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/ipip.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 #ifdef CONFIG_IPV6_TUNNEL
  60 #include <net/ip6_tunnel.h>
  61 #endif
  62
  63 #include <asm/uaccess.h>
  64 #include <linux/mroute6.h>
  65
  66 MODULE_AUTHOR("Cast of dozens");
  67 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  68 MODULE_LICENSE("GPL");
  69
  70 /* The inetsw6 table contains everything that inet6_create needs to
  71  * build a new socket.
  72  */
  73 static struct list_head inetsw6[SOCK_MAX];
  74 static DEFINE_SPINLOCK(inetsw6_lock);
  75
  76 struct ipv6_params ipv6_defaults = {
  77         .disable_ipv6 = 0,
  78         .autoconf = 1,
  79 };
  80
  81 static int disable_ipv6_mod;
  82
  83 module_param_named(disable, disable_ipv6_mod, int, 0444);
  84 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
  85
  86 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
  87 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
  88
  89 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
  90 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
  91
  92 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
  93 {
  94         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
  95
  96         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
  97 }
  98
  99 static int inet6_create(struct net *net, struct socket *sock, int protocol,
 100                         int kern)
 101 {
 102         struct inet_sock *inet;
 103         struct ipv6_pinfo *np;
 104         struct sock *sk;
 105         struct inet_protosw *answer;
 106         struct proto *answer_prot;
 107         unsigned char answer_flags;
 108         char answer_no_check;
 109         int try_loading_module = 0;
 110         int err;
 111
 112         if (sock->type != SOCK_RAW &&
 113             sock->type != SOCK_DGRAM &&
 114             !inet_ehash_secret)
 115                 build_ehash_secret();
 116
 117         /* Look for the requested type/protocol pair. */
 118 lookup_protocol:
 119         err = -ESOCKTNOSUPPORT;
 120         rcu_read_lock();
 121         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 122
 123                 err = 0;
 124                 /* Check the non-wild match. */
 125                 if (protocol == answer->protocol) {
 126                         if (protocol != IPPROTO_IP)
 127                                 break;
 128                 } else {
 129                         /* Check for the two wild cases. */
 130                         if (IPPROTO_IP == protocol) {
 131                                 protocol = answer->protocol;
 132                                 break;
 133                         }
 134                         if (IPPROTO_IP == answer->protocol)
 135                                 break;
 136                 }
 137                 err = -EPROTONOSUPPORT;
 138         }
 139
 140         if (err) {
 141                 if (try_loading_module < 2) {
 142                         rcu_read_unlock();
 143                         /*
 144                          * Be more specific, e.g. net-pf-10-proto-132-type-1
 145                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
 146                          */
 147                         if (++try_loading_module == 1)
 148                                 request_module("net-pf-%d-proto-%d-type-%d",
 149                                                 PF_INET6, protocol, sock->type);
 150                         /*
 151                          * Fall back to generic, e.g. net-pf-10-proto-132
 152                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
 153                          */
 154                         else
 155                                 request_module("net-pf-%d-proto-%d",
 156                                                 PF_INET6, protocol);
 157                         goto lookup_protocol;
 158                 } else
 159                         goto out_rcu_unlock;
 160         }
 161
 162         err = -EPERM;
 163         if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
 164                 goto out_rcu_unlock;
 165
 166         sock->ops = answer->ops;
 167         answer_prot = answer->prot;
 168         answer_no_check = answer->no_check;
 169         answer_flags = answer->flags;
 170         rcu_read_unlock();
 171
 172         WARN_ON(answer_prot->slab == NULL);
 173
 174         err = -ENOBUFS;
 175         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
 176         if (sk == NULL)
 177                 goto out;
 178
 179         sock_init_data(sock, sk);
 180
 181         err = 0;
 182         sk->sk_no_check = answer_no_check;
 183         if (INET_PROTOSW_REUSE & answer_flags)
 184                 sk->sk_reuse = SK_CAN_REUSE;
 185
 186         inet = inet_sk(sk);
 187         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 188
 189         if (SOCK_RAW == sock->type) {
 190                 inet->inet_num = protocol;
 191                 if (IPPROTO_RAW == protocol)
 192                         inet->hdrincl = 1;
 193         }
 194
 195         sk->sk_destruct         = inet_sock_destruct;
 196         sk->sk_family           = PF_INET6;
 197         sk->sk_protocol         = protocol;
 198
 199         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
 200
 201         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
 202         np->hop_limit   = -1;
 203         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
 204         np->mc_loop     = 1;
 205         np->pmtudisc    = IPV6_PMTUDISC_WANT;
 206         np->ipv6only    = net->ipv6.sysctl.bindv6only;
 207
 208         /* Init the ipv4 part of the socket since we can have sockets
 209          * using v6 API for ipv4.
 210          */
 211         inet->uc_ttl    = -1;
 212
 213         inet->mc_loop   = 1;
 214         inet->mc_ttl    = 1;
 215         inet->mc_index  = 0;
 216         inet->mc_list   = NULL;
 217         inet->rcv_tos   = 0;
 218
 219         if (ipv4_config.no_pmtu_disc)
 220                 inet->pmtudisc = IP_PMTUDISC_DONT;
 221         else
 222                 inet->pmtudisc = IP_PMTUDISC_WANT;
 223         /*
 224          * Increment only the relevant sk_prot->socks debug field, this changes
 225          * the previous behaviour of incrementing both the equivalent to
 226          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
 227          *
 228          * This allows better debug granularity as we'll know exactly how many
 229          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
 230          * transport protocol socks. -acme
 231          */
 232         sk_refcnt_debug_inc(sk);
 233
 234         if (inet->inet_num) {
 235                 /* It assumes that any protocol which allows
 236                  * the user to assign a number at socket
 237                  * creation time automatically shares.
 238                  */
 239                 inet->inet_sport = htons(inet->inet_num);
 240                 sk->sk_prot->hash(sk);
 241         }
 242         if (sk->sk_prot->init) {
 243                 err = sk->sk_prot->init(sk);
 244                 if (err) {
 245                         sk_common_release(sk);
 246                         goto out;
 247                 }
 248         }
 249 out:
 250         return err;
 251 out_rcu_unlock:
 252         rcu_read_unlock();
 253         goto out;
 254 }
 255
 256
 257 /* bind for INET6 API */
 258 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 259 {
 260         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
 261         struct sock *sk = sock->sk;
 262         struct inet_sock *inet = inet_sk(sk);
 263         struct ipv6_pinfo *np = inet6_sk(sk);
 264         struct net *net = sock_net(sk);
 265         __be32 v4addr = 0;
 266         unsigned short snum;
 267         int addr_type = 0;
 268         int err = 0;
 269
 270         /* If the socket has its own bind function then use it. */
 271         if (sk->sk_prot->bind)
 272                 return sk->sk_prot->bind(sk, uaddr, addr_len);
 273
 274         if (addr_len < SIN6_LEN_RFC2133)
 275                 return -EINVAL;
 276
 277         if (addr->sin6_family != AF_INET6)
 278                 return -EAFNOSUPPORT;
 279
 280         addr_type = ipv6_addr_type(&addr->sin6_addr);
 281         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 282                 return -EINVAL;
 283
 284         snum = ntohs(addr->sin6_port);
 285         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 286                 return -EACCES;
 287
 288         lock_sock(sk);
 289
 290         /* Check these errors (active socket, double bind). */
 291         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
 292                 err = -EINVAL;
 293                 goto out;
 294         }
 295
 296         /* Check if the address belongs to the host. */
 297         if (addr_type == IPV6_ADDR_MAPPED) {
 298                 int chk_addr_ret;
 299
 300                 /* Binding to v4-mapped address on a v6-only socket
 301                  * makes no sense
 302                  */
 303                 if (np->ipv6only) {
 304                         err = -EINVAL;
 305                         goto out;
 306                 }
 307
 308                 /* Reproduce AF_INET checks to make the bindings consistent */
 309                 v4addr = addr->sin6_addr.s6_addr32[3];
 310                 chk_addr_ret = inet_addr_type(net, v4addr);
 311                 if (!sysctl_ip_nonlocal_bind &&
 312                     !(inet->freebind || inet->transparent) &&
 313                     v4addr != htonl(INADDR_ANY) &&
 314                     chk_addr_ret != RTN_LOCAL &&
 315                     chk_addr_ret != RTN_MULTICAST &&
 316                     chk_addr_ret != RTN_BROADCAST) {
 317                         err = -EADDRNOTAVAIL;
 318                         goto out;
 319                 }
 320         } else {
 321                 if (addr_type != IPV6_ADDR_ANY) {
 322                         struct net_device *dev = NULL;
 323
 324                         rcu_read_lock();
 325                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 326                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 327                                     addr->sin6_scope_id) {
 328                                         /* Override any existing binding, if another one
 329                                          * is supplied by user.
 330                                          */
 331                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 332                                 }
 333
 334                                 /* Binding to link-local address requires an interface */
 335                                 if (!sk->sk_bound_dev_if) {
 336                                         err = -EINVAL;
 337                                         goto out_unlock;
 338                                 }
 339                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 340                                 if (!dev) {
 341                                         err = -ENODEV;
 342                                         goto out_unlock;
 343                                 }
 344                         }
 345
 346                         /* ipv4 addr of the socket is invalid.  Only the
 347                          * unspecified and mapped address have a v4 equivalent.
 348                          */
 349                         v4addr = LOOPBACK4_IPV6;
 350                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 351                                 if (!(inet->freebind || inet->transparent) &&
 352                                     !ipv6_chk_addr(net, &addr->sin6_addr,
 353                                                    dev, 0)) {
 354                                         err = -EADDRNOTAVAIL;
 355                                         goto out_unlock;
 356                                 }
 357                         }
 358                         rcu_read_unlock();
 359                 }
 360         }
 361
 362         inet->inet_rcv_saddr = v4addr;
 363         inet->inet_saddr = v4addr;
 364
 365         np->rcv_saddr = addr->sin6_addr;
 366
 367         if (!(addr_type & IPV6_ADDR_MULTICAST))
 368                 np->saddr = addr->sin6_addr;
 369
 370         /* Make sure we are allowed to bind here. */
 371         if (sk->sk_prot->get_port(sk, snum)) {
 372                 inet_reset_saddr(sk);
 373                 err = -EADDRINUSE;
 374                 goto out;
 375         }
 376
 377         if (addr_type != IPV6_ADDR_ANY) {
 378                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 379                 if (addr_type != IPV6_ADDR_MAPPED)
 380                         np->ipv6only = 1;
 381         }
 382         if (snum)
 383                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 384         inet->inet_sport = htons(inet->inet_num);
 385         inet->inet_dport = 0;
 386         inet->inet_daddr = 0;
 387 out:
 388         release_sock(sk);
 389         return err;
 390 out_unlock:
 391         rcu_read_unlock();
 392         goto out;
 393 }
 394 EXPORT_SYMBOL(inet6_bind);
 395
 396 int inet6_release(struct socket *sock)
 397 {
 398         struct sock *sk = sock->sk;
 399
 400         if (sk == NULL)
 401                 return -EINVAL;
 402
 403         /* Free mc lists */
 404         ipv6_sock_mc_close(sk);
 405
 406         /* Free ac lists */
 407         ipv6_sock_ac_close(sk);
 408
 409         return inet_release(sock);
 410 }
 411 EXPORT_SYMBOL(inet6_release);
 412
 413 void inet6_destroy_sock(struct sock *sk)
 414 {
 415         struct ipv6_pinfo *np = inet6_sk(sk);
 416         struct sk_buff *skb;
 417         struct ipv6_txoptions *opt;
 418
 419         /* Release rx options */
 420
 421         skb = xchg(&np->pktoptions, NULL);
 422         if (skb != NULL)
 423                 kfree_skb(skb);
 424
 425         skb = xchg(&np->rxpmtu, NULL);
 426         if (skb != NULL)
 427                 kfree_skb(skb);
 428
 429         /* Free flowlabels */
 430         fl6_free_socklist(sk);
 431
 432         /* Free tx options */
 433
 434         opt = xchg(&np->opt, NULL);
 435         if (opt != NULL)
 436                 sock_kfree_s(sk, opt, opt->tot_len);
 437 }
 438 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
 439
 440 /*
 441  *      This does both peername and sockname.
 442  */
 443
 444 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 445                  int *uaddr_len, int peer)
 446 {
 447         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
 448         struct sock *sk = sock->sk;
 449         struct inet_sock *inet = inet_sk(sk);
 450         struct ipv6_pinfo *np = inet6_sk(sk);
 451
 452         sin->sin6_family = AF_INET6;
 453         sin->sin6_flowinfo = 0;
 454         sin->sin6_scope_id = 0;
 455         if (peer) {
 456                 if (!inet->inet_dport)
 457                         return -ENOTCONN;
 458                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
 459                     peer == 1)
 460                         return -ENOTCONN;
 461                 sin->sin6_port = inet->inet_dport;
 462                 sin->sin6_addr = np->daddr;
 463                 if (np->sndflow)
 464                         sin->sin6_flowinfo = np->flow_label;
 465         } else {
 466                 if (ipv6_addr_any(&np->rcv_saddr))
 467                         sin->sin6_addr = np->saddr;
 468                 else
 469                         sin->sin6_addr = np->rcv_saddr;
 470
 471                 sin->sin6_port = inet->inet_sport;
 472         }
 473         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 474                 sin->sin6_scope_id = sk->sk_bound_dev_if;
 475         *uaddr_len = sizeof(*sin);
 476         return 0;
 477 }
 478 EXPORT_SYMBOL(inet6_getname);
 479
 480 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 481 {
 482         struct sock *sk = sock->sk;
 483         struct net *net = sock_net(sk);
 484
 485         switch (cmd) {
 486         case SIOCGSTAMP:
 487                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
 488
 489         case SIOCGSTAMPNS:
 490                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
 491
 492         case SIOCADDRT:
 493         case SIOCDELRT:
 494
 495                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
 496
 497         case SIOCSIFADDR:
 498                 return addrconf_add_ifaddr(net, (void __user *) arg);
 499         case SIOCDIFADDR:
 500                 return addrconf_del_ifaddr(net, (void __user *) arg);
 501         case SIOCSIFDSTADDR:
 502                 return addrconf_set_dstaddr(net, (void __user *) arg);
 503         default:
 504                 if (!sk->sk_prot->ioctl)
 505                         return -ENOIOCTLCMD;
 506                 return sk->sk_prot->ioctl(sk, cmd, arg);
 507         }
 508         /*NOTREACHED*/
 509         return 0;
 510 }
 511 EXPORT_SYMBOL(inet6_ioctl);
 512
 513 const struct proto_ops inet6_stream_ops = {
 514         .family            = PF_INET6,
 515         .owner             = THIS_MODULE,
 516         .release           = inet6_release,
 517         .bind              = inet6_bind,
 518         .connect           = inet_stream_connect,       /* ok           */
 519         .socketpair        = sock_no_socketpair,        /* a do nothing */
 520         .accept            = inet_accept,               /* ok           */
 521         .getname           = inet6_getname,
 522         .poll              = tcp_poll,                  /* ok           */
 523         .ioctl             = inet6_ioctl,               /* must change  */
 524         .listen            = inet_listen,               /* ok           */
 525         .shutdown          = inet_shutdown,             /* ok           */
 526         .setsockopt        = sock_common_setsockopt,    /* ok           */
 527         .getsockopt        = sock_common_getsockopt,    /* ok           */
 528         .sendmsg           = inet_sendmsg,              /* ok           */
 529         .recvmsg           = inet_recvmsg,              /* ok           */
 530         .mmap              = sock_no_mmap,
 531         .sendpage          = inet_sendpage,
 532         .splice_read       = tcp_splice_read,
 533 #ifdef CONFIG_COMPAT
 534         .compat_setsockopt = compat_sock_common_setsockopt,
 535         .compat_getsockopt = compat_sock_common_getsockopt,
 536 #endif
 537 };
 538
 539 const struct proto_ops inet6_dgram_ops = {
 540         .family            = PF_INET6,
 541         .owner             = THIS_MODULE,
 542         .release           = inet6_release,
 543         .bind              = inet6_bind,
 544         .connect           = inet_dgram_connect,        /* ok           */
 545         .socketpair        = sock_no_socketpair,        /* a do nothing */
 546         .accept            = sock_no_accept,            /* a do nothing */
 547         .getname           = inet6_getname,
 548         .poll              = udp_poll,                  /* ok           */
 549         .ioctl             = inet6_ioctl,               /* must change  */
 550         .listen            = sock_no_listen,            /* ok           */
 551         .shutdown          = inet_shutdown,             /* ok           */
 552         .setsockopt        = sock_common_setsockopt,    /* ok           */
 553         .getsockopt        = sock_common_getsockopt,    /* ok           */
 554         .sendmsg           = inet_sendmsg,              /* ok           */
 555         .recvmsg           = inet_recvmsg,              /* ok           */
 556         .mmap              = sock_no_mmap,
 557         .sendpage          = sock_no_sendpage,
 558 #ifdef CONFIG_COMPAT
 559         .compat_setsockopt = compat_sock_common_setsockopt,
 560         .compat_getsockopt = compat_sock_common_getsockopt,
 561 #endif
 562 };
 563
 564 static const struct net_proto_family inet6_family_ops = {
 565         .family = PF_INET6,
 566         .create = inet6_create,
 567         .owner  = THIS_MODULE,
 568 };
 569
 570 int inet6_register_protosw(struct inet_protosw *p)
 571 {
 572         struct list_head *lh;
 573         struct inet_protosw *answer;
 574         struct list_head *last_perm;
 575         int protocol = p->protocol;
 576         int ret;
 577
 578         spin_lock_bh(&inetsw6_lock);
 579
 580         ret = -EINVAL;
 581         if (p->type >= SOCK_MAX)
 582                 goto out_illegal;
 583
 584         /* If we are trying to override a permanent protocol, bail. */
 585         answer = NULL;
 586         ret = -EPERM;
 587         last_perm = &inetsw6[p->type];
 588         list_for_each(lh, &inetsw6[p->type]) {
 589                 answer = list_entry(lh, struct inet_protosw, list);
 590
 591                 /* Check only the non-wild match. */
 592                 if (INET_PROTOSW_PERMANENT & answer->flags) {
 593                         if (protocol == answer->protocol)
 594                                 break;
 595                         last_perm = lh;
 596                 }
 597
 598                 answer = NULL;
 599         }
 600         if (answer)
 601                 goto out_permanent;
 602
 603         /* Add the new entry after the last permanent entry if any, so that
 604          * the new entry does not override a permanent entry when matched with
 605          * a wild-card protocol. But it is allowed to override any existing
 606          * non-permanent entry.  This means that when we remove this entry, the
 607          * system automatically returns to the old behavior.
 608          */
 609         list_add_rcu(&p->list, last_perm);
 610         ret = 0;
 611 out:
 612         spin_unlock_bh(&inetsw6_lock);
 613         return ret;
 614
 615 out_permanent:
 616         pr_err("Attempt to override permanent protocol %d\n", protocol);
 617         goto out;
 618
 619 out_illegal:
 620         pr_err("Ignoring attempt to register invalid socket type %d\n",
 621                p->type);
 622         goto out;
 623 }
 624 EXPORT_SYMBOL(inet6_register_protosw);
 625
 626 void
 627 inet6_unregister_protosw(struct inet_protosw *p)
 628 {
 629         if (INET_PROTOSW_PERMANENT & p->flags) {
 630                 pr_err("Attempt to unregister permanent protocol %d\n",
 631                        p->protocol);
 632         } else {
 633                 spin_lock_bh(&inetsw6_lock);
 634                 list_del_rcu(&p->list);
 635                 spin_unlock_bh(&inetsw6_lock);
 636
 637                 synchronize_net();
 638         }
 639 }
 640 EXPORT_SYMBOL(inet6_unregister_protosw);
 641
 642 int inet6_sk_rebuild_header(struct sock *sk)
 643 {
 644         struct ipv6_pinfo *np = inet6_sk(sk);
 645         struct dst_entry *dst;
 646
 647         dst = __sk_dst_check(sk, np->dst_cookie);
 648
 649         if (dst == NULL) {
 650                 struct inet_sock *inet = inet_sk(sk);
 651                 struct in6_addr *final_p, final;
 652                 struct flowi6 fl6;
 653
 654                 memset(&fl6, 0, sizeof(fl6));
 655                 fl6.flowi6_proto = sk->sk_protocol;
 656                 fl6.daddr = np->daddr;
 657                 fl6.saddr = np->saddr;
 658                 fl6.flowlabel = np->flow_label;
 659                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 660                 fl6.flowi6_mark = sk->sk_mark;
 661                 fl6.fl6_dport = inet->inet_dport;
 662                 fl6.fl6_sport = inet->inet_sport;
 663                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 664
 665                 final_p = fl6_update_dst(&fl6, np->opt, &final);
 666
 667                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
 668                 if (IS_ERR(dst)) {
 669                         sk->sk_route_caps = 0;
 670                         sk->sk_err_soft = -PTR_ERR(dst);
 671                         return PTR_ERR(dst);
 672                 }
 673
 674                 __ip6_dst_store(sk, dst, NULL, NULL);
 675         }
 676
 677         return 0;
 678 }
 679 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
 680
 681 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
 682 {
 683         const struct ipv6_pinfo *np = inet6_sk(sk);
 684         const struct inet6_skb_parm *opt = IP6CB(skb);
 685
 686         if (np->rxopt.all) {
 687                 if ((opt->hop && (np->rxopt.bits.hopopts ||
 688                                   np->rxopt.bits.ohopopts)) ||
 689                     ((IPV6_FLOWINFO_MASK &
 690                       *(__be32 *)skb_network_header(skb)) &&
 691                      np->rxopt.bits.rxflow) ||
 692                     (opt->srcrt && (np->rxopt.bits.srcrt ||
 693                      np->rxopt.bits.osrcrt)) ||
 694                     ((opt->dst1 || opt->dst0) &&
 695                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
 696                         return true;
 697         }
 698         return false;
 699 }
 700 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 701
 702 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
 703 {
 704         const struct inet6_protocol *ops = NULL;
 705
 706         for (;;) {
 707                 struct ipv6_opt_hdr *opth;
 708                 int len;
 709
 710                 if (proto != NEXTHDR_HOP) {
 711                         ops = rcu_dereference(inet6_protos[proto]);
 712
 713                         if (unlikely(!ops))
 714                                 break;
 715
 716                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
 717                                 break;
 718                 }
 719
 720                 if (unlikely(!pskb_may_pull(skb, 8)))
 721                         break;
 722
 723                 opth = (void *)skb->data;
 724                 len = ipv6_optlen(opth);
 725
 726                 if (unlikely(!pskb_may_pull(skb, len)))
 727                         break;
 728
 729                 proto = opth->nexthdr;
 730                 __skb_pull(skb, len);
 731         }
 732
 733         return proto;
 734 }
 735
 736 static int ipv6_gso_send_check(struct sk_buff *skb)
 737 {
 738         const struct ipv6hdr *ipv6h;
 739         const struct inet6_protocol *ops;
 740         int err = -EINVAL;
 741
 742         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 743                 goto out;
 744
 745         ipv6h = ipv6_hdr(skb);
 746         __skb_pull(skb, sizeof(*ipv6h));
 747         err = -EPROTONOSUPPORT;
 748
 749         rcu_read_lock();
 750         ops = rcu_dereference(inet6_protos[
 751                 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
 752
 753         if (likely(ops && ops->gso_send_check)) {
 754                 skb_reset_transport_header(skb);
 755                 err = ops->gso_send_check(skb);
 756         }
 757         rcu_read_unlock();
 758
 759 out:
 760         return err;
 761 }
 762
 763 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb,
 764         netdev_features_t features)
 765 {
 766         struct sk_buff *segs = ERR_PTR(-EINVAL);
 767         struct ipv6hdr *ipv6h;
 768         const struct inet6_protocol *ops;
 769         int proto;
 770         struct frag_hdr *fptr;
 771         unsigned int unfrag_ip6hlen;
 772         u8 *prevhdr;
 773         int offset = 0;
 774
 775         if (!(features & NETIF_F_V6_CSUM))
 776                 features &= ~NETIF_F_SG;
 777
 778         if (unlikely(skb_shinfo(skb)->gso_type &
 779                      ~(SKB_GSO_UDP |
 780                        SKB_GSO_DODGY |
 781                        SKB_GSO_TCP_ECN |
 782                        SKB_GSO_TCPV6 |
 783                        0)))
 784                 goto out;
 785
 786         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
 787                 goto out;
 788
 789         ipv6h = ipv6_hdr(skb);
 790         __skb_pull(skb, sizeof(*ipv6h));
 791         segs = ERR_PTR(-EPROTONOSUPPORT);
 792
 793         proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
 794         rcu_read_lock();
 795         ops = rcu_dereference(inet6_protos[proto]);
 796         if (likely(ops && ops->gso_segment)) {
 797                 skb_reset_transport_header(skb);
 798                 segs = ops->gso_segment(skb, features);
 799         }
 800         rcu_read_unlock();
 801
 802         if (IS_ERR(segs))
 803                 goto out;
 804
 805         for (skb = segs; skb; skb = skb->next) {
 806                 ipv6h = ipv6_hdr(skb);
 807                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
 808                                            sizeof(*ipv6h));
 809                 if (proto == IPPROTO_UDP) {
 810                         unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
 811                         fptr = (struct frag_hdr *)(skb_network_header(skb) +
 812                                 unfrag_ip6hlen);
 813                         fptr->frag_off = htons(offset);
 814                         if (skb->next != NULL)
 815                                 fptr->frag_off |= htons(IP6_MF);
 816                         offset += (ntohs(ipv6h->payload_len) -
 817                                    sizeof(struct frag_hdr));
 818                 }
 819         }
 820
 821 out:
 822         return segs;
 823 }
 824
 825 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
 826                                          struct sk_buff *skb)
 827 {
 828         const struct inet6_protocol *ops;
 829         struct sk_buff **pp = NULL;
 830         struct sk_buff *p;
 831         struct ipv6hdr *iph;
 832         unsigned int nlen;
 833         unsigned int hlen;
 834         unsigned int off;
 835         int flush = 1;
 836         int proto;
 837         __wsum csum;
 838
 839         off = skb_gro_offset(skb);
 840         hlen = off + sizeof(*iph);
 841         iph = skb_gro_header_fast(skb, off);
 842         if (skb_gro_header_hard(skb, hlen)) {
 843                 iph = skb_gro_header_slow(skb, hlen, off);
 844                 if (unlikely(!iph))
 845                         goto out;
 846         }
 847
 848         skb_gro_pull(skb, sizeof(*iph));
 849         skb_set_transport_header(skb, skb_gro_offset(skb));
 850
 851         flush += ntohs(iph->payload_len) != skb_gro_len(skb);
 852
 853         rcu_read_lock();
 854         proto = iph->nexthdr;
 855         ops = rcu_dereference(inet6_protos[proto]);
 856         if (!ops || !ops->gro_receive) {
 857                 __pskb_pull(skb, skb_gro_offset(skb));
 858                 proto = ipv6_gso_pull_exthdrs(skb, proto);
 859                 skb_gro_pull(skb, -skb_transport_offset(skb));
 860                 skb_reset_transport_header(skb);
 861                 __skb_push(skb, skb_gro_offset(skb));
 862
 863                 ops = rcu_dereference(inet6_protos[proto]);
 864                 if (!ops || !ops->gro_receive)
 865                         goto out_unlock;
 866
 867                 iph = ipv6_hdr(skb);
 868         }
 869
 870         NAPI_GRO_CB(skb)->proto = proto;
 871
 872         flush--;
 873         nlen = skb_network_header_len(skb);
 874
 875         for (p = *head; p; p = p->next) {
 876                 const struct ipv6hdr *iph2;
 877                 __be32 first_word; /* <Version:4><Traffic_Class:8><Flow_Label:20> */
 878
 879                 if (!NAPI_GRO_CB(p)->same_flow)
 880                         continue;
 881
 882                 iph2 = ipv6_hdr(p);
 883                 first_word = *(__be32 *)iph ^ *(__be32 *)iph2 ;
 884
 885                 /* All fields must match except length and Traffic Class. */
 886                 if (nlen != skb_network_header_len(p) ||
 887                     (first_word & htonl(0xF00FFFFF)) ||
 888                     memcmp(&iph->nexthdr, &iph2->nexthdr,
 889                            nlen - offsetof(struct ipv6hdr, nexthdr))) {
 890                         NAPI_GRO_CB(p)->same_flow = 0;
 891                         continue;
 892                 }
 893                 /* flush if Traffic Class fields are different */
 894                 NAPI_GRO_CB(p)->flush |= !!(first_word & htonl(0x0FF00000));
 895                 NAPI_GRO_CB(p)->flush |= flush;
 896         }
 897
 898         NAPI_GRO_CB(skb)->flush |= flush;
 899
 900         csum = skb->csum;
 901         skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
 902
 903         pp = ops->gro_receive(head, skb);
 904
 905         skb->csum = csum;
 906
 907 out_unlock:
 908         rcu_read_unlock();
 909
 910 out:
 911         NAPI_GRO_CB(skb)->flush |= flush;
 912
 913         return pp;
 914 }
 915
 916 static int ipv6_gro_complete(struct sk_buff *skb)
 917 {
 918         const struct inet6_protocol *ops;
 919         struct ipv6hdr *iph = ipv6_hdr(skb);
 920         int err = -ENOSYS;
 921
 922         iph->payload_len = htons(skb->len - skb_network_offset(skb) -
 923                                  sizeof(*iph));
 924
 925         rcu_read_lock();
 926         ops = rcu_dereference(inet6_protos[NAPI_GRO_CB(skb)->proto]);
 927         if (WARN_ON(!ops || !ops->gro_complete))
 928                 goto out_unlock;
 929
 930         err = ops->gro_complete(skb);
 931
 932 out_unlock:
 933         rcu_read_unlock();
 934
 935         return err;
 936 }
 937
 938 static struct packet_type ipv6_packet_type __read_mostly = {
 939         .type = cpu_to_be16(ETH_P_IPV6),
 940         .func = ipv6_rcv,
 941         .gso_send_check = ipv6_gso_send_check,
 942         .gso_segment = ipv6_gso_segment,
 943         .gro_receive = ipv6_gro_receive,
 944         .gro_complete = ipv6_gro_complete,
 945 };
 946
 947 static int __init ipv6_packet_init(void)
 948 {
 949         dev_add_pack(&ipv6_packet_type);
 950         return 0;
 951 }
 952
 953 static void ipv6_packet_cleanup(void)
 954 {
 955         dev_remove_pack(&ipv6_packet_type);
 956 }
 957
 958 static int __net_init ipv6_init_mibs(struct net *net)
 959 {
 960         if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
 961                           sizeof(struct udp_mib),
 962                           __alignof__(struct udp_mib)) < 0)
 963                 return -ENOMEM;
 964         if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
 965                           sizeof(struct udp_mib),
 966                           __alignof__(struct udp_mib)) < 0)
 967                 goto err_udplite_mib;
 968         if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
 969                           sizeof(struct ipstats_mib),
 970                           __alignof__(struct ipstats_mib)) < 0)
 971                 goto err_ip_mib;
 972         if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
 973                           sizeof(struct icmpv6_mib),
 974                           __alignof__(struct icmpv6_mib)) < 0)
 975                 goto err_icmp_mib;
 976         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
 977                                                 GFP_KERNEL);
 978         if (!net->mib.icmpv6msg_statistics)
 979                 goto err_icmpmsg_mib;
 980         return 0;
 981
 982 err_icmpmsg_mib:
 983         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 984 err_icmp_mib:
 985         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 986 err_ip_mib:
 987         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 988 err_udplite_mib:
 989         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 990         return -ENOMEM;
 991 }
 992
 993 static void ipv6_cleanup_mibs(struct net *net)
 994 {
 995         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
 996         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
 997         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 998         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 999         kfree(net->mib.icmpv6msg_statistics);
1000 }
1001
1002 static int __net_init inet6_net_init(struct net *net)
1003 {
1004         int err = 0;
1005
1006         net->ipv6.sysctl.bindv6only = 0;
1007         net->ipv6.sysctl.icmpv6_time = 1*HZ;
1008
1009         err = ipv6_init_mibs(net);
1010         if (err)
1011                 return err;
1012 #ifdef CONFIG_PROC_FS
1013         err = udp6_proc_init(net);
1014         if (err)
1015                 goto out;
1016         err = tcp6_proc_init(net);
1017         if (err)
1018                 goto proc_tcp6_fail;
1019         err = ac6_proc_init(net);
1020         if (err)
1021                 goto proc_ac6_fail;
1022 #endif
1023         return err;
1024
1025 #ifdef CONFIG_PROC_FS
1026 proc_ac6_fail:
1027         tcp6_proc_exit(net);
1028 proc_tcp6_fail:
1029         udp6_proc_exit(net);
1030 out:
1031         ipv6_cleanup_mibs(net);
1032         return err;
1033 #endif
1034 }
1035
1036 static void __net_exit inet6_net_exit(struct net *net)
1037 {
1038 #ifdef CONFIG_PROC_FS
1039         udp6_proc_exit(net);
1040         tcp6_proc_exit(net);
1041         ac6_proc_exit(net);
1042 #endif
1043         ipv6_cleanup_mibs(net);
1044 }
1045
1046 static struct pernet_operations inet6_net_ops = {
1047         .init = inet6_net_init,
1048         .exit = inet6_net_exit,
1049 };
1050
1051 static int __init inet6_init(void)
1052 {
1053         struct sk_buff *dummy_skb;
1054         struct list_head *r;
1055         int err = 0;
1056
1057         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1058
1059         /* Register the socket-side information for inet6_create.  */
1060         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1061                 INIT_LIST_HEAD(r);
1062
1063         if (disable_ipv6_mod) {
1064                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1065                 goto out;
1066         }
1067
1068         err = proto_register(&tcpv6_prot, 1);
1069         if (err)
1070                 goto out;
1071
1072         err = proto_register(&udpv6_prot, 1);
1073         if (err)
1074                 goto out_unregister_tcp_proto;
1075
1076         err = proto_register(&udplitev6_prot, 1);
1077         if (err)
1078                 goto out_unregister_udp_proto;
1079
1080         err = proto_register(&rawv6_prot, 1);
1081         if (err)
1082                 goto out_unregister_udplite_proto;
1083
1084
1085         /* We MUST register RAW sockets before we create the ICMP6,
1086          * IGMP6, or NDISC control sockets.
1087          */
1088         err = rawv6_init();
1089         if (err)
1090                 goto out_unregister_raw_proto;
1091
1092         /* Register the family here so that the init calls below will
1093          * be able to create sockets. (?? is this dangerous ??)
1094          */
1095         err = sock_register(&inet6_family_ops);
1096         if (err)
1097                 goto out_sock_register_fail;
1098
1099         tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
1100
1101         /*
1102          *      ipngwg API draft makes clear that the correct semantics
1103          *      for TCP and UDP is to consider one TCP and UDP instance
1104          *      in a host available by both INET and INET6 APIs and
1105          *      able to communicate via both network protocols.
1106          */
1107
1108         err = register_pernet_subsys(&inet6_net_ops);
1109         if (err)
1110                 goto register_pernet_fail;
1111         err = icmpv6_init();
1112         if (err)
1113                 goto icmp_fail;
1114         err = ip6_mr_init();
1115         if (err)
1116                 goto ipmr_fail;
1117         err = ndisc_init();
1118         if (err)
1119                 goto ndisc_fail;
1120         err = igmp6_init();
1121         if (err)
1122                 goto igmp_fail;
1123         err = ipv6_netfilter_init();
1124         if (err)
1125                 goto netfilter_fail;
1126         /* Create /proc/foo6 entries. */
1127 #ifdef CONFIG_PROC_FS
1128         err = -ENOMEM;
1129         if (raw6_proc_init())
1130                 goto proc_raw6_fail;
1131         if (udplite6_proc_init())
1132                 goto proc_udplite6_fail;
1133         if (ipv6_misc_proc_init())
1134                 goto proc_misc6_fail;
1135         if (if6_proc_init())
1136                 goto proc_if6_fail;
1137 #endif
1138         err = ip6_route_init();
1139         if (err)
1140                 goto ip6_route_fail;
1141         err = ip6_flowlabel_init();
1142         if (err)
1143                 goto ip6_flowlabel_fail;
1144         err = addrconf_init();
1145         if (err)
1146                 goto addrconf_fail;
1147
1148         /* Init v6 extension headers. */
1149         err = ipv6_exthdrs_init();
1150         if (err)
1151                 goto ipv6_exthdrs_fail;
1152
1153         err = ipv6_frag_init();
1154         if (err)
1155                 goto ipv6_frag_fail;
1156
1157         /* Init v6 transport protocols. */
1158         err = udpv6_init();
1159         if (err)
1160                 goto udpv6_fail;
1161
1162         err = udplitev6_init();
1163         if (err)
1164                 goto udplitev6_fail;
1165
1166         err = tcpv6_init();
1167         if (err)
1168                 goto tcpv6_fail;
1169
1170         err = ipv6_packet_init();
1171         if (err)
1172                 goto ipv6_packet_fail;
1173
1174 #ifdef CONFIG_SYSCTL
1175         err = ipv6_sysctl_register();
1176         if (err)
1177                 goto sysctl_fail;
1178 #endif
1179 out:
1180         return err;
1181
1182 #ifdef CONFIG_SYSCTL
1183 sysctl_fail:
1184         ipv6_packet_cleanup();
1185 #endif
1186 ipv6_packet_fail:
1187         tcpv6_exit();
1188 tcpv6_fail:
1189         udplitev6_exit();
1190 udplitev6_fail:
1191         udpv6_exit();
1192 udpv6_fail:
1193         ipv6_frag_exit();
1194 ipv6_frag_fail:
1195         ipv6_exthdrs_exit();
1196 ipv6_exthdrs_fail:
1197         addrconf_cleanup();
1198 addrconf_fail:
1199         ip6_flowlabel_cleanup();
1200 ip6_flowlabel_fail:
1201         ip6_route_cleanup();
1202 ip6_route_fail:
1203 #ifdef CONFIG_PROC_FS
1204         if6_proc_exit();
1205 proc_if6_fail:
1206         ipv6_misc_proc_exit();
1207 proc_misc6_fail:
1208         udplite6_proc_exit();
1209 proc_udplite6_fail:
1210         raw6_proc_exit();
1211 proc_raw6_fail:
1212 #endif
1213         ipv6_netfilter_fini();
1214 netfilter_fail:
1215         igmp6_cleanup();
1216 igmp_fail:
1217         ndisc_cleanup();
1218 ndisc_fail:
1219         ip6_mr_cleanup();
1220 ipmr_fail:
1221         icmpv6_cleanup();
1222 icmp_fail:
1223         unregister_pernet_subsys(&inet6_net_ops);
1224 register_pernet_fail:
1225         sock_unregister(PF_INET6);
1226         rtnl_unregister_all(PF_INET6);
1227 out_sock_register_fail:
1228         rawv6_exit();
1229 out_unregister_raw_proto:
1230         proto_unregister(&rawv6_prot);
1231 out_unregister_udplite_proto:
1232         proto_unregister(&udplitev6_prot);
1233 out_unregister_udp_proto:
1234         proto_unregister(&udpv6_prot);
1235 out_unregister_tcp_proto:
1236         proto_unregister(&tcpv6_prot);
1237         goto out;
1238 }
1239 module_init(inet6_init);
1240
1241 static void __exit inet6_exit(void)
1242 {
1243         if (disable_ipv6_mod)
1244                 return;
1245
1246         /* First of all disallow new sockets creation. */
1247         sock_unregister(PF_INET6);
1248         /* Disallow any further netlink messages */
1249         rtnl_unregister_all(PF_INET6);
1250
1251         udpv6_exit();
1252         udplitev6_exit();
1253         tcpv6_exit();
1254
1255         /* Cleanup code parts. */
1256         ipv6_packet_cleanup();
1257         ipv6_frag_exit();
1258         ipv6_exthdrs_exit();
1259         addrconf_cleanup();
1260         ip6_flowlabel_cleanup();
1261         ip6_route_cleanup();
1262 #ifdef CONFIG_PROC_FS
1263
1264         /* Cleanup code parts. */
1265         if6_proc_exit();
1266         ipv6_misc_proc_exit();
1267         udplite6_proc_exit();
1268         raw6_proc_exit();
1269 #endif
1270         ipv6_netfilter_fini();
1271         igmp6_cleanup();
1272         ndisc_cleanup();
1273         ip6_mr_cleanup();
1274         icmpv6_cleanup();
1275         rawv6_exit();
1276
1277         unregister_pernet_subsys(&inet6_net_ops);
1278         proto_unregister(&rawv6_prot);
1279         proto_unregister(&udplitev6_prot);
1280         proto_unregister(&udpv6_prot);
1281         proto_unregister(&tcpv6_prot);
1282
1283         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1284 }
1285 module_exit(inet6_exit);
1286
1287 MODULE_ALIAS_NETPROTO(PF_INET6);