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
  40 #include <net/sock.h>
  41 #include <net/snmp.h>
  42 #include <net/ipv6.h>
  43 #include <net/ndisc.h>
  44 #include <net/protocol.h>
  45 #include <net/transp_v6.h>
  46 #include <net/ip6_route.h>
  47 #include <net/addrconf.h>
  48 #include <net/inet_common.h>
  49 #include <net/tcp.h>
  50 #include <net/udp.h>
  51 #include <net/udplite.h>
  52 #include <net/xfrm.h>
  53 #include <net/compat.h>
  54
  55 #include <asm/uaccess.h>
  56
  57 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
  58
  59 struct ip6_ra_chain *ip6_ra_chain;
  60 DEFINE_RWLOCK(ip6_ra_lock);
  61
  62 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
  63 {
  64         struct ip6_ra_chain *ra, *new_ra, **rap;
  65
  66         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
  67         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
  68                 return -ENOPROTOOPT;
  69
  70         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
  71
  72         write_lock_bh(&ip6_ra_lock);
  73         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
  74                 if (ra->sk == sk) {
  75                         if (sel>=0) {
  76                                 write_unlock_bh(&ip6_ra_lock);
  77                                 kfree(new_ra);
  78                                 return -EADDRINUSE;
  79                         }
  80
  81                         *rap = ra->next;
  82                         write_unlock_bh(&ip6_ra_lock);
  83
  84                         if (ra->destructor)
  85                                 ra->destructor(sk);
  86                         sock_put(sk);
  87                         kfree(ra);
  88                         return 0;
  89                 }
  90         }
  91         if (new_ra == NULL) {
  92                 write_unlock_bh(&ip6_ra_lock);
  93                 return -ENOBUFS;
  94         }
  95         new_ra->sk = sk;
  96         new_ra->sel = sel;
  97         new_ra->destructor = destructor;
  98         new_ra->next = ra;
  99         *rap = new_ra;
 100         sock_hold(sk);
 101         write_unlock_bh(&ip6_ra_lock);
 102         return 0;
 103 }
 104
 105 static
 106 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
 107                                            struct ipv6_txoptions *opt)
 108 {
 109         if (inet_sk(sk)->is_icsk) {
 110                 if (opt &&
 111                     !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
 112                     inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
 113                         struct inet_connection_sock *icsk = inet_csk(sk);
 114                         icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
 115                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 116                 }
 117                 opt = xchg(&inet6_sk(sk)->opt, opt);
 118         } else {
 119                 write_lock(&sk->sk_dst_lock);
 120                 opt = xchg(&inet6_sk(sk)->opt, opt);
 121                 write_unlock(&sk->sk_dst_lock);
 122         }
 123         sk_dst_reset(sk);
 124
 125         return opt;
 126 }
 127
 128 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 129                     char __user *optval, int optlen)
 130 {
 131         struct ipv6_pinfo *np = inet6_sk(sk);
 132         struct net *net = sock_net(sk);
 133         int val, valbool;
 134         int retv = -ENOPROTOOPT;
 135
 136         if (optval == NULL)
 137                 val=0;
 138         else {
 139                 if (optlen >= sizeof(int)) {
 140                         if (get_user(val, (int __user *) optval))
 141                                 return -EFAULT;
 142                 } else
 143                         val = 0;
 144         }
 145
 146         valbool = (val!=0);
 147
 148         if (ip6_mroute_opt(optname))
 149                 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
 150
 151         lock_sock(sk);
 152
 153         switch (optname) {
 154
 155         case IPV6_ADDRFORM:
 156                 if (optlen < sizeof(int))
 157                         goto e_inval;
 158                 if (val == PF_INET) {
 159                         struct ipv6_txoptions *opt;
 160                         struct sk_buff *pktopt;
 161
 162                         if (sk->sk_type == SOCK_RAW)
 163                                 break;
 164
 165                         if (sk->sk_protocol == IPPROTO_UDP ||
 166                             sk->sk_protocol == IPPROTO_UDPLITE) {
 167                                 struct udp_sock *up = udp_sk(sk);
 168                                 if (up->pending == AF_INET6) {
 169                                         retv = -EBUSY;
 170                                         break;
 171                                 }
 172                         } else if (sk->sk_protocol != IPPROTO_TCP)
 173                                 break;
 174
 175                         if (sk->sk_state != TCP_ESTABLISHED) {
 176                                 retv = -ENOTCONN;
 177                                 break;
 178                         }
 179
 180                         if (ipv6_only_sock(sk) ||
 181                             !ipv6_addr_v4mapped(&np->daddr)) {
 182                                 retv = -EADDRNOTAVAIL;
 183                                 break;
 184                         }
 185
 186                         fl6_free_socklist(sk);
 187                         ipv6_sock_mc_close(sk);
 188
 189                         /*
 190                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
 191                          * remove it from the refcnt debug socks count in the
 192                          * original family...
 193                          */
 194                         sk_refcnt_debug_dec(sk);
 195
 196                         if (sk->sk_protocol == IPPROTO_TCP) {
 197                                 struct inet_connection_sock *icsk = inet_csk(sk);
 198                                 local_bh_disable();
 199                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 200                                 sock_prot_inuse_add(net, &tcp_prot, 1);
 201                                 local_bh_enable();
 202                                 sk->sk_prot = &tcp_prot;
 203                                 icsk->icsk_af_ops = &ipv4_specific;
 204                                 sk->sk_socket->ops = &inet_stream_ops;
 205                                 sk->sk_family = PF_INET;
 206                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 207                         } else {
 208                                 struct proto *prot = &udp_prot;
 209
 210                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
 211                                         prot = &udplite_prot;
 212                                 local_bh_disable();
 213                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 214                                 sock_prot_inuse_add(net, prot, 1);
 215                                 local_bh_enable();
 216                                 sk->sk_prot = prot;
 217                                 sk->sk_socket->ops = &inet_dgram_ops;
 218                                 sk->sk_family = PF_INET;
 219                         }
 220                         opt = xchg(&np->opt, NULL);
 221                         if (opt)
 222                                 sock_kfree_s(sk, opt, opt->tot_len);
 223                         pktopt = xchg(&np->pktoptions, NULL);
 224                         if (pktopt)
 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)->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                 np->tclass = val;
 323                 retv = 0;
 324                 break;
 325
 326         case IPV6_RECVTCLASS:
 327                 if (optlen < sizeof(int))
 328                         goto e_inval;
 329                 np->rxopt.bits.rxtclass = valbool;
 330                 retv = 0;
 331                 break;
 332
 333         case IPV6_FLOWINFO:
 334                 if (optlen < sizeof(int))
 335                         goto e_inval;
 336                 np->rxopt.bits.rxflow = valbool;
 337                 retv = 0;
 338                 break;
 339
 340         case IPV6_HOPOPTS:
 341         case IPV6_RTHDRDSTOPTS:
 342         case IPV6_RTHDR:
 343         case IPV6_DSTOPTS:
 344         {
 345                 struct ipv6_txoptions *opt;
 346
 347                 /* remove any sticky options header with a zero option
 348                  * length, per RFC3542.
 349                  */
 350                 if (optlen == 0)
 351                         optval = NULL;
 352                 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
 353                          optlen & 0x7 || optlen > 8 * 255)
 354                         goto e_inval;
 355
 356                 /* hop-by-hop / destination options are privileged option */
 357                 retv = -EPERM;
 358                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
 359                         break;
 360
 361                 opt = ipv6_renew_options(sk, np->opt, optname,
 362                                          (struct ipv6_opt_hdr __user *)optval,
 363                                          optlen);
 364                 if (IS_ERR(opt)) {
 365                         retv = PTR_ERR(opt);
 366                         break;
 367                 }
 368
 369                 /* routing header option needs extra check */
 370                 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
 371                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
 372                         switch (rthdr->type) {
 373 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 374                         case IPV6_SRCRT_TYPE_2:
 375                                 break;
 376 #endif
 377                         default:
 378                                 goto sticky_done;
 379                         }
 380
 381                         if ((rthdr->hdrlen & 1) ||
 382                             (rthdr->hdrlen >> 1) != rthdr->segments_left)
 383                                 goto sticky_done;
 384                 }
 385
 386                 retv = 0;
 387                 opt = ipv6_update_options(sk, opt);
 388 sticky_done:
 389                 if (opt)
 390                         sock_kfree_s(sk, opt, opt->tot_len);
 391                 break;
 392         }
 393
 394         case IPV6_2292PKTOPTIONS:
 395         {
 396                 struct ipv6_txoptions *opt = NULL;
 397                 struct msghdr msg;
 398                 struct flowi fl;
 399                 int junk;
 400
 401                 fl.fl6_flowlabel = 0;
 402                 fl.oif = sk->sk_bound_dev_if;
 403
 404                 if (optlen == 0)
 405                         goto update;
 406
 407                 /* 1K is probably excessive
 408                  * 1K is surely not enough, 2K per standard header is 16K.
 409                  */
 410                 retv = -EINVAL;
 411                 if (optlen > 64*1024)
 412                         break;
 413
 414                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 415                 retv = -ENOBUFS;
 416                 if (opt == NULL)
 417                         break;
 418
 419                 memset(opt, 0, sizeof(*opt));
 420                 opt->tot_len = sizeof(*opt) + optlen;
 421                 retv = -EFAULT;
 422                 if (copy_from_user(opt+1, optval, optlen))
 423                         goto done;
 424
 425                 msg.msg_controllen = optlen;
 426                 msg.msg_control = (void*)(opt+1);
 427
 428                 retv = datagram_send_ctl(net, &msg, &fl, opt, &junk, &junk);
 429                 if (retv)
 430                         goto done;
 431 update:
 432                 retv = 0;
 433                 opt = ipv6_update_options(sk, opt);
 434 done:
 435                 if (opt)
 436                         sock_kfree_s(sk, opt, opt->tot_len);
 437                 break;
 438         }
 439         case IPV6_UNICAST_HOPS:
 440                 if (optlen < sizeof(int))
 441                         goto e_inval;
 442                 if (val > 255 || val < -1)
 443                         goto e_inval;
 444                 np->hop_limit = val;
 445                 retv = 0;
 446                 break;
 447
 448         case IPV6_MULTICAST_HOPS:
 449                 if (sk->sk_type == SOCK_STREAM)
 450                         break;
 451                 if (optlen < sizeof(int))
 452                         goto e_inval;
 453                 if (val > 255 || val < -1)
 454                         goto e_inval;
 455                 np->mcast_hops = val;
 456                 retv = 0;
 457                 break;
 458
 459         case IPV6_MULTICAST_LOOP:
 460                 if (optlen < sizeof(int))
 461                         goto e_inval;
 462                 if (val != valbool)
 463                         goto e_inval;
 464                 np->mc_loop = valbool;
 465                 retv = 0;
 466                 break;
 467
 468         case IPV6_MULTICAST_IF:
 469                 if (sk->sk_type == SOCK_STREAM)
 470                         break;
 471                 if (optlen < sizeof(int))
 472                         goto e_inval;
 473
 474                 if (val) {
 475                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
 476                                 goto e_inval;
 477
 478                         if (__dev_get_by_index(net, val) == NULL) {
 479                                 retv = -ENODEV;
 480                                 break;
 481                         }
 482                 }
 483                 np->mcast_oif = val;
 484                 retv = 0;
 485                 break;
 486         case IPV6_ADD_MEMBERSHIP:
 487         case IPV6_DROP_MEMBERSHIP:
 488         {
 489                 struct ipv6_mreq mreq;
 490
 491                 if (optlen < sizeof(struct ipv6_mreq))
 492                         goto e_inval;
 493
 494                 retv = -EPROTO;
 495                 if (inet_sk(sk)->is_icsk)
 496                         break;
 497
 498                 retv = -EFAULT;
 499                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 500                         break;
 501
 502                 if (optname == IPV6_ADD_MEMBERSHIP)
 503                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 504                 else
 505                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 506                 break;
 507         }
 508         case IPV6_JOIN_ANYCAST:
 509         case IPV6_LEAVE_ANYCAST:
 510         {
 511                 struct ipv6_mreq mreq;
 512
 513                 if (optlen < sizeof(struct ipv6_mreq))
 514                         goto e_inval;
 515
 516                 retv = -EFAULT;
 517                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 518                         break;
 519
 520                 if (optname == IPV6_JOIN_ANYCAST)
 521                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 522                 else
 523                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 524                 break;
 525         }
 526         case MCAST_JOIN_GROUP:
 527         case MCAST_LEAVE_GROUP:
 528         {
 529                 struct group_req greq;
 530                 struct sockaddr_in6 *psin6;
 531
 532                 if (optlen < sizeof(struct group_req))
 533                         goto e_inval;
 534
 535                 retv = -EFAULT;
 536                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
 537                         break;
 538                 if (greq.gr_group.ss_family != AF_INET6) {
 539                         retv = -EADDRNOTAVAIL;
 540                         break;
 541                 }
 542                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 543                 if (optname == MCAST_JOIN_GROUP)
 544                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
 545                                 &psin6->sin6_addr);
 546                 else
 547                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
 548                                 &psin6->sin6_addr);
 549                 break;
 550         }
 551         case MCAST_JOIN_SOURCE_GROUP:
 552         case MCAST_LEAVE_SOURCE_GROUP:
 553         case MCAST_BLOCK_SOURCE:
 554         case MCAST_UNBLOCK_SOURCE:
 555         {
 556                 struct group_source_req greqs;
 557                 int omode, add;
 558
 559                 if (optlen < sizeof(struct group_source_req))
 560                         goto e_inval;
 561                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 562                         retv = -EFAULT;
 563                         break;
 564                 }
 565                 if (greqs.gsr_group.ss_family != AF_INET6 ||
 566                     greqs.gsr_source.ss_family != AF_INET6) {
 567                         retv = -EADDRNOTAVAIL;
 568                         break;
 569                 }
 570                 if (optname == MCAST_BLOCK_SOURCE) {
 571                         omode = MCAST_EXCLUDE;
 572                         add = 1;
 573                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 574                         omode = MCAST_EXCLUDE;
 575                         add = 0;
 576                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 577                         struct sockaddr_in6 *psin6;
 578
 579                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 580                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
 581                                 &psin6->sin6_addr);
 582                         /* prior join w/ different source is ok */
 583                         if (retv && retv != -EADDRINUSE)
 584                                 break;
 585                         omode = MCAST_INCLUDE;
 586                         add = 1;
 587                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 588                         omode = MCAST_INCLUDE;
 589                         add = 0;
 590                 }
 591                 retv = ip6_mc_source(add, omode, sk, &greqs);
 592                 break;
 593         }
 594         case MCAST_MSFILTER:
 595         {
 596                 extern int sysctl_mld_max_msf;
 597                 struct group_filter *gsf;
 598
 599                 if (optlen < GROUP_FILTER_SIZE(0))
 600                         goto e_inval;
 601                 if (optlen > sysctl_optmem_max) {
 602                         retv = -ENOBUFS;
 603                         break;
 604                 }
 605                 gsf = kmalloc(optlen,GFP_KERNEL);
 606                 if (!gsf) {
 607                         retv = -ENOBUFS;
 608                         break;
 609                 }
 610                 retv = -EFAULT;
 611                 if (copy_from_user(gsf, optval, optlen)) {
 612                         kfree(gsf);
 613                         break;
 614                 }
 615                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 616                 if (gsf->gf_numsrc >= 0x1ffffffU ||
 617                     gsf->gf_numsrc > sysctl_mld_max_msf) {
 618                         kfree(gsf);
 619                         retv = -ENOBUFS;
 620                         break;
 621                 }
 622                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 623                         kfree(gsf);
 624                         retv = -EINVAL;
 625                         break;
 626                 }
 627                 retv = ip6_mc_msfilter(sk, gsf);
 628                 kfree(gsf);
 629
 630                 break;
 631         }
 632         case IPV6_ROUTER_ALERT:
 633                 if (optlen < sizeof(int))
 634                         goto e_inval;
 635                 retv = ip6_ra_control(sk, val, NULL);
 636                 break;
 637         case IPV6_MTU_DISCOVER:
 638                 if (optlen < sizeof(int))
 639                         goto e_inval;
 640                 if (val<0 || val>3)
 641                         goto e_inval;
 642                 np->pmtudisc = val;
 643                 retv = 0;
 644                 break;
 645         case IPV6_MTU:
 646                 if (optlen < sizeof(int))
 647                         goto e_inval;
 648                 if (val && val < IPV6_MIN_MTU)
 649                         goto e_inval;
 650                 np->frag_size = val;
 651                 retv = 0;
 652                 break;
 653         case IPV6_RECVERR:
 654                 if (optlen < sizeof(int))
 655                         goto e_inval;
 656                 np->recverr = valbool;
 657                 if (!val)
 658                         skb_queue_purge(&sk->sk_error_queue);
 659                 retv = 0;
 660                 break;
 661         case IPV6_FLOWINFO_SEND:
 662                 if (optlen < sizeof(int))
 663                         goto e_inval;
 664                 np->sndflow = valbool;
 665                 retv = 0;
 666                 break;
 667         case IPV6_FLOWLABEL_MGR:
 668                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
 669                 break;
 670         case IPV6_IPSEC_POLICY:
 671         case IPV6_XFRM_POLICY:
 672                 retv = -EPERM;
 673                 if (!capable(CAP_NET_ADMIN))
 674                         break;
 675                 retv = xfrm_user_policy(sk, optname, optval, optlen);
 676                 break;
 677
 678         case IPV6_ADDR_PREFERENCES:
 679             {
 680                 unsigned int pref = 0;
 681                 unsigned int prefmask = ~0;
 682
 683                 if (optlen < sizeof(int))
 684                         goto e_inval;
 685
 686                 retv = -EINVAL;
 687
 688                 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
 689                 switch (val & (IPV6_PREFER_SRC_PUBLIC|
 690                                IPV6_PREFER_SRC_TMP|
 691                                IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
 692                 case IPV6_PREFER_SRC_PUBLIC:
 693                         pref |= IPV6_PREFER_SRC_PUBLIC;
 694                         break;
 695                 case IPV6_PREFER_SRC_TMP:
 696                         pref |= IPV6_PREFER_SRC_TMP;
 697                         break;
 698                 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
 699                         break;
 700                 case 0:
 701                         goto pref_skip_pubtmp;
 702                 default:
 703                         goto e_inval;
 704                 }
 705
 706                 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
 707                               IPV6_PREFER_SRC_TMP);
 708 pref_skip_pubtmp:
 709
 710                 /* check HOME/COA conflicts */
 711                 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
 712                 case IPV6_PREFER_SRC_HOME:
 713                         break;
 714                 case IPV6_PREFER_SRC_COA:
 715                         pref |= IPV6_PREFER_SRC_COA;
 716                 case 0:
 717                         goto pref_skip_coa;
 718                 default:
 719                         goto e_inval;
 720                 }
 721
 722                 prefmask &= ~IPV6_PREFER_SRC_COA;
 723 pref_skip_coa:
 724
 725                 /* check CGA/NONCGA conflicts */
 726                 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
 727                 case IPV6_PREFER_SRC_CGA:
 728                 case IPV6_PREFER_SRC_NONCGA:
 729                 case 0:
 730                         break;
 731                 default:
 732                         goto e_inval;
 733                 }
 734
 735                 np->srcprefs = (np->srcprefs & prefmask) | pref;
 736                 retv = 0;
 737
 738                 break;
 739             }
 740         }
 741
 742         release_sock(sk);
 743
 744         return retv;
 745
 746 e_inval:
 747         release_sock(sk);
 748         return -EINVAL;
 749 }
 750
 751 int ipv6_setsockopt(struct sock *sk, int level, int optname,
 752                     char __user *optval, int optlen)
 753 {
 754         int err;
 755
 756         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 757                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 758
 759         if (level != SOL_IPV6)
 760                 return -ENOPROTOOPT;
 761
 762         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 763 #ifdef CONFIG_NETFILTER
 764         /* we need to exclude all possible ENOPROTOOPTs except default case */
 765         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 766                         optname != IPV6_XFRM_POLICY) {
 767                 lock_sock(sk);
 768                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
 769                                 optlen);
 770                 release_sock(sk);
 771         }
 772 #endif
 773         return err;
 774 }
 775
 776 EXPORT_SYMBOL(ipv6_setsockopt);
 777
 778 #ifdef CONFIG_COMPAT
 779 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
 780                            char __user *optval, int optlen)
 781 {
 782         int err;
 783
 784         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
 785                 if (udp_prot.compat_setsockopt != NULL)
 786                         return udp_prot.compat_setsockopt(sk, level, optname,
 787                                                           optval, optlen);
 788                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 789         }
 790
 791         if (level != SOL_IPV6)
 792                 return -ENOPROTOOPT;
 793
 794         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
 795                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
 796                         ipv6_setsockopt);
 797
 798         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 799 #ifdef CONFIG_NETFILTER
 800         /* we need to exclude all possible ENOPROTOOPTs except default case */
 801         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 802             optname != IPV6_XFRM_POLICY) {
 803                 lock_sock(sk);
 804                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
 805                                            optval, optlen);
 806                 release_sock(sk);
 807         }
 808 #endif
 809         return err;
 810 }
 811
 812 EXPORT_SYMBOL(compat_ipv6_setsockopt);
 813 #endif
 814
 815 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
 816                                   int optname, char __user *optval, int len)
 817 {
 818         struct ipv6_opt_hdr *hdr;
 819
 820         if (!opt)
 821                 return 0;
 822
 823         switch(optname) {
 824         case IPV6_HOPOPTS:
 825                 hdr = opt->hopopt;
 826                 break;
 827         case IPV6_RTHDRDSTOPTS:
 828                 hdr = opt->dst0opt;
 829                 break;
 830         case IPV6_RTHDR:
 831                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
 832                 break;
 833         case IPV6_DSTOPTS:
 834                 hdr = opt->dst1opt;
 835                 break;
 836         default:
 837                 return -EINVAL; /* should not happen */
 838         }
 839
 840         if (!hdr)
 841                 return 0;
 842
 843         len = min_t(unsigned int, len, ipv6_optlen(hdr));
 844         if (copy_to_user(optval, hdr, len))
 845                 return -EFAULT;
 846         return len;
 847 }
 848
 849 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 850                     char __user *optval, int __user *optlen)
 851 {
 852         struct ipv6_pinfo *np = inet6_sk(sk);
 853         int len;
 854         int val;
 855
 856         if (ip6_mroute_opt(optname))
 857                 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
 858
 859         if (get_user(len, optlen))
 860                 return -EFAULT;
 861         switch (optname) {
 862         case IPV6_ADDRFORM:
 863                 if (sk->sk_protocol != IPPROTO_UDP &&
 864                     sk->sk_protocol != IPPROTO_UDPLITE &&
 865                     sk->sk_protocol != IPPROTO_TCP)
 866                         return -ENOPROTOOPT;
 867                 if (sk->sk_state != TCP_ESTABLISHED)
 868                         return -ENOTCONN;
 869                 val = sk->sk_family;
 870                 break;
 871         case MCAST_MSFILTER:
 872         {
 873                 struct group_filter gsf;
 874                 int err;
 875
 876                 if (len < GROUP_FILTER_SIZE(0))
 877                         return -EINVAL;
 878                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
 879                         return -EFAULT;
 880                 if (gsf.gf_group.ss_family != AF_INET6)
 881                         return -EADDRNOTAVAIL;
 882                 lock_sock(sk);
 883                 err = ip6_mc_msfget(sk, &gsf,
 884                         (struct group_filter __user *)optval, optlen);
 885                 release_sock(sk);
 886                 return err;
 887         }
 888
 889         case IPV6_2292PKTOPTIONS:
 890         {
 891                 struct msghdr msg;
 892                 struct sk_buff *skb;
 893
 894                 if (sk->sk_type != SOCK_STREAM)
 895                         return -ENOPROTOOPT;
 896
 897                 msg.msg_control = optval;
 898                 msg.msg_controllen = len;
 899                 msg.msg_flags = 0;
 900
 901                 lock_sock(sk);
 902                 skb = np->pktoptions;
 903                 if (skb)
 904                         atomic_inc(&skb->users);
 905                 release_sock(sk);
 906
 907                 if (skb) {
 908                         int err = datagram_recv_ctl(sk, &msg, skb);
 909                         kfree_skb(skb);
 910                         if (err)
 911                                 return err;
 912                 } else {
 913                         if (np->rxopt.bits.rxinfo) {
 914                                 struct in6_pktinfo src_info;
 915                                 src_info.ipi6_ifindex = np->mcast_oif;
 916                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 917                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 918                         }
 919                         if (np->rxopt.bits.rxhlim) {
 920                                 int hlim = np->mcast_hops;
 921                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 922                         }
 923                         if (np->rxopt.bits.rxoinfo) {
 924                                 struct in6_pktinfo src_info;
 925                                 src_info.ipi6_ifindex = np->mcast_oif;
 926                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 927                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 928                         }
 929                         if (np->rxopt.bits.rxohlim) {
 930                                 int hlim = np->mcast_hops;
 931                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 932                         }
 933                 }
 934                 len -= msg.msg_controllen;
 935                 return put_user(len, optlen);
 936         }
 937         case IPV6_MTU:
 938         {
 939                 struct dst_entry *dst;
 940                 val = 0;
 941                 lock_sock(sk);
 942                 dst = sk_dst_get(sk);
 943                 if (dst) {
 944                         val = dst_mtu(dst);
 945                         dst_release(dst);
 946                 }
 947                 release_sock(sk);
 948                 if (!val)
 949                         return -ENOTCONN;
 950                 break;
 951         }
 952
 953         case IPV6_V6ONLY:
 954                 val = np->ipv6only;
 955                 break;
 956
 957         case IPV6_RECVPKTINFO:
 958                 val = np->rxopt.bits.rxinfo;
 959                 break;
 960
 961         case IPV6_2292PKTINFO:
 962                 val = np->rxopt.bits.rxoinfo;
 963                 break;
 964
 965         case IPV6_RECVHOPLIMIT:
 966                 val = np->rxopt.bits.rxhlim;
 967                 break;
 968
 969         case IPV6_2292HOPLIMIT:
 970                 val = np->rxopt.bits.rxohlim;
 971                 break;
 972
 973         case IPV6_RECVRTHDR:
 974                 val = np->rxopt.bits.srcrt;
 975                 break;
 976
 977         case IPV6_2292RTHDR:
 978                 val = np->rxopt.bits.osrcrt;
 979                 break;
 980
 981         case IPV6_HOPOPTS:
 982         case IPV6_RTHDRDSTOPTS:
 983         case IPV6_RTHDR:
 984         case IPV6_DSTOPTS:
 985         {
 986
 987                 lock_sock(sk);
 988                 len = ipv6_getsockopt_sticky(sk, np->opt,
 989                                              optname, optval, len);
 990                 release_sock(sk);
 991                 /* check if ipv6_getsockopt_sticky() returns err code */
 992                 if (len < 0)
 993                         return len;
 994                 return put_user(len, optlen);
 995         }
 996
 997         case IPV6_RECVHOPOPTS:
 998                 val = np->rxopt.bits.hopopts;
 999                 break;
1000
1001         case IPV6_2292HOPOPTS:
1002                 val = np->rxopt.bits.ohopopts;
1003                 break;
1004
1005         case IPV6_RECVDSTOPTS:
1006                 val = np->rxopt.bits.dstopts;
1007                 break;
1008
1009         case IPV6_2292DSTOPTS:
1010                 val = np->rxopt.bits.odstopts;
1011                 break;
1012
1013         case IPV6_TCLASS:
1014                 val = np->tclass;
1015                 if (val < 0)
1016                         val = 0;
1017                 break;
1018
1019         case IPV6_RECVTCLASS:
1020                 val = np->rxopt.bits.rxtclass;
1021                 break;
1022
1023         case IPV6_FLOWINFO:
1024                 val = np->rxopt.bits.rxflow;
1025                 break;
1026
1027         case IPV6_UNICAST_HOPS:
1028         case IPV6_MULTICAST_HOPS:
1029         {
1030                 struct dst_entry *dst;
1031
1032                 if (optname == IPV6_UNICAST_HOPS)
1033                         val = np->hop_limit;
1034                 else
1035                         val = np->mcast_hops;
1036
1037                 dst = sk_dst_get(sk);
1038                 if (dst) {
1039                         if (val < 0)
1040                                 val = ip6_dst_hoplimit(dst);
1041                         dst_release(dst);
1042                 }
1043                 if (val < 0)
1044                         val = ipv6_devconf.hop_limit;
1045                 break;
1046         }
1047
1048         case IPV6_MULTICAST_LOOP:
1049                 val = np->mc_loop;
1050                 break;
1051
1052         case IPV6_MULTICAST_IF:
1053                 val = np->mcast_oif;
1054                 break;
1055
1056         case IPV6_MTU_DISCOVER:
1057                 val = np->pmtudisc;
1058                 break;
1059
1060         case IPV6_RECVERR:
1061                 val = np->recverr;
1062                 break;
1063
1064         case IPV6_FLOWINFO_SEND:
1065                 val = np->sndflow;
1066                 break;
1067
1068         case IPV6_ADDR_PREFERENCES:
1069                 val = 0;
1070
1071                 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1072                         val |= IPV6_PREFER_SRC_TMP;
1073                 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1074                         val |= IPV6_PREFER_SRC_PUBLIC;
1075                 else {
1076                         /* XXX: should we return system default? */
1077                         val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1078                 }
1079
1080                 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1081                         val |= IPV6_PREFER_SRC_COA;
1082                 else
1083                         val |= IPV6_PREFER_SRC_HOME;
1084                 break;
1085
1086         default:
1087                 return -ENOPROTOOPT;
1088         }
1089         len = min_t(unsigned int, sizeof(int), len);
1090         if(put_user(len, optlen))
1091                 return -EFAULT;
1092         if(copy_to_user(optval,&val,len))
1093                 return -EFAULT;
1094         return 0;
1095 }
1096
1097 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1098                     char __user *optval, int __user *optlen)
1099 {
1100         int err;
1101
1102         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1103                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1104
1105         if(level != SOL_IPV6)
1106                 return -ENOPROTOOPT;
1107
1108         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1109 #ifdef CONFIG_NETFILTER
1110         /* we need to exclude all possible ENOPROTOOPTs except default case */
1111         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1112                 int len;
1113
1114                 if (get_user(len, optlen))
1115                         return -EFAULT;
1116
1117                 lock_sock(sk);
1118                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1119                                 &len);
1120                 release_sock(sk);
1121                 if (err >= 0)
1122                         err = put_user(len, optlen);
1123         }
1124 #endif
1125         return err;
1126 }
1127
1128 EXPORT_SYMBOL(ipv6_getsockopt);
1129
1130 #ifdef CONFIG_COMPAT
1131 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1132                            char __user *optval, int __user *optlen)
1133 {
1134         int err;
1135
1136         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1137                 if (udp_prot.compat_getsockopt != NULL)
1138                         return udp_prot.compat_getsockopt(sk, level, optname,
1139                                                           optval, optlen);
1140                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1141         }
1142
1143         if (level != SOL_IPV6)
1144                 return -ENOPROTOOPT;
1145
1146         if (optname == MCAST_MSFILTER)
1147                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1148                         ipv6_getsockopt);
1149
1150         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1151 #ifdef CONFIG_NETFILTER
1152         /* we need to exclude all possible ENOPROTOOPTs except default case */
1153         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1154                 int len;
1155
1156                 if (get_user(len, optlen))
1157                         return -EFAULT;
1158
1159                 lock_sock(sk);
1160                 err = compat_nf_getsockopt(sk, PF_INET6,
1161                                            optname, optval, &len);
1162                 release_sock(sk);
1163                 if (err >= 0)
1164                         err = put_user(len, optlen);
1165         }
1166 #endif
1167         return err;
1168 }
1169
1170 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1171 #endif
1172