net/ipv6/ipv6_sockglue.c

   1 /*
   2  *      IPv6 BSD socket options interface
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Based on linux/net/ipv4/ip_sockglue.c
   9  *
  10  *      This program is free software; you can redistribute it and/or
  11  *      modify it under the terms of the GNU General Public License
  12  *      as published by the Free Software Foundation; either version
  13  *      2 of the License, or (at your option) any later version.
  14  *
  15  *      FIXME: Make the setsockopt code POSIX compliant: That is
  16  *
  17  *      o       Truncate getsockopt returns
  18  *      o       Return an optlen of the truncated length if need be
  19  *
  20  *      Changes:
  21  *      David L Stevens <dlstevens@us.ibm.com>:
  22  *              - added multicast source filtering API for MLDv2
  23  */
  24
  25 #include <linux/module.h>
  26 #include <linux/capability.h>
  27 #include <linux/errno.h>
  28 #include <linux/types.h>
  29 #include <linux/socket.h>
  30 #include <linux/sockios.h>
  31 #include <linux/net.h>
  32 #include <linux/in6.h>
  33 #include <linux/mroute6.h>
  34 #include <linux/netdevice.h>
  35 #include <linux/if_arp.h>
  36 #include <linux/init.h>
  37 #include <linux/sysctl.h>
  38 #include <linux/netfilter.h>
  39 #include <linux/slab.h>
  40
  41 #include <net/sock.h>
  42 #include <net/snmp.h>
  43 #include <net/ipv6.h>
  44 #include <net/ndisc.h>
  45 #include <net/protocol.h>
  46 #include <net/transp_v6.h>
  47 #include <net/ip6_route.h>
  48 #include <net/addrconf.h>
  49 #include <net/inet_common.h>
  50 #include <net/tcp.h>
  51 #include <net/udp.h>
  52 #include <net/udplite.h>
  53 #include <net/xfrm.h>
  54 #include <net/compat.h>
  55
  56 #include <asm/uaccess.h>
  57
  58 struct ip6_ra_chain *ip6_ra_chain;
  59 DEFINE_RWLOCK(ip6_ra_lock);
  60
  61 int ip6_ra_control(struct sock *sk, int sel)
  62 {
  63         struct ip6_ra_chain *ra, *new_ra, **rap;
  64
  65         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
  66         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
  67                 return -ENOPROTOOPT;
  68
  69         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
  70
  71         write_lock_bh(&ip6_ra_lock);
  72         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
  73                 if (ra->sk == sk) {
  74                         if (sel>=0) {
  75                                 write_unlock_bh(&ip6_ra_lock);
  76                                 kfree(new_ra);
  77                                 return -EADDRINUSE;
  78                         }
  79
  80                         *rap = ra->next;
  81                         write_unlock_bh(&ip6_ra_lock);
  82
  83                         sock_put(sk);
  84                         kfree(ra);
  85                         return 0;
  86                 }
  87         }
  88         if (new_ra == NULL) {
  89                 write_unlock_bh(&ip6_ra_lock);
  90                 return -ENOBUFS;
  91         }
  92         new_ra->sk = sk;
  93         new_ra->sel = sel;
  94         new_ra->next = ra;
  95         *rap = new_ra;
  96         sock_hold(sk);
  97         write_unlock_bh(&ip6_ra_lock);
  98         return 0;
  99 }
 100
 101 static
 102 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
 103                                            struct ipv6_txoptions *opt)
 104 {
 105         if (inet_sk(sk)->is_icsk) {
 106                 if (opt &&
 107                     !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
 108                     inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) {
 109                         struct inet_connection_sock *icsk = inet_csk(sk);
 110                         icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
 111                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 112                 }
 113                 opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt,
 114                            opt);
 115         } else {
 116                 spin_lock(&sk->sk_dst_lock);
 117                 opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt,
 118                            opt);
 119                 spin_unlock(&sk->sk_dst_lock);
 120         }
 121         sk_dst_reset(sk);
 122
 123         return opt;
 124 }
 125
 126 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 127                     char __user *optval, unsigned int optlen)
 128 {
 129         struct ipv6_pinfo *np = inet6_sk(sk);
 130         struct net *net = sock_net(sk);
 131         int val, valbool;
 132         int retv = -ENOPROTOOPT;
 133
 134         if (optval == NULL)
 135                 val=0;
 136         else {
 137                 if (optlen >= sizeof(int)) {
 138                         if (get_user(val, (int __user *) optval))
 139                                 return -EFAULT;
 140                 } else
 141                         val = 0;
 142         }
 143
 144         valbool = (val!=0);
 145
 146         if (ip6_mroute_opt(optname))
 147                 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
 148
 149         lock_sock(sk);
 150
 151         switch (optname) {
 152
 153         case IPV6_ADDRFORM:
 154                 if (optlen < sizeof(int))
 155                         goto e_inval;
 156                 if (val == PF_INET) {
 157                         struct ipv6_txoptions *opt;
 158                         struct sk_buff *pktopt;
 159
 160                         if (sk->sk_type == SOCK_RAW)
 161                                 break;
 162
 163                         if (sk->sk_protocol == IPPROTO_UDP ||
 164                             sk->sk_protocol == IPPROTO_UDPLITE) {
 165                                 struct udp_sock *up = udp_sk(sk);
 166                                 if (up->pending == AF_INET6) {
 167                                         retv = -EBUSY;
 168                                         break;
 169                                 }
 170                         } else if (sk->sk_protocol != IPPROTO_TCP)
 171                                 break;
 172
 173                         if (sk->sk_state != TCP_ESTABLISHED) {
 174                                 retv = -ENOTCONN;
 175                                 break;
 176                         }
 177
 178                         if (ipv6_only_sock(sk) ||
 179                             !ipv6_addr_v4mapped(&np->daddr)) {
 180                                 retv = -EADDRNOTAVAIL;
 181                                 break;
 182                         }
 183
 184                         fl6_free_socklist(sk);
 185                         ipv6_sock_mc_close(sk);
 186
 187                         /*
 188                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
 189                          * remove it from the refcnt debug socks count in the
 190                          * original family...
 191                          */
 192                         sk_refcnt_debug_dec(sk);
 193
 194                         if (sk->sk_protocol == IPPROTO_TCP) {
 195                                 struct inet_connection_sock *icsk = inet_csk(sk);
 196                                 local_bh_disable();
 197                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 198                                 sock_prot_inuse_add(net, &tcp_prot, 1);
 199                                 local_bh_enable();
 200                                 sk->sk_prot = &tcp_prot;
 201                                 icsk->icsk_af_ops = &ipv4_specific;
 202                                 sk->sk_socket->ops = &inet_stream_ops;
 203                                 sk->sk_family = PF_INET;
 204                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 205                         } else {
 206                                 struct proto *prot = &udp_prot;
 207
 208                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
 209                                         prot = &udplite_prot;
 210                                 local_bh_disable();
 211                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 212                                 sock_prot_inuse_add(net, prot, 1);
 213                                 local_bh_enable();
 214                                 sk->sk_prot = prot;
 215                                 sk->sk_socket->ops = &inet_dgram_ops;
 216                                 sk->sk_family = PF_INET;
 217                         }
 218                         opt = xchg((__force struct ipv6_txoptions **)&np->opt,
 219                                    NULL);
 220                         if (opt) {
 221                                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 222                                 txopt_put(opt);
 223                         }
 224                         pktopt = xchg(&np->pktoptions, NULL);
 225                         kfree_skb(pktopt);
 226
 227                         sk->sk_destruct = inet_sock_destruct;
 228                         /*
 229                          * ... and add it to the refcnt debug socks count
 230                          * in the new family. -acme
 231                          */
 232                         sk_refcnt_debug_inc(sk);
 233                         module_put(THIS_MODULE);
 234                         retv = 0;
 235                         break;
 236                 }
 237                 goto e_inval;
 238
 239         case IPV6_V6ONLY:
 240                 if (optlen < sizeof(int) ||
 241                     inet_sk(sk)->inet_num)
 242                         goto e_inval;
 243                 np->ipv6only = valbool;
 244                 retv = 0;
 245                 break;
 246
 247         case IPV6_RECVPKTINFO:
 248                 if (optlen < sizeof(int))
 249                         goto e_inval;
 250                 np->rxopt.bits.rxinfo = valbool;
 251                 retv = 0;
 252                 break;
 253
 254         case IPV6_2292PKTINFO:
 255                 if (optlen < sizeof(int))
 256                         goto e_inval;
 257                 np->rxopt.bits.rxoinfo = valbool;
 258                 retv = 0;
 259                 break;
 260
 261         case IPV6_RECVHOPLIMIT:
 262                 if (optlen < sizeof(int))
 263                         goto e_inval;
 264                 np->rxopt.bits.rxhlim = valbool;
 265                 retv = 0;
 266                 break;
 267
 268         case IPV6_2292HOPLIMIT:
 269                 if (optlen < sizeof(int))
 270                         goto e_inval;
 271                 np->rxopt.bits.rxohlim = valbool;
 272                 retv = 0;
 273                 break;
 274
 275         case IPV6_RECVRTHDR:
 276                 if (optlen < sizeof(int))
 277                         goto e_inval;
 278                 np->rxopt.bits.srcrt = valbool;
 279                 retv = 0;
 280                 break;
 281
 282         case IPV6_2292RTHDR:
 283                 if (optlen < sizeof(int))
 284                         goto e_inval;
 285                 np->rxopt.bits.osrcrt = valbool;
 286                 retv = 0;
 287                 break;
 288
 289         case IPV6_RECVHOPOPTS:
 290                 if (optlen < sizeof(int))
 291                         goto e_inval;
 292                 np->rxopt.bits.hopopts = valbool;
 293                 retv = 0;
 294                 break;
 295
 296         case IPV6_2292HOPOPTS:
 297                 if (optlen < sizeof(int))
 298                         goto e_inval;
 299                 np->rxopt.bits.ohopopts = valbool;
 300                 retv = 0;
 301                 break;
 302
 303         case IPV6_RECVDSTOPTS:
 304                 if (optlen < sizeof(int))
 305                         goto e_inval;
 306                 np->rxopt.bits.dstopts = valbool;
 307                 retv = 0;
 308                 break;
 309
 310         case IPV6_2292DSTOPTS:
 311                 if (optlen < sizeof(int))
 312                         goto e_inval;
 313                 np->rxopt.bits.odstopts = valbool;
 314                 retv = 0;
 315                 break;
 316
 317         case IPV6_TCLASS:
 318                 if (optlen < sizeof(int))
 319                         goto e_inval;
 320                 if (val < -1 || val > 0xff)
 321                         goto e_inval;
 322                 /* RFC 3542, 6.5: default traffic class of 0x0 */
 323                 if (val == -1)
 324                         val = 0;
 325                 np->tclass = val;
 326                 retv = 0;
 327                 break;
 328
 329         case IPV6_RECVTCLASS:
 330                 if (optlen < sizeof(int))
 331                         goto e_inval;
 332                 np->rxopt.bits.rxtclass = valbool;
 333                 retv = 0;
 334                 break;
 335
 336         case IPV6_FLOWINFO:
 337                 if (optlen < sizeof(int))
 338                         goto e_inval;
 339                 np->rxopt.bits.rxflow = valbool;
 340                 retv = 0;
 341                 break;
 342
 343         case IPV6_RECVPATHMTU:
 344                 if (optlen < sizeof(int))
 345                         goto e_inval;
 346                 np->rxopt.bits.rxpmtu = valbool;
 347                 retv = 0;
 348                 break;
 349
 350         case IPV6_TRANSPARENT:
 351                 if (valbool && !capable(CAP_NET_ADMIN) && !capable(CAP_NET_RAW)) {
 352                         retv = -EPERM;
 353                         break;
 354                 }
 355                 if (optlen < sizeof(int))
 356                         goto e_inval;
 357                 /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
 358                 inet_sk(sk)->transparent = valbool;
 359                 retv = 0;
 360                 break;
 361
 362         case IPV6_RECVORIGDSTADDR:
 363                 if (optlen < sizeof(int))
 364                         goto e_inval;
 365                 np->rxopt.bits.rxorigdstaddr = valbool;
 366                 retv = 0;
 367                 break;
 368
 369         case IPV6_HOPOPTS:
 370         case IPV6_RTHDRDSTOPTS:
 371         case IPV6_RTHDR:
 372         case IPV6_DSTOPTS:
 373         {
 374                 struct ipv6_txoptions *opt;
 375
 376                 /* remove any sticky options header with a zero option
 377                  * length, per RFC3542.
 378                  */
 379                 if (optlen == 0)
 380                         optval = NULL;
 381                 else if (optval == NULL)
 382                         goto e_inval;
 383                 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
 384                          optlen & 0x7 || optlen > 8 * 255)
 385                         goto e_inval;
 386
 387                 /* hop-by-hop / destination options are privileged option */
 388                 retv = -EPERM;
 389                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
 390                         break;
 391
 392                 opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk));
 393                 opt = ipv6_renew_options(sk, opt, optname,
 394                                          (struct ipv6_opt_hdr __user *)optval,
 395                                          optlen);
 396                 if (IS_ERR(opt)) {
 397                         retv = PTR_ERR(opt);
 398                         break;
 399                 }
 400
 401                 /* routing header option needs extra check */
 402                 retv = -EINVAL;
 403                 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
 404                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
 405                         switch (rthdr->type) {
 406 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 407                         case IPV6_SRCRT_TYPE_2:
 408                                 if (rthdr->hdrlen != 2 ||
 409                                     rthdr->segments_left != 1)
 410                                         goto sticky_done;
 411
 412                                 break;
 413 #endif
 414                         default:
 415                                 goto sticky_done;
 416                         }
 417                 }
 418
 419                 retv = 0;
 420                 opt = ipv6_update_options(sk, opt);
 421 sticky_done:
 422                 if (opt) {
 423                         atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 424                         txopt_put(opt);
 425                 }
 426                 break;
 427         }
 428
 429         case IPV6_PKTINFO:
 430         {
 431                 struct in6_pktinfo pkt;
 432
 433                 if (optlen == 0)
 434                         goto e_inval;
 435                 else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
 436                         goto e_inval;
 437
 438                 if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
 439                                 retv = -EFAULT;
 440                                 break;
 441                 }
 442                 if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
 443                         goto e_inval;
 444
 445                 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
 446                 ipv6_addr_copy(&np->sticky_pktinfo.ipi6_addr, &pkt.ipi6_addr);
 447                 retv = 0;
 448                 break;
 449         }
 450
 451         case IPV6_2292PKTOPTIONS:
 452         {
 453                 struct ipv6_txoptions *opt = NULL;
 454                 struct msghdr msg;
 455                 struct flowi6 fl6;
 456                 int junk;
 457
 458                 memset(&fl6, 0, sizeof(fl6));
 459                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 460                 fl6.flowi6_mark = sk->sk_mark;
 461
 462                 if (optlen == 0)
 463                         goto update;
 464
 465                 /* 1K is probably excessive
 466                  * 1K is surely not enough, 2K per standard header is 16K.
 467                  */
 468                 retv = -EINVAL;
 469                 if (optlen > 64*1024)
 470                         break;
 471
 472                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 473                 retv = -ENOBUFS;
 474                 if (opt == NULL)
 475                         break;
 476
 477                 memset(opt, 0, sizeof(*opt));
 478                 atomic_set(&opt->refcnt, 1);
 479                 opt->tot_len = sizeof(*opt) + optlen;
 480                 retv = -EFAULT;
 481                 if (copy_from_user(opt+1, optval, optlen))
 482                         goto done;
 483
 484                 msg.msg_controllen = optlen;
 485                 msg.msg_control = (void*)(opt+1);
 486
 487                 retv = datagram_send_ctl(net, sk, &msg, &fl6, opt, &junk, &junk,
 488                                          &junk);
 489                 if (retv)
 490                         goto done;
 491 update:
 492                 retv = 0;
 493                 opt = ipv6_update_options(sk, opt);
 494 done:
 495                 if (opt) {
 496                         atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 497                         txopt_put(opt);
 498                 }
 499                 break;
 500         }
 501         case IPV6_UNICAST_HOPS:
 502                 if (optlen < sizeof(int))
 503                         goto e_inval;
 504                 if (val > 255 || val < -1)
 505                         goto e_inval;
 506                 np->hop_limit = val;
 507                 retv = 0;
 508                 break;
 509
 510         case IPV6_MULTICAST_HOPS:
 511                 if (sk->sk_type == SOCK_STREAM)
 512                         break;
 513                 if (optlen < sizeof(int))
 514                         goto e_inval;
 515                 if (val > 255 || val < -1)
 516                         goto e_inval;
 517                 np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
 518                 retv = 0;
 519                 break;
 520
 521         case IPV6_MULTICAST_LOOP:
 522                 if (optlen < sizeof(int))
 523                         goto e_inval;
 524                 if (val != valbool)
 525                         goto e_inval;
 526                 np->mc_loop = valbool;
 527                 retv = 0;
 528                 break;
 529
 530         case IPV6_MULTICAST_IF:
 531                 if (sk->sk_type == SOCK_STREAM)
 532                         break;
 533                 if (optlen < sizeof(int))
 534                         goto e_inval;
 535
 536                 if (val) {
 537                         struct net_device *dev;
 538
 539                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
 540                                 goto e_inval;
 541
 542                         dev = dev_get_by_index(net, val);
 543                         if (!dev) {
 544                                 retv = -ENODEV;
 545                                 break;
 546                         }
 547                         dev_put(dev);
 548                 }
 549                 np->mcast_oif = val;
 550                 retv = 0;
 551                 break;
 552         case IPV6_ADD_MEMBERSHIP:
 553         case IPV6_DROP_MEMBERSHIP:
 554         {
 555                 struct ipv6_mreq mreq;
 556
 557                 if (optlen < sizeof(struct ipv6_mreq))
 558                         goto e_inval;
 559
 560                 retv = -EPROTO;
 561                 if (inet_sk(sk)->is_icsk)
 562                         break;
 563
 564                 retv = -EFAULT;
 565                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 566                         break;
 567
 568                 if (optname == IPV6_ADD_MEMBERSHIP)
 569                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 570                 else
 571                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 572                 break;
 573         }
 574         case IPV6_JOIN_ANYCAST:
 575         case IPV6_LEAVE_ANYCAST:
 576         {
 577                 struct ipv6_mreq mreq;
 578
 579                 if (optlen < sizeof(struct ipv6_mreq))
 580                         goto e_inval;
 581
 582                 retv = -EFAULT;
 583                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 584                         break;
 585
 586                 if (optname == IPV6_JOIN_ANYCAST)
 587                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 588                 else
 589                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 590                 break;
 591         }
 592         case MCAST_JOIN_GROUP:
 593         case MCAST_LEAVE_GROUP:
 594         {
 595                 struct group_req greq;
 596                 struct sockaddr_in6 *psin6;
 597
 598                 if (optlen < sizeof(struct group_req))
 599                         goto e_inval;
 600
 601                 retv = -EFAULT;
 602                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
 603                         break;
 604                 if (greq.gr_group.ss_family != AF_INET6) {
 605                         retv = -EADDRNOTAVAIL;
 606                         break;
 607                 }
 608                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 609                 if (optname == MCAST_JOIN_GROUP)
 610                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
 611                                 &psin6->sin6_addr);
 612                 else
 613                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
 614                                 &psin6->sin6_addr);
 615                 break;
 616         }
 617         case MCAST_JOIN_SOURCE_GROUP:
 618         case MCAST_LEAVE_SOURCE_GROUP:
 619         case MCAST_BLOCK_SOURCE:
 620         case MCAST_UNBLOCK_SOURCE:
 621         {
 622                 struct group_source_req greqs;
 623                 int omode, add;
 624
 625                 if (optlen < sizeof(struct group_source_req))
 626                         goto e_inval;
 627                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 628                         retv = -EFAULT;
 629                         break;
 630                 }
 631                 if (greqs.gsr_group.ss_family != AF_INET6 ||
 632                     greqs.gsr_source.ss_family != AF_INET6) {
 633                         retv = -EADDRNOTAVAIL;
 634                         break;
 635                 }
 636                 if (optname == MCAST_BLOCK_SOURCE) {
 637                         omode = MCAST_EXCLUDE;
 638                         add = 1;
 639                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 640                         omode = MCAST_EXCLUDE;
 641                         add = 0;
 642                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 643                         struct sockaddr_in6 *psin6;
 644
 645                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 646                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
 647                                 &psin6->sin6_addr);
 648                         /* prior join w/ different source is ok */
 649                         if (retv && retv != -EADDRINUSE)
 650                                 break;
 651                         omode = MCAST_INCLUDE;
 652                         add = 1;
 653                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 654                         omode = MCAST_INCLUDE;
 655                         add = 0;
 656                 }
 657                 retv = ip6_mc_source(add, omode, sk, &greqs);
 658                 break;
 659         }
 660         case MCAST_MSFILTER:
 661         {
 662                 extern int sysctl_mld_max_msf;
 663                 struct group_filter *gsf;
 664
 665                 if (optlen < GROUP_FILTER_SIZE(0))
 666                         goto e_inval;
 667                 if (optlen > sysctl_optmem_max) {
 668                         retv = -ENOBUFS;
 669                         break;
 670                 }
 671                 gsf = kmalloc(optlen,GFP_KERNEL);
 672                 if (!gsf) {
 673                         retv = -ENOBUFS;
 674                         break;
 675                 }
 676                 retv = -EFAULT;
 677                 if (copy_from_user(gsf, optval, optlen)) {
 678                         kfree(gsf);
 679                         break;
 680                 }
 681                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 682                 if (gsf->gf_numsrc >= 0x1ffffffU ||
 683                     gsf->gf_numsrc > sysctl_mld_max_msf) {
 684                         kfree(gsf);
 685                         retv = -ENOBUFS;
 686                         break;
 687                 }
 688                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 689                         kfree(gsf);
 690                         retv = -EINVAL;
 691                         break;
 692                 }
 693                 retv = ip6_mc_msfilter(sk, gsf);
 694                 kfree(gsf);
 695
 696                 break;
 697         }
 698         case IPV6_ROUTER_ALERT:
 699                 if (optlen < sizeof(int))
 700                         goto e_inval;
 701                 retv = ip6_ra_control(sk, val);
 702                 break;
 703         case IPV6_MTU_DISCOVER:
 704                 if (optlen < sizeof(int))
 705                         goto e_inval;
 706                 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
 707                         goto e_inval;
 708                 np->pmtudisc = val;
 709                 retv = 0;
 710                 break;
 711         case IPV6_MTU:
 712                 if (optlen < sizeof(int))
 713                         goto e_inval;
 714                 if (val && val < IPV6_MIN_MTU)
 715                         goto e_inval;
 716                 np->frag_size = val;
 717                 retv = 0;
 718                 break;
 719         case IPV6_RECVERR:
 720                 if (optlen < sizeof(int))
 721                         goto e_inval;
 722                 np->recverr = valbool;
 723                 if (!val)
 724                         skb_queue_purge(&sk->sk_error_queue);
 725                 retv = 0;
 726                 break;
 727         case IPV6_FLOWINFO_SEND:
 728                 if (optlen < sizeof(int))
 729                         goto e_inval;
 730                 np->sndflow = valbool;
 731                 retv = 0;
 732                 break;
 733         case IPV6_FLOWLABEL_MGR:
 734                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
 735                 break;
 736         case IPV6_IPSEC_POLICY:
 737         case IPV6_XFRM_POLICY:
 738                 retv = -EPERM;
 739                 if (!capable(CAP_NET_ADMIN))
 740                         break;
 741                 retv = xfrm_user_policy(sk, optname, optval, optlen);
 742                 break;
 743
 744         case IPV6_ADDR_PREFERENCES:
 745             {
 746                 unsigned int pref = 0;
 747                 unsigned int prefmask = ~0;
 748
 749                 if (optlen < sizeof(int))
 750                         goto e_inval;
 751
 752                 retv = -EINVAL;
 753
 754                 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
 755                 switch (val & (IPV6_PREFER_SRC_PUBLIC|
 756                                IPV6_PREFER_SRC_TMP|
 757                                IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
 758                 case IPV6_PREFER_SRC_PUBLIC:
 759                         pref |= IPV6_PREFER_SRC_PUBLIC;
 760                         break;
 761                 case IPV6_PREFER_SRC_TMP:
 762                         pref |= IPV6_PREFER_SRC_TMP;
 763                         break;
 764                 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
 765                         break;
 766                 case 0:
 767                         goto pref_skip_pubtmp;
 768                 default:
 769                         goto e_inval;
 770                 }
 771
 772                 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
 773                               IPV6_PREFER_SRC_TMP);
 774 pref_skip_pubtmp:
 775
 776                 /* check HOME/COA conflicts */
 777                 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
 778                 case IPV6_PREFER_SRC_HOME:
 779                         break;
 780                 case IPV6_PREFER_SRC_COA:
 781                         pref |= IPV6_PREFER_SRC_COA;
 782                 case 0:
 783                         goto pref_skip_coa;
 784                 default:
 785                         goto e_inval;
 786                 }
 787
 788                 prefmask &= ~IPV6_PREFER_SRC_COA;
 789 pref_skip_coa:
 790
 791                 /* check CGA/NONCGA conflicts */
 792                 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
 793                 case IPV6_PREFER_SRC_CGA:
 794                 case IPV6_PREFER_SRC_NONCGA:
 795                 case 0:
 796                         break;
 797                 default:
 798                         goto e_inval;
 799                 }
 800
 801                 np->srcprefs = (np->srcprefs & prefmask) | pref;
 802                 retv = 0;
 803
 804                 break;
 805             }
 806         case IPV6_MINHOPCOUNT:
 807                 if (optlen < sizeof(int))
 808                         goto e_inval;
 809                 if (val < 0 || val > 255)
 810                         goto e_inval;
 811                 np->min_hopcount = val;
 812                 retv = 0;
 813                 break;
 814         case IPV6_DONTFRAG:
 815                 np->dontfrag = valbool;
 816                 retv = 0;
 817                 break;
 818         }
 819
 820         release_sock(sk);
 821
 822         return retv;
 823
 824 e_inval:
 825         release_sock(sk);
 826         return -EINVAL;
 827 }
 828
 829 int ipv6_setsockopt(struct sock *sk, int level, int optname,
 830                     char __user *optval, unsigned int optlen)
 831 {
 832         int err;
 833
 834         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 835                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 836
 837         if (level != SOL_IPV6)
 838                 return -ENOPROTOOPT;
 839
 840         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 841 #ifdef CONFIG_NETFILTER
 842         /* we need to exclude all possible ENOPROTOOPTs except default case */
 843         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 844                         optname != IPV6_XFRM_POLICY)
 845                 err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
 846 #endif
 847         return err;
 848 }
 849
 850 EXPORT_SYMBOL(ipv6_setsockopt);
 851
 852 #ifdef CONFIG_COMPAT
 853 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
 854                            char __user *optval, unsigned int optlen)
 855 {
 856         int err;
 857
 858         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
 859                 if (udp_prot.compat_setsockopt != NULL)
 860                         return udp_prot.compat_setsockopt(sk, level, optname,
 861                                                           optval, optlen);
 862                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 863         }
 864
 865         if (level != SOL_IPV6)
 866                 return -ENOPROTOOPT;
 867
 868         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
 869                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
 870                         ipv6_setsockopt);
 871
 872         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 873 #ifdef CONFIG_NETFILTER
 874         /* we need to exclude all possible ENOPROTOOPTs except default case */
 875         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 876             optname != IPV6_XFRM_POLICY)
 877                 err = compat_nf_setsockopt(sk, PF_INET6, optname, optval,
 878                                            optlen);
 879 #endif
 880         return err;
 881 }
 882
 883 EXPORT_SYMBOL(compat_ipv6_setsockopt);
 884 #endif
 885
 886 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
 887                                   int optname, char __user *optval, int len)
 888 {
 889         struct ipv6_opt_hdr *hdr;
 890
 891         if (!opt)
 892                 return 0;
 893
 894         switch(optname) {
 895         case IPV6_HOPOPTS:
 896                 hdr = opt->hopopt;
 897                 break;
 898         case IPV6_RTHDRDSTOPTS:
 899                 hdr = opt->dst0opt;
 900                 break;
 901         case IPV6_RTHDR:
 902                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
 903                 break;
 904         case IPV6_DSTOPTS:
 905                 hdr = opt->dst1opt;
 906                 break;
 907         default:
 908                 return -EINVAL; /* should not happen */
 909         }
 910
 911         if (!hdr)
 912                 return 0;
 913
 914         len = min_t(unsigned int, len, ipv6_optlen(hdr));
 915         if (copy_to_user(optval, hdr, len))
 916                 return -EFAULT;
 917         return len;
 918 }
 919
 920 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 921                     char __user *optval, int __user *optlen, unsigned flags)
 922 {
 923         struct ipv6_pinfo *np = inet6_sk(sk);
 924         int len;
 925         int val;
 926
 927         if (ip6_mroute_opt(optname))
 928                 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
 929
 930         if (get_user(len, optlen))
 931                 return -EFAULT;
 932         switch (optname) {
 933         case IPV6_ADDRFORM:
 934                 if (sk->sk_protocol != IPPROTO_UDP &&
 935                     sk->sk_protocol != IPPROTO_UDPLITE &&
 936                     sk->sk_protocol != IPPROTO_TCP)
 937                         return -ENOPROTOOPT;
 938                 if (sk->sk_state != TCP_ESTABLISHED)
 939                         return -ENOTCONN;
 940                 val = sk->sk_family;
 941                 break;
 942         case MCAST_MSFILTER:
 943         {
 944                 struct group_filter gsf;
 945                 int err;
 946
 947                 if (len < GROUP_FILTER_SIZE(0))
 948                         return -EINVAL;
 949                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
 950                         return -EFAULT;
 951                 if (gsf.gf_group.ss_family != AF_INET6)
 952                         return -EADDRNOTAVAIL;
 953                 lock_sock(sk);
 954                 err = ip6_mc_msfget(sk, &gsf,
 955                         (struct group_filter __user *)optval, optlen);
 956                 release_sock(sk);
 957                 return err;
 958         }
 959
 960         case IPV6_2292PKTOPTIONS:
 961         {
 962                 struct msghdr msg;
 963                 struct sk_buff *skb;
 964
 965                 if (sk->sk_type != SOCK_STREAM)
 966                         return -ENOPROTOOPT;
 967
 968                 msg.msg_control = optval;
 969                 msg.msg_controllen = len;
 970                 msg.msg_flags = flags;
 971
 972                 lock_sock(sk);
 973                 skb = np->pktoptions;
 974                 if (skb)
 975                         atomic_inc(&skb->users);
 976                 release_sock(sk);
 977
 978                 if (skb) {
 979                         int err = datagram_recv_ctl(sk, &msg, skb);
 980                         kfree_skb(skb);
 981                         if (err)
 982                                 return err;
 983                 } else {
 984                         if (np->rxopt.bits.rxinfo) {
 985                                 struct in6_pktinfo src_info;
 986                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
 987                                         np->sticky_pktinfo.ipi6_ifindex;
 988                                 np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
 989                                         ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
 990                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 991                         }
 992                         if (np->rxopt.bits.rxhlim) {
 993                                 int hlim = np->mcast_hops;
 994                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 995                         }
 996                         if (np->rxopt.bits.rxoinfo) {
 997                                 struct in6_pktinfo src_info;
 998                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
 999                                         np->sticky_pktinfo.ipi6_ifindex;
1000                                 np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
1001                                         ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
1002                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1003                         }
1004                         if (np->rxopt.bits.rxohlim) {
1005                                 int hlim = np->mcast_hops;
1006                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1007                         }
1008                 }
1009                 len -= msg.msg_controllen;
1010                 return put_user(len, optlen);
1011         }
1012         case IPV6_MTU:
1013         {
1014                 struct dst_entry *dst;
1015
1016                 val = 0;
1017                 rcu_read_lock();
1018                 dst = __sk_dst_get(sk);
1019                 if (dst)
1020                         val = dst_mtu(dst);
1021                 rcu_read_unlock();
1022                 if (!val)
1023                         return -ENOTCONN;
1024                 break;
1025         }
1026
1027         case IPV6_V6ONLY:
1028                 val = np->ipv6only;
1029                 break;
1030
1031         case IPV6_RECVPKTINFO:
1032                 val = np->rxopt.bits.rxinfo;
1033                 break;
1034
1035         case IPV6_2292PKTINFO:
1036                 val = np->rxopt.bits.rxoinfo;
1037                 break;
1038
1039         case IPV6_RECVHOPLIMIT:
1040                 val = np->rxopt.bits.rxhlim;
1041                 break;
1042
1043         case IPV6_2292HOPLIMIT:
1044                 val = np->rxopt.bits.rxohlim;
1045                 break;
1046
1047         case IPV6_RECVRTHDR:
1048                 val = np->rxopt.bits.srcrt;
1049                 break;
1050
1051         case IPV6_2292RTHDR:
1052                 val = np->rxopt.bits.osrcrt;
1053                 break;
1054
1055         case IPV6_HOPOPTS:
1056         case IPV6_RTHDRDSTOPTS:
1057         case IPV6_RTHDR:
1058         case IPV6_DSTOPTS:
1059         {
1060                 struct ipv6_txoptions *opt;
1061
1062                 lock_sock(sk);
1063                 opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk));
1064                 len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1065                 release_sock(sk);
1066                 /* check if ipv6_getsockopt_sticky() returns err code */
1067                 if (len < 0)
1068                         return len;
1069                 return put_user(len, optlen);
1070         }
1071
1072         case IPV6_RECVHOPOPTS:
1073                 val = np->rxopt.bits.hopopts;
1074                 break;
1075
1076         case IPV6_2292HOPOPTS:
1077                 val = np->rxopt.bits.ohopopts;
1078                 break;
1079
1080         case IPV6_RECVDSTOPTS:
1081                 val = np->rxopt.bits.dstopts;
1082                 break;
1083
1084         case IPV6_2292DSTOPTS:
1085                 val = np->rxopt.bits.odstopts;
1086                 break;
1087
1088         case IPV6_TCLASS:
1089                 val = np->tclass;
1090                 break;
1091
1092         case IPV6_RECVTCLASS:
1093                 val = np->rxopt.bits.rxtclass;
1094                 break;
1095
1096         case IPV6_FLOWINFO:
1097                 val = np->rxopt.bits.rxflow;
1098                 break;
1099
1100         case IPV6_RECVPATHMTU:
1101                 val = np->rxopt.bits.rxpmtu;
1102                 break;
1103
1104         case IPV6_PATHMTU:
1105         {
1106                 struct dst_entry *dst;
1107                 struct ip6_mtuinfo mtuinfo;
1108
1109                 if (len < sizeof(mtuinfo))
1110                         return -EINVAL;
1111
1112                 len = sizeof(mtuinfo);
1113                 memset(&mtuinfo, 0, sizeof(mtuinfo));
1114
1115                 rcu_read_lock();
1116                 dst = __sk_dst_get(sk);
1117                 if (dst)
1118                         mtuinfo.ip6m_mtu = dst_mtu(dst);
1119                 rcu_read_unlock();
1120                 if (!mtuinfo.ip6m_mtu)
1121                         return -ENOTCONN;
1122
1123                 if (put_user(len, optlen))
1124                         return -EFAULT;
1125                 if (copy_to_user(optval, &mtuinfo, len))
1126                         return -EFAULT;
1127
1128                 return 0;
1129                 break;
1130         }
1131
1132         case IPV6_TRANSPARENT:
1133                 val = inet_sk(sk)->transparent;
1134                 break;
1135
1136         case IPV6_RECVORIGDSTADDR:
1137                 val = np->rxopt.bits.rxorigdstaddr;
1138                 break;
1139
1140         case IPV6_UNICAST_HOPS:
1141         case IPV6_MULTICAST_HOPS:
1142         {
1143                 struct dst_entry *dst;
1144
1145                 if (optname == IPV6_UNICAST_HOPS)
1146                         val = np->hop_limit;
1147                 else
1148                         val = np->mcast_hops;
1149
1150                 if (val < 0) {
1151                         rcu_read_lock();
1152                         dst = __sk_dst_get(sk);
1153                         if (dst)
1154                                 val = ip6_dst_hoplimit(dst);
1155                         rcu_read_unlock();
1156                 }
1157
1158                 if (val < 0)
1159                         val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1160                 break;
1161         }
1162
1163         case IPV6_MULTICAST_LOOP:
1164                 val = np->mc_loop;
1165                 break;
1166
1167         case IPV6_MULTICAST_IF:
1168                 val = np->mcast_oif;
1169                 break;
1170
1171         case IPV6_MTU_DISCOVER:
1172                 val = np->pmtudisc;
1173                 break;
1174
1175         case IPV6_RECVERR:
1176                 val = np->recverr;
1177                 break;
1178
1179         case IPV6_FLOWINFO_SEND:
1180                 val = np->sndflow;
1181                 break;
1182
1183         case IPV6_ADDR_PREFERENCES:
1184                 val = 0;
1185
1186                 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1187                         val |= IPV6_PREFER_SRC_TMP;
1188                 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1189                         val |= IPV6_PREFER_SRC_PUBLIC;
1190                 else {
1191                         /* XXX: should we return system default? */
1192                         val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1193                 }
1194
1195                 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1196                         val |= IPV6_PREFER_SRC_COA;
1197                 else
1198                         val |= IPV6_PREFER_SRC_HOME;
1199                 break;
1200
1201         case IPV6_MINHOPCOUNT:
1202                 val = np->min_hopcount;
1203                 break;
1204
1205         case IPV6_DONTFRAG:
1206                 val = np->dontfrag;
1207                 break;
1208
1209         default:
1210                 return -ENOPROTOOPT;
1211         }
1212         len = min_t(unsigned int, sizeof(int), len);
1213         if(put_user(len, optlen))
1214                 return -EFAULT;
1215         if(copy_to_user(optval,&val,len))
1216                 return -EFAULT;
1217         return 0;
1218 }
1219
1220 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1221                     char __user *optval, int __user *optlen)
1222 {
1223         int err;
1224
1225         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1226                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1227
1228         if(level != SOL_IPV6)
1229                 return -ENOPROTOOPT;
1230
1231         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen, 0);
1232 #ifdef CONFIG_NETFILTER
1233         /* we need to exclude all possible ENOPROTOOPTs except default case */
1234         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1235                 int len;
1236
1237                 if (get_user(len, optlen))
1238                         return -EFAULT;
1239
1240                 err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1241                 if (err >= 0)
1242                         err = put_user(len, optlen);
1243         }
1244 #endif
1245         return err;
1246 }
1247
1248 EXPORT_SYMBOL(ipv6_getsockopt);
1249
1250 #ifdef CONFIG_COMPAT
1251 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1252                            char __user *optval, int __user *optlen)
1253 {
1254         int err;
1255
1256         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1257                 if (udp_prot.compat_getsockopt != NULL)
1258                         return udp_prot.compat_getsockopt(sk, level, optname,
1259                                                           optval, optlen);
1260                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1261         }
1262
1263         if (level != SOL_IPV6)
1264                 return -ENOPROTOOPT;
1265
1266         if (optname == MCAST_MSFILTER)
1267                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1268                         ipv6_getsockopt);
1269
1270         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen,
1271                                  MSG_CMSG_COMPAT);
1272 #ifdef CONFIG_NETFILTER
1273         /* we need to exclude all possible ENOPROTOOPTs except default case */
1274         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1275                 int len;
1276
1277                 if (get_user(len, optlen))
1278                         return -EFAULT;
1279
1280                 err = compat_nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1281                 if (err >= 0)
1282                         err = put_user(len, optlen);
1283         }
1284 #endif
1285         return err;
1286 }
1287
1288 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1289 #endif
1290