sys/netinet6/icmp6.c

   1 /*      $FreeBSD: src/sys/netinet6/icmp6.c,v 1.6.2.13 2003/05/06 06:46:58 suz Exp $     */
   2 /*      $DragonFly: src/sys/netinet6/icmp6.c,v 1.31 2008/10/27 02:56:30 sephe Exp $     */
   3 /*      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $ */
   4
   5 /*
   6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   7  * All rights reserved.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. Neither the name of the project nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  */
  33
  34 /*
  35  * Copyright (c) 1982, 1986, 1988, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. All advertising materials mentioning features or use of this software
  47  *    must display the following acknowledgement:
  48  *      This product includes software developed by the University of
  49  *      California, Berkeley and its contributors.
  50  * 4. Neither the name of the University nor the names of its contributors
  51  *    may be used to endorse or promote products derived from this software
  52  *    without specific prior written permission.
  53  *
  54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  64  * SUCH DAMAGE.
  65  *
  66  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
  67  */
  68
  69 #include "opt_inet.h"
  70 #include "opt_inet6.h"
  71 #include "opt_ipsec.h"
  72
  73 #include <sys/param.h>
  74 #include <sys/systm.h>
  75 #include <sys/malloc.h>
  76 #include <sys/mbuf.h>
  77 #include <sys/protosw.h>
  78 #include <sys/socket.h>
  79 #include <sys/socketvar.h>
  80 #include <sys/socketops.h>
  81 #include <sys/time.h>
  82 #include <sys/kernel.h>
  83 #include <sys/syslog.h>
  84 #include <sys/domain.h>
  85
  86 #include <sys/thread2.h>
  87 #include <sys/msgport2.h>
  88
  89 #include <net/if.h>
  90 #include <net/route.h>
  91 #include <net/if_dl.h>
  92 #include <net/if_types.h>
  93
  94 #include <netinet/in.h>
  95 #include <netinet/in_var.h>
  96 #include <netinet/ip6.h>
  97 #include <netinet6/ip6_var.h>
  98 #include <netinet/icmp6.h>
  99 #include <netinet6/mld6_var.h>
 100 #include <netinet/in_pcb.h>
 101 #include <netinet6/in6_pcb.h>
 102 #include <netinet6/nd6.h>
 103 #include <netinet6/in6_ifattach.h>
 104 #include <netinet6/ip6protosw.h>
 105
 106 #ifdef IPSEC
 107 #include <netinet6/ipsec.h>
 108 #include <netproto/key/key.h>
 109 #endif
 110
 111 #ifdef FAST_IPSEC
 112 #include <netproto/ipsec/ipsec.h>
 113 #include <netproto/ipsec/key.h>
 114 #define IPSEC
 115 #endif
 116
 117 #include <net/net_osdep.h>
 118
 119 #ifdef HAVE_NRL_INPCB
 120 /* inpcb members */
 121 #define in6pcb          inpcb
 122 #define in6p_laddr      inp_laddr6
 123 #define in6p_faddr      inp_faddr6
 124 #define in6p_icmp6filt  inp_icmp6filt
 125 #define in6p_route      inp_route
 126 #define in6p_socket     inp_socket
 127 #define in6p_flags      inp_flags
 128 #define in6p_moptions   inp_moptions6
 129 #define in6p_outputopts inp_outputopts6
 130 #define in6p_ip6        inp_ipv6
 131 #define in6p_flowinfo   inp_flowinfo
 132 #define in6p_sp         inp_sp
 133 #define in6p_next       inp_next
 134 #define in6p_prev       inp_prev
 135 /* function names */
 136 #define in6_pcbdetach   in_pcbdetach
 137 #define in6_rtchange    in_rtchange
 138
 139 /*
 140  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
 141  * others...
 142  */
 143 #define in6p_ip6_nxt    inp_ipv6.ip6_nxt
 144 #endif
 145
 146 /*
 147  * ppratecheck() is available, so define it here.
 148  */
 149 #define HAVE_PPSRATECHECK
 150
 151 extern struct domain inet6domain;
 152 extern struct protosw inet6sw[];
 153 extern u_char ip6_protox[];
 154
 155 struct icmp6stat icmp6stat;
 156
 157 extern struct inpcbinfo ripcbinfo;
 158 extern int icmp6errppslim;
 159 static int icmp6errpps_count = 0;
 160 static struct timeval icmp6errppslim_last;
 161 extern int icmp6_nodeinfo;
 162
 163 static void icmp6_errcount (struct icmp6errstat *, int, int);
 164 static int icmp6_rip6_input (struct mbuf **, int);
 165 static int icmp6_ratelimit (const struct in6_addr *, const int, const int);
 166 static const char *icmp6_redirect_diag (struct in6_addr *,
 167         struct in6_addr *, struct in6_addr *);
 168 #ifndef HAVE_PPSRATECHECK
 169 static int ppsratecheck (struct timeval *, int *, int);
 170 #endif
 171 static struct mbuf *ni6_input (struct mbuf *, int);
 172 static struct mbuf *ni6_nametodns (const char *, int, int);
 173 static int ni6_dnsmatch (const char *, int, const char *, int);
 174 static int ni6_addrs (struct icmp6_nodeinfo *, struct mbuf *,
 175                           struct ifnet **, char *);
 176 static int ni6_store_addrs (struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
 177                                 struct ifnet *, int);
 178 static int icmp6_notify_error (struct mbuf *, int, int, int);
 179
 180 #ifdef COMPAT_RFC1885
 181 static struct route_in6 icmp6_reflect_rt;
 182 #endif
 183
 184
 185 void
 186 icmp6_init(void)
 187 {
 188         mld6_init();
 189 }
 190
 191 static void
 192 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
 193 {
 194         switch (type) {
 195         case ICMP6_DST_UNREACH:
 196                 switch (code) {
 197                 case ICMP6_DST_UNREACH_NOROUTE:
 198                         stat->icp6errs_dst_unreach_noroute++;
 199                         return;
 200                 case ICMP6_DST_UNREACH_ADMIN:
 201                         stat->icp6errs_dst_unreach_admin++;
 202                         return;
 203                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 204                         stat->icp6errs_dst_unreach_beyondscope++;
 205                         return;
 206                 case ICMP6_DST_UNREACH_ADDR:
 207                         stat->icp6errs_dst_unreach_addr++;
 208                         return;
 209                 case ICMP6_DST_UNREACH_NOPORT:
 210                         stat->icp6errs_dst_unreach_noport++;
 211                         return;
 212                 }
 213                 break;
 214         case ICMP6_PACKET_TOO_BIG:
 215                 stat->icp6errs_packet_too_big++;
 216                 return;
 217         case ICMP6_TIME_EXCEEDED:
 218                 switch (code) {
 219                 case ICMP6_TIME_EXCEED_TRANSIT:
 220                         stat->icp6errs_time_exceed_transit++;
 221                         return;
 222                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 223                         stat->icp6errs_time_exceed_reassembly++;
 224                         return;
 225                 }
 226                 break;
 227         case ICMP6_PARAM_PROB:
 228                 switch (code) {
 229                 case ICMP6_PARAMPROB_HEADER:
 230                         stat->icp6errs_paramprob_header++;
 231                         return;
 232                 case ICMP6_PARAMPROB_NEXTHEADER:
 233                         stat->icp6errs_paramprob_nextheader++;
 234                         return;
 235                 case ICMP6_PARAMPROB_OPTION:
 236                         stat->icp6errs_paramprob_option++;
 237                         return;
 238                 }
 239                 break;
 240         case ND_REDIRECT:
 241                 stat->icp6errs_redirect++;
 242                 return;
 243         }
 244         stat->icp6errs_unknown++;
 245 }
 246
 247 /*
 248  * Generate an error packet of type error in response to bad IP6 packet.
 249  */
 250 void
 251 icmp6_error(struct mbuf *m, int type, int code, int param)
 252 {
 253         struct ip6_hdr *oip6, *nip6;
 254         struct icmp6_hdr *icmp6;
 255         u_int preplen;
 256         int off;
 257         int nxt;
 258
 259         icmp6stat.icp6s_error++;
 260
 261         /* count per-type-code statistics */
 262         icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
 263
 264 #ifdef M_DECRYPTED      /*not openbsd*/
 265         if (m->m_flags & M_DECRYPTED) {
 266                 icmp6stat.icp6s_canterror++;
 267                 goto freeit;
 268         }
 269 #endif
 270
 271 #ifndef PULLDOWN_TEST
 272         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 273 #else
 274         if (m->m_len < sizeof(struct ip6_hdr)) {
 275                 m = m_pullup(m, sizeof(struct ip6_hdr));
 276                 if (m == NULL)
 277                         return;
 278         }
 279 #endif
 280         oip6 = mtod(m, struct ip6_hdr *);
 281
 282         /*
 283          * Multicast destination check. For unrecognized option errors,
 284          * this check has already done in ip6_unknown_opt(), so we can
 285          * check only for other errors.
 286          */
 287         if ((m->m_flags & (M_BCAST|M_MCAST) ||
 288              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
 289             (type != ICMP6_PACKET_TOO_BIG &&
 290              (type != ICMP6_PARAM_PROB ||
 291               code != ICMP6_PARAMPROB_OPTION)))
 292                 goto freeit;
 293
 294         /* Source address check. XXX: the case of anycast source? */
 295         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
 296             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
 297                 goto freeit;
 298
 299         /*
 300          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
 301          * don't do it.
 302          */
 303         nxt = -1;
 304         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
 305         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
 306                 struct icmp6_hdr *icp;
 307
 308 #ifndef PULLDOWN_TEST
 309                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
 310                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 311 #else
 312                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
 313                         sizeof(*icp));
 314                 if (icp == NULL) {
 315                         icmp6stat.icp6s_tooshort++;
 316                         return;
 317                 }
 318 #endif
 319                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
 320                     icp->icmp6_type == ND_REDIRECT) {
 321                         /*
 322                          * ICMPv6 error
 323                          * Special case: for redirect (which is
 324                          * informational) we must not send icmp6 error.
 325                          */
 326                         icmp6stat.icp6s_canterror++;
 327                         goto freeit;
 328                 } else {
 329                         /* ICMPv6 informational - send the error */
 330                 }
 331         } else {
 332                 /* non-ICMPv6 - send the error */
 333         }
 334
 335         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
 336
 337         /* Finally, do rate limitation check. */
 338         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
 339                 icmp6stat.icp6s_toofreq++;
 340                 goto freeit;
 341         }
 342
 343         /*
 344          * OK, ICMP6 can be generated.
 345          */
 346
 347         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
 348                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
 349
 350         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
 351         M_PREPEND(m, preplen, MB_DONTWAIT);
 352         if (m && m->m_len < preplen)
 353                 m = m_pullup(m, preplen);
 354         if (m == NULL) {
 355                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
 356                 return;
 357         }
 358
 359         nip6 = mtod(m, struct ip6_hdr *);
 360         nip6->ip6_src  = oip6->ip6_src;
 361         nip6->ip6_dst  = oip6->ip6_dst;
 362
 363         if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
 364                 oip6->ip6_src.s6_addr16[1] = 0;
 365         if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
 366                 oip6->ip6_dst.s6_addr16[1] = 0;
 367
 368         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
 369         icmp6->icmp6_type = type;
 370         icmp6->icmp6_code = code;
 371         icmp6->icmp6_pptr = htonl((u_int32_t)param);
 372
 373         /*
 374          * icmp6_reflect() is designed to be in the input path.
 375          * icmp6_error() can be called from both input and outut path,
 376          * and if we are in output path rcvif could contain bogus value.
 377          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
 378          * information in ip header (nip6).
 379          */
 380         m->m_pkthdr.rcvif = NULL;
 381
 382         icmp6stat.icp6s_outhist[type]++;
 383         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
 384
 385         return;
 386
 387 freeit:
 388         /*
 389          * If we can't tell wheter or not we can generate ICMP6, free it.
 390          */
 391         m_freem(m);
 392 }
 393
 394 /*
 395  * Process a received ICMP6 message.
 396  */
 397 int
 398 icmp6_input(struct mbuf **mp, int *offp, int proto)
 399 {
 400         struct mbuf *m = *mp, *n;
 401         struct ip6_hdr *ip6, *nip6;
 402         struct icmp6_hdr *icmp6, *nicmp6;
 403         int off = *offp;
 404         int icmp6len = m->m_pkthdr.len - *offp;
 405         int code, sum, noff;
 406
 407 #ifndef PULLDOWN_TEST
 408         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
 409         /* m might change if M_LOOP.  So, call mtod after this */
 410 #endif
 411
 412         /*
 413          * Locate icmp6 structure in mbuf, and check
 414          * that not corrupted and of at least minimum length
 415          */
 416
 417         ip6 = mtod(m, struct ip6_hdr *);
 418         if (icmp6len < sizeof(struct icmp6_hdr)) {
 419                 icmp6stat.icp6s_tooshort++;
 420                 goto freeit;
 421         }
 422
 423         /*
 424          * calculate the checksum
 425          */
 426 #ifndef PULLDOWN_TEST
 427         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
 428 #else
 429         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
 430         if (icmp6 == NULL) {
 431                 icmp6stat.icp6s_tooshort++;
 432                 return IPPROTO_DONE;
 433         }
 434 #endif
 435         code = icmp6->icmp6_code;
 436
 437         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
 438                 nd6log((LOG_ERR,
 439                     "ICMP6 checksum error(%d|%x) %s\n",
 440                     icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
 441                 icmp6stat.icp6s_checksum++;
 442                 goto freeit;
 443         }
 444
 445         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
 446                 /*
 447                  * Deliver very specific ICMP6 type only.
 448                  * This is important to deilver TOOBIG.  Otherwise PMTUD
 449                  * will not work.
 450                  */
 451                 switch (icmp6->icmp6_type) {
 452                 case ICMP6_DST_UNREACH:
 453                 case ICMP6_PACKET_TOO_BIG:
 454                 case ICMP6_TIME_EXCEEDED:
 455                         break;
 456                 default:
 457                         goto freeit;
 458                 }
 459         }
 460
 461         icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
 462         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
 463         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
 464                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
 465
 466         switch (icmp6->icmp6_type) {
 467         case ICMP6_DST_UNREACH:
 468                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
 469                 switch (code) {
 470                 case ICMP6_DST_UNREACH_NOROUTE:
 471                         code = PRC_UNREACH_NET;
 472                         break;
 473                 case ICMP6_DST_UNREACH_ADMIN:
 474                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
 475                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
 476                         break;
 477                 case ICMP6_DST_UNREACH_ADDR:
 478                         code = PRC_HOSTDEAD;
 479                         break;
 480 #ifdef COMPAT_RFC1885
 481                 case ICMP6_DST_UNREACH_NOTNEIGHBOR:
 482                         code = PRC_UNREACH_SRCFAIL;
 483                         break;
 484 #else
 485                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 486                         /* I mean "source address was incorrect." */
 487                         code = PRC_PARAMPROB;
 488                         break;
 489 #endif
 490                 case ICMP6_DST_UNREACH_NOPORT:
 491                         code = PRC_UNREACH_PORT;
 492                         break;
 493                 default:
 494                         goto badcode;
 495                 }
 496                 goto deliver;
 497                 break;
 498
 499         case ICMP6_PACKET_TOO_BIG:
 500                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
 501                 if (code != 0)
 502                         goto badcode;
 503
 504                 code = PRC_MSGSIZE;
 505
 506                 /*
 507                  * Updating the path MTU will be done after examining
 508                  * intermediate extension headers.
 509                  */
 510                 goto deliver;
 511                 break;
 512
 513         case ICMP6_TIME_EXCEEDED:
 514                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
 515                 switch (code) {
 516                 case ICMP6_TIME_EXCEED_TRANSIT:
 517                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 518                         code += PRC_TIMXCEED_INTRANS;
 519                         break;
 520                 default:
 521                         goto badcode;
 522                 }
 523                 goto deliver;
 524                 break;
 525
 526         case ICMP6_PARAM_PROB:
 527                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
 528                 switch (code) {
 529                 case ICMP6_PARAMPROB_NEXTHEADER:
 530                         code = PRC_UNREACH_PROTOCOL;
 531                         break;
 532                 case ICMP6_PARAMPROB_HEADER:
 533                 case ICMP6_PARAMPROB_OPTION:
 534                         code = PRC_PARAMPROB;
 535                         break;
 536                 default:
 537                         goto badcode;
 538                 }
 539                 goto deliver;
 540                 break;
 541
 542         case ICMP6_ECHO_REQUEST:
 543                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
 544                 if (code != 0)
 545                         goto badcode;
 546                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
 547                         /* Give up remote */
 548                         break;
 549                 }
 550                 if ((n->m_flags & M_EXT) ||
 551                     n->m_len < off + sizeof(struct icmp6_hdr)) {
 552                         struct mbuf *n0 = n;
 553                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
 554                         int n0len;
 555
 556                         /*
 557                          * Prepare an internal mbuf. m_pullup() doesn't
 558                          * always copy the length we specified.
 559                          */
 560                         if (maxlen >= MCLBYTES) {
 561                                 /* Give up remote */
 562                                 m_freem(n0);
 563                                 break;
 564                         }
 565                         n = m_getb(maxlen, MB_DONTWAIT, n0->m_type, M_PKTHDR);
 566                         if (n == NULL) {
 567                                 /* Give up remote */
 568                                 m_freem(n0);
 569                                 break;
 570                         }
 571                         n0len = n0->m_pkthdr.len;       /* save for use below */
 572                         M_MOVE_PKTHDR(n, n0);
 573                         /*
 574                          * Copy IPv6 and ICMPv6 only.
 575                          */
 576                         nip6 = mtod(n, struct ip6_hdr *);
 577                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 578                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 579                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 580                         noff = sizeof(struct ip6_hdr);
 581                         /* new mbuf contains only ipv6+icmpv6 headers */
 582                         n->m_len = noff + sizeof(struct icmp6_hdr);
 583                         /*
 584                          * Adjust mbuf. ip6_plen will be adjusted in
 585                          * ip6_output().
 586                          */
 587                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 588                         /* recalculate complete packet size */
 589                         n->m_pkthdr.len = n0len + (noff - off);
 590                         n->m_next = n0;
 591                 } else {
 592                         nip6 = mtod(n, struct ip6_hdr *);
 593                         nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
 594                         noff = off;
 595                 }
 596                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 597                 nicmp6->icmp6_code = 0;
 598                 if (n) {
 599                         icmp6stat.icp6s_reflect++;
 600                         icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
 601                         icmp6_reflect(n, noff);
 602                 }
 603                 break;
 604
 605         case ICMP6_ECHO_REPLY:
 606                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
 607                 if (code != 0)
 608                         goto badcode;
 609                 break;
 610
 611         case MLD_LISTENER_QUERY:
 612         case MLD_LISTENER_REPORT:
 613                 if (icmp6len < sizeof(struct mld_hdr))
 614                         goto badlen;
 615                 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
 616                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
 617                 else
 618                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
 619                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 620                         /* give up local */
 621                         mld6_input(m, off);
 622                         m = NULL;
 623                         goto freeit;
 624                 }
 625                 mld6_input(n, off);
 626                 /* m stays. */
 627                 break;
 628
 629         case MLD_LISTENER_DONE:
 630                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
 631                 if (icmp6len < sizeof(struct mld_hdr))  /* necessary? */
 632                         goto badlen;
 633                 break;          /* nothing to be done in kernel */
 634
 635         case MLD_MTRACE_RESP:
 636         case MLD_MTRACE:
 637                 /* XXX: these two are experimental.  not officially defind. */
 638                 /* XXX: per-interface statistics? */
 639                 break;          /* just pass it to applications */
 640
 641         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 642             {
 643                 enum { WRU, FQDN } mode;
 644
 645                 if (!icmp6_nodeinfo)
 646                         break;
 647
 648                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 649                         mode = WRU;
 650                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 651                         mode = FQDN;
 652                 else
 653                         goto badlen;
 654
 655 #define hostnamelen     strlen(hostname)
 656                 if (mode == FQDN) {
 657 #ifndef PULLDOWN_TEST
 658                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
 659                                          IPPROTO_DONE);
 660 #endif
 661                         n = m_copy(m, 0, M_COPYALL);
 662                         if (n)
 663                                 n = ni6_input(n, off);
 664                         /* XXX meaningless if n == NULL */
 665                         noff = sizeof(struct ip6_hdr);
 666                 } else {
 667                         u_char *p;
 668                         int maxlen, maxhlen;
 669
 670                         if ((icmp6_nodeinfo & 5) != 5)
 671                                 break;
 672
 673                         if (code != 0)
 674                                 goto badcode;
 675                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
 676                         if (maxlen >= MCLBYTES) {
 677                                 /* Give up remote */
 678                                 break;
 679                         }
 680                         n = m_getb(maxlen, MB_DONTWAIT, m->m_type, M_PKTHDR);
 681                         if (n == NULL) {
 682                                 /* Give up remote */
 683                                 break;
 684                         }
 685                         if (!m_dup_pkthdr(n, m, MB_DONTWAIT)) {
 686                                 /*
 687                                  * Previous code did a blind M_COPY_PKTHDR
 688                                  * and said "just for rcvif".  If true, then
 689                                  * we could tolerate the dup failing (due to
 690                                  * the deep copy of the tag chain).  For now
 691                                  * be conservative and just fail.
 692                                  */
 693                                 m_free(n);
 694                                 break;
 695                         }
 696                         n->m_len = 0;
 697                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
 698                         if (maxhlen > hostnamelen)
 699                                 maxhlen = hostnamelen;
 700                         /*
 701                          * Copy IPv6 and ICMPv6 only.
 702                          */
 703                         nip6 = mtod(n, struct ip6_hdr *);
 704                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 705                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 706                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 707                         p = (u_char *)(nicmp6 + 1);
 708                         bzero(p, 4);
 709                         bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
 710                         noff = sizeof(struct ip6_hdr);
 711                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 712                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 713                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 714                         nicmp6->icmp6_code = 0;
 715                 }
 716 #undef hostnamelen
 717                 if (n) {
 718                         icmp6stat.icp6s_reflect++;
 719                         icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
 720                         icmp6_reflect(n, noff);
 721                 }
 722                 break;
 723             }
 724
 725         case ICMP6_WRUREPLY:
 726                 if (code != 0)
 727                         goto badcode;
 728                 break;
 729
 730         case ND_ROUTER_SOLICIT:
 731                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
 732                 if (code != 0)
 733                         goto badcode;
 734                 if (icmp6len < sizeof(struct nd_router_solicit))
 735                         goto badlen;
 736                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 737                         /* give up local */
 738                         nd6_rs_input(m, off, icmp6len);
 739                         m = NULL;
 740                         goto freeit;
 741                 }
 742                 nd6_rs_input(n, off, icmp6len);
 743                 /* m stays. */
 744                 break;
 745
 746         case ND_ROUTER_ADVERT:
 747                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
 748                 if (code != 0)
 749                         goto badcode;
 750                 if (icmp6len < sizeof(struct nd_router_advert))
 751                         goto badlen;
 752                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 753                         /* give up local */
 754                         nd6_ra_input(m, off, icmp6len);
 755                         m = NULL;
 756                         goto freeit;
 757                 }
 758                 nd6_ra_input(n, off, icmp6len);
 759                 /* m stays. */
 760                 break;
 761
 762         case ND_NEIGHBOR_SOLICIT:
 763                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
 764                 if (code != 0)
 765                         goto badcode;
 766                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 767                         goto badlen;
 768                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 769                         /* give up local */
 770                         nd6_ns_input(m, off, icmp6len);
 771                         m = NULL;
 772                         goto freeit;
 773                 }
 774                 nd6_ns_input(n, off, icmp6len);
 775                 /* m stays. */
 776                 break;
 777
 778         case ND_NEIGHBOR_ADVERT:
 779                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
 780                 if (code != 0)
 781                         goto badcode;
 782                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 783                         goto badlen;
 784                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 785                         /* give up local */
 786                         nd6_na_input(m, off, icmp6len);
 787                         m = NULL;
 788                         goto freeit;
 789                 }
 790                 nd6_na_input(n, off, icmp6len);
 791                 /* m stays. */
 792                 break;
 793
 794         case ND_REDIRECT:
 795                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
 796                 if (code != 0)
 797                         goto badcode;
 798                 if (icmp6len < sizeof(struct nd_redirect))
 799                         goto badlen;
 800                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 801                         /* give up local */
 802                         icmp6_redirect_input(m, off);
 803                         m = NULL;
 804                         goto freeit;
 805                 }
 806                 icmp6_redirect_input(n, off);
 807                 /* m stays. */
 808                 break;
 809
 810         case ICMP6_ROUTER_RENUMBERING:
 811                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 812                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 813                         goto badcode;
 814                 if (icmp6len < sizeof(struct icmp6_router_renum))
 815                         goto badlen;
 816                 break;
 817
 818         default:
 819                 nd6log((LOG_DEBUG,
 820                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 821                     icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
 822                     ip6_sprintf(&ip6->ip6_dst),
 823                     m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
 824                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 825                         /* ICMPv6 error: MUST deliver it by spec... */
 826                         code = PRC_NCMDS;
 827                         /* deliver */
 828                 } else {
 829                         /* ICMPv6 informational: MUST not deliver */
 830                         break;
 831                 }
 832         deliver:
 833                 if (icmp6_notify_error(m, off, icmp6len, code)) {
 834                         /* In this case, m should've been freed. */
 835                         return (IPPROTO_DONE);
 836                 }
 837                 break;
 838
 839         badcode:
 840                 icmp6stat.icp6s_badcode++;
 841                 break;
 842
 843         badlen:
 844                 icmp6stat.icp6s_badlen++;
 845                 break;
 846         }
 847
 848         /* deliver the packet to appropriate sockets */
 849         icmp6_rip6_input(&m, *offp);
 850
 851         return IPPROTO_DONE;
 852
 853 freeit:
 854         m_freem(m);
 855         return IPPROTO_DONE;
 856 }
 857
 858 static int
 859 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
 860 {
 861         struct icmp6_hdr *icmp6;
 862         struct ip6_hdr *eip6;
 863         u_int32_t notifymtu;
 864         struct sockaddr_in6 icmp6src, icmp6dst;
 865
 866         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 867                 icmp6stat.icp6s_tooshort++;
 868                 goto freeit;
 869         }
 870 #ifndef PULLDOWN_TEST
 871         IP6_EXTHDR_CHECK(m, off,
 872                          sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
 873                          -1);
 874         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 875 #else
 876         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
 877                        sizeof(*icmp6) + sizeof(struct ip6_hdr));
 878         if (icmp6 == NULL) {
 879                 icmp6stat.icp6s_tooshort++;
 880                 return (-1);
 881         }
 882 #endif
 883         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 884
 885         /* Detect the upper level protocol */
 886         {
 887                 u_int8_t nxt = eip6->ip6_nxt;
 888                 int eoff = off + sizeof(struct icmp6_hdr) +
 889                         sizeof(struct ip6_hdr);
 890                 struct ip6ctlparam ip6cp;
 891                 struct in6_addr *finaldst = NULL;
 892                 int icmp6type = icmp6->icmp6_type;
 893                 struct ip6_frag *fh;
 894                 struct ip6_rthdr *rth;
 895                 int rthlen;
 896
 897                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 898                         struct ip6_ext *eh;
 899
 900                         switch (nxt) {
 901                         case IPPROTO_HOPOPTS:
 902                         case IPPROTO_DSTOPTS:
 903                         case IPPROTO_AH:
 904 #ifndef PULLDOWN_TEST
 905                                 IP6_EXTHDR_CHECK(m, 0, eoff +
 906                                                  sizeof(struct ip6_ext),
 907                                                  -1);
 908                                 eh = (struct ip6_ext *)(mtod(m, caddr_t)
 909                                                         + eoff);
 910 #else
 911                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
 912                                                eoff, sizeof(*eh));
 913                                 if (eh == NULL) {
 914                                         icmp6stat.icp6s_tooshort++;
 915                                         return (-1);
 916                                 }
 917 #endif
 918
 919                                 if (nxt == IPPROTO_AH)
 920                                         eoff += (eh->ip6e_len + 2) << 2;
 921                                 else
 922                                         eoff += (eh->ip6e_len + 1) << 3;
 923                                 nxt = eh->ip6e_nxt;
 924                                 break;
 925                         case IPPROTO_ROUTING:
 926                                 /*
 927                                  * When the erroneous packet contains a
 928                                  * routing header, we should examine the
 929                                  * header to determine the final destination.
 930                                  * Otherwise, we can't properly update
 931                                  * information that depends on the final
 932                                  * destination (e.g. path MTU).
 933                                  */
 934 #ifndef PULLDOWN_TEST
 935                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
 936                                                  -1);
 937                                 rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
 938                                                            + eoff);
 939 #else
 940                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
 941                                                eoff, sizeof(*rth));
 942                                 if (rth == NULL) {
 943                                         icmp6stat.icp6s_tooshort++;
 944                                         return (-1);
 945                                 }
 946 #endif
 947                                 rthlen = (rth->ip6r_len + 1) << 3;
 948                                 eoff += rthlen;
 949                                 nxt = rth->ip6r_nxt;
 950                                 break;
 951                         case IPPROTO_FRAGMENT:
 952 #ifndef PULLDOWN_TEST
 953                                 IP6_EXTHDR_CHECK(m, 0, eoff +
 954                                                  sizeof(struct ip6_frag),
 955                                                  -1);
 956                                 fh = (struct ip6_frag *)(mtod(m, caddr_t)
 957                                                          + eoff);
 958 #else
 959                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
 960                                                eoff, sizeof(*fh));
 961                                 if (fh == NULL) {
 962                                         icmp6stat.icp6s_tooshort++;
 963                                         return (-1);
 964                                 }
 965 #endif
 966                                 /*
 967                                  * Data after a fragment header is meaningless
 968                                  * unless it is the first fragment, but
 969                                  * we'll go to the notify label for path MTU
 970                                  * discovery.
 971                                  */
 972                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
 973                                         goto notify;
 974
 975                                 eoff += sizeof(struct ip6_frag);
 976                                 nxt = fh->ip6f_nxt;
 977                                 break;
 978                         default:
 979                                 /*
 980                                  * This case includes ESP and the No Next
 981                                  * Header.  In such cases going to the notify
 982                                  * label does not have any meaning
 983                                  * (i.e. pr_ctlinput will be NULL), but we go
 984                                  * anyway since we might have to update
 985                                  * path MTU information.
 986                                  */
 987                                 goto notify;
 988                         }
 989                 }
 990           notify:
 991 #ifndef PULLDOWN_TEST
 992                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 993 #else
 994                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
 995                                sizeof(*icmp6) + sizeof(struct ip6_hdr));
 996                 if (icmp6 == NULL) {
 997                         icmp6stat.icp6s_tooshort++;
 998                         return (-1);
 999                 }
1000 #endif
1001
1002                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1003                 bzero(&icmp6dst, sizeof(icmp6dst));
1004                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1005                 icmp6dst.sin6_family = AF_INET6;
1006                 if (finaldst == NULL)
1007                         icmp6dst.sin6_addr = eip6->ip6_dst;
1008                 else
1009                         icmp6dst.sin6_addr = *finaldst;
1010                 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1011                                                           &icmp6dst.sin6_addr);
1012                 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1013                                    NULL, NULL)) {
1014                         /* should be impossbile */
1015                         nd6log((LOG_DEBUG,
1016                             "icmp6_notify_error: in6_embedscope failed\n"));
1017                         goto freeit;
1018                 }
1019
1020                 /*
1021                  * retrieve parameters from the inner IPv6 header, and convert
1022                  * them into sockaddr structures.
1023                  */
1024                 bzero(&icmp6src, sizeof(icmp6src));
1025                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1026                 icmp6src.sin6_family = AF_INET6;
1027                 icmp6src.sin6_addr = eip6->ip6_src;
1028                 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1029                                                           &icmp6src.sin6_addr);
1030                 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1031                                    NULL, NULL)) {
1032                         /* should be impossbile */
1033                         nd6log((LOG_DEBUG,
1034                             "icmp6_notify_error: in6_embedscope failed\n"));
1035                         goto freeit;
1036                 }
1037                 icmp6src.sin6_flowinfo =
1038                         (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1039
1040                 if (finaldst == NULL)
1041                         finaldst = &eip6->ip6_dst;
1042                 ip6cp.ip6c_m = m;
1043                 ip6cp.ip6c_icmp6 = icmp6;
1044                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1045                 ip6cp.ip6c_off = eoff;
1046                 ip6cp.ip6c_finaldst = finaldst;
1047                 ip6cp.ip6c_src = &icmp6src;
1048                 ip6cp.ip6c_nxt = nxt;
1049
1050                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1051                         notifymtu = ntohl(icmp6->icmp6_mtu);
1052                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1053                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1054                 }
1055
1056                 so_pru_ctlinput(
1057                         (struct protosw *)&inet6sw[ip6_protox[nxt]],
1058                         code, (struct sockaddr *)&icmp6dst, &ip6cp);
1059         }
1060         return (0);
1061
1062 freeit:
1063         m_freem(m);
1064         return (-1);
1065 }
1066
1067 void
1068 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1069 {
1070         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1071         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1072         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1073         u_int mtu = ntohl(icmp6->icmp6_mtu);
1074         struct rtentry *rt = NULL;
1075         struct sockaddr_in6 sin6;
1076
1077         if (!validated)
1078                 return;
1079
1080         bzero(&sin6, sizeof(sin6));
1081         sin6.sin6_family = PF_INET6;
1082         sin6.sin6_len = sizeof(struct sockaddr_in6);
1083         sin6.sin6_addr = *dst;
1084         /* XXX normally, this won't happen */
1085         if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1086                 sin6.sin6_addr.s6_addr16[1] =
1087                     htons(m->m_pkthdr.rcvif->if_index);
1088         }
1089         /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1090         rt = rtpurelookup((struct sockaddr *)&sin6);
1091         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1092             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1093                 if (mtu < IPV6_MMTU) {                          /* XXX */
1094                         rt->rt_rmx.rmx_locks |= RTV_MTU;
1095                 } else if (mtu < rt->rt_ifp->if_mtu &&
1096                            rt->rt_rmx.rmx_mtu > mtu) {
1097                         icmp6stat.icp6s_pmtuchg++;
1098                         rt->rt_rmx.rmx_mtu = mtu;
1099                 }
1100         }
1101         if (rt)
1102                 --rt->rt_refcnt;
1103 }
1104
1105 /*
1106  * Process a Node Information Query packet, based on
1107  * draft-ietf-ipngwg-icmp-name-lookups-07.
1108  *
1109  * Spec incompatibilities:
1110  * - IPv6 Subject address handling
1111  * - IPv4 Subject address handling support missing
1112  * - Proxy reply (answer even if it's not for me)
1113  * - joins NI group address at in6_ifattach() time only, does not cope
1114  *   with hostname changes by sethostname(3)
1115  */
1116 #define hostnamelen     strlen(hostname)
1117 static struct mbuf *
1118 ni6_input(struct mbuf *m, int off)
1119 {
1120         struct icmp6_nodeinfo *ni6, *nni6;
1121         struct mbuf *n = NULL;
1122         u_int16_t qtype;
1123         int subjlen;
1124         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1125         struct ni_reply_fqdn *fqdn;
1126         int addrs;              /* for NI_QTYPE_NODEADDR */
1127         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1128         struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1129         struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1130         struct ip6_hdr *ip6;
1131         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1132         char *subj = NULL;
1133         struct in6_ifaddr *ia6 = NULL;
1134
1135         ip6 = mtod(m, struct ip6_hdr *);
1136 #ifndef PULLDOWN_TEST
1137         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1138 #else
1139         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1140         if (ni6 == NULL) {
1141                 /* m is already reclaimed */
1142                 return NULL;
1143         }
1144 #endif
1145
1146         /*
1147          * Validate IPv6 destination address.
1148          *
1149          * The Responder must discard the Query without further processing
1150          * unless it is one of the Responder's unicast or anycast addresses, or
1151          * a link-local scope multicast address which the Responder has joined.
1152          * [icmp-name-lookups-07, Section 4.]
1153          */
1154         bzero(&sin6, sizeof(sin6));
1155         sin6.sin6_family = AF_INET6;
1156         sin6.sin6_len = sizeof(struct sockaddr_in6);
1157         bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1158         /* XXX scopeid */
1159         if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1160                 /* unicast/anycast, fine */
1161                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1162                     (icmp6_nodeinfo & 4) == 0) {
1163                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1164                                 "a temporary address in %s:%d",
1165                                __FILE__, __LINE__));
1166                         goto bad;
1167                 }
1168         } else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1169                 ; /* link-local multicast, fine */
1170         else
1171                 goto bad;
1172
1173         /* validate query Subject field. */
1174         qtype = ntohs(ni6->ni_qtype);
1175         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1176         switch (qtype) {
1177         case NI_QTYPE_NOOP:
1178         case NI_QTYPE_SUPTYPES:
1179                 /* 07 draft */
1180                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1181                         break;
1182                 /* FALLTHROUGH */
1183         case NI_QTYPE_FQDN:
1184         case NI_QTYPE_NODEADDR:
1185                 switch (ni6->ni_code) {
1186                 case ICMP6_NI_SUBJ_IPV6:
1187 #if ICMP6_NI_SUBJ_IPV6 != 0
1188                 case 0:
1189 #endif
1190                         /*
1191                          * backward compatibility - try to accept 03 draft
1192                          * format, where no Subject is present.
1193                          */
1194                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1195                             subjlen == 0) {
1196                                 oldfqdn++;
1197                                 break;
1198                         }
1199 #if ICMP6_NI_SUBJ_IPV6 != 0
1200                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1201                                 goto bad;
1202 #endif
1203
1204                         if (subjlen != sizeof(sin6.sin6_addr))
1205                                 goto bad;
1206
1207                         /*
1208                          * Validate Subject address.
1209                          *
1210                          * Not sure what exactly "address belongs to the node"
1211                          * means in the spec, is it just unicast, or what?
1212                          *
1213                          * At this moment we consider Subject address as
1214                          * "belong to the node" if the Subject address equals
1215                          * to the IPv6 destination address; validation for
1216                          * IPv6 destination address should have done enough
1217                          * check for us.
1218                          *
1219                          * We do not do proxy at this moment.
1220                          */
1221                         /* m_pulldown instead of copy? */
1222                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1223                             subjlen, (caddr_t)&sin6.sin6_addr);
1224                         sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1225                                                               &sin6.sin6_addr);
1226                         in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1227                         bzero(&sin6_d, sizeof(sin6_d));
1228                         sin6_d.sin6_family = AF_INET6; /* not used, actually */
1229                         sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1230                         sin6_d.sin6_addr = ip6->ip6_dst;
1231                         sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1232                                                                 &ip6->ip6_dst);
1233                         in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1234                         subj = (char *)&sin6;
1235                         if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1236                                 break;
1237
1238                         /*
1239                          * XXX if we are to allow other cases, we should really
1240                          * be careful about scope here.
1241                          * basically, we should disallow queries toward IPv6
1242                          * destination X with subject Y, if scope(X) > scope(Y).
1243                          * if we allow scope(X) > scope(Y), it will result in
1244                          * information leakage across scope boundary.
1245                          */
1246                         goto bad;
1247
1248                 case ICMP6_NI_SUBJ_FQDN:
1249                         /*
1250                          * Validate Subject name with gethostname(3).
1251                          *
1252                          * The behavior may need some debate, since:
1253                          * - we are not sure if the node has FQDN as
1254                          *   hostname (returned by gethostname(3)).
1255                          * - the code does wildcard match for truncated names.
1256                          *   however, we are not sure if we want to perform
1257                          *   wildcard match, if gethostname(3) side has
1258                          *   truncated hostname.
1259                          */
1260                         n = ni6_nametodns(hostname, hostnamelen, 0);
1261                         if (!n || n->m_next || n->m_len == 0)
1262                                 goto bad;
1263                         IP6_EXTHDR_GET(subj, char *, m,
1264                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1265                         if (subj == NULL)
1266                                 goto bad;
1267                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1268                                         n->m_len)) {
1269                                 goto bad;
1270                         }
1271                         m_freem(n);
1272                         n = NULL;
1273                         break;
1274
1275                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1276                 default:
1277                         goto bad;
1278                 }
1279                 break;
1280         }
1281
1282         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1283         switch (qtype) {
1284         case NI_QTYPE_FQDN:
1285                 if ((icmp6_nodeinfo & 1) == 0)
1286                         goto bad;
1287                 break;
1288         case NI_QTYPE_NODEADDR:
1289                 if ((icmp6_nodeinfo & 2) == 0)
1290                         goto bad;
1291                 break;
1292         }
1293
1294         /* guess reply length */
1295         switch (qtype) {
1296         case NI_QTYPE_NOOP:
1297                 break;          /* no reply data */
1298         case NI_QTYPE_SUPTYPES:
1299                 replylen += sizeof(u_int32_t);
1300                 break;
1301         case NI_QTYPE_FQDN:
1302                 /* XXX will append an mbuf */
1303                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1304                 break;
1305         case NI_QTYPE_NODEADDR:
1306                 addrs = ni6_addrs(ni6, m, &ifp, subj);
1307                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1308                                           sizeof(u_int32_t))) > MCLBYTES)
1309                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1310                 break;
1311         default:
1312                 /*
1313                  * XXX: We must return a reply with the ICMP6 code
1314                  * `unknown Qtype' in this case. However we regard the case
1315                  * as an FQDN query for backward compatibility.
1316                  * Older versions set a random value to this field,
1317                  * so it rarely varies in the defined qtypes.
1318                  * But the mechanism is not reliable...
1319                  * maybe we should obsolete older versions.
1320                  */
1321                 qtype = NI_QTYPE_FQDN;
1322                 /* XXX will append an mbuf */
1323                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1324                 oldfqdn++;
1325                 break;
1326         }
1327
1328         /* allocate an mbuf to reply. */
1329         if (replylen > MCLBYTES) {
1330                 /*
1331                  * XXX: should we try to allocate more? But MCLBYTES
1332                  * is probably much larger than IPV6_MMTU...
1333                  */
1334                 goto bad;
1335         }
1336         n = m_getb(replylen, MB_DONTWAIT, m->m_type, M_PKTHDR);
1337         if (n == NULL) {
1338                 m_freem(m);
1339                 return (NULL);
1340         }
1341         M_MOVE_PKTHDR(n, m); /* just for recvif */
1342         n->m_pkthdr.len = n->m_len = replylen;
1343
1344         /* copy mbuf header and IPv6 + Node Information base headers */
1345         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1346         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1347         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1348
1349         /* qtype dependent procedure */
1350         switch (qtype) {
1351         case NI_QTYPE_NOOP:
1352                 nni6->ni_code = ICMP6_NI_SUCCESS;
1353                 nni6->ni_flags = 0;
1354                 break;
1355         case NI_QTYPE_SUPTYPES:
1356         {
1357                 u_int32_t v;
1358                 nni6->ni_code = ICMP6_NI_SUCCESS;
1359                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1360                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1361                 v = (u_int32_t)htonl(0x0000000f);
1362                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1363                 break;
1364         }
1365         case NI_QTYPE_FQDN:
1366                 nni6->ni_code = ICMP6_NI_SUCCESS;
1367                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1368                                                 sizeof(struct ip6_hdr) +
1369                                                 sizeof(struct icmp6_nodeinfo));
1370                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1371                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1372                 /*
1373                  * XXX do we really have FQDN in variable "hostname"?
1374                  */
1375                 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1376                 if (n->m_next == NULL)
1377                         goto bad;
1378                 /* XXX we assume that n->m_next is not a chain */
1379                 if (n->m_next->m_next != NULL)
1380                         goto bad;
1381                 n->m_pkthdr.len += n->m_next->m_len;
1382                 break;
1383         case NI_QTYPE_NODEADDR:
1384         {
1385                 int lenlim, copied;
1386
1387                 nni6->ni_code = ICMP6_NI_SUCCESS;
1388                 n->m_pkthdr.len = n->m_len =
1389                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1390                 lenlim = M_TRAILINGSPACE(n);
1391                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1392                 /* XXX: reset mbuf length */
1393                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1394                         sizeof(struct icmp6_nodeinfo) + copied;
1395                 break;
1396         }
1397         default:
1398                 break;          /* XXX impossible! */
1399         }
1400
1401         nni6->ni_type = ICMP6_NI_REPLY;
1402         m_freem(m);
1403         return (n);
1404
1405 bad:
1406         m_freem(m);
1407         if (n)
1408                 m_freem(n);
1409         return (NULL);
1410 }
1411 #undef hostnamelen
1412
1413 /*
1414  * make a mbuf with DNS-encoded string.  no compression support.
1415  *
1416  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1417  * treated as truncated name (two \0 at the end).  this is a wild guess.
1418  */
1419 static struct mbuf *
1420 ni6_nametodns(const char *name, int namelen,
1421               int old) /* return pascal string if non-zero */
1422 {
1423         struct mbuf *m;
1424         char *cp, *ep;
1425         const char *p, *q;
1426         int i, len, nterm;
1427
1428         if (old)
1429                 len = namelen + 1;
1430         else
1431                 len = MCLBYTES;
1432
1433         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1434         m = m_getb(len, MB_DONTWAIT, MT_DATA, 0);
1435         if (!m)
1436                 goto fail;
1437
1438         if (old) {
1439                 m->m_len = len;
1440                 *mtod(m, char *) = namelen;
1441                 bcopy(name, mtod(m, char *) + 1, namelen);
1442                 return m;
1443         } else {
1444                 m->m_len = 0;
1445                 cp = mtod(m, char *);
1446                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1447
1448                 /* if not certain about my name, return empty buffer */
1449                 if (namelen == 0)
1450                         return m;
1451
1452                 /*
1453                  * guess if it looks like shortened hostname, or FQDN.
1454                  * shortened hostname needs two trailing "\0".
1455                  */
1456                 i = 0;
1457                 for (p = name; p < name + namelen; p++) {
1458                         if (*p && *p == '.')
1459                                 i++;
1460                 }
1461                 if (i < 2)
1462                         nterm = 2;
1463                 else
1464                         nterm = 1;
1465
1466                 p = name;
1467                 while (cp < ep && p < name + namelen) {
1468                         i = 0;
1469                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1470                                 i++;
1471                         /* result does not fit into mbuf */
1472                         if (cp + i + 1 >= ep)
1473                                 goto fail;
1474                         /*
1475                          * DNS label length restriction, RFC1035 page 8.
1476                          * "i == 0" case is included here to avoid returning
1477                          * 0-length label on "foo..bar".
1478                          */
1479                         if (i <= 0 || i >= 64)
1480                                 goto fail;
1481                         *cp++ = i;
1482                         bcopy(p, cp, i);
1483                         cp += i;
1484                         p = q;
1485                         if (p < name + namelen && *p == '.')
1486                                 p++;
1487                 }
1488                 /* termination */
1489                 if (cp + nterm >= ep)
1490                         goto fail;
1491                 while (nterm-- > 0)
1492                         *cp++ = '\0';
1493                 m->m_len = cp - mtod(m, char *);
1494                 return m;
1495         }
1496
1497         panic("should not reach here");
1498         /* NOTREACHED */
1499
1500 fail:
1501         if (m)
1502                 m_freem(m);
1503         return NULL;
1504 }
1505
1506 /*
1507  * check if two DNS-encoded string matches.  takes care of truncated
1508  * form (with \0\0 at the end).  no compression support.
1509  * XXX upper/lowercase match (see RFC2065)
1510  */
1511 static int
1512 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1513 {
1514         const char *a0, *b0;
1515         int l;
1516
1517         /* simplest case - need validation? */
1518         if (alen == blen && bcmp(a, b, alen) == 0)
1519                 return 1;
1520
1521         a0 = a;
1522         b0 = b;
1523
1524         /* termination is mandatory */
1525         if (alen < 2 || blen < 2)
1526                 return 0;
1527         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1528                 return 0;
1529         alen--;
1530         blen--;
1531
1532         while (a - a0 < alen && b - b0 < blen) {
1533                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1534                         return 0;
1535
1536                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1537                         return 0;
1538                 /* we don't support compression yet */
1539                 if (a[0] >= 64 || b[0] >= 64)
1540                         return 0;
1541
1542                 /* truncated case */
1543                 if (a[0] == 0 && a - a0 == alen - 1)
1544                         return 1;
1545                 if (b[0] == 0 && b - b0 == blen - 1)
1546                         return 1;
1547                 if (a[0] == 0 || b[0] == 0)
1548                         return 0;
1549
1550                 if (a[0] != b[0])
1551                         return 0;
1552                 l = a[0];
1553                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1554                         return 0;
1555                 if (bcmp(a + 1, b + 1, l) != 0)
1556                         return 0;
1557
1558                 a += 1 + l;
1559                 b += 1 + l;
1560         }
1561
1562         if (a - a0 == alen && b - b0 == blen)
1563                 return 1;
1564         else
1565                 return 0;
1566 }
1567
1568 /*
1569  * calculate the number of addresses to be returned in the node info reply.
1570  */
1571 static int
1572 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1573           char *subj)
1574 {
1575         struct ifnet *ifp;
1576         struct in6_ifaddr *ifa6;
1577         struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1578         int addrs = 0, addrsofif, iffound = 0;
1579         int niflags = ni6->ni_flags;
1580
1581         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1582                 switch (ni6->ni_code) {
1583                 case ICMP6_NI_SUBJ_IPV6:
1584                         if (subj == NULL) /* must be impossible... */
1585                                 return (0);
1586                         subj_ip6 = (struct sockaddr_in6 *)subj;
1587                         break;
1588                 default:
1589                         /*
1590                          * XXX: we only support IPv6 subject address for
1591                          * this Qtype.
1592                          */
1593                         return (0);
1594                 }
1595         }
1596
1597         for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1598         {
1599                 struct ifaddr_container *ifac;
1600
1601                 addrsofif = 0;
1602                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link)
1603                 {
1604                         struct ifaddr *ifa = ifac->ifa;
1605
1606                         if (ifa->ifa_addr->sa_family != AF_INET6)
1607                                 continue;
1608                         ifa6 = (struct in6_ifaddr *)ifa;
1609
1610                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1611                             IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1612                                                &ifa6->ia_addr.sin6_addr))
1613                                 iffound = 1;
1614
1615                         /*
1616                          * IPv4-mapped addresses can only be returned by a
1617                          * Node Information proxy, since they represent
1618                          * addresses of IPv4-only nodes, which perforce do
1619                          * not implement this protocol.
1620                          * [icmp-name-lookups-07, Section 5.4]
1621                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1622                          * this function at this moment.
1623                          */
1624
1625                         /* What do we have to do about ::1? */
1626                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1627                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1628                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1629                                         continue;
1630                                 break;
1631                         case IPV6_ADDR_SCOPE_SITELOCAL:
1632                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1633                                         continue;
1634                                 break;
1635                         case IPV6_ADDR_SCOPE_GLOBAL:
1636                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1637                                         continue;
1638                                 break;
1639                         default:
1640                                 continue;
1641                         }
1642
1643                         /*
1644                          * check if anycast is okay.
1645                          * XXX: just experimental.  not in the spec.
1646                          */
1647                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1648                             !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1649                                 continue; /* we need only unicast addresses */
1650                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1651                             (icmp6_nodeinfo & 4) == 0) {
1652                                 continue;
1653                         }
1654                         addrsofif++; /* count the address */
1655                 }
1656                 if (iffound) {
1657                         *ifpp = ifp;
1658                         return (addrsofif);
1659                 }
1660
1661                 addrs += addrsofif;
1662         }
1663
1664         return (addrs);
1665 }
1666
1667 static int
1668 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1669                 struct ifnet *ifp0, int resid)
1670 {
1671         struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1672         struct in6_ifaddr *ifa6;
1673         struct ifnet *ifp_dep = NULL;
1674         int copied = 0, allow_deprecated = 0;
1675         u_char *cp = (u_char *)(nni6 + 1);
1676         int niflags = ni6->ni_flags;
1677         u_int32_t ltime;
1678
1679         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1680                 return (0);     /* needless to copy */
1681
1682 again:
1683
1684         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1685         {
1686                 struct ifaddr_container *ifac;
1687
1688                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1689                         struct ifaddr *ifa = ifac->ifa;
1690
1691                         if (ifa->ifa_addr->sa_family != AF_INET6)
1692                                 continue;
1693                         ifa6 = (struct in6_ifaddr *)ifa;
1694
1695                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1696                             allow_deprecated == 0) {
1697                                 /*
1698                                  * prefererred address should be put before
1699                                  * deprecated addresses.
1700                                  */
1701
1702                                 /* record the interface for later search */
1703                                 if (ifp_dep == NULL)
1704                                         ifp_dep = ifp;
1705
1706                                 continue;
1707                         }
1708                         else if (!(ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1709                                  allow_deprecated != 0)
1710                                 continue; /* we now collect deprecated addrs */
1711
1712                         /* What do we have to do about ::1? */
1713                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1714                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1715                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1716                                         continue;
1717                                 break;
1718                         case IPV6_ADDR_SCOPE_SITELOCAL:
1719                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1720                                         continue;
1721                                 break;
1722                         case IPV6_ADDR_SCOPE_GLOBAL:
1723                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1724                                         continue;
1725                                 break;
1726                         default:
1727                                 continue;
1728                         }
1729
1730                         /*
1731                          * check if anycast is okay.
1732                          * XXX: just experimental. not in the spec.
1733                          */
1734                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1735                             !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1736                                 continue;
1737                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1738                             (icmp6_nodeinfo & 4) == 0) {
1739                                 continue;
1740                         }
1741
1742                         /* now we can copy the address */
1743                         if (resid < sizeof(struct in6_addr) +
1744                             sizeof(u_int32_t)) {
1745                                 /*
1746                                  * We give up much more copy.
1747                                  * Set the truncate flag and return.
1748                                  */
1749                                 nni6->ni_flags |=
1750                                         NI_NODEADDR_FLAG_TRUNCATE;
1751                                 return (copied);
1752                         }
1753
1754                         /*
1755                          * Set the TTL of the address.
1756                          * The TTL value should be one of the following
1757                          * according to the specification:
1758                          *
1759                          * 1. The remaining lifetime of a DHCP lease on the
1760                          *    address, or
1761                          * 2. The remaining Valid Lifetime of a prefix from
1762                          *    which the address was derived through Stateless
1763                          *    Autoconfiguration.
1764                          *
1765                          * Note that we currently do not support stateful
1766                          * address configuration by DHCPv6, so the former
1767                          * case can't happen.
1768                          */
1769                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1770                                 ltime = ND6_INFINITE_LIFETIME;
1771                         else {
1772                                 if (ifa6->ia6_lifetime.ia6t_expire >
1773                                     time_second)
1774                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1775                                 else
1776                                         ltime = 0;
1777                         }
1778
1779                         bcopy(&ltime, cp, sizeof(u_int32_t));
1780                         cp += sizeof(u_int32_t);
1781
1782                         /* copy the address itself */
1783                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1784                               sizeof(struct in6_addr));
1785                         /* XXX: KAME link-local hack; remove ifindex */
1786                         if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1787                                 ((struct in6_addr *)cp)->s6_addr16[1] = 0;
1788                         cp += sizeof(struct in6_addr);
1789
1790                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1791                         copied += (sizeof(struct in6_addr) +
1792                                    sizeof(u_int32_t));
1793                 }
1794                 if (ifp0)       /* we need search only on the specified IF */
1795                         break;
1796         }
1797
1798         if (allow_deprecated == 0 && ifp_dep != NULL) {
1799                 ifp = ifp_dep;
1800                 allow_deprecated = 1;
1801
1802                 goto again;
1803         }
1804
1805         return (copied);
1806 }
1807
1808 /*
1809  * XXX almost dup'ed code with rip6_input.
1810  */
1811 static int
1812 icmp6_rip6_input(struct mbuf **mp, int  off)
1813 {
1814         struct mbuf *m = *mp;
1815         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1816         struct in6pcb *in6p;
1817         struct in6pcb *last = NULL;
1818         struct sockaddr_in6 rip6src;
1819         struct icmp6_hdr *icmp6;
1820         struct mbuf *opts = NULL;
1821
1822 #ifndef PULLDOWN_TEST
1823         /* this is assumed to be safe. */
1824         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1825 #else
1826         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1827         if (icmp6 == NULL) {
1828                 /* m is already reclaimed */
1829                 return IPPROTO_DONE;
1830         }
1831 #endif
1832
1833         bzero(&rip6src, sizeof(rip6src));
1834         rip6src.sin6_len = sizeof(struct sockaddr_in6);
1835         rip6src.sin6_family = AF_INET6;
1836         /* KAME hack: recover scopeid */
1837         in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1838
1839         LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1840         {
1841                 if (in6p->inp_flags & INP_PLACEMARKER)
1842                         continue;
1843                 if ((in6p->inp_vflag & INP_IPV6) == 0)
1844                         continue;
1845 #ifdef HAVE_NRL_INPCB
1846                 if (!(in6p->in6p_flags & INP_IPV6))
1847                         continue;
1848 #endif
1849                 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1850                         continue;
1851                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1852                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1853                         continue;
1854                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1855                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1856                         continue;
1857                 if (in6p->in6p_icmp6filt
1858                     && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1859                                  in6p->in6p_icmp6filt))
1860                         continue;
1861                 if (last) {
1862                         struct  mbuf *n;
1863                         if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1864                                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1865                                         ip6_savecontrol(last, &opts, ip6, n);
1866                                 /* strip intermediate headers */
1867                                 m_adj(n, off);
1868                                 if (ssb_appendaddr(&last->in6p_socket->so_rcv,
1869                                                  (struct sockaddr *)&rip6src,
1870                                                  n, opts) == 0) {
1871                                         /* should notify about lost packet */
1872                                         m_freem(n);
1873                                         if (opts) {
1874                                                 m_freem(opts);
1875                                         }
1876                                 } else
1877                                         sorwakeup(last->in6p_socket);
1878                                 opts = NULL;
1879                         }
1880                 }
1881                 last = in6p;
1882         }
1883         if (last) {
1884                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1885                         ip6_savecontrol(last, &opts, ip6, m);
1886                 /* strip intermediate headers */
1887                 m_adj(m, off);
1888                 if (ssb_appendaddr(&last->in6p_socket->so_rcv,
1889                                  (struct sockaddr *)&rip6src, m, opts) == 0) {
1890                         m_freem(m);
1891                         if (opts)
1892                                 m_freem(opts);
1893                 } else
1894                         sorwakeup(last->in6p_socket);
1895         } else {
1896                 m_freem(m);
1897                 ip6stat.ip6s_delivered--;
1898         }
1899         return IPPROTO_DONE;
1900 }
1901
1902 /*
1903  * Reflect the ip6 packet back to the source.
1904  * OFF points to the icmp6 header, counted from the top of the mbuf.
1905  */
1906 void
1907 icmp6_reflect(struct mbuf *m, size_t off)
1908 {
1909         struct ip6_hdr *ip6;
1910         struct icmp6_hdr *icmp6;
1911         struct in6_ifaddr *ia;
1912         struct in6_addr t, *src = 0;
1913         int plen;
1914         int type, code;
1915         struct ifnet *outif = NULL;
1916         struct sockaddr_in6 sa6_src, sa6_dst;
1917 #ifdef COMPAT_RFC1885
1918         int mtu = IPV6_MMTU;
1919         struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1920 #endif
1921
1922         /* too short to reflect */
1923         if (off < sizeof(struct ip6_hdr)) {
1924                 nd6log((LOG_DEBUG,
1925                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1926                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
1927                     __FILE__, __LINE__));
1928                 goto bad;
1929         }
1930
1931         /*
1932          * If there are extra headers between IPv6 and ICMPv6, strip
1933          * off that header first.
1934          */
1935 #ifdef DIAGNOSTIC
1936         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1937                 panic("assumption failed in icmp6_reflect");
1938 #endif
1939         if (off > sizeof(struct ip6_hdr)) {
1940                 size_t l;
1941                 struct ip6_hdr nip6;
1942
1943                 l = off - sizeof(struct ip6_hdr);
1944                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1945                 m_adj(m, l);
1946                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1947                 if (m->m_len < l) {
1948                         if ((m = m_pullup(m, l)) == NULL)
1949                                 return;
1950                 }
1951                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1952         } else /* off == sizeof(struct ip6_hdr) */ {
1953                 size_t l;
1954                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1955                 if (m->m_len < l) {
1956                         if ((m = m_pullup(m, l)) == NULL)
1957                                 return;
1958                 }
1959         }
1960         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1961         ip6 = mtod(m, struct ip6_hdr *);
1962         ip6->ip6_nxt = IPPROTO_ICMPV6;
1963         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1964         type = icmp6->icmp6_type; /* keep type for statistics */
1965         code = icmp6->icmp6_code; /* ditto. */
1966
1967         t = ip6->ip6_dst;
1968         /*
1969          * ip6_input() drops a packet if its src is multicast.
1970          * So, the src is never multicast.
1971          */
1972         ip6->ip6_dst = ip6->ip6_src;
1973
1974         /*
1975          * XXX: make sure to embed scope zone information, using
1976          * already embedded IDs or the received interface (if any).
1977          * Note that rcvif may be NULL.
1978          * TODO: scoped routing case (XXX).
1979          */
1980         bzero(&sa6_src, sizeof(sa6_src));
1981         sa6_src.sin6_family = AF_INET6;
1982         sa6_src.sin6_len = sizeof(sa6_src);
1983         sa6_src.sin6_addr = ip6->ip6_dst;
1984         in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
1985         in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
1986         bzero(&sa6_dst, sizeof(sa6_dst));
1987         sa6_dst.sin6_family = AF_INET6;
1988         sa6_dst.sin6_len = sizeof(sa6_dst);
1989         sa6_dst.sin6_addr = t;
1990         in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
1991         in6_embedscope(&t, &sa6_dst, NULL, NULL);
1992
1993 #ifdef COMPAT_RFC1885
1994         /*
1995          * xxx guess MTU
1996          * RFC 1885 requires that echo reply should be truncated if it
1997          * does not fit in with (return) path MTU, but the description was
1998          * removed in the new spec.
1999          */
2000         if (icmp6_reflect_rt.ro_rt == 0 ||
2001             ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2002                 if (icmp6_reflect_rt.ro_rt) {
2003                         RTFREE(icmp6_reflect_rt.ro_rt);
2004                         icmp6_reflect_rt.ro_rt = 0;
2005                 }
2006                 bzero(sin6, sizeof(*sin6));
2007                 sin6->sin6_family = PF_INET6;
2008                 sin6->sin6_len = sizeof(struct sockaddr_in6);
2009                 sin6->sin6_addr = ip6->ip6_dst;
2010
2011                 rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2012                             RTF_PRCLONING);
2013         }
2014
2015         if (icmp6_reflect_rt.ro_rt == 0)
2016                 goto bad;
2017
2018         if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2019             && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2020                 mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2021
2022         if (mtu < m->m_pkthdr.len) {
2023                 plen -= (m->m_pkthdr.len - mtu);
2024                 m_adj(m, mtu - m->m_pkthdr.len);
2025         }
2026 #endif
2027         /*
2028          * If the incoming packet was addressed directly to us(i.e. unicast),
2029          * use dst as the src for the reply.
2030          * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2031          * (for example) when we encounter an error while forwarding procedure
2032          * destined to a duplicated address of ours.
2033          */
2034         for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2035                 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2036                     (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2037                         src = &t;
2038                         break;
2039                 }
2040         if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2041                 /*
2042                  * This is the case if the dst is our link-local address
2043                  * and the sender is also ourselves.
2044                  */
2045                 src = &t;
2046         }
2047
2048         if (src == 0) {
2049                 int e;
2050                 struct route_in6 ro;
2051
2052                 /*
2053                  * This case matches to multicasts, our anycast, or unicasts
2054                  * that we do not own.  Select a source address based on the
2055                  * source address of the erroneous packet.
2056                  */
2057                 bzero(&ro, sizeof(ro));
2058                 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e, NULL);
2059                 if (ro.ro_rt)
2060                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2061                 if (src == NULL) {
2062                         nd6log((LOG_DEBUG,
2063                             "icmp6_reflect: source can't be determined: "
2064                             "dst=%s, error=%d\n",
2065                             ip6_sprintf(&sa6_src.sin6_addr), e));
2066                         goto bad;
2067                 }
2068         }
2069
2070         ip6->ip6_src = *src;
2071
2072         ip6->ip6_flow = 0;
2073         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2074         ip6->ip6_vfc |= IPV6_VERSION;
2075         ip6->ip6_nxt = IPPROTO_ICMPV6;
2076         if (m->m_pkthdr.rcvif) {
2077                 /* XXX: This may not be the outgoing interface */
2078                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2079         } else
2080                 ip6->ip6_hlim = ip6_defhlim;
2081
2082         icmp6->icmp6_cksum = 0;
2083         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2084                                         sizeof(struct ip6_hdr), plen);
2085
2086         /*
2087          * XXX option handling
2088          */
2089
2090         m->m_flags &= ~(M_BCAST|M_MCAST);
2091
2092 #ifdef COMPAT_RFC1885
2093         ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2094 #else
2095         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2096 #endif
2097         if (outif)
2098                 icmp6_ifoutstat_inc(outif, type, code);
2099
2100         return;
2101
2102 bad:
2103         m_freem(m);
2104         return;
2105 }
2106
2107 void
2108 icmp6_fasttimo(void)
2109 {
2110
2111         mld6_fasttimeo();
2112 }
2113
2114 static const char *
2115 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2116                     struct in6_addr *tgt6)
2117 {
2118         static char buf[1024];
2119         ksnprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2120                 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2121         return buf;
2122 }
2123
2124 void
2125 icmp6_redirect_input(struct mbuf *m, int off)
2126 {
2127         struct ifnet *ifp = m->m_pkthdr.rcvif;
2128         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2129         struct nd_redirect *nd_rd;
2130         int icmp6len = ntohs(ip6->ip6_plen);
2131         char *lladdr = NULL;
2132         int lladdrlen = 0;
2133         u_char *redirhdr = NULL;
2134         int redirhdrlen = 0;
2135         struct rtentry *rt = NULL;
2136         int is_router;
2137         int is_onlink;
2138         struct in6_addr src6 = ip6->ip6_src;
2139         struct in6_addr redtgt6;
2140         struct in6_addr reddst6;
2141         union nd_opts ndopts;
2142
2143         if (!m || !ifp)
2144                 return;
2145
2146         /* XXX if we are router, we don't update route by icmp6 redirect */
2147         if (ip6_forwarding)
2148                 goto freeit;
2149         if (!icmp6_rediraccept)
2150                 goto freeit;
2151
2152 #ifndef PULLDOWN_TEST
2153         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2154         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2155 #else
2156         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2157         if (nd_rd == NULL) {
2158                 icmp6stat.icp6s_tooshort++;
2159                 return;
2160         }
2161 #endif
2162         redtgt6 = nd_rd->nd_rd_target;
2163         reddst6 = nd_rd->nd_rd_dst;
2164
2165         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2166                 redtgt6.s6_addr16[1] = htons(ifp->if_index);
2167         if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2168                 reddst6.s6_addr16[1] = htons(ifp->if_index);
2169
2170         /* validation */
2171         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2172                 nd6log((LOG_ERR,
2173                         "ICMP6 redirect sent from %s rejected; "
2174                         "must be from linklocal\n", ip6_sprintf(&src6)));
2175                 goto bad;
2176         }
2177         if (ip6->ip6_hlim != 255) {
2178                 nd6log((LOG_ERR,
2179                         "ICMP6 redirect sent from %s rejected; "
2180                         "hlim=%d (must be 255)\n",
2181                         ip6_sprintf(&src6), ip6->ip6_hlim));
2182                 goto bad;
2183         }
2184     {
2185         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2186         struct sockaddr_in6 sin6;
2187         struct in6_addr *gw6;
2188
2189         bzero(&sin6, sizeof(sin6));
2190         sin6.sin6_family = AF_INET6;
2191         sin6.sin6_len = sizeof(struct sockaddr_in6);
2192         bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2193         rt = rtpurelookup((struct sockaddr *)&sin6);
2194         if (rt != NULL) {
2195                 if (rt->rt_gateway == NULL ||
2196                     rt->rt_gateway->sa_family != AF_INET6) {
2197                         nd6log((LOG_ERR,
2198                             "ICMP6 redirect rejected; no route "
2199                             "with inet6 gateway found for redirect dst: %s\n",
2200                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2201                         --rt->rt_refcnt;
2202                         goto bad;
2203                 }
2204
2205                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2206                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2207                         nd6log((LOG_ERR,
2208                                 "ICMP6 redirect rejected; "
2209                                 "not equal to gw-for-src=%s (must be same): "
2210                                 "%s\n",
2211                                 ip6_sprintf(gw6),
2212                                 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2213                         --rt->rt_refcnt;
2214                         goto bad;
2215                 }
2216         } else {
2217                 nd6log((LOG_ERR,
2218                         "ICMP6 redirect rejected; "
2219                         "no route found for redirect dst: %s\n",
2220                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2221                 goto bad;
2222         }
2223         --rt->rt_refcnt;
2224         rt = NULL;
2225     }
2226         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2227                 nd6log((LOG_ERR,
2228                         "ICMP6 redirect rejected; "
2229                         "redirect dst must be unicast: %s\n",
2230                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2231                 goto bad;
2232         }
2233
2234         is_router = is_onlink = 0;
2235         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2236                 is_router = 1;  /* router case */
2237         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2238                 is_onlink = 1;  /* on-link destination case */
2239         if (!is_router && !is_onlink) {
2240                 nd6log((LOG_ERR,
2241                         "ICMP6 redirect rejected; "
2242                         "neither router case nor onlink case: %s\n",
2243                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2244                 goto bad;
2245         }
2246         /* validation passed */
2247
2248         icmp6len -= sizeof(*nd_rd);
2249         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2250         if (nd6_options(&ndopts) < 0) {
2251                 nd6log((LOG_INFO, "icmp6_redirect_input: "
2252                         "invalid ND option, rejected: %s\n",
2253                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2254                 /* nd6_options have incremented stats */
2255                 goto freeit;
2256         }
2257
2258         if (ndopts.nd_opts_tgt_lladdr) {
2259                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2260                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2261         }
2262
2263         if (ndopts.nd_opts_rh) {
2264                 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2265                 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2266         }
2267
2268         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2269                 nd6log((LOG_INFO,
2270                         "icmp6_redirect_input: lladdrlen mismatch for %s "
2271                         "(if %d, icmp6 packet %d): %s\n",
2272                         ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2273                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2274                 goto bad;
2275         }
2276
2277         /* RFC 2461 8.3 */
2278         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2279                          is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2280
2281         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2282                 /* perform rtredirect */
2283                 struct sockaddr_in6 sdst;
2284                 struct sockaddr_in6 sgw;
2285                 struct sockaddr_in6 ssrc;
2286
2287                 bzero(&sdst, sizeof(sdst));
2288                 bzero(&sgw, sizeof(sgw));
2289                 bzero(&ssrc, sizeof(ssrc));
2290                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2291                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2292                         sizeof(struct sockaddr_in6);
2293                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2294                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2295                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2296                 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2297                            NULL, RTF_GATEWAY | RTF_HOST,
2298                            (struct sockaddr *)&ssrc);
2299         }
2300         /* finally update cached route in each socket via kpfctlinput */
2301     {
2302         struct sockaddr_in6 sdst;
2303
2304         bzero(&sdst, sizeof(sdst));
2305         sdst.sin6_family = AF_INET6;
2306         sdst.sin6_len = sizeof(struct sockaddr_in6);
2307         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2308         kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2309 #ifdef IPSEC
2310         key_sa_routechange((struct sockaddr *)&sdst);
2311 #endif
2312     }
2313
2314 freeit:
2315         m_freem(m);
2316         return;
2317
2318 bad:
2319         icmp6stat.icp6s_badredirect++;
2320         m_freem(m);
2321 }
2322
2323 void
2324 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2325 {
2326         struct ifnet *ifp;      /* my outgoing interface */
2327         struct in6_addr *ifp_ll6;
2328         struct in6_addr *router_ll6;
2329         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2330         struct mbuf *m = NULL;  /* newly allocated one */
2331         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2332         struct nd_redirect *nd_rd;
2333         size_t maxlen;
2334         u_char *p;
2335         struct ifnet *outif = NULL;
2336         struct sockaddr_in6 src_sa;
2337
2338         icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2339
2340         /* if we are not router, we don't send icmp6 redirect */
2341         if (!ip6_forwarding || ip6_accept_rtadv)
2342                 goto fail;
2343
2344         /* sanity check */
2345         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2346                 goto fail;
2347
2348         /*
2349          * Address check:
2350          *  the source address must identify a neighbor, and
2351          *  the destination address must not be a multicast address
2352          *  [RFC 2461, sec 8.2]
2353          */
2354         sip6 = mtod(m0, struct ip6_hdr *);
2355         bzero(&src_sa, sizeof(src_sa));
2356         src_sa.sin6_family = AF_INET6;
2357         src_sa.sin6_len = sizeof(src_sa);
2358         src_sa.sin6_addr = sip6->ip6_src;
2359         /* we don't currently use sin6_scope_id, but eventually use it */
2360         src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2361         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2362                 goto fail;
2363         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2364                 goto fail;      /* what should we do here? */
2365
2366         /* rate limit */
2367         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2368                 goto fail;
2369
2370         /*
2371          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2372          * we almost always ask for an mbuf cluster for simplicity.
2373          * (MHLEN < IPV6_MMTU is almost always true)
2374          */
2375 #if IPV6_MMTU >= MCLBYTES
2376 # error assumption failed about IPV6_MMTU and MCLBYTES
2377 #endif
2378         m = m_getb(IPV6_MMTU, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
2379         if (!m)
2380                 goto fail;
2381         m->m_len = 0;
2382         maxlen = M_TRAILINGSPACE(m);
2383         maxlen = min(IPV6_MMTU, maxlen);
2384         /* just for safety */
2385         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2386             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2387                 goto fail;
2388         }
2389
2390         {
2391                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2392                 struct in6_ifaddr *ia;
2393                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2394                                                  IN6_IFF_NOTREADY|
2395                                                  IN6_IFF_ANYCAST)) == NULL)
2396                         goto fail;
2397                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2398         }
2399
2400         /* get ip6 linklocal address for the router. */
2401         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2402                 struct sockaddr_in6 *sin6;
2403                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2404                 router_ll6 = &sin6->sin6_addr;
2405                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2406                         router_ll6 = NULL;
2407         } else
2408                 router_ll6 = NULL;
2409
2410         /* ip6 */
2411         ip6 = mtod(m, struct ip6_hdr *);
2412         ip6->ip6_flow = 0;
2413         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2414         ip6->ip6_vfc |= IPV6_VERSION;
2415         /* ip6->ip6_plen will be set later */
2416         ip6->ip6_nxt = IPPROTO_ICMPV6;
2417         ip6->ip6_hlim = 255;
2418         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2419         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2420         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2421
2422         /* ND Redirect */
2423         nd_rd = (struct nd_redirect *)(ip6 + 1);
2424         nd_rd->nd_rd_type = ND_REDIRECT;
2425         nd_rd->nd_rd_code = 0;
2426         nd_rd->nd_rd_reserved = 0;
2427         if (rt->rt_flags & RTF_GATEWAY) {
2428                 /*
2429                  * nd_rd->nd_rd_target must be a link-local address in
2430                  * better router cases.
2431                  */
2432                 if (!router_ll6)
2433                         goto fail;
2434                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2435                       sizeof(nd_rd->nd_rd_target));
2436                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2437                       sizeof(nd_rd->nd_rd_dst));
2438         } else {
2439                 /* make sure redtgt == reddst */
2440                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2441                       sizeof(nd_rd->nd_rd_target));
2442                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2443                       sizeof(nd_rd->nd_rd_dst));
2444         }
2445
2446         p = (u_char *)(nd_rd + 1);
2447
2448         if (!router_ll6)
2449                 goto nolladdropt;
2450
2451     {
2452         /* target lladdr option */
2453         struct rtentry *rt_router = NULL;
2454         int len;
2455         struct sockaddr_dl *sdl;
2456         struct nd_opt_hdr *nd_opt;
2457         char *lladdr;
2458
2459         rt_router = nd6_lookup(router_ll6, 0, ifp);
2460         if (!rt_router)
2461                 goto nolladdropt;
2462         len = sizeof(*nd_opt) + ifp->if_addrlen;
2463         len = (len + 7) & ~7;   /* round by 8 */
2464         /* safety check */
2465         if (len + (p - (u_char *)ip6) > maxlen)
2466                 goto nolladdropt;
2467         if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2468             (rt_router->rt_flags & RTF_LLINFO) &&
2469             (rt_router->rt_gateway->sa_family == AF_LINK) &&
2470             (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2471             sdl->sdl_alen) {
2472                 nd_opt = (struct nd_opt_hdr *)p;
2473                 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2474                 nd_opt->nd_opt_len = len >> 3;
2475                 lladdr = (char *)(nd_opt + 1);
2476                 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2477                 p += len;
2478         }
2479     }
2480 nolladdropt:;
2481
2482         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2483
2484         /* just to be safe */
2485 #ifdef M_DECRYPTED      /*not openbsd*/
2486         if (m0->m_flags & M_DECRYPTED)
2487                 goto noredhdropt;
2488 #endif
2489         if (p - (u_char *)ip6 > maxlen)
2490                 goto noredhdropt;
2491
2492     {
2493         /* redirected header option */
2494         int len;
2495         struct nd_opt_rd_hdr *nd_opt_rh;
2496
2497         /*
2498          * compute the maximum size for icmp6 redirect header option.
2499          * XXX room for auth header?
2500          */
2501         len = maxlen - (p - (u_char *)ip6);
2502         len &= ~7;
2503
2504         /* This is just for simplicity. */
2505         if (m0->m_pkthdr.len != m0->m_len) {
2506                 if (m0->m_next) {
2507                         m_freem(m0->m_next);
2508                         m0->m_next = NULL;
2509                 }
2510                 m0->m_pkthdr.len = m0->m_len;
2511         }
2512
2513         /*
2514          * Redirected header option spec (RFC2461 4.6.3) talks nothing
2515          * about padding/truncate rule for the original IP packet.
2516          * From the discussion on IPv6imp in Feb 1999, the consensus was:
2517          * - "attach as much as possible" is the goal
2518          * - pad if not aligned (original size can be guessed by original
2519          *   ip6 header)
2520          * Following code adds the padding if it is simple enough,
2521          * and truncates if not.
2522          */
2523         if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2524                 panic("assumption failed in %s:%d", __FILE__, __LINE__);
2525
2526         if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2527                 /* not enough room, truncate */
2528                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2529         } else {
2530                 /* enough room, pad or truncate */
2531                 size_t extra;
2532
2533                 extra = m0->m_pkthdr.len % 8;
2534                 if (extra) {
2535                         /* pad if easy enough, truncate if not */
2536                         if (8 - extra <= M_TRAILINGSPACE(m0)) {
2537                                 /* pad */
2538                                 m0->m_len += (8 - extra);
2539                                 m0->m_pkthdr.len += (8 - extra);
2540                         } else {
2541                                 /* truncate */
2542                                 m0->m_pkthdr.len -= extra;
2543                                 m0->m_len -= extra;
2544                         }
2545                 }
2546                 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2547                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2548         }
2549
2550         nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2551         bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2552         nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2553         nd_opt_rh->nd_opt_rh_len = len >> 3;
2554         p += sizeof(*nd_opt_rh);
2555         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2556
2557         /* connect m0 to m */
2558         m->m_next = m0;
2559         m->m_pkthdr.len = m->m_len + m0->m_len;
2560     }
2561 noredhdropt:;
2562
2563         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2564                 sip6->ip6_src.s6_addr16[1] = 0;
2565         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2566                 sip6->ip6_dst.s6_addr16[1] = 0;
2567 #if 0
2568         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2569                 ip6->ip6_src.s6_addr16[1] = 0;
2570         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2571                 ip6->ip6_dst.s6_addr16[1] = 0;
2572 #endif
2573         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2574                 nd_rd->nd_rd_target.s6_addr16[1] = 0;
2575         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2576                 nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2577
2578         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2579
2580         nd_rd->nd_rd_cksum = 0;
2581         nd_rd->nd_rd_cksum
2582                 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2583
2584         /* send the packet to outside... */
2585         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2586         if (outif) {
2587                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2588                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2589         }
2590         icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2591
2592         return;
2593
2594 fail:
2595         if (m)
2596                 m_freem(m);
2597         if (m0)
2598                 m_freem(m0);
2599 }
2600
2601 #ifdef HAVE_NRL_INPCB
2602 #define in6pcb          inpcb
2603 #define in6p_icmp6filt  inp_icmp6filt
2604 #endif
2605 /*
2606  * ICMPv6 socket option processing.
2607  */
2608 void
2609 icmp6_ctloutput(netmsg_t msg)
2610 {
2611         struct socket *so = msg->ctloutput.base.nm_so;
2612         struct sockopt *sopt = msg->ctloutput.nm_sopt;
2613         int error = 0;
2614         int optlen;
2615         struct inpcb *inp = so->so_pcb;
2616         int level, op, optname;
2617
2618         if (sopt) {
2619                 level = sopt->sopt_level;
2620                 op = sopt->sopt_dir;
2621                 optname = sopt->sopt_name;
2622                 optlen = sopt->sopt_valsize;
2623         } else
2624                 level = op = optname = optlen = 0;
2625
2626         if (level != IPPROTO_ICMPV6) {
2627                 error = EINVAL;
2628                 goto out;
2629         }
2630
2631         switch (op) {
2632         case PRCO_SETOPT:
2633                 switch (optname) {
2634                 case ICMP6_FILTER:
2635                     {
2636                         struct icmp6_filter *p;
2637
2638                         if (optlen != sizeof(*p)) {
2639                                 error = EMSGSIZE;
2640                                 break;
2641                         }
2642                         if (inp->in6p_icmp6filt == NULL) {
2643                                 error = EINVAL;
2644                                 break;
2645                         }
2646                         error = soopt_to_kbuf(sopt, inp->in6p_icmp6filt, optlen,
2647                                 optlen);
2648                         break;
2649                     }
2650
2651                 default:
2652                         error = ENOPROTOOPT;
2653                         break;
2654                 }
2655                 break;
2656
2657         case PRCO_GETOPT:
2658                 switch (optname) {
2659                 case ICMP6_FILTER:
2660                     {
2661                         if (inp->in6p_icmp6filt == NULL) {
2662                                 error = EINVAL;
2663                                 break;
2664                         }
2665                         soopt_from_kbuf(sopt, inp->in6p_icmp6filt,
2666                                 sizeof(struct icmp6_filter));
2667                         break;
2668                     }
2669
2670                 default:
2671                         error = ENOPROTOOPT;
2672                         break;
2673                 }
2674                 break;
2675         }
2676 out:
2677         lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
2678 }
2679 #ifdef HAVE_NRL_INPCB
2680 #undef in6pcb
2681 #undef in6p_icmp6filt
2682 #endif
2683
2684 #ifndef HAVE_PPSRATECHECK
2685 #ifndef timersub
2686 #define timersub(tvp, uvp, vvp)                                         \
2687         do {                                                            \
2688                 (vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;          \
2689                 (vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;       \
2690                 if ((vvp)->tv_usec < 0) {                               \
2691                         (vvp)->tv_sec--;                                \
2692                         (vvp)->tv_usec += 1000000;                      \
2693                 }                                                       \
2694         } while (0)
2695 #endif
2696
2697 /*
2698  * ppsratecheck(): packets (or events) per second limitation.
2699  */
2700 static int
2701 ppsratecheck(struct timeval *lasttime, int *curpps,
2702              int maxpps)        /* maximum pps allowed */
2703 {
2704         struct timeval tv, delta;
2705         int rv;
2706
2707         microtime(&tv);
2708
2709         timersub(&tv, lasttime, &delta);
2710
2711         /*
2712          * Check for 0,0 so that the message will be seen at least once.
2713          * If more than one second has passed since the last update of
2714          * lasttime, reset the counter.
2715          *
2716          * We do increment *curpps even in *curpps < maxpps case, as some may
2717          * try to use *curpps for stat purposes as well.
2718          */
2719         if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2720             delta.tv_sec >= 1) {
2721                 *lasttime = tv;
2722                 *curpps = 0;
2723                 rv = 1;
2724         } else if (maxpps < 0)
2725                 rv = 1;
2726         else if (*curpps < maxpps)
2727                 rv = 1;
2728         else
2729                 rv = 0;
2730
2731 #if 1 /* DIAGNOSTIC? */
2732         /* be careful about wrap-around */
2733         if (*curpps + 1 > *curpps)
2734                 *curpps = *curpps + 1;
2735 #else
2736         /*
2737          * assume that there's not too many calls to this function.
2738          * not sure if the assumption holds, as it depends on *caller's*
2739          * behavior, not the behavior of this function.
2740          * IMHO it is wrong to make assumption on the caller's behavior,
2741          * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2742          */
2743         *curpps = *curpps + 1;
2744 #endif
2745
2746         return (rv);
2747 }
2748 #endif
2749
2750 /*
2751  * Perform rate limit check.
2752  * Returns 0 if it is okay to send the icmp6 packet.
2753  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2754  * limitation.
2755  *
2756  * XXX per-destination/type check necessary?
2757  */
2758 static int
2759 icmp6_ratelimit(const struct in6_addr *dst,     /* not used at this moment */
2760                 const int type,                 /* not used at this moment */
2761                 const int code)                 /* not used at this moment */
2762 {
2763         int ret;
2764
2765         ret = 0;        /* okay to send */
2766
2767         /* PPS limit */
2768         if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2769             icmp6errppslim)) {
2770                 /* The packet is subject to rate limit */
2771                 ret++;
2772         }
2773
2774         return ret;
2775 }