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 #include <net/seg6.h>
  65
  66 #include <asm/uaccess.h>
  67 #include <linux/mroute6.h>
  68
  69 #include "ip6_offload.h"
  70
  71 MODULE_AUTHOR("Cast of dozens");
  72 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  73 MODULE_LICENSE("GPL");
  74
  75 /* The inetsw6 table contains everything that inet6_create needs to
  76  * build a new socket.
  77  */
  78 static struct list_head inetsw6[SOCK_MAX];
  79 static DEFINE_SPINLOCK(inetsw6_lock);
  80
  81 struct ipv6_params ipv6_defaults = {
  82         .disable_ipv6 = 0,
  83         .autoconf = 1,
  84 };
  85
  86 static int disable_ipv6_mod;
  87
  88 module_param_named(disable, disable_ipv6_mod, int, 0444);
  89 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  90
  91 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  92 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  93
  94 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  95 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  96
  97 bool ipv6_mod_enabled(void)
  98 {
  99         return disable_ipv6_mod == 0;
 100 }
 101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
 102
 103 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
 104 {
 105         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
 106
 107         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
 108 }
 109
 110 static int inet6_create(struct net *net, struct socket *sock, int protocol,
 111                         int kern)
 112 {
 113         struct inet_sock *inet;
 114         struct ipv6_pinfo *np;
 115         struct sock *sk;
 116         struct inet_protosw *answer;
 117         struct proto *answer_prot;
 118         unsigned char answer_flags;
 119         int try_loading_module = 0;
 120         int err;
 121
 122         if (protocol < 0 || protocol >= IPPROTO_MAX)
 123                 return -EINVAL;
 124
 125         /* Look for the requested type/protocol pair. */
 126 lookup_protocol:
 127         err = -ESOCKTNOSUPPORT;
 128         rcu_read_lock();
 129         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 130
 131                 err = 0;
 132                 /* Check the non-wild match. */
 133                 if (protocol == answer->protocol) {
 134                         if (protocol != IPPROTO_IP)
 135                                 break;
 136                 } else {
 137                         /* Check for the two wild cases. */
 138                         if (IPPROTO_IP == protocol) {
 139                                 protocol = answer->protocol;
 140                                 break;
 141                         }
 142                         if (IPPROTO_IP == answer->protocol)
 143                                 break;
 144                 }
 145                 err = -EPROTONOSUPPORT;
 146         }
 147
 148         if (err) {
 149                 if (try_loading_module < 2) {
 150                         rcu_read_unlock();
 151                         /*
 152                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 153                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 154                          */
 155                         if (++try_loading_module == 1)
 156                                 request_module("net-pf-%d-proto-%d-type-%d",
 157                                                 PF_INET6, protocol, sock->type);
 158                         /*
 159                          * Fall back to generic, e.g. net-pf-10-proto-132
 160                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 161                          */
 162                         else
 163                                 request_module("net-pf-%d-proto-%d",
 164                                                 PF_INET6, protocol);
 165                         goto lookup_protocol;
 166                 } else
 167                         goto out_rcu_unlock;
 168         }
 169
 170         err = -EPERM;
 171         if (sock->type == SOCK_RAW && !kern &&
 172             !ns_capable(net->user_ns, CAP_NET_RAW))
 173                 goto out_rcu_unlock;
 174
 175         sock->ops = answer->ops;
 176         answer_prot = answer->prot;
 177         answer_flags = answer->flags;
 178         rcu_read_unlock();
 179
 180         WARN_ON(!answer_prot->slab);
 181
 182         err = -ENOBUFS;
 183         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
 184         if (!sk)
 185                 goto out;
 186
 187         sock_init_data(sock, sk);
 188
 189         err = 0;
 190         if (INET_PROTOSW_REUSE & answer_flags)
 191                 sk->sk_reuse = SK_CAN_REUSE;
 192
 193         inet = inet_sk(sk);
 194         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 195
 196         if (SOCK_RAW == sock->type) {
 197                 inet->inet_num = protocol;
 198                 if (IPPROTO_RAW == protocol)
 199                         inet->hdrincl = 1;
 200         }
 201
 202         sk->sk_destruct         = inet_sock_destruct;
 203         sk->sk_family           = PF_INET6;
 204         sk->sk_protocol         = protocol;
 205
 206         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 207
 208         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 209         np->hop_limit   = -1;
 210         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 211         np->mc_loop     = 1;
 212         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 213         np->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
 214         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
 215
 216         /* Init the ipv4 part of the socket since we can have sockets
 217          * using v6 API for ipv4.
 218          */
 219         inet->uc_ttl    = -1;
 220
 221         inet->mc_loop   = 1;
 222         inet->mc_ttl    = 1;
 223         inet->mc_index  = 0;
 224         inet->mc_list   = NULL;
 225         inet->rcv_tos   = 0;
 226
 227         if (net->ipv4.sysctl_ip_no_pmtu_disc)
 228                 inet->pmtudisc = IP_PMTUDISC_DONT;
 229         else
 230                 inet->pmtudisc = IP_PMTUDISC_WANT;
 231         /*
 232          * Increment only the relevant sk_prot->socks debug field, this changes
 233          * the previous behaviour of incrementing both the equivalent to
 234          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 235          *
 236          * This allows better debug granularity as we'll know exactly how many
 237          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 238          * transport protocol socks. -acme
 239          */
 240         sk_refcnt_debug_inc(sk);
 241
 242         if (inet->inet_num) {
 243                 /* It assumes that any protocol which allows
 244                  * the user to assign a number at socket
 245                  * creation time automatically shares.
 246                  */
 247                 inet->inet_sport = htons(inet->inet_num);
 248                 err = sk->sk_prot->hash(sk);
 249                 if (err) {
 250                         sk_common_release(sk);
 251                         goto out;
 252                 }
 253         }
 254         if (sk->sk_prot->init) {
 255                 err = sk->sk_prot->init(sk);
 256                 if (err) {
 257                         sk_common_release(sk);
 258                         goto out;
 259                 }
 260         }
 261 out:
 262         return err;
 263 out_rcu_unlock:
 264         rcu_read_unlock();
 265         goto out;
 266 }
 267
 268
 269 /* bind for INET6 API */
 270 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 271 {
 272         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 273         struct sock *sk = sock->sk;
 274         struct inet_sock *inet = inet_sk(sk);
 275         struct ipv6_pinfo *np = inet6_sk(sk);
 276         struct net *net = sock_net(sk);
 277         __be32 v4addr = 0;
 278         unsigned short snum;
 279         int addr_type = 0;
 280         int err = 0;
 281
 282         /* If the socket has its own bind function then use it. */
 283         if (sk->sk_prot->bind)
 284                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 285
 286         if (addr_len < SIN6_LEN_RFC2133)
 287                 return -EINVAL;
 288
 289         if (addr->sin6_family != AF_INET6)
 290                 return -EAFNOSUPPORT;
 291
 292         addr_type = ipv6_addr_type(&addr->sin6_addr);
 293         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 294                 return -EINVAL;
 295
 296         snum = ntohs(addr->sin6_port);
 297         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 298                 return -EACCES;
 299
 300         lock_sock(sk);
 301
 302         /* Check these errors (active socket, double bind). */
 303         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 304                 err = -EINVAL;
 305                 goto out;
 306         }
 307
 308         /* Check if the address belongs to the host. */
 309         if (addr_type == IPV6_ADDR_MAPPED) {
 310                 int chk_addr_ret;
 311
 312                 /* Binding to v4-mapped address on a v6-only socket
 313                  * makes no sense
 314                  */
 315                 if (sk->sk_ipv6only) {
 316                         err = -EINVAL;
 317                         goto out;
 318                 }
 319
 320                 /* Reproduce AF_INET checks to make the bindings consistent */
 321                 v4addr = addr->sin6_addr.s6_addr32[3];
 322                 chk_addr_ret = inet_addr_type(net, v4addr);
 323                 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
 324                     !(inet->freebind || inet->transparent) &&
 325                     v4addr != htonl(INADDR_ANY) &&
 326                     chk_addr_ret != RTN_LOCAL &&
 327                     chk_addr_ret != RTN_MULTICAST &&
 328                     chk_addr_ret != RTN_BROADCAST) {
 329                         err = -EADDRNOTAVAIL;
 330                         goto out;
 331                 }
 332         } else {
 333                 if (addr_type != IPV6_ADDR_ANY) {
 334                         struct net_device *dev = NULL;
 335
 336                         rcu_read_lock();
 337                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 338                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 339                                     addr->sin6_scope_id) {
 340                                         /* Override any existing binding, if another one
 341                                          * is supplied by user.
 342                                          */
 343                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 344                                 }
 345
 346                                 /* Binding to link-local address requires an interface */
 347                                 if (!sk->sk_bound_dev_if) {
 348                                         err = -EINVAL;
 349                                         goto out_unlock;
 350                                 }
 351                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 352                                 if (!dev) {
 353                                         err = -ENODEV;
 354                                         goto out_unlock;
 355                                 }
 356                         }
 357
 358                         /* ipv4 addr of the socket is invalid.  Only the
 359                          * unspecified and mapped address have a v4 equivalent.
 360                          */
 361                         v4addr = LOOPBACK4_IPV6;
 362                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 363                                 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
 364                                     !(inet->freebind || inet->transparent) &&
 365                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 366                                                    dev, 0)) {
 367                                         err = -EADDRNOTAVAIL;
 368                                         goto out_unlock;
 369                                 }
 370                         }
 371                         rcu_read_unlock();
 372                 }
 373         }
 374
 375         inet->inet_rcv_saddr = v4addr;
 376         inet->inet_saddr = v4addr;
 377
 378         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 379
 380         if (!(addr_type & IPV6_ADDR_MULTICAST))
 381                 np->saddr = addr->sin6_addr;
 382
 383         /* Make sure we are allowed to bind here. */
 384         if ((snum || !inet->bind_address_no_port) &&
 385             sk->sk_prot->get_port(sk, snum)) {
 386                 inet_reset_saddr(sk);
 387                 err = -EADDRINUSE;
 388                 goto out;
 389         }
 390
 391         if (addr_type != IPV6_ADDR_ANY) {
 392                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 393                 if (addr_type != IPV6_ADDR_MAPPED)
 394                         sk->sk_ipv6only = 1;
 395         }
 396         if (snum)
 397                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 398         inet->inet_sport = htons(inet->inet_num);
 399         inet->inet_dport = 0;
 400         inet->inet_daddr = 0;
 401 out:
 402         release_sock(sk);
 403         return err;
 404 out_unlock:
 405         rcu_read_unlock();
 406         goto out;
 407 }
 408 EXPORT_SYMBOL(inet6_bind);
 409
 410 int inet6_release(struct socket *sock)
 411 {
 412         struct sock *sk = sock->sk;
 413
 414         if (!sk)
 415                 return -EINVAL;
 416
 417         /* Free mc lists */
 418         ipv6_sock_mc_close(sk);
 419
 420         /* Free ac lists */
 421         ipv6_sock_ac_close(sk);
 422
 423         return inet_release(sock);
 424 }
 425 EXPORT_SYMBOL(inet6_release);
 426
 427 void inet6_destroy_sock(struct sock *sk)
 428 {
 429         struct ipv6_pinfo *np = inet6_sk(sk);
 430         struct sk_buff *skb;
 431         struct ipv6_txoptions *opt;
 432
 433         /* Release rx options */
 434
 435         skb = xchg(&np->pktoptions, NULL);
 436         if (skb)
 437                 kfree_skb(skb);
 438
 439         skb = xchg(&np->rxpmtu, NULL);
 440         if (skb)
 441                 kfree_skb(skb);
 442
 443         /* Free flowlabels */
 444         fl6_free_socklist(sk);
 445
 446         /* Free tx options */
 447
 448         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 449         if (opt) {
 450                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 451                 txopt_put(opt);
 452         }
 453 }
 454 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 455
 456 /*
 457  *      This does both peername and sockname.
 458  */
 459
 460 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 461                  int *uaddr_len, int peer)
 462 {
 463         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 464         struct sock *sk = sock->sk;
 465         struct inet_sock *inet = inet_sk(sk);
 466         struct ipv6_pinfo *np = inet6_sk(sk);
 467
 468         sin->sin6_family = AF_INET6;
 469         sin->sin6_flowinfo = 0;
 470         sin->sin6_scope_id = 0;
 471         if (peer) {
 472                 if (!inet->inet_dport)
 473                         return -ENOTCONN;
 474                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 475                     peer == 1)
 476                         return -ENOTCONN;
 477                 sin->sin6_port = inet->inet_dport;
 478                 sin->sin6_addr = sk->sk_v6_daddr;
 479                 if (np->sndflow)
 480                         sin->sin6_flowinfo = np->flow_label;
 481         } else {
 482                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 483                         sin->sin6_addr = np->saddr;
 484                 else
 485                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 486
 487                 sin->sin6_port = inet->inet_sport;
 488         }
 489         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 490                                                  sk->sk_bound_dev_if);
 491         *uaddr_len = sizeof(*sin);
 492         return 0;
 493 }
 494 EXPORT_SYMBOL(inet6_getname);
 495
 496 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 497 {
 498         struct sock *sk = sock->sk;
 499         struct net *net = sock_net(sk);
 500
 501         switch (cmd) {
 502         case SIOCGSTAMP:
 503                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 504
 505         case SIOCGSTAMPNS:
 506                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 507
 508         case SIOCADDRT:
 509         case SIOCDELRT:
 510
 511                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 512
 513         case SIOCSIFADDR:
 514                 return addrconf_add_ifaddr(net, (void __user *) arg);
 515         case SIOCDIFADDR:
 516                 return addrconf_del_ifaddr(net, (void __user *) arg);
 517         case SIOCSIFDSTADDR:
 518                 return addrconf_set_dstaddr(net, (void __user *) arg);
 519         default:
 520                 if (!sk->sk_prot->ioctl)
 521                         return -ENOIOCTLCMD;
 522                 return sk->sk_prot->ioctl(sk, cmd, arg);
 523         }
 524         /*NOTREACHED*/
 525         return 0;
 526 }
 527 EXPORT_SYMBOL(inet6_ioctl);
 528
 529 const struct proto_ops inet6_stream_ops = {
 530         .family            = PF_INET6,
 531         .owner             = THIS_MODULE,
 532         .release           = inet6_release,
 533         .bind              = inet6_bind,
 534         .connect           = inet_stream_connect,       /* ok           */
 535         .socketpair        = sock_no_socketpair,        /* a do nothing */
 536         .accept            = inet_accept,               /* ok           */
 537         .getname           = inet6_getname,
 538         .poll              = tcp_poll,                  /* ok           */
 539         .ioctl             = inet6_ioctl,               /* must change  */
 540         .listen            = inet_listen,               /* ok           */
 541         .shutdown          = inet_shutdown,             /* ok           */
 542         .setsockopt        = sock_common_setsockopt,    /* ok           */
 543         .getsockopt        = sock_common_getsockopt,    /* ok           */
 544         .sendmsg           = inet_sendmsg,              /* ok           */
 545         .recvmsg           = inet_recvmsg,              /* ok           */
 546         .mmap              = sock_no_mmap,
 547         .sendpage          = inet_sendpage,
 548         .splice_read       = tcp_splice_read,
 549         .read_sock         = tcp_read_sock,
 550         .peek_len          = tcp_peek_len,
 551 #ifdef CONFIG_COMPAT
 552         .compat_setsockopt = compat_sock_common_setsockopt,
 553         .compat_getsockopt = compat_sock_common_getsockopt,
 554 #endif
 555 };
 556
 557 const struct proto_ops inet6_dgram_ops = {
 558         .family            = PF_INET6,
 559         .owner             = THIS_MODULE,
 560         .release           = inet6_release,
 561         .bind              = inet6_bind,
 562         .connect           = inet_dgram_connect,        /* ok           */
 563         .socketpair        = sock_no_socketpair,        /* a do nothing */
 564         .accept            = sock_no_accept,            /* a do nothing */
 565         .getname           = inet6_getname,
 566         .poll              = udp_poll,                  /* ok           */
 567         .ioctl             = inet6_ioctl,               /* must change  */
 568         .listen            = sock_no_listen,            /* ok           */
 569         .shutdown          = inet_shutdown,             /* ok           */
 570         .setsockopt        = sock_common_setsockopt,    /* ok           */
 571         .getsockopt        = sock_common_getsockopt,    /* ok           */
 572         .sendmsg           = inet_sendmsg,              /* ok           */
 573         .recvmsg           = inet_recvmsg,              /* ok           */
 574         .mmap              = sock_no_mmap,
 575         .sendpage          = sock_no_sendpage,
 576         .set_peek_off      = sk_set_peek_off,
 577 #ifdef CONFIG_COMPAT
 578         .compat_setsockopt = compat_sock_common_setsockopt,
 579         .compat_getsockopt = compat_sock_common_getsockopt,
 580 #endif
 581 };
 582
 583 static const struct net_proto_family inet6_family_ops = {
 584         .family = PF_INET6,
 585         .create = inet6_create,
 586         .owner  = THIS_MODULE,
 587 };
 588
 589 int inet6_register_protosw(struct inet_protosw *p)
 590 {
 591         struct list_head *lh;
 592         struct inet_protosw *answer;
 593         struct list_head *last_perm;
 594         int protocol = p->protocol;
 595         int ret;
 596
 597         spin_lock_bh(&inetsw6_lock);
 598
 599         ret = -EINVAL;
 600         if (p->type >= SOCK_MAX)
 601                 goto out_illegal;
 602
 603         /* If we are trying to override a permanent protocol, bail. */
 604         answer = NULL;
 605         ret = -EPERM;
 606         last_perm = &inetsw6[p->type];
 607         list_for_each(lh, &inetsw6[p->type]) {
 608                 answer = list_entry(lh, struct inet_protosw, list);
 609
 610                 /* Check only the non-wild match. */
 611                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 612                         if (protocol == answer->protocol)
 613                                 break;
 614                         last_perm = lh;
 615                 }
 616
 617                 answer = NULL;
 618         }
 619         if (answer)
 620                 goto out_permanent;
 621
 622         /* Add the new entry after the last permanent entry if any, so that
 623          * the new entry does not override a permanent entry when matched with
 624          * a wild-card protocol. But it is allowed to override any existing
 625          * non-permanent entry.  This means that when we remove this entry, the
 626          * system automatically returns to the old behavior.
 627          */
 628         list_add_rcu(&p->list, last_perm);
 629         ret = 0;
 630 out:
 631         spin_unlock_bh(&inetsw6_lock);
 632         return ret;
 633
 634 out_permanent:
 635         pr_err("Attempt to override permanent protocol %d\n", protocol);
 636         goto out;
 637
 638 out_illegal:
 639         pr_err("Ignoring attempt to register invalid socket type %d\n",
 640                p->type);
 641         goto out;
 642 }
 643 EXPORT_SYMBOL(inet6_register_protosw);
 644
 645 void
 646 inet6_unregister_protosw(struct inet_protosw *p)
 647 {
 648         if (INET_PROTOSW_PERMANENT & p->flags) {
 649                 pr_err("Attempt to unregister permanent protocol %d\n",
 650                        p->protocol);
 651         } else {
 652                 spin_lock_bh(&inetsw6_lock);
 653                 list_del_rcu(&p->list);
 654                 spin_unlock_bh(&inetsw6_lock);
 655
 656                 synchronize_net();
 657         }
 658 }
 659 EXPORT_SYMBOL(inet6_unregister_protosw);
 660
 661 int inet6_sk_rebuild_header(struct sock *sk)
 662 {
 663         struct ipv6_pinfo *np = inet6_sk(sk);
 664         struct dst_entry *dst;
 665
 666         dst = __sk_dst_check(sk, np->dst_cookie);
 667
 668         if (!dst) {
 669                 struct inet_sock *inet = inet_sk(sk);
 670                 struct in6_addr *final_p, final;
 671                 struct flowi6 fl6;
 672
 673                 memset(&fl6, 0, sizeof(fl6));
 674                 fl6.flowi6_proto = sk->sk_protocol;
 675                 fl6.daddr = sk->sk_v6_daddr;
 676                 fl6.saddr = np->saddr;
 677                 fl6.flowlabel = np->flow_label;
 678                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 679                 fl6.flowi6_mark = sk->sk_mark;
 680                 fl6.fl6_dport = inet->inet_dport;
 681                 fl6.fl6_sport = inet->inet_sport;
 682                 fl6.flowi6_uid = sk->sk_uid;
 683                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 684
 685                 rcu_read_lock();
 686                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 687                                          &final);
 688                 rcu_read_unlock();
 689
 690                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 691                 if (IS_ERR(dst)) {
 692                         sk->sk_route_caps = 0;
 693                         sk->sk_err_soft = -PTR_ERR(dst);
 694                         return PTR_ERR(dst);
 695                 }
 696
 697                 ip6_dst_store(sk, dst, NULL, NULL);
 698         }
 699
 700         return 0;
 701 }
 702 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 703
 704 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 705                        const struct inet6_skb_parm *opt)
 706 {
 707         const struct ipv6_pinfo *np = inet6_sk(sk);
 708
 709         if (np->rxopt.all) {
 710                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
 711                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 712                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 713                      np->rxopt.bits.rxflow) ||
 714                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 715                      np->rxopt.bits.osrcrt)) ||
 716                     ((opt->dst1 || opt->dst0) &&
 717                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 718                         return true;
 719         }
 720         return false;
 721 }
 722 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 723
 724 static struct packet_type ipv6_packet_type __read_mostly = {
 725         .type = cpu_to_be16(ETH_P_IPV6),
 726         .func = ipv6_rcv,
 727 };
 728
 729 static int __init ipv6_packet_init(void)
 730 {
 731         dev_add_pack(&ipv6_packet_type);
 732         return 0;
 733 }
 734
 735 static void ipv6_packet_cleanup(void)
 736 {
 737         dev_remove_pack(&ipv6_packet_type);
 738 }
 739
 740 static int __net_init ipv6_init_mibs(struct net *net)
 741 {
 742         int i;
 743
 744         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 745         if (!net->mib.udp_stats_in6)
 746                 return -ENOMEM;
 747         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 748         if (!net->mib.udplite_stats_in6)
 749                 goto err_udplite_mib;
 750         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 751         if (!net->mib.ipv6_statistics)
 752                 goto err_ip_mib;
 753
 754         for_each_possible_cpu(i) {
 755                 struct ipstats_mib *af_inet6_stats;
 756                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 757                 u64_stats_init(&af_inet6_stats->syncp);
 758         }
 759
 760
 761         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 762         if (!net->mib.icmpv6_statistics)
 763                 goto err_icmp_mib;
 764         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 765                                                 GFP_KERNEL);
 766         if (!net->mib.icmpv6msg_statistics)
 767                 goto err_icmpmsg_mib;
 768         return 0;
 769
 770 err_icmpmsg_mib:
 771         free_percpu(net->mib.icmpv6_statistics);
 772 err_icmp_mib:
 773         free_percpu(net->mib.ipv6_statistics);
 774 err_ip_mib:
 775         free_percpu(net->mib.udplite_stats_in6);
 776 err_udplite_mib:
 777         free_percpu(net->mib.udp_stats_in6);
 778         return -ENOMEM;
 779 }
 780
 781 static void ipv6_cleanup_mibs(struct net *net)
 782 {
 783         free_percpu(net->mib.udp_stats_in6);
 784         free_percpu(net->mib.udplite_stats_in6);
 785         free_percpu(net->mib.ipv6_statistics);
 786         free_percpu(net->mib.icmpv6_statistics);
 787         kfree(net->mib.icmpv6msg_statistics);
 788 }
 789
 790 static int __net_init inet6_net_init(struct net *net)
 791 {
 792         int err = 0;
 793
 794         net->ipv6.sysctl.bindv6only = 0;
 795         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 796         net->ipv6.sysctl.flowlabel_consistency = 1;
 797         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
 798         net->ipv6.sysctl.idgen_retries = 3;
 799         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 800         net->ipv6.sysctl.flowlabel_state_ranges = 0;
 801         atomic_set(&net->ipv6.fib6_sernum, 1);
 802
 803         err = ipv6_init_mibs(net);
 804         if (err)
 805                 return err;
 806 #ifdef CONFIG_PROC_FS
 807         err = udp6_proc_init(net);
 808         if (err)
 809                 goto out;
 810         err = tcp6_proc_init(net);
 811         if (err)
 812                 goto proc_tcp6_fail;
 813         err = ac6_proc_init(net);
 814         if (err)
 815                 goto proc_ac6_fail;
 816 #endif
 817         return err;
 818
 819 #ifdef CONFIG_PROC_FS
 820 proc_ac6_fail:
 821         tcp6_proc_exit(net);
 822 proc_tcp6_fail:
 823         udp6_proc_exit(net);
 824 out:
 825         ipv6_cleanup_mibs(net);
 826         return err;
 827 #endif
 828 }
 829
 830 static void __net_exit inet6_net_exit(struct net *net)
 831 {
 832 #ifdef CONFIG_PROC_FS
 833         udp6_proc_exit(net);
 834         tcp6_proc_exit(net);
 835         ac6_proc_exit(net);
 836 #endif
 837         ipv6_cleanup_mibs(net);
 838 }
 839
 840 static struct pernet_operations inet6_net_ops = {
 841         .init = inet6_net_init,
 842         .exit = inet6_net_exit,
 843 };
 844
 845 static const struct ipv6_stub ipv6_stub_impl = {
 846         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 847         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 848         .ipv6_dst_lookup = ip6_dst_lookup,
 849         .udpv6_encap_enable = udpv6_encap_enable,
 850         .ndisc_send_na = ndisc_send_na,
 851         .nd_tbl = &nd_tbl,
 852 };
 853
 854 static int __init inet6_init(void)
 855 {
 856         struct list_head *r;
 857         int err = 0;
 858
 859         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 860
 861         /* Register the socket-side information for inet6_create.  */
 862         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 863                 INIT_LIST_HEAD(r);
 864
 865         if (disable_ipv6_mod) {
 866                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 867                 goto out;
 868         }
 869
 870         err = proto_register(&tcpv6_prot, 1);
 871         if (err)
 872                 goto out;
 873
 874         err = proto_register(&udpv6_prot, 1);
 875         if (err)
 876                 goto out_unregister_tcp_proto;
 877
 878         err = proto_register(&udplitev6_prot, 1);
 879         if (err)
 880                 goto out_unregister_udp_proto;
 881
 882         err = proto_register(&rawv6_prot, 1);
 883         if (err)
 884                 goto out_unregister_udplite_proto;
 885
 886         err = proto_register(&pingv6_prot, 1);
 887         if (err)
 888                 goto out_unregister_ping_proto;
 889
 890         /* We MUST register RAW sockets before we create the ICMP6,
 891          * IGMP6, or NDISC control sockets.
 892          */
 893         err = rawv6_init();
 894         if (err)
 895                 goto out_unregister_raw_proto;
 896
 897         /* Register the family here so that the init calls below will
 898          * be able to create sockets. (?? is this dangerous ??)
 899          */
 900         err = sock_register(&inet6_family_ops);
 901         if (err)
 902                 goto out_sock_register_fail;
 903
 904         /*
 905          *      ipngwg API draft makes clear that the correct semantics
 906          *      for TCP and UDP is to consider one TCP and UDP instance
 907          *      in a host available by both INET and INET6 APIs and
 908          *      able to communicate via both network protocols.
 909          */
 910
 911         err = register_pernet_subsys(&inet6_net_ops);
 912         if (err)
 913                 goto register_pernet_fail;
 914         err = icmpv6_init();
 915         if (err)
 916                 goto icmp_fail;
 917         err = ip6_mr_init();
 918         if (err)
 919                 goto ipmr_fail;
 920         err = ndisc_init();
 921         if (err)
 922                 goto ndisc_fail;
 923         err = igmp6_init();
 924         if (err)
 925                 goto igmp_fail;
 926
 927         ipv6_stub = &ipv6_stub_impl;
 928
 929         err = ipv6_netfilter_init();
 930         if (err)
 931                 goto netfilter_fail;
 932         /* Create /proc/foo6 entries. */
 933 #ifdef CONFIG_PROC_FS
 934         err = -ENOMEM;
 935         if (raw6_proc_init())
 936                 goto proc_raw6_fail;
 937         if (udplite6_proc_init())
 938                 goto proc_udplite6_fail;
 939         if (ipv6_misc_proc_init())
 940                 goto proc_misc6_fail;
 941         if (if6_proc_init())
 942                 goto proc_if6_fail;
 943 #endif
 944         err = ip6_route_init();
 945         if (err)
 946                 goto ip6_route_fail;
 947         err = ndisc_late_init();
 948         if (err)
 949                 goto ndisc_late_fail;
 950         err = ip6_flowlabel_init();
 951         if (err)
 952                 goto ip6_flowlabel_fail;
 953         err = addrconf_init();
 954         if (err)
 955                 goto addrconf_fail;
 956
 957         /* Init v6 extension headers. */
 958         err = ipv6_exthdrs_init();
 959         if (err)
 960                 goto ipv6_exthdrs_fail;
 961
 962         err = ipv6_frag_init();
 963         if (err)
 964                 goto ipv6_frag_fail;
 965
 966         /* Init v6 transport protocols. */
 967         err = udpv6_init();
 968         if (err)
 969                 goto udpv6_fail;
 970
 971         err = udplitev6_init();
 972         if (err)
 973                 goto udplitev6_fail;
 974
 975         err = udpv6_offload_init();
 976         if (err)
 977                 goto udpv6_offload_fail;
 978
 979         err = tcpv6_init();
 980         if (err)
 981                 goto tcpv6_fail;
 982
 983         err = ipv6_packet_init();
 984         if (err)
 985                 goto ipv6_packet_fail;
 986
 987         err = pingv6_init();
 988         if (err)
 989                 goto pingv6_fail;
 990
 991         err = calipso_init();
 992         if (err)
 993                 goto calipso_fail;
 994
 995         err = seg6_init();
 996         if (err)
 997                 goto seg6_fail;
 998
 999 #ifdef CONFIG_SYSCTL
1000         err = ipv6_sysctl_register();
1001         if (err)
1002                 goto sysctl_fail;
1003 #endif
1004 out:
1005         return err;
1006
1007 #ifdef CONFIG_SYSCTL
1008 sysctl_fail:
1009         seg6_exit();
1010 #endif
1011 seg6_fail:
1012         calipso_exit();
1013 calipso_fail:
1014         pingv6_exit();
1015 pingv6_fail:
1016         ipv6_packet_cleanup();
1017 ipv6_packet_fail:
1018         tcpv6_exit();
1019 tcpv6_fail:
1020         udpv6_offload_exit();
1021 udpv6_offload_fail:
1022         udplitev6_exit();
1023 udplitev6_fail:
1024         udpv6_exit();
1025 udpv6_fail:
1026         ipv6_frag_exit();
1027 ipv6_frag_fail:
1028         ipv6_exthdrs_exit();
1029 ipv6_exthdrs_fail:
1030         addrconf_cleanup();
1031 addrconf_fail:
1032         ip6_flowlabel_cleanup();
1033 ip6_flowlabel_fail:
1034         ndisc_late_cleanup();
1035 ndisc_late_fail:
1036         ip6_route_cleanup();
1037 ip6_route_fail:
1038 #ifdef CONFIG_PROC_FS
1039         if6_proc_exit();
1040 proc_if6_fail:
1041         ipv6_misc_proc_exit();
1042 proc_misc6_fail:
1043         udplite6_proc_exit();
1044 proc_udplite6_fail:
1045         raw6_proc_exit();
1046 proc_raw6_fail:
1047 #endif
1048         ipv6_netfilter_fini();
1049 netfilter_fail:
1050         igmp6_cleanup();
1051 igmp_fail:
1052         ndisc_cleanup();
1053 ndisc_fail:
1054         ip6_mr_cleanup();
1055 ipmr_fail:
1056         icmpv6_cleanup();
1057 icmp_fail:
1058         unregister_pernet_subsys(&inet6_net_ops);
1059 register_pernet_fail:
1060         sock_unregister(PF_INET6);
1061         rtnl_unregister_all(PF_INET6);
1062 out_sock_register_fail:
1063         rawv6_exit();
1064 out_unregister_ping_proto:
1065         proto_unregister(&pingv6_prot);
1066 out_unregister_raw_proto:
1067         proto_unregister(&rawv6_prot);
1068 out_unregister_udplite_proto:
1069         proto_unregister(&udplitev6_prot);
1070 out_unregister_udp_proto:
1071         proto_unregister(&udpv6_prot);
1072 out_unregister_tcp_proto:
1073         proto_unregister(&tcpv6_prot);
1074         goto out;
1075 }
1076 module_init(inet6_init);
1077
1078 MODULE_ALIAS_NETPROTO(PF_INET6);