sys/netinet6/icmp6.c

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