sys/netinet6/ip6_output.c

   1 /*      $FreeBSD: src/sys/netinet6/ip6_output.c,v 1.13.2.18 2003/01/24 05:11:35 sam Exp $       */
   2 /*      $KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $    */
   3
   4 /*
   5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   6  * All rights reserved.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the name of the project nor the names of its contributors
  17  *    may be used to endorse or promote products derived from this software
  18  *    without specific prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  30  * SUCH DAMAGE.
  31  */
  32
  33 /*
  34  * Copyright (c) 1982, 1986, 1988, 1990, 1993
  35  *      The Regents of the University of California.  All rights reserved.
  36  *
  37  * Redistribution and use in source and binary forms, with or without
  38  * modification, are permitted provided that the following conditions
  39  * are met:
  40  * 1. Redistributions of source code must retain the above copyright
  41  *    notice, this list of conditions and the following disclaimer.
  42  * 2. Redistributions in binary form must reproduce the above copyright
  43  *    notice, this list of conditions and the following disclaimer in the
  44  *    documentation and/or other materials provided with the distribution.
  45  * 3. Neither the name of the University nor the names of its contributors
  46  *    may be used to endorse or promote products derived from this software
  47  *    without specific prior written permission.
  48  *
  49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  59  * SUCH DAMAGE.
  60  *
  61  *      @(#)ip_output.c 8.3 (Berkeley) 1/21/94
  62  */
  63
  64 #include "opt_ip6fw.h"
  65 #include "opt_inet.h"
  66 #include "opt_inet6.h"
  67
  68 #include <sys/param.h>
  69 #include <sys/malloc.h>
  70 #include <sys/mbuf.h>
  71 #include <sys/errno.h>
  72 #include <sys/protosw.h>
  73 #include <sys/socket.h>
  74 #include <sys/socketvar.h>
  75 #include <sys/systm.h>
  76 #include <sys/kernel.h>
  77 #include <sys/proc.h>
  78 #include <sys/priv.h>
  79
  80 #include <sys/msgport2.h>
  81
  82 #include <net/if.h>
  83 #include <net/route.h>
  84 #include <net/pfil.h>
  85
  86 #include <netinet/in.h>
  87 #include <netinet/in_var.h>
  88 #include <netinet6/in6_var.h>
  89 #include <netinet/ip6.h>
  90 #include <netinet/icmp6.h>
  91 #include <netinet6/ip6_var.h>
  92 #include <netinet/in_pcb.h>
  93 #include <netinet6/nd6.h>
  94 #include <netinet6/ip6protosw.h>
  95
  96 #include <net/ip6fw/ip6_fw.h>
  97
  98 #include <net/net_osdep.h>
  99
 100 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
 101
 102 struct ip6_exthdrs {
 103         struct mbuf *ip6e_ip6;
 104         struct mbuf *ip6e_hbh;
 105         struct mbuf *ip6e_dest1;
 106         struct mbuf *ip6e_rthdr;
 107         struct mbuf *ip6e_dest2;
 108 };
 109
 110 static int ip6_pcbopt (int, u_char *, int, struct ip6_pktopts **, int);
 111 static int ip6_setpktoption (int, u_char *, int, struct ip6_pktopts *,
 112         int, int, int, int);
 113 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, struct socket *,
 114         struct sockopt *);
 115 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
 116 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *);
 117 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **);
 118 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
 119         struct ifnet *, struct in6_addr *, u_long *, int *);
 120 static int copyexthdr (void *, struct mbuf **);
 121 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 122         struct ip6_frag **);
 123 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t);
 124 static struct mbuf *ip6_splithdr (struct mbuf *);
 125 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 126
 127 /*
 128  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
 129  * header (with pri, len, nxt, hlim, src, dst).
 130  * This function may modify ver and hlim only.
 131  * The mbuf chain containing the packet will be freed.
 132  * The mbuf opt, if present, will not be freed.
 133  *
 134  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
 135  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
 136  * which is rt_rmx.rmx_mtu.
 137  */
 138 int
 139 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
 140            int flags, struct ip6_moptions *im6o,
 141            struct ifnet **ifpp,         /* XXX: just for statistics */
 142            struct inpcb *inp)
 143 {
 144         struct ip6_hdr *ip6, *mhip6;
 145         struct ifnet *ifp, *origifp;
 146         struct mbuf *m = m0;
 147         struct mbuf *mprev;
 148         u_char *nexthdrp;
 149         int hlen, tlen, len, off;
 150         struct route_in6 ip6route;
 151         struct sockaddr_in6 *dst;
 152         int error = 0;
 153         struct in6_ifaddr *ia = NULL;
 154         u_long mtu;
 155         int alwaysfrag, dontfrag;
 156         u_int32_t optlen, plen = 0, unfragpartlen;
 157         struct ip6_exthdrs exthdrs;
 158         struct in6_addr finaldst;
 159         struct route_in6 *ro_pmtu = NULL;
 160         boolean_t hdrsplit = FALSE;
 161
 162         bzero(&exthdrs, sizeof exthdrs);
 163
 164         if (opt) {
 165                 if ((error = copyexthdr(opt->ip6po_hbh, &exthdrs.ip6e_hbh)))
 166                         goto freehdrs;
 167                 if ((error = copyexthdr(opt->ip6po_dest1, &exthdrs.ip6e_dest1)))
 168                         goto freehdrs;
 169                 if ((error = copyexthdr(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr)))
 170                         goto freehdrs;
 171                 if ((error = copyexthdr(opt->ip6po_dest2, &exthdrs.ip6e_dest2)))
 172                         goto freehdrs;
 173         }
 174
 175         /*
 176          * Calculate the total length of the extension header chain.
 177          * Keep the length of the unfragmentable part for fragmentation.
 178          */
 179         optlen = m_lengthm(exthdrs.ip6e_hbh, NULL) +
 180             m_lengthm(exthdrs.ip6e_dest1, NULL) +
 181             m_lengthm(exthdrs.ip6e_rthdr, NULL);
 182
 183         unfragpartlen = optlen + sizeof(struct ip6_hdr);
 184
 185         /* NOTE: we don't add AH/ESP length here. do that later. */
 186         optlen += m_lengthm(exthdrs.ip6e_dest2, NULL);
 187
 188         /*
 189          * If there is at least one extension header,
 190          * separate IP6 header from the payload.
 191          */
 192         if (optlen && !hdrsplit) {
 193                 exthdrs.ip6e_ip6 = ip6_splithdr(m);
 194                 if (exthdrs.ip6e_ip6 == NULL) {
 195                         error = ENOBUFS;
 196                         goto freehdrs;
 197                 }
 198                 m = exthdrs.ip6e_ip6;
 199                 hdrsplit = TRUE;
 200         }
 201
 202         /* adjust pointer */
 203         ip6 = mtod(m, struct ip6_hdr *);
 204
 205         /* adjust mbuf packet header length */
 206         m->m_pkthdr.len += optlen;
 207         plen = m->m_pkthdr.len - sizeof(*ip6);
 208
 209         /* If this is a jumbo payload, insert a jumbo payload option. */
 210         if (plen > IPV6_MAXPACKET) {
 211                 if (!hdrsplit) {
 212                         exthdrs.ip6e_ip6 = ip6_splithdr(m);
 213                         if (exthdrs.ip6e_ip6 == NULL) {
 214                                 error = ENOBUFS;
 215                                 goto freehdrs;
 216                         }
 217                         m = exthdrs.ip6e_ip6;
 218                         hdrsplit = TRUE;
 219                 }
 220                 /* adjust pointer */
 221                 ip6 = mtod(m, struct ip6_hdr *);
 222                 if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
 223                         goto freehdrs;
 224                 ip6->ip6_plen = 0;
 225         } else
 226                 ip6->ip6_plen = htons(plen);
 227
 228         /*
 229          * Concatenate headers and fill in next header fields.
 230          * Here we have, on "m"
 231          *      IPv6 payload
 232          * and we insert headers accordingly.  Finally, we should be getting:
 233          *      IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
 234          *
 235          * during the header composing process, "m" points to IPv6 header.
 236          * "mprev" points to an extension header prior to esp.
 237          */
 238
 239         nexthdrp = &ip6->ip6_nxt;
 240         mprev = m;
 241
 242         /*
 243          * we treat dest2 specially. the goal here is to make mprev point the
 244          * mbuf prior to dest2.
 245          *
 246          * result: IPv6 dest2 payload
 247          * m and mprev will point to IPv6 header.
 248          */
 249         if (exthdrs.ip6e_dest2) {
 250                 if (!hdrsplit)
 251                         panic("assumption failed: hdr not split");
 252                 exthdrs.ip6e_dest2->m_next = m->m_next;
 253                 m->m_next = exthdrs.ip6e_dest2;
 254                 *mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
 255                 ip6->ip6_nxt = IPPROTO_DSTOPTS;
 256         }
 257
 258 /*
 259  * Place m1 after mprev.
 260  */
 261 #define MAKE_CHAIN(m1, mprev, nexthdrp, i)\
 262     do {\
 263         if (m1) {\
 264                 if (!hdrsplit)\
 265                         panic("assumption failed: hdr not split");\
 266                 *mtod(m1, u_char *) = *nexthdrp;\
 267                 *nexthdrp = (i);\
 268                 nexthdrp = mtod(m1, u_char *);\
 269                 m1->m_next = mprev->m_next;\
 270                 mprev->m_next = m1;\
 271                 mprev = m1;\
 272         }\
 273     } while (0)
 274
 275         /*
 276          * result: IPv6 hbh dest1 rthdr dest2 payload
 277          * m will point to IPv6 header.  mprev will point to the
 278          * extension header prior to dest2 (rthdr in the above case).
 279          */
 280         MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
 281         MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp, IPPROTO_DSTOPTS);
 282         MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp, IPPROTO_ROUTING);
 283
 284         /*
 285          * If there is a routing header, replace the destination address field
 286          * with the first hop of the routing header.
 287          */
 288         if (exthdrs.ip6e_rthdr) {
 289                 struct ip6_rthdr *rh;
 290
 291                 finaldst = ip6->ip6_dst;
 292                 rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
 293                 switch (rh->ip6r_type) {
 294                 default:        /* is it possible? */
 295                          error = EINVAL;
 296                          goto bad;
 297                 }
 298         }
 299
 300         /* Source address validation */
 301         if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
 302             !(flags & IPV6_DADOUTPUT)) {
 303                 error = EOPNOTSUPP;
 304                 ip6stat.ip6s_badscope++;
 305                 goto bad;
 306         }
 307         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
 308                 error = EOPNOTSUPP;
 309                 ip6stat.ip6s_badscope++;
 310                 goto bad;
 311         }
 312
 313         ip6stat.ip6s_localout++;
 314
 315         /*
 316          * Route packet.
 317          */
 318         if (ro == NULL) {
 319                 ro = &ip6route;
 320                 bzero(ro, sizeof(*ro));
 321         }
 322         ro_pmtu = ro;
 323         if (opt && opt->ip6po_rthdr)
 324                 ro = &opt->ip6po_route;
 325         dst = (struct sockaddr_in6 *)&ro->ro_dst;
 326
 327         /*
 328          * If there is a cached route,
 329          * check that it is to the same destination
 330          * and is still up. If not, free it and try again.
 331          */
 332         if (ro->ro_rt != NULL &&
 333             (!(ro->ro_rt->rt_flags & RTF_UP) || dst->sin6_family != AF_INET6 ||
 334              !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
 335                 RTFREE(ro->ro_rt);
 336                 ro->ro_rt = NULL;
 337         }
 338         if (ro->ro_rt == NULL) {
 339                 bzero(dst, sizeof(*dst));
 340                 dst->sin6_family = AF_INET6;
 341                 dst->sin6_len = sizeof(struct sockaddr_in6);
 342                 dst->sin6_addr = ip6->ip6_dst;
 343         }
 344         if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
 345                 /* Unicast */
 346
 347 #define ifatoia6(ifa)   ((struct in6_ifaddr *)(ifa))
 348 #define sin6tosa(sin6)  ((struct sockaddr *)(sin6))
 349                 /* xxx
 350                  * interface selection comes here
 351                  * if an interface is specified from an upper layer,
 352                  * ifp must point it.
 353                  */
 354                 if (ro->ro_rt == NULL) {
 355                         /*
 356                          * non-bsdi always clone routes, if parent is
 357                          * PRF_CLONING.
 358                          */
 359                         rtalloc((struct route *)ro);
 360                 }
 361                 if (ro->ro_rt == NULL) {
 362                         ip6stat.ip6s_noroute++;
 363                         error = EHOSTUNREACH;
 364                         /* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
 365                         goto bad;
 366                 }
 367                 ia = ifatoia6(ro->ro_rt->rt_ifa);
 368                 ifp = ro->ro_rt->rt_ifp;
 369                 ro->ro_rt->rt_use++;
 370                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
 371                         dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
 372                 m->m_flags &= ~(M_BCAST | M_MCAST);     /* just in case */
 373
 374                 in6_ifstat_inc(ifp, ifs6_out_request);
 375
 376                 /*
 377                  * Check if the outgoing interface conflicts with
 378                  * the interface specified by ifi6_ifindex (if specified).
 379                  * Note that loopback interface is always okay.
 380                  * (this may happen when we are sending a packet to one of
 381                  *  our own addresses.)
 382                  */
 383                 if (opt && opt->ip6po_pktinfo
 384                  && opt->ip6po_pktinfo->ipi6_ifindex) {
 385                         if (!(ifp->if_flags & IFF_LOOPBACK)
 386                          && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
 387                                 ip6stat.ip6s_noroute++;
 388                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 389                                 error = EHOSTUNREACH;
 390                                 goto bad;
 391                         }
 392                 }
 393
 394                 if (opt && opt->ip6po_hlim != -1)
 395                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 396         } else {
 397                 /* Multicast */
 398                 struct  in6_multi *in6m;
 399
 400                 m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
 401
 402                 /*
 403                  * See if the caller provided any multicast options
 404                  */
 405                 ifp = NULL;
 406                 if (im6o != NULL) {
 407                         ip6->ip6_hlim = im6o->im6o_multicast_hlim;
 408                         if (im6o->im6o_multicast_ifp != NULL)
 409                                 ifp = im6o->im6o_multicast_ifp;
 410                 } else
 411                         ip6->ip6_hlim = ip6_defmcasthlim;
 412
 413                 /*
 414                  * See if the caller provided the outgoing interface
 415                  * as an ancillary data.
 416                  * Boundary check for ifindex is assumed to be already done.
 417                  */
 418                 if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
 419                         ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
 420
 421                 /*
 422                  * If the destination is a node-local scope multicast,
 423                  * the packet should be loop-backed only.
 424                  */
 425                 if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
 426                         /*
 427                          * If the outgoing interface is already specified,
 428                          * it should be a loopback interface.
 429                          */
 430                         if (ifp && !(ifp->if_flags & IFF_LOOPBACK)) {
 431                                 ip6stat.ip6s_badscope++;
 432                                 error = ENETUNREACH; /* XXX: better error? */
 433                                 /* XXX correct ifp? */
 434                                 in6_ifstat_inc(ifp, ifs6_out_discard);
 435                                 goto bad;
 436                         } else {
 437                                 ifp = loif;
 438                         }
 439                 }
 440
 441                 if (opt && opt->ip6po_hlim != -1)
 442                         ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
 443
 444                 /*
 445                  * If caller did not provide an interface lookup a
 446                  * default in the routing table.  This is either a
 447                  * default for the speicfied group (i.e. a host
 448                  * route), or a multicast default (a route for the
 449                  * ``net'' ff00::/8).
 450                  */
 451                 if (ifp == NULL) {
 452                         if (ro->ro_rt == NULL) {
 453                                 ro->ro_rt =
 454                                   rtpurelookup((struct sockaddr *)&ro->ro_dst);
 455                         }
 456                         if (ro->ro_rt == NULL) {
 457                                 ip6stat.ip6s_noroute++;
 458                                 error = EHOSTUNREACH;
 459                                 /* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
 460                                 goto bad;
 461                         }
 462                         ia = ifatoia6(ro->ro_rt->rt_ifa);
 463                         ifp = ro->ro_rt->rt_ifp;
 464                         ro->ro_rt->rt_use++;
 465                 }
 466
 467                 if (!(flags & IPV6_FORWARDING))
 468                         in6_ifstat_inc(ifp, ifs6_out_request);
 469                 in6_ifstat_inc(ifp, ifs6_out_mcast);
 470
 471                 /*
 472                  * Confirm that the outgoing interface supports multicast.
 473                  */
 474                 if (!(ifp->if_flags & IFF_MULTICAST)) {
 475                         ip6stat.ip6s_noroute++;
 476                         in6_ifstat_inc(ifp, ifs6_out_discard);
 477                         error = ENETUNREACH;
 478                         goto bad;
 479                 }
 480                 in6m = IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp);
 481                 if (in6m != NULL &&
 482                    (im6o == NULL || im6o->im6o_multicast_loop)) {
 483                         /*
 484                          * If we belong to the destination multicast group
 485                          * on the outgoing interface, and the caller did not
 486                          * forbid loopback, loop back a copy.
 487                          */
 488                         ip6_mloopback(ifp, m, dst);
 489                 } else {
 490                         /*
 491                          * If we are acting as a multicast router, perform
 492                          * multicast forwarding as if the packet had just
 493                          * arrived on the interface to which we are about
 494                          * to send.  The multicast forwarding function
 495                          * recursively calls this function, using the
 496                          * IPV6_FORWARDING flag to prevent infinite recursion.
 497                          *
 498                          * Multicasts that are looped back by ip6_mloopback(),
 499                          * above, will be forwarded by the ip6_input() routine,
 500                          * if necessary.
 501                          */
 502                         if (ip6_mrouter && !(flags & IPV6_FORWARDING)) {
 503                                 if (ip6_mforward(ip6, ifp, m) != 0) {
 504                                         m_freem(m);
 505                                         goto done;
 506                                 }
 507                         }
 508                 }
 509                 /*
 510                  * Multicasts with a hoplimit of zero may be looped back,
 511                  * above, but must not be transmitted on a network.
 512                  * Also, multicasts addressed to the loopback interface
 513                  * are not sent -- the above call to ip6_mloopback() will
 514                  * loop back a copy if this host actually belongs to the
 515                  * destination group on the loopback interface.
 516                  */
 517                 if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
 518                         m_freem(m);
 519                         goto done;
 520                 }
 521         }
 522
 523         /*
 524          * Fill the outgoing inteface to tell the upper layer
 525          * to increment per-interface statistics.
 526          */
 527         if (ifpp)
 528                 *ifpp = ifp;
 529
 530         /* Determine path MTU. */
 531         if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
 532             &alwaysfrag)) != 0)
 533                 goto bad;
 534
 535         /*
 536          * The caller of this function may specify to use the minimum MTU
 537          * in some cases.
 538          * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
 539          * setting.  The logic is a bit complicated; by default, unicast
 540          * packets will follow path MTU while multicast packets will be sent at
 541          * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
 542          * including unicast ones will be sent at the minimum MTU.  Multicast
 543          * packets will always be sent at the minimum MTU unless
 544          * IP6PO_MINMTU_DISABLE is explicitly specified.
 545          * See RFC 3542 for more details.
 546          */
 547         if (mtu > IPV6_MMTU) {
 548                 if ((flags & IPV6_MINMTU))
 549                         mtu = IPV6_MMTU;
 550                 else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
 551                         mtu = IPV6_MMTU;
 552                 else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
 553                          (opt == NULL ||
 554                           opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
 555                         mtu = IPV6_MMTU;
 556                 }
 557         }
 558
 559         /* Fake scoped addresses */
 560         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
 561                 /*
 562                  * If source or destination address is a scoped address, and
 563                  * the packet is going to be sent to a loopback interface,
 564                  * we should keep the original interface.
 565                  */
 566
 567                 /*
 568                  * XXX: this is a very experimental and temporary solution.
 569                  * We eventually have sockaddr_in6 and use the sin6_scope_id
 570                  * field of the structure here.
 571                  * We rely on the consistency between two scope zone ids
 572                  * of source and destination, which should already be assured.
 573                  * Larger scopes than link will be supported in the future.
 574                  */
 575                 origifp = NULL;
 576                 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
 577                         origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
 578                 else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
 579                         origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
 580                 /*
 581                  * XXX: origifp can be NULL even in those two cases above.
 582                  * For example, if we remove the (only) link-local address
 583                  * from the loopback interface, and try to send a link-local
 584                  * address without link-id information.  Then the source
 585                  * address is ::1, and the destination address is the
 586                  * link-local address with its s6_addr16[1] being zero.
 587                  * What is worse, if the packet goes to the loopback interface
 588                  * by a default rejected route, the null pointer would be
 589                  * passed to looutput, and the kernel would hang.
 590                  * The following last resort would prevent such disaster.
 591                  */
 592                 if (origifp == NULL)
 593                         origifp = ifp;
 594         }
 595         else
 596                 origifp = ifp;
 597         /*
 598          * clear embedded scope identifiers if necessary.
 599          * in6_clearscope will touch the addresses only when necessary.
 600          */
 601         in6_clearscope(&ip6->ip6_src);
 602         in6_clearscope(&ip6->ip6_dst);
 603
 604         /*
 605          * Check with the firewall...
 606          */
 607         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 608                 u_short port = 0;
 609
 610                 m->m_pkthdr.rcvif = NULL;       /* XXX */
 611                 /* If ipfw says divert, we have to just drop packet */
 612                 if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
 613                         m_freem(m);
 614                         goto done;
 615                 }
 616                 if (!m) {
 617                         error = EACCES;
 618                         goto done;
 619                 }
 620         }
 621
 622         /*
 623          * If the outgoing packet contains a hop-by-hop options header,
 624          * it must be examined and processed even by the source node.
 625          * (RFC 2460, section 4.)
 626          */
 627         if (exthdrs.ip6e_hbh) {
 628                 struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
 629                 u_int32_t dummy1; /* XXX unused */
 630                 u_int32_t dummy2; /* XXX unused */
 631
 632 #ifdef DIAGNOSTIC
 633                 if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
 634                         panic("ip6e_hbh is not continuous");
 635 #endif
 636                 /*
 637                  *  XXX: if we have to send an ICMPv6 error to the sender,
 638                  *       we need the M_LOOP flag since icmp6_error() expects
 639                  *       the IPv6 and the hop-by-hop options header are
 640                  *       continuous unless the flag is set.
 641                  */
 642                 m->m_flags |= M_LOOP;
 643                 m->m_pkthdr.rcvif = ifp;
 644                 if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
 645                     ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
 646                     &dummy1, &dummy2) < 0) {
 647                         /* m was already freed at this point */
 648                         error = EINVAL;/* better error? */
 649                         goto done;
 650                 }
 651                 m->m_flags &= ~M_LOOP; /* XXX */
 652                 m->m_pkthdr.rcvif = NULL;
 653         }
 654
 655         /*
 656          * Run through list of hooks for output packets.
 657          */
 658         if (pfil_has_hooks(&inet6_pfil_hook)) {
 659                 error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
 660                 if (error != 0 || m == NULL)
 661                         goto done;
 662                 ip6 = mtod(m, struct ip6_hdr *);
 663         }
 664
 665         /*
 666          * Send the packet to the outgoing interface.
 667          * If necessary, do IPv6 fragmentation before sending.
 668          *
 669          * the logic here is rather complex:
 670          * 1: normal case (dontfrag == 0, alwaysfrag == 0)
 671          * 1-a: send as is if tlen <= path mtu
 672          * 1-b: fragment if tlen > path mtu
 673          *
 674          * 2: if user asks us not to fragment (dontfrag == 1)
 675          * 2-a: send as is if tlen <= interface mtu
 676          * 2-b: error if tlen > interface mtu
 677          *
 678          * 3: if we always need to attach fragment header (alwaysfrag == 1)
 679          *      always fragment
 680          *
 681          * 4: if dontfrag == 1 && alwaysfrag == 1
 682          *      error, as we cannot handle this conflicting request
 683          */
 684         tlen = m->m_pkthdr.len;
 685
 686         if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
 687                 dontfrag = 1;
 688         else
 689                 dontfrag = 0;
 690         if (dontfrag && alwaysfrag) {   /* case 4 */
 691                 /* conflicting request - can't transmit */
 692                 error = EMSGSIZE;
 693                 goto bad;
 694         }
 695         if (dontfrag && tlen > IN6_LINKMTU(ifp)) {      /* case 2-b */
 696                 /*
 697                  * Even if the DONTFRAG option is specified, we cannot send the
 698                  * packet when the data length is larger than the MTU of the
 699                  * outgoing interface.
 700                  * Notify the error by sending IPV6_PATHMTU ancillary data as
 701                  * well as returning an error code (the latter is not described
 702                  * in the API spec.)
 703                  */
 704                 u_int32_t mtu32;
 705                 struct ip6ctlparam ip6cp;
 706
 707                 mtu32 = (u_int32_t)mtu;
 708                 bzero(&ip6cp, sizeof(ip6cp));
 709                 ip6cp.ip6c_cmdarg = (void *)&mtu32;
 710                 kpfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
 711                     (void *)&ip6cp);
 712
 713                 error = EMSGSIZE;
 714                 goto bad;
 715         }
 716
 717         /*
 718          * transmit packet without fragmentation
 719          */
 720         if (dontfrag || (!alwaysfrag && tlen <= mtu)) { /* case 1-a and 2-a */
 721                 struct in6_ifaddr *ia6;
 722
 723                 ip6 = mtod(m, struct ip6_hdr *);
 724                 ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
 725                 if (ia6) {
 726                         /* Record statistics for this interface address. */
 727                         IFA_STAT_INC(&ia6->ia_ifa, opackets, 1);
 728                         IFA_STAT_INC(&ia6->ia_ifa, obytes, m->m_pkthdr.len);
 729                 }
 730                 error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 731                 goto done;
 732         }
 733
 734         /*
 735          * try to fragment the packet.  case 1-b and 3
 736          */
 737         if (mtu < IPV6_MMTU) {
 738                 /*
 739                  * note that path MTU is never less than IPV6_MMTU
 740                  * (see icmp6_input).
 741                  */
 742                 error = EMSGSIZE;
 743                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 744                 goto bad;
 745         } else if (ip6->ip6_plen == 0) {
 746                 /* jumbo payload cannot be fragmented */
 747                 error = EMSGSIZE;
 748                 in6_ifstat_inc(ifp, ifs6_out_fragfail);
 749                 goto bad;
 750         } else {
 751                 struct mbuf **mnext, *m_frgpart;
 752                 struct ip6_frag *ip6f;
 753                 u_int32_t id = htonl(ip6_id++);
 754                 u_char nextproto;
 755
 756                 /*
 757                  * Too large for the destination or interface;
 758                  * fragment if possible.
 759                  * Must be able to put at least 8 bytes per fragment.
 760                  */
 761                 hlen = unfragpartlen;
 762                 if (mtu > IPV6_MAXPACKET)
 763                         mtu = IPV6_MAXPACKET;
 764
 765                 len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
 766                 if (len < 8) {
 767                         error = EMSGSIZE;
 768                         in6_ifstat_inc(ifp, ifs6_out_fragfail);
 769                         goto bad;
 770                 }
 771
 772                 mnext = &m->m_nextpkt;
 773
 774                 /*
 775                  * Change the next header field of the last header in the
 776                  * unfragmentable part.
 777                  */
 778                 if (exthdrs.ip6e_rthdr) {
 779                         nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
 780                         *mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
 781                 } else if (exthdrs.ip6e_dest1) {
 782                         nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
 783                         *mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
 784                 } else if (exthdrs.ip6e_hbh) {
 785                         nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
 786                         *mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
 787                 } else {
 788                         nextproto = ip6->ip6_nxt;
 789                         ip6->ip6_nxt = IPPROTO_FRAGMENT;
 790                 }
 791
 792                 /*
 793                  * Loop through length of segment after first fragment,
 794                  * make new header and copy data of each part and link onto
 795                  * chain.
 796                  */
 797                 m0 = m;
 798                 for (off = hlen; off < tlen; off += len) {
 799                         MGETHDR(m, M_NOWAIT, MT_HEADER);
 800                         if (!m) {
 801                                 error = ENOBUFS;
 802                                 ip6stat.ip6s_odropped++;
 803                                 goto sendorfree;
 804                         }
 805                         m->m_pkthdr.rcvif = NULL;
 806                         m->m_flags = m0->m_flags & M_COPYFLAGS;
 807                         *mnext = m;
 808                         mnext = &m->m_nextpkt;
 809                         m->m_data += max_linkhdr;
 810                         mhip6 = mtod(m, struct ip6_hdr *);
 811                         *mhip6 = *ip6;
 812                         m->m_len = sizeof(*mhip6);
 813                         error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
 814                         if (error) {
 815                                 ip6stat.ip6s_odropped++;
 816                                 goto sendorfree;
 817                         }
 818                         ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
 819                         if (off + len >= tlen)
 820                                 len = tlen - off;
 821                         else
 822                                 ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
 823                         mhip6->ip6_plen = htons((u_short)(len + hlen +
 824                             sizeof(*ip6f) - sizeof(struct ip6_hdr)));
 825                         if ((m_frgpart = m_copy(m0, off, len)) == NULL) {
 826                                 error = ENOBUFS;
 827                                 ip6stat.ip6s_odropped++;
 828                                 goto sendorfree;
 829                         }
 830                         m_cat(m, m_frgpart);
 831                         m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
 832                         m->m_pkthdr.rcvif = NULL;
 833                         ip6f->ip6f_reserved = 0;
 834                         ip6f->ip6f_ident = id;
 835                         ip6f->ip6f_nxt = nextproto;
 836                         ip6stat.ip6s_ofragments++;
 837                         in6_ifstat_inc(ifp, ifs6_out_fragcreat);
 838                 }
 839
 840                 in6_ifstat_inc(ifp, ifs6_out_fragok);
 841         }
 842
 843         /*
 844          * Remove leading garbages.
 845          */
 846 sendorfree:
 847         m = m0->m_nextpkt;
 848         m0->m_nextpkt = NULL;
 849         m_freem(m0);
 850         for (m0 = m; m; m = m0) {
 851                 m0 = m->m_nextpkt;
 852                 m->m_nextpkt = NULL;
 853                 if (error == 0) {
 854                         /* Record statistics for this interface address. */
 855                         if (ia) {
 856                                 IFA_STAT_INC(&ia->ia_ifa, opackets, 1);
 857                                 IFA_STAT_INC(&ia->ia_ifa, obytes,
 858                                     m->m_pkthdr.len);
 859                         }
 860                         error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
 861                 } else
 862                         m_freem(m);
 863         }
 864
 865         if (error == 0)
 866                 ip6stat.ip6s_fragmented++;
 867
 868 done:
 869         if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
 870                 RTFREE(ro->ro_rt);
 871         } else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
 872                 RTFREE(ro_pmtu->ro_rt);
 873         }
 874
 875         return (error);
 876
 877 freehdrs:
 878         m_freem(exthdrs.ip6e_hbh);      /* m_freem will check if mbuf is 0 */
 879         m_freem(exthdrs.ip6e_dest1);
 880         m_freem(exthdrs.ip6e_rthdr);
 881         m_freem(exthdrs.ip6e_dest2);
 882         /* FALLTHROUGH */
 883 bad:
 884         m_freem(m);
 885         goto done;
 886 }
 887
 888 static int
 889 copyexthdr(void *h, struct mbuf **mp)
 890 {
 891         struct ip6_ext *hdr = h;
 892         int hlen;
 893         struct mbuf *m;
 894
 895         if (hdr == NULL)
 896                 return 0;
 897
 898         hlen = (hdr->ip6e_len + 1) * 8;
 899         if (hlen > MCLBYTES)
 900                 return ENOBUFS; /* XXX */
 901
 902         m = m_getb(hlen, M_NOWAIT, MT_DATA, 0);
 903         if (!m)
 904                 return ENOBUFS;
 905         m->m_len = hlen;
 906
 907         bcopy(hdr, mtod(m, caddr_t), hlen);
 908
 909         *mp = m;
 910         return 0;
 911 }
 912
 913 /*
 914  * Insert jumbo payload option.
 915  */
 916 static int
 917 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
 918 {
 919         struct mbuf *mopt;
 920         u_char *optbuf;
 921         u_int32_t v;
 922
 923 #define JUMBOOPTLEN     8       /* length of jumbo payload option and padding */
 924
 925         /*
 926          * If there is no hop-by-hop options header, allocate new one.
 927          * If there is one but it doesn't have enough space to store the
 928          * jumbo payload option, allocate a cluster to store the whole options.
 929          * Otherwise, use it to store the options.
 930          */
 931         if (exthdrs->ip6e_hbh == NULL) {
 932                 MGET(mopt, M_NOWAIT, MT_DATA);
 933                 if (mopt == NULL)
 934                         return (ENOBUFS);
 935                 mopt->m_len = JUMBOOPTLEN;
 936                 optbuf = mtod(mopt, u_char *);
 937                 optbuf[1] = 0;  /* = ((JUMBOOPTLEN) >> 3) - 1 */
 938                 exthdrs->ip6e_hbh = mopt;
 939         } else {
 940                 struct ip6_hbh *hbh;
 941
 942                 mopt = exthdrs->ip6e_hbh;
 943                 if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
 944                         /*
 945                          * XXX assumption:
 946                          * - exthdrs->ip6e_hbh is not referenced from places
 947                          *   other than exthdrs.
 948                          * - exthdrs->ip6e_hbh is not an mbuf chain.
 949                          */
 950                         int oldoptlen = mopt->m_len;
 951                         struct mbuf *n;
 952
 953                         /*
 954                          * XXX: give up if the whole (new) hbh header does
 955                          * not fit even in an mbuf cluster.
 956                          */
 957                         if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
 958                                 return (ENOBUFS);
 959
 960                         /*
 961                          * As a consequence, we must always prepare a cluster
 962                          * at this point.
 963                          */
 964                         n = m_getcl(M_NOWAIT, MT_DATA, 0);
 965                         if (!n)
 966                                 return (ENOBUFS);
 967                         n->m_len = oldoptlen + JUMBOOPTLEN;
 968                         bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t), oldoptlen);
 969                         optbuf = mtod(n, caddr_t) + oldoptlen;
 970                         m_freem(mopt);
 971                         mopt = exthdrs->ip6e_hbh = n;
 972                 } else {
 973                         optbuf = mtod(mopt, u_char *) + mopt->m_len;
 974                         mopt->m_len += JUMBOOPTLEN;
 975                 }
 976                 optbuf[0] = IP6OPT_PADN;
 977                 optbuf[1] = 1;
 978
 979                 /*
 980                  * Adjust the header length according to the pad and
 981                  * the jumbo payload option.
 982                  */
 983                 hbh = mtod(mopt, struct ip6_hbh *);
 984                 hbh->ip6h_len += (JUMBOOPTLEN >> 3);
 985         }
 986
 987         /* fill in the option. */
 988         optbuf[2] = IP6OPT_JUMBO;
 989         optbuf[3] = 4;
 990         v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
 991         bcopy(&v, &optbuf[4], sizeof(u_int32_t));
 992
 993         /* finally, adjust the packet header length */
 994         exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
 995
 996         return (0);
 997 #undef JUMBOOPTLEN
 998 }
 999
1000 /*
1001  * Insert fragment header and copy unfragmentable header portions.
1002  */
1003 static int
1004 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1005                   struct ip6_frag **frghdrp)
1006 {
1007         struct mbuf *n, *mlast;
1008
1009         if (hlen > sizeof(struct ip6_hdr)) {
1010                 n = m_copym(m0, sizeof(struct ip6_hdr),
1011                             hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1012                 if (n == NULL)
1013                         return (ENOBUFS);
1014                 m->m_next = n;
1015         } else
1016                 n = m;
1017
1018         /* Search for the last mbuf of unfragmentable part. */
1019         for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1020                 ;
1021
1022         if (!(mlast->m_flags & M_EXT) &&
1023             M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1024                 /* use the trailing space of the last mbuf for the fragment hdr */
1025                 *frghdrp = (struct ip6_frag *)
1026                     (mtod(mlast, caddr_t) + mlast->m_len);
1027                 mlast->m_len += sizeof(struct ip6_frag);
1028                 m->m_pkthdr.len += sizeof(struct ip6_frag);
1029         } else {
1030                 /* allocate a new mbuf for the fragment header */
1031                 struct mbuf *mfrg;
1032
1033                 MGET(mfrg, M_NOWAIT, MT_DATA);
1034                 if (mfrg == NULL)
1035                         return (ENOBUFS);
1036                 mfrg->m_len = sizeof(struct ip6_frag);
1037                 *frghdrp = mtod(mfrg, struct ip6_frag *);
1038                 mlast->m_next = mfrg;
1039         }
1040
1041         return (0);
1042 }
1043
1044 static int
1045 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1046     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1047     int *alwaysfragp)
1048 {
1049         u_int32_t mtu = 0;
1050         int alwaysfrag = 0;
1051         int error = 0;
1052
1053         if (ro_pmtu != ro) {
1054                 /* The first hop and the final destination may differ. */
1055                 struct sockaddr_in6 *sa6_dst =
1056                     (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1057                 if (ro_pmtu->ro_rt &&
1058                     ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1059                      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1060                         RTFREE(ro_pmtu->ro_rt);
1061                         ro_pmtu->ro_rt = NULL;
1062                 }
1063                 if (ro_pmtu->ro_rt == NULL) {
1064                         bzero(sa6_dst, sizeof(*sa6_dst));
1065                         sa6_dst->sin6_family = AF_INET6;
1066                         sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1067                         sa6_dst->sin6_addr = *dst;
1068
1069                         rtalloc((struct route *)ro_pmtu);
1070                 }
1071         }
1072         if (ro_pmtu->ro_rt) {
1073                 u_int32_t ifmtu;
1074
1075                 if (ifp == NULL)
1076                         ifp = ro_pmtu->ro_rt->rt_ifp;
1077                 ifmtu = IN6_LINKMTU(ifp);
1078                 mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1079                 if (mtu == 0) {
1080                         mtu = ifmtu;
1081                 } else if (mtu < IPV6_MMTU) {
1082                         /*
1083                          * RFC2460 section 5, last paragraph:
1084                          * if we record ICMPv6 too big message with
1085                          * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1086                          * or smaller, with framgent header attached.
1087                          * (fragment header is needed regardless from the
1088                          * packet size, for translators to identify packets)
1089                          */
1090                         alwaysfrag = 1;
1091                         mtu = IPV6_MMTU;
1092                 } else if (mtu > ifmtu) {
1093                         /*
1094                          * The MTU on the route is larger than the MTU on
1095                          * the interface!  This shouldn't happen, unless the
1096                          * MTU of the interface has been changed after the
1097                          * interface was brought up.  Change the MTU in the
1098                          * route to match the interface MTU (as long as the
1099                          * field isn't locked).
1100                          */
1101                         mtu = ifmtu;
1102                         ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1103                 }
1104         } else if (ifp) {
1105                 mtu = IN6_LINKMTU(ifp);
1106         } else {
1107                 error = EHOSTUNREACH; /* XXX */
1108         }
1109
1110         *mtup = mtu;
1111         if (alwaysfragp)
1112                 *alwaysfragp = alwaysfrag;
1113         return (error);
1114 }
1115
1116 /*
1117  * IP6 socket option processing.
1118  */
1119 void
1120 ip6_ctloutput_dispatch(netmsg_t msg)
1121 {
1122         int error;
1123
1124         error = ip6_ctloutput(msg->ctloutput.base.nm_so,
1125                               msg->ctloutput.nm_sopt);
1126         lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
1127 }
1128
1129 int
1130 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1131 {
1132         int optdatalen,uproto;
1133         int privileged;
1134         struct inpcb *in6p = so->so_pcb;
1135         void *optdata;
1136         int error, optval;
1137         int level, op, optname;
1138         int optlen;
1139         struct thread *td;
1140
1141         if (sopt) {
1142                 level = sopt->sopt_level;
1143                 op = sopt->sopt_dir;
1144                 optname = sopt->sopt_name;
1145                 optlen = sopt->sopt_valsize;
1146                 td = sopt->sopt_td;
1147         } else {
1148                 panic("ip6_ctloutput: arg soopt is NULL");
1149                 /* NOT REACHED */
1150                 td = NULL;
1151         }
1152         error = optval = 0;
1153
1154         uproto = (int)so->so_proto->pr_protocol;
1155         privileged = (td == NULL || priv_check(td, PRIV_ROOT)) ? 0 : 1;
1156
1157         if (level == IPPROTO_IPV6) {
1158                 switch (op) {
1159
1160                 case SOPT_SET:
1161                         switch (optname) {
1162                         case IPV6_2292PKTOPTIONS:
1163 #ifdef IPV6_PKTOPTIONS
1164                         case IPV6_PKTOPTIONS:
1165 #endif
1166                         {
1167                                 struct mbuf *m;
1168
1169                                 error = soopt_getm(sopt, &m); /* XXX */
1170                                 if (error != 0)
1171                                         break;
1172                                 soopt_to_mbuf(sopt, m); /* XXX */
1173                                 error = ip6_pcbopts(&in6p->in6p_outputopts,
1174                                                     m, so, sopt);
1175                                 m_freem(m); /* XXX */
1176                                 break;
1177                         }
1178
1179                         /*
1180                          * Use of some Hop-by-Hop options or some
1181                          * Destination options, might require special
1182                          * privilege.  That is, normal applications
1183                          * (without special privilege) might be forbidden
1184                          * from setting certain options in outgoing packets,
1185                          * and might never see certain options in received
1186                          * packets. [RFC 2292 Section 6]
1187                          * KAME specific note:
1188                          *  KAME prevents non-privileged users from sending or
1189                          *  receiving ANY hbh/dst options in order to avoid
1190                          *  overhead of parsing options in the kernel.
1191                          */
1192                         case IPV6_RECVHOPOPTS:
1193                         case IPV6_RECVDSTOPTS:
1194                         case IPV6_RECVRTHDRDSTOPTS:
1195                                 if (!privileged)
1196                                         return (EPERM);
1197                         case IPV6_RECVPKTINFO:
1198                         case IPV6_RECVHOPLIMIT:
1199                         case IPV6_RECVRTHDR:
1200                         case IPV6_RECVPATHMTU:
1201                         case IPV6_RECVTCLASS:
1202                         case IPV6_AUTOFLOWLABEL:
1203                         case IPV6_HOPLIMIT:
1204                         /* FALLTHROUGH */
1205                         case IPV6_UNICAST_HOPS:
1206
1207                         case IPV6_V6ONLY:
1208                                 if (optlen != sizeof(int)) {
1209                                         error = EINVAL;
1210                                         break;
1211                                 }
1212                                 error = soopt_to_kbuf(sopt, &optval,
1213                                         sizeof optval, sizeof optval);
1214                                 if (error)
1215                                         break;
1216                                 switch (optname) {
1217
1218                                 case IPV6_UNICAST_HOPS:
1219                                         if (optval < -1 || optval >= 256)
1220                                                 error = EINVAL;
1221                                         else {
1222                                                 /* -1 = kernel default */
1223                                                 in6p->in6p_hops = optval;
1224                                         }
1225                                         break;
1226 #define OPTSET(bit) \
1227 do { \
1228         if (optval) \
1229                 in6p->in6p_flags |= (bit); \
1230         else \
1231                 in6p->in6p_flags &= ~(bit); \
1232 } while (0)
1233 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1234 /*
1235  * Although changed to RFC3542, It's better to also support RFC2292 API
1236  */
1237 #define OPTSET2292(bit) \
1238 do { \
1239         in6p->in6p_flags |= IN6P_RFC2292; \
1240         if (optval) \
1241                 in6p->in6p_flags |= (bit); \
1242         else \
1243                 in6p->in6p_flags &= ~(bit); \
1244 } while (/*CONSTCOND*/ 0)
1245
1246                                 case IPV6_RECVPKTINFO:
1247                                         /* cannot mix with RFC2292 */
1248                                         if (OPTBIT(IN6P_RFC2292)) {
1249                                                 error = EINVAL;
1250                                                 break;
1251                                         }
1252                                         OPTSET(IN6P_PKTINFO);
1253                                         break;
1254
1255                                 case IPV6_HOPLIMIT:
1256                                 {
1257                                         struct ip6_pktopts **optp;
1258
1259                                         /* cannot mix with RFC2292 */
1260                                         if (OPTBIT(IN6P_RFC2292)) {
1261                                                 error = EINVAL;
1262                                                 break;
1263                                         }
1264                                         optp = &in6p->in6p_outputopts;
1265                                         error = ip6_pcbopt(IPV6_HOPLIMIT,
1266                                             (u_char *)&optval, sizeof(optval),
1267                                             optp, uproto);
1268                                         break;
1269                                 }
1270
1271                                 case IPV6_RECVHOPLIMIT:
1272                                         /* cannot mix with RFC2292 */
1273                                         if (OPTBIT(IN6P_RFC2292)) {
1274                                                 error = EINVAL;
1275                                                 break;
1276                                         }
1277                                         OPTSET(IN6P_HOPLIMIT);
1278                                         break;
1279
1280                                 case IPV6_RECVHOPOPTS:
1281                                         /* cannot mix with RFC2292 */
1282                                         if (OPTBIT(IN6P_RFC2292)) {
1283                                                 error = EINVAL;
1284                                                 break;
1285                                         }
1286                                         OPTSET(IN6P_HOPOPTS);
1287                                         break;
1288
1289                                 case IPV6_RECVDSTOPTS:
1290                                         /* cannot mix with RFC2292 */
1291                                         if (OPTBIT(IN6P_RFC2292)) {
1292                                                 error = EINVAL;
1293                                                 break;
1294                                         }
1295                                         OPTSET(IN6P_DSTOPTS);
1296                                         break;
1297
1298                                 case IPV6_RECVRTHDRDSTOPTS:
1299                                         /* cannot mix with RFC2292 */
1300                                         if (OPTBIT(IN6P_RFC2292)) {
1301                                                 error = EINVAL;
1302                                                 break;
1303                                         }
1304                                         OPTSET(IN6P_RTHDRDSTOPTS);
1305                                         break;
1306
1307                                 case IPV6_RECVRTHDR:
1308                                         /* cannot mix with RFC2292 */
1309                                         if (OPTBIT(IN6P_RFC2292)) {
1310                                                 error = EINVAL;
1311                                                 break;
1312                                         }
1313                                         OPTSET(IN6P_RTHDR);
1314                                         break;
1315
1316                                 case IPV6_RECVPATHMTU:
1317                                         /*
1318                                          * We ignore this option for TCP
1319                                          * sockets.
1320                                          * (RFC3542 leaves this case
1321                                          * unspecified.)
1322                                          */
1323                                         if (uproto != IPPROTO_TCP)
1324                                                 OPTSET(IN6P_MTU);
1325                                         break;
1326
1327                                 case IPV6_RECVTCLASS:
1328                                         /* cannot mix with RFC2292 XXX */
1329                                         if (OPTBIT(IN6P_RFC2292)) {
1330                                                 error = EINVAL;
1331                                                 break;
1332                                         }
1333                                         OPTSET(IN6P_TCLASS);
1334                                         break;
1335
1336                                 case IPV6_AUTOFLOWLABEL:
1337                                         OPTSET(IN6P_AUTOFLOWLABEL);
1338                                         break;
1339
1340                                 case IPV6_V6ONLY:
1341                                         /*
1342                                          * make setsockopt(IPV6_V6ONLY)
1343                                          * available only prior to bind(2).
1344                                          */
1345                                         if (in6p->in6p_lport ||
1346                                             !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1347                                         {
1348                                                 error = EINVAL;
1349                                                 break;
1350                                         }
1351                                         if (!optval) {
1352                                                 /* Don't allow v4-mapped */
1353                                                 error = EOPNOTSUPP;
1354                                         }
1355                                         break;
1356                                 }
1357                                 break;
1358
1359                         case IPV6_TCLASS:
1360                         case IPV6_DONTFRAG:
1361                         case IPV6_USE_MIN_MTU:
1362                         case IPV6_PREFER_TEMPADDR:
1363                                 if (optlen != sizeof(optval)) {
1364                                         error = EINVAL;
1365                                         break;
1366                                 }
1367                                 error = soopt_to_kbuf(sopt, &optval,
1368                                         sizeof optval, sizeof optval);
1369                                 if (error)
1370                                         break;
1371                                 {
1372                                         struct ip6_pktopts **optp;
1373                                         optp = &in6p->in6p_outputopts;
1374                                         error = ip6_pcbopt(optname,
1375                                             (u_char *)&optval, sizeof(optval),
1376                                             optp, uproto);
1377                                         break;
1378                                 }
1379
1380                         case IPV6_2292PKTINFO:
1381                         case IPV6_2292HOPLIMIT:
1382                         case IPV6_2292HOPOPTS:
1383                         case IPV6_2292DSTOPTS:
1384                         case IPV6_2292RTHDR:
1385                                 /* RFC 2292 */
1386                                 if (optlen != sizeof(int)) {
1387                                         error = EINVAL;
1388                                         break;
1389                                 }
1390                                 error = soopt_to_kbuf(sopt, &optval,
1391                                         sizeof optval, sizeof optval);
1392                                 if (error)
1393                                         break;
1394                                 switch (optname) {
1395                                 case IPV6_2292PKTINFO:
1396                                         OPTSET2292(IN6P_PKTINFO);
1397                                         break;
1398                                 case IPV6_2292HOPLIMIT:
1399                                         OPTSET2292(IN6P_HOPLIMIT);
1400                                         break;
1401                                 case IPV6_2292HOPOPTS:
1402                                         /*
1403                                          * Check super-user privilege.
1404                                          * See comments for IPV6_RECVHOPOPTS.
1405                                          */
1406                                         if (!privileged)
1407                                                 return (EPERM);
1408                                         OPTSET2292(IN6P_HOPOPTS);
1409                                         break;
1410                                 case IPV6_2292DSTOPTS:
1411                                         if (!privileged)
1412                                                 return (EPERM);
1413                                         OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1414                                         break;
1415                                 case IPV6_2292RTHDR:
1416                                         OPTSET2292(IN6P_RTHDR);
1417                                         break;
1418                                 }
1419                                 break;
1420
1421                         case IPV6_PKTINFO:
1422                         case IPV6_HOPOPTS:
1423                         case IPV6_RTHDR:
1424                         case IPV6_DSTOPTS:
1425                         case IPV6_RTHDRDSTOPTS:
1426                         case IPV6_NEXTHOP:
1427                         {
1428                                 /*
1429                                  * New advanced API (RFC3542)
1430                                  */
1431                                 u_char *optbuf;
1432                                 u_char optbuf_storage[MCLBYTES];
1433                                 int optlen;
1434                                 struct ip6_pktopts **optp;
1435
1436                                 /* cannot mix with RFC2292 */
1437                                 if (OPTBIT(IN6P_RFC2292)) {
1438                                         error = EINVAL;
1439                                         break;
1440                                 }
1441
1442                                 /*
1443                                  * We only ensure valsize is not too large
1444                                  * here.  Further validation will be done
1445                                  * later.
1446                                  */
1447                                 error = soopt_to_kbuf(sopt, optbuf_storage,
1448                                     sizeof(optbuf_storage), 0);
1449                                 if (error)
1450                                         break;
1451                                 optlen = sopt->sopt_valsize;
1452                                 optbuf = optbuf_storage;
1453                                 optp = &in6p->in6p_outputopts;
1454                                 error = ip6_pcbopt(optname, optbuf, optlen,
1455                                     optp, uproto);
1456                                 break;
1457                         }
1458 #undef OPTSET
1459
1460                         case IPV6_MULTICAST_IF:
1461                         case IPV6_MULTICAST_HOPS:
1462                         case IPV6_MULTICAST_LOOP:
1463                         case IPV6_JOIN_GROUP:
1464                         case IPV6_LEAVE_GROUP:
1465                             {
1466                                 struct mbuf *m;
1467
1468                                 if (sopt->sopt_valsize > MLEN) {
1469                                         error = EMSGSIZE;
1470                                         break;
1471                                 }
1472                                 /* XXX */
1473                                 MGET(m, sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_HEADER);
1474                                 if (m == NULL) {
1475                                         error = ENOBUFS;
1476                                         break;
1477                                 }
1478                                 m->m_len = sopt->sopt_valsize;
1479                                 error = soopt_to_kbuf(sopt, mtod(m, char *),
1480                                                     m->m_len, m->m_len);
1481                                 error = ip6_setmoptions(sopt->sopt_name,
1482                                                         &in6p->in6p_moptions,
1483                                                         m);
1484                                 m_free(m);
1485                             }
1486                                 break;
1487
1488                         case IPV6_PORTRANGE:
1489                                 error = soopt_to_kbuf(sopt, &optval,
1490                                     sizeof optval, sizeof optval);
1491                                 if (error)
1492                                         break;
1493
1494                                 switch (optval) {
1495                                 case IPV6_PORTRANGE_DEFAULT:
1496                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1497                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1498                                         break;
1499
1500                                 case IPV6_PORTRANGE_HIGH:
1501                                         in6p->in6p_flags &= ~(IN6P_LOWPORT);
1502                                         in6p->in6p_flags |= IN6P_HIGHPORT;
1503                                         break;
1504
1505                                 case IPV6_PORTRANGE_LOW:
1506                                         in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1507                                         in6p->in6p_flags |= IN6P_LOWPORT;
1508                                         break;
1509
1510                                 default:
1511                                         error = EINVAL;
1512                                         break;
1513                                 }
1514                                 break;
1515
1516                         case IPV6_FW_ADD:
1517                         case IPV6_FW_DEL:
1518                         case IPV6_FW_FLUSH:
1519                         case IPV6_FW_ZERO:
1520                             {
1521                                 struct mbuf *m;
1522                                 struct mbuf **mp = &m;
1523
1524                                 if (ip6_fw_ctl_ptr == NULL)
1525                                         return EINVAL;
1526                                 /* XXX */
1527                                 if ((error = soopt_getm(sopt, &m)) != 0)
1528                                         break;
1529                                 /* XXX */
1530                                 soopt_to_mbuf(sopt, m);
1531                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1532                                 m = *mp;
1533                             }
1534                                 break;
1535
1536                         default:
1537                                 error = ENOPROTOOPT;
1538                                 break;
1539                         }
1540                         break;
1541
1542                 case SOPT_GET:
1543                         switch (optname) {
1544                         case IPV6_2292PKTOPTIONS:
1545 #ifdef IPV6_PKTOPTIONS
1546                         case IPV6_PKTOPTIONS:
1547 #endif
1548                                 /*
1549                                  * RFC3542 (effectively) deprecated the
1550                                  * semantics of the 2292-style pktoptions.
1551                                  * Since it was not reliable in nature (i.e.,
1552                                  * applications had to expect the lack of some
1553                                  * information after all), it would make sense
1554                                  * to simplify this part by always returning
1555                                  * empty data.
1556                                  */
1557                                 if (in6p->in6p_options) {
1558                                         struct mbuf *m;
1559                                         m = m_copym(in6p->in6p_options,
1560                                             0, M_COPYALL, M_WAITOK);
1561                                         error = soopt_from_mbuf(sopt, m);
1562                                         if (error == 0)
1563                                                 m_freem(m);
1564                                 } else
1565                                         sopt->sopt_valsize = 0;
1566                                 break;
1567
1568                         case IPV6_RECVHOPOPTS:
1569                         case IPV6_RECVDSTOPTS:
1570                         case IPV6_RECVRTHDRDSTOPTS:
1571                         case IPV6_UNICAST_HOPS:
1572                         case IPV6_RECVPKTINFO:
1573                         case IPV6_RECVHOPLIMIT:
1574                         case IPV6_RECVRTHDR:
1575                         case IPV6_RECVPATHMTU:
1576                         case IPV6_RECVTCLASS:
1577                         case IPV6_AUTOFLOWLABEL:
1578                         case IPV6_V6ONLY:
1579                         case IPV6_PORTRANGE:
1580                                 switch (optname) {
1581
1582                                 case IPV6_RECVHOPOPTS:
1583                                         optval = OPTBIT(IN6P_HOPOPTS);
1584                                         break;
1585
1586                                 case IPV6_RECVDSTOPTS:
1587                                         optval = OPTBIT(IN6P_DSTOPTS);
1588                                         break;
1589
1590                                 case IPV6_RECVRTHDRDSTOPTS:
1591                                         optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1592                                         break;
1593
1594                                 case IPV6_RECVPKTINFO:
1595                                         optval = OPTBIT(IN6P_PKTINFO);
1596                                         break;
1597
1598                                 case IPV6_RECVHOPLIMIT:
1599                                         optval = OPTBIT(IN6P_HOPLIMIT);
1600                                         break;
1601
1602                                 case IPV6_RECVRTHDR:
1603                                         optval = OPTBIT(IN6P_RTHDR);
1604                                         break;
1605
1606                                 case IPV6_RECVPATHMTU:
1607                                         optval = OPTBIT(IN6P_MTU);
1608                                         break;
1609
1610                                 case IPV6_RECVTCLASS:
1611                                         optval = OPTBIT(IN6P_TCLASS);
1612                                         break;
1613
1614                                 case IPV6_AUTOFLOWLABEL:
1615                                         optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1616                                         break;
1617
1618
1619                                 case IPV6_UNICAST_HOPS:
1620                                         optval = in6p->in6p_hops;
1621                                         break;
1622
1623                                 case IPV6_V6ONLY:
1624                                         optval = 1;
1625                                         break;
1626
1627                                 case IPV6_PORTRANGE:
1628                                     {
1629                                         int flags;
1630                                         flags = in6p->in6p_flags;
1631                                         if (flags & IN6P_HIGHPORT)
1632                                                 optval = IPV6_PORTRANGE_HIGH;
1633                                         else if (flags & IN6P_LOWPORT)
1634                                                 optval = IPV6_PORTRANGE_LOW;
1635                                         else
1636                                                 optval = 0;
1637                                         break;
1638                                     }
1639                                 }
1640                                 soopt_from_kbuf(sopt, &optval,
1641                                         sizeof optval);
1642                                 break;
1643
1644                         case IPV6_PATHMTU:
1645                         {
1646                                 u_long pmtu = 0;
1647                                 struct ip6_mtuinfo mtuinfo;
1648                                 struct route_in6 sro;
1649
1650                                 bzero(&sro, sizeof(sro));
1651
1652                                 if (!(so->so_state & SS_ISCONNECTED))
1653                                         return (ENOTCONN);
1654                                 /*
1655                                  * XXX: we dot not consider the case of source
1656                                  * routing, or optional information to specify
1657                                  * the outgoing interface.
1658                                  */
1659                                 error = ip6_getpmtu(&sro, NULL, NULL,
1660                                     &in6p->in6p_faddr, &pmtu, NULL);
1661                                 if (sro.ro_rt)
1662                                         RTFREE(sro.ro_rt);
1663                                 if (error)
1664                                         break;
1665                                 if (pmtu > IPV6_MAXPACKET)
1666                                         pmtu = IPV6_MAXPACKET;
1667
1668                                 bzero(&mtuinfo, sizeof(mtuinfo));
1669                                 mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1670                                 optdata = (void *)&mtuinfo;
1671                                 optdatalen = sizeof(mtuinfo);
1672                                 soopt_from_kbuf(sopt, optdata,
1673                                     optdatalen);
1674                                 break;
1675                         }
1676
1677                         case IPV6_2292PKTINFO:
1678                         case IPV6_2292HOPLIMIT:
1679                         case IPV6_2292HOPOPTS:
1680                         case IPV6_2292RTHDR:
1681                         case IPV6_2292DSTOPTS:
1682                                 if (optname == IPV6_2292HOPOPTS ||
1683                                     optname == IPV6_2292DSTOPTS ||
1684                                     !privileged)
1685                                         return (EPERM);
1686                                 switch (optname) {
1687                                 case IPV6_2292PKTINFO:
1688                                         optval = OPTBIT(IN6P_PKTINFO);
1689                                         break;
1690                                 case IPV6_2292HOPLIMIT:
1691                                         optval = OPTBIT(IN6P_HOPLIMIT);
1692                                         break;
1693                                 case IPV6_2292HOPOPTS:
1694                                         if (!privileged)
1695                                                 return (EPERM);
1696                                         optval = OPTBIT(IN6P_HOPOPTS);
1697                                         break;
1698                                 case IPV6_2292RTHDR:
1699                                         optval = OPTBIT(IN6P_RTHDR);
1700                                         break;
1701                                 case IPV6_2292DSTOPTS:
1702                                         if (!privileged)
1703                                                 return (EPERM);
1704                                         optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1705                                         break;
1706                                 }
1707                                 soopt_from_kbuf(sopt, &optval,
1708                                         sizeof optval);
1709                                 break;
1710
1711                         case IPV6_PKTINFO:
1712                         case IPV6_HOPOPTS:
1713                         case IPV6_RTHDR:
1714                         case IPV6_DSTOPTS:
1715                         case IPV6_RTHDRDSTOPTS:
1716                         case IPV6_NEXTHOP:
1717                         case IPV6_TCLASS:
1718                         case IPV6_DONTFRAG:
1719                         case IPV6_USE_MIN_MTU:
1720                         case IPV6_PREFER_TEMPADDR:
1721                                 error = ip6_getpcbopt(in6p->in6p_outputopts,
1722                                     optname, sopt);
1723                                 break;
1724
1725                         case IPV6_MULTICAST_IF:
1726                         case IPV6_MULTICAST_HOPS:
1727                         case IPV6_MULTICAST_LOOP:
1728                         case IPV6_JOIN_GROUP:
1729                         case IPV6_LEAVE_GROUP:
1730                             {
1731                                 struct mbuf *m;
1732                                 error = ip6_getmoptions(sopt->sopt_name,
1733                                     in6p->in6p_moptions, &m);
1734                                 if (error == 0) {
1735                                         soopt_from_kbuf(sopt,
1736                                             mtod(m, char *), m->m_len);
1737                                 }
1738                                 m_freem(m);
1739                             }
1740                                 break;
1741
1742                         case IPV6_FW_GET:
1743                           {
1744                                 struct mbuf *m;
1745                                 struct mbuf **mp = &m;
1746
1747                                 if (ip6_fw_ctl_ptr == NULL)
1748                                 {
1749                                         return EINVAL;
1750                                 }
1751                                 error = (*ip6_fw_ctl_ptr)(optname, mp);
1752                                 if (error == 0)
1753                                         error = soopt_from_mbuf(sopt, m); /* XXX */
1754                                 if (error == 0 && m != NULL)
1755                                         m_freem(m);
1756                           }
1757                                 break;
1758
1759                         default:
1760                                 error = ENOPROTOOPT;
1761                                 break;
1762                         }
1763                         break;
1764                 }
1765         } else {
1766                 error = EINVAL;
1767         }
1768         return (error);
1769 }
1770
1771 int
1772 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
1773 {
1774         int error = 0, optval, optlen;
1775         const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1776         struct in6pcb *in6p = sotoin6pcb(so);
1777         int level, op, optname;
1778
1779         if (sopt) {
1780                 level = sopt->sopt_level;
1781                 op = sopt->sopt_dir;
1782                 optname = sopt->sopt_name;
1783                 optlen = sopt->sopt_valsize;
1784         } else
1785                 panic("ip6_raw_ctloutput: arg soopt is NULL");
1786
1787         if (level != IPPROTO_IPV6) {
1788                 return (EINVAL);
1789         }
1790
1791         switch (optname) {
1792         case IPV6_CHECKSUM:
1793                 /*
1794                  * For ICMPv6 sockets, no modification allowed for checksum
1795                  * offset, permit "no change" values to help existing apps.
1796                  *
1797                  * RFC3542 says: "An attempt to set IPV6_CHECKSUM
1798                  * for an ICMPv6 socket will fail."
1799                  * The current behavior does not meet RFC3542.
1800                  */
1801                 switch (op) {
1802                 case SOPT_SET:
1803                         if (optlen != sizeof(int)) {
1804                                 error = EINVAL;
1805                                 break;
1806                         }
1807                         error = soopt_to_kbuf(sopt, &optval,
1808                                     sizeof optval, sizeof optval);
1809                         if (error)
1810                                 break;
1811                         if ((optval % 2) != 0) {
1812                                 /* the API assumes even offset values */
1813                                 error = EINVAL;
1814                         } else if (so->so_proto->pr_protocol ==
1815                             IPPROTO_ICMPV6) {
1816                                 if (optval != icmp6off)
1817                                         error = EINVAL;
1818                         } else
1819                                 in6p->in6p_cksum = optval;
1820                         break;
1821
1822                 case SOPT_GET:
1823                         if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1824                                 optval = icmp6off;
1825                         else
1826                                 optval = in6p->in6p_cksum;
1827
1828                         soopt_from_kbuf(sopt, &optval, sizeof(optval));
1829                         break;
1830
1831                 default:
1832                         error = EINVAL;
1833                         break;
1834                 }
1835                 break;
1836
1837         default:
1838                 error = ENOPROTOOPT;
1839                 break;
1840         }
1841
1842         return (error);
1843 }
1844
1845 /*
1846  * Set up IP6 options in pcb for insertion in output packets or
1847  * specifying behavior of outgoing packets.
1848  */
1849 static int
1850 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
1851     struct socket *so, struct sockopt *sopt)
1852 {
1853         int priv = 0;
1854         struct ip6_pktopts *opt = *pktopt;
1855         int error = 0;
1856
1857         /* turn off any old options. */
1858         if (opt) {
1859 #ifdef DIAGNOSTIC
1860                 if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1861                     opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1862                     opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1863                         kprintf("ip6_pcbopts: all specified options are cleared.\n");
1864 #endif
1865                 ip6_clearpktopts(opt, -1);
1866         } else
1867                 opt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1868         *pktopt = NULL;
1869
1870         if (!m || m->m_len == 0) {
1871                 /*
1872                  * Only turning off any previous options, regardless of
1873                  * whether the opt is just created or given.
1874                  */
1875                 kfree(opt, M_IP6OPT);
1876                 return (0);
1877         }
1878
1879         /*  set options specified by user. */
1880         if ((error = ip6_setpktoptions(m, opt, NULL, so->so_proto->pr_protocol, priv)) != 0) {
1881                 ip6_clearpktopts(opt, -1); /* XXX: discard all options */
1882                 kfree(opt, M_IP6OPT);
1883                 return (error);
1884         }
1885         *pktopt = opt;
1886         return (0);
1887 }
1888
1889
1890 /*
1891  * Below three functions are introduced by merge to RFC3542
1892  */
1893
1894 static int
1895 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
1896 {
1897         void *optdata = NULL;
1898         int optdatalen = 0;
1899         struct ip6_ext *ip6e;
1900         int error = 0;
1901         struct in6_pktinfo null_pktinfo;
1902         int deftclass = 0, on;
1903         int defminmtu = IP6PO_MINMTU_MCASTONLY;
1904         int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
1905
1906         switch (optname) {
1907         case IPV6_PKTINFO:
1908                 if (pktopt && pktopt->ip6po_pktinfo)
1909                         optdata = (void *)pktopt->ip6po_pktinfo;
1910                 else {
1911                         /* XXX: we don't have to do this every time... */
1912                         bzero(&null_pktinfo, sizeof(null_pktinfo));
1913                         optdata = (void *)&null_pktinfo;
1914                 }
1915                 optdatalen = sizeof(struct in6_pktinfo);
1916                 break;
1917         case IPV6_TCLASS:
1918                 if (pktopt && pktopt->ip6po_tclass >= 0)
1919                         optdata = (void *)&pktopt->ip6po_tclass;
1920                 else
1921                         optdata = (void *)&deftclass;
1922                 optdatalen = sizeof(int);
1923                 break;
1924         case IPV6_HOPOPTS:
1925                 if (pktopt && pktopt->ip6po_hbh) {
1926                         optdata = (void *)pktopt->ip6po_hbh;
1927                         ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
1928                         optdatalen = (ip6e->ip6e_len + 1) << 3;
1929                 }
1930                 break;
1931         case IPV6_RTHDR:
1932                 if (pktopt && pktopt->ip6po_rthdr) {
1933                         optdata = (void *)pktopt->ip6po_rthdr;
1934                         ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
1935                         optdatalen = (ip6e->ip6e_len + 1) << 3;
1936                 }
1937                 break;
1938         case IPV6_RTHDRDSTOPTS:
1939                 if (pktopt && pktopt->ip6po_dest1) {
1940                         optdata = (void *)pktopt->ip6po_dest1;
1941                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
1942                         optdatalen = (ip6e->ip6e_len + 1) << 3;
1943                 }
1944                 break;
1945         case IPV6_DSTOPTS:
1946                 if (pktopt && pktopt->ip6po_dest2) {
1947                         optdata = (void *)pktopt->ip6po_dest2;
1948                         ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
1949                         optdatalen = (ip6e->ip6e_len + 1) << 3;
1950                 }
1951                 break;
1952         case IPV6_NEXTHOP:
1953                 if (pktopt && pktopt->ip6po_nexthop) {
1954                         optdata = (void *)pktopt->ip6po_nexthop;
1955                         optdatalen = pktopt->ip6po_nexthop->sa_len;
1956                 }
1957                 break;
1958         case IPV6_USE_MIN_MTU:
1959                 if (pktopt)
1960                         optdata = (void *)&pktopt->ip6po_minmtu;
1961                 else
1962                         optdata = (void *)&defminmtu;
1963                 optdatalen = sizeof(int);
1964                 break;
1965         case IPV6_DONTFRAG:
1966                 if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
1967                         on = 1;
1968                 else
1969                         on = 0;
1970                 optdata = (void *)&on;
1971                 optdatalen = sizeof(on);
1972                 break;
1973         case IPV6_PREFER_TEMPADDR:
1974                 if (pktopt)
1975                         optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
1976                 else
1977                         optdata = (void *)&defpreftemp;
1978                 optdatalen = sizeof(int);
1979                 break;
1980         default:                /* should not happen */
1981 #ifdef DIAGNOSTIC
1982                 panic("ip6_getpcbopt: unexpected option");
1983 #endif
1984                 return (ENOPROTOOPT);
1985         }
1986
1987         soopt_from_kbuf(sopt, optdata, optdatalen);
1988
1989         return (error);
1990 }
1991
1992 /*
1993  * initialize ip6_pktopts.  beware that there are non-zero default values in
1994  * the struct.
1995  */
1996
1997 static int
1998 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt, int uproto)
1999 {
2000         struct ip6_pktopts *opt;
2001         int priv =0;
2002         if (*pktopt == NULL) {
2003                 *pktopt = kmalloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2004                 init_ip6pktopts(*pktopt);
2005         }
2006         opt = *pktopt;
2007
2008         return (ip6_setpktoption(optname, buf, len, opt, 1, 0, uproto, priv));
2009 }
2010
2011 /*
2012  * initialize ip6_pktopts.  beware that there are non-zero default values in
2013  * the struct.
2014  */
2015 void
2016 init_ip6pktopts(struct ip6_pktopts *opt)
2017 {
2018
2019         bzero(opt, sizeof(*opt));
2020         opt->ip6po_hlim = -1;   /* -1 means default hop limit */
2021         opt->ip6po_tclass = -1; /* -1 means default traffic class */
2022         opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2023         opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2024 }
2025
2026 void
2027 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2028 {
2029         if (pktopt == NULL)
2030                 return;
2031
2032         if (optname == -1 || optname == IPV6_PKTINFO) {
2033                 if (pktopt->ip6po_pktinfo)
2034                         kfree(pktopt->ip6po_pktinfo, M_IP6OPT);
2035                 pktopt->ip6po_pktinfo = NULL;
2036         }
2037         if (optname == -1 || optname == IPV6_HOPLIMIT)
2038                 pktopt->ip6po_hlim = -1;
2039         if (optname == -1 || optname == IPV6_TCLASS)
2040                 pktopt->ip6po_tclass = -1;
2041         if (optname == -1 || optname == IPV6_NEXTHOP) {
2042                 if (pktopt->ip6po_nextroute.ro_rt) {
2043                         RTFREE(pktopt->ip6po_nextroute.ro_rt);
2044                         pktopt->ip6po_nextroute.ro_rt = NULL;
2045                 }
2046                 if (pktopt->ip6po_nexthop)
2047                         kfree(pktopt->ip6po_nexthop, M_IP6OPT);
2048                 pktopt->ip6po_nexthop = NULL;
2049         }
2050         if (optname == -1 || optname == IPV6_HOPOPTS) {
2051                 if (pktopt->ip6po_hbh)
2052                         kfree(pktopt->ip6po_hbh, M_IP6OPT);
2053                 pktopt->ip6po_hbh = NULL;
2054         }
2055         if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2056                 if (pktopt->ip6po_dest1)
2057                         kfree(pktopt->ip6po_dest1, M_IP6OPT);
2058                 pktopt->ip6po_dest1 = NULL;
2059         }
2060         if (optname == -1 || optname == IPV6_RTHDR) {
2061                 if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2062                         kfree(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2063                 pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2064                 if (pktopt->ip6po_route.ro_rt) {
2065                         RTFREE(pktopt->ip6po_route.ro_rt);
2066                         pktopt->ip6po_route.ro_rt = NULL;
2067                 }
2068         }
2069         if (optname == -1 || optname == IPV6_DSTOPTS) {
2070                 if (pktopt->ip6po_dest2)
2071                         kfree(pktopt->ip6po_dest2, M_IP6OPT);
2072                 pktopt->ip6po_dest2 = NULL;
2073         }
2074 }
2075
2076 #define PKTOPT_EXTHDRCPY(type) \
2077 do {\
2078         if (src->type) {\
2079                 int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2080                 dst->type = kmalloc(hlen, M_IP6OPT, canwait);\
2081                 if (dst->type == NULL)\
2082                         goto bad;\
2083                 bcopy(src->type, dst->type, hlen);\
2084         }\
2085 } while (0)
2086
2087 struct ip6_pktopts *
2088 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2089 {
2090         struct ip6_pktopts *dst;
2091
2092         if (src == NULL) {
2093                 kprintf("ip6_clearpktopts: invalid argument\n");
2094                 return (NULL);
2095         }
2096
2097         dst = kmalloc(sizeof(*dst), M_IP6OPT, canwait | M_ZERO);
2098         if (dst == NULL)
2099                 return (NULL);
2100
2101         dst->ip6po_hlim = src->ip6po_hlim;
2102         if (src->ip6po_pktinfo) {
2103                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2104                     M_IP6OPT, canwait);
2105                 if (dst->ip6po_pktinfo == NULL)
2106                         goto bad;
2107                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2108         }
2109         if (src->ip6po_nexthop) {
2110                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2111                     M_IP6OPT, canwait);
2112                 if (dst->ip6po_nexthop == NULL)
2113                         goto bad;
2114                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2115                     src->ip6po_nexthop->sa_len);
2116         }
2117         PKTOPT_EXTHDRCPY(ip6po_hbh);
2118         PKTOPT_EXTHDRCPY(ip6po_dest1);
2119         PKTOPT_EXTHDRCPY(ip6po_dest2);
2120         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2121         return (dst);
2122
2123 bad:
2124         if (dst->ip6po_pktinfo) kfree(dst->ip6po_pktinfo, M_IP6OPT);
2125         if (dst->ip6po_nexthop) kfree(dst->ip6po_nexthop, M_IP6OPT);
2126         if (dst->ip6po_hbh) kfree(dst->ip6po_hbh, M_IP6OPT);
2127         if (dst->ip6po_dest1) kfree(dst->ip6po_dest1, M_IP6OPT);
2128         if (dst->ip6po_dest2) kfree(dst->ip6po_dest2, M_IP6OPT);
2129         if (dst->ip6po_rthdr) kfree(dst->ip6po_rthdr, M_IP6OPT);
2130         kfree(dst, M_IP6OPT);
2131         return (NULL);
2132 }
2133
2134 static int
2135 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2136 {
2137         if (dst == NULL || src == NULL)  {
2138 #ifdef DIAGNOSTIC
2139                 kprintf("ip6_clearpktopts: invalid argument\n");
2140 #endif
2141                 return (EINVAL);
2142         }
2143
2144         dst->ip6po_hlim = src->ip6po_hlim;
2145         dst->ip6po_tclass = src->ip6po_tclass;
2146         dst->ip6po_flags = src->ip6po_flags;
2147         if (src->ip6po_pktinfo) {
2148                 dst->ip6po_pktinfo = kmalloc(sizeof(*dst->ip6po_pktinfo),
2149                     M_IP6OPT, canwait);
2150                 if (dst->ip6po_pktinfo == NULL)
2151                         goto bad;
2152                 *dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2153         }
2154         if (src->ip6po_nexthop) {
2155                 dst->ip6po_nexthop = kmalloc(src->ip6po_nexthop->sa_len,
2156                     M_IP6OPT, canwait);
2157                 if (dst->ip6po_nexthop == NULL)
2158                         goto bad;
2159                 bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2160                     src->ip6po_nexthop->sa_len);
2161         }
2162         PKTOPT_EXTHDRCPY(ip6po_hbh);
2163         PKTOPT_EXTHDRCPY(ip6po_dest1);
2164         PKTOPT_EXTHDRCPY(ip6po_dest2);
2165         PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2166         return (0);
2167
2168   bad:
2169         ip6_clearpktopts(dst, -1);
2170         return (ENOBUFS);
2171 }
2172 #undef PKTOPT_EXTHDRCPY
2173
2174 void
2175 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2176 {
2177         if (pktopt == NULL)
2178                 return;
2179
2180         ip6_clearpktopts(pktopt, -1);
2181
2182         kfree(pktopt, M_IP6OPT);
2183 }
2184
2185 /*
2186  * Set the IP6 multicast options in response to user setsockopt().
2187  */
2188 static int
2189 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2190 {
2191         int error = 0;
2192         u_int loop, ifindex;
2193         struct ipv6_mreq *mreq;
2194         struct ifnet *ifp;
2195         struct ip6_moptions *im6o = *im6op;
2196         struct route_in6 ro;
2197         struct sockaddr_in6 *dst;
2198         struct in6_multi_mship *imm;
2199         struct thread *td = curthread;
2200
2201         if (im6o == NULL) {
2202                 /*
2203                  * No multicast option buffer attached to the pcb;
2204                  * allocate one and initialize to default values.
2205                  */
2206                 im6o = (struct ip6_moptions *)
2207                         kmalloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2208
2209                 *im6op = im6o;
2210                 im6o->im6o_multicast_ifp = NULL;
2211                 im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2212                 im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2213                 LIST_INIT(&im6o->im6o_memberships);
2214         }
2215
2216         switch (optname) {
2217
2218         case IPV6_MULTICAST_IF:
2219                 /*
2220                  * Select the interface for outgoing multicast packets.
2221                  */
2222                 if (m == NULL || m->m_len != sizeof(u_int)) {
2223                         error = EINVAL;
2224                         break;
2225                 }
2226                 bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2227                 if (ifindex < 0 || if_index < ifindex) {
2228                         error = ENXIO;  /* XXX EINVAL? */
2229                         break;
2230                 }
2231                 ifp = ifindex2ifnet[ifindex];
2232                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2233                         error = EADDRNOTAVAIL;
2234                         break;
2235                 }
2236                 im6o->im6o_multicast_ifp = ifp;
2237                 break;
2238
2239         case IPV6_MULTICAST_HOPS:
2240             {
2241                 /*
2242                  * Set the IP6 hoplimit for outgoing multicast packets.
2243                  */
2244                 int optval;
2245                 if (m == NULL || m->m_len != sizeof(int)) {
2246                         error = EINVAL;
2247                         break;
2248                 }
2249                 bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2250                 if (optval < -1 || optval >= 256)
2251                         error = EINVAL;
2252                 else if (optval == -1)
2253                         im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2254                 else
2255                         im6o->im6o_multicast_hlim = optval;
2256                 break;
2257             }
2258
2259         case IPV6_MULTICAST_LOOP:
2260                 /*
2261                  * Set the loopback flag for outgoing multicast packets.
2262                  * Must be zero or one.
2263                  */
2264                 if (m == NULL || m->m_len != sizeof(u_int)) {
2265                         error = EINVAL;
2266                         break;
2267                 }
2268                 bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2269                 if (loop > 1) {
2270                         error = EINVAL;
2271                         break;
2272                 }
2273                 im6o->im6o_multicast_loop = loop;
2274                 break;
2275
2276         case IPV6_JOIN_GROUP:
2277                 /*
2278                  * Add a multicast group membership.
2279                  * Group must be a valid IP6 multicast address.
2280                  */
2281                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2282                         error = EINVAL;
2283                         break;
2284                 }
2285                 mreq = mtod(m, struct ipv6_mreq *);
2286                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2287                         /*
2288                          * We use the unspecified address to specify to accept
2289                          * all multicast addresses. Only super user is allowed
2290                          * to do this.
2291                          */
2292                         if (priv_check(td, PRIV_ROOT)) {
2293                                 error = EACCES;
2294                                 break;
2295                         }
2296                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2297                         error = EINVAL;
2298                         break;
2299                 }
2300
2301                 /*
2302                  * If the interface is specified, validate it.
2303                  */
2304                 if (mreq->ipv6mr_interface < 0
2305                  || if_index < mreq->ipv6mr_interface) {
2306                         error = ENXIO;  /* XXX EINVAL? */
2307                         break;
2308                 }
2309                 /*
2310                  * If no interface was explicitly specified, choose an
2311                  * appropriate one according to the given multicast address.
2312                  */
2313                 if (mreq->ipv6mr_interface == 0) {
2314                         /*
2315                          * If the multicast address is in node-local scope,
2316                          * the interface should be a loopback interface.
2317                          * Otherwise, look up the routing table for the
2318                          * address, and choose the outgoing interface.
2319                          *   XXX: is it a good approach?
2320                          */
2321                         if (IN6_IS_ADDR_MC_INTFACELOCAL(&mreq->ipv6mr_multiaddr)) {
2322                                 ifp = loif;
2323                         } else {
2324                                 ro.ro_rt = NULL;
2325                                 dst = (struct sockaddr_in6 *)&ro.ro_dst;
2326                                 bzero(dst, sizeof(*dst));
2327                                 dst->sin6_len = sizeof(struct sockaddr_in6);
2328                                 dst->sin6_family = AF_INET6;
2329                                 dst->sin6_addr = mreq->ipv6mr_multiaddr;
2330                                 rtalloc((struct route *)&ro);
2331                                 if (ro.ro_rt == NULL) {
2332                                         error = EADDRNOTAVAIL;
2333                                         break;
2334                                 }
2335                                 ifp = ro.ro_rt->rt_ifp;
2336                                 rtfree(ro.ro_rt);
2337                         }
2338                 } else
2339                         ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2340
2341                 /*
2342                  * See if we found an interface, and confirm that it
2343                  * supports multicast
2344                  */
2345                 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) {
2346                         error = EADDRNOTAVAIL;
2347                         break;
2348                 }
2349                 /*
2350                  * Put interface index into the multicast address,
2351                  * if the address has link-local scope.
2352                  */
2353                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2354                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2355                                 = htons(mreq->ipv6mr_interface);
2356                 }
2357                 /*
2358                  * See if the membership already exists.
2359                  */
2360                 for (imm = im6o->im6o_memberships.lh_first;
2361                      imm != NULL; imm = imm->i6mm_chain.le_next)
2362                         if (imm->i6mm_maddr->in6m_ifp == ifp &&
2363                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2364                                                &mreq->ipv6mr_multiaddr))
2365                                 break;
2366                 if (imm != NULL) {
2367                         error = EADDRINUSE;
2368                         break;
2369                 }
2370                 /*
2371                  * Everything looks good; add a new record to the multicast
2372                  * address list for the given interface.
2373                  */
2374                 imm = kmalloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2375                 if ((imm->i6mm_maddr =
2376                      in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2377                         kfree(imm, M_IPMADDR);
2378                         break;
2379                 }
2380                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2381                 break;
2382
2383         case IPV6_LEAVE_GROUP:
2384                 /*
2385                  * Drop a multicast group membership.
2386                  * Group must be a valid IP6 multicast address.
2387                  */
2388                 if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2389                         error = EINVAL;
2390                         break;
2391                 }
2392                 mreq = mtod(m, struct ipv6_mreq *);
2393                 if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2394                         if (priv_check(td, PRIV_ROOT)) {
2395                                 error = EACCES;
2396                                 break;
2397                         }
2398                 } else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2399                         error = EINVAL;
2400                         break;
2401                 }
2402                 /*
2403                  * If an interface address was specified, get a pointer
2404                  * to its ifnet structure.
2405                  */
2406                 if (mreq->ipv6mr_interface < 0
2407                  || if_index < mreq->ipv6mr_interface) {
2408                         error = ENXIO;  /* XXX EINVAL? */
2409                         break;
2410                 }
2411                 ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2412                 /*
2413                  * Put interface index into the multicast address,
2414                  * if the address has link-local scope.
2415                  */
2416                 if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2417                         mreq->ipv6mr_multiaddr.s6_addr16[1]
2418                                 = htons(mreq->ipv6mr_interface);
2419                 }
2420
2421                 /*
2422                  * Find the membership in the membership list.
2423                  */
2424                 for (imm = im6o->im6o_memberships.lh_first;
2425                      imm != NULL; imm = imm->i6mm_chain.le_next) {
2426                         if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2427                             IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2428                             &mreq->ipv6mr_multiaddr))
2429                                 break;
2430                 }
2431                 if (imm == NULL) {
2432                         /* Unable to resolve interface */
2433                         error = EADDRNOTAVAIL;
2434                         break;
2435                 }
2436                 /*
2437                  * Give up the multicast address record to which the
2438                  * membership points.
2439                  */
2440                 LIST_REMOVE(imm, i6mm_chain);
2441                 in6_delmulti(imm->i6mm_maddr);
2442                 kfree(imm, M_IPMADDR);
2443                 break;
2444
2445         default:
2446                 error = EOPNOTSUPP;
2447                 break;
2448         }
2449
2450         /*
2451          * If all options have default values, no need to keep the mbuf.
2452          */
2453         if (im6o->im6o_multicast_ifp == NULL &&
2454             im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2455             im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2456             im6o->im6o_memberships.lh_first == NULL) {
2457                 kfree(*im6op, M_IPMOPTS);
2458                 *im6op = NULL;
2459         }
2460
2461         return (error);
2462 }
2463
2464 /*
2465  * Return the IP6 multicast options in response to user getsockopt().
2466  */
2467 static int
2468 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2469 {
2470         u_int *hlim, *loop, *ifindex;
2471
2472         *mp = m_get(M_WAITOK, MT_HEADER);               /* XXX */
2473
2474         switch (optname) {
2475
2476         case IPV6_MULTICAST_IF:
2477                 ifindex = mtod(*mp, u_int *);
2478                 (*mp)->m_len = sizeof(u_int);
2479                 if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2480                         *ifindex = 0;
2481                 else
2482                         *ifindex = im6o->im6o_multicast_ifp->if_index;
2483                 return (0);
2484
2485         case IPV6_MULTICAST_HOPS:
2486                 hlim = mtod(*mp, u_int *);
2487                 (*mp)->m_len = sizeof(u_int);
2488                 if (im6o == NULL)
2489                         *hlim = ip6_defmcasthlim;
2490                 else
2491                         *hlim = im6o->im6o_multicast_hlim;
2492                 return (0);
2493
2494         case IPV6_MULTICAST_LOOP:
2495                 loop = mtod(*mp, u_int *);
2496                 (*mp)->m_len = sizeof(u_int);
2497                 if (im6o == NULL)
2498                         *loop = ip6_defmcasthlim;
2499                 else
2500                         *loop = im6o->im6o_multicast_loop;
2501                 return (0);
2502
2503         default:
2504                 return (EOPNOTSUPP);
2505         }
2506 }
2507
2508 /*
2509  * Discard the IP6 multicast options.
2510  */
2511 void
2512 ip6_freemoptions(struct ip6_moptions *im6o)
2513 {
2514         struct in6_multi_mship *imm;
2515
2516         if (im6o == NULL)
2517                 return;
2518
2519         while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2520                 LIST_REMOVE(imm, i6mm_chain);
2521                 if (imm->i6mm_maddr)
2522                         in6_delmulti(imm->i6mm_maddr);
2523                 kfree(imm, M_IPMADDR);
2524         }
2525         kfree(im6o, M_IPMOPTS);
2526 }
2527
2528 /*
2529  * Set a particular packet option, as a sticky option or an ancillary data
2530  * item.  "len" can be 0 only when it's a sticky option.
2531  * We have 4 cases of combination of "sticky" and "cmsg":
2532  * "sticky=0, cmsg=0": impossible
2533  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2534  * "sticky=1, cmsg=0": RFC3542 socket option
2535  * "sticky=1, cmsg=1": RFC2292 socket option
2536  */
2537 static int
2538 ip6_setpktoption(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2539      int sticky, int cmsg, int uproto, int priv)
2540 {
2541         int minmtupolicy, preftemp;
2542         //int error;
2543
2544         if (!sticky && !cmsg) {
2545                 kprintf("ip6_setpktoption: impossible case\n");
2546                 return (EINVAL);
2547         }
2548
2549         /*
2550          * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2551          * not be specified in the context of RFC3542.  Conversely,
2552          * RFC3542 types should not be specified in the context of RFC2292.
2553          */
2554         if (!cmsg) {
2555                 switch (optname) {
2556                 case IPV6_2292PKTINFO:
2557                 case IPV6_2292HOPLIMIT:
2558                 case IPV6_2292NEXTHOP:
2559                 case IPV6_2292HOPOPTS:
2560                 case IPV6_2292DSTOPTS:
2561                 case IPV6_2292RTHDR:
2562                 case IPV6_2292PKTOPTIONS:
2563                         return (ENOPROTOOPT);
2564                 }
2565         }
2566         if (sticky && cmsg) {
2567                 switch (optname) {
2568                 case IPV6_PKTINFO:
2569                 case IPV6_HOPLIMIT:
2570                 case IPV6_NEXTHOP:
2571                 case IPV6_HOPOPTS:
2572                 case IPV6_DSTOPTS:
2573                 case IPV6_RTHDRDSTOPTS:
2574                 case IPV6_RTHDR:
2575                 case IPV6_USE_MIN_MTU:
2576                 case IPV6_DONTFRAG:
2577                 case IPV6_TCLASS:
2578                 case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2579                         return (ENOPROTOOPT);
2580                 }
2581         }
2582
2583         switch (optname) {
2584         case IPV6_2292PKTINFO:
2585         case IPV6_PKTINFO:
2586         {
2587                 struct in6_pktinfo *pktinfo;
2588                 if (len != sizeof(struct in6_pktinfo))
2589                         return (EINVAL);
2590                 pktinfo = (struct in6_pktinfo *)buf;
2591
2592                 /*
2593                  * An application can clear any sticky IPV6_PKTINFO option by
2594                  * doing a "regular" setsockopt with ipi6_addr being
2595                  * in6addr_any and ipi6_ifindex being zero.
2596                  * [RFC 3542, Section 6]
2597                  */
2598                 if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2599                     pktinfo->ipi6_ifindex == 0 &&
2600                     IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2601                         ip6_clearpktopts(opt, optname);
2602                         break;
2603                 }
2604
2605                 if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2606                     sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2607                         return (EINVAL);
2608                 }
2609
2610                 /* validate the interface index if specified. */
2611                 if (pktinfo->ipi6_ifindex > if_index ||
2612                     pktinfo->ipi6_ifindex < 0) {
2613                          return (ENXIO);
2614                 }
2615                 /*
2616                  * Check if the requested source address is indeed a
2617                  * unicast address assigned to the node, and can be
2618                  * used as the packet's source address.
2619                  */
2620                 if (opt->ip6po_pktinfo != NULL &&
2621                     !IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2622                         struct in6_ifaddr *ia6;
2623                         struct sockaddr_in6 sin6;
2624
2625                         bzero(&sin6, sizeof(sin6));
2626                         sin6.sin6_len = sizeof(sin6);
2627                         sin6.sin6_family = AF_INET6;
2628                         sin6.sin6_addr =
2629                         opt->ip6po_pktinfo->ipi6_addr;
2630                         ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2631                         if (ia6 == NULL ||
2632                                 (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2633                                         IN6_IFF_NOTREADY)) != 0)
2634                         return (EADDRNOTAVAIL);
2635                 }
2636
2637                 /*
2638                  * We store the address anyway, and let in6_selectsrc()
2639                  * validate the specified address.  This is because ipi6_addr
2640                  * may not have enough information about its scope zone, and
2641                  * we may need additional information (such as outgoing
2642                  * interface or the scope zone of a destination address) to
2643                  * disambiguate the scope.
2644                  * XXX: the delay of the validation may confuse the
2645                  * application when it is used as a sticky option.
2646                  */
2647                 if (opt->ip6po_pktinfo == NULL) {
2648                         opt->ip6po_pktinfo = kmalloc(sizeof(*pktinfo),
2649                             M_IP6OPT, M_NOWAIT);
2650                         if (opt->ip6po_pktinfo == NULL)
2651                                 return (ENOBUFS);
2652                 }
2653                 bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2654                 break;
2655         }
2656
2657         case IPV6_2292HOPLIMIT:
2658         case IPV6_HOPLIMIT:
2659         {
2660                 int *hlimp;
2661
2662                 /*
2663                  * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2664                  * to simplify the ordering among hoplimit options.
2665                  */
2666                 if (optname == IPV6_HOPLIMIT && sticky)
2667                         return (ENOPROTOOPT);
2668
2669                 if (len != sizeof(int))
2670                         return (EINVAL);
2671                 hlimp = (int *)buf;
2672                 if (*hlimp < -1 || *hlimp > 255)
2673                         return (EINVAL);
2674
2675                 opt->ip6po_hlim = *hlimp;
2676                 break;
2677         }
2678
2679         case IPV6_TCLASS:
2680         {
2681                 int tclass;
2682
2683                 if (len != sizeof(int))
2684                         return (EINVAL);
2685                 tclass = *(int *)buf;
2686                 if (tclass < -1 || tclass > 255)
2687                         return (EINVAL);
2688
2689                 opt->ip6po_tclass = tclass;
2690                 break;
2691         }
2692
2693         case IPV6_2292NEXTHOP:
2694         case IPV6_NEXTHOP:
2695                 if (!priv)
2696                         return (EPERM);
2697
2698                 if (len == 0) { /* just remove the option */
2699                         ip6_clearpktopts(opt, IPV6_NEXTHOP);
2700                         break;
2701                 }
2702
2703                 /* check if cmsg_len is large enough for sa_len */
2704                 if (len < sizeof(struct sockaddr) || len < *buf)
2705                         return (EINVAL);
2706
2707                 switch (((struct sockaddr *)buf)->sa_family) {
2708                 case AF_INET6:
2709                 {
2710                         struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2711                         //int error;
2712
2713                         if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2714                                 return (EINVAL);
2715
2716                         if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2717                             IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2718                                 return (EINVAL);
2719                         }
2720                         break;
2721                 }
2722                 case AF_LINK:   /* should eventually be supported */
2723                 default:
2724                         return (EAFNOSUPPORT);
2725                 }
2726
2727                 /* turn off the previous option, then set the new option. */
2728                 ip6_clearpktopts(opt, IPV6_NEXTHOP);
2729                 opt->ip6po_nexthop = kmalloc(*buf, M_IP6OPT, M_NOWAIT);
2730                 if (opt->ip6po_nexthop == NULL)
2731                         return (ENOBUFS);
2732                 bcopy(buf, opt->ip6po_nexthop, *buf);
2733                 break;
2734
2735         case IPV6_2292HOPOPTS:
2736         case IPV6_HOPOPTS:
2737         {
2738                 struct ip6_hbh *hbh;
2739                 int hbhlen;
2740
2741                 /*
2742                  * XXX: We don't allow a non-privileged user to set ANY HbH
2743                  * options, since per-option restriction has too much
2744                  * overhead.
2745                  */
2746                 if (!priv)
2747                         return (EPERM);
2748                 if (len == 0) {
2749                         ip6_clearpktopts(opt, IPV6_HOPOPTS);
2750                         break;  /* just remove the option */
2751                 }
2752
2753                 /* message length validation */
2754                 if (len < sizeof(struct ip6_hbh))
2755                         return (EINVAL);
2756                 hbh = (struct ip6_hbh *)buf;
2757                 hbhlen = (hbh->ip6h_len + 1) << 3;
2758                 if (len != hbhlen)
2759                         return (EINVAL);
2760
2761                 /* turn off the previous option, then set the new option. */
2762                 ip6_clearpktopts(opt, IPV6_HOPOPTS);
2763                 opt->ip6po_hbh = kmalloc(hbhlen, M_IP6OPT, M_NOWAIT);
2764                 if (opt->ip6po_hbh == NULL)
2765                         return (ENOBUFS);
2766                 bcopy(hbh, opt->ip6po_hbh, hbhlen);
2767
2768                 break;
2769         }
2770
2771         case IPV6_2292DSTOPTS:
2772         case IPV6_DSTOPTS:
2773         case IPV6_RTHDRDSTOPTS:
2774         {
2775                 struct ip6_dest *dest, **newdest = NULL;
2776                 int destlen;
2777                 if (!priv)
2778                         return (EPERM);
2779
2780                 if (len == 0) {
2781                         ip6_clearpktopts(opt, optname);
2782                         break;  /* just remove the option */
2783                 }
2784
2785                 /* message length validation */
2786                 if (len < sizeof(struct ip6_dest))
2787                         return (EINVAL);
2788                 dest = (struct ip6_dest *)buf;
2789                 destlen = (dest->ip6d_len + 1) << 3;
2790                 if (len != destlen)
2791                         return (EINVAL);
2792
2793                 /*
2794                  * Determine the position that the destination options header
2795                  * should be inserted; before or after the routing header.
2796                  */
2797                 switch (optname) {
2798                 case IPV6_2292DSTOPTS:
2799                         /*
2800                          * The old advacned API is ambiguous on this point.
2801                          * Our approach is to determine the position based
2802                          * according to the existence of a routing header.
2803                          * Note, however, that this depends on the order of the
2804                          * extension headers in the ancillary data; the 1st
2805                          * part of the destination options header must appear
2806                          * before the routing header in the ancillary data,
2807                          * too.
2808                          * RFC3542 solved the ambiguity by introducing
2809                          * separate ancillary data or option types.
2810                          */
2811                         if (opt->ip6po_rthdr == NULL)
2812                                 newdest = &opt->ip6po_dest1;
2813                         else
2814                                 newdest = &opt->ip6po_dest2;
2815                         break;
2816                 case IPV6_RTHDRDSTOPTS:
2817                         newdest = &opt->ip6po_dest1;
2818                         break;
2819                 case IPV6_DSTOPTS:
2820                         newdest = &opt->ip6po_dest2;
2821                         break;
2822                 }
2823
2824                 /* turn off the previous option, then set the new option. */
2825                 ip6_clearpktopts(opt, optname);
2826                 *newdest = kmalloc(destlen, M_IP6OPT, M_NOWAIT);
2827                 if (*newdest == NULL)
2828                         return (ENOBUFS);
2829                 bcopy(dest, *newdest, destlen);
2830
2831                 break;
2832         }
2833
2834         case IPV6_2292RTHDR:
2835         case IPV6_RTHDR:
2836         {
2837                 struct ip6_rthdr *rth;
2838                 int rthlen;
2839
2840                 if (len == 0) {
2841                         ip6_clearpktopts(opt, IPV6_RTHDR);
2842                         break;  /* just remove the option */
2843                 }
2844
2845                 /* message length validation */
2846                 if (len < sizeof(struct ip6_rthdr))
2847                         return (EINVAL);
2848                 rth = (struct ip6_rthdr *)buf;
2849                 rthlen = (rth->ip6r_len + 1) << 3;
2850                 if (len != rthlen)
2851                         return (EINVAL);
2852
2853                 switch (rth->ip6r_type) {
2854                 default:
2855                         return (EINVAL);        /* not supported */
2856                 }
2857
2858                 /* turn off the previous option */
2859                 ip6_clearpktopts(opt, IPV6_RTHDR);
2860                 opt->ip6po_rthdr = kmalloc(rthlen, M_IP6OPT, M_NOWAIT);
2861                 if (opt->ip6po_rthdr == NULL)
2862                         return (ENOBUFS);
2863                 bcopy(rth, opt->ip6po_rthdr, rthlen);
2864
2865                 break;
2866         }
2867
2868         case IPV6_USE_MIN_MTU:
2869                 if (len != sizeof(int))
2870                         return (EINVAL);
2871                 minmtupolicy = *(int *)buf;
2872                 if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2873                     minmtupolicy != IP6PO_MINMTU_DISABLE &&
2874                     minmtupolicy != IP6PO_MINMTU_ALL) {
2875                         return (EINVAL);
2876                 }
2877                 opt->ip6po_minmtu = minmtupolicy;
2878                 break;
2879
2880         case IPV6_DONTFRAG:
2881                 if (len != sizeof(int))
2882                         return (EINVAL);
2883
2884                 if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2885                         /*
2886                          * we ignore this option for TCP sockets.
2887                          * (RFC3542 leaves this case unspecified.)
2888                          */
2889                         opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2890                 } else
2891                         opt->ip6po_flags |= IP6PO_DONTFRAG;
2892                 break;
2893
2894         case IPV6_PREFER_TEMPADDR:
2895                 if (len != sizeof(int))
2896                         return (EINVAL);
2897                 preftemp = *(int *)buf;
2898                 if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2899                     preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2900                     preftemp != IP6PO_TEMPADDR_PREFER) {
2901                         return (EINVAL);
2902                 }
2903                 opt->ip6po_prefer_tempaddr = preftemp;
2904                 break;
2905
2906         default:
2907                 return (ENOPROTOOPT);
2908         } /* end of switch */
2909
2910         return (0);
2911 }
2912
2913
2914 /*
2915  * Set IPv6 outgoing packet options based on advanced API.
2916  */
2917 int
2918 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt,
2919     struct ip6_pktopts *stickyopt, int uproto, int priv)
2920 {
2921         struct cmsghdr *cm = NULL;
2922
2923         if (control == NULL || opt == NULL)
2924                 return (EINVAL);
2925
2926         init_ip6pktopts(opt);
2927
2928         /*
2929          * XXX: Currently, we assume all the optional information is stored
2930          * in a single mbuf.
2931          */
2932         if (stickyopt) {
2933                 int error;
2934
2935                 /*
2936                  * If stickyopt is provided, make a local copy of the options
2937                  * for this particular packet, then override them by ancillary
2938                  * objects.
2939                  * XXX: copypktopts() does not copy the cached route to a next
2940                  * hop (if any).  This is not very good in terms of efficiency,
2941                  * but we can allow this since this option should be rarely
2942                  * used.
2943                  */
2944                 if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2945                         return (error);
2946         }
2947
2948         /*
2949          * XXX: Currently, we assume all the optional information is stored
2950          * in a single mbuf.
2951          */
2952         if (control->m_next)
2953                 return (EINVAL);
2954
2955         for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2956             control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2957                 int error;
2958
2959                 if (control->m_len < CMSG_LEN(0))
2960                         return (EINVAL);
2961
2962                 cm = mtod(control, struct cmsghdr *);
2963                 if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2964                         return (EINVAL);
2965                 if (cm->cmsg_level != IPPROTO_IPV6)
2966                         continue;
2967
2968                 error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
2969                     cm->cmsg_len - CMSG_LEN(0), opt, 0, 1, uproto, priv);
2970                 if (error)
2971                         return (error);
2972         }
2973
2974         return (0);
2975 }
2976
2977 /*
2978  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2979  * packet to the input queue of a specified interface.  Note that this
2980  * calls the output routine of the loopback "driver", but with an interface
2981  * pointer that might NOT be loif -- easier than replicating that code here.
2982  */
2983 void
2984 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2985 {
2986         struct mbuf *copym;
2987         struct ip6_hdr *ip6;
2988
2989         copym = m_copy(m, 0, M_COPYALL);
2990         if (copym == NULL)
2991                 return;
2992
2993         /*
2994          * Make sure to deep-copy IPv6 header portion in case the data
2995          * is in an mbuf cluster, so that we can safely override the IPv6
2996          * header portion later.
2997          */
2998         if ((copym->m_flags & M_EXT) != 0 ||
2999             copym->m_len < sizeof(struct ip6_hdr)) {
3000                 copym = m_pullup(copym, sizeof(struct ip6_hdr));
3001                 if (copym == NULL)
3002                         return;
3003         }
3004
3005 #ifdef DIAGNOSTIC
3006         if (copym->m_len < sizeof(*ip6)) {
3007                 m_freem(copym);
3008                 return;
3009         }
3010 #endif
3011
3012         ip6 = mtod(copym, struct ip6_hdr *);
3013         /*
3014          * clear embedded scope identifiers if necessary.
3015          * in6_clearscope will touch the addresses only when necessary.
3016          */
3017         in6_clearscope(&ip6->ip6_src);
3018         in6_clearscope(&ip6->ip6_dst);
3019
3020         if_simloop(ifp, copym, dst->sin6_family, 0);
3021 }
3022
3023 /*
3024  * Separate the IPv6 header from the payload into its own mbuf.
3025  *
3026  * Returns the new mbuf chain or the original mbuf if no payload.
3027  * Returns NULL if can't allocate new mbuf for header.
3028  */
3029 static struct mbuf *
3030 ip6_splithdr(struct mbuf *m)
3031 {
3032         struct mbuf *mh;
3033
3034         if (m->m_len <= sizeof(struct ip6_hdr))         /* no payload */
3035                 return (m);
3036
3037         MGETHDR(mh, M_NOWAIT, MT_HEADER);
3038         if (mh == NULL)
3039                 return (NULL);
3040         mh->m_len = sizeof(struct ip6_hdr);
3041         M_MOVE_PKTHDR(mh, m);
3042         MH_ALIGN(mh, sizeof(struct ip6_hdr));
3043         bcopy(mtod(m, caddr_t), mtod(mh, caddr_t), sizeof(struct ip6_hdr));
3044         m->m_data += sizeof(struct ip6_hdr);
3045         m->m_len -= sizeof(struct ip6_hdr);
3046         mh->m_next = m;
3047         return (mh);
3048 }
3049
3050 /*
3051  * Compute IPv6 extension header length.
3052  */
3053 int
3054 ip6_optlen(struct in6pcb *in6p)
3055 {
3056         int len;
3057
3058         if (!in6p->in6p_outputopts)
3059                 return 0;
3060
3061         len = 0;
3062 #define elen(x) \
3063     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3064
3065         len += elen(in6p->in6p_outputopts->ip6po_hbh);
3066         if (in6p->in6p_outputopts->ip6po_rthdr)
3067                 /* dest1 is valid with rthdr only */
3068                 len += elen(in6p->in6p_outputopts->ip6po_dest1);
3069         len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3070         len += elen(in6p->in6p_outputopts->ip6po_dest2);
3071         return len;
3072 #undef elen
3073 }