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 <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 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
 277                         bool force_bind_address_no_port, bool with_lock)
 278 {
 279         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 280         struct inet_sock *inet = inet_sk(sk);
 281         struct ipv6_pinfo *np = inet6_sk(sk);
 282         struct net *net = sock_net(sk);
 283         __be32 v4addr = 0;
 284         unsigned short snum;
 285         bool saved_ipv6only;
 286         int addr_type = 0;
 287         int err = 0;
 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) && sk->sk_type == SOCK_STREAM)
 294                 return -EINVAL;
 295
 296         snum = ntohs(addr->sin6_port);
 297         if (snum && snum < inet_prot_sock(net) &&
 298             !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 299                 return -EACCES;
 300
 301         if (with_lock)
 302                 lock_sock(sk);
 303
 304         /* Check these errors (active socket, double bind). */
 305         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 306                 err = -EINVAL;
 307                 goto out;
 308         }
 309
 310         /* Check if the address belongs to the host. */
 311         if (addr_type == IPV6_ADDR_MAPPED) {
 312                 struct net_device *dev = NULL;
 313                 int chk_addr_ret;
 314
 315                 /* Binding to v4-mapped address on a v6-only socket
 316                  * makes no sense
 317                  */
 318                 if (sk->sk_ipv6only) {
 319                         err = -EINVAL;
 320                         goto out;
 321                 }
 322
 323                 rcu_read_lock();
 324                 if (sk->sk_bound_dev_if) {
 325                         dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 326                         if (!dev) {
 327                                 err = -ENODEV;
 328                                 goto out_unlock;
 329                         }
 330                 }
 331
 332                 /* Reproduce AF_INET checks to make the bindings consistent */
 333                 v4addr = addr->sin6_addr.s6_addr32[3];
 334                 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
 335                 rcu_read_unlock();
 336
 337                 if (!inet_can_nonlocal_bind(net, inet) &&
 338                     v4addr != htonl(INADDR_ANY) &&
 339                     chk_addr_ret != RTN_LOCAL &&
 340                     chk_addr_ret != RTN_MULTICAST &&
 341                     chk_addr_ret != RTN_BROADCAST) {
 342                         err = -EADDRNOTAVAIL;
 343                         goto out;
 344                 }
 345         } else {
 346                 if (addr_type != IPV6_ADDR_ANY) {
 347                         struct net_device *dev = NULL;
 348
 349                         rcu_read_lock();
 350                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 351                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 352                                     addr->sin6_scope_id) {
 353                                         /* Override any existing binding, if another one
 354                                          * is supplied by user.
 355                                          */
 356                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 357                                 }
 358
 359                                 /* Binding to link-local address requires an interface */
 360                                 if (!sk->sk_bound_dev_if) {
 361                                         err = -EINVAL;
 362                                         goto out_unlock;
 363                                 }
 364                         }
 365
 366                         if (sk->sk_bound_dev_if) {
 367                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 368                                 if (!dev) {
 369                                         err = -ENODEV;
 370                                         goto out_unlock;
 371                                 }
 372                         }
 373
 374                         /* ipv4 addr of the socket is invalid.  Only the
 375                          * unspecified and mapped address have a v4 equivalent.
 376                          */
 377                         v4addr = LOOPBACK4_IPV6;
 378                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 379                                 if (!ipv6_can_nonlocal_bind(net, inet) &&
 380                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 381                                                    dev, 0)) {
 382                                         err = -EADDRNOTAVAIL;
 383                                         goto out_unlock;
 384                                 }
 385                         }
 386                         rcu_read_unlock();
 387                 }
 388         }
 389
 390         inet->inet_rcv_saddr = v4addr;
 391         inet->inet_saddr = v4addr;
 392
 393         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 394
 395         if (!(addr_type & IPV6_ADDR_MULTICAST))
 396                 np->saddr = addr->sin6_addr;
 397
 398         saved_ipv6only = sk->sk_ipv6only;
 399         if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
 400                 sk->sk_ipv6only = 1;
 401
 402         /* Make sure we are allowed to bind here. */
 403         if (snum || !(inet->bind_address_no_port ||
 404                       force_bind_address_no_port)) {
 405                 if (sk->sk_prot->get_port(sk, snum)) {
 406                         sk->sk_ipv6only = saved_ipv6only;
 407                         inet_reset_saddr(sk);
 408                         err = -EADDRINUSE;
 409                         goto out;
 410                 }
 411                 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
 412                 if (err) {
 413                         sk->sk_ipv6only = saved_ipv6only;
 414                         inet_reset_saddr(sk);
 415                         goto out;
 416                 }
 417         }
 418
 419         if (addr_type != IPV6_ADDR_ANY)
 420                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 421         if (snum)
 422                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 423         inet->inet_sport = htons(inet->inet_num);
 424         inet->inet_dport = 0;
 425         inet->inet_daddr = 0;
 426 out:
 427         if (with_lock)
 428                 release_sock(sk);
 429         return err;
 430 out_unlock:
 431         rcu_read_unlock();
 432         goto out;
 433 }
 434
 435 /* bind for INET6 API */
 436 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 437 {
 438         struct sock *sk = sock->sk;
 439         int err = 0;
 440
 441         /* If the socket has its own bind function then use it. */
 442         if (sk->sk_prot->bind)
 443                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 444
 445         if (addr_len < SIN6_LEN_RFC2133)
 446                 return -EINVAL;
 447
 448         /* BPF prog is run before any checks are done so that if the prog
 449          * changes context in a wrong way it will be caught.
 450          */
 451         err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
 452         if (err)
 453                 return err;
 454
 455         return __inet6_bind(sk, uaddr, addr_len, false, true);
 456 }
 457 EXPORT_SYMBOL(inet6_bind);
 458
 459 int inet6_release(struct socket *sock)
 460 {
 461         struct sock *sk = sock->sk;
 462
 463         if (!sk)
 464                 return -EINVAL;
 465
 466         /* Free mc lists */
 467         ipv6_sock_mc_close(sk);
 468
 469         /* Free ac lists */
 470         ipv6_sock_ac_close(sk);
 471
 472         return inet_release(sock);
 473 }
 474 EXPORT_SYMBOL(inet6_release);
 475
 476 void inet6_destroy_sock(struct sock *sk)
 477 {
 478         struct ipv6_pinfo *np = inet6_sk(sk);
 479         struct sk_buff *skb;
 480         struct ipv6_txoptions *opt;
 481
 482         /* Release rx options */
 483
 484         skb = xchg(&np->pktoptions, NULL);
 485         if (skb)
 486                 kfree_skb(skb);
 487
 488         skb = xchg(&np->rxpmtu, NULL);
 489         if (skb)
 490                 kfree_skb(skb);
 491
 492         /* Free flowlabels */
 493         fl6_free_socklist(sk);
 494
 495         /* Free tx options */
 496
 497         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 498         if (opt) {
 499                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 500                 txopt_put(opt);
 501         }
 502 }
 503 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 504
 505 /*
 506  *      This does both peername and sockname.
 507  */
 508
 509 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 510                  int peer)
 511 {
 512         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 513         struct sock *sk = sock->sk;
 514         struct inet_sock *inet = inet_sk(sk);
 515         struct ipv6_pinfo *np = inet6_sk(sk);
 516
 517         sin->sin6_family = AF_INET6;
 518         sin->sin6_flowinfo = 0;
 519         sin->sin6_scope_id = 0;
 520         if (peer) {
 521                 if (!inet->inet_dport)
 522                         return -ENOTCONN;
 523                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 524                     peer == 1)
 525                         return -ENOTCONN;
 526                 sin->sin6_port = inet->inet_dport;
 527                 sin->sin6_addr = sk->sk_v6_daddr;
 528                 if (np->sndflow)
 529                         sin->sin6_flowinfo = np->flow_label;
 530         } else {
 531                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 532                         sin->sin6_addr = np->saddr;
 533                 else
 534                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 535
 536                 sin->sin6_port = inet->inet_sport;
 537         }
 538         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 539                                                  sk->sk_bound_dev_if);
 540         return sizeof(*sin);
 541 }
 542 EXPORT_SYMBOL(inet6_getname);
 543
 544 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 545 {
 546         struct sock *sk = sock->sk;
 547         struct net *net = sock_net(sk);
 548
 549         switch (cmd) {
 550         case SIOCGSTAMP:
 551                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 552
 553         case SIOCGSTAMPNS:
 554                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 555
 556         case SIOCADDRT:
 557         case SIOCDELRT:
 558
 559                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 560
 561         case SIOCSIFADDR:
 562                 return addrconf_add_ifaddr(net, (void __user *) arg);
 563         case SIOCDIFADDR:
 564                 return addrconf_del_ifaddr(net, (void __user *) arg);
 565         case SIOCSIFDSTADDR:
 566                 return addrconf_set_dstaddr(net, (void __user *) arg);
 567         default:
 568                 if (!sk->sk_prot->ioctl)
 569                         return -ENOIOCTLCMD;
 570                 return sk->sk_prot->ioctl(sk, cmd, arg);
 571         }
 572         /*NOTREACHED*/
 573         return 0;
 574 }
 575 EXPORT_SYMBOL(inet6_ioctl);
 576
 577 const struct proto_ops inet6_stream_ops = {
 578         .family            = PF_INET6,
 579         .owner             = THIS_MODULE,
 580         .release           = inet6_release,
 581         .bind              = inet6_bind,
 582         .connect           = inet_stream_connect,       /* ok           */
 583         .socketpair        = sock_no_socketpair,        /* a do nothing */
 584         .accept            = inet_accept,               /* ok           */
 585         .getname           = inet6_getname,
 586         .poll              = tcp_poll,                  /* ok           */
 587         .ioctl             = inet6_ioctl,               /* must change  */
 588         .listen            = inet_listen,               /* ok           */
 589         .shutdown          = inet_shutdown,             /* ok           */
 590         .setsockopt        = sock_common_setsockopt,    /* ok           */
 591         .getsockopt        = sock_common_getsockopt,    /* ok           */
 592         .sendmsg           = inet_sendmsg,              /* ok           */
 593         .recvmsg           = inet_recvmsg,              /* ok           */
 594 #ifdef CONFIG_MMU
 595         .mmap              = tcp_mmap,
 596 #endif
 597         .sendpage          = inet_sendpage,
 598         .sendmsg_locked    = tcp_sendmsg_locked,
 599         .sendpage_locked   = tcp_sendpage_locked,
 600         .splice_read       = tcp_splice_read,
 601         .read_sock         = tcp_read_sock,
 602         .peek_len          = tcp_peek_len,
 603 #ifdef CONFIG_COMPAT
 604         .compat_setsockopt = compat_sock_common_setsockopt,
 605         .compat_getsockopt = compat_sock_common_getsockopt,
 606 #endif
 607         .set_rcvlowat      = tcp_set_rcvlowat,
 608 };
 609
 610 const struct proto_ops inet6_dgram_ops = {
 611         .family            = PF_INET6,
 612         .owner             = THIS_MODULE,
 613         .release           = inet6_release,
 614         .bind              = inet6_bind,
 615         .connect           = inet_dgram_connect,        /* ok           */
 616         .socketpair        = sock_no_socketpair,        /* a do nothing */
 617         .accept            = sock_no_accept,            /* a do nothing */
 618         .getname           = inet6_getname,
 619         .poll              = udp_poll,                  /* ok           */
 620         .ioctl             = inet6_ioctl,               /* must change  */
 621         .listen            = sock_no_listen,            /* ok           */
 622         .shutdown          = inet_shutdown,             /* ok           */
 623         .setsockopt        = sock_common_setsockopt,    /* ok           */
 624         .getsockopt        = sock_common_getsockopt,    /* ok           */
 625         .sendmsg           = inet_sendmsg,              /* ok           */
 626         .recvmsg           = inet_recvmsg,              /* ok           */
 627         .mmap              = sock_no_mmap,
 628         .sendpage          = sock_no_sendpage,
 629         .set_peek_off      = sk_set_peek_off,
 630 #ifdef CONFIG_COMPAT
 631         .compat_setsockopt = compat_sock_common_setsockopt,
 632         .compat_getsockopt = compat_sock_common_getsockopt,
 633 #endif
 634 };
 635
 636 static const struct net_proto_family inet6_family_ops = {
 637         .family = PF_INET6,
 638         .create = inet6_create,
 639         .owner  = THIS_MODULE,
 640 };
 641
 642 int inet6_register_protosw(struct inet_protosw *p)
 643 {
 644         struct list_head *lh;
 645         struct inet_protosw *answer;
 646         struct list_head *last_perm;
 647         int protocol = p->protocol;
 648         int ret;
 649
 650         spin_lock_bh(&inetsw6_lock);
 651
 652         ret = -EINVAL;
 653         if (p->type >= SOCK_MAX)
 654                 goto out_illegal;
 655
 656         /* If we are trying to override a permanent protocol, bail. */
 657         answer = NULL;
 658         ret = -EPERM;
 659         last_perm = &inetsw6[p->type];
 660         list_for_each(lh, &inetsw6[p->type]) {
 661                 answer = list_entry(lh, struct inet_protosw, list);
 662
 663                 /* Check only the non-wild match. */
 664                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 665                         if (protocol == answer->protocol)
 666                                 break;
 667                         last_perm = lh;
 668                 }
 669
 670                 answer = NULL;
 671         }
 672         if (answer)
 673                 goto out_permanent;
 674
 675         /* Add the new entry after the last permanent entry if any, so that
 676          * the new entry does not override a permanent entry when matched with
 677          * a wild-card protocol. But it is allowed to override any existing
 678          * non-permanent entry.  This means that when we remove this entry, the
 679          * system automatically returns to the old behavior.
 680          */
 681         list_add_rcu(&p->list, last_perm);
 682         ret = 0;
 683 out:
 684         spin_unlock_bh(&inetsw6_lock);
 685         return ret;
 686
 687 out_permanent:
 688         pr_err("Attempt to override permanent protocol %d\n", protocol);
 689         goto out;
 690
 691 out_illegal:
 692         pr_err("Ignoring attempt to register invalid socket type %d\n",
 693                p->type);
 694         goto out;
 695 }
 696 EXPORT_SYMBOL(inet6_register_protosw);
 697
 698 void
 699 inet6_unregister_protosw(struct inet_protosw *p)
 700 {
 701         if (INET_PROTOSW_PERMANENT & p->flags) {
 702                 pr_err("Attempt to unregister permanent protocol %d\n",
 703                        p->protocol);
 704         } else {
 705                 spin_lock_bh(&inetsw6_lock);
 706                 list_del_rcu(&p->list);
 707                 spin_unlock_bh(&inetsw6_lock);
 708
 709                 synchronize_net();
 710         }
 711 }
 712 EXPORT_SYMBOL(inet6_unregister_protosw);
 713
 714 int inet6_sk_rebuild_header(struct sock *sk)
 715 {
 716         struct ipv6_pinfo *np = inet6_sk(sk);
 717         struct dst_entry *dst;
 718
 719         dst = __sk_dst_check(sk, np->dst_cookie);
 720
 721         if (!dst) {
 722                 struct inet_sock *inet = inet_sk(sk);
 723                 struct in6_addr *final_p, final;
 724                 struct flowi6 fl6;
 725
 726                 memset(&fl6, 0, sizeof(fl6));
 727                 fl6.flowi6_proto = sk->sk_protocol;
 728                 fl6.daddr = sk->sk_v6_daddr;
 729                 fl6.saddr = np->saddr;
 730                 fl6.flowlabel = np->flow_label;
 731                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 732                 fl6.flowi6_mark = sk->sk_mark;
 733                 fl6.fl6_dport = inet->inet_dport;
 734                 fl6.fl6_sport = inet->inet_sport;
 735                 fl6.flowi6_uid = sk->sk_uid;
 736                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 737
 738                 rcu_read_lock();
 739                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 740                                          &final);
 741                 rcu_read_unlock();
 742
 743                 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
 744                 if (IS_ERR(dst)) {
 745                         sk->sk_route_caps = 0;
 746                         sk->sk_err_soft = -PTR_ERR(dst);
 747                         return PTR_ERR(dst);
 748                 }
 749
 750                 ip6_dst_store(sk, dst, NULL, NULL);
 751         }
 752
 753         return 0;
 754 }
 755 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 756
 757 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 758                        const struct inet6_skb_parm *opt)
 759 {
 760         const struct ipv6_pinfo *np = inet6_sk(sk);
 761
 762         if (np->rxopt.all) {
 763                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
 764                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 765                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 766                      np->rxopt.bits.rxflow) ||
 767                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 768                      np->rxopt.bits.osrcrt)) ||
 769                     ((opt->dst1 || opt->dst0) &&
 770                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 771                         return true;
 772         }
 773         return false;
 774 }
 775 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 776
 777 static struct packet_type ipv6_packet_type __read_mostly = {
 778         .type = cpu_to_be16(ETH_P_IPV6),
 779         .func = ipv6_rcv,
 780         .list_func = ipv6_list_rcv,
 781 };
 782
 783 static int __init ipv6_packet_init(void)
 784 {
 785         dev_add_pack(&ipv6_packet_type);
 786         return 0;
 787 }
 788
 789 static void ipv6_packet_cleanup(void)
 790 {
 791         dev_remove_pack(&ipv6_packet_type);
 792 }
 793
 794 static int __net_init ipv6_init_mibs(struct net *net)
 795 {
 796         int i;
 797
 798         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 799         if (!net->mib.udp_stats_in6)
 800                 return -ENOMEM;
 801         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 802         if (!net->mib.udplite_stats_in6)
 803                 goto err_udplite_mib;
 804         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 805         if (!net->mib.ipv6_statistics)
 806                 goto err_ip_mib;
 807
 808         for_each_possible_cpu(i) {
 809                 struct ipstats_mib *af_inet6_stats;
 810                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 811                 u64_stats_init(&af_inet6_stats->syncp);
 812         }
 813
 814
 815         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 816         if (!net->mib.icmpv6_statistics)
 817                 goto err_icmp_mib;
 818         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 819                                                 GFP_KERNEL);
 820         if (!net->mib.icmpv6msg_statistics)
 821                 goto err_icmpmsg_mib;
 822         return 0;
 823
 824 err_icmpmsg_mib:
 825         free_percpu(net->mib.icmpv6_statistics);
 826 err_icmp_mib:
 827         free_percpu(net->mib.ipv6_statistics);
 828 err_ip_mib:
 829         free_percpu(net->mib.udplite_stats_in6);
 830 err_udplite_mib:
 831         free_percpu(net->mib.udp_stats_in6);
 832         return -ENOMEM;
 833 }
 834
 835 static void ipv6_cleanup_mibs(struct net *net)
 836 {
 837         free_percpu(net->mib.udp_stats_in6);
 838         free_percpu(net->mib.udplite_stats_in6);
 839         free_percpu(net->mib.ipv6_statistics);
 840         free_percpu(net->mib.icmpv6_statistics);
 841         kfree(net->mib.icmpv6msg_statistics);
 842 }
 843
 844 static int __net_init inet6_net_init(struct net *net)
 845 {
 846         int err = 0;
 847
 848         net->ipv6.sysctl.bindv6only = 0;
 849         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 850         net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
 851         net->ipv6.sysctl.flowlabel_consistency = 1;
 852         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
 853         net->ipv6.sysctl.idgen_retries = 3;
 854         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 855         net->ipv6.sysctl.flowlabel_state_ranges = 0;
 856         net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
 857         net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
 858         net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
 859         net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
 860         atomic_set(&net->ipv6.fib6_sernum, 1);
 861
 862         err = ipv6_init_mibs(net);
 863         if (err)
 864                 return err;
 865 #ifdef CONFIG_PROC_FS
 866         err = udp6_proc_init(net);
 867         if (err)
 868                 goto out;
 869         err = tcp6_proc_init(net);
 870         if (err)
 871                 goto proc_tcp6_fail;
 872         err = ac6_proc_init(net);
 873         if (err)
 874                 goto proc_ac6_fail;
 875 #endif
 876         return err;
 877
 878 #ifdef CONFIG_PROC_FS
 879 proc_ac6_fail:
 880         tcp6_proc_exit(net);
 881 proc_tcp6_fail:
 882         udp6_proc_exit(net);
 883 out:
 884         ipv6_cleanup_mibs(net);
 885         return err;
 886 #endif
 887 }
 888
 889 static void __net_exit inet6_net_exit(struct net *net)
 890 {
 891 #ifdef CONFIG_PROC_FS
 892         udp6_proc_exit(net);
 893         tcp6_proc_exit(net);
 894         ac6_proc_exit(net);
 895 #endif
 896         ipv6_cleanup_mibs(net);
 897 }
 898
 899 static struct pernet_operations inet6_net_ops = {
 900         .init = inet6_net_init,
 901         .exit = inet6_net_exit,
 902 };
 903
 904 static const struct ipv6_stub ipv6_stub_impl = {
 905         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 906         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 907         .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
 908         .fib6_get_table    = fib6_get_table,
 909         .fib6_table_lookup = fib6_table_lookup,
 910         .fib6_lookup       = fib6_lookup,
 911         .fib6_multipath_select = fib6_multipath_select,
 912         .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
 913         .udpv6_encap_enable = udpv6_encap_enable,
 914         .ndisc_send_na = ndisc_send_na,
 915         .nd_tbl = &nd_tbl,
 916 };
 917
 918 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
 919         .inet6_bind = __inet6_bind,
 920 };
 921
 922 static int __init inet6_init(void)
 923 {
 924         struct list_head *r;
 925         int err = 0;
 926
 927         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 928
 929         /* Register the socket-side information for inet6_create.  */
 930         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 931                 INIT_LIST_HEAD(r);
 932
 933         if (disable_ipv6_mod) {
 934                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 935                 goto out;
 936         }
 937
 938         err = proto_register(&tcpv6_prot, 1);
 939         if (err)
 940                 goto out;
 941
 942         err = proto_register(&udpv6_prot, 1);
 943         if (err)
 944                 goto out_unregister_tcp_proto;
 945
 946         err = proto_register(&udplitev6_prot, 1);
 947         if (err)
 948                 goto out_unregister_udp_proto;
 949
 950         err = proto_register(&rawv6_prot, 1);
 951         if (err)
 952                 goto out_unregister_udplite_proto;
 953
 954         err = proto_register(&pingv6_prot, 1);
 955         if (err)
 956                 goto out_unregister_raw_proto;
 957
 958         /* We MUST register RAW sockets before we create the ICMP6,
 959          * IGMP6, or NDISC control sockets.
 960          */
 961         err = rawv6_init();
 962         if (err)
 963                 goto out_unregister_ping_proto;
 964
 965         /* Register the family here so that the init calls below will
 966          * be able to create sockets. (?? is this dangerous ??)
 967          */
 968         err = sock_register(&inet6_family_ops);
 969         if (err)
 970                 goto out_sock_register_fail;
 971
 972         /*
 973          *      ipngwg API draft makes clear that the correct semantics
 974          *      for TCP and UDP is to consider one TCP and UDP instance
 975          *      in a host available by both INET and INET6 APIs and
 976          *      able to communicate via both network protocols.
 977          */
 978
 979         err = register_pernet_subsys(&inet6_net_ops);
 980         if (err)
 981                 goto register_pernet_fail;
 982         err = ip6_mr_init();
 983         if (err)
 984                 goto ipmr_fail;
 985         err = icmpv6_init();
 986         if (err)
 987                 goto icmp_fail;
 988         err = ndisc_init();
 989         if (err)
 990                 goto ndisc_fail;
 991         err = igmp6_init();
 992         if (err)
 993                 goto igmp_fail;
 994
 995         err = ipv6_netfilter_init();
 996         if (err)
 997                 goto netfilter_fail;
 998         /* Create /proc/foo6 entries. */
 999 #ifdef CONFIG_PROC_FS
1000         err = -ENOMEM;
1001         if (raw6_proc_init())
1002                 goto proc_raw6_fail;
1003         if (udplite6_proc_init())
1004                 goto proc_udplite6_fail;
1005         if (ipv6_misc_proc_init())
1006                 goto proc_misc6_fail;
1007         if (if6_proc_init())
1008                 goto proc_if6_fail;
1009 #endif
1010         err = ip6_route_init();
1011         if (err)
1012                 goto ip6_route_fail;
1013         err = ndisc_late_init();
1014         if (err)
1015                 goto ndisc_late_fail;
1016         err = ip6_flowlabel_init();
1017         if (err)
1018                 goto ip6_flowlabel_fail;
1019         err = addrconf_init();
1020         if (err)
1021                 goto addrconf_fail;
1022
1023         /* Init v6 extension headers. */
1024         err = ipv6_exthdrs_init();
1025         if (err)
1026                 goto ipv6_exthdrs_fail;
1027
1028         err = ipv6_frag_init();
1029         if (err)
1030                 goto ipv6_frag_fail;
1031
1032         /* Init v6 transport protocols. */
1033         err = udpv6_init();
1034         if (err)
1035                 goto udpv6_fail;
1036
1037         err = udplitev6_init();
1038         if (err)
1039                 goto udplitev6_fail;
1040
1041         err = udpv6_offload_init();
1042         if (err)
1043                 goto udpv6_offload_fail;
1044
1045         err = tcpv6_init();
1046         if (err)
1047                 goto tcpv6_fail;
1048
1049         err = ipv6_packet_init();
1050         if (err)
1051                 goto ipv6_packet_fail;
1052
1053         err = pingv6_init();
1054         if (err)
1055                 goto pingv6_fail;
1056
1057         err = calipso_init();
1058         if (err)
1059                 goto calipso_fail;
1060
1061         err = seg6_init();
1062         if (err)
1063                 goto seg6_fail;
1064
1065         err = igmp6_late_init();
1066         if (err)
1067                 goto igmp6_late_err;
1068
1069 #ifdef CONFIG_SYSCTL
1070         err = ipv6_sysctl_register();
1071         if (err)
1072                 goto sysctl_fail;
1073 #endif
1074
1075         /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1076         wmb();
1077         ipv6_stub = &ipv6_stub_impl;
1078         ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1079 out:
1080         return err;
1081
1082 #ifdef CONFIG_SYSCTL
1083 sysctl_fail:
1084         igmp6_late_cleanup();
1085 #endif
1086 igmp6_late_err:
1087         seg6_exit();
1088 seg6_fail:
1089         calipso_exit();
1090 calipso_fail:
1091         pingv6_exit();
1092 pingv6_fail:
1093         ipv6_packet_cleanup();
1094 ipv6_packet_fail:
1095         tcpv6_exit();
1096 tcpv6_fail:
1097         udpv6_offload_exit();
1098 udpv6_offload_fail:
1099         udplitev6_exit();
1100 udplitev6_fail:
1101         udpv6_exit();
1102 udpv6_fail:
1103         ipv6_frag_exit();
1104 ipv6_frag_fail:
1105         ipv6_exthdrs_exit();
1106 ipv6_exthdrs_fail:
1107         addrconf_cleanup();
1108 addrconf_fail:
1109         ip6_flowlabel_cleanup();
1110 ip6_flowlabel_fail:
1111         ndisc_late_cleanup();
1112 ndisc_late_fail:
1113         ip6_route_cleanup();
1114 ip6_route_fail:
1115 #ifdef CONFIG_PROC_FS
1116         if6_proc_exit();
1117 proc_if6_fail:
1118         ipv6_misc_proc_exit();
1119 proc_misc6_fail:
1120         udplite6_proc_exit();
1121 proc_udplite6_fail:
1122         raw6_proc_exit();
1123 proc_raw6_fail:
1124 #endif
1125         ipv6_netfilter_fini();
1126 netfilter_fail:
1127         igmp6_cleanup();
1128 igmp_fail:
1129         ndisc_cleanup();
1130 ndisc_fail:
1131         icmpv6_cleanup();
1132 icmp_fail:
1133         ip6_mr_cleanup();
1134 ipmr_fail:
1135         unregister_pernet_subsys(&inet6_net_ops);
1136 register_pernet_fail:
1137         sock_unregister(PF_INET6);
1138         rtnl_unregister_all(PF_INET6);
1139 out_sock_register_fail:
1140         rawv6_exit();
1141 out_unregister_ping_proto:
1142         proto_unregister(&pingv6_prot);
1143 out_unregister_raw_proto:
1144         proto_unregister(&rawv6_prot);
1145 out_unregister_udplite_proto:
1146         proto_unregister(&udplitev6_prot);
1147 out_unregister_udp_proto:
1148         proto_unregister(&udpv6_prot);
1149 out_unregister_tcp_proto:
1150         proto_unregister(&tcpv6_prot);
1151         goto out;
1152 }
1153 module_init(inet6_init);
1154
1155 MODULE_ALIAS_NETPROTO(PF_INET6);