icmp6.c revision 194118
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 194118 2009-06-13 15:39:12Z jamie $"); 65 66#include "opt_inet.h" 67#include "opt_inet6.h" 68#include "opt_ipsec.h" 69 70#include <sys/param.h> 71#include <sys/domain.h> 72#include <sys/jail.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 struct mbuf *m = *mp, *n; 405 struct ifnet *ifp; 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 ifp = m->m_pkthdr.rcvif; 414 415#ifndef PULLDOWN_TEST 416 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE); 417 /* m might change if M_LOOP. So, call mtod after this */ 418#endif 419 420 /* 421 * Locate icmp6 structure in mbuf, and check 422 * that not corrupted and of at least minimum length 423 */ 424 425 ip6 = mtod(m, struct ip6_hdr *); 426 if (icmp6len < sizeof(struct icmp6_hdr)) { 427 ICMP6STAT_INC(icp6s_tooshort); 428 goto freeit; 429 } 430 431 /* 432 * Check multicast group membership. 433 * Note: SSM filters are not applied for ICMPv6 traffic. 434 */ 435 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 436 struct in6_multi *inm; 437 438 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(ifp, ifs6_in_msg); 487 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK) 488 icmp6_ifstat_inc(ifp, ifs6_in_error); 489 490 switch (icmp6->icmp6_type) { 491 case ICMP6_DST_UNREACH: 492 icmp6_ifstat_inc(ifp, 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(ifp, 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(ifp, 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(ifp, 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(ifp, 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(ifp, 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(ifp, 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, has a hop limit 637 * of greater than 1 hop, or which does not have the 638 * IPv6 HBH Router Alert option. 639 * As IPv6 HBH options are stripped in ip6_input() we must 640 * check an mbuf header flag. 641 * XXX Should we also sanity check that these messages 642 * were directed to a link-local multicast prefix? 643 */ 644 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0) 645 goto freeit; 646 if (mld_input(m, off, icmp6len) != 0) 647 return (IPPROTO_DONE); 648 /* m stays. */ 649 break; 650 651 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 652 { 653 enum { WRU, FQDN } mode; 654 655 if (!V_icmp6_nodeinfo) 656 break; 657 658 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 659 mode = WRU; 660 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 661 mode = FQDN; 662 else 663 goto badlen; 664 665 if (mode == FQDN) { 666#ifndef PULLDOWN_TEST 667 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo), 668 IPPROTO_DONE); 669#endif 670 n = m_copy(m, 0, M_COPYALL); 671 if (n) 672 n = ni6_input(n, off); 673 /* XXX meaningless if n == NULL */ 674 noff = sizeof(struct ip6_hdr); 675 } else { 676 struct prison *pr; 677 u_char *p; 678 int maxlen, maxhlen, hlen; 679 680 /* 681 * XXX: this combination of flags is pointless, 682 * but should we keep this for compatibility? 683 */ 684 if ((V_icmp6_nodeinfo & 5) != 5) 685 break; 686 687 if (code != 0) 688 goto badcode; 689 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; 690 if (maxlen >= MCLBYTES) { 691 /* Give up remote */ 692 break; 693 } 694 MGETHDR(n, M_DONTWAIT, m->m_type); 695 if (n && maxlen > MHLEN) { 696 MCLGET(n, M_DONTWAIT); 697 if ((n->m_flags & M_EXT) == 0) { 698 m_free(n); 699 n = NULL; 700 } 701 } 702 if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) { 703 /* 704 * Previous code did a blind M_COPY_PKTHDR 705 * and said "just for rcvif". If true, then 706 * we could tolerate the dup failing (due to 707 * the deep copy of the tag chain). For now 708 * be conservative and just fail. 709 */ 710 m_free(n); 711 n = NULL; 712 } 713 if (n == NULL) { 714 /* Give up remote */ 715 break; 716 } 717 n->m_pkthdr.rcvif = NULL; 718 n->m_len = 0; 719 maxhlen = M_TRAILINGSPACE(n) - maxlen; 720 pr = curthread->td_ucred->cr_prison; 721 mtx_lock(&pr->pr_mtx); 722 hlen = strlen(pr->pr_hostname); 723 if (maxhlen > hlen) 724 maxhlen = hlen; 725 /* 726 * Copy IPv6 and ICMPv6 only. 727 */ 728 nip6 = mtod(n, struct ip6_hdr *); 729 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 730 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 731 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 732 p = (u_char *)(nicmp6 + 1); 733 bzero(p, 4); 734 /* meaningless TTL */ 735 bcopy(pr->pr_hostname, p + 4, maxhlen); 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_hostname, 1339 strlen(pr->pr_hostname), 0); 1340 mtx_unlock(&pr->pr_mtx); 1341 if (!n || n->m_next || n->m_len == 0) 1342 goto bad; 1343 IP6_EXTHDR_GET(subj, char *, m, 1344 off + sizeof(struct icmp6_nodeinfo), subjlen); 1345 if (subj == NULL) 1346 goto bad; 1347 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1348 n->m_len)) { 1349 goto bad; 1350 } 1351 m_freem(n); 1352 n = NULL; 1353 break; 1354 1355 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1356 default: 1357 goto bad; 1358 } 1359 break; 1360 } 1361 1362 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1363 switch (qtype) { 1364 case NI_QTYPE_FQDN: 1365 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0) 1366 goto bad; 1367 break; 1368 case NI_QTYPE_NODEADDR: 1369 case NI_QTYPE_IPV4ADDR: 1370 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0) 1371 goto bad; 1372 break; 1373 } 1374 1375 /* guess reply length */ 1376 switch (qtype) { 1377 case NI_QTYPE_NOOP: 1378 break; /* no reply data */ 1379 case NI_QTYPE_SUPTYPES: 1380 replylen += sizeof(u_int32_t); 1381 break; 1382 case NI_QTYPE_FQDN: 1383 /* XXX will append an mbuf */ 1384 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1385 break; 1386 case NI_QTYPE_NODEADDR: 1387 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj); 1388 if ((replylen += addrs * (sizeof(struct in6_addr) + 1389 sizeof(u_int32_t))) > MCLBYTES) 1390 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1391 break; 1392 case NI_QTYPE_IPV4ADDR: 1393 /* unsupported - should respond with unknown Qtype? */ 1394 break; 1395 default: 1396 /* 1397 * XXX: We must return a reply with the ICMP6 code 1398 * `unknown Qtype' in this case. However we regard the case 1399 * as an FQDN query for backward compatibility. 1400 * Older versions set a random value to this field, 1401 * so it rarely varies in the defined qtypes. 1402 * But the mechanism is not reliable... 1403 * maybe we should obsolete older versions. 1404 */ 1405 qtype = NI_QTYPE_FQDN; 1406 /* XXX will append an mbuf */ 1407 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1408 oldfqdn++; 1409 break; 1410 } 1411 1412 /* allocate an mbuf to reply. */ 1413 MGETHDR(n, M_DONTWAIT, m->m_type); 1414 if (n == NULL) { 1415 m_freem(m); 1416 return (NULL); 1417 } 1418 M_MOVE_PKTHDR(n, m); /* just for recvif */ 1419 if (replylen > MHLEN) { 1420 if (replylen > MCLBYTES) { 1421 /* 1422 * XXX: should we try to allocate more? But MCLBYTES 1423 * is probably much larger than IPV6_MMTU... 1424 */ 1425 goto bad; 1426 } 1427 MCLGET(n, M_DONTWAIT); 1428 if ((n->m_flags & M_EXT) == 0) { 1429 goto bad; 1430 } 1431 } 1432 n->m_pkthdr.len = n->m_len = replylen; 1433 1434 /* copy mbuf header and IPv6 + Node Information base headers */ 1435 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1436 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1437 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1438 1439 /* qtype dependent procedure */ 1440 switch (qtype) { 1441 case NI_QTYPE_NOOP: 1442 nni6->ni_code = ICMP6_NI_SUCCESS; 1443 nni6->ni_flags = 0; 1444 break; 1445 case NI_QTYPE_SUPTYPES: 1446 { 1447 u_int32_t v; 1448 nni6->ni_code = ICMP6_NI_SUCCESS; 1449 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1450 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1451 v = (u_int32_t)htonl(0x0000000f); 1452 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1453 break; 1454 } 1455 case NI_QTYPE_FQDN: 1456 nni6->ni_code = ICMP6_NI_SUCCESS; 1457 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1458 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo)); 1459 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1460 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1461 /* 1462 * XXX do we really have FQDN in hostname? 1463 */ 1464 pr = curthread->td_ucred->cr_prison; 1465 mtx_lock(&pr->pr_mtx); 1466 n->m_next = ni6_nametodns(pr->pr_hostname, 1467 strlen(pr->pr_hostname), oldfqdn); 1468 mtx_unlock(&pr->pr_mtx); 1469 if (n->m_next == NULL) 1470 goto bad; 1471 /* XXX we assume that n->m_next is not a chain */ 1472 if (n->m_next->m_next != NULL) 1473 goto bad; 1474 n->m_pkthdr.len += n->m_next->m_len; 1475 break; 1476 case NI_QTYPE_NODEADDR: 1477 { 1478 int lenlim, copied; 1479 1480 nni6->ni_code = ICMP6_NI_SUCCESS; 1481 n->m_pkthdr.len = n->m_len = 1482 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1483 lenlim = M_TRAILINGSPACE(n); 1484 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1485 /* XXX: reset mbuf length */ 1486 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1487 sizeof(struct icmp6_nodeinfo) + copied; 1488 break; 1489 } 1490 default: 1491 break; /* XXX impossible! */ 1492 } 1493 1494 nni6->ni_type = ICMP6_NI_REPLY; 1495 m_freem(m); 1496 return (n); 1497 1498 bad: 1499 m_freem(m); 1500 if (n) 1501 m_freem(n); 1502 return (NULL); 1503} 1504 1505/* 1506 * make a mbuf with DNS-encoded string. no compression support. 1507 * 1508 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1509 * treated as truncated name (two \0 at the end). this is a wild guess. 1510 * 1511 * old - return pascal string if non-zero 1512 */ 1513static struct mbuf * 1514ni6_nametodns(const char *name, int namelen, int old) 1515{ 1516 struct mbuf *m; 1517 char *cp, *ep; 1518 const char *p, *q; 1519 int i, len, nterm; 1520 1521 if (old) 1522 len = namelen + 1; 1523 else 1524 len = MCLBYTES; 1525 1526 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1527 MGET(m, M_DONTWAIT, MT_DATA); 1528 if (m && len > MLEN) { 1529 MCLGET(m, M_DONTWAIT); 1530 if ((m->m_flags & M_EXT) == 0) 1531 goto fail; 1532 } 1533 if (!m) 1534 goto fail; 1535 m->m_next = NULL; 1536 1537 if (old) { 1538 m->m_len = len; 1539 *mtod(m, char *) = namelen; 1540 bcopy(name, mtod(m, char *) + 1, namelen); 1541 return m; 1542 } else { 1543 m->m_len = 0; 1544 cp = mtod(m, char *); 1545 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1546 1547 /* if not certain about my name, return empty buffer */ 1548 if (namelen == 0) 1549 return m; 1550 1551 /* 1552 * guess if it looks like shortened hostname, or FQDN. 1553 * shortened hostname needs two trailing "\0". 1554 */ 1555 i = 0; 1556 for (p = name; p < name + namelen; p++) { 1557 if (*p && *p == '.') 1558 i++; 1559 } 1560 if (i < 2) 1561 nterm = 2; 1562 else 1563 nterm = 1; 1564 1565 p = name; 1566 while (cp < ep && p < name + namelen) { 1567 i = 0; 1568 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1569 i++; 1570 /* result does not fit into mbuf */ 1571 if (cp + i + 1 >= ep) 1572 goto fail; 1573 /* 1574 * DNS label length restriction, RFC1035 page 8. 1575 * "i == 0" case is included here to avoid returning 1576 * 0-length label on "foo..bar". 1577 */ 1578 if (i <= 0 || i >= 64) 1579 goto fail; 1580 *cp++ = i; 1581 bcopy(p, cp, i); 1582 cp += i; 1583 p = q; 1584 if (p < name + namelen && *p == '.') 1585 p++; 1586 } 1587 /* termination */ 1588 if (cp + nterm >= ep) 1589 goto fail; 1590 while (nterm-- > 0) 1591 *cp++ = '\0'; 1592 m->m_len = cp - mtod(m, char *); 1593 return m; 1594 } 1595 1596 panic("should not reach here"); 1597 /* NOTREACHED */ 1598 1599 fail: 1600 if (m) 1601 m_freem(m); 1602 return NULL; 1603} 1604 1605/* 1606 * check if two DNS-encoded string matches. takes care of truncated 1607 * form (with \0\0 at the end). no compression support. 1608 * XXX upper/lowercase match (see RFC2065) 1609 */ 1610static int 1611ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1612{ 1613 const char *a0, *b0; 1614 int l; 1615 1616 /* simplest case - need validation? */ 1617 if (alen == blen && bcmp(a, b, alen) == 0) 1618 return 1; 1619 1620 a0 = a; 1621 b0 = b; 1622 1623 /* termination is mandatory */ 1624 if (alen < 2 || blen < 2) 1625 return 0; 1626 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1627 return 0; 1628 alen--; 1629 blen--; 1630 1631 while (a - a0 < alen && b - b0 < blen) { 1632 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1633 return 0; 1634 1635 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1636 return 0; 1637 /* we don't support compression yet */ 1638 if (a[0] >= 64 || b[0] >= 64) 1639 return 0; 1640 1641 /* truncated case */ 1642 if (a[0] == 0 && a - a0 == alen - 1) 1643 return 1; 1644 if (b[0] == 0 && b - b0 == blen - 1) 1645 return 1; 1646 if (a[0] == 0 || b[0] == 0) 1647 return 0; 1648 1649 if (a[0] != b[0]) 1650 return 0; 1651 l = a[0]; 1652 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1653 return 0; 1654 if (bcmp(a + 1, b + 1, l) != 0) 1655 return 0; 1656 1657 a += 1 + l; 1658 b += 1 + l; 1659 } 1660 1661 if (a - a0 == alen && b - b0 == blen) 1662 return 1; 1663 else 1664 return 0; 1665} 1666 1667/* 1668 * calculate the number of addresses to be returned in the node info reply. 1669 */ 1670static int 1671ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp, 1672 struct in6_addr *subj) 1673{ 1674 INIT_VNET_NET(curvnet); 1675 INIT_VNET_INET6(curvnet); 1676 struct ifnet *ifp; 1677 struct in6_ifaddr *ifa6; 1678 struct ifaddr *ifa; 1679 int addrs = 0, addrsofif, iffound = 0; 1680 int niflags = ni6->ni_flags; 1681 1682 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1683 switch (ni6->ni_code) { 1684 case ICMP6_NI_SUBJ_IPV6: 1685 if (subj == NULL) /* must be impossible... */ 1686 return (0); 1687 break; 1688 default: 1689 /* 1690 * XXX: we only support IPv6 subject address for 1691 * this Qtype. 1692 */ 1693 return (0); 1694 } 1695 } 1696 1697 IFNET_RLOCK(); 1698 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) { 1699 addrsofif = 0; 1700 IF_ADDR_LOCK(ifp); 1701 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1702 if (ifa->ifa_addr->sa_family != AF_INET6) 1703 continue; 1704 ifa6 = (struct in6_ifaddr *)ifa; 1705 1706 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1707 IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr)) 1708 iffound = 1; 1709 1710 /* 1711 * IPv4-mapped addresses can only be returned by a 1712 * Node Information proxy, since they represent 1713 * addresses of IPv4-only nodes, which perforce do 1714 * not implement this protocol. 1715 * [icmp-name-lookups-07, Section 5.4] 1716 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1717 * this function at this moment. 1718 */ 1719 1720 /* What do we have to do about ::1? */ 1721 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1722 case IPV6_ADDR_SCOPE_LINKLOCAL: 1723 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1724 continue; 1725 break; 1726 case IPV6_ADDR_SCOPE_SITELOCAL: 1727 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1728 continue; 1729 break; 1730 case IPV6_ADDR_SCOPE_GLOBAL: 1731 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1732 continue; 1733 break; 1734 default: 1735 continue; 1736 } 1737 1738 /* 1739 * check if anycast is okay. 1740 * XXX: just experimental. not in the spec. 1741 */ 1742 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1743 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1744 continue; /* we need only unicast addresses */ 1745 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1746 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) { 1747 continue; 1748 } 1749 addrsofif++; /* count the address */ 1750 } 1751 IF_ADDR_UNLOCK(ifp); 1752 if (iffound) { 1753 *ifpp = ifp; 1754 IFNET_RUNLOCK(); 1755 return (addrsofif); 1756 } 1757 1758 addrs += addrsofif; 1759 } 1760 IFNET_RUNLOCK(); 1761 1762 return (addrs); 1763} 1764 1765static int 1766ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6, 1767 struct ifnet *ifp0, int resid) 1768{ 1769 INIT_VNET_NET(curvnet); 1770 INIT_VNET_INET6(curvnet); 1771 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet); 1772 struct in6_ifaddr *ifa6; 1773 struct ifaddr *ifa; 1774 struct ifnet *ifp_dep = NULL; 1775 int copied = 0, allow_deprecated = 0; 1776 u_char *cp = (u_char *)(nni6 + 1); 1777 int niflags = ni6->ni_flags; 1778 u_int32_t ltime; 1779 1780 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1781 return (0); /* needless to copy */ 1782 1783 IFNET_RLOCK(); 1784 again: 1785 1786 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) { 1787 IF_ADDR_LOCK(ifp); 1788 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1789 if (ifa->ifa_addr->sa_family != AF_INET6) 1790 continue; 1791 ifa6 = (struct in6_ifaddr *)ifa; 1792 1793 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1794 allow_deprecated == 0) { 1795 /* 1796 * prefererred address should be put before 1797 * deprecated addresses. 1798 */ 1799 1800 /* record the interface for later search */ 1801 if (ifp_dep == NULL) 1802 ifp_dep = ifp; 1803 1804 continue; 1805 } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1806 allow_deprecated != 0) 1807 continue; /* we now collect deprecated addrs */ 1808 1809 /* What do we have to do about ::1? */ 1810 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1811 case IPV6_ADDR_SCOPE_LINKLOCAL: 1812 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1813 continue; 1814 break; 1815 case IPV6_ADDR_SCOPE_SITELOCAL: 1816 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1817 continue; 1818 break; 1819 case IPV6_ADDR_SCOPE_GLOBAL: 1820 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1821 continue; 1822 break; 1823 default: 1824 continue; 1825 } 1826 1827 /* 1828 * check if anycast is okay. 1829 * XXX: just experimental. not in the spec. 1830 */ 1831 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1832 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1833 continue; 1834 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1835 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) { 1836 continue; 1837 } 1838 1839 /* now we can copy the address */ 1840 if (resid < sizeof(struct in6_addr) + 1841 sizeof(u_int32_t)) { 1842 IF_ADDR_UNLOCK(ifp); 1843 /* 1844 * We give up much more copy. 1845 * Set the truncate flag and return. 1846 */ 1847 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1848 IFNET_RUNLOCK(); 1849 return (copied); 1850 } 1851 1852 /* 1853 * Set the TTL of the address. 1854 * The TTL value should be one of the following 1855 * according to the specification: 1856 * 1857 * 1. The remaining lifetime of a DHCP lease on the 1858 * address, or 1859 * 2. The remaining Valid Lifetime of a prefix from 1860 * which the address was derived through Stateless 1861 * Autoconfiguration. 1862 * 1863 * Note that we currently do not support stateful 1864 * address configuration by DHCPv6, so the former 1865 * case can't happen. 1866 */ 1867 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1868 ltime = ND6_INFINITE_LIFETIME; 1869 else { 1870 if (ifa6->ia6_lifetime.ia6t_expire > 1871 time_second) 1872 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second); 1873 else 1874 ltime = 0; 1875 } 1876 1877 bcopy(<ime, cp, sizeof(u_int32_t)); 1878 cp += sizeof(u_int32_t); 1879 1880 /* copy the address itself */ 1881 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1882 sizeof(struct in6_addr)); 1883 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1884 cp += sizeof(struct in6_addr); 1885 1886 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1887 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1888 } 1889 IF_ADDR_UNLOCK(ifp); 1890 if (ifp0) /* we need search only on the specified IF */ 1891 break; 1892 } 1893 1894 if (allow_deprecated == 0 && ifp_dep != NULL) { 1895 ifp = ifp_dep; 1896 allow_deprecated = 1; 1897 1898 goto again; 1899 } 1900 1901 IFNET_RUNLOCK(); 1902 1903 return (copied); 1904} 1905 1906/* 1907 * XXX almost dup'ed code with rip6_input. 1908 */ 1909static int 1910icmp6_rip6_input(struct mbuf **mp, int off) 1911{ 1912 INIT_VNET_INET(curvnet); 1913 INIT_VNET_INET6(curvnet); 1914 struct mbuf *m = *mp; 1915 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1916 struct inpcb *in6p; 1917 struct inpcb *last = NULL; 1918 struct sockaddr_in6 fromsa; 1919 struct icmp6_hdr *icmp6; 1920 struct mbuf *opts = NULL; 1921 1922#ifndef PULLDOWN_TEST 1923 /* this is assumed to be safe. */ 1924 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 1925#else 1926 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1927 if (icmp6 == NULL) { 1928 /* m is already reclaimed */ 1929 return (IPPROTO_DONE); 1930 } 1931#endif 1932 1933 /* 1934 * XXX: the address may have embedded scope zone ID, which should be 1935 * hidden from applications. 1936 */ 1937 bzero(&fromsa, sizeof(fromsa)); 1938 fromsa.sin6_family = AF_INET6; 1939 fromsa.sin6_len = sizeof(struct sockaddr_in6); 1940 fromsa.sin6_addr = ip6->ip6_src; 1941 if (sa6_recoverscope(&fromsa)) { 1942 m_freem(m); 1943 return (IPPROTO_DONE); 1944 } 1945 1946 INP_INFO_RLOCK(&V_ripcbinfo); 1947 LIST_FOREACH(in6p, &V_ripcb, inp_list) { 1948 if ((in6p->inp_vflag & INP_IPV6) == 0) 1949 continue; 1950 if (in6p->inp_ip_p != IPPROTO_ICMPV6) 1951 continue; 1952 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1953 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1954 continue; 1955 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1956 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1957 continue; 1958 INP_RLOCK(in6p); 1959 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1960 in6p->in6p_icmp6filt)) { 1961 INP_RUNLOCK(in6p); 1962 continue; 1963 } 1964 if (last != NULL) { 1965 struct mbuf *n = NULL; 1966 1967 /* 1968 * Recent network drivers tend to allocate a single 1969 * mbuf cluster, rather than to make a couple of 1970 * mbufs without clusters. Also, since the IPv6 code 1971 * path tries to avoid m_pullup(), it is highly 1972 * probable that we still have an mbuf cluster here 1973 * even though the necessary length can be stored in an 1974 * mbuf's internal buffer. 1975 * Meanwhile, the default size of the receive socket 1976 * buffer for raw sockets is not so large. This means 1977 * the possibility of packet loss is relatively higher 1978 * than before. To avoid this scenario, we copy the 1979 * received data to a separate mbuf that does not use 1980 * a cluster, if possible. 1981 * XXX: it is better to copy the data after stripping 1982 * intermediate headers. 1983 */ 1984 if ((m->m_flags & M_EXT) && m->m_next == NULL && 1985 m->m_len <= MHLEN) { 1986 MGET(n, M_DONTWAIT, m->m_type); 1987 if (n != NULL) { 1988 if (m_dup_pkthdr(n, m, M_NOWAIT)) { 1989 bcopy(m->m_data, n->m_data, 1990 m->m_len); 1991 n->m_len = m->m_len; 1992 } else { 1993 m_free(n); 1994 n = NULL; 1995 } 1996 } 1997 } 1998 if (n != NULL || 1999 (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 2000 if (last->inp_flags & INP_CONTROLOPTS) 2001 ip6_savecontrol(last, n, &opts); 2002 /* strip intermediate headers */ 2003 m_adj(n, off); 2004 SOCKBUF_LOCK(&last->inp_socket->so_rcv); 2005 if (sbappendaddr_locked( 2006 &last->inp_socket->so_rcv, 2007 (struct sockaddr *)&fromsa, n, opts) 2008 == 0) { 2009 /* should notify about lost packet */ 2010 m_freem(n); 2011 if (opts) { 2012 m_freem(opts); 2013 } 2014 SOCKBUF_UNLOCK( 2015 &last->inp_socket->so_rcv); 2016 } else 2017 sorwakeup_locked(last->inp_socket); 2018 opts = NULL; 2019 } 2020 INP_RUNLOCK(last); 2021 } 2022 last = in6p; 2023 } 2024 INP_INFO_RUNLOCK(&V_ripcbinfo); 2025 if (last != NULL) { 2026 if (last->inp_flags & INP_CONTROLOPTS) 2027 ip6_savecontrol(last, m, &opts); 2028 /* strip intermediate headers */ 2029 m_adj(m, off); 2030 2031 /* avoid using mbuf clusters if possible (see above) */ 2032 if ((m->m_flags & M_EXT) && m->m_next == NULL && 2033 m->m_len <= MHLEN) { 2034 struct mbuf *n; 2035 2036 MGET(n, M_DONTWAIT, m->m_type); 2037 if (n != NULL) { 2038 if (m_dup_pkthdr(n, m, M_NOWAIT)) { 2039 bcopy(m->m_data, n->m_data, m->m_len); 2040 n->m_len = m->m_len; 2041 2042 m_freem(m); 2043 m = n; 2044 } else { 2045 m_freem(n); 2046 n = NULL; 2047 } 2048 } 2049 } 2050 SOCKBUF_LOCK(&last->inp_socket->so_rcv); 2051 if (sbappendaddr_locked(&last->inp_socket->so_rcv, 2052 (struct sockaddr *)&fromsa, m, opts) == 0) { 2053 m_freem(m); 2054 if (opts) 2055 m_freem(opts); 2056 SOCKBUF_UNLOCK(&last->inp_socket->so_rcv); 2057 } else 2058 sorwakeup_locked(last->inp_socket); 2059 INP_RUNLOCK(last); 2060 } else { 2061 m_freem(m); 2062 IP6STAT_DEC(ip6s_delivered); 2063 } 2064 return IPPROTO_DONE; 2065} 2066 2067/* 2068 * Reflect the ip6 packet back to the source. 2069 * OFF points to the icmp6 header, counted from the top of the mbuf. 2070 */ 2071void 2072icmp6_reflect(struct mbuf *m, size_t off) 2073{ 2074 INIT_VNET_INET6(curvnet); 2075 struct ip6_hdr *ip6; 2076 struct icmp6_hdr *icmp6; 2077 struct in6_ifaddr *ia; 2078 int plen; 2079 int type, code; 2080 struct ifnet *outif = NULL; 2081 struct in6_addr origdst, *src = NULL; 2082 2083 /* too short to reflect */ 2084 if (off < sizeof(struct ip6_hdr)) { 2085 nd6log((LOG_DEBUG, 2086 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 2087 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2088 __FILE__, __LINE__)); 2089 goto bad; 2090 } 2091 2092 /* 2093 * If there are extra headers between IPv6 and ICMPv6, strip 2094 * off that header first. 2095 */ 2096#ifdef DIAGNOSTIC 2097 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2098 panic("assumption failed in icmp6_reflect"); 2099#endif 2100 if (off > sizeof(struct ip6_hdr)) { 2101 size_t l; 2102 struct ip6_hdr nip6; 2103 2104 l = off - sizeof(struct ip6_hdr); 2105 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2106 m_adj(m, 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 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2113 } else /* off == sizeof(struct ip6_hdr) */ { 2114 size_t l; 2115 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2116 if (m->m_len < l) { 2117 if ((m = m_pullup(m, l)) == NULL) 2118 return; 2119 } 2120 } 2121 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2122 ip6 = mtod(m, struct ip6_hdr *); 2123 ip6->ip6_nxt = IPPROTO_ICMPV6; 2124 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2125 type = icmp6->icmp6_type; /* keep type for statistics */ 2126 code = icmp6->icmp6_code; /* ditto. */ 2127 2128 origdst = ip6->ip6_dst; 2129 /* 2130 * ip6_input() drops a packet if its src is multicast. 2131 * So, the src is never multicast. 2132 */ 2133 ip6->ip6_dst = ip6->ip6_src; 2134 2135 /* 2136 * If the incoming packet was addressed directly to us (i.e. unicast), 2137 * use dst as the src for the reply. 2138 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2139 * (for example) when we encounter an error while forwarding procedure 2140 * destined to a duplicated address of ours. 2141 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2142 * procedure of an outgoing packet of our own, in which case we need 2143 * to search in the ifaddr list. 2144 */ 2145 if (!IN6_IS_ADDR_MULTICAST(&origdst)) { 2146 if ((ia = ip6_getdstifaddr(m))) { 2147 if (!(ia->ia6_flags & 2148 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) 2149 src = &ia->ia_addr.sin6_addr; 2150 } else { 2151 struct sockaddr_in6 d; 2152 2153 bzero(&d, sizeof(d)); 2154 d.sin6_family = AF_INET6; 2155 d.sin6_len = sizeof(d); 2156 d.sin6_addr = origdst; 2157 ia = (struct in6_ifaddr *) 2158 ifa_ifwithaddr((struct sockaddr *)&d); 2159 if (ia && 2160 !(ia->ia6_flags & 2161 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) { 2162 src = &ia->ia_addr.sin6_addr; 2163 } 2164 } 2165 } 2166 2167 if (src == NULL) { 2168 int e; 2169 struct sockaddr_in6 sin6; 2170 struct route_in6 ro; 2171 2172 /* 2173 * This case matches to multicasts, our anycast, or unicasts 2174 * that we do not own. Select a source address based on the 2175 * source address of the erroneous packet. 2176 */ 2177 bzero(&sin6, sizeof(sin6)); 2178 sin6.sin6_family = AF_INET6; 2179 sin6.sin6_len = sizeof(sin6); 2180 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */ 2181 2182 bzero(&ro, sizeof(ro)); 2183 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2184 if (ro.ro_rt) 2185 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2186 if (src == NULL) { 2187 char ip6buf[INET6_ADDRSTRLEN]; 2188 nd6log((LOG_DEBUG, 2189 "icmp6_reflect: source can't be determined: " 2190 "dst=%s, error=%d\n", 2191 ip6_sprintf(ip6buf, &sin6.sin6_addr), e)); 2192 goto bad; 2193 } 2194 } 2195 2196 ip6->ip6_src = *src; 2197 ip6->ip6_flow = 0; 2198 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2199 ip6->ip6_vfc |= IPV6_VERSION; 2200 ip6->ip6_nxt = IPPROTO_ICMPV6; 2201 if (outif) 2202 ip6->ip6_hlim = ND_IFINFO(outif)->chlim; 2203 else if (m->m_pkthdr.rcvif) { 2204 /* XXX: This may not be the outgoing interface */ 2205 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2206 } else 2207 ip6->ip6_hlim = V_ip6_defhlim; 2208 2209 icmp6->icmp6_cksum = 0; 2210 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2211 sizeof(struct ip6_hdr), plen); 2212 2213 /* 2214 * XXX option handling 2215 */ 2216 2217 m->m_flags &= ~(M_BCAST|M_MCAST); 2218 2219 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL); 2220 if (outif) 2221 icmp6_ifoutstat_inc(outif, type, code); 2222 2223 return; 2224 2225 bad: 2226 m_freem(m); 2227 return; 2228} 2229 2230void 2231icmp6_fasttimo(void) 2232{ 2233 2234 mld_fasttimo(); 2235} 2236 2237void 2238icmp6_slowtimo(void) 2239{ 2240 2241 mld_slowtimo(); 2242} 2243 2244static const char * 2245icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2246 struct in6_addr *tgt6) 2247{ 2248 static char buf[1024]; 2249 char ip6bufs[INET6_ADDRSTRLEN]; 2250 char ip6bufd[INET6_ADDRSTRLEN]; 2251 char ip6buft[INET6_ADDRSTRLEN]; 2252 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2253 ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6), 2254 ip6_sprintf(ip6buft, tgt6)); 2255 return buf; 2256} 2257 2258void 2259icmp6_redirect_input(struct mbuf *m, int off) 2260{ 2261 INIT_VNET_INET6(curvnet); 2262 struct ifnet *ifp; 2263 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2264 struct nd_redirect *nd_rd; 2265 int icmp6len = ntohs(ip6->ip6_plen); 2266 char *lladdr = NULL; 2267 int lladdrlen = 0; 2268 u_char *redirhdr = NULL; 2269 int redirhdrlen = 0; 2270 struct rtentry *rt = NULL; 2271 int is_router; 2272 int is_onlink; 2273 struct in6_addr src6 = ip6->ip6_src; 2274 struct in6_addr redtgt6; 2275 struct in6_addr reddst6; 2276 union nd_opts ndopts; 2277 char ip6buf[INET6_ADDRSTRLEN]; 2278 2279 if (!m) 2280 return; 2281 2282 ifp = m->m_pkthdr.rcvif; 2283 2284 if (!ifp) 2285 return; 2286 2287 /* XXX if we are router, we don't update route by icmp6 redirect */ 2288 if (V_ip6_forwarding) 2289 goto freeit; 2290 if (!V_icmp6_rediraccept) 2291 goto freeit; 2292 2293#ifndef PULLDOWN_TEST 2294 IP6_EXTHDR_CHECK(m, off, icmp6len,); 2295 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off); 2296#else 2297 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2298 if (nd_rd == NULL) { 2299 ICMP6STAT_INC(icp6s_tooshort); 2300 return; 2301 } 2302#endif 2303 redtgt6 = nd_rd->nd_rd_target; 2304 reddst6 = nd_rd->nd_rd_dst; 2305 2306 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2307 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2308 goto freeit; 2309 } 2310 2311 /* validation */ 2312 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2313 nd6log((LOG_ERR, 2314 "ICMP6 redirect sent from %s rejected; " 2315 "must be from linklocal\n", 2316 ip6_sprintf(ip6buf, &src6))); 2317 goto bad; 2318 } 2319 if (ip6->ip6_hlim != 255) { 2320 nd6log((LOG_ERR, 2321 "ICMP6 redirect sent from %s rejected; " 2322 "hlim=%d (must be 255)\n", 2323 ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim)); 2324 goto bad; 2325 } 2326 { 2327 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2328 struct sockaddr_in6 sin6; 2329 struct in6_addr *gw6; 2330 2331 bzero(&sin6, sizeof(sin6)); 2332 sin6.sin6_family = AF_INET6; 2333 sin6.sin6_len = sizeof(struct sockaddr_in6); 2334 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2335 rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL); 2336 if (rt) { 2337 if (rt->rt_gateway == NULL || 2338 rt->rt_gateway->sa_family != AF_INET6) { 2339 nd6log((LOG_ERR, 2340 "ICMP6 redirect rejected; no route " 2341 "with inet6 gateway found for redirect dst: %s\n", 2342 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2343 RTFREE_LOCKED(rt); 2344 goto bad; 2345 } 2346 2347 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2348 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2349 nd6log((LOG_ERR, 2350 "ICMP6 redirect rejected; " 2351 "not equal to gw-for-src=%s (must be same): " 2352 "%s\n", 2353 ip6_sprintf(ip6buf, gw6), 2354 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2355 RTFREE_LOCKED(rt); 2356 goto bad; 2357 } 2358 } else { 2359 nd6log((LOG_ERR, 2360 "ICMP6 redirect rejected; " 2361 "no route found for redirect dst: %s\n", 2362 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2363 goto bad; 2364 } 2365 RTFREE_LOCKED(rt); 2366 rt = NULL; 2367 } 2368 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2369 nd6log((LOG_ERR, 2370 "ICMP6 redirect rejected; " 2371 "redirect dst must be unicast: %s\n", 2372 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2373 goto bad; 2374 } 2375 2376 is_router = is_onlink = 0; 2377 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2378 is_router = 1; /* router case */ 2379 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2380 is_onlink = 1; /* on-link destination case */ 2381 if (!is_router && !is_onlink) { 2382 nd6log((LOG_ERR, 2383 "ICMP6 redirect rejected; " 2384 "neither router case nor onlink case: %s\n", 2385 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2386 goto bad; 2387 } 2388 /* validation passed */ 2389 2390 icmp6len -= sizeof(*nd_rd); 2391 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2392 if (nd6_options(&ndopts) < 0) { 2393 nd6log((LOG_INFO, "icmp6_redirect_input: " 2394 "invalid ND option, rejected: %s\n", 2395 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2396 /* nd6_options have incremented stats */ 2397 goto freeit; 2398 } 2399 2400 if (ndopts.nd_opts_tgt_lladdr) { 2401 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2402 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2403 } 2404 2405 if (ndopts.nd_opts_rh) { 2406 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len; 2407 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */ 2408 } 2409 2410 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2411 nd6log((LOG_INFO, 2412 "icmp6_redirect_input: lladdrlen mismatch for %s " 2413 "(if %d, icmp6 packet %d): %s\n", 2414 ip6_sprintf(ip6buf, &redtgt6), 2415 ifp->if_addrlen, lladdrlen - 2, 2416 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2417 goto bad; 2418 } 2419 2420 /* RFC 2461 8.3 */ 2421 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2422 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2423 2424 if (!is_onlink) { /* better router case. perform rtredirect. */ 2425 /* perform rtredirect */ 2426 struct sockaddr_in6 sdst; 2427 struct sockaddr_in6 sgw; 2428 struct sockaddr_in6 ssrc; 2429 2430 bzero(&sdst, sizeof(sdst)); 2431 bzero(&sgw, sizeof(sgw)); 2432 bzero(&ssrc, sizeof(ssrc)); 2433 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2434 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2435 sizeof(struct sockaddr_in6); 2436 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2437 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2438 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2439 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2440 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2441 (struct sockaddr *)&ssrc); 2442 } 2443 /* finally update cached route in each socket via pfctlinput */ 2444 { 2445 struct sockaddr_in6 sdst; 2446 2447 bzero(&sdst, sizeof(sdst)); 2448 sdst.sin6_family = AF_INET6; 2449 sdst.sin6_len = sizeof(struct sockaddr_in6); 2450 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2451 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2452#ifdef IPSEC 2453 key_sa_routechange((struct sockaddr *)&sdst); 2454#endif /* IPSEC */ 2455 } 2456 2457 freeit: 2458 m_freem(m); 2459 return; 2460 2461 bad: 2462 ICMP6STAT_INC(icp6s_badredirect); 2463 m_freem(m); 2464} 2465 2466void 2467icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2468{ 2469 INIT_VNET_INET6(curvnet); 2470 struct ifnet *ifp; /* my outgoing interface */ 2471 struct in6_addr *ifp_ll6; 2472 struct in6_addr *router_ll6; 2473 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2474 struct mbuf *m = NULL; /* newly allocated one */ 2475 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2476 struct nd_redirect *nd_rd; 2477 struct llentry *ln = NULL; 2478 size_t maxlen; 2479 u_char *p; 2480 struct ifnet *outif = NULL; 2481 struct sockaddr_in6 src_sa; 2482 2483 icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2484 2485 /* if we are not router, we don't send icmp6 redirect */ 2486 if (!V_ip6_forwarding) 2487 goto fail; 2488 2489 /* sanity check */ 2490 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2491 goto fail; 2492 2493 /* 2494 * Address check: 2495 * the source address must identify a neighbor, and 2496 * the destination address must not be a multicast address 2497 * [RFC 2461, sec 8.2] 2498 */ 2499 sip6 = mtod(m0, struct ip6_hdr *); 2500 bzero(&src_sa, sizeof(src_sa)); 2501 src_sa.sin6_family = AF_INET6; 2502 src_sa.sin6_len = sizeof(src_sa); 2503 src_sa.sin6_addr = sip6->ip6_src; 2504 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2505 goto fail; 2506 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2507 goto fail; /* what should we do here? */ 2508 2509 /* rate limit */ 2510 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2511 goto fail; 2512 2513 /* 2514 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2515 * we almost always ask for an mbuf cluster for simplicity. 2516 * (MHLEN < IPV6_MMTU is almost always true) 2517 */ 2518#if IPV6_MMTU >= MCLBYTES 2519# error assumption failed about IPV6_MMTU and MCLBYTES 2520#endif 2521 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2522 if (m && IPV6_MMTU >= MHLEN) 2523 MCLGET(m, M_DONTWAIT); 2524 if (!m) 2525 goto fail; 2526 m->m_pkthdr.rcvif = NULL; 2527 m->m_len = 0; 2528 maxlen = M_TRAILINGSPACE(m); 2529 maxlen = min(IPV6_MMTU, maxlen); 2530 /* just for safety */ 2531 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2532 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2533 goto fail; 2534 } 2535 2536 { 2537 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2538 struct in6_ifaddr *ia; 2539 if ((ia = in6ifa_ifpforlinklocal(ifp, 2540 IN6_IFF_NOTREADY| 2541 IN6_IFF_ANYCAST)) == NULL) 2542 goto fail; 2543 ifp_ll6 = &ia->ia_addr.sin6_addr; 2544 } 2545 2546 /* get ip6 linklocal address for the router. */ 2547 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2548 struct sockaddr_in6 *sin6; 2549 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2550 router_ll6 = &sin6->sin6_addr; 2551 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6)) 2552 router_ll6 = (struct in6_addr *)NULL; 2553 } else 2554 router_ll6 = (struct in6_addr *)NULL; 2555 2556 /* ip6 */ 2557 ip6 = mtod(m, struct ip6_hdr *); 2558 ip6->ip6_flow = 0; 2559 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2560 ip6->ip6_vfc |= IPV6_VERSION; 2561 /* ip6->ip6_plen will be set later */ 2562 ip6->ip6_nxt = IPPROTO_ICMPV6; 2563 ip6->ip6_hlim = 255; 2564 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2565 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2566 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2567 2568 /* ND Redirect */ 2569 nd_rd = (struct nd_redirect *)(ip6 + 1); 2570 nd_rd->nd_rd_type = ND_REDIRECT; 2571 nd_rd->nd_rd_code = 0; 2572 nd_rd->nd_rd_reserved = 0; 2573 if (rt->rt_flags & RTF_GATEWAY) { 2574 /* 2575 * nd_rd->nd_rd_target must be a link-local address in 2576 * better router cases. 2577 */ 2578 if (!router_ll6) 2579 goto fail; 2580 bcopy(router_ll6, &nd_rd->nd_rd_target, 2581 sizeof(nd_rd->nd_rd_target)); 2582 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2583 sizeof(nd_rd->nd_rd_dst)); 2584 } else { 2585 /* make sure redtgt == reddst */ 2586 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2587 sizeof(nd_rd->nd_rd_target)); 2588 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2589 sizeof(nd_rd->nd_rd_dst)); 2590 } 2591 2592 p = (u_char *)(nd_rd + 1); 2593 2594 if (!router_ll6) 2595 goto nolladdropt; 2596 2597 { 2598 /* target lladdr option */ 2599 int len; 2600 struct nd_opt_hdr *nd_opt; 2601 char *lladdr; 2602 2603 IF_AFDATA_LOCK(ifp); 2604 ln = nd6_lookup(router_ll6, 0, ifp); 2605 IF_AFDATA_UNLOCK(ifp); 2606 if (ln == NULL) 2607 goto nolladdropt; 2608 2609 len = sizeof(*nd_opt) + ifp->if_addrlen; 2610 len = (len + 7) & ~7; /* round by 8 */ 2611 /* safety check */ 2612 if (len + (p - (u_char *)ip6) > maxlen) 2613 goto nolladdropt; 2614 2615 if (ln->la_flags & LLE_VALID) { 2616 nd_opt = (struct nd_opt_hdr *)p; 2617 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2618 nd_opt->nd_opt_len = len >> 3; 2619 lladdr = (char *)(nd_opt + 1); 2620 bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen); 2621 p += len; 2622 } 2623 } 2624nolladdropt: 2625 if (ln != NULL) 2626 LLE_RUNLOCK(ln); 2627 2628 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2629 2630 /* just to be safe */ 2631#ifdef M_DECRYPTED /*not openbsd*/ 2632 if (m0->m_flags & M_DECRYPTED) 2633 goto noredhdropt; 2634#endif 2635 if (p - (u_char *)ip6 > maxlen) 2636 goto noredhdropt; 2637 2638 { 2639 /* redirected header option */ 2640 int len; 2641 struct nd_opt_rd_hdr *nd_opt_rh; 2642 2643 /* 2644 * compute the maximum size for icmp6 redirect header option. 2645 * XXX room for auth header? 2646 */ 2647 len = maxlen - (p - (u_char *)ip6); 2648 len &= ~7; 2649 2650 /* This is just for simplicity. */ 2651 if (m0->m_pkthdr.len != m0->m_len) { 2652 if (m0->m_next) { 2653 m_freem(m0->m_next); 2654 m0->m_next = NULL; 2655 } 2656 m0->m_pkthdr.len = m0->m_len; 2657 } 2658 2659 /* 2660 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2661 * about padding/truncate rule for the original IP packet. 2662 * From the discussion on IPv6imp in Feb 1999, 2663 * the consensus was: 2664 * - "attach as much as possible" is the goal 2665 * - pad if not aligned (original size can be guessed by 2666 * original ip6 header) 2667 * Following code adds the padding if it is simple enough, 2668 * and truncates if not. 2669 */ 2670 if (m0->m_next || m0->m_pkthdr.len != m0->m_len) 2671 panic("assumption failed in %s:%d", __FILE__, 2672 __LINE__); 2673 2674 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2675 /* not enough room, truncate */ 2676 m0->m_pkthdr.len = m0->m_len = len - 2677 sizeof(*nd_opt_rh); 2678 } else { 2679 /* enough room, pad or truncate */ 2680 size_t extra; 2681 2682 extra = m0->m_pkthdr.len % 8; 2683 if (extra) { 2684 /* pad if easy enough, truncate if not */ 2685 if (8 - extra <= M_TRAILINGSPACE(m0)) { 2686 /* pad */ 2687 m0->m_len += (8 - extra); 2688 m0->m_pkthdr.len += (8 - extra); 2689 } else { 2690 /* truncate */ 2691 m0->m_pkthdr.len -= extra; 2692 m0->m_len -= extra; 2693 } 2694 } 2695 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2696 m0->m_pkthdr.len = m0->m_len = len - 2697 sizeof(*nd_opt_rh); 2698 } 2699 2700 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2701 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2702 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2703 nd_opt_rh->nd_opt_rh_len = len >> 3; 2704 p += sizeof(*nd_opt_rh); 2705 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2706 2707 /* connect m0 to m */ 2708 m_tag_delete_chain(m0, NULL); 2709 m0->m_flags &= ~M_PKTHDR; 2710 m->m_next = m0; 2711 m->m_pkthdr.len = m->m_len + m0->m_len; 2712 m0 = NULL; 2713 } 2714noredhdropt:; 2715 if (m0) { 2716 m_freem(m0); 2717 m0 = NULL; 2718 } 2719 2720 /* XXX: clear embedded link IDs in the inner header */ 2721 in6_clearscope(&sip6->ip6_src); 2722 in6_clearscope(&sip6->ip6_dst); 2723 in6_clearscope(&nd_rd->nd_rd_target); 2724 in6_clearscope(&nd_rd->nd_rd_dst); 2725 2726 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2727 2728 nd_rd->nd_rd_cksum = 0; 2729 nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2730 sizeof(*ip6), ntohs(ip6->ip6_plen)); 2731 2732 /* send the packet to outside... */ 2733 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL); 2734 if (outif) { 2735 icmp6_ifstat_inc(outif, ifs6_out_msg); 2736 icmp6_ifstat_inc(outif, ifs6_out_redirect); 2737 } 2738 ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]); 2739 2740 return; 2741 2742fail: 2743 if (m) 2744 m_freem(m); 2745 if (m0) 2746 m_freem(m0); 2747} 2748 2749/* 2750 * ICMPv6 socket option processing. 2751 */ 2752int 2753icmp6_ctloutput(struct socket *so, struct sockopt *sopt) 2754{ 2755 int error = 0; 2756 int optlen; 2757 struct inpcb *inp = sotoinpcb(so); 2758 int level, op, optname; 2759 2760 if (sopt) { 2761 level = sopt->sopt_level; 2762 op = sopt->sopt_dir; 2763 optname = sopt->sopt_name; 2764 optlen = sopt->sopt_valsize; 2765 } else 2766 level = op = optname = optlen = 0; 2767 2768 if (level != IPPROTO_ICMPV6) { 2769 return EINVAL; 2770 } 2771 2772 switch (op) { 2773 case PRCO_SETOPT: 2774 switch (optname) { 2775 case ICMP6_FILTER: 2776 { 2777 struct icmp6_filter ic6f; 2778 2779 if (optlen != sizeof(ic6f)) { 2780 error = EMSGSIZE; 2781 break; 2782 } 2783 error = sooptcopyin(sopt, &ic6f, optlen, optlen); 2784 if (error == 0) { 2785 INP_WLOCK(inp); 2786 *inp->in6p_icmp6filt = ic6f; 2787 INP_WUNLOCK(inp); 2788 } 2789 break; 2790 } 2791 2792 default: 2793 error = ENOPROTOOPT; 2794 break; 2795 } 2796 break; 2797 2798 case PRCO_GETOPT: 2799 switch (optname) { 2800 case ICMP6_FILTER: 2801 { 2802 struct icmp6_filter ic6f; 2803 2804 INP_RLOCK(inp); 2805 ic6f = *inp->in6p_icmp6filt; 2806 INP_RUNLOCK(inp); 2807 error = sooptcopyout(sopt, &ic6f, sizeof(ic6f)); 2808 break; 2809 } 2810 2811 default: 2812 error = ENOPROTOOPT; 2813 break; 2814 } 2815 break; 2816 } 2817 2818 return (error); 2819} 2820 2821/* 2822 * Perform rate limit check. 2823 * Returns 0 if it is okay to send the icmp6 packet. 2824 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2825 * limitation. 2826 * 2827 * XXX per-destination/type check necessary? 2828 * 2829 * dst - not used at this moment 2830 * type - not used at this moment 2831 * code - not used at this moment 2832 */ 2833static int 2834icmp6_ratelimit(const struct in6_addr *dst, const int type, 2835 const int code) 2836{ 2837 INIT_VNET_INET6(curvnet); 2838 int ret; 2839 2840 ret = 0; /* okay to send */ 2841 2842 /* PPS limit */ 2843 if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count, 2844 V_icmp6errppslim)) { 2845 /* The packet is subject to rate limit */ 2846 ret++; 2847 } 2848 2849 return ret; 2850} 2851