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 (!(inet->freebind || inet->transparent) &&
 346                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 347                                                    dev, 0)) {
 348                                         err = -EADDRNOTAVAIL;
 349                                         goto out_unlock;
 350                                 }
 351                         }
 352                         rcu_read_unlock();
 353                 }
 354         }
 355
 356         inet->inet_rcv_saddr = v4addr;
 357         inet->inet_saddr = v4addr;
 358
 359         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 360
 361         if (!(addr_type & IPV6_ADDR_MULTICAST))
 362                 np->saddr = addr->sin6_addr;
 363
 364         /* Make sure we are allowed to bind here. */
 365         if ((snum || !inet->bind_address_no_port) &&
 366             sk->sk_prot->get_port(sk, snum)) {
 367                 inet_reset_saddr(sk);
 368                 err = -EADDRINUSE;
 369                 goto out;
 370         }
 371
 372         if (addr_type != IPV6_ADDR_ANY) {
 373                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 374                 if (addr_type != IPV6_ADDR_MAPPED)
 375                         sk->sk_ipv6only = 1;
 376         }
 377         if (snum)
 378                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 379         inet->inet_sport = htons(inet->inet_num);
 380         inet->inet_dport = 0;
 381         inet->inet_daddr = 0;
 382 out:
 383         release_sock(sk);
 384         return err;
 385 out_unlock:
 386         rcu_read_unlock();
 387         goto out;
 388 }
 389 EXPORT_SYMBOL(inet6_bind);
 390
 391 int inet6_release(struct socket *sock)
 392 {
 393         struct sock *sk = sock->sk;
 394
 395         if (!sk)
 396                 return -EINVAL;
 397
 398         /* Free mc lists */
 399         ipv6_sock_mc_close(sk);
 400
 401         /* Free ac lists */
 402         ipv6_sock_ac_close(sk);
 403
 404         return inet_release(sock);
 405 }
 406 EXPORT_SYMBOL(inet6_release);
 407
 408 void inet6_destroy_sock(struct sock *sk)
 409 {
 410         struct ipv6_pinfo *np = inet6_sk(sk);
 411         struct sk_buff *skb;
 412         struct ipv6_txoptions *opt;
 413
 414         /* Release rx options */
 415
 416         skb = xchg(&np->pktoptions, NULL);
 417         if (skb)
 418                 kfree_skb(skb);
 419
 420         skb = xchg(&np->rxpmtu, NULL);
 421         if (skb)
 422                 kfree_skb(skb);
 423
 424         /* Free flowlabels */
 425         fl6_free_socklist(sk);
 426
 427         /* Free tx options */
 428
 429         opt = xchg(&np->opt, NULL);
 430         if (opt)
 431                 sock_kfree_s(sk, opt, opt->tot_len);
 432 }
 433 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 434
 435 /*
 436  *      This does both peername and sockname.
 437  */
 438
 439 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 440                  int *uaddr_len, int peer)
 441 {
 442         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 443         struct sock *sk = sock->sk;
 444         struct inet_sock *inet = inet_sk(sk);
 445         struct ipv6_pinfo *np = inet6_sk(sk);
 446
 447         sin->sin6_family = AF_INET6;
 448         sin->sin6_flowinfo = 0;
 449         sin->sin6_scope_id = 0;
 450         if (peer) {
 451                 if (!inet->inet_dport)
 452                         return -ENOTCONN;
 453                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 454                     peer == 1)
 455                         return -ENOTCONN;
 456                 sin->sin6_port = inet->inet_dport;
 457                 sin->sin6_addr = sk->sk_v6_daddr;
 458                 if (np->sndflow)
 459                         sin->sin6_flowinfo = np->flow_label;
 460         } else {
 461                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 462                         sin->sin6_addr = np->saddr;
 463                 else
 464                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 465
 466                 sin->sin6_port = inet->inet_sport;
 467         }
 468         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 469                                                  sk->sk_bound_dev_if);
 470         *uaddr_len = sizeof(*sin);
 471         return 0;
 472 }
 473 EXPORT_SYMBOL(inet6_getname);
 474
 475 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 476 {
 477         struct sock *sk = sock->sk;
 478         struct net *net = sock_net(sk);
 479
 480         switch (cmd) {
 481         case SIOCGSTAMP:
 482                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 483
 484         case SIOCGSTAMPNS:
 485                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 486
 487         case SIOCADDRT:
 488         case SIOCDELRT:
 489
 490                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 491
 492         case SIOCSIFADDR:
 493                 return addrconf_add_ifaddr(net, (void __user *) arg);
 494         case SIOCDIFADDR:
 495                 return addrconf_del_ifaddr(net, (void __user *) arg);
 496         case SIOCSIFDSTADDR:
 497                 return addrconf_set_dstaddr(net, (void __user *) arg);
 498         default:
 499                 if (!sk->sk_prot->ioctl)
 500                         return -ENOIOCTLCMD;
 501                 return sk->sk_prot->ioctl(sk, cmd, arg);
 502         }
 503         /*NOTREACHED*/
 504         return 0;
 505 }
 506 EXPORT_SYMBOL(inet6_ioctl);
 507
 508 const struct proto_ops inet6_stream_ops = {
 509         .family            = PF_INET6,
 510         .owner             = THIS_MODULE,
 511         .release           = inet6_release,
 512         .bind              = inet6_bind,
 513         .connect           = inet_stream_connect,       /* ok           */
 514         .socketpair        = sock_no_socketpair,        /* a do nothing */
 515         .accept            = inet_accept,               /* ok           */
 516         .getname           = inet6_getname,
 517         .poll              = tcp_poll,                  /* ok           */
 518         .ioctl             = inet6_ioctl,               /* must change  */
 519         .listen            = inet_listen,               /* ok           */
 520         .shutdown          = inet_shutdown,             /* ok           */
 521         .setsockopt        = sock_common_setsockopt,    /* ok           */
 522         .getsockopt        = sock_common_getsockopt,    /* ok           */
 523         .sendmsg           = inet_sendmsg,              /* ok           */
 524         .recvmsg           = inet_recvmsg,              /* ok           */
 525         .mmap              = sock_no_mmap,
 526         .sendpage          = inet_sendpage,
 527         .splice_read       = tcp_splice_read,
 528 #ifdef CONFIG_COMPAT
 529         .compat_setsockopt = compat_sock_common_setsockopt,
 530         .compat_getsockopt = compat_sock_common_getsockopt,
 531 #endif
 532 };
 533
 534 const struct proto_ops inet6_dgram_ops = {
 535         .family            = PF_INET6,
 536         .owner             = THIS_MODULE,
 537         .release           = inet6_release,
 538         .bind              = inet6_bind,
 539         .connect           = inet_dgram_connect,        /* ok           */
 540         .socketpair        = sock_no_socketpair,        /* a do nothing */
 541         .accept            = sock_no_accept,            /* a do nothing */
 542         .getname           = inet6_getname,
 543         .poll              = udp_poll,                  /* ok           */
 544         .ioctl             = inet6_ioctl,               /* must change  */
 545         .listen            = sock_no_listen,            /* ok           */
 546         .shutdown          = inet_shutdown,             /* ok           */
 547         .setsockopt        = sock_common_setsockopt,    /* ok           */
 548         .getsockopt        = sock_common_getsockopt,    /* ok           */
 549         .sendmsg           = inet_sendmsg,              /* ok           */
 550         .recvmsg           = inet_recvmsg,              /* ok           */
 551         .mmap              = sock_no_mmap,
 552         .sendpage          = sock_no_sendpage,
 553 #ifdef CONFIG_COMPAT
 554         .compat_setsockopt = compat_sock_common_setsockopt,
 555         .compat_getsockopt = compat_sock_common_getsockopt,
 556 #endif
 557 };
 558
 559 static const struct net_proto_family inet6_family_ops = {
 560         .family = PF_INET6,
 561         .create = inet6_create,
 562         .owner  = THIS_MODULE,
 563 };
 564
 565 int inet6_register_protosw(struct inet_protosw *p)
 566 {
 567         struct list_head *lh;
 568         struct inet_protosw *answer;
 569         struct list_head *last_perm;
 570         int protocol = p->protocol;
 571         int ret;
 572
 573         spin_lock_bh(&inetsw6_lock);
 574
 575         ret = -EINVAL;
 576         if (p->type >= SOCK_MAX)
 577                 goto out_illegal;
 578
 579         /* If we are trying to override a permanent protocol, bail. */
 580         answer = NULL;
 581         ret = -EPERM;
 582         last_perm = &inetsw6[p->type];
 583         list_for_each(lh, &inetsw6[p->type]) {
 584                 answer = list_entry(lh, struct inet_protosw, list);
 585
 586                 /* Check only the non-wild match. */
 587                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 588                         if (protocol == answer->protocol)
 589                                 break;
 590                         last_perm = lh;
 591                 }
 592
 593                 answer = NULL;
 594         }
 595         if (answer)
 596                 goto out_permanent;
 597
 598         /* Add the new entry after the last permanent entry if any, so that
 599          * the new entry does not override a permanent entry when matched with
 600          * a wild-card protocol. But it is allowed to override any existing
 601          * non-permanent entry.  This means that when we remove this entry, the
 602          * system automatically returns to the old behavior.
 603          */
 604         list_add_rcu(&p->list, last_perm);
 605         ret = 0;
 606 out:
 607         spin_unlock_bh(&inetsw6_lock);
 608         return ret;
 609
 610 out_permanent:
 611         pr_err("Attempt to override permanent protocol %d\n", protocol);
 612         goto out;
 613
 614 out_illegal:
 615         pr_err("Ignoring attempt to register invalid socket type %d\n",
 616                p->type);
 617         goto out;
 618 }
 619 EXPORT_SYMBOL(inet6_register_protosw);
 620
 621 void
 622 inet6_unregister_protosw(struct inet_protosw *p)
 623 {
 624         if (INET_PROTOSW_PERMANENT & p->flags) {
 625                 pr_err("Attempt to unregister permanent protocol %d\n",
 626                        p->protocol);
 627         } else {
 628                 spin_lock_bh(&inetsw6_lock);
 629                 list_del_rcu(&p->list);
 630                 spin_unlock_bh(&inetsw6_lock);
 631
 632                 synchronize_net();
 633         }
 634 }
 635 EXPORT_SYMBOL(inet6_unregister_protosw);
 636
 637 int inet6_sk_rebuild_header(struct sock *sk)
 638 {
 639         struct ipv6_pinfo *np = inet6_sk(sk);
 640         struct dst_entry *dst;
 641
 642         dst = __sk_dst_check(sk, np->dst_cookie);
 643
 644         if (!dst) {
 645                 struct inet_sock *inet = inet_sk(sk);
 646                 struct in6_addr *final_p, final;
 647                 struct flowi6 fl6;
 648
 649                 memset(&fl6, 0, sizeof(fl6));
 650                 fl6.flowi6_proto = sk->sk_protocol;
 651                 fl6.daddr = sk->sk_v6_daddr;
 652                 fl6.saddr = np->saddr;
 653                 fl6.flowlabel = np->flow_label;
 654                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 655                 fl6.flowi6_mark = sk->sk_mark;
 656                 fl6.fl6_dport = inet->inet_dport;
 657                 fl6.fl6_sport = inet->inet_sport;
 658                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 659
 660                 final_p = fl6_update_dst(&fl6, np->opt, &final);
 661
 662                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 663                 if (IS_ERR(dst)) {
 664                         sk->sk_route_caps = 0;
 665                         sk->sk_err_soft = -PTR_ERR(dst);
 666                         return PTR_ERR(dst);
 667                 }
 668
 669                 __ip6_dst_store(sk, dst, NULL, NULL);
 670         }
 671
 672         return 0;
 673 }
 674 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 675
 676 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 677                        const struct inet6_skb_parm *opt)
 678 {
 679         const struct ipv6_pinfo *np = inet6_sk(sk);
 680
 681         if (np->rxopt.all) {
 682                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 683                                   np->rxopt.bits.ohopopts)) ||
 684                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 685                      np->rxopt.bits.rxflow) ||
 686                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 687                      np->rxopt.bits.osrcrt)) ||
 688                     ((opt->dst1 || opt->dst0) &&
 689                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 690                         return true;
 691         }
 692         return false;
 693 }
 694 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 695
 696 static struct packet_type ipv6_packet_type __read_mostly = {
 697         .type = cpu_to_be16(ETH_P_IPV6),
 698         .func = ipv6_rcv,
 699 };
 700
 701 static int __init ipv6_packet_init(void)
 702 {
 703         dev_add_pack(&ipv6_packet_type);
 704         return 0;
 705 }
 706
 707 static void ipv6_packet_cleanup(void)
 708 {
 709         dev_remove_pack(&ipv6_packet_type);
 710 }
 711
 712 static int __net_init ipv6_init_mibs(struct net *net)
 713 {
 714         int i;
 715
 716         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 717         if (!net->mib.udp_stats_in6)
 718                 return -ENOMEM;
 719         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 720         if (!net->mib.udplite_stats_in6)
 721                 goto err_udplite_mib;
 722         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 723         if (!net->mib.ipv6_statistics)
 724                 goto err_ip_mib;
 725
 726         for_each_possible_cpu(i) {
 727                 struct ipstats_mib *af_inet6_stats;
 728                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 729                 u64_stats_init(&af_inet6_stats->syncp);
 730         }
 731
 732
 733         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 734         if (!net->mib.icmpv6_statistics)
 735                 goto err_icmp_mib;
 736         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 737                                                 GFP_KERNEL);
 738         if (!net->mib.icmpv6msg_statistics)
 739                 goto err_icmpmsg_mib;
 740         return 0;
 741
 742 err_icmpmsg_mib:
 743         free_percpu(net->mib.icmpv6_statistics);
 744 err_icmp_mib:
 745         free_percpu(net->mib.ipv6_statistics);
 746 err_ip_mib:
 747         free_percpu(net->mib.udplite_stats_in6);
 748 err_udplite_mib:
 749         free_percpu(net->mib.udp_stats_in6);
 750         return -ENOMEM;
 751 }
 752
 753 static void ipv6_cleanup_mibs(struct net *net)
 754 {
 755         free_percpu(net->mib.udp_stats_in6);
 756         free_percpu(net->mib.udplite_stats_in6);
 757         free_percpu(net->mib.ipv6_statistics);
 758         free_percpu(net->mib.icmpv6_statistics);
 759         kfree(net->mib.icmpv6msg_statistics);
 760 }
 761
 762 static int __net_init inet6_net_init(struct net *net)
 763 {
 764         int err = 0;
 765
 766         net->ipv6.sysctl.bindv6only = 0;
 767         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 768         net->ipv6.sysctl.flowlabel_consistency = 1;
 769         net->ipv6.sysctl.auto_flowlabels = 0;
 770         net->ipv6.sysctl.idgen_retries = 3;
 771         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 772         net->ipv6.sysctl.flowlabel_state_ranges = 1;
 773         atomic_set(&net->ipv6.fib6_sernum, 1);
 774
 775         err = ipv6_init_mibs(net);
 776         if (err)
 777                 return err;
 778 #ifdef CONFIG_PROC_FS
 779         err = udp6_proc_init(net);
 780         if (err)
 781                 goto out;
 782         err = tcp6_proc_init(net);
 783         if (err)
 784                 goto proc_tcp6_fail;
 785         err = ac6_proc_init(net);
 786         if (err)
 787                 goto proc_ac6_fail;
 788 #endif
 789         return err;
 790
 791 #ifdef CONFIG_PROC_FS
 792 proc_ac6_fail:
 793         tcp6_proc_exit(net);
 794 proc_tcp6_fail:
 795         udp6_proc_exit(net);
 796 out:
 797         ipv6_cleanup_mibs(net);
 798         return err;
 799 #endif
 800 }
 801
 802 static void __net_exit inet6_net_exit(struct net *net)
 803 {
 804 #ifdef CONFIG_PROC_FS
 805         udp6_proc_exit(net);
 806         tcp6_proc_exit(net);
 807         ac6_proc_exit(net);
 808 #endif
 809         ipv6_cleanup_mibs(net);
 810 }
 811
 812 static struct pernet_operations inet6_net_ops = {
 813         .init = inet6_net_init,
 814         .exit = inet6_net_exit,
 815 };
 816
 817 static const struct ipv6_stub ipv6_stub_impl = {
 818         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 819         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 820         .ipv6_dst_lookup = ip6_dst_lookup,
 821         .udpv6_encap_enable = udpv6_encap_enable,
 822         .ndisc_send_na = ndisc_send_na,
 823         .nd_tbl = &nd_tbl,
 824 };
 825
 826 static int __init inet6_init(void)
 827 {
 828         struct list_head *r;
 829         int err = 0;
 830
 831         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 832
 833         /* Register the socket-side information for inet6_create.  */
 834         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 835                 INIT_LIST_HEAD(r);
 836
 837         if (disable_ipv6_mod) {
 838                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 839                 goto out;
 840         }
 841
 842         err = proto_register(&tcpv6_prot, 1);
 843         if (err)
 844                 goto out;
 845
 846         err = proto_register(&udpv6_prot, 1);
 847         if (err)
 848                 goto out_unregister_tcp_proto;
 849
 850         err = proto_register(&udplitev6_prot, 1);
 851         if (err)
 852                 goto out_unregister_udp_proto;
 853
 854         err = proto_register(&rawv6_prot, 1);
 855         if (err)
 856                 goto out_unregister_udplite_proto;
 857
 858         err = proto_register(&pingv6_prot, 1);
 859         if (err)
 860                 goto out_unregister_ping_proto;
 861
 862         /* We MUST register RAW sockets before we create the ICMP6,
 863          * IGMP6, or NDISC control sockets.
 864          */
 865         err = rawv6_init();
 866         if (err)
 867                 goto out_unregister_raw_proto;
 868
 869         /* Register the family here so that the init calls below will
 870          * be able to create sockets. (?? is this dangerous ??)
 871          */
 872         err = sock_register(&inet6_family_ops);
 873         if (err)
 874                 goto out_sock_register_fail;
 875
 876         /*
 877          *      ipngwg API draft makes clear that the correct semantics
 878          *      for TCP and UDP is to consider one TCP and UDP instance
 879          *      in a host available by both INET and INET6 APIs and
 880          *      able to communicate via both network protocols.
 881          */
 882
 883         err = register_pernet_subsys(&inet6_net_ops);
 884         if (err)
 885                 goto register_pernet_fail;
 886         err = icmpv6_init();
 887         if (err)
 888                 goto icmp_fail;
 889         err = ip6_mr_init();
 890         if (err)
 891                 goto ipmr_fail;
 892         err = ndisc_init();
 893         if (err)
 894                 goto ndisc_fail;
 895         err = igmp6_init();
 896         if (err)
 897                 goto igmp_fail;
 898
 899         ipv6_stub = &ipv6_stub_impl;
 900
 901         err = ipv6_netfilter_init();
 902         if (err)
 903                 goto netfilter_fail;
 904         /* Create /proc/foo6 entries. */
 905 #ifdef CONFIG_PROC_FS
 906         err = -ENOMEM;
 907         if (raw6_proc_init())
 908                 goto proc_raw6_fail;
 909         if (udplite6_proc_init())
 910                 goto proc_udplite6_fail;
 911         if (ipv6_misc_proc_init())
 912                 goto proc_misc6_fail;
 913         if (if6_proc_init())
 914                 goto proc_if6_fail;
 915 #endif
 916         err = ip6_route_init();
 917         if (err)
 918                 goto ip6_route_fail;
 919         err = ndisc_late_init();
 920         if (err)
 921                 goto ndisc_late_fail;
 922         err = ip6_flowlabel_init();
 923         if (err)
 924                 goto ip6_flowlabel_fail;
 925         err = addrconf_init();
 926         if (err)
 927                 goto addrconf_fail;
 928
 929         /* Init v6 extension headers. */
 930         err = ipv6_exthdrs_init();
 931         if (err)
 932                 goto ipv6_exthdrs_fail;
 933
 934         err = ipv6_frag_init();
 935         if (err)
 936                 goto ipv6_frag_fail;
 937
 938         /* Init v6 transport protocols. */
 939         err = udpv6_init();
 940         if (err)
 941                 goto udpv6_fail;
 942
 943         err = udplitev6_init();
 944         if (err)
 945                 goto udplitev6_fail;
 946
 947         err = tcpv6_init();
 948         if (err)
 949                 goto tcpv6_fail;
 950
 951         err = ipv6_packet_init();
 952         if (err)
 953                 goto ipv6_packet_fail;
 954
 955         err = pingv6_init();
 956         if (err)
 957                 goto pingv6_fail;
 958
 959 #ifdef CONFIG_SYSCTL
 960         err = ipv6_sysctl_register();
 961         if (err)
 962                 goto sysctl_fail;
 963 #endif
 964 out:
 965         return err;
 966
 967 #ifdef CONFIG_SYSCTL
 968 sysctl_fail:
 969         pingv6_exit();
 970 #endif
 971 pingv6_fail:
 972         ipv6_packet_cleanup();
 973 ipv6_packet_fail:
 974         tcpv6_exit();
 975 tcpv6_fail:
 976         udplitev6_exit();
 977 udplitev6_fail:
 978         udpv6_exit();
 979 udpv6_fail:
 980         ipv6_frag_exit();
 981 ipv6_frag_fail:
 982         ipv6_exthdrs_exit();
 983 ipv6_exthdrs_fail:
 984         addrconf_cleanup();
 985 addrconf_fail:
 986         ip6_flowlabel_cleanup();
 987 ip6_flowlabel_fail:
 988         ndisc_late_cleanup();
 989 ndisc_late_fail:
 990         ip6_route_cleanup();
 991 ip6_route_fail:
 992 #ifdef CONFIG_PROC_FS
 993         if6_proc_exit();
 994 proc_if6_fail:
 995         ipv6_misc_proc_exit();
 996 proc_misc6_fail:
 997         udplite6_proc_exit();
 998 proc_udplite6_fail:
 999         raw6_proc_exit();
1000 proc_raw6_fail:
1001 #endif
1002         ipv6_netfilter_fini();
1003 netfilter_fail:
1004         igmp6_cleanup();
1005 igmp_fail:
1006         ndisc_cleanup();
1007 ndisc_fail:
1008         ip6_mr_cleanup();
1009 ipmr_fail:
1010         icmpv6_cleanup();
1011 icmp_fail:
1012         unregister_pernet_subsys(&inet6_net_ops);
1013 register_pernet_fail:
1014         sock_unregister(PF_INET6);
1015         rtnl_unregister_all(PF_INET6);
1016 out_sock_register_fail:
1017         rawv6_exit();
1018 out_unregister_ping_proto:
1019         proto_unregister(&pingv6_prot);
1020 out_unregister_raw_proto:
1021         proto_unregister(&rawv6_prot);
1022 out_unregister_udplite_proto:
1023         proto_unregister(&udplitev6_prot);
1024 out_unregister_udp_proto:
1025         proto_unregister(&udpv6_prot);
1026 out_unregister_tcp_proto:
1027         proto_unregister(&tcpv6_prot);
1028         goto out;
1029 }
1030 module_init(inet6_init);
1031
1032 MODULE_ALIAS_NETPROTO(PF_INET6);