1 /* $NetBSD: icmp6.c,v 1.77 2002/05/24 09:21:30 itojun 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. All advertising materials mentioning features or use of this software 46 * must display the following acknowledgement: 47 * This product includes software developed by the University of 48 * California, Berkeley and its contributors. 49 * 4. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 66 */ 67 68 #include <sys/cdefs.h> 69 __KERNEL_RCSID(0, "$NetBSD: icmp6.c,v 1.77 2002/05/24 09:21:30 itojun Exp $"); 70 71 #include "opt_inet.h" 72 #include "opt_ipsec.h" 73 74 #include <sys/param.h> 75 #include <sys/systm.h> 76 #include <sys/malloc.h> 77 #include <sys/mbuf.h> 78 #include <sys/protosw.h> 79 #include <sys/socket.h> 80 #include <sys/socketvar.h> 81 #include <sys/time.h> 82 #include <sys/kernel.h> 83 #include <sys/syslog.h> 84 #include <sys/domain.h> 85 #include <sys/sysctl.h> 86 87 #include <net/if.h> 88 #include <net/route.h> 89 #include <net/if_dl.h> 90 #include <net/if_types.h> 91 92 #include <netinet/in.h> 93 #include <netinet/in_var.h> 94 #include <netinet/ip6.h> 95 #include <netinet6/ip6_var.h> 96 #include <netinet/icmp6.h> 97 #include <netinet6/mld6_var.h> 98 #include <netinet6/in6_pcb.h> 99 #include <netinet6/nd6.h> 100 #include <netinet6/in6_ifattach.h> 101 #include <netinet6/ip6protosw.h> 102 103 #ifdef IPSEC 104 #include <netinet6/ipsec.h> 105 #include <netkey/key.h> 106 #endif 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 struct icmp6stat icmp6stat; 118 119 extern struct in6pcb rawin6pcb; 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) __P((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 __P((struct icmp6errstat *, int, int)); 153 static int icmp6_rip6_input __P((struct mbuf **, int)); 154 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int)); 155 static const char *icmp6_redirect_diag __P((struct in6_addr *, 156 struct in6_addr *, struct in6_addr *)); 157 static struct mbuf *ni6_input __P((struct mbuf *, int)); 158 static struct mbuf *ni6_nametodns __P((const char *, int, int)); 159 static int ni6_dnsmatch __P((const char *, int, const char *, int)); 160 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *, 161 struct ifnet **, char *)); 162 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 163 struct ifnet *, int)); 164 static int icmp6_notify_error __P((struct mbuf *, int, int, int)); 165 static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *)); 166 static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *)); 167 static void icmp6_redirect_timeout __P((struct rtentry *, struct rttimer *)); 168 169 void 170 icmp6_init() 171 { 172 mld6_init(); 173 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 174 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 175 } 176 177 static void 178 icmp6_errcount(stat, type, code) 179 struct icmp6errstat *stat; 180 int type, code; 181 { 182 switch (type) { 183 case ICMP6_DST_UNREACH: 184 switch (code) { 185 case ICMP6_DST_UNREACH_NOROUTE: 186 stat->icp6errs_dst_unreach_noroute++; 187 return; 188 case ICMP6_DST_UNREACH_ADMIN: 189 stat->icp6errs_dst_unreach_admin++; 190 return; 191 case ICMP6_DST_UNREACH_BEYONDSCOPE: 192 stat->icp6errs_dst_unreach_beyondscope++; 193 return; 194 case ICMP6_DST_UNREACH_ADDR: 195 stat->icp6errs_dst_unreach_addr++; 196 return; 197 case ICMP6_DST_UNREACH_NOPORT: 198 stat->icp6errs_dst_unreach_noport++; 199 return; 200 } 201 break; 202 case ICMP6_PACKET_TOO_BIG: 203 stat->icp6errs_packet_too_big++; 204 return; 205 case ICMP6_TIME_EXCEEDED: 206 switch (code) { 207 case ICMP6_TIME_EXCEED_TRANSIT: 208 stat->icp6errs_time_exceed_transit++; 209 return; 210 case ICMP6_TIME_EXCEED_REASSEMBLY: 211 stat->icp6errs_time_exceed_reassembly++; 212 return; 213 } 214 break; 215 case ICMP6_PARAM_PROB: 216 switch (code) { 217 case ICMP6_PARAMPROB_HEADER: 218 stat->icp6errs_paramprob_header++; 219 return; 220 case ICMP6_PARAMPROB_NEXTHEADER: 221 stat->icp6errs_paramprob_nextheader++; 222 return; 223 case ICMP6_PARAMPROB_OPTION: 224 stat->icp6errs_paramprob_option++; 225 return; 226 } 227 break; 228 case ND_REDIRECT: 229 stat->icp6errs_redirect++; 230 return; 231 } 232 stat->icp6errs_unknown++; 233 } 234 235 /* 236 * Register a Path MTU Discovery callback. 237 */ 238 void 239 icmp6_mtudisc_callback_register(func) 240 void (*func) __P((struct in6_addr *)); 241 { 242 struct icmp6_mtudisc_callback *mc; 243 244 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 245 mc = LIST_NEXT(mc, mc_list)) { 246 if (mc->mc_func == func) 247 return; 248 } 249 250 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 251 if (mc == NULL) 252 panic("icmp6_mtudisc_callback_register"); 253 254 mc->mc_func = func; 255 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 256 } 257 258 /* 259 * Generate an error packet of type error in response to bad IP6 packet. 260 */ 261 void 262 icmp6_error(m, type, code, param) 263 struct mbuf *m; 264 int type, code, param; 265 { 266 struct ip6_hdr *oip6, *nip6; 267 struct icmp6_hdr *icmp6; 268 u_int preplen; 269 int off; 270 int nxt; 271 272 icmp6stat.icp6s_error++; 273 274 /* count per-type-code statistics */ 275 icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code); 276 277 if (m->m_flags & M_DECRYPTED) { 278 icmp6stat.icp6s_canterror++; 279 goto freeit; 280 } 281 282 #ifndef PULLDOWN_TEST 283 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), ); 284 #else 285 if (m->m_len < sizeof(struct ip6_hdr)) { 286 m = m_pullup(m, sizeof(struct ip6_hdr)); 287 if (m == NULL) 288 return; 289 } 290 #endif 291 oip6 = mtod(m, struct ip6_hdr *); 292 293 /* 294 * If the destination address of the erroneous packet is a multicast 295 * address, or the packet was sent using link-layer multicast, 296 * we should basically suppress sending an error (RFC 2463, Section 297 * 2.4). 298 * We have two exceptions (the item e.2 in that section): 299 * - the Pakcet Too Big message can be sent for path MTU discovery. 300 * - the Parameter Problem Message that can be allowed an icmp6 error 301 * in the option type field. This check has been done in 302 * ip6_unknown_opt(), so we can just check the type and code. 303 */ 304 if ((m->m_flags & (M_BCAST|M_MCAST) || 305 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 306 (type != ICMP6_PACKET_TOO_BIG && 307 (type != ICMP6_PARAM_PROB || 308 code != ICMP6_PARAMPROB_OPTION))) 309 goto freeit; 310 311 /* 312 * RFC 2463, 2.4 (e.5): source address check. 313 * XXX: the case of anycast source? 314 */ 315 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 316 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 317 goto freeit; 318 319 /* 320 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 321 * don't do it. 322 */ 323 nxt = -1; 324 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 325 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 326 struct icmp6_hdr *icp; 327 328 #ifndef PULLDOWN_TEST 329 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), ); 330 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 331 #else 332 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 333 sizeof(*icp)); 334 if (icp == NULL) { 335 icmp6stat.icp6s_tooshort++; 336 return; 337 } 338 #endif 339 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 340 icp->icmp6_type == ND_REDIRECT) { 341 /* 342 * ICMPv6 error 343 * Special case: for redirect (which is 344 * informational) we must not send icmp6 error. 345 */ 346 icmp6stat.icp6s_canterror++; 347 goto freeit; 348 } else { 349 /* ICMPv6 informational - send the error */ 350 } 351 } 352 #if 0 /* controversial */ 353 else if (off >= 0 && nxt == IPPROTO_ESP) { 354 /* 355 * It could be ICMPv6 error inside ESP. Take a safer side, 356 * don't respond. 357 */ 358 icmp6stat.icp6s_canterror++; 359 goto freeit; 360 } 361 #endif 362 else { 363 /* non-ICMPv6 - send the error */ 364 } 365 366 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 367 368 /* Finally, do rate limitation check. */ 369 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 370 icmp6stat.icp6s_toofreq++; 371 goto freeit; 372 } 373 374 /* 375 * OK, ICMP6 can be generated. 376 */ 377 378 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 379 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 380 381 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 382 M_PREPEND(m, preplen, M_DONTWAIT); 383 if (m && m->m_len < preplen) 384 m = m_pullup(m, preplen); 385 if (m == NULL) { 386 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 387 return; 388 } 389 390 nip6 = mtod(m, struct ip6_hdr *); 391 nip6->ip6_src = oip6->ip6_src; 392 nip6->ip6_dst = oip6->ip6_dst; 393 394 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src)) 395 oip6->ip6_src.s6_addr16[1] = 0; 396 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst)) 397 oip6->ip6_dst.s6_addr16[1] = 0; 398 399 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 400 icmp6->icmp6_type = type; 401 icmp6->icmp6_code = code; 402 icmp6->icmp6_pptr = htonl((u_int32_t)param); 403 404 /* 405 * icmp6_reflect() is designed to be in the input path. 406 * icmp6_error() can be called from both input and outut path, 407 * and if we are in output path rcvif could contain bogus value. 408 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 409 * information in ip header (nip6). 410 */ 411 m->m_pkthdr.rcvif = NULL; 412 413 icmp6stat.icp6s_outhist[type]++; 414 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 415 416 return; 417 418 freeit: 419 /* 420 * If we can't tell wheter or not we can generate ICMP6, free it. 421 */ 422 m_freem(m); 423 } 424 425 /* 426 * Process a received ICMP6 message. 427 */ 428 int 429 icmp6_input(mp, offp, proto) 430 struct mbuf **mp; 431 int *offp, proto; 432 { 433 struct mbuf *m = *mp, *n; 434 struct ip6_hdr *ip6, *nip6; 435 struct icmp6_hdr *icmp6, *nicmp6; 436 int off = *offp; 437 int icmp6len = m->m_pkthdr.len - *offp; 438 int code, sum, noff; 439 440 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 441 442 #ifndef PULLDOWN_TEST 443 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE); 444 /* m might change if M_LOOP. So, call mtod after this */ 445 #endif 446 447 /* 448 * Locate icmp6 structure in mbuf, and check 449 * that not corrupted and of at least minimum length 450 */ 451 452 ip6 = mtod(m, struct ip6_hdr *); 453 if (icmp6len < sizeof(struct icmp6_hdr)) { 454 icmp6stat.icp6s_tooshort++; 455 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 456 goto freeit; 457 } 458 459 /* 460 * calculate the checksum 461 */ 462 #ifndef PULLDOWN_TEST 463 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 464 #else 465 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 466 if (icmp6 == NULL) { 467 icmp6stat.icp6s_tooshort++; 468 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 469 return IPPROTO_DONE; 470 } 471 #endif 472 code = icmp6->icmp6_code; 473 474 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 475 nd6log((LOG_ERR, 476 "ICMP6 checksum error(%d|%x) %s\n", 477 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 478 icmp6stat.icp6s_checksum++; 479 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 480 goto freeit; 481 } 482 483 #if defined(NFAITH) && 0 < NFAITH 484 if (faithprefix(&ip6->ip6_dst)) { 485 /* 486 * Deliver very specific ICMP6 type only. 487 * This is important to deilver TOOBIG. Otherwise PMTUD 488 * will not work. 489 */ 490 switch (icmp6->icmp6_type) { 491 case ICMP6_DST_UNREACH: 492 case ICMP6_PACKET_TOO_BIG: 493 case ICMP6_TIME_EXCEEDED: 494 break; 495 default: 496 goto freeit; 497 } 498 } 499 #endif 500 501 icmp6stat.icp6s_inhist[icmp6->icmp6_type]++; 502 503 switch (icmp6->icmp6_type) { 504 case ICMP6_DST_UNREACH: 505 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 506 switch (code) { 507 case ICMP6_DST_UNREACH_NOROUTE: 508 code = PRC_UNREACH_NET; 509 break; 510 case ICMP6_DST_UNREACH_ADMIN: 511 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 512 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 513 break; 514 case ICMP6_DST_UNREACH_ADDR: 515 code = PRC_HOSTDEAD; 516 break; 517 #ifdef COMPAT_RFC1885 518 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 519 code = PRC_UNREACH_SRCFAIL; 520 break; 521 #else 522 case ICMP6_DST_UNREACH_BEYONDSCOPE: 523 /* I mean "source address was incorrect." */ 524 code = PRC_UNREACH_NET; 525 break; 526 #endif 527 case ICMP6_DST_UNREACH_NOPORT: 528 code = PRC_UNREACH_PORT; 529 break; 530 default: 531 goto badcode; 532 } 533 goto deliver; 534 break; 535 536 case ICMP6_PACKET_TOO_BIG: 537 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 538 if (code != 0) 539 goto badcode; 540 541 code = PRC_MSGSIZE; 542 543 /* 544 * Updating the path MTU will be done after examining 545 * intermediate extension headers. 546 */ 547 goto deliver; 548 break; 549 550 case ICMP6_TIME_EXCEEDED: 551 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 552 switch (code) { 553 case ICMP6_TIME_EXCEED_TRANSIT: 554 code = PRC_TIMXCEED_INTRANS; 555 break; 556 case ICMP6_TIME_EXCEED_REASSEMBLY: 557 code = PRC_TIMXCEED_REASS; 558 break; 559 default: 560 goto badcode; 561 } 562 goto deliver; 563 break; 564 565 case ICMP6_PARAM_PROB: 566 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 567 switch (code) { 568 case ICMP6_PARAMPROB_NEXTHEADER: 569 code = PRC_UNREACH_PROTOCOL; 570 break; 571 case ICMP6_PARAMPROB_HEADER: 572 case ICMP6_PARAMPROB_OPTION: 573 code = PRC_PARAMPROB; 574 break; 575 default: 576 goto badcode; 577 } 578 goto deliver; 579 break; 580 581 case ICMP6_ECHO_REQUEST: 582 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 583 if (code != 0) 584 goto badcode; 585 /* 586 * Copy mbuf to send to two data paths: userland socket(s), 587 * and to the querier (echo reply). 588 * m: a copy for socket, n: a copy for querier 589 */ 590 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 591 /* Give up local */ 592 n = m; 593 m = NULL; 594 goto deliverecho; 595 } 596 /* 597 * If the first mbuf is shared, or the first mbuf is too short, 598 * copy the first part of the data into a fresh mbuf. 599 * Otherwise, we will wrongly overwrite both copies. 600 */ 601 if ((n->m_flags & M_EXT) != 0 || 602 n->m_len < off + sizeof(struct icmp6_hdr)) { 603 struct mbuf *n0 = n; 604 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6); 605 606 /* 607 * Prepare an internal mbuf. m_pullup() doesn't 608 * always copy the length we specified. 609 */ 610 if (maxlen >= MCLBYTES) { 611 /* Give up remote */ 612 m_freem(n0); 613 break; 614 } 615 MGETHDR(n, M_DONTWAIT, n0->m_type); 616 if (n && maxlen >= MHLEN) { 617 MCLGET(n, M_DONTWAIT); 618 if ((n->m_flags & M_EXT) == 0) { 619 m_free(n); 620 n = NULL; 621 } 622 } 623 if (n == NULL) { 624 /* Give up local */ 625 m_freem(n0); 626 n = m; 627 m = NULL; 628 goto deliverecho; 629 } 630 M_COPY_PKTHDR(n, n0); 631 /* 632 * Copy IPv6 and ICMPv6 only. 633 */ 634 nip6 = mtod(n, struct ip6_hdr *); 635 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 636 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 637 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 638 noff = sizeof(struct ip6_hdr); 639 n->m_len = noff + sizeof(struct icmp6_hdr); 640 /* 641 * Adjust mbuf. ip6_plen will be adjusted in 642 * ip6_output(). 643 * n->m_pkthdr.len == n0->m_pkthdr.len at this point. 644 */ 645 n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr); 646 n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr)); 647 m_adj(n0, off + sizeof(struct icmp6_hdr)); 648 n->m_next = n0; 649 n0->m_flags &= ~M_PKTHDR; 650 } else { 651 deliverecho: 652 nip6 = mtod(n, struct ip6_hdr *); 653 nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off); 654 noff = off; 655 } 656 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 657 nicmp6->icmp6_code = 0; 658 if (n) { 659 icmp6stat.icp6s_reflect++; 660 icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++; 661 icmp6_reflect(n, noff); 662 } 663 if (!m) 664 goto freeit; 665 break; 666 667 case ICMP6_ECHO_REPLY: 668 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 669 if (code != 0) 670 goto badcode; 671 break; 672 673 case MLD6_LISTENER_QUERY: 674 case MLD6_LISTENER_REPORT: 675 if (icmp6len < sizeof(struct mld6_hdr)) 676 goto badlen; 677 if (icmp6->icmp6_type == MLD6_LISTENER_QUERY) /* XXX: ugly... */ 678 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 679 else 680 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 681 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 682 /* give up local */ 683 mld6_input(m, off); 684 m = NULL; 685 goto freeit; 686 } 687 mld6_input(n, off); 688 /* m stays. */ 689 break; 690 691 case MLD6_LISTENER_DONE: 692 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 693 if (icmp6len < sizeof(struct mld6_hdr)) /* necessary? */ 694 goto badlen; 695 break; /* nothing to be done in kernel */ 696 697 case MLD6_MTRACE_RESP: 698 case MLD6_MTRACE: 699 /* XXX: these two are experimental. not officially defined. */ 700 /* XXX: per-interface statistics? */ 701 break; /* just pass it to applications */ 702 703 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 704 { 705 enum { WRU, FQDN } mode; 706 707 if (!icmp6_nodeinfo) 708 break; 709 710 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 711 mode = WRU; 712 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 713 mode = FQDN; 714 else 715 goto badlen; 716 717 if (mode == FQDN) { 718 #ifndef PULLDOWN_TEST 719 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo), 720 IPPROTO_DONE); 721 #endif 722 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 723 if (n) 724 n = ni6_input(n, off); 725 /* XXX meaningless if n == NULL */ 726 noff = sizeof(struct ip6_hdr); 727 } else { 728 u_char *p; 729 int maxlen, maxhlen; 730 731 if ((icmp6_nodeinfo & 5) != 5) 732 break; 733 734 if (code != 0) 735 goto badcode; 736 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; 737 if (maxlen >= MCLBYTES) { 738 /* Give up remote */ 739 break; 740 } 741 MGETHDR(n, M_DONTWAIT, m->m_type); 742 if (n && maxlen > MHLEN) { 743 MCLGET(n, M_DONTWAIT); 744 if ((n->m_flags & M_EXT) == 0) { 745 m_free(n); 746 n = NULL; 747 } 748 } 749 if (n == NULL) { 750 /* Give up remote */ 751 break; 752 } 753 n->m_pkthdr.rcvif = NULL; 754 n->m_len = 0; 755 maxhlen = M_TRAILINGSPACE(n) - maxlen; 756 if (maxhlen > hostnamelen) 757 maxhlen = hostnamelen; 758 /* 759 * Copy IPv6 and ICMPv6 only. 760 */ 761 nip6 = mtod(n, struct ip6_hdr *); 762 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 763 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 764 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 765 p = (u_char *)(nicmp6 + 1); 766 bzero(p, 4); 767 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 768 noff = sizeof(struct ip6_hdr); 769 M_COPY_PKTHDR(n, m); /* just for rcvif */ 770 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 771 sizeof(struct icmp6_hdr) + 4 + maxhlen; 772 nicmp6->icmp6_type = ICMP6_WRUREPLY; 773 nicmp6->icmp6_code = 0; 774 } 775 #undef hostnamelen 776 if (n) { 777 icmp6stat.icp6s_reflect++; 778 icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++; 779 icmp6_reflect(n, noff); 780 } 781 break; 782 } 783 784 case ICMP6_WRUREPLY: 785 if (code != 0) 786 goto badcode; 787 break; 788 789 case ND_ROUTER_SOLICIT: 790 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 791 if (code != 0) 792 goto badcode; 793 if (icmp6len < sizeof(struct nd_router_solicit)) 794 goto badlen; 795 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 796 /* give up local */ 797 nd6_rs_input(m, off, icmp6len); 798 m = NULL; 799 goto freeit; 800 } 801 nd6_rs_input(n, off, icmp6len); 802 /* m stays. */ 803 break; 804 805 case ND_ROUTER_ADVERT: 806 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 807 if (code != 0) 808 goto badcode; 809 if (icmp6len < sizeof(struct nd_router_advert)) 810 goto badlen; 811 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 812 /* give up local */ 813 nd6_ra_input(m, off, icmp6len); 814 m = NULL; 815 goto freeit; 816 } 817 nd6_ra_input(n, off, icmp6len); 818 /* m stays. */ 819 break; 820 821 case ND_NEIGHBOR_SOLICIT: 822 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 823 if (code != 0) 824 goto badcode; 825 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 826 goto badlen; 827 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 828 /* give up local */ 829 nd6_ns_input(m, off, icmp6len); 830 m = NULL; 831 goto freeit; 832 } 833 nd6_ns_input(n, off, icmp6len); 834 /* m stays. */ 835 break; 836 837 case ND_NEIGHBOR_ADVERT: 838 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 839 if (code != 0) 840 goto badcode; 841 if (icmp6len < sizeof(struct nd_neighbor_advert)) 842 goto badlen; 843 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 844 /* give up local */ 845 nd6_na_input(m, off, icmp6len); 846 m = NULL; 847 goto freeit; 848 } 849 nd6_na_input(n, off, icmp6len); 850 /* m stays. */ 851 break; 852 853 case ND_REDIRECT: 854 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 855 if (code != 0) 856 goto badcode; 857 if (icmp6len < sizeof(struct nd_redirect)) 858 goto badlen; 859 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 860 /* give up local */ 861 icmp6_redirect_input(m, off); 862 m = NULL; 863 goto freeit; 864 } 865 icmp6_redirect_input(n, off); 866 /* m stays. */ 867 break; 868 869 case ICMP6_ROUTER_RENUMBERING: 870 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 871 code != ICMP6_ROUTER_RENUMBERING_RESULT) 872 goto badcode; 873 if (icmp6len < sizeof(struct icmp6_router_renum)) 874 goto badlen; 875 break; 876 877 default: 878 nd6log((LOG_DEBUG, 879 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 880 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 881 ip6_sprintf(&ip6->ip6_dst), 882 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 883 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 884 /* ICMPv6 error: MUST deliver it by spec... */ 885 code = PRC_NCMDS; 886 /* deliver */ 887 } else { 888 /* ICMPv6 informational: MUST not deliver */ 889 break; 890 } 891 deliver: 892 if (icmp6_notify_error(m, off, icmp6len, code)) { 893 /* In this case, m should've been freed. */ 894 return(IPPROTO_DONE); 895 } 896 break; 897 898 badcode: 899 icmp6stat.icp6s_badcode++; 900 break; 901 902 badlen: 903 icmp6stat.icp6s_badlen++; 904 break; 905 } 906 907 /* deliver the packet to appropriate sockets */ 908 icmp6_rip6_input(&m, *offp); 909 910 return IPPROTO_DONE; 911 912 freeit: 913 m_freem(m); 914 return IPPROTO_DONE; 915 } 916 917 static int 918 icmp6_notify_error(m, off, icmp6len, code) 919 struct mbuf *m; 920 int off, icmp6len; 921 { 922 struct icmp6_hdr *icmp6; 923 struct ip6_hdr *eip6; 924 u_int32_t notifymtu; 925 struct sockaddr_in6 icmp6src, icmp6dst; 926 927 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 928 icmp6stat.icp6s_tooshort++; 929 goto freeit; 930 } 931 #ifndef PULLDOWN_TEST 932 IP6_EXTHDR_CHECK(m, off, 933 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), 934 -1); 935 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 936 #else 937 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 938 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 939 if (icmp6 == NULL) { 940 icmp6stat.icp6s_tooshort++; 941 return(-1); 942 } 943 #endif 944 eip6 = (struct ip6_hdr *)(icmp6 + 1); 945 946 /* Detect the upper level protocol */ 947 { 948 void (*ctlfunc) __P((int, struct sockaddr *, void *)); 949 u_int8_t nxt = eip6->ip6_nxt; 950 int eoff = off + sizeof(struct icmp6_hdr) + 951 sizeof(struct ip6_hdr); 952 struct ip6ctlparam ip6cp; 953 struct in6_addr *finaldst = NULL; 954 int icmp6type = icmp6->icmp6_type; 955 struct ip6_frag *fh; 956 struct ip6_rthdr *rth; 957 struct ip6_rthdr0 *rth0; 958 int rthlen; 959 960 while (1) { /* XXX: should avoid infinite loop explicitly? */ 961 struct ip6_ext *eh; 962 963 switch (nxt) { 964 case IPPROTO_HOPOPTS: 965 case IPPROTO_DSTOPTS: 966 case IPPROTO_AH: 967 #ifndef PULLDOWN_TEST 968 IP6_EXTHDR_CHECK(m, 0, eoff + 969 sizeof(struct ip6_ext), 970 -1); 971 eh = (struct ip6_ext *)(mtod(m, caddr_t) 972 + eoff); 973 #else 974 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 975 eoff, sizeof(*eh)); 976 if (eh == NULL) { 977 icmp6stat.icp6s_tooshort++; 978 return(-1); 979 } 980 #endif 981 982 if (nxt == IPPROTO_AH) 983 eoff += (eh->ip6e_len + 2) << 2; 984 else 985 eoff += (eh->ip6e_len + 1) << 3; 986 nxt = eh->ip6e_nxt; 987 break; 988 case IPPROTO_ROUTING: 989 /* 990 * When the erroneous packet contains a 991 * routing header, we should examine the 992 * header to determine the final destination. 993 * Otherwise, we can't properly update 994 * information that depends on the final 995 * destination (e.g. path MTU). 996 */ 997 #ifndef PULLDOWN_TEST 998 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), 999 -1); 1000 rth = (struct ip6_rthdr *)(mtod(m, caddr_t) 1001 + eoff); 1002 #else 1003 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 1004 eoff, sizeof(*rth)); 1005 if (rth == NULL) { 1006 icmp6stat.icp6s_tooshort++; 1007 return(-1); 1008 } 1009 #endif 1010 rthlen = (rth->ip6r_len + 1) << 3; 1011 /* 1012 * XXX: currently there is no 1013 * officially defined type other 1014 * than type-0. 1015 * Note that if the segment left field 1016 * is 0, all intermediate hops must 1017 * have been passed. 1018 */ 1019 if (rth->ip6r_segleft && 1020 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 1021 int hops; 1022 1023 #ifndef PULLDOWN_TEST 1024 IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, 1025 -1); 1026 rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff); 1027 #else 1028 IP6_EXTHDR_GET(rth0, 1029 struct ip6_rthdr0 *, m, 1030 eoff, rthlen); 1031 if (rth0 == NULL) { 1032 icmp6stat.icp6s_tooshort++; 1033 return(-1); 1034 } 1035 #endif 1036 /* just ignore a bogus header */ 1037 if ((rth0->ip6r0_len % 2) == 0 && 1038 (hops = rth0->ip6r0_len/2)) 1039 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 1040 } 1041 eoff += rthlen; 1042 nxt = rth->ip6r_nxt; 1043 break; 1044 case IPPROTO_FRAGMENT: 1045 #ifndef PULLDOWN_TEST 1046 IP6_EXTHDR_CHECK(m, 0, eoff + 1047 sizeof(struct ip6_frag), 1048 -1); 1049 fh = (struct ip6_frag *)(mtod(m, caddr_t) 1050 + eoff); 1051 #else 1052 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 1053 eoff, sizeof(*fh)); 1054 if (fh == NULL) { 1055 icmp6stat.icp6s_tooshort++; 1056 return(-1); 1057 } 1058 #endif 1059 /* 1060 * Data after a fragment header is meaningless 1061 * unless it is the first fragment, but 1062 * we'll go to the notify label for path MTU 1063 * discovery. 1064 */ 1065 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1066 goto notify; 1067 1068 eoff += sizeof(struct ip6_frag); 1069 nxt = fh->ip6f_nxt; 1070 break; 1071 default: 1072 /* 1073 * This case includes ESP and the No Next 1074 * Header. In such cases going to the notify 1075 * label does not have any meaning 1076 * (i.e. ctlfunc will be NULL), but we go 1077 * anyway since we might have to update 1078 * path MTU information. 1079 */ 1080 goto notify; 1081 } 1082 } 1083 notify: 1084 #ifndef PULLDOWN_TEST 1085 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 1086 #else 1087 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1088 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1089 if (icmp6 == NULL) { 1090 icmp6stat.icp6s_tooshort++; 1091 return(-1); 1092 } 1093 #endif 1094 1095 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1096 bzero(&icmp6dst, sizeof(icmp6dst)); 1097 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1098 icmp6dst.sin6_family = AF_INET6; 1099 if (finaldst == NULL) 1100 icmp6dst.sin6_addr = eip6->ip6_dst; 1101 else 1102 icmp6dst.sin6_addr = *finaldst; 1103 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1104 &icmp6dst.sin6_addr); 1105 #ifndef SCOPEDROUTING 1106 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst, 1107 NULL, NULL)) { 1108 /* should be impossbile */ 1109 nd6log((LOG_DEBUG, 1110 "icmp6_notify_error: in6_embedscope failed\n")); 1111 goto freeit; 1112 } 1113 #endif 1114 1115 /* 1116 * retrieve parameters from the inner IPv6 header, and convert 1117 * them into sockaddr structures. 1118 */ 1119 bzero(&icmp6src, sizeof(icmp6src)); 1120 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1121 icmp6src.sin6_family = AF_INET6; 1122 icmp6src.sin6_addr = eip6->ip6_src; 1123 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1124 &icmp6src.sin6_addr); 1125 #ifndef SCOPEDROUTING 1126 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src, 1127 NULL, NULL)) { 1128 /* should be impossbile */ 1129 nd6log((LOG_DEBUG, 1130 "icmp6_notify_error: in6_embedscope failed\n")); 1131 goto freeit; 1132 } 1133 #endif 1134 icmp6src.sin6_flowinfo = 1135 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1136 1137 if (finaldst == NULL) 1138 finaldst = &eip6->ip6_dst; 1139 ip6cp.ip6c_m = m; 1140 ip6cp.ip6c_icmp6 = icmp6; 1141 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1142 ip6cp.ip6c_off = eoff; 1143 ip6cp.ip6c_finaldst = finaldst; 1144 ip6cp.ip6c_src = &icmp6src; 1145 ip6cp.ip6c_nxt = nxt; 1146 1147 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1148 notifymtu = ntohl(icmp6->icmp6_mtu); 1149 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1150 } 1151 1152 ctlfunc = (void (*) __P((int, struct sockaddr *, void *))) 1153 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1154 if (ctlfunc) { 1155 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1156 &ip6cp); 1157 } 1158 } 1159 return(0); 1160 1161 freeit: 1162 m_freem(m); 1163 return(-1); 1164 } 1165 1166 void 1167 icmp6_mtudisc_update(ip6cp, validated) 1168 struct ip6ctlparam *ip6cp; 1169 int validated; 1170 { 1171 unsigned long rtcount; 1172 struct icmp6_mtudisc_callback *mc; 1173 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1174 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1175 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1176 u_int mtu = ntohl(icmp6->icmp6_mtu); 1177 struct rtentry *rt = NULL; 1178 struct sockaddr_in6 sin6; 1179 1180 /* 1181 * allow non-validated cases if memory is plenty, to make traffic 1182 * from non-connected pcb happy. 1183 */ 1184 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1185 if (validated) { 1186 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1187 return; 1188 else if (0 <= icmp6_mtudisc_lowat && 1189 rtcount > icmp6_mtudisc_lowat) { 1190 /* 1191 * XXX nuke a victim, install the new one. 1192 */ 1193 } 1194 } else { 1195 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1196 return; 1197 } 1198 1199 bzero(&sin6, sizeof(sin6)); 1200 sin6.sin6_family = PF_INET6; 1201 sin6.sin6_len = sizeof(struct sockaddr_in6); 1202 sin6.sin6_addr = *dst; 1203 /* XXX normally, this won't happen */ 1204 if (IN6_IS_ADDR_LINKLOCAL(dst)) { 1205 sin6.sin6_addr.s6_addr16[1] = 1206 htons(m->m_pkthdr.rcvif->if_index); 1207 } 1208 /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */ 1209 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1210 1211 if (rt && (rt->rt_flags & RTF_HOST) 1212 && !(rt->rt_rmx.rmx_locks & RTV_MTU)) { 1213 if (mtu < IPV6_MMTU) { 1214 /* xxx */ 1215 rt->rt_rmx.rmx_locks |= RTV_MTU; 1216 } else if (mtu < rt->rt_ifp->if_mtu && 1217 rt->rt_rmx.rmx_mtu > mtu) { 1218 icmp6stat.icp6s_pmtuchg++; 1219 rt->rt_rmx.rmx_mtu = mtu; 1220 } 1221 } 1222 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */ 1223 RTFREE(rt); 1224 } 1225 1226 /* 1227 * Notify protocols that the MTU for this destination 1228 * has changed. 1229 */ 1230 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1231 mc = LIST_NEXT(mc, mc_list)) 1232 (*mc->mc_func)(&sin6.sin6_addr); 1233 } 1234 1235 /* 1236 * Process a Node Information Query packet, based on 1237 * draft-ietf-ipngwg-icmp-name-lookups-07. 1238 * 1239 * Spec incompatibilities: 1240 * - IPv6 Subject address handling 1241 * - IPv4 Subject address handling support missing 1242 * - Proxy reply (answer even if it's not for me) 1243 * - joins NI group address at in6_ifattach() time only, does not cope 1244 * with hostname changes by sethostname(3) 1245 */ 1246 #ifndef offsetof /* XXX */ 1247 #define offsetof(type, member) ((size_t)(&((type *)0)->member)) 1248 #endif 1249 static struct mbuf * 1250 ni6_input(m, off) 1251 struct mbuf *m; 1252 int off; 1253 { 1254 struct icmp6_nodeinfo *ni6, *nni6; 1255 struct mbuf *n = NULL; 1256 u_int16_t qtype; 1257 int subjlen; 1258 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1259 struct ni_reply_fqdn *fqdn; 1260 int addrs; /* for NI_QTYPE_NODEADDR */ 1261 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1262 struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */ 1263 struct ip6_hdr *ip6; 1264 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1265 char *subj = NULL; 1266 1267 ip6 = mtod(m, struct ip6_hdr *); 1268 #ifndef PULLDOWN_TEST 1269 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off); 1270 #else 1271 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1272 if (ni6 == NULL) { 1273 /* m is already reclaimed */ 1274 return NULL; 1275 } 1276 #endif 1277 1278 /* 1279 * Validate IPv6 destination address. 1280 * 1281 * The Responder must discard the Query without further processing 1282 * unless it is one of the Responder's unicast or anycast addresses, or 1283 * a link-local scope multicast address which the Responder has joined. 1284 * [icmp-name-lookups-07, Section 4.] 1285 */ 1286 bzero(&sin6, sizeof(sin6)); 1287 sin6.sin6_family = AF_INET6; 1288 sin6.sin6_len = sizeof(struct sockaddr_in6); 1289 bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr)); 1290 /* XXX scopeid */ 1291 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1292 ; /* unicast/anycast, fine */ 1293 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1294 ; /* link-local multicast, fine */ 1295 else 1296 goto bad; 1297 1298 /* validate query Subject field. */ 1299 qtype = ntohs(ni6->ni_qtype); 1300 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1301 switch (qtype) { 1302 case NI_QTYPE_NOOP: 1303 case NI_QTYPE_SUPTYPES: 1304 /* 07 draft */ 1305 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1306 break; 1307 /* FALLTHROUGH */ 1308 case NI_QTYPE_FQDN: 1309 case NI_QTYPE_NODEADDR: 1310 switch (ni6->ni_code) { 1311 case ICMP6_NI_SUBJ_IPV6: 1312 #if ICMP6_NI_SUBJ_IPV6 != 0 1313 case 0: 1314 #endif 1315 /* 1316 * backward compatibility - try to accept 03 draft 1317 * format, where no Subject is present. 1318 */ 1319 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1320 subjlen == 0) { 1321 oldfqdn++; 1322 break; 1323 } 1324 #if ICMP6_NI_SUBJ_IPV6 != 0 1325 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1326 goto bad; 1327 #endif 1328 1329 if (subjlen != sizeof(sin6.sin6_addr)) 1330 goto bad; 1331 1332 /* 1333 * Validate Subject address. 1334 * 1335 * Not sure what exactly "address belongs to the node" 1336 * means in the spec, is it just unicast, or what? 1337 * 1338 * At this moment we consider Subject address as 1339 * "belong to the node" if the Subject address equals 1340 * to the IPv6 destination address; validation for 1341 * IPv6 destination address should have done enough 1342 * check for us. 1343 * 1344 * We do not do proxy at this moment. 1345 */ 1346 /* m_pulldown instead of copy? */ 1347 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1348 subjlen, (caddr_t)&sin6.sin6_addr); 1349 /* XXX kame scope hack */ 1350 if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) { 1351 if ((m->m_flags & M_PKTHDR) != 0 && 1352 m->m_pkthdr.rcvif) { 1353 sin6.sin6_addr.s6_addr16[1] = 1354 htons(m->m_pkthdr.rcvif->if_index); 1355 } 1356 } 1357 subj = (char *)&sin6; 1358 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr)) 1359 break; 1360 1361 /* 1362 * XXX if we are to allow other cases, we should really 1363 * be careful about scope here. 1364 * basically, we should disallow queries toward IPv6 1365 * destination X with subject Y, if scope(X) > scope(Y). 1366 * if we allow scope(X) > scope(Y), it will result in 1367 * information leakage across scope boundary. 1368 */ 1369 goto bad; 1370 1371 case ICMP6_NI_SUBJ_FQDN: 1372 /* 1373 * Validate Subject name with gethostname(3). 1374 * 1375 * The behavior may need some debate, since: 1376 * - we are not sure if the node has FQDN as 1377 * hostname (returned by gethostname(3)). 1378 * - the code does wildcard match for truncated names. 1379 * however, we are not sure if we want to perform 1380 * wildcard match, if gethostname(3) side has 1381 * truncated hostname. 1382 */ 1383 n = ni6_nametodns(hostname, hostnamelen, 0); 1384 if (!n || n->m_next || n->m_len == 0) 1385 goto bad; 1386 IP6_EXTHDR_GET(subj, char *, m, 1387 off + sizeof(struct icmp6_nodeinfo), subjlen); 1388 if (subj == NULL) 1389 goto bad; 1390 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1391 n->m_len)) { 1392 goto bad; 1393 } 1394 m_freem(n); 1395 n = NULL; 1396 break; 1397 1398 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1399 default: 1400 goto bad; 1401 } 1402 break; 1403 } 1404 1405 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1406 switch (qtype) { 1407 case NI_QTYPE_FQDN: 1408 if ((icmp6_nodeinfo & 1) == 0) 1409 goto bad; 1410 break; 1411 case NI_QTYPE_NODEADDR: 1412 if ((icmp6_nodeinfo & 2) == 0) 1413 goto bad; 1414 break; 1415 } 1416 1417 /* guess reply length */ 1418 switch (qtype) { 1419 case NI_QTYPE_NOOP: 1420 break; /* no reply data */ 1421 case NI_QTYPE_SUPTYPES: 1422 replylen += sizeof(u_int32_t); 1423 break; 1424 case NI_QTYPE_FQDN: 1425 /* XXX will append an mbuf */ 1426 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1427 break; 1428 case NI_QTYPE_NODEADDR: 1429 addrs = ni6_addrs(ni6, m, &ifp, subj); 1430 if ((replylen += addrs * (sizeof(struct in6_addr) + 1431 sizeof(u_int32_t))) > MCLBYTES) 1432 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1433 break; 1434 default: 1435 /* 1436 * XXX: We must return a reply with the ICMP6 code 1437 * `unknown Qtype' in this case. However we regard the case 1438 * as an FQDN query for backward compatibility. 1439 * Older versions set a random value to this field, 1440 * so it rarely varies in the defined qtypes. 1441 * But the mechanism is not reliable... 1442 * maybe we should obsolete older versions. 1443 */ 1444 qtype = NI_QTYPE_FQDN; 1445 /* XXX will append an mbuf */ 1446 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1447 oldfqdn++; 1448 break; 1449 } 1450 1451 /* allocate an mbuf to reply. */ 1452 MGETHDR(n, M_DONTWAIT, m->m_type); 1453 if (n == NULL) { 1454 m_freem(m); 1455 return(NULL); 1456 } 1457 M_COPY_PKTHDR(n, m); /* just for rcvif */ 1458 if (replylen > MHLEN) { 1459 if (replylen > MCLBYTES) { 1460 /* 1461 * XXX: should we try to allocate more? But MCLBYTES 1462 * is probably much larger than IPV6_MMTU... 1463 */ 1464 goto bad; 1465 } 1466 MCLGET(n, M_DONTWAIT); 1467 if ((n->m_flags & M_EXT) == 0) { 1468 goto bad; 1469 } 1470 } 1471 n->m_pkthdr.len = n->m_len = replylen; 1472 1473 /* copy mbuf header and IPv6 + Node Information base headers */ 1474 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1475 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1476 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1477 1478 /* qtype dependent procedure */ 1479 switch (qtype) { 1480 case NI_QTYPE_NOOP: 1481 nni6->ni_code = ICMP6_NI_SUCCESS; 1482 nni6->ni_flags = 0; 1483 break; 1484 case NI_QTYPE_SUPTYPES: 1485 { 1486 u_int32_t v; 1487 nni6->ni_code = ICMP6_NI_SUCCESS; 1488 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1489 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1490 v = (u_int32_t)htonl(0x0000000f); 1491 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1492 break; 1493 } 1494 case NI_QTYPE_FQDN: 1495 nni6->ni_code = ICMP6_NI_SUCCESS; 1496 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1497 sizeof(struct ip6_hdr) + 1498 sizeof(struct icmp6_nodeinfo)); 1499 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1500 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1501 /* 1502 * XXX do we really have FQDN in variable "hostname"? 1503 */ 1504 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1505 if (n->m_next == NULL) 1506 goto bad; 1507 /* XXX we assume that n->m_next is not a chain */ 1508 if (n->m_next->m_next != NULL) 1509 goto bad; 1510 n->m_pkthdr.len += n->m_next->m_len; 1511 break; 1512 case NI_QTYPE_NODEADDR: 1513 { 1514 int lenlim, copied; 1515 1516 nni6->ni_code = ICMP6_NI_SUCCESS; 1517 n->m_pkthdr.len = n->m_len = 1518 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1519 lenlim = M_TRAILINGSPACE(n); 1520 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1521 /* XXX: reset mbuf length */ 1522 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1523 sizeof(struct icmp6_nodeinfo) + copied; 1524 break; 1525 } 1526 default: 1527 break; /* XXX impossible! */ 1528 } 1529 1530 nni6->ni_type = ICMP6_NI_REPLY; 1531 m_freem(m); 1532 return(n); 1533 1534 bad: 1535 m_freem(m); 1536 if (n) 1537 m_freem(n); 1538 return(NULL); 1539 } 1540 #undef hostnamelen 1541 1542 #define isupper(x) ('A' <= (x) && (x) <= 'Z') 1543 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1544 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1545 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1546 1547 /* 1548 * make a mbuf with DNS-encoded string. no compression support. 1549 * 1550 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1551 * treated as truncated name (two \0 at the end). this is a wild guess. 1552 */ 1553 static struct mbuf * 1554 ni6_nametodns(name, namelen, old) 1555 const char *name; 1556 int namelen; 1557 int old; /* return pascal string if non-zero */ 1558 { 1559 struct mbuf *m; 1560 char *cp, *ep; 1561 const char *p, *q; 1562 int i, len, nterm; 1563 1564 if (old) 1565 len = namelen + 1; 1566 else 1567 len = MCLBYTES; 1568 1569 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1570 MGET(m, M_DONTWAIT, MT_DATA); 1571 if (m && len > MLEN) { 1572 MCLGET(m, M_DONTWAIT); 1573 if ((m->m_flags & M_EXT) == 0) 1574 goto fail; 1575 } 1576 if (!m) 1577 goto fail; 1578 m->m_next = NULL; 1579 1580 if (old) { 1581 m->m_len = len; 1582 *mtod(m, char *) = namelen; 1583 bcopy(name, mtod(m, char *) + 1, namelen); 1584 return m; 1585 } else { 1586 m->m_len = 0; 1587 cp = mtod(m, char *); 1588 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1589 1590 /* if not certain about my name, return empty buffer */ 1591 if (namelen == 0) 1592 return m; 1593 1594 /* 1595 * guess if it looks like shortened hostname, or FQDN. 1596 * shortened hostname needs two trailing "\0". 1597 */ 1598 i = 0; 1599 for (p = name; p < name + namelen; p++) { 1600 if (*p && *p == '.') 1601 i++; 1602 } 1603 if (i < 2) 1604 nterm = 2; 1605 else 1606 nterm = 1; 1607 1608 p = name; 1609 while (cp < ep && p < name + namelen) { 1610 i = 0; 1611 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1612 i++; 1613 /* result does not fit into mbuf */ 1614 if (cp + i + 1 >= ep) 1615 goto fail; 1616 /* 1617 * DNS label length restriction, RFC1035 page 8. 1618 * "i == 0" case is included here to avoid returning 1619 * 0-length label on "foo..bar". 1620 */ 1621 if (i <= 0 || i >= 64) 1622 goto fail; 1623 *cp++ = i; 1624 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1625 goto fail; 1626 while (i > 0) { 1627 if (!isalnum(*p) && *p != '-') 1628 goto fail; 1629 if (isupper(*p)) 1630 *cp++ = tolower(*p++); 1631 else 1632 *cp++ = *p++; 1633 i--; 1634 } 1635 p = q; 1636 if (p < name + namelen && *p == '.') 1637 p++; 1638 } 1639 /* termination */ 1640 if (cp + nterm >= ep) 1641 goto fail; 1642 while (nterm-- > 0) 1643 *cp++ = '\0'; 1644 m->m_len = cp - mtod(m, char *); 1645 return m; 1646 } 1647 1648 panic("should not reach here"); 1649 /* NOTREACHED */ 1650 1651 fail: 1652 if (m) 1653 m_freem(m); 1654 return NULL; 1655 } 1656 1657 /* 1658 * check if two DNS-encoded string matches. takes care of truncated 1659 * form (with \0\0 at the end). no compression support. 1660 * XXX upper/lowercase match (see RFC2065) 1661 */ 1662 static int 1663 ni6_dnsmatch(a, alen, b, blen) 1664 const char *a; 1665 int alen; 1666 const char *b; 1667 int blen; 1668 { 1669 const char *a0, *b0; 1670 int l; 1671 1672 /* simplest case - need validation? */ 1673 if (alen == blen && bcmp(a, b, alen) == 0) 1674 return 1; 1675 1676 a0 = a; 1677 b0 = b; 1678 1679 /* termination is mandatory */ 1680 if (alen < 2 || blen < 2) 1681 return 0; 1682 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1683 return 0; 1684 alen--; 1685 blen--; 1686 1687 while (a - a0 < alen && b - b0 < blen) { 1688 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1689 return 0; 1690 1691 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1692 return 0; 1693 /* we don't support compression yet */ 1694 if (a[0] >= 64 || b[0] >= 64) 1695 return 0; 1696 1697 /* truncated case */ 1698 if (a[0] == 0 && a - a0 == alen - 1) 1699 return 1; 1700 if (b[0] == 0 && b - b0 == blen - 1) 1701 return 1; 1702 if (a[0] == 0 || b[0] == 0) 1703 return 0; 1704 1705 if (a[0] != b[0]) 1706 return 0; 1707 l = a[0]; 1708 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1709 return 0; 1710 if (bcmp(a + 1, b + 1, l) != 0) 1711 return 0; 1712 1713 a += 1 + l; 1714 b += 1 + l; 1715 } 1716 1717 if (a - a0 == alen && b - b0 == blen) 1718 return 1; 1719 else 1720 return 0; 1721 } 1722 1723 /* 1724 * calculate the number of addresses to be returned in the node info reply. 1725 */ 1726 static int 1727 ni6_addrs(ni6, m, ifpp, subj) 1728 struct icmp6_nodeinfo *ni6; 1729 struct mbuf *m; 1730 struct ifnet **ifpp; 1731 char *subj; 1732 { 1733 struct ifnet *ifp; 1734 struct in6_ifaddr *ifa6; 1735 struct ifaddr *ifa; 1736 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1737 int addrs = 0, addrsofif, iffound = 0; 1738 int niflags = ni6->ni_flags; 1739 1740 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1741 switch (ni6->ni_code) { 1742 case ICMP6_NI_SUBJ_IPV6: 1743 if (subj == NULL) /* must be impossible... */ 1744 return(0); 1745 subj_ip6 = (struct sockaddr_in6 *)subj; 1746 break; 1747 default: 1748 /* 1749 * XXX: we only support IPv6 subject address for 1750 * this Qtype. 1751 */ 1752 return(0); 1753 } 1754 } 1755 1756 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1757 { 1758 addrsofif = 0; 1759 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1760 ifa = ifa->ifa_list.tqe_next) 1761 { 1762 if (ifa->ifa_addr->sa_family != AF_INET6) 1763 continue; 1764 ifa6 = (struct in6_ifaddr *)ifa; 1765 1766 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1767 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1768 &ifa6->ia_addr.sin6_addr)) 1769 iffound = 1; 1770 1771 /* 1772 * IPv4-mapped addresses can only be returned by a 1773 * Node Information proxy, since they represent 1774 * addresses of IPv4-only nodes, which perforce do 1775 * not implement this protocol. 1776 * [icmp-name-lookups-07, Section 5.4] 1777 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1778 * this function at this moment. 1779 */ 1780 1781 /* What do we have to do about ::1? */ 1782 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1783 case IPV6_ADDR_SCOPE_LINKLOCAL: 1784 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1785 continue; 1786 break; 1787 case IPV6_ADDR_SCOPE_SITELOCAL: 1788 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1789 continue; 1790 break; 1791 case IPV6_ADDR_SCOPE_GLOBAL: 1792 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1793 continue; 1794 break; 1795 default: 1796 continue; 1797 } 1798 1799 /* 1800 * check if anycast is okay. 1801 * XXX: just experimental. not in the spec. 1802 */ 1803 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1804 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1805 continue; /* we need only unicast addresses */ 1806 1807 addrsofif++; /* count the address */ 1808 } 1809 if (iffound) { 1810 *ifpp = ifp; 1811 return(addrsofif); 1812 } 1813 1814 addrs += addrsofif; 1815 } 1816 1817 return(addrs); 1818 } 1819 1820 static int 1821 ni6_store_addrs(ni6, nni6, ifp0, resid) 1822 struct icmp6_nodeinfo *ni6, *nni6; 1823 struct ifnet *ifp0; 1824 int resid; 1825 { 1826 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1827 struct in6_ifaddr *ifa6; 1828 struct ifaddr *ifa; 1829 struct ifnet *ifp_dep = NULL; 1830 int copied = 0, allow_deprecated = 0; 1831 u_char *cp = (u_char *)(nni6 + 1); 1832 int niflags = ni6->ni_flags; 1833 u_int32_t ltime; 1834 long time_second = time.tv_sec; 1835 1836 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1837 return(0); /* needless to copy */ 1838 1839 again: 1840 1841 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1842 { 1843 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1844 ifa = ifa->ifa_list.tqe_next) 1845 { 1846 if (ifa->ifa_addr->sa_family != AF_INET6) 1847 continue; 1848 ifa6 = (struct in6_ifaddr *)ifa; 1849 1850 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1851 allow_deprecated == 0) { 1852 /* 1853 * prefererred address should be put before 1854 * deprecated addresses. 1855 */ 1856 1857 /* record the interface for later search */ 1858 if (ifp_dep == NULL) 1859 ifp_dep = ifp; 1860 1861 continue; 1862 } 1863 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1864 allow_deprecated != 0) 1865 continue; /* we now collect deprecated addrs */ 1866 1867 /* What do we have to do about ::1? */ 1868 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1869 case IPV6_ADDR_SCOPE_LINKLOCAL: 1870 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1871 continue; 1872 break; 1873 case IPV6_ADDR_SCOPE_SITELOCAL: 1874 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1875 continue; 1876 break; 1877 case IPV6_ADDR_SCOPE_GLOBAL: 1878 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1879 continue; 1880 break; 1881 default: 1882 continue; 1883 } 1884 1885 /* 1886 * check if anycast is okay. 1887 * XXX: just experimental. not in the spec. 1888 */ 1889 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1890 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1891 continue; 1892 1893 /* now we can copy the address */ 1894 if (resid < sizeof(struct in6_addr) + 1895 sizeof(u_int32_t)) { 1896 /* 1897 * We give up much more copy. 1898 * Set the truncate flag and return. 1899 */ 1900 nni6->ni_flags |= 1901 NI_NODEADDR_FLAG_TRUNCATE; 1902 return(copied); 1903 } 1904 1905 /* 1906 * Set the TTL of the address. 1907 * The TTL value should be one of the following 1908 * according to the specification: 1909 * 1910 * 1. The remaining lifetime of a DHCP lease on the 1911 * address, or 1912 * 2. The remaining Valid Lifetime of a prefix from 1913 * which the address was derived through Stateless 1914 * Autoconfiguration. 1915 * 1916 * Note that we currently do not support stateful 1917 * address configuration by DHCPv6, so the former 1918 * case can't happen. 1919 * 1920 * TTL must be 2^31 > TTL >= 0. 1921 */ 1922 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1923 ltime = ND6_INFINITE_LIFETIME; 1924 else { 1925 if (ifa6->ia6_lifetime.ia6t_expire > 1926 time_second) 1927 ltime = ifa6->ia6_lifetime.ia6t_expire - time_second; 1928 else 1929 ltime = 0; 1930 } 1931 if (ltime > 0x7fffffff) 1932 ltime = 0x7fffffff; 1933 ltime = htonl(ltime); 1934 1935 bcopy(<ime, cp, sizeof(u_int32_t)); 1936 cp += sizeof(u_int32_t); 1937 1938 /* copy the address itself */ 1939 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1940 sizeof(struct in6_addr)); 1941 /* XXX: KAME link-local hack; remove ifindex */ 1942 if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr)) 1943 ((struct in6_addr *)cp)->s6_addr16[1] = 0; 1944 cp += sizeof(struct in6_addr); 1945 1946 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1947 copied += (sizeof(struct in6_addr) + 1948 sizeof(u_int32_t)); 1949 } 1950 if (ifp0) /* we need search only on the specified IF */ 1951 break; 1952 } 1953 1954 if (allow_deprecated == 0 && ifp_dep != NULL) { 1955 ifp = ifp_dep; 1956 allow_deprecated = 1; 1957 1958 goto again; 1959 } 1960 1961 return(copied); 1962 } 1963 1964 /* 1965 * XXX almost dup'ed code with rip6_input. 1966 */ 1967 static int 1968 icmp6_rip6_input(mp, off) 1969 struct mbuf **mp; 1970 int off; 1971 { 1972 struct mbuf *m = *mp; 1973 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1974 struct in6pcb *in6p; 1975 struct in6pcb *last = NULL; 1976 struct sockaddr_in6 rip6src; 1977 struct icmp6_hdr *icmp6; 1978 struct mbuf *opts = NULL; 1979 1980 #ifndef PULLDOWN_TEST 1981 /* this is assumed to be safe. */ 1982 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 1983 #else 1984 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1985 if (icmp6 == NULL) { 1986 /* m is already reclaimed */ 1987 return IPPROTO_DONE; 1988 } 1989 #endif 1990 1991 bzero(&rip6src, sizeof(rip6src)); 1992 rip6src.sin6_len = sizeof(struct sockaddr_in6); 1993 rip6src.sin6_family = AF_INET6; 1994 /* KAME hack: recover scopeid */ 1995 (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif); 1996 1997 for (in6p = rawin6pcb.in6p_next; 1998 in6p != &rawin6pcb; in6p = in6p->in6p_next) 1999 { 2000 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6) 2001 continue; 2002 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 2003 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 2004 continue; 2005 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 2006 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 2007 continue; 2008 if (in6p->in6p_icmp6filt 2009 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 2010 in6p->in6p_icmp6filt)) 2011 continue; 2012 if (last) { 2013 struct mbuf *n; 2014 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 2015 if (last->in6p_flags & IN6P_CONTROLOPTS) 2016 ip6_savecontrol(last, &opts, ip6, n); 2017 /* strip intermediate headers */ 2018 m_adj(n, off); 2019 if (sbappendaddr(&last->in6p_socket->so_rcv, 2020 (struct sockaddr *)&rip6src, 2021 n, opts) == 0) { 2022 /* should notify about lost packet */ 2023 m_freem(n); 2024 if (opts) 2025 m_freem(opts); 2026 } else 2027 sorwakeup(last->in6p_socket); 2028 opts = NULL; 2029 } 2030 } 2031 last = in6p; 2032 } 2033 if (last) { 2034 if (last->in6p_flags & IN6P_CONTROLOPTS) 2035 ip6_savecontrol(last, &opts, ip6, m); 2036 /* strip intermediate headers */ 2037 m_adj(m, off); 2038 if (sbappendaddr(&last->in6p_socket->so_rcv, 2039 (struct sockaddr *)&rip6src, m, opts) == 0) { 2040 m_freem(m); 2041 if (opts) 2042 m_freem(opts); 2043 } else 2044 sorwakeup(last->in6p_socket); 2045 } else { 2046 m_freem(m); 2047 ip6stat.ip6s_delivered--; 2048 } 2049 return IPPROTO_DONE; 2050 } 2051 2052 /* 2053 * Reflect the ip6 packet back to the source. 2054 * OFF points to the icmp6 header, counted from the top of the mbuf. 2055 * 2056 * Note: RFC 1885 required that an echo reply should be truncated if it 2057 * did not fit in with (return) path MTU, and KAME code supported the 2058 * behavior. However, as a clarification after the RFC, this limitation 2059 * was removed in a revised version of the spec, RFC 2463. We had kept the 2060 * old behavior, with a (non-default) ifdef block, while the new version of 2061 * the spec was an internet-draft status, and even after the new RFC was 2062 * published. But it would rather make sense to clean the obsoleted part 2063 * up, and to make the code simpler at this stage. 2064 */ 2065 void 2066 icmp6_reflect(m, off) 2067 struct mbuf *m; 2068 size_t off; 2069 { 2070 struct ip6_hdr *ip6; 2071 struct icmp6_hdr *icmp6; 2072 struct in6_ifaddr *ia; 2073 struct in6_addr t, *src = 0; 2074 int plen; 2075 int type, code; 2076 struct ifnet *outif = NULL; 2077 struct sockaddr_in6 sa6_src, sa6_dst; 2078 2079 /* too short to reflect */ 2080 if (off < sizeof(struct ip6_hdr)) { 2081 nd6log((LOG_DEBUG, 2082 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 2083 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2084 __FILE__, __LINE__)); 2085 goto bad; 2086 } 2087 2088 /* 2089 * If there are extra headers between IPv6 and ICMPv6, strip 2090 * off that header first. 2091 */ 2092 #ifdef DIAGNOSTIC 2093 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2094 panic("assumption failed in icmp6_reflect"); 2095 #endif 2096 if (off > sizeof(struct ip6_hdr)) { 2097 size_t l; 2098 struct ip6_hdr nip6; 2099 2100 l = off - sizeof(struct ip6_hdr); 2101 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2102 m_adj(m, l); 2103 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2104 if (m->m_len < l) { 2105 if ((m = m_pullup(m, l)) == NULL) 2106 return; 2107 } 2108 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2109 } else /* off == sizeof(struct ip6_hdr) */ { 2110 size_t l; 2111 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2112 if (m->m_len < l) { 2113 if ((m = m_pullup(m, l)) == NULL) 2114 return; 2115 } 2116 } 2117 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2118 ip6 = mtod(m, struct ip6_hdr *); 2119 ip6->ip6_nxt = IPPROTO_ICMPV6; 2120 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2121 type = icmp6->icmp6_type; /* keep type for statistics */ 2122 code = icmp6->icmp6_code; /* ditto. */ 2123 2124 t = ip6->ip6_dst; 2125 /* 2126 * ip6_input() drops a packet if its src is multicast. 2127 * So, the src is never multicast. 2128 */ 2129 ip6->ip6_dst = ip6->ip6_src; 2130 2131 /* 2132 * XXX: make sure to embed scope zone information, using 2133 * already embedded IDs or the received interface (if any). 2134 * Note that rcvif may be NULL. 2135 * TODO: scoped routing case (XXX). 2136 */ 2137 bzero(&sa6_src, sizeof(sa6_src)); 2138 sa6_src.sin6_family = AF_INET6; 2139 sa6_src.sin6_len = sizeof(sa6_src); 2140 sa6_src.sin6_addr = ip6->ip6_dst; 2141 in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif); 2142 in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL); 2143 ip6->ip6_dst = sa6_src.sin6_addr; 2144 2145 bzero(&sa6_dst, sizeof(sa6_dst)); 2146 sa6_dst.sin6_family = AF_INET6; 2147 sa6_dst.sin6_len = sizeof(sa6_dst); 2148 sa6_dst.sin6_addr = t; 2149 in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif); 2150 in6_embedscope(&t, &sa6_dst, NULL, NULL); 2151 2152 /* 2153 * If the incoming packet was addressed directly to us (i.e. unicast), 2154 * use dst as the src for the reply. 2155 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible 2156 * (for example) when we encounter an error while forwarding procedure 2157 * destined to a duplicated address of ours. 2158 */ 2159 for (ia = in6_ifaddr; ia; ia = ia->ia_next) 2160 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) && 2161 (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) { 2162 src = &t; 2163 break; 2164 } 2165 if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) { 2166 /* 2167 * This is the case if the dst is our link-local address 2168 * and the sender is also ourselves. 2169 */ 2170 src = &t; 2171 } 2172 2173 if (src == 0) { 2174 int e; 2175 struct route_in6 ro; 2176 2177 /* 2178 * This case matches to multicasts, our anycast, or unicasts 2179 * that we do not own. Select a source address based on the 2180 * source address of the erroneous packet. 2181 */ 2182 bzero(&ro, sizeof(ro)); 2183 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e); 2184 if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */ 2185 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2186 } 2187 if (src == NULL) { 2188 nd6log((LOG_DEBUG, 2189 "icmp6_reflect: source can't be determined: " 2190 "dst=%s, error=%d\n", 2191 ip6_sprintf(&sa6_src.sin6_addr), e)); 2192 goto bad; 2193 } 2194 } 2195 2196 ip6->ip6_src = *src; 2197 2198 ip6->ip6_flow = 0; 2199 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2200 ip6->ip6_vfc |= IPV6_VERSION; 2201 ip6->ip6_nxt = IPPROTO_ICMPV6; 2202 if (m->m_pkthdr.rcvif) { 2203 /* XXX: This may not be the outgoing interface */ 2204 ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim; 2205 } else 2206 ip6->ip6_hlim = ip6_defhlim; 2207 2208 icmp6->icmp6_cksum = 0; 2209 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2210 sizeof(struct ip6_hdr), plen); 2211 2212 /* 2213 * XXX option handling 2214 */ 2215 2216 m->m_flags &= ~(M_BCAST|M_MCAST); 2217 #ifdef IPSEC 2218 /* Don't lookup socket */ 2219 (void)ipsec_setsocket(m, NULL); 2220 #endif /* IPSEC */ 2221 2222 /* 2223 * To avoid a "too big" situation at an intermediate router 2224 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2225 * Note that only echo and node information replies are affected, 2226 * since the length of ICMP6 errors is limited to the minimum MTU. 2227 */ 2228 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, &outif) != 0 && outif) 2229 icmp6_ifstat_inc(outif, ifs6_out_error); 2230 2231 if (outif) 2232 icmp6_ifoutstat_inc(outif, type, code); 2233 2234 return; 2235 2236 bad: 2237 m_freem(m); 2238 return; 2239 } 2240 2241 void 2242 icmp6_fasttimo() 2243 { 2244 2245 mld6_fasttimeo(); 2246 } 2247 2248 static const char * 2249 icmp6_redirect_diag(src6, dst6, tgt6) 2250 struct in6_addr *src6; 2251 struct in6_addr *dst6; 2252 struct in6_addr *tgt6; 2253 { 2254 static char buf[1024]; 2255 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2256 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2257 return buf; 2258 } 2259 2260 void 2261 icmp6_redirect_input(m, off) 2262 struct mbuf *m; 2263 int off; 2264 { 2265 struct ifnet *ifp = m->m_pkthdr.rcvif; 2266 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2267 struct nd_redirect *nd_rd; 2268 int icmp6len = ntohs(ip6->ip6_plen); 2269 char *lladdr = NULL; 2270 int lladdrlen = 0; 2271 u_char *redirhdr = NULL; 2272 int redirhdrlen = 0; 2273 struct rtentry *rt = NULL; 2274 int is_router; 2275 int is_onlink; 2276 struct in6_addr src6 = ip6->ip6_src; 2277 struct in6_addr redtgt6; 2278 struct in6_addr reddst6; 2279 union nd_opts ndopts; 2280 2281 if (!m || !ifp) 2282 return; 2283 2284 /* XXX if we are router, we don't update route by icmp6 redirect */ 2285 if (ip6_forwarding) 2286 goto freeit; 2287 if (!icmp6_rediraccept) 2288 goto freeit; 2289 2290 #ifndef PULLDOWN_TEST 2291 IP6_EXTHDR_CHECK(m, off, icmp6len,); 2292 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off); 2293 #else 2294 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2295 if (nd_rd == NULL) { 2296 icmp6stat.icp6s_tooshort++; 2297 return; 2298 } 2299 #endif 2300 redtgt6 = nd_rd->nd_rd_target; 2301 reddst6 = nd_rd->nd_rd_dst; 2302 2303 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2304 redtgt6.s6_addr16[1] = htons(ifp->if_index); 2305 if (IN6_IS_ADDR_LINKLOCAL(&reddst6)) 2306 reddst6.s6_addr16[1] = htons(ifp->if_index); 2307 2308 /* validation */ 2309 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2310 nd6log((LOG_ERR, 2311 "ICMP6 redirect sent from %s rejected; " 2312 "must be from linklocal\n", ip6_sprintf(&src6))); 2313 goto bad; 2314 } 2315 if (ip6->ip6_hlim != 255) { 2316 nd6log((LOG_ERR, 2317 "ICMP6 redirect sent from %s rejected; " 2318 "hlim=%d (must be 255)\n", 2319 ip6_sprintf(&src6), ip6->ip6_hlim)); 2320 goto bad; 2321 } 2322 { 2323 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2324 struct sockaddr_in6 sin6; 2325 struct in6_addr *gw6; 2326 2327 bzero(&sin6, sizeof(sin6)); 2328 sin6.sin6_family = AF_INET6; 2329 sin6.sin6_len = sizeof(struct sockaddr_in6); 2330 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2331 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2332 if (rt) { 2333 if (rt->rt_gateway == NULL || 2334 rt->rt_gateway->sa_family != AF_INET6) { 2335 nd6log((LOG_ERR, 2336 "ICMP6 redirect rejected; no route " 2337 "with inet6 gateway found for redirect dst: %s\n", 2338 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2339 RTFREE(rt); 2340 goto bad; 2341 } 2342 2343 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2344 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2345 nd6log((LOG_ERR, 2346 "ICMP6 redirect rejected; " 2347 "not equal to gw-for-src=%s (must be same): " 2348 "%s\n", 2349 ip6_sprintf(gw6), 2350 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2351 RTFREE(rt); 2352 goto bad; 2353 } 2354 } else { 2355 nd6log((LOG_ERR, 2356 "ICMP6 redirect rejected; " 2357 "no route found for redirect dst: %s\n", 2358 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2359 goto bad; 2360 } 2361 RTFREE(rt); 2362 rt = NULL; 2363 } 2364 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2365 nd6log((LOG_ERR, 2366 "ICMP6 redirect rejected; " 2367 "redirect dst must be unicast: %s\n", 2368 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2369 goto bad; 2370 } 2371 2372 is_router = is_onlink = 0; 2373 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2374 is_router = 1; /* router case */ 2375 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2376 is_onlink = 1; /* on-link destination case */ 2377 if (!is_router && !is_onlink) { 2378 nd6log((LOG_ERR, 2379 "ICMP6 redirect rejected; " 2380 "neither router case nor onlink case: %s\n", 2381 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2382 goto bad; 2383 } 2384 /* validation passed */ 2385 2386 icmp6len -= sizeof(*nd_rd); 2387 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2388 if (nd6_options(&ndopts) < 0) { 2389 nd6log((LOG_INFO, "icmp6_redirect_input: " 2390 "invalid ND option, rejected: %s\n", 2391 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2392 /* nd6_options have incremented stats */ 2393 goto freeit; 2394 } 2395 2396 if (ndopts.nd_opts_tgt_lladdr) { 2397 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2398 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2399 } 2400 2401 if (ndopts.nd_opts_rh) { 2402 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len; 2403 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */ 2404 } 2405 2406 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2407 nd6log((LOG_INFO, 2408 "icmp6_redirect_input: lladdrlen mismatch for %s " 2409 "(if %d, icmp6 packet %d): %s\n", 2410 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2411 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2412 goto bad; 2413 } 2414 2415 /* RFC 2461 8.3 */ 2416 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2417 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2418 2419 if (!is_onlink) { /* better router case. perform rtredirect. */ 2420 /* perform rtredirect */ 2421 struct sockaddr_in6 sdst; 2422 struct sockaddr_in6 sgw; 2423 struct sockaddr_in6 ssrc; 2424 unsigned long rtcount; 2425 struct rtentry *newrt = NULL; 2426 2427 /* 2428 * do not install redirect route, if the number of entries 2429 * is too much (> hiwat). note that, the node (= host) will 2430 * work just fine even if we do not install redirect route 2431 * (there will be additional hops, though). 2432 */ 2433 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2434 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2435 return; 2436 else if (0 <= icmp6_redirect_lowat && 2437 rtcount > icmp6_redirect_lowat) { 2438 /* 2439 * XXX nuke a victim, install the new one. 2440 */ 2441 } 2442 2443 bzero(&sdst, sizeof(sdst)); 2444 bzero(&sgw, sizeof(sgw)); 2445 bzero(&ssrc, sizeof(ssrc)); 2446 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2447 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2448 sizeof(struct sockaddr_in6); 2449 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2450 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2451 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2452 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2453 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2454 (struct sockaddr *)&ssrc, 2455 &newrt); 2456 2457 if (newrt) { 2458 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2459 icmp6_redirect_timeout_q); 2460 rtfree(newrt); 2461 } 2462 } 2463 /* finally update cached route in each socket via pfctlinput */ 2464 { 2465 struct sockaddr_in6 sdst; 2466 2467 bzero(&sdst, sizeof(sdst)); 2468 sdst.sin6_family = AF_INET6; 2469 sdst.sin6_len = sizeof(struct sockaddr_in6); 2470 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2471 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2472 #ifdef IPSEC 2473 key_sa_routechange((struct sockaddr *)&sdst); 2474 #endif 2475 } 2476 2477 freeit: 2478 m_freem(m); 2479 return; 2480 2481 bad: 2482 icmp6stat.icp6s_badredirect++; 2483 m_freem(m); 2484 } 2485 2486 void 2487 icmp6_redirect_output(m0, rt) 2488 struct mbuf *m0; 2489 struct rtentry *rt; 2490 { 2491 struct ifnet *ifp; /* my outgoing interface */ 2492 struct in6_addr *ifp_ll6; 2493 struct in6_addr *nexthop; 2494 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2495 struct mbuf *m = NULL; /* newly allocated one */ 2496 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2497 struct nd_redirect *nd_rd; 2498 size_t maxlen; 2499 u_char *p; 2500 struct sockaddr_in6 src_sa; 2501 2502 icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2503 2504 /* if we are not router, we don't send icmp6 redirect */ 2505 if (!ip6_forwarding || ip6_accept_rtadv) 2506 goto fail; 2507 2508 /* sanity check */ 2509 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2510 goto fail; 2511 2512 /* 2513 * Address check: 2514 * the source address must identify a neighbor, and 2515 * the destination address must not be a multicast address 2516 * [RFC 2461, sec 8.2] 2517 */ 2518 sip6 = mtod(m0, struct ip6_hdr *); 2519 bzero(&src_sa, sizeof(src_sa)); 2520 src_sa.sin6_family = AF_INET6; 2521 src_sa.sin6_len = sizeof(src_sa); 2522 src_sa.sin6_addr = sip6->ip6_src; 2523 /* we don't currently use sin6_scope_id, but eventually use it */ 2524 src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src); 2525 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2526 goto fail; 2527 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2528 goto fail; /* what should we do here? */ 2529 2530 /* rate limit */ 2531 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2532 goto fail; 2533 2534 /* 2535 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2536 * we almost always ask for an mbuf cluster for simplicity. 2537 * (MHLEN < IPV6_MMTU is almost always true) 2538 */ 2539 #if IPV6_MMTU >= MCLBYTES 2540 # error assumption failed about IPV6_MMTU and MCLBYTES 2541 #endif 2542 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2543 if (m && IPV6_MMTU >= MHLEN) 2544 MCLGET(m, M_DONTWAIT); 2545 if (!m) 2546 goto fail; 2547 m->m_pkthdr.rcvif = NULL; 2548 m->m_len = 0; 2549 maxlen = M_TRAILINGSPACE(m); 2550 maxlen = min(IPV6_MMTU, maxlen); 2551 /* just for safety */ 2552 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2553 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2554 goto fail; 2555 } 2556 2557 { 2558 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2559 struct in6_ifaddr *ia; 2560 if ((ia = in6ifa_ifpforlinklocal(ifp, 2561 IN6_IFF_NOTREADY| 2562 IN6_IFF_ANYCAST)) == NULL) 2563 goto fail; 2564 ifp_ll6 = &ia->ia_addr.sin6_addr; 2565 } 2566 2567 /* get ip6 linklocal address for the router. */ 2568 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2569 struct sockaddr_in6 *sin6; 2570 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2571 nexthop = &sin6->sin6_addr; 2572 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2573 nexthop = NULL; 2574 } else 2575 nexthop = NULL; 2576 2577 /* ip6 */ 2578 ip6 = mtod(m, struct ip6_hdr *); 2579 ip6->ip6_flow = 0; 2580 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2581 ip6->ip6_vfc |= IPV6_VERSION; 2582 /* ip6->ip6_plen will be set later */ 2583 ip6->ip6_nxt = IPPROTO_ICMPV6; 2584 ip6->ip6_hlim = 255; 2585 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2586 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2587 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2588 2589 /* ND Redirect */ 2590 nd_rd = (struct nd_redirect *)(ip6 + 1); 2591 nd_rd->nd_rd_type = ND_REDIRECT; 2592 nd_rd->nd_rd_code = 0; 2593 nd_rd->nd_rd_reserved = 0; 2594 if (rt->rt_flags & RTF_GATEWAY) { 2595 /* 2596 * nd_rd->nd_rd_target must be a link-local address in 2597 * better router cases. 2598 */ 2599 if (!nexthop) 2600 goto fail; 2601 bcopy(nexthop, &nd_rd->nd_rd_target, 2602 sizeof(nd_rd->nd_rd_target)); 2603 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2604 sizeof(nd_rd->nd_rd_dst)); 2605 } else { 2606 /* make sure redtgt == reddst */ 2607 nexthop = &sip6->ip6_dst; 2608 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2609 sizeof(nd_rd->nd_rd_target)); 2610 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2611 sizeof(nd_rd->nd_rd_dst)); 2612 } 2613 2614 p = (u_char *)(nd_rd + 1); 2615 2616 { 2617 /* target lladdr option */ 2618 struct rtentry *rt_nexthop = NULL; 2619 int len; 2620 struct sockaddr_dl *sdl; 2621 struct nd_opt_hdr *nd_opt; 2622 char *lladdr; 2623 2624 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2625 if (!rt_nexthop) 2626 goto nolladdropt; 2627 len = sizeof(*nd_opt) + ifp->if_addrlen; 2628 len = (len + 7) & ~7; /* round by 8 */ 2629 /* safety check */ 2630 if (len + (p - (u_char *)ip6) > maxlen) 2631 goto nolladdropt; 2632 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2633 (rt_nexthop->rt_flags & RTF_LLINFO) && 2634 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2635 (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) && 2636 sdl->sdl_alen) { 2637 nd_opt = (struct nd_opt_hdr *)p; 2638 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2639 nd_opt->nd_opt_len = len >> 3; 2640 lladdr = (char *)(nd_opt + 1); 2641 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen); 2642 p += len; 2643 } 2644 } 2645 nolladdropt:; 2646 2647 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2648 2649 /* just to be safe */ 2650 if (m0->m_flags & M_DECRYPTED) 2651 goto noredhdropt; 2652 if (p - (u_char *)ip6 > maxlen) 2653 goto noredhdropt; 2654 2655 { 2656 /* redirected header option */ 2657 int len; 2658 struct nd_opt_rd_hdr *nd_opt_rh; 2659 2660 /* 2661 * compute the maximum size for icmp6 redirect header option. 2662 * XXX room for auth header? 2663 */ 2664 len = maxlen - (p - (u_char *)ip6); 2665 len &= ~7; 2666 2667 /* This is just for simplicity. */ 2668 if (m0->m_pkthdr.len != m0->m_len) { 2669 if (m0->m_next) { 2670 m_freem(m0->m_next); 2671 m0->m_next = NULL; 2672 } 2673 m0->m_pkthdr.len = m0->m_len; 2674 } 2675 2676 /* 2677 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2678 * about padding/truncate rule for the original IP packet. 2679 * From the discussion on IPv6imp in Feb 1999, the consensus was: 2680 * - "attach as much as possible" is the goal 2681 * - pad if not aligned (original size can be guessed by original 2682 * ip6 header) 2683 * Following code adds the padding if it is simple enough, 2684 * and truncates if not. 2685 */ 2686 if (m0->m_next || m0->m_pkthdr.len != m0->m_len) 2687 panic("assumption failed in %s:%d\n", __FILE__, __LINE__); 2688 2689 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2690 /* not enough room, truncate */ 2691 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh); 2692 } else { 2693 /* enough room, pad or truncate */ 2694 size_t extra; 2695 2696 extra = m0->m_pkthdr.len % 8; 2697 if (extra) { 2698 /* pad if easy enough, truncate if not */ 2699 if (8 - extra <= M_TRAILINGSPACE(m0)) { 2700 /* pad */ 2701 m0->m_len += (8 - extra); 2702 m0->m_pkthdr.len += (8 - extra); 2703 } else { 2704 /* truncate */ 2705 m0->m_pkthdr.len -= extra; 2706 m0->m_len -= extra; 2707 } 2708 } 2709 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2710 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh); 2711 } 2712 2713 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2714 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2715 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2716 nd_opt_rh->nd_opt_rh_len = len >> 3; 2717 p += sizeof(*nd_opt_rh); 2718 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2719 2720 /* connect m0 to m */ 2721 m->m_next = m0; 2722 m->m_pkthdr.len = m->m_len + m0->m_len; 2723 } 2724 noredhdropt:; 2725 2726 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src)) 2727 sip6->ip6_src.s6_addr16[1] = 0; 2728 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst)) 2729 sip6->ip6_dst.s6_addr16[1] = 0; 2730 #if 0 2731 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) 2732 ip6->ip6_src.s6_addr16[1] = 0; 2733 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) 2734 ip6->ip6_dst.s6_addr16[1] = 0; 2735 #endif 2736 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target)) 2737 nd_rd->nd_rd_target.s6_addr16[1] = 0; 2738 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst)) 2739 nd_rd->nd_rd_dst.s6_addr16[1] = 0; 2740 2741 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2742 2743 nd_rd->nd_rd_cksum = 0; 2744 nd_rd->nd_rd_cksum 2745 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2746 2747 /* send the packet to outside... */ 2748 #ifdef IPSEC 2749 /* Don't lookup socket */ 2750 (void)ipsec_setsocket(m, NULL); 2751 #endif /* IPSEC */ 2752 if (ip6_output(m, NULL, NULL, 0, NULL, NULL) != 0) 2753 icmp6_ifstat_inc(ifp, ifs6_out_error); 2754 2755 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2756 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2757 icmp6stat.icp6s_outhist[ND_REDIRECT]++; 2758 2759 return; 2760 2761 fail: 2762 if (m) 2763 m_freem(m); 2764 if (m0) 2765 m_freem(m0); 2766 } 2767 2768 /* 2769 * ICMPv6 socket option processing. 2770 */ 2771 int 2772 icmp6_ctloutput(op, so, level, optname, mp) 2773 int op; 2774 struct socket *so; 2775 int level, optname; 2776 struct mbuf **mp; 2777 { 2778 int error = 0; 2779 int optlen; 2780 struct in6pcb *in6p = sotoin6pcb(so); 2781 struct mbuf *m = *mp; 2782 2783 optlen = m ? m->m_len : 0; 2784 2785 if (level != IPPROTO_ICMPV6) { 2786 if (op == PRCO_SETOPT && m) 2787 (void)m_free(m); 2788 return EINVAL; 2789 } 2790 2791 switch (op) { 2792 case PRCO_SETOPT: 2793 switch (optname) { 2794 case ICMP6_FILTER: 2795 { 2796 struct icmp6_filter *p; 2797 2798 if (optlen != sizeof(*p)) { 2799 error = EMSGSIZE; 2800 break; 2801 } 2802 p = mtod(m, struct icmp6_filter *); 2803 if (!p || !in6p->in6p_icmp6filt) { 2804 error = EINVAL; 2805 break; 2806 } 2807 bcopy(p, in6p->in6p_icmp6filt, 2808 sizeof(struct icmp6_filter)); 2809 error = 0; 2810 break; 2811 } 2812 2813 default: 2814 error = ENOPROTOOPT; 2815 break; 2816 } 2817 if (m) 2818 (void)m_freem(m); 2819 break; 2820 2821 case PRCO_GETOPT: 2822 switch (optname) { 2823 case ICMP6_FILTER: 2824 { 2825 struct icmp6_filter *p; 2826 2827 if (!in6p->in6p_icmp6filt) { 2828 error = EINVAL; 2829 break; 2830 } 2831 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2832 m->m_len = sizeof(struct icmp6_filter); 2833 p = mtod(m, struct icmp6_filter *); 2834 bcopy(in6p->in6p_icmp6filt, p, 2835 sizeof(struct icmp6_filter)); 2836 error = 0; 2837 break; 2838 } 2839 2840 default: 2841 error = ENOPROTOOPT; 2842 break; 2843 } 2844 break; 2845 } 2846 2847 return(error); 2848 } 2849 2850 /* 2851 * Perform rate limit check. 2852 * Returns 0 if it is okay to send the icmp6 packet. 2853 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2854 * limitation. 2855 * 2856 * XXX per-destination/type check necessary? 2857 */ 2858 static int 2859 icmp6_ratelimit(dst, type, code) 2860 const struct in6_addr *dst; /* not used at this moment */ 2861 const int type; /* not used at this moment */ 2862 const int code; /* not used at this moment */ 2863 { 2864 int ret; 2865 2866 ret = 0; /* okay to send */ 2867 2868 /* PPS limit */ 2869 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2870 icmp6errppslim)) { 2871 /* The packet is subject to rate limit */ 2872 ret++; 2873 } 2874 2875 return ret; 2876 } 2877 2878 static struct rtentry * 2879 icmp6_mtudisc_clone(dst) 2880 struct sockaddr *dst; 2881 { 2882 struct rtentry *rt; 2883 int error; 2884 2885 rt = rtalloc1(dst, 1); 2886 if (rt == 0) 2887 return NULL; 2888 2889 /* If we didn't get a host route, allocate one */ 2890 if ((rt->rt_flags & RTF_HOST) == 0) { 2891 struct rtentry *nrt; 2892 2893 error = rtrequest((int) RTM_ADD, dst, 2894 (struct sockaddr *) rt->rt_gateway, 2895 (struct sockaddr *) 0, 2896 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2897 if (error) { 2898 rtfree(rt); 2899 return NULL; 2900 } 2901 nrt->rt_rmx = rt->rt_rmx; 2902 rtfree(rt); 2903 rt = nrt; 2904 } 2905 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2906 icmp6_mtudisc_timeout_q); 2907 if (error) { 2908 rtfree(rt); 2909 return NULL; 2910 } 2911 2912 return rt; /* caller need to call rtfree() */ 2913 } 2914 2915 static void 2916 icmp6_mtudisc_timeout(rt, r) 2917 struct rtentry *rt; 2918 struct rttimer *r; 2919 { 2920 if (rt == NULL) 2921 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2922 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2923 (RTF_DYNAMIC | RTF_HOST)) { 2924 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2925 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2926 } else { 2927 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 2928 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu; 2929 } 2930 } 2931 2932 static void 2933 icmp6_redirect_timeout(rt, r) 2934 struct rtentry *rt; 2935 struct rttimer *r; 2936 { 2937 if (rt == NULL) 2938 panic("icmp6_redirect_timeout: bad route to timeout"); 2939 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2940 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2941 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2942 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2943 } 2944 } 2945 2946 int 2947 icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 2948 int *name; 2949 u_int namelen; 2950 void *oldp; 2951 size_t *oldlenp; 2952 void *newp; 2953 size_t newlen; 2954 { 2955 2956 /* All sysctl names at this level are terminal. */ 2957 if (namelen != 1) 2958 return ENOTDIR; 2959 2960 switch (name[0]) { 2961 2962 case ICMPV6CTL_REDIRACCEPT: 2963 return sysctl_int(oldp, oldlenp, newp, newlen, 2964 &icmp6_rediraccept); 2965 case ICMPV6CTL_REDIRTIMEOUT: 2966 return sysctl_int(oldp, oldlenp, newp, newlen, 2967 &icmp6_redirtimeout); 2968 case ICMPV6CTL_STATS: 2969 return sysctl_rdstruct(oldp, oldlenp, newp, 2970 &icmp6stat, sizeof(icmp6stat)); 2971 case ICMPV6CTL_ND6_PRUNE: 2972 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune); 2973 case ICMPV6CTL_ND6_DELAY: 2974 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay); 2975 case ICMPV6CTL_ND6_UMAXTRIES: 2976 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries); 2977 case ICMPV6CTL_ND6_MMAXTRIES: 2978 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries); 2979 case ICMPV6CTL_ND6_USELOOPBACK: 2980 return sysctl_int(oldp, oldlenp, newp, newlen, 2981 &nd6_useloopback); 2982 case ICMPV6CTL_NODEINFO: 2983 return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo); 2984 case ICMPV6CTL_ERRPPSLIMIT: 2985 return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6errppslim); 2986 case ICMPV6CTL_ND6_MAXNUDHINT: 2987 return sysctl_int(oldp, oldlenp, newp, newlen, 2988 &nd6_maxnudhint); 2989 case ICMPV6CTL_MTUDISC_HIWAT: 2990 return sysctl_int(oldp, oldlenp, newp, newlen, 2991 &icmp6_mtudisc_hiwat); 2992 case ICMPV6CTL_MTUDISC_LOWAT: 2993 return sysctl_int(oldp, oldlenp, newp, newlen, 2994 &icmp6_mtudisc_lowat); 2995 case ICMPV6CTL_ND6_DEBUG: 2996 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_debug); 2997 default: 2998 return ENOPROTOOPT; 2999 } 3000 /* NOTREACHED */ 3001 } 3002