1 /* $NetBSD: icmp6.c,v 1.169 2014/06/06 01:02:47 rmind 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 1.169 2014/06/06 01:02:47 rmind 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/kmem.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 <netipsec/ipsec.h> 104 #include <netipsec/key.h> 105 #endif 106 107 108 #include "faith.h" 109 #if defined(NFAITH) && 0 < NFAITH 110 #include <net/if_faith.h> 111 #endif 112 113 #include <net/net_osdep.h> 114 115 extern struct domain inet6domain; 116 117 percpu_t *icmp6stat_percpu; 118 119 extern struct inpcbtable raw6cbtable; 120 extern int icmp6errppslim; 121 static int icmp6errpps_count = 0; 122 static struct timeval icmp6errppslim_last; 123 extern int icmp6_nodeinfo; 124 125 /* 126 * List of callbacks to notify when Path MTU changes are made. 127 */ 128 struct icmp6_mtudisc_callback { 129 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 130 void (*mc_func)(struct in6_addr *); 131 }; 132 133 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 134 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 135 136 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 137 extern int pmtu_expire; 138 139 /* XXX do these values make any sense? */ 140 static int icmp6_mtudisc_hiwat = 1280; 141 static int icmp6_mtudisc_lowat = 256; 142 143 /* 144 * keep track of # of redirect routes. 145 */ 146 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 147 148 /* XXX experimental, turned off */ 149 static int icmp6_redirect_hiwat = -1; 150 static int icmp6_redirect_lowat = -1; 151 152 static void icmp6_errcount(u_int, int, int); 153 static int icmp6_rip6_input(struct mbuf **, int); 154 static int icmp6_ratelimit(const struct in6_addr *, const int, const int); 155 static const char *icmp6_redirect_diag(struct in6_addr *, 156 struct in6_addr *, struct in6_addr *); 157 static struct mbuf *ni6_input(struct mbuf *, int); 158 static struct mbuf *ni6_nametodns(const char *, int, int); 159 static int ni6_dnsmatch(const char *, int, const char *, int); 160 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *, 161 struct ifnet **, char *); 162 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 163 struct ifnet *, int); 164 static int icmp6_notify_error(struct mbuf *, int, int, int); 165 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *); 166 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *); 167 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *); 168 static void sysctl_net_inet6_icmp6_setup(struct sysctllog **); 169 170 171 void 172 icmp6_init(void) 173 { 174 175 sysctl_net_inet6_icmp6_setup(NULL); 176 mld_init(); 177 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 178 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 179 180 icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS); 181 } 182 183 static void 184 icmp6_errcount(u_int base, int type, int code) 185 { 186 switch (type) { 187 case ICMP6_DST_UNREACH: 188 switch (code) { 189 case ICMP6_DST_UNREACH_NOROUTE: 190 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE); 191 return; 192 case ICMP6_DST_UNREACH_ADMIN: 193 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN); 194 return; 195 case ICMP6_DST_UNREACH_BEYONDSCOPE: 196 ICMP6_STATINC(base + 197 ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE); 198 return; 199 case ICMP6_DST_UNREACH_ADDR: 200 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR); 201 return; 202 case ICMP6_DST_UNREACH_NOPORT: 203 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT); 204 return; 205 } 206 break; 207 case ICMP6_PACKET_TOO_BIG: 208 ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG); 209 return; 210 case ICMP6_TIME_EXCEEDED: 211 switch (code) { 212 case ICMP6_TIME_EXCEED_TRANSIT: 213 ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT); 214 return; 215 case ICMP6_TIME_EXCEED_REASSEMBLY: 216 ICMP6_STATINC(base + 217 ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY); 218 return; 219 } 220 break; 221 case ICMP6_PARAM_PROB: 222 switch (code) { 223 case ICMP6_PARAMPROB_HEADER: 224 ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER); 225 return; 226 case ICMP6_PARAMPROB_NEXTHEADER: 227 ICMP6_STATINC(base + 228 ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER); 229 return; 230 case ICMP6_PARAMPROB_OPTION: 231 ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION); 232 return; 233 } 234 break; 235 case ND_REDIRECT: 236 ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT); 237 return; 238 } 239 ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN); 240 } 241 242 /* 243 * Register a Path MTU Discovery callback. 244 */ 245 void 246 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *)) 247 { 248 struct icmp6_mtudisc_callback *mc; 249 250 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 251 mc = LIST_NEXT(mc, mc_list)) { 252 if (mc->mc_func == func) 253 return; 254 } 255 256 mc = kmem_alloc(sizeof(*mc), KM_SLEEP); 257 mc->mc_func = func; 258 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 259 } 260 261 /* 262 * A wrapper function for icmp6_error() necessary when the erroneous packet 263 * may not contain enough scope zone information. 264 */ 265 void 266 icmp6_error2(struct mbuf *m, int type, int code, int param, 267 struct ifnet *ifp) 268 { 269 struct ip6_hdr *ip6; 270 271 if (ifp == NULL) 272 return; 273 274 if (m->m_len < sizeof(struct ip6_hdr)) { 275 m = m_pullup(m, sizeof(struct ip6_hdr)); 276 if (m == NULL) 277 return; 278 } 279 280 ip6 = mtod(m, struct ip6_hdr *); 281 282 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 283 return; 284 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 285 return; 286 287 icmp6_error(m, type, code, param); 288 } 289 290 /* 291 * Generate an error packet of type error in response to bad IP6 packet. 292 */ 293 void 294 icmp6_error(struct mbuf *m, int type, int code, int param) 295 { 296 struct ip6_hdr *oip6, *nip6; 297 struct icmp6_hdr *icmp6; 298 u_int preplen; 299 int off; 300 int nxt; 301 302 ICMP6_STATINC(ICMP6_STAT_ERROR); 303 304 /* count per-type-code statistics */ 305 icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code); 306 307 if (m->m_flags & M_DECRYPTED) { 308 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 309 goto freeit; 310 } 311 312 if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) && 313 (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) 314 return; 315 oip6 = mtod(m, struct ip6_hdr *); 316 317 /* 318 * If the destination address of the erroneous packet is a multicast 319 * address, or the packet was sent using link-layer multicast, 320 * we should basically suppress sending an error (RFC 2463, Section 321 * 2.4). 322 * We have two exceptions (the item e.2 in that section): 323 * - the Pakcet Too Big message can be sent for path MTU discovery. 324 * - the Parameter Problem Message that can be allowed an icmp6 error 325 * in the option type field. This check has been done in 326 * ip6_unknown_opt(), so we can just check the type and code. 327 */ 328 if ((m->m_flags & (M_BCAST|M_MCAST) || 329 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 330 (type != ICMP6_PACKET_TOO_BIG && 331 (type != ICMP6_PARAM_PROB || 332 code != ICMP6_PARAMPROB_OPTION))) 333 goto freeit; 334 335 /* 336 * RFC 2463, 2.4 (e.5): source address check. 337 * XXX: the case of anycast source? 338 */ 339 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 340 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 341 goto freeit; 342 343 /* 344 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 345 * don't do it. 346 */ 347 nxt = -1; 348 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 349 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 350 struct icmp6_hdr *icp; 351 352 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 353 sizeof(*icp)); 354 if (icp == NULL) { 355 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 356 return; 357 } 358 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 359 icp->icmp6_type == ND_REDIRECT) { 360 /* 361 * ICMPv6 error 362 * Special case: for redirect (which is 363 * informational) we must not send icmp6 error. 364 */ 365 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 366 goto freeit; 367 } else { 368 /* ICMPv6 informational - send the error */ 369 } 370 } 371 #if 0 /* controversial */ 372 else if (off >= 0 && nxt == IPPROTO_ESP) { 373 /* 374 * It could be ICMPv6 error inside ESP. Take a safer side, 375 * don't respond. 376 */ 377 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 378 goto freeit; 379 } 380 #endif 381 else { 382 /* non-ICMPv6 - send the error */ 383 } 384 385 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 386 387 /* Finally, do rate limitation check. */ 388 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 389 ICMP6_STATINC(ICMP6_STAT_TOOFREQ); 390 goto freeit; 391 } 392 393 /* 394 * OK, ICMP6 can be generated. 395 */ 396 397 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 398 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 399 400 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 401 M_PREPEND(m, preplen, M_DONTWAIT); 402 if (m && M_UNWRITABLE(m, preplen)) 403 m = m_pullup(m, preplen); 404 if (m == NULL) { 405 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 406 return; 407 } 408 409 nip6 = mtod(m, struct ip6_hdr *); 410 nip6->ip6_src = oip6->ip6_src; 411 nip6->ip6_dst = oip6->ip6_dst; 412 413 in6_clearscope(&oip6->ip6_src); 414 in6_clearscope(&oip6->ip6_dst); 415 416 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 417 icmp6->icmp6_type = type; 418 icmp6->icmp6_code = code; 419 icmp6->icmp6_pptr = htonl((u_int32_t)param); 420 421 /* 422 * icmp6_reflect() is designed to be in the input path. 423 * icmp6_error() can be called from both input and output path, 424 * and if we are in output path rcvif could contain bogus value. 425 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 426 * information in ip header (nip6). 427 */ 428 m->m_pkthdr.rcvif = NULL; 429 430 ICMP6_STATINC(ICMP6_STAT_OUTHIST + type); 431 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 432 433 return; 434 435 freeit: 436 /* 437 * If we can't tell whether or not we can generate ICMP6, free it. 438 */ 439 m_freem(m); 440 } 441 442 /* 443 * Process a received ICMP6 message. 444 */ 445 int 446 icmp6_input(struct mbuf **mp, int *offp, int proto) 447 { 448 struct mbuf *m = *mp, *n; 449 struct ip6_hdr *ip6, *nip6; 450 struct icmp6_hdr *icmp6, *nicmp6; 451 int off = *offp; 452 int icmp6len = m->m_pkthdr.len - *offp; 453 int code, sum, noff, i; 454 455 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4) 456 KASSERT(ICMP6_MAXLEN < MCLBYTES); 457 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 458 459 /* 460 * Locate icmp6 structure in mbuf, and check 461 * that not corrupted and of at least minimum length 462 */ 463 464 if (icmp6len < sizeof(struct icmp6_hdr)) { 465 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 466 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 467 goto freeit; 468 } 469 470 i = off + sizeof(*icmp6); 471 if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) { 472 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 473 #if 0 /* m is 0 here */ 474 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 475 #endif 476 goto freeit; 477 } 478 ip6 = mtod(m, struct ip6_hdr *); 479 /* 480 * calculate the checksum 481 */ 482 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 483 if (icmp6 == NULL) { 484 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 485 /* m is invalid */ 486 /*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/ 487 return IPPROTO_DONE; 488 } 489 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 490 code = icmp6->icmp6_code; 491 492 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 493 nd6log((LOG_ERR, 494 "ICMP6 checksum error(%d|%x) %s\n", 495 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 496 ICMP6_STATINC(ICMP6_STAT_CHECKSUM); 497 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 498 goto freeit; 499 } 500 501 #if defined(NFAITH) && 0 < NFAITH 502 if (faithprefix(&ip6->ip6_dst)) { 503 /* 504 * Deliver very specific ICMP6 type only. 505 * This is important to deliver TOOBIG. Otherwise PMTUD 506 * will not work. 507 */ 508 switch (icmp6->icmp6_type) { 509 case ICMP6_DST_UNREACH: 510 case ICMP6_PACKET_TOO_BIG: 511 case ICMP6_TIME_EXCEEDED: 512 break; 513 default: 514 goto freeit; 515 } 516 } 517 #endif 518 519 ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type); 520 521 switch (icmp6->icmp6_type) { 522 case ICMP6_DST_UNREACH: 523 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 524 switch (code) { 525 case ICMP6_DST_UNREACH_NOROUTE: 526 code = PRC_UNREACH_NET; 527 break; 528 case ICMP6_DST_UNREACH_ADMIN: 529 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 530 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 531 break; 532 case ICMP6_DST_UNREACH_ADDR: 533 code = PRC_HOSTDEAD; 534 break; 535 #ifdef COMPAT_RFC1885 536 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 537 code = PRC_UNREACH_SRCFAIL; 538 break; 539 #else 540 case ICMP6_DST_UNREACH_BEYONDSCOPE: 541 /* I mean "source address was incorrect." */ 542 code = PRC_UNREACH_NET; 543 break; 544 #endif 545 case ICMP6_DST_UNREACH_NOPORT: 546 code = PRC_UNREACH_PORT; 547 break; 548 default: 549 goto badcode; 550 } 551 goto deliver; 552 553 case ICMP6_PACKET_TOO_BIG: 554 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 555 556 /* 557 * MTU is checked in icmp6_mtudisc. 558 */ 559 code = PRC_MSGSIZE; 560 561 /* 562 * Updating the path MTU will be done after examining 563 * intermediate extension headers. 564 */ 565 goto deliver; 566 567 case ICMP6_TIME_EXCEEDED: 568 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 569 switch (code) { 570 case ICMP6_TIME_EXCEED_TRANSIT: 571 code = PRC_TIMXCEED_INTRANS; 572 break; 573 case ICMP6_TIME_EXCEED_REASSEMBLY: 574 code = PRC_TIMXCEED_REASS; 575 break; 576 default: 577 goto badcode; 578 } 579 goto deliver; 580 581 case ICMP6_PARAM_PROB: 582 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 583 switch (code) { 584 case ICMP6_PARAMPROB_NEXTHEADER: 585 code = PRC_UNREACH_PROTOCOL; 586 break; 587 case ICMP6_PARAMPROB_HEADER: 588 case ICMP6_PARAMPROB_OPTION: 589 code = PRC_PARAMPROB; 590 break; 591 default: 592 goto badcode; 593 } 594 goto deliver; 595 596 case ICMP6_ECHO_REQUEST: 597 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 598 if (code != 0) 599 goto badcode; 600 /* 601 * Copy mbuf to send to two data paths: userland socket(s), 602 * and to the querier (echo reply). 603 * m: a copy for socket, n: a copy for querier 604 * 605 * If the first mbuf is shared, or the first mbuf is too short, 606 * copy the first part of the data into a fresh mbuf. 607 * Otherwise, we will wrongly overwrite both copies. 608 */ 609 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 610 /* Give up local */ 611 n = m; 612 m = NULL; 613 } else if (M_READONLY(n) || 614 n->m_len < off + sizeof(struct icmp6_hdr)) { 615 struct mbuf *n0 = n; 616 617 /* 618 * Prepare an internal mbuf. m_pullup() doesn't 619 * always copy the length we specified. 620 */ 621 if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 622 /* Give up local */ 623 n = m; 624 m = NULL; 625 } 626 m_freem(n0); 627 } 628 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, 629 sizeof(*nicmp6)); 630 if (nicmp6 == NULL) 631 goto freeit; 632 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 633 nicmp6->icmp6_code = 0; 634 if (n) { 635 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 636 icmp6s[ICMP6_STAT_REFLECT]++; 637 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++; 638 ICMP6_STAT_PUTREF(); 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 memset(p, 0, 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 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 747 icmp6s[ICMP6_STAT_REFLECT]++; 748 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++; 749 ICMP6_STAT_PUTREF(); 750 icmp6_reflect(n, noff); 751 } 752 break; 753 } 754 755 case ICMP6_WRUREPLY: 756 if (code != 0) 757 goto badcode; 758 break; 759 760 case ND_ROUTER_SOLICIT: 761 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 762 if (code != 0) 763 goto badcode; 764 if (icmp6len < sizeof(struct nd_router_solicit)) 765 goto badlen; 766 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 767 /* give up local */ 768 nd6_rs_input(m, off, icmp6len); 769 m = NULL; 770 goto freeit; 771 } 772 nd6_rs_input(n, off, icmp6len); 773 /* m stays. */ 774 break; 775 776 case ND_ROUTER_ADVERT: 777 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 778 if (code != 0) 779 goto badcode; 780 if (icmp6len < sizeof(struct nd_router_advert)) 781 goto badlen; 782 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 783 /* give up local */ 784 nd6_ra_input(m, off, icmp6len); 785 m = NULL; 786 goto freeit; 787 } 788 nd6_ra_input(n, off, icmp6len); 789 /* m stays. */ 790 break; 791 792 case ND_NEIGHBOR_SOLICIT: 793 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 794 if (code != 0) 795 goto badcode; 796 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 797 goto badlen; 798 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 799 /* give up local */ 800 nd6_ns_input(m, off, icmp6len); 801 m = NULL; 802 goto freeit; 803 } 804 nd6_ns_input(n, off, icmp6len); 805 /* m stays. */ 806 break; 807 808 case ND_NEIGHBOR_ADVERT: 809 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 810 if (code != 0) 811 goto badcode; 812 if (icmp6len < sizeof(struct nd_neighbor_advert)) 813 goto badlen; 814 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 815 /* give up local */ 816 nd6_na_input(m, off, icmp6len); 817 m = NULL; 818 goto freeit; 819 } 820 nd6_na_input(n, off, icmp6len); 821 /* m stays. */ 822 break; 823 824 case ND_REDIRECT: 825 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 826 if (code != 0) 827 goto badcode; 828 if (icmp6len < sizeof(struct nd_redirect)) 829 goto badlen; 830 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 831 /* give up local */ 832 icmp6_redirect_input(m, off); 833 m = NULL; 834 goto freeit; 835 } 836 icmp6_redirect_input(n, off); 837 /* m stays. */ 838 break; 839 840 case ICMP6_ROUTER_RENUMBERING: 841 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 842 code != ICMP6_ROUTER_RENUMBERING_RESULT) 843 goto badcode; 844 if (icmp6len < sizeof(struct icmp6_router_renum)) 845 goto badlen; 846 break; 847 848 default: 849 nd6log((LOG_DEBUG, 850 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 851 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 852 ip6_sprintf(&ip6->ip6_dst), 853 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 854 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 855 /* ICMPv6 error: MUST deliver it by spec... */ 856 code = PRC_NCMDS; 857 /* deliver */ 858 } else { 859 /* ICMPv6 informational: MUST not deliver */ 860 break; 861 } 862 deliver: 863 if (icmp6_notify_error(m, off, icmp6len, code)) { 864 /* In this case, m should've been freed. */ 865 return (IPPROTO_DONE); 866 } 867 break; 868 869 badcode: 870 ICMP6_STATINC(ICMP6_STAT_BADCODE); 871 break; 872 873 badlen: 874 ICMP6_STATINC(ICMP6_STAT_BADLEN); 875 break; 876 } 877 878 /* deliver the packet to appropriate sockets */ 879 icmp6_rip6_input(&m, *offp); 880 881 return IPPROTO_DONE; 882 883 freeit: 884 m_freem(m); 885 return IPPROTO_DONE; 886 } 887 888 static int 889 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int 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 ICMP6_STATINC(ICMP6_STAT_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 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 904 return (-1); 905 } 906 eip6 = (struct ip6_hdr *)(icmp6 + 1); 907 908 /* Detect the upper level protocol */ 909 { 910 void (*ctlfunc)(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 ICMP6_STATINC(ICMP6_STAT_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 ICMP6_STATINC(ICMP6_STAT_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 ICMP6_STATINC(ICMP6_STAT_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 ICMP6_STATINC(ICMP6_STAT_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 ICMP6_STATINC(ICMP6_STAT_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 sockaddr_in6_init(&icmp6dst, 1035 (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0); 1036 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1037 goto freeit; 1038 sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0); 1039 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1040 goto freeit; 1041 icmp6src.sin6_flowinfo = 1042 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1043 1044 if (finaldst == NULL) 1045 finaldst = &eip6->ip6_dst; 1046 ip6cp.ip6c_m = m; 1047 ip6cp.ip6c_icmp6 = icmp6; 1048 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1049 ip6cp.ip6c_off = eoff; 1050 ip6cp.ip6c_finaldst = finaldst; 1051 ip6cp.ip6c_src = &icmp6src; 1052 ip6cp.ip6c_nxt = nxt; 1053 1054 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1055 notifymtu = ntohl(icmp6->icmp6_mtu); 1056 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1057 } 1058 1059 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1060 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1061 if (ctlfunc) { 1062 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1063 &ip6cp); 1064 } 1065 } 1066 return (0); 1067 1068 freeit: 1069 m_freem(m); 1070 return (-1); 1071 } 1072 1073 void 1074 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1075 { 1076 unsigned long rtcount; 1077 struct icmp6_mtudisc_callback *mc; 1078 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1079 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1080 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1081 u_int mtu = ntohl(icmp6->icmp6_mtu); 1082 struct rtentry *rt = NULL; 1083 struct sockaddr_in6 sin6; 1084 1085 /* 1086 * The MTU should not be less than the minimal IPv6 MTU except for the 1087 * hack in ip6_output/ip6_setpmtu where we always include a frag header. 1088 * In that one case, the MTU might be less than 1280. 1089 */ 1090 if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) { 1091 /* is the mtu even sane? */ 1092 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) 1093 return; 1094 if (!validated) 1095 return; 1096 mtu = IPV6_MMTU - sizeof(struct ip6_frag); 1097 } 1098 1099 /* 1100 * allow non-validated cases if memory is plenty, to make traffic 1101 * from non-connected pcb happy. 1102 */ 1103 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1104 if (validated) { 1105 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1106 return; 1107 else if (0 <= icmp6_mtudisc_lowat && 1108 rtcount > icmp6_mtudisc_lowat) { 1109 /* 1110 * XXX nuke a victim, install the new one. 1111 */ 1112 } 1113 } else { 1114 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1115 return; 1116 } 1117 1118 memset(&sin6, 0, sizeof(sin6)); 1119 sin6.sin6_family = PF_INET6; 1120 sin6.sin6_len = sizeof(struct sockaddr_in6); 1121 sin6.sin6_addr = *dst; 1122 if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1123 return; 1124 1125 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1126 1127 if (rt && (rt->rt_flags & RTF_HOST) && 1128 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1129 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1130 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1131 ICMP6_STATINC(ICMP6_STAT_PMTUCHG); 1132 rt->rt_rmx.rmx_mtu = mtu; 1133 } 1134 } 1135 if (rt) { 1136 rtfree(rt); 1137 } 1138 1139 /* 1140 * Notify protocols that the MTU for this destination 1141 * has changed. 1142 */ 1143 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1144 mc = LIST_NEXT(mc, mc_list)) 1145 (*mc->mc_func)(&sin6.sin6_addr); 1146 } 1147 1148 /* 1149 * Process a Node Information Query packet, based on 1150 * draft-ietf-ipngwg-icmp-name-lookups-07. 1151 * 1152 * Spec incompatibilities: 1153 * - IPv6 Subject address handling 1154 * - IPv4 Subject address handling support missing 1155 * - Proxy reply (answer even if it's not for me) 1156 * - joins NI group address at in6_ifattach() time only, does not cope 1157 * with hostname changes by sethostname(3) 1158 */ 1159 static struct mbuf * 1160 ni6_input(struct mbuf *m, int off) 1161 { 1162 struct icmp6_nodeinfo *ni6, *nni6; 1163 struct mbuf *n = NULL; 1164 u_int16_t qtype; 1165 int subjlen; 1166 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1167 struct ni_reply_fqdn *fqdn; 1168 int addrs; /* for NI_QTYPE_NODEADDR */ 1169 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1170 struct sockaddr_in6 sin6; /* ip6_dst */ 1171 struct in6_addr in6_subj; /* subject address */ 1172 struct ip6_hdr *ip6; 1173 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1174 char *subj = NULL; 1175 1176 ip6 = mtod(m, struct ip6_hdr *); 1177 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1178 if (ni6 == NULL) { 1179 /* m is already reclaimed */ 1180 return NULL; 1181 } 1182 1183 /* 1184 * Validate IPv6 destination address. 1185 * 1186 * The Responder must discard the Query without further processing 1187 * unless it is one of the Responder's unicast or anycast addresses, or 1188 * a link-local scope multicast address which the Responder has joined. 1189 * [icmp-name-lookups-07, Section 4.] 1190 */ 1191 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 1192 /* XXX scopeid */ 1193 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1194 ; /* unicast/anycast, fine */ 1195 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1196 ; /* link-local multicast, fine */ 1197 else 1198 goto bad; 1199 1200 /* validate query Subject field. */ 1201 qtype = ntohs(ni6->ni_qtype); 1202 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1203 switch (qtype) { 1204 case NI_QTYPE_NOOP: 1205 case NI_QTYPE_SUPTYPES: 1206 /* 07 draft */ 1207 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1208 break; 1209 /* FALLTHROUGH */ 1210 case NI_QTYPE_FQDN: 1211 case NI_QTYPE_NODEADDR: 1212 case NI_QTYPE_IPV4ADDR: 1213 switch (ni6->ni_code) { 1214 case ICMP6_NI_SUBJ_IPV6: 1215 #if ICMP6_NI_SUBJ_IPV6 != 0 1216 case 0: 1217 #endif 1218 /* 1219 * backward compatibility - try to accept 03 draft 1220 * format, where no Subject is present. 1221 */ 1222 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1223 subjlen == 0) { 1224 oldfqdn++; 1225 break; 1226 } 1227 #if ICMP6_NI_SUBJ_IPV6 != 0 1228 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1229 goto bad; 1230 #endif 1231 1232 if (subjlen != sizeof(sin6.sin6_addr)) 1233 goto bad; 1234 1235 /* 1236 * Validate Subject address. 1237 * 1238 * Not sure what exactly "address belongs to the node" 1239 * means in the spec, is it just unicast, or what? 1240 * 1241 * At this moment we consider Subject address as 1242 * "belong to the node" if the Subject address equals 1243 * to the IPv6 destination address; validation for 1244 * IPv6 destination address should have done enough 1245 * check for us. 1246 * 1247 * We do not do proxy at this moment. 1248 */ 1249 /* m_pulldown instead of copy? */ 1250 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1251 subjlen, (void *)&in6_subj); 1252 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1253 goto bad; 1254 1255 subj = (char *)&in6_subj; 1256 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1257 break; 1258 1259 /* 1260 * XXX if we are to allow other cases, we should really 1261 * be careful about scope here. 1262 * basically, we should disallow queries toward IPv6 1263 * destination X with subject Y, if scope(X) > scope(Y). 1264 * if we allow scope(X) > scope(Y), it will result in 1265 * information leakage across scope boundary. 1266 */ 1267 goto bad; 1268 1269 case ICMP6_NI_SUBJ_FQDN: 1270 /* 1271 * Validate Subject name with gethostname(3). 1272 * 1273 * The behavior may need some debate, since: 1274 * - we are not sure if the node has FQDN as 1275 * hostname (returned by gethostname(3)). 1276 * - the code does wildcard match for truncated names. 1277 * however, we are not sure if we want to perform 1278 * wildcard match, if gethostname(3) side has 1279 * truncated hostname. 1280 */ 1281 n = ni6_nametodns(hostname, hostnamelen, 0); 1282 if (!n || n->m_next || n->m_len == 0) 1283 goto bad; 1284 IP6_EXTHDR_GET(subj, char *, m, 1285 off + sizeof(struct icmp6_nodeinfo), subjlen); 1286 if (subj == NULL) 1287 goto bad; 1288 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1289 n->m_len)) { 1290 goto bad; 1291 } 1292 m_freem(n); 1293 n = NULL; 1294 break; 1295 1296 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1297 default: 1298 goto bad; 1299 } 1300 break; 1301 } 1302 1303 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1304 switch (qtype) { 1305 case NI_QTYPE_FQDN: 1306 if ((icmp6_nodeinfo & 1) == 0) 1307 goto bad; 1308 break; 1309 case NI_QTYPE_NODEADDR: 1310 case NI_QTYPE_IPV4ADDR: 1311 if ((icmp6_nodeinfo & 2) == 0) 1312 goto bad; 1313 break; 1314 } 1315 1316 /* guess reply length */ 1317 switch (qtype) { 1318 case NI_QTYPE_NOOP: 1319 break; /* no reply data */ 1320 case NI_QTYPE_SUPTYPES: 1321 replylen += sizeof(u_int32_t); 1322 break; 1323 case NI_QTYPE_FQDN: 1324 /* XXX will append an mbuf */ 1325 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1326 break; 1327 case NI_QTYPE_NODEADDR: 1328 addrs = ni6_addrs(ni6, m, &ifp, subj); 1329 if ((replylen += addrs * (sizeof(struct in6_addr) + 1330 sizeof(u_int32_t))) > MCLBYTES) 1331 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1332 break; 1333 case NI_QTYPE_IPV4ADDR: 1334 /* unsupported - should respond with unknown Qtype? */ 1335 goto bad; 1336 default: 1337 /* 1338 * XXX: We must return a reply with the ICMP6 code 1339 * `unknown Qtype' in this case. However we regard the case 1340 * as an FQDN query for backward compatibility. 1341 * Older versions set a random value to this field, 1342 * so it rarely varies in the defined qtypes. 1343 * But the mechanism is not reliable... 1344 * maybe we should obsolete older versions. 1345 */ 1346 qtype = NI_QTYPE_FQDN; 1347 /* XXX will append an mbuf */ 1348 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1349 oldfqdn++; 1350 break; 1351 } 1352 1353 /* allocate an mbuf to reply. */ 1354 MGETHDR(n, M_DONTWAIT, m->m_type); 1355 if (n == NULL) { 1356 m_freem(m); 1357 return (NULL); 1358 } 1359 M_MOVE_PKTHDR(n, m); /* just for rcvif */ 1360 if (replylen > MHLEN) { 1361 if (replylen > MCLBYTES) { 1362 /* 1363 * XXX: should we try to allocate more? But MCLBYTES 1364 * is probably much larger than IPV6_MMTU... 1365 */ 1366 goto bad; 1367 } 1368 MCLGET(n, M_DONTWAIT); 1369 if ((n->m_flags & M_EXT) == 0) { 1370 goto bad; 1371 } 1372 } 1373 n->m_pkthdr.len = n->m_len = replylen; 1374 1375 /* copy mbuf header and IPv6 + Node Information base headers */ 1376 bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr)); 1377 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1378 bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo)); 1379 1380 /* qtype dependent procedure */ 1381 switch (qtype) { 1382 case NI_QTYPE_NOOP: 1383 nni6->ni_code = ICMP6_NI_SUCCESS; 1384 nni6->ni_flags = 0; 1385 break; 1386 case NI_QTYPE_SUPTYPES: 1387 { 1388 u_int32_t v; 1389 nni6->ni_code = ICMP6_NI_SUCCESS; 1390 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1391 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1392 v = (u_int32_t)htonl(0x0000000f); 1393 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1394 break; 1395 } 1396 case NI_QTYPE_FQDN: 1397 nni6->ni_code = ICMP6_NI_SUCCESS; 1398 fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) + 1399 sizeof(struct ip6_hdr) + 1400 sizeof(struct icmp6_nodeinfo)); 1401 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1402 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1403 /* 1404 * XXX do we really have FQDN in variable "hostname"? 1405 */ 1406 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1407 if (n->m_next == NULL) 1408 goto bad; 1409 /* XXX we assume that n->m_next is not a chain */ 1410 if (n->m_next->m_next != NULL) 1411 goto bad; 1412 n->m_pkthdr.len += n->m_next->m_len; 1413 break; 1414 case NI_QTYPE_NODEADDR: 1415 { 1416 int lenlim, copied; 1417 1418 nni6->ni_code = ICMP6_NI_SUCCESS; 1419 n->m_pkthdr.len = n->m_len = 1420 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1421 lenlim = M_TRAILINGSPACE(n); 1422 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1423 /* XXX: reset mbuf length */ 1424 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1425 sizeof(struct icmp6_nodeinfo) + copied; 1426 break; 1427 } 1428 default: 1429 break; /* XXX impossible! */ 1430 } 1431 1432 nni6->ni_type = ICMP6_NI_REPLY; 1433 m_freem(m); 1434 return (n); 1435 1436 bad: 1437 m_freem(m); 1438 if (n) 1439 m_freem(n); 1440 return (NULL); 1441 } 1442 #undef hostnamelen 1443 1444 #define isupper(x) ('A' <= (x) && (x) <= 'Z') 1445 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1446 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1447 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1448 1449 /* 1450 * make a mbuf with DNS-encoded string. no compression support. 1451 * 1452 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1453 * treated as truncated name (two \0 at the end). this is a wild guess. 1454 * 1455 * old - return pascal string if non-zero 1456 */ 1457 static struct mbuf * 1458 ni6_nametodns(const char *name, int namelen, int old) 1459 { 1460 struct mbuf *m; 1461 char *cp, *ep; 1462 const char *p, *q; 1463 int i, len, nterm; 1464 1465 if (old) 1466 len = namelen + 1; 1467 else 1468 len = MCLBYTES; 1469 1470 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1471 MGET(m, M_DONTWAIT, MT_DATA); 1472 if (m && len > MLEN) { 1473 MCLGET(m, M_DONTWAIT); 1474 if ((m->m_flags & M_EXT) == 0) 1475 goto fail; 1476 } 1477 if (!m) 1478 goto fail; 1479 m->m_next = NULL; 1480 1481 if (old) { 1482 m->m_len = len; 1483 *mtod(m, char *) = namelen; 1484 bcopy(name, mtod(m, char *) + 1, namelen); 1485 return m; 1486 } else { 1487 m->m_len = 0; 1488 cp = mtod(m, char *); 1489 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1490 1491 /* if not certain about my name, return empty buffer */ 1492 if (namelen == 0) 1493 return m; 1494 1495 /* 1496 * guess if it looks like shortened hostname, or FQDN. 1497 * shortened hostname needs two trailing "\0". 1498 */ 1499 i = 0; 1500 for (p = name; p < name + namelen; p++) { 1501 if (*p && *p == '.') 1502 i++; 1503 } 1504 if (i < 2) 1505 nterm = 2; 1506 else 1507 nterm = 1; 1508 1509 p = name; 1510 while (cp < ep && p < name + namelen) { 1511 i = 0; 1512 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1513 i++; 1514 /* result does not fit into mbuf */ 1515 if (cp + i + 1 >= ep) 1516 goto fail; 1517 /* 1518 * DNS label length restriction, RFC1035 page 8. 1519 * "i == 0" case is included here to avoid returning 1520 * 0-length label on "foo..bar". 1521 */ 1522 if (i <= 0 || i >= 64) 1523 goto fail; 1524 *cp++ = i; 1525 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1526 goto fail; 1527 while (i > 0) { 1528 if (!isalnum(*p) && *p != '-') 1529 goto fail; 1530 if (isupper(*p)) { 1531 *cp++ = tolower(*p); 1532 p++; 1533 } else 1534 *cp++ = *p++; 1535 i--; 1536 } 1537 p = q; 1538 if (p < name + namelen && *p == '.') 1539 p++; 1540 } 1541 /* termination */ 1542 if (cp + nterm >= ep) 1543 goto fail; 1544 while (nterm-- > 0) 1545 *cp++ = '\0'; 1546 m->m_len = cp - mtod(m, char *); 1547 return m; 1548 } 1549 1550 panic("should not reach here"); 1551 /* NOTREACHED */ 1552 1553 fail: 1554 if (m) 1555 m_freem(m); 1556 return NULL; 1557 } 1558 1559 /* 1560 * check if two DNS-encoded string matches. takes care of truncated 1561 * form (with \0\0 at the end). no compression support. 1562 * XXX upper/lowercase match (see RFC2065) 1563 */ 1564 static int 1565 ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1566 { 1567 const char *a0, *b0; 1568 int l; 1569 1570 /* simplest case - need validation? */ 1571 if (alen == blen && memcmp(a, b, alen) == 0) 1572 return 1; 1573 1574 a0 = a; 1575 b0 = b; 1576 1577 /* termination is mandatory */ 1578 if (alen < 2 || blen < 2) 1579 return 0; 1580 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1581 return 0; 1582 alen--; 1583 blen--; 1584 1585 while (a - a0 < alen && b - b0 < blen) { 1586 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1587 return 0; 1588 1589 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1590 return 0; 1591 /* we don't support compression yet */ 1592 if (a[0] >= 64 || b[0] >= 64) 1593 return 0; 1594 1595 /* truncated case */ 1596 if (a[0] == 0 && a - a0 == alen - 1) 1597 return 1; 1598 if (b[0] == 0 && b - b0 == blen - 1) 1599 return 1; 1600 if (a[0] == 0 || b[0] == 0) 1601 return 0; 1602 1603 if (a[0] != b[0]) 1604 return 0; 1605 l = a[0]; 1606 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1607 return 0; 1608 if (memcmp(a + 1, b + 1, l) != 0) 1609 return 0; 1610 1611 a += 1 + l; 1612 b += 1 + l; 1613 } 1614 1615 if (a - a0 == alen && b - b0 == blen) 1616 return 1; 1617 else 1618 return 0; 1619 } 1620 1621 /* 1622 * calculate the number of addresses to be returned in the node info reply. 1623 */ 1624 static int 1625 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, 1626 struct ifnet **ifpp, char *subj) 1627 { 1628 struct ifnet *ifp; 1629 struct in6_ifaddr *ifa6; 1630 struct ifaddr *ifa; 1631 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1632 int addrs = 0, addrsofif, iffound = 0; 1633 int niflags = ni6->ni_flags; 1634 1635 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1636 switch (ni6->ni_code) { 1637 case ICMP6_NI_SUBJ_IPV6: 1638 if (subj == NULL) /* must be impossible... */ 1639 return (0); 1640 subj_ip6 = (struct sockaddr_in6 *)subj; 1641 break; 1642 default: 1643 /* 1644 * XXX: we only support IPv6 subject address for 1645 * this Qtype. 1646 */ 1647 return (0); 1648 } 1649 } 1650 1651 IFNET_FOREACH(ifp) { 1652 addrsofif = 0; 1653 IFADDR_FOREACH(ifa, ifp) { 1654 if (ifa->ifa_addr->sa_family != AF_INET6) 1655 continue; 1656 ifa6 = (struct in6_ifaddr *)ifa; 1657 1658 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1659 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1660 &ifa6->ia_addr.sin6_addr)) 1661 iffound = 1; 1662 1663 /* 1664 * IPv4-mapped addresses can only be returned by a 1665 * Node Information proxy, since they represent 1666 * addresses of IPv4-only nodes, which perforce do 1667 * not implement this protocol. 1668 * [icmp-name-lookups-07, Section 5.4] 1669 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1670 * this function at this moment. 1671 */ 1672 1673 /* What do we have to do about ::1? */ 1674 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1675 case IPV6_ADDR_SCOPE_LINKLOCAL: 1676 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1677 continue; 1678 break; 1679 case IPV6_ADDR_SCOPE_SITELOCAL: 1680 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1681 continue; 1682 break; 1683 case IPV6_ADDR_SCOPE_GLOBAL: 1684 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1685 continue; 1686 break; 1687 default: 1688 continue; 1689 } 1690 1691 /* 1692 * check if anycast is okay. 1693 * XXX: just experimental. not in the spec. 1694 */ 1695 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1696 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1697 continue; /* we need only unicast addresses */ 1698 1699 addrsofif++; /* count the address */ 1700 } 1701 if (iffound) { 1702 *ifpp = ifp; 1703 return (addrsofif); 1704 } 1705 1706 addrs += addrsofif; 1707 } 1708 1709 return (addrs); 1710 } 1711 1712 static int 1713 ni6_store_addrs(struct icmp6_nodeinfo *ni6, 1714 struct icmp6_nodeinfo *nni6, struct ifnet *ifp0, 1715 int resid) 1716 { 1717 struct ifnet *ifp = ifp0 ? ifp0 : IFNET_FIRST(); 1718 struct in6_ifaddr *ifa6; 1719 struct ifaddr *ifa; 1720 struct ifnet *ifp_dep = NULL; 1721 int copied = 0, allow_deprecated = 0; 1722 u_char *cp = (u_char *)(nni6 + 1); 1723 int niflags = ni6->ni_flags; 1724 u_int32_t ltime; 1725 1726 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1727 return (0); /* needless to copy */ 1728 1729 again: 1730 1731 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1732 { 1733 IFADDR_FOREACH(ifa, ifp) { 1734 if (ifa->ifa_addr->sa_family != AF_INET6) 1735 continue; 1736 ifa6 = (struct in6_ifaddr *)ifa; 1737 1738 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1739 allow_deprecated == 0) { 1740 /* 1741 * prefererred address should be put before 1742 * deprecated addresses. 1743 */ 1744 1745 /* record the interface for later search */ 1746 if (ifp_dep == NULL) 1747 ifp_dep = ifp; 1748 1749 continue; 1750 } 1751 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1752 allow_deprecated != 0) 1753 continue; /* we now collect deprecated addrs */ 1754 1755 /* What do we have to do about ::1? */ 1756 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1757 case IPV6_ADDR_SCOPE_LINKLOCAL: 1758 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1759 continue; 1760 break; 1761 case IPV6_ADDR_SCOPE_SITELOCAL: 1762 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1763 continue; 1764 break; 1765 case IPV6_ADDR_SCOPE_GLOBAL: 1766 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1767 continue; 1768 break; 1769 default: 1770 continue; 1771 } 1772 1773 /* 1774 * check if anycast is okay. 1775 * XXX: just experimental. not in the spec. 1776 */ 1777 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1778 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1779 continue; 1780 1781 /* now we can copy the address */ 1782 if (resid < sizeof(struct in6_addr) + 1783 sizeof(u_int32_t)) { 1784 /* 1785 * We give up much more copy. 1786 * Set the truncate flag and return. 1787 */ 1788 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1789 return (copied); 1790 } 1791 1792 /* 1793 * Set the TTL of the address. 1794 * The TTL value should be one of the following 1795 * according to the specification: 1796 * 1797 * 1. The remaining lifetime of a DHCP lease on the 1798 * address, or 1799 * 2. The remaining Valid Lifetime of a prefix from 1800 * which the address was derived through Stateless 1801 * Autoconfiguration. 1802 * 1803 * Note that we currently do not support stateful 1804 * address configuration by DHCPv6, so the former 1805 * case can't happen. 1806 * 1807 * TTL must be 2^31 > TTL >= 0. 1808 */ 1809 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1810 ltime = ND6_INFINITE_LIFETIME; 1811 else { 1812 if (ifa6->ia6_lifetime.ia6t_expire > 1813 time_second) 1814 ltime = ifa6->ia6_lifetime.ia6t_expire - 1815 time_second; 1816 else 1817 ltime = 0; 1818 } 1819 if (ltime > 0x7fffffff) 1820 ltime = 0x7fffffff; 1821 ltime = htonl(ltime); 1822 1823 bcopy(<ime, cp, sizeof(u_int32_t)); 1824 cp += sizeof(u_int32_t); 1825 1826 /* copy the address itself */ 1827 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1828 sizeof(struct in6_addr)); 1829 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1830 cp += sizeof(struct in6_addr); 1831 1832 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1833 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1834 } 1835 if (ifp0) /* we need search only on the specified IF */ 1836 break; 1837 } 1838 1839 if (allow_deprecated == 0 && ifp_dep != NULL) { 1840 ifp = ifp_dep; 1841 allow_deprecated = 1; 1842 1843 goto again; 1844 } 1845 1846 return (copied); 1847 } 1848 1849 /* 1850 * XXX almost dup'ed code with rip6_input. 1851 */ 1852 static int 1853 icmp6_rip6_input(struct mbuf **mp, int off) 1854 { 1855 struct mbuf *m = *mp; 1856 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1857 struct inpcb_hdr *inph; 1858 struct in6pcb *in6p; 1859 struct in6pcb *last = NULL; 1860 struct sockaddr_in6 rip6src; 1861 struct icmp6_hdr *icmp6; 1862 struct mbuf *opts = NULL; 1863 1864 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1865 if (icmp6 == NULL) { 1866 /* m is already reclaimed */ 1867 return IPPROTO_DONE; 1868 } 1869 1870 /* 1871 * XXX: the address may have embedded scope zone ID, which should be 1872 * hidden from applications. 1873 */ 1874 sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0); 1875 if (sa6_recoverscope(&rip6src)) { 1876 m_freem(m); 1877 return (IPPROTO_DONE); 1878 } 1879 1880 TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) { 1881 in6p = (struct in6pcb *)inph; 1882 if (in6p->in6p_af != AF_INET6) 1883 continue; 1884 if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6) 1885 continue; 1886 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1887 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1888 continue; 1889 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1890 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1891 continue; 1892 if (in6p->in6p_icmp6filt 1893 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1894 in6p->in6p_icmp6filt)) 1895 continue; 1896 if (last) { 1897 struct mbuf *n; 1898 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1899 if (last->in6p_flags & IN6P_CONTROLOPTS) 1900 ip6_savecontrol(last, &opts, ip6, n); 1901 /* strip intermediate headers */ 1902 m_adj(n, off); 1903 if (sbappendaddr(&last->in6p_socket->so_rcv, 1904 (struct sockaddr *)&rip6src, 1905 n, opts) == 0) { 1906 /* should notify about lost packet */ 1907 m_freem(n); 1908 if (opts) 1909 m_freem(opts); 1910 } else 1911 sorwakeup(last->in6p_socket); 1912 opts = NULL; 1913 } 1914 } 1915 last = in6p; 1916 } 1917 if (last) { 1918 if (last->in6p_flags & IN6P_CONTROLOPTS) 1919 ip6_savecontrol(last, &opts, ip6, m); 1920 /* strip intermediate headers */ 1921 m_adj(m, off); 1922 if (sbappendaddr(&last->in6p_socket->so_rcv, 1923 (struct sockaddr *)&rip6src, m, opts) == 0) { 1924 m_freem(m); 1925 if (opts) 1926 m_freem(opts); 1927 } else 1928 sorwakeup(last->in6p_socket); 1929 } else { 1930 m_freem(m); 1931 IP6_STATDEC(IP6_STAT_DELIVERED); 1932 } 1933 return IPPROTO_DONE; 1934 } 1935 1936 /* 1937 * Reflect the ip6 packet back to the source. 1938 * OFF points to the icmp6 header, counted from the top of the mbuf. 1939 * 1940 * Note: RFC 1885 required that an echo reply should be truncated if it 1941 * did not fit in with (return) path MTU, and KAME code supported the 1942 * behavior. However, as a clarification after the RFC, this limitation 1943 * was removed in a revised version of the spec, RFC 2463. We had kept the 1944 * old behavior, with a (non-default) ifdef block, while the new version of 1945 * the spec was an internet-draft status, and even after the new RFC was 1946 * published. But it would rather make sense to clean the obsoleted part 1947 * up, and to make the code simpler at this stage. 1948 */ 1949 void 1950 icmp6_reflect(struct mbuf *m, size_t off) 1951 { 1952 struct ip6_hdr *ip6; 1953 struct icmp6_hdr *icmp6; 1954 const struct in6_ifaddr *ia; 1955 const struct ip6aux *ip6a; 1956 int plen; 1957 int type, code; 1958 struct ifnet *outif = NULL; 1959 struct in6_addr origdst; 1960 const struct in6_addr *src = NULL; 1961 1962 /* too short to reflect */ 1963 if (off < sizeof(struct ip6_hdr)) { 1964 nd6log((LOG_DEBUG, 1965 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 1966 (u_long)off, (u_long)sizeof(struct ip6_hdr), 1967 __FILE__, __LINE__)); 1968 goto bad; 1969 } 1970 1971 /* 1972 * If there are extra headers between IPv6 and ICMPv6, strip 1973 * off that header first. 1974 */ 1975 #ifdef DIAGNOSTIC 1976 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 1977 panic("assumption failed in icmp6_reflect"); 1978 #endif 1979 if (off > sizeof(struct ip6_hdr)) { 1980 size_t l; 1981 struct ip6_hdr nip6; 1982 1983 l = off - sizeof(struct ip6_hdr); 1984 m_copydata(m, 0, sizeof(nip6), (void *)&nip6); 1985 m_adj(m, l); 1986 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1987 if (m->m_len < l) { 1988 if ((m = m_pullup(m, l)) == NULL) 1989 return; 1990 } 1991 bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6)); 1992 } else /* off == sizeof(struct ip6_hdr) */ { 1993 size_t l; 1994 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1995 if (m->m_len < l) { 1996 if ((m = m_pullup(m, l)) == NULL) 1997 return; 1998 } 1999 } 2000 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2001 ip6 = mtod(m, struct ip6_hdr *); 2002 ip6->ip6_nxt = IPPROTO_ICMPV6; 2003 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2004 type = icmp6->icmp6_type; /* keep type for statistics */ 2005 code = icmp6->icmp6_code; /* ditto. */ 2006 2007 origdst = ip6->ip6_dst; 2008 /* 2009 * ip6_input() drops a packet if its src is multicast. 2010 * So, the src is never multicast. 2011 */ 2012 ip6->ip6_dst = ip6->ip6_src; 2013 2014 /* 2015 * If the incoming packet was addressed directly to us (i.e. unicast), 2016 * use dst as the src for the reply. 2017 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2018 * (for example) when we encounter an error while forwarding procedure 2019 * destined to a duplicated address of ours. 2020 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2021 * procedure of an outgoing packet of our own, in which case we need 2022 * to search in the ifaddr list. 2023 */ 2024 if (IN6_IS_ADDR_MULTICAST(&origdst)) 2025 ; 2026 else if ((ip6a = ip6_getdstifaddr(m)) != NULL) { 2027 if ((ip6a->ip6a_flags & 2028 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2029 src = &ip6a->ip6a_src; 2030 } else { 2031 union { 2032 struct sockaddr_in6 sin6; 2033 struct sockaddr sa; 2034 } u; 2035 2036 sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0); 2037 2038 ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa); 2039 2040 if (ia == NULL) 2041 ; 2042 else if ((ia->ia6_flags & 2043 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2044 src = &ia->ia_addr.sin6_addr; 2045 } 2046 2047 if (src == NULL) { 2048 int e; 2049 struct sockaddr_in6 sin6; 2050 struct route ro; 2051 2052 /* 2053 * This case matches to multicasts, our anycast, or unicasts 2054 * that we do not own. Select a source address based on the 2055 * source address of the erroneous packet. 2056 */ 2057 /* zone ID should be embedded */ 2058 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 2059 2060 memset(&ro, 0, sizeof(ro)); 2061 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2062 rtcache_free(&ro); 2063 if (src == NULL) { 2064 nd6log((LOG_DEBUG, 2065 "icmp6_reflect: source can't be determined: " 2066 "dst=%s, error=%d\n", 2067 ip6_sprintf(&sin6.sin6_addr), e)); 2068 goto bad; 2069 } 2070 } 2071 2072 ip6->ip6_src = *src; 2073 ip6->ip6_flow = 0; 2074 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2075 ip6->ip6_vfc |= IPV6_VERSION; 2076 ip6->ip6_nxt = IPPROTO_ICMPV6; 2077 if (m->m_pkthdr.rcvif) { 2078 /* XXX: This may not be the outgoing interface */ 2079 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2080 } else 2081 ip6->ip6_hlim = ip6_defhlim; 2082 2083 m->m_pkthdr.csum_flags = 0; 2084 icmp6->icmp6_cksum = 0; 2085 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2086 sizeof(struct ip6_hdr), plen); 2087 2088 /* 2089 * XXX option handling 2090 */ 2091 2092 m->m_flags &= ~(M_BCAST|M_MCAST); 2093 2094 /* 2095 * To avoid a "too big" situation at an intermediate router 2096 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2097 * Note that only echo and node information replies are affected, 2098 * since the length of ICMP6 errors is limited to the minimum MTU. 2099 */ 2100 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 && 2101 outif) 2102 icmp6_ifstat_inc(outif, ifs6_out_error); 2103 2104 if (outif) 2105 icmp6_ifoutstat_inc(outif, type, code); 2106 2107 return; 2108 2109 bad: 2110 m_freem(m); 2111 return; 2112 } 2113 2114 static const char * 2115 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2116 struct in6_addr *tgt6) 2117 { 2118 static char buf[1024]; 2119 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2120 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2121 return buf; 2122 } 2123 2124 void 2125 icmp6_redirect_input(struct mbuf *m, int off) 2126 { 2127 struct ifnet *ifp = m->m_pkthdr.rcvif; 2128 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2129 struct nd_redirect *nd_rd; 2130 int icmp6len = ntohs(ip6->ip6_plen); 2131 char *lladdr = NULL; 2132 int lladdrlen = 0; 2133 struct rtentry *rt = NULL; 2134 int is_router; 2135 int is_onlink; 2136 struct in6_addr src6 = ip6->ip6_src; 2137 struct in6_addr redtgt6; 2138 struct in6_addr reddst6; 2139 union nd_opts ndopts; 2140 2141 if (!ifp) 2142 return; 2143 2144 /* XXX if we are router, we don't update route by icmp6 redirect */ 2145 if (ip6_forwarding) 2146 goto freeit; 2147 if (!icmp6_rediraccept) 2148 goto freeit; 2149 2150 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2151 if (nd_rd == NULL) { 2152 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 2153 return; 2154 } 2155 redtgt6 = nd_rd->nd_rd_target; 2156 reddst6 = nd_rd->nd_rd_dst; 2157 2158 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2159 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2160 goto freeit; 2161 } 2162 2163 /* validation */ 2164 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2165 nd6log((LOG_ERR, 2166 "ICMP6 redirect sent from %s rejected; " 2167 "must be from linklocal\n", ip6_sprintf(&src6))); 2168 goto bad; 2169 } 2170 if (ip6->ip6_hlim != 255) { 2171 nd6log((LOG_ERR, 2172 "ICMP6 redirect sent from %s rejected; " 2173 "hlim=%d (must be 255)\n", 2174 ip6_sprintf(&src6), ip6->ip6_hlim)); 2175 goto bad; 2176 } 2177 { 2178 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2179 struct sockaddr_in6 sin6; 2180 struct in6_addr *gw6; 2181 2182 sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0); 2183 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2184 if (rt) { 2185 if (rt->rt_gateway == NULL || 2186 rt->rt_gateway->sa_family != AF_INET6) { 2187 nd6log((LOG_ERR, 2188 "ICMP6 redirect rejected; no route " 2189 "with inet6 gateway found for redirect dst: %s\n", 2190 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2191 rtfree(rt); 2192 goto bad; 2193 } 2194 2195 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2196 if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2197 nd6log((LOG_ERR, 2198 "ICMP6 redirect rejected; " 2199 "not equal to gw-for-src=%s (must be same): " 2200 "%s\n", 2201 ip6_sprintf(gw6), 2202 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2203 rtfree(rt); 2204 goto bad; 2205 } 2206 } else { 2207 nd6log((LOG_ERR, 2208 "ICMP6 redirect rejected; " 2209 "no route found for redirect dst: %s\n", 2210 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2211 goto bad; 2212 } 2213 rtfree(rt); 2214 rt = NULL; 2215 } 2216 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2217 nd6log((LOG_ERR, 2218 "ICMP6 redirect rejected; " 2219 "redirect dst must be unicast: %s\n", 2220 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2221 goto bad; 2222 } 2223 2224 is_router = is_onlink = 0; 2225 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2226 is_router = 1; /* router case */ 2227 if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2228 is_onlink = 1; /* on-link destination case */ 2229 if (!is_router && !is_onlink) { 2230 nd6log((LOG_ERR, 2231 "ICMP6 redirect rejected; " 2232 "neither router case nor onlink case: %s\n", 2233 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2234 goto bad; 2235 } 2236 /* validation passed */ 2237 2238 icmp6len -= sizeof(*nd_rd); 2239 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2240 if (nd6_options(&ndopts) < 0) { 2241 nd6log((LOG_INFO, "icmp6_redirect_input: " 2242 "invalid ND option, rejected: %s\n", 2243 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2244 /* nd6_options have incremented stats */ 2245 goto freeit; 2246 } 2247 2248 if (ndopts.nd_opts_tgt_lladdr) { 2249 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2250 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2251 } 2252 2253 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2254 nd6log((LOG_INFO, 2255 "icmp6_redirect_input: lladdrlen mismatch for %s " 2256 "(if %d, icmp6 packet %d): %s\n", 2257 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2258 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2259 goto bad; 2260 } 2261 2262 /* RFC 2461 8.3 */ 2263 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2264 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2265 2266 if (!is_onlink) { /* better router case. perform rtredirect. */ 2267 /* perform rtredirect */ 2268 struct sockaddr_in6 sdst; 2269 struct sockaddr_in6 sgw; 2270 struct sockaddr_in6 ssrc; 2271 unsigned long rtcount; 2272 struct rtentry *newrt = NULL; 2273 2274 /* 2275 * do not install redirect route, if the number of entries 2276 * is too much (> hiwat). note that, the node (= host) will 2277 * work just fine even if we do not install redirect route 2278 * (there will be additional hops, though). 2279 */ 2280 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2281 if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes) 2282 goto freeit; 2283 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2284 return; 2285 else if (0 <= icmp6_redirect_lowat && 2286 rtcount > icmp6_redirect_lowat) { 2287 /* 2288 * XXX nuke a victim, install the new one. 2289 */ 2290 } 2291 2292 memset(&sdst, 0, sizeof(sdst)); 2293 memset(&sgw, 0, sizeof(sgw)); 2294 memset(&ssrc, 0, sizeof(ssrc)); 2295 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2296 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2297 sizeof(struct sockaddr_in6); 2298 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2299 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2300 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2301 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2302 NULL, RTF_GATEWAY | RTF_HOST, 2303 (struct sockaddr *)&ssrc, 2304 &newrt); 2305 2306 if (newrt) { 2307 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2308 icmp6_redirect_timeout_q); 2309 rtfree(newrt); 2310 } 2311 } 2312 /* finally update cached route in each socket via pfctlinput */ 2313 { 2314 struct sockaddr_in6 sdst; 2315 2316 sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0); 2317 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2318 #if defined(IPSEC) 2319 if (ipsec_used) 2320 key_sa_routechange((struct sockaddr *)&sdst); 2321 #endif 2322 } 2323 2324 freeit: 2325 m_freem(m); 2326 return; 2327 2328 bad: 2329 ICMP6_STATINC(ICMP6_STAT_BADREDIRECT); 2330 m_freem(m); 2331 } 2332 2333 void 2334 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2335 { 2336 struct ifnet *ifp; /* my outgoing interface */ 2337 struct in6_addr *ifp_ll6; 2338 struct in6_addr *nexthop; 2339 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2340 struct mbuf *m = NULL; /* newly allocated one */ 2341 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2342 struct nd_redirect *nd_rd; 2343 size_t maxlen; 2344 u_char *p; 2345 struct sockaddr_in6 src_sa; 2346 2347 icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0); 2348 2349 /* if we are not router, we don't send icmp6 redirect */ 2350 if (!ip6_forwarding) 2351 goto fail; 2352 2353 /* sanity check */ 2354 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2355 goto fail; 2356 2357 /* 2358 * Address check: 2359 * the source address must identify a neighbor, and 2360 * the destination address must not be a multicast address 2361 * [RFC 2461, sec 8.2] 2362 */ 2363 sip6 = mtod(m0, struct ip6_hdr *); 2364 sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0); 2365 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2366 goto fail; 2367 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2368 goto fail; /* what should we do here? */ 2369 2370 /* rate limit */ 2371 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2372 goto fail; 2373 2374 /* 2375 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2376 * we almost always ask for an mbuf cluster for simplicity. 2377 * (MHLEN < IPV6_MMTU is almost always true) 2378 */ 2379 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2380 if (m && IPV6_MMTU >= MHLEN) { 2381 #if IPV6_MMTU >= MCLBYTES 2382 _MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT); 2383 #else 2384 MCLGET(m, M_DONTWAIT); 2385 #endif 2386 } 2387 2388 if (!m) 2389 goto fail; 2390 m->m_pkthdr.rcvif = NULL; 2391 m->m_len = 0; 2392 maxlen = M_TRAILINGSPACE(m); 2393 maxlen = min(IPV6_MMTU, maxlen); 2394 /* just for safety */ 2395 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2396 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2397 goto fail; 2398 } 2399 2400 { 2401 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2402 struct in6_ifaddr *ia; 2403 if ((ia = in6ifa_ifpforlinklocal(ifp, 2404 IN6_IFF_NOTREADY| 2405 IN6_IFF_ANYCAST)) == NULL) 2406 goto fail; 2407 ifp_ll6 = &ia->ia_addr.sin6_addr; 2408 } 2409 2410 /* get ip6 linklocal address for the router. */ 2411 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2412 struct sockaddr_in6 *sin6; 2413 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2414 nexthop = &sin6->sin6_addr; 2415 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2416 nexthop = NULL; 2417 } else 2418 nexthop = NULL; 2419 2420 /* ip6 */ 2421 ip6 = mtod(m, struct ip6_hdr *); 2422 ip6->ip6_flow = 0; 2423 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2424 ip6->ip6_vfc |= IPV6_VERSION; 2425 /* ip6->ip6_plen will be set later */ 2426 ip6->ip6_nxt = IPPROTO_ICMPV6; 2427 ip6->ip6_hlim = 255; 2428 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2429 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2430 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2431 2432 /* ND Redirect */ 2433 nd_rd = (struct nd_redirect *)(ip6 + 1); 2434 nd_rd->nd_rd_type = ND_REDIRECT; 2435 nd_rd->nd_rd_code = 0; 2436 nd_rd->nd_rd_reserved = 0; 2437 if (rt->rt_flags & RTF_GATEWAY) { 2438 /* 2439 * nd_rd->nd_rd_target must be a link-local address in 2440 * better router cases. 2441 */ 2442 if (!nexthop) 2443 goto fail; 2444 bcopy(nexthop, &nd_rd->nd_rd_target, 2445 sizeof(nd_rd->nd_rd_target)); 2446 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2447 sizeof(nd_rd->nd_rd_dst)); 2448 } else { 2449 /* make sure redtgt == reddst */ 2450 nexthop = &sip6->ip6_dst; 2451 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2452 sizeof(nd_rd->nd_rd_target)); 2453 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2454 sizeof(nd_rd->nd_rd_dst)); 2455 } 2456 2457 p = (u_char *)(nd_rd + 1); 2458 2459 { 2460 /* target lladdr option */ 2461 struct rtentry *rt_nexthop = NULL; 2462 int len; 2463 const struct sockaddr_dl *sdl; 2464 struct nd_opt_hdr *nd_opt; 2465 char *lladdr; 2466 2467 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2468 if (!rt_nexthop) 2469 goto nolladdropt; 2470 len = sizeof(*nd_opt) + ifp->if_addrlen; 2471 len = (len + 7) & ~7; /* round by 8 */ 2472 /* safety check */ 2473 if (len + (p - (u_char *)ip6) > maxlen) 2474 goto nolladdropt; 2475 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2476 (rt_nexthop->rt_flags & RTF_LLINFO) && 2477 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2478 (sdl = satocsdl(rt_nexthop->rt_gateway)) && 2479 sdl->sdl_alen) { 2480 nd_opt = (struct nd_opt_hdr *)p; 2481 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2482 nd_opt->nd_opt_len = len >> 3; 2483 lladdr = (char *)(nd_opt + 1); 2484 memcpy(lladdr, CLLADDR(sdl), ifp->if_addrlen); 2485 p += len; 2486 } 2487 } 2488 nolladdropt:; 2489 2490 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2491 2492 /* just to be safe */ 2493 if (m0->m_flags & M_DECRYPTED) 2494 goto noredhdropt; 2495 if (p - (u_char *)ip6 > maxlen) 2496 goto noredhdropt; 2497 2498 { 2499 /* redirected header option */ 2500 int len; 2501 struct nd_opt_rd_hdr *nd_opt_rh; 2502 2503 /* 2504 * compute the maximum size for icmp6 redirect header option. 2505 * XXX room for auth header? 2506 */ 2507 len = maxlen - (p - (u_char *)ip6); 2508 len &= ~7; 2509 2510 /* 2511 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2512 * about padding/truncate rule for the original IP packet. 2513 * From the discussion on IPv6imp in Feb 1999, 2514 * the consensus was: 2515 * - "attach as much as possible" is the goal 2516 * - pad if not aligned (original size can be guessed by 2517 * original ip6 header) 2518 * Following code adds the padding if it is simple enough, 2519 * and truncates if not. 2520 */ 2521 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2522 /* not enough room, truncate */ 2523 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2524 m0->m_pkthdr.len); 2525 } else { 2526 /* 2527 * enough room, truncate if not aligned. 2528 * we don't pad here for simplicity. 2529 */ 2530 size_t extra; 2531 2532 extra = m0->m_pkthdr.len % 8; 2533 if (extra) { 2534 /* truncate */ 2535 m_adj(m0, -extra); 2536 } 2537 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2538 } 2539 2540 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2541 memset(nd_opt_rh, 0, sizeof(*nd_opt_rh)); 2542 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2543 nd_opt_rh->nd_opt_rh_len = len >> 3; 2544 p += sizeof(*nd_opt_rh); 2545 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2546 2547 /* connect m0 to m */ 2548 m->m_pkthdr.len += m0->m_pkthdr.len; 2549 m_cat(m, m0); 2550 m0 = NULL; 2551 } 2552 noredhdropt: 2553 if (m0) { 2554 m_freem(m0); 2555 m0 = NULL; 2556 } 2557 2558 /* XXX: clear embedded link IDs in the inner header */ 2559 in6_clearscope(&sip6->ip6_src); 2560 in6_clearscope(&sip6->ip6_dst); 2561 in6_clearscope(&nd_rd->nd_rd_target); 2562 in6_clearscope(&nd_rd->nd_rd_dst); 2563 2564 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2565 2566 nd_rd->nd_rd_cksum = 0; 2567 nd_rd->nd_rd_cksum 2568 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2569 2570 /* send the packet to outside... */ 2571 if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0) 2572 icmp6_ifstat_inc(ifp, ifs6_out_error); 2573 2574 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2575 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2576 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT); 2577 2578 return; 2579 2580 fail: 2581 if (m) 2582 m_freem(m); 2583 if (m0) 2584 m_freem(m0); 2585 } 2586 2587 /* 2588 * ICMPv6 socket option processing. 2589 */ 2590 int 2591 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 2592 { 2593 int error = 0; 2594 struct in6pcb *in6p = sotoin6pcb(so); 2595 2596 if (sopt->sopt_level != IPPROTO_ICMPV6) 2597 return rip6_ctloutput(op, so, sopt); 2598 2599 switch (op) { 2600 case PRCO_SETOPT: 2601 switch (sopt->sopt_name) { 2602 case ICMP6_FILTER: 2603 { 2604 struct icmp6_filter fil; 2605 2606 error = sockopt_get(sopt, &fil, sizeof(fil)); 2607 if (error) 2608 break; 2609 memcpy(in6p->in6p_icmp6filt, &fil, 2610 sizeof(struct icmp6_filter)); 2611 error = 0; 2612 break; 2613 } 2614 2615 default: 2616 error = ENOPROTOOPT; 2617 break; 2618 } 2619 break; 2620 2621 case PRCO_GETOPT: 2622 switch (sopt->sopt_name) { 2623 case ICMP6_FILTER: 2624 { 2625 if (in6p->in6p_icmp6filt == NULL) { 2626 error = EINVAL; 2627 break; 2628 } 2629 error = sockopt_set(sopt, in6p->in6p_icmp6filt, 2630 sizeof(struct icmp6_filter)); 2631 break; 2632 } 2633 2634 default: 2635 error = ENOPROTOOPT; 2636 break; 2637 } 2638 break; 2639 } 2640 2641 return (error); 2642 } 2643 2644 /* 2645 * Perform rate limit check. 2646 * Returns 0 if it is okay to send the icmp6 packet. 2647 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2648 * limitation. 2649 * 2650 * XXX per-destination/type check necessary? 2651 */ 2652 static int 2653 icmp6_ratelimit( 2654 const struct in6_addr *dst, /* not used at this moment */ 2655 const int type, /* not used at this moment */ 2656 const int code) /* not used at this moment */ 2657 { 2658 int ret; 2659 2660 ret = 0; /* okay to send */ 2661 2662 /* PPS limit */ 2663 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2664 icmp6errppslim)) { 2665 /* The packet is subject to rate limit */ 2666 ret++; 2667 } 2668 2669 return ret; 2670 } 2671 2672 static struct rtentry * 2673 icmp6_mtudisc_clone(struct sockaddr *dst) 2674 { 2675 struct rtentry *rt; 2676 int error; 2677 2678 rt = rtalloc1(dst, 1); 2679 if (rt == 0) 2680 return NULL; 2681 2682 /* If we didn't get a host route, allocate one */ 2683 if ((rt->rt_flags & RTF_HOST) == 0) { 2684 struct rtentry *nrt; 2685 2686 error = rtrequest((int) RTM_ADD, dst, 2687 (struct sockaddr *) rt->rt_gateway, NULL, 2688 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2689 if (error) { 2690 rtfree(rt); 2691 return NULL; 2692 } 2693 nrt->rt_rmx = rt->rt_rmx; 2694 rtfree(rt); 2695 rt = nrt; 2696 } 2697 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2698 icmp6_mtudisc_timeout_q); 2699 if (error) { 2700 rtfree(rt); 2701 return NULL; 2702 } 2703 2704 return rt; /* caller need to call rtfree() */ 2705 } 2706 2707 static void 2708 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 2709 { 2710 if (rt == NULL) 2711 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2712 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2713 (RTF_DYNAMIC | RTF_HOST)) { 2714 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2715 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2716 } else { 2717 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2718 rt->rt_rmx.rmx_mtu = 0; 2719 } 2720 } 2721 2722 static void 2723 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r) 2724 { 2725 if (rt == NULL) 2726 panic("icmp6_redirect_timeout: bad route to timeout"); 2727 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2728 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2729 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2730 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2731 } 2732 } 2733 2734 /* 2735 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes. silly? 2736 */ 2737 static int 2738 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS) 2739 { 2740 (void)&name; 2741 (void)&l; 2742 (void)&oname; 2743 2744 if (namelen != 0) 2745 return (EINVAL); 2746 2747 return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp, 2748 /*XXXUNCONST*/ 2749 __UNCONST(newp), newlen)); 2750 } 2751 2752 static int 2753 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS) 2754 { 2755 2756 return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS)); 2757 } 2758 2759 static void 2760 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog) 2761 { 2762 extern int nd6_maxqueuelen; /* defined in nd6.c */ 2763 2764 sysctl_createv(clog, 0, NULL, NULL, 2765 CTLFLAG_PERMANENT, 2766 CTLTYPE_NODE, "inet6", NULL, 2767 NULL, 0, NULL, 0, 2768 CTL_NET, PF_INET6, CTL_EOL); 2769 sysctl_createv(clog, 0, NULL, NULL, 2770 CTLFLAG_PERMANENT, 2771 CTLTYPE_NODE, "icmp6", 2772 SYSCTL_DESCR("ICMPv6 related settings"), 2773 NULL, 0, NULL, 0, 2774 CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL); 2775 2776 sysctl_createv(clog, 0, NULL, NULL, 2777 CTLFLAG_PERMANENT, 2778 CTLTYPE_STRUCT, "stats", 2779 SYSCTL_DESCR("ICMPv6 transmission statistics"), 2780 sysctl_net_inet6_icmp6_stats, 0, NULL, 0, 2781 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2782 ICMPV6CTL_STATS, CTL_EOL); 2783 sysctl_createv(clog, 0, NULL, NULL, 2784 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2785 CTLTYPE_INT, "rediraccept", 2786 SYSCTL_DESCR("Accept and process redirect messages"), 2787 NULL, 0, &icmp6_rediraccept, 0, 2788 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2789 ICMPV6CTL_REDIRACCEPT, CTL_EOL); 2790 sysctl_createv(clog, 0, NULL, NULL, 2791 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2792 CTLTYPE_INT, "redirtimeout", 2793 SYSCTL_DESCR("Redirect generated route lifetime"), 2794 NULL, 0, &icmp6_redirtimeout, 0, 2795 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2796 ICMPV6CTL_REDIRTIMEOUT, CTL_EOL); 2797 #if 0 /* obsoleted */ 2798 sysctl_createv(clog, 0, NULL, NULL, 2799 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2800 CTLTYPE_INT, "errratelimit", NULL, 2801 NULL, 0, &icmp6_errratelimit, 0, 2802 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2803 ICMPV6CTL_ERRRATELIMIT, CTL_EOL); 2804 #endif 2805 sysctl_createv(clog, 0, NULL, NULL, 2806 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2807 CTLTYPE_INT, "nd6_prune", 2808 SYSCTL_DESCR("Neighbor discovery prune interval"), 2809 NULL, 0, &nd6_prune, 0, 2810 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2811 ICMPV6CTL_ND6_PRUNE, CTL_EOL); 2812 sysctl_createv(clog, 0, NULL, NULL, 2813 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2814 CTLTYPE_INT, "nd6_delay", 2815 SYSCTL_DESCR("First probe delay time"), 2816 NULL, 0, &nd6_delay, 0, 2817 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2818 ICMPV6CTL_ND6_DELAY, CTL_EOL); 2819 sysctl_createv(clog, 0, NULL, NULL, 2820 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2821 CTLTYPE_INT, "nd6_umaxtries", 2822 SYSCTL_DESCR("Number of unicast discovery attempts"), 2823 NULL, 0, &nd6_umaxtries, 0, 2824 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2825 ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL); 2826 sysctl_createv(clog, 0, NULL, NULL, 2827 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2828 CTLTYPE_INT, "nd6_mmaxtries", 2829 SYSCTL_DESCR("Number of multicast discovery attempts"), 2830 NULL, 0, &nd6_mmaxtries, 0, 2831 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2832 ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL); 2833 sysctl_createv(clog, 0, NULL, NULL, 2834 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2835 CTLTYPE_INT, "nd6_useloopback", 2836 SYSCTL_DESCR("Use loopback interface for local traffic"), 2837 NULL, 0, &nd6_useloopback, 0, 2838 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2839 ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL); 2840 #if 0 /* obsoleted */ 2841 sysctl_createv(clog, 0, NULL, NULL, 2842 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2843 CTLTYPE_INT, "nd6_proxyall", NULL, 2844 NULL, 0, &nd6_proxyall, 0, 2845 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2846 ICMPV6CTL_ND6_PROXYALL, CTL_EOL); 2847 #endif 2848 sysctl_createv(clog, 0, NULL, NULL, 2849 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2850 CTLTYPE_INT, "nodeinfo", 2851 SYSCTL_DESCR("Respond to node information requests"), 2852 NULL, 0, &icmp6_nodeinfo, 0, 2853 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2854 ICMPV6CTL_NODEINFO, CTL_EOL); 2855 sysctl_createv(clog, 0, NULL, NULL, 2856 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2857 CTLTYPE_INT, "errppslimit", 2858 SYSCTL_DESCR("Maximum ICMP errors sent per second"), 2859 NULL, 0, &icmp6errppslim, 0, 2860 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2861 ICMPV6CTL_ERRPPSLIMIT, CTL_EOL); 2862 sysctl_createv(clog, 0, NULL, NULL, 2863 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2864 CTLTYPE_INT, "nd6_maxnudhint", 2865 SYSCTL_DESCR("Maximum neighbor unreachable hint count"), 2866 NULL, 0, &nd6_maxnudhint, 0, 2867 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2868 ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL); 2869 sysctl_createv(clog, 0, NULL, NULL, 2870 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2871 CTLTYPE_INT, "mtudisc_hiwat", 2872 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2873 NULL, 0, &icmp6_mtudisc_hiwat, 0, 2874 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2875 ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL); 2876 sysctl_createv(clog, 0, NULL, NULL, 2877 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2878 CTLTYPE_INT, "mtudisc_lowat", 2879 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2880 NULL, 0, &icmp6_mtudisc_lowat, 0, 2881 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2882 ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL); 2883 sysctl_createv(clog, 0, NULL, NULL, 2884 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2885 CTLTYPE_INT, "nd6_debug", 2886 SYSCTL_DESCR("Enable neighbor discovery debug output"), 2887 NULL, 0, &nd6_debug, 0, 2888 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2889 ICMPV6CTL_ND6_DEBUG, CTL_EOL); 2890 sysctl_createv(clog, 0, NULL, NULL, 2891 CTLFLAG_PERMANENT, 2892 CTLTYPE_STRUCT, "nd6_drlist", 2893 SYSCTL_DESCR("Default router list"), 2894 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2895 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2896 ICMPV6CTL_ND6_DRLIST, CTL_EOL); 2897 sysctl_createv(clog, 0, NULL, NULL, 2898 CTLFLAG_PERMANENT, 2899 CTLTYPE_STRUCT, "nd6_prlist", 2900 SYSCTL_DESCR("Prefix list"), 2901 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2902 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2903 ICMPV6CTL_ND6_PRLIST, CTL_EOL); 2904 sysctl_createv(clog, 0, NULL, NULL, 2905 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2906 CTLTYPE_INT, "maxqueuelen", 2907 SYSCTL_DESCR("max packet queue len for a unresolved ND"), 2908 NULL, 1, &nd6_maxqueuelen, 0, 2909 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2910 ICMPV6CTL_ND6_MAXQLEN, CTL_EOL); 2911 } 2912 2913 void 2914 icmp6_statinc(u_int stat) 2915 { 2916 2917 KASSERT(stat < ICMP6_NSTATS); 2918 ICMP6_STATINC(stat); 2919 } 2920