sys/netinet6/icmp6.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-3-Clause
   3  *
   4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 3. Neither the name of the project nor the names of its contributors
  16  *    may be used to endorse or promote products derived from this software
  17  *    without specific prior written permission.
  18  *
  19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  29  * SUCH DAMAGE.
  30  *
  31  *      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
  32  */
  33
  34 /*-
  35  * Copyright (c) 1982, 1986, 1988, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. Neither the name of the University nor the names of its contributors
  47  *    may be used to endorse or promote products derived from this software
  48  *    without specific prior written permission.
  49  *
  50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  60  * SUCH DAMAGE.
  61  *
  62  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
  63  */
  64
  65 #include <sys/cdefs.h>
  66 __FBSDID("$FreeBSD$");
  67
  68 #define MBUF_PRIVATE    /* XXXRW: Optimisation tries to avoid M_EXT mbufs */
  69
  70 #include "opt_inet.h"
  71 #include "opt_inet6.h"
  72
  73 #include <sys/param.h>
  74 #include <sys/domain.h>
  75 #include <sys/jail.h>
  76 #include <sys/kernel.h>
  77 #include <sys/lock.h>
  78 #include <sys/malloc.h>
  79 #include <sys/mbuf.h>
  80 #include <sys/proc.h>
  81 #include <sys/protosw.h>
  82 #include <sys/signalvar.h>
  83 #include <sys/socket.h>
  84 #include <sys/socketvar.h>
  85 #include <sys/sx.h>
  86 #include <sys/syslog.h>
  87 #include <sys/systm.h>
  88 #include <sys/time.h>
  89
  90 #include <net/if.h>
  91 #include <net/if_var.h>
  92 #include <net/if_dl.h>
  93 #include <net/if_llatbl.h>
  94 #include <net/if_types.h>
  95 #include <net/route.h>
  96 #include <net/route/route_ctl.h>
  97 #include <net/route/nhop.h>
  98 #include <net/vnet.h>
  99
 100 #include <netinet/in.h>
 101 #include <netinet/in_pcb.h>
 102 #include <netinet/in_var.h>
 103 #include <netinet/ip6.h>
 104 #include <netinet/icmp6.h>
 105 #include <netinet/tcp_var.h>
 106
 107 #include <netinet6/in6_fib.h>
 108 #include <netinet6/in6_ifattach.h>
 109 #include <netinet6/in6_pcb.h>
 110 #include <netinet6/ip6protosw.h>
 111 #include <netinet6/ip6_var.h>
 112 #include <netinet6/scope6_var.h>
 113 #include <netinet6/mld6_var.h>
 114 #include <netinet6/nd6.h>
 115 #include <netinet6/send.h>
 116
 117 extern struct domain inet6domain;
 118
 119 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
 120 VNET_PCPUSTAT_SYSINIT(icmp6stat);
 121
 122 #ifdef VIMAGE
 123 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
 124 #endif /* VIMAGE */
 125
 126 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
 127 VNET_DECLARE(int, icmp6errppslim);
 128 VNET_DEFINE_STATIC(int, icmp6errpps_count) = 0;
 129 VNET_DEFINE_STATIC(struct timeval, icmp6errppslim_last);
 130 VNET_DECLARE(int, icmp6_nodeinfo);
 131
 132 #define V_ripcbinfo                     VNET(ripcbinfo)
 133 #define V_icmp6errppslim                VNET(icmp6errppslim)
 134 #define V_icmp6errpps_count             VNET(icmp6errpps_count)
 135 #define V_icmp6errppslim_last           VNET(icmp6errppslim_last)
 136 #define V_icmp6_nodeinfo                VNET(icmp6_nodeinfo)
 137
 138 static void icmp6_errcount(int, int);
 139 static int icmp6_rip6_input(struct mbuf **, int);
 140 static void icmp6_reflect(struct mbuf *, size_t);
 141 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
 142 static const char *icmp6_redirect_diag(struct in6_addr *,
 143         struct in6_addr *, struct in6_addr *);
 144 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *);
 145 static struct mbuf *ni6_nametodns(const char *, int, int);
 146 static int ni6_dnsmatch(const char *, int, const char *, int);
 147 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
 148                           struct ifnet **, struct in6_addr *);
 149 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
 150                                 struct ifnet *, int);
 151 static int icmp6_notify_error(struct mbuf **, int, int, int);
 152
 153 /*
 154  * Kernel module interface for updating icmp6stat.  The argument is an index
 155  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
 156  * general layout of icmp6stat into the caller, it doesn't encode its
 157  * location, so that future changes to add, for example, per-CPU stats
 158  * support won't cause binary compatibility problems for kernel modules.
 159  */
 160 void
 161 kmod_icmp6stat_inc(int statnum)
 162 {
 163
 164         counter_u64_add(VNET(icmp6stat)[statnum], 1);
 165 }
 166
 167 static void
 168 icmp6_errcount(int type, int code)
 169 {
 170         switch (type) {
 171         case ICMP6_DST_UNREACH:
 172                 switch (code) {
 173                 case ICMP6_DST_UNREACH_NOROUTE:
 174                         ICMP6STAT_INC(icp6s_odst_unreach_noroute);
 175                         return;
 176                 case ICMP6_DST_UNREACH_ADMIN:
 177                         ICMP6STAT_INC(icp6s_odst_unreach_admin);
 178                         return;
 179                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 180                         ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
 181                         return;
 182                 case ICMP6_DST_UNREACH_ADDR:
 183                         ICMP6STAT_INC(icp6s_odst_unreach_addr);
 184                         return;
 185                 case ICMP6_DST_UNREACH_NOPORT:
 186                         ICMP6STAT_INC(icp6s_odst_unreach_noport);
 187                         return;
 188                 }
 189                 break;
 190         case ICMP6_PACKET_TOO_BIG:
 191                 ICMP6STAT_INC(icp6s_opacket_too_big);
 192                 return;
 193         case ICMP6_TIME_EXCEEDED:
 194                 switch (code) {
 195                 case ICMP6_TIME_EXCEED_TRANSIT:
 196                         ICMP6STAT_INC(icp6s_otime_exceed_transit);
 197                         return;
 198                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 199                         ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
 200                         return;
 201                 }
 202                 break;
 203         case ICMP6_PARAM_PROB:
 204                 switch (code) {
 205                 case ICMP6_PARAMPROB_HEADER:
 206                         ICMP6STAT_INC(icp6s_oparamprob_header);
 207                         return;
 208                 case ICMP6_PARAMPROB_NEXTHEADER:
 209                         ICMP6STAT_INC(icp6s_oparamprob_nextheader);
 210                         return;
 211                 case ICMP6_PARAMPROB_OPTION:
 212                         ICMP6STAT_INC(icp6s_oparamprob_option);
 213                         return;
 214                 }
 215                 break;
 216         case ND_REDIRECT:
 217                 ICMP6STAT_INC(icp6s_oredirect);
 218                 return;
 219         }
 220         ICMP6STAT_INC(icp6s_ounknown);
 221 }
 222
 223 /*
 224  * A wrapper function for icmp6_error() necessary when the erroneous packet
 225  * may not contain enough scope zone information.
 226  */
 227 void
 228 icmp6_error2(struct mbuf *m, int type, int code, int param,
 229     struct ifnet *ifp)
 230 {
 231         struct ip6_hdr *ip6;
 232
 233         if (ifp == NULL)
 234                 return;
 235
 236         if (m->m_len < sizeof(struct ip6_hdr)) {
 237                 m = m_pullup(m, sizeof(struct ip6_hdr));
 238                 if (m == NULL) {
 239                         IP6STAT_INC(ip6s_exthdrtoolong);
 240                         return;
 241                 }
 242         }
 243         ip6 = mtod(m, struct ip6_hdr *);
 244
 245         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
 246                 return;
 247         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
 248                 return;
 249
 250         icmp6_error(m, type, code, param);
 251 }
 252
 253 /*
 254  * Generate an error packet of type error in response to bad IP6 packet.
 255  */
 256 void
 257 icmp6_error(struct mbuf *m, int type, int code, int param)
 258 {
 259         struct ip6_hdr *oip6, *nip6;
 260         struct icmp6_hdr *icmp6;
 261         struct epoch_tracker et;
 262         u_int preplen;
 263         int off;
 264         int nxt;
 265
 266         ICMP6STAT_INC(icp6s_error);
 267
 268         /* count per-type-code statistics */
 269         icmp6_errcount(type, code);
 270
 271 #ifdef M_DECRYPTED      /*not openbsd*/
 272         if (m->m_flags & M_DECRYPTED) {
 273                 ICMP6STAT_INC(icp6s_canterror);
 274                 goto freeit;
 275         }
 276 #endif
 277
 278         if (m->m_len < sizeof(struct ip6_hdr)) {
 279                 m = m_pullup(m, sizeof(struct ip6_hdr));
 280                 if (m == NULL) {
 281                         IP6STAT_INC(ip6s_exthdrtoolong);
 282                         return;
 283                 }
 284         }
 285         oip6 = mtod(m, struct ip6_hdr *);
 286
 287         /*
 288          * If the destination address of the erroneous packet is a multicast
 289          * address, or the packet was sent using link-layer multicast,
 290          * we should basically suppress sending an error (RFC 2463, Section
 291          * 2.4).
 292          * We have two exceptions (the item e.2 in that section):
 293          * - the Packet Too Big message can be sent for path MTU discovery.
 294          * - the Parameter Problem Message that can be allowed an icmp6 error
 295          *   in the option type field.  This check has been done in
 296          *   ip6_unknown_opt(), so we can just check the type and code.
 297          */
 298         if ((m->m_flags & (M_BCAST|M_MCAST) ||
 299              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
 300             (type != ICMP6_PACKET_TOO_BIG &&
 301              (type != ICMP6_PARAM_PROB ||
 302               code != ICMP6_PARAMPROB_OPTION)))
 303                 goto freeit;
 304
 305         /*
 306          * RFC 2463, 2.4 (e.5): source address check.
 307          * XXX: the case of anycast source?
 308          */
 309         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
 310             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
 311                 goto freeit;
 312
 313         /*
 314          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
 315          * don't do it.
 316          */
 317         nxt = -1;
 318         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
 319         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
 320                 struct icmp6_hdr *icp;
 321
 322                 if (m->m_len < off + sizeof(struct icmp6_hdr)) {
 323                         m = m_pullup(m, off + sizeof(struct icmp6_hdr));
 324                         if (m == NULL) {
 325                                 IP6STAT_INC(ip6s_exthdrtoolong);
 326                                 return;
 327                         }
 328                 }
 329                 oip6 = mtod(m, struct ip6_hdr *);
 330                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 331
 332                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
 333                     icp->icmp6_type == ND_REDIRECT) {
 334                         /*
 335                          * ICMPv6 error
 336                          * Special case: for redirect (which is
 337                          * informational) we must not send icmp6 error.
 338                          */
 339                         ICMP6STAT_INC(icp6s_canterror);
 340                         goto freeit;
 341                 } else {
 342                         /* ICMPv6 informational - send the error */
 343                 }
 344         } else {
 345                 /* non-ICMPv6 - send the error */
 346         }
 347
 348         /* Finally, do rate limitation check. */
 349         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
 350                 ICMP6STAT_INC(icp6s_toofreq);
 351                 goto freeit;
 352         }
 353
 354         /*
 355          * OK, ICMP6 can be generated.
 356          */
 357
 358         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
 359                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
 360
 361         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
 362         M_PREPEND(m, preplen, M_NOWAIT);        /* FIB is also copied over. */
 363         if (m == NULL) {
 364                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
 365                 return;
 366         }
 367
 368         nip6 = mtod(m, struct ip6_hdr *);
 369         nip6->ip6_src  = oip6->ip6_src;
 370         nip6->ip6_dst  = oip6->ip6_dst;
 371
 372         in6_clearscope(&oip6->ip6_src);
 373         in6_clearscope(&oip6->ip6_dst);
 374
 375         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
 376         icmp6->icmp6_type = type;
 377         icmp6->icmp6_code = code;
 378         icmp6->icmp6_pptr = htonl((u_int32_t)param);
 379
 380         ICMP6STAT_INC(icp6s_outhist[type]);
 381         NET_EPOCH_ENTER(et);
 382         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
 383         NET_EPOCH_EXIT(et);
 384
 385         return;
 386
 387   freeit:
 388         /*
 389          * If we can't tell whether or not we can generate ICMP6, free it.
 390          */
 391         m_freem(m);
 392 }
 393
 394 /*
 395  * Process a received ICMP6 message.
 396  */
 397 int
 398 icmp6_input(struct mbuf **mp, int *offp, int proto)
 399 {
 400         struct mbuf *m, *n;
 401         struct ifnet *ifp;
 402         struct ip6_hdr *ip6, *nip6;
 403         struct icmp6_hdr *icmp6, *nicmp6;
 404         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
 405         int code, error, icmp6len, ip6len, noff, off, sum;
 406
 407         NET_EPOCH_ASSERT();
 408
 409         m = *mp;
 410         off = *offp;
 411
 412         if (m->m_len < off + sizeof(struct icmp6_hdr)) {
 413                 m = m_pullup(m, off + sizeof(struct icmp6_hdr));
 414                 if (m == NULL) {
 415                         IP6STAT_INC(ip6s_exthdrtoolong);
 416                         *mp = m;
 417                         return (IPPROTO_DONE);
 418                 }
 419         }
 420
 421         /*
 422          * Locate icmp6 structure in mbuf, and check
 423          * that not corrupted and of at least minimum length
 424          */
 425
 426         icmp6len = m->m_pkthdr.len - off;
 427         if (icmp6len < sizeof(struct icmp6_hdr)) {
 428                 ICMP6STAT_INC(icp6s_tooshort);
 429                 goto freeit;
 430         }
 431
 432         ip6 = mtod(m, struct ip6_hdr *);
 433         ifp = m->m_pkthdr.rcvif;
 434         /*
 435          * Check multicast group membership.
 436          * Note: SSM filters are not applied for ICMPv6 traffic.
 437          */
 438         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 439                 struct in6_multi        *inm;
 440
 441                 inm = in6m_lookup(ifp, &ip6->ip6_dst);
 442                 if (inm == NULL) {
 443                         IP6STAT_INC(ip6s_notmember);
 444                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
 445                         goto freeit;
 446                 }
 447         }
 448
 449         /* Calculate the checksum. */
 450         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
 451         code = icmp6->icmp6_code;
 452         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
 453                 nd6log((LOG_ERR,
 454                     "ICMP6 checksum error(%d|%x) %s\n",
 455                     icmp6->icmp6_type, sum,
 456                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
 457                 ICMP6STAT_INC(icp6s_checksum);
 458                 goto freeit;
 459         }
 460
 461         ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
 462         icmp6_ifstat_inc(ifp, ifs6_in_msg);
 463         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
 464                 icmp6_ifstat_inc(ifp, ifs6_in_error);
 465
 466         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
 467         switch (icmp6->icmp6_type) {
 468         case ICMP6_DST_UNREACH:
 469                 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
 470                 switch (code) {
 471                 case ICMP6_DST_UNREACH_NOROUTE:
 472                 case ICMP6_DST_UNREACH_ADDR:    /* PRC_HOSTDEAD is a DOS */
 473                         code = PRC_UNREACH_NET;
 474                         break;
 475                 case ICMP6_DST_UNREACH_ADMIN:
 476                         icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
 477                         code = PRC_UNREACH_ADMIN_PROHIB;
 478                         break;
 479                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 480                         /* I mean "source address was incorrect." */
 481                         code = PRC_PARAMPROB;
 482                         break;
 483                 case ICMP6_DST_UNREACH_NOPORT:
 484                         code = PRC_UNREACH_PORT;
 485                         break;
 486                 default:
 487                         goto badcode;
 488                 }
 489                 goto deliver;
 490                 break;
 491
 492         case ICMP6_PACKET_TOO_BIG:
 493                 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
 494
 495                 /* validation is made in icmp6_mtudisc_update */
 496
 497                 code = PRC_MSGSIZE;
 498
 499                 /*
 500                  * Updating the path MTU will be done after examining
 501                  * intermediate extension headers.
 502                  */
 503                 goto deliver;
 504                 break;
 505
 506         case ICMP6_TIME_EXCEEDED:
 507                 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
 508                 switch (code) {
 509                 case ICMP6_TIME_EXCEED_TRANSIT:
 510                         code = PRC_TIMXCEED_INTRANS;
 511                         break;
 512                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 513                         code = PRC_TIMXCEED_REASS;
 514                         break;
 515                 default:
 516                         goto badcode;
 517                 }
 518                 goto deliver;
 519                 break;
 520
 521         case ICMP6_PARAM_PROB:
 522                 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
 523                 switch (code) {
 524                 case ICMP6_PARAMPROB_NEXTHEADER:
 525                         code = PRC_UNREACH_PROTOCOL;
 526                         break;
 527                 case ICMP6_PARAMPROB_HEADER:
 528                 case ICMP6_PARAMPROB_OPTION:
 529                         code = PRC_PARAMPROB;
 530                         break;
 531                 default:
 532                         goto badcode;
 533                 }
 534                 goto deliver;
 535                 break;
 536
 537         case ICMP6_ECHO_REQUEST:
 538                 icmp6_ifstat_inc(ifp, ifs6_in_echo);
 539                 if (code != 0)
 540                         goto badcode;
 541                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
 542                         /* Give up remote */
 543                         break;
 544                 }
 545                 if (!M_WRITABLE(n)
 546                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
 547                         struct mbuf *n0 = n;
 548                         int n0len;
 549
 550                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
 551                         n = m_gethdr(M_NOWAIT, n0->m_type);
 552                         if (n == NULL) {
 553                                 /* Give up remote */
 554                                 m_freem(n0);
 555                                 break;
 556                         }
 557
 558                         m_move_pkthdr(n, n0);   /* FIB copied. */
 559                         n0len = n0->m_pkthdr.len;       /* save for use below */
 560                         /*
 561                          * Copy IPv6 and ICMPv6 only.
 562                          */
 563                         nip6 = mtod(n, struct ip6_hdr *);
 564                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 565                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 566                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 567                         noff = sizeof(struct ip6_hdr);
 568                         /* new mbuf contains only ipv6+icmpv6 headers */
 569                         n->m_len = noff + sizeof(struct icmp6_hdr);
 570                         /*
 571                          * Adjust mbuf.  ip6_plen will be adjusted in
 572                          * ip6_output().
 573                          */
 574                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 575                         /* recalculate complete packet size */
 576                         n->m_pkthdr.len = n0len + (noff - off);
 577                         n->m_next = n0;
 578                 } else {
 579                         if (n->m_len < off + sizeof(*nicmp6)) {
 580                                 n = m_pullup(n, off + sizeof(*nicmp6));
 581                                 if (n == NULL) {
 582                                         IP6STAT_INC(ip6s_exthdrtoolong);
 583                                         break;
 584                                 }
 585                         }
 586                         nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
 587                         noff = off;
 588                 }
 589                 if (n) {
 590                         nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 591                         nicmp6->icmp6_code = 0;
 592                         ICMP6STAT_INC(icp6s_reflect);
 593                         ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
 594                         icmp6_reflect(n, noff);
 595                 }
 596                 break;
 597
 598         case ICMP6_ECHO_REPLY:
 599                 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
 600                 if (code != 0)
 601                         goto badcode;
 602                 break;
 603
 604         case MLD_LISTENER_QUERY:
 605         case MLD_LISTENER_REPORT:
 606         case MLD_LISTENER_DONE:
 607         case MLDV2_LISTENER_REPORT:
 608                 /*
 609                  * Drop MLD traffic which is not link-local, has a hop limit
 610                  * of greater than 1 hop, or which does not have the
 611                  * IPv6 HBH Router Alert option.
 612                  * As IPv6 HBH options are stripped in ip6_input() we must
 613                  * check an mbuf header flag.
 614                  * XXX Should we also sanity check that these messages
 615                  * were directed to a link-local multicast prefix?
 616                  */
 617                 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
 618                         goto freeit;
 619                 if (mld_input(&m, off, icmp6len) != 0) {
 620                         *mp = NULL;
 621                         return (IPPROTO_DONE);
 622                 }
 623                 /* m stays. */
 624                 break;
 625
 626         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 627             {
 628                 enum { WRU, FQDN } mode;
 629                 struct prison *pr;
 630
 631                 if (!V_icmp6_nodeinfo)
 632                         break;
 633
 634                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 635                         mode = WRU;
 636                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 637                         mode = FQDN;
 638                 else
 639                         goto badlen;
 640
 641                 pr = NULL;
 642                 sx_slock(&allprison_lock);
 643                 TAILQ_FOREACH(pr, &allprison, pr_list)
 644                         if (pr->pr_vnet == ifp->if_vnet)
 645                                 break;
 646                 sx_sunlock(&allprison_lock);
 647                 if (pr == NULL)
 648                         pr = curthread->td_ucred->cr_prison;
 649                 if (mode == FQDN) {
 650                         if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
 651                                 m = m_pullup(m, off +
 652                                     sizeof(struct icmp6_nodeinfo));
 653                                 if (m == NULL) {
 654                                         IP6STAT_INC(ip6s_exthdrtoolong);
 655                                         *mp = m;
 656                                         return (IPPROTO_DONE);
 657                                 }
 658                         }
 659                         n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 660                         if (n)
 661                                 n = ni6_input(n, off, pr);
 662                         /* XXX meaningless if n == NULL */
 663                         noff = sizeof(struct ip6_hdr);
 664                 } else {
 665                         u_char *p;
 666                         int maxhlen, hlen;
 667
 668                         /*
 669                          * XXX: this combination of flags is pointless,
 670                          * but should we keep this for compatibility?
 671                          */
 672                         if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
 673                             ICMP6_NODEINFO_TMPADDROK)) !=
 674                             (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
 675                                 break;
 676
 677                         if (code != 0)
 678                                 goto badcode;
 679
 680                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
 681                         n = m_gethdr(M_NOWAIT, m->m_type);
 682                         if (n == NULL) {
 683                                 /* Give up remote */
 684                                 break;
 685                         }
 686                         if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
 687                                 /*
 688                                  * Previous code did a blind M_COPY_PKTHDR
 689                                  * and said "just for rcvif".  If true, then
 690                                  * we could tolerate the dup failing (due to
 691                                  * the deep copy of the tag chain).  For now
 692                                  * be conservative and just fail.
 693                                  */
 694                                 m_free(n);
 695                                 n = NULL;
 696                                 break;
 697                         }
 698                         /*
 699                          * Copy IPv6 and ICMPv6 only.
 700                          */
 701                         nip6 = mtod(n, struct ip6_hdr *);
 702                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 703                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 704                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 705                         p = (u_char *)(nicmp6 + 1);
 706                         bzero(p, 4);
 707
 708                         maxhlen = M_TRAILINGSPACE(n) -
 709                             (sizeof(*nip6) + sizeof(*nicmp6) + 4);
 710                         mtx_lock(&pr->pr_mtx);
 711                         hlen = strlen(pr->pr_hostname);
 712                         if (maxhlen > hlen)
 713                                 maxhlen = hlen;
 714                         /* meaningless TTL */
 715                         bcopy(pr->pr_hostname, p + 4, maxhlen);
 716                         mtx_unlock(&pr->pr_mtx);
 717                         noff = sizeof(struct ip6_hdr);
 718                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 719                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 720                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 721                         nicmp6->icmp6_code = 0;
 722                 }
 723                 if (n) {
 724                         ICMP6STAT_INC(icp6s_reflect);
 725                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
 726                         icmp6_reflect(n, noff);
 727                 }
 728                 break;
 729             }
 730
 731         case ICMP6_WRUREPLY:
 732                 if (code != 0)
 733                         goto badcode;
 734                 break;
 735
 736         case ND_ROUTER_SOLICIT:
 737                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
 738                 if (code != 0)
 739                         goto badcode;
 740                 if (icmp6len < sizeof(struct nd_router_solicit))
 741                         goto badlen;
 742                 if (send_sendso_input_hook != NULL) {
 743                         if (m->m_len < off + icmp6len) {
 744                                 m = m_pullup(m, off + icmp6len);
 745                                 if (m == NULL) {
 746                                         IP6STAT_INC(ip6s_exthdrtoolong);
 747                                         *mp = NULL;
 748                                         return (IPPROTO_DONE);
 749                                 }
 750                         }
 751                         error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
 752                         if (error == 0) {
 753                                 m = NULL;
 754                                 goto freeit;
 755                         }
 756                 }
 757                 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 758                 nd6_rs_input(m, off, icmp6len);
 759                 m = n;
 760                 if (m == NULL)
 761                         goto freeit;
 762                 break;
 763
 764         case ND_ROUTER_ADVERT:
 765                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
 766                 if (code != 0)
 767                         goto badcode;
 768                 if (icmp6len < sizeof(struct nd_router_advert))
 769                         goto badlen;
 770                 if (send_sendso_input_hook != NULL) {
 771                         error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
 772                         if (error == 0) {
 773                                 m = NULL;
 774                                 goto freeit;
 775                         }
 776                 }
 777                 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 778                 nd6_ra_input(m, off, icmp6len);
 779                 m = n;
 780                 if (m == NULL)
 781                         goto freeit;
 782                 break;
 783
 784         case ND_NEIGHBOR_SOLICIT:
 785                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
 786                 if (code != 0)
 787                         goto badcode;
 788                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 789                         goto badlen;
 790                 if (send_sendso_input_hook != NULL) {
 791                         error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
 792                         if (error == 0) {
 793                                 m = NULL;
 794                                 goto freeit;
 795                         }
 796                 }
 797                 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 798                 nd6_ns_input(m, off, icmp6len);
 799                 m = n;
 800                 if (m == NULL)
 801                         goto freeit;
 802                 break;
 803
 804         case ND_NEIGHBOR_ADVERT:
 805                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
 806                 if (code != 0)
 807                         goto badcode;
 808                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 809                         goto badlen;
 810                 if (send_sendso_input_hook != NULL) {
 811                         error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
 812                         if (error == 0) {
 813                                 m = NULL;
 814                                 goto freeit;
 815                         }
 816                 }
 817                 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 818                 nd6_na_input(m, off, icmp6len);
 819                 m = n;
 820                 if (m == NULL)
 821                         goto freeit;
 822                 break;
 823
 824         case ND_REDIRECT:
 825                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
 826                 if (code != 0)
 827                         goto badcode;
 828                 if (icmp6len < sizeof(struct nd_redirect))
 829                         goto badlen;
 830                 if (send_sendso_input_hook != NULL) {
 831                         error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
 832                         if (error == 0) {
 833                                 m = NULL;
 834                                 goto freeit;
 835                         }
 836                 }
 837                 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 838                 icmp6_redirect_input(m, off);
 839                 m = n;
 840                 if (m == NULL)
 841                         goto freeit;
 842                 break;
 843
 844         case ICMP6_ROUTER_RENUMBERING:
 845                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 846                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 847                         goto badcode;
 848                 if (icmp6len < sizeof(struct icmp6_router_renum))
 849                         goto badlen;
 850                 break;
 851
 852         default:
 853                 nd6log((LOG_DEBUG,
 854                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 855                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
 856                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
 857                     ifp ? ifp->if_index : 0));
 858                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 859                         /* ICMPv6 error: MUST deliver it by spec... */
 860                         code = PRC_NCMDS;
 861                         /* deliver */
 862                 } else {
 863                         /* ICMPv6 informational: MUST not deliver */
 864                         break;
 865                 }
 866         deliver:
 867                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
 868                         /* In this case, m should've been freed. */
 869                         *mp = NULL;
 870                         return (IPPROTO_DONE);
 871                 }
 872                 break;
 873
 874         badcode:
 875                 ICMP6STAT_INC(icp6s_badcode);
 876                 break;
 877
 878         badlen:
 879                 ICMP6STAT_INC(icp6s_badlen);
 880                 break;
 881         }
 882
 883         /* deliver the packet to appropriate sockets */
 884         icmp6_rip6_input(&m, *offp);
 885
 886         *mp = m;
 887         return (IPPROTO_DONE);
 888
 889  freeit:
 890         m_freem(m);
 891         *mp = NULL;
 892         return (IPPROTO_DONE);
 893 }
 894
 895 static int
 896 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
 897 {
 898         struct mbuf *m;
 899         struct icmp6_hdr *icmp6;
 900         struct ip6_hdr *eip6;
 901         u_int32_t notifymtu;
 902         struct sockaddr_in6 icmp6src, icmp6dst;
 903
 904         m = *mp;
 905
 906         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 907                 ICMP6STAT_INC(icp6s_tooshort);
 908                 goto freeit;
 909         }
 910
 911         if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
 912                 m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
 913                 if (m == NULL) {
 914                         IP6STAT_INC(ip6s_exthdrtoolong);
 915                         *mp = m;
 916                         return (-1);
 917                 }
 918         }
 919         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 920         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 921         bzero(&icmp6dst, sizeof(icmp6dst));
 922
 923         /* Detect the upper level protocol */
 924         {
 925                 void (*ctlfunc)(int, struct sockaddr *, void *);
 926                 u_int8_t nxt = eip6->ip6_nxt;
 927                 int eoff = off + sizeof(struct icmp6_hdr) +
 928                     sizeof(struct ip6_hdr);
 929                 struct ip6ctlparam ip6cp;
 930                 int icmp6type = icmp6->icmp6_type;
 931                 struct ip6_frag *fh;
 932                 struct ip6_rthdr *rth;
 933                 struct ip6_rthdr0 *rth0;
 934                 int rthlen;
 935
 936                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 937                         struct ip6_ext *eh;
 938
 939                         switch (nxt) {
 940                         case IPPROTO_HOPOPTS:
 941                         case IPPROTO_DSTOPTS:
 942                         case IPPROTO_AH:
 943                                 if (m->m_len < eoff + sizeof(struct ip6_ext)) {
 944                                         m = m_pullup(m, eoff +
 945                                             sizeof(struct ip6_ext));
 946                                         if (m == NULL) {
 947                                                 IP6STAT_INC(ip6s_exthdrtoolong);
 948                                                 *mp = m;
 949                                                 return (-1);
 950                                         }
 951                                 }
 952                                 eh = (struct ip6_ext *)
 953                                     (mtod(m, caddr_t) + eoff);
 954                                 if (nxt == IPPROTO_AH)
 955                                         eoff += (eh->ip6e_len + 2) << 2;
 956                                 else
 957                                         eoff += (eh->ip6e_len + 1) << 3;
 958                                 nxt = eh->ip6e_nxt;
 959                                 break;
 960                         case IPPROTO_ROUTING:
 961                                 /*
 962                                  * When the erroneous packet contains a
 963                                  * routing header, we should examine the
 964                                  * header to determine the final destination.
 965                                  * Otherwise, we can't properly update
 966                                  * information that depends on the final
 967                                  * destination (e.g. path MTU).
 968                                  */
 969                                 if (m->m_len < eoff + sizeof(*rth)) {
 970                                         m = m_pullup(m, eoff + sizeof(*rth));
 971                                         if (m == NULL) {
 972                                                 IP6STAT_INC(ip6s_exthdrtoolong);
 973                                                 *mp = m;
 974                                                 return (-1);
 975                                         }
 976                                 }
 977                                 rth = (struct ip6_rthdr *)
 978                                     (mtod(m, caddr_t) + eoff);
 979                                 rthlen = (rth->ip6r_len + 1) << 3;
 980                                 /*
 981                                  * XXX: currently there is no
 982                                  * officially defined type other
 983                                  * than type-0.
 984                                  * Note that if the segment left field
 985                                  * is 0, all intermediate hops must
 986                                  * have been passed.
 987                                  */
 988                                 if (rth->ip6r_segleft &&
 989                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
 990                                         int hops;
 991
 992                                         if (m->m_len < eoff + rthlen) {
 993                                                 m = m_pullup(m, eoff + rthlen);
 994                                                 if (m == NULL) {
 995                                                         IP6STAT_INC(
 996                                                             ip6s_exthdrtoolong);
 997                                                         *mp = m;
 998                                                         return (-1);
 999                                                 }
1000                                         }
1001                                         rth0 = (struct ip6_rthdr0 *)
1002                                             (mtod(m, caddr_t) + eoff);
1003
1004                                         /* just ignore a bogus header */
1005                                         if ((rth0->ip6r0_len % 2) == 0 &&
1006                                             (hops = rth0->ip6r0_len/2))
1007                                                 icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1008                                 }
1009                                 eoff += rthlen;
1010                                 nxt = rth->ip6r_nxt;
1011                                 break;
1012                         case IPPROTO_FRAGMENT:
1013                                 if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1014                                         m = m_pullup(m, eoff +
1015                                             sizeof(struct ip6_frag));
1016                                         if (m == NULL) {
1017                                                 IP6STAT_INC(ip6s_exthdrtoolong);
1018                                                 *mp = m;
1019                                                 return (-1);
1020                                         }
1021                                 }
1022                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1023                                     eoff);
1024                                 /*
1025                                  * Data after a fragment header is meaningless
1026                                  * unless it is the first fragment, but
1027                                  * we'll go to the notify label for path MTU
1028                                  * discovery.
1029                                  */
1030                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1031                                         goto notify;
1032
1033                                 eoff += sizeof(struct ip6_frag);
1034                                 nxt = fh->ip6f_nxt;
1035                                 break;
1036                         default:
1037                                 /*
1038                                  * This case includes ESP and the No Next
1039                                  * Header.  In such cases going to the notify
1040                                  * label does not have any meaning
1041                                  * (i.e. ctlfunc will be NULL), but we go
1042                                  * anyway since we might have to update
1043                                  * path MTU information.
1044                                  */
1045                                 goto notify;
1046                         }
1047                 }
1048           notify:
1049                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1050
1051                 /*
1052                  * retrieve parameters from the inner IPv6 header, and convert
1053                  * them into sockaddr structures.
1054                  * XXX: there is no guarantee that the source or destination
1055                  * addresses of the inner packet are in the same scope as
1056                  * the addresses of the icmp packet.  But there is no other
1057                  * way to determine the zone.
1058                  */
1059                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1060
1061                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1062                 icmp6dst.sin6_family = AF_INET6;
1063                 if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1064                         icmp6dst.sin6_addr = eip6->ip6_dst;
1065                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1066                         goto freeit;
1067                 bzero(&icmp6src, sizeof(icmp6src));
1068                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1069                 icmp6src.sin6_family = AF_INET6;
1070                 icmp6src.sin6_addr = eip6->ip6_src;
1071                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1072                         goto freeit;
1073                 icmp6src.sin6_flowinfo =
1074                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1075
1076                 ip6cp.ip6c_m = m;
1077                 ip6cp.ip6c_icmp6 = icmp6;
1078                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1079                 ip6cp.ip6c_off = eoff;
1080                 ip6cp.ip6c_finaldst = &icmp6dst.sin6_addr;
1081                 ip6cp.ip6c_src = &icmp6src;
1082                 ip6cp.ip6c_nxt = nxt;
1083
1084                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1085                         notifymtu = ntohl(icmp6->icmp6_mtu);
1086                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1087                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1088                 }
1089
1090                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1091                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1092                 if (ctlfunc) {
1093                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1094                             &ip6cp);
1095                 }
1096         }
1097         *mp = m;
1098         return (0);
1099
1100   freeit:
1101         m_freem(m);
1102         *mp = NULL;
1103         return (-1);
1104 }
1105
1106 void
1107 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1108 {
1109         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1110         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1111         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1112         u_int mtu = ntohl(icmp6->icmp6_mtu);
1113         struct in_conninfo inc;
1114
1115 #if 0
1116         /*
1117          * RFC2460 section 5, last paragraph.
1118          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1119          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1120          * due to packet translator in the middle.
1121          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1122          * special handling.
1123          */
1124         if (mtu < IPV6_MMTU)
1125                 return;
1126 #endif
1127
1128         /*
1129          * we reject ICMPv6 too big with abnormally small value.
1130          * XXX what is the good definition of "abnormally small"?
1131          */
1132         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1133                 return;
1134
1135         if (!validated)
1136                 return;
1137
1138         /*
1139          * In case the suggested mtu is less than IPV6_MMTU, we
1140          * only need to remember that it was for above mentioned
1141          * "alwaysfrag" case.
1142          * Try to be as close to the spec as possible.
1143          */
1144         if (mtu < IPV6_MMTU)
1145                 mtu = IPV6_MMTU - 8;
1146
1147         bzero(&inc, sizeof(inc));
1148         inc.inc_fibnum = M_GETFIB(m);
1149         inc.inc_flags |= INC_ISIPV6;
1150         inc.inc6_faddr = *dst;
1151         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1152                 return;
1153
1154         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1155                 tcp_hc_updatemtu(&inc, mtu);
1156                 ICMP6STAT_INC(icp6s_pmtuchg);
1157         }
1158 }
1159
1160 /*
1161  * Process a Node Information Query packet, based on
1162  * draft-ietf-ipngwg-icmp-name-lookups-07.
1163  *
1164  * Spec incompatibilities:
1165  * - IPv6 Subject address handling
1166  * - IPv4 Subject address handling support missing
1167  * - Proxy reply (answer even if it's not for me)
1168  * - joins NI group address at in6_ifattach() time only, does not cope
1169  *   with hostname changes by sethostname(3)
1170  */
1171 static struct mbuf *
1172 ni6_input(struct mbuf *m, int off, struct prison *pr)
1173 {
1174         struct icmp6_nodeinfo *ni6, *nni6;
1175         struct mbuf *n = NULL;
1176         u_int16_t qtype;
1177         int subjlen;
1178         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1179         struct ni_reply_fqdn *fqdn;
1180         int addrs;              /* for NI_QTYPE_NODEADDR */
1181         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1182         struct in6_addr in6_subj; /* subject address */
1183         struct ip6_hdr *ip6;
1184         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1185         char *subj = NULL;
1186         struct in6_ifaddr *ia6 = NULL;
1187
1188         ip6 = mtod(m, struct ip6_hdr *);
1189         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1190
1191         /*
1192          * Validate IPv6 source address.
1193          * The default configuration MUST be to refuse answering queries from
1194          * global-scope addresses according to RFC4602.
1195          * Notes:
1196          *  - it's not very clear what "refuse" means; this implementation
1197          *    simply drops it.
1198          *  - it's not very easy to identify global-scope (unicast) addresses
1199          *    since there are many prefixes for them.  It should be safer
1200          *    and in practice sufficient to check "all" but loopback and
1201          *    link-local (note that site-local unicast was deprecated and
1202          *    ULA is defined as global scope-wise)
1203          */
1204         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1205             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1206             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1207                 goto bad;
1208
1209         /*
1210          * Validate IPv6 destination address.
1211          *
1212          * The Responder must discard the Query without further processing
1213          * unless it is one of the Responder's unicast or anycast addresses, or
1214          * a link-local scope multicast address which the Responder has joined.
1215          * [RFC4602, Section 5.]
1216          */
1217         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1218                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1219                         goto bad;
1220                 /* else it's a link-local multicast, fine */
1221         } else {                /* unicast or anycast */
1222                 ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
1223                 if (ia6 == NULL)
1224                         goto bad; /* XXX impossible */
1225
1226                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1227                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1228                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1229                                 "a temporary address in %s:%d",
1230                                __FILE__, __LINE__));
1231                         goto bad;
1232                 }
1233         }
1234
1235         /* validate query Subject field. */
1236         qtype = ntohs(ni6->ni_qtype);
1237         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1238         switch (qtype) {
1239         case NI_QTYPE_NOOP:
1240         case NI_QTYPE_SUPTYPES:
1241                 /* 07 draft */
1242                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1243                         break;
1244                 /* FALLTHROUGH */
1245         case NI_QTYPE_FQDN:
1246         case NI_QTYPE_NODEADDR:
1247         case NI_QTYPE_IPV4ADDR:
1248                 switch (ni6->ni_code) {
1249                 case ICMP6_NI_SUBJ_IPV6:
1250 #if ICMP6_NI_SUBJ_IPV6 != 0
1251                 case 0:
1252 #endif
1253                         /*
1254                          * backward compatibility - try to accept 03 draft
1255                          * format, where no Subject is present.
1256                          */
1257                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1258                             subjlen == 0) {
1259                                 oldfqdn++;
1260                                 break;
1261                         }
1262 #if ICMP6_NI_SUBJ_IPV6 != 0
1263                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1264                                 goto bad;
1265 #endif
1266
1267                         if (subjlen != sizeof(struct in6_addr))
1268                                 goto bad;
1269
1270                         /*
1271                          * Validate Subject address.
1272                          *
1273                          * Not sure what exactly "address belongs to the node"
1274                          * means in the spec, is it just unicast, or what?
1275                          *
1276                          * At this moment we consider Subject address as
1277                          * "belong to the node" if the Subject address equals
1278                          * to the IPv6 destination address; validation for
1279                          * IPv6 destination address should have done enough
1280                          * check for us.
1281                          *
1282                          * We do not do proxy at this moment.
1283                          */
1284                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1285                             subjlen, (caddr_t)&in6_subj);
1286                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1287                                 goto bad;
1288
1289                         subj = (char *)&in6_subj;
1290                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1291                                 break;
1292
1293                         /*
1294                          * XXX if we are to allow other cases, we should really
1295                          * be careful about scope here.
1296                          * basically, we should disallow queries toward IPv6
1297                          * destination X with subject Y,
1298                          * if scope(X) > scope(Y).
1299                          * if we allow scope(X) > scope(Y), it will result in
1300                          * information leakage across scope boundary.
1301                          */
1302                         goto bad;
1303
1304                 case ICMP6_NI_SUBJ_FQDN:
1305                         /*
1306                          * Validate Subject name with gethostname(3).
1307                          *
1308                          * The behavior may need some debate, since:
1309                          * - we are not sure if the node has FQDN as
1310                          *   hostname (returned by gethostname(3)).
1311                          * - the code does wildcard match for truncated names.
1312                          *   however, we are not sure if we want to perform
1313                          *   wildcard match, if gethostname(3) side has
1314                          *   truncated hostname.
1315                          */
1316                         mtx_lock(&pr->pr_mtx);
1317                         n = ni6_nametodns(pr->pr_hostname,
1318                             strlen(pr->pr_hostname), 0);
1319                         mtx_unlock(&pr->pr_mtx);
1320                         if (!n || n->m_next || n->m_len == 0)
1321                                 goto bad;
1322                         if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1323                             subjlen) {
1324                                 m = m_pullup(m, off +
1325                                     sizeof(struct icmp6_nodeinfo) + subjlen);
1326                                 if (m == NULL) {
1327                                         IP6STAT_INC(ip6s_exthdrtoolong);
1328                                         goto bad;
1329                                 }
1330                         }
1331                         /* ip6 possibly invalid but not used after. */
1332                         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1333                         subj = (char *)(mtod(m, caddr_t) + off +
1334                             sizeof(struct icmp6_nodeinfo));
1335                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1336                             n->m_len)) {
1337                                 goto bad;
1338                         }
1339                         m_freem(n);
1340                         n = NULL;
1341                         break;
1342
1343                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1344                 default:
1345                         goto bad;
1346                 }
1347                 break;
1348         }
1349
1350         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1351         switch (qtype) {
1352         case NI_QTYPE_FQDN:
1353                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1354                         goto bad;
1355                 break;
1356         case NI_QTYPE_NODEADDR:
1357         case NI_QTYPE_IPV4ADDR:
1358                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1359                         goto bad;
1360                 break;
1361         }
1362
1363         /* guess reply length */
1364         switch (qtype) {
1365         case NI_QTYPE_NOOP:
1366                 break;          /* no reply data */
1367         case NI_QTYPE_SUPTYPES:
1368                 replylen += sizeof(u_int32_t);
1369                 break;
1370         case NI_QTYPE_FQDN:
1371                 /* XXX will append an mbuf */
1372                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1373                 break;
1374         case NI_QTYPE_NODEADDR:
1375                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1376                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1377                     sizeof(u_int32_t))) > MCLBYTES)
1378                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1379                 break;
1380         case NI_QTYPE_IPV4ADDR:
1381                 /* unsupported - should respond with unknown Qtype? */
1382                 break;
1383         default:
1384                 /*
1385                  * XXX: We must return a reply with the ICMP6 code
1386                  * `unknown Qtype' in this case.  However we regard the case
1387                  * as an FQDN query for backward compatibility.
1388                  * Older versions set a random value to this field,
1389                  * so it rarely varies in the defined qtypes.
1390                  * But the mechanism is not reliable...
1391                  * maybe we should obsolete older versions.
1392                  */
1393                 qtype = NI_QTYPE_FQDN;
1394                 /* XXX will append an mbuf */
1395                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1396                 oldfqdn++;
1397                 break;
1398         }
1399
1400         /* Allocate an mbuf to reply. */
1401         if (replylen > MCLBYTES) {
1402                 /*
1403                  * XXX: should we try to allocate more? But MCLBYTES
1404                  * is probably much larger than IPV6_MMTU...
1405                  */
1406                 goto bad;
1407         }
1408         if (replylen > MHLEN)
1409                 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1410         else
1411                 n = m_gethdr(M_NOWAIT, m->m_type);
1412         if (n == NULL) {
1413                 m_freem(m);
1414                 return (NULL);
1415         }
1416         m_move_pkthdr(n, m); /* just for recvif and FIB */
1417         n->m_pkthdr.len = n->m_len = replylen;
1418
1419         /* copy mbuf header and IPv6 + Node Information base headers */
1420         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1421         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1422         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1423
1424         /* qtype dependent procedure */
1425         switch (qtype) {
1426         case NI_QTYPE_NOOP:
1427                 nni6->ni_code = ICMP6_NI_SUCCESS;
1428                 nni6->ni_flags = 0;
1429                 break;
1430         case NI_QTYPE_SUPTYPES:
1431         {
1432                 u_int32_t v;
1433                 nni6->ni_code = ICMP6_NI_SUCCESS;
1434                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1435                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1436                 v = (u_int32_t)htonl(0x0000000f);
1437                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1438                 break;
1439         }
1440         case NI_QTYPE_FQDN:
1441                 nni6->ni_code = ICMP6_NI_SUCCESS;
1442                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1443                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1444                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1445                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1446                 /*
1447                  * XXX do we really have FQDN in hostname?
1448                  */
1449                 mtx_lock(&pr->pr_mtx);
1450                 n->m_next = ni6_nametodns(pr->pr_hostname,
1451                     strlen(pr->pr_hostname), oldfqdn);
1452                 mtx_unlock(&pr->pr_mtx);
1453                 if (n->m_next == NULL)
1454                         goto bad;
1455                 /* XXX we assume that n->m_next is not a chain */
1456                 if (n->m_next->m_next != NULL)
1457                         goto bad;
1458                 n->m_pkthdr.len += n->m_next->m_len;
1459                 break;
1460         case NI_QTYPE_NODEADDR:
1461         {
1462                 int lenlim, copied;
1463
1464                 nni6->ni_code = ICMP6_NI_SUCCESS;
1465                 n->m_pkthdr.len = n->m_len =
1466                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1467                 lenlim = M_TRAILINGSPACE(n);
1468                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1469                 /* XXX: reset mbuf length */
1470                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1471                     sizeof(struct icmp6_nodeinfo) + copied;
1472                 break;
1473         }
1474         default:
1475                 break;          /* XXX impossible! */
1476         }
1477
1478         nni6->ni_type = ICMP6_NI_REPLY;
1479         m_freem(m);
1480         return (n);
1481
1482   bad:
1483         m_freem(m);
1484         if (n)
1485                 m_freem(n);
1486         return (NULL);
1487 }
1488
1489 /*
1490  * make a mbuf with DNS-encoded string.  no compression support.
1491  *
1492  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1493  * treated as truncated name (two \0 at the end).  this is a wild guess.
1494  *
1495  * old - return pascal string if non-zero
1496  */
1497 static struct mbuf *
1498 ni6_nametodns(const char *name, int namelen, int old)
1499 {
1500         struct mbuf *m;
1501         char *cp, *ep;
1502         const char *p, *q;
1503         int i, len, nterm;
1504
1505         if (old)
1506                 len = namelen + 1;
1507         else
1508                 len = MCLBYTES;
1509
1510         /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1511         if (len > MLEN)
1512                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1513         else
1514                 m = m_get(M_NOWAIT, MT_DATA);
1515         if (m == NULL)
1516                 goto fail;
1517
1518         if (old) {
1519                 m->m_len = len;
1520                 *mtod(m, char *) = namelen;
1521                 bcopy(name, mtod(m, char *) + 1, namelen);
1522                 return m;
1523         } else {
1524                 m->m_len = 0;
1525                 cp = mtod(m, char *);
1526                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1527
1528                 /* if not certain about my name, return empty buffer */
1529                 if (namelen == 0)
1530                         return m;
1531
1532                 /*
1533                  * guess if it looks like shortened hostname, or FQDN.
1534                  * shortened hostname needs two trailing "\0".
1535                  */
1536                 i = 0;
1537                 for (p = name; p < name + namelen; p++) {
1538                         if (*p && *p == '.')
1539                                 i++;
1540                 }
1541                 if (i < 2)
1542                         nterm = 2;
1543                 else
1544                         nterm = 1;
1545
1546                 p = name;
1547                 while (cp < ep && p < name + namelen) {
1548                         i = 0;
1549                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1550                                 i++;
1551                         /* result does not fit into mbuf */
1552                         if (cp + i + 1 >= ep)
1553                                 goto fail;
1554                         /*
1555                          * DNS label length restriction, RFC1035 page 8.
1556                          * "i == 0" case is included here to avoid returning
1557                          * 0-length label on "foo..bar".
1558                          */
1559                         if (i <= 0 || i >= 64)
1560                                 goto fail;
1561                         *cp++ = i;
1562                         bcopy(p, cp, i);
1563                         cp += i;
1564                         p = q;
1565                         if (p < name + namelen && *p == '.')
1566                                 p++;
1567                 }
1568                 /* termination */
1569                 if (cp + nterm >= ep)
1570                         goto fail;
1571                 while (nterm-- > 0)
1572                         *cp++ = '\0';
1573                 m->m_len = cp - mtod(m, char *);
1574                 return m;
1575         }
1576
1577         panic("should not reach here");
1578         /* NOTREACHED */
1579
1580  fail:
1581         if (m)
1582                 m_freem(m);
1583         return NULL;
1584 }
1585
1586 /*
1587  * check if two DNS-encoded string matches.  takes care of truncated
1588  * form (with \0\0 at the end).  no compression support.
1589  * XXX upper/lowercase match (see RFC2065)
1590  */
1591 static int
1592 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1593 {
1594         const char *a0, *b0;
1595         int l;
1596
1597         /* simplest case - need validation? */
1598         if (alen == blen && bcmp(a, b, alen) == 0)
1599                 return 1;
1600
1601         a0 = a;
1602         b0 = b;
1603
1604         /* termination is mandatory */
1605         if (alen < 2 || blen < 2)
1606                 return 0;
1607         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1608                 return 0;
1609         alen--;
1610         blen--;
1611
1612         while (a - a0 < alen && b - b0 < blen) {
1613                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1614                         return 0;
1615
1616                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1617                         return 0;
1618                 /* we don't support compression yet */
1619                 if (a[0] >= 64 || b[0] >= 64)
1620                         return 0;
1621
1622                 /* truncated case */
1623                 if (a[0] == 0 && a - a0 == alen - 1)
1624                         return 1;
1625                 if (b[0] == 0 && b - b0 == blen - 1)
1626                         return 1;
1627                 if (a[0] == 0 || b[0] == 0)
1628                         return 0;
1629
1630                 if (a[0] != b[0])
1631                         return 0;
1632                 l = a[0];
1633                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1634                         return 0;
1635                 if (bcmp(a + 1, b + 1, l) != 0)
1636                         return 0;
1637
1638                 a += 1 + l;
1639                 b += 1 + l;
1640         }
1641
1642         if (a - a0 == alen && b - b0 == blen)
1643                 return 1;
1644         else
1645                 return 0;
1646 }
1647
1648 /*
1649  * calculate the number of addresses to be returned in the node info reply.
1650  */
1651 static int
1652 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1653     struct in6_addr *subj)
1654 {
1655         struct ifnet *ifp;
1656         struct in6_ifaddr *ifa6;
1657         struct ifaddr *ifa;
1658         int addrs = 0, addrsofif, iffound = 0;
1659         int niflags = ni6->ni_flags;
1660
1661         NET_EPOCH_ASSERT();
1662
1663         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1664                 switch (ni6->ni_code) {
1665                 case ICMP6_NI_SUBJ_IPV6:
1666                         if (subj == NULL) /* must be impossible... */
1667                                 return (0);
1668                         break;
1669                 default:
1670                         /*
1671                          * XXX: we only support IPv6 subject address for
1672                          * this Qtype.
1673                          */
1674                         return (0);
1675                 }
1676         }
1677
1678         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1679                 addrsofif = 0;
1680                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1681                         if (ifa->ifa_addr->sa_family != AF_INET6)
1682                                 continue;
1683                         ifa6 = (struct in6_ifaddr *)ifa;
1684
1685                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1686                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1687                                 iffound = 1;
1688
1689                         /*
1690                          * IPv4-mapped addresses can only be returned by a
1691                          * Node Information proxy, since they represent
1692                          * addresses of IPv4-only nodes, which perforce do
1693                          * not implement this protocol.
1694                          * [icmp-name-lookups-07, Section 5.4]
1695                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1696                          * this function at this moment.
1697                          */
1698
1699                         /* What do we have to do about ::1? */
1700                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1701                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1702                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1703                                         continue;
1704                                 break;
1705                         case IPV6_ADDR_SCOPE_SITELOCAL:
1706                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1707                                         continue;
1708                                 break;
1709                         case IPV6_ADDR_SCOPE_GLOBAL:
1710                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1711                                         continue;
1712                                 break;
1713                         default:
1714                                 continue;
1715                         }
1716
1717                         /*
1718                          * check if anycast is okay.
1719                          * XXX: just experimental.  not in the spec.
1720                          */
1721                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1722                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1723                                 continue; /* we need only unicast addresses */
1724                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1725                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1726                                 continue;
1727                         }
1728                         addrsofif++; /* count the address */
1729                 }
1730                 if (iffound) {
1731                         *ifpp = ifp;
1732                         return (addrsofif);
1733                 }
1734
1735                 addrs += addrsofif;
1736         }
1737
1738         return (addrs);
1739 }
1740
1741 static int
1742 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1743     struct ifnet *ifp0, int resid)
1744 {
1745         struct ifnet *ifp;
1746         struct in6_ifaddr *ifa6;
1747         struct ifaddr *ifa;
1748         struct ifnet *ifp_dep = NULL;
1749         int copied = 0, allow_deprecated = 0;
1750         u_char *cp = (u_char *)(nni6 + 1);
1751         int niflags = ni6->ni_flags;
1752         u_int32_t ltime;
1753
1754         NET_EPOCH_ASSERT();
1755
1756         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1757                 return (0);     /* needless to copy */
1758
1759         ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1760   again:
1761
1762         for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1763                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1764                         if (ifa->ifa_addr->sa_family != AF_INET6)
1765                                 continue;
1766                         ifa6 = (struct in6_ifaddr *)ifa;
1767
1768                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1769                             allow_deprecated == 0) {
1770                                 /*
1771                                  * prefererred address should be put before
1772                                  * deprecated addresses.
1773                                  */
1774
1775                                 /* record the interface for later search */
1776                                 if (ifp_dep == NULL)
1777                                         ifp_dep = ifp;
1778
1779                                 continue;
1780                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1781                             allow_deprecated != 0)
1782                                 continue; /* we now collect deprecated addrs */
1783
1784                         /* What do we have to do about ::1? */
1785                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1786                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1787                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1788                                         continue;
1789                                 break;
1790                         case IPV6_ADDR_SCOPE_SITELOCAL:
1791                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1792                                         continue;
1793                                 break;
1794                         case IPV6_ADDR_SCOPE_GLOBAL:
1795                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1796                                         continue;
1797                                 break;
1798                         default:
1799                                 continue;
1800                         }
1801
1802                         /*
1803                          * check if anycast is okay.
1804                          * XXX: just experimental.  not in the spec.
1805                          */
1806                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1807                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1808                                 continue;
1809                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1810                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1811                                 continue;
1812                         }
1813
1814                         /* now we can copy the address */
1815                         if (resid < sizeof(struct in6_addr) +
1816                             sizeof(u_int32_t)) {
1817                                 /*
1818                                  * We give up much more copy.
1819                                  * Set the truncate flag and return.
1820                                  */
1821                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1822                                 return (copied);
1823                         }
1824
1825                         /*
1826                          * Set the TTL of the address.
1827                          * The TTL value should be one of the following
1828                          * according to the specification:
1829                          *
1830                          * 1. The remaining lifetime of a DHCP lease on the
1831                          *    address, or
1832                          * 2. The remaining Valid Lifetime of a prefix from
1833                          *    which the address was derived through Stateless
1834                          *    Autoconfiguration.
1835                          *
1836                          * Note that we currently do not support stateful
1837                          * address configuration by DHCPv6, so the former
1838                          * case can't happen.
1839                          */
1840                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1841                                 ltime = ND6_INFINITE_LIFETIME;
1842                         else {
1843                                 if (ifa6->ia6_lifetime.ia6t_expire >
1844                                     time_uptime)
1845                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1846                                 else
1847                                         ltime = 0;
1848                         }
1849
1850                         bcopy(&ltime, cp, sizeof(u_int32_t));
1851                         cp += sizeof(u_int32_t);
1852
1853                         /* copy the address itself */
1854                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1855                             sizeof(struct in6_addr));
1856                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1857                         cp += sizeof(struct in6_addr);
1858
1859                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1860                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1861                 }
1862                 if (ifp0)       /* we need search only on the specified IF */
1863                         break;
1864         }
1865
1866         if (allow_deprecated == 0 && ifp_dep != NULL) {
1867                 ifp = ifp_dep;
1868                 allow_deprecated = 1;
1869
1870                 goto again;
1871         }
1872
1873         return (copied);
1874 }
1875
1876 static bool
1877 icmp6_rip6_match(const struct inpcb *inp, void *v)
1878 {
1879         struct ip6_hdr *ip6 = v;
1880
1881         if ((inp->inp_vflag & INP_IPV6) == 0)
1882                 return (false);
1883         if (inp->inp_ip_p != IPPROTO_ICMPV6)
1884                 return (false);
1885         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1886            !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1887                 return (false);
1888         if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1889            !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1890                 return (false);
1891         return (true);
1892 }
1893
1894 /*
1895  * XXX almost dup'ed code with rip6_input.
1896  */
1897 static int
1898 icmp6_rip6_input(struct mbuf **mp, int off)
1899 {
1900         struct mbuf *n, *m = *mp;
1901         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1902         struct inpcb_iterator inpi = INP_ITERATOR(&V_ripcbinfo,
1903             INPLOOKUP_RLOCKPCB, icmp6_rip6_match, ip6);
1904         struct inpcb *inp;
1905         struct sockaddr_in6 fromsa;
1906         struct icmp6_hdr *icmp6;
1907         struct mbuf *opts = NULL;
1908         int delivered = 0;
1909
1910         /* This is assumed to be safe; icmp6_input() does a pullup. */
1911         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1912
1913         /*
1914          * XXX: the address may have embedded scope zone ID, which should be
1915          * hidden from applications.
1916          */
1917         bzero(&fromsa, sizeof(fromsa));
1918         fromsa.sin6_family = AF_INET6;
1919         fromsa.sin6_len = sizeof(struct sockaddr_in6);
1920         fromsa.sin6_addr = ip6->ip6_src;
1921         if (sa6_recoverscope(&fromsa)) {
1922                 m_freem(m);
1923                 *mp = NULL;
1924                 return (IPPROTO_DONE);
1925         }
1926
1927         while ((inp = inp_next(&inpi)) != NULL) {
1928                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1929                     inp->in6p_icmp6filt))
1930                         continue;
1931                 /*
1932                  * Recent network drivers tend to allocate a single
1933                  * mbuf cluster, rather than to make a couple of
1934                  * mbufs without clusters.  Also, since the IPv6 code
1935                  * path tries to avoid m_pullup(), it is highly
1936                  * probable that we still have an mbuf cluster here
1937                  * even though the necessary length can be stored in an
1938                  * mbuf's internal buffer.
1939                  * Meanwhile, the default size of the receive socket
1940                  * buffer for raw sockets is not so large.  This means
1941                  * the possibility of packet loss is relatively higher
1942                  * than before.  To avoid this scenario, we copy the
1943                  * received data to a separate mbuf that does not use
1944                  * a cluster, if possible.
1945                  * XXX: it is better to copy the data after stripping
1946                  * intermediate headers.
1947                  */
1948                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1949                     m->m_len <= MHLEN) {
1950                         n = m_get(M_NOWAIT, m->m_type);
1951                         if (n != NULL) {
1952                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1953                                         bcopy(m->m_data, n->m_data, m->m_len);
1954                                         n->m_len = m->m_len;
1955                                 } else {
1956                                         m_free(n);
1957                                         n = NULL;
1958                                 }
1959                         }
1960                 } else
1961                         n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1962                 if (n == NULL)
1963                         continue;
1964                 if (inp->inp_flags & INP_CONTROLOPTS)
1965                         ip6_savecontrol(inp, n, &opts);
1966                 /* strip intermediate headers */
1967                 m_adj(n, off);
1968                 SOCKBUF_LOCK(&inp->inp_socket->so_rcv);
1969                 if (sbappendaddr_locked(&inp->inp_socket->so_rcv,
1970                     (struct sockaddr *)&fromsa, n, opts) == 0) {
1971                         soroverflow_locked(inp->inp_socket);
1972                         m_freem(n);
1973                         if (opts)
1974                                 m_freem(opts);
1975                 } else {
1976                         sorwakeup_locked(inp->inp_socket);
1977                         delivered++;
1978                 }
1979                 opts = NULL;
1980         }
1981         m_freem(m);
1982         *mp = NULL;
1983         if (delivered == 0)
1984                 IP6STAT_DEC(ip6s_delivered);
1985         return (IPPROTO_DONE);
1986 }
1987
1988 /*
1989  * Reflect the ip6 packet back to the source.
1990  * OFF points to the icmp6 header, counted from the top of the mbuf.
1991  */
1992 static void
1993 icmp6_reflect(struct mbuf *m, size_t off)
1994 {
1995         struct in6_addr src6, *srcp;
1996         struct ip6_hdr *ip6;
1997         struct icmp6_hdr *icmp6;
1998         struct in6_ifaddr *ia = NULL;
1999         struct ifnet *outif = NULL;
2000         int plen;
2001         int type, code, hlim;
2002
2003         /* too short to reflect */
2004         if (off < sizeof(struct ip6_hdr)) {
2005                 nd6log((LOG_DEBUG,
2006                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2007                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2008                     __FILE__, __LINE__));
2009                 goto bad;
2010         }
2011
2012         /*
2013          * If there are extra headers between IPv6 and ICMPv6, strip
2014          * off that header first.
2015          */
2016 #ifdef DIAGNOSTIC
2017         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2018                 panic("assumption failed in icmp6_reflect");
2019 #endif
2020         if (off > sizeof(struct ip6_hdr)) {
2021                 size_t l;
2022                 struct ip6_hdr nip6;
2023
2024                 l = off - sizeof(struct ip6_hdr);
2025                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2026                 m_adj(m, l);
2027                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2028                 if (m->m_len < l) {
2029                         if ((m = m_pullup(m, l)) == NULL)
2030                                 return;
2031                 }
2032                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2033         } else /* off == sizeof(struct ip6_hdr) */ {
2034                 size_t l;
2035                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2036                 if (m->m_len < l) {
2037                         if ((m = m_pullup(m, l)) == NULL)
2038                                 return;
2039                 }
2040         }
2041         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2042         ip6 = mtod(m, struct ip6_hdr *);
2043         ip6->ip6_nxt = IPPROTO_ICMPV6;
2044         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2045         type = icmp6->icmp6_type; /* keep type for statistics */
2046         code = icmp6->icmp6_code; /* ditto. */
2047         hlim = 0;
2048         srcp = NULL;
2049
2050         /*
2051          * If the incoming packet was addressed directly to us (i.e. unicast),
2052          * use dst as the src for the reply.
2053          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2054          * (for example) when we encounter an error while forwarding procedure
2055          * destined to a duplicated address of ours.
2056          */
2057         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2058                 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
2059                 if (ia != NULL && !(ia->ia6_flags &
2060                     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2061                         src6 = ia->ia_addr.sin6_addr;
2062                         srcp = &src6;
2063
2064                         if (m->m_pkthdr.rcvif != NULL) {
2065                                 /* XXX: This may not be the outgoing interface */
2066                                 hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2067                         } else
2068                                 hlim = V_ip6_defhlim;
2069                 }
2070         }
2071
2072         if (srcp == NULL) {
2073                 int error;
2074                 struct in6_addr dst6;
2075                 uint32_t scopeid;
2076
2077                 /*
2078                  * This case matches to multicasts, our anycast, or unicasts
2079                  * that we do not own.  Select a source address based on the
2080                  * source address of the erroneous packet.
2081                  */
2082                 in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2083                 error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2084                     scopeid, NULL, &src6, &hlim);
2085
2086                 if (error) {
2087                         char ip6buf[INET6_ADDRSTRLEN];
2088                         nd6log((LOG_DEBUG,
2089                             "icmp6_reflect: source can't be determined: "
2090                             "dst=%s, error=%d\n",
2091                             ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2092                         goto bad;
2093                 }
2094                 srcp = &src6;
2095         }
2096         /*
2097          * ip6_input() drops a packet if its src is multicast.
2098          * So, the src is never multicast.
2099          */
2100         ip6->ip6_dst = ip6->ip6_src;
2101         ip6->ip6_src = *srcp;
2102         ip6->ip6_flow = 0;
2103         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2104         ip6->ip6_vfc |= IPV6_VERSION;
2105         ip6->ip6_nxt = IPPROTO_ICMPV6;
2106         ip6->ip6_hlim = hlim;
2107
2108         icmp6->icmp6_cksum = 0;
2109         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2110             sizeof(struct ip6_hdr), plen);
2111
2112         /*
2113          * XXX option handling
2114          */
2115
2116         m->m_flags &= ~(M_BCAST|M_MCAST);
2117         m->m_pkthdr.rcvif = NULL;
2118         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2119         if (outif)
2120                 icmp6_ifoutstat_inc(outif, type, code);
2121
2122         return;
2123
2124  bad:
2125         m_freem(m);
2126         return;
2127 }
2128
2129 void
2130 icmp6_fasttimo(void)
2131 {
2132
2133         mld_fasttimo();
2134 }
2135
2136 void
2137 icmp6_slowtimo(void)
2138 {
2139
2140         mld_slowtimo();
2141 }
2142
2143 static const char *
2144 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2145     struct in6_addr *tgt6)
2146 {
2147         static char buf[1024];
2148         char ip6bufs[INET6_ADDRSTRLEN];
2149         char ip6bufd[INET6_ADDRSTRLEN];
2150         char ip6buft[INET6_ADDRSTRLEN];
2151         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2152             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2153             ip6_sprintf(ip6buft, tgt6));
2154         return buf;
2155 }
2156
2157 void
2158 icmp6_redirect_input(struct mbuf *m, int off)
2159 {
2160         struct ifnet *ifp;
2161         struct ip6_hdr *ip6;
2162         struct nd_redirect *nd_rd;
2163         struct in6_addr src6, redtgt6, reddst6;
2164         union nd_opts ndopts;
2165         char ip6buf[INET6_ADDRSTRLEN];
2166         char *lladdr;
2167         int icmp6len, is_onlink, is_router, lladdrlen;
2168
2169         M_ASSERTPKTHDR(m);
2170         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2171
2172         /* XXX if we are router, we don't update route by icmp6 redirect */
2173         if (V_ip6_forwarding)
2174                 goto freeit;
2175         if (!V_icmp6_rediraccept)
2176                 goto freeit;
2177
2178         /* RFC 6980: Nodes MUST silently ignore fragments */
2179         if(m->m_flags & M_FRAGMENTED)
2180                 goto freeit;
2181
2182         ip6 = mtod(m, struct ip6_hdr *);
2183         icmp6len = ntohs(ip6->ip6_plen);
2184         if (m->m_len < off + icmp6len) {
2185                 m = m_pullup(m, off + icmp6len);
2186                 if (m == NULL) {
2187                         IP6STAT_INC(ip6s_exthdrtoolong);
2188                         return;
2189                 }
2190         }
2191         ip6 = mtod(m, struct ip6_hdr *);
2192         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2193
2194         ifp = m->m_pkthdr.rcvif;
2195         redtgt6 = nd_rd->nd_rd_target;
2196         reddst6 = nd_rd->nd_rd_dst;
2197
2198         if (in6_setscope(&redtgt6, ifp, NULL) ||
2199             in6_setscope(&reddst6, ifp, NULL)) {
2200                 goto freeit;
2201         }
2202
2203         /* validation */
2204         src6 = ip6->ip6_src;
2205         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2206                 nd6log((LOG_ERR,
2207                     "ICMP6 redirect sent from %s rejected; "
2208                     "must be from linklocal\n",
2209                     ip6_sprintf(ip6buf, &src6)));
2210                 goto bad;
2211         }
2212         if (__predict_false(ip6->ip6_hlim != 255)) {
2213                 ICMP6STAT_INC(icp6s_invlhlim);
2214                 nd6log((LOG_ERR,
2215                     "ICMP6 redirect sent from %s rejected; "
2216                     "hlim=%d (must be 255)\n",
2217                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2218                 goto bad;
2219         }
2220     {
2221         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2222         struct nhop_object *nh;
2223         struct in6_addr kdst;
2224         uint32_t scopeid;
2225
2226         in6_splitscope(&reddst6, &kdst, &scopeid);
2227         NET_EPOCH_ASSERT();
2228         nh = fib6_lookup(ifp->if_fib, &kdst, scopeid, 0, 0);
2229         if (nh != NULL) {
2230                 struct in6_addr nh_addr;
2231                 nh_addr = ifatoia6(nh->nh_ifa)->ia_addr.sin6_addr;
2232                 if ((nh->nh_flags & NHF_GATEWAY) == 0) {
2233                         nd6log((LOG_ERR,
2234                             "ICMP6 redirect rejected; no route "
2235                             "with inet6 gateway found for redirect dst: %s\n",
2236                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2237                         goto bad;
2238                 }
2239
2240                 /*
2241                  * Embed scope zone id into next hop address.
2242                  */
2243                 nh_addr = nh->gw6_sa.sin6_addr;
2244
2245                 if (IN6_ARE_ADDR_EQUAL(&src6, &nh_addr) == 0) {
2246                         nd6log((LOG_ERR,
2247                             "ICMP6 redirect rejected; "
2248                             "not equal to gw-for-src=%s (must be same): "
2249                             "%s\n",
2250                             ip6_sprintf(ip6buf, &nh_addr),
2251                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2252                         goto bad;
2253                 }
2254         } else {
2255                 nd6log((LOG_ERR,
2256                     "ICMP6 redirect rejected; "
2257                     "no route found for redirect dst: %s\n",
2258                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2259                 goto bad;
2260         }
2261     }
2262         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2263                 nd6log((LOG_ERR,
2264                     "ICMP6 redirect rejected; "
2265                     "redirect dst must be unicast: %s\n",
2266                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2267                 goto bad;
2268         }
2269
2270         is_router = is_onlink = 0;
2271         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2272                 is_router = 1;  /* router case */
2273         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2274                 is_onlink = 1;  /* on-link destination case */
2275         if (!is_router && !is_onlink) {
2276                 nd6log((LOG_ERR,
2277                     "ICMP6 redirect rejected; "
2278                     "neither router case nor onlink case: %s\n",
2279                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2280                 goto bad;
2281         }
2282
2283         icmp6len -= sizeof(*nd_rd);
2284         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2285         if (nd6_options(&ndopts) < 0) {
2286                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2287                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2288                 /* nd6_options have incremented stats */
2289                 goto freeit;
2290         }
2291
2292         lladdr = NULL;
2293         lladdrlen = 0;
2294         if (ndopts.nd_opts_tgt_lladdr) {
2295                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2296                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2297         }
2298
2299         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2300                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2301                     "(if %d, icmp6 packet %d): %s\n",
2302                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2303                     ifp->if_addrlen, lladdrlen - 2,
2304                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2305                 goto bad;
2306         }
2307
2308         /* Validation passed. */
2309
2310         /* RFC 2461 8.3 */
2311         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2312             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2313
2314         /*
2315          * Install a gateway route in the better-router case or an interface
2316          * route in the on-link-destination case.
2317          */
2318         {
2319                 struct sockaddr_in6 sdst;
2320                 struct sockaddr_in6 sgw;
2321                 struct sockaddr_in6 ssrc;
2322                 struct sockaddr *gw;
2323                 int rt_flags;
2324                 u_int fibnum;
2325
2326                 bzero(&sdst, sizeof(sdst));
2327                 bzero(&ssrc, sizeof(ssrc));
2328                 sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2329                 sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2330                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2331                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2332                 rt_flags = 0;
2333                 if (is_router) {
2334                         bzero(&sgw, sizeof(sgw));
2335                         sgw.sin6_family = AF_INET6;
2336                         sgw.sin6_len = sizeof(struct sockaddr_in6);
2337                         bcopy(&redtgt6, &sgw.sin6_addr,
2338                                 sizeof(struct in6_addr));
2339                         gw = (struct sockaddr *)&sgw;
2340                         rt_flags |= RTF_GATEWAY;
2341                 } else
2342                         gw = ifp->if_addr->ifa_addr;
2343                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2344                         rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw,
2345                             (struct sockaddr *)&ssrc, ifp, rt_flags,
2346                             V_icmp6_redirtimeout);
2347         }
2348         /* finally update cached route in each socket via pfctlinput */
2349     {
2350         struct sockaddr_in6 sdst;
2351
2352         bzero(&sdst, sizeof(sdst));
2353         sdst.sin6_family = AF_INET6;
2354         sdst.sin6_len = sizeof(struct sockaddr_in6);
2355         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2356         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2357     }
2358
2359  freeit:
2360         m_freem(m);
2361         return;
2362
2363  bad:
2364         ICMP6STAT_INC(icp6s_badredirect);
2365         m_freem(m);
2366 }
2367
2368 void
2369 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh)
2370 {
2371         struct ifnet *ifp;      /* my outgoing interface */
2372         struct in6_addr *ifp_ll6;
2373         struct in6_addr *router_ll6;
2374         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2375         struct mbuf *m = NULL;  /* newly allocated one */
2376         struct m_tag *mtag;
2377         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2378         struct nd_redirect *nd_rd;
2379         struct llentry *ln = NULL;
2380         size_t maxlen;
2381         u_char *p;
2382         struct ifnet *outif = NULL;
2383         struct sockaddr_in6 src_sa;
2384
2385         icmp6_errcount(ND_REDIRECT, 0);
2386
2387         /* if we are not router, we don't send icmp6 redirect */
2388         if (!V_ip6_forwarding)
2389                 goto fail;
2390
2391         /* sanity check */
2392         if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp))
2393                 goto fail;
2394
2395         /*
2396          * Address check:
2397          *  the source address must identify a neighbor, and
2398          *  the destination address must not be a multicast address
2399          *  [RFC 2461, sec 8.2]
2400          */
2401         sip6 = mtod(m0, struct ip6_hdr *);
2402         bzero(&src_sa, sizeof(src_sa));
2403         src_sa.sin6_family = AF_INET6;
2404         src_sa.sin6_len = sizeof(src_sa);
2405         src_sa.sin6_addr = sip6->ip6_src;
2406         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2407                 goto fail;
2408         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2409                 goto fail;      /* what should we do here? */
2410
2411         /* rate limit */
2412         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2413                 goto fail;
2414
2415         /*
2416          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2417          * we almost always ask for an mbuf cluster for simplicity.
2418          * (MHLEN < IPV6_MMTU is almost always true)
2419          */
2420 #if IPV6_MMTU >= MCLBYTES
2421 # error assumption failed about IPV6_MMTU and MCLBYTES
2422 #endif
2423         m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2424         if (m == NULL)
2425                 goto fail;
2426         M_SETFIB(m, M_GETFIB(m0));
2427         maxlen = M_TRAILINGSPACE(m);
2428         maxlen = min(IPV6_MMTU, maxlen);
2429         /* just for safety */
2430         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2431             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2432                 goto fail;
2433         }
2434
2435         {
2436                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2437                 struct in6_ifaddr *ia;
2438                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2439                                                  IN6_IFF_NOTREADY|
2440                                                  IN6_IFF_ANYCAST)) == NULL)
2441                         goto fail;
2442                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2443                 /* XXXRW: reference released prematurely. */
2444                 ifa_free(&ia->ia_ifa);
2445         }
2446
2447         /* get ip6 linklocal address for the router. */
2448         if (nh->nh_flags & NHF_GATEWAY) {
2449                 struct sockaddr_in6 *sin6;
2450                 sin6 = &nh->gw6_sa;
2451                 router_ll6 = &sin6->sin6_addr;
2452                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2453                         router_ll6 = (struct in6_addr *)NULL;
2454         } else
2455                 router_ll6 = (struct in6_addr *)NULL;
2456
2457         /* ip6 */
2458         ip6 = mtod(m, struct ip6_hdr *);
2459         ip6->ip6_flow = 0;
2460         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2461         ip6->ip6_vfc |= IPV6_VERSION;
2462         /* ip6->ip6_plen will be set later */
2463         ip6->ip6_nxt = IPPROTO_ICMPV6;
2464         ip6->ip6_hlim = 255;
2465         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2466         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2467         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2468
2469         /* ND Redirect */
2470         nd_rd = (struct nd_redirect *)(ip6 + 1);
2471         nd_rd->nd_rd_type = ND_REDIRECT;
2472         nd_rd->nd_rd_code = 0;
2473         nd_rd->nd_rd_reserved = 0;
2474         if (nh->nh_flags & NHF_GATEWAY) {
2475                 /*
2476                  * nd_rd->nd_rd_target must be a link-local address in
2477                  * better router cases.
2478                  */
2479                 if (!router_ll6)
2480                         goto fail;
2481                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2482                     sizeof(nd_rd->nd_rd_target));
2483                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2484                     sizeof(nd_rd->nd_rd_dst));
2485         } else {
2486                 /* make sure redtgt == reddst */
2487                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2488                     sizeof(nd_rd->nd_rd_target));
2489                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2490                     sizeof(nd_rd->nd_rd_dst));
2491         }
2492
2493         p = (u_char *)(nd_rd + 1);
2494
2495         if (!router_ll6)
2496                 goto nolladdropt;
2497
2498         {
2499                 /* target lladdr option */
2500                 int len;
2501                 struct nd_opt_hdr *nd_opt;
2502                 char *lladdr;
2503
2504                 ln = nd6_lookup(router_ll6, LLE_SF(AF_INET6,  0), ifp);
2505                 if (ln == NULL)
2506                         goto nolladdropt;
2507
2508                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2509                 len = (len + 7) & ~7;   /* round by 8 */
2510                 /* safety check */
2511                 if (len + (p - (u_char *)ip6) > maxlen)
2512                         goto nolladdropt;
2513
2514                 if (ln->la_flags & LLE_VALID) {
2515                         nd_opt = (struct nd_opt_hdr *)p;
2516                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2517                         nd_opt->nd_opt_len = len >> 3;
2518                         lladdr = (char *)(nd_opt + 1);
2519                         bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2520                         p += len;
2521                 }
2522         }
2523 nolladdropt:
2524         if (ln != NULL)
2525                 LLE_RUNLOCK(ln);
2526
2527         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2528
2529         /* just to be safe */
2530 #ifdef M_DECRYPTED      /*not openbsd*/
2531         if (m0->m_flags & M_DECRYPTED)
2532                 goto noredhdropt;
2533 #endif
2534         if (p - (u_char *)ip6 > maxlen)
2535                 goto noredhdropt;
2536
2537         {
2538                 /* redirected header option */
2539                 int len;
2540                 struct nd_opt_rd_hdr *nd_opt_rh;
2541
2542                 /*
2543                  * compute the maximum size for icmp6 redirect header option.
2544                  * XXX room for auth header?
2545                  */
2546                 len = maxlen - (p - (u_char *)ip6);
2547                 len &= ~7;
2548
2549                 /* This is just for simplicity. */
2550                 if (m0->m_pkthdr.len != m0->m_len) {
2551                         if (m0->m_next) {
2552                                 m_freem(m0->m_next);
2553                                 m0->m_next = NULL;
2554                         }
2555                         m0->m_pkthdr.len = m0->m_len;
2556                 }
2557
2558                 /*
2559                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2560                  * about padding/truncate rule for the original IP packet.
2561                  * From the discussion on IPv6imp in Feb 1999,
2562                  * the consensus was:
2563                  * - "attach as much as possible" is the goal
2564                  * - pad if not aligned (original size can be guessed by
2565                  *   original ip6 header)
2566                  * Following code adds the padding if it is simple enough,
2567                  * and truncates if not.
2568                  */
2569                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2570                         panic("assumption failed in %s:%d", __FILE__,
2571                             __LINE__);
2572
2573                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2574                         /* not enough room, truncate */
2575                         m0->m_pkthdr.len = m0->m_len = len -
2576                             sizeof(*nd_opt_rh);
2577                 } else {
2578                         /* enough room, pad or truncate */
2579                         size_t extra;
2580
2581                         extra = m0->m_pkthdr.len % 8;
2582                         if (extra) {
2583                                 /* pad if easy enough, truncate if not */
2584                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2585                                         /* pad */
2586                                         m0->m_len += (8 - extra);
2587                                         m0->m_pkthdr.len += (8 - extra);
2588                                 } else {
2589                                         /* truncate */
2590                                         m0->m_pkthdr.len -= extra;
2591                                         m0->m_len -= extra;
2592                                 }
2593                         }
2594                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2595                         m0->m_pkthdr.len = m0->m_len = len -
2596                             sizeof(*nd_opt_rh);
2597                 }
2598
2599                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2600                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2601                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2602                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2603                 p += sizeof(*nd_opt_rh);
2604                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2605
2606                 /* connect m0 to m */
2607                 m_tag_delete_chain(m0, NULL);
2608                 m0->m_flags &= ~M_PKTHDR;
2609                 m->m_next = m0;
2610                 m->m_pkthdr.len = m->m_len + m0->m_len;
2611                 m0 = NULL;
2612         }
2613 noredhdropt:;
2614         if (m0) {
2615                 m_freem(m0);
2616                 m0 = NULL;
2617         }
2618
2619         /* XXX: clear embedded link IDs in the inner header */
2620         in6_clearscope(&sip6->ip6_src);
2621         in6_clearscope(&sip6->ip6_dst);
2622         in6_clearscope(&nd_rd->nd_rd_target);
2623         in6_clearscope(&nd_rd->nd_rd_dst);
2624
2625         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2626
2627         nd_rd->nd_rd_cksum = 0;
2628         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2629             sizeof(*ip6), ntohs(ip6->ip6_plen));
2630
2631         if (send_sendso_input_hook != NULL) {
2632                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2633                         M_NOWAIT);
2634                 if (mtag == NULL)
2635                         goto fail;
2636                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2637                 m_tag_prepend(m, mtag);
2638         }
2639
2640         /* send the packet to outside... */
2641         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2642         if (outif) {
2643                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2644                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2645         }
2646         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2647
2648         return;
2649
2650 fail:
2651         if (m)
2652                 m_freem(m);
2653         if (m0)
2654                 m_freem(m0);
2655 }
2656
2657 /*
2658  * ICMPv6 socket option processing.
2659  */
2660 int
2661 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2662 {
2663         int error = 0;
2664         int optlen;
2665         struct inpcb *inp = sotoinpcb(so);
2666         int level, op, optname;
2667
2668         if (sopt) {
2669                 level = sopt->sopt_level;
2670                 op = sopt->sopt_dir;
2671                 optname = sopt->sopt_name;
2672                 optlen = sopt->sopt_valsize;
2673         } else
2674                 level = op = optname = optlen = 0;
2675
2676         if (level != IPPROTO_ICMPV6) {
2677                 return EINVAL;
2678         }
2679
2680         switch (op) {
2681         case PRCO_SETOPT:
2682                 switch (optname) {
2683                 case ICMP6_FILTER:
2684                     {
2685                         struct icmp6_filter ic6f;
2686
2687                         if (optlen != sizeof(ic6f)) {
2688                                 error = EMSGSIZE;
2689                                 break;
2690                         }
2691                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2692                         if (error == 0) {
2693                                 INP_WLOCK(inp);
2694                                 *inp->in6p_icmp6filt = ic6f;
2695                                 INP_WUNLOCK(inp);
2696                         }
2697                         break;
2698                     }
2699
2700                 default:
2701                         error = ENOPROTOOPT;
2702                         break;
2703                 }
2704                 break;
2705
2706         case PRCO_GETOPT:
2707                 switch (optname) {
2708                 case ICMP6_FILTER:
2709                     {
2710                         struct icmp6_filter ic6f;
2711
2712                         INP_RLOCK(inp);
2713                         ic6f = *inp->in6p_icmp6filt;
2714                         INP_RUNLOCK(inp);
2715                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2716                         break;
2717                     }
2718
2719                 default:
2720                         error = ENOPROTOOPT;
2721                         break;
2722                 }
2723                 break;
2724         }
2725
2726         return (error);
2727 }
2728
2729 /*
2730  * Perform rate limit check.
2731  * Returns 0 if it is okay to send the icmp6 packet.
2732  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2733  * limitation.
2734  *
2735  * XXX per-destination/type check necessary?
2736  *
2737  * dst - not used at this moment
2738  * type - not used at this moment
2739  * code - not used at this moment
2740  */
2741 static int
2742 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2743     const int code)
2744 {
2745         int ret;
2746
2747         ret = 0;        /* okay to send */
2748
2749         /* PPS limit */
2750         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2751             V_icmp6errppslim)) {
2752                 /* The packet is subject to rate limit */
2753                 ret++;
2754         }
2755
2756         return ret;
2757 }