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/ping.h>
  53 #include <net/protocol.h>
  54 #include <net/inet_common.h>
  55 #include <net/route.h>
  56 #include <net/transp_v6.h>
  57 #include <net/ip6_route.h>
  58 #include <net/addrconf.h>
  59 #include <net/ndisc.h>
  60 #ifdef CONFIG_IPV6_TUNNEL
  61 #include <net/ip6_tunnel.h>
  62 #endif
  63 #include <net/calipso.h>
  64
  65 #include <asm/uaccess.h>
  66 #include <linux/mroute6.h>
  67
  68 #include "ip6_offload.h"
  69
  70 MODULE_AUTHOR("Cast of dozens");
  71 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  72 MODULE_LICENSE("GPL");
  73
  74 /* The inetsw6 table contains everything that inet6_create needs to
  75  * build a new socket.
  76  */
  77 static struct list_head inetsw6[SOCK_MAX];
  78 static DEFINE_SPINLOCK(inetsw6_lock);
  79
  80 struct ipv6_params ipv6_defaults = {
  81         .disable_ipv6 = 0,
  82         .autoconf = 1,
  83 };
  84
  85 static int disable_ipv6_mod;
  86
  87 module_param_named(disable, disable_ipv6_mod, int, 0444);
  88 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  89
  90 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  91 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  92
  93 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  94 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  95
  96 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  97 {
  98         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  99
 100         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
 101 }
 102
 103 static int inet6_create(struct net *net, struct socket *sock, int protocol,
 104                         int kern)
 105 {
 106         struct inet_sock *inet;
 107         struct ipv6_pinfo *np;
 108         struct sock *sk;
 109         struct inet_protosw *answer;
 110         struct proto *answer_prot;
 111         unsigned char answer_flags;
 112         int try_loading_module = 0;
 113         int err;
 114
 115         if (protocol < 0 || protocol >= IPPROTO_MAX)
 116                 return -EINVAL;
 117
 118         /* Look for the requested type/protocol pair. */
 119 lookup_protocol:
 120         err = -ESOCKTNOSUPPORT;
 121         rcu_read_lock();
 122         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 123
 124                 err = 0;
 125                 /* Check the non-wild match. */
 126                 if (protocol == answer->protocol) {
 127                         if (protocol != IPPROTO_IP)
 128                                 break;
 129                 } else {
 130                         /* Check for the two wild cases. */
 131                         if (IPPROTO_IP == protocol) {
 132                                 protocol = answer->protocol;
 133                                 break;
 134                         }
 135                         if (IPPROTO_IP == answer->protocol)
 136                                 break;
 137                 }
 138                 err = -EPROTONOSUPPORT;
 139         }
 140
 141         if (err) {
 142                 if (try_loading_module < 2) {
 143                         rcu_read_unlock();
 144                         /*
 145                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 146                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 147                          */
 148                         if (++try_loading_module == 1)
 149                                 request_module("net-pf-%d-proto-%d-type-%d",
 150                                                 PF_INET6, protocol, sock->type);
 151                         /*
 152                          * Fall back to generic, e.g. net-pf-10-proto-132
 153                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 154                          */
 155                         else
 156                                 request_module("net-pf-%d-proto-%d",
 157                                                 PF_INET6, protocol);
 158                         goto lookup_protocol;
 159                 } else
 160                         goto out_rcu_unlock;
 161         }
 162
 163         err = -EPERM;
 164         if (sock->type == SOCK_RAW && !kern &&
 165             !ns_capable(net->user_ns, CAP_NET_RAW))
 166                 goto out_rcu_unlock;
 167
 168         sock->ops = answer->ops;
 169         answer_prot = answer->prot;
 170         answer_flags = answer->flags;
 171         rcu_read_unlock();
 172
 173         WARN_ON(!answer_prot->slab);
 174
 175         err = -ENOBUFS;
 176         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
 177         if (!sk)
 178                 goto out;
 179
 180         sock_init_data(sock, sk);
 181
 182         err = 0;
 183         if (INET_PROTOSW_REUSE & answer_flags)
 184                 sk->sk_reuse = SK_CAN_REUSE;
 185
 186         inet = inet_sk(sk);
 187         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 188
 189         if (SOCK_RAW == sock->type) {
 190                 inet->inet_num = protocol;
 191                 if (IPPROTO_RAW == protocol)
 192                         inet->hdrincl = 1;
 193         }
 194
 195         sk->sk_destruct         = inet_sock_destruct;
 196         sk->sk_family           = PF_INET6;
 197         sk->sk_protocol         = protocol;
 198
 199         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 200
 201         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 202         np->hop_limit   = -1;
 203         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 204         np->mc_loop     = 1;
 205         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 206         np->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
 207         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
 208
 209         /* Init the ipv4 part of the socket since we can have sockets
 210          * using v6 API for ipv4.
 211          */
 212         inet->uc_ttl    = -1;
 213
 214         inet->mc_loop   = 1;
 215         inet->mc_ttl    = 1;
 216         inet->mc_index  = 0;
 217         inet->mc_list   = NULL;
 218         inet->rcv_tos   = 0;
 219
 220         if (net->ipv4.sysctl_ip_no_pmtu_disc)
 221                 inet->pmtudisc = IP_PMTUDISC_DONT;
 222         else
 223                 inet->pmtudisc = IP_PMTUDISC_WANT;
 224         /*
 225          * Increment only the relevant sk_prot->socks debug field, this changes
 226          * the previous behaviour of incrementing both the equivalent to
 227          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 228          *
 229          * This allows better debug granularity as we'll know exactly how many
 230          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 231          * transport protocol socks. -acme
 232          */
 233         sk_refcnt_debug_inc(sk);
 234
 235         if (inet->inet_num) {
 236                 /* It assumes that any protocol which allows
 237                  * the user to assign a number at socket
 238                  * creation time automatically shares.
 239                  */
 240                 inet->inet_sport = htons(inet->inet_num);
 241                 err = sk->sk_prot->hash(sk);
 242                 if (err) {
 243                         sk_common_release(sk);
 244                         goto out;
 245                 }
 246         }
 247         if (sk->sk_prot->init) {
 248                 err = sk->sk_prot->init(sk);
 249                 if (err) {
 250                         sk_common_release(sk);
 251                         goto out;
 252                 }
 253         }
 254 out:
 255         return err;
 256 out_rcu_unlock:
 257         rcu_read_unlock();
 258         goto out;
 259 }
 260
 261
 262 /* bind for INET6 API */
 263 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 264 {
 265         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 266         struct sock *sk = sock->sk;
 267         struct inet_sock *inet = inet_sk(sk);
 268         struct ipv6_pinfo *np = inet6_sk(sk);
 269         struct net *net = sock_net(sk);
 270         __be32 v4addr = 0;
 271         unsigned short snum;
 272         int addr_type = 0;
 273         int err = 0;
 274
 275         /* If the socket has its own bind function then use it. */
 276         if (sk->sk_prot->bind)
 277                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 278
 279         if (addr_len < SIN6_LEN_RFC2133)
 280                 return -EINVAL;
 281
 282         if (addr->sin6_family != AF_INET6)
 283                 return -EAFNOSUPPORT;
 284
 285         addr_type = ipv6_addr_type(&addr->sin6_addr);
 286         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 287                 return -EINVAL;
 288
 289         snum = ntohs(addr->sin6_port);
 290         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 291                 return -EACCES;
 292
 293         lock_sock(sk);
 294
 295         /* Check these errors (active socket, double bind). */
 296         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 297                 err = -EINVAL;
 298                 goto out;
 299         }
 300
 301         /* Check if the address belongs to the host. */
 302         if (addr_type == IPV6_ADDR_MAPPED) {
 303                 int chk_addr_ret;
 304
 305                 /* Binding to v4-mapped address on a v6-only socket
 306                  * makes no sense
 307                  */
 308                 if (sk->sk_ipv6only) {
 309                         err = -EINVAL;
 310                         goto out;
 311                 }
 312
 313                 /* Reproduce AF_INET checks to make the bindings consistent */
 314                 v4addr = addr->sin6_addr.s6_addr32[3];
 315                 chk_addr_ret = inet_addr_type(net, v4addr);
 316                 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
 317                     !(inet->freebind || inet->transparent) &&
 318                     v4addr != htonl(INADDR_ANY) &&
 319                     chk_addr_ret != RTN_LOCAL &&
 320                     chk_addr_ret != RTN_MULTICAST &&
 321                     chk_addr_ret != RTN_BROADCAST) {
 322                         err = -EADDRNOTAVAIL;
 323                         goto out;
 324                 }
 325         } else {
 326                 if (addr_type != IPV6_ADDR_ANY) {
 327                         struct net_device *dev = NULL;
 328
 329                         rcu_read_lock();
 330                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 331                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 332                                     addr->sin6_scope_id) {
 333                                         /* Override any existing binding, if another one
 334                                          * is supplied by user.
 335                                          */
 336                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 337                                 }
 338
 339                                 /* Binding to link-local address requires an interface */
 340                                 if (!sk->sk_bound_dev_if) {
 341                                         err = -EINVAL;
 342                                         goto out_unlock;
 343                                 }
 344                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 345                                 if (!dev) {
 346                                         err = -ENODEV;
 347                                         goto out_unlock;
 348                                 }
 349                         }
 350
 351                         /* ipv4 addr of the socket is invalid.  Only the
 352                          * unspecified and mapped address have a v4 equivalent.
 353                          */
 354                         v4addr = LOOPBACK4_IPV6;
 355                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 356                                 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
 357                                     !(inet->freebind || inet->transparent) &&
 358                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 359                                                    dev, 0)) {
 360                                         err = -EADDRNOTAVAIL;
 361                                         goto out_unlock;
 362                                 }
 363                         }
 364                         rcu_read_unlock();
 365                 }
 366         }
 367
 368         inet->inet_rcv_saddr = v4addr;
 369         inet->inet_saddr = v4addr;
 370
 371         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 372
 373         if (!(addr_type & IPV6_ADDR_MULTICAST))
 374                 np->saddr = addr->sin6_addr;
 375
 376         /* Make sure we are allowed to bind here. */
 377         if ((snum || !inet->bind_address_no_port) &&
 378             sk->sk_prot->get_port(sk, snum)) {
 379                 inet_reset_saddr(sk);
 380                 err = -EADDRINUSE;
 381                 goto out;
 382         }
 383
 384         if (addr_type != IPV6_ADDR_ANY) {
 385                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 386                 if (addr_type != IPV6_ADDR_MAPPED)
 387                         sk->sk_ipv6only = 1;
 388         }
 389         if (snum)
 390                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 391         inet->inet_sport = htons(inet->inet_num);
 392         inet->inet_dport = 0;
 393         inet->inet_daddr = 0;
 394 out:
 395         release_sock(sk);
 396         return err;
 397 out_unlock:
 398         rcu_read_unlock();
 399         goto out;
 400 }
 401 EXPORT_SYMBOL(inet6_bind);
 402
 403 int inet6_release(struct socket *sock)
 404 {
 405         struct sock *sk = sock->sk;
 406
 407         if (!sk)
 408                 return -EINVAL;
 409
 410         /* Free mc lists */
 411         ipv6_sock_mc_close(sk);
 412
 413         /* Free ac lists */
 414         ipv6_sock_ac_close(sk);
 415
 416         return inet_release(sock);
 417 }
 418 EXPORT_SYMBOL(inet6_release);
 419
 420 void inet6_destroy_sock(struct sock *sk)
 421 {
 422         struct ipv6_pinfo *np = inet6_sk(sk);
 423         struct sk_buff *skb;
 424         struct ipv6_txoptions *opt;
 425
 426         /* Release rx options */
 427
 428         skb = xchg(&np->pktoptions, NULL);
 429         if (skb)
 430                 kfree_skb(skb);
 431
 432         skb = xchg(&np->rxpmtu, NULL);
 433         if (skb)
 434                 kfree_skb(skb);
 435
 436         /* Free flowlabels */
 437         fl6_free_socklist(sk);
 438
 439         /* Free tx options */
 440
 441         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 442         if (opt) {
 443                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 444                 txopt_put(opt);
 445         }
 446 }
 447 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 448
 449 /*
 450  *      This does both peername and sockname.
 451  */
 452
 453 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 454                  int *uaddr_len, int peer)
 455 {
 456         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 457         struct sock *sk = sock->sk;
 458         struct inet_sock *inet = inet_sk(sk);
 459         struct ipv6_pinfo *np = inet6_sk(sk);
 460
 461         sin->sin6_family = AF_INET6;
 462         sin->sin6_flowinfo = 0;
 463         sin->sin6_scope_id = 0;
 464         if (peer) {
 465                 if (!inet->inet_dport)
 466                         return -ENOTCONN;
 467                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 468                     peer == 1)
 469                         return -ENOTCONN;
 470                 sin->sin6_port = inet->inet_dport;
 471                 sin->sin6_addr = sk->sk_v6_daddr;
 472                 if (np->sndflow)
 473                         sin->sin6_flowinfo = np->flow_label;
 474         } else {
 475                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 476                         sin->sin6_addr = np->saddr;
 477                 else
 478                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 479
 480                 sin->sin6_port = inet->inet_sport;
 481         }
 482         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 483                                                  sk->sk_bound_dev_if);
 484         *uaddr_len = sizeof(*sin);
 485         return 0;
 486 }
 487 EXPORT_SYMBOL(inet6_getname);
 488
 489 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 490 {
 491         struct sock *sk = sock->sk;
 492         struct net *net = sock_net(sk);
 493
 494         switch (cmd) {
 495         case SIOCGSTAMP:
 496                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 497
 498         case SIOCGSTAMPNS:
 499                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 500
 501         case SIOCADDRT:
 502         case SIOCDELRT:
 503
 504                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 505
 506         case SIOCSIFADDR:
 507                 return addrconf_add_ifaddr(net, (void __user *) arg);
 508         case SIOCDIFADDR:
 509                 return addrconf_del_ifaddr(net, (void __user *) arg);
 510         case SIOCSIFDSTADDR:
 511                 return addrconf_set_dstaddr(net, (void __user *) arg);
 512         default:
 513                 if (!sk->sk_prot->ioctl)
 514                         return -ENOIOCTLCMD;
 515                 return sk->sk_prot->ioctl(sk, cmd, arg);
 516         }
 517         /*NOTREACHED*/
 518         return 0;
 519 }
 520 EXPORT_SYMBOL(inet6_ioctl);
 521
 522 const struct proto_ops inet6_stream_ops = {
 523         .family            = PF_INET6,
 524         .owner             = THIS_MODULE,
 525         .release           = inet6_release,
 526         .bind              = inet6_bind,
 527         .connect           = inet_stream_connect,       /* ok           */
 528         .socketpair        = sock_no_socketpair,        /* a do nothing */
 529         .accept            = inet_accept,               /* ok           */
 530         .getname           = inet6_getname,
 531         .poll              = tcp_poll,                  /* ok           */
 532         .ioctl             = inet6_ioctl,               /* must change  */
 533         .listen            = inet_listen,               /* ok           */
 534         .shutdown          = inet_shutdown,             /* ok           */
 535         .setsockopt        = sock_common_setsockopt,    /* ok           */
 536         .getsockopt        = sock_common_getsockopt,    /* ok           */
 537         .sendmsg           = inet_sendmsg,              /* ok           */
 538         .recvmsg           = inet_recvmsg,              /* ok           */
 539         .mmap              = sock_no_mmap,
 540         .sendpage          = inet_sendpage,
 541         .splice_read       = tcp_splice_read,
 542 #ifdef CONFIG_COMPAT
 543         .compat_setsockopt = compat_sock_common_setsockopt,
 544         .compat_getsockopt = compat_sock_common_getsockopt,
 545 #endif
 546 };
 547
 548 const struct proto_ops inet6_dgram_ops = {
 549         .family            = PF_INET6,
 550         .owner             = THIS_MODULE,
 551         .release           = inet6_release,
 552         .bind              = inet6_bind,
 553         .connect           = inet_dgram_connect,        /* ok           */
 554         .socketpair        = sock_no_socketpair,        /* a do nothing */
 555         .accept            = sock_no_accept,            /* a do nothing */
 556         .getname           = inet6_getname,
 557         .poll              = udp_poll,                  /* ok           */
 558         .ioctl             = inet6_ioctl,               /* must change  */
 559         .listen            = sock_no_listen,            /* ok           */
 560         .shutdown          = inet_shutdown,             /* ok           */
 561         .setsockopt        = sock_common_setsockopt,    /* ok           */
 562         .getsockopt        = sock_common_getsockopt,    /* ok           */
 563         .sendmsg           = inet_sendmsg,              /* ok           */
 564         .recvmsg           = inet_recvmsg,              /* ok           */
 565         .mmap              = sock_no_mmap,
 566         .sendpage          = sock_no_sendpage,
 567         .set_peek_off      = sk_set_peek_off,
 568 #ifdef CONFIG_COMPAT
 569         .compat_setsockopt = compat_sock_common_setsockopt,
 570         .compat_getsockopt = compat_sock_common_getsockopt,
 571 #endif
 572 };
 573
 574 static const struct net_proto_family inet6_family_ops = {
 575         .family = PF_INET6,
 576         .create = inet6_create,
 577         .owner  = THIS_MODULE,
 578 };
 579
 580 int inet6_register_protosw(struct inet_protosw *p)
 581 {
 582         struct list_head *lh;
 583         struct inet_protosw *answer;
 584         struct list_head *last_perm;
 585         int protocol = p->protocol;
 586         int ret;
 587
 588         spin_lock_bh(&inetsw6_lock);
 589
 590         ret = -EINVAL;
 591         if (p->type >= SOCK_MAX)
 592                 goto out_illegal;
 593
 594         /* If we are trying to override a permanent protocol, bail. */
 595         answer = NULL;
 596         ret = -EPERM;
 597         last_perm = &inetsw6[p->type];
 598         list_for_each(lh, &inetsw6[p->type]) {
 599                 answer = list_entry(lh, struct inet_protosw, list);
 600
 601                 /* Check only the non-wild match. */
 602                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 603                         if (protocol == answer->protocol)
 604                                 break;
 605                         last_perm = lh;
 606                 }
 607
 608                 answer = NULL;
 609         }
 610         if (answer)
 611                 goto out_permanent;
 612
 613         /* Add the new entry after the last permanent entry if any, so that
 614          * the new entry does not override a permanent entry when matched with
 615          * a wild-card protocol. But it is allowed to override any existing
 616          * non-permanent entry.  This means that when we remove this entry, the
 617          * system automatically returns to the old behavior.
 618          */
 619         list_add_rcu(&p->list, last_perm);
 620         ret = 0;
 621 out:
 622         spin_unlock_bh(&inetsw6_lock);
 623         return ret;
 624
 625 out_permanent:
 626         pr_err("Attempt to override permanent protocol %d\n", protocol);
 627         goto out;
 628
 629 out_illegal:
 630         pr_err("Ignoring attempt to register invalid socket type %d\n",
 631                p->type);
 632         goto out;
 633 }
 634 EXPORT_SYMBOL(inet6_register_protosw);
 635
 636 void
 637 inet6_unregister_protosw(struct inet_protosw *p)
 638 {
 639         if (INET_PROTOSW_PERMANENT & p->flags) {
 640                 pr_err("Attempt to unregister permanent protocol %d\n",
 641                        p->protocol);
 642         } else {
 643                 spin_lock_bh(&inetsw6_lock);
 644                 list_del_rcu(&p->list);
 645                 spin_unlock_bh(&inetsw6_lock);
 646
 647                 synchronize_net();
 648         }
 649 }
 650 EXPORT_SYMBOL(inet6_unregister_protosw);
 651
 652 int inet6_sk_rebuild_header(struct sock *sk)
 653 {
 654         struct ipv6_pinfo *np = inet6_sk(sk);
 655         struct dst_entry *dst;
 656
 657         dst = __sk_dst_check(sk, np->dst_cookie);
 658
 659         if (!dst) {
 660                 struct inet_sock *inet = inet_sk(sk);
 661                 struct in6_addr *final_p, final;
 662                 struct flowi6 fl6;
 663
 664                 memset(&fl6, 0, sizeof(fl6));
 665                 fl6.flowi6_proto = sk->sk_protocol;
 666                 fl6.daddr = sk->sk_v6_daddr;
 667                 fl6.saddr = np->saddr;
 668                 fl6.flowlabel = np->flow_label;
 669                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 670                 fl6.flowi6_mark = sk->sk_mark;
 671                 fl6.fl6_dport = inet->inet_dport;
 672                 fl6.fl6_sport = inet->inet_sport;
 673                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 674
 675                 rcu_read_lock();
 676                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 677                                          &final);
 678                 rcu_read_unlock();
 679
 680                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 681                 if (IS_ERR(dst)) {
 682                         sk->sk_route_caps = 0;
 683                         sk->sk_err_soft = -PTR_ERR(dst);
 684                         return PTR_ERR(dst);
 685                 }
 686
 687                 ip6_dst_store(sk, dst, NULL, NULL);
 688         }
 689
 690         return 0;
 691 }
 692 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 693
 694 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 695                        const struct inet6_skb_parm *opt)
 696 {
 697         const struct ipv6_pinfo *np = inet6_sk(sk);
 698
 699         if (np->rxopt.all) {
 700                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
 701                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 702                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 703                      np->rxopt.bits.rxflow) ||
 704                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 705                      np->rxopt.bits.osrcrt)) ||
 706                     ((opt->dst1 || opt->dst0) &&
 707                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 708                         return true;
 709         }
 710         return false;
 711 }
 712 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 713
 714 static struct packet_type ipv6_packet_type __read_mostly = {
 715         .type = cpu_to_be16(ETH_P_IPV6),
 716         .func = ipv6_rcv,
 717 };
 718
 719 static int __init ipv6_packet_init(void)
 720 {
 721         dev_add_pack(&ipv6_packet_type);
 722         return 0;
 723 }
 724
 725 static void ipv6_packet_cleanup(void)
 726 {
 727         dev_remove_pack(&ipv6_packet_type);
 728 }
 729
 730 static int __net_init ipv6_init_mibs(struct net *net)
 731 {
 732         int i;
 733
 734         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 735         if (!net->mib.udp_stats_in6)
 736                 return -ENOMEM;
 737         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 738         if (!net->mib.udplite_stats_in6)
 739                 goto err_udplite_mib;
 740         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 741         if (!net->mib.ipv6_statistics)
 742                 goto err_ip_mib;
 743
 744         for_each_possible_cpu(i) {
 745                 struct ipstats_mib *af_inet6_stats;
 746                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 747                 u64_stats_init(&af_inet6_stats->syncp);
 748         }
 749
 750
 751         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 752         if (!net->mib.icmpv6_statistics)
 753                 goto err_icmp_mib;
 754         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 755                                                 GFP_KERNEL);
 756         if (!net->mib.icmpv6msg_statistics)
 757                 goto err_icmpmsg_mib;
 758         return 0;
 759
 760 err_icmpmsg_mib:
 761         free_percpu(net->mib.icmpv6_statistics);
 762 err_icmp_mib:
 763         free_percpu(net->mib.ipv6_statistics);
 764 err_ip_mib:
 765         free_percpu(net->mib.udplite_stats_in6);
 766 err_udplite_mib:
 767         free_percpu(net->mib.udp_stats_in6);
 768         return -ENOMEM;
 769 }
 770
 771 static void ipv6_cleanup_mibs(struct net *net)
 772 {
 773         free_percpu(net->mib.udp_stats_in6);
 774         free_percpu(net->mib.udplite_stats_in6);
 775         free_percpu(net->mib.ipv6_statistics);
 776         free_percpu(net->mib.icmpv6_statistics);
 777         kfree(net->mib.icmpv6msg_statistics);
 778 }
 779
 780 static int __net_init inet6_net_init(struct net *net)
 781 {
 782         int err = 0;
 783
 784         net->ipv6.sysctl.bindv6only = 0;
 785         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 786         net->ipv6.sysctl.flowlabel_consistency = 1;
 787         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
 788         net->ipv6.sysctl.idgen_retries = 3;
 789         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 790         net->ipv6.sysctl.flowlabel_state_ranges = 0;
 791         atomic_set(&net->ipv6.fib6_sernum, 1);
 792
 793         err = ipv6_init_mibs(net);
 794         if (err)
 795                 return err;
 796 #ifdef CONFIG_PROC_FS
 797         err = udp6_proc_init(net);
 798         if (err)
 799                 goto out;
 800         err = tcp6_proc_init(net);
 801         if (err)
 802                 goto proc_tcp6_fail;
 803         err = ac6_proc_init(net);
 804         if (err)
 805                 goto proc_ac6_fail;
 806 #endif
 807         return err;
 808
 809 #ifdef CONFIG_PROC_FS
 810 proc_ac6_fail:
 811         tcp6_proc_exit(net);
 812 proc_tcp6_fail:
 813         udp6_proc_exit(net);
 814 out:
 815         ipv6_cleanup_mibs(net);
 816         return err;
 817 #endif
 818 }
 819
 820 static void __net_exit inet6_net_exit(struct net *net)
 821 {
 822 #ifdef CONFIG_PROC_FS
 823         udp6_proc_exit(net);
 824         tcp6_proc_exit(net);
 825         ac6_proc_exit(net);
 826 #endif
 827         ipv6_cleanup_mibs(net);
 828 }
 829
 830 static struct pernet_operations inet6_net_ops = {
 831         .init = inet6_net_init,
 832         .exit = inet6_net_exit,
 833 };
 834
 835 static const struct ipv6_stub ipv6_stub_impl = {
 836         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 837         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 838         .ipv6_dst_lookup = ip6_dst_lookup,
 839         .udpv6_encap_enable = udpv6_encap_enable,
 840         .ndisc_send_na = ndisc_send_na,
 841         .nd_tbl = &nd_tbl,
 842 };
 843
 844 static int __init inet6_init(void)
 845 {
 846         struct list_head *r;
 847         int err = 0;
 848
 849         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 850
 851         /* Register the socket-side information for inet6_create.  */
 852         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 853                 INIT_LIST_HEAD(r);
 854
 855         if (disable_ipv6_mod) {
 856                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 857                 goto out;
 858         }
 859
 860         err = proto_register(&tcpv6_prot, 1);
 861         if (err)
 862                 goto out;
 863
 864         err = proto_register(&udpv6_prot, 1);
 865         if (err)
 866                 goto out_unregister_tcp_proto;
 867
 868         err = proto_register(&udplitev6_prot, 1);
 869         if (err)
 870                 goto out_unregister_udp_proto;
 871
 872         err = proto_register(&rawv6_prot, 1);
 873         if (err)
 874                 goto out_unregister_udplite_proto;
 875
 876         err = proto_register(&pingv6_prot, 1);
 877         if (err)
 878                 goto out_unregister_ping_proto;
 879
 880         /* We MUST register RAW sockets before we create the ICMP6,
 881          * IGMP6, or NDISC control sockets.
 882          */
 883         err = rawv6_init();
 884         if (err)
 885                 goto out_unregister_raw_proto;
 886
 887         /* Register the family here so that the init calls below will
 888          * be able to create sockets. (?? is this dangerous ??)
 889          */
 890         err = sock_register(&inet6_family_ops);
 891         if (err)
 892                 goto out_sock_register_fail;
 893
 894         /*
 895          *      ipngwg API draft makes clear that the correct semantics
 896          *      for TCP and UDP is to consider one TCP and UDP instance
 897          *      in a host available by both INET and INET6 APIs and
 898          *      able to communicate via both network protocols.
 899          */
 900
 901         err = register_pernet_subsys(&inet6_net_ops);
 902         if (err)
 903                 goto register_pernet_fail;
 904         err = icmpv6_init();
 905         if (err)
 906                 goto icmp_fail;
 907         err = ip6_mr_init();
 908         if (err)
 909                 goto ipmr_fail;
 910         err = ndisc_init();
 911         if (err)
 912                 goto ndisc_fail;
 913         err = igmp6_init();
 914         if (err)
 915                 goto igmp_fail;
 916
 917         ipv6_stub = &ipv6_stub_impl;
 918
 919         err = ipv6_netfilter_init();
 920         if (err)
 921                 goto netfilter_fail;
 922         /* Create /proc/foo6 entries. */
 923 #ifdef CONFIG_PROC_FS
 924         err = -ENOMEM;
 925         if (raw6_proc_init())
 926                 goto proc_raw6_fail;
 927         if (udplite6_proc_init())
 928                 goto proc_udplite6_fail;
 929         if (ipv6_misc_proc_init())
 930                 goto proc_misc6_fail;
 931         if (if6_proc_init())
 932                 goto proc_if6_fail;
 933 #endif
 934         err = ip6_route_init();
 935         if (err)
 936                 goto ip6_route_fail;
 937         err = ndisc_late_init();
 938         if (err)
 939                 goto ndisc_late_fail;
 940         err = ip6_flowlabel_init();
 941         if (err)
 942                 goto ip6_flowlabel_fail;
 943         err = addrconf_init();
 944         if (err)
 945                 goto addrconf_fail;
 946
 947         /* Init v6 extension headers. */
 948         err = ipv6_exthdrs_init();
 949         if (err)
 950                 goto ipv6_exthdrs_fail;
 951
 952         err = ipv6_frag_init();
 953         if (err)
 954                 goto ipv6_frag_fail;
 955
 956         /* Init v6 transport protocols. */
 957         err = udpv6_init();
 958         if (err)
 959                 goto udpv6_fail;
 960
 961         err = udplitev6_init();
 962         if (err)
 963                 goto udplitev6_fail;
 964
 965         err = udpv6_offload_init();
 966         if (err)
 967                 goto udpv6_offload_fail;
 968
 969         err = tcpv6_init();
 970         if (err)
 971                 goto tcpv6_fail;
 972
 973         err = ipv6_packet_init();
 974         if (err)
 975                 goto ipv6_packet_fail;
 976
 977         err = pingv6_init();
 978         if (err)
 979                 goto pingv6_fail;
 980
 981         err = calipso_init();
 982         if (err)
 983                 goto calipso_fail;
 984
 985 #ifdef CONFIG_SYSCTL
 986         err = ipv6_sysctl_register();
 987         if (err)
 988                 goto sysctl_fail;
 989 #endif
 990 out:
 991         return err;
 992
 993 #ifdef CONFIG_SYSCTL
 994 sysctl_fail:
 995         calipso_exit();
 996 #endif
 997 calipso_fail:
 998         pingv6_exit();
 999 pingv6_fail:
1000         ipv6_packet_cleanup();
1001 ipv6_packet_fail:
1002         tcpv6_exit();
1003 tcpv6_fail:
1004         udpv6_offload_exit();
1005 udpv6_offload_fail:
1006         udplitev6_exit();
1007 udplitev6_fail:
1008         udpv6_exit();
1009 udpv6_fail:
1010         ipv6_frag_exit();
1011 ipv6_frag_fail:
1012         ipv6_exthdrs_exit();
1013 ipv6_exthdrs_fail:
1014         addrconf_cleanup();
1015 addrconf_fail:
1016         ip6_flowlabel_cleanup();
1017 ip6_flowlabel_fail:
1018         ndisc_late_cleanup();
1019 ndisc_late_fail:
1020         ip6_route_cleanup();
1021 ip6_route_fail:
1022 #ifdef CONFIG_PROC_FS
1023         if6_proc_exit();
1024 proc_if6_fail:
1025         ipv6_misc_proc_exit();
1026 proc_misc6_fail:
1027         udplite6_proc_exit();
1028 proc_udplite6_fail:
1029         raw6_proc_exit();
1030 proc_raw6_fail:
1031 #endif
1032         ipv6_netfilter_fini();
1033 netfilter_fail:
1034         igmp6_cleanup();
1035 igmp_fail:
1036         ndisc_cleanup();
1037 ndisc_fail:
1038         ip6_mr_cleanup();
1039 ipmr_fail:
1040         icmpv6_cleanup();
1041 icmp_fail:
1042         unregister_pernet_subsys(&inet6_net_ops);
1043 register_pernet_fail:
1044         sock_unregister(PF_INET6);
1045         rtnl_unregister_all(PF_INET6);
1046 out_sock_register_fail:
1047         rawv6_exit();
1048 out_unregister_ping_proto:
1049         proto_unregister(&pingv6_prot);
1050 out_unregister_raw_proto:
1051         proto_unregister(&rawv6_prot);
1052 out_unregister_udplite_proto:
1053         proto_unregister(&udplitev6_prot);
1054 out_unregister_udp_proto:
1055         proto_unregister(&udpv6_prot);
1056 out_unregister_tcp_proto:
1057         proto_unregister(&tcpv6_prot);
1058         goto out;
1059 }
1060 module_init(inet6_init);
1061
1062 MODULE_ALIAS_NETPROTO(PF_INET6);