1 /* $OpenBSD: icmp6.c,v 1.73 2003/06/24 07:55:12 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. 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/param.h> 65 #include <sys/systm.h> 66 #include <sys/malloc.h> 67 #include <sys/mbuf.h> 68 #include <sys/protosw.h> 69 #include <sys/socket.h> 70 #include <sys/socketvar.h> 71 #include <sys/time.h> 72 #include <sys/kernel.h> 73 #include <sys/syslog.h> 74 #include <sys/domain.h> 75 76 #include <net/if.h> 77 #include <net/route.h> 78 #include <net/if_dl.h> 79 #include <net/if_types.h> 80 81 #include <netinet/in.h> 82 #include <netinet/in_var.h> 83 #include <netinet/in_systm.h> 84 #include <netinet/ip.h> 85 #include <netinet/ip6.h> 86 #include <netinet6/ip6_var.h> 87 #include <netinet/icmp6.h> 88 #include <netinet6/mld6_var.h> 89 #include <netinet/in_pcb.h> 90 #include <netinet6/nd6.h> 91 #include <netinet6/in6_ifattach.h> 92 #include <netinet6/ip6protosw.h> 93 94 #include "faith.h" 95 96 /* inpcb members */ 97 #define in6pcb inpcb 98 #define in6p_laddr inp_laddr6 99 #define in6p_faddr inp_faddr6 100 #define in6p_icmp6filt inp_icmp6filt 101 #define in6p_route inp_route 102 #define in6p_socket inp_socket 103 #define in6p_flags inp_flags 104 #define in6p_moptions inp_moptions6 105 #define in6p_outputopts inp_outputopts6 106 #define in6p_ip6 inp_ipv6 107 #define in6p_flowinfo inp_flowinfo 108 #define in6p_sp inp_sp 109 #define in6p_next inp_next 110 #define in6p_prev inp_prev 111 /* macro names */ 112 #define sotoin6pcb sotoinpcb 113 /* function names */ 114 #define in6_pcbdetach in_pcbdetach 115 #define in6_rtchange in_rtchange 116 117 /* 118 * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from 119 * others... 120 */ 121 #define in6p_ip6_nxt inp_ipv6.ip6_nxt 122 123 extern struct domain inet6domain; 124 extern struct ip6protosw inet6sw[]; 125 extern u_char ip6_protox[]; 126 127 struct icmp6stat icmp6stat; 128 129 extern struct inpcbtable rawin6pcbtable; 130 extern int icmp6errppslim; 131 static int icmp6errpps_count = 0; 132 static struct timeval icmp6errppslim_last; 133 extern int icmp6_nodeinfo; 134 135 /* 136 * List of callbacks to notify when Path MTU changes are made. 137 */ 138 struct icmp6_mtudisc_callback { 139 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 140 void (*mc_func)(struct in6_addr *); 141 }; 142 143 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 144 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 145 146 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 147 extern int pmtu_expire; 148 149 /* XXX do these values make any sense? */ 150 static int icmp6_mtudisc_hiwat = 1280; 151 static int icmp6_mtudisc_lowat = 256; 152 153 /* 154 * keep track of # of redirect routes. 155 */ 156 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 157 158 /* XXX experimental, turned off */ 159 static int icmp6_redirect_hiwat = -1; 160 static int icmp6_redirect_lowat = -1; 161 162 static void icmp6_errcount(struct icmp6errstat *, int, int); 163 static int icmp6_rip6_input(struct mbuf **, int); 164 static int icmp6_ratelimit(const struct in6_addr *, const int, const int); 165 static const char *icmp6_redirect_diag(struct in6_addr *, 166 struct in6_addr *, struct in6_addr *); 167 static struct mbuf *ni6_input(struct mbuf *, int); 168 static struct mbuf *ni6_nametodns(const char *, int, int); 169 static int ni6_dnsmatch(const char *, int, const char *, int); 170 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *, 171 struct ifnet **, char *); 172 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 173 struct ifnet *, int); 174 static int icmp6_notify_error(struct mbuf *, int, int, int); 175 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *); 176 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *); 177 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *); 178 179 void 180 icmp6_init() 181 { 182 mld6_init(); 183 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 184 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 185 } 186 187 static void 188 icmp6_errcount(stat, type, code) 189 struct icmp6errstat *stat; 190 int type, code; 191 { 192 switch (type) { 193 case ICMP6_DST_UNREACH: 194 switch (code) { 195 case ICMP6_DST_UNREACH_NOROUTE: 196 stat->icp6errs_dst_unreach_noroute++; 197 return; 198 case ICMP6_DST_UNREACH_ADMIN: 199 stat->icp6errs_dst_unreach_admin++; 200 return; 201 case ICMP6_DST_UNREACH_BEYONDSCOPE: 202 stat->icp6errs_dst_unreach_beyondscope++; 203 return; 204 case ICMP6_DST_UNREACH_ADDR: 205 stat->icp6errs_dst_unreach_addr++; 206 return; 207 case ICMP6_DST_UNREACH_NOPORT: 208 stat->icp6errs_dst_unreach_noport++; 209 return; 210 } 211 break; 212 case ICMP6_PACKET_TOO_BIG: 213 stat->icp6errs_packet_too_big++; 214 return; 215 case ICMP6_TIME_EXCEEDED: 216 switch (code) { 217 case ICMP6_TIME_EXCEED_TRANSIT: 218 stat->icp6errs_time_exceed_transit++; 219 return; 220 case ICMP6_TIME_EXCEED_REASSEMBLY: 221 stat->icp6errs_time_exceed_reassembly++; 222 return; 223 } 224 break; 225 case ICMP6_PARAM_PROB: 226 switch (code) { 227 case ICMP6_PARAMPROB_HEADER: 228 stat->icp6errs_paramprob_header++; 229 return; 230 case ICMP6_PARAMPROB_NEXTHEADER: 231 stat->icp6errs_paramprob_nextheader++; 232 return; 233 case ICMP6_PARAMPROB_OPTION: 234 stat->icp6errs_paramprob_option++; 235 return; 236 } 237 break; 238 case ND_REDIRECT: 239 stat->icp6errs_redirect++; 240 return; 241 } 242 stat->icp6errs_unknown++; 243 } 244 245 /* 246 * Register a Path MTU Discovery callback. 247 */ 248 void 249 icmp6_mtudisc_callback_register(func) 250 void (*func)(struct in6_addr *); 251 { 252 struct icmp6_mtudisc_callback *mc; 253 254 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 255 mc = LIST_NEXT(mc, mc_list)) { 256 if (mc->mc_func == func) 257 return; 258 } 259 260 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 261 if (mc == NULL) 262 panic("icmp6_mtudisc_callback_register"); 263 264 mc->mc_func = func; 265 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 266 } 267 268 /* 269 * Generate an error packet of type error in response to bad IP6 packet. 270 */ 271 void 272 icmp6_error(m, type, code, param) 273 struct mbuf *m; 274 int type, code, param; 275 { 276 struct ip6_hdr *oip6, *nip6; 277 struct icmp6_hdr *icmp6; 278 u_int preplen; 279 int off; 280 int nxt; 281 282 icmp6stat.icp6s_error++; 283 284 /* count per-type-code statistics */ 285 icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code); 286 287 if (m->m_len < sizeof(struct ip6_hdr)) { 288 m = m_pullup(m, sizeof(struct ip6_hdr)); 289 if (m == NULL) 290 return; 291 } 292 oip6 = mtod(m, struct ip6_hdr *); 293 294 /* 295 * If the destination address of the erroneous packet is a multicast 296 * address, or the packet was sent using link-layer multicast, 297 * we should basically suppress sending an error (RFC 2463, Section 298 * 2.4). 299 * We have two exceptions (the item e.2 in that section): 300 * - the Pakcet Too Big message can be sent for path MTU discovery. 301 * - the Parameter Problem Message that can be allowed an icmp6 error 302 * in the option type field. This check has been done in 303 * ip6_unknown_opt(), so we can just check the type and code. 304 */ 305 if ((m->m_flags & (M_BCAST|M_MCAST) || 306 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 307 (type != ICMP6_PACKET_TOO_BIG && 308 (type != ICMP6_PARAM_PROB || 309 code != ICMP6_PARAMPROB_OPTION))) 310 goto freeit; 311 312 /* 313 * RFC 2463, 2.4 (e.5): source address check. 314 * XXX: the case of anycast source? 315 */ 316 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 317 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 318 goto freeit; 319 320 /* 321 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 322 * don't do it. 323 */ 324 nxt = -1; 325 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 326 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 327 struct icmp6_hdr *icp; 328 329 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 330 sizeof(*icp)); 331 if (icp == NULL) { 332 icmp6stat.icp6s_tooshort++; 333 return; 334 } 335 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 336 icp->icmp6_type == ND_REDIRECT) { 337 /* 338 * ICMPv6 error 339 * Special case: for redirect (which is 340 * informational) we must not send icmp6 error. 341 */ 342 icmp6stat.icp6s_canterror++; 343 goto freeit; 344 } else { 345 /* ICMPv6 informational - send the error */ 346 } 347 } 348 else { 349 /* non-ICMPv6 - send the error */ 350 } 351 352 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 353 354 /* Finally, do rate limitation check. */ 355 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 356 icmp6stat.icp6s_toofreq++; 357 goto freeit; 358 } 359 360 /* 361 * OK, ICMP6 can be generated. 362 */ 363 364 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 365 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 366 367 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 368 M_PREPEND(m, preplen, M_DONTWAIT); 369 if (m && m->m_len < preplen) 370 m = m_pullup(m, preplen); 371 if (m == NULL) { 372 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 373 return; 374 } 375 376 nip6 = mtod(m, struct ip6_hdr *); 377 nip6->ip6_src = oip6->ip6_src; 378 nip6->ip6_dst = oip6->ip6_dst; 379 380 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src)) 381 oip6->ip6_src.s6_addr16[1] = 0; 382 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst)) 383 oip6->ip6_dst.s6_addr16[1] = 0; 384 385 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 386 icmp6->icmp6_type = type; 387 icmp6->icmp6_code = code; 388 icmp6->icmp6_pptr = htonl((u_int32_t)param); 389 390 /* 391 * icmp6_reflect() is designed to be in the input path. 392 * icmp6_error() can be called from both input and outut path, 393 * and if we are in output path rcvif could contain bogus value. 394 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 395 * information in ip header (nip6). 396 */ 397 m->m_pkthdr.rcvif = NULL; 398 399 icmp6stat.icp6s_outhist[type]++; 400 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 401 402 return; 403 404 freeit: 405 /* 406 * If we can't tell wheter or not we can generate ICMP6, free it. 407 */ 408 m_freem(m); 409 } 410 411 /* 412 * Process a received ICMP6 message. 413 */ 414 int 415 icmp6_input(mp, offp, proto) 416 struct mbuf **mp; 417 int *offp, proto; 418 { 419 struct mbuf *m = *mp, *n; 420 struct ip6_hdr *ip6, *nip6; 421 struct icmp6_hdr *icmp6, *nicmp6; 422 int off = *offp; 423 int icmp6len = m->m_pkthdr.len - *offp; 424 int code, sum, noff; 425 426 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 427 428 /* 429 * Locate icmp6 structure in mbuf, and check 430 * that not corrupted and of at least minimum length 431 */ 432 433 ip6 = mtod(m, struct ip6_hdr *); 434 if (icmp6len < sizeof(struct icmp6_hdr)) { 435 icmp6stat.icp6s_tooshort++; 436 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 437 goto freeit; 438 } 439 440 /* 441 * calculate the checksum 442 */ 443 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 444 if (icmp6 == NULL) { 445 icmp6stat.icp6s_tooshort++; 446 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 447 return IPPROTO_DONE; 448 } 449 code = icmp6->icmp6_code; 450 451 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 452 nd6log((LOG_ERR, 453 "ICMP6 checksum error(%d|%x) %s\n", 454 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 455 icmp6stat.icp6s_checksum++; 456 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 457 goto freeit; 458 } 459 460 #if defined(NFAITH) && 0 < NFAITH 461 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 462 /* 463 * Deliver very specific ICMP6 type only. 464 * This is important to deilver TOOBIG. Otherwise PMTUD 465 * will not work. 466 */ 467 switch (icmp6->icmp6_type) { 468 case ICMP6_DST_UNREACH: 469 case ICMP6_PACKET_TOO_BIG: 470 case ICMP6_TIME_EXCEEDED: 471 break; 472 default: 473 goto freeit; 474 } 475 } 476 #endif 477 478 icmp6stat.icp6s_inhist[icmp6->icmp6_type]++; 479 480 switch (icmp6->icmp6_type) { 481 case ICMP6_DST_UNREACH: 482 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 483 switch (code) { 484 case ICMP6_DST_UNREACH_NOROUTE: 485 code = PRC_UNREACH_NET; 486 break; 487 case ICMP6_DST_UNREACH_ADMIN: 488 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 489 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 490 break; 491 case ICMP6_DST_UNREACH_ADDR: 492 code = PRC_HOSTDEAD; 493 break; 494 #ifdef COMPAT_RFC1885 495 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 496 code = PRC_UNREACH_SRCFAIL; 497 break; 498 #else 499 case ICMP6_DST_UNREACH_BEYONDSCOPE: 500 /* I mean "source address was incorrect." */ 501 code = PRC_PARAMPROB; 502 break; 503 #endif 504 case ICMP6_DST_UNREACH_NOPORT: 505 code = PRC_UNREACH_PORT; 506 break; 507 default: 508 goto badcode; 509 } 510 goto deliver; 511 512 case ICMP6_PACKET_TOO_BIG: 513 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 514 if (code != 0) 515 goto badcode; 516 517 code = PRC_MSGSIZE; 518 519 /* 520 * Updating the path MTU will be done after examining 521 * intermediate extension headers. 522 */ 523 goto deliver; 524 525 case ICMP6_TIME_EXCEEDED: 526 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 527 switch (code) { 528 case ICMP6_TIME_EXCEED_TRANSIT: 529 code = PRC_TIMXCEED_INTRANS; 530 break; 531 case ICMP6_TIME_EXCEED_REASSEMBLY: 532 code = PRC_TIMXCEED_REASS; 533 break; 534 default: 535 goto badcode; 536 } 537 goto deliver; 538 539 case ICMP6_PARAM_PROB: 540 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 541 switch (code) { 542 case ICMP6_PARAMPROB_NEXTHEADER: 543 code = PRC_UNREACH_PROTOCOL; 544 break; 545 case ICMP6_PARAMPROB_HEADER: 546 case ICMP6_PARAMPROB_OPTION: 547 code = PRC_PARAMPROB; 548 break; 549 default: 550 goto badcode; 551 } 552 goto deliver; 553 554 case ICMP6_ECHO_REQUEST: 555 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 556 if (code != 0) 557 goto badcode; 558 /* 559 * Copy mbuf to send to two data paths: userland socket(s), 560 * and to the querier (echo reply). 561 * m: a copy for socket, n: a copy for querier 562 */ 563 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 564 /* Give up local */ 565 n = m; 566 m = NULL; 567 goto deliverecho; 568 } 569 /* 570 * If the first mbuf is shared, or the first mbuf is too short, 571 * copy the first part of the data into a fresh mbuf. 572 * Otherwise, we will wrongly overwrite both copies. 573 */ 574 if ((n->m_flags & M_EXT) != 0 || 575 n->m_len < off + sizeof(struct icmp6_hdr)) { 576 struct mbuf *n0 = n; 577 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6); 578 579 /* 580 * Prepare an internal mbuf. m_pullup() doesn't 581 * always copy the length we specified. 582 */ 583 if (maxlen >= MCLBYTES) { 584 /* Give up remote */ 585 m_freem(n0); 586 break; 587 } 588 MGETHDR(n, M_DONTWAIT, n0->m_type); 589 if (n && maxlen >= MHLEN) { 590 MCLGET(n, M_DONTWAIT); 591 if ((n->m_flags & M_EXT) == 0) { 592 m_free(n); 593 n = NULL; 594 } 595 } 596 if (n == NULL) { 597 /* Give up local */ 598 m_freem(n0); 599 n = m; 600 m = NULL; 601 goto deliverecho; 602 } 603 M_MOVE_PKTHDR(n, n0); 604 /* 605 * Copy IPv6 and ICMPv6 only. 606 */ 607 nip6 = mtod(n, struct ip6_hdr *); 608 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 609 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 610 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 611 noff = sizeof(struct ip6_hdr); 612 n->m_len = noff + sizeof(struct icmp6_hdr); 613 /* 614 * Adjust mbuf. ip6_plen will be adjusted in 615 * ip6_output(). 616 * n->m_pkthdr.len == n0->m_pkthdr.len at this point. 617 */ 618 n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr); 619 n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr)); 620 m_adj(n0, off + sizeof(struct icmp6_hdr)); 621 n->m_next = n0; 622 } else { 623 deliverecho: 624 nip6 = mtod(n, struct ip6_hdr *); 625 nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off); 626 noff = off; 627 } 628 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 629 nicmp6->icmp6_code = 0; 630 if (n) { 631 icmp6stat.icp6s_reflect++; 632 icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++; 633 icmp6_reflect(n, noff); 634 } 635 if (!m) 636 goto freeit; 637 break; 638 639 case ICMP6_ECHO_REPLY: 640 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 641 if (code != 0) 642 goto badcode; 643 break; 644 645 case MLD_LISTENER_QUERY: 646 case MLD_LISTENER_REPORT: 647 if (icmp6len < sizeof(struct mld_hdr)) 648 goto badlen; 649 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */ 650 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 651 else 652 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 653 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 654 /* give up local */ 655 mld6_input(m, off); 656 m = NULL; 657 goto freeit; 658 } 659 mld6_input(n, off); 660 /* m stays. */ 661 break; 662 663 case MLD_LISTENER_DONE: 664 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 665 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */ 666 goto badlen; 667 break; /* nothing to be done in kernel */ 668 669 case MLD_MTRACE_RESP: 670 case MLD_MTRACE: 671 /* XXX: these two are experimental. not officially defined. */ 672 /* XXX: per-interface statistics? */ 673 break; /* just pass it to applications */ 674 675 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 676 { 677 enum { WRU, FQDN } mode; 678 679 if (!icmp6_nodeinfo) 680 break; 681 682 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 683 mode = WRU; 684 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 685 mode = FQDN; 686 else 687 goto badlen; 688 689 if (mode == FQDN) { 690 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 691 if (n) 692 n = ni6_input(n, off); 693 /* XXX meaningless if n == NULL */ 694 noff = sizeof(struct ip6_hdr); 695 } else { 696 u_char *p; 697 int maxlen, maxhlen; 698 699 if ((icmp6_nodeinfo & 1) == 0) 700 break; 701 702 if (code != 0) 703 goto badcode; 704 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; 705 if (maxlen >= MCLBYTES) { 706 /* Give up remote */ 707 break; 708 } 709 MGETHDR(n, M_DONTWAIT, m->m_type); 710 if (n && maxlen > MHLEN) { 711 MCLGET(n, M_DONTWAIT); 712 if ((n->m_flags & M_EXT) == 0) { 713 m_free(n); 714 n = NULL; 715 } 716 } 717 if (n == NULL) { 718 /* Give up remote */ 719 break; 720 } 721 n->m_pkthdr.rcvif = NULL; 722 n->m_len = 0; 723 maxhlen = M_TRAILINGSPACE(n) - maxlen; 724 if (maxhlen > hostnamelen) 725 maxhlen = hostnamelen; 726 /* 727 * Copy IPv6 and ICMPv6 only. 728 */ 729 nip6 = mtod(n, struct ip6_hdr *); 730 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 731 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 732 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 733 p = (u_char *)(nicmp6 + 1); 734 bzero(p, 4); 735 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 736 noff = sizeof(struct ip6_hdr); 737 M_DUP_PKTHDR(n, m); /* just for rcvif */ 738 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 739 sizeof(struct icmp6_hdr) + 4 + maxhlen; 740 nicmp6->icmp6_type = ICMP6_WRUREPLY; 741 nicmp6->icmp6_code = 0; 742 } 743 #undef hostnamelen 744 if (n) { 745 icmp6stat.icp6s_reflect++; 746 icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++; 747 icmp6_reflect(n, noff); 748 } 749 break; 750 } 751 752 case ICMP6_WRUREPLY: 753 if (code != 0) 754 goto badcode; 755 break; 756 757 case ND_ROUTER_SOLICIT: 758 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 759 if (code != 0) 760 goto badcode; 761 if (icmp6len < sizeof(struct nd_router_solicit)) 762 goto badlen; 763 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 764 /* give up local */ 765 nd6_rs_input(m, off, icmp6len); 766 m = NULL; 767 goto freeit; 768 } 769 nd6_rs_input(n, off, icmp6len); 770 /* m stays. */ 771 break; 772 773 case ND_ROUTER_ADVERT: 774 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 775 if (code != 0) 776 goto badcode; 777 if (icmp6len < sizeof(struct nd_router_advert)) 778 goto badlen; 779 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 780 /* give up local */ 781 nd6_ra_input(m, off, icmp6len); 782 m = NULL; 783 goto freeit; 784 } 785 nd6_ra_input(n, off, icmp6len); 786 /* m stays. */ 787 break; 788 789 case ND_NEIGHBOR_SOLICIT: 790 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 791 if (code != 0) 792 goto badcode; 793 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 794 goto badlen; 795 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 796 /* give up local */ 797 nd6_ns_input(m, off, icmp6len); 798 m = NULL; 799 goto freeit; 800 } 801 nd6_ns_input(n, off, icmp6len); 802 /* m stays. */ 803 break; 804 805 case ND_NEIGHBOR_ADVERT: 806 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 807 if (code != 0) 808 goto badcode; 809 if (icmp6len < sizeof(struct nd_neighbor_advert)) 810 goto badlen; 811 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 812 /* give up local */ 813 nd6_na_input(m, off, icmp6len); 814 m = NULL; 815 goto freeit; 816 } 817 nd6_na_input(n, off, icmp6len); 818 /* m stays. */ 819 break; 820 821 case ND_REDIRECT: 822 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 823 if (code != 0) 824 goto badcode; 825 if (icmp6len < sizeof(struct nd_redirect)) 826 goto badlen; 827 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 828 /* give up local */ 829 icmp6_redirect_input(m, off); 830 m = NULL; 831 goto freeit; 832 } 833 icmp6_redirect_input(n, off); 834 /* m stays. */ 835 break; 836 837 case ICMP6_ROUTER_RENUMBERING: 838 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 839 code != ICMP6_ROUTER_RENUMBERING_RESULT) 840 goto badcode; 841 if (icmp6len < sizeof(struct icmp6_router_renum)) 842 goto badlen; 843 break; 844 845 default: 846 nd6log((LOG_DEBUG, 847 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 848 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 849 ip6_sprintf(&ip6->ip6_dst), 850 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 851 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 852 /* ICMPv6 error: MUST deliver it by spec... */ 853 code = PRC_NCMDS; 854 /* deliver */ 855 } else { 856 /* ICMPv6 informational: MUST not deliver */ 857 break; 858 } 859 deliver: 860 if (icmp6_notify_error(m, off, icmp6len, code)) { 861 /* In this case, m should've been freed. */ 862 return (IPPROTO_DONE); 863 } 864 break; 865 866 badcode: 867 icmp6stat.icp6s_badcode++; 868 break; 869 870 badlen: 871 icmp6stat.icp6s_badlen++; 872 break; 873 } 874 875 /* deliver the packet to appropriate sockets */ 876 icmp6_rip6_input(&m, *offp); 877 878 return IPPROTO_DONE; 879 880 freeit: 881 m_freem(m); 882 return IPPROTO_DONE; 883 } 884 885 static int 886 icmp6_notify_error(m, off, icmp6len, code) 887 struct mbuf *m; 888 int off, icmp6len; 889 { 890 struct icmp6_hdr *icmp6; 891 struct ip6_hdr *eip6; 892 u_int32_t notifymtu; 893 struct sockaddr_in6 icmp6src, icmp6dst; 894 895 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 896 icmp6stat.icp6s_tooshort++; 897 goto freeit; 898 } 899 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 900 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 901 if (icmp6 == NULL) { 902 icmp6stat.icp6s_tooshort++; 903 return (-1); 904 } 905 eip6 = (struct ip6_hdr *)(icmp6 + 1); 906 907 /* Detect the upper level protocol */ 908 { 909 void (*ctlfunc)(int, struct sockaddr *, void *); 910 u_int8_t nxt = eip6->ip6_nxt; 911 int eoff = off + sizeof(struct icmp6_hdr) + 912 sizeof(struct ip6_hdr); 913 struct ip6ctlparam ip6cp; 914 struct in6_addr *finaldst = NULL; 915 int icmp6type = icmp6->icmp6_type; 916 struct ip6_frag *fh; 917 struct ip6_rthdr *rth; 918 struct ip6_rthdr0 *rth0; 919 int rthlen; 920 921 while (1) { /* XXX: should avoid infinite loop explicitly? */ 922 struct ip6_ext *eh; 923 924 switch (nxt) { 925 case IPPROTO_HOPOPTS: 926 case IPPROTO_DSTOPTS: 927 case IPPROTO_AH: 928 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 929 eoff, sizeof(*eh)); 930 if (eh == NULL) { 931 icmp6stat.icp6s_tooshort++; 932 return (-1); 933 } 934 935 if (nxt == IPPROTO_AH) 936 eoff += (eh->ip6e_len + 2) << 2; 937 else 938 eoff += (eh->ip6e_len + 1) << 3; 939 nxt = eh->ip6e_nxt; 940 break; 941 case IPPROTO_ROUTING: 942 /* 943 * When the erroneous packet contains a 944 * routing header, we should examine the 945 * header to determine the final destination. 946 * Otherwise, we can't properly update 947 * information that depends on the final 948 * destination (e.g. path MTU). 949 */ 950 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 951 eoff, sizeof(*rth)); 952 if (rth == NULL) { 953 icmp6stat.icp6s_tooshort++; 954 return (-1); 955 } 956 rthlen = (rth->ip6r_len + 1) << 3; 957 /* 958 * XXX: currently there is no 959 * officially defined type other 960 * than type-0. 961 * Note that if the segment left field 962 * is 0, all intermediate hops must 963 * have been passed. 964 */ 965 if (rth->ip6r_segleft && 966 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 967 int hops; 968 969 IP6_EXTHDR_GET(rth0, 970 struct ip6_rthdr0 *, m, 971 eoff, rthlen); 972 if (rth0 == NULL) { 973 icmp6stat.icp6s_tooshort++; 974 return (-1); 975 } 976 /* just ignore a bogus header */ 977 if ((rth0->ip6r0_len % 2) == 0 && 978 (hops = rth0->ip6r0_len/2)) 979 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 980 } 981 eoff += rthlen; 982 nxt = rth->ip6r_nxt; 983 break; 984 case IPPROTO_FRAGMENT: 985 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 986 eoff, sizeof(*fh)); 987 if (fh == NULL) { 988 icmp6stat.icp6s_tooshort++; 989 return (-1); 990 } 991 /* 992 * Data after a fragment header is meaningless 993 * unless it is the first fragment, but 994 * we'll go to the notify label for path MTU 995 * discovery. 996 */ 997 if (fh->ip6f_offlg & IP6F_OFF_MASK) 998 goto notify; 999 1000 eoff += sizeof(struct ip6_frag); 1001 nxt = fh->ip6f_nxt; 1002 break; 1003 default: 1004 /* 1005 * This case includes ESP and the No Next 1006 * Header. In such cases going to the notify 1007 * label does not have any meaning 1008 * (i.e. ctlfunc will be NULL), but we go 1009 * anyway since we might have to update 1010 * path MTU information. 1011 */ 1012 goto notify; 1013 } 1014 } 1015 notify: 1016 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1017 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1018 if (icmp6 == NULL) { 1019 icmp6stat.icp6s_tooshort++; 1020 return (-1); 1021 } 1022 1023 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1024 bzero(&icmp6dst, sizeof(icmp6dst)); 1025 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1026 icmp6dst.sin6_family = AF_INET6; 1027 if (finaldst == NULL) 1028 icmp6dst.sin6_addr = eip6->ip6_dst; 1029 else 1030 icmp6dst.sin6_addr = *finaldst; 1031 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1032 &icmp6dst.sin6_addr); 1033 #ifndef SCOPEDROUTING 1034 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst, 1035 NULL, NULL)) { 1036 /* should be impossbile */ 1037 nd6log((LOG_DEBUG, 1038 "icmp6_notify_error: in6_embedscope failed\n")); 1039 goto freeit; 1040 } 1041 #endif 1042 1043 /* 1044 * retrieve parameters from the inner IPv6 header, and convert 1045 * them into sockaddr structures. 1046 */ 1047 bzero(&icmp6src, sizeof(icmp6src)); 1048 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1049 icmp6src.sin6_family = AF_INET6; 1050 icmp6src.sin6_addr = eip6->ip6_src; 1051 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1052 &icmp6src.sin6_addr); 1053 #ifndef SCOPEDROUTING 1054 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src, 1055 NULL, NULL)) { 1056 /* should be impossbile */ 1057 nd6log((LOG_DEBUG, 1058 "icmp6_notify_error: in6_embedscope failed\n")); 1059 goto freeit; 1060 } 1061 #endif 1062 icmp6src.sin6_flowinfo = 1063 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1064 1065 if (finaldst == NULL) 1066 finaldst = &eip6->ip6_dst; 1067 ip6cp.ip6c_m = m; 1068 ip6cp.ip6c_icmp6 = icmp6; 1069 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1070 ip6cp.ip6c_off = eoff; 1071 ip6cp.ip6c_finaldst = finaldst; 1072 ip6cp.ip6c_src = &icmp6src; 1073 ip6cp.ip6c_nxt = nxt; 1074 1075 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1076 notifymtu = ntohl(icmp6->icmp6_mtu); 1077 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1078 } 1079 1080 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1081 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1082 if (ctlfunc) { 1083 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1084 &ip6cp); 1085 } 1086 } 1087 return (0); 1088 1089 freeit: 1090 m_freem(m); 1091 return (-1); 1092 } 1093 1094 void 1095 icmp6_mtudisc_update(ip6cp, validated) 1096 struct ip6ctlparam *ip6cp; 1097 int validated; 1098 { 1099 unsigned long rtcount; 1100 struct icmp6_mtudisc_callback *mc; 1101 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1102 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1103 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1104 u_int mtu = ntohl(icmp6->icmp6_mtu); 1105 struct rtentry *rt = NULL; 1106 struct sockaddr_in6 sin6; 1107 1108 /* 1109 * allow non-validated cases if memory is plenty, to make traffic 1110 * from non-connected pcb happy. 1111 */ 1112 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1113 if (validated) { 1114 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1115 return; 1116 else if (0 <= icmp6_mtudisc_lowat && 1117 rtcount > icmp6_mtudisc_lowat) { 1118 /* 1119 * XXX nuke a victim, install the new one. 1120 */ 1121 } 1122 } else { 1123 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1124 return; 1125 } 1126 1127 bzero(&sin6, sizeof(sin6)); 1128 sin6.sin6_family = PF_INET6; 1129 sin6.sin6_len = sizeof(struct sockaddr_in6); 1130 sin6.sin6_addr = *dst; 1131 /* XXX normally, this won't happen */ 1132 if (IN6_IS_ADDR_LINKLOCAL(dst)) { 1133 sin6.sin6_addr.s6_addr16[1] = 1134 htons(m->m_pkthdr.rcvif->if_index); 1135 } 1136 /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */ 1137 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1138 1139 if (rt && (rt->rt_flags & RTF_HOST) && 1140 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1141 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1142 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1143 icmp6stat.icp6s_pmtuchg++; 1144 rt->rt_rmx.rmx_mtu = mtu; 1145 } 1146 } 1147 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */ 1148 RTFREE(rt); 1149 } 1150 1151 /* 1152 * Notify protocols that the MTU for this destination 1153 * has changed. 1154 */ 1155 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1156 mc = LIST_NEXT(mc, mc_list)) 1157 (*mc->mc_func)(&sin6.sin6_addr); 1158 } 1159 1160 /* 1161 * Process a Node Information Query packet, based on 1162 * draft-ietf-ipngwg-icmp-name-lookups-07. 1163 * 1164 * Spec incompatibilities: 1165 * - IPv6 Subject address handling 1166 * - IPv4 Subject address handling support missing 1167 * - Proxy reply (answer even if it's not for me) 1168 * - joins NI group address at in6_ifattach() time only, does not cope 1169 * with hostname changes by sethostname(3) 1170 */ 1171 #ifndef offsetof /* XXX */ 1172 #define offsetof(type, member) ((size_t)(&((type *)0)->member)) 1173 #endif 1174 static struct mbuf * 1175 ni6_input(m, off) 1176 struct mbuf *m; 1177 int off; 1178 { 1179 struct icmp6_nodeinfo *ni6, *nni6; 1180 struct mbuf *n = NULL; 1181 u_int16_t qtype; 1182 int subjlen; 1183 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1184 struct ni_reply_fqdn *fqdn; 1185 int addrs; /* for NI_QTYPE_NODEADDR */ 1186 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1187 struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */ 1188 struct ip6_hdr *ip6; 1189 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1190 char *subj = NULL; 1191 1192 ip6 = mtod(m, struct ip6_hdr *); 1193 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1194 if (ni6 == NULL) { 1195 /* m is already reclaimed */ 1196 return NULL; 1197 } 1198 1199 /* 1200 * Validate IPv6 destination address. 1201 * 1202 * The Responder must discard the Query without further processing 1203 * unless it is one of the Responder's unicast or anycast addresses, or 1204 * a link-local scope multicast address which the Responder has joined. 1205 * [icmp-name-lookups-07, Section 4.] 1206 */ 1207 bzero(&sin6, sizeof(sin6)); 1208 sin6.sin6_family = AF_INET6; 1209 sin6.sin6_len = sizeof(struct sockaddr_in6); 1210 bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr)); 1211 /* XXX scopeid */ 1212 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1213 ; /* unicast/anycast, fine */ 1214 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1215 ; /* link-local multicast, fine */ 1216 else 1217 goto bad; 1218 1219 /* validate query Subject field. */ 1220 qtype = ntohs(ni6->ni_qtype); 1221 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1222 switch (qtype) { 1223 case NI_QTYPE_NOOP: 1224 case NI_QTYPE_SUPTYPES: 1225 /* 07 draft */ 1226 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1227 break; 1228 /* FALLTHROUGH */ 1229 case NI_QTYPE_FQDN: 1230 case NI_QTYPE_NODEADDR: 1231 switch (ni6->ni_code) { 1232 case ICMP6_NI_SUBJ_IPV6: 1233 #if ICMP6_NI_SUBJ_IPV6 != 0 1234 case 0: 1235 #endif 1236 /* 1237 * backward compatibility - try to accept 03 draft 1238 * format, where no Subject is present. 1239 */ 1240 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1241 subjlen == 0) { 1242 oldfqdn++; 1243 break; 1244 } 1245 #if ICMP6_NI_SUBJ_IPV6 != 0 1246 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1247 goto bad; 1248 #endif 1249 1250 if (subjlen != sizeof(sin6.sin6_addr)) 1251 goto bad; 1252 1253 /* 1254 * Validate Subject address. 1255 * 1256 * Not sure what exactly "address belongs to the node" 1257 * means in the spec, is it just unicast, or what? 1258 * 1259 * At this moment we consider Subject address as 1260 * "belong to the node" if the Subject address equals 1261 * to the IPv6 destination address; validation for 1262 * IPv6 destination address should have done enough 1263 * check for us. 1264 * 1265 * We do not do proxy at this moment. 1266 */ 1267 /* m_pulldown instead of copy? */ 1268 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1269 subjlen, (caddr_t)&sin6.sin6_addr); 1270 /* XXX kame scope hack */ 1271 if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) { 1272 if ((m->m_flags & M_PKTHDR) != 0 && 1273 m->m_pkthdr.rcvif) { 1274 sin6.sin6_addr.s6_addr16[1] = 1275 htons(m->m_pkthdr.rcvif->if_index); 1276 } 1277 } 1278 subj = (char *)&sin6; 1279 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr)) 1280 break; 1281 1282 /* 1283 * XXX if we are to allow other cases, we should really 1284 * be careful about scope here. 1285 * basically, we should disallow queries toward IPv6 1286 * destination X with subject Y, if scope(X) > scope(Y). 1287 * if we allow scope(X) > scope(Y), it will result in 1288 * information leakage across scope boundary. 1289 */ 1290 goto bad; 1291 1292 case ICMP6_NI_SUBJ_FQDN: 1293 /* 1294 * Validate Subject name with gethostname(3). 1295 * 1296 * The behavior may need some debate, since: 1297 * - we are not sure if the node has FQDN as 1298 * hostname (returned by gethostname(3)). 1299 * - the code does wildcard match for truncated names. 1300 * however, we are not sure if we want to perform 1301 * wildcard match, if gethostname(3) side has 1302 * truncated hostname. 1303 */ 1304 n = ni6_nametodns(hostname, hostnamelen, 0); 1305 if (!n || n->m_next || n->m_len == 0) 1306 goto bad; 1307 IP6_EXTHDR_GET(subj, char *, m, 1308 off + sizeof(struct icmp6_nodeinfo), subjlen); 1309 if (subj == NULL) 1310 goto bad; 1311 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1312 n->m_len)) { 1313 goto bad; 1314 } 1315 m_freem(n); 1316 n = NULL; 1317 break; 1318 1319 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1320 default: 1321 goto bad; 1322 } 1323 break; 1324 } 1325 1326 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1327 switch (qtype) { 1328 case NI_QTYPE_FQDN: 1329 if ((icmp6_nodeinfo & 1) == 0) 1330 goto bad; 1331 break; 1332 case NI_QTYPE_NODEADDR: 1333 if ((icmp6_nodeinfo & 2) == 0) 1334 goto bad; 1335 break; 1336 } 1337 1338 /* guess reply length */ 1339 switch (qtype) { 1340 case NI_QTYPE_NOOP: 1341 break; /* no reply data */ 1342 case NI_QTYPE_SUPTYPES: 1343 replylen += sizeof(u_int32_t); 1344 break; 1345 case NI_QTYPE_FQDN: 1346 /* XXX will append an mbuf */ 1347 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1348 break; 1349 case NI_QTYPE_NODEADDR: 1350 addrs = ni6_addrs(ni6, m, &ifp, subj); 1351 if ((replylen += addrs * (sizeof(struct in6_addr) + 1352 sizeof(u_int32_t))) > MCLBYTES) 1353 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1354 break; 1355 default: 1356 /* 1357 * XXX: We must return a reply with the ICMP6 code 1358 * `unknown Qtype' in this case. However we regard the case 1359 * as an FQDN query for backward compatibility. 1360 * Older versions set a random value to this field, 1361 * so it rarely varies in the defined qtypes. 1362 * But the mechanism is not reliable... 1363 * maybe we should obsolete older versions. 1364 */ 1365 qtype = NI_QTYPE_FQDN; 1366 /* XXX will append an mbuf */ 1367 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1368 oldfqdn++; 1369 break; 1370 } 1371 1372 /* allocate an mbuf to reply. */ 1373 MGETHDR(n, M_DONTWAIT, m->m_type); 1374 if (n == NULL) { 1375 m_freem(m); 1376 return (NULL); 1377 } 1378 M_DUP_PKTHDR(n, m); /* just for rcvif */ 1379 if (replylen > MHLEN) { 1380 if (replylen > MCLBYTES) { 1381 /* 1382 * XXX: should we try to allocate more? But MCLBYTES 1383 * is probably much larger than IPV6_MMTU... 1384 */ 1385 goto bad; 1386 } 1387 MCLGET(n, M_DONTWAIT); 1388 if ((n->m_flags & M_EXT) == 0) { 1389 goto bad; 1390 } 1391 } 1392 n->m_pkthdr.len = n->m_len = replylen; 1393 1394 /* copy mbuf header and IPv6 + Node Information base headers */ 1395 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1396 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1397 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1398 1399 /* qtype dependent procedure */ 1400 switch (qtype) { 1401 case NI_QTYPE_NOOP: 1402 nni6->ni_code = ICMP6_NI_SUCCESS; 1403 nni6->ni_flags = 0; 1404 break; 1405 case NI_QTYPE_SUPTYPES: 1406 { 1407 u_int32_t v; 1408 nni6->ni_code = ICMP6_NI_SUCCESS; 1409 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1410 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1411 v = (u_int32_t)htonl(0x0000000f); 1412 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1413 break; 1414 } 1415 case NI_QTYPE_FQDN: 1416 nni6->ni_code = ICMP6_NI_SUCCESS; 1417 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1418 sizeof(struct ip6_hdr) + 1419 sizeof(struct icmp6_nodeinfo)); 1420 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1421 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1422 /* 1423 * XXX do we really have FQDN in variable "hostname"? 1424 */ 1425 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1426 if (n->m_next == NULL) 1427 goto bad; 1428 /* XXX we assume that n->m_next is not a chain */ 1429 if (n->m_next->m_next != NULL) 1430 goto bad; 1431 n->m_pkthdr.len += n->m_next->m_len; 1432 break; 1433 case NI_QTYPE_NODEADDR: 1434 { 1435 int lenlim, copied; 1436 1437 nni6->ni_code = ICMP6_NI_SUCCESS; 1438 n->m_pkthdr.len = n->m_len = 1439 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1440 lenlim = M_TRAILINGSPACE(n); 1441 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1442 /* XXX: reset mbuf length */ 1443 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1444 sizeof(struct icmp6_nodeinfo) + copied; 1445 break; 1446 } 1447 default: 1448 break; /* XXX impossible! */ 1449 } 1450 1451 nni6->ni_type = ICMP6_NI_REPLY; 1452 m_freem(m); 1453 return (n); 1454 1455 bad: 1456 m_freem(m); 1457 if (n) 1458 m_freem(n); 1459 return (NULL); 1460 } 1461 #undef hostnamelen 1462 1463 #define isupper(x) ('A' <= (x) && (x) <= 'Z') 1464 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1465 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1466 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1467 1468 /* 1469 * make a mbuf with DNS-encoded string. no compression support. 1470 * 1471 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1472 * treated as truncated name (two \0 at the end). this is a wild guess. 1473 */ 1474 static struct mbuf * 1475 ni6_nametodns(name, namelen, old) 1476 const char *name; 1477 int namelen; 1478 int old; /* return pascal string if non-zero */ 1479 { 1480 struct mbuf *m; 1481 char *cp, *ep; 1482 const char *p, *q; 1483 int i, len, nterm; 1484 1485 if (old) 1486 len = namelen + 1; 1487 else 1488 len = MCLBYTES; 1489 1490 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1491 MGET(m, M_DONTWAIT, MT_DATA); 1492 if (m && len > MLEN) { 1493 MCLGET(m, M_DONTWAIT); 1494 if ((m->m_flags & M_EXT) == 0) 1495 goto fail; 1496 } 1497 if (!m) 1498 goto fail; 1499 m->m_next = NULL; 1500 1501 if (old) { 1502 m->m_len = len; 1503 *mtod(m, char *) = namelen; 1504 bcopy(name, mtod(m, char *) + 1, namelen); 1505 return m; 1506 } else { 1507 m->m_len = 0; 1508 cp = mtod(m, char *); 1509 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1510 1511 /* if not certain about my name, return empty buffer */ 1512 if (namelen == 0) 1513 return m; 1514 1515 /* 1516 * guess if it looks like shortened hostname, or FQDN. 1517 * shortened hostname needs two trailing "\0". 1518 */ 1519 i = 0; 1520 for (p = name; p < name + namelen; p++) { 1521 if (*p && *p == '.') 1522 i++; 1523 } 1524 if (i < 2) 1525 nterm = 2; 1526 else 1527 nterm = 1; 1528 1529 p = name; 1530 while (cp < ep && p < name + namelen) { 1531 i = 0; 1532 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1533 i++; 1534 /* result does not fit into mbuf */ 1535 if (cp + i + 1 >= ep) 1536 goto fail; 1537 /* 1538 * DNS label length restriction, RFC1035 page 8. 1539 * "i == 0" case is included here to avoid returning 1540 * 0-length label on "foo..bar". 1541 */ 1542 if (i <= 0 || i >= 64) 1543 goto fail; 1544 *cp++ = i; 1545 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1546 goto fail; 1547 while (i > 0) { 1548 if (!isalnum(*p) && *p != '-') 1549 goto fail; 1550 if (isupper(*p)) { 1551 *cp++ = tolower(*p); 1552 p++; 1553 } else 1554 *cp++ = *p++; 1555 i--; 1556 } 1557 p = q; 1558 if (p < name + namelen && *p == '.') 1559 p++; 1560 } 1561 /* termination */ 1562 if (cp + nterm >= ep) 1563 goto fail; 1564 while (nterm-- > 0) 1565 *cp++ = '\0'; 1566 m->m_len = cp - mtod(m, char *); 1567 return m; 1568 } 1569 1570 panic("should not reach here"); 1571 /* NOTREACHED */ 1572 1573 fail: 1574 if (m) 1575 m_freem(m); 1576 return NULL; 1577 } 1578 1579 /* 1580 * check if two DNS-encoded string matches. takes care of truncated 1581 * form (with \0\0 at the end). no compression support. 1582 * XXX upper/lowercase match (see RFC2065) 1583 */ 1584 static int 1585 ni6_dnsmatch(a, alen, b, blen) 1586 const char *a; 1587 int alen; 1588 const char *b; 1589 int blen; 1590 { 1591 const char *a0, *b0; 1592 int l; 1593 1594 /* simplest case - need validation? */ 1595 if (alen == blen && bcmp(a, b, alen) == 0) 1596 return 1; 1597 1598 a0 = a; 1599 b0 = b; 1600 1601 /* termination is mandatory */ 1602 if (alen < 2 || blen < 2) 1603 return 0; 1604 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1605 return 0; 1606 alen--; 1607 blen--; 1608 1609 while (a - a0 < alen && b - b0 < blen) { 1610 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1611 return 0; 1612 1613 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1614 return 0; 1615 /* we don't support compression yet */ 1616 if (a[0] >= 64 || b[0] >= 64) 1617 return 0; 1618 1619 /* truncated case */ 1620 if (a[0] == 0 && a - a0 == alen - 1) 1621 return 1; 1622 if (b[0] == 0 && b - b0 == blen - 1) 1623 return 1; 1624 if (a[0] == 0 || b[0] == 0) 1625 return 0; 1626 1627 if (a[0] != b[0]) 1628 return 0; 1629 l = a[0]; 1630 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1631 return 0; 1632 if (bcmp(a + 1, b + 1, l) != 0) 1633 return 0; 1634 1635 a += 1 + l; 1636 b += 1 + l; 1637 } 1638 1639 if (a - a0 == alen && b - b0 == blen) 1640 return 1; 1641 else 1642 return 0; 1643 } 1644 1645 /* 1646 * calculate the number of addresses to be returned in the node info reply. 1647 */ 1648 static int 1649 ni6_addrs(ni6, m, ifpp, subj) 1650 struct icmp6_nodeinfo *ni6; 1651 struct mbuf *m; 1652 struct ifnet **ifpp; 1653 char *subj; 1654 { 1655 struct ifnet *ifp; 1656 struct in6_ifaddr *ifa6; 1657 struct ifaddr *ifa; 1658 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1659 int addrs = 0, addrsofif, iffound = 0; 1660 int niflags = ni6->ni_flags; 1661 1662 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1663 switch (ni6->ni_code) { 1664 case ICMP6_NI_SUBJ_IPV6: 1665 if (subj == NULL) /* must be impossible... */ 1666 return (0); 1667 subj_ip6 = (struct sockaddr_in6 *)subj; 1668 break; 1669 default: 1670 /* 1671 * XXX: we only support IPv6 subject address for 1672 * this Qtype. 1673 */ 1674 return (0); 1675 } 1676 } 1677 1678 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1679 { 1680 addrsofif = 0; 1681 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1682 ifa = ifa->ifa_list.tqe_next) 1683 { 1684 if (ifa->ifa_addr->sa_family != AF_INET6) 1685 continue; 1686 ifa6 = (struct in6_ifaddr *)ifa; 1687 1688 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1689 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1690 &ifa6->ia_addr.sin6_addr)) 1691 iffound = 1; 1692 1693 /* 1694 * IPv4-mapped addresses can only be returned by a 1695 * Node Information proxy, since they represent 1696 * addresses of IPv4-only nodes, which perforce do 1697 * not implement this protocol. 1698 * [icmp-name-lookups-07, Section 5.4] 1699 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1700 * this function at this moment. 1701 */ 1702 1703 /* What do we have to do about ::1? */ 1704 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1705 case IPV6_ADDR_SCOPE_LINKLOCAL: 1706 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1707 continue; 1708 break; 1709 case IPV6_ADDR_SCOPE_SITELOCAL: 1710 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1711 continue; 1712 break; 1713 case IPV6_ADDR_SCOPE_GLOBAL: 1714 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1715 continue; 1716 break; 1717 default: 1718 continue; 1719 } 1720 1721 /* 1722 * check if anycast is okay. 1723 * XXX: just experimental. not in the spec. 1724 */ 1725 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1726 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1727 continue; /* we need only unicast addresses */ 1728 1729 addrsofif++; /* count the address */ 1730 } 1731 if (iffound) { 1732 *ifpp = ifp; 1733 return (addrsofif); 1734 } 1735 1736 addrs += addrsofif; 1737 } 1738 1739 return (addrs); 1740 } 1741 1742 static int 1743 ni6_store_addrs(ni6, nni6, ifp0, resid) 1744 struct icmp6_nodeinfo *ni6, *nni6; 1745 struct ifnet *ifp0; 1746 int resid; 1747 { 1748 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1749 struct in6_ifaddr *ifa6; 1750 struct ifaddr *ifa; 1751 struct ifnet *ifp_dep = NULL; 1752 int copied = 0, allow_deprecated = 0; 1753 u_char *cp = (u_char *)(nni6 + 1); 1754 int niflags = ni6->ni_flags; 1755 u_int32_t ltime; 1756 1757 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1758 return (0); /* needless to copy */ 1759 1760 again: 1761 1762 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1763 { 1764 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1765 ifa = ifa->ifa_list.tqe_next) 1766 { 1767 if (ifa->ifa_addr->sa_family != AF_INET6) 1768 continue; 1769 ifa6 = (struct in6_ifaddr *)ifa; 1770 1771 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1772 allow_deprecated == 0) { 1773 /* 1774 * prefererred address should be put before 1775 * deprecated addresses. 1776 */ 1777 1778 /* record the interface for later search */ 1779 if (ifp_dep == NULL) 1780 ifp_dep = ifp; 1781 1782 continue; 1783 } 1784 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1785 allow_deprecated != 0) 1786 continue; /* we now collect deprecated addrs */ 1787 1788 /* What do we have to do about ::1? */ 1789 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1790 case IPV6_ADDR_SCOPE_LINKLOCAL: 1791 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1792 continue; 1793 break; 1794 case IPV6_ADDR_SCOPE_SITELOCAL: 1795 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1796 continue; 1797 break; 1798 case IPV6_ADDR_SCOPE_GLOBAL: 1799 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1800 continue; 1801 break; 1802 default: 1803 continue; 1804 } 1805 1806 /* 1807 * check if anycast is okay. 1808 * XXX: just experimental. not in the spec. 1809 */ 1810 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1811 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1812 continue; 1813 1814 /* now we can copy the address */ 1815 if (resid < sizeof(struct in6_addr) + 1816 sizeof(u_int32_t)) { 1817 /* 1818 * We give up much more copy. 1819 * Set the truncate flag and return. 1820 */ 1821 nni6->ni_flags |= 1822 NI_NODEADDR_FLAG_TRUNCATE; 1823 return (copied); 1824 } 1825 1826 /* 1827 * Set the TTL of the address. 1828 * The TTL value should be one of the following 1829 * according to the specification: 1830 * 1831 * 1. The remaining lifetime of a DHCP lease on the 1832 * address, or 1833 * 2. The remaining Valid Lifetime of a prefix from 1834 * which the address was derived through Stateless 1835 * Autoconfiguration. 1836 * 1837 * Note that we currently do not support stateful 1838 * address configuration by DHCPv6, so the former 1839 * case can't happen. 1840 * 1841 * TTL must be 2^31 > TTL >= 0. 1842 */ 1843 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1844 ltime = ND6_INFINITE_LIFETIME; 1845 else { 1846 if (ifa6->ia6_lifetime.ia6t_expire > 1847 time.tv_sec) 1848 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time.tv_sec); 1849 else 1850 ltime = 0; 1851 } 1852 1853 bcopy(<ime, cp, sizeof(u_int32_t)); 1854 cp += sizeof(u_int32_t); 1855 1856 /* copy the address itself */ 1857 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1858 sizeof(struct in6_addr)); 1859 /* XXX: KAME link-local hack; remove ifindex */ 1860 if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr)) 1861 ((struct in6_addr *)cp)->s6_addr16[1] = 0; 1862 cp += sizeof(struct in6_addr); 1863 1864 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1865 copied += (sizeof(struct in6_addr) + 1866 sizeof(u_int32_t)); 1867 } 1868 if (ifp0) /* we need search only on the specified IF */ 1869 break; 1870 } 1871 1872 if (allow_deprecated == 0 && ifp_dep != NULL) { 1873 ifp = ifp_dep; 1874 allow_deprecated = 1; 1875 1876 goto again; 1877 } 1878 1879 return (copied); 1880 } 1881 1882 /* 1883 * XXX almost dup'ed code with rip6_input. 1884 */ 1885 static int 1886 icmp6_rip6_input(mp, off) 1887 struct mbuf **mp; 1888 int off; 1889 { 1890 struct mbuf *m = *mp; 1891 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1892 struct in6pcb *in6p; 1893 struct in6pcb *last = NULL; 1894 struct sockaddr_in6 rip6src; 1895 struct icmp6_hdr *icmp6; 1896 struct mbuf *opts = NULL; 1897 1898 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1899 if (icmp6 == NULL) { 1900 /* m is already reclaimed */ 1901 return IPPROTO_DONE; 1902 } 1903 1904 bzero(&rip6src, sizeof(rip6src)); 1905 rip6src.sin6_len = sizeof(struct sockaddr_in6); 1906 rip6src.sin6_family = AF_INET6; 1907 /* KAME hack: recover scopeid */ 1908 (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif); 1909 1910 for (in6p = rawin6pcbtable.inpt_queue.cqh_first; 1911 in6p != (struct inpcb *)&rawin6pcbtable.inpt_queue; 1912 in6p = in6p->inp_queue.cqe_next) 1913 { 1914 if (!(in6p->in6p_flags & INP_IPV6)) 1915 continue; 1916 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6) 1917 continue; 1918 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1919 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1920 continue; 1921 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1922 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1923 continue; 1924 if (in6p->in6p_icmp6filt 1925 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1926 in6p->in6p_icmp6filt)) 1927 continue; 1928 if (last) { 1929 struct mbuf *n; 1930 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1931 if (last->in6p_flags & IN6P_CONTROLOPTS) 1932 ip6_savecontrol(last, &opts, ip6, n); 1933 /* strip intermediate headers */ 1934 m_adj(n, off); 1935 if (sbappendaddr(&last->in6p_socket->so_rcv, 1936 (struct sockaddr *)&rip6src, 1937 n, opts) == 0) { 1938 /* should notify about lost packet */ 1939 m_freem(n); 1940 if (opts) 1941 m_freem(opts); 1942 } else 1943 sorwakeup(last->in6p_socket); 1944 opts = NULL; 1945 } 1946 } 1947 last = in6p; 1948 } 1949 if (last) { 1950 if (last->in6p_flags & IN6P_CONTROLOPTS) 1951 ip6_savecontrol(last, &opts, ip6, m); 1952 /* strip intermediate headers */ 1953 m_adj(m, off); 1954 if (sbappendaddr(&last->in6p_socket->so_rcv, 1955 (struct sockaddr *)&rip6src, 1956 m, opts) == 0) { 1957 m_freem(m); 1958 if (opts) 1959 m_freem(opts); 1960 } else 1961 sorwakeup(last->in6p_socket); 1962 } else { 1963 m_freem(m); 1964 ip6stat.ip6s_delivered--; 1965 } 1966 return IPPROTO_DONE; 1967 } 1968 1969 /* 1970 * Reflect the ip6 packet back to the source. 1971 * OFF points to the icmp6 header, counted from the top of the mbuf. 1972 * 1973 * Note: RFC 1885 required that an echo reply should be truncated if it 1974 * did not fit in with (return) path MTU, and KAME code supported the 1975 * behavior. However, as a clarification after the RFC, this limitation 1976 * was removed in a revised version of the spec, RFC 2463. We had kept the 1977 * old behavior, with a (non-default) ifdef block, while the new version of 1978 * the spec was an internet-draft status, and even after the new RFC was 1979 * published. But it would rather make sense to clean the obsoleted part 1980 * up, and to make the code simpler at this stage. 1981 */ 1982 void 1983 icmp6_reflect(m, off) 1984 struct mbuf *m; 1985 size_t off; 1986 { 1987 struct ip6_hdr *ip6; 1988 struct icmp6_hdr *icmp6; 1989 struct in6_ifaddr *ia; 1990 struct in6_addr t, *src = 0; 1991 int plen; 1992 int type, code; 1993 struct ifnet *outif = NULL; 1994 struct sockaddr_in6 sa6_src, sa6_dst; 1995 1996 /* too short to reflect */ 1997 if (off < sizeof(struct ip6_hdr)) { 1998 nd6log((LOG_DEBUG, 1999 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 2000 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2001 __FILE__, __LINE__)); 2002 goto bad; 2003 } 2004 2005 /* 2006 * If there are extra headers between IPv6 and ICMPv6, strip 2007 * off that header first. 2008 */ 2009 #ifdef DIAGNOSTIC 2010 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2011 panic("assumption failed in icmp6_reflect"); 2012 #endif 2013 if (off > sizeof(struct ip6_hdr)) { 2014 size_t l; 2015 struct ip6_hdr nip6; 2016 2017 l = off - sizeof(struct ip6_hdr); 2018 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2019 m_adj(m, l); 2020 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2021 if (m->m_len < l) { 2022 if ((m = m_pullup(m, l)) == NULL) 2023 return; 2024 } 2025 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2026 } else /* off == sizeof(struct ip6_hdr) */ { 2027 size_t l; 2028 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2029 if (m->m_len < l) { 2030 if ((m = m_pullup(m, l)) == NULL) 2031 return; 2032 } 2033 } 2034 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2035 ip6 = mtod(m, struct ip6_hdr *); 2036 ip6->ip6_nxt = IPPROTO_ICMPV6; 2037 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2038 type = icmp6->icmp6_type; /* keep type for statistics */ 2039 code = icmp6->icmp6_code; /* ditto. */ 2040 2041 t = ip6->ip6_dst; 2042 /* 2043 * ip6_input() drops a packet if its src is multicast. 2044 * So, the src is never multicast. 2045 */ 2046 ip6->ip6_dst = ip6->ip6_src; 2047 2048 /* 2049 * XXX: make sure to embed scope zone information, using 2050 * already embedded IDs or the received interface (if any). 2051 * Note that rcvif may be NULL. 2052 * TODO: scoped routing case (XXX). 2053 */ 2054 bzero(&sa6_src, sizeof(sa6_src)); 2055 sa6_src.sin6_family = AF_INET6; 2056 sa6_src.sin6_len = sizeof(sa6_src); 2057 sa6_src.sin6_addr = ip6->ip6_dst; 2058 in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif); 2059 in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL); 2060 bzero(&sa6_dst, sizeof(sa6_dst)); 2061 sa6_dst.sin6_family = AF_INET6; 2062 sa6_dst.sin6_len = sizeof(sa6_dst); 2063 sa6_dst.sin6_addr = t; 2064 in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif); 2065 in6_embedscope(&t, &sa6_dst, NULL, NULL); 2066 2067 /* 2068 * If the incoming packet was addressed directly to us (i.e. unicast), 2069 * use dst as the src for the reply. 2070 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible 2071 * (for example) when we encounter an error while forwarding procedure 2072 * destined to a duplicated address of ours. 2073 */ 2074 for (ia = in6_ifaddr; ia; ia = ia->ia_next) 2075 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) && 2076 (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) { 2077 src = &t; 2078 break; 2079 } 2080 if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) { 2081 /* 2082 * This is the case if the dst is our link-local address 2083 * and the sender is also ourselves. 2084 */ 2085 src = &t; 2086 } 2087 2088 if (src == 0) { 2089 int e; 2090 struct route_in6 ro; 2091 2092 /* 2093 * This case matches to multicasts, our anycast, or unicasts 2094 * that we do not own. Select a source address based on the 2095 * source address of the erroneous packet. 2096 */ 2097 bzero(&ro, sizeof(ro)); 2098 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e); 2099 if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */ 2100 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2101 } 2102 if (src == NULL) { 2103 nd6log((LOG_DEBUG, 2104 "icmp6_reflect: source can't be determined: " 2105 "dst=%s, error=%d\n", 2106 ip6_sprintf(&sa6_src.sin6_addr), e)); 2107 goto bad; 2108 } 2109 } 2110 2111 ip6->ip6_src = *src; 2112 2113 ip6->ip6_flow = 0; 2114 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2115 ip6->ip6_vfc |= IPV6_VERSION; 2116 ip6->ip6_nxt = IPPROTO_ICMPV6; 2117 if (m->m_pkthdr.rcvif) { 2118 /* XXX: This may not be the outgoing interface */ 2119 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2120 } else 2121 ip6->ip6_hlim = ip6_defhlim; 2122 2123 icmp6->icmp6_cksum = 0; 2124 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2125 sizeof(struct ip6_hdr), plen); 2126 2127 /* 2128 * XXX option handling 2129 */ 2130 2131 m->m_flags &= ~(M_BCAST|M_MCAST); 2132 2133 /* 2134 * To avoid a "too big" situation at an intermediate router 2135 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2136 * Note that only echo and node information replies are affected, 2137 * since the length of ICMP6 errors is limited to the minimum MTU. 2138 */ 2139 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, &outif) != 0 && outif) 2140 icmp6_ifstat_inc(outif, ifs6_out_error); 2141 2142 if (outif) 2143 icmp6_ifoutstat_inc(outif, type, code); 2144 2145 return; 2146 2147 bad: 2148 m_freem(m); 2149 return; 2150 } 2151 2152 void 2153 icmp6_fasttimo() 2154 { 2155 2156 mld6_fasttimeo(); 2157 } 2158 2159 static const char * 2160 icmp6_redirect_diag(src6, dst6, tgt6) 2161 struct in6_addr *src6; 2162 struct in6_addr *dst6; 2163 struct in6_addr *tgt6; 2164 { 2165 static char buf[1024]; 2166 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2167 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2168 return buf; 2169 } 2170 2171 void 2172 icmp6_redirect_input(m, off) 2173 struct mbuf *m; 2174 int off; 2175 { 2176 struct ifnet *ifp = m->m_pkthdr.rcvif; 2177 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2178 struct nd_redirect *nd_rd; 2179 int icmp6len = ntohs(ip6->ip6_plen); 2180 char *lladdr = NULL; 2181 int lladdrlen = 0; 2182 u_char *redirhdr = NULL; 2183 int redirhdrlen = 0; 2184 struct rtentry *rt = NULL; 2185 int is_router; 2186 int is_onlink; 2187 struct in6_addr src6 = ip6->ip6_src; 2188 struct in6_addr redtgt6; 2189 struct in6_addr reddst6; 2190 union nd_opts ndopts; 2191 2192 if (!ifp) 2193 return; 2194 2195 /* XXX if we are router, we don't update route by icmp6 redirect */ 2196 if (ip6_forwarding) 2197 goto freeit; 2198 if (!icmp6_rediraccept) 2199 goto freeit; 2200 2201 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2202 if (nd_rd == NULL) { 2203 icmp6stat.icp6s_tooshort++; 2204 return; 2205 } 2206 redtgt6 = nd_rd->nd_rd_target; 2207 reddst6 = nd_rd->nd_rd_dst; 2208 2209 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2210 redtgt6.s6_addr16[1] = htons(ifp->if_index); 2211 if (IN6_IS_ADDR_LINKLOCAL(&reddst6)) 2212 reddst6.s6_addr16[1] = htons(ifp->if_index); 2213 2214 /* validation */ 2215 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2216 nd6log((LOG_ERR, 2217 "ICMP6 redirect sent from %s rejected; " 2218 "must be from linklocal\n", ip6_sprintf(&src6))); 2219 goto bad; 2220 } 2221 if (ip6->ip6_hlim != 255) { 2222 nd6log((LOG_ERR, 2223 "ICMP6 redirect sent from %s rejected; " 2224 "hlim=%d (must be 255)\n", 2225 ip6_sprintf(&src6), ip6->ip6_hlim)); 2226 goto bad; 2227 } 2228 { 2229 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2230 struct sockaddr_in6 sin6; 2231 struct in6_addr *gw6; 2232 2233 bzero(&sin6, sizeof(sin6)); 2234 sin6.sin6_family = AF_INET6; 2235 sin6.sin6_len = sizeof(struct sockaddr_in6); 2236 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2237 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2238 if (rt) { 2239 if (rt->rt_gateway == NULL || 2240 rt->rt_gateway->sa_family != AF_INET6) { 2241 nd6log((LOG_ERR, 2242 "ICMP6 redirect rejected; no route " 2243 "with inet6 gateway found for redirect dst: %s\n", 2244 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2245 RTFREE(rt); 2246 goto bad; 2247 } 2248 2249 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2250 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2251 nd6log((LOG_ERR, 2252 "ICMP6 redirect rejected; " 2253 "not equal to gw-for-src=%s (must be same): " 2254 "%s\n", 2255 ip6_sprintf(gw6), 2256 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2257 RTFREE(rt); 2258 goto bad; 2259 } 2260 } else { 2261 nd6log((LOG_ERR, 2262 "ICMP6 redirect rejected; " 2263 "no route found for redirect dst: %s\n", 2264 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2265 goto bad; 2266 } 2267 RTFREE(rt); 2268 rt = NULL; 2269 } 2270 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2271 nd6log((LOG_ERR, 2272 "ICMP6 redirect rejected; " 2273 "redirect dst must be unicast: %s\n", 2274 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2275 goto bad; 2276 } 2277 2278 is_router = is_onlink = 0; 2279 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2280 is_router = 1; /* router case */ 2281 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2282 is_onlink = 1; /* on-link destination case */ 2283 if (!is_router && !is_onlink) { 2284 nd6log((LOG_ERR, 2285 "ICMP6 redirect rejected; " 2286 "neither router case nor onlink case: %s\n", 2287 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2288 goto bad; 2289 } 2290 /* validation passed */ 2291 2292 icmp6len -= sizeof(*nd_rd); 2293 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2294 if (nd6_options(&ndopts) < 0) { 2295 nd6log((LOG_INFO, "icmp6_redirect_input: " 2296 "invalid ND option, rejected: %s\n", 2297 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2298 /* nd6_options have incremented stats */ 2299 goto freeit; 2300 } 2301 2302 if (ndopts.nd_opts_tgt_lladdr) { 2303 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2304 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2305 } 2306 2307 if (ndopts.nd_opts_rh) { 2308 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len; 2309 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */ 2310 } 2311 2312 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2313 nd6log((LOG_INFO, 2314 "icmp6_redirect_input: lladdrlen mismatch for %s " 2315 "(if %d, icmp6 packet %d): %s\n", 2316 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2317 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2318 goto bad; 2319 } 2320 2321 /* RFC 2461 8.3 */ 2322 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2323 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2324 2325 if (!is_onlink) { /* better router case. perform rtredirect. */ 2326 /* perform rtredirect */ 2327 struct sockaddr_in6 sdst; 2328 struct sockaddr_in6 sgw; 2329 struct sockaddr_in6 ssrc; 2330 unsigned long rtcount; 2331 struct rtentry *newrt = NULL; 2332 2333 /* 2334 * do not install redirect route, if the number of entries 2335 * is too much (> hiwat). note that, the node (= host) will 2336 * work just fine even if we do not install redirect route 2337 * (there will be additional hops, though). 2338 */ 2339 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2340 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2341 return; 2342 else if (0 <= icmp6_redirect_lowat && 2343 rtcount > icmp6_redirect_lowat) { 2344 /* 2345 * XXX nuke a victim, install the new one. 2346 */ 2347 } 2348 2349 bzero(&sdst, sizeof(sdst)); 2350 bzero(&sgw, sizeof(sgw)); 2351 bzero(&ssrc, sizeof(ssrc)); 2352 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2353 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2354 sizeof(struct sockaddr_in6); 2355 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2356 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2357 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2358 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2359 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2360 (struct sockaddr *)&ssrc, 2361 &newrt); 2362 2363 if (newrt) { 2364 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2365 icmp6_redirect_timeout_q); 2366 rtfree(newrt); 2367 } 2368 } 2369 /* finally update cached route in each socket via pfctlinput */ 2370 { 2371 struct sockaddr_in6 sdst; 2372 2373 bzero(&sdst, sizeof(sdst)); 2374 sdst.sin6_family = AF_INET6; 2375 sdst.sin6_len = sizeof(struct sockaddr_in6); 2376 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2377 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2378 } 2379 2380 freeit: 2381 m_freem(m); 2382 return; 2383 2384 bad: 2385 icmp6stat.icp6s_badredirect++; 2386 m_freem(m); 2387 } 2388 2389 void 2390 icmp6_redirect_output(m0, rt) 2391 struct mbuf *m0; 2392 struct rtentry *rt; 2393 { 2394 struct ifnet *ifp; /* my outgoing interface */ 2395 struct in6_addr *ifp_ll6; 2396 struct in6_addr *nexthop; 2397 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2398 struct mbuf *m = NULL; /* newly allocated one */ 2399 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2400 struct nd_redirect *nd_rd; 2401 size_t maxlen; 2402 u_char *p; 2403 struct sockaddr_in6 src_sa; 2404 2405 icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2406 2407 /* if we are not router, we don't send icmp6 redirect */ 2408 if (!ip6_forwarding) 2409 goto fail; 2410 2411 /* sanity check */ 2412 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2413 goto fail; 2414 2415 /* 2416 * Address check: 2417 * the source address must identify a neighbor, and 2418 * the destination address must not be a multicast address 2419 * [RFC 2461, sec 8.2] 2420 */ 2421 sip6 = mtod(m0, struct ip6_hdr *); 2422 bzero(&src_sa, sizeof(src_sa)); 2423 src_sa.sin6_family = AF_INET6; 2424 src_sa.sin6_len = sizeof(src_sa); 2425 src_sa.sin6_addr = sip6->ip6_src; 2426 /* we don't currently use sin6_scope_id, but eventually use it */ 2427 src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src); 2428 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2429 goto fail; 2430 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2431 goto fail; /* what should we do here? */ 2432 2433 /* rate limit */ 2434 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2435 goto fail; 2436 2437 /* 2438 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2439 * we almost always ask for an mbuf cluster for simplicity. 2440 * (MHLEN < IPV6_MMTU is almost always true) 2441 */ 2442 #if IPV6_MMTU >= MCLBYTES 2443 # error assumption failed about IPV6_MMTU and MCLBYTES 2444 #endif 2445 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2446 if (m && IPV6_MMTU >= MHLEN) 2447 MCLGET(m, M_DONTWAIT); 2448 if (!m) 2449 goto fail; 2450 m->m_pkthdr.rcvif = NULL; 2451 m->m_len = 0; 2452 maxlen = M_TRAILINGSPACE(m); 2453 maxlen = min(IPV6_MMTU, maxlen); 2454 /* just for safety */ 2455 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2456 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2457 goto fail; 2458 } 2459 2460 { 2461 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2462 struct in6_ifaddr *ia; 2463 if ((ia = in6ifa_ifpforlinklocal(ifp, 2464 IN6_IFF_NOTREADY| 2465 IN6_IFF_ANYCAST)) == NULL) 2466 goto fail; 2467 ifp_ll6 = &ia->ia_addr.sin6_addr; 2468 } 2469 2470 /* get ip6 linklocal address for the router. */ 2471 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2472 struct sockaddr_in6 *sin6; 2473 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2474 nexthop = &sin6->sin6_addr; 2475 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2476 nexthop = NULL; 2477 } else 2478 nexthop = NULL; 2479 2480 /* ip6 */ 2481 ip6 = mtod(m, struct ip6_hdr *); 2482 ip6->ip6_flow = 0; 2483 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2484 ip6->ip6_vfc |= IPV6_VERSION; 2485 /* ip6->ip6_plen will be set later */ 2486 ip6->ip6_nxt = IPPROTO_ICMPV6; 2487 ip6->ip6_hlim = 255; 2488 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2489 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2490 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2491 2492 /* ND Redirect */ 2493 nd_rd = (struct nd_redirect *)(ip6 + 1); 2494 nd_rd->nd_rd_type = ND_REDIRECT; 2495 nd_rd->nd_rd_code = 0; 2496 nd_rd->nd_rd_reserved = 0; 2497 if (rt->rt_flags & RTF_GATEWAY) { 2498 /* 2499 * nd_rd->nd_rd_target must be a link-local address in 2500 * better router cases. 2501 */ 2502 if (!nexthop) 2503 goto fail; 2504 bcopy(nexthop, &nd_rd->nd_rd_target, 2505 sizeof(nd_rd->nd_rd_target)); 2506 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2507 sizeof(nd_rd->nd_rd_dst)); 2508 } else { 2509 /* make sure redtgt == reddst */ 2510 nexthop = &sip6->ip6_dst; 2511 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2512 sizeof(nd_rd->nd_rd_target)); 2513 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2514 sizeof(nd_rd->nd_rd_dst)); 2515 } 2516 2517 p = (u_char *)(nd_rd + 1); 2518 2519 { 2520 /* target lladdr option */ 2521 struct rtentry *rt_nexthop = NULL; 2522 int len; 2523 struct sockaddr_dl *sdl; 2524 struct nd_opt_hdr *nd_opt; 2525 char *lladdr; 2526 2527 rt_nexthop = nd6_lookup(nexthop, 0, ifp); 2528 if (!rt_nexthop) 2529 goto nolladdropt; 2530 len = sizeof(*nd_opt) + ifp->if_addrlen; 2531 len = (len + 7) & ~7; /* round by 8 */ 2532 /* safety check */ 2533 if (len + (p - (u_char *)ip6) > maxlen) 2534 goto nolladdropt; 2535 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) && 2536 (rt_nexthop->rt_flags & RTF_LLINFO) && 2537 (rt_nexthop->rt_gateway->sa_family == AF_LINK) && 2538 (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) && 2539 sdl->sdl_alen) { 2540 nd_opt = (struct nd_opt_hdr *)p; 2541 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2542 nd_opt->nd_opt_len = len >> 3; 2543 lladdr = (char *)(nd_opt + 1); 2544 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen); 2545 p += len; 2546 } 2547 } 2548 nolladdropt:; 2549 2550 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2551 2552 /* just to be safe */ 2553 if (p - (u_char *)ip6 > maxlen) 2554 goto noredhdropt; 2555 2556 { 2557 /* redirected header option */ 2558 int len; 2559 struct nd_opt_rd_hdr *nd_opt_rh; 2560 2561 /* 2562 * compute the maximum size for icmp6 redirect header option. 2563 * XXX room for auth header? 2564 */ 2565 len = maxlen - (p - (u_char *)ip6); 2566 len &= ~7; 2567 2568 /* 2569 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2570 * about padding/truncate rule for the original IP packet. 2571 * From the discussion on IPv6imp in Feb 1999, 2572 * the consensus was: 2573 * - "attach as much as possible" is the goal 2574 * - pad if not aligned (original size can be guessed by 2575 * original ip6 header) 2576 * Following code adds the padding if it is simple enough, 2577 * and truncates if not. 2578 */ 2579 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2580 /* not enough room, truncate */ 2581 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2582 m0->m_pkthdr.len); 2583 } else { 2584 /* 2585 * enough room, truncate if not aligned. 2586 * we don't pad here for simplicity. 2587 */ 2588 size_t extra; 2589 2590 extra = m0->m_pkthdr.len % 8; 2591 if (extra) { 2592 /* truncate */ 2593 m_adj(m0, -extra); 2594 } 2595 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2596 } 2597 2598 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2599 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2600 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2601 nd_opt_rh->nd_opt_rh_len = len >> 3; 2602 p += sizeof(*nd_opt_rh); 2603 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2604 2605 /* connect m0 to m */ 2606 m_cat(m, m0); 2607 m->m_pkthdr.len += m0->m_pkthdr.len; 2608 m0 = NULL; 2609 } 2610 noredhdropt: 2611 if (m0) { 2612 m_freem(m0); 2613 m0 = NULL; 2614 } 2615 2616 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src)) 2617 sip6->ip6_src.s6_addr16[1] = 0; 2618 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst)) 2619 sip6->ip6_dst.s6_addr16[1] = 0; 2620 #if 0 2621 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) 2622 ip6->ip6_src.s6_addr16[1] = 0; 2623 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) 2624 ip6->ip6_dst.s6_addr16[1] = 0; 2625 #endif 2626 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target)) 2627 nd_rd->nd_rd_target.s6_addr16[1] = 0; 2628 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst)) 2629 nd_rd->nd_rd_dst.s6_addr16[1] = 0; 2630 2631 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2632 2633 nd_rd->nd_rd_cksum = 0; 2634 nd_rd->nd_rd_cksum 2635 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2636 2637 /* send the packet to outside... */ 2638 if (ip6_output(m, NULL, NULL, 0, NULL, NULL) != 0) 2639 icmp6_ifstat_inc(ifp, ifs6_out_error); 2640 2641 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2642 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2643 icmp6stat.icp6s_outhist[ND_REDIRECT]++; 2644 2645 return; 2646 2647 fail: 2648 if (m) 2649 m_freem(m); 2650 if (m0) 2651 m_freem(m0); 2652 } 2653 2654 /* NRL PCB */ 2655 #define sotoin6pcb sotoinpcb 2656 #define in6pcb inpcb 2657 #define in6p_icmp6filt inp_icmp6filt 2658 2659 /* 2660 * ICMPv6 socket option processing. 2661 */ 2662 int 2663 icmp6_ctloutput(op, so, level, optname, mp) 2664 int op; 2665 struct socket *so; 2666 int level, optname; 2667 struct mbuf **mp; 2668 { 2669 int error = 0; 2670 int optlen; 2671 struct in6pcb *in6p = sotoin6pcb(so); 2672 struct mbuf *m = *mp; 2673 2674 optlen = m ? m->m_len : 0; 2675 2676 if (level != IPPROTO_ICMPV6) { 2677 if (op == PRCO_SETOPT && m) 2678 (void)m_free(m); 2679 return EINVAL; 2680 } 2681 2682 switch (op) { 2683 case PRCO_SETOPT: 2684 switch (optname) { 2685 case ICMP6_FILTER: 2686 { 2687 struct icmp6_filter *p; 2688 2689 if (optlen != sizeof(*p)) { 2690 error = EMSGSIZE; 2691 break; 2692 } 2693 p = mtod(m, struct icmp6_filter *); 2694 if (!p || !in6p->in6p_icmp6filt) { 2695 error = EINVAL; 2696 break; 2697 } 2698 bcopy(p, in6p->in6p_icmp6filt, 2699 sizeof(struct icmp6_filter)); 2700 error = 0; 2701 break; 2702 } 2703 2704 default: 2705 error = ENOPROTOOPT; 2706 break; 2707 } 2708 if (m) 2709 (void)m_freem(m); 2710 break; 2711 2712 case PRCO_GETOPT: 2713 switch (optname) { 2714 case ICMP6_FILTER: 2715 { 2716 struct icmp6_filter *p; 2717 2718 if (!in6p->in6p_icmp6filt) { 2719 error = EINVAL; 2720 break; 2721 } 2722 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2723 m->m_len = sizeof(struct icmp6_filter); 2724 p = mtod(m, struct icmp6_filter *); 2725 bcopy(in6p->in6p_icmp6filt, p, 2726 sizeof(struct icmp6_filter)); 2727 error = 0; 2728 break; 2729 } 2730 2731 default: 2732 error = ENOPROTOOPT; 2733 break; 2734 } 2735 break; 2736 } 2737 2738 return (error); 2739 } 2740 2741 /* NRL PCB */ 2742 #undef sotoin6pcb 2743 #undef in6pcb 2744 #undef in6p_icmp6filt 2745 2746 /* 2747 * Perform rate limit check. 2748 * Returns 0 if it is okay to send the icmp6 packet. 2749 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2750 * limitation. 2751 * 2752 * XXX per-destination/type check necessary? 2753 */ 2754 static int 2755 icmp6_ratelimit(dst, type, code) 2756 const struct in6_addr *dst; /* not used at this moment */ 2757 const int type; /* not used at this moment */ 2758 const int code; /* not used at this moment */ 2759 { 2760 int ret; 2761 2762 ret = 0; /* okay to send */ 2763 2764 /* PPS limit */ 2765 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2766 icmp6errppslim)) { 2767 /* The packet is subject to rate limit */ 2768 ret++; 2769 } 2770 2771 return ret; 2772 } 2773 2774 static struct rtentry * 2775 icmp6_mtudisc_clone(dst) 2776 struct sockaddr *dst; 2777 { 2778 struct rtentry *rt; 2779 int error; 2780 2781 rt = rtalloc1(dst, 1); 2782 if (rt == 0) 2783 return NULL; 2784 2785 /* If we didn't get a host route, allocate one */ 2786 if ((rt->rt_flags & RTF_HOST) == 0) { 2787 struct rtentry *nrt; 2788 2789 error = rtrequest((int) RTM_ADD, dst, 2790 (struct sockaddr *) rt->rt_gateway, 2791 (struct sockaddr *) 0, 2792 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2793 if (error) { 2794 rtfree(rt); 2795 return NULL; 2796 } 2797 nrt->rt_rmx = rt->rt_rmx; 2798 rtfree(rt); 2799 rt = nrt; 2800 } 2801 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2802 icmp6_mtudisc_timeout_q); 2803 if (error) { 2804 rtfree(rt); 2805 return NULL; 2806 } 2807 2808 return rt; /* caller need to call rtfree() */ 2809 } 2810 2811 static void 2812 icmp6_mtudisc_timeout(rt, r) 2813 struct rtentry *rt; 2814 struct rttimer *r; 2815 { 2816 if (rt == NULL) 2817 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2818 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2819 (RTF_DYNAMIC | RTF_HOST)) { 2820 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2821 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2822 } else { 2823 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2824 rt->rt_rmx.rmx_mtu = 0; 2825 } 2826 } 2827 2828 static void 2829 icmp6_redirect_timeout(rt, r) 2830 struct rtentry *rt; 2831 struct rttimer *r; 2832 { 2833 if (rt == NULL) 2834 panic("icmp6_redirect_timeout: bad route to timeout"); 2835 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2836 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2837 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2838 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2839 } 2840 } 2841 2842 #include <uvm/uvm_extern.h> 2843 #include <sys/sysctl.h> 2844 int 2845 icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 2846 int *name; 2847 u_int namelen; 2848 void *oldp; 2849 size_t *oldlenp; 2850 void *newp; 2851 size_t newlen; 2852 { 2853 2854 /* All sysctl names at this level are terminal. */ 2855 if (namelen != 1) 2856 return ENOTDIR; 2857 2858 switch (name[0]) { 2859 2860 case ICMPV6CTL_REDIRACCEPT: 2861 return sysctl_int(oldp, oldlenp, newp, newlen, 2862 &icmp6_rediraccept); 2863 case ICMPV6CTL_REDIRTIMEOUT: 2864 return sysctl_int(oldp, oldlenp, newp, newlen, 2865 &icmp6_redirtimeout); 2866 case ICMPV6CTL_STATS: 2867 return sysctl_rdstruct(oldp, oldlenp, newp, 2868 &icmp6stat, sizeof(icmp6stat)); 2869 case ICMPV6CTL_ND6_PRUNE: 2870 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune); 2871 case ICMPV6CTL_ND6_DELAY: 2872 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay); 2873 case ICMPV6CTL_ND6_UMAXTRIES: 2874 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries); 2875 case ICMPV6CTL_ND6_MMAXTRIES: 2876 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries); 2877 case ICMPV6CTL_ND6_USELOOPBACK: 2878 return sysctl_int(oldp, oldlenp, newp, newlen, 2879 &nd6_useloopback); 2880 case ICMPV6CTL_NODEINFO: 2881 return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo); 2882 case ICMPV6CTL_ERRPPSLIMIT: 2883 return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6errppslim); 2884 case ICMPV6CTL_ND6_MAXNUDHINT: 2885 return sysctl_int(oldp, oldlenp, newp, newlen, 2886 &nd6_maxnudhint); 2887 case ICMPV6CTL_MTUDISC_HIWAT: 2888 return sysctl_int(oldp, oldlenp, newp, newlen, 2889 &icmp6_mtudisc_hiwat); 2890 case ICMPV6CTL_MTUDISC_LOWAT: 2891 return sysctl_int(oldp, oldlenp, newp, newlen, 2892 &icmp6_mtudisc_lowat); 2893 case ICMPV6CTL_ND6_DEBUG: 2894 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_debug); 2895 case ICMPV6CTL_ND6_DRLIST: 2896 case ICMPV6CTL_ND6_PRLIST: 2897 return nd6_sysctl(name[0], oldp, oldlenp, newp, newlen); 2898 default: 2899 return ENOPROTOOPT; 2900 } 2901 /* NOTREACHED */ 2902 } 2903