linux/net/ipv6/af_inet6.c
<<
>>
Prefs
   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#define pr_fmt(fmt) "IPv6: " fmt
  22
  23#include <linux/module.h>
  24#include <linux/capability.h>
  25#include <linux/errno.h>
  26#include <linux/types.h>
  27#include <linux/socket.h>
  28#include <linux/in.h>
  29#include <linux/kernel.h>
  30#include <linux/timer.h>
  31#include <linux/string.h>
  32#include <linux/sockios.h>
  33#include <linux/net.h>
  34#include <linux/fcntl.h>
  35#include <linux/mm.h>
  36#include <linux/interrupt.h>
  37#include <linux/proc_fs.h>
  38#include <linux/stat.h>
  39#include <linux/init.h>
  40#include <linux/slab.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/protocol.h>
  53#include <net/inet_common.h>
  54#include <net/route.h>
  55#include <net/transp_v6.h>
  56#include <net/ip6_route.h>
  57#include <net/addrconf.h>
  58#include <net/ndisc.h>
  59#ifdef CONFIG_IPV6_TUNNEL
  60#include <net/ip6_tunnel.h>
  61#endif
  62
  63#include <asm/uaccess.h>
  64#include <linux/mroute6.h>
  65
  66#include "ip6_offload.h"
  67
  68MODULE_AUTHOR("Cast of dozens");
  69MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  70MODULE_LICENSE("GPL");
  71
  72/* The inetsw6 table contains everything that inet6_create needs to
  73 * build a new socket.
  74 */
  75static struct list_head inetsw6[SOCK_MAX];
  76static DEFINE_SPINLOCK(inetsw6_lock);
  77
  78struct ipv6_params ipv6_defaults = {
  79        .disable_ipv6 = 0,
  80        .autoconf = 1,
  81};
  82
  83static int disable_ipv6_mod;
  84
  85module_param_named(disable, disable_ipv6_mod, int, 0444);
  86MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  87
  88module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  89MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  90
  91module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  92MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  93
  94bool ipv6_mod_enabled(void)
  95{
  96        return disable_ipv6_mod == 0;
  97}
  98EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
  99
 100static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
 101{
 102        const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
 103
 104        return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
 105}
 106
 107static int inet6_create(struct net *net, struct socket *sock, int protocol,
 108                        int kern)
 109{
 110        struct inet_sock *inet;
 111        struct ipv6_pinfo *np;
 112        struct sock *sk;
 113        struct inet_protosw *answer;
 114        struct proto *answer_prot;
 115        unsigned char answer_flags;
 116        int try_loading_module = 0;
 117        int err;
 118
 119        if (protocol < 0 || protocol >= IPPROTO_MAX)
 120                return -EINVAL;
 121
 122        /* Look for the requested type/protocol pair. */
 123lookup_protocol:
 124        err = -ESOCKTNOSUPPORT;
 125        rcu_read_lock();
 126        list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 127
 128                err = 0;
 129                /* Check the non-wild match. */
 130                if (protocol == answer->protocol) {
 131                        if (protocol != IPPROTO_IP)
 132                                break;
 133                } else {
 134                        /* Check for the two wild cases. */
 135                        if (IPPROTO_IP == protocol) {
 136                                protocol = answer->protocol;
 137                                break;
 138                        }
 139                        if (IPPROTO_IP == answer->protocol)
 140                                break;
 141                }
 142                err = -EPROTONOSUPPORT;
 143        }
 144
 145        if (err) {
 146                if (try_loading_module < 2) {
 147                        rcu_read_unlock();
 148                        /*
 149                         * Be more specific, e.g. net-pf-10-proto-132-type-1
 150                         * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 151                         */
 152                        if (++try_loading_module == 1)
 153                                request_module("net-pf-%d-proto-%d-type-%d",
 154                                                PF_INET6, protocol, sock->type);
 155                        /*
 156                         * Fall back to generic, e.g. net-pf-10-proto-132
 157                         * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 158                         */
 159                        else
 160                                request_module("net-pf-%d-proto-%d",
 161                                                PF_INET6, protocol);
 162                        goto lookup_protocol;
 163                } else
 164                        goto out_rcu_unlock;
 165        }
 166
 167        err = -EPERM;
 168        if (sock->type == SOCK_RAW && !kern &&
 169            !ns_capable(net->user_ns, CAP_NET_RAW))
 170                goto out_rcu_unlock;
 171
 172        sock->ops = answer->ops;
 173        answer_prot = answer->prot;
 174        answer_flags = answer->flags;
 175        rcu_read_unlock();
 176
 177        WARN_ON(answer_prot->slab == NULL);
 178
 179        err = -ENOBUFS;
 180        sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 181        if (sk == NULL)
 182                goto out;
 183
 184        sock_init_data(sock, sk);
 185
 186        err = 0;
 187        if (INET_PROTOSW_REUSE & answer_flags)
 188                sk->sk_reuse = SK_CAN_REUSE;
 189
 190        inet = inet_sk(sk);
 191        inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 192
 193        if (SOCK_RAW == sock->type) {
 194                inet->inet_num = protocol;
 195                if (IPPROTO_RAW == protocol)
 196                        inet->hdrincl = 1;
 197        }
 198
 199        sk->sk_destruct         = inet_sock_destruct;
 200        sk->sk_family           = PF_INET6;
 201        sk->sk_protocol         = protocol;
 202
 203        sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 204
 205        inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 206        np->hop_limit   = -1;
 207        np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 208        np->mc_loop     = 1;
 209        np->pmtudisc    = IPV6_PMTUDISC_WANT;
 210        np->ipv6only    = net->ipv6.sysctl.bindv6only;
 211
 212        /* Init the ipv4 part of the socket since we can have sockets
 213         * using v6 API for ipv4.
 214         */
 215        inet->uc_ttl    = -1;
 216
 217        inet->mc_loop   = 1;
 218        inet->mc_ttl    = 1;
 219        inet->mc_index  = 0;
 220        inet->mc_list   = NULL;
 221        inet->rcv_tos   = 0;
 222
 223        if (net->sysctl_ip_no_pmtu_disc)
 224                inet->pmtudisc = IP_PMTUDISC_DONT;
 225        else
 226                inet->pmtudisc = IP_PMTUDISC_WANT;
 227        /*
 228         * Increment only the relevant sk_prot->socks debug field, this changes
 229         * the previous behaviour of incrementing both the equivalent to
 230         * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 231         *
 232         * This allows better debug granularity as we'll know exactly how many
 233         * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 234         * transport protocol socks. -acme
 235         */
 236        sk_refcnt_debug_inc(sk);
 237
 238        if (inet->inet_num) {
 239                /* It assumes that any protocol which allows
 240                 * the user to assign a number at socket
 241                 * creation time automatically shares.
 242                 */
 243                inet->inet_sport = htons(inet->inet_num);
 244                sk->sk_prot->hash(sk);
 245        }
 246        if (sk->sk_prot->init) {
 247                err = sk->sk_prot->init(sk);
 248                if (err) {
 249                        sk_common_release(sk);
 250                        goto out;
 251                }
 252        }
 253out:
 254        return err;
 255out_rcu_unlock:
 256        rcu_read_unlock();
 257        goto out;
 258}
 259
 260
 261/* bind for INET6 API */
 262int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 263{
 264        struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 265        struct sock *sk = sock->sk;
 266        struct inet_sock *inet = inet_sk(sk);
 267        struct ipv6_pinfo *np = inet6_sk(sk);
 268        struct net *net = sock_net(sk);
 269        __be32 v4addr = 0;
 270        unsigned short snum;
 271        int addr_type = 0;
 272        int err = 0;
 273
 274        /* If the socket has its own bind function then use it. */
 275        if (sk->sk_prot->bind)
 276                return sk->sk_prot->bind(sk, uaddr, addr_len);
 277
 278        if (addr_len < SIN6_LEN_RFC2133)
 279                return -EINVAL;
 280
 281        if (addr->sin6_family != AF_INET6)
 282                return -EAFNOSUPPORT;
 283
 284        addr_type = ipv6_addr_type(&addr->sin6_addr);
 285        if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 286                return -EINVAL;
 287
 288        snum = ntohs(addr->sin6_port);
 289        if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 290                return -EACCES;
 291
 292        lock_sock(sk);
 293
 294        /* Check these errors (active socket, double bind). */
 295        if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 296                err = -EINVAL;
 297                goto out;
 298        }
 299
 300        /* Check if the address belongs to the host. */
 301        if (addr_type == IPV6_ADDR_MAPPED) {
 302                int chk_addr_ret;
 303
 304                /* Binding to v4-mapped address on a v6-only socket
 305                 * makes no sense
 306                 */
 307                if (np->ipv6only) {
 308                        err = -EINVAL;
 309                        goto out;
 310                }
 311
 312                /* Reproduce AF_INET checks to make the bindings consistent */
 313                v4addr = addr->sin6_addr.s6_addr32[3];
 314                chk_addr_ret = inet_addr_type(net, v4addr);
 315                if (!net->ipv4_sysctl_ip_nonlocal_bind &&
 316                    !(inet->freebind || inet->transparent) &&
 317                    v4addr != htonl(INADDR_ANY) &&
 318                    chk_addr_ret != RTN_LOCAL &&
 319                    chk_addr_ret != RTN_MULTICAST &&
 320                    chk_addr_ret != RTN_BROADCAST) {
 321                        err = -EADDRNOTAVAIL;
 322                        goto out;
 323                }
 324        } else {
 325                if (addr_type != IPV6_ADDR_ANY) {
 326                        struct net_device *dev = NULL;
 327
 328                        rcu_read_lock();
 329                        if (__ipv6_addr_needs_scope_id(addr_type)) {
 330                                if (addr_len >= sizeof(struct sockaddr_in6) &&
 331                                    addr->sin6_scope_id) {
 332                                        /* Override any existing binding, if another one
 333                                         * is supplied by user.
 334                                         */
 335                                        sk->sk_bound_dev_if = addr->sin6_scope_id;
 336                                }
 337
 338                                /* Binding to link-local address requires an interface */
 339                                if (!sk->sk_bound_dev_if) {
 340                                        err = -EINVAL;
 341                                        goto out_unlock;
 342                                }
 343                                dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 344                                if (!dev) {
 345                                        err = -ENODEV;
 346                                        goto out_unlock;
 347                                }
 348                        }
 349
 350                        /* ipv4 addr of the socket is invalid.  Only the
 351                         * unspecified and mapped address have a v4 equivalent.
 352                         */
 353                        v4addr = LOOPBACK4_IPV6;
 354                        if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 355                                if (!net->ipv6_sysctl_ip_nonlocal_bind &&
 356                                    !(inet->freebind || inet->transparent) &&
 357                                    !ipv6_chk_addr(net, &addr->sin6_addr,
 358                                                   dev, 0)) {
 359                                        err = -EADDRNOTAVAIL;
 360                                        goto out_unlock;
 361                                }
 362                        }
 363                        rcu_read_unlock();
 364                }
 365        }
 366
 367        inet->inet_rcv_saddr = v4addr;
 368        inet->inet_saddr = v4addr;
 369
 370        sk->sk_v6_rcv_saddr = addr->sin6_addr;
 371
 372        if (!(addr_type & IPV6_ADDR_MULTICAST))
 373                np->saddr = addr->sin6_addr;
 374
 375        /* Make sure we are allowed to bind here. */
 376        if ((snum || !inet->bind_address_no_port) &&
 377            sk->sk_prot->get_port(sk, snum)) {
 378                inet_reset_saddr(sk);
 379                err = -EADDRINUSE;
 380                goto out;
 381        }
 382
 383        if (addr_type != IPV6_ADDR_ANY) {
 384                sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 385                if (addr_type != IPV6_ADDR_MAPPED)
 386                        np->ipv6only = 1;
 387        }
 388        if (snum)
 389                sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 390        inet->inet_sport = htons(inet->inet_num);
 391        inet->inet_dport = 0;
 392        inet->inet_daddr = 0;
 393out:
 394        release_sock(sk);
 395        return err;
 396out_unlock:
 397        rcu_read_unlock();
 398        goto out;
 399}
 400EXPORT_SYMBOL(inet6_bind);
 401
 402int inet6_release(struct socket *sock)
 403{
 404        struct sock *sk = sock->sk;
 405
 406        if (sk == NULL)
 407                return -EINVAL;
 408
 409        /* Free mc lists */
 410        ipv6_sock_mc_close(sk);
 411
 412        /* Free ac lists */
 413        ipv6_sock_ac_close(sk);
 414
 415        return inet_release(sock);
 416}
 417EXPORT_SYMBOL(inet6_release);
 418
 419void inet6_destroy_sock(struct sock *sk)
 420{
 421        struct ipv6_pinfo *np = inet6_sk(sk);
 422        struct sk_buff *skb;
 423        struct ipv6_txoptions *opt;
 424
 425        /* Release rx options */
 426
 427        skb = xchg(&np->pktoptions, NULL);
 428        if (skb != NULL)
 429                kfree_skb(skb);
 430
 431        skb = xchg(&np->rxpmtu, NULL);
 432        if (skb != NULL)
 433                kfree_skb(skb);
 434
 435        /* Free flowlabels */
 436        fl6_free_socklist(sk);
 437
 438        /* Free tx options */
 439
 440        opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 441        if (opt) {
 442                atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 443                txopt_put(opt);
 444        }
 445}
 446EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 447
 448/*
 449 *      This does both peername and sockname.
 450 */
 451
 452int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 453                 int *uaddr_len, int peer)
 454{
 455        struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 456        struct sock *sk = sock->sk;
 457        struct inet_sock *inet = inet_sk(sk);
 458        struct ipv6_pinfo *np = inet6_sk(sk);
 459
 460        sin->sin6_family = AF_INET6;
 461        sin->sin6_flowinfo = 0;
 462        sin->sin6_scope_id = 0;
 463        if (peer) {
 464                if (!inet->inet_dport)
 465                        return -ENOTCONN;
 466                if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 467                    peer == 1)
 468                        return -ENOTCONN;
 469                sin->sin6_port = inet->inet_dport;
 470                sin->sin6_addr = sk->sk_v6_daddr;
 471                if (np->sndflow)
 472                        sin->sin6_flowinfo = np->flow_label;
 473        } else {
 474                if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 475                        sin->sin6_addr = np->saddr;
 476                else
 477                        sin->sin6_addr = sk->sk_v6_rcv_saddr;
 478
 479                sin->sin6_port = inet->inet_sport;
 480        }
 481        sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 482                                                 sk->sk_bound_dev_if);
 483        *uaddr_len = sizeof(*sin);
 484        return 0;
 485}
 486EXPORT_SYMBOL(inet6_getname);
 487
 488int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 489{
 490        struct sock *sk = sock->sk;
 491        struct net *net = sock_net(sk);
 492
 493        switch (cmd) {
 494        case SIOCGSTAMP:
 495                return sock_get_timestamp(sk, (struct timeval __user *)arg);
 496
 497        case SIOCGSTAMPNS:
 498                return sock_get_timestampns(sk, (struct timespec __user *)arg);
 499
 500        case SIOCADDRT:
 501        case SIOCDELRT:
 502
 503                return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 504
 505        case SIOCSIFADDR:
 506                return addrconf_add_ifaddr(net, (void __user *) arg);
 507        case SIOCDIFADDR:
 508                return addrconf_del_ifaddr(net, (void __user *) arg);
 509        case SIOCSIFDSTADDR:
 510                return addrconf_set_dstaddr(net, (void __user *) arg);
 511        default:
 512                if (!sk->sk_prot->ioctl)
 513                        return -ENOIOCTLCMD;
 514                return sk->sk_prot->ioctl(sk, cmd, arg);
 515        }
 516        /*NOTREACHED*/
 517        return 0;
 518}
 519EXPORT_SYMBOL(inet6_ioctl);
 520
 521const struct proto_ops inet6_stream_ops = {
 522        .family            = PF_INET6,
 523        .owner             = THIS_MODULE,
 524        .release           = inet6_release,
 525        .bind              = inet6_bind,
 526        .connect           = inet_stream_connect,       /* ok           */
 527        .socketpair        = sock_no_socketpair,        /* a do nothing */
 528        .accept            = inet_accept,               /* ok           */
 529        .getname           = inet6_getname,
 530        .poll              = tcp_poll,                  /* ok           */
 531        .ioctl             = inet6_ioctl,               /* must change  */
 532        .listen            = inet_listen,               /* ok           */
 533        .shutdown          = inet_shutdown,             /* ok           */
 534        .setsockopt        = sock_common_setsockopt,    /* ok           */
 535        .getsockopt        = sock_common_getsockopt,    /* ok           */
 536        .sendmsg           = inet_sendmsg,              /* ok           */
 537        .recvmsg           = inet_recvmsg,              /* ok           */
 538        .mmap              = sock_no_mmap,
 539        .sendpage          = inet_sendpage,
 540        .splice_read       = tcp_splice_read,
 541#ifdef CONFIG_COMPAT
 542        .compat_setsockopt = compat_sock_common_setsockopt,
 543        .compat_getsockopt = compat_sock_common_getsockopt,
 544#endif
 545};
 546
 547const struct proto_ops inet6_dgram_ops = {
 548        .family            = PF_INET6,
 549        .owner             = THIS_MODULE,
 550        .release           = inet6_release,
 551        .bind              = inet6_bind,
 552        .connect           = inet_dgram_connect,        /* ok           */
 553        .socketpair        = sock_no_socketpair,        /* a do nothing */
 554        .accept            = sock_no_accept,            /* a do nothing */
 555        .getname           = inet6_getname,
 556        .poll              = udp_poll,                  /* ok           */
 557        .ioctl             = inet6_ioctl,               /* must change  */
 558        .listen            = sock_no_listen,            /* ok           */
 559        .shutdown          = inet_shutdown,             /* ok           */
 560        .setsockopt        = sock_common_setsockopt,    /* ok           */
 561        .getsockopt        = sock_common_getsockopt,    /* ok           */
 562        .sendmsg           = inet_sendmsg,              /* ok           */
 563        .recvmsg           = inet_recvmsg,              /* ok           */
 564        .mmap              = sock_no_mmap,
 565        .sendpage          = sock_no_sendpage,
 566#ifdef CONFIG_COMPAT
 567        .compat_setsockopt = compat_sock_common_setsockopt,
 568        .compat_getsockopt = compat_sock_common_getsockopt,
 569#endif
 570};
 571
 572static const struct net_proto_family inet6_family_ops = {
 573        .family = PF_INET6,
 574        .create = inet6_create,
 575        .owner  = THIS_MODULE,
 576};
 577
 578int inet6_register_protosw(struct inet_protosw *p)
 579{
 580        struct list_head *lh;
 581        struct inet_protosw *answer;
 582        struct list_head *last_perm;
 583        int protocol = p->protocol;
 584        int ret;
 585
 586        spin_lock_bh(&inetsw6_lock);
 587
 588        ret = -EINVAL;
 589        if (p->type >= SOCK_MAX)
 590                goto out_illegal;
 591
 592        /* If we are trying to override a permanent protocol, bail. */
 593        answer = NULL;
 594        ret = -EPERM;
 595        last_perm = &inetsw6[p->type];
 596        list_for_each(lh, &inetsw6[p->type]) {
 597                answer = list_entry(lh, struct inet_protosw, list);
 598
 599                /* Check only the non-wild match. */
 600                if (INET_PROTOSW_PERMANENT & answer->flags) {
 601                        if (protocol == answer->protocol)
 602                                break;
 603                        last_perm = lh;
 604                }
 605
 606                answer = NULL;
 607        }
 608        if (answer)
 609                goto out_permanent;
 610
 611        /* Add the new entry after the last permanent entry if any, so that
 612         * the new entry does not override a permanent entry when matched with
 613         * a wild-card protocol. But it is allowed to override any existing
 614         * non-permanent entry.  This means that when we remove this entry, the
 615         * system automatically returns to the old behavior.
 616         */
 617        list_add_rcu(&p->list, last_perm);
 618        ret = 0;
 619out:
 620        spin_unlock_bh(&inetsw6_lock);
 621        return ret;
 622
 623out_permanent:
 624        pr_err("Attempt to override permanent protocol %d\n", protocol);
 625        goto out;
 626
 627out_illegal:
 628        pr_err("Ignoring attempt to register invalid socket type %d\n",
 629               p->type);
 630        goto out;
 631}
 632EXPORT_SYMBOL(inet6_register_protosw);
 633
 634void
 635inet6_unregister_protosw(struct inet_protosw *p)
 636{
 637        if (INET_PROTOSW_PERMANENT & p->flags) {
 638                pr_err("Attempt to unregister permanent protocol %d\n",
 639                       p->protocol);
 640        } else {
 641                spin_lock_bh(&inetsw6_lock);
 642                list_del_rcu(&p->list);
 643                spin_unlock_bh(&inetsw6_lock);
 644
 645                synchronize_net();
 646        }
 647}
 648EXPORT_SYMBOL(inet6_unregister_protosw);
 649
 650int inet6_sk_rebuild_header(struct sock *sk)
 651{
 652        struct ipv6_pinfo *np = inet6_sk(sk);
 653        struct dst_entry *dst;
 654
 655        dst = __sk_dst_check(sk, np->dst_cookie);
 656
 657        if (dst == NULL) {
 658                struct inet_sock *inet = inet_sk(sk);
 659                struct in6_addr *final_p, final;
 660                struct flowi6 fl6;
 661
 662                memset(&fl6, 0, sizeof(fl6));
 663                fl6.flowi6_proto = sk->sk_protocol;
 664                fl6.daddr = sk->sk_v6_daddr;
 665                fl6.saddr = np->saddr;
 666                fl6.flowlabel = np->flow_label;
 667                fl6.flowi6_oif = sk->sk_bound_dev_if;
 668                fl6.flowi6_mark = sk->sk_mark;
 669                fl6.fl6_dport = inet->inet_dport;
 670                fl6.fl6_sport = inet->inet_sport;
 671                security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 672
 673                rcu_read_lock();
 674                final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 675                                         &final);
 676                rcu_read_unlock();
 677
 678                dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 679                if (IS_ERR(dst)) {
 680                        sk->sk_route_caps = 0;
 681                        sk->sk_err_soft = -PTR_ERR(dst);
 682                        return PTR_ERR(dst);
 683                }
 684
 685                ip6_dst_store(sk, dst, NULL, NULL);
 686        }
 687
 688        return 0;
 689}
 690EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 691
 692bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
 693{
 694        const struct ipv6_pinfo *np = inet6_sk(sk);
 695        const struct inet6_skb_parm *opt = IP6CB(skb);
 696
 697        if (np->rxopt.all) {
 698                if ((opt->hop && (np->rxopt.bits.hopopts ||
 699                                  np->rxopt.bits.ohopopts)) ||
 700                    ((IPV6_FLOWINFO_MASK &
 701                      *(__be32 *)skb_network_header(skb)) &&
 702                     np->rxopt.bits.rxflow) ||
 703                    (opt->srcrt && (np->rxopt.bits.srcrt ||
 704                     np->rxopt.bits.osrcrt)) ||
 705                    ((opt->dst1 || opt->dst0) &&
 706                     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 707                        return true;
 708        }
 709        return false;
 710}
 711EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 712
 713static struct packet_type ipv6_packet_type __read_mostly = {
 714        .type = cpu_to_be16(ETH_P_IPV6),
 715        .func = ipv6_rcv,
 716};
 717
 718static int __init ipv6_packet_init(void)
 719{
 720        dev_add_pack(&ipv6_packet_type);
 721        return 0;
 722}
 723
 724static void ipv6_packet_cleanup(void)
 725{
 726        dev_remove_pack(&ipv6_packet_type);
 727}
 728
 729static int __net_init ipv6_init_mibs(struct net *net)
 730{
 731        if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
 732                          sizeof(struct udp_mib),
 733                          __alignof__(struct udp_mib)) < 0)
 734                return -ENOMEM;
 735        if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
 736                          sizeof(struct udp_mib),
 737                          __alignof__(struct udp_mib)) < 0)
 738                goto err_udplite_mib;
 739        if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
 740                          sizeof(struct ipstats_mib),
 741                          __alignof__(struct ipstats_mib)) < 0)
 742                goto err_ip_mib;
 743        if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
 744                          sizeof(struct icmpv6_mib),
 745                          __alignof__(struct icmpv6_mib)) < 0)
 746                goto err_icmp_mib;
 747        net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 748                                                GFP_KERNEL);
 749        if (!net->mib.icmpv6msg_statistics)
 750                goto err_icmpmsg_mib;
 751        return 0;
 752
 753err_icmpmsg_mib:
 754        snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 755err_icmp_mib:
 756        snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 757err_ip_mib:
 758        snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 759err_udplite_mib:
 760        snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 761        return -ENOMEM;
 762}
 763
 764static void ipv6_cleanup_mibs(struct net *net)
 765{
 766        snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 767        snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 768        snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 769        snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 770        kfree(net->mib.icmpv6msg_statistics);
 771}
 772
 773static int __net_init inet6_net_init(struct net *net)
 774{
 775        int err = 0;
 776
 777        net->ipv6.sysctl.bindv6only = 0;
 778        net->ipv6.sysctl.icmpv6_time = 1*HZ;
 779        net->idgen_retries = 3;
 780        net->idgen_delay = 1 * HZ;
 781
 782        err = ipv6_init_mibs(net);
 783        if (err)
 784                return err;
 785#ifdef CONFIG_PROC_FS
 786        err = udp6_proc_init(net);
 787        if (err)
 788                goto out;
 789        err = tcp6_proc_init(net);
 790        if (err)
 791                goto proc_tcp6_fail;
 792        err = ac6_proc_init(net);
 793        if (err)
 794                goto proc_ac6_fail;
 795#endif
 796        return err;
 797
 798#ifdef CONFIG_PROC_FS
 799proc_ac6_fail:
 800        tcp6_proc_exit(net);
 801proc_tcp6_fail:
 802        udp6_proc_exit(net);
 803out:
 804        ipv6_cleanup_mibs(net);
 805        return err;
 806#endif
 807}
 808
 809static void __net_exit inet6_net_exit(struct net *net)
 810{
 811#ifdef CONFIG_PROC_FS
 812        udp6_proc_exit(net);
 813        tcp6_proc_exit(net);
 814        ac6_proc_exit(net);
 815#endif
 816        ipv6_cleanup_mibs(net);
 817}
 818
 819static struct pernet_operations inet6_net_ops = {
 820        .init = inet6_net_init,
 821        .exit = inet6_net_exit,
 822};
 823
 824static const struct ipv6_stub ipv6_stub_impl = {
 825        .ipv6_sock_mc_join = ipv6_sock_mc_join,
 826        .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 827        .ipv6_dst_lookup = ip6_dst_lookup,
 828        .udpv6_encap_enable = udpv6_encap_enable,
 829        .ndisc_send_na = ndisc_send_na,
 830        .nd_tbl = &nd_tbl,
 831};
 832
 833static int __init inet6_init(void)
 834{
 835        struct list_head *r;
 836        int err = 0;
 837
 838        BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
 839
 840        /* Register the socket-side information for inet6_create.  */
 841        for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 842                INIT_LIST_HEAD(r);
 843
 844        if (disable_ipv6_mod) {
 845                pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 846                goto out;
 847        }
 848
 849        err = proto_register(&tcpv6_prot, 1);
 850        if (err)
 851                goto out;
 852
 853        err = proto_register(&udpv6_prot, 1);
 854        if (err)
 855                goto out_unregister_tcp_proto;
 856
 857        err = proto_register(&udplitev6_prot, 1);
 858        if (err)
 859                goto out_unregister_udp_proto;
 860
 861        err = proto_register(&rawv6_prot, 1);
 862        if (err)
 863                goto out_unregister_udplite_proto;
 864
 865
 866        /* We MUST register RAW sockets before we create the ICMP6,
 867         * IGMP6, or NDISC control sockets.
 868         */
 869        err = rawv6_init();
 870        if (err)
 871                goto out_unregister_raw_proto;
 872
 873        /* Register the family here so that the init calls below will
 874         * be able to create sockets. (?? is this dangerous ??)
 875         */
 876        err = sock_register(&inet6_family_ops);
 877        if (err)
 878                goto out_sock_register_fail;
 879
 880        tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
 881
 882        /*
 883         *      ipngwg API draft makes clear that the correct semantics
 884         *      for TCP and UDP is to consider one TCP and UDP instance
 885         *      in a host available by both INET and INET6 APIs and
 886         *      able to communicate via both network protocols.
 887         */
 888
 889        err = register_pernet_subsys(&inet6_net_ops);
 890        if (err)
 891                goto register_pernet_fail;
 892        err = icmpv6_init();
 893        if (err)
 894                goto icmp_fail;
 895        err = ip6_mr_init();
 896        if (err)
 897                goto ipmr_fail;
 898        err = ndisc_init();
 899        if (err)
 900                goto ndisc_fail;
 901        err = igmp6_init();
 902        if (err)
 903                goto igmp_fail;
 904
 905        ipv6_stub = &ipv6_stub_impl;
 906
 907        err = ipv6_netfilter_init();
 908        if (err)
 909                goto netfilter_fail;
 910        /* Create /proc/foo6 entries. */
 911#ifdef CONFIG_PROC_FS
 912        err = -ENOMEM;
 913        if (raw6_proc_init())
 914                goto proc_raw6_fail;
 915        if (udplite6_proc_init())
 916                goto proc_udplite6_fail;
 917        if (ipv6_misc_proc_init())
 918                goto proc_misc6_fail;
 919        if (if6_proc_init())
 920                goto proc_if6_fail;
 921#endif
 922        err = ip6_route_init();
 923        if (err)
 924                goto ip6_route_fail;
 925        err = ndisc_late_init();
 926        if (err)
 927                goto ndisc_late_fail;
 928        err = ip6_flowlabel_init();
 929        if (err)
 930                goto ip6_flowlabel_fail;
 931        err = addrconf_init();
 932        if (err)
 933                goto addrconf_fail;
 934
 935        /* Init v6 extension headers. */
 936        err = ipv6_exthdrs_init();
 937        if (err)
 938                goto ipv6_exthdrs_fail;
 939
 940        err = ipv6_frag_init();
 941        if (err)
 942                goto ipv6_frag_fail;
 943
 944        /* Init v6 transport protocols. */
 945        err = udpv6_init();
 946        if (err)
 947                goto udpv6_fail;
 948
 949        err = udplitev6_init();
 950        if (err)
 951                goto udplitev6_fail;
 952
 953        err = udpv6_offload_init();
 954        if (err)
 955                goto udpv6_offload_fail;
 956
 957        err = tcpv6_init();
 958        if (err)
 959                goto tcpv6_fail;
 960
 961        err = ipv6_packet_init();
 962        if (err)
 963                goto ipv6_packet_fail;
 964
 965#ifdef CONFIG_SYSCTL
 966        err = ipv6_sysctl_register();
 967        if (err)
 968                goto sysctl_fail;
 969#endif
 970out:
 971        return err;
 972
 973#ifdef CONFIG_SYSCTL
 974sysctl_fail:
 975        ipv6_packet_cleanup();
 976#endif
 977ipv6_packet_fail:
 978        tcpv6_exit();
 979tcpv6_fail:
 980        udpv6_offload_exit();
 981udpv6_offload_fail:
 982        udplitev6_exit();
 983udplitev6_fail:
 984        udpv6_exit();
 985udpv6_fail:
 986        ipv6_frag_exit();
 987ipv6_frag_fail:
 988        ipv6_exthdrs_exit();
 989ipv6_exthdrs_fail:
 990        addrconf_cleanup();
 991addrconf_fail:
 992        ip6_flowlabel_cleanup();
 993ip6_flowlabel_fail:
 994        ndisc_late_cleanup();
 995ndisc_late_fail:
 996        ip6_route_cleanup();
 997ip6_route_fail:
 998#ifdef CONFIG_PROC_FS
 999        if6_proc_exit();
1000proc_if6_fail:
1001        ipv6_misc_proc_exit();
1002proc_misc6_fail:
1003        udplite6_proc_exit();
1004proc_udplite6_fail:
1005        raw6_proc_exit();
1006proc_raw6_fail:
1007#endif
1008        ipv6_netfilter_fini();
1009netfilter_fail:
1010        igmp6_cleanup();
1011igmp_fail:
1012        ndisc_cleanup();
1013ndisc_fail:
1014        ip6_mr_cleanup();
1015ipmr_fail:
1016        icmpv6_cleanup();
1017icmp_fail:
1018        unregister_pernet_subsys(&inet6_net_ops);
1019register_pernet_fail:
1020        sock_unregister(PF_INET6);
1021        rtnl_unregister_all(PF_INET6);
1022out_sock_register_fail:
1023        rawv6_exit();
1024out_unregister_raw_proto:
1025        proto_unregister(&rawv6_prot);
1026out_unregister_udplite_proto:
1027        proto_unregister(&udplitev6_prot);
1028out_unregister_udp_proto:
1029        proto_unregister(&udpv6_prot);
1030out_unregister_tcp_proto:
1031        proto_unregister(&tcpv6_prot);
1032        goto out;
1033}
1034module_init(inet6_init);
1035
1036static void __exit inet6_exit(void)
1037{
1038        if (disable_ipv6_mod)
1039                return;
1040
1041        /* First of all disallow new sockets creation. */
1042        sock_unregister(PF_INET6);
1043        /* Disallow any further netlink messages */
1044        rtnl_unregister_all(PF_INET6);
1045
1046        udpv6_exit();
1047        udplitev6_exit();
1048        tcpv6_exit();
1049
1050        /* Cleanup code parts. */
1051        ipv6_packet_cleanup();
1052        ipv6_frag_exit();
1053        ipv6_exthdrs_exit();
1054        addrconf_cleanup();
1055        ip6_flowlabel_cleanup();
1056        ndisc_late_cleanup();
1057        ip6_route_cleanup();
1058#ifdef CONFIG_PROC_FS
1059
1060        /* Cleanup code parts. */
1061        if6_proc_exit();
1062        ipv6_misc_proc_exit();
1063        udplite6_proc_exit();
1064        raw6_proc_exit();
1065#endif
1066        ipv6_netfilter_fini();
1067        ipv6_stub = NULL;
1068        igmp6_cleanup();
1069        ndisc_cleanup();
1070        ip6_mr_cleanup();
1071        icmpv6_cleanup();
1072        rawv6_exit();
1073
1074        unregister_pernet_subsys(&inet6_net_ops);
1075        proto_unregister(&rawv6_prot);
1076        proto_unregister(&udplitev6_prot);
1077        proto_unregister(&udpv6_prot);
1078        proto_unregister(&tcpv6_prot);
1079
1080        rcu_barrier(); /* Wait for completion of call_rcu()'s */
1081}
1082module_exit(inet6_exit);
1083
1084MODULE_ALIAS_NETPROTO(PF_INET6);
1085