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         /* Look for the requested type/protocol pair. */
 113 lookup_protocol:
 114         err = -ESOCKTNOSUPPORT;
 115         rcu_read_lock();
 116         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 117
 118                 err = 0;
 119                 /* Check the non-wild match. */
 120                 if (protocol == answer->protocol) {
 121                         if (protocol != IPPROTO_IP)
 122                                 break;
 123                 } else {
 124                         /* Check for the two wild cases. */
 125                         if (IPPROTO_IP == protocol) {
 126                                 protocol = answer->protocol;
 127                                 break;
 128                         }
 129                         if (IPPROTO_IP == answer->protocol)
 130                                 break;
 131                 }
 132                 err = -EPROTONOSUPPORT;
 133         }
 134
 135         if (err) {
 136                 if (try_loading_module < 2) {
 137                         rcu_read_unlock();
 138                         /*
 139                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 140                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 141                          */
 142                         if (++try_loading_module == 1)
 143                                 request_module("net-pf-%d-proto-%d-type-%d",
 144                                                 PF_INET6, protocol, sock->type);
 145                         /*
 146                          * Fall back to generic, e.g. net-pf-10-proto-132
 147                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 148                          */
 149                         else
 150                                 request_module("net-pf-%d-proto-%d",
 151                                                 PF_INET6, protocol);
 152                         goto lookup_protocol;
 153                 } else
 154                         goto out_rcu_unlock;
 155         }
 156
 157         err = -EPERM;
 158         if (sock->type == SOCK_RAW && !kern &&
 159             !ns_capable(net->user_ns, CAP_NET_RAW))
 160                 goto out_rcu_unlock;
 161
 162         sock->ops = answer->ops;
 163         answer_prot = answer->prot;
 164         answer_flags = answer->flags;
 165         rcu_read_unlock();
 166
 167         WARN_ON(!answer_prot->slab);
 168
 169         err = -ENOBUFS;
 170         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
 171         if (!sk)
 172                 goto out;
 173
 174         sock_init_data(sock, sk);
 175
 176         err = 0;
 177         if (INET_PROTOSW_REUSE & answer_flags)
 178                 sk->sk_reuse = SK_CAN_REUSE;
 179
 180         inet = inet_sk(sk);
 181         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 182
 183         if (SOCK_RAW == sock->type) {
 184                 inet->inet_num = protocol;
 185                 if (IPPROTO_RAW == protocol)
 186                         inet->hdrincl = 1;
 187         }
 188
 189         sk->sk_destruct         = inet_sock_destruct;
 190         sk->sk_family           = PF_INET6;
 191         sk->sk_protocol         = protocol;
 192
 193         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 194
 195         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 196         np->hop_limit   = -1;
 197         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 198         np->mc_loop     = 1;
 199         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 200         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
 201
 202         /* Init the ipv4 part of the socket since we can have sockets
 203          * using v6 API for ipv4.
 204          */
 205         inet->uc_ttl    = -1;
 206
 207         inet->mc_loop   = 1;
 208         inet->mc_ttl    = 1;
 209         inet->mc_index  = 0;
 210         inet->mc_list   = NULL;
 211         inet->rcv_tos   = 0;
 212
 213         if (net->ipv4.sysctl_ip_no_pmtu_disc)
 214                 inet->pmtudisc = IP_PMTUDISC_DONT;
 215         else
 216                 inet->pmtudisc = IP_PMTUDISC_WANT;
 217         /*
 218          * Increment only the relevant sk_prot->socks debug field, this changes
 219          * the previous behaviour of incrementing both the equivalent to
 220          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 221          *
 222          * This allows better debug granularity as we'll know exactly how many
 223          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 224          * transport protocol socks. -acme
 225          */
 226         sk_refcnt_debug_inc(sk);
 227
 228         if (inet->inet_num) {
 229                 /* It assumes that any protocol which allows
 230                  * the user to assign a number at socket
 231                  * creation time automatically shares.
 232                  */
 233                 inet->inet_sport = htons(inet->inet_num);
 234                 sk->sk_prot->hash(sk);
 235         }
 236         if (sk->sk_prot->init) {
 237                 err = sk->sk_prot->init(sk);
 238                 if (err) {
 239                         sk_common_release(sk);
 240                         goto out;
 241                 }
 242         }
 243 out:
 244         return err;
 245 out_rcu_unlock:
 246         rcu_read_unlock();
 247         goto out;
 248 }
 249
 250
 251 /* bind for INET6 API */
 252 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 253 {
 254         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 255         struct sock *sk = sock->sk;
 256         struct inet_sock *inet = inet_sk(sk);
 257         struct ipv6_pinfo *np = inet6_sk(sk);
 258         struct net *net = sock_net(sk);
 259         __be32 v4addr = 0;
 260         unsigned short snum;
 261         int addr_type = 0;
 262         int err = 0;
 263
 264         /* If the socket has its own bind function then use it. */
 265         if (sk->sk_prot->bind)
 266                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 267
 268         if (addr_len < SIN6_LEN_RFC2133)
 269                 return -EINVAL;
 270
 271         if (addr->sin6_family != AF_INET6)
 272                 return -EAFNOSUPPORT;
 273
 274         addr_type = ipv6_addr_type(&addr->sin6_addr);
 275         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 276                 return -EINVAL;
 277
 278         snum = ntohs(addr->sin6_port);
 279         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 280                 return -EACCES;
 281
 282         lock_sock(sk);
 283
 284         /* Check these errors (active socket, double bind). */
 285         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 286                 err = -EINVAL;
 287                 goto out;
 288         }
 289
 290         /* Check if the address belongs to the host. */
 291         if (addr_type == IPV6_ADDR_MAPPED) {
 292                 int chk_addr_ret;
 293
 294                 /* Binding to v4-mapped address on a v6-only socket
 295                  * makes no sense
 296                  */
 297                 if (sk->sk_ipv6only) {
 298                         err = -EINVAL;
 299                         goto out;
 300                 }
 301
 302                 /* Reproduce AF_INET checks to make the bindings consistent */
 303                 v4addr = addr->sin6_addr.s6_addr32[3];
 304                 chk_addr_ret = inet_addr_type(net, v4addr);
 305                 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
 306                     !(inet->freebind || inet->transparent) &&
 307                     v4addr != htonl(INADDR_ANY) &&
 308                     chk_addr_ret != RTN_LOCAL &&
 309                     chk_addr_ret != RTN_MULTICAST &&
 310                     chk_addr_ret != RTN_BROADCAST) {
 311                         err = -EADDRNOTAVAIL;
 312                         goto out;
 313                 }
 314         } else {
 315                 if (addr_type != IPV6_ADDR_ANY) {
 316                         struct net_device *dev = NULL;
 317
 318                         rcu_read_lock();
 319                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 320                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 321                                     addr->sin6_scope_id) {
 322                                         /* Override any existing binding, if another one
 323                                          * is supplied by user.
 324                                          */
 325                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 326                                 }
 327
 328                                 /* Binding to link-local address requires an interface */
 329                                 if (!sk->sk_bound_dev_if) {
 330                                         err = -EINVAL;
 331                                         goto out_unlock;
 332                                 }
 333                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 334                                 if (!dev) {
 335                                         err = -ENODEV;
 336                                         goto out_unlock;
 337                                 }
 338                         }
 339
 340                         /* ipv4 addr of the socket is invalid.  Only the
 341                          * unspecified and mapped address have a v4 equivalent.
 342                          */
 343                         v4addr = LOOPBACK4_IPV6;
 344                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 345                                 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
 346                                     !(inet->freebind || inet->transparent) &&
 347                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 348                                                    dev, 0)) {
 349                                         err = -EADDRNOTAVAIL;
 350                                         goto out_unlock;
 351                                 }
 352                         }
 353                         rcu_read_unlock();
 354                 }
 355         }
 356
 357         inet->inet_rcv_saddr = v4addr;
 358         inet->inet_saddr = v4addr;
 359
 360         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 361
 362         if (!(addr_type & IPV6_ADDR_MULTICAST))
 363                 np->saddr = addr->sin6_addr;
 364
 365         /* Make sure we are allowed to bind here. */
 366         if ((snum || !inet->bind_address_no_port) &&
 367             sk->sk_prot->get_port(sk, snum)) {
 368                 inet_reset_saddr(sk);
 369                 err = -EADDRINUSE;
 370                 goto out;
 371         }
 372
 373         if (addr_type != IPV6_ADDR_ANY) {
 374                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 375                 if (addr_type != IPV6_ADDR_MAPPED)
 376                         sk->sk_ipv6only = 1;
 377         }
 378         if (snum)
 379                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 380         inet->inet_sport = htons(inet->inet_num);
 381         inet->inet_dport = 0;
 382         inet->inet_daddr = 0;
 383 out:
 384         release_sock(sk);
 385         return err;
 386 out_unlock:
 387         rcu_read_unlock();
 388         goto out;
 389 }
 390 EXPORT_SYMBOL(inet6_bind);
 391
 392 int inet6_release(struct socket *sock)
 393 {
 394         struct sock *sk = sock->sk;
 395
 396         if (!sk)
 397                 return -EINVAL;
 398
 399         /* Free mc lists */
 400         ipv6_sock_mc_close(sk);
 401
 402         /* Free ac lists */
 403         ipv6_sock_ac_close(sk);
 404
 405         return inet_release(sock);
 406 }
 407 EXPORT_SYMBOL(inet6_release);
 408
 409 void inet6_destroy_sock(struct sock *sk)
 410 {
 411         struct ipv6_pinfo *np = inet6_sk(sk);
 412         struct sk_buff *skb;
 413         struct ipv6_txoptions *opt;
 414
 415         /* Release rx options */
 416
 417         skb = xchg(&np->pktoptions, NULL);
 418         if (skb)
 419                 kfree_skb(skb);
 420
 421         skb = xchg(&np->rxpmtu, NULL);
 422         if (skb)
 423                 kfree_skb(skb);
 424
 425         /* Free flowlabels */
 426         fl6_free_socklist(sk);
 427
 428         /* Free tx options */
 429
 430         opt = xchg(&np->opt, NULL);
 431         if (opt)
 432                 sock_kfree_s(sk, opt, opt->tot_len);
 433 }
 434 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 435
 436 /*
 437  *      This does both peername and sockname.
 438  */
 439
 440 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 441                  int *uaddr_len, int peer)
 442 {
 443         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 444         struct sock *sk = sock->sk;
 445         struct inet_sock *inet = inet_sk(sk);
 446         struct ipv6_pinfo *np = inet6_sk(sk);
 447
 448         sin->sin6_family = AF_INET6;
 449         sin->sin6_flowinfo = 0;
 450         sin->sin6_scope_id = 0;
 451         if (peer) {
 452                 if (!inet->inet_dport)
 453                         return -ENOTCONN;
 454                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 455                     peer == 1)
 456                         return -ENOTCONN;
 457                 sin->sin6_port = inet->inet_dport;
 458                 sin->sin6_addr = sk->sk_v6_daddr;
 459                 if (np->sndflow)
 460                         sin->sin6_flowinfo = np->flow_label;
 461         } else {
 462                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 463                         sin->sin6_addr = np->saddr;
 464                 else
 465                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 466
 467                 sin->sin6_port = inet->inet_sport;
 468         }
 469         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 470                                                  sk->sk_bound_dev_if);
 471         *uaddr_len = sizeof(*sin);
 472         return 0;
 473 }
 474 EXPORT_SYMBOL(inet6_getname);
 475
 476 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 477 {
 478         struct sock *sk = sock->sk;
 479         struct net *net = sock_net(sk);
 480
 481         switch (cmd) {
 482         case SIOCGSTAMP:
 483                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 484
 485         case SIOCGSTAMPNS:
 486                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 487
 488         case SIOCADDRT:
 489         case SIOCDELRT:
 490
 491                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 492
 493         case SIOCSIFADDR:
 494                 return addrconf_add_ifaddr(net, (void __user *) arg);
 495         case SIOCDIFADDR:
 496                 return addrconf_del_ifaddr(net, (void __user *) arg);
 497         case SIOCSIFDSTADDR:
 498                 return addrconf_set_dstaddr(net, (void __user *) arg);
 499         default:
 500                 if (!sk->sk_prot->ioctl)
 501                         return -ENOIOCTLCMD;
 502                 return sk->sk_prot->ioctl(sk, cmd, arg);
 503         }
 504         /*NOTREACHED*/
 505         return 0;
 506 }
 507 EXPORT_SYMBOL(inet6_ioctl);
 508
 509 const struct proto_ops inet6_stream_ops = {
 510         .family            = PF_INET6,
 511         .owner             = THIS_MODULE,
 512         .release           = inet6_release,
 513         .bind              = inet6_bind,
 514         .connect           = inet_stream_connect,       /* ok           */
 515         .socketpair        = sock_no_socketpair,        /* a do nothing */
 516         .accept            = inet_accept,               /* ok           */
 517         .getname           = inet6_getname,
 518         .poll              = tcp_poll,                  /* ok           */
 519         .ioctl             = inet6_ioctl,               /* must change  */
 520         .listen            = inet_listen,               /* ok           */
 521         .shutdown          = inet_shutdown,             /* ok           */
 522         .setsockopt        = sock_common_setsockopt,    /* ok           */
 523         .getsockopt        = sock_common_getsockopt,    /* ok           */
 524         .sendmsg           = inet_sendmsg,              /* ok           */
 525         .recvmsg           = inet_recvmsg,              /* ok           */
 526         .mmap              = sock_no_mmap,
 527         .sendpage          = inet_sendpage,
 528         .splice_read       = tcp_splice_read,
 529 #ifdef CONFIG_COMPAT
 530         .compat_setsockopt = compat_sock_common_setsockopt,
 531         .compat_getsockopt = compat_sock_common_getsockopt,
 532 #endif
 533 };
 534
 535 const struct proto_ops inet6_dgram_ops = {
 536         .family            = PF_INET6,
 537         .owner             = THIS_MODULE,
 538         .release           = inet6_release,
 539         .bind              = inet6_bind,
 540         .connect           = inet_dgram_connect,        /* ok           */
 541         .socketpair        = sock_no_socketpair,        /* a do nothing */
 542         .accept            = sock_no_accept,            /* a do nothing */
 543         .getname           = inet6_getname,
 544         .poll              = udp_poll,                  /* ok           */
 545         .ioctl             = inet6_ioctl,               /* must change  */
 546         .listen            = sock_no_listen,            /* ok           */
 547         .shutdown          = inet_shutdown,             /* ok           */
 548         .setsockopt        = sock_common_setsockopt,    /* ok           */
 549         .getsockopt        = sock_common_getsockopt,    /* ok           */
 550         .sendmsg           = inet_sendmsg,              /* ok           */
 551         .recvmsg           = inet_recvmsg,              /* ok           */
 552         .mmap              = sock_no_mmap,
 553         .sendpage          = sock_no_sendpage,
 554 #ifdef CONFIG_COMPAT
 555         .compat_setsockopt = compat_sock_common_setsockopt,
 556         .compat_getsockopt = compat_sock_common_getsockopt,
 557 #endif
 558 };
 559
 560 static const struct net_proto_family inet6_family_ops = {
 561         .family = PF_INET6,
 562         .create = inet6_create,
 563         .owner  = THIS_MODULE,
 564 };
 565
 566 int inet6_register_protosw(struct inet_protosw *p)
 567 {
 568         struct list_head *lh;
 569         struct inet_protosw *answer;
 570         struct list_head *last_perm;
 571         int protocol = p->protocol;
 572         int ret;
 573
 574         spin_lock_bh(&inetsw6_lock);
 575
 576         ret = -EINVAL;
 577         if (p->type >= SOCK_MAX)
 578                 goto out_illegal;
 579
 580         /* If we are trying to override a permanent protocol, bail. */
 581         answer = NULL;
 582         ret = -EPERM;
 583         last_perm = &inetsw6[p->type];
 584         list_for_each(lh, &inetsw6[p->type]) {
 585                 answer = list_entry(lh, struct inet_protosw, list);
 586
 587                 /* Check only the non-wild match. */
 588                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 589                         if (protocol == answer->protocol)
 590                                 break;
 591                         last_perm = lh;
 592                 }
 593
 594                 answer = NULL;
 595         }
 596         if (answer)
 597                 goto out_permanent;
 598
 599         /* Add the new entry after the last permanent entry if any, so that
 600          * the new entry does not override a permanent entry when matched with
 601          * a wild-card protocol. But it is allowed to override any existing
 602          * non-permanent entry.  This means that when we remove this entry, the
 603          * system automatically returns to the old behavior.
 604          */
 605         list_add_rcu(&p->list, last_perm);
 606         ret = 0;
 607 out:
 608         spin_unlock_bh(&inetsw6_lock);
 609         return ret;
 610
 611 out_permanent:
 612         pr_err("Attempt to override permanent protocol %d\n", protocol);
 613         goto out;
 614
 615 out_illegal:
 616         pr_err("Ignoring attempt to register invalid socket type %d\n",
 617                p->type);
 618         goto out;
 619 }
 620 EXPORT_SYMBOL(inet6_register_protosw);
 621
 622 void
 623 inet6_unregister_protosw(struct inet_protosw *p)
 624 {
 625         if (INET_PROTOSW_PERMANENT & p->flags) {
 626                 pr_err("Attempt to unregister permanent protocol %d\n",
 627                        p->protocol);
 628         } else {
 629                 spin_lock_bh(&inetsw6_lock);
 630                 list_del_rcu(&p->list);
 631                 spin_unlock_bh(&inetsw6_lock);
 632
 633                 synchronize_net();
 634         }
 635 }
 636 EXPORT_SYMBOL(inet6_unregister_protosw);
 637
 638 int inet6_sk_rebuild_header(struct sock *sk)
 639 {
 640         struct ipv6_pinfo *np = inet6_sk(sk);
 641         struct dst_entry *dst;
 642
 643         dst = __sk_dst_check(sk, np->dst_cookie);
 644
 645         if (!dst) {
 646                 struct inet_sock *inet = inet_sk(sk);
 647                 struct in6_addr *final_p, final;
 648                 struct flowi6 fl6;
 649
 650                 memset(&fl6, 0, sizeof(fl6));
 651                 fl6.flowi6_proto = sk->sk_protocol;
 652                 fl6.daddr = sk->sk_v6_daddr;
 653                 fl6.saddr = np->saddr;
 654                 fl6.flowlabel = np->flow_label;
 655                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 656                 fl6.flowi6_mark = sk->sk_mark;
 657                 fl6.fl6_dport = inet->inet_dport;
 658                 fl6.fl6_sport = inet->inet_sport;
 659                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 660
 661                 final_p = fl6_update_dst(&fl6, np->opt, &final);
 662
 663                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 664                 if (IS_ERR(dst)) {
 665                         sk->sk_route_caps = 0;
 666                         sk->sk_err_soft = -PTR_ERR(dst);
 667                         return PTR_ERR(dst);
 668                 }
 669
 670                 __ip6_dst_store(sk, dst, NULL, NULL);
 671         }
 672
 673         return 0;
 674 }
 675 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 676
 677 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 678                        const struct inet6_skb_parm *opt)
 679 {
 680         const struct ipv6_pinfo *np = inet6_sk(sk);
 681
 682         if (np->rxopt.all) {
 683                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
 684                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 685                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 686                      np->rxopt.bits.rxflow) ||
 687                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 688                      np->rxopt.bits.osrcrt)) ||
 689                     ((opt->dst1 || opt->dst0) &&
 690                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 691                         return true;
 692         }
 693         return false;
 694 }
 695 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 696
 697 static struct packet_type ipv6_packet_type __read_mostly = {
 698         .type = cpu_to_be16(ETH_P_IPV6),
 699         .func = ipv6_rcv,
 700 };
 701
 702 static int __init ipv6_packet_init(void)
 703 {
 704         dev_add_pack(&ipv6_packet_type);
 705         return 0;
 706 }
 707
 708 static void ipv6_packet_cleanup(void)
 709 {
 710         dev_remove_pack(&ipv6_packet_type);
 711 }
 712
 713 static int __net_init ipv6_init_mibs(struct net *net)
 714 {
 715         int i;
 716
 717         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 718         if (!net->mib.udp_stats_in6)
 719                 return -ENOMEM;
 720         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 721         if (!net->mib.udplite_stats_in6)
 722                 goto err_udplite_mib;
 723         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 724         if (!net->mib.ipv6_statistics)
 725                 goto err_ip_mib;
 726
 727         for_each_possible_cpu(i) {
 728                 struct ipstats_mib *af_inet6_stats;
 729                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 730                 u64_stats_init(&af_inet6_stats->syncp);
 731         }
 732
 733
 734         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 735         if (!net->mib.icmpv6_statistics)
 736                 goto err_icmp_mib;
 737         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 738                                                 GFP_KERNEL);
 739         if (!net->mib.icmpv6msg_statistics)
 740                 goto err_icmpmsg_mib;
 741         return 0;
 742
 743 err_icmpmsg_mib:
 744         free_percpu(net->mib.icmpv6_statistics);
 745 err_icmp_mib:
 746         free_percpu(net->mib.ipv6_statistics);
 747 err_ip_mib:
 748         free_percpu(net->mib.udplite_stats_in6);
 749 err_udplite_mib:
 750         free_percpu(net->mib.udp_stats_in6);
 751         return -ENOMEM;
 752 }
 753
 754 static void ipv6_cleanup_mibs(struct net *net)
 755 {
 756         free_percpu(net->mib.udp_stats_in6);
 757         free_percpu(net->mib.udplite_stats_in6);
 758         free_percpu(net->mib.ipv6_statistics);
 759         free_percpu(net->mib.icmpv6_statistics);
 760         kfree(net->mib.icmpv6msg_statistics);
 761 }
 762
 763 static int __net_init inet6_net_init(struct net *net)
 764 {
 765         int err = 0;
 766
 767         net->ipv6.sysctl.bindv6only = 0;
 768         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 769         net->ipv6.sysctl.flowlabel_consistency = 1;
 770         net->ipv6.sysctl.auto_flowlabels = 0;
 771         net->ipv6.sysctl.idgen_retries = 3;
 772         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 773         net->ipv6.sysctl.flowlabel_state_ranges = 1;
 774         atomic_set(&net->ipv6.fib6_sernum, 1);
 775
 776         err = ipv6_init_mibs(net);
 777         if (err)
 778                 return err;
 779 #ifdef CONFIG_PROC_FS
 780         err = udp6_proc_init(net);
 781         if (err)
 782                 goto out;
 783         err = tcp6_proc_init(net);
 784         if (err)
 785                 goto proc_tcp6_fail;
 786         err = ac6_proc_init(net);
 787         if (err)
 788                 goto proc_ac6_fail;
 789 #endif
 790         return err;
 791
 792 #ifdef CONFIG_PROC_FS
 793 proc_ac6_fail:
 794         tcp6_proc_exit(net);
 795 proc_tcp6_fail:
 796         udp6_proc_exit(net);
 797 out:
 798         ipv6_cleanup_mibs(net);
 799         return err;
 800 #endif
 801 }
 802
 803 static void __net_exit inet6_net_exit(struct net *net)
 804 {
 805 #ifdef CONFIG_PROC_FS
 806         udp6_proc_exit(net);
 807         tcp6_proc_exit(net);
 808         ac6_proc_exit(net);
 809 #endif
 810         ipv6_cleanup_mibs(net);
 811 }
 812
 813 static struct pernet_operations inet6_net_ops = {
 814         .init = inet6_net_init,
 815         .exit = inet6_net_exit,
 816 };
 817
 818 static const struct ipv6_stub ipv6_stub_impl = {
 819         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 820         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 821         .ipv6_dst_lookup = ip6_dst_lookup,
 822         .udpv6_encap_enable = udpv6_encap_enable,
 823         .ndisc_send_na = ndisc_send_na,
 824         .nd_tbl = &nd_tbl,
 825 };
 826
 827 static int __init inet6_init(void)
 828 {
 829         struct list_head *r;
 830         int err = 0;
 831
 832         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 833
 834         /* Register the socket-side information for inet6_create.  */
 835         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 836                 INIT_LIST_HEAD(r);
 837
 838         if (disable_ipv6_mod) {
 839                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 840                 goto out;
 841         }
 842
 843         err = proto_register(&tcpv6_prot, 1);
 844         if (err)
 845                 goto out;
 846
 847         err = proto_register(&udpv6_prot, 1);
 848         if (err)
 849                 goto out_unregister_tcp_proto;
 850
 851         err = proto_register(&udplitev6_prot, 1);
 852         if (err)
 853                 goto out_unregister_udp_proto;
 854
 855         err = proto_register(&rawv6_prot, 1);
 856         if (err)
 857                 goto out_unregister_udplite_proto;
 858
 859         err = proto_register(&pingv6_prot, 1);
 860         if (err)
 861                 goto out_unregister_ping_proto;
 862
 863         /* We MUST register RAW sockets before we create the ICMP6,
 864          * IGMP6, or NDISC control sockets.
 865          */
 866         err = rawv6_init();
 867         if (err)
 868                 goto out_unregister_raw_proto;
 869
 870         /* Register the family here so that the init calls below will
 871          * be able to create sockets. (?? is this dangerous ??)
 872          */
 873         err = sock_register(&inet6_family_ops);
 874         if (err)
 875                 goto out_sock_register_fail;
 876
 877         /*
 878          *      ipngwg API draft makes clear that the correct semantics
 879          *      for TCP and UDP is to consider one TCP and UDP instance
 880          *      in a host available by both INET and INET6 APIs and
 881          *      able to communicate via both network protocols.
 882          */
 883
 884         err = register_pernet_subsys(&inet6_net_ops);
 885         if (err)
 886                 goto register_pernet_fail;
 887         err = icmpv6_init();
 888         if (err)
 889                 goto icmp_fail;
 890         err = ip6_mr_init();
 891         if (err)
 892                 goto ipmr_fail;
 893         err = ndisc_init();
 894         if (err)
 895                 goto ndisc_fail;
 896         err = igmp6_init();
 897         if (err)
 898                 goto igmp_fail;
 899
 900         ipv6_stub = &ipv6_stub_impl;
 901
 902         err = ipv6_netfilter_init();
 903         if (err)
 904                 goto netfilter_fail;
 905         /* Create /proc/foo6 entries. */
 906 #ifdef CONFIG_PROC_FS
 907         err = -ENOMEM;
 908         if (raw6_proc_init())
 909                 goto proc_raw6_fail;
 910         if (udplite6_proc_init())
 911                 goto proc_udplite6_fail;
 912         if (ipv6_misc_proc_init())
 913                 goto proc_misc6_fail;
 914         if (if6_proc_init())
 915                 goto proc_if6_fail;
 916 #endif
 917         err = ip6_route_init();
 918         if (err)
 919                 goto ip6_route_fail;
 920         err = ndisc_late_init();
 921         if (err)
 922                 goto ndisc_late_fail;
 923         err = ip6_flowlabel_init();
 924         if (err)
 925                 goto ip6_flowlabel_fail;
 926         err = addrconf_init();
 927         if (err)
 928                 goto addrconf_fail;
 929
 930         /* Init v6 extension headers. */
 931         err = ipv6_exthdrs_init();
 932         if (err)
 933                 goto ipv6_exthdrs_fail;
 934
 935         err = ipv6_frag_init();
 936         if (err)
 937                 goto ipv6_frag_fail;
 938
 939         /* Init v6 transport protocols. */
 940         err = udpv6_init();
 941         if (err)
 942                 goto udpv6_fail;
 943
 944         err = udplitev6_init();
 945         if (err)
 946                 goto udplitev6_fail;
 947
 948         err = tcpv6_init();
 949         if (err)
 950                 goto tcpv6_fail;
 951
 952         err = ipv6_packet_init();
 953         if (err)
 954                 goto ipv6_packet_fail;
 955
 956         err = pingv6_init();
 957         if (err)
 958                 goto pingv6_fail;
 959
 960 #ifdef CONFIG_SYSCTL
 961         err = ipv6_sysctl_register();
 962         if (err)
 963                 goto sysctl_fail;
 964 #endif
 965 out:
 966         return err;
 967
 968 #ifdef CONFIG_SYSCTL
 969 sysctl_fail:
 970         pingv6_exit();
 971 #endif
 972 pingv6_fail:
 973         ipv6_packet_cleanup();
 974 ipv6_packet_fail:
 975         tcpv6_exit();
 976 tcpv6_fail:
 977         udplitev6_exit();
 978 udplitev6_fail:
 979         udpv6_exit();
 980 udpv6_fail:
 981         ipv6_frag_exit();
 982 ipv6_frag_fail:
 983         ipv6_exthdrs_exit();
 984 ipv6_exthdrs_fail:
 985         addrconf_cleanup();
 986 addrconf_fail:
 987         ip6_flowlabel_cleanup();
 988 ip6_flowlabel_fail:
 989         ndisc_late_cleanup();
 990 ndisc_late_fail:
 991         ip6_route_cleanup();
 992 ip6_route_fail:
 993 #ifdef CONFIG_PROC_FS
 994         if6_proc_exit();
 995 proc_if6_fail:
 996         ipv6_misc_proc_exit();
 997 proc_misc6_fail:
 998         udplite6_proc_exit();
 999 proc_udplite6_fail:
1000         raw6_proc_exit();
1001 proc_raw6_fail:
1002 #endif
1003         ipv6_netfilter_fini();
1004 netfilter_fail:
1005         igmp6_cleanup();
1006 igmp_fail:
1007         ndisc_cleanup();
1008 ndisc_fail:
1009         ip6_mr_cleanup();
1010 ipmr_fail:
1011         icmpv6_cleanup();
1012 icmp_fail:
1013         unregister_pernet_subsys(&inet6_net_ops);
1014 register_pernet_fail:
1015         sock_unregister(PF_INET6);
1016         rtnl_unregister_all(PF_INET6);
1017 out_sock_register_fail:
1018         rawv6_exit();
1019 out_unregister_ping_proto:
1020         proto_unregister(&pingv6_prot);
1021 out_unregister_raw_proto:
1022         proto_unregister(&rawv6_prot);
1023 out_unregister_udplite_proto:
1024         proto_unregister(&udplitev6_prot);
1025 out_unregister_udp_proto:
1026         proto_unregister(&udpv6_prot);
1027 out_unregister_tcp_proto:
1028         proto_unregister(&tcpv6_prot);
1029         goto out;
1030 }
1031 module_init(inet6_init);
1032
1033 MODULE_ALIAS_NETPROTO(PF_INET6);