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  *      $Id: af_inet6.c,v 1.66 2002/02/01 22:01:04 davem Exp $
  11  *
  12  *      Fixes:
  13  *      piggy, Karl Knutson     :       Socket protocol table
  14  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
  15  *      Arnaldo Melo            :       check proc_net_create return, cleanups
  16  *
  17  *      This program is free software; you can redistribute it and/or
  18  *      modify it under the terms of the GNU General Public License
  19  *      as published by the Free Software Foundation; either version
  20  *      2 of the License, or (at your option) any later version.
  21  */
  22
  23
  24 #include <linux/module.h>
  25 #include <linux/capability.h>
  26 #include <linux/errno.h>
  27 #include <linux/types.h>
  28 #include <linux/socket.h>
  29 #include <linux/in.h>
  30 #include <linux/kernel.h>
  31 #include <linux/timer.h>
  32 #include <linux/string.h>
  33 #include <linux/sockios.h>
  34 #include <linux/net.h>
  35 #include <linux/fcntl.h>
  36 #include <linux/mm.h>
  37 #include <linux/interrupt.h>
  38 #include <linux/proc_fs.h>
  39 #include <linux/stat.h>
  40 #include <linux/init.h>
  41
  42 #include <linux/inet.h>
  43 #include <linux/netdevice.h>
  44 #include <linux/icmpv6.h>
  45 #include <linux/netfilter_ipv6.h>
  46
  47 #include <net/ip.h>
  48 #include <net/ipv6.h>
  49 #include <net/udp.h>
  50 #include <net/udplite.h>
  51 #include <net/tcp.h>
  52 #include <net/ipip.h>
  53 #include <net/protocol.h>
  54 #include <net/inet_common.h>
  55 #include <net/transp_v6.h>
  56 #include <net/ip6_route.h>
  57 #include <net/addrconf.h>
  58 #ifdef CONFIG_IPV6_TUNNEL
  59 #include <net/ip6_tunnel.h>
  60 #endif
  61
  62 #include <asm/uaccess.h>
  63 #include <asm/system.h>
  64 #ifdef CONFIG_IPV6_MROUTE
  65 #include <linux/mroute6.h>
  66 #endif
  67
  68 MODULE_AUTHOR("Cast of dozens");
  69 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  70 MODULE_LICENSE("GPL");
  71
  72 /* The inetsw6 table contains everything that inet6_create needs to
  73  * build a new socket.
  74  */
  75 static struct list_head inetsw6[SOCK_MAX];
  76 static DEFINE_SPINLOCK(inetsw6_lock);
  77
  78 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  79 {
  80         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  81
  82         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  83 }
  84
  85 static int inet6_create(struct net *net, struct socket *sock, int protocol)
  86 {
  87         struct inet_sock *inet;
  88         struct ipv6_pinfo *np;
  89         struct sock *sk;
  90         struct list_head *p;
  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         answer = NULL;
 105 lookup_protocol:
 106         err = -ESOCKTNOSUPPORT;
 107         rcu_read_lock();
 108         list_for_each_rcu(p, &inetsw6[sock->type]) {
 109                 answer = list_entry(p, struct inet_protosw, list);
 110
 111                 /* Check the non-wild match. */
 112                 if (protocol == answer->protocol) {
 113                         if (protocol != IPPROTO_IP)
 114                                 break;
 115                 } else {
 116                         /* Check for the two wild cases. */
 117                         if (IPPROTO_IP == protocol) {
 118                                 protocol = answer->protocol;
 119                                 break;
 120                         }
 121                         if (IPPROTO_IP == answer->protocol)
 122                                 break;
 123                 }
 124                 err = -EPROTONOSUPPORT;
 125                 answer = NULL;
 126         }
 127
 128         if (!answer) {
 129                 if (try_loading_module < 2) {
 130                         rcu_read_unlock();
 131                         /*
 132                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 133                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 134                          */
 135                         if (++try_loading_module == 1)
 136                                 request_module("net-pf-%d-proto-%d-type-%d",
 137                                                 PF_INET6, protocol, sock->type);
 138                         /*
 139                          * Fall back to generic, e.g. net-pf-10-proto-132
 140                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 141                          */
 142                         else
 143                                 request_module("net-pf-%d-proto-%d",
 144                                                 PF_INET6, protocol);
 145                         goto lookup_protocol;
 146                 } else
 147                         goto out_rcu_unlock;
 148         }
 149
 150         err = -EPERM;
 151         if (answer->capability > 0 && !capable(answer->capability))
 152                 goto out_rcu_unlock;
 153
 154         sock->ops = answer->ops;
 155         answer_prot = answer->prot;
 156         answer_no_check = answer->no_check;
 157         answer_flags = answer->flags;
 158         rcu_read_unlock();
 159
 160         BUG_TRAP(answer_prot->slab != NULL);
 161
 162         err = -ENOBUFS;
 163         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 164         if (sk == NULL)
 165                 goto out;
 166
 167         sock_init_data(sock, sk);
 168
 169         err = 0;
 170         sk->sk_no_check = answer_no_check;
 171         if (INET_PROTOSW_REUSE & answer_flags)
 172                 sk->sk_reuse = 1;
 173
 174         inet = inet_sk(sk);
 175         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 176
 177         if (SOCK_RAW == sock->type) {
 178                 inet->num = protocol;
 179                 if (IPPROTO_RAW == protocol)
 180                         inet->hdrincl = 1;
 181         }
 182
 183         sk->sk_destruct         = inet_sock_destruct;
 184         sk->sk_family           = PF_INET6;
 185         sk->sk_protocol         = protocol;
 186
 187         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 188
 189         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 190         np->hop_limit   = -1;
 191         np->mcast_hops  = -1;
 192         np->mc_loop     = 1;
 193         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 194         np->ipv6only    = init_net.ipv6.sysctl.bindv6only;
 195
 196         /* Init the ipv4 part of the socket since we can have sockets
 197          * using v6 API for ipv4.
 198          */
 199         inet->uc_ttl    = -1;
 200
 201         inet->mc_loop   = 1;
 202         inet->mc_ttl    = 1;
 203         inet->mc_index  = 0;
 204         inet->mc_list   = NULL;
 205
 206         if (ipv4_config.no_pmtu_disc)
 207                 inet->pmtudisc = IP_PMTUDISC_DONT;
 208         else
 209                 inet->pmtudisc = IP_PMTUDISC_WANT;
 210         /*
 211          * Increment only the relevant sk_prot->socks debug field, this changes
 212          * the previous behaviour of incrementing both the equivalent to
 213          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 214          *
 215          * This allows better debug granularity as we'll know exactly how many
 216          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 217          * transport protocol socks. -acme
 218          */
 219         sk_refcnt_debug_inc(sk);
 220
 221         if (inet->num) {
 222                 /* It assumes that any protocol which allows
 223                  * the user to assign a number at socket
 224                  * creation time automatically shares.
 225                  */
 226                 inet->sport = htons(inet->num);
 227                 sk->sk_prot->hash(sk);
 228         }
 229         if (sk->sk_prot->init) {
 230                 err = sk->sk_prot->init(sk);
 231                 if (err) {
 232                         sk_common_release(sk);
 233                         goto out;
 234                 }
 235         }
 236 out:
 237         return err;
 238 out_rcu_unlock:
 239         rcu_read_unlock();
 240         goto out;
 241 }
 242
 243
 244 /* bind for INET6 API */
 245 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 246 {
 247         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 248         struct sock *sk = sock->sk;
 249         struct inet_sock *inet = inet_sk(sk);
 250         struct ipv6_pinfo *np = inet6_sk(sk);
 251         struct net *net = sock_net(sk);
 252         __be32 v4addr = 0;
 253         unsigned short snum;
 254         int addr_type = 0;
 255         int err = 0;
 256
 257         /* If the socket has its own bind function then use it. */
 258         if (sk->sk_prot->bind)
 259                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 260
 261         if (addr_len < SIN6_LEN_RFC2133)
 262                 return -EINVAL;
 263         addr_type = ipv6_addr_type(&addr->sin6_addr);
 264         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 265                 return -EINVAL;
 266
 267         snum = ntohs(addr->sin6_port);
 268         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 269                 return -EACCES;
 270
 271         lock_sock(sk);
 272
 273         /* Check these errors (active socket, double bind). */
 274         if (sk->sk_state != TCP_CLOSE || inet->num) {
 275                 err = -EINVAL;
 276                 goto out;
 277         }
 278
 279         /* Check if the address belongs to the host. */
 280         if (addr_type == IPV6_ADDR_MAPPED) {
 281                 v4addr = addr->sin6_addr.s6_addr32[3];
 282                 if (inet_addr_type(net, v4addr) != RTN_LOCAL) {
 283                         err = -EADDRNOTAVAIL;
 284                         goto out;
 285                 }
 286         } else {
 287                 if (addr_type != IPV6_ADDR_ANY) {
 288                         struct net_device *dev = NULL;
 289
 290                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 291                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 292                                     addr->sin6_scope_id) {
 293                                         /* Override any existing binding, if another one
 294                                          * is supplied by user.
 295                                          */
 296                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 297                                 }
 298
 299                                 /* Binding to link-local address requires an interface */
 300                                 if (!sk->sk_bound_dev_if) {
 301                                         err = -EINVAL;
 302                                         goto out;
 303                                 }
 304                                 dev = dev_get_by_index(net, sk->sk_bound_dev_if);
 305                                 if (!dev) {
 306                                         err = -ENODEV;
 307                                         goto out;
 308                                 }
 309                         }
 310
 311                         /* ipv4 addr of the socket is invalid.  Only the
 312                          * unspecified and mapped address have a v4 equivalent.
 313                          */
 314                         v4addr = LOOPBACK4_IPV6;
 315                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 316                                 if (!ipv6_chk_addr(net, &addr->sin6_addr,
 317                                                    dev, 0)) {
 318                                         if (dev)
 319                                                 dev_put(dev);
 320                                         err = -EADDRNOTAVAIL;
 321                                         goto out;
 322                                 }
 323                         }
 324                         if (dev)
 325                                 dev_put(dev);
 326                 }
 327         }
 328
 329         inet->rcv_saddr = v4addr;
 330         inet->saddr = v4addr;
 331
 332         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 333
 334         if (!(addr_type & IPV6_ADDR_MULTICAST))
 335                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 336
 337         /* Make sure we are allowed to bind here. */
 338         if (sk->sk_prot->get_port(sk, snum)) {
 339                 inet_reset_saddr(sk);
 340                 err = -EADDRINUSE;
 341                 goto out;
 342         }
 343
 344         if (addr_type != IPV6_ADDR_ANY)
 345                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 346         if (snum)
 347                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 348         inet->sport = htons(inet->num);
 349         inet->dport = 0;
 350         inet->daddr = 0;
 351 out:
 352         release_sock(sk);
 353         return err;
 354 }
 355
 356 EXPORT_SYMBOL(inet6_bind);
 357
 358 int inet6_release(struct socket *sock)
 359 {
 360         struct sock *sk = sock->sk;
 361
 362         if (sk == NULL)
 363                 return -EINVAL;
 364
 365         /* Free mc lists */
 366         ipv6_sock_mc_close(sk);
 367
 368         /* Free ac lists */
 369         ipv6_sock_ac_close(sk);
 370
 371         return inet_release(sock);
 372 }
 373
 374 EXPORT_SYMBOL(inet6_release);
 375
 376 int inet6_destroy_sock(struct sock *sk)
 377 {
 378         struct ipv6_pinfo *np = inet6_sk(sk);
 379         struct sk_buff *skb;
 380         struct ipv6_txoptions *opt;
 381
 382         /* Release rx options */
 383
 384         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 385                 kfree_skb(skb);
 386
 387         /* Free flowlabels */
 388         fl6_free_socklist(sk);
 389
 390         /* Free tx options */
 391
 392         if ((opt = xchg(&np->opt, NULL)) != NULL)
 393                 sock_kfree_s(sk, opt, opt->tot_len);
 394
 395         return 0;
 396 }
 397
 398 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 399
 400 /*
 401  *      This does both peername and sockname.
 402  */
 403
 404 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 405                  int *uaddr_len, int peer)
 406 {
 407         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 408         struct sock *sk = sock->sk;
 409         struct inet_sock *inet = inet_sk(sk);
 410         struct ipv6_pinfo *np = inet6_sk(sk);
 411
 412         sin->sin6_family = AF_INET6;
 413         sin->sin6_flowinfo = 0;
 414         sin->sin6_scope_id = 0;
 415         if (peer) {
 416                 if (!inet->dport)
 417                         return -ENOTCONN;
 418                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 419                     peer == 1)
 420                         return -ENOTCONN;
 421                 sin->sin6_port = inet->dport;
 422                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 423                 if (np->sndflow)
 424                         sin->sin6_flowinfo = np->flow_label;
 425         } else {
 426                 if (ipv6_addr_any(&np->rcv_saddr))
 427                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 428                 else
 429                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 430
 431                 sin->sin6_port = inet->sport;
 432         }
 433         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 434                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 435         *uaddr_len = sizeof(*sin);
 436         return(0);
 437 }
 438
 439 EXPORT_SYMBOL(inet6_getname);
 440
 441 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 442 {
 443         struct sock *sk = sock->sk;
 444         struct net *net = sock_net(sk);
 445
 446         switch(cmd)
 447         {
 448         case SIOCGSTAMP:
 449                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 450
 451         case SIOCGSTAMPNS:
 452                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 453
 454         case SIOCADDRT:
 455         case SIOCDELRT:
 456
 457                 return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
 458
 459         case SIOCSIFADDR:
 460                 return addrconf_add_ifaddr(net, (void __user *) arg);
 461         case SIOCDIFADDR:
 462                 return addrconf_del_ifaddr(net, (void __user *) arg);
 463         case SIOCSIFDSTADDR:
 464                 return addrconf_set_dstaddr(net, (void __user *) arg);
 465         default:
 466                 if (!sk->sk_prot->ioctl)
 467                         return -ENOIOCTLCMD;
 468                 return sk->sk_prot->ioctl(sk, cmd, arg);
 469         }
 470         /*NOTREACHED*/
 471         return(0);
 472 }
 473
 474 EXPORT_SYMBOL(inet6_ioctl);
 475
 476 const struct proto_ops inet6_stream_ops = {
 477         .family            = PF_INET6,
 478         .owner             = THIS_MODULE,
 479         .release           = inet6_release,
 480         .bind              = inet6_bind,
 481         .connect           = inet_stream_connect,       /* ok           */
 482         .socketpair        = sock_no_socketpair,        /* a do nothing */
 483         .accept            = inet_accept,               /* ok           */
 484         .getname           = inet6_getname,
 485         .poll              = tcp_poll,                  /* ok           */
 486         .ioctl             = inet6_ioctl,               /* must change  */
 487         .listen            = inet_listen,               /* ok           */
 488         .shutdown          = inet_shutdown,             /* ok           */
 489         .setsockopt        = sock_common_setsockopt,    /* ok           */
 490         .getsockopt        = sock_common_getsockopt,    /* ok           */
 491         .sendmsg           = tcp_sendmsg,               /* ok           */
 492         .recvmsg           = sock_common_recvmsg,       /* ok           */
 493         .mmap              = sock_no_mmap,
 494         .sendpage          = tcp_sendpage,
 495         .splice_read       = tcp_splice_read,
 496 #ifdef CONFIG_COMPAT
 497         .compat_setsockopt = compat_sock_common_setsockopt,
 498         .compat_getsockopt = compat_sock_common_getsockopt,
 499 #endif
 500 };
 501
 502 const struct proto_ops inet6_dgram_ops = {
 503         .family            = PF_INET6,
 504         .owner             = THIS_MODULE,
 505         .release           = inet6_release,
 506         .bind              = inet6_bind,
 507         .connect           = inet_dgram_connect,        /* ok           */
 508         .socketpair        = sock_no_socketpair,        /* a do nothing */
 509         .accept            = sock_no_accept,            /* a do nothing */
 510         .getname           = inet6_getname,
 511         .poll              = udp_poll,                  /* ok           */
 512         .ioctl             = inet6_ioctl,               /* must change  */
 513         .listen            = sock_no_listen,            /* ok           */
 514         .shutdown          = inet_shutdown,             /* ok           */
 515         .setsockopt        = sock_common_setsockopt,    /* ok           */
 516         .getsockopt        = sock_common_getsockopt,    /* ok           */
 517         .sendmsg           = inet_sendmsg,              /* ok           */
 518         .recvmsg           = sock_common_recvmsg,       /* ok           */
 519         .mmap              = sock_no_mmap,
 520         .sendpage          = sock_no_sendpage,
 521 #ifdef CONFIG_COMPAT
 522         .compat_setsockopt = compat_sock_common_setsockopt,
 523         .compat_getsockopt = compat_sock_common_getsockopt,
 524 #endif
 525 };
 526
 527 static struct net_proto_family inet6_family_ops = {
 528         .family = PF_INET6,
 529         .create = inet6_create,
 530         .owner  = THIS_MODULE,
 531 };
 532
 533 int inet6_register_protosw(struct inet_protosw *p)
 534 {
 535         struct list_head *lh;
 536         struct inet_protosw *answer;
 537         struct list_head *last_perm;
 538         int protocol = p->protocol;
 539         int ret;
 540
 541         spin_lock_bh(&inetsw6_lock);
 542
 543         ret = -EINVAL;
 544         if (p->type >= SOCK_MAX)
 545                 goto out_illegal;
 546
 547         /* If we are trying to override a permanent protocol, bail. */
 548         answer = NULL;
 549         ret = -EPERM;
 550         last_perm = &inetsw6[p->type];
 551         list_for_each(lh, &inetsw6[p->type]) {
 552                 answer = list_entry(lh, struct inet_protosw, list);
 553
 554                 /* Check only the non-wild match. */
 555                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 556                         if (protocol == answer->protocol)
 557                                 break;
 558                         last_perm = lh;
 559                 }
 560
 561                 answer = NULL;
 562         }
 563         if (answer)
 564                 goto out_permanent;
 565
 566         /* Add the new entry after the last permanent entry if any, so that
 567          * the new entry does not override a permanent entry when matched with
 568          * a wild-card protocol. But it is allowed to override any existing
 569          * non-permanent entry.  This means that when we remove this entry, the
 570          * system automatically returns to the old behavior.
 571          */
 572         list_add_rcu(&p->list, last_perm);
 573         ret = 0;
 574 out:
 575         spin_unlock_bh(&inetsw6_lock);
 576         return ret;
 577
 578 out_permanent:
 579         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 580                protocol);
 581         goto out;
 582
 583 out_illegal:
 584         printk(KERN_ERR
 585                "Ignoring attempt to register invalid socket type %d.\n",
 586                p->type);
 587         goto out;
 588 }
 589
 590 EXPORT_SYMBOL(inet6_register_protosw);
 591
 592 void
 593 inet6_unregister_protosw(struct inet_protosw *p)
 594 {
 595         if (INET_PROTOSW_PERMANENT & p->flags) {
 596                 printk(KERN_ERR
 597                        "Attempt to unregister permanent protocol %d.\n",
 598                        p->protocol);
 599         } else {
 600                 spin_lock_bh(&inetsw6_lock);
 601                 list_del_rcu(&p->list);
 602                 spin_unlock_bh(&inetsw6_lock);
 603
 604                 synchronize_net();
 605         }
 606 }
 607
 608 EXPORT_SYMBOL(inet6_unregister_protosw);
 609
 610 int inet6_sk_rebuild_header(struct sock *sk)
 611 {
 612         int err;
 613         struct dst_entry *dst;
 614         struct ipv6_pinfo *np = inet6_sk(sk);
 615
 616         dst = __sk_dst_check(sk, np->dst_cookie);
 617
 618         if (dst == NULL) {
 619                 struct inet_sock *inet = inet_sk(sk);
 620                 struct in6_addr *final_p = NULL, final;
 621                 struct flowi fl;
 622
 623                 memset(&fl, 0, sizeof(fl));
 624                 fl.proto = sk->sk_protocol;
 625                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 626                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 627                 fl.fl6_flowlabel = np->flow_label;
 628                 fl.oif = sk->sk_bound_dev_if;
 629                 fl.fl_ip_dport = inet->dport;
 630                 fl.fl_ip_sport = inet->sport;
 631                 security_sk_classify_flow(sk, &fl);
 632
 633                 if (np->opt && np->opt->srcrt) {
 634                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 635                         ipv6_addr_copy(&final, &fl.fl6_dst);
 636                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 637                         final_p = &final;
 638                 }
 639
 640                 err = ip6_dst_lookup(sk, &dst, &fl);
 641                 if (err) {
 642                         sk->sk_route_caps = 0;
 643                         return err;
 644                 }
 645                 if (final_p)
 646                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 647
 648                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
 649                         sk->sk_err_soft = -err;
 650                         return err;
 651                 }
 652
 653                 __ip6_dst_store(sk, dst, NULL, NULL);
 654         }
 655
 656         return 0;
 657 }
 658
 659 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 660
 661 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 662 {
 663         struct ipv6_pinfo *np = inet6_sk(sk);
 664         struct inet6_skb_parm *opt = IP6CB(skb);
 665
 666         if (np->rxopt.all) {
 667                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 668                                   np->rxopt.bits.ohopopts)) ||
 669                     ((IPV6_FLOWINFO_MASK &
 670                       *(__be32 *)skb_network_header(skb)) &&
 671                      np->rxopt.bits.rxflow) ||
 672                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 673                      np->rxopt.bits.osrcrt)) ||
 674                     ((opt->dst1 || opt->dst0) &&
 675                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 676                         return 1;
 677         }
 678         return 0;
 679 }
 680
 681 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 682
 683 static struct inet6_protocol *ipv6_gso_pull_exthdrs(struct sk_buff *skb,
 684                                                     int proto)
 685 {
 686         struct inet6_protocol *ops = NULL;
 687
 688         for (;;) {
 689                 struct ipv6_opt_hdr *opth;
 690                 int len;
 691
 692                 if (proto != NEXTHDR_HOP) {
 693                         ops = rcu_dereference(inet6_protos[proto]);
 694
 695                         if (unlikely(!ops))
 696                                 break;
 697
 698                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
 699                                 break;
 700                 }
 701
 702                 if (unlikely(!pskb_may_pull(skb, 8)))
 703                         break;
 704
 705                 opth = (void *)skb->data;
 706                 len = ipv6_optlen(opth);
 707
 708                 if (unlikely(!pskb_may_pull(skb, len)))
 709                         break;
 710
 711                 proto = opth->nexthdr;
 712                 __skb_pull(skb, len);
 713         }
 714
 715         return ops;
 716 }
 717
 718 static int ipv6_gso_send_check(struct sk_buff *skb)
 719 {
 720         struct ipv6hdr *ipv6h;
 721         struct inet6_protocol *ops;
 722         int err = -EINVAL;
 723
 724         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 725                 goto out;
 726
 727         ipv6h = ipv6_hdr(skb);
 728         __skb_pull(skb, sizeof(*ipv6h));
 729         err = -EPROTONOSUPPORT;
 730
 731         rcu_read_lock();
 732         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
 733         if (likely(ops && ops->gso_send_check)) {
 734                 skb_reset_transport_header(skb);
 735                 err = ops->gso_send_check(skb);
 736         }
 737         rcu_read_unlock();
 738
 739 out:
 740         return err;
 741 }
 742
 743 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
 744 {
 745         struct sk_buff *segs = ERR_PTR(-EINVAL);
 746         struct ipv6hdr *ipv6h;
 747         struct inet6_protocol *ops;
 748
 749         if (!(features & NETIF_F_V6_CSUM))
 750                 features &= ~NETIF_F_SG;
 751
 752         if (unlikely(skb_shinfo(skb)->gso_type &
 753                      ~(SKB_GSO_UDP |
 754                        SKB_GSO_DODGY |
 755                        SKB_GSO_TCP_ECN |
 756                        SKB_GSO_TCPV6 |
 757                        0)))
 758                 goto out;
 759
 760         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 761                 goto out;
 762
 763         ipv6h = ipv6_hdr(skb);
 764         __skb_pull(skb, sizeof(*ipv6h));
 765         segs = ERR_PTR(-EPROTONOSUPPORT);
 766
 767         rcu_read_lock();
 768         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
 769         if (likely(ops && ops->gso_segment)) {
 770                 skb_reset_transport_header(skb);
 771                 segs = ops->gso_segment(skb, features);
 772         }
 773         rcu_read_unlock();
 774
 775         if (unlikely(IS_ERR(segs)))
 776                 goto out;
 777
 778         for (skb = segs; skb; skb = skb->next) {
 779                 ipv6h = ipv6_hdr(skb);
 780                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
 781                                            sizeof(*ipv6h));
 782         }
 783
 784 out:
 785         return segs;
 786 }
 787
 788 static struct packet_type ipv6_packet_type = {
 789         .type = __constant_htons(ETH_P_IPV6),
 790         .func = ipv6_rcv,
 791         .gso_send_check = ipv6_gso_send_check,
 792         .gso_segment = ipv6_gso_segment,
 793 };
 794
 795 static int __init ipv6_packet_init(void)
 796 {
 797         dev_add_pack(&ipv6_packet_type);
 798         return 0;
 799 }
 800
 801 static void ipv6_packet_cleanup(void)
 802 {
 803         dev_remove_pack(&ipv6_packet_type);
 804 }
 805
 806 static int __init init_ipv6_mibs(void)
 807 {
 808         if (snmp_mib_init((void **)ipv6_statistics,
 809                           sizeof(struct ipstats_mib)) < 0)
 810                 goto err_ip_mib;
 811         if (snmp_mib_init((void **)icmpv6_statistics,
 812                           sizeof(struct icmpv6_mib)) < 0)
 813                 goto err_icmp_mib;
 814         if (snmp_mib_init((void **)icmpv6msg_statistics,
 815                           sizeof(struct icmpv6msg_mib)) < 0)
 816                 goto err_icmpmsg_mib;
 817         if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib)) < 0)
 818                 goto err_udp_mib;
 819         if (snmp_mib_init((void **)udplite_stats_in6,
 820                           sizeof (struct udp_mib)) < 0)
 821                 goto err_udplite_mib;
 822         return 0;
 823
 824 err_udplite_mib:
 825         snmp_mib_free((void **)udp_stats_in6);
 826 err_udp_mib:
 827         snmp_mib_free((void **)icmpv6msg_statistics);
 828 err_icmpmsg_mib:
 829         snmp_mib_free((void **)icmpv6_statistics);
 830 err_icmp_mib:
 831         snmp_mib_free((void **)ipv6_statistics);
 832 err_ip_mib:
 833         return -ENOMEM;
 834
 835 }
 836
 837 static void cleanup_ipv6_mibs(void)
 838 {
 839         snmp_mib_free((void **)ipv6_statistics);
 840         snmp_mib_free((void **)icmpv6_statistics);
 841         snmp_mib_free((void **)icmpv6msg_statistics);
 842         snmp_mib_free((void **)udp_stats_in6);
 843         snmp_mib_free((void **)udplite_stats_in6);
 844 }
 845
 846 static int inet6_net_init(struct net *net)
 847 {
 848         int err = 0;
 849
 850         net->ipv6.sysctl.bindv6only = 0;
 851         net->ipv6.sysctl.flush_delay = 0;
 852         net->ipv6.sysctl.ip6_rt_max_size = 4096;
 853         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
 854         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
 855         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
 856         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
 857         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
 858         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
 859         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 860
 861 #ifdef CONFIG_PROC_FS
 862         err = udp6_proc_init(net);
 863         if (err)
 864                 goto out;
 865         err = tcp6_proc_init(net);
 866         if (err)
 867                 goto proc_tcp6_fail;
 868         err = ac6_proc_init(net);
 869         if (err)
 870                 goto proc_ac6_fail;
 871 out:
 872 #endif
 873         return err;
 874
 875 #ifdef CONFIG_PROC_FS
 876 proc_ac6_fail:
 877         tcp6_proc_exit(net);
 878 proc_tcp6_fail:
 879         udp6_proc_exit(net);
 880         goto out;
 881 #endif
 882 }
 883
 884 static void inet6_net_exit(struct net *net)
 885 {
 886 #ifdef CONFIG_PROC_FS
 887         udp6_proc_exit(net);
 888         tcp6_proc_exit(net);
 889         ac6_proc_exit(net);
 890 #endif
 891 }
 892
 893 static struct pernet_operations inet6_net_ops = {
 894         .init = inet6_net_init,
 895         .exit = inet6_net_exit,
 896 };
 897
 898 static int __init inet6_init(void)
 899 {
 900         struct sk_buff *dummy_skb;
 901         struct list_head *r;
 902         int err;
 903
 904         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
 905
 906         err = proto_register(&tcpv6_prot, 1);
 907         if (err)
 908                 goto out;
 909
 910         err = proto_register(&udpv6_prot, 1);
 911         if (err)
 912                 goto out_unregister_tcp_proto;
 913
 914         err = proto_register(&udplitev6_prot, 1);
 915         if (err)
 916                 goto out_unregister_udp_proto;
 917
 918         err = proto_register(&rawv6_prot, 1);
 919         if (err)
 920                 goto out_unregister_udplite_proto;
 921
 922
 923         /* Register the socket-side information for inet6_create.  */
 924         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 925                 INIT_LIST_HEAD(r);
 926
 927         /* We MUST register RAW sockets before we create the ICMP6,
 928          * IGMP6, or NDISC control sockets.
 929          */
 930         err = rawv6_init();
 931         if (err)
 932                 goto out_unregister_raw_proto;
 933
 934         /* Register the family here so that the init calls below will
 935          * be able to create sockets. (?? is this dangerous ??)
 936          */
 937         err = sock_register(&inet6_family_ops);
 938         if (err)
 939                 goto out_sock_register_fail;
 940
 941         /* Initialise ipv6 mibs */
 942         err = init_ipv6_mibs();
 943         if (err)
 944                 goto out_unregister_sock;
 945
 946         /*
 947          *      ipngwg API draft makes clear that the correct semantics
 948          *      for TCP and UDP is to consider one TCP and UDP instance
 949          *      in a host availiable by both INET and INET6 APIs and
 950          *      able to communicate via both network protocols.
 951          */
 952
 953         err = register_pernet_subsys(&inet6_net_ops);
 954         if (err)
 955                 goto register_pernet_fail;
 956         err = icmpv6_init();
 957         if (err)
 958                 goto icmp_fail;
 959 #ifdef CONFIG_IPV6_MROUTE
 960         ip6_mr_init();
 961 #endif
 962         err = ndisc_init();
 963         if (err)
 964                 goto ndisc_fail;
 965         err = igmp6_init();
 966         if (err)
 967                 goto igmp_fail;
 968         err = ipv6_netfilter_init();
 969         if (err)
 970                 goto netfilter_fail;
 971         /* Create /proc/foo6 entries. */
 972 #ifdef CONFIG_PROC_FS
 973         err = -ENOMEM;
 974         if (raw6_proc_init())
 975                 goto proc_raw6_fail;
 976         if (udplite6_proc_init())
 977                 goto proc_udplite6_fail;
 978         if (ipv6_misc_proc_init())
 979                 goto proc_misc6_fail;
 980         if (if6_proc_init())
 981                 goto proc_if6_fail;
 982 #endif
 983         err = ip6_route_init();
 984         if (err)
 985                 goto ip6_route_fail;
 986         err = ip6_flowlabel_init();
 987         if (err)
 988                 goto ip6_flowlabel_fail;
 989         err = addrconf_init();
 990         if (err)
 991                 goto addrconf_fail;
 992
 993         /* Init v6 extension headers. */
 994         err = ipv6_exthdrs_init();
 995         if (err)
 996                 goto ipv6_exthdrs_fail;
 997
 998         err = ipv6_frag_init();
 999         if (err)
1000                 goto ipv6_frag_fail;
1001
1002         /* Init v6 transport protocols. */
1003         err = udpv6_init();
1004         if (err)
1005                 goto udpv6_fail;
1006
1007         err = udplitev6_init();
1008         if (err)
1009                 goto udplitev6_fail;
1010
1011         err = tcpv6_init();
1012         if (err)
1013                 goto tcpv6_fail;
1014
1015         err = ipv6_packet_init();
1016         if (err)
1017                 goto ipv6_packet_fail;
1018
1019 #ifdef CONFIG_SYSCTL
1020         err = ipv6_sysctl_register();
1021         if (err)
1022                 goto sysctl_fail;
1023 #endif
1024 out:
1025         return err;
1026
1027 #ifdef CONFIG_SYSCTL
1028 sysctl_fail:
1029         ipv6_packet_cleanup();
1030 #endif
1031 ipv6_packet_fail:
1032         tcpv6_exit();
1033 tcpv6_fail:
1034         udplitev6_exit();
1035 udplitev6_fail:
1036         udpv6_exit();
1037 udpv6_fail:
1038         ipv6_frag_exit();
1039 ipv6_frag_fail:
1040         ipv6_exthdrs_exit();
1041 ipv6_exthdrs_fail:
1042         addrconf_cleanup();
1043 addrconf_fail:
1044         ip6_flowlabel_cleanup();
1045 ip6_flowlabel_fail:
1046         ip6_route_cleanup();
1047 ip6_route_fail:
1048 #ifdef CONFIG_PROC_FS
1049         if6_proc_exit();
1050 proc_if6_fail:
1051         ipv6_misc_proc_exit();
1052 proc_misc6_fail:
1053         udplite6_proc_exit();
1054 proc_udplite6_fail:
1055         raw6_proc_exit();
1056 proc_raw6_fail:
1057 #endif
1058         ipv6_netfilter_fini();
1059 netfilter_fail:
1060         igmp6_cleanup();
1061 igmp_fail:
1062         ndisc_cleanup();
1063 ndisc_fail:
1064         icmpv6_cleanup();
1065 icmp_fail:
1066         unregister_pernet_subsys(&inet6_net_ops);
1067 register_pernet_fail:
1068         cleanup_ipv6_mibs();
1069 out_unregister_sock:
1070         sock_unregister(PF_INET6);
1071         rtnl_unregister_all(PF_INET6);
1072 out_sock_register_fail:
1073         rawv6_exit();
1074 out_unregister_raw_proto:
1075         proto_unregister(&rawv6_prot);
1076 out_unregister_udplite_proto:
1077         proto_unregister(&udplitev6_prot);
1078 out_unregister_udp_proto:
1079         proto_unregister(&udpv6_prot);
1080 out_unregister_tcp_proto:
1081         proto_unregister(&tcpv6_prot);
1082         goto out;
1083 }
1084 module_init(inet6_init);
1085
1086 static void __exit inet6_exit(void)
1087 {
1088         /* First of all disallow new sockets creation. */
1089         sock_unregister(PF_INET6);
1090         /* Disallow any further netlink messages */
1091         rtnl_unregister_all(PF_INET6);
1092
1093 #ifdef CONFIG_SYSCTL
1094         ipv6_sysctl_unregister();
1095 #endif
1096         udpv6_exit();
1097         udplitev6_exit();
1098         tcpv6_exit();
1099
1100         /* Cleanup code parts. */
1101         ipv6_packet_cleanup();
1102         ipv6_frag_exit();
1103         ipv6_exthdrs_exit();
1104         addrconf_cleanup();
1105         ip6_flowlabel_cleanup();
1106         ip6_route_cleanup();
1107 #ifdef CONFIG_PROC_FS
1108
1109         /* Cleanup code parts. */
1110         if6_proc_exit();
1111         ipv6_misc_proc_exit();
1112         udplite6_proc_exit();
1113         raw6_proc_exit();
1114 #endif
1115         ipv6_netfilter_fini();
1116         igmp6_cleanup();
1117         ndisc_cleanup();
1118         icmpv6_cleanup();
1119         rawv6_exit();
1120
1121         unregister_pernet_subsys(&inet6_net_ops);
1122         cleanup_ipv6_mibs();
1123         proto_unregister(&rawv6_prot);
1124         proto_unregister(&udplitev6_prot);
1125         proto_unregister(&udpv6_prot);
1126         proto_unregister(&tcpv6_prot);
1127 }
1128 module_exit(inet6_exit);
1129
1130 MODULE_ALIAS_NETPROTO(PF_INET6);