1 /* $NetBSD: icmp6.c,v 1.170 2014/11/25 19:51:17 christos 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 1.170 2014/11/25 19:51:17 christos 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 < 0) 726 break; 727 if (maxhlen > hostnamelen) 728 maxhlen = hostnamelen; 729 /* 730 * Copy IPv6 and ICMPv6 only. 731 */ 732 nip6 = mtod(n, struct ip6_hdr *); 733 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 734 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 735 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 736 p = (u_char *)(nicmp6 + 1); 737 memset(p, 0, 4); 738 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 739 noff = sizeof(struct ip6_hdr); 740 M_COPY_PKTHDR(n, m); /* just for rcvif */ 741 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 742 sizeof(struct icmp6_hdr) + 4 + maxhlen; 743 nicmp6->icmp6_type = ICMP6_WRUREPLY; 744 nicmp6->icmp6_code = 0; 745 } 746 #undef hostnamelen 747 if (n) { 748 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 749 icmp6s[ICMP6_STAT_REFLECT]++; 750 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++; 751 ICMP6_STAT_PUTREF(); 752 icmp6_reflect(n, noff); 753 } 754 break; 755 } 756 757 case ICMP6_WRUREPLY: 758 if (code != 0) 759 goto badcode; 760 break; 761 762 case ND_ROUTER_SOLICIT: 763 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 764 if (code != 0) 765 goto badcode; 766 if (icmp6len < sizeof(struct nd_router_solicit)) 767 goto badlen; 768 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 769 /* give up local */ 770 nd6_rs_input(m, off, icmp6len); 771 m = NULL; 772 goto freeit; 773 } 774 nd6_rs_input(n, off, icmp6len); 775 /* m stays. */ 776 break; 777 778 case ND_ROUTER_ADVERT: 779 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 780 if (code != 0) 781 goto badcode; 782 if (icmp6len < sizeof(struct nd_router_advert)) 783 goto badlen; 784 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 785 /* give up local */ 786 nd6_ra_input(m, off, icmp6len); 787 m = NULL; 788 goto freeit; 789 } 790 nd6_ra_input(n, off, icmp6len); 791 /* m stays. */ 792 break; 793 794 case ND_NEIGHBOR_SOLICIT: 795 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 796 if (code != 0) 797 goto badcode; 798 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 799 goto badlen; 800 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 801 /* give up local */ 802 nd6_ns_input(m, off, icmp6len); 803 m = NULL; 804 goto freeit; 805 } 806 nd6_ns_input(n, off, icmp6len); 807 /* m stays. */ 808 break; 809 810 case ND_NEIGHBOR_ADVERT: 811 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 812 if (code != 0) 813 goto badcode; 814 if (icmp6len < sizeof(struct nd_neighbor_advert)) 815 goto badlen; 816 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 817 /* give up local */ 818 nd6_na_input(m, off, icmp6len); 819 m = NULL; 820 goto freeit; 821 } 822 nd6_na_input(n, off, icmp6len); 823 /* m stays. */ 824 break; 825 826 case ND_REDIRECT: 827 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 828 if (code != 0) 829 goto badcode; 830 if (icmp6len < sizeof(struct nd_redirect)) 831 goto badlen; 832 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 833 /* give up local */ 834 icmp6_redirect_input(m, off); 835 m = NULL; 836 goto freeit; 837 } 838 icmp6_redirect_input(n, off); 839 /* m stays. */ 840 break; 841 842 case ICMP6_ROUTER_RENUMBERING: 843 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 844 code != ICMP6_ROUTER_RENUMBERING_RESULT) 845 goto badcode; 846 if (icmp6len < sizeof(struct icmp6_router_renum)) 847 goto badlen; 848 break; 849 850 default: 851 nd6log((LOG_DEBUG, 852 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 853 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 854 ip6_sprintf(&ip6->ip6_dst), 855 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 856 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 857 /* ICMPv6 error: MUST deliver it by spec... */ 858 code = PRC_NCMDS; 859 /* deliver */ 860 } else { 861 /* ICMPv6 informational: MUST not deliver */ 862 break; 863 } 864 deliver: 865 if (icmp6_notify_error(m, off, icmp6len, code)) { 866 /* In this case, m should've been freed. */ 867 return (IPPROTO_DONE); 868 } 869 break; 870 871 badcode: 872 ICMP6_STATINC(ICMP6_STAT_BADCODE); 873 break; 874 875 badlen: 876 ICMP6_STATINC(ICMP6_STAT_BADLEN); 877 break; 878 } 879 880 /* deliver the packet to appropriate sockets */ 881 icmp6_rip6_input(&m, *offp); 882 883 return IPPROTO_DONE; 884 885 freeit: 886 m_freem(m); 887 return IPPROTO_DONE; 888 } 889 890 static int 891 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code) 892 { 893 struct icmp6_hdr *icmp6; 894 struct ip6_hdr *eip6; 895 u_int32_t notifymtu; 896 struct sockaddr_in6 icmp6src, icmp6dst; 897 898 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 899 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 900 goto freeit; 901 } 902 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 903 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 904 if (icmp6 == NULL) { 905 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 906 return (-1); 907 } 908 eip6 = (struct ip6_hdr *)(icmp6 + 1); 909 910 /* Detect the upper level protocol */ 911 { 912 void (*ctlfunc)(int, struct sockaddr *, void *); 913 u_int8_t nxt = eip6->ip6_nxt; 914 int eoff = off + sizeof(struct icmp6_hdr) + 915 sizeof(struct ip6_hdr); 916 struct ip6ctlparam ip6cp; 917 struct in6_addr *finaldst = NULL; 918 int icmp6type = icmp6->icmp6_type; 919 struct ip6_frag *fh; 920 struct ip6_rthdr *rth; 921 struct ip6_rthdr0 *rth0; 922 int rthlen; 923 924 while (1) { /* XXX: should avoid infinite loop explicitly? */ 925 struct ip6_ext *eh; 926 927 switch (nxt) { 928 case IPPROTO_HOPOPTS: 929 case IPPROTO_DSTOPTS: 930 case IPPROTO_AH: 931 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 932 eoff, sizeof(*eh)); 933 if (eh == NULL) { 934 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 935 return (-1); 936 } 937 938 if (nxt == IPPROTO_AH) 939 eoff += (eh->ip6e_len + 2) << 2; 940 else 941 eoff += (eh->ip6e_len + 1) << 3; 942 nxt = eh->ip6e_nxt; 943 break; 944 case IPPROTO_ROUTING: 945 /* 946 * When the erroneous packet contains a 947 * routing header, we should examine the 948 * header to determine the final destination. 949 * Otherwise, we can't properly update 950 * information that depends on the final 951 * destination (e.g. path MTU). 952 */ 953 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 954 eoff, sizeof(*rth)); 955 if (rth == NULL) { 956 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 957 return (-1); 958 } 959 rthlen = (rth->ip6r_len + 1) << 3; 960 /* 961 * XXX: currently there is no 962 * officially defined type other 963 * than type-0. 964 * Note that if the segment left field 965 * is 0, all intermediate hops must 966 * have been passed. 967 */ 968 if (rth->ip6r_segleft && 969 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 970 int hops; 971 972 IP6_EXTHDR_GET(rth0, 973 struct ip6_rthdr0 *, m, 974 eoff, rthlen); 975 if (rth0 == NULL) { 976 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 977 return (-1); 978 } 979 /* just ignore a bogus header */ 980 if ((rth0->ip6r0_len % 2) == 0 && 981 (hops = rth0->ip6r0_len/2)) 982 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 983 } 984 eoff += rthlen; 985 nxt = rth->ip6r_nxt; 986 break; 987 case IPPROTO_FRAGMENT: 988 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 989 eoff, sizeof(*fh)); 990 if (fh == NULL) { 991 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 992 return (-1); 993 } 994 /* 995 * Data after a fragment header is meaningless 996 * unless it is the first fragment, but 997 * we'll go to the notify label for path MTU 998 * discovery. 999 */ 1000 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1001 goto notify; 1002 1003 eoff += sizeof(struct ip6_frag); 1004 nxt = fh->ip6f_nxt; 1005 break; 1006 default: 1007 /* 1008 * This case includes ESP and the No Next 1009 * Header. In such cases going to the notify 1010 * label does not have any meaning 1011 * (i.e. ctlfunc will be NULL), but we go 1012 * anyway since we might have to update 1013 * path MTU information. 1014 */ 1015 goto notify; 1016 } 1017 } 1018 notify: 1019 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1020 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1021 if (icmp6 == NULL) { 1022 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 1023 return (-1); 1024 } 1025 1026 /* 1027 * retrieve parameters from the inner IPv6 header, and convert 1028 * them into sockaddr structures. 1029 * XXX: there is no guarantee that the source or destination 1030 * addresses of the inner packet are in the same scope zone as 1031 * the addresses of the icmp packet. But there is no other 1032 * way to determine the zone. 1033 */ 1034 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1035 1036 sockaddr_in6_init(&icmp6dst, 1037 (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0); 1038 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1039 goto freeit; 1040 sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0); 1041 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1042 goto freeit; 1043 icmp6src.sin6_flowinfo = 1044 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1045 1046 if (finaldst == NULL) 1047 finaldst = &eip6->ip6_dst; 1048 ip6cp.ip6c_m = m; 1049 ip6cp.ip6c_icmp6 = icmp6; 1050 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1051 ip6cp.ip6c_off = eoff; 1052 ip6cp.ip6c_finaldst = finaldst; 1053 ip6cp.ip6c_src = &icmp6src; 1054 ip6cp.ip6c_nxt = nxt; 1055 1056 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1057 notifymtu = ntohl(icmp6->icmp6_mtu); 1058 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1059 } 1060 1061 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1062 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1063 if (ctlfunc) { 1064 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1065 &ip6cp); 1066 } 1067 } 1068 return (0); 1069 1070 freeit: 1071 m_freem(m); 1072 return (-1); 1073 } 1074 1075 void 1076 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1077 { 1078 unsigned long rtcount; 1079 struct icmp6_mtudisc_callback *mc; 1080 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1081 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1082 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1083 u_int mtu = ntohl(icmp6->icmp6_mtu); 1084 struct rtentry *rt = NULL; 1085 struct sockaddr_in6 sin6; 1086 1087 /* 1088 * The MTU should not be less than the minimal IPv6 MTU except for the 1089 * hack in ip6_output/ip6_setpmtu where we always include a frag header. 1090 * In that one case, the MTU might be less than 1280. 1091 */ 1092 if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) { 1093 /* is the mtu even sane? */ 1094 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) 1095 return; 1096 if (!validated) 1097 return; 1098 mtu = IPV6_MMTU - sizeof(struct ip6_frag); 1099 } 1100 1101 /* 1102 * allow non-validated cases if memory is plenty, to make traffic 1103 * from non-connected pcb happy. 1104 */ 1105 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1106 if (validated) { 1107 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1108 return; 1109 else if (0 <= icmp6_mtudisc_lowat && 1110 rtcount > icmp6_mtudisc_lowat) { 1111 /* 1112 * XXX nuke a victim, install the new one. 1113 */ 1114 } 1115 } else { 1116 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1117 return; 1118 } 1119 1120 memset(&sin6, 0, sizeof(sin6)); 1121 sin6.sin6_family = PF_INET6; 1122 sin6.sin6_len = sizeof(struct sockaddr_in6); 1123 sin6.sin6_addr = *dst; 1124 if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1125 return; 1126 1127 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1128 1129 if (rt && (rt->rt_flags & RTF_HOST) && 1130 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1131 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1132 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1133 ICMP6_STATINC(ICMP6_STAT_PMTUCHG); 1134 rt->rt_rmx.rmx_mtu = mtu; 1135 } 1136 } 1137 if (rt) { 1138 rtfree(rt); 1139 } 1140 1141 /* 1142 * Notify protocols that the MTU for this destination 1143 * has changed. 1144 */ 1145 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1146 mc = LIST_NEXT(mc, mc_list)) 1147 (*mc->mc_func)(&sin6.sin6_addr); 1148 } 1149 1150 /* 1151 * Process a Node Information Query packet, based on 1152 * draft-ietf-ipngwg-icmp-name-lookups-07. 1153 * 1154 * Spec incompatibilities: 1155 * - IPv6 Subject address handling 1156 * - IPv4 Subject address handling support missing 1157 * - Proxy reply (answer even if it's not for me) 1158 * - joins NI group address at in6_ifattach() time only, does not cope 1159 * with hostname changes by sethostname(3) 1160 */ 1161 static struct mbuf * 1162 ni6_input(struct mbuf *m, int off) 1163 { 1164 struct icmp6_nodeinfo *ni6, *nni6; 1165 struct mbuf *n = NULL; 1166 u_int16_t qtype; 1167 int subjlen; 1168 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1169 struct ni_reply_fqdn *fqdn; 1170 int addrs; /* for NI_QTYPE_NODEADDR */ 1171 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1172 struct sockaddr_in6 sin6; /* ip6_dst */ 1173 struct in6_addr in6_subj; /* subject address */ 1174 struct ip6_hdr *ip6; 1175 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1176 char *subj = NULL; 1177 1178 ip6 = mtod(m, struct ip6_hdr *); 1179 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1180 if (ni6 == NULL) { 1181 /* m is already reclaimed */ 1182 return NULL; 1183 } 1184 1185 /* 1186 * Validate IPv6 destination address. 1187 * 1188 * The Responder must discard the Query without further processing 1189 * unless it is one of the Responder's unicast or anycast addresses, or 1190 * a link-local scope multicast address which the Responder has joined. 1191 * [icmp-name-lookups-07, Section 4.] 1192 */ 1193 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 1194 /* XXX scopeid */ 1195 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1196 ; /* unicast/anycast, fine */ 1197 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1198 ; /* link-local multicast, fine */ 1199 else 1200 goto bad; 1201 1202 /* validate query Subject field. */ 1203 qtype = ntohs(ni6->ni_qtype); 1204 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1205 switch (qtype) { 1206 case NI_QTYPE_NOOP: 1207 case NI_QTYPE_SUPTYPES: 1208 /* 07 draft */ 1209 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1210 break; 1211 /* FALLTHROUGH */ 1212 case NI_QTYPE_FQDN: 1213 case NI_QTYPE_NODEADDR: 1214 case NI_QTYPE_IPV4ADDR: 1215 switch (ni6->ni_code) { 1216 case ICMP6_NI_SUBJ_IPV6: 1217 #if ICMP6_NI_SUBJ_IPV6 != 0 1218 case 0: 1219 #endif 1220 /* 1221 * backward compatibility - try to accept 03 draft 1222 * format, where no Subject is present. 1223 */ 1224 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1225 subjlen == 0) { 1226 oldfqdn++; 1227 break; 1228 } 1229 #if ICMP6_NI_SUBJ_IPV6 != 0 1230 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1231 goto bad; 1232 #endif 1233 1234 if (subjlen != sizeof(sin6.sin6_addr)) 1235 goto bad; 1236 1237 /* 1238 * Validate Subject address. 1239 * 1240 * Not sure what exactly "address belongs to the node" 1241 * means in the spec, is it just unicast, or what? 1242 * 1243 * At this moment we consider Subject address as 1244 * "belong to the node" if the Subject address equals 1245 * to the IPv6 destination address; validation for 1246 * IPv6 destination address should have done enough 1247 * check for us. 1248 * 1249 * We do not do proxy at this moment. 1250 */ 1251 /* m_pulldown instead of copy? */ 1252 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1253 subjlen, (void *)&in6_subj); 1254 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1255 goto bad; 1256 1257 subj = (char *)&in6_subj; 1258 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1259 break; 1260 1261 /* 1262 * XXX if we are to allow other cases, we should really 1263 * be careful about scope here. 1264 * basically, we should disallow queries toward IPv6 1265 * destination X with subject Y, if scope(X) > scope(Y). 1266 * if we allow scope(X) > scope(Y), it will result in 1267 * information leakage across scope boundary. 1268 */ 1269 goto bad; 1270 1271 case ICMP6_NI_SUBJ_FQDN: 1272 /* 1273 * Validate Subject name with gethostname(3). 1274 * 1275 * The behavior may need some debate, since: 1276 * - we are not sure if the node has FQDN as 1277 * hostname (returned by gethostname(3)). 1278 * - the code does wildcard match for truncated names. 1279 * however, we are not sure if we want to perform 1280 * wildcard match, if gethostname(3) side has 1281 * truncated hostname. 1282 */ 1283 n = ni6_nametodns(hostname, hostnamelen, 0); 1284 if (!n || n->m_next || n->m_len == 0) 1285 goto bad; 1286 IP6_EXTHDR_GET(subj, char *, m, 1287 off + sizeof(struct icmp6_nodeinfo), subjlen); 1288 if (subj == NULL) 1289 goto bad; 1290 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1291 n->m_len)) { 1292 goto bad; 1293 } 1294 m_freem(n); 1295 n = NULL; 1296 break; 1297 1298 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1299 default: 1300 goto bad; 1301 } 1302 break; 1303 } 1304 1305 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1306 switch (qtype) { 1307 case NI_QTYPE_FQDN: 1308 if ((icmp6_nodeinfo & 1) == 0) 1309 goto bad; 1310 break; 1311 case NI_QTYPE_NODEADDR: 1312 case NI_QTYPE_IPV4ADDR: 1313 if ((icmp6_nodeinfo & 2) == 0) 1314 goto bad; 1315 break; 1316 } 1317 1318 /* guess reply length */ 1319 switch (qtype) { 1320 case NI_QTYPE_NOOP: 1321 break; /* no reply data */ 1322 case NI_QTYPE_SUPTYPES: 1323 replylen += sizeof(u_int32_t); 1324 break; 1325 case NI_QTYPE_FQDN: 1326 /* XXX will append an mbuf */ 1327 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1328 break; 1329 case NI_QTYPE_NODEADDR: 1330 addrs = ni6_addrs(ni6, m, &ifp, subj); 1331 if ((replylen += addrs * (sizeof(struct in6_addr) + 1332 sizeof(u_int32_t))) > MCLBYTES) 1333 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1334 break; 1335 case NI_QTYPE_IPV4ADDR: 1336 /* unsupported - should respond with unknown Qtype? */ 1337 goto bad; 1338 default: 1339 /* 1340 * XXX: We must return a reply with the ICMP6 code 1341 * `unknown Qtype' in this case. However we regard the case 1342 * as an FQDN query for backward compatibility. 1343 * Older versions set a random value to this field, 1344 * so it rarely varies in the defined qtypes. 1345 * But the mechanism is not reliable... 1346 * maybe we should obsolete older versions. 1347 */ 1348 qtype = NI_QTYPE_FQDN; 1349 /* XXX will append an mbuf */ 1350 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1351 oldfqdn++; 1352 break; 1353 } 1354 1355 /* allocate an mbuf to reply. */ 1356 MGETHDR(n, M_DONTWAIT, m->m_type); 1357 if (n == NULL) { 1358 m_freem(m); 1359 return (NULL); 1360 } 1361 M_MOVE_PKTHDR(n, m); /* just for rcvif */ 1362 if (replylen > MHLEN) { 1363 if (replylen > MCLBYTES) { 1364 /* 1365 * XXX: should we try to allocate more? But MCLBYTES 1366 * is probably much larger than IPV6_MMTU... 1367 */ 1368 goto bad; 1369 } 1370 MCLGET(n, M_DONTWAIT); 1371 if ((n->m_flags & M_EXT) == 0) { 1372 goto bad; 1373 } 1374 } 1375 n->m_pkthdr.len = n->m_len = replylen; 1376 1377 /* copy mbuf header and IPv6 + Node Information base headers */ 1378 bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr)); 1379 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1380 bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo)); 1381 1382 /* qtype dependent procedure */ 1383 switch (qtype) { 1384 case NI_QTYPE_NOOP: 1385 nni6->ni_code = ICMP6_NI_SUCCESS; 1386 nni6->ni_flags = 0; 1387 break; 1388 case NI_QTYPE_SUPTYPES: 1389 { 1390 u_int32_t v; 1391 nni6->ni_code = ICMP6_NI_SUCCESS; 1392 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1393 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1394 v = (u_int32_t)htonl(0x0000000f); 1395 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1396 break; 1397 } 1398 case NI_QTYPE_FQDN: 1399 nni6->ni_code = ICMP6_NI_SUCCESS; 1400 fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) + 1401 sizeof(struct ip6_hdr) + 1402 sizeof(struct icmp6_nodeinfo)); 1403 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1404 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1405 /* 1406 * XXX do we really have FQDN in variable "hostname"? 1407 */ 1408 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1409 if (n->m_next == NULL) 1410 goto bad; 1411 /* XXX we assume that n->m_next is not a chain */ 1412 if (n->m_next->m_next != NULL) 1413 goto bad; 1414 n->m_pkthdr.len += n->m_next->m_len; 1415 break; 1416 case NI_QTYPE_NODEADDR: 1417 { 1418 int lenlim, copied; 1419 1420 nni6->ni_code = ICMP6_NI_SUCCESS; 1421 n->m_pkthdr.len = n->m_len = 1422 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1423 lenlim = M_TRAILINGSPACE(n); 1424 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1425 /* XXX: reset mbuf length */ 1426 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1427 sizeof(struct icmp6_nodeinfo) + copied; 1428 break; 1429 } 1430 default: 1431 break; /* XXX impossible! */ 1432 } 1433 1434 nni6->ni_type = ICMP6_NI_REPLY; 1435 m_freem(m); 1436 return (n); 1437 1438 bad: 1439 m_freem(m); 1440 if (n) 1441 m_freem(n); 1442 return (NULL); 1443 } 1444 #undef hostnamelen 1445 1446 #define isupper(x) ('A' <= (x) && (x) <= 'Z') 1447 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1448 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1449 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1450 1451 /* 1452 * make a mbuf with DNS-encoded string. no compression support. 1453 * 1454 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1455 * treated as truncated name (two \0 at the end). this is a wild guess. 1456 * 1457 * old - return pascal string if non-zero 1458 */ 1459 static struct mbuf * 1460 ni6_nametodns(const char *name, int namelen, int old) 1461 { 1462 struct mbuf *m; 1463 char *cp, *ep; 1464 const char *p, *q; 1465 int i, len, nterm; 1466 1467 if (old) 1468 len = namelen + 1; 1469 else 1470 len = MCLBYTES; 1471 1472 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1473 MGET(m, M_DONTWAIT, MT_DATA); 1474 if (m && len > MLEN) { 1475 MCLGET(m, M_DONTWAIT); 1476 if ((m->m_flags & M_EXT) == 0) 1477 goto fail; 1478 } 1479 if (!m) 1480 goto fail; 1481 m->m_next = NULL; 1482 1483 if (old) { 1484 m->m_len = len; 1485 *mtod(m, char *) = namelen; 1486 bcopy(name, mtod(m, char *) + 1, namelen); 1487 return m; 1488 } else { 1489 m->m_len = 0; 1490 cp = mtod(m, char *); 1491 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1492 1493 /* if not certain about my name, return empty buffer */ 1494 if (namelen == 0) 1495 return m; 1496 1497 /* 1498 * guess if it looks like shortened hostname, or FQDN. 1499 * shortened hostname needs two trailing "\0". 1500 */ 1501 i = 0; 1502 for (p = name; p < name + namelen; p++) { 1503 if (*p && *p == '.') 1504 i++; 1505 } 1506 if (i < 2) 1507 nterm = 2; 1508 else 1509 nterm = 1; 1510 1511 p = name; 1512 while (cp < ep && p < name + namelen) { 1513 i = 0; 1514 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1515 i++; 1516 /* result does not fit into mbuf */ 1517 if (cp + i + 1 >= ep) 1518 goto fail; 1519 /* 1520 * DNS label length restriction, RFC1035 page 8. 1521 * "i == 0" case is included here to avoid returning 1522 * 0-length label on "foo..bar". 1523 */ 1524 if (i <= 0 || i >= 64) 1525 goto fail; 1526 *cp++ = i; 1527 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1528 goto fail; 1529 while (i > 0) { 1530 if (!isalnum(*p) && *p != '-') 1531 goto fail; 1532 if (isupper(*p)) { 1533 *cp++ = tolower(*p); 1534 p++; 1535 } else 1536 *cp++ = *p++; 1537 i--; 1538 } 1539 p = q; 1540 if (p < name + namelen && *p == '.') 1541 p++; 1542 } 1543 /* termination */ 1544 if (cp + nterm >= ep) 1545 goto fail; 1546 while (nterm-- > 0) 1547 *cp++ = '\0'; 1548 m->m_len = cp - mtod(m, char *); 1549 return m; 1550 } 1551 1552 panic("should not reach here"); 1553 /* NOTREACHED */ 1554 1555 fail: 1556 if (m) 1557 m_freem(m); 1558 return NULL; 1559 } 1560 1561 /* 1562 * check if two DNS-encoded string matches. takes care of truncated 1563 * form (with \0\0 at the end). no compression support. 1564 * XXX upper/lowercase match (see RFC2065) 1565 */ 1566 static int 1567 ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1568 { 1569 const char *a0, *b0; 1570 int l; 1571 1572 /* simplest case - need validation? */ 1573 if (alen == blen && memcmp(a, b, alen) == 0) 1574 return 1; 1575 1576 a0 = a; 1577 b0 = b; 1578 1579 /* termination is mandatory */ 1580 if (alen < 2 || blen < 2) 1581 return 0; 1582 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1583 return 0; 1584 alen--; 1585 blen--; 1586 1587 while (a - a0 < alen && b - b0 < blen) { 1588 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1589 return 0; 1590 1591 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1592 return 0; 1593 /* we don't support compression yet */ 1594 if (a[0] >= 64 || b[0] >= 64) 1595 return 0; 1596 1597 /* truncated case */ 1598 if (a[0] == 0 && a - a0 == alen - 1) 1599 return 1; 1600 if (b[0] == 0 && b - b0 == blen - 1) 1601 return 1; 1602 if (a[0] == 0 || b[0] == 0) 1603 return 0; 1604 1605 if (a[0] != b[0]) 1606 return 0; 1607 l = a[0]; 1608 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1609 return 0; 1610 if (memcmp(a + 1, b + 1, l) != 0) 1611 return 0; 1612 1613 a += 1 + l; 1614 b += 1 + l; 1615 } 1616 1617 if (a - a0 == alen && b - b0 == blen) 1618 return 1; 1619 else 1620 return 0; 1621 } 1622 1623 /* 1624 * calculate the number of addresses to be returned in the node info reply. 1625 */ 1626 static int 1627 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, 1628 struct ifnet **ifpp, char *subj) 1629 { 1630 struct ifnet *ifp; 1631 struct in6_ifaddr *ifa6; 1632 struct ifaddr *ifa; 1633 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1634 int addrs = 0, addrsofif, iffound = 0; 1635 int niflags = ni6->ni_flags; 1636 1637 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1638 switch (ni6->ni_code) { 1639 case ICMP6_NI_SUBJ_IPV6: 1640 if (subj == NULL) /* must be impossible... */ 1641 return (0); 1642 subj_ip6 = (struct sockaddr_in6 *)subj; 1643 break; 1644 default: 1645 /* 1646 * XXX: we only support IPv6 subject address for 1647 * this Qtype. 1648 */ 1649 return (0); 1650 } 1651 } 1652 1653 IFNET_FOREACH(ifp) { 1654 addrsofif = 0; 1655 IFADDR_FOREACH(ifa, ifp) { 1656 if (ifa->ifa_addr->sa_family != AF_INET6) 1657 continue; 1658 ifa6 = (struct in6_ifaddr *)ifa; 1659 1660 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1661 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1662 &ifa6->ia_addr.sin6_addr)) 1663 iffound = 1; 1664 1665 /* 1666 * IPv4-mapped addresses can only be returned by a 1667 * Node Information proxy, since they represent 1668 * addresses of IPv4-only nodes, which perforce do 1669 * not implement this protocol. 1670 * [icmp-name-lookups-07, Section 5.4] 1671 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1672 * this function at this moment. 1673 */ 1674 1675 /* What do we have to do about ::1? */ 1676 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1677 case IPV6_ADDR_SCOPE_LINKLOCAL: 1678 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1679 continue; 1680 break; 1681 case IPV6_ADDR_SCOPE_SITELOCAL: 1682 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1683 continue; 1684 break; 1685 case IPV6_ADDR_SCOPE_GLOBAL: 1686 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1687 continue; 1688 break; 1689 default: 1690 continue; 1691 } 1692 1693 /* 1694 * check if anycast is okay. 1695 * XXX: just experimental. not in the spec. 1696 */ 1697 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1698 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1699 continue; /* we need only unicast addresses */ 1700 1701 addrsofif++; /* count the address */ 1702 } 1703 if (iffound) { 1704 *ifpp = ifp; 1705 return (addrsofif); 1706 } 1707 1708 addrs += addrsofif; 1709 } 1710 1711 return (addrs); 1712 } 1713 1714 static int 1715 ni6_store_addrs(struct icmp6_nodeinfo *ni6, 1716 struct icmp6_nodeinfo *nni6, struct ifnet *ifp0, 1717 int resid) 1718 { 1719 struct ifnet *ifp = ifp0 ? ifp0 : IFNET_FIRST(); 1720 struct in6_ifaddr *ifa6; 1721 struct ifaddr *ifa; 1722 struct ifnet *ifp_dep = NULL; 1723 int copied = 0, allow_deprecated = 0; 1724 u_char *cp = (u_char *)(nni6 + 1); 1725 int niflags = ni6->ni_flags; 1726 u_int32_t ltime; 1727 1728 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1729 return (0); /* needless to copy */ 1730 1731 again: 1732 1733 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1734 { 1735 IFADDR_FOREACH(ifa, ifp) { 1736 if (ifa->ifa_addr->sa_family != AF_INET6) 1737 continue; 1738 ifa6 = (struct in6_ifaddr *)ifa; 1739 1740 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1741 allow_deprecated == 0) { 1742 /* 1743 * prefererred address should be put before 1744 * deprecated addresses. 1745 */ 1746 1747 /* record the interface for later search */ 1748 if (ifp_dep == NULL) 1749 ifp_dep = ifp; 1750 1751 continue; 1752 } 1753 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1754 allow_deprecated != 0) 1755 continue; /* we now collect deprecated addrs */ 1756 1757 /* What do we have to do about ::1? */ 1758 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1759 case IPV6_ADDR_SCOPE_LINKLOCAL: 1760 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1761 continue; 1762 break; 1763 case IPV6_ADDR_SCOPE_SITELOCAL: 1764 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1765 continue; 1766 break; 1767 case IPV6_ADDR_SCOPE_GLOBAL: 1768 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1769 continue; 1770 break; 1771 default: 1772 continue; 1773 } 1774 1775 /* 1776 * check if anycast is okay. 1777 * XXX: just experimental. not in the spec. 1778 */ 1779 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1780 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1781 continue; 1782 1783 /* now we can copy the address */ 1784 if (resid < sizeof(struct in6_addr) + 1785 sizeof(u_int32_t)) { 1786 /* 1787 * We give up much more copy. 1788 * Set the truncate flag and return. 1789 */ 1790 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1791 return (copied); 1792 } 1793 1794 /* 1795 * Set the TTL of the address. 1796 * The TTL value should be one of the following 1797 * according to the specification: 1798 * 1799 * 1. The remaining lifetime of a DHCP lease on the 1800 * address, or 1801 * 2. The remaining Valid Lifetime of a prefix from 1802 * which the address was derived through Stateless 1803 * Autoconfiguration. 1804 * 1805 * Note that we currently do not support stateful 1806 * address configuration by DHCPv6, so the former 1807 * case can't happen. 1808 * 1809 * TTL must be 2^31 > TTL >= 0. 1810 */ 1811 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1812 ltime = ND6_INFINITE_LIFETIME; 1813 else { 1814 if (ifa6->ia6_lifetime.ia6t_expire > 1815 time_second) 1816 ltime = ifa6->ia6_lifetime.ia6t_expire - 1817 time_second; 1818 else 1819 ltime = 0; 1820 } 1821 if (ltime > 0x7fffffff) 1822 ltime = 0x7fffffff; 1823 ltime = htonl(ltime); 1824 1825 bcopy(<ime, cp, sizeof(u_int32_t)); 1826 cp += sizeof(u_int32_t); 1827 1828 /* copy the address itself */ 1829 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1830 sizeof(struct in6_addr)); 1831 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1832 cp += sizeof(struct in6_addr); 1833 1834 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1835 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1836 } 1837 if (ifp0) /* we need search only on the specified IF */ 1838 break; 1839 } 1840 1841 if (allow_deprecated == 0 && ifp_dep != NULL) { 1842 ifp = ifp_dep; 1843 allow_deprecated = 1; 1844 1845 goto again; 1846 } 1847 1848 return (copied); 1849 } 1850 1851 /* 1852 * XXX almost dup'ed code with rip6_input. 1853 */ 1854 static int 1855 icmp6_rip6_input(struct mbuf **mp, int off) 1856 { 1857 struct mbuf *m = *mp; 1858 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1859 struct inpcb_hdr *inph; 1860 struct in6pcb *in6p; 1861 struct in6pcb *last = NULL; 1862 struct sockaddr_in6 rip6src; 1863 struct icmp6_hdr *icmp6; 1864 struct mbuf *opts = NULL; 1865 1866 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1867 if (icmp6 == NULL) { 1868 /* m is already reclaimed */ 1869 return IPPROTO_DONE; 1870 } 1871 1872 /* 1873 * XXX: the address may have embedded scope zone ID, which should be 1874 * hidden from applications. 1875 */ 1876 sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0); 1877 if (sa6_recoverscope(&rip6src)) { 1878 m_freem(m); 1879 return (IPPROTO_DONE); 1880 } 1881 1882 TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) { 1883 in6p = (struct in6pcb *)inph; 1884 if (in6p->in6p_af != AF_INET6) 1885 continue; 1886 if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6) 1887 continue; 1888 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1889 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1890 continue; 1891 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1892 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1893 continue; 1894 if (in6p->in6p_icmp6filt 1895 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1896 in6p->in6p_icmp6filt)) 1897 continue; 1898 if (last) { 1899 struct mbuf *n; 1900 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1901 if (last->in6p_flags & IN6P_CONTROLOPTS) 1902 ip6_savecontrol(last, &opts, ip6, n); 1903 /* strip intermediate headers */ 1904 m_adj(n, off); 1905 if (sbappendaddr(&last->in6p_socket->so_rcv, 1906 (struct sockaddr *)&rip6src, 1907 n, opts) == 0) { 1908 /* should notify about lost packet */ 1909 m_freem(n); 1910 if (opts) 1911 m_freem(opts); 1912 } else 1913 sorwakeup(last->in6p_socket); 1914 opts = NULL; 1915 } 1916 } 1917 last = in6p; 1918 } 1919 if (last) { 1920 if (last->in6p_flags & IN6P_CONTROLOPTS) 1921 ip6_savecontrol(last, &opts, ip6, m); 1922 /* strip intermediate headers */ 1923 m_adj(m, off); 1924 if (sbappendaddr(&last->in6p_socket->so_rcv, 1925 (struct sockaddr *)&rip6src, m, opts) == 0) { 1926 m_freem(m); 1927 if (opts) 1928 m_freem(opts); 1929 } else 1930 sorwakeup(last->in6p_socket); 1931 } else { 1932 m_freem(m); 1933 IP6_STATDEC(IP6_STAT_DELIVERED); 1934 } 1935 return IPPROTO_DONE; 1936 } 1937 1938 /* 1939 * Reflect the ip6 packet back to the source. 1940 * OFF points to the icmp6 header, counted from the top of the mbuf. 1941 * 1942 * Note: RFC 1885 required that an echo reply should be truncated if it 1943 * did not fit in with (return) path MTU, and KAME code supported the 1944 * behavior. However, as a clarification after the RFC, this limitation 1945 * was removed in a revised version of the spec, RFC 2463. We had kept the 1946 * old behavior, with a (non-default) ifdef block, while the new version of 1947 * the spec was an internet-draft status, and even after the new RFC was 1948 * published. But it would rather make sense to clean the obsoleted part 1949 * up, and to make the code simpler at this stage. 1950 */ 1951 void 1952 icmp6_reflect(struct mbuf *m, size_t off) 1953 { 1954 struct ip6_hdr *ip6; 1955 struct icmp6_hdr *icmp6; 1956 const struct in6_ifaddr *ia; 1957 const struct ip6aux *ip6a; 1958 int plen; 1959 int type, code; 1960 struct ifnet *outif = NULL; 1961 struct in6_addr origdst; 1962 const struct in6_addr *src = NULL; 1963 1964 /* too short to reflect */ 1965 if (off < sizeof(struct ip6_hdr)) { 1966 nd6log((LOG_DEBUG, 1967 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 1968 (u_long)off, (u_long)sizeof(struct ip6_hdr), 1969 __FILE__, __LINE__)); 1970 goto bad; 1971 } 1972 1973 /* 1974 * If there are extra headers between IPv6 and ICMPv6, strip 1975 * off that header first. 1976 */ 1977 #ifdef DIAGNOSTIC 1978 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 1979 panic("assumption failed in icmp6_reflect"); 1980 #endif 1981 if (off > sizeof(struct ip6_hdr)) { 1982 size_t l; 1983 struct ip6_hdr nip6; 1984 1985 l = off - sizeof(struct ip6_hdr); 1986 m_copydata(m, 0, sizeof(nip6), (void *)&nip6); 1987 m_adj(m, l); 1988 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1989 if (m->m_len < l) { 1990 if ((m = m_pullup(m, l)) == NULL) 1991 return; 1992 } 1993 bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6)); 1994 } else /* off == sizeof(struct ip6_hdr) */ { 1995 size_t l; 1996 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 1997 if (m->m_len < l) { 1998 if ((m = m_pullup(m, l)) == NULL) 1999 return; 2000 } 2001 } 2002 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2003 ip6 = mtod(m, struct ip6_hdr *); 2004 ip6->ip6_nxt = IPPROTO_ICMPV6; 2005 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2006 type = icmp6->icmp6_type; /* keep type for statistics */ 2007 code = icmp6->icmp6_code; /* ditto. */ 2008 2009 origdst = ip6->ip6_dst; 2010 /* 2011 * ip6_input() drops a packet if its src is multicast. 2012 * So, the src is never multicast. 2013 */ 2014 ip6->ip6_dst = ip6->ip6_src; 2015 2016 /* 2017 * If the incoming packet was addressed directly to us (i.e. unicast), 2018 * use dst as the src for the reply. 2019 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2020 * (for example) when we encounter an error while forwarding procedure 2021 * destined to a duplicated address of ours. 2022 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2023 * procedure of an outgoing packet of our own, in which case we need 2024 * to search in the ifaddr list. 2025 */ 2026 if (IN6_IS_ADDR_MULTICAST(&origdst)) 2027 ; 2028 else if ((ip6a = ip6_getdstifaddr(m)) != NULL) { 2029 if ((ip6a->ip6a_flags & 2030 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2031 src = &ip6a->ip6a_src; 2032 } else { 2033 union { 2034 struct sockaddr_in6 sin6; 2035 struct sockaddr sa; 2036 } u; 2037 2038 sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0); 2039 2040 ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa); 2041 2042 if (ia == NULL) 2043 ; 2044 else if ((ia->ia6_flags & 2045 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2046 src = &ia->ia_addr.sin6_addr; 2047 } 2048 2049 if (src == NULL) { 2050 int e; 2051 struct sockaddr_in6 sin6; 2052 struct route ro; 2053 2054 /* 2055 * This case matches to multicasts, our anycast, or unicasts 2056 * that we do not own. Select a source address based on the 2057 * source address of the erroneous packet. 2058 */ 2059 /* zone ID should be embedded */ 2060 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 2061 2062 memset(&ro, 0, sizeof(ro)); 2063 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2064 rtcache_free(&ro); 2065 if (src == NULL) { 2066 nd6log((LOG_DEBUG, 2067 "icmp6_reflect: source can't be determined: " 2068 "dst=%s, error=%d\n", 2069 ip6_sprintf(&sin6.sin6_addr), e)); 2070 goto bad; 2071 } 2072 } 2073 2074 ip6->ip6_src = *src; 2075 ip6->ip6_flow = 0; 2076 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2077 ip6->ip6_vfc |= IPV6_VERSION; 2078 ip6->ip6_nxt = IPPROTO_ICMPV6; 2079 if (m->m_pkthdr.rcvif) { 2080 /* XXX: This may not be the outgoing interface */ 2081 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2082 } else 2083 ip6->ip6_hlim = ip6_defhlim; 2084 2085 m->m_pkthdr.csum_flags = 0; 2086 icmp6->icmp6_cksum = 0; 2087 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2088 sizeof(struct ip6_hdr), plen); 2089 2090 /* 2091 * XXX option handling 2092 */ 2093 2094 m->m_flags &= ~(M_BCAST|M_MCAST); 2095 2096 /* 2097 * To avoid a "too big" situation at an intermediate router 2098 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2099 * Note that only echo and node information replies are affected, 2100 * since the length of ICMP6 errors is limited to the minimum MTU. 2101 */ 2102 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 && 2103 outif) 2104 icmp6_ifstat_inc(outif, ifs6_out_error); 2105 2106 if (outif) 2107 icmp6_ifoutstat_inc(outif, type, code); 2108 2109 return; 2110 2111 bad: 2112 m_freem(m); 2113 return; 2114 } 2115 2116 static const char * 2117 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2118 struct in6_addr *tgt6) 2119 { 2120 static char buf[1024]; 2121 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2122 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2123 return buf; 2124 } 2125 2126 void 2127 icmp6_redirect_input(struct mbuf *m, int off) 2128 { 2129 struct ifnet *ifp = m->m_pkthdr.rcvif; 2130 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2131 struct nd_redirect *nd_rd; 2132 int icmp6len = ntohs(ip6->ip6_plen); 2133 char *lladdr = NULL; 2134 int lladdrlen = 0; 2135 struct rtentry *rt = NULL; 2136 int is_router; 2137 int is_onlink; 2138 struct in6_addr src6 = ip6->ip6_src; 2139 struct in6_addr redtgt6; 2140 struct in6_addr reddst6; 2141 union nd_opts ndopts; 2142 2143 if (!ifp) 2144 return; 2145 2146 /* XXX if we are router, we don't update route by icmp6 redirect */ 2147 if (ip6_forwarding) 2148 goto freeit; 2149 if (!icmp6_rediraccept) 2150 goto freeit; 2151 2152 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2153 if (nd_rd == NULL) { 2154 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 2155 return; 2156 } 2157 redtgt6 = nd_rd->nd_rd_target; 2158 reddst6 = nd_rd->nd_rd_dst; 2159 2160 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2161 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2162 goto freeit; 2163 } 2164 2165 /* validation */ 2166 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2167 nd6log((LOG_ERR, 2168 "ICMP6 redirect sent from %s rejected; " 2169 "must be from linklocal\n", ip6_sprintf(&src6))); 2170 goto bad; 2171 } 2172 if (ip6->ip6_hlim != 255) { 2173 nd6log((LOG_ERR, 2174 "ICMP6 redirect sent from %s rejected; " 2175 "hlim=%d (must be 255)\n", 2176 ip6_sprintf(&src6), ip6->ip6_hlim)); 2177 goto bad; 2178 } 2179 { 2180 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2181 struct sockaddr_in6 sin6; 2182 struct in6_addr *gw6; 2183 2184 sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0); 2185 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2186 if (rt) { 2187 if (rt->rt_gateway == NULL || 2188 rt->rt_gateway->sa_family != AF_INET6) { 2189 nd6log((LOG_ERR, 2190 "ICMP6 redirect rejected; no route " 2191 "with inet6 gateway found for redirect dst: %s\n", 2192 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2193 rtfree(rt); 2194 goto bad; 2195 } 2196 2197 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2198 if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2199 nd6log((LOG_ERR, 2200 "ICMP6 redirect rejected; " 2201 "not equal to gw-for-src=%s (must be same): " 2202 "%s\n", 2203 ip6_sprintf(gw6), 2204 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2205 rtfree(rt); 2206 goto bad; 2207 } 2208 } else { 2209 nd6log((LOG_ERR, 2210 "ICMP6 redirect rejected; " 2211 "no route found for redirect dst: %s\n", 2212 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2213 goto bad; 2214 } 2215 rtfree(rt); 2216 rt = NULL; 2217 } 2218 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2219 nd6log((LOG_ERR, 2220 "ICMP6 redirect rejected; " 2221 "redirect dst must be unicast: %s\n", 2222 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2223 goto bad; 2224 } 2225 2226 is_router = is_onlink = 0; 2227 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2228 is_router = 1; /* router case */ 2229 if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2230 is_onlink = 1; /* on-link destination case */ 2231 if (!is_router && !is_onlink) { 2232 nd6log((LOG_ERR, 2233 "ICMP6 redirect rejected; " 2234 "neither router case nor onlink case: %s\n", 2235 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2236 goto bad; 2237 } 2238 /* validation passed */ 2239 2240 icmp6len -= sizeof(*nd_rd); 2241 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2242 if (nd6_options(&ndopts) < 0) { 2243 nd6log((LOG_INFO, "icmp6_redirect_input: " 2244 "invalid ND option, rejected: %s\n", 2245 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2246 /* nd6_options have incremented stats */ 2247 goto freeit; 2248 } 2249 2250 if (ndopts.nd_opts_tgt_lladdr) { 2251 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2252 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2253 } 2254 2255 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2256 nd6log((LOG_INFO, 2257 "icmp6_redirect_input: lladdrlen mismatch for %s " 2258 "(if %d, icmp6 packet %d): %s\n", 2259 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2260 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2261 goto bad; 2262 } 2263 2264 /* RFC 2461 8.3 */ 2265 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2266 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2267 2268 if (!is_onlink) { /* better router case. perform rtredirect. */ 2269 /* perform rtredirect */ 2270 struct sockaddr_in6 sdst; 2271 struct sockaddr_in6 sgw; 2272 struct sockaddr_in6 ssrc; 2273 unsigned long rtcount; 2274 struct rtentry *newrt = NULL; 2275 2276 /* 2277 * do not install redirect route, if the number of entries 2278 * is too much (> hiwat). note that, the node (= host) will 2279 * work just fine even if we do not install redirect route 2280 * (there will be additional hops, though). 2281 */ 2282 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2283 if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes) 2284 goto freeit; 2285 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2286 return; 2287 else if (0 <= icmp6_redirect_lowat && 2288 rtcount > icmp6_redirect_lowat) { 2289 /* 2290 * XXX nuke a victim, install the new one. 2291 */ 2292 } 2293 2294 memset(&sdst, 0, sizeof(sdst)); 2295 memset(&sgw, 0, sizeof(sgw)); 2296 memset(&ssrc, 0, sizeof(ssrc)); 2297 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2298 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2299 sizeof(struct sockaddr_in6); 2300 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2301 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2302 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2303 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2304 NULL, RTF_GATEWAY | RTF_HOST, 2305 (struct sockaddr *)&ssrc, 2306 &newrt); 2307 2308 if (newrt) { 2309 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2310 icmp6_redirect_timeout_q); 2311 rtfree(newrt); 2312 } 2313 } 2314 /* finally update cached route in each socket via pfctlinput */ 2315 { 2316 struct sockaddr_in6 sdst; 2317 2318 sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0); 2319 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2320 #if defined(IPSEC) 2321 if (ipsec_used) 2322 key_sa_routechange((struct sockaddr *)&sdst); 2323 #endif 2324 } 2325 2326 freeit: 2327 m_freem(m); 2328 return; 2329 2330 bad: 2331 ICMP6_STATINC(ICMP6_STAT_BADREDIRECT); 2332 m_freem(m); 2333 } 2334 2335 void 2336 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2337 { 2338 struct ifnet *ifp; /* my outgoing interface */ 2339 struct in6_addr *ifp_ll6; 2340 struct in6_addr *nexthop; 2341 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2342 struct mbuf *m = NULL; /* newly allocated one */ 2343 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2344 struct nd_redirect *nd_rd; 2345 size_t maxlen; 2346 u_char *p; 2347 struct sockaddr_in6 src_sa; 2348 2349 icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0); 2350 2351 /* if we are not router, we don't send icmp6 redirect */ 2352 if (!ip6_forwarding) 2353 goto fail; 2354 2355 /* sanity check */ 2356 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2357 goto fail; 2358 2359 /* 2360 * Address check: 2361 * the source address must identify a neighbor, and 2362 * the destination address must not be a multicast address 2363 * [RFC 2461, sec 8.2] 2364 */ 2365 sip6 = mtod(m0, struct ip6_hdr *); 2366 sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0); 2367 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2368 goto fail; 2369 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2370 goto fail; /* what should we do here? */ 2371 2372 /* rate limit */ 2373 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2374 goto fail; 2375 2376 /* 2377 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2378 * we almost always ask for an mbuf cluster for simplicity. 2379 * (MHLEN < IPV6_MMTU is almost always true) 2380 */ 2381 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2382 if (m && IPV6_MMTU >= MHLEN) { 2383 #if IPV6_MMTU >= MCLBYTES 2384 _MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT); 2385 #else 2386 MCLGET(m, M_DONTWAIT); 2387 #endif 2388 } 2389 2390 if (!m) 2391 goto fail; 2392 m->m_pkthdr.rcvif = NULL; 2393 m->m_len = 0; 2394 maxlen = M_TRAILINGSPACE(m); 2395 maxlen = min(IPV6_MMTU, maxlen); 2396 /* just for safety */ 2397 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2398 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2399 goto fail; 2400 } 2401 2402 { 2403 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2404 struct in6_ifaddr *ia; 2405 if ((ia = in6ifa_ifpforlinklocal(ifp, 2406 IN6_IFF_NOTREADY| 2407 IN6_IFF_ANYCAST)) == NULL) 2408 goto fail; 2409 ifp_ll6 = &ia->ia_addr.sin6_addr; 2410 } 2411 2412 /* get ip6 linklocal address for the router. */ 2413 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2414 struct sockaddr_in6 *sin6; 2415 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2416 nexthop = &sin6->sin6_addr; 2417 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2418 nexthop = NULL; 2419 } else 2420 nexthop = NULL; 2421 2422 /* ip6 */ 2423 ip6 = mtod(m, struct ip6_hdr *); 2424 ip6->ip6_flow = 0; 2425 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2426 ip6->ip6_vfc |= IPV6_VERSION; 2427 /* ip6->ip6_plen will be set later */ 2428 ip6->ip6_nxt = IPPROTO_ICMPV6; 2429 ip6->ip6_hlim = 255; 2430 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2431 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2432 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2433 2434 /* ND Redirect */ 2435 nd_rd = (struct nd_redirect *)(ip6 + 1); 2436 nd_rd->nd_rd_type = ND_REDIRECT; 2437 nd_rd->nd_rd_code = 0; 2438 nd_rd->nd_rd_reserved = 0; 2439 if (rt->rt_flags & RTF_GATEWAY) { 2440 /* 2441 * nd_rd->nd_rd_target must be a link-local address in 2442 * better router cases. 2443 */ 2444 if (!nexthop) 2445 goto fail; 2446 bcopy(nexthop, &nd_rd->nd_rd_target, 2447 sizeof(nd_rd->nd_rd_target)); 2448 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2449 sizeof(nd_rd->nd_rd_dst)); 2450 } else { 2451 /* make sure redtgt == reddst */ 2452 nexthop = &sip6->ip6_dst; 2453 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2454 sizeof(nd_rd->nd_rd_target)); 2455 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2456 sizeof(nd_rd->nd_rd_dst)); 2457 } 2458 2459 p = (u_char *)(nd_rd + 1); 2460 2461 { 2462 /* target lladdr option */ 2463 struct rtentry *rt_nexthop = NULL; 2464 int len; 2465 const struct sockaddr_dl *sdl; 2466 struct nd_opt_hdr *nd_opt; 2467 char *lladdr; 2468 2469 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2470 if (!rt_nexthop) 2471 goto nolladdropt; 2472 len = sizeof(*nd_opt) + ifp->if_addrlen; 2473 len = (len + 7) & ~7; /* round by 8 */ 2474 /* safety check */ 2475 if (len + (p - (u_char *)ip6) > maxlen) 2476 goto nolladdropt; 2477 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2478 (rt_nexthop->rt_flags & RTF_LLINFO) && 2479 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2480 (sdl = satocsdl(rt_nexthop->rt_gateway)) && 2481 sdl->sdl_alen) { 2482 nd_opt = (struct nd_opt_hdr *)p; 2483 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2484 nd_opt->nd_opt_len = len >> 3; 2485 lladdr = (char *)(nd_opt + 1); 2486 memcpy(lladdr, CLLADDR(sdl), ifp->if_addrlen); 2487 p += len; 2488 } 2489 } 2490 nolladdropt:; 2491 2492 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2493 2494 /* just to be safe */ 2495 if (m0->m_flags & M_DECRYPTED) 2496 goto noredhdropt; 2497 if (p - (u_char *)ip6 > maxlen) 2498 goto noredhdropt; 2499 2500 { 2501 /* redirected header option */ 2502 int len; 2503 struct nd_opt_rd_hdr *nd_opt_rh; 2504 2505 /* 2506 * compute the maximum size for icmp6 redirect header option. 2507 * XXX room for auth header? 2508 */ 2509 len = maxlen - (p - (u_char *)ip6); 2510 len &= ~7; 2511 2512 /* 2513 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2514 * about padding/truncate rule for the original IP packet. 2515 * From the discussion on IPv6imp in Feb 1999, 2516 * the consensus was: 2517 * - "attach as much as possible" is the goal 2518 * - pad if not aligned (original size can be guessed by 2519 * original ip6 header) 2520 * Following code adds the padding if it is simple enough, 2521 * and truncates if not. 2522 */ 2523 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2524 /* not enough room, truncate */ 2525 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2526 m0->m_pkthdr.len); 2527 } else { 2528 /* 2529 * enough room, truncate if not aligned. 2530 * we don't pad here for simplicity. 2531 */ 2532 size_t extra; 2533 2534 extra = m0->m_pkthdr.len % 8; 2535 if (extra) { 2536 /* truncate */ 2537 m_adj(m0, -extra); 2538 } 2539 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2540 } 2541 2542 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2543 memset(nd_opt_rh, 0, sizeof(*nd_opt_rh)); 2544 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2545 nd_opt_rh->nd_opt_rh_len = len >> 3; 2546 p += sizeof(*nd_opt_rh); 2547 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2548 2549 /* connect m0 to m */ 2550 m->m_pkthdr.len += m0->m_pkthdr.len; 2551 m_cat(m, m0); 2552 m0 = NULL; 2553 } 2554 noredhdropt: 2555 if (m0) { 2556 m_freem(m0); 2557 m0 = NULL; 2558 } 2559 2560 /* XXX: clear embedded link IDs in the inner header */ 2561 in6_clearscope(&sip6->ip6_src); 2562 in6_clearscope(&sip6->ip6_dst); 2563 in6_clearscope(&nd_rd->nd_rd_target); 2564 in6_clearscope(&nd_rd->nd_rd_dst); 2565 2566 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2567 2568 nd_rd->nd_rd_cksum = 0; 2569 nd_rd->nd_rd_cksum 2570 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2571 2572 /* send the packet to outside... */ 2573 if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0) 2574 icmp6_ifstat_inc(ifp, ifs6_out_error); 2575 2576 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2577 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2578 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT); 2579 2580 return; 2581 2582 fail: 2583 if (m) 2584 m_freem(m); 2585 if (m0) 2586 m_freem(m0); 2587 } 2588 2589 /* 2590 * ICMPv6 socket option processing. 2591 */ 2592 int 2593 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 2594 { 2595 int error = 0; 2596 struct in6pcb *in6p = sotoin6pcb(so); 2597 2598 if (sopt->sopt_level != IPPROTO_ICMPV6) 2599 return rip6_ctloutput(op, so, sopt); 2600 2601 switch (op) { 2602 case PRCO_SETOPT: 2603 switch (sopt->sopt_name) { 2604 case ICMP6_FILTER: 2605 { 2606 struct icmp6_filter fil; 2607 2608 error = sockopt_get(sopt, &fil, sizeof(fil)); 2609 if (error) 2610 break; 2611 memcpy(in6p->in6p_icmp6filt, &fil, 2612 sizeof(struct icmp6_filter)); 2613 error = 0; 2614 break; 2615 } 2616 2617 default: 2618 error = ENOPROTOOPT; 2619 break; 2620 } 2621 break; 2622 2623 case PRCO_GETOPT: 2624 switch (sopt->sopt_name) { 2625 case ICMP6_FILTER: 2626 { 2627 if (in6p->in6p_icmp6filt == NULL) { 2628 error = EINVAL; 2629 break; 2630 } 2631 error = sockopt_set(sopt, in6p->in6p_icmp6filt, 2632 sizeof(struct icmp6_filter)); 2633 break; 2634 } 2635 2636 default: 2637 error = ENOPROTOOPT; 2638 break; 2639 } 2640 break; 2641 } 2642 2643 return (error); 2644 } 2645 2646 /* 2647 * Perform rate limit check. 2648 * Returns 0 if it is okay to send the icmp6 packet. 2649 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2650 * limitation. 2651 * 2652 * XXX per-destination/type check necessary? 2653 */ 2654 static int 2655 icmp6_ratelimit( 2656 const struct in6_addr *dst, /* not used at this moment */ 2657 const int type, /* not used at this moment */ 2658 const int code) /* not used at this moment */ 2659 { 2660 int ret; 2661 2662 ret = 0; /* okay to send */ 2663 2664 /* PPS limit */ 2665 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2666 icmp6errppslim)) { 2667 /* The packet is subject to rate limit */ 2668 ret++; 2669 } 2670 2671 return ret; 2672 } 2673 2674 static struct rtentry * 2675 icmp6_mtudisc_clone(struct sockaddr *dst) 2676 { 2677 struct rtentry *rt; 2678 int error; 2679 2680 rt = rtalloc1(dst, 1); 2681 if (rt == 0) 2682 return NULL; 2683 2684 /* If we didn't get a host route, allocate one */ 2685 if ((rt->rt_flags & RTF_HOST) == 0) { 2686 struct rtentry *nrt; 2687 2688 error = rtrequest((int) RTM_ADD, dst, 2689 (struct sockaddr *) rt->rt_gateway, NULL, 2690 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2691 if (error) { 2692 rtfree(rt); 2693 return NULL; 2694 } 2695 nrt->rt_rmx = rt->rt_rmx; 2696 rtfree(rt); 2697 rt = nrt; 2698 } 2699 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2700 icmp6_mtudisc_timeout_q); 2701 if (error) { 2702 rtfree(rt); 2703 return NULL; 2704 } 2705 2706 return rt; /* caller need to call rtfree() */ 2707 } 2708 2709 static void 2710 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 2711 { 2712 if (rt == NULL) 2713 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2714 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2715 (RTF_DYNAMIC | RTF_HOST)) { 2716 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2717 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2718 } else { 2719 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2720 rt->rt_rmx.rmx_mtu = 0; 2721 } 2722 } 2723 2724 static void 2725 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r) 2726 { 2727 if (rt == NULL) 2728 panic("icmp6_redirect_timeout: bad route to timeout"); 2729 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2730 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2731 rtrequest((int) RTM_DELETE, rt_getkey(rt), 2732 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2733 } 2734 } 2735 2736 /* 2737 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes. silly? 2738 */ 2739 static int 2740 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS) 2741 { 2742 (void)&name; 2743 (void)&l; 2744 (void)&oname; 2745 2746 if (namelen != 0) 2747 return (EINVAL); 2748 2749 return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp, 2750 /*XXXUNCONST*/ 2751 __UNCONST(newp), newlen)); 2752 } 2753 2754 static int 2755 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS) 2756 { 2757 2758 return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS)); 2759 } 2760 2761 static void 2762 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog) 2763 { 2764 extern int nd6_maxqueuelen; /* defined in nd6.c */ 2765 2766 sysctl_createv(clog, 0, NULL, NULL, 2767 CTLFLAG_PERMANENT, 2768 CTLTYPE_NODE, "inet6", NULL, 2769 NULL, 0, NULL, 0, 2770 CTL_NET, PF_INET6, CTL_EOL); 2771 sysctl_createv(clog, 0, NULL, NULL, 2772 CTLFLAG_PERMANENT, 2773 CTLTYPE_NODE, "icmp6", 2774 SYSCTL_DESCR("ICMPv6 related settings"), 2775 NULL, 0, NULL, 0, 2776 CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL); 2777 2778 sysctl_createv(clog, 0, NULL, NULL, 2779 CTLFLAG_PERMANENT, 2780 CTLTYPE_STRUCT, "stats", 2781 SYSCTL_DESCR("ICMPv6 transmission statistics"), 2782 sysctl_net_inet6_icmp6_stats, 0, NULL, 0, 2783 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2784 ICMPV6CTL_STATS, CTL_EOL); 2785 sysctl_createv(clog, 0, NULL, NULL, 2786 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2787 CTLTYPE_INT, "rediraccept", 2788 SYSCTL_DESCR("Accept and process redirect messages"), 2789 NULL, 0, &icmp6_rediraccept, 0, 2790 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2791 ICMPV6CTL_REDIRACCEPT, CTL_EOL); 2792 sysctl_createv(clog, 0, NULL, NULL, 2793 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2794 CTLTYPE_INT, "redirtimeout", 2795 SYSCTL_DESCR("Redirect generated route lifetime"), 2796 NULL, 0, &icmp6_redirtimeout, 0, 2797 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2798 ICMPV6CTL_REDIRTIMEOUT, CTL_EOL); 2799 #if 0 /* obsoleted */ 2800 sysctl_createv(clog, 0, NULL, NULL, 2801 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2802 CTLTYPE_INT, "errratelimit", NULL, 2803 NULL, 0, &icmp6_errratelimit, 0, 2804 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2805 ICMPV6CTL_ERRRATELIMIT, CTL_EOL); 2806 #endif 2807 sysctl_createv(clog, 0, NULL, NULL, 2808 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2809 CTLTYPE_INT, "nd6_prune", 2810 SYSCTL_DESCR("Neighbor discovery prune interval"), 2811 NULL, 0, &nd6_prune, 0, 2812 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2813 ICMPV6CTL_ND6_PRUNE, CTL_EOL); 2814 sysctl_createv(clog, 0, NULL, NULL, 2815 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2816 CTLTYPE_INT, "nd6_delay", 2817 SYSCTL_DESCR("First probe delay time"), 2818 NULL, 0, &nd6_delay, 0, 2819 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2820 ICMPV6CTL_ND6_DELAY, CTL_EOL); 2821 sysctl_createv(clog, 0, NULL, NULL, 2822 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2823 CTLTYPE_INT, "nd6_umaxtries", 2824 SYSCTL_DESCR("Number of unicast discovery attempts"), 2825 NULL, 0, &nd6_umaxtries, 0, 2826 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2827 ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL); 2828 sysctl_createv(clog, 0, NULL, NULL, 2829 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2830 CTLTYPE_INT, "nd6_mmaxtries", 2831 SYSCTL_DESCR("Number of multicast discovery attempts"), 2832 NULL, 0, &nd6_mmaxtries, 0, 2833 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2834 ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL); 2835 sysctl_createv(clog, 0, NULL, NULL, 2836 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2837 CTLTYPE_INT, "nd6_useloopback", 2838 SYSCTL_DESCR("Use loopback interface for local traffic"), 2839 NULL, 0, &nd6_useloopback, 0, 2840 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2841 ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL); 2842 #if 0 /* obsoleted */ 2843 sysctl_createv(clog, 0, NULL, NULL, 2844 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2845 CTLTYPE_INT, "nd6_proxyall", NULL, 2846 NULL, 0, &nd6_proxyall, 0, 2847 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2848 ICMPV6CTL_ND6_PROXYALL, CTL_EOL); 2849 #endif 2850 sysctl_createv(clog, 0, NULL, NULL, 2851 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2852 CTLTYPE_INT, "nodeinfo", 2853 SYSCTL_DESCR("Respond to node information requests"), 2854 NULL, 0, &icmp6_nodeinfo, 0, 2855 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2856 ICMPV6CTL_NODEINFO, CTL_EOL); 2857 sysctl_createv(clog, 0, NULL, NULL, 2858 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2859 CTLTYPE_INT, "errppslimit", 2860 SYSCTL_DESCR("Maximum ICMP errors sent per second"), 2861 NULL, 0, &icmp6errppslim, 0, 2862 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2863 ICMPV6CTL_ERRPPSLIMIT, CTL_EOL); 2864 sysctl_createv(clog, 0, NULL, NULL, 2865 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2866 CTLTYPE_INT, "nd6_maxnudhint", 2867 SYSCTL_DESCR("Maximum neighbor unreachable hint count"), 2868 NULL, 0, &nd6_maxnudhint, 0, 2869 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2870 ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL); 2871 sysctl_createv(clog, 0, NULL, NULL, 2872 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2873 CTLTYPE_INT, "mtudisc_hiwat", 2874 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2875 NULL, 0, &icmp6_mtudisc_hiwat, 0, 2876 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2877 ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL); 2878 sysctl_createv(clog, 0, NULL, NULL, 2879 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2880 CTLTYPE_INT, "mtudisc_lowat", 2881 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2882 NULL, 0, &icmp6_mtudisc_lowat, 0, 2883 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2884 ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL); 2885 sysctl_createv(clog, 0, NULL, NULL, 2886 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2887 CTLTYPE_INT, "nd6_debug", 2888 SYSCTL_DESCR("Enable neighbor discovery debug output"), 2889 NULL, 0, &nd6_debug, 0, 2890 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2891 ICMPV6CTL_ND6_DEBUG, CTL_EOL); 2892 sysctl_createv(clog, 0, NULL, NULL, 2893 CTLFLAG_PERMANENT, 2894 CTLTYPE_STRUCT, "nd6_drlist", 2895 SYSCTL_DESCR("Default router list"), 2896 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2897 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2898 ICMPV6CTL_ND6_DRLIST, CTL_EOL); 2899 sysctl_createv(clog, 0, NULL, NULL, 2900 CTLFLAG_PERMANENT, 2901 CTLTYPE_STRUCT, "nd6_prlist", 2902 SYSCTL_DESCR("Prefix list"), 2903 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2904 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2905 ICMPV6CTL_ND6_PRLIST, CTL_EOL); 2906 sysctl_createv(clog, 0, NULL, NULL, 2907 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2908 CTLTYPE_INT, "maxqueuelen", 2909 SYSCTL_DESCR("max packet queue len for a unresolved ND"), 2910 NULL, 1, &nd6_maxqueuelen, 0, 2911 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2912 ICMPV6CTL_ND6_MAXQLEN, CTL_EOL); 2913 } 2914 2915 void 2916 icmp6_statinc(u_int stat) 2917 { 2918 2919 KASSERT(stat < ICMP6_NSTATS); 2920 ICMP6_STATINC(stat); 2921 } 2922