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/netfilter_ipv6.h>
  46
  47 #include <net/ip.h>
  48 #include <net/ipv6.h>
  49 #include <net/udp.h>
  50 #include <net/udplite.h>
  51 #include <net/tcp.h>
  52 #include <net/ipip.h>
  53 #include <net/protocol.h>
  54 #include <net/inet_common.h>
  55 #include <net/transp_v6.h>
  56 #include <net/ip6_route.h>
  57 #include <net/addrconf.h>
  58 #ifdef CONFIG_IPV6_TUNNEL
  59 #include <net/ip6_tunnel.h>
  60 #endif
  61
  62 #include <asm/uaccess.h>
  63 #include <asm/system.h>
  64
  65 MODULE_AUTHOR("Cast of dozens");
  66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  67 MODULE_LICENSE("GPL");
  68
  69 /* The inetsw6 table contains everything that inet6_create needs to
  70  * build a new socket.
  71  */
  72 static struct list_head inetsw6[SOCK_MAX];
  73 static DEFINE_SPINLOCK(inetsw6_lock);
  74
  75 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  76 {
  77         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  78
  79         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  80 }
  81
  82 static int inet6_create(struct net *net, struct socket *sock, int protocol)
  83 {
  84         struct inet_sock *inet;
  85         struct ipv6_pinfo *np;
  86         struct sock *sk;
  87         struct list_head *p;
  88         struct inet_protosw *answer;
  89         struct proto *answer_prot;
  90         unsigned char answer_flags;
  91         char answer_no_check;
  92         int try_loading_module = 0;
  93         int err;
  94
  95         if (net != &init_net)
  96                 return -EAFNOSUPPORT;
  97
  98         if (sock->type != SOCK_RAW &&
  99             sock->type != SOCK_DGRAM &&
 100             !inet_ehash_secret)
 101                 build_ehash_secret();
 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 (answer->capability > 0 && !capable(answer->capability))
 152                 goto out_rcu_unlock;
 153
 154         sock->ops = answer->ops;
 155         answer_prot = answer->prot;
 156         answer_no_check = answer->no_check;
 157         answer_flags = answer->flags;
 158         rcu_read_unlock();
 159
 160         BUG_TRAP(answer_prot->slab != NULL);
 161
 162         err = -ENOBUFS;
 163         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 164         if (sk == NULL)
 165                 goto out;
 166
 167         sock_init_data(sock, sk);
 168
 169         err = 0;
 170         sk->sk_no_check = answer_no_check;
 171         if (INET_PROTOSW_REUSE & answer_flags)
 172                 sk->sk_reuse = 1;
 173
 174         inet = inet_sk(sk);
 175         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 176
 177         if (SOCK_RAW == sock->type) {
 178                 inet->num = protocol;
 179                 if (IPPROTO_RAW == protocol)
 180                         inet->hdrincl = 1;
 181         }
 182
 183         sk->sk_destruct         = inet_sock_destruct;
 184         sk->sk_family           = PF_INET6;
 185         sk->sk_protocol         = protocol;
 186
 187         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 188
 189         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 190         np->hop_limit   = -1;
 191         np->mcast_hops  = -1;
 192         np->mc_loop     = 1;
 193         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 194         np->ipv6only    = init_net.ipv6.sysctl.bindv6only;
 195
 196         /* Init the ipv4 part of the socket since we can have sockets
 197          * using v6 API for ipv4.
 198          */
 199         inet->uc_ttl    = -1;
 200
 201         inet->mc_loop   = 1;
 202         inet->mc_ttl    = 1;
 203         inet->mc_index  = 0;
 204         inet->mc_list   = NULL;
 205
 206         if (ipv4_config.no_pmtu_disc)
 207                 inet->pmtudisc = IP_PMTUDISC_DONT;
 208         else
 209                 inet->pmtudisc = IP_PMTUDISC_WANT;
 210         /*
 211          * Increment only the relevant sk_prot->socks debug field, this changes
 212          * the previous behaviour of incrementing both the equivalent to
 213          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 214          *
 215          * This allows better debug granularity as we'll know exactly how many
 216          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 217          * transport protocol socks. -acme
 218          */
 219         sk_refcnt_debug_inc(sk);
 220
 221         if (inet->num) {
 222                 /* It assumes that any protocol which allows
 223                  * the user to assign a number at socket
 224                  * creation time automatically shares.
 225                  */
 226                 inet->sport = htons(inet->num);
 227                 sk->sk_prot->hash(sk);
 228         }
 229         if (sk->sk_prot->init) {
 230                 err = sk->sk_prot->init(sk);
 231                 if (err) {
 232                         sk_common_release(sk);
 233                         goto out;
 234                 }
 235         }
 236 out:
 237         return err;
 238 out_rcu_unlock:
 239         rcu_read_unlock();
 240         goto out;
 241 }
 242
 243
 244 /* bind for INET6 API */
 245 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 246 {
 247         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 248         struct sock *sk = sock->sk;
 249         struct inet_sock *inet = inet_sk(sk);
 250         struct ipv6_pinfo *np = inet6_sk(sk);
 251         __be32 v4addr = 0;
 252         unsigned short snum;
 253         int addr_type = 0;
 254         int err = 0;
 255
 256         /* If the socket has its own bind function then use it. */
 257         if (sk->sk_prot->bind)
 258                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 259
 260         if (addr_len < SIN6_LEN_RFC2133)
 261                 return -EINVAL;
 262         addr_type = ipv6_addr_type(&addr->sin6_addr);
 263         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 264                 return -EINVAL;
 265
 266         snum = ntohs(addr->sin6_port);
 267         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 268                 return -EACCES;
 269
 270         lock_sock(sk);
 271
 272         /* Check these errors (active socket, double bind). */
 273         if (sk->sk_state != TCP_CLOSE || inet->num) {
 274                 err = -EINVAL;
 275                 goto out;
 276         }
 277
 278         /* Check if the address belongs to the host. */
 279         if (addr_type == IPV6_ADDR_MAPPED) {
 280                 v4addr = addr->sin6_addr.s6_addr32[3];
 281                 if (inet_addr_type(&init_net, v4addr) != RTN_LOCAL) {
 282                         err = -EADDRNOTAVAIL;
 283                         goto out;
 284                 }
 285         } else {
 286                 if (addr_type != IPV6_ADDR_ANY) {
 287                         struct net_device *dev = NULL;
 288
 289                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 290                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 291                                     addr->sin6_scope_id) {
 292                                         /* Override any existing binding, if another one
 293                                          * is supplied by user.
 294                                          */
 295                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 296                                 }
 297
 298                                 /* Binding to link-local address requires an interface */
 299                                 if (!sk->sk_bound_dev_if) {
 300                                         err = -EINVAL;
 301                                         goto out;
 302                                 }
 303                                 dev = dev_get_by_index(&init_net, sk->sk_bound_dev_if);
 304                                 if (!dev) {
 305                                         err = -ENODEV;
 306                                         goto out;
 307                                 }
 308                         }
 309
 310                         /* ipv4 addr of the socket is invalid.  Only the
 311                          * unspecified and mapped address have a v4 equivalent.
 312                          */
 313                         v4addr = LOOPBACK4_IPV6;
 314                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 315                                 if (!ipv6_chk_addr(&init_net, &addr->sin6_addr,
 316                                                    dev, 0)) {
 317                                         if (dev)
 318                                                 dev_put(dev);
 319                                         err = -EADDRNOTAVAIL;
 320                                         goto out;
 321                                 }
 322                         }
 323                         if (dev)
 324                                 dev_put(dev);
 325                 }
 326         }
 327
 328         inet->rcv_saddr = v4addr;
 329         inet->saddr = v4addr;
 330
 331         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 332
 333         if (!(addr_type & IPV6_ADDR_MULTICAST))
 334                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
 335
 336         /* Make sure we are allowed to bind here. */
 337         if (sk->sk_prot->get_port(sk, snum)) {
 338                 inet_reset_saddr(sk);
 339                 err = -EADDRINUSE;
 340                 goto out;
 341         }
 342
 343         if (addr_type != IPV6_ADDR_ANY)
 344                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 345         if (snum)
 346                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 347         inet->sport = htons(inet->num);
 348         inet->dport = 0;
 349         inet->daddr = 0;
 350 out:
 351         release_sock(sk);
 352         return err;
 353 }
 354
 355 EXPORT_SYMBOL(inet6_bind);
 356
 357 int inet6_release(struct socket *sock)
 358 {
 359         struct sock *sk = sock->sk;
 360
 361         if (sk == NULL)
 362                 return -EINVAL;
 363
 364         /* Free mc lists */
 365         ipv6_sock_mc_close(sk);
 366
 367         /* Free ac lists */
 368         ipv6_sock_ac_close(sk);
 369
 370         return inet_release(sock);
 371 }
 372
 373 EXPORT_SYMBOL(inet6_release);
 374
 375 int inet6_destroy_sock(struct sock *sk)
 376 {
 377         struct ipv6_pinfo *np = inet6_sk(sk);
 378         struct sk_buff *skb;
 379         struct ipv6_txoptions *opt;
 380
 381         /* Release rx options */
 382
 383         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
 384                 kfree_skb(skb);
 385
 386         /* Free flowlabels */
 387         fl6_free_socklist(sk);
 388
 389         /* Free tx options */
 390
 391         if ((opt = xchg(&np->opt, NULL)) != NULL)
 392                 sock_kfree_s(sk, opt, opt->tot_len);
 393
 394         return 0;
 395 }
 396
 397 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 398
 399 /*
 400  *      This does both peername and sockname.
 401  */
 402
 403 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 404                  int *uaddr_len, int peer)
 405 {
 406         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 407         struct sock *sk = sock->sk;
 408         struct inet_sock *inet = inet_sk(sk);
 409         struct ipv6_pinfo *np = inet6_sk(sk);
 410
 411         sin->sin6_family = AF_INET6;
 412         sin->sin6_flowinfo = 0;
 413         sin->sin6_scope_id = 0;
 414         if (peer) {
 415                 if (!inet->dport)
 416                         return -ENOTCONN;
 417                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 418                     peer == 1)
 419                         return -ENOTCONN;
 420                 sin->sin6_port = inet->dport;
 421                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
 422                 if (np->sndflow)
 423                         sin->sin6_flowinfo = np->flow_label;
 424         } else {
 425                 if (ipv6_addr_any(&np->rcv_saddr))
 426                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
 427                 else
 428                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 429
 430                 sin->sin6_port = inet->sport;
 431         }
 432         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 433                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 434         *uaddr_len = sizeof(*sin);
 435         return(0);
 436 }
 437
 438 EXPORT_SYMBOL(inet6_getname);
 439
 440 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 441 {
 442         struct sock *sk = sock->sk;
 443
 444         switch(cmd)
 445         {
 446         case SIOCGSTAMP:
 447                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 448
 449         case SIOCGSTAMPNS:
 450                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 451
 452         case SIOCADDRT:
 453         case SIOCDELRT:
 454
 455                 return(ipv6_route_ioctl(cmd,(void __user *)arg));
 456
 457         case SIOCSIFADDR:
 458                 return addrconf_add_ifaddr((void __user *) arg);
 459         case SIOCDIFADDR:
 460                 return addrconf_del_ifaddr((void __user *) arg);
 461         case SIOCSIFDSTADDR:
 462                 return addrconf_set_dstaddr((void __user *) arg);
 463         default:
 464                 if (!sk->sk_prot->ioctl)
 465                         return -ENOIOCTLCMD;
 466                 return sk->sk_prot->ioctl(sk, cmd, arg);
 467         }
 468         /*NOTREACHED*/
 469         return(0);
 470 }
 471
 472 EXPORT_SYMBOL(inet6_ioctl);
 473
 474 const struct proto_ops inet6_stream_ops = {
 475         .family            = PF_INET6,
 476         .owner             = THIS_MODULE,
 477         .release           = inet6_release,
 478         .bind              = inet6_bind,
 479         .connect           = inet_stream_connect,       /* ok           */
 480         .socketpair        = sock_no_socketpair,        /* a do nothing */
 481         .accept            = inet_accept,               /* ok           */
 482         .getname           = inet6_getname,
 483         .poll              = tcp_poll,                  /* ok           */
 484         .ioctl             = inet6_ioctl,               /* must change  */
 485         .listen            = inet_listen,               /* ok           */
 486         .shutdown          = inet_shutdown,             /* ok           */
 487         .setsockopt        = sock_common_setsockopt,    /* ok           */
 488         .getsockopt        = sock_common_getsockopt,    /* ok           */
 489         .sendmsg           = tcp_sendmsg,               /* ok           */
 490         .recvmsg           = sock_common_recvmsg,       /* ok           */
 491         .mmap              = sock_no_mmap,
 492         .sendpage          = tcp_sendpage,
 493         .splice_read       = tcp_splice_read,
 494 #ifdef CONFIG_COMPAT
 495         .compat_setsockopt = compat_sock_common_setsockopt,
 496         .compat_getsockopt = compat_sock_common_getsockopt,
 497 #endif
 498 };
 499
 500 const struct proto_ops inet6_dgram_ops = {
 501         .family            = PF_INET6,
 502         .owner             = THIS_MODULE,
 503         .release           = inet6_release,
 504         .bind              = inet6_bind,
 505         .connect           = inet_dgram_connect,        /* ok           */
 506         .socketpair        = sock_no_socketpair,        /* a do nothing */
 507         .accept            = sock_no_accept,            /* a do nothing */
 508         .getname           = inet6_getname,
 509         .poll              = udp_poll,                  /* ok           */
 510         .ioctl             = inet6_ioctl,               /* must change  */
 511         .listen            = sock_no_listen,            /* ok           */
 512         .shutdown          = inet_shutdown,             /* ok           */
 513         .setsockopt        = sock_common_setsockopt,    /* ok           */
 514         .getsockopt        = sock_common_getsockopt,    /* ok           */
 515         .sendmsg           = inet_sendmsg,              /* ok           */
 516         .recvmsg           = sock_common_recvmsg,       /* ok           */
 517         .mmap              = sock_no_mmap,
 518         .sendpage          = sock_no_sendpage,
 519 #ifdef CONFIG_COMPAT
 520         .compat_setsockopt = compat_sock_common_setsockopt,
 521         .compat_getsockopt = compat_sock_common_getsockopt,
 522 #endif
 523 };
 524
 525 static struct net_proto_family inet6_family_ops = {
 526         .family = PF_INET6,
 527         .create = inet6_create,
 528         .owner  = THIS_MODULE,
 529 };
 530
 531 int inet6_register_protosw(struct inet_protosw *p)
 532 {
 533         struct list_head *lh;
 534         struct inet_protosw *answer;
 535         struct list_head *last_perm;
 536         int protocol = p->protocol;
 537         int ret;
 538
 539         spin_lock_bh(&inetsw6_lock);
 540
 541         ret = -EINVAL;
 542         if (p->type >= SOCK_MAX)
 543                 goto out_illegal;
 544
 545         /* If we are trying to override a permanent protocol, bail. */
 546         answer = NULL;
 547         ret = -EPERM;
 548         last_perm = &inetsw6[p->type];
 549         list_for_each(lh, &inetsw6[p->type]) {
 550                 answer = list_entry(lh, struct inet_protosw, list);
 551
 552                 /* Check only the non-wild match. */
 553                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 554                         if (protocol == answer->protocol)
 555                                 break;
 556                         last_perm = lh;
 557                 }
 558
 559                 answer = NULL;
 560         }
 561         if (answer)
 562                 goto out_permanent;
 563
 564         /* Add the new entry after the last permanent entry if any, so that
 565          * the new entry does not override a permanent entry when matched with
 566          * a wild-card protocol. But it is allowed to override any existing
 567          * non-permanent entry.  This means that when we remove this entry, the
 568          * system automatically returns to the old behavior.
 569          */
 570         list_add_rcu(&p->list, last_perm);
 571         ret = 0;
 572 out:
 573         spin_unlock_bh(&inetsw6_lock);
 574         return ret;
 575
 576 out_permanent:
 577         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
 578                protocol);
 579         goto out;
 580
 581 out_illegal:
 582         printk(KERN_ERR
 583                "Ignoring attempt to register invalid socket type %d.\n",
 584                p->type);
 585         goto out;
 586 }
 587
 588 EXPORT_SYMBOL(inet6_register_protosw);
 589
 590 void
 591 inet6_unregister_protosw(struct inet_protosw *p)
 592 {
 593         if (INET_PROTOSW_PERMANENT & p->flags) {
 594                 printk(KERN_ERR
 595                        "Attempt to unregister permanent protocol %d.\n",
 596                        p->protocol);
 597         } else {
 598                 spin_lock_bh(&inetsw6_lock);
 599                 list_del_rcu(&p->list);
 600                 spin_unlock_bh(&inetsw6_lock);
 601
 602                 synchronize_net();
 603         }
 604 }
 605
 606 EXPORT_SYMBOL(inet6_unregister_protosw);
 607
 608 int inet6_sk_rebuild_header(struct sock *sk)
 609 {
 610         int err;
 611         struct dst_entry *dst;
 612         struct ipv6_pinfo *np = inet6_sk(sk);
 613
 614         dst = __sk_dst_check(sk, np->dst_cookie);
 615
 616         if (dst == NULL) {
 617                 struct inet_sock *inet = inet_sk(sk);
 618                 struct in6_addr *final_p = NULL, final;
 619                 struct flowi fl;
 620
 621                 memset(&fl, 0, sizeof(fl));
 622                 fl.proto = sk->sk_protocol;
 623                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 624                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 625                 fl.fl6_flowlabel = np->flow_label;
 626                 fl.oif = sk->sk_bound_dev_if;
 627                 fl.fl_ip_dport = inet->dport;
 628                 fl.fl_ip_sport = inet->sport;
 629                 security_sk_classify_flow(sk, &fl);
 630
 631                 if (np->opt && np->opt->srcrt) {
 632                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
 633                         ipv6_addr_copy(&final, &fl.fl6_dst);
 634                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 635                         final_p = &final;
 636                 }
 637
 638                 err = ip6_dst_lookup(sk, &dst, &fl);
 639                 if (err) {
 640                         sk->sk_route_caps = 0;
 641                         return err;
 642                 }
 643                 if (final_p)
 644                         ipv6_addr_copy(&fl.fl6_dst, final_p);
 645
 646                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
 647                         sk->sk_err_soft = -err;
 648                         return err;
 649                 }
 650
 651                 __ip6_dst_store(sk, dst, NULL, NULL);
 652         }
 653
 654         return 0;
 655 }
 656
 657 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 658
 659 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 660 {
 661         struct ipv6_pinfo *np = inet6_sk(sk);
 662         struct inet6_skb_parm *opt = IP6CB(skb);
 663
 664         if (np->rxopt.all) {
 665                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 666                                   np->rxopt.bits.ohopopts)) ||
 667                     ((IPV6_FLOWINFO_MASK &
 668                       *(__be32 *)skb_network_header(skb)) &&
 669                      np->rxopt.bits.rxflow) ||
 670                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 671                      np->rxopt.bits.osrcrt)) ||
 672                     ((opt->dst1 || opt->dst0) &&
 673                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 674                         return 1;
 675         }
 676         return 0;
 677 }
 678
 679 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 680
 681 static int __init init_ipv6_mibs(void)
 682 {
 683         if (snmp_mib_init((void **)ipv6_statistics,
 684                           sizeof(struct ipstats_mib)) < 0)
 685                 goto err_ip_mib;
 686         if (snmp_mib_init((void **)icmpv6_statistics,
 687                           sizeof(struct icmpv6_mib)) < 0)
 688                 goto err_icmp_mib;
 689         if (snmp_mib_init((void **)icmpv6msg_statistics,
 690                           sizeof(struct icmpv6msg_mib)) < 0)
 691                 goto err_icmpmsg_mib;
 692         if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib)) < 0)
 693                 goto err_udp_mib;
 694         if (snmp_mib_init((void **)udplite_stats_in6,
 695                           sizeof (struct udp_mib)) < 0)
 696                 goto err_udplite_mib;
 697         return 0;
 698
 699 err_udplite_mib:
 700         snmp_mib_free((void **)udp_stats_in6);
 701 err_udp_mib:
 702         snmp_mib_free((void **)icmpv6msg_statistics);
 703 err_icmpmsg_mib:
 704         snmp_mib_free((void **)icmpv6_statistics);
 705 err_icmp_mib:
 706         snmp_mib_free((void **)ipv6_statistics);
 707 err_ip_mib:
 708         return -ENOMEM;
 709
 710 }
 711
 712 static void cleanup_ipv6_mibs(void)
 713 {
 714         snmp_mib_free((void **)ipv6_statistics);
 715         snmp_mib_free((void **)icmpv6_statistics);
 716         snmp_mib_free((void **)icmpv6msg_statistics);
 717         snmp_mib_free((void **)udp_stats_in6);
 718         snmp_mib_free((void **)udplite_stats_in6);
 719 }
 720
 721 static int inet6_net_init(struct net *net)
 722 {
 723         net->ipv6.sysctl.bindv6only = 0;
 724         net->ipv6.sysctl.flush_delay = 0;
 725         net->ipv6.sysctl.ip6_rt_max_size = 4096;
 726         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
 727         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
 728         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
 729         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
 730         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
 731         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
 732         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 733
 734         return 0;
 735 }
 736
 737 static void inet6_net_exit(struct net *net)
 738 {
 739         return;
 740 }
 741
 742 static struct pernet_operations inet6_net_ops = {
 743         .init = inet6_net_init,
 744         .exit = inet6_net_exit,
 745 };
 746
 747 static int __init inet6_init(void)
 748 {
 749         struct sk_buff *dummy_skb;
 750         struct list_head *r;
 751         int err;
 752
 753         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
 754
 755         err = proto_register(&tcpv6_prot, 1);
 756         if (err)
 757                 goto out;
 758
 759         err = proto_register(&udpv6_prot, 1);
 760         if (err)
 761                 goto out_unregister_tcp_proto;
 762
 763         err = proto_register(&udplitev6_prot, 1);
 764         if (err)
 765                 goto out_unregister_udp_proto;
 766
 767         err = proto_register(&rawv6_prot, 1);
 768         if (err)
 769                 goto out_unregister_udplite_proto;
 770
 771
 772         /* Register the socket-side information for inet6_create.  */
 773         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 774                 INIT_LIST_HEAD(r);
 775
 776         /* We MUST register RAW sockets before we create the ICMP6,
 777          * IGMP6, or NDISC control sockets.
 778          */
 779         err = rawv6_init();
 780         if (err)
 781                 goto out_unregister_raw_proto;
 782
 783         /* Register the family here so that the init calls below will
 784          * be able to create sockets. (?? is this dangerous ??)
 785          */
 786         err = sock_register(&inet6_family_ops);
 787         if (err)
 788                 goto out_sock_register_fail;
 789
 790         /* Initialise ipv6 mibs */
 791         err = init_ipv6_mibs();
 792         if (err)
 793                 goto out_unregister_sock;
 794
 795         /*
 796          *      ipngwg API draft makes clear that the correct semantics
 797          *      for TCP and UDP is to consider one TCP and UDP instance
 798          *      in a host availiable by both INET and INET6 APIs and
 799          *      able to communicate via both network protocols.
 800          */
 801
 802         err = register_pernet_subsys(&inet6_net_ops);
 803         if (err)
 804                 goto register_pernet_fail;
 805
 806 #ifdef CONFIG_SYSCTL
 807         err = ipv6_sysctl_register();
 808         if (err)
 809                 goto sysctl_fail;
 810 #endif
 811         err = icmpv6_init(&inet6_family_ops);
 812         if (err)
 813                 goto icmp_fail;
 814         err = ndisc_init(&inet6_family_ops);
 815         if (err)
 816                 goto ndisc_fail;
 817         err = igmp6_init(&inet6_family_ops);
 818         if (err)
 819                 goto igmp_fail;
 820         err = ipv6_netfilter_init();
 821         if (err)
 822                 goto netfilter_fail;
 823         /* Create /proc/foo6 entries. */
 824 #ifdef CONFIG_PROC_FS
 825         err = -ENOMEM;
 826         if (raw6_proc_init())
 827                 goto proc_raw6_fail;
 828         if (tcp6_proc_init())
 829                 goto proc_tcp6_fail;
 830         if (udp6_proc_init())
 831                 goto proc_udp6_fail;
 832         if (udplite6_proc_init())
 833                 goto proc_udplite6_fail;
 834         if (ipv6_misc_proc_init())
 835                 goto proc_misc6_fail;
 836
 837         if (ac6_proc_init())
 838                 goto proc_anycast6_fail;
 839         if (if6_proc_init())
 840                 goto proc_if6_fail;
 841 #endif
 842         err = ip6_route_init();
 843         if (err)
 844                 goto ip6_route_fail;
 845         err = ip6_flowlabel_init();
 846         if (err)
 847                 goto ip6_flowlabel_fail;
 848         err = addrconf_init();
 849         if (err)
 850                 goto addrconf_fail;
 851
 852         /* Init v6 extension headers. */
 853         err = ipv6_exthdrs_init();
 854         if (err)
 855                 goto ipv6_exthdrs_fail;
 856
 857         err = ipv6_frag_init();
 858         if (err)
 859                 goto ipv6_frag_fail;
 860
 861         /* Init v6 transport protocols. */
 862         err = udpv6_init();
 863         if (err)
 864                 goto udpv6_fail;
 865
 866         err = udplitev6_init();
 867         if (err)
 868                 goto udplitev6_fail;
 869
 870         err = tcpv6_init();
 871         if (err)
 872                 goto tcpv6_fail;
 873
 874         err = ipv6_packet_init();
 875         if (err)
 876                 goto ipv6_packet_fail;
 877 out:
 878         return err;
 879
 880 ipv6_packet_fail:
 881         tcpv6_exit();
 882 tcpv6_fail:
 883         udplitev6_exit();
 884 udplitev6_fail:
 885         udpv6_exit();
 886 udpv6_fail:
 887         ipv6_frag_exit();
 888 ipv6_frag_fail:
 889         ipv6_exthdrs_exit();
 890 ipv6_exthdrs_fail:
 891         addrconf_cleanup();
 892 addrconf_fail:
 893         ip6_flowlabel_cleanup();
 894 ip6_flowlabel_fail:
 895         ip6_route_cleanup();
 896 ip6_route_fail:
 897 #ifdef CONFIG_PROC_FS
 898         if6_proc_exit();
 899 proc_if6_fail:
 900         ac6_proc_exit();
 901 proc_anycast6_fail:
 902         ipv6_misc_proc_exit();
 903 proc_misc6_fail:
 904         udplite6_proc_exit();
 905 proc_udplite6_fail:
 906         udp6_proc_exit();
 907 proc_udp6_fail:
 908         tcp6_proc_exit();
 909 proc_tcp6_fail:
 910         raw6_proc_exit();
 911 proc_raw6_fail:
 912 #endif
 913         ipv6_netfilter_fini();
 914 netfilter_fail:
 915         igmp6_cleanup();
 916 igmp_fail:
 917         ndisc_cleanup();
 918 ndisc_fail:
 919         icmpv6_cleanup();
 920 icmp_fail:
 921 #ifdef CONFIG_SYSCTL
 922         ipv6_sysctl_unregister();
 923 sysctl_fail:
 924 #endif
 925         unregister_pernet_subsys(&inet6_net_ops);
 926 register_pernet_fail:
 927         cleanup_ipv6_mibs();
 928 out_unregister_sock:
 929         sock_unregister(PF_INET6);
 930         rtnl_unregister_all(PF_INET6);
 931 out_sock_register_fail:
 932         rawv6_exit();
 933 out_unregister_raw_proto:
 934         proto_unregister(&rawv6_prot);
 935 out_unregister_udplite_proto:
 936         proto_unregister(&udplitev6_prot);
 937 out_unregister_udp_proto:
 938         proto_unregister(&udpv6_prot);
 939 out_unregister_tcp_proto:
 940         proto_unregister(&tcpv6_prot);
 941         goto out;
 942 }
 943 module_init(inet6_init);
 944
 945 static void __exit inet6_exit(void)
 946 {
 947         /* First of all disallow new sockets creation. */
 948         sock_unregister(PF_INET6);
 949         /* Disallow any further netlink messages */
 950         rtnl_unregister_all(PF_INET6);
 951
 952         udpv6_exit();
 953         udplitev6_exit();
 954         tcpv6_exit();
 955
 956         /* Cleanup code parts. */
 957         ipv6_packet_cleanup();
 958         ipv6_frag_exit();
 959         ipv6_exthdrs_exit();
 960         addrconf_cleanup();
 961         ip6_flowlabel_cleanup();
 962         ip6_route_cleanup();
 963 #ifdef CONFIG_PROC_FS
 964
 965         /* Cleanup code parts. */
 966         if6_proc_exit();
 967         ac6_proc_exit();
 968         ipv6_misc_proc_exit();
 969         udplite6_proc_exit();
 970         udp6_proc_exit();
 971         tcp6_proc_exit();
 972         raw6_proc_exit();
 973 #endif
 974         ipv6_netfilter_fini();
 975         igmp6_cleanup();
 976         ndisc_cleanup();
 977         icmpv6_cleanup();
 978         rawv6_exit();
 979 #ifdef CONFIG_SYSCTL
 980         ipv6_sysctl_unregister();
 981 #endif
 982         unregister_pernet_subsys(&inet6_net_ops);
 983         cleanup_ipv6_mibs();
 984         proto_unregister(&rawv6_prot);
 985         proto_unregister(&udplitev6_prot);
 986         proto_unregister(&udpv6_prot);
 987         proto_unregister(&tcpv6_prot);
 988 }
 989 module_exit(inet6_exit);
 990
 991 MODULE_ALIAS_NETPROTO(PF_INET6);