net/ipv6/af_inet6.c

   1 /*
   2  *      PF_INET6 socket protocol family
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Adapted from linux/net/ipv4/af_inet.c
   9  *
  10  *      $Id: af_inet6.c,v 1.66 2002/02/01 22:01:04 davem Exp $
  11  *
  12  *      Fixes:
  13  *      piggy, Karl Knutson     :       Socket protocol table
  14  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
  15  *      Arnaldo Melo            :       check proc_net_create return, cleanups
  16  *
  17  *      This program is free software; you can redistribute it and/or
  18  *      modify it under the terms of the GNU General Public License
  19  *      as published by the Free Software Foundation; either version
  20  *      2 of the License, or (at your option) any later version.
  21  */
  22
  23
  24 #include <linux/module.h>
  25 #include <linux/capability.h>
  26 #include <linux/errno.h>
  27 #include <linux/types.h>
  28 #include <linux/socket.h>
  29 #include <linux/in.h>
  30 #include <linux/kernel.h>
  31 #include <linux/timer.h>
  32 #include <linux/string.h>
  33 #include <linux/sockios.h>
  34 #include <linux/net.h>
  35 #include <linux/fcntl.h>
  36 #include <linux/mm.h>
  37 #include <linux/interrupt.h>
  38 #include <linux/proc_fs.h>
  39 #include <linux/stat.h>
  40 #include <linux/init.h>
  41
  42 #include <linux/inet.h>
  43 #include <linux/netdevice.h>
  44 #include <linux/icmpv6.h>
  45 #include <linux/smp_lock.h>
  46 #include <linux/netfilter_ipv6.h>
  47
  48 #include <net/ip.h>
  49 #include <net/ipv6.h>
  50 #include <net/udp.h>
  51 #include <net/udplite.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) != 0;
 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 = htons(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 = htons(inet->num);
 345         inet->dport = 0;
 346         inet->daddr = 0;
 347 out:
 348         release_sock(sk);
 349         return err;
 350 }
 351
 352 EXPORT_SYMBOL(inet6_bind);
 353
 354 int inet6_release(struct socket *sock)
 355 {
 356         struct sock *sk = sock->sk;
 357
 358         if (sk == NULL)
 359                 return -EINVAL;
 360
 361         /* Free mc lists */
 362         ipv6_sock_mc_close(sk);
 363
 364         /* Free ac lists */
 365         ipv6_sock_ac_close(sk);
 366
 367         return inet_release(sock);
 368 }
 369
 370 EXPORT_SYMBOL(inet6_release);
 371
 372 int inet6_destroy_sock(struct sock *sk)
 373 {
 374         struct ipv6_pinfo *np = inet6_sk(sk);
 375         struct sk_buff *skb;
 376         struct ipv6_txoptions *opt;
 377
 378         /* Release rx options */
 379
 380         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 381                 kfree_skb(skb);
 382
 383         /* Free flowlabels */
 384         fl6_free_socklist(sk);
 385
 386         /* Free tx options */
 387
 388         if ((opt = xchg(&np->opt, NULL)) != NULL)
 389                 sock_kfree_s(sk, opt, opt->tot_len);
 390
 391         return 0;
 392 }
 393
 394 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 395
 396 /*
 397  *      This does both peername and sockname.
 398  */
 399
 400 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 401                  int *uaddr_len, int peer)
 402 {
 403         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 404         struct sock *sk = sock->sk;
 405         struct inet_sock *inet = inet_sk(sk);
 406         struct ipv6_pinfo *np = inet6_sk(sk);
 407
 408         sin->sin6_family = AF_INET6;
 409         sin->sin6_flowinfo = 0;
 410         sin->sin6_scope_id = 0;
 411         if (peer) {
 412                 if (!inet->dport)
 413                         return -ENOTCONN;
 414                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 415                     peer == 1)
 416                         return -ENOTCONN;
 417                 sin->sin6_port = inet->dport;
 418                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 419                 if (np->sndflow)
 420                         sin->sin6_flowinfo = np->flow_label;
 421         } else {
 422                 if (ipv6_addr_any(&np->rcv_saddr))
 423                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 424                 else
 425                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 426
 427                 sin->sin6_port = inet->sport;
 428         }
 429         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 430                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 431         *uaddr_len = sizeof(*sin);
 432         return(0);
 433 }
 434
 435 EXPORT_SYMBOL(inet6_getname);
 436
 437 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 438 {
 439         struct sock *sk = sock->sk;
 440
 441         switch(cmd)
 442         {
 443         case SIOCGSTAMP:
 444                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 445
 446         case SIOCADDRT:
 447         case SIOCDELRT:
 448
 449                 return(ipv6_route_ioctl(cmd,(void __user *)arg));
 450
 451         case SIOCSIFADDR:
 452                 return addrconf_add_ifaddr((void __user *) arg);
 453         case SIOCDIFADDR:
 454                 return addrconf_del_ifaddr((void __user *) arg);
 455         case SIOCSIFDSTADDR:
 456                 return addrconf_set_dstaddr((void __user *) arg);
 457         default:
 458                 if (!sk->sk_prot->ioctl)
 459                         return -ENOIOCTLCMD;
 460                 return sk->sk_prot->ioctl(sk, cmd, arg);
 461         }
 462         /*NOTREACHED*/
 463         return(0);
 464 }
 465
 466 EXPORT_SYMBOL(inet6_ioctl);
 467
 468 const struct proto_ops inet6_stream_ops = {
 469         .family            = PF_INET6,
 470         .owner             = THIS_MODULE,
 471         .release           = inet6_release,
 472         .bind              = inet6_bind,
 473         .connect           = inet_stream_connect,       /* ok           */
 474         .socketpair        = sock_no_socketpair,        /* a do nothing */
 475         .accept            = inet_accept,               /* ok           */
 476         .getname           = inet6_getname,
 477         .poll              = tcp_poll,                  /* ok           */
 478         .ioctl             = inet6_ioctl,               /* must change  */
 479         .listen            = inet_listen,               /* ok           */
 480         .shutdown          = inet_shutdown,             /* ok           */
 481         .setsockopt        = sock_common_setsockopt,    /* ok           */
 482         .getsockopt        = sock_common_getsockopt,    /* ok           */
 483         .sendmsg           = inet_sendmsg,              /* ok           */
 484         .recvmsg           = sock_common_recvmsg,       /* ok           */
 485         .mmap              = sock_no_mmap,
 486         .sendpage          = tcp_sendpage,
 487 #ifdef CONFIG_COMPAT
 488         .compat_setsockopt = compat_sock_common_setsockopt,
 489         .compat_getsockopt = compat_sock_common_getsockopt,
 490 #endif
 491 };
 492
 493 const struct proto_ops inet6_dgram_ops = {
 494         .family            = PF_INET6,
 495         .owner             = THIS_MODULE,
 496         .release           = inet6_release,
 497         .bind              = inet6_bind,
 498         .connect           = inet_dgram_connect,        /* ok           */
 499         .socketpair        = sock_no_socketpair,        /* a do nothing */
 500         .accept            = sock_no_accept,            /* a do nothing */
 501         .getname           = inet6_getname,
 502         .poll              = udp_poll,                  /* ok           */
 503         .ioctl             = inet6_ioctl,               /* must change  */
 504         .listen            = sock_no_listen,            /* ok           */
 505         .shutdown          = inet_shutdown,             /* ok           */
 506         .setsockopt        = sock_common_setsockopt,    /* ok           */
 507         .getsockopt        = sock_common_getsockopt,    /* ok           */
 508         .sendmsg           = inet_sendmsg,              /* ok           */
 509         .recvmsg           = sock_common_recvmsg,       /* ok           */
 510         .mmap              = sock_no_mmap,
 511         .sendpage          = sock_no_sendpage,
 512 #ifdef CONFIG_COMPAT
 513         .compat_setsockopt = compat_sock_common_setsockopt,
 514         .compat_getsockopt = compat_sock_common_getsockopt,
 515 #endif
 516 };
 517
 518 static struct net_proto_family inet6_family_ops = {
 519         .family = PF_INET6,
 520         .create = inet6_create,
 521         .owner  = THIS_MODULE,
 522 };
 523
 524 /* Same as inet6_dgram_ops, sans udp_poll.  */
 525 static const struct proto_ops inet6_sockraw_ops = {
 526         .family            = PF_INET6,
 527         .owner             = THIS_MODULE,
 528         .release           = inet6_release,
 529         .bind              = inet6_bind,
 530         .connect           = inet_dgram_connect,        /* ok           */
 531         .socketpair        = sock_no_socketpair,        /* a do nothing */
 532         .accept            = sock_no_accept,            /* a do nothing */
 533         .getname           = inet6_getname,
 534         .poll              = datagram_poll,             /* ok           */
 535         .ioctl             = inet6_ioctl,               /* must change  */
 536         .listen            = sock_no_listen,            /* ok           */
 537         .shutdown          = inet_shutdown,             /* ok           */
 538         .setsockopt        = sock_common_setsockopt,    /* ok           */
 539         .getsockopt        = sock_common_getsockopt,    /* ok           */
 540         .sendmsg           = inet_sendmsg,              /* ok           */
 541         .recvmsg           = sock_common_recvmsg,       /* ok           */
 542         .mmap              = sock_no_mmap,
 543         .sendpage          = sock_no_sendpage,
 544 #ifdef CONFIG_COMPAT
 545         .compat_setsockopt = compat_sock_common_setsockopt,
 546         .compat_getsockopt = compat_sock_common_getsockopt,
 547 #endif
 548 };
 549
 550 static struct inet_protosw rawv6_protosw = {
 551         .type           = SOCK_RAW,
 552         .protocol       = IPPROTO_IP,   /* wild card */
 553         .prot           = &rawv6_prot,
 554         .ops            = &inet6_sockraw_ops,
 555         .capability     = CAP_NET_RAW,
 556         .no_check       = UDP_CSUM_DEFAULT,
 557         .flags          = INET_PROTOSW_REUSE,
 558 };
 559
 560 void
 561 inet6_register_protosw(struct inet_protosw *p)
 562 {
 563         struct list_head *lh;
 564         struct inet_protosw *answer;
 565         int protocol = p->protocol;
 566         struct list_head *last_perm;
 567
 568         spin_lock_bh(&inetsw6_lock);
 569
 570         if (p->type >= SOCK_MAX)
 571                 goto out_illegal;
 572
 573         /* If we are trying to override a permanent protocol, bail. */
 574         answer = NULL;
 575         last_perm = &inetsw6[p->type];
 576         list_for_each(lh, &inetsw6[p->type]) {
 577                 answer = list_entry(lh, struct inet_protosw, list);
 578
 579                 /* Check only the non-wild match. */
 580                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 581                         if (protocol == answer->protocol)
 582                                 break;
 583                         last_perm = lh;
 584                 }
 585
 586                 answer = NULL;
 587         }
 588         if (answer)
 589                 goto out_permanent;
 590
 591         /* Add the new entry after the last permanent entry if any, so that
 592          * the new entry does not override a permanent entry when matched with
 593          * a wild-card protocol. But it is allowed to override any existing
 594          * non-permanent entry.  This means that when we remove this entry, the
 595          * system automatically returns to the old behavior.
 596          */
 597         list_add_rcu(&p->list, last_perm);
 598 out:
 599         spin_unlock_bh(&inetsw6_lock);
 600         return;
 601
 602 out_permanent:
 603         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 604                protocol);
 605         goto out;
 606
 607 out_illegal:
 608         printk(KERN_ERR
 609                "Ignoring attempt to register invalid socket type %d.\n",
 610                p->type);
 611         goto out;
 612 }
 613
 614 EXPORT_SYMBOL(inet6_register_protosw);
 615
 616 void
 617 inet6_unregister_protosw(struct inet_protosw *p)
 618 {
 619         if (INET_PROTOSW_PERMANENT & p->flags) {
 620                 printk(KERN_ERR
 621                        "Attempt to unregister permanent protocol %d.\n",
 622                        p->protocol);
 623         } else {
 624                 spin_lock_bh(&inetsw6_lock);
 625                 list_del_rcu(&p->list);
 626                 spin_unlock_bh(&inetsw6_lock);
 627
 628                 synchronize_net();
 629         }
 630 }
 631
 632 EXPORT_SYMBOL(inet6_unregister_protosw);
 633
 634 int inet6_sk_rebuild_header(struct sock *sk)
 635 {
 636         int err;
 637         struct dst_entry *dst;
 638         struct ipv6_pinfo *np = inet6_sk(sk);
 639
 640         dst = __sk_dst_check(sk, np->dst_cookie);
 641
 642         if (dst == NULL) {
 643                 struct inet_sock *inet = inet_sk(sk);
 644                 struct in6_addr *final_p = NULL, final;
 645                 struct flowi fl;
 646
 647                 memset(&fl, 0, sizeof(fl));
 648                 fl.proto = sk->sk_protocol;
 649                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 650                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 651                 fl.fl6_flowlabel = np->flow_label;
 652                 fl.oif = sk->sk_bound_dev_if;
 653                 fl.fl_ip_dport = inet->dport;
 654                 fl.fl_ip_sport = inet->sport;
 655                 security_sk_classify_flow(sk, &fl);
 656
 657                 if (np->opt && np->opt->srcrt) {
 658                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 659                         ipv6_addr_copy(&final, &fl.fl6_dst);
 660                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 661                         final_p = &final;
 662                 }
 663
 664                 err = ip6_dst_lookup(sk, &dst, &fl);
 665                 if (err) {
 666                         sk->sk_route_caps = 0;
 667                         return err;
 668                 }
 669                 if (final_p)
 670                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 671
 672                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
 673                         sk->sk_err_soft = -err;
 674                         return err;
 675                 }
 676
 677                 __ip6_dst_store(sk, dst, NULL, NULL);
 678         }
 679
 680         return 0;
 681 }
 682
 683 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 684
 685 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 686 {
 687         struct ipv6_pinfo *np = inet6_sk(sk);
 688         struct inet6_skb_parm *opt = IP6CB(skb);
 689
 690         if (np->rxopt.all) {
 691                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 692                                   np->rxopt.bits.ohopopts)) ||
 693                     ((IPV6_FLOWINFO_MASK & *(__be32*)skb->nh.raw) &&
 694                      np->rxopt.bits.rxflow) ||
 695                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 696                      np->rxopt.bits.osrcrt)) ||
 697                     ((opt->dst1 || opt->dst0) &&
 698                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 699                         return 1;
 700         }
 701         return 0;
 702 }
 703
 704 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 705
 706 int
 707 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
 708 {
 709         if (ptr == NULL)
 710                 return -EINVAL;
 711
 712         ptr[0] = __alloc_percpu(mibsize);
 713         if (!ptr[0])
 714                 goto err0;
 715
 716         ptr[1] = __alloc_percpu(mibsize);
 717         if (!ptr[1])
 718                 goto err1;
 719
 720         return 0;
 721
 722 err1:
 723         free_percpu(ptr[0]);
 724         ptr[0] = NULL;
 725 err0:
 726         return -ENOMEM;
 727 }
 728
 729 void
 730 snmp6_mib_free(void *ptr[2])
 731 {
 732         if (ptr == NULL)
 733                 return;
 734         free_percpu(ptr[0]);
 735         free_percpu(ptr[1]);
 736         ptr[0] = ptr[1] = NULL;
 737 }
 738
 739 static int __init init_ipv6_mibs(void)
 740 {
 741         if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
 742                            __alignof__(struct ipstats_mib)) < 0)
 743                 goto err_ip_mib;
 744         if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
 745                            __alignof__(struct icmpv6_mib)) < 0)
 746                 goto err_icmp_mib;
 747         if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
 748                            __alignof__(struct udp_mib)) < 0)
 749                 goto err_udp_mib;
 750         if (snmp6_mib_init((void **)udplite_stats_in6, sizeof (struct udp_mib),
 751                            __alignof__(struct udp_mib)) < 0)
 752                 goto err_udplite_mib;
 753         return 0;
 754
 755 err_udplite_mib:
 756         snmp6_mib_free((void **)udp_stats_in6);
 757 err_udp_mib:
 758         snmp6_mib_free((void **)icmpv6_statistics);
 759 err_icmp_mib:
 760         snmp6_mib_free((void **)ipv6_statistics);
 761 err_ip_mib:
 762         return -ENOMEM;
 763
 764 }
 765
 766 static void cleanup_ipv6_mibs(void)
 767 {
 768         snmp6_mib_free((void **)ipv6_statistics);
 769         snmp6_mib_free((void **)icmpv6_statistics);
 770         snmp6_mib_free((void **)udp_stats_in6);
 771         snmp6_mib_free((void **)udplite_stats_in6);
 772 }
 773
 774 static int __init inet6_init(void)
 775 {
 776         struct sk_buff *dummy_skb;
 777         struct list_head *r;
 778         int err;
 779
 780         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
 781
 782 #ifdef MODULE
 783 #if 0 /* FIXME --RR */
 784         if (!mod_member_present(&__this_module, can_unload))
 785           return -EINVAL;
 786
 787         __this_module.can_unload = &ipv6_unload;
 788 #endif
 789 #endif
 790
 791         err = proto_register(&tcpv6_prot, 1);
 792         if (err)
 793                 goto out;
 794
 795         err = proto_register(&udpv6_prot, 1);
 796         if (err)
 797                 goto out_unregister_tcp_proto;
 798
 799         err = proto_register(&udplitev6_prot, 1);
 800         if (err)
 801                 goto out_unregister_udp_proto;
 802
 803         err = proto_register(&rawv6_prot, 1);
 804         if (err)
 805                 goto out_unregister_udplite_proto;
 806
 807
 808         /* Register the socket-side information for inet6_create.  */
 809         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 810                 INIT_LIST_HEAD(r);
 811
 812         /* We MUST register RAW sockets before we create the ICMP6,
 813          * IGMP6, or NDISC control sockets.
 814          */
 815         inet6_register_protosw(&rawv6_protosw);
 816
 817         /* Register the family here so that the init calls below will
 818          * be able to create sockets. (?? is this dangerous ??)
 819          */
 820         err = sock_register(&inet6_family_ops);
 821         if (err)
 822                 goto out_unregister_raw_proto;
 823
 824         /* Initialise ipv6 mibs */
 825         err = init_ipv6_mibs();
 826         if (err)
 827                 goto out_unregister_sock;
 828
 829         /*
 830          *      ipngwg API draft makes clear that the correct semantics
 831          *      for TCP and UDP is to consider one TCP and UDP instance
 832          *      in a host availiable by both INET and INET6 APIs and
 833          *      able to communicate via both network protocols.
 834          */
 835
 836 #ifdef CONFIG_SYSCTL
 837         ipv6_sysctl_register();
 838 #endif
 839         err = icmpv6_init(&inet6_family_ops);
 840         if (err)
 841                 goto icmp_fail;
 842         err = ndisc_init(&inet6_family_ops);
 843         if (err)
 844                 goto ndisc_fail;
 845         err = igmp6_init(&inet6_family_ops);
 846         if (err)
 847                 goto igmp_fail;
 848         err = ipv6_netfilter_init();
 849         if (err)
 850                 goto netfilter_fail;
 851         /* Create /proc/foo6 entries. */
 852 #ifdef CONFIG_PROC_FS
 853         err = -ENOMEM;
 854         if (raw6_proc_init())
 855                 goto proc_raw6_fail;
 856         if (tcp6_proc_init())
 857                 goto proc_tcp6_fail;
 858         if (udp6_proc_init())
 859                 goto proc_udp6_fail;
 860         if (udplite6_proc_init())
 861                 goto proc_udplite6_fail;
 862         if (ipv6_misc_proc_init())
 863                 goto proc_misc6_fail;
 864
 865         if (ac6_proc_init())
 866                 goto proc_anycast6_fail;
 867         if (if6_proc_init())
 868                 goto proc_if6_fail;
 869 #endif
 870         ip6_route_init();
 871         ip6_flowlabel_init();
 872         err = addrconf_init();
 873         if (err)
 874                 goto addrconf_fail;
 875
 876         /* Init v6 extension headers. */
 877         ipv6_rthdr_init();
 878         ipv6_frag_init();
 879         ipv6_nodata_init();
 880         ipv6_destopt_init();
 881 #ifdef CONFIG_IPV6_MIP6
 882         mip6_init();
 883 #endif
 884
 885         /* Init v6 transport protocols. */
 886         udpv6_init();
 887         udplitev6_init();
 888         tcpv6_init();
 889
 890         ipv6_packet_init();
 891         err = 0;
 892 out:
 893         return err;
 894
 895 addrconf_fail:
 896         ip6_flowlabel_cleanup();
 897         ip6_route_cleanup();
 898 #ifdef CONFIG_PROC_FS
 899         if6_proc_exit();
 900 proc_if6_fail:
 901         ac6_proc_exit();
 902 proc_anycast6_fail:
 903         ipv6_misc_proc_exit();
 904 proc_misc6_fail:
 905         udplite6_proc_exit();
 906 proc_udplite6_fail:
 907         udp6_proc_exit();
 908 proc_udp6_fail:
 909         tcp6_proc_exit();
 910 proc_tcp6_fail:
 911         raw6_proc_exit();
 912 proc_raw6_fail:
 913 #endif
 914         ipv6_netfilter_fini();
 915 netfilter_fail:
 916         igmp6_cleanup();
 917 igmp_fail:
 918         ndisc_cleanup();
 919 ndisc_fail:
 920         icmpv6_cleanup();
 921 icmp_fail:
 922 #ifdef CONFIG_SYSCTL
 923         ipv6_sysctl_unregister();
 924 #endif
 925         cleanup_ipv6_mibs();
 926 out_unregister_sock:
 927         sock_unregister(PF_INET6);
 928 out_unregister_raw_proto:
 929         proto_unregister(&rawv6_prot);
 930 out_unregister_udplite_proto:
 931         proto_unregister(&udplitev6_prot);
 932 out_unregister_udp_proto:
 933         proto_unregister(&udpv6_prot);
 934 out_unregister_tcp_proto:
 935         proto_unregister(&tcpv6_prot);
 936         goto out;
 937 }
 938 module_init(inet6_init);
 939
 940 static void __exit inet6_exit(void)
 941 {
 942         /* First of all disallow new sockets creation. */
 943         sock_unregister(PF_INET6);
 944
 945         /* Cleanup code parts. */
 946         ipv6_packet_cleanup();
 947 #ifdef CONFIG_IPV6_MIP6
 948         mip6_fini();
 949 #endif
 950         addrconf_cleanup();
 951         ip6_flowlabel_cleanup();
 952         ip6_route_cleanup();
 953 #ifdef CONFIG_PROC_FS
 954
 955         /* Cleanup code parts. */
 956         if6_proc_exit();
 957         ac6_proc_exit();
 958         ipv6_misc_proc_exit();
 959         udplite6_proc_exit();
 960         udp6_proc_exit();
 961         tcp6_proc_exit();
 962         raw6_proc_exit();
 963 #endif
 964         ipv6_netfilter_fini();
 965         igmp6_cleanup();
 966         ndisc_cleanup();
 967         icmpv6_cleanup();
 968 #ifdef CONFIG_SYSCTL
 969         ipv6_sysctl_unregister();
 970 #endif
 971         cleanup_ipv6_mibs();
 972         proto_unregister(&rawv6_prot);
 973         proto_unregister(&udplitev6_prot);
 974         proto_unregister(&udpv6_prot);
 975         proto_unregister(&tcpv6_prot);
 976 }
 977 module_exit(inet6_exit);
 978
 979 MODULE_ALIAS_NETPROTO(PF_INET6);