git.oblomov.eu Git - linux-2.6/blob - net/ipv6/af_inet6.c

   1 /*
   2  *      PF_INET6 socket protocol family
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Adapted from linux/net/ipv4/af_inet.c
   9  *
  10  *      Fixes:
  11  *      piggy, Karl Knutson     :       Socket protocol table
  12  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
  13  *      Arnaldo Melo            :       check proc_net_create return, cleanups
  14  *
  15  *      This program is free software; you can redistribute it and/or
  16  *      modify it under the terms of the GNU General Public License
  17  *      as published by the Free Software Foundation; either version
  18  *      2 of the License, or (at your option) any later version.
  19  */
  20
  21
  22 #include <linux/module.h>
  23 #include <linux/capability.h>
  24 #include <linux/errno.h>
  25 #include <linux/types.h>
  26 #include <linux/socket.h>
  27 #include <linux/in.h>
  28 #include <linux/kernel.h>
  29 #include <linux/timer.h>
  30 #include <linux/string.h>
  31 #include <linux/sockios.h>
  32 #include <linux/net.h>
  33 #include <linux/fcntl.h>
  34 #include <linux/mm.h>
  35 #include <linux/interrupt.h>
  36 #include <linux/proc_fs.h>
  37 #include <linux/stat.h>
  38 #include <linux/init.h>
  39
  40 #include <linux/inet.h>
  41 #include <linux/netdevice.h>
  42 #include <linux/icmpv6.h>
  43 #include <linux/netfilter_ipv6.h>
  44
  45 #include <net/ip.h>
  46 #include <net/ipv6.h>
  47 #include <net/udp.h>
  48 #include <net/udplite.h>
  49 #include <net/tcp.h>
  50 #include <net/ipip.h>
  51 #include <net/protocol.h>
  52 #include <net/inet_common.h>
  53 #include <net/route.h>
  54 #include <net/transp_v6.h>
  55 #include <net/ip6_route.h>
  56 #include <net/addrconf.h>
  57 #ifdef CONFIG_IPV6_TUNNEL
  58 #include <net/ip6_tunnel.h>
  59 #endif
  60
  61 #include <asm/uaccess.h>
  62 #include <asm/system.h>
  63 #include <linux/mroute6.h>
  64
  65 MODULE_AUTHOR("Cast of dozens");
  66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  67 MODULE_LICENSE("GPL");
  68
  69 /* The inetsw6 table contains everything that inet6_create needs to
  70  * build a new socket.
  71  */
  72 static struct list_head inetsw6[SOCK_MAX];
  73 static DEFINE_SPINLOCK(inetsw6_lock);
  74
  75 static int disable_ipv6 = 0;
  76 module_param_named(disable, disable_ipv6, int, 0);
  77 MODULE_PARM_DESC(disable, "Disable IPv6 such that it is non-functional");
  78
  79 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  80 {
  81         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  82
  83         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  84 }
  85
  86 static int inet6_create(struct net *net, struct socket *sock, int protocol)
  87 {
  88         struct inet_sock *inet;
  89         struct ipv6_pinfo *np;
  90         struct sock *sk;
  91         struct inet_protosw *answer;
  92         struct proto *answer_prot;
  93         unsigned char answer_flags;
  94         char answer_no_check;
  95         int try_loading_module = 0;
  96         int err;
  97
  98         if (sock->type != SOCK_RAW &&
  99             sock->type != SOCK_DGRAM &&
 100             !inet_ehash_secret)
 101                 build_ehash_secret();
 102
 103         /* Look for the requested type/protocol pair. */
 104 lookup_protocol:
 105         err = -ESOCKTNOSUPPORT;
 106         rcu_read_lock();
 107         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 108
 109                 err = 0;
 110                 /* Check the non-wild match. */
 111                 if (protocol == answer->protocol) {
 112                         if (protocol != IPPROTO_IP)
 113                                 break;
 114                 } else {
 115                         /* Check for the two wild cases. */
 116                         if (IPPROTO_IP == protocol) {
 117                                 protocol = answer->protocol;
 118                                 break;
 119                         }
 120                         if (IPPROTO_IP == answer->protocol)
 121                                 break;
 122                 }
 123                 err = -EPROTONOSUPPORT;
 124         }
 125
 126         if (err) {
 127                 if (try_loading_module < 2) {
 128                         rcu_read_unlock();
 129                         /*
 130                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 131                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 132                          */
 133                         if (++try_loading_module == 1)
 134                                 request_module("net-pf-%d-proto-%d-type-%d",
 135                                                 PF_INET6, protocol, sock->type);
 136                         /*
 137                          * Fall back to generic, e.g. net-pf-10-proto-132
 138                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 139                          */
 140                         else
 141                                 request_module("net-pf-%d-proto-%d",
 142                                                 PF_INET6, protocol);
 143                         goto lookup_protocol;
 144                 } else
 145                         goto out_rcu_unlock;
 146         }
 147
 148         err = -EPERM;
 149         if (answer->capability > 0 && !capable(answer->capability))
 150                 goto out_rcu_unlock;
 151
 152         sock->ops = answer->ops;
 153         answer_prot = answer->prot;
 154         answer_no_check = answer->no_check;
 155         answer_flags = answer->flags;
 156         rcu_read_unlock();
 157
 158         WARN_ON(answer_prot->slab == NULL);
 159
 160         err = -ENOBUFS;
 161         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 162         if (sk == NULL)
 163                 goto out;
 164
 165         sock_init_data(sock, sk);
 166
 167         err = 0;
 168         sk->sk_no_check = answer_no_check;
 169         if (INET_PROTOSW_REUSE & answer_flags)
 170                 sk->sk_reuse = 1;
 171
 172         inet = inet_sk(sk);
 173         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 174
 175         if (SOCK_RAW == sock->type) {
 176                 inet->num = protocol;
 177                 if (IPPROTO_RAW == protocol)
 178                         inet->hdrincl = 1;
 179         }
 180
 181         sk->sk_destruct         = inet_sock_destruct;
 182         sk->sk_family           = PF_INET6;
 183         sk->sk_protocol         = protocol;
 184
 185         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 186
 187         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 188         np->hop_limit   = -1;
 189         np->mcast_hops  = -1;
 190         np->mc_loop     = 1;
 191         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 192         np->ipv6only    = net->ipv6.sysctl.bindv6only;
 193
 194         /* Init the ipv4 part of the socket since we can have sockets
 195          * using v6 API for ipv4.
 196          */
 197         inet->uc_ttl    = -1;
 198
 199         inet->mc_loop   = 1;
 200         inet->mc_ttl    = 1;
 201         inet->mc_index  = 0;
 202         inet->mc_list   = NULL;
 203
 204         if (ipv4_config.no_pmtu_disc)
 205                 inet->pmtudisc = IP_PMTUDISC_DONT;
 206         else
 207                 inet->pmtudisc = IP_PMTUDISC_WANT;
 208         /*
 209          * Increment only the relevant sk_prot->socks debug field, this changes
 210          * the previous behaviour of incrementing both the equivalent to
 211          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 212          *
 213          * This allows better debug granularity as we'll know exactly how many
 214          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 215          * transport protocol socks. -acme
 216          */
 217         sk_refcnt_debug_inc(sk);
 218
 219         if (inet->num) {
 220                 /* It assumes that any protocol which allows
 221                  * the user to assign a number at socket
 222                  * creation time automatically shares.
 223                  */
 224                 inet->sport = htons(inet->num);
 225                 sk->sk_prot->hash(sk);
 226         }
 227         if (sk->sk_prot->init) {
 228                 err = sk->sk_prot->init(sk);
 229                 if (err) {
 230                         sk_common_release(sk);
 231                         goto out;
 232                 }
 233         }
 234 out:
 235         return err;
 236 out_rcu_unlock:
 237         rcu_read_unlock();
 238         goto out;
 239 }
 240
 241
 242 /* bind for INET6 API */
 243 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 244 {
 245         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 246         struct sock *sk = sock->sk;
 247         struct inet_sock *inet = inet_sk(sk);
 248         struct ipv6_pinfo *np = inet6_sk(sk);
 249         struct net *net = sock_net(sk);
 250         __be32 v4addr = 0;
 251         unsigned short snum;
 252         int addr_type = 0;
 253         int err = 0;
 254
 255         /* If the socket has its own bind function then use it. */
 256         if (sk->sk_prot->bind)
 257                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 258
 259         if (addr_len < SIN6_LEN_RFC2133)
 260                 return -EINVAL;
 261         addr_type = ipv6_addr_type(&addr->sin6_addr);
 262         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 263                 return -EINVAL;
 264
 265         snum = ntohs(addr->sin6_port);
 266         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 267                 return -EACCES;
 268
 269         lock_sock(sk);
 270
 271         /* Check these errors (active socket, double bind). */
 272         if (sk->sk_state != TCP_CLOSE || inet->num) {
 273                 err = -EINVAL;
 274                 goto out;
 275         }
 276
 277         /* Check if the address belongs to the host. */
 278         if (addr_type == IPV6_ADDR_MAPPED) {
 279                 v4addr = addr->sin6_addr.s6_addr32[3];
 280                 if (inet_addr_type(net, v4addr) != RTN_LOCAL) {
 281                         err = -EADDRNOTAVAIL;
 282                         goto out;
 283                 }
 284         } else {
 285                 if (addr_type != IPV6_ADDR_ANY) {
 286                         struct net_device *dev = NULL;
 287
 288                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 289                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 290                                     addr->sin6_scope_id) {
 291                                         /* Override any existing binding, if another one
 292                                          * is supplied by user.
 293                                          */
 294                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 295                                 }
 296
 297                                 /* Binding to link-local address requires an interface */
 298                                 if (!sk->sk_bound_dev_if) {
 299                                         err = -EINVAL;
 300                                         goto out;
 301                                 }
 302                                 dev = dev_get_by_index(net, sk->sk_bound_dev_if);
 303                                 if (!dev) {
 304                                         err = -ENODEV;
 305                                         goto out;
 306                                 }
 307                         }
 308
 309                         /* ipv4 addr of the socket is invalid.  Only the
 310                          * unspecified and mapped address have a v4 equivalent.
 311                          */
 312                         v4addr = LOOPBACK4_IPV6;
 313                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 314                                 if (!ipv6_chk_addr(net, &addr->sin6_addr,
 315                                                    dev, 0)) {
 316                                         if (dev)
 317                                                 dev_put(dev);
 318                                         err = -EADDRNOTAVAIL;
 319                                         goto out;
 320                                 }
 321                         }
 322                         if (dev)
 323                                 dev_put(dev);
 324                 }
 325         }
 326
 327         inet->rcv_saddr = v4addr;
 328         inet->saddr = v4addr;
 329
 330         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 331
 332         if (!(addr_type & IPV6_ADDR_MULTICAST))
 333                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 334
 335         /* Make sure we are allowed to bind here. */
 336         if (sk->sk_prot->get_port(sk, snum)) {
 337                 inet_reset_saddr(sk);
 338                 err = -EADDRINUSE;
 339                 goto out;
 340         }
 341
 342         if (addr_type != IPV6_ADDR_ANY)
 343                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 344         if (snum)
 345                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 346         inet->sport = htons(inet->num);
 347         inet->dport = 0;
 348         inet->daddr = 0;
 349 out:
 350         release_sock(sk);
 351         return err;
 352 }
 353
 354 EXPORT_SYMBOL(inet6_bind);
 355
 356 int inet6_release(struct socket *sock)
 357 {
 358         struct sock *sk = sock->sk;
 359
 360         if (sk == NULL)
 361                 return -EINVAL;
 362
 363         /* Free mc lists */
 364         ipv6_sock_mc_close(sk);
 365
 366         /* Free ac lists */
 367         ipv6_sock_ac_close(sk);
 368
 369         return inet_release(sock);
 370 }
 371
 372 EXPORT_SYMBOL(inet6_release);
 373
 374 void inet6_destroy_sock(struct sock *sk)
 375 {
 376         struct ipv6_pinfo *np = inet6_sk(sk);
 377         struct sk_buff *skb;
 378         struct ipv6_txoptions *opt;
 379
 380         /* Release rx options */
 381
 382         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 383                 kfree_skb(skb);
 384
 385         /* Free flowlabels */
 386         fl6_free_socklist(sk);
 387
 388         /* Free tx options */
 389
 390         if ((opt = xchg(&np->opt, NULL)) != NULL)
 391                 sock_kfree_s(sk, opt, opt->tot_len);
 392 }
 393
 394 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 395
 396 /*
 397  *      This does both peername and sockname.
 398  */
 399
 400 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 401                  int *uaddr_len, int peer)
 402 {
 403         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 404         struct sock *sk = sock->sk;
 405         struct inet_sock *inet = inet_sk(sk);
 406         struct ipv6_pinfo *np = inet6_sk(sk);
 407
 408         sin->sin6_family = AF_INET6;
 409         sin->sin6_flowinfo = 0;
 410         sin->sin6_scope_id = 0;
 411         if (peer) {
 412                 if (!inet->dport)
 413                         return -ENOTCONN;
 414                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 415                     peer == 1)
 416                         return -ENOTCONN;
 417                 sin->sin6_port = inet->dport;
 418                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 419                 if (np->sndflow)
 420                         sin->sin6_flowinfo = np->flow_label;
 421         } else {
 422                 if (ipv6_addr_any(&np->rcv_saddr))
 423                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 424                 else
 425                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 426
 427                 sin->sin6_port = inet->sport;
 428         }
 429         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 430                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 431         *uaddr_len = sizeof(*sin);
 432         return(0);
 433 }
 434
 435 EXPORT_SYMBOL(inet6_getname);
 436
 437 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 438 {
 439         struct sock *sk = sock->sk;
 440         struct net *net = sock_net(sk);
 441
 442         switch(cmd)
 443         {
 444         case SIOCGSTAMP:
 445                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 446
 447         case SIOCGSTAMPNS:
 448                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 449
 450         case SIOCADDRT:
 451         case SIOCDELRT:
 452
 453                 return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
 454
 455         case SIOCSIFADDR:
 456                 return addrconf_add_ifaddr(net, (void __user *) arg);
 457         case SIOCDIFADDR:
 458                 return addrconf_del_ifaddr(net, (void __user *) arg);
 459         case SIOCSIFDSTADDR:
 460                 return addrconf_set_dstaddr(net, (void __user *) arg);
 461         default:
 462                 if (!sk->sk_prot->ioctl)
 463                         return -ENOIOCTLCMD;
 464                 return sk->sk_prot->ioctl(sk, cmd, arg);
 465         }
 466         /*NOTREACHED*/
 467         return(0);
 468 }
 469
 470 EXPORT_SYMBOL(inet6_ioctl);
 471
 472 const struct proto_ops inet6_stream_ops = {
 473         .family            = PF_INET6,
 474         .owner             = THIS_MODULE,
 475         .release           = inet6_release,
 476         .bind              = inet6_bind,
 477         .connect           = inet_stream_connect,       /* ok           */
 478         .socketpair        = sock_no_socketpair,        /* a do nothing */
 479         .accept            = inet_accept,               /* ok           */
 480         .getname           = inet6_getname,
 481         .poll              = tcp_poll,                  /* ok           */
 482         .ioctl             = inet6_ioctl,               /* must change  */
 483         .listen            = inet_listen,               /* ok           */
 484         .shutdown          = inet_shutdown,             /* ok           */
 485         .setsockopt        = sock_common_setsockopt,    /* ok           */
 486         .getsockopt        = sock_common_getsockopt,    /* ok           */
 487         .sendmsg           = tcp_sendmsg,               /* ok           */
 488         .recvmsg           = sock_common_recvmsg,       /* ok           */
 489         .mmap              = sock_no_mmap,
 490         .sendpage          = tcp_sendpage,
 491         .splice_read       = tcp_splice_read,
 492 #ifdef CONFIG_COMPAT
 493         .compat_setsockopt = compat_sock_common_setsockopt,
 494         .compat_getsockopt = compat_sock_common_getsockopt,
 495 #endif
 496 };
 497
 498 const struct proto_ops inet6_dgram_ops = {
 499         .family            = PF_INET6,
 500         .owner             = THIS_MODULE,
 501         .release           = inet6_release,
 502         .bind              = inet6_bind,
 503         .connect           = inet_dgram_connect,        /* ok           */
 504         .socketpair        = sock_no_socketpair,        /* a do nothing */
 505         .accept            = sock_no_accept,            /* a do nothing */
 506         .getname           = inet6_getname,
 507         .poll              = udp_poll,                  /* ok           */
 508         .ioctl             = inet6_ioctl,               /* must change  */
 509         .listen            = sock_no_listen,            /* ok           */
 510         .shutdown          = inet_shutdown,             /* ok           */
 511         .setsockopt        = sock_common_setsockopt,    /* ok           */
 512         .getsockopt        = sock_common_getsockopt,    /* ok           */
 513         .sendmsg           = inet_sendmsg,              /* ok           */
 514         .recvmsg           = sock_common_recvmsg,       /* ok           */
 515         .mmap              = sock_no_mmap,
 516         .sendpage          = sock_no_sendpage,
 517 #ifdef CONFIG_COMPAT
 518         .compat_setsockopt = compat_sock_common_setsockopt,
 519         .compat_getsockopt = compat_sock_common_getsockopt,
 520 #endif
 521 };
 522
 523 static struct net_proto_family inet6_family_ops = {
 524         .family = PF_INET6,
 525         .create = inet6_create,
 526         .owner  = THIS_MODULE,
 527 };
 528
 529 int inet6_register_protosw(struct inet_protosw *p)
 530 {
 531         struct list_head *lh;
 532         struct inet_protosw *answer;
 533         struct list_head *last_perm;
 534         int protocol = p->protocol;
 535         int ret;
 536
 537         spin_lock_bh(&inetsw6_lock);
 538
 539         ret = -EINVAL;
 540         if (p->type >= SOCK_MAX)
 541                 goto out_illegal;
 542
 543         /* If we are trying to override a permanent protocol, bail. */
 544         answer = NULL;
 545         ret = -EPERM;
 546         last_perm = &inetsw6[p->type];
 547         list_for_each(lh, &inetsw6[p->type]) {
 548                 answer = list_entry(lh, struct inet_protosw, list);
 549
 550                 /* Check only the non-wild match. */
 551                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 552                         if (protocol == answer->protocol)
 553                                 break;
 554                         last_perm = lh;
 555                 }
 556
 557                 answer = NULL;
 558         }
 559         if (answer)
 560                 goto out_permanent;
 561
 562         /* Add the new entry after the last permanent entry if any, so that
 563          * the new entry does not override a permanent entry when matched with
 564          * a wild-card protocol. But it is allowed to override any existing
 565          * non-permanent entry.  This means that when we remove this entry, the
 566          * system automatically returns to the old behavior.
 567          */
 568         list_add_rcu(&p->list, last_perm);
 569         ret = 0;
 570 out:
 571         spin_unlock_bh(&inetsw6_lock);
 572         return ret;
 573
 574 out_permanent:
 575         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 576                protocol);
 577         goto out;
 578
 579 out_illegal:
 580         printk(KERN_ERR
 581                "Ignoring attempt to register invalid socket type %d.\n",
 582                p->type);
 583         goto out;
 584 }
 585
 586 EXPORT_SYMBOL(inet6_register_protosw);
 587
 588 void
 589 inet6_unregister_protosw(struct inet_protosw *p)
 590 {
 591         if (INET_PROTOSW_PERMANENT & p->flags) {
 592                 printk(KERN_ERR
 593                        "Attempt to unregister permanent protocol %d.\n",
 594                        p->protocol);
 595         } else {
 596                 spin_lock_bh(&inetsw6_lock);
 597                 list_del_rcu(&p->list);
 598                 spin_unlock_bh(&inetsw6_lock);
 599
 600                 synchronize_net();
 601         }
 602 }
 603
 604 EXPORT_SYMBOL(inet6_unregister_protosw);
 605
 606 int inet6_sk_rebuild_header(struct sock *sk)
 607 {
 608         int err;
 609         struct dst_entry *dst;
 610         struct ipv6_pinfo *np = inet6_sk(sk);
 611
 612         dst = __sk_dst_check(sk, np->dst_cookie);
 613
 614         if (dst == NULL) {
 615                 struct inet_sock *inet = inet_sk(sk);
 616                 struct in6_addr *final_p = NULL, final;
 617                 struct flowi fl;
 618
 619                 memset(&fl, 0, sizeof(fl));
 620                 fl.proto = sk->sk_protocol;
 621                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 622                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 623                 fl.fl6_flowlabel = np->flow_label;
 624                 fl.oif = sk->sk_bound_dev_if;
 625                 fl.fl_ip_dport = inet->dport;
 626                 fl.fl_ip_sport = inet->sport;
 627                 security_sk_classify_flow(sk, &fl);
 628
 629                 if (np->opt && np->opt->srcrt) {
 630                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 631                         ipv6_addr_copy(&final, &fl.fl6_dst);
 632                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 633                         final_p = &final;
 634                 }
 635
 636                 err = ip6_dst_lookup(sk, &dst, &fl);
 637                 if (err) {
 638                         sk->sk_route_caps = 0;
 639                         return err;
 640                 }
 641                 if (final_p)
 642                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 643
 644                 if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
 645                         sk->sk_err_soft = -err;
 646                         return err;
 647                 }
 648
 649                 __ip6_dst_store(sk, dst, NULL, NULL);
 650         }
 651
 652         return 0;
 653 }
 654
 655 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 656
 657 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 658 {
 659         struct ipv6_pinfo *np = inet6_sk(sk);
 660         struct inet6_skb_parm *opt = IP6CB(skb);
 661
 662         if (np->rxopt.all) {
 663                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 664                                   np->rxopt.bits.ohopopts)) ||
 665                     ((IPV6_FLOWINFO_MASK &
 666                       *(__be32 *)skb_network_header(skb)) &&
 667                      np->rxopt.bits.rxflow) ||
 668                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 669                      np->rxopt.bits.osrcrt)) ||
 670                     ((opt->dst1 || opt->dst0) &&
 671                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 672                         return 1;
 673         }
 674         return 0;
 675 }
 676
 677 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 678
 679 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
 680 {
 681         struct inet6_protocol *ops = NULL;
 682
 683         for (;;) {
 684                 struct ipv6_opt_hdr *opth;
 685                 int len;
 686
 687                 if (proto != NEXTHDR_HOP) {
 688                         ops = rcu_dereference(inet6_protos[proto]);
 689
 690                         if (unlikely(!ops))
 691                                 break;
 692
 693                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
 694                                 break;
 695                 }
 696
 697                 if (unlikely(!pskb_may_pull(skb, 8)))
 698                         break;
 699
 700                 opth = (void *)skb->data;
 701                 len = ipv6_optlen(opth);
 702
 703                 if (unlikely(!pskb_may_pull(skb, len)))
 704                         break;
 705
 706                 proto = opth->nexthdr;
 707                 __skb_pull(skb, len);
 708         }
 709
 710         return proto;
 711 }
 712
 713 static int ipv6_gso_send_check(struct sk_buff *skb)
 714 {
 715         struct ipv6hdr *ipv6h;
 716         struct inet6_protocol *ops;
 717         int err = -EINVAL;
 718
 719         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 720                 goto out;
 721
 722         ipv6h = ipv6_hdr(skb);
 723         __skb_pull(skb, sizeof(*ipv6h));
 724         err = -EPROTONOSUPPORT;
 725
 726         rcu_read_lock();
 727         ops = rcu_dereference(inet6_protos[
 728                 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
 729
 730         if (likely(ops && ops->gso_send_check)) {
 731                 skb_reset_transport_header(skb);
 732                 err = ops->gso_send_check(skb);
 733         }
 734         rcu_read_unlock();
 735
 736 out:
 737         return err;
 738 }
 739
 740 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
 741 {
 742         struct sk_buff *segs = ERR_PTR(-EINVAL);
 743         struct ipv6hdr *ipv6h;
 744         struct inet6_protocol *ops;
 745
 746         if (!(features & NETIF_F_V6_CSUM))
 747                 features &= ~NETIF_F_SG;
 748
 749         if (unlikely(skb_shinfo(skb)->gso_type &
 750                      ~(SKB_GSO_UDP |
 751                        SKB_GSO_DODGY |
 752                        SKB_GSO_TCP_ECN |
 753                        SKB_GSO_TCPV6 |
 754                        0)))
 755                 goto out;
 756
 757         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 758                 goto out;
 759
 760         ipv6h = ipv6_hdr(skb);
 761         __skb_pull(skb, sizeof(*ipv6h));
 762         segs = ERR_PTR(-EPROTONOSUPPORT);
 763
 764         rcu_read_lock();
 765         ops = rcu_dereference(inet6_protos[
 766                 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
 767
 768         if (likely(ops && ops->gso_segment)) {
 769                 skb_reset_transport_header(skb);
 770                 segs = ops->gso_segment(skb, features);
 771         }
 772         rcu_read_unlock();
 773
 774         if (unlikely(IS_ERR(segs)))
 775                 goto out;
 776
 777         for (skb = segs; skb; skb = skb->next) {
 778                 ipv6h = ipv6_hdr(skb);
 779                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
 780                                            sizeof(*ipv6h));
 781         }
 782
 783 out:
 784         return segs;
 785 }
 786
 787 struct ipv6_gro_cb {
 788         struct napi_gro_cb napi;
 789         int proto;
 790 };
 791
 792 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
 793
 794 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
 795                                          struct sk_buff *skb)
 796 {
 797         struct inet6_protocol *ops;
 798         struct sk_buff **pp = NULL;
 799         struct sk_buff *p;
 800         struct ipv6hdr *iph;
 801         unsigned int nlen;
 802         int flush = 1;
 803         int proto;
 804         __wsum csum;
 805
 806         if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
 807                 goto out;
 808
 809         iph = ipv6_hdr(skb);
 810         __skb_pull(skb, sizeof(*iph));
 811
 812         flush += ntohs(iph->payload_len) != skb->len;
 813
 814         rcu_read_lock();
 815         proto = ipv6_gso_pull_exthdrs(skb, iph->nexthdr);
 816         iph = ipv6_hdr(skb);
 817         IPV6_GRO_CB(skb)->proto = proto;
 818         ops = rcu_dereference(inet6_protos[proto]);
 819         if (!ops || !ops->gro_receive)
 820                 goto out_unlock;
 821
 822         flush--;
 823         skb_reset_transport_header(skb);
 824         nlen = skb_network_header_len(skb);
 825
 826         for (p = *head; p; p = p->next) {
 827                 struct ipv6hdr *iph2;
 828
 829                 if (!NAPI_GRO_CB(p)->same_flow)
 830                         continue;
 831
 832                 iph2 = ipv6_hdr(p);
 833
 834                 /* All fields must match except length. */
 835                 if (nlen != skb_network_header_len(p) ||
 836                     memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
 837                     memcmp(&iph->nexthdr, &iph2->nexthdr,
 838                            nlen - offsetof(struct ipv6hdr, nexthdr))) {
 839                         NAPI_GRO_CB(p)->same_flow = 0;
 840                         continue;
 841                 }
 842
 843                 NAPI_GRO_CB(p)->flush |= flush;
 844         }
 845
 846         NAPI_GRO_CB(skb)->flush |= flush;
 847
 848         csum = skb->csum;
 849         skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
 850
 851         pp = ops->gro_receive(head, skb);
 852
 853         skb->csum = csum;
 854
 855 out_unlock:
 856         rcu_read_unlock();
 857
 858 out:
 859         NAPI_GRO_CB(skb)->flush |= flush;
 860
 861         return pp;
 862 }
 863
 864 static int ipv6_gro_complete(struct sk_buff *skb)
 865 {
 866         struct inet6_protocol *ops;
 867         struct ipv6hdr *iph = ipv6_hdr(skb);
 868         int err = -ENOSYS;
 869
 870         iph->payload_len = htons(skb->len - skb_network_offset(skb) -
 871                                  sizeof(*iph));
 872
 873         rcu_read_lock();
 874         ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
 875         if (WARN_ON(!ops || !ops->gro_complete))
 876                 goto out_unlock;
 877
 878         err = ops->gro_complete(skb);
 879
 880 out_unlock:
 881         rcu_read_unlock();
 882
 883         return err;
 884 }
 885
 886 static struct packet_type ipv6_packet_type = {
 887         .type = __constant_htons(ETH_P_IPV6),
 888         .func = ipv6_rcv,
 889         .gso_send_check = ipv6_gso_send_check,
 890         .gso_segment = ipv6_gso_segment,
 891         .gro_receive = ipv6_gro_receive,
 892         .gro_complete = ipv6_gro_complete,
 893 };
 894
 895 static int __init ipv6_packet_init(void)
 896 {
 897         dev_add_pack(&ipv6_packet_type);
 898         return 0;
 899 }
 900
 901 static void ipv6_packet_cleanup(void)
 902 {
 903         dev_remove_pack(&ipv6_packet_type);
 904 }
 905
 906 static int __net_init ipv6_init_mibs(struct net *net)
 907 {
 908         if (snmp_mib_init((void **)net->mib.udp_stats_in6,
 909                           sizeof (struct udp_mib)) < 0)
 910                 return -ENOMEM;
 911         if (snmp_mib_init((void **)net->mib.udplite_stats_in6,
 912                           sizeof (struct udp_mib)) < 0)
 913                 goto err_udplite_mib;
 914         if (snmp_mib_init((void **)net->mib.ipv6_statistics,
 915                           sizeof(struct ipstats_mib)) < 0)
 916                 goto err_ip_mib;
 917         if (snmp_mib_init((void **)net->mib.icmpv6_statistics,
 918                           sizeof(struct icmpv6_mib)) < 0)
 919                 goto err_icmp_mib;
 920         if (snmp_mib_init((void **)net->mib.icmpv6msg_statistics,
 921                           sizeof(struct icmpv6msg_mib)) < 0)
 922                 goto err_icmpmsg_mib;
 923         return 0;
 924
 925 err_icmpmsg_mib:
 926         snmp_mib_free((void **)net->mib.icmpv6_statistics);
 927 err_icmp_mib:
 928         snmp_mib_free((void **)net->mib.ipv6_statistics);
 929 err_ip_mib:
 930         snmp_mib_free((void **)net->mib.udplite_stats_in6);
 931 err_udplite_mib:
 932         snmp_mib_free((void **)net->mib.udp_stats_in6);
 933         return -ENOMEM;
 934 }
 935
 936 static void __net_exit ipv6_cleanup_mibs(struct net *net)
 937 {
 938         snmp_mib_free((void **)net->mib.udp_stats_in6);
 939         snmp_mib_free((void **)net->mib.udplite_stats_in6);
 940         snmp_mib_free((void **)net->mib.ipv6_statistics);
 941         snmp_mib_free((void **)net->mib.icmpv6_statistics);
 942         snmp_mib_free((void **)net->mib.icmpv6msg_statistics);
 943 }
 944
 945 static int __net_init inet6_net_init(struct net *net)
 946 {
 947         int err = 0;
 948
 949         net->ipv6.sysctl.bindv6only = 0;
 950         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 951
 952         err = ipv6_init_mibs(net);
 953         if (err)
 954                 return err;
 955 #ifdef CONFIG_PROC_FS
 956         err = udp6_proc_init(net);
 957         if (err)
 958                 goto out;
 959         err = tcp6_proc_init(net);
 960         if (err)
 961                 goto proc_tcp6_fail;
 962         err = ac6_proc_init(net);
 963         if (err)
 964                 goto proc_ac6_fail;
 965 #endif
 966         return err;
 967
 968 #ifdef CONFIG_PROC_FS
 969 proc_ac6_fail:
 970         tcp6_proc_exit(net);
 971 proc_tcp6_fail:
 972         udp6_proc_exit(net);
 973 out:
 974         ipv6_cleanup_mibs(net);
 975         return err;
 976 #endif
 977 }
 978
 979 static void inet6_net_exit(struct net *net)
 980 {
 981 #ifdef CONFIG_PROC_FS
 982         udp6_proc_exit(net);
 983         tcp6_proc_exit(net);
 984         ac6_proc_exit(net);
 985 #endif
 986         ipv6_cleanup_mibs(net);
 987 }
 988
 989 static struct pernet_operations inet6_net_ops = {
 990         .init = inet6_net_init,
 991         .exit = inet6_net_exit,
 992 };
 993
 994 static int __init inet6_init(void)
 995 {
 996         struct sk_buff *dummy_skb;
 997         struct list_head *r;
 998         int err = 0;
 999
1000         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1001
1002         /* Register the socket-side information for inet6_create.  */
1003         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1004                 INIT_LIST_HEAD(r);
1005
1006         if (disable_ipv6) {
1007                 printk(KERN_INFO
1008                        "IPv6: Loaded, but administratively disabled, "
1009                        "reboot required to enable\n");
1010                 goto out;
1011         }
1012
1013         err = proto_register(&tcpv6_prot, 1);
1014         if (err)
1015                 goto out;
1016
1017         err = proto_register(&udpv6_prot, 1);
1018         if (err)
1019                 goto out_unregister_tcp_proto;
1020
1021         err = proto_register(&udplitev6_prot, 1);
1022         if (err)
1023                 goto out_unregister_udp_proto;
1024
1025         err = proto_register(&rawv6_prot, 1);
1026         if (err)
1027                 goto out_unregister_udplite_proto;
1028
1029
1030         /* We MUST register RAW sockets before we create the ICMP6,
1031          * IGMP6, or NDISC control sockets.
1032          */
1033         err = rawv6_init();
1034         if (err)
1035                 goto out_unregister_raw_proto;
1036
1037         /* Register the family here so that the init calls below will
1038          * be able to create sockets. (?? is this dangerous ??)
1039          */
1040         err = sock_register(&inet6_family_ops);
1041         if (err)
1042                 goto out_sock_register_fail;
1043
1044 #ifdef CONFIG_SYSCTL
1045         err = ipv6_static_sysctl_register();
1046         if (err)
1047                 goto static_sysctl_fail;
1048 #endif
1049         /*
1050          *      ipngwg API draft makes clear that the correct semantics
1051          *      for TCP and UDP is to consider one TCP and UDP instance
1052          *      in a host availiable by both INET and INET6 APIs and
1053          *      able to communicate via both network protocols.
1054          */
1055
1056         err = register_pernet_subsys(&inet6_net_ops);
1057         if (err)
1058                 goto register_pernet_fail;
1059         err = icmpv6_init();
1060         if (err)
1061                 goto icmp_fail;
1062         err = ip6_mr_init();
1063         if (err)
1064                 goto ipmr_fail;
1065         err = ndisc_init();
1066         if (err)
1067                 goto ndisc_fail;
1068         err = igmp6_init();
1069         if (err)
1070                 goto igmp_fail;
1071         err = ipv6_netfilter_init();
1072         if (err)
1073                 goto netfilter_fail;
1074         /* Create /proc/foo6 entries. */
1075 #ifdef CONFIG_PROC_FS
1076         err = -ENOMEM;
1077         if (raw6_proc_init())
1078                 goto proc_raw6_fail;
1079         if (udplite6_proc_init())
1080                 goto proc_udplite6_fail;
1081         if (ipv6_misc_proc_init())
1082                 goto proc_misc6_fail;
1083         if (if6_proc_init())
1084                 goto proc_if6_fail;
1085 #endif
1086         err = ip6_route_init();
1087         if (err)
1088                 goto ip6_route_fail;
1089         err = ip6_flowlabel_init();
1090         if (err)
1091                 goto ip6_flowlabel_fail;
1092         err = addrconf_init();
1093         if (err)
1094                 goto addrconf_fail;
1095
1096         /* Init v6 extension headers. */
1097         err = ipv6_exthdrs_init();
1098         if (err)
1099                 goto ipv6_exthdrs_fail;
1100
1101         err = ipv6_frag_init();
1102         if (err)
1103                 goto ipv6_frag_fail;
1104
1105         /* Init v6 transport protocols. */
1106         err = udpv6_init();
1107         if (err)
1108                 goto udpv6_fail;
1109
1110         err = udplitev6_init();
1111         if (err)
1112                 goto udplitev6_fail;
1113
1114         err = tcpv6_init();
1115         if (err)
1116                 goto tcpv6_fail;
1117
1118         err = ipv6_packet_init();
1119         if (err)
1120                 goto ipv6_packet_fail;
1121
1122 #ifdef CONFIG_SYSCTL
1123         err = ipv6_sysctl_register();
1124         if (err)
1125                 goto sysctl_fail;
1126 #endif
1127 out:
1128         return err;
1129
1130 #ifdef CONFIG_SYSCTL
1131 sysctl_fail:
1132         ipv6_packet_cleanup();
1133 #endif
1134 ipv6_packet_fail:
1135         tcpv6_exit();
1136 tcpv6_fail:
1137         udplitev6_exit();
1138 udplitev6_fail:
1139         udpv6_exit();
1140 udpv6_fail:
1141         ipv6_frag_exit();
1142 ipv6_frag_fail:
1143         ipv6_exthdrs_exit();
1144 ipv6_exthdrs_fail:
1145         addrconf_cleanup();
1146 addrconf_fail:
1147         ip6_flowlabel_cleanup();
1148 ip6_flowlabel_fail:
1149         ip6_route_cleanup();
1150 ip6_route_fail:
1151 #ifdef CONFIG_PROC_FS
1152         if6_proc_exit();
1153 proc_if6_fail:
1154         ipv6_misc_proc_exit();
1155 proc_misc6_fail:
1156         udplite6_proc_exit();
1157 proc_udplite6_fail:
1158         raw6_proc_exit();
1159 proc_raw6_fail:
1160 #endif
1161         ipv6_netfilter_fini();
1162 netfilter_fail:
1163         igmp6_cleanup();
1164 igmp_fail:
1165         ndisc_cleanup();
1166 ndisc_fail:
1167         ip6_mr_cleanup();
1168 ipmr_fail:
1169         icmpv6_cleanup();
1170 icmp_fail:
1171         unregister_pernet_subsys(&inet6_net_ops);
1172 register_pernet_fail:
1173 #ifdef CONFIG_SYSCTL
1174         ipv6_static_sysctl_unregister();
1175 static_sysctl_fail:
1176 #endif
1177         sock_unregister(PF_INET6);
1178         rtnl_unregister_all(PF_INET6);
1179 out_sock_register_fail:
1180         rawv6_exit();
1181 out_unregister_raw_proto:
1182         proto_unregister(&rawv6_prot);
1183 out_unregister_udplite_proto:
1184         proto_unregister(&udplitev6_prot);
1185 out_unregister_udp_proto:
1186         proto_unregister(&udpv6_prot);
1187 out_unregister_tcp_proto:
1188         proto_unregister(&tcpv6_prot);
1189         goto out;
1190 }
1191 module_init(inet6_init);
1192
1193 static void __exit inet6_exit(void)
1194 {
1195         if (disable_ipv6)
1196                 return;
1197
1198         /* First of all disallow new sockets creation. */
1199         sock_unregister(PF_INET6);
1200         /* Disallow any further netlink messages */
1201         rtnl_unregister_all(PF_INET6);
1202
1203 #ifdef CONFIG_SYSCTL
1204         ipv6_sysctl_unregister();
1205 #endif
1206         udpv6_exit();
1207         udplitev6_exit();
1208         tcpv6_exit();
1209
1210         /* Cleanup code parts. */
1211         ipv6_packet_cleanup();
1212         ipv6_frag_exit();
1213         ipv6_exthdrs_exit();
1214         addrconf_cleanup();
1215         ip6_flowlabel_cleanup();
1216         ip6_route_cleanup();
1217 #ifdef CONFIG_PROC_FS
1218
1219         /* Cleanup code parts. */
1220         if6_proc_exit();
1221         ipv6_misc_proc_exit();
1222         udplite6_proc_exit();
1223         raw6_proc_exit();
1224 #endif
1225         ipv6_netfilter_fini();
1226         igmp6_cleanup();
1227         ndisc_cleanup();
1228         ip6_mr_cleanup();
1229         icmpv6_cleanup();
1230         rawv6_exit();
1231
1232         unregister_pernet_subsys(&inet6_net_ops);
1233 #ifdef CONFIG_SYSCTL
1234         ipv6_static_sysctl_unregister();
1235 #endif
1236         proto_unregister(&rawv6_prot);
1237         proto_unregister(&udplitev6_prot);
1238         proto_unregister(&udpv6_prot);
1239         proto_unregister(&tcpv6_prot);
1240 }
1241 module_exit(inet6_exit);
1242
1243 MODULE_ALIAS_NETPROTO(PF_INET6);