1 /* $NetBSD: icmp6.c,v 1.156 2010/09/12 16:04:57 drochner 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 1.156 2010/09/12 16:04:57 drochner 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 <netinet6/ip6_private.h> 93 #include <netinet/icmp6.h> 94 #include <netinet6/icmp6_private.h> 95 #include <netinet6/mld6_var.h> 96 #include <netinet6/in6_pcb.h> 97 #include <netinet6/nd6.h> 98 #include <netinet6/in6_ifattach.h> 99 #include <netinet6/ip6protosw.h> 100 #include <netinet6/scope6_var.h> 101 102 #ifdef IPSEC 103 #include <netinet6/ipsec.h> 104 #include <netkey/key.h> 105 #endif 106 107 #ifdef FAST_IPSEC 108 #include <netipsec/ipsec.h> 109 #include <netipsec/key.h> 110 #endif 111 112 113 #include "faith.h" 114 #if defined(NFAITH) && 0 < NFAITH 115 #include <net/if_faith.h> 116 #endif 117 118 #include <net/net_osdep.h> 119 120 extern struct domain inet6domain; 121 122 percpu_t *icmp6stat_percpu; 123 124 extern struct inpcbtable raw6cbtable; 125 extern int icmp6errppslim; 126 static int icmp6errpps_count = 0; 127 static struct timeval icmp6errppslim_last; 128 extern int icmp6_nodeinfo; 129 130 /* 131 * List of callbacks to notify when Path MTU changes are made. 132 */ 133 struct icmp6_mtudisc_callback { 134 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 135 void (*mc_func)(struct in6_addr *); 136 }; 137 138 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 139 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 140 141 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 142 extern int pmtu_expire; 143 144 /* XXX do these values make any sense? */ 145 static int icmp6_mtudisc_hiwat = 1280; 146 static int icmp6_mtudisc_lowat = 256; 147 148 /* 149 * keep track of # of redirect routes. 150 */ 151 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 152 153 /* XXX experimental, turned off */ 154 static int icmp6_redirect_hiwat = -1; 155 static int icmp6_redirect_lowat = -1; 156 157 static void icmp6_errcount(u_int, int, int); 158 static int icmp6_rip6_input(struct mbuf **, int); 159 static int icmp6_ratelimit(const struct in6_addr *, const int, const int); 160 static const char *icmp6_redirect_diag(struct in6_addr *, 161 struct in6_addr *, struct in6_addr *); 162 static struct mbuf *ni6_input(struct mbuf *, int); 163 static struct mbuf *ni6_nametodns(const char *, int, int); 164 static int ni6_dnsmatch(const char *, int, const char *, int); 165 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *, 166 struct ifnet **, char *); 167 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 168 struct ifnet *, int); 169 static int icmp6_notify_error(struct mbuf *, int, int, int); 170 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *); 171 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *); 172 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *); 173 static void sysctl_net_inet6_icmp6_setup(struct sysctllog **); 174 175 176 void 177 icmp6_init(void) 178 { 179 180 sysctl_net_inet6_icmp6_setup(NULL); 181 mld_init(); 182 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 183 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 184 185 icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS); 186 } 187 188 static void 189 icmp6_errcount(u_int base, int type, int code) 190 { 191 switch (type) { 192 case ICMP6_DST_UNREACH: 193 switch (code) { 194 case ICMP6_DST_UNREACH_NOROUTE: 195 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE); 196 return; 197 case ICMP6_DST_UNREACH_ADMIN: 198 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN); 199 return; 200 case ICMP6_DST_UNREACH_BEYONDSCOPE: 201 ICMP6_STATINC(base + 202 ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE); 203 return; 204 case ICMP6_DST_UNREACH_ADDR: 205 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR); 206 return; 207 case ICMP6_DST_UNREACH_NOPORT: 208 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT); 209 return; 210 } 211 break; 212 case ICMP6_PACKET_TOO_BIG: 213 ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG); 214 return; 215 case ICMP6_TIME_EXCEEDED: 216 switch (code) { 217 case ICMP6_TIME_EXCEED_TRANSIT: 218 ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT); 219 return; 220 case ICMP6_TIME_EXCEED_REASSEMBLY: 221 ICMP6_STATINC(base + 222 ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY); 223 return; 224 } 225 break; 226 case ICMP6_PARAM_PROB: 227 switch (code) { 228 case ICMP6_PARAMPROB_HEADER: 229 ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER); 230 return; 231 case ICMP6_PARAMPROB_NEXTHEADER: 232 ICMP6_STATINC(base + 233 ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER); 234 return; 235 case ICMP6_PARAMPROB_OPTION: 236 ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION); 237 return; 238 } 239 break; 240 case ND_REDIRECT: 241 ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT); 242 return; 243 } 244 ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN); 245 } 246 247 /* 248 * Register a Path MTU Discovery callback. 249 */ 250 void 251 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *)) 252 { 253 struct icmp6_mtudisc_callback *mc; 254 255 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 256 mc = LIST_NEXT(mc, mc_list)) { 257 if (mc->mc_func == func) 258 return; 259 } 260 261 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 262 if (mc == NULL) 263 panic("icmp6_mtudisc_callback_register"); 264 265 mc->mc_func = func; 266 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 267 } 268 269 /* 270 * A wrapper function for icmp6_error() necessary when the erroneous packet 271 * may not contain enough scope zone information. 272 */ 273 void 274 icmp6_error2(struct mbuf *m, int type, int code, int param, 275 struct ifnet *ifp) 276 { 277 struct ip6_hdr *ip6; 278 279 if (ifp == NULL) 280 return; 281 282 if (m->m_len < sizeof(struct ip6_hdr)) { 283 m = m_pullup(m, sizeof(struct ip6_hdr)); 284 if (m == NULL) 285 return; 286 } 287 288 ip6 = mtod(m, struct ip6_hdr *); 289 290 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 291 return; 292 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 293 return; 294 295 icmp6_error(m, type, code, param); 296 } 297 298 /* 299 * Generate an error packet of type error in response to bad IP6 packet. 300 */ 301 void 302 icmp6_error(struct mbuf *m, int type, int code, int param) 303 { 304 struct ip6_hdr *oip6, *nip6; 305 struct icmp6_hdr *icmp6; 306 u_int preplen; 307 int off; 308 int nxt; 309 310 ICMP6_STATINC(ICMP6_STAT_ERROR); 311 312 /* count per-type-code statistics */ 313 icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code); 314 315 if (m->m_flags & M_DECRYPTED) { 316 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 317 goto freeit; 318 } 319 320 if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) && 321 (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) 322 return; 323 oip6 = mtod(m, struct ip6_hdr *); 324 325 /* 326 * If the destination address of the erroneous packet is a multicast 327 * address, or the packet was sent using link-layer multicast, 328 * we should basically suppress sending an error (RFC 2463, Section 329 * 2.4). 330 * We have two exceptions (the item e.2 in that section): 331 * - the Pakcet Too Big message can be sent for path MTU discovery. 332 * - the Parameter Problem Message that can be allowed an icmp6 error 333 * in the option type field. This check has been done in 334 * ip6_unknown_opt(), so we can just check the type and code. 335 */ 336 if ((m->m_flags & (M_BCAST|M_MCAST) || 337 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 338 (type != ICMP6_PACKET_TOO_BIG && 339 (type != ICMP6_PARAM_PROB || 340 code != ICMP6_PARAMPROB_OPTION))) 341 goto freeit; 342 343 /* 344 * RFC 2463, 2.4 (e.5): source address check. 345 * XXX: the case of anycast source? 346 */ 347 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 348 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 349 goto freeit; 350 351 /* 352 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 353 * don't do it. 354 */ 355 nxt = -1; 356 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 357 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 358 struct icmp6_hdr *icp; 359 360 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 361 sizeof(*icp)); 362 if (icp == NULL) { 363 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 364 return; 365 } 366 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 367 icp->icmp6_type == ND_REDIRECT) { 368 /* 369 * ICMPv6 error 370 * Special case: for redirect (which is 371 * informational) we must not send icmp6 error. 372 */ 373 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 374 goto freeit; 375 } else { 376 /* ICMPv6 informational - send the error */ 377 } 378 } 379 #if 0 /* controversial */ 380 else if (off >= 0 && nxt == IPPROTO_ESP) { 381 /* 382 * It could be ICMPv6 error inside ESP. Take a safer side, 383 * don't respond. 384 */ 385 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 386 goto freeit; 387 } 388 #endif 389 else { 390 /* non-ICMPv6 - send the error */ 391 } 392 393 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 394 395 /* Finally, do rate limitation check. */ 396 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 397 ICMP6_STATINC(ICMP6_STAT_TOOFREQ); 398 goto freeit; 399 } 400 401 /* 402 * OK, ICMP6 can be generated. 403 */ 404 405 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 406 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 407 408 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 409 M_PREPEND(m, preplen, M_DONTWAIT); 410 if (m && M_UNWRITABLE(m, preplen)) 411 m = m_pullup(m, preplen); 412 if (m == NULL) { 413 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 414 return; 415 } 416 417 nip6 = mtod(m, struct ip6_hdr *); 418 nip6->ip6_src = oip6->ip6_src; 419 nip6->ip6_dst = oip6->ip6_dst; 420 421 in6_clearscope(&oip6->ip6_src); 422 in6_clearscope(&oip6->ip6_dst); 423 424 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 425 icmp6->icmp6_type = type; 426 icmp6->icmp6_code = code; 427 icmp6->icmp6_pptr = htonl((u_int32_t)param); 428 429 /* 430 * icmp6_reflect() is designed to be in the input path. 431 * icmp6_error() can be called from both input and output path, 432 * and if we are in output path rcvif could contain bogus value. 433 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 434 * information in ip header (nip6). 435 */ 436 m->m_pkthdr.rcvif = NULL; 437 438 ICMP6_STATINC(ICMP6_STAT_OUTHIST + type); 439 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 440 441 return; 442 443 freeit: 444 /* 445 * If we can't tell whether or not we can generate ICMP6, free it. 446 */ 447 m_freem(m); 448 } 449 450 /* 451 * Process a received ICMP6 message. 452 */ 453 int 454 icmp6_input(struct mbuf **mp, int *offp, int proto) 455 { 456 struct mbuf *m = *mp, *n; 457 struct ip6_hdr *ip6, *nip6; 458 struct icmp6_hdr *icmp6, *nicmp6; 459 int off = *offp; 460 int icmp6len = m->m_pkthdr.len - *offp; 461 int code, sum, noff, i; 462 463 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4) 464 KASSERT(ICMP6_MAXLEN < MCLBYTES); 465 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 466 467 /* 468 * Locate icmp6 structure in mbuf, and check 469 * that not corrupted and of at least minimum length 470 */ 471 472 if (icmp6len < sizeof(struct icmp6_hdr)) { 473 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 474 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 475 goto freeit; 476 } 477 478 i = off + sizeof(*icmp6); 479 if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) { 480 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 481 #if 0 /* m is 0 here */ 482 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 483 #endif 484 goto freeit; 485 } 486 ip6 = mtod(m, struct ip6_hdr *); 487 /* 488 * calculate the checksum 489 */ 490 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 491 if (icmp6 == NULL) { 492 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 493 /* m is invalid */ 494 /*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/ 495 return IPPROTO_DONE; 496 } 497 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 498 code = icmp6->icmp6_code; 499 500 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 501 nd6log((LOG_ERR, 502 "ICMP6 checksum error(%d|%x) %s\n", 503 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 504 ICMP6_STATINC(ICMP6_STAT_CHECKSUM); 505 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 506 goto freeit; 507 } 508 509 #if defined(NFAITH) && 0 < NFAITH 510 if (faithprefix(&ip6->ip6_dst)) { 511 /* 512 * Deliver very specific ICMP6 type only. 513 * This is important to deliver TOOBIG. Otherwise PMTUD 514 * will not work. 515 */ 516 switch (icmp6->icmp6_type) { 517 case ICMP6_DST_UNREACH: 518 case ICMP6_PACKET_TOO_BIG: 519 case ICMP6_TIME_EXCEEDED: 520 break; 521 default: 522 goto freeit; 523 } 524 } 525 #endif 526 527 ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type); 528 529 switch (icmp6->icmp6_type) { 530 case ICMP6_DST_UNREACH: 531 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 532 switch (code) { 533 case ICMP6_DST_UNREACH_NOROUTE: 534 code = PRC_UNREACH_NET; 535 break; 536 case ICMP6_DST_UNREACH_ADMIN: 537 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 538 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 539 break; 540 case ICMP6_DST_UNREACH_ADDR: 541 code = PRC_HOSTDEAD; 542 break; 543 #ifdef COMPAT_RFC1885 544 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 545 code = PRC_UNREACH_SRCFAIL; 546 break; 547 #else 548 case ICMP6_DST_UNREACH_BEYONDSCOPE: 549 /* I mean "source address was incorrect." */ 550 code = PRC_UNREACH_NET; 551 break; 552 #endif 553 case ICMP6_DST_UNREACH_NOPORT: 554 code = PRC_UNREACH_PORT; 555 break; 556 default: 557 goto badcode; 558 } 559 goto deliver; 560 561 case ICMP6_PACKET_TOO_BIG: 562 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 563 564 /* 565 * MTU is checked in icmp6_mtudisc. 566 */ 567 code = PRC_MSGSIZE; 568 569 /* 570 * Updating the path MTU will be done after examining 571 * intermediate extension headers. 572 */ 573 goto deliver; 574 575 case ICMP6_TIME_EXCEEDED: 576 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 577 switch (code) { 578 case ICMP6_TIME_EXCEED_TRANSIT: 579 code = PRC_TIMXCEED_INTRANS; 580 break; 581 case ICMP6_TIME_EXCEED_REASSEMBLY: 582 code = PRC_TIMXCEED_REASS; 583 break; 584 default: 585 goto badcode; 586 } 587 goto deliver; 588 589 case ICMP6_PARAM_PROB: 590 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 591 switch (code) { 592 case ICMP6_PARAMPROB_NEXTHEADER: 593 code = PRC_UNREACH_PROTOCOL; 594 break; 595 case ICMP6_PARAMPROB_HEADER: 596 case ICMP6_PARAMPROB_OPTION: 597 code = PRC_PARAMPROB; 598 break; 599 default: 600 goto badcode; 601 } 602 goto deliver; 603 604 case ICMP6_ECHO_REQUEST: 605 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 606 if (code != 0) 607 goto badcode; 608 /* 609 * Copy mbuf to send to two data paths: userland socket(s), 610 * and to the querier (echo reply). 611 * m: a copy for socket, n: a copy for querier 612 * 613 * If the first mbuf is shared, or the first mbuf is too short, 614 * copy the first part of the data into a fresh mbuf. 615 * Otherwise, we will wrongly overwrite both copies. 616 */ 617 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 618 /* Give up local */ 619 n = m; 620 m = NULL; 621 } else if (M_READONLY(n) || 622 n->m_len < off + sizeof(struct icmp6_hdr)) { 623 struct mbuf *n0 = n; 624 625 /* 626 * Prepare an internal mbuf. m_pullup() doesn't 627 * always copy the length we specified. 628 */ 629 if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 630 /* Give up local */ 631 n = m; 632 m = NULL; 633 } 634 m_freem(n0); 635 } 636 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, 637 sizeof(*nicmp6)); 638 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 639 nicmp6->icmp6_code = 0; 640 if (n) { 641 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 642 icmp6s[ICMP6_STAT_REFLECT]++; 643 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++; 644 ICMP6_STAT_PUTREF(); 645 icmp6_reflect(n, off); 646 } 647 if (!m) 648 goto freeit; 649 break; 650 651 case ICMP6_ECHO_REPLY: 652 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 653 if (code != 0) 654 goto badcode; 655 break; 656 657 case MLD_LISTENER_QUERY: 658 case MLD_LISTENER_REPORT: 659 if (icmp6len < sizeof(struct mld_hdr)) 660 goto badlen; 661 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */ 662 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 663 else 664 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 665 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 666 /* give up local */ 667 mld_input(m, off); 668 m = NULL; 669 goto freeit; 670 } 671 mld_input(n, off); 672 /* m stays. */ 673 break; 674 675 case MLD_LISTENER_DONE: 676 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 677 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */ 678 goto badlen; 679 break; /* nothing to be done in kernel */ 680 681 case MLD_MTRACE_RESP: 682 case MLD_MTRACE: 683 /* XXX: these two are experimental. not officially defined. */ 684 /* XXX: per-interface statistics? */ 685 break; /* just pass it to applications */ 686 687 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 688 { 689 enum { WRU, FQDN } mode; 690 691 if (!icmp6_nodeinfo) 692 break; 693 694 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 695 mode = WRU; 696 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 697 mode = FQDN; 698 else 699 goto badlen; 700 701 if (mode == FQDN) { 702 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 703 if (n) 704 n = ni6_input(n, off); 705 /* XXX meaningless if n == NULL */ 706 noff = sizeof(struct ip6_hdr); 707 } else { 708 u_char *p; 709 int maxhlen; 710 711 if ((icmp6_nodeinfo & 5) != 5) 712 break; 713 714 if (code != 0) 715 goto badcode; 716 MGETHDR(n, M_DONTWAIT, m->m_type); 717 if (n && ICMP6_MAXLEN > MHLEN) { 718 MCLGET(n, M_DONTWAIT); 719 if ((n->m_flags & M_EXT) == 0) { 720 m_free(n); 721 n = NULL; 722 } 723 } 724 if (n == NULL) { 725 /* Give up remote */ 726 break; 727 } 728 n->m_pkthdr.rcvif = NULL; 729 n->m_len = 0; 730 maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN; 731 if (maxhlen > hostnamelen) 732 maxhlen = hostnamelen; 733 /* 734 * Copy IPv6 and ICMPv6 only. 735 */ 736 nip6 = mtod(n, struct ip6_hdr *); 737 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 738 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 739 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 740 p = (u_char *)(nicmp6 + 1); 741 memset(p, 0, 4); 742 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 743 noff = sizeof(struct ip6_hdr); 744 M_COPY_PKTHDR(n, m); /* just for rcvif */ 745 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 746 sizeof(struct icmp6_hdr) + 4 + maxhlen; 747 nicmp6->icmp6_type = ICMP6_WRUREPLY; 748 nicmp6->icmp6_code = 0; 749 } 750 #undef hostnamelen 751 if (n) { 752 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 753 icmp6s[ICMP6_STAT_REFLECT]++; 754 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++; 755 ICMP6_STAT_PUTREF(); 756 icmp6_reflect(n, noff); 757 } 758 break; 759 } 760 761 case ICMP6_WRUREPLY: 762 if (code != 0) 763 goto badcode; 764 break; 765 766 case ND_ROUTER_SOLICIT: 767 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 768 if (code != 0) 769 goto badcode; 770 if (icmp6len < sizeof(struct nd_router_solicit)) 771 goto badlen; 772 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 773 /* give up local */ 774 nd6_rs_input(m, off, icmp6len); 775 m = NULL; 776 goto freeit; 777 } 778 nd6_rs_input(n, off, icmp6len); 779 /* m stays. */ 780 break; 781 782 case ND_ROUTER_ADVERT: 783 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 784 if (code != 0) 785 goto badcode; 786 if (icmp6len < sizeof(struct nd_router_advert)) 787 goto badlen; 788 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 789 /* give up local */ 790 nd6_ra_input(m, off, icmp6len); 791 m = NULL; 792 goto freeit; 793 } 794 nd6_ra_input(n, off, icmp6len); 795 /* m stays. */ 796 break; 797 798 case ND_NEIGHBOR_SOLICIT: 799 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 800 if (code != 0) 801 goto badcode; 802 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 803 goto badlen; 804 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 805 /* give up local */ 806 nd6_ns_input(m, off, icmp6len); 807 m = NULL; 808 goto freeit; 809 } 810 nd6_ns_input(n, off, icmp6len); 811 /* m stays. */ 812 break; 813 814 case ND_NEIGHBOR_ADVERT: 815 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 816 if (code != 0) 817 goto badcode; 818 if (icmp6len < sizeof(struct nd_neighbor_advert)) 819 goto badlen; 820 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 821 /* give up local */ 822 nd6_na_input(m, off, icmp6len); 823 m = NULL; 824 goto freeit; 825 } 826 nd6_na_input(n, off, icmp6len); 827 /* m stays. */ 828 break; 829 830 case ND_REDIRECT: 831 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 832 if (code != 0) 833 goto badcode; 834 if (icmp6len < sizeof(struct nd_redirect)) 835 goto badlen; 836 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 837 /* give up local */ 838 icmp6_redirect_input(m, off); 839 m = NULL; 840 goto freeit; 841 } 842 icmp6_redirect_input(n, off); 843 /* m stays. */ 844 break; 845 846 case ICMP6_ROUTER_RENUMBERING: 847 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 848 code != ICMP6_ROUTER_RENUMBERING_RESULT) 849 goto badcode; 850 if (icmp6len < sizeof(struct icmp6_router_renum)) 851 goto badlen; 852 break; 853 854 default: 855 nd6log((LOG_DEBUG, 856 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 857 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 858 ip6_sprintf(&ip6->ip6_dst), 859 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 860 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 861 /* ICMPv6 error: MUST deliver it by spec... */ 862 code = PRC_NCMDS; 863 /* deliver */ 864 } else { 865 /* ICMPv6 informational: MUST not deliver */ 866 break; 867 } 868 deliver: 869 if (icmp6_notify_error(m, off, icmp6len, code)) { 870 /* In this case, m should've been freed. */ 871 return (IPPROTO_DONE); 872 } 873 break; 874 875 badcode: 876 ICMP6_STATINC(ICMP6_STAT_BADCODE); 877 break; 878 879 badlen: 880 ICMP6_STATINC(ICMP6_STAT_BADLEN); 881 break; 882 } 883 884 /* deliver the packet to appropriate sockets */ 885 icmp6_rip6_input(&m, *offp); 886 887 return IPPROTO_DONE; 888 889 freeit: 890 m_freem(m); 891 return IPPROTO_DONE; 892 } 893 894 static int 895 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code) 896 { 897 struct icmp6_hdr *icmp6; 898 struct ip6_hdr *eip6; 899 u_int32_t notifymtu; 900 struct sockaddr_in6 icmp6src, icmp6dst; 901 902 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 903 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 904 goto freeit; 905 } 906 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 907 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 908 if (icmp6 == NULL) { 909 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 910 return (-1); 911 } 912 eip6 = (struct ip6_hdr *)(icmp6 + 1); 913 914 /* Detect the upper level protocol */ 915 { 916 void (*ctlfunc)(int, struct sockaddr *, void *); 917 u_int8_t nxt = eip6->ip6_nxt; 918 int eoff = off + sizeof(struct icmp6_hdr) + 919 sizeof(struct ip6_hdr); 920 struct ip6ctlparam ip6cp; 921 struct in6_addr *finaldst = NULL; 922 int icmp6type = icmp6->icmp6_type; 923 struct ip6_frag *fh; 924 struct ip6_rthdr *rth; 925 struct ip6_rthdr0 *rth0; 926 int rthlen; 927 928 while (1) { /* XXX: should avoid infinite loop explicitly? */ 929 struct ip6_ext *eh; 930 931 switch (nxt) { 932 case IPPROTO_HOPOPTS: 933 case IPPROTO_DSTOPTS: 934 case IPPROTO_AH: 935 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 936 eoff, sizeof(*eh)); 937 if (eh == NULL) { 938 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 939 return (-1); 940 } 941 942 if (nxt == IPPROTO_AH) 943 eoff += (eh->ip6e_len + 2) << 2; 944 else 945 eoff += (eh->ip6e_len + 1) << 3; 946 nxt = eh->ip6e_nxt; 947 break; 948 case IPPROTO_ROUTING: 949 /* 950 * When the erroneous packet contains a 951 * routing header, we should examine the 952 * header to determine the final destination. 953 * Otherwise, we can't properly update 954 * information that depends on the final 955 * destination (e.g. path MTU). 956 */ 957 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 958 eoff, sizeof(*rth)); 959 if (rth == NULL) { 960 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 961 return (-1); 962 } 963 rthlen = (rth->ip6r_len + 1) << 3; 964 /* 965 * XXX: currently there is no 966 * officially defined type other 967 * than type-0. 968 * Note that if the segment left field 969 * is 0, all intermediate hops must 970 * have been passed. 971 */ 972 if (rth->ip6r_segleft && 973 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 974 int hops; 975 976 IP6_EXTHDR_GET(rth0, 977 struct ip6_rthdr0 *, m, 978 eoff, rthlen); 979 if (rth0 == NULL) { 980 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 981 return (-1); 982 } 983 /* just ignore a bogus header */ 984 if ((rth0->ip6r0_len % 2) == 0 && 985 (hops = rth0->ip6r0_len/2)) 986 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 987 } 988 eoff += rthlen; 989 nxt = rth->ip6r_nxt; 990 break; 991 case IPPROTO_FRAGMENT: 992 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 993 eoff, sizeof(*fh)); 994 if (fh == NULL) { 995 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 996 return (-1); 997 } 998 /* 999 * Data after a fragment header is meaningless 1000 * unless it is the first fragment, but 1001 * we'll go to the notify label for path MTU 1002 * discovery. 1003 */ 1004 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1005 goto notify; 1006 1007 eoff += sizeof(struct ip6_frag); 1008 nxt = fh->ip6f_nxt; 1009 break; 1010 default: 1011 /* 1012 * This case includes ESP and the No Next 1013 * Header. In such cases going to the notify 1014 * label does not have any meaning 1015 * (i.e. ctlfunc will be NULL), but we go 1016 * anyway since we might have to update 1017 * path MTU information. 1018 */ 1019 goto notify; 1020 } 1021 } 1022 notify: 1023 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1024 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1025 if (icmp6 == NULL) { 1026 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 1027 return (-1); 1028 } 1029 1030 /* 1031 * retrieve parameters from the inner IPv6 header, and convert 1032 * them into sockaddr structures. 1033 * XXX: there is no guarantee that the source or destination 1034 * addresses of the inner packet are in the same scope zone as 1035 * the addresses of the icmp packet. But there is no other 1036 * way to determine the zone. 1037 */ 1038 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1039 1040 sockaddr_in6_init(&icmp6dst, 1041 (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0); 1042 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1043 goto freeit; 1044 sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0); 1045 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1046 goto freeit; 1047 icmp6src.sin6_flowinfo = 1048 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1049 1050 if (finaldst == NULL) 1051 finaldst = &eip6->ip6_dst; 1052 ip6cp.ip6c_m = m; 1053 ip6cp.ip6c_icmp6 = icmp6; 1054 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1055 ip6cp.ip6c_off = eoff; 1056 ip6cp.ip6c_finaldst = finaldst; 1057 ip6cp.ip6c_src = &icmp6src; 1058 ip6cp.ip6c_nxt = nxt; 1059 1060 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1061 notifymtu = ntohl(icmp6->icmp6_mtu); 1062 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1063 } 1064 1065 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1066 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1067 if (ctlfunc) { 1068 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1069 &ip6cp); 1070 } 1071 } 1072 return (0); 1073 1074 freeit: 1075 m_freem(m); 1076 return (-1); 1077 } 1078 1079 void 1080 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1081 { 1082 unsigned long rtcount; 1083 struct icmp6_mtudisc_callback *mc; 1084 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1085 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1086 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1087 u_int mtu = ntohl(icmp6->icmp6_mtu); 1088 struct rtentry *rt = NULL; 1089 struct sockaddr_in6 sin6; 1090 1091 /* 1092 * The MTU should not be less than the minimal IPv6 MTU except for the 1093 * hack in ip6_output/ip6_setpmtu where we always include a frag header. 1094 * In that one case, the MTU might be less than 1280. 1095 */ 1096 if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) { 1097 /* is the mtu even sane? */ 1098 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) 1099 return; 1100 if (!validated) 1101 return; 1102 mtu = IPV6_MMTU - sizeof(struct ip6_frag); 1103 } 1104 1105 /* 1106 * allow non-validated cases if memory is plenty, to make traffic 1107 * from non-connected pcb happy. 1108 */ 1109 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1110 if (validated) { 1111 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1112 return; 1113 else if (0 <= icmp6_mtudisc_lowat && 1114 rtcount > icmp6_mtudisc_lowat) { 1115 /* 1116 * XXX nuke a victim, install the new one. 1117 */ 1118 } 1119 } else { 1120 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1121 return; 1122 } 1123 1124 memset(&sin6, 0, sizeof(sin6)); 1125 sin6.sin6_family = PF_INET6; 1126 sin6.sin6_len = sizeof(struct sockaddr_in6); 1127 sin6.sin6_addr = *dst; 1128 if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1129 return; 1130 1131 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1132 1133 if (rt && (rt->rt_flags & RTF_HOST) && 1134 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1135 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1136 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1137 ICMP6_STATINC(ICMP6_STAT_PMTUCHG); 1138 rt->rt_rmx.rmx_mtu = mtu; 1139 } 1140 } 1141 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */ 1142 RTFREE(rt); 1143 } 1144 1145 /* 1146 * Notify protocols that the MTU for this destination 1147 * has changed. 1148 */ 1149 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1150 mc = LIST_NEXT(mc, mc_list)) 1151 (*mc->mc_func)(&sin6.sin6_addr); 1152 } 1153 1154 /* 1155 * Process a Node Information Query packet, based on 1156 * draft-ietf-ipngwg-icmp-name-lookups-07. 1157 * 1158 * Spec incompatibilities: 1159 * - IPv6 Subject address handling 1160 * - IPv4 Subject address handling support missing 1161 * - Proxy reply (answer even if it's not for me) 1162 * - joins NI group address at in6_ifattach() time only, does not cope 1163 * with hostname changes by sethostname(3) 1164 */ 1165 #ifndef offsetof /* XXX */ 1166 #define offsetof(type, member) ((size_t)(&((type *)0)->member)) 1167 #endif 1168 static struct mbuf * 1169 ni6_input(struct mbuf *m, int off) 1170 { 1171 struct icmp6_nodeinfo *ni6, *nni6; 1172 struct mbuf *n = NULL; 1173 u_int16_t qtype; 1174 int subjlen; 1175 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1176 struct ni_reply_fqdn *fqdn; 1177 int addrs; /* for NI_QTYPE_NODEADDR */ 1178 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1179 struct sockaddr_in6 sin6; /* ip6_dst */ 1180 struct in6_addr in6_subj; /* subject address */ 1181 struct ip6_hdr *ip6; 1182 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1183 char *subj = NULL; 1184 1185 ip6 = mtod(m, struct ip6_hdr *); 1186 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1187 if (ni6 == NULL) { 1188 /* m is already reclaimed */ 1189 return NULL; 1190 } 1191 1192 /* 1193 * Validate IPv6 destination address. 1194 * 1195 * The Responder must discard the Query without further processing 1196 * unless it is one of the Responder's unicast or anycast addresses, or 1197 * a link-local scope multicast address which the Responder has joined. 1198 * [icmp-name-lookups-07, Section 4.] 1199 */ 1200 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 1201 /* XXX scopeid */ 1202 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1203 ; /* unicast/anycast, fine */ 1204 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1205 ; /* link-local multicast, fine */ 1206 else 1207 goto bad; 1208 1209 /* validate query Subject field. */ 1210 qtype = ntohs(ni6->ni_qtype); 1211 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1212 switch (qtype) { 1213 case NI_QTYPE_NOOP: 1214 case NI_QTYPE_SUPTYPES: 1215 /* 07 draft */ 1216 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1217 break; 1218 /* FALLTHROUGH */ 1219 case NI_QTYPE_FQDN: 1220 case NI_QTYPE_NODEADDR: 1221 case NI_QTYPE_IPV4ADDR: 1222 switch (ni6->ni_code) { 1223 case ICMP6_NI_SUBJ_IPV6: 1224 #if ICMP6_NI_SUBJ_IPV6 != 0 1225 case 0: 1226 #endif 1227 /* 1228 * backward compatibility - try to accept 03 draft 1229 * format, where no Subject is present. 1230 */ 1231 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1232 subjlen == 0) { 1233 oldfqdn++; 1234 break; 1235 } 1236 #if ICMP6_NI_SUBJ_IPV6 != 0 1237 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1238 goto bad; 1239 #endif 1240 1241 if (subjlen != sizeof(sin6.sin6_addr)) 1242 goto bad; 1243 1244 /* 1245 * Validate Subject address. 1246 * 1247 * Not sure what exactly "address belongs to the node" 1248 * means in the spec, is it just unicast, or what? 1249 * 1250 * At this moment we consider Subject address as 1251 * "belong to the node" if the Subject address equals 1252 * to the IPv6 destination address; validation for 1253 * IPv6 destination address should have done enough 1254 * check for us. 1255 * 1256 * We do not do proxy at this moment. 1257 */ 1258 /* m_pulldown instead of copy? */ 1259 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1260 subjlen, (void *)&in6_subj); 1261 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1262 goto bad; 1263 1264 subj = (char *)&in6_subj; 1265 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1266 break; 1267 1268 /* 1269 * XXX if we are to allow other cases, we should really 1270 * be careful about scope here. 1271 * basically, we should disallow queries toward IPv6 1272 * destination X with subject Y, if scope(X) > scope(Y). 1273 * if we allow scope(X) > scope(Y), it will result in 1274 * information leakage across scope boundary. 1275 */ 1276 goto bad; 1277 1278 case ICMP6_NI_SUBJ_FQDN: 1279 /* 1280 * Validate Subject name with gethostname(3). 1281 * 1282 * The behavior may need some debate, since: 1283 * - we are not sure if the node has FQDN as 1284 * hostname (returned by gethostname(3)). 1285 * - the code does wildcard match for truncated names. 1286 * however, we are not sure if we want to perform 1287 * wildcard match, if gethostname(3) side has 1288 * truncated hostname. 1289 */ 1290 n = ni6_nametodns(hostname, hostnamelen, 0); 1291 if (!n || n->m_next || n->m_len == 0) 1292 goto bad; 1293 IP6_EXTHDR_GET(subj, char *, m, 1294 off + sizeof(struct icmp6_nodeinfo), subjlen); 1295 if (subj == NULL) 1296 goto bad; 1297 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1298 n->m_len)) { 1299 goto bad; 1300 } 1301 m_freem(n); 1302 n = NULL; 1303 break; 1304 1305 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1306 default: 1307 goto bad; 1308 } 1309 break; 1310 } 1311 1312 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1313 switch (qtype) { 1314 case NI_QTYPE_FQDN: 1315 if ((icmp6_nodeinfo & 1) == 0) 1316 goto bad; 1317 break; 1318 case NI_QTYPE_NODEADDR: 1319 case NI_QTYPE_IPV4ADDR: 1320 if ((icmp6_nodeinfo & 2) == 0) 1321 goto bad; 1322 break; 1323 } 1324 1325 /* guess reply length */ 1326 switch (qtype) { 1327 case NI_QTYPE_NOOP: 1328 break; /* no reply data */ 1329 case NI_QTYPE_SUPTYPES: 1330 replylen += sizeof(u_int32_t); 1331 break; 1332 case NI_QTYPE_FQDN: 1333 /* XXX will append an mbuf */ 1334 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1335 break; 1336 case NI_QTYPE_NODEADDR: 1337 addrs = ni6_addrs(ni6, m, &ifp, subj); 1338 if ((replylen += addrs * (sizeof(struct in6_addr) + 1339 sizeof(u_int32_t))) > MCLBYTES) 1340 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1341 break; 1342 case NI_QTYPE_IPV4ADDR: 1343 /* unsupported - should respond with unknown Qtype? */ 1344 goto bad; 1345 default: 1346 /* 1347 * XXX: We must return a reply with the ICMP6 code 1348 * `unknown Qtype' in this case. However we regard the case 1349 * as an FQDN query for backward compatibility. 1350 * Older versions set a random value to this field, 1351 * so it rarely varies in the defined qtypes. 1352 * But the mechanism is not reliable... 1353 * maybe we should obsolete older versions. 1354 */ 1355 qtype = NI_QTYPE_FQDN; 1356 /* XXX will append an mbuf */ 1357 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1358 oldfqdn++; 1359 break; 1360 } 1361 1362 /* allocate an mbuf to reply. */ 1363 MGETHDR(n, M_DONTWAIT, m->m_type); 1364 if (n == NULL) { 1365 m_freem(m); 1366 return (NULL); 1367 } 1368 M_MOVE_PKTHDR(n, m); /* just for rcvif */ 1369 if (replylen > MHLEN) { 1370 if (replylen > MCLBYTES) { 1371 /* 1372 * XXX: should we try to allocate more? But MCLBYTES 1373 * is probably much larger than IPV6_MMTU... 1374 */ 1375 goto bad; 1376 } 1377 MCLGET(n, M_DONTWAIT); 1378 if ((n->m_flags & M_EXT) == 0) { 1379 goto bad; 1380 } 1381 } 1382 n->m_pkthdr.len = n->m_len = replylen; 1383 1384 /* copy mbuf header and IPv6 + Node Information base headers */ 1385 bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr)); 1386 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1387 bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo)); 1388 1389 /* qtype dependent procedure */ 1390 switch (qtype) { 1391 case NI_QTYPE_NOOP: 1392 nni6->ni_code = ICMP6_NI_SUCCESS; 1393 nni6->ni_flags = 0; 1394 break; 1395 case NI_QTYPE_SUPTYPES: 1396 { 1397 u_int32_t v; 1398 nni6->ni_code = ICMP6_NI_SUCCESS; 1399 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1400 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1401 v = (u_int32_t)htonl(0x0000000f); 1402 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1403 break; 1404 } 1405 case NI_QTYPE_FQDN: 1406 nni6->ni_code = ICMP6_NI_SUCCESS; 1407 fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) + 1408 sizeof(struct ip6_hdr) + 1409 sizeof(struct icmp6_nodeinfo)); 1410 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1411 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1412 /* 1413 * XXX do we really have FQDN in variable "hostname"? 1414 */ 1415 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1416 if (n->m_next == NULL) 1417 goto bad; 1418 /* XXX we assume that n->m_next is not a chain */ 1419 if (n->m_next->m_next != NULL) 1420 goto bad; 1421 n->m_pkthdr.len += n->m_next->m_len; 1422 break; 1423 case NI_QTYPE_NODEADDR: 1424 { 1425 int lenlim, copied; 1426 1427 nni6->ni_code = ICMP6_NI_SUCCESS; 1428 n->m_pkthdr.len = n->m_len = 1429 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1430 lenlim = M_TRAILINGSPACE(n); 1431 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1432 /* XXX: reset mbuf length */ 1433 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1434 sizeof(struct icmp6_nodeinfo) + copied; 1435 break; 1436 } 1437 default: 1438 break; /* XXX impossible! */ 1439 } 1440 1441 nni6->ni_type = ICMP6_NI_REPLY; 1442 m_freem(m); 1443 return (n); 1444 1445 bad: 1446 m_freem(m); 1447 if (n) 1448 m_freem(n); 1449 return (NULL); 1450 } 1451 #undef hostnamelen 1452 1453 #define isupper(x) ('A' <= (x) && (x) <= 'Z') 1454 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1455 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1456 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1457 1458 /* 1459 * make a mbuf with DNS-encoded string. no compression support. 1460 * 1461 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1462 * treated as truncated name (two \0 at the end). this is a wild guess. 1463 * 1464 * old - return pascal string if non-zero 1465 */ 1466 static struct mbuf * 1467 ni6_nametodns(const char *name, int namelen, int old) 1468 { 1469 struct mbuf *m; 1470 char *cp, *ep; 1471 const char *p, *q; 1472 int i, len, nterm; 1473 1474 if (old) 1475 len = namelen + 1; 1476 else 1477 len = MCLBYTES; 1478 1479 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1480 MGET(m, M_DONTWAIT, MT_DATA); 1481 if (m && len > MLEN) { 1482 MCLGET(m, M_DONTWAIT); 1483 if ((m->m_flags & M_EXT) == 0) 1484 goto fail; 1485 } 1486 if (!m) 1487 goto fail; 1488 m->m_next = NULL; 1489 1490 if (old) { 1491 m->m_len = len; 1492 *mtod(m, char *) = namelen; 1493 bcopy(name, mtod(m, char *) + 1, namelen); 1494 return m; 1495 } else { 1496 m->m_len = 0; 1497 cp = mtod(m, char *); 1498 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1499 1500 /* if not certain about my name, return empty buffer */ 1501 if (namelen == 0) 1502 return m; 1503 1504 /* 1505 * guess if it looks like shortened hostname, or FQDN. 1506 * shortened hostname needs two trailing "\0". 1507 */ 1508 i = 0; 1509 for (p = name; p < name + namelen; p++) { 1510 if (*p && *p == '.') 1511 i++; 1512 } 1513 if (i < 2) 1514 nterm = 2; 1515 else 1516 nterm = 1; 1517 1518 p = name; 1519 while (cp < ep && p < name + namelen) { 1520 i = 0; 1521 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1522 i++; 1523 /* result does not fit into mbuf */ 1524 if (cp + i + 1 >= ep) 1525 goto fail; 1526 /* 1527 * DNS label length restriction, RFC1035 page 8. 1528 * "i == 0" case is included here to avoid returning 1529 * 0-length label on "foo..bar". 1530 */ 1531 if (i <= 0 || i >= 64) 1532 goto fail; 1533 *cp++ = i; 1534 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1535 goto fail; 1536 while (i > 0) { 1537 if (!isalnum(*p) && *p != '-') 1538 goto fail; 1539 if (isupper(*p)) { 1540 *cp++ = tolower(*p); 1541 p++; 1542 } else 1543 *cp++ = *p++; 1544 i--; 1545 } 1546 p = q; 1547 if (p < name + namelen && *p == '.') 1548 p++; 1549 } 1550 /* termination */ 1551 if (cp + nterm >= ep) 1552 goto fail; 1553 while (nterm-- > 0) 1554 *cp++ = '\0'; 1555 m->m_len = cp - mtod(m, char *); 1556 return m; 1557 } 1558 1559 panic("should not reach here"); 1560 /* NOTREACHED */ 1561 1562 fail: 1563 if (m) 1564 m_freem(m); 1565 return NULL; 1566 } 1567 1568 /* 1569 * check if two DNS-encoded string matches. takes care of truncated 1570 * form (with \0\0 at the end). no compression support. 1571 * XXX upper/lowercase match (see RFC2065) 1572 */ 1573 static int 1574 ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1575 { 1576 const char *a0, *b0; 1577 int l; 1578 1579 /* simplest case - need validation? */ 1580 if (alen == blen && memcmp(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 (memcmp(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(struct icmp6_nodeinfo *ni6, struct mbuf *m, 1635 struct ifnet **ifpp, char *subj) 1636 { 1637 struct ifnet *ifp; 1638 struct in6_ifaddr *ifa6; 1639 struct ifaddr *ifa; 1640 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1641 int addrs = 0, addrsofif, iffound = 0; 1642 int niflags = ni6->ni_flags; 1643 1644 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1645 switch (ni6->ni_code) { 1646 case ICMP6_NI_SUBJ_IPV6: 1647 if (subj == NULL) /* must be impossible... */ 1648 return (0); 1649 subj_ip6 = (struct sockaddr_in6 *)subj; 1650 break; 1651 default: 1652 /* 1653 * XXX: we only support IPv6 subject address for 1654 * this Qtype. 1655 */ 1656 return (0); 1657 } 1658 } 1659 1660 IFNET_FOREACH(ifp) { 1661 addrsofif = 0; 1662 IFADDR_FOREACH(ifa, ifp) { 1663 if (ifa->ifa_addr->sa_family != AF_INET6) 1664 continue; 1665 ifa6 = (struct in6_ifaddr *)ifa; 1666 1667 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1668 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1669 &ifa6->ia_addr.sin6_addr)) 1670 iffound = 1; 1671 1672 /* 1673 * IPv4-mapped addresses can only be returned by a 1674 * Node Information proxy, since they represent 1675 * addresses of IPv4-only nodes, which perforce do 1676 * not implement this protocol. 1677 * [icmp-name-lookups-07, Section 5.4] 1678 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1679 * this function at this moment. 1680 */ 1681 1682 /* What do we have to do about ::1? */ 1683 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1684 case IPV6_ADDR_SCOPE_LINKLOCAL: 1685 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1686 continue; 1687 break; 1688 case IPV6_ADDR_SCOPE_SITELOCAL: 1689 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1690 continue; 1691 break; 1692 case IPV6_ADDR_SCOPE_GLOBAL: 1693 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1694 continue; 1695 break; 1696 default: 1697 continue; 1698 } 1699 1700 /* 1701 * check if anycast is okay. 1702 * XXX: just experimental. not in the spec. 1703 */ 1704 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1705 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1706 continue; /* we need only unicast addresses */ 1707 1708 addrsofif++; /* count the address */ 1709 } 1710 if (iffound) { 1711 *ifpp = ifp; 1712 return (addrsofif); 1713 } 1714 1715 addrs += addrsofif; 1716 } 1717 1718 return (addrs); 1719 } 1720 1721 static int 1722 ni6_store_addrs(struct icmp6_nodeinfo *ni6, 1723 struct icmp6_nodeinfo *nni6, struct ifnet *ifp0, 1724 int resid) 1725 { 1726 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1727 struct in6_ifaddr *ifa6; 1728 struct ifaddr *ifa; 1729 struct ifnet *ifp_dep = NULL; 1730 int copied = 0, allow_deprecated = 0; 1731 u_char *cp = (u_char *)(nni6 + 1); 1732 int niflags = ni6->ni_flags; 1733 u_int32_t ltime; 1734 1735 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1736 return (0); /* needless to copy */ 1737 1738 again: 1739 1740 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1741 { 1742 IFADDR_FOREACH(ifa, ifp) { 1743 if (ifa->ifa_addr->sa_family != AF_INET6) 1744 continue; 1745 ifa6 = (struct in6_ifaddr *)ifa; 1746 1747 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1748 allow_deprecated == 0) { 1749 /* 1750 * prefererred address should be put before 1751 * deprecated addresses. 1752 */ 1753 1754 /* record the interface for later search */ 1755 if (ifp_dep == NULL) 1756 ifp_dep = ifp; 1757 1758 continue; 1759 } 1760 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1761 allow_deprecated != 0) 1762 continue; /* we now collect deprecated addrs */ 1763 1764 /* What do we have to do about ::1? */ 1765 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1766 case IPV6_ADDR_SCOPE_LINKLOCAL: 1767 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1768 continue; 1769 break; 1770 case IPV6_ADDR_SCOPE_SITELOCAL: 1771 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1772 continue; 1773 break; 1774 case IPV6_ADDR_SCOPE_GLOBAL: 1775 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1776 continue; 1777 break; 1778 default: 1779 continue; 1780 } 1781 1782 /* 1783 * check if anycast is okay. 1784 * XXX: just experimental. not in the spec. 1785 */ 1786 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1787 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1788 continue; 1789 1790 /* now we can copy the address */ 1791 if (resid < sizeof(struct in6_addr) + 1792 sizeof(u_int32_t)) { 1793 /* 1794 * We give up much more copy. 1795 * Set the truncate flag and return. 1796 */ 1797 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1798 return (copied); 1799 } 1800 1801 /* 1802 * Set the TTL of the address. 1803 * The TTL value should be one of the following 1804 * according to the specification: 1805 * 1806 * 1. The remaining lifetime of a DHCP lease on the 1807 * address, or 1808 * 2. The remaining Valid Lifetime of a prefix from 1809 * which the address was derived through Stateless 1810 * Autoconfiguration. 1811 * 1812 * Note that we currently do not support stateful 1813 * address configuration by DHCPv6, so the former 1814 * case can't happen. 1815 * 1816 * TTL must be 2^31 > TTL >= 0. 1817 */ 1818 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1819 ltime = ND6_INFINITE_LIFETIME; 1820 else { 1821 if (ifa6->ia6_lifetime.ia6t_expire > 1822 time_second) 1823 ltime = ifa6->ia6_lifetime.ia6t_expire - 1824 time_second; 1825 else 1826 ltime = 0; 1827 } 1828 if (ltime > 0x7fffffff) 1829 ltime = 0x7fffffff; 1830 ltime = htonl(ltime); 1831 1832 bcopy(<ime, cp, sizeof(u_int32_t)); 1833 cp += sizeof(u_int32_t); 1834 1835 /* copy the address itself */ 1836 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1837 sizeof(struct in6_addr)); 1838 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1839 cp += sizeof(struct in6_addr); 1840 1841 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1842 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1843 } 1844 if (ifp0) /* we need search only on the specified IF */ 1845 break; 1846 } 1847 1848 if (allow_deprecated == 0 && ifp_dep != NULL) { 1849 ifp = ifp_dep; 1850 allow_deprecated = 1; 1851 1852 goto again; 1853 } 1854 1855 return (copied); 1856 } 1857 1858 /* 1859 * XXX almost dup'ed code with rip6_input. 1860 */ 1861 static int 1862 icmp6_rip6_input(struct mbuf **mp, int off) 1863 { 1864 struct mbuf *m = *mp; 1865 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1866 struct inpcb_hdr *inph; 1867 struct in6pcb *in6p; 1868 struct in6pcb *last = NULL; 1869 struct sockaddr_in6 rip6src; 1870 struct icmp6_hdr *icmp6; 1871 struct mbuf *opts = NULL; 1872 1873 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1874 if (icmp6 == NULL) { 1875 /* m is already reclaimed */ 1876 return IPPROTO_DONE; 1877 } 1878 1879 /* 1880 * XXX: the address may have embedded scope zone ID, which should be 1881 * hidden from applications. 1882 */ 1883 sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0); 1884 if (sa6_recoverscope(&rip6src)) { 1885 m_freem(m); 1886 return (IPPROTO_DONE); 1887 } 1888 1889 CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) { 1890 in6p = (struct in6pcb *)inph; 1891 if (in6p->in6p_af != AF_INET6) 1892 continue; 1893 if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6) 1894 continue; 1895 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1896 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1897 continue; 1898 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1899 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1900 continue; 1901 if (in6p->in6p_icmp6filt 1902 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1903 in6p->in6p_icmp6filt)) 1904 continue; 1905 if (last) { 1906 struct mbuf *n; 1907 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1908 if (last->in6p_flags & IN6P_CONTROLOPTS) 1909 ip6_savecontrol(last, &opts, ip6, n); 1910 /* strip intermediate headers */ 1911 m_adj(n, off); 1912 if (sbappendaddr(&last->in6p_socket->so_rcv, 1913 (struct sockaddr *)&rip6src, 1914 n, opts) == 0) { 1915 /* should notify about lost packet */ 1916 m_freem(n); 1917 if (opts) 1918 m_freem(opts); 1919 } else 1920 sorwakeup(last->in6p_socket); 1921 opts = NULL; 1922 } 1923 } 1924 last = in6p; 1925 } 1926 if (last) { 1927 if (last->in6p_flags & IN6P_CONTROLOPTS) 1928 ip6_savecontrol(last, &opts, ip6, m); 1929 /* strip intermediate headers */ 1930 m_adj(m, off); 1931 if (sbappendaddr(&last->in6p_socket->so_rcv, 1932 (struct sockaddr *)&rip6src, m, opts) == 0) { 1933 m_freem(m); 1934 if (opts) 1935 m_freem(opts); 1936 } else 1937 sorwakeup(last->in6p_socket); 1938 } else { 1939 m_freem(m); 1940 IP6_STATDEC(IP6_STAT_DELIVERED); 1941 } 1942 return IPPROTO_DONE; 1943 } 1944 1945 /* 1946 * Reflect the ip6 packet back to the source. 1947 * OFF points to the icmp6 header, counted from the top of the mbuf. 1948 * 1949 * Note: RFC 1885 required that an echo reply should be truncated if it 1950 * did not fit in with (return) path MTU, and KAME code supported the 1951 * behavior. However, as a clarification after the RFC, this limitation 1952 * was removed in a revised version of the spec, RFC 2463. We had kept the 1953 * old behavior, with a (non-default) ifdef block, while the new version of 1954 * the spec was an internet-draft status, and even after the new RFC was 1955 * published. But it would rather make sense to clean the obsoleted part 1956 * up, and to make the code simpler at this stage. 1957 */ 1958 void 1959 icmp6_reflect(struct mbuf *m, size_t off) 1960 { 1961 struct ip6_hdr *ip6; 1962 struct icmp6_hdr *icmp6; 1963 const struct in6_ifaddr *ia; 1964 const struct ip6aux *ip6a; 1965 int plen; 1966 int type, code; 1967 struct ifnet *outif = NULL; 1968 struct in6_addr origdst; 1969 const struct in6_addr *src = NULL; 1970 1971 /* too short to reflect */ 1972 if (off < sizeof(struct ip6_hdr)) { 1973 nd6log((LOG_DEBUG, 1974 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 1975 (u_long)off, (u_long)sizeof(struct ip6_hdr), 1976 __FILE__, __LINE__)); 1977 goto bad; 1978 } 1979 1980 /* 1981 * If there are extra headers between IPv6 and ICMPv6, strip 1982 * off that header first. 1983 */ 1984 #ifdef DIAGNOSTIC 1985 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 1986 panic("assumption failed in icmp6_reflect"); 1987 #endif 1988 if (off > sizeof(struct ip6_hdr)) { 1989 size_t l; 1990 struct ip6_hdr nip6; 1991 1992 l = off - sizeof(struct ip6_hdr); 1993 m_copydata(m, 0, sizeof(nip6), (void *)&nip6); 1994 m_adj(m, l); 1995 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1996 if (m->m_len < l) { 1997 if ((m = m_pullup(m, l)) == NULL) 1998 return; 1999 } 2000 bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6)); 2001 } else /* off == sizeof(struct ip6_hdr) */ { 2002 size_t l; 2003 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2004 if (m->m_len < l) { 2005 if ((m = m_pullup(m, l)) == NULL) 2006 return; 2007 } 2008 } 2009 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2010 ip6 = mtod(m, struct ip6_hdr *); 2011 ip6->ip6_nxt = IPPROTO_ICMPV6; 2012 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2013 type = icmp6->icmp6_type; /* keep type for statistics */ 2014 code = icmp6->icmp6_code; /* ditto. */ 2015 2016 origdst = ip6->ip6_dst; 2017 /* 2018 * ip6_input() drops a packet if its src is multicast. 2019 * So, the src is never multicast. 2020 */ 2021 ip6->ip6_dst = ip6->ip6_src; 2022 2023 /* 2024 * If the incoming packet was addressed directly to us (i.e. unicast), 2025 * use dst as the src for the reply. 2026 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2027 * (for example) when we encounter an error while forwarding procedure 2028 * destined to a duplicated address of ours. 2029 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2030 * procedure of an outgoing packet of our own, in which case we need 2031 * to search in the ifaddr list. 2032 */ 2033 if (IN6_IS_ADDR_MULTICAST(&origdst)) 2034 ; 2035 else if ((ip6a = ip6_getdstifaddr(m)) != NULL) { 2036 if ((ip6a->ip6a_flags & 2037 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2038 src = &ip6a->ip6a_src; 2039 } else { 2040 union { 2041 struct sockaddr_in6 sin6; 2042 struct sockaddr sa; 2043 } u; 2044 2045 sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0); 2046 2047 ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa); 2048 2049 if (ia == NULL) 2050 ; 2051 else if ((ia->ia6_flags & 2052 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2053 src = &ia->ia_addr.sin6_addr; 2054 } 2055 2056 if (src == NULL) { 2057 int e; 2058 struct sockaddr_in6 sin6; 2059 struct route ro; 2060 2061 /* 2062 * This case matches to multicasts, our anycast, or unicasts 2063 * that we do not own. Select a source address based on the 2064 * source address of the erroneous packet. 2065 */ 2066 /* zone ID should be embedded */ 2067 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 2068 2069 memset(&ro, 0, sizeof(ro)); 2070 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2071 rtcache_free(&ro); 2072 if (src == NULL) { 2073 nd6log((LOG_DEBUG, 2074 "icmp6_reflect: source can't be determined: " 2075 "dst=%s, error=%d\n", 2076 ip6_sprintf(&sin6.sin6_addr), e)); 2077 goto bad; 2078 } 2079 } 2080 2081 ip6->ip6_src = *src; 2082 ip6->ip6_flow = 0; 2083 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2084 ip6->ip6_vfc |= IPV6_VERSION; 2085 ip6->ip6_nxt = IPPROTO_ICMPV6; 2086 if (m->m_pkthdr.rcvif) { 2087 /* XXX: This may not be the outgoing interface */ 2088 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2089 } else 2090 ip6->ip6_hlim = ip6_defhlim; 2091 2092 m->m_pkthdr.csum_flags = 0; 2093 icmp6->icmp6_cksum = 0; 2094 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2095 sizeof(struct ip6_hdr), plen); 2096 2097 /* 2098 * XXX option handling 2099 */ 2100 2101 m->m_flags &= ~(M_BCAST|M_MCAST); 2102 2103 /* 2104 * To avoid a "too big" situation at an intermediate router 2105 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2106 * Note that only echo and node information replies are affected, 2107 * since the length of ICMP6 errors is limited to the minimum MTU. 2108 */ 2109 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 && 2110 outif) 2111 icmp6_ifstat_inc(outif, ifs6_out_error); 2112 2113 if (outif) 2114 icmp6_ifoutstat_inc(outif, type, code); 2115 2116 return; 2117 2118 bad: 2119 m_freem(m); 2120 return; 2121 } 2122 2123 static const char * 2124 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2125 struct in6_addr *tgt6) 2126 { 2127 static char buf[1024]; 2128 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2129 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2130 return buf; 2131 } 2132 2133 void 2134 icmp6_redirect_input(struct mbuf *m, int off) 2135 { 2136 struct ifnet *ifp = m->m_pkthdr.rcvif; 2137 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2138 struct nd_redirect *nd_rd; 2139 int icmp6len = ntohs(ip6->ip6_plen); 2140 char *lladdr = NULL; 2141 int lladdrlen = 0; 2142 struct rtentry *rt = NULL; 2143 int is_router; 2144 int is_onlink; 2145 struct in6_addr src6 = ip6->ip6_src; 2146 struct in6_addr redtgt6; 2147 struct in6_addr reddst6; 2148 union nd_opts ndopts; 2149 2150 if (!ifp) 2151 return; 2152 2153 /* XXX if we are router, we don't update route by icmp6 redirect */ 2154 if (ip6_forwarding) 2155 goto freeit; 2156 if (!icmp6_rediraccept) 2157 goto freeit; 2158 2159 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2160 if (nd_rd == NULL) { 2161 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 2162 return; 2163 } 2164 redtgt6 = nd_rd->nd_rd_target; 2165 reddst6 = nd_rd->nd_rd_dst; 2166 2167 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2168 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2169 goto freeit; 2170 } 2171 2172 /* validation */ 2173 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2174 nd6log((LOG_ERR, 2175 "ICMP6 redirect sent from %s rejected; " 2176 "must be from linklocal\n", ip6_sprintf(&src6))); 2177 goto bad; 2178 } 2179 if (ip6->ip6_hlim != 255) { 2180 nd6log((LOG_ERR, 2181 "ICMP6 redirect sent from %s rejected; " 2182 "hlim=%d (must be 255)\n", 2183 ip6_sprintf(&src6), ip6->ip6_hlim)); 2184 goto bad; 2185 } 2186 { 2187 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2188 struct sockaddr_in6 sin6; 2189 struct in6_addr *gw6; 2190 2191 sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0); 2192 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2193 if (rt) { 2194 if (rt->rt_gateway == NULL || 2195 rt->rt_gateway->sa_family != AF_INET6) { 2196 nd6log((LOG_ERR, 2197 "ICMP6 redirect rejected; no route " 2198 "with inet6 gateway found for redirect dst: %s\n", 2199 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2200 RTFREE(rt); 2201 goto bad; 2202 } 2203 2204 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2205 if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2206 nd6log((LOG_ERR, 2207 "ICMP6 redirect rejected; " 2208 "not equal to gw-for-src=%s (must be same): " 2209 "%s\n", 2210 ip6_sprintf(gw6), 2211 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2212 RTFREE(rt); 2213 goto bad; 2214 } 2215 } else { 2216 nd6log((LOG_ERR, 2217 "ICMP6 redirect rejected; " 2218 "no route found for redirect dst: %s\n", 2219 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2220 goto bad; 2221 } 2222 RTFREE(rt); 2223 rt = NULL; 2224 } 2225 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2226 nd6log((LOG_ERR, 2227 "ICMP6 redirect rejected; " 2228 "redirect dst must be unicast: %s\n", 2229 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2230 goto bad; 2231 } 2232 2233 is_router = is_onlink = 0; 2234 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2235 is_router = 1; /* router case */ 2236 if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2237 is_onlink = 1; /* on-link destination case */ 2238 if (!is_router && !is_onlink) { 2239 nd6log((LOG_ERR, 2240 "ICMP6 redirect rejected; " 2241 "neither router case nor onlink case: %s\n", 2242 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2243 goto bad; 2244 } 2245 /* validation passed */ 2246 2247 icmp6len -= sizeof(*nd_rd); 2248 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2249 if (nd6_options(&ndopts) < 0) { 2250 nd6log((LOG_INFO, "icmp6_redirect_input: " 2251 "invalid ND option, rejected: %s\n", 2252 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2253 /* nd6_options have incremented stats */ 2254 goto freeit; 2255 } 2256 2257 if (ndopts.nd_opts_tgt_lladdr) { 2258 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2259 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2260 } 2261 2262 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2263 nd6log((LOG_INFO, 2264 "icmp6_redirect_input: lladdrlen mismatch for %s " 2265 "(if %d, icmp6 packet %d): %s\n", 2266 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2267 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2268 goto bad; 2269 } 2270 2271 /* RFC 2461 8.3 */ 2272 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2273 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2274 2275 if (!is_onlink) { /* better router case. perform rtredirect. */ 2276 /* perform rtredirect */ 2277 struct sockaddr_in6 sdst; 2278 struct sockaddr_in6 sgw; 2279 struct sockaddr_in6 ssrc; 2280 unsigned long rtcount; 2281 struct rtentry *newrt = NULL; 2282 2283 /* 2284 * do not install redirect route, if the number of entries 2285 * is too much (> hiwat). note that, the node (= host) will 2286 * work just fine even if we do not install redirect route 2287 * (there will be additional hops, though). 2288 */ 2289 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2290 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2291 return; 2292 else if (0 <= icmp6_redirect_lowat && 2293 rtcount > icmp6_redirect_lowat) { 2294 /* 2295 * XXX nuke a victim, install the new one. 2296 */ 2297 } 2298 2299 memset(&sdst, 0, sizeof(sdst)); 2300 memset(&sgw, 0, sizeof(sgw)); 2301 memset(&ssrc, 0, sizeof(ssrc)); 2302 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2303 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2304 sizeof(struct sockaddr_in6); 2305 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2306 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2307 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2308 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2309 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2310 (struct sockaddr *)&ssrc, 2311 &newrt); 2312 2313 if (newrt) { 2314 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2315 icmp6_redirect_timeout_q); 2316 rtfree(newrt); 2317 } 2318 } 2319 /* finally update cached route in each socket via pfctlinput */ 2320 { 2321 struct sockaddr_in6 sdst; 2322 2323 sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0); 2324 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2325 #if defined(IPSEC) || defined(FAST_IPSEC) 2326 key_sa_routechange((struct sockaddr *)&sdst); 2327 #endif 2328 } 2329 2330 freeit: 2331 m_freem(m); 2332 return; 2333 2334 bad: 2335 ICMP6_STATINC(ICMP6_STAT_BADREDIRECT); 2336 m_freem(m); 2337 } 2338 2339 void 2340 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2341 { 2342 struct ifnet *ifp; /* my outgoing interface */ 2343 struct in6_addr *ifp_ll6; 2344 struct in6_addr *nexthop; 2345 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2346 struct mbuf *m = NULL; /* newly allocated one */ 2347 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2348 struct nd_redirect *nd_rd; 2349 size_t maxlen; 2350 u_char *p; 2351 struct sockaddr_in6 src_sa; 2352 2353 icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0); 2354 2355 /* if we are not router, we don't send icmp6 redirect */ 2356 if (!ip6_forwarding) 2357 goto fail; 2358 2359 /* sanity check */ 2360 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2361 goto fail; 2362 2363 /* 2364 * Address check: 2365 * the source address must identify a neighbor, and 2366 * the destination address must not be a multicast address 2367 * [RFC 2461, sec 8.2] 2368 */ 2369 sip6 = mtod(m0, struct ip6_hdr *); 2370 sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0); 2371 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2372 goto fail; 2373 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2374 goto fail; /* what should we do here? */ 2375 2376 /* rate limit */ 2377 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2378 goto fail; 2379 2380 /* 2381 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2382 * we almost always ask for an mbuf cluster for simplicity. 2383 * (MHLEN < IPV6_MMTU is almost always true) 2384 */ 2385 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2386 if (m && IPV6_MMTU >= MHLEN) { 2387 #if IPV6_MMTU >= MCLBYTES 2388 _MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT); 2389 #else 2390 MCLGET(m, M_DONTWAIT); 2391 #endif 2392 } 2393 2394 if (!m) 2395 goto fail; 2396 m->m_pkthdr.rcvif = NULL; 2397 m->m_len = 0; 2398 maxlen = M_TRAILINGSPACE(m); 2399 maxlen = min(IPV6_MMTU, maxlen); 2400 /* just for safety */ 2401 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2402 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2403 goto fail; 2404 } 2405 2406 { 2407 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2408 struct in6_ifaddr *ia; 2409 if ((ia = in6ifa_ifpforlinklocal(ifp, 2410 IN6_IFF_NOTREADY| 2411 IN6_IFF_ANYCAST)) == NULL) 2412 goto fail; 2413 ifp_ll6 = &ia->ia_addr.sin6_addr; 2414 } 2415 2416 /* get ip6 linklocal address for the router. */ 2417 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2418 struct sockaddr_in6 *sin6; 2419 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2420 nexthop = &sin6->sin6_addr; 2421 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2422 nexthop = NULL; 2423 } else 2424 nexthop = NULL; 2425 2426 /* ip6 */ 2427 ip6 = mtod(m, struct ip6_hdr *); 2428 ip6->ip6_flow = 0; 2429 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2430 ip6->ip6_vfc |= IPV6_VERSION; 2431 /* ip6->ip6_plen will be set later */ 2432 ip6->ip6_nxt = IPPROTO_ICMPV6; 2433 ip6->ip6_hlim = 255; 2434 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2435 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2436 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2437 2438 /* ND Redirect */ 2439 nd_rd = (struct nd_redirect *)(ip6 + 1); 2440 nd_rd->nd_rd_type = ND_REDIRECT; 2441 nd_rd->nd_rd_code = 0; 2442 nd_rd->nd_rd_reserved = 0; 2443 if (rt->rt_flags & RTF_GATEWAY) { 2444 /* 2445 * nd_rd->nd_rd_target must be a link-local address in 2446 * better router cases. 2447 */ 2448 if (!nexthop) 2449 goto fail; 2450 bcopy(nexthop, &nd_rd->nd_rd_target, 2451 sizeof(nd_rd->nd_rd_target)); 2452 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2453 sizeof(nd_rd->nd_rd_dst)); 2454 } else { 2455 /* make sure redtgt == reddst */ 2456 nexthop = &sip6->ip6_dst; 2457 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2458 sizeof(nd_rd->nd_rd_target)); 2459 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2460 sizeof(nd_rd->nd_rd_dst)); 2461 } 2462 2463 p = (u_char *)(nd_rd + 1); 2464 2465 { 2466 /* target lladdr option */ 2467 struct rtentry *rt_nexthop = NULL; 2468 int len; 2469 const struct sockaddr_dl *sdl; 2470 struct nd_opt_hdr *nd_opt; 2471 char *lladdr; 2472 2473 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2474 if (!rt_nexthop) 2475 goto nolladdropt; 2476 len = sizeof(*nd_opt) + ifp->if_addrlen; 2477 len = (len + 7) & ~7; /* round by 8 */ 2478 /* safety check */ 2479 if (len + (p - (u_char *)ip6) > maxlen) 2480 goto nolladdropt; 2481 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2482 (rt_nexthop->rt_flags & RTF_LLINFO) && 2483 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2484 (sdl = satocsdl(rt_nexthop->rt_gateway)) && 2485 sdl->sdl_alen) { 2486 nd_opt = (struct nd_opt_hdr *)p; 2487 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2488 nd_opt->nd_opt_len = len >> 3; 2489 lladdr = (char *)(nd_opt + 1); 2490 memcpy(lladdr, CLLADDR(sdl), ifp->if_addrlen); 2491 p += len; 2492 } 2493 } 2494 nolladdropt:; 2495 2496 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2497 2498 /* just to be safe */ 2499 if (m0->m_flags & M_DECRYPTED) 2500 goto noredhdropt; 2501 if (p - (u_char *)ip6 > maxlen) 2502 goto noredhdropt; 2503 2504 { 2505 /* redirected header option */ 2506 int len; 2507 struct nd_opt_rd_hdr *nd_opt_rh; 2508 2509 /* 2510 * compute the maximum size for icmp6 redirect header option. 2511 * XXX room for auth header? 2512 */ 2513 len = maxlen - (p - (u_char *)ip6); 2514 len &= ~7; 2515 2516 /* 2517 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2518 * about padding/truncate rule for the original IP packet. 2519 * From the discussion on IPv6imp in Feb 1999, 2520 * the consensus was: 2521 * - "attach as much as possible" is the goal 2522 * - pad if not aligned (original size can be guessed by 2523 * original ip6 header) 2524 * Following code adds the padding if it is simple enough, 2525 * and truncates if not. 2526 */ 2527 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2528 /* not enough room, truncate */ 2529 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2530 m0->m_pkthdr.len); 2531 } else { 2532 /* 2533 * enough room, truncate if not aligned. 2534 * we don't pad here for simplicity. 2535 */ 2536 size_t extra; 2537 2538 extra = m0->m_pkthdr.len % 8; 2539 if (extra) { 2540 /* truncate */ 2541 m_adj(m0, -extra); 2542 } 2543 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2544 } 2545 2546 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2547 memset(nd_opt_rh, 0, sizeof(*nd_opt_rh)); 2548 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2549 nd_opt_rh->nd_opt_rh_len = len >> 3; 2550 p += sizeof(*nd_opt_rh); 2551 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2552 2553 /* connect m0 to m */ 2554 m->m_pkthdr.len += m0->m_pkthdr.len; 2555 m_cat(m, m0); 2556 m0 = NULL; 2557 } 2558 noredhdropt: 2559 if (m0) { 2560 m_freem(m0); 2561 m0 = NULL; 2562 } 2563 2564 /* XXX: clear embedded link IDs in the inner header */ 2565 in6_clearscope(&sip6->ip6_src); 2566 in6_clearscope(&sip6->ip6_dst); 2567 in6_clearscope(&nd_rd->nd_rd_target); 2568 in6_clearscope(&nd_rd->nd_rd_dst); 2569 2570 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2571 2572 nd_rd->nd_rd_cksum = 0; 2573 nd_rd->nd_rd_cksum 2574 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2575 2576 /* send the packet to outside... */ 2577 if (ip6_output(m, NULL, NULL, 0, 2578 (struct ip6_moptions *)NULL, (struct socket *)NULL, NULL) != 0) 2579 icmp6_ifstat_inc(ifp, ifs6_out_error); 2580 2581 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2582 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2583 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT); 2584 2585 return; 2586 2587 fail: 2588 if (m) 2589 m_freem(m); 2590 if (m0) 2591 m_freem(m0); 2592 } 2593 2594 /* 2595 * ICMPv6 socket option processing. 2596 */ 2597 int 2598 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 2599 { 2600 int error = 0; 2601 struct in6pcb *in6p = sotoin6pcb(so); 2602 2603 if (sopt->sopt_level != IPPROTO_ICMPV6) 2604 return rip6_ctloutput(op, so, sopt); 2605 2606 switch (op) { 2607 case PRCO_SETOPT: 2608 switch (sopt->sopt_name) { 2609 case ICMP6_FILTER: 2610 { 2611 struct icmp6_filter fil; 2612 2613 error = sockopt_get(sopt, &fil, sizeof(fil)); 2614 if (error) 2615 break; 2616 memcpy(in6p->in6p_icmp6filt, &fil, 2617 sizeof(struct icmp6_filter)); 2618 error = 0; 2619 break; 2620 } 2621 2622 default: 2623 error = ENOPROTOOPT; 2624 break; 2625 } 2626 break; 2627 2628 case PRCO_GETOPT: 2629 switch (sopt->sopt_name) { 2630 case ICMP6_FILTER: 2631 { 2632 if (in6p->in6p_icmp6filt == NULL) { 2633 error = EINVAL; 2634 break; 2635 } 2636 error = sockopt_set(sopt, in6p->in6p_icmp6filt, 2637 sizeof(struct icmp6_filter)); 2638 break; 2639 } 2640 2641 default: 2642 error = ENOPROTOOPT; 2643 break; 2644 } 2645 break; 2646 } 2647 2648 return (error); 2649 } 2650 2651 /* 2652 * Perform rate limit check. 2653 * Returns 0 if it is okay to send the icmp6 packet. 2654 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2655 * limitation. 2656 * 2657 * XXX per-destination/type check necessary? 2658 */ 2659 static int 2660 icmp6_ratelimit( 2661 const struct in6_addr *dst, /* not used at this moment */ 2662 const int type, /* not used at this moment */ 2663 const int code) /* not used at this moment */ 2664 { 2665 int ret; 2666 2667 ret = 0; /* okay to send */ 2668 2669 /* PPS limit */ 2670 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2671 icmp6errppslim)) { 2672 /* The packet is subject to rate limit */ 2673 ret++; 2674 } 2675 2676 return ret; 2677 } 2678 2679 static struct rtentry * 2680 icmp6_mtudisc_clone(struct sockaddr *dst) 2681 { 2682 struct rtentry *rt; 2683 int error; 2684 2685 rt = rtalloc1(dst, 1); 2686 if (rt == 0) 2687 return NULL; 2688 2689 /* If we didn't get a host route, allocate one */ 2690 if ((rt->rt_flags & RTF_HOST) == 0) { 2691 struct rtentry *nrt; 2692 2693 error = rtrequest((int) RTM_ADD, dst, 2694 (struct sockaddr *) rt->rt_gateway, 2695 (struct sockaddr *) 0, 2696 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2697 if (error) { 2698 rtfree(rt); 2699 return NULL; 2700 } 2701 nrt->rt_rmx = rt->rt_rmx; 2702 rtfree(rt); 2703 rt = nrt; 2704 } 2705 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2706 icmp6_mtudisc_timeout_q); 2707 if (error) { 2708 rtfree(rt); 2709 return NULL; 2710 } 2711 2712 return rt; /* caller need to call rtfree() */ 2713 } 2714 2715 static void 2716 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 2717 { 2718 if (rt == NULL) 2719 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2720 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2721 (RTF_DYNAMIC | RTF_HOST)) { 2722 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2723 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2724 } else { 2725 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2726 rt->rt_rmx.rmx_mtu = 0; 2727 } 2728 } 2729 2730 static void 2731 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r) 2732 { 2733 if (rt == NULL) 2734 panic("icmp6_redirect_timeout: bad route to timeout"); 2735 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2736 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2737 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2738 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2739 } 2740 } 2741 2742 /* 2743 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes. silly? 2744 */ 2745 static int 2746 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS) 2747 { 2748 (void)&name; 2749 (void)&l; 2750 (void)&oname; 2751 2752 if (namelen != 0) 2753 return (EINVAL); 2754 2755 return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp, 2756 /*XXXUNCONST*/ 2757 __UNCONST(newp), newlen)); 2758 } 2759 2760 static int 2761 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS) 2762 { 2763 2764 return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS)); 2765 } 2766 2767 static void 2768 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog) 2769 { 2770 extern int nd6_maxqueuelen; /* defined in nd6.c */ 2771 2772 sysctl_createv(clog, 0, NULL, NULL, 2773 CTLFLAG_PERMANENT, 2774 CTLTYPE_NODE, "net", NULL, 2775 NULL, 0, NULL, 0, 2776 CTL_NET, CTL_EOL); 2777 sysctl_createv(clog, 0, NULL, NULL, 2778 CTLFLAG_PERMANENT, 2779 CTLTYPE_NODE, "inet6", NULL, 2780 NULL, 0, NULL, 0, 2781 CTL_NET, PF_INET6, CTL_EOL); 2782 sysctl_createv(clog, 0, NULL, NULL, 2783 CTLFLAG_PERMANENT, 2784 CTLTYPE_NODE, "icmp6", 2785 SYSCTL_DESCR("ICMPv6 related settings"), 2786 NULL, 0, NULL, 0, 2787 CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL); 2788 2789 sysctl_createv(clog, 0, NULL, NULL, 2790 CTLFLAG_PERMANENT, 2791 CTLTYPE_STRUCT, "stats", 2792 SYSCTL_DESCR("ICMPv6 transmission statistics"), 2793 sysctl_net_inet6_icmp6_stats, 0, NULL, 0, 2794 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2795 ICMPV6CTL_STATS, CTL_EOL); 2796 sysctl_createv(clog, 0, NULL, NULL, 2797 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2798 CTLTYPE_INT, "rediraccept", 2799 SYSCTL_DESCR("Accept and process redirect messages"), 2800 NULL, 0, &icmp6_rediraccept, 0, 2801 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2802 ICMPV6CTL_REDIRACCEPT, CTL_EOL); 2803 sysctl_createv(clog, 0, NULL, NULL, 2804 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2805 CTLTYPE_INT, "redirtimeout", 2806 SYSCTL_DESCR("Redirect generated route lifetime"), 2807 NULL, 0, &icmp6_redirtimeout, 0, 2808 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2809 ICMPV6CTL_REDIRTIMEOUT, CTL_EOL); 2810 #if 0 /* obsoleted */ 2811 sysctl_createv(clog, 0, NULL, NULL, 2812 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2813 CTLTYPE_INT, "errratelimit", NULL, 2814 NULL, 0, &icmp6_errratelimit, 0, 2815 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2816 ICMPV6CTL_ERRRATELIMIT, CTL_EOL); 2817 #endif 2818 sysctl_createv(clog, 0, NULL, NULL, 2819 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2820 CTLTYPE_INT, "nd6_prune", 2821 SYSCTL_DESCR("Neighbor discovery prune interval"), 2822 NULL, 0, &nd6_prune, 0, 2823 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2824 ICMPV6CTL_ND6_PRUNE, CTL_EOL); 2825 sysctl_createv(clog, 0, NULL, NULL, 2826 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2827 CTLTYPE_INT, "nd6_delay", 2828 SYSCTL_DESCR("First probe delay time"), 2829 NULL, 0, &nd6_delay, 0, 2830 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2831 ICMPV6CTL_ND6_DELAY, CTL_EOL); 2832 sysctl_createv(clog, 0, NULL, NULL, 2833 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2834 CTLTYPE_INT, "nd6_umaxtries", 2835 SYSCTL_DESCR("Number of unicast discovery attempts"), 2836 NULL, 0, &nd6_umaxtries, 0, 2837 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2838 ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL); 2839 sysctl_createv(clog, 0, NULL, NULL, 2840 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2841 CTLTYPE_INT, "nd6_mmaxtries", 2842 SYSCTL_DESCR("Number of multicast discovery attempts"), 2843 NULL, 0, &nd6_mmaxtries, 0, 2844 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2845 ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL); 2846 sysctl_createv(clog, 0, NULL, NULL, 2847 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2848 CTLTYPE_INT, "nd6_useloopback", 2849 SYSCTL_DESCR("Use loopback interface for local traffic"), 2850 NULL, 0, &nd6_useloopback, 0, 2851 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2852 ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL); 2853 #if 0 /* obsoleted */ 2854 sysctl_createv(clog, 0, NULL, NULL, 2855 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2856 CTLTYPE_INT, "nd6_proxyall", NULL, 2857 NULL, 0, &nd6_proxyall, 0, 2858 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2859 ICMPV6CTL_ND6_PROXYALL, CTL_EOL); 2860 #endif 2861 sysctl_createv(clog, 0, NULL, NULL, 2862 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2863 CTLTYPE_INT, "nodeinfo", 2864 SYSCTL_DESCR("Respond to node information requests"), 2865 NULL, 0, &icmp6_nodeinfo, 0, 2866 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2867 ICMPV6CTL_NODEINFO, CTL_EOL); 2868 sysctl_createv(clog, 0, NULL, NULL, 2869 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2870 CTLTYPE_INT, "errppslimit", 2871 SYSCTL_DESCR("Maximum ICMP errors sent per second"), 2872 NULL, 0, &icmp6errppslim, 0, 2873 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2874 ICMPV6CTL_ERRPPSLIMIT, CTL_EOL); 2875 sysctl_createv(clog, 0, NULL, NULL, 2876 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2877 CTLTYPE_INT, "nd6_maxnudhint", 2878 SYSCTL_DESCR("Maximum neighbor unreachable hint count"), 2879 NULL, 0, &nd6_maxnudhint, 0, 2880 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2881 ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL); 2882 sysctl_createv(clog, 0, NULL, NULL, 2883 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2884 CTLTYPE_INT, "mtudisc_hiwat", 2885 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2886 NULL, 0, &icmp6_mtudisc_hiwat, 0, 2887 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2888 ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL); 2889 sysctl_createv(clog, 0, NULL, NULL, 2890 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2891 CTLTYPE_INT, "mtudisc_lowat", 2892 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2893 NULL, 0, &icmp6_mtudisc_lowat, 0, 2894 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2895 ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL); 2896 sysctl_createv(clog, 0, NULL, NULL, 2897 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2898 CTLTYPE_INT, "nd6_debug", 2899 SYSCTL_DESCR("Enable neighbor discovery debug output"), 2900 NULL, 0, &nd6_debug, 0, 2901 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2902 ICMPV6CTL_ND6_DEBUG, CTL_EOL); 2903 sysctl_createv(clog, 0, NULL, NULL, 2904 CTLFLAG_PERMANENT, 2905 CTLTYPE_STRUCT, "nd6_drlist", 2906 SYSCTL_DESCR("Default router list"), 2907 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2908 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2909 ICMPV6CTL_ND6_DRLIST, CTL_EOL); 2910 sysctl_createv(clog, 0, NULL, NULL, 2911 CTLFLAG_PERMANENT, 2912 CTLTYPE_STRUCT, "nd6_prlist", 2913 SYSCTL_DESCR("Prefix list"), 2914 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2915 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2916 ICMPV6CTL_ND6_PRLIST, CTL_EOL); 2917 sysctl_createv(clog, 0, NULL, NULL, 2918 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2919 CTLTYPE_INT, "maxqueuelen", 2920 SYSCTL_DESCR("max packet queue len for a unresolved ND"), 2921 NULL, 1, &nd6_maxqueuelen, 0, 2922 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2923 ICMPV6CTL_ND6_MAXQLEN, CTL_EOL); 2924 } 2925 2926 void 2927 icmp6_statinc(u_int stat) 2928 { 2929 2930 KASSERT(stat < ICMP6_NSTATS); 2931 ICMP6_STATINC(stat); 2932 } 2933