sys/netinet6/icmp6.c

   1 /*-
   2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. Neither the name of the project nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
  30  */
  31
  32 /*-
  33  * Copyright (c) 1982, 1986, 1988, 1993
  34  *      The Regents of the University of California.  All rights reserved.
  35  *
  36  * Redistribution and use in source and binary forms, with or without
  37  * modification, are permitted provided that the following conditions
  38  * are met:
  39  * 1. Redistributions of source code must retain the above copyright
  40  *    notice, this list of conditions and the following disclaimer.
  41  * 2. Redistributions in binary form must reproduce the above copyright
  42  *    notice, this list of conditions and the following disclaimer in the
  43  *    documentation and/or other materials provided with the distribution.
  44  * 4. Neither the name of the University nor the names of its contributors
  45  *    may be used to endorse or promote products derived from this software
  46  *    without specific prior written permission.
  47  *
  48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  58  * SUCH DAMAGE.
  59  *
  60  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
  61  */
  62
  63 #include <sys/cdefs.h>
  64 __FBSDID("$FreeBSD$");
  65
  66 #define MBUF_PRIVATE    /* XXXRW: Optimisation tries to avoid M_EXT mbufs */
  67
  68 #include "opt_inet.h"
  69 #include "opt_inet6.h"
  70 #include "opt_ipsec.h"
  71
  72 #include <sys/param.h>
  73 #include <sys/domain.h>
  74 #include <sys/jail.h>
  75 #include <sys/kernel.h>
  76 #include <sys/lock.h>
  77 #include <sys/malloc.h>
  78 #include <sys/mbuf.h>
  79 #include <sys/proc.h>
  80 #include <sys/protosw.h>
  81 #include <sys/signalvar.h>
  82 #include <sys/socket.h>
  83 #include <sys/socketvar.h>
  84 #include <sys/sx.h>
  85 #include <sys/syslog.h>
  86 #include <sys/systm.h>
  87 #include <sys/time.h>
  88
  89 #include <net/if.h>
  90 #include <net/if_var.h>
  91 #include <net/if_dl.h>
  92 #include <net/if_llatbl.h>
  93 #include <net/if_types.h>
  94 #include <net/route.h>
  95 #include <net/vnet.h>
  96
  97 #include <netinet/in.h>
  98 #include <netinet/in_pcb.h>
  99 #include <netinet/in_var.h>
 100 #include <netinet/ip6.h>
 101 #include <netinet/icmp6.h>
 102 #include <netinet/tcp_var.h>
 103
 104 #include <netinet6/in6_ifattach.h>
 105 #include <netinet6/in6_pcb.h>
 106 #include <netinet6/ip6protosw.h>
 107 #include <netinet6/ip6_var.h>
 108 #include <netinet6/scope6_var.h>
 109 #include <netinet6/mld6_var.h>
 110 #include <netinet6/nd6.h>
 111 #include <netinet6/send.h>
 112
 113 #ifdef IPSEC
 114 #include <netipsec/ipsec.h>
 115 #include <netipsec/key.h>
 116 #endif
 117
 118 extern struct domain inet6domain;
 119
 120 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
 121 VNET_PCPUSTAT_SYSINIT(icmp6stat);
 122
 123 #ifdef VIMAGE
 124 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
 125 #endif /* VIMAGE */
 126
 127 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
 128 VNET_DECLARE(struct inpcbhead, ripcb);
 129 VNET_DECLARE(int, icmp6errppslim);
 130 static VNET_DEFINE(int, icmp6errpps_count) = 0;
 131 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
 132 VNET_DECLARE(int, icmp6_nodeinfo);
 133
 134 #define V_ripcbinfo                     VNET(ripcbinfo)
 135 #define V_ripcb                         VNET(ripcb)
 136 #define V_icmp6errppslim                VNET(icmp6errppslim)
 137 #define V_icmp6errpps_count             VNET(icmp6errpps_count)
 138 #define V_icmp6errppslim_last           VNET(icmp6errppslim_last)
 139 #define V_icmp6_nodeinfo                VNET(icmp6_nodeinfo)
 140
 141 static void icmp6_errcount(int, int);
 142 static int icmp6_rip6_input(struct mbuf **, int);
 143 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
 144 static const char *icmp6_redirect_diag(struct in6_addr *,
 145         struct in6_addr *, struct in6_addr *);
 146 static struct mbuf *ni6_input(struct mbuf *, int);
 147 static struct mbuf *ni6_nametodns(const char *, int, int);
 148 static int ni6_dnsmatch(const char *, int, const char *, int);
 149 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
 150                           struct ifnet **, struct in6_addr *);
 151 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
 152                                 struct ifnet *, int);
 153 static int icmp6_notify_error(struct mbuf **, int, int, int);
 154
 155 /*
 156  * Kernel module interface for updating icmp6stat.  The argument is an index
 157  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
 158  * general layout of icmp6stat into the caller, it doesn't encode its
 159  * location, so that future changes to add, for example, per-CPU stats
 160  * support won't cause binary compatibility problems for kernel modules.
 161  */
 162 void
 163 kmod_icmp6stat_inc(int statnum)
 164 {
 165
 166         counter_u64_add(VNET(icmp6stat)[statnum], 1);
 167 }
 168
 169 static void
 170 icmp6_errcount(int type, int code)
 171 {
 172         switch (type) {
 173         case ICMP6_DST_UNREACH:
 174                 switch (code) {
 175                 case ICMP6_DST_UNREACH_NOROUTE:
 176                         ICMP6STAT_INC(icp6s_odst_unreach_noroute);
 177                         return;
 178                 case ICMP6_DST_UNREACH_ADMIN:
 179                         ICMP6STAT_INC(icp6s_odst_unreach_admin);
 180                         return;
 181                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 182                         ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
 183                         return;
 184                 case ICMP6_DST_UNREACH_ADDR:
 185                         ICMP6STAT_INC(icp6s_odst_unreach_addr);
 186                         return;
 187                 case ICMP6_DST_UNREACH_NOPORT:
 188                         ICMP6STAT_INC(icp6s_odst_unreach_noport);
 189                         return;
 190                 }
 191                 break;
 192         case ICMP6_PACKET_TOO_BIG:
 193                 ICMP6STAT_INC(icp6s_opacket_too_big);
 194                 return;
 195         case ICMP6_TIME_EXCEEDED:
 196                 switch (code) {
 197                 case ICMP6_TIME_EXCEED_TRANSIT:
 198                         ICMP6STAT_INC(icp6s_otime_exceed_transit);
 199                         return;
 200                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 201                         ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
 202                         return;
 203                 }
 204                 break;
 205         case ICMP6_PARAM_PROB:
 206                 switch (code) {
 207                 case ICMP6_PARAMPROB_HEADER:
 208                         ICMP6STAT_INC(icp6s_oparamprob_header);
 209                         return;
 210                 case ICMP6_PARAMPROB_NEXTHEADER:
 211                         ICMP6STAT_INC(icp6s_oparamprob_nextheader);
 212                         return;
 213                 case ICMP6_PARAMPROB_OPTION:
 214                         ICMP6STAT_INC(icp6s_oparamprob_option);
 215                         return;
 216                 }
 217                 break;
 218         case ND_REDIRECT:
 219                 ICMP6STAT_INC(icp6s_oredirect);
 220                 return;
 221         }
 222         ICMP6STAT_INC(icp6s_ounknown);
 223 }
 224
 225 /*
 226  * A wrapper function for icmp6_error() necessary when the erroneous packet
 227  * may not contain enough scope zone information.
 228  */
 229 void
 230 icmp6_error2(struct mbuf *m, int type, int code, int param,
 231     struct ifnet *ifp)
 232 {
 233         struct ip6_hdr *ip6;
 234
 235         if (ifp == NULL)
 236                 return;
 237
 238 #ifndef PULLDOWN_TEST
 239         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 240 #else
 241         if (m->m_len < sizeof(struct ip6_hdr)) {
 242                 m = m_pullup(m, sizeof(struct ip6_hdr));
 243                 if (m == NULL)
 244                         return;
 245         }
 246 #endif
 247
 248         ip6 = mtod(m, struct ip6_hdr *);
 249
 250         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
 251                 return;
 252         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
 253                 return;
 254
 255         icmp6_error(m, type, code, param);
 256 }
 257
 258 /*
 259  * Generate an error packet of type error in response to bad IP6 packet.
 260  */
 261 void
 262 icmp6_error(struct mbuf *m, int type, int code, int param)
 263 {
 264         struct ip6_hdr *oip6, *nip6;
 265         struct icmp6_hdr *icmp6;
 266         u_int preplen;
 267         int off;
 268         int nxt;
 269
 270         ICMP6STAT_INC(icp6s_error);
 271
 272         /* count per-type-code statistics */
 273         icmp6_errcount(type, code);
 274
 275 #ifdef M_DECRYPTED      /*not openbsd*/
 276         if (m->m_flags & M_DECRYPTED) {
 277                 ICMP6STAT_INC(icp6s_canterror);
 278                 goto freeit;
 279         }
 280 #endif
 281
 282 #ifndef PULLDOWN_TEST
 283         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
 284 #else
 285         if (m->m_len < sizeof(struct ip6_hdr)) {
 286                 m = m_pullup(m, sizeof(struct ip6_hdr));
 287                 if (m == NULL)
 288                         return;
 289         }
 290 #endif
 291         oip6 = mtod(m, struct ip6_hdr *);
 292
 293         /*
 294          * If the destination address of the erroneous packet is a multicast
 295          * address, or the packet was sent using link-layer multicast,
 296          * we should basically suppress sending an error (RFC 2463, Section
 297          * 2.4).
 298          * We have two exceptions (the item e.2 in that section):
 299          * - the Packet Too Big message can be sent for path MTU discovery.
 300          * - the Parameter Problem Message that can be allowed an icmp6 error
 301          *   in the option type field.  This check has been done in
 302          *   ip6_unknown_opt(), so we can just check the type and code.
 303          */
 304         if ((m->m_flags & (M_BCAST|M_MCAST) ||
 305              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
 306             (type != ICMP6_PACKET_TOO_BIG &&
 307              (type != ICMP6_PARAM_PROB ||
 308               code != ICMP6_PARAMPROB_OPTION)))
 309                 goto freeit;
 310
 311         /*
 312          * RFC 2463, 2.4 (e.5): source address check.
 313          * XXX: the case of anycast source?
 314          */
 315         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
 316             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
 317                 goto freeit;
 318
 319         /*
 320          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
 321          * don't do it.
 322          */
 323         nxt = -1;
 324         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
 325         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
 326                 struct icmp6_hdr *icp;
 327
 328 #ifndef PULLDOWN_TEST
 329                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
 330                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 331 #else
 332                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
 333                         sizeof(*icp));
 334                 if (icp == NULL) {
 335                         ICMP6STAT_INC(icp6s_tooshort);
 336                         return;
 337                 }
 338 #endif
 339                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
 340                     icp->icmp6_type == ND_REDIRECT) {
 341                         /*
 342                          * ICMPv6 error
 343                          * Special case: for redirect (which is
 344                          * informational) we must not send icmp6 error.
 345                          */
 346                         ICMP6STAT_INC(icp6s_canterror);
 347                         goto freeit;
 348                 } else {
 349                         /* ICMPv6 informational - send the error */
 350                 }
 351         } else {
 352                 /* non-ICMPv6 - send the error */
 353         }
 354
 355         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
 356
 357         /* Finally, do rate limitation check. */
 358         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
 359                 ICMP6STAT_INC(icp6s_toofreq);
 360                 goto freeit;
 361         }
 362
 363         /*
 364          * OK, ICMP6 can be generated.
 365          */
 366
 367         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
 368                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
 369
 370         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
 371         M_PREPEND(m, preplen, M_NOWAIT);        /* FIB is also copied over. */
 372         if (m && m->m_len < preplen)
 373                 m = m_pullup(m, preplen);
 374         if (m == NULL) {
 375                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
 376                 return;
 377         }
 378
 379         nip6 = mtod(m, struct ip6_hdr *);
 380         nip6->ip6_src  = oip6->ip6_src;
 381         nip6->ip6_dst  = oip6->ip6_dst;
 382
 383         in6_clearscope(&oip6->ip6_src);
 384         in6_clearscope(&oip6->ip6_dst);
 385
 386         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
 387         icmp6->icmp6_type = type;
 388         icmp6->icmp6_code = code;
 389         icmp6->icmp6_pptr = htonl((u_int32_t)param);
 390
 391         /*
 392          * icmp6_reflect() is designed to be in the input path.
 393          * icmp6_error() can be called from both input and output path,
 394          * and if we are in output path rcvif could contain bogus value.
 395          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
 396          * information in ip header (nip6).
 397          */
 398         m->m_pkthdr.rcvif = NULL;
 399
 400         ICMP6STAT_INC(icp6s_outhist[type]);
 401         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
 402
 403         return;
 404
 405   freeit:
 406         /*
 407          * If we can't tell whether or not we can generate ICMP6, free it.
 408          */
 409         m_freem(m);
 410 }
 411
 412 /*
 413  * Process a received ICMP6 message.
 414  */
 415 int
 416 icmp6_input(struct mbuf **mp, int *offp, int proto)
 417 {
 418         struct mbuf *m = *mp, *n;
 419         struct ifnet *ifp;
 420         struct ip6_hdr *ip6, *nip6;
 421         struct icmp6_hdr *icmp6, *nicmp6;
 422         int off = *offp;
 423         int icmp6len = m->m_pkthdr.len - *offp;
 424         int code, sum, noff;
 425         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
 426         int ip6len, error;
 427
 428         ifp = m->m_pkthdr.rcvif;
 429
 430 #ifndef PULLDOWN_TEST
 431         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
 432         /* m might change if M_LOOP.  So, call mtod after this */
 433 #endif
 434
 435         /*
 436          * Locate icmp6 structure in mbuf, and check
 437          * that not corrupted and of at least minimum length
 438          */
 439
 440         ip6 = mtod(m, struct ip6_hdr *);
 441         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
 442         if (icmp6len < sizeof(struct icmp6_hdr)) {
 443                 ICMP6STAT_INC(icp6s_tooshort);
 444                 goto freeit;
 445         }
 446
 447         /*
 448          * Check multicast group membership.
 449          * Note: SSM filters are not applied for ICMPv6 traffic.
 450          */
 451         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 452                 struct in6_multi        *inm;
 453
 454                 inm = in6m_lookup(ifp, &ip6->ip6_dst);
 455                 if (inm == NULL) {
 456                         IP6STAT_INC(ip6s_notmember);
 457                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
 458                         goto freeit;
 459                 }
 460         }
 461
 462         /*
 463          * calculate the checksum
 464          */
 465 #ifndef PULLDOWN_TEST
 466         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
 467 #else
 468         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
 469         if (icmp6 == NULL) {
 470                 ICMP6STAT_INC(icp6s_tooshort);
 471                 return IPPROTO_DONE;
 472         }
 473 #endif
 474         code = icmp6->icmp6_code;
 475
 476         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
 477                 nd6log((LOG_ERR,
 478                     "ICMP6 checksum error(%d|%x) %s\n",
 479                     icmp6->icmp6_type, sum,
 480                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
 481                 ICMP6STAT_INC(icp6s_checksum);
 482                 goto freeit;
 483         }
 484
 485         ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
 486         icmp6_ifstat_inc(ifp, ifs6_in_msg);
 487         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
 488                 icmp6_ifstat_inc(ifp, ifs6_in_error);
 489
 490         switch (icmp6->icmp6_type) {
 491         case ICMP6_DST_UNREACH:
 492                 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
 493                 switch (code) {
 494                 case ICMP6_DST_UNREACH_NOROUTE:
 495                         code = PRC_UNREACH_NET;
 496                         break;
 497                 case ICMP6_DST_UNREACH_ADMIN:
 498                         icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
 499                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
 500                         break;
 501                 case ICMP6_DST_UNREACH_ADDR:
 502                         code = PRC_HOSTDEAD;
 503                         break;
 504                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 505                         /* I mean "source address was incorrect." */
 506                         code = PRC_PARAMPROB;
 507                         break;
 508                 case ICMP6_DST_UNREACH_NOPORT:
 509                         code = PRC_UNREACH_PORT;
 510                         break;
 511                 default:
 512                         goto badcode;
 513                 }
 514                 goto deliver;
 515                 break;
 516
 517         case ICMP6_PACKET_TOO_BIG:
 518                 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
 519
 520                 /* validation is made in icmp6_mtudisc_update */
 521
 522                 code = PRC_MSGSIZE;
 523
 524                 /*
 525                  * Updating the path MTU will be done after examining
 526                  * intermediate extension headers.
 527                  */
 528                 goto deliver;
 529                 break;
 530
 531         case ICMP6_TIME_EXCEEDED:
 532                 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
 533                 switch (code) {
 534                 case ICMP6_TIME_EXCEED_TRANSIT:
 535                         code = PRC_TIMXCEED_INTRANS;
 536                         break;
 537                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 538                         code = PRC_TIMXCEED_REASS;
 539                         break;
 540                 default:
 541                         goto badcode;
 542                 }
 543                 goto deliver;
 544                 break;
 545
 546         case ICMP6_PARAM_PROB:
 547                 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
 548                 switch (code) {
 549                 case ICMP6_PARAMPROB_NEXTHEADER:
 550                         code = PRC_UNREACH_PROTOCOL;
 551                         break;
 552                 case ICMP6_PARAMPROB_HEADER:
 553                 case ICMP6_PARAMPROB_OPTION:
 554                         code = PRC_PARAMPROB;
 555                         break;
 556                 default:
 557                         goto badcode;
 558                 }
 559                 goto deliver;
 560                 break;
 561
 562         case ICMP6_ECHO_REQUEST:
 563                 icmp6_ifstat_inc(ifp, ifs6_in_echo);
 564                 if (code != 0)
 565                         goto badcode;
 566                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
 567                         /* Give up remote */
 568                         break;
 569                 }
 570                 if (!M_WRITABLE(n)
 571                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
 572                         struct mbuf *n0 = n;
 573                         int n0len;
 574
 575                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
 576                         n = m_gethdr(M_NOWAIT, n0->m_type);
 577                         if (n == NULL) {
 578                                 /* Give up remote */
 579                                 m_freem(n0);
 580                                 break;
 581                         }
 582
 583                         m_move_pkthdr(n, n0);   /* FIB copied. */
 584                         n0len = n0->m_pkthdr.len;       /* save for use below */
 585                         /*
 586                          * Copy IPv6 and ICMPv6 only.
 587                          */
 588                         nip6 = mtod(n, struct ip6_hdr *);
 589                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 590                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 591                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 592                         noff = sizeof(struct ip6_hdr);
 593                         /* new mbuf contains only ipv6+icmpv6 headers */
 594                         n->m_len = noff + sizeof(struct icmp6_hdr);
 595                         /*
 596                          * Adjust mbuf.  ip6_plen will be adjusted in
 597                          * ip6_output().
 598                          */
 599                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 600                         /* recalculate complete packet size */
 601                         n->m_pkthdr.len = n0len + (noff - off);
 602                         n->m_next = n0;
 603                 } else {
 604                         nip6 = mtod(n, struct ip6_hdr *);
 605                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
 606                             sizeof(*nicmp6));
 607                         noff = off;
 608                 }
 609                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 610                 nicmp6->icmp6_code = 0;
 611                 if (n) {
 612                         ICMP6STAT_INC(icp6s_reflect);
 613                         ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
 614                         icmp6_reflect(n, noff);
 615                 }
 616                 break;
 617
 618         case ICMP6_ECHO_REPLY:
 619                 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
 620                 if (code != 0)
 621                         goto badcode;
 622                 break;
 623
 624         case MLD_LISTENER_QUERY:
 625         case MLD_LISTENER_REPORT:
 626         case MLD_LISTENER_DONE:
 627         case MLDV2_LISTENER_REPORT:
 628                 /*
 629                  * Drop MLD traffic which is not link-local, has a hop limit
 630                  * of greater than 1 hop, or which does not have the
 631                  * IPv6 HBH Router Alert option.
 632                  * As IPv6 HBH options are stripped in ip6_input() we must
 633                  * check an mbuf header flag.
 634                  * XXX Should we also sanity check that these messages
 635                  * were directed to a link-local multicast prefix?
 636                  */
 637                 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
 638                         goto freeit;
 639                 if (mld_input(m, off, icmp6len) != 0)
 640                         return (IPPROTO_DONE);
 641                 /* m stays. */
 642                 break;
 643
 644         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 645             {
 646                 enum { WRU, FQDN } mode;
 647
 648                 if (!V_icmp6_nodeinfo)
 649                         break;
 650
 651                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 652                         mode = WRU;
 653                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 654                         mode = FQDN;
 655                 else
 656                         goto badlen;
 657
 658                 if (mode == FQDN) {
 659 #ifndef PULLDOWN_TEST
 660                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
 661                             IPPROTO_DONE);
 662 #endif
 663                         n = m_copy(m, 0, M_COPYALL);
 664                         if (n)
 665                                 n = ni6_input(n, off);
 666                         /* XXX meaningless if n == NULL */
 667                         noff = sizeof(struct ip6_hdr);
 668                 } else {
 669                         struct prison *pr;
 670                         u_char *p;
 671                         int maxhlen, hlen;
 672
 673                         /*
 674                          * XXX: this combination of flags is pointless,
 675                          * but should we keep this for compatibility?
 676                          */
 677                         if ((V_icmp6_nodeinfo & 5) != 5)
 678                                 break;
 679
 680                         if (code != 0)
 681                                 goto badcode;
 682
 683                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
 684                         n = m_gethdr(M_NOWAIT, m->m_type);
 685                         if (n == NULL) {
 686                                 /* Give up remote */
 687                                 break;
 688                         }
 689                         if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
 690                                 /*
 691                                  * Previous code did a blind M_COPY_PKTHDR
 692                                  * and said "just for rcvif".  If true, then
 693                                  * we could tolerate the dup failing (due to
 694                                  * the deep copy of the tag chain).  For now
 695                                  * be conservative and just fail.
 696                                  */
 697                                 m_free(n);
 698                                 n = NULL;
 699                         }
 700                         maxhlen = M_TRAILINGSPACE(n) -
 701                             (sizeof(*nip6) + sizeof(*nicmp6) + 4);
 702                         pr = curthread->td_ucred->cr_prison;
 703                         mtx_lock(&pr->pr_mtx);
 704                         hlen = strlen(pr->pr_hostname);
 705                         if (maxhlen > hlen)
 706                                 maxhlen = hlen;
 707                         /*
 708                          * Copy IPv6 and ICMPv6 only.
 709                          */
 710                         nip6 = mtod(n, struct ip6_hdr *);
 711                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 712                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 713                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 714                         p = (u_char *)(nicmp6 + 1);
 715                         bzero(p, 4);
 716                         /* meaningless TTL */
 717                         bcopy(pr->pr_hostname, p + 4, maxhlen);
 718                         mtx_unlock(&pr->pr_mtx);
 719                         noff = sizeof(struct ip6_hdr);
 720                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 721                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 722                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 723                         nicmp6->icmp6_code = 0;
 724                 }
 725                 if (n) {
 726                         ICMP6STAT_INC(icp6s_reflect);
 727                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
 728                         icmp6_reflect(n, noff);
 729                 }
 730                 break;
 731             }
 732
 733         case ICMP6_WRUREPLY:
 734                 if (code != 0)
 735                         goto badcode;
 736                 break;
 737
 738         case ND_ROUTER_SOLICIT:
 739                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
 740                 if (code != 0)
 741                         goto badcode;
 742                 if (icmp6len < sizeof(struct nd_router_solicit))
 743                         goto badlen;
 744                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
 745                         /* give up local */
 746
 747                         /* Send incoming SeND packet to user space. */
 748                         if (send_sendso_input_hook != NULL) {
 749                                 IP6_EXTHDR_CHECK(m, off,
 750                                     icmp6len, IPPROTO_DONE);
 751                                 error = send_sendso_input_hook(m, ifp,
 752                                     SND_IN, ip6len);
 753                                 /* -1 == no app on SEND socket */
 754                                 if (error == 0)
 755                                         return (IPPROTO_DONE);
 756                                 nd6_rs_input(m, off, icmp6len);
 757                         } else
 758                                 nd6_rs_input(m, off, icmp6len);
 759                         m = NULL;
 760                         goto freeit;
 761                 }
 762                 if (send_sendso_input_hook != NULL) {
 763                         IP6_EXTHDR_CHECK(n, off,
 764                             icmp6len, IPPROTO_DONE);
 765                         error = send_sendso_input_hook(n, ifp,
 766                             SND_IN, ip6len);
 767                         if (error == 0)
 768                                 goto freeit;
 769                         /* -1 == no app on SEND socket */
 770                         nd6_rs_input(n, off, icmp6len);
 771                 } else
 772                         nd6_rs_input(n, off, icmp6len);
 773                 /* m stays. */
 774                 break;
 775
 776         case ND_ROUTER_ADVERT:
 777                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
 778                 if (code != 0)
 779                         goto badcode;
 780                 if (icmp6len < sizeof(struct nd_router_advert))
 781                         goto badlen;
 782                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
 783
 784                         /* Send incoming SeND-protected/ND packet to user space. */
 785                         if (send_sendso_input_hook != NULL) {
 786                                 error = send_sendso_input_hook(m, ifp,
 787                                     SND_IN, ip6len);
 788                                 if (error == 0)
 789                                         return (IPPROTO_DONE);
 790                                 nd6_ra_input(m, off, icmp6len);
 791                         } else
 792                                 nd6_ra_input(m, off, icmp6len);
 793                         m = NULL;
 794                         goto freeit;
 795                 }
 796                 if (send_sendso_input_hook != NULL) {
 797                         error = send_sendso_input_hook(n, ifp,
 798                             SND_IN, ip6len);
 799                         if (error == 0)
 800                                 goto freeit;
 801                         nd6_ra_input(n, off, icmp6len);
 802                 } else
 803                         nd6_ra_input(n, off, icmp6len);
 804                 /* m stays. */
 805                 break;
 806
 807         case ND_NEIGHBOR_SOLICIT:
 808                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
 809                 if (code != 0)
 810                         goto badcode;
 811                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 812                         goto badlen;
 813                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
 814                         if (send_sendso_input_hook != NULL) {
 815                                 error = send_sendso_input_hook(m, ifp,
 816                                     SND_IN, ip6len);
 817                                 if (error == 0)
 818                                         return (IPPROTO_DONE);
 819                                 nd6_ns_input(m, off, icmp6len);
 820                         } else
 821                                 nd6_ns_input(m, off, icmp6len);
 822                         m = NULL;
 823                         goto freeit;
 824                 }
 825                 if (send_sendso_input_hook != NULL) {
 826                         error = send_sendso_input_hook(n, ifp,
 827                             SND_IN, ip6len);
 828                         if (error == 0)
 829                                 goto freeit;
 830                         nd6_ns_input(n, off, icmp6len);
 831                 } else
 832                         nd6_ns_input(n, off, icmp6len);
 833                 /* m stays. */
 834                 break;
 835
 836         case ND_NEIGHBOR_ADVERT:
 837                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
 838                 if (code != 0)
 839                         goto badcode;
 840                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 841                         goto badlen;
 842                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
 843
 844                         /* Send incoming SeND-protected/ND packet to user space. */
 845                         if (send_sendso_input_hook != NULL) {
 846                                 error = send_sendso_input_hook(m, ifp,
 847                                     SND_IN, ip6len);
 848                                 if (error == 0)
 849                                         return (IPPROTO_DONE);
 850                                 nd6_na_input(m, off, icmp6len);
 851                         } else
 852                                 nd6_na_input(m, off, icmp6len);
 853                         m = NULL;
 854                         goto freeit;
 855                 }
 856                 if (send_sendso_input_hook != NULL) {
 857                         error = send_sendso_input_hook(n, ifp,
 858                             SND_IN, ip6len);
 859                         if (error == 0)
 860                                 goto freeit;
 861                         nd6_na_input(n, off, icmp6len);
 862                 } else
 863                         nd6_na_input(n, off, icmp6len);
 864                 /* m stays. */
 865                 break;
 866
 867         case ND_REDIRECT:
 868                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
 869                 if (code != 0)
 870                         goto badcode;
 871                 if (icmp6len < sizeof(struct nd_redirect))
 872                         goto badlen;
 873                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
 874                         if (send_sendso_input_hook != NULL) {
 875                                 error = send_sendso_input_hook(m, ifp,
 876                                     SND_IN, ip6len);
 877                                 if (error == 0)
 878                                         return (IPPROTO_DONE);
 879                             icmp6_redirect_input(m, off);
 880                         } else
 881                                 icmp6_redirect_input(m, off);
 882                         m = NULL;
 883                         goto freeit;
 884                 }
 885                 if (send_sendso_input_hook != NULL) {
 886                         error = send_sendso_input_hook(n, ifp,
 887                             SND_IN, ip6len);
 888                         if (error == 0)
 889                                 goto freeit;
 890                         icmp6_redirect_input(n, off);
 891                 } else
 892                         icmp6_redirect_input(n, off);
 893                 /* m stays. */
 894                 break;
 895
 896         case ICMP6_ROUTER_RENUMBERING:
 897                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 898                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 899                         goto badcode;
 900                 if (icmp6len < sizeof(struct icmp6_router_renum))
 901                         goto badlen;
 902                 break;
 903
 904         default:
 905                 nd6log((LOG_DEBUG,
 906                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 907                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
 908                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
 909                     ifp ? ifp->if_index : 0));
 910                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 911                         /* ICMPv6 error: MUST deliver it by spec... */
 912                         code = PRC_NCMDS;
 913                         /* deliver */
 914                 } else {
 915                         /* ICMPv6 informational: MUST not deliver */
 916                         break;
 917                 }
 918         deliver:
 919                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
 920                         /* In this case, m should've been freed. */
 921                         return (IPPROTO_DONE);
 922                 }
 923                 break;
 924
 925         badcode:
 926                 ICMP6STAT_INC(icp6s_badcode);
 927                 break;
 928
 929         badlen:
 930                 ICMP6STAT_INC(icp6s_badlen);
 931                 break;
 932         }
 933
 934         /* deliver the packet to appropriate sockets */
 935         icmp6_rip6_input(&m, *offp);
 936
 937         return IPPROTO_DONE;
 938
 939  freeit:
 940         m_freem(m);
 941         return IPPROTO_DONE;
 942 }
 943
 944 static int
 945 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
 946 {
 947         struct mbuf *m = *mp;
 948         struct icmp6_hdr *icmp6;
 949         struct ip6_hdr *eip6;
 950         u_int32_t notifymtu;
 951         struct sockaddr_in6 icmp6src, icmp6dst;
 952
 953         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 954                 ICMP6STAT_INC(icp6s_tooshort);
 955                 goto freeit;
 956         }
 957 #ifndef PULLDOWN_TEST
 958         IP6_EXTHDR_CHECK(m, off,
 959             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
 960         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 961 #else
 962         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
 963             sizeof(*icmp6) + sizeof(struct ip6_hdr));
 964         if (icmp6 == NULL) {
 965                 ICMP6STAT_INC(icp6s_tooshort);
 966                 return (-1);
 967         }
 968 #endif
 969         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 970
 971         /* Detect the upper level protocol */
 972         {
 973                 void (*ctlfunc)(int, struct sockaddr *, void *);
 974                 u_int8_t nxt = eip6->ip6_nxt;
 975                 int eoff = off + sizeof(struct icmp6_hdr) +
 976                     sizeof(struct ip6_hdr);
 977                 struct ip6ctlparam ip6cp;
 978                 struct in6_addr *finaldst = NULL;
 979                 int icmp6type = icmp6->icmp6_type;
 980                 struct ip6_frag *fh;
 981                 struct ip6_rthdr *rth;
 982                 struct ip6_rthdr0 *rth0;
 983                 int rthlen;
 984
 985                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 986                         struct ip6_ext *eh;
 987
 988                         switch (nxt) {
 989                         case IPPROTO_HOPOPTS:
 990                         case IPPROTO_DSTOPTS:
 991                         case IPPROTO_AH:
 992 #ifndef PULLDOWN_TEST
 993                                 IP6_EXTHDR_CHECK(m, 0,
 994                                     eoff + sizeof(struct ip6_ext), -1);
 995                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
 996 #else
 997                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
 998                                     eoff, sizeof(*eh));
 999                                 if (eh == NULL) {
1000                                         ICMP6STAT_INC(icp6s_tooshort);
1001                                         return (-1);
1002                                 }
1003 #endif
1004
1005                                 if (nxt == IPPROTO_AH)
1006                                         eoff += (eh->ip6e_len + 2) << 2;
1007                                 else
1008                                         eoff += (eh->ip6e_len + 1) << 3;
1009                                 nxt = eh->ip6e_nxt;
1010                                 break;
1011                         case IPPROTO_ROUTING:
1012                                 /*
1013                                  * When the erroneous packet contains a
1014                                  * routing header, we should examine the
1015                                  * header to determine the final destination.
1016                                  * Otherwise, we can't properly update
1017                                  * information that depends on the final
1018                                  * destination (e.g. path MTU).
1019                                  */
1020 #ifndef PULLDOWN_TEST
1021                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
1022                                 rth = (struct ip6_rthdr *)
1023                                     (mtod(m, caddr_t) + eoff);
1024 #else
1025                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1026                                     eoff, sizeof(*rth));
1027                                 if (rth == NULL) {
1028                                         ICMP6STAT_INC(icp6s_tooshort);
1029                                         return (-1);
1030                                 }
1031 #endif
1032                                 rthlen = (rth->ip6r_len + 1) << 3;
1033                                 /*
1034                                  * XXX: currently there is no
1035                                  * officially defined type other
1036                                  * than type-0.
1037                                  * Note that if the segment left field
1038                                  * is 0, all intermediate hops must
1039                                  * have been passed.
1040                                  */
1041                                 if (rth->ip6r_segleft &&
1042                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1043                                         int hops;
1044
1045 #ifndef PULLDOWN_TEST
1046                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1047                                         rth0 = (struct ip6_rthdr0 *)
1048                                             (mtod(m, caddr_t) + eoff);
1049 #else
1050                                         IP6_EXTHDR_GET(rth0,
1051                                             struct ip6_rthdr0 *, m,
1052                                             eoff, rthlen);
1053                                         if (rth0 == NULL) {
1054                                                 ICMP6STAT_INC(icp6s_tooshort);
1055                                                 return (-1);
1056                                         }
1057 #endif
1058                                         /* just ignore a bogus header */
1059                                         if ((rth0->ip6r0_len % 2) == 0 &&
1060                                             (hops = rth0->ip6r0_len/2))
1061                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1062                                 }
1063                                 eoff += rthlen;
1064                                 nxt = rth->ip6r_nxt;
1065                                 break;
1066                         case IPPROTO_FRAGMENT:
1067 #ifndef PULLDOWN_TEST
1068                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1069                                     sizeof(struct ip6_frag), -1);
1070                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1071                                     eoff);
1072 #else
1073                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1074                                     eoff, sizeof(*fh));
1075                                 if (fh == NULL) {
1076                                         ICMP6STAT_INC(icp6s_tooshort);
1077                                         return (-1);
1078                                 }
1079 #endif
1080                                 /*
1081                                  * Data after a fragment header is meaningless
1082                                  * unless it is the first fragment, but
1083                                  * we'll go to the notify label for path MTU
1084                                  * discovery.
1085                                  */
1086                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1087                                         goto notify;
1088
1089                                 eoff += sizeof(struct ip6_frag);
1090                                 nxt = fh->ip6f_nxt;
1091                                 break;
1092                         default:
1093                                 /*
1094                                  * This case includes ESP and the No Next
1095                                  * Header.  In such cases going to the notify
1096                                  * label does not have any meaning
1097                                  * (i.e. ctlfunc will be NULL), but we go
1098                                  * anyway since we might have to update
1099                                  * path MTU information.
1100                                  */
1101                                 goto notify;
1102                         }
1103                 }
1104           notify:
1105 #ifndef PULLDOWN_TEST
1106                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1107 #else
1108                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1109                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1110                 if (icmp6 == NULL) {
1111                         ICMP6STAT_INC(icp6s_tooshort);
1112                         return (-1);
1113                 }
1114 #endif
1115
1116                 /*
1117                  * retrieve parameters from the inner IPv6 header, and convert
1118                  * them into sockaddr structures.
1119                  * XXX: there is no guarantee that the source or destination
1120                  * addresses of the inner packet are in the same scope as
1121                  * the addresses of the icmp packet.  But there is no other
1122                  * way to determine the zone.
1123                  */
1124                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1125
1126                 bzero(&icmp6dst, sizeof(icmp6dst));
1127                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1128                 icmp6dst.sin6_family = AF_INET6;
1129                 if (finaldst == NULL)
1130                         icmp6dst.sin6_addr = eip6->ip6_dst;
1131                 else
1132                         icmp6dst.sin6_addr = *finaldst;
1133                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1134                         goto freeit;
1135                 bzero(&icmp6src, sizeof(icmp6src));
1136                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1137                 icmp6src.sin6_family = AF_INET6;
1138                 icmp6src.sin6_addr = eip6->ip6_src;
1139                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1140                         goto freeit;
1141                 icmp6src.sin6_flowinfo =
1142                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1143
1144                 if (finaldst == NULL)
1145                         finaldst = &eip6->ip6_dst;
1146                 ip6cp.ip6c_m = m;
1147                 ip6cp.ip6c_icmp6 = icmp6;
1148                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1149                 ip6cp.ip6c_off = eoff;
1150                 ip6cp.ip6c_finaldst = finaldst;
1151                 ip6cp.ip6c_src = &icmp6src;
1152                 ip6cp.ip6c_nxt = nxt;
1153
1154                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1155                         notifymtu = ntohl(icmp6->icmp6_mtu);
1156                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1157                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1158                 }
1159
1160                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1161                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1162                 if (ctlfunc) {
1163                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1164                             &ip6cp);
1165                 }
1166         }
1167         *mp = m;
1168         return (0);
1169
1170   freeit:
1171         m_freem(m);
1172         return (-1);
1173 }
1174
1175 void
1176 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1177 {
1178         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1179         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1180         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1181         u_int mtu = ntohl(icmp6->icmp6_mtu);
1182         struct in_conninfo inc;
1183
1184 #if 0
1185         /*
1186          * RFC2460 section 5, last paragraph.
1187          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1188          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1189          * due to packet translator in the middle.
1190          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1191          * special handling.
1192          */
1193         if (mtu < IPV6_MMTU)
1194                 return;
1195 #endif
1196
1197         /*
1198          * we reject ICMPv6 too big with abnormally small value.
1199          * XXX what is the good definition of "abnormally small"?
1200          */
1201         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1202                 return;
1203
1204         if (!validated)
1205                 return;
1206
1207         /*
1208          * In case the suggested mtu is less than IPV6_MMTU, we
1209          * only need to remember that it was for above mentioned
1210          * "alwaysfrag" case.
1211          * Try to be as close to the spec as possible.
1212          */
1213         if (mtu < IPV6_MMTU)
1214                 mtu = IPV6_MMTU - 8;
1215
1216         bzero(&inc, sizeof(inc));
1217         inc.inc_fibnum = M_GETFIB(m);
1218         inc.inc_flags |= INC_ISIPV6;
1219         inc.inc6_faddr = *dst;
1220         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1221                 return;
1222
1223         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1224                 tcp_hc_updatemtu(&inc, mtu);
1225                 ICMP6STAT_INC(icp6s_pmtuchg);
1226         }
1227 }
1228
1229 /*
1230  * Process a Node Information Query packet, based on
1231  * draft-ietf-ipngwg-icmp-name-lookups-07.
1232  *
1233  * Spec incompatibilities:
1234  * - IPv6 Subject address handling
1235  * - IPv4 Subject address handling support missing
1236  * - Proxy reply (answer even if it's not for me)
1237  * - joins NI group address at in6_ifattach() time only, does not cope
1238  *   with hostname changes by sethostname(3)
1239  */
1240 static struct mbuf *
1241 ni6_input(struct mbuf *m, int off)
1242 {
1243         struct icmp6_nodeinfo *ni6, *nni6;
1244         struct mbuf *n = NULL;
1245         struct prison *pr;
1246         u_int16_t qtype;
1247         int subjlen;
1248         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1249         struct ni_reply_fqdn *fqdn;
1250         int addrs;              /* for NI_QTYPE_NODEADDR */
1251         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1252         struct in6_addr in6_subj; /* subject address */
1253         struct ip6_hdr *ip6;
1254         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1255         char *subj = NULL;
1256         struct in6_ifaddr *ia6 = NULL;
1257
1258         ip6 = mtod(m, struct ip6_hdr *);
1259 #ifndef PULLDOWN_TEST
1260         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1261 #else
1262         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1263         if (ni6 == NULL) {
1264                 /* m is already reclaimed */
1265                 return (NULL);
1266         }
1267 #endif
1268
1269         /*
1270          * Validate IPv6 source address.
1271          * The default configuration MUST be to refuse answering queries from
1272          * global-scope addresses according to RFC4602.
1273          * Notes:
1274          *  - it's not very clear what "refuse" means; this implementation
1275          *    simply drops it.
1276          *  - it's not very easy to identify global-scope (unicast) addresses
1277          *    since there are many prefixes for them.  It should be safer
1278          *    and in practice sufficient to check "all" but loopback and
1279          *    link-local (note that site-local unicast was deprecated and
1280          *    ULA is defined as global scope-wise)
1281          */
1282         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1283             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1284             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1285                 goto bad;
1286
1287         /*
1288          * Validate IPv6 destination address.
1289          *
1290          * The Responder must discard the Query without further processing
1291          * unless it is one of the Responder's unicast or anycast addresses, or
1292          * a link-local scope multicast address which the Responder has joined.
1293          * [RFC4602, Section 5.]
1294          */
1295         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1296                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1297                         goto bad;
1298                 /* else it's a link-local multicast, fine */
1299         } else {                /* unicast or anycast */
1300                 ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
1301                 if (ia6 == NULL)
1302                         goto bad; /* XXX impossible */
1303
1304                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1305                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1306                         ifa_free(&ia6->ia_ifa);
1307                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1308                                 "a temporary address in %s:%d",
1309                                __FILE__, __LINE__));
1310                         goto bad;
1311                 }
1312                 ifa_free(&ia6->ia_ifa);
1313         }
1314
1315         /* validate query Subject field. */
1316         qtype = ntohs(ni6->ni_qtype);
1317         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1318         switch (qtype) {
1319         case NI_QTYPE_NOOP:
1320         case NI_QTYPE_SUPTYPES:
1321                 /* 07 draft */
1322                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1323                         break;
1324                 /* FALLTHROUGH */
1325         case NI_QTYPE_FQDN:
1326         case NI_QTYPE_NODEADDR:
1327         case NI_QTYPE_IPV4ADDR:
1328                 switch (ni6->ni_code) {
1329                 case ICMP6_NI_SUBJ_IPV6:
1330 #if ICMP6_NI_SUBJ_IPV6 != 0
1331                 case 0:
1332 #endif
1333                         /*
1334                          * backward compatibility - try to accept 03 draft
1335                          * format, where no Subject is present.
1336                          */
1337                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1338                             subjlen == 0) {
1339                                 oldfqdn++;
1340                                 break;
1341                         }
1342 #if ICMP6_NI_SUBJ_IPV6 != 0
1343                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1344                                 goto bad;
1345 #endif
1346
1347                         if (subjlen != sizeof(struct in6_addr))
1348                                 goto bad;
1349
1350                         /*
1351                          * Validate Subject address.
1352                          *
1353                          * Not sure what exactly "address belongs to the node"
1354                          * means in the spec, is it just unicast, or what?
1355                          *
1356                          * At this moment we consider Subject address as
1357                          * "belong to the node" if the Subject address equals
1358                          * to the IPv6 destination address; validation for
1359                          * IPv6 destination address should have done enough
1360                          * check for us.
1361                          *
1362                          * We do not do proxy at this moment.
1363                          */
1364                         /* m_pulldown instead of copy? */
1365                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1366                             subjlen, (caddr_t)&in6_subj);
1367                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1368                                 goto bad;
1369
1370                         subj = (char *)&in6_subj;
1371                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1372                                 break;
1373
1374                         /*
1375                          * XXX if we are to allow other cases, we should really
1376                          * be careful about scope here.
1377                          * basically, we should disallow queries toward IPv6
1378                          * destination X with subject Y,
1379                          * if scope(X) > scope(Y).
1380                          * if we allow scope(X) > scope(Y), it will result in
1381                          * information leakage across scope boundary.
1382                          */
1383                         goto bad;
1384
1385                 case ICMP6_NI_SUBJ_FQDN:
1386                         /*
1387                          * Validate Subject name with gethostname(3).
1388                          *
1389                          * The behavior may need some debate, since:
1390                          * - we are not sure if the node has FQDN as
1391                          *   hostname (returned by gethostname(3)).
1392                          * - the code does wildcard match for truncated names.
1393                          *   however, we are not sure if we want to perform
1394                          *   wildcard match, if gethostname(3) side has
1395                          *   truncated hostname.
1396                          */
1397                         pr = curthread->td_ucred->cr_prison;
1398                         mtx_lock(&pr->pr_mtx);
1399                         n = ni6_nametodns(pr->pr_hostname,
1400                             strlen(pr->pr_hostname), 0);
1401                         mtx_unlock(&pr->pr_mtx);
1402                         if (!n || n->m_next || n->m_len == 0)
1403                                 goto bad;
1404                         IP6_EXTHDR_GET(subj, char *, m,
1405                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1406                         if (subj == NULL)
1407                                 goto bad;
1408                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1409                             n->m_len)) {
1410                                 goto bad;
1411                         }
1412                         m_freem(n);
1413                         n = NULL;
1414                         break;
1415
1416                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1417                 default:
1418                         goto bad;
1419                 }
1420                 break;
1421         }
1422
1423         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1424         switch (qtype) {
1425         case NI_QTYPE_FQDN:
1426                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1427                         goto bad;
1428                 break;
1429         case NI_QTYPE_NODEADDR:
1430         case NI_QTYPE_IPV4ADDR:
1431                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1432                         goto bad;
1433                 break;
1434         }
1435
1436         /* guess reply length */
1437         switch (qtype) {
1438         case NI_QTYPE_NOOP:
1439                 break;          /* no reply data */
1440         case NI_QTYPE_SUPTYPES:
1441                 replylen += sizeof(u_int32_t);
1442                 break;
1443         case NI_QTYPE_FQDN:
1444                 /* XXX will append an mbuf */
1445                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1446                 break;
1447         case NI_QTYPE_NODEADDR:
1448                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1449                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1450                     sizeof(u_int32_t))) > MCLBYTES)
1451                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1452                 break;
1453         case NI_QTYPE_IPV4ADDR:
1454                 /* unsupported - should respond with unknown Qtype? */
1455                 break;
1456         default:
1457                 /*
1458                  * XXX: We must return a reply with the ICMP6 code
1459                  * `unknown Qtype' in this case.  However we regard the case
1460                  * as an FQDN query for backward compatibility.
1461                  * Older versions set a random value to this field,
1462                  * so it rarely varies in the defined qtypes.
1463                  * But the mechanism is not reliable...
1464                  * maybe we should obsolete older versions.
1465                  */
1466                 qtype = NI_QTYPE_FQDN;
1467                 /* XXX will append an mbuf */
1468                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1469                 oldfqdn++;
1470                 break;
1471         }
1472
1473         /* Allocate an mbuf to reply. */
1474         if (replylen > MCLBYTES) {
1475                 /*
1476                  * XXX: should we try to allocate more? But MCLBYTES
1477                  * is probably much larger than IPV6_MMTU...
1478                  */
1479                 goto bad;
1480         }
1481         if (replylen > MHLEN)
1482                 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1483         else
1484                 n = m_gethdr(M_NOWAIT, m->m_type);
1485         if (n == NULL) {
1486                 m_freem(m);
1487                 return (NULL);
1488         }
1489         m_move_pkthdr(n, m); /* just for recvif and FIB */
1490         n->m_pkthdr.len = n->m_len = replylen;
1491
1492         /* copy mbuf header and IPv6 + Node Information base headers */
1493         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1494         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1495         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1496
1497         /* qtype dependent procedure */
1498         switch (qtype) {
1499         case NI_QTYPE_NOOP:
1500                 nni6->ni_code = ICMP6_NI_SUCCESS;
1501                 nni6->ni_flags = 0;
1502                 break;
1503         case NI_QTYPE_SUPTYPES:
1504         {
1505                 u_int32_t v;
1506                 nni6->ni_code = ICMP6_NI_SUCCESS;
1507                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1508                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1509                 v = (u_int32_t)htonl(0x0000000f);
1510                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1511                 break;
1512         }
1513         case NI_QTYPE_FQDN:
1514                 nni6->ni_code = ICMP6_NI_SUCCESS;
1515                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1516                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1517                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1518                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1519                 /*
1520                  * XXX do we really have FQDN in hostname?
1521                  */
1522                 pr = curthread->td_ucred->cr_prison;
1523                 mtx_lock(&pr->pr_mtx);
1524                 n->m_next = ni6_nametodns(pr->pr_hostname,
1525                     strlen(pr->pr_hostname), oldfqdn);
1526                 mtx_unlock(&pr->pr_mtx);
1527                 if (n->m_next == NULL)
1528                         goto bad;
1529                 /* XXX we assume that n->m_next is not a chain */
1530                 if (n->m_next->m_next != NULL)
1531                         goto bad;
1532                 n->m_pkthdr.len += n->m_next->m_len;
1533                 break;
1534         case NI_QTYPE_NODEADDR:
1535         {
1536                 int lenlim, copied;
1537
1538                 nni6->ni_code = ICMP6_NI_SUCCESS;
1539                 n->m_pkthdr.len = n->m_len =
1540                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1541                 lenlim = M_TRAILINGSPACE(n);
1542                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1543                 /* XXX: reset mbuf length */
1544                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1545                     sizeof(struct icmp6_nodeinfo) + copied;
1546                 break;
1547         }
1548         default:
1549                 break;          /* XXX impossible! */
1550         }
1551
1552         nni6->ni_type = ICMP6_NI_REPLY;
1553         m_freem(m);
1554         return (n);
1555
1556   bad:
1557         m_freem(m);
1558         if (n)
1559                 m_freem(n);
1560         return (NULL);
1561 }
1562
1563 /*
1564  * make a mbuf with DNS-encoded string.  no compression support.
1565  *
1566  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1567  * treated as truncated name (two \0 at the end).  this is a wild guess.
1568  *
1569  * old - return pascal string if non-zero
1570  */
1571 static struct mbuf *
1572 ni6_nametodns(const char *name, int namelen, int old)
1573 {
1574         struct mbuf *m;
1575         char *cp, *ep;
1576         const char *p, *q;
1577         int i, len, nterm;
1578
1579         if (old)
1580                 len = namelen + 1;
1581         else
1582                 len = MCLBYTES;
1583
1584         /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1585         if (len > MLEN)
1586                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1587         else
1588                 m = m_get(M_NOWAIT, MT_DATA);
1589         if (m == NULL)
1590                 goto fail;
1591
1592         if (old) {
1593                 m->m_len = len;
1594                 *mtod(m, char *) = namelen;
1595                 bcopy(name, mtod(m, char *) + 1, namelen);
1596                 return m;
1597         } else {
1598                 m->m_len = 0;
1599                 cp = mtod(m, char *);
1600                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1601
1602                 /* if not certain about my name, return empty buffer */
1603                 if (namelen == 0)
1604                         return m;
1605
1606                 /*
1607                  * guess if it looks like shortened hostname, or FQDN.
1608                  * shortened hostname needs two trailing "\0".
1609                  */
1610                 i = 0;
1611                 for (p = name; p < name + namelen; p++) {
1612                         if (*p && *p == '.')
1613                                 i++;
1614                 }
1615                 if (i < 2)
1616                         nterm = 2;
1617                 else
1618                         nterm = 1;
1619
1620                 p = name;
1621                 while (cp < ep && p < name + namelen) {
1622                         i = 0;
1623                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1624                                 i++;
1625                         /* result does not fit into mbuf */
1626                         if (cp + i + 1 >= ep)
1627                                 goto fail;
1628                         /*
1629                          * DNS label length restriction, RFC1035 page 8.
1630                          * "i == 0" case is included here to avoid returning
1631                          * 0-length label on "foo..bar".
1632                          */
1633                         if (i <= 0 || i >= 64)
1634                                 goto fail;
1635                         *cp++ = i;
1636                         bcopy(p, cp, i);
1637                         cp += i;
1638                         p = q;
1639                         if (p < name + namelen && *p == '.')
1640                                 p++;
1641                 }
1642                 /* termination */
1643                 if (cp + nterm >= ep)
1644                         goto fail;
1645                 while (nterm-- > 0)
1646                         *cp++ = '\0';
1647                 m->m_len = cp - mtod(m, char *);
1648                 return m;
1649         }
1650
1651         panic("should not reach here");
1652         /* NOTREACHED */
1653
1654  fail:
1655         if (m)
1656                 m_freem(m);
1657         return NULL;
1658 }
1659
1660 /*
1661  * check if two DNS-encoded string matches.  takes care of truncated
1662  * form (with \0\0 at the end).  no compression support.
1663  * XXX upper/lowercase match (see RFC2065)
1664  */
1665 static int
1666 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1667 {
1668         const char *a0, *b0;
1669         int l;
1670
1671         /* simplest case - need validation? */
1672         if (alen == blen && bcmp(a, b, alen) == 0)
1673                 return 1;
1674
1675         a0 = a;
1676         b0 = b;
1677
1678         /* termination is mandatory */
1679         if (alen < 2 || blen < 2)
1680                 return 0;
1681         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1682                 return 0;
1683         alen--;
1684         blen--;
1685
1686         while (a - a0 < alen && b - b0 < blen) {
1687                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1688                         return 0;
1689
1690                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1691                         return 0;
1692                 /* we don't support compression yet */
1693                 if (a[0] >= 64 || b[0] >= 64)
1694                         return 0;
1695
1696                 /* truncated case */
1697                 if (a[0] == 0 && a - a0 == alen - 1)
1698                         return 1;
1699                 if (b[0] == 0 && b - b0 == blen - 1)
1700                         return 1;
1701                 if (a[0] == 0 || b[0] == 0)
1702                         return 0;
1703
1704                 if (a[0] != b[0])
1705                         return 0;
1706                 l = a[0];
1707                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1708                         return 0;
1709                 if (bcmp(a + 1, b + 1, l) != 0)
1710                         return 0;
1711
1712                 a += 1 + l;
1713                 b += 1 + l;
1714         }
1715
1716         if (a - a0 == alen && b - b0 == blen)
1717                 return 1;
1718         else
1719                 return 0;
1720 }
1721
1722 /*
1723  * calculate the number of addresses to be returned in the node info reply.
1724  */
1725 static int
1726 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1727     struct in6_addr *subj)
1728 {
1729         struct ifnet *ifp;
1730         struct in6_ifaddr *ifa6;
1731         struct ifaddr *ifa;
1732         int addrs = 0, addrsofif, iffound = 0;
1733         int niflags = ni6->ni_flags;
1734
1735         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1736                 switch (ni6->ni_code) {
1737                 case ICMP6_NI_SUBJ_IPV6:
1738                         if (subj == NULL) /* must be impossible... */
1739                                 return (0);
1740                         break;
1741                 default:
1742                         /*
1743                          * XXX: we only support IPv6 subject address for
1744                          * this Qtype.
1745                          */
1746                         return (0);
1747                 }
1748         }
1749
1750         IFNET_RLOCK_NOSLEEP();
1751         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1752                 addrsofif = 0;
1753                 IF_ADDR_RLOCK(ifp);
1754                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1755                         if (ifa->ifa_addr->sa_family != AF_INET6)
1756                                 continue;
1757                         ifa6 = (struct in6_ifaddr *)ifa;
1758
1759                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1760                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1761                                 iffound = 1;
1762
1763                         /*
1764                          * IPv4-mapped addresses can only be returned by a
1765                          * Node Information proxy, since they represent
1766                          * addresses of IPv4-only nodes, which perforce do
1767                          * not implement this protocol.
1768                          * [icmp-name-lookups-07, Section 5.4]
1769                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1770                          * this function at this moment.
1771                          */
1772
1773                         /* What do we have to do about ::1? */
1774                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1775                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1776                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1777                                         continue;
1778                                 break;
1779                         case IPV6_ADDR_SCOPE_SITELOCAL:
1780                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1781                                         continue;
1782                                 break;
1783                         case IPV6_ADDR_SCOPE_GLOBAL:
1784                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1785                                         continue;
1786                                 break;
1787                         default:
1788                                 continue;
1789                         }
1790
1791                         /*
1792                          * check if anycast is okay.
1793                          * XXX: just experimental.  not in the spec.
1794                          */
1795                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1796                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1797                                 continue; /* we need only unicast addresses */
1798                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1799                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1800                                 continue;
1801                         }
1802                         addrsofif++; /* count the address */
1803                 }
1804                 IF_ADDR_RUNLOCK(ifp);
1805                 if (iffound) {
1806                         *ifpp = ifp;
1807                         IFNET_RUNLOCK_NOSLEEP();
1808                         return (addrsofif);
1809                 }
1810
1811                 addrs += addrsofif;
1812         }
1813         IFNET_RUNLOCK_NOSLEEP();
1814
1815         return (addrs);
1816 }
1817
1818 static int
1819 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1820     struct ifnet *ifp0, int resid)
1821 {
1822         struct ifnet *ifp;
1823         struct in6_ifaddr *ifa6;
1824         struct ifaddr *ifa;
1825         struct ifnet *ifp_dep = NULL;
1826         int copied = 0, allow_deprecated = 0;
1827         u_char *cp = (u_char *)(nni6 + 1);
1828         int niflags = ni6->ni_flags;
1829         u_int32_t ltime;
1830
1831         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1832                 return (0);     /* needless to copy */
1833
1834         IFNET_RLOCK_NOSLEEP();
1835         ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1836   again:
1837
1838         for (; ifp; ifp = TAILQ_NEXT(ifp, if_link)) {
1839                 IF_ADDR_RLOCK(ifp);
1840                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1841                         if (ifa->ifa_addr->sa_family != AF_INET6)
1842                                 continue;
1843                         ifa6 = (struct in6_ifaddr *)ifa;
1844
1845                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1846                             allow_deprecated == 0) {
1847                                 /*
1848                                  * prefererred address should be put before
1849                                  * deprecated addresses.
1850                                  */
1851
1852                                 /* record the interface for later search */
1853                                 if (ifp_dep == NULL)
1854                                         ifp_dep = ifp;
1855
1856                                 continue;
1857                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1858                             allow_deprecated != 0)
1859                                 continue; /* we now collect deprecated addrs */
1860
1861                         /* What do we have to do about ::1? */
1862                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1863                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1864                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1865                                         continue;
1866                                 break;
1867                         case IPV6_ADDR_SCOPE_SITELOCAL:
1868                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1869                                         continue;
1870                                 break;
1871                         case IPV6_ADDR_SCOPE_GLOBAL:
1872                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1873                                         continue;
1874                                 break;
1875                         default:
1876                                 continue;
1877                         }
1878
1879                         /*
1880                          * check if anycast is okay.
1881                          * XXX: just experimental.  not in the spec.
1882                          */
1883                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1884                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1885                                 continue;
1886                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1887                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1888                                 continue;
1889                         }
1890
1891                         /* now we can copy the address */
1892                         if (resid < sizeof(struct in6_addr) +
1893                             sizeof(u_int32_t)) {
1894                                 IF_ADDR_RUNLOCK(ifp);
1895                                 /*
1896                                  * We give up much more copy.
1897                                  * Set the truncate flag and return.
1898                                  */
1899                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1900                                 IFNET_RUNLOCK_NOSLEEP();
1901                                 return (copied);
1902                         }
1903
1904                         /*
1905                          * Set the TTL of the address.
1906                          * The TTL value should be one of the following
1907                          * according to the specification:
1908                          *
1909                          * 1. The remaining lifetime of a DHCP lease on the
1910                          *    address, or
1911                          * 2. The remaining Valid Lifetime of a prefix from
1912                          *    which the address was derived through Stateless
1913                          *    Autoconfiguration.
1914                          *
1915                          * Note that we currently do not support stateful
1916                          * address configuration by DHCPv6, so the former
1917                          * case can't happen.
1918                          */
1919                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1920                                 ltime = ND6_INFINITE_LIFETIME;
1921                         else {
1922                                 if (ifa6->ia6_lifetime.ia6t_expire >
1923                                     time_uptime)
1924                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1925                                 else
1926                                         ltime = 0;
1927                         }
1928
1929                         bcopy(&ltime, cp, sizeof(u_int32_t));
1930                         cp += sizeof(u_int32_t);
1931
1932                         /* copy the address itself */
1933                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1934                             sizeof(struct in6_addr));
1935                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1936                         cp += sizeof(struct in6_addr);
1937
1938                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1939                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1940                 }
1941                 IF_ADDR_RUNLOCK(ifp);
1942                 if (ifp0)       /* we need search only on the specified IF */
1943                         break;
1944         }
1945
1946         if (allow_deprecated == 0 && ifp_dep != NULL) {
1947                 ifp = ifp_dep;
1948                 allow_deprecated = 1;
1949
1950                 goto again;
1951         }
1952
1953         IFNET_RUNLOCK_NOSLEEP();
1954
1955         return (copied);
1956 }
1957
1958 /*
1959  * XXX almost dup'ed code with rip6_input.
1960  */
1961 static int
1962 icmp6_rip6_input(struct mbuf **mp, int off)
1963 {
1964         struct mbuf *m = *mp;
1965         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1966         struct inpcb *in6p;
1967         struct inpcb *last = NULL;
1968         struct sockaddr_in6 fromsa;
1969         struct icmp6_hdr *icmp6;
1970         struct mbuf *opts = NULL;
1971
1972 #ifndef PULLDOWN_TEST
1973         /* this is assumed to be safe. */
1974         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1975 #else
1976         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1977         if (icmp6 == NULL) {
1978                 /* m is already reclaimed */
1979                 return (IPPROTO_DONE);
1980         }
1981 #endif
1982
1983         /*
1984          * XXX: the address may have embedded scope zone ID, which should be
1985          * hidden from applications.
1986          */
1987         bzero(&fromsa, sizeof(fromsa));
1988         fromsa.sin6_family = AF_INET6;
1989         fromsa.sin6_len = sizeof(struct sockaddr_in6);
1990         fromsa.sin6_addr = ip6->ip6_src;
1991         if (sa6_recoverscope(&fromsa)) {
1992                 m_freem(m);
1993                 return (IPPROTO_DONE);
1994         }
1995
1996         INP_INFO_RLOCK(&V_ripcbinfo);
1997         LIST_FOREACH(in6p, &V_ripcb, inp_list) {
1998                 if ((in6p->inp_vflag & INP_IPV6) == 0)
1999                         continue;
2000                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
2001                         continue;
2002                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2003                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2004                         continue;
2005                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2006                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2007                         continue;
2008                 INP_RLOCK(in6p);
2009                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2010                     in6p->in6p_icmp6filt)) {
2011                         INP_RUNLOCK(in6p);
2012                         continue;
2013                 }
2014                 if (last != NULL) {
2015                         struct  mbuf *n = NULL;
2016
2017                         /*
2018                          * Recent network drivers tend to allocate a single
2019                          * mbuf cluster, rather than to make a couple of
2020                          * mbufs without clusters.  Also, since the IPv6 code
2021                          * path tries to avoid m_pullup(), it is highly
2022                          * probable that we still have an mbuf cluster here
2023                          * even though the necessary length can be stored in an
2024                          * mbuf's internal buffer.
2025                          * Meanwhile, the default size of the receive socket
2026                          * buffer for raw sockets is not so large.  This means
2027                          * the possibility of packet loss is relatively higher
2028                          * than before.  To avoid this scenario, we copy the
2029                          * received data to a separate mbuf that does not use
2030                          * a cluster, if possible.
2031                          * XXX: it is better to copy the data after stripping
2032                          * intermediate headers.
2033                          */
2034                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2035                             m->m_len <= MHLEN) {
2036                                 n = m_get(M_NOWAIT, m->m_type);
2037                                 if (n != NULL) {
2038                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2039                                                 bcopy(m->m_data, n->m_data,
2040                                                       m->m_len);
2041                                                 n->m_len = m->m_len;
2042                                         } else {
2043                                                 m_free(n);
2044                                                 n = NULL;
2045                                         }
2046                                 }
2047                         }
2048                         if (n != NULL ||
2049                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2050                                 if (last->inp_flags & INP_CONTROLOPTS)
2051                                         ip6_savecontrol(last, n, &opts);
2052                                 /* strip intermediate headers */
2053                                 m_adj(n, off);
2054                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2055                                 if (sbappendaddr_locked(
2056                                     &last->inp_socket->so_rcv,
2057                                     (struct sockaddr *)&fromsa, n, opts)
2058                                     == 0) {
2059                                         /* should notify about lost packet */
2060                                         m_freem(n);
2061                                         if (opts) {
2062                                                 m_freem(opts);
2063                                         }
2064                                         SOCKBUF_UNLOCK(
2065                                             &last->inp_socket->so_rcv);
2066                                 } else
2067                                         sorwakeup_locked(last->inp_socket);
2068                                 opts = NULL;
2069                         }
2070                         INP_RUNLOCK(last);
2071                 }
2072                 last = in6p;
2073         }
2074         INP_INFO_RUNLOCK(&V_ripcbinfo);
2075         if (last != NULL) {
2076                 if (last->inp_flags & INP_CONTROLOPTS)
2077                         ip6_savecontrol(last, m, &opts);
2078                 /* strip intermediate headers */
2079                 m_adj(m, off);
2080
2081                 /* avoid using mbuf clusters if possible (see above) */
2082                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2083                     m->m_len <= MHLEN) {
2084                         struct mbuf *n;
2085
2086                         n = m_get(M_NOWAIT, m->m_type);
2087                         if (n != NULL) {
2088                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2089                                         bcopy(m->m_data, n->m_data, m->m_len);
2090                                         n->m_len = m->m_len;
2091
2092                                         m_freem(m);
2093                                         m = n;
2094                                 } else {
2095                                         m_freem(n);
2096                                         n = NULL;
2097                                 }
2098                         }
2099                 }
2100                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2101                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2102                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2103                         m_freem(m);
2104                         if (opts)
2105                                 m_freem(opts);
2106                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2107                 } else
2108                         sorwakeup_locked(last->inp_socket);
2109                 INP_RUNLOCK(last);
2110         } else {
2111                 m_freem(m);
2112                 IP6STAT_DEC(ip6s_delivered);
2113         }
2114         return IPPROTO_DONE;
2115 }
2116
2117 /*
2118  * Reflect the ip6 packet back to the source.
2119  * OFF points to the icmp6 header, counted from the top of the mbuf.
2120  */
2121 void
2122 icmp6_reflect(struct mbuf *m, size_t off)
2123 {
2124         struct in6_addr src, *srcp = NULL;
2125         struct ip6_hdr *ip6;
2126         struct icmp6_hdr *icmp6;
2127         struct in6_ifaddr *ia = NULL;
2128         struct ifnet *outif = NULL;
2129         int plen;
2130         int type, code;
2131
2132         /* too short to reflect */
2133         if (off < sizeof(struct ip6_hdr)) {
2134                 nd6log((LOG_DEBUG,
2135                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2136                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2137                     __FILE__, __LINE__));
2138                 goto bad;
2139         }
2140
2141         /*
2142          * If there are extra headers between IPv6 and ICMPv6, strip
2143          * off that header first.
2144          */
2145 #ifdef DIAGNOSTIC
2146         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2147                 panic("assumption failed in icmp6_reflect");
2148 #endif
2149         if (off > sizeof(struct ip6_hdr)) {
2150                 size_t l;
2151                 struct ip6_hdr nip6;
2152
2153                 l = off - sizeof(struct ip6_hdr);
2154                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2155                 m_adj(m, l);
2156                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2157                 if (m->m_len < l) {
2158                         if ((m = m_pullup(m, l)) == NULL)
2159                                 return;
2160                 }
2161                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2162         } else /* off == sizeof(struct ip6_hdr) */ {
2163                 size_t l;
2164                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2165                 if (m->m_len < l) {
2166                         if ((m = m_pullup(m, l)) == NULL)
2167                                 return;
2168                 }
2169         }
2170         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2171         ip6 = mtod(m, struct ip6_hdr *);
2172         ip6->ip6_nxt = IPPROTO_ICMPV6;
2173         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2174         type = icmp6->icmp6_type; /* keep type for statistics */
2175         code = icmp6->icmp6_code; /* ditto. */
2176
2177         /*
2178          * If the incoming packet was addressed directly to us (i.e. unicast),
2179          * use dst as the src for the reply.
2180          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2181          * (for example) when we encounter an error while forwarding procedure
2182          * destined to a duplicated address of ours.
2183          */
2184         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2185                 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
2186                 if (ia != NULL && !(ia->ia6_flags &
2187                     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2188                         srcp = &ia->ia_addr.sin6_addr;
2189         }
2190
2191         if (srcp == NULL) {
2192                 int e;
2193                 struct sockaddr_in6 sin6;
2194                 struct route_in6 ro;
2195
2196                 /*
2197                  * This case matches to multicasts, our anycast, or unicasts
2198                  * that we do not own.  Select a source address based on the
2199                  * source address of the erroneous packet.
2200                  */
2201                 bzero(&sin6, sizeof(sin6));
2202                 sin6.sin6_family = AF_INET6;
2203                 sin6.sin6_len = sizeof(sin6);
2204                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2205
2206                 bzero(&ro, sizeof(ro));
2207                 e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2208                 if (ro.ro_rt)
2209                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2210                 if (e) {
2211                         char ip6buf[INET6_ADDRSTRLEN];
2212                         nd6log((LOG_DEBUG,
2213                             "icmp6_reflect: source can't be determined: "
2214                             "dst=%s, error=%d\n",
2215                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2216                         goto bad;
2217                 }
2218                 srcp = &src;
2219         }
2220         /*
2221          * ip6_input() drops a packet if its src is multicast.
2222          * So, the src is never multicast.
2223          */
2224         ip6->ip6_dst = ip6->ip6_src;
2225         ip6->ip6_src = *srcp;
2226         ip6->ip6_flow = 0;
2227         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2228         ip6->ip6_vfc |= IPV6_VERSION;
2229         ip6->ip6_nxt = IPPROTO_ICMPV6;
2230         if (outif)
2231                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2232         else if (m->m_pkthdr.rcvif) {
2233                 /* XXX: This may not be the outgoing interface */
2234                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2235         } else
2236                 ip6->ip6_hlim = V_ip6_defhlim;
2237
2238         icmp6->icmp6_cksum = 0;
2239         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2240             sizeof(struct ip6_hdr), plen);
2241
2242         /*
2243          * XXX option handling
2244          */
2245
2246         m->m_flags &= ~(M_BCAST|M_MCAST);
2247
2248         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2249         if (outif)
2250                 icmp6_ifoutstat_inc(outif, type, code);
2251
2252         if (ia != NULL)
2253                 ifa_free(&ia->ia_ifa);
2254         return;
2255
2256  bad:
2257         if (ia != NULL)
2258                 ifa_free(&ia->ia_ifa);
2259         m_freem(m);
2260         return;
2261 }
2262
2263 void
2264 icmp6_fasttimo(void)
2265 {
2266
2267         mld_fasttimo();
2268 }
2269
2270 void
2271 icmp6_slowtimo(void)
2272 {
2273
2274         mld_slowtimo();
2275 }
2276
2277 static const char *
2278 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2279     struct in6_addr *tgt6)
2280 {
2281         static char buf[1024];
2282         char ip6bufs[INET6_ADDRSTRLEN];
2283         char ip6bufd[INET6_ADDRSTRLEN];
2284         char ip6buft[INET6_ADDRSTRLEN];
2285         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2286             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2287             ip6_sprintf(ip6buft, tgt6));
2288         return buf;
2289 }
2290
2291 void
2292 icmp6_redirect_input(struct mbuf *m, int off)
2293 {
2294         struct ifnet *ifp;
2295         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2296         struct nd_redirect *nd_rd;
2297         int icmp6len = ntohs(ip6->ip6_plen);
2298         char *lladdr = NULL;
2299         int lladdrlen = 0;
2300         struct rtentry *rt = NULL;
2301         int is_router;
2302         int is_onlink;
2303         struct in6_addr src6 = ip6->ip6_src;
2304         struct in6_addr redtgt6;
2305         struct in6_addr reddst6;
2306         union nd_opts ndopts;
2307         char ip6buf[INET6_ADDRSTRLEN];
2308
2309         M_ASSERTPKTHDR(m);
2310         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2311
2312         ifp = m->m_pkthdr.rcvif;
2313
2314         /* XXX if we are router, we don't update route by icmp6 redirect */
2315         if (V_ip6_forwarding)
2316                 goto freeit;
2317         if (!V_icmp6_rediraccept)
2318                 goto freeit;
2319
2320 #ifndef PULLDOWN_TEST
2321         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2322         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2323 #else
2324         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2325         if (nd_rd == NULL) {
2326                 ICMP6STAT_INC(icp6s_tooshort);
2327                 return;
2328         }
2329 #endif
2330         redtgt6 = nd_rd->nd_rd_target;
2331         reddst6 = nd_rd->nd_rd_dst;
2332
2333         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2334             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2335                 goto freeit;
2336         }
2337
2338         /* validation */
2339         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2340                 nd6log((LOG_ERR,
2341                     "ICMP6 redirect sent from %s rejected; "
2342                     "must be from linklocal\n",
2343                     ip6_sprintf(ip6buf, &src6)));
2344                 goto bad;
2345         }
2346         if (ip6->ip6_hlim != 255) {
2347                 nd6log((LOG_ERR,
2348                     "ICMP6 redirect sent from %s rejected; "
2349                     "hlim=%d (must be 255)\n",
2350                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2351                 goto bad;
2352         }
2353     {
2354         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2355         struct sockaddr_in6 sin6;
2356         struct in6_addr *gw6;
2357
2358         bzero(&sin6, sizeof(sin6));
2359         sin6.sin6_family = AF_INET6;
2360         sin6.sin6_len = sizeof(struct sockaddr_in6);
2361         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2362         rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL, RT_DEFAULT_FIB);
2363         if (rt) {
2364                 if (rt->rt_gateway == NULL ||
2365                     rt->rt_gateway->sa_family != AF_INET6) {
2366                         RTFREE_LOCKED(rt);
2367                         nd6log((LOG_ERR,
2368                             "ICMP6 redirect rejected; no route "
2369                             "with inet6 gateway found for redirect dst: %s\n",
2370                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2371                         goto bad;
2372                 }
2373
2374                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2375                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2376                         RTFREE_LOCKED(rt);
2377                         nd6log((LOG_ERR,
2378                             "ICMP6 redirect rejected; "
2379                             "not equal to gw-for-src=%s (must be same): "
2380                             "%s\n",
2381                             ip6_sprintf(ip6buf, gw6),
2382                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2383                         goto bad;
2384                 }
2385         } else {
2386                 nd6log((LOG_ERR,
2387                     "ICMP6 redirect rejected; "
2388                     "no route found for redirect dst: %s\n",
2389                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2390                 goto bad;
2391         }
2392         RTFREE_LOCKED(rt);
2393         rt = NULL;
2394     }
2395         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2396                 nd6log((LOG_ERR,
2397                     "ICMP6 redirect rejected; "
2398                     "redirect dst must be unicast: %s\n",
2399                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2400                 goto bad;
2401         }
2402
2403         is_router = is_onlink = 0;
2404         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2405                 is_router = 1;  /* router case */
2406         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2407                 is_onlink = 1;  /* on-link destination case */
2408         if (!is_router && !is_onlink) {
2409                 nd6log((LOG_ERR,
2410                     "ICMP6 redirect rejected; "
2411                     "neither router case nor onlink case: %s\n",
2412                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2413                 goto bad;
2414         }
2415         /* validation passed */
2416
2417         icmp6len -= sizeof(*nd_rd);
2418         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2419         if (nd6_options(&ndopts) < 0) {
2420                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2421                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2422                 /* nd6_options have incremented stats */
2423                 goto freeit;
2424         }
2425
2426         if (ndopts.nd_opts_tgt_lladdr) {
2427                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2428                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2429         }
2430
2431         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2432                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2433                     "(if %d, icmp6 packet %d): %s\n",
2434                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2435                     ifp->if_addrlen, lladdrlen - 2,
2436                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2437                 goto bad;
2438         }
2439
2440         /* RFC 2461 8.3 */
2441         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2442             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2443
2444         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2445                 /* perform rtredirect */
2446                 struct sockaddr_in6 sdst;
2447                 struct sockaddr_in6 sgw;
2448                 struct sockaddr_in6 ssrc;
2449                 u_int fibnum;
2450
2451                 bzero(&sdst, sizeof(sdst));
2452                 bzero(&sgw, sizeof(sgw));
2453                 bzero(&ssrc, sizeof(ssrc));
2454                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2455                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2456                         sizeof(struct sockaddr_in6);
2457                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2458                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2459                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2460                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2461                         in6_rtredirect((struct sockaddr *)&sdst,
2462                             (struct sockaddr *)&sgw, (struct sockaddr *)NULL,
2463                             RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&ssrc,
2464                             fibnum);
2465         }
2466         /* finally update cached route in each socket via pfctlinput */
2467     {
2468         struct sockaddr_in6 sdst;
2469
2470         bzero(&sdst, sizeof(sdst));
2471         sdst.sin6_family = AF_INET6;
2472         sdst.sin6_len = sizeof(struct sockaddr_in6);
2473         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2474         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2475 #ifdef IPSEC
2476         key_sa_routechange((struct sockaddr *)&sdst);
2477 #endif /* IPSEC */
2478     }
2479
2480  freeit:
2481         m_freem(m);
2482         return;
2483
2484  bad:
2485         ICMP6STAT_INC(icp6s_badredirect);
2486         m_freem(m);
2487 }
2488
2489 void
2490 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2491 {
2492         struct ifnet *ifp;      /* my outgoing interface */
2493         struct in6_addr *ifp_ll6;
2494         struct in6_addr *router_ll6;
2495         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2496         struct mbuf *m = NULL;  /* newly allocated one */
2497         struct m_tag *mtag;
2498         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2499         struct nd_redirect *nd_rd;
2500         struct llentry *ln = NULL;
2501         size_t maxlen;
2502         u_char *p;
2503         struct ifnet *outif = NULL;
2504         struct sockaddr_in6 src_sa;
2505
2506         icmp6_errcount(ND_REDIRECT, 0);
2507
2508         /* if we are not router, we don't send icmp6 redirect */
2509         if (!V_ip6_forwarding)
2510                 goto fail;
2511
2512         /* sanity check */
2513         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2514                 goto fail;
2515
2516         /*
2517          * Address check:
2518          *  the source address must identify a neighbor, and
2519          *  the destination address must not be a multicast address
2520          *  [RFC 2461, sec 8.2]
2521          */
2522         sip6 = mtod(m0, struct ip6_hdr *);
2523         bzero(&src_sa, sizeof(src_sa));
2524         src_sa.sin6_family = AF_INET6;
2525         src_sa.sin6_len = sizeof(src_sa);
2526         src_sa.sin6_addr = sip6->ip6_src;
2527         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2528                 goto fail;
2529         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2530                 goto fail;      /* what should we do here? */
2531
2532         /* rate limit */
2533         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2534                 goto fail;
2535
2536         /*
2537          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2538          * we almost always ask for an mbuf cluster for simplicity.
2539          * (MHLEN < IPV6_MMTU is almost always true)
2540          */
2541 #if IPV6_MMTU >= MCLBYTES
2542 # error assumption failed about IPV6_MMTU and MCLBYTES
2543 #endif
2544         m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2545         if (m == NULL)
2546                 goto fail;
2547         M_SETFIB(m, rt->rt_fibnum);
2548         maxlen = M_TRAILINGSPACE(m);
2549         maxlen = min(IPV6_MMTU, maxlen);
2550         /* just for safety */
2551         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2552             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2553                 goto fail;
2554         }
2555
2556         {
2557                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2558                 struct in6_ifaddr *ia;
2559                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2560                                                  IN6_IFF_NOTREADY|
2561                                                  IN6_IFF_ANYCAST)) == NULL)
2562                         goto fail;
2563                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2564                 /* XXXRW: reference released prematurely. */
2565                 ifa_free(&ia->ia_ifa);
2566         }
2567
2568         /* get ip6 linklocal address for the router. */
2569         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2570                 struct sockaddr_in6 *sin6;
2571                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2572                 router_ll6 = &sin6->sin6_addr;
2573                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2574                         router_ll6 = (struct in6_addr *)NULL;
2575         } else
2576                 router_ll6 = (struct in6_addr *)NULL;
2577
2578         /* ip6 */
2579         ip6 = mtod(m, struct ip6_hdr *);
2580         ip6->ip6_flow = 0;
2581         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2582         ip6->ip6_vfc |= IPV6_VERSION;
2583         /* ip6->ip6_plen will be set later */
2584         ip6->ip6_nxt = IPPROTO_ICMPV6;
2585         ip6->ip6_hlim = 255;
2586         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2587         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2588         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2589
2590         /* ND Redirect */
2591         nd_rd = (struct nd_redirect *)(ip6 + 1);
2592         nd_rd->nd_rd_type = ND_REDIRECT;
2593         nd_rd->nd_rd_code = 0;
2594         nd_rd->nd_rd_reserved = 0;
2595         if (rt->rt_flags & RTF_GATEWAY) {
2596                 /*
2597                  * nd_rd->nd_rd_target must be a link-local address in
2598                  * better router cases.
2599                  */
2600                 if (!router_ll6)
2601                         goto fail;
2602                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2603                     sizeof(nd_rd->nd_rd_target));
2604                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2605                     sizeof(nd_rd->nd_rd_dst));
2606         } else {
2607                 /* make sure redtgt == reddst */
2608                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2609                     sizeof(nd_rd->nd_rd_target));
2610                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2611                     sizeof(nd_rd->nd_rd_dst));
2612         }
2613
2614         p = (u_char *)(nd_rd + 1);
2615
2616         if (!router_ll6)
2617                 goto nolladdropt;
2618
2619         {
2620                 /* target lladdr option */
2621                 int len;
2622                 struct nd_opt_hdr *nd_opt;
2623                 char *lladdr;
2624
2625                 IF_AFDATA_RLOCK(ifp);
2626                 ln = nd6_lookup(router_ll6, 0, ifp);
2627                 IF_AFDATA_RUNLOCK(ifp);
2628                 if (ln == NULL)
2629                         goto nolladdropt;
2630
2631                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2632                 len = (len + 7) & ~7;   /* round by 8 */
2633                 /* safety check */
2634                 if (len + (p - (u_char *)ip6) > maxlen)
2635                         goto nolladdropt;
2636
2637                 if (ln->la_flags & LLE_VALID) {
2638                         nd_opt = (struct nd_opt_hdr *)p;
2639                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2640                         nd_opt->nd_opt_len = len >> 3;
2641                         lladdr = (char *)(nd_opt + 1);
2642                         bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2643                         p += len;
2644                 }
2645         }
2646 nolladdropt:
2647         if (ln != NULL)
2648                 LLE_RUNLOCK(ln);
2649
2650         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2651
2652         /* just to be safe */
2653 #ifdef M_DECRYPTED      /*not openbsd*/
2654         if (m0->m_flags & M_DECRYPTED)
2655                 goto noredhdropt;
2656 #endif
2657         if (p - (u_char *)ip6 > maxlen)
2658                 goto noredhdropt;
2659
2660         {
2661                 /* redirected header option */
2662                 int len;
2663                 struct nd_opt_rd_hdr *nd_opt_rh;
2664
2665                 /*
2666                  * compute the maximum size for icmp6 redirect header option.
2667                  * XXX room for auth header?
2668                  */
2669                 len = maxlen - (p - (u_char *)ip6);
2670                 len &= ~7;
2671
2672                 /* This is just for simplicity. */
2673                 if (m0->m_pkthdr.len != m0->m_len) {
2674                         if (m0->m_next) {
2675                                 m_freem(m0->m_next);
2676                                 m0->m_next = NULL;
2677                         }
2678                         m0->m_pkthdr.len = m0->m_len;
2679                 }
2680
2681                 /*
2682                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2683                  * about padding/truncate rule for the original IP packet.
2684                  * From the discussion on IPv6imp in Feb 1999,
2685                  * the consensus was:
2686                  * - "attach as much as possible" is the goal
2687                  * - pad if not aligned (original size can be guessed by
2688                  *   original ip6 header)
2689                  * Following code adds the padding if it is simple enough,
2690                  * and truncates if not.
2691                  */
2692                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2693                         panic("assumption failed in %s:%d", __FILE__,
2694                             __LINE__);
2695
2696                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2697                         /* not enough room, truncate */
2698                         m0->m_pkthdr.len = m0->m_len = len -
2699                             sizeof(*nd_opt_rh);
2700                 } else {
2701                         /* enough room, pad or truncate */
2702                         size_t extra;
2703
2704                         extra = m0->m_pkthdr.len % 8;
2705                         if (extra) {
2706                                 /* pad if easy enough, truncate if not */
2707                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2708                                         /* pad */
2709                                         m0->m_len += (8 - extra);
2710                                         m0->m_pkthdr.len += (8 - extra);
2711                                 } else {
2712                                         /* truncate */
2713                                         m0->m_pkthdr.len -= extra;
2714                                         m0->m_len -= extra;
2715                                 }
2716                         }
2717                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2718                         m0->m_pkthdr.len = m0->m_len = len -
2719                             sizeof(*nd_opt_rh);
2720                 }
2721
2722                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2723                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2724                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2725                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2726                 p += sizeof(*nd_opt_rh);
2727                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2728
2729                 /* connect m0 to m */
2730                 m_tag_delete_chain(m0, NULL);
2731                 m0->m_flags &= ~M_PKTHDR;
2732                 m->m_next = m0;
2733                 m->m_pkthdr.len = m->m_len + m0->m_len;
2734                 m0 = NULL;
2735         }
2736 noredhdropt:;
2737         if (m0) {
2738                 m_freem(m0);
2739                 m0 = NULL;
2740         }
2741
2742         /* XXX: clear embedded link IDs in the inner header */
2743         in6_clearscope(&sip6->ip6_src);
2744         in6_clearscope(&sip6->ip6_dst);
2745         in6_clearscope(&nd_rd->nd_rd_target);
2746         in6_clearscope(&nd_rd->nd_rd_dst);
2747
2748         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2749
2750         nd_rd->nd_rd_cksum = 0;
2751         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2752             sizeof(*ip6), ntohs(ip6->ip6_plen));
2753
2754         if (send_sendso_input_hook != NULL) {
2755                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2756                         M_NOWAIT);
2757                 if (mtag == NULL)
2758                         goto fail;
2759                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2760                 m_tag_prepend(m, mtag);
2761         }
2762
2763         /* send the packet to outside... */
2764         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2765         if (outif) {
2766                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2767                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2768         }
2769         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2770
2771         return;
2772
2773 fail:
2774         if (m)
2775                 m_freem(m);
2776         if (m0)
2777                 m_freem(m0);
2778 }
2779
2780 /*
2781  * ICMPv6 socket option processing.
2782  */
2783 int
2784 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2785 {
2786         int error = 0;
2787         int optlen;
2788         struct inpcb *inp = sotoinpcb(so);
2789         int level, op, optname;
2790
2791         if (sopt) {
2792                 level = sopt->sopt_level;
2793                 op = sopt->sopt_dir;
2794                 optname = sopt->sopt_name;
2795                 optlen = sopt->sopt_valsize;
2796         } else
2797                 level = op = optname = optlen = 0;
2798
2799         if (level != IPPROTO_ICMPV6) {
2800                 return EINVAL;
2801         }
2802
2803         switch (op) {
2804         case PRCO_SETOPT:
2805                 switch (optname) {
2806                 case ICMP6_FILTER:
2807                     {
2808                         struct icmp6_filter ic6f;
2809
2810                         if (optlen != sizeof(ic6f)) {
2811                                 error = EMSGSIZE;
2812                                 break;
2813                         }
2814                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2815                         if (error == 0) {
2816                                 INP_WLOCK(inp);
2817                                 *inp->in6p_icmp6filt = ic6f;
2818                                 INP_WUNLOCK(inp);
2819                         }
2820                         break;
2821                     }
2822
2823                 default:
2824                         error = ENOPROTOOPT;
2825                         break;
2826                 }
2827                 break;
2828
2829         case PRCO_GETOPT:
2830                 switch (optname) {
2831                 case ICMP6_FILTER:
2832                     {
2833                         struct icmp6_filter ic6f;
2834
2835                         INP_RLOCK(inp);
2836                         ic6f = *inp->in6p_icmp6filt;
2837                         INP_RUNLOCK(inp);
2838                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2839                         break;
2840                     }
2841
2842                 default:
2843                         error = ENOPROTOOPT;
2844                         break;
2845                 }
2846                 break;
2847         }
2848
2849         return (error);
2850 }
2851
2852 /*
2853  * Perform rate limit check.
2854  * Returns 0 if it is okay to send the icmp6 packet.
2855  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2856  * limitation.
2857  *
2858  * XXX per-destination/type check necessary?
2859  *
2860  * dst - not used at this moment
2861  * type - not used at this moment
2862  * code - not used at this moment
2863  */
2864 static int
2865 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2866     const int code)
2867 {
2868         int ret;
2869
2870         ret = 0;        /* okay to send */
2871
2872         /* PPS limit */
2873         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2874             V_icmp6errppslim)) {
2875                 /* The packet is subject to rate limit */
2876                 ret++;
2877         }
2878
2879         return ret;
2880 }