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