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