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(&inet6_sk(sk)->opt, opt);
 114         } else {
 115                 spin_lock(&sk->sk_dst_lock);
 116                 opt = xchg(&inet6_sk(sk)->opt, opt);
 117                 spin_unlock(&sk->sk_dst_lock);
 118         }
 119         sk_dst_reset(sk);
 120
 121         return opt;
 122 }
 123
 124 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 125                     char __user *optval, unsigned int optlen)
 126 {
 127         struct ipv6_pinfo *np = inet6_sk(sk);
 128         struct net *net = sock_net(sk);
 129         int val, valbool;
 130         int retv = -ENOPROTOOPT;
 131
 132         if (optval == NULL)
 133                 val=0;
 134         else {
 135                 if (optlen >= sizeof(int)) {
 136                         if (get_user(val, (int __user *) optval))
 137                                 return -EFAULT;
 138                 } else
 139                         val = 0;
 140         }
 141
 142         valbool = (val!=0);
 143
 144         if (ip6_mroute_opt(optname))
 145                 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
 146
 147         lock_sock(sk);
 148
 149         switch (optname) {
 150
 151         case IPV6_ADDRFORM:
 152                 if (optlen < sizeof(int))
 153                         goto e_inval;
 154                 if (val == PF_INET) {
 155                         struct ipv6_txoptions *opt;
 156                         struct sk_buff *pktopt;
 157
 158                         if (sk->sk_type == SOCK_RAW)
 159                                 break;
 160
 161                         if (sk->sk_protocol == IPPROTO_UDP ||
 162                             sk->sk_protocol == IPPROTO_UDPLITE) {
 163                                 struct udp_sock *up = udp_sk(sk);
 164                                 if (up->pending == AF_INET6) {
 165                                         retv = -EBUSY;
 166                                         break;
 167                                 }
 168                         } else if (sk->sk_protocol != IPPROTO_TCP)
 169                                 break;
 170
 171                         if (sk->sk_state != TCP_ESTABLISHED) {
 172                                 retv = -ENOTCONN;
 173                                 break;
 174                         }
 175
 176                         if (ipv6_only_sock(sk) ||
 177                             !ipv6_addr_v4mapped(&np->daddr)) {
 178                                 retv = -EADDRNOTAVAIL;
 179                                 break;
 180                         }
 181
 182                         fl6_free_socklist(sk);
 183                         ipv6_sock_mc_close(sk);
 184
 185                         /*
 186                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
 187                          * remove it from the refcnt debug socks count in the
 188                          * original family...
 189                          */
 190                         sk_refcnt_debug_dec(sk);
 191
 192                         if (sk->sk_protocol == IPPROTO_TCP) {
 193                                 struct inet_connection_sock *icsk = inet_csk(sk);
 194                                 local_bh_disable();
 195                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 196                                 sock_prot_inuse_add(net, &tcp_prot, 1);
 197                                 local_bh_enable();
 198                                 sk->sk_prot = &tcp_prot;
 199                                 icsk->icsk_af_ops = &ipv4_specific;
 200                                 sk->sk_socket->ops = &inet_stream_ops;
 201                                 sk->sk_family = PF_INET;
 202                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 203                         } else {
 204                                 struct proto *prot = &udp_prot;
 205
 206                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
 207                                         prot = &udplite_prot;
 208                                 local_bh_disable();
 209                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 210                                 sock_prot_inuse_add(net, prot, 1);
 211                                 local_bh_enable();
 212                                 sk->sk_prot = prot;
 213                                 sk->sk_socket->ops = &inet_dgram_ops;
 214                                 sk->sk_family = PF_INET;
 215                         }
 216                         opt = xchg(&np->opt, NULL);
 217                         if (opt)
 218                                 sock_kfree_s(sk, opt, opt->tot_len);
 219                         pktopt = xchg(&np->pktoptions, NULL);
 220                         kfree_skb(pktopt);
 221
 222                         sk->sk_destruct = inet_sock_destruct;
 223                         /*
 224                          * ... and add it to the refcnt debug socks count
 225                          * in the new family. -acme
 226                          */
 227                         sk_refcnt_debug_inc(sk);
 228                         module_put(THIS_MODULE);
 229                         retv = 0;
 230                         break;
 231                 }
 232                 goto e_inval;
 233
 234         case IPV6_V6ONLY:
 235                 if (optlen < sizeof(int) ||
 236                     inet_sk(sk)->inet_num)
 237                         goto e_inval;
 238                 np->ipv6only = valbool;
 239                 retv = 0;
 240                 break;
 241
 242         case IPV6_RECVPKTINFO:
 243                 if (optlen < sizeof(int))
 244                         goto e_inval;
 245                 np->rxopt.bits.rxinfo = valbool;
 246                 retv = 0;
 247                 break;
 248
 249         case IPV6_2292PKTINFO:
 250                 if (optlen < sizeof(int))
 251                         goto e_inval;
 252                 np->rxopt.bits.rxoinfo = valbool;
 253                 retv = 0;
 254                 break;
 255
 256         case IPV6_RECVHOPLIMIT:
 257                 if (optlen < sizeof(int))
 258                         goto e_inval;
 259                 np->rxopt.bits.rxhlim = valbool;
 260                 retv = 0;
 261                 break;
 262
 263         case IPV6_2292HOPLIMIT:
 264                 if (optlen < sizeof(int))
 265                         goto e_inval;
 266                 np->rxopt.bits.rxohlim = valbool;
 267                 retv = 0;
 268                 break;
 269
 270         case IPV6_RECVRTHDR:
 271                 if (optlen < sizeof(int))
 272                         goto e_inval;
 273                 np->rxopt.bits.srcrt = valbool;
 274                 retv = 0;
 275                 break;
 276
 277         case IPV6_2292RTHDR:
 278                 if (optlen < sizeof(int))
 279                         goto e_inval;
 280                 np->rxopt.bits.osrcrt = valbool;
 281                 retv = 0;
 282                 break;
 283
 284         case IPV6_RECVHOPOPTS:
 285                 if (optlen < sizeof(int))
 286                         goto e_inval;
 287                 np->rxopt.bits.hopopts = valbool;
 288                 retv = 0;
 289                 break;
 290
 291         case IPV6_2292HOPOPTS:
 292                 if (optlen < sizeof(int))
 293                         goto e_inval;
 294                 np->rxopt.bits.ohopopts = valbool;
 295                 retv = 0;
 296                 break;
 297
 298         case IPV6_RECVDSTOPTS:
 299                 if (optlen < sizeof(int))
 300                         goto e_inval;
 301                 np->rxopt.bits.dstopts = valbool;
 302                 retv = 0;
 303                 break;
 304
 305         case IPV6_2292DSTOPTS:
 306                 if (optlen < sizeof(int))
 307                         goto e_inval;
 308                 np->rxopt.bits.odstopts = valbool;
 309                 retv = 0;
 310                 break;
 311
 312         case IPV6_TCLASS:
 313                 if (optlen < sizeof(int))
 314                         goto e_inval;
 315                 if (val < -1 || val > 0xff)
 316                         goto e_inval;
 317                 /* RFC 3542, 6.5: default traffic class of 0x0 */
 318                 if (val == -1)
 319                         val = 0;
 320                 np->tclass = val;
 321                 retv = 0;
 322                 break;
 323
 324         case IPV6_RECVTCLASS:
 325                 if (optlen < sizeof(int))
 326                         goto e_inval;
 327                 np->rxopt.bits.rxtclass = valbool;
 328                 retv = 0;
 329                 break;
 330
 331         case IPV6_FLOWINFO:
 332                 if (optlen < sizeof(int))
 333                         goto e_inval;
 334                 np->rxopt.bits.rxflow = valbool;
 335                 retv = 0;
 336                 break;
 337
 338         case IPV6_RECVPATHMTU:
 339                 if (optlen < sizeof(int))
 340                         goto e_inval;
 341                 np->rxopt.bits.rxpmtu = valbool;
 342                 retv = 0;
 343                 break;
 344
 345         case IPV6_HOPOPTS:
 346         case IPV6_RTHDRDSTOPTS:
 347         case IPV6_RTHDR:
 348         case IPV6_DSTOPTS:
 349         {
 350                 struct ipv6_txoptions *opt;
 351
 352                 /* remove any sticky options header with a zero option
 353                  * length, per RFC3542.
 354                  */
 355                 if (optlen == 0)
 356                         optval = NULL;
 357                 else if (optval == NULL)
 358                         goto e_inval;
 359                 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
 360                          optlen & 0x7 || optlen > 8 * 255)
 361                         goto e_inval;
 362
 363                 /* hop-by-hop / destination options are privileged option */
 364                 retv = -EPERM;
 365                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
 366                         break;
 367
 368                 opt = ipv6_renew_options(sk, np->opt, optname,
 369                                          (struct ipv6_opt_hdr __user *)optval,
 370                                          optlen);
 371                 if (IS_ERR(opt)) {
 372                         retv = PTR_ERR(opt);
 373                         break;
 374                 }
 375
 376                 /* routing header option needs extra check */
 377                 retv = -EINVAL;
 378                 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
 379                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
 380                         switch (rthdr->type) {
 381 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 382                         case IPV6_SRCRT_TYPE_2:
 383                                 if (rthdr->hdrlen != 2 ||
 384                                     rthdr->segments_left != 1)
 385                                         goto sticky_done;
 386
 387                                 break;
 388 #endif
 389                         default:
 390                                 goto sticky_done;
 391                         }
 392                 }
 393
 394                 retv = 0;
 395                 opt = ipv6_update_options(sk, opt);
 396 sticky_done:
 397                 if (opt)
 398                         sock_kfree_s(sk, opt, opt->tot_len);
 399                 break;
 400         }
 401
 402         case IPV6_PKTINFO:
 403         {
 404                 struct in6_pktinfo pkt;
 405
 406                 if (optlen == 0)
 407                         goto e_inval;
 408                 else if (optlen < sizeof(struct in6_pktinfo) || optval == NULL)
 409                         goto e_inval;
 410
 411                 if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
 412                                 retv = -EFAULT;
 413                                 break;
 414                 }
 415                 if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
 416                         goto e_inval;
 417
 418                 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
 419                 ipv6_addr_copy(&np->sticky_pktinfo.ipi6_addr, &pkt.ipi6_addr);
 420                 retv = 0;
 421                 break;
 422         }
 423
 424         case IPV6_2292PKTOPTIONS:
 425         {
 426                 struct ipv6_txoptions *opt = NULL;
 427                 struct msghdr msg;
 428                 struct flowi fl;
 429                 int junk;
 430
 431                 fl.fl6_flowlabel = 0;
 432                 fl.oif = sk->sk_bound_dev_if;
 433                 fl.mark = sk->sk_mark;
 434
 435                 if (optlen == 0)
 436                         goto update;
 437
 438                 /* 1K is probably excessive
 439                  * 1K is surely not enough, 2K per standard header is 16K.
 440                  */
 441                 retv = -EINVAL;
 442                 if (optlen > 64*1024)
 443                         break;
 444
 445                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 446                 retv = -ENOBUFS;
 447                 if (opt == NULL)
 448                         break;
 449
 450                 memset(opt, 0, sizeof(*opt));
 451                 opt->tot_len = sizeof(*opt) + optlen;
 452                 retv = -EFAULT;
 453                 if (copy_from_user(opt+1, optval, optlen))
 454                         goto done;
 455
 456                 msg.msg_controllen = optlen;
 457                 msg.msg_control = (void*)(opt+1);
 458
 459                 retv = datagram_send_ctl(net, &msg, &fl, opt, &junk, &junk,
 460                                          &junk);
 461                 if (retv)
 462                         goto done;
 463 update:
 464                 retv = 0;
 465                 opt = ipv6_update_options(sk, opt);
 466 done:
 467                 if (opt)
 468                         sock_kfree_s(sk, opt, opt->tot_len);
 469                 break;
 470         }
 471         case IPV6_UNICAST_HOPS:
 472                 if (optlen < sizeof(int))
 473                         goto e_inval;
 474                 if (val > 255 || val < -1)
 475                         goto e_inval;
 476                 np->hop_limit = val;
 477                 retv = 0;
 478                 break;
 479
 480         case IPV6_MULTICAST_HOPS:
 481                 if (sk->sk_type == SOCK_STREAM)
 482                         break;
 483                 if (optlen < sizeof(int))
 484                         goto e_inval;
 485                 if (val > 255 || val < -1)
 486                         goto e_inval;
 487                 np->mcast_hops = val;
 488                 retv = 0;
 489                 break;
 490
 491         case IPV6_MULTICAST_LOOP:
 492                 if (optlen < sizeof(int))
 493                         goto e_inval;
 494                 if (val != valbool)
 495                         goto e_inval;
 496                 np->mc_loop = valbool;
 497                 retv = 0;
 498                 break;
 499
 500         case IPV6_MULTICAST_IF:
 501                 if (sk->sk_type == SOCK_STREAM)
 502                         break;
 503                 if (optlen < sizeof(int))
 504                         goto e_inval;
 505
 506                 if (val) {
 507                         struct net_device *dev;
 508
 509                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
 510                                 goto e_inval;
 511
 512                         dev = dev_get_by_index(net, val);
 513                         if (!dev) {
 514                                 retv = -ENODEV;
 515                                 break;
 516                         }
 517                         dev_put(dev);
 518                 }
 519                 np->mcast_oif = val;
 520                 retv = 0;
 521                 break;
 522         case IPV6_ADD_MEMBERSHIP:
 523         case IPV6_DROP_MEMBERSHIP:
 524         {
 525                 struct ipv6_mreq mreq;
 526
 527                 if (optlen < sizeof(struct ipv6_mreq))
 528                         goto e_inval;
 529
 530                 retv = -EPROTO;
 531                 if (inet_sk(sk)->is_icsk)
 532                         break;
 533
 534                 retv = -EFAULT;
 535                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 536                         break;
 537
 538                 if (optname == IPV6_ADD_MEMBERSHIP)
 539                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 540                 else
 541                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 542                 break;
 543         }
 544         case IPV6_JOIN_ANYCAST:
 545         case IPV6_LEAVE_ANYCAST:
 546         {
 547                 struct ipv6_mreq mreq;
 548
 549                 if (optlen < sizeof(struct ipv6_mreq))
 550                         goto e_inval;
 551
 552                 retv = -EFAULT;
 553                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 554                         break;
 555
 556                 if (optname == IPV6_JOIN_ANYCAST)
 557                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 558                 else
 559                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 560                 break;
 561         }
 562         case MCAST_JOIN_GROUP:
 563         case MCAST_LEAVE_GROUP:
 564         {
 565                 struct group_req greq;
 566                 struct sockaddr_in6 *psin6;
 567
 568                 if (optlen < sizeof(struct group_req))
 569                         goto e_inval;
 570
 571                 retv = -EFAULT;
 572                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
 573                         break;
 574                 if (greq.gr_group.ss_family != AF_INET6) {
 575                         retv = -EADDRNOTAVAIL;
 576                         break;
 577                 }
 578                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 579                 if (optname == MCAST_JOIN_GROUP)
 580                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
 581                                 &psin6->sin6_addr);
 582                 else
 583                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
 584                                 &psin6->sin6_addr);
 585                 break;
 586         }
 587         case MCAST_JOIN_SOURCE_GROUP:
 588         case MCAST_LEAVE_SOURCE_GROUP:
 589         case MCAST_BLOCK_SOURCE:
 590         case MCAST_UNBLOCK_SOURCE:
 591         {
 592                 struct group_source_req greqs;
 593                 int omode, add;
 594
 595                 if (optlen < sizeof(struct group_source_req))
 596                         goto e_inval;
 597                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 598                         retv = -EFAULT;
 599                         break;
 600                 }
 601                 if (greqs.gsr_group.ss_family != AF_INET6 ||
 602                     greqs.gsr_source.ss_family != AF_INET6) {
 603                         retv = -EADDRNOTAVAIL;
 604                         break;
 605                 }
 606                 if (optname == MCAST_BLOCK_SOURCE) {
 607                         omode = MCAST_EXCLUDE;
 608                         add = 1;
 609                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 610                         omode = MCAST_EXCLUDE;
 611                         add = 0;
 612                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 613                         struct sockaddr_in6 *psin6;
 614
 615                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 616                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
 617                                 &psin6->sin6_addr);
 618                         /* prior join w/ different source is ok */
 619                         if (retv && retv != -EADDRINUSE)
 620                                 break;
 621                         omode = MCAST_INCLUDE;
 622                         add = 1;
 623                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 624                         omode = MCAST_INCLUDE;
 625                         add = 0;
 626                 }
 627                 retv = ip6_mc_source(add, omode, sk, &greqs);
 628                 break;
 629         }
 630         case MCAST_MSFILTER:
 631         {
 632                 extern int sysctl_mld_max_msf;
 633                 struct group_filter *gsf;
 634
 635                 if (optlen < GROUP_FILTER_SIZE(0))
 636                         goto e_inval;
 637                 if (optlen > sysctl_optmem_max) {
 638                         retv = -ENOBUFS;
 639                         break;
 640                 }
 641                 gsf = kmalloc(optlen,GFP_KERNEL);
 642                 if (!gsf) {
 643                         retv = -ENOBUFS;
 644                         break;
 645                 }
 646                 retv = -EFAULT;
 647                 if (copy_from_user(gsf, optval, optlen)) {
 648                         kfree(gsf);
 649                         break;
 650                 }
 651                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 652                 if (gsf->gf_numsrc >= 0x1ffffffU ||
 653                     gsf->gf_numsrc > sysctl_mld_max_msf) {
 654                         kfree(gsf);
 655                         retv = -ENOBUFS;
 656                         break;
 657                 }
 658                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 659                         kfree(gsf);
 660                         retv = -EINVAL;
 661                         break;
 662                 }
 663                 retv = ip6_mc_msfilter(sk, gsf);
 664                 kfree(gsf);
 665
 666                 break;
 667         }
 668         case IPV6_ROUTER_ALERT:
 669                 if (optlen < sizeof(int))
 670                         goto e_inval;
 671                 retv = ip6_ra_control(sk, val);
 672                 break;
 673         case IPV6_MTU_DISCOVER:
 674                 if (optlen < sizeof(int))
 675                         goto e_inval;
 676                 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE)
 677                         goto e_inval;
 678                 np->pmtudisc = val;
 679                 retv = 0;
 680                 break;
 681         case IPV6_MTU:
 682                 if (optlen < sizeof(int))
 683                         goto e_inval;
 684                 if (val && val < IPV6_MIN_MTU)
 685                         goto e_inval;
 686                 np->frag_size = val;
 687                 retv = 0;
 688                 break;
 689         case IPV6_RECVERR:
 690                 if (optlen < sizeof(int))
 691                         goto e_inval;
 692                 np->recverr = valbool;
 693                 if (!val)
 694                         skb_queue_purge(&sk->sk_error_queue);
 695                 retv = 0;
 696                 break;
 697         case IPV6_FLOWINFO_SEND:
 698                 if (optlen < sizeof(int))
 699                         goto e_inval;
 700                 np->sndflow = valbool;
 701                 retv = 0;
 702                 break;
 703         case IPV6_FLOWLABEL_MGR:
 704                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
 705                 break;
 706         case IPV6_IPSEC_POLICY:
 707         case IPV6_XFRM_POLICY:
 708                 retv = -EPERM;
 709                 if (!capable(CAP_NET_ADMIN))
 710                         break;
 711                 retv = xfrm_user_policy(sk, optname, optval, optlen);
 712                 break;
 713
 714         case IPV6_ADDR_PREFERENCES:
 715             {
 716                 unsigned int pref = 0;
 717                 unsigned int prefmask = ~0;
 718
 719                 if (optlen < sizeof(int))
 720                         goto e_inval;
 721
 722                 retv = -EINVAL;
 723
 724                 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
 725                 switch (val & (IPV6_PREFER_SRC_PUBLIC|
 726                                IPV6_PREFER_SRC_TMP|
 727                                IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
 728                 case IPV6_PREFER_SRC_PUBLIC:
 729                         pref |= IPV6_PREFER_SRC_PUBLIC;
 730                         break;
 731                 case IPV6_PREFER_SRC_TMP:
 732                         pref |= IPV6_PREFER_SRC_TMP;
 733                         break;
 734                 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
 735                         break;
 736                 case 0:
 737                         goto pref_skip_pubtmp;
 738                 default:
 739                         goto e_inval;
 740                 }
 741
 742                 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
 743                               IPV6_PREFER_SRC_TMP);
 744 pref_skip_pubtmp:
 745
 746                 /* check HOME/COA conflicts */
 747                 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
 748                 case IPV6_PREFER_SRC_HOME:
 749                         break;
 750                 case IPV6_PREFER_SRC_COA:
 751                         pref |= IPV6_PREFER_SRC_COA;
 752                 case 0:
 753                         goto pref_skip_coa;
 754                 default:
 755                         goto e_inval;
 756                 }
 757
 758                 prefmask &= ~IPV6_PREFER_SRC_COA;
 759 pref_skip_coa:
 760
 761                 /* check CGA/NONCGA conflicts */
 762                 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
 763                 case IPV6_PREFER_SRC_CGA:
 764                 case IPV6_PREFER_SRC_NONCGA:
 765                 case 0:
 766                         break;
 767                 default:
 768                         goto e_inval;
 769                 }
 770
 771                 np->srcprefs = (np->srcprefs & prefmask) | pref;
 772                 retv = 0;
 773
 774                 break;
 775             }
 776         case IPV6_MINHOPCOUNT:
 777                 if (optlen < sizeof(int))
 778                         goto e_inval;
 779                 if (val < 0 || val > 255)
 780                         goto e_inval;
 781                 np->min_hopcount = val;
 782                 break;
 783         case IPV6_DONTFRAG:
 784                 np->dontfrag = valbool;
 785                 retv = 0;
 786                 break;
 787         }
 788
 789         release_sock(sk);
 790
 791         return retv;
 792
 793 e_inval:
 794         release_sock(sk);
 795         return -EINVAL;
 796 }
 797
 798 int ipv6_setsockopt(struct sock *sk, int level, int optname,
 799                     char __user *optval, unsigned int optlen)
 800 {
 801         int err;
 802
 803         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 804                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 805
 806         if (level != SOL_IPV6)
 807                 return -ENOPROTOOPT;
 808
 809         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 810 #ifdef CONFIG_NETFILTER
 811         /* we need to exclude all possible ENOPROTOOPTs except default case */
 812         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 813                         optname != IPV6_XFRM_POLICY) {
 814                 lock_sock(sk);
 815                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
 816                                 optlen);
 817                 release_sock(sk);
 818         }
 819 #endif
 820         return err;
 821 }
 822
 823 EXPORT_SYMBOL(ipv6_setsockopt);
 824
 825 #ifdef CONFIG_COMPAT
 826 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
 827                            char __user *optval, unsigned int optlen)
 828 {
 829         int err;
 830
 831         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
 832                 if (udp_prot.compat_setsockopt != NULL)
 833                         return udp_prot.compat_setsockopt(sk, level, optname,
 834                                                           optval, optlen);
 835                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 836         }
 837
 838         if (level != SOL_IPV6)
 839                 return -ENOPROTOOPT;
 840
 841         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
 842                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
 843                         ipv6_setsockopt);
 844
 845         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 846 #ifdef CONFIG_NETFILTER
 847         /* we need to exclude all possible ENOPROTOOPTs except default case */
 848         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 849             optname != IPV6_XFRM_POLICY) {
 850                 lock_sock(sk);
 851                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
 852                                            optval, optlen);
 853                 release_sock(sk);
 854         }
 855 #endif
 856         return err;
 857 }
 858
 859 EXPORT_SYMBOL(compat_ipv6_setsockopt);
 860 #endif
 861
 862 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
 863                                   int optname, char __user *optval, int len)
 864 {
 865         struct ipv6_opt_hdr *hdr;
 866
 867         if (!opt)
 868                 return 0;
 869
 870         switch(optname) {
 871         case IPV6_HOPOPTS:
 872                 hdr = opt->hopopt;
 873                 break;
 874         case IPV6_RTHDRDSTOPTS:
 875                 hdr = opt->dst0opt;
 876                 break;
 877         case IPV6_RTHDR:
 878                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
 879                 break;
 880         case IPV6_DSTOPTS:
 881                 hdr = opt->dst1opt;
 882                 break;
 883         default:
 884                 return -EINVAL; /* should not happen */
 885         }
 886
 887         if (!hdr)
 888                 return 0;
 889
 890         len = min_t(unsigned int, len, ipv6_optlen(hdr));
 891         if (copy_to_user(optval, hdr, len))
 892                 return -EFAULT;
 893         return len;
 894 }
 895
 896 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 897                     char __user *optval, int __user *optlen)
 898 {
 899         struct ipv6_pinfo *np = inet6_sk(sk);
 900         int len;
 901         int val;
 902
 903         if (ip6_mroute_opt(optname))
 904                 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
 905
 906         if (get_user(len, optlen))
 907                 return -EFAULT;
 908         switch (optname) {
 909         case IPV6_ADDRFORM:
 910                 if (sk->sk_protocol != IPPROTO_UDP &&
 911                     sk->sk_protocol != IPPROTO_UDPLITE &&
 912                     sk->sk_protocol != IPPROTO_TCP)
 913                         return -ENOPROTOOPT;
 914                 if (sk->sk_state != TCP_ESTABLISHED)
 915                         return -ENOTCONN;
 916                 val = sk->sk_family;
 917                 break;
 918         case MCAST_MSFILTER:
 919         {
 920                 struct group_filter gsf;
 921                 int err;
 922
 923                 if (len < GROUP_FILTER_SIZE(0))
 924                         return -EINVAL;
 925                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
 926                         return -EFAULT;
 927                 if (gsf.gf_group.ss_family != AF_INET6)
 928                         return -EADDRNOTAVAIL;
 929                 lock_sock(sk);
 930                 err = ip6_mc_msfget(sk, &gsf,
 931                         (struct group_filter __user *)optval, optlen);
 932                 release_sock(sk);
 933                 return err;
 934         }
 935
 936         case IPV6_2292PKTOPTIONS:
 937         {
 938                 struct msghdr msg;
 939                 struct sk_buff *skb;
 940
 941                 if (sk->sk_type != SOCK_STREAM)
 942                         return -ENOPROTOOPT;
 943
 944                 msg.msg_control = optval;
 945                 msg.msg_controllen = len;
 946                 msg.msg_flags = 0;
 947
 948                 lock_sock(sk);
 949                 skb = np->pktoptions;
 950                 if (skb)
 951                         atomic_inc(&skb->users);
 952                 release_sock(sk);
 953
 954                 if (skb) {
 955                         int err = datagram_recv_ctl(sk, &msg, skb);
 956                         kfree_skb(skb);
 957                         if (err)
 958                                 return err;
 959                 } else {
 960                         if (np->rxopt.bits.rxinfo) {
 961                                 struct in6_pktinfo src_info;
 962                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
 963                                         np->sticky_pktinfo.ipi6_ifindex;
 964                                 np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
 965                                         ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
 966                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 967                         }
 968                         if (np->rxopt.bits.rxhlim) {
 969                                 int hlim = np->mcast_hops;
 970                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 971                         }
 972                         if (np->rxopt.bits.rxoinfo) {
 973                                 struct in6_pktinfo src_info;
 974                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
 975                                         np->sticky_pktinfo.ipi6_ifindex;
 976                                 np->mcast_oif? ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr) :
 977                                         ipv6_addr_copy(&src_info.ipi6_addr, &(np->sticky_pktinfo.ipi6_addr));
 978                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 979                         }
 980                         if (np->rxopt.bits.rxohlim) {
 981                                 int hlim = np->mcast_hops;
 982                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 983                         }
 984                 }
 985                 len -= msg.msg_controllen;
 986                 return put_user(len, optlen);
 987         }
 988         case IPV6_MTU:
 989         {
 990                 struct dst_entry *dst;
 991
 992                 val = 0;
 993                 rcu_read_lock();
 994                 dst = __sk_dst_get(sk);
 995                 if (dst)
 996                         val = dst_mtu(dst);
 997                 rcu_read_unlock();
 998                 if (!val)
 999                         return -ENOTCONN;
1000                 break;
1001         }
1002
1003         case IPV6_V6ONLY:
1004                 val = np->ipv6only;
1005                 break;
1006
1007         case IPV6_RECVPKTINFO:
1008                 val = np->rxopt.bits.rxinfo;
1009                 break;
1010
1011         case IPV6_2292PKTINFO:
1012                 val = np->rxopt.bits.rxoinfo;
1013                 break;
1014
1015         case IPV6_RECVHOPLIMIT:
1016                 val = np->rxopt.bits.rxhlim;
1017                 break;
1018
1019         case IPV6_2292HOPLIMIT:
1020                 val = np->rxopt.bits.rxohlim;
1021                 break;
1022
1023         case IPV6_RECVRTHDR:
1024                 val = np->rxopt.bits.srcrt;
1025                 break;
1026
1027         case IPV6_2292RTHDR:
1028                 val = np->rxopt.bits.osrcrt;
1029                 break;
1030
1031         case IPV6_HOPOPTS:
1032         case IPV6_RTHDRDSTOPTS:
1033         case IPV6_RTHDR:
1034         case IPV6_DSTOPTS:
1035         {
1036
1037                 lock_sock(sk);
1038                 len = ipv6_getsockopt_sticky(sk, np->opt,
1039                                              optname, optval, len);
1040                 release_sock(sk);
1041                 /* check if ipv6_getsockopt_sticky() returns err code */
1042                 if (len < 0)
1043                         return len;
1044                 return put_user(len, optlen);
1045         }
1046
1047         case IPV6_RECVHOPOPTS:
1048                 val = np->rxopt.bits.hopopts;
1049                 break;
1050
1051         case IPV6_2292HOPOPTS:
1052                 val = np->rxopt.bits.ohopopts;
1053                 break;
1054
1055         case IPV6_RECVDSTOPTS:
1056                 val = np->rxopt.bits.dstopts;
1057                 break;
1058
1059         case IPV6_2292DSTOPTS:
1060                 val = np->rxopt.bits.odstopts;
1061                 break;
1062
1063         case IPV6_TCLASS:
1064                 val = np->tclass;
1065                 break;
1066
1067         case IPV6_RECVTCLASS:
1068                 val = np->rxopt.bits.rxtclass;
1069                 break;
1070
1071         case IPV6_FLOWINFO:
1072                 val = np->rxopt.bits.rxflow;
1073                 break;
1074
1075         case IPV6_RECVPATHMTU:
1076                 val = np->rxopt.bits.rxpmtu;
1077                 break;
1078
1079         case IPV6_PATHMTU:
1080         {
1081                 struct dst_entry *dst;
1082                 struct ip6_mtuinfo mtuinfo;
1083
1084                 if (len < sizeof(mtuinfo))
1085                         return -EINVAL;
1086
1087                 len = sizeof(mtuinfo);
1088                 memset(&mtuinfo, 0, sizeof(mtuinfo));
1089
1090                 rcu_read_lock();
1091                 dst = __sk_dst_get(sk);
1092                 if (dst)
1093                         mtuinfo.ip6m_mtu = dst_mtu(dst);
1094                 rcu_read_unlock();
1095                 if (!mtuinfo.ip6m_mtu)
1096                         return -ENOTCONN;
1097
1098                 if (put_user(len, optlen))
1099                         return -EFAULT;
1100                 if (copy_to_user(optval, &mtuinfo, len))
1101                         return -EFAULT;
1102
1103                 return 0;
1104                 break;
1105         }
1106
1107         case IPV6_UNICAST_HOPS:
1108         case IPV6_MULTICAST_HOPS:
1109         {
1110                 struct dst_entry *dst;
1111
1112                 if (optname == IPV6_UNICAST_HOPS)
1113                         val = np->hop_limit;
1114                 else
1115                         val = np->mcast_hops;
1116
1117                 if (val < 0) {
1118                         rcu_read_lock();
1119                         dst = __sk_dst_get(sk);
1120                         if (dst)
1121                                 val = ip6_dst_hoplimit(dst);
1122                         rcu_read_unlock();
1123                 }
1124
1125                 if (val < 0)
1126                         val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1127                 break;
1128         }
1129
1130         case IPV6_MULTICAST_LOOP:
1131                 val = np->mc_loop;
1132                 break;
1133
1134         case IPV6_MULTICAST_IF:
1135                 val = np->mcast_oif;
1136                 break;
1137
1138         case IPV6_MTU_DISCOVER:
1139                 val = np->pmtudisc;
1140                 break;
1141
1142         case IPV6_RECVERR:
1143                 val = np->recverr;
1144                 break;
1145
1146         case IPV6_FLOWINFO_SEND:
1147                 val = np->sndflow;
1148                 break;
1149
1150         case IPV6_ADDR_PREFERENCES:
1151                 val = 0;
1152
1153                 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1154                         val |= IPV6_PREFER_SRC_TMP;
1155                 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1156                         val |= IPV6_PREFER_SRC_PUBLIC;
1157                 else {
1158                         /* XXX: should we return system default? */
1159                         val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1160                 }
1161
1162                 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1163                         val |= IPV6_PREFER_SRC_COA;
1164                 else
1165                         val |= IPV6_PREFER_SRC_HOME;
1166                 break;
1167
1168         case IPV6_MINHOPCOUNT:
1169                 val = np->min_hopcount;
1170                 break;
1171
1172         case IPV6_DONTFRAG:
1173                 val = np->dontfrag;
1174                 break;
1175
1176         default:
1177                 return -ENOPROTOOPT;
1178         }
1179         len = min_t(unsigned int, sizeof(int), len);
1180         if(put_user(len, optlen))
1181                 return -EFAULT;
1182         if(copy_to_user(optval,&val,len))
1183                 return -EFAULT;
1184         return 0;
1185 }
1186
1187 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1188                     char __user *optval, int __user *optlen)
1189 {
1190         int err;
1191
1192         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1193                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1194
1195         if(level != SOL_IPV6)
1196                 return -ENOPROTOOPT;
1197
1198         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1199 #ifdef CONFIG_NETFILTER
1200         /* we need to exclude all possible ENOPROTOOPTs except default case */
1201         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1202                 int len;
1203
1204                 if (get_user(len, optlen))
1205                         return -EFAULT;
1206
1207                 lock_sock(sk);
1208                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1209                                 &len);
1210                 release_sock(sk);
1211                 if (err >= 0)
1212                         err = put_user(len, optlen);
1213         }
1214 #endif
1215         return err;
1216 }
1217
1218 EXPORT_SYMBOL(ipv6_getsockopt);
1219
1220 #ifdef CONFIG_COMPAT
1221 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1222                            char __user *optval, int __user *optlen)
1223 {
1224         int err;
1225
1226         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1227                 if (udp_prot.compat_getsockopt != NULL)
1228                         return udp_prot.compat_getsockopt(sk, level, optname,
1229                                                           optval, optlen);
1230                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1231         }
1232
1233         if (level != SOL_IPV6)
1234                 return -ENOPROTOOPT;
1235
1236         if (optname == MCAST_MSFILTER)
1237                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1238                         ipv6_getsockopt);
1239
1240         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1241 #ifdef CONFIG_NETFILTER
1242         /* we need to exclude all possible ENOPROTOOPTs except default case */
1243         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1244                 int len;
1245
1246                 if (get_user(len, optlen))
1247                         return -EFAULT;
1248
1249                 lock_sock(sk);
1250                 err = compat_nf_getsockopt(sk, PF_INET6,
1251                                            optname, optval, &len);
1252                 release_sock(sk);
1253                 if (err >= 0)
1254                         err = put_user(len, optlen);
1255         }
1256 #endif
1257         return err;
1258 }
1259
1260 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1261 #endif
1262