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->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
 214         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
 215
 216         /* Init the ipv4 part of the socket since we can have sockets
 217          * using v6 API for ipv4.
 218          */
 219         inet->uc_ttl    = -1;
 220
 221         inet->mc_loop   = 1;
 222         inet->mc_ttl    = 1;
 223         inet->mc_index  = 0;
 224         inet->mc_list   = NULL;
 225         inet->rcv_tos   = 0;
 226
 227         if (net->ipv4.sysctl_ip_no_pmtu_disc)
 228                 inet->pmtudisc = IP_PMTUDISC_DONT;
 229         else
 230                 inet->pmtudisc = IP_PMTUDISC_WANT;
 231         /*
 232          * Increment only the relevant sk_prot->socks debug field, this changes
 233          * the previous behaviour of incrementing both the equivalent to
 234          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 235          *
 236          * This allows better debug granularity as we'll know exactly how many
 237          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 238          * transport protocol socks. -acme
 239          */
 240         sk_refcnt_debug_inc(sk);
 241
 242         if (inet->inet_num) {
 243                 /* It assumes that any protocol which allows
 244                  * the user to assign a number at socket
 245                  * creation time automatically shares.
 246                  */
 247                 inet->inet_sport = htons(inet->inet_num);
 248                 err = sk->sk_prot->hash(sk);
 249                 if (err) {
 250                         sk_common_release(sk);
 251                         goto out;
 252                 }
 253         }
 254         if (sk->sk_prot->init) {
 255                 err = sk->sk_prot->init(sk);
 256                 if (err) {
 257                         sk_common_release(sk);
 258                         goto out;
 259                 }
 260         }
 261
 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         int addr_type = 0;
 288         int err = 0;
 289
 290         /* If the socket has its own bind function then use it. */
 291         if (sk->sk_prot->bind)
 292                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 293
 294         if (addr_len < SIN6_LEN_RFC2133)
 295                 return -EINVAL;
 296
 297         if (addr->sin6_family != AF_INET6)
 298                 return -EAFNOSUPPORT;
 299
 300         addr_type = ipv6_addr_type(&addr->sin6_addr);
 301         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 302                 return -EINVAL;
 303
 304         snum = ntohs(addr->sin6_port);
 305         if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 306                 return -EACCES;
 307
 308         lock_sock(sk);
 309
 310         /* Check these errors (active socket, double bind). */
 311         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 312                 err = -EINVAL;
 313                 goto out;
 314         }
 315
 316         /* Check if the address belongs to the host. */
 317         if (addr_type == IPV6_ADDR_MAPPED) {
 318                 int chk_addr_ret;
 319
 320                 /* Binding to v4-mapped address on a v6-only socket
 321                  * makes no sense
 322                  */
 323                 if (sk->sk_ipv6only) {
 324                         err = -EINVAL;
 325                         goto out;
 326                 }
 327
 328                 /* Reproduce AF_INET checks to make the bindings consistent */
 329                 v4addr = addr->sin6_addr.s6_addr32[3];
 330                 chk_addr_ret = inet_addr_type(net, v4addr);
 331                 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
 332                     !(inet->freebind || inet->transparent) &&
 333                     v4addr != htonl(INADDR_ANY) &&
 334                     chk_addr_ret != RTN_LOCAL &&
 335                     chk_addr_ret != RTN_MULTICAST &&
 336                     chk_addr_ret != RTN_BROADCAST) {
 337                         err = -EADDRNOTAVAIL;
 338                         goto out;
 339                 }
 340         } else {
 341                 if (addr_type != IPV6_ADDR_ANY) {
 342                         struct net_device *dev = NULL;
 343
 344                         rcu_read_lock();
 345                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 346                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 347                                     addr->sin6_scope_id) {
 348                                         /* Override any existing binding, if another one
 349                                          * is supplied by user.
 350                                          */
 351                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 352                                 }
 353
 354                                 /* Binding to link-local address requires an interface */
 355                                 if (!sk->sk_bound_dev_if) {
 356                                         err = -EINVAL;
 357                                         goto out_unlock;
 358                                 }
 359                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 360                                 if (!dev) {
 361                                         err = -ENODEV;
 362                                         goto out_unlock;
 363                                 }
 364                         }
 365
 366                         /* ipv4 addr of the socket is invalid.  Only the
 367                          * unspecified and mapped address have a v4 equivalent.
 368                          */
 369                         v4addr = LOOPBACK4_IPV6;
 370                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 371                                 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
 372                                     !(inet->freebind || inet->transparent) &&
 373                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 374                                                    dev, 0)) {
 375                                         err = -EADDRNOTAVAIL;
 376                                         goto out_unlock;
 377                                 }
 378                         }
 379                         rcu_read_unlock();
 380                 }
 381         }
 382
 383         inet->inet_rcv_saddr = v4addr;
 384         inet->inet_saddr = v4addr;
 385
 386         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 387
 388         if (!(addr_type & IPV6_ADDR_MULTICAST))
 389                 np->saddr = addr->sin6_addr;
 390
 391         /* Make sure we are allowed to bind here. */
 392         if ((snum || !inet->bind_address_no_port) &&
 393             sk->sk_prot->get_port(sk, snum)) {
 394                 inet_reset_saddr(sk);
 395                 err = -EADDRINUSE;
 396                 goto out;
 397         }
 398
 399         if (addr_type != IPV6_ADDR_ANY) {
 400                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 401                 if (addr_type != IPV6_ADDR_MAPPED)
 402                         sk->sk_ipv6only = 1;
 403         }
 404         if (snum)
 405                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 406         inet->inet_sport = htons(inet->inet_num);
 407         inet->inet_dport = 0;
 408         inet->inet_daddr = 0;
 409 out:
 410         release_sock(sk);
 411         return err;
 412 out_unlock:
 413         rcu_read_unlock();
 414         goto out;
 415 }
 416 EXPORT_SYMBOL(inet6_bind);
 417
 418 int inet6_release(struct socket *sock)
 419 {
 420         struct sock *sk = sock->sk;
 421
 422         if (!sk)
 423                 return -EINVAL;
 424
 425         /* Free mc lists */
 426         ipv6_sock_mc_close(sk);
 427
 428         /* Free ac lists */
 429         ipv6_sock_ac_close(sk);
 430
 431         return inet_release(sock);
 432 }
 433 EXPORT_SYMBOL(inet6_release);
 434
 435 void inet6_destroy_sock(struct sock *sk)
 436 {
 437         struct ipv6_pinfo *np = inet6_sk(sk);
 438         struct sk_buff *skb;
 439         struct ipv6_txoptions *opt;
 440
 441         /* Release rx options */
 442
 443         skb = xchg(&np->pktoptions, NULL);
 444         if (skb)
 445                 kfree_skb(skb);
 446
 447         skb = xchg(&np->rxpmtu, NULL);
 448         if (skb)
 449                 kfree_skb(skb);
 450
 451         /* Free flowlabels */
 452         fl6_free_socklist(sk);
 453
 454         /* Free tx options */
 455
 456         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 457         if (opt) {
 458                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 459                 txopt_put(opt);
 460         }
 461 }
 462 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 463
 464 /*
 465  *      This does both peername and sockname.
 466  */
 467
 468 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 469                  int *uaddr_len, int peer)
 470 {
 471         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 472         struct sock *sk = sock->sk;
 473         struct inet_sock *inet = inet_sk(sk);
 474         struct ipv6_pinfo *np = inet6_sk(sk);
 475
 476         sin->sin6_family = AF_INET6;
 477         sin->sin6_flowinfo = 0;
 478         sin->sin6_scope_id = 0;
 479         if (peer) {
 480                 if (!inet->inet_dport)
 481                         return -ENOTCONN;
 482                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 483                     peer == 1)
 484                         return -ENOTCONN;
 485                 sin->sin6_port = inet->inet_dport;
 486                 sin->sin6_addr = sk->sk_v6_daddr;
 487                 if (np->sndflow)
 488                         sin->sin6_flowinfo = np->flow_label;
 489         } else {
 490                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 491                         sin->sin6_addr = np->saddr;
 492                 else
 493                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 494
 495                 sin->sin6_port = inet->inet_sport;
 496         }
 497         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 498                                                  sk->sk_bound_dev_if);
 499         *uaddr_len = sizeof(*sin);
 500         return 0;
 501 }
 502 EXPORT_SYMBOL(inet6_getname);
 503
 504 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 505 {
 506         struct sock *sk = sock->sk;
 507         struct net *net = sock_net(sk);
 508
 509         switch (cmd) {
 510         case SIOCGSTAMP:
 511                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 512
 513         case SIOCGSTAMPNS:
 514                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 515
 516         case SIOCADDRT:
 517         case SIOCDELRT:
 518
 519                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 520
 521         case SIOCSIFADDR:
 522                 return addrconf_add_ifaddr(net, (void __user *) arg);
 523         case SIOCDIFADDR:
 524                 return addrconf_del_ifaddr(net, (void __user *) arg);
 525         case SIOCSIFDSTADDR:
 526                 return addrconf_set_dstaddr(net, (void __user *) arg);
 527         default:
 528                 if (!sk->sk_prot->ioctl)
 529                         return -ENOIOCTLCMD;
 530                 return sk->sk_prot->ioctl(sk, cmd, arg);
 531         }
 532         /*NOTREACHED*/
 533         return 0;
 534 }
 535 EXPORT_SYMBOL(inet6_ioctl);
 536
 537 const struct proto_ops inet6_stream_ops = {
 538         .family            = PF_INET6,
 539         .owner             = THIS_MODULE,
 540         .release           = inet6_release,
 541         .bind              = inet6_bind,
 542         .connect           = inet_stream_connect,       /* ok           */
 543         .socketpair        = sock_no_socketpair,        /* a do nothing */
 544         .accept            = inet_accept,               /* ok           */
 545         .getname           = inet6_getname,
 546         .poll              = tcp_poll,                  /* ok           */
 547         .ioctl             = inet6_ioctl,               /* must change  */
 548         .listen            = inet_listen,               /* ok           */
 549         .shutdown          = inet_shutdown,             /* ok           */
 550         .setsockopt        = sock_common_setsockopt,    /* ok           */
 551         .getsockopt        = sock_common_getsockopt,    /* ok           */
 552         .sendmsg           = inet_sendmsg,              /* ok           */
 553         .recvmsg           = inet_recvmsg,              /* ok           */
 554         .mmap              = sock_no_mmap,
 555         .sendpage          = inet_sendpage,
 556         .splice_read       = tcp_splice_read,
 557         .read_sock         = tcp_read_sock,
 558         .peek_len          = tcp_peek_len,
 559 #ifdef CONFIG_COMPAT
 560         .compat_setsockopt = compat_sock_common_setsockopt,
 561         .compat_getsockopt = compat_sock_common_getsockopt,
 562 #endif
 563 };
 564
 565 const struct proto_ops inet6_dgram_ops = {
 566         .family            = PF_INET6,
 567         .owner             = THIS_MODULE,
 568         .release           = inet6_release,
 569         .bind              = inet6_bind,
 570         .connect           = inet_dgram_connect,        /* ok           */
 571         .socketpair        = sock_no_socketpair,        /* a do nothing */
 572         .accept            = sock_no_accept,            /* a do nothing */
 573         .getname           = inet6_getname,
 574         .poll              = udp_poll,                  /* ok           */
 575         .ioctl             = inet6_ioctl,               /* must change  */
 576         .listen            = sock_no_listen,            /* ok           */
 577         .shutdown          = inet_shutdown,             /* ok           */
 578         .setsockopt        = sock_common_setsockopt,    /* ok           */
 579         .getsockopt        = sock_common_getsockopt,    /* ok           */
 580         .sendmsg           = inet_sendmsg,              /* ok           */
 581         .recvmsg           = inet_recvmsg,              /* ok           */
 582         .mmap              = sock_no_mmap,
 583         .sendpage          = sock_no_sendpage,
 584         .set_peek_off      = sk_set_peek_off,
 585 #ifdef CONFIG_COMPAT
 586         .compat_setsockopt = compat_sock_common_setsockopt,
 587         .compat_getsockopt = compat_sock_common_getsockopt,
 588 #endif
 589 };
 590
 591 static const struct net_proto_family inet6_family_ops = {
 592         .family = PF_INET6,
 593         .create = inet6_create,
 594         .owner  = THIS_MODULE,
 595 };
 596
 597 int inet6_register_protosw(struct inet_protosw *p)
 598 {
 599         struct list_head *lh;
 600         struct inet_protosw *answer;
 601         struct list_head *last_perm;
 602         int protocol = p->protocol;
 603         int ret;
 604
 605         spin_lock_bh(&inetsw6_lock);
 606
 607         ret = -EINVAL;
 608         if (p->type >= SOCK_MAX)
 609                 goto out_illegal;
 610
 611         /* If we are trying to override a permanent protocol, bail. */
 612         answer = NULL;
 613         ret = -EPERM;
 614         last_perm = &inetsw6[p->type];
 615         list_for_each(lh, &inetsw6[p->type]) {
 616                 answer = list_entry(lh, struct inet_protosw, list);
 617
 618                 /* Check only the non-wild match. */
 619                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 620                         if (protocol == answer->protocol)
 621                                 break;
 622                         last_perm = lh;
 623                 }
 624
 625                 answer = NULL;
 626         }
 627         if (answer)
 628                 goto out_permanent;
 629
 630         /* Add the new entry after the last permanent entry if any, so that
 631          * the new entry does not override a permanent entry when matched with
 632          * a wild-card protocol. But it is allowed to override any existing
 633          * non-permanent entry.  This means that when we remove this entry, the
 634          * system automatically returns to the old behavior.
 635          */
 636         list_add_rcu(&p->list, last_perm);
 637         ret = 0;
 638 out:
 639         spin_unlock_bh(&inetsw6_lock);
 640         return ret;
 641
 642 out_permanent:
 643         pr_err("Attempt to override permanent protocol %d\n", protocol);
 644         goto out;
 645
 646 out_illegal:
 647         pr_err("Ignoring attempt to register invalid socket type %d\n",
 648                p->type);
 649         goto out;
 650 }
 651 EXPORT_SYMBOL(inet6_register_protosw);
 652
 653 void
 654 inet6_unregister_protosw(struct inet_protosw *p)
 655 {
 656         if (INET_PROTOSW_PERMANENT & p->flags) {
 657                 pr_err("Attempt to unregister permanent protocol %d\n",
 658                        p->protocol);
 659         } else {
 660                 spin_lock_bh(&inetsw6_lock);
 661                 list_del_rcu(&p->list);
 662                 spin_unlock_bh(&inetsw6_lock);
 663
 664                 synchronize_net();
 665         }
 666 }
 667 EXPORT_SYMBOL(inet6_unregister_protosw);
 668
 669 int inet6_sk_rebuild_header(struct sock *sk)
 670 {
 671         struct ipv6_pinfo *np = inet6_sk(sk);
 672         struct dst_entry *dst;
 673
 674         dst = __sk_dst_check(sk, np->dst_cookie);
 675
 676         if (!dst) {
 677                 struct inet_sock *inet = inet_sk(sk);
 678                 struct in6_addr *final_p, final;
 679                 struct flowi6 fl6;
 680
 681                 memset(&fl6, 0, sizeof(fl6));
 682                 fl6.flowi6_proto = sk->sk_protocol;
 683                 fl6.daddr = sk->sk_v6_daddr;
 684                 fl6.saddr = np->saddr;
 685                 fl6.flowlabel = np->flow_label;
 686                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 687                 fl6.flowi6_mark = sk->sk_mark;
 688                 fl6.fl6_dport = inet->inet_dport;
 689                 fl6.fl6_sport = inet->inet_sport;
 690                 fl6.flowi6_uid = sk->sk_uid;
 691                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 692
 693                 rcu_read_lock();
 694                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 695                                          &final);
 696                 rcu_read_unlock();
 697
 698                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 699                 if (IS_ERR(dst)) {
 700                         sk->sk_route_caps = 0;
 701                         sk->sk_err_soft = -PTR_ERR(dst);
 702                         return PTR_ERR(dst);
 703                 }
 704
 705                 ip6_dst_store(sk, dst, NULL, NULL);
 706         }
 707
 708         return 0;
 709 }
 710 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 711
 712 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 713                        const struct inet6_skb_parm *opt)
 714 {
 715         const struct ipv6_pinfo *np = inet6_sk(sk);
 716
 717         if (np->rxopt.all) {
 718                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
 719                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 720                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 721                      np->rxopt.bits.rxflow) ||
 722                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 723                      np->rxopt.bits.osrcrt)) ||
 724                     ((opt->dst1 || opt->dst0) &&
 725                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 726                         return true;
 727         }
 728         return false;
 729 }
 730 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 731
 732 static struct packet_type ipv6_packet_type __read_mostly = {
 733         .type = cpu_to_be16(ETH_P_IPV6),
 734         .func = ipv6_rcv,
 735 };
 736
 737 static int __init ipv6_packet_init(void)
 738 {
 739         dev_add_pack(&ipv6_packet_type);
 740         return 0;
 741 }
 742
 743 static void ipv6_packet_cleanup(void)
 744 {
 745         dev_remove_pack(&ipv6_packet_type);
 746 }
 747
 748 static int __net_init ipv6_init_mibs(struct net *net)
 749 {
 750         int i;
 751
 752         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 753         if (!net->mib.udp_stats_in6)
 754                 return -ENOMEM;
 755         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 756         if (!net->mib.udplite_stats_in6)
 757                 goto err_udplite_mib;
 758         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 759         if (!net->mib.ipv6_statistics)
 760                 goto err_ip_mib;
 761
 762         for_each_possible_cpu(i) {
 763                 struct ipstats_mib *af_inet6_stats;
 764                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 765                 u64_stats_init(&af_inet6_stats->syncp);
 766         }
 767
 768
 769         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 770         if (!net->mib.icmpv6_statistics)
 771                 goto err_icmp_mib;
 772         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 773                                                 GFP_KERNEL);
 774         if (!net->mib.icmpv6msg_statistics)
 775                 goto err_icmpmsg_mib;
 776         return 0;
 777
 778 err_icmpmsg_mib:
 779         free_percpu(net->mib.icmpv6_statistics);
 780 err_icmp_mib:
 781         free_percpu(net->mib.ipv6_statistics);
 782 err_ip_mib:
 783         free_percpu(net->mib.udplite_stats_in6);
 784 err_udplite_mib:
 785         free_percpu(net->mib.udp_stats_in6);
 786         return -ENOMEM;
 787 }
 788
 789 static void ipv6_cleanup_mibs(struct net *net)
 790 {
 791         free_percpu(net->mib.udp_stats_in6);
 792         free_percpu(net->mib.udplite_stats_in6);
 793         free_percpu(net->mib.ipv6_statistics);
 794         free_percpu(net->mib.icmpv6_statistics);
 795         kfree(net->mib.icmpv6msg_statistics);
 796 }
 797
 798 static int __net_init inet6_net_init(struct net *net)
 799 {
 800         int err = 0;
 801
 802         net->ipv6.sysctl.bindv6only = 0;
 803         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 804         net->ipv6.sysctl.flowlabel_consistency = 1;
 805         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
 806         net->ipv6.sysctl.idgen_retries = 3;
 807         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 808         net->ipv6.sysctl.flowlabel_state_ranges = 0;
 809         atomic_set(&net->ipv6.fib6_sernum, 1);
 810
 811         err = ipv6_init_mibs(net);
 812         if (err)
 813                 return err;
 814 #ifdef CONFIG_PROC_FS
 815         err = udp6_proc_init(net);
 816         if (err)
 817                 goto out;
 818         err = tcp6_proc_init(net);
 819         if (err)
 820                 goto proc_tcp6_fail;
 821         err = ac6_proc_init(net);
 822         if (err)
 823                 goto proc_ac6_fail;
 824 #endif
 825         return err;
 826
 827 #ifdef CONFIG_PROC_FS
 828 proc_ac6_fail:
 829         tcp6_proc_exit(net);
 830 proc_tcp6_fail:
 831         udp6_proc_exit(net);
 832 out:
 833         ipv6_cleanup_mibs(net);
 834         return err;
 835 #endif
 836 }
 837
 838 static void __net_exit inet6_net_exit(struct net *net)
 839 {
 840 #ifdef CONFIG_PROC_FS
 841         udp6_proc_exit(net);
 842         tcp6_proc_exit(net);
 843         ac6_proc_exit(net);
 844 #endif
 845         ipv6_cleanup_mibs(net);
 846 }
 847
 848 static struct pernet_operations inet6_net_ops = {
 849         .init = inet6_net_init,
 850         .exit = inet6_net_exit,
 851 };
 852
 853 static const struct ipv6_stub ipv6_stub_impl = {
 854         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 855         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 856         .ipv6_dst_lookup = ip6_dst_lookup,
 857         .udpv6_encap_enable = udpv6_encap_enable,
 858         .ndisc_send_na = ndisc_send_na,
 859         .nd_tbl = &nd_tbl,
 860 };
 861
 862 static int __init inet6_init(void)
 863 {
 864         struct list_head *r;
 865         int err = 0;
 866
 867         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 868
 869         /* Register the socket-side information for inet6_create.  */
 870         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 871                 INIT_LIST_HEAD(r);
 872
 873         if (disable_ipv6_mod) {
 874                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 875                 goto out;
 876         }
 877
 878         err = proto_register(&tcpv6_prot, 1);
 879         if (err)
 880                 goto out;
 881
 882         err = proto_register(&udpv6_prot, 1);
 883         if (err)
 884                 goto out_unregister_tcp_proto;
 885
 886         err = proto_register(&udplitev6_prot, 1);
 887         if (err)
 888                 goto out_unregister_udp_proto;
 889
 890         err = proto_register(&rawv6_prot, 1);
 891         if (err)
 892                 goto out_unregister_udplite_proto;
 893
 894         err = proto_register(&pingv6_prot, 1);
 895         if (err)
 896                 goto out_unregister_ping_proto;
 897
 898         /* We MUST register RAW sockets before we create the ICMP6,
 899          * IGMP6, or NDISC control sockets.
 900          */
 901         err = rawv6_init();
 902         if (err)
 903                 goto out_unregister_raw_proto;
 904
 905         /* Register the family here so that the init calls below will
 906          * be able to create sockets. (?? is this dangerous ??)
 907          */
 908         err = sock_register(&inet6_family_ops);
 909         if (err)
 910                 goto out_sock_register_fail;
 911
 912         /*
 913          *      ipngwg API draft makes clear that the correct semantics
 914          *      for TCP and UDP is to consider one TCP and UDP instance
 915          *      in a host available by both INET and INET6 APIs and
 916          *      able to communicate via both network protocols.
 917          */
 918
 919         err = register_pernet_subsys(&inet6_net_ops);
 920         if (err)
 921                 goto register_pernet_fail;
 922         err = icmpv6_init();
 923         if (err)
 924                 goto icmp_fail;
 925         err = ip6_mr_init();
 926         if (err)
 927                 goto ipmr_fail;
 928         err = ndisc_init();
 929         if (err)
 930                 goto ndisc_fail;
 931         err = igmp6_init();
 932         if (err)
 933                 goto igmp_fail;
 934
 935         ipv6_stub = &ipv6_stub_impl;
 936
 937         err = ipv6_netfilter_init();
 938         if (err)
 939                 goto netfilter_fail;
 940         /* Create /proc/foo6 entries. */
 941 #ifdef CONFIG_PROC_FS
 942         err = -ENOMEM;
 943         if (raw6_proc_init())
 944                 goto proc_raw6_fail;
 945         if (udplite6_proc_init())
 946                 goto proc_udplite6_fail;
 947         if (ipv6_misc_proc_init())
 948                 goto proc_misc6_fail;
 949         if (if6_proc_init())
 950                 goto proc_if6_fail;
 951 #endif
 952         err = ip6_route_init();
 953         if (err)
 954                 goto ip6_route_fail;
 955         err = ndisc_late_init();
 956         if (err)
 957                 goto ndisc_late_fail;
 958         err = ip6_flowlabel_init();
 959         if (err)
 960                 goto ip6_flowlabel_fail;
 961         err = addrconf_init();
 962         if (err)
 963                 goto addrconf_fail;
 964
 965         /* Init v6 extension headers. */
 966         err = ipv6_exthdrs_init();
 967         if (err)
 968                 goto ipv6_exthdrs_fail;
 969
 970         err = ipv6_frag_init();
 971         if (err)
 972                 goto ipv6_frag_fail;
 973
 974         /* Init v6 transport protocols. */
 975         err = udpv6_init();
 976         if (err)
 977                 goto udpv6_fail;
 978
 979         err = udplitev6_init();
 980         if (err)
 981                 goto udplitev6_fail;
 982
 983         err = udpv6_offload_init();
 984         if (err)
 985                 goto udpv6_offload_fail;
 986
 987         err = tcpv6_init();
 988         if (err)
 989                 goto tcpv6_fail;
 990
 991         err = ipv6_packet_init();
 992         if (err)
 993                 goto ipv6_packet_fail;
 994
 995         err = pingv6_init();
 996         if (err)
 997                 goto pingv6_fail;
 998
 999         err = calipso_init();
1000         if (err)
1001                 goto calipso_fail;
1002
1003         err = seg6_init();
1004         if (err)
1005                 goto seg6_fail;
1006
1007 #ifdef CONFIG_SYSCTL
1008         err = ipv6_sysctl_register();
1009         if (err)
1010                 goto sysctl_fail;
1011 #endif
1012 out:
1013         return err;
1014
1015 #ifdef CONFIG_SYSCTL
1016 sysctl_fail:
1017         seg6_exit();
1018 #endif
1019 seg6_fail:
1020         calipso_exit();
1021 calipso_fail:
1022         pingv6_exit();
1023 pingv6_fail:
1024         ipv6_packet_cleanup();
1025 ipv6_packet_fail:
1026         tcpv6_exit();
1027 tcpv6_fail:
1028         udpv6_offload_exit();
1029 udpv6_offload_fail:
1030         udplitev6_exit();
1031 udplitev6_fail:
1032         udpv6_exit();
1033 udpv6_fail:
1034         ipv6_frag_exit();
1035 ipv6_frag_fail:
1036         ipv6_exthdrs_exit();
1037 ipv6_exthdrs_fail:
1038         addrconf_cleanup();
1039 addrconf_fail:
1040         ip6_flowlabel_cleanup();
1041 ip6_flowlabel_fail:
1042         ndisc_late_cleanup();
1043 ndisc_late_fail:
1044         ip6_route_cleanup();
1045 ip6_route_fail:
1046 #ifdef CONFIG_PROC_FS
1047         if6_proc_exit();
1048 proc_if6_fail:
1049         ipv6_misc_proc_exit();
1050 proc_misc6_fail:
1051         udplite6_proc_exit();
1052 proc_udplite6_fail:
1053         raw6_proc_exit();
1054 proc_raw6_fail:
1055 #endif
1056         ipv6_netfilter_fini();
1057 netfilter_fail:
1058         igmp6_cleanup();
1059 igmp_fail:
1060         ndisc_cleanup();
1061 ndisc_fail:
1062         ip6_mr_cleanup();
1063 ipmr_fail:
1064         icmpv6_cleanup();
1065 icmp_fail:
1066         unregister_pernet_subsys(&inet6_net_ops);
1067 register_pernet_fail:
1068         sock_unregister(PF_INET6);
1069         rtnl_unregister_all(PF_INET6);
1070 out_sock_register_fail:
1071         rawv6_exit();
1072 out_unregister_ping_proto:
1073         proto_unregister(&pingv6_prot);
1074 out_unregister_raw_proto:
1075         proto_unregister(&rawv6_prot);
1076 out_unregister_udplite_proto:
1077         proto_unregister(&udplitev6_prot);
1078 out_unregister_udp_proto:
1079         proto_unregister(&udpv6_prot);
1080 out_unregister_tcp_proto:
1081         proto_unregister(&tcpv6_prot);
1082         goto out;
1083 }
1084 module_init(inet6_init);
1085
1086 MODULE_ALIAS_NETPROTO(PF_INET6);