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 #include <net/calipso.h>
  64
  65 #include <asm/uaccess.h>
  66 #include <linux/mroute6.h>
  67
  68 #include "ip6_offload.h"
  69
  70 MODULE_AUTHOR("Cast of dozens");
  71 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  72 MODULE_LICENSE("GPL");
  73
  74 /* The inetsw6 table contains everything that inet6_create needs to
  75  * build a new socket.
  76  */
  77 static struct list_head inetsw6[SOCK_MAX];
  78 static DEFINE_SPINLOCK(inetsw6_lock);
  79
  80 struct ipv6_params ipv6_defaults = {
  81         .disable_ipv6 = 0,
  82         .autoconf = 1,
  83 };
  84
  85 static int disable_ipv6_mod;
  86
  87 module_param_named(disable, disable_ipv6_mod, int, 0444);
  88 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  89
  90 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  91 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  92
  93 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  94 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  95
  96 bool ipv6_mod_enabled(void)
  97 {
  98         return disable_ipv6_mod == 0;
  99 }
 100 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
 101
 102 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
 103 {
 104         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
 105
 106         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
 107 }
 108
 109 static int inet6_create(struct net *net, struct socket *sock, int protocol,
 110                         int kern)
 111 {
 112         struct inet_sock *inet;
 113         struct ipv6_pinfo *np;
 114         struct sock *sk;
 115         struct inet_protosw *answer;
 116         struct proto *answer_prot;
 117         unsigned char answer_flags;
 118         int try_loading_module = 0;
 119         int err;
 120
 121         if (protocol < 0 || protocol >= IPPROTO_MAX)
 122                 return -EINVAL;
 123
 124         /* Look for the requested type/protocol pair. */
 125 lookup_protocol:
 126         err = -ESOCKTNOSUPPORT;
 127         rcu_read_lock();
 128         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 129
 130                 err = 0;
 131                 /* Check the non-wild match. */
 132                 if (protocol == answer->protocol) {
 133                         if (protocol != IPPROTO_IP)
 134                                 break;
 135                 } else {
 136                         /* Check for the two wild cases. */
 137                         if (IPPROTO_IP == protocol) {
 138                                 protocol = answer->protocol;
 139                                 break;
 140                         }
 141                         if (IPPROTO_IP == answer->protocol)
 142                                 break;
 143                 }
 144                 err = -EPROTONOSUPPORT;
 145         }
 146
 147         if (err) {
 148                 if (try_loading_module < 2) {
 149                         rcu_read_unlock();
 150                         /*
 151                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 152                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 153                          */
 154                         if (++try_loading_module == 1)
 155                                 request_module("net-pf-%d-proto-%d-type-%d",
 156                                                 PF_INET6, protocol, sock->type);
 157                         /*
 158                          * Fall back to generic, e.g. net-pf-10-proto-132
 159                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 160                          */
 161                         else
 162                                 request_module("net-pf-%d-proto-%d",
 163                                                 PF_INET6, protocol);
 164                         goto lookup_protocol;
 165                 } else
 166                         goto out_rcu_unlock;
 167         }
 168
 169         err = -EPERM;
 170         if (sock->type == SOCK_RAW && !kern &&
 171             !ns_capable(net->user_ns, CAP_NET_RAW))
 172                 goto out_rcu_unlock;
 173
 174         sock->ops = answer->ops;
 175         answer_prot = answer->prot;
 176         answer_flags = answer->flags;
 177         rcu_read_unlock();
 178
 179         WARN_ON(!answer_prot->slab);
 180
 181         err = -ENOBUFS;
 182         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
 183         if (!sk)
 184                 goto out;
 185
 186         sock_init_data(sock, sk);
 187
 188         err = 0;
 189         if (INET_PROTOSW_REUSE & answer_flags)
 190                 sk->sk_reuse = SK_CAN_REUSE;
 191
 192         inet = inet_sk(sk);
 193         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 194
 195         if (SOCK_RAW == sock->type) {
 196                 inet->inet_num = protocol;
 197                 if (IPPROTO_RAW == protocol)
 198                         inet->hdrincl = 1;
 199         }
 200
 201         sk->sk_destruct         = inet_sock_destruct;
 202         sk->sk_family           = PF_INET6;
 203         sk->sk_protocol         = protocol;
 204
 205         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 206
 207         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 208         np->hop_limit   = -1;
 209         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 210         np->mc_loop     = 1;
 211         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 212         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
 213
 214         /* Init the ipv4 part of the socket since we can have sockets
 215          * using v6 API for ipv4.
 216          */
 217         inet->uc_ttl    = -1;
 218
 219         inet->mc_loop   = 1;
 220         inet->mc_ttl    = 1;
 221         inet->mc_index  = 0;
 222         inet->mc_list   = NULL;
 223         inet->rcv_tos   = 0;
 224
 225         if (net->ipv4.sysctl_ip_no_pmtu_disc)
 226                 inet->pmtudisc = IP_PMTUDISC_DONT;
 227         else
 228                 inet->pmtudisc = IP_PMTUDISC_WANT;
 229         /*
 230          * Increment only the relevant sk_prot->socks debug field, this changes
 231          * the previous behaviour of incrementing both the equivalent to
 232          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 233          *
 234          * This allows better debug granularity as we'll know exactly how many
 235          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 236          * transport protocol socks. -acme
 237          */
 238         sk_refcnt_debug_inc(sk);
 239
 240         if (inet->inet_num) {
 241                 /* It assumes that any protocol which allows
 242                  * the user to assign a number at socket
 243                  * creation time automatically shares.
 244                  */
 245                 inet->inet_sport = htons(inet->inet_num);
 246                 err = sk->sk_prot->hash(sk);
 247                 if (err) {
 248                         sk_common_release(sk);
 249                         goto out;
 250                 }
 251         }
 252         if (sk->sk_prot->init) {
 253                 err = sk->sk_prot->init(sk);
 254                 if (err) {
 255                         sk_common_release(sk);
 256                         goto out;
 257                 }
 258         }
 259 out:
 260         return err;
 261 out_rcu_unlock:
 262         rcu_read_unlock();
 263         goto out;
 264 }
 265
 266
 267 /* bind for INET6 API */
 268 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 269 {
 270         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 271         struct sock *sk = sock->sk;
 272         struct inet_sock *inet = inet_sk(sk);
 273         struct ipv6_pinfo *np = inet6_sk(sk);
 274         struct net *net = sock_net(sk);
 275         __be32 v4addr = 0;
 276         unsigned short snum;
 277         bool saved_ipv6only;
 278         int addr_type = 0;
 279         int err = 0;
 280
 281         /* If the socket has its own bind function then use it. */
 282         if (sk->sk_prot->bind)
 283                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 284
 285         if (addr_len < SIN6_LEN_RFC2133)
 286                 return -EINVAL;
 287
 288         if (addr->sin6_family != AF_INET6)
 289                 return -EAFNOSUPPORT;
 290
 291         addr_type = ipv6_addr_type(&addr->sin6_addr);
 292         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 293                 return -EINVAL;
 294
 295         snum = ntohs(addr->sin6_port);
 296         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 297                 return -EACCES;
 298
 299         lock_sock(sk);
 300
 301         /* Check these errors (active socket, double bind). */
 302         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 303                 err = -EINVAL;
 304                 goto out;
 305         }
 306
 307         /* Check if the address belongs to the host. */
 308         if (addr_type == IPV6_ADDR_MAPPED) {
 309                 struct net_device *dev = NULL;
 310                 int chk_addr_ret;
 311
 312                 /* Binding to v4-mapped address on a v6-only socket
 313                  * makes no sense
 314                  */
 315                 if (sk->sk_ipv6only) {
 316                         err = -EINVAL;
 317                         goto out;
 318                 }
 319
 320                 rcu_read_lock();
 321                 if (sk->sk_bound_dev_if) {
 322                         dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 323                         if (!dev) {
 324                                 err = -ENODEV;
 325                                 goto out_unlock;
 326                         }
 327                 }
 328
 329                 /* Reproduce AF_INET checks to make the bindings consistent */
 330                 v4addr = addr->sin6_addr.s6_addr32[3];
 331                 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
 332                 rcu_read_unlock();
 333
 334                 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
 335                     !(inet->freebind || inet->transparent) &&
 336                     v4addr != htonl(INADDR_ANY) &&
 337                     chk_addr_ret != RTN_LOCAL &&
 338                     chk_addr_ret != RTN_MULTICAST &&
 339                     chk_addr_ret != RTN_BROADCAST) {
 340                         err = -EADDRNOTAVAIL;
 341                         goto out;
 342                 }
 343         } else {
 344                 if (addr_type != IPV6_ADDR_ANY) {
 345                         struct net_device *dev = NULL;
 346
 347                         rcu_read_lock();
 348                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 349                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 350                                     addr->sin6_scope_id) {
 351                                         /* Override any existing binding, if another one
 352                                          * is supplied by user.
 353                                          */
 354                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 355                                 }
 356
 357                                 /* Binding to link-local address requires an interface */
 358                                 if (!sk->sk_bound_dev_if) {
 359                                         err = -EINVAL;
 360                                         goto out_unlock;
 361                                 }
 362                         }
 363
 364                         if (sk->sk_bound_dev_if) {
 365                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 366                                 if (!dev) {
 367                                         err = -ENODEV;
 368                                         goto out_unlock;
 369                                 }
 370                         }
 371
 372                         /* ipv4 addr of the socket is invalid.  Only the
 373                          * unspecified and mapped address have a v4 equivalent.
 374                          */
 375                         v4addr = LOOPBACK4_IPV6;
 376                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 377                                 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
 378                                     !(inet->freebind || inet->transparent) &&
 379                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 380                                                    dev, 0)) {
 381                                         err = -EADDRNOTAVAIL;
 382                                         goto out_unlock;
 383                                 }
 384                         }
 385                         rcu_read_unlock();
 386                 }
 387         }
 388
 389         inet->inet_rcv_saddr = v4addr;
 390         inet->inet_saddr = v4addr;
 391
 392         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 393
 394         if (!(addr_type & IPV6_ADDR_MULTICAST))
 395                 np->saddr = addr->sin6_addr;
 396
 397         saved_ipv6only = sk->sk_ipv6only;
 398         if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
 399                 sk->sk_ipv6only = 1;
 400
 401         /* Make sure we are allowed to bind here. */
 402         if ((snum || !inet->bind_address_no_port) &&
 403             sk->sk_prot->get_port(sk, snum)) {
 404                 sk->sk_ipv6only = saved_ipv6only;
 405                 inet_reset_saddr(sk);
 406                 err = -EADDRINUSE;
 407                 goto out;
 408         }
 409
 410         if (addr_type != IPV6_ADDR_ANY)
 411                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 412         if (snum)
 413                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 414         inet->inet_sport = htons(inet->inet_num);
 415         inet->inet_dport = 0;
 416         inet->inet_daddr = 0;
 417 out:
 418         release_sock(sk);
 419         return err;
 420 out_unlock:
 421         rcu_read_unlock();
 422         goto out;
 423 }
 424 EXPORT_SYMBOL(inet6_bind);
 425
 426 int inet6_release(struct socket *sock)
 427 {
 428         struct sock *sk = sock->sk;
 429
 430         if (!sk)
 431                 return -EINVAL;
 432
 433         /* Free mc lists */
 434         ipv6_sock_mc_close(sk);
 435
 436         /* Free ac lists */
 437         ipv6_sock_ac_close(sk);
 438
 439         return inet_release(sock);
 440 }
 441 EXPORT_SYMBOL(inet6_release);
 442
 443 void inet6_destroy_sock(struct sock *sk)
 444 {
 445         struct ipv6_pinfo *np = inet6_sk(sk);
 446         struct sk_buff *skb;
 447         struct ipv6_txoptions *opt;
 448
 449         /* Release rx options */
 450
 451         skb = xchg(&np->pktoptions, NULL);
 452         if (skb)
 453                 kfree_skb(skb);
 454
 455         skb = xchg(&np->rxpmtu, NULL);
 456         if (skb)
 457                 kfree_skb(skb);
 458
 459         /* Free flowlabels */
 460         fl6_free_socklist(sk);
 461
 462         /* Free tx options */
 463
 464         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 465         if (opt) {
 466                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 467                 txopt_put(opt);
 468         }
 469 }
 470 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 471
 472 /*
 473  *      This does both peername and sockname.
 474  */
 475
 476 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 477                  int *uaddr_len, int peer)
 478 {
 479         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 480         struct sock *sk = sock->sk;
 481         struct inet_sock *inet = inet_sk(sk);
 482         struct ipv6_pinfo *np = inet6_sk(sk);
 483
 484         sin->sin6_family = AF_INET6;
 485         sin->sin6_flowinfo = 0;
 486         sin->sin6_scope_id = 0;
 487         if (peer) {
 488                 if (!inet->inet_dport)
 489                         return -ENOTCONN;
 490                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 491                     peer == 1)
 492                         return -ENOTCONN;
 493                 sin->sin6_port = inet->inet_dport;
 494                 sin->sin6_addr = sk->sk_v6_daddr;
 495                 if (np->sndflow)
 496                         sin->sin6_flowinfo = np->flow_label;
 497         } else {
 498                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 499                         sin->sin6_addr = np->saddr;
 500                 else
 501                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 502
 503                 sin->sin6_port = inet->inet_sport;
 504         }
 505         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 506                                                  sk->sk_bound_dev_if);
 507         *uaddr_len = sizeof(*sin);
 508         return 0;
 509 }
 510 EXPORT_SYMBOL(inet6_getname);
 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         default:
 536                 if (!sk->sk_prot->ioctl)
 537                         return -ENOIOCTLCMD;
 538                 return sk->sk_prot->ioctl(sk, cmd, arg);
 539         }
 540         /*NOTREACHED*/
 541         return 0;
 542 }
 543 EXPORT_SYMBOL(inet6_ioctl);
 544
 545 const struct proto_ops inet6_stream_ops = {
 546         .family            = PF_INET6,
 547         .owner             = THIS_MODULE,
 548         .release           = inet6_release,
 549         .bind              = inet6_bind,
 550         .connect           = inet_stream_connect,       /* ok           */
 551         .socketpair        = sock_no_socketpair,        /* a do nothing */
 552         .accept            = inet_accept,               /* ok           */
 553         .getname           = inet6_getname,
 554         .poll              = tcp_poll,                  /* ok           */
 555         .ioctl             = inet6_ioctl,               /* must change  */
 556         .listen            = inet_listen,               /* ok           */
 557         .shutdown          = inet_shutdown,             /* ok           */
 558         .setsockopt        = sock_common_setsockopt,    /* ok           */
 559         .getsockopt        = sock_common_getsockopt,    /* ok           */
 560         .sendmsg           = inet_sendmsg,              /* ok           */
 561         .recvmsg           = inet_recvmsg,              /* ok           */
 562         .mmap              = sock_no_mmap,
 563         .sendpage          = inet_sendpage,
 564         .splice_read       = tcp_splice_read,
 565         .read_sock         = tcp_read_sock,
 566         .peek_len          = tcp_peek_len,
 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         .set_peek_off      = sk_set_peek_off,
 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                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 699
 700                 rcu_read_lock();
 701                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 702                                          &final);
 703                 rcu_read_unlock();
 704
 705                 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
 706                 if (IS_ERR(dst)) {
 707                         sk->sk_route_caps = 0;
 708                         sk->sk_err_soft = -PTR_ERR(dst);
 709                         return PTR_ERR(dst);
 710                 }
 711
 712                 ip6_dst_store(sk, dst, NULL, NULL);
 713         }
 714
 715         return 0;
 716 }
 717 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 718
 719 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 720                        const struct inet6_skb_parm *opt)
 721 {
 722         const struct ipv6_pinfo *np = inet6_sk(sk);
 723
 724         if (np->rxopt.all) {
 725                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
 726                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 727                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 728                      np->rxopt.bits.rxflow) ||
 729                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 730                      np->rxopt.bits.osrcrt)) ||
 731                     ((opt->dst1 || opt->dst0) &&
 732                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 733                         return true;
 734         }
 735         return false;
 736 }
 737 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 738
 739 static struct packet_type ipv6_packet_type __read_mostly = {
 740         .type = cpu_to_be16(ETH_P_IPV6),
 741         .func = ipv6_rcv,
 742 };
 743
 744 static int __init ipv6_packet_init(void)
 745 {
 746         dev_add_pack(&ipv6_packet_type);
 747         return 0;
 748 }
 749
 750 static void ipv6_packet_cleanup(void)
 751 {
 752         dev_remove_pack(&ipv6_packet_type);
 753 }
 754
 755 static int __net_init ipv6_init_mibs(struct net *net)
 756 {
 757         int i;
 758
 759         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 760         if (!net->mib.udp_stats_in6)
 761                 return -ENOMEM;
 762         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 763         if (!net->mib.udplite_stats_in6)
 764                 goto err_udplite_mib;
 765         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 766         if (!net->mib.ipv6_statistics)
 767                 goto err_ip_mib;
 768
 769         for_each_possible_cpu(i) {
 770                 struct ipstats_mib *af_inet6_stats;
 771                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 772                 u64_stats_init(&af_inet6_stats->syncp);
 773         }
 774
 775
 776         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 777         if (!net->mib.icmpv6_statistics)
 778                 goto err_icmp_mib;
 779         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 780                                                 GFP_KERNEL);
 781         if (!net->mib.icmpv6msg_statistics)
 782                 goto err_icmpmsg_mib;
 783         return 0;
 784
 785 err_icmpmsg_mib:
 786         free_percpu(net->mib.icmpv6_statistics);
 787 err_icmp_mib:
 788         free_percpu(net->mib.ipv6_statistics);
 789 err_ip_mib:
 790         free_percpu(net->mib.udplite_stats_in6);
 791 err_udplite_mib:
 792         free_percpu(net->mib.udp_stats_in6);
 793         return -ENOMEM;
 794 }
 795
 796 static void ipv6_cleanup_mibs(struct net *net)
 797 {
 798         free_percpu(net->mib.udp_stats_in6);
 799         free_percpu(net->mib.udplite_stats_in6);
 800         free_percpu(net->mib.ipv6_statistics);
 801         free_percpu(net->mib.icmpv6_statistics);
 802         kfree(net->mib.icmpv6msg_statistics);
 803 }
 804
 805 static int __net_init inet6_net_init(struct net *net)
 806 {
 807         int err = 0;
 808
 809         net->ipv6.sysctl.bindv6only = 0;
 810         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 811         net->ipv6.sysctl.flowlabel_consistency = 1;
 812         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
 813         net->ipv6.sysctl.idgen_retries = 3;
 814         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 815         net->ipv6.sysctl.flowlabel_state_ranges = 0;
 816         atomic_set(&net->ipv6.fib6_sernum, 1);
 817
 818         err = ipv6_init_mibs(net);
 819         if (err)
 820                 return err;
 821 #ifdef CONFIG_PROC_FS
 822         err = udp6_proc_init(net);
 823         if (err)
 824                 goto out;
 825         err = tcp6_proc_init(net);
 826         if (err)
 827                 goto proc_tcp6_fail;
 828         err = ac6_proc_init(net);
 829         if (err)
 830                 goto proc_ac6_fail;
 831 #endif
 832         return err;
 833
 834 #ifdef CONFIG_PROC_FS
 835 proc_ac6_fail:
 836         tcp6_proc_exit(net);
 837 proc_tcp6_fail:
 838         udp6_proc_exit(net);
 839 out:
 840         ipv6_cleanup_mibs(net);
 841         return err;
 842 #endif
 843 }
 844
 845 static void __net_exit inet6_net_exit(struct net *net)
 846 {
 847 #ifdef CONFIG_PROC_FS
 848         udp6_proc_exit(net);
 849         tcp6_proc_exit(net);
 850         ac6_proc_exit(net);
 851 #endif
 852         ipv6_cleanup_mibs(net);
 853 }
 854
 855 static struct pernet_operations inet6_net_ops = {
 856         .init = inet6_net_init,
 857         .exit = inet6_net_exit,
 858 };
 859
 860 static const struct ipv6_stub ipv6_stub_impl = {
 861         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 862         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 863         .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
 864         .udpv6_encap_enable = udpv6_encap_enable,
 865         .ndisc_send_na = ndisc_send_na,
 866         .nd_tbl = &nd_tbl,
 867 };
 868
 869 static int __init inet6_init(void)
 870 {
 871         struct list_head *r;
 872         int err = 0;
 873
 874         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 875
 876         /* Register the socket-side information for inet6_create.  */
 877         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 878                 INIT_LIST_HEAD(r);
 879
 880         if (disable_ipv6_mod) {
 881                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 882                 goto out;
 883         }
 884
 885         err = proto_register(&tcpv6_prot, 1);
 886         if (err)
 887                 goto out;
 888
 889         err = proto_register(&udpv6_prot, 1);
 890         if (err)
 891                 goto out_unregister_tcp_proto;
 892
 893         err = proto_register(&udplitev6_prot, 1);
 894         if (err)
 895                 goto out_unregister_udp_proto;
 896
 897         err = proto_register(&rawv6_prot, 1);
 898         if (err)
 899                 goto out_unregister_udplite_proto;
 900
 901         err = proto_register(&pingv6_prot, 1);
 902         if (err)
 903                 goto out_unregister_ping_proto;
 904
 905         /* We MUST register RAW sockets before we create the ICMP6,
 906          * IGMP6, or NDISC control sockets.
 907          */
 908         err = rawv6_init();
 909         if (err)
 910                 goto out_unregister_raw_proto;
 911
 912         /* Register the family here so that the init calls below will
 913          * be able to create sockets. (?? is this dangerous ??)
 914          */
 915         err = sock_register(&inet6_family_ops);
 916         if (err)
 917                 goto out_sock_register_fail;
 918
 919         /*
 920          *      ipngwg API draft makes clear that the correct semantics
 921          *      for TCP and UDP is to consider one TCP and UDP instance
 922          *      in a host available by both INET and INET6 APIs and
 923          *      able to communicate via both network protocols.
 924          */
 925
 926         err = register_pernet_subsys(&inet6_net_ops);
 927         if (err)
 928                 goto register_pernet_fail;
 929         err = ip6_mr_init();
 930         if (err)
 931                 goto ipmr_fail;
 932         err = icmpv6_init();
 933         if (err)
 934                 goto icmp_fail;
 935         err = ndisc_init();
 936         if (err)
 937                 goto ndisc_fail;
 938         err = igmp6_init();
 939         if (err)
 940                 goto igmp_fail;
 941
 942         ipv6_stub = &ipv6_stub_impl;
 943
 944         err = ipv6_netfilter_init();
 945         if (err)
 946                 goto netfilter_fail;
 947         /* Create /proc/foo6 entries. */
 948 #ifdef CONFIG_PROC_FS
 949         err = -ENOMEM;
 950         if (raw6_proc_init())
 951                 goto proc_raw6_fail;
 952         if (udplite6_proc_init())
 953                 goto proc_udplite6_fail;
 954         if (ipv6_misc_proc_init())
 955                 goto proc_misc6_fail;
 956         if (if6_proc_init())
 957                 goto proc_if6_fail;
 958 #endif
 959         err = ip6_route_init();
 960         if (err)
 961                 goto ip6_route_fail;
 962         err = ndisc_late_init();
 963         if (err)
 964                 goto ndisc_late_fail;
 965         err = ip6_flowlabel_init();
 966         if (err)
 967                 goto ip6_flowlabel_fail;
 968         err = addrconf_init();
 969         if (err)
 970                 goto addrconf_fail;
 971
 972         /* Init v6 extension headers. */
 973         err = ipv6_exthdrs_init();
 974         if (err)
 975                 goto ipv6_exthdrs_fail;
 976
 977         err = ipv6_frag_init();
 978         if (err)
 979                 goto ipv6_frag_fail;
 980
 981         /* Init v6 transport protocols. */
 982         err = udpv6_init();
 983         if (err)
 984                 goto udpv6_fail;
 985
 986         err = udplitev6_init();
 987         if (err)
 988                 goto udplitev6_fail;
 989
 990         err = udpv6_offload_init();
 991         if (err)
 992                 goto udpv6_offload_fail;
 993
 994         err = tcpv6_init();
 995         if (err)
 996                 goto tcpv6_fail;
 997
 998         err = ipv6_packet_init();
 999         if (err)
1000                 goto ipv6_packet_fail;
1001
1002         err = pingv6_init();
1003         if (err)
1004                 goto pingv6_fail;
1005
1006         err = calipso_init();
1007         if (err)
1008                 goto calipso_fail;
1009
1010 #ifdef CONFIG_SYSCTL
1011         err = ipv6_sysctl_register();
1012         if (err)
1013                 goto sysctl_fail;
1014 #endif
1015 out:
1016         return err;
1017
1018 #ifdef CONFIG_SYSCTL
1019 sysctl_fail:
1020         calipso_exit();
1021 #endif
1022 calipso_fail:
1023         pingv6_exit();
1024 pingv6_fail:
1025         ipv6_packet_cleanup();
1026 ipv6_packet_fail:
1027         tcpv6_exit();
1028 tcpv6_fail:
1029         udpv6_offload_exit();
1030 udpv6_offload_fail:
1031         udplitev6_exit();
1032 udplitev6_fail:
1033         udpv6_exit();
1034 udpv6_fail:
1035         ipv6_frag_exit();
1036 ipv6_frag_fail:
1037         ipv6_exthdrs_exit();
1038 ipv6_exthdrs_fail:
1039         addrconf_cleanup();
1040 addrconf_fail:
1041         ip6_flowlabel_cleanup();
1042 ip6_flowlabel_fail:
1043         ndisc_late_cleanup();
1044 ndisc_late_fail:
1045         ip6_route_cleanup();
1046 ip6_route_fail:
1047 #ifdef CONFIG_PROC_FS
1048         if6_proc_exit();
1049 proc_if6_fail:
1050         ipv6_misc_proc_exit();
1051 proc_misc6_fail:
1052         udplite6_proc_exit();
1053 proc_udplite6_fail:
1054         raw6_proc_exit();
1055 proc_raw6_fail:
1056 #endif
1057         ipv6_netfilter_fini();
1058 netfilter_fail:
1059         igmp6_cleanup();
1060 igmp_fail:
1061         ndisc_cleanup();
1062 ndisc_fail:
1063         icmpv6_cleanup();
1064 icmp_fail:
1065         ip6_mr_cleanup();
1066 ipmr_fail:
1067         unregister_pernet_subsys(&inet6_net_ops);
1068 register_pernet_fail:
1069         sock_unregister(PF_INET6);
1070         rtnl_unregister_all(PF_INET6);
1071 out_sock_register_fail:
1072         rawv6_exit();
1073 out_unregister_ping_proto:
1074         proto_unregister(&pingv6_prot);
1075 out_unregister_raw_proto:
1076         proto_unregister(&rawv6_prot);
1077 out_unregister_udplite_proto:
1078         proto_unregister(&udplitev6_prot);
1079 out_unregister_udp_proto:
1080         proto_unregister(&udpv6_prot);
1081 out_unregister_tcp_proto:
1082         proto_unregister(&tcpv6_prot);
1083         goto out;
1084 }
1085 module_init(inet6_init);
1086
1087 MODULE_ALIAS_NETPROTO(PF_INET6);