sys/netinet6/icmp6.c

   1 /*      $FreeBSD$       */
   2 /*      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun 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  * 4. Neither the name of the University nor the names of its contributors
  46  *    may be used to endorse or promote products derived from this software
  47  *    without specific prior written permission.
  48  *
  49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  59  * SUCH DAMAGE.
  60  *
  61  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
  62  */
  63
  64 #include "opt_inet.h"
  65 #include "opt_inet6.h"
  66 #include "opt_ipsec.h"
  67
  68 #include <sys/param.h>
  69 #include <sys/domain.h>
  70 #include <sys/kernel.h>
  71 #include <sys/lock.h>
  72 #include <sys/malloc.h>
  73 #include <sys/mbuf.h>
  74 #include <sys/protosw.h>
  75 #include <sys/signalvar.h>
  76 #include <sys/socket.h>
  77 #include <sys/socketvar.h>
  78 #include <sys/sx.h>
  79 #include <sys/syslog.h>
  80 #include <sys/systm.h>
  81 #include <sys/time.h>
  82
  83 #include <net/if.h>
  84 #include <net/if_dl.h>
  85 #include <net/if_types.h>
  86 #include <net/route.h>
  87
  88 #include <netinet/in.h>
  89 #include <netinet/in_pcb.h>
  90 #include <netinet/in_var.h>
  91 #include <netinet/ip6.h>
  92 #include <netinet/icmp6.h>
  93 #include <netinet/tcp_var.h>
  94 #include <netinet6/in6_ifattach.h>
  95 #include <netinet6/in6_pcb.h>
  96 #include <netinet6/ip6protosw.h>
  97 #include <netinet6/ip6_var.h>
  98 #include <netinet6/mld6_var.h>
  99 #include <netinet6/nd6.h>
 100
 101 #ifdef IPSEC
 102 #include <netinet6/ipsec.h>
 103 #include <netkey/key.h>
 104 #endif
 105
 106 #ifdef FAST_IPSEC
 107 #include <netipsec/ipsec.h>
 108 #include <netipsec/key.h>
 109 #endif
 110
 111 #include <net/net_osdep.h>
 112
 113 extern struct domain inet6domain;
 114
 115 struct icmp6stat icmp6stat;
 116
 117 extern struct inpcbinfo ripcbinfo;
 118 extern struct inpcbhead ripcb;
 119 extern int icmp6errppslim;
 120 static int icmp6errpps_count = 0;
 121 static struct timeval icmp6errppslim_last;
 122 extern int icmp6_nodeinfo;
 123
 124 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
 125 static int icmp6_rip6_input __P((struct mbuf **, int));
 126 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
 127 static const char *icmp6_redirect_diag __P((struct in6_addr *,
 128         struct in6_addr *, struct in6_addr *));
 129 static struct mbuf *ni6_input __P((struct mbuf *, int));
 130 static struct mbuf *ni6_nametodns __P((const char *, int, int));
 131 static int ni6_dnsmatch __P((const char *, int, const char *, int));
 132 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
 133                           struct ifnet **, char *));
 134 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
 135                                 struct ifnet *, int));
 136 static int icmp6_notify_error __P((struct mbuf **, int, int, int));
 137
 138 #ifdef COMPAT_RFC1885
 139 static struct route_in6 icmp6_reflect_rt;
 140 #endif
 141
 142
 143 void
 144 icmp6_init()
 145 {
 146         mld6_init();
 147 }
 148
 149 static void
 150 icmp6_errcount(stat, type, code)
 151         struct icmp6errstat *stat;
 152         int type, code;
 153 {
 154         switch (type) {
 155         case ICMP6_DST_UNREACH:
 156                 switch (code) {
 157                 case ICMP6_DST_UNREACH_NOROUTE:
 158                         stat->icp6errs_dst_unreach_noroute++;
 159                         return;
 160                 case ICMP6_DST_UNREACH_ADMIN:
 161                         stat->icp6errs_dst_unreach_admin++;
 162                         return;
 163                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 164                         stat->icp6errs_dst_unreach_beyondscope++;
 165                         return;
 166                 case ICMP6_DST_UNREACH_ADDR:
 167                         stat->icp6errs_dst_unreach_addr++;
 168                         return;
 169                 case ICMP6_DST_UNREACH_NOPORT:
 170                         stat->icp6errs_dst_unreach_noport++;
 171                         return;
 172                 }
 173                 break;
 174         case ICMP6_PACKET_TOO_BIG:
 175                 stat->icp6errs_packet_too_big++;
 176                 return;
 177         case ICMP6_TIME_EXCEEDED:
 178                 switch (code) {
 179                 case ICMP6_TIME_EXCEED_TRANSIT:
 180                         stat->icp6errs_time_exceed_transit++;
 181                         return;
 182                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 183                         stat->icp6errs_time_exceed_reassembly++;
 184                         return;
 185                 }
 186                 break;
 187         case ICMP6_PARAM_PROB:
 188                 switch (code) {
 189                 case ICMP6_PARAMPROB_HEADER:
 190                         stat->icp6errs_paramprob_header++;
 191                         return;
 192                 case ICMP6_PARAMPROB_NEXTHEADER:
 193                         stat->icp6errs_paramprob_nextheader++;
 194                         return;
 195                 case ICMP6_PARAMPROB_OPTION:
 196                         stat->icp6errs_paramprob_option++;
 197                         return;
 198                 }
 199                 break;
 200         case ND_REDIRECT:
 201                 stat->icp6errs_redirect++;
 202                 return;
 203         }
 204         stat->icp6errs_unknown++;
 205 }
 206
 207 /*
 208  * Generate an error packet of type error in response to bad IP6 packet.
 209  */
 210 void
 211 icmp6_error(m, type, code, param)
 212         struct mbuf *m;
 213         int type, code, param;
 214 {
 215         struct ip6_hdr *oip6, *nip6;
 216         struct icmp6_hdr *icmp6;
 217         u_int preplen;
 218         int off;
 219         int nxt;
 220
 221         icmp6stat.icp6s_error++;
 222
 223         /* count per-type-code statistics */
 224         icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
 225
 226 #ifdef M_DECRYPTED      /*not openbsd*/
 227         if (m->m_flags & M_DECRYPTED) {
 228                 icmp6stat.icp6s_canterror++;
 229                 goto freeit;
 230         }
 231 #endif
 232
 233 #ifndef PULLDOWN_TEST
 234         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 235 #else
 236         if (m->m_len < sizeof(struct ip6_hdr)) {
 237                 m = m_pullup(m, sizeof(struct ip6_hdr));
 238                 if (m == NULL)
 239                         return;
 240         }
 241 #endif
 242         oip6 = mtod(m, struct ip6_hdr *);
 243
 244         /*
 245          * If the destination address of the erroneous packet is a multicast
 246          * address, or the packet was sent using link-layer multicast,
 247          * we should basically suppress sending an error (RFC 2463, Section
 248          * 2.4).
 249          * We have two exceptions (the item e.2 in that section):
 250          * - the Pakcet Too Big message can be sent for path MTU discovery.
 251          * - the Parameter Problem Message that can be allowed an icmp6 error
 252          *   in the option type field.  This check has been done in
 253          *   ip6_unknown_opt(), so we can just check the type and code.
 254          */
 255         if ((m->m_flags & (M_BCAST|M_MCAST) ||
 256              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
 257             (type != ICMP6_PACKET_TOO_BIG &&
 258              (type != ICMP6_PARAM_PROB ||
 259               code != ICMP6_PARAMPROB_OPTION)))
 260                 goto freeit;
 261
 262         /*
 263          * RFC 2463, 2.4 (e.5): source address check.
 264          * XXX: the case of anycast source?
 265          */
 266         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
 267             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
 268                 goto freeit;
 269
 270         /*
 271          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
 272          * don't do it.
 273          */
 274         nxt = -1;
 275         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
 276         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
 277                 struct icmp6_hdr *icp;
 278
 279 #ifndef PULLDOWN_TEST
 280                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
 281                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 282 #else
 283                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
 284                         sizeof(*icp));
 285                 if (icp == NULL) {
 286                         icmp6stat.icp6s_tooshort++;
 287                         return;
 288                 }
 289 #endif
 290                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
 291                     icp->icmp6_type == ND_REDIRECT) {
 292                         /*
 293                          * ICMPv6 error
 294                          * Special case: for redirect (which is
 295                          * informational) we must not send icmp6 error.
 296                          */
 297                         icmp6stat.icp6s_canterror++;
 298                         goto freeit;
 299                 } else {
 300                         /* ICMPv6 informational - send the error */
 301                 }
 302         } else {
 303                 /* non-ICMPv6 - send the error */
 304         }
 305
 306         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
 307
 308         /* Finally, do rate limitation check. */
 309         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
 310                 icmp6stat.icp6s_toofreq++;
 311                 goto freeit;
 312         }
 313
 314         /*
 315          * OK, ICMP6 can be generated.
 316          */
 317
 318         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
 319                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
 320
 321         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
 322         M_PREPEND(m, preplen, M_DONTWAIT);
 323         if (m && m->m_len < preplen)
 324                 m = m_pullup(m, preplen);
 325         if (m == NULL) {
 326                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
 327                 return;
 328         }
 329
 330         nip6 = mtod(m, struct ip6_hdr *);
 331         nip6->ip6_src  = oip6->ip6_src;
 332         nip6->ip6_dst  = oip6->ip6_dst;
 333
 334         in6_clearscope(&oip6->ip6_src);
 335         in6_clearscope(&oip6->ip6_dst);
 336
 337         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
 338         icmp6->icmp6_type = type;
 339         icmp6->icmp6_code = code;
 340         icmp6->icmp6_pptr = htonl((u_int32_t)param);
 341
 342         /*
 343          * icmp6_reflect() is designed to be in the input path.
 344          * icmp6_error() can be called from both input and outut path,
 345          * and if we are in output path rcvif could contain bogus value.
 346          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
 347          * information in ip header (nip6).
 348          */
 349         m->m_pkthdr.rcvif = NULL;
 350
 351         icmp6stat.icp6s_outhist[type]++;
 352         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
 353
 354         return;
 355
 356   freeit:
 357         /*
 358          * If we can't tell wheter or not we can generate ICMP6, free it.
 359          */
 360         m_freem(m);
 361 }
 362
 363 /*
 364  * Process a received ICMP6 message.
 365  */
 366 int
 367 icmp6_input(mp, offp, proto)
 368         struct mbuf **mp;
 369         int *offp, proto;
 370 {
 371         struct mbuf *m = *mp, *n;
 372         struct ip6_hdr *ip6, *nip6;
 373         struct icmp6_hdr *icmp6, *nicmp6;
 374         int off = *offp;
 375         int icmp6len = m->m_pkthdr.len - *offp;
 376         int code, sum, noff;
 377
 378 #ifndef PULLDOWN_TEST
 379         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
 380         /* m might change if M_LOOP.  So, call mtod after this */
 381 #endif
 382
 383         /*
 384          * Locate icmp6 structure in mbuf, and check
 385          * that not corrupted and of at least minimum length
 386          */
 387
 388         ip6 = mtod(m, struct ip6_hdr *);
 389         if (icmp6len < sizeof(struct icmp6_hdr)) {
 390                 icmp6stat.icp6s_tooshort++;
 391                 goto freeit;
 392         }
 393
 394         /*
 395          * calculate the checksum
 396          */
 397 #ifndef PULLDOWN_TEST
 398         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
 399 #else
 400         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
 401         if (icmp6 == NULL) {
 402                 icmp6stat.icp6s_tooshort++;
 403                 return IPPROTO_DONE;
 404         }
 405 #endif
 406         code = icmp6->icmp6_code;
 407
 408         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
 409                 nd6log((LOG_ERR,
 410                     "ICMP6 checksum error(%d|%x) %s\n",
 411                     icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
 412                 icmp6stat.icp6s_checksum++;
 413                 goto freeit;
 414         }
 415
 416         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
 417                 /*
 418                  * Deliver very specific ICMP6 type only.
 419                  * This is important to deilver TOOBIG.  Otherwise PMTUD
 420                  * will not work.
 421                  */
 422                 switch (icmp6->icmp6_type) {
 423                 case ICMP6_DST_UNREACH:
 424                 case ICMP6_PACKET_TOO_BIG:
 425                 case ICMP6_TIME_EXCEEDED:
 426                         break;
 427                 default:
 428                         goto freeit;
 429                 }
 430         }
 431
 432         icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
 433         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
 434         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
 435                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
 436
 437         switch (icmp6->icmp6_type) {
 438         case ICMP6_DST_UNREACH:
 439                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
 440                 switch (code) {
 441                 case ICMP6_DST_UNREACH_NOROUTE:
 442                         code = PRC_UNREACH_NET;
 443                         break;
 444                 case ICMP6_DST_UNREACH_ADMIN:
 445                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
 446                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
 447                         break;
 448                 case ICMP6_DST_UNREACH_ADDR:
 449                         code = PRC_HOSTDEAD;
 450                         break;
 451 #ifdef COMPAT_RFC1885
 452                 case ICMP6_DST_UNREACH_NOTNEIGHBOR:
 453                         code = PRC_UNREACH_SRCFAIL;
 454                         break;
 455 #else
 456                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 457                         /* I mean "source address was incorrect." */
 458                         code = PRC_PARAMPROB;
 459                         break;
 460 #endif
 461                 case ICMP6_DST_UNREACH_NOPORT:
 462                         code = PRC_UNREACH_PORT;
 463                         break;
 464                 default:
 465                         goto badcode;
 466                 }
 467                 goto deliver;
 468                 break;
 469
 470         case ICMP6_PACKET_TOO_BIG:
 471                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
 472
 473                 /* validation is made in icmp6_mtudisc_update */
 474
 475                 code = PRC_MSGSIZE;
 476
 477                 /*
 478                  * Updating the path MTU will be done after examining
 479                  * intermediate extension headers.
 480                  */
 481                 goto deliver;
 482                 break;
 483
 484         case ICMP6_TIME_EXCEEDED:
 485                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
 486                 switch (code) {
 487                 case ICMP6_TIME_EXCEED_TRANSIT:
 488                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 489                         code += PRC_TIMXCEED_INTRANS;
 490                         break;
 491                 default:
 492                         goto badcode;
 493                 }
 494                 goto deliver;
 495                 break;
 496
 497         case ICMP6_PARAM_PROB:
 498                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
 499                 switch (code) {
 500                 case ICMP6_PARAMPROB_NEXTHEADER:
 501                         code = PRC_UNREACH_PROTOCOL;
 502                         break;
 503                 case ICMP6_PARAMPROB_HEADER:
 504                 case ICMP6_PARAMPROB_OPTION:
 505                         code = PRC_PARAMPROB;
 506                         break;
 507                 default:
 508                         goto badcode;
 509                 }
 510                 goto deliver;
 511                 break;
 512
 513         case ICMP6_ECHO_REQUEST:
 514                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
 515                 if (code != 0)
 516                         goto badcode;
 517                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
 518                         /* Give up remote */
 519                         break;
 520                 }
 521                 if ((n->m_flags & M_EXT) != 0
 522                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
 523                         struct mbuf *n0 = n;
 524                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
 525                         int n0len;
 526
 527                         MGETHDR(n, M_DONTWAIT, n0->m_type);
 528                         n0len = n0->m_pkthdr.len;       /* save for use below */
 529                         if (n)
 530                                 M_MOVE_PKTHDR(n, n0);
 531                         if (n && maxlen >= MHLEN) {
 532                                 MCLGET(n, M_DONTWAIT);
 533                                 if ((n->m_flags & M_EXT) == 0) {
 534                                         m_free(n);
 535                                         n = NULL;
 536                                 }
 537                         }
 538                         if (n == NULL) {
 539                                 /* Give up remote */
 540                                 m_freem(n0);
 541                                 break;
 542                         }
 543                         /*
 544                          * Copy IPv6 and ICMPv6 only.
 545                          */
 546                         nip6 = mtod(n, struct ip6_hdr *);
 547                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 548                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 549                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 550                         noff = sizeof(struct ip6_hdr);
 551                         /* new mbuf contains only ipv6+icmpv6 headers */
 552                         n->m_len = noff + sizeof(struct icmp6_hdr);
 553                         /*
 554                          * Adjust mbuf.  ip6_plen will be adjusted in
 555                          * ip6_output().
 556                          */
 557                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 558                         /* recalculate complete packet size */
 559                         n->m_pkthdr.len = n0len + (noff - off);
 560                         n->m_next = n0;
 561                 } else {
 562                         nip6 = mtod(n, struct ip6_hdr *);
 563                         nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
 564                         noff = off;
 565                 }
 566                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 567                 nicmp6->icmp6_code = 0;
 568                 if (n) {
 569                         icmp6stat.icp6s_reflect++;
 570                         icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
 571                         icmp6_reflect(n, noff);
 572                 }
 573                 break;
 574
 575         case ICMP6_ECHO_REPLY:
 576                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
 577                 if (code != 0)
 578                         goto badcode;
 579                 break;
 580
 581         case MLD_LISTENER_QUERY:
 582         case MLD_LISTENER_REPORT:
 583                 if (icmp6len < sizeof(struct mld_hdr))
 584                         goto badlen;
 585                 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
 586                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
 587                 else
 588                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
 589                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 590                         /* give up local */
 591                         mld6_input(m, off);
 592                         m = NULL;
 593                         goto freeit;
 594                 }
 595                 mld6_input(n, off);
 596                 /* m stays. */
 597                 break;
 598
 599         case MLD_LISTENER_DONE:
 600                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
 601                 if (icmp6len < sizeof(struct mld_hdr))  /* necessary? */
 602                         goto badlen;
 603                 break;          /* nothing to be done in kernel */
 604
 605         case MLD_MTRACE_RESP:
 606         case MLD_MTRACE:
 607                 /* XXX: these two are experimental.  not officially defined. */
 608                 /* XXX: per-interface statistics? */
 609                 break;          /* just pass it to applications */
 610
 611         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 612             {
 613                 enum { WRU, FQDN } mode;
 614
 615                 if (!icmp6_nodeinfo)
 616                         break;
 617
 618                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 619                         mode = WRU;
 620                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 621                         mode = FQDN;
 622                 else
 623                         goto badlen;
 624
 625 #define hostnamelen     strlen(hostname)
 626                 if (mode == FQDN) {
 627 #ifndef PULLDOWN_TEST
 628                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
 629                             IPPROTO_DONE);
 630 #endif
 631                         n = m_copy(m, 0, M_COPYALL);
 632                         if (n)
 633                                 n = ni6_input(n, off);
 634                         /* XXX meaningless if n == NULL */
 635                         noff = sizeof(struct ip6_hdr);
 636                 } else {
 637                         u_char *p;
 638                         int maxlen, maxhlen;
 639
 640                         if ((icmp6_nodeinfo & 5) != 5)
 641                                 break;
 642
 643                         if (code != 0)
 644                                 goto badcode;
 645                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
 646                         if (maxlen >= MCLBYTES) {
 647                                 /* Give up remote */
 648                                 break;
 649                         }
 650                         MGETHDR(n, M_DONTWAIT, m->m_type);
 651                         if (n && maxlen > MHLEN) {
 652                                 MCLGET(n, M_DONTWAIT);
 653                                 if ((n->m_flags & M_EXT) == 0) {
 654                                         m_free(n);
 655                                         n = NULL;
 656                                 }
 657                         }
 658                         if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
 659                                 /*
 660                                  * Previous code did a blind M_COPY_PKTHDR
 661                                  * and said "just for rcvif".  If true, then
 662                                  * we could tolerate the dup failing (due to
 663                                  * the deep copy of the tag chain).  For now
 664                                  * be conservative and just fail.
 665                                  */
 666                                 m_free(n);
 667                                 n = NULL;
 668                         }
 669                         if (n == NULL) {
 670                                 /* Give up remote */
 671                                 break;
 672                         }
 673                         n->m_pkthdr.rcvif = NULL;
 674                         n->m_len = 0;
 675                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
 676                         if (maxhlen > hostnamelen)
 677                                 maxhlen = hostnamelen;
 678                         /*
 679                          * Copy IPv6 and ICMPv6 only.
 680                          */
 681                         nip6 = mtod(n, struct ip6_hdr *);
 682                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 683                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 684                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 685                         p = (u_char *)(nicmp6 + 1);
 686                         bzero(p, 4);
 687                         bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
 688                         noff = sizeof(struct ip6_hdr);
 689                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 690                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 691                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 692                         nicmp6->icmp6_code = 0;
 693                 }
 694 #undef hostnamelen
 695                 if (n) {
 696                         icmp6stat.icp6s_reflect++;
 697                         icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
 698                         icmp6_reflect(n, noff);
 699                 }
 700                 break;
 701             }
 702
 703         case ICMP6_WRUREPLY:
 704                 if (code != 0)
 705                         goto badcode;
 706                 break;
 707
 708         case ND_ROUTER_SOLICIT:
 709                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
 710                 if (code != 0)
 711                         goto badcode;
 712                 if (icmp6len < sizeof(struct nd_router_solicit))
 713                         goto badlen;
 714                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 715                         /* give up local */
 716                         nd6_rs_input(m, off, icmp6len);
 717                         m = NULL;
 718                         goto freeit;
 719                 }
 720                 nd6_rs_input(n, off, icmp6len);
 721                 /* m stays. */
 722                 break;
 723
 724         case ND_ROUTER_ADVERT:
 725                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
 726                 if (code != 0)
 727                         goto badcode;
 728                 if (icmp6len < sizeof(struct nd_router_advert))
 729                         goto badlen;
 730                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 731                         /* give up local */
 732                         nd6_ra_input(m, off, icmp6len);
 733                         m = NULL;
 734                         goto freeit;
 735                 }
 736                 nd6_ra_input(n, off, icmp6len);
 737                 /* m stays. */
 738                 break;
 739
 740         case ND_NEIGHBOR_SOLICIT:
 741                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
 742                 if (code != 0)
 743                         goto badcode;
 744                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 745                         goto badlen;
 746                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 747                         /* give up local */
 748                         nd6_ns_input(m, off, icmp6len);
 749                         m = NULL;
 750                         goto freeit;
 751                 }
 752                 nd6_ns_input(n, off, icmp6len);
 753                 /* m stays. */
 754                 break;
 755
 756         case ND_NEIGHBOR_ADVERT:
 757                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
 758                 if (code != 0)
 759                         goto badcode;
 760                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 761                         goto badlen;
 762                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 763                         /* give up local */
 764                         nd6_na_input(m, off, icmp6len);
 765                         m = NULL;
 766                         goto freeit;
 767                 }
 768                 nd6_na_input(n, off, icmp6len);
 769                 /* m stays. */
 770                 break;
 771
 772         case ND_REDIRECT:
 773                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
 774                 if (code != 0)
 775                         goto badcode;
 776                 if (icmp6len < sizeof(struct nd_redirect))
 777                         goto badlen;
 778                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 779                         /* give up local */
 780                         icmp6_redirect_input(m, off);
 781                         m = NULL;
 782                         goto freeit;
 783                 }
 784                 icmp6_redirect_input(n, off);
 785                 /* m stays. */
 786                 break;
 787
 788         case ICMP6_ROUTER_RENUMBERING:
 789                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 790                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 791                         goto badcode;
 792                 if (icmp6len < sizeof(struct icmp6_router_renum))
 793                         goto badlen;
 794                 break;
 795
 796         default:
 797                 nd6log((LOG_DEBUG,
 798                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 799                     icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
 800                     ip6_sprintf(&ip6->ip6_dst),
 801                     m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
 802                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 803                         /* ICMPv6 error: MUST deliver it by spec... */
 804                         code = PRC_NCMDS;
 805                         /* deliver */
 806                 } else {
 807                         /* ICMPv6 informational: MUST not deliver */
 808                         break;
 809                 }
 810         deliver:
 811                 if (icmp6_notify_error(&m, off, icmp6len, code)) {
 812                         /* In this case, m should've been freed. */
 813                         return (IPPROTO_DONE);
 814                 }
 815                 break;
 816
 817         badcode:
 818                 icmp6stat.icp6s_badcode++;
 819                 break;
 820
 821         badlen:
 822                 icmp6stat.icp6s_badlen++;
 823                 break;
 824         }
 825
 826         /* deliver the packet to appropriate sockets */
 827         icmp6_rip6_input(&m, *offp);
 828
 829         return IPPROTO_DONE;
 830
 831  freeit:
 832         m_freem(m);
 833         return IPPROTO_DONE;
 834 }
 835
 836 static int
 837 icmp6_notify_error(mp, off, icmp6len, code)
 838         struct mbuf **mp;
 839         int off, icmp6len, code;
 840 {
 841         struct mbuf *m = *mp;
 842         struct icmp6_hdr *icmp6;
 843         struct ip6_hdr *eip6;
 844         u_int32_t notifymtu;
 845         struct sockaddr_in6 icmp6src, icmp6dst;
 846
 847         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 848                 icmp6stat.icp6s_tooshort++;
 849                 goto freeit;
 850         }
 851 #ifndef PULLDOWN_TEST
 852         IP6_EXTHDR_CHECK(m, off,
 853             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
 854         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 855 #else
 856         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
 857             sizeof(*icmp6) + sizeof(struct ip6_hdr));
 858         if (icmp6 == NULL) {
 859                 icmp6stat.icp6s_tooshort++;
 860                 return (-1);
 861         }
 862 #endif
 863         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 864
 865         /* Detect the upper level protocol */
 866         {
 867                 void (*ctlfunc) __P((int, struct sockaddr *, void *));
 868                 u_int8_t nxt = eip6->ip6_nxt;
 869                 int eoff = off + sizeof(struct icmp6_hdr) +
 870                     sizeof(struct ip6_hdr);
 871                 struct ip6ctlparam ip6cp;
 872                 struct in6_addr *finaldst = NULL;
 873                 int icmp6type = icmp6->icmp6_type;
 874                 struct ip6_frag *fh;
 875                 struct ip6_rthdr *rth;
 876                 struct ip6_rthdr0 *rth0;
 877                 int rthlen;
 878
 879                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 880                         struct ip6_ext *eh;
 881
 882                         switch (nxt) {
 883                         case IPPROTO_HOPOPTS:
 884                         case IPPROTO_DSTOPTS:
 885                         case IPPROTO_AH:
 886 #ifndef PULLDOWN_TEST
 887                                 IP6_EXTHDR_CHECK(m, 0,
 888                                     eoff + sizeof(struct ip6_ext), -1);
 889                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
 890 #else
 891                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
 892                                     eoff, sizeof(*eh));
 893                                 if (eh == NULL) {
 894                                         icmp6stat.icp6s_tooshort++;
 895                                         return (-1);
 896                                 }
 897 #endif
 898
 899                                 if (nxt == IPPROTO_AH)
 900                                         eoff += (eh->ip6e_len + 2) << 2;
 901                                 else
 902                                         eoff += (eh->ip6e_len + 1) << 3;
 903                                 nxt = eh->ip6e_nxt;
 904                                 break;
 905                         case IPPROTO_ROUTING:
 906                                 /*
 907                                  * When the erroneous packet contains a
 908                                  * routing header, we should examine the
 909                                  * header to determine the final destination.
 910                                  * Otherwise, we can't properly update
 911                                  * information that depends on the final
 912                                  * destination (e.g. path MTU).
 913                                  */
 914 #ifndef PULLDOWN_TEST
 915                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
 916                                 rth = (struct ip6_rthdr *)
 917                                     (mtod(m, caddr_t) + eoff);
 918 #else
 919                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
 920                                     eoff, sizeof(*rth));
 921                                 if (rth == NULL) {
 922                                         icmp6stat.icp6s_tooshort++;
 923                                         return (-1);
 924                                 }
 925 #endif
 926                                 rthlen = (rth->ip6r_len + 1) << 3;
 927                                 /*
 928                                  * XXX: currently there is no
 929                                  * officially defined type other
 930                                  * than type-0.
 931                                  * Note that if the segment left field
 932                                  * is 0, all intermediate hops must
 933                                  * have been passed.
 934                                  */
 935                                 if (rth->ip6r_segleft &&
 936                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
 937                                         int hops;
 938
 939 #ifndef PULLDOWN_TEST
 940                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
 941                                         rth0 = (struct ip6_rthdr0 *)
 942                                             (mtod(m, caddr_t) + eoff);
 943 #else
 944                                         IP6_EXTHDR_GET(rth0,
 945                                             struct ip6_rthdr0 *, m,
 946                                             eoff, rthlen);
 947                                         if (rth0 == NULL) {
 948                                                 icmp6stat.icp6s_tooshort++;
 949                                                 return (-1);
 950                                         }
 951 #endif
 952                                         /* just ignore a bogus header */
 953                                         if ((rth0->ip6r0_len % 2) == 0 &&
 954                                             (hops = rth0->ip6r0_len/2))
 955                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
 956                                 }
 957                                 eoff += rthlen;
 958                                 nxt = rth->ip6r_nxt;
 959                                 break;
 960                         case IPPROTO_FRAGMENT:
 961 #ifndef PULLDOWN_TEST
 962                                 IP6_EXTHDR_CHECK(m, 0, eoff +
 963                                     sizeof(struct ip6_frag), -1);
 964                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
 965                                     eoff);
 966 #else
 967                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
 968                                     eoff, sizeof(*fh));
 969                                 if (fh == NULL) {
 970                                         icmp6stat.icp6s_tooshort++;
 971                                         return (-1);
 972                                 }
 973 #endif
 974                                 /*
 975                                  * Data after a fragment header is meaningless
 976                                  * unless it is the first fragment, but
 977                                  * we'll go to the notify label for path MTU
 978                                  * discovery.
 979                                  */
 980                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
 981                                         goto notify;
 982
 983                                 eoff += sizeof(struct ip6_frag);
 984                                 nxt = fh->ip6f_nxt;
 985                                 break;
 986                         default:
 987                                 /*
 988                                  * This case includes ESP and the No Next
 989                                  * Header.  In such cases going to the notify
 990                                  * label does not have any meaning
 991                                  * (i.e. ctlfunc will be NULL), but we go
 992                                  * anyway since we might have to update
 993                                  * path MTU information.
 994                                  */
 995                                 goto notify;
 996                         }
 997                 }
 998           notify:
 999 #ifndef PULLDOWN_TEST
1000                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1001 #else
1002                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1003                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1004                 if (icmp6 == NULL) {
1005                         icmp6stat.icp6s_tooshort++;
1006                         return (-1);
1007                 }
1008 #endif
1009
1010                 /*
1011                  * retrieve parameters from the inner IPv6 header, and convert
1012                  * them into sockaddr structures.
1013                  * XXX: there is no guarantee that the source or destination
1014                  * addresses of the inner packet are in the same scope as
1015                  * the addresses of the icmp packet.  But there is no other
1016                  * way to determine the zone.
1017                  */
1018                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1019
1020                 bzero(&icmp6dst, sizeof(icmp6dst));
1021                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1022                 icmp6dst.sin6_family = AF_INET6;
1023                 if (finaldst == NULL)
1024                         icmp6dst.sin6_addr = eip6->ip6_dst;
1025                 else
1026                         icmp6dst.sin6_addr = *finaldst;
1027                 if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6dst.sin6_addr,
1028                     &icmp6dst.sin6_scope_id))
1029                         goto freeit;
1030                 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1031                                    NULL, NULL)) {
1032                         /* should be impossible */
1033                         nd6log((LOG_DEBUG,
1034                             "icmp6_notify_error: in6_embedscope failed\n"));
1035                         goto freeit;
1036                 }
1037
1038                 /*
1039                  * retrieve parameters from the inner IPv6 header, and convert
1040                  * them into sockaddr structures.
1041                  */
1042                 bzero(&icmp6src, sizeof(icmp6src));
1043                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1044                 icmp6src.sin6_family = AF_INET6;
1045                 icmp6src.sin6_addr = eip6->ip6_src;
1046                 if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6src.sin6_addr,
1047                     &icmp6src.sin6_scope_id)) {
1048                         goto freeit;
1049                 }
1050                 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1051                                    NULL, NULL)) {
1052                         /* should be impossible */
1053                         nd6log((LOG_DEBUG,
1054                             "icmp6_notify_error: in6_embedscope failed\n"));
1055                         goto freeit;
1056                 }
1057                 icmp6src.sin6_flowinfo = (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1058
1059                 if (finaldst == NULL)
1060                         finaldst = &eip6->ip6_dst;
1061                 ip6cp.ip6c_m = m;
1062                 ip6cp.ip6c_icmp6 = icmp6;
1063                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1064                 ip6cp.ip6c_off = eoff;
1065                 ip6cp.ip6c_finaldst = finaldst;
1066                 ip6cp.ip6c_src = &icmp6src;
1067                 ip6cp.ip6c_nxt = nxt;
1068
1069                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1070                         notifymtu = ntohl(icmp6->icmp6_mtu);
1071                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1072                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1073                 }
1074
1075                 ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1076                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1077                 if (ctlfunc) {
1078                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1079                             &ip6cp);
1080                 }
1081         }
1082         *mp = m;
1083         return (0);
1084
1085   freeit:
1086         m_freem(m);
1087         return (-1);
1088 }
1089
1090 void
1091 icmp6_mtudisc_update(ip6cp, validated)
1092         struct ip6ctlparam *ip6cp;
1093         int validated;
1094 {
1095         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1096         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1097         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1098         u_int mtu = ntohl(icmp6->icmp6_mtu);
1099         struct in_conninfo inc;
1100
1101 #if 0
1102         /*
1103          * RFC2460 section 5, last paragraph.
1104          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1105          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1106          * due to packet translator in the middle.
1107          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1108          * special handling.
1109          */
1110         if (mtu < IPV6_MMTU)
1111                 return;
1112 #endif
1113
1114         /*
1115          * we reject ICMPv6 too big with abnormally small value.
1116          * XXX what is the good definition of "abnormally small"?
1117          */
1118         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1119                 return;
1120
1121         if (!validated)
1122                 return;
1123
1124         bzero(&inc, sizeof(inc));
1125         inc.inc_flags = 1; /* IPv6 */
1126         inc.inc6_faddr = *dst;
1127         /* XXX normally, this won't happen */
1128         if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1129                 inc.inc6_faddr.s6_addr16[1] =
1130                     htons(m->m_pkthdr.rcvif->if_index);
1131         }
1132
1133         if (mtu < tcp_maxmtu6(&inc)) {
1134                 tcp_hc_updatemtu(&inc, mtu);
1135                 icmp6stat.icp6s_pmtuchg++;
1136         }
1137 }
1138
1139 /*
1140  * Process a Node Information Query packet, based on
1141  * draft-ietf-ipngwg-icmp-name-lookups-07.
1142  *
1143  * Spec incompatibilities:
1144  * - IPv6 Subject address handling
1145  * - IPv4 Subject address handling support missing
1146  * - Proxy reply (answer even if it's not for me)
1147  * - joins NI group address at in6_ifattach() time only, does not cope
1148  *   with hostname changes by sethostname(3)
1149  */
1150 #define hostnamelen     strlen(hostname)
1151 static struct mbuf *
1152 ni6_input(m, off)
1153         struct mbuf *m;
1154         int off;
1155 {
1156         struct icmp6_nodeinfo *ni6, *nni6;
1157         struct mbuf *n = NULL;
1158         u_int16_t qtype;
1159         int subjlen;
1160         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1161         struct ni_reply_fqdn *fqdn;
1162         int addrs;              /* for NI_QTYPE_NODEADDR */
1163         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1164         struct sockaddr_in6 sin6_sbj; /* subject address */
1165         struct ip6_hdr *ip6;
1166         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1167         char *subj = NULL;
1168         struct in6_ifaddr *ia6 = NULL;
1169
1170         ip6 = mtod(m, struct ip6_hdr *);
1171 #ifndef PULLDOWN_TEST
1172         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1173 #else
1174         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1175         if (ni6 == NULL) {
1176                 /* m is already reclaimed */
1177                 return (NULL);
1178         }
1179 #endif
1180
1181         /*
1182          * Validate IPv6 destination address.
1183          *
1184          * The Responder must discard the Query without further processing
1185          * unless it is one of the Responder's unicast or anycast addresses, or
1186          * a link-local scope multicast address which the Responder has joined.
1187          * [icmp-name-lookups-08, Section 4.]
1188          */
1189         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1190                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1191                         goto bad;
1192                 /* else it's a link-local multicast, fine */
1193         } else {                /* unicast or anycast */
1194                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1195                         goto bad; /* XXX impossible */
1196
1197                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1198                     !(icmp6_nodeinfo & 4)) {
1199                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1200                                 "a temporary address in %s:%d",
1201                                __FILE__, __LINE__));
1202                         goto bad;
1203                 }
1204         }
1205
1206         /* validate query Subject field. */
1207         qtype = ntohs(ni6->ni_qtype);
1208         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1209         switch (qtype) {
1210         case NI_QTYPE_NOOP:
1211         case NI_QTYPE_SUPTYPES:
1212                 /* 07 draft */
1213                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1214                         break;
1215                 /* FALLTHROUGH */
1216         case NI_QTYPE_FQDN:
1217         case NI_QTYPE_NODEADDR:
1218                 switch (ni6->ni_code) {
1219                 case ICMP6_NI_SUBJ_IPV6:
1220 #if ICMP6_NI_SUBJ_IPV6 != 0
1221                 case 0:
1222 #endif
1223                         /*
1224                          * backward compatibility - try to accept 03 draft
1225                          * format, where no Subject is present.
1226                          */
1227                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1228                             subjlen == 0) {
1229                                 oldfqdn++;
1230                                 break;
1231                         }
1232 #if ICMP6_NI_SUBJ_IPV6 != 0
1233                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1234                                 goto bad;
1235 #endif
1236
1237                         if (subjlen != sizeof(struct in6_addr))
1238                                 goto bad;
1239
1240                         /*
1241                          * Validate Subject address.
1242                          *
1243                          * Not sure what exactly "address belongs to the node"
1244                          * means in the spec, is it just unicast, or what?
1245                          *
1246                          * At this moment we consider Subject address as
1247                          * "belong to the node" if the Subject address equals
1248                          * to the IPv6 destination address; validation for
1249                          * IPv6 destination address should have done enough
1250                          * check for us.
1251                          *
1252                          * We do not do proxy at this moment.
1253                          */
1254                         /* m_pulldown instead of copy? */
1255                         bzero(&sin6_sbj, sizeof(sin6_sbj));
1256                         sin6_sbj.sin6_family = AF_INET6;
1257                         sin6_sbj.sin6_len = sizeof(sin6_sbj);
1258                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1259                             subjlen, (caddr_t)&sin6_sbj.sin6_addr);
1260                         if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1261                             &sin6_sbj.sin6_addr, &sin6_sbj.sin6_scope_id)) {
1262                                 goto bad;
1263                         }
1264                         if (in6_embedscope(&sin6_sbj.sin6_addr, &sin6_sbj,
1265                             NULL, NULL))
1266                                 goto bad; /* XXX should not happen */
1267
1268                         subj = (char *)&sin6_sbj;
1269                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1270                             &sin6_sbj.sin6_addr))
1271                                 break;
1272
1273                         /*
1274                          * XXX if we are to allow other cases, we should really
1275                          * be careful about scope here.
1276                          * basically, we should disallow queries toward IPv6
1277                          * destination X with subject Y,
1278                          * if scope(X) > scope(Y).
1279                          * if we allow scope(X) > scope(Y), it will result in
1280                          * information leakage across scope boundary.
1281                          */
1282                         goto bad;
1283
1284                 case ICMP6_NI_SUBJ_FQDN:
1285                         /*
1286                          * Validate Subject name with gethostname(3).
1287                          *
1288                          * The behavior may need some debate, since:
1289                          * - we are not sure if the node has FQDN as
1290                          *   hostname (returned by gethostname(3)).
1291                          * - the code does wildcard match for truncated names.
1292                          *   however, we are not sure if we want to perform
1293                          *   wildcard match, if gethostname(3) side has
1294                          *   truncated hostname.
1295                          */
1296                         n = ni6_nametodns(hostname, hostnamelen, 0);
1297                         if (!n || n->m_next || n->m_len == 0)
1298                                 goto bad;
1299                         IP6_EXTHDR_GET(subj, char *, m,
1300                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1301                         if (subj == NULL)
1302                                 goto bad;
1303                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1304                             n->m_len)) {
1305                                 goto bad;
1306                         }
1307                         m_freem(n);
1308                         n = NULL;
1309                         break;
1310
1311                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1312                 default:
1313                         goto bad;
1314                 }
1315                 break;
1316         }
1317
1318         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1319         switch (qtype) {
1320         case NI_QTYPE_FQDN:
1321                 if ((icmp6_nodeinfo & 1) == 0)
1322                         goto bad;
1323                 break;
1324         case NI_QTYPE_NODEADDR:
1325                 if ((icmp6_nodeinfo & 2) == 0)
1326                         goto bad;
1327                 break;
1328         }
1329
1330         /* guess reply length */
1331         switch (qtype) {
1332         case NI_QTYPE_NOOP:
1333                 break;          /* no reply data */
1334         case NI_QTYPE_SUPTYPES:
1335                 replylen += sizeof(u_int32_t);
1336                 break;
1337         case NI_QTYPE_FQDN:
1338                 /* XXX will append an mbuf */
1339                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1340                 break;
1341         case NI_QTYPE_NODEADDR:
1342                 addrs = ni6_addrs(ni6, m, &ifp, subj);
1343                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1344                     sizeof(u_int32_t))) > MCLBYTES)
1345                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1346                 break;
1347         default:
1348                 /*
1349                  * XXX: We must return a reply with the ICMP6 code
1350                  * `unknown Qtype' in this case.  However we regard the case
1351                  * as an FQDN query for backward compatibility.
1352                  * Older versions set a random value to this field,
1353                  * so it rarely varies in the defined qtypes.
1354                  * But the mechanism is not reliable...
1355                  * maybe we should obsolete older versions.
1356                  */
1357                 qtype = NI_QTYPE_FQDN;
1358                 /* XXX will append an mbuf */
1359                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1360                 oldfqdn++;
1361                 break;
1362         }
1363
1364         /* allocate an mbuf to reply. */
1365         MGETHDR(n, M_DONTWAIT, m->m_type);
1366         if (n == NULL) {
1367                 m_freem(m);
1368                 return (NULL);
1369         }
1370         M_MOVE_PKTHDR(n, m); /* just for recvif */
1371         if (replylen > MHLEN) {
1372                 if (replylen > MCLBYTES) {
1373                         /*
1374                          * XXX: should we try to allocate more? But MCLBYTES
1375                          * is probably much larger than IPV6_MMTU...
1376                          */
1377                         goto bad;
1378                 }
1379                 MCLGET(n, M_DONTWAIT);
1380                 if ((n->m_flags & M_EXT) == 0) {
1381                         goto bad;
1382                 }
1383         }
1384         n->m_pkthdr.len = n->m_len = replylen;
1385
1386         /* copy mbuf header and IPv6 + Node Information base headers */
1387         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1388         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1389         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1390
1391         /* qtype dependent procedure */
1392         switch (qtype) {
1393         case NI_QTYPE_NOOP:
1394                 nni6->ni_code = ICMP6_NI_SUCCESS;
1395                 nni6->ni_flags = 0;
1396                 break;
1397         case NI_QTYPE_SUPTYPES:
1398         {
1399                 u_int32_t v;
1400                 nni6->ni_code = ICMP6_NI_SUCCESS;
1401                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1402                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1403                 v = (u_int32_t)htonl(0x0000000f);
1404                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1405                 break;
1406         }
1407         case NI_QTYPE_FQDN:
1408                 nni6->ni_code = ICMP6_NI_SUCCESS;
1409                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1410                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1411                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1412                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1413                 /*
1414                  * XXX do we really have FQDN in variable "hostname"?
1415                  */
1416                 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1417                 if (n->m_next == NULL)
1418                         goto bad;
1419                 /* XXX we assume that n->m_next is not a chain */
1420                 if (n->m_next->m_next != NULL)
1421                         goto bad;
1422                 n->m_pkthdr.len += n->m_next->m_len;
1423                 break;
1424         case NI_QTYPE_NODEADDR:
1425         {
1426                 int lenlim, copied;
1427
1428                 nni6->ni_code = ICMP6_NI_SUCCESS;
1429                 n->m_pkthdr.len = n->m_len =
1430                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1431                 lenlim = M_TRAILINGSPACE(n);
1432                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1433                 /* XXX: reset mbuf length */
1434                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1435                     sizeof(struct icmp6_nodeinfo) + copied;
1436                 break;
1437         }
1438         default:
1439                 break;          /* XXX impossible! */
1440         }
1441
1442         nni6->ni_type = ICMP6_NI_REPLY;
1443         m_freem(m);
1444         return (n);
1445
1446   bad:
1447         m_freem(m);
1448         if (n)
1449                 m_freem(n);
1450         return (NULL);
1451 }
1452 #undef hostnamelen
1453
1454 /*
1455  * make a mbuf with DNS-encoded string.  no compression support.
1456  *
1457  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1458  * treated as truncated name (two \0 at the end).  this is a wild guess.
1459  */
1460 static struct mbuf *
1461 ni6_nametodns(name, namelen, old)
1462         const char *name;
1463         int namelen;
1464         int old;        /* return pascal string if non-zero */
1465 {
1466         struct mbuf *m;
1467         char *cp, *ep;
1468         const char *p, *q;
1469         int i, len, nterm;
1470
1471         if (old)
1472                 len = namelen + 1;
1473         else
1474                 len = MCLBYTES;
1475
1476         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1477         MGET(m, M_DONTWAIT, MT_DATA);
1478         if (m && len > MLEN) {
1479                 MCLGET(m, M_DONTWAIT);
1480                 if ((m->m_flags & M_EXT) == 0)
1481                         goto fail;
1482         }
1483         if (!m)
1484                 goto fail;
1485         m->m_next = NULL;
1486
1487         if (old) {
1488                 m->m_len = len;
1489                 *mtod(m, char *) = namelen;
1490                 bcopy(name, mtod(m, char *) + 1, namelen);
1491                 return m;
1492         } else {
1493                 m->m_len = 0;
1494                 cp = mtod(m, char *);
1495                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1496
1497                 /* if not certain about my name, return empty buffer */
1498                 if (namelen == 0)
1499                         return m;
1500
1501                 /*
1502                  * guess if it looks like shortened hostname, or FQDN.
1503                  * shortened hostname needs two trailing "\0".
1504                  */
1505                 i = 0;
1506                 for (p = name; p < name + namelen; p++) {
1507                         if (*p && *p == '.')
1508                                 i++;
1509                 }
1510                 if (i < 2)
1511                         nterm = 2;
1512                 else
1513                         nterm = 1;
1514
1515                 p = name;
1516                 while (cp < ep && p < name + namelen) {
1517                         i = 0;
1518                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1519                                 i++;
1520                         /* result does not fit into mbuf */
1521                         if (cp + i + 1 >= ep)
1522                                 goto fail;
1523                         /*
1524                          * DNS label length restriction, RFC1035 page 8.
1525                          * "i == 0" case is included here to avoid returning
1526                          * 0-length label on "foo..bar".
1527                          */
1528                         if (i <= 0 || i >= 64)
1529                                 goto fail;
1530                         *cp++ = i;
1531                         bcopy(p, cp, i);
1532                         cp += i;
1533                         p = q;
1534                         if (p < name + namelen && *p == '.')
1535                                 p++;
1536                 }
1537                 /* termination */
1538                 if (cp + nterm >= ep)
1539                         goto fail;
1540                 while (nterm-- > 0)
1541                         *cp++ = '\0';
1542                 m->m_len = cp - mtod(m, char *);
1543                 return m;
1544         }
1545
1546         panic("should not reach here");
1547         /* NOTREACHED */
1548
1549  fail:
1550         if (m)
1551                 m_freem(m);
1552         return NULL;
1553 }
1554
1555 /*
1556  * check if two DNS-encoded string matches.  takes care of truncated
1557  * form (with \0\0 at the end).  no compression support.
1558  * XXX upper/lowercase match (see RFC2065)
1559  */
1560 static int
1561 ni6_dnsmatch(a, alen, b, blen)
1562         const char *a;
1563         int alen;
1564         const char *b;
1565         int blen;
1566 {
1567         const char *a0, *b0;
1568         int l;
1569
1570         /* simplest case - need validation? */
1571         if (alen == blen && bcmp(a, b, alen) == 0)
1572                 return 1;
1573
1574         a0 = a;
1575         b0 = b;
1576
1577         /* termination is mandatory */
1578         if (alen < 2 || blen < 2)
1579                 return 0;
1580         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1581                 return 0;
1582         alen--;
1583         blen--;
1584
1585         while (a - a0 < alen && b - b0 < blen) {
1586                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1587                         return 0;
1588
1589                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1590                         return 0;
1591                 /* we don't support compression yet */
1592                 if (a[0] >= 64 || b[0] >= 64)
1593                         return 0;
1594
1595                 /* truncated case */
1596                 if (a[0] == 0 && a - a0 == alen - 1)
1597                         return 1;
1598                 if (b[0] == 0 && b - b0 == blen - 1)
1599                         return 1;
1600                 if (a[0] == 0 || b[0] == 0)
1601                         return 0;
1602
1603                 if (a[0] != b[0])
1604                         return 0;
1605                 l = a[0];
1606                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1607                         return 0;
1608                 if (bcmp(a + 1, b + 1, l) != 0)
1609                         return 0;
1610
1611                 a += 1 + l;
1612                 b += 1 + l;
1613         }
1614
1615         if (a - a0 == alen && b - b0 == blen)
1616                 return 1;
1617         else
1618                 return 0;
1619 }
1620
1621 /*
1622  * calculate the number of addresses to be returned in the node info reply.
1623  */
1624 static int
1625 ni6_addrs(ni6, m, ifpp, subj)
1626         struct icmp6_nodeinfo *ni6;
1627         struct mbuf *m;
1628         struct ifnet **ifpp;
1629         char *subj;
1630 {
1631         struct ifnet *ifp;
1632         struct in6_ifaddr *ifa6;
1633         struct ifaddr *ifa;
1634         struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1635         int addrs = 0, addrsofif, iffound = 0;
1636         int niflags = ni6->ni_flags;
1637
1638         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1639                 switch (ni6->ni_code) {
1640                 case ICMP6_NI_SUBJ_IPV6:
1641                         if (subj == NULL) /* must be impossible... */
1642                                 return (0);
1643                         subj_ip6 = (struct sockaddr_in6 *)subj;
1644                         break;
1645                 default:
1646                         /*
1647                          * XXX: we only support IPv6 subject address for
1648                          * this Qtype.
1649                          */
1650                         return (0);
1651                 }
1652         }
1653
1654         IFNET_RLOCK();
1655         for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1656                 addrsofif = 0;
1657                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1658                         if (ifa->ifa_addr->sa_family != AF_INET6)
1659                                 continue;
1660                         ifa6 = (struct in6_ifaddr *)ifa;
1661
1662                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1663                             IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1664                             &ifa6->ia_addr.sin6_addr))
1665                                 iffound = 1;
1666
1667                         /*
1668                          * IPv4-mapped addresses can only be returned by a
1669                          * Node Information proxy, since they represent
1670                          * addresses of IPv4-only nodes, which perforce do
1671                          * not implement this protocol.
1672                          * [icmp-name-lookups-07, Section 5.4]
1673                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1674                          * this function at this moment.
1675                          */
1676
1677                         /* What do we have to do about ::1? */
1678                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1679                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1680                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1681                                         continue;
1682                                 break;
1683                         case IPV6_ADDR_SCOPE_SITELOCAL:
1684                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1685                                         continue;
1686                                 break;
1687                         case IPV6_ADDR_SCOPE_GLOBAL:
1688                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1689                                         continue;
1690                                 break;
1691                         default:
1692                                 continue;
1693                         }
1694
1695                         /*
1696                          * check if anycast is okay.
1697                          * XXX: just experimental.  not in the spec.
1698                          */
1699                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1700                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1701                                 continue; /* we need only unicast addresses */
1702                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1703                             (icmp6_nodeinfo & 4) == 0) {
1704                                 continue;
1705                         }
1706                         addrsofif++; /* count the address */
1707                 }
1708                 if (iffound) {
1709                         *ifpp = ifp;
1710                         IFNET_RUNLOCK();
1711                         return (addrsofif);
1712                 }
1713
1714                 addrs += addrsofif;
1715         }
1716         IFNET_RUNLOCK();
1717
1718         return (addrs);
1719 }
1720
1721 static int
1722 ni6_store_addrs(ni6, nni6, ifp0, resid)
1723         struct icmp6_nodeinfo *ni6, *nni6;
1724         struct ifnet *ifp0;
1725         int resid;
1726 {
1727         struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1728         struct in6_ifaddr *ifa6;
1729         struct ifaddr *ifa;
1730         struct ifnet *ifp_dep = NULL;
1731         int copied = 0, allow_deprecated = 0;
1732         u_char *cp = (u_char *)(nni6 + 1);
1733         int niflags = ni6->ni_flags;
1734         u_int32_t ltime;
1735
1736         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1737                 return (0);     /* needless to copy */
1738
1739         IFNET_RLOCK();
1740   again:
1741
1742         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1743                 for (ifa = ifp->if_addrlist.tqh_first; ifa;
1744                      ifa = ifa->ifa_list.tqe_next) {
1745                         if (ifa->ifa_addr->sa_family != AF_INET6)
1746                                 continue;
1747                         ifa6 = (struct in6_ifaddr *)ifa;
1748
1749                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1750                             allow_deprecated == 0) {
1751                                 /*
1752                                  * prefererred address should be put before
1753                                  * deprecated addresses.
1754                                  */
1755
1756                                 /* record the interface for later search */
1757                                 if (ifp_dep == NULL)
1758                                         ifp_dep = ifp;
1759
1760                                 continue;
1761                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1762                             allow_deprecated != 0)
1763                                 continue; /* we now collect deprecated addrs */
1764
1765                         /* What do we have to do about ::1? */
1766                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1767                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1768                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1769                                         continue;
1770                                 break;
1771                         case IPV6_ADDR_SCOPE_SITELOCAL:
1772                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1773                                         continue;
1774                                 break;
1775                         case IPV6_ADDR_SCOPE_GLOBAL:
1776                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1777                                         continue;
1778                                 break;
1779                         default:
1780                                 continue;
1781                         }
1782
1783                         /*
1784                          * check if anycast is okay.
1785                          * XXX: just experimental.  not in the spec.
1786                          */
1787                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1788                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1789                                 continue;
1790                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1791                             (icmp6_nodeinfo & 4) == 0) {
1792                                 continue;
1793                         }
1794
1795                         /* now we can copy the address */
1796                         if (resid < sizeof(struct in6_addr) +
1797                             sizeof(u_int32_t)) {
1798                                 /*
1799                                  * We give up much more copy.
1800                                  * Set the truncate flag and return.
1801                                  */
1802                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1803                                 IFNET_RUNLOCK();
1804                                 return (copied);
1805                         }
1806
1807                         /*
1808                          * Set the TTL of the address.
1809                          * The TTL value should be one of the following
1810                          * according to the specification:
1811                          *
1812                          * 1. The remaining lifetime of a DHCP lease on the
1813                          *    address, or
1814                          * 2. The remaining Valid Lifetime of a prefix from
1815                          *    which the address was derived through Stateless
1816                          *    Autoconfiguration.
1817                          *
1818                          * Note that we currently do not support stateful
1819                          * address configuration by DHCPv6, so the former
1820                          * case can't happen.
1821                          */
1822                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1823                                 ltime = ND6_INFINITE_LIFETIME;
1824                         else {
1825                                 if (ifa6->ia6_lifetime.ia6t_expire >
1826                                     time_second)
1827                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1828                                 else
1829                                         ltime = 0;
1830                         }
1831
1832                         bcopy(&ltime, cp, sizeof(u_int32_t));
1833                         cp += sizeof(u_int32_t);
1834
1835                         /* copy the address itself */
1836                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1837                             sizeof(struct in6_addr));
1838                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1839                         cp += sizeof(struct in6_addr);
1840
1841                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1842                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1843                 }
1844                 if (ifp0)       /* we need search only on the specified IF */
1845                         break;
1846         }
1847
1848         if (allow_deprecated == 0 && ifp_dep != NULL) {
1849                 ifp = ifp_dep;
1850                 allow_deprecated = 1;
1851
1852                 goto again;
1853         }
1854
1855         IFNET_RUNLOCK();
1856
1857         return (copied);
1858 }
1859
1860 /*
1861  * XXX almost dup'ed code with rip6_input.
1862  */
1863 static int
1864 icmp6_rip6_input(mp, off)
1865         struct  mbuf **mp;
1866         int     off;
1867 {
1868         struct mbuf *m = *mp;
1869         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1870         struct in6pcb *in6p;
1871         struct in6pcb *last = NULL;
1872         struct sockaddr_in6 fromsa;
1873         struct icmp6_hdr *icmp6;
1874         struct mbuf *opts = NULL;
1875
1876 #ifndef PULLDOWN_TEST
1877         /* this is assumed to be safe. */
1878         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1879 #else
1880         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1881         if (icmp6 == NULL) {
1882                 /* m is already reclaimed */
1883                 return (IPPROTO_DONE);
1884         }
1885 #endif
1886
1887         bzero(&fromsa, sizeof(fromsa));
1888         fromsa.sin6_len = sizeof(struct sockaddr_in6);
1889         fromsa.sin6_family = AF_INET6;
1890         /* KAME hack: recover scopeid */
1891         (void)in6_recoverscope(&fromsa, &ip6->ip6_src, m->m_pkthdr.rcvif);
1892
1893         INP_INFO_RLOCK(&ripcbinfo);
1894         LIST_FOREACH(in6p, &ripcb, inp_list) {
1895                 INP_LOCK(in6p);
1896                 if ((in6p->inp_vflag & INP_IPV6) == 0) {
1897         docontinue:
1898                         INP_UNLOCK(in6p);
1899                         continue;
1900                 }
1901                 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1902                         goto docontinue;
1903                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1904                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1905                         goto docontinue;
1906                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1907                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1908                         goto docontinue;
1909                 if (in6p->in6p_icmp6filt
1910                     && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1911                                  in6p->in6p_icmp6filt))
1912                         goto docontinue;
1913                 if (last) {
1914                         struct  mbuf *n = NULL;
1915
1916                         /*
1917                          * Recent network drivers tend to allocate a single
1918                          * mbuf cluster, rather than to make a couple of
1919                          * mbufs without clusters.  Also, since the IPv6 code
1920                          * path tries to avoid m_pullup(), it is highly
1921                          * probable that we still have an mbuf cluster here
1922                          * even though the necessary length can be stored in an
1923                          * mbuf's internal buffer.
1924                          * Meanwhile, the default size of the receive socket
1925                          * buffer for raw sockets is not so large.  This means
1926                          * the possibility of packet loss is relatively higher
1927                          * than before.  To avoid this scenario, we copy the
1928                          * received data to a separate mbuf that does not use
1929                          * a cluster, if possible.
1930                          * XXX: it is better to copy the data after stripping
1931                          * intermediate headers.
1932                          */
1933                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1934                             m->m_len <= MHLEN) {
1935                                 MGET(n, M_DONTWAIT, m->m_type);
1936                                 if (n != NULL) {
1937                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1938                                                 bcopy(m->m_data, n->m_data,
1939                                                       m->m_len);
1940                                                 n->m_len = m->m_len;
1941                                         } else {
1942                                                 m_free(n);
1943                                                 n = NULL;
1944                                         }
1945                                 }
1946                         }
1947                         if (n != NULL ||
1948                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1949                                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1950                                         ip6_savecontrol(last, n, &opts);
1951                                 /* strip intermediate headers */
1952                                 m_adj(n, off);
1953                                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1954                                     (struct sockaddr *)&fromsa, n, opts)
1955                                     == 0) {
1956                                         /* should notify about lost packet */
1957                                         m_freem(n);
1958                                         if (opts) {
1959                                                 m_freem(opts);
1960                                         }
1961                                 } else
1962                                         sorwakeup(last->in6p_socket);
1963                                 opts = NULL;
1964                         }
1965                         INP_UNLOCK(last);
1966                 }
1967                 last = in6p;
1968         }
1969         if (last) {
1970                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1971                         ip6_savecontrol(last, m, &opts);
1972                 /* strip intermediate headers */
1973                 m_adj(m, off);
1974
1975                 /* avoid using mbuf clusters if possible (see above) */
1976                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1977                     m->m_len <= MHLEN) {
1978                         struct mbuf *n;
1979
1980                         MGET(n, M_DONTWAIT, m->m_type);
1981                         if (n != NULL) {
1982                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1983                                         bcopy(m->m_data, n->m_data, m->m_len);
1984                                         n->m_len = m->m_len;
1985
1986                                         m_freem(m);
1987                                         m = n;
1988                                 } else {
1989                                         m_freem(n);
1990                                         n = NULL;
1991                                 }
1992                         }
1993                 }
1994                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1995                     (struct sockaddr *)&fromsa, m, opts) == 0) {
1996                         m_freem(m);
1997                         if (opts)
1998                                 m_freem(opts);
1999                 } else
2000                         sorwakeup(last->in6p_socket);
2001                 INP_UNLOCK(last);
2002         } else {
2003                 m_freem(m);
2004                 ip6stat.ip6s_delivered--;
2005         }
2006         INP_INFO_RUNLOCK(&ripcbinfo);
2007         return IPPROTO_DONE;
2008 }
2009
2010 /*
2011  * Reflect the ip6 packet back to the source.
2012  * OFF points to the icmp6 header, counted from the top of the mbuf.
2013  */
2014 void
2015 icmp6_reflect(m, off)
2016         struct  mbuf *m;
2017         size_t off;
2018 {
2019         struct ip6_hdr *ip6;
2020         struct icmp6_hdr *icmp6;
2021         struct in6_ifaddr *ia;
2022         struct in6_addr t, *src = 0;
2023         int plen;
2024         int type, code;
2025         struct ifnet *outif = NULL;
2026         struct sockaddr_in6 sa6_src, sa6_dst;
2027 #ifdef COMPAT_RFC1885
2028         int mtu = IPV6_MMTU;
2029         struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
2030 #endif
2031
2032         /* too short to reflect */
2033         if (off < sizeof(struct ip6_hdr)) {
2034                 nd6log((LOG_DEBUG,
2035                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2036                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2037                     __FILE__, __LINE__));
2038                 goto bad;
2039         }
2040
2041         /*
2042          * If there are extra headers between IPv6 and ICMPv6, strip
2043          * off that header first.
2044          */
2045 #ifdef DIAGNOSTIC
2046         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2047                 panic("assumption failed in icmp6_reflect");
2048 #endif
2049         if (off > sizeof(struct ip6_hdr)) {
2050                 size_t l;
2051                 struct ip6_hdr nip6;
2052
2053                 l = off - sizeof(struct ip6_hdr);
2054                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2055                 m_adj(m, l);
2056                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2057                 if (m->m_len < l) {
2058                         if ((m = m_pullup(m, l)) == NULL)
2059                                 return;
2060                 }
2061                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2062         } else /* off == sizeof(struct ip6_hdr) */ {
2063                 size_t l;
2064                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2065                 if (m->m_len < l) {
2066                         if ((m = m_pullup(m, l)) == NULL)
2067                                 return;
2068                 }
2069         }
2070         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2071         ip6 = mtod(m, struct ip6_hdr *);
2072         ip6->ip6_nxt = IPPROTO_ICMPV6;
2073         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2074         type = icmp6->icmp6_type; /* keep type for statistics */
2075         code = icmp6->icmp6_code; /* ditto. */
2076
2077         t = ip6->ip6_dst;
2078         /*
2079          * ip6_input() drops a packet if its src is multicast.
2080          * So, the src is never multicast.
2081          */
2082         ip6->ip6_dst = ip6->ip6_src;
2083
2084         /*
2085          * XXX: make sure to embed scope zone information, using
2086          * already embedded IDs or the received interface (if any).
2087          * Note that rcvif may be NULL.
2088          * TODO: scoped routing case (XXX).
2089          */
2090         bzero(&sa6_src, sizeof(sa6_src));
2091         sa6_src.sin6_family = AF_INET6;
2092         sa6_src.sin6_len = sizeof(sa6_src);
2093         sa6_src.sin6_addr = ip6->ip6_dst;
2094         in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2095         if (in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL))
2096                 goto bad;
2097
2098         bzero(&sa6_dst, sizeof(sa6_dst));
2099         sa6_dst.sin6_family = AF_INET6;
2100         sa6_dst.sin6_len = sizeof(sa6_dst);
2101         sa6_dst.sin6_addr = t;
2102         in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2103         if (in6_embedscope(&t, &sa6_dst, NULL, NULL))
2104                 goto bad;
2105
2106 #ifdef COMPAT_RFC1885
2107         /*
2108          * xxx guess MTU
2109          * RFC 1885 requires that echo reply should be truncated if it
2110          * does not fit in with (return) path MTU, but the description was
2111          * removed in the new spec.
2112          */
2113         if (icmp6_reflect_rt.ro_rt == 0 ||
2114             ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2115                 if (icmp6_reflect_rt.ro_rt) {
2116                         RTFREE(icmp6_reflect_rt.ro_rt);
2117                         icmp6_reflect_rt.ro_rt = 0;
2118                 }
2119                 bzero(sin6, sizeof(*sin6));
2120                 sin6->sin6_family = PF_INET6;
2121                 sin6->sin6_len = sizeof(struct sockaddr_in6);
2122                 sin6->sin6_addr = ip6->ip6_dst;
2123
2124                 rtalloc((struct route *)&icmp6_reflect_rt.ro_rt);
2125         }
2126
2127         if (icmp6_reflect_rt.ro_rt == 0)
2128                 goto bad;
2129
2130         if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2131             && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2132                 mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2133
2134         if (mtu < m->m_pkthdr.len) {
2135                 plen -= (m->m_pkthdr.len - mtu);
2136                 m_adj(m, mtu - m->m_pkthdr.len);
2137         }
2138 #endif
2139         /*
2140          * If the incoming packet was addressed directly to us (i.e. unicast),
2141          * use dst as the src for the reply.
2142          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2143          * (for example) when we encounter an error while forwarding procedure
2144          * destined to a duplicated address of ours.
2145          */
2146         for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2147                 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2148                     (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2149                         src = &t;
2150                         break;
2151                 }
2152         if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2153                 /*
2154                  * This is the case if the dst is our link-local address
2155                  * and the sender is also ourselves.
2156                  */
2157                 src = &t;
2158         }
2159
2160         if (src == 0) {
2161                 int e;
2162                 struct route_in6 ro;
2163
2164                 /*
2165                  * This case matches to multicasts, our anycast, or unicasts
2166                  * that we do not own.  Select a source address based on the
2167                  * source address of the erroneous packet.
2168                  */
2169                 bzero(&ro, sizeof(ro));
2170                 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2171                 if (ro.ro_rt)
2172                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2173                 if (src == NULL) {
2174                         nd6log((LOG_DEBUG,
2175                             "icmp6_reflect: source can't be determined: "
2176                             "dst=%s, error=%d\n",
2177                             ip6_sprintf(&sa6_src.sin6_addr), e));
2178                         goto bad;
2179                 }
2180         }
2181
2182         ip6->ip6_src = *src;
2183
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 = 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 #ifdef COMPAT_RFC1885
2207         ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2208 #else
2209         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2210 #endif
2211         if (outif)
2212                 icmp6_ifoutstat_inc(outif, type, code);
2213
2214         return;
2215
2216  bad:
2217         m_freem(m);
2218         return;
2219 }
2220
2221 void
2222 icmp6_fasttimo()
2223 {
2224
2225         mld6_fasttimeo();
2226 }
2227
2228 static const char *
2229 icmp6_redirect_diag(src6, dst6, tgt6)
2230         struct in6_addr *src6;
2231         struct in6_addr *dst6;
2232         struct in6_addr *tgt6;
2233 {
2234         static char buf[1024];
2235         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2236             ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2237         return buf;
2238 }
2239
2240 void
2241 icmp6_redirect_input(m, off)
2242         struct mbuf *m;
2243         int off;
2244 {
2245         struct ifnet *ifp = m->m_pkthdr.rcvif;
2246         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2247         struct nd_redirect *nd_rd;
2248         int icmp6len = ntohs(ip6->ip6_plen);
2249         char *lladdr = NULL;
2250         int lladdrlen = 0;
2251         u_char *redirhdr = NULL;
2252         int redirhdrlen = 0;
2253         struct rtentry *rt = NULL;
2254         int is_router;
2255         int is_onlink;
2256         struct in6_addr src6 = ip6->ip6_src;
2257         struct in6_addr redtgt6;
2258         struct in6_addr reddst6;
2259         union nd_opts ndopts;
2260
2261         if (!m || !ifp)
2262                 return;
2263
2264         /* XXX if we are router, we don't update route by icmp6 redirect */
2265         if (ip6_forwarding)
2266                 goto freeit;
2267         if (!icmp6_rediraccept)
2268                 goto freeit;
2269
2270 #ifndef PULLDOWN_TEST
2271         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2272         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2273 #else
2274         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2275         if (nd_rd == NULL) {
2276                 icmp6stat.icp6s_tooshort++;
2277                 return;
2278         }
2279 #endif
2280         redtgt6 = nd_rd->nd_rd_target;
2281         reddst6 = nd_rd->nd_rd_dst;
2282
2283         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2284                 redtgt6.s6_addr16[1] = htons(ifp->if_index);
2285         if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2286                 reddst6.s6_addr16[1] = htons(ifp->if_index);
2287
2288         /* validation */
2289         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2290                 nd6log((LOG_ERR,
2291                     "ICMP6 redirect sent from %s rejected; "
2292                     "must be from linklocal\n",
2293                     ip6_sprintf(&src6)));
2294                 goto bad;
2295         }
2296         if (ip6->ip6_hlim != 255) {
2297                 nd6log((LOG_ERR,
2298                     "ICMP6 redirect sent from %s rejected; "
2299                     "hlim=%d (must be 255)\n",
2300                     ip6_sprintf(&src6), ip6->ip6_hlim));
2301                 goto bad;
2302         }
2303     {
2304         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2305         struct sockaddr_in6 sin6;
2306         struct in6_addr *gw6;
2307
2308         bzero(&sin6, sizeof(sin6));
2309         sin6.sin6_family = AF_INET6;
2310         sin6.sin6_len = sizeof(struct sockaddr_in6);
2311         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2312         rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2313         if (rt) {
2314                 if (rt->rt_gateway == NULL ||
2315                     rt->rt_gateway->sa_family != AF_INET6) {
2316                         nd6log((LOG_ERR,
2317                             "ICMP6 redirect rejected; no route "
2318                             "with inet6 gateway found for redirect dst: %s\n",
2319                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2320                         RTFREE_LOCKED(rt);
2321                         goto bad;
2322                 }
2323
2324                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2325                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2326                         nd6log((LOG_ERR,
2327                             "ICMP6 redirect rejected; "
2328                             "not equal to gw-for-src=%s (must be same): "
2329                             "%s\n",
2330                             ip6_sprintf(gw6),
2331                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2332                         RTFREE_LOCKED(rt);
2333                         goto bad;
2334                 }
2335         } else {
2336                 nd6log((LOG_ERR,
2337                     "ICMP6 redirect rejected; "
2338                     "no route found for redirect dst: %s\n",
2339                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2340                 goto bad;
2341         }
2342         RTFREE_LOCKED(rt);
2343         rt = NULL;
2344     }
2345         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2346                 nd6log((LOG_ERR,
2347                     "ICMP6 redirect rejected; "
2348                     "redirect dst must be unicast: %s\n",
2349                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2350                 goto bad;
2351         }
2352
2353         is_router = is_onlink = 0;
2354         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2355                 is_router = 1;  /* router case */
2356         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2357                 is_onlink = 1;  /* on-link destination case */
2358         if (!is_router && !is_onlink) {
2359                 nd6log((LOG_ERR,
2360                     "ICMP6 redirect rejected; "
2361                     "neither router case nor onlink case: %s\n",
2362                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2363                 goto bad;
2364         }
2365         /* validation passed */
2366
2367         icmp6len -= sizeof(*nd_rd);
2368         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2369         if (nd6_options(&ndopts) < 0) {
2370                 nd6log((LOG_INFO, "icmp6_redirect_input: "
2371                     "invalid ND option, rejected: %s\n",
2372                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2373                 /* nd6_options have incremented stats */
2374                 goto freeit;
2375         }
2376
2377         if (ndopts.nd_opts_tgt_lladdr) {
2378                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2379                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2380         }
2381
2382         if (ndopts.nd_opts_rh) {
2383                 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2384                 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2385         }
2386
2387         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2388                 nd6log((LOG_INFO,
2389                     "icmp6_redirect_input: lladdrlen mismatch for %s "
2390                     "(if %d, icmp6 packet %d): %s\n",
2391                     ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2392                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2393                 goto bad;
2394         }
2395
2396         /* RFC 2461 8.3 */
2397         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2398             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2399
2400         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2401                 /* perform rtredirect */
2402                 struct sockaddr_in6 sdst;
2403                 struct sockaddr_in6 sgw;
2404                 struct sockaddr_in6 ssrc;
2405
2406                 bzero(&sdst, sizeof(sdst));
2407                 bzero(&sgw, sizeof(sgw));
2408                 bzero(&ssrc, sizeof(ssrc));
2409                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2410                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2411                         sizeof(struct sockaddr_in6);
2412                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2413                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2414                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2415                 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2416                     (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2417                     (struct sockaddr *)&ssrc);
2418         }
2419         /* finally update cached route in each socket via pfctlinput */
2420     {
2421         struct sockaddr_in6 sdst;
2422
2423         bzero(&sdst, sizeof(sdst));
2424         sdst.sin6_family = AF_INET6;
2425         sdst.sin6_len = sizeof(struct sockaddr_in6);
2426         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2427         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2428 #if defined(IPSEC) || defined(FAST_IPSEC)
2429         key_sa_routechange((struct sockaddr *)&sdst);
2430 #endif
2431     }
2432
2433  freeit:
2434         m_freem(m);
2435         return;
2436
2437  bad:
2438         icmp6stat.icp6s_badredirect++;
2439         m_freem(m);
2440 }
2441
2442 void
2443 icmp6_redirect_output(m0, rt)
2444         struct mbuf *m0;
2445         struct rtentry *rt;
2446 {
2447         struct ifnet *ifp;      /* my outgoing interface */
2448         struct in6_addr *ifp_ll6;
2449         struct in6_addr *router_ll6;
2450         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2451         struct mbuf *m = NULL;  /* newly allocated one */
2452         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2453         struct nd_redirect *nd_rd;
2454         size_t maxlen;
2455         u_char *p;
2456         struct ifnet *outif = NULL;
2457         struct sockaddr_in6 src_sa;
2458
2459         icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2460
2461         /* if we are not router, we don't send icmp6 redirect */
2462         if (!ip6_forwarding || ip6_accept_rtadv)
2463                 goto fail;
2464
2465         /* sanity check */
2466         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2467                 goto fail;
2468
2469         /*
2470          * Address check:
2471          *  the source address must identify a neighbor, and
2472          *  the destination address must not be a multicast address
2473          *  [RFC 2461, sec 8.2]
2474          */
2475         sip6 = mtod(m0, struct ip6_hdr *);
2476         bzero(&src_sa, sizeof(src_sa));
2477         src_sa.sin6_family = AF_INET6;
2478         src_sa.sin6_len = sizeof(src_sa);
2479         src_sa.sin6_addr = sip6->ip6_src;
2480         /* we don't currently use sin6_scope_id, but eventually use it */
2481         if (in6_addr2zoneid(ifp, &sip6->ip6_src, &src_sa.sin6_scope_id))
2482                 goto fail;
2483         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2484                 goto fail;
2485         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2486                 goto fail;      /* what should we do here? */
2487
2488         /* rate limit */
2489         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2490                 goto fail;
2491
2492         /*
2493          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2494          * we almost always ask for an mbuf cluster for simplicity.
2495          * (MHLEN < IPV6_MMTU is almost always true)
2496          */
2497 #if IPV6_MMTU >= MCLBYTES
2498 # error assumption failed about IPV6_MMTU and MCLBYTES
2499 #endif
2500         MGETHDR(m, M_DONTWAIT, MT_HEADER);
2501         if (m && IPV6_MMTU >= MHLEN)
2502                 MCLGET(m, M_DONTWAIT);
2503         if (!m)
2504                 goto fail;
2505         m->m_pkthdr.rcvif = NULL;
2506         m->m_len = 0;
2507         maxlen = M_TRAILINGSPACE(m);
2508         maxlen = min(IPV6_MMTU, maxlen);
2509         /* just for safety */
2510         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2511             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2512                 goto fail;
2513         }
2514
2515         {
2516                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2517                 struct in6_ifaddr *ia;
2518                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2519                                                  IN6_IFF_NOTREADY|
2520                                                  IN6_IFF_ANYCAST)) == NULL)
2521                         goto fail;
2522                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2523         }
2524
2525         /* get ip6 linklocal address for the router. */
2526         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2527                 struct sockaddr_in6 *sin6;
2528                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2529                 router_ll6 = &sin6->sin6_addr;
2530                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2531                         router_ll6 = (struct in6_addr *)NULL;
2532         } else
2533                 router_ll6 = (struct in6_addr *)NULL;
2534
2535         /* ip6 */
2536         ip6 = mtod(m, struct ip6_hdr *);
2537         ip6->ip6_flow = 0;
2538         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2539         ip6->ip6_vfc |= IPV6_VERSION;
2540         /* ip6->ip6_plen will be set later */
2541         ip6->ip6_nxt = IPPROTO_ICMPV6;
2542         ip6->ip6_hlim = 255;
2543         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2544         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2545         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2546
2547         /* ND Redirect */
2548         nd_rd = (struct nd_redirect *)(ip6 + 1);
2549         nd_rd->nd_rd_type = ND_REDIRECT;
2550         nd_rd->nd_rd_code = 0;
2551         nd_rd->nd_rd_reserved = 0;
2552         if (rt->rt_flags & RTF_GATEWAY) {
2553                 /*
2554                  * nd_rd->nd_rd_target must be a link-local address in
2555                  * better router cases.
2556                  */
2557                 if (!router_ll6)
2558                         goto fail;
2559                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2560                     sizeof(nd_rd->nd_rd_target));
2561                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2562                     sizeof(nd_rd->nd_rd_dst));
2563         } else {
2564                 /* make sure redtgt == reddst */
2565                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2566                     sizeof(nd_rd->nd_rd_target));
2567                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2568                     sizeof(nd_rd->nd_rd_dst));
2569         }
2570
2571         p = (u_char *)(nd_rd + 1);
2572
2573         if (!router_ll6)
2574                 goto nolladdropt;
2575
2576         {
2577                 /* target lladdr option */
2578                 struct rtentry *rt_router = NULL;
2579                 int len;
2580                 struct sockaddr_dl *sdl;
2581                 struct nd_opt_hdr *nd_opt;
2582                 char *lladdr;
2583
2584                 rt_router = nd6_lookup(router_ll6, 0, ifp);
2585                 if (!rt_router)
2586                         goto nolladdropt;
2587                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2588                 len = (len + 7) & ~7;   /* round by 8 */
2589                 /* safety check */
2590                 if (len + (p - (u_char *)ip6) > maxlen)
2591                         goto nolladdropt;
2592                 if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2593                     (rt_router->rt_flags & RTF_LLINFO) &&
2594                     (rt_router->rt_gateway->sa_family == AF_LINK) &&
2595                     (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2596                     sdl->sdl_alen) {
2597                         nd_opt = (struct nd_opt_hdr *)p;
2598                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2599                         nd_opt->nd_opt_len = len >> 3;
2600                         lladdr = (char *)(nd_opt + 1);
2601                         bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2602                         p += len;
2603                 }
2604         }
2605 nolladdropt:;
2606
2607         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2608
2609         /* just to be safe */
2610 #ifdef M_DECRYPTED      /*not openbsd*/
2611         if (m0->m_flags & M_DECRYPTED)
2612                 goto noredhdropt;
2613 #endif
2614         if (p - (u_char *)ip6 > maxlen)
2615                 goto noredhdropt;
2616
2617         {
2618                 /* redirected header option */
2619                 int len;
2620                 struct nd_opt_rd_hdr *nd_opt_rh;
2621
2622                 /*
2623                  * compute the maximum size for icmp6 redirect header option.
2624                  * XXX room for auth header?
2625                  */
2626                 len = maxlen - (p - (u_char *)ip6);
2627                 len &= ~7;
2628
2629                 /* This is just for simplicity. */
2630                 if (m0->m_pkthdr.len != m0->m_len) {
2631                         if (m0->m_next) {
2632                                 m_freem(m0->m_next);
2633                                 m0->m_next = NULL;
2634                         }
2635                         m0->m_pkthdr.len = m0->m_len;
2636                 }
2637
2638                 /*
2639                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2640                  * about padding/truncate rule for the original IP packet.
2641                  * From the discussion on IPv6imp in Feb 1999,
2642                  * the consensus was:
2643                  * - "attach as much as possible" is the goal
2644                  * - pad if not aligned (original size can be guessed by
2645                  *   original ip6 header)
2646                  * Following code adds the padding if it is simple enough,
2647                  * and truncates if not.
2648                  */
2649                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2650                         panic("assumption failed in %s:%d", __FILE__,
2651                             __LINE__);
2652
2653                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2654                         /* not enough room, truncate */
2655                         m0->m_pkthdr.len = m0->m_len = len -
2656                             sizeof(*nd_opt_rh);
2657                 } else {
2658                         /* enough room, pad or truncate */
2659                         size_t extra;
2660
2661                         extra = m0->m_pkthdr.len % 8;
2662                         if (extra) {
2663                                 /* pad if easy enough, truncate if not */
2664                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2665                                         /* pad */
2666                                         m0->m_len += (8 - extra);
2667                                         m0->m_pkthdr.len += (8 - extra);
2668                                 } else {
2669                                         /* truncate */
2670                                         m0->m_pkthdr.len -= extra;
2671                                         m0->m_len -= extra;
2672                                 }
2673                         }
2674                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2675                         m0->m_pkthdr.len = m0->m_len = len -
2676                             sizeof(*nd_opt_rh);
2677                 }
2678
2679                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2680                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2681                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2682                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2683                 p += sizeof(*nd_opt_rh);
2684                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2685
2686                 /* connect m0 to m */
2687                 m_tag_delete_chain(m0, NULL);
2688                 m0->m_flags &= ~M_PKTHDR;
2689                 m->m_next = m0;
2690                 m->m_pkthdr.len = m->m_len + m0->m_len;
2691                 m0 = NULL;
2692         }
2693 noredhdropt:;
2694         if (m0) {
2695                 m_freem(m0);
2696                 m0 = NULL;
2697         }
2698
2699         /* XXX: clear embedded link IDs in the inner header */
2700         in6_clearscope(&sip6->ip6_src);
2701         in6_clearscope(&sip6->ip6_dst);
2702         in6_clearscope(&nd_rd->nd_rd_target);
2703         in6_clearscope(&nd_rd->nd_rd_dst);
2704
2705         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2706
2707         nd_rd->nd_rd_cksum = 0;
2708         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2709             sizeof(*ip6), ntohs(ip6->ip6_plen));
2710
2711         /* send the packet to outside... */
2712         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2713         if (outif) {
2714                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2715                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2716         }
2717         icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2718
2719         return;
2720
2721 fail:
2722         if (m)
2723                 m_freem(m);
2724         if (m0)
2725                 m_freem(m0);
2726 }
2727
2728 /*
2729  * ICMPv6 socket option processing.
2730  */
2731 int
2732 icmp6_ctloutput(so, sopt)
2733         struct socket *so;
2734         struct sockopt *sopt;
2735 {
2736         int error = 0;
2737         int optlen;
2738         struct inpcb *inp = sotoinpcb(so);
2739         int level, op, optname;
2740
2741         if (sopt) {
2742                 level = sopt->sopt_level;
2743                 op = sopt->sopt_dir;
2744                 optname = sopt->sopt_name;
2745                 optlen = sopt->sopt_valsize;
2746         } else
2747                 level = op = optname = optlen = 0;
2748
2749         if (level != IPPROTO_ICMPV6) {
2750                 return EINVAL;
2751         }
2752
2753         switch (op) {
2754         case PRCO_SETOPT:
2755                 switch (optname) {
2756                 case ICMP6_FILTER:
2757                     {
2758                         struct icmp6_filter *p;
2759
2760                         if (optlen != sizeof(*p)) {
2761                                 error = EMSGSIZE;
2762                                 break;
2763                         }
2764                         if (inp->in6p_icmp6filt == NULL) {
2765                                 error = EINVAL;
2766                                 break;
2767                         }
2768                         error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2769                                 optlen);
2770                         break;
2771                     }
2772
2773                 default:
2774                         error = ENOPROTOOPT;
2775                         break;
2776                 }
2777                 break;
2778
2779         case PRCO_GETOPT:
2780                 switch (optname) {
2781                 case ICMP6_FILTER:
2782                     {
2783                         if (inp->in6p_icmp6filt == NULL) {
2784                                 error = EINVAL;
2785                                 break;
2786                         }
2787                         error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2788                                 sizeof(struct icmp6_filter));
2789                         break;
2790                     }
2791
2792                 default:
2793                         error = ENOPROTOOPT;
2794                         break;
2795                 }
2796                 break;
2797         }
2798
2799         return (error);
2800 }
2801
2802 /*
2803  * Perform rate limit check.
2804  * Returns 0 if it is okay to send the icmp6 packet.
2805  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2806  * limitation.
2807  *
2808  * XXX per-destination/type check necessary?
2809  */
2810 static int
2811 icmp6_ratelimit(dst, type, code)
2812         const struct in6_addr *dst;     /* not used at this moment */
2813         const int type;                 /* not used at this moment */
2814         const int code;                 /* not used at this moment */
2815 {
2816         int ret;
2817
2818         ret = 0;        /* okay to send */
2819
2820         /* PPS limit */
2821         if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2822             icmp6errppslim)) {
2823                 /* The packet is subject to rate limit */
2824                 ret++;
2825         }
2826
2827         return ret;
2828 }