git.oblomov.eu Git - linux-2.6/blob - net/ipv6/ipv6_sockglue.c

   1 /*
   2  *      IPv6 BSD socket options interface
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Based on linux/net/ipv4/ip_sockglue.c
   9  *
  10  *      $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 inet6_protocol *ipv6_gso_pull_exthdrs(struct sk_buff *skb,
  61                                                     int proto)
  62 {
  63         struct inet6_protocol *ops = NULL;
  64
  65         for (;;) {
  66                 struct ipv6_opt_hdr *opth;
  67                 int len;
  68
  69                 if (proto != NEXTHDR_HOP) {
  70                         ops = rcu_dereference(inet6_protos[proto]);
  71
  72                         if (unlikely(!ops))
  73                                 break;
  74
  75                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
  76                                 break;
  77                 }
  78
  79                 if (unlikely(!pskb_may_pull(skb, 8)))
  80                         break;
  81
  82                 opth = (void *)skb->data;
  83                 len = opth->hdrlen * 8 + 8;
  84
  85                 if (unlikely(!pskb_may_pull(skb, len)))
  86                         break;
  87
  88                 proto = opth->nexthdr;
  89                 __skb_pull(skb, len);
  90         }
  91
  92         return ops;
  93 }
  94
  95 static int ipv6_gso_send_check(struct sk_buff *skb)
  96 {
  97         struct ipv6hdr *ipv6h;
  98         struct inet6_protocol *ops;
  99         int err = -EINVAL;
 100
 101         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 102                 goto out;
 103
 104         ipv6h = skb->nh.ipv6h;
 105         __skb_pull(skb, sizeof(*ipv6h));
 106         err = -EPROTONOSUPPORT;
 107
 108         rcu_read_lock();
 109         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
 110         if (likely(ops && ops->gso_send_check)) {
 111                 skb->h.raw = skb->data;
 112                 err = ops->gso_send_check(skb);
 113         }
 114         rcu_read_unlock();
 115
 116 out:
 117         return err;
 118 }
 119
 120 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
 121 {
 122         struct sk_buff *segs = ERR_PTR(-EINVAL);
 123         struct ipv6hdr *ipv6h;
 124         struct inet6_protocol *ops;
 125
 126         if (unlikely(skb_shinfo(skb)->gso_type &
 127                      ~(SKB_GSO_UDP |
 128                        SKB_GSO_DODGY |
 129                        SKB_GSO_TCP_ECN |
 130                        SKB_GSO_TCPV6 |
 131                        0)))
 132                 goto out;
 133
 134         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 135                 goto out;
 136
 137         ipv6h = skb->nh.ipv6h;
 138         __skb_pull(skb, sizeof(*ipv6h));
 139         segs = ERR_PTR(-EPROTONOSUPPORT);
 140
 141         rcu_read_lock();
 142         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
 143         if (likely(ops && ops->gso_segment)) {
 144                 skb->h.raw = skb->data;
 145                 segs = ops->gso_segment(skb, features);
 146         }
 147         rcu_read_unlock();
 148
 149         if (unlikely(IS_ERR(segs)))
 150                 goto out;
 151
 152         for (skb = segs; skb; skb = skb->next) {
 153                 ipv6h = skb->nh.ipv6h;
 154                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
 155                                            sizeof(*ipv6h));
 156         }
 157
 158 out:
 159         return segs;
 160 }
 161
 162 static struct packet_type ipv6_packet_type = {
 163         .type = __constant_htons(ETH_P_IPV6),
 164         .func = ipv6_rcv,
 165         .gso_send_check = ipv6_gso_send_check,
 166         .gso_segment = ipv6_gso_segment,
 167 };
 168
 169 struct ip6_ra_chain *ip6_ra_chain;
 170 DEFINE_RWLOCK(ip6_ra_lock);
 171
 172 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
 173 {
 174         struct ip6_ra_chain *ra, *new_ra, **rap;
 175
 176         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
 177         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
 178                 return -EINVAL;
 179
 180         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
 181
 182         write_lock_bh(&ip6_ra_lock);
 183         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
 184                 if (ra->sk == sk) {
 185                         if (sel>=0) {
 186                                 write_unlock_bh(&ip6_ra_lock);
 187                                 kfree(new_ra);
 188                                 return -EADDRINUSE;
 189                         }
 190
 191                         *rap = ra->next;
 192                         write_unlock_bh(&ip6_ra_lock);
 193
 194                         if (ra->destructor)
 195                                 ra->destructor(sk);
 196                         sock_put(sk);
 197                         kfree(ra);
 198                         return 0;
 199                 }
 200         }
 201         if (new_ra == NULL) {
 202                 write_unlock_bh(&ip6_ra_lock);
 203                 return -ENOBUFS;
 204         }
 205         new_ra->sk = sk;
 206         new_ra->sel = sel;
 207         new_ra->destructor = destructor;
 208         new_ra->next = ra;
 209         *rap = new_ra;
 210         sock_hold(sk);
 211         write_unlock_bh(&ip6_ra_lock);
 212         return 0;
 213 }
 214
 215 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 216                     char __user *optval, int optlen)
 217 {
 218         struct ipv6_pinfo *np = inet6_sk(sk);
 219         int val, valbool;
 220         int retv = -ENOPROTOOPT;
 221
 222         if (optval == NULL)
 223                 val=0;
 224         else if (get_user(val, (int __user *) optval))
 225                 return -EFAULT;
 226
 227         valbool = (val!=0);
 228
 229         lock_sock(sk);
 230
 231         switch (optname) {
 232
 233         case IPV6_ADDRFORM:
 234                 if (val == PF_INET) {
 235                         struct ipv6_txoptions *opt;
 236                         struct sk_buff *pktopt;
 237
 238                         if (sk->sk_protocol != IPPROTO_UDP &&
 239                             sk->sk_protocol != IPPROTO_TCP)
 240                                 break;
 241
 242                         if (sk->sk_state != TCP_ESTABLISHED) {
 243                                 retv = -ENOTCONN;
 244                                 break;
 245                         }
 246
 247                         if (ipv6_only_sock(sk) ||
 248                             !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
 249                                 retv = -EADDRNOTAVAIL;
 250                                 break;
 251                         }
 252
 253                         fl6_free_socklist(sk);
 254                         ipv6_sock_mc_close(sk);
 255
 256                         /*
 257                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
 258                          * remove it from the refcnt debug socks count in the
 259                          * original family...
 260                          */
 261                         sk_refcnt_debug_dec(sk);
 262
 263                         if (sk->sk_protocol == IPPROTO_TCP) {
 264                                 struct inet_connection_sock *icsk = inet_csk(sk);
 265
 266                                 local_bh_disable();
 267                                 sock_prot_dec_use(sk->sk_prot);
 268                                 sock_prot_inc_use(&tcp_prot);
 269                                 local_bh_enable();
 270                                 sk->sk_prot = &tcp_prot;
 271                                 icsk->icsk_af_ops = &ipv4_specific;
 272                                 sk->sk_socket->ops = &inet_stream_ops;
 273                                 sk->sk_family = PF_INET;
 274                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 275                         } else {
 276                                 local_bh_disable();
 277                                 sock_prot_dec_use(sk->sk_prot);
 278                                 sock_prot_inc_use(&udp_prot);
 279                                 local_bh_enable();
 280                                 sk->sk_prot = &udp_prot;
 281                                 sk->sk_socket->ops = &inet_dgram_ops;
 282                                 sk->sk_family = PF_INET;
 283                         }
 284                         opt = xchg(&np->opt, NULL);
 285                         if (opt)
 286                                 sock_kfree_s(sk, opt, opt->tot_len);
 287                         pktopt = xchg(&np->pktoptions, NULL);
 288                         if (pktopt)
 289                                 kfree_skb(pktopt);
 290
 291                         sk->sk_destruct = inet_sock_destruct;
 292                         /*
 293                          * ... and add it to the refcnt debug socks count
 294                          * in the new family. -acme
 295                          */
 296                         sk_refcnt_debug_inc(sk);
 297                         module_put(THIS_MODULE);
 298                         retv = 0;
 299                         break;
 300                 }
 301                 goto e_inval;
 302
 303         case IPV6_V6ONLY:
 304                 if (inet_sk(sk)->num)
 305                         goto e_inval;
 306                 np->ipv6only = valbool;
 307                 retv = 0;
 308                 break;
 309
 310         case IPV6_RECVPKTINFO:
 311                 np->rxopt.bits.rxinfo = valbool;
 312                 retv = 0;
 313                 break;
 314
 315         case IPV6_2292PKTINFO:
 316                 np->rxopt.bits.rxoinfo = valbool;
 317                 retv = 0;
 318                 break;
 319
 320         case IPV6_RECVHOPLIMIT:
 321                 np->rxopt.bits.rxhlim = valbool;
 322                 retv = 0;
 323                 break;
 324
 325         case IPV6_2292HOPLIMIT:
 326                 np->rxopt.bits.rxohlim = valbool;
 327                 retv = 0;
 328                 break;
 329
 330         case IPV6_RECVRTHDR:
 331                 if (val < 0 || val > 2)
 332                         goto e_inval;
 333                 np->rxopt.bits.srcrt = val;
 334                 retv = 0;
 335                 break;
 336
 337         case IPV6_2292RTHDR:
 338                 if (val < 0 || val > 2)
 339                         goto e_inval;
 340                 np->rxopt.bits.osrcrt = val;
 341                 retv = 0;
 342                 break;
 343
 344         case IPV6_RECVHOPOPTS:
 345                 np->rxopt.bits.hopopts = valbool;
 346                 retv = 0;
 347                 break;
 348
 349         case IPV6_2292HOPOPTS:
 350                 np->rxopt.bits.ohopopts = valbool;
 351                 retv = 0;
 352                 break;
 353
 354         case IPV6_RECVDSTOPTS:
 355                 np->rxopt.bits.dstopts = valbool;
 356                 retv = 0;
 357                 break;
 358
 359         case IPV6_2292DSTOPTS:
 360                 np->rxopt.bits.odstopts = valbool;
 361                 retv = 0;
 362                 break;
 363
 364         case IPV6_TCLASS:
 365                 if (val < 0 || val > 0xff)
 366                         goto e_inval;
 367                 np->tclass = val;
 368                 retv = 0;
 369                 break;
 370
 371         case IPV6_RECVTCLASS:
 372                 np->rxopt.bits.rxtclass = valbool;
 373                 retv = 0;
 374                 break;
 375
 376         case IPV6_FLOWINFO:
 377                 np->rxopt.bits.rxflow = valbool;
 378                 retv = 0;
 379                 break;
 380
 381         case IPV6_HOPOPTS:
 382         case IPV6_RTHDRDSTOPTS:
 383         case IPV6_RTHDR:
 384         case IPV6_DSTOPTS:
 385         {
 386                 struct ipv6_txoptions *opt;
 387                 if (optlen == 0)
 388                         optval = NULL;
 389
 390                 /* hop-by-hop / destination options are privileged option */
 391                 retv = -EPERM;
 392                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
 393                         break;
 394
 395                 retv = -EINVAL;
 396                 if (optlen & 0x7 || optlen > 8 * 255)
 397                         break;
 398
 399                 opt = ipv6_renew_options(sk, np->opt, optname,
 400                                          (struct ipv6_opt_hdr __user *)optval,
 401                                          optlen);
 402                 if (IS_ERR(opt)) {
 403                         retv = PTR_ERR(opt);
 404                         break;
 405                 }
 406
 407                 /* routing header option needs extra check */
 408                 if (optname == IPV6_RTHDR && opt->srcrt) {
 409                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
 410                         if (rthdr->type)
 411                                 goto sticky_done;
 412                         if ((rthdr->hdrlen & 1) ||
 413                             (rthdr->hdrlen >> 1) != rthdr->segments_left)
 414                                 goto sticky_done;
 415                 }
 416
 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 sticky_done:
 438                 if (opt)
 439                         sock_kfree_s(sk, opt, opt->tot_len);
 440                 break;
 441         }
 442
 443         case IPV6_2292PKTOPTIONS:
 444         {
 445                 struct ipv6_txoptions *opt = NULL;
 446                 struct msghdr msg;
 447                 struct flowi fl;
 448                 int junk;
 449
 450                 fl.fl6_flowlabel = 0;
 451                 fl.oif = sk->sk_bound_dev_if;
 452
 453                 if (optlen == 0)
 454                         goto update;
 455
 456                 /* 1K is probably excessive
 457                  * 1K is surely not enough, 2K per standard header is 16K.
 458                  */
 459                 retv = -EINVAL;
 460                 if (optlen > 64*1024)
 461                         break;
 462
 463                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 464                 retv = -ENOBUFS;
 465                 if (opt == NULL)
 466                         break;
 467
 468                 memset(opt, 0, sizeof(*opt));
 469                 opt->tot_len = sizeof(*opt) + optlen;
 470                 retv = -EFAULT;
 471                 if (copy_from_user(opt+1, optval, optlen))
 472                         goto done;
 473
 474                 msg.msg_controllen = optlen;
 475                 msg.msg_control = (void*)(opt+1);
 476
 477                 retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
 478                 if (retv)
 479                         goto done;
 480 update:
 481                 retv = 0;
 482                 if (inet_sk(sk)->is_icsk) {
 483                         if (opt) {
 484                                 struct inet_connection_sock *icsk = inet_csk(sk);
 485                                 if (!((1 << sk->sk_state) &
 486                                       (TCPF_LISTEN | TCPF_CLOSE))
 487                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
 488                                         icsk->icsk_ext_hdr_len =
 489                                                 opt->opt_flen + opt->opt_nflen;
 490                                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 491                                 }
 492                         }
 493                         opt = xchg(&np->opt, opt);
 494                         sk_dst_reset(sk);
 495                 } else {
 496                         write_lock(&sk->sk_dst_lock);
 497                         opt = xchg(&np->opt, opt);
 498                         write_unlock(&sk->sk_dst_lock);
 499                         sk_dst_reset(sk);
 500                 }
 501
 502 done:
 503                 if (opt)
 504                         sock_kfree_s(sk, opt, opt->tot_len);
 505                 break;
 506         }
 507         case IPV6_UNICAST_HOPS:
 508                 if (val > 255 || val < -1)
 509                         goto e_inval;
 510                 np->hop_limit = val;
 511                 retv = 0;
 512                 break;
 513
 514         case IPV6_MULTICAST_HOPS:
 515                 if (sk->sk_type == SOCK_STREAM)
 516                         goto e_inval;
 517                 if (val > 255 || val < -1)
 518                         goto e_inval;
 519                 np->mcast_hops = val;
 520                 retv = 0;
 521                 break;
 522
 523         case IPV6_MULTICAST_LOOP:
 524                 np->mc_loop = valbool;
 525                 retv = 0;
 526                 break;
 527
 528         case IPV6_MULTICAST_IF:
 529                 if (sk->sk_type == SOCK_STREAM)
 530                         goto e_inval;
 531                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
 532                         goto e_inval;
 533
 534                 if (__dev_get_by_index(val) == NULL) {
 535                         retv = -ENODEV;
 536                         break;
 537                 }
 538                 np->mcast_oif = val;
 539                 retv = 0;
 540                 break;
 541         case IPV6_ADD_MEMBERSHIP:
 542         case IPV6_DROP_MEMBERSHIP:
 543         {
 544                 struct ipv6_mreq mreq;
 545
 546                 retv = -EFAULT;
 547                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 548                         break;
 549
 550                 if (optname == IPV6_ADD_MEMBERSHIP)
 551                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 552                 else
 553                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 554                 break;
 555         }
 556         case IPV6_JOIN_ANYCAST:
 557         case IPV6_LEAVE_ANYCAST:
 558         {
 559                 struct ipv6_mreq mreq;
 560
 561                 if (optlen != sizeof(struct ipv6_mreq))
 562                         goto e_inval;
 563
 564                 retv = -EFAULT;
 565                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 566                         break;
 567
 568                 if (optname == IPV6_JOIN_ANYCAST)
 569                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 570                 else
 571                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 572                 break;
 573         }
 574         case MCAST_JOIN_GROUP:
 575         case MCAST_LEAVE_GROUP:
 576         {
 577                 struct group_req greq;
 578                 struct sockaddr_in6 *psin6;
 579
 580                 retv = -EFAULT;
 581                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
 582                         break;
 583                 if (greq.gr_group.ss_family != AF_INET6) {
 584                         retv = -EADDRNOTAVAIL;
 585                         break;
 586                 }
 587                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 588                 if (optname == MCAST_JOIN_GROUP)
 589                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
 590                                 &psin6->sin6_addr);
 591                 else
 592                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
 593                                 &psin6->sin6_addr);
 594                 break;
 595         }
 596         case MCAST_JOIN_SOURCE_GROUP:
 597         case MCAST_LEAVE_SOURCE_GROUP:
 598         case MCAST_BLOCK_SOURCE:
 599         case MCAST_UNBLOCK_SOURCE:
 600         {
 601                 struct group_source_req greqs;
 602                 int omode, add;
 603
 604                 if (optlen != sizeof(struct group_source_req))
 605                         goto e_inval;
 606                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 607                         retv = -EFAULT;
 608                         break;
 609                 }
 610                 if (greqs.gsr_group.ss_family != AF_INET6 ||
 611                     greqs.gsr_source.ss_family != AF_INET6) {
 612                         retv = -EADDRNOTAVAIL;
 613                         break;
 614                 }
 615                 if (optname == MCAST_BLOCK_SOURCE) {
 616                         omode = MCAST_EXCLUDE;
 617                         add = 1;
 618                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 619                         omode = MCAST_EXCLUDE;
 620                         add = 0;
 621                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 622                         struct sockaddr_in6 *psin6;
 623
 624                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 625                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
 626                                 &psin6->sin6_addr);
 627                         /* prior join w/ different source is ok */
 628                         if (retv && retv != -EADDRINUSE)
 629                                 break;
 630                         omode = MCAST_INCLUDE;
 631                         add = 1;
 632                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 633                         omode = MCAST_INCLUDE;
 634                         add = 0;
 635                 }
 636                 retv = ip6_mc_source(add, omode, sk, &greqs);
 637                 break;
 638         }
 639         case MCAST_MSFILTER:
 640         {
 641                 extern int sysctl_mld_max_msf;
 642                 struct group_filter *gsf;
 643
 644                 if (optlen < GROUP_FILTER_SIZE(0))
 645                         goto e_inval;
 646                 if (optlen > sysctl_optmem_max) {
 647                         retv = -ENOBUFS;
 648                         break;
 649                 }
 650                 gsf = kmalloc(optlen,GFP_KERNEL);
 651                 if (gsf == 0) {
 652                         retv = -ENOBUFS;
 653                         break;
 654                 }
 655                 retv = -EFAULT;
 656                 if (copy_from_user(gsf, optval, optlen)) {
 657                         kfree(gsf);
 658                         break;
 659                 }
 660                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 661                 if (gsf->gf_numsrc >= 0x1ffffffU ||
 662                     gsf->gf_numsrc > sysctl_mld_max_msf) {
 663                         kfree(gsf);
 664                         retv = -ENOBUFS;
 665                         break;
 666                 }
 667                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 668                         kfree(gsf);
 669                         retv = -EINVAL;
 670                         break;
 671                 }
 672                 retv = ip6_mc_msfilter(sk, gsf);
 673                 kfree(gsf);
 674
 675                 break;
 676         }
 677         case IPV6_ROUTER_ALERT:
 678                 retv = ip6_ra_control(sk, val, NULL);
 679                 break;
 680         case IPV6_MTU_DISCOVER:
 681                 if (val<0 || val>2)
 682                         goto e_inval;
 683                 np->pmtudisc = val;
 684                 retv = 0;
 685                 break;
 686         case IPV6_MTU:
 687                 if (val && val < IPV6_MIN_MTU)
 688                         goto e_inval;
 689                 np->frag_size = val;
 690                 retv = 0;
 691                 break;
 692         case IPV6_RECVERR:
 693                 np->recverr = valbool;
 694                 if (!val)
 695                         skb_queue_purge(&sk->sk_error_queue);
 696                 retv = 0;
 697                 break;
 698         case IPV6_FLOWINFO_SEND:
 699                 np->sndflow = valbool;
 700                 retv = 0;
 701                 break;
 702         case IPV6_FLOWLABEL_MGR:
 703                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
 704                 break;
 705         case IPV6_IPSEC_POLICY:
 706         case IPV6_XFRM_POLICY:
 707                 retv = -EPERM;
 708                 if (!capable(CAP_NET_ADMIN))
 709                         break;
 710                 retv = xfrm_user_policy(sk, optname, optval, optlen);
 711                 break;
 712
 713         }
 714         release_sock(sk);
 715
 716         return retv;
 717
 718 e_inval:
 719         release_sock(sk);
 720         return -EINVAL;
 721 }
 722
 723 int ipv6_setsockopt(struct sock *sk, int level, int optname,
 724                     char __user *optval, int optlen)
 725 {
 726         int err;
 727
 728         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 729                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 730
 731         if (level != SOL_IPV6)
 732                 return -ENOPROTOOPT;
 733
 734         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 735 #ifdef CONFIG_NETFILTER
 736         /* we need to exclude all possible ENOPROTOOPTs except default case */
 737         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 738                         optname != IPV6_XFRM_POLICY) {
 739                 lock_sock(sk);
 740                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
 741                                 optlen);
 742                 release_sock(sk);
 743         }
 744 #endif
 745         return err;
 746 }
 747
 748
 749 #ifdef CONFIG_COMPAT
 750 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
 751                            char __user *optval, int optlen)
 752 {
 753         int err;
 754
 755         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
 756                 if (udp_prot.compat_setsockopt != NULL)
 757                         return udp_prot.compat_setsockopt(sk, level, optname,
 758                                                           optval, optlen);
 759                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 760         }
 761
 762         if (level != SOL_IPV6)
 763                 return -ENOPROTOOPT;
 764
 765         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 766 #ifdef CONFIG_NETFILTER
 767         /* we need to exclude all possible ENOPROTOOPTs except default case */
 768         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 769             optname != IPV6_XFRM_POLICY) {
 770                 lock_sock(sk);
 771                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
 772                                            optval, optlen);
 773                 release_sock(sk);
 774         }
 775 #endif
 776         return err;
 777 }
 778
 779 EXPORT_SYMBOL(compat_ipv6_setsockopt);
 780 #endif
 781
 782 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_opt_hdr *hdr,
 783                                   char __user *optval, int len)
 784 {
 785         if (!hdr)
 786                 return 0;
 787         len = min_t(int, len, ipv6_optlen(hdr));
 788         if (copy_to_user(optval, hdr, ipv6_optlen(hdr)))
 789                 return -EFAULT;
 790         return len;
 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 (get_user(len, optlen))
 801                 return -EFAULT;
 802         switch (optname) {
 803         case IPV6_ADDRFORM:
 804                 if (sk->sk_protocol != IPPROTO_UDP &&
 805                     sk->sk_protocol != IPPROTO_TCP)
 806                         return -EINVAL;
 807                 if (sk->sk_state != TCP_ESTABLISHED)
 808                         return -ENOTCONN;
 809                 val = sk->sk_family;
 810                 break;
 811         case MCAST_MSFILTER:
 812         {
 813                 struct group_filter gsf;
 814                 int err;
 815
 816                 if (len < GROUP_FILTER_SIZE(0))
 817                         return -EINVAL;
 818                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
 819                         return -EFAULT;
 820                 lock_sock(sk);
 821                 err = ip6_mc_msfget(sk, &gsf,
 822                         (struct group_filter __user *)optval, optlen);
 823                 release_sock(sk);
 824                 return err;
 825         }
 826
 827         case IPV6_2292PKTOPTIONS:
 828         {
 829                 struct msghdr msg;
 830                 struct sk_buff *skb;
 831
 832                 if (sk->sk_type != SOCK_STREAM)
 833                         return -ENOPROTOOPT;
 834
 835                 msg.msg_control = optval;
 836                 msg.msg_controllen = len;
 837                 msg.msg_flags = 0;
 838
 839                 lock_sock(sk);
 840                 skb = np->pktoptions;
 841                 if (skb)
 842                         atomic_inc(&skb->users);
 843                 release_sock(sk);
 844
 845                 if (skb) {
 846                         int err = datagram_recv_ctl(sk, &msg, skb);
 847                         kfree_skb(skb);
 848                         if (err)
 849                                 return err;
 850                 } else {
 851                         if (np->rxopt.bits.rxinfo) {
 852                                 struct in6_pktinfo src_info;
 853                                 src_info.ipi6_ifindex = np->mcast_oif;
 854                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 855                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 856                         }
 857                         if (np->rxopt.bits.rxhlim) {
 858                                 int hlim = np->mcast_hops;
 859                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 860                         }
 861                         if (np->rxopt.bits.rxoinfo) {
 862                                 struct in6_pktinfo src_info;
 863                                 src_info.ipi6_ifindex = np->mcast_oif;
 864                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 865                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 866                         }
 867                         if (np->rxopt.bits.rxohlim) {
 868                                 int hlim = np->mcast_hops;
 869                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 870                         }
 871                 }
 872                 len -= msg.msg_controllen;
 873                 return put_user(len, optlen);
 874         }
 875         case IPV6_MTU:
 876         {
 877                 struct dst_entry *dst;
 878                 val = 0;
 879                 lock_sock(sk);
 880                 dst = sk_dst_get(sk);
 881                 if (dst) {
 882                         val = dst_mtu(dst);
 883                         dst_release(dst);
 884                 }
 885                 release_sock(sk);
 886                 if (!val)
 887                         return -ENOTCONN;
 888                 break;
 889         }
 890
 891         case IPV6_V6ONLY:
 892                 val = np->ipv6only;
 893                 break;
 894
 895         case IPV6_RECVPKTINFO:
 896                 val = np->rxopt.bits.rxinfo;
 897                 break;
 898
 899         case IPV6_2292PKTINFO:
 900                 val = np->rxopt.bits.rxoinfo;
 901                 break;
 902
 903         case IPV6_RECVHOPLIMIT:
 904                 val = np->rxopt.bits.rxhlim;
 905                 break;
 906
 907         case IPV6_2292HOPLIMIT:
 908                 val = np->rxopt.bits.rxohlim;
 909                 break;
 910
 911         case IPV6_RECVRTHDR:
 912                 val = np->rxopt.bits.srcrt;
 913                 break;
 914
 915         case IPV6_2292RTHDR:
 916                 val = np->rxopt.bits.osrcrt;
 917                 break;
 918
 919         case IPV6_HOPOPTS:
 920         case IPV6_RTHDRDSTOPTS:
 921         case IPV6_RTHDR:
 922         case IPV6_DSTOPTS:
 923         {
 924
 925                 lock_sock(sk);
 926                 len = ipv6_getsockopt_sticky(sk, np->opt->hopopt,
 927                                              optval, len);
 928                 release_sock(sk);
 929                 return put_user(len, optlen);
 930         }
 931
 932         case IPV6_RECVHOPOPTS:
 933                 val = np->rxopt.bits.hopopts;
 934                 break;
 935
 936         case IPV6_2292HOPOPTS:
 937                 val = np->rxopt.bits.ohopopts;
 938                 break;
 939
 940         case IPV6_RECVDSTOPTS:
 941                 val = np->rxopt.bits.dstopts;
 942                 break;
 943
 944         case IPV6_2292DSTOPTS:
 945                 val = np->rxopt.bits.odstopts;
 946                 break;
 947
 948         case IPV6_TCLASS:
 949                 val = np->tclass;
 950                 break;
 951
 952         case IPV6_RECVTCLASS:
 953                 val = np->rxopt.bits.rxtclass;
 954                 break;
 955
 956         case IPV6_FLOWINFO:
 957                 val = np->rxopt.bits.rxflow;
 958                 break;
 959
 960         case IPV6_UNICAST_HOPS:
 961                 val = np->hop_limit;
 962                 break;
 963
 964         case IPV6_MULTICAST_HOPS:
 965                 val = np->mcast_hops;
 966                 break;
 967
 968         case IPV6_MULTICAST_LOOP:
 969                 val = np->mc_loop;
 970                 break;
 971
 972         case IPV6_MULTICAST_IF:
 973                 val = np->mcast_oif;
 974                 break;
 975
 976         case IPV6_MTU_DISCOVER:
 977                 val = np->pmtudisc;
 978                 break;
 979
 980         case IPV6_RECVERR:
 981                 val = np->recverr;
 982                 break;
 983
 984         case IPV6_FLOWINFO_SEND:
 985                 val = np->sndflow;
 986                 break;
 987
 988         default:
 989                 return -EINVAL;
 990         }
 991         len = min_t(unsigned int, sizeof(int), len);
 992         if(put_user(len, optlen))
 993                 return -EFAULT;
 994         if(copy_to_user(optval,&val,len))
 995                 return -EFAULT;
 996         return 0;
 997 }
 998
 999 int 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                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1006
1007         if(level != SOL_IPV6)
1008                 return -ENOPROTOOPT;
1009
1010         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1011 #ifdef CONFIG_NETFILTER
1012         /* we need to exclude all possible EINVALs except default case */
1013         if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1014                         optname != MCAST_MSFILTER) {
1015                 int len;
1016
1017                 if (get_user(len, optlen))
1018                         return -EFAULT;
1019
1020                 lock_sock(sk);
1021                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1022                                 &len);
1023                 release_sock(sk);
1024                 if (err >= 0)
1025                         err = put_user(len, optlen);
1026         }
1027 #endif
1028         return err;
1029 }
1030
1031 #ifdef CONFIG_COMPAT
1032 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1033                            char __user *optval, int __user *optlen)
1034 {
1035         int err;
1036
1037         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1038                 if (udp_prot.compat_getsockopt != NULL)
1039                         return udp_prot.compat_getsockopt(sk, level, optname,
1040                                                           optval, optlen);
1041                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1042         }
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 EINVALs except default case */
1050         if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1051                         optname != MCAST_MSFILTER) {
1052                 int len;
1053
1054                 if (get_user(len, optlen))
1055                         return -EFAULT;
1056
1057                 lock_sock(sk);
1058                 err = compat_nf_getsockopt(sk, PF_INET6,
1059                                            optname, optval, &len);
1060                 release_sock(sk);
1061                 if (err >= 0)
1062                         err = put_user(len, optlen);
1063         }
1064 #endif
1065         return err;
1066 }
1067
1068 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1069 #endif
1070
1071 void __init ipv6_packet_init(void)
1072 {
1073         dev_add_pack(&ipv6_packet_type);
1074 }
1075
1076 void ipv6_packet_cleanup(void)
1077 {
1078         dev_remove_pack(&ipv6_packet_type);
1079 }