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/vnet.h>
  97
  98 #include <netinet/in.h>
  99 #include <netinet/in_pcb.h>
 100 #include <netinet/in_var.h>
 101 #include <netinet/ip6.h>
 102 #include <netinet/icmp6.h>
 103 #include <netinet/tcp_var.h>
 104
 105 #include <netinet6/in6_fib.h>
 106 #include <netinet6/in6_ifattach.h>
 107 #include <netinet6/in6_pcb.h>
 108 #include <netinet6/ip6protosw.h>
 109 #include <netinet6/ip6_var.h>
 110 #include <netinet6/scope6_var.h>
 111 #include <netinet6/mld6_var.h>
 112 #include <netinet6/nd6.h>
 113 #include <netinet6/send.h>
 114
 115 extern struct domain inet6domain;
 116
 117 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
 118 VNET_PCPUSTAT_SYSINIT(icmp6stat);
 119
 120 #ifdef VIMAGE
 121 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
 122 #endif /* VIMAGE */
 123
 124 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
 125 VNET_DECLARE(struct inpcbhead, ripcb);
 126 VNET_DECLARE(int, icmp6errppslim);
 127 VNET_DEFINE_STATIC(int, icmp6errpps_count) = 0;
 128 VNET_DEFINE_STATIC(struct timeval, icmp6errppslim_last);
 129 VNET_DECLARE(int, icmp6_nodeinfo);
 130
 131 #define V_ripcbinfo                     VNET(ripcbinfo)
 132 #define V_ripcb                         VNET(ripcb)
 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 int icmp6_ratelimit(const struct in6_addr *, const int, const int);
 141 static const char *icmp6_redirect_diag(struct in6_addr *,
 142         struct in6_addr *, struct in6_addr *);
 143 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *);
 144 static struct mbuf *ni6_nametodns(const char *, int, int);
 145 static int ni6_dnsmatch(const char *, int, const char *, int);
 146 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
 147                           struct ifnet **, struct in6_addr *);
 148 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
 149                                 struct ifnet *, int);
 150 static int icmp6_notify_error(struct mbuf **, int, int, int);
 151
 152 /*
 153  * Kernel module interface for updating icmp6stat.  The argument is an index
 154  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
 155  * general layout of icmp6stat into the caller, it doesn't encode its
 156  * location, so that future changes to add, for example, per-CPU stats
 157  * support won't cause binary compatibility problems for kernel modules.
 158  */
 159 void
 160 kmod_icmp6stat_inc(int statnum)
 161 {
 162
 163         counter_u64_add(VNET(icmp6stat)[statnum], 1);
 164 }
 165
 166 static void
 167 icmp6_errcount(int type, int code)
 168 {
 169         switch (type) {
 170         case ICMP6_DST_UNREACH:
 171                 switch (code) {
 172                 case ICMP6_DST_UNREACH_NOROUTE:
 173                         ICMP6STAT_INC(icp6s_odst_unreach_noroute);
 174                         return;
 175                 case ICMP6_DST_UNREACH_ADMIN:
 176                         ICMP6STAT_INC(icp6s_odst_unreach_admin);
 177                         return;
 178                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 179                         ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
 180                         return;
 181                 case ICMP6_DST_UNREACH_ADDR:
 182                         ICMP6STAT_INC(icp6s_odst_unreach_addr);
 183                         return;
 184                 case ICMP6_DST_UNREACH_NOPORT:
 185                         ICMP6STAT_INC(icp6s_odst_unreach_noport);
 186                         return;
 187                 }
 188                 break;
 189         case ICMP6_PACKET_TOO_BIG:
 190                 ICMP6STAT_INC(icp6s_opacket_too_big);
 191                 return;
 192         case ICMP6_TIME_EXCEEDED:
 193                 switch (code) {
 194                 case ICMP6_TIME_EXCEED_TRANSIT:
 195                         ICMP6STAT_INC(icp6s_otime_exceed_transit);
 196                         return;
 197                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 198                         ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
 199                         return;
 200                 }
 201                 break;
 202         case ICMP6_PARAM_PROB:
 203                 switch (code) {
 204                 case ICMP6_PARAMPROB_HEADER:
 205                         ICMP6STAT_INC(icp6s_oparamprob_header);
 206                         return;
 207                 case ICMP6_PARAMPROB_NEXTHEADER:
 208                         ICMP6STAT_INC(icp6s_oparamprob_nextheader);
 209                         return;
 210                 case ICMP6_PARAMPROB_OPTION:
 211                         ICMP6STAT_INC(icp6s_oparamprob_option);
 212                         return;
 213                 }
 214                 break;
 215         case ND_REDIRECT:
 216                 ICMP6STAT_INC(icp6s_oredirect);
 217                 return;
 218         }
 219         ICMP6STAT_INC(icp6s_ounknown);
 220 }
 221
 222 /*
 223  * A wrapper function for icmp6_error() necessary when the erroneous packet
 224  * may not contain enough scope zone information.
 225  */
 226 void
 227 icmp6_error2(struct mbuf *m, int type, int code, int param,
 228     struct ifnet *ifp)
 229 {
 230         struct ip6_hdr *ip6;
 231
 232         if (ifp == NULL)
 233                 return;
 234
 235         if (m->m_len < sizeof(struct ip6_hdr)) {
 236                 m = m_pullup(m, sizeof(struct ip6_hdr));
 237                 if (m == NULL) {
 238                         IP6STAT_INC(ip6s_exthdrtoolong);
 239                         return;
 240                 }
 241         }
 242         ip6 = mtod(m, struct ip6_hdr *);
 243
 244         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
 245                 return;
 246         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
 247                 return;
 248
 249         icmp6_error(m, type, code, param);
 250 }
 251
 252 /*
 253  * Generate an error packet of type error in response to bad IP6 packet.
 254  */
 255 void
 256 icmp6_error(struct mbuf *m, int type, int code, int param)
 257 {
 258         struct ip6_hdr *oip6, *nip6;
 259         struct icmp6_hdr *icmp6;
 260         u_int preplen;
 261         int off;
 262         int nxt;
 263
 264         ICMP6STAT_INC(icp6s_error);
 265
 266         /* count per-type-code statistics */
 267         icmp6_errcount(type, code);
 268
 269 #ifdef M_DECRYPTED      /*not openbsd*/
 270         if (m->m_flags & M_DECRYPTED) {
 271                 ICMP6STAT_INC(icp6s_canterror);
 272                 goto freeit;
 273         }
 274 #endif
 275
 276         if (m->m_len < sizeof(struct ip6_hdr)) {
 277                 m = m_pullup(m, sizeof(struct ip6_hdr));
 278                 if (m == NULL) {
 279                         IP6STAT_INC(ip6s_exthdrtoolong);
 280                         return;
 281                 }
 282         }
 283         oip6 = mtod(m, struct ip6_hdr *);
 284
 285         /*
 286          * If the destination address of the erroneous packet is a multicast
 287          * address, or the packet was sent using link-layer multicast,
 288          * we should basically suppress sending an error (RFC 2463, Section
 289          * 2.4).
 290          * We have two exceptions (the item e.2 in that section):
 291          * - the Packet Too Big message can be sent for path MTU discovery.
 292          * - the Parameter Problem Message that can be allowed an icmp6 error
 293          *   in the option type field.  This check has been done in
 294          *   ip6_unknown_opt(), so we can just check the type and code.
 295          */
 296         if ((m->m_flags & (M_BCAST|M_MCAST) ||
 297              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
 298             (type != ICMP6_PACKET_TOO_BIG &&
 299              (type != ICMP6_PARAM_PROB ||
 300               code != ICMP6_PARAMPROB_OPTION)))
 301                 goto freeit;
 302
 303         /*
 304          * RFC 2463, 2.4 (e.5): source address check.
 305          * XXX: the case of anycast source?
 306          */
 307         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
 308             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
 309                 goto freeit;
 310
 311         /*
 312          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
 313          * don't do it.
 314          */
 315         nxt = -1;
 316         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
 317         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
 318                 struct icmp6_hdr *icp;
 319
 320                 if (m->m_len < off + sizeof(struct icmp6_hdr)) {
 321                         m = m_pullup(m, off + sizeof(struct icmp6_hdr));
 322                         if (m == NULL) {
 323                                 IP6STAT_INC(ip6s_exthdrtoolong);
 324                                 return;
 325                         }
 326                 }
 327                 oip6 = mtod(m, struct ip6_hdr *);
 328                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 329
 330                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
 331                     icp->icmp6_type == ND_REDIRECT) {
 332                         /*
 333                          * ICMPv6 error
 334                          * Special case: for redirect (which is
 335                          * informational) we must not send icmp6 error.
 336                          */
 337                         ICMP6STAT_INC(icp6s_canterror);
 338                         goto freeit;
 339                 } else {
 340                         /* ICMPv6 informational - send the error */
 341                 }
 342         } else {
 343                 /* non-ICMPv6 - send the error */
 344         }
 345
 346         /* Finally, do rate limitation check. */
 347         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
 348                 ICMP6STAT_INC(icp6s_toofreq);
 349                 goto freeit;
 350         }
 351
 352         /*
 353          * OK, ICMP6 can be generated.
 354          */
 355
 356         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
 357                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
 358
 359         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
 360         M_PREPEND(m, preplen, M_NOWAIT);        /* FIB is also copied over. */
 361         if (m == NULL) {
 362                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
 363                 return;
 364         }
 365
 366         nip6 = mtod(m, struct ip6_hdr *);
 367         nip6->ip6_src  = oip6->ip6_src;
 368         nip6->ip6_dst  = oip6->ip6_dst;
 369
 370         in6_clearscope(&oip6->ip6_src);
 371         in6_clearscope(&oip6->ip6_dst);
 372
 373         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
 374         icmp6->icmp6_type = type;
 375         icmp6->icmp6_code = code;
 376         icmp6->icmp6_pptr = htonl((u_int32_t)param);
 377
 378         ICMP6STAT_INC(icp6s_outhist[type]);
 379         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
 380
 381         return;
 382
 383   freeit:
 384         /*
 385          * If we can't tell whether or not we can generate ICMP6, free it.
 386          */
 387         m_freem(m);
 388 }
 389
 390 /*
 391  * Process a received ICMP6 message.
 392  */
 393 int
 394 icmp6_input(struct mbuf **mp, int *offp, int proto)
 395 {
 396         struct mbuf *m, *n;
 397         struct ifnet *ifp;
 398         struct ip6_hdr *ip6, *nip6;
 399         struct icmp6_hdr *icmp6, *nicmp6;
 400         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
 401         int code, error, icmp6len, ip6len, noff, off, sum;
 402
 403         m = *mp;
 404         off = *offp;
 405
 406         if (m->m_len < off + sizeof(struct icmp6_hdr)) {
 407                 m = m_pullup(m, off + sizeof(struct icmp6_hdr));
 408                 if (m == NULL) {
 409                         IP6STAT_INC(ip6s_exthdrtoolong);
 410                         *mp = m;
 411                         return (IPPROTO_DONE);
 412                 }
 413         }
 414
 415         /*
 416          * Locate icmp6 structure in mbuf, and check
 417          * that not corrupted and of at least minimum length
 418          */
 419
 420         icmp6len = m->m_pkthdr.len - off;
 421         if (icmp6len < sizeof(struct icmp6_hdr)) {
 422                 ICMP6STAT_INC(icp6s_tooshort);
 423                 goto freeit;
 424         }
 425
 426         ip6 = mtod(m, struct ip6_hdr *);
 427         ifp = m->m_pkthdr.rcvif;
 428         /*
 429          * Check multicast group membership.
 430          * Note: SSM filters are not applied for ICMPv6 traffic.
 431          */
 432         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 433                 struct in6_multi        *inm;
 434
 435                 inm = in6m_lookup(ifp, &ip6->ip6_dst);
 436                 if (inm == NULL) {
 437                         IP6STAT_INC(ip6s_notmember);
 438                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
 439                         goto freeit;
 440                 }
 441         }
 442
 443         /* Calculate the checksum. */
 444         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
 445         code = icmp6->icmp6_code;
 446         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
 447                 nd6log((LOG_ERR,
 448                     "ICMP6 checksum error(%d|%x) %s\n",
 449                     icmp6->icmp6_type, sum,
 450                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
 451                 ICMP6STAT_INC(icp6s_checksum);
 452                 goto freeit;
 453         }
 454
 455         ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
 456         icmp6_ifstat_inc(ifp, ifs6_in_msg);
 457         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
 458                 icmp6_ifstat_inc(ifp, ifs6_in_error);
 459
 460         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
 461         switch (icmp6->icmp6_type) {
 462         case ICMP6_DST_UNREACH:
 463                 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
 464                 switch (code) {
 465                 case ICMP6_DST_UNREACH_NOROUTE:
 466                 case ICMP6_DST_UNREACH_ADDR:    /* PRC_HOSTDEAD is a DOS */
 467                         code = PRC_UNREACH_NET;
 468                         break;
 469                 case ICMP6_DST_UNREACH_ADMIN:
 470                         icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
 471                         code = PRC_UNREACH_ADMIN_PROHIB;
 472                         break;
 473                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
 474                         /* I mean "source address was incorrect." */
 475                         code = PRC_PARAMPROB;
 476                         break;
 477                 case ICMP6_DST_UNREACH_NOPORT:
 478                         code = PRC_UNREACH_PORT;
 479                         break;
 480                 default:
 481                         goto badcode;
 482                 }
 483                 goto deliver;
 484                 break;
 485
 486         case ICMP6_PACKET_TOO_BIG:
 487                 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
 488
 489                 /* validation is made in icmp6_mtudisc_update */
 490
 491                 code = PRC_MSGSIZE;
 492
 493                 /*
 494                  * Updating the path MTU will be done after examining
 495                  * intermediate extension headers.
 496                  */
 497                 goto deliver;
 498                 break;
 499
 500         case ICMP6_TIME_EXCEEDED:
 501                 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
 502                 switch (code) {
 503                 case ICMP6_TIME_EXCEED_TRANSIT:
 504                         code = PRC_TIMXCEED_INTRANS;
 505                         break;
 506                 case ICMP6_TIME_EXCEED_REASSEMBLY:
 507                         code = PRC_TIMXCEED_REASS;
 508                         break;
 509                 default:
 510                         goto badcode;
 511                 }
 512                 goto deliver;
 513                 break;
 514
 515         case ICMP6_PARAM_PROB:
 516                 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
 517                 switch (code) {
 518                 case ICMP6_PARAMPROB_NEXTHEADER:
 519                         code = PRC_UNREACH_PROTOCOL;
 520                         break;
 521                 case ICMP6_PARAMPROB_HEADER:
 522                 case ICMP6_PARAMPROB_OPTION:
 523                         code = PRC_PARAMPROB;
 524                         break;
 525                 default:
 526                         goto badcode;
 527                 }
 528                 goto deliver;
 529                 break;
 530
 531         case ICMP6_ECHO_REQUEST:
 532                 icmp6_ifstat_inc(ifp, ifs6_in_echo);
 533                 if (code != 0)
 534                         goto badcode;
 535                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
 536                         /* Give up remote */
 537                         break;
 538                 }
 539                 if (!M_WRITABLE(n)
 540                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
 541                         struct mbuf *n0 = n;
 542                         int n0len;
 543
 544                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
 545                         n = m_gethdr(M_NOWAIT, n0->m_type);
 546                         if (n == NULL) {
 547                                 /* Give up remote */
 548                                 m_freem(n0);
 549                                 break;
 550                         }
 551
 552                         m_move_pkthdr(n, n0);   /* FIB copied. */
 553                         n0len = n0->m_pkthdr.len;       /* save for use below */
 554                         /*
 555                          * Copy IPv6 and ICMPv6 only.
 556                          */
 557                         nip6 = mtod(n, struct ip6_hdr *);
 558                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 559                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 560                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 561                         noff = sizeof(struct ip6_hdr);
 562                         /* new mbuf contains only ipv6+icmpv6 headers */
 563                         n->m_len = noff + sizeof(struct icmp6_hdr);
 564                         /*
 565                          * Adjust mbuf.  ip6_plen will be adjusted in
 566                          * ip6_output().
 567                          */
 568                         m_adj(n0, off + sizeof(struct icmp6_hdr));
 569                         /* recalculate complete packet size */
 570                         n->m_pkthdr.len = n0len + (noff - off);
 571                         n->m_next = n0;
 572                 } else {
 573                         if (n->m_len < off + sizeof(*nicmp6)) {
 574                                 n = m_pullup(n, off + sizeof(*nicmp6));
 575                                 if (n == NULL) {
 576                                         IP6STAT_INC(ip6s_exthdrtoolong);
 577                                         break;
 578                                 }
 579                         }
 580                         nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
 581                         noff = off;
 582                 }
 583                 if (n) {
 584                         nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
 585                         nicmp6->icmp6_code = 0;
 586                         ICMP6STAT_INC(icp6s_reflect);
 587                         ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
 588                         icmp6_reflect(n, noff);
 589                 }
 590                 break;
 591
 592         case ICMP6_ECHO_REPLY:
 593                 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
 594                 if (code != 0)
 595                         goto badcode;
 596                 break;
 597
 598         case MLD_LISTENER_QUERY:
 599         case MLD_LISTENER_REPORT:
 600         case MLD_LISTENER_DONE:
 601         case MLDV2_LISTENER_REPORT:
 602                 /*
 603                  * Drop MLD traffic which is not link-local, has a hop limit
 604                  * of greater than 1 hop, or which does not have the
 605                  * IPv6 HBH Router Alert option.
 606                  * As IPv6 HBH options are stripped in ip6_input() we must
 607                  * check an mbuf header flag.
 608                  * XXX Should we also sanity check that these messages
 609                  * were directed to a link-local multicast prefix?
 610                  */
 611                 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
 612                         goto freeit;
 613                 if (mld_input(&m, off, icmp6len) != 0) {
 614                         *mp = NULL;
 615                         return (IPPROTO_DONE);
 616                 }
 617                 /* m stays. */
 618                 break;
 619
 620         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
 621             {
 622                 enum { WRU, FQDN } mode;
 623                 struct prison *pr;
 624
 625                 if (!V_icmp6_nodeinfo)
 626                         break;
 627
 628                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
 629                         mode = WRU;
 630                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
 631                         mode = FQDN;
 632                 else
 633                         goto badlen;
 634
 635                 pr = NULL;
 636                 sx_slock(&allprison_lock);
 637                 TAILQ_FOREACH(pr, &allprison, pr_list)
 638                         if (pr->pr_vnet == ifp->if_vnet)
 639                                 break;
 640                 sx_sunlock(&allprison_lock);
 641                 if (pr == NULL)
 642                         pr = curthread->td_ucred->cr_prison;
 643                 if (mode == FQDN) {
 644                         if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
 645                                 m = m_pullup(m, off +
 646                                     sizeof(struct icmp6_nodeinfo));
 647                                 if (m == NULL) {
 648                                         IP6STAT_INC(ip6s_exthdrtoolong);
 649                                         *mp = m;
 650                                         return (IPPROTO_DONE);
 651                                 }
 652                         }
 653                         n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 654                         if (n)
 655                                 n = ni6_input(n, off, pr);
 656                         /* XXX meaningless if n == NULL */
 657                         noff = sizeof(struct ip6_hdr);
 658                 } else {
 659                         u_char *p;
 660                         int maxhlen, hlen;
 661
 662                         /*
 663                          * XXX: this combination of flags is pointless,
 664                          * but should we keep this for compatibility?
 665                          */
 666                         if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
 667                             ICMP6_NODEINFO_TMPADDROK)) !=
 668                             (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
 669                                 break;
 670
 671                         if (code != 0)
 672                                 goto badcode;
 673
 674                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
 675                         n = m_gethdr(M_NOWAIT, m->m_type);
 676                         if (n == NULL) {
 677                                 /* Give up remote */
 678                                 break;
 679                         }
 680                         if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
 681                                 /*
 682                                  * Previous code did a blind M_COPY_PKTHDR
 683                                  * and said "just for rcvif".  If true, then
 684                                  * we could tolerate the dup failing (due to
 685                                  * the deep copy of the tag chain).  For now
 686                                  * be conservative and just fail.
 687                                  */
 688                                 m_free(n);
 689                                 n = NULL;
 690                                 break;
 691                         }
 692                         /*
 693                          * Copy IPv6 and ICMPv6 only.
 694                          */
 695                         nip6 = mtod(n, struct ip6_hdr *);
 696                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
 697                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
 698                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
 699                         p = (u_char *)(nicmp6 + 1);
 700                         bzero(p, 4);
 701
 702                         maxhlen = M_TRAILINGSPACE(n) -
 703                             (sizeof(*nip6) + sizeof(*nicmp6) + 4);
 704                         mtx_lock(&pr->pr_mtx);
 705                         hlen = strlen(pr->pr_hostname);
 706                         if (maxhlen > hlen)
 707                                 maxhlen = hlen;
 708                         /* meaningless TTL */
 709                         bcopy(pr->pr_hostname, p + 4, maxhlen);
 710                         mtx_unlock(&pr->pr_mtx);
 711                         noff = sizeof(struct ip6_hdr);
 712                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
 713                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
 714                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
 715                         nicmp6->icmp6_code = 0;
 716                 }
 717                 if (n) {
 718                         ICMP6STAT_INC(icp6s_reflect);
 719                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
 720                         icmp6_reflect(n, noff);
 721                 }
 722                 break;
 723             }
 724
 725         case ICMP6_WRUREPLY:
 726                 if (code != 0)
 727                         goto badcode;
 728                 break;
 729
 730         case ND_ROUTER_SOLICIT:
 731                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
 732                 if (code != 0)
 733                         goto badcode;
 734                 if (icmp6len < sizeof(struct nd_router_solicit))
 735                         goto badlen;
 736                 if (send_sendso_input_hook != NULL) {
 737                         if (m->m_len < off + icmp6len) {
 738                                 m = m_pullup(m, off + icmp6len);
 739                                 if (m == NULL) {
 740                                         IP6STAT_INC(ip6s_exthdrtoolong);
 741                                         *mp = NULL;
 742                                         return (IPPROTO_DONE);
 743                                 }
 744                         }
 745                         error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
 746                         if (error == 0) {
 747                                 m = NULL;
 748                                 goto freeit;
 749                         }
 750                 }
 751                 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 752                 nd6_rs_input(m, off, icmp6len);
 753                 m = n;
 754                 if (m == NULL)
 755                         goto freeit;
 756                 break;
 757
 758         case ND_ROUTER_ADVERT:
 759                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
 760                 if (code != 0)
 761                         goto badcode;
 762                 if (icmp6len < sizeof(struct nd_router_advert))
 763                         goto badlen;
 764                 if (send_sendso_input_hook != NULL) {
 765                         error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
 766                         if (error == 0) {
 767                                 m = NULL;
 768                                 goto freeit;
 769                         }
 770                 }
 771                 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 772                 nd6_ra_input(m, off, icmp6len);
 773                 m = n;
 774                 if (m == NULL)
 775                         goto freeit;
 776                 break;
 777
 778         case ND_NEIGHBOR_SOLICIT:
 779                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
 780                 if (code != 0)
 781                         goto badcode;
 782                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
 783                         goto badlen;
 784                 if (send_sendso_input_hook != NULL) {
 785                         error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
 786                         if (error == 0) {
 787                                 m = NULL;
 788                                 goto freeit;
 789                         }
 790                 }
 791                 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 792                 nd6_ns_input(m, off, icmp6len);
 793                 m = n;
 794                 if (m == NULL)
 795                         goto freeit;
 796                 break;
 797
 798         case ND_NEIGHBOR_ADVERT:
 799                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
 800                 if (code != 0)
 801                         goto badcode;
 802                 if (icmp6len < sizeof(struct nd_neighbor_advert))
 803                         goto badlen;
 804                 if (send_sendso_input_hook != NULL) {
 805                         error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
 806                         if (error == 0) {
 807                                 m = NULL;
 808                                 goto freeit;
 809                         }
 810                 }
 811                 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 812                 nd6_na_input(m, off, icmp6len);
 813                 m = n;
 814                 if (m == NULL)
 815                         goto freeit;
 816                 break;
 817
 818         case ND_REDIRECT:
 819                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
 820                 if (code != 0)
 821                         goto badcode;
 822                 if (icmp6len < sizeof(struct nd_redirect))
 823                         goto badlen;
 824                 if (send_sendso_input_hook != NULL) {
 825                         error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
 826                         if (error == 0) {
 827                                 m = NULL;
 828                                 goto freeit;
 829                         }
 830                 }
 831                 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
 832                 icmp6_redirect_input(m, off);
 833                 m = n;
 834                 if (m == NULL)
 835                         goto freeit;
 836                 break;
 837
 838         case ICMP6_ROUTER_RENUMBERING:
 839                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
 840                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
 841                         goto badcode;
 842                 if (icmp6len < sizeof(struct icmp6_router_renum))
 843                         goto badlen;
 844                 break;
 845
 846         default:
 847                 nd6log((LOG_DEBUG,
 848                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
 849                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
 850                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
 851                     ifp ? ifp->if_index : 0));
 852                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
 853                         /* ICMPv6 error: MUST deliver it by spec... */
 854                         code = PRC_NCMDS;
 855                         /* deliver */
 856                 } else {
 857                         /* ICMPv6 informational: MUST not deliver */
 858                         break;
 859                 }
 860         deliver:
 861                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
 862                         /* In this case, m should've been freed. */
 863                         *mp = NULL;
 864                         return (IPPROTO_DONE);
 865                 }
 866                 break;
 867
 868         badcode:
 869                 ICMP6STAT_INC(icp6s_badcode);
 870                 break;
 871
 872         badlen:
 873                 ICMP6STAT_INC(icp6s_badlen);
 874                 break;
 875         }
 876
 877         /* deliver the packet to appropriate sockets */
 878         icmp6_rip6_input(&m, *offp);
 879
 880         *mp = m;
 881         return (IPPROTO_DONE);
 882
 883  freeit:
 884         m_freem(m);
 885         *mp = NULL;
 886         return (IPPROTO_DONE);
 887 }
 888
 889 static int
 890 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
 891 {
 892         struct mbuf *m;
 893         struct icmp6_hdr *icmp6;
 894         struct ip6_hdr *eip6;
 895         u_int32_t notifymtu;
 896         struct sockaddr_in6 icmp6src, icmp6dst;
 897
 898         m = *mp;
 899
 900         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
 901                 ICMP6STAT_INC(icp6s_tooshort);
 902                 goto freeit;
 903         }
 904
 905         if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
 906                 m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
 907                 if (m == NULL) {
 908                         IP6STAT_INC(ip6s_exthdrtoolong);
 909                         *mp = m;
 910                         return (-1);
 911                 }
 912         }
 913         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
 914         eip6 = (struct ip6_hdr *)(icmp6 + 1);
 915
 916         /* Detect the upper level protocol */
 917         {
 918                 void (*ctlfunc)(int, struct sockaddr *, void *);
 919                 u_int8_t nxt = eip6->ip6_nxt;
 920                 int eoff = off + sizeof(struct icmp6_hdr) +
 921                     sizeof(struct ip6_hdr);
 922                 struct ip6ctlparam ip6cp;
 923                 struct in6_addr *finaldst = NULL;
 924                 int icmp6type = icmp6->icmp6_type;
 925                 struct ip6_frag *fh;
 926                 struct ip6_rthdr *rth;
 927                 struct ip6_rthdr0 *rth0;
 928                 int rthlen;
 929
 930                 while (1) { /* XXX: should avoid infinite loop explicitly? */
 931                         struct ip6_ext *eh;
 932
 933                         switch (nxt) {
 934                         case IPPROTO_HOPOPTS:
 935                         case IPPROTO_DSTOPTS:
 936                         case IPPROTO_AH:
 937                                 if (m->m_len < eoff + sizeof(struct ip6_ext)) {
 938                                         m = m_pullup(m, eoff +
 939                                             sizeof(struct ip6_ext));
 940                                         if (m == NULL) {
 941                                                 IP6STAT_INC(ip6s_exthdrtoolong);
 942                                                 *mp = m;
 943                                                 return (-1);
 944                                         }
 945                                 }
 946                                 eh = (struct ip6_ext *)
 947                                     (mtod(m, caddr_t) + eoff);
 948                                 if (nxt == IPPROTO_AH)
 949                                         eoff += (eh->ip6e_len + 2) << 2;
 950                                 else
 951                                         eoff += (eh->ip6e_len + 1) << 3;
 952                                 nxt = eh->ip6e_nxt;
 953                                 break;
 954                         case IPPROTO_ROUTING:
 955                                 /*
 956                                  * When the erroneous packet contains a
 957                                  * routing header, we should examine the
 958                                  * header to determine the final destination.
 959                                  * Otherwise, we can't properly update
 960                                  * information that depends on the final
 961                                  * destination (e.g. path MTU).
 962                                  */
 963                                 if (m->m_len < eoff + sizeof(*rth)) {
 964                                         m = m_pullup(m, eoff + sizeof(*rth));
 965                                         if (m == NULL) {
 966                                                 IP6STAT_INC(ip6s_exthdrtoolong);
 967                                                 *mp = m;
 968                                                 return (-1);
 969                                         }
 970                                 }
 971                                 rth = (struct ip6_rthdr *)
 972                                     (mtod(m, caddr_t) + eoff);
 973                                 rthlen = (rth->ip6r_len + 1) << 3;
 974                                 /*
 975                                  * XXX: currently there is no
 976                                  * officially defined type other
 977                                  * than type-0.
 978                                  * Note that if the segment left field
 979                                  * is 0, all intermediate hops must
 980                                  * have been passed.
 981                                  */
 982                                 if (rth->ip6r_segleft &&
 983                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
 984                                         int hops;
 985
 986                                         if (m->m_len < eoff + rthlen) {
 987                                                 m = m_pullup(m, eoff + rthlen);
 988                                                 if (m == NULL) {
 989                                                         IP6STAT_INC(
 990                                                             ip6s_exthdrtoolong);
 991                                                         *mp = m;
 992                                                         return (-1);
 993                                                 }
 994                                         }
 995                                         rth0 = (struct ip6_rthdr0 *)
 996                                             (mtod(m, caddr_t) + eoff);
 997                                         /* just ignore a bogus header */
 998                                         if ((rth0->ip6r0_len % 2) == 0 &&
 999                                             (hops = rth0->ip6r0_len/2))
1000                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1001                                 }
1002                                 eoff += rthlen;
1003                                 nxt = rth->ip6r_nxt;
1004                                 break;
1005                         case IPPROTO_FRAGMENT:
1006                                 if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1007                                         m = m_pullup(m, eoff +
1008                                             sizeof(struct ip6_frag));
1009                                         if (m == NULL) {
1010                                                 IP6STAT_INC(ip6s_exthdrtoolong);
1011                                                 *mp = m;
1012                                                 return (-1);
1013                                         }
1014                                 }
1015                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1016                                     eoff);
1017                                 /*
1018                                  * Data after a fragment header is meaningless
1019                                  * unless it is the first fragment, but
1020                                  * we'll go to the notify label for path MTU
1021                                  * discovery.
1022                                  */
1023                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1024                                         goto notify;
1025
1026                                 eoff += sizeof(struct ip6_frag);
1027                                 nxt = fh->ip6f_nxt;
1028                                 break;
1029                         default:
1030                                 /*
1031                                  * This case includes ESP and the No Next
1032                                  * Header.  In such cases going to the notify
1033                                  * label does not have any meaning
1034                                  * (i.e. ctlfunc will be NULL), but we go
1035                                  * anyway since we might have to update
1036                                  * path MTU information.
1037                                  */
1038                                 goto notify;
1039                         }
1040                 }
1041           notify:
1042                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1043
1044                 /*
1045                  * retrieve parameters from the inner IPv6 header, and convert
1046                  * them into sockaddr structures.
1047                  * XXX: there is no guarantee that the source or destination
1048                  * addresses of the inner packet are in the same scope as
1049                  * the addresses of the icmp packet.  But there is no other
1050                  * way to determine the zone.
1051                  */
1052                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1053
1054                 bzero(&icmp6dst, sizeof(icmp6dst));
1055                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1056                 icmp6dst.sin6_family = AF_INET6;
1057                 if (finaldst == NULL)
1058                         icmp6dst.sin6_addr = eip6->ip6_dst;
1059                 else
1060                         icmp6dst.sin6_addr = *finaldst;
1061                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1062                         goto freeit;
1063                 bzero(&icmp6src, sizeof(icmp6src));
1064                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1065                 icmp6src.sin6_family = AF_INET6;
1066                 icmp6src.sin6_addr = eip6->ip6_src;
1067                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1068                         goto freeit;
1069                 icmp6src.sin6_flowinfo =
1070                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1071
1072                 if (finaldst == NULL)
1073                         finaldst = &eip6->ip6_dst;
1074                 ip6cp.ip6c_m = m;
1075                 ip6cp.ip6c_icmp6 = icmp6;
1076                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1077                 ip6cp.ip6c_off = eoff;
1078                 ip6cp.ip6c_finaldst = finaldst;
1079                 ip6cp.ip6c_src = &icmp6src;
1080                 ip6cp.ip6c_nxt = nxt;
1081
1082                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1083                         notifymtu = ntohl(icmp6->icmp6_mtu);
1084                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1085                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1086                 }
1087
1088                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1089                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1090                 if (ctlfunc) {
1091                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1092                             &ip6cp);
1093                 }
1094         }
1095         *mp = m;
1096         return (0);
1097
1098   freeit:
1099         m_freem(m);
1100         *mp = NULL;
1101         return (-1);
1102 }
1103
1104 void
1105 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1106 {
1107         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1108         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1109         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1110         u_int mtu = ntohl(icmp6->icmp6_mtu);
1111         struct in_conninfo inc;
1112
1113 #if 0
1114         /*
1115          * RFC2460 section 5, last paragraph.
1116          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1117          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1118          * due to packet translator in the middle.
1119          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1120          * special handling.
1121          */
1122         if (mtu < IPV6_MMTU)
1123                 return;
1124 #endif
1125
1126         /*
1127          * we reject ICMPv6 too big with abnormally small value.
1128          * XXX what is the good definition of "abnormally small"?
1129          */
1130         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1131                 return;
1132
1133         if (!validated)
1134                 return;
1135
1136         /*
1137          * In case the suggested mtu is less than IPV6_MMTU, we
1138          * only need to remember that it was for above mentioned
1139          * "alwaysfrag" case.
1140          * Try to be as close to the spec as possible.
1141          */
1142         if (mtu < IPV6_MMTU)
1143                 mtu = IPV6_MMTU - 8;
1144
1145         bzero(&inc, sizeof(inc));
1146         inc.inc_fibnum = M_GETFIB(m);
1147         inc.inc_flags |= INC_ISIPV6;
1148         inc.inc6_faddr = *dst;
1149         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1150                 return;
1151
1152         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1153                 tcp_hc_updatemtu(&inc, mtu);
1154                 ICMP6STAT_INC(icp6s_pmtuchg);
1155         }
1156 }
1157
1158 /*
1159  * Process a Node Information Query packet, based on
1160  * draft-ietf-ipngwg-icmp-name-lookups-07.
1161  *
1162  * Spec incompatibilities:
1163  * - IPv6 Subject address handling
1164  * - IPv4 Subject address handling support missing
1165  * - Proxy reply (answer even if it's not for me)
1166  * - joins NI group address at in6_ifattach() time only, does not cope
1167  *   with hostname changes by sethostname(3)
1168  */
1169 static struct mbuf *
1170 ni6_input(struct mbuf *m, int off, struct prison *pr)
1171 {
1172         struct icmp6_nodeinfo *ni6, *nni6;
1173         struct mbuf *n = NULL;
1174         u_int16_t qtype;
1175         int subjlen;
1176         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1177         struct ni_reply_fqdn *fqdn;
1178         int addrs;              /* for NI_QTYPE_NODEADDR */
1179         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1180         struct in6_addr in6_subj; /* subject address */
1181         struct ip6_hdr *ip6;
1182         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1183         char *subj = NULL;
1184         struct in6_ifaddr *ia6 = NULL;
1185
1186         ip6 = mtod(m, struct ip6_hdr *);
1187         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1188
1189         /*
1190          * Validate IPv6 source address.
1191          * The default configuration MUST be to refuse answering queries from
1192          * global-scope addresses according to RFC4602.
1193          * Notes:
1194          *  - it's not very clear what "refuse" means; this implementation
1195          *    simply drops it.
1196          *  - it's not very easy to identify global-scope (unicast) addresses
1197          *    since there are many prefixes for them.  It should be safer
1198          *    and in practice sufficient to check "all" but loopback and
1199          *    link-local (note that site-local unicast was deprecated and
1200          *    ULA is defined as global scope-wise)
1201          */
1202         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1203             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1204             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1205                 goto bad;
1206
1207         /*
1208          * Validate IPv6 destination address.
1209          *
1210          * The Responder must discard the Query without further processing
1211          * unless it is one of the Responder's unicast or anycast addresses, or
1212          * a link-local scope multicast address which the Responder has joined.
1213          * [RFC4602, Section 5.]
1214          */
1215         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1216                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1217                         goto bad;
1218                 /* else it's a link-local multicast, fine */
1219         } else {                /* unicast or anycast */
1220                 ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
1221                 if (ia6 == NULL)
1222                         goto bad; /* XXX impossible */
1223
1224                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1225                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1226                         ifa_free(&ia6->ia_ifa);
1227                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1228                                 "a temporary address in %s:%d",
1229                                __FILE__, __LINE__));
1230                         goto bad;
1231                 }
1232                 ifa_free(&ia6->ia_ifa);
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         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1662                 switch (ni6->ni_code) {
1663                 case ICMP6_NI_SUBJ_IPV6:
1664                         if (subj == NULL) /* must be impossible... */
1665                                 return (0);
1666                         break;
1667                 default:
1668                         /*
1669                          * XXX: we only support IPv6 subject address for
1670                          * this Qtype.
1671                          */
1672                         return (0);
1673                 }
1674         }
1675
1676         IFNET_RLOCK_NOSLEEP();
1677         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1678                 addrsofif = 0;
1679                 IF_ADDR_RLOCK(ifp);
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_ADDR_RUNLOCK(ifp);
1731                 if (iffound) {
1732                         *ifpp = ifp;
1733                         IFNET_RUNLOCK_NOSLEEP();
1734                         return (addrsofif);
1735                 }
1736
1737                 addrs += addrsofif;
1738         }
1739         IFNET_RUNLOCK_NOSLEEP();
1740
1741         return (addrs);
1742 }
1743
1744 static int
1745 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1746     struct ifnet *ifp0, int resid)
1747 {
1748         struct ifnet *ifp;
1749         struct in6_ifaddr *ifa6;
1750         struct ifaddr *ifa;
1751         struct ifnet *ifp_dep = NULL;
1752         int copied = 0, allow_deprecated = 0;
1753         u_char *cp = (u_char *)(nni6 + 1);
1754         int niflags = ni6->ni_flags;
1755         u_int32_t ltime;
1756
1757         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1758                 return (0);     /* needless to copy */
1759
1760         IFNET_RLOCK_NOSLEEP();
1761         ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1762   again:
1763
1764         for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1765                 IF_ADDR_RLOCK(ifp);
1766                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1767                         if (ifa->ifa_addr->sa_family != AF_INET6)
1768                                 continue;
1769                         ifa6 = (struct in6_ifaddr *)ifa;
1770
1771                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1772                             allow_deprecated == 0) {
1773                                 /*
1774                                  * prefererred address should be put before
1775                                  * deprecated addresses.
1776                                  */
1777
1778                                 /* record the interface for later search */
1779                                 if (ifp_dep == NULL)
1780                                         ifp_dep = ifp;
1781
1782                                 continue;
1783                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1784                             allow_deprecated != 0)
1785                                 continue; /* we now collect deprecated addrs */
1786
1787                         /* What do we have to do about ::1? */
1788                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1789                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1790                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1791                                         continue;
1792                                 break;
1793                         case IPV6_ADDR_SCOPE_SITELOCAL:
1794                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1795                                         continue;
1796                                 break;
1797                         case IPV6_ADDR_SCOPE_GLOBAL:
1798                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1799                                         continue;
1800                                 break;
1801                         default:
1802                                 continue;
1803                         }
1804
1805                         /*
1806                          * check if anycast is okay.
1807                          * XXX: just experimental.  not in the spec.
1808                          */
1809                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1810                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1811                                 continue;
1812                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1813                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1814                                 continue;
1815                         }
1816
1817                         /* now we can copy the address */
1818                         if (resid < sizeof(struct in6_addr) +
1819                             sizeof(u_int32_t)) {
1820                                 IF_ADDR_RUNLOCK(ifp);
1821                                 /*
1822                                  * We give up much more copy.
1823                                  * Set the truncate flag and return.
1824                                  */
1825                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1826                                 IFNET_RUNLOCK_NOSLEEP();
1827                                 return (copied);
1828                         }
1829
1830                         /*
1831                          * Set the TTL of the address.
1832                          * The TTL value should be one of the following
1833                          * according to the specification:
1834                          *
1835                          * 1. The remaining lifetime of a DHCP lease on the
1836                          *    address, or
1837                          * 2. The remaining Valid Lifetime of a prefix from
1838                          *    which the address was derived through Stateless
1839                          *    Autoconfiguration.
1840                          *
1841                          * Note that we currently do not support stateful
1842                          * address configuration by DHCPv6, so the former
1843                          * case can't happen.
1844                          */
1845                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1846                                 ltime = ND6_INFINITE_LIFETIME;
1847                         else {
1848                                 if (ifa6->ia6_lifetime.ia6t_expire >
1849                                     time_uptime)
1850                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1851                                 else
1852                                         ltime = 0;
1853                         }
1854
1855                         bcopy(&ltime, cp, sizeof(u_int32_t));
1856                         cp += sizeof(u_int32_t);
1857
1858                         /* copy the address itself */
1859                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1860                             sizeof(struct in6_addr));
1861                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1862                         cp += sizeof(struct in6_addr);
1863
1864                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1865                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1866                 }
1867                 IF_ADDR_RUNLOCK(ifp);
1868                 if (ifp0)       /* we need search only on the specified IF */
1869                         break;
1870         }
1871
1872         if (allow_deprecated == 0 && ifp_dep != NULL) {
1873                 ifp = ifp_dep;
1874                 allow_deprecated = 1;
1875
1876                 goto again;
1877         }
1878
1879         IFNET_RUNLOCK_NOSLEEP();
1880
1881         return (copied);
1882 }
1883
1884 /*
1885  * XXX almost dup'ed code with rip6_input.
1886  */
1887 static int
1888 icmp6_rip6_input(struct mbuf **mp, int off)
1889 {
1890         struct mbuf *m = *mp;
1891         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1892         struct inpcb *inp;
1893         struct inpcb *last = NULL;
1894         struct sockaddr_in6 fromsa;
1895         struct icmp6_hdr *icmp6;
1896         struct epoch_tracker et;
1897         struct mbuf *opts = NULL;
1898
1899         /* This is assumed to be safe; icmp6_input() does a pullup. */
1900         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1901
1902         /*
1903          * XXX: the address may have embedded scope zone ID, which should be
1904          * hidden from applications.
1905          */
1906         bzero(&fromsa, sizeof(fromsa));
1907         fromsa.sin6_family = AF_INET6;
1908         fromsa.sin6_len = sizeof(struct sockaddr_in6);
1909         fromsa.sin6_addr = ip6->ip6_src;
1910         if (sa6_recoverscope(&fromsa)) {
1911                 m_freem(m);
1912                 *mp = NULL;
1913                 return (IPPROTO_DONE);
1914         }
1915
1916         INP_INFO_RLOCK_ET(&V_ripcbinfo, et);
1917         CK_LIST_FOREACH(inp, &V_ripcb, inp_list) {
1918                 if ((inp->inp_vflag & INP_IPV6) == 0)
1919                         continue;
1920                 if (inp->inp_ip_p != IPPROTO_ICMPV6)
1921                         continue;
1922                 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1923                    !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1924                         continue;
1925                 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1926                    !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1927                         continue;
1928                 INP_RLOCK(inp);
1929                 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1930                         INP_RUNLOCK(inp);
1931                         continue;
1932                 }
1933                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1934                     inp->in6p_icmp6filt)) {
1935                         INP_RUNLOCK(inp);
1936                         continue;
1937                 }
1938                 if (last != NULL) {
1939                         struct  mbuf *n = NULL;
1940
1941                         /*
1942                          * Recent network drivers tend to allocate a single
1943                          * mbuf cluster, rather than to make a couple of
1944                          * mbufs without clusters.  Also, since the IPv6 code
1945                          * path tries to avoid m_pullup(), it is highly
1946                          * probable that we still have an mbuf cluster here
1947                          * even though the necessary length can be stored in an
1948                          * mbuf's internal buffer.
1949                          * Meanwhile, the default size of the receive socket
1950                          * buffer for raw sockets is not so large.  This means
1951                          * the possibility of packet loss is relatively higher
1952                          * than before.  To avoid this scenario, we copy the
1953                          * received data to a separate mbuf that does not use
1954                          * a cluster, if possible.
1955                          * XXX: it is better to copy the data after stripping
1956                          * intermediate headers.
1957                          */
1958                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1959                             m->m_len <= MHLEN) {
1960                                 n = m_get(M_NOWAIT, m->m_type);
1961                                 if (n != NULL) {
1962                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1963                                                 bcopy(m->m_data, n->m_data,
1964                                                       m->m_len);
1965                                                 n->m_len = m->m_len;
1966                                         } else {
1967                                                 m_free(n);
1968                                                 n = NULL;
1969                                         }
1970                                 }
1971                         }
1972                         if (n != NULL ||
1973                             (n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
1974                                 if (last->inp_flags & INP_CONTROLOPTS)
1975                                         ip6_savecontrol(last, n, &opts);
1976                                 /* strip intermediate headers */
1977                                 m_adj(n, off);
1978                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1979                                 if (sbappendaddr_locked(
1980                                     &last->inp_socket->so_rcv,
1981                                     (struct sockaddr *)&fromsa, n, opts)
1982                                     == 0) {
1983                                         /* should notify about lost packet */
1984                                         m_freem(n);
1985                                         if (opts) {
1986                                                 m_freem(opts);
1987                                         }
1988                                         SOCKBUF_UNLOCK(
1989                                             &last->inp_socket->so_rcv);
1990                                 } else
1991                                         sorwakeup_locked(last->inp_socket);
1992                                 opts = NULL;
1993                         }
1994                         INP_RUNLOCK(last);
1995                 }
1996                 last = inp;
1997         }
1998         INP_INFO_RUNLOCK_ET(&V_ripcbinfo, et);
1999         if (last != NULL) {
2000                 if (last->inp_flags & INP_CONTROLOPTS)
2001                         ip6_savecontrol(last, m, &opts);
2002                 /* strip intermediate headers */
2003                 m_adj(m, off);
2004
2005                 /* avoid using mbuf clusters if possible (see above) */
2006                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2007                     m->m_len <= MHLEN) {
2008                         struct mbuf *n;
2009
2010                         n = m_get(M_NOWAIT, m->m_type);
2011                         if (n != NULL) {
2012                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2013                                         bcopy(m->m_data, n->m_data, m->m_len);
2014                                         n->m_len = m->m_len;
2015
2016                                         m_freem(m);
2017                                         m = n;
2018                                 } else {
2019                                         m_freem(n);
2020                                         n = NULL;
2021                                 }
2022                         }
2023                 }
2024                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2025                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2026                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2027                         m_freem(m);
2028                         if (opts)
2029                                 m_freem(opts);
2030                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2031                 } else
2032                         sorwakeup_locked(last->inp_socket);
2033                 INP_RUNLOCK(last);
2034         } else {
2035                 m_freem(m);
2036                 IP6STAT_DEC(ip6s_delivered);
2037         }
2038         *mp = NULL;
2039         return (IPPROTO_DONE);
2040 }
2041
2042 /*
2043  * Reflect the ip6 packet back to the source.
2044  * OFF points to the icmp6 header, counted from the top of the mbuf.
2045  */
2046 void
2047 icmp6_reflect(struct mbuf *m, size_t off)
2048 {
2049         struct in6_addr src6, *srcp;
2050         struct ip6_hdr *ip6;
2051         struct icmp6_hdr *icmp6;
2052         struct in6_ifaddr *ia = NULL;
2053         struct ifnet *outif = NULL;
2054         int plen;
2055         int type, code, hlim;
2056
2057         /* too short to reflect */
2058         if (off < sizeof(struct ip6_hdr)) {
2059                 nd6log((LOG_DEBUG,
2060                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2061                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2062                     __FILE__, __LINE__));
2063                 goto bad;
2064         }
2065
2066         /*
2067          * If there are extra headers between IPv6 and ICMPv6, strip
2068          * off that header first.
2069          */
2070 #ifdef DIAGNOSTIC
2071         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2072                 panic("assumption failed in icmp6_reflect");
2073 #endif
2074         if (off > sizeof(struct ip6_hdr)) {
2075                 size_t l;
2076                 struct ip6_hdr nip6;
2077
2078                 l = off - sizeof(struct ip6_hdr);
2079                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2080                 m_adj(m, l);
2081                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2082                 if (m->m_len < l) {
2083                         if ((m = m_pullup(m, l)) == NULL)
2084                                 return;
2085                 }
2086                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2087         } else /* off == sizeof(struct ip6_hdr) */ {
2088                 size_t l;
2089                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2090                 if (m->m_len < l) {
2091                         if ((m = m_pullup(m, l)) == NULL)
2092                                 return;
2093                 }
2094         }
2095         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2096         ip6 = mtod(m, struct ip6_hdr *);
2097         ip6->ip6_nxt = IPPROTO_ICMPV6;
2098         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2099         type = icmp6->icmp6_type; /* keep type for statistics */
2100         code = icmp6->icmp6_code; /* ditto. */
2101         hlim = 0;
2102         srcp = NULL;
2103
2104         /*
2105          * If the incoming packet was addressed directly to us (i.e. unicast),
2106          * use dst as the src for the reply.
2107          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2108          * (for example) when we encounter an error while forwarding procedure
2109          * destined to a duplicated address of ours.
2110          */
2111         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2112                 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
2113                 if (ia != NULL && !(ia->ia6_flags &
2114                     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2115                         src6 = ia->ia_addr.sin6_addr;
2116                         srcp = &src6;
2117
2118                         if (m->m_pkthdr.rcvif != NULL) {
2119                                 /* XXX: This may not be the outgoing interface */
2120                                 hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2121                         } else
2122                                 hlim = V_ip6_defhlim;
2123                 }
2124                 if (ia != NULL)
2125                         ifa_free(&ia->ia_ifa);
2126         }
2127
2128         if (srcp == NULL) {
2129                 int error;
2130                 struct in6_addr dst6;
2131                 uint32_t scopeid;
2132
2133                 /*
2134                  * This case matches to multicasts, our anycast, or unicasts
2135                  * that we do not own.  Select a source address based on the
2136                  * source address of the erroneous packet.
2137                  */
2138                 in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2139                 error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2140                     scopeid, NULL, &src6, &hlim);
2141
2142                 if (error) {
2143                         char ip6buf[INET6_ADDRSTRLEN];
2144                         nd6log((LOG_DEBUG,
2145                             "icmp6_reflect: source can't be determined: "
2146                             "dst=%s, error=%d\n",
2147                             ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2148                         goto bad;
2149                 }
2150                 srcp = &src6;
2151         }
2152         /*
2153          * ip6_input() drops a packet if its src is multicast.
2154          * So, the src is never multicast.
2155          */
2156         ip6->ip6_dst = ip6->ip6_src;
2157         ip6->ip6_src = *srcp;
2158         ip6->ip6_flow = 0;
2159         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2160         ip6->ip6_vfc |= IPV6_VERSION;
2161         ip6->ip6_nxt = IPPROTO_ICMPV6;
2162         ip6->ip6_hlim = hlim;
2163
2164         icmp6->icmp6_cksum = 0;
2165         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2166             sizeof(struct ip6_hdr), plen);
2167
2168         /*
2169          * XXX option handling
2170          */
2171
2172         m->m_flags &= ~(M_BCAST|M_MCAST);
2173         m->m_pkthdr.rcvif = NULL;
2174         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2175         if (outif)
2176                 icmp6_ifoutstat_inc(outif, type, code);
2177
2178         return;
2179
2180  bad:
2181         m_freem(m);
2182         return;
2183 }
2184
2185 void
2186 icmp6_fasttimo(void)
2187 {
2188
2189         mld_fasttimo();
2190 }
2191
2192 void
2193 icmp6_slowtimo(void)
2194 {
2195
2196         mld_slowtimo();
2197 }
2198
2199 static const char *
2200 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2201     struct in6_addr *tgt6)
2202 {
2203         static char buf[1024];
2204         char ip6bufs[INET6_ADDRSTRLEN];
2205         char ip6bufd[INET6_ADDRSTRLEN];
2206         char ip6buft[INET6_ADDRSTRLEN];
2207         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2208             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2209             ip6_sprintf(ip6buft, tgt6));
2210         return buf;
2211 }
2212
2213 void
2214 icmp6_redirect_input(struct mbuf *m, int off)
2215 {
2216         struct ifnet *ifp;
2217         struct ip6_hdr *ip6;
2218         struct nd_redirect *nd_rd;
2219         struct in6_addr src6, redtgt6, reddst6;
2220         union nd_opts ndopts;
2221         char ip6buf[INET6_ADDRSTRLEN];
2222         char *lladdr;
2223         int icmp6len, is_onlink, is_router, lladdrlen;
2224
2225         M_ASSERTPKTHDR(m);
2226         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2227
2228         /* XXX if we are router, we don't update route by icmp6 redirect */
2229         if (V_ip6_forwarding)
2230                 goto freeit;
2231         if (!V_icmp6_rediraccept)
2232                 goto freeit;
2233
2234         /* RFC 6980: Nodes MUST silently ignore fragments */
2235         if(m->m_flags & M_FRAGMENTED)
2236                 goto freeit;
2237
2238         ip6 = mtod(m, struct ip6_hdr *);
2239         icmp6len = ntohs(ip6->ip6_plen);
2240         if (m->m_len < off + icmp6len) {
2241                 m = m_pullup(m, off + icmp6len);
2242                 if (m == NULL) {
2243                         IP6STAT_INC(ip6s_exthdrtoolong);
2244                         return;
2245                 }
2246         }
2247         ip6 = mtod(m, struct ip6_hdr *);
2248         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2249
2250         ifp = m->m_pkthdr.rcvif;
2251         redtgt6 = nd_rd->nd_rd_target;
2252         reddst6 = nd_rd->nd_rd_dst;
2253
2254         if (in6_setscope(&redtgt6, ifp, NULL) ||
2255             in6_setscope(&reddst6, ifp, NULL)) {
2256                 goto freeit;
2257         }
2258
2259         /* validation */
2260         src6 = ip6->ip6_src;
2261         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2262                 nd6log((LOG_ERR,
2263                     "ICMP6 redirect sent from %s rejected; "
2264                     "must be from linklocal\n",
2265                     ip6_sprintf(ip6buf, &src6)));
2266                 goto bad;
2267         }
2268         if (ip6->ip6_hlim != 255) {
2269                 nd6log((LOG_ERR,
2270                     "ICMP6 redirect sent from %s rejected; "
2271                     "hlim=%d (must be 255)\n",
2272                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2273                 goto bad;
2274         }
2275     {
2276         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2277         struct nhop6_basic nh6;
2278         struct in6_addr kdst;
2279         uint32_t scopeid;
2280
2281         in6_splitscope(&reddst6, &kdst, &scopeid);
2282         if (fib6_lookup_nh_basic(ifp->if_fib, &kdst, scopeid, 0, 0,&nh6)==0){
2283                 if ((nh6.nh_flags & NHF_GATEWAY) == 0) {
2284                         nd6log((LOG_ERR,
2285                             "ICMP6 redirect rejected; no route "
2286                             "with inet6 gateway found for redirect dst: %s\n",
2287                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2288                         goto bad;
2289                 }
2290
2291                 /*
2292                  * Embed scope zone id into next hop address, since
2293                  * fib6_lookup_nh_basic() returns address without embedded
2294                  * scope zone id.
2295                  */
2296                 if (in6_setscope(&nh6.nh_addr, m->m_pkthdr.rcvif, NULL))
2297                         goto freeit;
2298
2299                 if (IN6_ARE_ADDR_EQUAL(&src6, &nh6.nh_addr) == 0) {
2300                         nd6log((LOG_ERR,
2301                             "ICMP6 redirect rejected; "
2302                             "not equal to gw-for-src=%s (must be same): "
2303                             "%s\n",
2304                             ip6_sprintf(ip6buf, &nh6.nh_addr),
2305                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2306                         goto bad;
2307                 }
2308         } else {
2309                 nd6log((LOG_ERR,
2310                     "ICMP6 redirect rejected; "
2311                     "no route found for redirect dst: %s\n",
2312                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2313                 goto bad;
2314         }
2315     }
2316         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2317                 nd6log((LOG_ERR,
2318                     "ICMP6 redirect rejected; "
2319                     "redirect dst must be unicast: %s\n",
2320                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2321                 goto bad;
2322         }
2323
2324         is_router = is_onlink = 0;
2325         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2326                 is_router = 1;  /* router case */
2327         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2328                 is_onlink = 1;  /* on-link destination case */
2329         if (!is_router && !is_onlink) {
2330                 nd6log((LOG_ERR,
2331                     "ICMP6 redirect rejected; "
2332                     "neither router case nor onlink case: %s\n",
2333                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2334                 goto bad;
2335         }
2336
2337         icmp6len -= sizeof(*nd_rd);
2338         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2339         if (nd6_options(&ndopts) < 0) {
2340                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2341                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2342                 /* nd6_options have incremented stats */
2343                 goto freeit;
2344         }
2345
2346         lladdr = NULL;
2347         lladdrlen = 0;
2348         if (ndopts.nd_opts_tgt_lladdr) {
2349                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2350                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2351         }
2352
2353         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2354                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2355                     "(if %d, icmp6 packet %d): %s\n",
2356                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2357                     ifp->if_addrlen, lladdrlen - 2,
2358                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2359                 goto bad;
2360         }
2361
2362         /* Validation passed. */
2363
2364         /* RFC 2461 8.3 */
2365         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2366             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2367
2368         /*
2369          * Install a gateway route in the better-router case or an interface
2370          * route in the on-link-destination case.
2371          */
2372         {
2373                 struct sockaddr_in6 sdst;
2374                 struct sockaddr_in6 sgw;
2375                 struct sockaddr_in6 ssrc;
2376                 struct sockaddr *gw;
2377                 int rt_flags;
2378                 u_int fibnum;
2379
2380                 bzero(&sdst, sizeof(sdst));
2381                 bzero(&ssrc, sizeof(ssrc));
2382                 sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2383                 sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2384                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2385                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2386                 rt_flags = RTF_HOST;
2387                 if (is_router) {
2388                         bzero(&sgw, sizeof(sgw));
2389                         sgw.sin6_family = AF_INET6;
2390                         sgw.sin6_len = sizeof(struct sockaddr_in6);
2391                         bcopy(&redtgt6, &sgw.sin6_addr,
2392                                 sizeof(struct in6_addr));
2393                         gw = (struct sockaddr *)&sgw;
2394                         rt_flags |= RTF_GATEWAY;
2395                 } else
2396                         gw = ifp->if_addr->ifa_addr;
2397                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2398                         in6_rtredirect((struct sockaddr *)&sdst, gw,
2399                             (struct sockaddr *)NULL, rt_flags,
2400                             (struct sockaddr *)&ssrc, fibnum);
2401         }
2402         /* finally update cached route in each socket via pfctlinput */
2403     {
2404         struct sockaddr_in6 sdst;
2405
2406         bzero(&sdst, sizeof(sdst));
2407         sdst.sin6_family = AF_INET6;
2408         sdst.sin6_len = sizeof(struct sockaddr_in6);
2409         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2410         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2411     }
2412
2413  freeit:
2414         m_freem(m);
2415         return;
2416
2417  bad:
2418         ICMP6STAT_INC(icp6s_badredirect);
2419         m_freem(m);
2420 }
2421
2422 void
2423 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2424 {
2425         struct ifnet *ifp;      /* my outgoing interface */
2426         struct in6_addr *ifp_ll6;
2427         struct in6_addr *router_ll6;
2428         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2429         struct mbuf *m = NULL;  /* newly allocated one */
2430         struct m_tag *mtag;
2431         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2432         struct nd_redirect *nd_rd;
2433         struct llentry *ln = NULL;
2434         size_t maxlen;
2435         u_char *p;
2436         struct ifnet *outif = NULL;
2437         struct sockaddr_in6 src_sa;
2438
2439         icmp6_errcount(ND_REDIRECT, 0);
2440
2441         /* if we are not router, we don't send icmp6 redirect */
2442         if (!V_ip6_forwarding)
2443                 goto fail;
2444
2445         /* sanity check */
2446         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2447                 goto fail;
2448
2449         /*
2450          * Address check:
2451          *  the source address must identify a neighbor, and
2452          *  the destination address must not be a multicast address
2453          *  [RFC 2461, sec 8.2]
2454          */
2455         sip6 = mtod(m0, struct ip6_hdr *);
2456         bzero(&src_sa, sizeof(src_sa));
2457         src_sa.sin6_family = AF_INET6;
2458         src_sa.sin6_len = sizeof(src_sa);
2459         src_sa.sin6_addr = sip6->ip6_src;
2460         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2461                 goto fail;
2462         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2463                 goto fail;      /* what should we do here? */
2464
2465         /* rate limit */
2466         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2467                 goto fail;
2468
2469         /*
2470          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2471          * we almost always ask for an mbuf cluster for simplicity.
2472          * (MHLEN < IPV6_MMTU is almost always true)
2473          */
2474 #if IPV6_MMTU >= MCLBYTES
2475 # error assumption failed about IPV6_MMTU and MCLBYTES
2476 #endif
2477         m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2478         if (m == NULL)
2479                 goto fail;
2480         M_SETFIB(m, rt->rt_fibnum);
2481         maxlen = M_TRAILINGSPACE(m);
2482         maxlen = min(IPV6_MMTU, maxlen);
2483         /* just for safety */
2484         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2485             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2486                 goto fail;
2487         }
2488
2489         {
2490                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2491                 struct in6_ifaddr *ia;
2492                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2493                                                  IN6_IFF_NOTREADY|
2494                                                  IN6_IFF_ANYCAST)) == NULL)
2495                         goto fail;
2496                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2497                 /* XXXRW: reference released prematurely. */
2498                 ifa_free(&ia->ia_ifa);
2499         }
2500
2501         /* get ip6 linklocal address for the router. */
2502         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2503                 struct sockaddr_in6 *sin6;
2504                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2505                 router_ll6 = &sin6->sin6_addr;
2506                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2507                         router_ll6 = (struct in6_addr *)NULL;
2508         } else
2509                 router_ll6 = (struct in6_addr *)NULL;
2510
2511         /* ip6 */
2512         ip6 = mtod(m, struct ip6_hdr *);
2513         ip6->ip6_flow = 0;
2514         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2515         ip6->ip6_vfc |= IPV6_VERSION;
2516         /* ip6->ip6_plen will be set later */
2517         ip6->ip6_nxt = IPPROTO_ICMPV6;
2518         ip6->ip6_hlim = 255;
2519         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2520         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2521         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2522
2523         /* ND Redirect */
2524         nd_rd = (struct nd_redirect *)(ip6 + 1);
2525         nd_rd->nd_rd_type = ND_REDIRECT;
2526         nd_rd->nd_rd_code = 0;
2527         nd_rd->nd_rd_reserved = 0;
2528         if (rt->rt_flags & RTF_GATEWAY) {
2529                 /*
2530                  * nd_rd->nd_rd_target must be a link-local address in
2531                  * better router cases.
2532                  */
2533                 if (!router_ll6)
2534                         goto fail;
2535                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2536                     sizeof(nd_rd->nd_rd_target));
2537                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2538                     sizeof(nd_rd->nd_rd_dst));
2539         } else {
2540                 /* make sure redtgt == reddst */
2541                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2542                     sizeof(nd_rd->nd_rd_target));
2543                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2544                     sizeof(nd_rd->nd_rd_dst));
2545         }
2546
2547         p = (u_char *)(nd_rd + 1);
2548
2549         if (!router_ll6)
2550                 goto nolladdropt;
2551
2552         {
2553                 /* target lladdr option */
2554                 int len;
2555                 struct nd_opt_hdr *nd_opt;
2556                 char *lladdr;
2557
2558                 IF_AFDATA_RLOCK(ifp);
2559                 ln = nd6_lookup(router_ll6, 0, ifp);
2560                 IF_AFDATA_RUNLOCK(ifp);
2561                 if (ln == NULL)
2562                         goto nolladdropt;
2563
2564                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2565                 len = (len + 7) & ~7;   /* round by 8 */
2566                 /* safety check */
2567                 if (len + (p - (u_char *)ip6) > maxlen)
2568                         goto nolladdropt;
2569
2570                 if (ln->la_flags & LLE_VALID) {
2571                         nd_opt = (struct nd_opt_hdr *)p;
2572                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2573                         nd_opt->nd_opt_len = len >> 3;
2574                         lladdr = (char *)(nd_opt + 1);
2575                         bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2576                         p += len;
2577                 }
2578         }
2579 nolladdropt:
2580         if (ln != NULL)
2581                 LLE_RUNLOCK(ln);
2582
2583         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2584
2585         /* just to be safe */
2586 #ifdef M_DECRYPTED      /*not openbsd*/
2587         if (m0->m_flags & M_DECRYPTED)
2588                 goto noredhdropt;
2589 #endif
2590         if (p - (u_char *)ip6 > maxlen)
2591                 goto noredhdropt;
2592
2593         {
2594                 /* redirected header option */
2595                 int len;
2596                 struct nd_opt_rd_hdr *nd_opt_rh;
2597
2598                 /*
2599                  * compute the maximum size for icmp6 redirect header option.
2600                  * XXX room for auth header?
2601                  */
2602                 len = maxlen - (p - (u_char *)ip6);
2603                 len &= ~7;
2604
2605                 /* This is just for simplicity. */
2606                 if (m0->m_pkthdr.len != m0->m_len) {
2607                         if (m0->m_next) {
2608                                 m_freem(m0->m_next);
2609                                 m0->m_next = NULL;
2610                         }
2611                         m0->m_pkthdr.len = m0->m_len;
2612                 }
2613
2614                 /*
2615                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2616                  * about padding/truncate rule for the original IP packet.
2617                  * From the discussion on IPv6imp in Feb 1999,
2618                  * the consensus was:
2619                  * - "attach as much as possible" is the goal
2620                  * - pad if not aligned (original size can be guessed by
2621                  *   original ip6 header)
2622                  * Following code adds the padding if it is simple enough,
2623                  * and truncates if not.
2624                  */
2625                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2626                         panic("assumption failed in %s:%d", __FILE__,
2627                             __LINE__);
2628
2629                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2630                         /* not enough room, truncate */
2631                         m0->m_pkthdr.len = m0->m_len = len -
2632                             sizeof(*nd_opt_rh);
2633                 } else {
2634                         /* enough room, pad or truncate */
2635                         size_t extra;
2636
2637                         extra = m0->m_pkthdr.len % 8;
2638                         if (extra) {
2639                                 /* pad if easy enough, truncate if not */
2640                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2641                                         /* pad */
2642                                         m0->m_len += (8 - extra);
2643                                         m0->m_pkthdr.len += (8 - extra);
2644                                 } else {
2645                                         /* truncate */
2646                                         m0->m_pkthdr.len -= extra;
2647                                         m0->m_len -= extra;
2648                                 }
2649                         }
2650                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2651                         m0->m_pkthdr.len = m0->m_len = len -
2652                             sizeof(*nd_opt_rh);
2653                 }
2654
2655                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2656                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2657                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2658                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2659                 p += sizeof(*nd_opt_rh);
2660                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2661
2662                 /* connect m0 to m */
2663                 m_tag_delete_chain(m0, NULL);
2664                 m0->m_flags &= ~M_PKTHDR;
2665                 m->m_next = m0;
2666                 m->m_pkthdr.len = m->m_len + m0->m_len;
2667                 m0 = NULL;
2668         }
2669 noredhdropt:;
2670         if (m0) {
2671                 m_freem(m0);
2672                 m0 = NULL;
2673         }
2674
2675         /* XXX: clear embedded link IDs in the inner header */
2676         in6_clearscope(&sip6->ip6_src);
2677         in6_clearscope(&sip6->ip6_dst);
2678         in6_clearscope(&nd_rd->nd_rd_target);
2679         in6_clearscope(&nd_rd->nd_rd_dst);
2680
2681         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2682
2683         nd_rd->nd_rd_cksum = 0;
2684         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2685             sizeof(*ip6), ntohs(ip6->ip6_plen));
2686
2687         if (send_sendso_input_hook != NULL) {
2688                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2689                         M_NOWAIT);
2690                 if (mtag == NULL)
2691                         goto fail;
2692                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2693                 m_tag_prepend(m, mtag);
2694         }
2695
2696         /* send the packet to outside... */
2697         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2698         if (outif) {
2699                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2700                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2701         }
2702         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2703
2704         return;
2705
2706 fail:
2707         if (m)
2708                 m_freem(m);
2709         if (m0)
2710                 m_freem(m0);
2711 }
2712
2713 /*
2714  * ICMPv6 socket option processing.
2715  */
2716 int
2717 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2718 {
2719         int error = 0;
2720         int optlen;
2721         struct inpcb *inp = sotoinpcb(so);
2722         int level, op, optname;
2723
2724         if (sopt) {
2725                 level = sopt->sopt_level;
2726                 op = sopt->sopt_dir;
2727                 optname = sopt->sopt_name;
2728                 optlen = sopt->sopt_valsize;
2729         } else
2730                 level = op = optname = optlen = 0;
2731
2732         if (level != IPPROTO_ICMPV6) {
2733                 return EINVAL;
2734         }
2735
2736         switch (op) {
2737         case PRCO_SETOPT:
2738                 switch (optname) {
2739                 case ICMP6_FILTER:
2740                     {
2741                         struct icmp6_filter ic6f;
2742
2743                         if (optlen != sizeof(ic6f)) {
2744                                 error = EMSGSIZE;
2745                                 break;
2746                         }
2747                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2748                         if (error == 0) {
2749                                 INP_WLOCK(inp);
2750                                 *inp->in6p_icmp6filt = ic6f;
2751                                 INP_WUNLOCK(inp);
2752                         }
2753                         break;
2754                     }
2755
2756                 default:
2757                         error = ENOPROTOOPT;
2758                         break;
2759                 }
2760                 break;
2761
2762         case PRCO_GETOPT:
2763                 switch (optname) {
2764                 case ICMP6_FILTER:
2765                     {
2766                         struct icmp6_filter ic6f;
2767
2768                         INP_RLOCK(inp);
2769                         ic6f = *inp->in6p_icmp6filt;
2770                         INP_RUNLOCK(inp);
2771                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2772                         break;
2773                     }
2774
2775                 default:
2776                         error = ENOPROTOOPT;
2777                         break;
2778                 }
2779                 break;
2780         }
2781
2782         return (error);
2783 }
2784
2785 /*
2786  * Perform rate limit check.
2787  * Returns 0 if it is okay to send the icmp6 packet.
2788  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2789  * limitation.
2790  *
2791  * XXX per-destination/type check necessary?
2792  *
2793  * dst - not used at this moment
2794  * type - not used at this moment
2795  * code - not used at this moment
2796  */
2797 static int
2798 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2799     const int code)
2800 {
2801         int ret;
2802
2803         ret = 0;        /* okay to send */
2804
2805         /* PPS limit */
2806         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2807             V_icmp6errppslim)) {
2808                 /* The packet is subject to rate limit */
2809                 ret++;
2810         }
2811
2812         return ret;
2813 }