icmp6.c revision 1.91
1/* $NetBSD: icmp6.c,v 1.91 2003/06/03 05:20:06 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 1.91 2003/06/03 05:20:06 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 if (m->m_len < sizeof(struct ip6_hdr)) { 283 m = m_pullup(m, sizeof(struct ip6_hdr)); 284 if (m == NULL) 285 return; 286 } 287 oip6 = mtod(m, struct ip6_hdr *); 288 289 /* 290 * If the destination address of the erroneous packet is a multicast 291 * address, or the packet was sent using link-layer multicast, 292 * we should basically suppress sending an error (RFC 2463, Section 293 * 2.4). 294 * We have two exceptions (the item e.2 in that section): 295 * - the Pakcet Too Big message can be sent for path MTU discovery. 296 * - the Parameter Problem Message that can be allowed an icmp6 error 297 * in the option type field. This check has been done in 298 * ip6_unknown_opt(), so we can just check the type and code. 299 */ 300 if ((m->m_flags & (M_BCAST|M_MCAST) || 301 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 302 (type != ICMP6_PACKET_TOO_BIG && 303 (type != ICMP6_PARAM_PROB || 304 code != ICMP6_PARAMPROB_OPTION))) 305 goto freeit; 306 307 /* 308 * RFC 2463, 2.4 (e.5): source address check. 309 * XXX: the case of anycast source? 310 */ 311 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 312 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 313 goto freeit; 314 315 /* 316 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 317 * don't do it. 318 */ 319 nxt = -1; 320 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 321 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 322 struct icmp6_hdr *icp; 323 324 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 325 sizeof(*icp)); 326 if (icp == NULL) { 327 icmp6stat.icp6s_tooshort++; 328 return; 329 } 330 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 331 icp->icmp6_type == ND_REDIRECT) { 332 /* 333 * ICMPv6 error 334 * Special case: for redirect (which is 335 * informational) we must not send icmp6 error. 336 */ 337 icmp6stat.icp6s_canterror++; 338 goto freeit; 339 } else { 340 /* ICMPv6 informational - send the error */ 341 } 342 } 343#if 0 /* controversial */ 344 else if (off >= 0 && nxt == IPPROTO_ESP) { 345 /* 346 * It could be ICMPv6 error inside ESP. Take a safer side, 347 * don't respond. 348 */ 349 icmp6stat.icp6s_canterror++; 350 goto freeit; 351 } 352#endif 353 else { 354 /* non-ICMPv6 - send the error */ 355 } 356 357 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 358 359 /* Finally, do rate limitation check. */ 360 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 361 icmp6stat.icp6s_toofreq++; 362 goto freeit; 363 } 364 365 /* 366 * OK, ICMP6 can be generated. 367 */ 368 369 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 370 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 371 372 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 373 M_PREPEND(m, preplen, M_DONTWAIT); 374 if (m && m->m_len < preplen) 375 m = m_pullup(m, preplen); 376 if (m == NULL) { 377 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 378 return; 379 } 380 381 nip6 = mtod(m, struct ip6_hdr *); 382 nip6->ip6_src = oip6->ip6_src; 383 nip6->ip6_dst = oip6->ip6_dst; 384 385 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src)) 386 oip6->ip6_src.s6_addr16[1] = 0; 387 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst)) 388 oip6->ip6_dst.s6_addr16[1] = 0; 389 390 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 391 icmp6->icmp6_type = type; 392 icmp6->icmp6_code = code; 393 icmp6->icmp6_pptr = htonl((u_int32_t)param); 394 395 /* 396 * icmp6_reflect() is designed to be in the input path. 397 * icmp6_error() can be called from both input and outut path, 398 * and if we are in output path rcvif could contain bogus value. 399 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 400 * information in ip header (nip6). 401 */ 402 m->m_pkthdr.rcvif = NULL; 403 404 icmp6stat.icp6s_outhist[type]++; 405 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 406 407 return; 408 409 freeit: 410 /* 411 * If we can't tell wheter or not we can generate ICMP6, free it. 412 */ 413 m_freem(m); 414} 415 416/* 417 * Process a received ICMP6 message. 418 */ 419int 420icmp6_input(mp, offp, proto) 421 struct mbuf **mp; 422 int *offp, proto; 423{ 424 struct mbuf *m = *mp, *n; 425 struct ip6_hdr *ip6, *nip6; 426 struct icmp6_hdr *icmp6, *nicmp6; 427 int off = *offp; 428 int icmp6len = m->m_pkthdr.len - *offp; 429 int code, sum, noff; 430 431 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 432 433 /* 434 * Locate icmp6 structure in mbuf, and check 435 * that not corrupted and of at least minimum length 436 */ 437 438 ip6 = mtod(m, struct ip6_hdr *); 439 if (icmp6len < sizeof(struct icmp6_hdr)) { 440 icmp6stat.icp6s_tooshort++; 441 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 442 goto freeit; 443 } 444 445 /* 446 * calculate the checksum 447 */ 448 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 449 if (icmp6 == NULL) { 450 icmp6stat.icp6s_tooshort++; 451 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 452 return IPPROTO_DONE; 453 } 454 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 455 code = icmp6->icmp6_code; 456 457 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 458 nd6log((LOG_ERR, 459 "ICMP6 checksum error(%d|%x) %s\n", 460 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 461 icmp6stat.icp6s_checksum++; 462 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 463 goto freeit; 464 } 465 466#if defined(NFAITH) && 0 < NFAITH 467 if (faithprefix(&ip6->ip6_dst)) { 468 /* 469 * Deliver very specific ICMP6 type only. 470 * This is important to deilver TOOBIG. Otherwise PMTUD 471 * will not work. 472 */ 473 switch (icmp6->icmp6_type) { 474 case ICMP6_DST_UNREACH: 475 case ICMP6_PACKET_TOO_BIG: 476 case ICMP6_TIME_EXCEEDED: 477 break; 478 default: 479 goto freeit; 480 } 481 } 482#endif 483 484 icmp6stat.icp6s_inhist[icmp6->icmp6_type]++; 485 486 switch (icmp6->icmp6_type) { 487 case ICMP6_DST_UNREACH: 488 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 489 switch (code) { 490 case ICMP6_DST_UNREACH_NOROUTE: 491 code = PRC_UNREACH_NET; 492 break; 493 case ICMP6_DST_UNREACH_ADMIN: 494 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 495 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 496 break; 497 case ICMP6_DST_UNREACH_ADDR: 498 code = PRC_HOSTDEAD; 499 break; 500#ifdef COMPAT_RFC1885 501 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 502 code = PRC_UNREACH_SRCFAIL; 503 break; 504#else 505 case ICMP6_DST_UNREACH_BEYONDSCOPE: 506 /* I mean "source address was incorrect." */ 507 code = PRC_UNREACH_NET; 508 break; 509#endif 510 case ICMP6_DST_UNREACH_NOPORT: 511 code = PRC_UNREACH_PORT; 512 break; 513 default: 514 goto badcode; 515 } 516 goto deliver; 517 518 case ICMP6_PACKET_TOO_BIG: 519 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 520 if (code != 0) 521 goto badcode; 522 523 code = PRC_MSGSIZE; 524 525 /* 526 * Updating the path MTU will be done after examining 527 * intermediate extension headers. 528 */ 529 goto deliver; 530 531 case ICMP6_TIME_EXCEEDED: 532 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 533 switch (code) { 534 case ICMP6_TIME_EXCEED_TRANSIT: 535 code = PRC_TIMXCEED_INTRANS; 536 break; 537 case ICMP6_TIME_EXCEED_REASSEMBLY: 538 code = PRC_TIMXCEED_REASS; 539 break; 540 default: 541 goto badcode; 542 } 543 goto deliver; 544 545 case ICMP6_PARAM_PROB: 546 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 547 switch (code) { 548 case ICMP6_PARAMPROB_NEXTHEADER: 549 code = PRC_UNREACH_PROTOCOL; 550 break; 551 case ICMP6_PARAMPROB_HEADER: 552 case ICMP6_PARAMPROB_OPTION: 553 code = PRC_PARAMPROB; 554 break; 555 default: 556 goto badcode; 557 } 558 goto deliver; 559 560 case ICMP6_ECHO_REQUEST: 561 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 562 if (code != 0) 563 goto badcode; 564 /* 565 * Copy mbuf to send to two data paths: userland socket(s), 566 * and to the querier (echo reply). 567 * m: a copy for socket, n: a copy for querier 568 */ 569 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 570 /* Give up local */ 571 n = m; 572 m = NULL; 573 goto deliverecho; 574 } 575 /* 576 * If the first mbuf is shared, or the first mbuf is too short, 577 * copy the first part of the data into a fresh mbuf. 578 * Otherwise, we will wrongly overwrite both copies. 579 */ 580 if ((n->m_flags & M_EXT) != 0 || 581 n->m_len < off + sizeof(struct icmp6_hdr)) { 582 struct mbuf *n0 = n; 583 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6); 584 585 /* 586 * Prepare an internal mbuf. m_pullup() doesn't 587 * always copy the length we specified. 588 */ 589 if (maxlen >= MCLBYTES) { 590 /* Give up remote */ 591 m_freem(n0); 592 break; 593 } 594 MGETHDR(n, M_DONTWAIT, n0->m_type); 595 if (n && maxlen >= MHLEN) { 596 MCLGET(n, M_DONTWAIT); 597 if ((n->m_flags & M_EXT) == 0) { 598 m_free(n); 599 n = NULL; 600 } 601 } 602 if (n == NULL) { 603 /* Give up local */ 604 m_freem(n0); 605 n = m; 606 m = NULL; 607 goto deliverecho; 608 } 609 M_COPY_PKTHDR(n, n0); 610 /* 611 * Copy IPv6 and ICMPv6 only. 612 */ 613 nip6 = mtod(n, struct ip6_hdr *); 614 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 615 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 616 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 617 noff = sizeof(struct ip6_hdr); 618 n->m_len = noff + sizeof(struct icmp6_hdr); 619 /* 620 * Adjust mbuf. ip6_plen will be adjusted in 621 * ip6_output(). 622 * n->m_pkthdr.len == n0->m_pkthdr.len at this point. 623 */ 624 n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr); 625 n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr)); 626 m_adj(n0, off + sizeof(struct icmp6_hdr)); 627 n->m_next = n0; 628 n0->m_flags &= ~M_PKTHDR; 629 } else { 630 deliverecho: 631 nip6 = mtod(n, struct ip6_hdr *); 632 nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off); 633 noff = off; 634 } 635 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 636 nicmp6->icmp6_code = 0; 637 if (n) { 638 icmp6stat.icp6s_reflect++; 639 icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++; 640 icmp6_reflect(n, noff); 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 MLD6_LISTENER_QUERY: 653 case MLD6_LISTENER_REPORT: 654 if (icmp6len < sizeof(struct mld6_hdr)) 655 goto badlen; 656 if (icmp6->icmp6_type == MLD6_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 mld6_input(m, off); 663 m = NULL; 664 goto freeit; 665 } 666 mld6_input(n, off); 667 /* m stays. */ 668 break; 669 670 case MLD6_LISTENER_DONE: 671 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 672 if (icmp6len < sizeof(struct mld6_hdr)) /* necessary? */ 673 goto badlen; 674 break; /* nothing to be done in kernel */ 675 676 case MLD6_MTRACE_RESP: 677 case MLD6_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 maxlen, maxhlen; 705 706 if ((icmp6_nodeinfo & 5) != 5) 707 break; 708 709 if (code != 0) 710 goto badcode; 711 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; 712 if (maxlen >= MCLBYTES) { 713 /* Give up remote */ 714 break; 715 } 716 MGETHDR(n, M_DONTWAIT, m->m_type); 717 if (n && maxlen > MHLEN) { 718 MCLGET(n, M_DONTWAIT); 719 if ((n->m_flags & M_EXT) == 0) { 720 m_free(n); 721 n = NULL; 722 } 723 } 724 if (n == NULL) { 725 /* Give up remote */ 726 break; 727 } 728 n->m_pkthdr.rcvif = NULL; 729 n->m_len = 0; 730 maxhlen = M_TRAILINGSPACE(n) - maxlen; 731 if (maxhlen > hostnamelen) 732 maxhlen = hostnamelen; 733 /* 734 * Copy IPv6 and ICMPv6 only. 735 */ 736 nip6 = mtod(n, struct ip6_hdr *); 737 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 738 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 739 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 740 p = (u_char *)(nicmp6 + 1); 741 bzero(p, 4); 742 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 743 noff = sizeof(struct ip6_hdr); 744 M_COPY_PKTHDR(n, m); /* just for rcvif */ 745 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 746 sizeof(struct icmp6_hdr) + 4 + maxhlen; 747 nicmp6->icmp6_type = ICMP6_WRUREPLY; 748 nicmp6->icmp6_code = 0; 749 } 750#undef hostnamelen 751 if (n) { 752 icmp6stat.icp6s_reflect++; 753 icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++; 754 icmp6_reflect(n, noff); 755 } 756 break; 757 } 758 759 case ICMP6_WRUREPLY: 760 if (code != 0) 761 goto badcode; 762 break; 763 764 case ND_ROUTER_SOLICIT: 765 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 766 if (code != 0) 767 goto badcode; 768 if (icmp6len < sizeof(struct nd_router_solicit)) 769 goto badlen; 770 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 771 /* give up local */ 772 nd6_rs_input(m, off, icmp6len); 773 m = NULL; 774 goto freeit; 775 } 776 nd6_rs_input(n, off, icmp6len); 777 /* m stays. */ 778 break; 779 780 case ND_ROUTER_ADVERT: 781 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 782 if (code != 0) 783 goto badcode; 784 if (icmp6len < sizeof(struct nd_router_advert)) 785 goto badlen; 786 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 787 /* give up local */ 788 nd6_ra_input(m, off, icmp6len); 789 m = NULL; 790 goto freeit; 791 } 792 nd6_ra_input(n, off, icmp6len); 793 /* m stays. */ 794 break; 795 796 case ND_NEIGHBOR_SOLICIT: 797 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 798 if (code != 0) 799 goto badcode; 800 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 801 goto badlen; 802 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 803 /* give up local */ 804 nd6_ns_input(m, off, icmp6len); 805 m = NULL; 806 goto freeit; 807 } 808 nd6_ns_input(n, off, icmp6len); 809 /* m stays. */ 810 break; 811 812 case ND_NEIGHBOR_ADVERT: 813 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 814 if (code != 0) 815 goto badcode; 816 if (icmp6len < sizeof(struct nd_neighbor_advert)) 817 goto badlen; 818 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 819 /* give up local */ 820 nd6_na_input(m, off, icmp6len); 821 m = NULL; 822 goto freeit; 823 } 824 nd6_na_input(n, off, icmp6len); 825 /* m stays. */ 826 break; 827 828 case ND_REDIRECT: 829 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 830 if (code != 0) 831 goto badcode; 832 if (icmp6len < sizeof(struct nd_redirect)) 833 goto badlen; 834 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 835 /* give up local */ 836 icmp6_redirect_input(m, off); 837 m = NULL; 838 goto freeit; 839 } 840 icmp6_redirect_input(n, off); 841 /* m stays. */ 842 break; 843 844 case ICMP6_ROUTER_RENUMBERING: 845 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 846 code != ICMP6_ROUTER_RENUMBERING_RESULT) 847 goto badcode; 848 if (icmp6len < sizeof(struct icmp6_router_renum)) 849 goto badlen; 850 break; 851 852 default: 853 nd6log((LOG_DEBUG, 854 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 855 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 856 ip6_sprintf(&ip6->ip6_dst), 857 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 858 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 859 /* ICMPv6 error: MUST deliver it by spec... */ 860 code = PRC_NCMDS; 861 /* deliver */ 862 } else { 863 /* ICMPv6 informational: MUST not deliver */ 864 break; 865 } 866 deliver: 867 if (icmp6_notify_error(m, off, icmp6len, code)) { 868 /* In this case, m should've been freed. */ 869 return (IPPROTO_DONE); 870 } 871 break; 872 873 badcode: 874 icmp6stat.icp6s_badcode++; 875 break; 876 877 badlen: 878 icmp6stat.icp6s_badlen++; 879 break; 880 } 881 882 /* deliver the packet to appropriate sockets */ 883 icmp6_rip6_input(&m, *offp); 884 885 return IPPROTO_DONE; 886 887 freeit: 888 m_freem(m); 889 return IPPROTO_DONE; 890} 891 892static int 893icmp6_notify_error(m, off, icmp6len, code) 894 struct mbuf *m; 895 int off, icmp6len; 896{ 897 struct icmp6_hdr *icmp6; 898 struct ip6_hdr *eip6; 899 u_int32_t notifymtu; 900 struct sockaddr_in6 icmp6src, icmp6dst; 901 902 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 903 icmp6stat.icp6s_tooshort++; 904 goto freeit; 905 } 906 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 907 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 908 if (icmp6 == NULL) { 909 icmp6stat.icp6s_tooshort++; 910 return (-1); 911 } 912 eip6 = (struct ip6_hdr *)(icmp6 + 1); 913 914 /* Detect the upper level protocol */ 915 { 916 void (*ctlfunc) __P((int, struct sockaddr *, void *)); 917 u_int8_t nxt = eip6->ip6_nxt; 918 int eoff = off + sizeof(struct icmp6_hdr) + 919 sizeof(struct ip6_hdr); 920 struct ip6ctlparam ip6cp; 921 struct in6_addr *finaldst = NULL; 922 int icmp6type = icmp6->icmp6_type; 923 struct ip6_frag *fh; 924 struct ip6_rthdr *rth; 925 struct ip6_rthdr0 *rth0; 926 int rthlen; 927 928 while (1) { /* XXX: should avoid infinite loop explicitly? */ 929 struct ip6_ext *eh; 930 931 switch (nxt) { 932 case IPPROTO_HOPOPTS: 933 case IPPROTO_DSTOPTS: 934 case IPPROTO_AH: 935 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 936 eoff, sizeof(*eh)); 937 if (eh == NULL) { 938 icmp6stat.icp6s_tooshort++; 939 return (-1); 940 } 941 942 if (nxt == IPPROTO_AH) 943 eoff += (eh->ip6e_len + 2) << 2; 944 else 945 eoff += (eh->ip6e_len + 1) << 3; 946 nxt = eh->ip6e_nxt; 947 break; 948 case IPPROTO_ROUTING: 949 /* 950 * When the erroneous packet contains a 951 * routing header, we should examine the 952 * header to determine the final destination. 953 * Otherwise, we can't properly update 954 * information that depends on the final 955 * destination (e.g. path MTU). 956 */ 957 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 958 eoff, sizeof(*rth)); 959 if (rth == NULL) { 960 icmp6stat.icp6s_tooshort++; 961 return (-1); 962 } 963 rthlen = (rth->ip6r_len + 1) << 3; 964 /* 965 * XXX: currently there is no 966 * officially defined type other 967 * than type-0. 968 * Note that if the segment left field 969 * is 0, all intermediate hops must 970 * have been passed. 971 */ 972 if (rth->ip6r_segleft && 973 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 974 int hops; 975 976 IP6_EXTHDR_GET(rth0, 977 struct ip6_rthdr0 *, m, 978 eoff, rthlen); 979 if (rth0 == NULL) { 980 icmp6stat.icp6s_tooshort++; 981 return (-1); 982 } 983 /* just ignore a bogus header */ 984 if ((rth0->ip6r0_len % 2) == 0 && 985 (hops = rth0->ip6r0_len/2)) 986 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 987 } 988 eoff += rthlen; 989 nxt = rth->ip6r_nxt; 990 break; 991 case IPPROTO_FRAGMENT: 992 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 993 eoff, sizeof(*fh)); 994 if (fh == NULL) { 995 icmp6stat.icp6s_tooshort++; 996 return (-1); 997 } 998 /* 999 * Data after a fragment header is meaningless 1000 * unless it is the first fragment, but 1001 * we'll go to the notify label for path MTU 1002 * discovery. 1003 */ 1004 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1005 goto notify; 1006 1007 eoff += sizeof(struct ip6_frag); 1008 nxt = fh->ip6f_nxt; 1009 break; 1010 default: 1011 /* 1012 * This case includes ESP and the No Next 1013 * Header. In such cases going to the notify 1014 * label does not have any meaning 1015 * (i.e. ctlfunc will be NULL), but we go 1016 * anyway since we might have to update 1017 * path MTU information. 1018 */ 1019 goto notify; 1020 } 1021 } 1022 notify: 1023 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1024 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1025 if (icmp6 == NULL) { 1026 icmp6stat.icp6s_tooshort++; 1027 return (-1); 1028 } 1029 1030 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1031 bzero(&icmp6dst, sizeof(icmp6dst)); 1032 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1033 icmp6dst.sin6_family = AF_INET6; 1034 if (finaldst == NULL) 1035 icmp6dst.sin6_addr = eip6->ip6_dst; 1036 else 1037 icmp6dst.sin6_addr = *finaldst; 1038 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1039 &icmp6dst.sin6_addr); 1040#ifndef SCOPEDROUTING 1041 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst, 1042 NULL, NULL)) { 1043 /* should be impossbile */ 1044 nd6log((LOG_DEBUG, 1045 "icmp6_notify_error: in6_embedscope failed\n")); 1046 goto freeit; 1047 } 1048#endif 1049 1050 /* 1051 * retrieve parameters from the inner IPv6 header, and convert 1052 * them into sockaddr structures. 1053 */ 1054 bzero(&icmp6src, sizeof(icmp6src)); 1055 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1056 icmp6src.sin6_family = AF_INET6; 1057 icmp6src.sin6_addr = eip6->ip6_src; 1058 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1059 &icmp6src.sin6_addr); 1060#ifndef SCOPEDROUTING 1061 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src, 1062 NULL, NULL)) { 1063 /* should be impossbile */ 1064 nd6log((LOG_DEBUG, 1065 "icmp6_notify_error: in6_embedscope failed\n")); 1066 goto freeit; 1067 } 1068#endif 1069 icmp6src.sin6_flowinfo = 1070 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1071 1072 if (finaldst == NULL) 1073 finaldst = &eip6->ip6_dst; 1074 ip6cp.ip6c_m = m; 1075 ip6cp.ip6c_icmp6 = icmp6; 1076 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1077 ip6cp.ip6c_off = eoff; 1078 ip6cp.ip6c_finaldst = finaldst; 1079 ip6cp.ip6c_src = &icmp6src; 1080 ip6cp.ip6c_nxt = nxt; 1081 1082 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1083 notifymtu = ntohl(icmp6->icmp6_mtu); 1084 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1085 } 1086 1087 ctlfunc = (void (*) __P((int, struct sockaddr *, void *))) 1088 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1089 if (ctlfunc) { 1090 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1091 &ip6cp); 1092 } 1093 } 1094 return (0); 1095 1096 freeit: 1097 m_freem(m); 1098 return (-1); 1099} 1100 1101void 1102icmp6_mtudisc_update(ip6cp, validated) 1103 struct ip6ctlparam *ip6cp; 1104 int validated; 1105{ 1106 unsigned long rtcount; 1107 struct icmp6_mtudisc_callback *mc; 1108 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1109 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1110 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1111 u_int mtu = ntohl(icmp6->icmp6_mtu); 1112 struct rtentry *rt = NULL; 1113 struct sockaddr_in6 sin6; 1114 1115 /* 1116 * allow non-validated cases if memory is plenty, to make traffic 1117 * from non-connected pcb happy. 1118 */ 1119 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1120 if (validated) { 1121 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1122 return; 1123 else if (0 <= icmp6_mtudisc_lowat && 1124 rtcount > icmp6_mtudisc_lowat) { 1125 /* 1126 * XXX nuke a victim, install the new one. 1127 */ 1128 } 1129 } else { 1130 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1131 return; 1132 } 1133 1134 bzero(&sin6, sizeof(sin6)); 1135 sin6.sin6_family = PF_INET6; 1136 sin6.sin6_len = sizeof(struct sockaddr_in6); 1137 sin6.sin6_addr = *dst; 1138 /* XXX normally, this won't happen */ 1139 if (IN6_IS_ADDR_LINKLOCAL(dst)) { 1140 sin6.sin6_addr.s6_addr16[1] = 1141 htons(m->m_pkthdr.rcvif->if_index); 1142 } 1143 /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */ 1144 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1145 1146 if (rt && (rt->rt_flags & RTF_HOST) && 1147 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1148 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1149 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1150 icmp6stat.icp6s_pmtuchg++; 1151 rt->rt_rmx.rmx_mtu = mtu; 1152 } 1153 } 1154 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */ 1155 RTFREE(rt); 1156 } 1157 1158 /* 1159 * Notify protocols that the MTU for this destination 1160 * has changed. 1161 */ 1162 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1163 mc = LIST_NEXT(mc, mc_list)) 1164 (*mc->mc_func)(&sin6.sin6_addr); 1165} 1166 1167/* 1168 * Process a Node Information Query packet, based on 1169 * draft-ietf-ipngwg-icmp-name-lookups-07. 1170 * 1171 * Spec incompatibilities: 1172 * - IPv6 Subject address handling 1173 * - IPv4 Subject address handling support missing 1174 * - Proxy reply (answer even if it's not for me) 1175 * - joins NI group address at in6_ifattach() time only, does not cope 1176 * with hostname changes by sethostname(3) 1177 */ 1178#ifndef offsetof /* XXX */ 1179#define offsetof(type, member) ((size_t)(&((type *)0)->member)) 1180#endif 1181static struct mbuf * 1182ni6_input(m, off) 1183 struct mbuf *m; 1184 int off; 1185{ 1186 struct icmp6_nodeinfo *ni6, *nni6; 1187 struct mbuf *n = NULL; 1188 u_int16_t qtype; 1189 int subjlen; 1190 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1191 struct ni_reply_fqdn *fqdn; 1192 int addrs; /* for NI_QTYPE_NODEADDR */ 1193 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1194 struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */ 1195 struct ip6_hdr *ip6; 1196 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1197 char *subj = NULL; 1198 1199 ip6 = mtod(m, struct ip6_hdr *); 1200 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1201 if (ni6 == NULL) { 1202 /* m is already reclaimed */ 1203 return NULL; 1204 } 1205 1206 /* 1207 * Validate IPv6 destination address. 1208 * 1209 * The Responder must discard the Query without further processing 1210 * unless it is one of the Responder's unicast or anycast addresses, or 1211 * a link-local scope multicast address which the Responder has joined. 1212 * [icmp-name-lookups-07, Section 4.] 1213 */ 1214 bzero(&sin6, sizeof(sin6)); 1215 sin6.sin6_family = AF_INET6; 1216 sin6.sin6_len = sizeof(struct sockaddr_in6); 1217 bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr)); 1218 /* XXX scopeid */ 1219 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1220 ; /* unicast/anycast, fine */ 1221 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1222 ; /* link-local multicast, fine */ 1223 else 1224 goto bad; 1225 1226 /* validate query Subject field. */ 1227 qtype = ntohs(ni6->ni_qtype); 1228 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1229 switch (qtype) { 1230 case NI_QTYPE_NOOP: 1231 case NI_QTYPE_SUPTYPES: 1232 /* 07 draft */ 1233 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1234 break; 1235 /* FALLTHROUGH */ 1236 case NI_QTYPE_FQDN: 1237 case NI_QTYPE_NODEADDR: 1238 switch (ni6->ni_code) { 1239 case ICMP6_NI_SUBJ_IPV6: 1240#if ICMP6_NI_SUBJ_IPV6 != 0 1241 case 0: 1242#endif 1243 /* 1244 * backward compatibility - try to accept 03 draft 1245 * format, where no Subject is present. 1246 */ 1247 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1248 subjlen == 0) { 1249 oldfqdn++; 1250 break; 1251 } 1252#if ICMP6_NI_SUBJ_IPV6 != 0 1253 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1254 goto bad; 1255#endif 1256 1257 if (subjlen != sizeof(sin6.sin6_addr)) 1258 goto bad; 1259 1260 /* 1261 * Validate Subject address. 1262 * 1263 * Not sure what exactly "address belongs to the node" 1264 * means in the spec, is it just unicast, or what? 1265 * 1266 * At this moment we consider Subject address as 1267 * "belong to the node" if the Subject address equals 1268 * to the IPv6 destination address; validation for 1269 * IPv6 destination address should have done enough 1270 * check for us. 1271 * 1272 * We do not do proxy at this moment. 1273 */ 1274 /* m_pulldown instead of copy? */ 1275 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1276 subjlen, (caddr_t)&sin6.sin6_addr); 1277 /* XXX kame scope hack */ 1278 if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) { 1279 if ((m->m_flags & M_PKTHDR) != 0 && 1280 m->m_pkthdr.rcvif) { 1281 sin6.sin6_addr.s6_addr16[1] = 1282 htons(m->m_pkthdr.rcvif->if_index); 1283 } 1284 } 1285 subj = (char *)&sin6; 1286 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr)) 1287 break; 1288 1289 /* 1290 * XXX if we are to allow other cases, we should really 1291 * be careful about scope here. 1292 * basically, we should disallow queries toward IPv6 1293 * destination X with subject Y, if scope(X) > scope(Y). 1294 * if we allow scope(X) > scope(Y), it will result in 1295 * information leakage across scope boundary. 1296 */ 1297 goto bad; 1298 1299 case ICMP6_NI_SUBJ_FQDN: 1300 /* 1301 * Validate Subject name with gethostname(3). 1302 * 1303 * The behavior may need some debate, since: 1304 * - we are not sure if the node has FQDN as 1305 * hostname (returned by gethostname(3)). 1306 * - the code does wildcard match for truncated names. 1307 * however, we are not sure if we want to perform 1308 * wildcard match, if gethostname(3) side has 1309 * truncated hostname. 1310 */ 1311 n = ni6_nametodns(hostname, hostnamelen, 0); 1312 if (!n || n->m_next || n->m_len == 0) 1313 goto bad; 1314 IP6_EXTHDR_GET(subj, char *, m, 1315 off + sizeof(struct icmp6_nodeinfo), subjlen); 1316 if (subj == NULL) 1317 goto bad; 1318 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1319 n->m_len)) { 1320 goto bad; 1321 } 1322 m_freem(n); 1323 n = NULL; 1324 break; 1325 1326 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1327 default: 1328 goto bad; 1329 } 1330 break; 1331 } 1332 1333 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1334 switch (qtype) { 1335 case NI_QTYPE_FQDN: 1336 if ((icmp6_nodeinfo & 1) == 0) 1337 goto bad; 1338 break; 1339 case NI_QTYPE_NODEADDR: 1340 if ((icmp6_nodeinfo & 2) == 0) 1341 goto bad; 1342 break; 1343 } 1344 1345 /* guess reply length */ 1346 switch (qtype) { 1347 case NI_QTYPE_NOOP: 1348 break; /* no reply data */ 1349 case NI_QTYPE_SUPTYPES: 1350 replylen += sizeof(u_int32_t); 1351 break; 1352 case NI_QTYPE_FQDN: 1353 /* XXX will append an mbuf */ 1354 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1355 break; 1356 case NI_QTYPE_NODEADDR: 1357 addrs = ni6_addrs(ni6, m, &ifp, subj); 1358 if ((replylen += addrs * (sizeof(struct in6_addr) + 1359 sizeof(u_int32_t))) > MCLBYTES) 1360 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1361 break; 1362 default: 1363 /* 1364 * XXX: We must return a reply with the ICMP6 code 1365 * `unknown Qtype' in this case. However we regard the case 1366 * as an FQDN query for backward compatibility. 1367 * Older versions set a random value to this field, 1368 * so it rarely varies in the defined qtypes. 1369 * But the mechanism is not reliable... 1370 * maybe we should obsolete older versions. 1371 */ 1372 qtype = NI_QTYPE_FQDN; 1373 /* XXX will append an mbuf */ 1374 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1375 oldfqdn++; 1376 break; 1377 } 1378 1379 /* allocate an mbuf to reply. */ 1380 MGETHDR(n, M_DONTWAIT, m->m_type); 1381 if (n == NULL) { 1382 m_freem(m); 1383 return (NULL); 1384 } 1385 M_COPY_PKTHDR(n, m); /* just for rcvif */ 1386 if (replylen > MHLEN) { 1387 if (replylen > MCLBYTES) { 1388 /* 1389 * XXX: should we try to allocate more? But MCLBYTES 1390 * is probably much larger than IPV6_MMTU... 1391 */ 1392 goto bad; 1393 } 1394 MCLGET(n, M_DONTWAIT); 1395 if ((n->m_flags & M_EXT) == 0) { 1396 goto bad; 1397 } 1398 } 1399 n->m_pkthdr.len = n->m_len = replylen; 1400 1401 /* copy mbuf header and IPv6 + Node Information base headers */ 1402 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1403 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1404 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1405 1406 /* qtype dependent procedure */ 1407 switch (qtype) { 1408 case NI_QTYPE_NOOP: 1409 nni6->ni_code = ICMP6_NI_SUCCESS; 1410 nni6->ni_flags = 0; 1411 break; 1412 case NI_QTYPE_SUPTYPES: 1413 { 1414 u_int32_t v; 1415 nni6->ni_code = ICMP6_NI_SUCCESS; 1416 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1417 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1418 v = (u_int32_t)htonl(0x0000000f); 1419 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1420 break; 1421 } 1422 case NI_QTYPE_FQDN: 1423 nni6->ni_code = ICMP6_NI_SUCCESS; 1424 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1425 sizeof(struct ip6_hdr) + 1426 sizeof(struct icmp6_nodeinfo)); 1427 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1428 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1429 /* 1430 * XXX do we really have FQDN in variable "hostname"? 1431 */ 1432 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1433 if (n->m_next == NULL) 1434 goto bad; 1435 /* XXX we assume that n->m_next is not a chain */ 1436 if (n->m_next->m_next != NULL) 1437 goto bad; 1438 n->m_pkthdr.len += n->m_next->m_len; 1439 break; 1440 case NI_QTYPE_NODEADDR: 1441 { 1442 int lenlim, copied; 1443 1444 nni6->ni_code = ICMP6_NI_SUCCESS; 1445 n->m_pkthdr.len = n->m_len = 1446 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1447 lenlim = M_TRAILINGSPACE(n); 1448 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1449 /* XXX: reset mbuf length */ 1450 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1451 sizeof(struct icmp6_nodeinfo) + copied; 1452 break; 1453 } 1454 default: 1455 break; /* XXX impossible! */ 1456 } 1457 1458 nni6->ni_type = ICMP6_NI_REPLY; 1459 m_freem(m); 1460 return (n); 1461 1462 bad: 1463 m_freem(m); 1464 if (n) 1465 m_freem(n); 1466 return (NULL); 1467} 1468#undef hostnamelen 1469 1470#define isupper(x) ('A' <= (x) && (x) <= 'Z') 1471#define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1472#define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1473#define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1474 1475/* 1476 * make a mbuf with DNS-encoded string. no compression support. 1477 * 1478 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1479 * treated as truncated name (two \0 at the end). this is a wild guess. 1480 */ 1481static struct mbuf * 1482ni6_nametodns(name, namelen, old) 1483 const char *name; 1484 int namelen; 1485 int old; /* return pascal string if non-zero */ 1486{ 1487 struct mbuf *m; 1488 char *cp, *ep; 1489 const char *p, *q; 1490 int i, len, nterm; 1491 1492 if (old) 1493 len = namelen + 1; 1494 else 1495 len = MCLBYTES; 1496 1497 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1498 MGET(m, M_DONTWAIT, MT_DATA); 1499 if (m && len > MLEN) { 1500 MCLGET(m, M_DONTWAIT); 1501 if ((m->m_flags & M_EXT) == 0) 1502 goto fail; 1503 } 1504 if (!m) 1505 goto fail; 1506 m->m_next = NULL; 1507 1508 if (old) { 1509 m->m_len = len; 1510 *mtod(m, char *) = namelen; 1511 bcopy(name, mtod(m, char *) + 1, namelen); 1512 return m; 1513 } else { 1514 m->m_len = 0; 1515 cp = mtod(m, char *); 1516 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1517 1518 /* if not certain about my name, return empty buffer */ 1519 if (namelen == 0) 1520 return m; 1521 1522 /* 1523 * guess if it looks like shortened hostname, or FQDN. 1524 * shortened hostname needs two trailing "\0". 1525 */ 1526 i = 0; 1527 for (p = name; p < name + namelen; p++) { 1528 if (*p && *p == '.') 1529 i++; 1530 } 1531 if (i < 2) 1532 nterm = 2; 1533 else 1534 nterm = 1; 1535 1536 p = name; 1537 while (cp < ep && p < name + namelen) { 1538 i = 0; 1539 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1540 i++; 1541 /* result does not fit into mbuf */ 1542 if (cp + i + 1 >= ep) 1543 goto fail; 1544 /* 1545 * DNS label length restriction, RFC1035 page 8. 1546 * "i == 0" case is included here to avoid returning 1547 * 0-length label on "foo..bar". 1548 */ 1549 if (i <= 0 || i >= 64) 1550 goto fail; 1551 *cp++ = i; 1552 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1553 goto fail; 1554 while (i > 0) { 1555 if (!isalnum(*p) && *p != '-') 1556 goto fail; 1557 if (isupper(*p)) { 1558 *cp++ = tolower(*p); 1559 p++; 1560 } else 1561 *cp++ = *p++; 1562 i--; 1563 } 1564 p = q; 1565 if (p < name + namelen && *p == '.') 1566 p++; 1567 } 1568 /* termination */ 1569 if (cp + nterm >= ep) 1570 goto fail; 1571 while (nterm-- > 0) 1572 *cp++ = '\0'; 1573 m->m_len = cp - mtod(m, char *); 1574 return m; 1575 } 1576 1577 panic("should not reach here"); 1578 /* NOTREACHED */ 1579 1580 fail: 1581 if (m) 1582 m_freem(m); 1583 return NULL; 1584} 1585 1586/* 1587 * check if two DNS-encoded string matches. takes care of truncated 1588 * form (with \0\0 at the end). no compression support. 1589 * XXX upper/lowercase match (see RFC2065) 1590 */ 1591static int 1592ni6_dnsmatch(a, alen, b, blen) 1593 const char *a; 1594 int alen; 1595 const char *b; 1596 int blen; 1597{ 1598 const char *a0, *b0; 1599 int l; 1600 1601 /* simplest case - need validation? */ 1602 if (alen == blen && bcmp(a, b, alen) == 0) 1603 return 1; 1604 1605 a0 = a; 1606 b0 = b; 1607 1608 /* termination is mandatory */ 1609 if (alen < 2 || blen < 2) 1610 return 0; 1611 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1612 return 0; 1613 alen--; 1614 blen--; 1615 1616 while (a - a0 < alen && b - b0 < blen) { 1617 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1618 return 0; 1619 1620 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1621 return 0; 1622 /* we don't support compression yet */ 1623 if (a[0] >= 64 || b[0] >= 64) 1624 return 0; 1625 1626 /* truncated case */ 1627 if (a[0] == 0 && a - a0 == alen - 1) 1628 return 1; 1629 if (b[0] == 0 && b - b0 == blen - 1) 1630 return 1; 1631 if (a[0] == 0 || b[0] == 0) 1632 return 0; 1633 1634 if (a[0] != b[0]) 1635 return 0; 1636 l = a[0]; 1637 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1638 return 0; 1639 if (bcmp(a + 1, b + 1, l) != 0) 1640 return 0; 1641 1642 a += 1 + l; 1643 b += 1 + l; 1644 } 1645 1646 if (a - a0 == alen && b - b0 == blen) 1647 return 1; 1648 else 1649 return 0; 1650} 1651 1652/* 1653 * calculate the number of addresses to be returned in the node info reply. 1654 */ 1655static int 1656ni6_addrs(ni6, m, ifpp, subj) 1657 struct icmp6_nodeinfo *ni6; 1658 struct mbuf *m; 1659 struct ifnet **ifpp; 1660 char *subj; 1661{ 1662 struct ifnet *ifp; 1663 struct in6_ifaddr *ifa6; 1664 struct ifaddr *ifa; 1665 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1666 int addrs = 0, addrsofif, iffound = 0; 1667 int niflags = ni6->ni_flags; 1668 1669 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1670 switch (ni6->ni_code) { 1671 case ICMP6_NI_SUBJ_IPV6: 1672 if (subj == NULL) /* must be impossible... */ 1673 return (0); 1674 subj_ip6 = (struct sockaddr_in6 *)subj; 1675 break; 1676 default: 1677 /* 1678 * XXX: we only support IPv6 subject address for 1679 * this Qtype. 1680 */ 1681 return (0); 1682 } 1683 } 1684 1685 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1686 { 1687 addrsofif = 0; 1688 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1689 ifa = ifa->ifa_list.tqe_next) 1690 { 1691 if (ifa->ifa_addr->sa_family != AF_INET6) 1692 continue; 1693 ifa6 = (struct in6_ifaddr *)ifa; 1694 1695 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1696 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1697 &ifa6->ia_addr.sin6_addr)) 1698 iffound = 1; 1699 1700 /* 1701 * IPv4-mapped addresses can only be returned by a 1702 * Node Information proxy, since they represent 1703 * addresses of IPv4-only nodes, which perforce do 1704 * not implement this protocol. 1705 * [icmp-name-lookups-07, Section 5.4] 1706 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1707 * this function at this moment. 1708 */ 1709 1710 /* What do we have to do about ::1? */ 1711 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1712 case IPV6_ADDR_SCOPE_LINKLOCAL: 1713 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1714 continue; 1715 break; 1716 case IPV6_ADDR_SCOPE_SITELOCAL: 1717 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1718 continue; 1719 break; 1720 case IPV6_ADDR_SCOPE_GLOBAL: 1721 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1722 continue; 1723 break; 1724 default: 1725 continue; 1726 } 1727 1728 /* 1729 * check if anycast is okay. 1730 * XXX: just experimental. not in the spec. 1731 */ 1732 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1733 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1734 continue; /* we need only unicast addresses */ 1735 1736 addrsofif++; /* count the address */ 1737 } 1738 if (iffound) { 1739 *ifpp = ifp; 1740 return (addrsofif); 1741 } 1742 1743 addrs += addrsofif; 1744 } 1745 1746 return (addrs); 1747} 1748 1749static int 1750ni6_store_addrs(ni6, nni6, ifp0, resid) 1751 struct icmp6_nodeinfo *ni6, *nni6; 1752 struct ifnet *ifp0; 1753 int resid; 1754{ 1755 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1756 struct in6_ifaddr *ifa6; 1757 struct ifaddr *ifa; 1758 struct ifnet *ifp_dep = NULL; 1759 int copied = 0, allow_deprecated = 0; 1760 u_char *cp = (u_char *)(nni6 + 1); 1761 int niflags = ni6->ni_flags; 1762 u_int32_t ltime; 1763 long time_second = time.tv_sec; 1764 1765 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1766 return (0); /* needless to copy */ 1767 1768 again: 1769 1770 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1771 { 1772 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1773 ifa = ifa->ifa_list.tqe_next) 1774 { 1775 if (ifa->ifa_addr->sa_family != AF_INET6) 1776 continue; 1777 ifa6 = (struct in6_ifaddr *)ifa; 1778 1779 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1780 allow_deprecated == 0) { 1781 /* 1782 * prefererred address should be put before 1783 * deprecated addresses. 1784 */ 1785 1786 /* record the interface for later search */ 1787 if (ifp_dep == NULL) 1788 ifp_dep = ifp; 1789 1790 continue; 1791 } 1792 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1793 allow_deprecated != 0) 1794 continue; /* we now collect deprecated addrs */ 1795 1796 /* What do we have to do about ::1? */ 1797 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1798 case IPV6_ADDR_SCOPE_LINKLOCAL: 1799 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1800 continue; 1801 break; 1802 case IPV6_ADDR_SCOPE_SITELOCAL: 1803 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1804 continue; 1805 break; 1806 case IPV6_ADDR_SCOPE_GLOBAL: 1807 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1808 continue; 1809 break; 1810 default: 1811 continue; 1812 } 1813 1814 /* 1815 * check if anycast is okay. 1816 * XXX: just experimental. not in the spec. 1817 */ 1818 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1819 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1820 continue; 1821 1822 /* now we can copy the address */ 1823 if (resid < sizeof(struct in6_addr) + 1824 sizeof(u_int32_t)) { 1825 /* 1826 * We give up much more copy. 1827 * Set the truncate flag and return. 1828 */ 1829 nni6->ni_flags |= 1830 NI_NODEADDR_FLAG_TRUNCATE; 1831 return (copied); 1832 } 1833 1834 /* 1835 * Set the TTL of the address. 1836 * The TTL value should be one of the following 1837 * according to the specification: 1838 * 1839 * 1. The remaining lifetime of a DHCP lease on the 1840 * address, or 1841 * 2. The remaining Valid Lifetime of a prefix from 1842 * which the address was derived through Stateless 1843 * Autoconfiguration. 1844 * 1845 * Note that we currently do not support stateful 1846 * address configuration by DHCPv6, so the former 1847 * case can't happen. 1848 * 1849 * TTL must be 2^31 > TTL >= 0. 1850 */ 1851 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1852 ltime = ND6_INFINITE_LIFETIME; 1853 else { 1854 if (ifa6->ia6_lifetime.ia6t_expire > 1855 time_second) 1856 ltime = ifa6->ia6_lifetime.ia6t_expire - time_second; 1857 else 1858 ltime = 0; 1859 } 1860 if (ltime > 0x7fffffff) 1861 ltime = 0x7fffffff; 1862 ltime = htonl(ltime); 1863 1864 bcopy(<ime, cp, sizeof(u_int32_t)); 1865 cp += sizeof(u_int32_t); 1866 1867 /* copy the address itself */ 1868 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1869 sizeof(struct in6_addr)); 1870 /* XXX: KAME link-local hack; remove ifindex */ 1871 if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr)) 1872 ((struct in6_addr *)cp)->s6_addr16[1] = 0; 1873 cp += sizeof(struct in6_addr); 1874 1875 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1876 copied += (sizeof(struct in6_addr) + 1877 sizeof(u_int32_t)); 1878 } 1879 if (ifp0) /* we need search only on the specified IF */ 1880 break; 1881 } 1882 1883 if (allow_deprecated == 0 && ifp_dep != NULL) { 1884 ifp = ifp_dep; 1885 allow_deprecated = 1; 1886 1887 goto again; 1888 } 1889 1890 return (copied); 1891} 1892 1893/* 1894 * XXX almost dup'ed code with rip6_input. 1895 */ 1896static int 1897icmp6_rip6_input(mp, off) 1898 struct mbuf **mp; 1899 int off; 1900{ 1901 struct mbuf *m = *mp; 1902 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1903 struct in6pcb *in6p; 1904 struct in6pcb *last = NULL; 1905 struct sockaddr_in6 rip6src; 1906 struct icmp6_hdr *icmp6; 1907 struct mbuf *opts = NULL; 1908 1909 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1910 if (icmp6 == NULL) { 1911 /* m is already reclaimed */ 1912 return IPPROTO_DONE; 1913 } 1914 1915 bzero(&rip6src, sizeof(rip6src)); 1916 rip6src.sin6_len = sizeof(struct sockaddr_in6); 1917 rip6src.sin6_family = AF_INET6; 1918 /* KAME hack: recover scopeid */ 1919 (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif); 1920 1921 for (in6p = rawin6pcb.in6p_next; 1922 in6p != &rawin6pcb; in6p = in6p->in6p_next) 1923 { 1924 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6) 1925 continue; 1926 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1927 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1928 continue; 1929 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1930 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1931 continue; 1932 if (in6p->in6p_icmp6filt 1933 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1934 in6p->in6p_icmp6filt)) 1935 continue; 1936 if (last) { 1937 struct mbuf *n; 1938 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1939 if (last->in6p_flags & IN6P_CONTROLOPTS) 1940 ip6_savecontrol(last, &opts, ip6, n); 1941 /* strip intermediate headers */ 1942 m_adj(n, off); 1943 if (sbappendaddr(&last->in6p_socket->so_rcv, 1944 (struct sockaddr *)&rip6src, 1945 n, opts) == 0) { 1946 /* should notify about lost packet */ 1947 m_freem(n); 1948 if (opts) 1949 m_freem(opts); 1950 } else 1951 sorwakeup(last->in6p_socket); 1952 opts = NULL; 1953 } 1954 } 1955 last = in6p; 1956 } 1957 if (last) { 1958 if (last->in6p_flags & IN6P_CONTROLOPTS) 1959 ip6_savecontrol(last, &opts, ip6, m); 1960 /* strip intermediate headers */ 1961 m_adj(m, off); 1962 if (sbappendaddr(&last->in6p_socket->so_rcv, 1963 (struct sockaddr *)&rip6src, m, opts) == 0) { 1964 m_freem(m); 1965 if (opts) 1966 m_freem(opts); 1967 } else 1968 sorwakeup(last->in6p_socket); 1969 } else { 1970 m_freem(m); 1971 ip6stat.ip6s_delivered--; 1972 } 1973 return IPPROTO_DONE; 1974} 1975 1976/* 1977 * Reflect the ip6 packet back to the source. 1978 * OFF points to the icmp6 header, counted from the top of the mbuf. 1979 * 1980 * Note: RFC 1885 required that an echo reply should be truncated if it 1981 * did not fit in with (return) path MTU, and KAME code supported the 1982 * behavior. However, as a clarification after the RFC, this limitation 1983 * was removed in a revised version of the spec, RFC 2463. We had kept the 1984 * old behavior, with a (non-default) ifdef block, while the new version of 1985 * the spec was an internet-draft status, and even after the new RFC was 1986 * published. But it would rather make sense to clean the obsoleted part 1987 * up, and to make the code simpler at this stage. 1988 */ 1989void 1990icmp6_reflect(m, off) 1991 struct mbuf *m; 1992 size_t off; 1993{ 1994 struct ip6_hdr *ip6; 1995 struct icmp6_hdr *icmp6; 1996 struct in6_ifaddr *ia; 1997 struct in6_addr t, *src = 0; 1998 int plen; 1999 int type, code; 2000 struct ifnet *outif = NULL; 2001 struct sockaddr_in6 sa6_src, sa6_dst; 2002 2003 /* too short to reflect */ 2004 if (off < sizeof(struct ip6_hdr)) { 2005 nd6log((LOG_DEBUG, 2006 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 2007 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2008 __FILE__, __LINE__)); 2009 goto bad; 2010 } 2011 2012 /* 2013 * If there are extra headers between IPv6 and ICMPv6, strip 2014 * off that header first. 2015 */ 2016#ifdef DIAGNOSTIC 2017 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2018 panic("assumption failed in icmp6_reflect"); 2019#endif 2020 if (off > sizeof(struct ip6_hdr)) { 2021 size_t l; 2022 struct ip6_hdr nip6; 2023 2024 l = off - sizeof(struct ip6_hdr); 2025 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2026 m_adj(m, l); 2027 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2028 if (m->m_len < l) { 2029 if ((m = m_pullup(m, l)) == NULL) 2030 return; 2031 } 2032 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2033 } else /* off == sizeof(struct ip6_hdr) */ { 2034 size_t l; 2035 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2036 if (m->m_len < l) { 2037 if ((m = m_pullup(m, l)) == NULL) 2038 return; 2039 } 2040 } 2041 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2042 ip6 = mtod(m, struct ip6_hdr *); 2043 ip6->ip6_nxt = IPPROTO_ICMPV6; 2044 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2045 type = icmp6->icmp6_type; /* keep type for statistics */ 2046 code = icmp6->icmp6_code; /* ditto. */ 2047 2048 t = ip6->ip6_dst; 2049 /* 2050 * ip6_input() drops a packet if its src is multicast. 2051 * So, the src is never multicast. 2052 */ 2053 ip6->ip6_dst = ip6->ip6_src; 2054 2055 /* 2056 * XXX: make sure to embed scope zone information, using 2057 * already embedded IDs or the received interface (if any). 2058 * Note that rcvif may be NULL. 2059 * TODO: scoped routing case (XXX). 2060 */ 2061 bzero(&sa6_src, sizeof(sa6_src)); 2062 sa6_src.sin6_family = AF_INET6; 2063 sa6_src.sin6_len = sizeof(sa6_src); 2064 sa6_src.sin6_addr = ip6->ip6_dst; 2065 in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif); 2066 in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL); 2067 ip6->ip6_dst = sa6_src.sin6_addr; 2068 2069 bzero(&sa6_dst, sizeof(sa6_dst)); 2070 sa6_dst.sin6_family = AF_INET6; 2071 sa6_dst.sin6_len = sizeof(sa6_dst); 2072 sa6_dst.sin6_addr = t; 2073 in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif); 2074 in6_embedscope(&t, &sa6_dst, NULL, NULL); 2075 2076 /* 2077 * If the incoming packet was addressed directly to us (i.e. unicast), 2078 * use dst as the src for the reply. 2079 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible 2080 * (for example) when we encounter an error while forwarding procedure 2081 * destined to a duplicated address of ours. 2082 */ 2083 for (ia = in6_ifaddr; ia; ia = ia->ia_next) 2084 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) && 2085 (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) { 2086 src = &t; 2087 break; 2088 } 2089 if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) { 2090 /* 2091 * This is the case if the dst is our link-local address 2092 * and the sender is also ourselves. 2093 */ 2094 src = &t; 2095 } 2096 2097 if (src == 0) { 2098 int e; 2099 struct route_in6 ro; 2100 2101 /* 2102 * This case matches to multicasts, our anycast, or unicasts 2103 * that we do not own. Select a source address based on the 2104 * source address of the erroneous packet. 2105 */ 2106 bzero(&ro, sizeof(ro)); 2107 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e); 2108 if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */ 2109 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2110 } 2111 if (src == NULL) { 2112 nd6log((LOG_DEBUG, 2113 "icmp6_reflect: source can't be determined: " 2114 "dst=%s, error=%d\n", 2115 ip6_sprintf(&sa6_src.sin6_addr), e)); 2116 goto bad; 2117 } 2118 } 2119 2120 ip6->ip6_src = *src; 2121 2122 ip6->ip6_flow = 0; 2123 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2124 ip6->ip6_vfc |= IPV6_VERSION; 2125 ip6->ip6_nxt = IPPROTO_ICMPV6; 2126 if (m->m_pkthdr.rcvif) { 2127 /* XXX: This may not be the outgoing interface */ 2128 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2129 } else 2130 ip6->ip6_hlim = ip6_defhlim; 2131 2132 icmp6->icmp6_cksum = 0; 2133 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2134 sizeof(struct ip6_hdr), plen); 2135 2136 /* 2137 * XXX option handling 2138 */ 2139 2140 m->m_flags &= ~(M_BCAST|M_MCAST); 2141#ifdef IPSEC 2142 /* Don't lookup socket */ 2143 (void)ipsec_setsocket(m, NULL); 2144#endif /* IPSEC */ 2145 2146 /* 2147 * To avoid a "too big" situation at an intermediate router 2148 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2149 * Note that only echo and node information replies are affected, 2150 * since the length of ICMP6 errors is limited to the minimum MTU. 2151 */ 2152 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, &outif) != 0 && outif) 2153 icmp6_ifstat_inc(outif, ifs6_out_error); 2154 2155 if (outif) 2156 icmp6_ifoutstat_inc(outif, type, code); 2157 2158 return; 2159 2160 bad: 2161 m_freem(m); 2162 return; 2163} 2164 2165void 2166icmp6_fasttimo() 2167{ 2168 2169 mld6_fasttimeo(); 2170} 2171 2172static const char * 2173icmp6_redirect_diag(src6, dst6, tgt6) 2174 struct in6_addr *src6; 2175 struct in6_addr *dst6; 2176 struct in6_addr *tgt6; 2177{ 2178 static char buf[1024]; 2179 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2180 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2181 return buf; 2182} 2183 2184void 2185icmp6_redirect_input(m, off) 2186 struct mbuf *m; 2187 int off; 2188{ 2189 struct ifnet *ifp = m->m_pkthdr.rcvif; 2190 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2191 struct nd_redirect *nd_rd; 2192 int icmp6len = ntohs(ip6->ip6_plen); 2193 char *lladdr = NULL; 2194 int lladdrlen = 0; 2195 u_char *redirhdr = NULL; 2196 int redirhdrlen = 0; 2197 struct rtentry *rt = NULL; 2198 int is_router; 2199 int is_onlink; 2200 struct in6_addr src6 = ip6->ip6_src; 2201 struct in6_addr redtgt6; 2202 struct in6_addr reddst6; 2203 union nd_opts ndopts; 2204 2205 if (!ifp) 2206 return; 2207 2208 /* XXX if we are router, we don't update route by icmp6 redirect */ 2209 if (ip6_forwarding) 2210 goto freeit; 2211 if (!icmp6_rediraccept) 2212 goto freeit; 2213 2214 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2215 if (nd_rd == NULL) { 2216 icmp6stat.icp6s_tooshort++; 2217 return; 2218 } 2219 redtgt6 = nd_rd->nd_rd_target; 2220 reddst6 = nd_rd->nd_rd_dst; 2221 2222 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2223 redtgt6.s6_addr16[1] = htons(ifp->if_index); 2224 if (IN6_IS_ADDR_LINKLOCAL(&reddst6)) 2225 reddst6.s6_addr16[1] = htons(ifp->if_index); 2226 2227 /* validation */ 2228 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2229 nd6log((LOG_ERR, 2230 "ICMP6 redirect sent from %s rejected; " 2231 "must be from linklocal\n", ip6_sprintf(&src6))); 2232 goto bad; 2233 } 2234 if (ip6->ip6_hlim != 255) { 2235 nd6log((LOG_ERR, 2236 "ICMP6 redirect sent from %s rejected; " 2237 "hlim=%d (must be 255)\n", 2238 ip6_sprintf(&src6), ip6->ip6_hlim)); 2239 goto bad; 2240 } 2241 { 2242 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2243 struct sockaddr_in6 sin6; 2244 struct in6_addr *gw6; 2245 2246 bzero(&sin6, sizeof(sin6)); 2247 sin6.sin6_family = AF_INET6; 2248 sin6.sin6_len = sizeof(struct sockaddr_in6); 2249 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2250 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2251 if (rt) { 2252 if (rt->rt_gateway == NULL || 2253 rt->rt_gateway->sa_family != AF_INET6) { 2254 nd6log((LOG_ERR, 2255 "ICMP6 redirect rejected; no route " 2256 "with inet6 gateway found for redirect dst: %s\n", 2257 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2258 RTFREE(rt); 2259 goto bad; 2260 } 2261 2262 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2263 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2264 nd6log((LOG_ERR, 2265 "ICMP6 redirect rejected; " 2266 "not equal to gw-for-src=%s (must be same): " 2267 "%s\n", 2268 ip6_sprintf(gw6), 2269 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2270 RTFREE(rt); 2271 goto bad; 2272 } 2273 } else { 2274 nd6log((LOG_ERR, 2275 "ICMP6 redirect rejected; " 2276 "no route found for redirect dst: %s\n", 2277 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2278 goto bad; 2279 } 2280 RTFREE(rt); 2281 rt = NULL; 2282 } 2283 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2284 nd6log((LOG_ERR, 2285 "ICMP6 redirect rejected; " 2286 "redirect dst must be unicast: %s\n", 2287 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2288 goto bad; 2289 } 2290 2291 is_router = is_onlink = 0; 2292 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2293 is_router = 1; /* router case */ 2294 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2295 is_onlink = 1; /* on-link destination case */ 2296 if (!is_router && !is_onlink) { 2297 nd6log((LOG_ERR, 2298 "ICMP6 redirect rejected; " 2299 "neither router case nor onlink case: %s\n", 2300 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2301 goto bad; 2302 } 2303 /* validation passed */ 2304 2305 icmp6len -= sizeof(*nd_rd); 2306 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2307 if (nd6_options(&ndopts) < 0) { 2308 nd6log((LOG_INFO, "icmp6_redirect_input: " 2309 "invalid ND option, rejected: %s\n", 2310 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2311 /* nd6_options have incremented stats */ 2312 goto freeit; 2313 } 2314 2315 if (ndopts.nd_opts_tgt_lladdr) { 2316 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2317 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2318 } 2319 2320 if (ndopts.nd_opts_rh) { 2321 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len; 2322 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */ 2323 } 2324 2325 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2326 nd6log((LOG_INFO, 2327 "icmp6_redirect_input: lladdrlen mismatch for %s " 2328 "(if %d, icmp6 packet %d): %s\n", 2329 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2330 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2331 goto bad; 2332 } 2333 2334 /* RFC 2461 8.3 */ 2335 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2336 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2337 2338 if (!is_onlink) { /* better router case. perform rtredirect. */ 2339 /* perform rtredirect */ 2340 struct sockaddr_in6 sdst; 2341 struct sockaddr_in6 sgw; 2342 struct sockaddr_in6 ssrc; 2343 unsigned long rtcount; 2344 struct rtentry *newrt = NULL; 2345 2346 /* 2347 * do not install redirect route, if the number of entries 2348 * is too much (> hiwat). note that, the node (= host) will 2349 * work just fine even if we do not install redirect route 2350 * (there will be additional hops, though). 2351 */ 2352 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2353 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2354 return; 2355 else if (0 <= icmp6_redirect_lowat && 2356 rtcount > icmp6_redirect_lowat) { 2357 /* 2358 * XXX nuke a victim, install the new one. 2359 */ 2360 } 2361 2362 bzero(&sdst, sizeof(sdst)); 2363 bzero(&sgw, sizeof(sgw)); 2364 bzero(&ssrc, sizeof(ssrc)); 2365 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2366 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2367 sizeof(struct sockaddr_in6); 2368 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2369 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2370 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2371 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2372 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2373 (struct sockaddr *)&ssrc, 2374 &newrt); 2375 2376 if (newrt) { 2377 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2378 icmp6_redirect_timeout_q); 2379 rtfree(newrt); 2380 } 2381 } 2382 /* finally update cached route in each socket via pfctlinput */ 2383 { 2384 struct sockaddr_in6 sdst; 2385 2386 bzero(&sdst, sizeof(sdst)); 2387 sdst.sin6_family = AF_INET6; 2388 sdst.sin6_len = sizeof(struct sockaddr_in6); 2389 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2390 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2391#ifdef IPSEC 2392 key_sa_routechange((struct sockaddr *)&sdst); 2393#endif 2394 } 2395 2396 freeit: 2397 m_freem(m); 2398 return; 2399 2400 bad: 2401 icmp6stat.icp6s_badredirect++; 2402 m_freem(m); 2403} 2404 2405void 2406icmp6_redirect_output(m0, rt) 2407 struct mbuf *m0; 2408 struct rtentry *rt; 2409{ 2410 struct ifnet *ifp; /* my outgoing interface */ 2411 struct in6_addr *ifp_ll6; 2412 struct in6_addr *nexthop; 2413 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2414 struct mbuf *m = NULL; /* newly allocated one */ 2415 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2416 struct nd_redirect *nd_rd; 2417 size_t maxlen; 2418 u_char *p; 2419 struct sockaddr_in6 src_sa; 2420 2421 icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2422 2423 /* if we are not router, we don't send icmp6 redirect */ 2424 if (!ip6_forwarding || ip6_accept_rtadv) 2425 goto fail; 2426 2427 /* sanity check */ 2428 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2429 goto fail; 2430 2431 /* 2432 * Address check: 2433 * the source address must identify a neighbor, and 2434 * the destination address must not be a multicast address 2435 * [RFC 2461, sec 8.2] 2436 */ 2437 sip6 = mtod(m0, struct ip6_hdr *); 2438 bzero(&src_sa, sizeof(src_sa)); 2439 src_sa.sin6_family = AF_INET6; 2440 src_sa.sin6_len = sizeof(src_sa); 2441 src_sa.sin6_addr = sip6->ip6_src; 2442 /* we don't currently use sin6_scope_id, but eventually use it */ 2443 src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src); 2444 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2445 goto fail; 2446 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2447 goto fail; /* what should we do here? */ 2448 2449 /* rate limit */ 2450 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2451 goto fail; 2452 2453 /* 2454 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2455 * we almost always ask for an mbuf cluster for simplicity. 2456 * (MHLEN < IPV6_MMTU is almost always true) 2457 */ 2458#if IPV6_MMTU >= MCLBYTES 2459# error assumption failed about IPV6_MMTU and MCLBYTES 2460#endif 2461 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2462 if (m && IPV6_MMTU >= MHLEN) 2463 MCLGET(m, M_DONTWAIT); 2464 if (!m) 2465 goto fail; 2466 m->m_pkthdr.rcvif = NULL; 2467 m->m_len = 0; 2468 maxlen = M_TRAILINGSPACE(m); 2469 maxlen = min(IPV6_MMTU, maxlen); 2470 /* just for safety */ 2471 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2472 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2473 goto fail; 2474 } 2475 2476 { 2477 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2478 struct in6_ifaddr *ia; 2479 if ((ia = in6ifa_ifpforlinklocal(ifp, 2480 IN6_IFF_NOTREADY| 2481 IN6_IFF_ANYCAST)) == NULL) 2482 goto fail; 2483 ifp_ll6 = &ia->ia_addr.sin6_addr; 2484 } 2485 2486 /* get ip6 linklocal address for the router. */ 2487 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2488 struct sockaddr_in6 *sin6; 2489 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2490 nexthop = &sin6->sin6_addr; 2491 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2492 nexthop = NULL; 2493 } else 2494 nexthop = NULL; 2495 2496 /* ip6 */ 2497 ip6 = mtod(m, struct ip6_hdr *); 2498 ip6->ip6_flow = 0; 2499 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2500 ip6->ip6_vfc |= IPV6_VERSION; 2501 /* ip6->ip6_plen will be set later */ 2502 ip6->ip6_nxt = IPPROTO_ICMPV6; 2503 ip6->ip6_hlim = 255; 2504 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2505 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2506 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2507 2508 /* ND Redirect */ 2509 nd_rd = (struct nd_redirect *)(ip6 + 1); 2510 nd_rd->nd_rd_type = ND_REDIRECT; 2511 nd_rd->nd_rd_code = 0; 2512 nd_rd->nd_rd_reserved = 0; 2513 if (rt->rt_flags & RTF_GATEWAY) { 2514 /* 2515 * nd_rd->nd_rd_target must be a link-local address in 2516 * better router cases. 2517 */ 2518 if (!nexthop) 2519 goto fail; 2520 bcopy(nexthop, &nd_rd->nd_rd_target, 2521 sizeof(nd_rd->nd_rd_target)); 2522 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2523 sizeof(nd_rd->nd_rd_dst)); 2524 } else { 2525 /* make sure redtgt == reddst */ 2526 nexthop = &sip6->ip6_dst; 2527 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2528 sizeof(nd_rd->nd_rd_target)); 2529 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2530 sizeof(nd_rd->nd_rd_dst)); 2531 } 2532 2533 p = (u_char *)(nd_rd + 1); 2534 2535 { 2536 /* target lladdr option */ 2537 struct rtentry *rt_nexthop = NULL; 2538 int len; 2539 struct sockaddr_dl *sdl; 2540 struct nd_opt_hdr *nd_opt; 2541 char *lladdr; 2542 2543 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2544 if (!rt_nexthop) 2545 goto nolladdropt; 2546 len = sizeof(*nd_opt) + ifp->if_addrlen; 2547 len = (len + 7) & ~7; /* round by 8 */ 2548 /* safety check */ 2549 if (len + (p - (u_char *)ip6) > maxlen) 2550 goto nolladdropt; 2551 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2552 (rt_nexthop->rt_flags & RTF_LLINFO) && 2553 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2554 (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) && 2555 sdl->sdl_alen) { 2556 nd_opt = (struct nd_opt_hdr *)p; 2557 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2558 nd_opt->nd_opt_len = len >> 3; 2559 lladdr = (char *)(nd_opt + 1); 2560 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen); 2561 p += len; 2562 } 2563 } 2564 nolladdropt:; 2565 2566 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2567 2568 /* just to be safe */ 2569 if (m0->m_flags & M_DECRYPTED) 2570 goto noredhdropt; 2571 if (p - (u_char *)ip6 > maxlen) 2572 goto noredhdropt; 2573 2574 { 2575 /* redirected header option */ 2576 int len; 2577 struct nd_opt_rd_hdr *nd_opt_rh; 2578 2579 /* 2580 * compute the maximum size for icmp6 redirect header option. 2581 * XXX room for auth header? 2582 */ 2583 len = maxlen - (p - (u_char *)ip6); 2584 len &= ~7; 2585 2586 /* 2587 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2588 * about padding/truncate rule for the original IP packet. 2589 * From the discussion on IPv6imp in Feb 1999, 2590 * the consensus was: 2591 * - "attach as much as possible" is the goal 2592 * - pad if not aligned (original size can be guessed by 2593 * original ip6 header) 2594 * Following code adds the padding if it is simple enough, 2595 * and truncates if not. 2596 */ 2597 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2598 /* not enough room, truncate */ 2599 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2600 m0->m_pkthdr.len); 2601 } else { 2602 /* 2603 * enough room, truncate if not aligned. 2604 * we don't pad here for simplicity. 2605 */ 2606 size_t extra; 2607 2608 extra = m0->m_pkthdr.len % 8; 2609 if (extra) { 2610 /* truncate */ 2611 m_adj(m0, -extra); 2612 } 2613 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2614 } 2615 2616 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2617 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2618 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2619 nd_opt_rh->nd_opt_rh_len = len >> 3; 2620 p += sizeof(*nd_opt_rh); 2621 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2622 2623 /* connect m0 to m */ 2624 m_cat(m, m0); 2625 m->m_pkthdr.len += m0->m_pkthdr.len; 2626 m0 = NULL; 2627 } 2628noredhdropt: 2629 if (m0) { 2630 m_freem(m0); 2631 m0 = NULL; 2632 } 2633 2634 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src)) 2635 sip6->ip6_src.s6_addr16[1] = 0; 2636 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst)) 2637 sip6->ip6_dst.s6_addr16[1] = 0; 2638#if 0 2639 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) 2640 ip6->ip6_src.s6_addr16[1] = 0; 2641 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) 2642 ip6->ip6_dst.s6_addr16[1] = 0; 2643#endif 2644 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target)) 2645 nd_rd->nd_rd_target.s6_addr16[1] = 0; 2646 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst)) 2647 nd_rd->nd_rd_dst.s6_addr16[1] = 0; 2648 2649 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2650 2651 nd_rd->nd_rd_cksum = 0; 2652 nd_rd->nd_rd_cksum 2653 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2654 2655 /* send the packet to outside... */ 2656#ifdef IPSEC 2657 /* Don't lookup socket */ 2658 (void)ipsec_setsocket(m, NULL); 2659#endif /* IPSEC */ 2660 if (ip6_output(m, NULL, NULL, 0, NULL, NULL) != 0) 2661 icmp6_ifstat_inc(ifp, ifs6_out_error); 2662 2663 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2664 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2665 icmp6stat.icp6s_outhist[ND_REDIRECT]++; 2666 2667 return; 2668 2669fail: 2670 if (m) 2671 m_freem(m); 2672 if (m0) 2673 m_freem(m0); 2674} 2675 2676/* 2677 * ICMPv6 socket option processing. 2678 */ 2679int 2680icmp6_ctloutput(op, so, level, optname, mp) 2681 int op; 2682 struct socket *so; 2683 int level, optname; 2684 struct mbuf **mp; 2685{ 2686 int error = 0; 2687 int optlen; 2688 struct in6pcb *in6p = sotoin6pcb(so); 2689 struct mbuf *m = *mp; 2690 2691 optlen = m ? m->m_len : 0; 2692 2693 if (level != IPPROTO_ICMPV6) { 2694 if (op == PRCO_SETOPT && m) 2695 (void)m_free(m); 2696 return EINVAL; 2697 } 2698 2699 switch (op) { 2700 case PRCO_SETOPT: 2701 switch (optname) { 2702 case ICMP6_FILTER: 2703 { 2704 struct icmp6_filter *p; 2705 2706 if (optlen != sizeof(*p)) { 2707 error = EMSGSIZE; 2708 break; 2709 } 2710 p = mtod(m, struct icmp6_filter *); 2711 if (!p || !in6p->in6p_icmp6filt) { 2712 error = EINVAL; 2713 break; 2714 } 2715 bcopy(p, in6p->in6p_icmp6filt, 2716 sizeof(struct icmp6_filter)); 2717 error = 0; 2718 break; 2719 } 2720 2721 default: 2722 error = ENOPROTOOPT; 2723 break; 2724 } 2725 if (m) 2726 (void)m_freem(m); 2727 break; 2728 2729 case PRCO_GETOPT: 2730 switch (optname) { 2731 case ICMP6_FILTER: 2732 { 2733 struct icmp6_filter *p; 2734 2735 if (!in6p->in6p_icmp6filt) { 2736 error = EINVAL; 2737 break; 2738 } 2739 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2740 m->m_len = sizeof(struct icmp6_filter); 2741 p = mtod(m, struct icmp6_filter *); 2742 bcopy(in6p->in6p_icmp6filt, p, 2743 sizeof(struct icmp6_filter)); 2744 error = 0; 2745 break; 2746 } 2747 2748 default: 2749 error = ENOPROTOOPT; 2750 break; 2751 } 2752 break; 2753 } 2754 2755 return (error); 2756} 2757 2758/* 2759 * Perform rate limit check. 2760 * Returns 0 if it is okay to send the icmp6 packet. 2761 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2762 * limitation. 2763 * 2764 * XXX per-destination/type check necessary? 2765 */ 2766static int 2767icmp6_ratelimit(dst, type, code) 2768 const struct in6_addr *dst; /* not used at this moment */ 2769 const int type; /* not used at this moment */ 2770 const int code; /* not used at this moment */ 2771{ 2772 int ret; 2773 2774 ret = 0; /* okay to send */ 2775 2776 /* PPS limit */ 2777 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2778 icmp6errppslim)) { 2779 /* The packet is subject to rate limit */ 2780 ret++; 2781 } 2782 2783 return ret; 2784} 2785 2786static struct rtentry * 2787icmp6_mtudisc_clone(dst) 2788 struct sockaddr *dst; 2789{ 2790 struct rtentry *rt; 2791 int error; 2792 2793 rt = rtalloc1(dst, 1); 2794 if (rt == 0) 2795 return NULL; 2796 2797 /* If we didn't get a host route, allocate one */ 2798 if ((rt->rt_flags & RTF_HOST) == 0) { 2799 struct rtentry *nrt; 2800 2801 error = rtrequest((int) RTM_ADD, dst, 2802 (struct sockaddr *) rt->rt_gateway, 2803 (struct sockaddr *) 0, 2804 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2805 if (error) { 2806 rtfree(rt); 2807 return NULL; 2808 } 2809 nrt->rt_rmx = rt->rt_rmx; 2810 rtfree(rt); 2811 rt = nrt; 2812 } 2813 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2814 icmp6_mtudisc_timeout_q); 2815 if (error) { 2816 rtfree(rt); 2817 return NULL; 2818 } 2819 2820 return rt; /* caller need to call rtfree() */ 2821} 2822 2823static void 2824icmp6_mtudisc_timeout(rt, r) 2825 struct rtentry *rt; 2826 struct rttimer *r; 2827{ 2828 if (rt == NULL) 2829 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2830 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2831 (RTF_DYNAMIC | RTF_HOST)) { 2832 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2833 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2834 } else { 2835 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2836 rt->rt_rmx.rmx_mtu = 0; 2837 } 2838} 2839 2840static void 2841icmp6_redirect_timeout(rt, r) 2842 struct rtentry *rt; 2843 struct rttimer *r; 2844{ 2845 if (rt == NULL) 2846 panic("icmp6_redirect_timeout: bad route to timeout"); 2847 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2848 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2849 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2850 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2851 } 2852} 2853 2854int 2855icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 2856 int *name; 2857 u_int namelen; 2858 void *oldp; 2859 size_t *oldlenp; 2860 void *newp; 2861 size_t newlen; 2862{ 2863 2864 /* All sysctl names at this level are terminal. */ 2865 if (namelen != 1) 2866 return ENOTDIR; 2867 2868 switch (name[0]) { 2869 2870 case ICMPV6CTL_REDIRACCEPT: 2871 return sysctl_int(oldp, oldlenp, newp, newlen, 2872 &icmp6_rediraccept); 2873 case ICMPV6CTL_REDIRTIMEOUT: 2874 return sysctl_int(oldp, oldlenp, newp, newlen, 2875 &icmp6_redirtimeout); 2876 case ICMPV6CTL_STATS: 2877 return sysctl_rdstruct(oldp, oldlenp, newp, 2878 &icmp6stat, sizeof(icmp6stat)); 2879 case ICMPV6CTL_ND6_PRUNE: 2880 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune); 2881 case ICMPV6CTL_ND6_DELAY: 2882 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay); 2883 case ICMPV6CTL_ND6_UMAXTRIES: 2884 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries); 2885 case ICMPV6CTL_ND6_MMAXTRIES: 2886 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries); 2887 case ICMPV6CTL_ND6_USELOOPBACK: 2888 return sysctl_int(oldp, oldlenp, newp, newlen, 2889 &nd6_useloopback); 2890 case ICMPV6CTL_NODEINFO: 2891 return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo); 2892 case ICMPV6CTL_ERRPPSLIMIT: 2893 return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6errppslim); 2894 case ICMPV6CTL_ND6_MAXNUDHINT: 2895 return sysctl_int(oldp, oldlenp, newp, newlen, 2896 &nd6_maxnudhint); 2897 case ICMPV6CTL_MTUDISC_HIWAT: 2898 return sysctl_int(oldp, oldlenp, newp, newlen, 2899 &icmp6_mtudisc_hiwat); 2900 case ICMPV6CTL_MTUDISC_LOWAT: 2901 return sysctl_int(oldp, oldlenp, newp, newlen, 2902 &icmp6_mtudisc_lowat); 2903 case ICMPV6CTL_ND6_DEBUG: 2904 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_debug); 2905 case ICMPV6CTL_ND6_DRLIST: 2906 case ICMPV6CTL_ND6_PRLIST: 2907 return nd6_sysctl(name[0], oldp, oldlenp, newp, newlen); 2908 default: 2909 return ENOPROTOOPT; 2910 } 2911 /* NOTREACHED */ 2912} 2913