git.oblomov.eu Git - linux-2.6/blob - 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)
  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                         sock_put(sk);
  85                         kfree(ra);
  86                         return 0;
  87                 }
  88         }
  89         if (new_ra == NULL) {
  90                 write_unlock_bh(&ip6_ra_lock);
  91                 return -ENOBUFS;
  92         }
  93         new_ra->sk = sk;
  94         new_ra->sel = sel;
  95         new_ra->next = ra;
  96         *rap = new_ra;
  97         sock_hold(sk);
  98         write_unlock_bh(&ip6_ra_lock);
  99         return 0;
 100 }
 101
 102 static
 103 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
 104                                            struct ipv6_txoptions *opt)
 105 {
 106         if (inet_sk(sk)->is_icsk) {
 107                 if (opt &&
 108                     !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
 109                     inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
 110                         struct inet_connection_sock *icsk = inet_csk(sk);
 111                         icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
 112                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 113                 }
 114                 opt = xchg(&inet6_sk(sk)->opt, opt);
 115         } else {
 116                 write_lock(&sk->sk_dst_lock);
 117                 opt = xchg(&inet6_sk(sk)->opt, opt);
 118                 write_unlock(&sk->sk_dst_lock);
 119         }
 120         sk_dst_reset(sk);
 121
 122         return opt;
 123 }
 124
 125 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 126                     char __user *optval, int optlen)
 127 {
 128         struct ipv6_pinfo *np = inet6_sk(sk);
 129         struct net *net = sock_net(sk);
 130         int val, valbool;
 131         int retv = -ENOPROTOOPT;
 132
 133         if (optval == NULL)
 134                 val=0;
 135         else {
 136                 if (optlen >= sizeof(int)) {
 137                         if (get_user(val, (int __user *) optval))
 138                                 return -EFAULT;
 139                 } else
 140                         val = 0;
 141         }
 142
 143         valbool = (val!=0);
 144
 145         if (ip6_mroute_opt(optname))
 146                 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
 147
 148         lock_sock(sk);
 149
 150         switch (optname) {
 151
 152         case IPV6_ADDRFORM:
 153                 if (optlen < sizeof(int))
 154                         goto e_inval;
 155                 if (val == PF_INET) {
 156                         struct ipv6_txoptions *opt;
 157                         struct sk_buff *pktopt;
 158
 159                         if (sk->sk_type == SOCK_RAW)
 160                                 break;
 161
 162                         if (sk->sk_protocol == IPPROTO_UDP ||
 163                             sk->sk_protocol == IPPROTO_UDPLITE) {
 164                                 struct udp_sock *up = udp_sk(sk);
 165                                 if (up->pending == AF_INET6) {
 166                                         retv = -EBUSY;
 167                                         break;
 168                                 }
 169                         } else if (sk->sk_protocol != IPPROTO_TCP)
 170                                 break;
 171
 172                         if (sk->sk_state != TCP_ESTABLISHED) {
 173                                 retv = -ENOTCONN;
 174                                 break;
 175                         }
 176
 177                         if (ipv6_only_sock(sk) ||
 178                             !ipv6_addr_v4mapped(&np->daddr)) {
 179                                 retv = -EADDRNOTAVAIL;
 180                                 break;
 181                         }
 182
 183                         fl6_free_socklist(sk);
 184                         ipv6_sock_mc_close(sk);
 185
 186                         /*
 187                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
 188                          * remove it from the refcnt debug socks count in the
 189                          * original family...
 190                          */
 191                         sk_refcnt_debug_dec(sk);
 192
 193                         if (sk->sk_protocol == IPPROTO_TCP) {
 194                                 struct inet_connection_sock *icsk = inet_csk(sk);
 195                                 local_bh_disable();
 196                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 197                                 sock_prot_inuse_add(net, &tcp_prot, 1);
 198                                 local_bh_enable();
 199                                 sk->sk_prot = &tcp_prot;
 200                                 icsk->icsk_af_ops = &ipv4_specific;
 201                                 sk->sk_socket->ops = &inet_stream_ops;
 202                                 sk->sk_family = PF_INET;
 203                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 204                         } else {
 205                                 struct proto *prot = &udp_prot;
 206
 207                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
 208                                         prot = &udplite_prot;
 209                                 local_bh_disable();
 210                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 211                                 sock_prot_inuse_add(net, prot, 1);
 212                                 local_bh_enable();
 213                                 sk->sk_prot = prot;
 214                                 sk->sk_socket->ops = &inet_dgram_ops;
 215                                 sk->sk_family = PF_INET;
 216                         }
 217                         opt = xchg(&np->opt, NULL);
 218                         if (opt)
 219                                 sock_kfree_s(sk, opt, opt->tot_len);
 220                         pktopt = xchg(&np->pktoptions, NULL);
 221                         if (pktopt)
 222                                 kfree_skb(pktopt);
 223
 224                         sk->sk_destruct = inet_sock_destruct;
 225                         /*
 226                          * ... and add it to the refcnt debug socks count
 227                          * in the new family. -acme
 228                          */
 229                         sk_refcnt_debug_inc(sk);
 230                         module_put(THIS_MODULE);
 231                         retv = 0;
 232                         break;
 233                 }
 234                 goto e_inval;
 235
 236         case IPV6_V6ONLY:
 237                 if (optlen < sizeof(int) ||
 238                     inet_sk(sk)->num)
 239                         goto e_inval;
 240                 np->ipv6only = valbool;
 241                 retv = 0;
 242                 break;
 243
 244         case IPV6_RECVPKTINFO:
 245                 if (optlen < sizeof(int))
 246                         goto e_inval;
 247                 np->rxopt.bits.rxinfo = valbool;
 248                 retv = 0;
 249                 break;
 250
 251         case IPV6_2292PKTINFO:
 252                 if (optlen < sizeof(int))
 253                         goto e_inval;
 254                 np->rxopt.bits.rxoinfo = valbool;
 255                 retv = 0;
 256                 break;
 257
 258         case IPV6_RECVHOPLIMIT:
 259                 if (optlen < sizeof(int))
 260                         goto e_inval;
 261                 np->rxopt.bits.rxhlim = valbool;
 262                 retv = 0;
 263                 break;
 264
 265         case IPV6_2292HOPLIMIT:
 266                 if (optlen < sizeof(int))
 267                         goto e_inval;
 268                 np->rxopt.bits.rxohlim = valbool;
 269                 retv = 0;
 270                 break;
 271
 272         case IPV6_RECVRTHDR:
 273                 if (optlen < sizeof(int))
 274                         goto e_inval;
 275                 np->rxopt.bits.srcrt = valbool;
 276                 retv = 0;
 277                 break;
 278
 279         case IPV6_2292RTHDR:
 280                 if (optlen < sizeof(int))
 281                         goto e_inval;
 282                 np->rxopt.bits.osrcrt = valbool;
 283                 retv = 0;
 284                 break;
 285
 286         case IPV6_RECVHOPOPTS:
 287                 if (optlen < sizeof(int))
 288                         goto e_inval;
 289                 np->rxopt.bits.hopopts = valbool;
 290                 retv = 0;
 291                 break;
 292
 293         case IPV6_2292HOPOPTS:
 294                 if (optlen < sizeof(int))
 295                         goto e_inval;
 296                 np->rxopt.bits.ohopopts = valbool;
 297                 retv = 0;
 298                 break;
 299
 300         case IPV6_RECVDSTOPTS:
 301                 if (optlen < sizeof(int))
 302                         goto e_inval;
 303                 np->rxopt.bits.dstopts = valbool;
 304                 retv = 0;
 305                 break;
 306
 307         case IPV6_2292DSTOPTS:
 308                 if (optlen < sizeof(int))
 309                         goto e_inval;
 310                 np->rxopt.bits.odstopts = valbool;
 311                 retv = 0;
 312                 break;
 313
 314         case IPV6_TCLASS:
 315                 if (optlen < sizeof(int))
 316                         goto e_inval;
 317                 if (val < -1 || val > 0xff)
 318                         goto e_inval;
 319                 np->tclass = val;
 320                 retv = 0;
 321                 break;
 322
 323         case IPV6_RECVTCLASS:
 324                 if (optlen < sizeof(int))
 325                         goto e_inval;
 326                 np->rxopt.bits.rxtclass = valbool;
 327                 retv = 0;
 328                 break;
 329
 330         case IPV6_FLOWINFO:
 331                 if (optlen < sizeof(int))
 332                         goto e_inval;
 333                 np->rxopt.bits.rxflow = valbool;
 334                 retv = 0;
 335                 break;
 336
 337         case IPV6_HOPOPTS:
 338         case IPV6_RTHDRDSTOPTS:
 339         case IPV6_RTHDR:
 340         case IPV6_DSTOPTS:
 341         {
 342                 struct ipv6_txoptions *opt;
 343
 344                 /* remove any sticky options header with a zero option
 345                  * length, per RFC3542.
 346                  */
 347                 if (optlen == 0)
 348                         optval = NULL;
 349                 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
 350                          optlen & 0x7 || optlen > 8 * 255)
 351                         goto e_inval;
 352
 353                 /* hop-by-hop / destination options are privileged option */
 354                 retv = -EPERM;
 355                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
 356                         break;
 357
 358                 opt = ipv6_renew_options(sk, np->opt, optname,
 359                                          (struct ipv6_opt_hdr __user *)optval,
 360                                          optlen);
 361                 if (IS_ERR(opt)) {
 362                         retv = PTR_ERR(opt);
 363                         break;
 364                 }
 365
 366                 /* routing header option needs extra check */
 367                 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
 368                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
 369                         switch (rthdr->type) {
 370 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 371                         case IPV6_SRCRT_TYPE_2:
 372                                 break;
 373 #endif
 374                         default:
 375                                 goto sticky_done;
 376                         }
 377
 378                         if ((rthdr->hdrlen & 1) ||
 379                             (rthdr->hdrlen >> 1) != rthdr->segments_left)
 380                                 goto sticky_done;
 381                 }
 382
 383                 retv = 0;
 384                 opt = ipv6_update_options(sk, opt);
 385 sticky_done:
 386                 if (opt)
 387                         sock_kfree_s(sk, opt, opt->tot_len);
 388                 break;
 389         }
 390
 391         case IPV6_2292PKTOPTIONS:
 392         {
 393                 struct ipv6_txoptions *opt = NULL;
 394                 struct msghdr msg;
 395                 struct flowi fl;
 396                 int junk;
 397
 398                 fl.fl6_flowlabel = 0;
 399                 fl.oif = sk->sk_bound_dev_if;
 400
 401                 if (optlen == 0)
 402                         goto update;
 403
 404                 /* 1K is probably excessive
 405                  * 1K is surely not enough, 2K per standard header is 16K.
 406                  */
 407                 retv = -EINVAL;
 408                 if (optlen > 64*1024)
 409                         break;
 410
 411                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 412                 retv = -ENOBUFS;
 413                 if (opt == NULL)
 414                         break;
 415
 416                 memset(opt, 0, sizeof(*opt));
 417                 opt->tot_len = sizeof(*opt) + optlen;
 418                 retv = -EFAULT;
 419                 if (copy_from_user(opt+1, optval, optlen))
 420                         goto done;
 421
 422                 msg.msg_controllen = optlen;
 423                 msg.msg_control = (void*)(opt+1);
 424
 425                 retv = datagram_send_ctl(net, &msg, &fl, opt, &junk, &junk);
 426                 if (retv)
 427                         goto done;
 428 update:
 429                 retv = 0;
 430                 opt = ipv6_update_options(sk, opt);
 431 done:
 432                 if (opt)
 433                         sock_kfree_s(sk, opt, opt->tot_len);
 434                 break;
 435         }
 436         case IPV6_UNICAST_HOPS:
 437                 if (optlen < sizeof(int))
 438                         goto e_inval;
 439                 if (val > 255 || val < -1)
 440                         goto e_inval;
 441                 np->hop_limit = val;
 442                 retv = 0;
 443                 break;
 444
 445         case IPV6_MULTICAST_HOPS:
 446                 if (sk->sk_type == SOCK_STREAM)
 447                         break;
 448                 if (optlen < sizeof(int))
 449                         goto e_inval;
 450                 if (val > 255 || val < -1)
 451                         goto e_inval;
 452                 np->mcast_hops = val;
 453                 retv = 0;
 454                 break;
 455
 456         case IPV6_MULTICAST_LOOP:
 457                 if (optlen < sizeof(int))
 458                         goto e_inval;
 459                 if (val != valbool)
 460                         goto e_inval;
 461                 np->mc_loop = valbool;
 462                 retv = 0;
 463                 break;
 464
 465         case IPV6_MULTICAST_IF:
 466                 if (sk->sk_type == SOCK_STREAM)
 467                         break;
 468                 if (optlen < sizeof(int))
 469                         goto e_inval;
 470
 471                 if (val) {
 472                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
 473                                 goto e_inval;
 474
 475                         if (__dev_get_by_index(net, val) == NULL) {
 476                                 retv = -ENODEV;
 477                                 break;
 478                         }
 479                 }
 480                 np->mcast_oif = val;
 481                 retv = 0;
 482                 break;
 483         case IPV6_ADD_MEMBERSHIP:
 484         case IPV6_DROP_MEMBERSHIP:
 485         {
 486                 struct ipv6_mreq mreq;
 487
 488                 if (optlen < sizeof(struct ipv6_mreq))
 489                         goto e_inval;
 490
 491                 retv = -EPROTO;
 492                 if (inet_sk(sk)->is_icsk)
 493                         break;
 494
 495                 retv = -EFAULT;
 496                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 497                         break;
 498
 499                 if (optname == IPV6_ADD_MEMBERSHIP)
 500                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 501                 else
 502                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 503                 break;
 504         }
 505         case IPV6_JOIN_ANYCAST:
 506         case IPV6_LEAVE_ANYCAST:
 507         {
 508                 struct ipv6_mreq mreq;
 509
 510                 if (optlen < sizeof(struct ipv6_mreq))
 511                         goto e_inval;
 512
 513                 retv = -EFAULT;
 514                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 515                         break;
 516
 517                 if (optname == IPV6_JOIN_ANYCAST)
 518                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 519                 else
 520                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 521                 break;
 522         }
 523         case MCAST_JOIN_GROUP:
 524         case MCAST_LEAVE_GROUP:
 525         {
 526                 struct group_req greq;
 527                 struct sockaddr_in6 *psin6;
 528
 529                 if (optlen < sizeof(struct group_req))
 530                         goto e_inval;
 531
 532                 retv = -EFAULT;
 533                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
 534                         break;
 535                 if (greq.gr_group.ss_family != AF_INET6) {
 536                         retv = -EADDRNOTAVAIL;
 537                         break;
 538                 }
 539                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 540                 if (optname == MCAST_JOIN_GROUP)
 541                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
 542                                 &psin6->sin6_addr);
 543                 else
 544                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
 545                                 &psin6->sin6_addr);
 546                 break;
 547         }
 548         case MCAST_JOIN_SOURCE_GROUP:
 549         case MCAST_LEAVE_SOURCE_GROUP:
 550         case MCAST_BLOCK_SOURCE:
 551         case MCAST_UNBLOCK_SOURCE:
 552         {
 553                 struct group_source_req greqs;
 554                 int omode, add;
 555
 556                 if (optlen < sizeof(struct group_source_req))
 557                         goto e_inval;
 558                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 559                         retv = -EFAULT;
 560                         break;
 561                 }
 562                 if (greqs.gsr_group.ss_family != AF_INET6 ||
 563                     greqs.gsr_source.ss_family != AF_INET6) {
 564                         retv = -EADDRNOTAVAIL;
 565                         break;
 566                 }
 567                 if (optname == MCAST_BLOCK_SOURCE) {
 568                         omode = MCAST_EXCLUDE;
 569                         add = 1;
 570                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 571                         omode = MCAST_EXCLUDE;
 572                         add = 0;
 573                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 574                         struct sockaddr_in6 *psin6;
 575
 576                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 577                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
 578                                 &psin6->sin6_addr);
 579                         /* prior join w/ different source is ok */
 580                         if (retv && retv != -EADDRINUSE)
 581                                 break;
 582                         omode = MCAST_INCLUDE;
 583                         add = 1;
 584                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 585                         omode = MCAST_INCLUDE;
 586                         add = 0;
 587                 }
 588                 retv = ip6_mc_source(add, omode, sk, &greqs);
 589                 break;
 590         }
 591         case MCAST_MSFILTER:
 592         {
 593                 extern int sysctl_mld_max_msf;
 594                 struct group_filter *gsf;
 595
 596                 if (optlen < GROUP_FILTER_SIZE(0))
 597                         goto e_inval;
 598                 if (optlen > sysctl_optmem_max) {
 599                         retv = -ENOBUFS;
 600                         break;
 601                 }
 602                 gsf = kmalloc(optlen,GFP_KERNEL);
 603                 if (!gsf) {
 604                         retv = -ENOBUFS;
 605                         break;
 606                 }
 607                 retv = -EFAULT;
 608                 if (copy_from_user(gsf, optval, optlen)) {
 609                         kfree(gsf);
 610                         break;
 611                 }
 612                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 613                 if (gsf->gf_numsrc >= 0x1ffffffU ||
 614                     gsf->gf_numsrc > sysctl_mld_max_msf) {
 615                         kfree(gsf);
 616                         retv = -ENOBUFS;
 617                         break;
 618                 }
 619                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 620                         kfree(gsf);
 621                         retv = -EINVAL;
 622                         break;
 623                 }
 624                 retv = ip6_mc_msfilter(sk, gsf);
 625                 kfree(gsf);
 626
 627                 break;
 628         }
 629         case IPV6_ROUTER_ALERT:
 630                 if (optlen < sizeof(int))
 631                         goto e_inval;
 632                 retv = ip6_ra_control(sk, val);
 633                 break;
 634         case IPV6_MTU_DISCOVER:
 635                 if (optlen < sizeof(int))
 636                         goto e_inval;
 637                 if (val<0 || val>3)
 638                         goto e_inval;
 639                 np->pmtudisc = val;
 640                 retv = 0;
 641                 break;
 642         case IPV6_MTU:
 643                 if (optlen < sizeof(int))
 644                         goto e_inval;
 645                 if (val && val < IPV6_MIN_MTU)
 646                         goto e_inval;
 647                 np->frag_size = val;
 648                 retv = 0;
 649                 break;
 650         case IPV6_RECVERR:
 651                 if (optlen < sizeof(int))
 652                         goto e_inval;
 653                 np->recverr = valbool;
 654                 if (!val)
 655                         skb_queue_purge(&sk->sk_error_queue);
 656                 retv = 0;
 657                 break;
 658         case IPV6_FLOWINFO_SEND:
 659                 if (optlen < sizeof(int))
 660                         goto e_inval;
 661                 np->sndflow = valbool;
 662                 retv = 0;
 663                 break;
 664         case IPV6_FLOWLABEL_MGR:
 665                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
 666                 break;
 667         case IPV6_IPSEC_POLICY:
 668         case IPV6_XFRM_POLICY:
 669                 retv = -EPERM;
 670                 if (!capable(CAP_NET_ADMIN))
 671                         break;
 672                 retv = xfrm_user_policy(sk, optname, optval, optlen);
 673                 break;
 674
 675         case IPV6_ADDR_PREFERENCES:
 676             {
 677                 unsigned int pref = 0;
 678                 unsigned int prefmask = ~0;
 679
 680                 if (optlen < sizeof(int))
 681                         goto e_inval;
 682
 683                 retv = -EINVAL;
 684
 685                 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
 686                 switch (val & (IPV6_PREFER_SRC_PUBLIC|
 687                                IPV6_PREFER_SRC_TMP|
 688                                IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
 689                 case IPV6_PREFER_SRC_PUBLIC:
 690                         pref |= IPV6_PREFER_SRC_PUBLIC;
 691                         break;
 692                 case IPV6_PREFER_SRC_TMP:
 693                         pref |= IPV6_PREFER_SRC_TMP;
 694                         break;
 695                 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
 696                         break;
 697                 case 0:
 698                         goto pref_skip_pubtmp;
 699                 default:
 700                         goto e_inval;
 701                 }
 702
 703                 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
 704                               IPV6_PREFER_SRC_TMP);
 705 pref_skip_pubtmp:
 706
 707                 /* check HOME/COA conflicts */
 708                 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
 709                 case IPV6_PREFER_SRC_HOME:
 710                         break;
 711                 case IPV6_PREFER_SRC_COA:
 712                         pref |= IPV6_PREFER_SRC_COA;
 713                 case 0:
 714                         goto pref_skip_coa;
 715                 default:
 716                         goto e_inval;
 717                 }
 718
 719                 prefmask &= ~IPV6_PREFER_SRC_COA;
 720 pref_skip_coa:
 721
 722                 /* check CGA/NONCGA conflicts */
 723                 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
 724                 case IPV6_PREFER_SRC_CGA:
 725                 case IPV6_PREFER_SRC_NONCGA:
 726                 case 0:
 727                         break;
 728                 default:
 729                         goto e_inval;
 730                 }
 731
 732                 np->srcprefs = (np->srcprefs & prefmask) | pref;
 733                 retv = 0;
 734
 735                 break;
 736             }
 737         }
 738
 739         release_sock(sk);
 740
 741         return retv;
 742
 743 e_inval:
 744         release_sock(sk);
 745         return -EINVAL;
 746 }
 747
 748 int ipv6_setsockopt(struct sock *sk, int level, int optname,
 749                     char __user *optval, int optlen)
 750 {
 751         int err;
 752
 753         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 754                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 755
 756         if (level != SOL_IPV6)
 757                 return -ENOPROTOOPT;
 758
 759         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 760 #ifdef CONFIG_NETFILTER
 761         /* we need to exclude all possible ENOPROTOOPTs except default case */
 762         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 763                         optname != IPV6_XFRM_POLICY) {
 764                 lock_sock(sk);
 765                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
 766                                 optlen);
 767                 release_sock(sk);
 768         }
 769 #endif
 770         return err;
 771 }
 772
 773 EXPORT_SYMBOL(ipv6_setsockopt);
 774
 775 #ifdef CONFIG_COMPAT
 776 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
 777                            char __user *optval, int optlen)
 778 {
 779         int err;
 780
 781         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
 782                 if (udp_prot.compat_setsockopt != NULL)
 783                         return udp_prot.compat_setsockopt(sk, level, optname,
 784                                                           optval, optlen);
 785                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 786         }
 787
 788         if (level != SOL_IPV6)
 789                 return -ENOPROTOOPT;
 790
 791         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
 792                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
 793                         ipv6_setsockopt);
 794
 795         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 796 #ifdef CONFIG_NETFILTER
 797         /* we need to exclude all possible ENOPROTOOPTs except default case */
 798         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 799             optname != IPV6_XFRM_POLICY) {
 800                 lock_sock(sk);
 801                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
 802                                            optval, optlen);
 803                 release_sock(sk);
 804         }
 805 #endif
 806         return err;
 807 }
 808
 809 EXPORT_SYMBOL(compat_ipv6_setsockopt);
 810 #endif
 811
 812 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
 813                                   int optname, char __user *optval, int len)
 814 {
 815         struct ipv6_opt_hdr *hdr;
 816
 817         if (!opt)
 818                 return 0;
 819
 820         switch(optname) {
 821         case IPV6_HOPOPTS:
 822                 hdr = opt->hopopt;
 823                 break;
 824         case IPV6_RTHDRDSTOPTS:
 825                 hdr = opt->dst0opt;
 826                 break;
 827         case IPV6_RTHDR:
 828                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
 829                 break;
 830         case IPV6_DSTOPTS:
 831                 hdr = opt->dst1opt;
 832                 break;
 833         default:
 834                 return -EINVAL; /* should not happen */
 835         }
 836
 837         if (!hdr)
 838                 return 0;
 839
 840         len = min_t(unsigned int, len, ipv6_optlen(hdr));
 841         if (copy_to_user(optval, hdr, len))
 842                 return -EFAULT;
 843         return len;
 844 }
 845
 846 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 847                     char __user *optval, int __user *optlen)
 848 {
 849         struct ipv6_pinfo *np = inet6_sk(sk);
 850         int len;
 851         int val;
 852
 853         if (ip6_mroute_opt(optname))
 854                 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
 855
 856         if (get_user(len, optlen))
 857                 return -EFAULT;
 858         switch (optname) {
 859         case IPV6_ADDRFORM:
 860                 if (sk->sk_protocol != IPPROTO_UDP &&
 861                     sk->sk_protocol != IPPROTO_UDPLITE &&
 862                     sk->sk_protocol != IPPROTO_TCP)
 863                         return -ENOPROTOOPT;
 864                 if (sk->sk_state != TCP_ESTABLISHED)
 865                         return -ENOTCONN;
 866                 val = sk->sk_family;
 867                 break;
 868         case MCAST_MSFILTER:
 869         {
 870                 struct group_filter gsf;
 871                 int err;
 872
 873                 if (len < GROUP_FILTER_SIZE(0))
 874                         return -EINVAL;
 875                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
 876                         return -EFAULT;
 877                 if (gsf.gf_group.ss_family != AF_INET6)
 878                         return -EADDRNOTAVAIL;
 879                 lock_sock(sk);
 880                 err = ip6_mc_msfget(sk, &gsf,
 881                         (struct group_filter __user *)optval, optlen);
 882                 release_sock(sk);
 883                 return err;
 884         }
 885
 886         case IPV6_2292PKTOPTIONS:
 887         {
 888                 struct msghdr msg;
 889                 struct sk_buff *skb;
 890
 891                 if (sk->sk_type != SOCK_STREAM)
 892                         return -ENOPROTOOPT;
 893
 894                 msg.msg_control = optval;
 895                 msg.msg_controllen = len;
 896                 msg.msg_flags = 0;
 897
 898                 lock_sock(sk);
 899                 skb = np->pktoptions;
 900                 if (skb)
 901                         atomic_inc(&skb->users);
 902                 release_sock(sk);
 903
 904                 if (skb) {
 905                         int err = datagram_recv_ctl(sk, &msg, skb);
 906                         kfree_skb(skb);
 907                         if (err)
 908                                 return err;
 909                 } else {
 910                         if (np->rxopt.bits.rxinfo) {
 911                                 struct in6_pktinfo src_info;
 912                                 src_info.ipi6_ifindex = np->mcast_oif;
 913                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 914                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 915                         }
 916                         if (np->rxopt.bits.rxhlim) {
 917                                 int hlim = np->mcast_hops;
 918                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 919                         }
 920                         if (np->rxopt.bits.rxoinfo) {
 921                                 struct in6_pktinfo src_info;
 922                                 src_info.ipi6_ifindex = np->mcast_oif;
 923                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 924                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 925                         }
 926                         if (np->rxopt.bits.rxohlim) {
 927                                 int hlim = np->mcast_hops;
 928                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 929                         }
 930                 }
 931                 len -= msg.msg_controllen;
 932                 return put_user(len, optlen);
 933         }
 934         case IPV6_MTU:
 935         {
 936                 struct dst_entry *dst;
 937                 val = 0;
 938                 lock_sock(sk);
 939                 dst = sk_dst_get(sk);
 940                 if (dst) {
 941                         val = dst_mtu(dst);
 942                         dst_release(dst);
 943                 }
 944                 release_sock(sk);
 945                 if (!val)
 946                         return -ENOTCONN;
 947                 break;
 948         }
 949
 950         case IPV6_V6ONLY:
 951                 val = np->ipv6only;
 952                 break;
 953
 954         case IPV6_RECVPKTINFO:
 955                 val = np->rxopt.bits.rxinfo;
 956                 break;
 957
 958         case IPV6_2292PKTINFO:
 959                 val = np->rxopt.bits.rxoinfo;
 960                 break;
 961
 962         case IPV6_RECVHOPLIMIT:
 963                 val = np->rxopt.bits.rxhlim;
 964                 break;
 965
 966         case IPV6_2292HOPLIMIT:
 967                 val = np->rxopt.bits.rxohlim;
 968                 break;
 969
 970         case IPV6_RECVRTHDR:
 971                 val = np->rxopt.bits.srcrt;
 972                 break;
 973
 974         case IPV6_2292RTHDR:
 975                 val = np->rxopt.bits.osrcrt;
 976                 break;
 977
 978         case IPV6_HOPOPTS:
 979         case IPV6_RTHDRDSTOPTS:
 980         case IPV6_RTHDR:
 981         case IPV6_DSTOPTS:
 982         {
 983
 984                 lock_sock(sk);
 985                 len = ipv6_getsockopt_sticky(sk, np->opt,
 986                                              optname, optval, len);
 987                 release_sock(sk);
 988                 /* check if ipv6_getsockopt_sticky() returns err code */
 989                 if (len < 0)
 990                         return len;
 991                 return put_user(len, optlen);
 992         }
 993
 994         case IPV6_RECVHOPOPTS:
 995                 val = np->rxopt.bits.hopopts;
 996                 break;
 997
 998         case IPV6_2292HOPOPTS:
 999                 val = np->rxopt.bits.ohopopts;
1000                 break;
1001
1002         case IPV6_RECVDSTOPTS:
1003                 val = np->rxopt.bits.dstopts;
1004                 break;
1005
1006         case IPV6_2292DSTOPTS:
1007                 val = np->rxopt.bits.odstopts;
1008                 break;
1009
1010         case IPV6_TCLASS:
1011                 val = np->tclass;
1012                 if (val < 0)
1013                         val = 0;
1014                 break;
1015
1016         case IPV6_RECVTCLASS:
1017                 val = np->rxopt.bits.rxtclass;
1018                 break;
1019
1020         case IPV6_FLOWINFO:
1021                 val = np->rxopt.bits.rxflow;
1022                 break;
1023
1024         case IPV6_UNICAST_HOPS:
1025         case IPV6_MULTICAST_HOPS:
1026         {
1027                 struct dst_entry *dst;
1028
1029                 if (optname == IPV6_UNICAST_HOPS)
1030                         val = np->hop_limit;
1031                 else
1032                         val = np->mcast_hops;
1033
1034                 dst = sk_dst_get(sk);
1035                 if (dst) {
1036                         if (val < 0)
1037                                 val = ip6_dst_hoplimit(dst);
1038                         dst_release(dst);
1039                 }
1040                 if (val < 0)
1041                         val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1042                 break;
1043         }
1044
1045         case IPV6_MULTICAST_LOOP:
1046                 val = np->mc_loop;
1047                 break;
1048
1049         case IPV6_MULTICAST_IF:
1050                 val = np->mcast_oif;
1051                 break;
1052
1053         case IPV6_MTU_DISCOVER:
1054                 val = np->pmtudisc;
1055                 break;
1056
1057         case IPV6_RECVERR:
1058                 val = np->recverr;
1059                 break;
1060
1061         case IPV6_FLOWINFO_SEND:
1062                 val = np->sndflow;
1063                 break;
1064
1065         case IPV6_ADDR_PREFERENCES:
1066                 val = 0;
1067
1068                 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1069                         val |= IPV6_PREFER_SRC_TMP;
1070                 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1071                         val |= IPV6_PREFER_SRC_PUBLIC;
1072                 else {
1073                         /* XXX: should we return system default? */
1074                         val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1075                 }
1076
1077                 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1078                         val |= IPV6_PREFER_SRC_COA;
1079                 else
1080                         val |= IPV6_PREFER_SRC_HOME;
1081                 break;
1082
1083         default:
1084                 return -ENOPROTOOPT;
1085         }
1086         len = min_t(unsigned int, sizeof(int), len);
1087         if(put_user(len, optlen))
1088                 return -EFAULT;
1089         if(copy_to_user(optval,&val,len))
1090                 return -EFAULT;
1091         return 0;
1092 }
1093
1094 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1095                     char __user *optval, int __user *optlen)
1096 {
1097         int err;
1098
1099         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1100                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1101
1102         if(level != SOL_IPV6)
1103                 return -ENOPROTOOPT;
1104
1105         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1106 #ifdef CONFIG_NETFILTER
1107         /* we need to exclude all possible ENOPROTOOPTs except default case */
1108         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1109                 int len;
1110
1111                 if (get_user(len, optlen))
1112                         return -EFAULT;
1113
1114                 lock_sock(sk);
1115                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1116                                 &len);
1117                 release_sock(sk);
1118                 if (err >= 0)
1119                         err = put_user(len, optlen);
1120         }
1121 #endif
1122         return err;
1123 }
1124
1125 EXPORT_SYMBOL(ipv6_getsockopt);
1126
1127 #ifdef CONFIG_COMPAT
1128 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1129                            char __user *optval, int __user *optlen)
1130 {
1131         int err;
1132
1133         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1134                 if (udp_prot.compat_getsockopt != NULL)
1135                         return udp_prot.compat_getsockopt(sk, level, optname,
1136                                                           optval, optlen);
1137                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1138         }
1139
1140         if (level != SOL_IPV6)
1141                 return -ENOPROTOOPT;
1142
1143         if (optname == MCAST_MSFILTER)
1144                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1145                         ipv6_getsockopt);
1146
1147         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1148 #ifdef CONFIG_NETFILTER
1149         /* we need to exclude all possible ENOPROTOOPTs except default case */
1150         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1151                 int len;
1152
1153                 if (get_user(len, optlen))
1154                         return -EFAULT;
1155
1156                 lock_sock(sk);
1157                 err = compat_nf_getsockopt(sk, PF_INET6,
1158                                            optname, optval, &len);
1159                 release_sock(sk);
1160                 if (err >= 0)
1161                         err = put_user(len, optlen);
1162         }
1163 #endif
1164         return err;
1165 }
1166
1167 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1168 #endif
1169