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/transp_v6.h>
  54 #include <net/ip6_route.h>
  55 #include <net/addrconf.h>
  56 #ifdef CONFIG_IPV6_TUNNEL
  57 #include <net/ip6_tunnel.h>
  58 #endif
  59
  60 #include <asm/uaccess.h>
  61 #include <asm/system.h>
  62 #ifdef CONFIG_IPV6_MROUTE
  63 #include <linux/mroute6.h>
  64 #endif
  65
  66 MODULE_AUTHOR("Cast of dozens");
  67 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  68 MODULE_LICENSE("GPL");
  69
  70 /* The inetsw6 table contains everything that inet6_create needs to
  71  * build a new socket.
  72  */
  73 static struct list_head inetsw6[SOCK_MAX];
  74 static DEFINE_SPINLOCK(inetsw6_lock);
  75
  76 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  77 {
  78         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  79
  80         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  81 }
  82
  83 static int inet6_create(struct net *net, struct socket *sock, int protocol)
  84 {
  85         struct inet_sock *inet;
  86         struct ipv6_pinfo *np;
  87         struct sock *sk;
  88         struct list_head *p;
  89         struct inet_protosw *answer;
  90         struct proto *answer_prot;
  91         unsigned char answer_flags;
  92         char answer_no_check;
  93         int try_loading_module = 0;
  94         int err;
  95
  96         if (sock->type != SOCK_RAW &&
  97             sock->type != SOCK_DGRAM &&
  98             !inet_ehash_secret)
  99                 build_ehash_secret();
 100
 101         /* Look for the requested type/protocol pair. */
 102         answer = NULL;
 103 lookup_protocol:
 104         err = -ESOCKTNOSUPPORT;
 105         rcu_read_lock();
 106         list_for_each_rcu(p, &inetsw6[sock->type]) {
 107                 answer = list_entry(p, struct inet_protosw, list);
 108
 109                 /* Check the non-wild match. */
 110                 if (protocol == answer->protocol) {
 111                         if (protocol != IPPROTO_IP)
 112                                 break;
 113                 } else {
 114                         /* Check for the two wild cases. */
 115                         if (IPPROTO_IP == protocol) {
 116                                 protocol = answer->protocol;
 117                                 break;
 118                         }
 119                         if (IPPROTO_IP == answer->protocol)
 120                                 break;
 121                 }
 122                 err = -EPROTONOSUPPORT;
 123                 answer = NULL;
 124         }
 125
 126         if (!answer) {
 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         BUG_TRAP(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 int 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         return 0;
 394 }
 395
 396 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 397
 398 /*
 399  *      This does both peername and sockname.
 400  */
 401
 402 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 403                  int *uaddr_len, int peer)
 404 {
 405         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 406         struct sock *sk = sock->sk;
 407         struct inet_sock *inet = inet_sk(sk);
 408         struct ipv6_pinfo *np = inet6_sk(sk);
 409
 410         sin->sin6_family = AF_INET6;
 411         sin->sin6_flowinfo = 0;
 412         sin->sin6_scope_id = 0;
 413         if (peer) {
 414                 if (!inet->dport)
 415                         return -ENOTCONN;
 416                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 417                     peer == 1)
 418                         return -ENOTCONN;
 419                 sin->sin6_port = inet->dport;
 420                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 421                 if (np->sndflow)
 422                         sin->sin6_flowinfo = np->flow_label;
 423         } else {
 424                 if (ipv6_addr_any(&np->rcv_saddr))
 425                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 426                 else
 427                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 428
 429                 sin->sin6_port = inet->sport;
 430         }
 431         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 432                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 433         *uaddr_len = sizeof(*sin);
 434         return(0);
 435 }
 436
 437 EXPORT_SYMBOL(inet6_getname);
 438
 439 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 440 {
 441         struct sock *sk = sock->sk;
 442         struct net *net = sock_net(sk);
 443
 444         switch(cmd)
 445         {
 446         case SIOCGSTAMP:
 447                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 448
 449         case SIOCGSTAMPNS:
 450                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 451
 452         case SIOCADDRT:
 453         case SIOCDELRT:
 454
 455                 return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
 456
 457         case SIOCSIFADDR:
 458                 return addrconf_add_ifaddr(net, (void __user *) arg);
 459         case SIOCDIFADDR:
 460                 return addrconf_del_ifaddr(net, (void __user *) arg);
 461         case SIOCSIFDSTADDR:
 462                 return addrconf_set_dstaddr(net, (void __user *) arg);
 463         default:
 464                 if (!sk->sk_prot->ioctl)
 465                         return -ENOIOCTLCMD;
 466                 return sk->sk_prot->ioctl(sk, cmd, arg);
 467         }
 468         /*NOTREACHED*/
 469         return(0);
 470 }
 471
 472 EXPORT_SYMBOL(inet6_ioctl);
 473
 474 const struct proto_ops inet6_stream_ops = {
 475         .family            = PF_INET6,
 476         .owner             = THIS_MODULE,
 477         .release           = inet6_release,
 478         .bind              = inet6_bind,
 479         .connect           = inet_stream_connect,       /* ok           */
 480         .socketpair        = sock_no_socketpair,        /* a do nothing */
 481         .accept            = inet_accept,               /* ok           */
 482         .getname           = inet6_getname,
 483         .poll              = tcp_poll,                  /* ok           */
 484         .ioctl             = inet6_ioctl,               /* must change  */
 485         .listen            = inet_listen,               /* ok           */
 486         .shutdown          = inet_shutdown,             /* ok           */
 487         .setsockopt        = sock_common_setsockopt,    /* ok           */
 488         .getsockopt        = sock_common_getsockopt,    /* ok           */
 489         .sendmsg           = tcp_sendmsg,               /* ok           */
 490         .recvmsg           = sock_common_recvmsg,       /* ok           */
 491         .mmap              = sock_no_mmap,
 492         .sendpage          = tcp_sendpage,
 493         .splice_read       = tcp_splice_read,
 494 #ifdef CONFIG_COMPAT
 495         .compat_setsockopt = compat_sock_common_setsockopt,
 496         .compat_getsockopt = compat_sock_common_getsockopt,
 497 #endif
 498 };
 499
 500 const struct proto_ops inet6_dgram_ops = {
 501         .family            = PF_INET6,
 502         .owner             = THIS_MODULE,
 503         .release           = inet6_release,
 504         .bind              = inet6_bind,
 505         .connect           = inet_dgram_connect,        /* ok           */
 506         .socketpair        = sock_no_socketpair,        /* a do nothing */
 507         .accept            = sock_no_accept,            /* a do nothing */
 508         .getname           = inet6_getname,
 509         .poll              = udp_poll,                  /* ok           */
 510         .ioctl             = inet6_ioctl,               /* must change  */
 511         .listen            = sock_no_listen,            /* ok           */
 512         .shutdown          = inet_shutdown,             /* ok           */
 513         .setsockopt        = sock_common_setsockopt,    /* ok           */
 514         .getsockopt        = sock_common_getsockopt,    /* ok           */
 515         .sendmsg           = inet_sendmsg,              /* ok           */
 516         .recvmsg           = sock_common_recvmsg,       /* ok           */
 517         .mmap              = sock_no_mmap,
 518         .sendpage          = sock_no_sendpage,
 519 #ifdef CONFIG_COMPAT
 520         .compat_setsockopt = compat_sock_common_setsockopt,
 521         .compat_getsockopt = compat_sock_common_getsockopt,
 522 #endif
 523 };
 524
 525 static struct net_proto_family inet6_family_ops = {
 526         .family = PF_INET6,
 527         .create = inet6_create,
 528         .owner  = THIS_MODULE,
 529 };
 530
 531 int inet6_register_protosw(struct inet_protosw *p)
 532 {
 533         struct list_head *lh;
 534         struct inet_protosw *answer;
 535         struct list_head *last_perm;
 536         int protocol = p->protocol;
 537         int ret;
 538
 539         spin_lock_bh(&inetsw6_lock);
 540
 541         ret = -EINVAL;
 542         if (p->type >= SOCK_MAX)
 543                 goto out_illegal;
 544
 545         /* If we are trying to override a permanent protocol, bail. */
 546         answer = NULL;
 547         ret = -EPERM;
 548         last_perm = &inetsw6[p->type];
 549         list_for_each(lh, &inetsw6[p->type]) {
 550                 answer = list_entry(lh, struct inet_protosw, list);
 551
 552                 /* Check only the non-wild match. */
 553                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 554                         if (protocol == answer->protocol)
 555                                 break;
 556                         last_perm = lh;
 557                 }
 558
 559                 answer = NULL;
 560         }
 561         if (answer)
 562                 goto out_permanent;
 563
 564         /* Add the new entry after the last permanent entry if any, so that
 565          * the new entry does not override a permanent entry when matched with
 566          * a wild-card protocol. But it is allowed to override any existing
 567          * non-permanent entry.  This means that when we remove this entry, the
 568          * system automatically returns to the old behavior.
 569          */
 570         list_add_rcu(&p->list, last_perm);
 571         ret = 0;
 572 out:
 573         spin_unlock_bh(&inetsw6_lock);
 574         return ret;
 575
 576 out_permanent:
 577         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 578                protocol);
 579         goto out;
 580
 581 out_illegal:
 582         printk(KERN_ERR
 583                "Ignoring attempt to register invalid socket type %d.\n",
 584                p->type);
 585         goto out;
 586 }
 587
 588 EXPORT_SYMBOL(inet6_register_protosw);
 589
 590 void
 591 inet6_unregister_protosw(struct inet_protosw *p)
 592 {
 593         if (INET_PROTOSW_PERMANENT & p->flags) {
 594                 printk(KERN_ERR
 595                        "Attempt to unregister permanent protocol %d.\n",
 596                        p->protocol);
 597         } else {
 598                 spin_lock_bh(&inetsw6_lock);
 599                 list_del_rcu(&p->list);
 600                 spin_unlock_bh(&inetsw6_lock);
 601
 602                 synchronize_net();
 603         }
 604 }
 605
 606 EXPORT_SYMBOL(inet6_unregister_protosw);
 607
 608 int inet6_sk_rebuild_header(struct sock *sk)
 609 {
 610         int err;
 611         struct dst_entry *dst;
 612         struct ipv6_pinfo *np = inet6_sk(sk);
 613
 614         dst = __sk_dst_check(sk, np->dst_cookie);
 615
 616         if (dst == NULL) {
 617                 struct inet_sock *inet = inet_sk(sk);
 618                 struct in6_addr *final_p = NULL, final;
 619                 struct flowi fl;
 620
 621                 memset(&fl, 0, sizeof(fl));
 622                 fl.proto = sk->sk_protocol;
 623                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 624                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 625                 fl.fl6_flowlabel = np->flow_label;
 626                 fl.oif = sk->sk_bound_dev_if;
 627                 fl.fl_ip_dport = inet->dport;
 628                 fl.fl_ip_sport = inet->sport;
 629                 security_sk_classify_flow(sk, &fl);
 630
 631                 if (np->opt && np->opt->srcrt) {
 632                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 633                         ipv6_addr_copy(&final, &fl.fl6_dst);
 634                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 635                         final_p = &final;
 636                 }
 637
 638                 err = ip6_dst_lookup(sk, &dst, &fl);
 639                 if (err) {
 640                         sk->sk_route_caps = 0;
 641                         return err;
 642                 }
 643                 if (final_p)
 644                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 645
 646                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
 647                         sk->sk_err_soft = -err;
 648                         return err;
 649                 }
 650
 651                 __ip6_dst_store(sk, dst, NULL, NULL);
 652         }
 653
 654         return 0;
 655 }
 656
 657 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 658
 659 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 660 {
 661         struct ipv6_pinfo *np = inet6_sk(sk);
 662         struct inet6_skb_parm *opt = IP6CB(skb);
 663
 664         if (np->rxopt.all) {
 665                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 666                                   np->rxopt.bits.ohopopts)) ||
 667                     ((IPV6_FLOWINFO_MASK &
 668                       *(__be32 *)skb_network_header(skb)) &&
 669                      np->rxopt.bits.rxflow) ||
 670                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 671                      np->rxopt.bits.osrcrt)) ||
 672                     ((opt->dst1 || opt->dst0) &&
 673                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 674                         return 1;
 675         }
 676         return 0;
 677 }
 678
 679 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 680
 681 static struct inet6_protocol *ipv6_gso_pull_exthdrs(struct sk_buff *skb,
 682                                                     int proto)
 683 {
 684         struct inet6_protocol *ops = NULL;
 685
 686         for (;;) {
 687                 struct ipv6_opt_hdr *opth;
 688                 int len;
 689
 690                 if (proto != NEXTHDR_HOP) {
 691                         ops = rcu_dereference(inet6_protos[proto]);
 692
 693                         if (unlikely(!ops))
 694                                 break;
 695
 696                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
 697                                 break;
 698                 }
 699
 700                 if (unlikely(!pskb_may_pull(skb, 8)))
 701                         break;
 702
 703                 opth = (void *)skb->data;
 704                 len = ipv6_optlen(opth);
 705
 706                 if (unlikely(!pskb_may_pull(skb, len)))
 707                         break;
 708
 709                 proto = opth->nexthdr;
 710                 __skb_pull(skb, len);
 711         }
 712
 713         return ops;
 714 }
 715
 716 static int ipv6_gso_send_check(struct sk_buff *skb)
 717 {
 718         struct ipv6hdr *ipv6h;
 719         struct inet6_protocol *ops;
 720         int err = -EINVAL;
 721
 722         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 723                 goto out;
 724
 725         ipv6h = ipv6_hdr(skb);
 726         __skb_pull(skb, sizeof(*ipv6h));
 727         err = -EPROTONOSUPPORT;
 728
 729         rcu_read_lock();
 730         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
 731         if (likely(ops && ops->gso_send_check)) {
 732                 skb_reset_transport_header(skb);
 733                 err = ops->gso_send_check(skb);
 734         }
 735         rcu_read_unlock();
 736
 737 out:
 738         return err;
 739 }
 740
 741 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
 742 {
 743         struct sk_buff *segs = ERR_PTR(-EINVAL);
 744         struct ipv6hdr *ipv6h;
 745         struct inet6_protocol *ops;
 746
 747         if (!(features & NETIF_F_V6_CSUM))
 748                 features &= ~NETIF_F_SG;
 749
 750         if (unlikely(skb_shinfo(skb)->gso_type &
 751                      ~(SKB_GSO_UDP |
 752                        SKB_GSO_DODGY |
 753                        SKB_GSO_TCP_ECN |
 754                        SKB_GSO_TCPV6 |
 755                        0)))
 756                 goto out;
 757
 758         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 759                 goto out;
 760
 761         ipv6h = ipv6_hdr(skb);
 762         __skb_pull(skb, sizeof(*ipv6h));
 763         segs = ERR_PTR(-EPROTONOSUPPORT);
 764
 765         rcu_read_lock();
 766         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
 767         if (likely(ops && ops->gso_segment)) {
 768                 skb_reset_transport_header(skb);
 769                 segs = ops->gso_segment(skb, features);
 770         }
 771         rcu_read_unlock();
 772
 773         if (unlikely(IS_ERR(segs)))
 774                 goto out;
 775
 776         for (skb = segs; skb; skb = skb->next) {
 777                 ipv6h = ipv6_hdr(skb);
 778                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
 779                                            sizeof(*ipv6h));
 780         }
 781
 782 out:
 783         return segs;
 784 }
 785
 786 static struct packet_type ipv6_packet_type = {
 787         .type = __constant_htons(ETH_P_IPV6),
 788         .func = ipv6_rcv,
 789         .gso_send_check = ipv6_gso_send_check,
 790         .gso_segment = ipv6_gso_segment,
 791 };
 792
 793 static int __init ipv6_packet_init(void)
 794 {
 795         dev_add_pack(&ipv6_packet_type);
 796         return 0;
 797 }
 798
 799 static void ipv6_packet_cleanup(void)
 800 {
 801         dev_remove_pack(&ipv6_packet_type);
 802 }
 803
 804 static int __init init_ipv6_mibs(void)
 805 {
 806         if (snmp_mib_init((void **)ipv6_statistics,
 807                           sizeof(struct ipstats_mib)) < 0)
 808                 goto err_ip_mib;
 809         if (snmp_mib_init((void **)icmpv6_statistics,
 810                           sizeof(struct icmpv6_mib)) < 0)
 811                 goto err_icmp_mib;
 812         if (snmp_mib_init((void **)icmpv6msg_statistics,
 813                           sizeof(struct icmpv6msg_mib)) < 0)
 814                 goto err_icmpmsg_mib;
 815         if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib)) < 0)
 816                 goto err_udp_mib;
 817         if (snmp_mib_init((void **)udplite_stats_in6,
 818                           sizeof (struct udp_mib)) < 0)
 819                 goto err_udplite_mib;
 820         return 0;
 821
 822 err_udplite_mib:
 823         snmp_mib_free((void **)udp_stats_in6);
 824 err_udp_mib:
 825         snmp_mib_free((void **)icmpv6msg_statistics);
 826 err_icmpmsg_mib:
 827         snmp_mib_free((void **)icmpv6_statistics);
 828 err_icmp_mib:
 829         snmp_mib_free((void **)ipv6_statistics);
 830 err_ip_mib:
 831         return -ENOMEM;
 832
 833 }
 834
 835 static void cleanup_ipv6_mibs(void)
 836 {
 837         snmp_mib_free((void **)ipv6_statistics);
 838         snmp_mib_free((void **)icmpv6_statistics);
 839         snmp_mib_free((void **)icmpv6msg_statistics);
 840         snmp_mib_free((void **)udp_stats_in6);
 841         snmp_mib_free((void **)udplite_stats_in6);
 842 }
 843
 844 static int inet6_net_init(struct net *net)
 845 {
 846         int err = 0;
 847
 848         net->ipv6.sysctl.bindv6only = 0;
 849         net->ipv6.sysctl.flush_delay = 0;
 850         net->ipv6.sysctl.ip6_rt_max_size = 4096;
 851         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
 852         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
 853         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
 854         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
 855         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
 856         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
 857         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 858
 859 #ifdef CONFIG_PROC_FS
 860         err = udp6_proc_init(net);
 861         if (err)
 862                 goto out;
 863         err = tcp6_proc_init(net);
 864         if (err)
 865                 goto proc_tcp6_fail;
 866         err = ac6_proc_init(net);
 867         if (err)
 868                 goto proc_ac6_fail;
 869 out:
 870 #endif
 871         return err;
 872
 873 #ifdef CONFIG_PROC_FS
 874 proc_ac6_fail:
 875         tcp6_proc_exit(net);
 876 proc_tcp6_fail:
 877         udp6_proc_exit(net);
 878         goto out;
 879 #endif
 880 }
 881
 882 static void inet6_net_exit(struct net *net)
 883 {
 884 #ifdef CONFIG_PROC_FS
 885         udp6_proc_exit(net);
 886         tcp6_proc_exit(net);
 887         ac6_proc_exit(net);
 888 #endif
 889 }
 890
 891 static struct pernet_operations inet6_net_ops = {
 892         .init = inet6_net_init,
 893         .exit = inet6_net_exit,
 894 };
 895
 896 static int __init inet6_init(void)
 897 {
 898         struct sk_buff *dummy_skb;
 899         struct list_head *r;
 900         int err;
 901
 902         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
 903
 904         err = proto_register(&tcpv6_prot, 1);
 905         if (err)
 906                 goto out;
 907
 908         err = proto_register(&udpv6_prot, 1);
 909         if (err)
 910                 goto out_unregister_tcp_proto;
 911
 912         err = proto_register(&udplitev6_prot, 1);
 913         if (err)
 914                 goto out_unregister_udp_proto;
 915
 916         err = proto_register(&rawv6_prot, 1);
 917         if (err)
 918                 goto out_unregister_udplite_proto;
 919
 920
 921         /* Register the socket-side information for inet6_create.  */
 922         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 923                 INIT_LIST_HEAD(r);
 924
 925         /* We MUST register RAW sockets before we create the ICMP6,
 926          * IGMP6, or NDISC control sockets.
 927          */
 928         err = rawv6_init();
 929         if (err)
 930                 goto out_unregister_raw_proto;
 931
 932         /* Register the family here so that the init calls below will
 933          * be able to create sockets. (?? is this dangerous ??)
 934          */
 935         err = sock_register(&inet6_family_ops);
 936         if (err)
 937                 goto out_sock_register_fail;
 938
 939         /* Initialise ipv6 mibs */
 940         err = init_ipv6_mibs();
 941         if (err)
 942                 goto out_unregister_sock;
 943
 944         /*
 945          *      ipngwg API draft makes clear that the correct semantics
 946          *      for TCP and UDP is to consider one TCP and UDP instance
 947          *      in a host availiable by both INET and INET6 APIs and
 948          *      able to communicate via both network protocols.
 949          */
 950
 951         err = register_pernet_subsys(&inet6_net_ops);
 952         if (err)
 953                 goto register_pernet_fail;
 954         err = icmpv6_init();
 955         if (err)
 956                 goto icmp_fail;
 957 #ifdef CONFIG_IPV6_MROUTE
 958         ip6_mr_init();
 959 #endif
 960         err = ndisc_init();
 961         if (err)
 962                 goto ndisc_fail;
 963         err = igmp6_init();
 964         if (err)
 965                 goto igmp_fail;
 966         err = ipv6_netfilter_init();
 967         if (err)
 968                 goto netfilter_fail;
 969         /* Create /proc/foo6 entries. */
 970 #ifdef CONFIG_PROC_FS
 971         err = -ENOMEM;
 972         if (raw6_proc_init())
 973                 goto proc_raw6_fail;
 974         if (udplite6_proc_init())
 975                 goto proc_udplite6_fail;
 976         if (ipv6_misc_proc_init())
 977                 goto proc_misc6_fail;
 978         if (if6_proc_init())
 979                 goto proc_if6_fail;
 980 #endif
 981         err = ip6_route_init();
 982         if (err)
 983                 goto ip6_route_fail;
 984         err = ip6_flowlabel_init();
 985         if (err)
 986                 goto ip6_flowlabel_fail;
 987         err = addrconf_init();
 988         if (err)
 989                 goto addrconf_fail;
 990
 991         /* Init v6 extension headers. */
 992         err = ipv6_exthdrs_init();
 993         if (err)
 994                 goto ipv6_exthdrs_fail;
 995
 996         err = ipv6_frag_init();
 997         if (err)
 998                 goto ipv6_frag_fail;
 999
1000         /* Init v6 transport protocols. */
1001         err = udpv6_init();
1002         if (err)
1003                 goto udpv6_fail;
1004
1005         err = udplitev6_init();
1006         if (err)
1007                 goto udplitev6_fail;
1008
1009         err = tcpv6_init();
1010         if (err)
1011                 goto tcpv6_fail;
1012
1013         err = ipv6_packet_init();
1014         if (err)
1015                 goto ipv6_packet_fail;
1016
1017 #ifdef CONFIG_SYSCTL
1018         err = ipv6_sysctl_register();
1019         if (err)
1020                 goto sysctl_fail;
1021 #endif
1022 out:
1023         return err;
1024
1025 #ifdef CONFIG_SYSCTL
1026 sysctl_fail:
1027         ipv6_packet_cleanup();
1028 #endif
1029 ipv6_packet_fail:
1030         tcpv6_exit();
1031 tcpv6_fail:
1032         udplitev6_exit();
1033 udplitev6_fail:
1034         udpv6_exit();
1035 udpv6_fail:
1036         ipv6_frag_exit();
1037 ipv6_frag_fail:
1038         ipv6_exthdrs_exit();
1039 ipv6_exthdrs_fail:
1040         addrconf_cleanup();
1041 addrconf_fail:
1042         ip6_flowlabel_cleanup();
1043 ip6_flowlabel_fail:
1044         ip6_route_cleanup();
1045 ip6_route_fail:
1046 #ifdef CONFIG_PROC_FS
1047         if6_proc_exit();
1048 proc_if6_fail:
1049         ipv6_misc_proc_exit();
1050 proc_misc6_fail:
1051         udplite6_proc_exit();
1052 proc_udplite6_fail:
1053         raw6_proc_exit();
1054 proc_raw6_fail:
1055 #endif
1056         ipv6_netfilter_fini();
1057 netfilter_fail:
1058         igmp6_cleanup();
1059 igmp_fail:
1060         ndisc_cleanup();
1061 ndisc_fail:
1062         icmpv6_cleanup();
1063 icmp_fail:
1064         unregister_pernet_subsys(&inet6_net_ops);
1065 register_pernet_fail:
1066         cleanup_ipv6_mibs();
1067 out_unregister_sock:
1068         sock_unregister(PF_INET6);
1069         rtnl_unregister_all(PF_INET6);
1070 out_sock_register_fail:
1071         rawv6_exit();
1072 out_unregister_raw_proto:
1073         proto_unregister(&rawv6_prot);
1074 out_unregister_udplite_proto:
1075         proto_unregister(&udplitev6_prot);
1076 out_unregister_udp_proto:
1077         proto_unregister(&udpv6_prot);
1078 out_unregister_tcp_proto:
1079         proto_unregister(&tcpv6_prot);
1080         goto out;
1081 }
1082 module_init(inet6_init);
1083
1084 static void __exit inet6_exit(void)
1085 {
1086         /* First of all disallow new sockets creation. */
1087         sock_unregister(PF_INET6);
1088         /* Disallow any further netlink messages */
1089         rtnl_unregister_all(PF_INET6);
1090
1091 #ifdef CONFIG_SYSCTL
1092         ipv6_sysctl_unregister();
1093 #endif
1094         udpv6_exit();
1095         udplitev6_exit();
1096         tcpv6_exit();
1097
1098         /* Cleanup code parts. */
1099         ipv6_packet_cleanup();
1100         ipv6_frag_exit();
1101         ipv6_exthdrs_exit();
1102         addrconf_cleanup();
1103         ip6_flowlabel_cleanup();
1104         ip6_route_cleanup();
1105 #ifdef CONFIG_PROC_FS
1106
1107         /* Cleanup code parts. */
1108         if6_proc_exit();
1109         ipv6_misc_proc_exit();
1110         udplite6_proc_exit();
1111         raw6_proc_exit();
1112 #endif
1113         ipv6_netfilter_fini();
1114         igmp6_cleanup();
1115         ndisc_cleanup();
1116         icmpv6_cleanup();
1117         rawv6_exit();
1118
1119         unregister_pernet_subsys(&inet6_net_ops);
1120         cleanup_ipv6_mibs();
1121         proto_unregister(&rawv6_prot);
1122         proto_unregister(&udplitev6_prot);
1123         proto_unregister(&udpv6_prot);
1124         proto_unregister(&tcpv6_prot);
1125 }
1126 module_exit(inet6_exit);
1127
1128 MODULE_ALIAS_NETPROTO(PF_INET6);