SAFE public projects git trees. - safe/jmp/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  *      $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/sched.h>
  35 #include <linux/net.h>
  36 #include <linux/in6.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/xfrm.h>
  55
  56 #include <asm/uaccess.h>
  57
  58 DEFINE_SNMP_STAT(struct ipstats_mib, ipv6_statistics) __read_mostly;
  59
  60 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
  61 {
  62         struct sk_buff *segs = ERR_PTR(-EINVAL);
  63         struct ipv6hdr *ipv6h;
  64         struct inet6_protocol *ops;
  65         int proto;
  66
  67         if (unlikely(skb_shinfo(skb)->gso_type &
  68                      ~(SKB_GSO_UDP |
  69                        SKB_GSO_DODGY |
  70                        SKB_GSO_TCP_ECN |
  71                        SKB_GSO_TCPV6 |
  72                        0)))
  73                 goto out;
  74
  75         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
  76                 goto out;
  77
  78         ipv6h = skb->nh.ipv6h;
  79         proto = ipv6h->nexthdr;
  80         __skb_pull(skb, sizeof(*ipv6h));
  81
  82         rcu_read_lock();
  83         for (;;) {
  84                 struct ipv6_opt_hdr *opth;
  85                 int len;
  86
  87                 if (proto != NEXTHDR_HOP) {
  88                         ops = rcu_dereference(inet6_protos[proto]);
  89
  90                         if (unlikely(!ops))
  91                                 goto unlock;
  92
  93                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
  94                                 break;
  95                 }
  96
  97                 if (unlikely(!pskb_may_pull(skb, 8)))
  98                         goto unlock;
  99
 100                 opth = (void *)skb->data;
 101                 len = opth->hdrlen * 8 + 8;
 102
 103                 if (unlikely(!pskb_may_pull(skb, len)))
 104                         goto unlock;
 105
 106                 proto = opth->nexthdr;
 107                 __skb_pull(skb, len);
 108         }
 109
 110         skb->h.raw = skb->data;
 111         if (likely(ops->gso_segment))
 112                 segs = ops->gso_segment(skb, features);
 113
 114 unlock:
 115         rcu_read_unlock();
 116
 117         if (unlikely(IS_ERR(segs)))
 118                 goto out;
 119
 120         for (skb = segs; skb; skb = skb->next) {
 121                 ipv6h = skb->nh.ipv6h;
 122                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
 123                                            sizeof(*ipv6h));
 124         }
 125
 126 out:
 127         return segs;
 128 }
 129
 130 static struct packet_type ipv6_packet_type = {
 131         .type = __constant_htons(ETH_P_IPV6),
 132         .func = ipv6_rcv,
 133         .gso_segment = ipv6_gso_segment,
 134 };
 135
 136 struct ip6_ra_chain *ip6_ra_chain;
 137 DEFINE_RWLOCK(ip6_ra_lock);
 138
 139 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
 140 {
 141         struct ip6_ra_chain *ra, *new_ra, **rap;
 142
 143         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
 144         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
 145                 return -EINVAL;
 146
 147         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
 148
 149         write_lock_bh(&ip6_ra_lock);
 150         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
 151                 if (ra->sk == sk) {
 152                         if (sel>=0) {
 153                                 write_unlock_bh(&ip6_ra_lock);
 154                                 kfree(new_ra);
 155                                 return -EADDRINUSE;
 156                         }
 157
 158                         *rap = ra->next;
 159                         write_unlock_bh(&ip6_ra_lock);
 160
 161                         if (ra->destructor)
 162                                 ra->destructor(sk);
 163                         sock_put(sk);
 164                         kfree(ra);
 165                         return 0;
 166                 }
 167         }
 168         if (new_ra == NULL) {
 169                 write_unlock_bh(&ip6_ra_lock);
 170                 return -ENOBUFS;
 171         }
 172         new_ra->sk = sk;
 173         new_ra->sel = sel;
 174         new_ra->destructor = destructor;
 175         new_ra->next = ra;
 176         *rap = new_ra;
 177         sock_hold(sk);
 178         write_unlock_bh(&ip6_ra_lock);
 179         return 0;
 180 }
 181
 182 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 183                     char __user *optval, int optlen)
 184 {
 185         struct ipv6_pinfo *np = inet6_sk(sk);
 186         int val, valbool;
 187         int retv = -ENOPROTOOPT;
 188
 189         if (optval == NULL)
 190                 val=0;
 191         else if (get_user(val, (int __user *) optval))
 192                 return -EFAULT;
 193
 194         valbool = (val!=0);
 195
 196         lock_sock(sk);
 197
 198         switch (optname) {
 199
 200         case IPV6_ADDRFORM:
 201                 if (val == PF_INET) {
 202                         struct ipv6_txoptions *opt;
 203                         struct sk_buff *pktopt;
 204
 205                         if (sk->sk_protocol != IPPROTO_UDP &&
 206                             sk->sk_protocol != IPPROTO_TCP)
 207                                 break;
 208
 209                         if (sk->sk_state != TCP_ESTABLISHED) {
 210                                 retv = -ENOTCONN;
 211                                 break;
 212                         }
 213
 214                         if (ipv6_only_sock(sk) ||
 215                             !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
 216                                 retv = -EADDRNOTAVAIL;
 217                                 break;
 218                         }
 219
 220                         fl6_free_socklist(sk);
 221                         ipv6_sock_mc_close(sk);
 222
 223                         /*
 224                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
 225                          * remove it from the refcnt debug socks count in the
 226                          * original family...
 227                          */
 228                         sk_refcnt_debug_dec(sk);
 229
 230                         if (sk->sk_protocol == IPPROTO_TCP) {
 231                                 struct inet_connection_sock *icsk = inet_csk(sk);
 232
 233                                 local_bh_disable();
 234                                 sock_prot_dec_use(sk->sk_prot);
 235                                 sock_prot_inc_use(&tcp_prot);
 236                                 local_bh_enable();
 237                                 sk->sk_prot = &tcp_prot;
 238                                 icsk->icsk_af_ops = &ipv4_specific;
 239                                 sk->sk_socket->ops = &inet_stream_ops;
 240                                 sk->sk_family = PF_INET;
 241                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 242                         } else {
 243                                 local_bh_disable();
 244                                 sock_prot_dec_use(sk->sk_prot);
 245                                 sock_prot_inc_use(&udp_prot);
 246                                 local_bh_enable();
 247                                 sk->sk_prot = &udp_prot;
 248                                 sk->sk_socket->ops = &inet_dgram_ops;
 249                                 sk->sk_family = PF_INET;
 250                         }
 251                         opt = xchg(&np->opt, NULL);
 252                         if (opt)
 253                                 sock_kfree_s(sk, opt, opt->tot_len);
 254                         pktopt = xchg(&np->pktoptions, NULL);
 255                         if (pktopt)
 256                                 kfree_skb(pktopt);
 257
 258                         sk->sk_destruct = inet_sock_destruct;
 259                         /*
 260                          * ... and add it to the refcnt debug socks count
 261                          * in the new family. -acme
 262                          */
 263                         sk_refcnt_debug_inc(sk);
 264                         module_put(THIS_MODULE);
 265                         retv = 0;
 266                         break;
 267                 }
 268                 goto e_inval;
 269
 270         case IPV6_V6ONLY:
 271                 if (inet_sk(sk)->num)
 272                         goto e_inval;
 273                 np->ipv6only = valbool;
 274                 retv = 0;
 275                 break;
 276
 277         case IPV6_RECVPKTINFO:
 278                 np->rxopt.bits.rxinfo = valbool;
 279                 retv = 0;
 280                 break;
 281
 282         case IPV6_2292PKTINFO:
 283                 np->rxopt.bits.rxoinfo = valbool;
 284                 retv = 0;
 285                 break;
 286
 287         case IPV6_RECVHOPLIMIT:
 288                 np->rxopt.bits.rxhlim = valbool;
 289                 retv = 0;
 290                 break;
 291
 292         case IPV6_2292HOPLIMIT:
 293                 np->rxopt.bits.rxohlim = valbool;
 294                 retv = 0;
 295                 break;
 296
 297         case IPV6_RECVRTHDR:
 298                 if (val < 0 || val > 2)
 299                         goto e_inval;
 300                 np->rxopt.bits.srcrt = val;
 301                 retv = 0;
 302                 break;
 303
 304         case IPV6_2292RTHDR:
 305                 if (val < 0 || val > 2)
 306                         goto e_inval;
 307                 np->rxopt.bits.osrcrt = val;
 308                 retv = 0;
 309                 break;
 310
 311         case IPV6_RECVHOPOPTS:
 312                 np->rxopt.bits.hopopts = valbool;
 313                 retv = 0;
 314                 break;
 315
 316         case IPV6_2292HOPOPTS:
 317                 np->rxopt.bits.ohopopts = valbool;
 318                 retv = 0;
 319                 break;
 320
 321         case IPV6_RECVDSTOPTS:
 322                 np->rxopt.bits.dstopts = valbool;
 323                 retv = 0;
 324                 break;
 325
 326         case IPV6_2292DSTOPTS:
 327                 np->rxopt.bits.odstopts = valbool;
 328                 retv = 0;
 329                 break;
 330
 331         case IPV6_TCLASS:
 332                 if (val < 0 || val > 0xff)
 333                         goto e_inval;
 334                 np->tclass = val;
 335                 retv = 0;
 336                 break;
 337
 338         case IPV6_RECVTCLASS:
 339                 np->rxopt.bits.rxtclass = valbool;
 340                 retv = 0;
 341                 break;
 342
 343         case IPV6_FLOWINFO:
 344                 np->rxopt.bits.rxflow = valbool;
 345                 retv = 0;
 346                 break;
 347
 348         case IPV6_HOPOPTS:
 349         case IPV6_RTHDRDSTOPTS:
 350         case IPV6_RTHDR:
 351         case IPV6_DSTOPTS:
 352         {
 353                 struct ipv6_txoptions *opt;
 354                 if (optlen == 0)
 355                         optval = NULL;
 356
 357                 /* hop-by-hop / destination options are privileged option */
 358                 retv = -EPERM;
 359                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
 360                         break;
 361
 362                 retv = -EINVAL;
 363                 if (optlen & 0x7 || optlen > 8 * 255)
 364                         break;
 365
 366                 opt = ipv6_renew_options(sk, np->opt, optname,
 367                                          (struct ipv6_opt_hdr __user *)optval,
 368                                          optlen);
 369                 if (IS_ERR(opt)) {
 370                         retv = PTR_ERR(opt);
 371                         break;
 372                 }
 373
 374                 /* routing header option needs extra check */
 375                 if (optname == IPV6_RTHDR && opt->srcrt) {
 376                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
 377                         if (rthdr->type)
 378                                 goto sticky_done;
 379                         if ((rthdr->hdrlen & 1) ||
 380                             (rthdr->hdrlen >> 1) != rthdr->segments_left)
 381                                 goto sticky_done;
 382                 }
 383
 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 sticky_done:
 405                 if (opt)
 406                         sock_kfree_s(sk, opt, opt->tot_len);
 407                 break;
 408         }
 409
 410         case IPV6_2292PKTOPTIONS:
 411         {
 412                 struct ipv6_txoptions *opt = NULL;
 413                 struct msghdr msg;
 414                 struct flowi fl;
 415                 int junk;
 416
 417                 fl.fl6_flowlabel = 0;
 418                 fl.oif = sk->sk_bound_dev_if;
 419
 420                 if (optlen == 0)
 421                         goto update;
 422
 423                 /* 1K is probably excessive
 424                  * 1K is surely not enough, 2K per standard header is 16K.
 425                  */
 426                 retv = -EINVAL;
 427                 if (optlen > 64*1024)
 428                         break;
 429
 430                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 431                 retv = -ENOBUFS;
 432                 if (opt == NULL)
 433                         break;
 434
 435                 memset(opt, 0, sizeof(*opt));
 436                 opt->tot_len = sizeof(*opt) + optlen;
 437                 retv = -EFAULT;
 438                 if (copy_from_user(opt+1, optval, optlen))
 439                         goto done;
 440
 441                 msg.msg_controllen = optlen;
 442                 msg.msg_control = (void*)(opt+1);
 443
 444                 retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
 445                 if (retv)
 446                         goto done;
 447 update:
 448                 retv = 0;
 449                 if (inet_sk(sk)->is_icsk) {
 450                         if (opt) {
 451                                 struct inet_connection_sock *icsk = inet_csk(sk);
 452                                 if (!((1 << sk->sk_state) &
 453                                       (TCPF_LISTEN | TCPF_CLOSE))
 454                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
 455                                         icsk->icsk_ext_hdr_len =
 456                                                 opt->opt_flen + opt->opt_nflen;
 457                                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 458                                 }
 459                         }
 460                         opt = xchg(&np->opt, opt);
 461                         sk_dst_reset(sk);
 462                 } else {
 463                         write_lock(&sk->sk_dst_lock);
 464                         opt = xchg(&np->opt, opt);
 465                         write_unlock(&sk->sk_dst_lock);
 466                         sk_dst_reset(sk);
 467                 }
 468
 469 done:
 470                 if (opt)
 471                         sock_kfree_s(sk, opt, opt->tot_len);
 472                 break;
 473         }
 474         case IPV6_UNICAST_HOPS:
 475                 if (val > 255 || val < -1)
 476                         goto e_inval;
 477                 np->hop_limit = val;
 478                 retv = 0;
 479                 break;
 480
 481         case IPV6_MULTICAST_HOPS:
 482                 if (sk->sk_type == SOCK_STREAM)
 483                         goto e_inval;
 484                 if (val > 255 || val < -1)
 485                         goto e_inval;
 486                 np->mcast_hops = val;
 487                 retv = 0;
 488                 break;
 489
 490         case IPV6_MULTICAST_LOOP:
 491                 np->mc_loop = valbool;
 492                 retv = 0;
 493                 break;
 494
 495         case IPV6_MULTICAST_IF:
 496                 if (sk->sk_type == SOCK_STREAM)
 497                         goto e_inval;
 498                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
 499                         goto e_inval;
 500
 501                 if (__dev_get_by_index(val) == NULL) {
 502                         retv = -ENODEV;
 503                         break;
 504                 }
 505                 np->mcast_oif = val;
 506                 retv = 0;
 507                 break;
 508         case IPV6_ADD_MEMBERSHIP:
 509         case IPV6_DROP_MEMBERSHIP:
 510         {
 511                 struct ipv6_mreq mreq;
 512
 513                 retv = -EFAULT;
 514                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 515                         break;
 516
 517                 if (optname == IPV6_ADD_MEMBERSHIP)
 518                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 519                 else
 520                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 521                 break;
 522         }
 523         case IPV6_JOIN_ANYCAST:
 524         case IPV6_LEAVE_ANYCAST:
 525         {
 526                 struct ipv6_mreq mreq;
 527
 528                 if (optlen != sizeof(struct ipv6_mreq))
 529                         goto e_inval;
 530
 531                 retv = -EFAULT;
 532                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 533                         break;
 534
 535                 if (optname == IPV6_JOIN_ANYCAST)
 536                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 537                 else
 538                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 539                 break;
 540         }
 541         case MCAST_JOIN_GROUP:
 542         case MCAST_LEAVE_GROUP:
 543         {
 544                 struct group_req greq;
 545                 struct sockaddr_in6 *psin6;
 546
 547                 retv = -EFAULT;
 548                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
 549                         break;
 550                 if (greq.gr_group.ss_family != AF_INET6) {
 551                         retv = -EADDRNOTAVAIL;
 552                         break;
 553                 }
 554                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 555                 if (optname == MCAST_JOIN_GROUP)
 556                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
 557                                 &psin6->sin6_addr);
 558                 else
 559                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
 560                                 &psin6->sin6_addr);
 561                 break;
 562         }
 563         case MCAST_JOIN_SOURCE_GROUP:
 564         case MCAST_LEAVE_SOURCE_GROUP:
 565         case MCAST_BLOCK_SOURCE:
 566         case MCAST_UNBLOCK_SOURCE:
 567         {
 568                 struct group_source_req greqs;
 569                 int omode, add;
 570
 571                 if (optlen != sizeof(struct group_source_req))
 572                         goto e_inval;
 573                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 574                         retv = -EFAULT;
 575                         break;
 576                 }
 577                 if (greqs.gsr_group.ss_family != AF_INET6 ||
 578                     greqs.gsr_source.ss_family != AF_INET6) {
 579                         retv = -EADDRNOTAVAIL;
 580                         break;
 581                 }
 582                 if (optname == MCAST_BLOCK_SOURCE) {
 583                         omode = MCAST_EXCLUDE;
 584                         add = 1;
 585                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 586                         omode = MCAST_EXCLUDE;
 587                         add = 0;
 588                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 589                         struct sockaddr_in6 *psin6;
 590
 591                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 592                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
 593                                 &psin6->sin6_addr);
 594                         /* prior join w/ different source is ok */
 595                         if (retv && retv != -EADDRINUSE)
 596                                 break;
 597                         omode = MCAST_INCLUDE;
 598                         add = 1;
 599                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 600                         omode = MCAST_INCLUDE;
 601                         add = 0;
 602                 }
 603                 retv = ip6_mc_source(add, omode, sk, &greqs);
 604                 break;
 605         }
 606         case MCAST_MSFILTER:
 607         {
 608                 extern int sysctl_mld_max_msf;
 609                 struct group_filter *gsf;
 610
 611                 if (optlen < GROUP_FILTER_SIZE(0))
 612                         goto e_inval;
 613                 if (optlen > sysctl_optmem_max) {
 614                         retv = -ENOBUFS;
 615                         break;
 616                 }
 617                 gsf = kmalloc(optlen,GFP_KERNEL);
 618                 if (gsf == 0) {
 619                         retv = -ENOBUFS;
 620                         break;
 621                 }
 622                 retv = -EFAULT;
 623                 if (copy_from_user(gsf, optval, optlen)) {
 624                         kfree(gsf);
 625                         break;
 626                 }
 627                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 628                 if (gsf->gf_numsrc >= 0x1ffffffU ||
 629                     gsf->gf_numsrc > sysctl_mld_max_msf) {
 630                         kfree(gsf);
 631                         retv = -ENOBUFS;
 632                         break;
 633                 }
 634                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 635                         kfree(gsf);
 636                         retv = -EINVAL;
 637                         break;
 638                 }
 639                 retv = ip6_mc_msfilter(sk, gsf);
 640                 kfree(gsf);
 641
 642                 break;
 643         }
 644         case IPV6_ROUTER_ALERT:
 645                 retv = ip6_ra_control(sk, val, NULL);
 646                 break;
 647         case IPV6_MTU_DISCOVER:
 648                 if (val<0 || val>2)
 649                         goto e_inval;
 650                 np->pmtudisc = val;
 651                 retv = 0;
 652                 break;
 653         case IPV6_MTU:
 654                 if (val && val < IPV6_MIN_MTU)
 655                         goto e_inval;
 656                 np->frag_size = val;
 657                 retv = 0;
 658                 break;
 659         case IPV6_RECVERR:
 660                 np->recverr = valbool;
 661                 if (!val)
 662                         skb_queue_purge(&sk->sk_error_queue);
 663                 retv = 0;
 664                 break;
 665         case IPV6_FLOWINFO_SEND:
 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         }
 681         release_sock(sk);
 682
 683         return retv;
 684
 685 e_inval:
 686         release_sock(sk);
 687         return -EINVAL;
 688 }
 689
 690 int ipv6_setsockopt(struct sock *sk, int level, int optname,
 691                     char __user *optval, int optlen)
 692 {
 693         int err;
 694
 695         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 696                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 697
 698         if (level != SOL_IPV6)
 699                 return -ENOPROTOOPT;
 700
 701         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 702 #ifdef CONFIG_NETFILTER
 703         /* we need to exclude all possible ENOPROTOOPTs except default case */
 704         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 705                         optname != IPV6_XFRM_POLICY) {
 706                 lock_sock(sk);
 707                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
 708                                 optlen);
 709                 release_sock(sk);
 710         }
 711 #endif
 712         return err;
 713 }
 714
 715
 716 #ifdef CONFIG_COMPAT
 717 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
 718                            char __user *optval, int optlen)
 719 {
 720         int err;
 721
 722         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
 723                 if (udp_prot.compat_setsockopt != NULL)
 724                         return udp_prot.compat_setsockopt(sk, level, optname,
 725                                                           optval, optlen);
 726                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 727         }
 728
 729         if (level != SOL_IPV6)
 730                 return -ENOPROTOOPT;
 731
 732         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 733 #ifdef CONFIG_NETFILTER
 734         /* we need to exclude all possible ENOPROTOOPTs except default case */
 735         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 736             optname != IPV6_XFRM_POLICY) {
 737                 lock_sock(sk);
 738                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
 739                                            optval, optlen);
 740                 release_sock(sk);
 741         }
 742 #endif
 743         return err;
 744 }
 745
 746 EXPORT_SYMBOL(compat_ipv6_setsockopt);
 747 #endif
 748
 749 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_opt_hdr *hdr,
 750                                   char __user *optval, int len)
 751 {
 752         if (!hdr)
 753                 return 0;
 754         len = min_t(int, len, ipv6_optlen(hdr));
 755         if (copy_to_user(optval, hdr, ipv6_optlen(hdr)))
 756                 return -EFAULT;
 757         return len;
 758 }
 759
 760 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 761                     char __user *optval, int __user *optlen)
 762 {
 763         struct ipv6_pinfo *np = inet6_sk(sk);
 764         int len;
 765         int val;
 766
 767         if (get_user(len, optlen))
 768                 return -EFAULT;
 769         switch (optname) {
 770         case IPV6_ADDRFORM:
 771                 if (sk->sk_protocol != IPPROTO_UDP &&
 772                     sk->sk_protocol != IPPROTO_TCP)
 773                         return -EINVAL;
 774                 if (sk->sk_state != TCP_ESTABLISHED)
 775                         return -ENOTCONN;
 776                 val = sk->sk_family;
 777                 break;
 778         case MCAST_MSFILTER:
 779         {
 780                 struct group_filter gsf;
 781                 int err;
 782
 783                 if (len < GROUP_FILTER_SIZE(0))
 784                         return -EINVAL;
 785                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
 786                         return -EFAULT;
 787                 lock_sock(sk);
 788                 err = ip6_mc_msfget(sk, &gsf,
 789                         (struct group_filter __user *)optval, optlen);
 790                 release_sock(sk);
 791                 return err;
 792         }
 793
 794         case IPV6_2292PKTOPTIONS:
 795         {
 796                 struct msghdr msg;
 797                 struct sk_buff *skb;
 798
 799                 if (sk->sk_type != SOCK_STREAM)
 800                         return -ENOPROTOOPT;
 801
 802                 msg.msg_control = optval;
 803                 msg.msg_controllen = len;
 804                 msg.msg_flags = 0;
 805
 806                 lock_sock(sk);
 807                 skb = np->pktoptions;
 808                 if (skb)
 809                         atomic_inc(&skb->users);
 810                 release_sock(sk);
 811
 812                 if (skb) {
 813                         int err = datagram_recv_ctl(sk, &msg, skb);
 814                         kfree_skb(skb);
 815                         if (err)
 816                                 return err;
 817                 } else {
 818                         if (np->rxopt.bits.rxinfo) {
 819                                 struct in6_pktinfo src_info;
 820                                 src_info.ipi6_ifindex = np->mcast_oif;
 821                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 822                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 823                         }
 824                         if (np->rxopt.bits.rxhlim) {
 825                                 int hlim = np->mcast_hops;
 826                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 827                         }
 828                         if (np->rxopt.bits.rxoinfo) {
 829                                 struct in6_pktinfo src_info;
 830                                 src_info.ipi6_ifindex = np->mcast_oif;
 831                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 832                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 833                         }
 834                         if (np->rxopt.bits.rxohlim) {
 835                                 int hlim = np->mcast_hops;
 836                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 837                         }
 838                 }
 839                 len -= msg.msg_controllen;
 840                 return put_user(len, optlen);
 841         }
 842         case IPV6_MTU:
 843         {
 844                 struct dst_entry *dst;
 845                 val = 0;
 846                 lock_sock(sk);
 847                 dst = sk_dst_get(sk);
 848                 if (dst) {
 849                         val = dst_mtu(dst);
 850                         dst_release(dst);
 851                 }
 852                 release_sock(sk);
 853                 if (!val)
 854                         return -ENOTCONN;
 855                 break;
 856         }
 857
 858         case IPV6_V6ONLY:
 859                 val = np->ipv6only;
 860                 break;
 861
 862         case IPV6_RECVPKTINFO:
 863                 val = np->rxopt.bits.rxinfo;
 864                 break;
 865
 866         case IPV6_2292PKTINFO:
 867                 val = np->rxopt.bits.rxoinfo;
 868                 break;
 869
 870         case IPV6_RECVHOPLIMIT:
 871                 val = np->rxopt.bits.rxhlim;
 872                 break;
 873
 874         case IPV6_2292HOPLIMIT:
 875                 val = np->rxopt.bits.rxohlim;
 876                 break;
 877
 878         case IPV6_RECVRTHDR:
 879                 val = np->rxopt.bits.srcrt;
 880                 break;
 881
 882         case IPV6_2292RTHDR:
 883                 val = np->rxopt.bits.osrcrt;
 884                 break;
 885
 886         case IPV6_HOPOPTS:
 887         case IPV6_RTHDRDSTOPTS:
 888         case IPV6_RTHDR:
 889         case IPV6_DSTOPTS:
 890         {
 891
 892                 lock_sock(sk);
 893                 len = ipv6_getsockopt_sticky(sk, np->opt->hopopt,
 894                                              optval, len);
 895                 release_sock(sk);
 896                 return put_user(len, optlen);
 897         }
 898
 899         case IPV6_RECVHOPOPTS:
 900                 val = np->rxopt.bits.hopopts;
 901                 break;
 902
 903         case IPV6_2292HOPOPTS:
 904                 val = np->rxopt.bits.ohopopts;
 905                 break;
 906
 907         case IPV6_RECVDSTOPTS:
 908                 val = np->rxopt.bits.dstopts;
 909                 break;
 910
 911         case IPV6_2292DSTOPTS:
 912                 val = np->rxopt.bits.odstopts;
 913                 break;
 914
 915         case IPV6_TCLASS:
 916                 val = np->tclass;
 917                 break;
 918
 919         case IPV6_RECVTCLASS:
 920                 val = np->rxopt.bits.rxtclass;
 921                 break;
 922
 923         case IPV6_FLOWINFO:
 924                 val = np->rxopt.bits.rxflow;
 925                 break;
 926
 927         case IPV6_UNICAST_HOPS:
 928                 val = np->hop_limit;
 929                 break;
 930
 931         case IPV6_MULTICAST_HOPS:
 932                 val = np->mcast_hops;
 933                 break;
 934
 935         case IPV6_MULTICAST_LOOP:
 936                 val = np->mc_loop;
 937                 break;
 938
 939         case IPV6_MULTICAST_IF:
 940                 val = np->mcast_oif;
 941                 break;
 942
 943         case IPV6_MTU_DISCOVER:
 944                 val = np->pmtudisc;
 945                 break;
 946
 947         case IPV6_RECVERR:
 948                 val = np->recverr;
 949                 break;
 950
 951         case IPV6_FLOWINFO_SEND:
 952                 val = np->sndflow;
 953                 break;
 954
 955         default:
 956                 return -EINVAL;
 957         }
 958         len = min_t(unsigned int, sizeof(int), len);
 959         if(put_user(len, optlen))
 960                 return -EFAULT;
 961         if(copy_to_user(optval,&val,len))
 962                 return -EFAULT;
 963         return 0;
 964 }
 965
 966 int ipv6_getsockopt(struct sock *sk, int level, int optname,
 967                     char __user *optval, int __user *optlen)
 968 {
 969         int err;
 970
 971         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 972                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
 973
 974         if(level != SOL_IPV6)
 975                 return -ENOPROTOOPT;
 976
 977         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
 978 #ifdef CONFIG_NETFILTER
 979         /* we need to exclude all possible EINVALs except default case */
 980         if (err == -EINVAL && optname != IPV6_ADDRFORM &&
 981                         optname != MCAST_MSFILTER) {
 982                 int len;
 983
 984                 if (get_user(len, optlen))
 985                         return -EFAULT;
 986
 987                 lock_sock(sk);
 988                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
 989                                 &len);
 990                 release_sock(sk);
 991                 if (err >= 0)
 992                         err = put_user(len, optlen);
 993         }
 994 #endif
 995         return err;
 996 }
 997
 998 #ifdef CONFIG_COMPAT
 999 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1000                            char __user *optval, int __user *optlen)
1001 {
1002         int err;
1003
1004         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1005                 if (udp_prot.compat_getsockopt != NULL)
1006                         return udp_prot.compat_getsockopt(sk, level, optname,
1007                                                           optval, optlen);
1008                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1009         }
1010
1011         if (level != SOL_IPV6)
1012                 return -ENOPROTOOPT;
1013
1014         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1015 #ifdef CONFIG_NETFILTER
1016         /* we need to exclude all possible EINVALs except default case */
1017         if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1018                         optname != MCAST_MSFILTER) {
1019                 int len;
1020
1021                 if (get_user(len, optlen))
1022                         return -EFAULT;
1023
1024                 lock_sock(sk);
1025                 err = compat_nf_getsockopt(sk, PF_INET6,
1026                                            optname, optval, &len);
1027                 release_sock(sk);
1028                 if (err >= 0)
1029                         err = put_user(len, optlen);
1030         }
1031 #endif
1032         return err;
1033 }
1034
1035 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1036 #endif
1037
1038 void __init ipv6_packet_init(void)
1039 {
1040         dev_add_pack(&ipv6_packet_type);
1041 }
1042
1043 void ipv6_packet_cleanup(void)
1044 {
1045         dev_remove_pack(&ipv6_packet_type);
1046 }