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/protocol.h>
  53 #include <net/inet_common.h>
  54 #include <net/route.h>
  55 #include <net/transp_v6.h>
  56 #include <net/ip6_route.h>
  57 #include <net/addrconf.h>
  58 #ifdef CONFIG_IPV6_TUNNEL
  59 #include <net/ip6_tunnel.h>
  60 #endif
  61
  62 #include <asm/uaccess.h>
  63 #include <linux/mroute6.h>
  64
  65 #ifdef CONFIG_ANDROID_PARANOID_NETWORK
  66 #include <linux/android_aid.h>
  67
  68 static inline int current_has_network(void)
  69 {
  70         return in_egroup_p(AID_INET) || capable(CAP_NET_RAW);
  71 }
  72 #else
  73 static inline int current_has_network(void)
  74 {
  75         return 1;
  76 }
  77 #endif
  78
  79 MODULE_AUTHOR("Cast of dozens");
  80 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  81 MODULE_LICENSE("GPL");
  82
  83 /* The inetsw6 table contains everything that inet6_create needs to
  84  * build a new socket.
  85  */
  86 static struct list_head inetsw6[SOCK_MAX];
  87 static DEFINE_SPINLOCK(inetsw6_lock);
  88
  89 struct ipv6_params ipv6_defaults = {
  90         .disable_ipv6 = 0,
  91         .autoconf = 1,
  92 };
  93
  94 static int disable_ipv6_mod;
  95
  96 module_param_named(disable, disable_ipv6_mod, int, 0444);
  97 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  98
  99 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
 100 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
 101
 102 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
 103 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
 104
 105 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
 106 {
 107         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
 108
 109         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
 110 }
 111
 112 static int inet6_create(struct net *net, struct socket *sock, int protocol,
 113                         int kern)
 114 {
 115         struct inet_sock *inet;
 116         struct ipv6_pinfo *np;
 117         struct sock *sk;
 118         struct inet_protosw *answer;
 119         struct proto *answer_prot;
 120         unsigned char answer_flags;
 121         char answer_no_check;
 122         int try_loading_module = 0;
 123         int err;
 124
 125         if (!current_has_network())
 126                 return -EACCES;
 127
 128         if (sock->type != SOCK_RAW &&
 129             sock->type != SOCK_DGRAM &&
 130             !inet_ehash_secret)
 131                 build_ehash_secret();
 132
 133         /* Look for the requested type/protocol pair. */
 134 lookup_protocol:
 135         err = -ESOCKTNOSUPPORT;
 136         rcu_read_lock();
 137         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 138
 139                 err = 0;
 140                 /* Check the non-wild match. */
 141                 if (protocol == answer->protocol) {
 142                         if (protocol != IPPROTO_IP)
 143                                 break;
 144                 } else {
 145                         /* Check for the two wild cases. */
 146                         if (IPPROTO_IP == protocol) {
 147                                 protocol = answer->protocol;
 148                                 break;
 149                         }
 150                         if (IPPROTO_IP == answer->protocol)
 151                                 break;
 152                 }
 153                 err = -EPROTONOSUPPORT;
 154         }
 155
 156         if (err) {
 157                 if (try_loading_module < 2) {
 158                         rcu_read_unlock();
 159                         /*
 160                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 161                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 162                          */
 163                         if (++try_loading_module == 1)
 164                                 request_module("net-pf-%d-proto-%d-type-%d",
 165                                                 PF_INET6, protocol, sock->type);
 166                         /*
 167                          * Fall back to generic, e.g. net-pf-10-proto-132
 168                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 169                          */
 170                         else
 171                                 request_module("net-pf-%d-proto-%d",
 172                                                 PF_INET6, protocol);
 173                         goto lookup_protocol;
 174                 } else
 175                         goto out_rcu_unlock;
 176         }
 177
 178         err = -EPERM;
 179         if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
 180                 goto out_rcu_unlock;
 181
 182         sock->ops = answer->ops;
 183         answer_prot = answer->prot;
 184         answer_no_check = answer->no_check;
 185         answer_flags = answer->flags;
 186         rcu_read_unlock();
 187
 188         WARN_ON(answer_prot->slab == NULL);
 189
 190         err = -ENOBUFS;
 191         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 192         if (sk == NULL)
 193                 goto out;
 194
 195         sock_init_data(sock, sk);
 196
 197         err = 0;
 198         sk->sk_no_check = answer_no_check;
 199         if (INET_PROTOSW_REUSE & answer_flags)
 200                 sk->sk_reuse = SK_CAN_REUSE;
 201
 202         inet = inet_sk(sk);
 203         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 204
 205         if (SOCK_RAW == sock->type) {
 206                 inet->inet_num = protocol;
 207                 if (IPPROTO_RAW == protocol)
 208                         inet->hdrincl = 1;
 209         }
 210
 211         sk->sk_destruct         = inet_sock_destruct;
 212         sk->sk_family           = PF_INET6;
 213         sk->sk_protocol         = protocol;
 214
 215         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 216
 217         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 218         np->hop_limit   = -1;
 219         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 220         np->mc_loop     = 1;
 221         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 222         np->ipv6only    = net->ipv6.sysctl.bindv6only;
 223
 224         /* Init the ipv4 part of the socket since we can have sockets
 225          * using v6 API for ipv4.
 226          */
 227         inet->uc_ttl    = -1;
 228
 229         inet->mc_loop   = 1;
 230         inet->mc_ttl    = 1;
 231         inet->mc_index  = 0;
 232         inet->mc_list   = NULL;
 233         inet->rcv_tos   = 0;
 234
 235         if (ipv4_config.no_pmtu_disc)
 236                 inet->pmtudisc = IP_PMTUDISC_DONT;
 237         else
 238                 inet->pmtudisc = IP_PMTUDISC_WANT;
 239         /*
 240          * Increment only the relevant sk_prot->socks debug field, this changes
 241          * the previous behaviour of incrementing both the equivalent to
 242          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 243          *
 244          * This allows better debug granularity as we'll know exactly how many
 245          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 246          * transport protocol socks. -acme
 247          */
 248         sk_refcnt_debug_inc(sk);
 249
 250         if (inet->inet_num) {
 251                 /* It assumes that any protocol which allows
 252                  * the user to assign a number at socket
 253                  * creation time automatically shares.
 254                  */
 255                 inet->inet_sport = htons(inet->inet_num);
 256                 sk->sk_prot->hash(sk);
 257         }
 258         if (sk->sk_prot->init) {
 259                 err = sk->sk_prot->init(sk);
 260                 if (err) {
 261                         sk_common_release(sk);
 262                         goto out;
 263                 }
 264         }
 265 out:
 266         return err;
 267 out_rcu_unlock:
 268         rcu_read_unlock();
 269         goto out;
 270 }
 271
 272
 273 /* bind for INET6 API */
 274 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 275 {
 276         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 277         struct sock *sk = sock->sk;
 278         struct inet_sock *inet = inet_sk(sk);
 279         struct ipv6_pinfo *np = inet6_sk(sk);
 280         struct net *net = sock_net(sk);
 281         __be32 v4addr = 0;
 282         unsigned short snum;
 283         int addr_type = 0;
 284         int err = 0;
 285
 286         /* If the socket has its own bind function then use it. */
 287         if (sk->sk_prot->bind)
 288                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 289
 290         if (addr_len < SIN6_LEN_RFC2133)
 291                 return -EINVAL;
 292
 293         if (addr->sin6_family != AF_INET6)
 294                 return -EAFNOSUPPORT;
 295
 296         addr_type = ipv6_addr_type(&addr->sin6_addr);
 297         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 298                 return -EINVAL;
 299
 300         snum = ntohs(addr->sin6_port);
 301         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 302                 return -EACCES;
 303
 304         lock_sock(sk);
 305
 306         /* Check these errors (active socket, double bind). */
 307         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 308                 err = -EINVAL;
 309                 goto out;
 310         }
 311
 312         /* Check if the address belongs to the host. */
 313         if (addr_type == IPV6_ADDR_MAPPED) {
 314                 int chk_addr_ret;
 315
 316                 /* Binding to v4-mapped address on a v6-only socket
 317                  * makes no sense
 318                  */
 319                 if (np->ipv6only) {
 320                         err = -EINVAL;
 321                         goto out;
 322                 }
 323
 324                 /* Reproduce AF_INET checks to make the bindings consistent */
 325                 v4addr = addr->sin6_addr.s6_addr32[3];
 326                 chk_addr_ret = inet_addr_type(net, v4addr);
 327                 if (!sysctl_ip_nonlocal_bind &&
 328                     !(inet->freebind || inet->transparent) &&
 329                     v4addr != htonl(INADDR_ANY) &&
 330                     chk_addr_ret != RTN_LOCAL &&
 331                     chk_addr_ret != RTN_MULTICAST &&
 332                     chk_addr_ret != RTN_BROADCAST) {
 333                         err = -EADDRNOTAVAIL;
 334                         goto out;
 335                 }
 336         } else {
 337                 if (addr_type != IPV6_ADDR_ANY) {
 338                         struct net_device *dev = NULL;
 339
 340                         rcu_read_lock();
 341                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 342                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 343                                     addr->sin6_scope_id) {
 344                                         /* Override any existing binding, if another one
 345                                          * is supplied by user.
 346                                          */
 347                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 348                                 }
 349
 350                                 /* Binding to link-local address requires an interface */
 351                                 if (!sk->sk_bound_dev_if) {
 352                                         err = -EINVAL;
 353                                         goto out_unlock;
 354                                 }
 355                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 356                                 if (!dev) {
 357                                         err = -ENODEV;
 358                                         goto out_unlock;
 359                                 }
 360                         }
 361
 362                         /* ipv4 addr of the socket is invalid.  Only the
 363                          * unspecified and mapped address have a v4 equivalent.
 364                          */
 365                         v4addr = LOOPBACK4_IPV6;
 366                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 367                                 if (!(inet->freebind || inet->transparent) &&
 368                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 369                                                    dev, 0)) {
 370                                         err = -EADDRNOTAVAIL;
 371                                         goto out_unlock;
 372                                 }
 373                         }
 374                         rcu_read_unlock();
 375                 }
 376         }
 377
 378         inet->inet_rcv_saddr = v4addr;
 379         inet->inet_saddr = v4addr;
 380
 381         np->rcv_saddr = addr->sin6_addr;
 382
 383         if (!(addr_type & IPV6_ADDR_MULTICAST))
 384                 np->saddr = addr->sin6_addr;
 385
 386         /* Make sure we are allowed to bind here. */
 387         if (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                         np->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 == NULL)
 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 != NULL)
 439                 kfree_skb(skb);
 440
 441         skb = xchg(&np->rxpmtu, NULL);
 442         if (skb != NULL)
 443                 kfree_skb(skb);
 444
 445         /* Free flowlabels */
 446         fl6_free_socklist(sk);
 447
 448         /* Free tx options */
 449
 450         opt = xchg(&np->opt, NULL);
 451         if (opt != NULL)
 452                 sock_kfree_s(sk, opt, opt->tot_len);
 453 }
 454 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 455
 456 /*
 457  *      This does both peername and sockname.
 458  */
 459
 460 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 461                  int *uaddr_len, int peer)
 462 {
 463         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 464         struct sock *sk = sock->sk;
 465         struct inet_sock *inet = inet_sk(sk);
 466         struct ipv6_pinfo *np = inet6_sk(sk);
 467
 468         sin->sin6_family = AF_INET6;
 469         sin->sin6_flowinfo = 0;
 470         sin->sin6_scope_id = 0;
 471         if (peer) {
 472                 if (!inet->inet_dport)
 473                         return -ENOTCONN;
 474                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 475                     peer == 1)
 476                         return -ENOTCONN;
 477                 sin->sin6_port = inet->inet_dport;
 478                 sin->sin6_addr = np->daddr;
 479                 if (np->sndflow)
 480                         sin->sin6_flowinfo = np->flow_label;
 481         } else {
 482                 if (ipv6_addr_any(&np->rcv_saddr))
 483                         sin->sin6_addr = np->saddr;
 484                 else
 485                         sin->sin6_addr = np->rcv_saddr;
 486
 487                 sin->sin6_port = inet->inet_sport;
 488         }
 489         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 490                                                  sk->sk_bound_dev_if);
 491         *uaddr_len = sizeof(*sin);
 492         return 0;
 493 }
 494 EXPORT_SYMBOL(inet6_getname);
 495
 496 int inet6_killaddr_ioctl(struct net *net, void __user *arg) {
 497         struct in6_ifreq ireq;
 498         struct sockaddr_in6 sin6;
 499
 500         if (!capable(CAP_NET_ADMIN))
 501                 return -EACCES;
 502
 503         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
 504                 return -EFAULT;
 505
 506         sin6.sin6_family = AF_INET6;
 507         sin6.sin6_addr = ireq.ifr6_addr;
 508         return tcp_nuke_addr(net, (struct sockaddr *) &sin6);
 509 }
 510
 511 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 512 {
 513         struct sock *sk = sock->sk;
 514         struct net *net = sock_net(sk);
 515
 516         switch (cmd) {
 517         case SIOCGSTAMP:
 518                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 519
 520         case SIOCGSTAMPNS:
 521                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 522
 523         case SIOCADDRT:
 524         case SIOCDELRT:
 525
 526                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 527
 528         case SIOCSIFADDR:
 529                 return addrconf_add_ifaddr(net, (void __user *) arg);
 530         case SIOCDIFADDR:
 531                 return addrconf_del_ifaddr(net, (void __user *) arg);
 532         case SIOCSIFDSTADDR:
 533                 return addrconf_set_dstaddr(net, (void __user *) arg);
 534         case SIOCKILLADDR:
 535                 return inet6_killaddr_ioctl(net, (void __user *) arg);
 536         default:
 537                 if (!sk->sk_prot->ioctl)
 538                         return -ENOIOCTLCMD;
 539                 return sk->sk_prot->ioctl(sk, cmd, arg);
 540         }
 541         /*NOTREACHED*/
 542         return 0;
 543 }
 544 EXPORT_SYMBOL(inet6_ioctl);
 545
 546 const struct proto_ops inet6_stream_ops = {
 547         .family            = PF_INET6,
 548         .owner             = THIS_MODULE,
 549         .release           = inet6_release,
 550         .bind              = inet6_bind,
 551         .connect           = inet_stream_connect,       /* ok           */
 552         .socketpair        = sock_no_socketpair,        /* a do nothing */
 553         .accept            = inet_accept,               /* ok           */
 554         .getname           = inet6_getname,
 555         .poll              = tcp_poll,                  /* ok           */
 556         .ioctl             = inet6_ioctl,               /* must change  */
 557         .listen            = inet_listen,               /* ok           */
 558         .shutdown          = inet_shutdown,             /* ok           */
 559         .setsockopt        = sock_common_setsockopt,    /* ok           */
 560         .getsockopt        = sock_common_getsockopt,    /* ok           */
 561         .sendmsg           = inet_sendmsg,              /* ok           */
 562         .recvmsg           = inet_recvmsg,              /* ok           */
 563         .mmap              = sock_no_mmap,
 564         .sendpage          = inet_sendpage,
 565         .splice_read       = tcp_splice_read,
 566 #ifdef CONFIG_COMPAT
 567         .compat_setsockopt = compat_sock_common_setsockopt,
 568         .compat_getsockopt = compat_sock_common_getsockopt,
 569 #endif
 570 };
 571
 572 const struct proto_ops inet6_dgram_ops = {
 573         .family            = PF_INET6,
 574         .owner             = THIS_MODULE,
 575         .release           = inet6_release,
 576         .bind              = inet6_bind,
 577         .connect           = inet_dgram_connect,        /* ok           */
 578         .socketpair        = sock_no_socketpair,        /* a do nothing */
 579         .accept            = sock_no_accept,            /* a do nothing */
 580         .getname           = inet6_getname,
 581         .poll              = udp_poll,                  /* ok           */
 582         .ioctl             = inet6_ioctl,               /* must change  */
 583         .listen            = sock_no_listen,            /* ok           */
 584         .shutdown          = inet_shutdown,             /* ok           */
 585         .setsockopt        = sock_common_setsockopt,    /* ok           */
 586         .getsockopt        = sock_common_getsockopt,    /* ok           */
 587         .sendmsg           = inet_sendmsg,              /* ok           */
 588         .recvmsg           = inet_recvmsg,              /* ok           */
 589         .mmap              = sock_no_mmap,
 590         .sendpage          = sock_no_sendpage,
 591 #ifdef CONFIG_COMPAT
 592         .compat_setsockopt = compat_sock_common_setsockopt,
 593         .compat_getsockopt = compat_sock_common_getsockopt,
 594 #endif
 595 };
 596
 597 static const struct net_proto_family inet6_family_ops = {
 598         .family = PF_INET6,
 599         .create = inet6_create,
 600         .owner  = THIS_MODULE,
 601 };
 602
 603 int inet6_register_protosw(struct inet_protosw *p)
 604 {
 605         struct list_head *lh;
 606         struct inet_protosw *answer;
 607         struct list_head *last_perm;
 608         int protocol = p->protocol;
 609         int ret;
 610
 611         spin_lock_bh(&inetsw6_lock);
 612
 613         ret = -EINVAL;
 614         if (p->type >= SOCK_MAX)
 615                 goto out_illegal;
 616
 617         /* If we are trying to override a permanent protocol, bail. */
 618         answer = NULL;
 619         ret = -EPERM;
 620         last_perm = &inetsw6[p->type];
 621         list_for_each(lh, &inetsw6[p->type]) {
 622                 answer = list_entry(lh, struct inet_protosw, list);
 623
 624                 /* Check only the non-wild match. */
 625                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 626                         if (protocol == answer->protocol)
 627                                 break;
 628                         last_perm = lh;
 629                 }
 630
 631                 answer = NULL;
 632         }
 633         if (answer)
 634                 goto out_permanent;
 635
 636         /* Add the new entry after the last permanent entry if any, so that
 637          * the new entry does not override a permanent entry when matched with
 638          * a wild-card protocol. But it is allowed to override any existing
 639          * non-permanent entry.  This means that when we remove this entry, the
 640          * system automatically returns to the old behavior.
 641          */
 642         list_add_rcu(&p->list, last_perm);
 643         ret = 0;
 644 out:
 645         spin_unlock_bh(&inetsw6_lock);
 646         return ret;
 647
 648 out_permanent:
 649         pr_err("Attempt to override permanent protocol %d\n", protocol);
 650         goto out;
 651
 652 out_illegal:
 653         pr_err("Ignoring attempt to register invalid socket type %d\n",
 654                p->type);
 655         goto out;
 656 }
 657 EXPORT_SYMBOL(inet6_register_protosw);
 658
 659 void
 660 inet6_unregister_protosw(struct inet_protosw *p)
 661 {
 662         if (INET_PROTOSW_PERMANENT & p->flags) {
 663                 pr_err("Attempt to unregister permanent protocol %d\n",
 664                        p->protocol);
 665         } else {
 666                 spin_lock_bh(&inetsw6_lock);
 667                 list_del_rcu(&p->list);
 668                 spin_unlock_bh(&inetsw6_lock);
 669
 670                 synchronize_net();
 671         }
 672 }
 673 EXPORT_SYMBOL(inet6_unregister_protosw);
 674
 675 int inet6_sk_rebuild_header(struct sock *sk)
 676 {
 677         struct ipv6_pinfo *np = inet6_sk(sk);
 678         struct dst_entry *dst;
 679
 680         dst = __sk_dst_check(sk, np->dst_cookie);
 681
 682         if (dst == NULL) {
 683                 struct inet_sock *inet = inet_sk(sk);
 684                 struct in6_addr *final_p, final;
 685                 struct flowi6 fl6;
 686
 687                 memset(&fl6, 0, sizeof(fl6));
 688                 fl6.flowi6_proto = sk->sk_protocol;
 689                 fl6.daddr = np->daddr;
 690                 fl6.saddr = np->saddr;
 691                 fl6.flowlabel = np->flow_label;
 692                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 693                 fl6.flowi6_mark = sk->sk_mark;
 694                 fl6.fl6_dport = inet->inet_dport;
 695                 fl6.fl6_sport = inet->inet_sport;
 696                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 697
 698                 final_p = fl6_update_dst(&fl6, np->opt, &final);
 699
 700                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
 701                 if (IS_ERR(dst)) {
 702                         sk->sk_route_caps = 0;
 703                         sk->sk_err_soft = -PTR_ERR(dst);
 704                         return PTR_ERR(dst);
 705                 }
 706
 707                 __ip6_dst_store(sk, dst, NULL, NULL);
 708         }
 709
 710         return 0;
 711 }
 712 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 713
 714 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
 715 {
 716         const struct ipv6_pinfo *np = inet6_sk(sk);
 717         const struct inet6_skb_parm *opt = IP6CB(skb);
 718
 719         if (np->rxopt.all) {
 720                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 721                                   np->rxopt.bits.ohopopts)) ||
 722                     ((IPV6_FLOWINFO_MASK &
 723                       *(__be32 *)skb_network_header(skb)) &&
 724                      np->rxopt.bits.rxflow) ||
 725                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 726                      np->rxopt.bits.osrcrt)) ||
 727                     ((opt->dst1 || opt->dst0) &&
 728                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 729                         return true;
 730         }
 731         return false;
 732 }
 733 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 734
 735 static struct packet_type ipv6_packet_type __read_mostly = {
 736         .type = cpu_to_be16(ETH_P_IPV6),
 737         .func = ipv6_rcv,
 738 };
 739
 740 static int __init ipv6_packet_init(void)
 741 {
 742         dev_add_pack(&ipv6_packet_type);
 743         return 0;
 744 }
 745
 746 static void ipv6_packet_cleanup(void)
 747 {
 748         dev_remove_pack(&ipv6_packet_type);
 749 }
 750
 751 static int __net_init ipv6_init_mibs(struct net *net)
 752 {
 753         if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
 754                           sizeof(struct udp_mib),
 755                           __alignof__(struct udp_mib)) < 0)
 756                 return -ENOMEM;
 757         if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
 758                           sizeof(struct udp_mib),
 759                           __alignof__(struct udp_mib)) < 0)
 760                 goto err_udplite_mib;
 761         if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
 762                           sizeof(struct ipstats_mib),
 763                           __alignof__(struct ipstats_mib)) < 0)
 764                 goto err_ip_mib;
 765         if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
 766                           sizeof(struct icmpv6_mib),
 767                           __alignof__(struct icmpv6_mib)) < 0)
 768                 goto err_icmp_mib;
 769         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 770                                                 GFP_KERNEL);
 771         if (!net->mib.icmpv6msg_statistics)
 772                 goto err_icmpmsg_mib;
 773         return 0;
 774
 775 err_icmpmsg_mib:
 776         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 777 err_icmp_mib:
 778         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 779 err_ip_mib:
 780         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 781 err_udplite_mib:
 782         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 783         return -ENOMEM;
 784 }
 785
 786 static void ipv6_cleanup_mibs(struct net *net)
 787 {
 788         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 789         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 790         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 791         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 792         kfree(net->mib.icmpv6msg_statistics);
 793 }
 794
 795 static int __net_init inet6_net_init(struct net *net)
 796 {
 797         int err = 0;
 798
 799         net->ipv6.sysctl.bindv6only = 0;
 800         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 801
 802         err = ipv6_init_mibs(net);
 803         if (err)
 804                 return err;
 805 #ifdef CONFIG_PROC_FS
 806         err = udp6_proc_init(net);
 807         if (err)
 808                 goto out;
 809         err = tcp6_proc_init(net);
 810         if (err)
 811                 goto proc_tcp6_fail;
 812         err = ac6_proc_init(net);
 813         if (err)
 814                 goto proc_ac6_fail;
 815 #endif
 816         return err;
 817
 818 #ifdef CONFIG_PROC_FS
 819 proc_ac6_fail:
 820         tcp6_proc_exit(net);
 821 proc_tcp6_fail:
 822         udp6_proc_exit(net);
 823 out:
 824         ipv6_cleanup_mibs(net);
 825         return err;
 826 #endif
 827 }
 828
 829 static void __net_exit inet6_net_exit(struct net *net)
 830 {
 831 #ifdef CONFIG_PROC_FS
 832         udp6_proc_exit(net);
 833         tcp6_proc_exit(net);
 834         ac6_proc_exit(net);
 835 #endif
 836         ipv6_cleanup_mibs(net);
 837 }
 838
 839 static struct pernet_operations inet6_net_ops = {
 840         .init = inet6_net_init,
 841         .exit = inet6_net_exit,
 842 };
 843
 844 static int __init inet6_init(void)
 845 {
 846         struct list_head *r;
 847         int err = 0;
 848
 849         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
 850
 851         /* Register the socket-side information for inet6_create.  */
 852         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 853                 INIT_LIST_HEAD(r);
 854
 855         if (disable_ipv6_mod) {
 856                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 857                 goto out;
 858         }
 859
 860         err = proto_register(&tcpv6_prot, 1);
 861         if (err)
 862                 goto out;
 863
 864         err = proto_register(&udpv6_prot, 1);
 865         if (err)
 866                 goto out_unregister_tcp_proto;
 867
 868         err = proto_register(&udplitev6_prot, 1);
 869         if (err)
 870                 goto out_unregister_udp_proto;
 871
 872         err = proto_register(&rawv6_prot, 1);
 873         if (err)
 874                 goto out_unregister_udplite_proto;
 875
 876
 877         /* We MUST register RAW sockets before we create the ICMP6,
 878          * IGMP6, or NDISC control sockets.
 879          */
 880         err = rawv6_init();
 881         if (err)
 882                 goto out_unregister_raw_proto;
 883
 884         /* Register the family here so that the init calls below will
 885          * be able to create sockets. (?? is this dangerous ??)
 886          */
 887         err = sock_register(&inet6_family_ops);
 888         if (err)
 889                 goto out_sock_register_fail;
 890
 891         tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
 892
 893         /*
 894          *      ipngwg API draft makes clear that the correct semantics
 895          *      for TCP and UDP is to consider one TCP and UDP instance
 896          *      in a host available by both INET and INET6 APIs and
 897          *      able to communicate via both network protocols.
 898          */
 899
 900         err = register_pernet_subsys(&inet6_net_ops);
 901         if (err)
 902                 goto register_pernet_fail;
 903         err = icmpv6_init();
 904         if (err)
 905                 goto icmp_fail;
 906         err = ip6_mr_init();
 907         if (err)
 908                 goto ipmr_fail;
 909         err = ndisc_init();
 910         if (err)
 911                 goto ndisc_fail;
 912         err = igmp6_init();
 913         if (err)
 914                 goto igmp_fail;
 915         err = ipv6_netfilter_init();
 916         if (err)
 917                 goto netfilter_fail;
 918         /* Create /proc/foo6 entries. */
 919 #ifdef CONFIG_PROC_FS
 920         err = -ENOMEM;
 921         if (raw6_proc_init())
 922                 goto proc_raw6_fail;
 923         if (udplite6_proc_init())
 924                 goto proc_udplite6_fail;
 925         if (ipv6_misc_proc_init())
 926                 goto proc_misc6_fail;
 927         if (if6_proc_init())
 928                 goto proc_if6_fail;
 929 #endif
 930         err = ip6_route_init();
 931         if (err)
 932                 goto ip6_route_fail;
 933         err = ip6_flowlabel_init();
 934         if (err)
 935                 goto ip6_flowlabel_fail;
 936         err = addrconf_init();
 937         if (err)
 938                 goto addrconf_fail;
 939
 940         /* Init v6 extension headers. */
 941         err = ipv6_exthdrs_init();
 942         if (err)
 943                 goto ipv6_exthdrs_fail;
 944
 945         err = ipv6_frag_init();
 946         if (err)
 947                 goto ipv6_frag_fail;
 948
 949         /* Init v6 transport protocols. */
 950         err = udpv6_init();
 951         if (err)
 952                 goto udpv6_fail;
 953
 954         err = udplitev6_init();
 955         if (err)
 956                 goto udplitev6_fail;
 957
 958         err = tcpv6_init();
 959         if (err)
 960                 goto tcpv6_fail;
 961
 962         err = ipv6_packet_init();
 963         if (err)
 964                 goto ipv6_packet_fail;
 965
 966 #ifdef CONFIG_SYSCTL
 967         err = ipv6_sysctl_register();
 968         if (err)
 969                 goto sysctl_fail;
 970 #endif
 971 out:
 972         return err;
 973
 974 #ifdef CONFIG_SYSCTL
 975 sysctl_fail:
 976         ipv6_packet_cleanup();
 977 #endif
 978 ipv6_packet_fail:
 979         tcpv6_exit();
 980 tcpv6_fail:
 981         udplitev6_exit();
 982 udplitev6_fail:
 983         udpv6_exit();
 984 udpv6_fail:
 985         ipv6_frag_exit();
 986 ipv6_frag_fail:
 987         ipv6_exthdrs_exit();
 988 ipv6_exthdrs_fail:
 989         addrconf_cleanup();
 990 addrconf_fail:
 991         ip6_flowlabel_cleanup();
 992 ip6_flowlabel_fail:
 993         ip6_route_cleanup();
 994 ip6_route_fail:
 995 #ifdef CONFIG_PROC_FS
 996         if6_proc_exit();
 997 proc_if6_fail:
 998         ipv6_misc_proc_exit();
 999 proc_misc6_fail:
1000         udplite6_proc_exit();
1001 proc_udplite6_fail:
1002         raw6_proc_exit();
1003 proc_raw6_fail:
1004 #endif
1005         ipv6_netfilter_fini();
1006 netfilter_fail:
1007         igmp6_cleanup();
1008 igmp_fail:
1009         ndisc_cleanup();
1010 ndisc_fail:
1011         ip6_mr_cleanup();
1012 ipmr_fail:
1013         icmpv6_cleanup();
1014 icmp_fail:
1015         unregister_pernet_subsys(&inet6_net_ops);
1016 register_pernet_fail:
1017         sock_unregister(PF_INET6);
1018         rtnl_unregister_all(PF_INET6);
1019 out_sock_register_fail:
1020         rawv6_exit();
1021 out_unregister_raw_proto:
1022         proto_unregister(&rawv6_prot);
1023 out_unregister_udplite_proto:
1024         proto_unregister(&udplitev6_prot);
1025 out_unregister_udp_proto:
1026         proto_unregister(&udpv6_prot);
1027 out_unregister_tcp_proto:
1028         proto_unregister(&tcpv6_prot);
1029         goto out;
1030 }
1031 module_init(inet6_init);
1032
1033 static void __exit inet6_exit(void)
1034 {
1035         if (disable_ipv6_mod)
1036                 return;
1037
1038         /* First of all disallow new sockets creation. */
1039         sock_unregister(PF_INET6);
1040         /* Disallow any further netlink messages */
1041         rtnl_unregister_all(PF_INET6);
1042
1043         udpv6_exit();
1044         udplitev6_exit();
1045         tcpv6_exit();
1046
1047         /* Cleanup code parts. */
1048         ipv6_packet_cleanup();
1049         ipv6_frag_exit();
1050         ipv6_exthdrs_exit();
1051         addrconf_cleanup();
1052         ip6_flowlabel_cleanup();
1053         ip6_route_cleanup();
1054 #ifdef CONFIG_PROC_FS
1055
1056         /* Cleanup code parts. */
1057         if6_proc_exit();
1058         ipv6_misc_proc_exit();
1059         udplite6_proc_exit();
1060         raw6_proc_exit();
1061 #endif
1062         ipv6_netfilter_fini();
1063         igmp6_cleanup();
1064         ndisc_cleanup();
1065         ip6_mr_cleanup();
1066         icmpv6_cleanup();
1067         rawv6_exit();
1068
1069         unregister_pernet_subsys(&inet6_net_ops);
1070         proto_unregister(&rawv6_prot);
1071         proto_unregister(&udplitev6_prot);
1072         proto_unregister(&udpv6_prot);
1073         proto_unregister(&tcpv6_prot);
1074
1075         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1076 }
1077 module_exit(inet6_exit);
1078
1079 MODULE_ALIAS_NETPROTO(PF_INET6);