net/ipv6/af_inet6.c

   1 /*
   2  *      PF_INET6 socket protocol family
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Adapted from linux/net/ipv4/af_inet.c
   9  *
  10  *      Fixes:
  11  *      piggy, Karl Knutson     :       Socket protocol table
  12  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
  13  *      Arnaldo Melo            :       check proc_net_create return, cleanups
  14  *
  15  *      This program is free software; you can redistribute it and/or
  16  *      modify it under the terms of the GNU General Public License
  17  *      as published by the Free Software Foundation; either version
  18  *      2 of the License, or (at your option) any later version.
  19  */
  20
  21 #define pr_fmt(fmt) "IPv6: " fmt
  22
  23 #include <linux/module.h>
  24 #include <linux/capability.h>
  25 #include <linux/errno.h>
  26 #include <linux/types.h>
  27 #include <linux/socket.h>
  28 #include <linux/in.h>
  29 #include <linux/kernel.h>
  30 #include <linux/timer.h>
  31 #include <linux/string.h>
  32 #include <linux/sockios.h>
  33 #include <linux/net.h>
  34 #include <linux/fcntl.h>
  35 #include <linux/mm.h>
  36 #include <linux/interrupt.h>
  37 #include <linux/proc_fs.h>
  38 #include <linux/stat.h>
  39 #include <linux/init.h>
  40 #include <linux/slab.h>
  41
  42 #include <linux/inet.h>
  43 #include <linux/netdevice.h>
  44 #include <linux/icmpv6.h>
  45 #include <linux/netfilter_ipv6.h>
  46
  47 #include <net/ip.h>
  48 #include <net/ipv6.h>
  49 #include <net/udp.h>
  50 #include <net/udplite.h>
  51 #include <net/tcp.h>
  52 #include <net/ping.h>
  53 #include <net/protocol.h>
  54 #include <net/inet_common.h>
  55 #include <net/route.h>
  56 #include <net/transp_v6.h>
  57 #include <net/ip6_route.h>
  58 #include <net/addrconf.h>
  59 #ifdef CONFIG_IPV6_TUNNEL
  60 #include <net/ip6_tunnel.h>
  61 #endif
  62
  63 #include <asm/uaccess.h>
  64 #include <linux/mroute6.h>
  65
  66 #ifdef CONFIG_ANDROID_PARANOID_NETWORK
  67 #include <linux/android_aid.h>
  68
  69 static inline int current_has_network(void)
  70 {
  71         return in_egroup_p(AID_INET) || capable(CAP_NET_RAW);
  72 }
  73 #else
  74 static inline int current_has_network(void)
  75 {
  76         return 1;
  77 }
  78 #endif
  79
  80 MODULE_AUTHOR("Cast of dozens");
  81 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  82 MODULE_LICENSE("GPL");
  83
  84 /* The inetsw6 table contains everything that inet6_create needs to
  85  * build a new socket.
  86  */
  87 static struct list_head inetsw6[SOCK_MAX];
  88 static DEFINE_SPINLOCK(inetsw6_lock);
  89
  90 struct ipv6_params ipv6_defaults = {
  91         .disable_ipv6 = 0,
  92         .autoconf = 1,
  93 };
  94
  95 static int disable_ipv6_mod;
  96
  97 module_param_named(disable, disable_ipv6_mod, int, 0444);
  98 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  99
 100 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
 101 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
 102
 103 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
 104 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
 105
 106 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
 107 {
 108         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
 109
 110         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
 111 }
 112
 113 static int inet6_create(struct net *net, struct socket *sock, int protocol,
 114                         int kern)
 115 {
 116         struct inet_sock *inet;
 117         struct ipv6_pinfo *np;
 118         struct sock *sk;
 119         struct inet_protosw *answer;
 120         struct proto *answer_prot;
 121         unsigned char answer_flags;
 122         char answer_no_check;
 123         int try_loading_module = 0;
 124         int err;
 125
 126         if (!current_has_network())
 127                 return -EACCES;
 128
 129         if (sock->type != SOCK_RAW &&
 130             sock->type != SOCK_DGRAM &&
 131             !inet_ehash_secret)
 132                 build_ehash_secret();
 133
 134         if (protocol < 0 || protocol >= IPPROTO_MAX)
 135                 return -EINVAL;
 136
 137         /* Look for the requested type/protocol pair. */
 138 lookup_protocol:
 139         err = -ESOCKTNOSUPPORT;
 140         rcu_read_lock();
 141         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 142
 143                 err = 0;
 144                 /* Check the non-wild match. */
 145                 if (protocol == answer->protocol) {
 146                         if (protocol != IPPROTO_IP)
 147                                 break;
 148                 } else {
 149                         /* Check for the two wild cases. */
 150                         if (IPPROTO_IP == protocol) {
 151                                 protocol = answer->protocol;
 152                                 break;
 153                         }
 154                         if (IPPROTO_IP == answer->protocol)
 155                                 break;
 156                 }
 157                 err = -EPROTONOSUPPORT;
 158         }
 159
 160         if (err) {
 161                 if (try_loading_module < 2) {
 162                         rcu_read_unlock();
 163                         /*
 164                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 165                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 166                          */
 167                         if (++try_loading_module == 1)
 168                                 request_module("net-pf-%d-proto-%d-type-%d",
 169                                                 PF_INET6, protocol, sock->type);
 170                         /*
 171                          * Fall back to generic, e.g. net-pf-10-proto-132
 172                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 173                          */
 174                         else
 175                                 request_module("net-pf-%d-proto-%d",
 176                                                 PF_INET6, protocol);
 177                         goto lookup_protocol;
 178                 } else
 179                         goto out_rcu_unlock;
 180         }
 181
 182         err = -EPERM;
 183         if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
 184                 goto out_rcu_unlock;
 185
 186         sock->ops = answer->ops;
 187         answer_prot = answer->prot;
 188         answer_no_check = answer->no_check;
 189         answer_flags = answer->flags;
 190         rcu_read_unlock();
 191
 192         WARN_ON(answer_prot->slab == NULL);
 193
 194         err = -ENOBUFS;
 195         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 196         if (sk == NULL)
 197                 goto out;
 198
 199         sock_init_data(sock, sk);
 200
 201         err = 0;
 202         sk->sk_no_check = answer_no_check;
 203         if (INET_PROTOSW_REUSE & answer_flags)
 204                 sk->sk_reuse = SK_CAN_REUSE;
 205
 206         inet = inet_sk(sk);
 207         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 208
 209         if (SOCK_RAW == sock->type) {
 210                 inet->inet_num = protocol;
 211                 if (IPPROTO_RAW == protocol)
 212                         inet->hdrincl = 1;
 213         }
 214
 215         sk->sk_destruct         = inet_sock_destruct;
 216         sk->sk_family           = PF_INET6;
 217         sk->sk_protocol         = protocol;
 218
 219         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 220
 221         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 222         np->hop_limit   = -1;
 223         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 224         np->mc_loop     = 1;
 225         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 226         np->ipv6only    = net->ipv6.sysctl.bindv6only;
 227
 228         /* Init the ipv4 part of the socket since we can have sockets
 229          * using v6 API for ipv4.
 230          */
 231         inet->uc_ttl    = -1;
 232
 233         inet->mc_loop   = 1;
 234         inet->mc_ttl    = 1;
 235         inet->mc_index  = 0;
 236         inet->mc_list   = NULL;
 237         inet->rcv_tos   = 0;
 238
 239         if (ipv4_config.no_pmtu_disc)
 240                 inet->pmtudisc = IP_PMTUDISC_DONT;
 241         else
 242                 inet->pmtudisc = IP_PMTUDISC_WANT;
 243         /*
 244          * Increment only the relevant sk_prot->socks debug field, this changes
 245          * the previous behaviour of incrementing both the equivalent to
 246          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 247          *
 248          * This allows better debug granularity as we'll know exactly how many
 249          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 250          * transport protocol socks. -acme
 251          */
 252         sk_refcnt_debug_inc(sk);
 253
 254         if (inet->inet_num) {
 255                 /* It assumes that any protocol which allows
 256                  * the user to assign a number at socket
 257                  * creation time automatically shares.
 258                  */
 259                 inet->inet_sport = htons(inet->inet_num);
 260                 sk->sk_prot->hash(sk);
 261         }
 262         if (sk->sk_prot->init) {
 263                 err = sk->sk_prot->init(sk);
 264                 if (err) {
 265                         sk_common_release(sk);
 266                         goto out;
 267                 }
 268         }
 269 out:
 270         return err;
 271 out_rcu_unlock:
 272         rcu_read_unlock();
 273         goto out;
 274 }
 275
 276
 277 /* bind for INET6 API */
 278 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 279 {
 280         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 281         struct sock *sk = sock->sk;
 282         struct inet_sock *inet = inet_sk(sk);
 283         struct ipv6_pinfo *np = inet6_sk(sk);
 284         struct net *net = sock_net(sk);
 285         __be32 v4addr = 0;
 286         unsigned short snum;
 287         int addr_type = 0;
 288         int err = 0;
 289
 290         /* If the socket has its own bind function then use it. */
 291         if (sk->sk_prot->bind)
 292                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 293
 294         if (addr_len < SIN6_LEN_RFC2133)
 295                 return -EINVAL;
 296
 297         if (addr->sin6_family != AF_INET6)
 298                 return -EAFNOSUPPORT;
 299
 300         addr_type = ipv6_addr_type(&addr->sin6_addr);
 301         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 302                 return -EINVAL;
 303
 304         snum = ntohs(addr->sin6_port);
 305         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 306                 return -EACCES;
 307
 308         lock_sock(sk);
 309
 310         /* Check these errors (active socket, double bind). */
 311         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 312                 err = -EINVAL;
 313                 goto out;
 314         }
 315
 316         /* Check if the address belongs to the host. */
 317         if (addr_type == IPV6_ADDR_MAPPED) {
 318                 int chk_addr_ret;
 319
 320                 /* Binding to v4-mapped address on a v6-only socket
 321                  * makes no sense
 322                  */
 323                 if (np->ipv6only) {
 324                         err = -EINVAL;
 325                         goto out;
 326                 }
 327
 328                 /* Reproduce AF_INET checks to make the bindings consistent */
 329                 v4addr = addr->sin6_addr.s6_addr32[3];
 330                 chk_addr_ret = inet_addr_type(net, v4addr);
 331                 if (!sysctl_ip_nonlocal_bind &&
 332                     !(inet->freebind || inet->transparent) &&
 333                     v4addr != htonl(INADDR_ANY) &&
 334                     chk_addr_ret != RTN_LOCAL &&
 335                     chk_addr_ret != RTN_MULTICAST &&
 336                     chk_addr_ret != RTN_BROADCAST) {
 337                         err = -EADDRNOTAVAIL;
 338                         goto out;
 339                 }
 340         } else {
 341                 if (addr_type != IPV6_ADDR_ANY) {
 342                         struct net_device *dev = NULL;
 343
 344                         rcu_read_lock();
 345                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 346                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 347                                     addr->sin6_scope_id) {
 348                                         /* Override any existing binding, if another one
 349                                          * is supplied by user.
 350                                          */
 351                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 352                                 }
 353
 354                                 /* Binding to link-local address requires an interface */
 355                                 if (!sk->sk_bound_dev_if) {
 356                                         err = -EINVAL;
 357                                         goto out_unlock;
 358                                 }
 359                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 360                                 if (!dev) {
 361                                         err = -ENODEV;
 362                                         goto out_unlock;
 363                                 }
 364                         }
 365
 366                         /* ipv4 addr of the socket is invalid.  Only the
 367                          * unspecified and mapped address have a v4 equivalent.
 368                          */
 369                         v4addr = LOOPBACK4_IPV6;
 370                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 371                                 if (!(inet->freebind || inet->transparent) &&
 372                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 373                                                    dev, 0)) {
 374                                         err = -EADDRNOTAVAIL;
 375                                         goto out_unlock;
 376                                 }
 377                         }
 378                         rcu_read_unlock();
 379                 }
 380         }
 381
 382         inet->inet_rcv_saddr = v4addr;
 383         inet->inet_saddr = v4addr;
 384
 385         np->rcv_saddr = addr->sin6_addr;
 386
 387         if (!(addr_type & IPV6_ADDR_MULTICAST))
 388                 np->saddr = addr->sin6_addr;
 389
 390         /* Make sure we are allowed to bind here. */
 391         if (sk->sk_prot->get_port(sk, snum)) {
 392                 inet_reset_saddr(sk);
 393                 err = -EADDRINUSE;
 394                 goto out;
 395         }
 396
 397         if (addr_type != IPV6_ADDR_ANY) {
 398                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 399                 if (addr_type != IPV6_ADDR_MAPPED)
 400                         np->ipv6only = 1;
 401         }
 402         if (snum)
 403                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 404         inet->inet_sport = htons(inet->inet_num);
 405         inet->inet_dport = 0;
 406         inet->inet_daddr = 0;
 407 out:
 408         release_sock(sk);
 409         return err;
 410 out_unlock:
 411         rcu_read_unlock();
 412         goto out;
 413 }
 414 EXPORT_SYMBOL(inet6_bind);
 415
 416 int inet6_release(struct socket *sock)
 417 {
 418         struct sock *sk = sock->sk;
 419
 420         if (sk == NULL)
 421                 return -EINVAL;
 422
 423         /* Free mc lists */
 424         ipv6_sock_mc_close(sk);
 425
 426         /* Free ac lists */
 427         ipv6_sock_ac_close(sk);
 428
 429         return inet_release(sock);
 430 }
 431 EXPORT_SYMBOL(inet6_release);
 432
 433 void inet6_destroy_sock(struct sock *sk)
 434 {
 435         struct ipv6_pinfo *np = inet6_sk(sk);
 436         struct sk_buff *skb;
 437         struct ipv6_txoptions *opt;
 438
 439         /* Release rx options */
 440
 441         skb = xchg(&np->pktoptions, NULL);
 442         if (skb != NULL)
 443                 kfree_skb(skb);
 444
 445         skb = xchg(&np->rxpmtu, NULL);
 446         if (skb != NULL)
 447                 kfree_skb(skb);
 448
 449         /* Free flowlabels */
 450         fl6_free_socklist(sk);
 451
 452         /* Free tx options */
 453
 454         opt = xchg(&np->opt, NULL);
 455         if (opt != NULL)
 456                 sock_kfree_s(sk, opt, opt->tot_len);
 457 }
 458 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 459
 460 /*
 461  *      This does both peername and sockname.
 462  */
 463
 464 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 465                  int *uaddr_len, int peer)
 466 {
 467         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 468         struct sock *sk = sock->sk;
 469         struct inet_sock *inet = inet_sk(sk);
 470         struct ipv6_pinfo *np = inet6_sk(sk);
 471
 472         sin->sin6_family = AF_INET6;
 473         sin->sin6_flowinfo = 0;
 474         sin->sin6_scope_id = 0;
 475         if (peer) {
 476                 if (!inet->inet_dport)
 477                         return -ENOTCONN;
 478                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 479                     peer == 1)
 480                         return -ENOTCONN;
 481                 sin->sin6_port = inet->inet_dport;
 482                 sin->sin6_addr = np->daddr;
 483                 if (np->sndflow)
 484                         sin->sin6_flowinfo = np->flow_label;
 485         } else {
 486                 if (ipv6_addr_any(&np->rcv_saddr))
 487                         sin->sin6_addr = np->saddr;
 488                 else
 489                         sin->sin6_addr = np->rcv_saddr;
 490
 491                 sin->sin6_port = inet->inet_sport;
 492         }
 493         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 494                                                  sk->sk_bound_dev_if);
 495         *uaddr_len = sizeof(*sin);
 496         return 0;
 497 }
 498 EXPORT_SYMBOL(inet6_getname);
 499
 500 int inet6_killaddr_ioctl(struct net *net, void __user *arg) {
 501         struct in6_ifreq ireq;
 502         struct sockaddr_in6 sin6;
 503
 504         if (!capable(CAP_NET_ADMIN))
 505                 return -EACCES;
 506
 507         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
 508                 return -EFAULT;
 509
 510         sin6.sin6_family = AF_INET6;
 511         sin6.sin6_addr = ireq.ifr6_addr;
 512         return tcp_nuke_addr(net, (struct sockaddr *) &sin6);
 513 }
 514
 515 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 516 {
 517         struct sock *sk = sock->sk;
 518         struct net *net = sock_net(sk);
 519
 520         switch (cmd) {
 521         case SIOCGSTAMP:
 522                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 523
 524         case SIOCGSTAMPNS:
 525                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 526
 527         case SIOCADDRT:
 528         case SIOCDELRT:
 529
 530                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 531
 532         case SIOCSIFADDR:
 533                 return addrconf_add_ifaddr(net, (void __user *) arg);
 534         case SIOCDIFADDR:
 535                 return addrconf_del_ifaddr(net, (void __user *) arg);
 536         case SIOCSIFDSTADDR:
 537                 return addrconf_set_dstaddr(net, (void __user *) arg);
 538         case SIOCKILLADDR:
 539                 return inet6_killaddr_ioctl(net, (void __user *) arg);
 540         default:
 541                 if (!sk->sk_prot->ioctl)
 542                         return -ENOIOCTLCMD;
 543                 return sk->sk_prot->ioctl(sk, cmd, arg);
 544         }
 545         /*NOTREACHED*/
 546         return 0;
 547 }
 548 EXPORT_SYMBOL(inet6_ioctl);
 549
 550 const struct proto_ops inet6_stream_ops = {
 551         .family            = PF_INET6,
 552         .owner             = THIS_MODULE,
 553         .release           = inet6_release,
 554         .bind              = inet6_bind,
 555         .connect           = inet_stream_connect,       /* ok           */
 556         .socketpair        = sock_no_socketpair,        /* a do nothing */
 557         .accept            = inet_accept,               /* ok           */
 558         .getname           = inet6_getname,
 559         .poll              = tcp_poll,                  /* ok           */
 560         .ioctl             = inet6_ioctl,               /* must change  */
 561         .listen            = inet_listen,               /* ok           */
 562         .shutdown          = inet_shutdown,             /* ok           */
 563         .setsockopt        = sock_common_setsockopt,    /* ok           */
 564         .getsockopt        = sock_common_getsockopt,    /* ok           */
 565         .sendmsg           = inet_sendmsg,              /* ok           */
 566         .recvmsg           = inet_recvmsg,              /* ok           */
 567         .mmap              = sock_no_mmap,
 568         .sendpage          = inet_sendpage,
 569         .splice_read       = tcp_splice_read,
 570 #ifdef CONFIG_COMPAT
 571         .compat_setsockopt = compat_sock_common_setsockopt,
 572         .compat_getsockopt = compat_sock_common_getsockopt,
 573 #endif
 574 };
 575
 576 const struct proto_ops inet6_dgram_ops = {
 577         .family            = PF_INET6,
 578         .owner             = THIS_MODULE,
 579         .release           = inet6_release,
 580         .bind              = inet6_bind,
 581         .connect           = inet_dgram_connect,        /* ok           */
 582         .socketpair        = sock_no_socketpair,        /* a do nothing */
 583         .accept            = sock_no_accept,            /* a do nothing */
 584         .getname           = inet6_getname,
 585         .poll              = udp_poll,                  /* ok           */
 586         .ioctl             = inet6_ioctl,               /* must change  */
 587         .listen            = sock_no_listen,            /* ok           */
 588         .shutdown          = inet_shutdown,             /* ok           */
 589         .setsockopt        = sock_common_setsockopt,    /* ok           */
 590         .getsockopt        = sock_common_getsockopt,    /* ok           */
 591         .sendmsg           = inet_sendmsg,              /* ok           */
 592         .recvmsg           = inet_recvmsg,              /* ok           */
 593         .mmap              = sock_no_mmap,
 594         .sendpage          = sock_no_sendpage,
 595 #ifdef CONFIG_COMPAT
 596         .compat_setsockopt = compat_sock_common_setsockopt,
 597         .compat_getsockopt = compat_sock_common_getsockopt,
 598 #endif
 599 };
 600
 601 static const struct net_proto_family inet6_family_ops = {
 602         .family = PF_INET6,
 603         .create = inet6_create,
 604         .owner  = THIS_MODULE,
 605 };
 606
 607 int inet6_register_protosw(struct inet_protosw *p)
 608 {
 609         struct list_head *lh;
 610         struct inet_protosw *answer;
 611         struct list_head *last_perm;
 612         int protocol = p->protocol;
 613         int ret;
 614
 615         spin_lock_bh(&inetsw6_lock);
 616
 617         ret = -EINVAL;
 618         if (p->type >= SOCK_MAX)
 619                 goto out_illegal;
 620
 621         /* If we are trying to override a permanent protocol, bail. */
 622         answer = NULL;
 623         ret = -EPERM;
 624         last_perm = &inetsw6[p->type];
 625         list_for_each(lh, &inetsw6[p->type]) {
 626                 answer = list_entry(lh, struct inet_protosw, list);
 627
 628                 /* Check only the non-wild match. */
 629                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 630                         if (protocol == answer->protocol)
 631                                 break;
 632                         last_perm = lh;
 633                 }
 634
 635                 answer = NULL;
 636         }
 637         if (answer)
 638                 goto out_permanent;
 639
 640         /* Add the new entry after the last permanent entry if any, so that
 641          * the new entry does not override a permanent entry when matched with
 642          * a wild-card protocol. But it is allowed to override any existing
 643          * non-permanent entry.  This means that when we remove this entry, the
 644          * system automatically returns to the old behavior.
 645          */
 646         list_add_rcu(&p->list, last_perm);
 647         ret = 0;
 648 out:
 649         spin_unlock_bh(&inetsw6_lock);
 650         return ret;
 651
 652 out_permanent:
 653         pr_err("Attempt to override permanent protocol %d\n", protocol);
 654         goto out;
 655
 656 out_illegal:
 657         pr_err("Ignoring attempt to register invalid socket type %d\n",
 658                p->type);
 659         goto out;
 660 }
 661 EXPORT_SYMBOL(inet6_register_protosw);
 662
 663 void
 664 inet6_unregister_protosw(struct inet_protosw *p)
 665 {
 666         if (INET_PROTOSW_PERMANENT & p->flags) {
 667                 pr_err("Attempt to unregister permanent protocol %d\n",
 668                        p->protocol);
 669         } else {
 670                 spin_lock_bh(&inetsw6_lock);
 671                 list_del_rcu(&p->list);
 672                 spin_unlock_bh(&inetsw6_lock);
 673
 674                 synchronize_net();
 675         }
 676 }
 677 EXPORT_SYMBOL(inet6_unregister_protosw);
 678
 679 int inet6_sk_rebuild_header(struct sock *sk)
 680 {
 681         struct ipv6_pinfo *np = inet6_sk(sk);
 682         struct dst_entry *dst;
 683
 684         dst = __sk_dst_check(sk, np->dst_cookie);
 685
 686         if (dst == NULL) {
 687                 struct inet_sock *inet = inet_sk(sk);
 688                 struct in6_addr *final_p, final;
 689                 struct flowi6 fl6;
 690
 691                 memset(&fl6, 0, sizeof(fl6));
 692                 fl6.flowi6_proto = sk->sk_protocol;
 693                 fl6.daddr = np->daddr;
 694                 fl6.saddr = np->saddr;
 695                 fl6.flowlabel = np->flow_label;
 696                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 697                 fl6.flowi6_mark = sk->sk_mark;
 698                 fl6.fl6_dport = inet->inet_dport;
 699                 fl6.fl6_sport = inet->inet_sport;
 700                 fl6.flowi6_uid = sock_i_uid(sk);
 701                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 702
 703                 final_p = fl6_update_dst(&fl6, np->opt, &final);
 704
 705                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
 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 {
 721         const struct ipv6_pinfo *np = inet6_sk(sk);
 722         const struct inet6_skb_parm *opt = IP6CB(skb);
 723
 724         if (np->rxopt.all) {
 725                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 726                                   np->rxopt.bits.ohopopts)) ||
 727                     ((IPV6_FLOWINFO_MASK &
 728                       *(__be32 *)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         if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
 759                           sizeof(struct udp_mib),
 760                           __alignof__(struct udp_mib)) < 0)
 761                 return -ENOMEM;
 762         if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
 763                           sizeof(struct udp_mib),
 764                           __alignof__(struct udp_mib)) < 0)
 765                 goto err_udplite_mib;
 766         if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
 767                           sizeof(struct ipstats_mib),
 768                           __alignof__(struct ipstats_mib)) < 0)
 769                 goto err_ip_mib;
 770         if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
 771                           sizeof(struct icmpv6_mib),
 772                           __alignof__(struct icmpv6_mib)) < 0)
 773                 goto err_icmp_mib;
 774         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 775                                                 GFP_KERNEL);
 776         if (!net->mib.icmpv6msg_statistics)
 777                 goto err_icmpmsg_mib;
 778         return 0;
 779
 780 err_icmpmsg_mib:
 781         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 782 err_icmp_mib:
 783         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 784 err_ip_mib:
 785         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 786 err_udplite_mib:
 787         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 788         return -ENOMEM;
 789 }
 790
 791 static void ipv6_cleanup_mibs(struct net *net)
 792 {
 793         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 794         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 795         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 796         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 797         kfree(net->mib.icmpv6msg_statistics);
 798 }
 799
 800 static int __net_init inet6_net_init(struct net *net)
 801 {
 802         int err = 0;
 803
 804         net->ipv6.sysctl.bindv6only = 0;
 805         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 806
 807         err = ipv6_init_mibs(net);
 808         if (err)
 809                 return err;
 810 #ifdef CONFIG_PROC_FS
 811         err = udp6_proc_init(net);
 812         if (err)
 813                 goto out;
 814         err = tcp6_proc_init(net);
 815         if (err)
 816                 goto proc_tcp6_fail;
 817         err = ac6_proc_init(net);
 818         if (err)
 819                 goto proc_ac6_fail;
 820 #endif
 821         return err;
 822
 823 #ifdef CONFIG_PROC_FS
 824 proc_ac6_fail:
 825         tcp6_proc_exit(net);
 826 proc_tcp6_fail:
 827         udp6_proc_exit(net);
 828 out:
 829         ipv6_cleanup_mibs(net);
 830         return err;
 831 #endif
 832 }
 833
 834 static void __net_exit inet6_net_exit(struct net *net)
 835 {
 836 #ifdef CONFIG_PROC_FS
 837         udp6_proc_exit(net);
 838         tcp6_proc_exit(net);
 839         ac6_proc_exit(net);
 840 #endif
 841         ipv6_cleanup_mibs(net);
 842 }
 843
 844 static struct pernet_operations inet6_net_ops = {
 845         .init = inet6_net_init,
 846         .exit = inet6_net_exit,
 847 };
 848
 849 static int __init inet6_init(void)
 850 {
 851         struct list_head *r;
 852         int err = 0;
 853
 854         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
 855
 856         /* Register the socket-side information for inet6_create.  */
 857         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 858                 INIT_LIST_HEAD(r);
 859
 860         if (disable_ipv6_mod) {
 861                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 862                 goto out;
 863         }
 864
 865         err = proto_register(&tcpv6_prot, 1);
 866         if (err)
 867                 goto out;
 868
 869         err = proto_register(&udpv6_prot, 1);
 870         if (err)
 871                 goto out_unregister_tcp_proto;
 872
 873         err = proto_register(&udplitev6_prot, 1);
 874         if (err)
 875                 goto out_unregister_udp_proto;
 876
 877         err = proto_register(&rawv6_prot, 1);
 878         if (err)
 879                 goto out_unregister_udplite_proto;
 880
 881         err = proto_register(&pingv6_prot, 1);
 882         if (err)
 883                 goto out_unregister_ping_proto;
 884
 885         /* We MUST register RAW sockets before we create the ICMP6,
 886          * IGMP6, or NDISC control sockets.
 887          */
 888         err = rawv6_init();
 889         if (err)
 890                 goto out_unregister_raw_proto;
 891
 892         /* Register the family here so that the init calls below will
 893          * be able to create sockets. (?? is this dangerous ??)
 894          */
 895         err = sock_register(&inet6_family_ops);
 896         if (err)
 897                 goto out_sock_register_fail;
 898
 899         tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
 900
 901         /*
 902          *      ipngwg API draft makes clear that the correct semantics
 903          *      for TCP and UDP is to consider one TCP and UDP instance
 904          *      in a host available by both INET and INET6 APIs and
 905          *      able to communicate via both network protocols.
 906          */
 907
 908         err = register_pernet_subsys(&inet6_net_ops);
 909         if (err)
 910                 goto register_pernet_fail;
 911         err = icmpv6_init();
 912         if (err)
 913                 goto icmp_fail;
 914         err = ip6_mr_init();
 915         if (err)
 916                 goto ipmr_fail;
 917         err = ndisc_init();
 918         if (err)
 919                 goto ndisc_fail;
 920         err = igmp6_init();
 921         if (err)
 922                 goto igmp_fail;
 923         err = ipv6_netfilter_init();
 924         if (err)
 925                 goto netfilter_fail;
 926         /* Create /proc/foo6 entries. */
 927 #ifdef CONFIG_PROC_FS
 928         err = -ENOMEM;
 929         if (raw6_proc_init())
 930                 goto proc_raw6_fail;
 931         if (udplite6_proc_init())
 932                 goto proc_udplite6_fail;
 933         if (ipv6_misc_proc_init())
 934                 goto proc_misc6_fail;
 935         if (if6_proc_init())
 936                 goto proc_if6_fail;
 937 #endif
 938         err = ip6_route_init();
 939         if (err)
 940                 goto ip6_route_fail;
 941         err = ip6_flowlabel_init();
 942         if (err)
 943                 goto ip6_flowlabel_fail;
 944         err = addrconf_init();
 945         if (err)
 946                 goto addrconf_fail;
 947
 948         /* Init v6 extension headers. */
 949         err = ipv6_exthdrs_init();
 950         if (err)
 951                 goto ipv6_exthdrs_fail;
 952
 953         err = ipv6_frag_init();
 954         if (err)
 955                 goto ipv6_frag_fail;
 956
 957         /* Init v6 transport protocols. */
 958         err = udpv6_init();
 959         if (err)
 960                 goto udpv6_fail;
 961
 962         err = udplitev6_init();
 963         if (err)
 964                 goto udplitev6_fail;
 965
 966         err = tcpv6_init();
 967         if (err)
 968                 goto tcpv6_fail;
 969
 970         err = ipv6_packet_init();
 971         if (err)
 972                 goto ipv6_packet_fail;
 973
 974         err = pingv6_init();
 975         if (err)
 976                 goto pingv6_fail;
 977
 978 #ifdef CONFIG_SYSCTL
 979         err = ipv6_sysctl_register();
 980         if (err)
 981                 goto sysctl_fail;
 982 #endif
 983 out:
 984         return err;
 985
 986 #ifdef CONFIG_SYSCTL
 987 sysctl_fail:
 988         ipv6_packet_cleanup();
 989 #endif
 990 pingv6_fail:
 991         pingv6_exit();
 992 ipv6_packet_fail:
 993         tcpv6_exit();
 994 tcpv6_fail:
 995         udplitev6_exit();
 996 udplitev6_fail:
 997         udpv6_exit();
 998 udpv6_fail:
 999         ipv6_frag_exit();
1000 ipv6_frag_fail:
1001         ipv6_exthdrs_exit();
1002 ipv6_exthdrs_fail:
1003         addrconf_cleanup();
1004 addrconf_fail:
1005         ip6_flowlabel_cleanup();
1006 ip6_flowlabel_fail:
1007         ip6_route_cleanup();
1008 ip6_route_fail:
1009 #ifdef CONFIG_PROC_FS
1010         if6_proc_exit();
1011 proc_if6_fail:
1012         ipv6_misc_proc_exit();
1013 proc_misc6_fail:
1014         udplite6_proc_exit();
1015 proc_udplite6_fail:
1016         raw6_proc_exit();
1017 proc_raw6_fail:
1018 #endif
1019         ipv6_netfilter_fini();
1020 netfilter_fail:
1021         igmp6_cleanup();
1022 igmp_fail:
1023         ndisc_cleanup();
1024 ndisc_fail:
1025         ip6_mr_cleanup();
1026 ipmr_fail:
1027         icmpv6_cleanup();
1028 icmp_fail:
1029         unregister_pernet_subsys(&inet6_net_ops);
1030 register_pernet_fail:
1031         sock_unregister(PF_INET6);
1032         rtnl_unregister_all(PF_INET6);
1033 out_sock_register_fail:
1034         rawv6_exit();
1035 out_unregister_ping_proto:
1036         proto_unregister(&pingv6_prot);
1037 out_unregister_raw_proto:
1038         proto_unregister(&rawv6_prot);
1039 out_unregister_udplite_proto:
1040         proto_unregister(&udplitev6_prot);
1041 out_unregister_udp_proto:
1042         proto_unregister(&udpv6_prot);
1043 out_unregister_tcp_proto:
1044         proto_unregister(&tcpv6_prot);
1045         goto out;
1046 }
1047 module_init(inet6_init);
1048
1049 static void __exit inet6_exit(void)
1050 {
1051         if (disable_ipv6_mod)
1052                 return;
1053
1054         /* First of all disallow new sockets creation. */
1055         sock_unregister(PF_INET6);
1056         /* Disallow any further netlink messages */
1057         rtnl_unregister_all(PF_INET6);
1058
1059         udpv6_exit();
1060         udplitev6_exit();
1061         tcpv6_exit();
1062
1063         /* Cleanup code parts. */
1064         ipv6_packet_cleanup();
1065         ipv6_frag_exit();
1066         ipv6_exthdrs_exit();
1067         addrconf_cleanup();
1068         ip6_flowlabel_cleanup();
1069         ip6_route_cleanup();
1070 #ifdef CONFIG_PROC_FS
1071
1072         /* Cleanup code parts. */
1073         if6_proc_exit();
1074         ipv6_misc_proc_exit();
1075         udplite6_proc_exit();
1076         raw6_proc_exit();
1077 #endif
1078         ipv6_netfilter_fini();
1079         igmp6_cleanup();
1080         ndisc_cleanup();
1081         ip6_mr_cleanup();
1082         icmpv6_cleanup();
1083         rawv6_exit();
1084
1085         unregister_pernet_subsys(&inet6_net_ops);
1086         proto_unregister(&rawv6_prot);
1087         proto_unregister(&udplitev6_prot);
1088         proto_unregister(&udpv6_prot);
1089         proto_unregister(&tcpv6_prot);
1090
1091         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1092 }
1093 module_exit(inet6_exit);
1094
1095 MODULE_ALIAS_NETPROTO(PF_INET6);