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 #ifdef CONFIG_IPV6_TUNNEL
  60 #include <net/ip6_tunnel.h>
  61 #endif
  62
  63 #include <asm/uaccess.h>
  64 #include <linux/mroute6.h>
  65
  66 #ifdef CONFIG_ANDROID_PARANOID_NETWORK
  67 #include <linux/android_aid.h>
  68
  69 static inline int current_has_network(void)
  70 {
  71         return in_egroup_p(AID_INET) || capable(CAP_NET_RAW);
  72 }
  73 #else
  74 static inline int current_has_network(void)
  75 {
  76         return 1;
  77 }
  78 #endif
  79
  80 MODULE_AUTHOR("Cast of dozens");
  81 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  82 MODULE_LICENSE("GPL");
  83
  84 /* The inetsw6 table contains everything that inet6_create needs to
  85  * build a new socket.
  86  */
  87 static struct list_head inetsw6[SOCK_MAX];
  88 static DEFINE_SPINLOCK(inetsw6_lock);
  89
  90 struct ipv6_params ipv6_defaults = {
  91         .disable_ipv6 = 0,
  92         .autoconf = 1,
  93 };
  94
  95 static int disable_ipv6_mod;
  96
  97 module_param_named(disable, disable_ipv6_mod, int, 0444);
  98 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  99
 100 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
 101 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
 102
 103 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
 104 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
 105
 106 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
 107 {
 108         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
 109
 110         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
 111 }
 112
 113 static int inet6_create(struct net *net, struct socket *sock, int protocol,
 114                         int kern)
 115 {
 116         struct inet_sock *inet;
 117         struct ipv6_pinfo *np;
 118         struct sock *sk;
 119         struct inet_protosw *answer;
 120         struct proto *answer_prot;
 121         unsigned char answer_flags;
 122         char answer_no_check;
 123         int try_loading_module = 0;
 124         int err;
 125
 126         if (!current_has_network())
 127                 return -EACCES;
 128
 129         if (sock->type != SOCK_RAW &&
 130             sock->type != SOCK_DGRAM &&
 131             !inet_ehash_secret)
 132                 build_ehash_secret();
 133
 134         /* Look for the requested type/protocol pair. */
 135 lookup_protocol:
 136         err = -ESOCKTNOSUPPORT;
 137         rcu_read_lock();
 138         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 139
 140                 err = 0;
 141                 /* Check the non-wild match. */
 142                 if (protocol == answer->protocol) {
 143                         if (protocol != IPPROTO_IP)
 144                                 break;
 145                 } else {
 146                         /* Check for the two wild cases. */
 147                         if (IPPROTO_IP == protocol) {
 148                                 protocol = answer->protocol;
 149                                 break;
 150                         }
 151                         if (IPPROTO_IP == answer->protocol)
 152                                 break;
 153                 }
 154                 err = -EPROTONOSUPPORT;
 155         }
 156
 157         if (err) {
 158                 if (try_loading_module < 2) {
 159                         rcu_read_unlock();
 160                         /*
 161                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 162                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 163                          */
 164                         if (++try_loading_module == 1)
 165                                 request_module("net-pf-%d-proto-%d-type-%d",
 166                                                 PF_INET6, protocol, sock->type);
 167                         /*
 168                          * Fall back to generic, e.g. net-pf-10-proto-132
 169                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 170                          */
 171                         else
 172                                 request_module("net-pf-%d-proto-%d",
 173                                                 PF_INET6, protocol);
 174                         goto lookup_protocol;
 175                 } else
 176                         goto out_rcu_unlock;
 177         }
 178
 179         err = -EPERM;
 180         if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
 181                 goto out_rcu_unlock;
 182
 183         sock->ops = answer->ops;
 184         answer_prot = answer->prot;
 185         answer_no_check = answer->no_check;
 186         answer_flags = answer->flags;
 187         rcu_read_unlock();
 188
 189         WARN_ON(answer_prot->slab == NULL);
 190
 191         err = -ENOBUFS;
 192         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 193         if (sk == NULL)
 194                 goto out;
 195
 196         sock_init_data(sock, sk);
 197
 198         err = 0;
 199         sk->sk_no_check = answer_no_check;
 200         if (INET_PROTOSW_REUSE & answer_flags)
 201                 sk->sk_reuse = SK_CAN_REUSE;
 202
 203         inet = inet_sk(sk);
 204         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 205
 206         if (SOCK_RAW == sock->type) {
 207                 inet->inet_num = protocol;
 208                 if (IPPROTO_RAW == protocol)
 209                         inet->hdrincl = 1;
 210         }
 211
 212         sk->sk_destruct         = inet_sock_destruct;
 213         sk->sk_family           = PF_INET6;
 214         sk->sk_protocol         = protocol;
 215
 216         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 217
 218         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 219         np->hop_limit   = -1;
 220         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 221         np->mc_loop     = 1;
 222         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 223         np->ipv6only    = net->ipv6.sysctl.bindv6only;
 224
 225         /* Init the ipv4 part of the socket since we can have sockets
 226          * using v6 API for ipv4.
 227          */
 228         inet->uc_ttl    = -1;
 229
 230         inet->mc_loop   = 1;
 231         inet->mc_ttl    = 1;
 232         inet->mc_index  = 0;
 233         inet->mc_list   = NULL;
 234         inet->rcv_tos   = 0;
 235
 236         if (ipv4_config.no_pmtu_disc)
 237                 inet->pmtudisc = IP_PMTUDISC_DONT;
 238         else
 239                 inet->pmtudisc = IP_PMTUDISC_WANT;
 240         /*
 241          * Increment only the relevant sk_prot->socks debug field, this changes
 242          * the previous behaviour of incrementing both the equivalent to
 243          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 244          *
 245          * This allows better debug granularity as we'll know exactly how many
 246          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 247          * transport protocol socks. -acme
 248          */
 249         sk_refcnt_debug_inc(sk);
 250
 251         if (inet->inet_num) {
 252                 /* It assumes that any protocol which allows
 253                  * the user to assign a number at socket
 254                  * creation time automatically shares.
 255                  */
 256                 inet->inet_sport = htons(inet->inet_num);
 257                 sk->sk_prot->hash(sk);
 258         }
 259         if (sk->sk_prot->init) {
 260                 err = sk->sk_prot->init(sk);
 261                 if (err) {
 262                         sk_common_release(sk);
 263                         goto out;
 264                 }
 265         }
 266 out:
 267         return err;
 268 out_rcu_unlock:
 269         rcu_read_unlock();
 270         goto out;
 271 }
 272
 273
 274 /* bind for INET6 API */
 275 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 276 {
 277         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 278         struct sock *sk = sock->sk;
 279         struct inet_sock *inet = inet_sk(sk);
 280         struct ipv6_pinfo *np = inet6_sk(sk);
 281         struct net *net = sock_net(sk);
 282         __be32 v4addr = 0;
 283         unsigned short snum;
 284         int addr_type = 0;
 285         int err = 0;
 286
 287         /* If the socket has its own bind function then use it. */
 288         if (sk->sk_prot->bind)
 289                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 290
 291         if (addr_len < SIN6_LEN_RFC2133)
 292                 return -EINVAL;
 293
 294         if (addr->sin6_family != AF_INET6)
 295                 return -EAFNOSUPPORT;
 296
 297         addr_type = ipv6_addr_type(&addr->sin6_addr);
 298         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 299                 return -EINVAL;
 300
 301         snum = ntohs(addr->sin6_port);
 302         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 303                 return -EACCES;
 304
 305         lock_sock(sk);
 306
 307         /* Check these errors (active socket, double bind). */
 308         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 309                 err = -EINVAL;
 310                 goto out;
 311         }
 312
 313         /* Check if the address belongs to the host. */
 314         if (addr_type == IPV6_ADDR_MAPPED) {
 315                 int chk_addr_ret;
 316
 317                 /* Binding to v4-mapped address on a v6-only socket
 318                  * makes no sense
 319                  */
 320                 if (np->ipv6only) {
 321                         err = -EINVAL;
 322                         goto out;
 323                 }
 324
 325                 /* Reproduce AF_INET checks to make the bindings consistent */
 326                 v4addr = addr->sin6_addr.s6_addr32[3];
 327                 chk_addr_ret = inet_addr_type(net, v4addr);
 328                 if (!sysctl_ip_nonlocal_bind &&
 329                     !(inet->freebind || inet->transparent) &&
 330                     v4addr != htonl(INADDR_ANY) &&
 331                     chk_addr_ret != RTN_LOCAL &&
 332                     chk_addr_ret != RTN_MULTICAST &&
 333                     chk_addr_ret != RTN_BROADCAST) {
 334                         err = -EADDRNOTAVAIL;
 335                         goto out;
 336                 }
 337         } else {
 338                 if (addr_type != IPV6_ADDR_ANY) {
 339                         struct net_device *dev = NULL;
 340
 341                         rcu_read_lock();
 342                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 343                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 344                                     addr->sin6_scope_id) {
 345                                         /* Override any existing binding, if another one
 346                                          * is supplied by user.
 347                                          */
 348                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 349                                 }
 350
 351                                 /* Binding to link-local address requires an interface */
 352                                 if (!sk->sk_bound_dev_if) {
 353                                         err = -EINVAL;
 354                                         goto out_unlock;
 355                                 }
 356                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 357                                 if (!dev) {
 358                                         err = -ENODEV;
 359                                         goto out_unlock;
 360                                 }
 361                         }
 362
 363                         /* ipv4 addr of the socket is invalid.  Only the
 364                          * unspecified and mapped address have a v4 equivalent.
 365                          */
 366                         v4addr = LOOPBACK4_IPV6;
 367                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 368                                 if (!(inet->freebind || inet->transparent) &&
 369                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 370                                                    dev, 0)) {
 371                                         err = -EADDRNOTAVAIL;
 372                                         goto out_unlock;
 373                                 }
 374                         }
 375                         rcu_read_unlock();
 376                 }
 377         }
 378
 379         inet->inet_rcv_saddr = v4addr;
 380         inet->inet_saddr = v4addr;
 381
 382         np->rcv_saddr = addr->sin6_addr;
 383
 384         if (!(addr_type & IPV6_ADDR_MULTICAST))
 385                 np->saddr = addr->sin6_addr;
 386
 387         /* Make sure we are allowed to bind here. */
 388         if (sk->sk_prot->get_port(sk, snum)) {
 389                 inet_reset_saddr(sk);
 390                 err = -EADDRINUSE;
 391                 goto out;
 392         }
 393
 394         if (addr_type != IPV6_ADDR_ANY) {
 395                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 396                 if (addr_type != IPV6_ADDR_MAPPED)
 397                         np->ipv6only = 1;
 398         }
 399         if (snum)
 400                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 401         inet->inet_sport = htons(inet->inet_num);
 402         inet->inet_dport = 0;
 403         inet->inet_daddr = 0;
 404 out:
 405         release_sock(sk);
 406         return err;
 407 out_unlock:
 408         rcu_read_unlock();
 409         goto out;
 410 }
 411 EXPORT_SYMBOL(inet6_bind);
 412
 413 int inet6_release(struct socket *sock)
 414 {
 415         struct sock *sk = sock->sk;
 416
 417         if (sk == NULL)
 418                 return -EINVAL;
 419
 420         /* Free mc lists */
 421         ipv6_sock_mc_close(sk);
 422
 423         /* Free ac lists */
 424         ipv6_sock_ac_close(sk);
 425
 426         return inet_release(sock);
 427 }
 428 EXPORT_SYMBOL(inet6_release);
 429
 430 void inet6_destroy_sock(struct sock *sk)
 431 {
 432         struct ipv6_pinfo *np = inet6_sk(sk);
 433         struct sk_buff *skb;
 434         struct ipv6_txoptions *opt;
 435
 436         /* Release rx options */
 437
 438         skb = xchg(&np->pktoptions, NULL);
 439         if (skb != NULL)
 440                 kfree_skb(skb);
 441
 442         skb = xchg(&np->rxpmtu, NULL);
 443         if (skb != NULL)
 444                 kfree_skb(skb);
 445
 446         /* Free flowlabels */
 447         fl6_free_socklist(sk);
 448
 449         /* Free tx options */
 450
 451         opt = xchg(&np->opt, NULL);
 452         if (opt != NULL)
 453                 sock_kfree_s(sk, opt, opt->tot_len);
 454 }
 455 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 456
 457 /*
 458  *      This does both peername and sockname.
 459  */
 460
 461 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 462                  int *uaddr_len, int peer)
 463 {
 464         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 465         struct sock *sk = sock->sk;
 466         struct inet_sock *inet = inet_sk(sk);
 467         struct ipv6_pinfo *np = inet6_sk(sk);
 468
 469         sin->sin6_family = AF_INET6;
 470         sin->sin6_flowinfo = 0;
 471         sin->sin6_scope_id = 0;
 472         if (peer) {
 473                 if (!inet->inet_dport)
 474                         return -ENOTCONN;
 475                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 476                     peer == 1)
 477                         return -ENOTCONN;
 478                 sin->sin6_port = inet->inet_dport;
 479                 sin->sin6_addr = np->daddr;
 480                 if (np->sndflow)
 481                         sin->sin6_flowinfo = np->flow_label;
 482         } else {
 483                 if (ipv6_addr_any(&np->rcv_saddr))
 484                         sin->sin6_addr = np->saddr;
 485                 else
 486                         sin->sin6_addr = np->rcv_saddr;
 487
 488                 sin->sin6_port = inet->inet_sport;
 489         }
 490         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 491                                                  sk->sk_bound_dev_if);
 492         *uaddr_len = sizeof(*sin);
 493         return 0;
 494 }
 495 EXPORT_SYMBOL(inet6_getname);
 496
 497 int inet6_killaddr_ioctl(struct net *net, void __user *arg) {
 498         struct in6_ifreq ireq;
 499         struct sockaddr_in6 sin6;
 500
 501         if (!capable(CAP_NET_ADMIN))
 502                 return -EACCES;
 503
 504         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
 505                 return -EFAULT;
 506
 507         sin6.sin6_family = AF_INET6;
 508         sin6.sin6_addr = ireq.ifr6_addr;
 509         return tcp_nuke_addr(net, (struct sockaddr *) &sin6);
 510 }
 511
 512 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 513 {
 514         struct sock *sk = sock->sk;
 515         struct net *net = sock_net(sk);
 516
 517         switch (cmd) {
 518         case SIOCGSTAMP:
 519                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 520
 521         case SIOCGSTAMPNS:
 522                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 523
 524         case SIOCADDRT:
 525         case SIOCDELRT:
 526
 527                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 528
 529         case SIOCSIFADDR:
 530                 return addrconf_add_ifaddr(net, (void __user *) arg);
 531         case SIOCDIFADDR:
 532                 return addrconf_del_ifaddr(net, (void __user *) arg);
 533         case SIOCSIFDSTADDR:
 534                 return addrconf_set_dstaddr(net, (void __user *) arg);
 535         case SIOCKILLADDR:
 536                 return inet6_killaddr_ioctl(net, (void __user *) arg);
 537         default:
 538                 if (!sk->sk_prot->ioctl)
 539                         return -ENOIOCTLCMD;
 540                 return sk->sk_prot->ioctl(sk, cmd, arg);
 541         }
 542         /*NOTREACHED*/
 543         return 0;
 544 }
 545 EXPORT_SYMBOL(inet6_ioctl);
 546
 547 const struct proto_ops inet6_stream_ops = {
 548         .family            = PF_INET6,
 549         .owner             = THIS_MODULE,
 550         .release           = inet6_release,
 551         .bind              = inet6_bind,
 552         .connect           = inet_stream_connect,       /* ok           */
 553         .socketpair        = sock_no_socketpair,        /* a do nothing */
 554         .accept            = inet_accept,               /* ok           */
 555         .getname           = inet6_getname,
 556         .poll              = tcp_poll,                  /* ok           */
 557         .ioctl             = inet6_ioctl,               /* must change  */
 558         .listen            = inet_listen,               /* ok           */
 559         .shutdown          = inet_shutdown,             /* ok           */
 560         .setsockopt        = sock_common_setsockopt,    /* ok           */
 561         .getsockopt        = sock_common_getsockopt,    /* ok           */
 562         .sendmsg           = inet_sendmsg,              /* ok           */
 563         .recvmsg           = inet_recvmsg,              /* ok           */
 564         .mmap              = sock_no_mmap,
 565         .sendpage          = inet_sendpage,
 566         .splice_read       = tcp_splice_read,
 567 #ifdef CONFIG_COMPAT
 568         .compat_setsockopt = compat_sock_common_setsockopt,
 569         .compat_getsockopt = compat_sock_common_getsockopt,
 570 #endif
 571 };
 572
 573 const struct proto_ops inet6_dgram_ops = {
 574         .family            = PF_INET6,
 575         .owner             = THIS_MODULE,
 576         .release           = inet6_release,
 577         .bind              = inet6_bind,
 578         .connect           = inet_dgram_connect,        /* ok           */
 579         .socketpair        = sock_no_socketpair,        /* a do nothing */
 580         .accept            = sock_no_accept,            /* a do nothing */
 581         .getname           = inet6_getname,
 582         .poll              = udp_poll,                  /* ok           */
 583         .ioctl             = inet6_ioctl,               /* must change  */
 584         .listen            = sock_no_listen,            /* ok           */
 585         .shutdown          = inet_shutdown,             /* ok           */
 586         .setsockopt        = sock_common_setsockopt,    /* ok           */
 587         .getsockopt        = sock_common_getsockopt,    /* ok           */
 588         .sendmsg           = inet_sendmsg,              /* ok           */
 589         .recvmsg           = inet_recvmsg,              /* ok           */
 590         .mmap              = sock_no_mmap,
 591         .sendpage          = sock_no_sendpage,
 592 #ifdef CONFIG_COMPAT
 593         .compat_setsockopt = compat_sock_common_setsockopt,
 594         .compat_getsockopt = compat_sock_common_getsockopt,
 595 #endif
 596 };
 597
 598 static const struct net_proto_family inet6_family_ops = {
 599         .family = PF_INET6,
 600         .create = inet6_create,
 601         .owner  = THIS_MODULE,
 602 };
 603
 604 int inet6_register_protosw(struct inet_protosw *p)
 605 {
 606         struct list_head *lh;
 607         struct inet_protosw *answer;
 608         struct list_head *last_perm;
 609         int protocol = p->protocol;
 610         int ret;
 611
 612         spin_lock_bh(&inetsw6_lock);
 613
 614         ret = -EINVAL;
 615         if (p->type >= SOCK_MAX)
 616                 goto out_illegal;
 617
 618         /* If we are trying to override a permanent protocol, bail. */
 619         answer = NULL;
 620         ret = -EPERM;
 621         last_perm = &inetsw6[p->type];
 622         list_for_each(lh, &inetsw6[p->type]) {
 623                 answer = list_entry(lh, struct inet_protosw, list);
 624
 625                 /* Check only the non-wild match. */
 626                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 627                         if (protocol == answer->protocol)
 628                                 break;
 629                         last_perm = lh;
 630                 }
 631
 632                 answer = NULL;
 633         }
 634         if (answer)
 635                 goto out_permanent;
 636
 637         /* Add the new entry after the last permanent entry if any, so that
 638          * the new entry does not override a permanent entry when matched with
 639          * a wild-card protocol. But it is allowed to override any existing
 640          * non-permanent entry.  This means that when we remove this entry, the
 641          * system automatically returns to the old behavior.
 642          */
 643         list_add_rcu(&p->list, last_perm);
 644         ret = 0;
 645 out:
 646         spin_unlock_bh(&inetsw6_lock);
 647         return ret;
 648
 649 out_permanent:
 650         pr_err("Attempt to override permanent protocol %d\n", protocol);
 651         goto out;
 652
 653 out_illegal:
 654         pr_err("Ignoring attempt to register invalid socket type %d\n",
 655                p->type);
 656         goto out;
 657 }
 658 EXPORT_SYMBOL(inet6_register_protosw);
 659
 660 void
 661 inet6_unregister_protosw(struct inet_protosw *p)
 662 {
 663         if (INET_PROTOSW_PERMANENT & p->flags) {
 664                 pr_err("Attempt to unregister permanent protocol %d\n",
 665                        p->protocol);
 666         } else {
 667                 spin_lock_bh(&inetsw6_lock);
 668                 list_del_rcu(&p->list);
 669                 spin_unlock_bh(&inetsw6_lock);
 670
 671                 synchronize_net();
 672         }
 673 }
 674 EXPORT_SYMBOL(inet6_unregister_protosw);
 675
 676 int inet6_sk_rebuild_header(struct sock *sk)
 677 {
 678         struct ipv6_pinfo *np = inet6_sk(sk);
 679         struct dst_entry *dst;
 680
 681         dst = __sk_dst_check(sk, np->dst_cookie);
 682
 683         if (dst == NULL) {
 684                 struct inet_sock *inet = inet_sk(sk);
 685                 struct in6_addr *final_p, final;
 686                 struct flowi6 fl6;
 687
 688                 memset(&fl6, 0, sizeof(fl6));
 689                 fl6.flowi6_proto = sk->sk_protocol;
 690                 fl6.daddr = np->daddr;
 691                 fl6.saddr = np->saddr;
 692                 fl6.flowlabel = np->flow_label;
 693                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 694                 fl6.flowi6_mark = sk->sk_mark;
 695                 fl6.fl6_dport = inet->inet_dport;
 696                 fl6.fl6_sport = inet->inet_sport;
 697                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 698
 699                 final_p = fl6_update_dst(&fl6, np->opt, &final);
 700
 701                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
 702                 if (IS_ERR(dst)) {
 703                         sk->sk_route_caps = 0;
 704                         sk->sk_err_soft = -PTR_ERR(dst);
 705                         return PTR_ERR(dst);
 706                 }
 707
 708                 __ip6_dst_store(sk, dst, NULL, NULL);
 709         }
 710
 711         return 0;
 712 }
 713 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 714
 715 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
 716 {
 717         const struct ipv6_pinfo *np = inet6_sk(sk);
 718         const struct inet6_skb_parm *opt = IP6CB(skb);
 719
 720         if (np->rxopt.all) {
 721                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 722                                   np->rxopt.bits.ohopopts)) ||
 723                     ((IPV6_FLOWINFO_MASK &
 724                       *(__be32 *)skb_network_header(skb)) &&
 725                      np->rxopt.bits.rxflow) ||
 726                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 727                      np->rxopt.bits.osrcrt)) ||
 728                     ((opt->dst1 || opt->dst0) &&
 729                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 730                         return true;
 731         }
 732         return false;
 733 }
 734 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 735
 736 static struct packet_type ipv6_packet_type __read_mostly = {
 737         .type = cpu_to_be16(ETH_P_IPV6),
 738         .func = ipv6_rcv,
 739 };
 740
 741 static int __init ipv6_packet_init(void)
 742 {
 743         dev_add_pack(&ipv6_packet_type);
 744         return 0;
 745 }
 746
 747 static void ipv6_packet_cleanup(void)
 748 {
 749         dev_remove_pack(&ipv6_packet_type);
 750 }
 751
 752 static int __net_init ipv6_init_mibs(struct net *net)
 753 {
 754         if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
 755                           sizeof(struct udp_mib),
 756                           __alignof__(struct udp_mib)) < 0)
 757                 return -ENOMEM;
 758         if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
 759                           sizeof(struct udp_mib),
 760                           __alignof__(struct udp_mib)) < 0)
 761                 goto err_udplite_mib;
 762         if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
 763                           sizeof(struct ipstats_mib),
 764                           __alignof__(struct ipstats_mib)) < 0)
 765                 goto err_ip_mib;
 766         if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
 767                           sizeof(struct icmpv6_mib),
 768                           __alignof__(struct icmpv6_mib)) < 0)
 769                 goto err_icmp_mib;
 770         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 771                                                 GFP_KERNEL);
 772         if (!net->mib.icmpv6msg_statistics)
 773                 goto err_icmpmsg_mib;
 774         return 0;
 775
 776 err_icmpmsg_mib:
 777         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 778 err_icmp_mib:
 779         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 780 err_ip_mib:
 781         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 782 err_udplite_mib:
 783         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 784         return -ENOMEM;
 785 }
 786
 787 static void ipv6_cleanup_mibs(struct net *net)
 788 {
 789         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 790         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 791         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 792         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 793         kfree(net->mib.icmpv6msg_statistics);
 794 }
 795
 796 static int __net_init inet6_net_init(struct net *net)
 797 {
 798         int err = 0;
 799
 800         net->ipv6.sysctl.bindv6only = 0;
 801         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 802
 803         err = ipv6_init_mibs(net);
 804         if (err)
 805                 return err;
 806 #ifdef CONFIG_PROC_FS
 807         err = udp6_proc_init(net);
 808         if (err)
 809                 goto out;
 810         err = tcp6_proc_init(net);
 811         if (err)
 812                 goto proc_tcp6_fail;
 813         err = ac6_proc_init(net);
 814         if (err)
 815                 goto proc_ac6_fail;
 816 #endif
 817         return err;
 818
 819 #ifdef CONFIG_PROC_FS
 820 proc_ac6_fail:
 821         tcp6_proc_exit(net);
 822 proc_tcp6_fail:
 823         udp6_proc_exit(net);
 824 out:
 825         ipv6_cleanup_mibs(net);
 826         return err;
 827 #endif
 828 }
 829
 830 static void __net_exit inet6_net_exit(struct net *net)
 831 {
 832 #ifdef CONFIG_PROC_FS
 833         udp6_proc_exit(net);
 834         tcp6_proc_exit(net);
 835         ac6_proc_exit(net);
 836 #endif
 837         ipv6_cleanup_mibs(net);
 838 }
 839
 840 static struct pernet_operations inet6_net_ops = {
 841         .init = inet6_net_init,
 842         .exit = inet6_net_exit,
 843 };
 844
 845 static int __init inet6_init(void)
 846 {
 847         struct list_head *r;
 848         int err = 0;
 849
 850         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
 851
 852         /* Register the socket-side information for inet6_create.  */
 853         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 854                 INIT_LIST_HEAD(r);
 855
 856         if (disable_ipv6_mod) {
 857                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 858                 goto out;
 859         }
 860
 861         err = proto_register(&tcpv6_prot, 1);
 862         if (err)
 863                 goto out;
 864
 865         err = proto_register(&udpv6_prot, 1);
 866         if (err)
 867                 goto out_unregister_tcp_proto;
 868
 869         err = proto_register(&udplitev6_prot, 1);
 870         if (err)
 871                 goto out_unregister_udp_proto;
 872
 873         err = proto_register(&rawv6_prot, 1);
 874         if (err)
 875                 goto out_unregister_udplite_proto;
 876
 877         err = proto_register(&pingv6_prot, 1);
 878         if (err)
 879                 goto out_unregister_ping_proto;
 880
 881         /* We MUST register RAW sockets before we create the ICMP6,
 882          * IGMP6, or NDISC control sockets.
 883          */
 884         err = rawv6_init();
 885         if (err)
 886                 goto out_unregister_raw_proto;
 887
 888         /* Register the family here so that the init calls below will
 889          * be able to create sockets. (?? is this dangerous ??)
 890          */
 891         err = sock_register(&inet6_family_ops);
 892         if (err)
 893                 goto out_sock_register_fail;
 894
 895         tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
 896
 897         /*
 898          *      ipngwg API draft makes clear that the correct semantics
 899          *      for TCP and UDP is to consider one TCP and UDP instance
 900          *      in a host available by both INET and INET6 APIs and
 901          *      able to communicate via both network protocols.
 902          */
 903
 904         err = register_pernet_subsys(&inet6_net_ops);
 905         if (err)
 906                 goto register_pernet_fail;
 907         err = icmpv6_init();
 908         if (err)
 909                 goto icmp_fail;
 910         err = ip6_mr_init();
 911         if (err)
 912                 goto ipmr_fail;
 913         err = ndisc_init();
 914         if (err)
 915                 goto ndisc_fail;
 916         err = igmp6_init();
 917         if (err)
 918                 goto igmp_fail;
 919         err = ipv6_netfilter_init();
 920         if (err)
 921                 goto netfilter_fail;
 922         /* Create /proc/foo6 entries. */
 923 #ifdef CONFIG_PROC_FS
 924         err = -ENOMEM;
 925         if (raw6_proc_init())
 926                 goto proc_raw6_fail;
 927         if (udplite6_proc_init())
 928                 goto proc_udplite6_fail;
 929         if (ipv6_misc_proc_init())
 930                 goto proc_misc6_fail;
 931         if (if6_proc_init())
 932                 goto proc_if6_fail;
 933 #endif
 934         err = ip6_route_init();
 935         if (err)
 936                 goto ip6_route_fail;
 937         err = ip6_flowlabel_init();
 938         if (err)
 939                 goto ip6_flowlabel_fail;
 940         err = addrconf_init();
 941         if (err)
 942                 goto addrconf_fail;
 943
 944         /* Init v6 extension headers. */
 945         err = ipv6_exthdrs_init();
 946         if (err)
 947                 goto ipv6_exthdrs_fail;
 948
 949         err = ipv6_frag_init();
 950         if (err)
 951                 goto ipv6_frag_fail;
 952
 953         /* Init v6 transport protocols. */
 954         err = udpv6_init();
 955         if (err)
 956                 goto udpv6_fail;
 957
 958         err = udplitev6_init();
 959         if (err)
 960                 goto udplitev6_fail;
 961
 962         err = tcpv6_init();
 963         if (err)
 964                 goto tcpv6_fail;
 965
 966         err = ipv6_packet_init();
 967         if (err)
 968                 goto ipv6_packet_fail;
 969
 970         err = pingv6_init();
 971         if (err)
 972                 goto pingv6_fail;
 973
 974 #ifdef CONFIG_SYSCTL
 975         err = ipv6_sysctl_register();
 976         if (err)
 977                 goto sysctl_fail;
 978 #endif
 979 out:
 980         return err;
 981
 982 #ifdef CONFIG_SYSCTL
 983 sysctl_fail:
 984         ipv6_packet_cleanup();
 985 #endif
 986 pingv6_fail:
 987         pingv6_exit();
 988 ipv6_packet_fail:
 989         tcpv6_exit();
 990 tcpv6_fail:
 991         udplitev6_exit();
 992 udplitev6_fail:
 993         udpv6_exit();
 994 udpv6_fail:
 995         ipv6_frag_exit();
 996 ipv6_frag_fail:
 997         ipv6_exthdrs_exit();
 998 ipv6_exthdrs_fail:
 999         addrconf_cleanup();
1000 addrconf_fail:
1001         ip6_flowlabel_cleanup();
1002 ip6_flowlabel_fail:
1003         ip6_route_cleanup();
1004 ip6_route_fail:
1005 #ifdef CONFIG_PROC_FS
1006         if6_proc_exit();
1007 proc_if6_fail:
1008         ipv6_misc_proc_exit();
1009 proc_misc6_fail:
1010         udplite6_proc_exit();
1011 proc_udplite6_fail:
1012         raw6_proc_exit();
1013 proc_raw6_fail:
1014 #endif
1015         ipv6_netfilter_fini();
1016 netfilter_fail:
1017         igmp6_cleanup();
1018 igmp_fail:
1019         ndisc_cleanup();
1020 ndisc_fail:
1021         ip6_mr_cleanup();
1022 ipmr_fail:
1023         icmpv6_cleanup();
1024 icmp_fail:
1025         unregister_pernet_subsys(&inet6_net_ops);
1026 register_pernet_fail:
1027         sock_unregister(PF_INET6);
1028         rtnl_unregister_all(PF_INET6);
1029 out_sock_register_fail:
1030         rawv6_exit();
1031 out_unregister_ping_proto:
1032         proto_unregister(&pingv6_prot);
1033 out_unregister_raw_proto:
1034         proto_unregister(&rawv6_prot);
1035 out_unregister_udplite_proto:
1036         proto_unregister(&udplitev6_prot);
1037 out_unregister_udp_proto:
1038         proto_unregister(&udpv6_prot);
1039 out_unregister_tcp_proto:
1040         proto_unregister(&tcpv6_prot);
1041         goto out;
1042 }
1043 module_init(inet6_init);
1044
1045 static void __exit inet6_exit(void)
1046 {
1047         if (disable_ipv6_mod)
1048                 return;
1049
1050         /* First of all disallow new sockets creation. */
1051         sock_unregister(PF_INET6);
1052         /* Disallow any further netlink messages */
1053         rtnl_unregister_all(PF_INET6);
1054
1055         udpv6_exit();
1056         udplitev6_exit();
1057         tcpv6_exit();
1058
1059         /* Cleanup code parts. */
1060         ipv6_packet_cleanup();
1061         ipv6_frag_exit();
1062         ipv6_exthdrs_exit();
1063         addrconf_cleanup();
1064         ip6_flowlabel_cleanup();
1065         ip6_route_cleanup();
1066 #ifdef CONFIG_PROC_FS
1067
1068         /* Cleanup code parts. */
1069         if6_proc_exit();
1070         ipv6_misc_proc_exit();
1071         udplite6_proc_exit();
1072         raw6_proc_exit();
1073 #endif
1074         ipv6_netfilter_fini();
1075         igmp6_cleanup();
1076         ndisc_cleanup();
1077         ip6_mr_cleanup();
1078         icmpv6_cleanup();
1079         rawv6_exit();
1080
1081         unregister_pernet_subsys(&inet6_net_ops);
1082         proto_unregister(&rawv6_prot);
1083         proto_unregister(&udplitev6_prot);
1084         proto_unregister(&udpv6_prot);
1085         proto_unregister(&tcpv6_prot);
1086
1087         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1088 }
1089 module_exit(inet6_exit);
1090
1091 MODULE_ALIAS_NETPROTO(PF_INET6);