1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $ 30 */ 31 32 /*- 33 * Copyright (c) 1982, 1986, 1988, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 4. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 61 */ 62 63 #include <sys/cdefs.h> 64 __FBSDID("$FreeBSD$"); 65 66 #include "opt_inet.h" 67 #include "opt_inet6.h" 68 #include "opt_ipsec.h" 69 #include "opt_route.h" 70 71 #include <sys/param.h> 72 #include <sys/domain.h> 73 #include <sys/kernel.h> 74 #include <sys/lock.h> 75 #include <sys/malloc.h> 76 #include <sys/mbuf.h> 77 #include <sys/protosw.h> 78 #include <sys/signalvar.h> 79 #include <sys/socket.h> 80 #include <sys/socketvar.h> 81 #include <sys/sx.h> 82 #include <sys/syslog.h> 83 #include <sys/systm.h> 84 #include <sys/time.h> 85 #include <sys/vimage.h> 86 87 #include <net/if.h> 88 #include <net/if_dl.h> 89 #include <net/if_llatbl.h> 90 #include <net/if_types.h> 91 #include <net/route.h> 92 #include <net/vnet.h> 93 94 #include <netinet/in.h> 95 #include <netinet/in_pcb.h> 96 #include <netinet/in_var.h> 97 #include <netinet/ip6.h> 98 #include <netinet/icmp6.h> 99 #include <netinet/tcp_var.h> 100 #include <netinet/vinet.h> 101 102 #include <netinet6/in6_ifattach.h> 103 #include <netinet6/in6_pcb.h> 104 #include <netinet6/ip6protosw.h> 105 #include <netinet6/ip6_var.h> 106 #include <netinet6/scope6_var.h> 107 #include <netinet6/mld6_var.h> 108 #include <netinet6/nd6.h> 109 #include <netinet6/vinet6.h> 110 111 #ifdef IPSEC 112 #include <netipsec/ipsec.h> 113 #include <netipsec/key.h> 114 #endif 115 116 extern struct domain inet6domain; 117 118 #ifdef VIMAGE_GLOBALS 119 extern struct inpcbinfo ripcbinfo; 120 extern struct inpcbhead ripcb; 121 extern int icmp6errppslim; 122 extern int icmp6_nodeinfo; 123 124 struct icmp6stat icmp6stat; 125 static int icmp6errpps_count; 126 static struct timeval icmp6errppslim_last; 127 #endif 128 129 static void icmp6_errcount(struct icmp6errstat *, int, int); 130 static int icmp6_rip6_input(struct mbuf **, int); 131 static int icmp6_ratelimit(const struct in6_addr *, const int, const int); 132 static const char *icmp6_redirect_diag __P((struct in6_addr *, 133 struct in6_addr *, struct in6_addr *)); 134 static struct mbuf *ni6_input(struct mbuf *, int); 135 static struct mbuf *ni6_nametodns(const char *, int, int); 136 static int ni6_dnsmatch(const char *, int, const char *, int); 137 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *, 138 struct ifnet **, struct in6_addr *)); 139 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 140 struct ifnet *, int)); 141 static int icmp6_notify_error(struct mbuf **, int, int, int); 142 143 144 void 145 icmp6_init(void) 146 { 147 INIT_VNET_INET6(curvnet); 148 149 V_icmp6errpps_count = 0; 150 151 mld6_init(); 152 } 153 154 static void 155 icmp6_errcount(struct icmp6errstat *stat, int type, int code) 156 { 157 switch (type) { 158 case ICMP6_DST_UNREACH: 159 switch (code) { 160 case ICMP6_DST_UNREACH_NOROUTE: 161 stat->icp6errs_dst_unreach_noroute++; 162 return; 163 case ICMP6_DST_UNREACH_ADMIN: 164 stat->icp6errs_dst_unreach_admin++; 165 return; 166 case ICMP6_DST_UNREACH_BEYONDSCOPE: 167 stat->icp6errs_dst_unreach_beyondscope++; 168 return; 169 case ICMP6_DST_UNREACH_ADDR: 170 stat->icp6errs_dst_unreach_addr++; 171 return; 172 case ICMP6_DST_UNREACH_NOPORT: 173 stat->icp6errs_dst_unreach_noport++; 174 return; 175 } 176 break; 177 case ICMP6_PACKET_TOO_BIG: 178 stat->icp6errs_packet_too_big++; 179 return; 180 case ICMP6_TIME_EXCEEDED: 181 switch (code) { 182 case ICMP6_TIME_EXCEED_TRANSIT: 183 stat->icp6errs_time_exceed_transit++; 184 return; 185 case ICMP6_TIME_EXCEED_REASSEMBLY: 186 stat->icp6errs_time_exceed_reassembly++; 187 return; 188 } 189 break; 190 case ICMP6_PARAM_PROB: 191 switch (code) { 192 case ICMP6_PARAMPROB_HEADER: 193 stat->icp6errs_paramprob_header++; 194 return; 195 case ICMP6_PARAMPROB_NEXTHEADER: 196 stat->icp6errs_paramprob_nextheader++; 197 return; 198 case ICMP6_PARAMPROB_OPTION: 199 stat->icp6errs_paramprob_option++; 200 return; 201 } 202 break; 203 case ND_REDIRECT: 204 stat->icp6errs_redirect++; 205 return; 206 } 207 stat->icp6errs_unknown++; 208 } 209 210 /* 211 * A wrapper function for icmp6_error() necessary when the erroneous packet 212 * may not contain enough scope zone information. 213 */ 214 void 215 icmp6_error2(struct mbuf *m, int type, int code, int param, 216 struct ifnet *ifp) 217 { 218 INIT_VNET_INET6(curvnet); 219 struct ip6_hdr *ip6; 220 221 if (ifp == NULL) 222 return; 223 224 #ifndef PULLDOWN_TEST 225 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), ); 226 #else 227 if (m->m_len < sizeof(struct ip6_hdr)) { 228 m = m_pullup(m, sizeof(struct ip6_hdr)); 229 if (m == NULL) 230 return; 231 } 232 #endif 233 234 ip6 = mtod(m, struct ip6_hdr *); 235 236 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 237 return; 238 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 239 return; 240 241 icmp6_error(m, type, code, param); 242 } 243 244 /* 245 * Generate an error packet of type error in response to bad IP6 packet. 246 */ 247 void 248 icmp6_error(struct mbuf *m, int type, int code, int param) 249 { 250 INIT_VNET_INET6(curvnet); 251 struct ip6_hdr *oip6, *nip6; 252 struct icmp6_hdr *icmp6; 253 u_int preplen; 254 int off; 255 int nxt; 256 257 V_icmp6stat.icp6s_error++; 258 259 /* count per-type-code statistics */ 260 icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, type, code); 261 262 #ifdef M_DECRYPTED /*not openbsd*/ 263 if (m->m_flags & M_DECRYPTED) { 264 V_icmp6stat.icp6s_canterror++; 265 goto freeit; 266 } 267 #endif 268 269 #ifndef PULLDOWN_TEST 270 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), ); 271 #else 272 if (m->m_len < sizeof(struct ip6_hdr)) { 273 m = m_pullup(m, sizeof(struct ip6_hdr)); 274 if (m == NULL) 275 return; 276 } 277 #endif 278 oip6 = mtod(m, struct ip6_hdr *); 279 280 /* 281 * If the destination address of the erroneous packet is a multicast 282 * address, or the packet was sent using link-layer multicast, 283 * we should basically suppress sending an error (RFC 2463, Section 284 * 2.4). 285 * We have two exceptions (the item e.2 in that section): 286 * - the Pakcet Too Big message can be sent for path MTU discovery. 287 * - the Parameter Problem Message that can be allowed an icmp6 error 288 * in the option type field. This check has been done in 289 * ip6_unknown_opt(), so we can just check the type and code. 290 */ 291 if ((m->m_flags & (M_BCAST|M_MCAST) || 292 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 293 (type != ICMP6_PACKET_TOO_BIG && 294 (type != ICMP6_PARAM_PROB || 295 code != ICMP6_PARAMPROB_OPTION))) 296 goto freeit; 297 298 /* 299 * RFC 2463, 2.4 (e.5): source address check. 300 * XXX: the case of anycast source? 301 */ 302 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 303 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 304 goto freeit; 305 306 /* 307 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 308 * don't do it. 309 */ 310 nxt = -1; 311 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 312 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 313 struct icmp6_hdr *icp; 314 315 #ifndef PULLDOWN_TEST 316 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), ); 317 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 318 #else 319 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 320 sizeof(*icp)); 321 if (icp == NULL) { 322 V_icmp6stat.icp6s_tooshort++; 323 return; 324 } 325 #endif 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 V_icmp6stat.icp6s_canterror++; 334 goto freeit; 335 } else { 336 /* ICMPv6 informational - send the error */ 337 } 338 } else { 339 /* non-ICMPv6 - send the error */ 340 } 341 342 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 343 344 /* Finally, do rate limitation check. */ 345 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 346 V_icmp6stat.icp6s_toofreq++; 347 goto freeit; 348 } 349 350 /* 351 * OK, ICMP6 can be generated. 352 */ 353 354 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 355 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 356 357 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 358 M_PREPEND(m, preplen, M_DONTWAIT); 359 if (m && m->m_len < preplen) 360 m = m_pullup(m, preplen); 361 if (m == NULL) { 362 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 363 return; 364 } 365 366 nip6 = mtod(m, struct ip6_hdr *); 367 nip6->ip6_src = oip6->ip6_src; 368 nip6->ip6_dst = oip6->ip6_dst; 369 370 in6_clearscope(&oip6->ip6_src); 371 in6_clearscope(&oip6->ip6_dst); 372 373 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 374 icmp6->icmp6_type = type; 375 icmp6->icmp6_code = code; 376 icmp6->icmp6_pptr = htonl((u_int32_t)param); 377 378 /* 379 * icmp6_reflect() is designed to be in the input path. 380 * icmp6_error() can be called from both input and output path, 381 * and if we are in output path rcvif could contain bogus value. 382 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 383 * information in ip header (nip6). 384 */ 385 m->m_pkthdr.rcvif = NULL; 386 387 V_icmp6stat.icp6s_outhist[type]++; 388 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 389 390 return; 391 392 freeit: 393 /* 394 * If we can't tell whether or not we can generate ICMP6, free it. 395 */ 396 m_freem(m); 397 } 398 399 /* 400 * Process a received ICMP6 message. 401 */ 402 int 403 icmp6_input(struct mbuf **mp, int *offp, int proto) 404 { 405 INIT_VNET_INET6(curvnet); 406 INIT_VPROCG(TD_TO_VPROCG(curthread)); /* XXX V_hostname needs this */ 407 struct mbuf *m = *mp, *n; 408 struct ip6_hdr *ip6, *nip6; 409 struct icmp6_hdr *icmp6, *nicmp6; 410 int off = *offp; 411 int icmp6len = m->m_pkthdr.len - *offp; 412 int code, sum, noff; 413 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 414 415 #ifndef PULLDOWN_TEST 416 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE); 417 /* m might change if M_LOOP. So, call mtod after this */ 418 #endif 419 420 /* 421 * Locate icmp6 structure in mbuf, and check 422 * that not corrupted and of at least minimum length 423 */ 424 425 ip6 = mtod(m, struct ip6_hdr *); 426 if (icmp6len < sizeof(struct icmp6_hdr)) { 427 V_icmp6stat.icp6s_tooshort++; 428 goto freeit; 429 } 430 431 /* 432 * calculate the checksum 433 */ 434 #ifndef PULLDOWN_TEST 435 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 436 #else 437 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 438 if (icmp6 == NULL) { 439 V_icmp6stat.icp6s_tooshort++; 440 return IPPROTO_DONE; 441 } 442 #endif 443 code = icmp6->icmp6_code; 444 445 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 446 nd6log((LOG_ERR, 447 "ICMP6 checksum error(%d|%x) %s\n", 448 icmp6->icmp6_type, sum, 449 ip6_sprintf(ip6bufs, &ip6->ip6_src))); 450 V_icmp6stat.icp6s_checksum++; 451 goto freeit; 452 } 453 454 if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) { 455 /* 456 * Deliver very specific ICMP6 type only. 457 * This is important to deliver TOOBIG. Otherwise PMTUD 458 * will not work. 459 */ 460 switch (icmp6->icmp6_type) { 461 case ICMP6_DST_UNREACH: 462 case ICMP6_PACKET_TOO_BIG: 463 case ICMP6_TIME_EXCEEDED: 464 break; 465 default: 466 goto freeit; 467 } 468 } 469 470 V_icmp6stat.icp6s_inhist[icmp6->icmp6_type]++; 471 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 472 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK) 473 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 474 475 switch (icmp6->icmp6_type) { 476 case ICMP6_DST_UNREACH: 477 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 478 switch (code) { 479 case ICMP6_DST_UNREACH_NOROUTE: 480 code = PRC_UNREACH_NET; 481 break; 482 case ICMP6_DST_UNREACH_ADMIN: 483 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 484 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 485 break; 486 case ICMP6_DST_UNREACH_ADDR: 487 code = PRC_HOSTDEAD; 488 break; 489 case ICMP6_DST_UNREACH_BEYONDSCOPE: 490 /* I mean "source address was incorrect." */ 491 code = PRC_PARAMPROB; 492 break; 493 case ICMP6_DST_UNREACH_NOPORT: 494 code = PRC_UNREACH_PORT; 495 break; 496 default: 497 goto badcode; 498 } 499 goto deliver; 500 break; 501 502 case ICMP6_PACKET_TOO_BIG: 503 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig); 504 505 /* validation is made in icmp6_mtudisc_update */ 506 507 code = PRC_MSGSIZE; 508 509 /* 510 * Updating the path MTU will be done after examining 511 * intermediate extension headers. 512 */ 513 goto deliver; 514 break; 515 516 case ICMP6_TIME_EXCEEDED: 517 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 518 switch (code) { 519 case ICMP6_TIME_EXCEED_TRANSIT: 520 code = PRC_TIMXCEED_INTRANS; 521 break; 522 case ICMP6_TIME_EXCEED_REASSEMBLY: 523 code = PRC_TIMXCEED_REASS; 524 break; 525 default: 526 goto badcode; 527 } 528 goto deliver; 529 break; 530 531 case ICMP6_PARAM_PROB: 532 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob); 533 switch (code) { 534 case ICMP6_PARAMPROB_NEXTHEADER: 535 code = PRC_UNREACH_PROTOCOL; 536 break; 537 case ICMP6_PARAMPROB_HEADER: 538 case ICMP6_PARAMPROB_OPTION: 539 code = PRC_PARAMPROB; 540 break; 541 default: 542 goto badcode; 543 } 544 goto deliver; 545 break; 546 547 case ICMP6_ECHO_REQUEST: 548 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 549 if (code != 0) 550 goto badcode; 551 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) { 552 /* Give up remote */ 553 break; 554 } 555 if ((n->m_flags & M_EXT) != 0 556 || n->m_len < off + sizeof(struct icmp6_hdr)) { 557 struct mbuf *n0 = n; 558 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6); 559 int n0len; 560 561 MGETHDR(n, M_DONTWAIT, n0->m_type); 562 n0len = n0->m_pkthdr.len; /* save for use below */ 563 if (n) 564 M_MOVE_PKTHDR(n, n0); 565 if (n && maxlen >= MHLEN) { 566 MCLGET(n, M_DONTWAIT); 567 if ((n->m_flags & M_EXT) == 0) { 568 m_free(n); 569 n = NULL; 570 } 571 } 572 if (n == NULL) { 573 /* Give up remote */ 574 m_freem(n0); 575 break; 576 } 577 /* 578 * Copy IPv6 and ICMPv6 only. 579 */ 580 nip6 = mtod(n, struct ip6_hdr *); 581 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 582 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 583 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 584 noff = sizeof(struct ip6_hdr); 585 /* new mbuf contains only ipv6+icmpv6 headers */ 586 n->m_len = noff + sizeof(struct icmp6_hdr); 587 /* 588 * Adjust mbuf. ip6_plen will be adjusted in 589 * ip6_output(). 590 */ 591 m_adj(n0, off + sizeof(struct icmp6_hdr)); 592 /* recalculate complete packet size */ 593 n->m_pkthdr.len = n0len + (noff - off); 594 n->m_next = n0; 595 } else { 596 nip6 = mtod(n, struct ip6_hdr *); 597 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, 598 sizeof(*nicmp6)); 599 noff = off; 600 } 601 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 602 nicmp6->icmp6_code = 0; 603 if (n) { 604 V_icmp6stat.icp6s_reflect++; 605 V_icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++; 606 icmp6_reflect(n, noff); 607 } 608 break; 609 610 case ICMP6_ECHO_REPLY: 611 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 612 if (code != 0) 613 goto badcode; 614 break; 615 616 case MLD_LISTENER_QUERY: 617 case MLD_LISTENER_REPORT: 618 if (icmp6len < sizeof(struct mld_hdr)) 619 goto badlen; 620 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */ 621 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 622 else 623 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 624 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 625 /* give up local */ 626 mld6_input(m, off); 627 m = NULL; 628 goto freeit; 629 } 630 mld6_input(n, off); 631 /* m stays. */ 632 break; 633 634 case MLD_LISTENER_DONE: 635 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 636 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */ 637 goto badlen; 638 break; /* nothing to be done in kernel */ 639 640 case MLD_MTRACE_RESP: 641 case MLD_MTRACE: 642 /* XXX: these two are experimental. not officially defined. */ 643 /* XXX: per-interface statistics? */ 644 break; /* just pass it to applications */ 645 646 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 647 { 648 enum { WRU, FQDN } mode; 649 650 if (!V_icmp6_nodeinfo) 651 break; 652 653 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 654 mode = WRU; 655 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 656 mode = FQDN; 657 else 658 goto badlen; 659 660 #define hostnamelen strlen(V_hostname) 661 if (mode == FQDN) { 662 #ifndef PULLDOWN_TEST 663 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo), 664 IPPROTO_DONE); 665 #endif 666 n = m_copy(m, 0, M_COPYALL); 667 if (n) 668 n = ni6_input(n, off); 669 /* XXX meaningless if n == NULL */ 670 noff = sizeof(struct ip6_hdr); 671 } else { 672 u_char *p; 673 int maxlen, maxhlen; 674 675 /* 676 * XXX: this combination of flags is pointless, 677 * but should we keep this for compatibility? 678 */ 679 if ((V_icmp6_nodeinfo & 5) != 5) 680 break; 681 682 if (code != 0) 683 goto badcode; 684 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; 685 if (maxlen >= MCLBYTES) { 686 /* Give up remote */ 687 break; 688 } 689 MGETHDR(n, M_DONTWAIT, m->m_type); 690 if (n && maxlen > MHLEN) { 691 MCLGET(n, M_DONTWAIT); 692 if ((n->m_flags & M_EXT) == 0) { 693 m_free(n); 694 n = NULL; 695 } 696 } 697 if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) { 698 /* 699 * Previous code did a blind M_COPY_PKTHDR 700 * and said "just for rcvif". If true, then 701 * we could tolerate the dup failing (due to 702 * the deep copy of the tag chain). For now 703 * be conservative and just fail. 704 */ 705 m_free(n); 706 n = NULL; 707 } 708 if (n == NULL) { 709 /* Give up remote */ 710 break; 711 } 712 n->m_pkthdr.rcvif = NULL; 713 n->m_len = 0; 714 maxhlen = M_TRAILINGSPACE(n) - maxlen; 715 mtx_lock(&hostname_mtx); 716 if (maxhlen > hostnamelen) 717 maxhlen = hostnamelen; 718 /* 719 * Copy IPv6 and ICMPv6 only. 720 */ 721 nip6 = mtod(n, struct ip6_hdr *); 722 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 723 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 724 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 725 p = (u_char *)(nicmp6 + 1); 726 bzero(p, 4); 727 bcopy(V_hostname, p + 4, maxhlen); /* meaningless TTL */ 728 mtx_unlock(&hostname_mtx); 729 noff = sizeof(struct ip6_hdr); 730 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 731 sizeof(struct icmp6_hdr) + 4 + maxhlen; 732 nicmp6->icmp6_type = ICMP6_WRUREPLY; 733 nicmp6->icmp6_code = 0; 734 } 735 #undef hostnamelen 736 if (n) { 737 V_icmp6stat.icp6s_reflect++; 738 V_icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++; 739 icmp6_reflect(n, noff); 740 } 741 break; 742 } 743 744 case ICMP6_WRUREPLY: 745 if (code != 0) 746 goto badcode; 747 break; 748 749 case ND_ROUTER_SOLICIT: 750 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 751 if (code != 0) 752 goto badcode; 753 if (icmp6len < sizeof(struct nd_router_solicit)) 754 goto badlen; 755 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 756 /* give up local */ 757 nd6_rs_input(m, off, icmp6len); 758 m = NULL; 759 goto freeit; 760 } 761 nd6_rs_input(n, off, icmp6len); 762 /* m stays. */ 763 break; 764 765 case ND_ROUTER_ADVERT: 766 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 767 if (code != 0) 768 goto badcode; 769 if (icmp6len < sizeof(struct nd_router_advert)) 770 goto badlen; 771 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 772 /* give up local */ 773 nd6_ra_input(m, off, icmp6len); 774 m = NULL; 775 goto freeit; 776 } 777 nd6_ra_input(n, off, icmp6len); 778 /* m stays. */ 779 break; 780 781 case ND_NEIGHBOR_SOLICIT: 782 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 783 if (code != 0) 784 goto badcode; 785 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 786 goto badlen; 787 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 788 /* give up local */ 789 nd6_ns_input(m, off, icmp6len); 790 m = NULL; 791 goto freeit; 792 } 793 nd6_ns_input(n, off, icmp6len); 794 /* m stays. */ 795 break; 796 797 case ND_NEIGHBOR_ADVERT: 798 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 799 if (code != 0) 800 goto badcode; 801 if (icmp6len < sizeof(struct nd_neighbor_advert)) 802 goto badlen; 803 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 804 /* give up local */ 805 nd6_na_input(m, off, icmp6len); 806 m = NULL; 807 goto freeit; 808 } 809 nd6_na_input(n, off, icmp6len); 810 /* m stays. */ 811 break; 812 813 case ND_REDIRECT: 814 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 815 if (code != 0) 816 goto badcode; 817 if (icmp6len < sizeof(struct nd_redirect)) 818 goto badlen; 819 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 820 /* give up local */ 821 icmp6_redirect_input(m, off); 822 m = NULL; 823 goto freeit; 824 } 825 icmp6_redirect_input(n, off); 826 /* m stays. */ 827 break; 828 829 case ICMP6_ROUTER_RENUMBERING: 830 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 831 code != ICMP6_ROUTER_RENUMBERING_RESULT) 832 goto badcode; 833 if (icmp6len < sizeof(struct icmp6_router_renum)) 834 goto badlen; 835 break; 836 837 default: 838 nd6log((LOG_DEBUG, 839 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 840 icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src), 841 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 842 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 843 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 844 /* ICMPv6 error: MUST deliver it by spec... */ 845 code = PRC_NCMDS; 846 /* deliver */ 847 } else { 848 /* ICMPv6 informational: MUST not deliver */ 849 break; 850 } 851 deliver: 852 if (icmp6_notify_error(&m, off, icmp6len, code)) { 853 /* In this case, m should've been freed. */ 854 return (IPPROTO_DONE); 855 } 856 break; 857 858 badcode: 859 V_icmp6stat.icp6s_badcode++; 860 break; 861 862 badlen: 863 V_icmp6stat.icp6s_badlen++; 864 break; 865 } 866 867 /* deliver the packet to appropriate sockets */ 868 icmp6_rip6_input(&m, *offp); 869 870 return IPPROTO_DONE; 871 872 freeit: 873 m_freem(m); 874 return IPPROTO_DONE; 875 } 876 877 static int 878 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code) 879 { 880 INIT_VNET_INET6(curvnet); 881 struct mbuf *m = *mp; 882 struct icmp6_hdr *icmp6; 883 struct ip6_hdr *eip6; 884 u_int32_t notifymtu; 885 struct sockaddr_in6 icmp6src, icmp6dst; 886 887 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 888 V_icmp6stat.icp6s_tooshort++; 889 goto freeit; 890 } 891 #ifndef PULLDOWN_TEST 892 IP6_EXTHDR_CHECK(m, off, 893 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1); 894 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 895 #else 896 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 897 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 898 if (icmp6 == NULL) { 899 V_icmp6stat.icp6s_tooshort++; 900 return (-1); 901 } 902 #endif 903 eip6 = (struct ip6_hdr *)(icmp6 + 1); 904 905 /* Detect the upper level protocol */ 906 { 907 void (*ctlfunc)(int, struct sockaddr *, void *); 908 u_int8_t nxt = eip6->ip6_nxt; 909 int eoff = off + sizeof(struct icmp6_hdr) + 910 sizeof(struct ip6_hdr); 911 struct ip6ctlparam ip6cp; 912 struct in6_addr *finaldst = NULL; 913 int icmp6type = icmp6->icmp6_type; 914 struct ip6_frag *fh; 915 struct ip6_rthdr *rth; 916 struct ip6_rthdr0 *rth0; 917 int rthlen; 918 919 while (1) { /* XXX: should avoid infinite loop explicitly? */ 920 struct ip6_ext *eh; 921 922 switch (nxt) { 923 case IPPROTO_HOPOPTS: 924 case IPPROTO_DSTOPTS: 925 case IPPROTO_AH: 926 #ifndef PULLDOWN_TEST 927 IP6_EXTHDR_CHECK(m, 0, 928 eoff + sizeof(struct ip6_ext), -1); 929 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff); 930 #else 931 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 932 eoff, sizeof(*eh)); 933 if (eh == NULL) { 934 V_icmp6stat.icp6s_tooshort++; 935 return (-1); 936 } 937 #endif 938 939 if (nxt == IPPROTO_AH) 940 eoff += (eh->ip6e_len + 2) << 2; 941 else 942 eoff += (eh->ip6e_len + 1) << 3; 943 nxt = eh->ip6e_nxt; 944 break; 945 case IPPROTO_ROUTING: 946 /* 947 * When the erroneous packet contains a 948 * routing header, we should examine the 949 * header to determine the final destination. 950 * Otherwise, we can't properly update 951 * information that depends on the final 952 * destination (e.g. path MTU). 953 */ 954 #ifndef PULLDOWN_TEST 955 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1); 956 rth = (struct ip6_rthdr *) 957 (mtod(m, caddr_t) + eoff); 958 #else 959 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 960 eoff, sizeof(*rth)); 961 if (rth == NULL) { 962 V_icmp6stat.icp6s_tooshort++; 963 return (-1); 964 } 965 #endif 966 rthlen = (rth->ip6r_len + 1) << 3; 967 /* 968 * XXX: currently there is no 969 * officially defined type other 970 * than type-0. 971 * Note that if the segment left field 972 * is 0, all intermediate hops must 973 * have been passed. 974 */ 975 if (rth->ip6r_segleft && 976 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 977 int hops; 978 979 #ifndef PULLDOWN_TEST 980 IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1); 981 rth0 = (struct ip6_rthdr0 *) 982 (mtod(m, caddr_t) + eoff); 983 #else 984 IP6_EXTHDR_GET(rth0, 985 struct ip6_rthdr0 *, m, 986 eoff, rthlen); 987 if (rth0 == NULL) { 988 V_icmp6stat.icp6s_tooshort++; 989 return (-1); 990 } 991 #endif 992 /* just ignore a bogus header */ 993 if ((rth0->ip6r0_len % 2) == 0 && 994 (hops = rth0->ip6r0_len/2)) 995 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 996 } 997 eoff += rthlen; 998 nxt = rth->ip6r_nxt; 999 break; 1000 case IPPROTO_FRAGMENT: 1001 #ifndef PULLDOWN_TEST 1002 IP6_EXTHDR_CHECK(m, 0, eoff + 1003 sizeof(struct ip6_frag), -1); 1004 fh = (struct ip6_frag *)(mtod(m, caddr_t) + 1005 eoff); 1006 #else 1007 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 1008 eoff, sizeof(*fh)); 1009 if (fh == NULL) { 1010 V_icmp6stat.icp6s_tooshort++; 1011 return (-1); 1012 } 1013 #endif 1014 /* 1015 * Data after a fragment header is meaningless 1016 * unless it is the first fragment, but 1017 * we'll go to the notify label for path MTU 1018 * discovery. 1019 */ 1020 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1021 goto notify; 1022 1023 eoff += sizeof(struct ip6_frag); 1024 nxt = fh->ip6f_nxt; 1025 break; 1026 default: 1027 /* 1028 * This case includes ESP and the No Next 1029 * Header. In such cases going to the notify 1030 * label does not have any meaning 1031 * (i.e. ctlfunc will be NULL), but we go 1032 * anyway since we might have to update 1033 * path MTU information. 1034 */ 1035 goto notify; 1036 } 1037 } 1038 notify: 1039 #ifndef PULLDOWN_TEST 1040 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 1041 #else 1042 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1043 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1044 if (icmp6 == NULL) { 1045 V_icmp6stat.icp6s_tooshort++; 1046 return (-1); 1047 } 1048 #endif 1049 1050 /* 1051 * retrieve parameters from the inner IPv6 header, and convert 1052 * them into sockaddr structures. 1053 * XXX: there is no guarantee that the source or destination 1054 * addresses of the inner packet are in the same scope as 1055 * the addresses of the icmp packet. But there is no other 1056 * way to determine the zone. 1057 */ 1058 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1059 1060 bzero(&icmp6dst, sizeof(icmp6dst)); 1061 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1062 icmp6dst.sin6_family = AF_INET6; 1063 if (finaldst == NULL) 1064 icmp6dst.sin6_addr = eip6->ip6_dst; 1065 else 1066 icmp6dst.sin6_addr = *finaldst; 1067 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1068 goto freeit; 1069 bzero(&icmp6src, sizeof(icmp6src)); 1070 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1071 icmp6src.sin6_family = AF_INET6; 1072 icmp6src.sin6_addr = eip6->ip6_src; 1073 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL)) 1074 goto freeit; 1075 icmp6src.sin6_flowinfo = 1076 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1077 1078 if (finaldst == NULL) 1079 finaldst = &eip6->ip6_dst; 1080 ip6cp.ip6c_m = m; 1081 ip6cp.ip6c_icmp6 = icmp6; 1082 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1083 ip6cp.ip6c_off = eoff; 1084 ip6cp.ip6c_finaldst = finaldst; 1085 ip6cp.ip6c_src = &icmp6src; 1086 ip6cp.ip6c_nxt = nxt; 1087 1088 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1089 notifymtu = ntohl(icmp6->icmp6_mtu); 1090 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1091 icmp6_mtudisc_update(&ip6cp, 1); /*XXX*/ 1092 } 1093 1094 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1095 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1096 if (ctlfunc) { 1097 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1098 &ip6cp); 1099 } 1100 } 1101 *mp = m; 1102 return (0); 1103 1104 freeit: 1105 m_freem(m); 1106 return (-1); 1107 } 1108 1109 void 1110 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1111 { 1112 INIT_VNET_INET6(curvnet); 1113 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1114 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1115 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1116 u_int mtu = ntohl(icmp6->icmp6_mtu); 1117 struct in_conninfo inc; 1118 1119 #if 0 1120 /* 1121 * RFC2460 section 5, last paragraph. 1122 * even though minimum link MTU for IPv6 is IPV6_MMTU, 1123 * we may see ICMPv6 too big with mtu < IPV6_MMTU 1124 * due to packet translator in the middle. 1125 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for 1126 * special handling. 1127 */ 1128 if (mtu < IPV6_MMTU) 1129 return; 1130 #endif 1131 1132 /* 1133 * we reject ICMPv6 too big with abnormally small value. 1134 * XXX what is the good definition of "abnormally small"? 1135 */ 1136 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) 1137 return; 1138 1139 if (!validated) 1140 return; 1141 1142 /* 1143 * In case the suggested mtu is less than IPV6_MMTU, we 1144 * only need to remember that it was for above mentioned 1145 * "alwaysfrag" case. 1146 * Try to be as close to the spec as possible. 1147 */ 1148 if (mtu < IPV6_MMTU) 1149 mtu = IPV6_MMTU - 8; 1150 1151 bzero(&inc, sizeof(inc)); 1152 inc.inc_flags |= INC_ISIPV6; 1153 inc.inc6_faddr = *dst; 1154 if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL)) 1155 return; 1156 1157 if (mtu < tcp_maxmtu6(&inc, NULL)) { 1158 tcp_hc_updatemtu(&inc, mtu); 1159 V_icmp6stat.icp6s_pmtuchg++; 1160 } 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 #define hostnamelen strlen(V_hostname) 1175 static struct mbuf * 1176 ni6_input(struct mbuf *m, int off) 1177 { 1178 INIT_VNET_INET6(curvnet); 1179 INIT_VPROCG(TD_TO_VPROCG(curthread)); /* XXX V_hostname needs this */ 1180 struct icmp6_nodeinfo *ni6, *nni6; 1181 struct mbuf *n = NULL; 1182 u_int16_t qtype; 1183 int subjlen; 1184 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1185 struct ni_reply_fqdn *fqdn; 1186 int addrs; /* for NI_QTYPE_NODEADDR */ 1187 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1188 struct in6_addr in6_subj; /* subject address */ 1189 struct ip6_hdr *ip6; 1190 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1191 char *subj = NULL; 1192 struct in6_ifaddr *ia6 = NULL; 1193 1194 ip6 = mtod(m, struct ip6_hdr *); 1195 #ifndef PULLDOWN_TEST 1196 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off); 1197 #else 1198 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1199 if (ni6 == NULL) { 1200 /* m is already reclaimed */ 1201 return (NULL); 1202 } 1203 #endif 1204 1205 /* 1206 * Validate IPv6 source address. 1207 * The default configuration MUST be to refuse answering queries from 1208 * global-scope addresses according to RFC4602. 1209 * Notes: 1210 * - it's not very clear what "refuse" means; this implementation 1211 * simply drops it. 1212 * - it's not very easy to identify global-scope (unicast) addresses 1213 * since there are many prefixes for them. It should be safer 1214 * and in practice sufficient to check "all" but loopback and 1215 * link-local (note that site-local unicast was deprecated and 1216 * ULA is defined as global scope-wise) 1217 */ 1218 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 && 1219 !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) && 1220 !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) 1221 goto bad; 1222 1223 /* 1224 * Validate IPv6 destination address. 1225 * 1226 * The Responder must discard the Query without further processing 1227 * unless it is one of the Responder's unicast or anycast addresses, or 1228 * a link-local scope multicast address which the Responder has joined. 1229 * [RFC4602, Section 5.] 1230 */ 1231 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { 1232 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst)) 1233 goto bad; 1234 /* else it's a link-local multicast, fine */ 1235 } else { /* unicast or anycast */ 1236 if ((ia6 = ip6_getdstifaddr(m)) == NULL) 1237 goto bad; /* XXX impossible */ 1238 1239 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) && 1240 !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) { 1241 nd6log((LOG_DEBUG, "ni6_input: ignore node info to " 1242 "a temporary address in %s:%d", 1243 __FILE__, __LINE__)); 1244 goto bad; 1245 } 1246 } 1247 1248 /* validate query Subject field. */ 1249 qtype = ntohs(ni6->ni_qtype); 1250 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1251 switch (qtype) { 1252 case NI_QTYPE_NOOP: 1253 case NI_QTYPE_SUPTYPES: 1254 /* 07 draft */ 1255 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1256 break; 1257 /* FALLTHROUGH */ 1258 case NI_QTYPE_FQDN: 1259 case NI_QTYPE_NODEADDR: 1260 case NI_QTYPE_IPV4ADDR: 1261 switch (ni6->ni_code) { 1262 case ICMP6_NI_SUBJ_IPV6: 1263 #if ICMP6_NI_SUBJ_IPV6 != 0 1264 case 0: 1265 #endif 1266 /* 1267 * backward compatibility - try to accept 03 draft 1268 * format, where no Subject is present. 1269 */ 1270 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1271 subjlen == 0) { 1272 oldfqdn++; 1273 break; 1274 } 1275 #if ICMP6_NI_SUBJ_IPV6 != 0 1276 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1277 goto bad; 1278 #endif 1279 1280 if (subjlen != sizeof(struct in6_addr)) 1281 goto bad; 1282 1283 /* 1284 * Validate Subject address. 1285 * 1286 * Not sure what exactly "address belongs to the node" 1287 * means in the spec, is it just unicast, or what? 1288 * 1289 * At this moment we consider Subject address as 1290 * "belong to the node" if the Subject address equals 1291 * to the IPv6 destination address; validation for 1292 * IPv6 destination address should have done enough 1293 * check for us. 1294 * 1295 * We do not do proxy at this moment. 1296 */ 1297 /* m_pulldown instead of copy? */ 1298 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1299 subjlen, (caddr_t)&in6_subj); 1300 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL)) 1301 goto bad; 1302 1303 subj = (char *)&in6_subj; 1304 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1305 break; 1306 1307 /* 1308 * XXX if we are to allow other cases, we should really 1309 * be careful about scope here. 1310 * basically, we should disallow queries toward IPv6 1311 * destination X with subject Y, 1312 * if scope(X) > scope(Y). 1313 * if we allow scope(X) > scope(Y), it will result in 1314 * information leakage across scope boundary. 1315 */ 1316 goto bad; 1317 1318 case ICMP6_NI_SUBJ_FQDN: 1319 /* 1320 * Validate Subject name with gethostname(3). 1321 * 1322 * The behavior may need some debate, since: 1323 * - we are not sure if the node has FQDN as 1324 * hostname (returned by gethostname(3)). 1325 * - the code does wildcard match for truncated names. 1326 * however, we are not sure if we want to perform 1327 * wildcard match, if gethostname(3) side has 1328 * truncated hostname. 1329 */ 1330 mtx_lock(&hostname_mtx); 1331 n = ni6_nametodns(V_hostname, hostnamelen, 0); 1332 mtx_unlock(&hostname_mtx); 1333 if (!n || n->m_next || n->m_len == 0) 1334 goto bad; 1335 IP6_EXTHDR_GET(subj, char *, m, 1336 off + sizeof(struct icmp6_nodeinfo), subjlen); 1337 if (subj == NULL) 1338 goto bad; 1339 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1340 n->m_len)) { 1341 goto bad; 1342 } 1343 m_freem(n); 1344 n = NULL; 1345 break; 1346 1347 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1348 default: 1349 goto bad; 1350 } 1351 break; 1352 } 1353 1354 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1355 switch (qtype) { 1356 case NI_QTYPE_FQDN: 1357 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0) 1358 goto bad; 1359 break; 1360 case NI_QTYPE_NODEADDR: 1361 case NI_QTYPE_IPV4ADDR: 1362 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0) 1363 goto bad; 1364 break; 1365 } 1366 1367 /* guess reply length */ 1368 switch (qtype) { 1369 case NI_QTYPE_NOOP: 1370 break; /* no reply data */ 1371 case NI_QTYPE_SUPTYPES: 1372 replylen += sizeof(u_int32_t); 1373 break; 1374 case NI_QTYPE_FQDN: 1375 /* XXX will append an mbuf */ 1376 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1377 break; 1378 case NI_QTYPE_NODEADDR: 1379 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj); 1380 if ((replylen += addrs * (sizeof(struct in6_addr) + 1381 sizeof(u_int32_t))) > MCLBYTES) 1382 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1383 break; 1384 case NI_QTYPE_IPV4ADDR: 1385 /* unsupported - should respond with unknown Qtype? */ 1386 break; 1387 default: 1388 /* 1389 * XXX: We must return a reply with the ICMP6 code 1390 * `unknown Qtype' in this case. However we regard the case 1391 * as an FQDN query for backward compatibility. 1392 * Older versions set a random value to this field, 1393 * so it rarely varies in the defined qtypes. 1394 * But the mechanism is not reliable... 1395 * maybe we should obsolete older versions. 1396 */ 1397 qtype = NI_QTYPE_FQDN; 1398 /* XXX will append an mbuf */ 1399 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1400 oldfqdn++; 1401 break; 1402 } 1403 1404 /* allocate an mbuf to reply. */ 1405 MGETHDR(n, M_DONTWAIT, m->m_type); 1406 if (n == NULL) { 1407 m_freem(m); 1408 return (NULL); 1409 } 1410 M_MOVE_PKTHDR(n, m); /* just for recvif */ 1411 if (replylen > MHLEN) { 1412 if (replylen > MCLBYTES) { 1413 /* 1414 * XXX: should we try to allocate more? But MCLBYTES 1415 * is probably much larger than IPV6_MMTU... 1416 */ 1417 goto bad; 1418 } 1419 MCLGET(n, M_DONTWAIT); 1420 if ((n->m_flags & M_EXT) == 0) { 1421 goto bad; 1422 } 1423 } 1424 n->m_pkthdr.len = n->m_len = replylen; 1425 1426 /* copy mbuf header and IPv6 + Node Information base headers */ 1427 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1428 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1429 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1430 1431 /* qtype dependent procedure */ 1432 switch (qtype) { 1433 case NI_QTYPE_NOOP: 1434 nni6->ni_code = ICMP6_NI_SUCCESS; 1435 nni6->ni_flags = 0; 1436 break; 1437 case NI_QTYPE_SUPTYPES: 1438 { 1439 u_int32_t v; 1440 nni6->ni_code = ICMP6_NI_SUCCESS; 1441 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1442 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1443 v = (u_int32_t)htonl(0x0000000f); 1444 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1445 break; 1446 } 1447 case NI_QTYPE_FQDN: 1448 nni6->ni_code = ICMP6_NI_SUCCESS; 1449 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1450 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo)); 1451 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1452 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1453 /* 1454 * XXX do we really have FQDN in variable "hostname"? 1455 */ 1456 mtx_lock(&hostname_mtx); 1457 n->m_next = ni6_nametodns(V_hostname, hostnamelen, oldfqdn); 1458 mtx_unlock(&hostname_mtx); 1459 if (n->m_next == NULL) 1460 goto bad; 1461 /* XXX we assume that n->m_next is not a chain */ 1462 if (n->m_next->m_next != NULL) 1463 goto bad; 1464 n->m_pkthdr.len += n->m_next->m_len; 1465 break; 1466 case NI_QTYPE_NODEADDR: 1467 { 1468 int lenlim, copied; 1469 1470 nni6->ni_code = ICMP6_NI_SUCCESS; 1471 n->m_pkthdr.len = n->m_len = 1472 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1473 lenlim = M_TRAILINGSPACE(n); 1474 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1475 /* XXX: reset mbuf length */ 1476 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1477 sizeof(struct icmp6_nodeinfo) + copied; 1478 break; 1479 } 1480 default: 1481 break; /* XXX impossible! */ 1482 } 1483 1484 nni6->ni_type = ICMP6_NI_REPLY; 1485 m_freem(m); 1486 return (n); 1487 1488 bad: 1489 m_freem(m); 1490 if (n) 1491 m_freem(n); 1492 return (NULL); 1493 } 1494 #undef hostnamelen 1495 1496 /* 1497 * make a mbuf with DNS-encoded string. no compression support. 1498 * 1499 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1500 * treated as truncated name (two \0 at the end). this is a wild guess. 1501 * 1502 * old - return pascal string if non-zero 1503 */ 1504 static struct mbuf * 1505 ni6_nametodns(const char *name, int namelen, int old) 1506 { 1507 struct mbuf *m; 1508 char *cp, *ep; 1509 const char *p, *q; 1510 int i, len, nterm; 1511 1512 if (old) 1513 len = namelen + 1; 1514 else 1515 len = MCLBYTES; 1516 1517 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1518 MGET(m, M_DONTWAIT, MT_DATA); 1519 if (m && len > MLEN) { 1520 MCLGET(m, M_DONTWAIT); 1521 if ((m->m_flags & M_EXT) == 0) 1522 goto fail; 1523 } 1524 if (!m) 1525 goto fail; 1526 m->m_next = NULL; 1527 1528 if (old) { 1529 m->m_len = len; 1530 *mtod(m, char *) = namelen; 1531 bcopy(name, mtod(m, char *) + 1, namelen); 1532 return m; 1533 } else { 1534 m->m_len = 0; 1535 cp = mtod(m, char *); 1536 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1537 1538 /* if not certain about my name, return empty buffer */ 1539 if (namelen == 0) 1540 return m; 1541 1542 /* 1543 * guess if it looks like shortened hostname, or FQDN. 1544 * shortened hostname needs two trailing "\0". 1545 */ 1546 i = 0; 1547 for (p = name; p < name + namelen; p++) { 1548 if (*p && *p == '.') 1549 i++; 1550 } 1551 if (i < 2) 1552 nterm = 2; 1553 else 1554 nterm = 1; 1555 1556 p = name; 1557 while (cp < ep && p < name + namelen) { 1558 i = 0; 1559 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1560 i++; 1561 /* result does not fit into mbuf */ 1562 if (cp + i + 1 >= ep) 1563 goto fail; 1564 /* 1565 * DNS label length restriction, RFC1035 page 8. 1566 * "i == 0" case is included here to avoid returning 1567 * 0-length label on "foo..bar". 1568 */ 1569 if (i <= 0 || i >= 64) 1570 goto fail; 1571 *cp++ = i; 1572 bcopy(p, cp, i); 1573 cp += i; 1574 p = q; 1575 if (p < name + namelen && *p == '.') 1576 p++; 1577 } 1578 /* termination */ 1579 if (cp + nterm >= ep) 1580 goto fail; 1581 while (nterm-- > 0) 1582 *cp++ = '\0'; 1583 m->m_len = cp - mtod(m, char *); 1584 return m; 1585 } 1586 1587 panic("should not reach here"); 1588 /* NOTREACHED */ 1589 1590 fail: 1591 if (m) 1592 m_freem(m); 1593 return NULL; 1594 } 1595 1596 /* 1597 * check if two DNS-encoded string matches. takes care of truncated 1598 * form (with \0\0 at the end). no compression support. 1599 * XXX upper/lowercase match (see RFC2065) 1600 */ 1601 static int 1602 ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1603 { 1604 const char *a0, *b0; 1605 int l; 1606 1607 /* simplest case - need validation? */ 1608 if (alen == blen && bcmp(a, b, alen) == 0) 1609 return 1; 1610 1611 a0 = a; 1612 b0 = b; 1613 1614 /* termination is mandatory */ 1615 if (alen < 2 || blen < 2) 1616 return 0; 1617 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1618 return 0; 1619 alen--; 1620 blen--; 1621 1622 while (a - a0 < alen && b - b0 < blen) { 1623 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1624 return 0; 1625 1626 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1627 return 0; 1628 /* we don't support compression yet */ 1629 if (a[0] >= 64 || b[0] >= 64) 1630 return 0; 1631 1632 /* truncated case */ 1633 if (a[0] == 0 && a - a0 == alen - 1) 1634 return 1; 1635 if (b[0] == 0 && b - b0 == blen - 1) 1636 return 1; 1637 if (a[0] == 0 || b[0] == 0) 1638 return 0; 1639 1640 if (a[0] != b[0]) 1641 return 0; 1642 l = a[0]; 1643 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1644 return 0; 1645 if (bcmp(a + 1, b + 1, l) != 0) 1646 return 0; 1647 1648 a += 1 + l; 1649 b += 1 + l; 1650 } 1651 1652 if (a - a0 == alen && b - b0 == blen) 1653 return 1; 1654 else 1655 return 0; 1656 } 1657 1658 /* 1659 * calculate the number of addresses to be returned in the node info reply. 1660 */ 1661 static int 1662 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp, 1663 struct in6_addr *subj) 1664 { 1665 INIT_VNET_NET(curvnet); 1666 INIT_VNET_INET6(curvnet); 1667 struct ifnet *ifp; 1668 struct in6_ifaddr *ifa6; 1669 struct ifaddr *ifa; 1670 int addrs = 0, addrsofif, iffound = 0; 1671 int niflags = ni6->ni_flags; 1672 1673 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1674 switch (ni6->ni_code) { 1675 case ICMP6_NI_SUBJ_IPV6: 1676 if (subj == NULL) /* must be impossible... */ 1677 return (0); 1678 break; 1679 default: 1680 /* 1681 * XXX: we only support IPv6 subject address for 1682 * this Qtype. 1683 */ 1684 return (0); 1685 } 1686 } 1687 1688 IFNET_RLOCK(); 1689 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) { 1690 addrsofif = 0; 1691 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) { 1692 if (ifa->ifa_addr->sa_family != AF_INET6) 1693 continue; 1694 ifa6 = (struct in6_ifaddr *)ifa; 1695 1696 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1697 IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr)) 1698 iffound = 1; 1699 1700 /* 1701 * IPv4-mapped addresses can only be returned by a 1702 * Node Information proxy, since they represent 1703 * addresses of IPv4-only nodes, which perforce do 1704 * not implement this protocol. 1705 * [icmp-name-lookups-07, Section 5.4] 1706 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1707 * this function at this moment. 1708 */ 1709 1710 /* What do we have to do about ::1? */ 1711 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1712 case IPV6_ADDR_SCOPE_LINKLOCAL: 1713 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1714 continue; 1715 break; 1716 case IPV6_ADDR_SCOPE_SITELOCAL: 1717 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1718 continue; 1719 break; 1720 case IPV6_ADDR_SCOPE_GLOBAL: 1721 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1722 continue; 1723 break; 1724 default: 1725 continue; 1726 } 1727 1728 /* 1729 * check if anycast is okay. 1730 * XXX: just experimental. not in the spec. 1731 */ 1732 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1733 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1734 continue; /* we need only unicast addresses */ 1735 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1736 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) { 1737 continue; 1738 } 1739 addrsofif++; /* count the address */ 1740 } 1741 if (iffound) { 1742 *ifpp = ifp; 1743 IFNET_RUNLOCK(); 1744 return (addrsofif); 1745 } 1746 1747 addrs += addrsofif; 1748 } 1749 IFNET_RUNLOCK(); 1750 1751 return (addrs); 1752 } 1753 1754 static int 1755 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6, 1756 struct ifnet *ifp0, int resid) 1757 { 1758 INIT_VNET_NET(curvnet); 1759 INIT_VNET_INET6(curvnet); 1760 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet); 1761 struct in6_ifaddr *ifa6; 1762 struct ifaddr *ifa; 1763 struct ifnet *ifp_dep = NULL; 1764 int copied = 0, allow_deprecated = 0; 1765 u_char *cp = (u_char *)(nni6 + 1); 1766 int niflags = ni6->ni_flags; 1767 u_int32_t ltime; 1768 1769 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1770 return (0); /* needless to copy */ 1771 1772 IFNET_RLOCK(); 1773 again: 1774 1775 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) { 1776 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1777 ifa = ifa->ifa_list.tqe_next) { 1778 if (ifa->ifa_addr->sa_family != AF_INET6) 1779 continue; 1780 ifa6 = (struct in6_ifaddr *)ifa; 1781 1782 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1783 allow_deprecated == 0) { 1784 /* 1785 * prefererred address should be put before 1786 * deprecated addresses. 1787 */ 1788 1789 /* record the interface for later search */ 1790 if (ifp_dep == NULL) 1791 ifp_dep = ifp; 1792 1793 continue; 1794 } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1795 allow_deprecated != 0) 1796 continue; /* we now collect deprecated addrs */ 1797 1798 /* What do we have to do about ::1? */ 1799 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1800 case IPV6_ADDR_SCOPE_LINKLOCAL: 1801 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1802 continue; 1803 break; 1804 case IPV6_ADDR_SCOPE_SITELOCAL: 1805 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1806 continue; 1807 break; 1808 case IPV6_ADDR_SCOPE_GLOBAL: 1809 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1810 continue; 1811 break; 1812 default: 1813 continue; 1814 } 1815 1816 /* 1817 * check if anycast is okay. 1818 * XXX: just experimental. not in the spec. 1819 */ 1820 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1821 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1822 continue; 1823 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1824 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) { 1825 continue; 1826 } 1827 1828 /* now we can copy the address */ 1829 if (resid < sizeof(struct in6_addr) + 1830 sizeof(u_int32_t)) { 1831 /* 1832 * We give up much more copy. 1833 * Set the truncate flag and return. 1834 */ 1835 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1836 IFNET_RUNLOCK(); 1837 return (copied); 1838 } 1839 1840 /* 1841 * Set the TTL of the address. 1842 * The TTL value should be one of the following 1843 * according to the specification: 1844 * 1845 * 1. The remaining lifetime of a DHCP lease on the 1846 * address, or 1847 * 2. The remaining Valid Lifetime of a prefix from 1848 * which the address was derived through Stateless 1849 * Autoconfiguration. 1850 * 1851 * Note that we currently do not support stateful 1852 * address configuration by DHCPv6, so the former 1853 * case can't happen. 1854 */ 1855 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1856 ltime = ND6_INFINITE_LIFETIME; 1857 else { 1858 if (ifa6->ia6_lifetime.ia6t_expire > 1859 time_second) 1860 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second); 1861 else 1862 ltime = 0; 1863 } 1864 1865 bcopy(<ime, cp, sizeof(u_int32_t)); 1866 cp += sizeof(u_int32_t); 1867 1868 /* copy the address itself */ 1869 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1870 sizeof(struct in6_addr)); 1871 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1872 cp += sizeof(struct in6_addr); 1873 1874 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1875 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1876 } 1877 if (ifp0) /* we need search only on the specified IF */ 1878 break; 1879 } 1880 1881 if (allow_deprecated == 0 && ifp_dep != NULL) { 1882 ifp = ifp_dep; 1883 allow_deprecated = 1; 1884 1885 goto again; 1886 } 1887 1888 IFNET_RUNLOCK(); 1889 1890 return (copied); 1891 } 1892 1893 /* 1894 * XXX almost dup'ed code with rip6_input. 1895 */ 1896 static int 1897 icmp6_rip6_input(struct mbuf **mp, int off) 1898 { 1899 INIT_VNET_INET(curvnet); 1900 INIT_VNET_INET6(curvnet); 1901 struct mbuf *m = *mp; 1902 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1903 struct inpcb *in6p; 1904 struct inpcb *last = NULL; 1905 struct sockaddr_in6 fromsa; 1906 struct icmp6_hdr *icmp6; 1907 struct mbuf *opts = NULL; 1908 1909 #ifndef PULLDOWN_TEST 1910 /* this is assumed to be safe. */ 1911 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 1912 #else 1913 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1914 if (icmp6 == NULL) { 1915 /* m is already reclaimed */ 1916 return (IPPROTO_DONE); 1917 } 1918 #endif 1919 1920 /* 1921 * XXX: the address may have embedded scope zone ID, which should be 1922 * hidden from applications. 1923 */ 1924 bzero(&fromsa, sizeof(fromsa)); 1925 fromsa.sin6_family = AF_INET6; 1926 fromsa.sin6_len = sizeof(struct sockaddr_in6); 1927 fromsa.sin6_addr = ip6->ip6_src; 1928 if (sa6_recoverscope(&fromsa)) { 1929 m_freem(m); 1930 return (IPPROTO_DONE); 1931 } 1932 1933 INP_INFO_RLOCK(&V_ripcbinfo); 1934 LIST_FOREACH(in6p, &V_ripcb, inp_list) { 1935 if ((in6p->inp_vflag & INP_IPV6) == 0) 1936 continue; 1937 if (in6p->inp_ip_p != IPPROTO_ICMPV6) 1938 continue; 1939 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1940 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1941 continue; 1942 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1943 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1944 continue; 1945 INP_RLOCK(in6p); 1946 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1947 in6p->in6p_icmp6filt)) { 1948 INP_RUNLOCK(in6p); 1949 continue; 1950 } 1951 if (last != NULL) { 1952 struct mbuf *n = NULL; 1953 1954 /* 1955 * Recent network drivers tend to allocate a single 1956 * mbuf cluster, rather than to make a couple of 1957 * mbufs without clusters. Also, since the IPv6 code 1958 * path tries to avoid m_pullup(), it is highly 1959 * probable that we still have an mbuf cluster here 1960 * even though the necessary length can be stored in an 1961 * mbuf's internal buffer. 1962 * Meanwhile, the default size of the receive socket 1963 * buffer for raw sockets is not so large. This means 1964 * the possibility of packet loss is relatively higher 1965 * than before. To avoid this scenario, we copy the 1966 * received data to a separate mbuf that does not use 1967 * a cluster, if possible. 1968 * XXX: it is better to copy the data after stripping 1969 * intermediate headers. 1970 */ 1971 if ((m->m_flags & M_EXT) && m->m_next == NULL && 1972 m->m_len <= MHLEN) { 1973 MGET(n, M_DONTWAIT, m->m_type); 1974 if (n != NULL) { 1975 if (m_dup_pkthdr(n, m, M_NOWAIT)) { 1976 bcopy(m->m_data, n->m_data, 1977 m->m_len); 1978 n->m_len = m->m_len; 1979 } else { 1980 m_free(n); 1981 n = NULL; 1982 } 1983 } 1984 } 1985 if (n != NULL || 1986 (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1987 if (last->inp_flags & INP_CONTROLOPTS) 1988 ip6_savecontrol(last, n, &opts); 1989 /* strip intermediate headers */ 1990 m_adj(n, off); 1991 SOCKBUF_LOCK(&last->inp_socket->so_rcv); 1992 if (sbappendaddr_locked( 1993 &last->inp_socket->so_rcv, 1994 (struct sockaddr *)&fromsa, n, opts) 1995 == 0) { 1996 /* should notify about lost packet */ 1997 m_freem(n); 1998 if (opts) { 1999 m_freem(opts); 2000 } 2001 SOCKBUF_UNLOCK( 2002 &last->inp_socket->so_rcv); 2003 } else 2004 sorwakeup_locked(last->inp_socket); 2005 opts = NULL; 2006 } 2007 INP_RUNLOCK(last); 2008 } 2009 last = in6p; 2010 } 2011 INP_INFO_RUNLOCK(&V_ripcbinfo); 2012 if (last != NULL) { 2013 if (last->inp_flags & INP_CONTROLOPTS) 2014 ip6_savecontrol(last, m, &opts); 2015 /* strip intermediate headers */ 2016 m_adj(m, off); 2017 2018 /* avoid using mbuf clusters if possible (see above) */ 2019 if ((m->m_flags & M_EXT) && m->m_next == NULL && 2020 m->m_len <= MHLEN) { 2021 struct mbuf *n; 2022 2023 MGET(n, M_DONTWAIT, m->m_type); 2024 if (n != NULL) { 2025 if (m_dup_pkthdr(n, m, M_NOWAIT)) { 2026 bcopy(m->m_data, n->m_data, m->m_len); 2027 n->m_len = m->m_len; 2028 2029 m_freem(m); 2030 m = n; 2031 } else { 2032 m_freem(n); 2033 n = NULL; 2034 } 2035 } 2036 } 2037 SOCKBUF_LOCK(&last->inp_socket->so_rcv); 2038 if (sbappendaddr_locked(&last->inp_socket->so_rcv, 2039 (struct sockaddr *)&fromsa, m, opts) == 0) { 2040 m_freem(m); 2041 if (opts) 2042 m_freem(opts); 2043 SOCKBUF_UNLOCK(&last->inp_socket->so_rcv); 2044 } else 2045 sorwakeup_locked(last->inp_socket); 2046 INP_RUNLOCK(last); 2047 } else { 2048 m_freem(m); 2049 V_ip6stat.ip6s_delivered--; 2050 } 2051 return IPPROTO_DONE; 2052 } 2053 2054 /* 2055 * Reflect the ip6 packet back to the source. 2056 * OFF points to the icmp6 header, counted from the top of the mbuf. 2057 */ 2058 void 2059 icmp6_reflect(struct mbuf *m, size_t off) 2060 { 2061 INIT_VNET_INET6(curvnet); 2062 struct ip6_hdr *ip6; 2063 struct icmp6_hdr *icmp6; 2064 struct in6_ifaddr *ia; 2065 int plen; 2066 int type, code; 2067 struct ifnet *outif = NULL; 2068 struct in6_addr origdst, *src = NULL; 2069 2070 /* too short to reflect */ 2071 if (off < sizeof(struct ip6_hdr)) { 2072 nd6log((LOG_DEBUG, 2073 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 2074 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2075 __FILE__, __LINE__)); 2076 goto bad; 2077 } 2078 2079 /* 2080 * If there are extra headers between IPv6 and ICMPv6, strip 2081 * off that header first. 2082 */ 2083 #ifdef DIAGNOSTIC 2084 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2085 panic("assumption failed in icmp6_reflect"); 2086 #endif 2087 if (off > sizeof(struct ip6_hdr)) { 2088 size_t l; 2089 struct ip6_hdr nip6; 2090 2091 l = off - sizeof(struct ip6_hdr); 2092 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2093 m_adj(m, l); 2094 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2095 if (m->m_len < l) { 2096 if ((m = m_pullup(m, l)) == NULL) 2097 return; 2098 } 2099 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2100 } else /* off == sizeof(struct ip6_hdr) */ { 2101 size_t l; 2102 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2103 if (m->m_len < l) { 2104 if ((m = m_pullup(m, l)) == NULL) 2105 return; 2106 } 2107 } 2108 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2109 ip6 = mtod(m, struct ip6_hdr *); 2110 ip6->ip6_nxt = IPPROTO_ICMPV6; 2111 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2112 type = icmp6->icmp6_type; /* keep type for statistics */ 2113 code = icmp6->icmp6_code; /* ditto. */ 2114 2115 origdst = ip6->ip6_dst; 2116 /* 2117 * ip6_input() drops a packet if its src is multicast. 2118 * So, the src is never multicast. 2119 */ 2120 ip6->ip6_dst = ip6->ip6_src; 2121 2122 /* 2123 * If the incoming packet was addressed directly to us (i.e. unicast), 2124 * use dst as the src for the reply. 2125 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2126 * (for example) when we encounter an error while forwarding procedure 2127 * destined to a duplicated address of ours. 2128 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2129 * procedure of an outgoing packet of our own, in which case we need 2130 * to search in the ifaddr list. 2131 */ 2132 if (!IN6_IS_ADDR_MULTICAST(&origdst)) { 2133 if ((ia = ip6_getdstifaddr(m))) { 2134 if (!(ia->ia6_flags & 2135 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) 2136 src = &ia->ia_addr.sin6_addr; 2137 } else { 2138 struct sockaddr_in6 d; 2139 2140 bzero(&d, sizeof(d)); 2141 d.sin6_family = AF_INET6; 2142 d.sin6_len = sizeof(d); 2143 d.sin6_addr = origdst; 2144 ia = (struct in6_ifaddr *) 2145 ifa_ifwithaddr((struct sockaddr *)&d); 2146 if (ia && 2147 !(ia->ia6_flags & 2148 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) { 2149 src = &ia->ia_addr.sin6_addr; 2150 } 2151 } 2152 } 2153 2154 if (src == NULL) { 2155 int e; 2156 struct sockaddr_in6 sin6; 2157 struct route_in6 ro; 2158 2159 /* 2160 * This case matches to multicasts, our anycast, or unicasts 2161 * that we do not own. Select a source address based on the 2162 * source address of the erroneous packet. 2163 */ 2164 bzero(&sin6, sizeof(sin6)); 2165 sin6.sin6_family = AF_INET6; 2166 sin6.sin6_len = sizeof(sin6); 2167 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */ 2168 2169 bzero(&ro, sizeof(ro)); 2170 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e); 2171 if (ro.ro_rt) 2172 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2173 if (src == NULL) { 2174 char ip6buf[INET6_ADDRSTRLEN]; 2175 nd6log((LOG_DEBUG, 2176 "icmp6_reflect: source can't be determined: " 2177 "dst=%s, error=%d\n", 2178 ip6_sprintf(ip6buf, &sin6.sin6_addr), e)); 2179 goto bad; 2180 } 2181 } 2182 2183 ip6->ip6_src = *src; 2184 ip6->ip6_flow = 0; 2185 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2186 ip6->ip6_vfc |= IPV6_VERSION; 2187 ip6->ip6_nxt = IPPROTO_ICMPV6; 2188 if (outif) 2189 ip6->ip6_hlim = ND_IFINFO(outif)->chlim; 2190 else if (m->m_pkthdr.rcvif) { 2191 /* XXX: This may not be the outgoing interface */ 2192 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim; 2193 } else 2194 ip6->ip6_hlim = V_ip6_defhlim; 2195 2196 icmp6->icmp6_cksum = 0; 2197 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2198 sizeof(struct ip6_hdr), plen); 2199 2200 /* 2201 * XXX option handling 2202 */ 2203 2204 m->m_flags &= ~(M_BCAST|M_MCAST); 2205 2206 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL); 2207 if (outif) 2208 icmp6_ifoutstat_inc(outif, type, code); 2209 2210 return; 2211 2212 bad: 2213 m_freem(m); 2214 return; 2215 } 2216 2217 void 2218 icmp6_fasttimo(void) 2219 { 2220 2221 return; 2222 } 2223 2224 static const char * 2225 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2226 struct in6_addr *tgt6) 2227 { 2228 static char buf[1024]; 2229 char ip6bufs[INET6_ADDRSTRLEN]; 2230 char ip6bufd[INET6_ADDRSTRLEN]; 2231 char ip6buft[INET6_ADDRSTRLEN]; 2232 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2233 ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6), 2234 ip6_sprintf(ip6buft, tgt6)); 2235 return buf; 2236 } 2237 2238 void 2239 icmp6_redirect_input(struct mbuf *m, int off) 2240 { 2241 INIT_VNET_INET6(curvnet); 2242 struct ifnet *ifp; 2243 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2244 struct nd_redirect *nd_rd; 2245 int icmp6len = ntohs(ip6->ip6_plen); 2246 char *lladdr = NULL; 2247 int lladdrlen = 0; 2248 u_char *redirhdr = NULL; 2249 int redirhdrlen = 0; 2250 struct rtentry *rt = NULL; 2251 int is_router; 2252 int is_onlink; 2253 struct in6_addr src6 = ip6->ip6_src; 2254 struct in6_addr redtgt6; 2255 struct in6_addr reddst6; 2256 union nd_opts ndopts; 2257 char ip6buf[INET6_ADDRSTRLEN]; 2258 2259 if (!m) 2260 return; 2261 2262 ifp = m->m_pkthdr.rcvif; 2263 2264 if (!ifp) 2265 return; 2266 2267 /* XXX if we are router, we don't update route by icmp6 redirect */ 2268 if (V_ip6_forwarding) 2269 goto freeit; 2270 if (!V_icmp6_rediraccept) 2271 goto freeit; 2272 2273 #ifndef PULLDOWN_TEST 2274 IP6_EXTHDR_CHECK(m, off, icmp6len,); 2275 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off); 2276 #else 2277 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2278 if (nd_rd == NULL) { 2279 V_icmp6stat.icp6s_tooshort++; 2280 return; 2281 } 2282 #endif 2283 redtgt6 = nd_rd->nd_rd_target; 2284 reddst6 = nd_rd->nd_rd_dst; 2285 2286 if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) || 2287 in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) { 2288 goto freeit; 2289 } 2290 2291 /* validation */ 2292 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2293 nd6log((LOG_ERR, 2294 "ICMP6 redirect sent from %s rejected; " 2295 "must be from linklocal\n", 2296 ip6_sprintf(ip6buf, &src6))); 2297 goto bad; 2298 } 2299 if (ip6->ip6_hlim != 255) { 2300 nd6log((LOG_ERR, 2301 "ICMP6 redirect sent from %s rejected; " 2302 "hlim=%d (must be 255)\n", 2303 ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim)); 2304 goto bad; 2305 } 2306 { 2307 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2308 struct sockaddr_in6 sin6; 2309 struct in6_addr *gw6; 2310 2311 bzero(&sin6, sizeof(sin6)); 2312 sin6.sin6_family = AF_INET6; 2313 sin6.sin6_len = sizeof(struct sockaddr_in6); 2314 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6)); 2315 rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL); 2316 if (rt) { 2317 if (rt->rt_gateway == NULL || 2318 rt->rt_gateway->sa_family != AF_INET6) { 2319 nd6log((LOG_ERR, 2320 "ICMP6 redirect rejected; no route " 2321 "with inet6 gateway found for redirect dst: %s\n", 2322 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2323 RTFREE_LOCKED(rt); 2324 goto bad; 2325 } 2326 2327 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2328 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2329 nd6log((LOG_ERR, 2330 "ICMP6 redirect rejected; " 2331 "not equal to gw-for-src=%s (must be same): " 2332 "%s\n", 2333 ip6_sprintf(ip6buf, gw6), 2334 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2335 RTFREE_LOCKED(rt); 2336 goto bad; 2337 } 2338 } else { 2339 nd6log((LOG_ERR, 2340 "ICMP6 redirect rejected; " 2341 "no route found for redirect dst: %s\n", 2342 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2343 goto bad; 2344 } 2345 RTFREE_LOCKED(rt); 2346 rt = NULL; 2347 } 2348 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2349 nd6log((LOG_ERR, 2350 "ICMP6 redirect rejected; " 2351 "redirect dst must be unicast: %s\n", 2352 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2353 goto bad; 2354 } 2355 2356 is_router = is_onlink = 0; 2357 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2358 is_router = 1; /* router case */ 2359 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2360 is_onlink = 1; /* on-link destination case */ 2361 if (!is_router && !is_onlink) { 2362 nd6log((LOG_ERR, 2363 "ICMP6 redirect rejected; " 2364 "neither router case nor onlink case: %s\n", 2365 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2366 goto bad; 2367 } 2368 /* validation passed */ 2369 2370 icmp6len -= sizeof(*nd_rd); 2371 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2372 if (nd6_options(&ndopts) < 0) { 2373 nd6log((LOG_INFO, "icmp6_redirect_input: " 2374 "invalid ND option, rejected: %s\n", 2375 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2376 /* nd6_options have incremented stats */ 2377 goto freeit; 2378 } 2379 2380 if (ndopts.nd_opts_tgt_lladdr) { 2381 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2382 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2383 } 2384 2385 if (ndopts.nd_opts_rh) { 2386 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len; 2387 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */ 2388 } 2389 2390 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2391 nd6log((LOG_INFO, 2392 "icmp6_redirect_input: lladdrlen mismatch for %s " 2393 "(if %d, icmp6 packet %d): %s\n", 2394 ip6_sprintf(ip6buf, &redtgt6), 2395 ifp->if_addrlen, lladdrlen - 2, 2396 icmp6_redirect_diag(&src6, &reddst6, &redtgt6))); 2397 goto bad; 2398 } 2399 2400 /* RFC 2461 8.3 */ 2401 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2402 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2403 2404 if (!is_onlink) { /* better router case. perform rtredirect. */ 2405 /* perform rtredirect */ 2406 struct sockaddr_in6 sdst; 2407 struct sockaddr_in6 sgw; 2408 struct sockaddr_in6 ssrc; 2409 2410 bzero(&sdst, sizeof(sdst)); 2411 bzero(&sgw, sizeof(sgw)); 2412 bzero(&ssrc, sizeof(ssrc)); 2413 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2414 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2415 sizeof(struct sockaddr_in6); 2416 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2417 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2418 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2419 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2420 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST, 2421 (struct sockaddr *)&ssrc); 2422 } 2423 /* finally update cached route in each socket via pfctlinput */ 2424 { 2425 struct sockaddr_in6 sdst; 2426 2427 bzero(&sdst, sizeof(sdst)); 2428 sdst.sin6_family = AF_INET6; 2429 sdst.sin6_len = sizeof(struct sockaddr_in6); 2430 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2431 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2432 #ifdef IPSEC 2433 key_sa_routechange((struct sockaddr *)&sdst); 2434 #endif /* IPSEC */ 2435 } 2436 2437 freeit: 2438 m_freem(m); 2439 return; 2440 2441 bad: 2442 V_icmp6stat.icp6s_badredirect++; 2443 m_freem(m); 2444 } 2445 2446 void 2447 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2448 { 2449 INIT_VNET_INET6(curvnet); 2450 struct ifnet *ifp; /* my outgoing interface */ 2451 struct in6_addr *ifp_ll6; 2452 struct in6_addr *router_ll6; 2453 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2454 struct mbuf *m = NULL; /* newly allocated one */ 2455 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2456 struct nd_redirect *nd_rd; 2457 struct llentry *ln = NULL; 2458 size_t maxlen; 2459 u_char *p; 2460 struct ifnet *outif = NULL; 2461 struct sockaddr_in6 src_sa; 2462 2463 icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2464 2465 /* if we are not router, we don't send icmp6 redirect */ 2466 if (!V_ip6_forwarding) 2467 goto fail; 2468 2469 /* sanity check */ 2470 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2471 goto fail; 2472 2473 /* 2474 * Address check: 2475 * the source address must identify a neighbor, and 2476 * the destination address must not be a multicast address 2477 * [RFC 2461, sec 8.2] 2478 */ 2479 sip6 = mtod(m0, struct ip6_hdr *); 2480 bzero(&src_sa, sizeof(src_sa)); 2481 src_sa.sin6_family = AF_INET6; 2482 src_sa.sin6_len = sizeof(src_sa); 2483 src_sa.sin6_addr = sip6->ip6_src; 2484 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2485 goto fail; 2486 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2487 goto fail; /* what should we do here? */ 2488 2489 /* rate limit */ 2490 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2491 goto fail; 2492 2493 /* 2494 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2495 * we almost always ask for an mbuf cluster for simplicity. 2496 * (MHLEN < IPV6_MMTU is almost always true) 2497 */ 2498 #if IPV6_MMTU >= MCLBYTES 2499 # error assumption failed about IPV6_MMTU and MCLBYTES 2500 #endif 2501 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2502 if (m && IPV6_MMTU >= MHLEN) 2503 MCLGET(m, M_DONTWAIT); 2504 if (!m) 2505 goto fail; 2506 m->m_pkthdr.rcvif = NULL; 2507 m->m_len = 0; 2508 maxlen = M_TRAILINGSPACE(m); 2509 maxlen = min(IPV6_MMTU, maxlen); 2510 /* just for safety */ 2511 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2512 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2513 goto fail; 2514 } 2515 2516 { 2517 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2518 struct in6_ifaddr *ia; 2519 if ((ia = in6ifa_ifpforlinklocal(ifp, 2520 IN6_IFF_NOTREADY| 2521 IN6_IFF_ANYCAST)) == NULL) 2522 goto fail; 2523 ifp_ll6 = &ia->ia_addr.sin6_addr; 2524 } 2525 2526 /* get ip6 linklocal address for the router. */ 2527 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2528 struct sockaddr_in6 *sin6; 2529 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2530 router_ll6 = &sin6->sin6_addr; 2531 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6)) 2532 router_ll6 = (struct in6_addr *)NULL; 2533 } else 2534 router_ll6 = (struct in6_addr *)NULL; 2535 2536 /* ip6 */ 2537 ip6 = mtod(m, struct ip6_hdr *); 2538 ip6->ip6_flow = 0; 2539 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2540 ip6->ip6_vfc |= IPV6_VERSION; 2541 /* ip6->ip6_plen will be set later */ 2542 ip6->ip6_nxt = IPPROTO_ICMPV6; 2543 ip6->ip6_hlim = 255; 2544 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2545 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2546 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2547 2548 /* ND Redirect */ 2549 nd_rd = (struct nd_redirect *)(ip6 + 1); 2550 nd_rd->nd_rd_type = ND_REDIRECT; 2551 nd_rd->nd_rd_code = 0; 2552 nd_rd->nd_rd_reserved = 0; 2553 if (rt->rt_flags & RTF_GATEWAY) { 2554 /* 2555 * nd_rd->nd_rd_target must be a link-local address in 2556 * better router cases. 2557 */ 2558 if (!router_ll6) 2559 goto fail; 2560 bcopy(router_ll6, &nd_rd->nd_rd_target, 2561 sizeof(nd_rd->nd_rd_target)); 2562 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2563 sizeof(nd_rd->nd_rd_dst)); 2564 } else { 2565 /* make sure redtgt == reddst */ 2566 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2567 sizeof(nd_rd->nd_rd_target)); 2568 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2569 sizeof(nd_rd->nd_rd_dst)); 2570 } 2571 2572 p = (u_char *)(nd_rd + 1); 2573 2574 if (!router_ll6) 2575 goto nolladdropt; 2576 2577 { 2578 /* target lladdr option */ 2579 int len; 2580 struct nd_opt_hdr *nd_opt; 2581 char *lladdr; 2582 2583 IF_AFDATA_LOCK(ifp); 2584 ln = nd6_lookup(router_ll6, 0, ifp); 2585 IF_AFDATA_UNLOCK(ifp); 2586 if (ln == NULL) 2587 goto nolladdropt; 2588 2589 len = sizeof(*nd_opt) + ifp->if_addrlen; 2590 len = (len + 7) & ~7; /* round by 8 */ 2591 /* safety check */ 2592 if (len + (p - (u_char *)ip6) > maxlen) 2593 goto nolladdropt; 2594 2595 if (ln->la_flags & LLE_VALID) { 2596 nd_opt = (struct nd_opt_hdr *)p; 2597 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2598 nd_opt->nd_opt_len = len >> 3; 2599 lladdr = (char *)(nd_opt + 1); 2600 bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen); 2601 p += len; 2602 } 2603 } 2604 nolladdropt: 2605 if (ln != NULL) 2606 LLE_RUNLOCK(ln); 2607 2608 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2609 2610 /* just to be safe */ 2611 #ifdef M_DECRYPTED /*not openbsd*/ 2612 if (m0->m_flags & M_DECRYPTED) 2613 goto noredhdropt; 2614 #endif 2615 if (p - (u_char *)ip6 > maxlen) 2616 goto noredhdropt; 2617 2618 { 2619 /* redirected header option */ 2620 int len; 2621 struct nd_opt_rd_hdr *nd_opt_rh; 2622 2623 /* 2624 * compute the maximum size for icmp6 redirect header option. 2625 * XXX room for auth header? 2626 */ 2627 len = maxlen - (p - (u_char *)ip6); 2628 len &= ~7; 2629 2630 /* This is just for simplicity. */ 2631 if (m0->m_pkthdr.len != m0->m_len) { 2632 if (m0->m_next) { 2633 m_freem(m0->m_next); 2634 m0->m_next = NULL; 2635 } 2636 m0->m_pkthdr.len = m0->m_len; 2637 } 2638 2639 /* 2640 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2641 * about padding/truncate rule for the original IP packet. 2642 * From the discussion on IPv6imp in Feb 1999, 2643 * the consensus was: 2644 * - "attach as much as possible" is the goal 2645 * - pad if not aligned (original size can be guessed by 2646 * original ip6 header) 2647 * Following code adds the padding if it is simple enough, 2648 * and truncates if not. 2649 */ 2650 if (m0->m_next || m0->m_pkthdr.len != m0->m_len) 2651 panic("assumption failed in %s:%d", __FILE__, 2652 __LINE__); 2653 2654 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2655 /* not enough room, truncate */ 2656 m0->m_pkthdr.len = m0->m_len = len - 2657 sizeof(*nd_opt_rh); 2658 } else { 2659 /* enough room, pad or truncate */ 2660 size_t extra; 2661 2662 extra = m0->m_pkthdr.len % 8; 2663 if (extra) { 2664 /* pad if easy enough, truncate if not */ 2665 if (8 - extra <= M_TRAILINGSPACE(m0)) { 2666 /* pad */ 2667 m0->m_len += (8 - extra); 2668 m0->m_pkthdr.len += (8 - extra); 2669 } else { 2670 /* truncate */ 2671 m0->m_pkthdr.len -= extra; 2672 m0->m_len -= extra; 2673 } 2674 } 2675 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2676 m0->m_pkthdr.len = m0->m_len = len - 2677 sizeof(*nd_opt_rh); 2678 } 2679 2680 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2681 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2682 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2683 nd_opt_rh->nd_opt_rh_len = len >> 3; 2684 p += sizeof(*nd_opt_rh); 2685 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2686 2687 /* connect m0 to m */ 2688 m_tag_delete_chain(m0, NULL); 2689 m0->m_flags &= ~M_PKTHDR; 2690 m->m_next = m0; 2691 m->m_pkthdr.len = m->m_len + m0->m_len; 2692 m0 = NULL; 2693 } 2694 noredhdropt:; 2695 if (m0) { 2696 m_freem(m0); 2697 m0 = NULL; 2698 } 2699 2700 /* XXX: clear embedded link IDs in the inner header */ 2701 in6_clearscope(&sip6->ip6_src); 2702 in6_clearscope(&sip6->ip6_dst); 2703 in6_clearscope(&nd_rd->nd_rd_target); 2704 in6_clearscope(&nd_rd->nd_rd_dst); 2705 2706 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2707 2708 nd_rd->nd_rd_cksum = 0; 2709 nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2710 sizeof(*ip6), ntohs(ip6->ip6_plen)); 2711 2712 /* send the packet to outside... */ 2713 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL); 2714 if (outif) { 2715 icmp6_ifstat_inc(outif, ifs6_out_msg); 2716 icmp6_ifstat_inc(outif, ifs6_out_redirect); 2717 } 2718 V_icmp6stat.icp6s_outhist[ND_REDIRECT]++; 2719 2720 return; 2721 2722 fail: 2723 if (m) 2724 m_freem(m); 2725 if (m0) 2726 m_freem(m0); 2727 } 2728 2729 /* 2730 * ICMPv6 socket option processing. 2731 */ 2732 int 2733 icmp6_ctloutput(struct socket *so, struct sockopt *sopt) 2734 { 2735 int error = 0; 2736 int optlen; 2737 struct inpcb *inp = sotoinpcb(so); 2738 int level, op, optname; 2739 2740 if (sopt) { 2741 level = sopt->sopt_level; 2742 op = sopt->sopt_dir; 2743 optname = sopt->sopt_name; 2744 optlen = sopt->sopt_valsize; 2745 } else 2746 level = op = optname = optlen = 0; 2747 2748 if (level != IPPROTO_ICMPV6) { 2749 return EINVAL; 2750 } 2751 2752 switch (op) { 2753 case PRCO_SETOPT: 2754 switch (optname) { 2755 case ICMP6_FILTER: 2756 { 2757 struct icmp6_filter ic6f; 2758 2759 if (optlen != sizeof(ic6f)) { 2760 error = EMSGSIZE; 2761 break; 2762 } 2763 error = sooptcopyin(sopt, &ic6f, optlen, optlen); 2764 if (error == 0) { 2765 INP_WLOCK(inp); 2766 *inp->in6p_icmp6filt = ic6f; 2767 INP_WUNLOCK(inp); 2768 } 2769 break; 2770 } 2771 2772 default: 2773 error = ENOPROTOOPT; 2774 break; 2775 } 2776 break; 2777 2778 case PRCO_GETOPT: 2779 switch (optname) { 2780 case ICMP6_FILTER: 2781 { 2782 struct icmp6_filter ic6f; 2783 2784 INP_RLOCK(inp); 2785 ic6f = *inp->in6p_icmp6filt; 2786 INP_RUNLOCK(inp); 2787 error = sooptcopyout(sopt, &ic6f, sizeof(ic6f)); 2788 break; 2789 } 2790 2791 default: 2792 error = ENOPROTOOPT; 2793 break; 2794 } 2795 break; 2796 } 2797 2798 return (error); 2799 } 2800 2801 /* 2802 * Perform rate limit check. 2803 * Returns 0 if it is okay to send the icmp6 packet. 2804 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2805 * limitation. 2806 * 2807 * XXX per-destination/type check necessary? 2808 * 2809 * dst - not used at this moment 2810 * type - not used at this moment 2811 * code - not used at this moment 2812 */ 2813 static int 2814 icmp6_ratelimit(const struct in6_addr *dst, const int type, 2815 const int code) 2816 { 2817 INIT_VNET_INET6(curvnet); 2818 int ret; 2819 2820 ret = 0; /* okay to send */ 2821 2822 /* PPS limit */ 2823 if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count, 2824 V_icmp6errppslim)) { 2825 /* The packet is subject to rate limit */ 2826 ret++; 2827 } 2828 2829 return ret; 2830 } 2831