 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1918 ltime = ND6_INFINITE_LIFETIME; 1919 else { 1920 if (ifa6->ia6_lifetime.ia6t_expire > 1921 time_second) 1922 ltime = ifa6->ia6_lifetime.ia6t_expire - time_second; 1923 else 1924 ltime = 0; 1925 } 1926 if (ltime > 0x7fffffff) 1927 ltime = 0x7fffffff; 1928 ltime = htonl(ltime); 1929 1930 bcopy(<ime, cp, sizeof(u_int32_t)); 1931 cp += sizeof(u_int32_t); 1932 1933 /* copy the address itself */ 1934 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1935 sizeof(struct in6_addr)); 1936 /* XXX: KAME link-local hack; remove ifindex */ 1937 if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr)) 1938 ((struct in6_addr *)cp)->s6_addr16[1] = 0; 1939 cp += sizeof(struct in6_addr); 1940 1941 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1942 copied += (sizeof(struct in6_addr) + 1943 sizeof(u_int32_t)); 1944 } 1945 if (ifp0) /* we need search only on the specified IF */ 1946 break; 1947 } 1948 1949 if (allow_deprecated == 0 && ifp_dep != NULL) { 1950 ifp = ifp_dep; 1951 allow_deprecated = 1; 1952 1953 goto again; 1954 } 1955 1956 return (copied); 1957} 1958 1959/* 1960 * XXX almost dup'ed code with rip6_input. 1961 */ 1962static int 1963icmp6_rip6_input(mp, off) 1964 struct mbuf **mp; 1965 int off; 1966{ 1967 struct mbuf *m = *mp; 1968 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1969 struct in6pcb *in6p; 1970 struct in6pcb *last = NULL; 1971 struct sockaddr_in6 rip6src; 1972 struct icmp6_hdr *icmp6; 1973 struct mbuf *opts = NULL; 1974 1975#ifndef PULLDOWN_TEST 1976 /* this is assumed to be safe. */ 1977 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 1978#else 1979 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1980 if (icmp6 == NULL) { 1981 /* m is already reclaimed */ 1982 return IPPROTO_DONE; 1983 } 1984#endif 1985 1986 bzero(&rip6src, sizeof(rip6src)); 1987 rip6src.sin6_len = sizeof(struct sockaddr_in6); 1988 rip6src.sin6_family = AF_INET6; 1989 /* KAME hack: recover scopeid */ 1990 (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif); 1991 1992 for (in6p = rawin6pcb.in6p_next; 1993 in6p != &rawin6pcb; in6p = in6p->in6p_next) 1994 { 1995 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6) 1996 continue; 1997 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1998 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1999 continue; 2000 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 2001 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 2002 continue; 2003 if (in6p->in6p_icmp6filt 2004 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 2005 in6p->in6p_icmp6filt)) 2006 continue; 2007 if (last) { 2008 struct mbuf *n; 2009 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 2010 if (last->in6p_flags & IN6P_CONTROLOPTS) 2011 ip6_savecontrol(last, &opts, ip6, n); 2012 /* strip intermediate headers */ 2013 m_adj(n, off); 2014 if (sbappendaddr(&last->in6p_socket->so_rcv, 2015 (struct sockaddr *)&rip6src, 2016 n, opts) == 0) { 2017 /* should notify about lost packet */ 2018 m_freem(n); 2019 if (opts) 2020 m_freem(opts); 2021 } else 2022 sorwakeup(last->in6p_socket); 2023 opts = NULL; 2024 } 2025 } 2026 last = in6p; 2027 } 2028 if (last) { 2029 if (last->in6p_flags & IN6P_CONTROLOPTS) 2030 ip6_savecontrol(last, &opts, ip6, m); 2031 /* strip intermediate headers */ 2032 m_adj(m, off); 2033 if (sbappendaddr(&last->in6p_socket->so_rcv, 2034 (struct sockaddr *)&rip6src, m, opts) == 0) { 2035 m_freem(m); 2036 if (opts) 2037 m_freem(opts); 2038 } else 2039 sorwakeup(last->in6p_socket); 2040 } else { 2041 m_freem(m); 2042 ip6stat.ip6s_delivered--; 2043 } 2044 return IPPROTO_DONE; 2045} 2046 2047/* 2048 * Reflect the ip6 packet back to the source. 2049 * OFF points to the icmp6 header, counted from the top of the mbuf. 2050 * 2051 * Note: RFC 1885 required that an echo reply should be truncated if it 2052 * did not fit in with (return) path MTU, and KAME code supported the 2053 * behavior. However, as a clarification after the RFC, this limitation 2054 * was removed in a revised version of the spec, RFC 2463. We had kept the 2055 * old behavior, with a (non-default) ifdef block, while the new version of 2056 * the spec was an internet-draft status, and even after the new RFC was 2057 * published. But it would rather make sense to clean the obsoleted part 2058 * up, and to make the code simpler at this stage. 2059 */ 2060void 2061icmp6_reflect(m, off) 2062 struct mbuf *m; 2063 size_t off; 2064{ 2065 struct ip6_hdr *ip6; 2066 struct icmp6_hdr *icmp6; 2067 struct in6_ifaddr *ia; 2068 struct in6_addr t, *src = 0; 2069 int plen; 2070 int type, code; 2071 struct ifnet *outif = NULL; 2072 struct sockaddr_in6 sa6_src, sa6_dst; 2073 2074 /* too short to reflect */ 2075 if (off < sizeof(struct ip6_hdr)) { 2076 nd6log((LOG_DEBUG, 2077 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 2078 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2079 __FILE__, __LINE__)); 2080 goto bad; 2081 } 2082 2083 /* 2084 * If there are extra headers between IPv6 and ICMPv6, strip 2085 * off that header first. 2086 */ 2087#ifdef DIAGNOSTIC 2088 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2089 panic("assumption failed in icmp6_reflect"); 2090#endif 2091 if (off > sizeof(struct ip6_hdr)) { 2092 size_t l; 2093 struct ip6_hdr nip6; 2094 2095 l = off - sizeof(struct ip6_hdr); 2096 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2097 m_adj(m, l); 2098 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2099 if (m->m_len < l) { 2100 if ((m = m_pullup(m, l)) == NULL) 2101 return; 2102 } 2103 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2104 } else /* off == sizeof(struct ip6_hdr) */ { 2105 size_t l; 2106 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2107 if (m->m_len < l) { 2108 if ((m = m_pullup(m, l)) == NULL) 2109 return; 2110 } 2111 } 2112 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2113 ip6 = mtod(m, struct ip6_hdr *); 2114 ip6->ip6_nxt = IPPROTO_ICMPV6; 2115 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2116 type = icmp6->icmp6_type; /* keep type for statistics */ 2117 code = icmp6->icmp6_code; /* ditto. */ 2118 2119 t = ip6->ip6_dst; 2120 /* 2121 * ip6_input() drops a packet if its src is multicast. 2122 * So, the src is never multicast. 2123 */ 2124 ip6->ip6_dst = ip6->ip6_src; 2125 2126 /* 2127 * XXX: make sure to embed scope zone information, using 2128 * already embedded IDs or the received interface (if any). 2129 * Note that rcvif may be NULL. 2130 * TODO: scoped routing case (XXX). 2131 */ 2132 bzero(&sa6_src, sizeof(sa6_src)); 2133 sa6_src.sin6_family = AF_INET6; 2134 sa6_src.sin6_len = sizeof(sa6_src); 2135 sa6_src.sin6_addr = ip6->ip6_dst; 2136 in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif); 2137 in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL); 2138 ip6->ip6_dst = sa6_src.sin6_addr; 2139 2140 bzero(&sa6_dst, sizeof(sa6_dst)); 2141 sa6_dst.sin6_family = AF_INET6; 2142 sa6_dst.sin6_len = sizeof(sa6_dst); 2143 sa6_dst.sin6_addr = t; 2144 in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif); 2145 in6_embedscope(&t, &sa6_dst, NULL, NULL); 2146 2147 /* 2148 * If the incoming packet was addressed directly to us (i.e. unicast), 2149 * use dst as the src for the reply. 2150 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible 2151 * (for example) when we encounter an error while forwarding procedure 2152 * destined to a duplicated address of ours. 2153 */ 2154 for (ia = in6_ifaddr; ia; ia = ia->ia_next) 2155 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) && 2156 (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) { 2157 src = &t; 2158 break; 2159 } 2160 if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) { 2161 /* 2162 * This is the case if the dst is our link-local address 2163 * and the sender is also ourselves. 2164 */ 2165 src = &t; 2166 } 2167 2168 if (src == 0) { 2169 int e; 2170 struct route_in6 ro; 2171 2172 /* 2173 * This case matches to multicasts, our anycast, or unicasts 2174 * that we do not own. Select a source address based on the 2175 * source address of the erroneous packet. 2176 */ 2177 bzero(&ro, sizeof(ro)); 2178 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e); 2179 if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */ 2180 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2181 } 2182 if (src == NULL) { 2183 nd6log((LOG_DEBUG, 2184 "icmp6_reflect: source can't be determined: " 2185 "dst=%s, error=%d\n", 2186 ip6_sprintf(&sa6_src.sin6_addr), e)); 2187 goto bad; 2188 } 2189 } 2190 2191 ip6->ip6_src = *src; 2192 2193 ip6->ip6_flow = 0; 2194 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2195 ip6->ip6_vfc |= IPV6_VERSION; 2196 ip6->ip6_nxt = IPPROTO_ICMPV6; 2197 if (m->m_pkthdr.rcvif) { 2198 /* XXX: This may not be the outgoing interface */ 2199 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2200 } else 2201 ip6->ip6_hlim = ip6_defhlim; 2202 2203 icmp6->icmp6_cksum = 0; 2204 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2205 sizeof(struct ip6_hdr), plen); 2206 2207 /* 2208 * XXX option handling 2209 */ 2210 2211 m->m_flags &= ~(M_BCAST|M_MCAST); 2212#ifdef IPSEC 2213 /* Don't lookup socket */ 2214 (void)ipsec_setsocket(m, NULL); 2215#endif /* IPSEC */ 2216 2217 /* 2218 * To avoid a "too big" situation at an intermediate router 2219 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2220 * Note that only echo and node information replies are affected, 2221 * since the length of ICMP6 errors is limited to the minimum MTU. 2222 */ 2223 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, &outif) != 0 && outif) 2224 icmp6_ifstat_inc(outif, ifs6_out_error); 2225 2226 if (outif) 2227 icmp6_ifoutstat_inc(outif, type, code); 2228 2229 return; 2230 2231 bad: 2232 m_freem(m); 2233 return; 2234} 2235 2236void 2237icmp6_fasttimo() 2238{ 2239 2240 mld6_fasttimeo(); 2241} 2242 2243static const char * 2244icmp6_redirect_diag(src6, dst6, tgt6) 2245 struct in6_addr *src6; 2246 struct in6_addr *dst6; 2247 struct in6_addr *tgt6; 2248{ 2249 static char buf[1024]; 2250 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2251 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2252 return buf; 2253} 2254 2255void 2256icmp6_redirect_input(m, off) 2257 struct mbuf *m; 2258 int off; 2259{ 2260 struct ifnet *ifp = m->m_pkthdr.rcvif; 2261 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2262 struct nd_redirect *nd_rd; 2263 int icmp6len = ntohs(ip6->ip6_plen); 2264 char *lladdr = NULL; 2265 int lladdrlen = 0; 2266 u_char *redirhdr = NULL; 2267 int redirhdrlen = 0; 2268 struct rtentry *rt = NULL; 2269 int is_router; 2270 int is_onlink; 2271 struct in6_addr src6 = ip6->ip6_src; 2272 struct in6_addr redtgt6; 2273 struct in6_addr reddst6; 2274 union nd_opts ndopts; 2275 2276 if (!ifp) 2277 return; 2278 2279 /* XXX if we are router, we don't update route by icmp6 redirect */ 2280 if (ip6_forwarding) 2281 goto freeit; 2282 if (!icmp6_rediraccept) 2283 goto freeit; 2284 2285#ifndef PULLDOWN_TEST 2286 IP6_EXTHDR_CHECK(m, off, icmp6len,); 2287 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off); 2288#else 2289 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2290 if (nd_rd == NULL) { 2291 icmp6stat.icp6s_tooshort++; 2292 return; 2293 } 2294#endif 2295 redtgt6 = nd_rd->nd_rd_target; 2296 reddst6 = nd_rd->nd_rd_dst; 2297 2298 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2299 redtgt6.s6_addr16[1] = htons(ifp->if_index); 2300 if (IN6_IS_ADDR_LINKLOCAL(&reddst6)) 2301 reddst6.s6_addr16[1] = htons(ifp->if_index); 2302 2303 /* validation */ 2304 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2305 nd6log((LOG_ERR, 2306 "ICMP6 redirect sent from %s rejected; " 2307 "must be from linklocal\n", ip6_sprintf(&src6))); 2308 goto bad; 2309 } 2310 if (ip6->ip6_hlim != 255) { 2311 nd6log((LOG_ERR, 2312 "ICMP6 redirect sent from %s rejected; " 2313 "hlim=%d (must be 255)\n", 2314 ip6_sprintf(&src6), ip6->ip6_hlim)); 2315 goto bad; 2316 } 2317 { 2318 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2319 struct sockaddr_in6 sin6; 2320 struct in6_addr *gw6; 2321 2322 bzero(&sin6, sizeof(sin6)); 2323 sin6.sin6_family = AF_INET6; 2324 sin6.sin6_len = sizeof(struct sockaddr_in6); 2325 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2326 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2327 if (rt) { 2328 if (rt->rt_gateway == NULL || 2329 rt->rt_gateway->sa_family != AF_INET6) { 2330 nd6log((LOG_ERR, 2331 "ICMP6 redirect rejected; no route " 2332 "with inet6 gateway found for redirect dst: %s\n", 2333 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2334 RTFREE(rt); 2335 goto bad; 2336 } 2337 2338 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2339 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2340 nd6log((LOG_ERR, 2341 "ICMP6 redirect rejected; " 2342 "not equal to gw-for-src=%s (must be same): " 2343 "%s\n", 2344 ip6_sprintf(gw6), 2345 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2346 RTFREE(rt); 2347 goto bad; 2348 } 2349 } else { 2350 nd6log((LOG_ERR, 2351 "ICMP6 redirect rejected; " 2352 "no route found for redirect dst: %s\n", 2353 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2354 goto bad; 2355 } 2356 RTFREE(rt); 2357 rt = NULL; 2358 } 2359 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2360 nd6log((LOG_ERR, 2361 "ICMP6 redirect rejected; " 2362 "redirect dst must be unicast: %s\n", 2363 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2364 goto bad; 2365 } 2366 2367 is_router = is_onlink = 0; 2368 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2369 is_router = 1; /* router case */ 2370 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2371 is_onlink = 1; /* on-link destination case */ 2372 if (!is_router && !is_onlink) { 2373 nd6log((LOG_ERR, 2374 "ICMP6 redirect rejected; " 2375 "neither router case nor onlink case: %s\n", 2376 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2377 goto bad; 2378 } 2379 /* validation passed */ 2380 2381 icmp6len -= sizeof(*nd_rd); 2382 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2383 if (nd6_options(&ndopts) < 0) { 2384 nd6log((LOG_INFO, "icmp6_redirect_input: " 2385 "invalid ND option, rejected: %s\n", 2386 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2387 /* nd6_options have incremented stats */ 2388 goto freeit; 2389 } 2390 2391 if (ndopts.nd_opts_tgt_lladdr) { 2392 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2393 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2394 } 2395 2396 if (ndopts.nd_opts_rh) { 2397 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len; 2398 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */ 2399 } 2400 2401 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2402 nd6log((LOG_INFO, 2403 "icmp6_redirect_input: lladdrlen mismatch for %s " 2404 "(if %d, icmp6 packet %d): %s\n", 2405 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2406 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2407 goto bad; 2408 } 2409 2410 /* RFC 2461 8.3 */ 2411 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2412 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2413 2414 if (!is_onlink) { /* better router case. perform rtredirect. */ 2415 /* perform rtredirect */ 2416 struct sockaddr_in6 sdst; 2417 struct sockaddr_in6 sgw; 2418 struct sockaddr_in6 ssrc; 2419 unsigned long rtcount; 2420 struct rtentry *newrt = NULL; 2421 2422 /* 2423 * do not install redirect route, if the number of entries 2424 * is too much (> hiwat). note that, the node (= host) will 2425 * work just fine even if we do not install redirect route 2426 * (there will be additional hops, though). 2427 */ 2428 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2429 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2430 return; 2431 else if (0 <= icmp6_redirect_lowat && 2432 rtcount > icmp6_redirect_lowat) { 2433 /* 2434 * XXX nuke a victim, install the new one. 2435 */ 2436 } 2437 2438 bzero(&sdst, sizeof(sdst)); 2439 bzero(&sgw, sizeof(sgw)); 2440 bzero(&ssrc, sizeof(ssrc)); 2441 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2442 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2443 sizeof(struct sockaddr_in6); 2444 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2445 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2446 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2447 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2448 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2449 (struct sockaddr *)&ssrc, 2450 &newrt); 2451 2452 if (newrt) { 2453 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2454 icmp6_redirect_timeout_q); 2455 rtfree(newrt); 2456 } 2457 } 2458 /* finally update cached route in each socket via pfctlinput */ 2459 { 2460 struct sockaddr_in6 sdst; 2461 2462 bzero(&sdst, sizeof(sdst)); 2463 sdst.sin6_family = AF_INET6; 2464 sdst.sin6_len = sizeof(struct sockaddr_in6); 2465 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2466 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2467#ifdef IPSEC 2468 key_sa_routechange((struct sockaddr *)&sdst); 2469#endif 2470 } 2471 2472 freeit: 2473 m_freem(m); 2474 return; 2475 2476 bad: 2477 icmp6stat.icp6s_badredirect++; 2478 m_freem(m); 2479} 2480 2481void 2482icmp6_redirect_output(m0, rt) 2483 struct mbuf *m0; 2484 struct rtentry *rt; 2485{ 2486 struct ifnet *ifp; /* my outgoing interface */ 2487 struct in6_addr *ifp_ll6; 2488 struct in6_addr *nexthop; 2489 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2490 struct mbuf *m = NULL; /* newly allocated one */ 2491 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2492 struct nd_redirect *nd_rd; 2493 size_t maxlen; 2494 u_char *p; 2495 struct sockaddr_in6 src_sa; 2496 2497 icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2498 2499 /* if we are not router, we don't send icmp6 redirect */ 2500 if (!ip6_forwarding || ip6_accept_rtadv) 2501 goto fail; 2502 2503 /* sanity check */ 2504 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2505 goto fail; 2506 2507 /* 2508 * Address check: 2509 * the source address must identify a neighbor, and 2510 * the destination address must not be a multicast address 2511 * [RFC 2461, sec 8.2] 2512 */ 2513 sip6 = mtod(m0, struct ip6_hdr *); 2514 bzero(&src_sa, sizeof(src_sa)); 2515 src_sa.sin6_family = AF_INET6; 2516 src_sa.sin6_len = sizeof(src_sa); 2517 src_sa.sin6_addr = sip6->ip6_src; 2518 /* we don't currently use sin6_scope_id, but eventually use it */ 2519 src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src); 2520 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2521 goto fail; 2522 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2523 goto fail; /* what should we do here? */ 2524 2525 /* rate limit */ 2526 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2527 goto fail; 2528 2529 /* 2530 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2531 * we almost always ask for an mbuf cluster for simplicity. 2532 * (MHLEN < IPV6_MMTU is almost always true) 2533 */ 2534#if IPV6_MMTU >= MCLBYTES 2535# error assumption failed about IPV6_MMTU and MCLBYTES 2536#endif 2537 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2538 if (m && IPV6_MMTU >= MHLEN) 2539 MCLGET(m, M_DONTWAIT); 2540 if (!m) 2541 goto fail; 2542 m->m_pkthdr.rcvif = NULL; 2543 m->m_len = 0; 2544 maxlen = M_TRAILINGSPACE(m); 2545 maxlen = min(IPV6_MMTU, maxlen); 2546 /* just for safety */ 2547 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2548 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2549 goto fail; 2550 } 2551 2552 { 2553 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2554 struct in6_ifaddr *ia; 2555 if ((ia = in6ifa_ifpforlinklocal(ifp, 2556 IN6_IFF_NOTREADY| 2557 IN6_IFF_ANYCAST)) == NULL) 2558 goto fail; 2559 ifp_ll6 = &ia->ia_addr.sin6_addr; 2560 } 2561 2562 /* get ip6 linklocal address for the router. */ 2563 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2564 struct sockaddr_in6 *sin6; 2565 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2566 nexthop = &sin6->sin6_addr; 2567 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2568 nexthop = NULL; 2569 } else 2570 nexthop = NULL; 2571 2572 /* ip6 */ 2573 ip6 = mtod(m, struct ip6_hdr *); 2574 ip6->ip6_flow = 0; 2575 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2576 ip6->ip6_vfc |= IPV6_VERSION; 2577 /* ip6->ip6_plen will be set later */ 2578 ip6->ip6_nxt = IPPROTO_ICMPV6; 2579 ip6->ip6_hlim = 255; 2580 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2581 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2582 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2583 2584 /* ND Redirect */ 2585 nd_rd = (struct nd_redirect *)(ip6 + 1); 2586 nd_rd->nd_rd_type = ND_REDIRECT; 2587 nd_rd->nd_rd_code = 0; 2588 nd_rd->nd_rd_reserved = 0; 2589 if (rt->rt_flags & RTF_GATEWAY) { 2590 /* 2591 * nd_rd->nd_rd_target must be a link-local address in 2592 * better router cases. 2593 */ 2594 if (!nexthop) 2595 goto fail; 2596 bcopy(nexthop, &nd_rd->nd_rd_target, 2597 sizeof(nd_rd->nd_rd_target)); 2598 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2599 sizeof(nd_rd->nd_rd_dst)); 2600 } else { 2601 /* make sure redtgt == reddst */ 2602 nexthop = &sip6->ip6_dst; 2603 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2604 sizeof(nd_rd->nd_rd_target)); 2605 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2606 sizeof(nd_rd->nd_rd_dst)); 2607 } 2608 2609 p = (u_char *)(nd_rd + 1); 2610 2611 { 2612 /* target lladdr option */ 2613 struct rtentry *rt_nexthop = NULL; 2614 int len; 2615 struct sockaddr_dl *sdl; 2616 struct nd_opt_hdr *nd_opt; 2617 char *lladdr; 2618 2619 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2620 if (!rt_nexthop) 2621 goto nolladdropt; 2622 len = sizeof(*nd_opt) + ifp->if_addrlen; 2623 len = (len + 7) & ~7; /* round by 8 */ 2624 /* safety check */ 2625 if (len + (p - (u_char *)ip6) > maxlen) 2626 goto nolladdropt; 2627 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2628 (rt_nexthop->rt_flags & RTF_LLINFO) && 2629 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2630 (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) && 2631 sdl->sdl_alen) { 2632 nd_opt = (struct nd_opt_hdr *)p; 2633 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2634 nd_opt->nd_opt_len = len >> 3; 2635 lladdr = (char *)(nd_opt + 1); 2636 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen); 2637 p += len; 2638 } 2639 } 2640 nolladdropt:; 2641 2642 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2643 2644 /* just to be safe */ 2645 if (m0->m_flags & M_DECRYPTED) 2646 goto noredhdropt; 2647 if (p - (u_char *)ip6 > maxlen) 2648 goto noredhdropt; 2649 2650 { 2651 /* redirected header option */ 2652 int len; 2653 struct nd_opt_rd_hdr *nd_opt_rh; 2654 2655 /* 2656 * compute the maximum size for icmp6 redirect header option. 2657 * XXX room for auth header? 2658 */ 2659 len = maxlen - (p - (u_char *)ip6); 2660 len &= ~7; 2661 2662 /* This is just for simplicity. */ 2663 if (m0->m_pkthdr.len != m0->m_len) { 2664 if (m0->m_next) { 2665 m_freem(m0->m_next); 2666 m0->m_next = NULL; 2667 } 2668 m0->m_pkthdr.len = m0->m_len; 2669 } 2670 2671 /* 2672 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2673 * about padding/truncate rule for the original IP packet. 2674 * From the discussion on IPv6imp in Feb 1999, the consensus was: 2675 * - "attach as much as possible" is the goal 2676 * - pad if not aligned (original size can be guessed by original 2677 * ip6 header) 2678 * Following code adds the padding if it is simple enough, 2679 * and truncates if not. 2680 */ 2681 if (m0->m_next || m0->m_pkthdr.len != m0->m_len) 2682 panic("assumption failed in %s:%d", __FILE__, __LINE__); 2683 2684 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2685 /* not enough room, truncate */ 2686 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh); 2687 } else { 2688 /* enough room, pad or truncate */ 2689 size_t extra; 2690 2691 extra = m0->m_pkthdr.len % 8; 2692 if (extra) { 2693 /* pad if easy enough, truncate if not */ 2694 if (8 - extra <= M_TRAILINGSPACE(m0)) { 2695 /* pad */ 2696 m0->m_len += (8 - extra); 2697 m0->m_pkthdr.len += (8 - extra); 2698 } else { 2699 /* truncate */ 2700 m0->m_pkthdr.len -= extra; 2701 m0->m_len -= extra; 2702 } 2703 } 2704 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2705 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh); 2706 } 2707 2708 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2709 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2710 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2711 nd_opt_rh->nd_opt_rh_len = len >> 3; 2712 p += sizeof(*nd_opt_rh); 2713 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2714 2715 /* connect m0 to m */ 2716 m->m_next = m0; 2717 m->m_pkthdr.len = m->m_len + m0->m_len; 2718 m0 = NULL; 2719 } 2720noredhdropt:; 2721 if (m0) { 2722 m_freem(m0); 2723 m0 = NULL; 2724 } 2725 2726 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src)) 2727 sip6->ip6_src.s6_addr16[1] = 0; 2728 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst)) 2729 sip6->ip6_dst.s6_addr16[1] = 0; 2730#if 0 2731 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) 2732 ip6->ip6_src.s6_addr16[1] = 0; 2733 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) 2734 ip6->ip6_dst.s6_addr16[1] = 0; 2735#endif 2736 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target)) 2737 nd_rd->nd_rd_target.s6_addr16[1] = 0; 2738 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst)) 2739 nd_rd->nd_rd_dst.s6_addr16[1] = 0; 2740 2741 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2742 2743 nd_rd->nd_rd_cksum = 0; 2744 nd_rd->nd_rd_cksum 2745 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2746 2747 /* send the packet to outside... */ 2748#ifdef IPSEC 2749 /* Don't lookup socket */ 2750 (void)ipsec_setsocket(m, NULL); 2751#endif /* IPSEC */ 2752 if (ip6_output(m, NULL, NULL, 0, NULL, NULL) != 0) 2753 icmp6_ifstat_inc(ifp, ifs6_out_error); 2754 2755 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2756 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2757 icmp6stat.icp6s_outhist[ND_REDIRECT]++; 2758 2759 return; 2760 2761fail: 2762 if (m) 2763 m_freem(m); 2764 if (m0) 2765 m_freem(m0); 2766} 2767 2768/* 2769 * ICMPv6 socket option processing. 2770 */ 2771int 2772icmp6_ctloutput(op, so, level, optname, mp) 2773 int op; 2774 struct socket *so; 2775 int level, optname; 2776 struct mbuf **mp; 2777{ 2778 int error = 0; 2779 int optlen; 2780 struct in6pcb *in6p = sotoin6pcb(so); 2781 struct mbuf *m = *mp; 2782 2783 optlen = m ? m->m_len : 0; 2784 2785 if (level != IPPROTO_ICMPV6) { 2786 if (op == PRCO_SETOPT && m) 2787 (void)m_free(m); 2788 return EINVAL; 2789 } 2790 2791 switch (op) { 2792 case PRCO_SETOPT: 2793 switch (optname) { 2794 case ICMP6_FILTER: 2795 { 2796 struct icmp6_filter *p; 2797 2798 if (optlen != sizeof(*p)) { 2799 error = EMSGSIZE; 2800 break; 2801 } 2802 p = mtod(m, struct icmp6_filter *); 2803 if (!p || !in6p->in6p_icmp6filt) { 2804 error = EINVAL; 2805 break; 2806 } 2807 bcopy(p, in6p->in6p_icmp6filt, 2808 sizeof(struct icmp6_filter)); 2809 error = 0; 2810 break; 2811 } 2812 2813 default: 2814 error = ENOPROTOOPT; 2815 break; 2816 } 2817 if (m) 2818 (void)m_freem(m); 2819 break; 2820 2821 case PRCO_GETOPT: 2822 switch (optname) { 2823 case ICMP6_FILTER: 2824 { 2825 struct icmp6_filter *p; 2826 2827 if (!in6p->in6p_icmp6filt) { 2828 error = EINVAL; 2829 break; 2830 } 2831 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2832 m->m_len = sizeof(struct icmp6_filter); 2833 p = mtod(m, struct icmp6_filter *); 2834 bcopy(in6p->in6p_icmp6filt, p, 2835 sizeof(struct icmp6_filter)); 2836 error = 0; 2837 break; 2838 } 2839 2840 default: 2841 error = ENOPROTOOPT; 2842 break; 2843 } 2844 break; 2845 } 2846 2847 return (error); 2848} 2849 2850/* 2851 * Perform rate limit check. 2852 * Returns 0 if it is okay to send the icmp6 packet. 2853 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2854 * limitation. 2855 * 2856 * XXX per-destination/type check necessary? 2857 */ 2858static int 2859icmp6_ratelimit(dst, type, code) 2860 const struct in6_addr *dst; /* not used at this moment */ 2861 const int type; /* not used at this moment */ 2862 const int code; /* not used at this moment */ 2863{ 2864 int ret; 2865 2866 ret = 0; /* okay to send */ 2867 2868 /* PPS limit */ 2869 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2870 icmp6errppslim)) { 2871 /* The packet is subject to rate limit */ 2872 ret++; 2873 } 2874 2875 return ret; 2876} 2877 2878static struct rtentry * 2879icmp6_mtudisc_clone(dst) 2880 struct sockaddr *dst; 2881{ 2882 struct rtentry *rt; 2883 int error; 2884 2885 rt = rtalloc1(dst, 1); 2886 if (rt == 0) 2887 return NULL; 2888 2889 /* If we didn't get a host route, allocate one */ 2890 if ((rt->rt_flags & RTF_HOST) == 0) { 2891 struct rtentry *nrt; 2892 2893 error = rtrequest((int) RTM_ADD, dst, 2894 (struct sockaddr *) rt->rt_gateway, 2895 (struct sockaddr *) 0, 2896 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2897 if (error) { 2898 rtfree(rt); 2899 return NULL; 2900 } 2901 nrt->rt_rmx = rt->rt_rmx; 2902 rtfree(rt); 2903 rt = nrt; 2904 } 2905 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2906 icmp6_mtudisc_timeout_q); 2907 if (error) { 2908 rtfree(rt); 2909 return NULL; 2910 } 2911 2912 return rt; /* caller need to call rtfree() */ 2913} 2914 2915static void 2916icmp6_mtudisc_timeout(rt, r) 2917 struct rtentry *rt; 2918 struct rttimer *r; 2919{ 2920 if (rt == NULL) 2921 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2922 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2923 (RTF_DYNAMIC | RTF_HOST)) { 2924 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2925 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2926 } else { 2927 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2928 rt->rt_rmx.rmx_mtu = 0; 2929 } 2930} 2931 2932static void 2933icmp6_redirect_timeout(rt, r) 2934 struct rtentry *rt; 2935 struct rttimer *r; 2936{ 2937 if (rt == NULL) 2938 panic("icmp6_redirect_timeout: bad route to timeout"); 2939 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2940 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2941 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2942 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2943 } 2944} 2945 2946int 2947icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 2948 int *name; 2949 u_int namelen; 2950 void *oldp; 2951 size_t *oldlenp; 2952 void *newp; 2953 size_t newlen; 2954{ 2955 2956 /* All sysctl names at this level are terminal. */ 2957 if (namelen != 1) 2958 return ENOTDIR; 2959 2960 switch (name[0]) { 2961 2962 case ICMPV6CTL_REDIRACCEPT: 2963 return sysctl_int(oldp, oldlenp, newp, newlen, 2964 &icmp6_rediraccept); 2965 case ICMPV6CTL_REDIRTIMEOUT: 2966 return sysctl_int(oldp, oldlenp, newp, newlen, 2967 &icmp6_redirtimeout); 2968 case ICMPV6CTL_STATS: 2969 return sysctl_rdstruct(oldp, oldlenp, newp, 2970 &icmp6stat, sizeof(icmp6stat)); 2971 case ICMPV6CTL_ND6_PRUNE: 2972 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune); 2973 case ICMPV6CTL_ND6_DELAY: 2974 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay); 2975 case ICMPV6CTL_ND6_UMAXTRIES: 2976 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries); 2977 case ICMPV6CTL_ND6_MMAXTRIES: 2978 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries); 2979 case ICMPV6CTL_ND6_USELOOPBACK: 2980 return sysctl_int(oldp, oldlenp, newp, newlen, 2981 &nd6_useloopback); 2982 case ICMPV6CTL_NODEINFO: 2983 return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo); 2984 case ICMPV6CTL_ERRPPSLIMIT: 2985 return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6errppslim); 2986 case ICMPV6CTL_ND6_MAXNUDHINT: 2987 return sysctl_int(oldp, oldlenp, newp, newlen, 2988 &nd6_maxnudhint); 2989 case ICMPV6CTL_MTUDISC_HIWAT: 2990 return sysctl_int(oldp, oldlenp, newp, newlen, 2991 &icmp6_mtudisc_hiwat); 2992 case ICMPV6CTL_MTUDISC_LOWAT: 2993 return sysctl_int(oldp, oldlenp, newp, newlen, 2994 &icmp6_mtudisc_lowat); 2995 case ICMPV6CTL_ND6_DEBUG: 2996 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_debug); 2997 case ICMPV6CTL_ND6_DRLIST: 2998 case ICMPV6CTL_ND6_PRLIST: 2999 return nd6_sysctl(name[0], oldp, oldlenp, newp, newlen); 3000 default: 3001 return ENOPROTOOPT; 3002 } 3003 /* NOTREACHED */ 3004} 3005