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  *      $Id: af_inet6.c,v 1.66 2002/02/01 22:01:04 davem Exp $
  11  *
  12  *      Fixes:
  13  *      piggy, Karl Knutson     :       Socket protocol table
  14  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
  15  *      Arnaldo Melo            :       check proc_net_create return, cleanups
  16  *
  17  *      This program is free software; you can redistribute it and/or
  18  *      modify it under the terms of the GNU General Public License
  19  *      as published by the Free Software Foundation; either version
  20  *      2 of the License, or (at your option) any later version.
  21  */
  22
  23
  24 #include <linux/module.h>
  25 #include <linux/capability.h>
  26 #include <linux/errno.h>
  27 #include <linux/types.h>
  28 #include <linux/socket.h>
  29 #include <linux/in.h>
  30 #include <linux/kernel.h>
  31 #include <linux/timer.h>
  32 #include <linux/string.h>
  33 #include <linux/sockios.h>
  34 #include <linux/net.h>
  35 #include <linux/fcntl.h>
  36 #include <linux/mm.h>
  37 #include <linux/interrupt.h>
  38 #include <linux/proc_fs.h>
  39 #include <linux/stat.h>
  40 #include <linux/init.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/ipip.h>
  53 #include <net/protocol.h>
  54 #include <net/inet_common.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 <asm/system.h>
  64
  65 MODULE_AUTHOR("Cast of dozens");
  66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  67 MODULE_LICENSE("GPL");
  68
  69 int sysctl_ipv6_bindv6only __read_mostly;
  70
  71 /* The inetsw6 table contains everything that inet6_create needs to
  72  * build a new socket.
  73  */
  74 static struct list_head inetsw6[SOCK_MAX];
  75 static DEFINE_SPINLOCK(inetsw6_lock);
  76
  77 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  78 {
  79         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  80
  81         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  82 }
  83
  84 static int inet6_create(struct net *net, struct socket *sock, int protocol)
  85 {
  86         struct inet_sock *inet;
  87         struct ipv6_pinfo *np;
  88         struct sock *sk;
  89         struct list_head *p;
  90         struct inet_protosw *answer;
  91         struct proto *answer_prot;
  92         unsigned char answer_flags;
  93         char answer_no_check;
  94         int try_loading_module = 0;
  95         int err;
  96
  97         if (net != &init_net)
  98                 return -EAFNOSUPPORT;
  99
 100         if (sock->type != SOCK_RAW &&
 101             sock->type != SOCK_DGRAM &&
 102             !inet_ehash_secret)
 103                 build_ehash_secret();
 104
 105         /* Look for the requested type/protocol pair. */
 106         answer = NULL;
 107 lookup_protocol:
 108         err = -ESOCKTNOSUPPORT;
 109         rcu_read_lock();
 110         list_for_each_rcu(p, &inetsw6[sock->type]) {
 111                 answer = list_entry(p, struct inet_protosw, list);
 112
 113                 /* Check the non-wild match. */
 114                 if (protocol == answer->protocol) {
 115                         if (protocol != IPPROTO_IP)
 116                                 break;
 117                 } else {
 118                         /* Check for the two wild cases. */
 119                         if (IPPROTO_IP == protocol) {
 120                                 protocol = answer->protocol;
 121                                 break;
 122                         }
 123                         if (IPPROTO_IP == answer->protocol)
 124                                 break;
 125                 }
 126                 err = -EPROTONOSUPPORT;
 127                 answer = NULL;
 128         }
 129
 130         if (!answer) {
 131                 if (try_loading_module < 2) {
 132                         rcu_read_unlock();
 133                         /*
 134                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 135                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 136                          */
 137                         if (++try_loading_module == 1)
 138                                 request_module("net-pf-%d-proto-%d-type-%d",
 139                                                 PF_INET6, protocol, sock->type);
 140                         /*
 141                          * Fall back to generic, e.g. net-pf-10-proto-132
 142                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 143                          */
 144                         else
 145                                 request_module("net-pf-%d-proto-%d",
 146                                                 PF_INET6, protocol);
 147                         goto lookup_protocol;
 148                 } else
 149                         goto out_rcu_unlock;
 150         }
 151
 152         err = -EPERM;
 153         if (answer->capability > 0 && !capable(answer->capability))
 154                 goto out_rcu_unlock;
 155
 156         sock->ops = answer->ops;
 157         answer_prot = answer->prot;
 158         answer_no_check = answer->no_check;
 159         answer_flags = answer->flags;
 160         rcu_read_unlock();
 161
 162         BUG_TRAP(answer_prot->slab != NULL);
 163
 164         err = -ENOBUFS;
 165         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 166         if (sk == NULL)
 167                 goto out;
 168
 169         sock_init_data(sock, sk);
 170
 171         err = 0;
 172         sk->sk_no_check = answer_no_check;
 173         if (INET_PROTOSW_REUSE & answer_flags)
 174                 sk->sk_reuse = 1;
 175
 176         inet = inet_sk(sk);
 177         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 178
 179         if (SOCK_RAW == sock->type) {
 180                 inet->num = protocol;
 181                 if (IPPROTO_RAW == protocol)
 182                         inet->hdrincl = 1;
 183         }
 184
 185         sk->sk_destruct         = inet_sock_destruct;
 186         sk->sk_family           = PF_INET6;
 187         sk->sk_protocol         = protocol;
 188
 189         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 190
 191         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 192         np->hop_limit   = -1;
 193         np->mcast_hops  = -1;
 194         np->mc_loop     = 1;
 195         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 196         np->ipv6only    = sysctl_ipv6_bindv6only;
 197
 198         /* Init the ipv4 part of the socket since we can have sockets
 199          * using v6 API for ipv4.
 200          */
 201         inet->uc_ttl    = -1;
 202
 203         inet->mc_loop   = 1;
 204         inet->mc_ttl    = 1;
 205         inet->mc_index  = 0;
 206         inet->mc_list   = NULL;
 207
 208         if (ipv4_config.no_pmtu_disc)
 209                 inet->pmtudisc = IP_PMTUDISC_DONT;
 210         else
 211                 inet->pmtudisc = IP_PMTUDISC_WANT;
 212         /*
 213          * Increment only the relevant sk_prot->socks debug field, this changes
 214          * the previous behaviour of incrementing both the equivalent to
 215          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 216          *
 217          * This allows better debug granularity as we'll know exactly how many
 218          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 219          * transport protocol socks. -acme
 220          */
 221         sk_refcnt_debug_inc(sk);
 222
 223         if (inet->num) {
 224                 /* It assumes that any protocol which allows
 225                  * the user to assign a number at socket
 226                  * creation time automatically shares.
 227                  */
 228                 inet->sport = htons(inet->num);
 229                 sk->sk_prot->hash(sk);
 230         }
 231         if (sk->sk_prot->init) {
 232                 err = sk->sk_prot->init(sk);
 233                 if (err) {
 234                         sk_common_release(sk);
 235                         goto out;
 236                 }
 237         }
 238 out:
 239         return err;
 240 out_rcu_unlock:
 241         rcu_read_unlock();
 242         goto out;
 243 }
 244
 245
 246 /* bind for INET6 API */
 247 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 248 {
 249         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 250         struct sock *sk = sock->sk;
 251         struct inet_sock *inet = inet_sk(sk);
 252         struct ipv6_pinfo *np = inet6_sk(sk);
 253         __be32 v4addr = 0;
 254         unsigned short snum;
 255         int addr_type = 0;
 256         int err = 0;
 257
 258         /* If the socket has its own bind function then use it. */
 259         if (sk->sk_prot->bind)
 260                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 261
 262         if (addr_len < SIN6_LEN_RFC2133)
 263                 return -EINVAL;
 264         addr_type = ipv6_addr_type(&addr->sin6_addr);
 265         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 266                 return -EINVAL;
 267
 268         snum = ntohs(addr->sin6_port);
 269         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 270                 return -EACCES;
 271
 272         lock_sock(sk);
 273
 274         /* Check these errors (active socket, double bind). */
 275         if (sk->sk_state != TCP_CLOSE || inet->num) {
 276                 err = -EINVAL;
 277                 goto out;
 278         }
 279
 280         /* Check if the address belongs to the host. */
 281         if (addr_type == IPV6_ADDR_MAPPED) {
 282                 v4addr = addr->sin6_addr.s6_addr32[3];
 283                 if (inet_addr_type(v4addr) != RTN_LOCAL) {
 284                         err = -EADDRNOTAVAIL;
 285                         goto out;
 286                 }
 287         } else {
 288                 if (addr_type != IPV6_ADDR_ANY) {
 289                         struct net_device *dev = NULL;
 290
 291                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 292                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 293                                     addr->sin6_scope_id) {
 294                                         /* Override any existing binding, if another one
 295                                          * is supplied by user.
 296                                          */
 297                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 298                                 }
 299
 300                                 /* Binding to link-local address requires an interface */
 301                                 if (!sk->sk_bound_dev_if) {
 302                                         err = -EINVAL;
 303                                         goto out;
 304                                 }
 305                                 dev = dev_get_by_index(&init_net, sk->sk_bound_dev_if);
 306                                 if (!dev) {
 307                                         err = -ENODEV;
 308                                         goto out;
 309                                 }
 310                         }
 311
 312                         /* ipv4 addr of the socket is invalid.  Only the
 313                          * unspecified and mapped address have a v4 equivalent.
 314                          */
 315                         v4addr = LOOPBACK4_IPV6;
 316                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 317                                 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
 318                                         if (dev)
 319                                                 dev_put(dev);
 320                                         err = -EADDRNOTAVAIL;
 321                                         goto out;
 322                                 }
 323                         }
 324                         if (dev)
 325                                 dev_put(dev);
 326                 }
 327         }
 328
 329         inet->rcv_saddr = v4addr;
 330         inet->saddr = v4addr;
 331
 332         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 333
 334         if (!(addr_type & IPV6_ADDR_MULTICAST))
 335                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 336
 337         /* Make sure we are allowed to bind here. */
 338         if (sk->sk_prot->get_port(sk, snum)) {
 339                 inet_reset_saddr(sk);
 340                 err = -EADDRINUSE;
 341                 goto out;
 342         }
 343
 344         if (addr_type != IPV6_ADDR_ANY)
 345                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 346         if (snum)
 347                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 348         inet->sport = htons(inet->num);
 349         inet->dport = 0;
 350         inet->daddr = 0;
 351 out:
 352         release_sock(sk);
 353         return err;
 354 }
 355
 356 EXPORT_SYMBOL(inet6_bind);
 357
 358 int inet6_release(struct socket *sock)
 359 {
 360         struct sock *sk = sock->sk;
 361
 362         if (sk == NULL)
 363                 return -EINVAL;
 364
 365         /* Free mc lists */
 366         ipv6_sock_mc_close(sk);
 367
 368         /* Free ac lists */
 369         ipv6_sock_ac_close(sk);
 370
 371         return inet_release(sock);
 372 }
 373
 374 EXPORT_SYMBOL(inet6_release);
 375
 376 int inet6_destroy_sock(struct sock *sk)
 377 {
 378         struct ipv6_pinfo *np = inet6_sk(sk);
 379         struct sk_buff *skb;
 380         struct ipv6_txoptions *opt;
 381
 382         /* Release rx options */
 383
 384         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 385                 kfree_skb(skb);
 386
 387         /* Free flowlabels */
 388         fl6_free_socklist(sk);
 389
 390         /* Free tx options */
 391
 392         if ((opt = xchg(&np->opt, NULL)) != NULL)
 393                 sock_kfree_s(sk, opt, opt->tot_len);
 394
 395         return 0;
 396 }
 397
 398 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 399
 400 /*
 401  *      This does both peername and sockname.
 402  */
 403
 404 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 405                  int *uaddr_len, int peer)
 406 {
 407         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 408         struct sock *sk = sock->sk;
 409         struct inet_sock *inet = inet_sk(sk);
 410         struct ipv6_pinfo *np = inet6_sk(sk);
 411
 412         sin->sin6_family = AF_INET6;
 413         sin->sin6_flowinfo = 0;
 414         sin->sin6_scope_id = 0;
 415         if (peer) {
 416                 if (!inet->dport)
 417                         return -ENOTCONN;
 418                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 419                     peer == 1)
 420                         return -ENOTCONN;
 421                 sin->sin6_port = inet->dport;
 422                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 423                 if (np->sndflow)
 424                         sin->sin6_flowinfo = np->flow_label;
 425         } else {
 426                 if (ipv6_addr_any(&np->rcv_saddr))
 427                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 428                 else
 429                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 430
 431                 sin->sin6_port = inet->sport;
 432         }
 433         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 434                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 435         *uaddr_len = sizeof(*sin);
 436         return(0);
 437 }
 438
 439 EXPORT_SYMBOL(inet6_getname);
 440
 441 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 442 {
 443         struct sock *sk = sock->sk;
 444
 445         switch(cmd)
 446         {
 447         case SIOCGSTAMP:
 448                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 449
 450         case SIOCGSTAMPNS:
 451                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 452
 453         case SIOCADDRT:
 454         case SIOCDELRT:
 455
 456                 return(ipv6_route_ioctl(cmd,(void __user *)arg));
 457
 458         case SIOCSIFADDR:
 459                 return addrconf_add_ifaddr((void __user *) arg);
 460         case SIOCDIFADDR:
 461                 return addrconf_del_ifaddr((void __user *) arg);
 462         case SIOCSIFDSTADDR:
 463                 return addrconf_set_dstaddr((void __user *) arg);
 464         default:
 465                 if (!sk->sk_prot->ioctl)
 466                         return -ENOIOCTLCMD;
 467                 return sk->sk_prot->ioctl(sk, cmd, arg);
 468         }
 469         /*NOTREACHED*/
 470         return(0);
 471 }
 472
 473 EXPORT_SYMBOL(inet6_ioctl);
 474
 475 const struct proto_ops inet6_stream_ops = {
 476         .family            = PF_INET6,
 477         .owner             = THIS_MODULE,
 478         .release           = inet6_release,
 479         .bind              = inet6_bind,
 480         .connect           = inet_stream_connect,       /* ok           */
 481         .socketpair        = sock_no_socketpair,        /* a do nothing */
 482         .accept            = inet_accept,               /* ok           */
 483         .getname           = inet6_getname,
 484         .poll              = tcp_poll,                  /* ok           */
 485         .ioctl             = inet6_ioctl,               /* must change  */
 486         .listen            = inet_listen,               /* ok           */
 487         .shutdown          = inet_shutdown,             /* ok           */
 488         .setsockopt        = sock_common_setsockopt,    /* ok           */
 489         .getsockopt        = sock_common_getsockopt,    /* ok           */
 490         .sendmsg           = tcp_sendmsg,               /* ok           */
 491         .recvmsg           = sock_common_recvmsg,       /* ok           */
 492         .mmap              = sock_no_mmap,
 493         .sendpage          = tcp_sendpage,
 494         .splice_read       = tcp_splice_read,
 495 #ifdef CONFIG_COMPAT
 496         .compat_setsockopt = compat_sock_common_setsockopt,
 497         .compat_getsockopt = compat_sock_common_getsockopt,
 498 #endif
 499 };
 500
 501 const struct proto_ops inet6_dgram_ops = {
 502         .family            = PF_INET6,
 503         .owner             = THIS_MODULE,
 504         .release           = inet6_release,
 505         .bind              = inet6_bind,
 506         .connect           = inet_dgram_connect,        /* ok           */
 507         .socketpair        = sock_no_socketpair,        /* a do nothing */
 508         .accept            = sock_no_accept,            /* a do nothing */
 509         .getname           = inet6_getname,
 510         .poll              = udp_poll,                  /* ok           */
 511         .ioctl             = inet6_ioctl,               /* must change  */
 512         .listen            = sock_no_listen,            /* ok           */
 513         .shutdown          = inet_shutdown,             /* ok           */
 514         .setsockopt        = sock_common_setsockopt,    /* ok           */
 515         .getsockopt        = sock_common_getsockopt,    /* ok           */
 516         .sendmsg           = inet_sendmsg,              /* ok           */
 517         .recvmsg           = sock_common_recvmsg,       /* ok           */
 518         .mmap              = sock_no_mmap,
 519         .sendpage          = sock_no_sendpage,
 520 #ifdef CONFIG_COMPAT
 521         .compat_setsockopt = compat_sock_common_setsockopt,
 522         .compat_getsockopt = compat_sock_common_getsockopt,
 523 #endif
 524 };
 525
 526 static struct net_proto_family inet6_family_ops = {
 527         .family = PF_INET6,
 528         .create = inet6_create,
 529         .owner  = THIS_MODULE,
 530 };
 531
 532 /* Same as inet6_dgram_ops, sans udp_poll.  */
 533 static const struct proto_ops inet6_sockraw_ops = {
 534         .family            = PF_INET6,
 535         .owner             = THIS_MODULE,
 536         .release           = inet6_release,
 537         .bind              = inet6_bind,
 538         .connect           = inet_dgram_connect,        /* ok           */
 539         .socketpair        = sock_no_socketpair,        /* a do nothing */
 540         .accept            = sock_no_accept,            /* a do nothing */
 541         .getname           = inet6_getname,
 542         .poll              = datagram_poll,             /* ok           */
 543         .ioctl             = inet6_ioctl,               /* must change  */
 544         .listen            = sock_no_listen,            /* ok           */
 545         .shutdown          = inet_shutdown,             /* ok           */
 546         .setsockopt        = sock_common_setsockopt,    /* ok           */
 547         .getsockopt        = sock_common_getsockopt,    /* ok           */
 548         .sendmsg           = inet_sendmsg,              /* ok           */
 549         .recvmsg           = sock_common_recvmsg,       /* ok           */
 550         .mmap              = sock_no_mmap,
 551         .sendpage          = sock_no_sendpage,
 552 #ifdef CONFIG_COMPAT
 553         .compat_setsockopt = compat_sock_common_setsockopt,
 554         .compat_getsockopt = compat_sock_common_getsockopt,
 555 #endif
 556 };
 557
 558 static struct inet_protosw rawv6_protosw = {
 559         .type           = SOCK_RAW,
 560         .protocol       = IPPROTO_IP,   /* wild card */
 561         .prot           = &rawv6_prot,
 562         .ops            = &inet6_sockraw_ops,
 563         .capability     = CAP_NET_RAW,
 564         .no_check       = UDP_CSUM_DEFAULT,
 565         .flags          = INET_PROTOSW_REUSE,
 566 };
 567
 568 int inet6_register_protosw(struct inet_protosw *p)
 569 {
 570         struct list_head *lh;
 571         struct inet_protosw *answer;
 572         struct list_head *last_perm;
 573         int protocol = p->protocol;
 574         int ret;
 575
 576         spin_lock_bh(&inetsw6_lock);
 577
 578         ret = -EINVAL;
 579         if (p->type >= SOCK_MAX)
 580                 goto out_illegal;
 581
 582         /* If we are trying to override a permanent protocol, bail. */
 583         answer = NULL;
 584         ret = -EPERM;
 585         last_perm = &inetsw6[p->type];
 586         list_for_each(lh, &inetsw6[p->type]) {
 587                 answer = list_entry(lh, struct inet_protosw, list);
 588
 589                 /* Check only the non-wild match. */
 590                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 591                         if (protocol == answer->protocol)
 592                                 break;
 593                         last_perm = lh;
 594                 }
 595
 596                 answer = NULL;
 597         }
 598         if (answer)
 599                 goto out_permanent;
 600
 601         /* Add the new entry after the last permanent entry if any, so that
 602          * the new entry does not override a permanent entry when matched with
 603          * a wild-card protocol. But it is allowed to override any existing
 604          * non-permanent entry.  This means that when we remove this entry, the
 605          * system automatically returns to the old behavior.
 606          */
 607         list_add_rcu(&p->list, last_perm);
 608         ret = 0;
 609 out:
 610         spin_unlock_bh(&inetsw6_lock);
 611         return ret;
 612
 613 out_permanent:
 614         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 615                protocol);
 616         goto out;
 617
 618 out_illegal:
 619         printk(KERN_ERR
 620                "Ignoring attempt to register invalid socket type %d.\n",
 621                p->type);
 622         goto out;
 623 }
 624
 625 EXPORT_SYMBOL(inet6_register_protosw);
 626
 627 void
 628 inet6_unregister_protosw(struct inet_protosw *p)
 629 {
 630         if (INET_PROTOSW_PERMANENT & p->flags) {
 631                 printk(KERN_ERR
 632                        "Attempt to unregister permanent protocol %d.\n",
 633                        p->protocol);
 634         } else {
 635                 spin_lock_bh(&inetsw6_lock);
 636                 list_del_rcu(&p->list);
 637                 spin_unlock_bh(&inetsw6_lock);
 638
 639                 synchronize_net();
 640         }
 641 }
 642
 643 EXPORT_SYMBOL(inet6_unregister_protosw);
 644
 645 int inet6_sk_rebuild_header(struct sock *sk)
 646 {
 647         int err;
 648         struct dst_entry *dst;
 649         struct ipv6_pinfo *np = inet6_sk(sk);
 650
 651         dst = __sk_dst_check(sk, np->dst_cookie);
 652
 653         if (dst == NULL) {
 654                 struct inet_sock *inet = inet_sk(sk);
 655                 struct in6_addr *final_p = NULL, final;
 656                 struct flowi fl;
 657
 658                 memset(&fl, 0, sizeof(fl));
 659                 fl.proto = sk->sk_protocol;
 660                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 661                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 662                 fl.fl6_flowlabel = np->flow_label;
 663                 fl.oif = sk->sk_bound_dev_if;
 664                 fl.fl_ip_dport = inet->dport;
 665                 fl.fl_ip_sport = inet->sport;
 666                 security_sk_classify_flow(sk, &fl);
 667
 668                 if (np->opt && np->opt->srcrt) {
 669                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 670                         ipv6_addr_copy(&final, &fl.fl6_dst);
 671                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 672                         final_p = &final;
 673                 }
 674
 675                 err = ip6_dst_lookup(sk, &dst, &fl);
 676                 if (err) {
 677                         sk->sk_route_caps = 0;
 678                         return err;
 679                 }
 680                 if (final_p)
 681                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 682
 683                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
 684                         sk->sk_err_soft = -err;
 685                         return err;
 686                 }
 687
 688                 __ip6_dst_store(sk, dst, NULL, NULL);
 689         }
 690
 691         return 0;
 692 }
 693
 694 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 695
 696 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 697 {
 698         struct ipv6_pinfo *np = inet6_sk(sk);
 699         struct inet6_skb_parm *opt = IP6CB(skb);
 700
 701         if (np->rxopt.all) {
 702                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 703                                   np->rxopt.bits.ohopopts)) ||
 704                     ((IPV6_FLOWINFO_MASK &
 705                       *(__be32 *)skb_network_header(skb)) &&
 706                      np->rxopt.bits.rxflow) ||
 707                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 708                      np->rxopt.bits.osrcrt)) ||
 709                     ((opt->dst1 || opt->dst0) &&
 710                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 711                         return 1;
 712         }
 713         return 0;
 714 }
 715
 716 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 717
 718 static int __init init_ipv6_mibs(void)
 719 {
 720         if (snmp_mib_init((void **)ipv6_statistics,
 721                           sizeof(struct ipstats_mib)) < 0)
 722                 goto err_ip_mib;
 723         if (snmp_mib_init((void **)icmpv6_statistics,
 724                           sizeof(struct icmpv6_mib)) < 0)
 725                 goto err_icmp_mib;
 726         if (snmp_mib_init((void **)icmpv6msg_statistics,
 727                           sizeof(struct icmpv6msg_mib)) < 0)
 728                 goto err_icmpmsg_mib;
 729         if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib)) < 0)
 730                 goto err_udp_mib;
 731         if (snmp_mib_init((void **)udplite_stats_in6,
 732                           sizeof (struct udp_mib)) < 0)
 733                 goto err_udplite_mib;
 734         return 0;
 735
 736 err_udplite_mib:
 737         snmp_mib_free((void **)udp_stats_in6);
 738 err_udp_mib:
 739         snmp_mib_free((void **)icmpv6msg_statistics);
 740 err_icmpmsg_mib:
 741         snmp_mib_free((void **)icmpv6_statistics);
 742 err_icmp_mib:
 743         snmp_mib_free((void **)ipv6_statistics);
 744 err_ip_mib:
 745         return -ENOMEM;
 746
 747 }
 748
 749 static void cleanup_ipv6_mibs(void)
 750 {
 751         snmp_mib_free((void **)ipv6_statistics);
 752         snmp_mib_free((void **)icmpv6_statistics);
 753         snmp_mib_free((void **)icmpv6msg_statistics);
 754         snmp_mib_free((void **)udp_stats_in6);
 755         snmp_mib_free((void **)udplite_stats_in6);
 756 }
 757
 758 static int __init inet6_init(void)
 759 {
 760         struct sk_buff *dummy_skb;
 761         struct list_head *r;
 762         int err;
 763
 764         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
 765
 766 #ifdef MODULE
 767 #if 0 /* FIXME --RR */
 768         if (!mod_member_present(&__this_module, can_unload))
 769           return -EINVAL;
 770
 771         __this_module.can_unload = &ipv6_unload;
 772 #endif
 773 #endif
 774
 775         err = proto_register(&tcpv6_prot, 1);
 776         if (err)
 777                 goto out;
 778
 779         err = proto_register(&udpv6_prot, 1);
 780         if (err)
 781                 goto out_unregister_tcp_proto;
 782
 783         err = proto_register(&udplitev6_prot, 1);
 784         if (err)
 785                 goto out_unregister_udp_proto;
 786
 787         err = proto_register(&rawv6_prot, 1);
 788         if (err)
 789                 goto out_unregister_udplite_proto;
 790
 791
 792         /* Register the socket-side information for inet6_create.  */
 793         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 794                 INIT_LIST_HEAD(r);
 795
 796         /* We MUST register RAW sockets before we create the ICMP6,
 797          * IGMP6, or NDISC control sockets.
 798          */
 799         inet6_register_protosw(&rawv6_protosw);
 800
 801         /* Register the family here so that the init calls below will
 802          * be able to create sockets. (?? is this dangerous ??)
 803          */
 804         err = sock_register(&inet6_family_ops);
 805         if (err)
 806                 goto out_unregister_raw_proto;
 807
 808         /* Initialise ipv6 mibs */
 809         err = init_ipv6_mibs();
 810         if (err)
 811                 goto out_unregister_sock;
 812
 813         /*
 814          *      ipngwg API draft makes clear that the correct semantics
 815          *      for TCP and UDP is to consider one TCP and UDP instance
 816          *      in a host availiable by both INET and INET6 APIs and
 817          *      able to communicate via both network protocols.
 818          */
 819
 820 #ifdef CONFIG_SYSCTL
 821         ipv6_sysctl_register();
 822 #endif
 823         err = icmpv6_init(&inet6_family_ops);
 824         if (err)
 825                 goto icmp_fail;
 826         err = ndisc_init(&inet6_family_ops);
 827         if (err)
 828                 goto ndisc_fail;
 829         err = igmp6_init(&inet6_family_ops);
 830         if (err)
 831                 goto igmp_fail;
 832         err = ipv6_netfilter_init();
 833         if (err)
 834                 goto netfilter_fail;
 835         /* Create /proc/foo6 entries. */
 836 #ifdef CONFIG_PROC_FS
 837         err = -ENOMEM;
 838         if (raw6_proc_init())
 839                 goto proc_raw6_fail;
 840         if (tcp6_proc_init())
 841                 goto proc_tcp6_fail;
 842         if (udp6_proc_init())
 843                 goto proc_udp6_fail;
 844         if (udplite6_proc_init())
 845                 goto proc_udplite6_fail;
 846         if (ipv6_misc_proc_init())
 847                 goto proc_misc6_fail;
 848
 849         if (ac6_proc_init())
 850                 goto proc_anycast6_fail;
 851         if (if6_proc_init())
 852                 goto proc_if6_fail;
 853 #endif
 854         err = ip6_route_init();
 855         if (err)
 856                 goto ip6_route_fail;
 857         err = ip6_flowlabel_init();
 858         if (err)
 859                 goto ip6_flowlabel_fail;
 860         err = addrconf_init();
 861         if (err)
 862                 goto addrconf_fail;
 863
 864         /* Init v6 extension headers. */
 865         err = ipv6_exthdrs_init();
 866         if (err)
 867                 goto ipv6_exthdrs_fail;
 868
 869         err = ipv6_frag_init();
 870         if (err)
 871                 goto ipv6_frag_fail;
 872
 873         /* Init v6 transport protocols. */
 874         udpv6_init();
 875         udplitev6_init();
 876         tcpv6_init();
 877
 878         ipv6_packet_init();
 879         err = 0;
 880 out:
 881         return err;
 882
 883 ipv6_frag_fail:
 884         ipv6_exthdrs_exit();
 885 ipv6_exthdrs_fail:
 886         addrconf_cleanup();
 887 addrconf_fail:
 888         ip6_flowlabel_cleanup();
 889 ip6_flowlabel_fail:
 890         ip6_route_cleanup();
 891 ip6_route_fail:
 892 #ifdef CONFIG_PROC_FS
 893         if6_proc_exit();
 894 proc_if6_fail:
 895         ac6_proc_exit();
 896 proc_anycast6_fail:
 897         ipv6_misc_proc_exit();
 898 proc_misc6_fail:
 899         udplite6_proc_exit();
 900 proc_udplite6_fail:
 901         udp6_proc_exit();
 902 proc_udp6_fail:
 903         tcp6_proc_exit();
 904 proc_tcp6_fail:
 905         raw6_proc_exit();
 906 proc_raw6_fail:
 907 #endif
 908         ipv6_netfilter_fini();
 909 netfilter_fail:
 910         igmp6_cleanup();
 911 igmp_fail:
 912         ndisc_cleanup();
 913 ndisc_fail:
 914         icmpv6_cleanup();
 915 icmp_fail:
 916 #ifdef CONFIG_SYSCTL
 917         ipv6_sysctl_unregister();
 918 #endif
 919         cleanup_ipv6_mibs();
 920 out_unregister_sock:
 921         sock_unregister(PF_INET6);
 922         rtnl_unregister_all(PF_INET6);
 923 out_unregister_raw_proto:
 924         proto_unregister(&rawv6_prot);
 925 out_unregister_udplite_proto:
 926         proto_unregister(&udplitev6_prot);
 927 out_unregister_udp_proto:
 928         proto_unregister(&udpv6_prot);
 929 out_unregister_tcp_proto:
 930         proto_unregister(&tcpv6_prot);
 931         goto out;
 932 }
 933 module_init(inet6_init);
 934
 935 static void __exit inet6_exit(void)
 936 {
 937         /* First of all disallow new sockets creation. */
 938         sock_unregister(PF_INET6);
 939         /* Disallow any further netlink messages */
 940         rtnl_unregister_all(PF_INET6);
 941
 942         /* Cleanup code parts. */
 943         ipv6_packet_cleanup();
 944         ipv6_frag_exit();
 945         ipv6_exthdrs_exit();
 946         addrconf_cleanup();
 947         ip6_flowlabel_cleanup();
 948         ip6_route_cleanup();
 949 #ifdef CONFIG_PROC_FS
 950
 951         /* Cleanup code parts. */
 952         if6_proc_exit();
 953         ac6_proc_exit();
 954         ipv6_misc_proc_exit();
 955         udplite6_proc_exit();
 956         udp6_proc_exit();
 957         tcp6_proc_exit();
 958         raw6_proc_exit();
 959 #endif
 960         ipv6_netfilter_fini();
 961         igmp6_cleanup();
 962         ndisc_cleanup();
 963         icmpv6_cleanup();
 964 #ifdef CONFIG_SYSCTL
 965         ipv6_sysctl_unregister();
 966 #endif
 967         cleanup_ipv6_mibs();
 968         proto_unregister(&rawv6_prot);
 969         proto_unregister(&udplitev6_prot);
 970         proto_unregister(&udpv6_prot);
 971         proto_unregister(&tcpv6_prot);
 972 }
 973 module_exit(inet6_exit);
 974
 975 MODULE_ALIAS_NETPROTO(PF_INET6);