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/sched.h>
  32 #include <linux/timer.h>
  33 #include <linux/string.h>
  34 #include <linux/sockios.h>
  35 #include <linux/net.h>
  36 #include <linux/fcntl.h>
  37 #include <linux/mm.h>
  38 #include <linux/interrupt.h>
  39 #include <linux/proc_fs.h>
  40 #include <linux/stat.h>
  41 #include <linux/init.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <linux/icmpv6.h>
  46 #include <linux/smp_lock.h>
  47 #include <linux/netfilter_ipv6.h>
  48
  49 #include <net/ip.h>
  50 #include <net/ipv6.h>
  51 #include <net/udp.h>
  52 #include <net/tcp.h>
  53 #include <net/ipip.h>
  54 #include <net/protocol.h>
  55 #include <net/inet_common.h>
  56 #include <net/transp_v6.h>
  57 #include <net/ip6_route.h>
  58 #include <net/addrconf.h>
  59 #ifdef CONFIG_IPV6_TUNNEL
  60 #include <net/ip6_tunnel.h>
  61 #endif
  62
  63 #include <asm/uaccess.h>
  64 #include <asm/system.h>
  65
  66 MODULE_AUTHOR("Cast of dozens");
  67 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  68 MODULE_LICENSE("GPL");
  69
  70 int sysctl_ipv6_bindv6only;
  71
  72 /* The inetsw table contains everything that inet_create needs to
  73  * build a new socket.
  74  */
  75 static struct list_head inetsw6[SOCK_MAX];
  76 static DEFINE_SPINLOCK(inetsw6_lock);
  77
  78 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  79 {
  80         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  81
  82         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  83 }
  84
  85 static int inet6_create(struct socket *sock, int protocol)
  86 {
  87         struct inet_sock *inet;
  88         struct ipv6_pinfo *np;
  89         struct sock *sk;
  90         struct list_head *p;
  91         struct inet_protosw *answer;
  92         struct proto *answer_prot;
  93         unsigned char answer_flags;
  94         char answer_no_check;
  95         int try_loading_module = 0;
  96         int err;
  97
  98         /* Look for the requested type/protocol pair. */
  99         answer = NULL;
 100 lookup_protocol:
 101         err = -ESOCKTNOSUPPORT;
 102         rcu_read_lock();
 103         list_for_each_rcu(p, &inetsw6[sock->type]) {
 104                 answer = list_entry(p, struct inet_protosw, list);
 105
 106                 /* Check the non-wild match. */
 107                 if (protocol == answer->protocol) {
 108                         if (protocol != IPPROTO_IP)
 109                                 break;
 110                 } else {
 111                         /* Check for the two wild cases. */
 112                         if (IPPROTO_IP == protocol) {
 113                                 protocol = answer->protocol;
 114                                 break;
 115                         }
 116                         if (IPPROTO_IP == answer->protocol)
 117                                 break;
 118                 }
 119                 err = -EPROTONOSUPPORT;
 120                 answer = NULL;
 121         }
 122
 123         if (!answer) {
 124                 if (try_loading_module < 2) {
 125                         rcu_read_unlock();
 126                         /*
 127                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 128                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 129                          */
 130                         if (++try_loading_module == 1)
 131                                 request_module("net-pf-%d-proto-%d-type-%d",
 132                                                 PF_INET6, protocol, sock->type);
 133                         /*
 134                          * Fall back to generic, e.g. net-pf-10-proto-132
 135                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 136                          */
 137                         else
 138                                 request_module("net-pf-%d-proto-%d",
 139                                                 PF_INET6, protocol);
 140                         goto lookup_protocol;
 141                 } else
 142                         goto out_rcu_unlock;
 143         }
 144
 145         err = -EPERM;
 146         if (answer->capability > 0 && !capable(answer->capability))
 147                 goto out_rcu_unlock;
 148
 149         sock->ops = answer->ops;
 150         answer_prot = answer->prot;
 151         answer_no_check = answer->no_check;
 152         answer_flags = answer->flags;
 153         rcu_read_unlock();
 154
 155         BUG_TRAP(answer_prot->slab != NULL);
 156
 157         err = -ENOBUFS;
 158         sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
 159         if (sk == NULL)
 160                 goto out;
 161
 162         sock_init_data(sock, sk);
 163
 164         err = 0;
 165         sk->sk_no_check = answer_no_check;
 166         if (INET_PROTOSW_REUSE & answer_flags)
 167                 sk->sk_reuse = 1;
 168
 169         inet = inet_sk(sk);
 170         inet->is_icsk = INET_PROTOSW_ICSK & answer_flags;
 171
 172         if (SOCK_RAW == sock->type) {
 173                 inet->num = protocol;
 174                 if (IPPROTO_RAW == protocol)
 175                         inet->hdrincl = 1;
 176         }
 177
 178         sk->sk_destruct         = inet_sock_destruct;
 179         sk->sk_family           = PF_INET6;
 180         sk->sk_protocol         = protocol;
 181
 182         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 183
 184         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 185         np->hop_limit   = -1;
 186         np->mcast_hops  = -1;
 187         np->mc_loop     = 1;
 188         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 189         np->ipv6only    = sysctl_ipv6_bindv6only;
 190
 191         /* Init the ipv4 part of the socket since we can have sockets
 192          * using v6 API for ipv4.
 193          */
 194         inet->uc_ttl    = -1;
 195
 196         inet->mc_loop   = 1;
 197         inet->mc_ttl    = 1;
 198         inet->mc_index  = 0;
 199         inet->mc_list   = NULL;
 200
 201         if (ipv4_config.no_pmtu_disc)
 202                 inet->pmtudisc = IP_PMTUDISC_DONT;
 203         else
 204                 inet->pmtudisc = IP_PMTUDISC_WANT;
 205         /*
 206          * Increment only the relevant sk_prot->socks debug field, this changes
 207          * the previous behaviour of incrementing both the equivalent to
 208          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 209          *
 210          * This allows better debug granularity as we'll know exactly how many
 211          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 212          * transport protocol socks. -acme
 213          */
 214         sk_refcnt_debug_inc(sk);
 215
 216         if (inet->num) {
 217                 /* It assumes that any protocol which allows
 218                  * the user to assign a number at socket
 219                  * creation time automatically shares.
 220                  */
 221                 inet->sport = ntohs(inet->num);
 222                 sk->sk_prot->hash(sk);
 223         }
 224         if (sk->sk_prot->init) {
 225                 err = sk->sk_prot->init(sk);
 226                 if (err) {
 227                         sk_common_release(sk);
 228                         goto out;
 229                 }
 230         }
 231 out:
 232         return err;
 233 out_rcu_unlock:
 234         rcu_read_unlock();
 235         goto out;
 236 }
 237
 238
 239 /* bind for INET6 API */
 240 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 241 {
 242         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 243         struct sock *sk = sock->sk;
 244         struct inet_sock *inet = inet_sk(sk);
 245         struct ipv6_pinfo *np = inet6_sk(sk);
 246         __u32 v4addr = 0;
 247         unsigned short snum;
 248         int addr_type = 0;
 249         int err = 0;
 250
 251         /* If the socket has its own bind function then use it. */
 252         if (sk->sk_prot->bind)
 253                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 254
 255         if (addr_len < SIN6_LEN_RFC2133)
 256                 return -EINVAL;
 257         addr_type = ipv6_addr_type(&addr->sin6_addr);
 258         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 259                 return -EINVAL;
 260
 261         snum = ntohs(addr->sin6_port);
 262         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 263                 return -EACCES;
 264
 265         lock_sock(sk);
 266
 267         /* Check these errors (active socket, double bind). */
 268         if (sk->sk_state != TCP_CLOSE || inet->num) {
 269                 err = -EINVAL;
 270                 goto out;
 271         }
 272
 273         /* Check if the address belongs to the host. */
 274         if (addr_type == IPV6_ADDR_MAPPED) {
 275                 v4addr = addr->sin6_addr.s6_addr32[3];
 276                 if (inet_addr_type(v4addr) != RTN_LOCAL) {
 277                         err = -EADDRNOTAVAIL;
 278                         goto out;
 279                 }
 280         } else {
 281                 if (addr_type != IPV6_ADDR_ANY) {
 282                         struct net_device *dev = NULL;
 283
 284                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 285                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 286                                     addr->sin6_scope_id) {
 287                                         /* Override any existing binding, if another one
 288                                          * is supplied by user.
 289                                          */
 290                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 291                                 }
 292
 293                                 /* Binding to link-local address requires an interface */
 294                                 if (!sk->sk_bound_dev_if) {
 295                                         err = -EINVAL;
 296                                         goto out;
 297                                 }
 298                                 dev = dev_get_by_index(sk->sk_bound_dev_if);
 299                                 if (!dev) {
 300                                         err = -ENODEV;
 301                                         goto out;
 302                                 }
 303                         }
 304
 305                         /* ipv4 addr of the socket is invalid.  Only the
 306                          * unspecified and mapped address have a v4 equivalent.
 307                          */
 308                         v4addr = LOOPBACK4_IPV6;
 309                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 310                                 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
 311                                         if (dev)
 312                                                 dev_put(dev);
 313                                         err = -EADDRNOTAVAIL;
 314                                         goto out;
 315                                 }
 316                         }
 317                         if (dev)
 318                                 dev_put(dev);
 319                 }
 320         }
 321
 322         inet->rcv_saddr = v4addr;
 323         inet->saddr = v4addr;
 324
 325         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 326
 327         if (!(addr_type & IPV6_ADDR_MULTICAST))
 328                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 329
 330         /* Make sure we are allowed to bind here. */
 331         if (sk->sk_prot->get_port(sk, snum)) {
 332                 inet_reset_saddr(sk);
 333                 err = -EADDRINUSE;
 334                 goto out;
 335         }
 336
 337         if (addr_type != IPV6_ADDR_ANY)
 338                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 339         if (snum)
 340                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 341         inet->sport = ntohs(inet->num);
 342         inet->dport = 0;
 343         inet->daddr = 0;
 344 out:
 345         release_sock(sk);
 346         return err;
 347 }
 348
 349 int inet6_release(struct socket *sock)
 350 {
 351         struct sock *sk = sock->sk;
 352
 353         if (sk == NULL)
 354                 return -EINVAL;
 355
 356         /* Free mc lists */
 357         ipv6_sock_mc_close(sk);
 358
 359         /* Free ac lists */
 360         ipv6_sock_ac_close(sk);
 361
 362         return inet_release(sock);
 363 }
 364
 365 int inet6_destroy_sock(struct sock *sk)
 366 {
 367         struct ipv6_pinfo *np = inet6_sk(sk);
 368         struct sk_buff *skb;
 369         struct ipv6_txoptions *opt;
 370
 371         /* Release rx options */
 372
 373         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 374                 kfree_skb(skb);
 375
 376         /* Free flowlabels */
 377         fl6_free_socklist(sk);
 378
 379         /* Free tx options */
 380
 381         if ((opt = xchg(&np->opt, NULL)) != NULL)
 382                 sock_kfree_s(sk, opt, opt->tot_len);
 383
 384         return 0;
 385 }
 386
 387 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 388
 389 /*
 390  *      This does both peername and sockname.
 391  */
 392
 393 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 394                  int *uaddr_len, int peer)
 395 {
 396         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 397         struct sock *sk = sock->sk;
 398         struct inet_sock *inet = inet_sk(sk);
 399         struct ipv6_pinfo *np = inet6_sk(sk);
 400
 401         sin->sin6_family = AF_INET6;
 402         sin->sin6_flowinfo = 0;
 403         sin->sin6_scope_id = 0;
 404         if (peer) {
 405                 if (!inet->dport)
 406                         return -ENOTCONN;
 407                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 408                     peer == 1)
 409                         return -ENOTCONN;
 410                 sin->sin6_port = inet->dport;
 411                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 412                 if (np->sndflow)
 413                         sin->sin6_flowinfo = np->flow_label;
 414         } else {
 415                 if (ipv6_addr_any(&np->rcv_saddr))
 416                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 417                 else
 418                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 419
 420                 sin->sin6_port = inet->sport;
 421         }
 422         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 423                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 424         *uaddr_len = sizeof(*sin);
 425         return(0);
 426 }
 427
 428 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 429 {
 430         struct sock *sk = sock->sk;
 431
 432         switch(cmd)
 433         {
 434         case SIOCGSTAMP:
 435                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 436
 437         case SIOCADDRT:
 438         case SIOCDELRT:
 439
 440                 return(ipv6_route_ioctl(cmd,(void __user *)arg));
 441
 442         case SIOCSIFADDR:
 443                 return addrconf_add_ifaddr((void __user *) arg);
 444         case SIOCDIFADDR:
 445                 return addrconf_del_ifaddr((void __user *) arg);
 446         case SIOCSIFDSTADDR:
 447                 return addrconf_set_dstaddr((void __user *) arg);
 448         default:
 449                 if (!sk->sk_prot->ioctl)
 450                         return -ENOIOCTLCMD;
 451                 return sk->sk_prot->ioctl(sk, cmd, arg);
 452         }
 453         /*NOTREACHED*/
 454         return(0);
 455 }
 456
 457 const struct proto_ops inet6_stream_ops = {
 458         .family            = PF_INET6,
 459         .owner             = THIS_MODULE,
 460         .release           = inet6_release,
 461         .bind              = inet6_bind,
 462         .connect           = inet_stream_connect,       /* ok           */
 463         .socketpair        = sock_no_socketpair,        /* a do nothing */
 464         .accept            = inet_accept,               /* ok           */
 465         .getname           = inet6_getname,
 466         .poll              = tcp_poll,                  /* ok           */
 467         .ioctl             = inet6_ioctl,               /* must change  */
 468         .listen            = inet_listen,               /* ok           */
 469         .shutdown          = inet_shutdown,             /* ok           */
 470         .setsockopt        = sock_common_setsockopt,    /* ok           */
 471         .getsockopt        = sock_common_getsockopt,    /* ok           */
 472         .sendmsg           = inet_sendmsg,              /* ok           */
 473         .recvmsg           = sock_common_recvmsg,       /* ok           */
 474         .mmap              = sock_no_mmap,
 475         .sendpage          = tcp_sendpage,
 476 #ifdef CONFIG_COMPAT
 477         .compat_setsockopt = compat_sock_common_setsockopt,
 478         .compat_getsockopt = compat_sock_common_getsockopt,
 479 #endif
 480 };
 481
 482 const struct proto_ops inet6_dgram_ops = {
 483         .family            = PF_INET6,
 484         .owner             = THIS_MODULE,
 485         .release           = inet6_release,
 486         .bind              = inet6_bind,
 487         .connect           = inet_dgram_connect,        /* ok           */
 488         .socketpair        = sock_no_socketpair,        /* a do nothing */
 489         .accept            = sock_no_accept,            /* a do nothing */
 490         .getname           = inet6_getname,
 491         .poll              = udp_poll,                  /* ok           */
 492         .ioctl             = inet6_ioctl,               /* must change  */
 493         .listen            = sock_no_listen,            /* ok           */
 494         .shutdown          = inet_shutdown,             /* ok           */
 495         .setsockopt        = sock_common_setsockopt,    /* ok           */
 496         .getsockopt        = sock_common_getsockopt,    /* ok           */
 497         .sendmsg           = inet_sendmsg,              /* ok           */
 498         .recvmsg           = sock_common_recvmsg,       /* ok           */
 499         .mmap              = sock_no_mmap,
 500         .sendpage          = sock_no_sendpage,
 501 #ifdef CONFIG_COMPAT
 502         .compat_setsockopt = compat_sock_common_setsockopt,
 503         .compat_getsockopt = compat_sock_common_getsockopt,
 504 #endif
 505 };
 506
 507 static struct net_proto_family inet6_family_ops = {
 508         .family = PF_INET6,
 509         .create = inet6_create,
 510         .owner  = THIS_MODULE,
 511 };
 512
 513 /* Same as inet6_dgram_ops, sans udp_poll.  */
 514 static const struct proto_ops inet6_sockraw_ops = {
 515         .family            = PF_INET6,
 516         .owner             = THIS_MODULE,
 517         .release           = inet6_release,
 518         .bind              = inet6_bind,
 519         .connect           = inet_dgram_connect,        /* ok           */
 520         .socketpair        = sock_no_socketpair,        /* a do nothing */
 521         .accept            = sock_no_accept,            /* a do nothing */
 522         .getname           = inet6_getname,
 523         .poll              = datagram_poll,             /* ok           */
 524         .ioctl             = inet6_ioctl,               /* must change  */
 525         .listen            = sock_no_listen,            /* ok           */
 526         .shutdown          = inet_shutdown,             /* ok           */
 527         .setsockopt        = sock_common_setsockopt,    /* ok           */
 528         .getsockopt        = sock_common_getsockopt,    /* ok           */
 529         .sendmsg           = inet_sendmsg,              /* ok           */
 530         .recvmsg           = sock_common_recvmsg,       /* ok           */
 531         .mmap              = sock_no_mmap,
 532         .sendpage          = sock_no_sendpage,
 533 #ifdef CONFIG_COMPAT
 534         .compat_setsockopt = compat_sock_common_setsockopt,
 535         .compat_getsockopt = compat_sock_common_getsockopt,
 536 #endif
 537 };
 538
 539 static struct inet_protosw rawv6_protosw = {
 540         .type           = SOCK_RAW,
 541         .protocol       = IPPROTO_IP,   /* wild card */
 542         .prot           = &rawv6_prot,
 543         .ops            = &inet6_sockraw_ops,
 544         .capability     = CAP_NET_RAW,
 545         .no_check       = UDP_CSUM_DEFAULT,
 546         .flags          = INET_PROTOSW_REUSE,
 547 };
 548
 549 void
 550 inet6_register_protosw(struct inet_protosw *p)
 551 {
 552         struct list_head *lh;
 553         struct inet_protosw *answer;
 554         int protocol = p->protocol;
 555         struct list_head *last_perm;
 556
 557         spin_lock_bh(&inetsw6_lock);
 558
 559         if (p->type >= SOCK_MAX)
 560                 goto out_illegal;
 561
 562         /* If we are trying to override a permanent protocol, bail. */
 563         answer = NULL;
 564         last_perm = &inetsw6[p->type];
 565         list_for_each(lh, &inetsw6[p->type]) {
 566                 answer = list_entry(lh, struct inet_protosw, list);
 567
 568                 /* Check only the non-wild match. */
 569                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 570                         if (protocol == answer->protocol)
 571                                 break;
 572                         last_perm = lh;
 573                 }
 574
 575                 answer = NULL;
 576         }
 577         if (answer)
 578                 goto out_permanent;
 579
 580         /* Add the new entry after the last permanent entry if any, so that
 581          * the new entry does not override a permanent entry when matched with
 582          * a wild-card protocol. But it is allowed to override any existing
 583          * non-permanent entry.  This means that when we remove this entry, the
 584          * system automatically returns to the old behavior.
 585          */
 586         list_add_rcu(&p->list, last_perm);
 587 out:
 588         spin_unlock_bh(&inetsw6_lock);
 589         return;
 590
 591 out_permanent:
 592         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 593                protocol);
 594         goto out;
 595
 596 out_illegal:
 597         printk(KERN_ERR
 598                "Ignoring attempt to register invalid socket type %d.\n",
 599                p->type);
 600         goto out;
 601 }
 602
 603 void
 604 inet6_unregister_protosw(struct inet_protosw *p)
 605 {
 606         if (INET_PROTOSW_PERMANENT & p->flags) {
 607                 printk(KERN_ERR
 608                        "Attempt to unregister permanent protocol %d.\n",
 609                        p->protocol);
 610         } else {
 611                 spin_lock_bh(&inetsw6_lock);
 612                 list_del_rcu(&p->list);
 613                 spin_unlock_bh(&inetsw6_lock);
 614
 615                 synchronize_net();
 616         }
 617 }
 618
 619 int inet6_sk_rebuild_header(struct sock *sk)
 620 {
 621         int err;
 622         struct dst_entry *dst;
 623         struct ipv6_pinfo *np = inet6_sk(sk);
 624
 625         dst = __sk_dst_check(sk, np->dst_cookie);
 626
 627         if (dst == NULL) {
 628                 struct inet_sock *inet = inet_sk(sk);
 629                 struct in6_addr *final_p = NULL, final;
 630                 struct flowi fl;
 631
 632                 memset(&fl, 0, sizeof(fl));
 633                 fl.proto = sk->sk_protocol;
 634                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 635                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 636                 fl.fl6_flowlabel = np->flow_label;
 637                 fl.oif = sk->sk_bound_dev_if;
 638                 fl.fl_ip_dport = inet->dport;
 639                 fl.fl_ip_sport = inet->sport;
 640
 641                 if (np->opt && np->opt->srcrt) {
 642                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 643                         ipv6_addr_copy(&final, &fl.fl6_dst);
 644                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 645                         final_p = &final;
 646                 }
 647
 648                 err = ip6_dst_lookup(sk, &dst, &fl);
 649                 if (err) {
 650                         sk->sk_route_caps = 0;
 651                         return err;
 652                 }
 653                 if (final_p)
 654                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 655
 656                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
 657                         sk->sk_err_soft = -err;
 658                         return err;
 659                 }
 660
 661                 __ip6_dst_store(sk, dst, NULL);
 662         }
 663
 664         return 0;
 665 }
 666
 667 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 668
 669 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 670 {
 671         struct ipv6_pinfo *np = inet6_sk(sk);
 672         struct inet6_skb_parm *opt = IP6CB(skb);
 673
 674         if (np->rxopt.all) {
 675                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 676                                   np->rxopt.bits.ohopopts)) ||
 677                     ((IPV6_FLOWINFO_MASK & *(u32*)skb->nh.raw) &&
 678                      np->rxopt.bits.rxflow) ||
 679                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 680                      np->rxopt.bits.osrcrt)) ||
 681                     ((opt->dst1 || opt->dst0) &&
 682                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 683                         return 1;
 684         }
 685         return 0;
 686 }
 687
 688 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 689
 690 int
 691 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
 692 {
 693         if (ptr == NULL)
 694                 return -EINVAL;
 695
 696         ptr[0] = __alloc_percpu(mibsize);
 697         if (!ptr[0])
 698                 goto err0;
 699
 700         ptr[1] = __alloc_percpu(mibsize);
 701         if (!ptr[1])
 702                 goto err1;
 703
 704         return 0;
 705
 706 err1:
 707         free_percpu(ptr[0]);
 708         ptr[0] = NULL;
 709 err0:
 710         return -ENOMEM;
 711 }
 712
 713 void
 714 snmp6_mib_free(void *ptr[2])
 715 {
 716         if (ptr == NULL)
 717                 return;
 718         if (ptr[0])
 719                 free_percpu(ptr[0]);
 720         if (ptr[1])
 721                 free_percpu(ptr[1]);
 722         ptr[0] = ptr[1] = NULL;
 723 }
 724
 725 static int __init init_ipv6_mibs(void)
 726 {
 727         if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
 728                            __alignof__(struct ipstats_mib)) < 0)
 729                 goto err_ip_mib;
 730         if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
 731                            __alignof__(struct icmpv6_mib)) < 0)
 732                 goto err_icmp_mib;
 733         if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
 734                            __alignof__(struct udp_mib)) < 0)
 735                 goto err_udp_mib;
 736         return 0;
 737
 738 err_udp_mib:
 739         snmp6_mib_free((void **)icmpv6_statistics);
 740 err_icmp_mib:
 741         snmp6_mib_free((void **)ipv6_statistics);
 742 err_ip_mib:
 743         return -ENOMEM;
 744
 745 }
 746
 747 static void cleanup_ipv6_mibs(void)
 748 {
 749         snmp6_mib_free((void **)ipv6_statistics);
 750         snmp6_mib_free((void **)icmpv6_statistics);
 751         snmp6_mib_free((void **)udp_stats_in6);
 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 #ifdef MODULE
 761 #if 0 /* FIXME --RR */
 762         if (!mod_member_present(&__this_module, can_unload))
 763           return -EINVAL;
 764
 765         __this_module.can_unload = &ipv6_unload;
 766 #endif
 767 #endif
 768
 769         if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb)) {
 770                 printk(KERN_CRIT "inet6_proto_init: size fault\n");
 771                 return -EINVAL;
 772         }
 773
 774         err = proto_register(&tcpv6_prot, 1);
 775         if (err)
 776                 goto out;
 777
 778         err = proto_register(&udpv6_prot, 1);
 779         if (err)
 780                 goto out_unregister_tcp_proto;
 781
 782         err = proto_register(&rawv6_prot, 1);
 783         if (err)
 784                 goto out_unregister_udp_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         inet6_register_protosw(&rawv6_protosw);
 795
 796         /* Register the family here so that the init calls below will
 797          * be able to create sockets. (?? is this dangerous ??)
 798          */
 799         err = sock_register(&inet6_family_ops);
 800         if (err)
 801                 goto out_unregister_raw_proto;
 802
 803         /* Initialise ipv6 mibs */
 804         err = init_ipv6_mibs();
 805         if (err)
 806                 goto out_unregister_sock;
 807
 808         /*
 809          *      ipngwg API draft makes clear that the correct semantics
 810          *      for TCP and UDP is to consider one TCP and UDP instance
 811          *      in a host availiable by both INET and INET6 APIs and
 812          *      able to communicate via both network protocols.
 813          */
 814
 815 #ifdef CONFIG_SYSCTL
 816         ipv6_sysctl_register();
 817 #endif
 818         err = icmpv6_init(&inet6_family_ops);
 819         if (err)
 820                 goto icmp_fail;
 821         err = ndisc_init(&inet6_family_ops);
 822         if (err)
 823                 goto ndisc_fail;
 824         err = igmp6_init(&inet6_family_ops);
 825         if (err)
 826                 goto igmp_fail;
 827         err = ipv6_netfilter_init();
 828         if (err)
 829                 goto netfilter_fail;
 830         /* Create /proc/foo6 entries. */
 831 #ifdef CONFIG_PROC_FS
 832         err = -ENOMEM;
 833         if (raw6_proc_init())
 834                 goto proc_raw6_fail;
 835         if (tcp6_proc_init())
 836                 goto proc_tcp6_fail;
 837         if (udp6_proc_init())
 838                 goto proc_udp6_fail;
 839         if (ipv6_misc_proc_init())
 840                 goto proc_misc6_fail;
 841
 842         if (ac6_proc_init())
 843                 goto proc_anycast6_fail;
 844         if (if6_proc_init())
 845                 goto proc_if6_fail;
 846 #endif
 847         ip6_route_init();
 848         ip6_flowlabel_init();
 849         err = addrconf_init();
 850         if (err)
 851                 goto addrconf_fail;
 852         sit_init();
 853
 854         /* Init v6 extension headers. */
 855         ipv6_rthdr_init();
 856         ipv6_frag_init();
 857         ipv6_nodata_init();
 858         ipv6_destopt_init();
 859
 860         /* Init v6 transport protocols. */
 861         udpv6_init();
 862         tcpv6_init();
 863
 864         ipv6_packet_init();
 865         err = 0;
 866 out:
 867         return err;
 868
 869 addrconf_fail:
 870         ip6_flowlabel_cleanup();
 871         ip6_route_cleanup();
 872 #ifdef CONFIG_PROC_FS
 873         if6_proc_exit();
 874 proc_if6_fail:
 875         ac6_proc_exit();
 876 proc_anycast6_fail:
 877         ipv6_misc_proc_exit();
 878 proc_misc6_fail:
 879         udp6_proc_exit();
 880 proc_udp6_fail:
 881         tcp6_proc_exit();
 882 proc_tcp6_fail:
 883         raw6_proc_exit();
 884 proc_raw6_fail:
 885 #endif
 886         ipv6_netfilter_fini();
 887 netfilter_fail:
 888         igmp6_cleanup();
 889 igmp_fail:
 890         ndisc_cleanup();
 891 ndisc_fail:
 892         icmpv6_cleanup();
 893 icmp_fail:
 894 #ifdef CONFIG_SYSCTL
 895         ipv6_sysctl_unregister();
 896 #endif
 897         cleanup_ipv6_mibs();
 898 out_unregister_sock:
 899         sock_unregister(PF_INET6);
 900 out_unregister_raw_proto:
 901         proto_unregister(&rawv6_prot);
 902 out_unregister_udp_proto:
 903         proto_unregister(&udpv6_prot);
 904 out_unregister_tcp_proto:
 905         proto_unregister(&tcpv6_prot);
 906         goto out;
 907 }
 908 module_init(inet6_init);
 909
 910 static void __exit inet6_exit(void)
 911 {
 912         /* First of all disallow new sockets creation. */
 913         sock_unregister(PF_INET6);
 914 #ifdef CONFIG_PROC_FS
 915         if6_proc_exit();
 916         ac6_proc_exit();
 917         ipv6_misc_proc_exit();
 918         udp6_proc_exit();
 919         tcp6_proc_exit();
 920         raw6_proc_exit();
 921 #endif
 922         /* Cleanup code parts. */
 923         sit_cleanup();
 924         ip6_flowlabel_cleanup();
 925         addrconf_cleanup();
 926         ip6_route_cleanup();
 927         ipv6_packet_cleanup();
 928         igmp6_cleanup();
 929         ipv6_netfilter_fini();
 930         ndisc_cleanup();
 931         icmpv6_cleanup();
 932 #ifdef CONFIG_SYSCTL
 933         ipv6_sysctl_unregister();
 934 #endif
 935         cleanup_ipv6_mibs();
 936         proto_unregister(&rawv6_prot);
 937         proto_unregister(&udpv6_prot);
 938         proto_unregister(&tcpv6_prot);
 939 }
 940 module_exit(inet6_exit);
 941
 942 MODULE_ALIAS_NETPROTO(PF_INET6);