icmp6.c revision 193066
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 193066 2009-05-29 21:27:12Z jamie $"); 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/jail.h> 74#include <sys/kernel.h> 75#include <sys/lock.h> 76#include <sys/malloc.h> 77#include <sys/mbuf.h> 78#include <sys/protosw.h> 79#include <sys/signalvar.h> 80#include <sys/socket.h> 81#include <sys/socketvar.h> 82#include <sys/sx.h> 83#include <sys/syslog.h> 84#include <sys/systm.h> 85#include <sys/time.h> 86#include <sys/vimage.h> 87 88#include <net/if.h> 89#include <net/if_dl.h> 90#include <net/if_llatbl.h> 91#include <net/if_types.h> 92#include <net/route.h> 93#include <net/vnet.h> 94 95#include <netinet/in.h> 96#include <netinet/in_pcb.h> 97#include <netinet/in_var.h> 98#include <netinet/ip6.h> 99#include <netinet/icmp6.h> 100#include <netinet/tcp_var.h> 101#include <netinet/vinet.h> 102 103#include <netinet6/in6_ifattach.h> 104#include <netinet6/in6_pcb.h> 105#include <netinet6/ip6protosw.h> 106#include <netinet6/ip6_var.h> 107#include <netinet6/scope6_var.h> 108#include <netinet6/mld6_var.h> 109#include <netinet6/nd6.h> 110#include <netinet6/vinet6.h> 111 112#ifdef IPSEC 113#include <netipsec/ipsec.h> 114#include <netipsec/key.h> 115#endif 116 117extern struct domain inet6domain; 118 119#ifdef VIMAGE_GLOBALS 120extern struct inpcbinfo ripcbinfo; 121extern struct inpcbhead ripcb; 122extern int icmp6errppslim; 123extern int icmp6_nodeinfo; 124 125struct icmp6stat icmp6stat; 126static int icmp6errpps_count; 127static struct timeval icmp6errppslim_last; 128#endif 129 130static void icmp6_errcount(struct icmp6errstat *, int, int); 131static int icmp6_rip6_input(struct mbuf **, int); 132static int icmp6_ratelimit(const struct in6_addr *, const int, const int); 133static const char *icmp6_redirect_diag __P((struct in6_addr *, 134 struct in6_addr *, struct in6_addr *)); 135static struct mbuf *ni6_input(struct mbuf *, int); 136static struct mbuf *ni6_nametodns(const char *, int, int); 137static int ni6_dnsmatch(const char *, int, const char *, int); 138static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *, 139 struct ifnet **, struct in6_addr *)); 140static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 141 struct ifnet *, int)); 142static int icmp6_notify_error(struct mbuf **, int, int, int); 143 144 145void 146icmp6_init(void) 147{ 148 INIT_VNET_INET6(curvnet); 149 150 V_icmp6errpps_count = 0; 151} 152 153static void 154icmp6_errcount(struct icmp6errstat *stat, int type, int code) 155{ 156 switch (type) { 157 case ICMP6_DST_UNREACH: 158 switch (code) { 159 case ICMP6_DST_UNREACH_NOROUTE: 160 stat->icp6errs_dst_unreach_noroute++; 161 return; 162 case ICMP6_DST_UNREACH_ADMIN: 163 stat->icp6errs_dst_unreach_admin++; 164 return; 165 case ICMP6_DST_UNREACH_BEYONDSCOPE: 166 stat->icp6errs_dst_unreach_beyondscope++; 167 return; 168 case ICMP6_DST_UNREACH_ADDR: 169 stat->icp6errs_dst_unreach_addr++; 170 return; 171 case ICMP6_DST_UNREACH_NOPORT: 172 stat->icp6errs_dst_unreach_noport++; 173 return; 174 } 175 break; 176 case ICMP6_PACKET_TOO_BIG: 177 stat->icp6errs_packet_too_big++; 178 return; 179 case ICMP6_TIME_EXCEEDED: 180 switch (code) { 181 case ICMP6_TIME_EXCEED_TRANSIT: 182 stat->icp6errs_time_exceed_transit++; 183 return; 184 case ICMP6_TIME_EXCEED_REASSEMBLY: 185 stat->icp6errs_time_exceed_reassembly++; 186 return; 187 } 188 break; 189 case ICMP6_PARAM_PROB: 190 switch (code) { 191 case ICMP6_PARAMPROB_HEADER: 192 stat->icp6errs_paramprob_header++; 193 return; 194 case ICMP6_PARAMPROB_NEXTHEADER: 195 stat->icp6errs_paramprob_nextheader++; 196 return; 197 case ICMP6_PARAMPROB_OPTION: 198 stat->icp6errs_paramprob_option++; 199 return; 200 } 201 break; 202 case ND_REDIRECT: 203 stat->icp6errs_redirect++; 204 return; 205 } 206 stat->icp6errs_unknown++; 207} 208 209/* 210 * A wrapper function for icmp6_error() necessary when the erroneous packet 211 * may not contain enough scope zone information. 212 */ 213void 214icmp6_error2(struct mbuf *m, int type, int code, int param, 215 struct ifnet *ifp) 216{ 217 INIT_VNET_INET6(curvnet); 218 struct ip6_hdr *ip6; 219 220 if (ifp == NULL) 221 return; 222 223#ifndef PULLDOWN_TEST 224 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), ); 225#else 226 if (m->m_len < sizeof(struct ip6_hdr)) { 227 m = m_pullup(m, sizeof(struct ip6_hdr)); 228 if (m == NULL) 229 return; 230 } 231#endif 232 233 ip6 = mtod(m, struct ip6_hdr *); 234 235 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 236 return; 237 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 238 return; 239 240 icmp6_error(m, type, code, param); 241} 242 243/* 244 * Generate an error packet of type error in response to bad IP6 packet. 245 */ 246void 247icmp6_error(struct mbuf *m, int type, int code, int param) 248{ 249 INIT_VNET_INET6(curvnet); 250 struct ip6_hdr *oip6, *nip6; 251 struct icmp6_hdr *icmp6; 252 u_int preplen; 253 int off; 254 int nxt; 255 256 ICMP6STAT_INC(icp6s_error); 257 258 /* count per-type-code statistics */ 259 icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, type, code); 260 261#ifdef M_DECRYPTED /*not openbsd*/ 262 if (m->m_flags & M_DECRYPTED) { 263 ICMP6STAT_INC(icp6s_canterror); 264 goto freeit; 265 } 266#endif 267 268#ifndef PULLDOWN_TEST 269 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), ); 270#else 271 if (m->m_len < sizeof(struct ip6_hdr)) { 272 m = m_pullup(m, sizeof(struct ip6_hdr)); 273 if (m == NULL) 274 return; 275 } 276#endif 277 oip6 = mtod(m, struct ip6_hdr *); 278 279 /* 280 * If the destination address of the erroneous packet is a multicast 281 * address, or the packet was sent using link-layer multicast, 282 * we should basically suppress sending an error (RFC 2463, Section 283 * 2.4). 284 * We have two exceptions (the item e.2 in that section): 285 * - the Pakcet Too Big message can be sent for path MTU discovery. 286 * - the Parameter Problem Message that can be allowed an icmp6 error 287 * in the option type field. This check has been done in 288 * ip6_unknown_opt(), so we can just check the type and code. 289 */ 290 if ((m->m_flags & (M_BCAST|M_MCAST) || 291 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 292 (type != ICMP6_PACKET_TOO_BIG && 293 (type != ICMP6_PARAM_PROB || 294 code != ICMP6_PARAMPROB_OPTION))) 295 goto freeit; 296 297 /* 298 * RFC 2463, 2.4 (e.5): source address check. 299 * XXX: the case of anycast source? 300 */ 301 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 302 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 303 goto freeit; 304 305 /* 306 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 307 * don't do it. 308 */ 309 nxt = -1; 310 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 311 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 312 struct icmp6_hdr *icp; 313 314#ifndef PULLDOWN_TEST 315 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), ); 316 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 317#else 318 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 319 sizeof(*icp)); 320 if (icp == NULL) { 321 ICMP6STAT_INC(icp6s_tooshort); 322 return; 323 } 324#endif 325 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 326 icp->icmp6_type == ND_REDIRECT) { 327 /* 328 * ICMPv6 error 329 * Special case: for redirect (which is 330 * informational) we must not send icmp6 error. 331 */ 332 ICMP6STAT_INC(icp6s_canterror); 333 goto freeit; 334 } else { 335 /* ICMPv6 informational - send the error */ 336 } 337 } else { 338 /* non-ICMPv6 - send the error */ 339 } 340 341 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 342 343 /* Finally, do rate limitation check. */ 344 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 345 ICMP6STAT_INC(icp6s_toofreq); 346 goto freeit; 347 } 348 349 /* 350 * OK, ICMP6 can be generated. 351 */ 352 353 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 354 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 355 356 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 357 M_PREPEND(m, preplen, M_DONTWAIT); 358 if (m && m->m_len < preplen) 359 m = m_pullup(m, preplen); 360 if (m == NULL) { 361 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 362 return; 363 } 364 365 nip6 = mtod(m, struct ip6_hdr *); 366 nip6->ip6_src = oip6->ip6_src; 367 nip6->ip6_dst = oip6->ip6_dst; 368 369 in6_clearscope(&oip6->ip6_src); 370 in6_clearscope(&oip6->ip6_dst); 371 372 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 373 icmp6->icmp6_type = type; 374 icmp6->icmp6_code = code; 375 icmp6->icmp6_pptr = htonl((u_int32_t)param); 376 377 /* 378 * icmp6_reflect() is designed to be in the input path. 379 * icmp6_error() can be called from both input and output path, 380 * and if we are in output path rcvif could contain bogus value. 381 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 382 * information in ip header (nip6). 383 */ 384 m->m_pkthdr.rcvif = NULL; 385 386 ICMP6STAT_INC(icp6s_outhist[type]); 387 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 388 389 return; 390 391 freeit: 392 /* 393 * If we can't tell whether or not we can generate ICMP6, free it. 394 */ 395 m_freem(m); 396} 397 398/* 399 * Process a received ICMP6 message. 400 */ 401int 402icmp6_input(struct mbuf **mp, int *offp, int proto) 403{ 404 INIT_VNET_INET6(curvnet); 405 struct mbuf *m = *mp, *n; 406 struct ifnet *ifp; 407 struct ip6_hdr *ip6, *nip6; 408 struct icmp6_hdr *icmp6, *nicmp6; 409 int off = *offp; 410 int icmp6len = m->m_pkthdr.len - *offp; 411 int code, sum, noff; 412 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 413 414 ifp = m->m_pkthdr.rcvif; 415 416#ifndef PULLDOWN_TEST 417 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE); 418 /* m might change if M_LOOP. So, call mtod after this */ 419#endif 420 421 /* 422 * Locate icmp6 structure in mbuf, and check 423 * that not corrupted and of at least minimum length 424 */ 425 426 ip6 = mtod(m, struct ip6_hdr *); 427 if (icmp6len < sizeof(struct icmp6_hdr)) { 428 ICMP6STAT_INC(icp6s_tooshort); 429 goto freeit; 430 } 431 432 /* 433 * Check multicast group membership. 434 * Note: SSM filters are not applied for ICMPv6 traffic. 435 */ 436 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 437 struct in6_multi *inm; 438 439 inm = in6m_lookup(ifp, &ip6->ip6_dst); 440 if (inm == NULL) { 441 IP6STAT_INC(ip6s_notmember); 442 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard); 443 goto freeit; 444 } 445 } 446 447 /* 448 * calculate the checksum 449 */ 450#ifndef PULLDOWN_TEST 451 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 452#else 453 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 454 if (icmp6 == NULL) { 455 ICMP6STAT_INC(icp6s_tooshort); 456 return IPPROTO_DONE; 457 } 458#endif 459 code = icmp6->icmp6_code; 460 461 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 462 nd6log((LOG_ERR, 463 "ICMP6 checksum error(%d|%x) %s\n", 464 icmp6->icmp6_type, sum, 465 ip6_sprintf(ip6bufs, &ip6->ip6_src))); 466 ICMP6STAT_INC(icp6s_checksum); 467 goto freeit; 468 } 469 470 if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) { 471 /* 472 * Deliver very specific ICMP6 type only. 473 * This is important to deliver TOOBIG. Otherwise PMTUD 474 * will not work. 475 */ 476 switch (icmp6->icmp6_type) { 477 case ICMP6_DST_UNREACH: 478 case ICMP6_PACKET_TOO_BIG: 479 case ICMP6_TIME_EXCEEDED: 480 break; 481 default: 482 goto freeit; 483 } 484 } 485 486 ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]); 487 icmp6_ifstat_inc(ifp, ifs6_in_msg); 488 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK) 489 icmp6_ifstat_inc(ifp, ifs6_in_error); 490 491 switch (icmp6->icmp6_type) { 492 case ICMP6_DST_UNREACH: 493 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach); 494 switch (code) { 495 case ICMP6_DST_UNREACH_NOROUTE: 496 code = PRC_UNREACH_NET; 497 break; 498 case ICMP6_DST_UNREACH_ADMIN: 499 icmp6_ifstat_inc(ifp, ifs6_in_adminprohib); 500 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 501 break; 502 case ICMP6_DST_UNREACH_ADDR: 503 code = PRC_HOSTDEAD; 504 break; 505 case ICMP6_DST_UNREACH_BEYONDSCOPE: 506 /* I mean "source address was incorrect." */ 507 code = PRC_PARAMPROB; 508 break; 509 case ICMP6_DST_UNREACH_NOPORT: 510 code = PRC_UNREACH_PORT; 511 break; 512 default: 513 goto badcode; 514 } 515 goto deliver; 516 break; 517 518 case ICMP6_PACKET_TOO_BIG: 519 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig); 520 521 /* validation is made in icmp6_mtudisc_update */ 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 break; 531 532 case ICMP6_TIME_EXCEEDED: 533 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed); 534 switch (code) { 535 case ICMP6_TIME_EXCEED_TRANSIT: 536 code = PRC_TIMXCEED_INTRANS; 537 break; 538 case ICMP6_TIME_EXCEED_REASSEMBLY: 539 code = PRC_TIMXCEED_REASS; 540 break; 541 default: 542 goto badcode; 543 } 544 goto deliver; 545 break; 546 547 case ICMP6_PARAM_PROB: 548 icmp6_ifstat_inc(ifp, ifs6_in_paramprob); 549 switch (code) { 550 case ICMP6_PARAMPROB_NEXTHEADER: 551 code = PRC_UNREACH_PROTOCOL; 552 break; 553 case ICMP6_PARAMPROB_HEADER: 554 case ICMP6_PARAMPROB_OPTION: 555 code = PRC_PARAMPROB; 556 break; 557 default: 558 goto badcode; 559 } 560 goto deliver; 561 break; 562 563 case ICMP6_ECHO_REQUEST: 564 icmp6_ifstat_inc(ifp, ifs6_in_echo); 565 if (code != 0) 566 goto badcode; 567 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) { 568 /* Give up remote */ 569 break; 570 } 571 if ((n->m_flags & M_EXT) != 0 572 || n->m_len < off + sizeof(struct icmp6_hdr)) { 573 struct mbuf *n0 = n; 574 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6); 575 int n0len; 576 577 MGETHDR(n, M_DONTWAIT, n0->m_type); 578 n0len = n0->m_pkthdr.len; /* save for use below */ 579 if (n) 580 M_MOVE_PKTHDR(n, n0); 581 if (n && maxlen >= MHLEN) { 582 MCLGET(n, M_DONTWAIT); 583 if ((n->m_flags & M_EXT) == 0) { 584 m_free(n); 585 n = NULL; 586 } 587 } 588 if (n == NULL) { 589 /* Give up remote */ 590 m_freem(n0); 591 break; 592 } 593 /* 594 * Copy IPv6 and ICMPv6 only. 595 */ 596 nip6 = mtod(n, struct ip6_hdr *); 597 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 598 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 599 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 600 noff = sizeof(struct ip6_hdr); 601 /* new mbuf contains only ipv6+icmpv6 headers */ 602 n->m_len = noff + sizeof(struct icmp6_hdr); 603 /* 604 * Adjust mbuf. ip6_plen will be adjusted in 605 * ip6_output(). 606 */ 607 m_adj(n0, off + sizeof(struct icmp6_hdr)); 608 /* recalculate complete packet size */ 609 n->m_pkthdr.len = n0len + (noff - off); 610 n->m_next = n0; 611 } else { 612 nip6 = mtod(n, struct ip6_hdr *); 613 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, 614 sizeof(*nicmp6)); 615 noff = off; 616 } 617 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 618 nicmp6->icmp6_code = 0; 619 if (n) { 620 ICMP6STAT_INC(icp6s_reflect); 621 ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]); 622 icmp6_reflect(n, noff); 623 } 624 break; 625 626 case ICMP6_ECHO_REPLY: 627 icmp6_ifstat_inc(ifp, ifs6_in_echoreply); 628 if (code != 0) 629 goto badcode; 630 break; 631 632 case MLD_LISTENER_QUERY: 633 case MLD_LISTENER_REPORT: 634 case MLD_LISTENER_DONE: 635 case MLDV2_LISTENER_REPORT: 636 /* 637 * Drop MLD traffic which is not link-local, has a hop limit 638 * of greater than 1 hop, or which does not have the 639 * IPv6 HBH Router Alert option. 640 * As IPv6 HBH options are stripped in ip6_input() we must 641 * check an mbuf header flag. 642 * XXX Should we also sanity check that these messages 643 * were directed to a link-local multicast prefix? 644 */ 645 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0) 646 goto freeit; 647 if (mld_input(m, off, icmp6len) != 0) 648 return (IPPROTO_DONE); 649 /* m stays. */ 650 break; 651 652 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 653 { 654 enum { WRU, FQDN } mode; 655 656 if (!V_icmp6_nodeinfo) 657 break; 658 659 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 660 mode = WRU; 661 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 662 mode = FQDN; 663 else 664 goto badlen; 665 666 if (mode == FQDN) { 667#ifndef PULLDOWN_TEST 668 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo), 669 IPPROTO_DONE); 670#endif 671 n = m_copy(m, 0, M_COPYALL); 672 if (n) 673 n = ni6_input(n, off); 674 /* XXX meaningless if n == NULL */ 675 noff = sizeof(struct ip6_hdr); 676 } else { 677 struct prison *pr; 678 u_char *p; 679 int maxlen, maxhlen, hlen; 680 681 /* 682 * XXX: this combination of flags is pointless, 683 * but should we keep this for compatibility? 684 */ 685 if ((V_icmp6_nodeinfo & 5) != 5) 686 break; 687 688 if (code != 0) 689 goto badcode; 690 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; 691 if (maxlen >= MCLBYTES) { 692 /* Give up remote */ 693 break; 694 } 695 MGETHDR(n, M_DONTWAIT, m->m_type); 696 if (n && maxlen > MHLEN) { 697 MCLGET(n, M_DONTWAIT); 698 if ((n->m_flags & M_EXT) == 0) { 699 m_free(n); 700 n = NULL; 701 } 702 } 703 if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) { 704 /* 705 * Previous code did a blind M_COPY_PKTHDR 706 * and said "just for rcvif". If true, then 707 * we could tolerate the dup failing (due to 708 * the deep copy of the tag chain). For now 709 * be conservative and just fail. 710 */ 711 m_free(n); 712 n = NULL; 713 } 714 if (n == NULL) { 715 /* Give up remote */ 716 break; 717 } 718 n->m_pkthdr.rcvif = NULL; 719 n->m_len = 0; 720 maxhlen = M_TRAILINGSPACE(n) - maxlen; 721 pr = curthread->td_ucred->cr_prison; 722 mtx_lock(&pr->pr_mtx); 723 hlen = strlen(pr->pr_host); 724 if (maxhlen > hlen) 725 maxhlen = hlen; 726 /* 727 * Copy IPv6 and ICMPv6 only. 728 */ 729 nip6 = mtod(n, struct ip6_hdr *); 730 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 731 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 732 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 733 p = (u_char *)(nicmp6 + 1); 734 bzero(p, 4); 735 bcopy(pr->pr_host, p + 4, maxhlen); /* meaningless TTL */ 736 mtx_unlock(&pr->pr_mtx); 737 noff = sizeof(struct ip6_hdr); 738 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 739 sizeof(struct icmp6_hdr) + 4 + maxhlen; 740 nicmp6->icmp6_type = ICMP6_WRUREPLY; 741 nicmp6->icmp6_code = 0; 742 } 743 if (n) { 744 ICMP6STAT_INC(icp6s_reflect); 745 ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]); 746 icmp6_reflect(n, noff); 747 } 748 break; 749 } 750 751 case ICMP6_WRUREPLY: 752 if (code != 0) 753 goto badcode; 754 break; 755 756 case ND_ROUTER_SOLICIT: 757 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit); 758 if (code != 0) 759 goto badcode; 760 if (icmp6len < sizeof(struct nd_router_solicit)) 761 goto badlen; 762 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 763 /* give up local */ 764 nd6_rs_input(m, off, icmp6len); 765 m = NULL; 766 goto freeit; 767 } 768 nd6_rs_input(n, off, icmp6len); 769 /* m stays. */ 770 break; 771 772 case ND_ROUTER_ADVERT: 773 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert); 774 if (code != 0) 775 goto badcode; 776 if (icmp6len < sizeof(struct nd_router_advert)) 777 goto badlen; 778 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 779 /* give up local */ 780 nd6_ra_input(m, off, icmp6len); 781 m = NULL; 782 goto freeit; 783 } 784 nd6_ra_input(n, off, icmp6len); 785 /* m stays. */ 786 break; 787 788 case ND_NEIGHBOR_SOLICIT: 789 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit); 790 if (code != 0) 791 goto badcode; 792 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 793 goto badlen; 794 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 795 /* give up local */ 796 nd6_ns_input(m, off, icmp6len); 797 m = NULL; 798 goto freeit; 799 } 800 nd6_ns_input(n, off, icmp6len); 801 /* m stays. */ 802 break; 803 804 case ND_NEIGHBOR_ADVERT: 805 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert); 806 if (code != 0) 807 goto badcode; 808 if (icmp6len < sizeof(struct nd_neighbor_advert)) 809 goto badlen; 810 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 811 /* give up local */ 812 nd6_na_input(m, off, icmp6len); 813 m = NULL; 814 goto freeit; 815 } 816 nd6_na_input(n, off, icmp6len); 817 /* m stays. */ 818 break; 819 820 case ND_REDIRECT: 821 icmp6_ifstat_inc(ifp, ifs6_in_redirect); 822 if (code != 0) 823 goto badcode; 824 if (icmp6len < sizeof(struct nd_redirect)) 825 goto badlen; 826 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 827 /* give up local */ 828 icmp6_redirect_input(m, off); 829 m = NULL; 830 goto freeit; 831 } 832 icmp6_redirect_input(n, off); 833 /* m stays. */ 834 break; 835 836 case ICMP6_ROUTER_RENUMBERING: 837 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 838 code != ICMP6_ROUTER_RENUMBERING_RESULT) 839 goto badcode; 840 if (icmp6len < sizeof(struct icmp6_router_renum)) 841 goto badlen; 842 break; 843 844 default: 845 nd6log((LOG_DEBUG, 846 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 847 icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src), 848 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 849 ifp ? ifp->if_index : 0)); 850 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 851 /* ICMPv6 error: MUST deliver it by spec... */ 852 code = PRC_NCMDS; 853 /* deliver */ 854 } else { 855 /* ICMPv6 informational: MUST not deliver */ 856 break; 857 } 858 deliver: 859 if (icmp6_notify_error(&m, off, icmp6len, code)) { 860 /* In this case, m should've been freed. */ 861 return (IPPROTO_DONE); 862 } 863 break; 864 865 badcode: 866 ICMP6STAT_INC(icp6s_badcode); 867 break; 868 869 badlen: 870 ICMP6STAT_INC(icp6s_badlen); 871 break; 872 } 873 874 /* deliver the packet to appropriate sockets */ 875 icmp6_rip6_input(&m, *offp); 876 877 return IPPROTO_DONE; 878 879 freeit: 880 m_freem(m); 881 return IPPROTO_DONE; 882} 883 884static int 885icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code) 886{ 887 INIT_VNET_INET6(curvnet); 888 struct mbuf *m = *mp; 889 struct icmp6_hdr *icmp6; 890 struct ip6_hdr *eip6; 891 u_int32_t notifymtu; 892 struct sockaddr_in6 icmp6src, icmp6dst; 893 894 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 895 ICMP6STAT_INC(icp6s_tooshort); 896 goto freeit; 897 } 898#ifndef PULLDOWN_TEST 899 IP6_EXTHDR_CHECK(m, off, 900 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1); 901 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 902#else 903 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 904 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 905 if (icmp6 == NULL) { 906 ICMP6STAT_INC(icp6s_tooshort); 907 return (-1); 908 } 909#endif 910 eip6 = (struct ip6_hdr *)(icmp6 + 1); 911 912 /* Detect the upper level protocol */ 913 { 914 void (*ctlfunc)(int, struct sockaddr *, void *); 915 u_int8_t nxt = eip6->ip6_nxt; 916 int eoff = off + sizeof(struct icmp6_hdr) + 917 sizeof(struct ip6_hdr); 918 struct ip6ctlparam ip6cp; 919 struct in6_addr *finaldst = NULL; 920 int icmp6type = icmp6->icmp6_type; 921 struct ip6_frag *fh; 922 struct ip6_rthdr *rth; 923 struct ip6_rthdr0 *rth0; 924 int rthlen; 925 926 while (1) { /* XXX: should avoid infinite loop explicitly? */ 927 struct ip6_ext *eh; 928 929 switch (nxt) { 930 case IPPROTO_HOPOPTS: 931 case IPPROTO_DSTOPTS: 932 case IPPROTO_AH: 933#ifndef PULLDOWN_TEST 934 IP6_EXTHDR_CHECK(m, 0, 935 eoff + sizeof(struct ip6_ext), -1); 936 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff); 937#else 938 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 939 eoff, sizeof(*eh)); 940 if (eh == NULL) { 941 ICMP6STAT_INC(icp6s_tooshort); 942 return (-1); 943 } 944#endif 945 946 if (nxt == IPPROTO_AH) 947 eoff += (eh->ip6e_len + 2) << 2; 948 else 949 eoff += (eh->ip6e_len + 1) << 3; 950 nxt = eh->ip6e_nxt; 951 break; 952 case IPPROTO_ROUTING: 953 /* 954 * When the erroneous packet contains a 955 * routing header, we should examine the 956 * header to determine the final destination. 957 * Otherwise, we can't properly update 958 * information that depends on the final 959 * destination (e.g. path MTU). 960 */ 961#ifndef PULLDOWN_TEST 962 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1); 963 rth = (struct ip6_rthdr *) 964 (mtod(m, caddr_t) + eoff); 965#else 966 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 967 eoff, sizeof(*rth)); 968 if (rth == NULL) { 969 ICMP6STAT_INC(icp6s_tooshort); 970 return (-1); 971 } 972#endif 973 rthlen = (rth->ip6r_len + 1) << 3; 974 /* 975 * XXX: currently there is no 976 * officially defined type other 977 * than type-0. 978 * Note that if the segment left field 979 * is 0, all intermediate hops must 980 * have been passed. 981 */ 982 if (rth->ip6r_segleft && 983 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 984 int hops; 985 986#ifndef PULLDOWN_TEST 987 IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1); 988 rth0 = (struct ip6_rthdr0 *) 989 (mtod(m, caddr_t) + eoff); 990#else 991 IP6_EXTHDR_GET(rth0, 992 struct ip6_rthdr0 *, m, 993 eoff, rthlen); 994 if (rth0 == NULL) { 995 ICMP6STAT_INC(icp6s_tooshort); 996 return (-1); 997 } 998#endif 999 /* just ignore a bogus header */ 1000 if ((rth0->ip6r0_len % 2) == 0 && 1001 (hops = rth0->ip6r0_len/2)) 1002 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 1003 } 1004 eoff += rthlen; 1005 nxt = rth->ip6r_nxt; 1006 break; 1007 case IPPROTO_FRAGMENT: 1008#ifndef PULLDOWN_TEST 1009 IP6_EXTHDR_CHECK(m, 0, eoff + 1010 sizeof(struct ip6_frag), -1); 1011 fh = (struct ip6_frag *)(mtod(m, caddr_t) + 1012 eoff); 1013#else 1014 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 1015 eoff, sizeof(*fh)); 1016 if (fh == NULL) { 1017 ICMP6STAT_INC(icp6s_tooshort); 1018 return (-1); 1019 } 1020#endif 1021 /* 1022 * Data after a fragment header is meaningless 1023 * unless it is the first fragment, but 1024 * we'll go to the notify label for path MTU 1025 * discovery. 1026 */ 1027 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1028 goto notify; 1029 1030 eoff += sizeof(struct ip6_frag); 1031 nxt = fh->ip6f_nxt; 1032 break; 1033 default: 1034 /* 1035 * This case includes ESP and the No Next 1036 * Header. In such cases going to the notify 1037 * label does not have any meaning 1038 * (i.e. ctlfunc will be NULL), but we go 1039 * anyway since we might have to update 1040 * path MTU information. 1041 */ 1042 goto notify; 1043 } 1044 } 1045 notify: 1046#ifndef PULLDOWN_TEST 1047 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 1048#else 1049 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1050 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1051 if (icmp6 == NULL) { 1052 ICMP6STAT_INC(icp6s_tooshort); 1053 return (-1); 1054 } 1055#endif 1056 1057 /* 1058 * retrieve parameters from the inner IPv6 header, and convert 1059 * them into sockaddr structures. 1060 * XXX: there is no guarantee that the source or destination 1061 * addresses of the inner packet are in the same scope as 1062 * the addresses of the icmp packet. But there is no other 1063 * way to determine the zone. 1064 */ 1065 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1066 1067 bzero(&icmp6dst, sizeof(icmp6dst)); 1068 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1069 icmp6dst.sin6_family = AF_INET6; 1070 if (finaldst == NULL) 1071 icmp6dst.sin6_addr = eip6->ip6_dst; 1072 else 1073 icmp6dst.sin6_addr = *finaldst; 1074 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1075 goto freeit; 1076 bzero(&icmp6src, sizeof(icmp6src)); 1077 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1078 icmp6src.sin6_family = AF_INET6; 1079 icmp6src.sin6_addr = eip6->ip6_src; 1080 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1081 goto freeit; 1082 icmp6src.sin6_flowinfo = 1083 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1084 1085 if (finaldst == NULL) 1086 finaldst = &eip6->ip6_dst; 1087 ip6cp.ip6c_m = m; 1088 ip6cp.ip6c_icmp6 = icmp6; 1089 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1090 ip6cp.ip6c_off = eoff; 1091 ip6cp.ip6c_finaldst = finaldst; 1092 ip6cp.ip6c_src = &icmp6src; 1093 ip6cp.ip6c_nxt = nxt; 1094 1095 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1096 notifymtu = ntohl(icmp6->icmp6_mtu); 1097 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1098 icmp6_mtudisc_update(&ip6cp, 1); /*XXX*/ 1099 } 1100 1101 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1102 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1103 if (ctlfunc) { 1104 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1105 &ip6cp); 1106 } 1107 } 1108 *mp = m; 1109 return (0); 1110 1111 freeit: 1112 m_freem(m); 1113 return (-1); 1114} 1115 1116void 1117icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1118{ 1119 INIT_VNET_INET6(curvnet); 1120 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1121 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1122 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1123 u_int mtu = ntohl(icmp6->icmp6_mtu); 1124 struct in_conninfo inc; 1125 1126#if 0 1127 /* 1128 * RFC2460 section 5, last paragraph. 1129 * even though minimum link MTU for IPv6 is IPV6_MMTU, 1130 * we may see ICMPv6 too big with mtu < IPV6_MMTU 1131 * due to packet translator in the middle. 1132 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for 1133 * special handling. 1134 */ 1135 if (mtu < IPV6_MMTU) 1136 return; 1137#endif 1138 1139 /* 1140 * we reject ICMPv6 too big with abnormally small value. 1141 * XXX what is the good definition of "abnormally small"? 1142 */ 1143 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) 1144 return; 1145 1146 if (!validated) 1147 return; 1148 1149 /* 1150 * In case the suggested mtu is less than IPV6_MMTU, we 1151 * only need to remember that it was for above mentioned 1152 * "alwaysfrag" case. 1153 * Try to be as close to the spec as possible. 1154 */ 1155 if (mtu < IPV6_MMTU) 1156 mtu = IPV6_MMTU - 8; 1157 1158 bzero(&inc, sizeof(inc)); 1159 inc.inc_flags |= INC_ISIPV6; 1160 inc.inc6_faddr = *dst; 1161 if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL)) 1162 return; 1163 1164 if (mtu < tcp_maxmtu6(&inc, NULL)) { 1165 tcp_hc_updatemtu(&inc, mtu); 1166 ICMP6STAT_INC(icp6s_pmtuchg); 1167 } 1168} 1169 1170/* 1171 * Process a Node Information Query packet, based on 1172 * draft-ietf-ipngwg-icmp-name-lookups-07. 1173 * 1174 * Spec incompatibilities: 1175 * - IPv6 Subject address handling 1176 * - IPv4 Subject address handling support missing 1177 * - Proxy reply (answer even if it's not for me) 1178 * - joins NI group address at in6_ifattach() time only, does not cope 1179 * with hostname changes by sethostname(3) 1180 */ 1181static struct mbuf * 1182ni6_input(struct mbuf *m, int off) 1183{ 1184 INIT_VNET_INET6(curvnet); 1185 struct icmp6_nodeinfo *ni6, *nni6; 1186 struct mbuf *n = NULL; 1187 struct prison *pr; 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 in6_addr in6_subj; /* subject address */ 1195 struct ip6_hdr *ip6; 1196 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1197 char *subj = NULL; 1198 struct in6_ifaddr *ia6 = NULL; 1199 1200 ip6 = mtod(m, struct ip6_hdr *); 1201#ifndef PULLDOWN_TEST 1202 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off); 1203#else 1204 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1205 if (ni6 == NULL) { 1206 /* m is already reclaimed */ 1207 return (NULL); 1208 } 1209#endif 1210 1211 /* 1212 * Validate IPv6 source address. 1213 * The default configuration MUST be to refuse answering queries from 1214 * global-scope addresses according to RFC4602. 1215 * Notes: 1216 * - it's not very clear what "refuse" means; this implementation 1217 * simply drops it. 1218 * - it's not very easy to identify global-scope (unicast) addresses 1219 * since there are many prefixes for them. It should be safer 1220 * and in practice sufficient to check "all" but loopback and 1221 * link-local (note that site-local unicast was deprecated and 1222 * ULA is defined as global scope-wise) 1223 */ 1224 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 && 1225 !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) && 1226 !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) 1227 goto bad; 1228 1229 /* 1230 * Validate IPv6 destination address. 1231 * 1232 * The Responder must discard the Query without further processing 1233 * unless it is one of the Responder's unicast or anycast addresses, or 1234 * a link-local scope multicast address which the Responder has joined. 1235 * [RFC4602, Section 5.] 1236 */ 1237 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 1238 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst)) 1239 goto bad; 1240 /* else it's a link-local multicast, fine */ 1241 } else { /* unicast or anycast */ 1242 if ((ia6 = ip6_getdstifaddr(m)) == NULL) 1243 goto bad; /* XXX impossible */ 1244 1245 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) && 1246 !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) { 1247 nd6log((LOG_DEBUG, "ni6_input: ignore node info to " 1248 "a temporary address in %s:%d", 1249 __FILE__, __LINE__)); 1250 goto bad; 1251 } 1252 } 1253 1254 /* validate query Subject field. */ 1255 qtype = ntohs(ni6->ni_qtype); 1256 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1257 switch (qtype) { 1258 case NI_QTYPE_NOOP: 1259 case NI_QTYPE_SUPTYPES: 1260 /* 07 draft */ 1261 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1262 break; 1263 /* FALLTHROUGH */ 1264 case NI_QTYPE_FQDN: 1265 case NI_QTYPE_NODEADDR: 1266 case NI_QTYPE_IPV4ADDR: 1267 switch (ni6->ni_code) { 1268 case ICMP6_NI_SUBJ_IPV6: 1269#if ICMP6_NI_SUBJ_IPV6 != 0 1270 case 0: 1271#endif 1272 /* 1273 * backward compatibility - try to accept 03 draft 1274 * format, where no Subject is present. 1275 */ 1276 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1277 subjlen == 0) { 1278 oldfqdn++; 1279 break; 1280 } 1281#if ICMP6_NI_SUBJ_IPV6 != 0 1282 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1283 goto bad; 1284#endif 1285 1286 if (subjlen != sizeof(struct in6_addr)) 1287 goto bad; 1288 1289 /* 1290 * Validate Subject address. 1291 * 1292 * Not sure what exactly "address belongs to the node" 1293 * means in the spec, is it just unicast, or what? 1294 * 1295 * At this moment we consider Subject address as 1296 * "belong to the node" if the Subject address equals 1297 * to the IPv6 destination address; validation for 1298 * IPv6 destination address should have done enough 1299 * check for us. 1300 * 1301 * We do not do proxy at this moment. 1302 */ 1303 /* m_pulldown instead of copy? */ 1304 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1305 subjlen, (caddr_t)&in6_subj); 1306 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1307 goto bad; 1308 1309 subj = (char *)&in6_subj; 1310 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1311 break; 1312 1313 /* 1314 * XXX if we are to allow other cases, we should really 1315 * be careful about scope here. 1316 * basically, we should disallow queries toward IPv6 1317 * destination X with subject Y, 1318 * if scope(X) > scope(Y). 1319 * if we allow scope(X) > scope(Y), it will result in 1320 * information leakage across scope boundary. 1321 */ 1322 goto bad; 1323 1324 case ICMP6_NI_SUBJ_FQDN: 1325 /* 1326 * Validate Subject name with gethostname(3). 1327 * 1328 * The behavior may need some debate, since: 1329 * - we are not sure if the node has FQDN as 1330 * hostname (returned by gethostname(3)). 1331 * - the code does wildcard match for truncated names. 1332 * however, we are not sure if we want to perform 1333 * wildcard match, if gethostname(3) side has 1334 * truncated hostname. 1335 */ 1336 pr = curthread->td_ucred->cr_prison; 1337 mtx_lock(&pr->pr_mtx); 1338 n = ni6_nametodns(pr->pr_host, strlen(pr->pr_host), 0); 1339 mtx_unlock(&pr->pr_mtx); 1340 if (!n || n->m_next || n->m_len == 0) 1341 goto bad; 1342 IP6_EXTHDR_GET(subj, char *, m, 1343 off + sizeof(struct icmp6_nodeinfo), subjlen); 1344 if (subj == NULL) 1345 goto bad; 1346 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1347 n->m_len)) { 1348 goto bad; 1349 } 1350 m_freem(n); 1351 n = NULL; 1352 break; 1353 1354 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1355 default: 1356 goto bad; 1357 } 1358 break; 1359 } 1360 1361 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1362 switch (qtype) { 1363 case NI_QTYPE_FQDN: 1364 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0) 1365 goto bad; 1366 break; 1367 case NI_QTYPE_NODEADDR: 1368 case NI_QTYPE_IPV4ADDR: 1369 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0) 1370 goto bad; 1371 break; 1372 } 1373 1374 /* guess reply length */ 1375 switch (qtype) { 1376 case NI_QTYPE_NOOP: 1377 break; /* no reply data */ 1378 case NI_QTYPE_SUPTYPES: 1379 replylen += sizeof(u_int32_t); 1380 break; 1381 case NI_QTYPE_FQDN: 1382 /* XXX will append an mbuf */ 1383 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1384 break; 1385 case NI_QTYPE_NODEADDR: 1386 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj); 1387 if ((replylen += addrs * (sizeof(struct in6_addr) + 1388 sizeof(u_int32_t))) > MCLBYTES) 1389 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1390 break; 1391 case NI_QTYPE_IPV4ADDR: 1392 /* unsupported - should respond with unknown Qtype? */ 1393 break; 1394 default: 1395 /* 1396 * XXX: We must return a reply with the ICMP6 code 1397 * `unknown Qtype' in this case. However we regard the case 1398 * as an FQDN query for backward compatibility. 1399 * Older versions set a random value to this field, 1400 * so it rarely varies in the defined qtypes. 1401 * But the mechanism is not reliable... 1402 * maybe we should obsolete older versions. 1403 */ 1404 qtype = NI_QTYPE_FQDN; 1405 /* XXX will append an mbuf */ 1406 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1407 oldfqdn++; 1408 break; 1409 } 1410 1411 /* allocate an mbuf to reply. */ 1412 MGETHDR(n, M_DONTWAIT, m->m_type); 1413 if (n == NULL) { 1414 m_freem(m); 1415 return (NULL); 1416 } 1417 M_MOVE_PKTHDR(n, m); /* just for recvif */ 1418 if (replylen > MHLEN) { 1419 if (replylen > MCLBYTES) { 1420 /* 1421 * XXX: should we try to allocate more? But MCLBYTES 1422 * is probably much larger than IPV6_MMTU... 1423 */ 1424 goto bad; 1425 } 1426 MCLGET(n, M_DONTWAIT); 1427 if ((n->m_flags & M_EXT) == 0) { 1428 goto bad; 1429 } 1430 } 1431 n->m_pkthdr.len = n->m_len = replylen; 1432 1433 /* copy mbuf header and IPv6 + Node Information base headers */ 1434 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1435 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1436 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1437 1438 /* qtype dependent procedure */ 1439 switch (qtype) { 1440 case NI_QTYPE_NOOP: 1441 nni6->ni_code = ICMP6_NI_SUCCESS; 1442 nni6->ni_flags = 0; 1443 break; 1444 case NI_QTYPE_SUPTYPES: 1445 { 1446 u_int32_t v; 1447 nni6->ni_code = ICMP6_NI_SUCCESS; 1448 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1449 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1450 v = (u_int32_t)htonl(0x0000000f); 1451 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1452 break; 1453 } 1454 case NI_QTYPE_FQDN: 1455 nni6->ni_code = ICMP6_NI_SUCCESS; 1456 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1457 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo)); 1458 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1459 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1460 /* 1461 * XXX do we really have FQDN in hostname? 1462 */ 1463 pr = curthread->td_ucred->cr_prison; 1464 mtx_lock(&pr->pr_mtx); 1465 n->m_next = 1466 ni6_nametodns(pr->pr_host, strlen(pr->pr_host), oldfqdn); 1467 mtx_unlock(&pr->pr_mtx); 1468 if (n->m_next == NULL) 1469 goto bad; 1470 /* XXX we assume that n->m_next is not a chain */ 1471 if (n->m_next->m_next != NULL) 1472 goto bad; 1473 n->m_pkthdr.len += n->m_next->m_len; 1474 break; 1475 case NI_QTYPE_NODEADDR: 1476 { 1477 int lenlim, copied; 1478 1479 nni6->ni_code = ICMP6_NI_SUCCESS; 1480 n->m_pkthdr.len = n->m_len = 1481 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1482 lenlim = M_TRAILINGSPACE(n); 1483 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1484 /* XXX: reset mbuf length */ 1485 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1486 sizeof(struct icmp6_nodeinfo) + copied; 1487 break; 1488 } 1489 default: 1490 break; /* XXX impossible! */ 1491 } 1492 1493 nni6->ni_type = ICMP6_NI_REPLY; 1494 m_freem(m); 1495 return (n); 1496 1497 bad: 1498 m_freem(m); 1499 if (n) 1500 m_freem(n); 1501 return (NULL); 1502} 1503 1504/* 1505 * make a mbuf with DNS-encoded string. no compression support. 1506 * 1507 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1508 * treated as truncated name (two \0 at the end). this is a wild guess. 1509 * 1510 * old - return pascal string if non-zero 1511 */ 1512static struct mbuf * 1513ni6_nametodns(const char *name, int namelen, int old) 1514{ 1515 struct mbuf *m; 1516 char *cp, *ep; 1517 const char *p, *q; 1518 int i, len, nterm; 1519 1520 if (old) 1521 len = namelen + 1; 1522 else 1523 len = MCLBYTES; 1524 1525 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1526 MGET(m, M_DONTWAIT, MT_DATA); 1527 if (m && len > MLEN) { 1528 MCLGET(m, M_DONTWAIT); 1529 if ((m->m_flags & M_EXT) == 0) 1530 goto fail; 1531 } 1532 if (!m) 1533 goto fail; 1534 m->m_next = NULL; 1535 1536 if (old) { 1537 m->m_len = len; 1538 *mtod(m, char *) = namelen; 1539 bcopy(name, mtod(m, char *) + 1, namelen); 1540 return m; 1541 } else { 1542 m->m_len = 0; 1543 cp = mtod(m, char *); 1544 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1545 1546 /* if not certain about my name, return empty buffer */ 1547 if (namelen == 0) 1548 return m; 1549 1550 /* 1551 * guess if it looks like shortened hostname, or FQDN. 1552 * shortened hostname needs two trailing "\0". 1553 */ 1554 i = 0; 1555 for (p = name; p < name + namelen; p++) { 1556 if (*p && *p == '.') 1557 i++; 1558 } 1559 if (i < 2) 1560 nterm = 2; 1561 else 1562 nterm = 1; 1563 1564 p = name; 1565 while (cp < ep && p < name + namelen) { 1566 i = 0; 1567 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1568 i++; 1569 /* result does not fit into mbuf */ 1570 if (cp + i + 1 >= ep) 1571 goto fail; 1572 /* 1573 * DNS label length restriction, RFC1035 page 8. 1574 * "i == 0" case is included here to avoid returning 1575 * 0-length label on "foo..bar". 1576 */ 1577 if (i <= 0 || i >= 64) 1578 goto fail; 1579 *cp++ = i; 1580 bcopy(p, cp, i); 1581 cp += i; 1582 p = q; 1583 if (p < name + namelen && *p == '.') 1584 p++; 1585 } 1586 /* termination */ 1587 if (cp + nterm >= ep) 1588 goto fail; 1589 while (nterm-- > 0) 1590 *cp++ = '\0'; 1591 m->m_len = cp - mtod(m, char *); 1592 return m; 1593 } 1594 1595 panic("should not reach here"); 1596 /* NOTREACHED */ 1597 1598 fail: 1599 if (m) 1600 m_freem(m); 1601 return NULL; 1602} 1603 1604/* 1605 * check if two DNS-encoded string matches. takes care of truncated 1606 * form (with \0\0 at the end). no compression support. 1607 * XXX upper/lowercase match (see RFC2065) 1608 */ 1609static int 1610ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1611{ 1612 const char *a0, *b0; 1613 int l; 1614 1615 /* simplest case - need validation? */ 1616 if (alen == blen && bcmp(a, b, alen) == 0) 1617 return 1; 1618 1619 a0 = a; 1620 b0 = b; 1621 1622 /* termination is mandatory */ 1623 if (alen < 2 || blen < 2) 1624 return 0; 1625 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1626 return 0; 1627 alen--; 1628 blen--; 1629 1630 while (a - a0 < alen && b - b0 < blen) { 1631 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1632 return 0; 1633 1634 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1635 return 0; 1636 /* we don't support compression yet */ 1637 if (a[0] >= 64 || b[0] >= 64) 1638 return 0; 1639 1640 /* truncated case */ 1641 if (a[0] == 0 && a - a0 == alen - 1) 1642 return 1; 1643 if (b[0] == 0 && b - b0 == blen - 1) 1644 return 1; 1645 if (a[0] == 0 || b[0] == 0) 1646 return 0; 1647 1648 if (a[0] != b[0]) 1649 return 0; 1650 l = a[0]; 1651 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1652 return 0; 1653 if (bcmp(a + 1, b + 1, l) != 0) 1654 return 0; 1655 1656 a += 1 + l; 1657 b += 1 + l; 1658 } 1659 1660 if (a - a0 == alen && b - b0 == blen) 1661 return 1; 1662 else 1663 return 0; 1664} 1665 1666/* 1667 * calculate the number of addresses to be returned in the node info reply. 1668 */ 1669static int 1670ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp, 1671 struct in6_addr *subj) 1672{ 1673 INIT_VNET_NET(curvnet); 1674 INIT_VNET_INET6(curvnet); 1675 struct ifnet *ifp; 1676 struct in6_ifaddr *ifa6; 1677 struct ifaddr *ifa; 1678 int addrs = 0, addrsofif, iffound = 0; 1679 int niflags = ni6->ni_flags; 1680 1681 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1682 switch (ni6->ni_code) { 1683 case ICMP6_NI_SUBJ_IPV6: 1684 if (subj == NULL) /* must be impossible... */ 1685 return (0); 1686 break; 1687 default: 1688 /* 1689 * XXX: we only support IPv6 subject address for 1690 * this Qtype. 1691 */ 1692 return (0); 1693 } 1694 } 1695 1696 IFNET_RLOCK(); 1697 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) { 1698 addrsofif = 0; 1699 IF_ADDR_LOCK(ifp); 1700 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1701 if (ifa->ifa_addr->sa_family != AF_INET6) 1702 continue; 1703 ifa6 = (struct in6_ifaddr *)ifa; 1704 1705 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1706 IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr)) 1707 iffound = 1; 1708 1709 /* 1710 * IPv4-mapped addresses can only be returned by a 1711 * Node Information proxy, since they represent 1712 * addresses of IPv4-only nodes, which perforce do 1713 * not implement this protocol. 1714 * [icmp-name-lookups-07, Section 5.4] 1715 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1716 * this function at this moment. 1717 */ 1718 1719 /* What do we have to do about ::1? */ 1720 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1721 case IPV6_ADDR_SCOPE_LINKLOCAL: 1722 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1723 continue; 1724 break; 1725 case IPV6_ADDR_SCOPE_SITELOCAL: 1726 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1727 continue; 1728 break; 1729 case IPV6_ADDR_SCOPE_GLOBAL: 1730 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1731 continue; 1732 break; 1733 default: 1734 continue; 1735 } 1736 1737 /* 1738 * check if anycast is okay. 1739 * XXX: just experimental. not in the spec. 1740 */ 1741 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1742 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1743 continue; /* we need only unicast addresses */ 1744 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1745 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) { 1746 continue; 1747 } 1748 addrsofif++; /* count the address */ 1749 } 1750 IF_ADDR_UNLOCK(ifp); 1751 if (iffound) { 1752 *ifpp = ifp; 1753 IFNET_RUNLOCK(); 1754 return (addrsofif); 1755 } 1756 1757 addrs += addrsofif; 1758 } 1759 IFNET_RUNLOCK(); 1760 1761 return (addrs); 1762} 1763 1764static int 1765ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6, 1766 struct ifnet *ifp0, int resid) 1767{ 1768 INIT_VNET_NET(curvnet); 1769 INIT_VNET_INET6(curvnet); 1770 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet); 1771 struct in6_ifaddr *ifa6; 1772 struct ifaddr *ifa; 1773 struct ifnet *ifp_dep = NULL; 1774 int copied = 0, allow_deprecated = 0; 1775 u_char *cp = (u_char *)(nni6 + 1); 1776 int niflags = ni6->ni_flags; 1777 u_int32_t ltime; 1778 1779 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1780 return (0); /* needless to copy */ 1781 1782 IFNET_RLOCK(); 1783 again: 1784 1785 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) { 1786 IF_ADDR_LOCK(ifp); 1787 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1788 if (ifa->ifa_addr->sa_family != AF_INET6) 1789 continue; 1790 ifa6 = (struct in6_ifaddr *)ifa; 1791 1792 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1793 allow_deprecated == 0) { 1794 /* 1795 * prefererred address should be put before 1796 * deprecated addresses. 1797 */ 1798 1799 /* record the interface for later search */ 1800 if (ifp_dep == NULL) 1801 ifp_dep = ifp; 1802 1803 continue; 1804 } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1805 allow_deprecated != 0) 1806 continue; /* we now collect deprecated addrs */ 1807 1808 /* What do we have to do about ::1? */ 1809 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1810 case IPV6_ADDR_SCOPE_LINKLOCAL: 1811 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1812 continue; 1813 break; 1814 case IPV6_ADDR_SCOPE_SITELOCAL: 1815 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1816 continue; 1817 break; 1818 case IPV6_ADDR_SCOPE_GLOBAL: 1819 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1820 continue; 1821 break; 1822 default: 1823 continue; 1824 } 1825 1826 /* 1827 * check if anycast is okay. 1828 * XXX: just experimental. not in the spec. 1829 */ 1830 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1831 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1832 continue; 1833 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1834 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) { 1835 continue; 1836 } 1837 1838 /* now we can copy the address */ 1839 if (resid < sizeof(struct in6_addr) + 1840 sizeof(u_int32_t)) { 1841 IF_ADDR_UNLOCK(ifp); 1842 /* 1843 * We give up much more copy. 1844 * Set the truncate flag and return. 1845 */ 1846 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1847 IFNET_RUNLOCK(); 1848 return (copied); 1849 } 1850 1851 /* 1852 * Set the TTL of the address. 1853 * The TTL value should be one of the following 1854 * according to the specification: 1855 * 1856 * 1. The remaining lifetime of a DHCP lease on the 1857 * address, or 1858 * 2. The remaining Valid Lifetime of a prefix from 1859 * which the address was derived through Stateless 1860 * Autoconfiguration. 1861 * 1862 * Note that we currently do not support stateful 1863 * address configuration by DHCPv6, so the former 1864 * case can't happen. 1865 */ 1866 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1867 ltime = ND6_INFINITE_LIFETIME; 1868 else { 1869 if (ifa6->ia6_lifetime.ia6t_expire > 1870 time_second) 1871 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second); 1872 else 1873 ltime = 0; 1874 } 1875 1876 bcopy(<ime, cp, sizeof(u_int32_t)); 1877 cp += sizeof(u_int32_t); 1878 1879 /* copy the address itself */ 1880 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1881 sizeof(struct in6_addr)); 1882 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1883 cp += sizeof(struct in6_addr); 1884 1885 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1886 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1887 } 1888 IF_ADDR_UNLOCK(ifp); 1889 if (ifp0) /* we need search only on the specified IF */ 1890 break; 1891 } 1892 1893 if (allow_deprecated == 0 && ifp_dep != NULL) { 1894 ifp = ifp_dep; 1895 allow_deprecated = 1; 1896 1897 goto again; 1898 } 1899 1900 IFNET_RUNLOCK(); 1901 1902 return (copied); 1903} 1904 1905/* 1906 * XXX almost dup'ed code with rip6_input. 1907 */ 1908static int 1909icmp6_rip6_input(struct mbuf **mp, int off) 1910{ 1911 INIT_VNET_INET(curvnet); 1912 INIT_VNET_INET6(curvnet); 1913 struct mbuf *m = *mp; 1914 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1915 struct inpcb *in6p; 1916 struct inpcb *last = NULL; 1917 struct sockaddr_in6 fromsa; 1918 struct icmp6_hdr *icmp6; 1919 struct mbuf *opts = NULL; 1920 1921#ifndef PULLDOWN_TEST 1922 /* this is assumed to be safe. */ 1923 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 1924#else 1925 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1926 if (icmp6 == NULL) { 1927 /* m is already reclaimed */ 1928 return (IPPROTO_DONE); 1929 } 1930#endif 1931 1932 /* 1933 * XXX: the address may have embedded scope zone ID, which should be 1934 * hidden from applications. 1935 */ 1936 bzero(&fromsa, sizeof(fromsa)); 1937 fromsa.sin6_family = AF_INET6; 1938 fromsa.sin6_len = sizeof(struct sockaddr_in6); 1939 fromsa.sin6_addr = ip6->ip6_src; 1940 if (sa6_recoverscope(&fromsa)) { 1941 m_freem(m); 1942 return (IPPROTO_DONE); 1943 } 1944 1945 INP_INFO_RLOCK(&V_ripcbinfo); 1946 LIST_FOREACH(in6p, &V_ripcb, inp_list) { 1947 if ((in6p->inp_vflag & INP_IPV6) == 0) 1948 continue; 1949 if (in6p->inp_ip_p != IPPROTO_ICMPV6) 1950 continue; 1951 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1952 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1953 continue; 1954 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1955 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1956 continue; 1957 INP_RLOCK(in6p); 1958 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1959 in6p->in6p_icmp6filt)) { 1960 INP_RUNLOCK(in6p); 1961 continue; 1962 } 1963 if (last != NULL) { 1964 struct mbuf *n = NULL; 1965 1966 /* 1967 * Recent network drivers tend to allocate a single 1968 * mbuf cluster, rather than to make a couple of 1969 * mbufs without clusters. Also, since the IPv6 code 1970 * path tries to avoid m_pullup(), it is highly 1971 * probable that we still have an mbuf cluster here 1972 * even though the necessary length can be stored in an 1973 * mbuf's internal buffer. 1974 * Meanwhile, the default size of the receive socket 1975 * buffer for raw sockets is not so large. This means 1976 * the possibility of packet loss is relatively higher 1977 * than before. To avoid this scenario, we copy the 1978 * received data to a separate mbuf that does not use 1979 * a cluster, if possible. 1980 * XXX: it is better to copy the data after stripping 1981 * intermediate headers. 1982 */ 1983 if ((m->m_flags & M_EXT) && m->m_next == NULL && 1984 m->m_len <= MHLEN) { 1985 MGET(n, M_DONTWAIT, m->m_type); 1986 if (n != NULL) { 1987 if (m_dup_pkthdr(n, m, M_NOWAIT)) { 1988 bcopy(m->m_data, n->m_data, 1989 m->m_len); 1990 n->m_len = m->m_len; 1991 } else { 1992 m_free(n); 1993 n = NULL; 1994 } 1995 } 1996 } 1997 if (n != NULL || 1998 (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1999 if (last->inp_flags & INP_CONTROLOPTS) 2000 ip6_savecontrol(last, n, &opts); 2001 /* strip intermediate headers */ 2002 m_adj(n, off); 2003 SOCKBUF_LOCK(&last->inp_socket->so_rcv); 2004 if (sbappendaddr_locked( 2005 &last->inp_socket->so_rcv, 2006 (struct sockaddr *)&fromsa, n, opts) 2007 == 0) { 2008 /* should notify about lost packet */ 2009 m_freem(n); 2010 if (opts) { 2011 m_freem(opts); 2012 } 2013 SOCKBUF_UNLOCK( 2014 &last->inp_socket->so_rcv); 2015 } else 2016 sorwakeup_locked(last->inp_socket); 2017 opts = NULL; 2018 } 2019 INP_RUNLOCK(last); 2020 } 2021 last = in6p; 2022 } 2023 INP_INFO_RUNLOCK(&V_ripcbinfo); 2024 if (last != NULL) { 2025 if (last->inp_flags & INP_CONTROLOPTS) 2026 ip6_savecontrol(last, m, &opts); 2027 /* strip intermediate headers */ 2028 m_adj(m, off); 2029 2030 /* avoid using mbuf clusters if possible (see above) */ 2031 if ((m->m_flags & M_EXT) && m->m_next == NULL && 2032 m->m_len <= MHLEN) { 2033 struct mbuf *n; 2034 2035 MGET(n, M_DONTWAIT, m->m_type); 2036 if (n != NULL) { 2037 if (m_dup_pkthdr(n, m, M_NOWAIT)) { 2038 bcopy(m->m_data, n->m_data, m->m_len); 2039 n->m_len = m->m_len; 2040 2041 m_freem(m); 2042 m = n; 2043 } else { 2044 m_freem(n); 2045 n = NULL; 2046 } 2047 } 2048 } 2049 SOCKBUF_LOCK(&last->inp_socket->so_rcv); 2050 if (sbappendaddr_locked(&last->inp_socket->so_rcv, 2051 (struct sockaddr *)&fromsa, m, opts) == 0) { 2052 m_freem(m); 2053 if (opts) 2054 m_freem(opts); 2055 SOCKBUF_UNLOCK(&last->inp_socket->so_rcv); 2056 } else 2057 sorwakeup_locked(last->inp_socket); 2058 INP_RUNLOCK(last); 2059 } else { 2060 m_freem(m); 2061 IP6STAT_DEC(ip6s_delivered); 2062 } 2063 return IPPROTO_DONE; 2064} 2065 2066/* 2067 * Reflect the ip6 packet back to the source. 2068 * OFF points to the icmp6 header, counted from the top of the mbuf. 2069 */ 2070void 2071icmp6_reflect(struct mbuf *m, size_t off) 2072{ 2073 INIT_VNET_INET6(curvnet); 2074 struct ip6_hdr *ip6; 2075 struct icmp6_hdr *icmp6; 2076 struct in6_ifaddr *ia; 2077 int plen; 2078 int type, code; 2079 struct ifnet *outif = NULL; 2080 struct in6_addr origdst, *src = NULL; 2081 2082 /* too short to reflect */ 2083 if (off < sizeof(struct ip6_hdr)) { 2084 nd6log((LOG_DEBUG, 2085 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 2086 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2087 __FILE__, __LINE__)); 2088 goto bad; 2089 } 2090 2091 /* 2092 * If there are extra headers between IPv6 and ICMPv6, strip 2093 * off that header first. 2094 */ 2095#ifdef DIAGNOSTIC 2096 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2097 panic("assumption failed in icmp6_reflect"); 2098#endif 2099 if (off > sizeof(struct ip6_hdr)) { 2100 size_t l; 2101 struct ip6_hdr nip6; 2102 2103 l = off - sizeof(struct ip6_hdr); 2104 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2105 m_adj(m, l); 2106 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2107 if (m->m_len < l) { 2108 if ((m = m_pullup(m, l)) == NULL) 2109 return; 2110 } 2111 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2112 } else /* off == sizeof(struct ip6_hdr) */ { 2113 size_t l; 2114 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2115 if (m->m_len < l) { 2116 if ((m = m_pullup(m, l)) == NULL) 2117 return; 2118 } 2119 } 2120 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2121 ip6 = mtod(m, struct ip6_hdr *); 2122 ip6->ip6_nxt = IPPROTO_ICMPV6; 2123 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2124 type = icmp6->icmp6_type; /* keep type for statistics */ 2125 code = icmp6->icmp6_code; /* ditto. */ 2126 2127 origdst = ip6->ip6_dst; 2128 /* 2129 * ip6_input() drops a packet if its src is multicast. 2130 * So, the src is never multicast. 2131 */ 2132 ip6->ip6_dst = ip6->ip6_src; 2133 2134 /* 2135 * If the incoming packet was addressed directly to us (i.e. unicast), 2136 * use dst as the src for the reply. 2137 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2138 * (for example) when we encounter an error while forwarding procedure 2139 * destined to a duplicated address of ours. 2140 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2141 * procedure of an outgoing packet of our own, in which case we need 2142 * to search in the ifaddr list. 2143 */ 2144 if (!IN6_IS_ADDR_MULTICAST(&origdst)) { 2145 if ((ia = ip6_getdstifaddr(m))) { 2146 if (!(ia->ia6_flags & 2147 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) 2148 src = &ia->ia_addr.sin6_addr; 2149 } else { 2150 struct sockaddr_in6 d; 2151 2152 bzero(&d, sizeof(d)); 2153 d.sin6_family = AF_INET6; 2154 d.sin6_len = sizeof(d); 2155 d.sin6_addr = origdst; 2156 ia = (struct in6_ifaddr *) 2157 ifa_ifwithaddr((struct sockaddr *)&d); 2158 if (ia && 2159 !(ia->ia6_flags & 2160 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) { 2161 src = &ia->ia_addr.sin6_addr; 2162 } 2163 } 2164 } 2165 2166 if (src == NULL) { 2167 int e; 2168 struct sockaddr_in6 sin6; 2169 struct route_in6 ro; 2170 2171 /* 2172 * This case matches to multicasts, our anycast, or unicasts 2173 * that we do not own. Select a source address based on the 2174 * source address of the erroneous packet. 2175 */ 2176 bzero(&sin6, sizeof(sin6)); 2177 sin6.sin6_family = AF_INET6; 2178 sin6.sin6_len = sizeof(sin6); 2179 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */ 2180 2181 bzero(&ro, sizeof(ro)); 2182 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2183 if (ro.ro_rt) 2184 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2185 if (src == NULL) { 2186 char ip6buf[INET6_ADDRSTRLEN]; 2187 nd6log((LOG_DEBUG, 2188 "icmp6_reflect: source can't be determined: " 2189 "dst=%s, error=%d\n", 2190 ip6_sprintf(ip6buf, &sin6.sin6_addr), e)); 2191 goto bad; 2192 } 2193 } 2194 2195 ip6->ip6_src = *src; 2196 ip6->ip6_flow = 0; 2197 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2198 ip6->ip6_vfc |= IPV6_VERSION; 2199 ip6->ip6_nxt = IPPROTO_ICMPV6; 2200 if (outif) 2201 ip6->ip6_hlim = ND_IFINFO(outif)->chlim; 2202 else if (m->m_pkthdr.rcvif) { 2203 /* XXX: This may not be the outgoing interface */ 2204 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2205 } else 2206 ip6->ip6_hlim = V_ip6_defhlim; 2207 2208 icmp6->icmp6_cksum = 0; 2209 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2210 sizeof(struct ip6_hdr), plen); 2211 2212 /* 2213 * XXX option handling 2214 */ 2215 2216 m->m_flags &= ~(M_BCAST|M_MCAST); 2217 2218 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL); 2219 if (outif) 2220 icmp6_ifoutstat_inc(outif, type, code); 2221 2222 return; 2223 2224 bad: 2225 m_freem(m); 2226 return; 2227} 2228 2229void 2230icmp6_fasttimo(void) 2231{ 2232 2233 mld_fasttimo(); 2234} 2235 2236void 2237icmp6_slowtimo(void) 2238{ 2239 2240 mld_slowtimo(); 2241} 2242 2243static const char * 2244icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2245 struct in6_addr *tgt6) 2246{ 2247 static char buf[1024]; 2248 char ip6bufs[INET6_ADDRSTRLEN]; 2249 char ip6bufd[INET6_ADDRSTRLEN]; 2250 char ip6buft[INET6_ADDRSTRLEN]; 2251 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2252 ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6), 2253 ip6_sprintf(ip6buft, tgt6)); 2254 return buf; 2255} 2256 2257void 2258icmp6_redirect_input(struct mbuf *m, int off) 2259{ 2260 INIT_VNET_INET6(curvnet); 2261 struct ifnet *ifp; 2262 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2263 struct nd_redirect *nd_rd; 2264 int icmp6len = ntohs(ip6->ip6_plen); 2265 char *lladdr = NULL; 2266 int lladdrlen = 0; 2267 u_char *redirhdr = NULL; 2268 int redirhdrlen = 0; 2269 struct rtentry *rt = NULL; 2270 int is_router; 2271 int is_onlink; 2272 struct in6_addr src6 = ip6->ip6_src; 2273 struct in6_addr redtgt6; 2274 struct in6_addr reddst6; 2275 union nd_opts ndopts; 2276 char ip6buf[INET6_ADDRSTRLEN]; 2277 2278 if (!m) 2279 return; 2280 2281 ifp = m->m_pkthdr.rcvif; 2282 2283 if (!ifp) 2284 return; 2285 2286 /* XXX if we are router, we don't update route by icmp6 redirect */ 2287 if (V_ip6_forwarding) 2288 goto freeit; 2289 if (!V_icmp6_rediraccept) 2290 goto freeit; 2291 2292#ifndef PULLDOWN_TEST 2293 IP6_EXTHDR_CHECK(m, off, icmp6len,); 2294 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off); 2295#else 2296 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2297 if (nd_rd == NULL) { 2298 ICMP6STAT_INC(icp6s_tooshort); 2299 return; 2300 } 2301#endif 2302 redtgt6 = nd_rd->nd_rd_target; 2303 reddst6 = nd_rd->nd_rd_dst; 2304 2305 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2306 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2307 goto freeit; 2308 } 2309 2310 /* validation */ 2311 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2312 nd6log((LOG_ERR, 2313 "ICMP6 redirect sent from %s rejected; " 2314 "must be from linklocal\n", 2315 ip6_sprintf(ip6buf, &src6))); 2316 goto bad; 2317 } 2318 if (ip6->ip6_hlim != 255) { 2319 nd6log((LOG_ERR, 2320 "ICMP6 redirect sent from %s rejected; " 2321 "hlim=%d (must be 255)\n", 2322 ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim)); 2323 goto bad; 2324 } 2325 { 2326 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2327 struct sockaddr_in6 sin6; 2328 struct in6_addr *gw6; 2329 2330 bzero(&sin6, sizeof(sin6)); 2331 sin6.sin6_family = AF_INET6; 2332 sin6.sin6_len = sizeof(struct sockaddr_in6); 2333 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2334 rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL); 2335 if (rt) { 2336 if (rt->rt_gateway == NULL || 2337 rt->rt_gateway->sa_family != AF_INET6) { 2338 nd6log((LOG_ERR, 2339 "ICMP6 redirect rejected; no route " 2340 "with inet6 gateway found for redirect dst: %s\n", 2341 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2342 RTFREE_LOCKED(rt); 2343 goto bad; 2344 } 2345 2346 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2347 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2348 nd6log((LOG_ERR, 2349 "ICMP6 redirect rejected; " 2350 "not equal to gw-for-src=%s (must be same): " 2351 "%s\n", 2352 ip6_sprintf(ip6buf, gw6), 2353 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2354 RTFREE_LOCKED(rt); 2355 goto bad; 2356 } 2357 } else { 2358 nd6log((LOG_ERR, 2359 "ICMP6 redirect rejected; " 2360 "no route found for redirect dst: %s\n", 2361 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2362 goto bad; 2363 } 2364 RTFREE_LOCKED(rt); 2365 rt = NULL; 2366 } 2367 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2368 nd6log((LOG_ERR, 2369 "ICMP6 redirect rejected; " 2370 "redirect dst must be unicast: %s\n", 2371 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2372 goto bad; 2373 } 2374 2375 is_router = is_onlink = 0; 2376 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2377 is_router = 1; /* router case */ 2378 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2379 is_onlink = 1; /* on-link destination case */ 2380 if (!is_router && !is_onlink) { 2381 nd6log((LOG_ERR, 2382 "ICMP6 redirect rejected; " 2383 "neither router case nor onlink case: %s\n", 2384 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2385 goto bad; 2386 } 2387 /* validation passed */ 2388 2389 icmp6len -= sizeof(*nd_rd); 2390 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2391 if (nd6_options(&ndopts) < 0) { 2392 nd6log((LOG_INFO, "icmp6_redirect_input: " 2393 "invalid ND option, rejected: %s\n", 2394 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2395 /* nd6_options have incremented stats */ 2396 goto freeit; 2397 } 2398 2399 if (ndopts.nd_opts_tgt_lladdr) { 2400 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2401 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2402 } 2403 2404 if (ndopts.nd_opts_rh) { 2405 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len; 2406 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */ 2407 } 2408 2409 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2410 nd6log((LOG_INFO, 2411 "icmp6_redirect_input: lladdrlen mismatch for %s " 2412 "(if %d, icmp6 packet %d): %s\n", 2413 ip6_sprintf(ip6buf, &redtgt6), 2414 ifp->if_addrlen, lladdrlen - 2, 2415 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2416 goto bad; 2417 } 2418 2419 /* RFC 2461 8.3 */ 2420 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2421 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2422 2423 if (!is_onlink) { /* better router case. perform rtredirect. */ 2424 /* perform rtredirect */ 2425 struct sockaddr_in6 sdst; 2426 struct sockaddr_in6 sgw; 2427 struct sockaddr_in6 ssrc; 2428 2429 bzero(&sdst, sizeof(sdst)); 2430 bzero(&sgw, sizeof(sgw)); 2431 bzero(&ssrc, sizeof(ssrc)); 2432 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2433 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2434 sizeof(struct sockaddr_in6); 2435 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2436 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2437 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2438 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2439 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2440 (struct sockaddr *)&ssrc); 2441 } 2442 /* finally update cached route in each socket via pfctlinput */ 2443 { 2444 struct sockaddr_in6 sdst; 2445 2446 bzero(&sdst, sizeof(sdst)); 2447 sdst.sin6_family = AF_INET6; 2448 sdst.sin6_len = sizeof(struct sockaddr_in6); 2449 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2450 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2451#ifdef IPSEC 2452 key_sa_routechange((struct sockaddr *)&sdst); 2453#endif /* IPSEC */ 2454 } 2455 2456 freeit: 2457 m_freem(m); 2458 return; 2459 2460 bad: 2461 ICMP6STAT_INC(icp6s_badredirect); 2462 m_freem(m); 2463} 2464 2465void 2466icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2467{ 2468 INIT_VNET_INET6(curvnet); 2469 struct ifnet *ifp; /* my outgoing interface */ 2470 struct in6_addr *ifp_ll6; 2471 struct in6_addr *router_ll6; 2472 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2473 struct mbuf *m = NULL; /* newly allocated one */ 2474 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2475 struct nd_redirect *nd_rd; 2476 struct llentry *ln = NULL; 2477 size_t maxlen; 2478 u_char *p; 2479 struct ifnet *outif = NULL; 2480 struct sockaddr_in6 src_sa; 2481 2482 icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2483 2484 /* if we are not router, we don't send icmp6 redirect */ 2485 if (!V_ip6_forwarding) 2486 goto fail; 2487 2488 /* sanity check */ 2489 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2490 goto fail; 2491 2492 /* 2493 * Address check: 2494 * the source address must identify a neighbor, and 2495 * the destination address must not be a multicast address 2496 * [RFC 2461, sec 8.2] 2497 */ 2498 sip6 = mtod(m0, struct ip6_hdr *); 2499 bzero(&src_sa, sizeof(src_sa)); 2500 src_sa.sin6_family = AF_INET6; 2501 src_sa.sin6_len = sizeof(src_sa); 2502 src_sa.sin6_addr = sip6->ip6_src; 2503 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2504 goto fail; 2505 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2506 goto fail; /* what should we do here? */ 2507 2508 /* rate limit */ 2509 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2510 goto fail; 2511 2512 /* 2513 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2514 * we almost always ask for an mbuf cluster for simplicity. 2515 * (MHLEN < IPV6_MMTU is almost always true) 2516 */ 2517#if IPV6_MMTU >= MCLBYTES 2518# error assumption failed about IPV6_MMTU and MCLBYTES 2519#endif 2520 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2521 if (m && IPV6_MMTU >= MHLEN) 2522 MCLGET(m, M_DONTWAIT); 2523 if (!m) 2524 goto fail; 2525 m->m_pkthdr.rcvif = NULL; 2526 m->m_len = 0; 2527 maxlen = M_TRAILINGSPACE(m); 2528 maxlen = min(IPV6_MMTU, maxlen); 2529 /* just for safety */ 2530 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2531 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2532 goto fail; 2533 } 2534 2535 { 2536 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2537 struct in6_ifaddr *ia; 2538 if ((ia = in6ifa_ifpforlinklocal(ifp, 2539 IN6_IFF_NOTREADY| 2540 IN6_IFF_ANYCAST)) == NULL) 2541 goto fail; 2542 ifp_ll6 = &ia->ia_addr.sin6_addr; 2543 } 2544 2545 /* get ip6 linklocal address for the router. */ 2546 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2547 struct sockaddr_in6 *sin6; 2548 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2549 router_ll6 = &sin6->sin6_addr; 2550 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6)) 2551 router_ll6 = (struct in6_addr *)NULL; 2552 } else 2553 router_ll6 = (struct in6_addr *)NULL; 2554 2555 /* ip6 */ 2556 ip6 = mtod(m, struct ip6_hdr *); 2557 ip6->ip6_flow = 0; 2558 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2559 ip6->ip6_vfc |= IPV6_VERSION; 2560 /* ip6->ip6_plen will be set later */ 2561 ip6->ip6_nxt = IPPROTO_ICMPV6; 2562 ip6->ip6_hlim = 255; 2563 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2564 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2565 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2566 2567 /* ND Redirect */ 2568 nd_rd = (struct nd_redirect *)(ip6 + 1); 2569 nd_rd->nd_rd_type = ND_REDIRECT; 2570 nd_rd->nd_rd_code = 0; 2571 nd_rd->nd_rd_reserved = 0; 2572 if (rt->rt_flags & RTF_GATEWAY) { 2573 /* 2574 * nd_rd->nd_rd_target must be a link-local address in 2575 * better router cases. 2576 */ 2577 if (!router_ll6) 2578 goto fail; 2579 bcopy(router_ll6, &nd_rd->nd_rd_target, 2580 sizeof(nd_rd->nd_rd_target)); 2581 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2582 sizeof(nd_rd->nd_rd_dst)); 2583 } else { 2584 /* make sure redtgt == reddst */ 2585 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2586 sizeof(nd_rd->nd_rd_target)); 2587 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2588 sizeof(nd_rd->nd_rd_dst)); 2589 } 2590 2591 p = (u_char *)(nd_rd + 1); 2592 2593 if (!router_ll6) 2594 goto nolladdropt; 2595 2596 { 2597 /* target lladdr option */ 2598 int len; 2599 struct nd_opt_hdr *nd_opt; 2600 char *lladdr; 2601 2602 IF_AFDATA_LOCK(ifp); 2603 ln = nd6_lookup(router_ll6, 0, ifp); 2604 IF_AFDATA_UNLOCK(ifp); 2605 if (ln == NULL) 2606 goto nolladdropt; 2607 2608 len = sizeof(*nd_opt) + ifp->if_addrlen; 2609 len = (len + 7) & ~7; /* round by 8 */ 2610 /* safety check */ 2611 if (len + (p - (u_char *)ip6) > maxlen) 2612 goto nolladdropt; 2613 2614 if (ln->la_flags & LLE_VALID) { 2615 nd_opt = (struct nd_opt_hdr *)p; 2616 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2617 nd_opt->nd_opt_len = len >> 3; 2618 lladdr = (char *)(nd_opt + 1); 2619 bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen); 2620 p += len; 2621 } 2622 } 2623nolladdropt: 2624 if (ln != NULL) 2625 LLE_RUNLOCK(ln); 2626 2627 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2628 2629 /* just to be safe */ 2630#ifdef M_DECRYPTED /*not openbsd*/ 2631 if (m0->m_flags & M_DECRYPTED) 2632 goto noredhdropt; 2633#endif 2634 if (p - (u_char *)ip6 > maxlen) 2635 goto noredhdropt; 2636 2637 { 2638 /* redirected header option */ 2639 int len; 2640 struct nd_opt_rd_hdr *nd_opt_rh; 2641 2642 /* 2643 * compute the maximum size for icmp6 redirect header option. 2644 * XXX room for auth header? 2645 */ 2646 len = maxlen - (p - (u_char *)ip6); 2647 len &= ~7; 2648 2649 /* This is just for simplicity. */ 2650 if (m0->m_pkthdr.len != m0->m_len) { 2651 if (m0->m_next) { 2652 m_freem(m0->m_next); 2653 m0->m_next = NULL; 2654 } 2655 m0->m_pkthdr.len = m0->m_len; 2656 } 2657 2658 /* 2659 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2660 * about padding/truncate rule for the original IP packet. 2661 * From the discussion on IPv6imp in Feb 1999, 2662 * the consensus was: 2663 * - "attach as much as possible" is the goal 2664 * - pad if not aligned (original size can be guessed by 2665 * original ip6 header) 2666 * Following code adds the padding if it is simple enough, 2667 * and truncates if not. 2668 */ 2669 if (m0->m_next || m0->m_pkthdr.len != m0->m_len) 2670 panic("assumption failed in %s:%d", __FILE__, 2671 __LINE__); 2672 2673 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2674 /* not enough room, truncate */ 2675 m0->m_pkthdr.len = m0->m_len = len - 2676 sizeof(*nd_opt_rh); 2677 } else { 2678 /* enough room, pad or truncate */ 2679 size_t extra; 2680 2681 extra = m0->m_pkthdr.len % 8; 2682 if (extra) { 2683 /* pad if easy enough, truncate if not */ 2684 if (8 - extra <= M_TRAILINGSPACE(m0)) { 2685 /* pad */ 2686 m0->m_len += (8 - extra); 2687 m0->m_pkthdr.len += (8 - extra); 2688 } else { 2689 /* truncate */ 2690 m0->m_pkthdr.len -= extra; 2691 m0->m_len -= extra; 2692 } 2693 } 2694 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2695 m0->m_pkthdr.len = m0->m_len = len - 2696 sizeof(*nd_opt_rh); 2697 } 2698 2699 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2700 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2701 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2702 nd_opt_rh->nd_opt_rh_len = len >> 3; 2703 p += sizeof(*nd_opt_rh); 2704 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2705 2706 /* connect m0 to m */ 2707 m_tag_delete_chain(m0, NULL); 2708 m0->m_flags &= ~M_PKTHDR; 2709 m->m_next = m0; 2710 m->m_pkthdr.len = m->m_len + m0->m_len; 2711 m0 = NULL; 2712 } 2713noredhdropt:; 2714 if (m0) { 2715 m_freem(m0); 2716 m0 = NULL; 2717 } 2718 2719 /* XXX: clear embedded link IDs in the inner header */ 2720 in6_clearscope(&sip6->ip6_src); 2721 in6_clearscope(&sip6->ip6_dst); 2722 in6_clearscope(&nd_rd->nd_rd_target); 2723 in6_clearscope(&nd_rd->nd_rd_dst); 2724 2725 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2726 2727 nd_rd->nd_rd_cksum = 0; 2728 nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2729 sizeof(*ip6), ntohs(ip6->ip6_plen)); 2730 2731 /* send the packet to outside... */ 2732 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL); 2733 if (outif) { 2734 icmp6_ifstat_inc(outif, ifs6_out_msg); 2735 icmp6_ifstat_inc(outif, ifs6_out_redirect); 2736 } 2737 ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]); 2738 2739 return; 2740 2741fail: 2742 if (m) 2743 m_freem(m); 2744 if (m0) 2745 m_freem(m0); 2746} 2747 2748/* 2749 * ICMPv6 socket option processing. 2750 */ 2751int 2752icmp6_ctloutput(struct socket *so, struct sockopt *sopt) 2753{ 2754 int error = 0; 2755 int optlen; 2756 struct inpcb *inp = sotoinpcb(so); 2757 int level, op, optname; 2758 2759 if (sopt) { 2760 level = sopt->sopt_level; 2761 op = sopt->sopt_dir; 2762 optname = sopt->sopt_name; 2763 optlen = sopt->sopt_valsize; 2764 } else 2765 level = op = optname = optlen = 0; 2766 2767 if (level != IPPROTO_ICMPV6) { 2768 return EINVAL; 2769 } 2770 2771 switch (op) { 2772 case PRCO_SETOPT: 2773 switch (optname) { 2774 case ICMP6_FILTER: 2775 { 2776 struct icmp6_filter ic6f; 2777 2778 if (optlen != sizeof(ic6f)) { 2779 error = EMSGSIZE; 2780 break; 2781 } 2782 error = sooptcopyin(sopt, &ic6f, optlen, optlen); 2783 if (error == 0) { 2784 INP_WLOCK(inp); 2785 *inp->in6p_icmp6filt = ic6f; 2786 INP_WUNLOCK(inp); 2787 } 2788 break; 2789 } 2790 2791 default: 2792 error = ENOPROTOOPT; 2793 break; 2794 } 2795 break; 2796 2797 case PRCO_GETOPT: 2798 switch (optname) { 2799 case ICMP6_FILTER: 2800 { 2801 struct icmp6_filter ic6f; 2802 2803 INP_RLOCK(inp); 2804 ic6f = *inp->in6p_icmp6filt; 2805 INP_RUNLOCK(inp); 2806 error = sooptcopyout(sopt, &ic6f, sizeof(ic6f)); 2807 break; 2808 } 2809 2810 default: 2811 error = ENOPROTOOPT; 2812 break; 2813 } 2814 break; 2815 } 2816 2817 return (error); 2818} 2819 2820/* 2821 * Perform rate limit check. 2822 * Returns 0 if it is okay to send the icmp6 packet. 2823 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2824 * limitation. 2825 * 2826 * XXX per-destination/type check necessary? 2827 * 2828 * dst - not used at this moment 2829 * type - not used at this moment 2830 * code - not used at this moment 2831 */ 2832static int 2833icmp6_ratelimit(const struct in6_addr *dst, const int type, 2834 const int code) 2835{ 2836 INIT_VNET_INET6(curvnet); 2837 int ret; 2838 2839 ret = 0; /* okay to send */ 2840 2841 /* PPS limit */ 2842 if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count, 2843 V_icmp6errppslim)) { 2844 /* The packet is subject to rate limit */ 2845 ret++; 2846 } 2847 2848 return ret; 2849} 2850