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 MODULE_AUTHOR("Cast of dozens");
  66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  67 MODULE_LICENSE("GPL");
  68
  69 /* The inetsw6 table contains everything that inet6_create needs to
  70  * build a new socket.
  71  */
  72 static struct list_head inetsw6[SOCK_MAX];
  73 static DEFINE_SPINLOCK(inetsw6_lock);
  74
  75 struct ipv6_params ipv6_defaults = {
  76         .disable_ipv6 = 0,
  77         .autoconf = 1,
  78 };
  79
  80 static int disable_ipv6_mod;
  81
  82 module_param_named(disable, disable_ipv6_mod, int, 0444);
  83 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  84
  85 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  86 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  87
  88 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  89 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  90
  91 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  92 {
  93         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  94
  95         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  96 }
  97
  98 static int inet6_create(struct net *net, struct socket *sock, int protocol,
  99                         int kern)
 100 {
 101         struct inet_sock *inet;
 102         struct ipv6_pinfo *np;
 103         struct sock *sk;
 104         struct inet_protosw *answer;
 105         struct proto *answer_prot;
 106         unsigned char answer_flags;
 107         char answer_no_check;
 108         int try_loading_module = 0;
 109         int err;
 110
 111         if (sock->type != SOCK_RAW &&
 112             sock->type != SOCK_DGRAM &&
 113             !inet_ehash_secret)
 114                 build_ehash_secret();
 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_no_check = answer->no_check;
 169         answer_flags = answer->flags;
 170         rcu_read_unlock();
 171
 172         WARN_ON(answer_prot->slab == NULL);
 173
 174         err = -ENOBUFS;
 175         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 176         if (sk == NULL)
 177                 goto out;
 178
 179         sock_init_data(sock, sk);
 180
 181         err = 0;
 182         sk->sk_no_check = answer_no_check;
 183         if (INET_PROTOSW_REUSE & answer_flags)
 184                 sk->sk_reuse = SK_CAN_REUSE;
 185
 186         inet = inet_sk(sk);
 187         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 188
 189         if (SOCK_RAW == sock->type) {
 190                 inet->inet_num = protocol;
 191                 if (IPPROTO_RAW == protocol)
 192                         inet->hdrincl = 1;
 193         }
 194
 195         sk->sk_destruct         = inet_sock_destruct;
 196         sk->sk_family           = PF_INET6;
 197         sk->sk_protocol         = protocol;
 198
 199         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 200
 201         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 202         np->hop_limit   = -1;
 203         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 204         np->mc_loop     = 1;
 205         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 206         np->ipv6only    = net->ipv6.sysctl.bindv6only;
 207
 208         /* Init the ipv4 part of the socket since we can have sockets
 209          * using v6 API for ipv4.
 210          */
 211         inet->uc_ttl    = -1;
 212
 213         inet->mc_loop   = 1;
 214         inet->mc_ttl    = 1;
 215         inet->mc_index  = 0;
 216         inet->mc_list   = NULL;
 217         inet->rcv_tos   = 0;
 218
 219         if (ipv4_config.no_pmtu_disc)
 220                 inet->pmtudisc = IP_PMTUDISC_DONT;
 221         else
 222                 inet->pmtudisc = IP_PMTUDISC_WANT;
 223         /*
 224          * Increment only the relevant sk_prot->socks debug field, this changes
 225          * the previous behaviour of incrementing both the equivalent to
 226          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 227          *
 228          * This allows better debug granularity as we'll know exactly how many
 229          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 230          * transport protocol socks. -acme
 231          */
 232         sk_refcnt_debug_inc(sk);
 233
 234         if (inet->inet_num) {
 235                 /* It assumes that any protocol which allows
 236                  * the user to assign a number at socket
 237                  * creation time automatically shares.
 238                  */
 239                 inet->inet_sport = htons(inet->inet_num);
 240                 sk->sk_prot->hash(sk);
 241         }
 242         if (sk->sk_prot->init) {
 243                 err = sk->sk_prot->init(sk);
 244                 if (err) {
 245                         sk_common_release(sk);
 246                         goto out;
 247                 }
 248         }
 249 out:
 250         return err;
 251 out_rcu_unlock:
 252         rcu_read_unlock();
 253         goto out;
 254 }
 255
 256
 257 /* bind for INET6 API */
 258 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 259 {
 260         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 261         struct sock *sk = sock->sk;
 262         struct inet_sock *inet = inet_sk(sk);
 263         struct ipv6_pinfo *np = inet6_sk(sk);
 264         struct net *net = sock_net(sk);
 265         __be32 v4addr = 0;
 266         unsigned short snum;
 267         int addr_type = 0;
 268         int err = 0;
 269
 270         /* If the socket has its own bind function then use it. */
 271         if (sk->sk_prot->bind)
 272                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 273
 274         if (addr_len < SIN6_LEN_RFC2133)
 275                 return -EINVAL;
 276
 277         if (addr->sin6_family != AF_INET6)
 278                 return -EAFNOSUPPORT;
 279
 280         addr_type = ipv6_addr_type(&addr->sin6_addr);
 281         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 282                 return -EINVAL;
 283
 284         snum = ntohs(addr->sin6_port);
 285         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 286                 return -EACCES;
 287
 288         lock_sock(sk);
 289
 290         /* Check these errors (active socket, double bind). */
 291         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 292                 err = -EINVAL;
 293                 goto out;
 294         }
 295
 296         /* Check if the address belongs to the host. */
 297         if (addr_type == IPV6_ADDR_MAPPED) {
 298                 int chk_addr_ret;
 299
 300                 /* Binding to v4-mapped address on a v6-only socket
 301                  * makes no sense
 302                  */
 303                 if (np->ipv6only) {
 304                         err = -EINVAL;
 305                         goto out;
 306                 }
 307
 308                 /* Reproduce AF_INET checks to make the bindings consistent */
 309                 v4addr = addr->sin6_addr.s6_addr32[3];
 310                 chk_addr_ret = inet_addr_type(net, v4addr);
 311                 if (!sysctl_ip_nonlocal_bind &&
 312                     !(inet->freebind || inet->transparent) &&
 313                     v4addr != htonl(INADDR_ANY) &&
 314                     chk_addr_ret != RTN_LOCAL &&
 315                     chk_addr_ret != RTN_MULTICAST &&
 316                     chk_addr_ret != RTN_BROADCAST) {
 317                         err = -EADDRNOTAVAIL;
 318                         goto out;
 319                 }
 320         } else {
 321                 if (addr_type != IPV6_ADDR_ANY) {
 322                         struct net_device *dev = NULL;
 323
 324                         rcu_read_lock();
 325                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 326                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 327                                     addr->sin6_scope_id) {
 328                                         /* Override any existing binding, if another one
 329                                          * is supplied by user.
 330                                          */
 331                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 332                                 }
 333
 334                                 /* Binding to link-local address requires an interface */
 335                                 if (!sk->sk_bound_dev_if) {
 336                                         err = -EINVAL;
 337                                         goto out_unlock;
 338                                 }
 339                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 340                                 if (!dev) {
 341                                         err = -ENODEV;
 342                                         goto out_unlock;
 343                                 }
 344                         }
 345
 346                         /* ipv4 addr of the socket is invalid.  Only the
 347                          * unspecified and mapped address have a v4 equivalent.
 348                          */
 349                         v4addr = LOOPBACK4_IPV6;
 350                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 351                                 if (!(inet->freebind || inet->transparent) &&
 352                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 353                                                    dev, 0)) {
 354                                         err = -EADDRNOTAVAIL;
 355                                         goto out_unlock;
 356                                 }
 357                         }
 358                         rcu_read_unlock();
 359                 }
 360         }
 361
 362         inet->inet_rcv_saddr = v4addr;
 363         inet->inet_saddr = v4addr;
 364
 365         np->rcv_saddr = addr->sin6_addr;
 366
 367         if (!(addr_type & IPV6_ADDR_MULTICAST))
 368                 np->saddr = addr->sin6_addr;
 369
 370         /* Make sure we are allowed to bind here. */
 371         if (sk->sk_prot->get_port(sk, snum)) {
 372                 inet_reset_saddr(sk);
 373                 err = -EADDRINUSE;
 374                 goto out;
 375         }
 376
 377         if (addr_type != IPV6_ADDR_ANY) {
 378                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 379                 if (addr_type != IPV6_ADDR_MAPPED)
 380                         np->ipv6only = 1;
 381         }
 382         if (snum)
 383                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 384         inet->inet_sport = htons(inet->inet_num);
 385         inet->inet_dport = 0;
 386         inet->inet_daddr = 0;
 387 out:
 388         release_sock(sk);
 389         return err;
 390 out_unlock:
 391         rcu_read_unlock();
 392         goto out;
 393 }
 394 EXPORT_SYMBOL(inet6_bind);
 395
 396 int inet6_release(struct socket *sock)
 397 {
 398         struct sock *sk = sock->sk;
 399
 400         if (sk == NULL)
 401                 return -EINVAL;
 402
 403         /* Free mc lists */
 404         ipv6_sock_mc_close(sk);
 405
 406         /* Free ac lists */
 407         ipv6_sock_ac_close(sk);
 408
 409         return inet_release(sock);
 410 }
 411 EXPORT_SYMBOL(inet6_release);
 412
 413 void inet6_destroy_sock(struct sock *sk)
 414 {
 415         struct ipv6_pinfo *np = inet6_sk(sk);
 416         struct sk_buff *skb;
 417         struct ipv6_txoptions *opt;
 418
 419         /* Release rx options */
 420
 421         skb = xchg(&np->pktoptions, NULL);
 422         if (skb != NULL)
 423                 kfree_skb(skb);
 424
 425         skb = xchg(&np->rxpmtu, NULL);
 426         if (skb != NULL)
 427                 kfree_skb(skb);
 428
 429         /* Free flowlabels */
 430         fl6_free_socklist(sk);
 431
 432         /* Free tx options */
 433
 434         opt = xchg(&np->opt, NULL);
 435         if (opt != NULL)
 436                 sock_kfree_s(sk, opt, opt->tot_len);
 437 }
 438 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 439
 440 /*
 441  *      This does both peername and sockname.
 442  */
 443
 444 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 445                  int *uaddr_len, int peer)
 446 {
 447         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 448         struct sock *sk = sock->sk;
 449         struct inet_sock *inet = inet_sk(sk);
 450         struct ipv6_pinfo *np = inet6_sk(sk);
 451
 452         sin->sin6_family = AF_INET6;
 453         sin->sin6_flowinfo = 0;
 454         sin->sin6_scope_id = 0;
 455         if (peer) {
 456                 if (!inet->inet_dport)
 457                         return -ENOTCONN;
 458                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 459                     peer == 1)
 460                         return -ENOTCONN;
 461                 sin->sin6_port = inet->inet_dport;
 462                 sin->sin6_addr = np->daddr;
 463                 if (np->sndflow)
 464                         sin->sin6_flowinfo = np->flow_label;
 465         } else {
 466                 if (ipv6_addr_any(&np->rcv_saddr))
 467                         sin->sin6_addr = np->saddr;
 468                 else
 469                         sin->sin6_addr = np->rcv_saddr;
 470
 471                 sin->sin6_port = inet->inet_sport;
 472         }
 473         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 474                                                  sk->sk_bound_dev_if);
 475         *uaddr_len = sizeof(*sin);
 476         return 0;
 477 }
 478 EXPORT_SYMBOL(inet6_getname);
 479
 480 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 481 {
 482         struct sock *sk = sock->sk;
 483         struct net *net = sock_net(sk);
 484
 485         switch (cmd) {
 486         case SIOCGSTAMP:
 487                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 488
 489         case SIOCGSTAMPNS:
 490                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 491
 492         case SIOCADDRT:
 493         case SIOCDELRT:
 494
 495                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 496
 497         case SIOCSIFADDR:
 498                 return addrconf_add_ifaddr(net, (void __user *) arg);
 499         case SIOCDIFADDR:
 500                 return addrconf_del_ifaddr(net, (void __user *) arg);
 501         case SIOCSIFDSTADDR:
 502                 return addrconf_set_dstaddr(net, (void __user *) arg);
 503         default:
 504                 if (!sk->sk_prot->ioctl)
 505                         return -ENOIOCTLCMD;
 506                 return sk->sk_prot->ioctl(sk, cmd, arg);
 507         }
 508         /*NOTREACHED*/
 509         return 0;
 510 }
 511 EXPORT_SYMBOL(inet6_ioctl);
 512
 513 const struct proto_ops inet6_stream_ops = {
 514         .family            = PF_INET6,
 515         .owner             = THIS_MODULE,
 516         .release           = inet6_release,
 517         .bind              = inet6_bind,
 518         .connect           = inet_stream_connect,       /* ok           */
 519         .socketpair        = sock_no_socketpair,        /* a do nothing */
 520         .accept            = inet_accept,               /* ok           */
 521         .getname           = inet6_getname,
 522         .poll              = tcp_poll,                  /* ok           */
 523         .ioctl             = inet6_ioctl,               /* must change  */
 524         .listen            = inet_listen,               /* ok           */
 525         .shutdown          = inet_shutdown,             /* ok           */
 526         .setsockopt        = sock_common_setsockopt,    /* ok           */
 527         .getsockopt        = sock_common_getsockopt,    /* ok           */
 528         .sendmsg           = inet_sendmsg,              /* ok           */
 529         .recvmsg           = inet_recvmsg,              /* ok           */
 530         .mmap              = sock_no_mmap,
 531         .sendpage          = inet_sendpage,
 532         .splice_read       = tcp_splice_read,
 533 #ifdef CONFIG_COMPAT
 534         .compat_setsockopt = compat_sock_common_setsockopt,
 535         .compat_getsockopt = compat_sock_common_getsockopt,
 536 #endif
 537 };
 538
 539 const struct proto_ops inet6_dgram_ops = {
 540         .family            = PF_INET6,
 541         .owner             = THIS_MODULE,
 542         .release           = inet6_release,
 543         .bind              = inet6_bind,
 544         .connect           = inet_dgram_connect,        /* ok           */
 545         .socketpair        = sock_no_socketpair,        /* a do nothing */
 546         .accept            = sock_no_accept,            /* a do nothing */
 547         .getname           = inet6_getname,
 548         .poll              = udp_poll,                  /* ok           */
 549         .ioctl             = inet6_ioctl,               /* must change  */
 550         .listen            = sock_no_listen,            /* ok           */
 551         .shutdown          = inet_shutdown,             /* ok           */
 552         .setsockopt        = sock_common_setsockopt,    /* ok           */
 553         .getsockopt        = sock_common_getsockopt,    /* ok           */
 554         .sendmsg           = inet_sendmsg,              /* ok           */
 555         .recvmsg           = inet_recvmsg,              /* ok           */
 556         .mmap              = sock_no_mmap,
 557         .sendpage          = sock_no_sendpage,
 558 #ifdef CONFIG_COMPAT
 559         .compat_setsockopt = compat_sock_common_setsockopt,
 560         .compat_getsockopt = compat_sock_common_getsockopt,
 561 #endif
 562 };
 563
 564 static const struct net_proto_family inet6_family_ops = {
 565         .family = PF_INET6,
 566         .create = inet6_create,
 567         .owner  = THIS_MODULE,
 568 };
 569
 570 int inet6_register_protosw(struct inet_protosw *p)
 571 {
 572         struct list_head *lh;
 573         struct inet_protosw *answer;
 574         struct list_head *last_perm;
 575         int protocol = p->protocol;
 576         int ret;
 577
 578         spin_lock_bh(&inetsw6_lock);
 579
 580         ret = -EINVAL;
 581         if (p->type >= SOCK_MAX)
 582                 goto out_illegal;
 583
 584         /* If we are trying to override a permanent protocol, bail. */
 585         answer = NULL;
 586         ret = -EPERM;
 587         last_perm = &inetsw6[p->type];
 588         list_for_each(lh, &inetsw6[p->type]) {
 589                 answer = list_entry(lh, struct inet_protosw, list);
 590
 591                 /* Check only the non-wild match. */
 592                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 593                         if (protocol == answer->protocol)
 594                                 break;
 595                         last_perm = lh;
 596                 }
 597
 598                 answer = NULL;
 599         }
 600         if (answer)
 601                 goto out_permanent;
 602
 603         /* Add the new entry after the last permanent entry if any, so that
 604          * the new entry does not override a permanent entry when matched with
 605          * a wild-card protocol. But it is allowed to override any existing
 606          * non-permanent entry.  This means that when we remove this entry, the
 607          * system automatically returns to the old behavior.
 608          */
 609         list_add_rcu(&p->list, last_perm);
 610         ret = 0;
 611 out:
 612         spin_unlock_bh(&inetsw6_lock);
 613         return ret;
 614
 615 out_permanent:
 616         pr_err("Attempt to override permanent protocol %d\n", protocol);
 617         goto out;
 618
 619 out_illegal:
 620         pr_err("Ignoring attempt to register invalid socket type %d\n",
 621                p->type);
 622         goto out;
 623 }
 624 EXPORT_SYMBOL(inet6_register_protosw);
 625
 626 void
 627 inet6_unregister_protosw(struct inet_protosw *p)
 628 {
 629         if (INET_PROTOSW_PERMANENT & p->flags) {
 630                 pr_err("Attempt to unregister permanent protocol %d\n",
 631                        p->protocol);
 632         } else {
 633                 spin_lock_bh(&inetsw6_lock);
 634                 list_del_rcu(&p->list);
 635                 spin_unlock_bh(&inetsw6_lock);
 636
 637                 synchronize_net();
 638         }
 639 }
 640 EXPORT_SYMBOL(inet6_unregister_protosw);
 641
 642 int inet6_sk_rebuild_header(struct sock *sk)
 643 {
 644         struct ipv6_pinfo *np = inet6_sk(sk);
 645         struct dst_entry *dst;
 646
 647         dst = __sk_dst_check(sk, np->dst_cookie);
 648
 649         if (dst == NULL) {
 650                 struct inet_sock *inet = inet_sk(sk);
 651                 struct in6_addr *final_p, final;
 652                 struct flowi6 fl6;
 653
 654                 memset(&fl6, 0, sizeof(fl6));
 655                 fl6.flowi6_proto = sk->sk_protocol;
 656                 fl6.daddr = np->daddr;
 657                 fl6.saddr = np->saddr;
 658                 fl6.flowlabel = np->flow_label;
 659                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 660                 fl6.flowi6_mark = sk->sk_mark;
 661                 fl6.fl6_dport = inet->inet_dport;
 662                 fl6.fl6_sport = inet->inet_sport;
 663                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 664
 665                 final_p = fl6_update_dst(&fl6, np->opt, &final);
 666
 667                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
 668                 if (IS_ERR(dst)) {
 669                         sk->sk_route_caps = 0;
 670                         sk->sk_err_soft = -PTR_ERR(dst);
 671                         return PTR_ERR(dst);
 672                 }
 673
 674                 __ip6_dst_store(sk, dst, NULL, NULL);
 675         }
 676
 677         return 0;
 678 }
 679 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 680
 681 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
 682 {
 683         const struct ipv6_pinfo *np = inet6_sk(sk);
 684         const struct inet6_skb_parm *opt = IP6CB(skb);
 685
 686         if (np->rxopt.all) {
 687                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 688                                   np->rxopt.bits.ohopopts)) ||
 689                     ((IPV6_FLOWINFO_MASK &
 690                       *(__be32 *)skb_network_header(skb)) &&
 691                      np->rxopt.bits.rxflow) ||
 692                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 693                      np->rxopt.bits.osrcrt)) ||
 694                     ((opt->dst1 || opt->dst0) &&
 695                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 696                         return true;
 697         }
 698         return false;
 699 }
 700 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 701
 702 static struct packet_type ipv6_packet_type __read_mostly = {
 703         .type = cpu_to_be16(ETH_P_IPV6),
 704         .func = ipv6_rcv,
 705 };
 706
 707 static int __init ipv6_packet_init(void)
 708 {
 709         dev_add_pack(&ipv6_packet_type);
 710         return 0;
 711 }
 712
 713 static void ipv6_packet_cleanup(void)
 714 {
 715         dev_remove_pack(&ipv6_packet_type);
 716 }
 717
 718 static int __net_init ipv6_init_mibs(struct net *net)
 719 {
 720         if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
 721                           sizeof(struct udp_mib),
 722                           __alignof__(struct udp_mib)) < 0)
 723                 return -ENOMEM;
 724         if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
 725                           sizeof(struct udp_mib),
 726                           __alignof__(struct udp_mib)) < 0)
 727                 goto err_udplite_mib;
 728         if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
 729                           sizeof(struct ipstats_mib),
 730                           __alignof__(struct ipstats_mib)) < 0)
 731                 goto err_ip_mib;
 732         if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
 733                           sizeof(struct icmpv6_mib),
 734                           __alignof__(struct icmpv6_mib)) < 0)
 735                 goto err_icmp_mib;
 736         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 737                                                 GFP_KERNEL);
 738         if (!net->mib.icmpv6msg_statistics)
 739                 goto err_icmpmsg_mib;
 740         return 0;
 741
 742 err_icmpmsg_mib:
 743         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 744 err_icmp_mib:
 745         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 746 err_ip_mib:
 747         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 748 err_udplite_mib:
 749         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 750         return -ENOMEM;
 751 }
 752
 753 static void ipv6_cleanup_mibs(struct net *net)
 754 {
 755         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 756         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 757         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 758         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 759         kfree(net->mib.icmpv6msg_statistics);
 760 }
 761
 762 static int __net_init inet6_net_init(struct net *net)
 763 {
 764         int err = 0;
 765
 766         net->ipv6.sysctl.bindv6only = 0;
 767         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 768
 769         err = ipv6_init_mibs(net);
 770         if (err)
 771                 return err;
 772 #ifdef CONFIG_PROC_FS
 773         err = udp6_proc_init(net);
 774         if (err)
 775                 goto out;
 776         err = tcp6_proc_init(net);
 777         if (err)
 778                 goto proc_tcp6_fail;
 779         err = ac6_proc_init(net);
 780         if (err)
 781                 goto proc_ac6_fail;
 782 #endif
 783         return err;
 784
 785 #ifdef CONFIG_PROC_FS
 786 proc_ac6_fail:
 787         tcp6_proc_exit(net);
 788 proc_tcp6_fail:
 789         udp6_proc_exit(net);
 790 out:
 791         ipv6_cleanup_mibs(net);
 792         return err;
 793 #endif
 794 }
 795
 796 static void __net_exit inet6_net_exit(struct net *net)
 797 {
 798 #ifdef CONFIG_PROC_FS
 799         udp6_proc_exit(net);
 800         tcp6_proc_exit(net);
 801         ac6_proc_exit(net);
 802 #endif
 803         ipv6_cleanup_mibs(net);
 804 }
 805
 806 static struct pernet_operations inet6_net_ops = {
 807         .init = inet6_net_init,
 808         .exit = inet6_net_exit,
 809 };
 810
 811 static int __init inet6_init(void)
 812 {
 813         struct list_head *r;
 814         int err = 0;
 815
 816         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
 817
 818         /* Register the socket-side information for inet6_create.  */
 819         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 820                 INIT_LIST_HEAD(r);
 821
 822         if (disable_ipv6_mod) {
 823                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 824                 goto out;
 825         }
 826
 827         err = proto_register(&tcpv6_prot, 1);
 828         if (err)
 829                 goto out;
 830
 831         err = proto_register(&udpv6_prot, 1);
 832         if (err)
 833                 goto out_unregister_tcp_proto;
 834
 835         err = proto_register(&udplitev6_prot, 1);
 836         if (err)
 837                 goto out_unregister_udp_proto;
 838
 839         err = proto_register(&rawv6_prot, 1);
 840         if (err)
 841                 goto out_unregister_udplite_proto;
 842
 843
 844         /* We MUST register RAW sockets before we create the ICMP6,
 845          * IGMP6, or NDISC control sockets.
 846          */
 847         err = rawv6_init();
 848         if (err)
 849                 goto out_unregister_raw_proto;
 850
 851         /* Register the family here so that the init calls below will
 852          * be able to create sockets. (?? is this dangerous ??)
 853          */
 854         err = sock_register(&inet6_family_ops);
 855         if (err)
 856                 goto out_sock_register_fail;
 857
 858         tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
 859
 860         /*
 861          *      ipngwg API draft makes clear that the correct semantics
 862          *      for TCP and UDP is to consider one TCP and UDP instance
 863          *      in a host available by both INET and INET6 APIs and
 864          *      able to communicate via both network protocols.
 865          */
 866
 867         err = register_pernet_subsys(&inet6_net_ops);
 868         if (err)
 869                 goto register_pernet_fail;
 870         err = icmpv6_init();
 871         if (err)
 872                 goto icmp_fail;
 873         err = ip6_mr_init();
 874         if (err)
 875                 goto ipmr_fail;
 876         err = ndisc_init();
 877         if (err)
 878                 goto ndisc_fail;
 879         err = igmp6_init();
 880         if (err)
 881                 goto igmp_fail;
 882         err = ipv6_netfilter_init();
 883         if (err)
 884                 goto netfilter_fail;
 885         /* Create /proc/foo6 entries. */
 886 #ifdef CONFIG_PROC_FS
 887         err = -ENOMEM;
 888         if (raw6_proc_init())
 889                 goto proc_raw6_fail;
 890         if (udplite6_proc_init())
 891                 goto proc_udplite6_fail;
 892         if (ipv6_misc_proc_init())
 893                 goto proc_misc6_fail;
 894         if (if6_proc_init())
 895                 goto proc_if6_fail;
 896 #endif
 897         err = ip6_route_init();
 898         if (err)
 899                 goto ip6_route_fail;
 900         err = ip6_flowlabel_init();
 901         if (err)
 902                 goto ip6_flowlabel_fail;
 903         err = addrconf_init();
 904         if (err)
 905                 goto addrconf_fail;
 906
 907         /* Init v6 extension headers. */
 908         err = ipv6_exthdrs_init();
 909         if (err)
 910                 goto ipv6_exthdrs_fail;
 911
 912         err = ipv6_frag_init();
 913         if (err)
 914                 goto ipv6_frag_fail;
 915
 916         /* Init v6 transport protocols. */
 917         err = udpv6_init();
 918         if (err)
 919                 goto udpv6_fail;
 920
 921         err = udplitev6_init();
 922         if (err)
 923                 goto udplitev6_fail;
 924
 925         err = tcpv6_init();
 926         if (err)
 927                 goto tcpv6_fail;
 928
 929         err = ipv6_packet_init();
 930         if (err)
 931                 goto ipv6_packet_fail;
 932
 933 #ifdef CONFIG_SYSCTL
 934         err = ipv6_sysctl_register();
 935         if (err)
 936                 goto sysctl_fail;
 937 #endif
 938 out:
 939         return err;
 940
 941 #ifdef CONFIG_SYSCTL
 942 sysctl_fail:
 943         ipv6_packet_cleanup();
 944 #endif
 945 ipv6_packet_fail:
 946         tcpv6_exit();
 947 tcpv6_fail:
 948         udplitev6_exit();
 949 udplitev6_fail:
 950         udpv6_exit();
 951 udpv6_fail:
 952         ipv6_frag_exit();
 953 ipv6_frag_fail:
 954         ipv6_exthdrs_exit();
 955 ipv6_exthdrs_fail:
 956         addrconf_cleanup();
 957 addrconf_fail:
 958         ip6_flowlabel_cleanup();
 959 ip6_flowlabel_fail:
 960         ip6_route_cleanup();
 961 ip6_route_fail:
 962 #ifdef CONFIG_PROC_FS
 963         if6_proc_exit();
 964 proc_if6_fail:
 965         ipv6_misc_proc_exit();
 966 proc_misc6_fail:
 967         udplite6_proc_exit();
 968 proc_udplite6_fail:
 969         raw6_proc_exit();
 970 proc_raw6_fail:
 971 #endif
 972         ipv6_netfilter_fini();
 973 netfilter_fail:
 974         igmp6_cleanup();
 975 igmp_fail:
 976         ndisc_cleanup();
 977 ndisc_fail:
 978         ip6_mr_cleanup();
 979 ipmr_fail:
 980         icmpv6_cleanup();
 981 icmp_fail:
 982         unregister_pernet_subsys(&inet6_net_ops);
 983 register_pernet_fail:
 984         sock_unregister(PF_INET6);
 985         rtnl_unregister_all(PF_INET6);
 986 out_sock_register_fail:
 987         rawv6_exit();
 988 out_unregister_raw_proto:
 989         proto_unregister(&rawv6_prot);
 990 out_unregister_udplite_proto:
 991         proto_unregister(&udplitev6_prot);
 992 out_unregister_udp_proto:
 993         proto_unregister(&udpv6_prot);
 994 out_unregister_tcp_proto:
 995         proto_unregister(&tcpv6_prot);
 996         goto out;
 997 }
 998 module_init(inet6_init);
 999
1000 static void __exit inet6_exit(void)
1001 {
1002         if (disable_ipv6_mod)
1003                 return;
1004
1005         /* First of all disallow new sockets creation. */
1006         sock_unregister(PF_INET6);
1007         /* Disallow any further netlink messages */
1008         rtnl_unregister_all(PF_INET6);
1009
1010         udpv6_exit();
1011         udplitev6_exit();
1012         tcpv6_exit();
1013
1014         /* Cleanup code parts. */
1015         ipv6_packet_cleanup();
1016         ipv6_frag_exit();
1017         ipv6_exthdrs_exit();
1018         addrconf_cleanup();
1019         ip6_flowlabel_cleanup();
1020         ip6_route_cleanup();
1021 #ifdef CONFIG_PROC_FS
1022
1023         /* Cleanup code parts. */
1024         if6_proc_exit();
1025         ipv6_misc_proc_exit();
1026         udplite6_proc_exit();
1027         raw6_proc_exit();
1028 #endif
1029         ipv6_netfilter_fini();
1030         igmp6_cleanup();
1031         ndisc_cleanup();
1032         ip6_mr_cleanup();
1033         icmpv6_cleanup();
1034         rawv6_exit();
1035
1036         unregister_pernet_subsys(&inet6_net_ops);
1037         proto_unregister(&rawv6_prot);
1038         proto_unregister(&udplitev6_prot);
1039         proto_unregister(&udpv6_prot);
1040         proto_unregister(&tcpv6_prot);
1041
1042         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1043 }
1044 module_exit(inet6_exit);
1045
1046 MODULE_ALIAS_NETPROTO(PF_INET6);