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 (!(features & NETIF_F_HW_CSUM))
 127                 features &= ~NETIF_F_SG;
 128
 129         if (unlikely(skb_shinfo(skb)->gso_type &
 130                      ~(SKB_GSO_UDP |
 131                        SKB_GSO_DODGY |
 132                        SKB_GSO_TCP_ECN |
 133                        SKB_GSO_TCPV6 |
 134                        0)))
 135                 goto out;
 136
 137         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 138                 goto out;
 139
 140         ipv6h = skb->nh.ipv6h;
 141         __skb_pull(skb, sizeof(*ipv6h));
 142         segs = ERR_PTR(-EPROTONOSUPPORT);
 143
 144         rcu_read_lock();
 145         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
 146         if (likely(ops && ops->gso_segment)) {
 147                 skb->h.raw = skb->data;
 148                 segs = ops->gso_segment(skb, features);
 149         }
 150         rcu_read_unlock();
 151
 152         if (unlikely(IS_ERR(segs)))
 153                 goto out;
 154
 155         for (skb = segs; skb; skb = skb->next) {
 156                 ipv6h = skb->nh.ipv6h;
 157                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
 158                                            sizeof(*ipv6h));
 159         }
 160
 161 out:
 162         return segs;
 163 }
 164
 165 static struct packet_type ipv6_packet_type = {
 166         .type = __constant_htons(ETH_P_IPV6),
 167         .func = ipv6_rcv,
 168         .gso_send_check = ipv6_gso_send_check,
 169         .gso_segment = ipv6_gso_segment,
 170 };
 171
 172 struct ip6_ra_chain *ip6_ra_chain;
 173 DEFINE_RWLOCK(ip6_ra_lock);
 174
 175 int ip6_ra_control(struct sock *sk, int sel, void (*destructor)(struct sock *))
 176 {
 177         struct ip6_ra_chain *ra, *new_ra, **rap;
 178
 179         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
 180         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num != IPPROTO_RAW)
 181                 return -EINVAL;
 182
 183         new_ra = (sel>=0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
 184
 185         write_lock_bh(&ip6_ra_lock);
 186         for (rap = &ip6_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
 187                 if (ra->sk == sk) {
 188                         if (sel>=0) {
 189                                 write_unlock_bh(&ip6_ra_lock);
 190                                 kfree(new_ra);
 191                                 return -EADDRINUSE;
 192                         }
 193
 194                         *rap = ra->next;
 195                         write_unlock_bh(&ip6_ra_lock);
 196
 197                         if (ra->destructor)
 198                                 ra->destructor(sk);
 199                         sock_put(sk);
 200                         kfree(ra);
 201                         return 0;
 202                 }
 203         }
 204         if (new_ra == NULL) {
 205                 write_unlock_bh(&ip6_ra_lock);
 206                 return -ENOBUFS;
 207         }
 208         new_ra->sk = sk;
 209         new_ra->sel = sel;
 210         new_ra->destructor = destructor;
 211         new_ra->next = ra;
 212         *rap = new_ra;
 213         sock_hold(sk);
 214         write_unlock_bh(&ip6_ra_lock);
 215         return 0;
 216 }
 217
 218 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 219                     char __user *optval, int optlen)
 220 {
 221         struct ipv6_pinfo *np = inet6_sk(sk);
 222         int val, valbool;
 223         int retv = -ENOPROTOOPT;
 224
 225         if (optval == NULL)
 226                 val=0;
 227         else if (get_user(val, (int __user *) optval))
 228                 return -EFAULT;
 229
 230         valbool = (val!=0);
 231
 232         lock_sock(sk);
 233
 234         switch (optname) {
 235
 236         case IPV6_ADDRFORM:
 237                 if (val == PF_INET) {
 238                         struct ipv6_txoptions *opt;
 239                         struct sk_buff *pktopt;
 240
 241                         if (sk->sk_protocol != IPPROTO_UDP &&
 242                             sk->sk_protocol != IPPROTO_TCP)
 243                                 break;
 244
 245                         if (sk->sk_state != TCP_ESTABLISHED) {
 246                                 retv = -ENOTCONN;
 247                                 break;
 248                         }
 249
 250                         if (ipv6_only_sock(sk) ||
 251                             !(ipv6_addr_type(&np->daddr) & IPV6_ADDR_MAPPED)) {
 252                                 retv = -EADDRNOTAVAIL;
 253                                 break;
 254                         }
 255
 256                         fl6_free_socklist(sk);
 257                         ipv6_sock_mc_close(sk);
 258
 259                         /*
 260                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
 261                          * remove it from the refcnt debug socks count in the
 262                          * original family...
 263                          */
 264                         sk_refcnt_debug_dec(sk);
 265
 266                         if (sk->sk_protocol == IPPROTO_TCP) {
 267                                 struct inet_connection_sock *icsk = inet_csk(sk);
 268
 269                                 local_bh_disable();
 270                                 sock_prot_dec_use(sk->sk_prot);
 271                                 sock_prot_inc_use(&tcp_prot);
 272                                 local_bh_enable();
 273                                 sk->sk_prot = &tcp_prot;
 274                                 icsk->icsk_af_ops = &ipv4_specific;
 275                                 sk->sk_socket->ops = &inet_stream_ops;
 276                                 sk->sk_family = PF_INET;
 277                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 278                         } else {
 279                                 local_bh_disable();
 280                                 sock_prot_dec_use(sk->sk_prot);
 281                                 sock_prot_inc_use(&udp_prot);
 282                                 local_bh_enable();
 283                                 sk->sk_prot = &udp_prot;
 284                                 sk->sk_socket->ops = &inet_dgram_ops;
 285                                 sk->sk_family = PF_INET;
 286                         }
 287                         opt = xchg(&np->opt, NULL);
 288                         if (opt)
 289                                 sock_kfree_s(sk, opt, opt->tot_len);
 290                         pktopt = xchg(&np->pktoptions, NULL);
 291                         if (pktopt)
 292                                 kfree_skb(pktopt);
 293
 294                         sk->sk_destruct = inet_sock_destruct;
 295                         /*
 296                          * ... and add it to the refcnt debug socks count
 297                          * in the new family. -acme
 298                          */
 299                         sk_refcnt_debug_inc(sk);
 300                         module_put(THIS_MODULE);
 301                         retv = 0;
 302                         break;
 303                 }
 304                 goto e_inval;
 305
 306         case IPV6_V6ONLY:
 307                 if (inet_sk(sk)->num)
 308                         goto e_inval;
 309                 np->ipv6only = valbool;
 310                 retv = 0;
 311                 break;
 312
 313         case IPV6_RECVPKTINFO:
 314                 np->rxopt.bits.rxinfo = valbool;
 315                 retv = 0;
 316                 break;
 317
 318         case IPV6_2292PKTINFO:
 319                 np->rxopt.bits.rxoinfo = valbool;
 320                 retv = 0;
 321                 break;
 322
 323         case IPV6_RECVHOPLIMIT:
 324                 np->rxopt.bits.rxhlim = valbool;
 325                 retv = 0;
 326                 break;
 327
 328         case IPV6_2292HOPLIMIT:
 329                 np->rxopt.bits.rxohlim = valbool;
 330                 retv = 0;
 331                 break;
 332
 333         case IPV6_RECVRTHDR:
 334                 if (val < 0 || val > 2)
 335                         goto e_inval;
 336                 np->rxopt.bits.srcrt = val;
 337                 retv = 0;
 338                 break;
 339
 340         case IPV6_2292RTHDR:
 341                 if (val < 0 || val > 2)
 342                         goto e_inval;
 343                 np->rxopt.bits.osrcrt = val;
 344                 retv = 0;
 345                 break;
 346
 347         case IPV6_RECVHOPOPTS:
 348                 np->rxopt.bits.hopopts = valbool;
 349                 retv = 0;
 350                 break;
 351
 352         case IPV6_2292HOPOPTS:
 353                 np->rxopt.bits.ohopopts = valbool;
 354                 retv = 0;
 355                 break;
 356
 357         case IPV6_RECVDSTOPTS:
 358                 np->rxopt.bits.dstopts = valbool;
 359                 retv = 0;
 360                 break;
 361
 362         case IPV6_2292DSTOPTS:
 363                 np->rxopt.bits.odstopts = valbool;
 364                 retv = 0;
 365                 break;
 366
 367         case IPV6_TCLASS:
 368                 if (val < -1 || val > 0xff)
 369                         goto e_inval;
 370                 np->tclass = val;
 371                 retv = 0;
 372                 break;
 373
 374         case IPV6_RECVTCLASS:
 375                 np->rxopt.bits.rxtclass = valbool;
 376                 retv = 0;
 377                 break;
 378
 379         case IPV6_FLOWINFO:
 380                 np->rxopt.bits.rxflow = valbool;
 381                 retv = 0;
 382                 break;
 383
 384         case IPV6_HOPOPTS:
 385         case IPV6_RTHDRDSTOPTS:
 386         case IPV6_RTHDR:
 387         case IPV6_DSTOPTS:
 388         {
 389                 struct ipv6_txoptions *opt;
 390                 if (optlen == 0)
 391                         optval = NULL;
 392
 393                 /* hop-by-hop / destination options are privileged option */
 394                 retv = -EPERM;
 395                 if (optname != IPV6_RTHDR && !capable(CAP_NET_RAW))
 396                         break;
 397
 398                 retv = -EINVAL;
 399                 if (optlen & 0x7 || optlen > 8 * 255)
 400                         break;
 401
 402                 opt = ipv6_renew_options(sk, np->opt, optname,
 403                                          (struct ipv6_opt_hdr __user *)optval,
 404                                          optlen);
 405                 if (IS_ERR(opt)) {
 406                         retv = PTR_ERR(opt);
 407                         break;
 408                 }
 409
 410                 /* routing header option needs extra check */
 411                 if (optname == IPV6_RTHDR && opt->srcrt) {
 412                         struct ipv6_rt_hdr *rthdr = opt->srcrt;
 413                         if (rthdr->type)
 414                                 goto sticky_done;
 415                         if ((rthdr->hdrlen & 1) ||
 416                             (rthdr->hdrlen >> 1) != rthdr->segments_left)
 417                                 goto sticky_done;
 418                 }
 419
 420                 retv = 0;
 421                 if (inet_sk(sk)->is_icsk) {
 422                         if (opt) {
 423                                 struct inet_connection_sock *icsk = inet_csk(sk);
 424                                 if (!((1 << sk->sk_state) &
 425                                       (TCPF_LISTEN | TCPF_CLOSE))
 426                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
 427                                         icsk->icsk_ext_hdr_len =
 428                                                 opt->opt_flen + opt->opt_nflen;
 429                                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 430                                 }
 431                         }
 432                         opt = xchg(&np->opt, opt);
 433                         sk_dst_reset(sk);
 434                 } else {
 435                         write_lock(&sk->sk_dst_lock);
 436                         opt = xchg(&np->opt, opt);
 437                         write_unlock(&sk->sk_dst_lock);
 438                         sk_dst_reset(sk);
 439                 }
 440 sticky_done:
 441                 if (opt)
 442                         sock_kfree_s(sk, opt, opt->tot_len);
 443                 break;
 444         }
 445
 446         case IPV6_2292PKTOPTIONS:
 447         {
 448                 struct ipv6_txoptions *opt = NULL;
 449                 struct msghdr msg;
 450                 struct flowi fl;
 451                 int junk;
 452
 453                 fl.fl6_flowlabel = 0;
 454                 fl.oif = sk->sk_bound_dev_if;
 455
 456                 if (optlen == 0)
 457                         goto update;
 458
 459                 /* 1K is probably excessive
 460                  * 1K is surely not enough, 2K per standard header is 16K.
 461                  */
 462                 retv = -EINVAL;
 463                 if (optlen > 64*1024)
 464                         break;
 465
 466                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 467                 retv = -ENOBUFS;
 468                 if (opt == NULL)
 469                         break;
 470
 471                 memset(opt, 0, sizeof(*opt));
 472                 opt->tot_len = sizeof(*opt) + optlen;
 473                 retv = -EFAULT;
 474                 if (copy_from_user(opt+1, optval, optlen))
 475                         goto done;
 476
 477                 msg.msg_controllen = optlen;
 478                 msg.msg_control = (void*)(opt+1);
 479
 480                 retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
 481                 if (retv)
 482                         goto done;
 483 update:
 484                 retv = 0;
 485                 if (inet_sk(sk)->is_icsk) {
 486                         if (opt) {
 487                                 struct inet_connection_sock *icsk = inet_csk(sk);
 488                                 if (!((1 << sk->sk_state) &
 489                                       (TCPF_LISTEN | TCPF_CLOSE))
 490                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
 491                                         icsk->icsk_ext_hdr_len =
 492                                                 opt->opt_flen + opt->opt_nflen;
 493                                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 494                                 }
 495                         }
 496                         opt = xchg(&np->opt, opt);
 497                         sk_dst_reset(sk);
 498                 } else {
 499                         write_lock(&sk->sk_dst_lock);
 500                         opt = xchg(&np->opt, opt);
 501                         write_unlock(&sk->sk_dst_lock);
 502                         sk_dst_reset(sk);
 503                 }
 504
 505 done:
 506                 if (opt)
 507                         sock_kfree_s(sk, opt, opt->tot_len);
 508                 break;
 509         }
 510         case IPV6_UNICAST_HOPS:
 511                 if (val > 255 || val < -1)
 512                         goto e_inval;
 513                 np->hop_limit = val;
 514                 retv = 0;
 515                 break;
 516
 517         case IPV6_MULTICAST_HOPS:
 518                 if (sk->sk_type == SOCK_STREAM)
 519                         goto e_inval;
 520                 if (val > 255 || val < -1)
 521                         goto e_inval;
 522                 np->mcast_hops = val;
 523                 retv = 0;
 524                 break;
 525
 526         case IPV6_MULTICAST_LOOP:
 527                 np->mc_loop = valbool;
 528                 retv = 0;
 529                 break;
 530
 531         case IPV6_MULTICAST_IF:
 532                 if (sk->sk_type == SOCK_STREAM)
 533                         goto e_inval;
 534                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
 535                         goto e_inval;
 536
 537                 if (__dev_get_by_index(val) == NULL) {
 538                         retv = -ENODEV;
 539                         break;
 540                 }
 541                 np->mcast_oif = val;
 542                 retv = 0;
 543                 break;
 544         case IPV6_ADD_MEMBERSHIP:
 545         case IPV6_DROP_MEMBERSHIP:
 546         {
 547                 struct ipv6_mreq mreq;
 548
 549                 retv = -EFAULT;
 550                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 551                         break;
 552
 553                 if (optname == IPV6_ADD_MEMBERSHIP)
 554                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 555                 else
 556                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 557                 break;
 558         }
 559         case IPV6_JOIN_ANYCAST:
 560         case IPV6_LEAVE_ANYCAST:
 561         {
 562                 struct ipv6_mreq mreq;
 563
 564                 if (optlen != sizeof(struct ipv6_mreq))
 565                         goto e_inval;
 566
 567                 retv = -EFAULT;
 568                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 569                         break;
 570
 571                 if (optname == IPV6_JOIN_ANYCAST)
 572                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 573                 else
 574                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 575                 break;
 576         }
 577         case MCAST_JOIN_GROUP:
 578         case MCAST_LEAVE_GROUP:
 579         {
 580                 struct group_req greq;
 581                 struct sockaddr_in6 *psin6;
 582
 583                 retv = -EFAULT;
 584                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
 585                         break;
 586                 if (greq.gr_group.ss_family != AF_INET6) {
 587                         retv = -EADDRNOTAVAIL;
 588                         break;
 589                 }
 590                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 591                 if (optname == MCAST_JOIN_GROUP)
 592                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
 593                                 &psin6->sin6_addr);
 594                 else
 595                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
 596                                 &psin6->sin6_addr);
 597                 break;
 598         }
 599         case MCAST_JOIN_SOURCE_GROUP:
 600         case MCAST_LEAVE_SOURCE_GROUP:
 601         case MCAST_BLOCK_SOURCE:
 602         case MCAST_UNBLOCK_SOURCE:
 603         {
 604                 struct group_source_req greqs;
 605                 int omode, add;
 606
 607                 if (optlen != sizeof(struct group_source_req))
 608                         goto e_inval;
 609                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 610                         retv = -EFAULT;
 611                         break;
 612                 }
 613                 if (greqs.gsr_group.ss_family != AF_INET6 ||
 614                     greqs.gsr_source.ss_family != AF_INET6) {
 615                         retv = -EADDRNOTAVAIL;
 616                         break;
 617                 }
 618                 if (optname == MCAST_BLOCK_SOURCE) {
 619                         omode = MCAST_EXCLUDE;
 620                         add = 1;
 621                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 622                         omode = MCAST_EXCLUDE;
 623                         add = 0;
 624                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 625                         struct sockaddr_in6 *psin6;
 626
 627                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 628                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
 629                                 &psin6->sin6_addr);
 630                         /* prior join w/ different source is ok */
 631                         if (retv && retv != -EADDRINUSE)
 632                                 break;
 633                         omode = MCAST_INCLUDE;
 634                         add = 1;
 635                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 636                         omode = MCAST_INCLUDE;
 637                         add = 0;
 638                 }
 639                 retv = ip6_mc_source(add, omode, sk, &greqs);
 640                 break;
 641         }
 642         case MCAST_MSFILTER:
 643         {
 644                 extern int sysctl_mld_max_msf;
 645                 struct group_filter *gsf;
 646
 647                 if (optlen < GROUP_FILTER_SIZE(0))
 648                         goto e_inval;
 649                 if (optlen > sysctl_optmem_max) {
 650                         retv = -ENOBUFS;
 651                         break;
 652                 }
 653                 gsf = kmalloc(optlen,GFP_KERNEL);
 654                 if (gsf == 0) {
 655                         retv = -ENOBUFS;
 656                         break;
 657                 }
 658                 retv = -EFAULT;
 659                 if (copy_from_user(gsf, optval, optlen)) {
 660                         kfree(gsf);
 661                         break;
 662                 }
 663                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 664                 if (gsf->gf_numsrc >= 0x1ffffffU ||
 665                     gsf->gf_numsrc > sysctl_mld_max_msf) {
 666                         kfree(gsf);
 667                         retv = -ENOBUFS;
 668                         break;
 669                 }
 670                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 671                         kfree(gsf);
 672                         retv = -EINVAL;
 673                         break;
 674                 }
 675                 retv = ip6_mc_msfilter(sk, gsf);
 676                 kfree(gsf);
 677
 678                 break;
 679         }
 680         case IPV6_ROUTER_ALERT:
 681                 retv = ip6_ra_control(sk, val, NULL);
 682                 break;
 683         case IPV6_MTU_DISCOVER:
 684                 if (val<0 || val>2)
 685                         goto e_inval;
 686                 np->pmtudisc = val;
 687                 retv = 0;
 688                 break;
 689         case IPV6_MTU:
 690                 if (val && val < IPV6_MIN_MTU)
 691                         goto e_inval;
 692                 np->frag_size = val;
 693                 retv = 0;
 694                 break;
 695         case IPV6_RECVERR:
 696                 np->recverr = valbool;
 697                 if (!val)
 698                         skb_queue_purge(&sk->sk_error_queue);
 699                 retv = 0;
 700                 break;
 701         case IPV6_FLOWINFO_SEND:
 702                 np->sndflow = valbool;
 703                 retv = 0;
 704                 break;
 705         case IPV6_FLOWLABEL_MGR:
 706                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
 707                 break;
 708         case IPV6_IPSEC_POLICY:
 709         case IPV6_XFRM_POLICY:
 710                 retv = -EPERM;
 711                 if (!capable(CAP_NET_ADMIN))
 712                         break;
 713                 retv = xfrm_user_policy(sk, optname, optval, optlen);
 714                 break;
 715
 716         }
 717         release_sock(sk);
 718
 719         return retv;
 720
 721 e_inval:
 722         release_sock(sk);
 723         return -EINVAL;
 724 }
 725
 726 int ipv6_setsockopt(struct sock *sk, int level, int optname,
 727                     char __user *optval, int optlen)
 728 {
 729         int err;
 730
 731         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 732                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 733
 734         if (level != SOL_IPV6)
 735                 return -ENOPROTOOPT;
 736
 737         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 738 #ifdef CONFIG_NETFILTER
 739         /* we need to exclude all possible ENOPROTOOPTs except default case */
 740         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 741                         optname != IPV6_XFRM_POLICY) {
 742                 lock_sock(sk);
 743                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
 744                                 optlen);
 745                 release_sock(sk);
 746         }
 747 #endif
 748         return err;
 749 }
 750
 751
 752 #ifdef CONFIG_COMPAT
 753 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
 754                            char __user *optval, int optlen)
 755 {
 756         int err;
 757
 758         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
 759                 if (udp_prot.compat_setsockopt != NULL)
 760                         return udp_prot.compat_setsockopt(sk, level, optname,
 761                                                           optval, optlen);
 762                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 763         }
 764
 765         if (level != SOL_IPV6)
 766                 return -ENOPROTOOPT;
 767
 768         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 769 #ifdef CONFIG_NETFILTER
 770         /* we need to exclude all possible ENOPROTOOPTs except default case */
 771         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 772             optname != IPV6_XFRM_POLICY) {
 773                 lock_sock(sk);
 774                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
 775                                            optval, optlen);
 776                 release_sock(sk);
 777         }
 778 #endif
 779         return err;
 780 }
 781
 782 EXPORT_SYMBOL(compat_ipv6_setsockopt);
 783 #endif
 784
 785 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_opt_hdr *hdr,
 786                                   char __user *optval, int len)
 787 {
 788         if (!hdr)
 789                 return 0;
 790         len = min_t(int, len, ipv6_optlen(hdr));
 791         if (copy_to_user(optval, hdr, ipv6_optlen(hdr)))
 792                 return -EFAULT;
 793         return len;
 794 }
 795
 796 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 797                     char __user *optval, int __user *optlen)
 798 {
 799         struct ipv6_pinfo *np = inet6_sk(sk);
 800         int len;
 801         int val;
 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_TCP)
 809                         return -EINVAL;
 810                 if (sk->sk_state != TCP_ESTABLISHED)
 811                         return -ENOTCONN;
 812                 val = sk->sk_family;
 813                 break;
 814         case MCAST_MSFILTER:
 815         {
 816                 struct group_filter gsf;
 817                 int err;
 818
 819                 if (len < GROUP_FILTER_SIZE(0))
 820                         return -EINVAL;
 821                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
 822                         return -EFAULT;
 823                 lock_sock(sk);
 824                 err = ip6_mc_msfget(sk, &gsf,
 825                         (struct group_filter __user *)optval, optlen);
 826                 release_sock(sk);
 827                 return err;
 828         }
 829
 830         case IPV6_2292PKTOPTIONS:
 831         {
 832                 struct msghdr msg;
 833                 struct sk_buff *skb;
 834
 835                 if (sk->sk_type != SOCK_STREAM)
 836                         return -ENOPROTOOPT;
 837
 838                 msg.msg_control = optval;
 839                 msg.msg_controllen = len;
 840                 msg.msg_flags = 0;
 841
 842                 lock_sock(sk);
 843                 skb = np->pktoptions;
 844                 if (skb)
 845                         atomic_inc(&skb->users);
 846                 release_sock(sk);
 847
 848                 if (skb) {
 849                         int err = datagram_recv_ctl(sk, &msg, skb);
 850                         kfree_skb(skb);
 851                         if (err)
 852                                 return err;
 853                 } else {
 854                         if (np->rxopt.bits.rxinfo) {
 855                                 struct in6_pktinfo src_info;
 856                                 src_info.ipi6_ifindex = np->mcast_oif;
 857                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 858                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 859                         }
 860                         if (np->rxopt.bits.rxhlim) {
 861                                 int hlim = np->mcast_hops;
 862                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 863                         }
 864                         if (np->rxopt.bits.rxoinfo) {
 865                                 struct in6_pktinfo src_info;
 866                                 src_info.ipi6_ifindex = np->mcast_oif;
 867                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 868                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 869                         }
 870                         if (np->rxopt.bits.rxohlim) {
 871                                 int hlim = np->mcast_hops;
 872                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 873                         }
 874                 }
 875                 len -= msg.msg_controllen;
 876                 return put_user(len, optlen);
 877         }
 878         case IPV6_MTU:
 879         {
 880                 struct dst_entry *dst;
 881                 val = 0;
 882                 lock_sock(sk);
 883                 dst = sk_dst_get(sk);
 884                 if (dst) {
 885                         val = dst_mtu(dst);
 886                         dst_release(dst);
 887                 }
 888                 release_sock(sk);
 889                 if (!val)
 890                         return -ENOTCONN;
 891                 break;
 892         }
 893
 894         case IPV6_V6ONLY:
 895                 val = np->ipv6only;
 896                 break;
 897
 898         case IPV6_RECVPKTINFO:
 899                 val = np->rxopt.bits.rxinfo;
 900                 break;
 901
 902         case IPV6_2292PKTINFO:
 903                 val = np->rxopt.bits.rxoinfo;
 904                 break;
 905
 906         case IPV6_RECVHOPLIMIT:
 907                 val = np->rxopt.bits.rxhlim;
 908                 break;
 909
 910         case IPV6_2292HOPLIMIT:
 911                 val = np->rxopt.bits.rxohlim;
 912                 break;
 913
 914         case IPV6_RECVRTHDR:
 915                 val = np->rxopt.bits.srcrt;
 916                 break;
 917
 918         case IPV6_2292RTHDR:
 919                 val = np->rxopt.bits.osrcrt;
 920                 break;
 921
 922         case IPV6_HOPOPTS:
 923         case IPV6_RTHDRDSTOPTS:
 924         case IPV6_RTHDR:
 925         case IPV6_DSTOPTS:
 926         {
 927
 928                 lock_sock(sk);
 929                 len = ipv6_getsockopt_sticky(sk, np->opt->hopopt,
 930                                              optval, len);
 931                 release_sock(sk);
 932                 return put_user(len, optlen);
 933         }
 934
 935         case IPV6_RECVHOPOPTS:
 936                 val = np->rxopt.bits.hopopts;
 937                 break;
 938
 939         case IPV6_2292HOPOPTS:
 940                 val = np->rxopt.bits.ohopopts;
 941                 break;
 942
 943         case IPV6_RECVDSTOPTS:
 944                 val = np->rxopt.bits.dstopts;
 945                 break;
 946
 947         case IPV6_2292DSTOPTS:
 948                 val = np->rxopt.bits.odstopts;
 949                 break;
 950
 951         case IPV6_TCLASS:
 952                 val = np->tclass;
 953                 if (val < 0)
 954                         val = 0;
 955                 break;
 956
 957         case IPV6_RECVTCLASS:
 958                 val = np->rxopt.bits.rxtclass;
 959                 break;
 960
 961         case IPV6_FLOWINFO:
 962                 val = np->rxopt.bits.rxflow;
 963                 break;
 964
 965         case IPV6_UNICAST_HOPS:
 966                 val = np->hop_limit;
 967                 break;
 968
 969         case IPV6_MULTICAST_HOPS:
 970                 val = np->mcast_hops;
 971                 break;
 972
 973         case IPV6_MULTICAST_LOOP:
 974                 val = np->mc_loop;
 975                 break;
 976
 977         case IPV6_MULTICAST_IF:
 978                 val = np->mcast_oif;
 979                 break;
 980
 981         case IPV6_MTU_DISCOVER:
 982                 val = np->pmtudisc;
 983                 break;
 984
 985         case IPV6_RECVERR:
 986                 val = np->recverr;
 987                 break;
 988
 989         case IPV6_FLOWINFO_SEND:
 990                 val = np->sndflow;
 991                 break;
 992
 993         default:
 994                 return -EINVAL;
 995         }
 996         len = min_t(unsigned int, sizeof(int), len);
 997         if(put_user(len, optlen))
 998                 return -EFAULT;
 999         if(copy_to_user(optval,&val,len))
1000                 return -EFAULT;
1001         return 0;
1002 }
1003
1004 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1005                     char __user *optval, int __user *optlen)
1006 {
1007         int err;
1008
1009         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1010                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1011
1012         if(level != SOL_IPV6)
1013                 return -ENOPROTOOPT;
1014
1015         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1016 #ifdef CONFIG_NETFILTER
1017         /* we need to exclude all possible EINVALs except default case */
1018         if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1019                         optname != MCAST_MSFILTER) {
1020                 int len;
1021
1022                 if (get_user(len, optlen))
1023                         return -EFAULT;
1024
1025                 lock_sock(sk);
1026                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1027                                 &len);
1028                 release_sock(sk);
1029                 if (err >= 0)
1030                         err = put_user(len, optlen);
1031         }
1032 #endif
1033         return err;
1034 }
1035
1036 #ifdef CONFIG_COMPAT
1037 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1038                            char __user *optval, int __user *optlen)
1039 {
1040         int err;
1041
1042         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1043                 if (udp_prot.compat_getsockopt != NULL)
1044                         return udp_prot.compat_getsockopt(sk, level, optname,
1045                                                           optval, optlen);
1046                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1047         }
1048
1049         if (level != SOL_IPV6)
1050                 return -ENOPROTOOPT;
1051
1052         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1053 #ifdef CONFIG_NETFILTER
1054         /* we need to exclude all possible EINVALs except default case */
1055         if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1056                         optname != MCAST_MSFILTER) {
1057                 int len;
1058
1059                 if (get_user(len, optlen))
1060                         return -EFAULT;
1061
1062                 lock_sock(sk);
1063                 err = compat_nf_getsockopt(sk, PF_INET6,
1064                                            optname, optval, &len);
1065                 release_sock(sk);
1066                 if (err >= 0)
1067                         err = put_user(len, optlen);
1068         }
1069 #endif
1070         return err;
1071 }
1072
1073 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1074 #endif
1075
1076 void __init ipv6_packet_init(void)
1077 {
1078         dev_add_pack(&ipv6_packet_type);
1079 }
1080
1081 void ipv6_packet_cleanup(void)
1082 {
1083         dev_remove_pack(&ipv6_packet_type);
1084 }