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 #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/ping.h>
  53 #include <net/protocol.h>
  54 #include <net/inet_common.h>
  55 #include <net/route.h>
  56 #include <net/transp_v6.h>
  57 #include <net/ip6_route.h>
  58 #include <net/addrconf.h>
  59 #include <net/ndisc.h>
  60 #ifdef CONFIG_IPV6_TUNNEL
  61 #include <net/ip6_tunnel.h>
  62 #endif
  63
  64 #include <asm/uaccess.h>
  65 #include <linux/mroute6.h>
  66
  67 MODULE_AUTHOR("Cast of dozens");
  68 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  69 MODULE_LICENSE("GPL");
  70
  71 /* The inetsw6 table contains everything that inet6_create needs to
  72  * build a new socket.
  73  */
  74 static struct list_head inetsw6[SOCK_MAX];
  75 static DEFINE_SPINLOCK(inetsw6_lock);
  76
  77 struct ipv6_params ipv6_defaults = {
  78         .disable_ipv6 = 0,
  79         .autoconf = 1,
  80 };
  81
  82 static int disable_ipv6_mod;
  83
  84 module_param_named(disable, disable_ipv6_mod, int, 0444);
  85 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  86
  87 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  88 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  89
  90 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  91 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  92
  93 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  94 {
  95         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  96
  97         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  98 }
  99
 100 static int inet6_create(struct net *net, struct socket *sock, int protocol,
 101                         int kern)
 102 {
 103         struct inet_sock *inet;
 104         struct ipv6_pinfo *np;
 105         struct sock *sk;
 106         struct inet_protosw *answer;
 107         struct proto *answer_prot;
 108         unsigned char answer_flags;
 109         int try_loading_module = 0;
 110         int err;
 111
 112         if (protocol < 0 || protocol >= IPPROTO_MAX)
 113                 return -EINVAL;
 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 (sock->type == SOCK_RAW && !kern &&
 162             !ns_capable(net->user_ns, CAP_NET_RAW))
 163                 goto out_rcu_unlock;
 164
 165         sock->ops = answer->ops;
 166         answer_prot = answer->prot;
 167         answer_flags = answer->flags;
 168         rcu_read_unlock();
 169
 170         WARN_ON(!answer_prot->slab);
 171
 172         err = -ENOBUFS;
 173         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
 174         if (!sk)
 175                 goto out;
 176
 177         sock_init_data(sock, sk);
 178
 179         err = 0;
 180         if (INET_PROTOSW_REUSE & answer_flags)
 181                 sk->sk_reuse = SK_CAN_REUSE;
 182
 183         inet = inet_sk(sk);
 184         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 185
 186         if (SOCK_RAW == sock->type) {
 187                 inet->inet_num = protocol;
 188                 if (IPPROTO_RAW == protocol)
 189                         inet->hdrincl = 1;
 190         }
 191
 192         sk->sk_destruct         = inet_sock_destruct;
 193         sk->sk_family           = PF_INET6;
 194         sk->sk_protocol         = protocol;
 195
 196         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 197
 198         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 199         np->hop_limit   = -1;
 200         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 201         np->mc_loop     = 1;
 202         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 203         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
 204
 205         /* Init the ipv4 part of the socket since we can have sockets
 206          * using v6 API for ipv4.
 207          */
 208         inet->uc_ttl    = -1;
 209
 210         inet->mc_loop   = 1;
 211         inet->mc_ttl    = 1;
 212         inet->mc_index  = 0;
 213         inet->mc_list   = NULL;
 214         inet->rcv_tos   = 0;
 215
 216         if (net->ipv4.sysctl_ip_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->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->inet_sport = htons(inet->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
 274         if (addr->sin6_family != AF_INET6)
 275                 return -EAFNOSUPPORT;
 276
 277         addr_type = ipv6_addr_type(&addr->sin6_addr);
 278         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 279                 return -EINVAL;
 280
 281         snum = ntohs(addr->sin6_port);
 282         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 283                 return -EACCES;
 284
 285         lock_sock(sk);
 286
 287         /* Check these errors (active socket, double bind). */
 288         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 289                 err = -EINVAL;
 290                 goto out;
 291         }
 292
 293         /* Check if the address belongs to the host. */
 294         if (addr_type == IPV6_ADDR_MAPPED) {
 295                 int chk_addr_ret;
 296
 297                 /* Binding to v4-mapped address on a v6-only socket
 298                  * makes no sense
 299                  */
 300                 if (sk->sk_ipv6only) {
 301                         err = -EINVAL;
 302                         goto out;
 303                 }
 304
 305                 /* Reproduce AF_INET checks to make the bindings consistent */
 306                 v4addr = addr->sin6_addr.s6_addr32[3];
 307                 chk_addr_ret = inet_addr_type(net, v4addr);
 308                 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
 309                     !(inet->freebind || inet->transparent) &&
 310                     v4addr != htonl(INADDR_ANY) &&
 311                     chk_addr_ret != RTN_LOCAL &&
 312                     chk_addr_ret != RTN_MULTICAST &&
 313                     chk_addr_ret != RTN_BROADCAST) {
 314                         err = -EADDRNOTAVAIL;
 315                         goto out;
 316                 }
 317         } else {
 318                 if (addr_type != IPV6_ADDR_ANY) {
 319                         struct net_device *dev = NULL;
 320
 321                         rcu_read_lock();
 322                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 323                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 324                                     addr->sin6_scope_id) {
 325                                         /* Override any existing binding, if another one
 326                                          * is supplied by user.
 327                                          */
 328                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 329                                 }
 330
 331                                 /* Binding to link-local address requires an interface */
 332                                 if (!sk->sk_bound_dev_if) {
 333                                         err = -EINVAL;
 334                                         goto out_unlock;
 335                                 }
 336                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 337                                 if (!dev) {
 338                                         err = -ENODEV;
 339                                         goto out_unlock;
 340                                 }
 341                         }
 342
 343                         /* ipv4 addr of the socket is invalid.  Only the
 344                          * unspecified and mapped address have a v4 equivalent.
 345                          */
 346                         v4addr = LOOPBACK4_IPV6;
 347                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 348                                 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
 349                                     !(inet->freebind || inet->transparent) &&
 350                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 351                                                    dev, 0)) {
 352                                         err = -EADDRNOTAVAIL;
 353                                         goto out_unlock;
 354                                 }
 355                         }
 356                         rcu_read_unlock();
 357                 }
 358         }
 359
 360         inet->inet_rcv_saddr = v4addr;
 361         inet->inet_saddr = v4addr;
 362
 363         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 364
 365         if (!(addr_type & IPV6_ADDR_MULTICAST))
 366                 np->saddr = addr->sin6_addr;
 367
 368         /* Make sure we are allowed to bind here. */
 369         if ((snum || !inet->bind_address_no_port) &&
 370             sk->sk_prot->get_port(sk, snum)) {
 371                 inet_reset_saddr(sk);
 372                 err = -EADDRINUSE;
 373                 goto out;
 374         }
 375
 376         if (addr_type != IPV6_ADDR_ANY) {
 377                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 378                 if (addr_type != IPV6_ADDR_MAPPED)
 379                         sk->sk_ipv6only = 1;
 380         }
 381         if (snum)
 382                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 383         inet->inet_sport = htons(inet->inet_num);
 384         inet->inet_dport = 0;
 385         inet->inet_daddr = 0;
 386 out:
 387         release_sock(sk);
 388         return err;
 389 out_unlock:
 390         rcu_read_unlock();
 391         goto out;
 392 }
 393 EXPORT_SYMBOL(inet6_bind);
 394
 395 int inet6_release(struct socket *sock)
 396 {
 397         struct sock *sk = sock->sk;
 398
 399         if (!sk)
 400                 return -EINVAL;
 401
 402         /* Free mc lists */
 403         ipv6_sock_mc_close(sk);
 404
 405         /* Free ac lists */
 406         ipv6_sock_ac_close(sk);
 407
 408         return inet_release(sock);
 409 }
 410 EXPORT_SYMBOL(inet6_release);
 411
 412 void inet6_destroy_sock(struct sock *sk)
 413 {
 414         struct ipv6_pinfo *np = inet6_sk(sk);
 415         struct sk_buff *skb;
 416         struct ipv6_txoptions *opt;
 417
 418         /* Release rx options */
 419
 420         skb = xchg(&np->pktoptions, NULL);
 421         if (skb)
 422                 kfree_skb(skb);
 423
 424         skb = xchg(&np->rxpmtu, NULL);
 425         if (skb)
 426                 kfree_skb(skb);
 427
 428         /* Free flowlabels */
 429         fl6_free_socklist(sk);
 430
 431         /* Free tx options */
 432
 433         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 434         if (opt) {
 435                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 436                 txopt_put(opt);
 437         }
 438 }
 439 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 440
 441 /*
 442  *      This does both peername and sockname.
 443  */
 444
 445 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 446                  int *uaddr_len, int peer)
 447 {
 448         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 449         struct sock *sk = sock->sk;
 450         struct inet_sock *inet = inet_sk(sk);
 451         struct ipv6_pinfo *np = inet6_sk(sk);
 452
 453         sin->sin6_family = AF_INET6;
 454         sin->sin6_flowinfo = 0;
 455         sin->sin6_scope_id = 0;
 456         if (peer) {
 457                 if (!inet->inet_dport)
 458                         return -ENOTCONN;
 459                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 460                     peer == 1)
 461                         return -ENOTCONN;
 462                 sin->sin6_port = inet->inet_dport;
 463                 sin->sin6_addr = sk->sk_v6_daddr;
 464                 if (np->sndflow)
 465                         sin->sin6_flowinfo = np->flow_label;
 466         } else {
 467                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 468                         sin->sin6_addr = np->saddr;
 469                 else
 470                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 471
 472                 sin->sin6_port = inet->inet_sport;
 473         }
 474         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 475                                                  sk->sk_bound_dev_if);
 476         *uaddr_len = sizeof(*sin);
 477         return 0;
 478 }
 479 EXPORT_SYMBOL(inet6_getname);
 480
 481 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 482 {
 483         struct sock *sk = sock->sk;
 484         struct net *net = sock_net(sk);
 485
 486         switch (cmd) {
 487         case SIOCGSTAMP:
 488                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 489
 490         case SIOCGSTAMPNS:
 491                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 492
 493         case SIOCADDRT:
 494         case SIOCDELRT:
 495
 496                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 497
 498         case SIOCSIFADDR:
 499                 return addrconf_add_ifaddr(net, (void __user *) arg);
 500         case SIOCDIFADDR:
 501                 return addrconf_del_ifaddr(net, (void __user *) arg);
 502         case SIOCSIFDSTADDR:
 503                 return addrconf_set_dstaddr(net, (void __user *) arg);
 504         default:
 505                 if (!sk->sk_prot->ioctl)
 506                         return -ENOIOCTLCMD;
 507                 return sk->sk_prot->ioctl(sk, cmd, arg);
 508         }
 509         /*NOTREACHED*/
 510         return 0;
 511 }
 512 EXPORT_SYMBOL(inet6_ioctl);
 513
 514 const struct proto_ops inet6_stream_ops = {
 515         .family            = PF_INET6,
 516         .owner             = THIS_MODULE,
 517         .release           = inet6_release,
 518         .bind              = inet6_bind,
 519         .connect           = inet_stream_connect,       /* ok           */
 520         .socketpair        = sock_no_socketpair,        /* a do nothing */
 521         .accept            = inet_accept,               /* ok           */
 522         .getname           = inet6_getname,
 523         .poll              = tcp_poll,                  /* ok           */
 524         .ioctl             = inet6_ioctl,               /* must change  */
 525         .listen            = inet_listen,               /* ok           */
 526         .shutdown          = inet_shutdown,             /* ok           */
 527         .setsockopt        = sock_common_setsockopt,    /* ok           */
 528         .getsockopt        = sock_common_getsockopt,    /* ok           */
 529         .sendmsg           = inet_sendmsg,              /* ok           */
 530         .recvmsg           = inet_recvmsg,              /* ok           */
 531         .mmap              = sock_no_mmap,
 532         .sendpage          = inet_sendpage,
 533         .splice_read       = tcp_splice_read,
 534 #ifdef CONFIG_COMPAT
 535         .compat_setsockopt = compat_sock_common_setsockopt,
 536         .compat_getsockopt = compat_sock_common_getsockopt,
 537 #endif
 538 };
 539
 540 const struct proto_ops inet6_dgram_ops = {
 541         .family            = PF_INET6,
 542         .owner             = THIS_MODULE,
 543         .release           = inet6_release,
 544         .bind              = inet6_bind,
 545         .connect           = inet_dgram_connect,        /* ok           */
 546         .socketpair        = sock_no_socketpair,        /* a do nothing */
 547         .accept            = sock_no_accept,            /* a do nothing */
 548         .getname           = inet6_getname,
 549         .poll              = udp_poll,                  /* ok           */
 550         .ioctl             = inet6_ioctl,               /* must change  */
 551         .listen            = sock_no_listen,            /* ok           */
 552         .shutdown          = inet_shutdown,             /* ok           */
 553         .setsockopt        = sock_common_setsockopt,    /* ok           */
 554         .getsockopt        = sock_common_getsockopt,    /* ok           */
 555         .sendmsg           = inet_sendmsg,              /* ok           */
 556         .recvmsg           = inet_recvmsg,              /* ok           */
 557         .mmap              = sock_no_mmap,
 558         .sendpage          = sock_no_sendpage,
 559 #ifdef CONFIG_COMPAT
 560         .compat_setsockopt = compat_sock_common_setsockopt,
 561         .compat_getsockopt = compat_sock_common_getsockopt,
 562 #endif
 563 };
 564
 565 static const struct net_proto_family inet6_family_ops = {
 566         .family = PF_INET6,
 567         .create = inet6_create,
 568         .owner  = THIS_MODULE,
 569 };
 570
 571 int inet6_register_protosw(struct inet_protosw *p)
 572 {
 573         struct list_head *lh;
 574         struct inet_protosw *answer;
 575         struct list_head *last_perm;
 576         int protocol = p->protocol;
 577         int ret;
 578
 579         spin_lock_bh(&inetsw6_lock);
 580
 581         ret = -EINVAL;
 582         if (p->type >= SOCK_MAX)
 583                 goto out_illegal;
 584
 585         /* If we are trying to override a permanent protocol, bail. */
 586         answer = NULL;
 587         ret = -EPERM;
 588         last_perm = &inetsw6[p->type];
 589         list_for_each(lh, &inetsw6[p->type]) {
 590                 answer = list_entry(lh, struct inet_protosw, list);
 591
 592                 /* Check only the non-wild match. */
 593                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 594                         if (protocol == answer->protocol)
 595                                 break;
 596                         last_perm = lh;
 597                 }
 598
 599                 answer = NULL;
 600         }
 601         if (answer)
 602                 goto out_permanent;
 603
 604         /* Add the new entry after the last permanent entry if any, so that
 605          * the new entry does not override a permanent entry when matched with
 606          * a wild-card protocol. But it is allowed to override any existing
 607          * non-permanent entry.  This means that when we remove this entry, the
 608          * system automatically returns to the old behavior.
 609          */
 610         list_add_rcu(&p->list, last_perm);
 611         ret = 0;
 612 out:
 613         spin_unlock_bh(&inetsw6_lock);
 614         return ret;
 615
 616 out_permanent:
 617         pr_err("Attempt to override permanent protocol %d\n", protocol);
 618         goto out;
 619
 620 out_illegal:
 621         pr_err("Ignoring attempt to register invalid socket type %d\n",
 622                p->type);
 623         goto out;
 624 }
 625 EXPORT_SYMBOL(inet6_register_protosw);
 626
 627 void
 628 inet6_unregister_protosw(struct inet_protosw *p)
 629 {
 630         if (INET_PROTOSW_PERMANENT & p->flags) {
 631                 pr_err("Attempt to unregister permanent protocol %d\n",
 632                        p->protocol);
 633         } else {
 634                 spin_lock_bh(&inetsw6_lock);
 635                 list_del_rcu(&p->list);
 636                 spin_unlock_bh(&inetsw6_lock);
 637
 638                 synchronize_net();
 639         }
 640 }
 641 EXPORT_SYMBOL(inet6_unregister_protosw);
 642
 643 int inet6_sk_rebuild_header(struct sock *sk)
 644 {
 645         struct ipv6_pinfo *np = inet6_sk(sk);
 646         struct dst_entry *dst;
 647
 648         dst = __sk_dst_check(sk, np->dst_cookie);
 649
 650         if (!dst) {
 651                 struct inet_sock *inet = inet_sk(sk);
 652                 struct in6_addr *final_p, final;
 653                 struct flowi6 fl6;
 654
 655                 memset(&fl6, 0, sizeof(fl6));
 656                 fl6.flowi6_proto = sk->sk_protocol;
 657                 fl6.daddr = sk->sk_v6_daddr;
 658                 fl6.saddr = np->saddr;
 659                 fl6.flowlabel = np->flow_label;
 660                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 661                 fl6.flowi6_mark = sk->sk_mark;
 662                 fl6.fl6_dport = inet->inet_dport;
 663                 fl6.fl6_sport = inet->inet_sport;
 664                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 665
 666                 rcu_read_lock();
 667                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 668                                          &final);
 669                 rcu_read_unlock();
 670
 671                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 672                 if (IS_ERR(dst)) {
 673                         sk->sk_route_caps = 0;
 674                         sk->sk_err_soft = -PTR_ERR(dst);
 675                         return PTR_ERR(dst);
 676                 }
 677
 678                 ip6_dst_store(sk, dst, NULL, NULL);
 679         }
 680
 681         return 0;
 682 }
 683 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 684
 685 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 686                        const struct inet6_skb_parm *opt)
 687 {
 688         const struct ipv6_pinfo *np = inet6_sk(sk);
 689
 690         if (np->rxopt.all) {
 691                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
 692                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 693                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 694                      np->rxopt.bits.rxflow) ||
 695                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 696                      np->rxopt.bits.osrcrt)) ||
 697                     ((opt->dst1 || opt->dst0) &&
 698                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 699                         return true;
 700         }
 701         return false;
 702 }
 703 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 704
 705 static struct packet_type ipv6_packet_type __read_mostly = {
 706         .type = cpu_to_be16(ETH_P_IPV6),
 707         .func = ipv6_rcv,
 708 };
 709
 710 static int __init ipv6_packet_init(void)
 711 {
 712         dev_add_pack(&ipv6_packet_type);
 713         return 0;
 714 }
 715
 716 static void ipv6_packet_cleanup(void)
 717 {
 718         dev_remove_pack(&ipv6_packet_type);
 719 }
 720
 721 static int __net_init ipv6_init_mibs(struct net *net)
 722 {
 723         int i;
 724
 725         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 726         if (!net->mib.udp_stats_in6)
 727                 return -ENOMEM;
 728         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 729         if (!net->mib.udplite_stats_in6)
 730                 goto err_udplite_mib;
 731         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 732         if (!net->mib.ipv6_statistics)
 733                 goto err_ip_mib;
 734
 735         for_each_possible_cpu(i) {
 736                 struct ipstats_mib *af_inet6_stats;
 737                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 738                 u64_stats_init(&af_inet6_stats->syncp);
 739         }
 740
 741
 742         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 743         if (!net->mib.icmpv6_statistics)
 744                 goto err_icmp_mib;
 745         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 746                                                 GFP_KERNEL);
 747         if (!net->mib.icmpv6msg_statistics)
 748                 goto err_icmpmsg_mib;
 749         return 0;
 750
 751 err_icmpmsg_mib:
 752         free_percpu(net->mib.icmpv6_statistics);
 753 err_icmp_mib:
 754         free_percpu(net->mib.ipv6_statistics);
 755 err_ip_mib:
 756         free_percpu(net->mib.udplite_stats_in6);
 757 err_udplite_mib:
 758         free_percpu(net->mib.udp_stats_in6);
 759         return -ENOMEM;
 760 }
 761
 762 static void ipv6_cleanup_mibs(struct net *net)
 763 {
 764         free_percpu(net->mib.udp_stats_in6);
 765         free_percpu(net->mib.udplite_stats_in6);
 766         free_percpu(net->mib.ipv6_statistics);
 767         free_percpu(net->mib.icmpv6_statistics);
 768         kfree(net->mib.icmpv6msg_statistics);
 769 }
 770
 771 static int __net_init inet6_net_init(struct net *net)
 772 {
 773         int err = 0;
 774
 775         net->ipv6.sysctl.bindv6only = 0;
 776         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 777         net->ipv6.sysctl.flowlabel_consistency = 1;
 778         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
 779         net->ipv6.sysctl.idgen_retries = 3;
 780         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 781         net->ipv6.sysctl.flowlabel_state_ranges = 0;
 782         atomic_set(&net->ipv6.fib6_sernum, 1);
 783
 784         err = ipv6_init_mibs(net);
 785         if (err)
 786                 return err;
 787 #ifdef CONFIG_PROC_FS
 788         err = udp6_proc_init(net);
 789         if (err)
 790                 goto out;
 791         err = tcp6_proc_init(net);
 792         if (err)
 793                 goto proc_tcp6_fail;
 794         err = ac6_proc_init(net);
 795         if (err)
 796                 goto proc_ac6_fail;
 797 #endif
 798         return err;
 799
 800 #ifdef CONFIG_PROC_FS
 801 proc_ac6_fail:
 802         tcp6_proc_exit(net);
 803 proc_tcp6_fail:
 804         udp6_proc_exit(net);
 805 out:
 806         ipv6_cleanup_mibs(net);
 807         return err;
 808 #endif
 809 }
 810
 811 static void __net_exit inet6_net_exit(struct net *net)
 812 {
 813 #ifdef CONFIG_PROC_FS
 814         udp6_proc_exit(net);
 815         tcp6_proc_exit(net);
 816         ac6_proc_exit(net);
 817 #endif
 818         ipv6_cleanup_mibs(net);
 819 }
 820
 821 static struct pernet_operations inet6_net_ops = {
 822         .init = inet6_net_init,
 823         .exit = inet6_net_exit,
 824 };
 825
 826 static const struct ipv6_stub ipv6_stub_impl = {
 827         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 828         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 829         .ipv6_dst_lookup = ip6_dst_lookup,
 830         .udpv6_encap_enable = udpv6_encap_enable,
 831         .ndisc_send_na = ndisc_send_na,
 832         .nd_tbl = &nd_tbl,
 833 };
 834
 835 static int __init inet6_init(void)
 836 {
 837         struct list_head *r;
 838         int err = 0;
 839
 840         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 841
 842         /* Register the socket-side information for inet6_create.  */
 843         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 844                 INIT_LIST_HEAD(r);
 845
 846         if (disable_ipv6_mod) {
 847                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 848                 goto out;
 849         }
 850
 851         err = proto_register(&tcpv6_prot, 1);
 852         if (err)
 853                 goto out;
 854
 855         err = proto_register(&udpv6_prot, 1);
 856         if (err)
 857                 goto out_unregister_tcp_proto;
 858
 859         err = proto_register(&udplitev6_prot, 1);
 860         if (err)
 861                 goto out_unregister_udp_proto;
 862
 863         err = proto_register(&rawv6_prot, 1);
 864         if (err)
 865                 goto out_unregister_udplite_proto;
 866
 867         err = proto_register(&pingv6_prot, 1);
 868         if (err)
 869                 goto out_unregister_ping_proto;
 870
 871         /* We MUST register RAW sockets before we create the ICMP6,
 872          * IGMP6, or NDISC control sockets.
 873          */
 874         err = rawv6_init();
 875         if (err)
 876                 goto out_unregister_raw_proto;
 877
 878         /* Register the family here so that the init calls below will
 879          * be able to create sockets. (?? is this dangerous ??)
 880          */
 881         err = sock_register(&inet6_family_ops);
 882         if (err)
 883                 goto out_sock_register_fail;
 884
 885         /*
 886          *      ipngwg API draft makes clear that the correct semantics
 887          *      for TCP and UDP is to consider one TCP and UDP instance
 888          *      in a host available by both INET and INET6 APIs and
 889          *      able to communicate via both network protocols.
 890          */
 891
 892         err = register_pernet_subsys(&inet6_net_ops);
 893         if (err)
 894                 goto register_pernet_fail;
 895         err = ip6_mr_init();
 896         if (err)
 897                 goto ipmr_fail;
 898         err = icmpv6_init();
 899         if (err)
 900                 goto icmp_fail;
 901         err = ndisc_init();
 902         if (err)
 903                 goto ndisc_fail;
 904         err = igmp6_init();
 905         if (err)
 906                 goto igmp_fail;
 907
 908         ipv6_stub = &ipv6_stub_impl;
 909
 910         err = ipv6_netfilter_init();
 911         if (err)
 912                 goto netfilter_fail;
 913         /* Create /proc/foo6 entries. */
 914 #ifdef CONFIG_PROC_FS
 915         err = -ENOMEM;
 916         if (raw6_proc_init())
 917                 goto proc_raw6_fail;
 918         if (udplite6_proc_init())
 919                 goto proc_udplite6_fail;
 920         if (ipv6_misc_proc_init())
 921                 goto proc_misc6_fail;
 922         if (if6_proc_init())
 923                 goto proc_if6_fail;
 924 #endif
 925         err = ip6_route_init();
 926         if (err)
 927                 goto ip6_route_fail;
 928         err = ndisc_late_init();
 929         if (err)
 930                 goto ndisc_late_fail;
 931         err = ip6_flowlabel_init();
 932         if (err)
 933                 goto ip6_flowlabel_fail;
 934         err = addrconf_init();
 935         if (err)
 936                 goto addrconf_fail;
 937
 938         /* Init v6 extension headers. */
 939         err = ipv6_exthdrs_init();
 940         if (err)
 941                 goto ipv6_exthdrs_fail;
 942
 943         err = ipv6_frag_init();
 944         if (err)
 945                 goto ipv6_frag_fail;
 946
 947         /* Init v6 transport protocols. */
 948         err = udpv6_init();
 949         if (err)
 950                 goto udpv6_fail;
 951
 952         err = udplitev6_init();
 953         if (err)
 954                 goto udplitev6_fail;
 955
 956         err = tcpv6_init();
 957         if (err)
 958                 goto tcpv6_fail;
 959
 960         err = ipv6_packet_init();
 961         if (err)
 962                 goto ipv6_packet_fail;
 963
 964         err = pingv6_init();
 965         if (err)
 966                 goto pingv6_fail;
 967
 968 #ifdef CONFIG_SYSCTL
 969         err = ipv6_sysctl_register();
 970         if (err)
 971                 goto sysctl_fail;
 972 #endif
 973 out:
 974         return err;
 975
 976 #ifdef CONFIG_SYSCTL
 977 sysctl_fail:
 978         pingv6_exit();
 979 #endif
 980 pingv6_fail:
 981         ipv6_packet_cleanup();
 982 ipv6_packet_fail:
 983         tcpv6_exit();
 984 tcpv6_fail:
 985         udplitev6_exit();
 986 udplitev6_fail:
 987         udpv6_exit();
 988 udpv6_fail:
 989         ipv6_frag_exit();
 990 ipv6_frag_fail:
 991         ipv6_exthdrs_exit();
 992 ipv6_exthdrs_fail:
 993         addrconf_cleanup();
 994 addrconf_fail:
 995         ip6_flowlabel_cleanup();
 996 ip6_flowlabel_fail:
 997         ndisc_late_cleanup();
 998 ndisc_late_fail:
 999         ip6_route_cleanup();
1000 ip6_route_fail:
1001 #ifdef CONFIG_PROC_FS
1002         if6_proc_exit();
1003 proc_if6_fail:
1004         ipv6_misc_proc_exit();
1005 proc_misc6_fail:
1006         udplite6_proc_exit();
1007 proc_udplite6_fail:
1008         raw6_proc_exit();
1009 proc_raw6_fail:
1010 #endif
1011         ipv6_netfilter_fini();
1012 netfilter_fail:
1013         igmp6_cleanup();
1014 igmp_fail:
1015         ndisc_cleanup();
1016 ndisc_fail:
1017         ip6_mr_cleanup();
1018 icmp_fail:
1019         unregister_pernet_subsys(&inet6_net_ops);
1020 ipmr_fail:
1021         icmpv6_cleanup();
1022 register_pernet_fail:
1023         sock_unregister(PF_INET6);
1024         rtnl_unregister_all(PF_INET6);
1025 out_sock_register_fail:
1026         rawv6_exit();
1027 out_unregister_ping_proto:
1028         proto_unregister(&pingv6_prot);
1029 out_unregister_raw_proto:
1030         proto_unregister(&rawv6_prot);
1031 out_unregister_udplite_proto:
1032         proto_unregister(&udplitev6_prot);
1033 out_unregister_udp_proto:
1034         proto_unregister(&udpv6_prot);
1035 out_unregister_tcp_proto:
1036         proto_unregister(&tcpv6_prot);
1037         goto out;
1038 }
1039 module_init(inet6_init);
1040
1041 MODULE_ALIAS_NETPROTO(PF_INET6);