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 1.24 2006/12/20 18:14:43 dillon 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) ||
 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                         n = m_getb(maxlen, MB_DONTWAIT, n0->m_type, M_PKTHDR);
 562                         if (n == NULL) {
 563                                 /* Give up remote */
 564                                 m_freem(n0);
 565                                 break;
 566                         }
 567                         n0len = n0->m_pkthdr.len;       /* save for use below */
 568                         M_MOVE_PKTHDR(n, n0);
 569                         /*
 570                          * Copy IPv6 and ICMPv6 only.
 571                          */
 572                         nip6 = mtod(n, struct ip6_hdr *);
 573                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 574                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 575                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 576                         noff = sizeof(struct ip6_hdr);
 577                         /* new mbuf contains only ipv6+icmpv6 headers */
 578                         n->m_len = noff + sizeof(struct icmp6_hdr);
 579                         /*
 580                          * Adjust mbuf. ip6_plen will be adjusted in
 581                          * ip6_output().
 582                          */
 583                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 584                         /* recalculate complete packet size */
 585                         n->m_pkthdr.len = n0len + (noff - off);
 586                         n->m_next = n0;
 587                 } else {
 588                         nip6 = mtod(n, struct ip6_hdr *);
 589                         nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
 590                         noff = off;
 591                 }
 592                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 593                 nicmp6->icmp6_code = 0;
 594                 if (n) {
 595                         icmp6stat.icp6s_reflect++;
 596                         icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
 597                         icmp6_reflect(n, noff);
 598                 }
 599                 break;
 600
 601         case ICMP6_ECHO_REPLY:
 602                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
 603                 if (code != 0)
 604                         goto badcode;
 605                 break;
 606
 607         case MLD_LISTENER_QUERY:
 608         case MLD_LISTENER_REPORT:
 609                 if (icmp6len < sizeof(struct mld_hdr))
 610                         goto badlen;
 611                 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
 612                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
 613                 else
 614                         icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
 615                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 616                         /* give up local */
 617                         mld6_input(m, off);
 618                         m = NULL;
 619                         goto freeit;
 620                 }
 621                 mld6_input(n, off);
 622                 /* m stays. */
 623                 break;
 624
 625         case MLD_LISTENER_DONE:
 626                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
 627                 if (icmp6len < sizeof(struct mld_hdr))  /* necessary? */
 628                         goto badlen;
 629                 break;          /* nothing to be done in kernel */
 630
 631         case MLD_MTRACE_RESP:
 632         case MLD_MTRACE:
 633                 /* XXX: these two are experimental.  not officially defind. */
 634                 /* XXX: per-interface statistics? */
 635                 break;          /* just pass it to applications */
 636
 637         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 638             {
 639                 enum { WRU, FQDN } mode;
 640
 641                 if (!icmp6_nodeinfo)
 642                         break;
 643
 644                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 645                         mode = WRU;
 646                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 647                         mode = FQDN;
 648                 else
 649                         goto badlen;
 650
 651 #define hostnamelen     strlen(hostname)
 652                 if (mode == FQDN) {
 653 #ifndef PULLDOWN_TEST
 654                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
 655                                          IPPROTO_DONE);
 656 #endif
 657                         n = m_copy(m, 0, M_COPYALL);
 658                         if (n)
 659                                 n = ni6_input(n, off);
 660                         /* XXX meaningless if n == NULL */
 661                         noff = sizeof(struct ip6_hdr);
 662                 } else {
 663                         u_char *p;
 664                         int maxlen, maxhlen;
 665
 666                         if ((icmp6_nodeinfo & 5) != 5)
 667                                 break;
 668
 669                         if (code != 0)
 670                                 goto badcode;
 671                         maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
 672                         if (maxlen >= MCLBYTES) {
 673                                 /* Give up remote */
 674                                 break;
 675                         }
 676                         n = m_getb(maxlen, MB_DONTWAIT, m->m_type, M_PKTHDR);
 677                         if (n == NULL) {
 678                                 /* Give up remote */
 679                                 break;
 680                         }
 681                         if (!m_dup_pkthdr(n, m, MB_DONTWAIT)) {
 682                                 /*
 683                                  * Previous code did a blind M_COPY_PKTHDR
 684                                  * and said "just for rcvif".  If true, then
 685                                  * we could tolerate the dup failing (due to
 686                                  * the deep copy of the tag chain).  For now
 687                                  * be conservative and just fail.
 688                                  */
 689                                 m_free(n);
 690                                 break;
 691                         }
 692                         n->m_len = 0;
 693                         maxhlen = M_TRAILINGSPACE(n) - maxlen;
 694                         if (maxhlen > hostnamelen)
 695                                 maxhlen = hostnamelen;
 696                         /*
 697                          * Copy IPv6 and ICMPv6 only.
 698                          */
 699                         nip6 = mtod(n, struct ip6_hdr *);
 700                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 701                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 702                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 703                         p = (u_char *)(nicmp6 + 1);
 704                         bzero(p, 4);
 705                         bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
 706                         noff = sizeof(struct ip6_hdr);
 707                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 708                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 709                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 710                         nicmp6->icmp6_code = 0;
 711                 }
 712 #undef hostnamelen
 713                 if (n) {
 714                         icmp6stat.icp6s_reflect++;
 715                         icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
 716                         icmp6_reflect(n, noff);
 717                 }
 718                 break;
 719             }
 720
 721         case ICMP6_WRUREPLY:
 722                 if (code != 0)
 723                         goto badcode;
 724                 break;
 725
 726         case ND_ROUTER_SOLICIT:
 727                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
 728                 if (code != 0)
 729                         goto badcode;
 730                 if (icmp6len < sizeof(struct nd_router_solicit))
 731                         goto badlen;
 732                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 733                         /* give up local */
 734                         nd6_rs_input(m, off, icmp6len);
 735                         m = NULL;
 736                         goto freeit;
 737                 }
 738                 nd6_rs_input(n, off, icmp6len);
 739                 /* m stays. */
 740                 break;
 741
 742         case ND_ROUTER_ADVERT:
 743                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
 744                 if (code != 0)
 745                         goto badcode;
 746                 if (icmp6len < sizeof(struct nd_router_advert))
 747                         goto badlen;
 748                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 749                         /* give up local */
 750                         nd6_ra_input(m, off, icmp6len);
 751                         m = NULL;
 752                         goto freeit;
 753                 }
 754                 nd6_ra_input(n, off, icmp6len);
 755                 /* m stays. */
 756                 break;
 757
 758         case ND_NEIGHBOR_SOLICIT:
 759                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
 760                 if (code != 0)
 761                         goto badcode;
 762                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 763                         goto badlen;
 764                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 765                         /* give up local */
 766                         nd6_ns_input(m, off, icmp6len);
 767                         m = NULL;
 768                         goto freeit;
 769                 }
 770                 nd6_ns_input(n, off, icmp6len);
 771                 /* m stays. */
 772                 break;
 773
 774         case ND_NEIGHBOR_ADVERT:
 775                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
 776                 if (code != 0)
 777                         goto badcode;
 778                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 779                         goto badlen;
 780                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 781                         /* give up local */
 782                         nd6_na_input(m, off, icmp6len);
 783                         m = NULL;
 784                         goto freeit;
 785                 }
 786                 nd6_na_input(n, off, icmp6len);
 787                 /* m stays. */
 788                 break;
 789
 790         case ND_REDIRECT:
 791                 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
 792                 if (code != 0)
 793                         goto badcode;
 794                 if (icmp6len < sizeof(struct nd_redirect))
 795                         goto badlen;
 796                 if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
 797                         /* give up local */
 798                         icmp6_redirect_input(m, off);
 799                         m = NULL;
 800                         goto freeit;
 801                 }
 802                 icmp6_redirect_input(n, off);
 803                 /* m stays. */
 804                 break;
 805
 806         case ICMP6_ROUTER_RENUMBERING:
 807                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 808                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 809                         goto badcode;
 810                 if (icmp6len < sizeof(struct icmp6_router_renum))
 811                         goto badlen;
 812                 break;
 813
 814         default:
 815                 nd6log((LOG_DEBUG,
 816                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 817                     icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
 818                     ip6_sprintf(&ip6->ip6_dst),
 819                     m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
 820                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 821                         /* ICMPv6 error: MUST deliver it by spec... */
 822                         code = PRC_NCMDS;
 823                         /* deliver */
 824                 } else {
 825                         /* ICMPv6 informational: MUST not deliver */
 826                         break;
 827                 }
 828         deliver:
 829                 if (icmp6_notify_error(m, off, icmp6len, code)) {
 830                         /* In this case, m should've been freed. */
 831                         return (IPPROTO_DONE);
 832                 }
 833                 break;
 834
 835         badcode:
 836                 icmp6stat.icp6s_badcode++;
 837                 break;
 838
 839         badlen:
 840                 icmp6stat.icp6s_badlen++;
 841                 break;
 842         }
 843
 844         /* deliver the packet to appropriate sockets */
 845         icmp6_rip6_input(&m, *offp);
 846
 847         return IPPROTO_DONE;
 848
 849 freeit:
 850         m_freem(m);
 851         return IPPROTO_DONE;
 852 }
 853
 854 static int
 855 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
 856 {
 857         struct icmp6_hdr *icmp6;
 858         struct ip6_hdr *eip6;
 859         u_int32_t notifymtu;
 860         struct sockaddr_in6 icmp6src, icmp6dst;
 861
 862         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 863                 icmp6stat.icp6s_tooshort++;
 864                 goto freeit;
 865         }
 866 #ifndef PULLDOWN_TEST
 867         IP6_EXTHDR_CHECK(m, off,
 868                          sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
 869                          -1);
 870         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 871 #else
 872         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
 873                        sizeof(*icmp6) + sizeof(struct ip6_hdr));
 874         if (icmp6 == NULL) {
 875                 icmp6stat.icp6s_tooshort++;
 876                 return (-1);
 877         }
 878 #endif
 879         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 880
 881         /* Detect the upper level protocol */
 882         {
 883                 void (*ctlfunc) (int, struct sockaddr *, void *);
 884                 u_int8_t nxt = eip6->ip6_nxt;
 885                 int eoff = off + sizeof(struct icmp6_hdr) +
 886                         sizeof(struct ip6_hdr);
 887                 struct ip6ctlparam ip6cp;
 888                 struct in6_addr *finaldst = NULL;
 889                 int icmp6type = icmp6->icmp6_type;
 890                 struct ip6_frag *fh;
 891                 struct ip6_rthdr *rth;
 892                 struct ip6_rthdr0 *rth0;
 893                 int rthlen;
 894
 895                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 896                         struct ip6_ext *eh;
 897
 898                         switch (nxt) {
 899                         case IPPROTO_HOPOPTS:
 900                         case IPPROTO_DSTOPTS:
 901                         case IPPROTO_AH:
 902 #ifndef PULLDOWN_TEST
 903                                 IP6_EXTHDR_CHECK(m, 0, eoff +
 904                                                  sizeof(struct ip6_ext),
 905                                                  -1);
 906                                 eh = (struct ip6_ext *)(mtod(m, caddr_t)
 907                                                         + eoff);
 908 #else
 909                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
 910                                                eoff, sizeof(*eh));
 911                                 if (eh == NULL) {
 912                                         icmp6stat.icp6s_tooshort++;
 913                                         return (-1);
 914                                 }
 915 #endif
 916
 917                                 if (nxt == IPPROTO_AH)
 918                                         eoff += (eh->ip6e_len + 2) << 2;
 919                                 else
 920                                         eoff += (eh->ip6e_len + 1) << 3;
 921                                 nxt = eh->ip6e_nxt;
 922                                 break;
 923                         case IPPROTO_ROUTING:
 924                                 /*
 925                                  * When the erroneous packet contains a
 926                                  * routing header, we should examine the
 927                                  * header to determine the final destination.
 928                                  * Otherwise, we can't properly update
 929                                  * information that depends on the final
 930                                  * destination (e.g. path MTU).
 931                                  */
 932 #ifndef PULLDOWN_TEST
 933                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
 934                                                  -1);
 935                                 rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
 936                                                            + eoff);
 937 #else
 938                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
 939                                                eoff, sizeof(*rth));
 940                                 if (rth == NULL) {
 941                                         icmp6stat.icp6s_tooshort++;
 942                                         return (-1);
 943                                 }
 944 #endif
 945                                 rthlen = (rth->ip6r_len + 1) << 3;
 946                                 /*
 947                                  * XXX: currently there is no
 948                                  * officially defined type other
 949                                  * than type-0.
 950                                  * Note that if the segment left field
 951                                  * is 0, all intermediate hops must
 952                                  * have been passed.
 953                                  */
 954                                 if (rth->ip6r_segleft &&
 955                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
 956                                         int hops;
 957
 958 #ifndef PULLDOWN_TEST
 959                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
 960                                                          -1);
 961                                         rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
 962 #else
 963                                         IP6_EXTHDR_GET(rth0,
 964                                                        struct ip6_rthdr0 *, m,
 965                                                        eoff, rthlen);
 966                                         if (rth0 == NULL) {
 967                                                 icmp6stat.icp6s_tooshort++;
 968                                                 return (-1);
 969                                         }
 970 #endif
 971                                         /* just ignore a bogus header */
 972                                         if ((rth0->ip6r0_len % 2) == 0 &&
 973                                             (hops = rth0->ip6r0_len/2))
 974                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
 975                                 }
 976                                 eoff += rthlen;
 977                                 nxt = rth->ip6r_nxt;
 978                                 break;
 979                         case IPPROTO_FRAGMENT:
 980 #ifndef PULLDOWN_TEST
 981                                 IP6_EXTHDR_CHECK(m, 0, eoff +
 982                                                  sizeof(struct ip6_frag),
 983                                                  -1);
 984                                 fh = (struct ip6_frag *)(mtod(m, caddr_t)
 985                                                          + eoff);
 986 #else
 987                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
 988                                                eoff, sizeof(*fh));
 989                                 if (fh == NULL) {
 990                                         icmp6stat.icp6s_tooshort++;
 991                                         return (-1);
 992                                 }
 993 #endif
 994                                 /*
 995                                  * Data after a fragment header is meaningless
 996                                  * unless it is the first fragment, but
 997                                  * we'll go to the notify label for path MTU
 998                                  * discovery.
 999                                  */
1000                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1001                                         goto notify;
1002
1003                                 eoff += sizeof(struct ip6_frag);
1004                                 nxt = fh->ip6f_nxt;
1005                                 break;
1006                         default:
1007                                 /*
1008                                  * This case includes ESP and the No Next
1009                                  * Header.  In such cases going to the notify
1010                                  * label does not have any meaning
1011                                  * (i.e. ctlfunc will be NULL), but we go
1012                                  * anyway since we might have to update
1013                                  * path MTU information.
1014                                  */
1015                                 goto notify;
1016                         }
1017                 }
1018           notify:
1019 #ifndef PULLDOWN_TEST
1020                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1021 #else
1022                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1023                                sizeof(*icmp6) + sizeof(struct ip6_hdr));
1024                 if (icmp6 == NULL) {
1025                         icmp6stat.icp6s_tooshort++;
1026                         return (-1);
1027                 }
1028 #endif
1029
1030                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1031                 bzero(&icmp6dst, sizeof(icmp6dst));
1032                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1033                 icmp6dst.sin6_family = AF_INET6;
1034                 if (finaldst == NULL)
1035                         icmp6dst.sin6_addr = eip6->ip6_dst;
1036                 else
1037                         icmp6dst.sin6_addr = *finaldst;
1038                 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1039                                                           &icmp6dst.sin6_addr);
1040                 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1041                                    NULL, NULL)) {
1042                         /* should be impossbile */
1043                         nd6log((LOG_DEBUG,
1044                             "icmp6_notify_error: in6_embedscope failed\n"));
1045                         goto freeit;
1046                 }
1047
1048                 /*
1049                  * retrieve parameters from the inner IPv6 header, and convert
1050                  * them into sockaddr structures.
1051                  */
1052                 bzero(&icmp6src, sizeof(icmp6src));
1053                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1054                 icmp6src.sin6_family = AF_INET6;
1055                 icmp6src.sin6_addr = eip6->ip6_src;
1056                 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1057                                                           &icmp6src.sin6_addr);
1058                 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1059                                    NULL, NULL)) {
1060                         /* should be impossbile */
1061                         nd6log((LOG_DEBUG,
1062                             "icmp6_notify_error: in6_embedscope failed\n"));
1063                         goto freeit;
1064                 }
1065                 icmp6src.sin6_flowinfo =
1066                         (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1067
1068                 if (finaldst == NULL)
1069                         finaldst = &eip6->ip6_dst;
1070                 ip6cp.ip6c_m = m;
1071                 ip6cp.ip6c_icmp6 = icmp6;
1072                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1073                 ip6cp.ip6c_off = eoff;
1074                 ip6cp.ip6c_finaldst = finaldst;
1075                 ip6cp.ip6c_src = &icmp6src;
1076                 ip6cp.ip6c_nxt = nxt;
1077
1078                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1079                         notifymtu = ntohl(icmp6->icmp6_mtu);
1080                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1081                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1082                 }
1083
1084                 ctlfunc = (void (*) (int, struct sockaddr *, void *))
1085                         (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1086                 if (ctlfunc) {
1087                         (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, &ip6cp);
1088                 }
1089         }
1090         return (0);
1091
1092 freeit:
1093         m_freem(m);
1094         return (-1);
1095 }
1096
1097 void
1098 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1099 {
1100         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1101         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1102         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1103         u_int mtu = ntohl(icmp6->icmp6_mtu);
1104         struct rtentry *rt = NULL;
1105         struct sockaddr_in6 sin6;
1106
1107         if (!validated)
1108                 return;
1109
1110         bzero(&sin6, sizeof(sin6));
1111         sin6.sin6_family = PF_INET6;
1112         sin6.sin6_len = sizeof(struct sockaddr_in6);
1113         sin6.sin6_addr = *dst;
1114         /* XXX normally, this won't happen */
1115         if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1116                 sin6.sin6_addr.s6_addr16[1] =
1117                     htons(m->m_pkthdr.rcvif->if_index);
1118         }
1119         /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1120         rt = rtpurelookup((struct sockaddr *)&sin6);
1121         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1122             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1123                 if (mtu < IPV6_MMTU) {                          /* XXX */
1124                         rt->rt_rmx.rmx_locks |= RTV_MTU;
1125                 } else if (mtu < rt->rt_ifp->if_mtu &&
1126                            rt->rt_rmx.rmx_mtu > mtu) {
1127                         icmp6stat.icp6s_pmtuchg++;
1128                         rt->rt_rmx.rmx_mtu = mtu;
1129                 }
1130         }
1131         if (rt)
1132                 --rt->rt_refcnt;
1133 }
1134
1135 /*
1136  * Process a Node Information Query packet, based on
1137  * draft-ietf-ipngwg-icmp-name-lookups-07.
1138  *
1139  * Spec incompatibilities:
1140  * - IPv6 Subject address handling
1141  * - IPv4 Subject address handling support missing
1142  * - Proxy reply (answer even if it's not for me)
1143  * - joins NI group address at in6_ifattach() time only, does not cope
1144  *   with hostname changes by sethostname(3)
1145  */
1146 #define hostnamelen     strlen(hostname)
1147 static struct mbuf *
1148 ni6_input(struct mbuf *m, int off)
1149 {
1150         struct icmp6_nodeinfo *ni6, *nni6;
1151         struct mbuf *n = NULL;
1152         u_int16_t qtype;
1153         int subjlen;
1154         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1155         struct ni_reply_fqdn *fqdn;
1156         int addrs;              /* for NI_QTYPE_NODEADDR */
1157         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1158         struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1159         struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1160         struct ip6_hdr *ip6;
1161         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1162         char *subj = NULL;
1163         struct in6_ifaddr *ia6 = NULL;
1164
1165         ip6 = mtod(m, struct ip6_hdr *);
1166 #ifndef PULLDOWN_TEST
1167         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1168 #else
1169         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1170         if (ni6 == NULL) {
1171                 /* m is already reclaimed */
1172                 return NULL;
1173         }
1174 #endif
1175
1176         /*
1177          * Validate IPv6 destination address.
1178          *
1179          * The Responder must discard the Query without further processing
1180          * unless it is one of the Responder's unicast or anycast addresses, or
1181          * a link-local scope multicast address which the Responder has joined.
1182          * [icmp-name-lookups-07, Section 4.]
1183          */
1184         bzero(&sin6, sizeof(sin6));
1185         sin6.sin6_family = AF_INET6;
1186         sin6.sin6_len = sizeof(struct sockaddr_in6);
1187         bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1188         /* XXX scopeid */
1189         if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1190                 /* unicast/anycast, fine */
1191                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1192                     (icmp6_nodeinfo & 4) == 0) {
1193                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1194                                 "a temporary address in %s:%d",
1195                                __FILE__, __LINE__));
1196                         goto bad;
1197                 }
1198         } else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1199                 ; /* link-local multicast, fine */
1200         else
1201                 goto bad;
1202
1203         /* validate query Subject field. */
1204         qtype = ntohs(ni6->ni_qtype);
1205         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1206         switch (qtype) {
1207         case NI_QTYPE_NOOP:
1208         case NI_QTYPE_SUPTYPES:
1209                 /* 07 draft */
1210                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1211                         break;
1212                 /* FALLTHROUGH */
1213         case NI_QTYPE_FQDN:
1214         case NI_QTYPE_NODEADDR:
1215                 switch (ni6->ni_code) {
1216                 case ICMP6_NI_SUBJ_IPV6:
1217 #if ICMP6_NI_SUBJ_IPV6 != 0
1218                 case 0:
1219 #endif
1220                         /*
1221                          * backward compatibility - try to accept 03 draft
1222                          * format, where no Subject is present.
1223                          */
1224                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1225                             subjlen == 0) {
1226                                 oldfqdn++;
1227                                 break;
1228                         }
1229 #if ICMP6_NI_SUBJ_IPV6 != 0
1230                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1231                                 goto bad;
1232 #endif
1233
1234                         if (subjlen != sizeof(sin6.sin6_addr))
1235                                 goto bad;
1236
1237                         /*
1238                          * Validate Subject address.
1239                          *
1240                          * Not sure what exactly "address belongs to the node"
1241                          * means in the spec, is it just unicast, or what?
1242                          *
1243                          * At this moment we consider Subject address as
1244                          * "belong to the node" if the Subject address equals
1245                          * to the IPv6 destination address; validation for
1246                          * IPv6 destination address should have done enough
1247                          * check for us.
1248                          *
1249                          * We do not do proxy at this moment.
1250                          */
1251                         /* m_pulldown instead of copy? */
1252                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1253                             subjlen, (caddr_t)&sin6.sin6_addr);
1254                         sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1255                                                               &sin6.sin6_addr);
1256                         in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1257                         bzero(&sin6_d, sizeof(sin6_d));
1258                         sin6_d.sin6_family = AF_INET6; /* not used, actually */
1259                         sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1260                         sin6_d.sin6_addr = ip6->ip6_dst;
1261                         sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1262                                                                 &ip6->ip6_dst);
1263                         in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1264                         subj = (char *)&sin6;
1265                         if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1266                                 break;
1267
1268                         /*
1269                          * XXX if we are to allow other cases, we should really
1270                          * be careful about scope here.
1271                          * basically, we should disallow queries toward IPv6
1272                          * destination X with subject Y, if scope(X) > scope(Y).
1273                          * if we allow scope(X) > scope(Y), it will result in
1274                          * information leakage across scope boundary.
1275                          */
1276                         goto bad;
1277
1278                 case ICMP6_NI_SUBJ_FQDN:
1279                         /*
1280                          * Validate Subject name with gethostname(3).
1281                          *
1282                          * The behavior may need some debate, since:
1283                          * - we are not sure if the node has FQDN as
1284                          *   hostname (returned by gethostname(3)).
1285                          * - the code does wildcard match for truncated names.
1286                          *   however, we are not sure if we want to perform
1287                          *   wildcard match, if gethostname(3) side has
1288                          *   truncated hostname.
1289                          */
1290                         n = ni6_nametodns(hostname, hostnamelen, 0);
1291                         if (!n || n->m_next || n->m_len == 0)
1292                                 goto bad;
1293                         IP6_EXTHDR_GET(subj, char *, m,
1294                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1295                         if (subj == NULL)
1296                                 goto bad;
1297                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1298                                         n->m_len)) {
1299                                 goto bad;
1300                         }
1301                         m_freem(n);
1302                         n = NULL;
1303                         break;
1304
1305                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1306                 default:
1307                         goto bad;
1308                 }
1309                 break;
1310         }
1311
1312         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1313         switch (qtype) {
1314         case NI_QTYPE_FQDN:
1315                 if ((icmp6_nodeinfo & 1) == 0)
1316                         goto bad;
1317                 break;
1318         case NI_QTYPE_NODEADDR:
1319                 if ((icmp6_nodeinfo & 2) == 0)
1320                         goto bad;
1321                 break;
1322         }
1323
1324         /* guess reply length */
1325         switch (qtype) {
1326         case NI_QTYPE_NOOP:
1327                 break;          /* no reply data */
1328         case NI_QTYPE_SUPTYPES:
1329                 replylen += sizeof(u_int32_t);
1330                 break;
1331         case NI_QTYPE_FQDN:
1332                 /* XXX will append an mbuf */
1333                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1334                 break;
1335         case NI_QTYPE_NODEADDR:
1336                 addrs = ni6_addrs(ni6, m, &ifp, subj);
1337                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1338                                           sizeof(u_int32_t))) > MCLBYTES)
1339                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1340                 break;
1341         default:
1342                 /*
1343                  * XXX: We must return a reply with the ICMP6 code
1344                  * `unknown Qtype' in this case. However we regard the case
1345                  * as an FQDN query for backward compatibility.
1346                  * Older versions set a random value to this field,
1347                  * so it rarely varies in the defined qtypes.
1348                  * But the mechanism is not reliable...
1349                  * maybe we should obsolete older versions.
1350                  */
1351                 qtype = NI_QTYPE_FQDN;
1352                 /* XXX will append an mbuf */
1353                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1354                 oldfqdn++;
1355                 break;
1356         }
1357
1358         /* allocate an mbuf to reply. */
1359         if (replylen > MCLBYTES) {
1360                 /*
1361                  * XXX: should we try to allocate more? But MCLBYTES
1362                  * is probably much larger than IPV6_MMTU...
1363                  */
1364                 goto bad;
1365         }
1366         n = m_getb(replylen, MB_DONTWAIT, m->m_type, M_PKTHDR);
1367         if (n == NULL) {
1368                 m_freem(m);
1369                 return (NULL);
1370         }
1371         M_MOVE_PKTHDR(n, m); /* just for recvif */
1372         n->m_pkthdr.len = n->m_len = replylen;
1373
1374         /* copy mbuf header and IPv6 + Node Information base headers */
1375         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1376         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1377         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1378
1379         /* qtype dependent procedure */
1380         switch (qtype) {
1381         case NI_QTYPE_NOOP:
1382                 nni6->ni_code = ICMP6_NI_SUCCESS;
1383                 nni6->ni_flags = 0;
1384                 break;
1385         case NI_QTYPE_SUPTYPES:
1386         {
1387                 u_int32_t v;
1388                 nni6->ni_code = ICMP6_NI_SUCCESS;
1389                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1390                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1391                 v = (u_int32_t)htonl(0x0000000f);
1392                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1393                 break;
1394         }
1395         case NI_QTYPE_FQDN:
1396                 nni6->ni_code = ICMP6_NI_SUCCESS;
1397                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1398                                                 sizeof(struct ip6_hdr) +
1399                                                 sizeof(struct icmp6_nodeinfo));
1400                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1401                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1402                 /*
1403                  * XXX do we really have FQDN in variable "hostname"?
1404                  */
1405                 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1406                 if (n->m_next == NULL)
1407                         goto bad;
1408                 /* XXX we assume that n->m_next is not a chain */
1409                 if (n->m_next->m_next != NULL)
1410                         goto bad;
1411                 n->m_pkthdr.len += n->m_next->m_len;
1412                 break;
1413         case NI_QTYPE_NODEADDR:
1414         {
1415                 int lenlim, copied;
1416
1417                 nni6->ni_code = ICMP6_NI_SUCCESS;
1418                 n->m_pkthdr.len = n->m_len =
1419                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1420                 lenlim = M_TRAILINGSPACE(n);
1421                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1422                 /* XXX: reset mbuf length */
1423                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1424                         sizeof(struct icmp6_nodeinfo) + copied;
1425                 break;
1426         }
1427         default:
1428                 break;          /* XXX impossible! */
1429         }
1430
1431         nni6->ni_type = ICMP6_NI_REPLY;
1432         m_freem(m);
1433         return (n);
1434
1435 bad:
1436         m_freem(m);
1437         if (n)
1438                 m_freem(n);
1439         return (NULL);
1440 }
1441 #undef hostnamelen
1442
1443 /*
1444  * make a mbuf with DNS-encoded string.  no compression support.
1445  *
1446  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1447  * treated as truncated name (two \0 at the end).  this is a wild guess.
1448  */
1449 static struct mbuf *
1450 ni6_nametodns(const char *name, int namelen,
1451               int old) /* return pascal string if non-zero */
1452 {
1453         struct mbuf *m;
1454         char *cp, *ep;
1455         const char *p, *q;
1456         int i, len, nterm;
1457
1458         if (old)
1459                 len = namelen + 1;
1460         else
1461                 len = MCLBYTES;
1462
1463         /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1464         m = m_getb(len, MB_DONTWAIT, MT_DATA, 0);
1465         if (!m)
1466                 goto fail;
1467
1468         if (old) {
1469                 m->m_len = len;
1470                 *mtod(m, char *) = namelen;
1471                 bcopy(name, mtod(m, char *) + 1, namelen);
1472                 return m;
1473         } else {
1474                 m->m_len = 0;
1475                 cp = mtod(m, char *);
1476                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1477
1478                 /* if not certain about my name, return empty buffer */
1479                 if (namelen == 0)
1480                         return m;
1481
1482                 /*
1483                  * guess if it looks like shortened hostname, or FQDN.
1484                  * shortened hostname needs two trailing "\0".
1485                  */
1486                 i = 0;
1487                 for (p = name; p < name + namelen; p++) {
1488                         if (*p && *p == '.')
1489                                 i++;
1490                 }
1491                 if (i < 2)
1492                         nterm = 2;
1493                 else
1494                         nterm = 1;
1495
1496                 p = name;
1497                 while (cp < ep && p < name + namelen) {
1498                         i = 0;
1499                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1500                                 i++;
1501                         /* result does not fit into mbuf */
1502                         if (cp + i + 1 >= ep)
1503                                 goto fail;
1504                         /*
1505                          * DNS label length restriction, RFC1035 page 8.
1506                          * "i == 0" case is included here to avoid returning
1507                          * 0-length label on "foo..bar".
1508                          */
1509                         if (i <= 0 || i >= 64)
1510                                 goto fail;
1511                         *cp++ = i;
1512                         bcopy(p, cp, i);
1513                         cp += i;
1514                         p = q;
1515                         if (p < name + namelen && *p == '.')
1516                                 p++;
1517                 }
1518                 /* termination */
1519                 if (cp + nterm >= ep)
1520                         goto fail;
1521                 while (nterm-- > 0)
1522                         *cp++ = '\0';
1523                 m->m_len = cp - mtod(m, char *);
1524                 return m;
1525         }
1526
1527         panic("should not reach here");
1528         /* NOTREACHED */
1529
1530 fail:
1531         if (m)
1532                 m_freem(m);
1533         return NULL;
1534 }
1535
1536 /*
1537  * check if two DNS-encoded string matches.  takes care of truncated
1538  * form (with \0\0 at the end).  no compression support.
1539  * XXX upper/lowercase match (see RFC2065)
1540  */
1541 static int
1542 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1543 {
1544         const char *a0, *b0;
1545         int l;
1546
1547         /* simplest case - need validation? */
1548         if (alen == blen && bcmp(a, b, alen) == 0)
1549                 return 1;
1550
1551         a0 = a;
1552         b0 = b;
1553
1554         /* termination is mandatory */
1555         if (alen < 2 || blen < 2)
1556                 return 0;
1557         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1558                 return 0;
1559         alen--;
1560         blen--;
1561
1562         while (a - a0 < alen && b - b0 < blen) {
1563                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1564                         return 0;
1565
1566                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1567                         return 0;
1568                 /* we don't support compression yet */
1569                 if (a[0] >= 64 || b[0] >= 64)
1570                         return 0;
1571
1572                 /* truncated case */
1573                 if (a[0] == 0 && a - a0 == alen - 1)
1574                         return 1;
1575                 if (b[0] == 0 && b - b0 == blen - 1)
1576                         return 1;
1577                 if (a[0] == 0 || b[0] == 0)
1578                         return 0;
1579
1580                 if (a[0] != b[0])
1581                         return 0;
1582                 l = a[0];
1583                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1584                         return 0;
1585                 if (bcmp(a + 1, b + 1, l) != 0)
1586                         return 0;
1587
1588                 a += 1 + l;
1589                 b += 1 + l;
1590         }
1591
1592         if (a - a0 == alen && b - b0 == blen)
1593                 return 1;
1594         else
1595                 return 0;
1596 }
1597
1598 /*
1599  * calculate the number of addresses to be returned in the node info reply.
1600  */
1601 static int
1602 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1603           char *subj)
1604 {
1605         struct ifnet *ifp;
1606         struct in6_ifaddr *ifa6;
1607         struct ifaddr *ifa;
1608         struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1609         int addrs = 0, addrsofif, iffound = 0;
1610         int niflags = ni6->ni_flags;
1611
1612         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1613                 switch (ni6->ni_code) {
1614                 case ICMP6_NI_SUBJ_IPV6:
1615                         if (subj == NULL) /* must be impossible... */
1616                                 return (0);
1617                         subj_ip6 = (struct sockaddr_in6 *)subj;
1618                         break;
1619                 default:
1620                         /*
1621                          * XXX: we only support IPv6 subject address for
1622                          * this Qtype.
1623                          */
1624                         return (0);
1625                 }
1626         }
1627
1628         for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1629         {
1630                 addrsofif = 0;
1631                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1632                 {
1633                         if (ifa->ifa_addr->sa_family != AF_INET6)
1634                                 continue;
1635                         ifa6 = (struct in6_ifaddr *)ifa;
1636
1637                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1638                             IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1639                                                &ifa6->ia_addr.sin6_addr))
1640                                 iffound = 1;
1641
1642                         /*
1643                          * IPv4-mapped addresses can only be returned by a
1644                          * Node Information proxy, since they represent
1645                          * addresses of IPv4-only nodes, which perforce do
1646                          * not implement this protocol.
1647                          * [icmp-name-lookups-07, Section 5.4]
1648                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1649                          * this function at this moment.
1650                          */
1651
1652                         /* What do we have to do about ::1? */
1653                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1654                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1655                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1656                                         continue;
1657                                 break;
1658                         case IPV6_ADDR_SCOPE_SITELOCAL:
1659                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1660                                         continue;
1661                                 break;
1662                         case IPV6_ADDR_SCOPE_GLOBAL:
1663                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1664                                         continue;
1665                                 break;
1666                         default:
1667                                 continue;
1668                         }
1669
1670                         /*
1671                          * check if anycast is okay.
1672                          * XXX: just experimental.  not in the spec.
1673                          */
1674                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1675                             !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1676                                 continue; /* we need only unicast addresses */
1677                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1678                             (icmp6_nodeinfo & 4) == 0) {
1679                                 continue;
1680                         }
1681                         addrsofif++; /* count the address */
1682                 }
1683                 if (iffound) {
1684                         *ifpp = ifp;
1685                         return (addrsofif);
1686                 }
1687
1688                 addrs += addrsofif;
1689         }
1690
1691         return (addrs);
1692 }
1693
1694 static int
1695 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1696                 struct ifnet *ifp0, int resid)
1697 {
1698         struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1699         struct in6_ifaddr *ifa6;
1700         struct ifaddr *ifa;
1701         struct ifnet *ifp_dep = NULL;
1702         int copied = 0, allow_deprecated = 0;
1703         u_char *cp = (u_char *)(nni6 + 1);
1704         int niflags = ni6->ni_flags;
1705         u_int32_t ltime;
1706
1707         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1708                 return (0);     /* needless to copy */
1709
1710 again:
1711
1712         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1713         {
1714                 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1715                         if (ifa->ifa_addr->sa_family != AF_INET6)
1716                                 continue;
1717                         ifa6 = (struct in6_ifaddr *)ifa;
1718
1719                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1720                             allow_deprecated == 0) {
1721                                 /*
1722                                  * prefererred address should be put before
1723                                  * deprecated addresses.
1724                                  */
1725
1726                                 /* record the interface for later search */
1727                                 if (ifp_dep == NULL)
1728                                         ifp_dep = ifp;
1729
1730                                 continue;
1731                         }
1732                         else if (!(ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1733                                  allow_deprecated != 0)
1734                                 continue; /* we now collect deprecated addrs */
1735
1736                         /* What do we have to do about ::1? */
1737                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1738                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1739                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1740                                         continue;
1741                                 break;
1742                         case IPV6_ADDR_SCOPE_SITELOCAL:
1743                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1744                                         continue;
1745                                 break;
1746                         case IPV6_ADDR_SCOPE_GLOBAL:
1747                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1748                                         continue;
1749                                 break;
1750                         default:
1751                                 continue;
1752                         }
1753
1754                         /*
1755                          * check if anycast is okay.
1756                          * XXX: just experimental. not in the spec.
1757                          */
1758                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1759                             !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1760                                 continue;
1761                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1762                             (icmp6_nodeinfo & 4) == 0) {
1763                                 continue;
1764                         }
1765
1766                         /* now we can copy the address */
1767                         if (resid < sizeof(struct in6_addr) +
1768                             sizeof(u_int32_t)) {
1769                                 /*
1770                                  * We give up much more copy.
1771                                  * Set the truncate flag and return.
1772                                  */
1773                                 nni6->ni_flags |=
1774                                         NI_NODEADDR_FLAG_TRUNCATE;
1775                                 return (copied);
1776                         }
1777
1778                         /*
1779                          * Set the TTL of the address.
1780                          * The TTL value should be one of the following
1781                          * according to the specification:
1782                          *
1783                          * 1. The remaining lifetime of a DHCP lease on the
1784                          *    address, or
1785                          * 2. The remaining Valid Lifetime of a prefix from
1786                          *    which the address was derived through Stateless
1787                          *    Autoconfiguration.
1788                          *
1789                          * Note that we currently do not support stateful
1790                          * address configuration by DHCPv6, so the former
1791                          * case can't happen.
1792                          */
1793                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1794                                 ltime = ND6_INFINITE_LIFETIME;
1795                         else {
1796                                 if (ifa6->ia6_lifetime.ia6t_expire >
1797                                     time_second)
1798                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1799                                 else
1800                                         ltime = 0;
1801                         }
1802
1803                         bcopy(&ltime, cp, sizeof(u_int32_t));
1804                         cp += sizeof(u_int32_t);
1805
1806                         /* copy the address itself */
1807                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1808                               sizeof(struct in6_addr));
1809                         /* XXX: KAME link-local hack; remove ifindex */
1810                         if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1811                                 ((struct in6_addr *)cp)->s6_addr16[1] = 0;
1812                         cp += sizeof(struct in6_addr);
1813
1814                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1815                         copied += (sizeof(struct in6_addr) +
1816                                    sizeof(u_int32_t));
1817                 }
1818                 if (ifp0)       /* we need search only on the specified IF */
1819                         break;
1820         }
1821
1822         if (allow_deprecated == 0 && ifp_dep != NULL) {
1823                 ifp = ifp_dep;
1824                 allow_deprecated = 1;
1825
1826                 goto again;
1827         }
1828
1829         return (copied);
1830 }
1831
1832 /*
1833  * XXX almost dup'ed code with rip6_input.
1834  */
1835 static int
1836 icmp6_rip6_input(struct mbuf **mp, int  off)
1837 {
1838         struct mbuf *m = *mp;
1839         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1840         struct in6pcb *in6p;
1841         struct in6pcb *last = NULL;
1842         struct sockaddr_in6 rip6src;
1843         struct icmp6_hdr *icmp6;
1844         struct mbuf *opts = NULL;
1845
1846 #ifndef PULLDOWN_TEST
1847         /* this is assumed to be safe. */
1848         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1849 #else
1850         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1851         if (icmp6 == NULL) {
1852                 /* m is already reclaimed */
1853                 return IPPROTO_DONE;
1854         }
1855 #endif
1856
1857         bzero(&rip6src, sizeof(rip6src));
1858         rip6src.sin6_len = sizeof(struct sockaddr_in6);
1859         rip6src.sin6_family = AF_INET6;
1860         /* KAME hack: recover scopeid */
1861         in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1862
1863         LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1864         {
1865                 if (in6p->inp_flags & INP_PLACEMARKER)
1866                         continue;
1867                 if ((in6p->inp_vflag & INP_IPV6) == 0)
1868                         continue;
1869 #ifdef HAVE_NRL_INPCB
1870                 if (!(in6p->in6p_flags & INP_IPV6))
1871                         continue;
1872 #endif
1873                 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1874                         continue;
1875                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1876                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1877                         continue;
1878                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1879                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1880                         continue;
1881                 if (in6p->in6p_icmp6filt
1882                     && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1883                                  in6p->in6p_icmp6filt))
1884                         continue;
1885                 if (last) {
1886                         struct  mbuf *n;
1887                         if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1888                                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1889                                         ip6_savecontrol(last, &opts, ip6, n);
1890                                 /* strip intermediate headers */
1891                                 m_adj(n, off);
1892                                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1893                                                  (struct sockaddr *)&rip6src,
1894                                                  n, opts) == 0) {
1895                                         /* should notify about lost packet */
1896                                         m_freem(n);
1897                                         if (opts) {
1898                                                 m_freem(opts);
1899                                         }
1900                                 } else
1901                                         sorwakeup(last->in6p_socket);
1902                                 opts = NULL;
1903                         }
1904                 }
1905                 last = in6p;
1906         }
1907         if (last) {
1908                 if (last->in6p_flags & IN6P_CONTROLOPTS)
1909                         ip6_savecontrol(last, &opts, ip6, m);
1910                 /* strip intermediate headers */
1911                 m_adj(m, off);
1912                 if (sbappendaddr(&last->in6p_socket->so_rcv,
1913                                  (struct sockaddr *)&rip6src, m, opts) == 0) {
1914                         m_freem(m);
1915                         if (opts)
1916                                 m_freem(opts);
1917                 } else
1918                         sorwakeup(last->in6p_socket);
1919         } else {
1920                 m_freem(m);
1921                 ip6stat.ip6s_delivered--;
1922         }
1923         return IPPROTO_DONE;
1924 }
1925
1926 /*
1927  * Reflect the ip6 packet back to the source.
1928  * OFF points to the icmp6 header, counted from the top of the mbuf.
1929  */
1930 void
1931 icmp6_reflect(struct mbuf *m, size_t off)
1932 {
1933         struct ip6_hdr *ip6;
1934         struct icmp6_hdr *icmp6;
1935         struct in6_ifaddr *ia;
1936         struct in6_addr t, *src = 0;
1937         int plen;
1938         int type, code;
1939         struct ifnet *outif = NULL;
1940         struct sockaddr_in6 sa6_src, sa6_dst;
1941 #ifdef COMPAT_RFC1885
1942         int mtu = IPV6_MMTU;
1943         struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1944 #endif
1945
1946         /* too short to reflect */
1947         if (off < sizeof(struct ip6_hdr)) {
1948                 nd6log((LOG_DEBUG,
1949                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1950                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
1951                     __FILE__, __LINE__));
1952                 goto bad;
1953         }
1954
1955         /*
1956          * If there are extra headers between IPv6 and ICMPv6, strip
1957          * off that header first.
1958          */
1959 #ifdef DIAGNOSTIC
1960         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1961                 panic("assumption failed in icmp6_reflect");
1962 #endif
1963         if (off > sizeof(struct ip6_hdr)) {
1964                 size_t l;
1965                 struct ip6_hdr nip6;
1966
1967                 l = off - sizeof(struct ip6_hdr);
1968                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1969                 m_adj(m, l);
1970                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1971                 if (m->m_len < l) {
1972                         if ((m = m_pullup(m, l)) == NULL)
1973                                 return;
1974                 }
1975                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1976         } else /* off == sizeof(struct ip6_hdr) */ {
1977                 size_t l;
1978                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1979                 if (m->m_len < l) {
1980                         if ((m = m_pullup(m, l)) == NULL)
1981                                 return;
1982                 }
1983         }
1984         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1985         ip6 = mtod(m, struct ip6_hdr *);
1986         ip6->ip6_nxt = IPPROTO_ICMPV6;
1987         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1988         type = icmp6->icmp6_type; /* keep type for statistics */
1989         code = icmp6->icmp6_code; /* ditto. */
1990
1991         t = ip6->ip6_dst;
1992         /*
1993          * ip6_input() drops a packet if its src is multicast.
1994          * So, the src is never multicast.
1995          */
1996         ip6->ip6_dst = ip6->ip6_src;
1997
1998         /*
1999          * XXX: make sure to embed scope zone information, using
2000          * already embedded IDs or the received interface (if any).
2001          * Note that rcvif may be NULL.
2002          * TODO: scoped routing case (XXX).
2003          */
2004         bzero(&sa6_src, sizeof(sa6_src));
2005         sa6_src.sin6_family = AF_INET6;
2006         sa6_src.sin6_len = sizeof(sa6_src);
2007         sa6_src.sin6_addr = ip6->ip6_dst;
2008         in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2009         in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2010         bzero(&sa6_dst, sizeof(sa6_dst));
2011         sa6_dst.sin6_family = AF_INET6;
2012         sa6_dst.sin6_len = sizeof(sa6_dst);
2013         sa6_dst.sin6_addr = t;
2014         in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2015         in6_embedscope(&t, &sa6_dst, NULL, NULL);
2016
2017 #ifdef COMPAT_RFC1885
2018         /*
2019          * xxx guess MTU
2020          * RFC 1885 requires that echo reply should be truncated if it
2021          * does not fit in with (return) path MTU, but the description was
2022          * removed in the new spec.
2023          */
2024         if (icmp6_reflect_rt.ro_rt == 0 ||
2025             ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2026                 if (icmp6_reflect_rt.ro_rt) {
2027                         RTFREE(icmp6_reflect_rt.ro_rt);
2028                         icmp6_reflect_rt.ro_rt = 0;
2029                 }
2030                 bzero(sin6, sizeof(*sin6));
2031                 sin6->sin6_family = PF_INET6;
2032                 sin6->sin6_len = sizeof(struct sockaddr_in6);
2033                 sin6->sin6_addr = ip6->ip6_dst;
2034
2035                 rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2036                             RTF_PRCLONING);
2037         }
2038
2039         if (icmp6_reflect_rt.ro_rt == 0)
2040                 goto bad;
2041
2042         if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2043             && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2044                 mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2045
2046         if (mtu < m->m_pkthdr.len) {
2047                 plen -= (m->m_pkthdr.len - mtu);
2048                 m_adj(m, mtu - m->m_pkthdr.len);
2049         }
2050 #endif
2051         /*
2052          * If the incoming packet was addressed directly to us(i.e. unicast),
2053          * use dst as the src for the reply.
2054          * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2055          * (for example) when we encounter an error while forwarding procedure
2056          * destined to a duplicated address of ours.
2057          */
2058         for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2059                 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2060                     (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2061                         src = &t;
2062                         break;
2063                 }
2064         if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2065                 /*
2066                  * This is the case if the dst is our link-local address
2067                  * and the sender is also ourselves.
2068                  */
2069                 src = &t;
2070         }
2071
2072         if (src == 0) {
2073                 int e;
2074                 struct route_in6 ro;
2075
2076                 /*
2077                  * This case matches to multicasts, our anycast, or unicasts
2078                  * that we do not own.  Select a source address based on the
2079                  * source address of the erroneous packet.
2080                  */
2081                 bzero(&ro, sizeof(ro));
2082                 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2083                 if (ro.ro_rt)
2084                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2085                 if (src == NULL) {
2086                         nd6log((LOG_DEBUG,
2087                             "icmp6_reflect: source can't be determined: "
2088                             "dst=%s, error=%d\n",
2089                             ip6_sprintf(&sa6_src.sin6_addr), e));
2090                         goto bad;
2091                 }
2092         }
2093
2094         ip6->ip6_src = *src;
2095
2096         ip6->ip6_flow = 0;
2097         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2098         ip6->ip6_vfc |= IPV6_VERSION;
2099         ip6->ip6_nxt = IPPROTO_ICMPV6;
2100         if (m->m_pkthdr.rcvif) {
2101                 /* XXX: This may not be the outgoing interface */
2102                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2103         } else
2104                 ip6->ip6_hlim = ip6_defhlim;
2105
2106         icmp6->icmp6_cksum = 0;
2107         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2108                                         sizeof(struct ip6_hdr), plen);
2109
2110         /*
2111          * XXX option handling
2112          */
2113
2114         m->m_flags &= ~(M_BCAST|M_MCAST);
2115
2116 #ifdef COMPAT_RFC1885
2117         ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2118 #else
2119         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2120 #endif
2121         if (outif)
2122                 icmp6_ifoutstat_inc(outif, type, code);
2123
2124         return;
2125
2126 bad:
2127         m_freem(m);
2128         return;
2129 }
2130
2131 void
2132 icmp6_fasttimo(void)
2133 {
2134
2135         mld6_fasttimeo();
2136 }
2137
2138 static const char *
2139 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2140                     struct in6_addr *tgt6)
2141 {
2142         static char buf[1024];
2143         ksnprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2144                 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2145         return buf;
2146 }
2147
2148 void
2149 icmp6_redirect_input(struct mbuf *m, int off)
2150 {
2151         struct ifnet *ifp = m->m_pkthdr.rcvif;
2152         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2153         struct nd_redirect *nd_rd;
2154         int icmp6len = ntohs(ip6->ip6_plen);
2155         char *lladdr = NULL;
2156         int lladdrlen = 0;
2157         u_char *redirhdr = NULL;
2158         int redirhdrlen = 0;
2159         struct rtentry *rt = NULL;
2160         int is_router;
2161         int is_onlink;
2162         struct in6_addr src6 = ip6->ip6_src;
2163         struct in6_addr redtgt6;
2164         struct in6_addr reddst6;
2165         union nd_opts ndopts;
2166
2167         if (!m || !ifp)
2168                 return;
2169
2170         /* XXX if we are router, we don't update route by icmp6 redirect */
2171         if (ip6_forwarding)
2172                 goto freeit;
2173         if (!icmp6_rediraccept)
2174                 goto freeit;
2175
2176 #ifndef PULLDOWN_TEST
2177         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2178         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2179 #else
2180         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2181         if (nd_rd == NULL) {
2182                 icmp6stat.icp6s_tooshort++;
2183                 return;
2184         }
2185 #endif
2186         redtgt6 = nd_rd->nd_rd_target;
2187         reddst6 = nd_rd->nd_rd_dst;
2188
2189         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2190                 redtgt6.s6_addr16[1] = htons(ifp->if_index);
2191         if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2192                 reddst6.s6_addr16[1] = htons(ifp->if_index);
2193
2194         /* validation */
2195         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2196                 nd6log((LOG_ERR,
2197                         "ICMP6 redirect sent from %s rejected; "
2198                         "must be from linklocal\n", ip6_sprintf(&src6)));
2199                 goto bad;
2200         }
2201         if (ip6->ip6_hlim != 255) {
2202                 nd6log((LOG_ERR,
2203                         "ICMP6 redirect sent from %s rejected; "
2204                         "hlim=%d (must be 255)\n",
2205                         ip6_sprintf(&src6), ip6->ip6_hlim));
2206                 goto bad;
2207         }
2208     {
2209         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2210         struct sockaddr_in6 sin6;
2211         struct in6_addr *gw6;
2212
2213         bzero(&sin6, sizeof(sin6));
2214         sin6.sin6_family = AF_INET6;
2215         sin6.sin6_len = sizeof(struct sockaddr_in6);
2216         bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2217         rt = rtpurelookup((struct sockaddr *)&sin6);
2218         if (rt != NULL) {
2219                 if (rt->rt_gateway == NULL ||
2220                     rt->rt_gateway->sa_family != AF_INET6) {
2221                         nd6log((LOG_ERR,
2222                             "ICMP6 redirect rejected; no route "
2223                             "with inet6 gateway found for redirect dst: %s\n",
2224                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2225                         --rt->rt_refcnt;
2226                         goto bad;
2227                 }
2228
2229                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2230                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2231                         nd6log((LOG_ERR,
2232                                 "ICMP6 redirect rejected; "
2233                                 "not equal to gw-for-src=%s (must be same): "
2234                                 "%s\n",
2235                                 ip6_sprintf(gw6),
2236                                 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2237                         --rt->rt_refcnt;
2238                         goto bad;
2239                 }
2240         } else {
2241                 nd6log((LOG_ERR,
2242                         "ICMP6 redirect rejected; "
2243                         "no route found for redirect dst: %s\n",
2244                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2245                 goto bad;
2246         }
2247         --rt->rt_refcnt;
2248         rt = NULL;
2249     }
2250         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2251                 nd6log((LOG_ERR,
2252                         "ICMP6 redirect rejected; "
2253                         "redirect dst must be unicast: %s\n",
2254                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2255                 goto bad;
2256         }
2257
2258         is_router = is_onlink = 0;
2259         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2260                 is_router = 1;  /* router case */
2261         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2262                 is_onlink = 1;  /* on-link destination case */
2263         if (!is_router && !is_onlink) {
2264                 nd6log((LOG_ERR,
2265                         "ICMP6 redirect rejected; "
2266                         "neither router case nor onlink case: %s\n",
2267                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2268                 goto bad;
2269         }
2270         /* validation passed */
2271
2272         icmp6len -= sizeof(*nd_rd);
2273         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2274         if (nd6_options(&ndopts) < 0) {
2275                 nd6log((LOG_INFO, "icmp6_redirect_input: "
2276                         "invalid ND option, rejected: %s\n",
2277                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2278                 /* nd6_options have incremented stats */
2279                 goto freeit;
2280         }
2281
2282         if (ndopts.nd_opts_tgt_lladdr) {
2283                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2284                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2285         }
2286
2287         if (ndopts.nd_opts_rh) {
2288                 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2289                 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2290         }
2291
2292         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2293                 nd6log((LOG_INFO,
2294                         "icmp6_redirect_input: lladdrlen mismatch for %s "
2295                         "(if %d, icmp6 packet %d): %s\n",
2296                         ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2297                         icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2298                 goto bad;
2299         }
2300
2301         /* RFC 2461 8.3 */
2302         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2303                          is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2304
2305         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2306                 /* perform rtredirect */
2307                 struct sockaddr_in6 sdst;
2308                 struct sockaddr_in6 sgw;
2309                 struct sockaddr_in6 ssrc;
2310
2311                 bzero(&sdst, sizeof(sdst));
2312                 bzero(&sgw, sizeof(sgw));
2313                 bzero(&ssrc, sizeof(ssrc));
2314                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2315                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2316                         sizeof(struct sockaddr_in6);
2317                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2318                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2319                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2320                 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2321                            (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2322                            (struct sockaddr *)&ssrc);
2323         }
2324         /* finally update cached route in each socket via pfctlinput */
2325     {
2326         struct sockaddr_in6 sdst;
2327
2328         bzero(&sdst, sizeof(sdst));
2329         sdst.sin6_family = AF_INET6;
2330         sdst.sin6_len = sizeof(struct sockaddr_in6);
2331         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2332         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2333 #ifdef IPSEC
2334         key_sa_routechange((struct sockaddr *)&sdst);
2335 #endif
2336     }
2337
2338 freeit:
2339         m_freem(m);
2340         return;
2341
2342 bad:
2343         icmp6stat.icp6s_badredirect++;
2344         m_freem(m);
2345 }
2346
2347 void
2348 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2349 {
2350         struct ifnet *ifp;      /* my outgoing interface */
2351         struct in6_addr *ifp_ll6;
2352         struct in6_addr *router_ll6;
2353         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2354         struct mbuf *m = NULL;  /* newly allocated one */
2355         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2356         struct nd_redirect *nd_rd;
2357         size_t maxlen;
2358         u_char *p;
2359         struct ifnet *outif = NULL;
2360         struct sockaddr_in6 src_sa;
2361
2362         icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2363
2364         /* if we are not router, we don't send icmp6 redirect */
2365         if (!ip6_forwarding || ip6_accept_rtadv)
2366                 goto fail;
2367
2368         /* sanity check */
2369         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2370                 goto fail;
2371
2372         /*
2373          * Address check:
2374          *  the source address must identify a neighbor, and
2375          *  the destination address must not be a multicast address
2376          *  [RFC 2461, sec 8.2]
2377          */
2378         sip6 = mtod(m0, struct ip6_hdr *);
2379         bzero(&src_sa, sizeof(src_sa));
2380         src_sa.sin6_family = AF_INET6;
2381         src_sa.sin6_len = sizeof(src_sa);
2382         src_sa.sin6_addr = sip6->ip6_src;
2383         /* we don't currently use sin6_scope_id, but eventually use it */
2384         src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2385         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2386                 goto fail;
2387         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2388                 goto fail;      /* what should we do here? */
2389
2390         /* rate limit */
2391         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2392                 goto fail;
2393
2394         /*
2395          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2396          * we almost always ask for an mbuf cluster for simplicity.
2397          * (MHLEN < IPV6_MMTU is almost always true)
2398          */
2399 #if IPV6_MMTU >= MCLBYTES
2400 # error assumption failed about IPV6_MMTU and MCLBYTES
2401 #endif
2402         m = m_getb(IPV6_MMTU, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
2403         if (!m)
2404                 goto fail;
2405         m->m_len = 0;
2406         maxlen = M_TRAILINGSPACE(m);
2407         maxlen = min(IPV6_MMTU, maxlen);
2408         /* just for safety */
2409         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2410             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2411                 goto fail;
2412         }
2413
2414         {
2415                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2416                 struct in6_ifaddr *ia;
2417                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2418                                                  IN6_IFF_NOTREADY|
2419                                                  IN6_IFF_ANYCAST)) == NULL)
2420                         goto fail;
2421                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2422         }
2423
2424         /* get ip6 linklocal address for the router. */
2425         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2426                 struct sockaddr_in6 *sin6;
2427                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2428                 router_ll6 = &sin6->sin6_addr;
2429                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2430                         router_ll6 = (struct in6_addr *)NULL;
2431         } else
2432                 router_ll6 = (struct in6_addr *)NULL;
2433
2434         /* ip6 */
2435         ip6 = mtod(m, struct ip6_hdr *);
2436         ip6->ip6_flow = 0;
2437         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2438         ip6->ip6_vfc |= IPV6_VERSION;
2439         /* ip6->ip6_plen will be set later */
2440         ip6->ip6_nxt = IPPROTO_ICMPV6;
2441         ip6->ip6_hlim = 255;
2442         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2443         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2444         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2445
2446         /* ND Redirect */
2447         nd_rd = (struct nd_redirect *)(ip6 + 1);
2448         nd_rd->nd_rd_type = ND_REDIRECT;
2449         nd_rd->nd_rd_code = 0;
2450         nd_rd->nd_rd_reserved = 0;
2451         if (rt->rt_flags & RTF_GATEWAY) {
2452                 /*
2453                  * nd_rd->nd_rd_target must be a link-local address in
2454                  * better router cases.
2455                  */
2456                 if (!router_ll6)
2457                         goto fail;
2458                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2459                       sizeof(nd_rd->nd_rd_target));
2460                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2461                       sizeof(nd_rd->nd_rd_dst));
2462         } else {
2463                 /* make sure redtgt == reddst */
2464                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2465                       sizeof(nd_rd->nd_rd_target));
2466                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2467                       sizeof(nd_rd->nd_rd_dst));
2468         }
2469
2470         p = (u_char *)(nd_rd + 1);
2471
2472         if (!router_ll6)
2473                 goto nolladdropt;
2474
2475     {
2476         /* target lladdr option */
2477         struct rtentry *rt_router = NULL;
2478         int len;
2479         struct sockaddr_dl *sdl;
2480         struct nd_opt_hdr *nd_opt;
2481         char *lladdr;
2482
2483         rt_router = nd6_lookup(router_ll6, 0, ifp);
2484         if (!rt_router)
2485                 goto nolladdropt;
2486         len = sizeof(*nd_opt) + ifp->if_addrlen;
2487         len = (len + 7) & ~7;   /* round by 8 */
2488         /* safety check */
2489         if (len + (p - (u_char *)ip6) > maxlen)
2490                 goto nolladdropt;
2491         if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2492             (rt_router->rt_flags & RTF_LLINFO) &&
2493             (rt_router->rt_gateway->sa_family == AF_LINK) &&
2494             (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2495             sdl->sdl_alen) {
2496                 nd_opt = (struct nd_opt_hdr *)p;
2497                 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2498                 nd_opt->nd_opt_len = len >> 3;
2499                 lladdr = (char *)(nd_opt + 1);
2500                 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2501                 p += len;
2502         }
2503     }
2504 nolladdropt:;
2505
2506         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2507
2508         /* just to be safe */
2509 #ifdef M_DECRYPTED      /*not openbsd*/
2510         if (m0->m_flags & M_DECRYPTED)
2511                 goto noredhdropt;
2512 #endif
2513         if (p - (u_char *)ip6 > maxlen)
2514                 goto noredhdropt;
2515
2516     {
2517         /* redirected header option */
2518         int len;
2519         struct nd_opt_rd_hdr *nd_opt_rh;
2520
2521         /*
2522          * compute the maximum size for icmp6 redirect header option.
2523          * XXX room for auth header?
2524          */
2525         len = maxlen - (p - (u_char *)ip6);
2526         len &= ~7;
2527
2528         /* This is just for simplicity. */
2529         if (m0->m_pkthdr.len != m0->m_len) {
2530                 if (m0->m_next) {
2531                         m_freem(m0->m_next);
2532                         m0->m_next = NULL;
2533                 }
2534                 m0->m_pkthdr.len = m0->m_len;
2535         }
2536
2537         /*
2538          * Redirected header option spec (RFC2461 4.6.3) talks nothing
2539          * about padding/truncate rule for the original IP packet.
2540          * From the discussion on IPv6imp in Feb 1999, the consensus was:
2541          * - "attach as much as possible" is the goal
2542          * - pad if not aligned (original size can be guessed by original
2543          *   ip6 header)
2544          * Following code adds the padding if it is simple enough,
2545          * and truncates if not.
2546          */
2547         if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2548                 panic("assumption failed in %s:%d", __FILE__, __LINE__);
2549
2550         if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2551                 /* not enough room, truncate */
2552                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2553         } else {
2554                 /* enough room, pad or truncate */
2555                 size_t extra;
2556
2557                 extra = m0->m_pkthdr.len % 8;
2558                 if (extra) {
2559                         /* pad if easy enough, truncate if not */
2560                         if (8 - extra <= M_TRAILINGSPACE(m0)) {
2561                                 /* pad */
2562                                 m0->m_len += (8 - extra);
2563                                 m0->m_pkthdr.len += (8 - extra);
2564                         } else {
2565                                 /* truncate */
2566                                 m0->m_pkthdr.len -= extra;
2567                                 m0->m_len -= extra;
2568                         }
2569                 }
2570                 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2571                 m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2572         }
2573
2574         nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2575         bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2576         nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2577         nd_opt_rh->nd_opt_rh_len = len >> 3;
2578         p += sizeof(*nd_opt_rh);
2579         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2580
2581         /* connect m0 to m */
2582         m->m_next = m0;
2583         m->m_pkthdr.len = m->m_len + m0->m_len;
2584     }
2585 noredhdropt:;
2586
2587         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2588                 sip6->ip6_src.s6_addr16[1] = 0;
2589         if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2590                 sip6->ip6_dst.s6_addr16[1] = 0;
2591 #if 0
2592         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2593                 ip6->ip6_src.s6_addr16[1] = 0;
2594         if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2595                 ip6->ip6_dst.s6_addr16[1] = 0;
2596 #endif
2597         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2598                 nd_rd->nd_rd_target.s6_addr16[1] = 0;
2599         if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2600                 nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2601
2602         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2603
2604         nd_rd->nd_rd_cksum = 0;
2605         nd_rd->nd_rd_cksum
2606                 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2607
2608         /* send the packet to outside... */
2609         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2610         if (outif) {
2611                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2612                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2613         }
2614         icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2615
2616         return;
2617
2618 fail:
2619         if (m)
2620                 m_freem(m);
2621         if (m0)
2622                 m_freem(m0);
2623 }
2624
2625 #ifdef HAVE_NRL_INPCB
2626 #define in6pcb          inpcb
2627 #define in6p_icmp6filt  inp_icmp6filt
2628 #endif
2629 /*
2630  * ICMPv6 socket option processing.
2631  */
2632 int
2633 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2634 {
2635         int error = 0;
2636         int optlen;
2637         struct inpcb *inp = so->so_pcb;
2638         int level, op, optname;
2639
2640         if (sopt) {
2641                 level = sopt->sopt_level;
2642                 op = sopt->sopt_dir;
2643                 optname = sopt->sopt_name;
2644                 optlen = sopt->sopt_valsize;
2645         } else
2646                 level = op = optname = optlen = 0;
2647
2648         if (level != IPPROTO_ICMPV6) {
2649                 return EINVAL;
2650         }
2651
2652         switch (op) {
2653         case PRCO_SETOPT:
2654                 switch (optname) {
2655                 case ICMP6_FILTER:
2656                     {
2657                         struct icmp6_filter *p;
2658
2659                         if (optlen != sizeof(*p)) {
2660                                 error = EMSGSIZE;
2661                                 break;
2662                         }
2663                         if (inp->in6p_icmp6filt == NULL) {
2664                                 error = EINVAL;
2665                                 break;
2666                         }
2667                         error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2668                                 optlen);
2669                         break;
2670                     }
2671
2672                 default:
2673                         error = ENOPROTOOPT;
2674                         break;
2675                 }
2676                 break;
2677
2678         case PRCO_GETOPT:
2679                 switch (optname) {
2680                 case ICMP6_FILTER:
2681                     {
2682                         if (inp->in6p_icmp6filt == NULL) {
2683                                 error = EINVAL;
2684                                 break;
2685                         }
2686                         error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2687                                 sizeof(struct icmp6_filter));
2688                         break;
2689                     }
2690
2691                 default:
2692                         error = ENOPROTOOPT;
2693                         break;
2694                 }
2695                 break;
2696         }
2697
2698         return (error);
2699 }
2700 #ifdef HAVE_NRL_INPCB
2701 #undef in6pcb
2702 #undef in6p_icmp6filt
2703 #endif
2704
2705 #ifndef HAVE_PPSRATECHECK
2706 #ifndef timersub
2707 #define timersub(tvp, uvp, vvp)                                         \
2708         do {                                                            \
2709                 (vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;          \
2710                 (vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;       \
2711                 if ((vvp)->tv_usec < 0) {                               \
2712                         (vvp)->tv_sec--;                                \
2713                         (vvp)->tv_usec += 1000000;                      \
2714                 }                                                       \
2715         } while (0)
2716 #endif
2717
2718 /*
2719  * ppsratecheck(): packets (or events) per second limitation.
2720  */
2721 static int
2722 ppsratecheck(struct timeval *lasttime, int *curpps,
2723              int maxpps)        /* maximum pps allowed */
2724 {
2725         struct timeval tv, delta;
2726         int rv;
2727
2728         microtime(&tv);
2729
2730         timersub(&tv, lasttime, &delta);
2731
2732         /*
2733          * Check for 0,0 so that the message will be seen at least once.
2734          * If more than one second has passed since the last update of
2735          * lasttime, reset the counter.
2736          *
2737          * We do increment *curpps even in *curpps < maxpps case, as some may
2738          * try to use *curpps for stat purposes as well.
2739          */
2740         if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2741             delta.tv_sec >= 1) {
2742                 *lasttime = tv;
2743                 *curpps = 0;
2744                 rv = 1;
2745         } else if (maxpps < 0)
2746                 rv = 1;
2747         else if (*curpps < maxpps)
2748                 rv = 1;
2749         else
2750                 rv = 0;
2751
2752 #if 1 /* DIAGNOSTIC? */
2753         /* be careful about wrap-around */
2754         if (*curpps + 1 > *curpps)
2755                 *curpps = *curpps + 1;
2756 #else
2757         /*
2758          * assume that there's not too many calls to this function.
2759          * not sure if the assumption holds, as it depends on *caller's*
2760          * behavior, not the behavior of this function.
2761          * IMHO it is wrong to make assumption on the caller's behavior,
2762          * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2763          */
2764         *curpps = *curpps + 1;
2765 #endif
2766
2767         return (rv);
2768 }
2769 #endif
2770
2771 /*
2772  * Perform rate limit check.
2773  * Returns 0 if it is okay to send the icmp6 packet.
2774  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2775  * limitation.
2776  *
2777  * XXX per-destination/type check necessary?
2778  */
2779 static int
2780 icmp6_ratelimit(const struct in6_addr *dst,     /* not used at this moment */
2781                 const int type,                 /* not used at this moment */
2782                 const int code)                 /* not used at this moment */
2783 {
2784         int ret;
2785
2786         ret = 0;        /* okay to send */
2787
2788         /* PPS limit */
2789         if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2790             icmp6errppslim)) {
2791                 /* The packet is subject to rate limit */
2792                 ret++;
2793         }
2794
2795         return ret;
2796 }