net/ipv6/af_inet6.c

   1 /*
   2  *      PF_INET6 socket protocol family
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <[email protected]>
   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 <linux/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->repflow     = net->ipv6.sysctl.flowlabel_reflect;
 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
 262         if (!kern) {
 263                 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
 264                 if (err) {
 265                         sk_common_release(sk);
 266                         goto out;
 267                 }
 268         }
 269 out:
 270         return err;
 271 out_rcu_unlock:
 272         rcu_read_unlock();
 273         goto out;
 274 }
 275
 276
 277 /* bind for INET6 API */
 278 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 279 {
 280         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 281         struct sock *sk = sock->sk;
 282         struct inet_sock *inet = inet_sk(sk);
 283         struct ipv6_pinfo *np = inet6_sk(sk);
 284         struct net *net = sock_net(sk);
 285         __be32 v4addr = 0;
 286         unsigned short snum;
 287         bool saved_ipv6only;
 288         int addr_type = 0;
 289         int err = 0;
 290
 291         /* If the socket has its own bind function then use it. */
 292         if (sk->sk_prot->bind)
 293                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 294
 295         if (addr_len < SIN6_LEN_RFC2133)
 296                 return -EINVAL;
 297
 298         if (addr->sin6_family != AF_INET6)
 299                 return -EAFNOSUPPORT;
 300
 301         addr_type = ipv6_addr_type(&addr->sin6_addr);
 302         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 303                 return -EINVAL;
 304
 305         snum = ntohs(addr->sin6_port);
 306         if (snum && snum < inet_prot_sock(net) &&
 307             !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 308                 return -EACCES;
 309
 310         lock_sock(sk);
 311
 312         /* Check these errors (active socket, double bind). */
 313         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 314                 err = -EINVAL;
 315                 goto out;
 316         }
 317
 318         /* Check if the address belongs to the host. */
 319         if (addr_type == IPV6_ADDR_MAPPED) {
 320                 int chk_addr_ret;
 321
 322                 /* Binding to v4-mapped address on a v6-only socket
 323                  * makes no sense
 324                  */
 325                 if (sk->sk_ipv6only) {
 326                         err = -EINVAL;
 327                         goto out;
 328                 }
 329
 330                 /* Reproduce AF_INET checks to make the bindings consistent */
 331                 v4addr = addr->sin6_addr.s6_addr32[3];
 332                 chk_addr_ret = inet_addr_type(net, v4addr);
 333                 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
 334                     !(inet->freebind || inet->transparent) &&
 335                     v4addr != htonl(INADDR_ANY) &&
 336                     chk_addr_ret != RTN_LOCAL &&
 337                     chk_addr_ret != RTN_MULTICAST &&
 338                     chk_addr_ret != RTN_BROADCAST) {
 339                         err = -EADDRNOTAVAIL;
 340                         goto out;
 341                 }
 342         } else {
 343                 if (addr_type != IPV6_ADDR_ANY) {
 344                         struct net_device *dev = NULL;
 345
 346                         rcu_read_lock();
 347                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 348                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 349                                     addr->sin6_scope_id) {
 350                                         /* Override any existing binding, if another one
 351                                          * is supplied by user.
 352                                          */
 353                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 354                                 }
 355
 356                                 /* Binding to link-local address requires an interface */
 357                                 if (!sk->sk_bound_dev_if) {
 358                                         err = -EINVAL;
 359                                         goto out_unlock;
 360                                 }
 361                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 362                                 if (!dev) {
 363                                         err = -ENODEV;
 364                                         goto out_unlock;
 365                                 }
 366                         }
 367
 368                         /* ipv4 addr of the socket is invalid.  Only the
 369                          * unspecified and mapped address have a v4 equivalent.
 370                          */
 371                         v4addr = LOOPBACK4_IPV6;
 372                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 373                                 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
 374                                     !(inet->freebind || inet->transparent) &&
 375                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 376                                                    dev, 0)) {
 377                                         err = -EADDRNOTAVAIL;
 378                                         goto out_unlock;
 379                                 }
 380                         }
 381                         rcu_read_unlock();
 382                 }
 383         }
 384
 385         inet->inet_rcv_saddr = v4addr;
 386         inet->inet_saddr = v4addr;
 387
 388         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 389
 390         if (!(addr_type & IPV6_ADDR_MULTICAST))
 391                 np->saddr = addr->sin6_addr;
 392
 393         saved_ipv6only = sk->sk_ipv6only;
 394         if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
 395                 sk->sk_ipv6only = 1;
 396
 397         /* Make sure we are allowed to bind here. */
 398         if ((snum || !inet->bind_address_no_port) &&
 399             sk->sk_prot->get_port(sk, snum)) {
 400                 sk->sk_ipv6only = saved_ipv6only;
 401                 inet_reset_saddr(sk);
 402                 err = -EADDRINUSE;
 403                 goto out;
 404         }
 405
 406         if (addr_type != IPV6_ADDR_ANY)
 407                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 408         if (snum)
 409                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 410         inet->inet_sport = htons(inet->inet_num);
 411         inet->inet_dport = 0;
 412         inet->inet_daddr = 0;
 413 out:
 414         release_sock(sk);
 415         return err;
 416 out_unlock:
 417         rcu_read_unlock();
 418         goto out;
 419 }
 420 EXPORT_SYMBOL(inet6_bind);
 421
 422 int inet6_release(struct socket *sock)
 423 {
 424         struct sock *sk = sock->sk;
 425
 426         if (!sk)
 427                 return -EINVAL;
 428
 429         /* Free mc lists */
 430         ipv6_sock_mc_close(sk);
 431
 432         /* Free ac lists */
 433         ipv6_sock_ac_close(sk);
 434
 435         return inet_release(sock);
 436 }
 437 EXPORT_SYMBOL(inet6_release);
 438
 439 void inet6_destroy_sock(struct sock *sk)
 440 {
 441         struct ipv6_pinfo *np = inet6_sk(sk);
 442         struct sk_buff *skb;
 443         struct ipv6_txoptions *opt;
 444
 445         /* Release rx options */
 446
 447         skb = xchg(&np->pktoptions, NULL);
 448         if (skb)
 449                 kfree_skb(skb);
 450
 451         skb = xchg(&np->rxpmtu, NULL);
 452         if (skb)
 453                 kfree_skb(skb);
 454
 455         /* Free flowlabels */
 456         fl6_free_socklist(sk);
 457
 458         /* Free tx options */
 459
 460         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 461         if (opt) {
 462                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 463                 txopt_put(opt);
 464         }
 465 }
 466 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 467
 468 /*
 469  *      This does both peername and sockname.
 470  */
 471
 472 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 473                  int peer)
 474 {
 475         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 476         struct sock *sk = sock->sk;
 477         struct inet_sock *inet = inet_sk(sk);
 478         struct ipv6_pinfo *np = inet6_sk(sk);
 479
 480         sin->sin6_family = AF_INET6;
 481         sin->sin6_flowinfo = 0;
 482         sin->sin6_scope_id = 0;
 483         if (peer) {
 484                 if (!inet->inet_dport)
 485                         return -ENOTCONN;
 486                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 487                     peer == 1)
 488                         return -ENOTCONN;
 489                 sin->sin6_port = inet->inet_dport;
 490                 sin->sin6_addr = sk->sk_v6_daddr;
 491                 if (np->sndflow)
 492                         sin->sin6_flowinfo = np->flow_label;
 493         } else {
 494                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 495                         sin->sin6_addr = np->saddr;
 496                 else
 497                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 498
 499                 sin->sin6_port = inet->inet_sport;
 500         }
 501         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 502                                                  sk->sk_bound_dev_if);
 503         return sizeof(*sin);
 504 }
 505 EXPORT_SYMBOL(inet6_getname);
 506
 507 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 508 {
 509         struct sock *sk = sock->sk;
 510         struct net *net = sock_net(sk);
 511
 512         switch (cmd) {
 513         case SIOCGSTAMP:
 514                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 515
 516         case SIOCGSTAMPNS:
 517                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 518
 519         case SIOCADDRT:
 520         case SIOCDELRT:
 521
 522                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 523
 524         case SIOCSIFADDR:
 525                 return addrconf_add_ifaddr(net, (void __user *) arg);
 526         case SIOCDIFADDR:
 527                 return addrconf_del_ifaddr(net, (void __user *) arg);
 528         case SIOCSIFDSTADDR:
 529                 return addrconf_set_dstaddr(net, (void __user *) arg);
 530         default:
 531                 if (!sk->sk_prot->ioctl)
 532                         return -ENOIOCTLCMD;
 533                 return sk->sk_prot->ioctl(sk, cmd, arg);
 534         }
 535         /*NOTREACHED*/
 536         return 0;
 537 }
 538 EXPORT_SYMBOL(inet6_ioctl);
 539
 540 const struct proto_ops inet6_stream_ops = {
 541         .family            = PF_INET6,
 542         .owner             = THIS_MODULE,
 543         .release           = inet6_release,
 544         .bind              = inet6_bind,
 545         .connect           = inet_stream_connect,       /* ok           */
 546         .socketpair        = sock_no_socketpair,        /* a do nothing */
 547         .accept            = inet_accept,               /* ok           */
 548         .getname           = inet6_getname,
 549         .poll              = tcp_poll,                  /* ok           */
 550         .ioctl             = inet6_ioctl,               /* must change  */
 551         .listen            = inet_listen,               /* ok           */
 552         .shutdown          = inet_shutdown,             /* ok           */
 553         .setsockopt        = sock_common_setsockopt,    /* ok           */
 554         .getsockopt        = sock_common_getsockopt,    /* ok           */
 555         .sendmsg           = inet_sendmsg,              /* ok           */
 556         .recvmsg           = inet_recvmsg,              /* ok           */
 557         .mmap              = sock_no_mmap,
 558         .sendpage          = inet_sendpage,
 559         .sendmsg_locked    = tcp_sendmsg_locked,
 560         .sendpage_locked   = tcp_sendpage_locked,
 561         .splice_read       = tcp_splice_read,
 562         .read_sock         = tcp_read_sock,
 563         .peek_len          = tcp_peek_len,
 564 #ifdef CONFIG_COMPAT
 565         .compat_setsockopt = compat_sock_common_setsockopt,
 566         .compat_getsockopt = compat_sock_common_getsockopt,
 567 #endif
 568 };
 569
 570 const struct proto_ops inet6_dgram_ops = {
 571         .family            = PF_INET6,
 572         .owner             = THIS_MODULE,
 573         .release           = inet6_release,
 574         .bind              = inet6_bind,
 575         .connect           = inet_dgram_connect,        /* ok           */
 576         .socketpair        = sock_no_socketpair,        /* a do nothing */
 577         .accept            = sock_no_accept,            /* a do nothing */
 578         .getname           = inet6_getname,
 579         .poll              = udp_poll,                  /* ok           */
 580         .ioctl             = inet6_ioctl,               /* must change  */
 581         .listen            = sock_no_listen,            /* ok           */
 582         .shutdown          = inet_shutdown,             /* ok           */
 583         .setsockopt        = sock_common_setsockopt,    /* ok           */
 584         .getsockopt        = sock_common_getsockopt,    /* ok           */
 585         .sendmsg           = inet_sendmsg,              /* ok           */
 586         .recvmsg           = inet_recvmsg,              /* ok           */
 587         .mmap              = sock_no_mmap,
 588         .sendpage          = sock_no_sendpage,
 589         .set_peek_off      = sk_set_peek_off,
 590 #ifdef CONFIG_COMPAT
 591         .compat_setsockopt = compat_sock_common_setsockopt,
 592         .compat_getsockopt = compat_sock_common_getsockopt,
 593 #endif
 594 };
 595
 596 static const struct net_proto_family inet6_family_ops = {
 597         .family = PF_INET6,
 598         .create = inet6_create,
 599         .owner  = THIS_MODULE,
 600 };
 601
 602 int inet6_register_protosw(struct inet_protosw *p)
 603 {
 604         struct list_head *lh;
 605         struct inet_protosw *answer;
 606         struct list_head *last_perm;
 607         int protocol = p->protocol;
 608         int ret;
 609
 610         spin_lock_bh(&inetsw6_lock);
 611
 612         ret = -EINVAL;
 613         if (p->type >= SOCK_MAX)
 614                 goto out_illegal;
 615
 616         /* If we are trying to override a permanent protocol, bail. */
 617         answer = NULL;
 618         ret = -EPERM;
 619         last_perm = &inetsw6[p->type];
 620         list_for_each(lh, &inetsw6[p->type]) {
 621                 answer = list_entry(lh, struct inet_protosw, list);
 622
 623                 /* Check only the non-wild match. */
 624                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 625                         if (protocol == answer->protocol)
 626                                 break;
 627                         last_perm = lh;
 628                 }
 629
 630                 answer = NULL;
 631         }
 632         if (answer)
 633                 goto out_permanent;
 634
 635         /* Add the new entry after the last permanent entry if any, so that
 636          * the new entry does not override a permanent entry when matched with
 637          * a wild-card protocol. But it is allowed to override any existing
 638          * non-permanent entry.  This means that when we remove this entry, the
 639          * system automatically returns to the old behavior.
 640          */
 641         list_add_rcu(&p->list, last_perm);
 642         ret = 0;
 643 out:
 644         spin_unlock_bh(&inetsw6_lock);
 645         return ret;
 646
 647 out_permanent:
 648         pr_err("Attempt to override permanent protocol %d\n", protocol);
 649         goto out;
 650
 651 out_illegal:
 652         pr_err("Ignoring attempt to register invalid socket type %d\n",
 653                p->type);
 654         goto out;
 655 }
 656 EXPORT_SYMBOL(inet6_register_protosw);
 657
 658 void
 659 inet6_unregister_protosw(struct inet_protosw *p)
 660 {
 661         if (INET_PROTOSW_PERMANENT & p->flags) {
 662                 pr_err("Attempt to unregister permanent protocol %d\n",
 663                        p->protocol);
 664         } else {
 665                 spin_lock_bh(&inetsw6_lock);
 666                 list_del_rcu(&p->list);
 667                 spin_unlock_bh(&inetsw6_lock);
 668
 669                 synchronize_net();
 670         }
 671 }
 672 EXPORT_SYMBOL(inet6_unregister_protosw);
 673
 674 int inet6_sk_rebuild_header(struct sock *sk)
 675 {
 676         struct ipv6_pinfo *np = inet6_sk(sk);
 677         struct dst_entry *dst;
 678
 679         dst = __sk_dst_check(sk, np->dst_cookie);
 680
 681         if (!dst) {
 682                 struct inet_sock *inet = inet_sk(sk);
 683                 struct in6_addr *final_p, final;
 684                 struct flowi6 fl6;
 685
 686                 memset(&fl6, 0, sizeof(fl6));
 687                 fl6.flowi6_proto = sk->sk_protocol;
 688                 fl6.daddr = sk->sk_v6_daddr;
 689                 fl6.saddr = np->saddr;
 690                 fl6.flowlabel = np->flow_label;
 691                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 692                 fl6.flowi6_mark = sk->sk_mark;
 693                 fl6.fl6_dport = inet->inet_dport;
 694                 fl6.fl6_sport = inet->inet_sport;
 695                 fl6.flowi6_uid = sk->sk_uid;
 696                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 697
 698                 rcu_read_lock();
 699                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 700                                          &final);
 701                 rcu_read_unlock();
 702
 703                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 704                 if (IS_ERR(dst)) {
 705                         sk->sk_route_caps = 0;
 706                         sk->sk_err_soft = -PTR_ERR(dst);
 707                         return PTR_ERR(dst);
 708                 }
 709
 710                 ip6_dst_store(sk, dst, NULL, NULL);
 711         }
 712
 713         return 0;
 714 }
 715 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 716
 717 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 718                        const struct inet6_skb_parm *opt)
 719 {
 720         const struct ipv6_pinfo *np = inet6_sk(sk);
 721
 722         if (np->rxopt.all) {
 723                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
 724                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 725                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 726                      np->rxopt.bits.rxflow) ||
 727                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 728                      np->rxopt.bits.osrcrt)) ||
 729                     ((opt->dst1 || opt->dst0) &&
 730                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 731                         return true;
 732         }
 733         return false;
 734 }
 735 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 736
 737 static struct packet_type ipv6_packet_type __read_mostly = {
 738         .type = cpu_to_be16(ETH_P_IPV6),
 739         .func = ipv6_rcv,
 740 };
 741
 742 static int __init ipv6_packet_init(void)
 743 {
 744         dev_add_pack(&ipv6_packet_type);
 745         return 0;
 746 }
 747
 748 static void ipv6_packet_cleanup(void)
 749 {
 750         dev_remove_pack(&ipv6_packet_type);
 751 }
 752
 753 static int __net_init ipv6_init_mibs(struct net *net)
 754 {
 755         int i;
 756
 757         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 758         if (!net->mib.udp_stats_in6)
 759                 return -ENOMEM;
 760         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 761         if (!net->mib.udplite_stats_in6)
 762                 goto err_udplite_mib;
 763         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 764         if (!net->mib.ipv6_statistics)
 765                 goto err_ip_mib;
 766
 767         for_each_possible_cpu(i) {
 768                 struct ipstats_mib *af_inet6_stats;
 769                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 770                 u64_stats_init(&af_inet6_stats->syncp);
 771         }
 772
 773
 774         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 775         if (!net->mib.icmpv6_statistics)
 776                 goto err_icmp_mib;
 777         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 778                                                 GFP_KERNEL);
 779         if (!net->mib.icmpv6msg_statistics)
 780                 goto err_icmpmsg_mib;
 781         return 0;
 782
 783 err_icmpmsg_mib:
 784         free_percpu(net->mib.icmpv6_statistics);
 785 err_icmp_mib:
 786         free_percpu(net->mib.ipv6_statistics);
 787 err_ip_mib:
 788         free_percpu(net->mib.udplite_stats_in6);
 789 err_udplite_mib:
 790         free_percpu(net->mib.udp_stats_in6);
 791         return -ENOMEM;
 792 }
 793
 794 static void ipv6_cleanup_mibs(struct net *net)
 795 {
 796         free_percpu(net->mib.udp_stats_in6);
 797         free_percpu(net->mib.udplite_stats_in6);
 798         free_percpu(net->mib.ipv6_statistics);
 799         free_percpu(net->mib.icmpv6_statistics);
 800         kfree(net->mib.icmpv6msg_statistics);
 801 }
 802
 803 static int __net_init inet6_net_init(struct net *net)
 804 {
 805         int err = 0;
 806
 807         net->ipv6.sysctl.bindv6only = 0;
 808         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 809         net->ipv6.sysctl.flowlabel_consistency = 1;
 810         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
 811         net->ipv6.sysctl.idgen_retries = 3;
 812         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 813         net->ipv6.sysctl.flowlabel_state_ranges = 0;
 814         net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
 815         net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
 816         net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
 817         net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
 818         atomic_set(&net->ipv6.fib6_sernum, 1);
 819
 820         err = ipv6_init_mibs(net);
 821         if (err)
 822                 return err;
 823 #ifdef CONFIG_PROC_FS
 824         err = udp6_proc_init(net);
 825         if (err)
 826                 goto out;
 827         err = tcp6_proc_init(net);
 828         if (err)
 829                 goto proc_tcp6_fail;
 830         err = ac6_proc_init(net);
 831         if (err)
 832                 goto proc_ac6_fail;
 833 #endif
 834         return err;
 835
 836 #ifdef CONFIG_PROC_FS
 837 proc_ac6_fail:
 838         tcp6_proc_exit(net);
 839 proc_tcp6_fail:
 840         udp6_proc_exit(net);
 841 out:
 842         ipv6_cleanup_mibs(net);
 843         return err;
 844 #endif
 845 }
 846
 847 static void __net_exit inet6_net_exit(struct net *net)
 848 {
 849 #ifdef CONFIG_PROC_FS
 850         udp6_proc_exit(net);
 851         tcp6_proc_exit(net);
 852         ac6_proc_exit(net);
 853 #endif
 854         ipv6_cleanup_mibs(net);
 855 }
 856
 857 static struct pernet_operations inet6_net_ops = {
 858         .init = inet6_net_init,
 859         .exit = inet6_net_exit,
 860 };
 861
 862 static const struct ipv6_stub ipv6_stub_impl = {
 863         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 864         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 865         .ipv6_dst_lookup = ip6_dst_lookup,
 866         .udpv6_encap_enable = udpv6_encap_enable,
 867         .ndisc_send_na = ndisc_send_na,
 868         .nd_tbl = &nd_tbl,
 869 };
 870
 871 static int __init inet6_init(void)
 872 {
 873         struct list_head *r;
 874         int err = 0;
 875
 876         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 877
 878         /* Register the socket-side information for inet6_create.  */
 879         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 880                 INIT_LIST_HEAD(r);
 881
 882         if (disable_ipv6_mod) {
 883                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 884                 goto out;
 885         }
 886
 887         err = proto_register(&tcpv6_prot, 1);
 888         if (err)
 889                 goto out;
 890
 891         err = proto_register(&udpv6_prot, 1);
 892         if (err)
 893                 goto out_unregister_tcp_proto;
 894
 895         err = proto_register(&udplitev6_prot, 1);
 896         if (err)
 897                 goto out_unregister_udp_proto;
 898
 899         err = proto_register(&rawv6_prot, 1);
 900         if (err)
 901                 goto out_unregister_udplite_proto;
 902
 903         err = proto_register(&pingv6_prot, 1);
 904         if (err)
 905                 goto out_unregister_ping_proto;
 906
 907         /* We MUST register RAW sockets before we create the ICMP6,
 908          * IGMP6, or NDISC control sockets.
 909          */
 910         err = rawv6_init();
 911         if (err)
 912                 goto out_unregister_raw_proto;
 913
 914         /* Register the family here so that the init calls below will
 915          * be able to create sockets. (?? is this dangerous ??)
 916          */
 917         err = sock_register(&inet6_family_ops);
 918         if (err)
 919                 goto out_sock_register_fail;
 920
 921         /*
 922          *      ipngwg API draft makes clear that the correct semantics
 923          *      for TCP and UDP is to consider one TCP and UDP instance
 924          *      in a host available by both INET and INET6 APIs and
 925          *      able to communicate via both network protocols.
 926          */
 927
 928         err = register_pernet_subsys(&inet6_net_ops);
 929         if (err)
 930                 goto register_pernet_fail;
 931         err = ip6_mr_init();
 932         if (err)
 933                 goto ipmr_fail;
 934         err = icmpv6_init();
 935         if (err)
 936                 goto icmp_fail;
 937         err = ndisc_init();
 938         if (err)
 939                 goto ndisc_fail;
 940         err = igmp6_init();
 941         if (err)
 942                 goto igmp_fail;
 943
 944         err = ipv6_netfilter_init();
 945         if (err)
 946                 goto netfilter_fail;
 947         /* Create /proc/foo6 entries. */
 948 #ifdef CONFIG_PROC_FS
 949         err = -ENOMEM;
 950         if (raw6_proc_init())
 951                 goto proc_raw6_fail;
 952         if (udplite6_proc_init())
 953                 goto proc_udplite6_fail;
 954         if (ipv6_misc_proc_init())
 955                 goto proc_misc6_fail;
 956         if (if6_proc_init())
 957                 goto proc_if6_fail;
 958 #endif
 959         err = ip6_route_init();
 960         if (err)
 961                 goto ip6_route_fail;
 962         err = ndisc_late_init();
 963         if (err)
 964                 goto ndisc_late_fail;
 965         err = ip6_flowlabel_init();
 966         if (err)
 967                 goto ip6_flowlabel_fail;
 968         err = addrconf_init();
 969         if (err)
 970                 goto addrconf_fail;
 971
 972         /* Init v6 extension headers. */
 973         err = ipv6_exthdrs_init();
 974         if (err)
 975                 goto ipv6_exthdrs_fail;
 976
 977         err = ipv6_frag_init();
 978         if (err)
 979                 goto ipv6_frag_fail;
 980
 981         /* Init v6 transport protocols. */
 982         err = udpv6_init();
 983         if (err)
 984                 goto udpv6_fail;
 985
 986         err = udplitev6_init();
 987         if (err)
 988                 goto udplitev6_fail;
 989
 990         err = udpv6_offload_init();
 991         if (err)
 992                 goto udpv6_offload_fail;
 993
 994         err = tcpv6_init();
 995         if (err)
 996                 goto tcpv6_fail;
 997
 998         err = ipv6_packet_init();
 999         if (err)
1000                 goto ipv6_packet_fail;
1001
1002         err = pingv6_init();
1003         if (err)
1004                 goto pingv6_fail;
1005
1006         err = calipso_init();
1007         if (err)
1008                 goto calipso_fail;
1009
1010         err = seg6_init();
1011         if (err)
1012                 goto seg6_fail;
1013
1014         err = igmp6_late_init();
1015         if (err)
1016                 goto igmp6_late_err;
1017
1018 #ifdef CONFIG_SYSCTL
1019         err = ipv6_sysctl_register();
1020         if (err)
1021                 goto sysctl_fail;
1022 #endif
1023
1024         /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1025         wmb();
1026         ipv6_stub = &ipv6_stub_impl;
1027 out:
1028         return err;
1029
1030 #ifdef CONFIG_SYSCTL
1031 sysctl_fail:
1032         igmp6_late_cleanup();
1033 #endif
1034 igmp6_late_err:
1035         seg6_exit();
1036 seg6_fail:
1037         calipso_exit();
1038 calipso_fail:
1039         pingv6_exit();
1040 pingv6_fail:
1041         ipv6_packet_cleanup();
1042 ipv6_packet_fail:
1043         tcpv6_exit();
1044 tcpv6_fail:
1045         udpv6_offload_exit();
1046 udpv6_offload_fail:
1047         udplitev6_exit();
1048 udplitev6_fail:
1049         udpv6_exit();
1050 udpv6_fail:
1051         ipv6_frag_exit();
1052 ipv6_frag_fail:
1053         ipv6_exthdrs_exit();
1054 ipv6_exthdrs_fail:
1055         addrconf_cleanup();
1056 addrconf_fail:
1057         ip6_flowlabel_cleanup();
1058 ip6_flowlabel_fail:
1059         ndisc_late_cleanup();
1060 ndisc_late_fail:
1061         ip6_route_cleanup();
1062 ip6_route_fail:
1063 #ifdef CONFIG_PROC_FS
1064         if6_proc_exit();
1065 proc_if6_fail:
1066         ipv6_misc_proc_exit();
1067 proc_misc6_fail:
1068         udplite6_proc_exit();
1069 proc_udplite6_fail:
1070         raw6_proc_exit();
1071 proc_raw6_fail:
1072 #endif
1073         ipv6_netfilter_fini();
1074 netfilter_fail:
1075         igmp6_cleanup();
1076 igmp_fail:
1077         ndisc_cleanup();
1078 ndisc_fail:
1079         ip6_mr_cleanup();
1080 icmp_fail:
1081         unregister_pernet_subsys(&inet6_net_ops);
1082 ipmr_fail:
1083         icmpv6_cleanup();
1084 register_pernet_fail:
1085         sock_unregister(PF_INET6);
1086         rtnl_unregister_all(PF_INET6);
1087 out_sock_register_fail:
1088         rawv6_exit();
1089 out_unregister_ping_proto:
1090         proto_unregister(&pingv6_prot);
1091 out_unregister_raw_proto:
1092         proto_unregister(&rawv6_prot);
1093 out_unregister_udplite_proto:
1094         proto_unregister(&udplitev6_prot);
1095 out_unregister_udp_proto:
1096         proto_unregister(&udpv6_prot);
1097 out_unregister_tcp_proto:
1098         proto_unregister(&tcpv6_prot);
1099         goto out;
1100 }
1101 module_init(inet6_init);
1102
1103 MODULE_ALIAS_NETPROTO(PF_INET6);