sys/netinet6/icmp6.c

   1 /*-
   2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. Neither the name of the project nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
  30  */
  31
  32 /*-
  33  * Copyright (c) 1982, 1986, 1988, 1993
  34  *      The Regents of the University of California.  All rights reserved.
  35  *
  36  * Redistribution and use in source and binary forms, with or without
  37  * modification, are permitted provided that the following conditions
  38  * are met:
  39  * 1. Redistributions of source code must retain the above copyright
  40  *    notice, this list of conditions and the following disclaimer.
  41  * 2. Redistributions in binary form must reproduce the above copyright
  42  *    notice, this list of conditions and the following disclaimer in the
  43  *    documentation and/or other materials provided with the distribution.
  44  * 4. Neither the name of the University nor the names of its contributors
  45  *    may be used to endorse or promote products derived from this software
  46  *    without specific prior written permission.
  47  *
  48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  58  * SUCH DAMAGE.
  59  *
  60  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
  61  */
  62
  63 #include <sys/cdefs.h>
  64 __FBSDID("$FreeBSD$");
  65
  66 #include "opt_inet.h"
  67 #include "opt_inet6.h"
  68 #include "opt_ipsec.h"
  69
  70 #include <sys/param.h>
  71 #include <sys/domain.h>
  72 #include <sys/kernel.h>
  73 #include <sys/lock.h>
  74 #include <sys/malloc.h>
  75 #include <sys/mbuf.h>
  76 #include <sys/protosw.h>
  77 #include <sys/signalvar.h>
  78 #include <sys/socket.h>
  79 #include <sys/socketvar.h>
  80 #include <sys/sx.h>
  81 #include <sys/syslog.h>
  82 #include <sys/systm.h>
  83 #include <sys/time.h>
  84
  85 #include <net/if.h>
  86 #include <net/if_dl.h>
  87 #include <net/if_types.h>
  88 #include <net/route.h>
  89
  90 #include <netinet/in.h>
  91 #include <netinet/in_pcb.h>
  92 #include <netinet/in_var.h>
  93 #include <netinet/ip6.h>
  94 #include <netinet/icmp6.h>
  95 #include <netinet/tcp_var.h>
  96 #include <netinet6/in6_ifattach.h>
  97 #include <netinet6/in6_pcb.h>
  98 #include <netinet6/ip6protosw.h>
  99 #include <netinet6/ip6_var.h>
 100 #include <netinet6/scope6_var.h>
 101 #include <netinet6/mld6_var.h>
 102 #include <netinet6/nd6.h>
 103
 104 #ifdef IPSEC
 105 #include <netinet6/ipsec.h>
 106 #include <netkey/key.h>
 107 #endif
 108
 109 #ifdef FAST_IPSEC
 110 #include <netipsec/ipsec.h>
 111 #include <netipsec/key.h>
 112 #endif
 113
 114 #include <net/net_osdep.h>
 115
 116 extern struct domain inet6domain;
 117
 118 struct icmp6stat icmp6stat;
 119
 120 extern struct inpcbinfo ripcbinfo;
 121 extern struct inpcbhead ripcb;
 122 extern int icmp6errppslim;
 123 static int icmp6errpps_count = 0;
 124 static struct timeval icmp6errppslim_last;
 125 extern int icmp6_nodeinfo;
 126
 127 static void icmp6_errcount(struct icmp6errstat *, int, int);
 128 static int icmp6_rip6_input(struct mbuf **, int);
 129 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
 130 static const char *icmp6_redirect_diag __P((struct in6_addr *,
 131         struct in6_addr *, struct in6_addr *));
 132 static struct mbuf *ni6_input(struct mbuf *, int);
 133 static struct mbuf *ni6_nametodns(const char *, int, int);
 134 static int ni6_dnsmatch(const char *, int, const char *, int);
 135 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
 136                           struct ifnet **, struct in6_addr *));
 137 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
 138                                 struct ifnet *, int));
 139 static int icmp6_notify_error(struct mbuf **, int, int, int);
 140
 141
 142 void
 143 icmp6_init()
 144 {
 145         mld6_init();
 146 }
 147
 148 static void
 149 icmp6_errcount(stat, type, code)
 150         struct icmp6errstat *stat;
 151         int type, code;
 152 {
 153         switch (type) {
 154         case ICMP6_DST_UNREACH:
 155                 switch (code) {
 156                 case ICMP6_DST_UNREACH_NOROUTE:
 157                         stat->icp6errs_dst_unreach_noroute++;
 158                         return;
 159                 case ICMP6_DST_UNREACH_ADMIN:
 160                         stat->icp6errs_dst_unreach_admin++;
 161                         return;
 162                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 163                         stat->icp6errs_dst_unreach_beyondscope++;
 164                         return;
 165                 case ICMP6_DST_UNREACH_ADDR:
 166                         stat->icp6errs_dst_unreach_addr++;
 167                         return;
 168                 case ICMP6_DST_UNREACH_NOPORT:
 169                         stat->icp6errs_dst_unreach_noport++;
 170                         return;
 171                 }
 172                 break;
 173         case ICMP6_PACKET_TOO_BIG:
 174                 stat->icp6errs_packet_too_big++;
 175                 return;
 176         case ICMP6_TIME_EXCEEDED:
 177                 switch (code) {
 178                 case ICMP6_TIME_EXCEED_TRANSIT:
 179                         stat->icp6errs_time_exceed_transit++;
 180                         return;
 181                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 182                         stat->icp6errs_time_exceed_reassembly++;
 183                         return;
 184                 }
 185                 break;
 186         case ICMP6_PARAM_PROB:
 187                 switch (code) {
 188                 case ICMP6_PARAMPROB_HEADER:
 189                         stat->icp6errs_paramprob_header++;
 190                         return;
 191                 case ICMP6_PARAMPROB_NEXTHEADER:
 192                         stat->icp6errs_paramprob_nextheader++;
 193                         return;
 194                 case ICMP6_PARAMPROB_OPTION:
 195                         stat->icp6errs_paramprob_option++;
 196                         return;
 197                 }
 198                 break;
 199         case ND_REDIRECT:
 200                 stat->icp6errs_redirect++;
 201                 return;
 202         }
 203         stat->icp6errs_unknown++;
 204 }
 205
 206 /*
 207  * A wrapper function for icmp6_error() necessary when the erroneous packet
 208  * may not contain enough scope zone information.
 209  */
 210 void
 211 icmp6_error2(m, type, code, param, ifp)
 212         struct mbuf *m;
 213         int type, code, param;
 214         struct ifnet *ifp;
 215 {
 216         struct ip6_hdr *ip6;
 217
 218         if (ifp == NULL)
 219                 return;
 220
 221 #ifndef PULLDOWN_TEST
 222         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 223 #else
 224         if (m->m_len < sizeof(struct ip6_hdr)) {
 225                 m = m_pullup(m, sizeof(struct ip6_hdr));
 226                 if (m == NULL)
 227                         return;
 228         }
 229 #endif
 230
 231         ip6 = mtod(m, struct ip6_hdr *);
 232
 233         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
 234                 return;
 235         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
 236                 return;
 237
 238         icmp6_error(m, type, code, param);
 239 }
 240
 241 /*
 242  * Generate an error packet of type error in response to bad IP6 packet.
 243  */
 244 void
 245 icmp6_error(m, type, code, param)
 246         struct mbuf *m;
 247         int type, code, param;
 248 {
 249         struct ip6_hdr *oip6, *nip6;
 250         struct icmp6_hdr *icmp6;
 251         u_int preplen;
 252         int off;
 253         int nxt;
 254
 255         icmp6stat.icp6s_error++;
 256
 257         /* count per-type-code statistics */
 258         icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
 259
 260 #ifdef M_DECRYPTED      /*not openbsd*/
 261         if (m->m_flags & M_DECRYPTED) {
 262                 icmp6stat.icp6s_canterror++;
 263                 goto freeit;
 264         }
 265 #endif
 266
 267 #ifndef PULLDOWN_TEST
 268         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 269 #else
 270         if (m->m_len < sizeof(struct ip6_hdr)) {
 271                 m = m_pullup(m, sizeof(struct ip6_hdr));
 272                 if (m == NULL)
 273                         return;
 274         }
 275 #endif
 276         oip6 = mtod(m, struct ip6_hdr *);
 277
 278         /*
 279          * If the destination address of the erroneous packet is a multicast
 280          * address, or the packet was sent using link-layer multicast,
 281          * we should basically suppress sending an error (RFC 2463, Section
 282          * 2.4).
 283          * We have two exceptions (the item e.2 in that section):
 284          * - the Pakcet Too Big message can be sent for path MTU discovery.
 285          * - the Parameter Problem Message that can be allowed an icmp6 error
 286          *   in the option type field.  This check has been done in
 287          *   ip6_unknown_opt(), so we can just check the type and code.
 288          */
 289         if ((m->m_flags & (M_BCAST|M_MCAST) ||
 290              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
 291             (type != ICMP6_PACKET_TOO_BIG &&
 292              (type != ICMP6_PARAM_PROB ||
 293               code != ICMP6_PARAMPROB_OPTION)))
 294                 goto freeit;
 295
 296         /*
 297          * RFC 2463, 2.4 (e.5): source address check.
 298          * XXX: the case of anycast source?
 299          */
 300         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
 301             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
 302                 goto freeit;
 303
 304         /*
 305          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
 306          * don't do it.
 307          */
 308         nxt = -1;
 309         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
 310         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
 311                 struct icmp6_hdr *icp;
 312
 313 #ifndef PULLDOWN_TEST
 314                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
 315                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 316 #else
 317                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
 318                         sizeof(*icp));
 319                 if (icp == NULL) {
 320                         icmp6stat.icp6s_tooshort++;
 321                         return;
 322                 }
 323 #endif
 324                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
 325                     icp->icmp6_type == ND_REDIRECT) {
 326                         /*
 327                          * ICMPv6 error
 328                          * Special case: for redirect (which is
 329                          * informational) we must not send icmp6 error.
 330                          */
 331                         icmp6stat.icp6s_canterror++;
 332                         goto freeit;
 333                 } else {
 334                         /* ICMPv6 informational - send the error */
 335                 }
 336         } else {
 337                 /* non-ICMPv6 - send the error */
 338         }
 339
 340         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
 341
 342         /* Finally, do rate limitation check. */
 343         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
 344                 icmp6stat.icp6s_toofreq++;
 345                 goto freeit;
 346         }
 347
 348         /*
 349          * OK, ICMP6 can be generated.
 350          */
 351
 352         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
 353                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
 354
 355         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
 356         M_PREPEND(m, preplen, M_DONTWAIT);
 357         if (m && m->m_len < preplen)
 358                 m = m_pullup(m, preplen);
 359         if (m == NULL) {
 360                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
 361                 return;
 362         }
 363
 364         nip6 = mtod(m, struct ip6_hdr *);
 365         nip6->ip6_src  = oip6->ip6_src;
 366         nip6->ip6_dst  = oip6->ip6_dst;
 367
 368         in6_clearscope(&oip6->ip6_src);
 369         in6_clearscope(&oip6->ip6_dst);
 370
 371         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
 372         icmp6->icmp6_type = type;
 373         icmp6->icmp6_code = code;
 374         icmp6->icmp6_pptr = htonl((u_int32_t)param);
 375
 376         /*
 377          * icmp6_reflect() is designed to be in the input path.
 378          * icmp6_error() can be called from both input and output path,
 379          * and if we are in output path rcvif could contain bogus value.
 380          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
 381          * information in ip header (nip6).
 382          */
 383         m->m_pkthdr.rcvif = NULL;
 384
 385         icmp6stat.icp6s_outhist[type]++;
 386         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
 387
 388         return;
 389
 390   freeit:
 391         /*
 392          * If we can't tell whether or not we can generate ICMP6, free it.
 393          */
 394         m_freem(m);
 395 }
 396
 397 /*
 398  * Process a received ICMP6 message.
 399  */
 400 int
 401 icmp6_input(mp, offp, proto)
 402         struct mbuf **mp;
 403         int *offp, proto;
 404 {
 405         struct mbuf *m = *mp, *n;
 406         struct ip6_hdr *ip6, *nip6;
 407         struct icmp6_hdr *icmp6, *nicmp6;
 408         int off = *offp;
 409         int icmp6len = m->m_pkthdr.len - *offp;
 410         int code, sum, noff;
 411
 412 #ifndef PULLDOWN_TEST
 413         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
 414         /* m might change if M_LOOP.  So, call mtod after this */
 415 #endif
 416
 417         /*
 418          * Locate icmp6 structure in mbuf, and check
 419          * that not corrupted and of at least minimum length
 420          */
 421
 422         ip6 = mtod(m, struct ip6_hdr *);
 423         if (icmp6len < sizeof(struct icmp6_hdr)) {
 424                 icmp6stat.icp6s_tooshort++;
 425                 goto freeit;
 426         }
 427
 428         /*
 429          * calculate the checksum
 430          */
 431 #ifndef PULLDOWN_TEST
 432         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
 433 #else
 434         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
 435         if (icmp6 == NULL) {
 436                 icmp6stat.icp6s_tooshort++;
 437                 return IPPROTO_DONE;
 438         }
 439 #endif
 440         code = icmp6->icmp6_code;
 441
 442         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
 443                 nd6log((LOG_ERR,
 444                     "ICMP6 checksum error(%d|%x) %s\n",
 445                     icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
 446                 icmp6stat.icp6s_checksum++;
 447                 goto freeit;
 448         }
 449
 450         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
 451                 /*
 452                  * Deliver very specific ICMP6 type only.
 453                  * This is important to deliver TOOBIG.  Otherwise PMTUD
 454                  * will not work.
 455                  */
 456                 switch (icmp6->icmp6_type) {
 457                 case ICMP6_DST_UNREACH:
 458                 case ICMP6_PACKET_TOO_BIG:
 459                 case ICMP6_TIME_EXCEEDED:
 460                         break;
 461                 default:
 462                         goto freeit;
 463                 }
 464         }
 465
 466         icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
 467         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
 468         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
 469                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
 470
 471         switch (icmp6->icmp6_type) {
 472         case ICMP6_DST_UNREACH:
 473                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
 474                 switch (code) {
 475                 case ICMP6_DST_UNREACH_NOROUTE:
 476                         code = PRC_UNREACH_NET;
 477                         break;
 478                 case ICMP6_DST_UNREACH_ADMIN:
 479                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
 480                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
 481                         break;
 482                 case ICMP6_DST_UNREACH_ADDR:
 483                         code = PRC_HOSTDEAD;
 484                         break;
 485                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 486                         /* I mean "source address was incorrect." */
 487                         code = PRC_PARAMPROB;
 488                         break;
 489                 case ICMP6_DST_UNREACH_NOPORT:
 490                         code = PRC_UNREACH_PORT;
 491                         break;
 492                 default:
 493                         goto badcode;
 494                 }
 495                 goto deliver;
 496                 break;
 497
 498         case ICMP6_PACKET_TOO_BIG:
 499                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
 500
 501                 /* validation is made in icmp6_mtudisc_update */
 502
 503                 code = PRC_MSGSIZE;
 504
 505                 /*
 506                  * Updating the path MTU will be done after examining
 507                  * intermediate extension headers.
 508                  */
 509                 goto deliver;
 510                 break;
 511
 512         case ICMP6_TIME_EXCEEDED:
 513                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
 514                 switch (code) {
 515                 case ICMP6_TIME_EXCEED_TRANSIT:
 516                         code = PRC_TIMXCEED_INTRANS;
 517                         break;
 518                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 519                         code = PRC_TIMXCEED_REASS;
 520                         break;
 521                 default:
 522                         goto badcode;
 523                 }
 524                 goto deliver;
 525                 break;
 526
 527         case ICMP6_PARAM_PROB:
 528                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
 529                 switch (code) {
 530                 case ICMP6_PARAMPROB_NEXTHEADER:
 531                         code = PRC_UNREACH_PROTOCOL;
 532                         break;
 533                 case ICMP6_PARAMPROB_HEADER:
 534                 case ICMP6_PARAMPROB_OPTION:
 535                         code = PRC_PARAMPROB;
 536                         break;
 537                 default:
 538                         goto badcode;
 539                 }
 540                 goto deliver;
 541                 break;
 542
 543         case ICMP6_ECHO_REQUEST:
 544                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
 545                 if (code != 0)
 546                         goto badcode;
 547                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
 548                         /* Give up remote */
 549                         break;
 550                 }
 551                 if ((n->m_flags & M_EXT) != 0
 552                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
 553                         struct mbuf *n0 = n;
 554                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
 555                         int n0len;
 556
 557                         MGETHDR(n, M_DONTWAIT, n0->m_type);
 558                         n0len = n0->m_pkthdr.len;       /* save for use below */
 559                         if (n)
 560                                 M_MOVE_PKTHDR(n, n0);
 561                         if (n && maxlen >= MHLEN) {
 562                                 MCLGET(n, M_DONTWAIT);
 563                                 if ((n->m_flags & M_EXT) == 0) {
 564                                         m_free(n);
 565                                         n = NULL;
 566                                 }
 567                         }
 568                         if (n == NULL) {
 569                                 /* Give up remote */
 570                                 m_freem(n0);
 571                                 break;
 572                         }
 573                         /*
 574                          * Copy IPv6 and ICMPv6 only.
 575                          */
 576                         nip6 = mtod(n, struct ip6_hdr *);
 577                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 578                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 579                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 580                         noff = sizeof(struct ip6_hdr);
 581                         /* new mbuf contains only ipv6+icmpv6 headers */
 582                         n->m_len = noff + sizeof(struct icmp6_hdr);
 583                         /*
 584                          * Adjust mbuf.  ip6_plen will be adjusted in
 585                          * ip6_output().
 586                          */
 587                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 588                         /* recalculate complete packet size */
 589                         n->m_pkthdr.len = n0len + (noff - off);
 590                         n->m_next = n0;
 591                 } else {
 592                         nip6 = mtod(n, struct ip6_hdr *);
 593                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
 594                             sizeof(*nicmp6));
 595                         noff = off;
 596                 }
 597                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 598                 nicmp6->icmp6_code = 0;
 599                 if (n) {
 600                         icmp6stat.icp6s_reflect++;
 601                         icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
 602                         icmp6_reflect(n, noff);
 603                 }
 604                 break;
 605
 606         case ICMP6_ECHO_REPLY:
 607                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
 608                 if (code != 0)
 609                         goto badcode;
 610                 break;
 611
 612         case MLD_LISTENER_QUERY:
 613         case MLD_LISTENER_REPORT:
 614                 if (icmp6len < sizeof(struct mld_hdr))
 615                         goto badlen;
 616                 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
 617                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
 618                 else
 619                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
 620                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 621                         /* give up local */
 622                         mld6_input(m, off);
 623                         m = NULL;
 624                         goto freeit;
 625                 }
 626                 mld6_input(n, off);
 627                 /* m stays. */
 628                 break;
 629
 630         case MLD_LISTENER_DONE:
 631                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
 632                 if (icmp6len < sizeof(struct mld_hdr))  /* necessary? */
 633                         goto badlen;
 634                 break;          /* nothing to be done in kernel */
 635
 636         case MLD_MTRACE_RESP:
 637         case MLD_MTRACE:
 638                 /* XXX: these two are experimental.  not officially defined. */
 639                 /* XXX: per-interface statistics? */
 640                 break;          /* just pass it to applications */
 641
 642         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 643             {
 644                 enum { WRU, FQDN } mode;
 645
 646                 if (!icmp6_nodeinfo)
 647                         break;
 648
 649                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 650                         mode = WRU;
 651                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 652                         mode = FQDN;
 653                 else
 654                         goto badlen;
 655
 656 #define hostnamelen     strlen(hostname)
 657                 if (mode == FQDN) {
 658 #ifndef PULLDOWN_TEST
 659                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
 660                             IPPROTO_DONE);
 661 #endif
 662                         n = m_copy(m, 0, M_COPYALL);
 663                         if (n)
 664                                 n = ni6_input(n, off);
 665                         /* XXX meaningless if n == NULL */
 666                         noff = sizeof(struct ip6_hdr);
 667                 } else {
 668                         u_char *p;
 669                         int maxlen, maxhlen;
 670
 671                         /*
 672                          * XXX: this combination of flags is pointless,
 673                          * but should we keep this for compatibility?
 674                          */
 675                         if ((icmp6_nodeinfo & 5) != 5)
 676                                 break;
 677
 678                         if (code != 0)
 679                                 goto badcode;
 680                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
 681                         if (maxlen >= MCLBYTES) {
 682                                 /* Give up remote */
 683                                 break;
 684                         }
 685                         MGETHDR(n, M_DONTWAIT, m->m_type);
 686                         if (n && maxlen > MHLEN) {
 687                                 MCLGET(n, M_DONTWAIT);
 688                                 if ((n->m_flags & M_EXT) == 0) {
 689                                         m_free(n);
 690                                         n = NULL;
 691                                 }
 692                         }
 693                         if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
 694                                 /*
 695                                  * Previous code did a blind M_COPY_PKTHDR
 696                                  * and said "just for rcvif".  If true, then
 697                                  * we could tolerate the dup failing (due to
 698                                  * the deep copy of the tag chain).  For now
 699                                  * be conservative and just fail.
 700                                  */
 701                                 m_free(n);
 702                                 n = NULL;
 703                         }
 704                         if (n == NULL) {
 705                                 /* Give up remote */
 706                                 break;
 707                         }
 708                         n->m_pkthdr.rcvif = NULL;
 709                         n->m_len = 0;
 710                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
 711                         if (maxhlen > hostnamelen)
 712                                 maxhlen = hostnamelen;
 713                         /*
 714                          * Copy IPv6 and ICMPv6 only.
 715                          */
 716                         nip6 = mtod(n, struct ip6_hdr *);
 717                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 718                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 719                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 720                         p = (u_char *)(nicmp6 + 1);
 721                         bzero(p, 4);
 722                         bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
 723                         noff = sizeof(struct ip6_hdr);
 724                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 725                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 726                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 727                         nicmp6->icmp6_code = 0;
 728                 }
 729 #undef hostnamelen
 730                 if (n) {
 731                         icmp6stat.icp6s_reflect++;
 732                         icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
 733                         icmp6_reflect(n, noff);
 734                 }
 735                 break;
 736             }
 737
 738         case ICMP6_WRUREPLY:
 739                 if (code != 0)
 740                         goto badcode;
 741                 break;
 742
 743         case ND_ROUTER_SOLICIT:
 744                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
 745                 if (code != 0)
 746                         goto badcode;
 747                 if (icmp6len < sizeof(struct nd_router_solicit))
 748                         goto badlen;
 749                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 750                         /* give up local */
 751                         nd6_rs_input(m, off, icmp6len);
 752                         m = NULL;
 753                         goto freeit;
 754                 }
 755                 nd6_rs_input(n, off, icmp6len);
 756                 /* m stays. */
 757                 break;
 758
 759         case ND_ROUTER_ADVERT:
 760                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
 761                 if (code != 0)
 762                         goto badcode;
 763                 if (icmp6len < sizeof(struct nd_router_advert))
 764                         goto badlen;
 765                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 766                         /* give up local */
 767                         nd6_ra_input(m, off, icmp6len);
 768                         m = NULL;
 769                         goto freeit;
 770                 }
 771                 nd6_ra_input(n, off, icmp6len);
 772                 /* m stays. */
 773                 break;
 774
 775         case ND_NEIGHBOR_SOLICIT:
 776                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
 777                 if (code != 0)
 778                         goto badcode;
 779                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 780                         goto badlen;
 781                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 782                         /* give up local */
 783                         nd6_ns_input(m, off, icmp6len);
 784                         m = NULL;
 785                         goto freeit;
 786                 }
 787                 nd6_ns_input(n, off, icmp6len);
 788                 /* m stays. */
 789                 break;
 790
 791         case ND_NEIGHBOR_ADVERT:
 792                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
 793                 if (code != 0)
 794                         goto badcode;
 795                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 796                         goto badlen;
 797                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 798                         /* give up local */
 799                         nd6_na_input(m, off, icmp6len);
 800                         m = NULL;
 801                         goto freeit;
 802                 }
 803                 nd6_na_input(n, off, icmp6len);
 804                 /* m stays. */
 805                 break;
 806
 807         case ND_REDIRECT:
 808                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
 809                 if (code != 0)
 810                         goto badcode;
 811                 if (icmp6len < sizeof(struct nd_redirect))
 812                         goto badlen;
 813                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 814                         /* give up local */
 815                         icmp6_redirect_input(m, off);
 816                         m = NULL;
 817                         goto freeit;
 818                 }
 819                 icmp6_redirect_input(n, off);
 820                 /* m stays. */
 821                 break;
 822
 823         case ICMP6_ROUTER_RENUMBERING:
 824                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 825                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 826                         goto badcode;
 827                 if (icmp6len < sizeof(struct icmp6_router_renum))
 828                         goto badlen;
 829                 break;
 830
 831         default:
 832                 nd6log((LOG_DEBUG,
 833                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 834                     icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
 835                     ip6_sprintf(&ip6->ip6_dst),
 836                     m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
 837                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 838                         /* ICMPv6 error: MUST deliver it by spec... */
 839                         code = PRC_NCMDS;
 840                         /* deliver */
 841                 } else {
 842                         /* ICMPv6 informational: MUST not deliver */
 843                         break;
 844                 }
 845         deliver:
 846                 if (icmp6_notify_error(&m, off, icmp6len, code)) {
 847                         /* In this case, m should've been freed. */
 848                         return (IPPROTO_DONE);
 849                 }
 850                 break;
 851
 852         badcode:
 853                 icmp6stat.icp6s_badcode++;
 854                 break;
 855
 856         badlen:
 857                 icmp6stat.icp6s_badlen++;
 858                 break;
 859         }
 860
 861         /* deliver the packet to appropriate sockets */
 862         icmp6_rip6_input(&m, *offp);
 863
 864         return IPPROTO_DONE;
 865
 866  freeit:
 867         m_freem(m);
 868         return IPPROTO_DONE;
 869 }
 870
 871 static int
 872 icmp6_notify_error(mp, off, icmp6len, code)
 873         struct mbuf **mp;
 874         int off, icmp6len, code;
 875 {
 876         struct mbuf *m = *mp;
 877         struct icmp6_hdr *icmp6;
 878         struct ip6_hdr *eip6;
 879         u_int32_t notifymtu;
 880         struct sockaddr_in6 icmp6src, icmp6dst;
 881
 882         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 883                 icmp6stat.icp6s_tooshort++;
 884                 goto freeit;
 885         }
 886 #ifndef PULLDOWN_TEST
 887         IP6_EXTHDR_CHECK(m, off,
 888             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
 889         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 890 #else
 891         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
 892             sizeof(*icmp6) + sizeof(struct ip6_hdr));
 893         if (icmp6 == NULL) {
 894                 icmp6stat.icp6s_tooshort++;
 895                 return (-1);
 896         }
 897 #endif
 898         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 899
 900         /* Detect the upper level protocol */
 901         {
 902                 void (*ctlfunc)(int, struct sockaddr *, void *);
 903                 u_int8_t nxt = eip6->ip6_nxt;
 904                 int eoff = off + sizeof(struct icmp6_hdr) +
 905                     sizeof(struct ip6_hdr);
 906                 struct ip6ctlparam ip6cp;
 907                 struct in6_addr *finaldst = NULL;
 908                 int icmp6type = icmp6->icmp6_type;
 909                 struct ip6_frag *fh;
 910                 struct ip6_rthdr *rth;
 911                 struct ip6_rthdr0 *rth0;
 912                 int rthlen;
 913
 914                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 915                         struct ip6_ext *eh;
 916
 917                         switch (nxt) {
 918                         case IPPROTO_HOPOPTS:
 919                         case IPPROTO_DSTOPTS:
 920                         case IPPROTO_AH:
 921 #ifndef PULLDOWN_TEST
 922                                 IP6_EXTHDR_CHECK(m, 0,
 923                                     eoff + sizeof(struct ip6_ext), -1);
 924                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
 925 #else
 926                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
 927                                     eoff, sizeof(*eh));
 928                                 if (eh == NULL) {
 929                                         icmp6stat.icp6s_tooshort++;
 930                                         return (-1);
 931                                 }
 932 #endif
 933
 934                                 if (nxt == IPPROTO_AH)
 935                                         eoff += (eh->ip6e_len + 2) << 2;
 936                                 else
 937                                         eoff += (eh->ip6e_len + 1) << 3;
 938                                 nxt = eh->ip6e_nxt;
 939                                 break;
 940                         case IPPROTO_ROUTING:
 941                                 /*
 942                                  * When the erroneous packet contains a
 943                                  * routing header, we should examine the
 944                                  * header to determine the final destination.
 945                                  * Otherwise, we can't properly update
 946                                  * information that depends on the final
 947                                  * destination (e.g. path MTU).
 948                                  */
 949 #ifndef PULLDOWN_TEST
 950                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
 951                                 rth = (struct ip6_rthdr *)
 952                                     (mtod(m, caddr_t) + eoff);
 953 #else
 954                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
 955                                     eoff, sizeof(*rth));
 956                                 if (rth == NULL) {
 957                                         icmp6stat.icp6s_tooshort++;
 958                                         return (-1);
 959                                 }
 960 #endif
 961                                 rthlen = (rth->ip6r_len + 1) << 3;
 962                                 /*
 963                                  * XXX: currently there is no
 964                                  * officially defined type other
 965                                  * than type-0.
 966                                  * Note that if the segment left field
 967                                  * is 0, all intermediate hops must
 968                                  * have been passed.
 969                                  */
 970                                 if (rth->ip6r_segleft &&
 971                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
 972                                         int hops;
 973
 974 #ifndef PULLDOWN_TEST
 975                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
 976                                         rth0 = (struct ip6_rthdr0 *)
 977                                             (mtod(m, caddr_t) + eoff);
 978 #else
 979                                         IP6_EXTHDR_GET(rth0,
 980                                             struct ip6_rthdr0 *, m,
 981                                             eoff, rthlen);
 982                                         if (rth0 == NULL) {
 983                                                 icmp6stat.icp6s_tooshort++;
 984                                                 return (-1);
 985                                         }
 986 #endif
 987                                         /* just ignore a bogus header */
 988                                         if ((rth0->ip6r0_len % 2) == 0 &&
 989                                             (hops = rth0->ip6r0_len/2))
 990                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
 991                                 }
 992                                 eoff += rthlen;
 993                                 nxt = rth->ip6r_nxt;
 994                                 break;
 995                         case IPPROTO_FRAGMENT:
 996 #ifndef PULLDOWN_TEST
 997                                 IP6_EXTHDR_CHECK(m, 0, eoff +
 998                                     sizeof(struct ip6_frag), -1);
 999                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1000                                     eoff);
1001 #else
1002                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1003                                     eoff, sizeof(*fh));
1004                                 if (fh == NULL) {
1005                                         icmp6stat.icp6s_tooshort++;
1006                                         return (-1);
1007                                 }
1008 #endif
1009                                 /*
1010                                  * Data after a fragment header is meaningless
1011                                  * unless it is the first fragment, but
1012                                  * we'll go to the notify label for path MTU
1013                                  * discovery.
1014                                  */
1015                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1016                                         goto notify;
1017
1018                                 eoff += sizeof(struct ip6_frag);
1019                                 nxt = fh->ip6f_nxt;
1020                                 break;
1021                         default:
1022                                 /*
1023                                  * This case includes ESP and the No Next
1024                                  * Header.  In such cases going to the notify
1025                                  * label does not have any meaning
1026                                  * (i.e. ctlfunc will be NULL), but we go
1027                                  * anyway since we might have to update
1028                                  * path MTU information.
1029                                  */
1030                                 goto notify;
1031                         }
1032                 }
1033           notify:
1034 #ifndef PULLDOWN_TEST
1035                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1036 #else
1037                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1038                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1039                 if (icmp6 == NULL) {
1040                         icmp6stat.icp6s_tooshort++;
1041                         return (-1);
1042                 }
1043 #endif
1044
1045                 /*
1046                  * retrieve parameters from the inner IPv6 header, and convert
1047                  * them into sockaddr structures.
1048                  * XXX: there is no guarantee that the source or destination
1049                  * addresses of the inner packet are in the same scope as
1050                  * the addresses of the icmp packet.  But there is no other
1051                  * way to determine the zone.
1052                  */
1053                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1054
1055                 bzero(&icmp6dst, sizeof(icmp6dst));
1056                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1057                 icmp6dst.sin6_family = AF_INET6;
1058                 if (finaldst == NULL)
1059                         icmp6dst.sin6_addr = eip6->ip6_dst;
1060                 else
1061                         icmp6dst.sin6_addr = *finaldst;
1062                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1063                         goto freeit;
1064                 bzero(&icmp6src, sizeof(icmp6src));
1065                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1066                 icmp6src.sin6_family = AF_INET6;
1067                 icmp6src.sin6_addr = eip6->ip6_src;
1068                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1069                         goto freeit;
1070                 icmp6src.sin6_flowinfo =
1071                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1072
1073                 if (finaldst == NULL)
1074                         finaldst = &eip6->ip6_dst;
1075                 ip6cp.ip6c_m = m;
1076                 ip6cp.ip6c_icmp6 = icmp6;
1077                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1078                 ip6cp.ip6c_off = eoff;
1079                 ip6cp.ip6c_finaldst = finaldst;
1080                 ip6cp.ip6c_src = &icmp6src;
1081                 ip6cp.ip6c_nxt = nxt;
1082
1083                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1084                         notifymtu = ntohl(icmp6->icmp6_mtu);
1085                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1086                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1087                 }
1088
1089                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1090                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1091                 if (ctlfunc) {
1092                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1093                             &ip6cp);
1094                 }
1095         }
1096         *mp = m;
1097         return (0);
1098
1099   freeit:
1100         m_freem(m);
1101         return (-1);
1102 }
1103
1104 void
1105 icmp6_mtudisc_update(ip6cp, validated)
1106         struct ip6ctlparam *ip6cp;
1107         int validated;
1108 {
1109         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1110         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1111         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1112         u_int mtu = ntohl(icmp6->icmp6_mtu);
1113         struct in_conninfo inc;
1114
1115 #if 0
1116         /*
1117          * RFC2460 section 5, last paragraph.
1118          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1119          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1120          * due to packet translator in the middle.
1121          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1122          * special handling.
1123          */
1124         if (mtu < IPV6_MMTU)
1125                 return;
1126 #endif
1127
1128         /*
1129          * we reject ICMPv6 too big with abnormally small value.
1130          * XXX what is the good definition of "abnormally small"?
1131          */
1132         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1133                 return;
1134
1135         if (!validated)
1136                 return;
1137
1138         /*
1139          * In case the suggested mtu is less than IPV6_MMTU, we
1140          * only need to remember that it was for above mentioned
1141          * "alwaysfrag" case.
1142          * Try to be as close to the spec as possible.
1143          */
1144         if (mtu < IPV6_MMTU)
1145                 mtu = IPV6_MMTU - 8;
1146
1147         bzero(&inc, sizeof(inc));
1148         inc.inc_flags = 1; /* IPv6 */
1149         inc.inc6_faddr = *dst;
1150         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1151                 return;
1152
1153         if (mtu < tcp_maxmtu6(&inc)) {
1154                 tcp_hc_updatemtu(&inc, mtu);
1155                 icmp6stat.icp6s_pmtuchg++;
1156         }
1157 }
1158
1159 /*
1160  * Process a Node Information Query packet, based on
1161  * draft-ietf-ipngwg-icmp-name-lookups-07.
1162  *
1163  * Spec incompatibilities:
1164  * - IPv6 Subject address handling
1165  * - IPv4 Subject address handling support missing
1166  * - Proxy reply (answer even if it's not for me)
1167  * - joins NI group address at in6_ifattach() time only, does not cope
1168  *   with hostname changes by sethostname(3)
1169  */
1170 #define hostnamelen     strlen(hostname)
1171 static struct mbuf *
1172 ni6_input(m, off)
1173         struct mbuf *m;
1174         int off;
1175 {
1176         struct icmp6_nodeinfo *ni6, *nni6;
1177         struct mbuf *n = NULL;
1178         u_int16_t qtype;
1179         int subjlen;
1180         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1181         struct ni_reply_fqdn *fqdn;
1182         int addrs;              /* for NI_QTYPE_NODEADDR */
1183         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1184         struct in6_addr in6_subj; /* subject address */
1185         struct ip6_hdr *ip6;
1186         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1187         char *subj = NULL;
1188         struct in6_ifaddr *ia6 = NULL;
1189
1190         ip6 = mtod(m, struct ip6_hdr *);
1191 #ifndef PULLDOWN_TEST
1192         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1193 #else
1194         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1195         if (ni6 == NULL) {
1196                 /* m is already reclaimed */
1197                 return (NULL);
1198         }
1199 #endif
1200
1201         /*
1202          * Validate IPv6 source address.
1203          * The default configuration MUST be to refuse answering queries from
1204          * global-scope addresses according to RFC4602.
1205          * Notes:
1206          *  - it's not very clear what "refuse" means; this implementation
1207          *    simply drops it.
1208          *  - it's not very easy to identify global-scope (unicast) addresses
1209          *    since there are many prefixes for them.  It should be safer
1210          *    and in practice sufficient to check "all" but loopback and
1211          *    link-local (note that site-local unicast was deprecated and
1212          *    ULA is defined as global scope-wise)
1213          */
1214         if ((icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1215             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1216             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1217                 goto bad;
1218
1219         /*
1220          * Validate IPv6 destination address.
1221          *
1222          * The Responder must discard the Query without further processing
1223          * unless it is one of the Responder's unicast or anycast addresses, or
1224          * a link-local scope multicast address which the Responder has joined.
1225          * [RFC4602, Section 5.]
1226          */
1227         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1228                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1229                         goto bad;
1230                 /* else it's a link-local multicast, fine */
1231         } else {                /* unicast or anycast */
1232                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1233                         goto bad; /* XXX impossible */
1234
1235                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1236                     !(icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1237                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1238                                 "a temporary address in %s:%d",
1239                                __FILE__, __LINE__));
1240                         goto bad;
1241                 }
1242         }
1243
1244         /* validate query Subject field. */
1245         qtype = ntohs(ni6->ni_qtype);
1246         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1247         switch (qtype) {
1248         case NI_QTYPE_NOOP:
1249         case NI_QTYPE_SUPTYPES:
1250                 /* 07 draft */
1251                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1252                         break;
1253                 /* FALLTHROUGH */
1254         case NI_QTYPE_FQDN:
1255         case NI_QTYPE_NODEADDR:
1256         case NI_QTYPE_IPV4ADDR:
1257                 switch (ni6->ni_code) {
1258                 case ICMP6_NI_SUBJ_IPV6:
1259 #if ICMP6_NI_SUBJ_IPV6 != 0
1260                 case 0:
1261 #endif
1262                         /*
1263                          * backward compatibility - try to accept 03 draft
1264                          * format, where no Subject is present.
1265                          */
1266                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1267                             subjlen == 0) {
1268                                 oldfqdn++;
1269                                 break;
1270                         }
1271 #if ICMP6_NI_SUBJ_IPV6 != 0
1272                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1273                                 goto bad;
1274 #endif
1275
1276                         if (subjlen != sizeof(struct in6_addr))
1277                                 goto bad;
1278
1279                         /*
1280                          * Validate Subject address.
1281                          *
1282                          * Not sure what exactly "address belongs to the node"
1283                          * means in the spec, is it just unicast, or what?
1284                          *
1285                          * At this moment we consider Subject address as
1286                          * "belong to the node" if the Subject address equals
1287                          * to the IPv6 destination address; validation for
1288                          * IPv6 destination address should have done enough
1289                          * check for us.
1290                          *
1291                          * We do not do proxy at this moment.
1292                          */
1293                         /* m_pulldown instead of copy? */
1294                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1295                             subjlen, (caddr_t)&in6_subj);
1296                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1297                                 goto bad;
1298
1299                         subj = (char *)&in6_subj;
1300                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1301                                 break;
1302
1303                         /*
1304                          * XXX if we are to allow other cases, we should really
1305                          * be careful about scope here.
1306                          * basically, we should disallow queries toward IPv6
1307                          * destination X with subject Y,
1308                          * if scope(X) > scope(Y).
1309                          * if we allow scope(X) > scope(Y), it will result in
1310                          * information leakage across scope boundary.
1311                          */
1312                         goto bad;
1313
1314                 case ICMP6_NI_SUBJ_FQDN:
1315                         /*
1316                          * Validate Subject name with gethostname(3).
1317                          *
1318                          * The behavior may need some debate, since:
1319                          * - we are not sure if the node has FQDN as
1320                          *   hostname (returned by gethostname(3)).
1321                          * - the code does wildcard match for truncated names.
1322                          *   however, we are not sure if we want to perform
1323                          *   wildcard match, if gethostname(3) side has
1324                          *   truncated hostname.
1325                          */
1326                         n = ni6_nametodns(hostname, hostnamelen, 0);
1327                         if (!n || n->m_next || n->m_len == 0)
1328                                 goto bad;
1329                         IP6_EXTHDR_GET(subj, char *, m,
1330                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1331                         if (subj == NULL)
1332                                 goto bad;
1333                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1334                             n->m_len)) {
1335                                 goto bad;
1336                         }
1337                         m_freem(n);
1338                         n = NULL;
1339                         break;
1340
1341                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1342                 default:
1343                         goto bad;
1344                 }
1345                 break;
1346         }
1347
1348         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1349         switch (qtype) {
1350         case NI_QTYPE_FQDN:
1351                 if ((icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1352                         goto bad;
1353                 break;
1354         case NI_QTYPE_NODEADDR:
1355         case NI_QTYPE_IPV4ADDR:
1356                 if ((icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1357                         goto bad;
1358                 break;
1359         }
1360
1361         /* guess reply length */
1362         switch (qtype) {
1363         case NI_QTYPE_NOOP:
1364                 break;          /* no reply data */
1365         case NI_QTYPE_SUPTYPES:
1366                 replylen += sizeof(u_int32_t);
1367                 break;
1368         case NI_QTYPE_FQDN:
1369                 /* XXX will append an mbuf */
1370                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1371                 break;
1372         case NI_QTYPE_NODEADDR:
1373                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1374                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1375                     sizeof(u_int32_t))) > MCLBYTES)
1376                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1377                 break;
1378         case NI_QTYPE_IPV4ADDR:
1379                 /* unsupported - should respond with unknown Qtype? */
1380                 break;
1381         default:
1382                 /*
1383                  * XXX: We must return a reply with the ICMP6 code
1384                  * `unknown Qtype' in this case.  However we regard the case
1385                  * as an FQDN query for backward compatibility.
1386                  * Older versions set a random value to this field,
1387                  * so it rarely varies in the defined qtypes.
1388                  * But the mechanism is not reliable...
1389                  * maybe we should obsolete older versions.
1390                  */
1391                 qtype = NI_QTYPE_FQDN;
1392                 /* XXX will append an mbuf */
1393                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1394                 oldfqdn++;
1395                 break;
1396         }
1397
1398         /* allocate an mbuf to reply. */
1399         MGETHDR(n, M_DONTWAIT, m->m_type);
1400         if (n == NULL) {
1401                 m_freem(m);
1402                 return (NULL);
1403         }
1404         M_MOVE_PKTHDR(n, m); /* just for recvif */
1405         if (replylen > MHLEN) {
1406                 if (replylen > MCLBYTES) {
1407                         /*
1408                          * XXX: should we try to allocate more? But MCLBYTES
1409                          * is probably much larger than IPV6_MMTU...
1410                          */
1411                         goto bad;
1412                 }
1413                 MCLGET(n, M_DONTWAIT);
1414                 if ((n->m_flags & M_EXT) == 0) {
1415                         goto bad;
1416                 }
1417         }
1418         n->m_pkthdr.len = n->m_len = replylen;
1419
1420         /* copy mbuf header and IPv6 + Node Information base headers */
1421         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1422         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1423         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1424
1425         /* qtype dependent procedure */
1426         switch (qtype) {
1427         case NI_QTYPE_NOOP:
1428                 nni6->ni_code = ICMP6_NI_SUCCESS;
1429                 nni6->ni_flags = 0;
1430                 break;
1431         case NI_QTYPE_SUPTYPES:
1432         {
1433                 u_int32_t v;
1434                 nni6->ni_code = ICMP6_NI_SUCCESS;
1435                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1436                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1437                 v = (u_int32_t)htonl(0x0000000f);
1438                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1439                 break;
1440         }
1441         case NI_QTYPE_FQDN:
1442                 nni6->ni_code = ICMP6_NI_SUCCESS;
1443                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1444                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1445                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1446                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1447                 /*
1448                  * XXX do we really have FQDN in variable "hostname"?
1449                  */
1450                 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1451                 if (n->m_next == NULL)
1452                         goto bad;
1453                 /* XXX we assume that n->m_next is not a chain */
1454                 if (n->m_next->m_next != NULL)
1455                         goto bad;
1456                 n->m_pkthdr.len += n->m_next->m_len;
1457                 break;
1458         case NI_QTYPE_NODEADDR:
1459         {
1460                 int lenlim, copied;
1461
1462                 nni6->ni_code = ICMP6_NI_SUCCESS;
1463                 n->m_pkthdr.len = n->m_len =
1464                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1465                 lenlim = M_TRAILINGSPACE(n);
1466                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1467                 /* XXX: reset mbuf length */
1468                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1469                     sizeof(struct icmp6_nodeinfo) + copied;
1470                 break;
1471         }
1472         default:
1473                 break;          /* XXX impossible! */
1474         }
1475
1476         nni6->ni_type = ICMP6_NI_REPLY;
1477         m_freem(m);
1478         return (n);
1479
1480   bad:
1481         m_freem(m);
1482         if (n)
1483                 m_freem(n);
1484         return (NULL);
1485 }
1486 #undef hostnamelen
1487
1488 /*
1489  * make a mbuf with DNS-encoded string.  no compression support.
1490  *
1491  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1492  * treated as truncated name (two \0 at the end).  this is a wild guess.
1493  */
1494 static struct mbuf *
1495 ni6_nametodns(name, namelen, old)
1496         const char *name;
1497         int namelen;
1498         int old;        /* return pascal string if non-zero */
1499 {
1500         struct mbuf *m;
1501         char *cp, *ep;
1502         const char *p, *q;
1503         int i, len, nterm;
1504
1505         if (old)
1506                 len = namelen + 1;
1507         else
1508                 len = MCLBYTES;
1509
1510         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1511         MGET(m, M_DONTWAIT, MT_DATA);
1512         if (m && len > MLEN) {
1513                 MCLGET(m, M_DONTWAIT);
1514                 if ((m->m_flags & M_EXT) == 0)
1515                         goto fail;
1516         }
1517         if (!m)
1518                 goto fail;
1519         m->m_next = NULL;
1520
1521         if (old) {
1522                 m->m_len = len;
1523                 *mtod(m, char *) = namelen;
1524                 bcopy(name, mtod(m, char *) + 1, namelen);
1525                 return m;
1526         } else {
1527                 m->m_len = 0;
1528                 cp = mtod(m, char *);
1529                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1530
1531                 /* if not certain about my name, return empty buffer */
1532                 if (namelen == 0)
1533                         return m;
1534
1535                 /*
1536                  * guess if it looks like shortened hostname, or FQDN.
1537                  * shortened hostname needs two trailing "\0".
1538                  */
1539                 i = 0;
1540                 for (p = name; p < name + namelen; p++) {
1541                         if (*p && *p == '.')
1542                                 i++;
1543                 }
1544                 if (i < 2)
1545                         nterm = 2;
1546                 else
1547                         nterm = 1;
1548
1549                 p = name;
1550                 while (cp < ep && p < name + namelen) {
1551                         i = 0;
1552                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1553                                 i++;
1554                         /* result does not fit into mbuf */
1555                         if (cp + i + 1 >= ep)
1556                                 goto fail;
1557                         /*
1558                          * DNS label length restriction, RFC1035 page 8.
1559                          * "i == 0" case is included here to avoid returning
1560                          * 0-length label on "foo..bar".
1561                          */
1562                         if (i <= 0 || i >= 64)
1563                                 goto fail;
1564                         *cp++ = i;
1565                         bcopy(p, cp, i);
1566                         cp += i;
1567                         p = q;
1568                         if (p < name + namelen && *p == '.')
1569                                 p++;
1570                 }
1571                 /* termination */
1572                 if (cp + nterm >= ep)
1573                         goto fail;
1574                 while (nterm-- > 0)
1575                         *cp++ = '\0';
1576                 m->m_len = cp - mtod(m, char *);
1577                 return m;
1578         }
1579
1580         panic("should not reach here");
1581         /* NOTREACHED */
1582
1583  fail:
1584         if (m)
1585                 m_freem(m);
1586         return NULL;
1587 }
1588
1589 /*
1590  * check if two DNS-encoded string matches.  takes care of truncated
1591  * form (with \0\0 at the end).  no compression support.
1592  * XXX upper/lowercase match (see RFC2065)
1593  */
1594 static int
1595 ni6_dnsmatch(a, alen, b, blen)
1596         const char *a;
1597         int alen;
1598         const char *b;
1599         int blen;
1600 {
1601         const char *a0, *b0;
1602         int l;
1603
1604         /* simplest case - need validation? */
1605         if (alen == blen && bcmp(a, b, alen) == 0)
1606                 return 1;
1607
1608         a0 = a;
1609         b0 = b;
1610
1611         /* termination is mandatory */
1612         if (alen < 2 || blen < 2)
1613                 return 0;
1614         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1615                 return 0;
1616         alen--;
1617         blen--;
1618
1619         while (a - a0 < alen && b - b0 < blen) {
1620                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1621                         return 0;
1622
1623                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1624                         return 0;
1625                 /* we don't support compression yet */
1626                 if (a[0] >= 64 || b[0] >= 64)
1627                         return 0;
1628
1629                 /* truncated case */
1630                 if (a[0] == 0 && a - a0 == alen - 1)
1631                         return 1;
1632                 if (b[0] == 0 && b - b0 == blen - 1)
1633                         return 1;
1634                 if (a[0] == 0 || b[0] == 0)
1635                         return 0;
1636
1637                 if (a[0] != b[0])
1638                         return 0;
1639                 l = a[0];
1640                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1641                         return 0;
1642                 if (bcmp(a + 1, b + 1, l) != 0)
1643                         return 0;
1644
1645                 a += 1 + l;
1646                 b += 1 + l;
1647         }
1648
1649         if (a - a0 == alen && b - b0 == blen)
1650                 return 1;
1651         else
1652                 return 0;
1653 }
1654
1655 /*
1656  * calculate the number of addresses to be returned in the node info reply.
1657  */
1658 static int
1659 ni6_addrs(ni6, m, ifpp, subj)
1660         struct icmp6_nodeinfo *ni6;
1661         struct mbuf *m;
1662         struct ifnet **ifpp;
1663         struct in6_addr *subj;
1664 {
1665         struct ifnet *ifp;
1666         struct in6_ifaddr *ifa6;
1667         struct ifaddr *ifa;
1668         int addrs = 0, addrsofif, iffound = 0;
1669         int niflags = ni6->ni_flags;
1670
1671         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1672                 switch (ni6->ni_code) {
1673                 case ICMP6_NI_SUBJ_IPV6:
1674                         if (subj == NULL) /* must be impossible... */
1675                                 return (0);
1676                         break;
1677                 default:
1678                         /*
1679                          * XXX: we only support IPv6 subject address for
1680                          * this Qtype.
1681                          */
1682                         return (0);
1683                 }
1684         }
1685
1686         IFNET_RLOCK();
1687         for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1688                 addrsofif = 0;
1689                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1690                         if (ifa->ifa_addr->sa_family != AF_INET6)
1691                                 continue;
1692                         ifa6 = (struct in6_ifaddr *)ifa;
1693
1694                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1695                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1696                                 iffound = 1;
1697
1698                         /*
1699                          * IPv4-mapped addresses can only be returned by a
1700                          * Node Information proxy, since they represent
1701                          * addresses of IPv4-only nodes, which perforce do
1702                          * not implement this protocol.
1703                          * [icmp-name-lookups-07, Section 5.4]
1704                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1705                          * this function at this moment.
1706                          */
1707
1708                         /* What do we have to do about ::1? */
1709                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1710                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1711                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1712                                         continue;
1713                                 break;
1714                         case IPV6_ADDR_SCOPE_SITELOCAL:
1715                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1716                                         continue;
1717                                 break;
1718                         case IPV6_ADDR_SCOPE_GLOBAL:
1719                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1720                                         continue;
1721                                 break;
1722                         default:
1723                                 continue;
1724                         }
1725
1726                         /*
1727                          * check if anycast is okay.
1728                          * XXX: just experimental.  not in the spec.
1729                          */
1730                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1731                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1732                                 continue; /* we need only unicast addresses */
1733                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1734                             (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1735                                 continue;
1736                         }
1737                         addrsofif++; /* count the address */
1738                 }
1739                 if (iffound) {
1740                         *ifpp = ifp;
1741                         IFNET_RUNLOCK();
1742                         return (addrsofif);
1743                 }
1744
1745                 addrs += addrsofif;
1746         }
1747         IFNET_RUNLOCK();
1748
1749         return (addrs);
1750 }
1751
1752 static int
1753 ni6_store_addrs(ni6, nni6, ifp0, resid)
1754         struct icmp6_nodeinfo *ni6, *nni6;
1755         struct ifnet *ifp0;
1756         int resid;
1757 {
1758         struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1759         struct in6_ifaddr *ifa6;
1760         struct ifaddr *ifa;
1761         struct ifnet *ifp_dep = NULL;
1762         int copied = 0, allow_deprecated = 0;
1763         u_char *cp = (u_char *)(nni6 + 1);
1764         int niflags = ni6->ni_flags;
1765         u_int32_t ltime;
1766
1767         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1768                 return (0);     /* needless to copy */
1769
1770         IFNET_RLOCK();
1771   again:
1772
1773         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1774                 for (ifa = ifp->if_addrlist.tqh_first; ifa;
1775                      ifa = ifa->ifa_list.tqe_next) {
1776                         if (ifa->ifa_addr->sa_family != AF_INET6)
1777                                 continue;
1778                         ifa6 = (struct in6_ifaddr *)ifa;
1779
1780                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1781                             allow_deprecated == 0) {
1782                                 /*
1783                                  * prefererred address should be put before
1784                                  * deprecated addresses.
1785                                  */
1786
1787                                 /* record the interface for later search */
1788                                 if (ifp_dep == NULL)
1789                                         ifp_dep = ifp;
1790
1791                                 continue;
1792                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1793                             allow_deprecated != 0)
1794                                 continue; /* we now collect deprecated addrs */
1795
1796                         /* What do we have to do about ::1? */
1797                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1798                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1799                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1800                                         continue;
1801                                 break;
1802                         case IPV6_ADDR_SCOPE_SITELOCAL:
1803                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1804                                         continue;
1805                                 break;
1806                         case IPV6_ADDR_SCOPE_GLOBAL:
1807                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1808                                         continue;
1809                                 break;
1810                         default:
1811                                 continue;
1812                         }
1813
1814                         /*
1815                          * check if anycast is okay.
1816                          * XXX: just experimental.  not in the spec.
1817                          */
1818                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1819                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1820                                 continue;
1821                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1822                             (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1823                                 continue;
1824                         }
1825
1826                         /* now we can copy the address */
1827                         if (resid < sizeof(struct in6_addr) +
1828                             sizeof(u_int32_t)) {
1829                                 /*
1830                                  * We give up much more copy.
1831                                  * Set the truncate flag and return.
1832                                  */
1833                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1834                                 IFNET_RUNLOCK();
1835                                 return (copied);
1836                         }
1837
1838                         /*
1839                          * Set the TTL of the address.
1840                          * The TTL value should be one of the following
1841                          * according to the specification:
1842                          *
1843                          * 1. The remaining lifetime of a DHCP lease on the
1844                          *    address, or
1845                          * 2. The remaining Valid Lifetime of a prefix from
1846                          *    which the address was derived through Stateless
1847                          *    Autoconfiguration.
1848                          *
1849                          * Note that we currently do not support stateful
1850                          * address configuration by DHCPv6, so the former
1851                          * case can't happen.
1852                          */
1853                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1854                                 ltime = ND6_INFINITE_LIFETIME;
1855                         else {
1856                                 if (ifa6->ia6_lifetime.ia6t_expire >
1857                                     time_second)
1858                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1859                                 else
1860                                         ltime = 0;
1861                         }
1862
1863                         bcopy(&ltime, cp, sizeof(u_int32_t));
1864                         cp += sizeof(u_int32_t);
1865
1866                         /* copy the address itself */
1867                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1868                             sizeof(struct in6_addr));
1869                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1870                         cp += sizeof(struct in6_addr);
1871
1872                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1873                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1874                 }
1875                 if (ifp0)       /* we need search only on the specified IF */
1876                         break;
1877         }
1878
1879         if (allow_deprecated == 0 && ifp_dep != NULL) {
1880                 ifp = ifp_dep;
1881                 allow_deprecated = 1;
1882
1883                 goto again;
1884         }
1885
1886         IFNET_RUNLOCK();
1887
1888         return (copied);
1889 }
1890
1891 /*
1892  * XXX almost dup'ed code with rip6_input.
1893  */
1894 static int
1895 icmp6_rip6_input(mp, off)
1896         struct  mbuf **mp;
1897         int     off;
1898 {
1899         struct mbuf *m = *mp;
1900         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1901         struct in6pcb *in6p;
1902         struct in6pcb *last = NULL;
1903         struct sockaddr_in6 fromsa;
1904         struct icmp6_hdr *icmp6;
1905         struct mbuf *opts = NULL;
1906
1907 #ifndef PULLDOWN_TEST
1908         /* this is assumed to be safe. */
1909         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1910 #else
1911         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1912         if (icmp6 == NULL) {
1913                 /* m is already reclaimed */
1914                 return (IPPROTO_DONE);
1915         }
1916 #endif
1917
1918         /*
1919          * XXX: the address may have embedded scope zone ID, which should be
1920          * hidden from applications.
1921          */
1922         bzero(&fromsa, sizeof(fromsa));
1923         fromsa.sin6_family = AF_INET6;
1924         fromsa.sin6_len = sizeof(struct sockaddr_in6);
1925         fromsa.sin6_addr = ip6->ip6_src;
1926         if (sa6_recoverscope(&fromsa)) {
1927                 m_freem(m);
1928                 return (IPPROTO_DONE);
1929         }
1930
1931         INP_INFO_RLOCK(&ripcbinfo);
1932         LIST_FOREACH(in6p, &ripcb, inp_list) {
1933                 INP_LOCK(in6p);
1934                 if ((in6p->inp_vflag & INP_IPV6) == 0) {
1935         docontinue:
1936                         INP_UNLOCK(in6p);
1937                         continue;
1938                 }
1939                 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1940                         goto docontinue;
1941                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1942                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1943                         goto docontinue;
1944                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1945                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1946                         goto docontinue;
1947                 if (in6p->in6p_icmp6filt
1948                     && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1949                                  in6p->in6p_icmp6filt))
1950                         goto docontinue;
1951                 if (last) {
1952                         struct  mbuf *n = NULL;
1953
1954                         /*
1955                          * Recent network drivers tend to allocate a single
1956                          * mbuf cluster, rather than to make a couple of
1957                          * mbufs without clusters.  Also, since the IPv6 code
1958                          * path tries to avoid m_pullup(), it is highly
1959                          * probable that we still have an mbuf cluster here
1960                          * even though the necessary length can be stored in an
1961                          * mbuf's internal buffer.
1962                          * Meanwhile, the default size of the receive socket
1963                          * buffer for raw sockets is not so large.  This means
1964                          * the possibility of packet loss is relatively higher
1965                          * than before.  To avoid this scenario, we copy the
1966                          * received data to a separate mbuf that does not use
1967                          * a cluster, if possible.
1968                          * XXX: it is better to copy the data after stripping
1969                          * intermediate headers.
1970                          */
1971                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1972                             m->m_len <= MHLEN) {
1973                                 MGET(n, M_DONTWAIT, m->m_type);
1974                                 if (n != NULL) {
1975                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1976                                                 bcopy(m->m_data, n->m_data,
1977                                                       m->m_len);
1978                                                 n->m_len = m->m_len;
1979                                         } else {
1980                                                 m_free(n);
1981                                                 n = NULL;
1982                                         }
1983                                 }
1984                         }
1985                         if (n != NULL ||
1986                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1987                                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1988                                         ip6_savecontrol(last, n, &opts);
1989                                 /* strip intermediate headers */
1990                                 m_adj(n, off);
1991                                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1992                                     (struct sockaddr *)&fromsa, n, opts)
1993                                     == 0) {
1994                                         /* should notify about lost packet */
1995                                         m_freem(n);
1996                                         if (opts) {
1997                                                 m_freem(opts);
1998                                         }
1999                                 } else
2000                                         sorwakeup(last->in6p_socket);
2001                                 opts = NULL;
2002                         }
2003                         INP_UNLOCK(last);
2004                 }
2005                 last = in6p;
2006         }
2007         if (last) {
2008                 if (last->in6p_flags & IN6P_CONTROLOPTS)
2009                         ip6_savecontrol(last, m, &opts);
2010                 /* strip intermediate headers */
2011                 m_adj(m, off);
2012
2013                 /* avoid using mbuf clusters if possible (see above) */
2014                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2015                     m->m_len <= MHLEN) {
2016                         struct mbuf *n;
2017
2018                         MGET(n, M_DONTWAIT, m->m_type);
2019                         if (n != NULL) {
2020                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2021                                         bcopy(m->m_data, n->m_data, m->m_len);
2022                                         n->m_len = m->m_len;
2023
2024                                         m_freem(m);
2025                                         m = n;
2026                                 } else {
2027                                         m_freem(n);
2028                                         n = NULL;
2029                                 }
2030                         }
2031                 }
2032                 if (sbappendaddr(&last->in6p_socket->so_rcv,
2033                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2034                         m_freem(m);
2035                         if (opts)
2036                                 m_freem(opts);
2037                 } else
2038                         sorwakeup(last->in6p_socket);
2039                 INP_UNLOCK(last);
2040         } else {
2041                 m_freem(m);
2042                 ip6stat.ip6s_delivered--;
2043         }
2044         INP_INFO_RUNLOCK(&ripcbinfo);
2045         return IPPROTO_DONE;
2046 }
2047
2048 /*
2049  * Reflect the ip6 packet back to the source.
2050  * OFF points to the icmp6 header, counted from the top of the mbuf.
2051  */
2052 void
2053 icmp6_reflect(m, off)
2054         struct  mbuf *m;
2055         size_t off;
2056 {
2057         struct ip6_hdr *ip6;
2058         struct icmp6_hdr *icmp6;
2059         struct in6_ifaddr *ia;
2060         int plen;
2061         int type, code;
2062         struct ifnet *outif = NULL;
2063         struct in6_addr origdst, *src = NULL;
2064
2065         /* too short to reflect */
2066         if (off < sizeof(struct ip6_hdr)) {
2067                 nd6log((LOG_DEBUG,
2068                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2069                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2070                     __FILE__, __LINE__));
2071                 goto bad;
2072         }
2073
2074         /*
2075          * If there are extra headers between IPv6 and ICMPv6, strip
2076          * off that header first.
2077          */
2078 #ifdef DIAGNOSTIC
2079         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2080                 panic("assumption failed in icmp6_reflect");
2081 #endif
2082         if (off > sizeof(struct ip6_hdr)) {
2083                 size_t l;
2084                 struct ip6_hdr nip6;
2085
2086                 l = off - sizeof(struct ip6_hdr);
2087                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2088                 m_adj(m, l);
2089                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2090                 if (m->m_len < l) {
2091                         if ((m = m_pullup(m, l)) == NULL)
2092                                 return;
2093                 }
2094                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2095         } else /* off == sizeof(struct ip6_hdr) */ {
2096                 size_t l;
2097                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2098                 if (m->m_len < l) {
2099                         if ((m = m_pullup(m, l)) == NULL)
2100                                 return;
2101                 }
2102         }
2103         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2104         ip6 = mtod(m, struct ip6_hdr *);
2105         ip6->ip6_nxt = IPPROTO_ICMPV6;
2106         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2107         type = icmp6->icmp6_type; /* keep type for statistics */
2108         code = icmp6->icmp6_code; /* ditto. */
2109
2110         origdst = ip6->ip6_dst;
2111         /*
2112          * ip6_input() drops a packet if its src is multicast.
2113          * So, the src is never multicast.
2114          */
2115         ip6->ip6_dst = ip6->ip6_src;
2116
2117         /*
2118          * If the incoming packet was addressed directly to us (i.e. unicast),
2119          * use dst as the src for the reply.
2120          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2121          * (for example) when we encounter an error while forwarding procedure
2122          * destined to a duplicated address of ours.
2123          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2124          * procedure of an outgoing packet of our own, in which case we need
2125          * to search in the ifaddr list.
2126          */
2127         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2128                 if ((ia = ip6_getdstifaddr(m))) {
2129                         if (!(ia->ia6_flags &
2130                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2131                                 src = &ia->ia_addr.sin6_addr;
2132                 } else {
2133                         struct sockaddr_in6 d;
2134
2135                         bzero(&d, sizeof(d));
2136                         d.sin6_family = AF_INET6;
2137                         d.sin6_len = sizeof(d);
2138                         d.sin6_addr = origdst;
2139                         ia = (struct in6_ifaddr *)
2140                             ifa_ifwithaddr((struct sockaddr *)&d);
2141                         if (ia &&
2142                             !(ia->ia6_flags &
2143                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2144                                 src = &ia->ia_addr.sin6_addr;
2145                         }
2146                 }
2147         }
2148
2149         if (src == NULL) {
2150                 int e;
2151                 struct sockaddr_in6 sin6;
2152                 struct route_in6 ro;
2153
2154                 /*
2155                  * This case matches to multicasts, our anycast, or unicasts
2156                  * that we do not own.  Select a source address based on the
2157                  * source address of the erroneous packet.
2158                  */
2159                 bzero(&sin6, sizeof(sin6));
2160                 sin6.sin6_family = AF_INET6;
2161                 sin6.sin6_len = sizeof(sin6);
2162                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2163
2164                 bzero(&ro, sizeof(ro));
2165                 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2166                 if (ro.ro_rt)
2167                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2168                 if (src == NULL) {
2169                         nd6log((LOG_DEBUG,
2170                             "icmp6_reflect: source can't be determined: "
2171                             "dst=%s, error=%d\n",
2172                             ip6_sprintf(&sin6.sin6_addr), e));
2173                         goto bad;
2174                 }
2175         }
2176
2177         ip6->ip6_src = *src;
2178         ip6->ip6_flow = 0;
2179         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2180         ip6->ip6_vfc |= IPV6_VERSION;
2181         ip6->ip6_nxt = IPPROTO_ICMPV6;
2182         if (outif)
2183                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2184         else if (m->m_pkthdr.rcvif) {
2185                 /* XXX: This may not be the outgoing interface */
2186                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2187         } else
2188                 ip6->ip6_hlim = ip6_defhlim;
2189
2190         icmp6->icmp6_cksum = 0;
2191         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2192             sizeof(struct ip6_hdr), plen);
2193
2194         /*
2195          * XXX option handling
2196          */
2197
2198         m->m_flags &= ~(M_BCAST|M_MCAST);
2199
2200         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2201         if (outif)
2202                 icmp6_ifoutstat_inc(outif, type, code);
2203
2204         return;
2205
2206  bad:
2207         m_freem(m);
2208         return;
2209 }
2210
2211 void
2212 icmp6_fasttimo()
2213 {
2214
2215         return;
2216 }
2217
2218 static const char *
2219 icmp6_redirect_diag(src6, dst6, tgt6)
2220         struct in6_addr *src6;
2221         struct in6_addr *dst6;
2222         struct in6_addr *tgt6;
2223 {
2224         static char buf[1024];
2225         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2226             ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2227         return buf;
2228 }
2229
2230 void
2231 icmp6_redirect_input(m, off)
2232         struct mbuf *m;
2233         int off;
2234 {
2235         struct ifnet *ifp = m->m_pkthdr.rcvif;
2236         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2237         struct nd_redirect *nd_rd;
2238         int icmp6len = ntohs(ip6->ip6_plen);
2239         char *lladdr = NULL;
2240         int lladdrlen = 0;
2241         u_char *redirhdr = NULL;
2242         int redirhdrlen = 0;
2243         struct rtentry *rt = NULL;
2244         int is_router;
2245         int is_onlink;
2246         struct in6_addr src6 = ip6->ip6_src;
2247         struct in6_addr redtgt6;
2248         struct in6_addr reddst6;
2249         union nd_opts ndopts;
2250
2251         if (!m || !ifp)
2252                 return;
2253
2254         /* XXX if we are router, we don't update route by icmp6 redirect */
2255         if (ip6_forwarding)
2256                 goto freeit;
2257         if (!icmp6_rediraccept)
2258                 goto freeit;
2259
2260 #ifndef PULLDOWN_TEST
2261         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2262         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2263 #else
2264         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2265         if (nd_rd == NULL) {
2266                 icmp6stat.icp6s_tooshort++;
2267                 return;
2268         }
2269 #endif
2270         redtgt6 = nd_rd->nd_rd_target;
2271         reddst6 = nd_rd->nd_rd_dst;
2272
2273         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2274             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2275                 goto freeit;
2276         }
2277
2278         /* validation */
2279         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2280                 nd6log((LOG_ERR,
2281                     "ICMP6 redirect sent from %s rejected; "
2282                     "must be from linklocal\n",
2283                     ip6_sprintf(&src6)));
2284                 goto bad;
2285         }
2286         if (ip6->ip6_hlim != 255) {
2287                 nd6log((LOG_ERR,
2288                     "ICMP6 redirect sent from %s rejected; "
2289                     "hlim=%d (must be 255)\n",
2290                     ip6_sprintf(&src6), ip6->ip6_hlim));
2291                 goto bad;
2292         }
2293     {
2294         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2295         struct sockaddr_in6 sin6;
2296         struct in6_addr *gw6;
2297
2298         bzero(&sin6, sizeof(sin6));
2299         sin6.sin6_family = AF_INET6;
2300         sin6.sin6_len = sizeof(struct sockaddr_in6);
2301         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2302         rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2303         if (rt) {
2304                 if (rt->rt_gateway == NULL ||
2305                     rt->rt_gateway->sa_family != AF_INET6) {
2306                         nd6log((LOG_ERR,
2307                             "ICMP6 redirect rejected; no route "
2308                             "with inet6 gateway found for redirect dst: %s\n",
2309                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2310                         RTFREE_LOCKED(rt);
2311                         goto bad;
2312                 }
2313
2314                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2315                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2316                         nd6log((LOG_ERR,
2317                             "ICMP6 redirect rejected; "
2318                             "not equal to gw-for-src=%s (must be same): "
2319                             "%s\n",
2320                             ip6_sprintf(gw6),
2321                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2322                         RTFREE_LOCKED(rt);
2323                         goto bad;
2324                 }
2325         } else {
2326                 nd6log((LOG_ERR,
2327                     "ICMP6 redirect rejected; "
2328                     "no route found for redirect dst: %s\n",
2329                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2330                 goto bad;
2331         }
2332         RTFREE_LOCKED(rt);
2333         rt = NULL;
2334     }
2335         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2336                 nd6log((LOG_ERR,
2337                     "ICMP6 redirect rejected; "
2338                     "redirect dst must be unicast: %s\n",
2339                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2340                 goto bad;
2341         }
2342
2343         is_router = is_onlink = 0;
2344         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2345                 is_router = 1;  /* router case */
2346         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2347                 is_onlink = 1;  /* on-link destination case */
2348         if (!is_router && !is_onlink) {
2349                 nd6log((LOG_ERR,
2350                     "ICMP6 redirect rejected; "
2351                     "neither router case nor onlink case: %s\n",
2352                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2353                 goto bad;
2354         }
2355         /* validation passed */
2356
2357         icmp6len -= sizeof(*nd_rd);
2358         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2359         if (nd6_options(&ndopts) < 0) {
2360                 nd6log((LOG_INFO, "icmp6_redirect_input: "
2361                     "invalid ND option, rejected: %s\n",
2362                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2363                 /* nd6_options have incremented stats */
2364                 goto freeit;
2365         }
2366
2367         if (ndopts.nd_opts_tgt_lladdr) {
2368                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2369                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2370         }
2371
2372         if (ndopts.nd_opts_rh) {
2373                 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2374                 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2375         }
2376
2377         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2378                 nd6log((LOG_INFO,
2379                     "icmp6_redirect_input: lladdrlen mismatch for %s "
2380                     "(if %d, icmp6 packet %d): %s\n",
2381                     ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2382                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2383                 goto bad;
2384         }
2385
2386         /* RFC 2461 8.3 */
2387         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2388             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2389
2390         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2391                 /* perform rtredirect */
2392                 struct sockaddr_in6 sdst;
2393                 struct sockaddr_in6 sgw;
2394                 struct sockaddr_in6 ssrc;
2395
2396                 bzero(&sdst, sizeof(sdst));
2397                 bzero(&sgw, sizeof(sgw));
2398                 bzero(&ssrc, sizeof(ssrc));
2399                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2400                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2401                         sizeof(struct sockaddr_in6);
2402                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2403                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2404                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2405                 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2406                     (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2407                     (struct sockaddr *)&ssrc);
2408         }
2409         /* finally update cached route in each socket via pfctlinput */
2410     {
2411         struct sockaddr_in6 sdst;
2412
2413         bzero(&sdst, sizeof(sdst));
2414         sdst.sin6_family = AF_INET6;
2415         sdst.sin6_len = sizeof(struct sockaddr_in6);
2416         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2417         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2418 #if defined(IPSEC) || defined(FAST_IPSEC)
2419         key_sa_routechange((struct sockaddr *)&sdst);
2420 #endif
2421     }
2422
2423  freeit:
2424         m_freem(m);
2425         return;
2426
2427  bad:
2428         icmp6stat.icp6s_badredirect++;
2429         m_freem(m);
2430 }
2431
2432 void
2433 icmp6_redirect_output(m0, rt)
2434         struct mbuf *m0;
2435         struct rtentry *rt;
2436 {
2437         struct ifnet *ifp;      /* my outgoing interface */
2438         struct in6_addr *ifp_ll6;
2439         struct in6_addr *router_ll6;
2440         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2441         struct mbuf *m = NULL;  /* newly allocated one */
2442         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2443         struct nd_redirect *nd_rd;
2444         size_t maxlen;
2445         u_char *p;
2446         struct ifnet *outif = NULL;
2447         struct sockaddr_in6 src_sa;
2448
2449         icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2450
2451         /* if we are not router, we don't send icmp6 redirect */
2452         if (!ip6_forwarding)
2453                 goto fail;
2454
2455         /* sanity check */
2456         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2457                 goto fail;
2458
2459         /*
2460          * Address check:
2461          *  the source address must identify a neighbor, and
2462          *  the destination address must not be a multicast address
2463          *  [RFC 2461, sec 8.2]
2464          */
2465         sip6 = mtod(m0, struct ip6_hdr *);
2466         bzero(&src_sa, sizeof(src_sa));
2467         src_sa.sin6_family = AF_INET6;
2468         src_sa.sin6_len = sizeof(src_sa);
2469         src_sa.sin6_addr = sip6->ip6_src;
2470         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2471                 goto fail;
2472         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2473                 goto fail;      /* what should we do here? */
2474
2475         /* rate limit */
2476         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2477                 goto fail;
2478
2479         /*
2480          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2481          * we almost always ask for an mbuf cluster for simplicity.
2482          * (MHLEN < IPV6_MMTU is almost always true)
2483          */
2484 #if IPV6_MMTU >= MCLBYTES
2485 # error assumption failed about IPV6_MMTU and MCLBYTES
2486 #endif
2487         MGETHDR(m, M_DONTWAIT, MT_HEADER);
2488         if (m && IPV6_MMTU >= MHLEN)
2489                 MCLGET(m, M_DONTWAIT);
2490         if (!m)
2491                 goto fail;
2492         m->m_pkthdr.rcvif = NULL;
2493         m->m_len = 0;
2494         maxlen = M_TRAILINGSPACE(m);
2495         maxlen = min(IPV6_MMTU, maxlen);
2496         /* just for safety */
2497         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2498             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2499                 goto fail;
2500         }
2501
2502         {
2503                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2504                 struct in6_ifaddr *ia;
2505                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2506                                                  IN6_IFF_NOTREADY|
2507                                                  IN6_IFF_ANYCAST)) == NULL)
2508                         goto fail;
2509                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2510         }
2511
2512         /* get ip6 linklocal address for the router. */
2513         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2514                 struct sockaddr_in6 *sin6;
2515                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2516                 router_ll6 = &sin6->sin6_addr;
2517                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2518                         router_ll6 = (struct in6_addr *)NULL;
2519         } else
2520                 router_ll6 = (struct in6_addr *)NULL;
2521
2522         /* ip6 */
2523         ip6 = mtod(m, struct ip6_hdr *);
2524         ip6->ip6_flow = 0;
2525         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2526         ip6->ip6_vfc |= IPV6_VERSION;
2527         /* ip6->ip6_plen will be set later */
2528         ip6->ip6_nxt = IPPROTO_ICMPV6;
2529         ip6->ip6_hlim = 255;
2530         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2531         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2532         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2533
2534         /* ND Redirect */
2535         nd_rd = (struct nd_redirect *)(ip6 + 1);
2536         nd_rd->nd_rd_type = ND_REDIRECT;
2537         nd_rd->nd_rd_code = 0;
2538         nd_rd->nd_rd_reserved = 0;
2539         if (rt->rt_flags & RTF_GATEWAY) {
2540                 /*
2541                  * nd_rd->nd_rd_target must be a link-local address in
2542                  * better router cases.
2543                  */
2544                 if (!router_ll6)
2545                         goto fail;
2546                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2547                     sizeof(nd_rd->nd_rd_target));
2548                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2549                     sizeof(nd_rd->nd_rd_dst));
2550         } else {
2551                 /* make sure redtgt == reddst */
2552                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2553                     sizeof(nd_rd->nd_rd_target));
2554                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2555                     sizeof(nd_rd->nd_rd_dst));
2556         }
2557
2558         p = (u_char *)(nd_rd + 1);
2559
2560         if (!router_ll6)
2561                 goto nolladdropt;
2562
2563         {
2564                 /* target lladdr option */
2565                 struct rtentry *rt_router = NULL;
2566                 int len;
2567                 struct sockaddr_dl *sdl;
2568                 struct nd_opt_hdr *nd_opt;
2569                 char *lladdr;
2570
2571                 rt_router = nd6_lookup(router_ll6, 0, ifp);
2572                 if (!rt_router)
2573                         goto nolladdropt;
2574                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2575                 len = (len + 7) & ~7;   /* round by 8 */
2576                 /* safety check */
2577                 if (len + (p - (u_char *)ip6) > maxlen)
2578                         goto nolladdropt;
2579                 if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2580                     (rt_router->rt_flags & RTF_LLINFO) &&
2581                     (rt_router->rt_gateway->sa_family == AF_LINK) &&
2582                     (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2583                     sdl->sdl_alen) {
2584                         nd_opt = (struct nd_opt_hdr *)p;
2585                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2586                         nd_opt->nd_opt_len = len >> 3;
2587                         lladdr = (char *)(nd_opt + 1);
2588                         bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2589                         p += len;
2590                 }
2591         }
2592 nolladdropt:;
2593
2594         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2595
2596         /* just to be safe */
2597 #ifdef M_DECRYPTED      /*not openbsd*/
2598         if (m0->m_flags & M_DECRYPTED)
2599                 goto noredhdropt;
2600 #endif
2601         if (p - (u_char *)ip6 > maxlen)
2602                 goto noredhdropt;
2603
2604         {
2605                 /* redirected header option */
2606                 int len;
2607                 struct nd_opt_rd_hdr *nd_opt_rh;
2608
2609                 /*
2610                  * compute the maximum size for icmp6 redirect header option.
2611                  * XXX room for auth header?
2612                  */
2613                 len = maxlen - (p - (u_char *)ip6);
2614                 len &= ~7;
2615
2616                 /* This is just for simplicity. */
2617                 if (m0->m_pkthdr.len != m0->m_len) {
2618                         if (m0->m_next) {
2619                                 m_freem(m0->m_next);
2620                                 m0->m_next = NULL;
2621                         }
2622                         m0->m_pkthdr.len = m0->m_len;
2623                 }
2624
2625                 /*
2626                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2627                  * about padding/truncate rule for the original IP packet.
2628                  * From the discussion on IPv6imp in Feb 1999,
2629                  * the consensus was:
2630                  * - "attach as much as possible" is the goal
2631                  * - pad if not aligned (original size can be guessed by
2632                  *   original ip6 header)
2633                  * Following code adds the padding if it is simple enough,
2634                  * and truncates if not.
2635                  */
2636                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2637                         panic("assumption failed in %s:%d", __FILE__,
2638                             __LINE__);
2639
2640                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2641                         /* not enough room, truncate */
2642                         m0->m_pkthdr.len = m0->m_len = len -
2643                             sizeof(*nd_opt_rh);
2644                 } else {
2645                         /* enough room, pad or truncate */
2646                         size_t extra;
2647
2648                         extra = m0->m_pkthdr.len % 8;
2649                         if (extra) {
2650                                 /* pad if easy enough, truncate if not */
2651                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2652                                         /* pad */
2653                                         m0->m_len += (8 - extra);
2654                                         m0->m_pkthdr.len += (8 - extra);
2655                                 } else {
2656                                         /* truncate */
2657                                         m0->m_pkthdr.len -= extra;
2658                                         m0->m_len -= extra;
2659                                 }
2660                         }
2661                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2662                         m0->m_pkthdr.len = m0->m_len = len -
2663                             sizeof(*nd_opt_rh);
2664                 }
2665
2666                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2667                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2668                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2669                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2670                 p += sizeof(*nd_opt_rh);
2671                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2672
2673                 /* connect m0 to m */
2674                 m_tag_delete_chain(m0, NULL);
2675                 m0->m_flags &= ~M_PKTHDR;
2676                 m->m_next = m0;
2677                 m->m_pkthdr.len = m->m_len + m0->m_len;
2678                 m0 = NULL;
2679         }
2680 noredhdropt:;
2681         if (m0) {
2682                 m_freem(m0);
2683                 m0 = NULL;
2684         }
2685
2686         /* XXX: clear embedded link IDs in the inner header */
2687         in6_clearscope(&sip6->ip6_src);
2688         in6_clearscope(&sip6->ip6_dst);
2689         in6_clearscope(&nd_rd->nd_rd_target);
2690         in6_clearscope(&nd_rd->nd_rd_dst);
2691
2692         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2693
2694         nd_rd->nd_rd_cksum = 0;
2695         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2696             sizeof(*ip6), ntohs(ip6->ip6_plen));
2697
2698         /* send the packet to outside... */
2699         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2700         if (outif) {
2701                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2702                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2703         }
2704         icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2705
2706         return;
2707
2708 fail:
2709         if (m)
2710                 m_freem(m);
2711         if (m0)
2712                 m_freem(m0);
2713 }
2714
2715 /*
2716  * ICMPv6 socket option processing.
2717  */
2718 int
2719 icmp6_ctloutput(so, sopt)
2720         struct socket *so;
2721         struct sockopt *sopt;
2722 {
2723         int error = 0;
2724         int optlen;
2725         struct inpcb *inp = sotoinpcb(so);
2726         int level, op, optname;
2727
2728         if (sopt) {
2729                 level = sopt->sopt_level;
2730                 op = sopt->sopt_dir;
2731                 optname = sopt->sopt_name;
2732                 optlen = sopt->sopt_valsize;
2733         } else
2734                 level = op = optname = optlen = 0;
2735
2736         if (level != IPPROTO_ICMPV6) {
2737                 return EINVAL;
2738         }
2739
2740         switch (op) {
2741         case PRCO_SETOPT:
2742                 switch (optname) {
2743                 case ICMP6_FILTER:
2744                     {
2745                         struct icmp6_filter *p;
2746
2747                         if (optlen != sizeof(*p)) {
2748                                 error = EMSGSIZE;
2749                                 break;
2750                         }
2751                         if (inp->in6p_icmp6filt == NULL) {
2752                                 error = EINVAL;
2753                                 break;
2754                         }
2755                         error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2756                                 optlen);
2757                         break;
2758                     }
2759
2760                 default:
2761                         error = ENOPROTOOPT;
2762                         break;
2763                 }
2764                 break;
2765
2766         case PRCO_GETOPT:
2767                 switch (optname) {
2768                 case ICMP6_FILTER:
2769                     {
2770                         if (inp->in6p_icmp6filt == NULL) {
2771                                 error = EINVAL;
2772                                 break;
2773                         }
2774                         error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2775                                 sizeof(struct icmp6_filter));
2776                         break;
2777                     }
2778
2779                 default:
2780                         error = ENOPROTOOPT;
2781                         break;
2782                 }
2783                 break;
2784         }
2785
2786         return (error);
2787 }
2788
2789 /*
2790  * Perform rate limit check.
2791  * Returns 0 if it is okay to send the icmp6 packet.
2792  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2793  * limitation.
2794  *
2795  * XXX per-destination/type check necessary?
2796  */
2797 static int
2798 icmp6_ratelimit(dst, type, code)
2799         const struct in6_addr *dst;     /* not used at this moment */
2800         const int type;                 /* not used at this moment */
2801         const int code;                 /* not used at this moment */
2802 {
2803         int ret;
2804
2805         ret = 0;        /* okay to send */
2806
2807         /* PPS limit */
2808         if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2809             icmp6errppslim)) {
2810                 /* The packet is subject to rate limit */
2811                 ret++;
2812         }
2813
2814         return ret;
2815 }