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 struct ipv6_params ipv6_defaults = {
  76         .disable_ipv6 = 0,
  77         .autoconf = 1,
  78 };
  79
  80 static int disable_ipv6_mod = 0;
  81
  82 module_param_named(disable, disable_ipv6_mod, int, 0444);
  83 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  84
  85 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  86 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  87
  88 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  89 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  90
  91 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  92 {
  93         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  94
  95         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  96 }
  97
  98 static int inet6_create(struct net *net, struct socket *sock, int protocol)
  99 {
 100         struct inet_sock *inet;
 101         struct ipv6_pinfo *np;
 102         struct sock *sk;
 103         struct inet_protosw *answer;
 104         struct proto *answer_prot;
 105         unsigned char answer_flags;
 106         char answer_no_check;
 107         int try_loading_module = 0;
 108         int err;
 109
 110         if (sock->type != SOCK_RAW &&
 111             sock->type != SOCK_DGRAM &&
 112             !inet_ehash_secret)
 113                 build_ehash_secret();
 114
 115         /* Look for the requested type/protocol pair. */
 116 lookup_protocol:
 117         err = -ESOCKTNOSUPPORT;
 118         rcu_read_lock();
 119         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 120
 121                 err = 0;
 122                 /* Check the non-wild match. */
 123                 if (protocol == answer->protocol) {
 124                         if (protocol != IPPROTO_IP)
 125                                 break;
 126                 } else {
 127                         /* Check for the two wild cases. */
 128                         if (IPPROTO_IP == protocol) {
 129                                 protocol = answer->protocol;
 130                                 break;
 131                         }
 132                         if (IPPROTO_IP == answer->protocol)
 133                                 break;
 134                 }
 135                 err = -EPROTONOSUPPORT;
 136         }
 137
 138         if (err) {
 139                 if (try_loading_module < 2) {
 140                         rcu_read_unlock();
 141                         /*
 142                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 143                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 144                          */
 145                         if (++try_loading_module == 1)
 146                                 request_module("net-pf-%d-proto-%d-type-%d",
 147                                                 PF_INET6, protocol, sock->type);
 148                         /*
 149                          * Fall back to generic, e.g. net-pf-10-proto-132
 150                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 151                          */
 152                         else
 153                                 request_module("net-pf-%d-proto-%d",
 154                                                 PF_INET6, protocol);
 155                         goto lookup_protocol;
 156                 } else
 157                         goto out_rcu_unlock;
 158         }
 159
 160         err = -EPERM;
 161         if (answer->capability > 0 && !capable(answer->capability))
 162                 goto out_rcu_unlock;
 163
 164         sock->ops = answer->ops;
 165         answer_prot = answer->prot;
 166         answer_no_check = answer->no_check;
 167         answer_flags = answer->flags;
 168         rcu_read_unlock();
 169
 170         WARN_ON(answer_prot->slab == NULL);
 171
 172         err = -ENOBUFS;
 173         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 174         if (sk == NULL)
 175                 goto out;
 176
 177         sock_init_data(sock, sk);
 178
 179         err = 0;
 180         sk->sk_no_check = answer_no_check;
 181         if (INET_PROTOSW_REUSE & answer_flags)
 182                 sk->sk_reuse = 1;
 183
 184         inet = inet_sk(sk);
 185         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 186
 187         if (SOCK_RAW == sock->type) {
 188                 inet->num = protocol;
 189                 if (IPPROTO_RAW == protocol)
 190                         inet->hdrincl = 1;
 191         }
 192
 193         sk->sk_destruct         = inet_sock_destruct;
 194         sk->sk_family           = PF_INET6;
 195         sk->sk_protocol         = protocol;
 196
 197         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 198
 199         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 200         np->hop_limit   = -1;
 201         np->mcast_hops  = -1;
 202         np->mc_loop     = 1;
 203         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 204         np->ipv6only    = net->ipv6.sysctl.bindv6only;
 205
 206         /* Init the ipv4 part of the socket since we can have sockets
 207          * using v6 API for ipv4.
 208          */
 209         inet->uc_ttl    = -1;
 210
 211         inet->mc_loop   = 1;
 212         inet->mc_ttl    = 1;
 213         inet->mc_index  = 0;
 214         inet->mc_list   = NULL;
 215
 216         if (ipv4_config.no_pmtu_disc)
 217                 inet->pmtudisc = IP_PMTUDISC_DONT;
 218         else
 219                 inet->pmtudisc = IP_PMTUDISC_WANT;
 220         /*
 221          * Increment only the relevant sk_prot->socks debug field, this changes
 222          * the previous behaviour of incrementing both the equivalent to
 223          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 224          *
 225          * This allows better debug granularity as we'll know exactly how many
 226          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 227          * transport protocol socks. -acme
 228          */
 229         sk_refcnt_debug_inc(sk);
 230
 231         if (inet->num) {
 232                 /* It assumes that any protocol which allows
 233                  * the user to assign a number at socket
 234                  * creation time automatically shares.
 235                  */
 236                 inet->sport = htons(inet->num);
 237                 sk->sk_prot->hash(sk);
 238         }
 239         if (sk->sk_prot->init) {
 240                 err = sk->sk_prot->init(sk);
 241                 if (err) {
 242                         sk_common_release(sk);
 243                         goto out;
 244                 }
 245         }
 246 out:
 247         return err;
 248 out_rcu_unlock:
 249         rcu_read_unlock();
 250         goto out;
 251 }
 252
 253
 254 /* bind for INET6 API */
 255 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 256 {
 257         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 258         struct sock *sk = sock->sk;
 259         struct inet_sock *inet = inet_sk(sk);
 260         struct ipv6_pinfo *np = inet6_sk(sk);
 261         struct net *net = sock_net(sk);
 262         __be32 v4addr = 0;
 263         unsigned short snum;
 264         int addr_type = 0;
 265         int err = 0;
 266
 267         /* If the socket has its own bind function then use it. */
 268         if (sk->sk_prot->bind)
 269                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 270
 271         if (addr_len < SIN6_LEN_RFC2133)
 272                 return -EINVAL;
 273         addr_type = ipv6_addr_type(&addr->sin6_addr);
 274         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 275                 return -EINVAL;
 276
 277         snum = ntohs(addr->sin6_port);
 278         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 279                 return -EACCES;
 280
 281         lock_sock(sk);
 282
 283         /* Check these errors (active socket, double bind). */
 284         if (sk->sk_state != TCP_CLOSE || inet->num) {
 285                 err = -EINVAL;
 286                 goto out;
 287         }
 288
 289         /* Check if the address belongs to the host. */
 290         if (addr_type == IPV6_ADDR_MAPPED) {
 291                 int chk_addr_ret;
 292
 293                 /* Binding to v4-mapped address on a v6-only socket
 294                  * makes no sense
 295                  */
 296                 if (np->ipv6only) {
 297                         err = -EINVAL;
 298                         goto out;
 299                 }
 300
 301                 /* Reproduce AF_INET checks to make the bindings consitant */
 302                 v4addr = addr->sin6_addr.s6_addr32[3];
 303                 chk_addr_ret = inet_addr_type(net, v4addr);
 304                 if (!sysctl_ip_nonlocal_bind &&
 305                     !(inet->freebind || inet->transparent) &&
 306                     v4addr != htonl(INADDR_ANY) &&
 307                     chk_addr_ret != RTN_LOCAL &&
 308                     chk_addr_ret != RTN_MULTICAST &&
 309                     chk_addr_ret != RTN_BROADCAST)
 310                         goto out;
 311         } else {
 312                 if (addr_type != IPV6_ADDR_ANY) {
 313                         struct net_device *dev = NULL;
 314
 315                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 316                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 317                                     addr->sin6_scope_id) {
 318                                         /* Override any existing binding, if another one
 319                                          * is supplied by user.
 320                                          */
 321                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 322                                 }
 323
 324                                 /* Binding to link-local address requires an interface */
 325                                 if (!sk->sk_bound_dev_if) {
 326                                         err = -EINVAL;
 327                                         goto out;
 328                                 }
 329                                 dev = dev_get_by_index(net, sk->sk_bound_dev_if);
 330                                 if (!dev) {
 331                                         err = -ENODEV;
 332                                         goto out;
 333                                 }
 334                         }
 335
 336                         /* ipv4 addr of the socket is invalid.  Only the
 337                          * unspecified and mapped address have a v4 equivalent.
 338                          */
 339                         v4addr = LOOPBACK4_IPV6;
 340                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 341                                 if (!ipv6_chk_addr(net, &addr->sin6_addr,
 342                                                    dev, 0)) {
 343                                         if (dev)
 344                                                 dev_put(dev);
 345                                         err = -EADDRNOTAVAIL;
 346                                         goto out;
 347                                 }
 348                         }
 349                         if (dev)
 350                                 dev_put(dev);
 351                 }
 352         }
 353
 354         inet->rcv_saddr = v4addr;
 355         inet->saddr = v4addr;
 356
 357         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 358
 359         if (!(addr_type & IPV6_ADDR_MULTICAST))
 360                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 361
 362         /* Make sure we are allowed to bind here. */
 363         if (sk->sk_prot->get_port(sk, snum)) {
 364                 inet_reset_saddr(sk);
 365                 err = -EADDRINUSE;
 366                 goto out;
 367         }
 368
 369         if (addr_type != IPV6_ADDR_ANY) {
 370                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 371                 if (addr_type != IPV6_ADDR_MAPPED)
 372                         np->ipv6only = 1;
 373         }
 374         if (snum)
 375                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 376         inet->sport = htons(inet->num);
 377         inet->dport = 0;
 378         inet->daddr = 0;
 379 out:
 380         release_sock(sk);
 381         return err;
 382 }
 383
 384 EXPORT_SYMBOL(inet6_bind);
 385
 386 int inet6_release(struct socket *sock)
 387 {
 388         struct sock *sk = sock->sk;
 389
 390         if (sk == NULL)
 391                 return -EINVAL;
 392
 393         /* Free mc lists */
 394         ipv6_sock_mc_close(sk);
 395
 396         /* Free ac lists */
 397         ipv6_sock_ac_close(sk);
 398
 399         return inet_release(sock);
 400 }
 401
 402 EXPORT_SYMBOL(inet6_release);
 403
 404 void inet6_destroy_sock(struct sock *sk)
 405 {
 406         struct ipv6_pinfo *np = inet6_sk(sk);
 407         struct sk_buff *skb;
 408         struct ipv6_txoptions *opt;
 409
 410         /* Release rx options */
 411
 412         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 413                 kfree_skb(skb);
 414
 415         /* Free flowlabels */
 416         fl6_free_socklist(sk);
 417
 418         /* Free tx options */
 419
 420         if ((opt = xchg(&np->opt, NULL)) != NULL)
 421                 sock_kfree_s(sk, opt, opt->tot_len);
 422 }
 423
 424 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 425
 426 /*
 427  *      This does both peername and sockname.
 428  */
 429
 430 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 431                  int *uaddr_len, int peer)
 432 {
 433         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 434         struct sock *sk = sock->sk;
 435         struct inet_sock *inet = inet_sk(sk);
 436         struct ipv6_pinfo *np = inet6_sk(sk);
 437
 438         sin->sin6_family = AF_INET6;
 439         sin->sin6_flowinfo = 0;
 440         sin->sin6_scope_id = 0;
 441         if (peer) {
 442                 if (!inet->dport)
 443                         return -ENOTCONN;
 444                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 445                     peer == 1)
 446                         return -ENOTCONN;
 447                 sin->sin6_port = inet->dport;
 448                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 449                 if (np->sndflow)
 450                         sin->sin6_flowinfo = np->flow_label;
 451         } else {
 452                 if (ipv6_addr_any(&np->rcv_saddr))
 453                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 454                 else
 455                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 456
 457                 sin->sin6_port = inet->sport;
 458         }
 459         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 460                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 461         *uaddr_len = sizeof(*sin);
 462         return(0);
 463 }
 464
 465 EXPORT_SYMBOL(inet6_getname);
 466
 467 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 468 {
 469         struct sock *sk = sock->sk;
 470         struct net *net = sock_net(sk);
 471
 472         switch(cmd)
 473         {
 474         case SIOCGSTAMP:
 475                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 476
 477         case SIOCGSTAMPNS:
 478                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 479
 480         case SIOCADDRT:
 481         case SIOCDELRT:
 482
 483                 return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
 484
 485         case SIOCSIFADDR:
 486                 return addrconf_add_ifaddr(net, (void __user *) arg);
 487         case SIOCDIFADDR:
 488                 return addrconf_del_ifaddr(net, (void __user *) arg);
 489         case SIOCSIFDSTADDR:
 490                 return addrconf_set_dstaddr(net, (void __user *) arg);
 491         default:
 492                 if (!sk->sk_prot->ioctl)
 493                         return -ENOIOCTLCMD;
 494                 return sk->sk_prot->ioctl(sk, cmd, arg);
 495         }
 496         /*NOTREACHED*/
 497         return(0);
 498 }
 499
 500 EXPORT_SYMBOL(inet6_ioctl);
 501
 502 const struct proto_ops inet6_stream_ops = {
 503         .family            = PF_INET6,
 504         .owner             = THIS_MODULE,
 505         .release           = inet6_release,
 506         .bind              = inet6_bind,
 507         .connect           = inet_stream_connect,       /* ok           */
 508         .socketpair        = sock_no_socketpair,        /* a do nothing */
 509         .accept            = inet_accept,               /* ok           */
 510         .getname           = inet6_getname,
 511         .poll              = tcp_poll,                  /* ok           */
 512         .ioctl             = inet6_ioctl,               /* must change  */
 513         .listen            = inet_listen,               /* ok           */
 514         .shutdown          = inet_shutdown,             /* ok           */
 515         .setsockopt        = sock_common_setsockopt,    /* ok           */
 516         .getsockopt        = sock_common_getsockopt,    /* ok           */
 517         .sendmsg           = tcp_sendmsg,               /* ok           */
 518         .recvmsg           = sock_common_recvmsg,       /* ok           */
 519         .mmap              = sock_no_mmap,
 520         .sendpage          = tcp_sendpage,
 521         .splice_read       = tcp_splice_read,
 522 #ifdef CONFIG_COMPAT
 523         .compat_setsockopt = compat_sock_common_setsockopt,
 524         .compat_getsockopt = compat_sock_common_getsockopt,
 525 #endif
 526 };
 527
 528 const struct proto_ops inet6_dgram_ops = {
 529         .family            = PF_INET6,
 530         .owner             = THIS_MODULE,
 531         .release           = inet6_release,
 532         .bind              = inet6_bind,
 533         .connect           = inet_dgram_connect,        /* ok           */
 534         .socketpair        = sock_no_socketpair,        /* a do nothing */
 535         .accept            = sock_no_accept,            /* a do nothing */
 536         .getname           = inet6_getname,
 537         .poll              = udp_poll,                  /* ok           */
 538         .ioctl             = inet6_ioctl,               /* must change  */
 539         .listen            = sock_no_listen,            /* ok           */
 540         .shutdown          = inet_shutdown,             /* ok           */
 541         .setsockopt        = sock_common_setsockopt,    /* ok           */
 542         .getsockopt        = sock_common_getsockopt,    /* ok           */
 543         .sendmsg           = inet_sendmsg,              /* ok           */
 544         .recvmsg           = sock_common_recvmsg,       /* ok           */
 545         .mmap              = sock_no_mmap,
 546         .sendpage          = sock_no_sendpage,
 547 #ifdef CONFIG_COMPAT
 548         .compat_setsockopt = compat_sock_common_setsockopt,
 549         .compat_getsockopt = compat_sock_common_getsockopt,
 550 #endif
 551 };
 552
 553 static struct net_proto_family inet6_family_ops = {
 554         .family = PF_INET6,
 555         .create = inet6_create,
 556         .owner  = THIS_MODULE,
 557 };
 558
 559 int inet6_register_protosw(struct inet_protosw *p)
 560 {
 561         struct list_head *lh;
 562         struct inet_protosw *answer;
 563         struct list_head *last_perm;
 564         int protocol = p->protocol;
 565         int ret;
 566
 567         spin_lock_bh(&inetsw6_lock);
 568
 569         ret = -EINVAL;
 570         if (p->type >= SOCK_MAX)
 571                 goto out_illegal;
 572
 573         /* If we are trying to override a permanent protocol, bail. */
 574         answer = NULL;
 575         ret = -EPERM;
 576         last_perm = &inetsw6[p->type];
 577         list_for_each(lh, &inetsw6[p->type]) {
 578                 answer = list_entry(lh, struct inet_protosw, list);
 579
 580                 /* Check only the non-wild match. */
 581                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 582                         if (protocol == answer->protocol)
 583                                 break;
 584                         last_perm = lh;
 585                 }
 586
 587                 answer = NULL;
 588         }
 589         if (answer)
 590                 goto out_permanent;
 591
 592         /* Add the new entry after the last permanent entry if any, so that
 593          * the new entry does not override a permanent entry when matched with
 594          * a wild-card protocol. But it is allowed to override any existing
 595          * non-permanent entry.  This means that when we remove this entry, the
 596          * system automatically returns to the old behavior.
 597          */
 598         list_add_rcu(&p->list, last_perm);
 599         ret = 0;
 600 out:
 601         spin_unlock_bh(&inetsw6_lock);
 602         return ret;
 603
 604 out_permanent:
 605         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 606                protocol);
 607         goto out;
 608
 609 out_illegal:
 610         printk(KERN_ERR
 611                "Ignoring attempt to register invalid socket type %d.\n",
 612                p->type);
 613         goto out;
 614 }
 615
 616 EXPORT_SYMBOL(inet6_register_protosw);
 617
 618 void
 619 inet6_unregister_protosw(struct inet_protosw *p)
 620 {
 621         if (INET_PROTOSW_PERMANENT & p->flags) {
 622                 printk(KERN_ERR
 623                        "Attempt to unregister permanent protocol %d.\n",
 624                        p->protocol);
 625         } else {
 626                 spin_lock_bh(&inetsw6_lock);
 627                 list_del_rcu(&p->list);
 628                 spin_unlock_bh(&inetsw6_lock);
 629
 630                 synchronize_net();
 631         }
 632 }
 633
 634 EXPORT_SYMBOL(inet6_unregister_protosw);
 635
 636 int inet6_sk_rebuild_header(struct sock *sk)
 637 {
 638         int err;
 639         struct dst_entry *dst;
 640         struct ipv6_pinfo *np = inet6_sk(sk);
 641
 642         dst = __sk_dst_check(sk, np->dst_cookie);
 643
 644         if (dst == NULL) {
 645                 struct inet_sock *inet = inet_sk(sk);
 646                 struct in6_addr *final_p = NULL, final;
 647                 struct flowi fl;
 648
 649                 memset(&fl, 0, sizeof(fl));
 650                 fl.proto = sk->sk_protocol;
 651                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 652                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 653                 fl.fl6_flowlabel = np->flow_label;
 654                 fl.oif = sk->sk_bound_dev_if;
 655                 fl.fl_ip_dport = inet->dport;
 656                 fl.fl_ip_sport = inet->sport;
 657                 security_sk_classify_flow(sk, &fl);
 658
 659                 if (np->opt && np->opt->srcrt) {
 660                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 661                         ipv6_addr_copy(&final, &fl.fl6_dst);
 662                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 663                         final_p = &final;
 664                 }
 665
 666                 err = ip6_dst_lookup(sk, &dst, &fl);
 667                 if (err) {
 668                         sk->sk_route_caps = 0;
 669                         return err;
 670                 }
 671                 if (final_p)
 672                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 673
 674                 if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
 675                         sk->sk_err_soft = -err;
 676                         return err;
 677                 }
 678
 679                 __ip6_dst_store(sk, dst, NULL, NULL);
 680         }
 681
 682         return 0;
 683 }
 684
 685 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 686
 687 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 688 {
 689         struct ipv6_pinfo *np = inet6_sk(sk);
 690         struct inet6_skb_parm *opt = IP6CB(skb);
 691
 692         if (np->rxopt.all) {
 693                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 694                                   np->rxopt.bits.ohopopts)) ||
 695                     ((IPV6_FLOWINFO_MASK &
 696                       *(__be32 *)skb_network_header(skb)) &&
 697                      np->rxopt.bits.rxflow) ||
 698                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 699                      np->rxopt.bits.osrcrt)) ||
 700                     ((opt->dst1 || opt->dst0) &&
 701                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 702                         return 1;
 703         }
 704         return 0;
 705 }
 706
 707 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 708
 709 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
 710 {
 711         struct inet6_protocol *ops = NULL;
 712
 713         for (;;) {
 714                 struct ipv6_opt_hdr *opth;
 715                 int len;
 716
 717                 if (proto != NEXTHDR_HOP) {
 718                         ops = rcu_dereference(inet6_protos[proto]);
 719
 720                         if (unlikely(!ops))
 721                                 break;
 722
 723                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
 724                                 break;
 725                 }
 726
 727                 if (unlikely(!pskb_may_pull(skb, 8)))
 728                         break;
 729
 730                 opth = (void *)skb->data;
 731                 len = ipv6_optlen(opth);
 732
 733                 if (unlikely(!pskb_may_pull(skb, len)))
 734                         break;
 735
 736                 proto = opth->nexthdr;
 737                 __skb_pull(skb, len);
 738         }
 739
 740         return proto;
 741 }
 742
 743 static int ipv6_gso_send_check(struct sk_buff *skb)
 744 {
 745         struct ipv6hdr *ipv6h;
 746         struct inet6_protocol *ops;
 747         int err = -EINVAL;
 748
 749         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 750                 goto out;
 751
 752         ipv6h = ipv6_hdr(skb);
 753         __skb_pull(skb, sizeof(*ipv6h));
 754         err = -EPROTONOSUPPORT;
 755
 756         rcu_read_lock();
 757         ops = rcu_dereference(inet6_protos[
 758                 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
 759
 760         if (likely(ops && ops->gso_send_check)) {
 761                 skb_reset_transport_header(skb);
 762                 err = ops->gso_send_check(skb);
 763         }
 764         rcu_read_unlock();
 765
 766 out:
 767         return err;
 768 }
 769
 770 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
 771 {
 772         struct sk_buff *segs = ERR_PTR(-EINVAL);
 773         struct ipv6hdr *ipv6h;
 774         struct inet6_protocol *ops;
 775         int proto;
 776         struct frag_hdr *fptr;
 777         unsigned int unfrag_ip6hlen;
 778         u8 *prevhdr;
 779         int offset = 0;
 780
 781         if (!(features & NETIF_F_V6_CSUM))
 782                 features &= ~NETIF_F_SG;
 783
 784         if (unlikely(skb_shinfo(skb)->gso_type &
 785                      ~(SKB_GSO_UDP |
 786                        SKB_GSO_DODGY |
 787                        SKB_GSO_TCP_ECN |
 788                        SKB_GSO_TCPV6 |
 789                        0)))
 790                 goto out;
 791
 792         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 793                 goto out;
 794
 795         ipv6h = ipv6_hdr(skb);
 796         __skb_pull(skb, sizeof(*ipv6h));
 797         segs = ERR_PTR(-EPROTONOSUPPORT);
 798
 799         proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
 800         rcu_read_lock();
 801         ops = rcu_dereference(inet6_protos[proto]);
 802         if (likely(ops && ops->gso_segment)) {
 803                 skb_reset_transport_header(skb);
 804                 segs = ops->gso_segment(skb, features);
 805         }
 806         rcu_read_unlock();
 807
 808         if (unlikely(IS_ERR(segs)))
 809                 goto out;
 810
 811         for (skb = segs; skb; skb = skb->next) {
 812                 ipv6h = ipv6_hdr(skb);
 813                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
 814                                            sizeof(*ipv6h));
 815                 if (proto == IPPROTO_UDP) {
 816                         unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
 817                         fptr = (struct frag_hdr *)(skb_network_header(skb) +
 818                                 unfrag_ip6hlen);
 819                         fptr->frag_off = htons(offset);
 820                         if (skb->next != NULL)
 821                                 fptr->frag_off |= htons(IP6_MF);
 822                         offset += (ntohs(ipv6h->payload_len) -
 823                                    sizeof(struct frag_hdr));
 824                 }
 825         }
 826
 827 out:
 828         return segs;
 829 }
 830
 831 struct ipv6_gro_cb {
 832         struct napi_gro_cb napi;
 833         int proto;
 834 };
 835
 836 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
 837
 838 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
 839                                          struct sk_buff *skb)
 840 {
 841         struct inet6_protocol *ops;
 842         struct sk_buff **pp = NULL;
 843         struct sk_buff *p;
 844         struct ipv6hdr *iph;
 845         unsigned int nlen;
 846         unsigned int hlen;
 847         unsigned int off;
 848         int flush = 1;
 849         int proto;
 850         __wsum csum;
 851
 852         off = skb_gro_offset(skb);
 853         hlen = off + sizeof(*iph);
 854         iph = skb_gro_header_fast(skb, off);
 855         if (skb_gro_header_hard(skb, hlen)) {
 856                 iph = skb_gro_header_slow(skb, hlen, off);
 857                 if (unlikely(!iph))
 858                         goto out;
 859         }
 860
 861         skb_gro_pull(skb, sizeof(*iph));
 862         skb_set_transport_header(skb, skb_gro_offset(skb));
 863
 864         flush += ntohs(iph->payload_len) != skb_gro_len(skb);
 865
 866         rcu_read_lock();
 867         proto = iph->nexthdr;
 868         ops = rcu_dereference(inet6_protos[proto]);
 869         if (!ops || !ops->gro_receive) {
 870                 __pskb_pull(skb, skb_gro_offset(skb));
 871                 proto = ipv6_gso_pull_exthdrs(skb, proto);
 872                 skb_gro_pull(skb, -skb_transport_offset(skb));
 873                 skb_reset_transport_header(skb);
 874                 __skb_push(skb, skb_gro_offset(skb));
 875
 876                 if (!ops || !ops->gro_receive)
 877                         goto out_unlock;
 878
 879                 iph = ipv6_hdr(skb);
 880         }
 881
 882         IPV6_GRO_CB(skb)->proto = proto;
 883
 884         flush--;
 885         nlen = skb_network_header_len(skb);
 886
 887         for (p = *head; p; p = p->next) {
 888                 struct ipv6hdr *iph2;
 889
 890                 if (!NAPI_GRO_CB(p)->same_flow)
 891                         continue;
 892
 893                 iph2 = ipv6_hdr(p);
 894
 895                 /* All fields must match except length. */
 896                 if (nlen != skb_network_header_len(p) ||
 897                     memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
 898                     memcmp(&iph->nexthdr, &iph2->nexthdr,
 899                            nlen - offsetof(struct ipv6hdr, nexthdr))) {
 900                         NAPI_GRO_CB(p)->same_flow = 0;
 901                         continue;
 902                 }
 903
 904                 NAPI_GRO_CB(p)->flush |= flush;
 905         }
 906
 907         NAPI_GRO_CB(skb)->flush |= flush;
 908
 909         csum = skb->csum;
 910         skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
 911
 912         pp = ops->gro_receive(head, skb);
 913
 914         skb->csum = csum;
 915
 916 out_unlock:
 917         rcu_read_unlock();
 918
 919 out:
 920         NAPI_GRO_CB(skb)->flush |= flush;
 921
 922         return pp;
 923 }
 924
 925 static int ipv6_gro_complete(struct sk_buff *skb)
 926 {
 927         struct inet6_protocol *ops;
 928         struct ipv6hdr *iph = ipv6_hdr(skb);
 929         int err = -ENOSYS;
 930
 931         iph->payload_len = htons(skb->len - skb_network_offset(skb) -
 932                                  sizeof(*iph));
 933
 934         rcu_read_lock();
 935         ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
 936         if (WARN_ON(!ops || !ops->gro_complete))
 937                 goto out_unlock;
 938
 939         err = ops->gro_complete(skb);
 940
 941 out_unlock:
 942         rcu_read_unlock();
 943
 944         return err;
 945 }
 946
 947 static struct packet_type ipv6_packet_type __read_mostly = {
 948         .type = cpu_to_be16(ETH_P_IPV6),
 949         .func = ipv6_rcv,
 950         .gso_send_check = ipv6_gso_send_check,
 951         .gso_segment = ipv6_gso_segment,
 952         .gro_receive = ipv6_gro_receive,
 953         .gro_complete = ipv6_gro_complete,
 954 };
 955
 956 static int __init ipv6_packet_init(void)
 957 {
 958         dev_add_pack(&ipv6_packet_type);
 959         return 0;
 960 }
 961
 962 static void ipv6_packet_cleanup(void)
 963 {
 964         dev_remove_pack(&ipv6_packet_type);
 965 }
 966
 967 static int __net_init ipv6_init_mibs(struct net *net)
 968 {
 969         if (snmp_mib_init((void **)net->mib.udp_stats_in6,
 970                           sizeof (struct udp_mib)) < 0)
 971                 return -ENOMEM;
 972         if (snmp_mib_init((void **)net->mib.udplite_stats_in6,
 973                           sizeof (struct udp_mib)) < 0)
 974                 goto err_udplite_mib;
 975         if (snmp_mib_init((void **)net->mib.ipv6_statistics,
 976                           sizeof(struct ipstats_mib)) < 0)
 977                 goto err_ip_mib;
 978         if (snmp_mib_init((void **)net->mib.icmpv6_statistics,
 979                           sizeof(struct icmpv6_mib)) < 0)
 980                 goto err_icmp_mib;
 981         if (snmp_mib_init((void **)net->mib.icmpv6msg_statistics,
 982                           sizeof(struct icmpv6msg_mib)) < 0)
 983                 goto err_icmpmsg_mib;
 984         return 0;
 985
 986 err_icmpmsg_mib:
 987         snmp_mib_free((void **)net->mib.icmpv6_statistics);
 988 err_icmp_mib:
 989         snmp_mib_free((void **)net->mib.ipv6_statistics);
 990 err_ip_mib:
 991         snmp_mib_free((void **)net->mib.udplite_stats_in6);
 992 err_udplite_mib:
 993         snmp_mib_free((void **)net->mib.udp_stats_in6);
 994         return -ENOMEM;
 995 }
 996
 997 static void __net_exit ipv6_cleanup_mibs(struct net *net)
 998 {
 999         snmp_mib_free((void **)net->mib.udp_stats_in6);
1000         snmp_mib_free((void **)net->mib.udplite_stats_in6);
1001         snmp_mib_free((void **)net->mib.ipv6_statistics);
1002         snmp_mib_free((void **)net->mib.icmpv6_statistics);
1003         snmp_mib_free((void **)net->mib.icmpv6msg_statistics);
1004 }
1005
1006 static int __net_init inet6_net_init(struct net *net)
1007 {
1008         int err = 0;
1009
1010         net->ipv6.sysctl.bindv6only = 0;
1011         net->ipv6.sysctl.icmpv6_time = 1*HZ;
1012
1013         err = ipv6_init_mibs(net);
1014         if (err)
1015                 return err;
1016 #ifdef CONFIG_PROC_FS
1017         err = udp6_proc_init(net);
1018         if (err)
1019                 goto out;
1020         err = tcp6_proc_init(net);
1021         if (err)
1022                 goto proc_tcp6_fail;
1023         err = ac6_proc_init(net);
1024         if (err)
1025                 goto proc_ac6_fail;
1026 #endif
1027         return err;
1028
1029 #ifdef CONFIG_PROC_FS
1030 proc_ac6_fail:
1031         tcp6_proc_exit(net);
1032 proc_tcp6_fail:
1033         udp6_proc_exit(net);
1034 out:
1035         ipv6_cleanup_mibs(net);
1036         return err;
1037 #endif
1038 }
1039
1040 static void inet6_net_exit(struct net *net)
1041 {
1042 #ifdef CONFIG_PROC_FS
1043         udp6_proc_exit(net);
1044         tcp6_proc_exit(net);
1045         ac6_proc_exit(net);
1046 #endif
1047         ipv6_cleanup_mibs(net);
1048 }
1049
1050 static struct pernet_operations inet6_net_ops = {
1051         .init = inet6_net_init,
1052         .exit = inet6_net_exit,
1053 };
1054
1055 static int __init inet6_init(void)
1056 {
1057         struct sk_buff *dummy_skb;
1058         struct list_head *r;
1059         int err = 0;
1060
1061         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1062
1063         /* Register the socket-side information for inet6_create.  */
1064         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1065                 INIT_LIST_HEAD(r);
1066
1067         if (disable_ipv6_mod) {
1068                 printk(KERN_INFO
1069                        "IPv6: Loaded, but administratively disabled, "
1070                        "reboot required to enable\n");
1071                 goto out;
1072         }
1073
1074         err = proto_register(&tcpv6_prot, 1);
1075         if (err)
1076                 goto out;
1077
1078         err = proto_register(&udpv6_prot, 1);
1079         if (err)
1080                 goto out_unregister_tcp_proto;
1081
1082         err = proto_register(&udplitev6_prot, 1);
1083         if (err)
1084                 goto out_unregister_udp_proto;
1085
1086         err = proto_register(&rawv6_prot, 1);
1087         if (err)
1088                 goto out_unregister_udplite_proto;
1089
1090
1091         /* We MUST register RAW sockets before we create the ICMP6,
1092          * IGMP6, or NDISC control sockets.
1093          */
1094         err = rawv6_init();
1095         if (err)
1096                 goto out_unregister_raw_proto;
1097
1098         /* Register the family here so that the init calls below will
1099          * be able to create sockets. (?? is this dangerous ??)
1100          */
1101         err = sock_register(&inet6_family_ops);
1102         if (err)
1103                 goto out_sock_register_fail;
1104
1105 #ifdef CONFIG_SYSCTL
1106         err = ipv6_static_sysctl_register();
1107         if (err)
1108                 goto static_sysctl_fail;
1109 #endif
1110         /*
1111          *      ipngwg API draft makes clear that the correct semantics
1112          *      for TCP and UDP is to consider one TCP and UDP instance
1113          *      in a host availiable by both INET and INET6 APIs and
1114          *      able to communicate via both network protocols.
1115          */
1116
1117         err = register_pernet_subsys(&inet6_net_ops);
1118         if (err)
1119                 goto register_pernet_fail;
1120         err = icmpv6_init();
1121         if (err)
1122                 goto icmp_fail;
1123         err = ip6_mr_init();
1124         if (err)
1125                 goto ipmr_fail;
1126         err = ndisc_init();
1127         if (err)
1128                 goto ndisc_fail;
1129         err = igmp6_init();
1130         if (err)
1131                 goto igmp_fail;
1132         err = ipv6_netfilter_init();
1133         if (err)
1134                 goto netfilter_fail;
1135         /* Create /proc/foo6 entries. */
1136 #ifdef CONFIG_PROC_FS
1137         err = -ENOMEM;
1138         if (raw6_proc_init())
1139                 goto proc_raw6_fail;
1140         if (udplite6_proc_init())
1141                 goto proc_udplite6_fail;
1142         if (ipv6_misc_proc_init())
1143                 goto proc_misc6_fail;
1144         if (if6_proc_init())
1145                 goto proc_if6_fail;
1146 #endif
1147         err = ip6_route_init();
1148         if (err)
1149                 goto ip6_route_fail;
1150         err = ip6_flowlabel_init();
1151         if (err)
1152                 goto ip6_flowlabel_fail;
1153         err = addrconf_init();
1154         if (err)
1155                 goto addrconf_fail;
1156
1157         /* Init v6 extension headers. */
1158         err = ipv6_exthdrs_init();
1159         if (err)
1160                 goto ipv6_exthdrs_fail;
1161
1162         err = ipv6_frag_init();
1163         if (err)
1164                 goto ipv6_frag_fail;
1165
1166         /* Init v6 transport protocols. */
1167         err = udpv6_init();
1168         if (err)
1169                 goto udpv6_fail;
1170
1171         err = udplitev6_init();
1172         if (err)
1173                 goto udplitev6_fail;
1174
1175         err = tcpv6_init();
1176         if (err)
1177                 goto tcpv6_fail;
1178
1179         err = ipv6_packet_init();
1180         if (err)
1181                 goto ipv6_packet_fail;
1182
1183 #ifdef CONFIG_SYSCTL
1184         err = ipv6_sysctl_register();
1185         if (err)
1186                 goto sysctl_fail;
1187 #endif
1188 out:
1189         return err;
1190
1191 #ifdef CONFIG_SYSCTL
1192 sysctl_fail:
1193         ipv6_packet_cleanup();
1194 #endif
1195 ipv6_packet_fail:
1196         tcpv6_exit();
1197 tcpv6_fail:
1198         udplitev6_exit();
1199 udplitev6_fail:
1200         udpv6_exit();
1201 udpv6_fail:
1202         ipv6_frag_exit();
1203 ipv6_frag_fail:
1204         ipv6_exthdrs_exit();
1205 ipv6_exthdrs_fail:
1206         addrconf_cleanup();
1207 addrconf_fail:
1208         ip6_flowlabel_cleanup();
1209 ip6_flowlabel_fail:
1210         ip6_route_cleanup();
1211 ip6_route_fail:
1212 #ifdef CONFIG_PROC_FS
1213         if6_proc_exit();
1214 proc_if6_fail:
1215         ipv6_misc_proc_exit();
1216 proc_misc6_fail:
1217         udplite6_proc_exit();
1218 proc_udplite6_fail:
1219         raw6_proc_exit();
1220 proc_raw6_fail:
1221 #endif
1222         ipv6_netfilter_fini();
1223 netfilter_fail:
1224         igmp6_cleanup();
1225 igmp_fail:
1226         ndisc_cleanup();
1227 ndisc_fail:
1228         ip6_mr_cleanup();
1229 ipmr_fail:
1230         icmpv6_cleanup();
1231 icmp_fail:
1232         unregister_pernet_subsys(&inet6_net_ops);
1233 register_pernet_fail:
1234 #ifdef CONFIG_SYSCTL
1235         ipv6_static_sysctl_unregister();
1236 static_sysctl_fail:
1237 #endif
1238         sock_unregister(PF_INET6);
1239         rtnl_unregister_all(PF_INET6);
1240 out_sock_register_fail:
1241         rawv6_exit();
1242 out_unregister_raw_proto:
1243         proto_unregister(&rawv6_prot);
1244 out_unregister_udplite_proto:
1245         proto_unregister(&udplitev6_prot);
1246 out_unregister_udp_proto:
1247         proto_unregister(&udpv6_prot);
1248 out_unregister_tcp_proto:
1249         proto_unregister(&tcpv6_prot);
1250         goto out;
1251 }
1252 module_init(inet6_init);
1253
1254 static void __exit inet6_exit(void)
1255 {
1256         if (disable_ipv6_mod)
1257                 return;
1258
1259         /* First of all disallow new sockets creation. */
1260         sock_unregister(PF_INET6);
1261         /* Disallow any further netlink messages */
1262         rtnl_unregister_all(PF_INET6);
1263
1264 #ifdef CONFIG_SYSCTL
1265         ipv6_sysctl_unregister();
1266 #endif
1267         udpv6_exit();
1268         udplitev6_exit();
1269         tcpv6_exit();
1270
1271         /* Cleanup code parts. */
1272         ipv6_packet_cleanup();
1273         ipv6_frag_exit();
1274         ipv6_exthdrs_exit();
1275         addrconf_cleanup();
1276         ip6_flowlabel_cleanup();
1277         ip6_route_cleanup();
1278 #ifdef CONFIG_PROC_FS
1279
1280         /* Cleanup code parts. */
1281         if6_proc_exit();
1282         ipv6_misc_proc_exit();
1283         udplite6_proc_exit();
1284         raw6_proc_exit();
1285 #endif
1286         ipv6_netfilter_fini();
1287         igmp6_cleanup();
1288         ndisc_cleanup();
1289         ip6_mr_cleanup();
1290         icmpv6_cleanup();
1291         rawv6_exit();
1292
1293         unregister_pernet_subsys(&inet6_net_ops);
1294 #ifdef CONFIG_SYSCTL
1295         ipv6_static_sysctl_unregister();
1296 #endif
1297         proto_unregister(&rawv6_prot);
1298         proto_unregister(&udplitev6_prot);
1299         proto_unregister(&udpv6_prot);
1300         proto_unregister(&tcpv6_prot);
1301
1302         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1303 }
1304 module_exit(inet6_exit);
1305
1306 MODULE_ALIAS_NETPROTO(PF_INET6);