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 void inet6_sock_destruct(struct sock *sk)
 111 {
 112         inet6_cleanup_sock(sk);
 113         inet_sock_destruct(sk);
 114 }
 115 EXPORT_SYMBOL_GPL(inet6_sock_destruct);
 116
 117 static int inet6_create(struct net *net, struct socket *sock, int protocol,
 118                         int kern)
 119 {
 120         struct inet_sock *inet;
 121         struct ipv6_pinfo *np;
 122         struct sock *sk;
 123         struct inet_protosw *answer;
 124         struct proto *answer_prot;
 125         unsigned char answer_flags;
 126         int try_loading_module = 0;
 127         int err;
 128
 129         if (protocol < 0 || protocol >= IPPROTO_MAX)
 130                 return -EINVAL;
 131
 132         /* Look for the requested type/protocol pair. */
 133 lookup_protocol:
 134         err = -ESOCKTNOSUPPORT;
 135         rcu_read_lock();
 136         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 137
 138                 err = 0;
 139                 /* Check the non-wild match. */
 140                 if (protocol == answer->protocol) {
 141                         if (protocol != IPPROTO_IP)
 142                                 break;
 143                 } else {
 144                         /* Check for the two wild cases. */
 145                         if (IPPROTO_IP == protocol) {
 146                                 protocol = answer->protocol;
 147                                 break;
 148                         }
 149                         if (IPPROTO_IP == answer->protocol)
 150                                 break;
 151                 }
 152                 err = -EPROTONOSUPPORT;
 153         }
 154
 155         if (err) {
 156                 if (try_loading_module < 2) {
 157                         rcu_read_unlock();
 158                         /*
 159                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 160                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 161                          */
 162                         if (++try_loading_module == 1)
 163                                 request_module("net-pf-%d-proto-%d-type-%d",
 164                                                 PF_INET6, protocol, sock->type);
 165                         /*
 166                          * Fall back to generic, e.g. net-pf-10-proto-132
 167                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 168                          */
 169                         else
 170                                 request_module("net-pf-%d-proto-%d",
 171                                                 PF_INET6, protocol);
 172                         goto lookup_protocol;
 173                 } else
 174                         goto out_rcu_unlock;
 175         }
 176
 177         err = -EPERM;
 178         if (sock->type == SOCK_RAW && !kern &&
 179             !ns_capable(net->user_ns, CAP_NET_RAW))
 180                 goto out_rcu_unlock;
 181
 182         sock->ops = answer->ops;
 183         answer_prot = answer->prot;
 184         answer_flags = answer->flags;
 185         rcu_read_unlock();
 186
 187         WARN_ON(!answer_prot->slab);
 188
 189         err = -ENOBUFS;
 190         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
 191         if (!sk)
 192                 goto out;
 193
 194         sock_init_data(sock, sk);
 195
 196         err = 0;
 197         if (INET_PROTOSW_REUSE & answer_flags)
 198                 sk->sk_reuse = SK_CAN_REUSE;
 199
 200         inet = inet_sk(sk);
 201         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 202
 203         if (SOCK_RAW == sock->type) {
 204                 inet->inet_num = protocol;
 205                 if (IPPROTO_RAW == protocol)
 206                         inet->hdrincl = 1;
 207         }
 208
 209         sk->sk_destruct         = inet6_sock_destruct;
 210         sk->sk_family           = PF_INET6;
 211         sk->sk_protocol         = protocol;
 212
 213         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 214
 215         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 216         np->hop_limit   = -1;
 217         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 218         np->mc_loop     = 1;
 219         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 220         np->repflow     = net->ipv6.sysctl.flowlabel_reflect;
 221         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
 222
 223         /* Init the ipv4 part of the socket since we can have sockets
 224          * using v6 API for ipv4.
 225          */
 226         inet->uc_ttl    = -1;
 227
 228         inet->mc_loop   = 1;
 229         inet->mc_ttl    = 1;
 230         inet->mc_index  = 0;
 231         inet->mc_list   = NULL;
 232         inet->rcv_tos   = 0;
 233
 234         if (net->ipv4.sysctl_ip_no_pmtu_disc)
 235                 inet->pmtudisc = IP_PMTUDISC_DONT;
 236         else
 237                 inet->pmtudisc = IP_PMTUDISC_WANT;
 238         /*
 239          * Increment only the relevant sk_prot->socks debug field, this changes
 240          * the previous behaviour of incrementing both the equivalent to
 241          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 242          *
 243          * This allows better debug granularity as we'll know exactly how many
 244          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 245          * transport protocol socks. -acme
 246          */
 247         sk_refcnt_debug_inc(sk);
 248
 249         if (inet->inet_num) {
 250                 /* It assumes that any protocol which allows
 251                  * the user to assign a number at socket
 252                  * creation time automatically shares.
 253                  */
 254                 inet->inet_sport = htons(inet->inet_num);
 255                 err = sk->sk_prot->hash(sk);
 256                 if (err) {
 257                         sk_common_release(sk);
 258                         goto out;
 259                 }
 260         }
 261         if (sk->sk_prot->init) {
 262                 err = sk->sk_prot->init(sk);
 263                 if (err) {
 264                         sk_common_release(sk);
 265                         goto out;
 266                 }
 267         }
 268
 269         if (!kern) {
 270                 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
 271                 if (err) {
 272                         sk_common_release(sk);
 273                         goto out;
 274                 }
 275         }
 276 out:
 277         return err;
 278 out_rcu_unlock:
 279         rcu_read_unlock();
 280         goto out;
 281 }
 282
 283
 284 /* bind for INET6 API */
 285 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 286 {
 287         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 288         struct sock *sk = sock->sk;
 289         struct inet_sock *inet = inet_sk(sk);
 290         struct ipv6_pinfo *np = inet6_sk(sk);
 291         struct net *net = sock_net(sk);
 292         __be32 v4addr = 0;
 293         unsigned short snum;
 294         bool saved_ipv6only;
 295         int addr_type = 0;
 296         int err = 0;
 297
 298         /* If the socket has its own bind function then use it. */
 299         if (sk->sk_prot->bind)
 300                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 301
 302         if (addr_len < SIN6_LEN_RFC2133)
 303                 return -EINVAL;
 304
 305         if (addr->sin6_family != AF_INET6)
 306                 return -EAFNOSUPPORT;
 307
 308         addr_type = ipv6_addr_type(&addr->sin6_addr);
 309         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 310                 return -EINVAL;
 311
 312         snum = ntohs(addr->sin6_port);
 313         if (snum && snum < inet_prot_sock(net) &&
 314             !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
 315                 return -EACCES;
 316
 317         lock_sock(sk);
 318
 319         /* Check these errors (active socket, double bind). */
 320         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 321                 err = -EINVAL;
 322                 goto out;
 323         }
 324
 325         /* Check if the address belongs to the host. */
 326         if (addr_type == IPV6_ADDR_MAPPED) {
 327                 struct net_device *dev = NULL;
 328                 int chk_addr_ret;
 329
 330                 /* Binding to v4-mapped address on a v6-only socket
 331                  * makes no sense
 332                  */
 333                 if (sk->sk_ipv6only) {
 334                         err = -EINVAL;
 335                         goto out;
 336                 }
 337
 338                 rcu_read_lock();
 339                 if (sk->sk_bound_dev_if) {
 340                         dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 341                         if (!dev) {
 342                                 err = -ENODEV;
 343                                 goto out_unlock;
 344                         }
 345                 }
 346
 347                 /* Reproduce AF_INET checks to make the bindings consistent */
 348                 v4addr = addr->sin6_addr.s6_addr32[3];
 349                 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
 350                 rcu_read_unlock();
 351
 352                 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
 353                     !(inet->freebind || inet->transparent) &&
 354                     v4addr != htonl(INADDR_ANY) &&
 355                     chk_addr_ret != RTN_LOCAL &&
 356                     chk_addr_ret != RTN_MULTICAST &&
 357                     chk_addr_ret != RTN_BROADCAST) {
 358                         err = -EADDRNOTAVAIL;
 359                         goto out;
 360                 }
 361         } else {
 362                 if (addr_type != IPV6_ADDR_ANY) {
 363                         struct net_device *dev = NULL;
 364
 365                         rcu_read_lock();
 366                         if (__ipv6_addr_needs_scope_id(addr_type)) {
 367                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 368                                     addr->sin6_scope_id) {
 369                                         /* Override any existing binding, if another one
 370                                          * is supplied by user.
 371                                          */
 372                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 373                                 }
 374
 375                                 /* Binding to link-local address requires an interface */
 376                                 if (!sk->sk_bound_dev_if) {
 377                                         err = -EINVAL;
 378                                         goto out_unlock;
 379                                 }
 380                         }
 381
 382                         if (sk->sk_bound_dev_if) {
 383                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 384                                 if (!dev) {
 385                                         err = -ENODEV;
 386                                         goto out_unlock;
 387                                 }
 388                         }
 389
 390                         /* ipv4 addr of the socket is invalid.  Only the
 391                          * unspecified and mapped address have a v4 equivalent.
 392                          */
 393                         v4addr = LOOPBACK4_IPV6;
 394                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 395                                 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
 396                                     !(inet->freebind || inet->transparent) &&
 397                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 398                                                    dev, 0)) {
 399                                         err = -EADDRNOTAVAIL;
 400                                         goto out_unlock;
 401                                 }
 402                         }
 403                         rcu_read_unlock();
 404                 }
 405         }
 406
 407         inet->inet_rcv_saddr = v4addr;
 408         inet->inet_saddr = v4addr;
 409
 410         sk->sk_v6_rcv_saddr = addr->sin6_addr;
 411
 412         if (!(addr_type & IPV6_ADDR_MULTICAST))
 413                 np->saddr = addr->sin6_addr;
 414
 415         saved_ipv6only = sk->sk_ipv6only;
 416         if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
 417                 sk->sk_ipv6only = 1;
 418
 419         /* Make sure we are allowed to bind here. */
 420         if ((snum || !inet->bind_address_no_port) &&
 421             sk->sk_prot->get_port(sk, snum)) {
 422                 sk->sk_ipv6only = saved_ipv6only;
 423                 inet_reset_saddr(sk);
 424                 err = -EADDRINUSE;
 425                 goto out;
 426         }
 427
 428         if (addr_type != IPV6_ADDR_ANY)
 429                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 430         if (snum)
 431                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 432         inet->inet_sport = htons(inet->inet_num);
 433         inet->inet_dport = 0;
 434         inet->inet_daddr = 0;
 435 out:
 436         release_sock(sk);
 437         return err;
 438 out_unlock:
 439         rcu_read_unlock();
 440         goto out;
 441 }
 442 EXPORT_SYMBOL(inet6_bind);
 443
 444 int inet6_release(struct socket *sock)
 445 {
 446         struct sock *sk = sock->sk;
 447
 448         if (!sk)
 449                 return -EINVAL;
 450
 451         /* Free mc lists */
 452         ipv6_sock_mc_close(sk);
 453
 454         /* Free ac lists */
 455         ipv6_sock_ac_close(sk);
 456
 457         return inet_release(sock);
 458 }
 459 EXPORT_SYMBOL(inet6_release);
 460
 461 void inet6_destroy_sock(struct sock *sk)
 462 {
 463         struct ipv6_pinfo *np = inet6_sk(sk);
 464         struct sk_buff *skb;
 465         struct ipv6_txoptions *opt;
 466
 467         /* Release rx options */
 468
 469         skb = xchg(&np->pktoptions, NULL);
 470         if (skb)
 471                 kfree_skb(skb);
 472
 473         skb = xchg(&np->rxpmtu, NULL);
 474         if (skb)
 475                 kfree_skb(skb);
 476
 477         /* Free flowlabels */
 478         fl6_free_socklist(sk);
 479
 480         /* Free tx options */
 481
 482         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
 483         if (opt) {
 484                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 485                 txopt_put(opt);
 486         }
 487 }
 488 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 489
 490 void inet6_cleanup_sock(struct sock *sk)
 491 {
 492         inet6_destroy_sock(sk);
 493 }
 494 EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
 495
 496 /*
 497  *      This does both peername and sockname.
 498  */
 499
 500 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 501                  int *uaddr_len, int peer)
 502 {
 503         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 504         struct sock *sk = sock->sk;
 505         struct inet_sock *inet = inet_sk(sk);
 506         struct ipv6_pinfo *np = inet6_sk(sk);
 507
 508         sin->sin6_family = AF_INET6;
 509         sin->sin6_flowinfo = 0;
 510         sin->sin6_scope_id = 0;
 511         if (peer) {
 512                 if (!inet->inet_dport)
 513                         return -ENOTCONN;
 514                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 515                     peer == 1)
 516                         return -ENOTCONN;
 517                 sin->sin6_port = inet->inet_dport;
 518                 sin->sin6_addr = sk->sk_v6_daddr;
 519                 if (np->sndflow)
 520                         sin->sin6_flowinfo = np->flow_label;
 521         } else {
 522                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 523                         sin->sin6_addr = np->saddr;
 524                 else
 525                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
 526
 527                 sin->sin6_port = inet->inet_sport;
 528         }
 529         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
 530                                                  sk->sk_bound_dev_if);
 531         *uaddr_len = sizeof(*sin);
 532         return 0;
 533 }
 534 EXPORT_SYMBOL(inet6_getname);
 535
 536 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 537 {
 538         struct sock *sk = sock->sk;
 539         struct net *net = sock_net(sk);
 540
 541         switch (cmd) {
 542         case SIOCGSTAMP:
 543                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 544
 545         case SIOCGSTAMPNS:
 546                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 547
 548         case SIOCADDRT:
 549         case SIOCDELRT:
 550
 551                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 552
 553         case SIOCSIFADDR:
 554                 return addrconf_add_ifaddr(net, (void __user *) arg);
 555         case SIOCDIFADDR:
 556                 return addrconf_del_ifaddr(net, (void __user *) arg);
 557         case SIOCSIFDSTADDR:
 558                 return addrconf_set_dstaddr(net, (void __user *) arg);
 559         default:
 560                 if (!sk->sk_prot->ioctl)
 561                         return -ENOIOCTLCMD;
 562                 return sk->sk_prot->ioctl(sk, cmd, arg);
 563         }
 564         /*NOTREACHED*/
 565         return 0;
 566 }
 567 EXPORT_SYMBOL(inet6_ioctl);
 568
 569 const struct proto_ops inet6_stream_ops = {
 570         .family            = PF_INET6,
 571         .owner             = THIS_MODULE,
 572         .release           = inet6_release,
 573         .bind              = inet6_bind,
 574         .connect           = inet_stream_connect,       /* ok           */
 575         .socketpair        = sock_no_socketpair,        /* a do nothing */
 576         .accept            = inet_accept,               /* ok           */
 577         .getname           = inet6_getname,
 578         .poll              = tcp_poll,                  /* ok           */
 579         .ioctl             = inet6_ioctl,               /* must change  */
 580         .listen            = inet_listen,               /* ok           */
 581         .shutdown          = inet_shutdown,             /* ok           */
 582         .setsockopt        = sock_common_setsockopt,    /* ok           */
 583         .getsockopt        = sock_common_getsockopt,    /* ok           */
 584         .sendmsg           = inet_sendmsg,              /* ok           */
 585         .recvmsg           = inet_recvmsg,              /* ok           */
 586         .mmap              = sock_no_mmap,
 587         .sendpage          = inet_sendpage,
 588         .sendmsg_locked    = tcp_sendmsg_locked,
 589         .sendpage_locked   = tcp_sendpage_locked,
 590         .splice_read       = tcp_splice_read,
 591         .read_sock         = tcp_read_sock,
 592         .peek_len          = tcp_peek_len,
 593 #ifdef CONFIG_COMPAT
 594         .compat_setsockopt = compat_sock_common_setsockopt,
 595         .compat_getsockopt = compat_sock_common_getsockopt,
 596 #endif
 597 };
 598
 599 const struct proto_ops inet6_dgram_ops = {
 600         .family            = PF_INET6,
 601         .owner             = THIS_MODULE,
 602         .release           = inet6_release,
 603         .bind              = inet6_bind,
 604         .connect           = inet_dgram_connect,        /* ok           */
 605         .socketpair        = sock_no_socketpair,        /* a do nothing */
 606         .accept            = sock_no_accept,            /* a do nothing */
 607         .getname           = inet6_getname,
 608         .poll              = udp_poll,                  /* ok           */
 609         .ioctl             = inet6_ioctl,               /* must change  */
 610         .listen            = sock_no_listen,            /* ok           */
 611         .shutdown          = inet_shutdown,             /* ok           */
 612         .setsockopt        = sock_common_setsockopt,    /* ok           */
 613         .getsockopt        = sock_common_getsockopt,    /* ok           */
 614         .sendmsg           = inet_sendmsg,              /* ok           */
 615         .recvmsg           = inet_recvmsg,              /* ok           */
 616         .mmap              = sock_no_mmap,
 617         .sendpage          = sock_no_sendpage,
 618         .set_peek_off      = sk_set_peek_off,
 619 #ifdef CONFIG_COMPAT
 620         .compat_setsockopt = compat_sock_common_setsockopt,
 621         .compat_getsockopt = compat_sock_common_getsockopt,
 622 #endif
 623 };
 624
 625 static const struct net_proto_family inet6_family_ops = {
 626         .family = PF_INET6,
 627         .create = inet6_create,
 628         .owner  = THIS_MODULE,
 629 };
 630
 631 int inet6_register_protosw(struct inet_protosw *p)
 632 {
 633         struct list_head *lh;
 634         struct inet_protosw *answer;
 635         struct list_head *last_perm;
 636         int protocol = p->protocol;
 637         int ret;
 638
 639         spin_lock_bh(&inetsw6_lock);
 640
 641         ret = -EINVAL;
 642         if (p->type >= SOCK_MAX)
 643                 goto out_illegal;
 644
 645         /* If we are trying to override a permanent protocol, bail. */
 646         answer = NULL;
 647         ret = -EPERM;
 648         last_perm = &inetsw6[p->type];
 649         list_for_each(lh, &inetsw6[p->type]) {
 650                 answer = list_entry(lh, struct inet_protosw, list);
 651
 652                 /* Check only the non-wild match. */
 653                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 654                         if (protocol == answer->protocol)
 655                                 break;
 656                         last_perm = lh;
 657                 }
 658
 659                 answer = NULL;
 660         }
 661         if (answer)
 662                 goto out_permanent;
 663
 664         /* Add the new entry after the last permanent entry if any, so that
 665          * the new entry does not override a permanent entry when matched with
 666          * a wild-card protocol. But it is allowed to override any existing
 667          * non-permanent entry.  This means that when we remove this entry, the
 668          * system automatically returns to the old behavior.
 669          */
 670         list_add_rcu(&p->list, last_perm);
 671         ret = 0;
 672 out:
 673         spin_unlock_bh(&inetsw6_lock);
 674         return ret;
 675
 676 out_permanent:
 677         pr_err("Attempt to override permanent protocol %d\n", protocol);
 678         goto out;
 679
 680 out_illegal:
 681         pr_err("Ignoring attempt to register invalid socket type %d\n",
 682                p->type);
 683         goto out;
 684 }
 685 EXPORT_SYMBOL(inet6_register_protosw);
 686
 687 void
 688 inet6_unregister_protosw(struct inet_protosw *p)
 689 {
 690         if (INET_PROTOSW_PERMANENT & p->flags) {
 691                 pr_err("Attempt to unregister permanent protocol %d\n",
 692                        p->protocol);
 693         } else {
 694                 spin_lock_bh(&inetsw6_lock);
 695                 list_del_rcu(&p->list);
 696                 spin_unlock_bh(&inetsw6_lock);
 697
 698                 synchronize_net();
 699         }
 700 }
 701 EXPORT_SYMBOL(inet6_unregister_protosw);
 702
 703 int inet6_sk_rebuild_header(struct sock *sk)
 704 {
 705         struct ipv6_pinfo *np = inet6_sk(sk);
 706         struct dst_entry *dst;
 707
 708         dst = __sk_dst_check(sk, np->dst_cookie);
 709
 710         if (!dst) {
 711                 struct inet_sock *inet = inet_sk(sk);
 712                 struct in6_addr *final_p, final;
 713                 struct flowi6 fl6;
 714
 715                 memset(&fl6, 0, sizeof(fl6));
 716                 fl6.flowi6_proto = sk->sk_protocol;
 717                 fl6.daddr = sk->sk_v6_daddr;
 718                 fl6.saddr = np->saddr;
 719                 fl6.flowlabel = np->flow_label;
 720                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 721                 fl6.flowi6_mark = sk->sk_mark;
 722                 fl6.fl6_dport = inet->inet_dport;
 723                 fl6.fl6_sport = inet->inet_sport;
 724                 fl6.flowi6_uid = sk->sk_uid;
 725                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 726
 727                 rcu_read_lock();
 728                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
 729                                          &final);
 730                 rcu_read_unlock();
 731
 732                 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
 733                 if (IS_ERR(dst)) {
 734                         sk->sk_route_caps = 0;
 735                         sk->sk_err_soft = -PTR_ERR(dst);
 736                         return PTR_ERR(dst);
 737                 }
 738
 739                 ip6_dst_store(sk, dst, NULL, NULL);
 740         }
 741
 742         return 0;
 743 }
 744 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 745
 746 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
 747                        const struct inet6_skb_parm *opt)
 748 {
 749         const struct ipv6_pinfo *np = inet6_sk(sk);
 750
 751         if (np->rxopt.all) {
 752                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
 753                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
 754                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
 755                      np->rxopt.bits.rxflow) ||
 756                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 757                      np->rxopt.bits.osrcrt)) ||
 758                     ((opt->dst1 || opt->dst0) &&
 759                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 760                         return true;
 761         }
 762         return false;
 763 }
 764 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 765
 766 static struct packet_type ipv6_packet_type __read_mostly = {
 767         .type = cpu_to_be16(ETH_P_IPV6),
 768         .func = ipv6_rcv,
 769 };
 770
 771 static int __init ipv6_packet_init(void)
 772 {
 773         dev_add_pack(&ipv6_packet_type);
 774         return 0;
 775 }
 776
 777 static void ipv6_packet_cleanup(void)
 778 {
 779         dev_remove_pack(&ipv6_packet_type);
 780 }
 781
 782 static int __net_init ipv6_init_mibs(struct net *net)
 783 {
 784         int i;
 785
 786         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
 787         if (!net->mib.udp_stats_in6)
 788                 return -ENOMEM;
 789         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
 790         if (!net->mib.udplite_stats_in6)
 791                 goto err_udplite_mib;
 792         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
 793         if (!net->mib.ipv6_statistics)
 794                 goto err_ip_mib;
 795
 796         for_each_possible_cpu(i) {
 797                 struct ipstats_mib *af_inet6_stats;
 798                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
 799                 u64_stats_init(&af_inet6_stats->syncp);
 800         }
 801
 802
 803         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
 804         if (!net->mib.icmpv6_statistics)
 805                 goto err_icmp_mib;
 806         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 807                                                 GFP_KERNEL);
 808         if (!net->mib.icmpv6msg_statistics)
 809                 goto err_icmpmsg_mib;
 810         return 0;
 811
 812 err_icmpmsg_mib:
 813         free_percpu(net->mib.icmpv6_statistics);
 814 err_icmp_mib:
 815         free_percpu(net->mib.ipv6_statistics);
 816 err_ip_mib:
 817         free_percpu(net->mib.udplite_stats_in6);
 818 err_udplite_mib:
 819         free_percpu(net->mib.udp_stats_in6);
 820         return -ENOMEM;
 821 }
 822
 823 static void ipv6_cleanup_mibs(struct net *net)
 824 {
 825         free_percpu(net->mib.udp_stats_in6);
 826         free_percpu(net->mib.udplite_stats_in6);
 827         free_percpu(net->mib.ipv6_statistics);
 828         free_percpu(net->mib.icmpv6_statistics);
 829         kfree(net->mib.icmpv6msg_statistics);
 830 }
 831
 832 static int __net_init inet6_net_init(struct net *net)
 833 {
 834         int err = 0;
 835
 836         net->ipv6.sysctl.bindv6only = 0;
 837         net->ipv6.sysctl.icmpv6_time = 1*HZ;
 838         net->ipv6.sysctl.flowlabel_consistency = 1;
 839         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
 840         net->ipv6.sysctl.idgen_retries = 3;
 841         net->ipv6.sysctl.idgen_delay = 1 * HZ;
 842         net->ipv6.sysctl.flowlabel_state_ranges = 0;
 843         atomic_set(&net->ipv6.fib6_sernum, 1);
 844
 845         err = ipv6_init_mibs(net);
 846         if (err)
 847                 return err;
 848 #ifdef CONFIG_PROC_FS
 849         err = udp6_proc_init(net);
 850         if (err)
 851                 goto out;
 852         err = tcp6_proc_init(net);
 853         if (err)
 854                 goto proc_tcp6_fail;
 855         err = ac6_proc_init(net);
 856         if (err)
 857                 goto proc_ac6_fail;
 858 #endif
 859         return err;
 860
 861 #ifdef CONFIG_PROC_FS
 862 proc_ac6_fail:
 863         tcp6_proc_exit(net);
 864 proc_tcp6_fail:
 865         udp6_proc_exit(net);
 866 out:
 867         ipv6_cleanup_mibs(net);
 868         return err;
 869 #endif
 870 }
 871
 872 static void __net_exit inet6_net_exit(struct net *net)
 873 {
 874 #ifdef CONFIG_PROC_FS
 875         udp6_proc_exit(net);
 876         tcp6_proc_exit(net);
 877         ac6_proc_exit(net);
 878 #endif
 879         ipv6_cleanup_mibs(net);
 880 }
 881
 882 static struct pernet_operations inet6_net_ops = {
 883         .init = inet6_net_init,
 884         .exit = inet6_net_exit,
 885 };
 886
 887 static const struct ipv6_stub ipv6_stub_impl = {
 888         .ipv6_sock_mc_join = ipv6_sock_mc_join,
 889         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
 890         .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
 891         .udpv6_encap_enable = udpv6_encap_enable,
 892         .ndisc_send_na = ndisc_send_na,
 893         .nd_tbl = &nd_tbl,
 894 };
 895
 896 static int __init inet6_init(void)
 897 {
 898         struct list_head *r;
 899         int err = 0;
 900
 901         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
 902
 903         /* Register the socket-side information for inet6_create.  */
 904         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
 905                 INIT_LIST_HEAD(r);
 906
 907         if (disable_ipv6_mod) {
 908                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
 909                 goto out;
 910         }
 911
 912         err = proto_register(&tcpv6_prot, 1);
 913         if (err)
 914                 goto out;
 915
 916         err = proto_register(&udpv6_prot, 1);
 917         if (err)
 918                 goto out_unregister_tcp_proto;
 919
 920         err = proto_register(&udplitev6_prot, 1);
 921         if (err)
 922                 goto out_unregister_udp_proto;
 923
 924         err = proto_register(&rawv6_prot, 1);
 925         if (err)
 926                 goto out_unregister_udplite_proto;
 927
 928         err = proto_register(&pingv6_prot, 1);
 929         if (err)
 930                 goto out_unregister_ping_proto;
 931
 932         /* We MUST register RAW sockets before we create the ICMP6,
 933          * IGMP6, or NDISC control sockets.
 934          */
 935         err = rawv6_init();
 936         if (err)
 937                 goto out_unregister_raw_proto;
 938
 939         /* Register the family here so that the init calls below will
 940          * be able to create sockets. (?? is this dangerous ??)
 941          */
 942         err = sock_register(&inet6_family_ops);
 943         if (err)
 944                 goto out_sock_register_fail;
 945
 946         /*
 947          *      ipngwg API draft makes clear that the correct semantics
 948          *      for TCP and UDP is to consider one TCP and UDP instance
 949          *      in a host available by both INET and INET6 APIs and
 950          *      able to communicate via both network protocols.
 951          */
 952
 953         err = register_pernet_subsys(&inet6_net_ops);
 954         if (err)
 955                 goto register_pernet_fail;
 956         err = ip6_mr_init();
 957         if (err)
 958                 goto ipmr_fail;
 959         err = icmpv6_init();
 960         if (err)
 961                 goto icmp_fail;
 962         err = ndisc_init();
 963         if (err)
 964                 goto ndisc_fail;
 965         err = igmp6_init();
 966         if (err)
 967                 goto igmp_fail;
 968
 969         err = ipv6_netfilter_init();
 970         if (err)
 971                 goto netfilter_fail;
 972         /* Create /proc/foo6 entries. */
 973 #ifdef CONFIG_PROC_FS
 974         err = -ENOMEM;
 975         if (raw6_proc_init())
 976                 goto proc_raw6_fail;
 977         if (udplite6_proc_init())
 978                 goto proc_udplite6_fail;
 979         if (ipv6_misc_proc_init())
 980                 goto proc_misc6_fail;
 981         if (if6_proc_init())
 982                 goto proc_if6_fail;
 983 #endif
 984         err = ip6_route_init();
 985         if (err)
 986                 goto ip6_route_fail;
 987         err = ndisc_late_init();
 988         if (err)
 989                 goto ndisc_late_fail;
 990         err = ip6_flowlabel_init();
 991         if (err)
 992                 goto ip6_flowlabel_fail;
 993         err = addrconf_init();
 994         if (err)
 995                 goto addrconf_fail;
 996
 997         /* Init v6 extension headers. */
 998         err = ipv6_exthdrs_init();
 999         if (err)
1000                 goto ipv6_exthdrs_fail;
1001
1002         err = ipv6_frag_init();
1003         if (err)
1004                 goto ipv6_frag_fail;
1005
1006         /* Init v6 transport protocols. */
1007         err = udpv6_init();
1008         if (err)
1009                 goto udpv6_fail;
1010
1011         err = udplitev6_init();
1012         if (err)
1013                 goto udplitev6_fail;
1014
1015         err = udpv6_offload_init();
1016         if (err)
1017                 goto udpv6_offload_fail;
1018
1019         err = tcpv6_init();
1020         if (err)
1021                 goto tcpv6_fail;
1022
1023         err = ipv6_packet_init();
1024         if (err)
1025                 goto ipv6_packet_fail;
1026
1027         err = pingv6_init();
1028         if (err)
1029                 goto pingv6_fail;
1030
1031         err = calipso_init();
1032         if (err)
1033                 goto calipso_fail;
1034
1035         err = seg6_init();
1036         if (err)
1037                 goto seg6_fail;
1038
1039         err = igmp6_late_init();
1040         if (err)
1041                 goto igmp6_late_err;
1042
1043 #ifdef CONFIG_SYSCTL
1044         err = ipv6_sysctl_register();
1045         if (err)
1046                 goto sysctl_fail;
1047 #endif
1048
1049         /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1050         wmb();
1051         ipv6_stub = &ipv6_stub_impl;
1052 out:
1053         return err;
1054
1055 #ifdef CONFIG_SYSCTL
1056 sysctl_fail:
1057         igmp6_late_cleanup();
1058 #endif
1059 igmp6_late_err:
1060         seg6_exit();
1061 seg6_fail:
1062         calipso_exit();
1063 calipso_fail:
1064         pingv6_exit();
1065 pingv6_fail:
1066         ipv6_packet_cleanup();
1067 ipv6_packet_fail:
1068         tcpv6_exit();
1069 tcpv6_fail:
1070         udpv6_offload_exit();
1071 udpv6_offload_fail:
1072         udplitev6_exit();
1073 udplitev6_fail:
1074         udpv6_exit();
1075 udpv6_fail:
1076         ipv6_frag_exit();
1077 ipv6_frag_fail:
1078         ipv6_exthdrs_exit();
1079 ipv6_exthdrs_fail:
1080         addrconf_cleanup();
1081 addrconf_fail:
1082         ip6_flowlabel_cleanup();
1083 ip6_flowlabel_fail:
1084         ndisc_late_cleanup();
1085 ndisc_late_fail:
1086         ip6_route_cleanup();
1087 ip6_route_fail:
1088 #ifdef CONFIG_PROC_FS
1089         if6_proc_exit();
1090 proc_if6_fail:
1091         ipv6_misc_proc_exit();
1092 proc_misc6_fail:
1093         udplite6_proc_exit();
1094 proc_udplite6_fail:
1095         raw6_proc_exit();
1096 proc_raw6_fail:
1097 #endif
1098         ipv6_netfilter_fini();
1099 netfilter_fail:
1100         igmp6_cleanup();
1101 igmp_fail:
1102         ndisc_cleanup();
1103 ndisc_fail:
1104         icmpv6_cleanup();
1105 icmp_fail:
1106         ip6_mr_cleanup();
1107 ipmr_fail:
1108         unregister_pernet_subsys(&inet6_net_ops);
1109 register_pernet_fail:
1110         sock_unregister(PF_INET6);
1111         rtnl_unregister_all(PF_INET6);
1112 out_sock_register_fail:
1113         rawv6_exit();
1114 out_unregister_ping_proto:
1115         proto_unregister(&pingv6_prot);
1116 out_unregister_raw_proto:
1117         proto_unregister(&rawv6_prot);
1118 out_unregister_udplite_proto:
1119         proto_unregister(&udplitev6_prot);
1120 out_unregister_udp_proto:
1121         proto_unregister(&udpv6_prot);
1122 out_unregister_tcp_proto:
1123         proto_unregister(&tcpv6_prot);
1124         goto out;
1125 }
1126 module_init(inet6_init);
1127
1128 MODULE_ALIAS_NETPROTO(PF_INET6);