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 #ifdef CONFIG_ANDROID_PARANOID_NETWORK
  68 #include <linux/android_aid.h>
  69
  70 static inline int current_has_network(void)
  71 {
  72         return in_egroup_p(AID_INET) || capable(CAP_NET_RAW);
  73 }
  74 #else
  75 static inline int current_has_network(void)
  76 {
  77         return 1;
  78 }
  79 #endif
  80
  81 MODULE_AUTHOR("Cast of dozens");
  82 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  83 MODULE_LICENSE("GPL");
  84
  85 /* The inetsw6 table contains everything that inet6_create needs to
  86  * build a new socket.
  87  */
  88 static struct list_head inetsw6[SOCK_MAX];
  89 static DEFINE_SPINLOCK(inetsw6_lock);
  90
  91 struct ipv6_params ipv6_defaults = {
  92         .disable_ipv6 = 0,
  93         .autoconf = 1,
  94 };
  95
  96 static int disable_ipv6_mod;
  97
  98 module_param_named(disable, disable_ipv6_mod, int, 0444);
  99 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
 100
 101 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
 102 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
 103
 104 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
 105 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
 106
 107 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
 108 {
 109         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
 110
 111         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
 112 }
 113
 114 static int inet6_create(struct net *net, struct socket *sock, int protocol,
 115                         int kern)
 116 {
 117         struct inet_sock *inet;
 118         struct ipv6_pinfo *np;
 119         struct sock *sk;
 120         struct inet_protosw *answer;
 121         struct proto *answer_prot;
 122         unsigned char answer_flags;
 123         int try_loading_module = 0;
 124         int err;
 125
 126         if (protocol < 0 || protocol >= IPPROTO_MAX)
 127                 return -EINVAL;
 128
 129         if (!current_has_network())
 130                 return -EACCES;
 131
 132         /* Look for the requested type/protocol pair. */
 133 lookup_protocol:
 134         err = -ESOCKTNOSUPPORT;
 135         rcu_read_lock();
 136         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 137
 138                 err = 0;
 139                 /* Check the non-wild match. */
 140                 if (protocol == answer->protocol) {
 141                         if (protocol != IPPROTO_IP)
 142                                 break;
 143                 } else {
 144                         /* Check for the two wild cases. */
 145                         if (IPPROTO_IP == protocol) {
 146                                 protocol = answer->protocol;
 147                                 break;
 148                         }
 149                         if (IPPROTO_IP == answer->protocol)
 150                                 break;
 151                 }
 152                 err = -EPROTONOSUPPORT;
 153         }
 154
 155         if (err) {
 156                 if (try_loading_module < 2) {
 157                         rcu_read_unlock();
 158                         /*
 159                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 160                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 161                          */
 162                         if (++try_loading_module == 1)
 163                                 request_module("net-pf-%d-proto-%d-type-%d",
 164                                                 PF_INET6, protocol, sock->type);
 165                         /*
 166                          * Fall back to generic, e.g. net-pf-10-proto-132
 167                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 168                          */
 169                         else
 170                                 request_module("net-pf-%d-proto-%d",
 171                                                 PF_INET6, protocol);
 172                         goto lookup_protocol;
 173                 } else
 174                         goto out_rcu_unlock;
 175         }
 176
 177         err = -EPERM;
 178         if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
 179                 goto out_rcu_unlock;
 180
 181         sock->ops = answer->ops;
 182         answer_prot = answer->prot;
 183         answer_flags = answer->flags;
 184         rcu_read_unlock();
 185
 186         WARN_ON(!answer_prot->slab);
 187
 188         err = -ENOBUFS;
 189         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
 190         if (!sk)
 191                 goto out;
 192
 193         sock_init_data(sock, sk);
 194
 195         err = 0;
 196         if (INET_PROTOSW_REUSE & answer_flags)
 197                 sk->sk_reuse = SK_CAN_REUSE;
 198
 199         inet = inet_sk(sk);
 200         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 201
 202         if (SOCK_RAW == sock->type) {
 203                 inet->inet_num = protocol;
 204                 if (IPPROTO_RAW == protocol)
 205                         inet->hdrincl = 1;
 206         }
 207
 208         sk->sk_destruct         = inet_sock_destruct;
 209         sk->sk_family           = PF_INET6;
 210         sk->sk_protocol         = protocol;
 211
 212         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 213
 214         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 215         np->hop_limit   = -1;
 216         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 217         np->mc_loop     = 1;
 218         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 219         np->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
 220         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
 221
 222         /* Init the ipv4 part of the socket since we can have sockets
 223          * using v6 API for ipv4.
 224          */
 225         inet->uc_ttl    = -1;
 226
 227         inet->mc_loop   = 1;
 228         inet->mc_ttl    = 1;
 229         inet->mc_index  = 0;
 230         inet->mc_list   = NULL;
 231         inet->rcv_tos   = 0;
 232
 233         if (net->ipv4.sysctl_ip_no_pmtu_disc)
 234                 inet->pmtudisc = IP_PMTUDISC_DONT;
 235         else
 236                 inet->pmtudisc = IP_PMTUDISC_WANT;
 237         /*
 238          * Increment only the relevant sk_prot->socks debug field, this changes
 239          * the previous behaviour of incrementing both the equivalent to
 240          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 241          *
 242          * This allows better debug granularity as we'll know exactly how many
 243          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 244          * transport protocol socks. -acme
 245          */
 246         sk_refcnt_debug_inc(sk);
 247
 248         if (inet->inet_num) {
 249                 /* It assumes that any protocol which allows
 250                  * the user to assign a number at socket
 251                  * creation time automatically shares.
 252                  */
 253                 inet->inet_sport = htons(inet->inet_num);
 254                 sk->sk_prot->hash(sk);
 255         }
 256         if (sk->sk_prot->init) {
 257                 err = sk->sk_prot->init(sk);
 258                 if (err) {
 259                         sk_common_release(sk);
 260                         goto out;
 261                 }
 262         }
 263 out:
 264         return err;
 265 out_rcu_unlock:
 266         rcu_read_unlock();
 267         goto out;
 268 }
 269
 270
 271 /* bind for INET6 API */
 272 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 273 {
 274         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 275         struct sock *sk = sock->sk;
 276         struct inet_sock *inet = inet_sk(sk);
 277         struct ipv6_pinfo *np = inet6_sk(sk);
 278         struct net *net = sock_net(sk);
 279         __be32 v4addr = 0;
 280         unsigned short snum;
 281         int addr_type = 0;
 282         int err = 0;
 283
 284         /* If the socket has its own bind function then use it. */
 285         if (sk->sk_prot->bind)
 286                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 287
 288         if (addr_len < SIN6_LEN_RFC2133)
 289                 return -EINVAL;
 290
 291         if (addr->sin6_family != AF_INET6)
 292                 return -EAFNOSUPPORT;
 293
 294         addr_type = ipv6_addr_type(&addr->sin6_addr);
 295         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 296                 return -EINVAL;
 297
 298         snum = ntohs(addr->sin6_port);
 299         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 300                 return -EACCES;
 301
 302         lock_sock(sk);
 303
 304         /* Check these errors (active socket, double bind). */
 305         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 306                 err = -EINVAL;
 307                 goto out;
 308         }
 309
 310         /* Check if the address belongs to the host. */
 311         if (addr_type == IPV6_ADDR_MAPPED) {
 312                 int chk_addr_ret;
 313
 314                 /* Binding to v4-mapped address on a v6-only socket
 315                  * makes no sense
 316                  */
 317                 if (sk->sk_ipv6only) {
 318                         err = -EINVAL;
 319                         goto out;
 320                 }
 321
 322                 /* Reproduce AF_INET checks to make the bindings consistent */
 323                 v4addr = addr->sin6_addr.s6_addr32[3];
 324                 chk_addr_ret = inet_addr_type(net, v4addr);
 325                 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
 326                     !(inet->freebind || inet->transparent) &&
 327                     v4addr != htonl(INADDR_ANY) &&
 328                     chk_addr_ret != RTN_LOCAL &&
 329                     chk_addr_ret != RTN_MULTICAST &&
 330                     chk_addr_ret != RTN_BROADCAST) {
 331                         err = -EADDRNOTAVAIL;
 332                         goto out;
 333                 }
 334         } else {
 335                 if (addr_type != IPV6_ADDR_ANY) {
 336                         struct net_device *dev = NULL;
 337
 338                         rcu_read_lock();
 339                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 340                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 341                                     addr->sin6_scope_id) {
 342                                         /* Override any existing binding, if another one
 343                                          * is supplied by user.
 344                                          */
 345                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 346                                 }
 347
 348                                 /* Binding to link-local address requires an interface */
 349                                 if (!sk->sk_bound_dev_if) {
 350                                         err = -EINVAL;
 351                                         goto out_unlock;
 352                                 }
 353                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 354                                 if (!dev) {
 355                                         err = -ENODEV;
 356                                         goto out_unlock;
 357                                 }
 358                         }
 359
 360                         /* ipv4 addr of the socket is invalid.  Only the
 361                          * unspecified and mapped address have a v4 equivalent.
 362                          */
 363                         v4addr = LOOPBACK4_IPV6;
 364                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 365                                 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
 366                                     !(inet->freebind || inet->transparent) &&
 367                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 368                                                    dev, 0)) {
 369                                         err = -EADDRNOTAVAIL;
 370                                         goto out_unlock;
 371                                 }
 372                         }
 373                         rcu_read_unlock();
 374                 }
 375         }
 376
 377         inet->inet_rcv_saddr = v4addr;
 378         inet->inet_saddr = v4addr;
 379
 380         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 381
 382         if (!(addr_type & IPV6_ADDR_MULTICAST))
 383                 np->saddr = addr->sin6_addr;
 384
 385         /* Make sure we are allowed to bind here. */
 386         if ((snum || !inet->bind_address_no_port) &&
 387             sk->sk_prot->get_port(sk, snum)) {
 388                 inet_reset_saddr(sk);
 389                 err = -EADDRINUSE;
 390                 goto out;
 391         }
 392
 393         if (addr_type != IPV6_ADDR_ANY) {
 394                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 395                 if (addr_type != IPV6_ADDR_MAPPED)
 396                         sk->sk_ipv6only = 1;
 397         }
 398         if (snum)
 399                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 400         inet->inet_sport = htons(inet->inet_num);
 401         inet->inet_dport = 0;
 402         inet->inet_daddr = 0;
 403 out:
 404         release_sock(sk);
 405         return err;
 406 out_unlock:
 407         rcu_read_unlock();
 408         goto out;
 409 }
 410 EXPORT_SYMBOL(inet6_bind);
 411
 412 int inet6_release(struct socket *sock)
 413 {
 414         struct sock *sk = sock->sk;
 415
 416         if (!sk)
 417                 return -EINVAL;
 418
 419         /* Free mc lists */
 420         ipv6_sock_mc_close(sk);
 421
 422         /* Free ac lists */
 423         ipv6_sock_ac_close(sk);
 424
 425         return inet_release(sock);
 426 }
 427 EXPORT_SYMBOL(inet6_release);
 428
 429 void inet6_destroy_sock(struct sock *sk)
 430 {
 431         struct ipv6_pinfo *np = inet6_sk(sk);
 432         struct sk_buff *skb;
 433         struct ipv6_txoptions *opt;
 434
 435         /* Release rx options */
 436
 437         skb = xchg(&np->pktoptions, NULL);
 438         if (skb)
 439                 kfree_skb(skb);
 440
 441         skb = xchg(&np->rxpmtu, NULL);
 442         if (skb)
 443                 kfree_skb(skb);
 444
 445         /* Free flowlabels */
 446         fl6_free_socklist(sk);
 447
 448         /* Free tx options */
 449
 450         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 451         if (opt) {
 452                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 453                 txopt_put(opt);
 454         }
 455 }
 456 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 457
 458 /*
 459  *      This does both peername and sockname.
 460  */
 461
 462 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 463                  int *uaddr_len, int peer)
 464 {
 465         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 466         struct sock *sk = sock->sk;
 467         struct inet_sock *inet = inet_sk(sk);
 468         struct ipv6_pinfo *np = inet6_sk(sk);
 469
 470         sin->sin6_family = AF_INET6;
 471         sin->sin6_flowinfo = 0;
 472         sin->sin6_scope_id = 0;
 473         if (peer) {
 474                 if (!inet->inet_dport)
 475                         return -ENOTCONN;
 476                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 477                     peer == 1)
 478                         return -ENOTCONN;
 479                 sin->sin6_port = inet->inet_dport;
 480                 sin->sin6_addr = sk->sk_v6_daddr;
 481                 if (np->sndflow)
 482                         sin->sin6_flowinfo = np->flow_label;
 483         } else {
 484                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 485                         sin->sin6_addr = np->saddr;
 486                 else
 487                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 488
 489                 sin->sin6_port = inet->inet_sport;
 490         }
 491         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 492                                                  sk->sk_bound_dev_if);
 493         *uaddr_len = sizeof(*sin);
 494         return 0;
 495 }
 496 EXPORT_SYMBOL(inet6_getname);
 497
 498 int inet6_killaddr_ioctl(struct net *net, void __user *arg) {
 499         struct in6_ifreq ireq;
 500         struct sockaddr_in6 sin6;
 501
 502         if (!capable(CAP_NET_ADMIN))
 503                 return -EACCES;
 504
 505         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
 506                 return -EFAULT;
 507
 508         sin6.sin6_family = AF_INET6;
 509         sin6.sin6_addr = ireq.ifr6_addr;
 510         return tcp_nuke_addr(net, (struct sockaddr *) &sin6);
 511 }
 512
 513 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 514 {
 515         struct sock *sk = sock->sk;
 516         struct net *net = sock_net(sk);
 517
 518         switch (cmd) {
 519         case SIOCGSTAMP:
 520                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 521
 522         case SIOCGSTAMPNS:
 523                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 524
 525         case SIOCADDRT:
 526         case SIOCDELRT:
 527
 528                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 529
 530         case SIOCSIFADDR:
 531                 return addrconf_add_ifaddr(net, (void __user *) arg);
 532         case SIOCDIFADDR:
 533                 return addrconf_del_ifaddr(net, (void __user *) arg);
 534         case SIOCSIFDSTADDR:
 535                 return addrconf_set_dstaddr(net, (void __user *) arg);
 536         case SIOCKILLADDR:
 537                 return inet6_killaddr_ioctl(net, (void __user *) arg);
 538         default:
 539                 if (!sk->sk_prot->ioctl)
 540                         return -ENOIOCTLCMD;
 541                 return sk->sk_prot->ioctl(sk, cmd, arg);
 542         }
 543         /*NOTREACHED*/
 544         return 0;
 545 }
 546 EXPORT_SYMBOL(inet6_ioctl);
 547
 548 const struct proto_ops inet6_stream_ops = {
 549         .family            = PF_INET6,
 550         .owner             = THIS_MODULE,
 551         .release           = inet6_release,
 552         .bind              = inet6_bind,
 553         .connect           = inet_stream_connect,       /* ok           */
 554         .socketpair        = sock_no_socketpair,        /* a do nothing */
 555         .accept            = inet_accept,               /* ok           */
 556         .getname           = inet6_getname,
 557         .poll              = tcp_poll,                  /* ok           */
 558         .ioctl             = inet6_ioctl,               /* must change  */
 559         .listen            = inet_listen,               /* ok           */
 560         .shutdown          = inet_shutdown,             /* ok           */
 561         .setsockopt        = sock_common_setsockopt,    /* ok           */
 562         .getsockopt        = sock_common_getsockopt,    /* ok           */
 563         .sendmsg           = inet_sendmsg,              /* ok           */
 564         .recvmsg           = inet_recvmsg,              /* ok           */
 565         .mmap              = sock_no_mmap,
 566         .sendpage          = inet_sendpage,
 567         .splice_read       = tcp_splice_read,
 568 #ifdef CONFIG_COMPAT
 569         .compat_setsockopt = compat_sock_common_setsockopt,
 570         .compat_getsockopt = compat_sock_common_getsockopt,
 571 #endif
 572 };
 573
 574 const struct proto_ops inet6_dgram_ops = {
 575         .family            = PF_INET6,
 576         .owner             = THIS_MODULE,
 577         .release           = inet6_release,
 578         .bind              = inet6_bind,
 579         .connect           = inet_dgram_connect,        /* ok           */
 580         .socketpair        = sock_no_socketpair,        /* a do nothing */
 581         .accept            = sock_no_accept,            /* a do nothing */
 582         .getname           = inet6_getname,
 583         .poll              = udp_poll,                  /* ok           */
 584         .ioctl             = inet6_ioctl,               /* must change  */
 585         .listen            = sock_no_listen,            /* ok           */
 586         .shutdown          = inet_shutdown,             /* ok           */
 587         .setsockopt        = sock_common_setsockopt,    /* ok           */
 588         .getsockopt        = sock_common_getsockopt,    /* ok           */
 589         .sendmsg           = inet_sendmsg,              /* ok           */
 590         .recvmsg           = inet_recvmsg,              /* ok           */
 591         .mmap              = sock_no_mmap,
 592         .sendpage          = sock_no_sendpage,
 593 #ifdef CONFIG_COMPAT
 594         .compat_setsockopt = compat_sock_common_setsockopt,
 595         .compat_getsockopt = compat_sock_common_getsockopt,
 596 #endif
 597 };
 598
 599 static const struct net_proto_family inet6_family_ops = {
 600         .family = PF_INET6,
 601         .create = inet6_create,
 602         .owner  = THIS_MODULE,
 603 };
 604
 605 int inet6_register_protosw(struct inet_protosw *p)
 606 {
 607         struct list_head *lh;
 608         struct inet_protosw *answer;
 609         struct list_head *last_perm;
 610         int protocol = p->protocol;
 611         int ret;
 612
 613         spin_lock_bh(&inetsw6_lock);
 614
 615         ret = -EINVAL;
 616         if (p->type >= SOCK_MAX)
 617                 goto out_illegal;
 618
 619         /* If we are trying to override a permanent protocol, bail. */
 620         answer = NULL;
 621         ret = -EPERM;
 622         last_perm = &inetsw6[p->type];
 623         list_for_each(lh, &inetsw6[p->type]) {
 624                 answer = list_entry(lh, struct inet_protosw, list);
 625
 626                 /* Check only the non-wild match. */
 627                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 628                         if (protocol == answer->protocol)
 629                                 break;
 630                         last_perm = lh;
 631                 }
 632
 633                 answer = NULL;
 634         }
 635         if (answer)
 636                 goto out_permanent;
 637
 638         /* Add the new entry after the last permanent entry if any, so that
 639          * the new entry does not override a permanent entry when matched with
 640          * a wild-card protocol. But it is allowed to override any existing
 641          * non-permanent entry.  This means that when we remove this entry, the
 642          * system automatically returns to the old behavior.
 643          */
 644         list_add_rcu(&p->list, last_perm);
 645         ret = 0;
 646 out:
 647         spin_unlock_bh(&inetsw6_lock);
 648         return ret;
 649
 650 out_permanent:
 651         pr_err("Attempt to override permanent protocol %d\n", protocol);
 652         goto out;
 653
 654 out_illegal:
 655         pr_err("Ignoring attempt to register invalid socket type %d\n",
 656                p->type);
 657         goto out;
 658 }
 659 EXPORT_SYMBOL(inet6_register_protosw);
 660
 661 void
 662 inet6_unregister_protosw(struct inet_protosw *p)
 663 {
 664         if (INET_PROTOSW_PERMANENT & p->flags) {
 665                 pr_err("Attempt to unregister permanent protocol %d\n",
 666                        p->protocol);
 667         } else {
 668                 spin_lock_bh(&inetsw6_lock);
 669                 list_del_rcu(&p->list);
 670                 spin_unlock_bh(&inetsw6_lock);
 671
 672                 synchronize_net();
 673         }
 674 }
 675 EXPORT_SYMBOL(inet6_unregister_protosw);
 676
 677 int inet6_sk_rebuild_header(struct sock *sk)
 678 {
 679         struct ipv6_pinfo *np = inet6_sk(sk);
 680         struct dst_entry *dst;
 681
 682         dst = __sk_dst_check(sk, np->dst_cookie);
 683
 684         if (!dst) {
 685                 struct inet_sock *inet = inet_sk(sk);
 686                 struct in6_addr *final_p, final;
 687                 struct flowi6 fl6;
 688
 689                 memset(&fl6, 0, sizeof(fl6));
 690                 fl6.flowi6_proto = sk->sk_protocol;
 691                 fl6.daddr = sk->sk_v6_daddr;
 692                 fl6.saddr = np->saddr;
 693                 fl6.flowlabel = np->flow_label;
 694                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 695                 fl6.flowi6_mark = sk->sk_mark;
 696                 fl6.fl6_dport = inet->inet_dport;
 697                 fl6.fl6_sport = inet->inet_sport;
 698                 fl6.flowi6_uid = sock_i_uid(sk);
 699                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 700
 701                 rcu_read_lock();
 702                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 703                                          &final);
 704                 rcu_read_unlock();
 705
 706                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 707                 if (IS_ERR(dst)) {
 708                         sk->sk_route_caps = 0;
 709                         sk->sk_err_soft = -PTR_ERR(dst);
 710                         return PTR_ERR(dst);
 711                 }
 712
 713                 ip6_dst_store(sk, dst, NULL, NULL);
 714         }
 715
 716         return 0;
 717 }
 718 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 719
 720 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 721                        const struct inet6_skb_parm *opt)
 722 {
 723         const struct ipv6_pinfo *np = inet6_sk(sk);
 724
 725         if (np->rxopt.all) {
 726                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
 727                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 728                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 729                      np->rxopt.bits.rxflow) ||
 730                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 731                      np->rxopt.bits.osrcrt)) ||
 732                     ((opt->dst1 || opt->dst0) &&
 733                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 734                         return true;
 735         }
 736         return false;
 737 }
 738 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 739
 740 static struct packet_type ipv6_packet_type __read_mostly = {
 741         .type = cpu_to_be16(ETH_P_IPV6),
 742         .func = ipv6_rcv,
 743 };
 744
 745 static int __init ipv6_packet_init(void)
 746 {
 747         dev_add_pack(&ipv6_packet_type);
 748         return 0;
 749 }
 750
 751 static void ipv6_packet_cleanup(void)
 752 {
 753         dev_remove_pack(&ipv6_packet_type);
 754 }
 755
 756 static int __net_init ipv6_init_mibs(struct net *net)
 757 {
 758         int i;
 759
 760         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 761         if (!net->mib.udp_stats_in6)
 762                 return -ENOMEM;
 763         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 764         if (!net->mib.udplite_stats_in6)
 765                 goto err_udplite_mib;
 766         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 767         if (!net->mib.ipv6_statistics)
 768                 goto err_ip_mib;
 769
 770         for_each_possible_cpu(i) {
 771                 struct ipstats_mib *af_inet6_stats;
 772                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 773                 u64_stats_init(&af_inet6_stats->syncp);
 774         }
 775
 776
 777         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 778         if (!net->mib.icmpv6_statistics)
 779                 goto err_icmp_mib;
 780         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 781                                                 GFP_KERNEL);
 782         if (!net->mib.icmpv6msg_statistics)
 783                 goto err_icmpmsg_mib;
 784         return 0;
 785
 786 err_icmpmsg_mib:
 787         free_percpu(net->mib.icmpv6_statistics);
 788 err_icmp_mib:
 789         free_percpu(net->mib.ipv6_statistics);
 790 err_ip_mib:
 791         free_percpu(net->mib.udplite_stats_in6);
 792 err_udplite_mib:
 793         free_percpu(net->mib.udp_stats_in6);
 794         return -ENOMEM;
 795 }
 796
 797 static void ipv6_cleanup_mibs(struct net *net)
 798 {
 799         free_percpu(net->mib.udp_stats_in6);
 800         free_percpu(net->mib.udplite_stats_in6);
 801         free_percpu(net->mib.ipv6_statistics);
 802         free_percpu(net->mib.icmpv6_statistics);
 803         kfree(net->mib.icmpv6msg_statistics);
 804 }
 805
 806 static int __net_init inet6_net_init(struct net *net)
 807 {
 808         int err = 0;
 809
 810         net->ipv6.sysctl.bindv6only = 0;
 811         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 812         net->ipv6.sysctl.flowlabel_consistency = 1;
 813         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
 814         net->ipv6.sysctl.idgen_retries = 3;
 815         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 816         net->ipv6.sysctl.flowlabel_state_ranges = 0;
 817         atomic_set(&net->ipv6.fib6_sernum, 1);
 818
 819         err = ipv6_init_mibs(net);
 820         if (err)
 821                 return err;
 822 #ifdef CONFIG_PROC_FS
 823         err = udp6_proc_init(net);
 824         if (err)
 825                 goto out;
 826         err = tcp6_proc_init(net);
 827         if (err)
 828                 goto proc_tcp6_fail;
 829         err = ac6_proc_init(net);
 830         if (err)
 831                 goto proc_ac6_fail;
 832 #endif
 833         return err;
 834
 835 #ifdef CONFIG_PROC_FS
 836 proc_ac6_fail:
 837         tcp6_proc_exit(net);
 838 proc_tcp6_fail:
 839         udp6_proc_exit(net);
 840 out:
 841         ipv6_cleanup_mibs(net);
 842         return err;
 843 #endif
 844 }
 845
 846 static void __net_exit inet6_net_exit(struct net *net)
 847 {
 848 #ifdef CONFIG_PROC_FS
 849         udp6_proc_exit(net);
 850         tcp6_proc_exit(net);
 851         ac6_proc_exit(net);
 852 #endif
 853         ipv6_cleanup_mibs(net);
 854 }
 855
 856 static struct pernet_operations inet6_net_ops = {
 857         .init = inet6_net_init,
 858         .exit = inet6_net_exit,
 859 };
 860
 861 static const struct ipv6_stub ipv6_stub_impl = {
 862         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 863         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 864         .ipv6_dst_lookup = ip6_dst_lookup,
 865         .udpv6_encap_enable = udpv6_encap_enable,
 866         .ndisc_send_na = ndisc_send_na,
 867         .nd_tbl = &nd_tbl,
 868 };
 869
 870 static int __init inet6_init(void)
 871 {
 872         struct list_head *r;
 873         int err = 0;
 874
 875         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 876
 877         /* Register the socket-side information for inet6_create.  */
 878         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 879                 INIT_LIST_HEAD(r);
 880
 881         if (disable_ipv6_mod) {
 882                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 883                 goto out;
 884         }
 885
 886         err = proto_register(&tcpv6_prot, 1);
 887         if (err)
 888                 goto out;
 889
 890         err = proto_register(&udpv6_prot, 1);
 891         if (err)
 892                 goto out_unregister_tcp_proto;
 893
 894         err = proto_register(&udplitev6_prot, 1);
 895         if (err)
 896                 goto out_unregister_udp_proto;
 897
 898         err = proto_register(&rawv6_prot, 1);
 899         if (err)
 900                 goto out_unregister_udplite_proto;
 901
 902         err = proto_register(&pingv6_prot, 1);
 903         if (err)
 904                 goto out_unregister_ping_proto;
 905
 906         /* We MUST register RAW sockets before we create the ICMP6,
 907          * IGMP6, or NDISC control sockets.
 908          */
 909         err = rawv6_init();
 910         if (err)
 911                 goto out_unregister_raw_proto;
 912
 913         /* Register the family here so that the init calls below will
 914          * be able to create sockets. (?? is this dangerous ??)
 915          */
 916         err = sock_register(&inet6_family_ops);
 917         if (err)
 918                 goto out_sock_register_fail;
 919
 920         /*
 921          *      ipngwg API draft makes clear that the correct semantics
 922          *      for TCP and UDP is to consider one TCP and UDP instance
 923          *      in a host available by both INET and INET6 APIs and
 924          *      able to communicate via both network protocols.
 925          */
 926
 927         err = register_pernet_subsys(&inet6_net_ops);
 928         if (err)
 929                 goto register_pernet_fail;
 930         err = icmpv6_init();
 931         if (err)
 932                 goto icmp_fail;
 933         err = ip6_mr_init();
 934         if (err)
 935                 goto ipmr_fail;
 936         err = ndisc_init();
 937         if (err)
 938                 goto ndisc_fail;
 939         err = igmp6_init();
 940         if (err)
 941                 goto igmp_fail;
 942
 943         ipv6_stub = &ipv6_stub_impl;
 944
 945         err = ipv6_netfilter_init();
 946         if (err)
 947                 goto netfilter_fail;
 948         /* Create /proc/foo6 entries. */
 949 #ifdef CONFIG_PROC_FS
 950         err = -ENOMEM;
 951         if (raw6_proc_init())
 952                 goto proc_raw6_fail;
 953         if (udplite6_proc_init())
 954                 goto proc_udplite6_fail;
 955         if (ipv6_misc_proc_init())
 956                 goto proc_misc6_fail;
 957         if (if6_proc_init())
 958                 goto proc_if6_fail;
 959 #endif
 960         err = ip6_route_init();
 961         if (err)
 962                 goto ip6_route_fail;
 963         err = ndisc_late_init();
 964         if (err)
 965                 goto ndisc_late_fail;
 966         err = ip6_flowlabel_init();
 967         if (err)
 968                 goto ip6_flowlabel_fail;
 969         err = addrconf_init();
 970         if (err)
 971                 goto addrconf_fail;
 972
 973         /* Init v6 extension headers. */
 974         err = ipv6_exthdrs_init();
 975         if (err)
 976                 goto ipv6_exthdrs_fail;
 977
 978         err = ipv6_frag_init();
 979         if (err)
 980                 goto ipv6_frag_fail;
 981
 982         /* Init v6 transport protocols. */
 983         err = udpv6_init();
 984         if (err)
 985                 goto udpv6_fail;
 986
 987         err = udplitev6_init();
 988         if (err)
 989                 goto udplitev6_fail;
 990
 991         err = tcpv6_init();
 992         if (err)
 993                 goto tcpv6_fail;
 994
 995         err = ipv6_packet_init();
 996         if (err)
 997                 goto ipv6_packet_fail;
 998
 999         err = pingv6_init();
1000         if (err)
1001                 goto pingv6_fail;
1002
1003 #ifdef CONFIG_SYSCTL
1004         err = ipv6_sysctl_register();
1005         if (err)
1006                 goto sysctl_fail;
1007 #endif
1008 out:
1009         return err;
1010
1011 #ifdef CONFIG_SYSCTL
1012 sysctl_fail:
1013         pingv6_exit();
1014 #endif
1015 pingv6_fail:
1016         ipv6_packet_cleanup();
1017 ipv6_packet_fail:
1018         tcpv6_exit();
1019 tcpv6_fail:
1020         udplitev6_exit();
1021 udplitev6_fail:
1022         udpv6_exit();
1023 udpv6_fail:
1024         ipv6_frag_exit();
1025 ipv6_frag_fail:
1026         ipv6_exthdrs_exit();
1027 ipv6_exthdrs_fail:
1028         addrconf_cleanup();
1029 addrconf_fail:
1030         ip6_flowlabel_cleanup();
1031 ip6_flowlabel_fail:
1032         ndisc_late_cleanup();
1033 ndisc_late_fail:
1034         ip6_route_cleanup();
1035 ip6_route_fail:
1036 #ifdef CONFIG_PROC_FS
1037         if6_proc_exit();
1038 proc_if6_fail:
1039         ipv6_misc_proc_exit();
1040 proc_misc6_fail:
1041         udplite6_proc_exit();
1042 proc_udplite6_fail:
1043         raw6_proc_exit();
1044 proc_raw6_fail:
1045 #endif
1046         ipv6_netfilter_fini();
1047 netfilter_fail:
1048         igmp6_cleanup();
1049 igmp_fail:
1050         ndisc_cleanup();
1051 ndisc_fail:
1052         ip6_mr_cleanup();
1053 ipmr_fail:
1054         icmpv6_cleanup();
1055 icmp_fail:
1056         unregister_pernet_subsys(&inet6_net_ops);
1057 register_pernet_fail:
1058         sock_unregister(PF_INET6);
1059         rtnl_unregister_all(PF_INET6);
1060 out_sock_register_fail:
1061         rawv6_exit();
1062 out_unregister_ping_proto:
1063         proto_unregister(&pingv6_prot);
1064 out_unregister_raw_proto:
1065         proto_unregister(&rawv6_prot);
1066 out_unregister_udplite_proto:
1067         proto_unregister(&udplitev6_prot);
1068 out_unregister_udp_proto:
1069         proto_unregister(&udpv6_prot);
1070 out_unregister_tcp_proto:
1071         proto_unregister(&tcpv6_prot);
1072         goto out;
1073 }
1074 module_init(inet6_init);
1075
1076 MODULE_ALIAS_NETPROTO(PF_INET6);