packages/net/bsd_tcpip/v2_0/src/sys/netinet6/ip6_output.c

   1 //==========================================================================
   2 //
   3 //      src/sys/netinet6/ip6_output.c
   4 //
   5 //==========================================================================
   6 //####BSDCOPYRIGHTBEGIN####
   7 //
   8 // -------------------------------------------
   9 //
  10 // Portions of this software may have been derived from OpenBSD,
  11 // FreeBSD or other sources, and are covered by the appropriate
  12 // copyright disclaimers included herein.
  13 //
  14 // Portions created by Red Hat are
  15 // Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
  16 //
  17 // -------------------------------------------
  18 //
  19 //####BSDCOPYRIGHTEND####
  20 //==========================================================================
  21
  22 /*      $KAME: ip6_output.c,v 1.272 2001/12/26 01:03:28 jinmei Exp $    */
  23
  24 /*
  25  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
  26  * All rights reserved.
  27  *
  28  * Redistribution and use in source and binary forms, with or without
  29  * modification, are permitted provided that the following conditions
  30  * are met:
  31  * 1. Redistributions of source code must retain the above copyright
  32  *    notice, this list of conditions and the following disclaimer.
  33  * 2. Redistributions in binary form must reproduce the above copyright
  34  *    notice, this list of conditions and the following disclaimer in the
  35  *    documentation and/or other materials provided with the distribution.
  36  * 3. Neither the name of the project nor the names of its contributors
  37  *    may be used to endorse or promote products derived from this software
  38  *    without specific prior written permission.
  39  *
  40  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  41  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  42  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  43  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  44  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  45  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  46  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  47  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  48  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  49  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  50  * SUCH DAMAGE.
  51  */
  52
  53 /*
  54  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  55  *      The Regents of the University of California.  All rights reserved.
  56  *
  57  * Redistribution and use in source and binary forms, with or without
  58  * modification, are permitted provided that the following conditions
  59  * are met:
  60  * 1. Redistributions of source code must retain the above copyright
  61  *    notice, this list of conditions and the following disclaimer.
  62  * 2. Redistributions in binary form must reproduce the above copyright
  63  *    notice, this list of conditions and the following disclaimer in the
  64  *    documentation and/or other materials provided with the distribution.
  65  * 3. All advertising materials mentioning features or use of this software
  66  *    must display the following acknowledgement:
  67  *      This product includes software developed by the University of
  68  *      California, Berkeley and its contributors.
  69  * 4. Neither the name of the University nor the names of its contributors
  70  *    may be used to endorse or promote products derived from this software
  71  *    without specific prior written permission.
  72  *
  73  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  74  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  75  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  76  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  77  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  78  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  79  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  80  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  81  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  82  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  83  * SUCH DAMAGE.
  84  *
  85  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  86  */
  87
  88 #include <sys/param.h>
  89 #include <sys/malloc.h>
  90 #include <sys/mbuf.h>
  91 #include <sys/errno.h>
  92 #include <sys/protosw.h>
  93 #include <sys/socket.h>
  94 #include <sys/socketvar.h>
  95
  96 #include <net/if.h>
  97 #include <net/route.h>
  98
  99 #include <netinet/in.h>
 100 #include <netinet/in_var.h>
 101 #include <netinet6/in6_var.h>
 102 #include <netinet/ip6.h>
 103 #include <netinet/icmp6.h>
 104 #include <netinet6/ip6_var.h>
 105 #include <netinet/in_pcb.h>
 106 #include <netinet6/nd6.h>
 107 #include <netinet6/ip6protosw.h>
 108 #include <netinet6/scope6_var.h>
 109
 110 #ifdef IPSEC
 111 #ifdef __OpenBSD__
 112 #include <netinet/ip_ah.h>
 113 #include <netinet/ip_esp.h>
 114 #include <netinet/udp.h>
 115 #include <netinet/tcp.h>
 116 #include <net/pfkeyv2.h>
 117
 118 extern u_int8_t get_sa_require  __P((struct inpcb *));
 119
 120 extern int ipsec_auth_default_level;
 121 extern int ipsec_esp_trans_default_level;
 122 extern int ipsec_esp_network_default_level;
 123 extern int ipsec_ipcomp_default_level;
 124 #else
 125 #include <netinet6/ipsec.h>
 126 #include <netkey/key.h>
 127 #endif
 128 #endif /* IPSEC */
 129
 130 #if defined(IPV6FIREWALL) || (defined(__FreeBSD__) && __FreeBSD__ >= 4)
 131 #include <netinet6/ip6_fw.h>
 132 #endif
 133
 134 #ifdef MIP6
 135 #include <sys/syslog.h>
 136 #include <netinet6/mip6.h>
 137 #endif /* MIP6 */
 138
 139 struct ip6_exthdrs {
 140         struct mbuf *ip6e_ip6;
 141         struct mbuf *ip6e_hbh;
 142         struct mbuf *ip6e_dest1;
 143         struct mbuf *ip6e_rthdr;
 144         struct mbuf *ip6e_haddr; /* for MIP6 */
 145         struct mbuf *ip6e_dest2;
 146 };
 147
 148 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **, int));
 149 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
 150 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
 151                             struct socket *, struct sockopt *));
 152 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct sockopt *));
 153 #else
 154 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
 155                             struct socket *));
 156 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct mbuf **));
 157 #endif
 158 static int ip6_setpktoption __P((int, u_char *, int, struct ip6_pktopts *, int,
 159                                  int, int));
 160 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
 161 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
 162 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
 163 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
 164                                   struct ip6_frag **));
 165 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
 166 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
 167 #ifdef NEW_STRUCT_ROUTE
 168 static int ip6_getpmtu __P((struct route *, struct route *, struct ifnet *,
 169                             struct in6_addr *, u_long *));
 170 #else
 171 static int ip6_getpmtu __P((struct route *, struct route *, struct ifnet *,
 172                             struct in6_addr *, u_long *));
 173 #endif
 174
 175 #ifdef __bsdi__
 176 #if _BSDI_VERSION < 199802
 177 extern struct ifnet loif;
 178 #else
 179 extern struct ifnet *loifp;
 180 #endif
 181 #endif
 182 #if defined(__NetBSD__)
 183 extern struct ifnet loif[NLOOP];
 184 #endif
 185
 186
 187 /*
 188  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 189  * header (with pri, len, nxt, hlim, src, dst).
 190  * This function may modify ver and hlim only.
 191  * The mbuf chain containing the packet will be freed.
 192  * The mbuf opt, if present, will not be freed.
 193  *
 194  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 195  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 196  * which is rt_rmx.rmx_mtu.
 197  */
 198 int
 199 ip6_output(m0, opt, ro, flags, im6o, ifpp)
 200         struct mbuf *m0;
 201         struct ip6_pktopts *opt;
 202 #ifdef NEW_STRUCT_ROUTE
 203         struct route *ro;
 204 #else
 205         struct route_in6 *ro;
 206 #endif
 207         int flags;
 208         struct ip6_moptions *im6o;
 209         struct ifnet **ifpp;            /* XXX: just for statistics */
 210 {
 211         struct ip6_hdr *ip6, *mhip6;
 212         struct ifnet *ifp, *origifp;
 213         struct mbuf *m = m0;
 214         int hlen, tlen, len, off;
 215 #ifdef NEW_STRUCT_ROUTE
 216         struct route ip6route;
 217 #else
 218         struct route_in6 ip6route;
 219 #endif
 220         struct rtentry *rt = NULL;
 221         struct sockaddr_in6 *dst;
 222         int error = 0;
 223         struct in6_ifaddr *ia = NULL;
 224         u_long mtu;
 225         u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
 226         struct ip6_exthdrs exthdrs;
 227         struct in6_addr finaldst;
 228 #ifdef NEW_STRUCT_ROUTE
 229         struct route *ro_pmtu = NULL;
 230 #else
 231         struct route_in6 *ro_pmtu = NULL;
 232 #endif
 233         int hdrsplit = 0;
 234 #ifdef __OpenBSD__
 235         u_int8_t sproto = 0;
 236 #else
 237         int needipsec = 0;
 238 #endif
 239 #if defined(__NetBSD__) && defined(PFIL_HOOKS)
 240         struct packet_filter_hook *pfh;
 241         struct mbuf *m1;
 242         int rv;
 243 #endif /* PFIL_HOOKS */
 244 #if defined(__bsdi__) && _BSDI_VERSION < 199802
 245         struct ifnet *loifp = &loif;
 246 #endif
 247 #ifdef MIP6
 248         struct mip6_pktopts mip6opt;
 249 #ifdef NEW_STRUCT_ROUTE
 250         struct route mip6_ip6route;
 251 #else
 252         struct route_in6 mip6_ip6route;
 253 #endif
 254 #endif /* MIP6 */
 255 #ifdef IPSEC
 256 #ifdef __OpenBSD__
 257         struct m_tag *mtag;
 258         union sockaddr_union sdst;
 259         struct tdb_ident *tdbi;
 260         u_int32_t sspi;
 261         struct inpcb *inp;
 262         struct tdb *tdb;
 263         int s;
 264
 265         inp = NULL;     /* XXX */
 266         if (inp && (inp->inp_flags & INP_IPV6) == 0)
 267                 panic("ip6_output: IPv4 pcb is passed");
 268 #else
 269         int needipsectun = 0;
 270         struct socket *so;
 271         struct secpolicy *sp = NULL;
 272
 273         /* for AH processing. stupid to have "socket" variable in IP layer... */
 274         so = ipsec_getsocket(m);
 275         (void)ipsec_setsocket(m, NULL);
 276         ip6 = mtod(m, struct ip6_hdr *);
 277 #endif
 278 #endif /* IPSEC */
 279
 280 #define MAKE_EXTHDR(hp, mp)                                             \
 281     do {                                                                \
 282         if (hp) {                                                       \
 283                 struct ip6_ext *eh = (struct ip6_ext *)(hp);            \
 284                 error = ip6_copyexthdr((mp), (caddr_t)(hp),             \
 285                                        ((eh)->ip6e_len + 1) << 3);      \
 286                 if (error)                                              \
 287                         goto freehdrs;                                  \
 288         }                                                               \
 289     } while (0)
 290
 291         bzero(&exthdrs, sizeof(exthdrs));
 292
 293         if (opt) {
 294                 /* Hop-by-Hop options header */
 295                 MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
 296                 /* Destination options header(1st part) */
 297                 if (opt->ip6po_rthdr) {
 298                         /*
 299                          * Destination options header(1st part)
 300                          * This only makes sence with a routing header.
 301                          * See Section 9.2 of
 302                          * draft-ietf-ipngwg-rfc2292bis-02.txt.
 303                          * Disabling this part just for MIP6 convenience is
 304                          * a bad idea.  We need to think carefully about a
 305                          * way to make the advanced API coexist with MIP6
 306                          * options, which might automatically be inserted in
 307                          * the kernel.
 308                          */
 309                         MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
 310                 }
 311                 /* Routing header */
 312                 MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
 313                 /* Destination options header(2nd part) */
 314                 MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
 315         }
 316 #ifdef MIP6
 317         bzero((caddr_t)&mip6opt, sizeof(mip6opt));
 318         if ((flags & IPV6_FORWARDING) == 0) {
 319                 /*
 320                  * XXX: reconsider the following routine.
 321                  */
 322                 /*
 323                  * MIP6 extention headers handling.
 324                  * insert HA, BU, BA, BR options if necessary.
 325                  */
 326                 if (mip6_exthdr_create(m, opt, &mip6opt))
 327                         goto freehdrs;
 328
 329                 if (((opt != NULL) && (opt->ip6po_rthdr != NULL))
 330                     || (mip6opt.mip6po_rthdr != NULL)) {
 331                         m_freem(exthdrs.ip6e_rthdr);
 332                         if (mip6opt.mip6po_rthdr != NULL) {
 333                                 /*
 334                                  * there is no rthdr specified in the
 335                                  * ip6_pktopts.  but mip6 create a
 336                                  * rthdr for the router optimization
 337                                  * purpose.
 338                                  */
 339                                 MAKE_EXTHDR(mip6opt.mip6po_rthdr,
 340                                             &exthdrs.ip6e_rthdr);
 341                         } else {
 342                                 /*
 343                                  * there is a rthdr specified in the
 344                                  * ip6_pktopts.  if mip6 require the
 345                                  * route optimization, the rthdr for
 346                                  * that purpose is already included in
 347                                  * the ip6po_rthdr in the
 348                                  * mip6_destopt_create().
 349                                  */
 350                                 MAKE_EXTHDR(opt->ip6po_rthdr,
 351                                             &exthdrs.ip6e_rthdr);
 352                         }
 353                         /*
 354                          * if a routing header exists dest1 must be
 355                          * inserted if it exists.
 356                          */
 357                         if ((opt != NULL) && (opt->ip6po_dest1)) {
 358                                 m_freem(exthdrs.ip6e_dest1);
 359                                 MAKE_EXTHDR(opt->ip6po_dest1,
 360                                             &exthdrs.ip6e_dest1);
 361                         }
 362                 }
 363                 MAKE_EXTHDR(mip6opt.mip6po_haddr, &exthdrs.ip6e_haddr);
 364                 if (mip6opt.mip6po_dest2) {
 365                         m_freem(exthdrs.ip6e_dest2);
 366                         MAKE_EXTHDR(mip6opt.mip6po_dest2, &exthdrs.ip6e_dest2);
 367                 }
 368         } else {
 369                 /*
 370                  * this is the forwarding packet.  do not modify any
 371                  * extension headers.
 372                  */
 373         }
 374 #endif /* MIP6 */
 375
 376 #ifdef IPSEC
 377 #ifdef __OpenBSD__
 378         /*
 379          * splnet is chosen over spltdb because we are not allowed to
 380          * lower the level, and udp6_output calls us in splnet(). XXX check
 381          */
 382         s = splnet();
 383
 384         /*
 385          * Check if there was an outgoing SA bound to the flow
 386          * from a transport protocol.
 387          */
 388         ip6 = mtod(m, struct ip6_hdr *);
 389
 390         /* Do we have any pending SAs to apply ? */
 391         mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
 392         if (mtag != NULL) {
 393 #ifdef DIAGNOSTIC
 394                 if (mtag->m_tag_len != sizeof (struct tdb_ident))
 395                         panic("ip6_output: tag of length %d (should be %d",
 396                             mtag->m_tag_len, sizeof (struct tdb_ident));
 397 #endif
 398                 tdbi = (struct tdb_ident *)(mtag + 1);
 399                 tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
 400                 if (tdb == NULL)
 401                         error = -EINVAL;
 402                 m_tag_delete(m, mtag);
 403         }
 404         else
 405                 tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
 406                     &error, IPSP_DIRECTION_OUT, NULL, inp);
 407
 408         if (tdb == NULL) {
 409                 splx(s);
 410
 411                 if (error == 0) {
 412                         /*
 413                          * No IPsec processing required, we'll just send the
 414                          * packet out.
 415                          */
 416                         sproto = 0;
 417
 418                         /* Fall through to routing/multicast handling */
 419                 } else {
 420                         /*
 421                          * -EINVAL is used to indicate that the packet should
 422                          * be silently dropped, typically because we've asked
 423                          * key management for an SA.
 424                          */
 425                         if (error == -EINVAL) /* Should silently drop packet */
 426                                 error = 0;
 427
 428                         goto freehdrs;
 429                 }
 430         } else {
 431                 /* Loop detection */
 432                 for (mtag = m_tag_first(m); mtag != NULL;
 433                     mtag = m_tag_next(m, mtag)) {
 434                         if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
 435                             mtag->m_tag_id !=
 436                             PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
 437                                 continue;
 438                         tdbi = (struct tdb_ident *)(mtag + 1);
 439                         if (tdbi->spi == tdb->tdb_spi &&
 440                             tdbi->proto == tdb->tdb_sproto &&
 441                             !bcmp(&tdbi->dst, &tdb->tdb_dst,
 442                             sizeof(union sockaddr_union))) {
 443                                 splx(s);
 444                                 sproto = 0; /* mark as no-IPsec-needed */
 445                                 goto done_spd;
 446                         }
 447                 }
 448
 449                 /* We need to do IPsec */
 450                 bcopy(&tdb->tdb_dst, &sdst, sizeof(sdst));
 451                 sspi = tdb->tdb_spi;
 452                 sproto = tdb->tdb_sproto;
 453                 splx(s);
 454
 455 #if 1 /* XXX */
 456                 /* if we have any extension header, we cannot perform IPsec */
 457                 if (exthdrs.ip6e_hbh || exthdrs.ip6e_dest1 ||
 458 #ifdef MIP6
 459                     exthdrs.ip6e_haddr ||
 460 #endif /* MIP6 */
 461                     exthdrs.ip6e_rthdr || exthdrs.ip6e_dest2) {
 462                         error = EHOSTUNREACH;
 463                         goto freehdrs;
 464                 }
 465 #endif
 466         }
 467
 468         /* Fall through to the routing/multicast handling code */
 469  done_spd:
 470 #else
 471         /* get a security policy for this packet */
 472         if (so == NULL)
 473                 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
 474         else
 475                 sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
 476
 477         if (sp == NULL) {
 478                 ipsec6stat.out_inval++;
 479                 goto freehdrs;
 480         }
 481
 482         error = 0;
 483
 484         /* check policy */
 485         switch (sp->policy) {
 486         case IPSEC_POLICY_DISCARD:
 487                 /*
 488                  * This packet is just discarded.
 489                  */
 490                 ipsec6stat.out_polvio++;
 491                 goto freehdrs;
 492
 493         case IPSEC_POLICY_BYPASS:
 494         case IPSEC_POLICY_NONE:
 495                 /* no need to do IPsec. */
 496                 needipsec = 0;
 497                 break;
 498
 499         case IPSEC_POLICY_IPSEC:
 500                 if (sp->req == NULL) {
 501                         /* acquire a policy */
 502                         error = key_spdacquire(sp);
 503                         goto freehdrs;
 504                 }
 505                 needipsec = 1;
 506                 break;
 507
 508         case IPSEC_POLICY_ENTRUST:
 509         default:
 510                 printf("ip6_output: Invalid policy found. %d\n", sp->policy);
 511         }
 512 #endif /* OpenBSD */
 513 #endif /* IPSEC */
 514
 515         /*
 516          * Calculate the total length of the extension header chain.
 517          * Keep the length of the unfragmentable part for fragmentation.
 518          */
 519         optlen = 0;
 520         if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
 521         if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
 522         if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
 523 #ifdef MIP6
 524         if (exthdrs.ip6e_haddr) optlen += exthdrs.ip6e_haddr->m_len;
 525 #endif /* MIP6 */
 526         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 527         /* NOTE: we don't add AH/ESP length here. do that later. */
 528         if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
 529
 530         /*
 531          * If we need IPsec, or there is at least one extension header,
 532          * separate IP6 header from the payload.
 533          */
 534 #ifdef __OpenBSD__
 535         if ((sproto || optlen) && !hdrsplit)
 536 #else
 537         if ((needipsec || optlen) && !hdrsplit)
 538 #endif
 539         {
 540                 if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 541                         m = NULL;
 542                         goto freehdrs;
 543                 }
 544                 m = exthdrs.ip6e_ip6;
 545                 hdrsplit++;
 546         }
 547
 548         /* adjust pointer */
 549         ip6 = mtod(m, struct ip6_hdr *);
 550
 551         /* adjust mbuf packet header length */
 552         m->m_pkthdr.len += optlen;
 553         plen = m->m_pkthdr.len - sizeof(*ip6);
 554
 555         /* If this is a jumbo payload, insert a jumbo payload option. */
 556         if (plen > IPV6_MAXPACKET) {
 557                 if (!hdrsplit) {
 558                         if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
 559                                 m = NULL;
 560                                 goto freehdrs;
 561                         }
 562                         m = exthdrs.ip6e_ip6;
 563                         hdrsplit++;
 564                 }
 565                 /* adjust pointer */
 566                 ip6 = mtod(m, struct ip6_hdr *);
 567                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 568                         goto freehdrs;
 569                 ip6->ip6_plen = 0;
 570         } else
 571                 ip6->ip6_plen = htons(plen);
 572
 573         /*
 574          * Concatenate headers and fill in next header fields.
 575          * Here we have, on "m"
 576          *      IPv6 payload
 577          * and we insert headers accordingly.  Finally, we should be getting:
 578          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 579          *
 580          * during the header composing process, "m" points to IPv6 header.
 581          * "mprev" points to an extension header prior to esp.
 582          */
 583         {
 584                 u_char *nexthdrp = &ip6->ip6_nxt;
 585                 struct mbuf *mprev = m;
 586
 587                 /*
 588                  * we treat dest2 specially.  this makes IPsec processing
 589                  * much easier.  the goal here is to make mprev point the
 590                  * mbuf prior to dest2.
 591                  *
 592                  * result: IPv6 dest2 payload
 593                  * m and mprev will point to IPv6 header.
 594                  */
 595                 if (exthdrs.ip6e_dest2) {
 596                         if (!hdrsplit)
 597                                 panic("assumption failed: hdr not split");
 598                         exthdrs.ip6e_dest2->m_next = m->m_next;
 599                         m->m_next = exthdrs.ip6e_dest2;
 600                         *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 601                         ip6->ip6_nxt = IPPROTO_DSTOPTS;
 602                 }
 603
 604 #define MAKE_CHAIN(m, mp, p, i)\
 605     do {\
 606         if (m) {\
 607                 if (!hdrsplit) \
 608                         panic("assumption failed: hdr not split"); \
 609                 *mtod((m), u_char *) = *(p);\
 610                 *(p) = (i);\
 611                 p = mtod((m), u_char *);\
 612                 (m)->m_next = (mp)->m_next;\
 613                 (mp)->m_next = (m);\
 614                 (mp) = (m);\
 615         }\
 616     } while (0)
 617                 /*
 618                  * result: IPv6 hbh dest1 rthdr dest2 payload
 619                  * m will point to IPv6 header.  mprev will point to the
 620                  * extension header prior to dest2 (rthdr in the above case).
 621                  */
 622                 MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
 623                            nexthdrp, IPPROTO_HOPOPTS);
 624                 MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
 625                            nexthdrp, IPPROTO_DSTOPTS);
 626                 MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
 627                            nexthdrp, IPPROTO_ROUTING);
 628 #ifdef MIP6
 629                 /*
 630                  * XXX
 631                  * MIP6 homeaddress destination option must reside
 632                  * after rthdr and before ah/esp/frag hdr.
 633                  * this order is not recommended in the ipv6 spec of course.
 634                  * result: IPv6 hbh dest1 rthdr ha dest2 payload.
 635                  */
 636                 MAKE_CHAIN(exthdrs.ip6e_haddr, mprev,
 637                            nexthdrp, IPPROTO_DSTOPTS);
 638 #endif /* MIP6 */
 639
 640 #if defined(IPSEC) && !defined(__OpenBSD__)
 641                 if (!needipsec)
 642                         goto skip_ipsec2;
 643
 644                 /*
 645                  * pointers after IPsec headers are not valid any more.
 646                  * other pointers need a great care too.
 647                  * (IPsec routines should not mangle mbufs prior to AH/ESP)
 648                  */
 649                 exthdrs.ip6e_dest2 = NULL;
 650
 651             {
 652                 struct ip6_rthdr *rh = NULL;
 653                 int segleft_org = 0;
 654                 struct ipsec_output_state state;
 655
 656                 if (exthdrs.ip6e_rthdr) {
 657                         rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 658                         segleft_org = rh->ip6r_segleft;
 659                         rh->ip6r_segleft = 0;
 660                 }
 661
 662                 bzero(&state, sizeof(state));
 663                 state.m = m;
 664                 error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
 665                         &needipsectun);
 666                 m = state.m;
 667                 if (error) {
 668                         /* mbuf is already reclaimed in ipsec6_output_trans. */
 669                         m = NULL;
 670                         switch (error) {
 671                         case EHOSTUNREACH:
 672                         case ENETUNREACH:
 673                         case EMSGSIZE:
 674                         case ENOBUFS:
 675                         case ENOMEM:
 676                                 break;
 677                         default:
 678                                 printf("ip6_output (ipsec): error code %d\n", error);
 679                                 /* fall through */
 680                         case ENOENT:
 681                                 /* don't show these error codes to the user */
 682                                 error = 0;
 683                                 break;
 684                         }
 685                         goto bad;
 686                 }
 687                 if (exthdrs.ip6e_rthdr) {
 688                         /* ah6_output doesn't modify mbuf chain */
 689                         rh->ip6r_segleft = segleft_org;
 690                 }
 691             }
 692 skip_ipsec2:;
 693 #endif
 694         }
 695
 696 #ifdef MIP6
 697         if ((flags & IPV6_FORWARDING) == 0) {
 698                 /*
 699                  * After the IPsec processing the IPv6 header source
 700                  * address (this is the homeaddress of this node) and
 701                  * the address currently stored in the Home Address
 702                  * destination option (this is the coa of this node)
 703                  * must be swapped.
 704                  */
 705                 if ((error = mip6_addr_exchange(m, exthdrs.ip6e_haddr)) != 0) {
 706                         mip6log((LOG_ERR,
 707                                  "%s:%d: "
 708                                  "addr exchange between haddr and "
 709                                  "coa failed.\n",
 710                                  __FILE__, __LINE__));
 711                         goto bad;
 712                 }
 713         } else {
 714                 /*
 715                  * this is the forwarding packet.  The typical (and
 716                  * only ?) case is multicast packet forwarding.  The
 717                  * swapping has been already done before (if
 718                  * necessary).  we must not touch any extension
 719                  * headers at all.
 720                  */
 721         }
 722 #endif /* MIP6 */
 723
 724         /*
 725          * If there is a routing header, replace destination address field
 726          * with the first hop of the routing header.
 727          */
 728         if (exthdrs.ip6e_rthdr) {
 729                 struct ip6_rthdr *rh =
 730                         (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
 731                                                   struct ip6_rthdr *));
 732                 struct ip6_rthdr0 *rh0;
 733                 struct in6_addr *addr;
 734
 735                 finaldst = ip6->ip6_dst;
 736                 switch (rh->ip6r_type) {
 737                 case IPV6_RTHDR_TYPE_0:
 738                          rh0 = (struct ip6_rthdr0 *)rh;
 739                          addr = (struct in6_addr *)(rh0 + 1);
 740
 741                          ip6->ip6_dst = *addr;
 742                          bcopy((caddr_t)(addr + 1), (caddr_t)addr,
 743                                  sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)
 744                                  );
 745                          *(addr + rh0->ip6r0_segleft - 1) = finaldst;
 746                          break;
 747                 default:        /* is it possible? */
 748                          error = EINVAL;
 749                          goto bad;
 750                 }
 751         }
 752
 753         /* Source address validation */
 754         if (!(flags & IPV6_UNSPECSRC) &&
 755             IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
 756                 /*
 757                  * XXX: we can probably assume validation in the caller, but
 758                  * we explicitly check the address here for safety.
 759                  */
 760                 error = EOPNOTSUPP;
 761                 ip6stat.ip6s_badscope++;
 762                 goto bad;
 763         }
 764         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 765                 error = EOPNOTSUPP;
 766                 ip6stat.ip6s_badscope++;
 767                 goto bad;
 768         }
 769
 770         ip6stat.ip6s_localout++;
 771
 772         /*
 773          * Route packet.
 774          */
 775         if (ro == 0) {
 776                 ro = &ip6route;
 777                 bzero((caddr_t)ro, sizeof(*ro));
 778         }
 779         ro_pmtu = ro;
 780         if (opt && opt->ip6po_rthdr)
 781                 ro = &opt->ip6po_route;
 782 #ifdef MIP6
 783         else if (exthdrs.ip6e_rthdr) {
 784                 struct sockaddr_in6 *firsthop;
 785                 struct ip6_hdr *ip6
 786                         = mtod(m, struct ip6_hdr *); /* needed ? */;
 787
 788                 ro = &mip6_ip6route;
 789                 bzero((caddr_t)ro, sizeof(*ro));
 790                 firsthop = (struct sockaddr_in6 *)&ro->ro_dst;
 791                 bzero(firsthop, sizeof(*firsthop));
 792                 firsthop->sin6_family = AF_INET6;
 793                 firsthop->sin6_len = sizeof(struct sockaddr_in6);
 794                 firsthop->sin6_addr = ip6->ip6_dst;
 795         }
 796 #endif /* MIP6 */
 797         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 798
 799 #ifdef IPSEC
 800 #ifdef __OpenBSD__
 801         /*
 802          * Check if the packet needs encapsulation.
 803          * ipsp_process_packet will never come back to here.
 804          */
 805         if (sproto != 0) {
 806                 s = splnet();
 807
 808                 /* fill in IPv6 header which would be filled later */
 809                 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 810                         if (opt && opt->ip6po_hlim != -1)
 811                                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 812                 } else {
 813                         if (im6o != NULL)
 814                                 ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 815                         else
 816                                 ip6->ip6_hlim = ip6_defmcasthlim;
 817                         if (opt && opt->ip6po_hlim != -1)
 818                                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 819
 820                         /*
 821                          * XXX what should we do if ip6_hlim == 0 and the packet
 822                          * gets tunnelled?
 823                          */
 824                 }
 825
 826                 tdb = gettdb(sspi, &sdst, sproto);
 827                 if (tdb == NULL) {
 828                         splx(s);
 829                         error = EHOSTUNREACH;
 830                         m_freem(m);
 831                         goto done;
 832                 }
 833
 834                 /* Latch to PCB */
 835                 if (inp)
 836                         tdb_add_inp(tdb, inp, 0);
 837
 838                 m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
 839
 840                 /* Callee frees mbuf */
 841                 error = ipsp_process_packet(m, tdb, AF_INET6, 0);
 842                 splx(s);
 843
 844                 return error;  /* Nothing more to be done */
 845         }
 846 #else
 847         if (needipsec && needipsectun) {
 848                 struct ipsec_output_state state;
 849
 850                 /*
 851                  * All the extension headers will become inaccessible
 852                  * (since they can be encrypted).
 853                  * Don't panic, we need no more updates to extension headers
 854                  * on inner IPv6 packet (since they are now encapsulated).
 855                  *
 856                  * IPv6 [ESP|AH] IPv6 [extension headers] payload
 857                  */
 858                 bzero(&exthdrs, sizeof(exthdrs));
 859                 exthdrs.ip6e_ip6 = m;
 860
 861                 bzero(&state, sizeof(state));
 862                 state.m = m;
 863                 state.ro = (struct route *)ro;
 864                 state.dst = (struct sockaddr *)dst;
 865
 866                 error = ipsec6_output_tunnel(&state, sp, flags);
 867
 868                 m = state.m;
 869 #ifdef NEW_STRUCT_ROUTE
 870                 ro = state.ro;
 871 #else
 872                 ro = (struct route_in6 *)state.ro;
 873 #endif
 874                 dst = (struct sockaddr_in6 *)state.dst;
 875                 if (error) {
 876                         /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 877                         m0 = m = NULL;
 878                         m = NULL;
 879                         switch (error) {
 880                         case EHOSTUNREACH:
 881                         case ENETUNREACH:
 882                         case EMSGSIZE:
 883                         case ENOBUFS:
 884                         case ENOMEM:
 885                                 break;
 886                         default:
 887                                 printf("ip6_output (ipsec): error code %d\n", error);
 888                                 /* fall through */
 889                         case ENOENT:
 890                                 /* don't show these error codes to the user */
 891                                 error = 0;
 892                                 break;
 893                         }
 894                         goto bad;
 895                 }
 896
 897                 exthdrs.ip6e_ip6 = m;
 898         }
 899 #endif /* OpenBSD */
 900 #endif /* IPSEC */
 901
 902         /* if specified, fill in the traffic class field. */
 903         if (opt) {
 904                 ip6->ip6_flow &= ~htonl(0xff << 20);
 905                 if (opt->ip6po_tclass >= 0)
 906                         ip6->ip6_flow |=
 907                             htonl((opt->ip6po_tclass & 0xff) << 20);
 908         }
 909         /* fill in or override the hop limit field, if necessary. */
 910         if (opt && opt->ip6po_hlim != -1)
 911                 ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 912         else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 913                 if (im6o != NULL)
 914                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 915                 else
 916                         ip6->ip6_hlim = ip6_defmcasthlim;
 917         }
 918
 919         {
 920                 /*
 921                  * XXX: using a block just to define a local variables is not
 922                  * a good style....
 923                  */
 924                 struct ifnet *ifp0 = NULL;
 925                 struct sockaddr_in6 src;
 926                 struct sockaddr_in6 dst0;
 927                 int clone = 0;
 928                 int64_t zone;
 929
 930                 /*
 931                  * XXX: sockaddr_in6 for the destination should be passed
 932                  * from the upper layer with a proper scope zone ID, in order
 933                  * to make a copy here.
 934                  */
 935                 bzero(&dst0, sizeof(dst0));
 936                 dst0.sin6_family = AF_INET6;
 937                 dst0.sin6_len = sizeof(dst0);
 938                 dst0.sin6_addr = ip6->ip6_dst;
 939 #ifdef SCOPEDROUTING
 940                 /* XXX: in6_recoverscope will clear the embedded ID */
 941                 error = in6_recoverscope(&dst0, &dst0.sin6_addr, NULL);
 942                 if (error != 0) {
 943                         ip6stat.ip6s_badscope++;
 944                         in6_ifstat_inc(ifp, ifs6_out_discard);
 945                         goto bad;
 946                 }
 947 #endif
 948
 949 #if defined(__bsdi__) || defined(__FreeBSD__)
 950                 if (ro != &ip6route && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
 951                         clone = 1;
 952 #endif
 953
 954                 if ((error = in6_selectroute(&dst0, opt, im6o, ro,
 955                                              &ifp, &rt, clone)) != 0) {
 956                         switch (error) {
 957                         case EHOSTUNREACH:
 958                                 ip6stat.ip6s_noroute++;
 959                                 break;
 960                         case EADDRNOTAVAIL:
 961                         default:
 962                                 break; /* XXX statistics? */
 963                         }
 964                         if (ifp != NULL)
 965                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 966                         goto bad;
 967                 }
 968                 if (rt == NULL) {
 969                         /*
 970                          * If in6_selectroute() does not return a route entry,
 971                          * dst may not have been updated.
 972                          */
 973                         *dst = dst0;    /* XXX */
 974                 }
 975
 976                 /*
 977                  * then rt (for unicast) and ifp must be non-NULL valid values.
 978                  */
 979                 if ((flags & IPV6_FORWARDING) == 0) {
 980                         /* XXX: the FORWARDING flag can be set for mrouting. */
 981                         in6_ifstat_inc(ifp, ifs6_out_request);
 982                 }
 983                 if (rt != NULL) {
 984                         ia = (struct in6_ifaddr *)(rt->rt_ifa);
 985                         rt->rt_use++;
 986                 }
 987
 988                 /*
 989                  * The outgoing interface must be in the zone of source and
 990                  * destination addresses.  We should use ia_ifp to support the
 991                  * case of sending packets to an address of our own.
 992                  */
 993                 if (ia != NULL && ia->ia_ifp)
 994                         ifp0 = ia->ia_ifp;
 995                 else
 996                         ifp0 = ifp;
 997                 /* XXX: we should not do this conversion for every packet. */
 998                 bzero(&src, sizeof(src));
 999                 src.sin6_family = AF_INET6;
1000                 src.sin6_len = sizeof(src);
1001                 src.sin6_addr = ip6->ip6_src;
1002                 if ((error = in6_recoverscope(&src, &ip6->ip6_src, NULL))
1003                     != 0) {
1004                         goto badscope;
1005                 }
1006                 if ((zone = in6_addr2zoneid(ifp0, &src.sin6_addr)) < 0 ||
1007                     zone != src.sin6_scope_id) {
1008 #ifdef SCOPEDEBUG               /* will be removed shortly */
1009                         printf("ip6 output: bad source scope %s for %s on %s\n",
1010                                ip6_sprintf(&ip6->ip6_src),
1011                                ip6_sprintf(&ip6->ip6_dst), if_name(ifp0));
1012 #endif
1013                         goto badscope;
1014                 }
1015                 /* XXX: in6_recoverscope will clear the embedded ID */
1016                 if ((error = in6_recoverscope(&dst0, &ip6->ip6_dst, NULL))
1017                     != 0) {
1018                         goto badscope;
1019                 }
1020                 if ((zone = in6_addr2zoneid(ifp0, &dst0.sin6_addr)) < 0 ||
1021                     zone != dst0.sin6_scope_id) {
1022 #ifdef SCOPEDEBUG               /* will be removed shortly */
1023                         printf("ip6 output: bad dst scope %s on %s\n",
1024                                ip6_sprintf(&dst0.sin6_addr), if_name(ifp0));
1025 #endif
1026                         goto badscope;
1027                 }
1028
1029                 /* scope check is done. */
1030                 goto routefound;
1031
1032           badscope:
1033                 ip6stat.ip6s_badscope++;
1034                 in6_ifstat_inc(ifp0, ifs6_out_discard);
1035                 if (error == 0)
1036                         error = EHOSTUNREACH; /* XXX */
1037                 goto bad;
1038         }
1039
1040   routefound:
1041         if (rt) {
1042                 if (opt && opt->ip6po_nextroute.ro_rt) {
1043                         /*
1044                          * The nexthop is explicitly specified by the
1045                          * application.  We assume the next hop is an IPv6
1046                          * address.
1047                          */
1048                         dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
1049                 }
1050                 else if ((rt->rt_flags & RTF_GATEWAY))
1051                         dst = (struct sockaddr_in6 *)rt->rt_gateway;
1052         }
1053
1054         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1055                 m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
1056         } else {
1057                 struct  in6_multi *in6m;
1058
1059                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
1060
1061                 in6_ifstat_inc(ifp, ifs6_out_mcast);
1062
1063                 /*
1064                  * Confirm that the outgoing interface supports multicast.
1065                  */
1066                 if (!(ifp->if_flags & IFF_MULTICAST)) {
1067                         ip6stat.ip6s_noroute++;
1068                         in6_ifstat_inc(ifp, ifs6_out_discard);
1069                         error = ENETUNREACH;
1070                         goto bad;
1071                 }
1072                 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1073                 if (in6m != NULL &&
1074                    (im6o == NULL || im6o->im6o_multicast_loop)) {
1075                         /*
1076                          * If we belong to the destination multicast group
1077                          * on the outgoing interface, and the caller did not
1078                          * forbid loopback, loop back a copy.
1079                          */
1080                         ip6_mloopback(ifp, m, dst);
1081                 } else {
1082                         /*
1083                          * If we are acting as a multicast router, perform
1084                          * multicast forwarding as if the packet had just
1085                          * arrived on the interface to which we are about
1086                          * to send.  The multicast forwarding function
1087                          * recursively calls this function, using the
1088                          * IPV6_FORWARDING flag to prevent infinite recursion.
1089                          *
1090                          * Multicasts that are looped back by ip6_mloopback(),
1091                          * above, will be forwarded by the ip6_input() routine,
1092                          * if necessary.
1093                          */
1094                         if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
1095                                 /*
1096                                  * XXX: ip6_mforward expects that rcvif is NULL
1097                                  * when it is called from the originating path.
1098                                  * However, it is not always the case, since
1099                                  * some versions of MGETHDR() does not
1100                                  * initialize the field.
1101                                  */
1102                                 m->m_pkthdr.rcvif = NULL;
1103                                 if (ip6_mforward(ip6, ifp, m) != 0) {
1104                                         m_freem(m);
1105                                         goto done;
1106                                 }
1107                         }
1108                 }
1109                 /*
1110                  * Multicasts with a hoplimit of zero may be looped back,
1111                  * above, but must not be transmitted on a network.
1112                  * Also, multicasts addressed to the loopback interface
1113                  * are not sent -- the above call to ip6_mloopback() will
1114                  * loop back a copy if this host actually belongs to the
1115                  * destination group on the loopback interface.
1116                  */
1117                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
1118                     IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
1119                         m_freem(m);
1120                         goto done;
1121                 }
1122         }
1123
1124         /*
1125          * Fill the outgoing inteface to tell the upper layer
1126          * to increment per-interface statistics.
1127          */
1128         if (ifpp)
1129                 *ifpp = ifp;
1130
1131         /*
1132          * Upper-layer reachability confirmation
1133          */
1134         if (opt && (opt->ip6po_flags & IP6PO_REACHCONF))
1135                 nd6_nud_hint(rt, NULL, 0);
1136
1137         /* Determine path MTU. */
1138         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu)) != 0)
1139                 goto bad;
1140
1141         /*
1142          * An advanced API option (IPV6_USE_MIN_MTU) overrides mtu setting.
1143          * We ignore the specified MTU if it is larger than the already-known
1144          * path MTU.
1145          */
1146         if (mtu > IPV6_MMTU && opt && (opt->ip6po_flags & IP6PO_MINMTU))
1147                 mtu = IPV6_MMTU;
1148
1149         /* Fake scoped addresses */
1150         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
1151                 /*
1152                  * If source or destination address is a scoped address, and
1153                  * the packet is going to be sent to a loopback interface,
1154                  * we should keep the original interface.
1155                  */
1156
1157                 /*
1158                  * XXX: this is a very experimental and temporary solution.
1159                  * We eventually have sockaddr_in6 and use the sin6_scope_id
1160                  * field of the structure here.
1161                  * We rely on the consistency between two scope zone ids
1162                  * of source and destination, which should already be assured.
1163                  * Larger scopes than link will be supported in the future.
1164                  */
1165                 origifp = NULL;
1166                 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
1167 #if defined(__FreeBSD__) && __FreeBSD__ >= 5
1168                         origifp = ifnet_byindex(ntohs(ip6->ip6_src.s6_addr16[1]));
1169 #else
1170                         origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
1171 #endif
1172                 } else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
1173 #if defined(__FreeBSD__) && __FreeBSD__ >= 5
1174                         origifp = ifnet_byindex(ntohs(ip6->ip6_dst.s6_addr16[1]));
1175 #else
1176                         origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
1177 #endif
1178                 } else if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
1179 #if defined(__FreeBSD__) && __FreeBSD__ >= 5
1180                         origifp = ifnet_byindex(ntohs(ip6->ip6_dst.s6_addr16[1]));
1181 #else
1182                         origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
1183 #endif
1184                 }
1185
1186                 /*
1187                  * XXX: origifp can be NULL even in those two cases above.
1188                  * For example, if we remove the (only) link-local address
1189                  * from the loopback interface, and try to send a link-local
1190                  * address without link-id information.  Then the source
1191                  * address is ::1, and the destination address is the
1192                  * link-local address with its s6_addr16[1] being zero.
1193                  * What is worse, if the packet goes to the loopback interface
1194                  * by a default rejected route, the null pointer would be
1195                  * passed to looutput, and the kernel would hang.
1196                  * The following last resort would prevent such disaster.
1197                  */
1198                 if (origifp == NULL)
1199                         origifp = ifp;
1200         }
1201         else
1202                 origifp = ifp;
1203 #ifndef SCOPEDROUTING
1204         /*
1205          * clear embedded scope identifiers if necessary.
1206          * in6_clearscope will touch the addresses only when necessary.
1207          */
1208         in6_clearscope(&ip6->ip6_src);
1209         in6_clearscope(&ip6->ip6_dst);
1210 #endif
1211
1212 #if defined(IPV6FIREWALL) || (defined(__FreeBSD__) && __FreeBSD__ >= 4)
1213         /*
1214          * Check with the firewall...
1215          */
1216 #if defined(__FreeBSD__) && __FreeBSD__ >= 4
1217         if (ip6_fw_enable && ip6_fw_chk_ptr) {
1218 #else
1219         if (ip6_fw_chk_ptr) {
1220 #endif
1221                 m->m_pkthdr.rcvif = NULL;       /* XXX */
1222                 /* If ipfw says divert, we have to just drop packet */
1223                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &m)) {
1224                         m_freem(m);
1225                         goto done;
1226                 }
1227                 if (!m) {
1228                         error = EACCES;
1229                         goto done;
1230                 }
1231         }
1232 #endif
1233
1234         /*
1235          * If the outgoing packet contains a hop-by-hop options header,
1236          * it must be examined and processed even by the source node.
1237          * (RFC 2460, section 4.)
1238          */
1239         if (exthdrs.ip6e_hbh) {
1240                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
1241                 u_int32_t dummy; /* XXX unused */
1242                 u_int32_t plen = 0; /* XXX: ip6_process will check the value */
1243
1244 #ifdef DIAGNOSTIC
1245                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
1246                         panic("ip6e_hbh is not continuous");
1247 #endif
1248                 /*
1249                  *  XXX: if we have to send an ICMPv6 error to the sender,
1250                  *       we need the M_LOOP flag since icmp6_error() expects
1251                  *       the IPv6 and the hop-by-hop options header are
1252                  *       continuous unless the flag is set.
1253                  */
1254                 m->m_flags |= M_LOOP;
1255                 m->m_pkthdr.rcvif = ifp;
1256                 if (ip6_process_hopopts(m,
1257                                         (u_int8_t *)(hbh + 1),
1258                                         ((hbh->ip6h_len + 1) << 3) -
1259                                         sizeof(struct ip6_hbh),
1260                                         &dummy, &plen) < 0) {
1261                         /* m was already freed at this point */
1262                         error = EINVAL;/* better error? */
1263                         goto done;
1264                 }
1265                 m->m_flags &= ~M_LOOP; /* XXX */
1266                 m->m_pkthdr.rcvif = NULL;
1267         }
1268
1269 #if defined(__NetBSD__) && defined(PFIL_HOOKS)
1270         /*
1271          * Run through list of hooks for output packets.
1272          */
1273         m1 = m;
1274         pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IPV6]].pr_pfh);
1275         for (; pfh; pfh = pfh->pfil_link.tqe_next)
1276                 if (pfh->pfil_func) {
1277                         rv = pfh->pfil_func(ip6, sizeof(*ip6), ifp, 1, &m1);
1278                         if (rv) {
1279                                 error = EHOSTUNREACH;
1280                                 goto done;
1281                         }
1282                         m = m1;
1283                         if (m == NULL)
1284                                 goto done;
1285                         ip6 = mtod(m, struct ip6_hdr *);
1286                 }
1287 #endif /* PFIL_HOOKS */
1288
1289 #if defined(__OpenBSD__) && NPF > 0
1290         if (pf_test6(PF_OUT, ifp, &m) != PF_PASS) {
1291                 error = EHOSTUNREACH;
1292                 m_freem(m);
1293                 goto done;
1294         }
1295         ip6 = mtod(m, struct ip6_hdr *);
1296 #endif
1297
1298         /*
1299          * Send the packet to the outgoing interface.
1300          * If necessary, do IPv6 fragmentation before sending.
1301          */
1302         tlen = m->m_pkthdr.len;
1303         /*
1304          * Even if the DONTFRAG option is specified, we cannot send the packet
1305          * when the data length is larger than the MTU of the outgoing
1306          * interface.
1307          * Notify the error by sending IPV6_PATHMTU ancillary data as well
1308          * as returning an error code (the latter is not described in the API
1309          * spec.)
1310          */
1311         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG) && tlen > ifp->if_mtu
1312 #ifdef notyet
1313             && !(ifp->if_flags & IFF_FRAGMENTABLE)
1314 #endif
1315                 ) {
1316                 u_int32_t mtu32;
1317                 struct ip6ctlparam ip6cp;
1318
1319                 mtu32 = (u_int32_t)mtu;
1320                 bzero(&ip6cp, sizeof(ip6cp));
1321                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1322                 pfctlinput2(PRC_MSGSIZE, &ro_pmtu->ro_dst, (void *)&ip6cp);
1323
1324                 error = EMSGSIZE;
1325                 goto bad;
1326         }
1327         if (tlen <= mtu || (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
1328 #ifdef notyet
1329             /*
1330              * On any link that cannot convey a 1280-octet packet in one piece,
1331              * link-specific fragmentation and reassembly must be provided at
1332              * a layer below IPv6. [RFC 2460, sec.5]
1333              * Thus if the interface has ability of link-level fragmentation,
1334              * we can just send the packet even if the packet size is
1335              * larger than the link's MTU.
1336              * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
1337              */
1338
1339             || (ifp->if_flags & IFF_FRAGMENTABLE)
1340 #endif
1341             )
1342         {
1343                 struct in6_ifaddr *ia6;
1344
1345                 ip6 = mtod(m, struct ip6_hdr *);
1346                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1347                 if (ia6) {
1348                         /* Record statistics for this interface address. */
1349 #if defined(__NetBSD__) && defined(IFA_STATS)
1350                         ia6->ia_ifa.ifa_data.ifad_outbytes +=
1351                                 m->m_pkthdr.len;
1352 #elif defined(__FreeBSD__) && __FreeBSD__ >= 4
1353                         ia6->ia_ifa.if_opackets++;
1354                         ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1355 #elif defined(__bsdi__) && _BSDI_VERSION >= 199802
1356                         ia6->ia_ifa.ifa_opackets++;
1357                         ia6->ia_ifa.ifa_obytes += m->m_pkthdr.len;
1358 #endif
1359                 }
1360 #if defined(IPSEC) && !defined(__OpenBSD__)
1361                 /* clean ipsec history once it goes out of the node */
1362                 ipsec_delaux(m);
1363 #endif
1364                 error = nd6_output(ifp, origifp, m, dst, rt);
1365                 goto done;
1366         } else if (mtu < IPV6_MMTU) {
1367                 /*
1368                  * note that path MTU is never less than IPV6_MMTU
1369                  * (see icmp6_input).
1370                  */
1371                 error = EMSGSIZE;
1372                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1373                 goto bad;
1374         } else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
1375                 error = EMSGSIZE;
1376                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
1377                 goto bad;
1378         } else {
1379                 struct mbuf **mnext, *m_frgpart;
1380                 struct ip6_frag *ip6f;
1381                 u_int32_t id = htonl(ip6_id++);
1382                 u_char nextproto;
1383                 struct ip6ctlparam ip6cp;
1384                 u_int32_t mtu32;
1385
1386                 /*
1387                  * Too large for the destination or interface;
1388                  * fragment if possible.
1389                  * Must be able to put at least 8 bytes per fragment.
1390                  */
1391                 hlen = unfragpartlen;
1392                 if (mtu > IPV6_MAXPACKET)
1393                         mtu = IPV6_MAXPACKET;
1394
1395                 /* Notify a proper path MTU to applications. */
1396                 mtu32 = (u_int32_t)mtu;
1397                 bzero(&ip6cp, sizeof(ip6cp));
1398                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
1399                 pfctlinput2(PRC_MSGSIZE, &ro_pmtu->ro_dst, (void *)&ip6cp);
1400
1401                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1402                 if (len < 8) {
1403                         error = EMSGSIZE;
1404                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
1405                         goto bad;
1406                 }
1407
1408                 mnext = &m->m_nextpkt;
1409
1410                 /*
1411                  * Change the next header field of the last header in the
1412                  * unfragmentable part.
1413                  */
1414 #ifdef MIP6
1415                 if (exthdrs.ip6e_haddr) {
1416                         nextproto = *mtod(exthdrs.ip6e_haddr, u_char *);
1417                         *mtod(exthdrs.ip6e_haddr, u_char *) = IPPROTO_FRAGMENT;
1418                 } else
1419 #endif /* MIP6 */
1420                 if (exthdrs.ip6e_rthdr) {
1421                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1422                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1423                 } else if (exthdrs.ip6e_dest1) {
1424                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1425                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1426                 } else if (exthdrs.ip6e_hbh) {
1427                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1428                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1429                 } else {
1430                         nextproto = ip6->ip6_nxt;
1431                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
1432                 }
1433
1434                 /*
1435                  * Loop through length of segment after first fragment,
1436                  * make new header and copy data of each part and link onto
1437                  * chain.
1438                  */
1439                 m0 = m;
1440                 for (off = hlen; off < tlen; off += len) {
1441                         MGETHDR(m, M_DONTWAIT, MT_HEADER);
1442                         if (!m) {
1443                                 error = ENOBUFS;
1444                                 ip6stat.ip6s_odropped++;
1445                                 goto sendorfree;
1446                         }
1447                         m->m_pkthdr.rcvif = NULL;
1448                         m->m_flags = m0->m_flags & M_COPYFLAGS;
1449                         *mnext = m;
1450                         mnext = &m->m_nextpkt;
1451                         m->m_data += max_linkhdr;
1452                         mhip6 = mtod(m, struct ip6_hdr *);
1453                         *mhip6 = *ip6;
1454                         m->m_len = sizeof(*mhip6);
1455                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1456                         if (error) {
1457                                 ip6stat.ip6s_odropped++;
1458                                 goto sendorfree;
1459                         }
1460                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1461                         if (off + len >= tlen)
1462                                 len = tlen - off;
1463                         else
1464                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1465                         mhip6->ip6_plen = htons((u_short)(len + hlen +
1466                                                           sizeof(*ip6f) -
1467                                                           sizeof(struct ip6_hdr)));
1468                         if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1469                                 error = ENOBUFS;
1470                                 ip6stat.ip6s_odropped++;
1471                                 goto sendorfree;
1472                         }
1473                         m_cat(m, m_frgpart);
1474                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1475                         m->m_pkthdr.rcvif = (struct ifnet *)0;
1476                         ip6f->ip6f_reserved = 0;
1477                         ip6f->ip6f_ident = id;
1478                         ip6f->ip6f_nxt = nextproto;
1479                         ip6stat.ip6s_ofragments++;
1480                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1481                 }
1482
1483                 in6_ifstat_inc(ifp, ifs6_out_fragok);
1484         }
1485
1486         /*
1487          * Remove leading garbages.
1488          */
1489 sendorfree:
1490         m = m0->m_nextpkt;
1491         m0->m_nextpkt = 0;
1492         m_freem(m0);
1493         for (m0 = m; m; m = m0) {
1494                 m0 = m->m_nextpkt;
1495                 m->m_nextpkt = 0;
1496                 if (error == 0) {
1497                         struct in6_ifaddr *ia6;
1498                         ip6 = mtod(m, struct ip6_hdr *);
1499                         ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1500                         if (ia6) {
1501                                 /*
1502                                  * Record statistics for this interface
1503                                  * address.
1504                                  */
1505 #if defined(__NetBSD__) && defined(IFA_STATS)
1506                                 ia6->ia_ifa.ifa_data.ifad_outbytes +=
1507                                         m->m_pkthdr.len;
1508 #elif defined(__FreeBSD__) && __FreeBSD__ >= 4
1509                                 ia6->ia_ifa.if_opackets++;
1510                                 ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1511 #elif defined(__bsdi__) && _BSDI_VERSION >= 199802
1512                                 ia6->ia_ifa.ifa_opackets++;
1513                                 ia6->ia_ifa.ifa_obytes += m->m_pkthdr.len;
1514 #endif
1515                         }
1516 #if defined(IPSEC) && !defined(__OpenBSD__)
1517                         /* clean ipsec history once it goes out of the node */
1518                         ipsec_delaux(m);
1519 #endif
1520                         error = nd6_output(ifp, origifp, m, dst, rt);
1521                 } else
1522                         m_freem(m);
1523         }
1524
1525         if (error == 0)
1526                 ip6stat.ip6s_fragmented++;
1527
1528 done:
1529         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1530                 RTFREE(ro->ro_rt);
1531         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1532                 RTFREE(ro_pmtu->ro_rt);
1533         }
1534
1535 #if defined(IPSEC) && !defined(__OpenBSD__)
1536         if (sp != NULL)
1537                 key_freesp(sp);
1538 #endif /* IPSEC */
1539
1540 #ifdef MIP6
1541         mip6_destopt_discard(&mip6opt);
1542 #endif /* MIP6 */
1543
1544         return(error);
1545
1546 freehdrs:
1547 #ifdef MIP6
1548         mip6_destopt_discard(&mip6opt);
1549 #endif /* MIP6 */
1550         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
1551         m_freem(exthdrs.ip6e_dest1);
1552         m_freem(exthdrs.ip6e_rthdr);
1553 #ifdef MIP6
1554         m_freem(exthdrs.ip6e_haddr);
1555 #endif /* MIP6 */
1556         m_freem(exthdrs.ip6e_dest2);
1557         /* fall through */
1558 bad:
1559         m_freem(m);
1560         goto done;
1561 }
1562
1563 static int
1564 ip6_copyexthdr(mp, hdr, hlen)
1565         struct mbuf **mp;
1566         caddr_t hdr;
1567         int hlen;
1568 {
1569         struct mbuf *m;
1570
1571         if (hlen > MCLBYTES)
1572                 return(ENOBUFS); /* XXX */
1573
1574         MGET(m, M_DONTWAIT, MT_DATA);
1575         if (!m)
1576                 return(ENOBUFS);
1577
1578         if (hlen > MLEN) {
1579                 MCLGET(m, M_DONTWAIT);
1580                 if ((m->m_flags & M_EXT) == 0) {
1581                         m_free(m);
1582                         return(ENOBUFS);
1583                 }
1584         }
1585         m->m_len = hlen;
1586         if (hdr)
1587                 bcopy(hdr, mtod(m, caddr_t), hlen);
1588
1589         *mp = m;
1590         return(0);
1591 }
1592
1593 /*
1594  * Insert jumbo payload option.
1595  */
1596 static int
1597 ip6_insert_jumboopt(exthdrs, plen)
1598         struct ip6_exthdrs *exthdrs;
1599         u_int32_t plen;
1600 {
1601         struct mbuf *mopt;
1602         u_char *optbuf;
1603         u_int32_t v;
1604
1605 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
1606
1607         /*
1608          * If there is no hop-by-hop options header, allocate new one.
1609          * If there is one but it doesn't have enough space to store the
1610          * jumbo payload option, allocate a cluster to store the whole options.
1611          * Otherwise, use it to store the options.
1612          */
1613         if (exthdrs->ip6e_hbh == 0) {
1614                 MGET(mopt, M_DONTWAIT, MT_DATA);
1615                 if (mopt == 0)
1616                         return(ENOBUFS);
1617                 mopt->m_len = JUMBOOPTLEN;
1618                 optbuf = mtod(mopt, u_char *);
1619                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
1620                 exthdrs->ip6e_hbh = mopt;
1621         } else {
1622                 struct ip6_hbh *hbh;
1623
1624                 mopt = exthdrs->ip6e_hbh;
1625                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1626                         /*
1627                          * XXX assumption:
1628                          * - exthdrs->ip6e_hbh is not referenced from places
1629                          *   other than exthdrs.
1630                          * - exthdrs->ip6e_hbh is not an mbuf chain.
1631                          */
1632                         int oldoptlen = mopt->m_len;
1633                         struct mbuf *n;
1634
1635                         /*
1636                          * XXX: give up if the whole (new) hbh header does
1637                          * not fit even in an mbuf cluster.
1638                          */
1639                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1640                                 return(ENOBUFS);
1641
1642                         /*
1643                          * As a consequence, we must always prepare a cluster
1644                          * at this point.
1645                          */
1646                         MGET(n, M_DONTWAIT, MT_DATA);
1647                         if (n) {
1648                                 MCLGET(n, M_DONTWAIT);
1649                                 if ((n->m_flags & M_EXT) == 0) {
1650                                         m_freem(n);
1651                                         n = NULL;
1652                                 }
1653                         }
1654                         if (!n)
1655                                 return(ENOBUFS);
1656                         n->m_len = oldoptlen + JUMBOOPTLEN;
1657                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1658                               oldoptlen);
1659                         optbuf = mtod(n, caddr_t) + oldoptlen;
1660                         m_freem(mopt);
1661                         mopt = exthdrs->ip6e_hbh = n;
1662                 } else {
1663                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
1664                         mopt->m_len += JUMBOOPTLEN;
1665                 }
1666                 optbuf[0] = IP6OPT_PADN;
1667                 optbuf[1] = 1;
1668
1669                 /*
1670                  * Adjust the header length according to the pad and
1671                  * the jumbo payload option.
1672                  */
1673                 hbh = mtod(mopt, struct ip6_hbh *);
1674                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1675         }
1676
1677         /* fill in the option. */
1678         optbuf[2] = IP6OPT_JUMBO;
1679         optbuf[3] = 4;
1680         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1681         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1682
1683         /* finally, adjust the packet header length */
1684         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1685
1686         return(0);
1687 #undef JUMBOOPTLEN
1688 }
1689
1690 /*
1691  * Insert fragment header and copy unfragmentable header portions.
1692  */
1693 static int
1694 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1695         struct mbuf *m0, *m;
1696         int hlen;
1697         struct ip6_frag **frghdrp;
1698 {
1699         struct mbuf *n, *mlast;
1700
1701         if (hlen > sizeof(struct ip6_hdr)) {
1702                 n = m_copym(m0, sizeof(struct ip6_hdr),
1703                             hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1704                 if (n == 0)
1705                         return(ENOBUFS);
1706                 m->m_next = n;
1707         } else
1708                 n = m;
1709
1710         /* Search for the last mbuf of unfragmentable part. */
1711         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1712                 ;
1713
1714         if ((mlast->m_flags & M_EXT) == 0 &&
1715             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1716                 /* use the trailing space of the last mbuf for the fragment hdr */
1717                 *frghdrp =
1718                         (struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1719                 mlast->m_len += sizeof(struct ip6_frag);
1720                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1721         } else {
1722                 /* allocate a new mbuf for the fragment header */
1723                 struct mbuf *mfrg;
1724
1725                 MGET(mfrg, M_DONTWAIT, MT_DATA);
1726                 if (mfrg == 0)
1727                         return(ENOBUFS);
1728                 mfrg->m_len = sizeof(struct ip6_frag);
1729                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1730                 mlast->m_next = mfrg;
1731         }
1732
1733         return(0);
1734 }
1735
1736 static int
1737 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup)
1738 #ifdef NEW_STRUCT_ROUTE
1739         struct route *ro_pmtu, *ro;
1740 #else
1741         struct route_in6 *ro_pmtu, *ro;
1742 #endif
1743         struct ifnet *ifp;
1744         struct in6_addr *dst;   /* XXX: should be sockaddr_in6 */
1745         u_long *mtup;
1746 {
1747         u_int32_t mtu = 0;
1748         int error = 0;
1749
1750         if (ro_pmtu != ro) {
1751                 /* The first hop and the final destination may differ. */
1752                 struct sockaddr_in6 *sa6_dst =
1753                         (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1754                 if (ro_pmtu->ro_rt && ((ro_pmtu->ro_rt->rt_flags & RTF_UP)
1755                                        == 0 ||
1756                                        !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr,
1757                                                            dst))) {
1758                         RTFREE(ro_pmtu->ro_rt);
1759                         ro_pmtu->ro_rt = (struct rtentry *)NULL;
1760                 }
1761                 if (ro_pmtu->ro_rt == NULL) {
1762                         bzero(sa6_dst, sizeof(*sa6_dst));
1763                         sa6_dst->sin6_family = AF_INET6;
1764                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1765                         sa6_dst->sin6_addr = *dst;
1766
1767 #ifdef __bsdi__                 /* bsdi needs rtcalloc to clone a route. */
1768                         rtcalloc((struct route *)ro_pmtu);
1769 #else
1770                         rtalloc((struct route *)ro_pmtu);
1771 #endif
1772                 }
1773         }
1774         if (ro_pmtu->ro_rt) {
1775                 u_int32_t ifmtu;
1776
1777                 if (ifp == NULL)
1778                         ifp = ro_pmtu->ro_rt->rt_ifp;
1779                 ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
1780                 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1781                 if (mtu > ifmtu || mtu == 0) {
1782                         /*
1783                          * The MTU on the route is larger than the MTU on
1784                          * the interface!  This shouldn't happen, unless the
1785                          * MTU of the interface has been changed after the
1786                          * interface was brought up.  Change the MTU in the
1787                          * route to match the interface MTU (as long as the
1788                          * field isn't locked).
1789                          *
1790                          * if MTU on the route is 0, we need to fix the MTU.
1791                          * this case happens with path MTU discovery timeouts.
1792                          */
1793                          mtu = ifmtu;
1794                          if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
1795                                  ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
1796                 }
1797         } else if (ifp) {
1798                 mtu = nd_ifinfo[ifp->if_index].linkmtu;
1799         } else
1800                 error = EHOSTUNREACH; /* XXX */
1801
1802         *mtup = mtu;
1803         return(error);
1804 }
1805
1806 /*
1807  * IP6 socket option processing.
1808  */
1809 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
1810 int
1811 ip6_ctloutput(so, sopt)
1812         struct socket *so;
1813         struct sockopt *sopt;
1814 #else
1815 int
1816 ip6_ctloutput(op, so, level, optname, mp)
1817         int op;
1818         struct socket *so;
1819         int level, optname;
1820         struct mbuf **mp;
1821 #endif
1822 {
1823         int privileged, optdatalen;
1824         void *optdata;
1825         struct ip6_recvpktopts *rcvopts;
1826 #if defined(IPSEC) && defined(__OpenBSD__)
1827         struct proc *p = curproc; /* XXX */
1828         struct tdb *tdb;
1829         struct tdb_ident *tdbip, tdbi;
1830         int s;
1831 #endif
1832 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
1833         struct inpcb *in6p = sotoinpcb(so);
1834         int error, optval;
1835         int level, op, optname;
1836         int optlen;
1837         struct proc *p;
1838
1839         if (!sopt) {
1840                 panic("ip6_ctloutput: arg soopt is NULL");
1841         }
1842         level = sopt->sopt_level;
1843         op = sopt->sopt_dir;
1844         optname = sopt->sopt_name;
1845         optlen = sopt->sopt_valsize;
1846         p = sopt->sopt_p;
1847 #else
1848 #ifdef HAVE_NRL_INPCB
1849         struct inpcb *inp = sotoinpcb(so);
1850 #define in6p inp
1851 #else  /* !NRL */
1852         struct in6pcb *in6p = sotoin6pcb(so);
1853 #endif /* HAVE_NRL_INPCB */
1854         struct mbuf *m = *mp;
1855         int error, optval;
1856         int optlen;
1857 #if defined(__NetBSD__) || (defined(__FreeBSD__) && __FreeBSD__ >= 3)
1858         struct proc *p = curproc;       /* XXX */
1859 #endif
1860
1861         optlen = m ? m->m_len : 0;
1862 #endif /* FreeBSD >= 3 */
1863         error = optval = 0;
1864         privileged = 1;
1865
1866         rcvopts = in6p->in6p_inputopts;
1867
1868         if (level == IPPROTO_IPV6) {
1869                 switch (op) {
1870
1871 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
1872                 case SOPT_SET:
1873 #else
1874                 case PRCO_SETOPT:
1875 #endif
1876                         switch (optname) {
1877                         case IPV6_2292PKTOPTIONS:
1878 #ifdef IPV6_PKTOPTIONS
1879                         case IPV6_PKTOPTIONS:
1880 #endif
1881                         {
1882 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
1883                                 struct mbuf *m;
1884
1885                                 error = soopt_getm(sopt, &m); /* XXX */
1886                                 if (error)
1887                                         break;
1888                                 error = soopt_mcopyin(sopt, m); /* XXX */
1889                                 if (error)
1890                                         break;
1891                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1892                                                     m, so, sopt);
1893                                 m_freem(m); /* XXX */
1894 #else
1895                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1896                                                     m, so);
1897 #endif /* FreeBSD >= 3 */
1898                                 break;
1899                         }
1900
1901                         /*
1902                          * Use of some Hop-by-Hop options or some
1903                          * Destination options, might require special
1904                          * privilege.  That is, normal applications
1905                          * (without special privilege) might be forbidden
1906                          * from setting certain options in outgoing packets,
1907                          * and might never see certain options in received
1908                          * packets. [RFC 2292 Section 6]
1909                          * KAME specific note:
1910                          *  KAME prevents non-privileged users from sending or
1911                          *  receiving ANY hbh/dst options in order to avoid
1912                          *  overhead of parsing options in the kernel.
1913                          */
1914                         case IPV6_RECVHOPOPTS:
1915                         case IPV6_RECVDSTOPTS:
1916                         case IPV6_RECVRTHDRDSTOPTS:
1917                                 /* fall through */
1918                         case IPV6_UNICAST_HOPS:
1919                         case IPV6_HOPLIMIT:
1920                         case IPV6_FAITH:
1921
1922                         case IPV6_RECVPKTINFO:
1923                         case IPV6_RECVHOPLIMIT:
1924                         case IPV6_RECVRTHDR:
1925                         case IPV6_RECVPATHMTU:
1926                         case IPV6_RECVTCLASS:
1927                         case IPV6_V6ONLY:
1928                         case IPV6_AUTOFLOWLABEL:
1929                                 if (optlen != sizeof(int)) {
1930                                         error = EINVAL;
1931                                         break;
1932                                 }
1933 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
1934                                 error = sooptcopyin(sopt, &optval,
1935                                         sizeof optval, sizeof optval);
1936                                 if (error)
1937                                         break;
1938 #else
1939                                 optval = *mtod(m, int *);
1940 #endif
1941                                 switch (optname) {
1942
1943                                 case IPV6_UNICAST_HOPS:
1944                                         if (optval < -1 || optval >= 256)
1945                                                 error = EINVAL;
1946                                         else {
1947                                                 /* -1 = kernel default */
1948                                                 in6p->in6p_hops = optval;
1949 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
1950                                                 if ((in6p->in6p_vflag &
1951                                                      INP_IPV4) != 0)
1952                                                         in6p->inp_ip_ttl = optval;
1953 #endif
1954                                         }
1955                                         break;
1956 #define OPTSET(bit) \
1957 do { \
1958         if (optval) \
1959                 in6p->in6p_flags |= (bit); \
1960         else \
1961                 in6p->in6p_flags &= ~(bit); \
1962 } while (0)
1963 #define OPTSET2292(bit) \
1964 do { \
1965         in6p->in6p_flags |= IN6P_RFC2292; \
1966         if (optval) \
1967                 in6p->in6p_flags |= (bit); \
1968         else \
1969                 in6p->in6p_flags &= ~(bit); \
1970 } while (0)
1971 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1972
1973                                 case IPV6_RECVPKTINFO:
1974                                         /* cannot mix with RFC2292 */
1975                                         if (OPTBIT(IN6P_RFC2292)) {
1976                                                 error = EINVAL;
1977                                                 break;
1978                                         }
1979                                         OPTSET(IN6P_PKTINFO);
1980                                         if (OPTBIT(IN6P_PKTINFO) == 0)
1981                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVPKTINFO);
1982                                         break;
1983
1984                                 case IPV6_HOPLIMIT:
1985                                 {
1986                                         struct ip6_pktopts **optp;
1987
1988                                         /* cannot mix with RFC2292 */
1989                                         if (OPTBIT(IN6P_RFC2292)) {
1990                                                 error = EINVAL;
1991                                                 break;
1992                                         }
1993                                         optp = &in6p->in6p_outputopts;
1994                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1995                                                            (u_char *)&optval,
1996                                                            sizeof(optval),
1997                                                            optp,
1998                                                            privileged);
1999                                         break;
2000                                 }
2001
2002                                 case IPV6_RECVHOPLIMIT:
2003                                         /* cannot mix with RFC2292 */
2004                                         if (OPTBIT(IN6P_RFC2292)) {
2005                                                 error = EINVAL;
2006                                                 break;
2007                                         }
2008                                         OPTSET(IN6P_HOPLIMIT);
2009                                         if (OPTBIT(IN6P_HOPLIMIT) == 0)
2010                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVHOPLIMIT);
2011                                         break;
2012
2013                                 case IPV6_RECVHOPOPTS:
2014                                         /* cannot mix with RFC2292 */
2015                                         if (OPTBIT(IN6P_RFC2292)) {
2016                                                 error = EINVAL;
2017                                                 break;
2018                                         }
2019                                         OPTSET(IN6P_HOPOPTS);
2020                                         if (OPTBIT(IN6P_HOPOPTS) == 0)
2021                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVHOPOPTS);
2022                                         break;
2023
2024                                 case IPV6_RECVDSTOPTS:
2025                                         /* cannot mix with RFC2292 */
2026                                         if (OPTBIT(IN6P_RFC2292)) {
2027                                                 error = EINVAL;
2028                                                 break;
2029                                         }
2030                                         OPTSET(IN6P_DSTOPTS);
2031                                         if (OPTBIT(IN6P_DSTOPTS) == 0)
2032                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVDSTOPTS);
2033                                         break;
2034
2035                                 case IPV6_RECVRTHDRDSTOPTS:
2036                                         /* cannot mix with RFC2292 */
2037                                         if (OPTBIT(IN6P_RFC2292)) {
2038                                                 error = EINVAL;
2039                                                 break;
2040                                         }
2041                                         OPTSET(IN6P_RTHDRDSTOPTS);
2042                                         if (OPTBIT(IN6P_RTHDRDSTOPTS) == 0)
2043                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVRTHDRDSTOPTS);
2044                                         break;
2045
2046                                 case IPV6_RECVRTHDR:
2047                                         /* cannot mix with RFC2292 */
2048                                         if (OPTBIT(IN6P_RFC2292)) {
2049                                                 error = EINVAL;
2050                                                 break;
2051                                         }
2052                                         OPTSET(IN6P_RTHDR);
2053                                         if (OPTBIT(IN6P_RTHDR) == 0)
2054                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVRTHDR);
2055                                         break;
2056
2057                                 case IPV6_FAITH:
2058                                         OPTSET(IN6P_FAITH);
2059                                         break;
2060
2061                                 case IPV6_RECVPATHMTU:
2062                                         OPTSET(IN6P_MTU);
2063                                         break;
2064
2065                                 case IPV6_V6ONLY:
2066                                         /*
2067                                          * make setsockopt(IPV6_V6ONLY)
2068                                          * available only prior to bind(2).
2069                                          * see ipng mailing list, Jun 22 2001.
2070                                          */
2071                                         if (in6p->in6p_lport ||
2072                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
2073                                         {
2074                                                 error = EINVAL;
2075                                                 break;
2076                                         }
2077 #ifdef __NetBSD__
2078 #ifdef INET6_BINDV6ONLY
2079                                         if (!optval)
2080                                                 error = EINVAL;
2081 #else
2082                                         OPTSET(IN6P_IPV6_V6ONLY);
2083 #endif
2084 #elif (defined(__FreeBSD__) && __FreeBSD__ >= 3) || (defined(__bsdi__) && _BSDI_VERSION >= 199802)
2085                                         OPTSET(IN6P_IPV6_V6ONLY);
2086 #else
2087                                         if ((ip6_v6only && optval) ||
2088                                             (!ip6_v6only && !optval))
2089                                                 error = 0;
2090                                         else
2091                                                 error = EINVAL;
2092 #endif
2093                                         break;
2094                                 case IPV6_RECVTCLASS:
2095                                         /* cannot mix with RFC2292 XXX */
2096                                         if (OPTBIT(IN6P_RFC2292)) {
2097                                                 error = EINVAL;
2098                                                 break;
2099                                         }
2100                                         OPTSET(IN6P_TCLASS);
2101                                         break;
2102                                 case IPV6_AUTOFLOWLABEL:
2103                                         OPTSET(IN6P_AUTOFLOWLABEL);
2104                                         break;
2105
2106                                 }
2107                                 break;
2108
2109                         case IPV6_OTCLASS:
2110                         {
2111                                 struct ip6_pktopts **optp;
2112                                 u_int8_t tclass;
2113
2114                                 if (optlen != sizeof(tclass)) {
2115                                         error = EINVAL;
2116                                         break;
2117                                 }
2118 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2119                                 error = sooptcopyin(sopt, &tclass,
2120                                         sizeof tclass, sizeof tclass);
2121                                 if (error)
2122                                         break;
2123 #else
2124                                 tclass = *mtod(m, u_int8_t *);
2125 #endif
2126                                 optp = &in6p->in6p_outputopts;
2127                                 error = ip6_pcbopt(optname,
2128                                                    (u_char *)&tclass,
2129                                                    sizeof(tclass),
2130                                                    optp,
2131                                                    privileged);
2132                                 break;
2133                         }
2134
2135                         case IPV6_TCLASS:
2136                         case IPV6_DONTFRAG:
2137                         case IPV6_USE_MIN_MTU:
2138                                 if (optlen != sizeof(optval)) {
2139                                         error = EINVAL;
2140                                         break;
2141                                 }
2142 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2143                                 error = sooptcopyin(sopt, &optval,
2144                                         sizeof optval, sizeof optval);
2145                                 if (error)
2146                                         break;
2147 #else
2148                                 optval = *mtod(m, int *);
2149 #endif
2150                                 {
2151                                         struct ip6_pktopts **optp;
2152                                         optp = &in6p->in6p_outputopts;
2153                                         error = ip6_pcbopt(optname,
2154                                                            (u_char *)&optval,
2155                                                            sizeof(optval),
2156                                                            optp,
2157                                                            privileged);
2158                                         break;
2159                                 }
2160
2161                         case IPV6_2292PKTINFO:
2162                         case IPV6_2292HOPLIMIT:
2163                         case IPV6_2292HOPOPTS:
2164                         case IPV6_2292DSTOPTS:
2165                         case IPV6_2292RTHDR:
2166                                 /* RFC 2292 */
2167                                 if (optlen != sizeof(int)) {
2168                                         error = EINVAL;
2169                                         break;
2170                                 }
2171 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2172                                 error = sooptcopyin(sopt, &optval,
2173                                         sizeof optval, sizeof optval);
2174                                 if (error)
2175                                         break;
2176 #else
2177                                 optval = *mtod(m, int *);
2178 #endif
2179                                 switch (optname) {
2180                                 case IPV6_2292PKTINFO:
2181                                         OPTSET2292(IN6P_PKTINFO);
2182                                         if (OPTBIT(IN6P_PKTINFO) == 0)
2183                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVPKTINFO);
2184                                         break;
2185                                 case IPV6_2292HOPLIMIT:
2186                                         OPTSET2292(IN6P_HOPLIMIT);
2187                                         if (OPTBIT(IN6P_HOPLIMIT) == 0)
2188                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVHOPLIMIT);
2189                                         break;
2190                                 case IPV6_2292HOPOPTS:
2191                                         /*
2192                                          * Check super-user privilege.
2193                                          * See comments for IPV6_RECVHOPOPTS.
2194                                          */
2195                                         OPTSET2292(IN6P_HOPOPTS);
2196                                         if (OPTBIT(IN6P_HOPOPTS) == 0)
2197                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVHOPOPTS);
2198                                         break;
2199                                 case IPV6_2292DSTOPTS:
2200                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
2201                                         if (OPTBIT(IN6P_DSTOPTS) == 0) {
2202                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVDSTOPTS);
2203                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVRTHDRDSTOPTS);
2204                                         }
2205                                         break;
2206                                 case IPV6_2292RTHDR:
2207                                         OPTSET2292(IN6P_RTHDR);
2208                                         if (OPTBIT(IN6P_RTHDR) == 0)
2209                                                 ip6_reset_rcvopt(rcvopts, IPV6_RECVRTHDR);
2210                                         break;
2211                                 }
2212                                 break;
2213                         case IPV6_PKTINFO:
2214                         case IPV6_HOPOPTS:
2215                         case IPV6_RTHDR:
2216                         case IPV6_DSTOPTS:
2217                         case IPV6_RTHDRDSTOPTS:
2218                         case IPV6_NEXTHOP:
2219                         {
2220                                 /* new advanced API (2292bis) */
2221                                 u_char *optbuf;
2222                                 int optlen;
2223                                 struct ip6_pktopts **optp;
2224
2225                                 /* cannot mix with RFC2292 */
2226                                 if (OPTBIT(IN6P_RFC2292)) {
2227                                         error = EINVAL;
2228                                         break;
2229                                 }
2230
2231 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2232                                 optbuf = sopt->sopt_val;
2233                                 optlen = sopt->sopt_valsize;
2234 #else  /* !fbsd3 */
2235                                 if (m && m->m_next) {
2236                                         error = EINVAL; /* XXX */
2237                                         break;
2238                                 }
2239                                 if (m) {
2240                                         optbuf = mtod(m, u_char *);
2241                                         optlen = m->m_len;
2242                                 } else {
2243                                         optbuf = NULL;
2244                                         optlen = 0;
2245                                 }
2246 #endif
2247                                 optp = &in6p->in6p_outputopts;
2248                                 error = ip6_pcbopt(optname,
2249                                                    optbuf, optlen,
2250                                                    optp, privileged);
2251                                 break;
2252                         }
2253 #undef OPTSET
2254
2255                         case IPV6_MULTICAST_IF:
2256                         case IPV6_MULTICAST_HOPS:
2257                         case IPV6_MULTICAST_LOOP:
2258                         case IPV6_JOIN_GROUP:
2259                         case IPV6_LEAVE_GROUP:
2260 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2261                             {
2262                                 struct mbuf *m;
2263                                 if (sopt->sopt_valsize > MLEN) {
2264                                         error = EMSGSIZE;
2265                                         break;
2266                                 }
2267                                 /* XXX */
2268                                 MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_HEADER);
2269                                 if (m == 0) {
2270                                         error = ENOBUFS;
2271                                         break;
2272                                 }
2273                                 m->m_len = sopt->sopt_valsize;
2274                                 error = sooptcopyin(sopt, mtod(m, char *),
2275                                                     m->m_len, m->m_len);
2276                                 error = ip6_setmoptions(sopt->sopt_name,
2277                                                         &in6p->in6p_moptions,
2278                                                         m);
2279                                 (void)m_free(m);
2280                             }
2281 #else
2282                                 error = ip6_setmoptions(optname,
2283                                                         &in6p->in6p_moptions,
2284                                                         m);
2285 #if defined(__bsdi__) && _BSDI_VERSION >= 199802
2286                                 if (in6p->in6p_moptions != NULL)
2287                                         in6p->in6p_flags |= INP_IPV6_MCAST; /* XXX */
2288 #endif
2289 #endif
2290                                 break;
2291
2292 #ifndef __bsdi__
2293                         case IPV6_PORTRANGE:
2294 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2295                                 error = sooptcopyin(sopt, &optval,
2296                                     sizeof optval, sizeof optval);
2297                                 if (error)
2298                                         break;
2299 #else
2300                                 optval = *mtod(m, int *);
2301 #endif
2302
2303                                 switch (optval) {
2304                                 case IPV6_PORTRANGE_DEFAULT:
2305                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
2306                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
2307                                         break;
2308
2309                                 case IPV6_PORTRANGE_HIGH:
2310                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
2311                                         in6p->in6p_flags |= IN6P_HIGHPORT;
2312                                         break;
2313
2314                                 case IPV6_PORTRANGE_LOW:
2315                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
2316                                         in6p->in6p_flags |= IN6P_LOWPORT;
2317                                         break;
2318
2319                                 default:
2320                                         error = EINVAL;
2321                                         break;
2322                                 }
2323                                 break;
2324 #endif
2325
2326 #ifdef __OpenBSD__
2327                         case IPSEC6_OUTSA:
2328 #ifndef IPSEC
2329                                 error = EINVAL;
2330 #else
2331                                 s = spltdb();
2332                                 if (m == 0 || m->m_len != sizeof(struct tdb_ident)) {
2333                                         error = EINVAL;
2334                                 } else {
2335                                         tdbip = mtod(m, struct tdb_ident *);
2336                                         tdb = gettdb(tdbip->spi, &tdbip->dst,
2337                                             tdbip->proto);
2338                                         if (tdb == NULL)
2339                                                 error = ESRCH;
2340                                         else
2341                                                 tdb_add_inp(tdb, inp, 0);
2342                                 }
2343                                 splx(s);
2344 #endif
2345                                 break;
2346
2347                         case IPV6_AUTH_LEVEL:
2348                         case IPV6_ESP_TRANS_LEVEL:
2349                         case IPV6_ESP_NETWORK_LEVEL:
2350                         case IPV6_IPCOMP_LEVEL:
2351 #ifndef IPSEC
2352                                 error = EINVAL;
2353 #else
2354                                 if (m == 0 || m->m_len != sizeof(int)) {
2355                                         error = EINVAL;
2356                                         break;
2357                                 }
2358                                 optval = *mtod(m, int *);
2359
2360                                 if (optval < IPSEC_LEVEL_BYPASS ||
2361                                     optval > IPSEC_LEVEL_UNIQUE) {
2362                                         error = EINVAL;
2363                                         break;
2364                                 }
2365
2366                                 switch (optname) {
2367                                 case IPV6_AUTH_LEVEL:
2368                                         inp->inp_seclevel[SL_AUTH] = optval;
2369                                         break;
2370
2371                                 case IPV6_ESP_TRANS_LEVEL:
2372                                         inp->inp_seclevel[SL_ESP_TRANS] = optval;
2373                                         break;
2374
2375                                 case IPV6_ESP_NETWORK_LEVEL:
2376                                         inp->inp_seclevel[SL_ESP_NETWORK] = optval;
2377                                         break;
2378
2379                                 case IPV6_IPCOMP_LEVEL:
2380                                         inp->inp_seclevel[SL_IPCOMP] = optval;
2381                                         break;
2382                                 }
2383                                 if (!error)
2384                                         inp->inp_secrequire = get_sa_require(inp);
2385 #endif
2386                                 break;
2387 #endif /* OpenBSD */
2388
2389 #if defined(IPSEC) && !defined(__OpenBSD__)
2390                         case IPV6_IPSEC_POLICY:
2391                             {
2392                                 caddr_t req = NULL;
2393                                 size_t len = 0;
2394 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2395                                 struct mbuf *m;
2396 #endif
2397
2398 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2399                                 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
2400                                         break;
2401                                 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
2402                                         break;
2403 #endif
2404                                 if (m) {
2405                                         req = mtod(m, caddr_t);
2406                                         len = m->m_len;
2407                                 }
2408                                 error = ipsec6_set_policy(in6p, optname, req,
2409                                                           len, privileged);
2410 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2411                                 m_freem(m);
2412 #endif
2413                             }
2414                                 break;
2415 #endif /* KAME IPSEC */
2416
2417 #if defined(IPV6FIREWALL) || (defined(__FreeBSD__) && __FreeBSD__ >= 4)
2418                         case IPV6_FW_ADD:
2419                         case IPV6_FW_DEL:
2420                         case IPV6_FW_FLUSH:
2421                         case IPV6_FW_ZERO:
2422                             {
2423 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2424                                 struct mbuf *m;
2425                                 struct mbuf **mp = &m;
2426 #endif
2427
2428 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2429                                 if (ip6_fw_ctl_ptr == NULL)
2430                                         return EINVAL;
2431                                 /* XXX */
2432                                 if ((error = soopt_getm(sopt, &m)) != 0)
2433                                         break;
2434                                 /* XXX */
2435                                 if ((error = soopt_mcopyin(sopt, m)) != 0)
2436                                         break;
2437 #else
2438                                 if (ip6_fw_ctl_ptr == NULL) {
2439                                         if (m) (void)m_free(m);
2440                                         return EINVAL;
2441                                 }
2442 #endif
2443                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
2444                                 m = *mp;
2445                             }
2446                                 break;
2447 #endif
2448
2449                         default:
2450                                 error = ENOPROTOOPT;
2451                                 break;
2452                         }
2453 #if !(defined(__FreeBSD__) && __FreeBSD__ >= 3)
2454                         if (m)
2455                                 (void)m_free(m);
2456 #endif
2457                         break;
2458
2459 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2460                 case SOPT_GET:
2461 #else
2462                 case PRCO_GETOPT:
2463 #endif
2464                         switch (optname) {
2465
2466                         case IPV6_2292PKTOPTIONS:
2467 #ifdef IPV6_PKTOPTIONS
2468                         case IPV6_PKTOPTIONS:
2469 #endif
2470 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2471                                 if (in6p->in6p_inputopts &&
2472                                     in6p->in6p_inputopts->head) {
2473                                         struct mbuf *m;
2474                                         m = m_copym(in6p->in6p_inputopts->head,
2475                                             0, M_COPYALL, M_WAIT);
2476                                         error = soopt_mcopyout(sopt, m);
2477                                         if (error == 0)
2478                                                 m_freem(m);
2479                                 } else
2480                                         sopt->sopt_valsize = 0;
2481 #else
2482                                 if (in6p->in6p_inputopts &&
2483                                     in6p->in6p_inputopts->head) {
2484                                         *mp = m_copym(in6p->in6p_inputopts->head,
2485                                                       0, M_COPYALL, M_WAIT);
2486                                 } else {
2487                                         *mp = m_get(M_WAIT, MT_SOOPTS);
2488                                         (*mp)->m_len = 0;
2489                                 }
2490 #endif
2491                                 break;
2492
2493                         case IPV6_RECVHOPOPTS:
2494                         case IPV6_RECVDSTOPTS:
2495                         case IPV6_RECVRTHDRDSTOPTS:
2496                         case IPV6_UNICAST_HOPS:
2497                         case IPV6_RECVPKTINFO:
2498                         case IPV6_RECVHOPLIMIT:
2499                         case IPV6_RECVRTHDR:
2500                         case IPV6_USE_MIN_MTU:
2501                         case IPV6_RECVPATHMTU:
2502                         case IPV6_DONTFRAG:
2503
2504                         case IPV6_FAITH:
2505                         case IPV6_V6ONLY:
2506 #ifndef __bsdi__
2507                         case IPV6_PORTRANGE:
2508 #endif
2509                         case IPV6_RECVTCLASS:
2510                         case IPV6_AUTOFLOWLABEL:
2511                                 switch (optname) {
2512
2513                                 case IPV6_UNICAST_HOPS:
2514                                         optval = in6p->in6p_hops;
2515                                         break;
2516
2517                                 case IPV6_RECVPKTINFO:
2518                                         optval = OPTBIT(IN6P_PKTINFO);
2519                                         break;
2520
2521                                 case IPV6_RECVHOPLIMIT:
2522                                         optval = OPTBIT(IN6P_HOPLIMIT);
2523                                         break;
2524
2525                                 case IPV6_RECVHOPOPTS:
2526                                         optval = OPTBIT(IN6P_HOPOPTS);
2527                                         break;
2528
2529                                 case IPV6_RECVDSTOPTS:
2530                                         optval = OPTBIT(IN6P_DSTOPTS);
2531                                         break;
2532
2533                                 case IPV6_RECVRTHDRDSTOPTS:
2534                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2535                                         break;
2536
2537                                 case IPV6_RECVPATHMTU:
2538                                         optval = OPTBIT(IN6P_MTU);
2539                                         break;
2540
2541                                 case IPV6_FAITH:
2542                                         optval = OPTBIT(IN6P_FAITH);
2543                                         break;
2544
2545                                 case IPV6_V6ONLY:
2546 #if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || defined(__NetBSD__) || (defined(__bsdi__) && _BSDI_VERSION >= 199802)
2547                                         optval = OPTBIT(IN6P_IPV6_V6ONLY);
2548 #else
2549                                         optval = (ip6_v6only != 0); /* XXX */
2550 #endif
2551                                         break;
2552
2553 #ifndef __bsdi__
2554                                 case IPV6_PORTRANGE:
2555                                     {
2556                                         int flags;
2557                                         flags = in6p->in6p_flags;
2558                                         if (flags & IN6P_HIGHPORT)
2559                                                 optval = IPV6_PORTRANGE_HIGH;
2560                                         else if (flags & IN6P_LOWPORT)
2561                                                 optval = IPV6_PORTRANGE_LOW;
2562                                         else
2563                                                 optval = 0;
2564                                         break;
2565                                     }
2566 #endif
2567                                 case IPV6_RECVTCLASS:
2568                                         optval = OPTBIT(IN6P_TCLASS);
2569                                         break;
2570
2571                                 case IPV6_AUTOFLOWLABEL:
2572                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2573                                         break;
2574
2575 #define PKTOPTBIT(bit) ((in6p->in6p_outputopts && \
2576                          (in6p->in6p_outputopts->ip6po_flags & (bit))) ? 1 : 0)
2577                                 case IPV6_DONTFRAG:
2578                                         optval = PKTOPTBIT(IP6PO_DONTFRAG);
2579                                         break;
2580
2581                                 case IPV6_USE_MIN_MTU:
2582                                         optval = PKTOPTBIT(IP6PO_MINMTU);
2583                                         break;
2584 #undef PKTOPTBIT
2585                                 }
2586                                 if (error)
2587                                         break;
2588 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2589                                 error = sooptcopyout(sopt, &optval,
2590                                         sizeof optval);
2591 #else
2592                                 *mp = m = m_get(M_WAIT, MT_SOOPTS);
2593                                 m->m_len = sizeof(int);
2594                                 *mtod(m, int *) = optval;
2595 #endif
2596                                 break;
2597
2598                         case IPV6_PATHMTU:
2599                         {
2600                                 u_long pmtu = 0;
2601                                 struct ip6_mtuinfo mtuinfo;
2602 #ifdef NEW_STRUCT_ROUTE
2603                                 struct route *ro = &in6p->in6p_route;
2604 #else
2605                                 struct route_in6 *ro = (struct route_in6 *)&in6p->in6p_route;
2606 #endif
2607
2608                                 if (!(so->so_state & SS_ISCONNECTED))
2609                                         return(ENOTCONN);
2610                                 /*
2611                                  * XXX: we dot not consider the case of source
2612                                  * routing, nor optional information to specify
2613                                  * the outgoing interface.
2614                                  */
2615                                 error = ip6_getpmtu(ro, NULL, NULL,
2616                                                     &in6p->in6p_faddr, &pmtu);
2617                                 if (error)
2618                                         break;
2619                                 if (pmtu > IPV6_MAXPACKET)
2620                                         pmtu = IPV6_MAXPACKET;
2621
2622                                 bzero(&mtuinfo, sizeof(mtuinfo));
2623                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2624                                 optdata = (void *)&mtuinfo;
2625                                 optdatalen = sizeof(mtuinfo);
2626 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2627                                 error = sooptcopyout(sopt, optdata,
2628                                                      optdatalen);
2629 #else  /* !FreeBSD3 */
2630                                         if (optdatalen > MCLBYTES)
2631                                                 return(EMSGSIZE); /* XXX */
2632                                         *mp = m = m_get(M_WAIT, MT_SOOPTS);
2633                                         if (optdatalen > MLEN)
2634                                                 MCLGET(m, M_WAIT);
2635                                         m->m_len = optdatalen;
2636                                         bcopy(optdata, mtod(m, void *),
2637                                               optdatalen);
2638 #endif /* FreeBSD3 */
2639                                 break;
2640                         }
2641
2642                         case IPV6_2292PKTINFO:
2643                         case IPV6_2292HOPLIMIT:
2644                         case IPV6_2292HOPOPTS:
2645                         case IPV6_2292RTHDR:
2646                         case IPV6_2292DSTOPTS:
2647                                 if (optname == IPV6_2292HOPOPTS ||
2648                                     optname == IPV6_2292DSTOPTS ||
2649                                     !privileged)
2650                                         return(EPERM);
2651                                 switch (optname) {
2652                                 case IPV6_2292PKTINFO:
2653                                         optval = OPTBIT(IN6P_PKTINFO);
2654                                         break;
2655                                 case IPV6_2292HOPLIMIT:
2656                                         optval = OPTBIT(IN6P_HOPLIMIT);
2657                                         break;
2658                                 case IPV6_2292HOPOPTS:
2659                                         optval = OPTBIT(IN6P_HOPOPTS);
2660                                         break;
2661                                 case IPV6_2292RTHDR:
2662                                         optval = OPTBIT(IN6P_RTHDR);
2663                                         break;
2664                                 case IPV6_2292DSTOPTS:
2665                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2666                                         break;
2667                                 }
2668 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2669                                 error = sooptcopyout(sopt, &optval,
2670                                         sizeof optval);
2671 #else
2672                                 *mp = m = m_get(M_WAIT, MT_SOOPTS);
2673                                 m->m_len = sizeof(int);
2674                                 *mtod(m, int *) = optval;
2675 #endif /* FreeBSD3 */
2676                                 break;
2677                         case IPV6_PKTINFO:
2678                         case IPV6_HOPOPTS:
2679                         case IPV6_RTHDR:
2680                         case IPV6_DSTOPTS:
2681                         case IPV6_RTHDRDSTOPTS:
2682                         case IPV6_NEXTHOP:
2683                         case IPV6_OTCLASS:
2684                         case IPV6_TCLASS:
2685 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2686                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2687                                                       optname, sopt);
2688 #else
2689                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
2690                                                       optname, mp);
2691 #endif
2692                                 break;
2693
2694                         case IPV6_MULTICAST_IF:
2695                         case IPV6_MULTICAST_HOPS:
2696                         case IPV6_MULTICAST_LOOP:
2697                         case IPV6_JOIN_GROUP:
2698                         case IPV6_LEAVE_GROUP:
2699 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2700                             {
2701                                 struct mbuf *m;
2702                                 error = ip6_getmoptions(sopt->sopt_name,
2703                                                 in6p->in6p_moptions, &m);
2704                                 if (error == 0)
2705                                         error = sooptcopyout(sopt,
2706                                                 mtod(m, char *), m->m_len);
2707                                 m_freem(m);
2708                             }
2709 #else
2710                                 error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
2711 #endif
2712                                 break;
2713
2714 #ifdef __OpenBSD__
2715                         case IPSEC6_OUTSA:
2716 #ifndef IPSEC
2717                                 error = EINVAL;
2718 #else
2719                                 s = spltdb();
2720                                 if (inp->inp_tdb_out == NULL) {
2721                                         error = ENOENT;
2722                                 } else {
2723                                         tdbi.spi = inp->inp_tdb_out->tdb_spi;
2724                                         tdbi.dst = inp->inp_tdb_out->tdb_dst;
2725                                         tdbi.proto = inp->inp_tdb_out->tdb_sproto;
2726                                         *mp = m = m_get(M_WAIT, MT_SOOPTS);
2727                                         m->m_len = sizeof(tdbi);
2728                                         bcopy((caddr_t)&tdbi, mtod(m, caddr_t),
2729                                             (unsigned)m->m_len);
2730                                 }
2731                                 splx(s);
2732 #endif
2733                                 break;
2734
2735                         case IPV6_AUTH_LEVEL:
2736                         case IPV6_ESP_TRANS_LEVEL:
2737                         case IPV6_ESP_NETWORK_LEVEL:
2738                         case IPV6_IPCOMP_LEVEL:
2739 #ifndef IPSEC
2740                                 m->m_len = sizeof(int);
2741                                 *mtod(m, int *) = IPSEC_LEVEL_NONE;
2742 #else
2743                                 m->m_len = sizeof(int);
2744                                 switch (optname) {
2745                                 case IPV6_AUTH_LEVEL:
2746                                         optval = inp->inp_seclevel[SL_AUTH];
2747                                         break;
2748
2749                                 case IPV6_ESP_TRANS_LEVEL:
2750                                         optval =
2751                                             inp->inp_seclevel[SL_ESP_TRANS];
2752                                         break;
2753
2754                                 case IPV6_ESP_NETWORK_LEVEL:
2755                                         optval =
2756                                             inp->inp_seclevel[SL_ESP_NETWORK];
2757                                         break;
2758
2759                                 case IPV6_IPCOMP_LEVEL:
2760                                         optval = inp->inp_seclevel[SL_IPCOMP];
2761                                         break;
2762                                 }
2763                                 *mtod(m, int *) = optval;
2764 #endif
2765                                 break;
2766 #endif /* OpenBSD */
2767
2768 #if defined(IPSEC) && !defined(__OpenBSD__)
2769                         case IPV6_IPSEC_POLICY:
2770                           {
2771                                 caddr_t req = NULL;
2772                                 size_t len = 0;
2773 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2774                                 struct mbuf *m = NULL;
2775                                 struct mbuf **mp = &m;
2776
2777                                 error = soopt_getm(sopt, &m); /* XXX */
2778                                 if (error != 0)
2779                                         break;
2780                                 error = soopt_mcopyin(sopt, m); /* XXX */
2781                                 if (error != 0)
2782                                         break;
2783 #endif
2784                                 if (m) {
2785                                         req = mtod(m, caddr_t);
2786                                         len = m->m_len;
2787                                 }
2788                                 error = ipsec6_get_policy(in6p, req, len, mp);
2789 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2790                                 if (error == 0)
2791                                         error = soopt_mcopyout(sopt, m); /* XXX */
2792                                 if (error == 0 && m)
2793                                         m_freem(m);
2794 #endif
2795                                 break;
2796                           }
2797 #endif /* KAME IPSEC */
2798
2799 #if defined(IPV6FIREWALL) || (defined(__FreeBSD__) && __FreeBSD__ >= 4)
2800                         case IPV6_FW_GET:
2801                           {
2802 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2803                                 struct mbuf *m;
2804                                 struct mbuf **mp = &m;
2805 #endif
2806
2807                                 if (ip6_fw_ctl_ptr == NULL)
2808                                 {
2809 #if !(defined(__FreeBSD__) && __FreeBSD__ >= 3)
2810                                         if (m)
2811                                                 (void)m_free(m);
2812 #endif
2813                                         return EINVAL;
2814                                 }
2815                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
2816 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2817                                 if (error == 0)
2818                                         error = soopt_mcopyout(sopt, m); /* XXX */
2819                                 if (error == 0 && m)
2820                                         m_freem(m);
2821 #endif
2822                           }
2823                                 break;
2824 #endif
2825
2826                         default:
2827                                 error = ENOPROTOOPT;
2828                                 break;
2829                         }
2830                         break;
2831                 }
2832         } else {                /* level != IPPROTO_IPV6 */
2833                 error = EINVAL;
2834 #if !(defined(__FreeBSD__) && __FreeBSD__ >= 3)
2835                 if (op == PRCO_SETOPT && *mp)
2836                         (void)m_free(*mp);
2837 #endif
2838         }
2839         return(error);
2840 }
2841
2842 #ifndef offsetof
2843 #define offsetof(type, member)  ((size_t)(&((type *)0)->member)) /* XXX */
2844 #endif
2845 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2846 int
2847 ip6_raw_ctloutput(so, sopt)
2848         struct socket *so;
2849         struct sockopt *sopt;
2850 #else
2851 int
2852 ip6_raw_ctloutput(op, so, level, optname, mp)
2853         int op;
2854         struct socket *so;
2855         int level, optname;
2856         struct mbuf **mp;
2857 #endif
2858 {
2859         int error = 0, optval, optlen;
2860         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2861 #ifdef HAVE_NRL_INPCB
2862         struct inpcb *inp = sotoinpcb(so);
2863 #else
2864         struct in6pcb *in6p = sotoin6pcb(so);
2865 #endif
2866 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2867         int level, op, optname;
2868         struct proc *p;
2869 #else
2870         struct mbuf *m = *mp;
2871 #endif /* FreeBSD >= 3 */
2872
2873 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2874         if (!sopt) {
2875                 panic("ip6_ctloutput: arg soopt is NULL");
2876         }
2877         level = sopt->sopt_level;
2878         op = sopt->sopt_dir;
2879         optname = sopt->sopt_name;
2880         optlen = sopt->sopt_valsize;
2881         p = sopt->sopt_p;
2882 #else
2883         optlen = m ? m->m_len : 0;
2884 #endif /* FreeBSD >= 3 */
2885
2886         if (level != IPPROTO_IPV6) {
2887 #if !(defined(__FreeBSD__) && __FreeBSD__ >= 3)
2888                 if (op == PRCO_SETOPT && *mp)
2889                         (void)m_free(*mp);
2890 #endif
2891                 return(EINVAL);
2892         }
2893
2894         switch (optname) {
2895         case IPV6_CHECKSUM:
2896                 /*
2897                  * For ICMPv6 sockets, no modification allowed for checksum
2898                  * offset, permit "no change" values to help existing apps.
2899                  *
2900                  * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM
2901                  * for an ICMPv6 socket will fail."
2902                  * The current behavior does not meet 2292bis.
2903                  */
2904                 switch (op) {
2905 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2906                 case SOPT_SET:
2907 #else
2908                 case PRCO_SETOPT:
2909 #endif
2910                         if (optlen != sizeof(int)) {
2911                                 error = EINVAL;
2912                                 break;
2913                         }
2914 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2915                         error = sooptcopyin(sopt, &optval, sizeof(optval),
2916                                             sizeof(optval));
2917                         if (error)
2918                                 break;
2919 #else
2920                         optval = *mtod(m, int *);
2921 #endif
2922                         if ((optval % 2) != 0) {
2923                                 /* the API assumes even offset values */
2924                                 error = EINVAL;
2925                         } else if (so->so_proto->pr_protocol ==
2926                                    IPPROTO_ICMPV6) {
2927                                 if (optval != icmp6off)
2928                                         error = EINVAL;
2929                         } else
2930                                 in6p->in6p_cksum = optval;
2931                         break;
2932
2933 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2934                 case SOPT_GET:
2935 #else
2936                 case PRCO_GETOPT:
2937 #endif
2938                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2939                                 optval = icmp6off;
2940                         else
2941                                 optval = in6p->in6p_cksum;
2942
2943 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2944                         error = sooptcopyout(sopt, &optval, sizeof(optval));
2945 #else
2946                         *mp = m = m_get(M_WAIT, MT_SOOPTS);
2947                         m->m_len = sizeof(int);
2948                         *mtod(m, int *) = optval;
2949 #endif /* FreeBSD3 */
2950                         break;
2951                 default:
2952                         error = EINVAL;
2953                         break;
2954                 }
2955
2956         default:
2957                 error = ENOPROTOOPT;
2958                 break;
2959         }
2960
2961 #if !(defined(__FreeBSD__) && __FreeBSD__ >= 3)
2962         if (op == PRCO_SETOPT && m)
2963                 (void)m_free(m);
2964 #endif
2965
2966         return(error);
2967 }
2968
2969 #ifdef HAVE_NRL_INPCB
2970 #undef in6p
2971 #endif
2972
2973 /*
2974  * Set up IP6 options in pcb for insertion in output packets or
2975  * specifying behavior of outgoing packets.
2976  */
2977 static int
2978 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2979 ip6_pcbopts(pktopt, m, so, sopt)
2980 #else
2981 ip6_pcbopts(pktopt, m, so)
2982 #endif
2983         struct ip6_pktopts **pktopt;
2984         struct mbuf *m;
2985         struct socket *so;
2986 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
2987         struct sockopt *sopt;
2988 #endif
2989 {
2990         struct ip6_pktopts *opt = *pktopt;
2991         int error = 0;
2992         int priv = 1;
2993
2994         /* turn off any old options. */
2995         if (opt) {
2996 #ifdef DIAGNOSTIC
2997             if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2998                 opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2999                 opt->ip6po_rhinfo.ip6po_rhi_rthdr)
3000                     printf("ip6_pcbopts: all specified options are cleared.\n");
3001 #endif
3002                 ip6_clearpktopts(opt, -1);
3003         } else
3004                 opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
3005         *pktopt = NULL;
3006
3007         if (!m || m->m_len == 0) {
3008                 /*
3009                  * Only turning off any previous options.
3010                  */
3011                 if (opt)
3012                         free(opt, M_IP6OPT);
3013                 return(0);
3014         }
3015
3016         if ((error = ip6_setpktoptions(m, opt, NULL, priv, 1)) != 0) {
3017                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
3018                 return(error);
3019         }
3020         *pktopt = opt;
3021         return(0);
3022 }
3023
3024 /*
3025  * initialize ip6_pktopts.  beware that there are non-zero default values in
3026  * the struct.
3027  */
3028 void
3029 init_ip6pktopts(opt)
3030         struct ip6_pktopts *opt;
3031 {
3032
3033         bzero(opt, sizeof(*opt));
3034         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
3035         opt->ip6po_tclass = -1; /* -1 means default traffic class */
3036 }
3037
3038 #define sin6tosa(sin6)  ((struct sockaddr *)(sin6)) /* XXX */
3039 static int
3040 ip6_pcbopt(optname, buf, len, pktopt, priv)
3041         int optname, len, priv;
3042         u_char *buf;
3043         struct ip6_pktopts **pktopt;
3044 {
3045         struct ip6_pktopts *opt;
3046
3047         if (*pktopt == NULL) {
3048                 *pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
3049                                  M_WAITOK);
3050                 init_ip6pktopts(*pktopt);
3051                 (*pktopt)->needfree = 1;
3052         }
3053         opt = *pktopt;
3054
3055         return(ip6_setpktoption(optname, buf, len, opt, priv, 1, 0));
3056 }
3057
3058 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
3059 static int
3060 ip6_getpcbopt(pktopt, optname, sopt)
3061         struct ip6_pktopts *pktopt;
3062         struct sockopt *sopt;
3063         int optname;
3064 #else
3065 static int
3066 ip6_getpcbopt(pktopt, optname, mp)
3067         struct ip6_pktopts *pktopt;
3068         int optname;
3069         struct mbuf **mp;
3070 #endif
3071 {
3072         void *optdata = NULL;
3073         int optdatalen = 0;
3074         struct ip6_ext *ip6e;
3075         int error = 0;
3076         struct in6_pktinfo null_pktinfo;
3077         int deftclass = 0;
3078 #if !(defined(__FreeBSD__) && __FreeBSD__ >= 3)
3079         struct mbuf *m;
3080 #endif
3081
3082         switch (optname) {
3083         case IPV6_PKTINFO:
3084                 if (pktopt && pktopt->ip6po_pktinfo)
3085                         optdata = (void *)pktopt->ip6po_pktinfo;
3086                 else {
3087                         /* XXX: we don't have to do this every time... */
3088                         bzero(&null_pktinfo, sizeof(null_pktinfo));
3089                         optdata = (void *)&null_pktinfo;
3090                 }
3091                 optdatalen = sizeof(struct in6_pktinfo);
3092                 break;
3093         case IPV6_OTCLASS:
3094                 /* XXX */
3095                 return(EINVAL);
3096         case IPV6_TCLASS:
3097                 if (pktopt && pktopt->ip6po_tclass >= 0)
3098                         optdata = (void *)&pktopt->ip6po_tclass;
3099                 else
3100                         optdata = (void *)&deftclass;
3101                 optdatalen = sizeof(int);
3102                 break;
3103         case IPV6_HOPOPTS:
3104                 if (pktopt && pktopt->ip6po_hbh) {
3105                         optdata = (void *)pktopt->ip6po_hbh;
3106                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
3107                         optdatalen = (ip6e->ip6e_len + 1) << 3;
3108                 }
3109                 break;
3110         case IPV6_RTHDR:
3111                 if (pktopt && pktopt->ip6po_rthdr) {
3112                         optdata = (void *)pktopt->ip6po_rthdr;
3113                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
3114                         optdatalen = (ip6e->ip6e_len + 1) << 3;
3115                 }
3116                 break;
3117         case IPV6_RTHDRDSTOPTS:
3118                 if (pktopt && pktopt->ip6po_dest1) {
3119                         optdata = (void *)pktopt->ip6po_dest1;
3120                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
3121                         optdatalen = (ip6e->ip6e_len + 1) << 3;
3122                 }
3123                 break;
3124         case IPV6_DSTOPTS:
3125                 if (pktopt && pktopt->ip6po_dest2) {
3126                         optdata = (void *)pktopt->ip6po_dest2;
3127                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
3128                         optdatalen = (ip6e->ip6e_len + 1) << 3;
3129                 }
3130                 break;
3131         case IPV6_NEXTHOP:
3132                 if (pktopt && pktopt->ip6po_nexthop) {
3133                         optdata = (void *)pktopt->ip6po_nexthop;
3134                         optdatalen = pktopt->ip6po_nexthop->sa_len;
3135                 }
3136                 break;
3137         default:                /* should not happen */
3138                 printf("ip6_getpcbopt: unexpected option: %d\n", optname);
3139                 return(ENOPROTOOPT);
3140         }
3141
3142 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
3143         error = sooptcopyout(sopt, optdata, optdatalen);
3144 #else  /* !FreeBSD3 */
3145         if (optdatalen > MCLBYTES)
3146                 return(EMSGSIZE); /* XXX */
3147         *mp = m = m_get(M_WAIT, MT_SOOPTS);
3148         if (optdatalen > MLEN)
3149                 MCLGET(m, M_WAIT);
3150         m->m_len = optdatalen;
3151         if (optdatalen)
3152                 bcopy(optdata, mtod(m, void *), optdatalen);
3153 #endif /* FreeBSD3 */
3154
3155         return(error);
3156 }
3157
3158 void
3159 ip6_clearpktopts(pktopt, optname)
3160         struct ip6_pktopts *pktopt;
3161         int optname;
3162 {
3163         int needfree = pktopt->needfree;
3164
3165         if (pktopt == NULL)
3166                 return;
3167
3168         if (optname == -1 || optname == IPV6_PKTINFO) {
3169                 if (needfree && pktopt->ip6po_pktinfo)
3170                         free(pktopt->ip6po_pktinfo, M_IP6OPT);
3171                 pktopt->ip6po_pktinfo = NULL;
3172         }
3173         if (optname == -1 || optname == IPV6_HOPLIMIT)
3174                 pktopt->ip6po_hlim = -1;
3175         if (optname == -1 || optname == IPV6_TCLASS)
3176                 pktopt->ip6po_tclass = -1;
3177         if (optname == -1 || optname == IPV6_NEXTHOP) {
3178                 if (pktopt->ip6po_nextroute.ro_rt) {
3179                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
3180                         pktopt->ip6po_nextroute.ro_rt = NULL;
3181                 }
3182                 if (needfree && pktopt->ip6po_nexthop)
3183                         free(pktopt->ip6po_nexthop, M_IP6OPT);
3184                 pktopt->ip6po_nexthop = NULL;
3185         }
3186         if (optname == -1 || optname == IPV6_HOPOPTS) {
3187                 if (needfree && pktopt->ip6po_hbh)
3188                         free(pktopt->ip6po_hbh, M_IP6OPT);
3189                 pktopt->ip6po_hbh = NULL;
3190         }
3191         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
3192                 if (needfree && pktopt->ip6po_dest1)
3193                         free(pktopt->ip6po_dest1, M_IP6OPT);
3194                 pktopt->ip6po_dest1 = NULL;
3195         }
3196         if (optname == -1 || optname == IPV6_RTHDR) {
3197                 if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
3198                         free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
3199                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
3200                 if (pktopt->ip6po_route.ro_rt) {
3201                         RTFREE(pktopt->ip6po_route.ro_rt);
3202                         pktopt->ip6po_route.ro_rt = NULL;
3203                 }
3204         }
3205         if (optname == -1 || optname == IPV6_DSTOPTS) {
3206                 if (needfree && pktopt->ip6po_dest2)
3207                         free(pktopt->ip6po_dest2, M_IP6OPT);
3208                 pktopt->ip6po_dest2 = NULL;
3209         }
3210 }
3211
3212 #define PKTOPT_EXTHDRCPY(type) \
3213 do {\
3214         if (src->type) {\
3215                 int hlen =\
3216                         (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
3217                 dst->type = malloc(hlen, M_IP6OPT, canwait);\
3218                 if (dst->type == NULL && canwait == M_NOWAIT)\
3219                         goto bad;\
3220                 bcopy(src->type, dst->type, hlen);\
3221         }\
3222 } while (0)
3223
3224 struct ip6_pktopts *
3225 ip6_copypktopts(src, canwait)
3226         struct ip6_pktopts *src;
3227         int canwait;
3228 {
3229         struct ip6_pktopts *dst;
3230
3231         if (src == NULL) {
3232                 printf("ip6_clearpktopts: invalid argument\n");
3233                 return(NULL);
3234         }
3235
3236         dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
3237         if (dst == NULL && canwait == M_NOWAIT)
3238                 goto bad;
3239         bzero(dst, sizeof(*dst));
3240         dst->needfree = 1;
3241
3242         dst->ip6po_hlim = src->ip6po_hlim;
3243         dst->ip6po_tclass = src->ip6po_tclass;
3244         dst->ip6po_flags = src->ip6po_flags;
3245         if (src->ip6po_pktinfo) {
3246                 dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
3247                                             M_IP6OPT, canwait);
3248                 if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
3249                         goto bad;
3250                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
3251         }
3252         if (src->ip6po_nexthop) {
3253                 dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
3254                                             M_IP6OPT, canwait);
3255                 if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
3256                         goto bad;
3257                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
3258                       src->ip6po_nexthop->sa_len);
3259         }
3260         PKTOPT_EXTHDRCPY(ip6po_hbh);
3261         PKTOPT_EXTHDRCPY(ip6po_dest1);
3262         PKTOPT_EXTHDRCPY(ip6po_dest2);
3263         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
3264         return(dst);
3265
3266   bad:
3267         printf("ip6_copypktopts: copy failed");
3268         if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
3269         if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
3270         if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
3271         if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
3272         if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
3273         if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
3274         return(NULL);
3275 }
3276 #undef PKTOPT_EXTHDRCPY
3277
3278 void
3279 ip6_freepcbopts(pktopt)
3280         struct ip6_pktopts *pktopt;
3281 {
3282         if (pktopt == NULL)
3283                 return;
3284
3285         ip6_clearpktopts(pktopt, -1);
3286
3287         free(pktopt, M_IP6OPT);
3288 }
3289
3290 /*
3291  * Set the IP6 multicast options in response to user setsockopt().
3292  */
3293 static int
3294 ip6_setmoptions(optname, im6op, m)
3295         int optname;
3296         struct ip6_moptions **im6op;
3297         struct mbuf *m;
3298 {
3299         int error = 0;
3300         u_int loop, ifindex;
3301         struct ipv6_mreq *mreq;
3302         struct ifnet *ifp;
3303         struct ip6_moptions *im6o = *im6op;
3304         struct route ro;
3305         struct sockaddr_in6 *dst;
3306         struct in6_multi_mship *imm;
3307
3308         if (im6o == NULL) {
3309                 /*
3310                  * No multicast option buffer attached to the pcb;
3311                  * allocate one and initialize to default values.
3312                  */
3313                 im6o = (struct ip6_moptions *)
3314                         malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
3315                 *im6op = im6o;
3316                 im6o->im6o_multicast_ifp = NULL;
3317                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
3318                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
3319                 LIST_INIT(&im6o->im6o_memberships);
3320         }
3321
3322         switch (optname) {
3323
3324         case IPV6_MULTICAST_IF:
3325                 /*
3326                  * Select the interface for outgoing multicast packets.
3327                  */
3328                 if (m == NULL || m->m_len != sizeof(u_int)) {
3329                         error = EINVAL;
3330                         break;
3331                 }
3332                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
3333                 if (ifindex < 0 || if_index < ifindex) {
3334                         error = ENXIO;  /* XXX EINVAL? */
3335                         break;
3336                 }
3337 #if defined(__FreeBSD__) && __FreeBSD__ >= 5
3338                 ifp = ifnet_byindex(ifindex);
3339 #else
3340                 ifp = ifindex2ifnet[ifindex];
3341 #endif
3342                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
3343                         error = EADDRNOTAVAIL;
3344                         break;
3345                 }
3346                 im6o->im6o_multicast_ifp = ifp;
3347                 break;
3348
3349         case IPV6_MULTICAST_HOPS:
3350             {
3351                 /*
3352                  * Set the IP6 hoplimit for outgoing multicast packets.
3353                  */
3354                 int optval;
3355                 if (m == NULL || m->m_len != sizeof(int)) {
3356                         error = EINVAL;
3357                         break;
3358                 }
3359                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
3360                 if (optval < -1 || optval >= 256)
3361                         error = EINVAL;
3362                 else if (optval == -1)
3363                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
3364                 else
3365                         im6o->im6o_multicast_hlim = optval;
3366                 break;
3367             }
3368
3369         case IPV6_MULTICAST_LOOP:
3370                 /*
3371                  * Set the loopback flag for outgoing multicast packets.
3372                  * Must be zero or one.
3373                  */
3374                 if (m == NULL || m->m_len != sizeof(u_int)) {
3375                         error = EINVAL;
3376                         break;
3377                 }
3378                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
3379                 if (loop > 1) {
3380                         error = EINVAL;
3381                         break;
3382                 }
3383                 im6o->im6o_multicast_loop = loop;
3384                 break;
3385
3386         case IPV6_JOIN_GROUP:
3387                 /*
3388                  * Add a multicast group membership.
3389                  * Group must be a valid IP6 multicast address.
3390                  */
3391                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
3392                         error = EINVAL;
3393                         break;
3394                 }
3395                 mreq = mtod(m, struct ipv6_mreq *);
3396                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
3397                         /*
3398                          * We use the unspecified address to specify to accept
3399                          * all multicast addresses. Only super user is allowed
3400                          * to do this.
3401                          */
3402                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
3403                         error = EINVAL;
3404                         break;
3405                 }
3406
3407                 /*
3408                  * If the interface is specified, validate it.
3409                  */
3410                 if (mreq->ipv6mr_interface < 0
3411                  || if_index < mreq->ipv6mr_interface) {
3412                         error = ENXIO;  /* XXX EINVAL? */
3413                         break;
3414                 }
3415                 /*
3416                  * If no interface was explicitly specified, choose an
3417                  * appropriate one according to the given multicast address.
3418                  */
3419                 if (mreq->ipv6mr_interface == 0) {
3420                         /*
3421                          * Look up the routing table for the
3422                          * address, and choose the outgoing interface.
3423                          *   XXX: is it a good approach?
3424                          */
3425                         ro.ro_rt = NULL;
3426                         dst = (struct sockaddr_in6 *)&ro.ro_dst;
3427                         bzero(dst, sizeof(*dst));
3428                         dst->sin6_len = sizeof(struct sockaddr_in6);
3429                         dst->sin6_family = AF_INET6;
3430                         dst->sin6_addr = mreq->ipv6mr_multiaddr;
3431                         rtalloc((struct route *)&ro);
3432                         if (ro.ro_rt == NULL) {
3433                                 error = EADDRNOTAVAIL;
3434                                 break;
3435                         }
3436                         ifp = ro.ro_rt->rt_ifp;
3437                         rtfree(ro.ro_rt);
3438                 } else {
3439 #if defined(__FreeBSD__) && __FreeBSD__ >= 5
3440                         ifp = ifnet_byindex(mreq->ipv6mr_interface);
3441 #else
3442                         ifp = ifindex2ifnet[mreq->ipv6mr_interface];
3443 #endif
3444                 }
3445
3446                 /*
3447                  * See if we found an interface, and confirm that it
3448                  * supports multicast
3449                  */
3450                 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
3451                         error = EADDRNOTAVAIL;
3452                         break;
3453                 }
3454                 /*
3455                  * Put interface index into the multicast address,
3456                  * if the address has interface/link-local scope.
3457                  * XXX: the embedded form is a KAME-local hack.
3458                  */
3459                 if (IN6_IS_ADDR_MC_INTFACELOCAL(&mreq->ipv6mr_multiaddr) ||
3460                     IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
3461                         mreq->ipv6mr_multiaddr.s6_addr16[1]
3462                                 = htons(mreq->ipv6mr_interface);
3463                 }
3464                 /*
3465                  * See if the membership already exists.
3466                  */
3467                 for (imm = im6o->im6o_memberships.lh_first;
3468                      imm != NULL; imm = imm->i6mm_chain.le_next)
3469                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
3470                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
3471                                                &mreq->ipv6mr_multiaddr))
3472                                 break;
3473                 if (imm != NULL) {
3474                         error = EADDRINUSE;
3475                         break;
3476                 }
3477                 /*
3478                  * Everything looks good; add a new record to the multicast
3479                  * address list for the given interface.
3480                  */
3481                 imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
3482                 if (!imm)
3483                         break;
3484                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
3485                 break;
3486
3487         case IPV6_LEAVE_GROUP:
3488                 /*
3489                  * Drop a multicast group membership.
3490                  * Group must be a valid IP6 multicast address.
3491                  */
3492                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
3493                         error = EINVAL;
3494                         break;
3495                 }
3496                 mreq = mtod(m, struct ipv6_mreq *);
3497                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
3498                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
3499                         error = EINVAL;
3500                         break;
3501                 }
3502                 /*
3503                  * If an interface address was specified, get a pointer
3504                  * to its ifnet structure.
3505                  */
3506                 if (mreq->ipv6mr_interface < 0
3507                  || if_index < mreq->ipv6mr_interface) {
3508                         error = ENXIO;  /* XXX EINVAL? */
3509                         break;
3510                 }
3511 #if defined(__FreeBSD__) && __FreeBSD__ >= 5
3512                 ifp = ifnet_byindex(mreq->ipv6mr_interface);
3513 #else
3514                 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
3515 #endif
3516                 /*
3517                  * Put interface index into the multicast address,
3518                  * if the address has interface/link-local scope.
3519                  */
3520                 if (IN6_IS_ADDR_MC_INTFACELOCAL(&mreq->ipv6mr_multiaddr) ||
3521                     IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
3522                         mreq->ipv6mr_multiaddr.s6_addr16[1]
3523                                 = htons(mreq->ipv6mr_interface);
3524                 }
3525                 /*
3526                  * Find the membership in the membership list.
3527                  */
3528                 for (imm = im6o->im6o_memberships.lh_first;
3529                      imm != NULL; imm = imm->i6mm_chain.le_next) {
3530                         if ((ifp == NULL ||
3531                              imm->i6mm_maddr->in6m_ifp == ifp) &&
3532                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
3533                                                &mreq->ipv6mr_multiaddr))
3534                                 break;
3535                 }
3536                 if (imm == NULL) {
3537                         /* Unable to resolve interface */
3538                         error = EADDRNOTAVAIL;
3539                         break;
3540                 }
3541                 /*
3542                  * Give up the multicast address record to which the
3543                  * membership points.
3544                  */
3545                 LIST_REMOVE(imm, i6mm_chain);
3546                 in6_leavegroup(imm);
3547                 break;
3548
3549         default:
3550                 error = EOPNOTSUPP;
3551                 break;
3552         }
3553
3554         /*
3555          * If all options have default values, no need to keep the mbuf.
3556          */
3557         if (im6o->im6o_multicast_ifp == NULL &&
3558             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
3559             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
3560             im6o->im6o_memberships.lh_first == NULL) {
3561                 free(*im6op, M_IPMOPTS);
3562                 *im6op = NULL;
3563         }
3564
3565         return(error);
3566 }
3567
3568 /*
3569  * Return the IP6 multicast options in response to user getsockopt().
3570  */
3571 static int
3572 ip6_getmoptions(optname, im6o, mp)
3573         int optname;
3574         struct ip6_moptions *im6o;
3575         struct mbuf **mp;
3576 {
3577         u_int *hlim, *loop, *ifindex;
3578
3579 #ifdef __FreeBSD__
3580         *mp = m_get(M_WAIT, MT_HEADER);         /* XXX */
3581 #else
3582         *mp = m_get(M_WAIT, MT_SOOPTS);
3583 #endif
3584
3585         switch (optname) {
3586
3587         case IPV6_MULTICAST_IF:
3588                 ifindex = mtod(*mp, u_int *);
3589                 (*mp)->m_len = sizeof(u_int);
3590                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
3591                         *ifindex = 0;
3592                 else
3593                         *ifindex = im6o->im6o_multicast_ifp->if_index;
3594                 return(0);
3595
3596         case IPV6_MULTICAST_HOPS:
3597                 hlim = mtod(*mp, u_int *);
3598                 (*mp)->m_len = sizeof(u_int);
3599                 if (im6o == NULL)
3600                         *hlim = ip6_defmcasthlim;
3601                 else
3602                         *hlim = im6o->im6o_multicast_hlim;
3603                 return(0);
3604
3605         case IPV6_MULTICAST_LOOP:
3606                 loop = mtod(*mp, u_int *);
3607                 (*mp)->m_len = sizeof(u_int);
3608                 if (im6o == NULL)
3609                         *loop = ip6_defmcasthlim;
3610                 else
3611                         *loop = im6o->im6o_multicast_loop;
3612                 return(0);
3613
3614         default:
3615                 return(EOPNOTSUPP);
3616         }
3617 }
3618
3619 /*
3620  * Discard the IP6 multicast options.
3621  */
3622 void
3623 ip6_freemoptions(im6o)
3624         struct ip6_moptions *im6o;
3625 {
3626         struct in6_multi_mship *imm;
3627
3628         if (im6o == NULL)
3629                 return;
3630
3631         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
3632                 LIST_REMOVE(imm, i6mm_chain);
3633                 in6_leavegroup(imm);
3634         }
3635         free(im6o, M_IPMOPTS);
3636 }
3637
3638 /*
3639  * Set IPv6 outgoing packet options based on advanced API.
3640  */
3641 int
3642 ip6_setpktoptions(control, opt, stickyopt, priv, needcopy)
3643         struct mbuf *control;
3644         struct ip6_pktopts *opt, *stickyopt;
3645         int priv, needcopy;
3646 {
3647         struct cmsghdr *cm = 0;
3648
3649         if (control == 0 || opt == 0)
3650                 return(EINVAL);
3651
3652         if (stickyopt) {
3653                 /*
3654                  * If stickyopt is provided, make a local copy of the options
3655                  * for this particular packet, then override them by ancillary
3656                  * objects.
3657                  * XXX: need to gain a reference for the cached route of the
3658                  * next hop in case of the overriding.
3659                  */
3660                 *opt = *stickyopt;
3661                 if (opt->ip6po_nextroute.ro_rt)
3662                         opt->ip6po_nextroute.ro_rt->rt_refcnt++;
3663         } else
3664                 init_ip6pktopts(opt);
3665         opt->needfree = needcopy;
3666
3667         /*
3668          * XXX: Currently, we assume all the optional information is stored
3669          * in a single mbuf.
3670          */
3671         if (control->m_next)
3672                 return(EINVAL);
3673
3674         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3675                      control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3676                 int error;
3677
3678                 if (control->m_len < CMSG_LEN(0))
3679                         return(EINVAL);
3680
3681                 cm = mtod(control, struct cmsghdr *);
3682                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3683                         return(EINVAL);
3684                 if (cm->cmsg_level != IPPROTO_IPV6)
3685                         continue;
3686
3687                 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3688                                          cm->cmsg_len - CMSG_LEN(0),
3689                                          opt, priv, needcopy, 1);
3690                 if (error)
3691                         return(error);
3692         }
3693
3694         return(0);
3695 }
3696
3697 /*
3698  * Set a particular packet option, as a sticky option or an ancillary data
3699  * item.  "len" can be 0 only when it's a sticky option.
3700  * We have 4 cases of combination of "sticky" and "cmsg":
3701  * "sticky=0, cmsg=0": impossible
3702  * "sticky=0, cmsg=1": RFC2292 or rfc2292bis ancillary data
3703  * "sticky=1, cmsg=0": rfc2292bis socket option
3704  * "sticky=1, cmsg=1": RFC2292 socket option
3705  */
3706 static int
3707 ip6_setpktoption(optname, buf, len, opt, priv, sticky, cmsg)
3708         int optname, len, priv, cmsg;
3709         u_char *buf;
3710         struct ip6_pktopts *opt;
3711         int sticky;
3712 {
3713         if (!sticky && !cmsg) {
3714 #ifdef DIAGNOSTIC
3715                 printf("ip6_setpktoption: impossible case\n");
3716 #endif
3717                 return(EINVAL);
3718         }
3719
3720         /*
3721          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
3722          * not be specified in the context of rfc2292bis.  Conversely,
3723          * rfc2292bis types should not be specified in the context of RFC2292.
3724          *
3725          */
3726         if (!cmsg) {
3727                 switch (optname) {
3728                 case IPV6_2292PKTINFO:
3729                 case IPV6_2292HOPLIMIT:
3730                 case IPV6_2292NEXTHOP:
3731                 case IPV6_2292HOPOPTS:
3732                 case IPV6_2292DSTOPTS:
3733                 case IPV6_2292RTHDR:
3734                 case IPV6_2292PKTOPTIONS:
3735                         return(ENOPROTOOPT);
3736                 }
3737         }
3738         if (sticky && cmsg) {
3739                 switch (optname) {
3740                 case IPV6_PKTINFO:
3741                 case IPV6_HOPLIMIT:
3742                 case IPV6_NEXTHOP:
3743                 case IPV6_HOPOPTS:
3744                 case IPV6_DSTOPTS:
3745                 case IPV6_RTHDRDSTOPTS:
3746                 case IPV6_RTHDR:
3747                 case IPV6_REACHCONF:
3748                 case IPV6_USE_MIN_MTU:
3749                 case IPV6_DONTFRAG:
3750                 case IPV6_OTCLASS:
3751                 case IPV6_TCLASS:
3752                         return(ENOPROTOOPT);
3753                 }
3754         }
3755
3756         switch (optname) {
3757         case IPV6_2292PKTINFO:
3758         case IPV6_PKTINFO:
3759         {
3760                 struct ifnet *ifp = NULL;
3761                 struct in6_pktinfo *pktinfo;
3762
3763                 if (len != sizeof(struct in6_pktinfo))
3764                         return(EINVAL);
3765
3766                 pktinfo = (struct in6_pktinfo *)buf;
3767
3768                 /*
3769                  * An application can clear any sticky IPV6_PKTINFO option by
3770                  * doing a "regular" setsockopt with ipi6_addr being
3771                  * in6addr_any and ipi6_ifindex being zero.
3772                  * [rfc2292bis-02, Section 6]
3773                  */
3774                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo) {
3775                         if (pktinfo->ipi6_ifindex == 0 &&
3776                             IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3777                                 ip6_clearpktopts(opt, optname);
3778                                 break;
3779                         }
3780                 }
3781
3782                 /* validate the interface index if specified. */
3783                 if (pktinfo->ipi6_ifindex > if_index ||
3784                     pktinfo->ipi6_ifindex < 0) {
3785                          return(ENXIO);
3786                 }
3787                 if (pktinfo->ipi6_ifindex) {
3788 #if defined(__FreeBSD__) && __FreeBSD__ >= 5
3789                         ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
3790 #else
3791                         ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
3792 #endif
3793                         if (ifp == NULL)
3794                                 return(ENXIO);
3795                 }
3796
3797                 /*
3798                  * We store the address anyway, and let in6_selectsrc()
3799                  * validate the specified address.  This is because ipi6_addr
3800                  * may not have enough information about its scope zone, and
3801                  * we may need additional information (such as outgoing
3802                  * interface or the scope zone of a destination address) to
3803                  * disambiguate the scope.
3804                  * XXX: the delay of the validation may confuse the
3805                  * application when it is used as a sticky option.
3806                  */
3807                 if (sticky) {
3808                         if (opt->ip6po_pktinfo == NULL) {
3809                                 opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3810                                                             M_IP6OPT,
3811                                                             M_WAITOK);
3812                         }
3813                         bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3814                 } else
3815                         opt->ip6po_pktinfo = pktinfo;
3816                 break;
3817         }
3818
3819         case IPV6_2292HOPLIMIT:
3820         case IPV6_HOPLIMIT:
3821         {
3822                 int *hlimp;
3823
3824                 /*
3825                  * rfc2292bis-03 obsoleted the usage of sticky IPV6_HOPLIMIT
3826                  * to simplify the ordering among hoplimit options.
3827                  */
3828                 if (optname == IPV6_HOPLIMIT && sticky)
3829                         return(ENOPROTOOPT);
3830
3831                 if (len != sizeof(int))
3832                         return(EINVAL);
3833                 hlimp = (int *)buf;
3834                 if (*hlimp < -1 || *hlimp > 255)
3835                         return(EINVAL);
3836
3837                 opt->ip6po_hlim = *hlimp;
3838                 break;
3839         }
3840
3841         case IPV6_OTCLASS:
3842                 if (len != sizeof(u_int8_t))
3843                         return(EINVAL);
3844
3845                 opt->ip6po_tclass = *(u_int8_t *)buf;
3846                 break;
3847
3848         case IPV6_TCLASS:
3849         {
3850                 int tclass;
3851
3852                 if (len != sizeof(int))
3853                         return(EINVAL);
3854                 tclass = *(int *)buf;
3855                 if (tclass < -1 || tclass > 255)
3856                         return(EINVAL);
3857
3858                 opt->ip6po_tclass = tclass;
3859                 break;
3860         }
3861
3862         case IPV6_2292NEXTHOP:
3863         case IPV6_NEXTHOP:
3864
3865                 if (len == 0) { /* just remove the option */
3866                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
3867                         break;
3868                 }
3869
3870                 /* check if cmsg_len is large enough for sa_len */
3871                 if (len < sizeof(struct sockaddr) || len < *buf)
3872                         return(EINVAL);
3873
3874                 switch (((struct sockaddr *)buf)->sa_family) {
3875                 case AF_INET6:
3876                 {
3877                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3878                         int error;
3879
3880                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3881                                 return(EINVAL);
3882
3883                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3884                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3885                                 return(EINVAL);
3886                         }
3887                         if ((error = scope6_check_id(sa6, ip6_use_defzone))
3888                             != 0) {
3889                                 return(error);
3890                         }
3891 #ifndef SCOPEDROUTING
3892                         sa6->sin6_scope_id = 0; /* XXX */
3893 #endif
3894                         break;
3895                 }
3896                 case AF_LINK:   /* should eventually be supported */
3897                 default:
3898                         return(EAFNOSUPPORT);
3899                 }
3900
3901                 /* turn off the previous option, then set the new option. */
3902                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
3903                 if (sticky) {
3904                         opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_WAITOK);
3905                         bcopy(buf, opt->ip6po_nexthop, *buf);
3906                 } else
3907                         opt->ip6po_nexthop = (struct sockaddr *)buf;
3908                 break;
3909
3910         case IPV6_2292HOPOPTS:
3911         case IPV6_HOPOPTS:
3912         {
3913                 struct ip6_hbh *hbh;
3914                 int hbhlen;
3915
3916                 /*
3917                  * XXX: We don't allow a non-privileged user to set ANY HbH
3918                  * options, since per-option restriction has too much
3919                  * overhead.
3920                  */
3921
3922                 if (len == 0) {
3923                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
3924                         break;  /* just remove the option */
3925                 }
3926
3927                 /* message length validation */
3928                 if (len < sizeof(struct ip6_hbh))
3929                         return(EINVAL);
3930                 hbh = (struct ip6_hbh *)buf;
3931                 hbhlen = (hbh->ip6h_len + 1) << 3;
3932                 if (len != hbhlen)
3933                         return(EINVAL);
3934
3935                 /* turn off the previous option, then set the new option. */
3936                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
3937                 if (sticky) {
3938                         opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_WAITOK);
3939                         bcopy(hbh, opt->ip6po_hbh, hbhlen);
3940                 } else
3941                         opt->ip6po_hbh = hbh;
3942
3943                 break;
3944         }
3945
3946         case IPV6_2292DSTOPTS:
3947         case IPV6_DSTOPTS:
3948         case IPV6_RTHDRDSTOPTS:
3949         {
3950                 struct ip6_dest *dest, **newdest = NULL;
3951                 int destlen;
3952
3953
3954                 if (len == 0) {
3955                         ip6_clearpktopts(opt, optname);
3956                         break;  /* just remove the option */
3957                 }
3958
3959                 /* message length validation */
3960                 if (len < sizeof(struct ip6_dest))
3961                         return(EINVAL);
3962                 dest = (struct ip6_dest *)buf;
3963                 destlen = (dest->ip6d_len + 1) << 3;
3964                 if (len != destlen)
3965                         return(EINVAL);
3966
3967                 /*
3968                  * Determine the position that the destination options header
3969                  * should be inserted; before or after the routing header.
3970                  */
3971                 switch (optname) {
3972                 case IPV6_2292DSTOPTS:
3973                         /*
3974                          * The old advacned API is ambiguous on this point.
3975                          * Our approach is to determine the position based
3976                          * according to the existence of a routing header.
3977                          * Note, however, that this depends on the order of the
3978                          * extension headers in the ancillary data; the 1st
3979                          * part of the destination options header must appear
3980                          * before the routing header in the ancillary data,
3981                          * too.
3982                          * RFC2292bis solved the ambiguity by introducing
3983                          * separate ancillary data or option types.
3984                          */
3985                         if (opt->ip6po_rthdr == NULL)
3986                                 newdest = &opt->ip6po_dest1;
3987                         else
3988                                 newdest = &opt->ip6po_dest2;
3989                         break;
3990                 case IPV6_RTHDRDSTOPTS:
3991                         newdest = &opt->ip6po_dest1;
3992                         break;
3993                 case IPV6_DSTOPTS:
3994                         newdest = &opt->ip6po_dest2;
3995                         break;
3996                 }
3997
3998                 /* turn off the previous option, then set the new option. */
3999                 ip6_clearpktopts(opt, optname);
4000                 if (sticky) {
4001                         *newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
4002                         bcopy(dest, *newdest, destlen);
4003                 } else
4004                         *newdest = dest;
4005
4006                 break;
4007         }
4008
4009         case IPV6_2292RTHDR:
4010         case IPV6_RTHDR:
4011         {
4012                 struct ip6_rthdr *rth;
4013                 int rthlen;
4014
4015                 if (len == 0) {
4016                         ip6_clearpktopts(opt, IPV6_RTHDR);
4017                         break;  /* just remove the option */
4018                 }
4019
4020                 /* message length validation */
4021                 if (len < sizeof(struct ip6_rthdr))
4022                         return(EINVAL);
4023                 rth = (struct ip6_rthdr *)buf;
4024                 rthlen = (rth->ip6r_len + 1) << 3;
4025                 if (len != rthlen)
4026                         return(EINVAL);
4027
4028                 switch (rth->ip6r_type) {
4029                 case IPV6_RTHDR_TYPE_0:
4030                         if (rth->ip6r_len == 0) /* must contain one addr */
4031                                 return(EINVAL);
4032                         if (rth->ip6r_len % 2) /* length must be even */
4033                                 return(EINVAL);
4034                         if (rth->ip6r_len / 2 != rth->ip6r_segleft)
4035                                 return(EINVAL);
4036                         break;
4037                 default:
4038                         return(EINVAL); /* not supported */
4039                 }
4040
4041                 /* turn off the previous option */
4042                 ip6_clearpktopts(opt, IPV6_RTHDR);
4043                 if (sticky) {
4044                         opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_WAITOK);
4045                         bcopy(rth, opt->ip6po_rthdr, rthlen);
4046                 } else
4047                         opt->ip6po_rthdr = rth;
4048
4049                 break;
4050         }
4051
4052         case IPV6_REACHCONF:
4053                 if (!cmsg)
4054                         return(ENOPROTOOPT);
4055
4056 #if 0
4057                 /*
4058                  * it looks dangerous to allow IPV6_REACHCONF to
4059                  * normal user.  it affects the ND state (system state)
4060                  * and can affect communication by others - jinmei
4061                  */
4062                 if (!priv)
4063                         return(EPERM);
4064 #else
4065                 /*
4066                  * we limit max # of subsequent userland reachability
4067                  * conformation by using ln->ln_byhint.
4068                  */
4069 #endif
4070                 if (len)
4071                         return(EINVAL);
4072                 opt->ip6po_flags |= IP6PO_REACHCONF;
4073                 break;
4074
4075         case IPV6_USE_MIN_MTU:
4076         case IPV6_DONTFRAG:
4077         {
4078                 int on, flag = 0;
4079
4080                 if (len != sizeof(int))
4081                         return(EINVAL);
4082                 on = *(int *)buf;
4083
4084                 switch (optname) {
4085                 case IPV6_USE_MIN_MTU:
4086                         flag = IP6PO_MINMTU;
4087                         break;
4088                 case IPV6_DONTFRAG:
4089                         flag = IP6PO_DONTFRAG;
4090                         break;
4091                 }
4092
4093                 if (on)
4094                         opt->ip6po_flags |= flag;
4095                 else
4096                         opt->ip6po_flags &= ~flag;
4097                 break;
4098         }
4099
4100         default:
4101                 return(ENOPROTOOPT);
4102         } /* end of switch */
4103
4104         return(0);
4105 }
4106
4107 /*
4108  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
4109  * packet to the input queue of a specified interface.  Note that this
4110  * calls the output routine of the loopback "driver", but with an interface
4111  * pointer that might NOT be &loif -- easier than replicating that code here.
4112  */
4113 void
4114 ip6_mloopback(ifp, m, dst)
4115         struct ifnet *ifp;
4116         struct mbuf *m;
4117         struct sockaddr_in6 *dst;
4118 {
4119         struct mbuf *copym;
4120         struct ip6_hdr *ip6;
4121
4122         copym = m_copy(m, 0, M_COPYALL);
4123         if (copym == NULL)
4124                 return;
4125
4126         /*
4127          * Make sure to deep-copy IPv6 header portion in case the data
4128          * is in an mbuf cluster, so that we can safely override the IPv6
4129          * header portion later.
4130          */
4131         if ((copym->m_flags & M_EXT) != 0 ||
4132             copym->m_len < sizeof(struct ip6_hdr)) {
4133                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
4134                 if (copym == NULL)
4135                         return;
4136         }
4137
4138 #ifdef DIAGNOSTIC
4139         if (copym->m_len < sizeof(*ip6)) {
4140                 m_freem(copym);
4141                 return;
4142         }
4143 #endif
4144
4145         ip6 = mtod(copym, struct ip6_hdr *);
4146 #ifndef SCOPEDROUTING
4147         /*
4148          * clear embedded scope identifiers if necessary.
4149          * in6_clearscope will touch the addresses only when necessary.
4150          */
4151         in6_clearscope(&ip6->ip6_src);
4152         in6_clearscope(&ip6->ip6_dst);
4153 #endif
4154
4155 #if defined(__FreeBSD__) && __FreeBSD__ >= 3
4156 #if (__FreeBSD_version >= 410000)
4157         (void)if_simloop(ifp, copym, dst->sin6_family, (int)NULL);
4158 #else
4159         (void)if_simloop(ifp, copym, (struct sockaddr *)dst, NULL);
4160 #endif
4161 #else
4162         (void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
4163 #endif
4164 }
4165
4166 /*
4167  * Chop IPv6 header off from the payload.
4168  */
4169 static int
4170 ip6_splithdr(m, exthdrs)
4171         struct mbuf *m;
4172         struct ip6_exthdrs *exthdrs;
4173 {
4174         struct mbuf *mh;
4175         struct ip6_hdr *ip6;
4176
4177         ip6 = mtod(m, struct ip6_hdr *);
4178         if (m->m_len > sizeof(*ip6)) {
4179                 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
4180                 if (mh == 0) {
4181                         m_freem(m);
4182                         return ENOBUFS;
4183                 }
4184 #ifdef __OpenBSD__
4185                 M_MOVE_PKTHDR(mh, m);
4186 #else
4187                 M_COPY_PKTHDR(mh, m);
4188 #endif
4189                 MH_ALIGN(mh, sizeof(*ip6));
4190                 m->m_flags &= ~M_PKTHDR;
4191                 m->m_len -= sizeof(*ip6);
4192                 m->m_data += sizeof(*ip6);
4193                 mh->m_next = m;
4194                 m = mh;
4195                 m->m_len = sizeof(*ip6);
4196                 bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
4197         }
4198         exthdrs->ip6e_ip6 = m;
4199         return 0;
4200 }
4201
4202 /*
4203  * Compute IPv6 extension header length.
4204  */
4205 #ifdef HAVE_NRL_INPCB
4206 # define in6pcb inpcb
4207 # define in6p_outputopts        inp_outputopts6
4208 #endif
4209 int
4210 ip6_optlen(in6p)
4211         struct in6pcb *in6p;
4212 {
4213         int len;
4214
4215         if (!in6p->in6p_outputopts)
4216                 return 0;
4217
4218         len = 0;
4219 #define elen(x) \
4220     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
4221
4222         len += elen(in6p->in6p_outputopts->ip6po_hbh);
4223         if (in6p->in6p_outputopts->ip6po_rthdr)
4224                 /* dest1 is valid with rthdr only */
4225                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
4226         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
4227         len += elen(in6p->in6p_outputopts->ip6po_dest2);
4228         return len;
4229 #undef elen
4230 }
4231 #ifdef HAVE_NRL_INPCB
4232 # undef in6pcb
4233 # undef in6p_outputopts
4234 #endif