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 #ifdef CONFIG_IPV6_MIP6
  63 #include <net/mip6.h>
  64 #endif
  65
  66 #include <asm/uaccess.h>
  67 #include <asm/system.h>
  68
  69 MODULE_AUTHOR("Cast of dozens");
  70 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  71 MODULE_LICENSE("GPL");
  72
  73 int sysctl_ipv6_bindv6only __read_mostly;
  74
  75 /* The inetsw table contains everything that inet_create needs to
  76  * build a new socket.
  77  */
  78 static struct list_head inetsw6[SOCK_MAX];
  79 static DEFINE_SPINLOCK(inetsw6_lock);
  80
  81 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  82 {
  83         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  84
  85         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  86 }
  87
  88 static int inet6_create(struct socket *sock, int protocol)
  89 {
  90         struct inet_sock *inet;
  91         struct ipv6_pinfo *np;
  92         struct sock *sk;
  93         struct list_head *p;
  94         struct inet_protosw *answer;
  95         struct proto *answer_prot;
  96         unsigned char answer_flags;
  97         char answer_no_check;
  98         int try_loading_module = 0;
  99         int err;
 100
 101         /* Look for the requested type/protocol pair. */
 102         answer = NULL;
 103 lookup_protocol:
 104         err = -ESOCKTNOSUPPORT;
 105         rcu_read_lock();
 106         list_for_each_rcu(p, &inetsw6[sock->type]) {
 107                 answer = list_entry(p, struct inet_protosw, list);
 108
 109                 /* Check the non-wild match. */
 110                 if (protocol == answer->protocol) {
 111                         if (protocol != IPPROTO_IP)
 112                                 break;
 113                 } else {
 114                         /* Check for the two wild cases. */
 115                         if (IPPROTO_IP == protocol) {
 116                                 protocol = answer->protocol;
 117                                 break;
 118                         }
 119                         if (IPPROTO_IP == answer->protocol)
 120                                 break;
 121                 }
 122                 err = -EPROTONOSUPPORT;
 123                 answer = NULL;
 124         }
 125
 126         if (!answer) {
 127                 if (try_loading_module < 2) {
 128                         rcu_read_unlock();
 129                         /*
 130                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 131                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 132                          */
 133                         if (++try_loading_module == 1)
 134                                 request_module("net-pf-%d-proto-%d-type-%d",
 135                                                 PF_INET6, protocol, sock->type);
 136                         /*
 137                          * Fall back to generic, e.g. net-pf-10-proto-132
 138                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 139                          */
 140                         else
 141                                 request_module("net-pf-%d-proto-%d",
 142                                                 PF_INET6, protocol);
 143                         goto lookup_protocol;
 144                 } else
 145                         goto out_rcu_unlock;
 146         }
 147
 148         err = -EPERM;
 149         if (answer->capability > 0 && !capable(answer->capability))
 150                 goto out_rcu_unlock;
 151
 152         sock->ops = answer->ops;
 153         answer_prot = answer->prot;
 154         answer_no_check = answer->no_check;
 155         answer_flags = answer->flags;
 156         rcu_read_unlock();
 157
 158         BUG_TRAP(answer_prot->slab != NULL);
 159
 160         err = -ENOBUFS;
 161         sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
 162         if (sk == NULL)
 163                 goto out;
 164
 165         sock_init_data(sock, sk);
 166
 167         err = 0;
 168         sk->sk_no_check = answer_no_check;
 169         if (INET_PROTOSW_REUSE & answer_flags)
 170                 sk->sk_reuse = 1;
 171
 172         inet = inet_sk(sk);
 173         inet->is_icsk = INET_PROTOSW_ICSK & answer_flags;
 174
 175         if (SOCK_RAW == sock->type) {
 176                 inet->num = protocol;
 177                 if (IPPROTO_RAW == protocol)
 178                         inet->hdrincl = 1;
 179         }
 180
 181         sk->sk_destruct         = inet_sock_destruct;
 182         sk->sk_family           = PF_INET6;
 183         sk->sk_protocol         = protocol;
 184
 185         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 186
 187         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 188         np->hop_limit   = -1;
 189         np->mcast_hops  = -1;
 190         np->mc_loop     = 1;
 191         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 192         np->ipv6only    = sysctl_ipv6_bindv6only;
 193
 194         /* Init the ipv4 part of the socket since we can have sockets
 195          * using v6 API for ipv4.
 196          */
 197         inet->uc_ttl    = -1;
 198
 199         inet->mc_loop   = 1;
 200         inet->mc_ttl    = 1;
 201         inet->mc_index  = 0;
 202         inet->mc_list   = NULL;
 203
 204         if (ipv4_config.no_pmtu_disc)
 205                 inet->pmtudisc = IP_PMTUDISC_DONT;
 206         else
 207                 inet->pmtudisc = IP_PMTUDISC_WANT;
 208         /*
 209          * Increment only the relevant sk_prot->socks debug field, this changes
 210          * the previous behaviour of incrementing both the equivalent to
 211          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 212          *
 213          * This allows better debug granularity as we'll know exactly how many
 214          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 215          * transport protocol socks. -acme
 216          */
 217         sk_refcnt_debug_inc(sk);
 218
 219         if (inet->num) {
 220                 /* It assumes that any protocol which allows
 221                  * the user to assign a number at socket
 222                  * creation time automatically shares.
 223                  */
 224                 inet->sport = ntohs(inet->num);
 225                 sk->sk_prot->hash(sk);
 226         }
 227         if (sk->sk_prot->init) {
 228                 err = sk->sk_prot->init(sk);
 229                 if (err) {
 230                         sk_common_release(sk);
 231                         goto out;
 232                 }
 233         }
 234 out:
 235         return err;
 236 out_rcu_unlock:
 237         rcu_read_unlock();
 238         goto out;
 239 }
 240
 241
 242 /* bind for INET6 API */
 243 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 244 {
 245         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 246         struct sock *sk = sock->sk;
 247         struct inet_sock *inet = inet_sk(sk);
 248         struct ipv6_pinfo *np = inet6_sk(sk);
 249         __be32 v4addr = 0;
 250         unsigned short snum;
 251         int addr_type = 0;
 252         int err = 0;
 253
 254         /* If the socket has its own bind function then use it. */
 255         if (sk->sk_prot->bind)
 256                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 257
 258         if (addr_len < SIN6_LEN_RFC2133)
 259                 return -EINVAL;
 260         addr_type = ipv6_addr_type(&addr->sin6_addr);
 261         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 262                 return -EINVAL;
 263
 264         snum = ntohs(addr->sin6_port);
 265         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 266                 return -EACCES;
 267
 268         lock_sock(sk);
 269
 270         /* Check these errors (active socket, double bind). */
 271         if (sk->sk_state != TCP_CLOSE || inet->num) {
 272                 err = -EINVAL;
 273                 goto out;
 274         }
 275
 276         /* Check if the address belongs to the host. */
 277         if (addr_type == IPV6_ADDR_MAPPED) {
 278                 v4addr = addr->sin6_addr.s6_addr32[3];
 279                 if (inet_addr_type(v4addr) != RTN_LOCAL) {
 280                         err = -EADDRNOTAVAIL;
 281                         goto out;
 282                 }
 283         } else {
 284                 if (addr_type != IPV6_ADDR_ANY) {
 285                         struct net_device *dev = NULL;
 286
 287                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 288                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 289                                     addr->sin6_scope_id) {
 290                                         /* Override any existing binding, if another one
 291                                          * is supplied by user.
 292                                          */
 293                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 294                                 }
 295
 296                                 /* Binding to link-local address requires an interface */
 297                                 if (!sk->sk_bound_dev_if) {
 298                                         err = -EINVAL;
 299                                         goto out;
 300                                 }
 301                                 dev = dev_get_by_index(sk->sk_bound_dev_if);
 302                                 if (!dev) {
 303                                         err = -ENODEV;
 304                                         goto out;
 305                                 }
 306                         }
 307
 308                         /* ipv4 addr of the socket is invalid.  Only the
 309                          * unspecified and mapped address have a v4 equivalent.
 310                          */
 311                         v4addr = LOOPBACK4_IPV6;
 312                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 313                                 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
 314                                         if (dev)
 315                                                 dev_put(dev);
 316                                         err = -EADDRNOTAVAIL;
 317                                         goto out;
 318                                 }
 319                         }
 320                         if (dev)
 321                                 dev_put(dev);
 322                 }
 323         }
 324
 325         inet->rcv_saddr = v4addr;
 326         inet->saddr = v4addr;
 327
 328         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 329
 330         if (!(addr_type & IPV6_ADDR_MULTICAST))
 331                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 332
 333         /* Make sure we are allowed to bind here. */
 334         if (sk->sk_prot->get_port(sk, snum)) {
 335                 inet_reset_saddr(sk);
 336                 err = -EADDRINUSE;
 337                 goto out;
 338         }
 339
 340         if (addr_type != IPV6_ADDR_ANY)
 341                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 342         if (snum)
 343                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 344         inet->sport = ntohs(inet->num);
 345         inet->dport = 0;
 346         inet->daddr = 0;
 347 out:
 348         release_sock(sk);
 349         return err;
 350 }
 351
 352 int inet6_release(struct socket *sock)
 353 {
 354         struct sock *sk = sock->sk;
 355
 356         if (sk == NULL)
 357                 return -EINVAL;
 358
 359         /* Free mc lists */
 360         ipv6_sock_mc_close(sk);
 361
 362         /* Free ac lists */
 363         ipv6_sock_ac_close(sk);
 364
 365         return inet_release(sock);
 366 }
 367
 368 int inet6_destroy_sock(struct sock *sk)
 369 {
 370         struct ipv6_pinfo *np = inet6_sk(sk);
 371         struct sk_buff *skb;
 372         struct ipv6_txoptions *opt;
 373
 374         /* Release rx options */
 375
 376         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 377                 kfree_skb(skb);
 378
 379         /* Free flowlabels */
 380         fl6_free_socklist(sk);
 381
 382         /* Free tx options */
 383
 384         if ((opt = xchg(&np->opt, NULL)) != NULL)
 385                 sock_kfree_s(sk, opt, opt->tot_len);
 386
 387         return 0;
 388 }
 389
 390 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 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
 435         switch(cmd)
 436         {
 437         case SIOCGSTAMP:
 438                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 439
 440         case SIOCADDRT:
 441         case SIOCDELRT:
 442
 443                 return(ipv6_route_ioctl(cmd,(void __user *)arg));
 444
 445         case SIOCSIFADDR:
 446                 return addrconf_add_ifaddr((void __user *) arg);
 447         case SIOCDIFADDR:
 448                 return addrconf_del_ifaddr((void __user *) arg);
 449         case SIOCSIFDSTADDR:
 450                 return addrconf_set_dstaddr((void __user *) arg);
 451         default:
 452                 if (!sk->sk_prot->ioctl)
 453                         return -ENOIOCTLCMD;
 454                 return sk->sk_prot->ioctl(sk, cmd, arg);
 455         }
 456         /*NOTREACHED*/
 457         return(0);
 458 }
 459
 460 const struct proto_ops inet6_stream_ops = {
 461         .family            = PF_INET6,
 462         .owner             = THIS_MODULE,
 463         .release           = inet6_release,
 464         .bind              = inet6_bind,
 465         .connect           = inet_stream_connect,       /* ok           */
 466         .socketpair        = sock_no_socketpair,        /* a do nothing */
 467         .accept            = inet_accept,               /* ok           */
 468         .getname           = inet6_getname,
 469         .poll              = tcp_poll,                  /* ok           */
 470         .ioctl             = inet6_ioctl,               /* must change  */
 471         .listen            = inet_listen,               /* ok           */
 472         .shutdown          = inet_shutdown,             /* ok           */
 473         .setsockopt        = sock_common_setsockopt,    /* ok           */
 474         .getsockopt        = sock_common_getsockopt,    /* ok           */
 475         .sendmsg           = inet_sendmsg,              /* ok           */
 476         .recvmsg           = sock_common_recvmsg,       /* ok           */
 477         .mmap              = sock_no_mmap,
 478         .sendpage          = tcp_sendpage,
 479 #ifdef CONFIG_COMPAT
 480         .compat_setsockopt = compat_sock_common_setsockopt,
 481         .compat_getsockopt = compat_sock_common_getsockopt,
 482 #endif
 483 };
 484
 485 const struct proto_ops inet6_dgram_ops = {
 486         .family            = PF_INET6,
 487         .owner             = THIS_MODULE,
 488         .release           = inet6_release,
 489         .bind              = inet6_bind,
 490         .connect           = inet_dgram_connect,        /* ok           */
 491         .socketpair        = sock_no_socketpair,        /* a do nothing */
 492         .accept            = sock_no_accept,            /* a do nothing */
 493         .getname           = inet6_getname,
 494         .poll              = udp_poll,                  /* ok           */
 495         .ioctl             = inet6_ioctl,               /* must change  */
 496         .listen            = sock_no_listen,            /* ok           */
 497         .shutdown          = inet_shutdown,             /* ok           */
 498         .setsockopt        = sock_common_setsockopt,    /* ok           */
 499         .getsockopt        = sock_common_getsockopt,    /* ok           */
 500         .sendmsg           = inet_sendmsg,              /* ok           */
 501         .recvmsg           = sock_common_recvmsg,       /* ok           */
 502         .mmap              = sock_no_mmap,
 503         .sendpage          = sock_no_sendpage,
 504 #ifdef CONFIG_COMPAT
 505         .compat_setsockopt = compat_sock_common_setsockopt,
 506         .compat_getsockopt = compat_sock_common_getsockopt,
 507 #endif
 508 };
 509
 510 static struct net_proto_family inet6_family_ops = {
 511         .family = PF_INET6,
 512         .create = inet6_create,
 513         .owner  = THIS_MODULE,
 514 };
 515
 516 /* Same as inet6_dgram_ops, sans udp_poll.  */
 517 static const struct proto_ops inet6_sockraw_ops = {
 518         .family            = PF_INET6,
 519         .owner             = THIS_MODULE,
 520         .release           = inet6_release,
 521         .bind              = inet6_bind,
 522         .connect           = inet_dgram_connect,        /* ok           */
 523         .socketpair        = sock_no_socketpair,        /* a do nothing */
 524         .accept            = sock_no_accept,            /* a do nothing */
 525         .getname           = inet6_getname,
 526         .poll              = datagram_poll,             /* ok           */
 527         .ioctl             = inet6_ioctl,               /* must change  */
 528         .listen            = sock_no_listen,            /* ok           */
 529         .shutdown          = inet_shutdown,             /* ok           */
 530         .setsockopt        = sock_common_setsockopt,    /* ok           */
 531         .getsockopt        = sock_common_getsockopt,    /* ok           */
 532         .sendmsg           = inet_sendmsg,              /* ok           */
 533         .recvmsg           = sock_common_recvmsg,       /* ok           */
 534         .mmap              = sock_no_mmap,
 535         .sendpage          = sock_no_sendpage,
 536 #ifdef CONFIG_COMPAT
 537         .compat_setsockopt = compat_sock_common_setsockopt,
 538         .compat_getsockopt = compat_sock_common_getsockopt,
 539 #endif
 540 };
 541
 542 static struct inet_protosw rawv6_protosw = {
 543         .type           = SOCK_RAW,
 544         .protocol       = IPPROTO_IP,   /* wild card */
 545         .prot           = &rawv6_prot,
 546         .ops            = &inet6_sockraw_ops,
 547         .capability     = CAP_NET_RAW,
 548         .no_check       = UDP_CSUM_DEFAULT,
 549         .flags          = INET_PROTOSW_REUSE,
 550 };
 551
 552 void
 553 inet6_register_protosw(struct inet_protosw *p)
 554 {
 555         struct list_head *lh;
 556         struct inet_protosw *answer;
 557         int protocol = p->protocol;
 558         struct list_head *last_perm;
 559
 560         spin_lock_bh(&inetsw6_lock);
 561
 562         if (p->type >= SOCK_MAX)
 563                 goto out_illegal;
 564
 565         /* If we are trying to override a permanent protocol, bail. */
 566         answer = NULL;
 567         last_perm = &inetsw6[p->type];
 568         list_for_each(lh, &inetsw6[p->type]) {
 569                 answer = list_entry(lh, struct inet_protosw, list);
 570
 571                 /* Check only the non-wild match. */
 572                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 573                         if (protocol == answer->protocol)
 574                                 break;
 575                         last_perm = lh;
 576                 }
 577
 578                 answer = NULL;
 579         }
 580         if (answer)
 581                 goto out_permanent;
 582
 583         /* Add the new entry after the last permanent entry if any, so that
 584          * the new entry does not override a permanent entry when matched with
 585          * a wild-card protocol. But it is allowed to override any existing
 586          * non-permanent entry.  This means that when we remove this entry, the
 587          * system automatically returns to the old behavior.
 588          */
 589         list_add_rcu(&p->list, last_perm);
 590 out:
 591         spin_unlock_bh(&inetsw6_lock);
 592         return;
 593
 594 out_permanent:
 595         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 596                protocol);
 597         goto out;
 598
 599 out_illegal:
 600         printk(KERN_ERR
 601                "Ignoring attempt to register invalid socket type %d.\n",
 602                p->type);
 603         goto out;
 604 }
 605
 606 void
 607 inet6_unregister_protosw(struct inet_protosw *p)
 608 {
 609         if (INET_PROTOSW_PERMANENT & p->flags) {
 610                 printk(KERN_ERR
 611                        "Attempt to unregister permanent protocol %d.\n",
 612                        p->protocol);
 613         } else {
 614                 spin_lock_bh(&inetsw6_lock);
 615                 list_del_rcu(&p->list);
 616                 spin_unlock_bh(&inetsw6_lock);
 617
 618                 synchronize_net();
 619         }
 620 }
 621
 622 int inet6_sk_rebuild_header(struct sock *sk)
 623 {
 624         int err;
 625         struct dst_entry *dst;
 626         struct ipv6_pinfo *np = inet6_sk(sk);
 627
 628         dst = __sk_dst_check(sk, np->dst_cookie);
 629
 630         if (dst == NULL) {
 631                 struct inet_sock *inet = inet_sk(sk);
 632                 struct in6_addr *final_p = NULL, final;
 633                 struct flowi fl;
 634
 635                 memset(&fl, 0, sizeof(fl));
 636                 fl.proto = sk->sk_protocol;
 637                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 638                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 639                 fl.fl6_flowlabel = np->flow_label;
 640                 fl.oif = sk->sk_bound_dev_if;
 641                 fl.fl_ip_dport = inet->dport;
 642                 fl.fl_ip_sport = inet->sport;
 643                 security_sk_classify_flow(sk, &fl);
 644
 645                 if (np->opt && np->opt->srcrt) {
 646                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 647                         ipv6_addr_copy(&final, &fl.fl6_dst);
 648                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 649                         final_p = &final;
 650                 }
 651
 652                 err = ip6_dst_lookup(sk, &dst, &fl);
 653                 if (err) {
 654                         sk->sk_route_caps = 0;
 655                         return err;
 656                 }
 657                 if (final_p)
 658                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 659
 660                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
 661                         sk->sk_err_soft = -err;
 662                         return err;
 663                 }
 664
 665                 __ip6_dst_store(sk, dst, NULL, NULL);
 666         }
 667
 668         return 0;
 669 }
 670
 671 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 672
 673 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 674 {
 675         struct ipv6_pinfo *np = inet6_sk(sk);
 676         struct inet6_skb_parm *opt = IP6CB(skb);
 677
 678         if (np->rxopt.all) {
 679                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 680                                   np->rxopt.bits.ohopopts)) ||
 681                     ((IPV6_FLOWINFO_MASK & *(u32*)skb->nh.raw) &&
 682                      np->rxopt.bits.rxflow) ||
 683                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 684                      np->rxopt.bits.osrcrt)) ||
 685                     ((opt->dst1 || opt->dst0) &&
 686                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 687                         return 1;
 688         }
 689         return 0;
 690 }
 691
 692 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 693
 694 int
 695 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
 696 {
 697         if (ptr == NULL)
 698                 return -EINVAL;
 699
 700         ptr[0] = __alloc_percpu(mibsize);
 701         if (!ptr[0])
 702                 goto err0;
 703
 704         ptr[1] = __alloc_percpu(mibsize);
 705         if (!ptr[1])
 706                 goto err1;
 707
 708         return 0;
 709
 710 err1:
 711         free_percpu(ptr[0]);
 712         ptr[0] = NULL;
 713 err0:
 714         return -ENOMEM;
 715 }
 716
 717 void
 718 snmp6_mib_free(void *ptr[2])
 719 {
 720         if (ptr == NULL)
 721                 return;
 722         if (ptr[0])
 723                 free_percpu(ptr[0]);
 724         if (ptr[1])
 725                 free_percpu(ptr[1]);
 726         ptr[0] = ptr[1] = NULL;
 727 }
 728
 729 static int __init init_ipv6_mibs(void)
 730 {
 731         if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
 732                            __alignof__(struct ipstats_mib)) < 0)
 733                 goto err_ip_mib;
 734         if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
 735                            __alignof__(struct icmpv6_mib)) < 0)
 736                 goto err_icmp_mib;
 737         if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
 738                            __alignof__(struct udp_mib)) < 0)
 739                 goto err_udp_mib;
 740         return 0;
 741
 742 err_udp_mib:
 743         snmp6_mib_free((void **)icmpv6_statistics);
 744 err_icmp_mib:
 745         snmp6_mib_free((void **)ipv6_statistics);
 746 err_ip_mib:
 747         return -ENOMEM;
 748
 749 }
 750
 751 static void cleanup_ipv6_mibs(void)
 752 {
 753         snmp6_mib_free((void **)ipv6_statistics);
 754         snmp6_mib_free((void **)icmpv6_statistics);
 755         snmp6_mib_free((void **)udp_stats_in6);
 756 }
 757
 758 static int __init inet6_init(void)
 759 {
 760         struct sk_buff *dummy_skb;
 761         struct list_head *r;
 762         int err;
 763
 764         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
 765
 766 #ifdef MODULE
 767 #if 0 /* FIXME --RR */
 768         if (!mod_member_present(&__this_module, can_unload))
 769           return -EINVAL;
 770
 771         __this_module.can_unload = &ipv6_unload;
 772 #endif
 773 #endif
 774
 775         err = proto_register(&tcpv6_prot, 1);
 776         if (err)
 777                 goto out;
 778
 779         err = proto_register(&udpv6_prot, 1);
 780         if (err)
 781                 goto out_unregister_tcp_proto;
 782
 783         err = proto_register(&rawv6_prot, 1);
 784         if (err)
 785                 goto out_unregister_udp_proto;
 786
 787
 788         /* Register the socket-side information for inet6_create.  */
 789         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 790                 INIT_LIST_HEAD(r);
 791
 792         /* We MUST register RAW sockets before we create the ICMP6,
 793          * IGMP6, or NDISC control sockets.
 794          */
 795         inet6_register_protosw(&rawv6_protosw);
 796
 797         /* Register the family here so that the init calls below will
 798          * be able to create sockets. (?? is this dangerous ??)
 799          */
 800         err = sock_register(&inet6_family_ops);
 801         if (err)
 802                 goto out_unregister_raw_proto;
 803
 804         /* Initialise ipv6 mibs */
 805         err = init_ipv6_mibs();
 806         if (err)
 807                 goto out_unregister_sock;
 808
 809         /*
 810          *      ipngwg API draft makes clear that the correct semantics
 811          *      for TCP and UDP is to consider one TCP and UDP instance
 812          *      in a host availiable by both INET and INET6 APIs and
 813          *      able to communicate via both network protocols.
 814          */
 815
 816 #ifdef CONFIG_SYSCTL
 817         ipv6_sysctl_register();
 818 #endif
 819         err = icmpv6_init(&inet6_family_ops);
 820         if (err)
 821                 goto icmp_fail;
 822         err = ndisc_init(&inet6_family_ops);
 823         if (err)
 824                 goto ndisc_fail;
 825         err = igmp6_init(&inet6_family_ops);
 826         if (err)
 827                 goto igmp_fail;
 828         err = ipv6_netfilter_init();
 829         if (err)
 830                 goto netfilter_fail;
 831         /* Create /proc/foo6 entries. */
 832 #ifdef CONFIG_PROC_FS
 833         err = -ENOMEM;
 834         if (raw6_proc_init())
 835                 goto proc_raw6_fail;
 836         if (tcp6_proc_init())
 837                 goto proc_tcp6_fail;
 838         if (udp6_proc_init())
 839                 goto proc_udp6_fail;
 840         if (ipv6_misc_proc_init())
 841                 goto proc_misc6_fail;
 842
 843         if (ac6_proc_init())
 844                 goto proc_anycast6_fail;
 845         if (if6_proc_init())
 846                 goto proc_if6_fail;
 847 #endif
 848         ip6_route_init();
 849         ip6_flowlabel_init();
 850         err = addrconf_init();
 851         if (err)
 852                 goto addrconf_fail;
 853
 854         /* Init v6 extension headers. */
 855         ipv6_rthdr_init();
 856         ipv6_frag_init();
 857         ipv6_nodata_init();
 858         ipv6_destopt_init();
 859 #ifdef CONFIG_IPV6_MIP6
 860         mip6_init();
 861 #endif
 862
 863         /* Init v6 transport protocols. */
 864         udpv6_init();
 865         tcpv6_init();
 866
 867         ipv6_packet_init();
 868         err = 0;
 869 out:
 870         return err;
 871
 872 addrconf_fail:
 873         ip6_flowlabel_cleanup();
 874         ip6_route_cleanup();
 875 #ifdef CONFIG_PROC_FS
 876         if6_proc_exit();
 877 proc_if6_fail:
 878         ac6_proc_exit();
 879 proc_anycast6_fail:
 880         ipv6_misc_proc_exit();
 881 proc_misc6_fail:
 882         udp6_proc_exit();
 883 proc_udp6_fail:
 884         tcp6_proc_exit();
 885 proc_tcp6_fail:
 886         raw6_proc_exit();
 887 proc_raw6_fail:
 888 #endif
 889         ipv6_netfilter_fini();
 890 netfilter_fail:
 891         igmp6_cleanup();
 892 igmp_fail:
 893         ndisc_cleanup();
 894 ndisc_fail:
 895         icmpv6_cleanup();
 896 icmp_fail:
 897 #ifdef CONFIG_SYSCTL
 898         ipv6_sysctl_unregister();
 899 #endif
 900         cleanup_ipv6_mibs();
 901 out_unregister_sock:
 902         sock_unregister(PF_INET6);
 903 out_unregister_raw_proto:
 904         proto_unregister(&rawv6_prot);
 905 out_unregister_udp_proto:
 906         proto_unregister(&udpv6_prot);
 907 out_unregister_tcp_proto:
 908         proto_unregister(&tcpv6_prot);
 909         goto out;
 910 }
 911 module_init(inet6_init);
 912
 913 static void __exit inet6_exit(void)
 914 {
 915         /* First of all disallow new sockets creation. */
 916         sock_unregister(PF_INET6);
 917 #ifdef CONFIG_PROC_FS
 918         if6_proc_exit();
 919         ac6_proc_exit();
 920         ipv6_misc_proc_exit();
 921         udp6_proc_exit();
 922         tcp6_proc_exit();
 923         raw6_proc_exit();
 924 #endif
 925 #ifdef CONFIG_IPV6_MIP6
 926         mip6_fini();
 927 #endif
 928         /* Cleanup code parts. */
 929         ip6_flowlabel_cleanup();
 930         addrconf_cleanup();
 931         ip6_route_cleanup();
 932         ipv6_packet_cleanup();
 933         igmp6_cleanup();
 934         ipv6_netfilter_fini();
 935         ndisc_cleanup();
 936         icmpv6_cleanup();
 937 #ifdef CONFIG_SYSCTL
 938         ipv6_sysctl_unregister();
 939 #endif
 940         cleanup_ipv6_mibs();
 941         proto_unregister(&rawv6_prot);
 942         proto_unregister(&udpv6_prot);
 943         proto_unregister(&tcpv6_prot);
 944 }
 945 module_exit(inet6_exit);
 946
 947 MODULE_ALIAS_NETPROTO(PF_INET6);