icmp6.c revision 191672
1/*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $ 30 */ 31 32/*- 33 * Copyright (c) 1982, 1986, 1988, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 4. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 61 */ 62 63#include <sys/cdefs.h> 64__FBSDID("$FreeBSD: head/sys/netinet6/icmp6.c 191672 2009-04-29 19:19:13Z bms $"); 65 66#include "opt_inet.h" 67#include "opt_inet6.h" 68#include "opt_ipsec.h" 69#include "opt_route.h" 70 71#include <sys/param.h> 72#include <sys/domain.h> 73#include <sys/kernel.h> 74#include <sys/lock.h> 75#include <sys/malloc.h> 76#include <sys/mbuf.h> 77#include <sys/protosw.h> 78#include <sys/signalvar.h> 79#include <sys/socket.h> 80#include <sys/socketvar.h> 81#include <sys/sx.h> 82#include <sys/syslog.h> 83#include <sys/systm.h> 84#include <sys/time.h> 85#include <sys/vimage.h> 86 87#include <net/if.h> 88#include <net/if_dl.h> 89#include <net/if_llatbl.h> 90#include <net/if_types.h> 91#include <net/route.h> 92#include <net/vnet.h> 93 94#include <netinet/in.h> 95#include <netinet/in_pcb.h> 96#include <netinet/in_var.h> 97#include <netinet/ip6.h> 98#include <netinet/icmp6.h> 99#include <netinet/tcp_var.h> 100#include <netinet/vinet.h> 101 102#include <netinet6/in6_ifattach.h> 103#include <netinet6/in6_pcb.h> 104#include <netinet6/ip6protosw.h> 105#include <netinet6/ip6_var.h> 106#include <netinet6/scope6_var.h> 107#include <netinet6/mld6_var.h> 108#include <netinet6/nd6.h> 109#include <netinet6/vinet6.h> 110 111#ifdef IPSEC 112#include <netipsec/ipsec.h> 113#include <netipsec/key.h> 114#endif 115 116extern struct domain inet6domain; 117 118#ifdef VIMAGE_GLOBALS 119extern struct inpcbinfo ripcbinfo; 120extern struct inpcbhead ripcb; 121extern int icmp6errppslim; 122extern int icmp6_nodeinfo; 123 124struct icmp6stat icmp6stat; 125static int icmp6errpps_count; 126static struct timeval icmp6errppslim_last; 127#endif 128 129static void icmp6_errcount(struct icmp6errstat *, int, int); 130static int icmp6_rip6_input(struct mbuf **, int); 131static int icmp6_ratelimit(const struct in6_addr *, const int, const int); 132static const char *icmp6_redirect_diag __P((struct in6_addr *, 133 struct in6_addr *, struct in6_addr *)); 134static struct mbuf *ni6_input(struct mbuf *, int); 135static struct mbuf *ni6_nametodns(const char *, int, int); 136static int ni6_dnsmatch(const char *, int, const char *, int); 137static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *, 138 struct ifnet **, struct in6_addr *)); 139static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 140 struct ifnet *, int)); 141static int icmp6_notify_error(struct mbuf **, int, int, int); 142 143 144void 145icmp6_init(void) 146{ 147 INIT_VNET_INET6(curvnet); 148 149 V_icmp6errpps_count = 0; 150} 151 152static void 153icmp6_errcount(struct icmp6errstat *stat, int type, int code) 154{ 155 switch (type) { 156 case ICMP6_DST_UNREACH: 157 switch (code) { 158 case ICMP6_DST_UNREACH_NOROUTE: 159 stat->icp6errs_dst_unreach_noroute++; 160 return; 161 case ICMP6_DST_UNREACH_ADMIN: 162 stat->icp6errs_dst_unreach_admin++; 163 return; 164 case ICMP6_DST_UNREACH_BEYONDSCOPE: 165 stat->icp6errs_dst_unreach_beyondscope++; 166 return; 167 case ICMP6_DST_UNREACH_ADDR: 168 stat->icp6errs_dst_unreach_addr++; 169 return; 170 case ICMP6_DST_UNREACH_NOPORT: 171 stat->icp6errs_dst_unreach_noport++; 172 return; 173 } 174 break; 175 case ICMP6_PACKET_TOO_BIG: 176 stat->icp6errs_packet_too_big++; 177 return; 178 case ICMP6_TIME_EXCEEDED: 179 switch (code) { 180 case ICMP6_TIME_EXCEED_TRANSIT: 181 stat->icp6errs_time_exceed_transit++; 182 return; 183 case ICMP6_TIME_EXCEED_REASSEMBLY: 184 stat->icp6errs_time_exceed_reassembly++; 185 return; 186 } 187 break; 188 case ICMP6_PARAM_PROB: 189 switch (code) { 190 case ICMP6_PARAMPROB_HEADER: 191 stat->icp6errs_paramprob_header++; 192 return; 193 case ICMP6_PARAMPROB_NEXTHEADER: 194 stat->icp6errs_paramprob_nextheader++; 195 return; 196 case ICMP6_PARAMPROB_OPTION: 197 stat->icp6errs_paramprob_option++; 198 return; 199 } 200 break; 201 case ND_REDIRECT: 202 stat->icp6errs_redirect++; 203 return; 204 } 205 stat->icp6errs_unknown++; 206} 207 208/* 209 * A wrapper function for icmp6_error() necessary when the erroneous packet 210 * may not contain enough scope zone information. 211 */ 212void 213icmp6_error2(struct mbuf *m, int type, int code, int param, 214 struct ifnet *ifp) 215{ 216 INIT_VNET_INET6(curvnet); 217 struct ip6_hdr *ip6; 218 219 if (ifp == NULL) 220 return; 221 222#ifndef PULLDOWN_TEST 223 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), ); 224#else 225 if (m->m_len < sizeof(struct ip6_hdr)) { 226 m = m_pullup(m, sizeof(struct ip6_hdr)); 227 if (m == NULL) 228 return; 229 } 230#endif 231 232 ip6 = mtod(m, struct ip6_hdr *); 233 234 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 235 return; 236 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 237 return; 238 239 icmp6_error(m, type, code, param); 240} 241 242/* 243 * Generate an error packet of type error in response to bad IP6 packet. 244 */ 245void 246icmp6_error(struct mbuf *m, int type, int code, int param) 247{ 248 INIT_VNET_INET6(curvnet); 249 struct ip6_hdr *oip6, *nip6; 250 struct icmp6_hdr *icmp6; 251 u_int preplen; 252 int off; 253 int nxt; 254 255 ICMP6STAT_INC(icp6s_error); 256 257 /* count per-type-code statistics */ 258 icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, type, code); 259 260#ifdef M_DECRYPTED /*not openbsd*/ 261 if (m->m_flags & M_DECRYPTED) { 262 ICMP6STAT_INC(icp6s_canterror); 263 goto freeit; 264 } 265#endif 266 267#ifndef PULLDOWN_TEST 268 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), ); 269#else 270 if (m->m_len < sizeof(struct ip6_hdr)) { 271 m = m_pullup(m, sizeof(struct ip6_hdr)); 272 if (m == NULL) 273 return; 274 } 275#endif 276 oip6 = mtod(m, struct ip6_hdr *); 277 278 /* 279 * If the destination address of the erroneous packet is a multicast 280 * address, or the packet was sent using link-layer multicast, 281 * we should basically suppress sending an error (RFC 2463, Section 282 * 2.4). 283 * We have two exceptions (the item e.2 in that section): 284 * - the Pakcet Too Big message can be sent for path MTU discovery. 285 * - the Parameter Problem Message that can be allowed an icmp6 error 286 * in the option type field. This check has been done in 287 * ip6_unknown_opt(), so we can just check the type and code. 288 */ 289 if ((m->m_flags & (M_BCAST|M_MCAST) || 290 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 291 (type != ICMP6_PACKET_TOO_BIG && 292 (type != ICMP6_PARAM_PROB || 293 code != ICMP6_PARAMPROB_OPTION))) 294 goto freeit; 295 296 /* 297 * RFC 2463, 2.4 (e.5): source address check. 298 * XXX: the case of anycast source? 299 */ 300 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 301 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 302 goto freeit; 303 304 /* 305 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 306 * don't do it. 307 */ 308 nxt = -1; 309 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 310 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 311 struct icmp6_hdr *icp; 312 313#ifndef PULLDOWN_TEST 314 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), ); 315 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 316#else 317 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 318 sizeof(*icp)); 319 if (icp == NULL) { 320 ICMP6STAT_INC(icp6s_tooshort); 321 return; 322 } 323#endif 324 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 325 icp->icmp6_type == ND_REDIRECT) { 326 /* 327 * ICMPv6 error 328 * Special case: for redirect (which is 329 * informational) we must not send icmp6 error. 330 */ 331 ICMP6STAT_INC(icp6s_canterror); 332 goto freeit; 333 } else { 334 /* ICMPv6 informational - send the error */ 335 } 336 } else { 337 /* non-ICMPv6 - send the error */ 338 } 339 340 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 341 342 /* Finally, do rate limitation check. */ 343 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 344 ICMP6STAT_INC(icp6s_toofreq); 345 goto freeit; 346 } 347 348 /* 349 * OK, ICMP6 can be generated. 350 */ 351 352 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 353 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 354 355 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 356 M_PREPEND(m, preplen, M_DONTWAIT); 357 if (m && m->m_len < preplen) 358 m = m_pullup(m, preplen); 359 if (m == NULL) { 360 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 361 return; 362 } 363 364 nip6 = mtod(m, struct ip6_hdr *); 365 nip6->ip6_src = oip6->ip6_src; 366 nip6->ip6_dst = oip6->ip6_dst; 367 368 in6_clearscope(&oip6->ip6_src); 369 in6_clearscope(&oip6->ip6_dst); 370 371 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 372 icmp6->icmp6_type = type; 373 icmp6->icmp6_code = code; 374 icmp6->icmp6_pptr = htonl((u_int32_t)param); 375 376 /* 377 * icmp6_reflect() is designed to be in the input path. 378 * icmp6_error() can be called from both input and output path, 379 * and if we are in output path rcvif could contain bogus value. 380 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 381 * information in ip header (nip6). 382 */ 383 m->m_pkthdr.rcvif = NULL; 384 385 ICMP6STAT_INC(icp6s_outhist[type]); 386 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 387 388 return; 389 390 freeit: 391 /* 392 * If we can't tell whether or not we can generate ICMP6, free it. 393 */ 394 m_freem(m); 395} 396 397/* 398 * Process a received ICMP6 message. 399 */ 400int 401icmp6_input(struct mbuf **mp, int *offp, int proto) 402{ 403 INIT_VNET_INET6(curvnet); 404 INIT_VPROCG(TD_TO_VPROCG(curthread)); /* XXX V_hostname needs this */ 405 struct mbuf *m = *mp, *n; 406 struct ip6_hdr *ip6, *nip6; 407 struct icmp6_hdr *icmp6, *nicmp6; 408 int off = *offp; 409 int icmp6len = m->m_pkthdr.len - *offp; 410 int code, sum, noff; 411 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 412 413#ifndef PULLDOWN_TEST 414 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE); 415 /* m might change if M_LOOP. So, call mtod after this */ 416#endif 417 418 /* 419 * Locate icmp6 structure in mbuf, and check 420 * that not corrupted and of at least minimum length 421 */ 422 423 ip6 = mtod(m, struct ip6_hdr *); 424 if (icmp6len < sizeof(struct icmp6_hdr)) { 425 ICMP6STAT_INC(icp6s_tooshort); 426 goto freeit; 427 } 428 429 /* 430 * Check multicast group membership. 431 * Note: SSM filters are not applied for ICMPv6 traffic. 432 */ 433 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 434 struct ifnet *ifp; 435 struct in6_multi *inm; 436 437 ifp = m->m_pkthdr.rcvif; 438 inm = in6m_lookup(ifp, &ip6->ip6_dst); 439 if (inm == NULL) { 440 IP6STAT_INC(ip6s_notmember); 441 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 442 goto freeit; 443 } 444 } 445 446 /* 447 * calculate the checksum 448 */ 449#ifndef PULLDOWN_TEST 450 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 451#else 452 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 453 if (icmp6 == NULL) { 454 ICMP6STAT_INC(icp6s_tooshort); 455 return IPPROTO_DONE; 456 } 457#endif 458 code = icmp6->icmp6_code; 459 460 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 461 nd6log((LOG_ERR, 462 "ICMP6 checksum error(%d|%x) %s\n", 463 icmp6->icmp6_type, sum, 464 ip6_sprintf(ip6bufs, &ip6->ip6_src))); 465 ICMP6STAT_INC(icp6s_checksum); 466 goto freeit; 467 } 468 469 if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) { 470 /* 471 * Deliver very specific ICMP6 type only. 472 * This is important to deliver TOOBIG. Otherwise PMTUD 473 * will not work. 474 */ 475 switch (icmp6->icmp6_type) { 476 case ICMP6_DST_UNREACH: 477 case ICMP6_PACKET_TOO_BIG: 478 case ICMP6_TIME_EXCEEDED: 479 break; 480 default: 481 goto freeit; 482 } 483 } 484 485 ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]); 486 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 487 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK) 488 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 489 490 switch (icmp6->icmp6_type) { 491 case ICMP6_DST_UNREACH: 492 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 493 switch (code) { 494 case ICMP6_DST_UNREACH_NOROUTE: 495 code = PRC_UNREACH_NET; 496 break; 497 case ICMP6_DST_UNREACH_ADMIN: 498 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 499 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 500 break; 501 case ICMP6_DST_UNREACH_ADDR: 502 code = PRC_HOSTDEAD; 503 break; 504 case ICMP6_DST_UNREACH_BEYONDSCOPE: 505 /* I mean "source address was incorrect." */ 506 code = PRC_PARAMPROB; 507 break; 508 case ICMP6_DST_UNREACH_NOPORT: 509 code = PRC_UNREACH_PORT; 510 break; 511 default: 512 goto badcode; 513 } 514 goto deliver; 515 break; 516 517 case ICMP6_PACKET_TOO_BIG: 518 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 519 520 /* validation is made in icmp6_mtudisc_update */ 521 522 code = PRC_MSGSIZE; 523 524 /* 525 * Updating the path MTU will be done after examining 526 * intermediate extension headers. 527 */ 528 goto deliver; 529 break; 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 break; 545 546 case ICMP6_PARAM_PROB: 547 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 548 switch (code) { 549 case ICMP6_PARAMPROB_NEXTHEADER: 550 code = PRC_UNREACH_PROTOCOL; 551 break; 552 case ICMP6_PARAMPROB_HEADER: 553 case ICMP6_PARAMPROB_OPTION: 554 code = PRC_PARAMPROB; 555 break; 556 default: 557 goto badcode; 558 } 559 goto deliver; 560 break; 561 562 case ICMP6_ECHO_REQUEST: 563 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 564 if (code != 0) 565 goto badcode; 566 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) { 567 /* Give up remote */ 568 break; 569 } 570 if ((n->m_flags & M_EXT) != 0 571 || n->m_len < off + sizeof(struct icmp6_hdr)) { 572 struct mbuf *n0 = n; 573 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6); 574 int n0len; 575 576 MGETHDR(n, M_DONTWAIT, n0->m_type); 577 n0len = n0->m_pkthdr.len; /* save for use below */ 578 if (n) 579 M_MOVE_PKTHDR(n, n0); 580 if (n && maxlen >= MHLEN) { 581 MCLGET(n, M_DONTWAIT); 582 if ((n->m_flags & M_EXT) == 0) { 583 m_free(n); 584 n = NULL; 585 } 586 } 587 if (n == NULL) { 588 /* Give up remote */ 589 m_freem(n0); 590 break; 591 } 592 /* 593 * Copy IPv6 and ICMPv6 only. 594 */ 595 nip6 = mtod(n, struct ip6_hdr *); 596 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 597 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 598 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 599 noff = sizeof(struct ip6_hdr); 600 /* new mbuf contains only ipv6+icmpv6 headers */ 601 n->m_len = noff + sizeof(struct icmp6_hdr); 602 /* 603 * Adjust mbuf. ip6_plen will be adjusted in 604 * ip6_output(). 605 */ 606 m_adj(n0, off + sizeof(struct icmp6_hdr)); 607 /* recalculate complete packet size */ 608 n->m_pkthdr.len = n0len + (noff - off); 609 n->m_next = n0; 610 } else { 611 nip6 = mtod(n, struct ip6_hdr *); 612 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, 613 sizeof(*nicmp6)); 614 noff = off; 615 } 616 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 617 nicmp6->icmp6_code = 0; 618 if (n) { 619 ICMP6STAT_INC(icp6s_reflect); 620 ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]); 621 icmp6_reflect(n, noff); 622 } 623 break; 624 625 case ICMP6_ECHO_REPLY: 626 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 627 if (code != 0) 628 goto badcode; 629 break; 630 631 case MLD_LISTENER_QUERY: 632 case MLD_LISTENER_REPORT: 633 case MLD_LISTENER_DONE: 634 case MLDV2_LISTENER_REPORT: 635 /* 636 * Drop MLD traffic which is not link-local. 637 * XXX Should we also sanity check that these messages 638 * were directed to a link-local multicast prefix? 639 */ 640 if (ip6->ip6_hlim != 1) 641 goto freeit; 642 if (mld_input(m, off, icmp6len) != 0) 643 return (IPPROTO_DONE); 644 /* m stays. */ 645 break; 646 647 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 648 { 649 enum { WRU, FQDN } mode; 650 651 if (!V_icmp6_nodeinfo) 652 break; 653 654 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 655 mode = WRU; 656 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 657 mode = FQDN; 658 else 659 goto badlen; 660 661#define hostnamelen strlen(V_hostname) 662 if (mode == FQDN) { 663#ifndef PULLDOWN_TEST 664 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo), 665 IPPROTO_DONE); 666#endif 667 n = m_copy(m, 0, M_COPYALL); 668 if (n) 669 n = ni6_input(n, off); 670 /* XXX meaningless if n == NULL */ 671 noff = sizeof(struct ip6_hdr); 672 } else { 673 u_char *p; 674 int maxlen, maxhlen; 675 676 /* 677 * XXX: this combination of flags is pointless, 678 * but should we keep this for compatibility? 679 */ 680 if ((V_icmp6_nodeinfo & 5) != 5) 681 break; 682 683 if (code != 0) 684 goto badcode; 685 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; 686 if (maxlen >= MCLBYTES) { 687 /* Give up remote */ 688 break; 689 } 690 MGETHDR(n, M_DONTWAIT, m->m_type); 691 if (n && maxlen > MHLEN) { 692 MCLGET(n, M_DONTWAIT); 693 if ((n->m_flags & M_EXT) == 0) { 694 m_free(n); 695 n = NULL; 696 } 697 } 698 if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) { 699 /* 700 * Previous code did a blind M_COPY_PKTHDR 701 * and said "just for rcvif". If true, then 702 * we could tolerate the dup failing (due to 703 * the deep copy of the tag chain). For now 704 * be conservative and just fail. 705 */ 706 m_free(n); 707 n = NULL; 708 } 709 if (n == NULL) { 710 /* Give up remote */ 711 break; 712 } 713 n->m_pkthdr.rcvif = NULL; 714 n->m_len = 0; 715 maxhlen = M_TRAILINGSPACE(n) - maxlen; 716 mtx_lock(&hostname_mtx); 717 if (maxhlen > hostnamelen) 718 maxhlen = hostnamelen; 719 /* 720 * Copy IPv6 and ICMPv6 only. 721 */ 722 nip6 = mtod(n, struct ip6_hdr *); 723 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 724 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 725 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 726 p = (u_char *)(nicmp6 + 1); 727 bzero(p, 4); 728 bcopy(V_hostname, p + 4, maxhlen); /* meaningless TTL */ 729 mtx_unlock(&hostname_mtx); 730 noff = sizeof(struct ip6_hdr); 731 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 732 sizeof(struct icmp6_hdr) + 4 + maxhlen; 733 nicmp6->icmp6_type = ICMP6_WRUREPLY; 734 nicmp6->icmp6_code = 0; 735 } 736#undef hostnamelen 737 if (n) { 738 ICMP6STAT_INC(icp6s_reflect); 739 ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]); 740 icmp6_reflect(n, noff); 741 } 742 break; 743 } 744 745 case ICMP6_WRUREPLY: 746 if (code != 0) 747 goto badcode; 748 break; 749 750 case ND_ROUTER_SOLICIT: 751 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 752 if (code != 0) 753 goto badcode; 754 if (icmp6len < sizeof(struct nd_router_solicit)) 755 goto badlen; 756 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 757 /* give up local */ 758 nd6_rs_input(m, off, icmp6len); 759 m = NULL; 760 goto freeit; 761 } 762 nd6_rs_input(n, off, icmp6len); 763 /* m stays. */ 764 break; 765 766 case ND_ROUTER_ADVERT: 767 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 768 if (code != 0) 769 goto badcode; 770 if (icmp6len < sizeof(struct nd_router_advert)) 771 goto badlen; 772 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 773 /* give up local */ 774 nd6_ra_input(m, off, icmp6len); 775 m = NULL; 776 goto freeit; 777 } 778 nd6_ra_input(n, off, icmp6len); 779 /* m stays. */ 780 break; 781 782 case ND_NEIGHBOR_SOLICIT: 783 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 784 if (code != 0) 785 goto badcode; 786 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 787 goto badlen; 788 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 789 /* give up local */ 790 nd6_ns_input(m, off, icmp6len); 791 m = NULL; 792 goto freeit; 793 } 794 nd6_ns_input(n, off, icmp6len); 795 /* m stays. */ 796 break; 797 798 case ND_NEIGHBOR_ADVERT: 799 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 800 if (code != 0) 801 goto badcode; 802 if (icmp6len < sizeof(struct nd_neighbor_advert)) 803 goto badlen; 804 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 805 /* give up local */ 806 nd6_na_input(m, off, icmp6len); 807 m = NULL; 808 goto freeit; 809 } 810 nd6_na_input(n, off, icmp6len); 811 /* m stays. */ 812 break; 813 814 case ND_REDIRECT: 815 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 816 if (code != 0) 817 goto badcode; 818 if (icmp6len < sizeof(struct nd_redirect)) 819 goto badlen; 820 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 821 /* give up local */ 822 icmp6_redirect_input(m, off); 823 m = NULL; 824 goto freeit; 825 } 826 icmp6_redirect_input(n, off); 827 /* m stays. */ 828 break; 829 830 case ICMP6_ROUTER_RENUMBERING: 831 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 832 code != ICMP6_ROUTER_RENUMBERING_RESULT) 833 goto badcode; 834 if (icmp6len < sizeof(struct icmp6_router_renum)) 835 goto badlen; 836 break; 837 838 default: 839 nd6log((LOG_DEBUG, 840 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 841 icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src), 842 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 843 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 844 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 845 /* ICMPv6 error: MUST deliver it by spec... */ 846 code = PRC_NCMDS; 847 /* deliver */ 848 } else { 849 /* ICMPv6 informational: MUST not deliver */ 850 break; 851 } 852 deliver: 853 if (icmp6_notify_error(&m, off, icmp6len, code)) { 854 /* In this case, m should've been freed. */ 855 return (IPPROTO_DONE); 856 } 857 break; 858 859 badcode: 860 ICMP6STAT_INC(icp6s_badcode); 861 break; 862 863 badlen: 864 ICMP6STAT_INC(icp6s_badlen); 865 break; 866 } 867 868 /* deliver the packet to appropriate sockets */ 869 icmp6_rip6_input(&m, *offp); 870 871 return IPPROTO_DONE; 872 873 freeit: 874 m_freem(m); 875 return IPPROTO_DONE; 876} 877 878static int 879icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code) 880{ 881 INIT_VNET_INET6(curvnet); 882 struct mbuf *m = *mp; 883 struct icmp6_hdr *icmp6; 884 struct ip6_hdr *eip6; 885 u_int32_t notifymtu; 886 struct sockaddr_in6 icmp6src, icmp6dst; 887 888 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 889 ICMP6STAT_INC(icp6s_tooshort); 890 goto freeit; 891 } 892#ifndef PULLDOWN_TEST 893 IP6_EXTHDR_CHECK(m, off, 894 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1); 895 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 896#else 897 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 898 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 899 if (icmp6 == NULL) { 900 ICMP6STAT_INC(icp6s_tooshort); 901 return (-1); 902 } 903#endif 904 eip6 = (struct ip6_hdr *)(icmp6 + 1); 905 906 /* Detect the upper level protocol */ 907 { 908 void (*ctlfunc)(int, struct sockaddr *, void *); 909 u_int8_t nxt = eip6->ip6_nxt; 910 int eoff = off + sizeof(struct icmp6_hdr) + 911 sizeof(struct ip6_hdr); 912 struct ip6ctlparam ip6cp; 913 struct in6_addr *finaldst = NULL; 914 int icmp6type = icmp6->icmp6_type; 915 struct ip6_frag *fh; 916 struct ip6_rthdr *rth; 917 struct ip6_rthdr0 *rth0; 918 int rthlen; 919 920 while (1) { /* XXX: should avoid infinite loop explicitly? */ 921 struct ip6_ext *eh; 922 923 switch (nxt) { 924 case IPPROTO_HOPOPTS: 925 case IPPROTO_DSTOPTS: 926 case IPPROTO_AH: 927#ifndef PULLDOWN_TEST 928 IP6_EXTHDR_CHECK(m, 0, 929 eoff + sizeof(struct ip6_ext), -1); 930 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff); 931#else 932 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 933 eoff, sizeof(*eh)); 934 if (eh == NULL) { 935 ICMP6STAT_INC(icp6s_tooshort); 936 return (-1); 937 } 938#endif 939 940 if (nxt == IPPROTO_AH) 941 eoff += (eh->ip6e_len + 2) << 2; 942 else 943 eoff += (eh->ip6e_len + 1) << 3; 944 nxt = eh->ip6e_nxt; 945 break; 946 case IPPROTO_ROUTING: 947 /* 948 * When the erroneous packet contains a 949 * routing header, we should examine the 950 * header to determine the final destination. 951 * Otherwise, we can't properly update 952 * information that depends on the final 953 * destination (e.g. path MTU). 954 */ 955#ifndef PULLDOWN_TEST 956 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1); 957 rth = (struct ip6_rthdr *) 958 (mtod(m, caddr_t) + eoff); 959#else 960 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 961 eoff, sizeof(*rth)); 962 if (rth == NULL) { 963 ICMP6STAT_INC(icp6s_tooshort); 964 return (-1); 965 } 966#endif 967 rthlen = (rth->ip6r_len + 1) << 3; 968 /* 969 * XXX: currently there is no 970 * officially defined type other 971 * than type-0. 972 * Note that if the segment left field 973 * is 0, all intermediate hops must 974 * have been passed. 975 */ 976 if (rth->ip6r_segleft && 977 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 978 int hops; 979 980#ifndef PULLDOWN_TEST 981 IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1); 982 rth0 = (struct ip6_rthdr0 *) 983 (mtod(m, caddr_t) + eoff); 984#else 985 IP6_EXTHDR_GET(rth0, 986 struct ip6_rthdr0 *, m, 987 eoff, rthlen); 988 if (rth0 == NULL) { 989 ICMP6STAT_INC(icp6s_tooshort); 990 return (-1); 991 } 992#endif 993 /* just ignore a bogus header */ 994 if ((rth0->ip6r0_len % 2) == 0 && 995 (hops = rth0->ip6r0_len/2)) 996 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 997 } 998 eoff += rthlen; 999 nxt = rth->ip6r_nxt; 1000 break; 1001 case IPPROTO_FRAGMENT: 1002#ifndef PULLDOWN_TEST 1003 IP6_EXTHDR_CHECK(m, 0, eoff + 1004 sizeof(struct ip6_frag), -1); 1005 fh = (struct ip6_frag *)(mtod(m, caddr_t) + 1006 eoff); 1007#else 1008 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 1009 eoff, sizeof(*fh)); 1010 if (fh == NULL) { 1011 ICMP6STAT_INC(icp6s_tooshort); 1012 return (-1); 1013 } 1014#endif 1015 /* 1016 * Data after a fragment header is meaningless 1017 * unless it is the first fragment, but 1018 * we'll go to the notify label for path MTU 1019 * discovery. 1020 */ 1021 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1022 goto notify; 1023 1024 eoff += sizeof(struct ip6_frag); 1025 nxt = fh->ip6f_nxt; 1026 break; 1027 default: 1028 /* 1029 * This case includes ESP and the No Next 1030 * Header. In such cases going to the notify 1031 * label does not have any meaning 1032 * (i.e. ctlfunc will be NULL), but we go 1033 * anyway since we might have to update 1034 * path MTU information. 1035 */ 1036 goto notify; 1037 } 1038 } 1039 notify: 1040#ifndef PULLDOWN_TEST 1041 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 1042#else 1043 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1044 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1045 if (icmp6 == NULL) { 1046 ICMP6STAT_INC(icp6s_tooshort); 1047 return (-1); 1048 } 1049#endif 1050 1051 /* 1052 * retrieve parameters from the inner IPv6 header, and convert 1053 * them into sockaddr structures. 1054 * XXX: there is no guarantee that the source or destination 1055 * addresses of the inner packet are in the same scope as 1056 * the addresses of the icmp packet. But there is no other 1057 * way to determine the zone. 1058 */ 1059 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1060 1061 bzero(&icmp6dst, sizeof(icmp6dst)); 1062 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1063 icmp6dst.sin6_family = AF_INET6; 1064 if (finaldst == NULL) 1065 icmp6dst.sin6_addr = eip6->ip6_dst; 1066 else 1067 icmp6dst.sin6_addr = *finaldst; 1068 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1069 goto freeit; 1070 bzero(&icmp6src, sizeof(icmp6src)); 1071 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1072 icmp6src.sin6_family = AF_INET6; 1073 icmp6src.sin6_addr = eip6->ip6_src; 1074 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1075 goto freeit; 1076 icmp6src.sin6_flowinfo = 1077 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1078 1079 if (finaldst == NULL) 1080 finaldst = &eip6->ip6_dst; 1081 ip6cp.ip6c_m = m; 1082 ip6cp.ip6c_icmp6 = icmp6; 1083 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1084 ip6cp.ip6c_off = eoff; 1085 ip6cp.ip6c_finaldst = finaldst; 1086 ip6cp.ip6c_src = &icmp6src; 1087 ip6cp.ip6c_nxt = nxt; 1088 1089 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1090 notifymtu = ntohl(icmp6->icmp6_mtu); 1091 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1092 icmp6_mtudisc_update(&ip6cp, 1); /*XXX*/ 1093 } 1094 1095 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1096 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1097 if (ctlfunc) { 1098 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1099 &ip6cp); 1100 } 1101 } 1102 *mp = m; 1103 return (0); 1104 1105 freeit: 1106 m_freem(m); 1107 return (-1); 1108} 1109 1110void 1111icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1112{ 1113 INIT_VNET_INET6(curvnet); 1114 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1115 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1116 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1117 u_int mtu = ntohl(icmp6->icmp6_mtu); 1118 struct in_conninfo inc; 1119 1120#if 0 1121 /* 1122 * RFC2460 section 5, last paragraph. 1123 * even though minimum link MTU for IPv6 is IPV6_MMTU, 1124 * we may see ICMPv6 too big with mtu < IPV6_MMTU 1125 * due to packet translator in the middle. 1126 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for 1127 * special handling. 1128 */ 1129 if (mtu < IPV6_MMTU) 1130 return; 1131#endif 1132 1133 /* 1134 * we reject ICMPv6 too big with abnormally small value. 1135 * XXX what is the good definition of "abnormally small"? 1136 */ 1137 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) 1138 return; 1139 1140 if (!validated) 1141 return; 1142 1143 /* 1144 * In case the suggested mtu is less than IPV6_MMTU, we 1145 * only need to remember that it was for above mentioned 1146 * "alwaysfrag" case. 1147 * Try to be as close to the spec as possible. 1148 */ 1149 if (mtu < IPV6_MMTU) 1150 mtu = IPV6_MMTU - 8; 1151 1152 bzero(&inc, sizeof(inc)); 1153 inc.inc_flags |= INC_ISIPV6; 1154 inc.inc6_faddr = *dst; 1155 if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL)) 1156 return; 1157 1158 if (mtu < tcp_maxmtu6(&inc, NULL)) { 1159 tcp_hc_updatemtu(&inc, mtu); 1160 ICMP6STAT_INC(icp6s_pmtuchg); 1161 } 1162} 1163 1164/* 1165 * Process a Node Information Query packet, based on 1166 * draft-ietf-ipngwg-icmp-name-lookups-07. 1167 * 1168 * Spec incompatibilities: 1169 * - IPv6 Subject address handling 1170 * - IPv4 Subject address handling support missing 1171 * - Proxy reply (answer even if it's not for me) 1172 * - joins NI group address at in6_ifattach() time only, does not cope 1173 * with hostname changes by sethostname(3) 1174 */ 1175#define hostnamelen strlen(V_hostname) 1176static struct mbuf * 1177ni6_input(struct mbuf *m, int off) 1178{ 1179 INIT_VNET_INET6(curvnet); 1180 INIT_VPROCG(TD_TO_VPROCG(curthread)); /* XXX V_hostname needs this */ 1181 struct icmp6_nodeinfo *ni6, *nni6; 1182 struct mbuf *n = NULL; 1183 u_int16_t qtype; 1184 int subjlen; 1185 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1186 struct ni_reply_fqdn *fqdn; 1187 int addrs; /* for NI_QTYPE_NODEADDR */ 1188 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1189 struct in6_addr in6_subj; /* subject address */ 1190 struct ip6_hdr *ip6; 1191 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1192 char *subj = NULL; 1193 struct in6_ifaddr *ia6 = NULL; 1194 1195 ip6 = mtod(m, struct ip6_hdr *); 1196#ifndef PULLDOWN_TEST 1197 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off); 1198#else 1199 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1200 if (ni6 == NULL) { 1201 /* m is already reclaimed */ 1202 return (NULL); 1203 } 1204#endif 1205 1206 /* 1207 * Validate IPv6 source address. 1208 * The default configuration MUST be to refuse answering queries from 1209 * global-scope addresses according to RFC4602. 1210 * Notes: 1211 * - it's not very clear what "refuse" means; this implementation 1212 * simply drops it. 1213 * - it's not very easy to identify global-scope (unicast) addresses 1214 * since there are many prefixes for them. It should be safer 1215 * and in practice sufficient to check "all" but loopback and 1216 * link-local (note that site-local unicast was deprecated and 1217 * ULA is defined as global scope-wise) 1218 */ 1219 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 && 1220 !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) && 1221 !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) 1222 goto bad; 1223 1224 /* 1225 * Validate IPv6 destination address. 1226 * 1227 * The Responder must discard the Query without further processing 1228 * unless it is one of the Responder's unicast or anycast addresses, or 1229 * a link-local scope multicast address which the Responder has joined. 1230 * [RFC4602, Section 5.] 1231 */ 1232 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 1233 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst)) 1234 goto bad; 1235 /* else it's a link-local multicast, fine */ 1236 } else { /* unicast or anycast */ 1237 if ((ia6 = ip6_getdstifaddr(m)) == NULL) 1238 goto bad; /* XXX impossible */ 1239 1240 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) && 1241 !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) { 1242 nd6log((LOG_DEBUG, "ni6_input: ignore node info to " 1243 "a temporary address in %s:%d", 1244 __FILE__, __LINE__)); 1245 goto bad; 1246 } 1247 } 1248 1249 /* validate query Subject field. */ 1250 qtype = ntohs(ni6->ni_qtype); 1251 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1252 switch (qtype) { 1253 case NI_QTYPE_NOOP: 1254 case NI_QTYPE_SUPTYPES: 1255 /* 07 draft */ 1256 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1257 break; 1258 /* FALLTHROUGH */ 1259 case NI_QTYPE_FQDN: 1260 case NI_QTYPE_NODEADDR: 1261 case NI_QTYPE_IPV4ADDR: 1262 switch (ni6->ni_code) { 1263 case ICMP6_NI_SUBJ_IPV6: 1264#if ICMP6_NI_SUBJ_IPV6 != 0 1265 case 0: 1266#endif 1267 /* 1268 * backward compatibility - try to accept 03 draft 1269 * format, where no Subject is present. 1270 */ 1271 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1272 subjlen == 0) { 1273 oldfqdn++; 1274 break; 1275 } 1276#if ICMP6_NI_SUBJ_IPV6 != 0 1277 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1278 goto bad; 1279#endif 1280 1281 if (subjlen != sizeof(struct in6_addr)) 1282 goto bad; 1283 1284 /* 1285 * Validate Subject address. 1286 * 1287 * Not sure what exactly "address belongs to the node" 1288 * means in the spec, is it just unicast, or what? 1289 * 1290 * At this moment we consider Subject address as 1291 * "belong to the node" if the Subject address equals 1292 * to the IPv6 destination address; validation for 1293 * IPv6 destination address should have done enough 1294 * check for us. 1295 * 1296 * We do not do proxy at this moment. 1297 */ 1298 /* m_pulldown instead of copy? */ 1299 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1300 subjlen, (caddr_t)&in6_subj); 1301 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1302 goto bad; 1303 1304 subj = (char *)&in6_subj; 1305 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1306 break; 1307 1308 /* 1309 * XXX if we are to allow other cases, we should really 1310 * be careful about scope here. 1311 * basically, we should disallow queries toward IPv6 1312 * destination X with subject Y, 1313 * if scope(X) > scope(Y). 1314 * if we allow scope(X) > scope(Y), it will result in 1315 * information leakage across scope boundary. 1316 */ 1317 goto bad; 1318 1319 case ICMP6_NI_SUBJ_FQDN: 1320 /* 1321 * Validate Subject name with gethostname(3). 1322 * 1323 * The behavior may need some debate, since: 1324 * - we are not sure if the node has FQDN as 1325 * hostname (returned by gethostname(3)). 1326 * - the code does wildcard match for truncated names. 1327 * however, we are not sure if we want to perform 1328 * wildcard match, if gethostname(3) side has 1329 * truncated hostname. 1330 */ 1331 mtx_lock(&hostname_mtx); 1332 n = ni6_nametodns(V_hostname, hostnamelen, 0); 1333 mtx_unlock(&hostname_mtx); 1334 if (!n || n->m_next || n->m_len == 0) 1335 goto bad; 1336 IP6_EXTHDR_GET(subj, char *, m, 1337 off + sizeof(struct icmp6_nodeinfo), subjlen); 1338 if (subj == NULL) 1339 goto bad; 1340 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1341 n->m_len)) { 1342 goto bad; 1343 } 1344 m_freem(n); 1345 n = NULL; 1346 break; 1347 1348 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1349 default: 1350 goto bad; 1351 } 1352 break; 1353 } 1354 1355 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1356 switch (qtype) { 1357 case NI_QTYPE_FQDN: 1358 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0) 1359 goto bad; 1360 break; 1361 case NI_QTYPE_NODEADDR: 1362 case NI_QTYPE_IPV4ADDR: 1363 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0) 1364 goto bad; 1365 break; 1366 } 1367 1368 /* guess reply length */ 1369 switch (qtype) { 1370 case NI_QTYPE_NOOP: 1371 break; /* no reply data */ 1372 case NI_QTYPE_SUPTYPES: 1373 replylen += sizeof(u_int32_t); 1374 break; 1375 case NI_QTYPE_FQDN: 1376 /* XXX will append an mbuf */ 1377 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1378 break; 1379 case NI_QTYPE_NODEADDR: 1380 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj); 1381 if ((replylen += addrs * (sizeof(struct in6_addr) + 1382 sizeof(u_int32_t))) > MCLBYTES) 1383 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1384 break; 1385 case NI_QTYPE_IPV4ADDR: 1386 /* unsupported - should respond with unknown Qtype? */ 1387 break; 1388 default: 1389 /* 1390 * XXX: We must return a reply with the ICMP6 code 1391 * `unknown Qtype' in this case. However we regard the case 1392 * as an FQDN query for backward compatibility. 1393 * Older versions set a random value to this field, 1394 * so it rarely varies in the defined qtypes. 1395 * But the mechanism is not reliable... 1396 * maybe we should obsolete older versions. 1397 */ 1398 qtype = NI_QTYPE_FQDN; 1399 /* XXX will append an mbuf */ 1400 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1401 oldfqdn++; 1402 break; 1403 } 1404 1405 /* allocate an mbuf to reply. */ 1406 MGETHDR(n, M_DONTWAIT, m->m_type); 1407 if (n == NULL) { 1408 m_freem(m); 1409 return (NULL); 1410 } 1411 M_MOVE_PKTHDR(n, m); /* just for recvif */ 1412 if (replylen > MHLEN) { 1413 if (replylen > MCLBYTES) { 1414 /* 1415 * XXX: should we try to allocate more? But MCLBYTES 1416 * is probably much larger than IPV6_MMTU... 1417 */ 1418 goto bad; 1419 } 1420 MCLGET(n, M_DONTWAIT); 1421 if ((n->m_flags & M_EXT) == 0) { 1422 goto bad; 1423 } 1424 } 1425 n->m_pkthdr.len = n->m_len = replylen; 1426 1427 /* copy mbuf header and IPv6 + Node Information base headers */ 1428 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1429 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1430 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1431 1432 /* qtype dependent procedure */ 1433 switch (qtype) { 1434 case NI_QTYPE_NOOP: 1435 nni6->ni_code = ICMP6_NI_SUCCESS; 1436 nni6->ni_flags = 0; 1437 break; 1438 case NI_QTYPE_SUPTYPES: 1439 { 1440 u_int32_t v; 1441 nni6->ni_code = ICMP6_NI_SUCCESS; 1442 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1443 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1444 v = (u_int32_t)htonl(0x0000000f); 1445 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1446 break; 1447 } 1448 case NI_QTYPE_FQDN: 1449 nni6->ni_code = ICMP6_NI_SUCCESS; 1450 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1451 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo)); 1452 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1453 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1454 /* 1455 * XXX do we really have FQDN in variable "hostname"? 1456 */ 1457 mtx_lock(&hostname_mtx); 1458 n->m_next = ni6_nametodns(V_hostname, hostnamelen, oldfqdn); 1459 mtx_unlock(&hostname_mtx); 1460 if (n->m_next == NULL) 1461 goto bad; 1462 /* XXX we assume that n->m_next is not a chain */ 1463 if (n->m_next->m_next != NULL) 1464 goto bad; 1465 n->m_pkthdr.len += n->m_next->m_len; 1466 break; 1467 case NI_QTYPE_NODEADDR: 1468 { 1469 int lenlim, copied; 1470 1471 nni6->ni_code = ICMP6_NI_SUCCESS; 1472 n->m_pkthdr.len = n->m_len = 1473 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1474 lenlim = M_TRAILINGSPACE(n); 1475 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1476 /* XXX: reset mbuf length */ 1477 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1478 sizeof(struct icmp6_nodeinfo) + copied; 1479 break; 1480 } 1481 default: 1482 break; /* XXX impossible! */ 1483 } 1484 1485 nni6->ni_type = ICMP6_NI_REPLY; 1486 m_freem(m); 1487 return (n); 1488 1489 bad: 1490 m_freem(m); 1491 if (n) 1492 m_freem(n); 1493 return (NULL); 1494} 1495#undef hostnamelen 1496 1497/* 1498 * make a mbuf with DNS-encoded string. no compression support. 1499 * 1500 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1501 * treated as truncated name (two \0 at the end). this is a wild guess. 1502 * 1503 * old - return pascal string if non-zero 1504 */ 1505static struct mbuf * 1506ni6_nametodns(const char *name, int namelen, int old) 1507{ 1508 struct mbuf *m; 1509 char *cp, *ep; 1510 const char *p, *q; 1511 int i, len, nterm; 1512 1513 if (old) 1514 len = namelen + 1; 1515 else 1516 len = MCLBYTES; 1517 1518 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1519 MGET(m, M_DONTWAIT, MT_DATA); 1520 if (m && len > MLEN) { 1521 MCLGET(m, M_DONTWAIT); 1522 if ((m->m_flags & M_EXT) == 0) 1523 goto fail; 1524 } 1525 if (!m) 1526 goto fail; 1527 m->m_next = NULL; 1528 1529 if (old) { 1530 m->m_len = len; 1531 *mtod(m, char *) = namelen; 1532 bcopy(name, mtod(m, char *) + 1, namelen); 1533 return m; 1534 } else { 1535 m->m_len = 0; 1536 cp = mtod(m, char *); 1537 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1538 1539 /* if not certain about my name, return empty buffer */ 1540 if (namelen == 0) 1541 return m; 1542 1543 /* 1544 * guess if it looks like shortened hostname, or FQDN. 1545 * shortened hostname needs two trailing "\0". 1546 */ 1547 i = 0; 1548 for (p = name; p < name + namelen; p++) { 1549 if (*p && *p == '.') 1550 i++; 1551 } 1552 if (i < 2) 1553 nterm = 2; 1554 else 1555 nterm = 1; 1556 1557 p = name; 1558 while (cp < ep && p < name + namelen) { 1559 i = 0; 1560 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1561 i++; 1562 /* result does not fit into mbuf */ 1563 if (cp + i + 1 >= ep) 1564 goto fail; 1565 /* 1566 * DNS label length restriction, RFC1035 page 8. 1567 * "i == 0" case is included here to avoid returning 1568 * 0-length label on "foo..bar". 1569 */ 1570 if (i <= 0 || i >= 64) 1571 goto fail; 1572 *cp++ = i; 1573 bcopy(p, cp, i); 1574 cp += i; 1575 p = q; 1576 if (p < name + namelen && *p == '.') 1577 p++; 1578 } 1579 /* termination */ 1580 if (cp + nterm >= ep) 1581 goto fail; 1582 while (nterm-- > 0) 1583 *cp++ = '\0'; 1584 m->m_len = cp - mtod(m, char *); 1585 return m; 1586 } 1587 1588 panic("should not reach here"); 1589 /* NOTREACHED */ 1590 1591 fail: 1592 if (m) 1593 m_freem(m); 1594 return NULL; 1595} 1596 1597/* 1598 * check if two DNS-encoded string matches. takes care of truncated 1599 * form (with \0\0 at the end). no compression support. 1600 * XXX upper/lowercase match (see RFC2065) 1601 */ 1602static int 1603ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1604{ 1605 const char *a0, *b0; 1606 int l; 1607 1608 /* simplest case - need validation? */ 1609 if (alen == blen && bcmp(a, b, alen) == 0) 1610 return 1; 1611 1612 a0 = a; 1613 b0 = b; 1614 1615 /* termination is mandatory */ 1616 if (alen < 2 || blen < 2) 1617 return 0; 1618 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1619 return 0; 1620 alen--; 1621 blen--; 1622 1623 while (a - a0 < alen && b - b0 < blen) { 1624 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1625 return 0; 1626 1627 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1628 return 0; 1629 /* we don't support compression yet */ 1630 if (a[0] >= 64 || b[0] >= 64) 1631 return 0; 1632 1633 /* truncated case */ 1634 if (a[0] == 0 && a - a0 == alen - 1) 1635 return 1; 1636 if (b[0] == 0 && b - b0 == blen - 1) 1637 return 1; 1638 if (a[0] == 0 || b[0] == 0) 1639 return 0; 1640 1641 if (a[0] != b[0]) 1642 return 0; 1643 l = a[0]; 1644 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1645 return 0; 1646 if (bcmp(a + 1, b + 1, l) != 0) 1647 return 0; 1648 1649 a += 1 + l; 1650 b += 1 + l; 1651 } 1652 1653 if (a - a0 == alen && b - b0 == blen) 1654 return 1; 1655 else 1656 return 0; 1657} 1658 1659/* 1660 * calculate the number of addresses to be returned in the node info reply. 1661 */ 1662static int 1663ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp, 1664 struct in6_addr *subj) 1665{ 1666 INIT_VNET_NET(curvnet); 1667 INIT_VNET_INET6(curvnet); 1668 struct ifnet *ifp; 1669 struct in6_ifaddr *ifa6; 1670 struct ifaddr *ifa; 1671 int addrs = 0, addrsofif, iffound = 0; 1672 int niflags = ni6->ni_flags; 1673 1674 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1675 switch (ni6->ni_code) { 1676 case ICMP6_NI_SUBJ_IPV6: 1677 if (subj == NULL) /* must be impossible... */ 1678 return (0); 1679 break; 1680 default: 1681 /* 1682 * XXX: we only support IPv6 subject address for 1683 * this Qtype. 1684 */ 1685 return (0); 1686 } 1687 } 1688 1689 IFNET_RLOCK(); 1690 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) { 1691 addrsofif = 0; 1692 IF_ADDR_LOCK(ifp); 1693 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1694 if (ifa->ifa_addr->sa_family != AF_INET6) 1695 continue; 1696 ifa6 = (struct in6_ifaddr *)ifa; 1697 1698 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1699 IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr)) 1700 iffound = 1; 1701 1702 /* 1703 * IPv4-mapped addresses can only be returned by a 1704 * Node Information proxy, since they represent 1705 * addresses of IPv4-only nodes, which perforce do 1706 * not implement this protocol. 1707 * [icmp-name-lookups-07, Section 5.4] 1708 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1709 * this function at this moment. 1710 */ 1711 1712 /* What do we have to do about ::1? */ 1713 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1714 case IPV6_ADDR_SCOPE_LINKLOCAL: 1715 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1716 continue; 1717 break; 1718 case IPV6_ADDR_SCOPE_SITELOCAL: 1719 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1720 continue; 1721 break; 1722 case IPV6_ADDR_SCOPE_GLOBAL: 1723 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1724 continue; 1725 break; 1726 default: 1727 continue; 1728 } 1729 1730 /* 1731 * check if anycast is okay. 1732 * XXX: just experimental. not in the spec. 1733 */ 1734 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1735 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1736 continue; /* we need only unicast addresses */ 1737 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1738 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) { 1739 continue; 1740 } 1741 addrsofif++; /* count the address */ 1742 } 1743 IF_ADDR_UNLOCK(ifp); 1744 if (iffound) { 1745 *ifpp = ifp; 1746 IFNET_RUNLOCK(); 1747 return (addrsofif); 1748 } 1749 1750 addrs += addrsofif; 1751 } 1752 IFNET_RUNLOCK(); 1753 1754 return (addrs); 1755} 1756 1757static int 1758ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6, 1759 struct ifnet *ifp0, int resid) 1760{ 1761 INIT_VNET_NET(curvnet); 1762 INIT_VNET_INET6(curvnet); 1763 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet); 1764 struct in6_ifaddr *ifa6; 1765 struct ifaddr *ifa; 1766 struct ifnet *ifp_dep = NULL; 1767 int copied = 0, allow_deprecated = 0; 1768 u_char *cp = (u_char *)(nni6 + 1); 1769 int niflags = ni6->ni_flags; 1770 u_int32_t ltime; 1771 1772 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1773 return (0); /* needless to copy */ 1774 1775 IFNET_RLOCK(); 1776 again: 1777 1778 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) { 1779 IF_ADDR_LOCK(ifp); 1780 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1781 if (ifa->ifa_addr->sa_family != AF_INET6) 1782 continue; 1783 ifa6 = (struct in6_ifaddr *)ifa; 1784 1785 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1786 allow_deprecated == 0) { 1787 /* 1788 * prefererred address should be put before 1789 * deprecated addresses. 1790 */ 1791 1792 /* record the interface for later search */ 1793 if (ifp_dep == NULL) 1794 ifp_dep = ifp; 1795 1796 continue; 1797 } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1798 allow_deprecated != 0) 1799 continue; /* we now collect deprecated addrs */ 1800 1801 /* What do we have to do about ::1? */ 1802 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1803 case IPV6_ADDR_SCOPE_LINKLOCAL: 1804 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1805 continue; 1806 break; 1807 case IPV6_ADDR_SCOPE_SITELOCAL: 1808 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1809 continue; 1810 break; 1811 case IPV6_ADDR_SCOPE_GLOBAL: 1812 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1813 continue; 1814 break; 1815 default: 1816 continue; 1817 } 1818 1819 /* 1820 * check if anycast is okay. 1821 * XXX: just experimental. not in the spec. 1822 */ 1823 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1824 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1825 continue; 1826 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1827 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) { 1828 continue; 1829 } 1830 1831 /* now we can copy the address */ 1832 if (resid < sizeof(struct in6_addr) + 1833 sizeof(u_int32_t)) { 1834 IF_ADDR_UNLOCK(ifp); 1835 /* 1836 * We give up much more copy. 1837 * Set the truncate flag and return. 1838 */ 1839 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1840 IFNET_RUNLOCK(); 1841 return (copied); 1842 } 1843 1844 /* 1845 * Set the TTL of the address. 1846 * The TTL value should be one of the following 1847 * according to the specification: 1848 * 1849 * 1. The remaining lifetime of a DHCP lease on the 1850 * address, or 1851 * 2. The remaining Valid Lifetime of a prefix from 1852 * which the address was derived through Stateless 1853 * Autoconfiguration. 1854 * 1855 * Note that we currently do not support stateful 1856 * address configuration by DHCPv6, so the former 1857 * case can't happen. 1858 */ 1859 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1860 ltime = ND6_INFINITE_LIFETIME; 1861 else { 1862 if (ifa6->ia6_lifetime.ia6t_expire > 1863 time_second) 1864 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second); 1865 else 1866 ltime = 0; 1867 } 1868 1869 bcopy(<ime, cp, sizeof(u_int32_t)); 1870 cp += sizeof(u_int32_t); 1871 1872 /* copy the address itself */ 1873 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1874 sizeof(struct in6_addr)); 1875 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1876 cp += sizeof(struct in6_addr); 1877 1878 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1879 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1880 } 1881 IF_ADDR_UNLOCK(ifp); 1882 if (ifp0) /* we need search only on the specified IF */ 1883 break; 1884 } 1885 1886 if (allow_deprecated == 0 && ifp_dep != NULL) { 1887 ifp = ifp_dep; 1888 allow_deprecated = 1; 1889 1890 goto again; 1891 } 1892 1893 IFNET_RUNLOCK(); 1894 1895 return (copied); 1896} 1897 1898/* 1899 * XXX almost dup'ed code with rip6_input. 1900 */ 1901static int 1902icmp6_rip6_input(struct mbuf **mp, int off) 1903{ 1904 INIT_VNET_INET(curvnet); 1905 INIT_VNET_INET6(curvnet); 1906 struct mbuf *m = *mp; 1907 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1908 struct inpcb *in6p; 1909 struct inpcb *last = NULL; 1910 struct sockaddr_in6 fromsa; 1911 struct icmp6_hdr *icmp6; 1912 struct mbuf *opts = NULL; 1913 1914#ifndef PULLDOWN_TEST 1915 /* this is assumed to be safe. */ 1916 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 1917#else 1918 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1919 if (icmp6 == NULL) { 1920 /* m is already reclaimed */ 1921 return (IPPROTO_DONE); 1922 } 1923#endif 1924 1925 /* 1926 * XXX: the address may have embedded scope zone ID, which should be 1927 * hidden from applications. 1928 */ 1929 bzero(&fromsa, sizeof(fromsa)); 1930 fromsa.sin6_family = AF_INET6; 1931 fromsa.sin6_len = sizeof(struct sockaddr_in6); 1932 fromsa.sin6_addr = ip6->ip6_src; 1933 if (sa6_recoverscope(&fromsa)) { 1934 m_freem(m); 1935 return (IPPROTO_DONE); 1936 } 1937 1938 INP_INFO_RLOCK(&V_ripcbinfo); 1939 LIST_FOREACH(in6p, &V_ripcb, inp_list) { 1940 if ((in6p->inp_vflag & INP_IPV6) == 0) 1941 continue; 1942 if (in6p->inp_ip_p != IPPROTO_ICMPV6) 1943 continue; 1944 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1945 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1946 continue; 1947 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1948 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1949 continue; 1950 INP_RLOCK(in6p); 1951 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1952 in6p->in6p_icmp6filt)) { 1953 INP_RUNLOCK(in6p); 1954 continue; 1955 } 1956 if (last != NULL) { 1957 struct mbuf *n = NULL; 1958 1959 /* 1960 * Recent network drivers tend to allocate a single 1961 * mbuf cluster, rather than to make a couple of 1962 * mbufs without clusters. Also, since the IPv6 code 1963 * path tries to avoid m_pullup(), it is highly 1964 * probable that we still have an mbuf cluster here 1965 * even though the necessary length can be stored in an 1966 * mbuf's internal buffer. 1967 * Meanwhile, the default size of the receive socket 1968 * buffer for raw sockets is not so large. This means 1969 * the possibility of packet loss is relatively higher 1970 * than before. To avoid this scenario, we copy the 1971 * received data to a separate mbuf that does not use 1972 * a cluster, if possible. 1973 * XXX: it is better to copy the data after stripping 1974 * intermediate headers. 1975 */ 1976 if ((m->m_flags & M_EXT) && m->m_next == NULL && 1977 m->m_len <= MHLEN) { 1978 MGET(n, M_DONTWAIT, m->m_type); 1979 if (n != NULL) { 1980 if (m_dup_pkthdr(n, m, M_NOWAIT)) { 1981 bcopy(m->m_data, n->m_data, 1982 m->m_len); 1983 n->m_len = m->m_len; 1984 } else { 1985 m_free(n); 1986 n = NULL; 1987 } 1988 } 1989 } 1990 if (n != NULL || 1991 (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1992 if (last->inp_flags & INP_CONTROLOPTS) 1993 ip6_savecontrol(last, n, &opts); 1994 /* strip intermediate headers */ 1995 m_adj(n, off); 1996 SOCKBUF_LOCK(&last->inp_socket->so_rcv); 1997 if (sbappendaddr_locked( 1998 &last->inp_socket->so_rcv, 1999 (struct sockaddr *)&fromsa, n, opts) 2000 == 0) { 2001 /* should notify about lost packet */ 2002 m_freem(n); 2003 if (opts) { 2004 m_freem(opts); 2005 } 2006 SOCKBUF_UNLOCK( 2007 &last->inp_socket->so_rcv); 2008 } else 2009 sorwakeup_locked(last->inp_socket); 2010 opts = NULL; 2011 } 2012 INP_RUNLOCK(last); 2013 } 2014 last = in6p; 2015 } 2016 INP_INFO_RUNLOCK(&V_ripcbinfo); 2017 if (last != NULL) { 2018 if (last->inp_flags & INP_CONTROLOPTS) 2019 ip6_savecontrol(last, m, &opts); 2020 /* strip intermediate headers */ 2021 m_adj(m, off); 2022 2023 /* avoid using mbuf clusters if possible (see above) */ 2024 if ((m->m_flags & M_EXT) && m->m_next == NULL && 2025 m->m_len <= MHLEN) { 2026 struct mbuf *n; 2027 2028 MGET(n, M_DONTWAIT, m->m_type); 2029 if (n != NULL) { 2030 if (m_dup_pkthdr(n, m, M_NOWAIT)) { 2031 bcopy(m->m_data, n->m_data, m->m_len); 2032 n->m_len = m->m_len; 2033 2034 m_freem(m); 2035 m = n; 2036 } else { 2037 m_freem(n); 2038 n = NULL; 2039 } 2040 } 2041 } 2042 SOCKBUF_LOCK(&last->inp_socket->so_rcv); 2043 if (sbappendaddr_locked(&last->inp_socket->so_rcv, 2044 (struct sockaddr *)&fromsa, m, opts) == 0) { 2045 m_freem(m); 2046 if (opts) 2047 m_freem(opts); 2048 SOCKBUF_UNLOCK(&last->inp_socket->so_rcv); 2049 } else 2050 sorwakeup_locked(last->inp_socket); 2051 INP_RUNLOCK(last); 2052 } else { 2053 m_freem(m); 2054 IP6STAT_DEC(ip6s_delivered); 2055 } 2056 return IPPROTO_DONE; 2057} 2058 2059/* 2060 * Reflect the ip6 packet back to the source. 2061 * OFF points to the icmp6 header, counted from the top of the mbuf. 2062 */ 2063void 2064icmp6_reflect(struct mbuf *m, size_t off) 2065{ 2066 INIT_VNET_INET6(curvnet); 2067 struct ip6_hdr *ip6; 2068 struct icmp6_hdr *icmp6; 2069 struct in6_ifaddr *ia; 2070 int plen; 2071 int type, code; 2072 struct ifnet *outif = NULL; 2073 struct in6_addr origdst, *src = NULL; 2074 2075 /* too short to reflect */ 2076 if (off < sizeof(struct ip6_hdr)) { 2077 nd6log((LOG_DEBUG, 2078 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 2079 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2080 __FILE__, __LINE__)); 2081 goto bad; 2082 } 2083 2084 /* 2085 * If there are extra headers between IPv6 and ICMPv6, strip 2086 * off that header first. 2087 */ 2088#ifdef DIAGNOSTIC 2089 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2090 panic("assumption failed in icmp6_reflect"); 2091#endif 2092 if (off > sizeof(struct ip6_hdr)) { 2093 size_t l; 2094 struct ip6_hdr nip6; 2095 2096 l = off - sizeof(struct ip6_hdr); 2097 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2098 m_adj(m, l); 2099 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2100 if (m->m_len < l) { 2101 if ((m = m_pullup(m, l)) == NULL) 2102 return; 2103 } 2104 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2105 } else /* off == sizeof(struct ip6_hdr) */ { 2106 size_t l; 2107 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2108 if (m->m_len < l) { 2109 if ((m = m_pullup(m, l)) == NULL) 2110 return; 2111 } 2112 } 2113 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2114 ip6 = mtod(m, struct ip6_hdr *); 2115 ip6->ip6_nxt = IPPROTO_ICMPV6; 2116 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2117 type = icmp6->icmp6_type; /* keep type for statistics */ 2118 code = icmp6->icmp6_code; /* ditto. */ 2119 2120 origdst = ip6->ip6_dst; 2121 /* 2122 * ip6_input() drops a packet if its src is multicast. 2123 * So, the src is never multicast. 2124 */ 2125 ip6->ip6_dst = ip6->ip6_src; 2126 2127 /* 2128 * If the incoming packet was addressed directly to us (i.e. unicast), 2129 * use dst as the src for the reply. 2130 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2131 * (for example) when we encounter an error while forwarding procedure 2132 * destined to a duplicated address of ours. 2133 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2134 * procedure of an outgoing packet of our own, in which case we need 2135 * to search in the ifaddr list. 2136 */ 2137 if (!IN6_IS_ADDR_MULTICAST(&origdst)) { 2138 if ((ia = ip6_getdstifaddr(m))) { 2139 if (!(ia->ia6_flags & 2140 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) 2141 src = &ia->ia_addr.sin6_addr; 2142 } else { 2143 struct sockaddr_in6 d; 2144 2145 bzero(&d, sizeof(d)); 2146 d.sin6_family = AF_INET6; 2147 d.sin6_len = sizeof(d); 2148 d.sin6_addr = origdst; 2149 ia = (struct in6_ifaddr *) 2150 ifa_ifwithaddr((struct sockaddr *)&d); 2151 if (ia && 2152 !(ia->ia6_flags & 2153 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) { 2154 src = &ia->ia_addr.sin6_addr; 2155 } 2156 } 2157 } 2158 2159 if (src == NULL) { 2160 int e; 2161 struct sockaddr_in6 sin6; 2162 struct route_in6 ro; 2163 2164 /* 2165 * This case matches to multicasts, our anycast, or unicasts 2166 * that we do not own. Select a source address based on the 2167 * source address of the erroneous packet. 2168 */ 2169 bzero(&sin6, sizeof(sin6)); 2170 sin6.sin6_family = AF_INET6; 2171 sin6.sin6_len = sizeof(sin6); 2172 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */ 2173 2174 bzero(&ro, sizeof(ro)); 2175 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2176 if (ro.ro_rt) 2177 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2178 if (src == NULL) { 2179 char ip6buf[INET6_ADDRSTRLEN]; 2180 nd6log((LOG_DEBUG, 2181 "icmp6_reflect: source can't be determined: " 2182 "dst=%s, error=%d\n", 2183 ip6_sprintf(ip6buf, &sin6.sin6_addr), e)); 2184 goto bad; 2185 } 2186 } 2187 2188 ip6->ip6_src = *src; 2189 ip6->ip6_flow = 0; 2190 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2191 ip6->ip6_vfc |= IPV6_VERSION; 2192 ip6->ip6_nxt = IPPROTO_ICMPV6; 2193 if (outif) 2194 ip6->ip6_hlim = ND_IFINFO(outif)->chlim; 2195 else if (m->m_pkthdr.rcvif) { 2196 /* XXX: This may not be the outgoing interface */ 2197 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2198 } else 2199 ip6->ip6_hlim = V_ip6_defhlim; 2200 2201 icmp6->icmp6_cksum = 0; 2202 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2203 sizeof(struct ip6_hdr), plen); 2204 2205 /* 2206 * XXX option handling 2207 */ 2208 2209 m->m_flags &= ~(M_BCAST|M_MCAST); 2210 2211 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL); 2212 if (outif) 2213 icmp6_ifoutstat_inc(outif, type, code); 2214 2215 return; 2216 2217 bad: 2218 m_freem(m); 2219 return; 2220} 2221 2222void 2223icmp6_fasttimo(void) 2224{ 2225 2226 mld_fasttimo(); 2227} 2228 2229void 2230icmp6_slowtimo(void) 2231{ 2232 2233 mld_slowtimo(); 2234} 2235 2236static const char * 2237icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2238 struct in6_addr *tgt6) 2239{ 2240 static char buf[1024]; 2241 char ip6bufs[INET6_ADDRSTRLEN]; 2242 char ip6bufd[INET6_ADDRSTRLEN]; 2243 char ip6buft[INET6_ADDRSTRLEN]; 2244 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2245 ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6), 2246 ip6_sprintf(ip6buft, tgt6)); 2247 return buf; 2248} 2249 2250void 2251icmp6_redirect_input(struct mbuf *m, int off) 2252{ 2253 INIT_VNET_INET6(curvnet); 2254 struct ifnet *ifp; 2255 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2256 struct nd_redirect *nd_rd; 2257 int icmp6len = ntohs(ip6->ip6_plen); 2258 char *lladdr = NULL; 2259 int lladdrlen = 0; 2260 u_char *redirhdr = NULL; 2261 int redirhdrlen = 0; 2262 struct rtentry *rt = NULL; 2263 int is_router; 2264 int is_onlink; 2265 struct in6_addr src6 = ip6->ip6_src; 2266 struct in6_addr redtgt6; 2267 struct in6_addr reddst6; 2268 union nd_opts ndopts; 2269 char ip6buf[INET6_ADDRSTRLEN]; 2270 2271 if (!m) 2272 return; 2273 2274 ifp = m->m_pkthdr.rcvif; 2275 2276 if (!ifp) 2277 return; 2278 2279 /* XXX if we are router, we don't update route by icmp6 redirect */ 2280 if (V_ip6_forwarding) 2281 goto freeit; 2282 if (!V_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_INC(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_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2299 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2300 goto freeit; 2301 } 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", 2308 ip6_sprintf(ip6buf, &src6))); 2309 goto bad; 2310 } 2311 if (ip6->ip6_hlim != 255) { 2312 nd6log((LOG_ERR, 2313 "ICMP6 redirect sent from %s rejected; " 2314 "hlim=%d (must be 255)\n", 2315 ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim)); 2316 goto bad; 2317 } 2318 { 2319 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2320 struct sockaddr_in6 sin6; 2321 struct in6_addr *gw6; 2322 2323 bzero(&sin6, sizeof(sin6)); 2324 sin6.sin6_family = AF_INET6; 2325 sin6.sin6_len = sizeof(struct sockaddr_in6); 2326 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2327 rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL); 2328 if (rt) { 2329 if (rt->rt_gateway == NULL || 2330 rt->rt_gateway->sa_family != AF_INET6) { 2331 nd6log((LOG_ERR, 2332 "ICMP6 redirect rejected; no route " 2333 "with inet6 gateway found for redirect dst: %s\n", 2334 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2335 RTFREE_LOCKED(rt); 2336 goto bad; 2337 } 2338 2339 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2340 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2341 nd6log((LOG_ERR, 2342 "ICMP6 redirect rejected; " 2343 "not equal to gw-for-src=%s (must be same): " 2344 "%s\n", 2345 ip6_sprintf(ip6buf, gw6), 2346 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2347 RTFREE_LOCKED(rt); 2348 goto bad; 2349 } 2350 } else { 2351 nd6log((LOG_ERR, 2352 "ICMP6 redirect rejected; " 2353 "no route found for redirect dst: %s\n", 2354 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2355 goto bad; 2356 } 2357 RTFREE_LOCKED(rt); 2358 rt = NULL; 2359 } 2360 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2361 nd6log((LOG_ERR, 2362 "ICMP6 redirect rejected; " 2363 "redirect dst must be unicast: %s\n", 2364 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2365 goto bad; 2366 } 2367 2368 is_router = is_onlink = 0; 2369 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2370 is_router = 1; /* router case */ 2371 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2372 is_onlink = 1; /* on-link destination case */ 2373 if (!is_router && !is_onlink) { 2374 nd6log((LOG_ERR, 2375 "ICMP6 redirect rejected; " 2376 "neither router case nor onlink case: %s\n", 2377 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2378 goto bad; 2379 } 2380 /* validation passed */ 2381 2382 icmp6len -= sizeof(*nd_rd); 2383 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2384 if (nd6_options(&ndopts) < 0) { 2385 nd6log((LOG_INFO, "icmp6_redirect_input: " 2386 "invalid ND option, rejected: %s\n", 2387 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2388 /* nd6_options have incremented stats */ 2389 goto freeit; 2390 } 2391 2392 if (ndopts.nd_opts_tgt_lladdr) { 2393 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2394 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2395 } 2396 2397 if (ndopts.nd_opts_rh) { 2398 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len; 2399 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */ 2400 } 2401 2402 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2403 nd6log((LOG_INFO, 2404 "icmp6_redirect_input: lladdrlen mismatch for %s " 2405 "(if %d, icmp6 packet %d): %s\n", 2406 ip6_sprintf(ip6buf, &redtgt6), 2407 ifp->if_addrlen, lladdrlen - 2, 2408 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2409 goto bad; 2410 } 2411 2412 /* RFC 2461 8.3 */ 2413 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2414 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2415 2416 if (!is_onlink) { /* better router case. perform rtredirect. */ 2417 /* perform rtredirect */ 2418 struct sockaddr_in6 sdst; 2419 struct sockaddr_in6 sgw; 2420 struct sockaddr_in6 ssrc; 2421 2422 bzero(&sdst, sizeof(sdst)); 2423 bzero(&sgw, sizeof(sgw)); 2424 bzero(&ssrc, sizeof(ssrc)); 2425 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2426 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2427 sizeof(struct sockaddr_in6); 2428 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2429 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2430 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2431 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2432 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2433 (struct sockaddr *)&ssrc); 2434 } 2435 /* finally update cached route in each socket via pfctlinput */ 2436 { 2437 struct sockaddr_in6 sdst; 2438 2439 bzero(&sdst, sizeof(sdst)); 2440 sdst.sin6_family = AF_INET6; 2441 sdst.sin6_len = sizeof(struct sockaddr_in6); 2442 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2443 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2444#ifdef IPSEC 2445 key_sa_routechange((struct sockaddr *)&sdst); 2446#endif /* IPSEC */ 2447 } 2448 2449 freeit: 2450 m_freem(m); 2451 return; 2452 2453 bad: 2454 ICMP6STAT_INC(icp6s_badredirect); 2455 m_freem(m); 2456} 2457 2458void 2459icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2460{ 2461 INIT_VNET_INET6(curvnet); 2462 struct ifnet *ifp; /* my outgoing interface */ 2463 struct in6_addr *ifp_ll6; 2464 struct in6_addr *router_ll6; 2465 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2466 struct mbuf *m = NULL; /* newly allocated one */ 2467 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2468 struct nd_redirect *nd_rd; 2469 struct llentry *ln = NULL; 2470 size_t maxlen; 2471 u_char *p; 2472 struct ifnet *outif = NULL; 2473 struct sockaddr_in6 src_sa; 2474 2475 icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2476 2477 /* if we are not router, we don't send icmp6 redirect */ 2478 if (!V_ip6_forwarding) 2479 goto fail; 2480 2481 /* sanity check */ 2482 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2483 goto fail; 2484 2485 /* 2486 * Address check: 2487 * the source address must identify a neighbor, and 2488 * the destination address must not be a multicast address 2489 * [RFC 2461, sec 8.2] 2490 */ 2491 sip6 = mtod(m0, struct ip6_hdr *); 2492 bzero(&src_sa, sizeof(src_sa)); 2493 src_sa.sin6_family = AF_INET6; 2494 src_sa.sin6_len = sizeof(src_sa); 2495 src_sa.sin6_addr = sip6->ip6_src; 2496 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2497 goto fail; 2498 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2499 goto fail; /* what should we do here? */ 2500 2501 /* rate limit */ 2502 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2503 goto fail; 2504 2505 /* 2506 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2507 * we almost always ask for an mbuf cluster for simplicity. 2508 * (MHLEN < IPV6_MMTU is almost always true) 2509 */ 2510#if IPV6_MMTU >= MCLBYTES 2511# error assumption failed about IPV6_MMTU and MCLBYTES 2512#endif 2513 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2514 if (m && IPV6_MMTU >= MHLEN) 2515 MCLGET(m, M_DONTWAIT); 2516 if (!m) 2517 goto fail; 2518 m->m_pkthdr.rcvif = NULL; 2519 m->m_len = 0; 2520 maxlen = M_TRAILINGSPACE(m); 2521 maxlen = min(IPV6_MMTU, maxlen); 2522 /* just for safety */ 2523 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2524 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2525 goto fail; 2526 } 2527 2528 { 2529 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2530 struct in6_ifaddr *ia; 2531 if ((ia = in6ifa_ifpforlinklocal(ifp, 2532 IN6_IFF_NOTREADY| 2533 IN6_IFF_ANYCAST)) == NULL) 2534 goto fail; 2535 ifp_ll6 = &ia->ia_addr.sin6_addr; 2536 } 2537 2538 /* get ip6 linklocal address for the router. */ 2539 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2540 struct sockaddr_in6 *sin6; 2541 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2542 router_ll6 = &sin6->sin6_addr; 2543 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6)) 2544 router_ll6 = (struct in6_addr *)NULL; 2545 } else 2546 router_ll6 = (struct in6_addr *)NULL; 2547 2548 /* ip6 */ 2549 ip6 = mtod(m, struct ip6_hdr *); 2550 ip6->ip6_flow = 0; 2551 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2552 ip6->ip6_vfc |= IPV6_VERSION; 2553 /* ip6->ip6_plen will be set later */ 2554 ip6->ip6_nxt = IPPROTO_ICMPV6; 2555 ip6->ip6_hlim = 255; 2556 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2557 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2558 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2559 2560 /* ND Redirect */ 2561 nd_rd = (struct nd_redirect *)(ip6 + 1); 2562 nd_rd->nd_rd_type = ND_REDIRECT; 2563 nd_rd->nd_rd_code = 0; 2564 nd_rd->nd_rd_reserved = 0; 2565 if (rt->rt_flags & RTF_GATEWAY) { 2566 /* 2567 * nd_rd->nd_rd_target must be a link-local address in 2568 * better router cases. 2569 */ 2570 if (!router_ll6) 2571 goto fail; 2572 bcopy(router_ll6, &nd_rd->nd_rd_target, 2573 sizeof(nd_rd->nd_rd_target)); 2574 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2575 sizeof(nd_rd->nd_rd_dst)); 2576 } else { 2577 /* make sure redtgt == reddst */ 2578 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2579 sizeof(nd_rd->nd_rd_target)); 2580 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2581 sizeof(nd_rd->nd_rd_dst)); 2582 } 2583 2584 p = (u_char *)(nd_rd + 1); 2585 2586 if (!router_ll6) 2587 goto nolladdropt; 2588 2589 { 2590 /* target lladdr option */ 2591 int len; 2592 struct nd_opt_hdr *nd_opt; 2593 char *lladdr; 2594 2595 IF_AFDATA_LOCK(ifp); 2596 ln = nd6_lookup(router_ll6, 0, ifp); 2597 IF_AFDATA_UNLOCK(ifp); 2598 if (ln == NULL) 2599 goto nolladdropt; 2600 2601 len = sizeof(*nd_opt) + ifp->if_addrlen; 2602 len = (len + 7) & ~7; /* round by 8 */ 2603 /* safety check */ 2604 if (len + (p - (u_char *)ip6) > maxlen) 2605 goto nolladdropt; 2606 2607 if (ln->la_flags & LLE_VALID) { 2608 nd_opt = (struct nd_opt_hdr *)p; 2609 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2610 nd_opt->nd_opt_len = len >> 3; 2611 lladdr = (char *)(nd_opt + 1); 2612 bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen); 2613 p += len; 2614 } 2615 } 2616nolladdropt: 2617 if (ln != NULL) 2618 LLE_RUNLOCK(ln); 2619 2620 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2621 2622 /* just to be safe */ 2623#ifdef M_DECRYPTED /*not openbsd*/ 2624 if (m0->m_flags & M_DECRYPTED) 2625 goto noredhdropt; 2626#endif 2627 if (p - (u_char *)ip6 > maxlen) 2628 goto noredhdropt; 2629 2630 { 2631 /* redirected header option */ 2632 int len; 2633 struct nd_opt_rd_hdr *nd_opt_rh; 2634 2635 /* 2636 * compute the maximum size for icmp6 redirect header option. 2637 * XXX room for auth header? 2638 */ 2639 len = maxlen - (p - (u_char *)ip6); 2640 len &= ~7; 2641 2642 /* This is just for simplicity. */ 2643 if (m0->m_pkthdr.len != m0->m_len) { 2644 if (m0->m_next) { 2645 m_freem(m0->m_next); 2646 m0->m_next = NULL; 2647 } 2648 m0->m_pkthdr.len = m0->m_len; 2649 } 2650 2651 /* 2652 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2653 * about padding/truncate rule for the original IP packet. 2654 * From the discussion on IPv6imp in Feb 1999, 2655 * the consensus was: 2656 * - "attach as much as possible" is the goal 2657 * - pad if not aligned (original size can be guessed by 2658 * original ip6 header) 2659 * Following code adds the padding if it is simple enough, 2660 * and truncates if not. 2661 */ 2662 if (m0->m_next || m0->m_pkthdr.len != m0->m_len) 2663 panic("assumption failed in %s:%d", __FILE__, 2664 __LINE__); 2665 2666 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2667 /* not enough room, truncate */ 2668 m0->m_pkthdr.len = m0->m_len = len - 2669 sizeof(*nd_opt_rh); 2670 } else { 2671 /* enough room, pad or truncate */ 2672 size_t extra; 2673 2674 extra = m0->m_pkthdr.len % 8; 2675 if (extra) { 2676 /* pad if easy enough, truncate if not */ 2677 if (8 - extra <= M_TRAILINGSPACE(m0)) { 2678 /* pad */ 2679 m0->m_len += (8 - extra); 2680 m0->m_pkthdr.len += (8 - extra); 2681 } else { 2682 /* truncate */ 2683 m0->m_pkthdr.len -= extra; 2684 m0->m_len -= extra; 2685 } 2686 } 2687 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2688 m0->m_pkthdr.len = m0->m_len = len - 2689 sizeof(*nd_opt_rh); 2690 } 2691 2692 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2693 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2694 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2695 nd_opt_rh->nd_opt_rh_len = len >> 3; 2696 p += sizeof(*nd_opt_rh); 2697 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2698 2699 /* connect m0 to m */ 2700 m_tag_delete_chain(m0, NULL); 2701 m0->m_flags &= ~M_PKTHDR; 2702 m->m_next = m0; 2703 m->m_pkthdr.len = m->m_len + m0->m_len; 2704 m0 = NULL; 2705 } 2706noredhdropt:; 2707 if (m0) { 2708 m_freem(m0); 2709 m0 = NULL; 2710 } 2711 2712 /* XXX: clear embedded link IDs in the inner header */ 2713 in6_clearscope(&sip6->ip6_src); 2714 in6_clearscope(&sip6->ip6_dst); 2715 in6_clearscope(&nd_rd->nd_rd_target); 2716 in6_clearscope(&nd_rd->nd_rd_dst); 2717 2718 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2719 2720 nd_rd->nd_rd_cksum = 0; 2721 nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2722 sizeof(*ip6), ntohs(ip6->ip6_plen)); 2723 2724 /* send the packet to outside... */ 2725 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL); 2726 if (outif) { 2727 icmp6_ifstat_inc(outif, ifs6_out_msg); 2728 icmp6_ifstat_inc(outif, ifs6_out_redirect); 2729 } 2730 ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]); 2731 2732 return; 2733 2734fail: 2735 if (m) 2736 m_freem(m); 2737 if (m0) 2738 m_freem(m0); 2739} 2740 2741/* 2742 * ICMPv6 socket option processing. 2743 */ 2744int 2745icmp6_ctloutput(struct socket *so, struct sockopt *sopt) 2746{ 2747 int error = 0; 2748 int optlen; 2749 struct inpcb *inp = sotoinpcb(so); 2750 int level, op, optname; 2751 2752 if (sopt) { 2753 level = sopt->sopt_level; 2754 op = sopt->sopt_dir; 2755 optname = sopt->sopt_name; 2756 optlen = sopt->sopt_valsize; 2757 } else 2758 level = op = optname = optlen = 0; 2759 2760 if (level != IPPROTO_ICMPV6) { 2761 return EINVAL; 2762 } 2763 2764 switch (op) { 2765 case PRCO_SETOPT: 2766 switch (optname) { 2767 case ICMP6_FILTER: 2768 { 2769 struct icmp6_filter ic6f; 2770 2771 if (optlen != sizeof(ic6f)) { 2772 error = EMSGSIZE; 2773 break; 2774 } 2775 error = sooptcopyin(sopt, &ic6f, optlen, optlen); 2776 if (error == 0) { 2777 INP_WLOCK(inp); 2778 *inp->in6p_icmp6filt = ic6f; 2779 INP_WUNLOCK(inp); 2780 } 2781 break; 2782 } 2783 2784 default: 2785 error = ENOPROTOOPT; 2786 break; 2787 } 2788 break; 2789 2790 case PRCO_GETOPT: 2791 switch (optname) { 2792 case ICMP6_FILTER: 2793 { 2794 struct icmp6_filter ic6f; 2795 2796 INP_RLOCK(inp); 2797 ic6f = *inp->in6p_icmp6filt; 2798 INP_RUNLOCK(inp); 2799 error = sooptcopyout(sopt, &ic6f, sizeof(ic6f)); 2800 break; 2801 } 2802 2803 default: 2804 error = ENOPROTOOPT; 2805 break; 2806 } 2807 break; 2808 } 2809 2810 return (error); 2811} 2812 2813/* 2814 * Perform rate limit check. 2815 * Returns 0 if it is okay to send the icmp6 packet. 2816 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2817 * limitation. 2818 * 2819 * XXX per-destination/type check necessary? 2820 * 2821 * dst - not used at this moment 2822 * type - not used at this moment 2823 * code - not used at this moment 2824 */ 2825static int 2826icmp6_ratelimit(const struct in6_addr *dst, const int type, 2827 const int code) 2828{ 2829 INIT_VNET_INET6(curvnet); 2830 int ret; 2831 2832 ret = 0; /* okay to send */ 2833 2834 /* PPS limit */ 2835 if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count, 2836 V_icmp6errppslim)) { 2837 /* The packet is subject to rate limit */ 2838 ret++; 2839 } 2840 2841 return ret; 2842} 2843