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                         switch (rthdr->type) {
 414                         case IPV6_SRCRT_TYPE_0:
 415 #ifdef CONFIG_IPV6_MIP6
 416                         case IPV6_SRCRT_TYPE_2:
 417 #endif
 418                                 break;
 419                         default:
 420                                 goto sticky_done;
 421                         }
 422
 423                         if ((rthdr->hdrlen & 1) ||
 424                             (rthdr->hdrlen >> 1) != rthdr->segments_left)
 425                                 goto sticky_done;
 426                 }
 427
 428                 retv = 0;
 429                 if (inet_sk(sk)->is_icsk) {
 430                         if (opt) {
 431                                 struct inet_connection_sock *icsk = inet_csk(sk);
 432                                 if (!((1 << sk->sk_state) &
 433                                       (TCPF_LISTEN | TCPF_CLOSE))
 434                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
 435                                         icsk->icsk_ext_hdr_len =
 436                                                 opt->opt_flen + opt->opt_nflen;
 437                                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 438                                 }
 439                         }
 440                         opt = xchg(&np->opt, opt);
 441                         sk_dst_reset(sk);
 442                 } else {
 443                         write_lock(&sk->sk_dst_lock);
 444                         opt = xchg(&np->opt, opt);
 445                         write_unlock(&sk->sk_dst_lock);
 446                         sk_dst_reset(sk);
 447                 }
 448 sticky_done:
 449                 if (opt)
 450                         sock_kfree_s(sk, opt, opt->tot_len);
 451                 break;
 452         }
 453
 454         case IPV6_2292PKTOPTIONS:
 455         {
 456                 struct ipv6_txoptions *opt = NULL;
 457                 struct msghdr msg;
 458                 struct flowi fl;
 459                 int junk;
 460
 461                 fl.fl6_flowlabel = 0;
 462                 fl.oif = sk->sk_bound_dev_if;
 463
 464                 if (optlen == 0)
 465                         goto update;
 466
 467                 /* 1K is probably excessive
 468                  * 1K is surely not enough, 2K per standard header is 16K.
 469                  */
 470                 retv = -EINVAL;
 471                 if (optlen > 64*1024)
 472                         break;
 473
 474                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 475                 retv = -ENOBUFS;
 476                 if (opt == NULL)
 477                         break;
 478
 479                 memset(opt, 0, sizeof(*opt));
 480                 opt->tot_len = sizeof(*opt) + optlen;
 481                 retv = -EFAULT;
 482                 if (copy_from_user(opt+1, optval, optlen))
 483                         goto done;
 484
 485                 msg.msg_controllen = optlen;
 486                 msg.msg_control = (void*)(opt+1);
 487
 488                 retv = datagram_send_ctl(&msg, &fl, opt, &junk, &junk);
 489                 if (retv)
 490                         goto done;
 491 update:
 492                 retv = 0;
 493                 if (inet_sk(sk)->is_icsk) {
 494                         if (opt) {
 495                                 struct inet_connection_sock *icsk = inet_csk(sk);
 496                                 if (!((1 << sk->sk_state) &
 497                                       (TCPF_LISTEN | TCPF_CLOSE))
 498                                     && inet_sk(sk)->daddr != LOOPBACK4_IPV6) {
 499                                         icsk->icsk_ext_hdr_len =
 500                                                 opt->opt_flen + opt->opt_nflen;
 501                                         icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 502                                 }
 503                         }
 504                         opt = xchg(&np->opt, opt);
 505                         sk_dst_reset(sk);
 506                 } else {
 507                         write_lock(&sk->sk_dst_lock);
 508                         opt = xchg(&np->opt, opt);
 509                         write_unlock(&sk->sk_dst_lock);
 510                         sk_dst_reset(sk);
 511                 }
 512
 513 done:
 514                 if (opt)
 515                         sock_kfree_s(sk, opt, opt->tot_len);
 516                 break;
 517         }
 518         case IPV6_UNICAST_HOPS:
 519                 if (val > 255 || val < -1)
 520                         goto e_inval;
 521                 np->hop_limit = val;
 522                 retv = 0;
 523                 break;
 524
 525         case IPV6_MULTICAST_HOPS:
 526                 if (sk->sk_type == SOCK_STREAM)
 527                         goto e_inval;
 528                 if (val > 255 || val < -1)
 529                         goto e_inval;
 530                 np->mcast_hops = val;
 531                 retv = 0;
 532                 break;
 533
 534         case IPV6_MULTICAST_LOOP:
 535                 np->mc_loop = valbool;
 536                 retv = 0;
 537                 break;
 538
 539         case IPV6_MULTICAST_IF:
 540                 if (sk->sk_type == SOCK_STREAM)
 541                         goto e_inval;
 542                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != val)
 543                         goto e_inval;
 544
 545                 if (__dev_get_by_index(val) == NULL) {
 546                         retv = -ENODEV;
 547                         break;
 548                 }
 549                 np->mcast_oif = val;
 550                 retv = 0;
 551                 break;
 552         case IPV6_ADD_MEMBERSHIP:
 553         case IPV6_DROP_MEMBERSHIP:
 554         {
 555                 struct ipv6_mreq mreq;
 556
 557                 retv = -EFAULT;
 558                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 559                         break;
 560
 561                 if (optname == IPV6_ADD_MEMBERSHIP)
 562                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 563                 else
 564                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 565                 break;
 566         }
 567         case IPV6_JOIN_ANYCAST:
 568         case IPV6_LEAVE_ANYCAST:
 569         {
 570                 struct ipv6_mreq mreq;
 571
 572                 if (optlen != sizeof(struct ipv6_mreq))
 573                         goto e_inval;
 574
 575                 retv = -EFAULT;
 576                 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
 577                         break;
 578
 579                 if (optname == IPV6_JOIN_ANYCAST)
 580                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 581                 else
 582                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 583                 break;
 584         }
 585         case MCAST_JOIN_GROUP:
 586         case MCAST_LEAVE_GROUP:
 587         {
 588                 struct group_req greq;
 589                 struct sockaddr_in6 *psin6;
 590
 591                 retv = -EFAULT;
 592                 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
 593                         break;
 594                 if (greq.gr_group.ss_family != AF_INET6) {
 595                         retv = -EADDRNOTAVAIL;
 596                         break;
 597                 }
 598                 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 599                 if (optname == MCAST_JOIN_GROUP)
 600                         retv = ipv6_sock_mc_join(sk, greq.gr_interface,
 601                                 &psin6->sin6_addr);
 602                 else
 603                         retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
 604                                 &psin6->sin6_addr);
 605                 break;
 606         }
 607         case MCAST_JOIN_SOURCE_GROUP:
 608         case MCAST_LEAVE_SOURCE_GROUP:
 609         case MCAST_BLOCK_SOURCE:
 610         case MCAST_UNBLOCK_SOURCE:
 611         {
 612                 struct group_source_req greqs;
 613                 int omode, add;
 614
 615                 if (optlen != sizeof(struct group_source_req))
 616                         goto e_inval;
 617                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 618                         retv = -EFAULT;
 619                         break;
 620                 }
 621                 if (greqs.gsr_group.ss_family != AF_INET6 ||
 622                     greqs.gsr_source.ss_family != AF_INET6) {
 623                         retv = -EADDRNOTAVAIL;
 624                         break;
 625                 }
 626                 if (optname == MCAST_BLOCK_SOURCE) {
 627                         omode = MCAST_EXCLUDE;
 628                         add = 1;
 629                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 630                         omode = MCAST_EXCLUDE;
 631                         add = 0;
 632                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 633                         struct sockaddr_in6 *psin6;
 634
 635                         psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 636                         retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
 637                                 &psin6->sin6_addr);
 638                         /* prior join w/ different source is ok */
 639                         if (retv && retv != -EADDRINUSE)
 640                                 break;
 641                         omode = MCAST_INCLUDE;
 642                         add = 1;
 643                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 644                         omode = MCAST_INCLUDE;
 645                         add = 0;
 646                 }
 647                 retv = ip6_mc_source(add, omode, sk, &greqs);
 648                 break;
 649         }
 650         case MCAST_MSFILTER:
 651         {
 652                 extern int sysctl_mld_max_msf;
 653                 struct group_filter *gsf;
 654
 655                 if (optlen < GROUP_FILTER_SIZE(0))
 656                         goto e_inval;
 657                 if (optlen > sysctl_optmem_max) {
 658                         retv = -ENOBUFS;
 659                         break;
 660                 }
 661                 gsf = kmalloc(optlen,GFP_KERNEL);
 662                 if (gsf == 0) {
 663                         retv = -ENOBUFS;
 664                         break;
 665                 }
 666                 retv = -EFAULT;
 667                 if (copy_from_user(gsf, optval, optlen)) {
 668                         kfree(gsf);
 669                         break;
 670                 }
 671                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 672                 if (gsf->gf_numsrc >= 0x1ffffffU ||
 673                     gsf->gf_numsrc > sysctl_mld_max_msf) {
 674                         kfree(gsf);
 675                         retv = -ENOBUFS;
 676                         break;
 677                 }
 678                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 679                         kfree(gsf);
 680                         retv = -EINVAL;
 681                         break;
 682                 }
 683                 retv = ip6_mc_msfilter(sk, gsf);
 684                 kfree(gsf);
 685
 686                 break;
 687         }
 688         case IPV6_ROUTER_ALERT:
 689                 retv = ip6_ra_control(sk, val, NULL);
 690                 break;
 691         case IPV6_MTU_DISCOVER:
 692                 if (val<0 || val>2)
 693                         goto e_inval;
 694                 np->pmtudisc = val;
 695                 retv = 0;
 696                 break;
 697         case IPV6_MTU:
 698                 if (val && val < IPV6_MIN_MTU)
 699                         goto e_inval;
 700                 np->frag_size = val;
 701                 retv = 0;
 702                 break;
 703         case IPV6_RECVERR:
 704                 np->recverr = valbool;
 705                 if (!val)
 706                         skb_queue_purge(&sk->sk_error_queue);
 707                 retv = 0;
 708                 break;
 709         case IPV6_FLOWINFO_SEND:
 710                 np->sndflow = valbool;
 711                 retv = 0;
 712                 break;
 713         case IPV6_FLOWLABEL_MGR:
 714                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
 715                 break;
 716         case IPV6_IPSEC_POLICY:
 717         case IPV6_XFRM_POLICY:
 718                 retv = -EPERM;
 719                 if (!capable(CAP_NET_ADMIN))
 720                         break;
 721                 retv = xfrm_user_policy(sk, optname, optval, optlen);
 722                 break;
 723
 724         }
 725         release_sock(sk);
 726
 727         return retv;
 728
 729 e_inval:
 730         release_sock(sk);
 731         return -EINVAL;
 732 }
 733
 734 int ipv6_setsockopt(struct sock *sk, int level, int optname,
 735                     char __user *optval, int optlen)
 736 {
 737         int err;
 738
 739         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
 740                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 741
 742         if (level != SOL_IPV6)
 743                 return -ENOPROTOOPT;
 744
 745         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 746 #ifdef CONFIG_NETFILTER
 747         /* we need to exclude all possible ENOPROTOOPTs except default case */
 748         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 749                         optname != IPV6_XFRM_POLICY) {
 750                 lock_sock(sk);
 751                 err = nf_setsockopt(sk, PF_INET6, optname, optval,
 752                                 optlen);
 753                 release_sock(sk);
 754         }
 755 #endif
 756         return err;
 757 }
 758
 759
 760 #ifdef CONFIG_COMPAT
 761 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
 762                            char __user *optval, int optlen)
 763 {
 764         int err;
 765
 766         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
 767                 if (udp_prot.compat_setsockopt != NULL)
 768                         return udp_prot.compat_setsockopt(sk, level, optname,
 769                                                           optval, optlen);
 770                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
 771         }
 772
 773         if (level != SOL_IPV6)
 774                 return -ENOPROTOOPT;
 775
 776         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
 777 #ifdef CONFIG_NETFILTER
 778         /* we need to exclude all possible ENOPROTOOPTs except default case */
 779         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
 780             optname != IPV6_XFRM_POLICY) {
 781                 lock_sock(sk);
 782                 err = compat_nf_setsockopt(sk, PF_INET6, optname,
 783                                            optval, optlen);
 784                 release_sock(sk);
 785         }
 786 #endif
 787         return err;
 788 }
 789
 790 EXPORT_SYMBOL(compat_ipv6_setsockopt);
 791 #endif
 792
 793 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_opt_hdr *hdr,
 794                                   char __user *optval, int len)
 795 {
 796         if (!hdr)
 797                 return 0;
 798         len = min_t(int, len, ipv6_optlen(hdr));
 799         if (copy_to_user(optval, hdr, ipv6_optlen(hdr)))
 800                 return -EFAULT;
 801         return len;
 802 }
 803
 804 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
 805                     char __user *optval, int __user *optlen)
 806 {
 807         struct ipv6_pinfo *np = inet6_sk(sk);
 808         int len;
 809         int val;
 810
 811         if (get_user(len, optlen))
 812                 return -EFAULT;
 813         switch (optname) {
 814         case IPV6_ADDRFORM:
 815                 if (sk->sk_protocol != IPPROTO_UDP &&
 816                     sk->sk_protocol != IPPROTO_TCP)
 817                         return -EINVAL;
 818                 if (sk->sk_state != TCP_ESTABLISHED)
 819                         return -ENOTCONN;
 820                 val = sk->sk_family;
 821                 break;
 822         case MCAST_MSFILTER:
 823         {
 824                 struct group_filter gsf;
 825                 int err;
 826
 827                 if (len < GROUP_FILTER_SIZE(0))
 828                         return -EINVAL;
 829                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
 830                         return -EFAULT;
 831                 lock_sock(sk);
 832                 err = ip6_mc_msfget(sk, &gsf,
 833                         (struct group_filter __user *)optval, optlen);
 834                 release_sock(sk);
 835                 return err;
 836         }
 837
 838         case IPV6_2292PKTOPTIONS:
 839         {
 840                 struct msghdr msg;
 841                 struct sk_buff *skb;
 842
 843                 if (sk->sk_type != SOCK_STREAM)
 844                         return -ENOPROTOOPT;
 845
 846                 msg.msg_control = optval;
 847                 msg.msg_controllen = len;
 848                 msg.msg_flags = 0;
 849
 850                 lock_sock(sk);
 851                 skb = np->pktoptions;
 852                 if (skb)
 853                         atomic_inc(&skb->users);
 854                 release_sock(sk);
 855
 856                 if (skb) {
 857                         int err = datagram_recv_ctl(sk, &msg, skb);
 858                         kfree_skb(skb);
 859                         if (err)
 860                                 return err;
 861                 } else {
 862                         if (np->rxopt.bits.rxinfo) {
 863                                 struct in6_pktinfo src_info;
 864                                 src_info.ipi6_ifindex = np->mcast_oif;
 865                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 866                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
 867                         }
 868                         if (np->rxopt.bits.rxhlim) {
 869                                 int hlim = np->mcast_hops;
 870                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
 871                         }
 872                         if (np->rxopt.bits.rxoinfo) {
 873                                 struct in6_pktinfo src_info;
 874                                 src_info.ipi6_ifindex = np->mcast_oif;
 875                                 ipv6_addr_copy(&src_info.ipi6_addr, &np->daddr);
 876                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
 877                         }
 878                         if (np->rxopt.bits.rxohlim) {
 879                                 int hlim = np->mcast_hops;
 880                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
 881                         }
 882                 }
 883                 len -= msg.msg_controllen;
 884                 return put_user(len, optlen);
 885         }
 886         case IPV6_MTU:
 887         {
 888                 struct dst_entry *dst;
 889                 val = 0;
 890                 lock_sock(sk);
 891                 dst = sk_dst_get(sk);
 892                 if (dst) {
 893                         val = dst_mtu(dst);
 894                         dst_release(dst);
 895                 }
 896                 release_sock(sk);
 897                 if (!val)
 898                         return -ENOTCONN;
 899                 break;
 900         }
 901
 902         case IPV6_V6ONLY:
 903                 val = np->ipv6only;
 904                 break;
 905
 906         case IPV6_RECVPKTINFO:
 907                 val = np->rxopt.bits.rxinfo;
 908                 break;
 909
 910         case IPV6_2292PKTINFO:
 911                 val = np->rxopt.bits.rxoinfo;
 912                 break;
 913
 914         case IPV6_RECVHOPLIMIT:
 915                 val = np->rxopt.bits.rxhlim;
 916                 break;
 917
 918         case IPV6_2292HOPLIMIT:
 919                 val = np->rxopt.bits.rxohlim;
 920                 break;
 921
 922         case IPV6_RECVRTHDR:
 923                 val = np->rxopt.bits.srcrt;
 924                 break;
 925
 926         case IPV6_2292RTHDR:
 927                 val = np->rxopt.bits.osrcrt;
 928                 break;
 929
 930         case IPV6_HOPOPTS:
 931         case IPV6_RTHDRDSTOPTS:
 932         case IPV6_RTHDR:
 933         case IPV6_DSTOPTS:
 934         {
 935
 936                 lock_sock(sk);
 937                 len = ipv6_getsockopt_sticky(sk, np->opt->hopopt,
 938                                              optval, len);
 939                 release_sock(sk);
 940                 return put_user(len, optlen);
 941         }
 942
 943         case IPV6_RECVHOPOPTS:
 944                 val = np->rxopt.bits.hopopts;
 945                 break;
 946
 947         case IPV6_2292HOPOPTS:
 948                 val = np->rxopt.bits.ohopopts;
 949                 break;
 950
 951         case IPV6_RECVDSTOPTS:
 952                 val = np->rxopt.bits.dstopts;
 953                 break;
 954
 955         case IPV6_2292DSTOPTS:
 956                 val = np->rxopt.bits.odstopts;
 957                 break;
 958
 959         case IPV6_TCLASS:
 960                 val = np->tclass;
 961                 if (val < 0)
 962                         val = 0;
 963                 break;
 964
 965         case IPV6_RECVTCLASS:
 966                 val = np->rxopt.bits.rxtclass;
 967                 break;
 968
 969         case IPV6_FLOWINFO:
 970                 val = np->rxopt.bits.rxflow;
 971                 break;
 972
 973         case IPV6_UNICAST_HOPS:
 974                 val = np->hop_limit;
 975                 break;
 976
 977         case IPV6_MULTICAST_HOPS:
 978                 val = np->mcast_hops;
 979                 break;
 980
 981         case IPV6_MULTICAST_LOOP:
 982                 val = np->mc_loop;
 983                 break;
 984
 985         case IPV6_MULTICAST_IF:
 986                 val = np->mcast_oif;
 987                 break;
 988
 989         case IPV6_MTU_DISCOVER:
 990                 val = np->pmtudisc;
 991                 break;
 992
 993         case IPV6_RECVERR:
 994                 val = np->recverr;
 995                 break;
 996
 997         case IPV6_FLOWINFO_SEND:
 998                 val = np->sndflow;
 999                 break;
1000
1001         default:
1002                 return -EINVAL;
1003         }
1004         len = min_t(unsigned int, sizeof(int), len);
1005         if(put_user(len, optlen))
1006                 return -EFAULT;
1007         if(copy_to_user(optval,&val,len))
1008                 return -EFAULT;
1009         return 0;
1010 }
1011
1012 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1013                     char __user *optval, int __user *optlen)
1014 {
1015         int err;
1016
1017         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1018                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1019
1020         if(level != SOL_IPV6)
1021                 return -ENOPROTOOPT;
1022
1023         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1024 #ifdef CONFIG_NETFILTER
1025         /* we need to exclude all possible EINVALs except default case */
1026         if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1027                         optname != MCAST_MSFILTER) {
1028                 int len;
1029
1030                 if (get_user(len, optlen))
1031                         return -EFAULT;
1032
1033                 lock_sock(sk);
1034                 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1035                                 &len);
1036                 release_sock(sk);
1037                 if (err >= 0)
1038                         err = put_user(len, optlen);
1039         }
1040 #endif
1041         return err;
1042 }
1043
1044 #ifdef CONFIG_COMPAT
1045 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1046                            char __user *optval, int __user *optlen)
1047 {
1048         int err;
1049
1050         if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1051                 if (udp_prot.compat_getsockopt != NULL)
1052                         return udp_prot.compat_getsockopt(sk, level, optname,
1053                                                           optval, optlen);
1054                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1055         }
1056
1057         if (level != SOL_IPV6)
1058                 return -ENOPROTOOPT;
1059
1060         err = do_ipv6_getsockopt(sk, level, optname, optval, optlen);
1061 #ifdef CONFIG_NETFILTER
1062         /* we need to exclude all possible EINVALs except default case */
1063         if (err == -EINVAL && optname != IPV6_ADDRFORM &&
1064                         optname != MCAST_MSFILTER) {
1065                 int len;
1066
1067                 if (get_user(len, optlen))
1068                         return -EFAULT;
1069
1070                 lock_sock(sk);
1071                 err = compat_nf_getsockopt(sk, PF_INET6,
1072                                            optname, optval, &len);
1073                 release_sock(sk);
1074                 if (err >= 0)
1075                         err = put_user(len, optlen);
1076         }
1077 #endif
1078         return err;
1079 }
1080
1081 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1082 #endif
1083
1084 void __init ipv6_packet_init(void)
1085 {
1086         dev_add_pack(&ipv6_packet_type);
1087 }
1088
1089 void ipv6_packet_cleanup(void)
1090 {
1091         dev_remove_pack(&ipv6_packet_type);
1092 }