linux/net/socket.c
<<
>>
Prefs
   1/*
   2 * NET          An implementation of the SOCKET network access protocol.
   3 *
   4 * Version:     @(#)socket.c    1.1.93  18/02/95
   5 *
   6 * Authors:     Orest Zborowski, <obz@Kodak.COM>
   7 *              Ross Biro
   8 *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
   9 *
  10 * Fixes:
  11 *              Anonymous       :       NOTSOCK/BADF cleanup. Error fix in
  12 *                                      shutdown()
  13 *              Alan Cox        :       verify_area() fixes
  14 *              Alan Cox        :       Removed DDI
  15 *              Jonathan Kamens :       SOCK_DGRAM reconnect bug
  16 *              Alan Cox        :       Moved a load of checks to the very
  17 *                                      top level.
  18 *              Alan Cox        :       Move address structures to/from user
  19 *                                      mode above the protocol layers.
  20 *              Rob Janssen     :       Allow 0 length sends.
  21 *              Alan Cox        :       Asynchronous I/O support (cribbed from the
  22 *                                      tty drivers).
  23 *              Niibe Yutaka    :       Asynchronous I/O for writes (4.4BSD style)
  24 *              Jeff Uphoff     :       Made max number of sockets command-line
  25 *                                      configurable.
  26 *              Matti Aarnio    :       Made the number of sockets dynamic,
  27 *                                      to be allocated when needed, and mr.
  28 *                                      Uphoff's max is used as max to be
  29 *                                      allowed to allocate.
  30 *              Linus           :       Argh. removed all the socket allocation
  31 *                                      altogether: it's in the inode now.
  32 *              Alan Cox        :       Made sock_alloc()/sock_release() public
  33 *                                      for NetROM and future kernel nfsd type
  34 *                                      stuff.
  35 *              Alan Cox        :       sendmsg/recvmsg basics.
  36 *              Tom Dyas        :       Export net symbols.
  37 *              Marcin Dalecki  :       Fixed problems with CONFIG_NET="n".
  38 *              Alan Cox        :       Added thread locking to sys_* calls
  39 *                                      for sockets. May have errors at the
  40 *                                      moment.
  41 *              Kevin Buhr      :       Fixed the dumb errors in the above.
  42 *              Andi Kleen      :       Some small cleanups, optimizations,
  43 *                                      and fixed a copy_from_user() bug.
  44 *              Tigran Aivazian :       sys_send(args) calls sys_sendto(args, NULL, 0)
  45 *              Tigran Aivazian :       Made listen(2) backlog sanity checks
  46 *                                      protocol-independent
  47 *
  48 *
  49 *              This program is free software; you can redistribute it and/or
  50 *              modify it under the terms of the GNU General Public License
  51 *              as published by the Free Software Foundation; either version
  52 *              2 of the License, or (at your option) any later version.
  53 *
  54 *
  55 *      This module is effectively the top level interface to the BSD socket
  56 *      paradigm.
  57 *
  58 *      Based upon Swansea University Computer Society NET3.039
  59 */
  60
  61#include <linux/mm.h>
  62#include <linux/socket.h>
  63#include <linux/file.h>
  64#include <linux/net.h>
  65#include <linux/interrupt.h>
  66#include <linux/thread_info.h>
  67#include <linux/rcupdate.h>
  68#include <linux/netdevice.h>
  69#include <linux/proc_fs.h>
  70#include <linux/seq_file.h>
  71#include <linux/mutex.h>
  72#include <linux/wanrouter.h>
  73#include <linux/if_bridge.h>
  74#include <linux/if_frad.h>
  75#include <linux/if_vlan.h>
  76#include <linux/init.h>
  77#include <linux/poll.h>
  78#include <linux/cache.h>
  79#include <linux/module.h>
  80#include <linux/highmem.h>
  81#include <linux/mount.h>
  82#include <linux/security.h>
  83#include <linux/syscalls.h>
  84#include <linux/compat.h>
  85#include <linux/kmod.h>
  86#include <linux/audit.h>
  87#include <linux/wireless.h>
  88#include <linux/nsproxy.h>
  89#include <linux/magic.h>
  90#include <linux/slab.h>
  91
  92#include <asm/uaccess.h>
  93#include <asm/unistd.h>
  94
  95#include <net/compat.h>
  96#include <net/wext.h>
  97#include <net/cls_cgroup.h>
  98
  99#include <net/sock.h>
 100#include <linux/netfilter.h>
 101
 102#include <linux/if_tun.h>
 103#include <linux/ipv6_route.h>
 104#include <linux/route.h>
 105#include <linux/sockios.h>
 106#include <linux/atalk.h>
 107
 108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
 109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 110                         unsigned long nr_segs, loff_t pos);
 111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 112                          unsigned long nr_segs, loff_t pos);
 113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
 114
 115static int sock_close(struct inode *inode, struct file *file);
 116static unsigned int sock_poll(struct file *file,
 117                              struct poll_table_struct *wait);
 118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 119#ifdef CONFIG_COMPAT
 120static long compat_sock_ioctl(struct file *file,
 121                              unsigned int cmd, unsigned long arg);
 122#endif
 123static int sock_fasync(int fd, struct file *filp, int on);
 124static ssize_t sock_sendpage(struct file *file, struct page *page,
 125                             int offset, size_t size, loff_t *ppos, int more);
 126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 127                                struct pipe_inode_info *pipe, size_t len,
 128                                unsigned int flags);
 129
 130/*
 131 *      Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
 132 *      in the operation structures but are done directly via the socketcall() multiplexor.
 133 */
 134
 135static const struct file_operations socket_file_ops = {
 136        .owner =        THIS_MODULE,
 137        .llseek =       no_llseek,
 138        .aio_read =     sock_aio_read,
 139        .aio_write =    sock_aio_write,
 140        .poll =         sock_poll,
 141        .unlocked_ioctl = sock_ioctl,
 142#ifdef CONFIG_COMPAT
 143        .compat_ioctl = compat_sock_ioctl,
 144#endif
 145        .mmap =         sock_mmap,
 146        .open =         sock_no_open,   /* special open code to disallow open via /proc */
 147        .release =      sock_close,
 148        .fasync =       sock_fasync,
 149        .sendpage =     sock_sendpage,
 150        .splice_write = generic_splice_sendpage,
 151        .splice_read =  sock_splice_read,
 152};
 153
 154/*
 155 *      The protocol list. Each protocol is registered in here.
 156 */
 157
 158static DEFINE_SPINLOCK(net_family_lock);
 159static const struct net_proto_family *net_families[NPROTO] __read_mostly;
 160
 161/*
 162 *      Statistics counters of the socket lists
 163 */
 164
 165static DEFINE_PER_CPU(int, sockets_in_use) = 0;
 166
 167/*
 168 * Support routines.
 169 * Move socket addresses back and forth across the kernel/user
 170 * divide and look after the messy bits.
 171 */
 172
 173#define MAX_SOCK_ADDR   128             /* 108 for Unix domain -
 174                                           16 for IP, 16 for IPX,
 175                                           24 for IPv6,
 176                                           about 80 for AX.25
 177                                           must be at least one bigger than
 178                                           the AF_UNIX size (see net/unix/af_unix.c
 179                                           :unix_mkname()).
 180                                         */
 181
 182/**
 183 *      move_addr_to_kernel     -       copy a socket address into kernel space
 184 *      @uaddr: Address in user space
 185 *      @kaddr: Address in kernel space
 186 *      @ulen: Length in user space
 187 *
 188 *      The address is copied into kernel space. If the provided address is
 189 *      too long an error code of -EINVAL is returned. If the copy gives
 190 *      invalid addresses -EFAULT is returned. On a success 0 is returned.
 191 */
 192
 193int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
 194{
 195        if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
 196                return -EINVAL;
 197        if (ulen == 0)
 198                return 0;
 199        if (copy_from_user(kaddr, uaddr, ulen))
 200                return -EFAULT;
 201        return audit_sockaddr(ulen, kaddr);
 202}
 203
 204/**
 205 *      move_addr_to_user       -       copy an address to user space
 206 *      @kaddr: kernel space address
 207 *      @klen: length of address in kernel
 208 *      @uaddr: user space address
 209 *      @ulen: pointer to user length field
 210 *
 211 *      The value pointed to by ulen on entry is the buffer length available.
 212 *      This is overwritten with the buffer space used. -EINVAL is returned
 213 *      if an overlong buffer is specified or a negative buffer size. -EFAULT
 214 *      is returned if either the buffer or the length field are not
 215 *      accessible.
 216 *      After copying the data up to the limit the user specifies, the true
 217 *      length of the data is written over the length limit the user
 218 *      specified. Zero is returned for a success.
 219 */
 220
 221int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
 222                      int __user *ulen)
 223{
 224        int err;
 225        int len;
 226
 227        err = get_user(len, ulen);
 228        if (err)
 229                return err;
 230        if (len > klen)
 231                len = klen;
 232        if (len < 0 || len > sizeof(struct sockaddr_storage))
 233                return -EINVAL;
 234        if (len) {
 235                if (audit_sockaddr(klen, kaddr))
 236                        return -ENOMEM;
 237                if (copy_to_user(uaddr, kaddr, len))
 238                        return -EFAULT;
 239        }
 240        /*
 241         *      "fromlen shall refer to the value before truncation.."
 242         *                      1003.1g
 243         */
 244        return __put_user(klen, ulen);
 245}
 246
 247static struct kmem_cache *sock_inode_cachep __read_mostly;
 248
 249static struct inode *sock_alloc_inode(struct super_block *sb)
 250{
 251        struct socket_alloc *ei;
 252
 253        ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
 254        if (!ei)
 255                return NULL;
 256        ei->socket.wq = kmalloc(sizeof(struct socket_wq), GFP_KERNEL);
 257        if (!ei->socket.wq) {
 258                kmem_cache_free(sock_inode_cachep, ei);
 259                return NULL;
 260        }
 261        init_waitqueue_head(&ei->socket.wq->wait);
 262        ei->socket.wq->fasync_list = NULL;
 263
 264        ei->socket.state = SS_UNCONNECTED;
 265        ei->socket.flags = 0;
 266        ei->socket.ops = NULL;
 267        ei->socket.sk = NULL;
 268        ei->socket.file = NULL;
 269
 270        return &ei->vfs_inode;
 271}
 272
 273
 274static void wq_free_rcu(struct rcu_head *head)
 275{
 276        struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
 277
 278        kfree(wq);
 279}
 280
 281static void sock_destroy_inode(struct inode *inode)
 282{
 283        struct socket_alloc *ei;
 284
 285        ei = container_of(inode, struct socket_alloc, vfs_inode);
 286        call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
 287        kmem_cache_free(sock_inode_cachep, ei);
 288}
 289
 290static void init_once(void *foo)
 291{
 292        struct socket_alloc *ei = (struct socket_alloc *)foo;
 293
 294        inode_init_once(&ei->vfs_inode);
 295}
 296
 297static int init_inodecache(void)
 298{
 299        sock_inode_cachep = kmem_cache_create("sock_inode_cache",
 300                                              sizeof(struct socket_alloc),
 301                                              0,
 302                                              (SLAB_HWCACHE_ALIGN |
 303                                               SLAB_RECLAIM_ACCOUNT |
 304                                               SLAB_MEM_SPREAD),
 305                                              init_once);
 306        if (sock_inode_cachep == NULL)
 307                return -ENOMEM;
 308        return 0;
 309}
 310
 311static const struct super_operations sockfs_ops = {
 312        .alloc_inode =  sock_alloc_inode,
 313        .destroy_inode =sock_destroy_inode,
 314        .statfs =       simple_statfs,
 315};
 316
 317static int sockfs_get_sb(struct file_system_type *fs_type,
 318                         int flags, const char *dev_name, void *data,
 319                         struct vfsmount *mnt)
 320{
 321        return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
 322                             mnt);
 323}
 324
 325static struct vfsmount *sock_mnt __read_mostly;
 326
 327static struct file_system_type sock_fs_type = {
 328        .name =         "sockfs",
 329        .get_sb =       sockfs_get_sb,
 330        .kill_sb =      kill_anon_super,
 331};
 332
 333/*
 334 * sockfs_dname() is called from d_path().
 335 */
 336static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
 337{
 338        return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
 339                                dentry->d_inode->i_ino);
 340}
 341
 342static const struct dentry_operations sockfs_dentry_operations = {
 343        .d_dname  = sockfs_dname,
 344};
 345
 346/*
 347 *      Obtains the first available file descriptor and sets it up for use.
 348 *
 349 *      These functions create file structures and maps them to fd space
 350 *      of the current process. On success it returns file descriptor
 351 *      and file struct implicitly stored in sock->file.
 352 *      Note that another thread may close file descriptor before we return
 353 *      from this function. We use the fact that now we do not refer
 354 *      to socket after mapping. If one day we will need it, this
 355 *      function will increment ref. count on file by 1.
 356 *
 357 *      In any case returned fd MAY BE not valid!
 358 *      This race condition is unavoidable
 359 *      with shared fd spaces, we cannot solve it inside kernel,
 360 *      but we take care of internal coherence yet.
 361 */
 362
 363static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
 364{
 365        struct qstr name = { .name = "" };
 366        struct path path;
 367        struct file *file;
 368        int fd;
 369
 370        fd = get_unused_fd_flags(flags);
 371        if (unlikely(fd < 0))
 372                return fd;
 373
 374        path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
 375        if (unlikely(!path.dentry)) {
 376                put_unused_fd(fd);
 377                return -ENOMEM;
 378        }
 379        path.mnt = mntget(sock_mnt);
 380
 381        path.dentry->d_op = &sockfs_dentry_operations;
 382        d_instantiate(path.dentry, SOCK_INODE(sock));
 383        SOCK_INODE(sock)->i_fop = &socket_file_ops;
 384
 385        file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
 386                  &socket_file_ops);
 387        if (unlikely(!file)) {
 388                /* drop dentry, keep inode */
 389                atomic_inc(&path.dentry->d_inode->i_count);
 390                path_put(&path);
 391                put_unused_fd(fd);
 392                return -ENFILE;
 393        }
 394
 395        sock->file = file;
 396        file->f_flags = O_RDWR | (flags & O_NONBLOCK);
 397        file->f_pos = 0;
 398        file->private_data = sock;
 399
 400        *f = file;
 401        return fd;
 402}
 403
 404int sock_map_fd(struct socket *sock, int flags)
 405{
 406        struct file *newfile;
 407        int fd = sock_alloc_file(sock, &newfile, flags);
 408
 409        if (likely(fd >= 0))
 410                fd_install(fd, newfile);
 411
 412        return fd;
 413}
 414
 415static struct socket *sock_from_file(struct file *file, int *err)
 416{
 417        if (file->f_op == &socket_file_ops)
 418                return file->private_data;      /* set in sock_map_fd */
 419
 420        *err = -ENOTSOCK;
 421        return NULL;
 422}
 423
 424/**
 425 *      sockfd_lookup   -       Go from a file number to its socket slot
 426 *      @fd: file handle
 427 *      @err: pointer to an error code return
 428 *
 429 *      The file handle passed in is locked and the socket it is bound
 430 *      too is returned. If an error occurs the err pointer is overwritten
 431 *      with a negative errno code and NULL is returned. The function checks
 432 *      for both invalid handles and passing a handle which is not a socket.
 433 *
 434 *      On a success the socket object pointer is returned.
 435 */
 436
 437struct socket *sockfd_lookup(int fd, int *err)
 438{
 439        struct file *file;
 440        struct socket *sock;
 441
 442        file = fget(fd);
 443        if (!file) {
 444                *err = -EBADF;
 445                return NULL;
 446        }
 447
 448        sock = sock_from_file(file, err);
 449        if (!sock)
 450                fput(file);
 451        return sock;
 452}
 453
 454static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 455{
 456        struct file *file;
 457        struct socket *sock;
 458
 459        *err = -EBADF;
 460        file = fget_light(fd, fput_needed);
 461        if (file) {
 462                sock = sock_from_file(file, err);
 463                if (sock)
 464                        return sock;
 465                fput_light(file, *fput_needed);
 466        }
 467        return NULL;
 468}
 469
 470/**
 471 *      sock_alloc      -       allocate a socket
 472 *
 473 *      Allocate a new inode and socket object. The two are bound together
 474 *      and initialised. The socket is then returned. If we are out of inodes
 475 *      NULL is returned.
 476 */
 477
 478static struct socket *sock_alloc(void)
 479{
 480        struct inode *inode;
 481        struct socket *sock;
 482
 483        inode = new_inode(sock_mnt->mnt_sb);
 484        if (!inode)
 485                return NULL;
 486
 487        sock = SOCKET_I(inode);
 488
 489        kmemcheck_annotate_bitfield(sock, type);
 490        inode->i_mode = S_IFSOCK | S_IRWXUGO;
 491        inode->i_uid = current_fsuid();
 492        inode->i_gid = current_fsgid();
 493
 494        percpu_add(sockets_in_use, 1);
 495        return sock;
 496}
 497
 498/*
 499 *      In theory you can't get an open on this inode, but /proc provides
 500 *      a back door. Remember to keep it shut otherwise you'll let the
 501 *      creepy crawlies in.
 502 */
 503
 504static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
 505{
 506        return -ENXIO;
 507}
 508
 509const struct file_operations bad_sock_fops = {
 510        .owner = THIS_MODULE,
 511        .open = sock_no_open,
 512};
 513
 514/**
 515 *      sock_release    -       close a socket
 516 *      @sock: socket to close
 517 *
 518 *      The socket is released from the protocol stack if it has a release
 519 *      callback, and the inode is then released if the socket is bound to
 520 *      an inode not a file.
 521 */
 522
 523void sock_release(struct socket *sock)
 524{
 525        if (sock->ops) {
 526                struct module *owner = sock->ops->owner;
 527
 528                sock->ops->release(sock);
 529                sock->ops = NULL;
 530                module_put(owner);
 531        }
 532
 533        if (sock->wq->fasync_list)
 534                printk(KERN_ERR "sock_release: fasync list not empty!\n");
 535
 536        percpu_sub(sockets_in_use, 1);
 537        if (!sock->file) {
 538                iput(SOCK_INODE(sock));
 539                return;
 540        }
 541        sock->file = NULL;
 542}
 543
 544int sock_tx_timestamp(struct msghdr *msg, struct sock *sk,
 545                      union skb_shared_tx *shtx)
 546{
 547        shtx->flags = 0;
 548        if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
 549                shtx->hardware = 1;
 550        if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
 551                shtx->software = 1;
 552        return 0;
 553}
 554EXPORT_SYMBOL(sock_tx_timestamp);
 555
 556static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 557                                 struct msghdr *msg, size_t size)
 558{
 559        struct sock_iocb *si = kiocb_to_siocb(iocb);
 560        int err;
 561
 562        sock_update_classid(sock->sk);
 563
 564        si->sock = sock;
 565        si->scm = NULL;
 566        si->msg = msg;
 567        si->size = size;
 568
 569        err = security_socket_sendmsg(sock, msg, size);
 570        if (err)
 571                return err;
 572
 573        return sock->ops->sendmsg(iocb, sock, msg, size);
 574}
 575
 576int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 577{
 578        struct kiocb iocb;
 579        struct sock_iocb siocb;
 580        int ret;
 581
 582        init_sync_kiocb(&iocb, NULL);
 583        iocb.private = &siocb;
 584        ret = __sock_sendmsg(&iocb, sock, msg, size);
 585        if (-EIOCBQUEUED == ret)
 586                ret = wait_on_sync_kiocb(&iocb);
 587        return ret;
 588}
 589
 590int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 591                   struct kvec *vec, size_t num, size_t size)
 592{
 593        mm_segment_t oldfs = get_fs();
 594        int result;
 595
 596        set_fs(KERNEL_DS);
 597        /*
 598         * the following is safe, since for compiler definitions of kvec and
 599         * iovec are identical, yielding the same in-core layout and alignment
 600         */
 601        msg->msg_iov = (struct iovec *)vec;
 602        msg->msg_iovlen = num;
 603        result = sock_sendmsg(sock, msg, size);
 604        set_fs(oldfs);
 605        return result;
 606}
 607
 608static int ktime2ts(ktime_t kt, struct timespec *ts)
 609{
 610        if (kt.tv64) {
 611                *ts = ktime_to_timespec(kt);
 612                return 1;
 613        } else {
 614                return 0;
 615        }
 616}
 617
 618/*
 619 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 620 */
 621void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 622        struct sk_buff *skb)
 623{
 624        int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 625        struct timespec ts[3];
 626        int empty = 1;
 627        struct skb_shared_hwtstamps *shhwtstamps =
 628                skb_hwtstamps(skb);
 629
 630        /* Race occurred between timestamp enabling and packet
 631           receiving.  Fill in the current time for now. */
 632        if (need_software_tstamp && skb->tstamp.tv64 == 0)
 633                __net_timestamp(skb);
 634
 635        if (need_software_tstamp) {
 636                if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 637                        struct timeval tv;
 638                        skb_get_timestamp(skb, &tv);
 639                        put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 640                                 sizeof(tv), &tv);
 641                } else {
 642                        skb_get_timestampns(skb, &ts[0]);
 643                        put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 644                                 sizeof(ts[0]), &ts[0]);
 645                }
 646        }
 647
 648
 649        memset(ts, 0, sizeof(ts));
 650        if (skb->tstamp.tv64 &&
 651            sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
 652                skb_get_timestampns(skb, ts + 0);
 653                empty = 0;
 654        }
 655        if (shhwtstamps) {
 656                if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 657                    ktime2ts(shhwtstamps->syststamp, ts + 1))
 658                        empty = 0;
 659                if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 660                    ktime2ts(shhwtstamps->hwtstamp, ts + 2))
 661                        empty = 0;
 662        }
 663        if (!empty)
 664                put_cmsg(msg, SOL_SOCKET,
 665                         SCM_TIMESTAMPING, sizeof(ts), &ts);
 666}
 667
 668EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 669
 670inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
 671{
 672        if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 673                put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 674                        sizeof(__u32), &skb->dropcount);
 675}
 676
 677void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 678        struct sk_buff *skb)
 679{
 680        sock_recv_timestamp(msg, sk, skb);
 681        sock_recv_drops(msg, sk, skb);
 682}
 683EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 684
 685static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 686                                       struct msghdr *msg, size_t size, int flags)
 687{
 688        struct sock_iocb *si = kiocb_to_siocb(iocb);
 689
 690        sock_update_classid(sock->sk);
 691
 692        si->sock = sock;
 693        si->scm = NULL;
 694        si->msg = msg;
 695        si->size = size;
 696        si->flags = flags;
 697
 698        return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 699}
 700
 701static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 702                                 struct msghdr *msg, size_t size, int flags)
 703{
 704        int err = security_socket_recvmsg(sock, msg, size, flags);
 705
 706        return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 707}
 708
 709int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 710                 size_t size, int flags)
 711{
 712        struct kiocb iocb;
 713        struct sock_iocb siocb;
 714        int ret;
 715
 716        init_sync_kiocb(&iocb, NULL);
 717        iocb.private = &siocb;
 718        ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 719        if (-EIOCBQUEUED == ret)
 720                ret = wait_on_sync_kiocb(&iocb);
 721        return ret;
 722}
 723
 724static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 725                              size_t size, int flags)
 726{
 727        struct kiocb iocb;
 728        struct sock_iocb siocb;
 729        int ret;
 730
 731        init_sync_kiocb(&iocb, NULL);
 732        iocb.private = &siocb;
 733        ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 734        if (-EIOCBQUEUED == ret)
 735                ret = wait_on_sync_kiocb(&iocb);
 736        return ret;
 737}
 738
 739int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 740                   struct kvec *vec, size_t num, size_t size, int flags)
 741{
 742        mm_segment_t oldfs = get_fs();
 743        int result;
 744
 745        set_fs(KERNEL_DS);
 746        /*
 747         * the following is safe, since for compiler definitions of kvec and
 748         * iovec are identical, yielding the same in-core layout and alignment
 749         */
 750        msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 751        result = sock_recvmsg(sock, msg, size, flags);
 752        set_fs(oldfs);
 753        return result;
 754}
 755
 756static void sock_aio_dtor(struct kiocb *iocb)
 757{
 758        kfree(iocb->private);
 759}
 760
 761static ssize_t sock_sendpage(struct file *file, struct page *page,
 762                             int offset, size_t size, loff_t *ppos, int more)
 763{
 764        struct socket *sock;
 765        int flags;
 766
 767        sock = file->private_data;
 768
 769        flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
 770        if (more)
 771                flags |= MSG_MORE;
 772
 773        return kernel_sendpage(sock, page, offset, size, flags);
 774}
 775
 776static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 777                                struct pipe_inode_info *pipe, size_t len,
 778                                unsigned int flags)
 779{
 780        struct socket *sock = file->private_data;
 781
 782        if (unlikely(!sock->ops->splice_read))
 783                return -EINVAL;
 784
 785        sock_update_classid(sock->sk);
 786
 787        return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 788}
 789
 790static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 791                                         struct sock_iocb *siocb)
 792{
 793        if (!is_sync_kiocb(iocb)) {
 794                siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
 795                if (!siocb)
 796                        return NULL;
 797                iocb->ki_dtor = sock_aio_dtor;
 798        }
 799
 800        siocb->kiocb = iocb;
 801        iocb->private = siocb;
 802        return siocb;
 803}
 804
 805static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 806                struct file *file, const struct iovec *iov,
 807                unsigned long nr_segs)
 808{
 809        struct socket *sock = file->private_data;
 810        size_t size = 0;
 811        int i;
 812
 813        for (i = 0; i < nr_segs; i++)
 814                size += iov[i].iov_len;
 815
 816        msg->msg_name = NULL;
 817        msg->msg_namelen = 0;
 818        msg->msg_control = NULL;
 819        msg->msg_controllen = 0;
 820        msg->msg_iov = (struct iovec *)iov;
 821        msg->msg_iovlen = nr_segs;
 822        msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 823
 824        return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 825}
 826
 827static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 828                                unsigned long nr_segs, loff_t pos)
 829{
 830        struct sock_iocb siocb, *x;
 831
 832        if (pos != 0)
 833                return -ESPIPE;
 834
 835        if (iocb->ki_left == 0) /* Match SYS5 behaviour */
 836                return 0;
 837
 838
 839        x = alloc_sock_iocb(iocb, &siocb);
 840        if (!x)
 841                return -ENOMEM;
 842        return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 843}
 844
 845static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 846                        struct file *file, const struct iovec *iov,
 847                        unsigned long nr_segs)
 848{
 849        struct socket *sock = file->private_data;
 850        size_t size = 0;
 851        int i;
 852
 853        for (i = 0; i < nr_segs; i++)
 854                size += iov[i].iov_len;
 855
 856        msg->msg_name = NULL;
 857        msg->msg_namelen = 0;
 858        msg->msg_control = NULL;
 859        msg->msg_controllen = 0;
 860        msg->msg_iov = (struct iovec *)iov;
 861        msg->msg_iovlen = nr_segs;
 862        msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 863        if (sock->type == SOCK_SEQPACKET)
 864                msg->msg_flags |= MSG_EOR;
 865
 866        return __sock_sendmsg(iocb, sock, msg, size);
 867}
 868
 869static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 870                          unsigned long nr_segs, loff_t pos)
 871{
 872        struct sock_iocb siocb, *x;
 873
 874        if (pos != 0)
 875                return -ESPIPE;
 876
 877        x = alloc_sock_iocb(iocb, &siocb);
 878        if (!x)
 879                return -ENOMEM;
 880
 881        return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 882}
 883
 884/*
 885 * Atomic setting of ioctl hooks to avoid race
 886 * with module unload.
 887 */
 888
 889static DEFINE_MUTEX(br_ioctl_mutex);
 890static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
 891
 892void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 893{
 894        mutex_lock(&br_ioctl_mutex);
 895        br_ioctl_hook = hook;
 896        mutex_unlock(&br_ioctl_mutex);
 897}
 898
 899EXPORT_SYMBOL(brioctl_set);
 900
 901static DEFINE_MUTEX(vlan_ioctl_mutex);
 902static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 903
 904void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 905{
 906        mutex_lock(&vlan_ioctl_mutex);
 907        vlan_ioctl_hook = hook;
 908        mutex_unlock(&vlan_ioctl_mutex);
 909}
 910
 911EXPORT_SYMBOL(vlan_ioctl_set);
 912
 913static DEFINE_MUTEX(dlci_ioctl_mutex);
 914static int (*dlci_ioctl_hook) (unsigned int, void __user *);
 915
 916void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
 917{
 918        mutex_lock(&dlci_ioctl_mutex);
 919        dlci_ioctl_hook = hook;
 920        mutex_unlock(&dlci_ioctl_mutex);
 921}
 922
 923EXPORT_SYMBOL(dlci_ioctl_set);
 924
 925static long sock_do_ioctl(struct net *net, struct socket *sock,
 926                                 unsigned int cmd, unsigned long arg)
 927{
 928        int err;
 929        void __user *argp = (void __user *)arg;
 930
 931        err = sock->ops->ioctl(sock, cmd, arg);
 932
 933        /*
 934         * If this ioctl is unknown try to hand it down
 935         * to the NIC driver.
 936         */
 937        if (err == -ENOIOCTLCMD)
 938                err = dev_ioctl(net, cmd, argp);
 939
 940        return err;
 941}
 942
 943/*
 944 *      With an ioctl, arg may well be a user mode pointer, but we don't know
 945 *      what to do with it - that's up to the protocol still.
 946 */
 947
 948static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 949{
 950        struct socket *sock;
 951        struct sock *sk;
 952        void __user *argp = (void __user *)arg;
 953        int pid, err;
 954        struct net *net;
 955
 956        sock = file->private_data;
 957        sk = sock->sk;
 958        net = sock_net(sk);
 959        if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
 960                err = dev_ioctl(net, cmd, argp);
 961        } else
 962#ifdef CONFIG_WEXT_CORE
 963        if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
 964                err = dev_ioctl(net, cmd, argp);
 965        } else
 966#endif
 967                switch (cmd) {
 968                case FIOSETOWN:
 969                case SIOCSPGRP:
 970                        err = -EFAULT;
 971                        if (get_user(pid, (int __user *)argp))
 972                                break;
 973                        err = f_setown(sock->file, pid, 1);
 974                        break;
 975                case FIOGETOWN:
 976                case SIOCGPGRP:
 977                        err = put_user(f_getown(sock->file),
 978                                       (int __user *)argp);
 979                        break;
 980                case SIOCGIFBR:
 981                case SIOCSIFBR:
 982                case SIOCBRADDBR:
 983                case SIOCBRDELBR:
 984                        err = -ENOPKG;
 985                        if (!br_ioctl_hook)
 986                                request_module("bridge");
 987
 988                        mutex_lock(&br_ioctl_mutex);
 989                        if (br_ioctl_hook)
 990                                err = br_ioctl_hook(net, cmd, argp);
 991                        mutex_unlock(&br_ioctl_mutex);
 992                        break;
 993                case SIOCGIFVLAN:
 994                case SIOCSIFVLAN:
 995                        err = -ENOPKG;
 996                        if (!vlan_ioctl_hook)
 997                                request_module("8021q");
 998
 999                        mutex_lock(&vlan_ioctl_mutex);
1000                        if (vlan_ioctl_hook)
1001                                err = vlan_ioctl_hook(net, argp);
1002                        mutex_unlock(&vlan_ioctl_mutex);
1003                        break;
1004                case SIOCADDDLCI:
1005                case SIOCDELDLCI:
1006                        err = -ENOPKG;
1007                        if (!dlci_ioctl_hook)
1008                                request_module("dlci");
1009
1010                        mutex_lock(&dlci_ioctl_mutex);
1011                        if (dlci_ioctl_hook)
1012                                err = dlci_ioctl_hook(cmd, argp);
1013                        mutex_unlock(&dlci_ioctl_mutex);
1014                        break;
1015                default:
1016                        err = sock_do_ioctl(net, sock, cmd, arg);
1017                        break;
1018                }
1019        return err;
1020}
1021
1022int sock_create_lite(int family, int type, int protocol, struct socket **res)
1023{
1024        int err;
1025        struct socket *sock = NULL;
1026
1027        err = security_socket_create(family, type, protocol, 1);
1028        if (err)
1029                goto out;
1030
1031        sock = sock_alloc();
1032        if (!sock) {
1033                err = -ENOMEM;
1034                goto out;
1035        }
1036
1037        sock->type = type;
1038        err = security_socket_post_create(sock, family, type, protocol, 1);
1039        if (err)
1040                goto out_release;
1041
1042out:
1043        *res = sock;
1044        return err;
1045out_release:
1046        sock_release(sock);
1047        sock = NULL;
1048        goto out;
1049}
1050
1051/* No kernel lock held - perfect */
1052static unsigned int sock_poll(struct file *file, poll_table *wait)
1053{
1054        struct socket *sock;
1055
1056        /*
1057         *      We can't return errors to poll, so it's either yes or no.
1058         */
1059        sock = file->private_data;
1060        return sock->ops->poll(file, sock, wait);
1061}
1062
1063static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1064{
1065        struct socket *sock = file->private_data;
1066
1067        return sock->ops->mmap(file, sock, vma);
1068}
1069
1070static int sock_close(struct inode *inode, struct file *filp)
1071{
1072        /*
1073         *      It was possible the inode is NULL we were
1074         *      closing an unfinished socket.
1075         */
1076
1077        if (!inode) {
1078                printk(KERN_DEBUG "sock_close: NULL inode\n");
1079                return 0;
1080        }
1081        sock_release(SOCKET_I(inode));
1082        return 0;
1083}
1084
1085/*
1086 *      Update the socket async list
1087 *
1088 *      Fasync_list locking strategy.
1089 *
1090 *      1. fasync_list is modified only under process context socket lock
1091 *         i.e. under semaphore.
1092 *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1093 *         or under socket lock
1094 */
1095
1096static int sock_fasync(int fd, struct file *filp, int on)
1097{
1098        struct socket *sock = filp->private_data;
1099        struct sock *sk = sock->sk;
1100
1101        if (sk == NULL)
1102                return -EINVAL;
1103
1104        lock_sock(sk);
1105
1106        fasync_helper(fd, filp, on, &sock->wq->fasync_list);
1107
1108        if (!sock->wq->fasync_list)
1109                sock_reset_flag(sk, SOCK_FASYNC);
1110        else
1111                sock_set_flag(sk, SOCK_FASYNC);
1112
1113        release_sock(sk);
1114        return 0;
1115}
1116
1117/* This function may be called only under socket lock or callback_lock or rcu_lock */
1118
1119int sock_wake_async(struct socket *sock, int how, int band)
1120{
1121        struct socket_wq *wq;
1122
1123        if (!sock)
1124                return -1;
1125        rcu_read_lock();
1126        wq = rcu_dereference(sock->wq);
1127        if (!wq || !wq->fasync_list) {
1128                rcu_read_unlock();
1129                return -1;
1130        }
1131        switch (how) {
1132        case SOCK_WAKE_WAITD:
1133                if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1134                        break;
1135                goto call_kill;
1136        case SOCK_WAKE_SPACE:
1137                if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1138                        break;
1139                /* fall through */
1140        case SOCK_WAKE_IO:
1141call_kill:
1142                kill_fasync(&wq->fasync_list, SIGIO, band);
1143                break;
1144        case SOCK_WAKE_URG:
1145                kill_fasync(&wq->fasync_list, SIGURG, band);
1146        }
1147        rcu_read_unlock();
1148        return 0;
1149}
1150
1151static int __sock_create(struct net *net, int family, int type, int protocol,
1152                         struct socket **res, int kern)
1153{
1154        int err;
1155        struct socket *sock;
1156        const struct net_proto_family *pf;
1157
1158        /*
1159         *      Check protocol is in range
1160         */
1161        if (family < 0 || family >= NPROTO)
1162                return -EAFNOSUPPORT;
1163        if (type < 0 || type >= SOCK_MAX)
1164                return -EINVAL;
1165
1166        /* Compatibility.
1167
1168           This uglymoron is moved from INET layer to here to avoid
1169           deadlock in module load.
1170         */
1171        if (family == PF_INET && type == SOCK_PACKET) {
1172                static int warned;
1173                if (!warned) {
1174                        warned = 1;
1175                        printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1176                               current->comm);
1177                }
1178                family = PF_PACKET;
1179        }
1180
1181        err = security_socket_create(family, type, protocol, kern);
1182        if (err)
1183                return err;
1184
1185        /*
1186         *      Allocate the socket and allow the family to set things up. if
1187         *      the protocol is 0, the family is instructed to select an appropriate
1188         *      default.
1189         */
1190        sock = sock_alloc();
1191        if (!sock) {
1192                if (net_ratelimit())
1193                        printk(KERN_WARNING "socket: no more sockets\n");
1194                return -ENFILE; /* Not exactly a match, but its the
1195                                   closest posix thing */
1196        }
1197
1198        sock->type = type;
1199
1200#ifdef CONFIG_MODULES
1201        /* Attempt to load a protocol module if the find failed.
1202         *
1203         * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1204         * requested real, full-featured networking support upon configuration.
1205         * Otherwise module support will break!
1206         */
1207        if (net_families[family] == NULL)
1208                request_module("net-pf-%d", family);
1209#endif
1210
1211        rcu_read_lock();
1212        pf = rcu_dereference(net_families[family]);
1213        err = -EAFNOSUPPORT;
1214        if (!pf)
1215                goto out_release;
1216
1217        /*
1218         * We will call the ->create function, that possibly is in a loadable
1219         * module, so we have to bump that loadable module refcnt first.
1220         */
1221        if (!try_module_get(pf->owner))
1222                goto out_release;
1223
1224        /* Now protected by module ref count */
1225        rcu_read_unlock();
1226
1227        err = pf->create(net, sock, protocol, kern);
1228        if (err < 0)
1229                goto out_module_put;
1230
1231        /*
1232         * Now to bump the refcnt of the [loadable] module that owns this
1233         * socket at sock_release time we decrement its refcnt.
1234         */
1235        if (!try_module_get(sock->ops->owner))
1236                goto out_module_busy;
1237
1238        /*
1239         * Now that we're done with the ->create function, the [loadable]
1240         * module can have its refcnt decremented
1241         */
1242        module_put(pf->owner);
1243        err = security_socket_post_create(sock, family, type, protocol, kern);
1244        if (err)
1245                goto out_sock_release;
1246        *res = sock;
1247
1248        return 0;
1249
1250out_module_busy:
1251        err = -EAFNOSUPPORT;
1252out_module_put:
1253        sock->ops = NULL;
1254        module_put(pf->owner);
1255out_sock_release:
1256        sock_release(sock);
1257        return err;
1258
1259out_release:
1260        rcu_read_unlock();
1261        goto out_sock_release;
1262}
1263
1264int sock_create(int family, int type, int protocol, struct socket **res)
1265{
1266        return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1267}
1268
1269int sock_create_kern(int family, int type, int protocol, struct socket **res)
1270{
1271        return __sock_create(&init_net, family, type, protocol, res, 1);
1272}
1273
1274SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1275{
1276        int retval;
1277        struct socket *sock;
1278        int flags;
1279
1280        /* Check the SOCK_* constants for consistency.  */
1281        BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1282        BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1283        BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1284        BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1285
1286        flags = type & ~SOCK_TYPE_MASK;
1287        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1288                return -EINVAL;
1289        type &= SOCK_TYPE_MASK;
1290
1291        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1292                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1293
1294        retval = sock_create(family, type, protocol, &sock);
1295        if (retval < 0)
1296                goto out;
1297
1298        retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1299        if (retval < 0)
1300                goto out_release;
1301
1302out:
1303        /* It may be already another descriptor 8) Not kernel problem. */
1304        return retval;
1305
1306out_release:
1307        sock_release(sock);
1308        return retval;
1309}
1310
1311/*
1312 *      Create a pair of connected sockets.
1313 */
1314
1315SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1316                int __user *, usockvec)
1317{
1318        struct socket *sock1, *sock2;
1319        int fd1, fd2, err;
1320        struct file *newfile1, *newfile2;
1321        int flags;
1322
1323        flags = type & ~SOCK_TYPE_MASK;
1324        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1325                return -EINVAL;
1326        type &= SOCK_TYPE_MASK;
1327
1328        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1329                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1330
1331        /*
1332         * Obtain the first socket and check if the underlying protocol
1333         * supports the socketpair call.
1334         */
1335
1336        err = sock_create(family, type, protocol, &sock1);
1337        if (err < 0)
1338                goto out;
1339
1340        err = sock_create(family, type, protocol, &sock2);
1341        if (err < 0)
1342                goto out_release_1;
1343
1344        err = sock1->ops->socketpair(sock1, sock2);
1345        if (err < 0)
1346                goto out_release_both;
1347
1348        fd1 = sock_alloc_file(sock1, &newfile1, flags);
1349        if (unlikely(fd1 < 0)) {
1350                err = fd1;
1351                goto out_release_both;
1352        }
1353
1354        fd2 = sock_alloc_file(sock2, &newfile2, flags);
1355        if (unlikely(fd2 < 0)) {
1356                err = fd2;
1357                fput(newfile1);
1358                put_unused_fd(fd1);
1359                sock_release(sock2);
1360                goto out;
1361        }
1362
1363        audit_fd_pair(fd1, fd2);
1364        fd_install(fd1, newfile1);
1365        fd_install(fd2, newfile2);
1366        /* fd1 and fd2 may be already another descriptors.
1367         * Not kernel problem.
1368         */
1369
1370        err = put_user(fd1, &usockvec[0]);
1371        if (!err)
1372                err = put_user(fd2, &usockvec[1]);
1373        if (!err)
1374                return 0;
1375
1376        sys_close(fd2);
1377        sys_close(fd1);
1378        return err;
1379
1380out_release_both:
1381        sock_release(sock2);
1382out_release_1:
1383        sock_release(sock1);
1384out:
1385        return err;
1386}
1387
1388/*
1389 *      Bind a name to a socket. Nothing much to do here since it's
1390 *      the protocol's responsibility to handle the local address.
1391 *
1392 *      We move the socket address to kernel space before we call
1393 *      the protocol layer (having also checked the address is ok).
1394 */
1395
1396SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1397{
1398        struct socket *sock;
1399        struct sockaddr_storage address;
1400        int err, fput_needed;
1401
1402        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1403        if (sock) {
1404                err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1405                if (err >= 0) {
1406                        err = security_socket_bind(sock,
1407                                                   (struct sockaddr *)&address,
1408                                                   addrlen);
1409                        if (!err)
1410                                err = sock->ops->bind(sock,
1411                                                      (struct sockaddr *)
1412                                                      &address, addrlen);
1413                }
1414                fput_light(sock->file, fput_needed);
1415        }
1416        return err;
1417}
1418
1419/*
1420 *      Perform a listen. Basically, we allow the protocol to do anything
1421 *      necessary for a listen, and if that works, we mark the socket as
1422 *      ready for listening.
1423 */
1424
1425SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1426{
1427        struct socket *sock;
1428        int err, fput_needed;
1429        int somaxconn;
1430
1431        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1432        if (sock) {
1433                somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1434                if ((unsigned)backlog > somaxconn)
1435                        backlog = somaxconn;
1436
1437                err = security_socket_listen(sock, backlog);
1438                if (!err)
1439                        err = sock->ops->listen(sock, backlog);
1440
1441                fput_light(sock->file, fput_needed);
1442        }
1443        return err;
1444}
1445
1446/*
1447 *      For accept, we attempt to create a new socket, set up the link
1448 *      with the client, wake up the client, then return the new
1449 *      connected fd. We collect the address of the connector in kernel
1450 *      space and move it to user at the very end. This is unclean because
1451 *      we open the socket then return an error.
1452 *
1453 *      1003.1g adds the ability to recvmsg() to query connection pending
1454 *      status to recvmsg. We need to add that support in a way thats
1455 *      clean when we restucture accept also.
1456 */
1457
1458SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1459                int __user *, upeer_addrlen, int, flags)
1460{
1461        struct socket *sock, *newsock;
1462        struct file *newfile;
1463        int err, len, newfd, fput_needed;
1464        struct sockaddr_storage address;
1465
1466        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1467                return -EINVAL;
1468
1469        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1470                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1471
1472        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1473        if (!sock)
1474                goto out;
1475
1476        err = -ENFILE;
1477        if (!(newsock = sock_alloc()))
1478                goto out_put;
1479
1480        newsock->type = sock->type;
1481        newsock->ops = sock->ops;
1482
1483        /*
1484         * We don't need try_module_get here, as the listening socket (sock)
1485         * has the protocol module (sock->ops->owner) held.
1486         */
1487        __module_get(newsock->ops->owner);
1488
1489        newfd = sock_alloc_file(newsock, &newfile, flags);
1490        if (unlikely(newfd < 0)) {
1491                err = newfd;
1492                sock_release(newsock);
1493                goto out_put;
1494        }
1495
1496        err = security_socket_accept(sock, newsock);
1497        if (err)
1498                goto out_fd;
1499
1500        err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1501        if (err < 0)
1502                goto out_fd;
1503
1504        if (upeer_sockaddr) {
1505                if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1506                                          &len, 2) < 0) {
1507                        err = -ECONNABORTED;
1508                        goto out_fd;
1509                }
1510                err = move_addr_to_user((struct sockaddr *)&address,
1511                                        len, upeer_sockaddr, upeer_addrlen);
1512                if (err < 0)
1513                        goto out_fd;
1514        }
1515
1516        /* File flags are not inherited via accept() unlike another OSes. */
1517
1518        fd_install(newfd, newfile);
1519        err = newfd;
1520
1521out_put:
1522        fput_light(sock->file, fput_needed);
1523out:
1524        return err;
1525out_fd:
1526        fput(newfile);
1527        put_unused_fd(newfd);
1528        goto out_put;
1529}
1530
1531SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1532                int __user *, upeer_addrlen)
1533{
1534        return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1535}
1536
1537/*
1538 *      Attempt to connect to a socket with the server address.  The address
1539 *      is in user space so we verify it is OK and move it to kernel space.
1540 *
1541 *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1542 *      break bindings
1543 *
1544 *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1545 *      other SEQPACKET protocols that take time to connect() as it doesn't
1546 *      include the -EINPROGRESS status for such sockets.
1547 */
1548
1549SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1550                int, addrlen)
1551{
1552        struct socket *sock;
1553        struct sockaddr_storage address;
1554        int err, fput_needed;
1555
1556        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1557        if (!sock)
1558                goto out;
1559        err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1560        if (err < 0)
1561                goto out_put;
1562
1563        err =
1564            security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1565        if (err)
1566                goto out_put;
1567
1568        err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1569                                 sock->file->f_flags);
1570out_put:
1571        fput_light(sock->file, fput_needed);
1572out:
1573        return err;
1574}
1575
1576/*
1577 *      Get the local address ('name') of a socket object. Move the obtained
1578 *      name to user space.
1579 */
1580
1581SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1582                int __user *, usockaddr_len)
1583{
1584        struct socket *sock;
1585        struct sockaddr_storage address;
1586        int len, err, fput_needed;
1587
1588        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1589        if (!sock)
1590                goto out;
1591
1592        err = security_socket_getsockname(sock);
1593        if (err)
1594                goto out_put;
1595
1596        err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1597        if (err)
1598                goto out_put;
1599        err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1600
1601out_put:
1602        fput_light(sock->file, fput_needed);
1603out:
1604        return err;
1605}
1606
1607/*
1608 *      Get the remote address ('name') of a socket object. Move the obtained
1609 *      name to user space.
1610 */
1611
1612SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1613                int __user *, usockaddr_len)
1614{
1615        struct socket *sock;
1616        struct sockaddr_storage address;
1617        int len, err, fput_needed;
1618
1619        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1620        if (sock != NULL) {
1621                err = security_socket_getpeername(sock);
1622                if (err) {
1623                        fput_light(sock->file, fput_needed);
1624                        return err;
1625                }
1626
1627                err =
1628                    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1629                                       1);
1630                if (!err)
1631                        err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1632                                                usockaddr_len);
1633                fput_light(sock->file, fput_needed);
1634        }
1635        return err;
1636}
1637
1638/*
1639 *      Send a datagram to a given address. We move the address into kernel
1640 *      space and check the user space data area is readable before invoking
1641 *      the protocol.
1642 */
1643
1644SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1645                unsigned, flags, struct sockaddr __user *, addr,
1646                int, addr_len)
1647{
1648        struct socket *sock;
1649        struct sockaddr_storage address;
1650        int err;
1651        struct msghdr msg;
1652        struct iovec iov;
1653        int fput_needed;
1654
1655        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1656        if (!sock)
1657                goto out;
1658
1659        iov.iov_base = buff;
1660        iov.iov_len = len;
1661        msg.msg_name = NULL;
1662        msg.msg_iov = &iov;
1663        msg.msg_iovlen = 1;
1664        msg.msg_control = NULL;
1665        msg.msg_controllen = 0;
1666        msg.msg_namelen = 0;
1667        if (addr) {
1668                err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1669                if (err < 0)
1670                        goto out_put;
1671                msg.msg_name = (struct sockaddr *)&address;
1672                msg.msg_namelen = addr_len;
1673        }
1674        if (sock->file->f_flags & O_NONBLOCK)
1675                flags |= MSG_DONTWAIT;
1676        msg.msg_flags = flags;
1677        err = sock_sendmsg(sock, &msg, len);
1678
1679out_put:
1680        fput_light(sock->file, fput_needed);
1681out:
1682        return err;
1683}
1684
1685/*
1686 *      Send a datagram down a socket.
1687 */
1688
1689SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1690                unsigned, flags)
1691{
1692        return sys_sendto(fd, buff, len, flags, NULL, 0);
1693}
1694
1695/*
1696 *      Receive a frame from the socket and optionally record the address of the
1697 *      sender. We verify the buffers are writable and if needed move the
1698 *      sender address from kernel to user space.
1699 */
1700
1701SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1702                unsigned, flags, struct sockaddr __user *, addr,
1703                int __user *, addr_len)
1704{
1705        struct socket *sock;
1706        struct iovec iov;
1707        struct msghdr msg;
1708        struct sockaddr_storage address;
1709        int err, err2;
1710        int fput_needed;
1711
1712        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1713        if (!sock)
1714                goto out;
1715
1716        msg.msg_control = NULL;
1717        msg.msg_controllen = 0;
1718        msg.msg_iovlen = 1;
1719        msg.msg_iov = &iov;
1720        iov.iov_len = size;
1721        iov.iov_base = ubuf;
1722        msg.msg_name = (struct sockaddr *)&address;
1723        msg.msg_namelen = sizeof(address);
1724        if (sock->file->f_flags & O_NONBLOCK)
1725                flags |= MSG_DONTWAIT;
1726        err = sock_recvmsg(sock, &msg, size, flags);
1727
1728        if (err >= 0 && addr != NULL) {
1729                err2 = move_addr_to_user((struct sockaddr *)&address,
1730                                         msg.msg_namelen, addr, addr_len);
1731                if (err2 < 0)
1732                        err = err2;
1733        }
1734
1735        fput_light(sock->file, fput_needed);
1736out:
1737        return err;
1738}
1739
1740/*
1741 *      Receive a datagram from a socket.
1742 */
1743
1744asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1745                         unsigned flags)
1746{
1747        return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1748}
1749
1750/*
1751 *      Set a socket option. Because we don't know the option lengths we have
1752 *      to pass the user mode parameter for the protocols to sort out.
1753 */
1754
1755SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1756                char __user *, optval, int, optlen)
1757{
1758        int err, fput_needed;
1759        struct socket *sock;
1760
1761        if (optlen < 0)
1762                return -EINVAL;
1763
1764        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1765        if (sock != NULL) {
1766                err = security_socket_setsockopt(sock, level, optname);
1767                if (err)
1768                        goto out_put;
1769
1770                if (level == SOL_SOCKET)
1771                        err =
1772                            sock_setsockopt(sock, level, optname, optval,
1773                                            optlen);
1774                else
1775                        err =
1776                            sock->ops->setsockopt(sock, level, optname, optval,
1777                                                  optlen);
1778out_put:
1779                fput_light(sock->file, fput_needed);
1780        }
1781        return err;
1782}
1783
1784/*
1785 *      Get a socket option. Because we don't know the option lengths we have
1786 *      to pass a user mode parameter for the protocols to sort out.
1787 */
1788
1789SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1790                char __user *, optval, int __user *, optlen)
1791{
1792        int err, fput_needed;
1793        struct socket *sock;
1794
1795        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1796        if (sock != NULL) {
1797                err = security_socket_getsockopt(sock, level, optname);
1798                if (err)
1799                        goto out_put;
1800
1801                if (level == SOL_SOCKET)
1802                        err =
1803                            sock_getsockopt(sock, level, optname, optval,
1804                                            optlen);
1805                else
1806                        err =
1807                            sock->ops->getsockopt(sock, level, optname, optval,
1808                                                  optlen);
1809out_put:
1810                fput_light(sock->file, fput_needed);
1811        }
1812        return err;
1813}
1814
1815/*
1816 *      Shutdown a socket.
1817 */
1818
1819SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1820{
1821        int err, fput_needed;
1822        struct socket *sock;
1823
1824        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1825        if (sock != NULL) {
1826                err = security_socket_shutdown(sock, how);
1827                if (!err)
1828                        err = sock->ops->shutdown(sock, how);
1829                fput_light(sock->file, fput_needed);
1830        }
1831        return err;
1832}
1833
1834/* A couple of helpful macros for getting the address of the 32/64 bit
1835 * fields which are the same type (int / unsigned) on our platforms.
1836 */
1837#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1838#define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1839#define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1840
1841/*
1842 *      BSD sendmsg interface
1843 */
1844
1845SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1846{
1847        struct compat_msghdr __user *msg_compat =
1848            (struct compat_msghdr __user *)msg;
1849        struct socket *sock;
1850        struct sockaddr_storage address;
1851        struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1852        unsigned char ctl[sizeof(struct cmsghdr) + 20]
1853            __attribute__ ((aligned(sizeof(__kernel_size_t))));
1854        /* 20 is size of ipv6_pktinfo */
1855        unsigned char *ctl_buf = ctl;
1856        struct msghdr msg_sys;
1857        int err, ctl_len, iov_size, total_len;
1858        int fput_needed;
1859
1860        err = -EFAULT;
1861        if (MSG_CMSG_COMPAT & flags) {
1862                if (get_compat_msghdr(&msg_sys, msg_compat))
1863                        return -EFAULT;
1864        }
1865        else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1866                return -EFAULT;
1867
1868        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1869        if (!sock)
1870                goto out;
1871
1872        /* do not move before msg_sys is valid */
1873        err = -EMSGSIZE;
1874        if (msg_sys.msg_iovlen > UIO_MAXIOV)
1875                goto out_put;
1876
1877        /* Check whether to allocate the iovec area */
1878        err = -ENOMEM;
1879        iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1880        if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1881                iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1882                if (!iov)
1883                        goto out_put;
1884        }
1885
1886        /* This will also move the address data into kernel space */
1887        if (MSG_CMSG_COMPAT & flags) {
1888                err = verify_compat_iovec(&msg_sys, iov,
1889                                          (struct sockaddr *)&address,
1890                                          VERIFY_READ);
1891        } else
1892                err = verify_iovec(&msg_sys, iov,
1893                                   (struct sockaddr *)&address,
1894                                   VERIFY_READ);
1895        if (err < 0)
1896                goto out_freeiov;
1897        total_len = err;
1898
1899        err = -ENOBUFS;
1900
1901        if (msg_sys.msg_controllen > INT_MAX)
1902                goto out_freeiov;
1903        ctl_len = msg_sys.msg_controllen;
1904        if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1905                err =
1906                    cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1907                                                     sizeof(ctl));
1908                if (err)
1909                        goto out_freeiov;
1910                ctl_buf = msg_sys.msg_control;
1911                ctl_len = msg_sys.msg_controllen;
1912        } else if (ctl_len) {
1913                if (ctl_len > sizeof(ctl)) {
1914                        ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1915                        if (ctl_buf == NULL)
1916                                goto out_freeiov;
1917                }
1918                err = -EFAULT;
1919                /*
1920                 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1921                 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1922                 * checking falls down on this.
1923                 */
1924                if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1925                                   ctl_len))
1926                        goto out_freectl;
1927                msg_sys.msg_control = ctl_buf;
1928        }
1929        msg_sys.msg_flags = flags;
1930
1931        if (sock->file->f_flags & O_NONBLOCK)
1932                msg_sys.msg_flags |= MSG_DONTWAIT;
1933        err = sock_sendmsg(sock, &msg_sys, total_len);
1934
1935out_freectl:
1936        if (ctl_buf != ctl)
1937                sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1938out_freeiov:
1939        if (iov != iovstack)
1940                sock_kfree_s(sock->sk, iov, iov_size);
1941out_put:
1942        fput_light(sock->file, fput_needed);
1943out:
1944        return err;
1945}
1946
1947static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1948                         struct msghdr *msg_sys, unsigned flags, int nosec)
1949{
1950        struct compat_msghdr __user *msg_compat =
1951            (struct compat_msghdr __user *)msg;
1952        struct iovec iovstack[UIO_FASTIOV];
1953        struct iovec *iov = iovstack;
1954        unsigned long cmsg_ptr;
1955        int err, iov_size, total_len, len;
1956
1957        /* kernel mode address */
1958        struct sockaddr_storage addr;
1959
1960        /* user mode address pointers */
1961        struct sockaddr __user *uaddr;
1962        int __user *uaddr_len;
1963
1964        if (MSG_CMSG_COMPAT & flags) {
1965                if (get_compat_msghdr(msg_sys, msg_compat))
1966                        return -EFAULT;
1967        }
1968        else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1969                return -EFAULT;
1970
1971        err = -EMSGSIZE;
1972        if (msg_sys->msg_iovlen > UIO_MAXIOV)
1973                goto out;
1974
1975        /* Check whether to allocate the iovec area */
1976        err = -ENOMEM;
1977        iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1978        if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1979                iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1980                if (!iov)
1981                        goto out;
1982        }
1983
1984        /*
1985         *      Save the user-mode address (verify_iovec will change the
1986         *      kernel msghdr to use the kernel address space)
1987         */
1988
1989        uaddr = (__force void __user *)msg_sys->msg_name;
1990        uaddr_len = COMPAT_NAMELEN(msg);
1991        if (MSG_CMSG_COMPAT & flags) {
1992                err = verify_compat_iovec(msg_sys, iov,
1993                                          (struct sockaddr *)&addr,
1994                                          VERIFY_WRITE);
1995        } else
1996                err = verify_iovec(msg_sys, iov,
1997                                   (struct sockaddr *)&addr,
1998                                   VERIFY_WRITE);
1999        if (err < 0)
2000                goto out_freeiov;
2001        total_len = err;
2002
2003        cmsg_ptr = (unsigned long)msg_sys->msg_control;
2004        msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2005
2006        if (sock->file->f_flags & O_NONBLOCK)
2007                flags |= MSG_DONTWAIT;
2008        err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2009                                                          total_len, flags);
2010        if (err < 0)
2011                goto out_freeiov;
2012        len = err;
2013
2014        if (uaddr != NULL) {
2015                err = move_addr_to_user((struct sockaddr *)&addr,
2016                                        msg_sys->msg_namelen, uaddr,
2017                                        uaddr_len);
2018                if (err < 0)
2019                        goto out_freeiov;
2020        }
2021        err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2022                         COMPAT_FLAGS(msg));
2023        if (err)
2024                goto out_freeiov;
2025        if (MSG_CMSG_COMPAT & flags)
2026                err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2027                                 &msg_compat->msg_controllen);
2028        else
2029                err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2030                                 &msg->msg_controllen);
2031        if (err)
2032                goto out_freeiov;
2033        err = len;
2034
2035out_freeiov:
2036        if (iov != iovstack)
2037                sock_kfree_s(sock->sk, iov, iov_size);
2038out:
2039        return err;
2040}
2041
2042/*
2043 *      BSD recvmsg interface
2044 */
2045
2046SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2047                unsigned int, flags)
2048{
2049        int fput_needed, err;
2050        struct msghdr msg_sys;
2051        struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2052
2053        if (!sock)
2054                goto out;
2055
2056        err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2057
2058        fput_light(sock->file, fput_needed);
2059out:
2060        return err;
2061}
2062
2063/*
2064 *     Linux recvmmsg interface
2065 */
2066
2067int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2068                   unsigned int flags, struct timespec *timeout)
2069{
2070        int fput_needed, err, datagrams;
2071        struct socket *sock;
2072        struct mmsghdr __user *entry;
2073        struct compat_mmsghdr __user *compat_entry;
2074        struct msghdr msg_sys;
2075        struct timespec end_time;
2076
2077        if (timeout &&
2078            poll_select_set_timeout(&end_time, timeout->tv_sec,
2079                                    timeout->tv_nsec))
2080                return -EINVAL;
2081
2082        datagrams = 0;
2083
2084        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2085        if (!sock)
2086                return err;
2087
2088        err = sock_error(sock->sk);
2089        if (err)
2090                goto out_put;
2091
2092        entry = mmsg;
2093        compat_entry = (struct compat_mmsghdr __user *)mmsg;
2094
2095        while (datagrams < vlen) {
2096                /*
2097                 * No need to ask LSM for more than the first datagram.
2098                 */
2099                if (MSG_CMSG_COMPAT & flags) {
2100                        err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2101                                            &msg_sys, flags, datagrams);
2102                        if (err < 0)
2103                                break;
2104                        err = __put_user(err, &compat_entry->msg_len);
2105                        ++compat_entry;
2106                } else {
2107                        err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2108                                            &msg_sys, flags, datagrams);
2109                        if (err < 0)
2110                                break;
2111                        err = put_user(err, &entry->msg_len);
2112                        ++entry;
2113                }
2114
2115                if (err)
2116                        break;
2117                ++datagrams;
2118
2119                /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2120                if (flags & MSG_WAITFORONE)
2121                        flags |= MSG_DONTWAIT;
2122
2123                if (timeout) {
2124                        ktime_get_ts(timeout);
2125                        *timeout = timespec_sub(end_time, *timeout);
2126                        if (timeout->tv_sec < 0) {
2127                                timeout->tv_sec = timeout->tv_nsec = 0;
2128                                break;
2129                        }
2130
2131                        /* Timeout, return less than vlen datagrams */
2132                        if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2133                                break;
2134                }
2135
2136                /* Out of band data, return right away */
2137                if (msg_sys.msg_flags & MSG_OOB)
2138                        break;
2139        }
2140
2141out_put:
2142        fput_light(sock->file, fput_needed);
2143
2144        if (err == 0)
2145                return datagrams;
2146
2147        if (datagrams != 0) {
2148                /*
2149                 * We may return less entries than requested (vlen) if the
2150                 * sock is non block and there aren't enough datagrams...
2151                 */
2152                if (err != -EAGAIN) {
2153                        /*
2154                         * ... or  if recvmsg returns an error after we
2155                         * received some datagrams, where we record the
2156                         * error to return on the next call or if the
2157                         * app asks about it using getsockopt(SO_ERROR).
2158                         */
2159                        sock->sk->sk_err = -err;
2160                }
2161
2162                return datagrams;
2163        }
2164
2165        return err;
2166}
2167
2168SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2169                unsigned int, vlen, unsigned int, flags,
2170                struct timespec __user *, timeout)
2171{
2172        int datagrams;
2173        struct timespec timeout_sys;
2174
2175        if (!timeout)
2176                return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2177
2178        if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2179                return -EFAULT;
2180
2181        datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2182
2183        if (datagrams > 0 &&
2184            copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2185                datagrams = -EFAULT;
2186
2187        return datagrams;
2188}
2189
2190#ifdef __ARCH_WANT_SYS_SOCKETCALL
2191/* Argument list sizes for sys_socketcall */
2192#define AL(x) ((x) * sizeof(unsigned long))
2193static const unsigned char nargs[20] = {
2194        AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2195        AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2196        AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2197        AL(4),AL(5)
2198};
2199
2200#undef AL
2201
2202/*
2203 *      System call vectors.
2204 *
2205 *      Argument checking cleaned up. Saved 20% in size.
2206 *  This function doesn't need to set the kernel lock because
2207 *  it is set by the callees.
2208 */
2209
2210SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2211{
2212        unsigned long a[6];
2213        unsigned long a0, a1;
2214        int err;
2215        unsigned int len;
2216
2217        if (call < 1 || call > SYS_RECVMMSG)
2218                return -EINVAL;
2219
2220        len = nargs[call];
2221        if (len > sizeof(a))
2222                return -EINVAL;
2223
2224        /* copy_from_user should be SMP safe. */
2225        if (copy_from_user(a, args, len))
2226                return -EFAULT;
2227
2228        audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2229
2230        a0 = a[0];
2231        a1 = a[1];
2232
2233        switch (call) {
2234        case SYS_SOCKET:
2235                err = sys_socket(a0, a1, a[2]);
2236                break;
2237        case SYS_BIND:
2238                err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2239                break;
2240        case SYS_CONNECT:
2241                err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2242                break;
2243        case SYS_LISTEN:
2244                err = sys_listen(a0, a1);
2245                break;
2246        case SYS_ACCEPT:
2247                err = sys_accept4(a0, (struct sockaddr __user *)a1,
2248                                  (int __user *)a[2], 0);
2249                break;
2250        case SYS_GETSOCKNAME:
2251                err =
2252                    sys_getsockname(a0, (struct sockaddr __user *)a1,
2253                                    (int __user *)a[2]);
2254                break;
2255        case SYS_GETPEERNAME:
2256                err =
2257                    sys_getpeername(a0, (struct sockaddr __user *)a1,
2258                                    (int __user *)a[2]);
2259                break;
2260        case SYS_SOCKETPAIR:
2261                err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2262                break;
2263        case SYS_SEND:
2264                err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2265                break;
2266        case SYS_SENDTO:
2267                err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2268                                 (struct sockaddr __user *)a[4], a[5]);
2269                break;
2270        case SYS_RECV:
2271                err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2272                break;
2273        case SYS_RECVFROM:
2274                err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2275                                   (struct sockaddr __user *)a[4],
2276                                   (int __user *)a[5]);
2277                break;
2278        case SYS_SHUTDOWN:
2279                err = sys_shutdown(a0, a1);
2280                break;
2281        case SYS_SETSOCKOPT:
2282                err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2283                break;
2284        case SYS_GETSOCKOPT:
2285                err =
2286                    sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2287                                   (int __user *)a[4]);
2288                break;
2289        case SYS_SENDMSG:
2290                err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2291                break;
2292        case SYS_RECVMSG:
2293                err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2294                break;
2295        case SYS_RECVMMSG:
2296                err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2297                                   (struct timespec __user *)a[4]);
2298                break;
2299        case SYS_ACCEPT4:
2300                err = sys_accept4(a0, (struct sockaddr __user *)a1,
2301                                  (int __user *)a[2], a[3]);
2302                break;
2303        default:
2304                err = -EINVAL;
2305                break;
2306        }
2307        return err;
2308}
2309
2310#endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2311
2312/**
2313 *      sock_register - add a socket protocol handler
2314 *      @ops: description of protocol
2315 *
2316 *      This function is called by a protocol handler that wants to
2317 *      advertise its address family, and have it linked into the
2318 *      socket interface. The value ops->family coresponds to the
2319 *      socket system call protocol family.
2320 */
2321int sock_register(const struct net_proto_family *ops)
2322{
2323        int err;
2324
2325        if (ops->family >= NPROTO) {
2326                printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2327                       NPROTO);
2328                return -ENOBUFS;
2329        }
2330
2331        spin_lock(&net_family_lock);
2332        if (net_families[ops->family])
2333                err = -EEXIST;
2334        else {
2335                net_families[ops->family] = ops;
2336                err = 0;
2337        }
2338        spin_unlock(&net_family_lock);
2339
2340        printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2341        return err;
2342}
2343
2344/**
2345 *      sock_unregister - remove a protocol handler
2346 *      @family: protocol family to remove
2347 *
2348 *      This function is called by a protocol handler that wants to
2349 *      remove its address family, and have it unlinked from the
2350 *      new socket creation.
2351 *
2352 *      If protocol handler is a module, then it can use module reference
2353 *      counts to protect against new references. If protocol handler is not
2354 *      a module then it needs to provide its own protection in
2355 *      the ops->create routine.
2356 */
2357void sock_unregister(int family)
2358{
2359        BUG_ON(family < 0 || family >= NPROTO);
2360
2361        spin_lock(&net_family_lock);
2362        net_families[family] = NULL;
2363        spin_unlock(&net_family_lock);
2364
2365        synchronize_rcu();
2366
2367        printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2368}
2369
2370static int __init sock_init(void)
2371{
2372        /*
2373         *      Initialize sock SLAB cache.
2374         */
2375
2376        sk_init();
2377
2378        /*
2379         *      Initialize skbuff SLAB cache
2380         */
2381        skb_init();
2382
2383        /*
2384         *      Initialize the protocols module.
2385         */
2386
2387        init_inodecache();
2388        register_filesystem(&sock_fs_type);
2389        sock_mnt = kern_mount(&sock_fs_type);
2390
2391        /* The real protocol initialization is performed in later initcalls.
2392         */
2393
2394#ifdef CONFIG_NETFILTER
2395        netfilter_init();
2396#endif
2397
2398        return 0;
2399}
2400
2401core_initcall(sock_init);       /* early initcall */
2402
2403#ifdef CONFIG_PROC_FS
2404void socket_seq_show(struct seq_file *seq)
2405{
2406        int cpu;
2407        int counter = 0;
2408
2409        for_each_possible_cpu(cpu)
2410            counter += per_cpu(sockets_in_use, cpu);
2411
2412        /* It can be negative, by the way. 8) */
2413        if (counter < 0)
2414                counter = 0;
2415
2416        seq_printf(seq, "sockets: used %d\n", counter);
2417}
2418#endif                          /* CONFIG_PROC_FS */
2419
2420#ifdef CONFIG_COMPAT
2421static int do_siocgstamp(struct net *net, struct socket *sock,
2422                         unsigned int cmd, struct compat_timeval __user *up)
2423{
2424        mm_segment_t old_fs = get_fs();
2425        struct timeval ktv;
2426        int err;
2427
2428        set_fs(KERNEL_DS);
2429        err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2430        set_fs(old_fs);
2431        if (!err) {
2432                err = put_user(ktv.tv_sec, &up->tv_sec);
2433                err |= __put_user(ktv.tv_usec, &up->tv_usec);
2434        }
2435        return err;
2436}
2437
2438static int do_siocgstampns(struct net *net, struct socket *sock,
2439                         unsigned int cmd, struct compat_timespec __user *up)
2440{
2441        mm_segment_t old_fs = get_fs();
2442        struct timespec kts;
2443        int err;
2444
2445        set_fs(KERNEL_DS);
2446        err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2447        set_fs(old_fs);
2448        if (!err) {
2449                err = put_user(kts.tv_sec, &up->tv_sec);
2450                err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2451        }
2452        return err;
2453}
2454
2455static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2456{
2457        struct ifreq __user *uifr;
2458        int err;
2459
2460        uifr = compat_alloc_user_space(sizeof(struct ifreq));
2461        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2462                return -EFAULT;
2463
2464        err = dev_ioctl(net, SIOCGIFNAME, uifr);
2465        if (err)
2466                return err;
2467
2468        if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2469                return -EFAULT;
2470
2471        return 0;
2472}
2473
2474static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2475{
2476        struct compat_ifconf ifc32;
2477        struct ifconf ifc;
2478        struct ifconf __user *uifc;
2479        struct compat_ifreq __user *ifr32;
2480        struct ifreq __user *ifr;
2481        unsigned int i, j;
2482        int err;
2483
2484        if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2485                return -EFAULT;
2486
2487        if (ifc32.ifcbuf == 0) {
2488                ifc32.ifc_len = 0;
2489                ifc.ifc_len = 0;
2490                ifc.ifc_req = NULL;
2491                uifc = compat_alloc_user_space(sizeof(struct ifconf));
2492        } else {
2493                size_t len =((ifc32.ifc_len / sizeof (struct compat_ifreq)) + 1) *
2494                        sizeof (struct ifreq);
2495                uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2496                ifc.ifc_len = len;
2497                ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2498                ifr32 = compat_ptr(ifc32.ifcbuf);
2499                for (i = 0; i < ifc32.ifc_len; i += sizeof (struct compat_ifreq)) {
2500                        if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2501                                return -EFAULT;
2502                        ifr++;
2503                        ifr32++;
2504                }
2505        }
2506        if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2507                return -EFAULT;
2508
2509        err = dev_ioctl(net, SIOCGIFCONF, uifc);
2510        if (err)
2511                return err;
2512
2513        if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2514                return -EFAULT;
2515
2516        ifr = ifc.ifc_req;
2517        ifr32 = compat_ptr(ifc32.ifcbuf);
2518        for (i = 0, j = 0;
2519             i + sizeof (struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2520             i += sizeof (struct compat_ifreq), j += sizeof (struct ifreq)) {
2521                if (copy_in_user(ifr32, ifr, sizeof (struct compat_ifreq)))
2522                        return -EFAULT;
2523                ifr32++;
2524                ifr++;
2525        }
2526
2527        if (ifc32.ifcbuf == 0) {
2528                /* Translate from 64-bit structure multiple to
2529                 * a 32-bit one.
2530                 */
2531                i = ifc.ifc_len;
2532                i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2533                ifc32.ifc_len = i;
2534        } else {
2535                ifc32.ifc_len = i;
2536        }
2537        if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2538                return -EFAULT;
2539
2540        return 0;
2541}
2542
2543static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2544{
2545        struct ifreq __user *ifr;
2546        u32 data;
2547        void __user *datap;
2548
2549        ifr = compat_alloc_user_space(sizeof(*ifr));
2550
2551        if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2552                return -EFAULT;
2553
2554        if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2555                return -EFAULT;
2556
2557        datap = compat_ptr(data);
2558        if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2559                return -EFAULT;
2560
2561        return dev_ioctl(net, SIOCETHTOOL, ifr);
2562}
2563
2564static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2565{
2566        void __user *uptr;
2567        compat_uptr_t uptr32;
2568        struct ifreq __user *uifr;
2569
2570        uifr = compat_alloc_user_space(sizeof (*uifr));
2571        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2572                return -EFAULT;
2573
2574        if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2575                return -EFAULT;
2576
2577        uptr = compat_ptr(uptr32);
2578
2579        if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2580                return -EFAULT;
2581
2582        return dev_ioctl(net, SIOCWANDEV, uifr);
2583}
2584
2585static int bond_ioctl(struct net *net, unsigned int cmd,
2586                         struct compat_ifreq __user *ifr32)
2587{
2588        struct ifreq kifr;
2589        struct ifreq __user *uifr;
2590        mm_segment_t old_fs;
2591        int err;
2592        u32 data;
2593        void __user *datap;
2594
2595        switch (cmd) {
2596        case SIOCBONDENSLAVE:
2597        case SIOCBONDRELEASE:
2598        case SIOCBONDSETHWADDR:
2599        case SIOCBONDCHANGEACTIVE:
2600                if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2601                        return -EFAULT;
2602
2603                old_fs = get_fs();
2604                set_fs (KERNEL_DS);
2605                err = dev_ioctl(net, cmd, &kifr);
2606                set_fs (old_fs);
2607
2608                return err;
2609        case SIOCBONDSLAVEINFOQUERY:
2610        case SIOCBONDINFOQUERY:
2611                uifr = compat_alloc_user_space(sizeof(*uifr));
2612                if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2613                        return -EFAULT;
2614
2615                if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2616                        return -EFAULT;
2617
2618                datap = compat_ptr(data);
2619                if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2620                        return -EFAULT;
2621
2622                return dev_ioctl(net, cmd, uifr);
2623        default:
2624                return -EINVAL;
2625        }
2626}
2627
2628static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2629                                 struct compat_ifreq __user *u_ifreq32)
2630{
2631        struct ifreq __user *u_ifreq64;
2632        char tmp_buf[IFNAMSIZ];
2633        void __user *data64;
2634        u32 data32;
2635
2636        if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2637                           IFNAMSIZ))
2638                return -EFAULT;
2639        if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2640                return -EFAULT;
2641        data64 = compat_ptr(data32);
2642
2643        u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2644
2645        /* Don't check these user accesses, just let that get trapped
2646         * in the ioctl handler instead.
2647         */
2648        if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2649                         IFNAMSIZ))
2650                return -EFAULT;
2651        if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2652                return -EFAULT;
2653
2654        return dev_ioctl(net, cmd, u_ifreq64);
2655}
2656
2657static int dev_ifsioc(struct net *net, struct socket *sock,
2658                         unsigned int cmd, struct compat_ifreq __user *uifr32)
2659{
2660        struct ifreq __user *uifr;
2661        int err;
2662
2663        uifr = compat_alloc_user_space(sizeof(*uifr));
2664        if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2665                return -EFAULT;
2666
2667        err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2668
2669        if (!err) {
2670                switch (cmd) {
2671                case SIOCGIFFLAGS:
2672                case SIOCGIFMETRIC:
2673                case SIOCGIFMTU:
2674                case SIOCGIFMEM:
2675                case SIOCGIFHWADDR:
2676                case SIOCGIFINDEX:
2677                case SIOCGIFADDR:
2678                case SIOCGIFBRDADDR:
2679                case SIOCGIFDSTADDR:
2680                case SIOCGIFNETMASK:
2681                case SIOCGIFPFLAGS:
2682                case SIOCGIFTXQLEN:
2683                case SIOCGMIIPHY:
2684                case SIOCGMIIREG:
2685                        if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2686                                err = -EFAULT;
2687                        break;
2688                }
2689        }
2690        return err;
2691}
2692
2693static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2694                        struct compat_ifreq __user *uifr32)
2695{
2696        struct ifreq ifr;
2697        struct compat_ifmap __user *uifmap32;
2698        mm_segment_t old_fs;
2699        int err;
2700
2701        uifmap32 = &uifr32->ifr_ifru.ifru_map;
2702        err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2703        err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2704        err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2705        err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2706        err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2707        err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2708        err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2709        if (err)
2710                return -EFAULT;
2711
2712        old_fs = get_fs();
2713        set_fs (KERNEL_DS);
2714        err = dev_ioctl(net, cmd, (void __user *)&ifr);
2715        set_fs (old_fs);
2716
2717        if (cmd == SIOCGIFMAP && !err) {
2718                err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2719                err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2720                err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2721                err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2722                err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2723                err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2724                err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2725                if (err)
2726                        err = -EFAULT;
2727        }
2728        return err;
2729}
2730
2731static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2732{
2733        void __user *uptr;
2734        compat_uptr_t uptr32;
2735        struct ifreq __user *uifr;
2736
2737        uifr = compat_alloc_user_space(sizeof (*uifr));
2738        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2739                return -EFAULT;
2740
2741        if (get_user(uptr32, &uifr32->ifr_data))
2742                return -EFAULT;
2743
2744        uptr = compat_ptr(uptr32);
2745
2746        if (put_user(uptr, &uifr->ifr_data))
2747                return -EFAULT;
2748
2749        return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2750}
2751
2752struct rtentry32 {
2753        u32             rt_pad1;
2754        struct sockaddr rt_dst;         /* target address               */
2755        struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
2756        struct sockaddr rt_genmask;     /* target network mask (IP)     */
2757        unsigned short  rt_flags;
2758        short           rt_pad2;
2759        u32             rt_pad3;
2760        unsigned char   rt_tos;
2761        unsigned char   rt_class;
2762        short           rt_pad4;
2763        short           rt_metric;      /* +1 for binary compatibility! */
2764        /* char * */ u32 rt_dev;        /* forcing the device at add    */
2765        u32             rt_mtu;         /* per route MTU/Window         */
2766        u32             rt_window;      /* Window clamping              */
2767        unsigned short  rt_irtt;        /* Initial RTT                  */
2768};
2769
2770struct in6_rtmsg32 {
2771        struct in6_addr         rtmsg_dst;
2772        struct in6_addr         rtmsg_src;
2773        struct in6_addr         rtmsg_gateway;
2774        u32                     rtmsg_type;
2775        u16                     rtmsg_dst_len;
2776        u16                     rtmsg_src_len;
2777        u32                     rtmsg_metric;
2778        u32                     rtmsg_info;
2779        u32                     rtmsg_flags;
2780        s32                     rtmsg_ifindex;
2781};
2782
2783static int routing_ioctl(struct net *net, struct socket *sock,
2784                         unsigned int cmd, void __user *argp)
2785{
2786        int ret;
2787        void *r = NULL;
2788        struct in6_rtmsg r6;
2789        struct rtentry r4;
2790        char devname[16];
2791        u32 rtdev;
2792        mm_segment_t old_fs = get_fs();
2793
2794        if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2795                struct in6_rtmsg32 __user *ur6 = argp;
2796                ret = copy_from_user (&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2797                        3 * sizeof(struct in6_addr));
2798                ret |= __get_user (r6.rtmsg_type, &(ur6->rtmsg_type));
2799                ret |= __get_user (r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2800                ret |= __get_user (r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2801                ret |= __get_user (r6.rtmsg_metric, &(ur6->rtmsg_metric));
2802                ret |= __get_user (r6.rtmsg_info, &(ur6->rtmsg_info));
2803                ret |= __get_user (r6.rtmsg_flags, &(ur6->rtmsg_flags));
2804                ret |= __get_user (r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2805
2806                r = (void *) &r6;
2807        } else { /* ipv4 */
2808                struct rtentry32 __user *ur4 = argp;
2809                ret = copy_from_user (&r4.rt_dst, &(ur4->rt_dst),
2810                                        3 * sizeof(struct sockaddr));
2811                ret |= __get_user (r4.rt_flags, &(ur4->rt_flags));
2812                ret |= __get_user (r4.rt_metric, &(ur4->rt_metric));
2813                ret |= __get_user (r4.rt_mtu, &(ur4->rt_mtu));
2814                ret |= __get_user (r4.rt_window, &(ur4->rt_window));
2815                ret |= __get_user (r4.rt_irtt, &(ur4->rt_irtt));
2816                ret |= __get_user (rtdev, &(ur4->rt_dev));
2817                if (rtdev) {
2818                        ret |= copy_from_user (devname, compat_ptr(rtdev), 15);
2819                        r4.rt_dev = devname; devname[15] = 0;
2820                } else
2821                        r4.rt_dev = NULL;
2822
2823                r = (void *) &r4;
2824        }
2825
2826        if (ret) {
2827                ret = -EFAULT;
2828                goto out;
2829        }
2830
2831        set_fs (KERNEL_DS);
2832        ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
2833        set_fs (old_fs);
2834
2835out:
2836        return ret;
2837}
2838
2839/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2840 * for some operations; this forces use of the newer bridge-utils that
2841 * use compatiable ioctls
2842 */
2843static int old_bridge_ioctl(compat_ulong_t __user *argp)
2844{
2845        compat_ulong_t tmp;
2846
2847        if (get_user(tmp, argp))
2848                return -EFAULT;
2849        if (tmp == BRCTL_GET_VERSION)
2850                return BRCTL_VERSION + 1;
2851        return -EINVAL;
2852}
2853
2854static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2855                         unsigned int cmd, unsigned long arg)
2856{
2857        void __user *argp = compat_ptr(arg);
2858        struct sock *sk = sock->sk;
2859        struct net *net = sock_net(sk);
2860
2861        if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2862                return siocdevprivate_ioctl(net, cmd, argp);
2863
2864        switch (cmd) {
2865        case SIOCSIFBR:
2866        case SIOCGIFBR:
2867                return old_bridge_ioctl(argp);
2868        case SIOCGIFNAME:
2869                return dev_ifname32(net, argp);
2870        case SIOCGIFCONF:
2871                return dev_ifconf(net, argp);
2872        case SIOCETHTOOL:
2873                return ethtool_ioctl(net, argp);
2874        case SIOCWANDEV:
2875                return compat_siocwandev(net, argp);
2876        case SIOCGIFMAP:
2877        case SIOCSIFMAP:
2878                return compat_sioc_ifmap(net, cmd, argp);
2879        case SIOCBONDENSLAVE:
2880        case SIOCBONDRELEASE:
2881        case SIOCBONDSETHWADDR:
2882        case SIOCBONDSLAVEINFOQUERY:
2883        case SIOCBONDINFOQUERY:
2884        case SIOCBONDCHANGEACTIVE:
2885                return bond_ioctl(net, cmd, argp);
2886        case SIOCADDRT:
2887        case SIOCDELRT:
2888                return routing_ioctl(net, sock, cmd, argp);
2889        case SIOCGSTAMP:
2890                return do_siocgstamp(net, sock, cmd, argp);
2891        case SIOCGSTAMPNS:
2892                return do_siocgstampns(net, sock, cmd, argp);
2893        case SIOCSHWTSTAMP:
2894                return compat_siocshwtstamp(net, argp);
2895
2896        case FIOSETOWN:
2897        case SIOCSPGRP:
2898        case FIOGETOWN:
2899        case SIOCGPGRP:
2900        case SIOCBRADDBR:
2901        case SIOCBRDELBR:
2902        case SIOCGIFVLAN:
2903        case SIOCSIFVLAN:
2904        case SIOCADDDLCI:
2905        case SIOCDELDLCI:
2906                return sock_ioctl(file, cmd, arg);
2907
2908        case SIOCGIFFLAGS:
2909        case SIOCSIFFLAGS:
2910        case SIOCGIFMETRIC:
2911        case SIOCSIFMETRIC:
2912        case SIOCGIFMTU:
2913        case SIOCSIFMTU:
2914        case SIOCGIFMEM:
2915        case SIOCSIFMEM:
2916        case SIOCGIFHWADDR:
2917        case SIOCSIFHWADDR:
2918        case SIOCADDMULTI:
2919        case SIOCDELMULTI:
2920        case SIOCGIFINDEX:
2921        case SIOCGIFADDR:
2922        case SIOCSIFADDR:
2923        case SIOCSIFHWBROADCAST:
2924        case SIOCDIFADDR:
2925        case SIOCGIFBRDADDR:
2926        case SIOCSIFBRDADDR:
2927        case SIOCGIFDSTADDR:
2928        case SIOCSIFDSTADDR:
2929        case SIOCGIFNETMASK:
2930        case SIOCSIFNETMASK:
2931        case SIOCSIFPFLAGS:
2932        case SIOCGIFPFLAGS:
2933        case SIOCGIFTXQLEN:
2934        case SIOCSIFTXQLEN:
2935        case SIOCBRADDIF:
2936        case SIOCBRDELIF:
2937        case SIOCSIFNAME:
2938        case SIOCGMIIPHY:
2939        case SIOCGMIIREG:
2940        case SIOCSMIIREG:
2941                return dev_ifsioc(net, sock, cmd, argp);
2942
2943        case SIOCSARP:
2944        case SIOCGARP:
2945        case SIOCDARP:
2946        case SIOCATMARK:
2947                return sock_do_ioctl(net, sock, cmd, arg);
2948        }
2949
2950        /* Prevent warning from compat_sys_ioctl, these always
2951         * result in -EINVAL in the native case anyway. */
2952        switch (cmd) {
2953        case SIOCRTMSG:
2954        case SIOCGIFCOUNT:
2955        case SIOCSRARP:
2956        case SIOCGRARP:
2957        case SIOCDRARP:
2958        case SIOCSIFLINK:
2959        case SIOCGIFSLAVE:
2960        case SIOCSIFSLAVE:
2961                return -EINVAL;
2962        }
2963
2964        return -ENOIOCTLCMD;
2965}
2966
2967static long compat_sock_ioctl(struct file *file, unsigned cmd,
2968                              unsigned long arg)
2969{
2970        struct socket *sock = file->private_data;
2971        int ret = -ENOIOCTLCMD;
2972        struct sock *sk;
2973        struct net *net;
2974
2975        sk = sock->sk;
2976        net = sock_net(sk);
2977
2978        if (sock->ops->compat_ioctl)
2979                ret = sock->ops->compat_ioctl(sock, cmd, arg);
2980
2981        if (ret == -ENOIOCTLCMD &&
2982            (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2983                ret = compat_wext_handle_ioctl(net, cmd, arg);
2984
2985        if (ret == -ENOIOCTLCMD)
2986                ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
2987
2988        return ret;
2989}
2990#endif
2991
2992int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2993{
2994        return sock->ops->bind(sock, addr, addrlen);
2995}
2996
2997int kernel_listen(struct socket *sock, int backlog)
2998{
2999        return sock->ops->listen(sock, backlog);
3000}
3001
3002int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3003{
3004        struct sock *sk = sock->sk;
3005        int err;
3006
3007        err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3008                               newsock);
3009        if (err < 0)
3010                goto done;
3011
3012        err = sock->ops->accept(sock, *newsock, flags);
3013        if (err < 0) {
3014                sock_release(*newsock);
3015                *newsock = NULL;
3016                goto done;
3017        }
3018
3019        (*newsock)->ops = sock->ops;
3020        __module_get((*newsock)->ops->owner);
3021
3022done:
3023        return err;
3024}
3025
3026int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3027                   int flags)
3028{
3029        return sock->ops->connect(sock, addr, addrlen, flags);
3030}
3031
3032int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3033                         int *addrlen)
3034{
3035        return sock->ops->getname(sock, addr, addrlen, 0);
3036}
3037
3038int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3039                         int *addrlen)
3040{
3041        return sock->ops->getname(sock, addr, addrlen, 1);
3042}
3043
3044int kernel_getsockopt(struct socket *sock, int level, int optname,
3045                        char *optval, int *optlen)
3046{
3047        mm_segment_t oldfs = get_fs();
3048        int err;
3049
3050        set_fs(KERNEL_DS);
3051        if (level == SOL_SOCKET)
3052                err = sock_getsockopt(sock, level, optname, optval, optlen);
3053        else
3054                err = sock->ops->getsockopt(sock, level, optname, optval,
3055                                            optlen);
3056        set_fs(oldfs);
3057        return err;
3058}
3059
3060int kernel_setsockopt(struct socket *sock, int level, int optname,
3061                        char *optval, unsigned int optlen)
3062{
3063        mm_segment_t oldfs = get_fs();
3064        int err;
3065
3066        set_fs(KERNEL_DS);
3067        if (level == SOL_SOCKET)
3068                err = sock_setsockopt(sock, level, optname, optval, optlen);
3069        else
3070                err = sock->ops->setsockopt(sock, level, optname, optval,
3071                                            optlen);
3072        set_fs(oldfs);
3073        return err;
3074}
3075
3076int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3077                    size_t size, int flags)
3078{
3079        sock_update_classid(sock->sk);
3080
3081        if (sock->ops->sendpage)
3082                return sock->ops->sendpage(sock, page, offset, size, flags);
3083
3084        return sock_no_sendpage(sock, page, offset, size, flags);
3085}
3086
3087int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3088{
3089        mm_segment_t oldfs = get_fs();
3090        int err;
3091
3092        set_fs(KERNEL_DS);
3093        err = sock->ops->ioctl(sock, cmd, arg);
3094        set_fs(oldfs);
3095
3096        return err;
3097}
3098
3099int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3100{
3101        return sock->ops->shutdown(sock, how);
3102}
3103
3104EXPORT_SYMBOL(sock_create);
3105EXPORT_SYMBOL(sock_create_kern);
3106EXPORT_SYMBOL(sock_create_lite);
3107EXPORT_SYMBOL(sock_map_fd);
3108EXPORT_SYMBOL(sock_recvmsg);
3109EXPORT_SYMBOL(sock_register);
3110EXPORT_SYMBOL(sock_release);
3111EXPORT_SYMBOL(sock_sendmsg);
3112EXPORT_SYMBOL(sock_unregister);
3113EXPORT_SYMBOL(sock_wake_async);
3114EXPORT_SYMBOL(sockfd_lookup);
3115EXPORT_SYMBOL(kernel_sendmsg);
3116EXPORT_SYMBOL(kernel_recvmsg);
3117EXPORT_SYMBOL(kernel_bind);
3118EXPORT_SYMBOL(kernel_listen);
3119EXPORT_SYMBOL(kernel_accept);
3120EXPORT_SYMBOL(kernel_connect);
3121EXPORT_SYMBOL(kernel_getsockname);
3122EXPORT_SYMBOL(kernel_getpeername);
3123EXPORT_SYMBOL(kernel_getsockopt);
3124EXPORT_SYMBOL(kernel_setsockopt);
3125EXPORT_SYMBOL(kernel_sendpage);
3126EXPORT_SYMBOL(kernel_sock_ioctl);
3127EXPORT_SYMBOL(kernel_sock_shutdown);
3128