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/config.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
 171         if (SOCK_RAW == sock->type) {
 172                 inet->num = protocol;
 173                 if (IPPROTO_RAW == protocol)
 174                         inet->hdrincl = 1;
 175         }
 176
 177         sk->sk_destruct         = inet_sock_destruct;
 178         sk->sk_family           = PF_INET6;
 179         sk->sk_protocol         = protocol;
 180
 181         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 182
 183         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 184         np->hop_limit   = -1;
 185         np->mcast_hops  = -1;
 186         np->mc_loop     = 1;
 187         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 188         np->ipv6only    = sysctl_ipv6_bindv6only;
 189
 190         /* Init the ipv4 part of the socket since we can have sockets
 191          * using v6 API for ipv4.
 192          */
 193         inet->uc_ttl    = -1;
 194
 195         inet->mc_loop   = 1;
 196         inet->mc_ttl    = 1;
 197         inet->mc_index  = 0;
 198         inet->mc_list   = NULL;
 199
 200         if (ipv4_config.no_pmtu_disc)
 201                 inet->pmtudisc = IP_PMTUDISC_DONT;
 202         else
 203                 inet->pmtudisc = IP_PMTUDISC_WANT;
 204         /*
 205          * Increment only the relevant sk_prot->socks debug field, this changes
 206          * the previous behaviour of incrementing both the equivalent to
 207          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 208          *
 209          * This allows better debug granularity as we'll know exactly how many
 210          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 211          * transport protocol socks. -acme
 212          */
 213         sk_refcnt_debug_inc(sk);
 214
 215         if (inet->num) {
 216                 /* It assumes that any protocol which allows
 217                  * the user to assign a number at socket
 218                  * creation time automatically shares.
 219                  */
 220                 inet->sport = ntohs(inet->num);
 221                 sk->sk_prot->hash(sk);
 222         }
 223         if (sk->sk_prot->init) {
 224                 err = sk->sk_prot->init(sk);
 225                 if (err) {
 226                         sk_common_release(sk);
 227                         goto out;
 228                 }
 229         }
 230 out:
 231         return err;
 232 out_rcu_unlock:
 233         rcu_read_unlock();
 234         goto out;
 235 }
 236
 237
 238 /* bind for INET6 API */
 239 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 240 {
 241         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 242         struct sock *sk = sock->sk;
 243         struct inet_sock *inet = inet_sk(sk);
 244         struct ipv6_pinfo *np = inet6_sk(sk);
 245         __u32 v4addr = 0;
 246         unsigned short snum;
 247         int addr_type = 0;
 248         int err = 0;
 249
 250         /* If the socket has its own bind function then use it. */
 251         if (sk->sk_prot->bind)
 252                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 253
 254         if (addr_len < SIN6_LEN_RFC2133)
 255                 return -EINVAL;
 256         addr_type = ipv6_addr_type(&addr->sin6_addr);
 257         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 258                 return -EINVAL;
 259
 260         snum = ntohs(addr->sin6_port);
 261         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 262                 return -EACCES;
 263
 264         lock_sock(sk);
 265
 266         /* Check these errors (active socket, double bind). */
 267         if (sk->sk_state != TCP_CLOSE || inet->num) {
 268                 err = -EINVAL;
 269                 goto out;
 270         }
 271
 272         /* Check if the address belongs to the host. */
 273         if (addr_type == IPV6_ADDR_MAPPED) {
 274                 v4addr = addr->sin6_addr.s6_addr32[3];
 275                 if (inet_addr_type(v4addr) != RTN_LOCAL) {
 276                         err = -EADDRNOTAVAIL;
 277                         goto out;
 278                 }
 279         } else {
 280                 if (addr_type != IPV6_ADDR_ANY) {
 281                         struct net_device *dev = NULL;
 282
 283                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 284                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 285                                     addr->sin6_scope_id) {
 286                                         /* Override any existing binding, if another one
 287                                          * is supplied by user.
 288                                          */
 289                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 290                                 }
 291
 292                                 /* Binding to link-local address requires an interface */
 293                                 if (!sk->sk_bound_dev_if) {
 294                                         err = -EINVAL;
 295                                         goto out;
 296                                 }
 297                                 dev = dev_get_by_index(sk->sk_bound_dev_if);
 298                                 if (!dev) {
 299                                         err = -ENODEV;
 300                                         goto out;
 301                                 }
 302                         }
 303
 304                         /* ipv4 addr of the socket is invalid.  Only the
 305                          * unspecified and mapped address have a v4 equivalent.
 306                          */
 307                         v4addr = LOOPBACK4_IPV6;
 308                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 309                                 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
 310                                         if (dev)
 311                                                 dev_put(dev);
 312                                         err = -EADDRNOTAVAIL;
 313                                         goto out;
 314                                 }
 315                         }
 316                         if (dev)
 317                                 dev_put(dev);
 318                 }
 319         }
 320
 321         inet->rcv_saddr = v4addr;
 322         inet->saddr = v4addr;
 323
 324         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 325
 326         if (!(addr_type & IPV6_ADDR_MULTICAST))
 327                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 328
 329         /* Make sure we are allowed to bind here. */
 330         if (sk->sk_prot->get_port(sk, snum)) {
 331                 inet_reset_saddr(sk);
 332                 err = -EADDRINUSE;
 333                 goto out;
 334         }
 335
 336         if (addr_type != IPV6_ADDR_ANY)
 337                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 338         if (snum)
 339                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 340         inet->sport = ntohs(inet->num);
 341         inet->dport = 0;
 342         inet->daddr = 0;
 343 out:
 344         release_sock(sk);
 345         return err;
 346 }
 347
 348 int inet6_release(struct socket *sock)
 349 {
 350         struct sock *sk = sock->sk;
 351
 352         if (sk == NULL)
 353                 return -EINVAL;
 354
 355         /* Free mc lists */
 356         ipv6_sock_mc_close(sk);
 357
 358         /* Free ac lists */
 359         ipv6_sock_ac_close(sk);
 360
 361         return inet_release(sock);
 362 }
 363
 364 int inet6_destroy_sock(struct sock *sk)
 365 {
 366         struct ipv6_pinfo *np = inet6_sk(sk);
 367         struct sk_buff *skb;
 368         struct ipv6_txoptions *opt;
 369
 370         /*
 371          *      Release destination entry
 372          */
 373
 374         sk_dst_reset(sk);
 375
 376         /* Release rx options */
 377
 378         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 379                 kfree_skb(skb);
 380
 381         /* Free flowlabels */
 382         fl6_free_socklist(sk);
 383
 384         /* Free tx options */
 385
 386         if ((opt = xchg(&np->opt, NULL)) != NULL)
 387                 sock_kfree_s(sk, opt, opt->tot_len);
 388
 389         return 0;
 390 }
 391
 392 /*
 393  *      This does both peername and sockname.
 394  */
 395
 396 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 397                  int *uaddr_len, int peer)
 398 {
 399         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 400         struct sock *sk = sock->sk;
 401         struct inet_sock *inet = inet_sk(sk);
 402         struct ipv6_pinfo *np = inet6_sk(sk);
 403
 404         sin->sin6_family = AF_INET6;
 405         sin->sin6_flowinfo = 0;
 406         sin->sin6_scope_id = 0;
 407         if (peer) {
 408                 if (!inet->dport)
 409                         return -ENOTCONN;
 410                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 411                     peer == 1)
 412                         return -ENOTCONN;
 413                 sin->sin6_port = inet->dport;
 414                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 415                 if (np->sndflow)
 416                         sin->sin6_flowinfo = np->flow_label;
 417         } else {
 418                 if (ipv6_addr_any(&np->rcv_saddr))
 419                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 420                 else
 421                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 422
 423                 sin->sin6_port = inet->sport;
 424         }
 425         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 426                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 427         *uaddr_len = sizeof(*sin);
 428         return(0);
 429 }
 430
 431 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 432 {
 433         struct sock *sk = sock->sk;
 434         int err = -EINVAL;
 435
 436         switch(cmd)
 437         {
 438         case SIOCGSTAMP:
 439                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 440
 441         case SIOCADDRT:
 442         case SIOCDELRT:
 443
 444                 return(ipv6_route_ioctl(cmd,(void __user *)arg));
 445
 446         case SIOCSIFADDR:
 447                 return addrconf_add_ifaddr((void __user *) arg);
 448         case SIOCDIFADDR:
 449                 return addrconf_del_ifaddr((void __user *) arg);
 450         case SIOCSIFDSTADDR:
 451                 return addrconf_set_dstaddr((void __user *) arg);
 452         default:
 453                 if (!sk->sk_prot->ioctl ||
 454                     (err = sk->sk_prot->ioctl(sk, cmd, arg)) == -ENOIOCTLCMD)
 455                         return(dev_ioctl(cmd,(void __user *) arg));
 456                 return err;
 457         }
 458         /*NOTREACHED*/
 459         return(0);
 460 }
 461
 462 struct proto_ops inet6_stream_ops = {
 463         .family =       PF_INET6,
 464         .owner =        THIS_MODULE,
 465         .release =      inet6_release,
 466         .bind =         inet6_bind,
 467         .connect =      inet_stream_connect,            /* ok           */
 468         .socketpair =   sock_no_socketpair,             /* a do nothing */
 469         .accept =       inet_accept,                    /* ok           */
 470         .getname =      inet6_getname,
 471         .poll =         tcp_poll,                       /* ok           */
 472         .ioctl =        inet6_ioctl,                    /* must change  */
 473         .listen =       inet_listen,                    /* ok           */
 474         .shutdown =     inet_shutdown,                  /* ok           */
 475         .setsockopt =   sock_common_setsockopt,         /* ok           */
 476         .getsockopt =   sock_common_getsockopt,         /* ok           */
 477         .sendmsg =      inet_sendmsg,                   /* ok           */
 478         .recvmsg =      sock_common_recvmsg,            /* ok           */
 479         .mmap =         sock_no_mmap,
 480         .sendpage =     tcp_sendpage
 481 };
 482
 483 struct proto_ops inet6_dgram_ops = {
 484         .family =       PF_INET6,
 485         .owner =        THIS_MODULE,
 486         .release =      inet6_release,
 487         .bind =         inet6_bind,
 488         .connect =      inet_dgram_connect,             /* ok           */
 489         .socketpair =   sock_no_socketpair,             /* a do nothing */
 490         .accept =       sock_no_accept,                 /* a do nothing */
 491         .getname =      inet6_getname,
 492         .poll =         udp_poll,                       /* ok           */
 493         .ioctl =        inet6_ioctl,                    /* must change  */
 494         .listen =       sock_no_listen,                 /* ok           */
 495         .shutdown =     inet_shutdown,                  /* ok           */
 496         .setsockopt =   sock_common_setsockopt,         /* ok           */
 497         .getsockopt =   sock_common_getsockopt,         /* ok           */
 498         .sendmsg =      inet_sendmsg,                   /* ok           */
 499         .recvmsg =      sock_common_recvmsg,            /* ok           */
 500         .mmap =         sock_no_mmap,
 501         .sendpage =     sock_no_sendpage,
 502 };
 503
 504 static struct net_proto_family inet6_family_ops = {
 505         .family = PF_INET6,
 506         .create = inet6_create,
 507         .owner  = THIS_MODULE,
 508 };
 509
 510 /* Same as inet6_dgram_ops, sans udp_poll.  */
 511 static struct proto_ops inet6_sockraw_ops = {
 512         .family =       PF_INET6,
 513         .owner =        THIS_MODULE,
 514         .release =      inet6_release,
 515         .bind =         inet6_bind,
 516         .connect =      inet_dgram_connect,             /* ok           */
 517         .socketpair =   sock_no_socketpair,             /* a do nothing */
 518         .accept =       sock_no_accept,                 /* a do nothing */
 519         .getname =      inet6_getname,
 520         .poll =         datagram_poll,                  /* ok           */
 521         .ioctl =        inet6_ioctl,                    /* must change  */
 522         .listen =       sock_no_listen,                 /* ok           */
 523         .shutdown =     inet_shutdown,                  /* ok           */
 524         .setsockopt =   sock_common_setsockopt,         /* ok           */
 525         .getsockopt =   sock_common_getsockopt,         /* ok           */
 526         .sendmsg =      inet_sendmsg,                   /* ok           */
 527         .recvmsg =      sock_common_recvmsg,            /* ok           */
 528         .mmap =         sock_no_mmap,
 529         .sendpage =     sock_no_sendpage,
 530 };
 531
 532 static struct inet_protosw rawv6_protosw = {
 533         .type           = SOCK_RAW,
 534         .protocol       = IPPROTO_IP,   /* wild card */
 535         .prot           = &rawv6_prot,
 536         .ops            = &inet6_sockraw_ops,
 537         .capability     = CAP_NET_RAW,
 538         .no_check       = UDP_CSUM_DEFAULT,
 539         .flags          = INET_PROTOSW_REUSE,
 540 };
 541
 542 void
 543 inet6_register_protosw(struct inet_protosw *p)
 544 {
 545         struct list_head *lh;
 546         struct inet_protosw *answer;
 547         int protocol = p->protocol;
 548         struct list_head *last_perm;
 549
 550         spin_lock_bh(&inetsw6_lock);
 551
 552         if (p->type >= SOCK_MAX)
 553                 goto out_illegal;
 554
 555         /* If we are trying to override a permanent protocol, bail. */
 556         answer = NULL;
 557         last_perm = &inetsw6[p->type];
 558         list_for_each(lh, &inetsw6[p->type]) {
 559                 answer = list_entry(lh, struct inet_protosw, list);
 560
 561                 /* Check only the non-wild match. */
 562                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 563                         if (protocol == answer->protocol)
 564                                 break;
 565                         last_perm = lh;
 566                 }
 567
 568                 answer = NULL;
 569         }
 570         if (answer)
 571                 goto out_permanent;
 572
 573         /* Add the new entry after the last permanent entry if any, so that
 574          * the new entry does not override a permanent entry when matched with
 575          * a wild-card protocol. But it is allowed to override any existing
 576          * non-permanent entry.  This means that when we remove this entry, the
 577          * system automatically returns to the old behavior.
 578          */
 579         list_add_rcu(&p->list, last_perm);
 580 out:
 581         spin_unlock_bh(&inetsw6_lock);
 582         return;
 583
 584 out_permanent:
 585         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 586                protocol);
 587         goto out;
 588
 589 out_illegal:
 590         printk(KERN_ERR
 591                "Ignoring attempt to register invalid socket type %d.\n",
 592                p->type);
 593         goto out;
 594 }
 595
 596 void
 597 inet6_unregister_protosw(struct inet_protosw *p)
 598 {
 599         if (INET_PROTOSW_PERMANENT & p->flags) {
 600                 printk(KERN_ERR
 601                        "Attempt to unregister permanent protocol %d.\n",
 602                        p->protocol);
 603         } else {
 604                 spin_lock_bh(&inetsw6_lock);
 605                 list_del_rcu(&p->list);
 606                 spin_unlock_bh(&inetsw6_lock);
 607
 608                 synchronize_net();
 609         }
 610 }
 611
 612 int inet6_sk_rebuild_header(struct sock *sk)
 613 {
 614         int err;
 615         struct dst_entry *dst;
 616         struct ipv6_pinfo *np = inet6_sk(sk);
 617
 618         dst = __sk_dst_check(sk, np->dst_cookie);
 619
 620         if (dst == NULL) {
 621                 struct inet_sock *inet = inet_sk(sk);
 622                 struct in6_addr *final_p = NULL, final;
 623                 struct flowi fl;
 624
 625                 memset(&fl, 0, sizeof(fl));
 626                 fl.proto = sk->sk_protocol;
 627                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 628                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 629                 fl.fl6_flowlabel = np->flow_label;
 630                 fl.oif = sk->sk_bound_dev_if;
 631                 fl.fl_ip_dport = inet->dport;
 632                 fl.fl_ip_sport = inet->sport;
 633
 634                 if (np->opt && np->opt->srcrt) {
 635                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 636                         ipv6_addr_copy(&final, &fl.fl6_dst);
 637                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 638                         final_p = &final;
 639                 }
 640
 641                 err = ip6_dst_lookup(sk, &dst, &fl);
 642                 if (err) {
 643                         sk->sk_route_caps = 0;
 644                         return err;
 645                 }
 646                 if (final_p)
 647                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 648
 649                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
 650                         sk->sk_err_soft = -err;
 651                         return err;
 652                 }
 653
 654                 ip6_dst_store(sk, dst, NULL);
 655                 sk->sk_route_caps = dst->dev->features &
 656                         ~(NETIF_F_IP_CSUM | NETIF_F_TSO);
 657         }
 658
 659         return 0;
 660 }
 661
 662 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 663
 664 int
 665 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
 666 {
 667         if (ptr == NULL)
 668                 return -EINVAL;
 669
 670         ptr[0] = __alloc_percpu(mibsize, mibalign);
 671         if (!ptr[0])
 672                 goto err0;
 673
 674         ptr[1] = __alloc_percpu(mibsize, mibalign);
 675         if (!ptr[1])
 676                 goto err1;
 677
 678         return 0;
 679
 680 err1:
 681         free_percpu(ptr[0]);
 682         ptr[0] = NULL;
 683 err0:
 684         return -ENOMEM;
 685 }
 686
 687 void
 688 snmp6_mib_free(void *ptr[2])
 689 {
 690         if (ptr == NULL)
 691                 return;
 692         if (ptr[0])
 693                 free_percpu(ptr[0]);
 694         if (ptr[1])
 695                 free_percpu(ptr[1]);
 696         ptr[0] = ptr[1] = NULL;
 697 }
 698
 699 static int __init init_ipv6_mibs(void)
 700 {
 701         if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
 702                            __alignof__(struct ipstats_mib)) < 0)
 703                 goto err_ip_mib;
 704         if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
 705                            __alignof__(struct icmpv6_mib)) < 0)
 706                 goto err_icmp_mib;
 707         if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
 708                            __alignof__(struct udp_mib)) < 0)
 709                 goto err_udp_mib;
 710         return 0;
 711
 712 err_udp_mib:
 713         snmp6_mib_free((void **)icmpv6_statistics);
 714 err_icmp_mib:
 715         snmp6_mib_free((void **)ipv6_statistics);
 716 err_ip_mib:
 717         return -ENOMEM;
 718
 719 }
 720
 721 static void cleanup_ipv6_mibs(void)
 722 {
 723         snmp6_mib_free((void **)ipv6_statistics);
 724         snmp6_mib_free((void **)icmpv6_statistics);
 725         snmp6_mib_free((void **)udp_stats_in6);
 726 }
 727
 728 static int __init inet6_init(void)
 729 {
 730         struct sk_buff *dummy_skb;
 731         struct list_head *r;
 732         int err;
 733
 734 #ifdef MODULE
 735 #if 0 /* FIXME --RR */
 736         if (!mod_member_present(&__this_module, can_unload))
 737           return -EINVAL;
 738
 739         __this_module.can_unload = &ipv6_unload;
 740 #endif
 741 #endif
 742
 743         if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb)) {
 744                 printk(KERN_CRIT "inet6_proto_init: size fault\n");
 745                 return -EINVAL;
 746         }
 747
 748         err = proto_register(&tcpv6_prot, 1);
 749         if (err)
 750                 goto out;
 751
 752         err = proto_register(&udpv6_prot, 1);
 753         if (err)
 754                 goto out_unregister_tcp_proto;
 755
 756         err = proto_register(&rawv6_prot, 1);
 757         if (err)
 758                 goto out_unregister_udp_proto;
 759
 760
 761         /* Register the socket-side information for inet6_create.  */
 762         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 763                 INIT_LIST_HEAD(r);
 764
 765         /* We MUST register RAW sockets before we create the ICMP6,
 766          * IGMP6, or NDISC control sockets.
 767          */
 768         inet6_register_protosw(&rawv6_protosw);
 769
 770         /* Register the family here so that the init calls below will
 771          * be able to create sockets. (?? is this dangerous ??)
 772          */
 773         err = sock_register(&inet6_family_ops);
 774         if (err)
 775                 goto out_unregister_raw_proto;
 776
 777         /* Initialise ipv6 mibs */
 778         err = init_ipv6_mibs();
 779         if (err)
 780                 goto out_unregister_sock;
 781
 782         /*
 783          *      ipngwg API draft makes clear that the correct semantics
 784          *      for TCP and UDP is to consider one TCP and UDP instance
 785          *      in a host availiable by both INET and INET6 APIs and
 786          *      able to communicate via both network protocols.
 787          */
 788
 789 #ifdef CONFIG_SYSCTL
 790         ipv6_sysctl_register();
 791 #endif
 792         err = icmpv6_init(&inet6_family_ops);
 793         if (err)
 794                 goto icmp_fail;
 795         err = ndisc_init(&inet6_family_ops);
 796         if (err)
 797                 goto ndisc_fail;
 798         err = igmp6_init(&inet6_family_ops);
 799         if (err)
 800                 goto igmp_fail;
 801         err = ipv6_netfilter_init();
 802         if (err)
 803                 goto netfilter_fail;
 804         /* Create /proc/foo6 entries. */
 805 #ifdef CONFIG_PROC_FS
 806         err = -ENOMEM;
 807         if (raw6_proc_init())
 808                 goto proc_raw6_fail;
 809         if (tcp6_proc_init())
 810                 goto proc_tcp6_fail;
 811         if (udp6_proc_init())
 812                 goto proc_udp6_fail;
 813         if (ipv6_misc_proc_init())
 814                 goto proc_misc6_fail;
 815
 816         if (ac6_proc_init())
 817                 goto proc_anycast6_fail;
 818         if (if6_proc_init())
 819                 goto proc_if6_fail;
 820 #endif
 821         ip6_route_init();
 822         ip6_flowlabel_init();
 823         err = addrconf_init();
 824         if (err)
 825                 goto addrconf_fail;
 826         sit_init();
 827
 828         /* Init v6 extension headers. */
 829         ipv6_rthdr_init();
 830         ipv6_frag_init();
 831         ipv6_nodata_init();
 832         ipv6_destopt_init();
 833
 834         /* Init v6 transport protocols. */
 835         udpv6_init();
 836         tcpv6_init();
 837
 838         ipv6_packet_init();
 839         err = 0;
 840 out:
 841         return err;
 842
 843 addrconf_fail:
 844         ip6_flowlabel_cleanup();
 845         ip6_route_cleanup();
 846 #ifdef CONFIG_PROC_FS
 847         if6_proc_exit();
 848 proc_if6_fail:
 849         ac6_proc_exit();
 850 proc_anycast6_fail:
 851         ipv6_misc_proc_exit();
 852 proc_misc6_fail:
 853         udp6_proc_exit();
 854 proc_udp6_fail:
 855         tcp6_proc_exit();
 856 proc_tcp6_fail:
 857         raw6_proc_exit();
 858 proc_raw6_fail:
 859 #endif
 860         ipv6_netfilter_fini();
 861 netfilter_fail:
 862         igmp6_cleanup();
 863 igmp_fail:
 864         ndisc_cleanup();
 865 ndisc_fail:
 866         icmpv6_cleanup();
 867 icmp_fail:
 868 #ifdef CONFIG_SYSCTL
 869         ipv6_sysctl_unregister();
 870 #endif
 871         cleanup_ipv6_mibs();
 872 out_unregister_sock:
 873         sock_unregister(PF_INET6);
 874 out_unregister_raw_proto:
 875         proto_unregister(&rawv6_prot);
 876 out_unregister_udp_proto:
 877         proto_unregister(&udpv6_prot);
 878 out_unregister_tcp_proto:
 879         proto_unregister(&tcpv6_prot);
 880         goto out;
 881 }
 882 module_init(inet6_init);
 883
 884 static void __exit inet6_exit(void)
 885 {
 886         /* First of all disallow new sockets creation. */
 887         sock_unregister(PF_INET6);
 888 #ifdef CONFIG_PROC_FS
 889         if6_proc_exit();
 890         ac6_proc_exit();
 891         ipv6_misc_proc_exit();
 892         udp6_proc_exit();
 893         tcp6_proc_exit();
 894         raw6_proc_exit();
 895 #endif
 896         /* Cleanup code parts. */
 897         sit_cleanup();
 898         ip6_flowlabel_cleanup();
 899         addrconf_cleanup();
 900         ip6_route_cleanup();
 901         ipv6_packet_cleanup();
 902         igmp6_cleanup();
 903         ipv6_netfilter_fini();
 904         ndisc_cleanup();
 905         icmpv6_cleanup();
 906 #ifdef CONFIG_SYSCTL
 907         ipv6_sysctl_unregister();
 908 #endif
 909         cleanup_ipv6_mibs();
 910         proto_unregister(&rawv6_prot);
 911         proto_unregister(&udpv6_prot);
 912         proto_unregister(&tcpv6_prot);
 913 }
 914 module_exit(inet6_exit);
 915
 916 MODULE_ALIAS_NETPROTO(PF_INET6);