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  *      Fixes:
  11  *      piggy, Karl Knutson     :       Socket protocol table
  12  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
  13  *      Arnaldo Melo            :       check proc_net_create return, cleanups
  14  *
  15  *      This program is free software; you can redistribute it and/or
  16  *      modify it under the terms of the GNU General Public License
  17  *      as published by the Free Software Foundation; either version
  18  *      2 of the License, or (at your option) any later version.
  19  */
  20
  21 #define pr_fmt(fmt) "IPv6: " fmt
  22
  23 #include <linux/module.h>
  24 #include <linux/capability.h>
  25 #include <linux/errno.h>
  26 #include <linux/types.h>
  27 #include <linux/socket.h>
  28 #include <linux/in.h>
  29 #include <linux/kernel.h>
  30 #include <linux/timer.h>
  31 #include <linux/string.h>
  32 #include <linux/sockios.h>
  33 #include <linux/net.h>
  34 #include <linux/fcntl.h>
  35 #include <linux/mm.h>
  36 #include <linux/interrupt.h>
  37 #include <linux/proc_fs.h>
  38 #include <linux/stat.h>
  39 #include <linux/init.h>
  40 #include <linux/slab.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/protocol.h>
  53 #include <net/inet_common.h>
  54 #include <net/route.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 <linux/mroute6.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 struct ipv6_params ipv6_defaults = {
  76         .disable_ipv6 = 0,
  77         .autoconf = 1,
  78 };
  79
  80 static int disable_ipv6_mod;
  81
  82 module_param_named(disable, disable_ipv6_mod, int, 0444);
  83 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  84
  85 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  86 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  87
  88 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  89 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  90
  91 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  92 {
  93         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  94
  95         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  96 }
  97
  98 static int inet6_create(struct net *net, struct socket *sock, int protocol,
  99                         int kern)
 100 {
 101         struct inet_sock *inet;
 102         struct ipv6_pinfo *np;
 103         struct sock *sk;
 104         struct inet_protosw *answer;
 105         struct proto *answer_prot;
 106         unsigned char answer_flags;
 107         char answer_no_check;
 108         int try_loading_module = 0;
 109         int err;
 110
 111         if (sock->type != SOCK_RAW &&
 112             sock->type != SOCK_DGRAM &&
 113             !inet_ehash_secret)
 114                 build_ehash_secret();
 115
 116         if (protocol < 0 || protocol >= IPPROTO_MAX)
 117                 return -EINVAL;
 118
 119         /* Look for the requested type/protocol pair. */
 120 lookup_protocol:
 121         err = -ESOCKTNOSUPPORT;
 122         rcu_read_lock();
 123         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 124
 125                 err = 0;
 126                 /* Check the non-wild match. */
 127                 if (protocol == answer->protocol) {
 128                         if (protocol != IPPROTO_IP)
 129                                 break;
 130                 } else {
 131                         /* Check for the two wild cases. */
 132                         if (IPPROTO_IP == protocol) {
 133                                 protocol = answer->protocol;
 134                                 break;
 135                         }
 136                         if (IPPROTO_IP == answer->protocol)
 137                                 break;
 138                 }
 139                 err = -EPROTONOSUPPORT;
 140         }
 141
 142         if (err) {
 143                 if (try_loading_module < 2) {
 144                         rcu_read_unlock();
 145                         /*
 146                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 147                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 148                          */
 149                         if (++try_loading_module == 1)
 150                                 request_module("net-pf-%d-proto-%d-type-%d",
 151                                                 PF_INET6, protocol, sock->type);
 152                         /*
 153                          * Fall back to generic, e.g. net-pf-10-proto-132
 154                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 155                          */
 156                         else
 157                                 request_module("net-pf-%d-proto-%d",
 158                                                 PF_INET6, protocol);
 159                         goto lookup_protocol;
 160                 } else
 161                         goto out_rcu_unlock;
 162         }
 163
 164         err = -EPERM;
 165         if (sock->type == SOCK_RAW && !kern &&
 166             !ns_capable(net->user_ns, CAP_NET_RAW))
 167                 goto out_rcu_unlock;
 168
 169         sock->ops = answer->ops;
 170         answer_prot = answer->prot;
 171         answer_no_check = answer->no_check;
 172         answer_flags = answer->flags;
 173         rcu_read_unlock();
 174
 175         WARN_ON(answer_prot->slab == NULL);
 176
 177         err = -ENOBUFS;
 178         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 179         if (sk == NULL)
 180                 goto out;
 181
 182         sock_init_data(sock, sk);
 183
 184         err = 0;
 185         sk->sk_no_check = answer_no_check;
 186         if (INET_PROTOSW_REUSE & answer_flags)
 187                 sk->sk_reuse = SK_CAN_REUSE;
 188
 189         inet = inet_sk(sk);
 190         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 191
 192         if (SOCK_RAW == sock->type) {
 193                 inet->inet_num = protocol;
 194                 if (IPPROTO_RAW == protocol)
 195                         inet->hdrincl = 1;
 196         }
 197
 198         sk->sk_destruct         = inet_sock_destruct;
 199         sk->sk_family           = PF_INET6;
 200         sk->sk_protocol         = protocol;
 201
 202         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 203
 204         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 205         np->hop_limit   = -1;
 206         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 207         np->mc_loop     = 1;
 208         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 209         np->ipv6only    = net->ipv6.sysctl.bindv6only;
 210
 211         /* Init the ipv4 part of the socket since we can have sockets
 212          * using v6 API for ipv4.
 213          */
 214         inet->uc_ttl    = -1;
 215
 216         inet->mc_loop   = 1;
 217         inet->mc_ttl    = 1;
 218         inet->mc_index  = 0;
 219         inet->mc_list   = NULL;
 220         inet->rcv_tos   = 0;
 221
 222         if (ipv4_config.no_pmtu_disc)
 223                 inet->pmtudisc = IP_PMTUDISC_DONT;
 224         else
 225                 inet->pmtudisc = IP_PMTUDISC_WANT;
 226         /*
 227          * Increment only the relevant sk_prot->socks debug field, this changes
 228          * the previous behaviour of incrementing both the equivalent to
 229          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 230          *
 231          * This allows better debug granularity as we'll know exactly how many
 232          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 233          * transport protocol socks. -acme
 234          */
 235         sk_refcnt_debug_inc(sk);
 236
 237         if (inet->inet_num) {
 238                 /* It assumes that any protocol which allows
 239                  * the user to assign a number at socket
 240                  * creation time automatically shares.
 241                  */
 242                 inet->inet_sport = htons(inet->inet_num);
 243                 sk->sk_prot->hash(sk);
 244         }
 245         if (sk->sk_prot->init) {
 246                 err = sk->sk_prot->init(sk);
 247                 if (err) {
 248                         sk_common_release(sk);
 249                         goto out;
 250                 }
 251         }
 252 out:
 253         return err;
 254 out_rcu_unlock:
 255         rcu_read_unlock();
 256         goto out;
 257 }
 258
 259
 260 /* bind for INET6 API */
 261 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 262 {
 263         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 264         struct sock *sk = sock->sk;
 265         struct inet_sock *inet = inet_sk(sk);
 266         struct ipv6_pinfo *np = inet6_sk(sk);
 267         struct net *net = sock_net(sk);
 268         __be32 v4addr = 0;
 269         unsigned short snum;
 270         int addr_type = 0;
 271         int err = 0;
 272
 273         /* If the socket has its own bind function then use it. */
 274         if (sk->sk_prot->bind)
 275                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 276
 277         if (addr_len < SIN6_LEN_RFC2133)
 278                 return -EINVAL;
 279
 280         if (addr->sin6_family != AF_INET6)
 281                 return -EAFNOSUPPORT;
 282
 283         addr_type = ipv6_addr_type(&addr->sin6_addr);
 284         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 285                 return -EINVAL;
 286
 287         snum = ntohs(addr->sin6_port);
 288         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 289                 return -EACCES;
 290
 291         lock_sock(sk);
 292
 293         /* Check these errors (active socket, double bind). */
 294         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 295                 err = -EINVAL;
 296                 goto out;
 297         }
 298
 299         /* Check if the address belongs to the host. */
 300         if (addr_type == IPV6_ADDR_MAPPED) {
 301                 int chk_addr_ret;
 302
 303                 /* Binding to v4-mapped address on a v6-only socket
 304                  * makes no sense
 305                  */
 306                 if (np->ipv6only) {
 307                         err = -EINVAL;
 308                         goto out;
 309                 }
 310
 311                 /* Reproduce AF_INET checks to make the bindings consistent */
 312                 v4addr = addr->sin6_addr.s6_addr32[3];
 313                 chk_addr_ret = inet_addr_type(net, v4addr);
 314                 if (!sysctl_ip_nonlocal_bind &&
 315                     !(inet->freebind || inet->transparent) &&
 316                     v4addr != htonl(INADDR_ANY) &&
 317                     chk_addr_ret != RTN_LOCAL &&
 318                     chk_addr_ret != RTN_MULTICAST &&
 319                     chk_addr_ret != RTN_BROADCAST) {
 320                         err = -EADDRNOTAVAIL;
 321                         goto out;
 322                 }
 323         } else {
 324                 if (addr_type != IPV6_ADDR_ANY) {
 325                         struct net_device *dev = NULL;
 326
 327                         rcu_read_lock();
 328                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 329                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 330                                     addr->sin6_scope_id) {
 331                                         /* Override any existing binding, if another one
 332                                          * is supplied by user.
 333                                          */
 334                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 335                                 }
 336
 337                                 /* Binding to link-local address requires an interface */
 338                                 if (!sk->sk_bound_dev_if) {
 339                                         err = -EINVAL;
 340                                         goto out_unlock;
 341                                 }
 342                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 343                                 if (!dev) {
 344                                         err = -ENODEV;
 345                                         goto out_unlock;
 346                                 }
 347                         }
 348
 349                         /* ipv4 addr of the socket is invalid.  Only the
 350                          * unspecified and mapped address have a v4 equivalent.
 351                          */
 352                         v4addr = LOOPBACK4_IPV6;
 353                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 354                                 if (!(inet->freebind || inet->transparent) &&
 355                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 356                                                    dev, 0)) {
 357                                         err = -EADDRNOTAVAIL;
 358                                         goto out_unlock;
 359                                 }
 360                         }
 361                         rcu_read_unlock();
 362                 }
 363         }
 364
 365         inet->inet_rcv_saddr = v4addr;
 366         inet->inet_saddr = v4addr;
 367
 368         np->rcv_saddr = addr->sin6_addr;
 369
 370         if (!(addr_type & IPV6_ADDR_MULTICAST))
 371                 np->saddr = addr->sin6_addr;
 372
 373         /* Make sure we are allowed to bind here. */
 374         if (sk->sk_prot->get_port(sk, snum)) {
 375                 inet_reset_saddr(sk);
 376                 err = -EADDRINUSE;
 377                 goto out;
 378         }
 379
 380         if (addr_type != IPV6_ADDR_ANY) {
 381                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 382                 if (addr_type != IPV6_ADDR_MAPPED)
 383                         np->ipv6only = 1;
 384         }
 385         if (snum)
 386                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 387         inet->inet_sport = htons(inet->inet_num);
 388         inet->inet_dport = 0;
 389         inet->inet_daddr = 0;
 390 out:
 391         release_sock(sk);
 392         return err;
 393 out_unlock:
 394         rcu_read_unlock();
 395         goto out;
 396 }
 397 EXPORT_SYMBOL(inet6_bind);
 398
 399 int inet6_release(struct socket *sock)
 400 {
 401         struct sock *sk = sock->sk;
 402
 403         if (sk == NULL)
 404                 return -EINVAL;
 405
 406         /* Free mc lists */
 407         ipv6_sock_mc_close(sk);
 408
 409         /* Free ac lists */
 410         ipv6_sock_ac_close(sk);
 411
 412         return inet_release(sock);
 413 }
 414 EXPORT_SYMBOL(inet6_release);
 415
 416 void inet6_destroy_sock(struct sock *sk)
 417 {
 418         struct ipv6_pinfo *np = inet6_sk(sk);
 419         struct sk_buff *skb;
 420         struct ipv6_txoptions *opt;
 421
 422         /* Release rx options */
 423
 424         skb = xchg(&np->pktoptions, NULL);
 425         if (skb != NULL)
 426                 kfree_skb(skb);
 427
 428         skb = xchg(&np->rxpmtu, NULL);
 429         if (skb != NULL)
 430                 kfree_skb(skb);
 431
 432         /* Free flowlabels */
 433         fl6_free_socklist(sk);
 434
 435         /* Free tx options */
 436
 437         opt = xchg(&np->opt, NULL);
 438         if (opt != NULL)
 439                 sock_kfree_s(sk, opt, opt->tot_len);
 440 }
 441 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 442
 443 /*
 444  *      This does both peername and sockname.
 445  */
 446
 447 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 448                  int *uaddr_len, int peer)
 449 {
 450         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 451         struct sock *sk = sock->sk;
 452         struct inet_sock *inet = inet_sk(sk);
 453         struct ipv6_pinfo *np = inet6_sk(sk);
 454
 455         sin->sin6_family = AF_INET6;
 456         sin->sin6_flowinfo = 0;
 457         sin->sin6_scope_id = 0;
 458         if (peer) {
 459                 if (!inet->inet_dport)
 460                         return -ENOTCONN;
 461                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 462                     peer == 1)
 463                         return -ENOTCONN;
 464                 sin->sin6_port = inet->inet_dport;
 465                 sin->sin6_addr = np->daddr;
 466                 if (np->sndflow)
 467                         sin->sin6_flowinfo = np->flow_label;
 468         } else {
 469                 if (ipv6_addr_any(&np->rcv_saddr))
 470                         sin->sin6_addr = np->saddr;
 471                 else
 472                         sin->sin6_addr = np->rcv_saddr;
 473
 474                 sin->sin6_port = inet->inet_sport;
 475         }
 476         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 477                                                  sk->sk_bound_dev_if);
 478         *uaddr_len = sizeof(*sin);
 479         return 0;
 480 }
 481 EXPORT_SYMBOL(inet6_getname);
 482
 483 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 484 {
 485         struct sock *sk = sock->sk;
 486         struct net *net = sock_net(sk);
 487
 488         switch (cmd) {
 489         case SIOCGSTAMP:
 490                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 491
 492         case SIOCGSTAMPNS:
 493                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 494
 495         case SIOCADDRT:
 496         case SIOCDELRT:
 497
 498                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 499
 500         case SIOCSIFADDR:
 501                 return addrconf_add_ifaddr(net, (void __user *) arg);
 502         case SIOCDIFADDR:
 503                 return addrconf_del_ifaddr(net, (void __user *) arg);
 504         case SIOCSIFDSTADDR:
 505                 return addrconf_set_dstaddr(net, (void __user *) arg);
 506         default:
 507                 if (!sk->sk_prot->ioctl)
 508                         return -ENOIOCTLCMD;
 509                 return sk->sk_prot->ioctl(sk, cmd, arg);
 510         }
 511         /*NOTREACHED*/
 512         return 0;
 513 }
 514 EXPORT_SYMBOL(inet6_ioctl);
 515
 516 const struct proto_ops inet6_stream_ops = {
 517         .family            = PF_INET6,
 518         .owner             = THIS_MODULE,
 519         .release           = inet6_release,
 520         .bind              = inet6_bind,
 521         .connect           = inet_stream_connect,       /* ok           */
 522         .socketpair        = sock_no_socketpair,        /* a do nothing */
 523         .accept            = inet_accept,               /* ok           */
 524         .getname           = inet6_getname,
 525         .poll              = tcp_poll,                  /* ok           */
 526         .ioctl             = inet6_ioctl,               /* must change  */
 527         .listen            = inet_listen,               /* ok           */
 528         .shutdown          = inet_shutdown,             /* ok           */
 529         .setsockopt        = sock_common_setsockopt,    /* ok           */
 530         .getsockopt        = sock_common_getsockopt,    /* ok           */
 531         .sendmsg           = inet_sendmsg,              /* ok           */
 532         .recvmsg           = inet_recvmsg,              /* ok           */
 533         .mmap              = sock_no_mmap,
 534         .sendpage          = inet_sendpage,
 535         .splice_read       = tcp_splice_read,
 536 #ifdef CONFIG_COMPAT
 537         .compat_setsockopt = compat_sock_common_setsockopt,
 538         .compat_getsockopt = compat_sock_common_getsockopt,
 539 #endif
 540 };
 541
 542 const struct proto_ops inet6_dgram_ops = {
 543         .family            = PF_INET6,
 544         .owner             = THIS_MODULE,
 545         .release           = inet6_release,
 546         .bind              = inet6_bind,
 547         .connect           = inet_dgram_connect,        /* ok           */
 548         .socketpair        = sock_no_socketpair,        /* a do nothing */
 549         .accept            = sock_no_accept,            /* a do nothing */
 550         .getname           = inet6_getname,
 551         .poll              = udp_poll,                  /* ok           */
 552         .ioctl             = inet6_ioctl,               /* must change  */
 553         .listen            = sock_no_listen,            /* ok           */
 554         .shutdown          = inet_shutdown,             /* ok           */
 555         .setsockopt        = sock_common_setsockopt,    /* ok           */
 556         .getsockopt        = sock_common_getsockopt,    /* ok           */
 557         .sendmsg           = inet_sendmsg,              /* ok           */
 558         .recvmsg           = inet_recvmsg,              /* ok           */
 559         .mmap              = sock_no_mmap,
 560         .sendpage          = sock_no_sendpage,
 561 #ifdef CONFIG_COMPAT
 562         .compat_setsockopt = compat_sock_common_setsockopt,
 563         .compat_getsockopt = compat_sock_common_getsockopt,
 564 #endif
 565 };
 566
 567 static const struct net_proto_family inet6_family_ops = {
 568         .family = PF_INET6,
 569         .create = inet6_create,
 570         .owner  = THIS_MODULE,
 571 };
 572
 573 int inet6_register_protosw(struct inet_protosw *p)
 574 {
 575         struct list_head *lh;
 576         struct inet_protosw *answer;
 577         struct list_head *last_perm;
 578         int protocol = p->protocol;
 579         int ret;
 580
 581         spin_lock_bh(&inetsw6_lock);
 582
 583         ret = -EINVAL;
 584         if (p->type >= SOCK_MAX)
 585                 goto out_illegal;
 586
 587         /* If we are trying to override a permanent protocol, bail. */
 588         answer = NULL;
 589         ret = -EPERM;
 590         last_perm = &inetsw6[p->type];
 591         list_for_each(lh, &inetsw6[p->type]) {
 592                 answer = list_entry(lh, struct inet_protosw, list);
 593
 594                 /* Check only the non-wild match. */
 595                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 596                         if (protocol == answer->protocol)
 597                                 break;
 598                         last_perm = lh;
 599                 }
 600
 601                 answer = NULL;
 602         }
 603         if (answer)
 604                 goto out_permanent;
 605
 606         /* Add the new entry after the last permanent entry if any, so that
 607          * the new entry does not override a permanent entry when matched with
 608          * a wild-card protocol. But it is allowed to override any existing
 609          * non-permanent entry.  This means that when we remove this entry, the
 610          * system automatically returns to the old behavior.
 611          */
 612         list_add_rcu(&p->list, last_perm);
 613         ret = 0;
 614 out:
 615         spin_unlock_bh(&inetsw6_lock);
 616         return ret;
 617
 618 out_permanent:
 619         pr_err("Attempt to override permanent protocol %d\n", protocol);
 620         goto out;
 621
 622 out_illegal:
 623         pr_err("Ignoring attempt to register invalid socket type %d\n",
 624                p->type);
 625         goto out;
 626 }
 627 EXPORT_SYMBOL(inet6_register_protosw);
 628
 629 void
 630 inet6_unregister_protosw(struct inet_protosw *p)
 631 {
 632         if (INET_PROTOSW_PERMANENT & p->flags) {
 633                 pr_err("Attempt to unregister permanent protocol %d\n",
 634                        p->protocol);
 635         } else {
 636                 spin_lock_bh(&inetsw6_lock);
 637                 list_del_rcu(&p->list);
 638                 spin_unlock_bh(&inetsw6_lock);
 639
 640                 synchronize_net();
 641         }
 642 }
 643 EXPORT_SYMBOL(inet6_unregister_protosw);
 644
 645 int inet6_sk_rebuild_header(struct sock *sk)
 646 {
 647         struct ipv6_pinfo *np = inet6_sk(sk);
 648         struct dst_entry *dst;
 649
 650         dst = __sk_dst_check(sk, np->dst_cookie);
 651
 652         if (dst == NULL) {
 653                 struct inet_sock *inet = inet_sk(sk);
 654                 struct in6_addr *final_p, final;
 655                 struct flowi6 fl6;
 656
 657                 memset(&fl6, 0, sizeof(fl6));
 658                 fl6.flowi6_proto = sk->sk_protocol;
 659                 fl6.daddr = np->daddr;
 660                 fl6.saddr = np->saddr;
 661                 fl6.flowlabel = np->flow_label;
 662                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 663                 fl6.flowi6_mark = sk->sk_mark;
 664                 fl6.fl6_dport = inet->inet_dport;
 665                 fl6.fl6_sport = inet->inet_sport;
 666                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 667
 668                 final_p = fl6_update_dst(&fl6, np->opt, &final);
 669
 670                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
 671                 if (IS_ERR(dst)) {
 672                         sk->sk_route_caps = 0;
 673                         sk->sk_err_soft = -PTR_ERR(dst);
 674                         return PTR_ERR(dst);
 675                 }
 676
 677                 __ip6_dst_store(sk, dst, NULL, NULL);
 678         }
 679
 680         return 0;
 681 }
 682 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 683
 684 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
 685 {
 686         const struct ipv6_pinfo *np = inet6_sk(sk);
 687         const struct inet6_skb_parm *opt = IP6CB(skb);
 688
 689         if (np->rxopt.all) {
 690                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 691                                   np->rxopt.bits.ohopopts)) ||
 692                     ((IPV6_FLOWINFO_MASK &
 693                       *(__be32 *)skb_network_header(skb)) &&
 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 true;
 700         }
 701         return false;
 702 }
 703 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 704
 705 static struct packet_type ipv6_packet_type __read_mostly = {
 706         .type = cpu_to_be16(ETH_P_IPV6),
 707         .func = ipv6_rcv,
 708 };
 709
 710 static int __init ipv6_packet_init(void)
 711 {
 712         dev_add_pack(&ipv6_packet_type);
 713         return 0;
 714 }
 715
 716 static void ipv6_packet_cleanup(void)
 717 {
 718         dev_remove_pack(&ipv6_packet_type);
 719 }
 720
 721 static int __net_init ipv6_init_mibs(struct net *net)
 722 {
 723         if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
 724                           sizeof(struct udp_mib),
 725                           __alignof__(struct udp_mib)) < 0)
 726                 return -ENOMEM;
 727         if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
 728                           sizeof(struct udp_mib),
 729                           __alignof__(struct udp_mib)) < 0)
 730                 goto err_udplite_mib;
 731         if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
 732                           sizeof(struct ipstats_mib),
 733                           __alignof__(struct ipstats_mib)) < 0)
 734                 goto err_ip_mib;
 735         if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
 736                           sizeof(struct icmpv6_mib),
 737                           __alignof__(struct icmpv6_mib)) < 0)
 738                 goto err_icmp_mib;
 739         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 740                                                 GFP_KERNEL);
 741         if (!net->mib.icmpv6msg_statistics)
 742                 goto err_icmpmsg_mib;
 743         return 0;
 744
 745 err_icmpmsg_mib:
 746         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 747 err_icmp_mib:
 748         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 749 err_ip_mib:
 750         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 751 err_udplite_mib:
 752         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 753         return -ENOMEM;
 754 }
 755
 756 static void ipv6_cleanup_mibs(struct net *net)
 757 {
 758         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 759         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 760         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 761         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 762         kfree(net->mib.icmpv6msg_statistics);
 763 }
 764
 765 static int __net_init inet6_net_init(struct net *net)
 766 {
 767         int err = 0;
 768
 769         net->ipv6.sysctl.bindv6only = 0;
 770         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 771
 772         err = ipv6_init_mibs(net);
 773         if (err)
 774                 return err;
 775 #ifdef CONFIG_PROC_FS
 776         err = udp6_proc_init(net);
 777         if (err)
 778                 goto out;
 779         err = tcp6_proc_init(net);
 780         if (err)
 781                 goto proc_tcp6_fail;
 782         err = ac6_proc_init(net);
 783         if (err)
 784                 goto proc_ac6_fail;
 785 #endif
 786         return err;
 787
 788 #ifdef CONFIG_PROC_FS
 789 proc_ac6_fail:
 790         tcp6_proc_exit(net);
 791 proc_tcp6_fail:
 792         udp6_proc_exit(net);
 793 out:
 794         ipv6_cleanup_mibs(net);
 795         return err;
 796 #endif
 797 }
 798
 799 static void __net_exit inet6_net_exit(struct net *net)
 800 {
 801 #ifdef CONFIG_PROC_FS
 802         udp6_proc_exit(net);
 803         tcp6_proc_exit(net);
 804         ac6_proc_exit(net);
 805 #endif
 806         ipv6_cleanup_mibs(net);
 807 }
 808
 809 static struct pernet_operations inet6_net_ops = {
 810         .init = inet6_net_init,
 811         .exit = inet6_net_exit,
 812 };
 813
 814 static int __init inet6_init(void)
 815 {
 816         struct list_head *r;
 817         int err = 0;
 818
 819         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
 820
 821         /* Register the socket-side information for inet6_create.  */
 822         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 823                 INIT_LIST_HEAD(r);
 824
 825         if (disable_ipv6_mod) {
 826                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 827                 goto out;
 828         }
 829
 830         err = proto_register(&tcpv6_prot, 1);
 831         if (err)
 832                 goto out;
 833
 834         err = proto_register(&udpv6_prot, 1);
 835         if (err)
 836                 goto out_unregister_tcp_proto;
 837
 838         err = proto_register(&udplitev6_prot, 1);
 839         if (err)
 840                 goto out_unregister_udp_proto;
 841
 842         err = proto_register(&rawv6_prot, 1);
 843         if (err)
 844                 goto out_unregister_udplite_proto;
 845
 846
 847         /* We MUST register RAW sockets before we create the ICMP6,
 848          * IGMP6, or NDISC control sockets.
 849          */
 850         err = rawv6_init();
 851         if (err)
 852                 goto out_unregister_raw_proto;
 853
 854         /* Register the family here so that the init calls below will
 855          * be able to create sockets. (?? is this dangerous ??)
 856          */
 857         err = sock_register(&inet6_family_ops);
 858         if (err)
 859                 goto out_sock_register_fail;
 860
 861         tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
 862
 863         /*
 864          *      ipngwg API draft makes clear that the correct semantics
 865          *      for TCP and UDP is to consider one TCP and UDP instance
 866          *      in a host available by both INET and INET6 APIs and
 867          *      able to communicate via both network protocols.
 868          */
 869
 870         err = register_pernet_subsys(&inet6_net_ops);
 871         if (err)
 872                 goto register_pernet_fail;
 873         err = icmpv6_init();
 874         if (err)
 875                 goto icmp_fail;
 876         err = ip6_mr_init();
 877         if (err)
 878                 goto ipmr_fail;
 879         err = ndisc_init();
 880         if (err)
 881                 goto ndisc_fail;
 882         err = igmp6_init();
 883         if (err)
 884                 goto igmp_fail;
 885         err = ipv6_netfilter_init();
 886         if (err)
 887                 goto netfilter_fail;
 888         /* Create /proc/foo6 entries. */
 889 #ifdef CONFIG_PROC_FS
 890         err = -ENOMEM;
 891         if (raw6_proc_init())
 892                 goto proc_raw6_fail;
 893         if (udplite6_proc_init())
 894                 goto proc_udplite6_fail;
 895         if (ipv6_misc_proc_init())
 896                 goto proc_misc6_fail;
 897         if (if6_proc_init())
 898                 goto proc_if6_fail;
 899 #endif
 900         err = ip6_route_init();
 901         if (err)
 902                 goto ip6_route_fail;
 903         err = ip6_flowlabel_init();
 904         if (err)
 905                 goto ip6_flowlabel_fail;
 906         err = addrconf_init();
 907         if (err)
 908                 goto addrconf_fail;
 909
 910         /* Init v6 extension headers. */
 911         err = ipv6_exthdrs_init();
 912         if (err)
 913                 goto ipv6_exthdrs_fail;
 914
 915         err = ipv6_frag_init();
 916         if (err)
 917                 goto ipv6_frag_fail;
 918
 919         /* Init v6 transport protocols. */
 920         err = udpv6_init();
 921         if (err)
 922                 goto udpv6_fail;
 923
 924         err = udplitev6_init();
 925         if (err)
 926                 goto udplitev6_fail;
 927
 928         err = tcpv6_init();
 929         if (err)
 930                 goto tcpv6_fail;
 931
 932         err = ipv6_packet_init();
 933         if (err)
 934                 goto ipv6_packet_fail;
 935
 936 #ifdef CONFIG_SYSCTL
 937         err = ipv6_sysctl_register();
 938         if (err)
 939                 goto sysctl_fail;
 940 #endif
 941 out:
 942         return err;
 943
 944 #ifdef CONFIG_SYSCTL
 945 sysctl_fail:
 946         ipv6_packet_cleanup();
 947 #endif
 948 ipv6_packet_fail:
 949         tcpv6_exit();
 950 tcpv6_fail:
 951         udplitev6_exit();
 952 udplitev6_fail:
 953         udpv6_exit();
 954 udpv6_fail:
 955         ipv6_frag_exit();
 956 ipv6_frag_fail:
 957         ipv6_exthdrs_exit();
 958 ipv6_exthdrs_fail:
 959         addrconf_cleanup();
 960 addrconf_fail:
 961         ip6_flowlabel_cleanup();
 962 ip6_flowlabel_fail:
 963         ip6_route_cleanup();
 964 ip6_route_fail:
 965 #ifdef CONFIG_PROC_FS
 966         if6_proc_exit();
 967 proc_if6_fail:
 968         ipv6_misc_proc_exit();
 969 proc_misc6_fail:
 970         udplite6_proc_exit();
 971 proc_udplite6_fail:
 972         raw6_proc_exit();
 973 proc_raw6_fail:
 974 #endif
 975         ipv6_netfilter_fini();
 976 netfilter_fail:
 977         igmp6_cleanup();
 978 igmp_fail:
 979         ndisc_cleanup();
 980 ndisc_fail:
 981         ip6_mr_cleanup();
 982 ipmr_fail:
 983         icmpv6_cleanup();
 984 icmp_fail:
 985         unregister_pernet_subsys(&inet6_net_ops);
 986 register_pernet_fail:
 987         sock_unregister(PF_INET6);
 988         rtnl_unregister_all(PF_INET6);
 989 out_sock_register_fail:
 990         rawv6_exit();
 991 out_unregister_raw_proto:
 992         proto_unregister(&rawv6_prot);
 993 out_unregister_udplite_proto:
 994         proto_unregister(&udplitev6_prot);
 995 out_unregister_udp_proto:
 996         proto_unregister(&udpv6_prot);
 997 out_unregister_tcp_proto:
 998         proto_unregister(&tcpv6_prot);
 999         goto out;
1000 }
1001 module_init(inet6_init);
1002
1003 static void __exit inet6_exit(void)
1004 {
1005         if (disable_ipv6_mod)
1006                 return;
1007
1008         /* First of all disallow new sockets creation. */
1009         sock_unregister(PF_INET6);
1010         /* Disallow any further netlink messages */
1011         rtnl_unregister_all(PF_INET6);
1012
1013         udpv6_exit();
1014         udplitev6_exit();
1015         tcpv6_exit();
1016
1017         /* Cleanup code parts. */
1018         ipv6_packet_cleanup();
1019         ipv6_frag_exit();
1020         ipv6_exthdrs_exit();
1021         addrconf_cleanup();
1022         ip6_flowlabel_cleanup();
1023         ip6_route_cleanup();
1024 #ifdef CONFIG_PROC_FS
1025
1026         /* Cleanup code parts. */
1027         if6_proc_exit();
1028         ipv6_misc_proc_exit();
1029         udplite6_proc_exit();
1030         raw6_proc_exit();
1031 #endif
1032         ipv6_netfilter_fini();
1033         igmp6_cleanup();
1034         ndisc_cleanup();
1035         ip6_mr_cleanup();
1036         icmpv6_cleanup();
1037         rawv6_exit();
1038
1039         unregister_pernet_subsys(&inet6_net_ops);
1040         proto_unregister(&rawv6_prot);
1041         proto_unregister(&udplitev6_prot);
1042         proto_unregister(&udpv6_prot);
1043         proto_unregister(&tcpv6_prot);
1044
1045         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1046 }
1047 module_exit(inet6_exit);
1048
1049 MODULE_ALIAS_NETPROTO(PF_INET6);