1 /* $NetBSD: icmp6.c,v 1.66 2001/06/22 13:01:49 itojun Exp $ */ 2 /* $KAME: icmp6.c,v 1.217 2001/06/20 15:03:29 jinmei Exp $ */ 3 4 /* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. All advertising materials mentioning features or use of this software 46 * must display the following acknowledgement: 47 * This product includes software developed by the University of 48 * California, Berkeley and its contributors. 49 * 4. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 66 */ 67 68 #include "opt_inet.h" 69 #include "opt_ipsec.h" 70 71 #include <sys/param.h> 72 #include <sys/systm.h> 73 #include <sys/malloc.h> 74 #include <sys/mbuf.h> 75 #include <sys/protosw.h> 76 #include <sys/socket.h> 77 #include <sys/socketvar.h> 78 #include <sys/time.h> 79 #include <sys/kernel.h> 80 #include <sys/syslog.h> 81 #include <sys/domain.h> 82 83 #include <net/if.h> 84 #include <net/route.h> 85 #include <net/if_dl.h> 86 #include <net/if_types.h> 87 88 #include <netinet/in.h> 89 #include <netinet/in_var.h> 90 #include <netinet/ip6.h> 91 #include <netinet6/ip6_var.h> 92 #include <netinet/icmp6.h> 93 #include <netinet6/mld6_var.h> 94 #include <netinet6/in6_pcb.h> 95 #include <netinet6/nd6.h> 96 #include <netinet6/in6_ifattach.h> 97 #include <netinet6/ip6protosw.h> 98 99 100 #ifdef IPSEC 101 #include <netinet6/ipsec.h> 102 #include <netkey/key.h> 103 #endif 104 105 #include "faith.h" 106 #if defined(NFAITH) && 0 < NFAITH 107 #include <net/if_faith.h> 108 #endif 109 110 #include <net/net_osdep.h> 111 112 extern struct domain inet6domain; 113 114 struct icmp6stat icmp6stat; 115 116 extern struct in6pcb rawin6pcb; 117 extern int icmp6errppslim; 118 static int icmp6errpps_count = 0; 119 static struct timeval icmp6errppslim_last; 120 extern int icmp6_nodeinfo; 121 122 /* 123 * List of callbacks to notify when Path MTU changes are made. 124 */ 125 struct icmp6_mtudisc_callback { 126 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 127 void (*mc_func) __P((struct in6_addr *)); 128 }; 129 130 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 131 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 132 133 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 134 extern int pmtu_expire; 135 136 /* XXX do these values make any sense? */ 137 static int icmp6_mtudisc_hiwat = 1280; 138 static int icmp6_mtudisc_lowat = 256; 139 140 /* 141 * keep track of # of redirect routes. 142 */ 143 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 144 145 /* XXX experimental, turned off */ 146 static int icmp6_redirect_hiwat = -1; 147 static int icmp6_redirect_lowat = -1; 148 149 static void icmp6_errcount __P((struct icmp6errstat *, int, int)); 150 static int icmp6_rip6_input __P((struct mbuf **, int)); 151 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int)); 152 static const char *icmp6_redirect_diag __P((struct in6_addr *, 153 struct in6_addr *, struct in6_addr *)); 154 static struct mbuf *ni6_input __P((struct mbuf *, int)); 155 static struct mbuf *ni6_nametodns __P((const char *, int, int)); 156 static int ni6_dnsmatch __P((const char *, int, const char *, int)); 157 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *, 158 struct ifnet **, char *)); 159 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 160 struct ifnet *, int)); 161 static int icmp6_notify_error __P((struct mbuf *, int, int, int)); 162 static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *)); 163 static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *)); 164 static void icmp6_redirect_timeout __P((struct rtentry *, struct rttimer *)); 165 166 void 167 icmp6_init() 168 { 169 mld6_init(); 170 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 171 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 172 } 173 174 static void 175 icmp6_errcount(stat, type, code) 176 struct icmp6errstat *stat; 177 int type, code; 178 { 179 switch (type) { 180 case ICMP6_DST_UNREACH: 181 switch (code) { 182 case ICMP6_DST_UNREACH_NOROUTE: 183 stat->icp6errs_dst_unreach_noroute++; 184 return; 185 case ICMP6_DST_UNREACH_ADMIN: 186 stat->icp6errs_dst_unreach_admin++; 187 return; 188 case ICMP6_DST_UNREACH_BEYONDSCOPE: 189 stat->icp6errs_dst_unreach_beyondscope++; 190 return; 191 case ICMP6_DST_UNREACH_ADDR: 192 stat->icp6errs_dst_unreach_addr++; 193 return; 194 case ICMP6_DST_UNREACH_NOPORT: 195 stat->icp6errs_dst_unreach_noport++; 196 return; 197 } 198 break; 199 case ICMP6_PACKET_TOO_BIG: 200 stat->icp6errs_packet_too_big++; 201 return; 202 case ICMP6_TIME_EXCEEDED: 203 switch (code) { 204 case ICMP6_TIME_EXCEED_TRANSIT: 205 stat->icp6errs_time_exceed_transit++; 206 return; 207 case ICMP6_TIME_EXCEED_REASSEMBLY: 208 stat->icp6errs_time_exceed_reassembly++; 209 return; 210 } 211 break; 212 case ICMP6_PARAM_PROB: 213 switch (code) { 214 case ICMP6_PARAMPROB_HEADER: 215 stat->icp6errs_paramprob_header++; 216 return; 217 case ICMP6_PARAMPROB_NEXTHEADER: 218 stat->icp6errs_paramprob_nextheader++; 219 return; 220 case ICMP6_PARAMPROB_OPTION: 221 stat->icp6errs_paramprob_option++; 222 return; 223 } 224 break; 225 case ND_REDIRECT: 226 stat->icp6errs_redirect++; 227 return; 228 } 229 stat->icp6errs_unknown++; 230 } 231 232 /* 233 * Register a Path MTU Discovery callback. 234 */ 235 void 236 icmp6_mtudisc_callback_register(func) 237 void (*func) __P((struct in6_addr *)); 238 { 239 struct icmp6_mtudisc_callback *mc; 240 241 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 242 mc = LIST_NEXT(mc, mc_list)) { 243 if (mc->mc_func == func) 244 return; 245 } 246 247 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 248 if (mc == NULL) 249 panic("icmp6_mtudisc_callback_register"); 250 251 mc->mc_func = func; 252 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 253 } 254 255 /* 256 * Generate an error packet of type error in response to bad IP6 packet. 257 */ 258 void 259 icmp6_error(m, type, code, param) 260 struct mbuf *m; 261 int type, code, param; 262 { 263 struct ip6_hdr *oip6, *nip6; 264 struct icmp6_hdr *icmp6; 265 u_int preplen; 266 int off; 267 int nxt; 268 269 icmp6stat.icp6s_error++; 270 271 /* count per-type-code statistics */ 272 icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code); 273 274 if (m->m_flags & M_DECRYPTED) { 275 icmp6stat.icp6s_canterror++; 276 goto freeit; 277 } 278 279 #ifndef PULLDOWN_TEST 280 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), ); 281 #else 282 if (m->m_len < sizeof(struct ip6_hdr)) { 283 m = m_pullup(m, sizeof(struct ip6_hdr)); 284 if (m == NULL) 285 return; 286 } 287 #endif 288 oip6 = mtod(m, struct ip6_hdr *); 289 290 /* 291 * Multicast destination check. For unrecognized option errors, 292 * this check has already done in ip6_unknown_opt(), so we can 293 * check only for other errors. 294 */ 295 if ((m->m_flags & (M_BCAST|M_MCAST) || 296 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 297 (type != ICMP6_PACKET_TOO_BIG && 298 (type != ICMP6_PARAM_PROB || 299 code != ICMP6_PARAMPROB_OPTION))) 300 goto freeit; 301 302 /* Source address check. XXX: the case of anycast source? */ 303 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 304 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 305 goto freeit; 306 307 /* 308 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 309 * don't do it. 310 */ 311 nxt = -1; 312 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 313 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 314 struct icmp6_hdr *icp; 315 316 #ifndef PULLDOWN_TEST 317 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), ); 318 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 319 #else 320 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 321 sizeof(*icp)); 322 if (icp == NULL) { 323 icmp6stat.icp6s_tooshort++; 324 return; 325 } 326 #endif 327 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 328 icp->icmp6_type == ND_REDIRECT) { 329 /* 330 * ICMPv6 error 331 * Special case: for redirect (which is 332 * informational) we must not send icmp6 error. 333 */ 334 icmp6stat.icp6s_canterror++; 335 goto freeit; 336 } else { 337 /* ICMPv6 informational - send the error */ 338 } 339 } else { 340 /* non-ICMPv6 - send the error */ 341 } 342 343 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 344 345 /* Finally, do rate limitation check. */ 346 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 347 icmp6stat.icp6s_toofreq++; 348 goto freeit; 349 } 350 351 /* 352 * OK, ICMP6 can be generated. 353 */ 354 355 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 356 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 357 358 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 359 M_PREPEND(m, preplen, M_DONTWAIT); 360 if (m && m->m_len < preplen) 361 m = m_pullup(m, preplen); 362 if (m == NULL) { 363 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__)); 364 return; 365 } 366 367 nip6 = mtod(m, struct ip6_hdr *); 368 nip6->ip6_src = oip6->ip6_src; 369 nip6->ip6_dst = oip6->ip6_dst; 370 371 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src)) 372 oip6->ip6_src.s6_addr16[1] = 0; 373 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst)) 374 oip6->ip6_dst.s6_addr16[1] = 0; 375 376 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 377 icmp6->icmp6_type = type; 378 icmp6->icmp6_code = code; 379 icmp6->icmp6_pptr = htonl((u_int32_t)param); 380 381 /* 382 * icmp6_reflect() is designed to be in the input path. 383 * icmp6_error() can be called from both input and outut path, 384 * and if we are in output path rcvif could contain bogus value. 385 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 386 * information in ip header (nip6). 387 */ 388 m->m_pkthdr.rcvif = NULL; 389 390 icmp6stat.icp6s_outhist[type]++; 391 icmp6_reflect(m, sizeof(struct ip6_hdr)); /*header order: IPv6 - ICMPv6*/ 392 393 return; 394 395 freeit: 396 /* 397 * If we can't tell wheter or not we can generate ICMP6, free it. 398 */ 399 m_freem(m); 400 } 401 402 /* 403 * Process a received ICMP6 message. 404 */ 405 int 406 icmp6_input(mp, offp, proto) 407 struct mbuf **mp; 408 int *offp, proto; 409 { 410 struct mbuf *m = *mp, *n; 411 struct ip6_hdr *ip6, *nip6; 412 struct icmp6_hdr *icmp6, *nicmp6; 413 int off = *offp; 414 int icmp6len = m->m_pkthdr.len - *offp; 415 int code, sum, noff; 416 417 #ifndef PULLDOWN_TEST 418 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE); 419 /* m might change if M_LOOP. So, call mtod after this */ 420 #endif 421 422 /* 423 * Locate icmp6 structure in mbuf, and check 424 * that not corrupted and of at least minimum length 425 */ 426 427 ip6 = mtod(m, struct ip6_hdr *); 428 if (icmp6len < sizeof(struct icmp6_hdr)) { 429 icmp6stat.icp6s_tooshort++; 430 goto freeit; 431 } 432 433 /* 434 * calculate the checksum 435 */ 436 #ifndef PULLDOWN_TEST 437 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 438 #else 439 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 440 if (icmp6 == NULL) { 441 icmp6stat.icp6s_tooshort++; 442 return IPPROTO_DONE; 443 } 444 #endif 445 code = icmp6->icmp6_code; 446 447 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 448 nd6log((LOG_ERR, 449 "ICMP6 checksum error(%d|%x) %s\n", 450 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src))); 451 icmp6stat.icp6s_checksum++; 452 goto freeit; 453 } 454 455 #if defined(NFAITH) && 0 < NFAITH 456 if (faithprefix(&ip6->ip6_dst)) { 457 /* 458 * Deliver very specific ICMP6 type only. 459 * This is important to deilver TOOBIG. Otherwise PMTUD 460 * will not work. 461 */ 462 switch (icmp6->icmp6_type) { 463 case ICMP6_DST_UNREACH: 464 case ICMP6_PACKET_TOO_BIG: 465 case ICMP6_TIME_EXCEEDED: 466 break; 467 default: 468 goto freeit; 469 } 470 } 471 #endif 472 473 icmp6stat.icp6s_inhist[icmp6->icmp6_type]++; 474 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg); 475 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK) 476 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error); 477 478 switch (icmp6->icmp6_type) { 479 480 case ICMP6_DST_UNREACH: 481 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach); 482 switch (code) { 483 case ICMP6_DST_UNREACH_NOROUTE: 484 code = PRC_UNREACH_NET; 485 break; 486 case ICMP6_DST_UNREACH_ADMIN: 487 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib); 488 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 489 break; 490 case ICMP6_DST_UNREACH_ADDR: 491 code = PRC_HOSTDEAD; 492 break; 493 #ifdef COMPAT_RFC1885 494 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 495 code = PRC_UNREACH_SRCFAIL; 496 break; 497 #else 498 case ICMP6_DST_UNREACH_BEYONDSCOPE: 499 /* I mean "source address was incorrect." */ 500 code = PRC_PARAMPROB; 501 break; 502 #endif 503 case ICMP6_DST_UNREACH_NOPORT: 504 code = PRC_UNREACH_PORT; 505 break; 506 default: 507 goto badcode; 508 } 509 goto deliver; 510 break; 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 break; 525 526 case ICMP6_TIME_EXCEEDED: 527 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed); 528 switch (code) { 529 case ICMP6_TIME_EXCEED_TRANSIT: 530 case ICMP6_TIME_EXCEED_REASSEMBLY: 531 code += PRC_TIMXCEED_INTRANS; 532 break; 533 default: 534 goto badcode; 535 } 536 goto deliver; 537 break; 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 break; 554 555 case ICMP6_ECHO_REQUEST: 556 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo); 557 if (code != 0) 558 goto badcode; 559 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) { 560 /* Give up remote */ 561 break; 562 } 563 if ((n->m_flags & M_EXT) != 0 564 || n->m_len < off + sizeof(struct icmp6_hdr)) { 565 struct mbuf *n0 = n; 566 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6); 567 568 /* 569 * Prepare an internal mbuf. m_pullup() doesn't 570 * always copy the length we specified. 571 */ 572 if (maxlen >= MCLBYTES) { 573 /* Give up remote */ 574 m_freem(n0); 575 break; 576 } 577 MGETHDR(n, M_DONTWAIT, n0->m_type); 578 if (n && maxlen >= MHLEN) { 579 MCLGET(n, M_DONTWAIT); 580 if ((n->m_flags & M_EXT) == 0) { 581 m_free(n); 582 n = NULL; 583 } 584 } 585 if (n == NULL) { 586 /* Give up remote */ 587 m_freem(n0); 588 break; 589 } 590 M_COPY_PKTHDR(n, n0); 591 /* 592 * Copy IPv6 and ICMPv6 only. 593 */ 594 nip6 = mtod(n, struct ip6_hdr *); 595 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 596 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 597 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 598 noff = sizeof(struct ip6_hdr); 599 n->m_pkthdr.len = n->m_len = 600 noff + sizeof(struct icmp6_hdr); 601 /* 602 * Adjust mbuf. ip6_plen will be adjusted in 603 * ip6_output(). 604 */ 605 m_adj(n0, off + sizeof(struct icmp6_hdr)); 606 n->m_pkthdr.len += n0->m_pkthdr.len; 607 n->m_next = n0; 608 n0->m_flags &= ~M_PKTHDR; 609 } else { 610 nip6 = mtod(n, struct ip6_hdr *); 611 nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off); 612 noff = off; 613 } 614 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 615 nicmp6->icmp6_code = 0; 616 if (n) { 617 icmp6stat.icp6s_reflect++; 618 icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++; 619 icmp6_reflect(n, noff); 620 } 621 break; 622 623 case ICMP6_ECHO_REPLY: 624 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply); 625 if (code != 0) 626 goto badcode; 627 break; 628 629 case MLD6_LISTENER_QUERY: 630 case MLD6_LISTENER_REPORT: 631 if (icmp6len < sizeof(struct mld6_hdr)) 632 goto badlen; 633 if (icmp6->icmp6_type == MLD6_LISTENER_QUERY) /* XXX: ugly... */ 634 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery); 635 else 636 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport); 637 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 638 /* give up local */ 639 mld6_input(m, off); 640 m = NULL; 641 goto freeit; 642 } 643 mld6_input(n, off); 644 /* m stays. */ 645 break; 646 647 case MLD6_LISTENER_DONE: 648 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone); 649 if (icmp6len < sizeof(struct mld6_hdr)) /* necessary? */ 650 goto badlen; 651 break; /* nothing to be done in kernel */ 652 653 case MLD6_MTRACE_RESP: 654 case MLD6_MTRACE: 655 /* XXX: these two are experimental. not officially defind. */ 656 /* XXX: per-interface statistics? */ 657 break; /* just pass it to applications */ 658 659 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 660 { 661 enum { WRU, FQDN } mode; 662 663 if (!icmp6_nodeinfo) 664 break; 665 666 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 667 mode = WRU; 668 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 669 mode = FQDN; 670 else 671 goto badlen; 672 673 if (mode == FQDN) { 674 #ifndef PULLDOWN_TEST 675 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo), 676 IPPROTO_DONE); 677 #endif 678 n = m_copy(m, 0, M_COPYALL); 679 if (n) 680 n = ni6_input(n, off); 681 /* XXX meaningless if n == NULL */ 682 noff = sizeof(struct ip6_hdr); 683 } else { 684 u_char *p; 685 int maxlen, maxhlen; 686 687 if (code != 0) 688 goto badcode; 689 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4; 690 if (maxlen >= MCLBYTES) { 691 /* Give up remote */ 692 break; 693 } 694 MGETHDR(n, M_DONTWAIT, m->m_type); 695 if (n && maxlen > MHLEN) { 696 MCLGET(n, M_DONTWAIT); 697 if ((n->m_flags & M_EXT) == 0) { 698 m_free(n); 699 n = NULL; 700 } 701 } 702 if (n == NULL) { 703 /* Give up remote */ 704 break; 705 } 706 n->m_len = 0; 707 maxhlen = M_TRAILINGSPACE(n) - maxlen; 708 if (maxhlen > hostnamelen) 709 maxhlen = hostnamelen; 710 /* 711 * Copy IPv6 and ICMPv6 only. 712 */ 713 nip6 = mtod(n, struct ip6_hdr *); 714 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 715 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 716 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 717 p = (u_char *)(nicmp6 + 1); 718 bzero(p, 4); 719 bcopy(hostname, p + 4, maxhlen); /*meaningless TTL*/ 720 noff = sizeof(struct ip6_hdr); 721 M_COPY_PKTHDR(n, m); /* just for recvif */ 722 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 723 sizeof(struct icmp6_hdr) + 4 + maxhlen; 724 nicmp6->icmp6_type = ICMP6_WRUREPLY; 725 nicmp6->icmp6_code = 0; 726 } 727 #undef hostnamelen 728 if (n) { 729 icmp6stat.icp6s_reflect++; 730 icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++; 731 icmp6_reflect(n, noff); 732 } 733 break; 734 } 735 736 case ICMP6_WRUREPLY: 737 if (code != 0) 738 goto badcode; 739 break; 740 741 case ND_ROUTER_SOLICIT: 742 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit); 743 if (code != 0) 744 goto badcode; 745 if (icmp6len < sizeof(struct nd_router_solicit)) 746 goto badlen; 747 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 748 /* give up local */ 749 nd6_rs_input(m, off, icmp6len); 750 m = NULL; 751 goto freeit; 752 } 753 nd6_rs_input(n, off, icmp6len); 754 /* m stays. */ 755 break; 756 757 case ND_ROUTER_ADVERT: 758 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert); 759 if (code != 0) 760 goto badcode; 761 if (icmp6len < sizeof(struct nd_router_advert)) 762 goto badlen; 763 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 764 /* give up local */ 765 nd6_ra_input(m, off, icmp6len); 766 m = NULL; 767 goto freeit; 768 } 769 nd6_ra_input(n, off, icmp6len); 770 /* m stays. */ 771 break; 772 773 case ND_NEIGHBOR_SOLICIT: 774 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit); 775 if (code != 0) 776 goto badcode; 777 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 778 goto badlen; 779 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 780 /* give up local */ 781 nd6_ns_input(m, off, icmp6len); 782 m = NULL; 783 goto freeit; 784 } 785 nd6_ns_input(n, off, icmp6len); 786 /* m stays. */ 787 break; 788 789 case ND_NEIGHBOR_ADVERT: 790 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert); 791 if (code != 0) 792 goto badcode; 793 if (icmp6len < sizeof(struct nd_neighbor_advert)) 794 goto badlen; 795 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 796 /* give up local */ 797 nd6_na_input(m, off, icmp6len); 798 m = NULL; 799 goto freeit; 800 } 801 nd6_na_input(n, off, icmp6len); 802 /* m stays. */ 803 break; 804 805 case ND_REDIRECT: 806 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect); 807 if (code != 0) 808 goto badcode; 809 if (icmp6len < sizeof(struct nd_redirect)) 810 goto badlen; 811 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 812 /* give up local */ 813 icmp6_redirect_input(m, off); 814 m = NULL; 815 goto freeit; 816 } 817 icmp6_redirect_input(n, off); 818 /* m stays. */ 819 break; 820 821 case ICMP6_ROUTER_RENUMBERING: 822 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 823 code != ICMP6_ROUTER_RENUMBERING_RESULT) 824 goto badcode; 825 if (icmp6len < sizeof(struct icmp6_router_renum)) 826 goto badlen; 827 break; 828 829 default: 830 nd6log((LOG_DEBUG, 831 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n", 832 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 833 ip6_sprintf(&ip6->ip6_dst), 834 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0)); 835 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 836 /* ICMPv6 error: MUST deliver it by spec... */ 837 code = PRC_NCMDS; 838 /* deliver */ 839 } else { 840 /* ICMPv6 informational: MUST not deliver */ 841 break; 842 } 843 deliver: 844 if (icmp6_notify_error(m, off, icmp6len, code)) { 845 /* In this case, m should've been freed. */ 846 return(IPPROTO_DONE); 847 } 848 break; 849 850 badcode: 851 icmp6stat.icp6s_badcode++; 852 break; 853 854 badlen: 855 icmp6stat.icp6s_badlen++; 856 break; 857 } 858 859 /* deliver the packet to appropriate sockets */ 860 icmp6_rip6_input(&m, *offp); 861 862 return IPPROTO_DONE; 863 864 freeit: 865 m_freem(m); 866 return IPPROTO_DONE; 867 } 868 869 static int 870 icmp6_notify_error(m, off, icmp6len, code) 871 struct mbuf *m; 872 int off, icmp6len; 873 { 874 struct icmp6_hdr *icmp6; 875 struct ip6_hdr *eip6; 876 u_int32_t notifymtu; 877 struct sockaddr_in6 icmp6src, icmp6dst; 878 879 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 880 icmp6stat.icp6s_tooshort++; 881 goto freeit; 882 } 883 #ifndef PULLDOWN_TEST 884 IP6_EXTHDR_CHECK(m, off, 885 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), 886 -1); 887 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 888 #else 889 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 890 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 891 if (icmp6 == NULL) { 892 icmp6stat.icp6s_tooshort++; 893 return(-1); 894 } 895 #endif 896 eip6 = (struct ip6_hdr *)(icmp6 + 1); 897 898 /* Detect the upper level protocol */ 899 { 900 void (*ctlfunc) __P((int, struct sockaddr *, void *)); 901 u_int8_t nxt = eip6->ip6_nxt; 902 int eoff = off + sizeof(struct icmp6_hdr) + 903 sizeof(struct ip6_hdr); 904 struct ip6ctlparam ip6cp; 905 struct in6_addr *finaldst = NULL; 906 int icmp6type = icmp6->icmp6_type; 907 struct ip6_frag *fh; 908 struct ip6_rthdr *rth; 909 struct ip6_rthdr0 *rth0; 910 int rthlen; 911 912 while (1) { /* XXX: should avoid inf. loop explicitly? */ 913 struct ip6_ext *eh; 914 915 switch (nxt) { 916 case IPPROTO_HOPOPTS: 917 case IPPROTO_DSTOPTS: 918 case IPPROTO_AH: 919 #ifndef PULLDOWN_TEST 920 IP6_EXTHDR_CHECK(m, 0, eoff + 921 sizeof(struct ip6_ext), 922 -1); 923 eh = (struct ip6_ext *)(mtod(m, caddr_t) 924 + eoff); 925 #else 926 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 927 eoff, sizeof(*eh)); 928 if (eh == NULL) { 929 icmp6stat.icp6s_tooshort++; 930 return(-1); 931 } 932 #endif 933 934 if (nxt == IPPROTO_AH) 935 eoff += (eh->ip6e_len + 2) << 2; 936 else 937 eoff += (eh->ip6e_len + 1) << 3; 938 nxt = eh->ip6e_nxt; 939 break; 940 case IPPROTO_ROUTING: 941 /* 942 * When the erroneous packet contains a 943 * routing header, we should examine the 944 * header to determine the final destination. 945 * Otherwise, we can't properly update 946 * information that depends on the final 947 * destination (e.g. path MTU). 948 */ 949 #ifndef PULLDOWN_TEST 950 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), 951 -1); 952 rth = (struct ip6_rthdr *)(mtod(m, caddr_t) 953 + eoff); 954 #else 955 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 956 eoff, sizeof(*rth)); 957 if (rth == NULL) { 958 icmp6stat.icp6s_tooshort++; 959 return(-1); 960 } 961 #endif 962 rthlen = (rth->ip6r_len + 1) << 3; 963 /* 964 * XXX: currently there is no 965 * officially defined type other 966 * than type-0. 967 * Note that if the segment left field 968 * is 0, all intermediate hops must 969 * have been passed. 970 */ 971 if (rth->ip6r_segleft && 972 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 973 int hops; 974 975 #ifndef PULLDOWN_TEST 976 IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, 977 -1); 978 rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff); 979 #else 980 IP6_EXTHDR_GET(rth0, 981 struct ip6_rthdr0 *, m, 982 eoff, rthlen); 983 if (rth0 == NULL) { 984 icmp6stat.icp6s_tooshort++; 985 return(-1); 986 } 987 #endif 988 /* just ignore a bogus header */ 989 if ((rth0->ip6r0_len % 2) == 0 && 990 (hops = rth0->ip6r0_len/2)) 991 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 992 } 993 eoff += rthlen; 994 nxt = rth->ip6r_nxt; 995 break; 996 case IPPROTO_FRAGMENT: 997 #ifndef PULLDOWN_TEST 998 IP6_EXTHDR_CHECK(m, 0, eoff + 999 sizeof(struct ip6_frag), 1000 -1); 1001 fh = (struct ip6_frag *)(mtod(m, caddr_t) 1002 + eoff); 1003 #else 1004 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 1005 eoff, sizeof(*fh)); 1006 if (fh == NULL) { 1007 icmp6stat.icp6s_tooshort++; 1008 return(-1); 1009 } 1010 #endif 1011 /* 1012 * Data after a fragment header is meaningless 1013 * unless it is the first fragment, but 1014 * we'll go to the notify label for path MTU 1015 * discovery. 1016 */ 1017 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1018 goto notify; 1019 1020 eoff += sizeof(struct ip6_frag); 1021 nxt = fh->ip6f_nxt; 1022 break; 1023 default: 1024 /* 1025 * This case includes ESP and the No Next 1026 * Header. In such cases going to the notify 1027 * label does not have any meaning 1028 * (i.e. ctlfunc will be NULL), but we go 1029 * anyway since we might have to update 1030 * path MTU information. 1031 */ 1032 goto notify; 1033 } 1034 } 1035 notify: 1036 #ifndef PULLDOWN_TEST 1037 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off); 1038 #else 1039 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1040 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1041 if (icmp6 == NULL) { 1042 icmp6stat.icp6s_tooshort++; 1043 return(-1); 1044 } 1045 #endif 1046 1047 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1048 bzero(&icmp6dst, sizeof(icmp6dst)); 1049 icmp6dst.sin6_len = sizeof(struct sockaddr_in6); 1050 icmp6dst.sin6_family = AF_INET6; 1051 if (finaldst == NULL) 1052 icmp6dst.sin6_addr = eip6->ip6_dst; 1053 else 1054 icmp6dst.sin6_addr = *finaldst; 1055 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1056 &icmp6dst.sin6_addr); 1057 #ifndef SCOPEDROUTING 1058 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst, 1059 NULL, NULL)) { 1060 /* should be impossbile */ 1061 nd6log((LOG_DEBUG, 1062 "icmp6_notify_error: in6_embedscope failed\n")); 1063 goto freeit; 1064 } 1065 #endif 1066 1067 /* 1068 * retrieve parameters from the inner IPv6 header, and convert 1069 * them into sockaddr structures. 1070 */ 1071 bzero(&icmp6src, sizeof(icmp6src)); 1072 icmp6src.sin6_len = sizeof(struct sockaddr_in6); 1073 icmp6src.sin6_family = AF_INET6; 1074 icmp6src.sin6_addr = eip6->ip6_src; 1075 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif, 1076 &icmp6src.sin6_addr); 1077 #ifndef SCOPEDROUTING 1078 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src, 1079 NULL, NULL)) { 1080 /* should be impossbile */ 1081 nd6log((LOG_DEBUG, 1082 "icmp6_notify_error: in6_embedscope failed\n")); 1083 goto freeit; 1084 } 1085 #endif 1086 icmp6src.sin6_flowinfo = 1087 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1088 1089 if (finaldst == NULL) 1090 finaldst = &eip6->ip6_dst; 1091 ip6cp.ip6c_m = m; 1092 ip6cp.ip6c_icmp6 = icmp6; 1093 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1094 ip6cp.ip6c_off = eoff; 1095 ip6cp.ip6c_finaldst = finaldst; 1096 ip6cp.ip6c_src = &icmp6src; 1097 ip6cp.ip6c_nxt = nxt; 1098 1099 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1100 notifymtu = ntohl(icmp6->icmp6_mtu); 1101 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1102 } 1103 1104 ctlfunc = (void (*) __P((int, struct sockaddr *, void *))) 1105 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1106 if (ctlfunc) { 1107 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1108 &ip6cp); 1109 } 1110 } 1111 return(0); 1112 1113 freeit: 1114 m_freem(m); 1115 return(-1); 1116 } 1117 1118 void 1119 icmp6_mtudisc_update(ip6cp, validated) 1120 struct ip6ctlparam *ip6cp; 1121 int validated; 1122 { 1123 unsigned long rtcount; 1124 struct icmp6_mtudisc_callback *mc; 1125 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1126 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1127 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1128 u_int mtu = ntohl(icmp6->icmp6_mtu); 1129 struct rtentry *rt = NULL; 1130 struct sockaddr_in6 sin6; 1131 1132 /* 1133 * allow non-validated cases if memory is plenty, to make traffic 1134 * from non-connected pcb happy. 1135 */ 1136 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1137 if (validated) { 1138 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1139 return; 1140 else if (0 <= icmp6_mtudisc_lowat && 1141 rtcount > icmp6_mtudisc_lowat) { 1142 /* 1143 * XXX nuke a victim, install the new one. 1144 */ 1145 } 1146 } else { 1147 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1148 return; 1149 } 1150 1151 bzero(&sin6, sizeof(sin6)); 1152 sin6.sin6_family = PF_INET6; 1153 sin6.sin6_len = sizeof(struct sockaddr_in6); 1154 sin6.sin6_addr = *dst; 1155 /* XXX normally, this won't happen */ 1156 if (IN6_IS_ADDR_LINKLOCAL(dst)) { 1157 sin6.sin6_addr.s6_addr16[1] = 1158 htons(m->m_pkthdr.rcvif->if_index); 1159 } 1160 /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */ 1161 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1162 1163 if (rt && (rt->rt_flags & RTF_HOST) 1164 && !(rt->rt_rmx.rmx_locks & RTV_MTU)) { 1165 if (mtu < IPV6_MMTU) { 1166 /* xxx */ 1167 rt->rt_rmx.rmx_locks |= RTV_MTU; 1168 } else if (mtu < rt->rt_ifp->if_mtu && 1169 rt->rt_rmx.rmx_mtu > mtu) { 1170 icmp6stat.icp6s_pmtuchg++; 1171 rt->rt_rmx.rmx_mtu = mtu; 1172 } 1173 } 1174 if (rt) 1175 RTFREE(rt); 1176 1177 /* 1178 * Notify protocols that the MTU for this destination 1179 * has changed. 1180 */ 1181 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1182 mc = LIST_NEXT(mc, mc_list)) 1183 (*mc->mc_func)(&sin6.sin6_addr); 1184 } 1185 1186 /* 1187 * Process a Node Information Query packet, based on 1188 * draft-ietf-ipngwg-icmp-name-lookups-07. 1189 * 1190 * Spec incompatibilities: 1191 * - IPv6 Subject address handling 1192 * - IPv4 Subject address handling support missing 1193 * - Proxy reply (answer even if it's not for me) 1194 * - joins NI group address at in6_ifattach() time only, does not cope 1195 * with hostname changes by sethostname(3) 1196 */ 1197 #ifndef offsetof /* XXX */ 1198 #define offsetof(type, member) ((size_t)(&((type *)0)->member)) 1199 #endif 1200 static struct mbuf * 1201 ni6_input(m, off) 1202 struct mbuf *m; 1203 int off; 1204 { 1205 struct icmp6_nodeinfo *ni6, *nni6; 1206 struct mbuf *n = NULL; 1207 u_int16_t qtype; 1208 int subjlen; 1209 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1210 struct ni_reply_fqdn *fqdn; 1211 int addrs; /* for NI_QTYPE_NODEADDR */ 1212 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1213 struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */ 1214 struct ip6_hdr *ip6; 1215 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1216 char *subj = NULL; 1217 1218 ip6 = mtod(m, struct ip6_hdr *); 1219 #ifndef PULLDOWN_TEST 1220 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off); 1221 #else 1222 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1223 if (ni6 == NULL) { 1224 /* m is already reclaimed */ 1225 return NULL; 1226 } 1227 #endif 1228 1229 /* 1230 * Validate IPv6 destination address. 1231 * 1232 * The Responder must discard the Query without further processing 1233 * unless it is one of the Responder's unicast or anycast addresses, or 1234 * a link-local scope multicast address which the Responder has joined. 1235 * [icmp-name-lookups-07, Section 4.] 1236 */ 1237 bzero(&sin6, sizeof(sin6)); 1238 sin6.sin6_family = AF_INET6; 1239 sin6.sin6_len = sizeof(struct sockaddr_in6); 1240 bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr)); 1241 /* XXX scopeid */ 1242 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1243 ; /* unicast/anycast, fine */ 1244 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1245 ; /* link-local multicast, fine */ 1246 else 1247 goto bad; 1248 1249 /* validate query Subject field. */ 1250 qtype = ntohs(ni6->ni_qtype); 1251 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1252 switch (qtype) { 1253 case NI_QTYPE_NOOP: 1254 case NI_QTYPE_SUPTYPES: 1255 /* 07 draft */ 1256 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1257 break; 1258 /*FALLTHROUGH*/ 1259 case NI_QTYPE_FQDN: 1260 case NI_QTYPE_NODEADDR: 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(sin6.sin6_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)&sin6.sin6_addr); 1300 /* XXX kame scope hack */ 1301 if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) { 1302 if ((m->m_flags & M_PKTHDR) != 0 && 1303 m->m_pkthdr.rcvif) { 1304 sin6.sin6_addr.s6_addr16[1] = 1305 htons(m->m_pkthdr.rcvif->if_index); 1306 } 1307 } 1308 subj = (char *)&sin6; 1309 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr)) 1310 break; 1311 1312 /* 1313 * XXX if we are to allow other cases, we should really 1314 * be careful about scope here. 1315 * basically, we should disallow queries toward IPv6 1316 * destination X with subject Y, if scope(X) > scope(Y). 1317 * if we allow scope(X) > scope(Y), it will result in 1318 * information leakage across scope boundary. 1319 */ 1320 goto bad; 1321 1322 case ICMP6_NI_SUBJ_FQDN: 1323 /* 1324 * Validate Subject name with gethostname(3). 1325 * 1326 * The behavior may need some debate, since: 1327 * - we are not sure if the node has FQDN as 1328 * hostname (returned by gethostname(3)). 1329 * - the code does wildcard match for truncated names. 1330 * however, we are not sure if we want to perform 1331 * wildcard match, if gethostname(3) side has 1332 * truncated hostname. 1333 */ 1334 n = ni6_nametodns(hostname, hostnamelen, 0); 1335 if (!n || n->m_next || n->m_len == 0) 1336 goto bad; 1337 IP6_EXTHDR_GET(subj, char *, m, 1338 off + sizeof(struct icmp6_nodeinfo), subjlen); 1339 if (subj == NULL) 1340 goto bad; 1341 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1342 n->m_len)) { 1343 goto bad; 1344 } 1345 m_freem(n); 1346 n = NULL; 1347 break; 1348 1349 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1350 default: 1351 goto bad; 1352 } 1353 break; 1354 } 1355 1356 /* guess reply length */ 1357 switch (qtype) { 1358 case NI_QTYPE_NOOP: 1359 break; /* no reply data */ 1360 case NI_QTYPE_SUPTYPES: 1361 replylen += sizeof(u_int32_t); 1362 break; 1363 case NI_QTYPE_FQDN: 1364 /* XXX will append an mbuf */ 1365 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1366 break; 1367 case NI_QTYPE_NODEADDR: 1368 addrs = ni6_addrs(ni6, m, &ifp, subj); 1369 if ((replylen += addrs * (sizeof(struct in6_addr) + 1370 sizeof(u_int32_t))) > MCLBYTES) 1371 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1372 break; 1373 default: 1374 /* 1375 * XXX: We must return a reply with the ICMP6 code 1376 * `unknown Qtype' in this case. However we regard the case 1377 * as an FQDN query for backward compatibility. 1378 * Older versions set a random value to this field, 1379 * so it rarely varies in the defined qtypes. 1380 * But the mechanism is not reliable... 1381 * maybe we should obsolete older versions. 1382 */ 1383 qtype = NI_QTYPE_FQDN; 1384 /* XXX will append an mbuf */ 1385 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1386 oldfqdn++; 1387 break; 1388 } 1389 1390 /* allocate an mbuf to reply. */ 1391 MGETHDR(n, M_DONTWAIT, m->m_type); 1392 if (n == NULL) { 1393 m_freem(m); 1394 return(NULL); 1395 } 1396 M_COPY_PKTHDR(n, m); /* just for recvif */ 1397 if (replylen > MHLEN) { 1398 if (replylen > MCLBYTES) { 1399 /* 1400 * XXX: should we try to allocate more? But MCLBYTES 1401 * is probably much larger than IPV6_MMTU... 1402 */ 1403 goto bad; 1404 } 1405 MCLGET(n, M_DONTWAIT); 1406 if ((n->m_flags & M_EXT) == 0) { 1407 goto bad; 1408 } 1409 } 1410 n->m_pkthdr.len = n->m_len = replylen; 1411 1412 /* copy mbuf header and IPv6 + Node Information base headers */ 1413 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr)); 1414 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1415 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo)); 1416 1417 /* qtype dependent procedure */ 1418 switch (qtype) { 1419 case NI_QTYPE_NOOP: 1420 nni6->ni_code = ICMP6_NI_SUCCESS; 1421 nni6->ni_flags = 0; 1422 break; 1423 case NI_QTYPE_SUPTYPES: 1424 { 1425 u_int32_t v; 1426 nni6->ni_code = ICMP6_NI_SUCCESS; 1427 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1428 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1429 v = (u_int32_t)htonl(0x0000000f); 1430 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1431 break; 1432 } 1433 case NI_QTYPE_FQDN: 1434 nni6->ni_code = ICMP6_NI_SUCCESS; 1435 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) + 1436 sizeof(struct ip6_hdr) + 1437 sizeof(struct icmp6_nodeinfo)); 1438 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1439 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1440 /* 1441 * XXX do we really have FQDN in variable "hostname"? 1442 */ 1443 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1444 if (n->m_next == NULL) 1445 goto bad; 1446 /* XXX we assume that n->m_next is not a chain */ 1447 if (n->m_next->m_next != NULL) 1448 goto bad; 1449 n->m_pkthdr.len += n->m_next->m_len; 1450 break; 1451 case NI_QTYPE_NODEADDR: 1452 { 1453 int lenlim, copied; 1454 1455 nni6->ni_code = ICMP6_NI_SUCCESS; 1456 n->m_pkthdr.len = n->m_len = 1457 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1458 lenlim = M_TRAILINGSPACE(n); 1459 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1460 /* XXX: reset mbuf length */ 1461 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1462 sizeof(struct icmp6_nodeinfo) + copied; 1463 break; 1464 } 1465 default: 1466 break; /* XXX impossible! */ 1467 } 1468 1469 nni6->ni_type = ICMP6_NI_REPLY; 1470 m_freem(m); 1471 return(n); 1472 1473 bad: 1474 m_freem(m); 1475 if (n) 1476 m_freem(n); 1477 return(NULL); 1478 } 1479 #undef hostnamelen 1480 1481 /* 1482 * make a mbuf with DNS-encoded string. no compression support. 1483 * 1484 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1485 * treated as truncated name (two \0 at the end). this is a wild guess. 1486 */ 1487 static struct mbuf * 1488 ni6_nametodns(name, namelen, old) 1489 const char *name; 1490 int namelen; 1491 int old; /* return pascal string if non-zero */ 1492 { 1493 struct mbuf *m; 1494 char *cp, *ep; 1495 const char *p, *q; 1496 int i, len, nterm; 1497 1498 if (old) 1499 len = namelen + 1; 1500 else 1501 len = MCLBYTES; 1502 1503 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1504 MGET(m, M_DONTWAIT, MT_DATA); 1505 if (m && len > MLEN) { 1506 MCLGET(m, M_DONTWAIT); 1507 if ((m->m_flags & M_EXT) == 0) 1508 goto fail; 1509 } 1510 if (!m) 1511 goto fail; 1512 m->m_next = NULL; 1513 1514 if (old) { 1515 m->m_len = len; 1516 *mtod(m, char *) = namelen; 1517 bcopy(name, mtod(m, char *) + 1, namelen); 1518 return m; 1519 } else { 1520 m->m_len = 0; 1521 cp = mtod(m, char *); 1522 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1523 1524 /* if not certain about my name, return empty buffer */ 1525 if (namelen == 0) 1526 return m; 1527 1528 /* 1529 * guess if it looks like shortened hostname, or FQDN. 1530 * shortened hostname needs two trailing "\0". 1531 */ 1532 i = 0; 1533 for (p = name; p < name + namelen; p++) { 1534 if (*p && *p == '.') 1535 i++; 1536 } 1537 if (i < 2) 1538 nterm = 2; 1539 else 1540 nterm = 1; 1541 1542 p = name; 1543 while (cp < ep && p < name + namelen) { 1544 i = 0; 1545 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1546 i++; 1547 /* result does not fit into mbuf */ 1548 if (cp + i + 1 >= ep) 1549 goto fail; 1550 /* DNS label length restriction, RFC1035 page 8 */ 1551 if (i >= 64) 1552 goto fail; 1553 *cp++ = i; 1554 bcopy(p, cp, i); 1555 cp += i; 1556 p = q; 1557 if (p < name + namelen && *p == '.') 1558 p++; 1559 } 1560 /* termination */ 1561 if (cp + nterm >= ep) 1562 goto fail; 1563 while (nterm-- > 0) 1564 *cp++ = '\0'; 1565 m->m_len = cp - mtod(m, char *); 1566 return m; 1567 } 1568 1569 panic("should not reach here"); 1570 /*NOTREACHED*/ 1571 1572 fail: 1573 if (m) 1574 m_freem(m); 1575 return NULL; 1576 } 1577 1578 /* 1579 * check if two DNS-encoded string matches. takes care of truncated 1580 * form (with \0\0 at the end). no compression support. 1581 * XXX upper/lowercase match (see RFC2065) 1582 */ 1583 static int 1584 ni6_dnsmatch(a, alen, b, blen) 1585 const char *a; 1586 int alen; 1587 const char *b; 1588 int blen; 1589 { 1590 const char *a0, *b0; 1591 int l; 1592 1593 /* simplest case - need validation? */ 1594 if (alen == blen && bcmp(a, b, alen) == 0) 1595 return 1; 1596 1597 a0 = a; 1598 b0 = b; 1599 1600 /* termination is mandatory */ 1601 if (alen < 2 || blen < 2) 1602 return 0; 1603 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1604 return 0; 1605 alen--; 1606 blen--; 1607 1608 while (a - a0 < alen && b - b0 < blen) { 1609 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1610 return 0; 1611 1612 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1613 return 0; 1614 /* we don't support compression yet */ 1615 if (a[0] >= 64 || b[0] >= 64) 1616 return 0; 1617 1618 /* truncated case */ 1619 if (a[0] == 0 && a - a0 == alen - 1) 1620 return 1; 1621 if (b[0] == 0 && b - b0 == blen - 1) 1622 return 1; 1623 if (a[0] == 0 || b[0] == 0) 1624 return 0; 1625 1626 if (a[0] != b[0]) 1627 return 0; 1628 l = a[0]; 1629 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1630 return 0; 1631 if (bcmp(a + 1, b + 1, l) != 0) 1632 return 0; 1633 1634 a += 1 + l; 1635 b += 1 + l; 1636 } 1637 1638 if (a - a0 == alen && b - b0 == blen) 1639 return 1; 1640 else 1641 return 0; 1642 } 1643 1644 /* 1645 * calculate the number of addresses to be returned in the node info reply. 1646 */ 1647 static int 1648 ni6_addrs(ni6, m, ifpp, subj) 1649 struct icmp6_nodeinfo *ni6; 1650 struct mbuf *m; 1651 struct ifnet **ifpp; 1652 char *subj; 1653 { 1654 struct ifnet *ifp; 1655 struct in6_ifaddr *ifa6; 1656 struct ifaddr *ifa; 1657 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1658 int addrs = 0, addrsofif, iffound = 0; 1659 int niflags = ni6->ni_flags; 1660 1661 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1662 switch (ni6->ni_code) { 1663 case ICMP6_NI_SUBJ_IPV6: 1664 if (subj == NULL) /* must be impossible... */ 1665 return(0); 1666 subj_ip6 = (struct sockaddr_in6 *)subj; 1667 break; 1668 default: 1669 /* 1670 * XXX: we only support IPv6 subject address for 1671 * this Qtype. 1672 */ 1673 return(0); 1674 } 1675 } 1676 1677 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) 1678 { 1679 addrsofif = 0; 1680 for (ifa = ifp->if_addrlist.tqh_first; ifa; 1681 ifa = ifa->ifa_list.tqe_next) 1682 { 1683 if (ifa->ifa_addr->sa_family != AF_INET6) 1684 continue; 1685 ifa6 = (struct in6_ifaddr *)ifa; 1686 1687 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1688 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1689 &ifa6->ia_addr.sin6_addr)) 1690 iffound = 1; 1691 1692 /* 1693 * IPv4-mapped addresses can only be returned by a 1694 * Node Information proxy, since they represent 1695 * addresses of IPv4-only nodes, which perforce do 1696 * not implement this protocol. 1697 * [icmp-name-lookups-07, Section 5.4] 1698 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1699 * this function at this moment. 1700 */ 1701 1702 /* What do we have to do about ::1? */ 1703 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1704 case IPV6_ADDR_SCOPE_LINKLOCAL: 1705 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1706 continue; 1707 break; 1708 case IPV6_ADDR_SCOPE_SITELOCAL: 1709 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1710 continue; 1711 break; 1712 case IPV6_ADDR_SCOPE_GLOBAL: 1713 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1714 continue; 1715 break; 1716 default: 1717 continue; 1718 } 1719 1720 /* 1721 * check if anycast is okay. 1722 * XXX: just experimental. not in the spec. 1723 */ 1724 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1725 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1726 continue; /* we need only unicast addresses */ 1727 1728 addrsofif++; /* count the address */ 1729 } 1730 if (iffound) { 1731 *ifpp = ifp; 1732 return(addrsofif); 1733 } 1734 1735 addrs += addrsofif; 1736 } 1737 1738 return(addrs); 1739 } 1740 1741 static int 1742 ni6_store_addrs(ni6, nni6, ifp0, resid) 1743 struct icmp6_nodeinfo *ni6, *nni6; 1744 struct ifnet *ifp0; 1745 int resid; 1746 { 1747 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet); 1748 struct in6_ifaddr *ifa6; 1749 struct ifaddr *ifa; 1750 struct ifnet *ifp_dep = NULL; 1751 int copied = 0, allow_deprecated = 0; 1752 u_char *cp = (u_char *)(nni6 + 1); 1753 int niflags = ni6->ni_flags; 1754 u_int32_t ltime; 1755 long time_second = time.tv_sec; 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 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1842 ltime = ND6_INFINITE_LIFETIME; 1843 else { 1844 if (ifa6->ia6_lifetime.ia6t_expire > 1845 time_second) 1846 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second); 1847 else 1848 ltime = 0; 1849 } 1850 1851 bcopy(<ime, cp, sizeof(u_int32_t)); 1852 cp += sizeof(u_int32_t); 1853 1854 /* copy the address itself */ 1855 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1856 sizeof(struct in6_addr)); 1857 /* XXX: KAME link-local hack; remove ifindex */ 1858 if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr)) 1859 ((struct in6_addr *)cp)->s6_addr16[1] = 0; 1860 cp += sizeof(struct in6_addr); 1861 1862 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1863 copied += (sizeof(struct in6_addr) + 1864 sizeof(u_int32_t)); 1865 } 1866 if (ifp0) /* we need search only on the specified IF */ 1867 break; 1868 } 1869 1870 if (allow_deprecated == 0 && ifp_dep != NULL) { 1871 ifp = ifp_dep; 1872 allow_deprecated = 1; 1873 1874 goto again; 1875 } 1876 1877 return(copied); 1878 } 1879 1880 /* 1881 * XXX almost dup'ed code with rip6_input. 1882 */ 1883 static int 1884 icmp6_rip6_input(mp, off) 1885 struct mbuf **mp; 1886 int off; 1887 { 1888 struct mbuf *m = *mp; 1889 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1890 struct in6pcb *in6p; 1891 struct in6pcb *last = NULL; 1892 struct sockaddr_in6 rip6src; 1893 struct icmp6_hdr *icmp6; 1894 struct mbuf *opts = NULL; 1895 1896 #ifndef PULLDOWN_TEST 1897 /* this is assumed to be safe. */ 1898 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off); 1899 #else 1900 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1901 if (icmp6 == NULL) { 1902 /* m is already reclaimed */ 1903 return IPPROTO_DONE; 1904 } 1905 #endif 1906 1907 bzero(&rip6src, sizeof(rip6src)); 1908 rip6src.sin6_len = sizeof(struct sockaddr_in6); 1909 rip6src.sin6_family = AF_INET6; 1910 /* KAME hack: recover scopeid */ 1911 (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif); 1912 1913 for (in6p = rawin6pcb.in6p_next; 1914 in6p != &rawin6pcb; in6p = in6p->in6p_next) 1915 { 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, m, opts) == 0) { 1956 m_freem(m); 1957 if (opts) 1958 m_freem(opts); 1959 } else 1960 sorwakeup(last->in6p_socket); 1961 } else { 1962 m_freem(m); 1963 ip6stat.ip6s_delivered--; 1964 } 1965 return IPPROTO_DONE; 1966 } 1967 1968 /* 1969 * Reflect the ip6 packet back to the source. 1970 * OFF points to the icmp6 header, counted from the top of the mbuf. 1971 * 1972 * Note: RFC 1885 required that an echo reply should be truncated if it 1973 * did not fit in with (return) path MTU, and KAME code supported the 1974 * behavior. However, as a clarification after the RFC, this limitation 1975 * was removed in a revised version of the spec, RFC 2463. We had kept the 1976 * old behavior, with a (non-default) ifdef block, while the new version of 1977 * the spec was an internet-draft status, and even after the new RFC was 1978 * published. But it would rather make sense to clean the obsoleted part 1979 * up, and to make the code simpler at this stage. 1980 */ 1981 void 1982 icmp6_reflect(m, off) 1983 struct mbuf *m; 1984 size_t off; 1985 { 1986 struct ip6_hdr *ip6; 1987 struct icmp6_hdr *icmp6; 1988 struct in6_ifaddr *ia; 1989 struct in6_addr t, *src = 0; 1990 int plen; 1991 int type, code; 1992 struct ifnet *outif = NULL; 1993 struct sockaddr_in6 sa6_src, sa6_dst; 1994 1995 /* too short to reflect */ 1996 if (off < sizeof(struct ip6_hdr)) { 1997 nd6log((LOG_DEBUG, 1998 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 1999 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2000 __FILE__, __LINE__)); 2001 goto bad; 2002 } 2003 2004 /* 2005 * If there are extra headers between IPv6 and ICMPv6, strip 2006 * off that header first. 2007 */ 2008 #ifdef DIAGNOSTIC 2009 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2010 panic("assumption failed in icmp6_reflect"); 2011 #endif 2012 if (off > sizeof(struct ip6_hdr)) { 2013 size_t l; 2014 struct ip6_hdr nip6; 2015 2016 l = off - sizeof(struct ip6_hdr); 2017 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6); 2018 m_adj(m, l); 2019 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2020 if (m->m_len < l) { 2021 if ((m = m_pullup(m, l)) == NULL) 2022 return; 2023 } 2024 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6)); 2025 } else /* off == sizeof(struct ip6_hdr) */ { 2026 size_t l; 2027 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2028 if (m->m_len < l) { 2029 if ((m = m_pullup(m, l)) == NULL) 2030 return; 2031 } 2032 } 2033 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2034 ip6 = mtod(m, struct ip6_hdr *); 2035 ip6->ip6_nxt = IPPROTO_ICMPV6; 2036 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2037 type = icmp6->icmp6_type; /* keep type for statistics */ 2038 code = icmp6->icmp6_code; /* ditto. */ 2039 2040 t = ip6->ip6_dst; 2041 /* 2042 * ip6_input() drops a packet if its src is multicast. 2043 * So, the src is never multicast. 2044 */ 2045 ip6->ip6_dst = ip6->ip6_src; 2046 2047 /* 2048 * XXX: make sure to embed scope zone information, using 2049 * already embedded IDs or the received interface (if any). 2050 * Note that rcvif may be NULL. 2051 * TODO: scoped routing case (XXX). 2052 */ 2053 bzero(&sa6_src, sizeof(sa6_src)); 2054 sa6_src.sin6_family = AF_INET6; 2055 sa6_src.sin6_len = sizeof(sa6_src); 2056 sa6_src.sin6_addr = ip6->ip6_dst; 2057 in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif); 2058 in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL); 2059 bzero(&sa6_dst, sizeof(sa6_dst)); 2060 sa6_dst.sin6_family = AF_INET6; 2061 sa6_dst.sin6_len = sizeof(sa6_dst); 2062 sa6_dst.sin6_addr = t; 2063 in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif); 2064 in6_embedscope(&t, &sa6_dst, NULL, NULL); 2065 2066 /* 2067 * If the incoming packet was addressed directly to us(i.e. unicast), 2068 * use dst as the src for the reply. 2069 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible 2070 * (for example) when we encounter an error while forwarding procedure 2071 * destined to a duplicated address of ours. 2072 */ 2073 for (ia = in6_ifaddr; ia; ia = ia->ia_next) 2074 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) && 2075 (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) { 2076 src = &t; 2077 break; 2078 } 2079 if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) { 2080 /* 2081 * This is the case if the dst is our link-local address 2082 * and the sender is also ourseleves. 2083 */ 2084 src = &t; 2085 } 2086 2087 if (src == 0) { 2088 int e; 2089 struct route_in6 ro; 2090 2091 /* 2092 * This case matches to multicasts, our anycast, or unicasts 2093 * that we do not own. Select a source address based on the 2094 * source address of the erroneous packet. 2095 */ 2096 bzero(&ro, sizeof(ro)); 2097 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e); 2098 if (ro.ro_rt) 2099 RTFREE(ro.ro_rt); /* XXX: we could use this */ 2100 if (src == NULL) { 2101 nd6log((LOG_DEBUG, 2102 "icmp6_reflect: source can't be determined: " 2103 "dst=%s, error=%d\n", 2104 ip6_sprintf(&sa6_src.sin6_addr), e)); 2105 goto bad; 2106 } 2107 } 2108 2109 ip6->ip6_src = *src; 2110 2111 ip6->ip6_flow = 0; 2112 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2113 ip6->ip6_vfc |= IPV6_VERSION; 2114 ip6->ip6_nxt = IPPROTO_ICMPV6; 2115 if (m->m_pkthdr.rcvif) { 2116 /* XXX: This may not be the outgoing interface */ 2117 ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim; 2118 } else 2119 ip6->ip6_hlim = ip6_defhlim; 2120 2121 icmp6->icmp6_cksum = 0; 2122 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2123 sizeof(struct ip6_hdr), plen); 2124 2125 /* 2126 * xxx option handling 2127 */ 2128 2129 m->m_flags &= ~(M_BCAST|M_MCAST); 2130 #ifdef IPSEC 2131 /* Don't lookup socket */ 2132 (void)ipsec_setsocket(m, NULL); 2133 #endif /*IPSEC*/ 2134 2135 ip6_output(m, NULL, NULL, 0, NULL, &outif); 2136 2137 if (outif) 2138 icmp6_ifoutstat_inc(outif, type, code); 2139 2140 return; 2141 2142 bad: 2143 m_freem(m); 2144 return; 2145 } 2146 2147 void 2148 icmp6_fasttimo() 2149 { 2150 2151 mld6_fasttimeo(); 2152 } 2153 2154 static const char * 2155 icmp6_redirect_diag(src6, dst6, tgt6) 2156 struct in6_addr *src6; 2157 struct in6_addr *dst6; 2158 struct in6_addr *tgt6; 2159 { 2160 static char buf[1024]; 2161 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2162 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2163 return buf; 2164 } 2165 2166 void 2167 icmp6_redirect_input(m, off) 2168 struct mbuf *m; 2169 int off; 2170 { 2171 struct ifnet *ifp = m->m_pkthdr.rcvif; 2172 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2173 struct nd_redirect *nd_rd; 2174 int icmp6len = ntohs(ip6->ip6_plen); 2175 char *lladdr = NULL; 2176 int lladdrlen = 0; 2177 u_char *redirhdr = NULL; 2178 int redirhdrlen = 0; 2179 struct rtentry *rt = NULL; 2180 int is_router; 2181 int is_onlink; 2182 struct in6_addr src6 = ip6->ip6_src; 2183 struct in6_addr redtgt6; 2184 struct in6_addr reddst6; 2185 union nd_opts ndopts; 2186 2187 if (!m || !ifp) 2188 return; 2189 2190 /* XXX if we are router, we don't update route by icmp6 redirect */ 2191 if (ip6_forwarding) 2192 goto freeit; 2193 if (!icmp6_rediraccept) 2194 goto freeit; 2195 2196 #ifndef PULLDOWN_TEST 2197 IP6_EXTHDR_CHECK(m, off, icmp6len,); 2198 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off); 2199 #else 2200 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2201 if (nd_rd == NULL) { 2202 icmp6stat.icp6s_tooshort++; 2203 return; 2204 } 2205 #endif 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 #ifdef IPSEC 2379 key_sa_routechange((struct sockaddr *)&sdst); 2380 #endif 2381 } 2382 2383 freeit: 2384 m_freem(m); 2385 return; 2386 2387 bad: 2388 icmp6stat.icp6s_badredirect++; 2389 m_freem(m); 2390 } 2391 2392 void 2393 icmp6_redirect_output(m0, rt) 2394 struct mbuf *m0; 2395 struct rtentry *rt; 2396 { 2397 struct ifnet *ifp; /* my outgoing interface */ 2398 struct in6_addr *ifp_ll6; 2399 struct in6_addr *router_ll6; 2400 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2401 struct mbuf *m = NULL; /* newly allocated one */ 2402 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2403 struct nd_redirect *nd_rd; 2404 size_t maxlen; 2405 u_char *p; 2406 struct ifnet *outif = NULL; 2407 struct sockaddr_in6 src_sa; 2408 2409 icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0); 2410 2411 /* if we are not router, we don't send icmp6 redirect */ 2412 if (!ip6_forwarding || ip6_accept_rtadv) 2413 goto fail; 2414 2415 /* sanity check */ 2416 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2417 goto fail; 2418 2419 /* 2420 * Address check: 2421 * the source address must identify a neighbor, and 2422 * the destination address must not be a multicast address 2423 * [RFC 2461, sec 8.2] 2424 */ 2425 sip6 = mtod(m0, struct ip6_hdr *); 2426 bzero(&src_sa, sizeof(src_sa)); 2427 src_sa.sin6_family = AF_INET6; 2428 src_sa.sin6_len = sizeof(src_sa); 2429 src_sa.sin6_addr = sip6->ip6_src; 2430 /* we don't currently use sin6_scope_id, but eventually use it */ 2431 src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src); 2432 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2433 goto fail; 2434 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2435 goto fail; /* what should we do here? */ 2436 2437 /* rate limit */ 2438 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2439 goto fail; 2440 2441 /* 2442 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2443 * we almost always ask for an mbuf cluster for simplicity. 2444 * (MHLEN < IPV6_MMTU is almost always true) 2445 */ 2446 #if IPV6_MMTU >= MCLBYTES 2447 # error assumption failed about IPV6_MMTU and MCLBYTES 2448 #endif 2449 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2450 if (m && IPV6_MMTU >= MHLEN) 2451 MCLGET(m, M_DONTWAIT); 2452 if (!m) 2453 goto fail; 2454 m->m_len = 0; 2455 maxlen = M_TRAILINGSPACE(m); 2456 maxlen = min(IPV6_MMTU, maxlen); 2457 /* just for safety */ 2458 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2459 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2460 goto fail; 2461 } 2462 2463 { 2464 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2465 struct in6_ifaddr *ia; 2466 if ((ia = in6ifa_ifpforlinklocal(ifp, 2467 IN6_IFF_NOTREADY| 2468 IN6_IFF_ANYCAST)) == NULL) 2469 goto fail; 2470 ifp_ll6 = &ia->ia_addr.sin6_addr; 2471 } 2472 2473 /* get ip6 linklocal address for the router. */ 2474 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2475 struct sockaddr_in6 *sin6; 2476 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2477 router_ll6 = &sin6->sin6_addr; 2478 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6)) 2479 router_ll6 = (struct in6_addr *)NULL; 2480 } else 2481 router_ll6 = (struct in6_addr *)NULL; 2482 2483 /* ip6 */ 2484 ip6 = mtod(m, struct ip6_hdr *); 2485 ip6->ip6_flow = 0; 2486 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2487 ip6->ip6_vfc |= IPV6_VERSION; 2488 /* ip6->ip6_plen will be set later */ 2489 ip6->ip6_nxt = IPPROTO_ICMPV6; 2490 ip6->ip6_hlim = 255; 2491 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2492 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2493 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2494 2495 /* ND Redirect */ 2496 nd_rd = (struct nd_redirect *)(ip6 + 1); 2497 nd_rd->nd_rd_type = ND_REDIRECT; 2498 nd_rd->nd_rd_code = 0; 2499 nd_rd->nd_rd_reserved = 0; 2500 if (rt->rt_flags & RTF_GATEWAY) { 2501 /* 2502 * nd_rd->nd_rd_target must be a link-local address in 2503 * better router cases. 2504 */ 2505 if (!router_ll6) 2506 goto fail; 2507 bcopy(router_ll6, &nd_rd->nd_rd_target, 2508 sizeof(nd_rd->nd_rd_target)); 2509 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2510 sizeof(nd_rd->nd_rd_dst)); 2511 } else { 2512 /* make sure redtgt == reddst */ 2513 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2514 sizeof(nd_rd->nd_rd_target)); 2515 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2516 sizeof(nd_rd->nd_rd_dst)); 2517 } 2518 2519 p = (u_char *)(nd_rd + 1); 2520 2521 if (!router_ll6) 2522 goto nolladdropt; 2523 2524 { 2525 /* target lladdr option */ 2526 struct rtentry *rt_router = NULL; 2527 int len; 2528 struct sockaddr_dl *sdl; 2529 struct nd_opt_hdr *nd_opt; 2530 char *lladdr; 2531 2532 rt_router = nd6_lookup(router_ll6, 0, ifp); 2533 if (!rt_router) 2534 goto nolladdropt; 2535 len = sizeof(*nd_opt) + ifp->if_addrlen; 2536 len = (len + 7) & ~7; /*round by 8*/ 2537 /* safety check */ 2538 if (len + (p - (u_char *)ip6) > maxlen) 2539 goto nolladdropt; 2540 if (!(rt_router->rt_flags & RTF_GATEWAY) && 2541 (rt_router->rt_flags & RTF_LLINFO) && 2542 (rt_router->rt_gateway->sa_family == AF_LINK) && 2543 (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) && 2544 sdl->sdl_alen) { 2545 nd_opt = (struct nd_opt_hdr *)p; 2546 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2547 nd_opt->nd_opt_len = len >> 3; 2548 lladdr = (char *)(nd_opt + 1); 2549 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen); 2550 p += len; 2551 } 2552 } 2553 nolladdropt:; 2554 2555 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2556 2557 /* just to be safe */ 2558 if (m0->m_flags & M_DECRYPTED) 2559 goto noredhdropt; 2560 if (p - (u_char *)ip6 > maxlen) 2561 goto noredhdropt; 2562 2563 { 2564 /* redirected header option */ 2565 int len; 2566 struct nd_opt_rd_hdr *nd_opt_rh; 2567 2568 /* 2569 * compute the maximum size for icmp6 redirect header option. 2570 * XXX room for auth header? 2571 */ 2572 len = maxlen - (p - (u_char *)ip6); 2573 len &= ~7; 2574 2575 /* This is just for simplicity. */ 2576 if (m0->m_pkthdr.len != m0->m_len) { 2577 if (m0->m_next) { 2578 m_freem(m0->m_next); 2579 m0->m_next = NULL; 2580 } 2581 m0->m_pkthdr.len = m0->m_len; 2582 } 2583 2584 /* 2585 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2586 * about padding/truncate rule for the original IP packet. 2587 * From the discussion on IPv6imp in Feb 1999, the consensus was: 2588 * - "attach as much as possible" is the goal 2589 * - pad if not aligned (original size can be guessed by original 2590 * ip6 header) 2591 * Following code adds the padding if it is simple enough, 2592 * and truncates if not. 2593 */ 2594 if (m0->m_next || m0->m_pkthdr.len != m0->m_len) 2595 panic("assumption failed in %s:%d\n", __FILE__, __LINE__); 2596 2597 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2598 /* not enough room, truncate */ 2599 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh); 2600 } else { 2601 /* enough room, pad or truncate */ 2602 size_t extra; 2603 2604 extra = m0->m_pkthdr.len % 8; 2605 if (extra) { 2606 /* pad if easy enough, truncate if not */ 2607 if (8 - extra <= M_TRAILINGSPACE(m0)) { 2608 /* pad */ 2609 m0->m_len += (8 - extra); 2610 m0->m_pkthdr.len += (8 - extra); 2611 } else { 2612 /* truncate */ 2613 m0->m_pkthdr.len -= extra; 2614 m0->m_len -= extra; 2615 } 2616 } 2617 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2618 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh); 2619 } 2620 2621 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2622 bzero(nd_opt_rh, sizeof(*nd_opt_rh)); 2623 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2624 nd_opt_rh->nd_opt_rh_len = len >> 3; 2625 p += sizeof(*nd_opt_rh); 2626 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2627 2628 /* connect m0 to m */ 2629 m->m_next = m0; 2630 m->m_pkthdr.len = m->m_len + m0->m_len; 2631 } 2632 noredhdropt:; 2633 2634 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src)) 2635 sip6->ip6_src.s6_addr16[1] = 0; 2636 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst)) 2637 sip6->ip6_dst.s6_addr16[1] = 0; 2638 #if 0 2639 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) 2640 ip6->ip6_src.s6_addr16[1] = 0; 2641 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) 2642 ip6->ip6_dst.s6_addr16[1] = 0; 2643 #endif 2644 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target)) 2645 nd_rd->nd_rd_target.s6_addr16[1] = 0; 2646 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst)) 2647 nd_rd->nd_rd_dst.s6_addr16[1] = 0; 2648 2649 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2650 2651 nd_rd->nd_rd_cksum = 0; 2652 nd_rd->nd_rd_cksum 2653 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2654 2655 /* send the packet to outside... */ 2656 #ifdef IPSEC 2657 /* Don't lookup socket */ 2658 (void)ipsec_setsocket(m, NULL); 2659 #endif /*IPSEC*/ 2660 ip6_output(m, NULL, NULL, 0, NULL, &outif); 2661 if (outif) { 2662 icmp6_ifstat_inc(outif, ifs6_out_msg); 2663 icmp6_ifstat_inc(outif, ifs6_out_redirect); 2664 } 2665 icmp6stat.icp6s_outhist[ND_REDIRECT]++; 2666 2667 return; 2668 2669 fail: 2670 if (m) 2671 m_freem(m); 2672 if (m0) 2673 m_freem(m0); 2674 } 2675 2676 /* 2677 * ICMPv6 socket option processing. 2678 */ 2679 int 2680 icmp6_ctloutput(op, so, level, optname, mp) 2681 int op; 2682 struct socket *so; 2683 int level, optname; 2684 struct mbuf **mp; 2685 { 2686 int error = 0; 2687 int optlen; 2688 struct in6pcb *in6p = sotoin6pcb(so); 2689 struct mbuf *m = *mp; 2690 2691 optlen = m ? m->m_len : 0; 2692 2693 if (level != IPPROTO_ICMPV6) { 2694 if (op == PRCO_SETOPT && m) 2695 (void)m_free(m); 2696 return EINVAL; 2697 } 2698 2699 switch (op) { 2700 case PRCO_SETOPT: 2701 switch (optname) { 2702 case ICMP6_FILTER: 2703 { 2704 struct icmp6_filter *p; 2705 2706 if (optlen != sizeof(*p)) { 2707 error = EMSGSIZE; 2708 break; 2709 } 2710 p = mtod(m, struct icmp6_filter *); 2711 if (!p || !in6p->in6p_icmp6filt) { 2712 error = EINVAL; 2713 break; 2714 } 2715 bcopy(p, in6p->in6p_icmp6filt, 2716 sizeof(struct icmp6_filter)); 2717 error = 0; 2718 break; 2719 } 2720 2721 default: 2722 error = ENOPROTOOPT; 2723 break; 2724 } 2725 if (m) 2726 (void)m_freem(m); 2727 break; 2728 2729 case PRCO_GETOPT: 2730 switch (optname) { 2731 case ICMP6_FILTER: 2732 { 2733 struct icmp6_filter *p; 2734 2735 if (!in6p->in6p_icmp6filt) { 2736 error = EINVAL; 2737 break; 2738 } 2739 *mp = m = m_get(M_WAIT, MT_SOOPTS); 2740 m->m_len = sizeof(struct icmp6_filter); 2741 p = mtod(m, struct icmp6_filter *); 2742 bcopy(in6p->in6p_icmp6filt, p, 2743 sizeof(struct icmp6_filter)); 2744 error = 0; 2745 break; 2746 } 2747 2748 default: 2749 error = ENOPROTOOPT; 2750 break; 2751 } 2752 break; 2753 } 2754 2755 return(error); 2756 } 2757 2758 /* 2759 * Perform rate limit check. 2760 * Returns 0 if it is okay to send the icmp6 packet. 2761 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2762 * limitation. 2763 * 2764 * XXX per-destination/type check necessary? 2765 */ 2766 static int 2767 icmp6_ratelimit(dst, type, code) 2768 const struct in6_addr *dst; /* not used at this moment */ 2769 const int type; /* not used at this moment */ 2770 const int code; /* not used at this moment */ 2771 { 2772 int ret; 2773 2774 ret = 0; /*okay to send*/ 2775 2776 /* PPS limit */ 2777 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2778 icmp6errppslim)) { 2779 /* The packet is subject to rate limit */ 2780 ret++; 2781 } 2782 2783 return ret; 2784 } 2785 2786 static struct rtentry * 2787 icmp6_mtudisc_clone(dst) 2788 struct sockaddr *dst; 2789 { 2790 struct rtentry *rt; 2791 int error; 2792 2793 rt = rtalloc1(dst, 1); 2794 if (rt == 0) 2795 return NULL; 2796 2797 /* If we didn't get a host route, allocate one */ 2798 if ((rt->rt_flags & RTF_HOST) == 0) { 2799 struct rtentry *nrt; 2800 2801 error = rtrequest((int) RTM_ADD, dst, 2802 (struct sockaddr *) rt->rt_gateway, 2803 (struct sockaddr *) 0, 2804 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2805 if (error) { 2806 rtfree(rt); 2807 return NULL; 2808 } 2809 nrt->rt_rmx = rt->rt_rmx; 2810 rtfree(rt); 2811 rt = nrt; 2812 } 2813 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2814 icmp6_mtudisc_timeout_q); 2815 if (error) { 2816 rtfree(rt); 2817 return NULL; 2818 } 2819 2820 return rt; /* caller need to call rtfree() */ 2821 } 2822 2823 static void 2824 icmp6_mtudisc_timeout(rt, r) 2825 struct rtentry *rt; 2826 struct rttimer *r; 2827 { 2828 if (rt == NULL) 2829 panic("icmp6_mtudisc_timeout: bad route to timeout"); 2830 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2831 (RTF_DYNAMIC | RTF_HOST)) { 2832 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2833 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2834 } else { 2835 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 2836 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu; 2837 } 2838 } 2839 2840 static void 2841 icmp6_redirect_timeout(rt, r) 2842 struct rtentry *rt; 2843 struct rttimer *r; 2844 { 2845 if (rt == NULL) 2846 panic("icmp6_redirect_timeout: bad route to timeout"); 2847 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2848 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2849 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 2850 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 2851 } 2852 } 2853 2854 #include <uvm/uvm_extern.h> 2855 #include <sys/sysctl.h> 2856 2857 int 2858 icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen) 2859 int *name; 2860 u_int namelen; 2861 void *oldp; 2862 size_t *oldlenp; 2863 void *newp; 2864 size_t newlen; 2865 { 2866 2867 /* All sysctl names at this level are terminal. */ 2868 if (namelen != 1) 2869 return ENOTDIR; 2870 2871 switch (name[0]) { 2872 2873 case ICMPV6CTL_REDIRACCEPT: 2874 return sysctl_int(oldp, oldlenp, newp, newlen, 2875 &icmp6_rediraccept); 2876 case ICMPV6CTL_REDIRTIMEOUT: 2877 return sysctl_int(oldp, oldlenp, newp, newlen, 2878 &icmp6_redirtimeout); 2879 case ICMPV6CTL_STATS: 2880 return sysctl_rdstruct(oldp, oldlenp, newp, 2881 &icmp6stat, sizeof(icmp6stat)); 2882 case ICMPV6CTL_ND6_PRUNE: 2883 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune); 2884 case ICMPV6CTL_ND6_DELAY: 2885 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay); 2886 case ICMPV6CTL_ND6_UMAXTRIES: 2887 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries); 2888 case ICMPV6CTL_ND6_MMAXTRIES: 2889 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries); 2890 case ICMPV6CTL_ND6_USELOOPBACK: 2891 return sysctl_int(oldp, oldlenp, newp, newlen, 2892 &nd6_useloopback); 2893 case ICMPV6CTL_NODEINFO: 2894 return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo); 2895 case ICMPV6CTL_ERRPPSLIMIT: 2896 return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6errppslim); 2897 case ICMPV6CTL_ND6_MAXNUDHINT: 2898 return sysctl_int(oldp, oldlenp, newp, newlen, 2899 &nd6_maxnudhint); 2900 case ICMPV6CTL_MTUDISC_HIWAT: 2901 return sysctl_int(oldp, oldlenp, newp, newlen, 2902 &icmp6_mtudisc_hiwat); 2903 case ICMPV6CTL_MTUDISC_LOWAT: 2904 return sysctl_int(oldp, oldlenp, newp, newlen, 2905 &icmp6_mtudisc_lowat); 2906 case ICMPV6CTL_ND6_DEBUG: 2907 return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_debug); 2908 default: 2909 return ENOPROTOOPT; 2910 } 2911 /* NOTREACHED */ 2912 } 2913