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  *      $Id: ipv6_sockglue.c,v 1.41 2002/02/01 22:01:04 davem Exp $
  11  *
  12  *      This program is free software; you can redistribute it and/or
  13  *      modify it under the terms of the GNU General Public License
  14  *      as published by the Free Software Foundation; either version
  15  *      2 of the License, or (at your option) any later version.
  16  *
  17  *      FIXME: Make the setsockopt code POSIX compliant: That is
  18  *
  19  *      o       Truncate getsockopt returns
  20  *      o       Return an optlen of the truncated length if need be
  21  *
  22  *      Changes:
  23  *      David L Stevens <dlstevens@us.ibm.com>:
  24  *              - added multicast source filtering API for MLDv2
  25  */
  26
  27 #include <linux/module.h>
  28 #include <linux/capability.h>
  29 #include <linux/errno.h>
  30 #include <linux/types.h>
  31 #include <linux/socket.h>
  32 #include <linux/sockios.h>
  33 #include <linux/net.h>
  34 #include <linux/in6.h>
  35 #include <linux/mroute6.h>
  36 #include <linux/netdevice.h>
  37 #include <linux/if_arp.h>
  38 #include <linux/init.h>
  39 #include <linux/sysctl.h>
  40 #include <linux/netfilter.h>
  41
  42 #include <net/sock.h>
  43 #include <net/snmp.h>
  44 #include <net/ipv6.h>
  45 #include <net/ndisc.h>
  46 #include <net/protocol.h>
  47 #include <net/transp_v6.h>
  48 #include <net/ip6_route.h>
  49 #include <net/addrconf.h>
  50 #include <net/inet_common.h>
  51 #include <net/tcp.h>
  52 #include <net/udp.h>
  53 #include <net/udplite.h>
  54 #include <net/xfrm.h>
  55
  56 #include <asm/uaccess.h>
  57
  58 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
  59
  60 struct ip6_ra_chain *ip6_ra_chain;
  61 DEFINE_RWLOCK(ip6_ra_lock);
  62
  63 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
  64 {
  65         struct ip6_ra_chain *ra, *new_ra, **rap;
  66
  67         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
  68         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
  69                 return -EINVAL;
  70
  71         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
  72
  73         write_lock_bh(&ip6_ra_lock);
  74         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
  75                 if (ra->sk == sk) {
  76                         if (sel>=0) {
  77                                 write_unlock_bh(&ip6_ra_lock);
  78                                 kfree(new_ra);
  79                                 return -EADDRINUSE;
  80                         }
  81
  82                         *rap = ra->next;
  83                         write_unlock_bh(&ip6_ra_lock);
  84
  85                         if (ra->destructor)
  86                                 ra->destructor(sk);
  87                         sock_put(sk);
  88                         kfree(ra);
  89                         return 0;
  90                 }
  91         }
  92         if (new_ra == NULL) {
  93                 write_unlock_bh(&ip6_ra_lock);
  94                 return -ENOBUFS;
  95         }
  96         new_ra->sk = sk;
  97         new_ra->sel = sel;
  98         new_ra->destructor = destructor;
  99         new_ra->next = ra;
 100         *rap = new_ra;
 101         sock_hold(sk);
 102         write_unlock_bh(&ip6_ra_lock);
 103         return 0;
 104 }
 105
 106 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 107                     char __user *optval, int optlen)
 108 {
 109         struct ipv6_pinfo *np = inet6_sk(sk);
 110         struct net *net = sock_net(sk);
 111         int val, valbool;
 112         int retv = -ENOPROTOOPT;
 113
 114         if (optval == NULL)
 115                 val=0;
 116         else {
 117                 if (optlen >= sizeof(int)) {
 118                         if (get_user(val, (int __user *) optval))
 119                                 return -EFAULT;
 120                 } else
 121                         val = 0;
 122         }
 123
 124         valbool = (val!=0);
 125
 126         if (ip6_mroute_opt(optname))
 127                 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
 128
 129         lock_sock(sk);
 130
 131         switch (optname) {
 132
 133         case IPV6_ADDRFORM:
 134                 if (optlen < sizeof(int))
 135                         goto e_inval;
 136                 if (val == PF_INET) {
 137                         struct ipv6_txoptions *opt;
 138                         struct sk_buff *pktopt;
 139
 140                         if (sk->sk_protocol != IPPROTO_UDP &&
 141                             sk->sk_protocol != IPPROTO_UDPLITE &&
 142                             sk->sk_protocol != IPPROTO_TCP)
 143                                 break;
 144
 145                         if (sk->sk_state != TCP_ESTABLISHED) {
 146                                 retv = -ENOTCONN;
 147                                 break;
 148                         }
 149
 150                         if (ipv6_only_sock(sk) ||
 151                             !ipv6_addr_v4mapped(&np->daddr)) {
 152                                 retv = -EADDRNOTAVAIL;
 153                                 break;
 154                         }
 155
 156                         fl6_free_socklist(sk);
 157                         ipv6_sock_mc_close(sk);
 158
 159                         /*
 160                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
 161                          * remove it from the refcnt debug socks count in the
 162                          * original family...
 163                          */
 164                         sk_refcnt_debug_dec(sk);
 165
 166                         if (sk->sk_protocol == IPPROTO_TCP) {
 167                                 struct inet_connection_sock *icsk = inet_csk(sk);
 168                                 local_bh_disable();
 169                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 170                                 sock_prot_inuse_add(net, &tcp_prot, 1);
 171                                 local_bh_enable();
 172                                 sk->sk_prot = &tcp_prot;
 173                                 icsk->icsk_af_ops = &ipv4_specific;
 174                                 sk->sk_socket->ops = &inet_stream_ops;
 175                                 sk->sk_family = PF_INET;
 176                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 177                         } else {
 178                                 struct proto *prot = &udp_prot;
 179
 180                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
 181                                         prot = &udplite_prot;
 182                                 local_bh_disable();
 183                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 184                                 sock_prot_inuse_add(net, prot, 1);
 185                                 local_bh_enable();
 186                                 sk->sk_prot = prot;
 187                                 sk->sk_socket->ops = &inet_dgram_ops;
 188                                 sk->sk_family = PF_INET;
 189                         }
 190                         opt = xchg(&np->opt, NULL);
 191                         if (opt)
 192                                 sock_kfree_s(sk, opt, opt->tot_len);
 193                         pktopt = xchg(&np->pktoptions, NULL);
 194                         if (pktopt)
 195                                 kfree_skb(pktopt);
 196
 197                         sk->sk_destruct = inet_sock_destruct;
 198                         /*
 199                          * ... and add it to the refcnt debug socks count
 200                          * in the new family. -acme
 201                          */
 202                         sk_refcnt_debug_inc(sk);
 203                         module_put(THIS_MODULE);
 204                         retv = 0;
 205                         break;
 206                 }
 207                 goto e_inval;
 208
 209         case IPV6_V6ONLY:
 210                 if (optlen < sizeof(int) ||
 211                     inet_sk(sk)->num)
 212                         goto e_inval;
 213                 np->ipv6only = valbool;
 214                 retv = 0;
 215                 break;
 216
 217         case IPV6_RECVPKTINFO:
 218                 if (optlen < sizeof(int))
 219                         goto e_inval;
 220                 np->rxopt.bits.rxinfo = valbool;
 221                 retv = 0;
 222                 break;
 223
 224         case IPV6_2292PKTINFO:
 225                 if (optlen < sizeof(int))
 226                         goto e_inval;
 227                 np->rxopt.bits.rxoinfo = valbool;
 228                 retv = 0;
 229                 break;
 230
 231         case IPV6_RECVHOPLIMIT:
 232                 if (optlen < sizeof(int))
 233                         goto e_inval;
 234                 np->rxopt.bits.rxhlim = valbool;
 235                 retv = 0;
 236                 break;
 237
 238         case IPV6_2292HOPLIMIT:
 239                 if (optlen < sizeof(int))
 240                         goto e_inval;
 241                 np->rxopt.bits.rxohlim = valbool;
 242                 retv = 0;
 243                 break;
 244
 245         case IPV6_RECVRTHDR:
 246                 if (optlen < sizeof(int))
 247                         goto e_inval;
 248                 np->rxopt.bits.srcrt = valbool;
 249                 retv = 0;
 250                 break;
 251
 252         case IPV6_2292RTHDR:
 253                 if (optlen < sizeof(int))
 254                         goto e_inval;
 255                 np->rxopt.bits.osrcrt = valbool;
 256                 retv = 0;
 257                 break;
 258
 259         case IPV6_RECVHOPOPTS:
 260                 if (optlen < sizeof(int))
 261                         goto e_inval;
 262                 np->rxopt.bits.hopopts = valbool;
 263                 retv = 0;
 264                 break;
 265
 266         case IPV6_2292HOPOPTS:
 267                 if (optlen < sizeof(int))
 268                         goto e_inval;
 269                 np->rxopt.bits.ohopopts = valbool;
 270                 retv = 0;
 271                 break;
 272
 273         case IPV6_RECVDSTOPTS:
 274                 if (optlen < sizeof(int))
 275                         goto e_inval;
 276                 np->rxopt.bits.dstopts = valbool;
 277                 retv = 0;
 278                 break;
 279
 280         case IPV6_2292DSTOPTS:
 281                 if (optlen < sizeof(int))
 282                         goto e_inval;
 283                 np->rxopt.bits.odstopts = valbool;
 284                 retv = 0;
 285                 break;
 286
 287         case IPV6_TCLASS:
 288                 if (optlen < sizeof(int))
 289                         goto e_inval;
 290                 if (val < -1 || val > 0xff)
 291                         goto e_inval;
 292                 np->tclass = val;
 293                 retv = 0;
 294                 break;
 295
 296         case IPV6_RECVTCLASS:
 297                 if (optlen < sizeof(int))
 298                         goto e_inval;
 299                 np->rxopt.bits.rxtclass = valbool;
 300                 retv = 0;
 301                 break;
 302
 303         case IPV6_FLOWINFO:
 304                 if (optlen < sizeof(int))
 305                         goto e_inval;
 306                 np->rxopt.bits.rxflow = valbool;
 307                 retv = 0;
 308                 break;
 309
 310         case IPV6_HOPOPTS:
 311         case IPV6_RTHDRDSTOPTS:
 312         case IPV6_RTHDR:
 313         case IPV6_DSTOPTS:
 314         {
 315                 struct ipv6_txoptions *opt;
 316                 if (optlen == 0)
 317                         optval = NULL;
 318
 319                 /* hop-by-hop / destination options are privileged option */
 320                 retv = -EPERM;
 321                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
 322                         break;
 323
 324                 if (optlen < sizeof(struct ipv6_opt_hdr) ||
 325                     optlen & 0x7 || optlen > 8 * 255)
 326                         goto e_inval;
 327
 328                 opt = ipv6_renew_options(sk, np->opt, optname,
 329                                          (struct ipv6_opt_hdr __user *)optval,
 330                                          optlen);
 331                 if (IS_ERR(opt)) {
 332                         retv = PTR_ERR(opt);
 333                         break;
 334                 }
 335
 336                 /* routing header option needs extra check */
 337                 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
 338                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
 339                         switch (rthdr->type) {
 340 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
 341                         case IPV6_SRCRT_TYPE_2:
 342                                 break;
 343 #endif
 344                         default:
 345                                 goto sticky_done;
 346                         }
 347
 348                         if ((rthdr->hdrlen & 1) ||
 349                             (rthdr->hdrlen >> 1) != rthdr->segments_left)
 350                                 goto sticky_done;
 351                 }
 352
 353                 retv = 0;
 354                 if (inet_sk(sk)->is_icsk) {
 355                         if (opt) {
 356                                 struct inet_connection_sock *icsk = inet_csk(sk);
 357                                 if (!((1 << sk->sk_state) &
 358                                       (TCPF_LISTEN | TCPF_CLOSE))
 359                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
 360                                         icsk->icsk_ext_hdr_len =
 361                                                 opt->opt_flen + opt->opt_nflen;
 362                                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 363                                 }
 364                         }
 365                         opt = xchg(&np->opt, opt);
 366                         sk_dst_reset(sk);
 367                 } else {
 368                         write_lock(&sk->sk_dst_lock);
 369                         opt = xchg(&np->opt, opt);
 370                         write_unlock(&sk->sk_dst_lock);
 371                         sk_dst_reset(sk);
 372                 }
 373 sticky_done:
 374                 if (opt)
 375                         sock_kfree_s(sk, opt, opt->tot_len);
 376                 break;
 377         }
 378
 379         case IPV6_2292PKTOPTIONS:
 380         {
 381                 struct ipv6_txoptions *opt = NULL;
 382                 struct msghdr msg;
 383                 struct flowi fl;
 384                 int junk;
 385
 386                 fl.fl6_flowlabel = 0;
 387                 fl.oif = sk->sk_bound_dev_if;
 388
 389                 if (optlen == 0)
 390                         goto update;
 391
 392                 /* 1K is probably excessive
 393                  * 1K is surely not enough, 2K per standard header is 16K.
 394                  */
 395                 retv = -EINVAL;
 396                 if (optlen > 64*1024)
 397                         break;
 398
 399                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 400                 retv = -ENOBUFS;
 401                 if (opt == NULL)
 402                         break;
 403
 404                 memset(opt, 0, sizeof(*opt));
 405                 opt->tot_len = sizeof(*opt) + optlen;
 406                 retv = -EFAULT;
 407                 if (copy_from_user(opt+1, optval, optlen))
 408                         goto done;
 409
 410                 msg.msg_controllen = optlen;
 411                 msg.msg_control = (void*)(opt+1);
 412
 413                 retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
 414                 if (retv)
 415                         goto done;
 416 update:
 417                 retv = 0;
 418                 if (inet_sk(sk)->is_icsk) {
 419                         if (opt) {
 420                                 struct inet_connection_sock *icsk = inet_csk(sk);
 421                                 if (!((1 << sk->sk_state) &
 422                                       (TCPF_LISTEN | TCPF_CLOSE))
 423                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
 424                                         icsk->icsk_ext_hdr_len =
 425                                                 opt->opt_flen + opt->opt_nflen;
 426                                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 427                                 }
 428                         }
 429                         opt = xchg(&np->opt, opt);
 430                         sk_dst_reset(sk);
 431                 } else {
 432                         write_lock(&sk->sk_dst_lock);
 433                         opt = xchg(&np->opt, opt);
 434                         write_unlock(&sk->sk_dst_lock);
 435                         sk_dst_reset(sk);
 436                 }
 437
 438 done:
 439                 if (opt)
 440                         sock_kfree_s(sk, opt, opt->tot_len);
 441                 break;
 442         }
 443         case IPV6_UNICAST_HOPS:
 444                 if (optlen < sizeof(int))
 445                         goto e_inval;
 446                 if (val > 255 || val < -1)
 447                         goto e_inval;
 448                 np->hop_limit = val;
 449                 retv = 0;
 450                 break;
 451
 452         case IPV6_MULTICAST_HOPS:
 453                 if (sk->sk_type == SOCK_STREAM)
 454                         goto e_inval;
 455                 if (optlen < sizeof(int))
 456                         goto e_inval;
 457                 if (val > 255 || val < -1)
 458                         goto e_inval;
 459                 np->mcast_hops = val;
 460                 retv = 0;
 461                 break;
 462
 463         case IPV6_MULTICAST_LOOP:
 464                 if (optlen < sizeof(int))
 465                         goto e_inval;
 466                 np->mc_loop = valbool;
 467                 retv = 0;
 468                 break;
 469
 470         case IPV6_MULTICAST_IF:
 471                 if (sk->sk_type == SOCK_STREAM)
 472                         goto e_inval;
 473                 if (optlen < sizeof(int))
 474                         goto e_inval;
 475
 476                 if (val) {
 477                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
 478                                 goto e_inval;
 479
 480                         if (__dev_get_by_index(net, val) == NULL) {
 481                                 retv = -ENODEV;
 482                                 break;
 483                         }
 484                 }
 485                 np->mcast_oif = val;
 486                 retv = 0;
 487                 break;
 488         case IPV6_ADD_MEMBERSHIP:
 489         case IPV6_DROP_MEMBERSHIP:
 490         {
 491                 struct ipv6_mreq mreq;
 492
 493                 if (optlen < sizeof(struct ipv6_mreq))
 494                         goto e_inval;
 495
 496                 retv = -EPROTO;
 497                 if (inet_sk(sk)->is_icsk)
 498                         break;
 499
 500                 retv = -EFAULT;
 501                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 502                         break;
 503
 504                 if (optname == IPV6_ADD_MEMBERSHIP)
 505                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 506                 else
 507                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 508                 break;
 509         }
 510         case IPV6_JOIN_ANYCAST:
 511         case IPV6_LEAVE_ANYCAST:
 512         {
 513                 struct ipv6_mreq mreq;
 514
 515                 if (optlen < sizeof(struct ipv6_mreq))
 516                         goto e_inval;
 517
 518                 retv = -EFAULT;
 519                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 520                         break;
 521
 522                 if (optname == IPV6_JOIN_ANYCAST)
 523                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 524                 else
 525                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 526                 break;
 527         }
 528         case MCAST_JOIN_GROUP:
 529         case MCAST_LEAVE_GROUP:
 530         {
 531                 struct group_req greq;
 532                 struct sockaddr_in6 *psin6;
 533
 534                 if (optlen < sizeof(struct group_req))
 535                         goto e_inval;
 536
 537                 retv = -EFAULT;
 538                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
 539                         break;
 540                 if (greq.gr_group.ss_family != AF_INET6) {
 541                         retv = -EADDRNOTAVAIL;
 542                         break;
 543                 }
 544                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 545                 if (optname == MCAST_JOIN_GROUP)
 546                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
 547                                 &psin6->sin6_addr);
 548                 else
 549                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
 550                                 &psin6->sin6_addr);
 551                 break;
 552         }
 553         case MCAST_JOIN_SOURCE_GROUP:
 554         case MCAST_LEAVE_SOURCE_GROUP:
 555         case MCAST_BLOCK_SOURCE:
 556         case MCAST_UNBLOCK_SOURCE:
 557         {
 558                 struct group_source_req greqs;
 559                 int omode, add;
 560
 561                 if (optlen < sizeof(struct group_source_req))
 562                         goto e_inval;
 563                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 564                         retv = -EFAULT;
 565                         break;
 566                 }
 567                 if (greqs.gsr_group.ss_family != AF_INET6 ||
 568                     greqs.gsr_source.ss_family != AF_INET6) {
 569                         retv = -EADDRNOTAVAIL;
 570                         break;
 571                 }
 572                 if (optname == MCAST_BLOCK_SOURCE) {
 573                         omode = MCAST_EXCLUDE;
 574                         add = 1;
 575                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 576                         omode = MCAST_EXCLUDE;
 577                         add = 0;
 578                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 579                         struct sockaddr_in6 *psin6;
 580
 581                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 582                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
 583                                 &psin6->sin6_addr);
 584                         /* prior join w/ different source is ok */
 585                         if (retv && retv != -EADDRINUSE)
 586                                 break;
 587                         omode = MCAST_INCLUDE;
 588                         add = 1;
 589                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 590                         omode = MCAST_INCLUDE;
 591                         add = 0;
 592                 }
 593                 retv = ip6_mc_source(add, omode, sk, &greqs);
 594                 break;
 595         }
 596         case MCAST_MSFILTER:
 597         {
 598                 extern int sysctl_mld_max_msf;
 599                 struct group_filter *gsf;
 600
 601                 if (optlen < GROUP_FILTER_SIZE(0))
 602                         goto e_inval;
 603                 if (optlen > sysctl_optmem_max) {
 604                         retv = -ENOBUFS;
 605                         break;
 606                 }
 607                 gsf = kmalloc(optlen,GFP_KERNEL);
 608                 if (!gsf) {
 609                         retv = -ENOBUFS;
 610                         break;
 611                 }
 612                 retv = -EFAULT;
 613                 if (copy_from_user(gsf, optval, optlen)) {
 614                         kfree(gsf);
 615                         break;
 616                 }
 617                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 618                 if (gsf->gf_numsrc >= 0x1ffffffU ||
 619                     gsf->gf_numsrc > sysctl_mld_max_msf) {
 620                         kfree(gsf);
 621                         retv = -ENOBUFS;
 622                         break;
 623                 }
 624                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 625                         kfree(gsf);
 626                         retv = -EINVAL;
 627                         break;
 628                 }
 629                 retv = ip6_mc_msfilter(sk, gsf);
 630                 kfree(gsf);
 631
 632                 break;
 633         }
 634         case IPV6_ROUTER_ALERT:
 635                 if (optlen < sizeof(int))
 636                         goto e_inval;
 637                 retv = ip6_ra_control(sk, val, NULL);
 638                 break;
 639         case IPV6_MTU_DISCOVER:
 640                 if (optlen < sizeof(int))
 641                         goto e_inval;
 642                 if (val<0 || val>3)
 643                         goto e_inval;
 644                 np->pmtudisc = val;
 645                 retv = 0;
 646                 break;
 647         case IPV6_MTU:
 648                 if (optlen < sizeof(int))
 649                         goto e_inval;
 650                 if (val && val < IPV6_MIN_MTU)
 651                         goto e_inval;
 652                 np->frag_size = val;
 653                 retv = 0;
 654                 break;
 655         case IPV6_RECVERR:
 656                 if (optlen < sizeof(int))
 657                         goto e_inval;
 658                 np->recverr = valbool;
 659                 if (!val)
 660                         skb_queue_purge(&sk->sk_error_queue);
 661                 retv = 0;
 662                 break;
 663         case IPV6_FLOWINFO_SEND:
 664                 if (optlen < sizeof(int))
 665                         goto e_inval;
 666                 np->sndflow = valbool;
 667                 retv = 0;
 668                 break;
 669         case IPV6_FLOWLABEL_MGR:
 670                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
 671                 break;
 672         case IPV6_IPSEC_POLICY:
 673         case IPV6_XFRM_POLICY:
 674                 retv = -EPERM;
 675                 if (!capable(CAP_NET_ADMIN))
 676                         break;
 677                 retv = xfrm_user_policy(sk, optname, optval, optlen);
 678                 break;
 679
 680         case IPV6_ADDR_PREFERENCES:
 681             {
 682                 unsigned int pref = 0;
 683                 unsigned int prefmask = ~0;
 684
 685                 if (optlen < sizeof(int))
 686                         goto e_inval;
 687
 688                 retv = -EINVAL;
 689
 690                 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
 691                 switch (val & (IPV6_PREFER_SRC_PUBLIC|
 692                                IPV6_PREFER_SRC_TMP|
 693                                IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
 694                 case IPV6_PREFER_SRC_PUBLIC:
 695                         pref |= IPV6_PREFER_SRC_PUBLIC;
 696                         break;
 697                 case IPV6_PREFER_SRC_TMP:
 698                         pref |= IPV6_PREFER_SRC_TMP;
 699                         break;
 700                 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
 701                         break;
 702                 case 0:
 703                         goto pref_skip_pubtmp;
 704                 default:
 705                         goto e_inval;
 706                 }
 707
 708                 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
 709                               IPV6_PREFER_SRC_TMP);
 710 pref_skip_pubtmp:
 711
 712                 /* check HOME/COA conflicts */
 713                 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
 714                 case IPV6_PREFER_SRC_HOME:
 715                         break;
 716                 case IPV6_PREFER_SRC_COA:
 717                         pref |= IPV6_PREFER_SRC_COA;
 718                 case 0:
 719                         goto pref_skip_coa;
 720                 default:
 721                         goto e_inval;
 722                 }
 723
 724                 prefmask &= ~IPV6_PREFER_SRC_COA;
 725 pref_skip_coa:
 726
 727                 /* check CGA/NONCGA conflicts */
 728                 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
 729                 case IPV6_PREFER_SRC_CGA:
 730                 case IPV6_PREFER_SRC_NONCGA:
 731                 case 0:
 732                         break;
 733                 default:
 734                         goto e_inval;
 735                 }
 736
 737                 np->srcprefs = (np->srcprefs & prefmask) | pref;
 738                 retv = 0;
 739
 740                 break;
 741             }
 742         }
 743
 744         release_sock(sk);
 745
 746         return retv;
 747
 748 e_inval:
 749         release_sock(sk);
 750         return -EINVAL;
 751 }
 752
 753 int ipv6_setsockopt(struct sock *sk, int level, int optname,
 754                     char __user *optval, int optlen)
 755 {
 756         int err;
 757
 758         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 759                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 760
 761         if (level != SOL_IPV6)
 762                 return -ENOPROTOOPT;
 763
 764         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 765 #ifdef CONFIG_NETFILTER
 766         /* we need to exclude all possible ENOPROTOOPTs except default case */
 767         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 768                         optname != IPV6_XFRM_POLICY) {
 769                 lock_sock(sk);
 770                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
 771                                 optlen);
 772                 release_sock(sk);
 773         }
 774 #endif
 775         return err;
 776 }
 777
 778 EXPORT_SYMBOL(ipv6_setsockopt);
 779
 780 #ifdef CONFIG_COMPAT
 781 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
 782                            char __user *optval, int optlen)
 783 {
 784         int err;
 785
 786         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
 787                 if (udp_prot.compat_setsockopt != NULL)
 788                         return udp_prot.compat_setsockopt(sk, level, optname,
 789                                                           optval, optlen);
 790                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 791         }
 792
 793         if (level != SOL_IPV6)
 794                 return -ENOPROTOOPT;
 795
 796         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 797 #ifdef CONFIG_NETFILTER
 798         /* we need to exclude all possible ENOPROTOOPTs except default case */
 799         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 800             optname != IPV6_XFRM_POLICY) {
 801                 lock_sock(sk);
 802                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
 803                                            optval, optlen);
 804                 release_sock(sk);
 805         }
 806 #endif
 807         return err;
 808 }
 809
 810 EXPORT_SYMBOL(compat_ipv6_setsockopt);
 811 #endif
 812
 813 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
 814                                   int optname, char __user *optval, int len)
 815 {
 816         struct ipv6_opt_hdr *hdr;
 817
 818         if (!opt)
 819                 return 0;
 820
 821         switch(optname) {
 822         case IPV6_HOPOPTS:
 823                 hdr = opt->hopopt;
 824                 break;
 825         case IPV6_RTHDRDSTOPTS:
 826                 hdr = opt->dst0opt;
 827                 break;
 828         case IPV6_RTHDR:
 829                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
 830                 break;
 831         case IPV6_DSTOPTS:
 832                 hdr = opt->dst1opt;
 833                 break;
 834         default:
 835                 return -EINVAL; /* should not happen */
 836         }
 837
 838         if (!hdr)
 839                 return 0;
 840
 841         len = min_t(unsigned int, len, ipv6_optlen(hdr));
 842         if (copy_to_user(optval, hdr, len))
 843                 return -EFAULT;
 844         return ipv6_optlen(hdr);
 845 }
 846
 847 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 848                     char __user *optval, int __user *optlen)
 849 {
 850         struct ipv6_pinfo *np = inet6_sk(sk);
 851         int len;
 852         int val;
 853
 854         if (ip6_mroute_opt(optname))
 855                 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
 856
 857         if (get_user(len, optlen))
 858                 return -EFAULT;
 859         switch (optname) {
 860         case IPV6_ADDRFORM:
 861                 if (sk->sk_protocol != IPPROTO_UDP &&
 862                     sk->sk_protocol != IPPROTO_UDPLITE &&
 863                     sk->sk_protocol != IPPROTO_TCP)
 864                         return -EINVAL;
 865                 if (sk->sk_state != TCP_ESTABLISHED)
 866                         return -ENOTCONN;
 867                 val = sk->sk_family;
 868                 break;
 869         case MCAST_MSFILTER:
 870         {
 871                 struct group_filter gsf;
 872                 int err;
 873
 874                 if (len < GROUP_FILTER_SIZE(0))
 875                         return -EINVAL;
 876                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
 877                         return -EFAULT;
 878                 lock_sock(sk);
 879                 err = ip6_mc_msfget(sk, &gsf,
 880                         (struct group_filter __user *)optval, optlen);
 881                 release_sock(sk);
 882                 return err;
 883         }
 884
 885         case IPV6_2292PKTOPTIONS:
 886         {
 887                 struct msghdr msg;
 888                 struct sk_buff *skb;
 889
 890                 if (sk->sk_type != SOCK_STREAM)
 891                         return -ENOPROTOOPT;
 892
 893                 msg.msg_control = optval;
 894                 msg.msg_controllen = len;
 895                 msg.msg_flags = 0;
 896
 897                 lock_sock(sk);
 898                 skb = np->pktoptions;
 899                 if (skb)
 900                         atomic_inc(&skb->users);
 901                 release_sock(sk);
 902
 903                 if (skb) {
 904                         int err = datagram_recv_ctl(sk, &msg, skb);
 905                         kfree_skb(skb);
 906                         if (err)
 907                                 return err;
 908                 } else {
 909                         if (np->rxopt.bits.rxinfo) {
 910                                 struct in6_pktinfo src_info;
 911                                 src_info.ipi6_ifindex = np->mcast_oif;
 912                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 913                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 914                         }
 915                         if (np->rxopt.bits.rxhlim) {
 916                                 int hlim = np->mcast_hops;
 917                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 918                         }
 919                         if (np->rxopt.bits.rxoinfo) {
 920                                 struct in6_pktinfo src_info;
 921                                 src_info.ipi6_ifindex = np->mcast_oif;
 922                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 923                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 924                         }
 925                         if (np->rxopt.bits.rxohlim) {
 926                                 int hlim = np->mcast_hops;
 927                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 928                         }
 929                 }
 930                 len -= msg.msg_controllen;
 931                 return put_user(len, optlen);
 932         }
 933         case IPV6_MTU:
 934         {
 935                 struct dst_entry *dst;
 936                 val = 0;
 937                 lock_sock(sk);
 938                 dst = sk_dst_get(sk);
 939                 if (dst) {
 940                         val = dst_mtu(dst);
 941                         dst_release(dst);
 942                 }
 943                 release_sock(sk);
 944                 if (!val)
 945                         return -ENOTCONN;
 946                 break;
 947         }
 948
 949         case IPV6_V6ONLY:
 950                 val = np->ipv6only;
 951                 break;
 952
 953         case IPV6_RECVPKTINFO:
 954                 val = np->rxopt.bits.rxinfo;
 955                 break;
 956
 957         case IPV6_2292PKTINFO:
 958                 val = np->rxopt.bits.rxoinfo;
 959                 break;
 960
 961         case IPV6_RECVHOPLIMIT:
 962                 val = np->rxopt.bits.rxhlim;
 963                 break;
 964
 965         case IPV6_2292HOPLIMIT:
 966                 val = np->rxopt.bits.rxohlim;
 967                 break;
 968
 969         case IPV6_RECVRTHDR:
 970                 val = np->rxopt.bits.srcrt;
 971                 break;
 972
 973         case IPV6_2292RTHDR:
 974                 val = np->rxopt.bits.osrcrt;
 975                 break;
 976
 977         case IPV6_HOPOPTS:
 978         case IPV6_RTHDRDSTOPTS:
 979         case IPV6_RTHDR:
 980         case IPV6_DSTOPTS:
 981         {
 982
 983                 lock_sock(sk);
 984                 len = ipv6_getsockopt_sticky(sk, np->opt,
 985                                              optname, optval, len);
 986                 release_sock(sk);
 987                 return put_user(len, optlen);
 988         }
 989
 990         case IPV6_RECVHOPOPTS:
 991                 val = np->rxopt.bits.hopopts;
 992                 break;
 993
 994         case IPV6_2292HOPOPTS:
 995                 val = np->rxopt.bits.ohopopts;
 996                 break;
 997
 998         case IPV6_RECVDSTOPTS:
 999                 val = np->rxopt.bits.dstopts;
1000                 break;
1001
1002         case IPV6_2292DSTOPTS:
1003                 val = np->rxopt.bits.odstopts;
1004                 break;
1005
1006         case IPV6_TCLASS:
1007                 val = np->tclass;
1008                 if (val < 0)
1009                         val = 0;
1010                 break;
1011
1012         case IPV6_RECVTCLASS:
1013                 val = np->rxopt.bits.rxtclass;
1014                 break;
1015
1016         case IPV6_FLOWINFO:
1017                 val = np->rxopt.bits.rxflow;
1018                 break;
1019
1020         case IPV6_UNICAST_HOPS:
1021         case IPV6_MULTICAST_HOPS:
1022         {
1023                 struct dst_entry *dst;
1024
1025                 if (optname == IPV6_UNICAST_HOPS)
1026                         val = np->hop_limit;
1027                 else
1028                         val = np->mcast_hops;
1029
1030                 dst = sk_dst_get(sk);
1031                 if (dst) {
1032                         if (val < 0)
1033                                 val = ip6_dst_hoplimit(dst);
1034                         dst_release(dst);
1035                 }
1036                 if (val < 0)
1037                         val = ipv6_devconf.hop_limit;
1038                 break;
1039         }
1040
1041         case IPV6_MULTICAST_LOOP:
1042                 val = np->mc_loop;
1043                 break;
1044
1045         case IPV6_MULTICAST_IF:
1046                 val = np->mcast_oif;
1047                 break;
1048
1049         case IPV6_MTU_DISCOVER:
1050                 val = np->pmtudisc;
1051                 break;
1052
1053         case IPV6_RECVERR:
1054                 val = np->recverr;
1055                 break;
1056
1057         case IPV6_FLOWINFO_SEND:
1058                 val = np->sndflow;
1059                 break;
1060
1061         case IPV6_ADDR_PREFERENCES:
1062                 val = 0;
1063
1064                 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1065                         val |= IPV6_PREFER_SRC_TMP;
1066                 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1067                         val |= IPV6_PREFER_SRC_PUBLIC;
1068                 else {
1069                         /* XXX: should we return system default? */
1070                         val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1071                 }
1072
1073                 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1074                         val |= IPV6_PREFER_SRC_COA;
1075                 else
1076                         val |= IPV6_PREFER_SRC_HOME;
1077                 break;
1078
1079         default:
1080                 return -ENOPROTOOPT;
1081         }
1082         len = min_t(unsigned int, sizeof(int), len);
1083         if(put_user(len, optlen))
1084                 return -EFAULT;
1085         if(copy_to_user(optval,&val,len))
1086                 return -EFAULT;
1087         return 0;
1088 }
1089
1090 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1091                     char __user *optval, int __user *optlen)
1092 {
1093         int err;
1094
1095         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1096                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1097
1098         if(level != SOL_IPV6)
1099                 return -ENOPROTOOPT;
1100
1101         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1102 #ifdef CONFIG_NETFILTER
1103         /* we need to exclude all possible ENOPROTOOPTs except default case */
1104         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1105                 int len;
1106
1107                 if (get_user(len, optlen))
1108                         return -EFAULT;
1109
1110                 lock_sock(sk);
1111                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1112                                 &len);
1113                 release_sock(sk);
1114                 if (err >= 0)
1115                         err = put_user(len, optlen);
1116         }
1117 #endif
1118         return err;
1119 }
1120
1121 EXPORT_SYMBOL(ipv6_getsockopt);
1122
1123 #ifdef CONFIG_COMPAT
1124 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1125                            char __user *optval, int __user *optlen)
1126 {
1127         int err;
1128
1129         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1130                 if (udp_prot.compat_getsockopt != NULL)
1131                         return udp_prot.compat_getsockopt(sk, level, optname,
1132                                                           optval, optlen);
1133                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1134         }
1135
1136         if (level != SOL_IPV6)
1137                 return -ENOPROTOOPT;
1138
1139         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1140 #ifdef CONFIG_NETFILTER
1141         /* we need to exclude all possible ENOPROTOOPTs except default case */
1142         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1143                 int len;
1144
1145                 if (get_user(len, optlen))
1146                         return -EFAULT;
1147
1148                 lock_sock(sk);
1149                 err = compat_nf_getsockopt(sk, PF_INET6,
1150                                            optname, optval, &len);
1151                 release_sock(sk);
1152                 if (err >= 0)
1153                         err = put_user(len, optlen);
1154         }
1155 #endif
1156         return err;
1157 }
1158
1159 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1160 #endif
1161