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 1.17 2005/02/01 16:09:37 hrs 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/time.h>
  81 #include <sys/kernel.h>
  82 #include <sys/syslog.h>
  83 #include <sys/domain.h>
  84
  85 #include <net/if.h>
  86 #include <net/route.h>
  87 #include <net/if_dl.h>
  88 #include <net/if_types.h>
  89
  90 #include <netinet/in.h>
  91 #include <netinet/in_var.h>
  92 #include <netinet/ip6.h>
  93 #include <netinet6/ip6_var.h>
  94 #include <netinet/icmp6.h>
  95 #include <netinet6/mld6_var.h>
  96 #include <netinet/in_pcb.h>
  97 #include <netinet6/in6_pcb.h>
  98 #include <netinet6/nd6.h>
  99 #include <netinet6/in6_ifattach.h>
 100 #include <netinet6/ip6protosw.h>
 101
 102 #ifdef IPSEC
 103 #include <netinet6/ipsec.h>
 104 #include <netproto/key/key.h>
 105 #endif
 106
 107 #ifdef FAST_IPSEC
 108 #include <netproto/ipsec/ipsec.h>
 109 #include <netproto/ipsec/key.h>
 110 #define IPSEC
 111 #endif
 112
 113 #include <net/net_osdep.h>
 114
 115 #ifdef HAVE_NRL_INPCB
 116 /* inpcb members */
 117 #define in6pcb          inpcb
 118 #define in6p_laddr      inp_laddr6
 119 #define in6p_faddr      inp_faddr6
 120 #define in6p_icmp6filt  inp_icmp6filt
 121 #define in6p_route      inp_route
 122 #define in6p_socket     inp_socket
 123 #define in6p_flags      inp_flags
 124 #define in6p_moptions   inp_moptions6
 125 #define in6p_outputopts inp_outputopts6
 126 #define in6p_ip6        inp_ipv6
 127 #define in6p_flowinfo   inp_flowinfo
 128 #define in6p_sp         inp_sp
 129 #define in6p_next       inp_next
 130 #define in6p_prev       inp_prev
 131 /* macro names */
 132 #define sotoin6pcb      sotoinpcb
 133 /* function names */
 134 #define in6_pcbdetach   in_pcbdetach
 135 #define in6_rtchange    in_rtchange
 136
 137 /*
 138  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
 139  * others...
 140  */
 141 #define in6p_ip6_nxt    inp_ipv6.ip6_nxt
 142 #endif
 143
 144 /*
 145  * ppratecheck() is available, so define it here.
 146  */
 147 #define HAVE_PPSRATECHECK
 148
 149 extern struct domain inet6domain;
 150 extern struct ip6protosw inet6sw[];
 151 extern u_char ip6_protox[];
 152
 153 struct icmp6stat icmp6stat;
 154
 155 extern struct inpcbinfo ripcbinfo;
 156 extern int icmp6errppslim;
 157 static int icmp6errpps_count = 0;
 158 static struct timeval icmp6errppslim_last;
 159 extern int icmp6_nodeinfo;
 160
 161 static void icmp6_errcount (struct icmp6errstat *, int, int);
 162 static int icmp6_rip6_input (struct mbuf **, int);
 163 static int icmp6_ratelimit (const struct in6_addr *, const int, const int);
 164 static const char *icmp6_redirect_diag (struct in6_addr *,
 165         struct in6_addr *, struct in6_addr *);
 166 #ifndef HAVE_PPSRATECHECK
 167 static int ppsratecheck (struct timeval *, int *, int);
 168 #endif
 169 static struct mbuf *ni6_input (struct mbuf *, int);
 170 static struct mbuf *ni6_nametodns (const char *, int, int);
 171 static int ni6_dnsmatch (const char *, int, const char *, int);
 172 static int ni6_addrs (struct icmp6_nodeinfo *, struct mbuf *,
 173                           struct ifnet **, char *);
 174 static int ni6_store_addrs (struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
 175                                 struct ifnet *, int);
 176 static int icmp6_notify_error (struct mbuf *, int, int, int);
 177
 178 #ifdef COMPAT_RFC1885
 179 static struct route_in6 icmp6_reflect_rt;
 180 #endif
 181
 182
 183 void
 184 icmp6_init(void)
 185 {
 186         mld6_init();
 187 }
 188
 189 static void
 190 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
 191 {
 192         switch (type) {
 193         case ICMP6_DST_UNREACH:
 194                 switch (code) {
 195                 case ICMP6_DST_UNREACH_NOROUTE:
 196                         stat->icp6errs_dst_unreach_noroute++;
 197                         return;
 198                 case ICMP6_DST_UNREACH_ADMIN:
 199                         stat->icp6errs_dst_unreach_admin++;
 200                         return;
 201                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 202                         stat->icp6errs_dst_unreach_beyondscope++;
 203                         return;
 204                 case ICMP6_DST_UNREACH_ADDR:
 205                         stat->icp6errs_dst_unreach_addr++;
 206                         return;
 207                 case ICMP6_DST_UNREACH_NOPORT:
 208                         stat->icp6errs_dst_unreach_noport++;
 209                         return;
 210                 }
 211                 break;
 212         case ICMP6_PACKET_TOO_BIG:
 213                 stat->icp6errs_packet_too_big++;
 214                 return;
 215         case ICMP6_TIME_EXCEEDED:
 216                 switch (code) {
 217                 case ICMP6_TIME_EXCEED_TRANSIT:
 218                         stat->icp6errs_time_exceed_transit++;
 219                         return;
 220                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 221                         stat->icp6errs_time_exceed_reassembly++;
 222                         return;
 223                 }
 224                 break;
 225         case ICMP6_PARAM_PROB:
 226                 switch (code) {
 227                 case ICMP6_PARAMPROB_HEADER:
 228                         stat->icp6errs_paramprob_header++;
 229                         return;
 230                 case ICMP6_PARAMPROB_NEXTHEADER:
 231                         stat->icp6errs_paramprob_nextheader++;
 232                         return;
 233                 case ICMP6_PARAMPROB_OPTION:
 234                         stat->icp6errs_paramprob_option++;
 235                         return;
 236                 }
 237                 break;
 238         case ND_REDIRECT:
 239                 stat->icp6errs_redirect++;
 240                 return;
 241         }
 242         stat->icp6errs_unknown++;
 243 }
 244
 245 /*
 246  * Generate an error packet of type error in response to bad IP6 packet.
 247  */
 248 void
 249 icmp6_error(struct mbuf *m, int type, int code, int param)
 250 {
 251         struct ip6_hdr *oip6, *nip6;
 252         struct icmp6_hdr *icmp6;
 253         u_int preplen;
 254         int off;
 255         int nxt;
 256
 257         icmp6stat.icp6s_error++;
 258
 259         /* count per-type-code statistics */
 260         icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
 261
 262 #ifdef M_DECRYPTED      /*not openbsd*/
 263         if (m->m_flags & M_DECRYPTED) {
 264                 icmp6stat.icp6s_canterror++;
 265                 goto freeit;
 266         }
 267 #endif
 268
 269 #ifndef PULLDOWN_TEST
 270         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 271 #else
 272         if (m->m_len < sizeof(struct ip6_hdr)) {
 273                 m = m_pullup(m, sizeof(struct ip6_hdr));
 274                 if (m == NULL)
 275                         return;
 276         }
 277 #endif
 278         oip6 = mtod(m, struct ip6_hdr *);
 279
 280         /*
 281          * Multicast destination check. For unrecognized option errors,
 282          * this check has already done in ip6_unknown_opt(), so we can
 283          * check only for other errors.
 284          */
 285         if ((m->m_flags & (M_BCAST|M_MCAST) ||
 286              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
 287             (type != ICMP6_PACKET_TOO_BIG &&
 288              (type != ICMP6_PARAM_PROB ||
 289               code != ICMP6_PARAMPROB_OPTION)))
 290                 goto freeit;
 291
 292         /* Source address check. XXX: the case of anycast source? */
 293         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
 294             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
 295                 goto freeit;
 296
 297         /*
 298          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
 299          * don't do it.
 300          */
 301         nxt = -1;
 302         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
 303         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
 304                 struct icmp6_hdr *icp;
 305
 306 #ifndef PULLDOWN_TEST
 307                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
 308                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 309 #else
 310                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
 311                         sizeof(*icp));
 312                 if (icp == NULL) {
 313                         icmp6stat.icp6s_tooshort++;
 314                         return;
 315                 }
 316 #endif
 317                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
 318                     icp->icmp6_type == ND_REDIRECT) {
 319                         /*
 320                          * ICMPv6 error
 321                          * Special case: for redirect (which is
 322                          * informational) we must not send icmp6 error.
 323                          */
 324                         icmp6stat.icp6s_canterror++;
 325                         goto freeit;
 326                 } else {
 327                         /* ICMPv6 informational - send the error */
 328                 }
 329         } else {
 330                 /* non-ICMPv6 - send the error */
 331         }
 332
 333         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
 334
 335         /* Finally, do rate limitation check. */
 336         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
 337                 icmp6stat.icp6s_toofreq++;
 338                 goto freeit;
 339         }
 340
 341         /*
 342          * OK, ICMP6 can be generated.
 343          */
 344
 345         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
 346                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
 347
 348         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
 349         M_PREPEND(m, preplen, MB_DONTWAIT);
 350         if (m && m->m_len < preplen)
 351                 m = m_pullup(m, preplen);
 352         if (m == NULL) {
 353                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
 354                 return;
 355         }
 356
 357         nip6 = mtod(m, struct ip6_hdr *);
 358         nip6->ip6_src  = oip6->ip6_src;
 359         nip6->ip6_dst  = oip6->ip6_dst;
 360
 361         if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
 362                 oip6->ip6_src.s6_addr16[1] = 0;
 363         if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
 364                 oip6->ip6_dst.s6_addr16[1] = 0;
 365
 366         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
 367         icmp6->icmp6_type = type;
 368         icmp6->icmp6_code = code;
 369         icmp6->icmp6_pptr = htonl((u_int32_t)param);
 370
 371         /*
 372          * icmp6_reflect() is designed to be in the input path.
 373          * icmp6_error() can be called from both input and outut path,
 374          * and if we are in output path rcvif could contain bogus value.
 375          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
 376          * information in ip header (nip6).
 377          */
 378         m->m_pkthdr.rcvif = NULL;
 379
 380         icmp6stat.icp6s_outhist[type]++;
 381         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
 382
 383         return;
 384
 385   freeit:
 386         /*
 387          * If we can't tell wheter or not we can generate ICMP6, free it.
 388          */
 389         m_freem(m);
 390 }
 391
 392 /*
 393  * Process a received ICMP6 message.
 394  */
 395 int
 396 icmp6_input(struct mbuf **mp, int *offp, int proto)
 397 {
 398         struct mbuf *m = *mp, *n;
 399         struct ip6_hdr *ip6, *nip6;
 400         struct icmp6_hdr *icmp6, *nicmp6;
 401         int off = *offp;
 402         int icmp6len = m->m_pkthdr.len - *offp;
 403         int code, sum, noff;
 404
 405 #ifndef PULLDOWN_TEST
 406         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
 407         /* m might change if M_LOOP.  So, call mtod after this */
 408 #endif
 409
 410         /*
 411          * Locate icmp6 structure in mbuf, and check
 412          * that not corrupted and of at least minimum length
 413          */
 414
 415         ip6 = mtod(m, struct ip6_hdr *);
 416         if (icmp6len < sizeof(struct icmp6_hdr)) {
 417                 icmp6stat.icp6s_tooshort++;
 418                 goto freeit;
 419         }
 420
 421         /*
 422          * calculate the checksum
 423          */
 424 #ifndef PULLDOWN_TEST
 425         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
 426 #else
 427         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
 428         if (icmp6 == NULL) {
 429                 icmp6stat.icp6s_tooshort++;
 430                 return IPPROTO_DONE;
 431         }
 432 #endif
 433         code = icmp6->icmp6_code;
 434
 435         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
 436                 nd6log((LOG_ERR,
 437                     "ICMP6 checksum error(%d|%x) %s\n",
 438                     icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
 439                 icmp6stat.icp6s_checksum++;
 440                 goto freeit;
 441         }
 442
 443         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
 444                 /*
 445                  * Deliver very specific ICMP6 type only.
 446                  * This is important to deilver TOOBIG.  Otherwise PMTUD
 447                  * will not work.
 448                  */
 449                 switch (icmp6->icmp6_type) {
 450                 case ICMP6_DST_UNREACH:
 451                 case ICMP6_PACKET_TOO_BIG:
 452                 case ICMP6_TIME_EXCEEDED:
 453                         break;
 454                 default:
 455                         goto freeit;
 456                 }
 457         }
 458
 459         icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
 460         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
 461         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
 462                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
 463
 464         switch (icmp6->icmp6_type) {
 465         case ICMP6_DST_UNREACH:
 466                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
 467                 switch (code) {
 468                 case ICMP6_DST_UNREACH_NOROUTE:
 469                         code = PRC_UNREACH_NET;
 470                         break;
 471                 case ICMP6_DST_UNREACH_ADMIN:
 472                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
 473                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
 474                         break;
 475                 case ICMP6_DST_UNREACH_ADDR:
 476                         code = PRC_HOSTDEAD;
 477                         break;
 478 #ifdef COMPAT_RFC1885
 479                 case ICMP6_DST_UNREACH_NOTNEIGHBOR:
 480                         code = PRC_UNREACH_SRCFAIL;
 481                         break;
 482 #else
 483                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 484                         /* I mean "source address was incorrect." */
 485                         code = PRC_PARAMPROB;
 486                         break;
 487 #endif
 488                 case ICMP6_DST_UNREACH_NOPORT:
 489                         code = PRC_UNREACH_PORT;
 490                         break;
 491                 default:
 492                         goto badcode;
 493                 }
 494                 goto deliver;
 495                 break;
 496
 497         case ICMP6_PACKET_TOO_BIG:
 498                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
 499                 if (code != 0)
 500                         goto badcode;
 501
 502                 code = PRC_MSGSIZE;
 503
 504                 /*
 505                  * Updating the path MTU will be done after examining
 506                  * intermediate extension headers.
 507                  */
 508                 goto deliver;
 509                 break;
 510
 511         case ICMP6_TIME_EXCEEDED:
 512                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
 513                 switch (code) {
 514                 case ICMP6_TIME_EXCEED_TRANSIT:
 515                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 516                         code += PRC_TIMXCEED_INTRANS;
 517                         break;
 518                 default:
 519                         goto badcode;
 520                 }
 521                 goto deliver;
 522                 break;
 523
 524         case ICMP6_PARAM_PROB:
 525                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
 526                 switch (code) {
 527                 case ICMP6_PARAMPROB_NEXTHEADER:
 528                         code = PRC_UNREACH_PROTOCOL;
 529                         break;
 530                 case ICMP6_PARAMPROB_HEADER:
 531                 case ICMP6_PARAMPROB_OPTION:
 532                         code = PRC_PARAMPROB;
 533                         break;
 534                 default:
 535                         goto badcode;
 536                 }
 537                 goto deliver;
 538                 break;
 539
 540         case ICMP6_ECHO_REQUEST:
 541                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
 542                 if (code != 0)
 543                         goto badcode;
 544                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
 545                         /* Give up remote */
 546                         break;
 547                 }
 548                 if ((n->m_flags & M_EXT) != 0
 549                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
 550                         struct mbuf *n0 = n;
 551                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
 552                         int n0len;
 553
 554                         /*
 555                          * Prepare an internal mbuf. m_pullup() doesn't
 556                          * always copy the length we specified.
 557                          */
 558                         if (maxlen >= MCLBYTES) {
 559                                 /* Give up remote */
 560                                 m_freem(n0);
 561                                 break;
 562                         }
 563                         MGETHDR(n, MB_DONTWAIT, n0->m_type);
 564                         n0len = n0->m_pkthdr.len;       /* save for use below */
 565                         if (n)
 566                                 M_MOVE_PKTHDR(n, n0);
 567                         if (n && maxlen >= MHLEN) {
 568                                 MCLGET(n, MB_DONTWAIT);
 569                                 if ((n->m_flags & M_EXT) == 0) {
 570                                         m_free(n);
 571                                         n = NULL;
 572                                 }
 573                         }
 574                         if (n == NULL) {
 575                                 /* Give up remote */
 576                                 m_freem(n0);
 577                                 break;
 578                         }
 579                         /*
 580                          * Copy IPv6 and ICMPv6 only.
 581                          */
 582                         nip6 = mtod(n, struct ip6_hdr *);
 583                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 584                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 585                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 586                         noff = sizeof(struct ip6_hdr);
 587                         /* new mbuf contains only ipv6+icmpv6 headers */
 588                         n->m_len = noff + sizeof(struct icmp6_hdr);
 589                         /*
 590                          * Adjust mbuf. ip6_plen will be adjusted in
 591                          * ip6_output().
 592                          */
 593                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 594                         /* recalculate complete packet size */
 595                         n->m_pkthdr.len = n0len + (noff - off);
 596                         n->m_next = n0;
 597                 } else {
 598                         nip6 = mtod(n, struct ip6_hdr *);
 599                         nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
 600                         noff = off;
 601                 }
 602                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 603                 nicmp6->icmp6_code = 0;
 604                 if (n) {
 605                         icmp6stat.icp6s_reflect++;
 606                         icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
 607                         icmp6_reflect(n, noff);
 608                 }
 609                 break;
 610
 611         case ICMP6_ECHO_REPLY:
 612                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
 613                 if (code != 0)
 614                         goto badcode;
 615                 break;
 616
 617         case MLD_LISTENER_QUERY:
 618         case MLD_LISTENER_REPORT:
 619                 if (icmp6len < sizeof(struct mld_hdr))
 620                         goto badlen;
 621                 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
 622                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
 623                 else
 624                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
 625                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 626                         /* give up local */
 627                         mld6_input(m, off);
 628                         m = NULL;
 629                         goto freeit;
 630                 }
 631                 mld6_input(n, off);
 632                 /* m stays. */
 633                 break;
 634
 635         case MLD_LISTENER_DONE:
 636                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
 637                 if (icmp6len < sizeof(struct mld_hdr))  /* necessary? */
 638                         goto badlen;
 639                 break;          /* nothing to be done in kernel */
 640
 641         case MLD_MTRACE_RESP:
 642         case MLD_MTRACE:
 643                 /* XXX: these two are experimental.  not officially defind. */
 644                 /* XXX: per-interface statistics? */
 645                 break;          /* just pass it to applications */
 646
 647         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 648             {
 649                 enum { WRU, FQDN } mode;
 650
 651                 if (!icmp6_nodeinfo)
 652                         break;
 653
 654                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 655                         mode = WRU;
 656                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 657                         mode = FQDN;
 658                 else
 659                         goto badlen;
 660
 661 #define hostnamelen     strlen(hostname)
 662                 if (mode == FQDN) {
 663 #ifndef PULLDOWN_TEST
 664                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
 665                                          IPPROTO_DONE);
 666 #endif
 667                         n = m_copy(m, 0, M_COPYALL);
 668                         if (n)
 669                                 n = ni6_input(n, off);
 670                         /* XXX meaningless if n == NULL */
 671                         noff = sizeof(struct ip6_hdr);
 672                 } else {
 673                         u_char *p;
 674                         int maxlen, maxhlen;
 675
 676                         if ((icmp6_nodeinfo & 5) != 5)
 677                                 break;
 678
 679                         if (code != 0)
 680                                 goto badcode;
 681                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
 682                         if (maxlen >= MCLBYTES) {
 683                                 /* Give up remote */
 684                                 break;
 685                         }
 686                         MGETHDR(n, MB_DONTWAIT, m->m_type);
 687                         if (n && maxlen > MHLEN) {
 688                                 MCLGET(n, MB_DONTWAIT);
 689                                 if ((n->m_flags & M_EXT) == 0) {
 690                                         m_free(n);
 691                                         n = NULL;
 692                                 }
 693                         }
 694                         if (!m_dup_pkthdr(n, m, MB_DONTWAIT)) {
 695                                 /*
 696                                  * Previous code did a blind M_COPY_PKTHDR
 697                                  * and said "just for rcvif".  If true, then
 698                                  * we could tolerate the dup failing (due to
 699                                  * the deep copy of the tag chain).  For now
 700                                  * be conservative and just fail.
 701                                  */
 702                                 m_free(n);
 703                                 n = NULL;
 704                         }
 705                         if (n == NULL) {
 706                                 /* Give up remote */
 707                                 break;
 708                         }
 709                         n->m_pkthdr.rcvif = NULL;
 710                         n->m_len = 0;
 711                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
 712                         if (maxhlen > hostnamelen)
 713                                 maxhlen = hostnamelen;
 714                         /*
 715                          * Copy IPv6 and ICMPv6 only.
 716                          */
 717                         nip6 = mtod(n, struct ip6_hdr *);
 718                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 719                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 720                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 721                         p = (u_char *)(nicmp6 + 1);
 722                         bzero(p, 4);
 723                         bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
 724                         noff = sizeof(struct ip6_hdr);
 725                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 726                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 727                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 728                         nicmp6->icmp6_code = 0;
 729                 }
 730 #undef hostnamelen
 731                 if (n) {
 732                         icmp6stat.icp6s_reflect++;
 733                         icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
 734                         icmp6_reflect(n, noff);
 735                 }
 736                 break;
 737             }
 738
 739         case ICMP6_WRUREPLY:
 740                 if (code != 0)
 741                         goto badcode;
 742                 break;
 743
 744         case ND_ROUTER_SOLICIT:
 745                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
 746                 if (code != 0)
 747                         goto badcode;
 748                 if (icmp6len < sizeof(struct nd_router_solicit))
 749                         goto badlen;
 750                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 751                         /* give up local */
 752                         nd6_rs_input(m, off, icmp6len);
 753                         m = NULL;
 754                         goto freeit;
 755                 }
 756                 nd6_rs_input(n, off, icmp6len);
 757                 /* m stays. */
 758                 break;
 759
 760         case ND_ROUTER_ADVERT:
 761                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
 762                 if (code != 0)
 763                         goto badcode;
 764                 if (icmp6len < sizeof(struct nd_router_advert))
 765                         goto badlen;
 766                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 767                         /* give up local */
 768                         nd6_ra_input(m, off, icmp6len);
 769                         m = NULL;
 770                         goto freeit;
 771                 }
 772                 nd6_ra_input(n, off, icmp6len);
 773                 /* m stays. */
 774                 break;
 775
 776         case ND_NEIGHBOR_SOLICIT:
 777                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
 778                 if (code != 0)
 779                         goto badcode;
 780                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 781                         goto badlen;
 782                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 783                         /* give up local */
 784                         nd6_ns_input(m, off, icmp6len);
 785                         m = NULL;
 786                         goto freeit;
 787                 }
 788                 nd6_ns_input(n, off, icmp6len);
 789                 /* m stays. */
 790                 break;
 791
 792         case ND_NEIGHBOR_ADVERT:
 793                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
 794                 if (code != 0)
 795                         goto badcode;
 796                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 797                         goto badlen;
 798                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 799                         /* give up local */
 800                         nd6_na_input(m, off, icmp6len);
 801                         m = NULL;
 802                         goto freeit;
 803                 }
 804                 nd6_na_input(n, off, icmp6len);
 805                 /* m stays. */
 806                 break;
 807
 808         case ND_REDIRECT:
 809                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
 810                 if (code != 0)
 811                         goto badcode;
 812                 if (icmp6len < sizeof(struct nd_redirect))
 813                         goto badlen;
 814                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 815                         /* give up local */
 816                         icmp6_redirect_input(m, off);
 817                         m = NULL;
 818                         goto freeit;
 819                 }
 820                 icmp6_redirect_input(n, off);
 821                 /* m stays. */
 822                 break;
 823
 824         case ICMP6_ROUTER_RENUMBERING:
 825                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 826                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 827                         goto badcode;
 828                 if (icmp6len < sizeof(struct icmp6_router_renum))
 829                         goto badlen;
 830                 break;
 831
 832         default:
 833                 nd6log((LOG_DEBUG,
 834                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 835                     icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
 836                     ip6_sprintf(&ip6->ip6_dst),
 837                     m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
 838                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 839                         /* ICMPv6 error: MUST deliver it by spec... */
 840                         code = PRC_NCMDS;
 841                         /* deliver */
 842                 } else {
 843                         /* ICMPv6 informational: MUST not deliver */
 844                         break;
 845                 }
 846         deliver:
 847                 if (icmp6_notify_error(m, off, icmp6len, code)) {
 848                         /* In this case, m should've been freed. */
 849                         return(IPPROTO_DONE);
 850                 }
 851                 break;
 852
 853         badcode:
 854                 icmp6stat.icp6s_badcode++;
 855                 break;
 856
 857         badlen:
 858                 icmp6stat.icp6s_badlen++;
 859                 break;
 860         }
 861
 862         /* deliver the packet to appropriate sockets */
 863         icmp6_rip6_input(&m, *offp);
 864
 865         return IPPROTO_DONE;
 866
 867  freeit:
 868         m_freem(m);
 869         return IPPROTO_DONE;
 870 }
 871
 872 static int
 873 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
 874 {
 875         struct icmp6_hdr *icmp6;
 876         struct ip6_hdr *eip6;
 877         u_int32_t notifymtu;
 878         struct sockaddr_in6 icmp6src, icmp6dst;
 879
 880         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 881                 icmp6stat.icp6s_tooshort++;
 882                 goto freeit;
 883         }
 884 #ifndef PULLDOWN_TEST
 885         IP6_EXTHDR_CHECK(m, off,
 886                          sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
 887                          -1);
 888         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 889 #else
 890         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
 891                        sizeof(*icmp6) + sizeof(struct ip6_hdr));
 892         if (icmp6 == NULL) {
 893                 icmp6stat.icp6s_tooshort++;
 894                 return(-1);
 895         }
 896 #endif
 897         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 898
 899         /* Detect the upper level protocol */
 900         {
 901                 void (*ctlfunc) (int, struct sockaddr *, void *);
 902                 u_int8_t nxt = eip6->ip6_nxt;
 903                 int eoff = off + sizeof(struct icmp6_hdr) +
 904                         sizeof(struct ip6_hdr);
 905                 struct ip6ctlparam ip6cp;
 906                 struct in6_addr *finaldst = NULL;
 907                 int icmp6type = icmp6->icmp6_type;
 908                 struct ip6_frag *fh;
 909                 struct ip6_rthdr *rth;
 910                 struct ip6_rthdr0 *rth0;
 911                 int rthlen;
 912
 913                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 914                         struct ip6_ext *eh;
 915
 916                         switch (nxt) {
 917                         case IPPROTO_HOPOPTS:
 918                         case IPPROTO_DSTOPTS:
 919                         case IPPROTO_AH:
 920 #ifndef PULLDOWN_TEST
 921                                 IP6_EXTHDR_CHECK(m, 0, eoff +
 922                                                  sizeof(struct ip6_ext),
 923                                                  -1);
 924                                 eh = (struct ip6_ext *)(mtod(m, caddr_t)
 925                                                         + eoff);
 926 #else
 927                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
 928                                                eoff, sizeof(*eh));
 929                                 if (eh == NULL) {
 930                                         icmp6stat.icp6s_tooshort++;
 931                                         return(-1);
 932                                 }
 933 #endif
 934
 935                                 if (nxt == IPPROTO_AH)
 936                                         eoff += (eh->ip6e_len + 2) << 2;
 937                                 else
 938                                         eoff += (eh->ip6e_len + 1) << 3;
 939                                 nxt = eh->ip6e_nxt;
 940                                 break;
 941                         case IPPROTO_ROUTING:
 942                                 /*
 943                                  * When the erroneous packet contains a
 944                                  * routing header, we should examine the
 945                                  * header to determine the final destination.
 946                                  * Otherwise, we can't properly update
 947                                  * information that depends on the final
 948                                  * destination (e.g. path MTU).
 949                                  */
 950 #ifndef PULLDOWN_TEST
 951                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
 952                                                  -1);
 953                                 rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
 954                                                            + eoff);
 955 #else
 956                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
 957                                                eoff, sizeof(*rth));
 958                                 if (rth == NULL) {
 959                                         icmp6stat.icp6s_tooshort++;
 960                                         return(-1);
 961                                 }
 962 #endif
 963                                 rthlen = (rth->ip6r_len + 1) << 3;
 964                                 /*
 965                                  * XXX: currently there is no
 966                                  * officially defined type other
 967                                  * than type-0.
 968                                  * Note that if the segment left field
 969                                  * is 0, all intermediate hops must
 970                                  * have been passed.
 971                                  */
 972                                 if (rth->ip6r_segleft &&
 973                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
 974                                         int hops;
 975
 976 #ifndef PULLDOWN_TEST
 977                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
 978                                                          -1);
 979                                         rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
 980 #else
 981                                         IP6_EXTHDR_GET(rth0,
 982                                                        struct ip6_rthdr0 *, m,
 983                                                        eoff, rthlen);
 984                                         if (rth0 == NULL) {
 985                                                 icmp6stat.icp6s_tooshort++;
 986                                                 return(-1);
 987                                         }
 988 #endif
 989                                         /* just ignore a bogus header */
 990                                         if ((rth0->ip6r0_len % 2) == 0 &&
 991                                             (hops = rth0->ip6r0_len/2))
 992                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
 993                                 }
 994                                 eoff += rthlen;
 995                                 nxt = rth->ip6r_nxt;
 996                                 break;
 997                         case IPPROTO_FRAGMENT:
 998 #ifndef PULLDOWN_TEST
 999                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1000                                                  sizeof(struct ip6_frag),
1001                                                  -1);
1002                                 fh = (struct ip6_frag *)(mtod(m, caddr_t)
1003                                                          + eoff);
1004 #else
1005                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1006                                                eoff, sizeof(*fh));
1007                                 if (fh == NULL) {
1008                                         icmp6stat.icp6s_tooshort++;
1009                                         return(-1);
1010                                 }
1011 #endif
1012                                 /*
1013                                  * Data after a fragment header is meaningless
1014                                  * unless it is the first fragment, but
1015                                  * we'll go to the notify label for path MTU
1016                                  * discovery.
1017                                  */
1018                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1019                                         goto notify;
1020
1021                                 eoff += sizeof(struct ip6_frag);
1022                                 nxt = fh->ip6f_nxt;
1023                                 break;
1024                         default:
1025                                 /*
1026                                  * This case includes ESP and the No Next
1027                                  * Header.  In such cases going to the notify
1028                                  * label does not have any meaning
1029                                  * (i.e. ctlfunc will be NULL), but we go
1030                                  * anyway since we might have to update
1031                                  * path MTU information.
1032                                  */
1033                                 goto notify;
1034                         }
1035                 }
1036           notify:
1037 #ifndef PULLDOWN_TEST
1038                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1039 #else
1040                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1041                                sizeof(*icmp6) + sizeof(struct ip6_hdr));
1042                 if (icmp6 == NULL) {
1043                         icmp6stat.icp6s_tooshort++;
1044                         return(-1);
1045                 }
1046 #endif
1047
1048                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1049                 bzero(&icmp6dst, sizeof(icmp6dst));
1050                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1051                 icmp6dst.sin6_family = AF_INET6;
1052                 if (finaldst == NULL)
1053                         icmp6dst.sin6_addr = eip6->ip6_dst;
1054                 else
1055                         icmp6dst.sin6_addr = *finaldst;
1056                 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1057                                                           &icmp6dst.sin6_addr);
1058                 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1059                                    NULL, NULL)) {
1060                         /* should be impossbile */
1061                         nd6log((LOG_DEBUG,
1062                             "icmp6_notify_error: in6_embedscope failed\n"));
1063                         goto freeit;
1064                 }
1065
1066                 /*
1067                  * retrieve parameters from the inner IPv6 header, and convert
1068                  * them into sockaddr structures.
1069                  */
1070                 bzero(&icmp6src, sizeof(icmp6src));
1071                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1072                 icmp6src.sin6_family = AF_INET6;
1073                 icmp6src.sin6_addr = eip6->ip6_src;
1074                 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1075                                                           &icmp6src.sin6_addr);
1076                 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1077                                    NULL, NULL)) {
1078                         /* should be impossbile */
1079                         nd6log((LOG_DEBUG,
1080                             "icmp6_notify_error: in6_embedscope failed\n"));
1081                         goto freeit;
1082                 }
1083                 icmp6src.sin6_flowinfo =
1084                         (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1085
1086                 if (finaldst == NULL)
1087                         finaldst = &eip6->ip6_dst;
1088                 ip6cp.ip6c_m = m;
1089                 ip6cp.ip6c_icmp6 = icmp6;
1090                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1091                 ip6cp.ip6c_off = eoff;
1092                 ip6cp.ip6c_finaldst = finaldst;
1093                 ip6cp.ip6c_src = &icmp6src;
1094                 ip6cp.ip6c_nxt = nxt;
1095
1096                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1097                         notifymtu = ntohl(icmp6->icmp6_mtu);
1098                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1099                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1100                 }
1101
1102                 ctlfunc = (void (*) (int, struct sockaddr *, void *))
1103                         (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1104                 if (ctlfunc) {
1105                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1106                                           &ip6cp);
1107                 }
1108         }
1109         return(0);
1110
1111   freeit:
1112         m_freem(m);
1113         return(-1);
1114 }
1115
1116 void
1117 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1118 {
1119         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1120         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1121         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1122         u_int mtu = ntohl(icmp6->icmp6_mtu);
1123         struct rtentry *rt = NULL;
1124         struct sockaddr_in6 sin6;
1125
1126         if (!validated)
1127                 return;
1128
1129         bzero(&sin6, sizeof(sin6));
1130         sin6.sin6_family = PF_INET6;
1131         sin6.sin6_len = sizeof(struct sockaddr_in6);
1132         sin6.sin6_addr = *dst;
1133         /* XXX normally, this won't happen */
1134         if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1135                 sin6.sin6_addr.s6_addr16[1] =
1136                     htons(m->m_pkthdr.rcvif->if_index);
1137         }
1138         /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1139         rt = rtpurelookup((struct sockaddr *)&sin6);
1140
1141         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1142             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1143                 if (mtu < IPV6_MMTU) {                          /* XXX */
1144                         rt->rt_rmx.rmx_locks |= RTV_MTU;
1145                 } else if (mtu < rt->rt_ifp->if_mtu &&
1146                            rt->rt_rmx.rmx_mtu > mtu) {
1147                         icmp6stat.icp6s_pmtuchg++;
1148                         rt->rt_rmx.rmx_mtu = mtu;
1149                 }
1150         }
1151         if (rt)
1152                 --rt->rt_refcnt;
1153 }
1154
1155 /*
1156  * Process a Node Information Query packet, based on
1157  * draft-ietf-ipngwg-icmp-name-lookups-07.
1158  *
1159  * Spec incompatibilities:
1160  * - IPv6 Subject address handling
1161  * - IPv4 Subject address handling support missing
1162  * - Proxy reply (answer even if it's not for me)
1163  * - joins NI group address at in6_ifattach() time only, does not cope
1164  *   with hostname changes by sethostname(3)
1165  */
1166 #define hostnamelen     strlen(hostname)
1167 static struct mbuf *
1168 ni6_input(struct mbuf *m, int off)
1169 {
1170         struct icmp6_nodeinfo *ni6, *nni6;
1171         struct mbuf *n = NULL;
1172         u_int16_t qtype;
1173         int subjlen;
1174         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1175         struct ni_reply_fqdn *fqdn;
1176         int addrs;              /* for NI_QTYPE_NODEADDR */
1177         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1178         struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1179         struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1180         struct ip6_hdr *ip6;
1181         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1182         char *subj = NULL;
1183         struct in6_ifaddr *ia6 = NULL;
1184
1185         ip6 = mtod(m, struct ip6_hdr *);
1186 #ifndef PULLDOWN_TEST
1187         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1188 #else
1189         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1190         if (ni6 == NULL) {
1191                 /* m is already reclaimed */
1192                 return NULL;
1193         }
1194 #endif
1195
1196         /*
1197          * Validate IPv6 destination address.
1198          *
1199          * The Responder must discard the Query without further processing
1200          * unless it is one of the Responder's unicast or anycast addresses, or
1201          * a link-local scope multicast address which the Responder has joined.
1202          * [icmp-name-lookups-07, Section 4.]
1203          */
1204         bzero(&sin6, sizeof(sin6));
1205         sin6.sin6_family = AF_INET6;
1206         sin6.sin6_len = sizeof(struct sockaddr_in6);
1207         bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1208         /* XXX scopeid */
1209         if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1210                 /* unicast/anycast, fine */
1211                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1212                     (icmp6_nodeinfo & 4) == 0) {
1213                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1214                                 "a temporary address in %s:%d",
1215                                __FILE__, __LINE__));
1216                         goto bad;
1217                 }
1218         } else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1219                 ; /* link-local multicast, fine */
1220         else
1221                 goto bad;
1222
1223         /* validate query Subject field. */
1224         qtype = ntohs(ni6->ni_qtype);
1225         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1226         switch (qtype) {
1227         case NI_QTYPE_NOOP:
1228         case NI_QTYPE_SUPTYPES:
1229                 /* 07 draft */
1230                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1231                         break;
1232                 /* FALLTHROUGH */
1233         case NI_QTYPE_FQDN:
1234         case NI_QTYPE_NODEADDR:
1235                 switch (ni6->ni_code) {
1236                 case ICMP6_NI_SUBJ_IPV6:
1237 #if ICMP6_NI_SUBJ_IPV6 != 0
1238                 case 0:
1239 #endif
1240                         /*
1241                          * backward compatibility - try to accept 03 draft
1242                          * format, where no Subject is present.
1243                          */
1244                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1245                             subjlen == 0) {
1246                                 oldfqdn++;
1247                                 break;
1248                         }
1249 #if ICMP6_NI_SUBJ_IPV6 != 0
1250                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1251                                 goto bad;
1252 #endif
1253
1254                         if (subjlen != sizeof(sin6.sin6_addr))
1255                                 goto bad;
1256
1257                         /*
1258                          * Validate Subject address.
1259                          *
1260                          * Not sure what exactly "address belongs to the node"
1261                          * means in the spec, is it just unicast, or what?
1262                          *
1263                          * At this moment we consider Subject address as
1264                          * "belong to the node" if the Subject address equals
1265                          * to the IPv6 destination address; validation for
1266                          * IPv6 destination address should have done enough
1267                          * check for us.
1268                          *
1269                          * We do not do proxy at this moment.
1270                          */
1271                         /* m_pulldown instead of copy? */
1272                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1273                             subjlen, (caddr_t)&sin6.sin6_addr);
1274                         sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1275                                                               &sin6.sin6_addr);
1276                         in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1277                         bzero(&sin6_d, sizeof(sin6_d));
1278                         sin6_d.sin6_family = AF_INET6; /* not used, actually */
1279                         sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1280                         sin6_d.sin6_addr = ip6->ip6_dst;
1281                         sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1282                                                                 &ip6->ip6_dst);
1283                         in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1284                         subj = (char *)&sin6;
1285                         if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1286                                 break;
1287
1288                         /*
1289                          * XXX if we are to allow other cases, we should really
1290                          * be careful about scope here.
1291                          * basically, we should disallow queries toward IPv6
1292                          * destination X with subject Y, if scope(X) > scope(Y).
1293                          * if we allow scope(X) > scope(Y), it will result in
1294                          * information leakage across scope boundary.
1295                          */
1296                         goto bad;
1297
1298                 case ICMP6_NI_SUBJ_FQDN:
1299                         /*
1300                          * Validate Subject name with gethostname(3).
1301                          *
1302                          * The behavior may need some debate, since:
1303                          * - we are not sure if the node has FQDN as
1304                          *   hostname (returned by gethostname(3)).
1305                          * - the code does wildcard match for truncated names.
1306                          *   however, we are not sure if we want to perform
1307                          *   wildcard match, if gethostname(3) side has
1308                          *   truncated hostname.
1309                          */
1310                         n = ni6_nametodns(hostname, hostnamelen, 0);
1311                         if (!n || n->m_next || n->m_len == 0)
1312                                 goto bad;
1313                         IP6_EXTHDR_GET(subj, char *, m,
1314                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1315                         if (subj == NULL)
1316                                 goto bad;
1317                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1318                                         n->m_len)) {
1319                                 goto bad;
1320                         }
1321                         m_freem(n);
1322                         n = NULL;
1323                         break;
1324
1325                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1326                 default:
1327                         goto bad;
1328                 }
1329                 break;
1330         }
1331
1332         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1333         switch (qtype) {
1334         case NI_QTYPE_FQDN:
1335                 if ((icmp6_nodeinfo & 1) == 0)
1336                         goto bad;
1337                 break;
1338         case NI_QTYPE_NODEADDR:
1339                 if ((icmp6_nodeinfo & 2) == 0)
1340                         goto bad;
1341                 break;
1342         }
1343
1344         /* guess reply length */
1345         switch (qtype) {
1346         case NI_QTYPE_NOOP:
1347                 break;          /* no reply data */
1348         case NI_QTYPE_SUPTYPES:
1349                 replylen += sizeof(u_int32_t);
1350                 break;
1351         case NI_QTYPE_FQDN:
1352                 /* XXX will append an mbuf */
1353                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1354                 break;
1355         case NI_QTYPE_NODEADDR:
1356                 addrs = ni6_addrs(ni6, m, &ifp, subj);
1357                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1358                                           sizeof(u_int32_t))) > MCLBYTES)
1359                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1360                 break;
1361         default:
1362                 /*
1363                  * XXX: We must return a reply with the ICMP6 code
1364                  * `unknown Qtype' in this case. However we regard the case
1365                  * as an FQDN query for backward compatibility.
1366                  * Older versions set a random value to this field,
1367                  * so it rarely varies in the defined qtypes.
1368                  * But the mechanism is not reliable...
1369                  * maybe we should obsolete older versions.
1370                  */
1371                 qtype = NI_QTYPE_FQDN;
1372                 /* XXX will append an mbuf */
1373                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1374                 oldfqdn++;
1375                 break;
1376         }
1377
1378         /* allocate an mbuf to reply. */
1379         MGETHDR(n, MB_DONTWAIT, m->m_type);
1380         if (n == NULL) {
1381                 m_freem(m);
1382                 return(NULL);
1383         }
1384         M_MOVE_PKTHDR(n, m); /* just for recvif */
1385         if (replylen > MHLEN) {
1386                 if (replylen > MCLBYTES) {
1387                         /*
1388                          * XXX: should we try to allocate more? But MCLBYTES
1389                          * is probably much larger than IPV6_MMTU...
1390                          */
1391                         goto bad;
1392                 }
1393                 MCLGET(n, MB_DONTWAIT);
1394                 if ((n->m_flags & M_EXT) == 0) {
1395                         goto bad;
1396                 }
1397         }
1398         n->m_pkthdr.len = n->m_len = replylen;
1399
1400         /* copy mbuf header and IPv6 + Node Information base headers */
1401         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1402         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1403         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1404
1405         /* qtype dependent procedure */
1406         switch (qtype) {
1407         case NI_QTYPE_NOOP:
1408                 nni6->ni_code = ICMP6_NI_SUCCESS;
1409                 nni6->ni_flags = 0;
1410                 break;
1411         case NI_QTYPE_SUPTYPES:
1412         {
1413                 u_int32_t v;
1414                 nni6->ni_code = ICMP6_NI_SUCCESS;
1415                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1416                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1417                 v = (u_int32_t)htonl(0x0000000f);
1418                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1419                 break;
1420         }
1421         case NI_QTYPE_FQDN:
1422                 nni6->ni_code = ICMP6_NI_SUCCESS;
1423                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1424                                                 sizeof(struct ip6_hdr) +
1425                                                 sizeof(struct icmp6_nodeinfo));
1426                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1427                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1428                 /*
1429                  * XXX do we really have FQDN in variable "hostname"?
1430                  */
1431                 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1432                 if (n->m_next == NULL)
1433                         goto bad;
1434                 /* XXX we assume that n->m_next is not a chain */
1435                 if (n->m_next->m_next != NULL)
1436                         goto bad;
1437                 n->m_pkthdr.len += n->m_next->m_len;
1438                 break;
1439         case NI_QTYPE_NODEADDR:
1440         {
1441                 int lenlim, copied;
1442
1443                 nni6->ni_code = ICMP6_NI_SUCCESS;
1444                 n->m_pkthdr.len = n->m_len =
1445                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1446                 lenlim = M_TRAILINGSPACE(n);
1447                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1448                 /* XXX: reset mbuf length */
1449                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1450                         sizeof(struct icmp6_nodeinfo) + copied;
1451                 break;
1452         }
1453         default:
1454                 break;          /* XXX impossible! */
1455         }
1456
1457         nni6->ni_type = ICMP6_NI_REPLY;
1458         m_freem(m);
1459         return(n);
1460
1461   bad:
1462         m_freem(m);
1463         if (n)
1464                 m_freem(n);
1465         return(NULL);
1466 }
1467 #undef hostnamelen
1468
1469 /*
1470  * make a mbuf with DNS-encoded string.  no compression support.
1471  *
1472  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1473  * treated as truncated name (two \0 at the end).  this is a wild guess.
1474  */
1475 static struct mbuf *
1476 ni6_nametodns(const char *name, int namelen,
1477               int old) /* return pascal string if non-zero */
1478 {
1479         struct mbuf *m;
1480         char *cp, *ep;
1481         const char *p, *q;
1482         int i, len, nterm;
1483
1484         if (old)
1485                 len = namelen + 1;
1486         else
1487                 len = MCLBYTES;
1488
1489         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1490         MGET(m, MB_DONTWAIT, MT_DATA);
1491         if (m && len > MLEN) {
1492                 MCLGET(m, MB_DONTWAIT);
1493                 if ((m->m_flags & M_EXT) == 0)
1494                         goto fail;
1495         }
1496         if (!m)
1497                 goto fail;
1498         m->m_next = NULL;
1499
1500         if (old) {
1501                 m->m_len = len;
1502                 *mtod(m, char *) = namelen;
1503                 bcopy(name, mtod(m, char *) + 1, namelen);
1504                 return m;
1505         } else {
1506                 m->m_len = 0;
1507                 cp = mtod(m, char *);
1508                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1509
1510                 /* if not certain about my name, return empty buffer */
1511                 if (namelen == 0)
1512                         return m;
1513
1514                 /*
1515                  * guess if it looks like shortened hostname, or FQDN.
1516                  * shortened hostname needs two trailing "\0".
1517                  */
1518                 i = 0;
1519                 for (p = name; p < name + namelen; p++) {
1520                         if (*p && *p == '.')
1521                                 i++;
1522                 }
1523                 if (i < 2)
1524                         nterm = 2;
1525                 else
1526                         nterm = 1;
1527
1528                 p = name;
1529                 while (cp < ep && p < name + namelen) {
1530                         i = 0;
1531                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1532                                 i++;
1533                         /* result does not fit into mbuf */
1534                         if (cp + i + 1 >= ep)
1535                                 goto fail;
1536                         /*
1537                          * DNS label length restriction, RFC1035 page 8.
1538                          * "i == 0" case is included here to avoid returning
1539                          * 0-length label on "foo..bar".
1540                          */
1541                         if (i <= 0 || i >= 64)
1542                                 goto fail;
1543                         *cp++ = i;
1544                         bcopy(p, cp, i);
1545                         cp += i;
1546                         p = q;
1547                         if (p < name + namelen && *p == '.')
1548                                 p++;
1549                 }
1550                 /* termination */
1551                 if (cp + nterm >= ep)
1552                         goto fail;
1553                 while (nterm-- > 0)
1554                         *cp++ = '\0';
1555                 m->m_len = cp - mtod(m, char *);
1556                 return m;
1557         }
1558
1559         panic("should not reach here");
1560         /* NOTREACHED */
1561
1562  fail:
1563         if (m)
1564                 m_freem(m);
1565         return NULL;
1566 }
1567
1568 /*
1569  * check if two DNS-encoded string matches.  takes care of truncated
1570  * form (with \0\0 at the end).  no compression support.
1571  * XXX upper/lowercase match (see RFC2065)
1572  */
1573 static int
1574 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1575 {
1576         const char *a0, *b0;
1577         int l;
1578
1579         /* simplest case - need validation? */
1580         if (alen == blen && bcmp(a, b, alen) == 0)
1581                 return 1;
1582
1583         a0 = a;
1584         b0 = b;
1585
1586         /* termination is mandatory */
1587         if (alen < 2 || blen < 2)
1588                 return 0;
1589         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1590                 return 0;
1591         alen--;
1592         blen--;
1593
1594         while (a - a0 < alen && b - b0 < blen) {
1595                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1596                         return 0;
1597
1598                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1599                         return 0;
1600                 /* we don't support compression yet */
1601                 if (a[0] >= 64 || b[0] >= 64)
1602                         return 0;
1603
1604                 /* truncated case */
1605                 if (a[0] == 0 && a - a0 == alen - 1)
1606                         return 1;
1607                 if (b[0] == 0 && b - b0 == blen - 1)
1608                         return 1;
1609                 if (a[0] == 0 || b[0] == 0)
1610                         return 0;
1611
1612                 if (a[0] != b[0])
1613                         return 0;
1614                 l = a[0];
1615                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1616                         return 0;
1617                 if (bcmp(a + 1, b + 1, l) != 0)
1618                         return 0;
1619
1620                 a += 1 + l;
1621                 b += 1 + l;
1622         }
1623
1624         if (a - a0 == alen && b - b0 == blen)
1625                 return 1;
1626         else
1627                 return 0;
1628 }
1629
1630 /*
1631  * calculate the number of addresses to be returned in the node info reply.
1632  */
1633 static int
1634 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1635           char *subj)
1636 {
1637         struct ifnet *ifp;
1638         struct in6_ifaddr *ifa6;
1639         struct ifaddr *ifa;
1640         struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1641         int addrs = 0, addrsofif, iffound = 0;
1642         int niflags = ni6->ni_flags;
1643
1644         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1645                 switch (ni6->ni_code) {
1646                 case ICMP6_NI_SUBJ_IPV6:
1647                         if (subj == NULL) /* must be impossible... */
1648                                 return(0);
1649                         subj_ip6 = (struct sockaddr_in6 *)subj;
1650                         break;
1651                 default:
1652                         /*
1653                          * XXX: we only support IPv6 subject address for
1654                          * this Qtype.
1655                          */
1656                         return(0);
1657                 }
1658         }
1659
1660         for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1661         {
1662                 addrsofif = 0;
1663                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1664                 {
1665                         if (ifa->ifa_addr->sa_family != AF_INET6)
1666                                 continue;
1667                         ifa6 = (struct in6_ifaddr *)ifa;
1668
1669                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1670                             IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1671                                                &ifa6->ia_addr.sin6_addr))
1672                                 iffound = 1;
1673
1674                         /*
1675                          * IPv4-mapped addresses can only be returned by a
1676                          * Node Information proxy, since they represent
1677                          * addresses of IPv4-only nodes, which perforce do
1678                          * not implement this protocol.
1679                          * [icmp-name-lookups-07, Section 5.4]
1680                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1681                          * this function at this moment.
1682                          */
1683
1684                         /* What do we have to do about ::1? */
1685                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1686                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1687                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1688                                         continue;
1689                                 break;
1690                         case IPV6_ADDR_SCOPE_SITELOCAL:
1691                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1692                                         continue;
1693                                 break;
1694                         case IPV6_ADDR_SCOPE_GLOBAL:
1695                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1696                                         continue;
1697                                 break;
1698                         default:
1699                                 continue;
1700                         }
1701
1702                         /*
1703                          * check if anycast is okay.
1704                          * XXX: just experimental.  not in the spec.
1705                          */
1706                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1707                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1708                                 continue; /* we need only unicast addresses */
1709                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1710                             (icmp6_nodeinfo & 4) == 0) {
1711                                 continue;
1712                         }
1713                         addrsofif++; /* count the address */
1714                 }
1715                 if (iffound) {
1716                         *ifpp = ifp;
1717                         return(addrsofif);
1718                 }
1719
1720                 addrs += addrsofif;
1721         }
1722
1723         return(addrs);
1724 }
1725
1726 static int
1727 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1728                 struct ifnet *ifp0, int resid)
1729 {
1730         struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1731         struct in6_ifaddr *ifa6;
1732         struct ifaddr *ifa;
1733         struct ifnet *ifp_dep = NULL;
1734         int copied = 0, allow_deprecated = 0;
1735         u_char *cp = (u_char *)(nni6 + 1);
1736         int niflags = ni6->ni_flags;
1737         u_int32_t ltime;
1738
1739         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1740                 return(0);      /* needless to copy */
1741
1742   again:
1743
1744         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1745         {
1746                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1747                         if (ifa->ifa_addr->sa_family != AF_INET6)
1748                                 continue;
1749                         ifa6 = (struct in6_ifaddr *)ifa;
1750
1751                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1752                             allow_deprecated == 0) {
1753                                 /*
1754                                  * prefererred address should be put before
1755                                  * deprecated addresses.
1756                                  */
1757
1758                                 /* record the interface for later search */
1759                                 if (ifp_dep == NULL)
1760                                         ifp_dep = ifp;
1761
1762                                 continue;
1763                         }
1764                         else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1765                                  allow_deprecated != 0)
1766                                 continue; /* we now collect deprecated addrs */
1767
1768                         /* What do we have to do about ::1? */
1769                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1770                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1771                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1772                                         continue;
1773                                 break;
1774                         case IPV6_ADDR_SCOPE_SITELOCAL:
1775                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1776                                         continue;
1777                                 break;
1778                         case IPV6_ADDR_SCOPE_GLOBAL:
1779                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1780                                         continue;
1781                                 break;
1782                         default:
1783                                 continue;
1784                         }
1785
1786                         /*
1787                          * check if anycast is okay.
1788                          * XXX: just experimental. not in the spec.
1789                          */
1790                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1791                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1792                                 continue;
1793                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1794                             (icmp6_nodeinfo & 4) == 0) {
1795                                 continue;
1796                         }
1797
1798                         /* now we can copy the address */
1799                         if (resid < sizeof(struct in6_addr) +
1800                             sizeof(u_int32_t)) {
1801                                 /*
1802                                  * We give up much more copy.
1803                                  * Set the truncate flag and return.
1804                                  */
1805                                 nni6->ni_flags |=
1806                                         NI_NODEADDR_FLAG_TRUNCATE;
1807                                 return(copied);
1808                         }
1809
1810                         /*
1811                          * Set the TTL of the address.
1812                          * The TTL value should be one of the following
1813                          * according to the specification:
1814                          *
1815                          * 1. The remaining lifetime of a DHCP lease on the
1816                          *    address, or
1817                          * 2. The remaining Valid Lifetime of a prefix from
1818                          *    which the address was derived through Stateless
1819                          *    Autoconfiguration.
1820                          *
1821                          * Note that we currently do not support stateful
1822                          * address configuration by DHCPv6, so the former
1823                          * case can't happen.
1824                          */
1825                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1826                                 ltime = ND6_INFINITE_LIFETIME;
1827                         else {
1828                                 if (ifa6->ia6_lifetime.ia6t_expire >
1829                                     time_second)
1830                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1831                                 else
1832                                         ltime = 0;
1833                         }
1834
1835                         bcopy(&ltime, cp, sizeof(u_int32_t));
1836                         cp += sizeof(u_int32_t);
1837
1838                         /* copy the address itself */
1839                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1840                               sizeof(struct in6_addr));
1841                         /* XXX: KAME link-local hack; remove ifindex */
1842                         if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1843                                 ((struct in6_addr *)cp)->s6_addr16[1] = 0;
1844                         cp += sizeof(struct in6_addr);
1845
1846                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1847                         copied += (sizeof(struct in6_addr) +
1848                                    sizeof(u_int32_t));
1849                 }
1850                 if (ifp0)       /* we need search only on the specified IF */
1851                         break;
1852         }
1853
1854         if (allow_deprecated == 0 && ifp_dep != NULL) {
1855                 ifp = ifp_dep;
1856                 allow_deprecated = 1;
1857
1858                 goto again;
1859         }
1860
1861         return(copied);
1862 }
1863
1864 /*
1865  * XXX almost dup'ed code with rip6_input.
1866  */
1867 static int
1868 icmp6_rip6_input(struct mbuf **mp, int  off)
1869 {
1870         struct mbuf *m = *mp;
1871         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1872         struct in6pcb *in6p;
1873         struct in6pcb *last = NULL;
1874         struct sockaddr_in6 rip6src;
1875         struct icmp6_hdr *icmp6;
1876         struct mbuf *opts = NULL;
1877
1878 #ifndef PULLDOWN_TEST
1879         /* this is assumed to be safe. */
1880         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1881 #else
1882         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1883         if (icmp6 == NULL) {
1884                 /* m is already reclaimed */
1885                 return IPPROTO_DONE;
1886         }
1887 #endif
1888
1889         bzero(&rip6src, sizeof(rip6src));
1890         rip6src.sin6_len = sizeof(struct sockaddr_in6);
1891         rip6src.sin6_family = AF_INET6;
1892         /* KAME hack: recover scopeid */
1893         (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1894
1895         LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1896         {
1897                 if (in6p->inp_flags & INP_PLACEMARKER)
1898                         continue;
1899                 if ((in6p->inp_vflag & INP_IPV6) == 0)
1900                         continue;
1901 #ifdef HAVE_NRL_INPCB
1902                 if (!(in6p->in6p_flags & INP_IPV6))
1903                         continue;
1904 #endif
1905                 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1906                         continue;
1907                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1908                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1909                         continue;
1910                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1911                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1912                         continue;
1913                 if (in6p->in6p_icmp6filt
1914                     && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1915                                  in6p->in6p_icmp6filt))
1916                         continue;
1917                 if (last) {
1918                         struct  mbuf *n;
1919                         if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1920                                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1921                                         ip6_savecontrol(last, &opts, ip6, n);
1922                                 /* strip intermediate headers */
1923                                 m_adj(n, off);
1924                                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1925                                                  (struct sockaddr *)&rip6src,
1926                                                  n, opts) == 0) {
1927                                         /* should notify about lost packet */
1928                                         m_freem(n);
1929                                         if (opts) {
1930                                                 m_freem(opts);
1931                                         }
1932                                 } else
1933                                         sorwakeup(last->in6p_socket);
1934                                 opts = NULL;
1935                         }
1936                 }
1937                 last = in6p;
1938         }
1939         if (last) {
1940                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1941                         ip6_savecontrol(last, &opts, ip6, m);
1942                 /* strip intermediate headers */
1943                 m_adj(m, off);
1944                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1945                                  (struct sockaddr *)&rip6src, m, opts) == 0) {
1946                         m_freem(m);
1947                         if (opts)
1948                                 m_freem(opts);
1949                 } else
1950                         sorwakeup(last->in6p_socket);
1951         } else {
1952                 m_freem(m);
1953                 ip6stat.ip6s_delivered--;
1954         }
1955         return IPPROTO_DONE;
1956 }
1957
1958 /*
1959  * Reflect the ip6 packet back to the source.
1960  * OFF points to the icmp6 header, counted from the top of the mbuf.
1961  */
1962 void
1963 icmp6_reflect(struct mbuf *m, size_t off)
1964 {
1965         struct ip6_hdr *ip6;
1966         struct icmp6_hdr *icmp6;
1967         struct in6_ifaddr *ia;
1968         struct in6_addr t, *src = 0;
1969         int plen;
1970         int type, code;
1971         struct ifnet *outif = NULL;
1972         struct sockaddr_in6 sa6_src, sa6_dst;
1973 #ifdef COMPAT_RFC1885
1974         int mtu = IPV6_MMTU;
1975         struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1976 #endif
1977
1978         /* too short to reflect */
1979         if (off < sizeof(struct ip6_hdr)) {
1980                 nd6log((LOG_DEBUG,
1981                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1982                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
1983                     __FILE__, __LINE__));
1984                 goto bad;
1985         }
1986
1987         /*
1988          * If there are extra headers between IPv6 and ICMPv6, strip
1989          * off that header first.
1990          */
1991 #ifdef DIAGNOSTIC
1992         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1993                 panic("assumption failed in icmp6_reflect");
1994 #endif
1995         if (off > sizeof(struct ip6_hdr)) {
1996                 size_t l;
1997                 struct ip6_hdr nip6;
1998
1999                 l = off - sizeof(struct ip6_hdr);
2000                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2001                 m_adj(m, l);
2002                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2003                 if (m->m_len < l) {
2004                         if ((m = m_pullup(m, l)) == NULL)
2005                                 return;
2006                 }
2007                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2008         } else /* off == sizeof(struct ip6_hdr) */ {
2009                 size_t l;
2010                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2011                 if (m->m_len < l) {
2012                         if ((m = m_pullup(m, l)) == NULL)
2013                                 return;
2014                 }
2015         }
2016         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2017         ip6 = mtod(m, struct ip6_hdr *);
2018         ip6->ip6_nxt = IPPROTO_ICMPV6;
2019         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2020         type = icmp6->icmp6_type; /* keep type for statistics */
2021         code = icmp6->icmp6_code; /* ditto. */
2022
2023         t = ip6->ip6_dst;
2024         /*
2025          * ip6_input() drops a packet if its src is multicast.
2026          * So, the src is never multicast.
2027          */
2028         ip6->ip6_dst = ip6->ip6_src;
2029
2030         /*
2031          * XXX: make sure to embed scope zone information, using
2032          * already embedded IDs or the received interface (if any).
2033          * Note that rcvif may be NULL.
2034          * TODO: scoped routing case (XXX).
2035          */
2036         bzero(&sa6_src, sizeof(sa6_src));
2037         sa6_src.sin6_family = AF_INET6;
2038         sa6_src.sin6_len = sizeof(sa6_src);
2039         sa6_src.sin6_addr = ip6->ip6_dst;
2040         in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2041         in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2042         bzero(&sa6_dst, sizeof(sa6_dst));
2043         sa6_dst.sin6_family = AF_INET6;
2044         sa6_dst.sin6_len = sizeof(sa6_dst);
2045         sa6_dst.sin6_addr = t;
2046         in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2047         in6_embedscope(&t, &sa6_dst, NULL, NULL);
2048
2049 #ifdef COMPAT_RFC1885
2050         /*
2051          * xxx guess MTU
2052          * RFC 1885 requires that echo reply should be truncated if it
2053          * does not fit in with (return) path MTU, but the description was
2054          * removed in the new spec.
2055          */
2056         if (icmp6_reflect_rt.ro_rt == 0 ||
2057             ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2058                 if (icmp6_reflect_rt.ro_rt) {
2059                         RTFREE(icmp6_reflect_rt.ro_rt);
2060                         icmp6_reflect_rt.ro_rt = 0;
2061                 }
2062                 bzero(sin6, sizeof(*sin6));
2063                 sin6->sin6_family = PF_INET6;
2064                 sin6->sin6_len = sizeof(struct sockaddr_in6);
2065                 sin6->sin6_addr = ip6->ip6_dst;
2066
2067                 rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2068                             RTF_PRCLONING);
2069         }
2070
2071         if (icmp6_reflect_rt.ro_rt == 0)
2072                 goto bad;
2073
2074         if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2075             && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2076                 mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2077
2078         if (mtu < m->m_pkthdr.len) {
2079                 plen -= (m->m_pkthdr.len - mtu);
2080                 m_adj(m, mtu - m->m_pkthdr.len);
2081         }
2082 #endif
2083         /*
2084          * If the incoming packet was addressed directly to us(i.e. unicast),
2085          * use dst as the src for the reply.
2086          * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2087          * (for example) when we encounter an error while forwarding procedure
2088          * destined to a duplicated address of ours.
2089          */
2090         for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2091                 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2092                     (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2093                         src = &t;
2094                         break;
2095                 }
2096         if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2097                 /*
2098                  * This is the case if the dst is our link-local address
2099                  * and the sender is also ourselves.
2100                  */
2101                 src = &t;
2102         }
2103
2104         if (src == 0) {
2105                 int e;
2106                 struct route_in6 ro;
2107
2108                 /*
2109                  * This case matches to multicasts, our anycast, or unicasts
2110                  * that we do not own.  Select a source address based on the
2111                  * source address of the erroneous packet.
2112                  */
2113                 bzero(&ro, sizeof(ro));
2114                 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2115                 if (ro.ro_rt)
2116                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2117                 if (src == NULL) {
2118                         nd6log((LOG_DEBUG,
2119                             "icmp6_reflect: source can't be determined: "
2120                             "dst=%s, error=%d\n",
2121                             ip6_sprintf(&sa6_src.sin6_addr), e));
2122                         goto bad;
2123                 }
2124         }
2125
2126         ip6->ip6_src = *src;
2127
2128         ip6->ip6_flow = 0;
2129         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2130         ip6->ip6_vfc |= IPV6_VERSION;
2131         ip6->ip6_nxt = IPPROTO_ICMPV6;
2132         if (m->m_pkthdr.rcvif) {
2133                 /* XXX: This may not be the outgoing interface */
2134                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2135         } else
2136                 ip6->ip6_hlim = ip6_defhlim;
2137
2138         icmp6->icmp6_cksum = 0;
2139         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2140                                         sizeof(struct ip6_hdr), plen);
2141
2142         /*
2143          * XXX option handling
2144          */
2145
2146         m->m_flags &= ~(M_BCAST|M_MCAST);
2147
2148 #ifdef COMPAT_RFC1885
2149         ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2150 #else
2151         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2152 #endif
2153         if (outif)
2154                 icmp6_ifoutstat_inc(outif, type, code);
2155
2156         return;
2157
2158  bad:
2159         m_freem(m);
2160         return;
2161 }
2162
2163 void
2164 icmp6_fasttimo(void)
2165 {
2166
2167         mld6_fasttimeo();
2168 }
2169
2170 static const char *
2171 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2172                     struct in6_addr *tgt6)
2173 {
2174         static char buf[1024];
2175         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2176                 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2177         return buf;
2178 }
2179
2180 void
2181 icmp6_redirect_input(struct mbuf *m, int off)
2182 {
2183         struct ifnet *ifp = m->m_pkthdr.rcvif;
2184         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2185         struct nd_redirect *nd_rd;
2186         int icmp6len = ntohs(ip6->ip6_plen);
2187         char *lladdr = NULL;
2188         int lladdrlen = 0;
2189         u_char *redirhdr = NULL;
2190         int redirhdrlen = 0;
2191         struct rtentry *rt = NULL;
2192         int is_router;
2193         int is_onlink;
2194         struct in6_addr src6 = ip6->ip6_src;
2195         struct in6_addr redtgt6;
2196         struct in6_addr reddst6;
2197         union nd_opts ndopts;
2198
2199         if (!m || !ifp)
2200                 return;
2201
2202         /* XXX if we are router, we don't update route by icmp6 redirect */
2203         if (ip6_forwarding)
2204                 goto freeit;
2205         if (!icmp6_rediraccept)
2206                 goto freeit;
2207
2208 #ifndef PULLDOWN_TEST
2209         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2210         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2211 #else
2212         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2213         if (nd_rd == NULL) {
2214                 icmp6stat.icp6s_tooshort++;
2215                 return;
2216         }
2217 #endif
2218         redtgt6 = nd_rd->nd_rd_target;
2219         reddst6 = nd_rd->nd_rd_dst;
2220
2221         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2222                 redtgt6.s6_addr16[1] = htons(ifp->if_index);
2223         if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2224                 reddst6.s6_addr16[1] = htons(ifp->if_index);
2225
2226         /* validation */
2227         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2228                 nd6log((LOG_ERR,
2229                         "ICMP6 redirect sent from %s rejected; "
2230                         "must be from linklocal\n", ip6_sprintf(&src6)));
2231                 goto bad;
2232         }
2233         if (ip6->ip6_hlim != 255) {
2234                 nd6log((LOG_ERR,
2235                         "ICMP6 redirect sent from %s rejected; "
2236                         "hlim=%d (must be 255)\n",
2237                         ip6_sprintf(&src6), ip6->ip6_hlim));
2238                 goto bad;
2239         }
2240     {
2241         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2242         struct sockaddr_in6 sin6;
2243         struct in6_addr *gw6;
2244
2245         bzero(&sin6, sizeof(sin6));
2246         sin6.sin6_family = AF_INET6;
2247         sin6.sin6_len = sizeof(struct sockaddr_in6);
2248         bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2249         rt = rtpurelookup((struct sockaddr *)&sin6);
2250         if (rt != NULL) {
2251                 if (rt->rt_gateway == NULL ||
2252                     rt->rt_gateway->sa_family != AF_INET6) {
2253                         nd6log((LOG_ERR,
2254                             "ICMP6 redirect rejected; no route "
2255                             "with inet6 gateway found for redirect dst: %s\n",
2256                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2257                         --rt->rt_refcnt;
2258                         goto bad;
2259                 }
2260
2261                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2262                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2263                         nd6log((LOG_ERR,
2264                                 "ICMP6 redirect rejected; "
2265                                 "not equal to gw-for-src=%s (must be same): "
2266                                 "%s\n",
2267                                 ip6_sprintf(gw6),
2268                                 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2269                         --rt->rt_refcnt;
2270                         goto bad;
2271                 }
2272         } else {
2273                 nd6log((LOG_ERR,
2274                         "ICMP6 redirect rejected; "
2275                         "no route found for redirect dst: %s\n",
2276                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2277                 goto bad;
2278         }
2279         --rt->rt_refcnt;
2280         rt = NULL;
2281     }
2282         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2283                 nd6log((LOG_ERR,
2284                         "ICMP6 redirect rejected; "
2285                         "redirect dst must be unicast: %s\n",
2286                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2287                 goto bad;
2288         }
2289
2290         is_router = is_onlink = 0;
2291         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2292                 is_router = 1;  /* router case */
2293         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2294                 is_onlink = 1;  /* on-link destination case */
2295         if (!is_router && !is_onlink) {
2296                 nd6log((LOG_ERR,
2297                         "ICMP6 redirect rejected; "
2298                         "neither router case nor onlink case: %s\n",
2299                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2300                 goto bad;
2301         }
2302         /* validation passed */
2303
2304         icmp6len -= sizeof(*nd_rd);
2305         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2306         if (nd6_options(&ndopts) < 0) {
2307                 nd6log((LOG_INFO, "icmp6_redirect_input: "
2308                         "invalid ND option, rejected: %s\n",
2309                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2310                 /* nd6_options have incremented stats */
2311                 goto freeit;
2312         }
2313
2314         if (ndopts.nd_opts_tgt_lladdr) {
2315                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2316                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2317         }
2318
2319         if (ndopts.nd_opts_rh) {
2320                 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2321                 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2322         }
2323
2324         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2325                 nd6log((LOG_INFO,
2326                         "icmp6_redirect_input: lladdrlen mismatch for %s "
2327                         "(if %d, icmp6 packet %d): %s\n",
2328                         ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2329                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2330                 goto bad;
2331         }
2332
2333         /* RFC 2461 8.3 */
2334         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2335                          is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2336
2337         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2338                 /* perform rtredirect */
2339                 struct sockaddr_in6 sdst;
2340                 struct sockaddr_in6 sgw;
2341                 struct sockaddr_in6 ssrc;
2342
2343                 bzero(&sdst, sizeof(sdst));
2344                 bzero(&sgw, sizeof(sgw));
2345                 bzero(&ssrc, sizeof(ssrc));
2346                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2347                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2348                         sizeof(struct sockaddr_in6);
2349                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2350                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2351                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2352                 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2353                            (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2354                            (struct sockaddr *)&ssrc,
2355                            (struct rtentry **)NULL);
2356         }
2357         /* finally update cached route in each socket via pfctlinput */
2358     {
2359         struct sockaddr_in6 sdst;
2360
2361         bzero(&sdst, sizeof(sdst));
2362         sdst.sin6_family = AF_INET6;
2363         sdst.sin6_len = sizeof(struct sockaddr_in6);
2364         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2365         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2366 #ifdef IPSEC
2367         key_sa_routechange((struct sockaddr *)&sdst);
2368 #endif
2369     }
2370
2371  freeit:
2372         m_freem(m);
2373         return;
2374
2375  bad:
2376         icmp6stat.icp6s_badredirect++;
2377         m_freem(m);
2378 }
2379
2380 void
2381 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2382 {
2383         struct ifnet *ifp;      /* my outgoing interface */
2384         struct in6_addr *ifp_ll6;
2385         struct in6_addr *router_ll6;
2386         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2387         struct mbuf *m = NULL;  /* newly allocated one */
2388         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2389         struct nd_redirect *nd_rd;
2390         size_t maxlen;
2391         u_char *p;
2392         struct ifnet *outif = NULL;
2393         struct sockaddr_in6 src_sa;
2394
2395         icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2396
2397         /* if we are not router, we don't send icmp6 redirect */
2398         if (!ip6_forwarding || ip6_accept_rtadv)
2399                 goto fail;
2400
2401         /* sanity check */
2402         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2403                 goto fail;
2404
2405         /*
2406          * Address check:
2407          *  the source address must identify a neighbor, and
2408          *  the destination address must not be a multicast address
2409          *  [RFC 2461, sec 8.2]
2410          */
2411         sip6 = mtod(m0, struct ip6_hdr *);
2412         bzero(&src_sa, sizeof(src_sa));
2413         src_sa.sin6_family = AF_INET6;
2414         src_sa.sin6_len = sizeof(src_sa);
2415         src_sa.sin6_addr = sip6->ip6_src;
2416         /* we don't currently use sin6_scope_id, but eventually use it */
2417         src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2418         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2419                 goto fail;
2420         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2421                 goto fail;      /* what should we do here? */
2422
2423         /* rate limit */
2424         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2425                 goto fail;
2426
2427         /*
2428          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2429          * we almost always ask for an mbuf cluster for simplicity.
2430          * (MHLEN < IPV6_MMTU is almost always true)
2431          */
2432 #if IPV6_MMTU >= MCLBYTES
2433 # error assumption failed about IPV6_MMTU and MCLBYTES
2434 #endif
2435         MGETHDR(m, MB_DONTWAIT, MT_HEADER);
2436         if (m && IPV6_MMTU >= MHLEN)
2437                 MCLGET(m, MB_DONTWAIT);
2438         if (!m)
2439                 goto fail;
2440         m->m_pkthdr.rcvif = NULL;
2441         m->m_len = 0;
2442         maxlen = M_TRAILINGSPACE(m);
2443         maxlen = min(IPV6_MMTU, maxlen);
2444         /* just for safety */
2445         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2446             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2447                 goto fail;
2448         }
2449
2450         {
2451                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2452                 struct in6_ifaddr *ia;
2453                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2454                                                  IN6_IFF_NOTREADY|
2455                                                  IN6_IFF_ANYCAST)) == NULL)
2456                         goto fail;
2457                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2458         }
2459
2460         /* get ip6 linklocal address for the router. */
2461         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2462                 struct sockaddr_in6 *sin6;
2463                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2464                 router_ll6 = &sin6->sin6_addr;
2465                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2466                         router_ll6 = (struct in6_addr *)NULL;
2467         } else
2468                 router_ll6 = (struct in6_addr *)NULL;
2469
2470         /* ip6 */
2471         ip6 = mtod(m, struct ip6_hdr *);
2472         ip6->ip6_flow = 0;
2473         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2474         ip6->ip6_vfc |= IPV6_VERSION;
2475         /* ip6->ip6_plen will be set later */
2476         ip6->ip6_nxt = IPPROTO_ICMPV6;
2477         ip6->ip6_hlim = 255;
2478         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2479         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2480         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2481
2482         /* ND Redirect */
2483         nd_rd = (struct nd_redirect *)(ip6 + 1);
2484         nd_rd->nd_rd_type = ND_REDIRECT;
2485         nd_rd->nd_rd_code = 0;
2486         nd_rd->nd_rd_reserved = 0;
2487         if (rt->rt_flags & RTF_GATEWAY) {
2488                 /*
2489                  * nd_rd->nd_rd_target must be a link-local address in
2490                  * better router cases.
2491                  */
2492                 if (!router_ll6)
2493                         goto fail;
2494                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2495                       sizeof(nd_rd->nd_rd_target));
2496                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2497                       sizeof(nd_rd->nd_rd_dst));
2498         } else {
2499                 /* make sure redtgt == reddst */
2500                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2501                       sizeof(nd_rd->nd_rd_target));
2502                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2503                       sizeof(nd_rd->nd_rd_dst));
2504         }
2505
2506         p = (u_char *)(nd_rd + 1);
2507
2508         if (!router_ll6)
2509                 goto nolladdropt;
2510
2511     {
2512         /* target lladdr option */
2513         struct rtentry *rt_router = NULL;
2514         int len;
2515         struct sockaddr_dl *sdl;
2516         struct nd_opt_hdr *nd_opt;
2517         char *lladdr;
2518
2519         rt_router = nd6_lookup(router_ll6, 0, ifp);
2520         if (!rt_router)
2521                 goto nolladdropt;
2522         len = sizeof(*nd_opt) + ifp->if_addrlen;
2523         len = (len + 7) & ~7;   /* round by 8 */
2524         /* safety check */
2525         if (len + (p - (u_char *)ip6) > maxlen)
2526                 goto nolladdropt;
2527         if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2528             (rt_router->rt_flags & RTF_LLINFO) &&
2529             (rt_router->rt_gateway->sa_family == AF_LINK) &&
2530             (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2531             sdl->sdl_alen) {
2532                 nd_opt = (struct nd_opt_hdr *)p;
2533                 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2534                 nd_opt->nd_opt_len = len >> 3;
2535                 lladdr = (char *)(nd_opt + 1);
2536                 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2537                 p += len;
2538         }
2539     }
2540 nolladdropt:;
2541
2542         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2543
2544         /* just to be safe */
2545 #ifdef M_DECRYPTED      /*not openbsd*/
2546         if (m0->m_flags & M_DECRYPTED)
2547                 goto noredhdropt;
2548 #endif
2549         if (p - (u_char *)ip6 > maxlen)
2550                 goto noredhdropt;
2551
2552     {
2553         /* redirected header option */
2554         int len;
2555         struct nd_opt_rd_hdr *nd_opt_rh;
2556
2557         /*
2558          * compute the maximum size for icmp6 redirect header option.
2559          * XXX room for auth header?
2560          */
2561         len = maxlen - (p - (u_char *)ip6);
2562         len &= ~7;
2563
2564         /* This is just for simplicity. */
2565         if (m0->m_pkthdr.len != m0->m_len) {
2566                 if (m0->m_next) {
2567                         m_freem(m0->m_next);
2568                         m0->m_next = NULL;
2569                 }
2570                 m0->m_pkthdr.len = m0->m_len;
2571         }
2572
2573         /*
2574          * Redirected header option spec (RFC2461 4.6.3) talks nothing
2575          * about padding/truncate rule for the original IP packet.
2576          * From the discussion on IPv6imp in Feb 1999, the consensus was:
2577          * - "attach as much as possible" is the goal
2578          * - pad if not aligned (original size can be guessed by original
2579          *   ip6 header)
2580          * Following code adds the padding if it is simple enough,
2581          * and truncates if not.
2582          */
2583         if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2584                 panic("assumption failed in %s:%d", __FILE__, __LINE__);
2585
2586         if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2587                 /* not enough room, truncate */
2588                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2589         } else {
2590                 /* enough room, pad or truncate */
2591                 size_t extra;
2592
2593                 extra = m0->m_pkthdr.len % 8;
2594                 if (extra) {
2595                         /* pad if easy enough, truncate if not */
2596                         if (8 - extra <= M_TRAILINGSPACE(m0)) {
2597                                 /* pad */
2598                                 m0->m_len += (8 - extra);
2599                                 m0->m_pkthdr.len += (8 - extra);
2600                         } else {
2601                                 /* truncate */
2602                                 m0->m_pkthdr.len -= extra;
2603                                 m0->m_len -= extra;
2604                         }
2605                 }
2606                 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2607                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2608         }
2609
2610         nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2611         bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2612         nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2613         nd_opt_rh->nd_opt_rh_len = len >> 3;
2614         p += sizeof(*nd_opt_rh);
2615         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2616
2617         /* connect m0 to m */
2618         m->m_next = m0;
2619         m->m_pkthdr.len = m->m_len + m0->m_len;
2620     }
2621 noredhdropt:;
2622
2623         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2624                 sip6->ip6_src.s6_addr16[1] = 0;
2625         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2626                 sip6->ip6_dst.s6_addr16[1] = 0;
2627 #if 0
2628         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2629                 ip6->ip6_src.s6_addr16[1] = 0;
2630         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2631                 ip6->ip6_dst.s6_addr16[1] = 0;
2632 #endif
2633         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2634                 nd_rd->nd_rd_target.s6_addr16[1] = 0;
2635         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2636                 nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2637
2638         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2639
2640         nd_rd->nd_rd_cksum = 0;
2641         nd_rd->nd_rd_cksum
2642                 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2643
2644         /* send the packet to outside... */
2645         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2646         if (outif) {
2647                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2648                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2649         }
2650         icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2651
2652         return;
2653
2654 fail:
2655         if (m)
2656                 m_freem(m);
2657         if (m0)
2658                 m_freem(m0);
2659 }
2660
2661 #ifdef HAVE_NRL_INPCB
2662 #define sotoin6pcb      sotoinpcb
2663 #define in6pcb          inpcb
2664 #define in6p_icmp6filt  inp_icmp6filt
2665 #endif
2666 /*
2667  * ICMPv6 socket option processing.
2668  */
2669 int
2670 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2671 {
2672         int error = 0;
2673         int optlen;
2674         struct inpcb *inp = sotoinpcb(so);
2675         int level, op, optname;
2676
2677         if (sopt) {
2678                 level = sopt->sopt_level;
2679                 op = sopt->sopt_dir;
2680                 optname = sopt->sopt_name;
2681                 optlen = sopt->sopt_valsize;
2682         } else
2683                 level = op = optname = optlen = 0;
2684
2685         if (level != IPPROTO_ICMPV6) {
2686                 return EINVAL;
2687         }
2688
2689         switch (op) {
2690         case PRCO_SETOPT:
2691                 switch (optname) {
2692                 case ICMP6_FILTER:
2693                     {
2694                         struct icmp6_filter *p;
2695
2696                         if (optlen != sizeof(*p)) {
2697                                 error = EMSGSIZE;
2698                                 break;
2699                         }
2700                         if (inp->in6p_icmp6filt == NULL) {
2701                                 error = EINVAL;
2702                                 break;
2703                         }
2704                         error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2705                                 optlen);
2706                         break;
2707                     }
2708
2709                 default:
2710                         error = ENOPROTOOPT;
2711                         break;
2712                 }
2713                 break;
2714
2715         case PRCO_GETOPT:
2716                 switch (optname) {
2717                 case ICMP6_FILTER:
2718                     {
2719                         if (inp->in6p_icmp6filt == NULL) {
2720                                 error = EINVAL;
2721                                 break;
2722                         }
2723                         error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2724                                 sizeof(struct icmp6_filter));
2725                         break;
2726                     }
2727
2728                 default:
2729                         error = ENOPROTOOPT;
2730                         break;
2731                 }
2732                 break;
2733         }
2734
2735         return(error);
2736 }
2737 #ifdef HAVE_NRL_INPCB
2738 #undef sotoin6pcb
2739 #undef in6pcb
2740 #undef in6p_icmp6filt
2741 #endif
2742
2743 #ifndef HAVE_PPSRATECHECK
2744 #ifndef timersub
2745 #define timersub(tvp, uvp, vvp)                                         \
2746         do {                                                            \
2747                 (vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;          \
2748                 (vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;       \
2749                 if ((vvp)->tv_usec < 0) {                               \
2750                         (vvp)->tv_sec--;                                \
2751                         (vvp)->tv_usec += 1000000;                      \
2752                 }                                                       \
2753         } while (0)
2754 #endif
2755
2756 /*
2757  * ppsratecheck(): packets (or events) per second limitation.
2758  */
2759 static int
2760 ppsratecheck(struct timeval *lasttime, int *curpps,
2761              int maxpps)        /* maximum pps allowed */
2762 {
2763         struct timeval tv, delta;
2764         int rv;
2765
2766         microtime(&tv);
2767
2768         timersub(&tv, lasttime, &delta);
2769
2770         /*
2771          * Check for 0,0 so that the message will be seen at least once.
2772          * If more than one second has passed since the last update of
2773          * lasttime, reset the counter.
2774          *
2775          * We do increment *curpps even in *curpps < maxpps case, as some may
2776          * try to use *curpps for stat purposes as well.
2777          */
2778         if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2779             delta.tv_sec >= 1) {
2780                 *lasttime = tv;
2781                 *curpps = 0;
2782                 rv = 1;
2783         } else if (maxpps < 0)
2784                 rv = 1;
2785         else if (*curpps < maxpps)
2786                 rv = 1;
2787         else
2788                 rv = 0;
2789
2790 #if 1 /* DIAGNOSTIC? */
2791         /* be careful about wrap-around */
2792         if (*curpps + 1 > *curpps)
2793                 *curpps = *curpps + 1;
2794 #else
2795         /*
2796          * assume that there's not too many calls to this function.
2797          * not sure if the assumption holds, as it depends on *caller's*
2798          * behavior, not the behavior of this function.
2799          * IMHO it is wrong to make assumption on the caller's behavior,
2800          * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2801          */
2802         *curpps = *curpps + 1;
2803 #endif
2804
2805         return (rv);
2806 }
2807 #endif
2808
2809 /*
2810  * Perform rate limit check.
2811  * Returns 0 if it is okay to send the icmp6 packet.
2812  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2813  * limitation.
2814  *
2815  * XXX per-destination/type check necessary?
2816  */
2817 static int
2818 icmp6_ratelimit(const struct in6_addr *dst,     /* not used at this moment */
2819                 const int type,                 /* not used at this moment */
2820                 const int code)                 /* not used at this moment */
2821 {
2822         int ret;
2823
2824         ret = 0;        /* okay to send */
2825
2826         /* PPS limit */
2827         if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2828             icmp6errppslim)) {
2829                 /* The packet is subject to rate limit */
2830                 ret++;
2831         }
2832
2833         return ret;
2834 }