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 #include <sys/vimage.h>
  85
  86 #include <net/if.h>
  87 #include <net/if_dl.h>
  88 #include <net/if_llatbl.h>
  89 #include <net/if_types.h>
  90 #include <net/route.h>
  91 #include <net/vnet.h>
  92
  93 #include <netinet/in.h>
  94 #include <netinet/in_pcb.h>
  95 #include <netinet/in_var.h>
  96 #include <netinet/ip6.h>
  97 #include <netinet/icmp6.h>
  98 #include <netinet/tcp_var.h>
  99 #include <netinet/vinet.h>
 100
 101 #include <netinet6/in6_ifattach.h>
 102 #include <netinet6/in6_pcb.h>
 103 #include <netinet6/ip6protosw.h>
 104 #include <netinet6/ip6_var.h>
 105 #include <netinet6/scope6_var.h>
 106 #include <netinet6/mld6_var.h>
 107 #include <netinet6/nd6.h>
 108 #include <netinet6/vinet6.h>
 109
 110 #ifdef IPSEC
 111 #include <netipsec/ipsec.h>
 112 #include <netipsec/key.h>
 113 #endif
 114
 115 extern struct domain inet6domain;
 116
 117 #ifdef VIMAGE_GLOBALS
 118 extern struct inpcbinfo ripcbinfo;
 119 extern struct inpcbhead ripcb;
 120 extern int icmp6errppslim;
 121 extern int icmp6_nodeinfo;
 122
 123 struct icmp6stat icmp6stat;
 124 static int icmp6errpps_count;
 125 static struct timeval icmp6errppslim_last;
 126 #endif
 127
 128 static void icmp6_errcount(struct icmp6errstat *, int, int);
 129 static int icmp6_rip6_input(struct mbuf **, int);
 130 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
 131 static const char *icmp6_redirect_diag __P((struct in6_addr *,
 132         struct in6_addr *, struct in6_addr *));
 133 static struct mbuf *ni6_input(struct mbuf *, int);
 134 static struct mbuf *ni6_nametodns(const char *, int, int);
 135 static int ni6_dnsmatch(const char *, int, const char *, int);
 136 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
 137                           struct ifnet **, struct in6_addr *));
 138 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
 139                                 struct ifnet *, int));
 140 static int icmp6_notify_error(struct mbuf **, int, int, int);
 141
 142
 143 void
 144 icmp6_init(void)
 145 {
 146         INIT_VNET_INET6(curvnet);
 147
 148         V_icmp6errpps_count = 0;
 149
 150         mld6_init();
 151 }
 152
 153 static void
 154 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
 155 {
 156         switch (type) {
 157         case ICMP6_DST_UNREACH:
 158                 switch (code) {
 159                 case ICMP6_DST_UNREACH_NOROUTE:
 160                         stat->icp6errs_dst_unreach_noroute++;
 161                         return;
 162                 case ICMP6_DST_UNREACH_ADMIN:
 163                         stat->icp6errs_dst_unreach_admin++;
 164                         return;
 165                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 166                         stat->icp6errs_dst_unreach_beyondscope++;
 167                         return;
 168                 case ICMP6_DST_UNREACH_ADDR:
 169                         stat->icp6errs_dst_unreach_addr++;
 170                         return;
 171                 case ICMP6_DST_UNREACH_NOPORT:
 172                         stat->icp6errs_dst_unreach_noport++;
 173                         return;
 174                 }
 175                 break;
 176         case ICMP6_PACKET_TOO_BIG:
 177                 stat->icp6errs_packet_too_big++;
 178                 return;
 179         case ICMP6_TIME_EXCEEDED:
 180                 switch (code) {
 181                 case ICMP6_TIME_EXCEED_TRANSIT:
 182                         stat->icp6errs_time_exceed_transit++;
 183                         return;
 184                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 185                         stat->icp6errs_time_exceed_reassembly++;
 186                         return;
 187                 }
 188                 break;
 189         case ICMP6_PARAM_PROB:
 190                 switch (code) {
 191                 case ICMP6_PARAMPROB_HEADER:
 192                         stat->icp6errs_paramprob_header++;
 193                         return;
 194                 case ICMP6_PARAMPROB_NEXTHEADER:
 195                         stat->icp6errs_paramprob_nextheader++;
 196                         return;
 197                 case ICMP6_PARAMPROB_OPTION:
 198                         stat->icp6errs_paramprob_option++;
 199                         return;
 200                 }
 201                 break;
 202         case ND_REDIRECT:
 203                 stat->icp6errs_redirect++;
 204                 return;
 205         }
 206         stat->icp6errs_unknown++;
 207 }
 208
 209 /*
 210  * A wrapper function for icmp6_error() necessary when the erroneous packet
 211  * may not contain enough scope zone information.
 212  */
 213 void
 214 icmp6_error2(struct mbuf *m, int type, int code, int param,
 215     struct ifnet *ifp)
 216 {
 217         INIT_VNET_INET6(curvnet);
 218         struct ip6_hdr *ip6;
 219
 220         if (ifp == NULL)
 221                 return;
 222
 223 #ifndef PULLDOWN_TEST
 224         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 225 #else
 226         if (m->m_len < sizeof(struct ip6_hdr)) {
 227                 m = m_pullup(m, sizeof(struct ip6_hdr));
 228                 if (m == NULL)
 229                         return;
 230         }
 231 #endif
 232
 233         ip6 = mtod(m, struct ip6_hdr *);
 234
 235         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
 236                 return;
 237         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
 238                 return;
 239
 240         icmp6_error(m, type, code, param);
 241 }
 242
 243 /*
 244  * Generate an error packet of type error in response to bad IP6 packet.
 245  */
 246 void
 247 icmp6_error(struct mbuf *m, int type, int code, int param)
 248 {
 249         INIT_VNET_INET6(curvnet);
 250         struct ip6_hdr *oip6, *nip6;
 251         struct icmp6_hdr *icmp6;
 252         u_int preplen;
 253         int off;
 254         int nxt;
 255
 256         V_icmp6stat.icp6s_error++;
 257
 258         /* count per-type-code statistics */
 259         icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, type, code);
 260
 261 #ifdef M_DECRYPTED      /*not openbsd*/
 262         if (m->m_flags & M_DECRYPTED) {
 263                 V_icmp6stat.icp6s_canterror++;
 264                 goto freeit;
 265         }
 266 #endif
 267
 268 #ifndef PULLDOWN_TEST
 269         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 270 #else
 271         if (m->m_len < sizeof(struct ip6_hdr)) {
 272                 m = m_pullup(m, sizeof(struct ip6_hdr));
 273                 if (m == NULL)
 274                         return;
 275         }
 276 #endif
 277         oip6 = mtod(m, struct ip6_hdr *);
 278
 279         /*
 280          * If the destination address of the erroneous packet is a multicast
 281          * address, or the packet was sent using link-layer multicast,
 282          * we should basically suppress sending an error (RFC 2463, Section
 283          * 2.4).
 284          * We have two exceptions (the item e.2 in that section):
 285          * - the Pakcet Too Big message can be sent for path MTU discovery.
 286          * - the Parameter Problem Message that can be allowed an icmp6 error
 287          *   in the option type field.  This check has been done in
 288          *   ip6_unknown_opt(), so we can just check the type and code.
 289          */
 290         if ((m->m_flags & (M_BCAST|M_MCAST) ||
 291              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
 292             (type != ICMP6_PACKET_TOO_BIG &&
 293              (type != ICMP6_PARAM_PROB ||
 294               code != ICMP6_PARAMPROB_OPTION)))
 295                 goto freeit;
 296
 297         /*
 298          * RFC 2463, 2.4 (e.5): source address check.
 299          * XXX: the case of anycast source?
 300          */
 301         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
 302             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
 303                 goto freeit;
 304
 305         /*
 306          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
 307          * don't do it.
 308          */
 309         nxt = -1;
 310         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
 311         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
 312                 struct icmp6_hdr *icp;
 313
 314 #ifndef PULLDOWN_TEST
 315                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
 316                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 317 #else
 318                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
 319                         sizeof(*icp));
 320                 if (icp == NULL) {
 321                         V_icmp6stat.icp6s_tooshort++;
 322                         return;
 323                 }
 324 #endif
 325                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
 326                     icp->icmp6_type == ND_REDIRECT) {
 327                         /*
 328                          * ICMPv6 error
 329                          * Special case: for redirect (which is
 330                          * informational) we must not send icmp6 error.
 331                          */
 332                         V_icmp6stat.icp6s_canterror++;
 333                         goto freeit;
 334                 } else {
 335                         /* ICMPv6 informational - send the error */
 336                 }
 337         } else {
 338                 /* non-ICMPv6 - send the error */
 339         }
 340
 341         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
 342
 343         /* Finally, do rate limitation check. */
 344         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
 345                 V_icmp6stat.icp6s_toofreq++;
 346                 goto freeit;
 347         }
 348
 349         /*
 350          * OK, ICMP6 can be generated.
 351          */
 352
 353         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
 354                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
 355
 356         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
 357         M_PREPEND(m, preplen, M_DONTWAIT);
 358         if (m && m->m_len < preplen)
 359                 m = m_pullup(m, preplen);
 360         if (m == NULL) {
 361                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
 362                 return;
 363         }
 364
 365         nip6 = mtod(m, struct ip6_hdr *);
 366         nip6->ip6_src  = oip6->ip6_src;
 367         nip6->ip6_dst  = oip6->ip6_dst;
 368
 369         in6_clearscope(&oip6->ip6_src);
 370         in6_clearscope(&oip6->ip6_dst);
 371
 372         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
 373         icmp6->icmp6_type = type;
 374         icmp6->icmp6_code = code;
 375         icmp6->icmp6_pptr = htonl((u_int32_t)param);
 376
 377         /*
 378          * icmp6_reflect() is designed to be in the input path.
 379          * icmp6_error() can be called from both input and output path,
 380          * and if we are in output path rcvif could contain bogus value.
 381          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
 382          * information in ip header (nip6).
 383          */
 384         m->m_pkthdr.rcvif = NULL;
 385
 386         V_icmp6stat.icp6s_outhist[type]++;
 387         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
 388
 389         return;
 390
 391   freeit:
 392         /*
 393          * If we can't tell whether or not we can generate ICMP6, free it.
 394          */
 395         m_freem(m);
 396 }
 397
 398 /*
 399  * Process a received ICMP6 message.
 400  */
 401 int
 402 icmp6_input(struct mbuf **mp, int *offp, int proto)
 403 {
 404         INIT_VNET_INET6(curvnet);
 405         INIT_VPROCG(TD_TO_VPROCG(curthread)); /* XXX V_hostname needs this */
 406         struct mbuf *m = *mp, *n;
 407         struct ip6_hdr *ip6, *nip6;
 408         struct icmp6_hdr *icmp6, *nicmp6;
 409         int off = *offp;
 410         int icmp6len = m->m_pkthdr.len - *offp;
 411         int code, sum, noff;
 412         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
 413
 414 #ifndef PULLDOWN_TEST
 415         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
 416         /* m might change if M_LOOP.  So, call mtod after this */
 417 #endif
 418
 419         /*
 420          * Locate icmp6 structure in mbuf, and check
 421          * that not corrupted and of at least minimum length
 422          */
 423
 424         ip6 = mtod(m, struct ip6_hdr *);
 425         if (icmp6len < sizeof(struct icmp6_hdr)) {
 426                 V_icmp6stat.icp6s_tooshort++;
 427                 goto freeit;
 428         }
 429
 430         /*
 431          * calculate the checksum
 432          */
 433 #ifndef PULLDOWN_TEST
 434         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
 435 #else
 436         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
 437         if (icmp6 == NULL) {
 438                 V_icmp6stat.icp6s_tooshort++;
 439                 return IPPROTO_DONE;
 440         }
 441 #endif
 442         code = icmp6->icmp6_code;
 443
 444         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
 445                 nd6log((LOG_ERR,
 446                     "ICMP6 checksum error(%d|%x) %s\n",
 447                     icmp6->icmp6_type, sum,
 448                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
 449                 V_icmp6stat.icp6s_checksum++;
 450                 goto freeit;
 451         }
 452
 453         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
 454                 /*
 455                  * Deliver very specific ICMP6 type only.
 456                  * This is important to deliver TOOBIG.  Otherwise PMTUD
 457                  * will not work.
 458                  */
 459                 switch (icmp6->icmp6_type) {
 460                 case ICMP6_DST_UNREACH:
 461                 case ICMP6_PACKET_TOO_BIG:
 462                 case ICMP6_TIME_EXCEEDED:
 463                         break;
 464                 default:
 465                         goto freeit;
 466                 }
 467         }
 468
 469         V_icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
 470         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
 471         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
 472                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
 473
 474         switch (icmp6->icmp6_type) {
 475         case ICMP6_DST_UNREACH:
 476                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
 477                 switch (code) {
 478                 case ICMP6_DST_UNREACH_NOROUTE:
 479                         code = PRC_UNREACH_NET;
 480                         break;
 481                 case ICMP6_DST_UNREACH_ADMIN:
 482                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
 483                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
 484                         break;
 485                 case ICMP6_DST_UNREACH_ADDR:
 486                         code = PRC_HOSTDEAD;
 487                         break;
 488                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 489                         /* I mean "source address was incorrect." */
 490                         code = PRC_PARAMPROB;
 491                         break;
 492                 case ICMP6_DST_UNREACH_NOPORT:
 493                         code = PRC_UNREACH_PORT;
 494                         break;
 495                 default:
 496                         goto badcode;
 497                 }
 498                 goto deliver;
 499                 break;
 500
 501         case ICMP6_PACKET_TOO_BIG:
 502                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
 503
 504                 /* validation is made in icmp6_mtudisc_update */
 505
 506                 code = PRC_MSGSIZE;
 507
 508                 /*
 509                  * Updating the path MTU will be done after examining
 510                  * intermediate extension headers.
 511                  */
 512                 goto deliver;
 513                 break;
 514
 515         case ICMP6_TIME_EXCEEDED:
 516                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
 517                 switch (code) {
 518                 case ICMP6_TIME_EXCEED_TRANSIT:
 519                         code = PRC_TIMXCEED_INTRANS;
 520                         break;
 521                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 522                         code = PRC_TIMXCEED_REASS;
 523                         break;
 524                 default:
 525                         goto badcode;
 526                 }
 527                 goto deliver;
 528                 break;
 529
 530         case ICMP6_PARAM_PROB:
 531                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
 532                 switch (code) {
 533                 case ICMP6_PARAMPROB_NEXTHEADER:
 534                         code = PRC_UNREACH_PROTOCOL;
 535                         break;
 536                 case ICMP6_PARAMPROB_HEADER:
 537                 case ICMP6_PARAMPROB_OPTION:
 538                         code = PRC_PARAMPROB;
 539                         break;
 540                 default:
 541                         goto badcode;
 542                 }
 543                 goto deliver;
 544                 break;
 545
 546         case ICMP6_ECHO_REQUEST:
 547                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
 548                 if (code != 0)
 549                         goto badcode;
 550                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
 551                         /* Give up remote */
 552                         break;
 553                 }
 554                 if ((n->m_flags & M_EXT) != 0
 555                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
 556                         struct mbuf *n0 = n;
 557                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
 558                         int n0len;
 559
 560                         MGETHDR(n, M_DONTWAIT, n0->m_type);
 561                         n0len = n0->m_pkthdr.len;       /* save for use below */
 562                         if (n)
 563                                 M_MOVE_PKTHDR(n, n0);
 564                         if (n && maxlen >= MHLEN) {
 565                                 MCLGET(n, M_DONTWAIT);
 566                                 if ((n->m_flags & M_EXT) == 0) {
 567                                         m_free(n);
 568                                         n = NULL;
 569                                 }
 570                         }
 571                         if (n == NULL) {
 572                                 /* Give up remote */
 573                                 m_freem(n0);
 574                                 break;
 575                         }
 576                         /*
 577                          * Copy IPv6 and ICMPv6 only.
 578                          */
 579                         nip6 = mtod(n, struct ip6_hdr *);
 580                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 581                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 582                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 583                         noff = sizeof(struct ip6_hdr);
 584                         /* new mbuf contains only ipv6+icmpv6 headers */
 585                         n->m_len = noff + sizeof(struct icmp6_hdr);
 586                         /*
 587                          * Adjust mbuf.  ip6_plen will be adjusted in
 588                          * ip6_output().
 589                          */
 590                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 591                         /* recalculate complete packet size */
 592                         n->m_pkthdr.len = n0len + (noff - off);
 593                         n->m_next = n0;
 594                 } else {
 595                         nip6 = mtod(n, struct ip6_hdr *);
 596                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
 597                             sizeof(*nicmp6));
 598                         noff = off;
 599                 }
 600                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 601                 nicmp6->icmp6_code = 0;
 602                 if (n) {
 603                         V_icmp6stat.icp6s_reflect++;
 604                         V_icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
 605                         icmp6_reflect(n, noff);
 606                 }
 607                 break;
 608
 609         case ICMP6_ECHO_REPLY:
 610                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
 611                 if (code != 0)
 612                         goto badcode;
 613                 break;
 614
 615         case MLD_LISTENER_QUERY:
 616         case MLD_LISTENER_REPORT:
 617                 if (icmp6len < sizeof(struct mld_hdr))
 618                         goto badlen;
 619                 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
 620                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
 621                 else
 622                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
 623                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 624                         /* give up local */
 625                         mld6_input(m, off);
 626                         m = NULL;
 627                         goto freeit;
 628                 }
 629                 mld6_input(n, off);
 630                 /* m stays. */
 631                 break;
 632
 633         case MLD_LISTENER_DONE:
 634                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
 635                 if (icmp6len < sizeof(struct mld_hdr))  /* necessary? */
 636                         goto badlen;
 637                 break;          /* nothing to be done in kernel */
 638
 639         case MLD_MTRACE_RESP:
 640         case MLD_MTRACE:
 641                 /* XXX: these two are experimental.  not officially defined. */
 642                 /* XXX: per-interface statistics? */
 643                 break;          /* just pass it to applications */
 644
 645         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 646             {
 647                 enum { WRU, FQDN } mode;
 648
 649                 if (!V_icmp6_nodeinfo)
 650                         break;
 651
 652                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 653                         mode = WRU;
 654                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 655                         mode = FQDN;
 656                 else
 657                         goto badlen;
 658
 659 #define hostnamelen     strlen(V_hostname)
 660                 if (mode == FQDN) {
 661 #ifndef PULLDOWN_TEST
 662                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
 663                             IPPROTO_DONE);
 664 #endif
 665                         n = m_copy(m, 0, M_COPYALL);
 666                         if (n)
 667                                 n = ni6_input(n, off);
 668                         /* XXX meaningless if n == NULL */
 669                         noff = sizeof(struct ip6_hdr);
 670                 } else {
 671                         u_char *p;
 672                         int maxlen, maxhlen;
 673
 674                         /*
 675                          * XXX: this combination of flags is pointless,
 676                          * but should we keep this for compatibility?
 677                          */
 678                         if ((V_icmp6_nodeinfo & 5) != 5)
 679                                 break;
 680
 681                         if (code != 0)
 682                                 goto badcode;
 683                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
 684                         if (maxlen >= MCLBYTES) {
 685                                 /* Give up remote */
 686                                 break;
 687                         }
 688                         MGETHDR(n, M_DONTWAIT, m->m_type);
 689                         if (n && maxlen > MHLEN) {
 690                                 MCLGET(n, M_DONTWAIT);
 691                                 if ((n->m_flags & M_EXT) == 0) {
 692                                         m_free(n);
 693                                         n = NULL;
 694                                 }
 695                         }
 696                         if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
 697                                 /*
 698                                  * Previous code did a blind M_COPY_PKTHDR
 699                                  * and said "just for rcvif".  If true, then
 700                                  * we could tolerate the dup failing (due to
 701                                  * the deep copy of the tag chain).  For now
 702                                  * be conservative and just fail.
 703                                  */
 704                                 m_free(n);
 705                                 n = NULL;
 706                         }
 707                         if (n == NULL) {
 708                                 /* Give up remote */
 709                                 break;
 710                         }
 711                         n->m_pkthdr.rcvif = NULL;
 712                         n->m_len = 0;
 713                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
 714                         mtx_lock(&hostname_mtx);
 715                         if (maxhlen > hostnamelen)
 716                                 maxhlen = hostnamelen;
 717                         /*
 718                          * Copy IPv6 and ICMPv6 only.
 719                          */
 720                         nip6 = mtod(n, struct ip6_hdr *);
 721                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 722                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 723                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 724                         p = (u_char *)(nicmp6 + 1);
 725                         bzero(p, 4);
 726                         bcopy(V_hostname, p + 4, maxhlen); /* meaningless TTL */
 727                         mtx_unlock(&hostname_mtx);
 728                         noff = sizeof(struct ip6_hdr);
 729                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 730                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 731                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 732                         nicmp6->icmp6_code = 0;
 733                 }
 734 #undef hostnamelen
 735                 if (n) {
 736                         V_icmp6stat.icp6s_reflect++;
 737                         V_icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
 738                         icmp6_reflect(n, noff);
 739                 }
 740                 break;
 741             }
 742
 743         case ICMP6_WRUREPLY:
 744                 if (code != 0)
 745                         goto badcode;
 746                 break;
 747
 748         case ND_ROUTER_SOLICIT:
 749                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
 750                 if (code != 0)
 751                         goto badcode;
 752                 if (icmp6len < sizeof(struct nd_router_solicit))
 753                         goto badlen;
 754                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 755                         /* give up local */
 756                         nd6_rs_input(m, off, icmp6len);
 757                         m = NULL;
 758                         goto freeit;
 759                 }
 760                 nd6_rs_input(n, off, icmp6len);
 761                 /* m stays. */
 762                 break;
 763
 764         case ND_ROUTER_ADVERT:
 765                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
 766                 if (code != 0)
 767                         goto badcode;
 768                 if (icmp6len < sizeof(struct nd_router_advert))
 769                         goto badlen;
 770                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 771                         /* give up local */
 772                         nd6_ra_input(m, off, icmp6len);
 773                         m = NULL;
 774                         goto freeit;
 775                 }
 776                 nd6_ra_input(n, off, icmp6len);
 777                 /* m stays. */
 778                 break;
 779
 780         case ND_NEIGHBOR_SOLICIT:
 781                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
 782                 if (code != 0)
 783                         goto badcode;
 784                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 785                         goto badlen;
 786                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 787                         /* give up local */
 788                         nd6_ns_input(m, off, icmp6len);
 789                         m = NULL;
 790                         goto freeit;
 791                 }
 792                 nd6_ns_input(n, off, icmp6len);
 793                 /* m stays. */
 794                 break;
 795
 796         case ND_NEIGHBOR_ADVERT:
 797                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
 798                 if (code != 0)
 799                         goto badcode;
 800                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 801                         goto badlen;
 802                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 803                         /* give up local */
 804                         nd6_na_input(m, off, icmp6len);
 805                         m = NULL;
 806                         goto freeit;
 807                 }
 808                 nd6_na_input(n, off, icmp6len);
 809                 /* m stays. */
 810                 break;
 811
 812         case ND_REDIRECT:
 813                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
 814                 if (code != 0)
 815                         goto badcode;
 816                 if (icmp6len < sizeof(struct nd_redirect))
 817                         goto badlen;
 818                 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
 819                         /* give up local */
 820                         icmp6_redirect_input(m, off);
 821                         m = NULL;
 822                         goto freeit;
 823                 }
 824                 icmp6_redirect_input(n, off);
 825                 /* m stays. */
 826                 break;
 827
 828         case ICMP6_ROUTER_RENUMBERING:
 829                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 830                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 831                         goto badcode;
 832                 if (icmp6len < sizeof(struct icmp6_router_renum))
 833                         goto badlen;
 834                 break;
 835
 836         default:
 837                 nd6log((LOG_DEBUG,
 838                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 839                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
 840                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
 841                     m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
 842                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 843                         /* ICMPv6 error: MUST deliver it by spec... */
 844                         code = PRC_NCMDS;
 845                         /* deliver */
 846                 } else {
 847                         /* ICMPv6 informational: MUST not deliver */
 848                         break;
 849                 }
 850         deliver:
 851                 if (icmp6_notify_error(&m, off, icmp6len, code)) {
 852                         /* In this case, m should've been freed. */
 853                         return (IPPROTO_DONE);
 854                 }
 855                 break;
 856
 857         badcode:
 858                 V_icmp6stat.icp6s_badcode++;
 859                 break;
 860
 861         badlen:
 862                 V_icmp6stat.icp6s_badlen++;
 863                 break;
 864         }
 865
 866         /* deliver the packet to appropriate sockets */
 867         icmp6_rip6_input(&m, *offp);
 868
 869         return IPPROTO_DONE;
 870
 871  freeit:
 872         m_freem(m);
 873         return IPPROTO_DONE;
 874 }
 875
 876 static int
 877 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
 878 {
 879         INIT_VNET_INET6(curvnet);
 880         struct mbuf *m = *mp;
 881         struct icmp6_hdr *icmp6;
 882         struct ip6_hdr *eip6;
 883         u_int32_t notifymtu;
 884         struct sockaddr_in6 icmp6src, icmp6dst;
 885
 886         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 887                 V_icmp6stat.icp6s_tooshort++;
 888                 goto freeit;
 889         }
 890 #ifndef PULLDOWN_TEST
 891         IP6_EXTHDR_CHECK(m, off,
 892             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
 893         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 894 #else
 895         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
 896             sizeof(*icmp6) + sizeof(struct ip6_hdr));
 897         if (icmp6 == NULL) {
 898                 V_icmp6stat.icp6s_tooshort++;
 899                 return (-1);
 900         }
 901 #endif
 902         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 903
 904         /* Detect the upper level protocol */
 905         {
 906                 void (*ctlfunc)(int, struct sockaddr *, void *);
 907                 u_int8_t nxt = eip6->ip6_nxt;
 908                 int eoff = off + sizeof(struct icmp6_hdr) +
 909                     sizeof(struct ip6_hdr);
 910                 struct ip6ctlparam ip6cp;
 911                 struct in6_addr *finaldst = NULL;
 912                 int icmp6type = icmp6->icmp6_type;
 913                 struct ip6_frag *fh;
 914                 struct ip6_rthdr *rth;
 915                 struct ip6_rthdr0 *rth0;
 916                 int rthlen;
 917
 918                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 919                         struct ip6_ext *eh;
 920
 921                         switch (nxt) {
 922                         case IPPROTO_HOPOPTS:
 923                         case IPPROTO_DSTOPTS:
 924                         case IPPROTO_AH:
 925 #ifndef PULLDOWN_TEST
 926                                 IP6_EXTHDR_CHECK(m, 0,
 927                                     eoff + sizeof(struct ip6_ext), -1);
 928                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
 929 #else
 930                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
 931                                     eoff, sizeof(*eh));
 932                                 if (eh == NULL) {
 933                                         V_icmp6stat.icp6s_tooshort++;
 934                                         return (-1);
 935                                 }
 936 #endif
 937
 938                                 if (nxt == IPPROTO_AH)
 939                                         eoff += (eh->ip6e_len + 2) << 2;
 940                                 else
 941                                         eoff += (eh->ip6e_len + 1) << 3;
 942                                 nxt = eh->ip6e_nxt;
 943                                 break;
 944                         case IPPROTO_ROUTING:
 945                                 /*
 946                                  * When the erroneous packet contains a
 947                                  * routing header, we should examine the
 948                                  * header to determine the final destination.
 949                                  * Otherwise, we can't properly update
 950                                  * information that depends on the final
 951                                  * destination (e.g. path MTU).
 952                                  */
 953 #ifndef PULLDOWN_TEST
 954                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
 955                                 rth = (struct ip6_rthdr *)
 956                                     (mtod(m, caddr_t) + eoff);
 957 #else
 958                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
 959                                     eoff, sizeof(*rth));
 960                                 if (rth == NULL) {
 961                                         V_icmp6stat.icp6s_tooshort++;
 962                                         return (-1);
 963                                 }
 964 #endif
 965                                 rthlen = (rth->ip6r_len + 1) << 3;
 966                                 /*
 967                                  * XXX: currently there is no
 968                                  * officially defined type other
 969                                  * than type-0.
 970                                  * Note that if the segment left field
 971                                  * is 0, all intermediate hops must
 972                                  * have been passed.
 973                                  */
 974                                 if (rth->ip6r_segleft &&
 975                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
 976                                         int hops;
 977
 978 #ifndef PULLDOWN_TEST
 979                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
 980                                         rth0 = (struct ip6_rthdr0 *)
 981                                             (mtod(m, caddr_t) + eoff);
 982 #else
 983                                         IP6_EXTHDR_GET(rth0,
 984                                             struct ip6_rthdr0 *, m,
 985                                             eoff, rthlen);
 986                                         if (rth0 == NULL) {
 987                                                 V_icmp6stat.icp6s_tooshort++;
 988                                                 return (-1);
 989                                         }
 990 #endif
 991                                         /* just ignore a bogus header */
 992                                         if ((rth0->ip6r0_len % 2) == 0 &&
 993                                             (hops = rth0->ip6r0_len/2))
 994                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
 995                                 }
 996                                 eoff += rthlen;
 997                                 nxt = rth->ip6r_nxt;
 998                                 break;
 999                         case IPPROTO_FRAGMENT:
1000 #ifndef PULLDOWN_TEST
1001                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1002                                     sizeof(struct ip6_frag), -1);
1003                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1004                                     eoff);
1005 #else
1006                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1007                                     eoff, sizeof(*fh));
1008                                 if (fh == NULL) {
1009                                         V_icmp6stat.icp6s_tooshort++;
1010                                         return (-1);
1011                                 }
1012 #endif
1013                                 /*
1014                                  * Data after a fragment header is meaningless
1015                                  * unless it is the first fragment, but
1016                                  * we'll go to the notify label for path MTU
1017                                  * discovery.
1018                                  */
1019                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1020                                         goto notify;
1021
1022                                 eoff += sizeof(struct ip6_frag);
1023                                 nxt = fh->ip6f_nxt;
1024                                 break;
1025                         default:
1026                                 /*
1027                                  * This case includes ESP and the No Next
1028                                  * Header.  In such cases going to the notify
1029                                  * label does not have any meaning
1030                                  * (i.e. ctlfunc will be NULL), but we go
1031                                  * anyway since we might have to update
1032                                  * path MTU information.
1033                                  */
1034                                 goto notify;
1035                         }
1036                 }
1037           notify:
1038 #ifndef PULLDOWN_TEST
1039                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1040 #else
1041                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1042                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1043                 if (icmp6 == NULL) {
1044                         V_icmp6stat.icp6s_tooshort++;
1045                         return (-1);
1046                 }
1047 #endif
1048
1049                 /*
1050                  * retrieve parameters from the inner IPv6 header, and convert
1051                  * them into sockaddr structures.
1052                  * XXX: there is no guarantee that the source or destination
1053                  * addresses of the inner packet are in the same scope as
1054                  * the addresses of the icmp packet.  But there is no other
1055                  * way to determine the zone.
1056                  */
1057                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1058
1059                 bzero(&icmp6dst, sizeof(icmp6dst));
1060                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1061                 icmp6dst.sin6_family = AF_INET6;
1062                 if (finaldst == NULL)
1063                         icmp6dst.sin6_addr = eip6->ip6_dst;
1064                 else
1065                         icmp6dst.sin6_addr = *finaldst;
1066                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1067                         goto freeit;
1068                 bzero(&icmp6src, sizeof(icmp6src));
1069                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1070                 icmp6src.sin6_family = AF_INET6;
1071                 icmp6src.sin6_addr = eip6->ip6_src;
1072                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1073                         goto freeit;
1074                 icmp6src.sin6_flowinfo =
1075                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1076
1077                 if (finaldst == NULL)
1078                         finaldst = &eip6->ip6_dst;
1079                 ip6cp.ip6c_m = m;
1080                 ip6cp.ip6c_icmp6 = icmp6;
1081                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1082                 ip6cp.ip6c_off = eoff;
1083                 ip6cp.ip6c_finaldst = finaldst;
1084                 ip6cp.ip6c_src = &icmp6src;
1085                 ip6cp.ip6c_nxt = nxt;
1086
1087                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1088                         notifymtu = ntohl(icmp6->icmp6_mtu);
1089                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1090                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1091                 }
1092
1093                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1094                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1095                 if (ctlfunc) {
1096                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1097                             &ip6cp);
1098                 }
1099         }
1100         *mp = m;
1101         return (0);
1102
1103   freeit:
1104         m_freem(m);
1105         return (-1);
1106 }
1107
1108 void
1109 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1110 {
1111         INIT_VNET_INET6(curvnet);
1112         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1113         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1114         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1115         u_int mtu = ntohl(icmp6->icmp6_mtu);
1116         struct in_conninfo inc;
1117
1118 #if 0
1119         /*
1120          * RFC2460 section 5, last paragraph.
1121          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1122          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1123          * due to packet translator in the middle.
1124          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1125          * special handling.
1126          */
1127         if (mtu < IPV6_MMTU)
1128                 return;
1129 #endif
1130
1131         /*
1132          * we reject ICMPv6 too big with abnormally small value.
1133          * XXX what is the good definition of "abnormally small"?
1134          */
1135         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1136                 return;
1137
1138         if (!validated)
1139                 return;
1140
1141         /*
1142          * In case the suggested mtu is less than IPV6_MMTU, we
1143          * only need to remember that it was for above mentioned
1144          * "alwaysfrag" case.
1145          * Try to be as close to the spec as possible.
1146          */
1147         if (mtu < IPV6_MMTU)
1148                 mtu = IPV6_MMTU - 8;
1149
1150         bzero(&inc, sizeof(inc));
1151         inc.inc_flags |= INC_ISIPV6;
1152         inc.inc6_faddr = *dst;
1153         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1154                 return;
1155
1156         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1157                 tcp_hc_updatemtu(&inc, mtu);
1158                 V_icmp6stat.icp6s_pmtuchg++;
1159         }
1160 }
1161
1162 /*
1163  * Process a Node Information Query packet, based on
1164  * draft-ietf-ipngwg-icmp-name-lookups-07.
1165  *
1166  * Spec incompatibilities:
1167  * - IPv6 Subject address handling
1168  * - IPv4 Subject address handling support missing
1169  * - Proxy reply (answer even if it's not for me)
1170  * - joins NI group address at in6_ifattach() time only, does not cope
1171  *   with hostname changes by sethostname(3)
1172  */
1173 #define hostnamelen     strlen(V_hostname)
1174 static struct mbuf *
1175 ni6_input(struct mbuf *m, int off)
1176 {
1177         INIT_VNET_INET6(curvnet);
1178         INIT_VPROCG(TD_TO_VPROCG(curthread)); /* XXX V_hostname needs this */
1179         struct icmp6_nodeinfo *ni6, *nni6;
1180         struct mbuf *n = NULL;
1181         u_int16_t qtype;
1182         int subjlen;
1183         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1184         struct ni_reply_fqdn *fqdn;
1185         int addrs;              /* for NI_QTYPE_NODEADDR */
1186         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1187         struct in6_addr in6_subj; /* subject address */
1188         struct ip6_hdr *ip6;
1189         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1190         char *subj = NULL;
1191         struct in6_ifaddr *ia6 = NULL;
1192
1193         ip6 = mtod(m, struct ip6_hdr *);
1194 #ifndef PULLDOWN_TEST
1195         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1196 #else
1197         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1198         if (ni6 == NULL) {
1199                 /* m is already reclaimed */
1200                 return (NULL);
1201         }
1202 #endif
1203
1204         /*
1205          * Validate IPv6 source address.
1206          * The default configuration MUST be to refuse answering queries from
1207          * global-scope addresses according to RFC4602.
1208          * Notes:
1209          *  - it's not very clear what "refuse" means; this implementation
1210          *    simply drops it.
1211          *  - it's not very easy to identify global-scope (unicast) addresses
1212          *    since there are many prefixes for them.  It should be safer
1213          *    and in practice sufficient to check "all" but loopback and
1214          *    link-local (note that site-local unicast was deprecated and
1215          *    ULA is defined as global scope-wise)
1216          */
1217         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1218             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1219             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1220                 goto bad;
1221
1222         /*
1223          * Validate IPv6 destination address.
1224          *
1225          * The Responder must discard the Query without further processing
1226          * unless it is one of the Responder's unicast or anycast addresses, or
1227          * a link-local scope multicast address which the Responder has joined.
1228          * [RFC4602, Section 5.]
1229          */
1230         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1231                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1232                         goto bad;
1233                 /* else it's a link-local multicast, fine */
1234         } else {                /* unicast or anycast */
1235                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1236                         goto bad; /* XXX impossible */
1237
1238                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1239                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1240                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1241                                 "a temporary address in %s:%d",
1242                                __FILE__, __LINE__));
1243                         goto bad;
1244                 }
1245         }
1246
1247         /* validate query Subject field. */
1248         qtype = ntohs(ni6->ni_qtype);
1249         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1250         switch (qtype) {
1251         case NI_QTYPE_NOOP:
1252         case NI_QTYPE_SUPTYPES:
1253                 /* 07 draft */
1254                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1255                         break;
1256                 /* FALLTHROUGH */
1257         case NI_QTYPE_FQDN:
1258         case NI_QTYPE_NODEADDR:
1259         case NI_QTYPE_IPV4ADDR:
1260                 switch (ni6->ni_code) {
1261                 case ICMP6_NI_SUBJ_IPV6:
1262 #if ICMP6_NI_SUBJ_IPV6 != 0
1263                 case 0:
1264 #endif
1265                         /*
1266                          * backward compatibility - try to accept 03 draft
1267                          * format, where no Subject is present.
1268                          */
1269                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1270                             subjlen == 0) {
1271                                 oldfqdn++;
1272                                 break;
1273                         }
1274 #if ICMP6_NI_SUBJ_IPV6 != 0
1275                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1276                                 goto bad;
1277 #endif
1278
1279                         if (subjlen != sizeof(struct in6_addr))
1280                                 goto bad;
1281
1282                         /*
1283                          * Validate Subject address.
1284                          *
1285                          * Not sure what exactly "address belongs to the node"
1286                          * means in the spec, is it just unicast, or what?
1287                          *
1288                          * At this moment we consider Subject address as
1289                          * "belong to the node" if the Subject address equals
1290                          * to the IPv6 destination address; validation for
1291                          * IPv6 destination address should have done enough
1292                          * check for us.
1293                          *
1294                          * We do not do proxy at this moment.
1295                          */
1296                         /* m_pulldown instead of copy? */
1297                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1298                             subjlen, (caddr_t)&in6_subj);
1299                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1300                                 goto bad;
1301
1302                         subj = (char *)&in6_subj;
1303                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1304                                 break;
1305
1306                         /*
1307                          * XXX if we are to allow other cases, we should really
1308                          * be careful about scope here.
1309                          * basically, we should disallow queries toward IPv6
1310                          * destination X with subject Y,
1311                          * if scope(X) > scope(Y).
1312                          * if we allow scope(X) > scope(Y), it will result in
1313                          * information leakage across scope boundary.
1314                          */
1315                         goto bad;
1316
1317                 case ICMP6_NI_SUBJ_FQDN:
1318                         /*
1319                          * Validate Subject name with gethostname(3).
1320                          *
1321                          * The behavior may need some debate, since:
1322                          * - we are not sure if the node has FQDN as
1323                          *   hostname (returned by gethostname(3)).
1324                          * - the code does wildcard match for truncated names.
1325                          *   however, we are not sure if we want to perform
1326                          *   wildcard match, if gethostname(3) side has
1327                          *   truncated hostname.
1328                          */
1329                         mtx_lock(&hostname_mtx);
1330                         n = ni6_nametodns(V_hostname, hostnamelen, 0);
1331                         mtx_unlock(&hostname_mtx);
1332                         if (!n || n->m_next || n->m_len == 0)
1333                                 goto bad;
1334                         IP6_EXTHDR_GET(subj, char *, m,
1335                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1336                         if (subj == NULL)
1337                                 goto bad;
1338                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1339                             n->m_len)) {
1340                                 goto bad;
1341                         }
1342                         m_freem(n);
1343                         n = NULL;
1344                         break;
1345
1346                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1347                 default:
1348                         goto bad;
1349                 }
1350                 break;
1351         }
1352
1353         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1354         switch (qtype) {
1355         case NI_QTYPE_FQDN:
1356                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1357                         goto bad;
1358                 break;
1359         case NI_QTYPE_NODEADDR:
1360         case NI_QTYPE_IPV4ADDR:
1361                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1362                         goto bad;
1363                 break;
1364         }
1365
1366         /* guess reply length */
1367         switch (qtype) {
1368         case NI_QTYPE_NOOP:
1369                 break;          /* no reply data */
1370         case NI_QTYPE_SUPTYPES:
1371                 replylen += sizeof(u_int32_t);
1372                 break;
1373         case NI_QTYPE_FQDN:
1374                 /* XXX will append an mbuf */
1375                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1376                 break;
1377         case NI_QTYPE_NODEADDR:
1378                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1379                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1380                     sizeof(u_int32_t))) > MCLBYTES)
1381                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1382                 break;
1383         case NI_QTYPE_IPV4ADDR:
1384                 /* unsupported - should respond with unknown Qtype? */
1385                 break;
1386         default:
1387                 /*
1388                  * XXX: We must return a reply with the ICMP6 code
1389                  * `unknown Qtype' in this case.  However we regard the case
1390                  * as an FQDN query for backward compatibility.
1391                  * Older versions set a random value to this field,
1392                  * so it rarely varies in the defined qtypes.
1393                  * But the mechanism is not reliable...
1394                  * maybe we should obsolete older versions.
1395                  */
1396                 qtype = NI_QTYPE_FQDN;
1397                 /* XXX will append an mbuf */
1398                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1399                 oldfqdn++;
1400                 break;
1401         }
1402
1403         /* allocate an mbuf to reply. */
1404         MGETHDR(n, M_DONTWAIT, m->m_type);
1405         if (n == NULL) {
1406                 m_freem(m);
1407                 return (NULL);
1408         }
1409         M_MOVE_PKTHDR(n, m); /* just for recvif */
1410         if (replylen > MHLEN) {
1411                 if (replylen > MCLBYTES) {
1412                         /*
1413                          * XXX: should we try to allocate more? But MCLBYTES
1414                          * is probably much larger than IPV6_MMTU...
1415                          */
1416                         goto bad;
1417                 }
1418                 MCLGET(n, M_DONTWAIT);
1419                 if ((n->m_flags & M_EXT) == 0) {
1420                         goto bad;
1421                 }
1422         }
1423         n->m_pkthdr.len = n->m_len = replylen;
1424
1425         /* copy mbuf header and IPv6 + Node Information base headers */
1426         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1427         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1428         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1429
1430         /* qtype dependent procedure */
1431         switch (qtype) {
1432         case NI_QTYPE_NOOP:
1433                 nni6->ni_code = ICMP6_NI_SUCCESS;
1434                 nni6->ni_flags = 0;
1435                 break;
1436         case NI_QTYPE_SUPTYPES:
1437         {
1438                 u_int32_t v;
1439                 nni6->ni_code = ICMP6_NI_SUCCESS;
1440                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1441                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1442                 v = (u_int32_t)htonl(0x0000000f);
1443                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1444                 break;
1445         }
1446         case NI_QTYPE_FQDN:
1447                 nni6->ni_code = ICMP6_NI_SUCCESS;
1448                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1449                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1450                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1451                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1452                 /*
1453                  * XXX do we really have FQDN in variable "hostname"?
1454                  */
1455                 mtx_lock(&hostname_mtx);
1456                 n->m_next = ni6_nametodns(V_hostname, hostnamelen, oldfqdn);
1457                 mtx_unlock(&hostname_mtx);
1458                 if (n->m_next == NULL)
1459                         goto bad;
1460                 /* XXX we assume that n->m_next is not a chain */
1461                 if (n->m_next->m_next != NULL)
1462                         goto bad;
1463                 n->m_pkthdr.len += n->m_next->m_len;
1464                 break;
1465         case NI_QTYPE_NODEADDR:
1466         {
1467                 int lenlim, copied;
1468
1469                 nni6->ni_code = ICMP6_NI_SUCCESS;
1470                 n->m_pkthdr.len = n->m_len =
1471                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1472                 lenlim = M_TRAILINGSPACE(n);
1473                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1474                 /* XXX: reset mbuf length */
1475                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1476                     sizeof(struct icmp6_nodeinfo) + copied;
1477                 break;
1478         }
1479         default:
1480                 break;          /* XXX impossible! */
1481         }
1482
1483         nni6->ni_type = ICMP6_NI_REPLY;
1484         m_freem(m);
1485         return (n);
1486
1487   bad:
1488         m_freem(m);
1489         if (n)
1490                 m_freem(n);
1491         return (NULL);
1492 }
1493 #undef hostnamelen
1494
1495 /*
1496  * make a mbuf with DNS-encoded string.  no compression support.
1497  *
1498  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1499  * treated as truncated name (two \0 at the end).  this is a wild guess.
1500  *
1501  * old - return pascal string if non-zero
1502  */
1503 static struct mbuf *
1504 ni6_nametodns(const char *name, int namelen, int old)
1505 {
1506         struct mbuf *m;
1507         char *cp, *ep;
1508         const char *p, *q;
1509         int i, len, nterm;
1510
1511         if (old)
1512                 len = namelen + 1;
1513         else
1514                 len = MCLBYTES;
1515
1516         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1517         MGET(m, M_DONTWAIT, MT_DATA);
1518         if (m && len > MLEN) {
1519                 MCLGET(m, M_DONTWAIT);
1520                 if ((m->m_flags & M_EXT) == 0)
1521                         goto fail;
1522         }
1523         if (!m)
1524                 goto fail;
1525         m->m_next = NULL;
1526
1527         if (old) {
1528                 m->m_len = len;
1529                 *mtod(m, char *) = namelen;
1530                 bcopy(name, mtod(m, char *) + 1, namelen);
1531                 return m;
1532         } else {
1533                 m->m_len = 0;
1534                 cp = mtod(m, char *);
1535                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1536
1537                 /* if not certain about my name, return empty buffer */
1538                 if (namelen == 0)
1539                         return m;
1540
1541                 /*
1542                  * guess if it looks like shortened hostname, or FQDN.
1543                  * shortened hostname needs two trailing "\0".
1544                  */
1545                 i = 0;
1546                 for (p = name; p < name + namelen; p++) {
1547                         if (*p && *p == '.')
1548                                 i++;
1549                 }
1550                 if (i < 2)
1551                         nterm = 2;
1552                 else
1553                         nterm = 1;
1554
1555                 p = name;
1556                 while (cp < ep && p < name + namelen) {
1557                         i = 0;
1558                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1559                                 i++;
1560                         /* result does not fit into mbuf */
1561                         if (cp + i + 1 >= ep)
1562                                 goto fail;
1563                         /*
1564                          * DNS label length restriction, RFC1035 page 8.
1565                          * "i == 0" case is included here to avoid returning
1566                          * 0-length label on "foo..bar".
1567                          */
1568                         if (i <= 0 || i >= 64)
1569                                 goto fail;
1570                         *cp++ = i;
1571                         bcopy(p, cp, i);
1572                         cp += i;
1573                         p = q;
1574                         if (p < name + namelen && *p == '.')
1575                                 p++;
1576                 }
1577                 /* termination */
1578                 if (cp + nterm >= ep)
1579                         goto fail;
1580                 while (nterm-- > 0)
1581                         *cp++ = '\0';
1582                 m->m_len = cp - mtod(m, char *);
1583                 return m;
1584         }
1585
1586         panic("should not reach here");
1587         /* NOTREACHED */
1588
1589  fail:
1590         if (m)
1591                 m_freem(m);
1592         return NULL;
1593 }
1594
1595 /*
1596  * check if two DNS-encoded string matches.  takes care of truncated
1597  * form (with \0\0 at the end).  no compression support.
1598  * XXX upper/lowercase match (see RFC2065)
1599  */
1600 static int
1601 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1602 {
1603         const char *a0, *b0;
1604         int l;
1605
1606         /* simplest case - need validation? */
1607         if (alen == blen && bcmp(a, b, alen) == 0)
1608                 return 1;
1609
1610         a0 = a;
1611         b0 = b;
1612
1613         /* termination is mandatory */
1614         if (alen < 2 || blen < 2)
1615                 return 0;
1616         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1617                 return 0;
1618         alen--;
1619         blen--;
1620
1621         while (a - a0 < alen && b - b0 < blen) {
1622                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1623                         return 0;
1624
1625                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1626                         return 0;
1627                 /* we don't support compression yet */
1628                 if (a[0] >= 64 || b[0] >= 64)
1629                         return 0;
1630
1631                 /* truncated case */
1632                 if (a[0] == 0 && a - a0 == alen - 1)
1633                         return 1;
1634                 if (b[0] == 0 && b - b0 == blen - 1)
1635                         return 1;
1636                 if (a[0] == 0 || b[0] == 0)
1637                         return 0;
1638
1639                 if (a[0] != b[0])
1640                         return 0;
1641                 l = a[0];
1642                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1643                         return 0;
1644                 if (bcmp(a + 1, b + 1, l) != 0)
1645                         return 0;
1646
1647                 a += 1 + l;
1648                 b += 1 + l;
1649         }
1650
1651         if (a - a0 == alen && b - b0 == blen)
1652                 return 1;
1653         else
1654                 return 0;
1655 }
1656
1657 /*
1658  * calculate the number of addresses to be returned in the node info reply.
1659  */
1660 static int
1661 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1662     struct in6_addr *subj)
1663 {
1664         INIT_VNET_NET(curvnet);
1665         INIT_VNET_INET6(curvnet);
1666         struct ifnet *ifp;
1667         struct in6_ifaddr *ifa6;
1668         struct ifaddr *ifa;
1669         int addrs = 0, addrsofif, iffound = 0;
1670         int niflags = ni6->ni_flags;
1671
1672         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1673                 switch (ni6->ni_code) {
1674                 case ICMP6_NI_SUBJ_IPV6:
1675                         if (subj == NULL) /* must be impossible... */
1676                                 return (0);
1677                         break;
1678                 default:
1679                         /*
1680                          * XXX: we only support IPv6 subject address for
1681                          * this Qtype.
1682                          */
1683                         return (0);
1684                 }
1685         }
1686
1687         IFNET_RLOCK();
1688         for (ifp = TAILQ_FIRST(&V_ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1689                 addrsofif = 0;
1690                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1691                         if (ifa->ifa_addr->sa_family != AF_INET6)
1692                                 continue;
1693                         ifa6 = (struct in6_ifaddr *)ifa;
1694
1695                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1696                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1697                                 iffound = 1;
1698
1699                         /*
1700                          * IPv4-mapped addresses can only be returned by a
1701                          * Node Information proxy, since they represent
1702                          * addresses of IPv4-only nodes, which perforce do
1703                          * not implement this protocol.
1704                          * [icmp-name-lookups-07, Section 5.4]
1705                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1706                          * this function at this moment.
1707                          */
1708
1709                         /* What do we have to do about ::1? */
1710                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1711                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1712                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1713                                         continue;
1714                                 break;
1715                         case IPV6_ADDR_SCOPE_SITELOCAL:
1716                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1717                                         continue;
1718                                 break;
1719                         case IPV6_ADDR_SCOPE_GLOBAL:
1720                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1721                                         continue;
1722                                 break;
1723                         default:
1724                                 continue;
1725                         }
1726
1727                         /*
1728                          * check if anycast is okay.
1729                          * XXX: just experimental.  not in the spec.
1730                          */
1731                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1732                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1733                                 continue; /* we need only unicast addresses */
1734                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1735                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1736                                 continue;
1737                         }
1738                         addrsofif++; /* count the address */
1739                 }
1740                 if (iffound) {
1741                         *ifpp = ifp;
1742                         IFNET_RUNLOCK();
1743                         return (addrsofif);
1744                 }
1745
1746                 addrs += addrsofif;
1747         }
1748         IFNET_RUNLOCK();
1749
1750         return (addrs);
1751 }
1752
1753 static int
1754 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1755     struct ifnet *ifp0, int resid)
1756 {
1757         INIT_VNET_NET(curvnet);
1758         INIT_VNET_INET6(curvnet);
1759         struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1760         struct in6_ifaddr *ifa6;
1761         struct ifaddr *ifa;
1762         struct ifnet *ifp_dep = NULL;
1763         int copied = 0, allow_deprecated = 0;
1764         u_char *cp = (u_char *)(nni6 + 1);
1765         int niflags = ni6->ni_flags;
1766         u_int32_t ltime;
1767
1768         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1769                 return (0);     /* needless to copy */
1770
1771         IFNET_RLOCK();
1772   again:
1773
1774         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1775                 for (ifa = ifp->if_addrlist.tqh_first; ifa;
1776                      ifa = ifa->ifa_list.tqe_next) {
1777                         if (ifa->ifa_addr->sa_family != AF_INET6)
1778                                 continue;
1779                         ifa6 = (struct in6_ifaddr *)ifa;
1780
1781                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1782                             allow_deprecated == 0) {
1783                                 /*
1784                                  * prefererred address should be put before
1785                                  * deprecated addresses.
1786                                  */
1787
1788                                 /* record the interface for later search */
1789                                 if (ifp_dep == NULL)
1790                                         ifp_dep = ifp;
1791
1792                                 continue;
1793                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1794                             allow_deprecated != 0)
1795                                 continue; /* we now collect deprecated addrs */
1796
1797                         /* What do we have to do about ::1? */
1798                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1799                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1800                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1801                                         continue;
1802                                 break;
1803                         case IPV6_ADDR_SCOPE_SITELOCAL:
1804                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1805                                         continue;
1806                                 break;
1807                         case IPV6_ADDR_SCOPE_GLOBAL:
1808                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1809                                         continue;
1810                                 break;
1811                         default:
1812                                 continue;
1813                         }
1814
1815                         /*
1816                          * check if anycast is okay.
1817                          * XXX: just experimental.  not in the spec.
1818                          */
1819                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1820                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1821                                 continue;
1822                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1823                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1824                                 continue;
1825                         }
1826
1827                         /* now we can copy the address */
1828                         if (resid < sizeof(struct in6_addr) +
1829                             sizeof(u_int32_t)) {
1830                                 /*
1831                                  * We give up much more copy.
1832                                  * Set the truncate flag and return.
1833                                  */
1834                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1835                                 IFNET_RUNLOCK();
1836                                 return (copied);
1837                         }
1838
1839                         /*
1840                          * Set the TTL of the address.
1841                          * The TTL value should be one of the following
1842                          * according to the specification:
1843                          *
1844                          * 1. The remaining lifetime of a DHCP lease on the
1845                          *    address, or
1846                          * 2. The remaining Valid Lifetime of a prefix from
1847                          *    which the address was derived through Stateless
1848                          *    Autoconfiguration.
1849                          *
1850                          * Note that we currently do not support stateful
1851                          * address configuration by DHCPv6, so the former
1852                          * case can't happen.
1853                          */
1854                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1855                                 ltime = ND6_INFINITE_LIFETIME;
1856                         else {
1857                                 if (ifa6->ia6_lifetime.ia6t_expire >
1858                                     time_second)
1859                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1860                                 else
1861                                         ltime = 0;
1862                         }
1863
1864                         bcopy(&ltime, cp, sizeof(u_int32_t));
1865                         cp += sizeof(u_int32_t);
1866
1867                         /* copy the address itself */
1868                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1869                             sizeof(struct in6_addr));
1870                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1871                         cp += sizeof(struct in6_addr);
1872
1873                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1874                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1875                 }
1876                 if (ifp0)       /* we need search only on the specified IF */
1877                         break;
1878         }
1879
1880         if (allow_deprecated == 0 && ifp_dep != NULL) {
1881                 ifp = ifp_dep;
1882                 allow_deprecated = 1;
1883
1884                 goto again;
1885         }
1886
1887         IFNET_RUNLOCK();
1888
1889         return (copied);
1890 }
1891
1892 /*
1893  * XXX almost dup'ed code with rip6_input.
1894  */
1895 static int
1896 icmp6_rip6_input(struct mbuf **mp, int off)
1897 {
1898         INIT_VNET_INET(curvnet);
1899         INIT_VNET_INET6(curvnet);
1900         struct mbuf *m = *mp;
1901         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1902         struct inpcb *in6p;
1903         struct inpcb *last = NULL;
1904         struct sockaddr_in6 fromsa;
1905         struct icmp6_hdr *icmp6;
1906         struct mbuf *opts = NULL;
1907
1908 #ifndef PULLDOWN_TEST
1909         /* this is assumed to be safe. */
1910         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1911 #else
1912         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1913         if (icmp6 == NULL) {
1914                 /* m is already reclaimed */
1915                 return (IPPROTO_DONE);
1916         }
1917 #endif
1918
1919         /*
1920          * XXX: the address may have embedded scope zone ID, which should be
1921          * hidden from applications.
1922          */
1923         bzero(&fromsa, sizeof(fromsa));
1924         fromsa.sin6_family = AF_INET6;
1925         fromsa.sin6_len = sizeof(struct sockaddr_in6);
1926         fromsa.sin6_addr = ip6->ip6_src;
1927         if (sa6_recoverscope(&fromsa)) {
1928                 m_freem(m);
1929                 return (IPPROTO_DONE);
1930         }
1931
1932         INP_INFO_RLOCK(&V_ripcbinfo);
1933         LIST_FOREACH(in6p, &V_ripcb, inp_list) {
1934                 if ((in6p->inp_vflag & INP_IPV6) == 0)
1935                         continue;
1936                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
1937                         continue;
1938                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1939                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1940                         continue;
1941                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1942                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1943                         continue;
1944                 INP_RLOCK(in6p);
1945                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1946                     in6p->in6p_icmp6filt)) {
1947                         INP_RUNLOCK(in6p);
1948                         continue;
1949                 }
1950                 if (last != NULL) {
1951                         struct  mbuf *n = NULL;
1952
1953                         /*
1954                          * Recent network drivers tend to allocate a single
1955                          * mbuf cluster, rather than to make a couple of
1956                          * mbufs without clusters.  Also, since the IPv6 code
1957                          * path tries to avoid m_pullup(), it is highly
1958                          * probable that we still have an mbuf cluster here
1959                          * even though the necessary length can be stored in an
1960                          * mbuf's internal buffer.
1961                          * Meanwhile, the default size of the receive socket
1962                          * buffer for raw sockets is not so large.  This means
1963                          * the possibility of packet loss is relatively higher
1964                          * than before.  To avoid this scenario, we copy the
1965                          * received data to a separate mbuf that does not use
1966                          * a cluster, if possible.
1967                          * XXX: it is better to copy the data after stripping
1968                          * intermediate headers.
1969                          */
1970                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1971                             m->m_len <= MHLEN) {
1972                                 MGET(n, M_DONTWAIT, m->m_type);
1973                                 if (n != NULL) {
1974                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1975                                                 bcopy(m->m_data, n->m_data,
1976                                                       m->m_len);
1977                                                 n->m_len = m->m_len;
1978                                         } else {
1979                                                 m_free(n);
1980                                                 n = NULL;
1981                                         }
1982                                 }
1983                         }
1984                         if (n != NULL ||
1985                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1986                                 if (last->inp_flags & INP_CONTROLOPTS)
1987                                         ip6_savecontrol(last, n, &opts);
1988                                 /* strip intermediate headers */
1989                                 m_adj(n, off);
1990                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1991                                 if (sbappendaddr_locked(
1992                                     &last->inp_socket->so_rcv,
1993                                     (struct sockaddr *)&fromsa, n, opts)
1994                                     == 0) {
1995                                         /* should notify about lost packet */
1996                                         m_freem(n);
1997                                         if (opts) {
1998                                                 m_freem(opts);
1999                                         }
2000                                         SOCKBUF_UNLOCK(
2001                                             &last->inp_socket->so_rcv);
2002                                 } else
2003                                         sorwakeup_locked(last->inp_socket);
2004                                 opts = NULL;
2005                         }
2006                         INP_RUNLOCK(last);
2007                 }
2008                 last = in6p;
2009         }
2010         INP_INFO_RUNLOCK(&V_ripcbinfo);
2011         if (last != NULL) {
2012                 if (last->inp_flags & INP_CONTROLOPTS)
2013                         ip6_savecontrol(last, m, &opts);
2014                 /* strip intermediate headers */
2015                 m_adj(m, off);
2016
2017                 /* avoid using mbuf clusters if possible (see above) */
2018                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2019                     m->m_len <= MHLEN) {
2020                         struct mbuf *n;
2021
2022                         MGET(n, M_DONTWAIT, m->m_type);
2023                         if (n != NULL) {
2024                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2025                                         bcopy(m->m_data, n->m_data, m->m_len);
2026                                         n->m_len = m->m_len;
2027
2028                                         m_freem(m);
2029                                         m = n;
2030                                 } else {
2031                                         m_freem(n);
2032                                         n = NULL;
2033                                 }
2034                         }
2035                 }
2036                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2037                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2038                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2039                         m_freem(m);
2040                         if (opts)
2041                                 m_freem(opts);
2042                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2043                 } else
2044                         sorwakeup_locked(last->inp_socket);
2045                 INP_RUNLOCK(last);
2046         } else {
2047                 m_freem(m);
2048                 V_ip6stat.ip6s_delivered--;
2049         }
2050         return IPPROTO_DONE;
2051 }
2052
2053 /*
2054  * Reflect the ip6 packet back to the source.
2055  * OFF points to the icmp6 header, counted from the top of the mbuf.
2056  */
2057 void
2058 icmp6_reflect(struct mbuf *m, size_t off)
2059 {
2060         INIT_VNET_INET6(curvnet);
2061         struct ip6_hdr *ip6;
2062         struct icmp6_hdr *icmp6;
2063         struct in6_ifaddr *ia;
2064         int plen;
2065         int type, code;
2066         struct ifnet *outif = NULL;
2067         struct in6_addr origdst, *src = NULL;
2068
2069         /* too short to reflect */
2070         if (off < sizeof(struct ip6_hdr)) {
2071                 nd6log((LOG_DEBUG,
2072                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2073                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2074                     __FILE__, __LINE__));
2075                 goto bad;
2076         }
2077
2078         /*
2079          * If there are extra headers between IPv6 and ICMPv6, strip
2080          * off that header first.
2081          */
2082 #ifdef DIAGNOSTIC
2083         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2084                 panic("assumption failed in icmp6_reflect");
2085 #endif
2086         if (off > sizeof(struct ip6_hdr)) {
2087                 size_t l;
2088                 struct ip6_hdr nip6;
2089
2090                 l = off - sizeof(struct ip6_hdr);
2091                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2092                 m_adj(m, l);
2093                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2094                 if (m->m_len < l) {
2095                         if ((m = m_pullup(m, l)) == NULL)
2096                                 return;
2097                 }
2098                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2099         } else /* off == sizeof(struct ip6_hdr) */ {
2100                 size_t l;
2101                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2102                 if (m->m_len < l) {
2103                         if ((m = m_pullup(m, l)) == NULL)
2104                                 return;
2105                 }
2106         }
2107         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2108         ip6 = mtod(m, struct ip6_hdr *);
2109         ip6->ip6_nxt = IPPROTO_ICMPV6;
2110         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2111         type = icmp6->icmp6_type; /* keep type for statistics */
2112         code = icmp6->icmp6_code; /* ditto. */
2113
2114         origdst = ip6->ip6_dst;
2115         /*
2116          * ip6_input() drops a packet if its src is multicast.
2117          * So, the src is never multicast.
2118          */
2119         ip6->ip6_dst = ip6->ip6_src;
2120
2121         /*
2122          * If the incoming packet was addressed directly to us (i.e. unicast),
2123          * use dst as the src for the reply.
2124          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2125          * (for example) when we encounter an error while forwarding procedure
2126          * destined to a duplicated address of ours.
2127          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2128          * procedure of an outgoing packet of our own, in which case we need
2129          * to search in the ifaddr list.
2130          */
2131         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2132                 if ((ia = ip6_getdstifaddr(m))) {
2133                         if (!(ia->ia6_flags &
2134                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2135                                 src = &ia->ia_addr.sin6_addr;
2136                 } else {
2137                         struct sockaddr_in6 d;
2138
2139                         bzero(&d, sizeof(d));
2140                         d.sin6_family = AF_INET6;
2141                         d.sin6_len = sizeof(d);
2142                         d.sin6_addr = origdst;
2143                         ia = (struct in6_ifaddr *)
2144                             ifa_ifwithaddr((struct sockaddr *)&d);
2145                         if (ia &&
2146                             !(ia->ia6_flags &
2147                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2148                                 src = &ia->ia_addr.sin6_addr;
2149                         }
2150                 }
2151         }
2152
2153         if (src == NULL) {
2154                 int e;
2155                 struct sockaddr_in6 sin6;
2156                 struct route_in6 ro;
2157
2158                 /*
2159                  * This case matches to multicasts, our anycast, or unicasts
2160                  * that we do not own.  Select a source address based on the
2161                  * source address of the erroneous packet.
2162                  */
2163                 bzero(&sin6, sizeof(sin6));
2164                 sin6.sin6_family = AF_INET6;
2165                 sin6.sin6_len = sizeof(sin6);
2166                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2167
2168                 bzero(&ro, sizeof(ro));
2169                 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2170                 if (ro.ro_rt)
2171                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2172                 if (src == NULL) {
2173                         char ip6buf[INET6_ADDRSTRLEN];
2174                         nd6log((LOG_DEBUG,
2175                             "icmp6_reflect: source can't be determined: "
2176                             "dst=%s, error=%d\n",
2177                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2178                         goto bad;
2179                 }
2180         }
2181
2182         ip6->ip6_src = *src;
2183         ip6->ip6_flow = 0;
2184         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2185         ip6->ip6_vfc |= IPV6_VERSION;
2186         ip6->ip6_nxt = IPPROTO_ICMPV6;
2187         if (outif)
2188                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2189         else if (m->m_pkthdr.rcvif) {
2190                 /* XXX: This may not be the outgoing interface */
2191                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2192         } else
2193                 ip6->ip6_hlim = V_ip6_defhlim;
2194
2195         icmp6->icmp6_cksum = 0;
2196         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2197             sizeof(struct ip6_hdr), plen);
2198
2199         /*
2200          * XXX option handling
2201          */
2202
2203         m->m_flags &= ~(M_BCAST|M_MCAST);
2204
2205         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2206         if (outif)
2207                 icmp6_ifoutstat_inc(outif, type, code);
2208
2209         return;
2210
2211  bad:
2212         m_freem(m);
2213         return;
2214 }
2215
2216 void
2217 icmp6_fasttimo(void)
2218 {
2219
2220         return;
2221 }
2222
2223 static const char *
2224 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2225     struct in6_addr *tgt6)
2226 {
2227         static char buf[1024];
2228         char ip6bufs[INET6_ADDRSTRLEN];
2229         char ip6bufd[INET6_ADDRSTRLEN];
2230         char ip6buft[INET6_ADDRSTRLEN];
2231         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2232             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2233             ip6_sprintf(ip6buft, tgt6));
2234         return buf;
2235 }
2236
2237 void
2238 icmp6_redirect_input(struct mbuf *m, int off)
2239 {
2240         INIT_VNET_INET6(curvnet);
2241         struct ifnet *ifp;
2242         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2243         struct nd_redirect *nd_rd;
2244         int icmp6len = ntohs(ip6->ip6_plen);
2245         char *lladdr = NULL;
2246         int lladdrlen = 0;
2247         u_char *redirhdr = NULL;
2248         int redirhdrlen = 0;
2249         struct rtentry *rt = NULL;
2250         int is_router;
2251         int is_onlink;
2252         struct in6_addr src6 = ip6->ip6_src;
2253         struct in6_addr redtgt6;
2254         struct in6_addr reddst6;
2255         union nd_opts ndopts;
2256         char ip6buf[INET6_ADDRSTRLEN];
2257
2258         if (!m)
2259                 return;
2260
2261         ifp = m->m_pkthdr.rcvif;
2262
2263         if (!ifp)
2264                 return;
2265
2266         /* XXX if we are router, we don't update route by icmp6 redirect */
2267         if (V_ip6_forwarding)
2268                 goto freeit;
2269         if (!V_icmp6_rediraccept)
2270                 goto freeit;
2271
2272 #ifndef PULLDOWN_TEST
2273         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2274         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2275 #else
2276         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2277         if (nd_rd == NULL) {
2278                 V_icmp6stat.icp6s_tooshort++;
2279                 return;
2280         }
2281 #endif
2282         redtgt6 = nd_rd->nd_rd_target;
2283         reddst6 = nd_rd->nd_rd_dst;
2284
2285         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2286             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2287                 goto freeit;
2288         }
2289
2290         /* validation */
2291         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2292                 nd6log((LOG_ERR,
2293                     "ICMP6 redirect sent from %s rejected; "
2294                     "must be from linklocal\n",
2295                     ip6_sprintf(ip6buf, &src6)));
2296                 goto bad;
2297         }
2298         if (ip6->ip6_hlim != 255) {
2299                 nd6log((LOG_ERR,
2300                     "ICMP6 redirect sent from %s rejected; "
2301                     "hlim=%d (must be 255)\n",
2302                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2303                 goto bad;
2304         }
2305     {
2306         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2307         struct sockaddr_in6 sin6;
2308         struct in6_addr *gw6;
2309
2310         bzero(&sin6, sizeof(sin6));
2311         sin6.sin6_family = AF_INET6;
2312         sin6.sin6_len = sizeof(struct sockaddr_in6);
2313         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2314         rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2315         if (rt) {
2316                 if (rt->rt_gateway == NULL ||
2317                     rt->rt_gateway->sa_family != AF_INET6) {
2318                         nd6log((LOG_ERR,
2319                             "ICMP6 redirect rejected; no route "
2320                             "with inet6 gateway found for redirect dst: %s\n",
2321                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2322                         RTFREE_LOCKED(rt);
2323                         goto bad;
2324                 }
2325
2326                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2327                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2328                         nd6log((LOG_ERR,
2329                             "ICMP6 redirect rejected; "
2330                             "not equal to gw-for-src=%s (must be same): "
2331                             "%s\n",
2332                             ip6_sprintf(ip6buf, gw6),
2333                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2334                         RTFREE_LOCKED(rt);
2335                         goto bad;
2336                 }
2337         } else {
2338                 nd6log((LOG_ERR,
2339                     "ICMP6 redirect rejected; "
2340                     "no route found for redirect dst: %s\n",
2341                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2342                 goto bad;
2343         }
2344         RTFREE_LOCKED(rt);
2345         rt = NULL;
2346     }
2347         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2348                 nd6log((LOG_ERR,
2349                     "ICMP6 redirect rejected; "
2350                     "redirect dst must be unicast: %s\n",
2351                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2352                 goto bad;
2353         }
2354
2355         is_router = is_onlink = 0;
2356         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2357                 is_router = 1;  /* router case */
2358         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2359                 is_onlink = 1;  /* on-link destination case */
2360         if (!is_router && !is_onlink) {
2361                 nd6log((LOG_ERR,
2362                     "ICMP6 redirect rejected; "
2363                     "neither router case nor onlink case: %s\n",
2364                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2365                 goto bad;
2366         }
2367         /* validation passed */
2368
2369         icmp6len -= sizeof(*nd_rd);
2370         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2371         if (nd6_options(&ndopts) < 0) {
2372                 nd6log((LOG_INFO, "icmp6_redirect_input: "
2373                     "invalid ND option, rejected: %s\n",
2374                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2375                 /* nd6_options have incremented stats */
2376                 goto freeit;
2377         }
2378
2379         if (ndopts.nd_opts_tgt_lladdr) {
2380                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2381                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2382         }
2383
2384         if (ndopts.nd_opts_rh) {
2385                 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2386                 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2387         }
2388
2389         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2390                 nd6log((LOG_INFO,
2391                     "icmp6_redirect_input: lladdrlen mismatch for %s "
2392                     "(if %d, icmp6 packet %d): %s\n",
2393                     ip6_sprintf(ip6buf, &redtgt6),
2394                     ifp->if_addrlen, lladdrlen - 2,
2395                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2396                 goto bad;
2397         }
2398
2399         /* RFC 2461 8.3 */
2400         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2401             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2402
2403         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2404                 /* perform rtredirect */
2405                 struct sockaddr_in6 sdst;
2406                 struct sockaddr_in6 sgw;
2407                 struct sockaddr_in6 ssrc;
2408
2409                 bzero(&sdst, sizeof(sdst));
2410                 bzero(&sgw, sizeof(sgw));
2411                 bzero(&ssrc, sizeof(ssrc));
2412                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2413                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2414                         sizeof(struct sockaddr_in6);
2415                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2416                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2417                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2418                 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2419                     (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2420                     (struct sockaddr *)&ssrc);
2421         }
2422         /* finally update cached route in each socket via pfctlinput */
2423     {
2424         struct sockaddr_in6 sdst;
2425
2426         bzero(&sdst, sizeof(sdst));
2427         sdst.sin6_family = AF_INET6;
2428         sdst.sin6_len = sizeof(struct sockaddr_in6);
2429         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2430         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2431 #ifdef IPSEC
2432         key_sa_routechange((struct sockaddr *)&sdst);
2433 #endif /* IPSEC */
2434     }
2435
2436  freeit:
2437         m_freem(m);
2438         return;
2439
2440  bad:
2441         V_icmp6stat.icp6s_badredirect++;
2442         m_freem(m);
2443 }
2444
2445 void
2446 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2447 {
2448         INIT_VNET_INET6(curvnet);
2449         struct ifnet *ifp;      /* my outgoing interface */
2450         struct in6_addr *ifp_ll6;
2451         struct in6_addr *router_ll6;
2452         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2453         struct mbuf *m = NULL;  /* newly allocated one */
2454         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2455         struct nd_redirect *nd_rd;
2456         struct llentry *ln = NULL;
2457         size_t maxlen;
2458         u_char *p;
2459         struct ifnet *outif = NULL;
2460         struct sockaddr_in6 src_sa;
2461
2462         icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2463
2464         /* if we are not router, we don't send icmp6 redirect */
2465         if (!V_ip6_forwarding)
2466                 goto fail;
2467
2468         /* sanity check */
2469         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2470                 goto fail;
2471
2472         /*
2473          * Address check:
2474          *  the source address must identify a neighbor, and
2475          *  the destination address must not be a multicast address
2476          *  [RFC 2461, sec 8.2]
2477          */
2478         sip6 = mtod(m0, struct ip6_hdr *);
2479         bzero(&src_sa, sizeof(src_sa));
2480         src_sa.sin6_family = AF_INET6;
2481         src_sa.sin6_len = sizeof(src_sa);
2482         src_sa.sin6_addr = sip6->ip6_src;
2483         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2484                 goto fail;
2485         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2486                 goto fail;      /* what should we do here? */
2487
2488         /* rate limit */
2489         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2490                 goto fail;
2491
2492         /*
2493          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2494          * we almost always ask for an mbuf cluster for simplicity.
2495          * (MHLEN < IPV6_MMTU is almost always true)
2496          */
2497 #if IPV6_MMTU >= MCLBYTES
2498 # error assumption failed about IPV6_MMTU and MCLBYTES
2499 #endif
2500         MGETHDR(m, M_DONTWAIT, MT_HEADER);
2501         if (m && IPV6_MMTU >= MHLEN)
2502                 MCLGET(m, M_DONTWAIT);
2503         if (!m)
2504                 goto fail;
2505         m->m_pkthdr.rcvif = NULL;
2506         m->m_len = 0;
2507         maxlen = M_TRAILINGSPACE(m);
2508         maxlen = min(IPV6_MMTU, maxlen);
2509         /* just for safety */
2510         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2511             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2512                 goto fail;
2513         }
2514
2515         {
2516                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2517                 struct in6_ifaddr *ia;
2518                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2519                                                  IN6_IFF_NOTREADY|
2520                                                  IN6_IFF_ANYCAST)) == NULL)
2521                         goto fail;
2522                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2523         }
2524
2525         /* get ip6 linklocal address for the router. */
2526         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2527                 struct sockaddr_in6 *sin6;
2528                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2529                 router_ll6 = &sin6->sin6_addr;
2530                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2531                         router_ll6 = (struct in6_addr *)NULL;
2532         } else
2533                 router_ll6 = (struct in6_addr *)NULL;
2534
2535         /* ip6 */
2536         ip6 = mtod(m, struct ip6_hdr *);
2537         ip6->ip6_flow = 0;
2538         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2539         ip6->ip6_vfc |= IPV6_VERSION;
2540         /* ip6->ip6_plen will be set later */
2541         ip6->ip6_nxt = IPPROTO_ICMPV6;
2542         ip6->ip6_hlim = 255;
2543         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2544         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2545         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2546
2547         /* ND Redirect */
2548         nd_rd = (struct nd_redirect *)(ip6 + 1);
2549         nd_rd->nd_rd_type = ND_REDIRECT;
2550         nd_rd->nd_rd_code = 0;
2551         nd_rd->nd_rd_reserved = 0;
2552         if (rt->rt_flags & RTF_GATEWAY) {
2553                 /*
2554                  * nd_rd->nd_rd_target must be a link-local address in
2555                  * better router cases.
2556                  */
2557                 if (!router_ll6)
2558                         goto fail;
2559                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2560                     sizeof(nd_rd->nd_rd_target));
2561                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2562                     sizeof(nd_rd->nd_rd_dst));
2563         } else {
2564                 /* make sure redtgt == reddst */
2565                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2566                     sizeof(nd_rd->nd_rd_target));
2567                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2568                     sizeof(nd_rd->nd_rd_dst));
2569         }
2570
2571         p = (u_char *)(nd_rd + 1);
2572
2573         if (!router_ll6)
2574                 goto nolladdropt;
2575
2576         {
2577                 /* target lladdr option */
2578                 int len;
2579                 struct nd_opt_hdr *nd_opt;
2580                 char *lladdr;
2581
2582                 IF_AFDATA_LOCK(ifp);
2583                 ln = nd6_lookup(router_ll6, 0, ifp);
2584                 IF_AFDATA_UNLOCK(ifp);
2585                 if (ln == NULL)
2586                         goto nolladdropt;
2587
2588                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2589                 len = (len + 7) & ~7;   /* round by 8 */
2590                 /* safety check */
2591                 if (len + (p - (u_char *)ip6) > maxlen)
2592                         goto nolladdropt;
2593
2594                 if (ln->la_flags & LLE_VALID) {
2595                         nd_opt = (struct nd_opt_hdr *)p;
2596                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2597                         nd_opt->nd_opt_len = len >> 3;
2598                         lladdr = (char *)(nd_opt + 1);
2599                         bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2600                         p += len;
2601                 }
2602         }
2603 nolladdropt:
2604         if (ln != NULL)
2605                 LLE_RUNLOCK(ln);
2606
2607         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2608
2609         /* just to be safe */
2610 #ifdef M_DECRYPTED      /*not openbsd*/
2611         if (m0->m_flags & M_DECRYPTED)
2612                 goto noredhdropt;
2613 #endif
2614         if (p - (u_char *)ip6 > maxlen)
2615                 goto noredhdropt;
2616
2617         {
2618                 /* redirected header option */
2619                 int len;
2620                 struct nd_opt_rd_hdr *nd_opt_rh;
2621
2622                 /*
2623                  * compute the maximum size for icmp6 redirect header option.
2624                  * XXX room for auth header?
2625                  */
2626                 len = maxlen - (p - (u_char *)ip6);
2627                 len &= ~7;
2628
2629                 /* This is just for simplicity. */
2630                 if (m0->m_pkthdr.len != m0->m_len) {
2631                         if (m0->m_next) {
2632                                 m_freem(m0->m_next);
2633                                 m0->m_next = NULL;
2634                         }
2635                         m0->m_pkthdr.len = m0->m_len;
2636                 }
2637
2638                 /*
2639                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2640                  * about padding/truncate rule for the original IP packet.
2641                  * From the discussion on IPv6imp in Feb 1999,
2642                  * the consensus was:
2643                  * - "attach as much as possible" is the goal
2644                  * - pad if not aligned (original size can be guessed by
2645                  *   original ip6 header)
2646                  * Following code adds the padding if it is simple enough,
2647                  * and truncates if not.
2648                  */
2649                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2650                         panic("assumption failed in %s:%d", __FILE__,
2651                             __LINE__);
2652
2653                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2654                         /* not enough room, truncate */
2655                         m0->m_pkthdr.len = m0->m_len = len -
2656                             sizeof(*nd_opt_rh);
2657                 } else {
2658                         /* enough room, pad or truncate */
2659                         size_t extra;
2660
2661                         extra = m0->m_pkthdr.len % 8;
2662                         if (extra) {
2663                                 /* pad if easy enough, truncate if not */
2664                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2665                                         /* pad */
2666                                         m0->m_len += (8 - extra);
2667                                         m0->m_pkthdr.len += (8 - extra);
2668                                 } else {
2669                                         /* truncate */
2670                                         m0->m_pkthdr.len -= extra;
2671                                         m0->m_len -= extra;
2672                                 }
2673                         }
2674                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2675                         m0->m_pkthdr.len = m0->m_len = len -
2676                             sizeof(*nd_opt_rh);
2677                 }
2678
2679                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2680                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2681                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2682                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2683                 p += sizeof(*nd_opt_rh);
2684                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2685
2686                 /* connect m0 to m */
2687                 m_tag_delete_chain(m0, NULL);
2688                 m0->m_flags &= ~M_PKTHDR;
2689                 m->m_next = m0;
2690                 m->m_pkthdr.len = m->m_len + m0->m_len;
2691                 m0 = NULL;
2692         }
2693 noredhdropt:;
2694         if (m0) {
2695                 m_freem(m0);
2696                 m0 = NULL;
2697         }
2698
2699         /* XXX: clear embedded link IDs in the inner header */
2700         in6_clearscope(&sip6->ip6_src);
2701         in6_clearscope(&sip6->ip6_dst);
2702         in6_clearscope(&nd_rd->nd_rd_target);
2703         in6_clearscope(&nd_rd->nd_rd_dst);
2704
2705         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2706
2707         nd_rd->nd_rd_cksum = 0;
2708         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2709             sizeof(*ip6), ntohs(ip6->ip6_plen));
2710
2711         /* send the packet to outside... */
2712         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2713         if (outif) {
2714                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2715                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2716         }
2717         V_icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2718
2719         return;
2720
2721 fail:
2722         if (m)
2723                 m_freem(m);
2724         if (m0)
2725                 m_freem(m0);
2726 }
2727
2728 /*
2729  * ICMPv6 socket option processing.
2730  */
2731 int
2732 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2733 {
2734         int error = 0;
2735         int optlen;
2736         struct inpcb *inp = sotoinpcb(so);
2737         int level, op, optname;
2738
2739         if (sopt) {
2740                 level = sopt->sopt_level;
2741                 op = sopt->sopt_dir;
2742                 optname = sopt->sopt_name;
2743                 optlen = sopt->sopt_valsize;
2744         } else
2745                 level = op = optname = optlen = 0;
2746
2747         if (level != IPPROTO_ICMPV6) {
2748                 return EINVAL;
2749         }
2750
2751         switch (op) {
2752         case PRCO_SETOPT:
2753                 switch (optname) {
2754                 case ICMP6_FILTER:
2755                     {
2756                         struct icmp6_filter ic6f;
2757
2758                         if (optlen != sizeof(ic6f)) {
2759                                 error = EMSGSIZE;
2760                                 break;
2761                         }
2762                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2763                         if (error == 0) {
2764                                 INP_WLOCK(inp);
2765                                 *inp->in6p_icmp6filt = ic6f;
2766                                 INP_WUNLOCK(inp);
2767                         }
2768                         break;
2769                     }
2770
2771                 default:
2772                         error = ENOPROTOOPT;
2773                         break;
2774                 }
2775                 break;
2776
2777         case PRCO_GETOPT:
2778                 switch (optname) {
2779                 case ICMP6_FILTER:
2780                     {
2781                         struct icmp6_filter ic6f;
2782
2783                         INP_RLOCK(inp);
2784                         ic6f = *inp->in6p_icmp6filt;
2785                         INP_RUNLOCK(inp);
2786                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2787                         break;
2788                     }
2789
2790                 default:
2791                         error = ENOPROTOOPT;
2792                         break;
2793                 }
2794                 break;
2795         }
2796
2797         return (error);
2798 }
2799
2800 /*
2801  * Perform rate limit check.
2802  * Returns 0 if it is okay to send the icmp6 packet.
2803  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2804  * limitation.
2805  *
2806  * XXX per-destination/type check necessary?
2807  *
2808  * dst - not used at this moment
2809  * type - not used at this moment
2810  * code - not used at this moment
2811  */
2812 static int
2813 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2814     const int code)
2815 {
2816         INIT_VNET_INET6(curvnet);
2817         int ret;
2818
2819         ret = 0;        /* okay to send */
2820
2821         /* PPS limit */
2822         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2823             V_icmp6errppslim)) {
2824                 /* The packet is subject to rate limit */
2825                 ret++;
2826         }
2827
2828         return ret;
2829 }