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 #include <linux/vs_context.h>
  43
  44 #include <linux/inet.h>
  45 #include <linux/netdevice.h>
  46 #include <linux/icmpv6.h>
  47 #include <linux/smp_lock.h>
  48 #include <linux/netfilter_ipv6.h>
  49
  50 #include <net/ip.h>
  51 #include <net/ipv6.h>
  52 #include <net/udp.h>
  53 #include <net/udplite.h>
  54 #include <net/tcp.h>
  55 #include <net/ipip.h>
  56 #include <net/protocol.h>
  57 #include <net/inet_common.h>
  58 #include <net/transp_v6.h>
  59 #include <net/ip6_route.h>
  60 #include <net/addrconf.h>
  61 #ifdef CONFIG_IPV6_TUNNEL
  62 #include <net/ip6_tunnel.h>
  63 #endif
  64 #ifdef CONFIG_IPV6_MIP6
  65 #include <net/mip6.h>
  66 #endif
  67
  68 #include <asm/uaccess.h>
  69 #include <asm/system.h>
  70
  71 MODULE_AUTHOR("Cast of dozens");
  72 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  73 MODULE_LICENSE("GPL");
  74
  75 int sysctl_ipv6_bindv6only __read_mostly;
  76
  77 /* The inetsw table contains everything that inet_create needs to
  78  * build a new socket.
  79  */
  80 static struct list_head inetsw6[SOCK_MAX];
  81 static DEFINE_SPINLOCK(inetsw6_lock);
  82
  83 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  84 {
  85         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  86
  87         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  88 }
  89
  90 static int inet6_create(struct socket *sock, int protocol)
  91 {
  92         struct inet_sock *inet;
  93         struct ipv6_pinfo *np;
  94         struct sock *sk;
  95         struct list_head *p;
  96         struct inet_protosw *answer;
  97         struct proto *answer_prot;
  98         unsigned char answer_flags;
  99         char answer_no_check;
 100         int try_loading_module = 0;
 101         int err;
 102
 103         /* Look for the requested type/protocol pair. */
 104         answer = NULL;
 105 lookup_protocol:
 106         err = -ESOCKTNOSUPPORT;
 107         rcu_read_lock();
 108         list_for_each_rcu(p, &inetsw6[sock->type]) {
 109                 answer = list_entry(p, struct inet_protosw, list);
 110
 111                 /* Check the non-wild match. */
 112                 if (protocol == answer->protocol) {
 113                         if (protocol != IPPROTO_IP)
 114                                 break;
 115                 } else {
 116                         /* Check for the two wild cases. */
 117                         if (IPPROTO_IP == protocol) {
 118                                 protocol = answer->protocol;
 119                                 break;
 120                         }
 121                         if (IPPROTO_IP == answer->protocol)
 122                                 break;
 123                 }
 124                 err = -EPROTONOSUPPORT;
 125                 answer = NULL;
 126         }
 127
 128         if (!answer) {
 129                 if (try_loading_module < 2) {
 130                         rcu_read_unlock();
 131                         /*
 132                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 133                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 134                          */
 135                         if (++try_loading_module == 1)
 136                                 request_module("net-pf-%d-proto-%d-type-%d",
 137                                                 PF_INET6, protocol, sock->type);
 138                         /*
 139                          * Fall back to generic, e.g. net-pf-10-proto-132
 140                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 141                          */
 142                         else
 143                                 request_module("net-pf-%d-proto-%d",
 144                                                 PF_INET6, protocol);
 145                         goto lookup_protocol;
 146                 } else
 147                         goto out_rcu_unlock;
 148         }
 149
 150         err = -EPERM;
 151         if ((protocol == IPPROTO_ICMPV6) && vx_ccaps(VXC_RAW_ICMP))
 152                 goto override;
 153         if (answer->capability > 0 && !capable(answer->capability))
 154                 goto out_rcu_unlock;
 155 override:
 156         sock->ops = answer->ops;
 157         answer_prot = answer->prot;
 158         answer_no_check = answer->no_check;
 159         answer_flags = answer->flags;
 160         rcu_read_unlock();
 161
 162         BUG_TRAP(answer_prot->slab != NULL);
 163
 164         err = -ENOBUFS;
 165         sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
 166         if (sk == NULL)
 167                 goto out;
 168
 169         sock_init_data(sock, sk);
 170
 171         err = 0;
 172         sk->sk_no_check = answer_no_check;
 173         if (INET_PROTOSW_REUSE & answer_flags)
 174                 sk->sk_reuse = 1;
 175
 176         inet = inet_sk(sk);
 177         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 178
 179         if (SOCK_RAW == sock->type) {
 180                 inet->num = protocol;
 181                 if (IPPROTO_RAW == protocol)
 182                         inet->hdrincl = 1;
 183         }
 184
 185         sk->sk_destruct         = inet_sock_destruct;
 186         sk->sk_family           = PF_INET6;
 187         sk->sk_protocol         = protocol;
 188
 189         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 190
 191         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 192         np->hop_limit   = -1;
 193         np->mcast_hops  = -1;
 194         np->mc_loop     = 1;
 195         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 196         np->ipv6only    = sysctl_ipv6_bindv6only;
 197
 198         /* Init the ipv4 part of the socket since we can have sockets
 199          * using v6 API for ipv4.
 200          */
 201         inet->uc_ttl    = -1;
 202
 203         inet->mc_loop   = 1;
 204         inet->mc_ttl    = 1;
 205         inet->mc_index  = 0;
 206         inet->mc_list   = NULL;
 207
 208         if (ipv4_config.no_pmtu_disc)
 209                 inet->pmtudisc = IP_PMTUDISC_DONT;
 210         else
 211                 inet->pmtudisc = IP_PMTUDISC_WANT;
 212         /*
 213          * Increment only the relevant sk_prot->socks debug field, this changes
 214          * the previous behaviour of incrementing both the equivalent to
 215          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 216          *
 217          * This allows better debug granularity as we'll know exactly how many
 218          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 219          * transport protocol socks. -acme
 220          */
 221         sk_refcnt_debug_inc(sk);
 222
 223         if (inet->num) {
 224                 /* It assumes that any protocol which allows
 225                  * the user to assign a number at socket
 226                  * creation time automatically shares.
 227                  */
 228                 inet->sport = htons(inet->num);
 229                 sk->sk_prot->hash(sk);
 230         }
 231         if (sk->sk_prot->init) {
 232                 err = sk->sk_prot->init(sk);
 233                 if (err) {
 234                         sk_common_release(sk);
 235                         goto out;
 236                 }
 237         }
 238 out:
 239         vxdprintk(VXD_CBIT(net, 4), "inet6_create(%p, %d): %d",
 240             sock, protocol, err);
 241         return err;
 242 out_rcu_unlock:
 243         rcu_read_unlock();
 244         goto out;
 245 }
 246
 247
 248 /* bind for INET6 API */
 249 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 250 {
 251         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 252         struct sock *sk = sock->sk;
 253         struct inet_sock *inet = inet_sk(sk);
 254         struct ipv6_pinfo *np = inet6_sk(sk);
 255         __be32 v4addr = 0;
 256         unsigned short snum;
 257         int addr_type = 0;
 258         int err = 0;
 259
 260         /* If the socket has its own bind function then use it. */
 261         if (sk->sk_prot->bind)
 262                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 263
 264         if (addr_len < SIN6_LEN_RFC2133)
 265                 return -EINVAL;
 266         addr_type = ipv6_addr_type(&addr->sin6_addr);
 267         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 268                 return -EINVAL;
 269
 270         snum = ntohs(addr->sin6_port);
 271         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 272                 return -EACCES;
 273
 274         lock_sock(sk);
 275
 276         /* Check these errors (active socket, double bind). */
 277         if (sk->sk_state != TCP_CLOSE || inet->num) {
 278                 err = -EINVAL;
 279                 goto out;
 280         }
 281
 282         /* Check if the address belongs to the host. */
 283         if (addr_type == IPV6_ADDR_MAPPED) {
 284                 v4addr = addr->sin6_addr.s6_addr32[3];
 285                 if (inet_addr_type(v4addr) != RTN_LOCAL) {
 286                         err = -EADDRNOTAVAIL;
 287                         goto out;
 288                 }
 289         } else {
 290                 if (addr_type != IPV6_ADDR_ANY) {
 291                         struct net_device *dev = NULL;
 292
 293                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 294                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 295                                     addr->sin6_scope_id) {
 296                                         /* Override any existing binding, if another one
 297                                          * is supplied by user.
 298                                          */
 299                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 300                                 }
 301
 302                                 /* Binding to link-local address requires an interface */
 303                                 if (!sk->sk_bound_dev_if) {
 304                                         err = -EINVAL;
 305                                         goto out;
 306                                 }
 307                                 dev = dev_get_by_index(sk->sk_bound_dev_if);
 308                                 if (!dev) {
 309                                         err = -ENODEV;
 310                                         goto out;
 311                                 }
 312                         }
 313
 314                         /* VServer: Only accept if we have the address available */
 315                         if (sk->sk_nx_info && !addr6_in_nx_info(sk->sk_nx_info, &addr->sin6_addr)) {
 316                                 vxdprintk(VXD_CBIT(net, 3), "inet6_bind(" NIP6_FMT ", %d): EADDRNOTAVAIL",
 317                                     NIP6(addr->sin6_addr), sk->sk_nx_info->nx_id);
 318                                 err = -EADDRNOTAVAIL;
 319                                 if (dev)
 320                                         dev_put(dev);
 321                                 goto out;
 322                         } else
 323                                 vxdprintk(VXD_CBIT(net, 3), "inet6_bind(" NIP6_FMT ", %d): PASS",
 324                                     NIP6(addr->sin6_addr), sk->sk_nx_info ? sk->sk_nx_info->nx_id : 0);
 325
 326                         /* ipv4 addr of the socket is invalid.  Only the
 327                          * unspecified and mapped address have a v4 equivalent.
 328                          */
 329                         v4addr = LOOPBACK4_IPV6;
 330                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 331                                 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
 332                                         if (dev)
 333                                                 dev_put(dev);
 334                                         err = -EADDRNOTAVAIL;
 335                                         goto out;
 336                                 }
 337                         }
 338                         if (dev)
 339                                 dev_put(dev);
 340                 }
 341         }
 342
 343         inet->rcv_saddr = v4addr;
 344         inet->saddr = v4addr;
 345
 346         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 347
 348         if (!(addr_type & IPV6_ADDR_MULTICAST))
 349                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 350
 351         /* Make sure we are allowed to bind here. */
 352         if (sk->sk_prot->get_port(sk, snum)) {
 353                 inet_reset_saddr(sk);
 354                 err = -EADDRINUSE;
 355                 goto out;
 356         }
 357
 358         if (addr_type != IPV6_ADDR_ANY)
 359                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 360         if (snum)
 361                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 362         inet->sport = htons(inet->num);
 363         inet->dport = 0;
 364         inet->daddr = 0;
 365 out:
 366         release_sock(sk);
 367         return err;
 368 }
 369
 370 int inet6_release(struct socket *sock)
 371 {
 372         struct sock *sk = sock->sk;
 373
 374         if (sk == NULL)
 375                 return -EINVAL;
 376
 377         /* Free mc lists */
 378         ipv6_sock_mc_close(sk);
 379
 380         /* Free ac lists */
 381         ipv6_sock_ac_close(sk);
 382
 383         return inet_release(sock);
 384 }
 385
 386 int inet6_destroy_sock(struct sock *sk)
 387 {
 388         struct ipv6_pinfo *np = inet6_sk(sk);
 389         struct sk_buff *skb;
 390         struct ipv6_txoptions *opt;
 391
 392         /* Release rx options */
 393
 394         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 395                 kfree_skb(skb);
 396
 397         /* Free flowlabels */
 398         fl6_free_socklist(sk);
 399
 400         /* Free tx options */
 401
 402         if ((opt = xchg(&np->opt, NULL)) != NULL)
 403                 sock_kfree_s(sk, opt, opt->tot_len);
 404
 405         return 0;
 406 }
 407
 408 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 409
 410 /*
 411  *      This does both peername and sockname.
 412  */
 413
 414 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 415                  int *uaddr_len, int peer)
 416 {
 417         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 418         struct sock *sk = sock->sk;
 419         struct inet_sock *inet = inet_sk(sk);
 420         struct ipv6_pinfo *np = inet6_sk(sk);
 421
 422         sin->sin6_family = AF_INET6;
 423         sin->sin6_flowinfo = 0;
 424         sin->sin6_scope_id = 0;
 425         if (peer) {
 426                 if (!inet->dport)
 427                         return -ENOTCONN;
 428                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 429                     peer == 1)
 430                         return -ENOTCONN;
 431                 sin->sin6_port = inet->dport;
 432                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 433                 if (np->sndflow)
 434                         sin->sin6_flowinfo = np->flow_label;
 435         } else {
 436                 if (ipv6_addr_any(&np->rcv_saddr))
 437                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 438                 else
 439                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 440
 441                 sin->sin6_port = inet->sport;
 442         }
 443         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 444                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 445         *uaddr_len = sizeof(*sin);
 446         return(0);
 447 }
 448
 449 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 450 {
 451         struct sock *sk = sock->sk;
 452
 453         switch(cmd)
 454         {
 455         case SIOCGSTAMP:
 456                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 457
 458         case SIOCADDRT:
 459         case SIOCDELRT:
 460
 461                 return(ipv6_route_ioctl(cmd,(void __user *)arg));
 462
 463         case SIOCSIFADDR:
 464                 return addrconf_add_ifaddr((void __user *) arg);
 465         case SIOCDIFADDR:
 466                 return addrconf_del_ifaddr((void __user *) arg);
 467         case SIOCSIFDSTADDR:
 468                 return addrconf_set_dstaddr((void __user *) arg);
 469         default:
 470                 if (!sk->sk_prot->ioctl)
 471                         return -ENOIOCTLCMD;
 472                 return sk->sk_prot->ioctl(sk, cmd, arg);
 473         }
 474         /*NOTREACHED*/
 475         return(0);
 476 }
 477
 478 const struct proto_ops inet6_stream_ops = {
 479         .family            = PF_INET6,
 480         .owner             = THIS_MODULE,
 481         .release           = inet6_release,
 482         .bind              = inet6_bind,
 483         .connect           = inet_stream_connect,       /* ok           */
 484         .socketpair        = sock_no_socketpair,        /* a do nothing */
 485         .accept            = inet_accept,               /* ok           */
 486         .getname           = inet6_getname,
 487         .poll              = tcp_poll,                  /* ok           */
 488         .ioctl             = inet6_ioctl,               /* must change  */
 489         .listen            = inet_listen,               /* ok           */
 490         .shutdown          = inet_shutdown,             /* ok           */
 491         .setsockopt        = sock_common_setsockopt,    /* ok           */
 492         .getsockopt        = sock_common_getsockopt,    /* ok           */
 493         .sendmsg           = inet_sendmsg,              /* ok           */
 494         .recvmsg           = sock_common_recvmsg,       /* ok           */
 495         .mmap              = sock_no_mmap,
 496         .sendpage          = tcp_sendpage,
 497 #ifdef CONFIG_COMPAT
 498         .compat_setsockopt = compat_sock_common_setsockopt,
 499         .compat_getsockopt = compat_sock_common_getsockopt,
 500 #endif
 501 };
 502
 503 const struct proto_ops inet6_dgram_ops = {
 504         .family            = PF_INET6,
 505         .owner             = THIS_MODULE,
 506         .release           = inet6_release,
 507         .bind              = inet6_bind,
 508         .connect           = inet_dgram_connect,        /* ok           */
 509         .socketpair        = sock_no_socketpair,        /* a do nothing */
 510         .accept            = sock_no_accept,            /* a do nothing */
 511         .getname           = inet6_getname,
 512         .poll              = udp_poll,                  /* ok           */
 513         .ioctl             = inet6_ioctl,               /* must change  */
 514         .listen            = sock_no_listen,            /* ok           */
 515         .shutdown          = inet_shutdown,             /* ok           */
 516         .setsockopt        = sock_common_setsockopt,    /* ok           */
 517         .getsockopt        = sock_common_getsockopt,    /* ok           */
 518         .sendmsg           = inet_sendmsg,              /* ok           */
 519         .recvmsg           = sock_common_recvmsg,       /* ok           */
 520         .mmap              = sock_no_mmap,
 521         .sendpage          = sock_no_sendpage,
 522 #ifdef CONFIG_COMPAT
 523         .compat_setsockopt = compat_sock_common_setsockopt,
 524         .compat_getsockopt = compat_sock_common_getsockopt,
 525 #endif
 526 };
 527
 528 static struct net_proto_family inet6_family_ops = {
 529         .family = PF_INET6,
 530         .create = inet6_create,
 531         .owner  = THIS_MODULE,
 532 };
 533
 534 /* Same as inet6_dgram_ops, sans udp_poll.  */
 535 static const struct proto_ops inet6_sockraw_ops = {
 536         .family            = PF_INET6,
 537         .owner             = THIS_MODULE,
 538         .release           = inet6_release,
 539         .bind              = inet6_bind,
 540         .connect           = inet_dgram_connect,        /* ok           */
 541         .socketpair        = sock_no_socketpair,        /* a do nothing */
 542         .accept            = sock_no_accept,            /* a do nothing */
 543         .getname           = inet6_getname,
 544         .poll              = datagram_poll,             /* ok           */
 545         .ioctl             = inet6_ioctl,               /* must change  */
 546         .listen            = sock_no_listen,            /* ok           */
 547         .shutdown          = inet_shutdown,             /* ok           */
 548         .setsockopt        = sock_common_setsockopt,    /* ok           */
 549         .getsockopt        = sock_common_getsockopt,    /* ok           */
 550         .sendmsg           = inet_sendmsg,              /* ok           */
 551         .recvmsg           = sock_common_recvmsg,       /* ok           */
 552         .mmap              = sock_no_mmap,
 553         .sendpage          = sock_no_sendpage,
 554 #ifdef CONFIG_COMPAT
 555         .compat_setsockopt = compat_sock_common_setsockopt,
 556         .compat_getsockopt = compat_sock_common_getsockopt,
 557 #endif
 558 };
 559
 560 static struct inet_protosw rawv6_protosw = {
 561         .type           = SOCK_RAW,
 562         .protocol       = IPPROTO_IP,   /* wild card */
 563         .prot           = &rawv6_prot,
 564         .ops            = &inet6_sockraw_ops,
 565         .capability     = CAP_NET_RAW,
 566         .no_check       = UDP_CSUM_DEFAULT,
 567         .flags          = INET_PROTOSW_REUSE,
 568 };
 569
 570 void
 571 inet6_register_protosw(struct inet_protosw *p)
 572 {
 573         struct list_head *lh;
 574         struct inet_protosw *answer;
 575         int protocol = p->protocol;
 576         struct list_head *last_perm;
 577
 578         spin_lock_bh(&inetsw6_lock);
 579
 580         if (p->type >= SOCK_MAX)
 581                 goto out_illegal;
 582
 583         /* If we are trying to override a permanent protocol, bail. */
 584         answer = NULL;
 585         last_perm = &inetsw6[p->type];
 586         list_for_each(lh, &inetsw6[p->type]) {
 587                 answer = list_entry(lh, struct inet_protosw, list);
 588
 589                 /* Check only the non-wild match. */
 590                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 591                         if (protocol == answer->protocol)
 592                                 break;
 593                         last_perm = lh;
 594                 }
 595
 596                 answer = NULL;
 597         }
 598         if (answer)
 599                 goto out_permanent;
 600
 601         /* Add the new entry after the last permanent entry if any, so that
 602          * the new entry does not override a permanent entry when matched with
 603          * a wild-card protocol. But it is allowed to override any existing
 604          * non-permanent entry.  This means that when we remove this entry, the
 605          * system automatically returns to the old behavior.
 606          */
 607         list_add_rcu(&p->list, last_perm);
 608 out:
 609         spin_unlock_bh(&inetsw6_lock);
 610         return;
 611
 612 out_permanent:
 613         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 614                protocol);
 615         goto out;
 616
 617 out_illegal:
 618         printk(KERN_ERR
 619                "Ignoring attempt to register invalid socket type %d.\n",
 620                p->type);
 621         goto out;
 622 }
 623
 624 void
 625 inet6_unregister_protosw(struct inet_protosw *p)
 626 {
 627         if (INET_PROTOSW_PERMANENT & p->flags) {
 628                 printk(KERN_ERR
 629                        "Attempt to unregister permanent protocol %d.\n",
 630                        p->protocol);
 631         } else {
 632                 spin_lock_bh(&inetsw6_lock);
 633                 list_del_rcu(&p->list);
 634                 spin_unlock_bh(&inetsw6_lock);
 635
 636                 synchronize_net();
 637         }
 638 }
 639
 640 int inet6_sk_rebuild_header(struct sock *sk)
 641 {
 642         int err;
 643         struct dst_entry *dst;
 644         struct ipv6_pinfo *np = inet6_sk(sk);
 645
 646         dst = __sk_dst_check(sk, np->dst_cookie);
 647
 648         if (dst == NULL) {
 649                 struct inet_sock *inet = inet_sk(sk);
 650                 struct in6_addr *final_p = NULL, final;
 651                 struct flowi fl;
 652
 653                 memset(&fl, 0, sizeof(fl));
 654                 fl.proto = sk->sk_protocol;
 655                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 656                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 657                 fl.fl6_flowlabel = np->flow_label;
 658                 fl.oif = sk->sk_bound_dev_if;
 659                 fl.fl_ip_dport = inet->dport;
 660                 fl.fl_ip_sport = inet->sport;
 661                 security_sk_classify_flow(sk, &fl);
 662
 663                 if (np->opt && np->opt->srcrt) {
 664                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 665                         ipv6_addr_copy(&final, &fl.fl6_dst);
 666                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 667                         final_p = &final;
 668                 }
 669
 670                 err = ip6_dst_lookup(sk, &dst, &fl);
 671                 if (err) {
 672                         sk->sk_route_caps = 0;
 673                         return err;
 674                 }
 675                 if (final_p)
 676                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 677
 678                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
 679                         sk->sk_err_soft = -err;
 680                         return err;
 681                 }
 682
 683                 __ip6_dst_store(sk, dst, NULL, NULL);
 684         }
 685
 686         return 0;
 687 }
 688
 689 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 690
 691 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 692 {
 693         struct ipv6_pinfo *np = inet6_sk(sk);
 694         struct inet6_skb_parm *opt = IP6CB(skb);
 695
 696         if (np->rxopt.all) {
 697                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 698                                   np->rxopt.bits.ohopopts)) ||
 699                     ((IPV6_FLOWINFO_MASK & *(__be32*)skb->nh.raw) &&
 700                      np->rxopt.bits.rxflow) ||
 701                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 702                      np->rxopt.bits.osrcrt)) ||
 703                     ((opt->dst1 || opt->dst0) &&
 704                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 705                         return 1;
 706         }
 707         return 0;
 708 }
 709
 710 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 711
 712 int ipv6_dev_in_nx_info6(struct net_device *dev, struct nx_info *nxi) {
 713         int ret = 0;
 714
 715         if (nxi->nbipv6 > 0) {
 716                 struct inet6_dev *in_dev;
 717                 struct inet6_ifaddr **ifap;
 718                 struct inet6_ifaddr *ifa;
 719
 720                 in_dev = in6_dev_get(dev);
 721                 if (in_dev) {
 722                         for (ifap = &in_dev->addr_list; (ifa = *ifap) != NULL;
 723                                 ifap = &ifa->if_next) {
 724                                 if (addr6_in_nx_info(nxi, &(ifa->addr))) {
 725                                         ret = 1;
 726                                         break;
 727                                 }
 728                         }
 729                         in6_dev_put(in_dev);
 730                 }
 731         }
 732         return ret;
 733 }
 734
 735 EXPORT_SYMBOL_GPL(ipv6_dev_in_nx_info6);
 736
 737 int
 738 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
 739 {
 740         if (ptr == NULL)
 741                 return -EINVAL;
 742
 743         ptr[0] = __alloc_percpu(mibsize);
 744         if (!ptr[0])
 745                 goto err0;
 746
 747         ptr[1] = __alloc_percpu(mibsize);
 748         if (!ptr[1])
 749                 goto err1;
 750
 751         return 0;
 752
 753 err1:
 754         free_percpu(ptr[0]);
 755         ptr[0] = NULL;
 756 err0:
 757         return -ENOMEM;
 758 }
 759
 760 void
 761 snmp6_mib_free(void *ptr[2])
 762 {
 763         if (ptr == NULL)
 764                 return;
 765         free_percpu(ptr[0]);
 766         free_percpu(ptr[1]);
 767         ptr[0] = ptr[1] = NULL;
 768 }
 769
 770 static int __init init_ipv6_mibs(void)
 771 {
 772         if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
 773                            __alignof__(struct ipstats_mib)) < 0)
 774                 goto err_ip_mib;
 775         if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
 776                            __alignof__(struct icmpv6_mib)) < 0)
 777                 goto err_icmp_mib;
 778         if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
 779                            __alignof__(struct udp_mib)) < 0)
 780                 goto err_udp_mib;
 781         if (snmp6_mib_init((void **)udplite_stats_in6, sizeof (struct udp_mib),
 782                            __alignof__(struct udp_mib)) < 0)
 783                 goto err_udplite_mib;
 784         return 0;
 785
 786 err_udplite_mib:
 787         snmp6_mib_free((void **)udp_stats_in6);
 788 err_udp_mib:
 789         snmp6_mib_free((void **)icmpv6_statistics);
 790 err_icmp_mib:
 791         snmp6_mib_free((void **)ipv6_statistics);
 792 err_ip_mib:
 793         return -ENOMEM;
 794
 795 }
 796
 797 static void cleanup_ipv6_mibs(void)
 798 {
 799         snmp6_mib_free((void **)ipv6_statistics);
 800         snmp6_mib_free((void **)icmpv6_statistics);
 801         snmp6_mib_free((void **)udp_stats_in6);
 802         snmp6_mib_free((void **)udplite_stats_in6);
 803 }
 804
 805 #ifdef CONFIG_IPV6_MODULE
 806 static void vc_net_register_init(void)
 807 {
 808         struct nx_ipv6_mod mymod = {
 809                 .dev_in_nx_info6 = ipv6_dev_in_nx_info6,
 810                 .owner = THIS_MODULE
 811         };
 812         vc_net_register_ipv6(&mymod);
 813 }
 814 #endif
 815
 816 static int __init inet6_init(void)
 817 {
 818         struct sk_buff *dummy_skb;
 819         struct list_head *r;
 820         int err;
 821
 822         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
 823
 824 #ifdef MODULE
 825 #if 0 /* FIXME --RR */
 826         if (!mod_member_present(&__this_module, can_unload))
 827           return -EINVAL;
 828
 829         __this_module.can_unload = &ipv6_unload;
 830 #endif
 831 #endif
 832
 833         err = proto_register(&tcpv6_prot, 1);
 834         if (err)
 835                 goto out;
 836
 837         err = proto_register(&udpv6_prot, 1);
 838         if (err)
 839                 goto out_unregister_tcp_proto;
 840
 841         err = proto_register(&udplitev6_prot, 1);
 842         if (err)
 843                 goto out_unregister_udp_proto;
 844
 845         err = proto_register(&rawv6_prot, 1);
 846         if (err)
 847                 goto out_unregister_udplite_proto;
 848
 849
 850         /* Register the socket-side information for inet6_create.  */
 851         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 852                 INIT_LIST_HEAD(r);
 853
 854         /* We MUST register RAW sockets before we create the ICMP6,
 855          * IGMP6, or NDISC control sockets.
 856          */
 857         inet6_register_protosw(&rawv6_protosw);
 858
 859         /* Register the family here so that the init calls below will
 860          * be able to create sockets. (?? is this dangerous ??)
 861          */
 862         err = sock_register(&inet6_family_ops);
 863         if (err)
 864                 goto out_unregister_raw_proto;
 865
 866         /* Initialise ipv6 mibs */
 867         err = init_ipv6_mibs();
 868         if (err)
 869                 goto out_unregister_sock;
 870
 871         /*
 872          *      ipngwg API draft makes clear that the correct semantics
 873          *      for TCP and UDP is to consider one TCP and UDP instance
 874          *      in a host availiable by both INET and INET6 APIs and
 875          *      able to communicate via both network protocols.
 876          */
 877
 878 #ifdef CONFIG_SYSCTL
 879         ipv6_sysctl_register();
 880 #endif
 881         err = icmpv6_init(&inet6_family_ops);
 882         if (err)
 883                 goto icmp_fail;
 884         err = ndisc_init(&inet6_family_ops);
 885         if (err)
 886                 goto ndisc_fail;
 887         err = igmp6_init(&inet6_family_ops);
 888         if (err)
 889                 goto igmp_fail;
 890         err = ipv6_netfilter_init();
 891         if (err)
 892                 goto netfilter_fail;
 893         /* Create /proc/foo6 entries. */
 894 #ifdef CONFIG_PROC_FS
 895         err = -ENOMEM;
 896         if (raw6_proc_init())
 897                 goto proc_raw6_fail;
 898         if (tcp6_proc_init())
 899                 goto proc_tcp6_fail;
 900         if (udp6_proc_init())
 901                 goto proc_udp6_fail;
 902         if (udplite6_proc_init())
 903                 goto proc_udplite6_fail;
 904         if (ipv6_misc_proc_init())
 905                 goto proc_misc6_fail;
 906
 907         if (ac6_proc_init())
 908                 goto proc_anycast6_fail;
 909         if (if6_proc_init())
 910                 goto proc_if6_fail;
 911 #endif
 912         ip6_route_init();
 913         ip6_flowlabel_init();
 914         err = addrconf_init();
 915         if (err)
 916                 goto addrconf_fail;
 917
 918         /* Init v6 extension headers. */
 919         ipv6_rthdr_init();
 920         ipv6_frag_init();
 921         ipv6_nodata_init();
 922         ipv6_destopt_init();
 923 #ifdef CONFIG_IPV6_MIP6
 924         mip6_init();
 925 #endif
 926
 927         /* Init v6 transport protocols. */
 928         udpv6_init();
 929         udplitev6_init();
 930         tcpv6_init();
 931
 932         ipv6_packet_init();
 933
 934 #ifdef CONFIG_IPV6_MODULE
 935         vc_net_register_init();
 936 #endif
 937
 938         err = 0;
 939 out:
 940         return err;
 941
 942 addrconf_fail:
 943         ip6_flowlabel_cleanup();
 944         ip6_route_cleanup();
 945 #ifdef CONFIG_PROC_FS
 946         if6_proc_exit();
 947 proc_if6_fail:
 948         ac6_proc_exit();
 949 proc_anycast6_fail:
 950         ipv6_misc_proc_exit();
 951 proc_misc6_fail:
 952         udplite6_proc_exit();
 953 proc_udplite6_fail:
 954         udp6_proc_exit();
 955 proc_udp6_fail:
 956         tcp6_proc_exit();
 957 proc_tcp6_fail:
 958         raw6_proc_exit();
 959 proc_raw6_fail:
 960 #endif
 961         ipv6_netfilter_fini();
 962 netfilter_fail:
 963         igmp6_cleanup();
 964 igmp_fail:
 965         ndisc_cleanup();
 966 ndisc_fail:
 967         icmpv6_cleanup();
 968 icmp_fail:
 969 #ifdef CONFIG_SYSCTL
 970         ipv6_sysctl_unregister();
 971 #endif
 972         cleanup_ipv6_mibs();
 973 out_unregister_sock:
 974         sock_unregister(PF_INET6);
 975 out_unregister_raw_proto:
 976         proto_unregister(&rawv6_prot);
 977 out_unregister_udplite_proto:
 978         proto_unregister(&udplitev6_prot);
 979 out_unregister_udp_proto:
 980         proto_unregister(&udpv6_prot);
 981 out_unregister_tcp_proto:
 982         proto_unregister(&tcpv6_prot);
 983         goto out;
 984 }
 985 module_init(inet6_init);
 986
 987 static void __exit inet6_exit(void)
 988 {
 989         /* First of all disallow new sockets creation. */
 990         sock_unregister(PF_INET6);
 991 #ifdef CONFIG_IPV6_MODULE
 992         vc_net_unregister_ipv6();
 993 #endif
 994 #ifdef CONFIG_PROC_FS
 995         if6_proc_exit();
 996         ac6_proc_exit();
 997         ipv6_misc_proc_exit();
 998         udp6_proc_exit();
 999         udplite6_proc_exit();
1000         tcp6_proc_exit();
1001         raw6_proc_exit();
1002 #endif
1003 #ifdef CONFIG_IPV6_MIP6
1004         mip6_fini();
1005 #endif
1006         /* Cleanup code parts. */
1007         ip6_flowlabel_cleanup();
1008         addrconf_cleanup();
1009         ip6_route_cleanup();
1010         ipv6_packet_cleanup();
1011         igmp6_cleanup();
1012         ipv6_netfilter_fini();
1013         ndisc_cleanup();
1014         icmpv6_cleanup();
1015 #ifdef CONFIG_SYSCTL
1016         ipv6_sysctl_unregister();
1017 #endif
1018         cleanup_ipv6_mibs();
1019         proto_unregister(&rawv6_prot);
1020         proto_unregister(&udpv6_prot);
1021         proto_unregister(&tcpv6_prot);
1022 }
1023 module_exit(inet6_exit);
1024
1025 MODULE_ALIAS_NETPROTO(PF_INET6);