icmp6.c revision 1.120
1 /* $NetBSD: icmp6.c,v 1.120 2006/09/01 02:44:46 dyoung Exp $ */ 2 /* $KAME: icmp6.c,v 1.217 2001/06/20 15:03:29 jinmei Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 62 */ 63 64 #include <sys/cdefs.h> 65 __KERNEL_RCSID(0, "$NetBSD: icmp6.c,v 1.120 2006/09/01 02:44:46 dyoung Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_ipsec.h" 69 70 #include <sys/param.h> 71 #include <sys/systm.h> 72 #include <sys/malloc.h> 73 #include <sys/mbuf.h> 74 #include <sys/protosw.h> 75 #include <sys/socket.h> 76 #include <sys/socketvar.h> 77 #include <sys/time.h> 78 #include <sys/kernel.h> 79 #include <sys/syslog.h> 80 #include <sys/domain.h> 81 #include <sys/sysctl.h> 82 83 #include <net/if.h> 84 #include <net/route.h> 85 #include <net/if_dl.h> 86 #include <net/if_types.h> 87 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip6.h> 91 #include <netinet6/ip6_var.h> 92 #include <netinet/icmp6.h> 93 #include <netinet6/mld6_var.h> 94 #include <netinet6/in6_pcb.h> 95 #include <netinet6/nd6.h> 96 #include <netinet6/in6_ifattach.h> 97 #include <netinet6/ip6protosw.h> 98 #include <netinet6/scope6_var.h> 99 100 #ifdef IPSEC 101 #include <netinet6/ipsec.h> 102 #include <netkey/key.h> 103 #endif 104 105 #include "faith.h" 106 #if defined(NFAITH) && 0 < NFAITH 107 #include <net/if_faith.h> 108 #endif 109 110 #include <net/net_osdep.h> 111 112 extern struct domain inet6domain; 113 114 struct icmp6stat icmp6stat; 115 116 extern struct inpcbtable raw6cbtable; 117 extern int icmp6errppslim; 118 static int icmp6errpps_count = 0; 119 static struct timeval icmp6errppslim_last; 120 extern int icmp6_nodeinfo; 121 122 /* 123 * List of callbacks to notify when Path MTU changes are made. 124 */ 125 struct icmp6_mtudisc_callback { 126 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 127 void (*mc_func) __P((struct in6_addr *)); 128 }; 129 130 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 131 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 132 133 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 134 extern int pmtu_expire; 135 136 /* XXX do these values make any sense? */ 137 static int icmp6_mtudisc_hiwat = 1280; 138 static int icmp6_mtudisc_lowat = 256; 139 140 /* 141 * keep track of # of redirect routes. 142 */ 143 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 144 145 /* XXX experimental, turned off */ 146 static int icmp6_redirect_hiwat = -1; 147 static int icmp6_redirect_lowat = -1; 148 149 static void icmp6_errcount __P((struct icmp6errstat *, int, int)); 150 static int icmp6_rip6_input __P((struct mbuf **, int)); 151 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int)); 152 static const char *icmp6_redirect_diag __P((struct in6_addr *, 153 struct in6_addr *, struct in6_addr *)); 154 static struct mbuf *ni6_input __P((struct mbuf *, int)); 155 static struct mbuf *ni6_nametodns __P((const char *, int, int)); 156 static int ni6_dnsmatch __P((const char *, int, const char *, int)); 157 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *, 158 struct ifnet **, char *)); 159 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 160 struct ifnet *, int)); 161 static int icmp6_notify_error __P((struct mbuf *, int, int, int)); 162 static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *)); 163 static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *)); 164 static void icmp6_redirect_timeout __P((struct rtentry *, struct rttimer *)); 165 166 167 void 168 icmp6_init() 169 { 170 mld_init(); 171 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 172 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 173 } 174 175 static void 176 icmp6_errcount(stat, type, code) 177 struct icmp6errstat *stat; 178 int type, code; 179 { 180 switch (type) { 181 case ICMP6_DST_UNREACH: 182 switch (code) { 183 case ICMP6_DST_UNREACH_NOROUTE: 184 stat->icp6errs_dst_unreach_noroute++; 185 return; 186 case ICMP6_DST_UNREACH_ADMIN: 187 stat->icp6errs_dst_unreach_admin++; 188 return; 189 case ICMP6_DST_UNREACH_BEYONDSCOPE: 190 stat->icp6errs_dst_unreach_beyondscope++; 191 return; 192 case ICMP6_DST_UNREACH_ADDR: 193 stat->icp6errs_dst_unreach_addr++; 194 return; 195 case ICMP6_DST_UNREACH_NOPORT: 196 stat->icp6errs_dst_unreach_noport++; 197 return; 198 } 199 break; 200 case ICMP6_PACKET_TOO_BIG: 201 stat->icp6errs_packet_too_big++; 202 return; 203 case ICMP6_TIME_EXCEEDED: 204 switch (code) { 205 case ICMP6_TIME_EXCEED_TRANSIT: 206 stat->icp6errs_time_exceed_transit++; 207 return; 208 case ICMP6_TIME_EXCEED_REASSEMBLY: 209 stat->icp6errs_time_exceed_reassembly++; 210 return; 211 } 212 break; 213 case ICMP6_PARAM_PROB: 214 switch (code) { 215 case ICMP6_PARAMPROB_HEADER: 216 stat->icp6errs_paramprob_header++; 217 return; 218 case ICMP6_PARAMPROB_NEXTHEADER: 219 stat->icp6errs_paramprob_nextheader++; 220 return; 221 case ICMP6_PARAMPROB_OPTION: 222 stat->icp6errs_paramprob_option++; 223 return; 224 } 225 break; 226 case ND_REDIRECT: 227 stat->icp6errs_redirect++; 228 return; 229 } 230 stat->icp6errs_unknown++; 231 } 232 233 /* 234 * Register a Path MTU Discovery callback. 235 */ 236 void 237 icmp6_mtudisc_callback_register(func) 238 void (*func) __P((struct in6_addr *)); 239 { 240 struct icmp6_mtudisc_callback *mc; 241 242 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 243 mc = LIST_NEXT(mc, mc_list)) { 244 if (mc->mc_func == func) 245 return; 246 } 247 248 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 249 if (mc == NULL) 250 panic("icmp6_mtudisc_callback_register"); 251 252 mc->mc_func = func; 253 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 254 } 255 256 /* 257 * A wrapper function for icmp6_error() necessary when the erroneous packet 258 * may not contain enough scope zone information. 259 */ 260 void 261 icmp6_error2(m, type, code, param, ifp) 262 struct mbuf *m; 263 int type, code, param; 264 struct ifnet *ifp; 265 { 266 struct ip6_hdr *ip6; 267 268 if (ifp == NULL) 269 return; 270 271 if (m->m_len < sizeof(struct ip6_hdr)) { 272 m = m_pullup(m, sizeof(struct ip6_hdr)); 273 if (m == NULL) 274 return; 275 } 276 277 ip6 = mtod(m, struct ip6_hdr *); 278 279 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 280 return; 281 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 282 return; 283 284 icmp6_error(m, type, code, param); 285 } 286 287 /* 288 * Generate an error packet of type error in response to bad IP6 packet. 289 */ 290 void 291 icmp6_error(m, type, code, param) 292 struct mbuf *m; 293 int type, code, param; 294 { 295 struct ip6_hdr *oip6, *nip6; 296 struct icmp6_hdr *icmp6; 297 u_int preplen; 298 int off; 299 int nxt; 300 301 icmp6stat.icp6s_error++; 302 303 /* count per-type-code statistics */ 304 icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code); 305 306 if (m->m_flags & M_DECRYPTED) { 307 icmp6stat.icp6s_canterror++; 308 goto freeit; 309 } 310 311 if (m->m_len < sizeof(struct ip6_hdr)) { 312 m = m_pullup(m, sizeof(struct ip6_hdr)); 313 if (m == NULL) 314 return; 315 } 316 oip6 = mtod(m, struct ip6_hdr *); 317 318 /* 319 * If the destination address of the erroneous packet is a multicast 320 * address, or the packet was sent using link-layer multicast, 321 * we should basically suppress sending an error (RFC 2463, Section 322 * 2.4). 323 * We have two exceptions (the item e.2 in that section): 324 * - the Pakcet Too Big message can be sent for path MTU discovery. 325 * - the Parameter Problem Message that can be allowed an icmp6 error 326 * in the option type field. This check has been done in 327 * ip6_unknown_opt(), so we can just check the type and code. 328 */ 329 if ((m->m_flags & (M_BCAST|M_MCAST) || 330 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 331 (type != ICMP6_PACKET_TOO_BIG && 332 (type != ICMP6_PARAM_PROB || 333 code != ICMP6_PARAMPROB_OPTION))) 334 goto freeit; 335 336 /* 337 * RFC 2463, 2.4 (e.5): source address check. 338 * XXX: the case of anycast source? 339 */ 340 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 341 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 342 goto freeit; 343 344 /* 345 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 346 * don't do it. 347 */ 348 nxt = -1; 349 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 350 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 351 struct icmp6_hdr *icp; 352 353 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 354 sizeof(*icp)); 355 if (icp == NULL) { 356 icmp6stat.icp6s_tooshort++; 357 return; 358 } 359 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 360 icp->icmp6_type == ND_REDIRECT) { 361 /* 362 * ICMPv6 error 363 * Special case: for redirect (which is 364 * informational) we must not send icmp6 error. 365 */ 366 icmp6stat.icp6s_canterror++; 367 goto freeit; 368 } else { 369 /* ICMPv6 informational - send the error */ 370 } 371 } 372 #if 0 /* controversial */ 373 else if (off >= 0 && nxt == IPPROTO_ESP) { 374 /* 375 * It could be ICMPv6 error inside ESP. Take a safer side, 376 * don't respond. 377 */ 378 icmp6stat.icp6s_canterror++; 379 goto freeit; 380 } 381 #endif 382 else { 383 /* non-ICMPv6 - send the error */ 384 } 385 386 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 387 388 /* Finally, do rate limitation check. */ 389 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 390 icmp6stat.icp6s_toofreq++; 391 goto freeit; 392 } 393 394 /* 395 * OK, ICMP6 can be generated. 396 */ 397 398 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 399 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 400 401 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 402 M_PREPEND(m, preplen, M_DONTWAIT); 403 if (m && m->m_len < preplen) 404 m = m_pullup(m, preplen); 405 if (m == NULL) { 406 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 407 return; 408 } 409 410 nip6 = mtod(m, struct ip6_hdr *); 411 nip6->ip6_src = oip6->ip6_src; 412 nip6->ip6_dst = oip6->ip6_dst; 413 414 in6_clearscope(&oip6->ip6_src); 415 in6_clearscope(&oip6->ip6_dst); 416 417 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 418 icmp6->icmp6_type = type; 419 icmp6->icmp6_code = code; 420 icmp6->icmp6_pptr = htonl((u_int32_t)param); 421 422 /* 423 * icmp6_reflect() is designed to be in the input path. 424 * icmp6_error() can be called from both input and output path, 425 * and if we are in output path rcvif could contain bogus value. 426 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 427 * information in ip header (nip6). 428 */ 429 m->m_pkthdr.rcvif = NULL; 430 431 icmp6stat.icp6s_outhist[type]++; 432 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 433 434 return; 435 436 freeit: 437 /* 438 * If we can't tell whether or not we can generate ICMP6, free it. 439 */ 440 m_freem(m); 441 } 442 443 /* 444 * Process a received ICMP6 message. 445 */ 446 int 447 icmp6_input(mp, offp, proto) 448 struct mbuf **mp; 449 int *offp, proto; 450 { 451 struct mbuf *m = *mp, *n; 452 struct ip6_hdr *ip6, *nip6; 453 struct icmp6_hdr *icmp6, *nicmp6; 454 int off = *offp; 455 int icmp6len = m->m_pkthdr.len - *offp; 456 int code, sum, noff; 457 458 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4) 459 KASSERT(ICMP6_MAXLEN < MCLBYTES); 460 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 461 462 /* 463 * Locate icmp6 structure in mbuf, and check 464 * that not corrupted and of at least minimum length 465 */ 466 467 ip6 = mtod(m, struct ip6_hdr *); 468 if (icmp6len < sizeof(struct icmp6_hdr)) { 469 icmp6stat.icp6s_tooshort++; 470 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 471 goto freeit; 472 } 473 474 /* 475 * calculate the checksum 476 */ 477 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 478 if (icmp6 == NULL) { 479 icmp6stat.icp6s_tooshort++; 480 /* m is invalid */ 481 /*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/ 482 return IPPROTO_DONE; 483 } 484 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 485 code = icmp6->icmp6_code; 486 487 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 488 nd6log((LOG_ERR, 489 "ICMP6 checksum error(%d|%x) %s\n", 490 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 491 icmp6stat.icp6s_checksum++; 492 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 493 goto freeit; 494 } 495 496 #if defined(NFAITH) && 0 < NFAITH 497 if (faithprefix(&ip6->ip6_dst)) { 498 /* 499 * Deliver very specific ICMP6 type only. 500 * This is important to deliver TOOBIG. Otherwise PMTUD 501 * will not work. 502 */ 503 switch (icmp6->icmp6_type) { 504 case ICMP6_DST_UNREACH: 505 case ICMP6_PACKET_TOO_BIG: 506 case ICMP6_TIME_EXCEEDED: 507 break; 508 default: 509 goto freeit; 510 } 511 } 512 #endif 513 514 icmp6stat.icp6s_inhist[icmp6->icmp6_type]++; 515 516 switch (icmp6->icmp6_type) { 517 case ICMP6_DST_UNREACH: 518 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 519 switch (code) { 520 case ICMP6_DST_UNREACH_NOROUTE: 521 code = PRC_UNREACH_NET; 522 break; 523 case ICMP6_DST_UNREACH_ADMIN: 524 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 525 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 526 break; 527 case ICMP6_DST_UNREACH_ADDR: 528 code = PRC_HOSTDEAD; 529 break; 530 #ifdef COMPAT_RFC1885 531 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 532 code = PRC_UNREACH_SRCFAIL; 533 break; 534 #else 535 case ICMP6_DST_UNREACH_BEYONDSCOPE: 536 /* I mean "source address was incorrect." */ 537 code = PRC_UNREACH_NET; 538 break; 539 #endif 540 case ICMP6_DST_UNREACH_NOPORT: 541 code = PRC_UNREACH_PORT; 542 break; 543 default: 544 goto badcode; 545 } 546 goto deliver; 547 548 case ICMP6_PACKET_TOO_BIG: 549 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 550 551 code = PRC_MSGSIZE; 552 553 /* 554 * Updating the path MTU will be done after examining 555 * intermediate extension headers. 556 */ 557 goto deliver; 558 559 case ICMP6_TIME_EXCEEDED: 560 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 561 switch (code) { 562 case ICMP6_TIME_EXCEED_TRANSIT: 563 code = PRC_TIMXCEED_INTRANS; 564 break; 565 case ICMP6_TIME_EXCEED_REASSEMBLY: 566 code = PRC_TIMXCEED_REASS; 567 break; 568 default: 569 goto badcode; 570 } 571 goto deliver; 572 573 case ICMP6_PARAM_PROB: 574 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 575 switch (code) { 576 case ICMP6_PARAMPROB_NEXTHEADER: 577 code = PRC_UNREACH_PROTOCOL; 578 break; 579 case ICMP6_PARAMPROB_HEADER: 580 case ICMP6_PARAMPROB_OPTION: 581 code = PRC_PARAMPROB; 582 break; 583 default: 584 goto badcode; 585 } 586 goto deliver; 587 588 case ICMP6_ECHO_REQUEST: 589 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 590 if (code != 0) 591 goto badcode; 592 /* 593 * Copy mbuf to send to two data paths: userland socket(s), 594 * and to the querier (echo reply). 595 * m: a copy for socket, n: a copy for querier 596 * 597 * If the first mbuf is shared, or the first mbuf is too short, 598 * copy the first part of the data into a fresh mbuf. 599 * Otherwise, we will wrongly overwrite both copies. 600 */ 601 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 602 /* Give up local */ 603 n = m; 604 m = NULL; 605 } else if (M_READONLY(n) || 606 n->m_len < off + sizeof(struct icmp6_hdr)) { 607 struct mbuf *n0 = n; 608 609 /* 610 * Prepare an internal mbuf. m_pullup() doesn't 611 * always copy the length we specified. 612 */ 613 MGETHDR(n, M_DONTWAIT, n0->m_type); 614 if (n && ICMP6_MAXLEN >= MHLEN) { 615 MCLGET(n, M_DONTWAIT); 616 if ((n->m_flags & M_EXT) == 0) { 617 m_free(n); 618 n = NULL; 619 } 620 } 621 if (n == NULL) { 622 /* Give up local */ 623 n = m; 624 m = NULL; 625 } else { 626 M_MOVE_PKTHDR(n, n0); 627 m_copydata(n0, 0, M_COPYALL, mtod(n, caddr_t)); 628 n->m_len = n->m_pkthdr.len = n0->m_pkthdr.len; 629 } 630 m_freem(n0); 631 } 632 nip6 = mtod(n, struct ip6_hdr *); 633 nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off); 634 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 635 nicmp6->icmp6_code = 0; 636 if (n) { 637 icmp6stat.icp6s_reflect++; 638 icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++; 639 icmp6_reflect(n, off); 640 } 641 if (!m) 642 goto freeit; 643 break; 644 645 case ICMP6_ECHO_REPLY: 646 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 647 if (code != 0) 648 goto badcode; 649 break; 650 651 case MLD_LISTENER_QUERY: 652 case MLD_LISTENER_REPORT: 653 if (icmp6len < sizeof(struct mld_hdr)) 654 goto badlen; 655 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */ 656 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 657 else 658 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 659 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 660 /* give up local */ 661 mld_input(m, off); 662 m = NULL; 663 goto freeit; 664 } 665 mld_input(n, off); 666 /* m stays. */ 667 break; 668 669 case MLD_LISTENER_DONE: 670 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 671 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */ 672 goto badlen; 673 break; /* nothing to be done in kernel */ 674 675 case MLD_MTRACE_RESP: 676 case MLD_MTRACE: 677 /* XXX: these two are experimental. not officially defined. */ 678 /* XXX: per-interface statistics? */ 679 break; /* just pass it to applications */ 680 681 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 682 { 683 enum { WRU, FQDN } mode; 684 685 if (!icmp6_nodeinfo) 686 break; 687 688 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 689 mode = WRU; 690 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 691 mode = FQDN; 692 else 693 goto badlen; 694 695 if (mode == FQDN) { 696 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 697 if (n) 698 n = ni6_input(n, off); 699 /* XXX meaningless if n == NULL */ 700 noff = sizeof(struct ip6_hdr); 701 } else { 702 u_char *p; 703 int maxhlen; 704 705 if ((icmp6_nodeinfo & 5) != 5) 706 break; 707 708 if (code != 0) 709 goto badcode; 710 MGETHDR(n, M_DONTWAIT, m->m_type); 711 if (n && ICMP6_MAXLEN > MHLEN) { 712 MCLGET(n, M_DONTWAIT); 713 if ((n->m_flags & M_EXT) == 0) { 714 m_free(n); 715 n = NULL; 716 } 717 } 718 if (n == NULL) { 719 /* Give up remote */ 720 break; 721 } 722 n->m_pkthdr.rcvif = NULL; 723 n->m_len = 0; 724 maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN; 725 if (maxhlen > hostnamelen) 726 maxhlen = hostnamelen; 727 /* 728 * Copy IPv6 and ICMPv6 only. 729 */ 730 nip6 = mtod(n, struct ip6_hdr *); 731 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 732 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 733 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 734 p = (u_char *)(nicmp6 + 1); 735 bzero(p, 4); 736 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 737 noff = sizeof(struct ip6_hdr); 738 M_COPY_PKTHDR(n, m); /* just for rcvif */ 739 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 740 sizeof(struct icmp6_hdr) + 4 + maxhlen; 741 nicmp6->icmp6_type = ICMP6_WRUREPLY; 742 nicmp6->icmp6_code = 0; 743 } 744 #undef hostnamelen 745 if (n) { 746 icmp6stat.icp6s_reflect++; 747 icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++; 748 icmp6_reflect(n, noff); 749 } 750 break; 751 } 752 753 case ICMP6_WRUREPLY: 754 if (code != 0) 755 goto badcode; 756 break; 757 758 case ND_ROUTER_SOLICIT: 759 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 760 if (code != 0) 761 goto badcode; 762 if (icmp6len < sizeof(struct nd_router_solicit)) 763 goto badlen; 764 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 765 /* give up local */ 766 nd6_rs_input(m, off, icmp6len); 767 m = NULL; 768 goto freeit; 769 } 770 nd6_rs_input(n, off, icmp6len); 771 /* m stays. */ 772 break; 773 774 case ND_ROUTER_ADVERT: 775 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 776 if (code != 0) 777 goto badcode; 778 if (icmp6len < sizeof(struct nd_router_advert)) 779 goto badlen; 780 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 781 /* give up local */ 782 nd6_ra_input(m, off, icmp6len); 783 m = NULL; 784 goto freeit; 785 } 786 nd6_ra_input(n, off, icmp6len); 787 /* m stays. */ 788 break; 789 790 case ND_NEIGHBOR_SOLICIT: 791 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 792 if (code != 0) 793 goto badcode; 794 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 795 goto badlen; 796 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 797 /* give up local */ 798 nd6_ns_input(m, off, icmp6len); 799 m = NULL; 800 goto freeit; 801 } 802 nd6_ns_input(n, off, icmp6len); 803 /* m stays. */ 804 break; 805 806 case ND_NEIGHBOR_ADVERT: 807 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 808 if (code != 0) 809 goto badcode; 810 if (icmp6len < sizeof(struct nd_neighbor_advert)) 811 goto badlen; 812 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 813 /* give up local */ 814 nd6_na_input(m, off, icmp6len); 815 m = NULL; 816 goto freeit; 817 } 818 nd6_na_input(n, off, icmp6len); 819 /* m stays. */ 820 break; 821 822 case ND_REDIRECT: 823 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 824 if (code != 0) 825 goto badcode; 826 if (icmp6len < sizeof(struct nd_redirect)) 827 goto badlen; 828 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 829 /* give up local */ 830 icmp6_redirect_input(m, off); 831 m = NULL; 832 goto freeit; 833 } 834 icmp6_redirect_input(n, off); 835 /* m stays. */ 836 break; 837 838 case ICMP6_ROUTER_RENUMBERING: 839 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 840 code != ICMP6_ROUTER_RENUMBERING_RESULT) 841 goto badcode; 842 if (icmp6len < sizeof(struct icmp6_router_renum)) 843 goto badlen; 844 break; 845 846 default: 847 nd6log((LOG_DEBUG, 848 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 849 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 850 ip6_sprintf(&ip6->ip6_dst), 851 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 852 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 853 /* ICMPv6 error: MUST deliver it by spec... */ 854 code = PRC_NCMDS; 855 /* deliver */ 856 } else { 857 /* ICMPv6 informational: MUST not deliver */ 858 break; 859 } 860 deliver: 861 if (icmp6_notify_error(m, off, icmp6len, code)) { 862 /* In this case, m should've been freed. */ 863 return (IPPROTO_DONE); 864 } 865 break; 866 867 badcode: 868 icmp6stat.icp6s_badcode++; 869 break; 870 871 badlen: 872 icmp6stat.icp6s_badlen++; 873 break; 874 } 875 876 /* deliver the packet to appropriate sockets */ 877 icmp6_rip6_input(&m, *offp); 878 879 return IPPROTO_DONE; 880 881 freeit: 882 m_freem(m); 883 return IPPROTO_DONE; 884 } 885 886 static int 887 icmp6_notify_error(m, off, icmp6len, code) 888 struct mbuf *m; 889 int off, icmp6len, code; 890 { 891 struct icmp6_hdr *icmp6; 892 struct ip6_hdr *eip6; 893 u_int32_t notifymtu; 894 struct sockaddr_in6 icmp6src, icmp6dst; 895 896 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 897 icmp6stat.icp6s_tooshort++; 898 goto freeit; 899 } 900 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 901 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 902 if (icmp6 == NULL) { 903 icmp6stat.icp6s_tooshort++; 904 return (-1); 905 } 906 eip6 = (struct ip6_hdr *)(icmp6 + 1); 907 908 /* Detect the upper level protocol */ 909 { 910 void (*ctlfunc) __P((int, struct sockaddr *, void *)); 911 u_int8_t nxt = eip6->ip6_nxt; 912 int eoff = off + sizeof(struct icmp6_hdr) + 913 sizeof(struct ip6_hdr); 914 struct ip6ctlparam ip6cp; 915 struct in6_addr *finaldst = NULL; 916 int icmp6type = icmp6->icmp6_type; 917 struct ip6_frag *fh; 918 struct ip6_rthdr *rth; 919 struct ip6_rthdr0 *rth0; 920 int rthlen; 921 922 while (1) { /* XXX: should avoid infinite loop explicitly? */ 923 struct ip6_ext *eh; 924 925 switch (nxt) { 926 case IPPROTO_HOPOPTS: 927 case IPPROTO_DSTOPTS: 928 case IPPROTO_AH: 929 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 930 eoff, sizeof(*eh)); 931 if (eh == NULL) { 932 icmp6stat.icp6s_tooshort++; 933 return (-1); 934 } 935 936 if (nxt == IPPROTO_AH) 937 eoff += (eh->ip6e_len + 2) << 2; 938 else 939 eoff += (eh->ip6e_len + 1) << 3; 940 nxt = eh->ip6e_nxt; 941 break; 942 case IPPROTO_ROUTING: 943 /* 944 * When the erroneous packet contains a 945 * routing header, we should examine the 946 * header to determine the final destination. 947 * Otherwise, we can't properly update 948 * information that depends on the final 949 * destination (e.g. path MTU). 950 */ 951 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 952 eoff, sizeof(*rth)); 953 if (rth == NULL) { 954 icmp6stat.icp6s_tooshort++; 955 return (-1); 956 } 957 rthlen = (rth->ip6r_len + 1) << 3; 958 /* 959 * XXX: currently there is no 960 * officially defined type other 961 * than type-0. 962 * Note that if the segment left field 963 * is 0, all intermediate hops must 964 * have been passed. 965 */ 966 if (rth->ip6r_segleft && 967 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 968 int hops; 969 970 IP6_EXTHDR_GET(rth0, 971 struct ip6_rthdr0 *, m, 972 eoff, rthlen); 973 if (rth0 == NULL) { 974 icmp6stat.icp6s_tooshort++; 975 return (-1); 976 } 977 /* just ignore a bogus header */ 978 if ((rth0->ip6r0_len % 2) == 0 && 979 (hops = rth0->ip6r0_len/2)) 980 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 981 } 982 eoff += rthlen; 983 nxt = rth->ip6r_nxt; 984 break; 985 case IPPROTO_FRAGMENT: 986 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 987 eoff, sizeof(*fh)); 988 if (fh == NULL) { 989 icmp6stat.icp6s_tooshort++; 990 return (-1); 991 } 992 /* 993 * Data after a fragment header is meaningless 994 * unless it is the first fragment, but 995 * we'll go to the notify label for path MTU 996 * discovery. 997 */ 998 if (fh->ip6f_offlg & IP6F_OFF_MASK) 999 goto notify; 1000 1001 eoff += sizeof(struct ip6_frag); 1002 nxt = fh->ip6f_nxt; 1003 break; 1004 default: 1005 /* 1006 * This case includes ESP and the No Next 1007 * Header. In such cases going to the notify 1008 * label does not have any meaning 1009 * (i.e. ctlfunc will be NULL), but we go 1010 * anyway since we might have to update 1011 * path MTU information. 1012 */ 1013 goto notify; 1014 } 1015 } 1016 notify: 1017 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1018 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1019 if (icmp6 == NULL) { 1020 icmp6stat.icp6s_tooshort++; 1021 return (-1); 1022 } 1023 1024 /* 1025 * retrieve parameters from the inner IPv6 header, and convert 1026 * them into sockaddr structures. 1027 * XXX: there is no guarantee that the source or destination 1028 * addresses of the inner packet are in the same scope zone as 1029 * the addresses of the icmp packet. But there is no other 1030 * way to determine the zone. 1031 */ 1032 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1033 1034 bzero(&icmp6dst, sizeof(icmp6dst)); 1035 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1036 icmp6dst.sin6_family = AF_INET6; 1037 if (finaldst == NULL) 1038 icmp6dst.sin6_addr = eip6->ip6_dst; 1039 else 1040 icmp6dst.sin6_addr = *finaldst; 1041 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1042 goto freeit; 1043 bzero(&icmp6src, sizeof(icmp6src)); 1044 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1045 icmp6src.sin6_family = AF_INET6; 1046 icmp6src.sin6_addr = eip6->ip6_src; 1047 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1048 goto freeit; 1049 icmp6src.sin6_flowinfo = 1050 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1051 1052 if (finaldst == NULL) 1053 finaldst = &eip6->ip6_dst; 1054 ip6cp.ip6c_m = m; 1055 ip6cp.ip6c_icmp6 = icmp6; 1056 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1057 ip6cp.ip6c_off = eoff; 1058 ip6cp.ip6c_finaldst = finaldst; 1059 ip6cp.ip6c_src = &icmp6src; 1060 ip6cp.ip6c_nxt = nxt; 1061 1062 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1063 notifymtu = ntohl(icmp6->icmp6_mtu); 1064 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1065 } 1066 1067 ctlfunc = (void (*) __P((int, struct sockaddr *, void *))) 1068 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1069 if (ctlfunc) { 1070 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1071 &ip6cp); 1072 } 1073 } 1074 return (0); 1075 1076 freeit: 1077 m_freem(m); 1078 return (-1); 1079 } 1080 1081 void 1082 icmp6_mtudisc_update(ip6cp, validated) 1083 struct ip6ctlparam *ip6cp; 1084 int validated; 1085 { 1086 unsigned long rtcount; 1087 struct icmp6_mtudisc_callback *mc; 1088 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1089 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1090 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1091 u_int mtu = ntohl(icmp6->icmp6_mtu); 1092 struct rtentry *rt = NULL; 1093 struct sockaddr_in6 sin6; 1094 1095 /* 1096 * allow non-validated cases if memory is plenty, to make traffic 1097 * from non-connected pcb happy. 1098 */ 1099 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1100 if (validated) { 1101 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1102 return; 1103 else if (0 <= icmp6_mtudisc_lowat && 1104 rtcount > icmp6_mtudisc_lowat) { 1105 /* 1106 * XXX nuke a victim, install the new one. 1107 */ 1108 } 1109 } else { 1110 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1111 return; 1112 } 1113 1114 bzero(&sin6, sizeof(sin6)); 1115 sin6.sin6_family = PF_INET6; 1116 sin6.sin6_len = sizeof(struct sockaddr_in6); 1117 sin6.sin6_addr = *dst; 1118 if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1119 return; 1120 1121 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1122 1123 if (rt && (rt->rt_flags & RTF_HOST) && 1124 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1125 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1126 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1127 icmp6stat.icp6s_pmtuchg++; 1128 rt->rt_rmx.rmx_mtu = mtu; 1129 } 1130 } 1131 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */ 1132 RTFREE(rt); 1133 } 1134 1135 /* 1136 * Notify protocols that the MTU for this destination 1137 * has changed. 1138 */ 1139 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1140 mc = LIST_NEXT(mc, mc_list)) 1141 (*mc->mc_func)(&sin6.sin6_addr); 1142 } 1143 1144 /* 1145 * Process a Node Information Query packet, based on 1146 * draft-ietf-ipngwg-icmp-name-lookups-07. 1147 * 1148 * Spec incompatibilities: 1149 * - IPv6 Subject address handling 1150 * - IPv4 Subject address handling support missing 1151 * - Proxy reply (answer even if it's not for me) 1152 * - joins NI group address at in6_ifattach() time only, does not cope 1153 * with hostname changes by sethostname(3) 1154 */ 1155 #ifndef offsetof /* XXX */ 1156 #define offsetof(type, member) ((size_t)(&((type *)0)->member)) 1157 #endif 1158 static struct mbuf * 1159 ni6_input(m, off) 1160 struct mbuf *m; 1161 int off; 1162 { 1163 struct icmp6_nodeinfo *ni6, *nni6; 1164 struct mbuf *n = NULL; 1165 u_int16_t qtype; 1166 int subjlen; 1167 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1168 struct ni_reply_fqdn *fqdn; 1169 int addrs; /* for NI_QTYPE_NODEADDR */ 1170 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1171 struct sockaddr_in6 sin6; /* ip6_dst */ 1172 struct in6_addr in6_subj; /* subject address */ 1173 struct ip6_hdr *ip6; 1174 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1175 char *subj = NULL; 1176 1177 ip6 = mtod(m, struct ip6_hdr *); 1178 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1179 if (ni6 == NULL) { 1180 /* m is already reclaimed */ 1181 return NULL; 1182 } 1183 1184 /* 1185 * Validate IPv6 destination address. 1186 * 1187 * The Responder must discard the Query without further processing 1188 * unless it is one of the Responder's unicast or anycast addresses, or 1189 * a link-local scope multicast address which the Responder has joined. 1190 * [icmp-name-lookups-07, Section 4.] 1191 */ 1192 bzero(&sin6, sizeof(sin6)); 1193 sin6.sin6_family = AF_INET6; 1194 sin6.sin6_len = sizeof(struct sockaddr_in6); 1195 bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr)); 1196 /* XXX scopeid */ 1197 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1198 ; /* unicast/anycast, fine */ 1199 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1200 ; /* link-local multicast, fine */ 1201 else 1202 goto bad; 1203 1204 /* validate query Subject field. */ 1205 qtype = ntohs(ni6->ni_qtype); 1206 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1207 switch (qtype) { 1208 case NI_QTYPE_NOOP: 1209 case NI_QTYPE_SUPTYPES: 1210 /* 07 draft */ 1211 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1212 break; 1213 /* FALLTHROUGH */ 1214 case NI_QTYPE_FQDN: 1215 case NI_QTYPE_NODEADDR: 1216 case NI_QTYPE_IPV4ADDR: 1217 switch (ni6->ni_code) { 1218 case ICMP6_NI_SUBJ_IPV6: 1219 #if ICMP6_NI_SUBJ_IPV6 != 0 1220 case 0: 1221 #endif 1222 /* 1223 * backward compatibility - try to accept 03 draft 1224 * format, where no Subject is present. 1225 */ 1226 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1227 subjlen == 0) { 1228 oldfqdn++; 1229 break; 1230 } 1231 #if ICMP6_NI_SUBJ_IPV6 != 0 1232 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1233 goto bad; 1234 #endif 1235 1236 if (subjlen != sizeof(sin6.sin6_addr)) 1237 goto bad; 1238 1239 /* 1240 * Validate Subject address. 1241 * 1242 * Not sure what exactly "address belongs to the node" 1243 * means in the spec, is it just unicast, or what? 1244 * 1245 * At this moment we consider Subject address as 1246 * "belong to the node" if the Subject address equals 1247 * to the IPv6 destination address; validation for 1248 * IPv6 destination address should have done enough 1249 * check for us. 1250 * 1251 * We do not do proxy at this moment. 1252 */ 1253 /* m_pulldown instead of copy? */ 1254 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1255 subjlen, (caddr_t)&in6_subj); 1256 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1257 goto bad; 1258 1259 subj = (char *)&in6_subj; 1260 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1261 break; 1262 1263 /* 1264 * XXX if we are to allow other cases, we should really 1265 * be careful about scope here. 1266 * basically, we should disallow queries toward IPv6 1267 * destination X with subject Y, if scope(X) > scope(Y). 1268 * if we allow scope(X) > scope(Y), it will result in 1269 * information leakage across scope boundary. 1270 */ 1271 goto bad; 1272 1273 case ICMP6_NI_SUBJ_FQDN: 1274 /* 1275 * Validate Subject name with gethostname(3). 1276 * 1277 * The behavior may need some debate, since: 1278 * - we are not sure if the node has FQDN as 1279 * hostname (returned by gethostname(3)). 1280 * - the code does wildcard match for truncated names. 1281 * however, we are not sure if we want to perform 1282 * wildcard match, if gethostname(3) side has 1283 * truncated hostname. 1284 */ 1285 n = ni6_nametodns(hostname, hostnamelen, 0); 1286 if (!n || n->m_next || n->m_len == 0) 1287 goto bad; 1288 IP6_EXTHDR_GET(subj, char *, m, 1289 off + sizeof(struct icmp6_nodeinfo), subjlen); 1290 if (subj == NULL) 1291 goto bad; 1292 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1293 n->m_len)) { 1294 goto bad; 1295 } 1296 m_freem(n); 1297 n = NULL; 1298 break; 1299 1300 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1301 default: 1302 goto bad; 1303 } 1304 break; 1305 } 1306 1307 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1308 switch (qtype) { 1309 case NI_QTYPE_FQDN: 1310 if ((icmp6_nodeinfo & 1) == 0) 1311 goto bad; 1312 break; 1313 case NI_QTYPE_NODEADDR: 1314 case NI_QTYPE_IPV4ADDR: 1315 if ((icmp6_nodeinfo & 2) == 0) 1316 goto bad; 1317 break; 1318 } 1319 1320 /* guess reply length */ 1321 switch (qtype) { 1322 case NI_QTYPE_NOOP: 1323 break; /* no reply data */ 1324 case NI_QTYPE_SUPTYPES: 1325 replylen += sizeof(u_int32_t); 1326 break; 1327 case NI_QTYPE_FQDN: 1328 /* XXX will append an mbuf */ 1329 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1330 break; 1331 case NI_QTYPE_NODEADDR: 1332 addrs = ni6_addrs(ni6, m, &ifp, subj); 1333 if ((replylen += addrs * (sizeof(struct in6_addr) + 1334 sizeof(u_int32_t))) > MCLBYTES) 1335 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1336 break; 1337 case NI_QTYPE_IPV4ADDR: 1338 /* unsupported - should respond with unknown Qtype? */ 1339 goto bad; 1340 default: 1341 /* 1342 * XXX: We must return a reply with the ICMP6 code 1343 * `unknown Qtype' in this case. However we regard the case 1344 * as an FQDN query for backward compatibility. 1345 * Older versions set a random value to this field, 1346 * so it rarely varies in the defined qtypes. 1347 * But the mechanism is not reliable... 1348 * maybe we should obsolete older versions. 1349 */ 1350 qtype = NI_QTYPE_FQDN; 1351 /* XXX will append an mbuf */ 1352 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1353 oldfqdn++; 1354 break; 1355 } 1356 1357 /* allocate an mbuf to reply. */ 1358 MGETHDR(n, M_DONTWAIT, m->m_type); 1359 if (n == NULL) { 1360 m_freem(m); 1361 return (NULL); 1362 } 1363 M_MOVE_PKTHDR(n, m); /* just for rcvif */ 1364 if (replylen > MHLEN) { 1365 if (replylen > MCLBYTES) { 1366 /* 1367 * XXX: should we try to allocate more? But MCLBYTES 1368 * is probably much larger than IPV6_MMTU... 1369 */ 1370 goto bad; 1371 } 1372 MCLGET(n, M_DONTWAIT); 1373 if ((n->m_flags & M_EXT) == 0) { 1374 goto bad; 1375 } 1376 } 1377 n->m_pkthdr.len = n->m_len = replylen; 1378 1379 /* copy mbuf header and IPv6 + Node Information base headers */ 1380 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1381 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1382 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1383 1384 /* qtype dependent procedure */ 1385 switch (qtype) { 1386 case NI_QTYPE_NOOP: 1387 nni6->ni_code = ICMP6_NI_SUCCESS; 1388 nni6->ni_flags = 0; 1389 break; 1390 case NI_QTYPE_SUPTYPES: 1391 { 1392 u_int32_t v; 1393 nni6->ni_code = ICMP6_NI_SUCCESS; 1394 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1395 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1396 v = (u_int32_t)htonl(0x0000000f); 1397 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1398 break; 1399 } 1400 case NI_QTYPE_FQDN: 1401 nni6->ni_code = ICMP6_NI_SUCCESS; 1402 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1403 sizeof(struct ip6_hdr) + 1404 sizeof(struct icmp6_nodeinfo)); 1405 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1406 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1407 /* 1408 * XXX do we really have FQDN in variable "hostname"? 1409 */ 1410 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1411 if (n->m_next == NULL) 1412 goto bad; 1413 /* XXX we assume that n->m_next is not a chain */ 1414 if (n->m_next->m_next != NULL) 1415 goto bad; 1416 n->m_pkthdr.len += n->m_next->m_len; 1417 break; 1418 case NI_QTYPE_NODEADDR: 1419 { 1420 int lenlim, copied; 1421 1422 nni6->ni_code = ICMP6_NI_SUCCESS; 1423 n->m_pkthdr.len = n->m_len = 1424 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1425 lenlim = M_TRAILINGSPACE(n); 1426 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1427 /* XXX: reset mbuf length */ 1428 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1429 sizeof(struct icmp6_nodeinfo) + copied; 1430 break; 1431 } 1432 default: 1433 break; /* XXX impossible! */ 1434 } 1435 1436 nni6->ni_type = ICMP6_NI_REPLY; 1437 m_freem(m); 1438 return (n); 1439 1440 bad: 1441 m_freem(m); 1442 if (n) 1443 m_freem(n); 1444 return (NULL); 1445 } 1446 #undef hostnamelen 1447 1448 #define isupper(x) ('A' <= (x) && (x) <= 'Z') 1449 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1450 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1451 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1452 1453 /* 1454 * make a mbuf with DNS-encoded string. no compression support. 1455 * 1456 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1457 * treated as truncated name (two \0 at the end). this is a wild guess. 1458 */ 1459 static struct mbuf * 1460 ni6_nametodns(name, namelen, old) 1461 const char *name; 1462 int namelen; 1463 int old; /* return pascal string if non-zero */ 1464 { 1465 struct mbuf *m; 1466 char *cp, *ep; 1467 const char *p, *q; 1468 int i, len, nterm; 1469 1470 if (old) 1471 len = namelen + 1; 1472 else 1473 len = MCLBYTES; 1474 1475 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1476 MGET(m, M_DONTWAIT, MT_DATA); 1477 if (m && len > MLEN) { 1478 MCLGET(m, M_DONTWAIT); 1479 if ((m->m_flags & M_EXT) == 0) 1480 goto fail; 1481 } 1482 if (!m) 1483 goto fail; 1484 m->m_next = NULL; 1485 1486 if (old) { 1487 m->m_len = len; 1488 *mtod(m, char *) = namelen; 1489 bcopy(name, mtod(m, char *) + 1, namelen); 1490 return m; 1491 } else { 1492 m->m_len = 0; 1493 cp = mtod(m, char *); 1494 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1495 1496 /* if not certain about my name, return empty buffer */ 1497 if (namelen == 0) 1498 return m; 1499 1500 /* 1501 * guess if it looks like shortened hostname, or FQDN. 1502 * shortened hostname needs two trailing "\0". 1503 */ 1504 i = 0; 1505 for (p = name; p < name + namelen; p++) { 1506 if (*p && *p == '.') 1507 i++; 1508 } 1509 if (i < 2) 1510 nterm = 2; 1511 else 1512 nterm = 1; 1513 1514 p = name; 1515 while (cp < ep && p < name + namelen) { 1516 i = 0; 1517 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1518 i++; 1519 /* result does not fit into mbuf */ 1520 if (cp + i + 1 >= ep) 1521 goto fail; 1522 /* 1523 * DNS label length restriction, RFC1035 page 8. 1524 * "i == 0" case is included here to avoid returning 1525 * 0-length label on "foo..bar". 1526 */ 1527 if (i <= 0 || i >= 64) 1528 goto fail; 1529 *cp++ = i; 1530 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1531 goto fail; 1532 while (i > 0) { 1533 if (!isalnum(*p) && *p != '-') 1534 goto fail; 1535 if (isupper(*p)) { 1536 *cp++ = tolower(*p); 1537 p++; 1538 } else 1539 *cp++ = *p++; 1540 i--; 1541 } 1542 p = q; 1543 if (p < name + namelen && *p == '.') 1544 p++; 1545 } 1546 /* termination */ 1547 if (cp + nterm >= ep) 1548 goto fail; 1549 while (nterm-- > 0) 1550 *cp++ = '\0'; 1551 m->m_len = cp - mtod(m, char *); 1552 return m; 1553 } 1554 1555 panic("should not reach here"); 1556 /* NOTREACHED */ 1557 1558 fail: 1559 if (m) 1560 m_freem(m); 1561 return NULL; 1562 } 1563 1564 /* 1565 * check if two DNS-encoded string matches. takes care of truncated 1566 * form (with \0\0 at the end). no compression support. 1567 * XXX upper/lowercase match (see RFC2065) 1568 */ 1569 static int 1570 ni6_dnsmatch(a, alen, b, blen) 1571 const char *a; 1572 int alen; 1573 const char *b; 1574 int blen; 1575 { 1576 const char *a0, *b0; 1577 int l; 1578 1579 /* simplest case - need validation? */ 1580 if (alen == blen && bcmp(a, b, alen) == 0) 1581 return 1; 1582 1583 a0 = a; 1584 b0 = b; 1585 1586 /* termination is mandatory */ 1587 if (alen < 2 || blen < 2) 1588 return 0; 1589 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1590 return 0; 1591 alen--; 1592 blen--; 1593 1594 while (a - a0 < alen && b - b0 < blen) { 1595 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1596 return 0; 1597 1598 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1599 return 0; 1600 /* we don't support compression yet */ 1601 if (a[0] >= 64 || b[0] >= 64) 1602 return 0; 1603 1604 /* truncated case */ 1605 if (a[0] == 0 && a - a0 == alen - 1) 1606 return 1; 1607 if (b[0] == 0 && b - b0 == blen - 1) 1608 return 1; 1609 if (a[0] == 0 || b[0] == 0) 1610 return 0; 1611 1612 if (a[0] != b[0]) 1613 return 0; 1614 l = a[0]; 1615 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1616 return 0; 1617 if (bcmp(a + 1, b + 1, l) != 0) 1618 return 0; 1619 1620 a += 1 + l; 1621 b += 1 + l; 1622 } 1623 1624 if (a - a0 == alen && b - b0 == blen) 1625 return 1; 1626 else 1627 return 0; 1628 } 1629 1630 /* 1631 * calculate the number of addresses to be returned in the node info reply. 1632 */ 1633 static int 1634 ni6_addrs(ni6, m, ifpp, subj) 1635 struct icmp6_nodeinfo *ni6; 1636 struct mbuf *m; 1637 struct ifnet **ifpp; 1638 char *subj; 1639 { 1640 struct ifnet *ifp; 1641 struct in6_ifaddr *ifa6; 1642 struct ifaddr *ifa; 1643 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1644 int addrs = 0, addrsofif, iffound = 0; 1645 int niflags = ni6->ni_flags; 1646 1647 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1648 switch (ni6->ni_code) { 1649 case ICMP6_NI_SUBJ_IPV6: 1650 if (subj == NULL) /* must be impossible... */ 1651 return (0); 1652 subj_ip6 = (struct sockaddr_in6 *)subj; 1653 break; 1654 default: 1655 /* 1656 * XXX: we only support IPv6 subject address for 1657 * this Qtype. 1658 */ 1659 return (0); 1660 } 1661 } 1662 1663 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1664 { 1665 addrsofif = 0; 1666 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1667 ifa = ifa->ifa_list.tqe_next) 1668 { 1669 if (ifa->ifa_addr->sa_family != AF_INET6) 1670 continue; 1671 ifa6 = (struct in6_ifaddr *)ifa; 1672 1673 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1674 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1675 &ifa6->ia_addr.sin6_addr)) 1676 iffound = 1; 1677 1678 /* 1679 * IPv4-mapped addresses can only be returned by a 1680 * Node Information proxy, since they represent 1681 * addresses of IPv4-only nodes, which perforce do 1682 * not implement this protocol. 1683 * [icmp-name-lookups-07, Section 5.4] 1684 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1685 * this function at this moment. 1686 */ 1687 1688 /* What do we have to do about ::1? */ 1689 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1690 case IPV6_ADDR_SCOPE_LINKLOCAL: 1691 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1692 continue; 1693 break; 1694 case IPV6_ADDR_SCOPE_SITELOCAL: 1695 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1696 continue; 1697 break; 1698 case IPV6_ADDR_SCOPE_GLOBAL: 1699 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1700 continue; 1701 break; 1702 default: 1703 continue; 1704 } 1705 1706 /* 1707 * check if anycast is okay. 1708 * XXX: just experimental. not in the spec. 1709 */ 1710 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1711 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1712 continue; /* we need only unicast addresses */ 1713 1714 addrsofif++; /* count the address */ 1715 } 1716 if (iffound) { 1717 *ifpp = ifp; 1718 return (addrsofif); 1719 } 1720 1721 addrs += addrsofif; 1722 } 1723 1724 return (addrs); 1725 } 1726 1727 static int 1728 ni6_store_addrs(ni6, nni6, ifp0, resid) 1729 struct icmp6_nodeinfo *ni6, *nni6; 1730 struct ifnet *ifp0; 1731 int resid; 1732 { 1733 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1734 struct in6_ifaddr *ifa6; 1735 struct ifaddr *ifa; 1736 struct ifnet *ifp_dep = NULL; 1737 int copied = 0, allow_deprecated = 0; 1738 u_char *cp = (u_char *)(nni6 + 1); 1739 int niflags = ni6->ni_flags; 1740 u_int32_t ltime; 1741 1742 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1743 return (0); /* needless to copy */ 1744 1745 again: 1746 1747 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1748 { 1749 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1750 ifa = ifa->ifa_list.tqe_next) 1751 { 1752 if (ifa->ifa_addr->sa_family != AF_INET6) 1753 continue; 1754 ifa6 = (struct in6_ifaddr *)ifa; 1755 1756 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1757 allow_deprecated == 0) { 1758 /* 1759 * prefererred address should be put before 1760 * deprecated addresses. 1761 */ 1762 1763 /* record the interface for later search */ 1764 if (ifp_dep == NULL) 1765 ifp_dep = ifp; 1766 1767 continue; 1768 } 1769 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1770 allow_deprecated != 0) 1771 continue; /* we now collect deprecated addrs */ 1772 1773 /* What do we have to do about ::1? */ 1774 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1775 case IPV6_ADDR_SCOPE_LINKLOCAL: 1776 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1777 continue; 1778 break; 1779 case IPV6_ADDR_SCOPE_SITELOCAL: 1780 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1781 continue; 1782 break; 1783 case IPV6_ADDR_SCOPE_GLOBAL: 1784 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1785 continue; 1786 break; 1787 default: 1788 continue; 1789 } 1790 1791 /* 1792 * check if anycast is okay. 1793 * XXX: just experimental. not in the spec. 1794 */ 1795 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1796 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1797 continue; 1798 1799 /* now we can copy the address */ 1800 if (resid < sizeof(struct in6_addr) + 1801 sizeof(u_int32_t)) { 1802 /* 1803 * We give up much more copy. 1804 * Set the truncate flag and return. 1805 */ 1806 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1807 return (copied); 1808 } 1809 1810 /* 1811 * Set the TTL of the address. 1812 * The TTL value should be one of the following 1813 * according to the specification: 1814 * 1815 * 1. The remaining lifetime of a DHCP lease on the 1816 * address, or 1817 * 2. The remaining Valid Lifetime of a prefix from 1818 * which the address was derived through Stateless 1819 * Autoconfiguration. 1820 * 1821 * Note that we currently do not support stateful 1822 * address configuration by DHCPv6, so the former 1823 * case can't happen. 1824 * 1825 * TTL must be 2^31 > TTL >= 0. 1826 */ 1827 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1828 ltime = ND6_INFINITE_LIFETIME; 1829 else { 1830 if (ifa6->ia6_lifetime.ia6t_expire > 1831 time_second) 1832 ltime = ifa6->ia6_lifetime.ia6t_expire - 1833 time_second; 1834 else 1835 ltime = 0; 1836 } 1837 if (ltime > 0x7fffffff) 1838 ltime = 0x7fffffff; 1839 ltime = htonl(ltime); 1840 1841 bcopy(<ime, cp, sizeof(u_int32_t)); 1842 cp += sizeof(u_int32_t); 1843 1844 /* copy the address itself */ 1845 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1846 sizeof(struct in6_addr)); 1847 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1848 cp += sizeof(struct in6_addr); 1849 1850 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1851 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1852 } 1853 if (ifp0) /* we need search only on the specified IF */ 1854 break; 1855 } 1856 1857 if (allow_deprecated == 0 && ifp_dep != NULL) { 1858 ifp = ifp_dep; 1859 allow_deprecated = 1; 1860 1861 goto again; 1862 } 1863 1864 return (copied); 1865 } 1866 1867 /* 1868 * XXX almost dup'ed code with rip6_input. 1869 */ 1870 static int 1871 icmp6_rip6_input(mp, off) 1872 struct mbuf **mp; 1873 int off; 1874 { 1875 struct mbuf *m = *mp; 1876 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1877 struct inpcb_hdr *inph; 1878 struct in6pcb *in6p; 1879 struct in6pcb *last = NULL; 1880 struct sockaddr_in6 rip6src; 1881 struct icmp6_hdr *icmp6; 1882 struct mbuf *opts = NULL; 1883 1884 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1885 if (icmp6 == NULL) { 1886 /* m is already reclaimed */ 1887 return IPPROTO_DONE; 1888 } 1889 1890 /* 1891 * XXX: the address may have embedded scope zone ID, which should be 1892 * hidden from applications. 1893 */ 1894 bzero(&rip6src, sizeof(rip6src)); 1895 rip6src.sin6_len = sizeof(struct sockaddr_in6); 1896 rip6src.sin6_family = AF_INET6; 1897 rip6src.sin6_addr = ip6->ip6_src; 1898 if (sa6_recoverscope(&rip6src)) { 1899 m_freem(m); 1900 return (IPPROTO_DONE); 1901 } 1902 1903 CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) { 1904 in6p = (struct in6pcb *)inph; 1905 if (in6p->in6p_af != AF_INET6) 1906 continue; 1907 if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6) 1908 continue; 1909 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1910 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1911 continue; 1912 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1913 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1914 continue; 1915 if (in6p->in6p_icmp6filt 1916 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1917 in6p->in6p_icmp6filt)) 1918 continue; 1919 if (last) { 1920 struct mbuf *n; 1921 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1922 if (last->in6p_flags & IN6P_CONTROLOPTS) 1923 ip6_savecontrol(last, &opts, ip6, n); 1924 /* strip intermediate headers */ 1925 m_adj(n, off); 1926 if (sbappendaddr(&last->in6p_socket->so_rcv, 1927 (struct sockaddr *)&rip6src, 1928 n, opts) == 0) { 1929 /* should notify about lost packet */ 1930 m_freem(n); 1931 if (opts) 1932 m_freem(opts); 1933 } else 1934 sorwakeup(last->in6p_socket); 1935 opts = NULL; 1936 } 1937 } 1938 last = in6p; 1939 } 1940 if (last) { 1941 if (last->in6p_flags & IN6P_CONTROLOPTS) 1942 ip6_savecontrol(last, &opts, ip6, m); 1943 /* strip intermediate headers */ 1944 m_adj(m, off); 1945 if (sbappendaddr(&last->in6p_socket->so_rcv, 1946 (struct sockaddr *)&rip6src, m, opts) == 0) { 1947 m_freem(m); 1948 if (opts) 1949 m_freem(opts); 1950 } else 1951 sorwakeup(last->in6p_socket); 1952 } else { 1953 m_freem(m); 1954 ip6stat.ip6s_delivered--; 1955 } 1956 return IPPROTO_DONE; 1957 } 1958 1959 /* 1960 * Reflect the ip6 packet back to the source. 1961 * OFF points to the icmp6 header, counted from the top of the mbuf. 1962 * 1963 * Note: RFC 1885 required that an echo reply should be truncated if it 1964 * did not fit in with (return) path MTU, and KAME code supported the 1965 * behavior. However, as a clarification after the RFC, this limitation 1966 * was removed in a revised version of the spec, RFC 2463. We had kept the 1967 * old behavior, with a (non-default) ifdef block, while the new version of 1968 * the spec was an internet-draft status, and even after the new RFC was 1969 * published. But it would rather make sense to clean the obsoleted part 1970 * up, and to make the code simpler at this stage. 1971 */ 1972 void 1973 icmp6_reflect(m, off) 1974 struct mbuf *m; 1975 size_t off; 1976 { 1977 struct ip6_hdr *ip6; 1978 struct icmp6_hdr *icmp6; 1979 struct in6_ifaddr *ia; 1980 int plen; 1981 int type, code; 1982 struct ifnet *outif = NULL; 1983 struct in6_addr origdst, *src = NULL; 1984 1985 /* too short to reflect */ 1986 if (off < sizeof(struct ip6_hdr)) { 1987 nd6log((LOG_DEBUG, 1988 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 1989 (u_long)off, (u_long)sizeof(struct ip6_hdr), 1990 __FILE__, __LINE__)); 1991 goto bad; 1992 } 1993 1994 /* 1995 * If there are extra headers between IPv6 and ICMPv6, strip 1996 * off that header first. 1997 */ 1998 #ifdef DIAGNOSTIC 1999 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2000 panic("assumption failed in icmp6_reflect"); 2001 #endif 2002 if (off > sizeof(struct ip6_hdr)) { 2003 size_t l; 2004 struct ip6_hdr nip6; 2005 2006 l = off - sizeof(struct ip6_hdr); 2007 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2008 m_adj(m, l); 2009 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2010 if (m->m_len < l) { 2011 if ((m = m_pullup(m, l)) == NULL) 2012 return; 2013 } 2014 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2015 } else /* off == sizeof(struct ip6_hdr) */ { 2016 size_t l; 2017 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2018 if (m->m_len < l) { 2019 if ((m = m_pullup(m, l)) == NULL) 2020 return; 2021 } 2022 } 2023 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2024 ip6 = mtod(m, struct ip6_hdr *); 2025 ip6->ip6_nxt = IPPROTO_ICMPV6; 2026 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2027 type = icmp6->icmp6_type; /* keep type for statistics */ 2028 code = icmp6->icmp6_code; /* ditto. */ 2029 2030 origdst = ip6->ip6_dst; 2031 /* 2032 * ip6_input() drops a packet if its src is multicast. 2033 * So, the src is never multicast. 2034 */ 2035 ip6->ip6_dst = ip6->ip6_src; 2036 2037 /* 2038 * If the incoming packet was addressed directly to us (i.e. unicast), 2039 * use dst as the src for the reply. 2040 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2041 * (for example) when we encounter an error while forwarding procedure 2042 * destined to a duplicated address of ours. 2043 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2044 * procedure of an outgoing packet of our own, in which case we need 2045 * to search in the ifaddr list. 2046 */ 2047 if (!IN6_IS_ADDR_MULTICAST(&origdst)) { 2048 if ((ia = ip6_getdstifaddr(m))) { 2049 if (!(ia->ia6_flags & 2050 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) 2051 src = &ia->ia_addr.sin6_addr; 2052 } else { 2053 struct sockaddr_in6 d; 2054 2055 bzero(&d, sizeof(d)); 2056 d.sin6_family = AF_INET6; 2057 d.sin6_len = sizeof(d); 2058 d.sin6_addr = origdst; 2059 ia = (struct in6_ifaddr *) 2060 ifa_ifwithaddr((struct sockaddr *)&d); 2061 if (ia && 2062 !(ia->ia6_flags & 2063 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) { 2064 src = &ia->ia_addr.sin6_addr; 2065 } 2066 } 2067 } 2068 2069 if (src == NULL) { 2070 int e; 2071 struct sockaddr_in6 sin6; 2072 struct route_in6 ro; 2073 2074 /* 2075 * This case matches to multicasts, our anycast, or unicasts 2076 * that we do not own. Select a source address based on the 2077 * source address of the erroneous packet. 2078 */ 2079 bzero(&sin6, sizeof(sin6)); 2080 sin6.sin6_family = AF_INET6; 2081 sin6.sin6_len = sizeof(sin6); 2082 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */ 2083 2084 bzero(&ro, sizeof(ro)); 2085 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2086 if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */ 2087 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2088 } 2089 if (src == NULL) { 2090 nd6log((LOG_DEBUG, 2091 "icmp6_reflect: source can't be determined: " 2092 "dst=%s, error=%d\n", 2093 ip6_sprintf(&sin6.sin6_addr), e)); 2094 goto bad; 2095 } 2096 } 2097 2098 ip6->ip6_src = *src; 2099 ip6->ip6_flow = 0; 2100 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2101 ip6->ip6_vfc |= IPV6_VERSION; 2102 ip6->ip6_nxt = IPPROTO_ICMPV6; 2103 if (m->m_pkthdr.rcvif) { 2104 /* XXX: This may not be the outgoing interface */ 2105 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2106 } else 2107 ip6->ip6_hlim = ip6_defhlim; 2108 2109 m->m_pkthdr.csum_flags = 0; 2110 icmp6->icmp6_cksum = 0; 2111 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2112 sizeof(struct ip6_hdr), plen); 2113 2114 /* 2115 * XXX option handling 2116 */ 2117 2118 m->m_flags &= ~(M_BCAST|M_MCAST); 2119 2120 /* 2121 * To avoid a "too big" situation at an intermediate router 2122 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2123 * Note that only echo and node information replies are affected, 2124 * since the length of ICMP6 errors is limited to the minimum MTU. 2125 */ 2126 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 && 2127 outif) 2128 icmp6_ifstat_inc(outif, ifs6_out_error); 2129 2130 if (outif) 2131 icmp6_ifoutstat_inc(outif, type, code); 2132 2133 return; 2134 2135 bad: 2136 m_freem(m); 2137 return; 2138 } 2139 2140 static const char * 2141 icmp6_redirect_diag(src6, dst6, tgt6) 2142 struct in6_addr *src6; 2143 struct in6_addr *dst6; 2144 struct in6_addr *tgt6; 2145 { 2146 static char buf[1024]; 2147 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2148 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2149 return buf; 2150 } 2151 2152 void 2153 icmp6_redirect_input(m, off) 2154 struct mbuf *m; 2155 int off; 2156 { 2157 struct ifnet *ifp = m->m_pkthdr.rcvif; 2158 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2159 struct nd_redirect *nd_rd; 2160 int icmp6len = ntohs(ip6->ip6_plen); 2161 char *lladdr = NULL; 2162 int lladdrlen = 0; 2163 struct rtentry *rt = NULL; 2164 int is_router; 2165 int is_onlink; 2166 struct in6_addr src6 = ip6->ip6_src; 2167 struct in6_addr redtgt6; 2168 struct in6_addr reddst6; 2169 union nd_opts ndopts; 2170 2171 if (!ifp) 2172 return; 2173 2174 /* XXX if we are router, we don't update route by icmp6 redirect */ 2175 if (ip6_forwarding) 2176 goto freeit; 2177 if (!icmp6_rediraccept) 2178 goto freeit; 2179 2180 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2181 if (nd_rd == NULL) { 2182 icmp6stat.icp6s_tooshort++; 2183 return; 2184 } 2185 redtgt6 = nd_rd->nd_rd_target; 2186 reddst6 = nd_rd->nd_rd_dst; 2187 2188 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2189 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2190 goto freeit; 2191 } 2192 2193 /* validation */ 2194 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2195 nd6log((LOG_ERR, 2196 "ICMP6 redirect sent from %s rejected; " 2197 "must be from linklocal\n", ip6_sprintf(&src6))); 2198 goto bad; 2199 } 2200 if (ip6->ip6_hlim != 255) { 2201 nd6log((LOG_ERR, 2202 "ICMP6 redirect sent from %s rejected; " 2203 "hlim=%d (must be 255)\n", 2204 ip6_sprintf(&src6), ip6->ip6_hlim)); 2205 goto bad; 2206 } 2207 { 2208 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2209 struct sockaddr_in6 sin6; 2210 struct in6_addr *gw6; 2211 2212 bzero(&sin6, sizeof(sin6)); 2213 sin6.sin6_family = AF_INET6; 2214 sin6.sin6_len = sizeof(struct sockaddr_in6); 2215 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2216 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2217 if (rt) { 2218 if (rt->rt_gateway == NULL || 2219 rt->rt_gateway->sa_family != AF_INET6) { 2220 nd6log((LOG_ERR, 2221 "ICMP6 redirect rejected; no route " 2222 "with inet6 gateway found for redirect dst: %s\n", 2223 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2224 RTFREE(rt); 2225 goto bad; 2226 } 2227 2228 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2229 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2230 nd6log((LOG_ERR, 2231 "ICMP6 redirect rejected; " 2232 "not equal to gw-for-src=%s (must be same): " 2233 "%s\n", 2234 ip6_sprintf(gw6), 2235 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2236 RTFREE(rt); 2237 goto bad; 2238 } 2239 } else { 2240 nd6log((LOG_ERR, 2241 "ICMP6 redirect rejected; " 2242 "no route found for redirect dst: %s\n", 2243 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2244 goto bad; 2245 } 2246 RTFREE(rt); 2247 rt = NULL; 2248 } 2249 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2250 nd6log((LOG_ERR, 2251 "ICMP6 redirect rejected; " 2252 "redirect dst must be unicast: %s\n", 2253 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2254 goto bad; 2255 } 2256 2257 is_router = is_onlink = 0; 2258 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2259 is_router = 1; /* router case */ 2260 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2261 is_onlink = 1; /* on-link destination case */ 2262 if (!is_router && !is_onlink) { 2263 nd6log((LOG_ERR, 2264 "ICMP6 redirect rejected; " 2265 "neither router case nor onlink case: %s\n", 2266 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2267 goto bad; 2268 } 2269 /* validation passed */ 2270 2271 icmp6len -= sizeof(*nd_rd); 2272 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2273 if (nd6_options(&ndopts) < 0) { 2274 nd6log((LOG_INFO, "icmp6_redirect_input: " 2275 "invalid ND option, rejected: %s\n", 2276 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2277 /* nd6_options have incremented stats */ 2278 goto freeit; 2279 } 2280 2281 if (ndopts.nd_opts_tgt_lladdr) { 2282 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2283 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2284 } 2285 2286 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2287 nd6log((LOG_INFO, 2288 "icmp6_redirect_input: lladdrlen mismatch for %s " 2289 "(if %d, icmp6 packet %d): %s\n", 2290 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2291 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2292 goto bad; 2293 } 2294 2295 /* RFC 2461 8.3 */ 2296 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2297 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2298 2299 if (!is_onlink) { /* better router case. perform rtredirect. */ 2300 /* perform rtredirect */ 2301 struct sockaddr_in6 sdst; 2302 struct sockaddr_in6 sgw; 2303 struct sockaddr_in6 ssrc; 2304 unsigned long rtcount; 2305 struct rtentry *newrt = NULL; 2306 2307 /* 2308 * do not install redirect route, if the number of entries 2309 * is too much (> hiwat). note that, the node (= host) will 2310 * work just fine even if we do not install redirect route 2311 * (there will be additional hops, though). 2312 */ 2313 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2314 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2315 return; 2316 else if (0 <= icmp6_redirect_lowat && 2317 rtcount > icmp6_redirect_lowat) { 2318 /* 2319 * XXX nuke a victim, install the new one. 2320 */ 2321 } 2322 2323 bzero(&sdst, sizeof(sdst)); 2324 bzero(&sgw, sizeof(sgw)); 2325 bzero(&ssrc, sizeof(ssrc)); 2326 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2327 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2328 sizeof(struct sockaddr_in6); 2329 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2330 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2331 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2332 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2333 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2334 (struct sockaddr *)&ssrc, 2335 &newrt); 2336 2337 if (newrt) { 2338 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2339 icmp6_redirect_timeout_q); 2340 rtfree(newrt); 2341 } 2342 } 2343 /* finally update cached route in each socket via pfctlinput */ 2344 { 2345 struct sockaddr_in6 sdst; 2346 2347 bzero(&sdst, sizeof(sdst)); 2348 sdst.sin6_family = AF_INET6; 2349 sdst.sin6_len = sizeof(struct sockaddr_in6); 2350 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2351 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2352 #ifdef IPSEC 2353 key_sa_routechange((struct sockaddr *)&sdst); 2354 #endif 2355 } 2356 2357 freeit: 2358 m_freem(m); 2359 return; 2360 2361 bad: 2362 icmp6stat.icp6s_badredirect++; 2363 m_freem(m); 2364 } 2365 2366 void 2367 icmp6_redirect_output(m0, rt) 2368 struct mbuf *m0; 2369 struct rtentry *rt; 2370 { 2371 struct ifnet *ifp; /* my outgoing interface */ 2372 struct in6_addr *ifp_ll6; 2373 struct in6_addr *nexthop; 2374 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2375 struct mbuf *m = NULL; /* newly allocated one */ 2376 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2377 struct nd_redirect *nd_rd; 2378 size_t maxlen; 2379 u_char *p; 2380 struct sockaddr_in6 src_sa; 2381 2382 icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2383 2384 /* if we are not router, we don't send icmp6 redirect */ 2385 if (!ip6_forwarding) 2386 goto fail; 2387 2388 /* sanity check */ 2389 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2390 goto fail; 2391 2392 /* 2393 * Address check: 2394 * the source address must identify a neighbor, and 2395 * the destination address must not be a multicast address 2396 * [RFC 2461, sec 8.2] 2397 */ 2398 sip6 = mtod(m0, struct ip6_hdr *); 2399 bzero(&src_sa, sizeof(src_sa)); 2400 src_sa.sin6_family = AF_INET6; 2401 src_sa.sin6_len = sizeof(src_sa); 2402 src_sa.sin6_addr = sip6->ip6_src; 2403 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2404 goto fail; 2405 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2406 goto fail; /* what should we do here? */ 2407 2408 /* rate limit */ 2409 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2410 goto fail; 2411 2412 /* 2413 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2414 * we almost always ask for an mbuf cluster for simplicity. 2415 * (MHLEN < IPV6_MMTU is almost always true) 2416 */ 2417 #if IPV6_MMTU >= MCLBYTES 2418 # error assumption failed about IPV6_MMTU and MCLBYTES 2419 #endif 2420 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2421 if (m && IPV6_MMTU >= MHLEN) 2422 MCLGET(m, M_DONTWAIT); 2423 if (!m) 2424 goto fail; 2425 m->m_pkthdr.rcvif = NULL; 2426 m->m_len = 0; 2427 maxlen = M_TRAILINGSPACE(m); 2428 maxlen = min(IPV6_MMTU, maxlen); 2429 /* just for safety */ 2430 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2431 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2432 goto fail; 2433 } 2434 2435 { 2436 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2437 struct in6_ifaddr *ia; 2438 if ((ia = in6ifa_ifpforlinklocal(ifp, 2439 IN6_IFF_NOTREADY| 2440 IN6_IFF_ANYCAST)) == NULL) 2441 goto fail; 2442 ifp_ll6 = &ia->ia_addr.sin6_addr; 2443 } 2444 2445 /* get ip6 linklocal address for the router. */ 2446 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2447 struct sockaddr_in6 *sin6; 2448 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2449 nexthop = &sin6->sin6_addr; 2450 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2451 nexthop = NULL; 2452 } else 2453 nexthop = NULL; 2454 2455 /* ip6 */ 2456 ip6 = mtod(m, struct ip6_hdr *); 2457 ip6->ip6_flow = 0; 2458 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2459 ip6->ip6_vfc |= IPV6_VERSION; 2460 /* ip6->ip6_plen will be set later */ 2461 ip6->ip6_nxt = IPPROTO_ICMPV6; 2462 ip6->ip6_hlim = 255; 2463 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2464 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2465 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2466 2467 /* ND Redirect */ 2468 nd_rd = (struct nd_redirect *)(ip6 + 1); 2469 nd_rd->nd_rd_type = ND_REDIRECT; 2470 nd_rd->nd_rd_code = 0; 2471 nd_rd->nd_rd_reserved = 0; 2472 if (rt->rt_flags & RTF_GATEWAY) { 2473 /* 2474 * nd_rd->nd_rd_target must be a link-local address in 2475 * better router cases. 2476 */ 2477 if (!nexthop) 2478 goto fail; 2479 bcopy(nexthop, &nd_rd->nd_rd_target, 2480 sizeof(nd_rd->nd_rd_target)); 2481 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2482 sizeof(nd_rd->nd_rd_dst)); 2483 } else { 2484 /* make sure redtgt == reddst */ 2485 nexthop = &sip6->ip6_dst; 2486 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2487 sizeof(nd_rd->nd_rd_target)); 2488 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2489 sizeof(nd_rd->nd_rd_dst)); 2490 } 2491 2492 p = (u_char *)(nd_rd + 1); 2493 2494 { 2495 /* target lladdr option */ 2496 struct rtentry *rt_nexthop = NULL; 2497 int len; 2498 struct sockaddr_dl *sdl; 2499 struct nd_opt_hdr *nd_opt; 2500 char *lladdr; 2501 2502 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2503 if (!rt_nexthop) 2504 goto nolladdropt; 2505 len = sizeof(*nd_opt) + ifp->if_addrlen; 2506 len = (len + 7) & ~7; /* round by 8 */ 2507 /* safety check */ 2508 if (len + (p - (u_char *)ip6) > maxlen) 2509 goto nolladdropt; 2510 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2511 (rt_nexthop->rt_flags & RTF_LLINFO) && 2512 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2513 (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) && 2514 sdl->sdl_alen) { 2515 nd_opt = (struct nd_opt_hdr *)p; 2516 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2517 nd_opt->nd_opt_len = len >> 3; 2518 lladdr = (char *)(nd_opt + 1); 2519 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen); 2520 p += len; 2521 } 2522 } 2523 nolladdropt:; 2524 2525 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2526 2527 /* just to be safe */ 2528 if (m0->m_flags & M_DECRYPTED) 2529 goto noredhdropt; 2530 if (p - (u_char *)ip6 > maxlen) 2531 goto noredhdropt; 2532 2533 { 2534 /* redirected header option */ 2535 int len; 2536 struct nd_opt_rd_hdr *nd_opt_rh; 2537 2538 /* 2539 * compute the maximum size for icmp6 redirect header option. 2540 * XXX room for auth header? 2541 */ 2542 len = maxlen - (p - (u_char *)ip6); 2543 len &= ~7; 2544 2545 /* 2546 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2547 * about padding/truncate rule for the original IP packet. 2548 * From the discussion on IPv6imp in Feb 1999, 2549 * the consensus was: 2550 * - "attach as much as possible" is the goal 2551 * - pad if not aligned (original size can be guessed by 2552 * original ip6 header) 2553 * Following code adds the padding if it is simple enough, 2554 * and truncates if not. 2555 */ 2556 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2557 /* not enough room, truncate */ 2558 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2559 m0->m_pkthdr.len); 2560 } else { 2561 /* 2562 * enough room, truncate if not aligned. 2563 * we don't pad here for simplicity. 2564 */ 2565 size_t extra; 2566 2567 extra = m0->m_pkthdr.len % 8; 2568 if (extra) { 2569 /* truncate */ 2570 m_adj(m0, -extra); 2571 } 2572 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2573 } 2574 2575 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2576 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2577 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2578 nd_opt_rh->nd_opt_rh_len = len >> 3; 2579 p += sizeof(*nd_opt_rh); 2580 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2581 2582 /* connect m0 to m */ 2583 m->m_pkthdr.len += m0->m_pkthdr.len; 2584 m_cat(m, m0); 2585 m0 = NULL; 2586 } 2587 noredhdropt: 2588 if (m0) { 2589 m_freem(m0); 2590 m0 = NULL; 2591 } 2592 2593 /* XXX: clear embedded link IDs in the inner header */ 2594 in6_clearscope(&sip6->ip6_src); 2595 in6_clearscope(&sip6->ip6_dst); 2596 in6_clearscope(&nd_rd->nd_rd_target); 2597 in6_clearscope(&nd_rd->nd_rd_dst); 2598 2599 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2600 2601 nd_rd->nd_rd_cksum = 0; 2602 nd_rd->nd_rd_cksum 2603 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2604 2605 /* send the packet to outside... */ 2606 if (ip6_output(m, NULL, NULL, 0, 2607 (struct ip6_moptions *)NULL, (struct socket *)NULL, NULL) != 0) 2608 icmp6_ifstat_inc(ifp, ifs6_out_error); 2609 2610 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2611 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2612 icmp6stat.icp6s_outhist[ND_REDIRECT]++; 2613 2614 return; 2615 2616 fail: 2617 if (m) 2618 m_freem(m); 2619 if (m0) 2620 m_freem(m0); 2621 } 2622 2623 /* 2624 * ICMPv6 socket option processing. 2625 */ 2626 int 2627 icmp6_ctloutput(op, so, level, optname, mp) 2628 int op; 2629 struct socket *so; 2630 int level, optname; 2631 struct mbuf **mp; 2632 { 2633 int error = 0; 2634 int optlen; 2635 struct in6pcb *in6p = sotoin6pcb(so); 2636 struct mbuf *m = *mp; 2637 2638 optlen = m ? m->m_len : 0; 2639 2640 if (level != IPPROTO_ICMPV6) { 2641 if (op == PRCO_SETOPT && m) 2642 (void)m_free(m); 2643 return EINVAL; 2644 } 2645 2646 switch (op) { 2647 case PRCO_SETOPT: 2648 switch (optname) { 2649 case ICMP6_FILTER: 2650 { 2651 struct icmp6_filter *p; 2652 2653 if (optlen != sizeof(*p)) { 2654 error = EMSGSIZE; 2655 break; 2656 } 2657 p = mtod(m, struct icmp6_filter *); 2658 if (!p || !in6p->in6p_icmp6filt) { 2659 error = EINVAL; 2660 break; 2661 } 2662 bcopy(p, in6p->in6p_icmp6filt, 2663 sizeof(struct icmp6_filter)); 2664 error = 0; 2665 break; 2666 } 2667 2668 default: 2669 error = ENOPROTOOPT; 2670 break; 2671 } 2672 if (m) 2673 (void)m_freem(m); 2674 break; 2675 2676 case PRCO_GETOPT: 2677 switch (optname) { 2678 case ICMP6_FILTER: 2679 { 2680 struct icmp6_filter *p; 2681 2682 if (!in6p->in6p_icmp6filt) { 2683 error = EINVAL; 2684 break; 2685 } 2686 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2687 m->m_len = sizeof(struct icmp6_filter); 2688 p = mtod(m, struct icmp6_filter *); 2689 bcopy(in6p->in6p_icmp6filt, p, 2690 sizeof(struct icmp6_filter)); 2691 error = 0; 2692 break; 2693 } 2694 2695 default: 2696 error = ENOPROTOOPT; 2697 break; 2698 } 2699 break; 2700 } 2701 2702 return (error); 2703 } 2704 2705 /* 2706 * Perform rate limit check. 2707 * Returns 0 if it is okay to send the icmp6 packet. 2708 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2709 * limitation. 2710 * 2711 * XXX per-destination/type check necessary? 2712 */ 2713 static int 2714 icmp6_ratelimit(dst, type, code) 2715 const struct in6_addr *dst; /* not used at this moment */ 2716 const int type; /* not used at this moment */ 2717 const int code; /* not used at this moment */ 2718 { 2719 int ret; 2720 2721 ret = 0; /* okay to send */ 2722 2723 /* PPS limit */ 2724 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2725 icmp6errppslim)) { 2726 /* The packet is subject to rate limit */ 2727 ret++; 2728 } 2729 2730 return ret; 2731 } 2732 2733 static struct rtentry * 2734 icmp6_mtudisc_clone(dst) 2735 struct sockaddr *dst; 2736 { 2737 struct rtentry *rt; 2738 int error; 2739 2740 rt = rtalloc1(dst, 1); 2741 if (rt == 0) 2742 return NULL; 2743 2744 /* If we didn't get a host route, allocate one */ 2745 if ((rt->rt_flags & RTF_HOST) == 0) { 2746 struct rtentry *nrt; 2747 2748 error = rtrequest((int) RTM_ADD, dst, 2749 (struct sockaddr *) rt->rt_gateway, 2750 (struct sockaddr *) 0, 2751 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2752 if (error) { 2753 rtfree(rt); 2754 return NULL; 2755 } 2756 nrt->rt_rmx = rt->rt_rmx; 2757 rtfree(rt); 2758 rt = nrt; 2759 } 2760 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2761 icmp6_mtudisc_timeout_q); 2762 if (error) { 2763 rtfree(rt); 2764 return NULL; 2765 } 2766 2767 return rt; /* caller need to call rtfree() */ 2768 } 2769 2770 static void 2771 icmp6_mtudisc_timeout(rt, r) 2772 struct rtentry *rt; 2773 struct rttimer *r; 2774 { 2775 if (rt == NULL) 2776 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2777 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2778 (RTF_DYNAMIC | RTF_HOST)) { 2779 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2780 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2781 } else { 2782 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2783 rt->rt_rmx.rmx_mtu = 0; 2784 } 2785 } 2786 2787 static void 2788 icmp6_redirect_timeout(rt, r) 2789 struct rtentry *rt; 2790 struct rttimer *r; 2791 { 2792 if (rt == NULL) 2793 panic("icmp6_redirect_timeout: bad route to timeout"); 2794 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2795 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2796 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2797 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2798 } 2799 } 2800 2801 /* 2802 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes. silly? 2803 */ 2804 static int 2805 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS) 2806 { 2807 2808 if (namelen != 0) 2809 return (EINVAL); 2810 2811 return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp, 2812 /*XXXUNCONST*/ 2813 __UNCONST(newp), newlen)); 2814 } 2815 2816 SYSCTL_SETUP(sysctl_net_inet6_icmp6_setup, 2817 "sysctl net.inet6.icmp6 subtree setup") 2818 { 2819 extern int nd6_maxqueuelen; /* defined in nd6.c */ 2820 2821 sysctl_createv(clog, 0, NULL, NULL, 2822 CTLFLAG_PERMANENT, 2823 CTLTYPE_NODE, "net", NULL, 2824 NULL, 0, NULL, 0, 2825 CTL_NET, CTL_EOL); 2826 sysctl_createv(clog, 0, NULL, NULL, 2827 CTLFLAG_PERMANENT, 2828 CTLTYPE_NODE, "inet6", NULL, 2829 NULL, 0, NULL, 0, 2830 CTL_NET, PF_INET6, CTL_EOL); 2831 sysctl_createv(clog, 0, NULL, NULL, 2832 CTLFLAG_PERMANENT, 2833 CTLTYPE_NODE, "icmp6", 2834 SYSCTL_DESCR("ICMPv6 related settings"), 2835 NULL, 0, NULL, 0, 2836 CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL); 2837 2838 sysctl_createv(clog, 0, NULL, NULL, 2839 CTLFLAG_PERMANENT, 2840 CTLTYPE_STRUCT, "stats", 2841 SYSCTL_DESCR("ICMPv6 transmission statistics"), 2842 NULL, 0, &icmp6stat, sizeof(icmp6stat), 2843 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2844 ICMPV6CTL_STATS, CTL_EOL); 2845 sysctl_createv(clog, 0, NULL, NULL, 2846 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2847 CTLTYPE_INT, "rediraccept", 2848 SYSCTL_DESCR("Accept and process redirect messages"), 2849 NULL, 0, &icmp6_rediraccept, 0, 2850 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2851 ICMPV6CTL_REDIRACCEPT, CTL_EOL); 2852 sysctl_createv(clog, 0, NULL, NULL, 2853 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2854 CTLTYPE_INT, "redirtimeout", 2855 SYSCTL_DESCR("Redirect generated route lifetime"), 2856 NULL, 0, &icmp6_redirtimeout, 0, 2857 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2858 ICMPV6CTL_REDIRTIMEOUT, CTL_EOL); 2859 #if 0 /* obsoleted */ 2860 sysctl_createv(clog, 0, NULL, NULL, 2861 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2862 CTLTYPE_INT, "errratelimit", NULL, 2863 NULL, 0, &icmp6_errratelimit, 0, 2864 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2865 ICMPV6CTL_ERRRATELIMIT, CTL_EOL); 2866 #endif 2867 sysctl_createv(clog, 0, NULL, NULL, 2868 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2869 CTLTYPE_INT, "nd6_prune", 2870 SYSCTL_DESCR("Neighbor discovery prune interval"), 2871 NULL, 0, &nd6_prune, 0, 2872 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2873 ICMPV6CTL_ND6_PRUNE, CTL_EOL); 2874 sysctl_createv(clog, 0, NULL, NULL, 2875 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2876 CTLTYPE_INT, "nd6_delay", 2877 SYSCTL_DESCR("First probe delay time"), 2878 NULL, 0, &nd6_delay, 0, 2879 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2880 ICMPV6CTL_ND6_DELAY, CTL_EOL); 2881 sysctl_createv(clog, 0, NULL, NULL, 2882 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2883 CTLTYPE_INT, "nd6_umaxtries", 2884 SYSCTL_DESCR("Number of unicast discovery attempts"), 2885 NULL, 0, &nd6_umaxtries, 0, 2886 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2887 ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL); 2888 sysctl_createv(clog, 0, NULL, NULL, 2889 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2890 CTLTYPE_INT, "nd6_mmaxtries", 2891 SYSCTL_DESCR("Number of multicast discovery attempts"), 2892 NULL, 0, &nd6_mmaxtries, 0, 2893 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2894 ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL); 2895 sysctl_createv(clog, 0, NULL, NULL, 2896 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2897 CTLTYPE_INT, "nd6_useloopback", 2898 SYSCTL_DESCR("Use loopback interface for local traffic"), 2899 NULL, 0, &nd6_useloopback, 0, 2900 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2901 ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL); 2902 #if 0 /* obsoleted */ 2903 sysctl_createv(clog, 0, NULL, NULL, 2904 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2905 CTLTYPE_INT, "nd6_proxyall", NULL, 2906 NULL, 0, &nd6_proxyall, 0, 2907 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2908 ICMPV6CTL_ND6_PROXYALL, CTL_EOL); 2909 #endif 2910 sysctl_createv(clog, 0, NULL, NULL, 2911 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2912 CTLTYPE_INT, "nodeinfo", 2913 SYSCTL_DESCR("Respond to node information requests"), 2914 NULL, 0, &icmp6_nodeinfo, 0, 2915 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2916 ICMPV6CTL_NODEINFO, CTL_EOL); 2917 sysctl_createv(clog, 0, NULL, NULL, 2918 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2919 CTLTYPE_INT, "errppslimit", 2920 SYSCTL_DESCR("Maximum ICMP errors sent per second"), 2921 NULL, 0, &icmp6errppslim, 0, 2922 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2923 ICMPV6CTL_ERRPPSLIMIT, CTL_EOL); 2924 sysctl_createv(clog, 0, NULL, NULL, 2925 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2926 CTLTYPE_INT, "nd6_maxnudhint", 2927 SYSCTL_DESCR("Maximum neighbor unreachable hint count"), 2928 NULL, 0, &nd6_maxnudhint, 0, 2929 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2930 ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL); 2931 sysctl_createv(clog, 0, NULL, NULL, 2932 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2933 CTLTYPE_INT, "mtudisc_hiwat", 2934 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2935 NULL, 0, &icmp6_mtudisc_hiwat, 0, 2936 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2937 ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL); 2938 sysctl_createv(clog, 0, NULL, NULL, 2939 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2940 CTLTYPE_INT, "mtudisc_lowat", 2941 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2942 NULL, 0, &icmp6_mtudisc_lowat, 0, 2943 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2944 ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL); 2945 sysctl_createv(clog, 0, NULL, NULL, 2946 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2947 CTLTYPE_INT, "nd6_debug", 2948 SYSCTL_DESCR("Enable neighbor discovery debug output"), 2949 NULL, 0, &nd6_debug, 0, 2950 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2951 ICMPV6CTL_ND6_DEBUG, CTL_EOL); 2952 sysctl_createv(clog, 0, NULL, NULL, 2953 CTLFLAG_PERMANENT, 2954 CTLTYPE_STRUCT, "nd6_drlist", 2955 SYSCTL_DESCR("Default router list"), 2956 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2957 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2958 ICMPV6CTL_ND6_DRLIST, CTL_EOL); 2959 sysctl_createv(clog, 0, NULL, NULL, 2960 CTLFLAG_PERMANENT, 2961 CTLTYPE_STRUCT, "nd6_prlist", 2962 SYSCTL_DESCR("Prefix list"), 2963 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2964 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2965 ICMPV6CTL_ND6_PRLIST, CTL_EOL); 2966 sysctl_createv(clog, 0, NULL, NULL, 2967 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2968 CTLTYPE_INT, "maxqueuelen", 2969 SYSCTL_DESCR("max packet queue len for a unresolved ND"), 2970 NULL, 1, &nd6_maxqueuelen, 0, 2971 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2972 ICMPV6CTL_ND6_MAXQLEN, CTL_EOL); 2973 } 2974
Indexes created Tue Oct 28 12:10:06 GMT 2025