1 /* $NetBSD: icmp6.c,v 1.104 2003/12/17 18:49:38 lha 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 1.104 2003/12/17 18:49:38 lha Exp $"); 66 67 #include "opt_inet.h" 68 #include "opt_ipsec.h" 69 70 #include <sys/param.h> 71 #include <sys/systm.h> 72 #include <sys/malloc.h> 73 #include <sys/mbuf.h> 74 #include <sys/protosw.h> 75 #include <sys/socket.h> 76 #include <sys/socketvar.h> 77 #include <sys/time.h> 78 #include <sys/kernel.h> 79 #include <sys/syslog.h> 80 #include <sys/domain.h> 81 #include <sys/sysctl.h> 82 83 #include <net/if.h> 84 #include <net/route.h> 85 #include <net/if_dl.h> 86 #include <net/if_types.h> 87 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip6.h> 91 #include <netinet6/ip6_var.h> 92 #include <netinet/icmp6.h> 93 #include <netinet6/mld6_var.h> 94 #include <netinet6/in6_pcb.h> 95 #include <netinet6/nd6.h> 96 #include <netinet6/in6_ifattach.h> 97 #include <netinet6/ip6protosw.h> 98 99 #ifdef IPSEC 100 #include <netinet6/ipsec.h> 101 #include <netkey/key.h> 102 #endif 103 104 #include "faith.h" 105 #if defined(NFAITH) && 0 < NFAITH 106 #include <net/if_faith.h> 107 #endif 108 109 #include <net/net_osdep.h> 110 111 extern struct domain inet6domain; 112 113 struct icmp6stat icmp6stat; 114 115 extern struct inpcbtable raw6cbtable; 116 extern int icmp6errppslim; 117 static int icmp6errpps_count = 0; 118 static struct timeval icmp6errppslim_last; 119 extern int icmp6_nodeinfo; 120 121 /* 122 * List of callbacks to notify when Path MTU changes are made. 123 */ 124 struct icmp6_mtudisc_callback { 125 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 126 void (*mc_func) __P((struct in6_addr *)); 127 }; 128 129 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 130 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 131 132 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 133 extern int pmtu_expire; 134 135 /* XXX do these values make any sense? */ 136 static int icmp6_mtudisc_hiwat = 1280; 137 static int icmp6_mtudisc_lowat = 256; 138 139 /* 140 * keep track of # of redirect routes. 141 */ 142 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 143 144 /* XXX experimental, turned off */ 145 static int icmp6_redirect_hiwat = -1; 146 static int icmp6_redirect_lowat = -1; 147 148 static void icmp6_errcount __P((struct icmp6errstat *, int, int)); 149 static int icmp6_rip6_input __P((struct mbuf **, int)); 150 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int)); 151 static const char *icmp6_redirect_diag __P((struct in6_addr *, 152 struct in6_addr *, struct in6_addr *)); 153 static struct mbuf *ni6_input __P((struct mbuf *, int)); 154 static struct mbuf *ni6_nametodns __P((const char *, int, int)); 155 static int ni6_dnsmatch __P((const char *, int, const char *, int)); 156 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *, 157 struct ifnet **, char *)); 158 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 159 struct ifnet *, int)); 160 static int icmp6_notify_error __P((struct mbuf *, int, int, int)); 161 static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *)); 162 static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *)); 163 static void icmp6_redirect_timeout __P((struct rtentry *, struct rttimer *)); 164 165 void 166 icmp6_init() 167 { 168 mld6_init(); 169 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 170 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 171 } 172 173 static void 174 icmp6_errcount(stat, type, code) 175 struct icmp6errstat *stat; 176 int type, code; 177 { 178 switch (type) { 179 case ICMP6_DST_UNREACH: 180 switch (code) { 181 case ICMP6_DST_UNREACH_NOROUTE: 182 stat->icp6errs_dst_unreach_noroute++; 183 return; 184 case ICMP6_DST_UNREACH_ADMIN: 185 stat->icp6errs_dst_unreach_admin++; 186 return; 187 case ICMP6_DST_UNREACH_BEYONDSCOPE: 188 stat->icp6errs_dst_unreach_beyondscope++; 189 return; 190 case ICMP6_DST_UNREACH_ADDR: 191 stat->icp6errs_dst_unreach_addr++; 192 return; 193 case ICMP6_DST_UNREACH_NOPORT: 194 stat->icp6errs_dst_unreach_noport++; 195 return; 196 } 197 break; 198 case ICMP6_PACKET_TOO_BIG: 199 stat->icp6errs_packet_too_big++; 200 return; 201 case ICMP6_TIME_EXCEEDED: 202 switch (code) { 203 case ICMP6_TIME_EXCEED_TRANSIT: 204 stat->icp6errs_time_exceed_transit++; 205 return; 206 case ICMP6_TIME_EXCEED_REASSEMBLY: 207 stat->icp6errs_time_exceed_reassembly++; 208 return; 209 } 210 break; 211 case ICMP6_PARAM_PROB: 212 switch (code) { 213 case ICMP6_PARAMPROB_HEADER: 214 stat->icp6errs_paramprob_header++; 215 return; 216 case ICMP6_PARAMPROB_NEXTHEADER: 217 stat->icp6errs_paramprob_nextheader++; 218 return; 219 case ICMP6_PARAMPROB_OPTION: 220 stat->icp6errs_paramprob_option++; 221 return; 222 } 223 break; 224 case ND_REDIRECT: 225 stat->icp6errs_redirect++; 226 return; 227 } 228 stat->icp6errs_unknown++; 229 } 230 231 /* 232 * Register a Path MTU Discovery callback. 233 */ 234 void 235 icmp6_mtudisc_callback_register(func) 236 void (*func) __P((struct in6_addr *)); 237 { 238 struct icmp6_mtudisc_callback *mc; 239 240 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 241 mc = LIST_NEXT(mc, mc_list)) { 242 if (mc->mc_func == func) 243 return; 244 } 245 246 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 247 if (mc == NULL) 248 panic("icmp6_mtudisc_callback_register"); 249 250 mc->mc_func = func; 251 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 252 } 253 254 /* 255 * Generate an error packet of type error in response to bad IP6 packet. 256 */ 257 void 258 icmp6_error(m, type, code, param) 259 struct mbuf *m; 260 int type, code, param; 261 { 262 struct ip6_hdr *oip6, *nip6; 263 struct icmp6_hdr *icmp6; 264 u_int preplen; 265 int off; 266 int nxt; 267 268 icmp6stat.icp6s_error++; 269 270 /* count per-type-code statistics */ 271 icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code); 272 273 if (m->m_flags & M_DECRYPTED) { 274 icmp6stat.icp6s_canterror++; 275 goto freeit; 276 } 277 278 if (m->m_len < sizeof(struct ip6_hdr)) { 279 m = m_pullup(m, sizeof(struct ip6_hdr)); 280 if (m == NULL) 281 return; 282 } 283 oip6 = mtod(m, struct ip6_hdr *); 284 285 /* 286 * If the destination address of the erroneous packet is a multicast 287 * address, or the packet was sent using link-layer multicast, 288 * we should basically suppress sending an error (RFC 2463, Section 289 * 2.4). 290 * We have two exceptions (the item e.2 in that section): 291 * - the Pakcet Too Big message can be sent for path MTU discovery. 292 * - the Parameter Problem Message that can be allowed an icmp6 error 293 * in the option type field. This check has been done in 294 * ip6_unknown_opt(), so we can just check the type and code. 295 */ 296 if ((m->m_flags & (M_BCAST|M_MCAST) || 297 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 298 (type != ICMP6_PACKET_TOO_BIG && 299 (type != ICMP6_PARAM_PROB || 300 code != ICMP6_PARAMPROB_OPTION))) 301 goto freeit; 302 303 /* 304 * RFC 2463, 2.4 (e.5): source address check. 305 * XXX: the case of anycast source? 306 */ 307 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 308 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 309 goto freeit; 310 311 /* 312 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 313 * don't do it. 314 */ 315 nxt = -1; 316 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 317 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 318 struct icmp6_hdr *icp; 319 320 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 321 sizeof(*icp)); 322 if (icp == NULL) { 323 icmp6stat.icp6s_tooshort++; 324 return; 325 } 326 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 327 icp->icmp6_type == ND_REDIRECT) { 328 /* 329 * ICMPv6 error 330 * Special case: for redirect (which is 331 * informational) we must not send icmp6 error. 332 */ 333 icmp6stat.icp6s_canterror++; 334 goto freeit; 335 } else { 336 /* ICMPv6 informational - send the error */ 337 } 338 } 339 #if 0 /* controversial */ 340 else if (off >= 0 && nxt == IPPROTO_ESP) { 341 /* 342 * It could be ICMPv6 error inside ESP. Take a safer side, 343 * don't respond. 344 */ 345 icmp6stat.icp6s_canterror++; 346 goto freeit; 347 } 348 #endif 349 else { 350 /* non-ICMPv6 - send the error */ 351 } 352 353 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 354 355 /* Finally, do rate limitation check. */ 356 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 357 icmp6stat.icp6s_toofreq++; 358 goto freeit; 359 } 360 361 /* 362 * OK, ICMP6 can be generated. 363 */ 364 365 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 366 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 367 368 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 369 M_PREPEND(m, preplen, M_DONTWAIT); 370 if (m && m->m_len < preplen) 371 m = m_pullup(m, preplen); 372 if (m == NULL) { 373 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 374 return; 375 } 376 377 nip6 = mtod(m, struct ip6_hdr *); 378 nip6->ip6_src = oip6->ip6_src; 379 nip6->ip6_dst = oip6->ip6_dst; 380 381 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src)) 382 oip6->ip6_src.s6_addr16[1] = 0; 383 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst)) 384 oip6->ip6_dst.s6_addr16[1] = 0; 385 386 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 387 icmp6->icmp6_type = type; 388 icmp6->icmp6_code = code; 389 icmp6->icmp6_pptr = htonl((u_int32_t)param); 390 391 /* 392 * icmp6_reflect() is designed to be in the input path. 393 * icmp6_error() can be called from both input and outut path, 394 * and if we are in output path rcvif could contain bogus value. 395 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 396 * information in ip header (nip6). 397 */ 398 m->m_pkthdr.rcvif = NULL; 399 400 icmp6stat.icp6s_outhist[type]++; 401 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 402 403 return; 404 405 freeit: 406 /* 407 * If we can't tell wheter or not we can generate ICMP6, free it. 408 */ 409 m_freem(m); 410 } 411 412 /* 413 * Process a received ICMP6 message. 414 */ 415 int 416 icmp6_input(mp, offp, proto) 417 struct mbuf **mp; 418 int *offp, proto; 419 { 420 struct mbuf *m = *mp, *n; 421 struct ip6_hdr *ip6, *nip6; 422 struct icmp6_hdr *icmp6, *nicmp6; 423 int off = *offp; 424 int icmp6len = m->m_pkthdr.len - *offp; 425 int code, sum, noff; 426 427 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 428 429 /* 430 * Locate icmp6 structure in mbuf, and check 431 * that not corrupted and of at least minimum length 432 */ 433 434 ip6 = mtod(m, struct ip6_hdr *); 435 if (icmp6len < sizeof(struct icmp6_hdr)) { 436 icmp6stat.icp6s_tooshort++; 437 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 438 goto freeit; 439 } 440 441 /* 442 * calculate the checksum 443 */ 444 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 445 if (icmp6 == NULL) { 446 icmp6stat.icp6s_tooshort++; 447 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 448 return IPPROTO_DONE; 449 } 450 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 451 code = icmp6->icmp6_code; 452 453 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 454 nd6log((LOG_ERR, 455 "ICMP6 checksum error(%d|%x) %s\n", 456 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 457 icmp6stat.icp6s_checksum++; 458 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 459 goto freeit; 460 } 461 462 #if defined(NFAITH) && 0 < NFAITH 463 if (faithprefix(&ip6->ip6_dst)) { 464 /* 465 * Deliver very specific ICMP6 type only. 466 * This is important to deilver TOOBIG. Otherwise PMTUD 467 * will not work. 468 */ 469 switch (icmp6->icmp6_type) { 470 case ICMP6_DST_UNREACH: 471 case ICMP6_PACKET_TOO_BIG: 472 case ICMP6_TIME_EXCEEDED: 473 break; 474 default: 475 goto freeit; 476 } 477 } 478 #endif 479 480 icmp6stat.icp6s_inhist[icmp6->icmp6_type]++; 481 482 switch (icmp6->icmp6_type) { 483 case ICMP6_DST_UNREACH: 484 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 485 switch (code) { 486 case ICMP6_DST_UNREACH_NOROUTE: 487 code = PRC_UNREACH_NET; 488 break; 489 case ICMP6_DST_UNREACH_ADMIN: 490 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 491 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 492 break; 493 case ICMP6_DST_UNREACH_ADDR: 494 code = PRC_HOSTDEAD; 495 break; 496 #ifdef COMPAT_RFC1885 497 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 498 code = PRC_UNREACH_SRCFAIL; 499 break; 500 #else 501 case ICMP6_DST_UNREACH_BEYONDSCOPE: 502 /* I mean "source address was incorrect." */ 503 code = PRC_UNREACH_NET; 504 break; 505 #endif 506 case ICMP6_DST_UNREACH_NOPORT: 507 code = PRC_UNREACH_PORT; 508 break; 509 default: 510 goto badcode; 511 } 512 goto deliver; 513 514 case ICMP6_PACKET_TOO_BIG: 515 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 516 if (code != 0) 517 goto badcode; 518 519 code = PRC_MSGSIZE; 520 521 /* 522 * Updating the path MTU will be done after examining 523 * intermediate extension headers. 524 */ 525 goto deliver; 526 527 case ICMP6_TIME_EXCEEDED: 528 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 529 switch (code) { 530 case ICMP6_TIME_EXCEED_TRANSIT: 531 code = PRC_TIMXCEED_INTRANS; 532 break; 533 case ICMP6_TIME_EXCEED_REASSEMBLY: 534 code = PRC_TIMXCEED_REASS; 535 break; 536 default: 537 goto badcode; 538 } 539 goto deliver; 540 541 case ICMP6_PARAM_PROB: 542 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 543 switch (code) { 544 case ICMP6_PARAMPROB_NEXTHEADER: 545 code = PRC_UNREACH_PROTOCOL; 546 break; 547 case ICMP6_PARAMPROB_HEADER: 548 case ICMP6_PARAMPROB_OPTION: 549 code = PRC_PARAMPROB; 550 break; 551 default: 552 goto badcode; 553 } 554 goto deliver; 555 556 case ICMP6_ECHO_REQUEST: 557 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 558 if (code != 0) 559 goto badcode; 560 /* 561 * Copy mbuf to send to two data paths: userland socket(s), 562 * and to the querier (echo reply). 563 * m: a copy for socket, n: a copy for querier 564 */ 565 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 566 /* Give up local */ 567 n = m; 568 m = NULL; 569 goto deliverecho; 570 } 571 /* 572 * If the first mbuf is shared, or the first mbuf is too short, 573 * copy the first part of the data into a fresh mbuf. 574 * Otherwise, we will wrongly overwrite both copies. 575 */ 576 if ((n->m_flags & M_EXT) != 0 || 577 n->m_len < off + sizeof(struct icmp6_hdr)) { 578 struct mbuf *n0 = n; 579 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6); 580 581 /* 582 * Prepare an internal mbuf. m_pullup() doesn't 583 * always copy the length we specified. 584 */ 585 if (maxlen >= MCLBYTES) { 586 /* Give up remote */ 587 m_freem(n0); 588 break; 589 } 590 MGETHDR(n, M_DONTWAIT, n0->m_type); 591 if (n && maxlen >= MHLEN) { 592 MCLGET(n, M_DONTWAIT); 593 if ((n->m_flags & M_EXT) == 0) { 594 m_free(n); 595 n = NULL; 596 } 597 } 598 if (n == NULL) { 599 /* Give up local */ 600 m_freem(n0); 601 n = m; 602 m = NULL; 603 goto deliverecho; 604 } 605 M_COPY_PKTHDR(n, n0); 606 /* 607 * Copy IPv6 and ICMPv6 only. 608 */ 609 nip6 = mtod(n, struct ip6_hdr *); 610 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 611 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 612 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 613 noff = sizeof(struct ip6_hdr); 614 n->m_len = noff + sizeof(struct icmp6_hdr); 615 /* 616 * Adjust mbuf. ip6_plen will be adjusted in 617 * ip6_output(). 618 * n->m_pkthdr.len == n0->m_pkthdr.len at this point. 619 */ 620 n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr); 621 n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr)); 622 m_adj(n0, off + sizeof(struct icmp6_hdr)); 623 n->m_next = n0; 624 n0->m_flags &= ~M_PKTHDR; 625 } else { 626 deliverecho: 627 nip6 = mtod(n, struct ip6_hdr *); 628 nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off); 629 noff = off; 630 } 631 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 632 nicmp6->icmp6_code = 0; 633 if (n) { 634 icmp6stat.icp6s_reflect++; 635 icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++; 636 icmp6_reflect(n, noff); 637 } 638 if (!m) 639 goto freeit; 640 break; 641 642 case ICMP6_ECHO_REPLY: 643 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 644 if (code != 0) 645 goto badcode; 646 break; 647 648 case MLD_LISTENER_QUERY: 649 case MLD_LISTENER_REPORT: 650 if (icmp6len < sizeof(struct mld_hdr)) 651 goto badlen; 652 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */ 653 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 654 else 655 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 656 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 657 /* give up local */ 658 mld6_input(m, off); 659 m = NULL; 660 goto freeit; 661 } 662 mld6_input(n, off); 663 /* m stays. */ 664 break; 665 666 case MLD_LISTENER_DONE: 667 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 668 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */ 669 goto badlen; 670 break; /* nothing to be done in kernel */ 671 672 case MLD_MTRACE_RESP: 673 case MLD_MTRACE: 674 /* XXX: these two are experimental. not officially defined. */ 675 /* XXX: per-interface statistics? */ 676 break; /* just pass it to applications */ 677 678 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 679 { 680 enum { WRU, FQDN } mode; 681 682 if (!icmp6_nodeinfo) 683 break; 684 685 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 686 mode = WRU; 687 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 688 mode = FQDN; 689 else 690 goto badlen; 691 692 if (mode == FQDN) { 693 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 694 if (n) 695 n = ni6_input(n, off); 696 /* XXX meaningless if n == NULL */ 697 noff = sizeof(struct ip6_hdr); 698 } else { 699 u_char *p; 700 int maxlen, maxhlen; 701 702 if ((icmp6_nodeinfo & 5) != 5) 703 break; 704 705 if (code != 0) 706 goto badcode; 707 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; 708 if (maxlen >= MCLBYTES) { 709 /* Give up remote */ 710 break; 711 } 712 MGETHDR(n, M_DONTWAIT, m->m_type); 713 if (n && maxlen > MHLEN) { 714 MCLGET(n, M_DONTWAIT); 715 if ((n->m_flags & M_EXT) == 0) { 716 m_free(n); 717 n = NULL; 718 } 719 } 720 if (n == NULL) { 721 /* Give up remote */ 722 break; 723 } 724 n->m_pkthdr.rcvif = NULL; 725 n->m_len = 0; 726 maxhlen = M_TRAILINGSPACE(n) - maxlen; 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 bzero(p, 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 icmp6stat.icp6s_reflect++; 749 icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++; 750 icmp6_reflect(n, noff); 751 } 752 break; 753 } 754 755 case ICMP6_WRUREPLY: 756 if (code != 0) 757 goto badcode; 758 break; 759 760 case ND_ROUTER_SOLICIT: 761 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 762 if (code != 0) 763 goto badcode; 764 if (icmp6len < sizeof(struct nd_router_solicit)) 765 goto badlen; 766 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 767 /* give up local */ 768 nd6_rs_input(m, off, icmp6len); 769 m = NULL; 770 goto freeit; 771 } 772 nd6_rs_input(n, off, icmp6len); 773 /* m stays. */ 774 break; 775 776 case ND_ROUTER_ADVERT: 777 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 778 if (code != 0) 779 goto badcode; 780 if (icmp6len < sizeof(struct nd_router_advert)) 781 goto badlen; 782 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 783 /* give up local */ 784 nd6_ra_input(m, off, icmp6len); 785 m = NULL; 786 goto freeit; 787 } 788 nd6_ra_input(n, off, icmp6len); 789 /* m stays. */ 790 break; 791 792 case ND_NEIGHBOR_SOLICIT: 793 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 794 if (code != 0) 795 goto badcode; 796 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 797 goto badlen; 798 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 799 /* give up local */ 800 nd6_ns_input(m, off, icmp6len); 801 m = NULL; 802 goto freeit; 803 } 804 nd6_ns_input(n, off, icmp6len); 805 /* m stays. */ 806 break; 807 808 case ND_NEIGHBOR_ADVERT: 809 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 810 if (code != 0) 811 goto badcode; 812 if (icmp6len < sizeof(struct nd_neighbor_advert)) 813 goto badlen; 814 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 815 /* give up local */ 816 nd6_na_input(m, off, icmp6len); 817 m = NULL; 818 goto freeit; 819 } 820 nd6_na_input(n, off, icmp6len); 821 /* m stays. */ 822 break; 823 824 case ND_REDIRECT: 825 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 826 if (code != 0) 827 goto badcode; 828 if (icmp6len < sizeof(struct nd_redirect)) 829 goto badlen; 830 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 831 /* give up local */ 832 icmp6_redirect_input(m, off); 833 m = NULL; 834 goto freeit; 835 } 836 icmp6_redirect_input(n, off); 837 /* m stays. */ 838 break; 839 840 case ICMP6_ROUTER_RENUMBERING: 841 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 842 code != ICMP6_ROUTER_RENUMBERING_RESULT) 843 goto badcode; 844 if (icmp6len < sizeof(struct icmp6_router_renum)) 845 goto badlen; 846 break; 847 848 default: 849 nd6log((LOG_DEBUG, 850 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 851 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 852 ip6_sprintf(&ip6->ip6_dst), 853 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 854 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 855 /* ICMPv6 error: MUST deliver it by spec... */ 856 code = PRC_NCMDS; 857 /* deliver */ 858 } else { 859 /* ICMPv6 informational: MUST not deliver */ 860 break; 861 } 862 deliver: 863 if (icmp6_notify_error(m, off, icmp6len, code)) { 864 /* In this case, m should've been freed. */ 865 return (IPPROTO_DONE); 866 } 867 break; 868 869 badcode: 870 icmp6stat.icp6s_badcode++; 871 break; 872 873 badlen: 874 icmp6stat.icp6s_badlen++; 875 break; 876 } 877 878 /* deliver the packet to appropriate sockets */ 879 icmp6_rip6_input(&m, *offp); 880 881 return IPPROTO_DONE; 882 883 freeit: 884 m_freem(m); 885 return IPPROTO_DONE; 886 } 887 888 static int 889 icmp6_notify_error(m, off, icmp6len, code) 890 struct mbuf *m; 891 int off, icmp6len; 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 icmp6stat.icp6s_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 icmp6stat.icp6s_tooshort++; 906 return (-1); 907 } 908 eip6 = (struct ip6_hdr *)(icmp6 + 1); 909 910 /* Detect the upper level protocol */ 911 { 912 void (*ctlfunc) __P((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 icmp6stat.icp6s_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 icmp6stat.icp6s_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 icmp6stat.icp6s_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 icmp6stat.icp6s_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 icmp6stat.icp6s_tooshort++; 1023 return (-1); 1024 } 1025 1026 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1027 bzero(&icmp6dst, sizeof(icmp6dst)); 1028 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1029 icmp6dst.sin6_family = AF_INET6; 1030 if (finaldst == NULL) 1031 icmp6dst.sin6_addr = eip6->ip6_dst; 1032 else 1033 icmp6dst.sin6_addr = *finaldst; 1034 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1035 &icmp6dst.sin6_addr); 1036 #ifndef SCOPEDROUTING 1037 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst, 1038 NULL, NULL)) { 1039 /* should be impossbile */ 1040 nd6log((LOG_DEBUG, 1041 "icmp6_notify_error: in6_embedscope failed\n")); 1042 goto freeit; 1043 } 1044 #endif 1045 1046 /* 1047 * retrieve parameters from the inner IPv6 header, and convert 1048 * them into sockaddr structures. 1049 */ 1050 bzero(&icmp6src, sizeof(icmp6src)); 1051 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1052 icmp6src.sin6_family = AF_INET6; 1053 icmp6src.sin6_addr = eip6->ip6_src; 1054 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1055 &icmp6src.sin6_addr); 1056 #ifndef SCOPEDROUTING 1057 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src, 1058 NULL, NULL)) { 1059 /* should be impossbile */ 1060 nd6log((LOG_DEBUG, 1061 "icmp6_notify_error: in6_embedscope failed\n")); 1062 goto freeit; 1063 } 1064 #endif 1065 icmp6src.sin6_flowinfo = 1066 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1067 1068 if (finaldst == NULL) 1069 finaldst = &eip6->ip6_dst; 1070 ip6cp.ip6c_m = m; 1071 ip6cp.ip6c_icmp6 = icmp6; 1072 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1073 ip6cp.ip6c_off = eoff; 1074 ip6cp.ip6c_finaldst = finaldst; 1075 ip6cp.ip6c_src = &icmp6src; 1076 ip6cp.ip6c_nxt = nxt; 1077 1078 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1079 notifymtu = ntohl(icmp6->icmp6_mtu); 1080 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1081 } 1082 1083 ctlfunc = (void (*) __P((int, struct sockaddr *, void *))) 1084 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1085 if (ctlfunc) { 1086 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1087 &ip6cp); 1088 } 1089 } 1090 return (0); 1091 1092 freeit: 1093 m_freem(m); 1094 return (-1); 1095 } 1096 1097 void 1098 icmp6_mtudisc_update(ip6cp, validated) 1099 struct ip6ctlparam *ip6cp; 1100 int validated; 1101 { 1102 unsigned long rtcount; 1103 struct icmp6_mtudisc_callback *mc; 1104 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1105 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1106 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1107 u_int mtu = ntohl(icmp6->icmp6_mtu); 1108 struct rtentry *rt = NULL; 1109 struct sockaddr_in6 sin6; 1110 1111 /* 1112 * allow non-validated cases if memory is plenty, to make traffic 1113 * from non-connected pcb happy. 1114 */ 1115 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1116 if (validated) { 1117 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1118 return; 1119 else if (0 <= icmp6_mtudisc_lowat && 1120 rtcount > icmp6_mtudisc_lowat) { 1121 /* 1122 * XXX nuke a victim, install the new one. 1123 */ 1124 } 1125 } else { 1126 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1127 return; 1128 } 1129 1130 bzero(&sin6, sizeof(sin6)); 1131 sin6.sin6_family = PF_INET6; 1132 sin6.sin6_len = sizeof(struct sockaddr_in6); 1133 sin6.sin6_addr = *dst; 1134 /* XXX normally, this won't happen */ 1135 if (IN6_IS_ADDR_LINKLOCAL(dst)) { 1136 sin6.sin6_addr.s6_addr16[1] = 1137 htons(m->m_pkthdr.rcvif->if_index); 1138 } 1139 /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */ 1140 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1141 1142 if (rt && (rt->rt_flags & RTF_HOST) && 1143 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1144 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1145 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1146 icmp6stat.icp6s_pmtuchg++; 1147 rt->rt_rmx.rmx_mtu = mtu; 1148 } 1149 } 1150 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */ 1151 RTFREE(rt); 1152 } 1153 1154 /* 1155 * Notify protocols that the MTU for this destination 1156 * has changed. 1157 */ 1158 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1159 mc = LIST_NEXT(mc, mc_list)) 1160 (*mc->mc_func)(&sin6.sin6_addr); 1161 } 1162 1163 /* 1164 * Process a Node Information Query packet, based on 1165 * draft-ietf-ipngwg-icmp-name-lookups-07. 1166 * 1167 * Spec incompatibilities: 1168 * - IPv6 Subject address handling 1169 * - IPv4 Subject address handling support missing 1170 * - Proxy reply (answer even if it's not for me) 1171 * - joins NI group address at in6_ifattach() time only, does not cope 1172 * with hostname changes by sethostname(3) 1173 */ 1174 #ifndef offsetof /* XXX */ 1175 #define offsetof(type, member) ((size_t)(&((type *)0)->member)) 1176 #endif 1177 static struct mbuf * 1178 ni6_input(m, off) 1179 struct mbuf *m; 1180 int off; 1181 { 1182 struct icmp6_nodeinfo *ni6, *nni6; 1183 struct mbuf *n = NULL; 1184 u_int16_t qtype; 1185 int subjlen; 1186 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1187 struct ni_reply_fqdn *fqdn; 1188 int addrs; /* for NI_QTYPE_NODEADDR */ 1189 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1190 struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */ 1191 struct ip6_hdr *ip6; 1192 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1193 char *subj = NULL; 1194 1195 ip6 = mtod(m, struct ip6_hdr *); 1196 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1197 if (ni6 == NULL) { 1198 /* m is already reclaimed */ 1199 return NULL; 1200 } 1201 1202 /* 1203 * Validate IPv6 destination address. 1204 * 1205 * The Responder must discard the Query without further processing 1206 * unless it is one of the Responder's unicast or anycast addresses, or 1207 * a link-local scope multicast address which the Responder has joined. 1208 * [icmp-name-lookups-07, Section 4.] 1209 */ 1210 bzero(&sin6, sizeof(sin6)); 1211 sin6.sin6_family = AF_INET6; 1212 sin6.sin6_len = sizeof(struct sockaddr_in6); 1213 bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr)); 1214 /* XXX scopeid */ 1215 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1216 ; /* unicast/anycast, fine */ 1217 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1218 ; /* link-local multicast, fine */ 1219 else 1220 goto bad; 1221 1222 /* validate query Subject field. */ 1223 qtype = ntohs(ni6->ni_qtype); 1224 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1225 switch (qtype) { 1226 case NI_QTYPE_NOOP: 1227 case NI_QTYPE_SUPTYPES: 1228 /* 07 draft */ 1229 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1230 break; 1231 /* FALLTHROUGH */ 1232 case NI_QTYPE_FQDN: 1233 case NI_QTYPE_NODEADDR: 1234 switch (ni6->ni_code) { 1235 case ICMP6_NI_SUBJ_IPV6: 1236 #if ICMP6_NI_SUBJ_IPV6 != 0 1237 case 0: 1238 #endif 1239 /* 1240 * backward compatibility - try to accept 03 draft 1241 * format, where no Subject is present. 1242 */ 1243 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1244 subjlen == 0) { 1245 oldfqdn++; 1246 break; 1247 } 1248 #if ICMP6_NI_SUBJ_IPV6 != 0 1249 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1250 goto bad; 1251 #endif 1252 1253 if (subjlen != sizeof(sin6.sin6_addr)) 1254 goto bad; 1255 1256 /* 1257 * Validate Subject address. 1258 * 1259 * Not sure what exactly "address belongs to the node" 1260 * means in the spec, is it just unicast, or what? 1261 * 1262 * At this moment we consider Subject address as 1263 * "belong to the node" if the Subject address equals 1264 * to the IPv6 destination address; validation for 1265 * IPv6 destination address should have done enough 1266 * check for us. 1267 * 1268 * We do not do proxy at this moment. 1269 */ 1270 /* m_pulldown instead of copy? */ 1271 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1272 subjlen, (caddr_t)&sin6.sin6_addr); 1273 /* XXX kame scope hack */ 1274 if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) { 1275 if ((m->m_flags & M_PKTHDR) != 0 && 1276 m->m_pkthdr.rcvif) { 1277 sin6.sin6_addr.s6_addr16[1] = 1278 htons(m->m_pkthdr.rcvif->if_index); 1279 } 1280 } 1281 subj = (char *)&sin6; 1282 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr)) 1283 break; 1284 1285 /* 1286 * XXX if we are to allow other cases, we should really 1287 * be careful about scope here. 1288 * basically, we should disallow queries toward IPv6 1289 * destination X with subject Y, if scope(X) > scope(Y). 1290 * if we allow scope(X) > scope(Y), it will result in 1291 * information leakage across scope boundary. 1292 */ 1293 goto bad; 1294 1295 case ICMP6_NI_SUBJ_FQDN: 1296 /* 1297 * Validate Subject name with gethostname(3). 1298 * 1299 * The behavior may need some debate, since: 1300 * - we are not sure if the node has FQDN as 1301 * hostname (returned by gethostname(3)). 1302 * - the code does wildcard match for truncated names. 1303 * however, we are not sure if we want to perform 1304 * wildcard match, if gethostname(3) side has 1305 * truncated hostname. 1306 */ 1307 n = ni6_nametodns(hostname, hostnamelen, 0); 1308 if (!n || n->m_next || n->m_len == 0) 1309 goto bad; 1310 IP6_EXTHDR_GET(subj, char *, m, 1311 off + sizeof(struct icmp6_nodeinfo), subjlen); 1312 if (subj == NULL) 1313 goto bad; 1314 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1315 n->m_len)) { 1316 goto bad; 1317 } 1318 m_freem(n); 1319 n = NULL; 1320 break; 1321 1322 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1323 default: 1324 goto bad; 1325 } 1326 break; 1327 } 1328 1329 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1330 switch (qtype) { 1331 case NI_QTYPE_FQDN: 1332 if ((icmp6_nodeinfo & 1) == 0) 1333 goto bad; 1334 break; 1335 case NI_QTYPE_NODEADDR: 1336 if ((icmp6_nodeinfo & 2) == 0) 1337 goto bad; 1338 break; 1339 } 1340 1341 /* guess reply length */ 1342 switch (qtype) { 1343 case NI_QTYPE_NOOP: 1344 break; /* no reply data */ 1345 case NI_QTYPE_SUPTYPES: 1346 replylen += sizeof(u_int32_t); 1347 break; 1348 case NI_QTYPE_FQDN: 1349 /* XXX will append an mbuf */ 1350 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1351 break; 1352 case NI_QTYPE_NODEADDR: 1353 addrs = ni6_addrs(ni6, m, &ifp, subj); 1354 if ((replylen += addrs * (sizeof(struct in6_addr) + 1355 sizeof(u_int32_t))) > MCLBYTES) 1356 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1357 break; 1358 default: 1359 /* 1360 * XXX: We must return a reply with the ICMP6 code 1361 * `unknown Qtype' in this case. However we regard the case 1362 * as an FQDN query for backward compatibility. 1363 * Older versions set a random value to this field, 1364 * so it rarely varies in the defined qtypes. 1365 * But the mechanism is not reliable... 1366 * maybe we should obsolete older versions. 1367 */ 1368 qtype = NI_QTYPE_FQDN; 1369 /* XXX will append an mbuf */ 1370 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1371 oldfqdn++; 1372 break; 1373 } 1374 1375 /* allocate an mbuf to reply. */ 1376 MGETHDR(n, M_DONTWAIT, m->m_type); 1377 if (n == NULL) { 1378 m_freem(m); 1379 return (NULL); 1380 } 1381 M_COPY_PKTHDR(n, m); /* just for rcvif */ 1382 if (replylen > MHLEN) { 1383 if (replylen > MCLBYTES) { 1384 /* 1385 * XXX: should we try to allocate more? But MCLBYTES 1386 * is probably much larger than IPV6_MMTU... 1387 */ 1388 goto bad; 1389 } 1390 MCLGET(n, M_DONTWAIT); 1391 if ((n->m_flags & M_EXT) == 0) { 1392 goto bad; 1393 } 1394 } 1395 n->m_pkthdr.len = n->m_len = replylen; 1396 1397 /* copy mbuf header and IPv6 + Node Information base headers */ 1398 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1399 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1400 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1401 1402 /* qtype dependent procedure */ 1403 switch (qtype) { 1404 case NI_QTYPE_NOOP: 1405 nni6->ni_code = ICMP6_NI_SUCCESS; 1406 nni6->ni_flags = 0; 1407 break; 1408 case NI_QTYPE_SUPTYPES: 1409 { 1410 u_int32_t v; 1411 nni6->ni_code = ICMP6_NI_SUCCESS; 1412 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1413 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1414 v = (u_int32_t)htonl(0x0000000f); 1415 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1416 break; 1417 } 1418 case NI_QTYPE_FQDN: 1419 nni6->ni_code = ICMP6_NI_SUCCESS; 1420 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1421 sizeof(struct ip6_hdr) + 1422 sizeof(struct icmp6_nodeinfo)); 1423 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1424 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1425 /* 1426 * XXX do we really have FQDN in variable "hostname"? 1427 */ 1428 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1429 if (n->m_next == NULL) 1430 goto bad; 1431 /* XXX we assume that n->m_next is not a chain */ 1432 if (n->m_next->m_next != NULL) 1433 goto bad; 1434 n->m_pkthdr.len += n->m_next->m_len; 1435 break; 1436 case NI_QTYPE_NODEADDR: 1437 { 1438 int lenlim, copied; 1439 1440 nni6->ni_code = ICMP6_NI_SUCCESS; 1441 n->m_pkthdr.len = n->m_len = 1442 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1443 lenlim = M_TRAILINGSPACE(n); 1444 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1445 /* XXX: reset mbuf length */ 1446 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1447 sizeof(struct icmp6_nodeinfo) + copied; 1448 break; 1449 } 1450 default: 1451 break; /* XXX impossible! */ 1452 } 1453 1454 nni6->ni_type = ICMP6_NI_REPLY; 1455 m_freem(m); 1456 return (n); 1457 1458 bad: 1459 m_freem(m); 1460 if (n) 1461 m_freem(n); 1462 return (NULL); 1463 } 1464 #undef hostnamelen 1465 1466 #define isupper(x) ('A' <= (x) && (x) <= 'Z') 1467 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1468 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1469 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1470 1471 /* 1472 * make a mbuf with DNS-encoded string. no compression support. 1473 * 1474 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1475 * treated as truncated name (two \0 at the end). this is a wild guess. 1476 */ 1477 static struct mbuf * 1478 ni6_nametodns(name, namelen, old) 1479 const char *name; 1480 int namelen; 1481 int old; /* return pascal string if non-zero */ 1482 { 1483 struct mbuf *m; 1484 char *cp, *ep; 1485 const char *p, *q; 1486 int i, len, nterm; 1487 1488 if (old) 1489 len = namelen + 1; 1490 else 1491 len = MCLBYTES; 1492 1493 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1494 MGET(m, M_DONTWAIT, MT_DATA); 1495 if (m && len > MLEN) { 1496 MCLGET(m, M_DONTWAIT); 1497 if ((m->m_flags & M_EXT) == 0) 1498 goto fail; 1499 } 1500 if (!m) 1501 goto fail; 1502 m->m_next = NULL; 1503 1504 if (old) { 1505 m->m_len = len; 1506 *mtod(m, char *) = namelen; 1507 bcopy(name, mtod(m, char *) + 1, namelen); 1508 return m; 1509 } else { 1510 m->m_len = 0; 1511 cp = mtod(m, char *); 1512 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1513 1514 /* if not certain about my name, return empty buffer */ 1515 if (namelen == 0) 1516 return m; 1517 1518 /* 1519 * guess if it looks like shortened hostname, or FQDN. 1520 * shortened hostname needs two trailing "\0". 1521 */ 1522 i = 0; 1523 for (p = name; p < name + namelen; p++) { 1524 if (*p && *p == '.') 1525 i++; 1526 } 1527 if (i < 2) 1528 nterm = 2; 1529 else 1530 nterm = 1; 1531 1532 p = name; 1533 while (cp < ep && p < name + namelen) { 1534 i = 0; 1535 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1536 i++; 1537 /* result does not fit into mbuf */ 1538 if (cp + i + 1 >= ep) 1539 goto fail; 1540 /* 1541 * DNS label length restriction, RFC1035 page 8. 1542 * "i == 0" case is included here to avoid returning 1543 * 0-length label on "foo..bar". 1544 */ 1545 if (i <= 0 || i >= 64) 1546 goto fail; 1547 *cp++ = i; 1548 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1549 goto fail; 1550 while (i > 0) { 1551 if (!isalnum(*p) && *p != '-') 1552 goto fail; 1553 if (isupper(*p)) { 1554 *cp++ = tolower(*p); 1555 p++; 1556 } else 1557 *cp++ = *p++; 1558 i--; 1559 } 1560 p = q; 1561 if (p < name + namelen && *p == '.') 1562 p++; 1563 } 1564 /* termination */ 1565 if (cp + nterm >= ep) 1566 goto fail; 1567 while (nterm-- > 0) 1568 *cp++ = '\0'; 1569 m->m_len = cp - mtod(m, char *); 1570 return m; 1571 } 1572 1573 panic("should not reach here"); 1574 /* NOTREACHED */ 1575 1576 fail: 1577 if (m) 1578 m_freem(m); 1579 return NULL; 1580 } 1581 1582 /* 1583 * check if two DNS-encoded string matches. takes care of truncated 1584 * form (with \0\0 at the end). no compression support. 1585 * XXX upper/lowercase match (see RFC2065) 1586 */ 1587 static int 1588 ni6_dnsmatch(a, alen, b, blen) 1589 const char *a; 1590 int alen; 1591 const char *b; 1592 int blen; 1593 { 1594 const char *a0, *b0; 1595 int l; 1596 1597 /* simplest case - need validation? */ 1598 if (alen == blen && bcmp(a, b, alen) == 0) 1599 return 1; 1600 1601 a0 = a; 1602 b0 = b; 1603 1604 /* termination is mandatory */ 1605 if (alen < 2 || blen < 2) 1606 return 0; 1607 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1608 return 0; 1609 alen--; 1610 blen--; 1611 1612 while (a - a0 < alen && b - b0 < blen) { 1613 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1614 return 0; 1615 1616 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1617 return 0; 1618 /* we don't support compression yet */ 1619 if (a[0] >= 64 || b[0] >= 64) 1620 return 0; 1621 1622 /* truncated case */ 1623 if (a[0] == 0 && a - a0 == alen - 1) 1624 return 1; 1625 if (b[0] == 0 && b - b0 == blen - 1) 1626 return 1; 1627 if (a[0] == 0 || b[0] == 0) 1628 return 0; 1629 1630 if (a[0] != b[0]) 1631 return 0; 1632 l = a[0]; 1633 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1634 return 0; 1635 if (bcmp(a + 1, b + 1, l) != 0) 1636 return 0; 1637 1638 a += 1 + l; 1639 b += 1 + l; 1640 } 1641 1642 if (a - a0 == alen && b - b0 == blen) 1643 return 1; 1644 else 1645 return 0; 1646 } 1647 1648 /* 1649 * calculate the number of addresses to be returned in the node info reply. 1650 */ 1651 static int 1652 ni6_addrs(ni6, m, ifpp, subj) 1653 struct icmp6_nodeinfo *ni6; 1654 struct mbuf *m; 1655 struct ifnet **ifpp; 1656 char *subj; 1657 { 1658 struct ifnet *ifp; 1659 struct in6_ifaddr *ifa6; 1660 struct ifaddr *ifa; 1661 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1662 int addrs = 0, addrsofif, iffound = 0; 1663 int niflags = ni6->ni_flags; 1664 1665 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1666 switch (ni6->ni_code) { 1667 case ICMP6_NI_SUBJ_IPV6: 1668 if (subj == NULL) /* must be impossible... */ 1669 return (0); 1670 subj_ip6 = (struct sockaddr_in6 *)subj; 1671 break; 1672 default: 1673 /* 1674 * XXX: we only support IPv6 subject address for 1675 * this Qtype. 1676 */ 1677 return (0); 1678 } 1679 } 1680 1681 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1682 { 1683 addrsofif = 0; 1684 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1685 ifa = ifa->ifa_list.tqe_next) 1686 { 1687 if (ifa->ifa_addr->sa_family != AF_INET6) 1688 continue; 1689 ifa6 = (struct in6_ifaddr *)ifa; 1690 1691 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1692 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1693 &ifa6->ia_addr.sin6_addr)) 1694 iffound = 1; 1695 1696 /* 1697 * IPv4-mapped addresses can only be returned by a 1698 * Node Information proxy, since they represent 1699 * addresses of IPv4-only nodes, which perforce do 1700 * not implement this protocol. 1701 * [icmp-name-lookups-07, Section 5.4] 1702 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1703 * this function at this moment. 1704 */ 1705 1706 /* What do we have to do about ::1? */ 1707 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1708 case IPV6_ADDR_SCOPE_LINKLOCAL: 1709 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1710 continue; 1711 break; 1712 case IPV6_ADDR_SCOPE_SITELOCAL: 1713 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1714 continue; 1715 break; 1716 case IPV6_ADDR_SCOPE_GLOBAL: 1717 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1718 continue; 1719 break; 1720 default: 1721 continue; 1722 } 1723 1724 /* 1725 * check if anycast is okay. 1726 * XXX: just experimental. not in the spec. 1727 */ 1728 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1729 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1730 continue; /* we need only unicast addresses */ 1731 1732 addrsofif++; /* count the address */ 1733 } 1734 if (iffound) { 1735 *ifpp = ifp; 1736 return (addrsofif); 1737 } 1738 1739 addrs += addrsofif; 1740 } 1741 1742 return (addrs); 1743 } 1744 1745 static int 1746 ni6_store_addrs(ni6, nni6, ifp0, resid) 1747 struct icmp6_nodeinfo *ni6, *nni6; 1748 struct ifnet *ifp0; 1749 int resid; 1750 { 1751 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1752 struct in6_ifaddr *ifa6; 1753 struct ifaddr *ifa; 1754 struct ifnet *ifp_dep = NULL; 1755 int copied = 0, allow_deprecated = 0; 1756 u_char *cp = (u_char *)(nni6 + 1); 1757 int niflags = ni6->ni_flags; 1758 u_int32_t ltime; 1759 1760 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1761 return (0); /* needless to copy */ 1762 1763 again: 1764 1765 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1766 { 1767 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1768 ifa = ifa->ifa_list.tqe_next) 1769 { 1770 if (ifa->ifa_addr->sa_family != AF_INET6) 1771 continue; 1772 ifa6 = (struct in6_ifaddr *)ifa; 1773 1774 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1775 allow_deprecated == 0) { 1776 /* 1777 * prefererred address should be put before 1778 * deprecated addresses. 1779 */ 1780 1781 /* record the interface for later search */ 1782 if (ifp_dep == NULL) 1783 ifp_dep = ifp; 1784 1785 continue; 1786 } 1787 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1788 allow_deprecated != 0) 1789 continue; /* we now collect deprecated addrs */ 1790 1791 /* What do we have to do about ::1? */ 1792 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1793 case IPV6_ADDR_SCOPE_LINKLOCAL: 1794 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1795 continue; 1796 break; 1797 case IPV6_ADDR_SCOPE_SITELOCAL: 1798 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1799 continue; 1800 break; 1801 case IPV6_ADDR_SCOPE_GLOBAL: 1802 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1803 continue; 1804 break; 1805 default: 1806 continue; 1807 } 1808 1809 /* 1810 * check if anycast is okay. 1811 * XXX: just experimental. not in the spec. 1812 */ 1813 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1814 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1815 continue; 1816 1817 /* now we can copy the address */ 1818 if (resid < sizeof(struct in6_addr) + 1819 sizeof(u_int32_t)) { 1820 /* 1821 * We give up much more copy. 1822 * Set the truncate flag and return. 1823 */ 1824 nni6->ni_flags |= 1825 NI_NODEADDR_FLAG_TRUNCATE; 1826 return (copied); 1827 } 1828 1829 /* 1830 * Set the TTL of the address. 1831 * The TTL value should be one of the following 1832 * according to the specification: 1833 * 1834 * 1. The remaining lifetime of a DHCP lease on the 1835 * address, or 1836 * 2. The remaining Valid Lifetime of a prefix from 1837 * which the address was derived through Stateless 1838 * Autoconfiguration. 1839 * 1840 * Note that we currently do not support stateful 1841 * address configuration by DHCPv6, so the former 1842 * case can't happen. 1843 * 1844 * TTL must be 2^31 > TTL >= 0. 1845 */ 1846 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1847 ltime = ND6_INFINITE_LIFETIME; 1848 else { 1849 if (ifa6->ia6_lifetime.ia6t_expire > 1850 time.tv_sec) 1851 ltime = ifa6->ia6_lifetime.ia6t_expire - time.tv_sec; 1852 else 1853 ltime = 0; 1854 } 1855 if (ltime > 0x7fffffff) 1856 ltime = 0x7fffffff; 1857 ltime = htonl(ltime); 1858 1859 bcopy(<ime, cp, sizeof(u_int32_t)); 1860 cp += sizeof(u_int32_t); 1861 1862 /* copy the address itself */ 1863 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1864 sizeof(struct in6_addr)); 1865 /* XXX: KAME link-local hack; remove ifindex */ 1866 if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr)) 1867 ((struct in6_addr *)cp)->s6_addr16[1] = 0; 1868 cp += sizeof(struct in6_addr); 1869 1870 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1871 copied += (sizeof(struct in6_addr) + 1872 sizeof(u_int32_t)); 1873 } 1874 if (ifp0) /* we need search only on the specified IF */ 1875 break; 1876 } 1877 1878 if (allow_deprecated == 0 && ifp_dep != NULL) { 1879 ifp = ifp_dep; 1880 allow_deprecated = 1; 1881 1882 goto again; 1883 } 1884 1885 return (copied); 1886 } 1887 1888 /* 1889 * XXX almost dup'ed code with rip6_input. 1890 */ 1891 static int 1892 icmp6_rip6_input(mp, off) 1893 struct mbuf **mp; 1894 int off; 1895 { 1896 struct mbuf *m = *mp; 1897 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1898 struct inpcb_hdr *inph; 1899 struct in6pcb *in6p; 1900 struct in6pcb *last = NULL; 1901 struct sockaddr_in6 rip6src; 1902 struct icmp6_hdr *icmp6; 1903 struct mbuf *opts = NULL; 1904 1905 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1906 if (icmp6 == NULL) { 1907 /* m is already reclaimed */ 1908 return IPPROTO_DONE; 1909 } 1910 1911 bzero(&rip6src, sizeof(rip6src)); 1912 rip6src.sin6_len = sizeof(struct sockaddr_in6); 1913 rip6src.sin6_family = AF_INET6; 1914 /* KAME hack: recover scopeid */ 1915 (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif); 1916 1917 CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) { 1918 in6p = (struct in6pcb *)inph; 1919 if (in6p->in6p_af != AF_INET6) 1920 continue; 1921 if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6) 1922 continue; 1923 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1924 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1925 continue; 1926 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1927 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1928 continue; 1929 if (in6p->in6p_icmp6filt 1930 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1931 in6p->in6p_icmp6filt)) 1932 continue; 1933 if (last) { 1934 struct mbuf *n; 1935 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1936 if (last->in6p_flags & IN6P_CONTROLOPTS) 1937 ip6_savecontrol(last, &opts, ip6, n); 1938 /* strip intermediate headers */ 1939 m_adj(n, off); 1940 if (sbappendaddr(&last->in6p_socket->so_rcv, 1941 (struct sockaddr *)&rip6src, 1942 n, opts) == 0) { 1943 /* should notify about lost packet */ 1944 m_freem(n); 1945 if (opts) 1946 m_freem(opts); 1947 } else 1948 sorwakeup(last->in6p_socket); 1949 opts = NULL; 1950 } 1951 } 1952 last = in6p; 1953 } 1954 if (last) { 1955 if (last->in6p_flags & IN6P_CONTROLOPTS) 1956 ip6_savecontrol(last, &opts, ip6, m); 1957 /* strip intermediate headers */ 1958 m_adj(m, off); 1959 if (sbappendaddr(&last->in6p_socket->so_rcv, 1960 (struct sockaddr *)&rip6src, m, opts) == 0) { 1961 m_freem(m); 1962 if (opts) 1963 m_freem(opts); 1964 } else 1965 sorwakeup(last->in6p_socket); 1966 } else { 1967 m_freem(m); 1968 ip6stat.ip6s_delivered--; 1969 } 1970 return IPPROTO_DONE; 1971 } 1972 1973 /* 1974 * Reflect the ip6 packet back to the source. 1975 * OFF points to the icmp6 header, counted from the top of the mbuf. 1976 * 1977 * Note: RFC 1885 required that an echo reply should be truncated if it 1978 * did not fit in with (return) path MTU, and KAME code supported the 1979 * behavior. However, as a clarification after the RFC, this limitation 1980 * was removed in a revised version of the spec, RFC 2463. We had kept the 1981 * old behavior, with a (non-default) ifdef block, while the new version of 1982 * the spec was an internet-draft status, and even after the new RFC was 1983 * published. But it would rather make sense to clean the obsoleted part 1984 * up, and to make the code simpler at this stage. 1985 */ 1986 void 1987 icmp6_reflect(m, off) 1988 struct mbuf *m; 1989 size_t off; 1990 { 1991 struct ip6_hdr *ip6; 1992 struct icmp6_hdr *icmp6; 1993 struct in6_ifaddr *ia; 1994 struct in6_addr t, *src = 0; 1995 int plen; 1996 int type, code; 1997 struct ifnet *outif = NULL; 1998 struct sockaddr_in6 sa6_src, sa6_dst; 1999 2000 /* too short to reflect */ 2001 if (off < sizeof(struct ip6_hdr)) { 2002 nd6log((LOG_DEBUG, 2003 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 2004 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2005 __FILE__, __LINE__)); 2006 goto bad; 2007 } 2008 2009 /* 2010 * If there are extra headers between IPv6 and ICMPv6, strip 2011 * off that header first. 2012 */ 2013 #ifdef DIAGNOSTIC 2014 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2015 panic("assumption failed in icmp6_reflect"); 2016 #endif 2017 if (off > sizeof(struct ip6_hdr)) { 2018 size_t l; 2019 struct ip6_hdr nip6; 2020 2021 l = off - sizeof(struct ip6_hdr); 2022 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2023 m_adj(m, l); 2024 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2025 if (m->m_len < l) { 2026 if ((m = m_pullup(m, l)) == NULL) 2027 return; 2028 } 2029 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2030 } else /* off == sizeof(struct ip6_hdr) */ { 2031 size_t l; 2032 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2033 if (m->m_len < l) { 2034 if ((m = m_pullup(m, l)) == NULL) 2035 return; 2036 } 2037 } 2038 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2039 ip6 = mtod(m, struct ip6_hdr *); 2040 ip6->ip6_nxt = IPPROTO_ICMPV6; 2041 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2042 type = icmp6->icmp6_type; /* keep type for statistics */ 2043 code = icmp6->icmp6_code; /* ditto. */ 2044 2045 t = ip6->ip6_dst; 2046 /* 2047 * ip6_input() drops a packet if its src is multicast. 2048 * So, the src is never multicast. 2049 */ 2050 ip6->ip6_dst = ip6->ip6_src; 2051 2052 /* 2053 * XXX: make sure to embed scope zone information, using 2054 * already embedded IDs or the received interface (if any). 2055 * Note that rcvif may be NULL. 2056 * TODO: scoped routing case (XXX). 2057 */ 2058 bzero(&sa6_src, sizeof(sa6_src)); 2059 sa6_src.sin6_family = AF_INET6; 2060 sa6_src.sin6_len = sizeof(sa6_src); 2061 sa6_src.sin6_addr = ip6->ip6_dst; 2062 in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif); 2063 in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL); 2064 ip6->ip6_dst = sa6_src.sin6_addr; 2065 2066 bzero(&sa6_dst, sizeof(sa6_dst)); 2067 sa6_dst.sin6_family = AF_INET6; 2068 sa6_dst.sin6_len = sizeof(sa6_dst); 2069 sa6_dst.sin6_addr = t; 2070 in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif); 2071 in6_embedscope(&t, &sa6_dst, NULL, NULL); 2072 2073 /* 2074 * If the incoming packet was addressed directly to us (i.e. unicast), 2075 * use dst as the src for the reply. 2076 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible 2077 * (for example) when we encounter an error while forwarding procedure 2078 * destined to a duplicated address of ours. 2079 */ 2080 for (ia = in6_ifaddr; ia; ia = ia->ia_next) 2081 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) && 2082 (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) { 2083 src = &t; 2084 break; 2085 } 2086 if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) { 2087 /* 2088 * This is the case if the dst is our link-local address 2089 * and the sender is also ourselves. 2090 */ 2091 src = &t; 2092 } 2093 2094 if (src == 0) { 2095 int e; 2096 struct route_in6 ro; 2097 2098 /* 2099 * This case matches to multicasts, our anycast, or unicasts 2100 * that we do not own. Select a source address based on the 2101 * source address of the erroneous packet. 2102 */ 2103 bzero(&ro, sizeof(ro)); 2104 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e); 2105 if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */ 2106 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2107 } 2108 if (src == NULL) { 2109 nd6log((LOG_DEBUG, 2110 "icmp6_reflect: source can't be determined: " 2111 "dst=%s, error=%d\n", 2112 ip6_sprintf(&sa6_src.sin6_addr), e)); 2113 goto bad; 2114 } 2115 } 2116 2117 ip6->ip6_src = *src; 2118 2119 ip6->ip6_flow = 0; 2120 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2121 ip6->ip6_vfc |= IPV6_VERSION; 2122 ip6->ip6_nxt = IPPROTO_ICMPV6; 2123 if (m->m_pkthdr.rcvif) { 2124 /* XXX: This may not be the outgoing interface */ 2125 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2126 } else 2127 ip6->ip6_hlim = ip6_defhlim; 2128 2129 icmp6->icmp6_cksum = 0; 2130 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2131 sizeof(struct ip6_hdr), plen); 2132 2133 /* 2134 * XXX option handling 2135 */ 2136 2137 m->m_flags &= ~(M_BCAST|M_MCAST); 2138 2139 /* 2140 * To avoid a "too big" situation at an intermediate router 2141 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2142 * Note that only echo and node information replies are affected, 2143 * since the length of ICMP6 errors is limited to the minimum MTU. 2144 */ 2145 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, 2146 (struct ip6_moptions *)NULL, (struct socket *)NULL, &outif) != 0 2147 && outif) 2148 icmp6_ifstat_inc(outif, ifs6_out_error); 2149 2150 if (outif) 2151 icmp6_ifoutstat_inc(outif, type, code); 2152 2153 return; 2154 2155 bad: 2156 m_freem(m); 2157 return; 2158 } 2159 2160 void 2161 icmp6_fasttimo() 2162 { 2163 2164 mld6_fasttimeo(); 2165 } 2166 2167 static const char * 2168 icmp6_redirect_diag(src6, dst6, tgt6) 2169 struct in6_addr *src6; 2170 struct in6_addr *dst6; 2171 struct in6_addr *tgt6; 2172 { 2173 static char buf[1024]; 2174 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2175 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2176 return buf; 2177 } 2178 2179 void 2180 icmp6_redirect_input(m, off) 2181 struct mbuf *m; 2182 int off; 2183 { 2184 struct ifnet *ifp = m->m_pkthdr.rcvif; 2185 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2186 struct nd_redirect *nd_rd; 2187 int icmp6len = ntohs(ip6->ip6_plen); 2188 char *lladdr = NULL; 2189 int lladdrlen = 0; 2190 struct rtentry *rt = NULL; 2191 int is_router; 2192 int is_onlink; 2193 struct in6_addr src6 = ip6->ip6_src; 2194 struct in6_addr redtgt6; 2195 struct in6_addr reddst6; 2196 union nd_opts ndopts; 2197 2198 if (!ifp) 2199 return; 2200 2201 /* XXX if we are router, we don't update route by icmp6 redirect */ 2202 if (ip6_forwarding) 2203 goto freeit; 2204 if (!icmp6_rediraccept) 2205 goto freeit; 2206 2207 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2208 if (nd_rd == NULL) { 2209 icmp6stat.icp6s_tooshort++; 2210 return; 2211 } 2212 redtgt6 = nd_rd->nd_rd_target; 2213 reddst6 = nd_rd->nd_rd_dst; 2214 2215 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2216 redtgt6.s6_addr16[1] = htons(ifp->if_index); 2217 if (IN6_IS_ADDR_LINKLOCAL(&reddst6)) 2218 reddst6.s6_addr16[1] = htons(ifp->if_index); 2219 2220 /* validation */ 2221 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2222 nd6log((LOG_ERR, 2223 "ICMP6 redirect sent from %s rejected; " 2224 "must be from linklocal\n", ip6_sprintf(&src6))); 2225 goto bad; 2226 } 2227 if (ip6->ip6_hlim != 255) { 2228 nd6log((LOG_ERR, 2229 "ICMP6 redirect sent from %s rejected; " 2230 "hlim=%d (must be 255)\n", 2231 ip6_sprintf(&src6), ip6->ip6_hlim)); 2232 goto bad; 2233 } 2234 { 2235 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2236 struct sockaddr_in6 sin6; 2237 struct in6_addr *gw6; 2238 2239 bzero(&sin6, sizeof(sin6)); 2240 sin6.sin6_family = AF_INET6; 2241 sin6.sin6_len = sizeof(struct sockaddr_in6); 2242 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2243 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2244 if (rt) { 2245 if (rt->rt_gateway == NULL || 2246 rt->rt_gateway->sa_family != AF_INET6) { 2247 nd6log((LOG_ERR, 2248 "ICMP6 redirect rejected; no route " 2249 "with inet6 gateway found for redirect dst: %s\n", 2250 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2251 RTFREE(rt); 2252 goto bad; 2253 } 2254 2255 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2256 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2257 nd6log((LOG_ERR, 2258 "ICMP6 redirect rejected; " 2259 "not equal to gw-for-src=%s (must be same): " 2260 "%s\n", 2261 ip6_sprintf(gw6), 2262 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2263 RTFREE(rt); 2264 goto bad; 2265 } 2266 } else { 2267 nd6log((LOG_ERR, 2268 "ICMP6 redirect rejected; " 2269 "no route found for redirect dst: %s\n", 2270 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2271 goto bad; 2272 } 2273 RTFREE(rt); 2274 rt = NULL; 2275 } 2276 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2277 nd6log((LOG_ERR, 2278 "ICMP6 redirect rejected; " 2279 "redirect dst must be unicast: %s\n", 2280 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2281 goto bad; 2282 } 2283 2284 is_router = is_onlink = 0; 2285 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2286 is_router = 1; /* router case */ 2287 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2288 is_onlink = 1; /* on-link destination case */ 2289 if (!is_router && !is_onlink) { 2290 nd6log((LOG_ERR, 2291 "ICMP6 redirect rejected; " 2292 "neither router case nor onlink case: %s\n", 2293 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2294 goto bad; 2295 } 2296 /* validation passed */ 2297 2298 icmp6len -= sizeof(*nd_rd); 2299 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2300 if (nd6_options(&ndopts) < 0) { 2301 nd6log((LOG_INFO, "icmp6_redirect_input: " 2302 "invalid ND option, rejected: %s\n", 2303 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2304 /* nd6_options have incremented stats */ 2305 goto freeit; 2306 } 2307 2308 if (ndopts.nd_opts_tgt_lladdr) { 2309 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2310 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2311 } 2312 2313 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2314 nd6log((LOG_INFO, 2315 "icmp6_redirect_input: lladdrlen mismatch for %s " 2316 "(if %d, icmp6 packet %d): %s\n", 2317 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2318 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2319 goto bad; 2320 } 2321 2322 /* RFC 2461 8.3 */ 2323 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2324 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2325 2326 if (!is_onlink) { /* better router case. perform rtredirect. */ 2327 /* perform rtredirect */ 2328 struct sockaddr_in6 sdst; 2329 struct sockaddr_in6 sgw; 2330 struct sockaddr_in6 ssrc; 2331 unsigned long rtcount; 2332 struct rtentry *newrt = NULL; 2333 2334 /* 2335 * do not install redirect route, if the number of entries 2336 * is too much (> hiwat). note that, the node (= host) will 2337 * work just fine even if we do not install redirect route 2338 * (there will be additional hops, though). 2339 */ 2340 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2341 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2342 return; 2343 else if (0 <= icmp6_redirect_lowat && 2344 rtcount > icmp6_redirect_lowat) { 2345 /* 2346 * XXX nuke a victim, install the new one. 2347 */ 2348 } 2349 2350 bzero(&sdst, sizeof(sdst)); 2351 bzero(&sgw, sizeof(sgw)); 2352 bzero(&ssrc, sizeof(ssrc)); 2353 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2354 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2355 sizeof(struct sockaddr_in6); 2356 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2357 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2358 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2359 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2360 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2361 (struct sockaddr *)&ssrc, 2362 &newrt); 2363 2364 if (newrt) { 2365 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2366 icmp6_redirect_timeout_q); 2367 rtfree(newrt); 2368 } 2369 } 2370 /* finally update cached route in each socket via pfctlinput */ 2371 { 2372 struct sockaddr_in6 sdst; 2373 2374 bzero(&sdst, sizeof(sdst)); 2375 sdst.sin6_family = AF_INET6; 2376 sdst.sin6_len = sizeof(struct sockaddr_in6); 2377 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2378 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2379 #ifdef IPSEC 2380 key_sa_routechange((struct sockaddr *)&sdst); 2381 #endif 2382 } 2383 2384 freeit: 2385 m_freem(m); 2386 return; 2387 2388 bad: 2389 icmp6stat.icp6s_badredirect++; 2390 m_freem(m); 2391 } 2392 2393 void 2394 icmp6_redirect_output(m0, rt) 2395 struct mbuf *m0; 2396 struct rtentry *rt; 2397 { 2398 struct ifnet *ifp; /* my outgoing interface */ 2399 struct in6_addr *ifp_ll6; 2400 struct in6_addr *nexthop; 2401 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2402 struct mbuf *m = NULL; /* newly allocated one */ 2403 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2404 struct nd_redirect *nd_rd; 2405 size_t maxlen; 2406 u_char *p; 2407 struct sockaddr_in6 src_sa; 2408 2409 icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2410 2411 /* if we are not router, we don't send icmp6 redirect */ 2412 if (!ip6_forwarding) 2413 goto fail; 2414 2415 /* sanity check */ 2416 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2417 goto fail; 2418 2419 /* 2420 * Address check: 2421 * the source address must identify a neighbor, and 2422 * the destination address must not be a multicast address 2423 * [RFC 2461, sec 8.2] 2424 */ 2425 sip6 = mtod(m0, struct ip6_hdr *); 2426 bzero(&src_sa, sizeof(src_sa)); 2427 src_sa.sin6_family = AF_INET6; 2428 src_sa.sin6_len = sizeof(src_sa); 2429 src_sa.sin6_addr = sip6->ip6_src; 2430 /* we don't currently use sin6_scope_id, but eventually use it */ 2431 src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src); 2432 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2433 goto fail; 2434 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2435 goto fail; /* what should we do here? */ 2436 2437 /* rate limit */ 2438 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2439 goto fail; 2440 2441 /* 2442 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2443 * we almost always ask for an mbuf cluster for simplicity. 2444 * (MHLEN < IPV6_MMTU is almost always true) 2445 */ 2446 #if IPV6_MMTU >= MCLBYTES 2447 # error assumption failed about IPV6_MMTU and MCLBYTES 2448 #endif 2449 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2450 if (m && IPV6_MMTU >= MHLEN) 2451 MCLGET(m, M_DONTWAIT); 2452 if (!m) 2453 goto fail; 2454 m->m_pkthdr.rcvif = NULL; 2455 m->m_len = 0; 2456 maxlen = M_TRAILINGSPACE(m); 2457 maxlen = min(IPV6_MMTU, maxlen); 2458 /* just for safety */ 2459 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2460 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2461 goto fail; 2462 } 2463 2464 { 2465 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2466 struct in6_ifaddr *ia; 2467 if ((ia = in6ifa_ifpforlinklocal(ifp, 2468 IN6_IFF_NOTREADY| 2469 IN6_IFF_ANYCAST)) == NULL) 2470 goto fail; 2471 ifp_ll6 = &ia->ia_addr.sin6_addr; 2472 } 2473 2474 /* get ip6 linklocal address for the router. */ 2475 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2476 struct sockaddr_in6 *sin6; 2477 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2478 nexthop = &sin6->sin6_addr; 2479 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2480 nexthop = NULL; 2481 } else 2482 nexthop = NULL; 2483 2484 /* ip6 */ 2485 ip6 = mtod(m, struct ip6_hdr *); 2486 ip6->ip6_flow = 0; 2487 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2488 ip6->ip6_vfc |= IPV6_VERSION; 2489 /* ip6->ip6_plen will be set later */ 2490 ip6->ip6_nxt = IPPROTO_ICMPV6; 2491 ip6->ip6_hlim = 255; 2492 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2493 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2494 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2495 2496 /* ND Redirect */ 2497 nd_rd = (struct nd_redirect *)(ip6 + 1); 2498 nd_rd->nd_rd_type = ND_REDIRECT; 2499 nd_rd->nd_rd_code = 0; 2500 nd_rd->nd_rd_reserved = 0; 2501 if (rt->rt_flags & RTF_GATEWAY) { 2502 /* 2503 * nd_rd->nd_rd_target must be a link-local address in 2504 * better router cases. 2505 */ 2506 if (!nexthop) 2507 goto fail; 2508 bcopy(nexthop, &nd_rd->nd_rd_target, 2509 sizeof(nd_rd->nd_rd_target)); 2510 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2511 sizeof(nd_rd->nd_rd_dst)); 2512 } else { 2513 /* make sure redtgt == reddst */ 2514 nexthop = &sip6->ip6_dst; 2515 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2516 sizeof(nd_rd->nd_rd_target)); 2517 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2518 sizeof(nd_rd->nd_rd_dst)); 2519 } 2520 2521 p = (u_char *)(nd_rd + 1); 2522 2523 { 2524 /* target lladdr option */ 2525 struct rtentry *rt_nexthop = NULL; 2526 int len; 2527 struct sockaddr_dl *sdl; 2528 struct nd_opt_hdr *nd_opt; 2529 char *lladdr; 2530 2531 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2532 if (!rt_nexthop) 2533 goto nolladdropt; 2534 len = sizeof(*nd_opt) + ifp->if_addrlen; 2535 len = (len + 7) & ~7; /* round by 8 */ 2536 /* safety check */ 2537 if (len + (p - (u_char *)ip6) > maxlen) 2538 goto nolladdropt; 2539 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2540 (rt_nexthop->rt_flags & RTF_LLINFO) && 2541 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2542 (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) && 2543 sdl->sdl_alen) { 2544 nd_opt = (struct nd_opt_hdr *)p; 2545 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2546 nd_opt->nd_opt_len = len >> 3; 2547 lladdr = (char *)(nd_opt + 1); 2548 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen); 2549 p += len; 2550 } 2551 } 2552 nolladdropt:; 2553 2554 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2555 2556 /* just to be safe */ 2557 if (m0->m_flags & M_DECRYPTED) 2558 goto noredhdropt; 2559 if (p - (u_char *)ip6 > maxlen) 2560 goto noredhdropt; 2561 2562 { 2563 /* redirected header option */ 2564 int len; 2565 struct nd_opt_rd_hdr *nd_opt_rh; 2566 2567 /* 2568 * compute the maximum size for icmp6 redirect header option. 2569 * XXX room for auth header? 2570 */ 2571 len = maxlen - (p - (u_char *)ip6); 2572 len &= ~7; 2573 2574 /* 2575 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2576 * about padding/truncate rule for the original IP packet. 2577 * From the discussion on IPv6imp in Feb 1999, 2578 * the consensus was: 2579 * - "attach as much as possible" is the goal 2580 * - pad if not aligned (original size can be guessed by 2581 * original ip6 header) 2582 * Following code adds the padding if it is simple enough, 2583 * and truncates if not. 2584 */ 2585 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2586 /* not enough room, truncate */ 2587 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2588 m0->m_pkthdr.len); 2589 } else { 2590 /* 2591 * enough room, truncate if not aligned. 2592 * we don't pad here for simplicity. 2593 */ 2594 size_t extra; 2595 2596 extra = m0->m_pkthdr.len % 8; 2597 if (extra) { 2598 /* truncate */ 2599 m_adj(m0, -extra); 2600 } 2601 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2602 } 2603 2604 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2605 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2606 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2607 nd_opt_rh->nd_opt_rh_len = len >> 3; 2608 p += sizeof(*nd_opt_rh); 2609 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2610 2611 /* connect m0 to m */ 2612 m->m_pkthdr.len += m0->m_pkthdr.len; 2613 m_cat(m, m0); 2614 m0 = NULL; 2615 } 2616 noredhdropt: 2617 if (m0) { 2618 m_freem(m0); 2619 m0 = NULL; 2620 } 2621 2622 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src)) 2623 sip6->ip6_src.s6_addr16[1] = 0; 2624 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst)) 2625 sip6->ip6_dst.s6_addr16[1] = 0; 2626 #if 0 2627 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) 2628 ip6->ip6_src.s6_addr16[1] = 0; 2629 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) 2630 ip6->ip6_dst.s6_addr16[1] = 0; 2631 #endif 2632 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target)) 2633 nd_rd->nd_rd_target.s6_addr16[1] = 0; 2634 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst)) 2635 nd_rd->nd_rd_dst.s6_addr16[1] = 0; 2636 2637 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2638 2639 nd_rd->nd_rd_cksum = 0; 2640 nd_rd->nd_rd_cksum 2641 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2642 2643 /* send the packet to outside... */ 2644 if (ip6_output(m, NULL, NULL, 0, 2645 (struct ip6_moptions *)NULL, (struct socket *)NULL, NULL) != 0) 2646 icmp6_ifstat_inc(ifp, ifs6_out_error); 2647 2648 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2649 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2650 icmp6stat.icp6s_outhist[ND_REDIRECT]++; 2651 2652 return; 2653 2654 fail: 2655 if (m) 2656 m_freem(m); 2657 if (m0) 2658 m_freem(m0); 2659 } 2660 2661 /* 2662 * ICMPv6 socket option processing. 2663 */ 2664 int 2665 icmp6_ctloutput(op, so, level, optname, mp) 2666 int op; 2667 struct socket *so; 2668 int level, optname; 2669 struct mbuf **mp; 2670 { 2671 int error = 0; 2672 int optlen; 2673 struct in6pcb *in6p = sotoin6pcb(so); 2674 struct mbuf *m = *mp; 2675 2676 optlen = m ? m->m_len : 0; 2677 2678 if (level != IPPROTO_ICMPV6) { 2679 if (op == PRCO_SETOPT && m) 2680 (void)m_free(m); 2681 return EINVAL; 2682 } 2683 2684 switch (op) { 2685 case PRCO_SETOPT: 2686 switch (optname) { 2687 case ICMP6_FILTER: 2688 { 2689 struct icmp6_filter *p; 2690 2691 if (optlen != sizeof(*p)) { 2692 error = EMSGSIZE; 2693 break; 2694 } 2695 p = mtod(m, struct icmp6_filter *); 2696 if (!p || !in6p->in6p_icmp6filt) { 2697 error = EINVAL; 2698 break; 2699 } 2700 bcopy(p, in6p->in6p_icmp6filt, 2701 sizeof(struct icmp6_filter)); 2702 error = 0; 2703 break; 2704 } 2705 2706 default: 2707 error = ENOPROTOOPT; 2708 break; 2709 } 2710 if (m) 2711 (void)m_freem(m); 2712 break; 2713 2714 case PRCO_GETOPT: 2715 switch (optname) { 2716 case ICMP6_FILTER: 2717 { 2718 struct icmp6_filter *p; 2719 2720 if (!in6p->in6p_icmp6filt) { 2721 error = EINVAL; 2722 break; 2723 } 2724 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2725 m->m_len = sizeof(struct icmp6_filter); 2726 p = mtod(m, struct icmp6_filter *); 2727 bcopy(in6p->in6p_icmp6filt, p, 2728 sizeof(struct icmp6_filter)); 2729 error = 0; 2730 break; 2731 } 2732 2733 default: 2734 error = ENOPROTOOPT; 2735 break; 2736 } 2737 break; 2738 } 2739 2740 return (error); 2741 } 2742 2743 /* 2744 * Perform rate limit check. 2745 * Returns 0 if it is okay to send the icmp6 packet. 2746 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2747 * limitation. 2748 * 2749 * XXX per-destination/type check necessary? 2750 */ 2751 static int 2752 icmp6_ratelimit(dst, type, code) 2753 const struct in6_addr *dst; /* not used at this moment */ 2754 const int type; /* not used at this moment */ 2755 const int code; /* not used at this moment */ 2756 { 2757 int ret; 2758 2759 ret = 0; /* okay to send */ 2760 2761 /* PPS limit */ 2762 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2763 icmp6errppslim)) { 2764 /* The packet is subject to rate limit */ 2765 ret++; 2766 } 2767 2768 return ret; 2769 } 2770 2771 static struct rtentry * 2772 icmp6_mtudisc_clone(dst) 2773 struct sockaddr *dst; 2774 { 2775 struct rtentry *rt; 2776 int error; 2777 2778 rt = rtalloc1(dst, 1); 2779 if (rt == 0) 2780 return NULL; 2781 2782 /* If we didn't get a host route, allocate one */ 2783 if ((rt->rt_flags & RTF_HOST) == 0) { 2784 struct rtentry *nrt; 2785 2786 error = rtrequest((int) RTM_ADD, dst, 2787 (struct sockaddr *) rt->rt_gateway, 2788 (struct sockaddr *) 0, 2789 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2790 if (error) { 2791 rtfree(rt); 2792 return NULL; 2793 } 2794 nrt->rt_rmx = rt->rt_rmx; 2795 rtfree(rt); 2796 rt = nrt; 2797 } 2798 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2799 icmp6_mtudisc_timeout_q); 2800 if (error) { 2801 rtfree(rt); 2802 return NULL; 2803 } 2804 2805 return rt; /* caller need to call rtfree() */ 2806 } 2807 2808 static void 2809 icmp6_mtudisc_timeout(rt, r) 2810 struct rtentry *rt; 2811 struct rttimer *r; 2812 { 2813 if (rt == NULL) 2814 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2815 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2816 (RTF_DYNAMIC | RTF_HOST)) { 2817 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2818 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2819 } else { 2820 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2821 rt->rt_rmx.rmx_mtu = 0; 2822 } 2823 } 2824 2825 static void 2826 icmp6_redirect_timeout(rt, r) 2827 struct rtentry *rt; 2828 struct rttimer *r; 2829 { 2830 if (rt == NULL) 2831 panic("icmp6_redirect_timeout: bad route to timeout"); 2832 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2833 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2834 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2835 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2836 } 2837 } 2838 2839 /* 2840 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes. silly? 2841 */ 2842 static int 2843 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS) 2844 { 2845 2846 if (namelen != 0) 2847 return (EINVAL); 2848 2849 return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp, 2850 (void*)newp, newlen)); 2851 } 2852 2853 SYSCTL_SETUP(sysctl_net_inet6_icmp6_setup, 2854 "sysctl net.inet6.icmp6 subtree setup") 2855 { 2856 2857 sysctl_createv(SYSCTL_PERMANENT, 2858 CTLTYPE_NODE, "net", NULL, 2859 NULL, 0, NULL, 0, 2860 CTL_NET, CTL_EOL); 2861 sysctl_createv(SYSCTL_PERMANENT, 2862 CTLTYPE_NODE, "inet6", NULL, 2863 NULL, 0, NULL, 0, 2864 CTL_NET, PF_INET6, CTL_EOL); 2865 sysctl_createv(SYSCTL_PERMANENT, 2866 CTLTYPE_NODE, "icmp6", NULL, 2867 NULL, 0, NULL, 0, 2868 CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL); 2869 2870 sysctl_createv(SYSCTL_PERMANENT, 2871 CTLTYPE_STRUCT, "stats", NULL, 2872 NULL, 0, &icmp6stat, sizeof(icmp6stat), 2873 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2874 ICMPV6CTL_STATS, CTL_EOL); 2875 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2876 CTLTYPE_INT, "rediraccept", NULL, 2877 NULL, 0, &icmp6_rediraccept, 0, 2878 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2879 ICMPV6CTL_REDIRACCEPT, CTL_EOL); 2880 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2881 CTLTYPE_INT, "redirtimeout", NULL, 2882 NULL, 0, &icmp6_redirtimeout, 0, 2883 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2884 ICMPV6CTL_REDIRTIMEOUT, CTL_EOL); 2885 #if 0 /* obsoleted */ 2886 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2887 CTLTYPE_INT, "errratelimit", NULL, 2888 NULL, 0, &icmp6_errratelimit, 0, 2889 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2890 ICMPV6CTL_ERRRATELIMIT, CTL_EOL); 2891 #endif 2892 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2893 CTLTYPE_INT, "nd6_prune", NULL, 2894 NULL, 0, &nd6_prune, 0, 2895 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2896 ICMPV6CTL_ND6_PRUNE, CTL_EOL); 2897 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2898 CTLTYPE_INT, "nd6_delay", NULL, 2899 NULL, 0, &nd6_delay, 0, 2900 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2901 ICMPV6CTL_ND6_DELAY, CTL_EOL); 2902 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2903 CTLTYPE_INT, "nd6_umaxtries", NULL, 2904 NULL, 0, &nd6_umaxtries, 0, 2905 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2906 ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL); 2907 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2908 CTLTYPE_INT, "nd6_mmaxtries", NULL, 2909 NULL, 0, &nd6_mmaxtries, 0, 2910 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2911 ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL); 2912 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2913 CTLTYPE_INT, "nd6_useloopback", NULL, 2914 NULL, 0, &nd6_useloopback, 0, 2915 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2916 ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL); 2917 #if 0 /* obsoleted */ 2918 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2919 CTLTYPE_INT, "nd6_proxyall", NULL, 2920 NULL, 0, &nd6_proxyall, 0, 2921 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2922 ICMPV6CTL_ND6_PROXYALL, CTL_EOL); 2923 #endif 2924 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2925 CTLTYPE_INT, "nodeinfo", NULL, 2926 NULL, 0, &icmp6_nodeinfo, 0, 2927 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2928 ICMPV6CTL_NODEINFO, CTL_EOL); 2929 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2930 CTLTYPE_INT, "errppslimit", NULL, 2931 NULL, 0, &icmp6errppslim, 0, 2932 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2933 ICMPV6CTL_ERRPPSLIMIT, CTL_EOL); 2934 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2935 CTLTYPE_INT, "nd6_maxnudhint", NULL, 2936 NULL, 0, &nd6_maxnudhint, 0, 2937 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2938 ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL); 2939 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2940 CTLTYPE_INT, "mtudisc_hiwat", NULL, 2941 NULL, 0, &icmp6_mtudisc_hiwat, 0, 2942 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2943 ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL); 2944 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2945 CTLTYPE_INT, "mtudisc_lowat", NULL, 2946 NULL, 0, &icmp6_mtudisc_lowat, 0, 2947 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2948 ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL); 2949 sysctl_createv(SYSCTL_PERMANENT|SYSCTL_READWRITE, 2950 CTLTYPE_INT, "nd6_debug", NULL, 2951 NULL, 0, &nd6_debug, 0, 2952 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2953 ICMPV6CTL_ND6_DEBUG, CTL_EOL); 2954 sysctl_createv(SYSCTL_PERMANENT, 2955 CTLTYPE_STRUCT, "nd6_drlist", NULL, 2956 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2957 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2958 ICMPV6CTL_ND6_DRLIST, CTL_EOL); 2959 sysctl_createv(SYSCTL_PERMANENT, 2960 CTLTYPE_STRUCT, "nd6_prlist", NULL, 2961 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2962 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2963 ICMPV6CTL_ND6_PRLIST, CTL_EOL); 2964 } 2965