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