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 /* 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 void ipv6_frag_sysctl_init(struct net *net);
  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    = init_net.ipv6.sysctl.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(&init_net, 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(&init_net, &addr->sin6_addr,
 318                                                    dev, 0)) {
 319                                         if (dev)
 320                                                 dev_put(dev);
 321                                         err = -EADDRNOTAVAIL;
 322                                         goto out;
 323                                 }
 324                         }
 325                         if (dev)
 326                                 dev_put(dev);
 327                 }
 328         }
 329
 330         inet->rcv_saddr = v4addr;
 331         inet->saddr = v4addr;
 332
 333         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 334
 335         if (!(addr_type & IPV6_ADDR_MULTICAST))
 336                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 337
 338         /* Make sure we are allowed to bind here. */
 339         if (sk->sk_prot->get_port(sk, snum)) {
 340                 inet_reset_saddr(sk);
 341                 err = -EADDRINUSE;
 342                 goto out;
 343         }
 344
 345         if (addr_type != IPV6_ADDR_ANY)
 346                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 347         if (snum)
 348                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 349         inet->sport = htons(inet->num);
 350         inet->dport = 0;
 351         inet->daddr = 0;
 352 out:
 353         release_sock(sk);
 354         return err;
 355 }
 356
 357 EXPORT_SYMBOL(inet6_bind);
 358
 359 int inet6_release(struct socket *sock)
 360 {
 361         struct sock *sk = sock->sk;
 362
 363         if (sk == NULL)
 364                 return -EINVAL;
 365
 366         /* Free mc lists */
 367         ipv6_sock_mc_close(sk);
 368
 369         /* Free ac lists */
 370         ipv6_sock_ac_close(sk);
 371
 372         return inet_release(sock);
 373 }
 374
 375 EXPORT_SYMBOL(inet6_release);
 376
 377 int inet6_destroy_sock(struct sock *sk)
 378 {
 379         struct ipv6_pinfo *np = inet6_sk(sk);
 380         struct sk_buff *skb;
 381         struct ipv6_txoptions *opt;
 382
 383         /* Release rx options */
 384
 385         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 386                 kfree_skb(skb);
 387
 388         /* Free flowlabels */
 389         fl6_free_socklist(sk);
 390
 391         /* Free tx options */
 392
 393         if ((opt = xchg(&np->opt, NULL)) != NULL)
 394                 sock_kfree_s(sk, opt, opt->tot_len);
 395
 396         return 0;
 397 }
 398
 399 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 400
 401 /*
 402  *      This does both peername and sockname.
 403  */
 404
 405 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 406                  int *uaddr_len, int peer)
 407 {
 408         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 409         struct sock *sk = sock->sk;
 410         struct inet_sock *inet = inet_sk(sk);
 411         struct ipv6_pinfo *np = inet6_sk(sk);
 412
 413         sin->sin6_family = AF_INET6;
 414         sin->sin6_flowinfo = 0;
 415         sin->sin6_scope_id = 0;
 416         if (peer) {
 417                 if (!inet->dport)
 418                         return -ENOTCONN;
 419                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 420                     peer == 1)
 421                         return -ENOTCONN;
 422                 sin->sin6_port = inet->dport;
 423                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 424                 if (np->sndflow)
 425                         sin->sin6_flowinfo = np->flow_label;
 426         } else {
 427                 if (ipv6_addr_any(&np->rcv_saddr))
 428                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 429                 else
 430                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 431
 432                 sin->sin6_port = inet->sport;
 433         }
 434         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 435                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 436         *uaddr_len = sizeof(*sin);
 437         return(0);
 438 }
 439
 440 EXPORT_SYMBOL(inet6_getname);
 441
 442 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 443 {
 444         struct sock *sk = sock->sk;
 445
 446         switch(cmd)
 447         {
 448         case SIOCGSTAMP:
 449                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 450
 451         case SIOCGSTAMPNS:
 452                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 453
 454         case SIOCADDRT:
 455         case SIOCDELRT:
 456
 457                 return(ipv6_route_ioctl(cmd,(void __user *)arg));
 458
 459         case SIOCSIFADDR:
 460                 return addrconf_add_ifaddr((void __user *) arg);
 461         case SIOCDIFADDR:
 462                 return addrconf_del_ifaddr((void __user *) arg);
 463         case SIOCSIFDSTADDR:
 464                 return addrconf_set_dstaddr((void __user *) arg);
 465         default:
 466                 if (!sk->sk_prot->ioctl)
 467                         return -ENOIOCTLCMD;
 468                 return sk->sk_prot->ioctl(sk, cmd, arg);
 469         }
 470         /*NOTREACHED*/
 471         return(0);
 472 }
 473
 474 EXPORT_SYMBOL(inet6_ioctl);
 475
 476 const struct proto_ops inet6_stream_ops = {
 477         .family            = PF_INET6,
 478         .owner             = THIS_MODULE,
 479         .release           = inet6_release,
 480         .bind              = inet6_bind,
 481         .connect           = inet_stream_connect,       /* ok           */
 482         .socketpair        = sock_no_socketpair,        /* a do nothing */
 483         .accept            = inet_accept,               /* ok           */
 484         .getname           = inet6_getname,
 485         .poll              = tcp_poll,                  /* ok           */
 486         .ioctl             = inet6_ioctl,               /* must change  */
 487         .listen            = inet_listen,               /* ok           */
 488         .shutdown          = inet_shutdown,             /* ok           */
 489         .setsockopt        = sock_common_setsockopt,    /* ok           */
 490         .getsockopt        = sock_common_getsockopt,    /* ok           */
 491         .sendmsg           = tcp_sendmsg,               /* ok           */
 492         .recvmsg           = sock_common_recvmsg,       /* ok           */
 493         .mmap              = sock_no_mmap,
 494         .sendpage          = tcp_sendpage,
 495         .splice_read       = tcp_splice_read,
 496 #ifdef CONFIG_COMPAT
 497         .compat_setsockopt = compat_sock_common_setsockopt,
 498         .compat_getsockopt = compat_sock_common_getsockopt,
 499 #endif
 500 };
 501
 502 const struct proto_ops inet6_dgram_ops = {
 503         .family            = PF_INET6,
 504         .owner             = THIS_MODULE,
 505         .release           = inet6_release,
 506         .bind              = inet6_bind,
 507         .connect           = inet_dgram_connect,        /* ok           */
 508         .socketpair        = sock_no_socketpair,        /* a do nothing */
 509         .accept            = sock_no_accept,            /* a do nothing */
 510         .getname           = inet6_getname,
 511         .poll              = udp_poll,                  /* ok           */
 512         .ioctl             = inet6_ioctl,               /* must change  */
 513         .listen            = sock_no_listen,            /* ok           */
 514         .shutdown          = inet_shutdown,             /* ok           */
 515         .setsockopt        = sock_common_setsockopt,    /* ok           */
 516         .getsockopt        = sock_common_getsockopt,    /* ok           */
 517         .sendmsg           = inet_sendmsg,              /* ok           */
 518         .recvmsg           = sock_common_recvmsg,       /* ok           */
 519         .mmap              = sock_no_mmap,
 520         .sendpage          = sock_no_sendpage,
 521 #ifdef CONFIG_COMPAT
 522         .compat_setsockopt = compat_sock_common_setsockopt,
 523         .compat_getsockopt = compat_sock_common_getsockopt,
 524 #endif
 525 };
 526
 527 static struct net_proto_family inet6_family_ops = {
 528         .family = PF_INET6,
 529         .create = inet6_create,
 530         .owner  = THIS_MODULE,
 531 };
 532
 533 int inet6_register_protosw(struct inet_protosw *p)
 534 {
 535         struct list_head *lh;
 536         struct inet_protosw *answer;
 537         struct list_head *last_perm;
 538         int protocol = p->protocol;
 539         int ret;
 540
 541         spin_lock_bh(&inetsw6_lock);
 542
 543         ret = -EINVAL;
 544         if (p->type >= SOCK_MAX)
 545                 goto out_illegal;
 546
 547         /* If we are trying to override a permanent protocol, bail. */
 548         answer = NULL;
 549         ret = -EPERM;
 550         last_perm = &inetsw6[p->type];
 551         list_for_each(lh, &inetsw6[p->type]) {
 552                 answer = list_entry(lh, struct inet_protosw, list);
 553
 554                 /* Check only the non-wild match. */
 555                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 556                         if (protocol == answer->protocol)
 557                                 break;
 558                         last_perm = lh;
 559                 }
 560
 561                 answer = NULL;
 562         }
 563         if (answer)
 564                 goto out_permanent;
 565
 566         /* Add the new entry after the last permanent entry if any, so that
 567          * the new entry does not override a permanent entry when matched with
 568          * a wild-card protocol. But it is allowed to override any existing
 569          * non-permanent entry.  This means that when we remove this entry, the
 570          * system automatically returns to the old behavior.
 571          */
 572         list_add_rcu(&p->list, last_perm);
 573         ret = 0;
 574 out:
 575         spin_unlock_bh(&inetsw6_lock);
 576         return ret;
 577
 578 out_permanent:
 579         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 580                protocol);
 581         goto out;
 582
 583 out_illegal:
 584         printk(KERN_ERR
 585                "Ignoring attempt to register invalid socket type %d.\n",
 586                p->type);
 587         goto out;
 588 }
 589
 590 EXPORT_SYMBOL(inet6_register_protosw);
 591
 592 void
 593 inet6_unregister_protosw(struct inet_protosw *p)
 594 {
 595         if (INET_PROTOSW_PERMANENT & p->flags) {
 596                 printk(KERN_ERR
 597                        "Attempt to unregister permanent protocol %d.\n",
 598                        p->protocol);
 599         } else {
 600                 spin_lock_bh(&inetsw6_lock);
 601                 list_del_rcu(&p->list);
 602                 spin_unlock_bh(&inetsw6_lock);
 603
 604                 synchronize_net();
 605         }
 606 }
 607
 608 EXPORT_SYMBOL(inet6_unregister_protosw);
 609
 610 int inet6_sk_rebuild_header(struct sock *sk)
 611 {
 612         int err;
 613         struct dst_entry *dst;
 614         struct ipv6_pinfo *np = inet6_sk(sk);
 615
 616         dst = __sk_dst_check(sk, np->dst_cookie);
 617
 618         if (dst == NULL) {
 619                 struct inet_sock *inet = inet_sk(sk);
 620                 struct in6_addr *final_p = NULL, final;
 621                 struct flowi fl;
 622
 623                 memset(&fl, 0, sizeof(fl));
 624                 fl.proto = sk->sk_protocol;
 625                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 626                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 627                 fl.fl6_flowlabel = np->flow_label;
 628                 fl.oif = sk->sk_bound_dev_if;
 629                 fl.fl_ip_dport = inet->dport;
 630                 fl.fl_ip_sport = inet->sport;
 631                 security_sk_classify_flow(sk, &fl);
 632
 633                 if (np->opt && np->opt->srcrt) {
 634                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 635                         ipv6_addr_copy(&final, &fl.fl6_dst);
 636                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 637                         final_p = &final;
 638                 }
 639
 640                 err = ip6_dst_lookup(sk, &dst, &fl);
 641                 if (err) {
 642                         sk->sk_route_caps = 0;
 643                         return err;
 644                 }
 645                 if (final_p)
 646                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 647
 648                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
 649                         sk->sk_err_soft = -err;
 650                         return err;
 651                 }
 652
 653                 __ip6_dst_store(sk, dst, NULL, NULL);
 654         }
 655
 656         return 0;
 657 }
 658
 659 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 660
 661 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 662 {
 663         struct ipv6_pinfo *np = inet6_sk(sk);
 664         struct inet6_skb_parm *opt = IP6CB(skb);
 665
 666         if (np->rxopt.all) {
 667                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 668                                   np->rxopt.bits.ohopopts)) ||
 669                     ((IPV6_FLOWINFO_MASK &
 670                       *(__be32 *)skb_network_header(skb)) &&
 671                      np->rxopt.bits.rxflow) ||
 672                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 673                      np->rxopt.bits.osrcrt)) ||
 674                     ((opt->dst1 || opt->dst0) &&
 675                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 676                         return 1;
 677         }
 678         return 0;
 679 }
 680
 681 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 682
 683 static int __init init_ipv6_mibs(void)
 684 {
 685         if (snmp_mib_init((void **)ipv6_statistics,
 686                           sizeof(struct ipstats_mib)) < 0)
 687                 goto err_ip_mib;
 688         if (snmp_mib_init((void **)icmpv6_statistics,
 689                           sizeof(struct icmpv6_mib)) < 0)
 690                 goto err_icmp_mib;
 691         if (snmp_mib_init((void **)icmpv6msg_statistics,
 692                           sizeof(struct icmpv6msg_mib)) < 0)
 693                 goto err_icmpmsg_mib;
 694         if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib)) < 0)
 695                 goto err_udp_mib;
 696         if (snmp_mib_init((void **)udplite_stats_in6,
 697                           sizeof (struct udp_mib)) < 0)
 698                 goto err_udplite_mib;
 699         return 0;
 700
 701 err_udplite_mib:
 702         snmp_mib_free((void **)udp_stats_in6);
 703 err_udp_mib:
 704         snmp_mib_free((void **)icmpv6msg_statistics);
 705 err_icmpmsg_mib:
 706         snmp_mib_free((void **)icmpv6_statistics);
 707 err_icmp_mib:
 708         snmp_mib_free((void **)ipv6_statistics);
 709 err_ip_mib:
 710         return -ENOMEM;
 711
 712 }
 713
 714 static void cleanup_ipv6_mibs(void)
 715 {
 716         snmp_mib_free((void **)ipv6_statistics);
 717         snmp_mib_free((void **)icmpv6_statistics);
 718         snmp_mib_free((void **)icmpv6msg_statistics);
 719         snmp_mib_free((void **)udp_stats_in6);
 720         snmp_mib_free((void **)udplite_stats_in6);
 721 }
 722
 723 static int inet6_net_init(struct net *net)
 724 {
 725         net->ipv6.sysctl.bindv6only = 0;
 726         net->ipv6.sysctl.frags.high_thresh = 256 * 1024;
 727         net->ipv6.sysctl.frags.low_thresh = 192 * 1024;
 728         net->ipv6.sysctl.frags.timeout = IPV6_FRAG_TIMEOUT;
 729         net->ipv6.sysctl.frags.secret_interval = 10 * 60 * HZ;
 730         net->ipv6.sysctl.flush_delay = 0;
 731         net->ipv6.sysctl.ip6_rt_max_size = 4096;
 732         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
 733         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
 734         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
 735         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
 736         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
 737         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
 738         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 739         ipv6_frag_sysctl_init(net);
 740
 741         return 0;
 742 }
 743
 744 static void inet6_net_exit(struct net *net)
 745 {
 746         return;
 747 }
 748
 749 static struct pernet_operations inet6_net_ops = {
 750         .init = inet6_net_init,
 751         .exit = inet6_net_exit,
 752 };
 753
 754 static int __init inet6_init(void)
 755 {
 756         struct sk_buff *dummy_skb;
 757         struct list_head *r;
 758         int err;
 759
 760         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
 761
 762 #ifdef MODULE
 763 #if 0 /* FIXME --RR */
 764         if (!mod_member_present(&__this_module, can_unload))
 765           return -EINVAL;
 766
 767         __this_module.can_unload = &ipv6_unload;
 768 #endif
 769 #endif
 770         err = proto_register(&tcpv6_prot, 1);
 771         if (err)
 772                 goto out;
 773
 774         err = proto_register(&udpv6_prot, 1);
 775         if (err)
 776                 goto out_unregister_tcp_proto;
 777
 778         err = proto_register(&udplitev6_prot, 1);
 779         if (err)
 780                 goto out_unregister_udp_proto;
 781
 782         err = proto_register(&rawv6_prot, 1);
 783         if (err)
 784                 goto out_unregister_udplite_proto;
 785
 786
 787         /* Register the socket-side information for inet6_create.  */
 788         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 789                 INIT_LIST_HEAD(r);
 790
 791         /* We MUST register RAW sockets before we create the ICMP6,
 792          * IGMP6, or NDISC control sockets.
 793          */
 794         err = rawv6_init();
 795         if (err)
 796                 goto out_unregister_raw_proto;
 797
 798         /* Register the family here so that the init calls below will
 799          * be able to create sockets. (?? is this dangerous ??)
 800          */
 801         err = sock_register(&inet6_family_ops);
 802         if (err)
 803                 goto out_sock_register_fail;
 804
 805         /* Initialise ipv6 mibs */
 806         err = init_ipv6_mibs();
 807         if (err)
 808                 goto out_unregister_sock;
 809
 810         /*
 811          *      ipngwg API draft makes clear that the correct semantics
 812          *      for TCP and UDP is to consider one TCP and UDP instance
 813          *      in a host availiable by both INET and INET6 APIs and
 814          *      able to communicate via both network protocols.
 815          */
 816
 817         err = register_pernet_subsys(&inet6_net_ops);
 818         if (err)
 819                 goto register_pernet_fail;
 820
 821 #ifdef CONFIG_SYSCTL
 822         err = ipv6_sysctl_register();
 823         if (err)
 824                 goto sysctl_fail;
 825 #endif
 826         err = icmpv6_init(&inet6_family_ops);
 827         if (err)
 828                 goto icmp_fail;
 829         err = ndisc_init(&inet6_family_ops);
 830         if (err)
 831                 goto ndisc_fail;
 832         err = igmp6_init(&inet6_family_ops);
 833         if (err)
 834                 goto igmp_fail;
 835         err = ipv6_netfilter_init();
 836         if (err)
 837                 goto netfilter_fail;
 838         /* Create /proc/foo6 entries. */
 839 #ifdef CONFIG_PROC_FS
 840         err = -ENOMEM;
 841         if (raw6_proc_init())
 842                 goto proc_raw6_fail;
 843         if (tcp6_proc_init())
 844                 goto proc_tcp6_fail;
 845         if (udp6_proc_init())
 846                 goto proc_udp6_fail;
 847         if (udplite6_proc_init())
 848                 goto proc_udplite6_fail;
 849         if (ipv6_misc_proc_init())
 850                 goto proc_misc6_fail;
 851
 852         if (ac6_proc_init())
 853                 goto proc_anycast6_fail;
 854         if (if6_proc_init())
 855                 goto proc_if6_fail;
 856 #endif
 857         err = ip6_route_init();
 858         if (err)
 859                 goto ip6_route_fail;
 860         err = ip6_flowlabel_init();
 861         if (err)
 862                 goto ip6_flowlabel_fail;
 863         err = addrconf_init();
 864         if (err)
 865                 goto addrconf_fail;
 866
 867         /* Init v6 extension headers. */
 868         err = ipv6_exthdrs_init();
 869         if (err)
 870                 goto ipv6_exthdrs_fail;
 871
 872         err = ipv6_frag_init();
 873         if (err)
 874                 goto ipv6_frag_fail;
 875
 876         /* Init v6 transport protocols. */
 877         err = udpv6_init();
 878         if (err)
 879                 goto udpv6_fail;
 880
 881         err = udplitev6_init();
 882         if (err)
 883                 goto udplitev6_fail;
 884
 885         err = tcpv6_init();
 886         if (err)
 887                 goto tcpv6_fail;
 888
 889         err = ipv6_packet_init();
 890         if (err)
 891                 goto ipv6_packet_fail;
 892 out:
 893         return err;
 894
 895 ipv6_packet_fail:
 896         tcpv6_exit();
 897 tcpv6_fail:
 898         udplitev6_exit();
 899 udplitev6_fail:
 900         udpv6_exit();
 901 udpv6_fail:
 902         ipv6_frag_exit();
 903 ipv6_frag_fail:
 904         ipv6_exthdrs_exit();
 905 ipv6_exthdrs_fail:
 906         addrconf_cleanup();
 907 addrconf_fail:
 908         ip6_flowlabel_cleanup();
 909 ip6_flowlabel_fail:
 910         ip6_route_cleanup();
 911 ip6_route_fail:
 912 #ifdef CONFIG_PROC_FS
 913         if6_proc_exit();
 914 proc_if6_fail:
 915         ac6_proc_exit();
 916 proc_anycast6_fail:
 917         ipv6_misc_proc_exit();
 918 proc_misc6_fail:
 919         udplite6_proc_exit();
 920 proc_udplite6_fail:
 921         udp6_proc_exit();
 922 proc_udp6_fail:
 923         tcp6_proc_exit();
 924 proc_tcp6_fail:
 925         raw6_proc_exit();
 926 proc_raw6_fail:
 927 #endif
 928         ipv6_netfilter_fini();
 929 netfilter_fail:
 930         igmp6_cleanup();
 931 igmp_fail:
 932         ndisc_cleanup();
 933 ndisc_fail:
 934         icmpv6_cleanup();
 935 icmp_fail:
 936 #ifdef CONFIG_SYSCTL
 937         ipv6_sysctl_unregister();
 938 sysctl_fail:
 939 #endif
 940         unregister_pernet_subsys(&inet6_net_ops);
 941 register_pernet_fail:
 942         cleanup_ipv6_mibs();
 943 out_unregister_sock:
 944         sock_unregister(PF_INET6);
 945         rtnl_unregister_all(PF_INET6);
 946 out_sock_register_fail:
 947         rawv6_exit();
 948 out_unregister_raw_proto:
 949         proto_unregister(&rawv6_prot);
 950 out_unregister_udplite_proto:
 951         proto_unregister(&udplitev6_prot);
 952 out_unregister_udp_proto:
 953         proto_unregister(&udpv6_prot);
 954 out_unregister_tcp_proto:
 955         proto_unregister(&tcpv6_prot);
 956         goto out;
 957 }
 958 module_init(inet6_init);
 959
 960 static void __exit inet6_exit(void)
 961 {
 962         /* First of all disallow new sockets creation. */
 963         sock_unregister(PF_INET6);
 964         /* Disallow any further netlink messages */
 965         rtnl_unregister_all(PF_INET6);
 966
 967         udpv6_exit();
 968         udplitev6_exit();
 969         tcpv6_exit();
 970
 971         /* Cleanup code parts. */
 972         ipv6_packet_cleanup();
 973         ipv6_frag_exit();
 974         ipv6_exthdrs_exit();
 975         addrconf_cleanup();
 976         ip6_flowlabel_cleanup();
 977         ip6_route_cleanup();
 978 #ifdef CONFIG_PROC_FS
 979
 980         /* Cleanup code parts. */
 981         if6_proc_exit();
 982         ac6_proc_exit();
 983         ipv6_misc_proc_exit();
 984         udplite6_proc_exit();
 985         udp6_proc_exit();
 986         tcp6_proc_exit();
 987         raw6_proc_exit();
 988 #endif
 989         ipv6_netfilter_fini();
 990         igmp6_cleanup();
 991         ndisc_cleanup();
 992         icmpv6_cleanup();
 993         rawv6_exit();
 994 #ifdef CONFIG_SYSCTL
 995         ipv6_sysctl_unregister();
 996 #endif
 997         unregister_pernet_subsys(&inet6_net_ops);
 998         cleanup_ipv6_mibs();
 999         proto_unregister(&rawv6_prot);
1000         proto_unregister(&udplitev6_prot);
1001         proto_unregister(&udpv6_prot);
1002         proto_unregister(&tcpv6_prot);
1003 }
1004 module_exit(inet6_exit);
1005
1006 MODULE_ALIAS_NETPROTO(PF_INET6);