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 <netipsec/ipsec.h>
 106 #include <netipsec/key.h>
 107 #endif
 108
 109 extern struct domain inet6domain;
 110
 111 struct icmp6stat icmp6stat;
 112
 113 extern struct inpcbinfo ripcbinfo;
 114 extern struct inpcbhead ripcb;
 115 extern int icmp6errppslim;
 116 static int icmp6errpps_count = 0;
 117 static struct timeval icmp6errppslim_last;
 118 extern int icmp6_nodeinfo;
 119
 120 static void icmp6_errcount(struct icmp6errstat *, int, int);
 121 static int icmp6_rip6_input(struct mbuf **, int);
 122 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
 123 static const char *icmp6_redirect_diag __P((struct in6_addr *,
 124         struct in6_addr *, struct in6_addr *));
 125 static struct mbuf *ni6_input(struct mbuf *, int);
 126 static struct mbuf *ni6_nametodns(const char *, int, int);
 127 static int ni6_dnsmatch(const char *, int, const char *, int);
 128 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
 129                           struct ifnet **, struct in6_addr *));
 130 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
 131                                 struct ifnet *, int));
 132 static int icmp6_notify_error(struct mbuf **, int, int, int);
 133
 134
 135 void
 136 icmp6_init(void)
 137 {
 138
 139         mld6_init();
 140 }
 141
 142 static void
 143 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
 144 {
 145         switch (type) {
 146         case ICMP6_DST_UNREACH:
 147                 switch (code) {
 148                 case ICMP6_DST_UNREACH_NOROUTE:
 149                         stat->icp6errs_dst_unreach_noroute++;
 150                         return;
 151                 case ICMP6_DST_UNREACH_ADMIN:
 152                         stat->icp6errs_dst_unreach_admin++;
 153                         return;
 154                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 155                         stat->icp6errs_dst_unreach_beyondscope++;
 156                         return;
 157                 case ICMP6_DST_UNREACH_ADDR:
 158                         stat->icp6errs_dst_unreach_addr++;
 159                         return;
 160                 case ICMP6_DST_UNREACH_NOPORT:
 161                         stat->icp6errs_dst_unreach_noport++;
 162                         return;
 163                 }
 164                 break;
 165         case ICMP6_PACKET_TOO_BIG:
 166                 stat->icp6errs_packet_too_big++;
 167                 return;
 168         case ICMP6_TIME_EXCEEDED:
 169                 switch (code) {
 170                 case ICMP6_TIME_EXCEED_TRANSIT:
 171                         stat->icp6errs_time_exceed_transit++;
 172                         return;
 173                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 174                         stat->icp6errs_time_exceed_reassembly++;
 175                         return;
 176                 }
 177                 break;
 178         case ICMP6_PARAM_PROB:
 179                 switch (code) {
 180                 case ICMP6_PARAMPROB_HEADER:
 181                         stat->icp6errs_paramprob_header++;
 182                         return;
 183                 case ICMP6_PARAMPROB_NEXTHEADER:
 184                         stat->icp6errs_paramprob_nextheader++;
 185                         return;
 186                 case ICMP6_PARAMPROB_OPTION:
 187                         stat->icp6errs_paramprob_option++;
 188                         return;
 189                 }
 190                 break;
 191         case ND_REDIRECT:
 192                 stat->icp6errs_redirect++;
 193                 return;
 194         }
 195         stat->icp6errs_unknown++;
 196 }
 197
 198 /*
 199  * A wrapper function for icmp6_error() necessary when the erroneous packet
 200  * may not contain enough scope zone information.
 201  */
 202 void
 203 icmp6_error2(struct mbuf *m, int type, int code, int param,
 204     struct ifnet *ifp)
 205 {
 206         struct ip6_hdr *ip6;
 207
 208         if (ifp == NULL)
 209                 return;
 210
 211 #ifndef PULLDOWN_TEST
 212         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 213 #else
 214         if (m->m_len < sizeof(struct ip6_hdr)) {
 215                 m = m_pullup(m, sizeof(struct ip6_hdr));
 216                 if (m == NULL)
 217                         return;
 218         }
 219 #endif
 220
 221         ip6 = mtod(m, struct ip6_hdr *);
 222
 223         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
 224                 return;
 225         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
 226                 return;
 227
 228         icmp6_error(m, type, code, param);
 229 }
 230
 231 /*
 232  * Generate an error packet of type error in response to bad IP6 packet.
 233  */
 234 void
 235 icmp6_error(struct mbuf *m, int type, int code, int param)
 236 {
 237         struct ip6_hdr *oip6, *nip6;
 238         struct icmp6_hdr *icmp6;
 239         u_int preplen;
 240         int off;
 241         int nxt;
 242
 243         icmp6stat.icp6s_error++;
 244
 245         /* count per-type-code statistics */
 246         icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
 247
 248 #ifdef M_DECRYPTED      /*not openbsd*/
 249         if (m->m_flags & M_DECRYPTED) {
 250                 icmp6stat.icp6s_canterror++;
 251                 goto freeit;
 252         }
 253 #endif
 254
 255 #ifndef PULLDOWN_TEST
 256         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 257 #else
 258         if (m->m_len < sizeof(struct ip6_hdr)) {
 259                 m = m_pullup(m, sizeof(struct ip6_hdr));
 260                 if (m == NULL)
 261                         return;
 262         }
 263 #endif
 264         oip6 = mtod(m, struct ip6_hdr *);
 265
 266         /*
 267          * If the destination address of the erroneous packet is a multicast
 268          * address, or the packet was sent using link-layer multicast,
 269          * we should basically suppress sending an error (RFC 2463, Section
 270          * 2.4).
 271          * We have two exceptions (the item e.2 in that section):
 272          * - the Pakcet Too Big message can be sent for path MTU discovery.
 273          * - the Parameter Problem Message that can be allowed an icmp6 error
 274          *   in the option type field.  This check has been done in
 275          *   ip6_unknown_opt(), so we can just check the type and code.
 276          */
 277         if ((m->m_flags & (M_BCAST|M_MCAST) ||
 278              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
 279             (type != ICMP6_PACKET_TOO_BIG &&
 280              (type != ICMP6_PARAM_PROB ||
 281               code != ICMP6_PARAMPROB_OPTION)))
 282                 goto freeit;
 283
 284         /*
 285          * RFC 2463, 2.4 (e.5): source address check.
 286          * XXX: the case of anycast source?
 287          */
 288         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
 289             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
 290                 goto freeit;
 291
 292         /*
 293          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
 294          * don't do it.
 295          */
 296         nxt = -1;
 297         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
 298         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
 299                 struct icmp6_hdr *icp;
 300
 301 #ifndef PULLDOWN_TEST
 302                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
 303                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 304 #else
 305                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
 306                         sizeof(*icp));
 307                 if (icp == NULL) {
 308                         icmp6stat.icp6s_tooshort++;
 309                         return;
 310                 }
 311 #endif
 312                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
 313                     icp->icmp6_type == ND_REDIRECT) {
 314                         /*
 315                          * ICMPv6 error
 316                          * Special case: for redirect (which is
 317                          * informational) we must not send icmp6 error.
 318                          */
 319                         icmp6stat.icp6s_canterror++;
 320                         goto freeit;
 321                 } else {
 322                         /* ICMPv6 informational - send the error */
 323                 }
 324         } else {
 325                 /* non-ICMPv6 - send the error */
 326         }
 327
 328         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
 329
 330         /* Finally, do rate limitation check. */
 331         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
 332                 icmp6stat.icp6s_toofreq++;
 333                 goto freeit;
 334         }
 335
 336         /*
 337          * OK, ICMP6 can be generated.
 338          */
 339
 340         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
 341                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
 342
 343         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
 344         M_PREPEND(m, preplen, M_DONTWAIT);
 345         if (m && m->m_len < preplen)
 346                 m = m_pullup(m, preplen);
 347         if (m == NULL) {
 348                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
 349                 return;
 350         }
 351
 352         nip6 = mtod(m, struct ip6_hdr *);
 353         nip6->ip6_src  = oip6->ip6_src;
 354         nip6->ip6_dst  = oip6->ip6_dst;
 355
 356         in6_clearscope(&oip6->ip6_src);
 357         in6_clearscope(&oip6->ip6_dst);
 358
 359         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
 360         icmp6->icmp6_type = type;
 361         icmp6->icmp6_code = code;
 362         icmp6->icmp6_pptr = htonl((u_int32_t)param);
 363
 364         /*
 365          * icmp6_reflect() is designed to be in the input path.
 366          * icmp6_error() can be called from both input and output path,
 367          * and if we are in output path rcvif could contain bogus value.
 368          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
 369          * information in ip header (nip6).
 370          */
 371         m->m_pkthdr.rcvif = NULL;
 372
 373         icmp6stat.icp6s_outhist[type]++;
 374         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
 375
 376         return;
 377
 378   freeit:
 379         /*
 380          * If we can't tell whether or not we can generate ICMP6, free it.
 381          */
 382         m_freem(m);
 383 }
 384
 385 /*
 386  * Process a received ICMP6 message.
 387  */
 388 int
 389 icmp6_input(struct mbuf **mp, int *offp, int proto)
 390 {
 391         struct mbuf *m = *mp, *n;
 392         struct ip6_hdr *ip6, *nip6;
 393         struct icmp6_hdr *icmp6, *nicmp6;
 394         int off = *offp;
 395         int icmp6len = m->m_pkthdr.len - *offp;
 396         int code, sum, noff;
 397         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
 398
 399 #ifndef PULLDOWN_TEST
 400         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
 401         /* m might change if M_LOOP.  So, call mtod after this */
 402 #endif
 403
 404         /*
 405          * Locate icmp6 structure in mbuf, and check
 406          * that not corrupted and of at least minimum length
 407          */
 408
 409         ip6 = mtod(m, struct ip6_hdr *);
 410         if (icmp6len < sizeof(struct icmp6_hdr)) {
 411                 icmp6stat.icp6s_tooshort++;
 412                 goto freeit;
 413         }
 414
 415         /*
 416          * calculate the checksum
 417          */
 418 #ifndef PULLDOWN_TEST
 419         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
 420 #else
 421         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
 422         if (icmp6 == NULL) {
 423                 icmp6stat.icp6s_tooshort++;
 424                 return IPPROTO_DONE;
 425         }
 426 #endif
 427         code = icmp6->icmp6_code;
 428
 429         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
 430                 nd6log((LOG_ERR,
 431                     "ICMP6 checksum error(%d|%x) %s\n",
 432                     icmp6->icmp6_type, sum,
 433                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
 434                 icmp6stat.icp6s_checksum++;
 435                 goto freeit;
 436         }
 437
 438         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
 439                 /*
 440                  * Deliver very specific ICMP6 type only.
 441                  * This is important to deliver TOOBIG.  Otherwise PMTUD
 442                  * will not work.
 443                  */
 444                 switch (icmp6->icmp6_type) {
 445                 case ICMP6_DST_UNREACH:
 446                 case ICMP6_PACKET_TOO_BIG:
 447                 case ICMP6_TIME_EXCEEDED:
 448                         break;
 449                 default:
 450                         goto freeit;
 451                 }
 452         }
 453
 454         icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
 455         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
 456         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
 457                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
 458
 459         switch (icmp6->icmp6_type) {
 460         case ICMP6_DST_UNREACH:
 461                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
 462                 switch (code) {
 463                 case ICMP6_DST_UNREACH_NOROUTE:
 464                         code = PRC_UNREACH_NET;
 465                         break;
 466                 case ICMP6_DST_UNREACH_ADMIN:
 467                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
 468                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
 469                         break;
 470                 case ICMP6_DST_UNREACH_ADDR:
 471                         code = PRC_HOSTDEAD;
 472                         break;
 473                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 474                         /* I mean "source address was incorrect." */
 475                         code = PRC_PARAMPROB;
 476                         break;
 477                 case ICMP6_DST_UNREACH_NOPORT:
 478                         code = PRC_UNREACH_PORT;
 479                         break;
 480                 default:
 481                         goto badcode;
 482                 }
 483                 goto deliver;
 484                 break;
 485
 486         case ICMP6_PACKET_TOO_BIG:
 487                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
 488
 489                 /* validation is made in icmp6_mtudisc_update */
 490
 491                 code = PRC_MSGSIZE;
 492
 493                 /*
 494                  * Updating the path MTU will be done after examining
 495                  * intermediate extension headers.
 496                  */
 497                 goto deliver;
 498                 break;
 499
 500         case ICMP6_TIME_EXCEEDED:
 501                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
 502                 switch (code) {
 503                 case ICMP6_TIME_EXCEED_TRANSIT:
 504                         code = PRC_TIMXCEED_INTRANS;
 505                         break;
 506                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 507                         code = PRC_TIMXCEED_REASS;
 508                         break;
 509                 default:
 510                         goto badcode;
 511                 }
 512                 goto deliver;
 513                 break;
 514
 515         case ICMP6_PARAM_PROB:
 516                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
 517                 switch (code) {
 518                 case ICMP6_PARAMPROB_NEXTHEADER:
 519                         code = PRC_UNREACH_PROTOCOL;
 520                         break;
 521                 case ICMP6_PARAMPROB_HEADER:
 522                 case ICMP6_PARAMPROB_OPTION:
 523                         code = PRC_PARAMPROB;
 524                         break;
 525                 default:
 526                         goto badcode;
 527                 }
 528                 goto deliver;
 529                 break;
 530
 531         case ICMP6_ECHO_REQUEST:
 532                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
 533                 if (code != 0)
 534                         goto badcode;
 535                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
 536                         /* Give up remote */
 537                         break;
 538                 }
 539                 if ((n->m_flags & M_EXT) != 0
 540                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
 541                         struct mbuf *n0 = n;
 542                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
 543                         int n0len;
 544
 545                         MGETHDR(n, M_DONTWAIT, n0->m_type);
 546                         n0len = n0->m_pkthdr.len;       /* save for use below */
 547                         if (n)
 548                                 M_MOVE_PKTHDR(n, n0);
 549                         if (n && maxlen >= MHLEN) {
 550                                 MCLGET(n, M_DONTWAIT);
 551                                 if ((n->m_flags & M_EXT) == 0) {
 552                                         m_free(n);
 553                                         n = NULL;
 554                                 }
 555                         }
 556                         if (n == NULL) {
 557                                 /* Give up remote */
 558                                 m_freem(n0);
 559                                 break;
 560                         }
 561                         /*
 562                          * Copy IPv6 and ICMPv6 only.
 563                          */
 564                         nip6 = mtod(n, struct ip6_hdr *);
 565                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 566                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 567                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 568                         noff = sizeof(struct ip6_hdr);
 569                         /* new mbuf contains only ipv6+icmpv6 headers */
 570                         n->m_len = noff + sizeof(struct icmp6_hdr);
 571                         /*
 572                          * Adjust mbuf.  ip6_plen will be adjusted in
 573                          * ip6_output().
 574                          */
 575                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 576                         /* recalculate complete packet size */
 577                         n->m_pkthdr.len = n0len + (noff - off);
 578                         n->m_next = n0;
 579                 } else {
 580                         nip6 = mtod(n, struct ip6_hdr *);
 581                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
 582                             sizeof(*nicmp6));
 583                         noff = off;
 584                 }
 585                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 586                 nicmp6->icmp6_code = 0;
 587                 if (n) {
 588                         icmp6stat.icp6s_reflect++;
 589                         icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
 590                         icmp6_reflect(n, noff);
 591                 }
 592                 break;
 593
 594         case ICMP6_ECHO_REPLY:
 595                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
 596                 if (code != 0)
 597                         goto badcode;
 598                 break;
 599
 600         case MLD_LISTENER_QUERY:
 601         case MLD_LISTENER_REPORT:
 602                 if (icmp6len < sizeof(struct mld_hdr))
 603                         goto badlen;
 604                 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
 605                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
 606                 else
 607                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
 608                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 609                         /* give up local */
 610                         mld6_input(m, off);
 611                         m = NULL;
 612                         goto freeit;
 613                 }
 614                 mld6_input(n, off);
 615                 /* m stays. */
 616                 break;
 617
 618         case MLD_LISTENER_DONE:
 619                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
 620                 if (icmp6len < sizeof(struct mld_hdr))  /* necessary? */
 621                         goto badlen;
 622                 break;          /* nothing to be done in kernel */
 623
 624         case MLD_MTRACE_RESP:
 625         case MLD_MTRACE:
 626                 /* XXX: these two are experimental.  not officially defined. */
 627                 /* XXX: per-interface statistics? */
 628                 break;          /* just pass it to applications */
 629
 630         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 631             {
 632                 enum { WRU, FQDN } mode;
 633
 634                 if (!icmp6_nodeinfo)
 635                         break;
 636
 637                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 638                         mode = WRU;
 639                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 640                         mode = FQDN;
 641                 else
 642                         goto badlen;
 643
 644 #define hostnamelen     strlen(hostname)
 645                 if (mode == FQDN) {
 646 #ifndef PULLDOWN_TEST
 647                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
 648                             IPPROTO_DONE);
 649 #endif
 650                         n = m_copy(m, 0, M_COPYALL);
 651                         if (n)
 652                                 n = ni6_input(n, off);
 653                         /* XXX meaningless if n == NULL */
 654                         noff = sizeof(struct ip6_hdr);
 655                 } else {
 656                         u_char *p;
 657                         int maxlen, maxhlen;
 658
 659                         /*
 660                          * XXX: this combination of flags is pointless,
 661                          * but should we keep this for compatibility?
 662                          */
 663                         if ((icmp6_nodeinfo & 5) != 5)
 664                                 break;
 665
 666                         if (code != 0)
 667                                 goto badcode;
 668                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
 669                         if (maxlen >= MCLBYTES) {
 670                                 /* Give up remote */
 671                                 break;
 672                         }
 673                         MGETHDR(n, M_DONTWAIT, m->m_type);
 674                         if (n && maxlen > MHLEN) {
 675                                 MCLGET(n, M_DONTWAIT);
 676                                 if ((n->m_flags & M_EXT) == 0) {
 677                                         m_free(n);
 678                                         n = NULL;
 679                                 }
 680                         }
 681                         if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
 682                                 /*
 683                                  * Previous code did a blind M_COPY_PKTHDR
 684                                  * and said "just for rcvif".  If true, then
 685                                  * we could tolerate the dup failing (due to
 686                                  * the deep copy of the tag chain).  For now
 687                                  * be conservative and just fail.
 688                                  */
 689                                 m_free(n);
 690                                 n = NULL;
 691                         }
 692                         if (n == NULL) {
 693                                 /* Give up remote */
 694                                 break;
 695                         }
 696                         n->m_pkthdr.rcvif = NULL;
 697                         n->m_len = 0;
 698                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
 699                         if (maxhlen > hostnamelen)
 700                                 maxhlen = hostnamelen;
 701                         /*
 702                          * Copy IPv6 and ICMPv6 only.
 703                          */
 704                         nip6 = mtod(n, struct ip6_hdr *);
 705                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 706                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 707                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 708                         p = (u_char *)(nicmp6 + 1);
 709                         bzero(p, 4);
 710                         bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
 711                         noff = sizeof(struct ip6_hdr);
 712                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 713                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 714                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 715                         nicmp6->icmp6_code = 0;
 716                 }
 717 #undef hostnamelen
 718                 if (n) {
 719                         icmp6stat.icp6s_reflect++;
 720                         icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
 721                         icmp6_reflect(n, noff);
 722                 }
 723                 break;
 724             }
 725
 726         case ICMP6_WRUREPLY:
 727                 if (code != 0)
 728                         goto badcode;
 729                 break;
 730
 731         case ND_ROUTER_SOLICIT:
 732                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
 733                 if (code != 0)
 734                         goto badcode;
 735                 if (icmp6len < sizeof(struct nd_router_solicit))
 736                         goto badlen;
 737                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 738                         /* give up local */
 739                         nd6_rs_input(m, off, icmp6len);
 740                         m = NULL;
 741                         goto freeit;
 742                 }
 743                 nd6_rs_input(n, off, icmp6len);
 744                 /* m stays. */
 745                 break;
 746
 747         case ND_ROUTER_ADVERT:
 748                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
 749                 if (code != 0)
 750                         goto badcode;
 751                 if (icmp6len < sizeof(struct nd_router_advert))
 752                         goto badlen;
 753                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 754                         /* give up local */
 755                         nd6_ra_input(m, off, icmp6len);
 756                         m = NULL;
 757                         goto freeit;
 758                 }
 759                 nd6_ra_input(n, off, icmp6len);
 760                 /* m stays. */
 761                 break;
 762
 763         case ND_NEIGHBOR_SOLICIT:
 764                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
 765                 if (code != 0)
 766                         goto badcode;
 767                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 768                         goto badlen;
 769                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 770                         /* give up local */
 771                         nd6_ns_input(m, off, icmp6len);
 772                         m = NULL;
 773                         goto freeit;
 774                 }
 775                 nd6_ns_input(n, off, icmp6len);
 776                 /* m stays. */
 777                 break;
 778
 779         case ND_NEIGHBOR_ADVERT:
 780                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
 781                 if (code != 0)
 782                         goto badcode;
 783                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 784                         goto badlen;
 785                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 786                         /* give up local */
 787                         nd6_na_input(m, off, icmp6len);
 788                         m = NULL;
 789                         goto freeit;
 790                 }
 791                 nd6_na_input(n, off, icmp6len);
 792                 /* m stays. */
 793                 break;
 794
 795         case ND_REDIRECT:
 796                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
 797                 if (code != 0)
 798                         goto badcode;
 799                 if (icmp6len < sizeof(struct nd_redirect))
 800                         goto badlen;
 801                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 802                         /* give up local */
 803                         icmp6_redirect_input(m, off);
 804                         m = NULL;
 805                         goto freeit;
 806                 }
 807                 icmp6_redirect_input(n, off);
 808                 /* m stays. */
 809                 break;
 810
 811         case ICMP6_ROUTER_RENUMBERING:
 812                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 813                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 814                         goto badcode;
 815                 if (icmp6len < sizeof(struct icmp6_router_renum))
 816                         goto badlen;
 817                 break;
 818
 819         default:
 820                 nd6log((LOG_DEBUG,
 821                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 822                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
 823                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
 824                     m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
 825                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 826                         /* ICMPv6 error: MUST deliver it by spec... */
 827                         code = PRC_NCMDS;
 828                         /* deliver */
 829                 } else {
 830                         /* ICMPv6 informational: MUST not deliver */
 831                         break;
 832                 }
 833         deliver:
 834                 if (icmp6_notify_error(&m, off, icmp6len, code)) {
 835                         /* In this case, m should've been freed. */
 836                         return (IPPROTO_DONE);
 837                 }
 838                 break;
 839
 840         badcode:
 841                 icmp6stat.icp6s_badcode++;
 842                 break;
 843
 844         badlen:
 845                 icmp6stat.icp6s_badlen++;
 846                 break;
 847         }
 848
 849         /* deliver the packet to appropriate sockets */
 850         icmp6_rip6_input(&m, *offp);
 851
 852         return IPPROTO_DONE;
 853
 854  freeit:
 855         m_freem(m);
 856         return IPPROTO_DONE;
 857 }
 858
 859 static int
 860 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
 861 {
 862         struct mbuf *m = *mp;
 863         struct icmp6_hdr *icmp6;
 864         struct ip6_hdr *eip6;
 865         u_int32_t notifymtu;
 866         struct sockaddr_in6 icmp6src, icmp6dst;
 867
 868         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 869                 icmp6stat.icp6s_tooshort++;
 870                 goto freeit;
 871         }
 872 #ifndef PULLDOWN_TEST
 873         IP6_EXTHDR_CHECK(m, off,
 874             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
 875         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 876 #else
 877         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
 878             sizeof(*icmp6) + sizeof(struct ip6_hdr));
 879         if (icmp6 == NULL) {
 880                 icmp6stat.icp6s_tooshort++;
 881                 return (-1);
 882         }
 883 #endif
 884         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 885
 886         /* Detect the upper level protocol */
 887         {
 888                 void (*ctlfunc)(int, struct sockaddr *, void *);
 889                 u_int8_t nxt = eip6->ip6_nxt;
 890                 int eoff = off + sizeof(struct icmp6_hdr) +
 891                     sizeof(struct ip6_hdr);
 892                 struct ip6ctlparam ip6cp;
 893                 struct in6_addr *finaldst = NULL;
 894                 int icmp6type = icmp6->icmp6_type;
 895                 struct ip6_frag *fh;
 896                 struct ip6_rthdr *rth;
 897                 struct ip6_rthdr0 *rth0;
 898                 int rthlen;
 899
 900                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 901                         struct ip6_ext *eh;
 902
 903                         switch (nxt) {
 904                         case IPPROTO_HOPOPTS:
 905                         case IPPROTO_DSTOPTS:
 906                         case IPPROTO_AH:
 907 #ifndef PULLDOWN_TEST
 908                                 IP6_EXTHDR_CHECK(m, 0,
 909                                     eoff + sizeof(struct ip6_ext), -1);
 910                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
 911 #else
 912                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
 913                                     eoff, sizeof(*eh));
 914                                 if (eh == NULL) {
 915                                         icmp6stat.icp6s_tooshort++;
 916                                         return (-1);
 917                                 }
 918 #endif
 919
 920                                 if (nxt == IPPROTO_AH)
 921                                         eoff += (eh->ip6e_len + 2) << 2;
 922                                 else
 923                                         eoff += (eh->ip6e_len + 1) << 3;
 924                                 nxt = eh->ip6e_nxt;
 925                                 break;
 926                         case IPPROTO_ROUTING:
 927                                 /*
 928                                  * When the erroneous packet contains a
 929                                  * routing header, we should examine the
 930                                  * header to determine the final destination.
 931                                  * Otherwise, we can't properly update
 932                                  * information that depends on the final
 933                                  * destination (e.g. path MTU).
 934                                  */
 935 #ifndef PULLDOWN_TEST
 936                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
 937                                 rth = (struct ip6_rthdr *)
 938                                     (mtod(m, caddr_t) + eoff);
 939 #else
 940                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
 941                                     eoff, sizeof(*rth));
 942                                 if (rth == NULL) {
 943                                         icmp6stat.icp6s_tooshort++;
 944                                         return (-1);
 945                                 }
 946 #endif
 947                                 rthlen = (rth->ip6r_len + 1) << 3;
 948                                 /*
 949                                  * XXX: currently there is no
 950                                  * officially defined type other
 951                                  * than type-0.
 952                                  * Note that if the segment left field
 953                                  * is 0, all intermediate hops must
 954                                  * have been passed.
 955                                  */
 956                                 if (rth->ip6r_segleft &&
 957                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
 958                                         int hops;
 959
 960 #ifndef PULLDOWN_TEST
 961                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
 962                                         rth0 = (struct ip6_rthdr0 *)
 963                                             (mtod(m, caddr_t) + eoff);
 964 #else
 965                                         IP6_EXTHDR_GET(rth0,
 966                                             struct ip6_rthdr0 *, m,
 967                                             eoff, rthlen);
 968                                         if (rth0 == NULL) {
 969                                                 icmp6stat.icp6s_tooshort++;
 970                                                 return (-1);
 971                                         }
 972 #endif
 973                                         /* just ignore a bogus header */
 974                                         if ((rth0->ip6r0_len % 2) == 0 &&
 975                                             (hops = rth0->ip6r0_len/2))
 976                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
 977                                 }
 978                                 eoff += rthlen;
 979                                 nxt = rth->ip6r_nxt;
 980                                 break;
 981                         case IPPROTO_FRAGMENT:
 982 #ifndef PULLDOWN_TEST
 983                                 IP6_EXTHDR_CHECK(m, 0, eoff +
 984                                     sizeof(struct ip6_frag), -1);
 985                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
 986                                     eoff);
 987 #else
 988                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
 989                                     eoff, sizeof(*fh));
 990                                 if (fh == NULL) {
 991                                         icmp6stat.icp6s_tooshort++;
 992                                         return (-1);
 993                                 }
 994 #endif
 995                                 /*
 996                                  * Data after a fragment header is meaningless
 997                                  * unless it is the first fragment, but
 998                                  * we'll go to the notify label for path MTU
 999                                  * discovery.
1000                                  */
1001                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1002                                         goto notify;
1003
1004                                 eoff += sizeof(struct ip6_frag);
1005                                 nxt = fh->ip6f_nxt;
1006                                 break;
1007                         default:
1008                                 /*
1009                                  * This case includes ESP and the No Next
1010                                  * Header.  In such cases going to the notify
1011                                  * label does not have any meaning
1012                                  * (i.e. ctlfunc will be NULL), but we go
1013                                  * anyway since we might have to update
1014                                  * path MTU information.
1015                                  */
1016                                 goto notify;
1017                         }
1018                 }
1019           notify:
1020 #ifndef PULLDOWN_TEST
1021                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1022 #else
1023                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1024                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1025                 if (icmp6 == NULL) {
1026                         icmp6stat.icp6s_tooshort++;
1027                         return (-1);
1028                 }
1029 #endif
1030
1031                 /*
1032                  * retrieve parameters from the inner IPv6 header, and convert
1033                  * them into sockaddr structures.
1034                  * XXX: there is no guarantee that the source or destination
1035                  * addresses of the inner packet are in the same scope as
1036                  * the addresses of the icmp packet.  But there is no other
1037                  * way to determine the zone.
1038                  */
1039                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1040
1041                 bzero(&icmp6dst, sizeof(icmp6dst));
1042                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1043                 icmp6dst.sin6_family = AF_INET6;
1044                 if (finaldst == NULL)
1045                         icmp6dst.sin6_addr = eip6->ip6_dst;
1046                 else
1047                         icmp6dst.sin6_addr = *finaldst;
1048                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1049                         goto freeit;
1050                 bzero(&icmp6src, sizeof(icmp6src));
1051                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1052                 icmp6src.sin6_family = AF_INET6;
1053                 icmp6src.sin6_addr = eip6->ip6_src;
1054                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1055                         goto freeit;
1056                 icmp6src.sin6_flowinfo =
1057                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1058
1059                 if (finaldst == NULL)
1060                         finaldst = &eip6->ip6_dst;
1061                 ip6cp.ip6c_m = m;
1062                 ip6cp.ip6c_icmp6 = icmp6;
1063                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1064                 ip6cp.ip6c_off = eoff;
1065                 ip6cp.ip6c_finaldst = finaldst;
1066                 ip6cp.ip6c_src = &icmp6src;
1067                 ip6cp.ip6c_nxt = nxt;
1068
1069                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1070                         notifymtu = ntohl(icmp6->icmp6_mtu);
1071                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1072                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1073                 }
1074
1075                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1076                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1077                 if (ctlfunc) {
1078                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1079                             &ip6cp);
1080                 }
1081         }
1082         *mp = m;
1083         return (0);
1084
1085   freeit:
1086         m_freem(m);
1087         return (-1);
1088 }
1089
1090 void
1091 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1092 {
1093         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1094         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1095         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1096         u_int mtu = ntohl(icmp6->icmp6_mtu);
1097         struct in_conninfo inc;
1098
1099 #if 0
1100         /*
1101          * RFC2460 section 5, last paragraph.
1102          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1103          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1104          * due to packet translator in the middle.
1105          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1106          * special handling.
1107          */
1108         if (mtu < IPV6_MMTU)
1109                 return;
1110 #endif
1111
1112         /*
1113          * we reject ICMPv6 too big with abnormally small value.
1114          * XXX what is the good definition of "abnormally small"?
1115          */
1116         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1117                 return;
1118
1119         if (!validated)
1120                 return;
1121
1122         /*
1123          * In case the suggested mtu is less than IPV6_MMTU, we
1124          * only need to remember that it was for above mentioned
1125          * "alwaysfrag" case.
1126          * Try to be as close to the spec as possible.
1127          */
1128         if (mtu < IPV6_MMTU)
1129                 mtu = IPV6_MMTU - 8;
1130
1131         bzero(&inc, sizeof(inc));
1132         inc.inc_flags |= INC_ISIPV6;
1133         inc.inc6_faddr = *dst;
1134         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1135                 return;
1136
1137         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1138                 tcp_hc_updatemtu(&inc, mtu);
1139                 icmp6stat.icp6s_pmtuchg++;
1140         }
1141 }
1142
1143 /*
1144  * Process a Node Information Query packet, based on
1145  * draft-ietf-ipngwg-icmp-name-lookups-07.
1146  *
1147  * Spec incompatibilities:
1148  * - IPv6 Subject address handling
1149  * - IPv4 Subject address handling support missing
1150  * - Proxy reply (answer even if it's not for me)
1151  * - joins NI group address at in6_ifattach() time only, does not cope
1152  *   with hostname changes by sethostname(3)
1153  */
1154 #define hostnamelen     strlen(hostname)
1155 static struct mbuf *
1156 ni6_input(struct mbuf *m, int off)
1157 {
1158         struct icmp6_nodeinfo *ni6, *nni6;
1159         struct mbuf *n = NULL;
1160         u_int16_t qtype;
1161         int subjlen;
1162         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1163         struct ni_reply_fqdn *fqdn;
1164         int addrs;              /* for NI_QTYPE_NODEADDR */
1165         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1166         struct in6_addr in6_subj; /* subject address */
1167         struct ip6_hdr *ip6;
1168         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1169         char *subj = NULL;
1170         struct in6_ifaddr *ia6 = NULL;
1171
1172         ip6 = mtod(m, struct ip6_hdr *);
1173 #ifndef PULLDOWN_TEST
1174         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1175 #else
1176         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1177         if (ni6 == NULL) {
1178                 /* m is already reclaimed */
1179                 return (NULL);
1180         }
1181 #endif
1182
1183         /*
1184          * Validate IPv6 source address.
1185          * The default configuration MUST be to refuse answering queries from
1186          * global-scope addresses according to RFC4602.
1187          * Notes:
1188          *  - it's not very clear what "refuse" means; this implementation
1189          *    simply drops it.
1190          *  - it's not very easy to identify global-scope (unicast) addresses
1191          *    since there are many prefixes for them.  It should be safer
1192          *    and in practice sufficient to check "all" but loopback and
1193          *    link-local (note that site-local unicast was deprecated and
1194          *    ULA is defined as global scope-wise)
1195          */
1196         if ((icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1197             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1198             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1199                 goto bad;
1200
1201         /*
1202          * Validate IPv6 destination address.
1203          *
1204          * The Responder must discard the Query without further processing
1205          * unless it is one of the Responder's unicast or anycast addresses, or
1206          * a link-local scope multicast address which the Responder has joined.
1207          * [RFC4602, Section 5.]
1208          */
1209         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1210                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1211                         goto bad;
1212                 /* else it's a link-local multicast, fine */
1213         } else {                /* unicast or anycast */
1214                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1215                         goto bad; /* XXX impossible */
1216
1217                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1218                     !(icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1219                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1220                                 "a temporary address in %s:%d",
1221                                __FILE__, __LINE__));
1222                         goto bad;
1223                 }
1224         }
1225
1226         /* validate query Subject field. */
1227         qtype = ntohs(ni6->ni_qtype);
1228         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1229         switch (qtype) {
1230         case NI_QTYPE_NOOP:
1231         case NI_QTYPE_SUPTYPES:
1232                 /* 07 draft */
1233                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1234                         break;
1235                 /* FALLTHROUGH */
1236         case NI_QTYPE_FQDN:
1237         case NI_QTYPE_NODEADDR:
1238         case NI_QTYPE_IPV4ADDR:
1239                 switch (ni6->ni_code) {
1240                 case ICMP6_NI_SUBJ_IPV6:
1241 #if ICMP6_NI_SUBJ_IPV6 != 0
1242                 case 0:
1243 #endif
1244                         /*
1245                          * backward compatibility - try to accept 03 draft
1246                          * format, where no Subject is present.
1247                          */
1248                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1249                             subjlen == 0) {
1250                                 oldfqdn++;
1251                                 break;
1252                         }
1253 #if ICMP6_NI_SUBJ_IPV6 != 0
1254                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1255                                 goto bad;
1256 #endif
1257
1258                         if (subjlen != sizeof(struct in6_addr))
1259                                 goto bad;
1260
1261                         /*
1262                          * Validate Subject address.
1263                          *
1264                          * Not sure what exactly "address belongs to the node"
1265                          * means in the spec, is it just unicast, or what?
1266                          *
1267                          * At this moment we consider Subject address as
1268                          * "belong to the node" if the Subject address equals
1269                          * to the IPv6 destination address; validation for
1270                          * IPv6 destination address should have done enough
1271                          * check for us.
1272                          *
1273                          * We do not do proxy at this moment.
1274                          */
1275                         /* m_pulldown instead of copy? */
1276                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1277                             subjlen, (caddr_t)&in6_subj);
1278                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1279                                 goto bad;
1280
1281                         subj = (char *)&in6_subj;
1282                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1283                                 break;
1284
1285                         /*
1286                          * XXX if we are to allow other cases, we should really
1287                          * be careful about scope here.
1288                          * basically, we should disallow queries toward IPv6
1289                          * destination X with subject Y,
1290                          * if scope(X) > scope(Y).
1291                          * if we allow scope(X) > scope(Y), it will result in
1292                          * information leakage across scope boundary.
1293                          */
1294                         goto bad;
1295
1296                 case ICMP6_NI_SUBJ_FQDN:
1297                         /*
1298                          * Validate Subject name with gethostname(3).
1299                          *
1300                          * The behavior may need some debate, since:
1301                          * - we are not sure if the node has FQDN as
1302                          *   hostname (returned by gethostname(3)).
1303                          * - the code does wildcard match for truncated names.
1304                          *   however, we are not sure if we want to perform
1305                          *   wildcard match, if gethostname(3) side has
1306                          *   truncated hostname.
1307                          */
1308                         n = ni6_nametodns(hostname, hostnamelen, 0);
1309                         if (!n || n->m_next || n->m_len == 0)
1310                                 goto bad;
1311                         IP6_EXTHDR_GET(subj, char *, m,
1312                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1313                         if (subj == NULL)
1314                                 goto bad;
1315                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1316                             n->m_len)) {
1317                                 goto bad;
1318                         }
1319                         m_freem(n);
1320                         n = NULL;
1321                         break;
1322
1323                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1324                 default:
1325                         goto bad;
1326                 }
1327                 break;
1328         }
1329
1330         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1331         switch (qtype) {
1332         case NI_QTYPE_FQDN:
1333                 if ((icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1334                         goto bad;
1335                 break;
1336         case NI_QTYPE_NODEADDR:
1337         case NI_QTYPE_IPV4ADDR:
1338                 if ((icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1339                         goto bad;
1340                 break;
1341         }
1342
1343         /* guess reply length */
1344         switch (qtype) {
1345         case NI_QTYPE_NOOP:
1346                 break;          /* no reply data */
1347         case NI_QTYPE_SUPTYPES:
1348                 replylen += sizeof(u_int32_t);
1349                 break;
1350         case NI_QTYPE_FQDN:
1351                 /* XXX will append an mbuf */
1352                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1353                 break;
1354         case NI_QTYPE_NODEADDR:
1355                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1356                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1357                     sizeof(u_int32_t))) > MCLBYTES)
1358                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1359                 break;
1360         case NI_QTYPE_IPV4ADDR:
1361                 /* unsupported - should respond with unknown Qtype? */
1362                 break;
1363         default:
1364                 /*
1365                  * XXX: We must return a reply with the ICMP6 code
1366                  * `unknown Qtype' in this case.  However we regard the case
1367                  * as an FQDN query for backward compatibility.
1368                  * Older versions set a random value to this field,
1369                  * so it rarely varies in the defined qtypes.
1370                  * But the mechanism is not reliable...
1371                  * maybe we should obsolete older versions.
1372                  */
1373                 qtype = NI_QTYPE_FQDN;
1374                 /* XXX will append an mbuf */
1375                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1376                 oldfqdn++;
1377                 break;
1378         }
1379
1380         /* allocate an mbuf to reply. */
1381         MGETHDR(n, M_DONTWAIT, m->m_type);
1382         if (n == NULL) {
1383                 m_freem(m);
1384                 return (NULL);
1385         }
1386         M_MOVE_PKTHDR(n, m); /* just for recvif */
1387         if (replylen > MHLEN) {
1388                 if (replylen > MCLBYTES) {
1389                         /*
1390                          * XXX: should we try to allocate more? But MCLBYTES
1391                          * is probably much larger than IPV6_MMTU...
1392                          */
1393                         goto bad;
1394                 }
1395                 MCLGET(n, M_DONTWAIT);
1396                 if ((n->m_flags & M_EXT) == 0) {
1397                         goto bad;
1398                 }
1399         }
1400         n->m_pkthdr.len = n->m_len = replylen;
1401
1402         /* copy mbuf header and IPv6 + Node Information base headers */
1403         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1404         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1405         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1406
1407         /* qtype dependent procedure */
1408         switch (qtype) {
1409         case NI_QTYPE_NOOP:
1410                 nni6->ni_code = ICMP6_NI_SUCCESS;
1411                 nni6->ni_flags = 0;
1412                 break;
1413         case NI_QTYPE_SUPTYPES:
1414         {
1415                 u_int32_t v;
1416                 nni6->ni_code = ICMP6_NI_SUCCESS;
1417                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1418                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1419                 v = (u_int32_t)htonl(0x0000000f);
1420                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1421                 break;
1422         }
1423         case NI_QTYPE_FQDN:
1424                 nni6->ni_code = ICMP6_NI_SUCCESS;
1425                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1426                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1427                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1428                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1429                 /*
1430                  * XXX do we really have FQDN in variable "hostname"?
1431                  */
1432                 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1433                 if (n->m_next == NULL)
1434                         goto bad;
1435                 /* XXX we assume that n->m_next is not a chain */
1436                 if (n->m_next->m_next != NULL)
1437                         goto bad;
1438                 n->m_pkthdr.len += n->m_next->m_len;
1439                 break;
1440         case NI_QTYPE_NODEADDR:
1441         {
1442                 int lenlim, copied;
1443
1444                 nni6->ni_code = ICMP6_NI_SUCCESS;
1445                 n->m_pkthdr.len = n->m_len =
1446                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1447                 lenlim = M_TRAILINGSPACE(n);
1448                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1449                 /* XXX: reset mbuf length */
1450                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1451                     sizeof(struct icmp6_nodeinfo) + copied;
1452                 break;
1453         }
1454         default:
1455                 break;          /* XXX impossible! */
1456         }
1457
1458         nni6->ni_type = ICMP6_NI_REPLY;
1459         m_freem(m);
1460         return (n);
1461
1462   bad:
1463         m_freem(m);
1464         if (n)
1465                 m_freem(n);
1466         return (NULL);
1467 }
1468 #undef hostnamelen
1469
1470 /*
1471  * make a mbuf with DNS-encoded string.  no compression support.
1472  *
1473  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1474  * treated as truncated name (two \0 at the end).  this is a wild guess.
1475  *
1476  * old - return pascal string if non-zero
1477  */
1478 static struct mbuf *
1479 ni6_nametodns(const char *name, int namelen, int old)
1480 {
1481         struct mbuf *m;
1482         char *cp, *ep;
1483         const char *p, *q;
1484         int i, len, nterm;
1485
1486         if (old)
1487                 len = namelen + 1;
1488         else
1489                 len = MCLBYTES;
1490
1491         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1492         MGET(m, M_DONTWAIT, MT_DATA);
1493         if (m && len > MLEN) {
1494                 MCLGET(m, M_DONTWAIT);
1495                 if ((m->m_flags & M_EXT) == 0)
1496                         goto fail;
1497         }
1498         if (!m)
1499                 goto fail;
1500         m->m_next = NULL;
1501
1502         if (old) {
1503                 m->m_len = len;
1504                 *mtod(m, char *) = namelen;
1505                 bcopy(name, mtod(m, char *) + 1, namelen);
1506                 return m;
1507         } else {
1508                 m->m_len = 0;
1509                 cp = mtod(m, char *);
1510                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1511
1512                 /* if not certain about my name, return empty buffer */
1513                 if (namelen == 0)
1514                         return m;
1515
1516                 /*
1517                  * guess if it looks like shortened hostname, or FQDN.
1518                  * shortened hostname needs two trailing "\0".
1519                  */
1520                 i = 0;
1521                 for (p = name; p < name + namelen; p++) {
1522                         if (*p && *p == '.')
1523                                 i++;
1524                 }
1525                 if (i < 2)
1526                         nterm = 2;
1527                 else
1528                         nterm = 1;
1529
1530                 p = name;
1531                 while (cp < ep && p < name + namelen) {
1532                         i = 0;
1533                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1534                                 i++;
1535                         /* result does not fit into mbuf */
1536                         if (cp + i + 1 >= ep)
1537                                 goto fail;
1538                         /*
1539                          * DNS label length restriction, RFC1035 page 8.
1540                          * "i == 0" case is included here to avoid returning
1541                          * 0-length label on "foo..bar".
1542                          */
1543                         if (i <= 0 || i >= 64)
1544                                 goto fail;
1545                         *cp++ = i;
1546                         bcopy(p, cp, i);
1547                         cp += i;
1548                         p = q;
1549                         if (p < name + namelen && *p == '.')
1550                                 p++;
1551                 }
1552                 /* termination */
1553                 if (cp + nterm >= ep)
1554                         goto fail;
1555                 while (nterm-- > 0)
1556                         *cp++ = '\0';
1557                 m->m_len = cp - mtod(m, char *);
1558                 return m;
1559         }
1560
1561         panic("should not reach here");
1562         /* NOTREACHED */
1563
1564  fail:
1565         if (m)
1566                 m_freem(m);
1567         return NULL;
1568 }
1569
1570 /*
1571  * check if two DNS-encoded string matches.  takes care of truncated
1572  * form (with \0\0 at the end).  no compression support.
1573  * XXX upper/lowercase match (see RFC2065)
1574  */
1575 static int
1576 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1577 {
1578         const char *a0, *b0;
1579         int l;
1580
1581         /* simplest case - need validation? */
1582         if (alen == blen && bcmp(a, b, alen) == 0)
1583                 return 1;
1584
1585         a0 = a;
1586         b0 = b;
1587
1588         /* termination is mandatory */
1589         if (alen < 2 || blen < 2)
1590                 return 0;
1591         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1592                 return 0;
1593         alen--;
1594         blen--;
1595
1596         while (a - a0 < alen && b - b0 < blen) {
1597                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1598                         return 0;
1599
1600                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1601                         return 0;
1602                 /* we don't support compression yet */
1603                 if (a[0] >= 64 || b[0] >= 64)
1604                         return 0;
1605
1606                 /* truncated case */
1607                 if (a[0] == 0 && a - a0 == alen - 1)
1608                         return 1;
1609                 if (b[0] == 0 && b - b0 == blen - 1)
1610                         return 1;
1611                 if (a[0] == 0 || b[0] == 0)
1612                         return 0;
1613
1614                 if (a[0] != b[0])
1615                         return 0;
1616                 l = a[0];
1617                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1618                         return 0;
1619                 if (bcmp(a + 1, b + 1, l) != 0)
1620                         return 0;
1621
1622                 a += 1 + l;
1623                 b += 1 + l;
1624         }
1625
1626         if (a - a0 == alen && b - b0 == blen)
1627                 return 1;
1628         else
1629                 return 0;
1630 }
1631
1632 /*
1633  * calculate the number of addresses to be returned in the node info reply.
1634  */
1635 static int
1636 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1637     struct in6_addr *subj)
1638 {
1639         struct ifnet *ifp;
1640         struct in6_ifaddr *ifa6;
1641         struct ifaddr *ifa;
1642         int addrs = 0, addrsofif, iffound = 0;
1643         int niflags = ni6->ni_flags;
1644
1645         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1646                 switch (ni6->ni_code) {
1647                 case ICMP6_NI_SUBJ_IPV6:
1648                         if (subj == NULL) /* must be impossible... */
1649                                 return (0);
1650                         break;
1651                 default:
1652                         /*
1653                          * XXX: we only support IPv6 subject address for
1654                          * this Qtype.
1655                          */
1656                         return (0);
1657                 }
1658         }
1659
1660         IFNET_RLOCK();
1661         for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1662                 addrsofif = 0;
1663                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1664                         if (ifa->ifa_addr->sa_family != AF_INET6)
1665                                 continue;
1666                         ifa6 = (struct in6_ifaddr *)ifa;
1667
1668                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1669                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1670                                 iffound = 1;
1671
1672                         /*
1673                          * IPv4-mapped addresses can only be returned by a
1674                          * Node Information proxy, since they represent
1675                          * addresses of IPv4-only nodes, which perforce do
1676                          * not implement this protocol.
1677                          * [icmp-name-lookups-07, Section 5.4]
1678                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1679                          * this function at this moment.
1680                          */
1681
1682                         /* What do we have to do about ::1? */
1683                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1684                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1685                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1686                                         continue;
1687                                 break;
1688                         case IPV6_ADDR_SCOPE_SITELOCAL:
1689                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1690                                         continue;
1691                                 break;
1692                         case IPV6_ADDR_SCOPE_GLOBAL:
1693                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1694                                         continue;
1695                                 break;
1696                         default:
1697                                 continue;
1698                         }
1699
1700                         /*
1701                          * check if anycast is okay.
1702                          * XXX: just experimental.  not in the spec.
1703                          */
1704                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1705                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1706                                 continue; /* we need only unicast addresses */
1707                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1708                             (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1709                                 continue;
1710                         }
1711                         addrsofif++; /* count the address */
1712                 }
1713                 if (iffound) {
1714                         *ifpp = ifp;
1715                         IFNET_RUNLOCK();
1716                         return (addrsofif);
1717                 }
1718
1719                 addrs += addrsofif;
1720         }
1721         IFNET_RUNLOCK();
1722
1723         return (addrs);
1724 }
1725
1726 static int
1727 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1728     struct ifnet *ifp0, int resid)
1729 {
1730         struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1731         struct in6_ifaddr *ifa6;
1732         struct ifaddr *ifa;
1733         struct ifnet *ifp_dep = NULL;
1734         int copied = 0, allow_deprecated = 0;
1735         u_char *cp = (u_char *)(nni6 + 1);
1736         int niflags = ni6->ni_flags;
1737         u_int32_t ltime;
1738
1739         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1740                 return (0);     /* needless to copy */
1741
1742         IFNET_RLOCK();
1743   again:
1744
1745         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1746                 for (ifa = ifp->if_addrlist.tqh_first; ifa;
1747                      ifa = ifa->ifa_list.tqe_next) {
1748                         if (ifa->ifa_addr->sa_family != AF_INET6)
1749                                 continue;
1750                         ifa6 = (struct in6_ifaddr *)ifa;
1751
1752                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1753                             allow_deprecated == 0) {
1754                                 /*
1755                                  * prefererred address should be put before
1756                                  * deprecated addresses.
1757                                  */
1758
1759                                 /* record the interface for later search */
1760                                 if (ifp_dep == NULL)
1761                                         ifp_dep = ifp;
1762
1763                                 continue;
1764                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1765                             allow_deprecated != 0)
1766                                 continue; /* we now collect deprecated addrs */
1767
1768                         /* What do we have to do about ::1? */
1769                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1770                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1771                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1772                                         continue;
1773                                 break;
1774                         case IPV6_ADDR_SCOPE_SITELOCAL:
1775                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1776                                         continue;
1777                                 break;
1778                         case IPV6_ADDR_SCOPE_GLOBAL:
1779                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1780                                         continue;
1781                                 break;
1782                         default:
1783                                 continue;
1784                         }
1785
1786                         /*
1787                          * check if anycast is okay.
1788                          * XXX: just experimental.  not in the spec.
1789                          */
1790                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1791                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1792                                 continue;
1793                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1794                             (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1795                                 continue;
1796                         }
1797
1798                         /* now we can copy the address */
1799                         if (resid < sizeof(struct in6_addr) +
1800                             sizeof(u_int32_t)) {
1801                                 /*
1802                                  * We give up much more copy.
1803                                  * Set the truncate flag and return.
1804                                  */
1805                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1806                                 IFNET_RUNLOCK();
1807                                 return (copied);
1808                         }
1809
1810                         /*
1811                          * Set the TTL of the address.
1812                          * The TTL value should be one of the following
1813                          * according to the specification:
1814                          *
1815                          * 1. The remaining lifetime of a DHCP lease on the
1816                          *    address, or
1817                          * 2. The remaining Valid Lifetime of a prefix from
1818                          *    which the address was derived through Stateless
1819                          *    Autoconfiguration.
1820                          *
1821                          * Note that we currently do not support stateful
1822                          * address configuration by DHCPv6, so the former
1823                          * case can't happen.
1824                          */
1825                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1826                                 ltime = ND6_INFINITE_LIFETIME;
1827                         else {
1828                                 if (ifa6->ia6_lifetime.ia6t_expire >
1829                                     time_second)
1830                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1831                                 else
1832                                         ltime = 0;
1833                         }
1834
1835                         bcopy(&ltime, cp, sizeof(u_int32_t));
1836                         cp += sizeof(u_int32_t);
1837
1838                         /* copy the address itself */
1839                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1840                             sizeof(struct in6_addr));
1841                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1842                         cp += sizeof(struct in6_addr);
1843
1844                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1845                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1846                 }
1847                 if (ifp0)       /* we need search only on the specified IF */
1848                         break;
1849         }
1850
1851         if (allow_deprecated == 0 && ifp_dep != NULL) {
1852                 ifp = ifp_dep;
1853                 allow_deprecated = 1;
1854
1855                 goto again;
1856         }
1857
1858         IFNET_RUNLOCK();
1859
1860         return (copied);
1861 }
1862
1863 /*
1864  * XXX almost dup'ed code with rip6_input.
1865  */
1866 static int
1867 icmp6_rip6_input(struct mbuf **mp, int off)
1868 {
1869         struct mbuf *m = *mp;
1870         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1871         struct inpcb *in6p;
1872         struct inpcb *last = NULL;
1873         struct sockaddr_in6 fromsa;
1874         struct icmp6_hdr *icmp6;
1875         struct mbuf *opts = NULL;
1876
1877 #ifndef PULLDOWN_TEST
1878         /* this is assumed to be safe. */
1879         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1880 #else
1881         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1882         if (icmp6 == NULL) {
1883                 /* m is already reclaimed */
1884                 return (IPPROTO_DONE);
1885         }
1886 #endif
1887
1888         /*
1889          * XXX: the address may have embedded scope zone ID, which should be
1890          * hidden from applications.
1891          */
1892         bzero(&fromsa, sizeof(fromsa));
1893         fromsa.sin6_family = AF_INET6;
1894         fromsa.sin6_len = sizeof(struct sockaddr_in6);
1895         fromsa.sin6_addr = ip6->ip6_src;
1896         if (sa6_recoverscope(&fromsa)) {
1897                 m_freem(m);
1898                 return (IPPROTO_DONE);
1899         }
1900
1901         INP_INFO_RLOCK(&ripcbinfo);
1902         LIST_FOREACH(in6p, &ripcb, inp_list) {
1903                 if ((in6p->inp_vflag & INP_IPV6) == 0)
1904                         continue;
1905                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
1906                         continue;
1907                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1908                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1909                         continue;
1910                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1911                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1912                         continue;
1913                 INP_RLOCK(in6p);
1914                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1915                     in6p->in6p_icmp6filt)) {
1916                         INP_RUNLOCK(in6p);
1917                         continue;
1918                 }
1919                 if (last) {
1920                         struct  mbuf *n = NULL;
1921
1922                         /*
1923                          * Recent network drivers tend to allocate a single
1924                          * mbuf cluster, rather than to make a couple of
1925                          * mbufs without clusters.  Also, since the IPv6 code
1926                          * path tries to avoid m_pullup(), it is highly
1927                          * probable that we still have an mbuf cluster here
1928                          * even though the necessary length can be stored in an
1929                          * mbuf's internal buffer.
1930                          * Meanwhile, the default size of the receive socket
1931                          * buffer for raw sockets is not so large.  This means
1932                          * the possibility of packet loss is relatively higher
1933                          * than before.  To avoid this scenario, we copy the
1934                          * received data to a separate mbuf that does not use
1935                          * a cluster, if possible.
1936                          * XXX: it is better to copy the data after stripping
1937                          * intermediate headers.
1938                          */
1939                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1940                             m->m_len <= MHLEN) {
1941                                 MGET(n, M_DONTWAIT, m->m_type);
1942                                 if (n != NULL) {
1943                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1944                                                 bcopy(m->m_data, n->m_data,
1945                                                       m->m_len);
1946                                                 n->m_len = m->m_len;
1947                                         } else {
1948                                                 m_free(n);
1949                                                 n = NULL;
1950                                         }
1951                                 }
1952                         }
1953                         if (n != NULL ||
1954                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1955                                 if (last->inp_flags & INP_CONTROLOPTS)
1956                                         ip6_savecontrol(last, n, &opts);
1957                                 /* strip intermediate headers */
1958                                 m_adj(n, off);
1959                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1960                                 if (sbappendaddr_locked(
1961                                     &last->inp_socket->so_rcv,
1962                                     (struct sockaddr *)&fromsa, n, opts)
1963                                     == 0) {
1964                                         /* should notify about lost packet */
1965                                         m_freem(n);
1966                                         if (opts) {
1967                                                 m_freem(opts);
1968                                         }
1969                                         SOCKBUF_UNLOCK(
1970                                             &last->inp_socket->so_rcv);
1971                                 } else
1972                                         sorwakeup_locked(last->inp_socket);
1973                                 opts = NULL;
1974                         }
1975                         INP_RUNLOCK(last);
1976                 }
1977                 last = in6p;
1978         }
1979         INP_INFO_RUNLOCK(&ripcbinfo);
1980         if (last) {
1981                 if (last->inp_flags & INP_CONTROLOPTS)
1982                         ip6_savecontrol(last, m, &opts);
1983                 /* strip intermediate headers */
1984                 m_adj(m, off);
1985
1986                 /* avoid using mbuf clusters if possible (see above) */
1987                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1988                     m->m_len <= MHLEN) {
1989                         struct mbuf *n;
1990
1991                         MGET(n, M_DONTWAIT, m->m_type);
1992                         if (n != NULL) {
1993                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1994                                         bcopy(m->m_data, n->m_data, m->m_len);
1995                                         n->m_len = m->m_len;
1996
1997                                         m_freem(m);
1998                                         m = n;
1999                                 } else {
2000                                         m_freem(n);
2001                                         n = NULL;
2002                                 }
2003                         }
2004                 }
2005                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2006                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2007                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2008                         m_freem(m);
2009                         if (opts)
2010                                 m_freem(opts);
2011                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2012                 } else
2013                         sorwakeup_locked(last->inp_socket);
2014                 INP_RUNLOCK(last);
2015         } else {
2016                 m_freem(m);
2017                 ip6stat.ip6s_delivered--;
2018         }
2019         return IPPROTO_DONE;
2020 }
2021
2022 /*
2023  * Reflect the ip6 packet back to the source.
2024  * OFF points to the icmp6 header, counted from the top of the mbuf.
2025  */
2026 void
2027 icmp6_reflect(struct mbuf *m, size_t off)
2028 {
2029         struct ip6_hdr *ip6;
2030         struct icmp6_hdr *icmp6;
2031         struct in6_ifaddr *ia;
2032         int plen;
2033         int type, code;
2034         struct ifnet *outif = NULL;
2035         struct in6_addr origdst, *src = NULL;
2036
2037         /* too short to reflect */
2038         if (off < sizeof(struct ip6_hdr)) {
2039                 nd6log((LOG_DEBUG,
2040                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2041                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2042                     __FILE__, __LINE__));
2043                 goto bad;
2044         }
2045
2046         /*
2047          * If there are extra headers between IPv6 and ICMPv6, strip
2048          * off that header first.
2049          */
2050 #ifdef DIAGNOSTIC
2051         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2052                 panic("assumption failed in icmp6_reflect");
2053 #endif
2054         if (off > sizeof(struct ip6_hdr)) {
2055                 size_t l;
2056                 struct ip6_hdr nip6;
2057
2058                 l = off - sizeof(struct ip6_hdr);
2059                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2060                 m_adj(m, l);
2061                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2062                 if (m->m_len < l) {
2063                         if ((m = m_pullup(m, l)) == NULL)
2064                                 return;
2065                 }
2066                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2067         } else /* off == sizeof(struct ip6_hdr) */ {
2068                 size_t l;
2069                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2070                 if (m->m_len < l) {
2071                         if ((m = m_pullup(m, l)) == NULL)
2072                                 return;
2073                 }
2074         }
2075         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2076         ip6 = mtod(m, struct ip6_hdr *);
2077         ip6->ip6_nxt = IPPROTO_ICMPV6;
2078         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2079         type = icmp6->icmp6_type; /* keep type for statistics */
2080         code = icmp6->icmp6_code; /* ditto. */
2081
2082         origdst = ip6->ip6_dst;
2083         /*
2084          * ip6_input() drops a packet if its src is multicast.
2085          * So, the src is never multicast.
2086          */
2087         ip6->ip6_dst = ip6->ip6_src;
2088
2089         /*
2090          * If the incoming packet was addressed directly to us (i.e. unicast),
2091          * use dst as the src for the reply.
2092          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2093          * (for example) when we encounter an error while forwarding procedure
2094          * destined to a duplicated address of ours.
2095          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2096          * procedure of an outgoing packet of our own, in which case we need
2097          * to search in the ifaddr list.
2098          */
2099         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2100                 if ((ia = ip6_getdstifaddr(m))) {
2101                         if (!(ia->ia6_flags &
2102                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2103                                 src = &ia->ia_addr.sin6_addr;
2104                 } else {
2105                         struct sockaddr_in6 d;
2106
2107                         bzero(&d, sizeof(d));
2108                         d.sin6_family = AF_INET6;
2109                         d.sin6_len = sizeof(d);
2110                         d.sin6_addr = origdst;
2111                         ia = (struct in6_ifaddr *)
2112                             ifa_ifwithaddr((struct sockaddr *)&d);
2113                         if (ia &&
2114                             !(ia->ia6_flags &
2115                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2116                                 src = &ia->ia_addr.sin6_addr;
2117                         }
2118                 }
2119         }
2120
2121         if (src == NULL) {
2122                 int e;
2123                 struct sockaddr_in6 sin6;
2124                 struct route_in6 ro;
2125
2126                 /*
2127                  * This case matches to multicasts, our anycast, or unicasts
2128                  * that we do not own.  Select a source address based on the
2129                  * source address of the erroneous packet.
2130                  */
2131                 bzero(&sin6, sizeof(sin6));
2132                 sin6.sin6_family = AF_INET6;
2133                 sin6.sin6_len = sizeof(sin6);
2134                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2135
2136                 bzero(&ro, sizeof(ro));
2137                 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2138                 if (ro.ro_rt)
2139                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2140                 if (src == NULL) {
2141                         char ip6buf[INET6_ADDRSTRLEN];
2142                         nd6log((LOG_DEBUG,
2143                             "icmp6_reflect: source can't be determined: "
2144                             "dst=%s, error=%d\n",
2145                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2146                         goto bad;
2147                 }
2148         }
2149
2150         ip6->ip6_src = *src;
2151         ip6->ip6_flow = 0;
2152         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2153         ip6->ip6_vfc |= IPV6_VERSION;
2154         ip6->ip6_nxt = IPPROTO_ICMPV6;
2155         if (outif)
2156                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2157         else if (m->m_pkthdr.rcvif) {
2158                 /* XXX: This may not be the outgoing interface */
2159                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2160         } else
2161                 ip6->ip6_hlim = ip6_defhlim;
2162
2163         icmp6->icmp6_cksum = 0;
2164         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2165             sizeof(struct ip6_hdr), plen);
2166
2167         /*
2168          * XXX option handling
2169          */
2170
2171         m->m_flags &= ~(M_BCAST|M_MCAST);
2172
2173         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2174         if (outif)
2175                 icmp6_ifoutstat_inc(outif, type, code);
2176
2177         return;
2178
2179  bad:
2180         m_freem(m);
2181         return;
2182 }
2183
2184 void
2185 icmp6_fasttimo(void)
2186 {
2187
2188         return;
2189 }
2190
2191 static const char *
2192 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2193     struct in6_addr *tgt6)
2194 {
2195         static char buf[1024];
2196         char ip6bufs[INET6_ADDRSTRLEN];
2197         char ip6bufd[INET6_ADDRSTRLEN];
2198         char ip6buft[INET6_ADDRSTRLEN];
2199         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2200             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2201             ip6_sprintf(ip6buft, tgt6));
2202         return buf;
2203 }
2204
2205 void
2206 icmp6_redirect_input(struct mbuf *m, int off)
2207 {
2208         struct ifnet *ifp;
2209         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2210         struct nd_redirect *nd_rd;
2211         int icmp6len = ntohs(ip6->ip6_plen);
2212         char *lladdr = NULL;
2213         int lladdrlen = 0;
2214         u_char *redirhdr = NULL;
2215         int redirhdrlen = 0;
2216         struct rtentry *rt = NULL;
2217         int is_router;
2218         int is_onlink;
2219         struct in6_addr src6 = ip6->ip6_src;
2220         struct in6_addr redtgt6;
2221         struct in6_addr reddst6;
2222         union nd_opts ndopts;
2223         char ip6buf[INET6_ADDRSTRLEN];
2224
2225         if (!m)
2226                 return;
2227
2228         ifp = m->m_pkthdr.rcvif;
2229
2230         if (!ifp)
2231                 return;
2232
2233         /* XXX if we are router, we don't update route by icmp6 redirect */
2234         if (ip6_forwarding)
2235                 goto freeit;
2236         if (!icmp6_rediraccept)
2237                 goto freeit;
2238
2239 #ifndef PULLDOWN_TEST
2240         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2241         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2242 #else
2243         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2244         if (nd_rd == NULL) {
2245                 icmp6stat.icp6s_tooshort++;
2246                 return;
2247         }
2248 #endif
2249         redtgt6 = nd_rd->nd_rd_target;
2250         reddst6 = nd_rd->nd_rd_dst;
2251
2252         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2253             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2254                 goto freeit;
2255         }
2256
2257         /* validation */
2258         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2259                 nd6log((LOG_ERR,
2260                     "ICMP6 redirect sent from %s rejected; "
2261                     "must be from linklocal\n",
2262                     ip6_sprintf(ip6buf, &src6)));
2263                 goto bad;
2264         }
2265         if (ip6->ip6_hlim != 255) {
2266                 nd6log((LOG_ERR,
2267                     "ICMP6 redirect sent from %s rejected; "
2268                     "hlim=%d (must be 255)\n",
2269                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2270                 goto bad;
2271         }
2272     {
2273         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2274         struct sockaddr_in6 sin6;
2275         struct in6_addr *gw6;
2276
2277         bzero(&sin6, sizeof(sin6));
2278         sin6.sin6_family = AF_INET6;
2279         sin6.sin6_len = sizeof(struct sockaddr_in6);
2280         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2281         rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2282         if (rt) {
2283                 if (rt->rt_gateway == NULL ||
2284                     rt->rt_gateway->sa_family != AF_INET6) {
2285                         nd6log((LOG_ERR,
2286                             "ICMP6 redirect rejected; no route "
2287                             "with inet6 gateway found for redirect dst: %s\n",
2288                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2289                         RTFREE_LOCKED(rt);
2290                         goto bad;
2291                 }
2292
2293                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2294                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2295                         nd6log((LOG_ERR,
2296                             "ICMP6 redirect rejected; "
2297                             "not equal to gw-for-src=%s (must be same): "
2298                             "%s\n",
2299                             ip6_sprintf(ip6buf, gw6),
2300                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2301                         RTFREE_LOCKED(rt);
2302                         goto bad;
2303                 }
2304         } else {
2305                 nd6log((LOG_ERR,
2306                     "ICMP6 redirect rejected; "
2307                     "no route found for redirect dst: %s\n",
2308                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2309                 goto bad;
2310         }
2311         RTFREE_LOCKED(rt);
2312         rt = NULL;
2313     }
2314         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2315                 nd6log((LOG_ERR,
2316                     "ICMP6 redirect rejected; "
2317                     "redirect dst must be unicast: %s\n",
2318                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2319                 goto bad;
2320         }
2321
2322         is_router = is_onlink = 0;
2323         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2324                 is_router = 1;  /* router case */
2325         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2326                 is_onlink = 1;  /* on-link destination case */
2327         if (!is_router && !is_onlink) {
2328                 nd6log((LOG_ERR,
2329                     "ICMP6 redirect rejected; "
2330                     "neither router case nor onlink case: %s\n",
2331                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2332                 goto bad;
2333         }
2334         /* validation passed */
2335
2336         icmp6len -= sizeof(*nd_rd);
2337         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2338         if (nd6_options(&ndopts) < 0) {
2339                 nd6log((LOG_INFO, "icmp6_redirect_input: "
2340                     "invalid ND option, rejected: %s\n",
2341                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2342                 /* nd6_options have incremented stats */
2343                 goto freeit;
2344         }
2345
2346         if (ndopts.nd_opts_tgt_lladdr) {
2347                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2348                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2349         }
2350
2351         if (ndopts.nd_opts_rh) {
2352                 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2353                 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2354         }
2355
2356         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2357                 nd6log((LOG_INFO,
2358                     "icmp6_redirect_input: lladdrlen mismatch for %s "
2359                     "(if %d, icmp6 packet %d): %s\n",
2360                     ip6_sprintf(ip6buf, &redtgt6),
2361                     ifp->if_addrlen, lladdrlen - 2,
2362                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2363                 goto bad;
2364         }
2365
2366         /* RFC 2461 8.3 */
2367         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2368             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2369
2370         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2371                 /* perform rtredirect */
2372                 struct sockaddr_in6 sdst;
2373                 struct sockaddr_in6 sgw;
2374                 struct sockaddr_in6 ssrc;
2375
2376                 bzero(&sdst, sizeof(sdst));
2377                 bzero(&sgw, sizeof(sgw));
2378                 bzero(&ssrc, sizeof(ssrc));
2379                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2380                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2381                         sizeof(struct sockaddr_in6);
2382                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2383                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2384                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2385                 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2386                     (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2387                     (struct sockaddr *)&ssrc);
2388         }
2389         /* finally update cached route in each socket via pfctlinput */
2390     {
2391         struct sockaddr_in6 sdst;
2392
2393         bzero(&sdst, sizeof(sdst));
2394         sdst.sin6_family = AF_INET6;
2395         sdst.sin6_len = sizeof(struct sockaddr_in6);
2396         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2397         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2398 #ifdef IPSEC
2399         key_sa_routechange((struct sockaddr *)&sdst);
2400 #endif /* IPSEC */
2401     }
2402
2403  freeit:
2404         m_freem(m);
2405         return;
2406
2407  bad:
2408         icmp6stat.icp6s_badredirect++;
2409         m_freem(m);
2410 }
2411
2412 void
2413 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2414 {
2415         struct ifnet *ifp;      /* my outgoing interface */
2416         struct in6_addr *ifp_ll6;
2417         struct in6_addr *router_ll6;
2418         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2419         struct mbuf *m = NULL;  /* newly allocated one */
2420         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2421         struct nd_redirect *nd_rd;
2422         size_t maxlen;
2423         u_char *p;
2424         struct ifnet *outif = NULL;
2425         struct sockaddr_in6 src_sa;
2426
2427         icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2428
2429         /* if we are not router, we don't send icmp6 redirect */
2430         if (!ip6_forwarding)
2431                 goto fail;
2432
2433         /* sanity check */
2434         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2435                 goto fail;
2436
2437         /*
2438          * Address check:
2439          *  the source address must identify a neighbor, and
2440          *  the destination address must not be a multicast address
2441          *  [RFC 2461, sec 8.2]
2442          */
2443         sip6 = mtod(m0, struct ip6_hdr *);
2444         bzero(&src_sa, sizeof(src_sa));
2445         src_sa.sin6_family = AF_INET6;
2446         src_sa.sin6_len = sizeof(src_sa);
2447         src_sa.sin6_addr = sip6->ip6_src;
2448         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2449                 goto fail;
2450         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2451                 goto fail;      /* what should we do here? */
2452
2453         /* rate limit */
2454         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2455                 goto fail;
2456
2457         /*
2458          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2459          * we almost always ask for an mbuf cluster for simplicity.
2460          * (MHLEN < IPV6_MMTU is almost always true)
2461          */
2462 #if IPV6_MMTU >= MCLBYTES
2463 # error assumption failed about IPV6_MMTU and MCLBYTES
2464 #endif
2465         MGETHDR(m, M_DONTWAIT, MT_HEADER);
2466         if (m && IPV6_MMTU >= MHLEN)
2467                 MCLGET(m, M_DONTWAIT);
2468         if (!m)
2469                 goto fail;
2470         m->m_pkthdr.rcvif = NULL;
2471         m->m_len = 0;
2472         maxlen = M_TRAILINGSPACE(m);
2473         maxlen = min(IPV6_MMTU, maxlen);
2474         /* just for safety */
2475         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2476             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2477                 goto fail;
2478         }
2479
2480         {
2481                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2482                 struct in6_ifaddr *ia;
2483                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2484                                                  IN6_IFF_NOTREADY|
2485                                                  IN6_IFF_ANYCAST)) == NULL)
2486                         goto fail;
2487                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2488         }
2489
2490         /* get ip6 linklocal address for the router. */
2491         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2492                 struct sockaddr_in6 *sin6;
2493                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2494                 router_ll6 = &sin6->sin6_addr;
2495                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2496                         router_ll6 = (struct in6_addr *)NULL;
2497         } else
2498                 router_ll6 = (struct in6_addr *)NULL;
2499
2500         /* ip6 */
2501         ip6 = mtod(m, struct ip6_hdr *);
2502         ip6->ip6_flow = 0;
2503         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2504         ip6->ip6_vfc |= IPV6_VERSION;
2505         /* ip6->ip6_plen will be set later */
2506         ip6->ip6_nxt = IPPROTO_ICMPV6;
2507         ip6->ip6_hlim = 255;
2508         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2509         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2510         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2511
2512         /* ND Redirect */
2513         nd_rd = (struct nd_redirect *)(ip6 + 1);
2514         nd_rd->nd_rd_type = ND_REDIRECT;
2515         nd_rd->nd_rd_code = 0;
2516         nd_rd->nd_rd_reserved = 0;
2517         if (rt->rt_flags & RTF_GATEWAY) {
2518                 /*
2519                  * nd_rd->nd_rd_target must be a link-local address in
2520                  * better router cases.
2521                  */
2522                 if (!router_ll6)
2523                         goto fail;
2524                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2525                     sizeof(nd_rd->nd_rd_target));
2526                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2527                     sizeof(nd_rd->nd_rd_dst));
2528         } else {
2529                 /* make sure redtgt == reddst */
2530                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2531                     sizeof(nd_rd->nd_rd_target));
2532                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2533                     sizeof(nd_rd->nd_rd_dst));
2534         }
2535
2536         p = (u_char *)(nd_rd + 1);
2537
2538         if (!router_ll6)
2539                 goto nolladdropt;
2540
2541         {
2542                 /* target lladdr option */
2543                 struct rtentry *rt_router = NULL;
2544                 int len;
2545                 struct sockaddr_dl *sdl;
2546                 struct nd_opt_hdr *nd_opt;
2547                 char *lladdr;
2548
2549                 rt_router = nd6_lookup(router_ll6, 0, ifp);
2550                 if (!rt_router)
2551                         goto nolladdropt;
2552                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2553                 len = (len + 7) & ~7;   /* round by 8 */
2554                 /* safety check */
2555                 if (len + (p - (u_char *)ip6) > maxlen)
2556                         goto nolladdropt;
2557                 if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2558                     (rt_router->rt_flags & RTF_LLINFO) &&
2559                     (rt_router->rt_gateway->sa_family == AF_LINK) &&
2560                     (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2561                     sdl->sdl_alen) {
2562                         nd_opt = (struct nd_opt_hdr *)p;
2563                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2564                         nd_opt->nd_opt_len = len >> 3;
2565                         lladdr = (char *)(nd_opt + 1);
2566                         bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2567                         p += len;
2568                 }
2569         }
2570 nolladdropt:;
2571
2572         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2573
2574         /* just to be safe */
2575 #ifdef M_DECRYPTED      /*not openbsd*/
2576         if (m0->m_flags & M_DECRYPTED)
2577                 goto noredhdropt;
2578 #endif
2579         if (p - (u_char *)ip6 > maxlen)
2580                 goto noredhdropt;
2581
2582         {
2583                 /* redirected header option */
2584                 int len;
2585                 struct nd_opt_rd_hdr *nd_opt_rh;
2586
2587                 /*
2588                  * compute the maximum size for icmp6 redirect header option.
2589                  * XXX room for auth header?
2590                  */
2591                 len = maxlen - (p - (u_char *)ip6);
2592                 len &= ~7;
2593
2594                 /* This is just for simplicity. */
2595                 if (m0->m_pkthdr.len != m0->m_len) {
2596                         if (m0->m_next) {
2597                                 m_freem(m0->m_next);
2598                                 m0->m_next = NULL;
2599                         }
2600                         m0->m_pkthdr.len = m0->m_len;
2601                 }
2602
2603                 /*
2604                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2605                  * about padding/truncate rule for the original IP packet.
2606                  * From the discussion on IPv6imp in Feb 1999,
2607                  * the consensus was:
2608                  * - "attach as much as possible" is the goal
2609                  * - pad if not aligned (original size can be guessed by
2610                  *   original ip6 header)
2611                  * Following code adds the padding if it is simple enough,
2612                  * and truncates if not.
2613                  */
2614                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2615                         panic("assumption failed in %s:%d", __FILE__,
2616                             __LINE__);
2617
2618                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2619                         /* not enough room, truncate */
2620                         m0->m_pkthdr.len = m0->m_len = len -
2621                             sizeof(*nd_opt_rh);
2622                 } else {
2623                         /* enough room, pad or truncate */
2624                         size_t extra;
2625
2626                         extra = m0->m_pkthdr.len % 8;
2627                         if (extra) {
2628                                 /* pad if easy enough, truncate if not */
2629                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2630                                         /* pad */
2631                                         m0->m_len += (8 - extra);
2632                                         m0->m_pkthdr.len += (8 - extra);
2633                                 } else {
2634                                         /* truncate */
2635                                         m0->m_pkthdr.len -= extra;
2636                                         m0->m_len -= extra;
2637                                 }
2638                         }
2639                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2640                         m0->m_pkthdr.len = m0->m_len = len -
2641                             sizeof(*nd_opt_rh);
2642                 }
2643
2644                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2645                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2646                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2647                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2648                 p += sizeof(*nd_opt_rh);
2649                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2650
2651                 /* connect m0 to m */
2652                 m_tag_delete_chain(m0, NULL);
2653                 m0->m_flags &= ~M_PKTHDR;
2654                 m->m_next = m0;
2655                 m->m_pkthdr.len = m->m_len + m0->m_len;
2656                 m0 = NULL;
2657         }
2658 noredhdropt:;
2659         if (m0) {
2660                 m_freem(m0);
2661                 m0 = NULL;
2662         }
2663
2664         /* XXX: clear embedded link IDs in the inner header */
2665         in6_clearscope(&sip6->ip6_src);
2666         in6_clearscope(&sip6->ip6_dst);
2667         in6_clearscope(&nd_rd->nd_rd_target);
2668         in6_clearscope(&nd_rd->nd_rd_dst);
2669
2670         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2671
2672         nd_rd->nd_rd_cksum = 0;
2673         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2674             sizeof(*ip6), ntohs(ip6->ip6_plen));
2675
2676         /* send the packet to outside... */
2677         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2678         if (outif) {
2679                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2680                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2681         }
2682         icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2683
2684         return;
2685
2686 fail:
2687         if (m)
2688                 m_freem(m);
2689         if (m0)
2690                 m_freem(m0);
2691 }
2692
2693 /*
2694  * ICMPv6 socket option processing.
2695  */
2696 int
2697 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2698 {
2699         int error = 0;
2700         int optlen;
2701         struct inpcb *inp = sotoinpcb(so);
2702         int level, op, optname;
2703
2704         if (sopt) {
2705                 level = sopt->sopt_level;
2706                 op = sopt->sopt_dir;
2707                 optname = sopt->sopt_name;
2708                 optlen = sopt->sopt_valsize;
2709         } else
2710                 level = op = optname = optlen = 0;
2711
2712         if (level != IPPROTO_ICMPV6) {
2713                 return EINVAL;
2714         }
2715
2716         switch (op) {
2717         case PRCO_SETOPT:
2718                 switch (optname) {
2719                 case ICMP6_FILTER:
2720                     {
2721                         struct icmp6_filter ic6f;
2722
2723                         if (optlen != sizeof(ic6f)) {
2724                                 error = EMSGSIZE;
2725                                 break;
2726                         }
2727                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2728                         if (error == 0) {
2729                                 INP_WLOCK(inp);
2730                                 *inp->in6p_icmp6filt = ic6f;
2731                                 INP_WUNLOCK(inp);
2732                         }
2733                         break;
2734                     }
2735
2736                 default:
2737                         error = ENOPROTOOPT;
2738                         break;
2739                 }
2740                 break;
2741
2742         case PRCO_GETOPT:
2743                 switch (optname) {
2744                 case ICMP6_FILTER:
2745                     {
2746                         struct icmp6_filter ic6f;
2747
2748                         INP_RLOCK(inp);
2749                         ic6f = *inp->in6p_icmp6filt;
2750                         INP_RUNLOCK(inp);
2751                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2752                         break;
2753                     }
2754
2755                 default:
2756                         error = ENOPROTOOPT;
2757                         break;
2758                 }
2759                 break;
2760         }
2761
2762         return (error);
2763 }
2764
2765 /*
2766  * Perform rate limit check.
2767  * Returns 0 if it is okay to send the icmp6 packet.
2768  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2769  * limitation.
2770  *
2771  * XXX per-destination/type check necessary?
2772  *
2773  * dst - not used at this moment
2774  * type - not used at this moment
2775  * code - not used at this moment
2776  */
2777 static int
2778 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2779     const int code)
2780 {
2781         int ret;
2782
2783         ret = 0;        /* okay to send */
2784
2785         /* PPS limit */
2786         if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2787             icmp6errppslim)) {
2788                 /* The packet is subject to rate limit */
2789                 ret++;
2790         }
2791
2792         return ret;
2793 }