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 1.21 2006/01/14 11:44:24 swildner Exp $  */
   3 /*      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $ */
   4
   5 /*
   6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   7  * All rights reserved.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. Neither the name of the project nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  */
  33
  34 /*
  35  * Copyright (c) 1982, 1986, 1988, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. All advertising materials mentioning features or use of this software
  47  *    must display the following acknowledgement:
  48  *      This product includes software developed by the University of
  49  *      California, Berkeley and its contributors.
  50  * 4. Neither the name of the University nor the names of its contributors
  51  *    may be used to endorse or promote products derived from this software
  52  *    without specific prior written permission.
  53  *
  54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  64  * SUCH DAMAGE.
  65  *
  66  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
  67  */
  68
  69 #include "opt_inet.h"
  70 #include "opt_inet6.h"
  71 #include "opt_ipsec.h"
  72
  73 #include <sys/param.h>
  74 #include <sys/systm.h>
  75 #include <sys/malloc.h>
  76 #include <sys/mbuf.h>
  77 #include <sys/protosw.h>
  78 #include <sys/socket.h>
  79 #include <sys/socketvar.h>
  80 #include <sys/time.h>
  81 #include <sys/kernel.h>
  82 #include <sys/syslog.h>
  83 #include <sys/domain.h>
  84
  85 #include <net/if.h>
  86 #include <net/route.h>
  87 #include <net/if_dl.h>
  88 #include <net/if_types.h>
  89
  90 #include <netinet/in.h>
  91 #include <netinet/in_var.h>
  92 #include <netinet/ip6.h>
  93 #include <netinet6/ip6_var.h>
  94 #include <netinet/icmp6.h>
  95 #include <netinet6/mld6_var.h>
  96 #include <netinet/in_pcb.h>
  97 #include <netinet6/in6_pcb.h>
  98 #include <netinet6/nd6.h>
  99 #include <netinet6/in6_ifattach.h>
 100 #include <netinet6/ip6protosw.h>
 101
 102 #ifdef IPSEC
 103 #include <netinet6/ipsec.h>
 104 #include <netproto/key/key.h>
 105 #endif
 106
 107 #ifdef FAST_IPSEC
 108 #include <netproto/ipsec/ipsec.h>
 109 #include <netproto/ipsec/key.h>
 110 #define IPSEC
 111 #endif
 112
 113 #include <net/net_osdep.h>
 114
 115 #ifdef HAVE_NRL_INPCB
 116 /* inpcb members */
 117 #define in6pcb          inpcb
 118 #define in6p_laddr      inp_laddr6
 119 #define in6p_faddr      inp_faddr6
 120 #define in6p_icmp6filt  inp_icmp6filt
 121 #define in6p_route      inp_route
 122 #define in6p_socket     inp_socket
 123 #define in6p_flags      inp_flags
 124 #define in6p_moptions   inp_moptions6
 125 #define in6p_outputopts inp_outputopts6
 126 #define in6p_ip6        inp_ipv6
 127 #define in6p_flowinfo   inp_flowinfo
 128 #define in6p_sp         inp_sp
 129 #define in6p_next       inp_next
 130 #define in6p_prev       inp_prev
 131 /* function names */
 132 #define in6_pcbdetach   in_pcbdetach
 133 #define in6_rtchange    in_rtchange
 134
 135 /*
 136  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
 137  * others...
 138  */
 139 #define in6p_ip6_nxt    inp_ipv6.ip6_nxt
 140 #endif
 141
 142 /*
 143  * ppratecheck() is available, so define it here.
 144  */
 145 #define HAVE_PPSRATECHECK
 146
 147 extern struct domain inet6domain;
 148 extern struct ip6protosw inet6sw[];
 149 extern u_char ip6_protox[];
 150
 151 struct icmp6stat icmp6stat;
 152
 153 extern struct inpcbinfo ripcbinfo;
 154 extern int icmp6errppslim;
 155 static int icmp6errpps_count = 0;
 156 static struct timeval icmp6errppslim_last;
 157 extern int icmp6_nodeinfo;
 158
 159 static void icmp6_errcount (struct icmp6errstat *, int, int);
 160 static int icmp6_rip6_input (struct mbuf **, int);
 161 static int icmp6_ratelimit (const struct in6_addr *, const int, const int);
 162 static const char *icmp6_redirect_diag (struct in6_addr *,
 163         struct in6_addr *, struct in6_addr *);
 164 #ifndef HAVE_PPSRATECHECK
 165 static int ppsratecheck (struct timeval *, int *, int);
 166 #endif
 167 static struct mbuf *ni6_input (struct mbuf *, int);
 168 static struct mbuf *ni6_nametodns (const char *, int, int);
 169 static int ni6_dnsmatch (const char *, int, const char *, int);
 170 static int ni6_addrs (struct icmp6_nodeinfo *, struct mbuf *,
 171                           struct ifnet **, char *);
 172 static int ni6_store_addrs (struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
 173                                 struct ifnet *, int);
 174 static int icmp6_notify_error (struct mbuf *, int, int, int);
 175
 176 #ifdef COMPAT_RFC1885
 177 static struct route_in6 icmp6_reflect_rt;
 178 #endif
 179
 180
 181 void
 182 icmp6_init(void)
 183 {
 184         mld6_init();
 185 }
 186
 187 static void
 188 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
 189 {
 190         switch (type) {
 191         case ICMP6_DST_UNREACH:
 192                 switch (code) {
 193                 case ICMP6_DST_UNREACH_NOROUTE:
 194                         stat->icp6errs_dst_unreach_noroute++;
 195                         return;
 196                 case ICMP6_DST_UNREACH_ADMIN:
 197                         stat->icp6errs_dst_unreach_admin++;
 198                         return;
 199                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 200                         stat->icp6errs_dst_unreach_beyondscope++;
 201                         return;
 202                 case ICMP6_DST_UNREACH_ADDR:
 203                         stat->icp6errs_dst_unreach_addr++;
 204                         return;
 205                 case ICMP6_DST_UNREACH_NOPORT:
 206                         stat->icp6errs_dst_unreach_noport++;
 207                         return;
 208                 }
 209                 break;
 210         case ICMP6_PACKET_TOO_BIG:
 211                 stat->icp6errs_packet_too_big++;
 212                 return;
 213         case ICMP6_TIME_EXCEEDED:
 214                 switch (code) {
 215                 case ICMP6_TIME_EXCEED_TRANSIT:
 216                         stat->icp6errs_time_exceed_transit++;
 217                         return;
 218                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 219                         stat->icp6errs_time_exceed_reassembly++;
 220                         return;
 221                 }
 222                 break;
 223         case ICMP6_PARAM_PROB:
 224                 switch (code) {
 225                 case ICMP6_PARAMPROB_HEADER:
 226                         stat->icp6errs_paramprob_header++;
 227                         return;
 228                 case ICMP6_PARAMPROB_NEXTHEADER:
 229                         stat->icp6errs_paramprob_nextheader++;
 230                         return;
 231                 case ICMP6_PARAMPROB_OPTION:
 232                         stat->icp6errs_paramprob_option++;
 233                         return;
 234                 }
 235                 break;
 236         case ND_REDIRECT:
 237                 stat->icp6errs_redirect++;
 238                 return;
 239         }
 240         stat->icp6errs_unknown++;
 241 }
 242
 243 /*
 244  * Generate an error packet of type error in response to bad IP6 packet.
 245  */
 246 void
 247 icmp6_error(struct mbuf *m, int type, int code, int param)
 248 {
 249         struct ip6_hdr *oip6, *nip6;
 250         struct icmp6_hdr *icmp6;
 251         u_int preplen;
 252         int off;
 253         int nxt;
 254
 255         icmp6stat.icp6s_error++;
 256
 257         /* count per-type-code statistics */
 258         icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
 259
 260 #ifdef M_DECRYPTED      /*not openbsd*/
 261         if (m->m_flags & M_DECRYPTED) {
 262                 icmp6stat.icp6s_canterror++;
 263                 goto freeit;
 264         }
 265 #endif
 266
 267 #ifndef PULLDOWN_TEST
 268         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 269 #else
 270         if (m->m_len < sizeof(struct ip6_hdr)) {
 271                 m = m_pullup(m, sizeof(struct ip6_hdr));
 272                 if (m == NULL)
 273                         return;
 274         }
 275 #endif
 276         oip6 = mtod(m, struct ip6_hdr *);
 277
 278         /*
 279          * Multicast destination check. For unrecognized option errors,
 280          * this check has already done in ip6_unknown_opt(), so we can
 281          * check only for other errors.
 282          */
 283         if ((m->m_flags & (M_BCAST|M_MCAST) ||
 284              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
 285             (type != ICMP6_PACKET_TOO_BIG &&
 286              (type != ICMP6_PARAM_PROB ||
 287               code != ICMP6_PARAMPROB_OPTION)))
 288                 goto freeit;
 289
 290         /* Source address check. XXX: the case of anycast source? */
 291         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
 292             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
 293                 goto freeit;
 294
 295         /*
 296          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
 297          * don't do it.
 298          */
 299         nxt = -1;
 300         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
 301         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
 302                 struct icmp6_hdr *icp;
 303
 304 #ifndef PULLDOWN_TEST
 305                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
 306                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 307 #else
 308                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
 309                         sizeof(*icp));
 310                 if (icp == NULL) {
 311                         icmp6stat.icp6s_tooshort++;
 312                         return;
 313                 }
 314 #endif
 315                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
 316                     icp->icmp6_type == ND_REDIRECT) {
 317                         /*
 318                          * ICMPv6 error
 319                          * Special case: for redirect (which is
 320                          * informational) we must not send icmp6 error.
 321                          */
 322                         icmp6stat.icp6s_canterror++;
 323                         goto freeit;
 324                 } else {
 325                         /* ICMPv6 informational - send the error */
 326                 }
 327         } else {
 328                 /* non-ICMPv6 - send the error */
 329         }
 330
 331         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
 332
 333         /* Finally, do rate limitation check. */
 334         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
 335                 icmp6stat.icp6s_toofreq++;
 336                 goto freeit;
 337         }
 338
 339         /*
 340          * OK, ICMP6 can be generated.
 341          */
 342
 343         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
 344                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
 345
 346         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
 347         M_PREPEND(m, preplen, MB_DONTWAIT);
 348         if (m && m->m_len < preplen)
 349                 m = m_pullup(m, preplen);
 350         if (m == NULL) {
 351                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
 352                 return;
 353         }
 354
 355         nip6 = mtod(m, struct ip6_hdr *);
 356         nip6->ip6_src  = oip6->ip6_src;
 357         nip6->ip6_dst  = oip6->ip6_dst;
 358
 359         if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
 360                 oip6->ip6_src.s6_addr16[1] = 0;
 361         if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
 362                 oip6->ip6_dst.s6_addr16[1] = 0;
 363
 364         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
 365         icmp6->icmp6_type = type;
 366         icmp6->icmp6_code = code;
 367         icmp6->icmp6_pptr = htonl((u_int32_t)param);
 368
 369         /*
 370          * icmp6_reflect() is designed to be in the input path.
 371          * icmp6_error() can be called from both input and outut path,
 372          * and if we are in output path rcvif could contain bogus value.
 373          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
 374          * information in ip header (nip6).
 375          */
 376         m->m_pkthdr.rcvif = NULL;
 377
 378         icmp6stat.icp6s_outhist[type]++;
 379         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
 380
 381         return;
 382
 383   freeit:
 384         /*
 385          * If we can't tell wheter or not we can generate ICMP6, free it.
 386          */
 387         m_freem(m);
 388 }
 389
 390 /*
 391  * Process a received ICMP6 message.
 392  */
 393 int
 394 icmp6_input(struct mbuf **mp, int *offp, int proto)
 395 {
 396         struct mbuf *m = *mp, *n;
 397         struct ip6_hdr *ip6, *nip6;
 398         struct icmp6_hdr *icmp6, *nicmp6;
 399         int off = *offp;
 400         int icmp6len = m->m_pkthdr.len - *offp;
 401         int code, sum, noff;
 402
 403 #ifndef PULLDOWN_TEST
 404         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
 405         /* m might change if M_LOOP.  So, call mtod after this */
 406 #endif
 407
 408         /*
 409          * Locate icmp6 structure in mbuf, and check
 410          * that not corrupted and of at least minimum length
 411          */
 412
 413         ip6 = mtod(m, struct ip6_hdr *);
 414         if (icmp6len < sizeof(struct icmp6_hdr)) {
 415                 icmp6stat.icp6s_tooshort++;
 416                 goto freeit;
 417         }
 418
 419         /*
 420          * calculate the checksum
 421          */
 422 #ifndef PULLDOWN_TEST
 423         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
 424 #else
 425         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
 426         if (icmp6 == NULL) {
 427                 icmp6stat.icp6s_tooshort++;
 428                 return IPPROTO_DONE;
 429         }
 430 #endif
 431         code = icmp6->icmp6_code;
 432
 433         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
 434                 nd6log((LOG_ERR,
 435                     "ICMP6 checksum error(%d|%x) %s\n",
 436                     icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
 437                 icmp6stat.icp6s_checksum++;
 438                 goto freeit;
 439         }
 440
 441         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
 442                 /*
 443                  * Deliver very specific ICMP6 type only.
 444                  * This is important to deilver TOOBIG.  Otherwise PMTUD
 445                  * will not work.
 446                  */
 447                 switch (icmp6->icmp6_type) {
 448                 case ICMP6_DST_UNREACH:
 449                 case ICMP6_PACKET_TOO_BIG:
 450                 case ICMP6_TIME_EXCEEDED:
 451                         break;
 452                 default:
 453                         goto freeit;
 454                 }
 455         }
 456
 457         icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
 458         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
 459         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
 460                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
 461
 462         switch (icmp6->icmp6_type) {
 463         case ICMP6_DST_UNREACH:
 464                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
 465                 switch (code) {
 466                 case ICMP6_DST_UNREACH_NOROUTE:
 467                         code = PRC_UNREACH_NET;
 468                         break;
 469                 case ICMP6_DST_UNREACH_ADMIN:
 470                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
 471                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
 472                         break;
 473                 case ICMP6_DST_UNREACH_ADDR:
 474                         code = PRC_HOSTDEAD;
 475                         break;
 476 #ifdef COMPAT_RFC1885
 477                 case ICMP6_DST_UNREACH_NOTNEIGHBOR:
 478                         code = PRC_UNREACH_SRCFAIL;
 479                         break;
 480 #else
 481                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 482                         /* I mean "source address was incorrect." */
 483                         code = PRC_PARAMPROB;
 484                         break;
 485 #endif
 486                 case ICMP6_DST_UNREACH_NOPORT:
 487                         code = PRC_UNREACH_PORT;
 488                         break;
 489                 default:
 490                         goto badcode;
 491                 }
 492                 goto deliver;
 493                 break;
 494
 495         case ICMP6_PACKET_TOO_BIG:
 496                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
 497                 if (code != 0)
 498                         goto badcode;
 499
 500                 code = PRC_MSGSIZE;
 501
 502                 /*
 503                  * Updating the path MTU will be done after examining
 504                  * intermediate extension headers.
 505                  */
 506                 goto deliver;
 507                 break;
 508
 509         case ICMP6_TIME_EXCEEDED:
 510                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
 511                 switch (code) {
 512                 case ICMP6_TIME_EXCEED_TRANSIT:
 513                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 514                         code += PRC_TIMXCEED_INTRANS;
 515                         break;
 516                 default:
 517                         goto badcode;
 518                 }
 519                 goto deliver;
 520                 break;
 521
 522         case ICMP6_PARAM_PROB:
 523                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
 524                 switch (code) {
 525                 case ICMP6_PARAMPROB_NEXTHEADER:
 526                         code = PRC_UNREACH_PROTOCOL;
 527                         break;
 528                 case ICMP6_PARAMPROB_HEADER:
 529                 case ICMP6_PARAMPROB_OPTION:
 530                         code = PRC_PARAMPROB;
 531                         break;
 532                 default:
 533                         goto badcode;
 534                 }
 535                 goto deliver;
 536                 break;
 537
 538         case ICMP6_ECHO_REQUEST:
 539                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
 540                 if (code != 0)
 541                         goto badcode;
 542                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
 543                         /* Give up remote */
 544                         break;
 545                 }
 546                 if ((n->m_flags & M_EXT) != 0
 547                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
 548                         struct mbuf *n0 = n;
 549                         const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
 550                         int n0len;
 551
 552                         /*
 553                          * Prepare an internal mbuf. m_pullup() doesn't
 554                          * always copy the length we specified.
 555                          */
 556                         if (maxlen >= MCLBYTES) {
 557                                 /* Give up remote */
 558                                 m_freem(n0);
 559                                 break;
 560                         }
 561                         MGETHDR(n, MB_DONTWAIT, n0->m_type);
 562                         n0len = n0->m_pkthdr.len;       /* save for use below */
 563                         if (n)
 564                                 M_MOVE_PKTHDR(n, n0);
 565                         if (n && maxlen >= MHLEN) {
 566                                 MCLGET(n, MB_DONTWAIT);
 567                                 if ((n->m_flags & M_EXT) == 0) {
 568                                         m_free(n);
 569                                         n = NULL;
 570                                 }
 571                         }
 572                         if (n == NULL) {
 573                                 /* Give up remote */
 574                                 m_freem(n0);
 575                                 break;
 576                         }
 577                         /*
 578                          * Copy IPv6 and ICMPv6 only.
 579                          */
 580                         nip6 = mtod(n, struct ip6_hdr *);
 581                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 582                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 583                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 584                         noff = sizeof(struct ip6_hdr);
 585                         /* new mbuf contains only ipv6+icmpv6 headers */
 586                         n->m_len = noff + sizeof(struct icmp6_hdr);
 587                         /*
 588                          * Adjust mbuf. ip6_plen will be adjusted in
 589                          * ip6_output().
 590                          */
 591                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 592                         /* recalculate complete packet size */
 593                         n->m_pkthdr.len = n0len + (noff - off);
 594                         n->m_next = n0;
 595                 } else {
 596                         nip6 = mtod(n, struct ip6_hdr *);
 597                         nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
 598                         noff = off;
 599                 }
 600                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 601                 nicmp6->icmp6_code = 0;
 602                 if (n) {
 603                         icmp6stat.icp6s_reflect++;
 604                         icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
 605                         icmp6_reflect(n, noff);
 606                 }
 607                 break;
 608
 609         case ICMP6_ECHO_REPLY:
 610                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
 611                 if (code != 0)
 612                         goto badcode;
 613                 break;
 614
 615         case MLD_LISTENER_QUERY:
 616         case MLD_LISTENER_REPORT:
 617                 if (icmp6len < sizeof(struct mld_hdr))
 618                         goto badlen;
 619                 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
 620                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
 621                 else
 622                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
 623                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 624                         /* give up local */
 625                         mld6_input(m, off);
 626                         m = NULL;
 627                         goto freeit;
 628                 }
 629                 mld6_input(n, off);
 630                 /* m stays. */
 631                 break;
 632
 633         case MLD_LISTENER_DONE:
 634                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
 635                 if (icmp6len < sizeof(struct mld_hdr))  /* necessary? */
 636                         goto badlen;
 637                 break;          /* nothing to be done in kernel */
 638
 639         case MLD_MTRACE_RESP:
 640         case MLD_MTRACE:
 641                 /* XXX: these two are experimental.  not officially defind. */
 642                 /* XXX: per-interface statistics? */
 643                 break;          /* just pass it to applications */
 644
 645         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 646             {
 647                 enum { WRU, FQDN } mode;
 648
 649                 if (!icmp6_nodeinfo)
 650                         break;
 651
 652                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 653                         mode = WRU;
 654                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 655                         mode = FQDN;
 656                 else
 657                         goto badlen;
 658
 659 #define hostnamelen     strlen(hostname)
 660                 if (mode == FQDN) {
 661 #ifndef PULLDOWN_TEST
 662                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
 663                                          IPPROTO_DONE);
 664 #endif
 665                         n = m_copy(m, 0, M_COPYALL);
 666                         if (n)
 667                                 n = ni6_input(n, off);
 668                         /* XXX meaningless if n == NULL */
 669                         noff = sizeof(struct ip6_hdr);
 670                 } else {
 671                         u_char *p;
 672                         int maxlen, maxhlen;
 673
 674                         if ((icmp6_nodeinfo & 5) != 5)
 675                                 break;
 676
 677                         if (code != 0)
 678                                 goto badcode;
 679                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
 680                         if (maxlen >= MCLBYTES) {
 681                                 /* Give up remote */
 682                                 break;
 683                         }
 684                         MGETHDR(n, MB_DONTWAIT, m->m_type);
 685                         if (n && maxlen > MHLEN) {
 686                                 MCLGET(n, MB_DONTWAIT);
 687                                 if ((n->m_flags & M_EXT) == 0) {
 688                                         m_free(n);
 689                                         n = NULL;
 690                                 }
 691                         }
 692                         if (!m_dup_pkthdr(n, m, MB_DONTWAIT)) {
 693                                 /*
 694                                  * Previous code did a blind M_COPY_PKTHDR
 695                                  * and said "just for rcvif".  If true, then
 696                                  * we could tolerate the dup failing (due to
 697                                  * the deep copy of the tag chain).  For now
 698                                  * be conservative and just fail.
 699                                  */
 700                                 m_free(n);
 701                                 n = NULL;
 702                         }
 703                         if (n == NULL) {
 704                                 /* Give up remote */
 705                                 break;
 706                         }
 707                         n->m_pkthdr.rcvif = NULL;
 708                         n->m_len = 0;
 709                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
 710                         if (maxhlen > hostnamelen)
 711                                 maxhlen = hostnamelen;
 712                         /*
 713                          * Copy IPv6 and ICMPv6 only.
 714                          */
 715                         nip6 = mtod(n, struct ip6_hdr *);
 716                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 717                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 718                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 719                         p = (u_char *)(nicmp6 + 1);
 720                         bzero(p, 4);
 721                         bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
 722                         noff = sizeof(struct ip6_hdr);
 723                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 724                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 725                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 726                         nicmp6->icmp6_code = 0;
 727                 }
 728 #undef hostnamelen
 729                 if (n) {
 730                         icmp6stat.icp6s_reflect++;
 731                         icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
 732                         icmp6_reflect(n, noff);
 733                 }
 734                 break;
 735             }
 736
 737         case ICMP6_WRUREPLY:
 738                 if (code != 0)
 739                         goto badcode;
 740                 break;
 741
 742         case ND_ROUTER_SOLICIT:
 743                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
 744                 if (code != 0)
 745                         goto badcode;
 746                 if (icmp6len < sizeof(struct nd_router_solicit))
 747                         goto badlen;
 748                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 749                         /* give up local */
 750                         nd6_rs_input(m, off, icmp6len);
 751                         m = NULL;
 752                         goto freeit;
 753                 }
 754                 nd6_rs_input(n, off, icmp6len);
 755                 /* m stays. */
 756                 break;
 757
 758         case ND_ROUTER_ADVERT:
 759                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
 760                 if (code != 0)
 761                         goto badcode;
 762                 if (icmp6len < sizeof(struct nd_router_advert))
 763                         goto badlen;
 764                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 765                         /* give up local */
 766                         nd6_ra_input(m, off, icmp6len);
 767                         m = NULL;
 768                         goto freeit;
 769                 }
 770                 nd6_ra_input(n, off, icmp6len);
 771                 /* m stays. */
 772                 break;
 773
 774         case ND_NEIGHBOR_SOLICIT:
 775                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
 776                 if (code != 0)
 777                         goto badcode;
 778                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 779                         goto badlen;
 780                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 781                         /* give up local */
 782                         nd6_ns_input(m, off, icmp6len);
 783                         m = NULL;
 784                         goto freeit;
 785                 }
 786                 nd6_ns_input(n, off, icmp6len);
 787                 /* m stays. */
 788                 break;
 789
 790         case ND_NEIGHBOR_ADVERT:
 791                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
 792                 if (code != 0)
 793                         goto badcode;
 794                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 795                         goto badlen;
 796                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 797                         /* give up local */
 798                         nd6_na_input(m, off, icmp6len);
 799                         m = NULL;
 800                         goto freeit;
 801                 }
 802                 nd6_na_input(n, off, icmp6len);
 803                 /* m stays. */
 804                 break;
 805
 806         case ND_REDIRECT:
 807                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
 808                 if (code != 0)
 809                         goto badcode;
 810                 if (icmp6len < sizeof(struct nd_redirect))
 811                         goto badlen;
 812                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 813                         /* give up local */
 814                         icmp6_redirect_input(m, off);
 815                         m = NULL;
 816                         goto freeit;
 817                 }
 818                 icmp6_redirect_input(n, off);
 819                 /* m stays. */
 820                 break;
 821
 822         case ICMP6_ROUTER_RENUMBERING:
 823                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 824                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 825                         goto badcode;
 826                 if (icmp6len < sizeof(struct icmp6_router_renum))
 827                         goto badlen;
 828                 break;
 829
 830         default:
 831                 nd6log((LOG_DEBUG,
 832                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 833                     icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
 834                     ip6_sprintf(&ip6->ip6_dst),
 835                     m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
 836                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 837                         /* ICMPv6 error: MUST deliver it by spec... */
 838                         code = PRC_NCMDS;
 839                         /* deliver */
 840                 } else {
 841                         /* ICMPv6 informational: MUST not deliver */
 842                         break;
 843                 }
 844         deliver:
 845                 if (icmp6_notify_error(m, off, icmp6len, code)) {
 846                         /* In this case, m should've been freed. */
 847                         return(IPPROTO_DONE);
 848                 }
 849                 break;
 850
 851         badcode:
 852                 icmp6stat.icp6s_badcode++;
 853                 break;
 854
 855         badlen:
 856                 icmp6stat.icp6s_badlen++;
 857                 break;
 858         }
 859
 860         /* deliver the packet to appropriate sockets */
 861         icmp6_rip6_input(&m, *offp);
 862
 863         return IPPROTO_DONE;
 864
 865  freeit:
 866         m_freem(m);
 867         return IPPROTO_DONE;
 868 }
 869
 870 static int
 871 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
 872 {
 873         struct icmp6_hdr *icmp6;
 874         struct ip6_hdr *eip6;
 875         u_int32_t notifymtu;
 876         struct sockaddr_in6 icmp6src, icmp6dst;
 877
 878         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 879                 icmp6stat.icp6s_tooshort++;
 880                 goto freeit;
 881         }
 882 #ifndef PULLDOWN_TEST
 883         IP6_EXTHDR_CHECK(m, off,
 884                          sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
 885                          -1);
 886         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 887 #else
 888         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
 889                        sizeof(*icmp6) + sizeof(struct ip6_hdr));
 890         if (icmp6 == NULL) {
 891                 icmp6stat.icp6s_tooshort++;
 892                 return(-1);
 893         }
 894 #endif
 895         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 896
 897         /* Detect the upper level protocol */
 898         {
 899                 void (*ctlfunc) (int, struct sockaddr *, void *);
 900                 u_int8_t nxt = eip6->ip6_nxt;
 901                 int eoff = off + sizeof(struct icmp6_hdr) +
 902                         sizeof(struct ip6_hdr);
 903                 struct ip6ctlparam ip6cp;
 904                 struct in6_addr *finaldst = NULL;
 905                 int icmp6type = icmp6->icmp6_type;
 906                 struct ip6_frag *fh;
 907                 struct ip6_rthdr *rth;
 908                 struct ip6_rthdr0 *rth0;
 909                 int rthlen;
 910
 911                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 912                         struct ip6_ext *eh;
 913
 914                         switch (nxt) {
 915                         case IPPROTO_HOPOPTS:
 916                         case IPPROTO_DSTOPTS:
 917                         case IPPROTO_AH:
 918 #ifndef PULLDOWN_TEST
 919                                 IP6_EXTHDR_CHECK(m, 0, eoff +
 920                                                  sizeof(struct ip6_ext),
 921                                                  -1);
 922                                 eh = (struct ip6_ext *)(mtod(m, caddr_t)
 923                                                         + eoff);
 924 #else
 925                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
 926                                                eoff, sizeof(*eh));
 927                                 if (eh == NULL) {
 928                                         icmp6stat.icp6s_tooshort++;
 929                                         return(-1);
 930                                 }
 931 #endif
 932
 933                                 if (nxt == IPPROTO_AH)
 934                                         eoff += (eh->ip6e_len + 2) << 2;
 935                                 else
 936                                         eoff += (eh->ip6e_len + 1) << 3;
 937                                 nxt = eh->ip6e_nxt;
 938                                 break;
 939                         case IPPROTO_ROUTING:
 940                                 /*
 941                                  * When the erroneous packet contains a
 942                                  * routing header, we should examine the
 943                                  * header to determine the final destination.
 944                                  * Otherwise, we can't properly update
 945                                  * information that depends on the final
 946                                  * destination (e.g. path MTU).
 947                                  */
 948 #ifndef PULLDOWN_TEST
 949                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
 950                                                  -1);
 951                                 rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
 952                                                            + eoff);
 953 #else
 954                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
 955                                                eoff, sizeof(*rth));
 956                                 if (rth == NULL) {
 957                                         icmp6stat.icp6s_tooshort++;
 958                                         return(-1);
 959                                 }
 960 #endif
 961                                 rthlen = (rth->ip6r_len + 1) << 3;
 962                                 /*
 963                                  * XXX: currently there is no
 964                                  * officially defined type other
 965                                  * than type-0.
 966                                  * Note that if the segment left field
 967                                  * is 0, all intermediate hops must
 968                                  * have been passed.
 969                                  */
 970                                 if (rth->ip6r_segleft &&
 971                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
 972                                         int hops;
 973
 974 #ifndef PULLDOWN_TEST
 975                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
 976                                                          -1);
 977                                         rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
 978 #else
 979                                         IP6_EXTHDR_GET(rth0,
 980                                                        struct ip6_rthdr0 *, m,
 981                                                        eoff, rthlen);
 982                                         if (rth0 == NULL) {
 983                                                 icmp6stat.icp6s_tooshort++;
 984                                                 return(-1);
 985                                         }
 986 #endif
 987                                         /* just ignore a bogus header */
 988                                         if ((rth0->ip6r0_len % 2) == 0 &&
 989                                             (hops = rth0->ip6r0_len/2))
 990                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
 991                                 }
 992                                 eoff += rthlen;
 993                                 nxt = rth->ip6r_nxt;
 994                                 break;
 995                         case IPPROTO_FRAGMENT:
 996 #ifndef PULLDOWN_TEST
 997                                 IP6_EXTHDR_CHECK(m, 0, eoff +
 998                                                  sizeof(struct ip6_frag),
 999                                                  -1);
1000                                 fh = (struct ip6_frag *)(mtod(m, caddr_t)
1001                                                          + eoff);
1002 #else
1003                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1004                                                eoff, sizeof(*fh));
1005                                 if (fh == NULL) {
1006                                         icmp6stat.icp6s_tooshort++;
1007                                         return(-1);
1008                                 }
1009 #endif
1010                                 /*
1011                                  * Data after a fragment header is meaningless
1012                                  * unless it is the first fragment, but
1013                                  * we'll go to the notify label for path MTU
1014                                  * discovery.
1015                                  */
1016                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1017                                         goto notify;
1018
1019                                 eoff += sizeof(struct ip6_frag);
1020                                 nxt = fh->ip6f_nxt;
1021                                 break;
1022                         default:
1023                                 /*
1024                                  * This case includes ESP and the No Next
1025                                  * Header.  In such cases going to the notify
1026                                  * label does not have any meaning
1027                                  * (i.e. ctlfunc will be NULL), but we go
1028                                  * anyway since we might have to update
1029                                  * path MTU information.
1030                                  */
1031                                 goto notify;
1032                         }
1033                 }
1034           notify:
1035 #ifndef PULLDOWN_TEST
1036                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1037 #else
1038                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1039                                sizeof(*icmp6) + sizeof(struct ip6_hdr));
1040                 if (icmp6 == NULL) {
1041                         icmp6stat.icp6s_tooshort++;
1042                         return(-1);
1043                 }
1044 #endif
1045
1046                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1047                 bzero(&icmp6dst, sizeof(icmp6dst));
1048                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1049                 icmp6dst.sin6_family = AF_INET6;
1050                 if (finaldst == NULL)
1051                         icmp6dst.sin6_addr = eip6->ip6_dst;
1052                 else
1053                         icmp6dst.sin6_addr = *finaldst;
1054                 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1055                                                           &icmp6dst.sin6_addr);
1056                 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1057                                    NULL, NULL)) {
1058                         /* should be impossbile */
1059                         nd6log((LOG_DEBUG,
1060                             "icmp6_notify_error: in6_embedscope failed\n"));
1061                         goto freeit;
1062                 }
1063
1064                 /*
1065                  * retrieve parameters from the inner IPv6 header, and convert
1066                  * them into sockaddr structures.
1067                  */
1068                 bzero(&icmp6src, sizeof(icmp6src));
1069                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1070                 icmp6src.sin6_family = AF_INET6;
1071                 icmp6src.sin6_addr = eip6->ip6_src;
1072                 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1073                                                           &icmp6src.sin6_addr);
1074                 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1075                                    NULL, NULL)) {
1076                         /* should be impossbile */
1077                         nd6log((LOG_DEBUG,
1078                             "icmp6_notify_error: in6_embedscope failed\n"));
1079                         goto freeit;
1080                 }
1081                 icmp6src.sin6_flowinfo =
1082                         (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1083
1084                 if (finaldst == NULL)
1085                         finaldst = &eip6->ip6_dst;
1086                 ip6cp.ip6c_m = m;
1087                 ip6cp.ip6c_icmp6 = icmp6;
1088                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1089                 ip6cp.ip6c_off = eoff;
1090                 ip6cp.ip6c_finaldst = finaldst;
1091                 ip6cp.ip6c_src = &icmp6src;
1092                 ip6cp.ip6c_nxt = nxt;
1093
1094                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1095                         notifymtu = ntohl(icmp6->icmp6_mtu);
1096                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1097                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1098                 }
1099
1100                 ctlfunc = (void (*) (int, struct sockaddr *, void *))
1101                         (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1102                 if (ctlfunc) {
1103                         (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, &ip6cp);
1104                 }
1105         }
1106         return(0);
1107
1108   freeit:
1109         m_freem(m);
1110         return(-1);
1111 }
1112
1113 void
1114 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1115 {
1116         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1117         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1118         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1119         u_int mtu = ntohl(icmp6->icmp6_mtu);
1120         struct rtentry *rt = NULL;
1121         struct sockaddr_in6 sin6;
1122
1123         if (!validated)
1124                 return;
1125
1126         bzero(&sin6, sizeof(sin6));
1127         sin6.sin6_family = PF_INET6;
1128         sin6.sin6_len = sizeof(struct sockaddr_in6);
1129         sin6.sin6_addr = *dst;
1130         /* XXX normally, this won't happen */
1131         if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1132                 sin6.sin6_addr.s6_addr16[1] =
1133                     htons(m->m_pkthdr.rcvif->if_index);
1134         }
1135         /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1136         rt = rtpurelookup((struct sockaddr *)&sin6);
1137         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1138             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1139                 if (mtu < IPV6_MMTU) {                          /* XXX */
1140                         rt->rt_rmx.rmx_locks |= RTV_MTU;
1141                 } else if (mtu < rt->rt_ifp->if_mtu &&
1142                            rt->rt_rmx.rmx_mtu > mtu) {
1143                         icmp6stat.icp6s_pmtuchg++;
1144                         rt->rt_rmx.rmx_mtu = mtu;
1145                 }
1146         }
1147         if (rt)
1148                 --rt->rt_refcnt;
1149 }
1150
1151 /*
1152  * Process a Node Information Query packet, based on
1153  * draft-ietf-ipngwg-icmp-name-lookups-07.
1154  *
1155  * Spec incompatibilities:
1156  * - IPv6 Subject address handling
1157  * - IPv4 Subject address handling support missing
1158  * - Proxy reply (answer even if it's not for me)
1159  * - joins NI group address at in6_ifattach() time only, does not cope
1160  *   with hostname changes by sethostname(3)
1161  */
1162 #define hostnamelen     strlen(hostname)
1163 static struct mbuf *
1164 ni6_input(struct mbuf *m, int off)
1165 {
1166         struct icmp6_nodeinfo *ni6, *nni6;
1167         struct mbuf *n = NULL;
1168         u_int16_t qtype;
1169         int subjlen;
1170         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1171         struct ni_reply_fqdn *fqdn;
1172         int addrs;              /* for NI_QTYPE_NODEADDR */
1173         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1174         struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1175         struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1176         struct ip6_hdr *ip6;
1177         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1178         char *subj = NULL;
1179         struct in6_ifaddr *ia6 = NULL;
1180
1181         ip6 = mtod(m, struct ip6_hdr *);
1182 #ifndef PULLDOWN_TEST
1183         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1184 #else
1185         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1186         if (ni6 == NULL) {
1187                 /* m is already reclaimed */
1188                 return NULL;
1189         }
1190 #endif
1191
1192         /*
1193          * Validate IPv6 destination address.
1194          *
1195          * The Responder must discard the Query without further processing
1196          * unless it is one of the Responder's unicast or anycast addresses, or
1197          * a link-local scope multicast address which the Responder has joined.
1198          * [icmp-name-lookups-07, Section 4.]
1199          */
1200         bzero(&sin6, sizeof(sin6));
1201         sin6.sin6_family = AF_INET6;
1202         sin6.sin6_len = sizeof(struct sockaddr_in6);
1203         bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1204         /* XXX scopeid */
1205         if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1206                 /* unicast/anycast, fine */
1207                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1208                     (icmp6_nodeinfo & 4) == 0) {
1209                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1210                                 "a temporary address in %s:%d",
1211                                __FILE__, __LINE__));
1212                         goto bad;
1213                 }
1214         } else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1215                 ; /* link-local multicast, fine */
1216         else
1217                 goto bad;
1218
1219         /* validate query Subject field. */
1220         qtype = ntohs(ni6->ni_qtype);
1221         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1222         switch (qtype) {
1223         case NI_QTYPE_NOOP:
1224         case NI_QTYPE_SUPTYPES:
1225                 /* 07 draft */
1226                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1227                         break;
1228                 /* FALLTHROUGH */
1229         case NI_QTYPE_FQDN:
1230         case NI_QTYPE_NODEADDR:
1231                 switch (ni6->ni_code) {
1232                 case ICMP6_NI_SUBJ_IPV6:
1233 #if ICMP6_NI_SUBJ_IPV6 != 0
1234                 case 0:
1235 #endif
1236                         /*
1237                          * backward compatibility - try to accept 03 draft
1238                          * format, where no Subject is present.
1239                          */
1240                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1241                             subjlen == 0) {
1242                                 oldfqdn++;
1243                                 break;
1244                         }
1245 #if ICMP6_NI_SUBJ_IPV6 != 0
1246                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1247                                 goto bad;
1248 #endif
1249
1250                         if (subjlen != sizeof(sin6.sin6_addr))
1251                                 goto bad;
1252
1253                         /*
1254                          * Validate Subject address.
1255                          *
1256                          * Not sure what exactly "address belongs to the node"
1257                          * means in the spec, is it just unicast, or what?
1258                          *
1259                          * At this moment we consider Subject address as
1260                          * "belong to the node" if the Subject address equals
1261                          * to the IPv6 destination address; validation for
1262                          * IPv6 destination address should have done enough
1263                          * check for us.
1264                          *
1265                          * We do not do proxy at this moment.
1266                          */
1267                         /* m_pulldown instead of copy? */
1268                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1269                             subjlen, (caddr_t)&sin6.sin6_addr);
1270                         sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1271                                                               &sin6.sin6_addr);
1272                         in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1273                         bzero(&sin6_d, sizeof(sin6_d));
1274                         sin6_d.sin6_family = AF_INET6; /* not used, actually */
1275                         sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1276                         sin6_d.sin6_addr = ip6->ip6_dst;
1277                         sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1278                                                                 &ip6->ip6_dst);
1279                         in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1280                         subj = (char *)&sin6;
1281                         if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1282                                 break;
1283
1284                         /*
1285                          * XXX if we are to allow other cases, we should really
1286                          * be careful about scope here.
1287                          * basically, we should disallow queries toward IPv6
1288                          * destination X with subject Y, if scope(X) > scope(Y).
1289                          * if we allow scope(X) > scope(Y), it will result in
1290                          * information leakage across scope boundary.
1291                          */
1292                         goto bad;
1293
1294                 case ICMP6_NI_SUBJ_FQDN:
1295                         /*
1296                          * Validate Subject name with gethostname(3).
1297                          *
1298                          * The behavior may need some debate, since:
1299                          * - we are not sure if the node has FQDN as
1300                          *   hostname (returned by gethostname(3)).
1301                          * - the code does wildcard match for truncated names.
1302                          *   however, we are not sure if we want to perform
1303                          *   wildcard match, if gethostname(3) side has
1304                          *   truncated hostname.
1305                          */
1306                         n = ni6_nametodns(hostname, hostnamelen, 0);
1307                         if (!n || n->m_next || n->m_len == 0)
1308                                 goto bad;
1309                         IP6_EXTHDR_GET(subj, char *, m,
1310                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1311                         if (subj == NULL)
1312                                 goto bad;
1313                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1314                                         n->m_len)) {
1315                                 goto bad;
1316                         }
1317                         m_freem(n);
1318                         n = NULL;
1319                         break;
1320
1321                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1322                 default:
1323                         goto bad;
1324                 }
1325                 break;
1326         }
1327
1328         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1329         switch (qtype) {
1330         case NI_QTYPE_FQDN:
1331                 if ((icmp6_nodeinfo & 1) == 0)
1332                         goto bad;
1333                 break;
1334         case NI_QTYPE_NODEADDR:
1335                 if ((icmp6_nodeinfo & 2) == 0)
1336                         goto bad;
1337                 break;
1338         }
1339
1340         /* guess reply length */
1341         switch (qtype) {
1342         case NI_QTYPE_NOOP:
1343                 break;          /* no reply data */
1344         case NI_QTYPE_SUPTYPES:
1345                 replylen += sizeof(u_int32_t);
1346                 break;
1347         case NI_QTYPE_FQDN:
1348                 /* XXX will append an mbuf */
1349                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1350                 break;
1351         case NI_QTYPE_NODEADDR:
1352                 addrs = ni6_addrs(ni6, m, &ifp, subj);
1353                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1354                                           sizeof(u_int32_t))) > MCLBYTES)
1355                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1356                 break;
1357         default:
1358                 /*
1359                  * XXX: We must return a reply with the ICMP6 code
1360                  * `unknown Qtype' in this case. However we regard the case
1361                  * as an FQDN query for backward compatibility.
1362                  * Older versions set a random value to this field,
1363                  * so it rarely varies in the defined qtypes.
1364                  * But the mechanism is not reliable...
1365                  * maybe we should obsolete older versions.
1366                  */
1367                 qtype = NI_QTYPE_FQDN;
1368                 /* XXX will append an mbuf */
1369                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1370                 oldfqdn++;
1371                 break;
1372         }
1373
1374         /* allocate an mbuf to reply. */
1375         MGETHDR(n, MB_DONTWAIT, m->m_type);
1376         if (n == NULL) {
1377                 m_freem(m);
1378                 return(NULL);
1379         }
1380         M_MOVE_PKTHDR(n, m); /* just for recvif */
1381         if (replylen > MHLEN) {
1382                 if (replylen > MCLBYTES) {
1383                         /*
1384                          * XXX: should we try to allocate more? But MCLBYTES
1385                          * is probably much larger than IPV6_MMTU...
1386                          */
1387                         goto bad;
1388                 }
1389                 MCLGET(n, MB_DONTWAIT);
1390                 if ((n->m_flags & M_EXT) == 0) {
1391                         goto bad;
1392                 }
1393         }
1394         n->m_pkthdr.len = n->m_len = replylen;
1395
1396         /* copy mbuf header and IPv6 + Node Information base headers */
1397         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1398         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1399         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1400
1401         /* qtype dependent procedure */
1402         switch (qtype) {
1403         case NI_QTYPE_NOOP:
1404                 nni6->ni_code = ICMP6_NI_SUCCESS;
1405                 nni6->ni_flags = 0;
1406                 break;
1407         case NI_QTYPE_SUPTYPES:
1408         {
1409                 u_int32_t v;
1410                 nni6->ni_code = ICMP6_NI_SUCCESS;
1411                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1412                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1413                 v = (u_int32_t)htonl(0x0000000f);
1414                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1415                 break;
1416         }
1417         case NI_QTYPE_FQDN:
1418                 nni6->ni_code = ICMP6_NI_SUCCESS;
1419                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1420                                                 sizeof(struct ip6_hdr) +
1421                                                 sizeof(struct icmp6_nodeinfo));
1422                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1423                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1424                 /*
1425                  * XXX do we really have FQDN in variable "hostname"?
1426                  */
1427                 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1428                 if (n->m_next == NULL)
1429                         goto bad;
1430                 /* XXX we assume that n->m_next is not a chain */
1431                 if (n->m_next->m_next != NULL)
1432                         goto bad;
1433                 n->m_pkthdr.len += n->m_next->m_len;
1434                 break;
1435         case NI_QTYPE_NODEADDR:
1436         {
1437                 int lenlim, copied;
1438
1439                 nni6->ni_code = ICMP6_NI_SUCCESS;
1440                 n->m_pkthdr.len = n->m_len =
1441                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1442                 lenlim = M_TRAILINGSPACE(n);
1443                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1444                 /* XXX: reset mbuf length */
1445                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1446                         sizeof(struct icmp6_nodeinfo) + copied;
1447                 break;
1448         }
1449         default:
1450                 break;          /* XXX impossible! */
1451         }
1452
1453         nni6->ni_type = ICMP6_NI_REPLY;
1454         m_freem(m);
1455         return(n);
1456
1457   bad:
1458         m_freem(m);
1459         if (n)
1460                 m_freem(n);
1461         return(NULL);
1462 }
1463 #undef hostnamelen
1464
1465 /*
1466  * make a mbuf with DNS-encoded string.  no compression support.
1467  *
1468  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1469  * treated as truncated name (two \0 at the end).  this is a wild guess.
1470  */
1471 static struct mbuf *
1472 ni6_nametodns(const char *name, int namelen,
1473               int old) /* return pascal string if non-zero */
1474 {
1475         struct mbuf *m;
1476         char *cp, *ep;
1477         const char *p, *q;
1478         int i, len, nterm;
1479
1480         if (old)
1481                 len = namelen + 1;
1482         else
1483                 len = MCLBYTES;
1484
1485         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1486         MGET(m, MB_DONTWAIT, MT_DATA);
1487         if (m && len > MLEN) {
1488                 MCLGET(m, MB_DONTWAIT);
1489                 if ((m->m_flags & M_EXT) == 0)
1490                         goto fail;
1491         }
1492         if (!m)
1493                 goto fail;
1494         m->m_next = NULL;
1495
1496         if (old) {
1497                 m->m_len = len;
1498                 *mtod(m, char *) = namelen;
1499                 bcopy(name, mtod(m, char *) + 1, namelen);
1500                 return m;
1501         } else {
1502                 m->m_len = 0;
1503                 cp = mtod(m, char *);
1504                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1505
1506                 /* if not certain about my name, return empty buffer */
1507                 if (namelen == 0)
1508                         return m;
1509
1510                 /*
1511                  * guess if it looks like shortened hostname, or FQDN.
1512                  * shortened hostname needs two trailing "\0".
1513                  */
1514                 i = 0;
1515                 for (p = name; p < name + namelen; p++) {
1516                         if (*p && *p == '.')
1517                                 i++;
1518                 }
1519                 if (i < 2)
1520                         nterm = 2;
1521                 else
1522                         nterm = 1;
1523
1524                 p = name;
1525                 while (cp < ep && p < name + namelen) {
1526                         i = 0;
1527                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1528                                 i++;
1529                         /* result does not fit into mbuf */
1530                         if (cp + i + 1 >= ep)
1531                                 goto fail;
1532                         /*
1533                          * DNS label length restriction, RFC1035 page 8.
1534                          * "i == 0" case is included here to avoid returning
1535                          * 0-length label on "foo..bar".
1536                          */
1537                         if (i <= 0 || i >= 64)
1538                                 goto fail;
1539                         *cp++ = i;
1540                         bcopy(p, cp, i);
1541                         cp += i;
1542                         p = q;
1543                         if (p < name + namelen && *p == '.')
1544                                 p++;
1545                 }
1546                 /* termination */
1547                 if (cp + nterm >= ep)
1548                         goto fail;
1549                 while (nterm-- > 0)
1550                         *cp++ = '\0';
1551                 m->m_len = cp - mtod(m, char *);
1552                 return m;
1553         }
1554
1555         panic("should not reach here");
1556         /* NOTREACHED */
1557
1558  fail:
1559         if (m)
1560                 m_freem(m);
1561         return NULL;
1562 }
1563
1564 /*
1565  * check if two DNS-encoded string matches.  takes care of truncated
1566  * form (with \0\0 at the end).  no compression support.
1567  * XXX upper/lowercase match (see RFC2065)
1568  */
1569 static int
1570 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1571 {
1572         const char *a0, *b0;
1573         int l;
1574
1575         /* simplest case - need validation? */
1576         if (alen == blen && bcmp(a, b, alen) == 0)
1577                 return 1;
1578
1579         a0 = a;
1580         b0 = b;
1581
1582         /* termination is mandatory */
1583         if (alen < 2 || blen < 2)
1584                 return 0;
1585         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1586                 return 0;
1587         alen--;
1588         blen--;
1589
1590         while (a - a0 < alen && b - b0 < blen) {
1591                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1592                         return 0;
1593
1594                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1595                         return 0;
1596                 /* we don't support compression yet */
1597                 if (a[0] >= 64 || b[0] >= 64)
1598                         return 0;
1599
1600                 /* truncated case */
1601                 if (a[0] == 0 && a - a0 == alen - 1)
1602                         return 1;
1603                 if (b[0] == 0 && b - b0 == blen - 1)
1604                         return 1;
1605                 if (a[0] == 0 || b[0] == 0)
1606                         return 0;
1607
1608                 if (a[0] != b[0])
1609                         return 0;
1610                 l = a[0];
1611                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1612                         return 0;
1613                 if (bcmp(a + 1, b + 1, l) != 0)
1614                         return 0;
1615
1616                 a += 1 + l;
1617                 b += 1 + l;
1618         }
1619
1620         if (a - a0 == alen && b - b0 == blen)
1621                 return 1;
1622         else
1623                 return 0;
1624 }
1625
1626 /*
1627  * calculate the number of addresses to be returned in the node info reply.
1628  */
1629 static int
1630 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1631           char *subj)
1632 {
1633         struct ifnet *ifp;
1634         struct in6_ifaddr *ifa6;
1635         struct ifaddr *ifa;
1636         struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1637         int addrs = 0, addrsofif, iffound = 0;
1638         int niflags = ni6->ni_flags;
1639
1640         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1641                 switch (ni6->ni_code) {
1642                 case ICMP6_NI_SUBJ_IPV6:
1643                         if (subj == NULL) /* must be impossible... */
1644                                 return(0);
1645                         subj_ip6 = (struct sockaddr_in6 *)subj;
1646                         break;
1647                 default:
1648                         /*
1649                          * XXX: we only support IPv6 subject address for
1650                          * this Qtype.
1651                          */
1652                         return(0);
1653                 }
1654         }
1655
1656         for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1657         {
1658                 addrsofif = 0;
1659                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1660                 {
1661                         if (ifa->ifa_addr->sa_family != AF_INET6)
1662                                 continue;
1663                         ifa6 = (struct in6_ifaddr *)ifa;
1664
1665                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1666                             IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1667                                                &ifa6->ia_addr.sin6_addr))
1668                                 iffound = 1;
1669
1670                         /*
1671                          * IPv4-mapped addresses can only be returned by a
1672                          * Node Information proxy, since they represent
1673                          * addresses of IPv4-only nodes, which perforce do
1674                          * not implement this protocol.
1675                          * [icmp-name-lookups-07, Section 5.4]
1676                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1677                          * this function at this moment.
1678                          */
1679
1680                         /* What do we have to do about ::1? */
1681                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1682                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1683                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1684                                         continue;
1685                                 break;
1686                         case IPV6_ADDR_SCOPE_SITELOCAL:
1687                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1688                                         continue;
1689                                 break;
1690                         case IPV6_ADDR_SCOPE_GLOBAL:
1691                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1692                                         continue;
1693                                 break;
1694                         default:
1695                                 continue;
1696                         }
1697
1698                         /*
1699                          * check if anycast is okay.
1700                          * XXX: just experimental.  not in the spec.
1701                          */
1702                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1703                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1704                                 continue; /* we need only unicast addresses */
1705                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1706                             (icmp6_nodeinfo & 4) == 0) {
1707                                 continue;
1708                         }
1709                         addrsofif++; /* count the address */
1710                 }
1711                 if (iffound) {
1712                         *ifpp = ifp;
1713                         return(addrsofif);
1714                 }
1715
1716                 addrs += addrsofif;
1717         }
1718
1719         return(addrs);
1720 }
1721
1722 static int
1723 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1724                 struct ifnet *ifp0, int resid)
1725 {
1726         struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1727         struct in6_ifaddr *ifa6;
1728         struct ifaddr *ifa;
1729         struct ifnet *ifp_dep = NULL;
1730         int copied = 0, allow_deprecated = 0;
1731         u_char *cp = (u_char *)(nni6 + 1);
1732         int niflags = ni6->ni_flags;
1733         u_int32_t ltime;
1734
1735         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1736                 return(0);      /* needless to copy */
1737
1738   again:
1739
1740         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1741         {
1742                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1743                         if (ifa->ifa_addr->sa_family != AF_INET6)
1744                                 continue;
1745                         ifa6 = (struct in6_ifaddr *)ifa;
1746
1747                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1748                             allow_deprecated == 0) {
1749                                 /*
1750                                  * prefererred address should be put before
1751                                  * deprecated addresses.
1752                                  */
1753
1754                                 /* record the interface for later search */
1755                                 if (ifp_dep == NULL)
1756                                         ifp_dep = ifp;
1757
1758                                 continue;
1759                         }
1760                         else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1761                                  allow_deprecated != 0)
1762                                 continue; /* we now collect deprecated addrs */
1763
1764                         /* What do we have to do about ::1? */
1765                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1766                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1767                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1768                                         continue;
1769                                 break;
1770                         case IPV6_ADDR_SCOPE_SITELOCAL:
1771                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1772                                         continue;
1773                                 break;
1774                         case IPV6_ADDR_SCOPE_GLOBAL:
1775                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1776                                         continue;
1777                                 break;
1778                         default:
1779                                 continue;
1780                         }
1781
1782                         /*
1783                          * check if anycast is okay.
1784                          * XXX: just experimental. not in the spec.
1785                          */
1786                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1787                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1788                                 continue;
1789                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1790                             (icmp6_nodeinfo & 4) == 0) {
1791                                 continue;
1792                         }
1793
1794                         /* now we can copy the address */
1795                         if (resid < sizeof(struct in6_addr) +
1796                             sizeof(u_int32_t)) {
1797                                 /*
1798                                  * We give up much more copy.
1799                                  * Set the truncate flag and return.
1800                                  */
1801                                 nni6->ni_flags |=
1802                                         NI_NODEADDR_FLAG_TRUNCATE;
1803                                 return(copied);
1804                         }
1805
1806                         /*
1807                          * Set the TTL of the address.
1808                          * The TTL value should be one of the following
1809                          * according to the specification:
1810                          *
1811                          * 1. The remaining lifetime of a DHCP lease on the
1812                          *    address, or
1813                          * 2. The remaining Valid Lifetime of a prefix from
1814                          *    which the address was derived through Stateless
1815                          *    Autoconfiguration.
1816                          *
1817                          * Note that we currently do not support stateful
1818                          * address configuration by DHCPv6, so the former
1819                          * case can't happen.
1820                          */
1821                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1822                                 ltime = ND6_INFINITE_LIFETIME;
1823                         else {
1824                                 if (ifa6->ia6_lifetime.ia6t_expire >
1825                                     time_second)
1826                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1827                                 else
1828                                         ltime = 0;
1829                         }
1830
1831                         bcopy(&ltime, cp, sizeof(u_int32_t));
1832                         cp += sizeof(u_int32_t);
1833
1834                         /* copy the address itself */
1835                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1836                               sizeof(struct in6_addr));
1837                         /* XXX: KAME link-local hack; remove ifindex */
1838                         if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1839                                 ((struct in6_addr *)cp)->s6_addr16[1] = 0;
1840                         cp += sizeof(struct in6_addr);
1841
1842                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1843                         copied += (sizeof(struct in6_addr) +
1844                                    sizeof(u_int32_t));
1845                 }
1846                 if (ifp0)       /* we need search only on the specified IF */
1847                         break;
1848         }
1849
1850         if (allow_deprecated == 0 && ifp_dep != NULL) {
1851                 ifp = ifp_dep;
1852                 allow_deprecated = 1;
1853
1854                 goto again;
1855         }
1856
1857         return(copied);
1858 }
1859
1860 /*
1861  * XXX almost dup'ed code with rip6_input.
1862  */
1863 static int
1864 icmp6_rip6_input(struct mbuf **mp, int  off)
1865 {
1866         struct mbuf *m = *mp;
1867         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1868         struct in6pcb *in6p;
1869         struct in6pcb *last = NULL;
1870         struct sockaddr_in6 rip6src;
1871         struct icmp6_hdr *icmp6;
1872         struct mbuf *opts = NULL;
1873
1874 #ifndef PULLDOWN_TEST
1875         /* this is assumed to be safe. */
1876         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1877 #else
1878         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1879         if (icmp6 == NULL) {
1880                 /* m is already reclaimed */
1881                 return IPPROTO_DONE;
1882         }
1883 #endif
1884
1885         bzero(&rip6src, sizeof(rip6src));
1886         rip6src.sin6_len = sizeof(struct sockaddr_in6);
1887         rip6src.sin6_family = AF_INET6;
1888         /* KAME hack: recover scopeid */
1889         in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1890
1891         LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1892         {
1893                 if (in6p->inp_flags & INP_PLACEMARKER)
1894                         continue;
1895                 if ((in6p->inp_vflag & INP_IPV6) == 0)
1896                         continue;
1897 #ifdef HAVE_NRL_INPCB
1898                 if (!(in6p->in6p_flags & INP_IPV6))
1899                         continue;
1900 #endif
1901                 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1902                         continue;
1903                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1904                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1905                         continue;
1906                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1907                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1908                         continue;
1909                 if (in6p->in6p_icmp6filt
1910                     && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1911                                  in6p->in6p_icmp6filt))
1912                         continue;
1913                 if (last) {
1914                         struct  mbuf *n;
1915                         if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1916                                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1917                                         ip6_savecontrol(last, &opts, ip6, n);
1918                                 /* strip intermediate headers */
1919                                 m_adj(n, off);
1920                                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1921                                                  (struct sockaddr *)&rip6src,
1922                                                  n, opts) == 0) {
1923                                         /* should notify about lost packet */
1924                                         m_freem(n);
1925                                         if (opts) {
1926                                                 m_freem(opts);
1927                                         }
1928                                 } else
1929                                         sorwakeup(last->in6p_socket);
1930                                 opts = NULL;
1931                         }
1932                 }
1933                 last = in6p;
1934         }
1935         if (last) {
1936                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1937                         ip6_savecontrol(last, &opts, ip6, m);
1938                 /* strip intermediate headers */
1939                 m_adj(m, off);
1940                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1941                                  (struct sockaddr *)&rip6src, m, opts) == 0) {
1942                         m_freem(m);
1943                         if (opts)
1944                                 m_freem(opts);
1945                 } else
1946                         sorwakeup(last->in6p_socket);
1947         } else {
1948                 m_freem(m);
1949                 ip6stat.ip6s_delivered--;
1950         }
1951         return IPPROTO_DONE;
1952 }
1953
1954 /*
1955  * Reflect the ip6 packet back to the source.
1956  * OFF points to the icmp6 header, counted from the top of the mbuf.
1957  */
1958 void
1959 icmp6_reflect(struct mbuf *m, size_t off)
1960 {
1961         struct ip6_hdr *ip6;
1962         struct icmp6_hdr *icmp6;
1963         struct in6_ifaddr *ia;
1964         struct in6_addr t, *src = 0;
1965         int plen;
1966         int type, code;
1967         struct ifnet *outif = NULL;
1968         struct sockaddr_in6 sa6_src, sa6_dst;
1969 #ifdef COMPAT_RFC1885
1970         int mtu = IPV6_MMTU;
1971         struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1972 #endif
1973
1974         /* too short to reflect */
1975         if (off < sizeof(struct ip6_hdr)) {
1976                 nd6log((LOG_DEBUG,
1977                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1978                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
1979                     __FILE__, __LINE__));
1980                 goto bad;
1981         }
1982
1983         /*
1984          * If there are extra headers between IPv6 and ICMPv6, strip
1985          * off that header first.
1986          */
1987 #ifdef DIAGNOSTIC
1988         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1989                 panic("assumption failed in icmp6_reflect");
1990 #endif
1991         if (off > sizeof(struct ip6_hdr)) {
1992                 size_t l;
1993                 struct ip6_hdr nip6;
1994
1995                 l = off - sizeof(struct ip6_hdr);
1996                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1997                 m_adj(m, l);
1998                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1999                 if (m->m_len < l) {
2000                         if ((m = m_pullup(m, l)) == NULL)
2001                                 return;
2002                 }
2003                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2004         } else /* off == sizeof(struct ip6_hdr) */ {
2005                 size_t l;
2006                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2007                 if (m->m_len < l) {
2008                         if ((m = m_pullup(m, l)) == NULL)
2009                                 return;
2010                 }
2011         }
2012         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2013         ip6 = mtod(m, struct ip6_hdr *);
2014         ip6->ip6_nxt = IPPROTO_ICMPV6;
2015         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2016         type = icmp6->icmp6_type; /* keep type for statistics */
2017         code = icmp6->icmp6_code; /* ditto. */
2018
2019         t = ip6->ip6_dst;
2020         /*
2021          * ip6_input() drops a packet if its src is multicast.
2022          * So, the src is never multicast.
2023          */
2024         ip6->ip6_dst = ip6->ip6_src;
2025
2026         /*
2027          * XXX: make sure to embed scope zone information, using
2028          * already embedded IDs or the received interface (if any).
2029          * Note that rcvif may be NULL.
2030          * TODO: scoped routing case (XXX).
2031          */
2032         bzero(&sa6_src, sizeof(sa6_src));
2033         sa6_src.sin6_family = AF_INET6;
2034         sa6_src.sin6_len = sizeof(sa6_src);
2035         sa6_src.sin6_addr = ip6->ip6_dst;
2036         in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2037         in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2038         bzero(&sa6_dst, sizeof(sa6_dst));
2039         sa6_dst.sin6_family = AF_INET6;
2040         sa6_dst.sin6_len = sizeof(sa6_dst);
2041         sa6_dst.sin6_addr = t;
2042         in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2043         in6_embedscope(&t, &sa6_dst, NULL, NULL);
2044
2045 #ifdef COMPAT_RFC1885
2046         /*
2047          * xxx guess MTU
2048          * RFC 1885 requires that echo reply should be truncated if it
2049          * does not fit in with (return) path MTU, but the description was
2050          * removed in the new spec.
2051          */
2052         if (icmp6_reflect_rt.ro_rt == 0 ||
2053             ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2054                 if (icmp6_reflect_rt.ro_rt) {
2055                         RTFREE(icmp6_reflect_rt.ro_rt);
2056                         icmp6_reflect_rt.ro_rt = 0;
2057                 }
2058                 bzero(sin6, sizeof(*sin6));
2059                 sin6->sin6_family = PF_INET6;
2060                 sin6->sin6_len = sizeof(struct sockaddr_in6);
2061                 sin6->sin6_addr = ip6->ip6_dst;
2062
2063                 rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2064                             RTF_PRCLONING);
2065         }
2066
2067         if (icmp6_reflect_rt.ro_rt == 0)
2068                 goto bad;
2069
2070         if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2071             && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2072                 mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2073
2074         if (mtu < m->m_pkthdr.len) {
2075                 plen -= (m->m_pkthdr.len - mtu);
2076                 m_adj(m, mtu - m->m_pkthdr.len);
2077         }
2078 #endif
2079         /*
2080          * If the incoming packet was addressed directly to us(i.e. unicast),
2081          * use dst as the src for the reply.
2082          * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2083          * (for example) when we encounter an error while forwarding procedure
2084          * destined to a duplicated address of ours.
2085          */
2086         for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2087                 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2088                     (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2089                         src = &t;
2090                         break;
2091                 }
2092         if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2093                 /*
2094                  * This is the case if the dst is our link-local address
2095                  * and the sender is also ourselves.
2096                  */
2097                 src = &t;
2098         }
2099
2100         if (src == 0) {
2101                 int e;
2102                 struct route_in6 ro;
2103
2104                 /*
2105                  * This case matches to multicasts, our anycast, or unicasts
2106                  * that we do not own.  Select a source address based on the
2107                  * source address of the erroneous packet.
2108                  */
2109                 bzero(&ro, sizeof(ro));
2110                 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2111                 if (ro.ro_rt)
2112                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2113                 if (src == NULL) {
2114                         nd6log((LOG_DEBUG,
2115                             "icmp6_reflect: source can't be determined: "
2116                             "dst=%s, error=%d\n",
2117                             ip6_sprintf(&sa6_src.sin6_addr), e));
2118                         goto bad;
2119                 }
2120         }
2121
2122         ip6->ip6_src = *src;
2123
2124         ip6->ip6_flow = 0;
2125         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2126         ip6->ip6_vfc |= IPV6_VERSION;
2127         ip6->ip6_nxt = IPPROTO_ICMPV6;
2128         if (m->m_pkthdr.rcvif) {
2129                 /* XXX: This may not be the outgoing interface */
2130                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2131         } else
2132                 ip6->ip6_hlim = ip6_defhlim;
2133
2134         icmp6->icmp6_cksum = 0;
2135         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2136                                         sizeof(struct ip6_hdr), plen);
2137
2138         /*
2139          * XXX option handling
2140          */
2141
2142         m->m_flags &= ~(M_BCAST|M_MCAST);
2143
2144 #ifdef COMPAT_RFC1885
2145         ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2146 #else
2147         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2148 #endif
2149         if (outif)
2150                 icmp6_ifoutstat_inc(outif, type, code);
2151
2152         return;
2153
2154  bad:
2155         m_freem(m);
2156         return;
2157 }
2158
2159 void
2160 icmp6_fasttimo(void)
2161 {
2162
2163         mld6_fasttimeo();
2164 }
2165
2166 static const char *
2167 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2168                     struct in6_addr *tgt6)
2169 {
2170         static char buf[1024];
2171         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2172                 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2173         return buf;
2174 }
2175
2176 void
2177 icmp6_redirect_input(struct mbuf *m, int off)
2178 {
2179         struct ifnet *ifp = m->m_pkthdr.rcvif;
2180         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2181         struct nd_redirect *nd_rd;
2182         int icmp6len = ntohs(ip6->ip6_plen);
2183         char *lladdr = NULL;
2184         int lladdrlen = 0;
2185         u_char *redirhdr = NULL;
2186         int redirhdrlen = 0;
2187         struct rtentry *rt = NULL;
2188         int is_router;
2189         int is_onlink;
2190         struct in6_addr src6 = ip6->ip6_src;
2191         struct in6_addr redtgt6;
2192         struct in6_addr reddst6;
2193         union nd_opts ndopts;
2194
2195         if (!m || !ifp)
2196                 return;
2197
2198         /* XXX if we are router, we don't update route by icmp6 redirect */
2199         if (ip6_forwarding)
2200                 goto freeit;
2201         if (!icmp6_rediraccept)
2202                 goto freeit;
2203
2204 #ifndef PULLDOWN_TEST
2205         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2206         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2207 #else
2208         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2209         if (nd_rd == NULL) {
2210                 icmp6stat.icp6s_tooshort++;
2211                 return;
2212         }
2213 #endif
2214         redtgt6 = nd_rd->nd_rd_target;
2215         reddst6 = nd_rd->nd_rd_dst;
2216
2217         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2218                 redtgt6.s6_addr16[1] = htons(ifp->if_index);
2219         if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2220                 reddst6.s6_addr16[1] = htons(ifp->if_index);
2221
2222         /* validation */
2223         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2224                 nd6log((LOG_ERR,
2225                         "ICMP6 redirect sent from %s rejected; "
2226                         "must be from linklocal\n", ip6_sprintf(&src6)));
2227                 goto bad;
2228         }
2229         if (ip6->ip6_hlim != 255) {
2230                 nd6log((LOG_ERR,
2231                         "ICMP6 redirect sent from %s rejected; "
2232                         "hlim=%d (must be 255)\n",
2233                         ip6_sprintf(&src6), ip6->ip6_hlim));
2234                 goto bad;
2235         }
2236     {
2237         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2238         struct sockaddr_in6 sin6;
2239         struct in6_addr *gw6;
2240
2241         bzero(&sin6, sizeof(sin6));
2242         sin6.sin6_family = AF_INET6;
2243         sin6.sin6_len = sizeof(struct sockaddr_in6);
2244         bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2245         rt = rtpurelookup((struct sockaddr *)&sin6);
2246         if (rt != NULL) {
2247                 if (rt->rt_gateway == NULL ||
2248                     rt->rt_gateway->sa_family != AF_INET6) {
2249                         nd6log((LOG_ERR,
2250                             "ICMP6 redirect rejected; no route "
2251                             "with inet6 gateway found for redirect dst: %s\n",
2252                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2253                         --rt->rt_refcnt;
2254                         goto bad;
2255                 }
2256
2257                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2258                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2259                         nd6log((LOG_ERR,
2260                                 "ICMP6 redirect rejected; "
2261                                 "not equal to gw-for-src=%s (must be same): "
2262                                 "%s\n",
2263                                 ip6_sprintf(gw6),
2264                                 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2265                         --rt->rt_refcnt;
2266                         goto bad;
2267                 }
2268         } else {
2269                 nd6log((LOG_ERR,
2270                         "ICMP6 redirect rejected; "
2271                         "no route found for redirect dst: %s\n",
2272                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2273                 goto bad;
2274         }
2275         --rt->rt_refcnt;
2276         rt = NULL;
2277     }
2278         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2279                 nd6log((LOG_ERR,
2280                         "ICMP6 redirect rejected; "
2281                         "redirect dst must be unicast: %s\n",
2282                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2283                 goto bad;
2284         }
2285
2286         is_router = is_onlink = 0;
2287         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2288                 is_router = 1;  /* router case */
2289         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2290                 is_onlink = 1;  /* on-link destination case */
2291         if (!is_router && !is_onlink) {
2292                 nd6log((LOG_ERR,
2293                         "ICMP6 redirect rejected; "
2294                         "neither router case nor onlink case: %s\n",
2295                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2296                 goto bad;
2297         }
2298         /* validation passed */
2299
2300         icmp6len -= sizeof(*nd_rd);
2301         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2302         if (nd6_options(&ndopts) < 0) {
2303                 nd6log((LOG_INFO, "icmp6_redirect_input: "
2304                         "invalid ND option, rejected: %s\n",
2305                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2306                 /* nd6_options have incremented stats */
2307                 goto freeit;
2308         }
2309
2310         if (ndopts.nd_opts_tgt_lladdr) {
2311                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2312                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2313         }
2314
2315         if (ndopts.nd_opts_rh) {
2316                 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2317                 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2318         }
2319
2320         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2321                 nd6log((LOG_INFO,
2322                         "icmp6_redirect_input: lladdrlen mismatch for %s "
2323                         "(if %d, icmp6 packet %d): %s\n",
2324                         ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2325                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2326                 goto bad;
2327         }
2328
2329         /* RFC 2461 8.3 */
2330         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2331                          is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2332
2333         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2334                 /* perform rtredirect */
2335                 struct sockaddr_in6 sdst;
2336                 struct sockaddr_in6 sgw;
2337                 struct sockaddr_in6 ssrc;
2338
2339                 bzero(&sdst, sizeof(sdst));
2340                 bzero(&sgw, sizeof(sgw));
2341                 bzero(&ssrc, sizeof(ssrc));
2342                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2343                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2344                         sizeof(struct sockaddr_in6);
2345                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2346                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2347                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2348                 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2349                            (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2350                            (struct sockaddr *)&ssrc);
2351         }
2352         /* finally update cached route in each socket via pfctlinput */
2353     {
2354         struct sockaddr_in6 sdst;
2355
2356         bzero(&sdst, sizeof(sdst));
2357         sdst.sin6_family = AF_INET6;
2358         sdst.sin6_len = sizeof(struct sockaddr_in6);
2359         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2360         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2361 #ifdef IPSEC
2362         key_sa_routechange((struct sockaddr *)&sdst);
2363 #endif
2364     }
2365
2366  freeit:
2367         m_freem(m);
2368         return;
2369
2370  bad:
2371         icmp6stat.icp6s_badredirect++;
2372         m_freem(m);
2373 }
2374
2375 void
2376 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2377 {
2378         struct ifnet *ifp;      /* my outgoing interface */
2379         struct in6_addr *ifp_ll6;
2380         struct in6_addr *router_ll6;
2381         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2382         struct mbuf *m = NULL;  /* newly allocated one */
2383         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2384         struct nd_redirect *nd_rd;
2385         size_t maxlen;
2386         u_char *p;
2387         struct ifnet *outif = NULL;
2388         struct sockaddr_in6 src_sa;
2389
2390         icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2391
2392         /* if we are not router, we don't send icmp6 redirect */
2393         if (!ip6_forwarding || ip6_accept_rtadv)
2394                 goto fail;
2395
2396         /* sanity check */
2397         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2398                 goto fail;
2399
2400         /*
2401          * Address check:
2402          *  the source address must identify a neighbor, and
2403          *  the destination address must not be a multicast address
2404          *  [RFC 2461, sec 8.2]
2405          */
2406         sip6 = mtod(m0, struct ip6_hdr *);
2407         bzero(&src_sa, sizeof(src_sa));
2408         src_sa.sin6_family = AF_INET6;
2409         src_sa.sin6_len = sizeof(src_sa);
2410         src_sa.sin6_addr = sip6->ip6_src;
2411         /* we don't currently use sin6_scope_id, but eventually use it */
2412         src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2413         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2414                 goto fail;
2415         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2416                 goto fail;      /* what should we do here? */
2417
2418         /* rate limit */
2419         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2420                 goto fail;
2421
2422         /*
2423          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2424          * we almost always ask for an mbuf cluster for simplicity.
2425          * (MHLEN < IPV6_MMTU is almost always true)
2426          */
2427 #if IPV6_MMTU >= MCLBYTES
2428 # error assumption failed about IPV6_MMTU and MCLBYTES
2429 #endif
2430         MGETHDR(m, MB_DONTWAIT, MT_HEADER);
2431         if (m && IPV6_MMTU >= MHLEN)
2432                 MCLGET(m, MB_DONTWAIT);
2433         if (!m)
2434                 goto fail;
2435         m->m_pkthdr.rcvif = NULL;
2436         m->m_len = 0;
2437         maxlen = M_TRAILINGSPACE(m);
2438         maxlen = min(IPV6_MMTU, maxlen);
2439         /* just for safety */
2440         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2441             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2442                 goto fail;
2443         }
2444
2445         {
2446                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2447                 struct in6_ifaddr *ia;
2448                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2449                                                  IN6_IFF_NOTREADY|
2450                                                  IN6_IFF_ANYCAST)) == NULL)
2451                         goto fail;
2452                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2453         }
2454
2455         /* get ip6 linklocal address for the router. */
2456         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2457                 struct sockaddr_in6 *sin6;
2458                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2459                 router_ll6 = &sin6->sin6_addr;
2460                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2461                         router_ll6 = (struct in6_addr *)NULL;
2462         } else
2463                 router_ll6 = (struct in6_addr *)NULL;
2464
2465         /* ip6 */
2466         ip6 = mtod(m, struct ip6_hdr *);
2467         ip6->ip6_flow = 0;
2468         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2469         ip6->ip6_vfc |= IPV6_VERSION;
2470         /* ip6->ip6_plen will be set later */
2471         ip6->ip6_nxt = IPPROTO_ICMPV6;
2472         ip6->ip6_hlim = 255;
2473         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2474         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2475         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2476
2477         /* ND Redirect */
2478         nd_rd = (struct nd_redirect *)(ip6 + 1);
2479         nd_rd->nd_rd_type = ND_REDIRECT;
2480         nd_rd->nd_rd_code = 0;
2481         nd_rd->nd_rd_reserved = 0;
2482         if (rt->rt_flags & RTF_GATEWAY) {
2483                 /*
2484                  * nd_rd->nd_rd_target must be a link-local address in
2485                  * better router cases.
2486                  */
2487                 if (!router_ll6)
2488                         goto fail;
2489                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2490                       sizeof(nd_rd->nd_rd_target));
2491                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2492                       sizeof(nd_rd->nd_rd_dst));
2493         } else {
2494                 /* make sure redtgt == reddst */
2495                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2496                       sizeof(nd_rd->nd_rd_target));
2497                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2498                       sizeof(nd_rd->nd_rd_dst));
2499         }
2500
2501         p = (u_char *)(nd_rd + 1);
2502
2503         if (!router_ll6)
2504                 goto nolladdropt;
2505
2506     {
2507         /* target lladdr option */
2508         struct rtentry *rt_router = NULL;
2509         int len;
2510         struct sockaddr_dl *sdl;
2511         struct nd_opt_hdr *nd_opt;
2512         char *lladdr;
2513
2514         rt_router = nd6_lookup(router_ll6, 0, ifp);
2515         if (!rt_router)
2516                 goto nolladdropt;
2517         len = sizeof(*nd_opt) + ifp->if_addrlen;
2518         len = (len + 7) & ~7;   /* round by 8 */
2519         /* safety check */
2520         if (len + (p - (u_char *)ip6) > maxlen)
2521                 goto nolladdropt;
2522         if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2523             (rt_router->rt_flags & RTF_LLINFO) &&
2524             (rt_router->rt_gateway->sa_family == AF_LINK) &&
2525             (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2526             sdl->sdl_alen) {
2527                 nd_opt = (struct nd_opt_hdr *)p;
2528                 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2529                 nd_opt->nd_opt_len = len >> 3;
2530                 lladdr = (char *)(nd_opt + 1);
2531                 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2532                 p += len;
2533         }
2534     }
2535 nolladdropt:;
2536
2537         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2538
2539         /* just to be safe */
2540 #ifdef M_DECRYPTED      /*not openbsd*/
2541         if (m0->m_flags & M_DECRYPTED)
2542                 goto noredhdropt;
2543 #endif
2544         if (p - (u_char *)ip6 > maxlen)
2545                 goto noredhdropt;
2546
2547     {
2548         /* redirected header option */
2549         int len;
2550         struct nd_opt_rd_hdr *nd_opt_rh;
2551
2552         /*
2553          * compute the maximum size for icmp6 redirect header option.
2554          * XXX room for auth header?
2555          */
2556         len = maxlen - (p - (u_char *)ip6);
2557         len &= ~7;
2558
2559         /* This is just for simplicity. */
2560         if (m0->m_pkthdr.len != m0->m_len) {
2561                 if (m0->m_next) {
2562                         m_freem(m0->m_next);
2563                         m0->m_next = NULL;
2564                 }
2565                 m0->m_pkthdr.len = m0->m_len;
2566         }
2567
2568         /*
2569          * Redirected header option spec (RFC2461 4.6.3) talks nothing
2570          * about padding/truncate rule for the original IP packet.
2571          * From the discussion on IPv6imp in Feb 1999, the consensus was:
2572          * - "attach as much as possible" is the goal
2573          * - pad if not aligned (original size can be guessed by original
2574          *   ip6 header)
2575          * Following code adds the padding if it is simple enough,
2576          * and truncates if not.
2577          */
2578         if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2579                 panic("assumption failed in %s:%d", __FILE__, __LINE__);
2580
2581         if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2582                 /* not enough room, truncate */
2583                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2584         } else {
2585                 /* enough room, pad or truncate */
2586                 size_t extra;
2587
2588                 extra = m0->m_pkthdr.len % 8;
2589                 if (extra) {
2590                         /* pad if easy enough, truncate if not */
2591                         if (8 - extra <= M_TRAILINGSPACE(m0)) {
2592                                 /* pad */
2593                                 m0->m_len += (8 - extra);
2594                                 m0->m_pkthdr.len += (8 - extra);
2595                         } else {
2596                                 /* truncate */
2597                                 m0->m_pkthdr.len -= extra;
2598                                 m0->m_len -= extra;
2599                         }
2600                 }
2601                 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2602                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2603         }
2604
2605         nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2606         bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2607         nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2608         nd_opt_rh->nd_opt_rh_len = len >> 3;
2609         p += sizeof(*nd_opt_rh);
2610         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2611
2612         /* connect m0 to m */
2613         m->m_next = m0;
2614         m->m_pkthdr.len = m->m_len + m0->m_len;
2615     }
2616 noredhdropt:;
2617
2618         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2619                 sip6->ip6_src.s6_addr16[1] = 0;
2620         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2621                 sip6->ip6_dst.s6_addr16[1] = 0;
2622 #if 0
2623         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2624                 ip6->ip6_src.s6_addr16[1] = 0;
2625         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2626                 ip6->ip6_dst.s6_addr16[1] = 0;
2627 #endif
2628         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2629                 nd_rd->nd_rd_target.s6_addr16[1] = 0;
2630         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2631                 nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2632
2633         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2634
2635         nd_rd->nd_rd_cksum = 0;
2636         nd_rd->nd_rd_cksum
2637                 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2638
2639         /* send the packet to outside... */
2640         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2641         if (outif) {
2642                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2643                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2644         }
2645         icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2646
2647         return;
2648
2649 fail:
2650         if (m)
2651                 m_freem(m);
2652         if (m0)
2653                 m_freem(m0);
2654 }
2655
2656 #ifdef HAVE_NRL_INPCB
2657 #define in6pcb          inpcb
2658 #define in6p_icmp6filt  inp_icmp6filt
2659 #endif
2660 /*
2661  * ICMPv6 socket option processing.
2662  */
2663 int
2664 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2665 {
2666         int error = 0;
2667         int optlen;
2668         struct inpcb *inp = so->so_pcb;
2669         int level, op, optname;
2670
2671         if (sopt) {
2672                 level = sopt->sopt_level;
2673                 op = sopt->sopt_dir;
2674                 optname = sopt->sopt_name;
2675                 optlen = sopt->sopt_valsize;
2676         } else
2677                 level = op = optname = optlen = 0;
2678
2679         if (level != IPPROTO_ICMPV6) {
2680                 return EINVAL;
2681         }
2682
2683         switch (op) {
2684         case PRCO_SETOPT:
2685                 switch (optname) {
2686                 case ICMP6_FILTER:
2687                     {
2688                         struct icmp6_filter *p;
2689
2690                         if (optlen != sizeof(*p)) {
2691                                 error = EMSGSIZE;
2692                                 break;
2693                         }
2694                         if (inp->in6p_icmp6filt == NULL) {
2695                                 error = EINVAL;
2696                                 break;
2697                         }
2698                         error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2699                                 optlen);
2700                         break;
2701                     }
2702
2703                 default:
2704                         error = ENOPROTOOPT;
2705                         break;
2706                 }
2707                 break;
2708
2709         case PRCO_GETOPT:
2710                 switch (optname) {
2711                 case ICMP6_FILTER:
2712                     {
2713                         if (inp->in6p_icmp6filt == NULL) {
2714                                 error = EINVAL;
2715                                 break;
2716                         }
2717                         error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2718                                 sizeof(struct icmp6_filter));
2719                         break;
2720                     }
2721
2722                 default:
2723                         error = ENOPROTOOPT;
2724                         break;
2725                 }
2726                 break;
2727         }
2728
2729         return(error);
2730 }
2731 #ifdef HAVE_NRL_INPCB
2732 #undef in6pcb
2733 #undef in6p_icmp6filt
2734 #endif
2735
2736 #ifndef HAVE_PPSRATECHECK
2737 #ifndef timersub
2738 #define timersub(tvp, uvp, vvp)                                         \
2739         do {                                                            \
2740                 (vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;          \
2741                 (vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;       \
2742                 if ((vvp)->tv_usec < 0) {                               \
2743                         (vvp)->tv_sec--;                                \
2744                         (vvp)->tv_usec += 1000000;                      \
2745                 }                                                       \
2746         } while (0)
2747 #endif
2748
2749 /*
2750  * ppsratecheck(): packets (or events) per second limitation.
2751  */
2752 static int
2753 ppsratecheck(struct timeval *lasttime, int *curpps,
2754              int maxpps)        /* maximum pps allowed */
2755 {
2756         struct timeval tv, delta;
2757         int rv;
2758
2759         microtime(&tv);
2760
2761         timersub(&tv, lasttime, &delta);
2762
2763         /*
2764          * Check for 0,0 so that the message will be seen at least once.
2765          * If more than one second has passed since the last update of
2766          * lasttime, reset the counter.
2767          *
2768          * We do increment *curpps even in *curpps < maxpps case, as some may
2769          * try to use *curpps for stat purposes as well.
2770          */
2771         if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2772             delta.tv_sec >= 1) {
2773                 *lasttime = tv;
2774                 *curpps = 0;
2775                 rv = 1;
2776         } else if (maxpps < 0)
2777                 rv = 1;
2778         else if (*curpps < maxpps)
2779                 rv = 1;
2780         else
2781                 rv = 0;
2782
2783 #if 1 /* DIAGNOSTIC? */
2784         /* be careful about wrap-around */
2785         if (*curpps + 1 > *curpps)
2786                 *curpps = *curpps + 1;
2787 #else
2788         /*
2789          * assume that there's not too many calls to this function.
2790          * not sure if the assumption holds, as it depends on *caller's*
2791          * behavior, not the behavior of this function.
2792          * IMHO it is wrong to make assumption on the caller's behavior,
2793          * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2794          */
2795         *curpps = *curpps + 1;
2796 #endif
2797
2798         return (rv);
2799 }
2800 #endif
2801
2802 /*
2803  * Perform rate limit check.
2804  * Returns 0 if it is okay to send the icmp6 packet.
2805  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2806  * limitation.
2807  *
2808  * XXX per-destination/type check necessary?
2809  */
2810 static int
2811 icmp6_ratelimit(const struct in6_addr *dst,     /* not used at this moment */
2812                 const int type,                 /* not used at this moment */
2813                 const int code)                 /* not used at this moment */
2814 {
2815         int ret;
2816
2817         ret = 0;        /* okay to send */
2818
2819         /* PPS limit */
2820         if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2821             icmp6errppslim)) {
2822                 /* The packet is subject to rate limit */
2823                 ret++;
2824         }
2825
2826         return ret;
2827 }