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);
 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/**
 174 *      move_addr_to_kernel     -       copy a socket address into kernel space
 175 *      @uaddr: Address in user space
 176 *      @kaddr: Address in kernel space
 177 *      @ulen: Length in user space
 178 *
 179 *      The address is copied into kernel space. If the provided address is
 180 *      too long an error code of -EINVAL is returned. If the copy gives
 181 *      invalid addresses -EFAULT is returned. On a success 0 is returned.
 182 */
 183
 184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
 185{
 186        if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
 187                return -EINVAL;
 188        if (ulen == 0)
 189                return 0;
 190        if (copy_from_user(kaddr, uaddr, ulen))
 191                return -EFAULT;
 192        return audit_sockaddr(ulen, kaddr);
 193}
 194
 195/**
 196 *      move_addr_to_user       -       copy an address to user space
 197 *      @kaddr: kernel space address
 198 *      @klen: length of address in kernel
 199 *      @uaddr: user space address
 200 *      @ulen: pointer to user length field
 201 *
 202 *      The value pointed to by ulen on entry is the buffer length available.
 203 *      This is overwritten with the buffer space used. -EINVAL is returned
 204 *      if an overlong buffer is specified or a negative buffer size. -EFAULT
 205 *      is returned if either the buffer or the length field are not
 206 *      accessible.
 207 *      After copying the data up to the limit the user specifies, the true
 208 *      length of the data is written over the length limit the user
 209 *      specified. Zero is returned for a success.
 210 */
 211
 212static int move_addr_to_user(struct sockaddr *kaddr, int klen,
 213                             void __user *uaddr, int __user *ulen)
 214{
 215        int err;
 216        int len;
 217
 218        err = get_user(len, ulen);
 219        if (err)
 220                return err;
 221        if (len > klen)
 222                len = klen;
 223        if (len < 0 || len > sizeof(struct sockaddr_storage))
 224                return -EINVAL;
 225        if (len) {
 226                if (audit_sockaddr(klen, kaddr))
 227                        return -ENOMEM;
 228                if (copy_to_user(uaddr, kaddr, len))
 229                        return -EFAULT;
 230        }
 231        /*
 232         *      "fromlen shall refer to the value before truncation.."
 233         *                      1003.1g
 234         */
 235        return __put_user(klen, ulen);
 236}
 237
 238static struct kmem_cache *sock_inode_cachep __read_mostly;
 239
 240static struct inode *sock_alloc_inode(struct super_block *sb)
 241{
 242        struct socket_alloc *ei;
 243
 244        ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
 245        if (!ei)
 246                return NULL;
 247        ei->socket.wq = kmalloc(sizeof(struct socket_wq), GFP_KERNEL);
 248        if (!ei->socket.wq) {
 249                kmem_cache_free(sock_inode_cachep, ei);
 250                return NULL;
 251        }
 252        init_waitqueue_head(&ei->socket.wq->wait);
 253        ei->socket.wq->fasync_list = NULL;
 254
 255        ei->socket.state = SS_UNCONNECTED;
 256        ei->socket.flags = 0;
 257        ei->socket.ops = NULL;
 258        ei->socket.sk = NULL;
 259        ei->socket.file = NULL;
 260
 261        return &ei->vfs_inode;
 262}
 263
 264
 265static void wq_free_rcu(struct rcu_head *head)
 266{
 267        struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
 268
 269        kfree(wq);
 270}
 271
 272static void sock_destroy_inode(struct inode *inode)
 273{
 274        struct socket_alloc *ei;
 275
 276        ei = container_of(inode, struct socket_alloc, vfs_inode);
 277        call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
 278        kmem_cache_free(sock_inode_cachep, ei);
 279}
 280
 281static void init_once(void *foo)
 282{
 283        struct socket_alloc *ei = (struct socket_alloc *)foo;
 284
 285        inode_init_once(&ei->vfs_inode);
 286}
 287
 288static int init_inodecache(void)
 289{
 290        sock_inode_cachep = kmem_cache_create("sock_inode_cache",
 291                                              sizeof(struct socket_alloc),
 292                                              0,
 293                                              (SLAB_HWCACHE_ALIGN |
 294                                               SLAB_RECLAIM_ACCOUNT |
 295                                               SLAB_MEM_SPREAD),
 296                                              init_once);
 297        if (sock_inode_cachep == NULL)
 298                return -ENOMEM;
 299        return 0;
 300}
 301
 302static const struct super_operations sockfs_ops = {
 303        .alloc_inode    = sock_alloc_inode,
 304        .destroy_inode  = sock_destroy_inode,
 305        .statfs         = simple_statfs,
 306};
 307
 308static struct dentry *sockfs_mount(struct file_system_type *fs_type,
 309                         int flags, const char *dev_name, void *data)
 310{
 311        return mount_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
 312}
 313
 314static struct vfsmount *sock_mnt __read_mostly;
 315
 316static struct file_system_type sock_fs_type = {
 317        .name =         "sockfs",
 318        .mount =        sockfs_mount,
 319        .kill_sb =      kill_anon_super,
 320};
 321
 322/*
 323 * sockfs_dname() is called from d_path().
 324 */
 325static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
 326{
 327        return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
 328                                dentry->d_inode->i_ino);
 329}
 330
 331static const struct dentry_operations sockfs_dentry_operations = {
 332        .d_dname  = sockfs_dname,
 333};
 334
 335/*
 336 *      Obtains the first available file descriptor and sets it up for use.
 337 *
 338 *      These functions create file structures and maps them to fd space
 339 *      of the current process. On success it returns file descriptor
 340 *      and file struct implicitly stored in sock->file.
 341 *      Note that another thread may close file descriptor before we return
 342 *      from this function. We use the fact that now we do not refer
 343 *      to socket after mapping. If one day we will need it, this
 344 *      function will increment ref. count on file by 1.
 345 *
 346 *      In any case returned fd MAY BE not valid!
 347 *      This race condition is unavoidable
 348 *      with shared fd spaces, we cannot solve it inside kernel,
 349 *      but we take care of internal coherence yet.
 350 */
 351
 352static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
 353{
 354        struct qstr name = { .name = "" };
 355        struct path path;
 356        struct file *file;
 357        int fd;
 358
 359        fd = get_unused_fd_flags(flags);
 360        if (unlikely(fd < 0))
 361                return fd;
 362
 363        path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
 364        if (unlikely(!path.dentry)) {
 365                put_unused_fd(fd);
 366                return -ENOMEM;
 367        }
 368        path.mnt = mntget(sock_mnt);
 369
 370        path.dentry->d_op = &sockfs_dentry_operations;
 371        d_instantiate(path.dentry, SOCK_INODE(sock));
 372        SOCK_INODE(sock)->i_fop = &socket_file_ops;
 373
 374        file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
 375                  &socket_file_ops);
 376        if (unlikely(!file)) {
 377                /* drop dentry, keep inode */
 378                ihold(path.dentry->d_inode);
 379                path_put(&path);
 380                put_unused_fd(fd);
 381                return -ENFILE;
 382        }
 383
 384        sock->file = file;
 385        file->f_flags = O_RDWR | (flags & O_NONBLOCK);
 386        file->f_pos = 0;
 387        file->private_data = sock;
 388
 389        *f = file;
 390        return fd;
 391}
 392
 393int sock_map_fd(struct socket *sock, int flags)
 394{
 395        struct file *newfile;
 396        int fd = sock_alloc_file(sock, &newfile, flags);
 397
 398        if (likely(fd >= 0))
 399                fd_install(fd, newfile);
 400
 401        return fd;
 402}
 403EXPORT_SYMBOL(sock_map_fd);
 404
 405static struct socket *sock_from_file(struct file *file, int *err)
 406{
 407        if (file->f_op == &socket_file_ops)
 408                return file->private_data;      /* set in sock_map_fd */
 409
 410        *err = -ENOTSOCK;
 411        return NULL;
 412}
 413
 414/**
 415 *      sockfd_lookup - Go from a file number to its socket slot
 416 *      @fd: file handle
 417 *      @err: pointer to an error code return
 418 *
 419 *      The file handle passed in is locked and the socket it is bound
 420 *      too is returned. If an error occurs the err pointer is overwritten
 421 *      with a negative errno code and NULL is returned. The function checks
 422 *      for both invalid handles and passing a handle which is not a socket.
 423 *
 424 *      On a success the socket object pointer is returned.
 425 */
 426
 427struct socket *sockfd_lookup(int fd, int *err)
 428{
 429        struct file *file;
 430        struct socket *sock;
 431
 432        file = fget(fd);
 433        if (!file) {
 434                *err = -EBADF;
 435                return NULL;
 436        }
 437
 438        sock = sock_from_file(file, err);
 439        if (!sock)
 440                fput(file);
 441        return sock;
 442}
 443EXPORT_SYMBOL(sockfd_lookup);
 444
 445static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 446{
 447        struct file *file;
 448        struct socket *sock;
 449
 450        *err = -EBADF;
 451        file = fget_light(fd, fput_needed);
 452        if (file) {
 453                sock = sock_from_file(file, err);
 454                if (sock)
 455                        return sock;
 456                fput_light(file, *fput_needed);
 457        }
 458        return NULL;
 459}
 460
 461/**
 462 *      sock_alloc      -       allocate a socket
 463 *
 464 *      Allocate a new inode and socket object. The two are bound together
 465 *      and initialised. The socket is then returned. If we are out of inodes
 466 *      NULL is returned.
 467 */
 468
 469static struct socket *sock_alloc(void)
 470{
 471        struct inode *inode;
 472        struct socket *sock;
 473
 474        inode = new_inode(sock_mnt->mnt_sb);
 475        if (!inode)
 476                return NULL;
 477
 478        sock = SOCKET_I(inode);
 479
 480        kmemcheck_annotate_bitfield(sock, type);
 481        inode->i_ino = get_next_ino();
 482        inode->i_mode = S_IFSOCK | S_IRWXUGO;
 483        inode->i_uid = current_fsuid();
 484        inode->i_gid = current_fsgid();
 485
 486        percpu_add(sockets_in_use, 1);
 487        return sock;
 488}
 489
 490/*
 491 *      In theory you can't get an open on this inode, but /proc provides
 492 *      a back door. Remember to keep it shut otherwise you'll let the
 493 *      creepy crawlies in.
 494 */
 495
 496static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
 497{
 498        return -ENXIO;
 499}
 500
 501const struct file_operations bad_sock_fops = {
 502        .owner = THIS_MODULE,
 503        .open = sock_no_open,
 504        .llseek = noop_llseek,
 505};
 506
 507/**
 508 *      sock_release    -       close a socket
 509 *      @sock: socket to close
 510 *
 511 *      The socket is released from the protocol stack if it has a release
 512 *      callback, and the inode is then released if the socket is bound to
 513 *      an inode not a file.
 514 */
 515
 516void sock_release(struct socket *sock)
 517{
 518        if (sock->ops) {
 519                struct module *owner = sock->ops->owner;
 520
 521                sock->ops->release(sock);
 522                sock->ops = NULL;
 523                module_put(owner);
 524        }
 525
 526        if (sock->wq->fasync_list)
 527                printk(KERN_ERR "sock_release: fasync list not empty!\n");
 528
 529        percpu_sub(sockets_in_use, 1);
 530        if (!sock->file) {
 531                iput(SOCK_INODE(sock));
 532                return;
 533        }
 534        sock->file = NULL;
 535}
 536EXPORT_SYMBOL(sock_release);
 537
 538int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
 539{
 540        *tx_flags = 0;
 541        if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
 542                *tx_flags |= SKBTX_HW_TSTAMP;
 543        if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
 544                *tx_flags |= SKBTX_SW_TSTAMP;
 545        return 0;
 546}
 547EXPORT_SYMBOL(sock_tx_timestamp);
 548
 549static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 550                                 struct msghdr *msg, size_t size)
 551{
 552        struct sock_iocb *si = kiocb_to_siocb(iocb);
 553        int err;
 554
 555        sock_update_classid(sock->sk);
 556
 557        si->sock = sock;
 558        si->scm = NULL;
 559        si->msg = msg;
 560        si->size = size;
 561
 562        err = security_socket_sendmsg(sock, msg, size);
 563        if (err)
 564                return err;
 565
 566        return sock->ops->sendmsg(iocb, sock, msg, size);
 567}
 568
 569int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 570{
 571        struct kiocb iocb;
 572        struct sock_iocb siocb;
 573        int ret;
 574
 575        init_sync_kiocb(&iocb, NULL);
 576        iocb.private = &siocb;
 577        ret = __sock_sendmsg(&iocb, sock, msg, size);
 578        if (-EIOCBQUEUED == ret)
 579                ret = wait_on_sync_kiocb(&iocb);
 580        return ret;
 581}
 582EXPORT_SYMBOL(sock_sendmsg);
 583
 584int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 585                   struct kvec *vec, size_t num, size_t size)
 586{
 587        mm_segment_t oldfs = get_fs();
 588        int result;
 589
 590        set_fs(KERNEL_DS);
 591        /*
 592         * the following is safe, since for compiler definitions of kvec and
 593         * iovec are identical, yielding the same in-core layout and alignment
 594         */
 595        msg->msg_iov = (struct iovec *)vec;
 596        msg->msg_iovlen = num;
 597        result = sock_sendmsg(sock, msg, size);
 598        set_fs(oldfs);
 599        return result;
 600}
 601EXPORT_SYMBOL(kernel_sendmsg);
 602
 603static int ktime2ts(ktime_t kt, struct timespec *ts)
 604{
 605        if (kt.tv64) {
 606                *ts = ktime_to_timespec(kt);
 607                return 1;
 608        } else {
 609                return 0;
 610        }
 611}
 612
 613/*
 614 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 615 */
 616void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 617        struct sk_buff *skb)
 618{
 619        int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 620        struct timespec ts[3];
 621        int empty = 1;
 622        struct skb_shared_hwtstamps *shhwtstamps =
 623                skb_hwtstamps(skb);
 624
 625        /* Race occurred between timestamp enabling and packet
 626           receiving.  Fill in the current time for now. */
 627        if (need_software_tstamp && skb->tstamp.tv64 == 0)
 628                __net_timestamp(skb);
 629
 630        if (need_software_tstamp) {
 631                if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 632                        struct timeval tv;
 633                        skb_get_timestamp(skb, &tv);
 634                        put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 635                                 sizeof(tv), &tv);
 636                } else {
 637                        skb_get_timestampns(skb, &ts[0]);
 638                        put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 639                                 sizeof(ts[0]), &ts[0]);
 640                }
 641        }
 642
 643
 644        memset(ts, 0, sizeof(ts));
 645        if (skb->tstamp.tv64 &&
 646            sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
 647                skb_get_timestampns(skb, ts + 0);
 648                empty = 0;
 649        }
 650        if (shhwtstamps) {
 651                if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 652                    ktime2ts(shhwtstamps->syststamp, ts + 1))
 653                        empty = 0;
 654                if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 655                    ktime2ts(shhwtstamps->hwtstamp, ts + 2))
 656                        empty = 0;
 657        }
 658        if (!empty)
 659                put_cmsg(msg, SOL_SOCKET,
 660                         SCM_TIMESTAMPING, sizeof(ts), &ts);
 661}
 662EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 663
 664static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 665                                   struct sk_buff *skb)
 666{
 667        if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 668                put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 669                        sizeof(__u32), &skb->dropcount);
 670}
 671
 672void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 673        struct sk_buff *skb)
 674{
 675        sock_recv_timestamp(msg, sk, skb);
 676        sock_recv_drops(msg, sk, skb);
 677}
 678EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 679
 680static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 681                                       struct msghdr *msg, size_t size, int flags)
 682{
 683        struct sock_iocb *si = kiocb_to_siocb(iocb);
 684
 685        sock_update_classid(sock->sk);
 686
 687        si->sock = sock;
 688        si->scm = NULL;
 689        si->msg = msg;
 690        si->size = size;
 691        si->flags = flags;
 692
 693        return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 694}
 695
 696static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 697                                 struct msghdr *msg, size_t size, int flags)
 698{
 699        int err = security_socket_recvmsg(sock, msg, size, flags);
 700
 701        return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 702}
 703
 704int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 705                 size_t size, int flags)
 706{
 707        struct kiocb iocb;
 708        struct sock_iocb siocb;
 709        int ret;
 710
 711        init_sync_kiocb(&iocb, NULL);
 712        iocb.private = &siocb;
 713        ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 714        if (-EIOCBQUEUED == ret)
 715                ret = wait_on_sync_kiocb(&iocb);
 716        return ret;
 717}
 718EXPORT_SYMBOL(sock_recvmsg);
 719
 720static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 721                              size_t size, int flags)
 722{
 723        struct kiocb iocb;
 724        struct sock_iocb siocb;
 725        int ret;
 726
 727        init_sync_kiocb(&iocb, NULL);
 728        iocb.private = &siocb;
 729        ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 730        if (-EIOCBQUEUED == ret)
 731                ret = wait_on_sync_kiocb(&iocb);
 732        return ret;
 733}
 734
 735/**
 736 * kernel_recvmsg - Receive a message from a socket (kernel space)
 737 * @sock:       The socket to receive the message from
 738 * @msg:        Received message
 739 * @vec:        Input s/g array for message data
 740 * @num:        Size of input s/g array
 741 * @size:       Number of bytes to read
 742 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 743 *
 744 * On return the msg structure contains the scatter/gather array passed in the
 745 * vec argument. The array is modified so that it consists of the unfilled
 746 * portion of the original array.
 747 *
 748 * The returned value is the total number of bytes received, or an error.
 749 */
 750int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 751                   struct kvec *vec, size_t num, size_t size, int flags)
 752{
 753        mm_segment_t oldfs = get_fs();
 754        int result;
 755
 756        set_fs(KERNEL_DS);
 757        /*
 758         * the following is safe, since for compiler definitions of kvec and
 759         * iovec are identical, yielding the same in-core layout and alignment
 760         */
 761        msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 762        result = sock_recvmsg(sock, msg, size, flags);
 763        set_fs(oldfs);
 764        return result;
 765}
 766EXPORT_SYMBOL(kernel_recvmsg);
 767
 768static void sock_aio_dtor(struct kiocb *iocb)
 769{
 770        kfree(iocb->private);
 771}
 772
 773static ssize_t sock_sendpage(struct file *file, struct page *page,
 774                             int offset, size_t size, loff_t *ppos, int more)
 775{
 776        struct socket *sock;
 777        int flags;
 778
 779        sock = file->private_data;
 780
 781        flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
 782        if (more)
 783                flags |= MSG_MORE;
 784
 785        return kernel_sendpage(sock, page, offset, size, flags);
 786}
 787
 788static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 789                                struct pipe_inode_info *pipe, size_t len,
 790                                unsigned int flags)
 791{
 792        struct socket *sock = file->private_data;
 793
 794        if (unlikely(!sock->ops->splice_read))
 795                return -EINVAL;
 796
 797        sock_update_classid(sock->sk);
 798
 799        return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 800}
 801
 802static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 803                                         struct sock_iocb *siocb)
 804{
 805        if (!is_sync_kiocb(iocb)) {
 806                siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
 807                if (!siocb)
 808                        return NULL;
 809                iocb->ki_dtor = sock_aio_dtor;
 810        }
 811
 812        siocb->kiocb = iocb;
 813        iocb->private = siocb;
 814        return siocb;
 815}
 816
 817static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 818                struct file *file, const struct iovec *iov,
 819                unsigned long nr_segs)
 820{
 821        struct socket *sock = file->private_data;
 822        size_t size = 0;
 823        int i;
 824
 825        for (i = 0; i < nr_segs; i++)
 826                size += iov[i].iov_len;
 827
 828        msg->msg_name = NULL;
 829        msg->msg_namelen = 0;
 830        msg->msg_control = NULL;
 831        msg->msg_controllen = 0;
 832        msg->msg_iov = (struct iovec *)iov;
 833        msg->msg_iovlen = nr_segs;
 834        msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 835
 836        return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 837}
 838
 839static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 840                                unsigned long nr_segs, loff_t pos)
 841{
 842        struct sock_iocb siocb, *x;
 843
 844        if (pos != 0)
 845                return -ESPIPE;
 846
 847        if (iocb->ki_left == 0) /* Match SYS5 behaviour */
 848                return 0;
 849
 850
 851        x = alloc_sock_iocb(iocb, &siocb);
 852        if (!x)
 853                return -ENOMEM;
 854        return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 855}
 856
 857static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 858                        struct file *file, const struct iovec *iov,
 859                        unsigned long nr_segs)
 860{
 861        struct socket *sock = file->private_data;
 862        size_t size = 0;
 863        int i;
 864
 865        for (i = 0; i < nr_segs; i++)
 866                size += iov[i].iov_len;
 867
 868        msg->msg_name = NULL;
 869        msg->msg_namelen = 0;
 870        msg->msg_control = NULL;
 871        msg->msg_controllen = 0;
 872        msg->msg_iov = (struct iovec *)iov;
 873        msg->msg_iovlen = nr_segs;
 874        msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 875        if (sock->type == SOCK_SEQPACKET)
 876                msg->msg_flags |= MSG_EOR;
 877
 878        return __sock_sendmsg(iocb, sock, msg, size);
 879}
 880
 881static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 882                          unsigned long nr_segs, loff_t pos)
 883{
 884        struct sock_iocb siocb, *x;
 885
 886        if (pos != 0)
 887                return -ESPIPE;
 888
 889        x = alloc_sock_iocb(iocb, &siocb);
 890        if (!x)
 891                return -ENOMEM;
 892
 893        return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 894}
 895
 896/*
 897 * Atomic setting of ioctl hooks to avoid race
 898 * with module unload.
 899 */
 900
 901static DEFINE_MUTEX(br_ioctl_mutex);
 902static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 903
 904void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 905{
 906        mutex_lock(&br_ioctl_mutex);
 907        br_ioctl_hook = hook;
 908        mutex_unlock(&br_ioctl_mutex);
 909}
 910EXPORT_SYMBOL(brioctl_set);
 911
 912static DEFINE_MUTEX(vlan_ioctl_mutex);
 913static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 914
 915void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 916{
 917        mutex_lock(&vlan_ioctl_mutex);
 918        vlan_ioctl_hook = hook;
 919        mutex_unlock(&vlan_ioctl_mutex);
 920}
 921EXPORT_SYMBOL(vlan_ioctl_set);
 922
 923static DEFINE_MUTEX(dlci_ioctl_mutex);
 924static int (*dlci_ioctl_hook) (unsigned int, void __user *);
 925
 926void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
 927{
 928        mutex_lock(&dlci_ioctl_mutex);
 929        dlci_ioctl_hook = hook;
 930        mutex_unlock(&dlci_ioctl_mutex);
 931}
 932EXPORT_SYMBOL(dlci_ioctl_set);
 933
 934static long sock_do_ioctl(struct net *net, struct socket *sock,
 935                                 unsigned int cmd, unsigned long arg)
 936{
 937        int err;
 938        void __user *argp = (void __user *)arg;
 939
 940        err = sock->ops->ioctl(sock, cmd, arg);
 941
 942        /*
 943         * If this ioctl is unknown try to hand it down
 944         * to the NIC driver.
 945         */
 946        if (err == -ENOIOCTLCMD)
 947                err = dev_ioctl(net, cmd, argp);
 948
 949        return err;
 950}
 951
 952/*
 953 *      With an ioctl, arg may well be a user mode pointer, but we don't know
 954 *      what to do with it - that's up to the protocol still.
 955 */
 956
 957static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 958{
 959        struct socket *sock;
 960        struct sock *sk;
 961        void __user *argp = (void __user *)arg;
 962        int pid, err;
 963        struct net *net;
 964
 965        sock = file->private_data;
 966        sk = sock->sk;
 967        net = sock_net(sk);
 968        if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
 969                err = dev_ioctl(net, cmd, argp);
 970        } else
 971#ifdef CONFIG_WEXT_CORE
 972        if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
 973                err = dev_ioctl(net, cmd, argp);
 974        } else
 975#endif
 976                switch (cmd) {
 977                case FIOSETOWN:
 978                case SIOCSPGRP:
 979                        err = -EFAULT;
 980                        if (get_user(pid, (int __user *)argp))
 981                                break;
 982                        err = f_setown(sock->file, pid, 1);
 983                        break;
 984                case FIOGETOWN:
 985                case SIOCGPGRP:
 986                        err = put_user(f_getown(sock->file),
 987                                       (int __user *)argp);
 988                        break;
 989                case SIOCGIFBR:
 990                case SIOCSIFBR:
 991                case SIOCBRADDBR:
 992                case SIOCBRDELBR:
 993                        err = -ENOPKG;
 994                        if (!br_ioctl_hook)
 995                                request_module("bridge");
 996
 997                        mutex_lock(&br_ioctl_mutex);
 998                        if (br_ioctl_hook)
 999                                err = br_ioctl_hook(net, cmd, argp);
1000                        mutex_unlock(&br_ioctl_mutex);
1001                        break;
1002                case SIOCGIFVLAN:
1003                case SIOCSIFVLAN:
1004                        err = -ENOPKG;
1005                        if (!vlan_ioctl_hook)
1006                                request_module("8021q");
1007
1008                        mutex_lock(&vlan_ioctl_mutex);
1009                        if (vlan_ioctl_hook)
1010                                err = vlan_ioctl_hook(net, argp);
1011                        mutex_unlock(&vlan_ioctl_mutex);
1012                        break;
1013                case SIOCADDDLCI:
1014                case SIOCDELDLCI:
1015                        err = -ENOPKG;
1016                        if (!dlci_ioctl_hook)
1017                                request_module("dlci");
1018
1019                        mutex_lock(&dlci_ioctl_mutex);
1020                        if (dlci_ioctl_hook)
1021                                err = dlci_ioctl_hook(cmd, argp);
1022                        mutex_unlock(&dlci_ioctl_mutex);
1023                        break;
1024                default:
1025                        err = sock_do_ioctl(net, sock, cmd, arg);
1026                        break;
1027                }
1028        return err;
1029}
1030
1031int sock_create_lite(int family, int type, int protocol, struct socket **res)
1032{
1033        int err;
1034        struct socket *sock = NULL;
1035
1036        err = security_socket_create(family, type, protocol, 1);
1037        if (err)
1038                goto out;
1039
1040        sock = sock_alloc();
1041        if (!sock) {
1042                err = -ENOMEM;
1043                goto out;
1044        }
1045
1046        sock->type = type;
1047        err = security_socket_post_create(sock, family, type, protocol, 1);
1048        if (err)
1049                goto out_release;
1050
1051out:
1052        *res = sock;
1053        return err;
1054out_release:
1055        sock_release(sock);
1056        sock = NULL;
1057        goto out;
1058}
1059EXPORT_SYMBOL(sock_create_lite);
1060
1061/* No kernel lock held - perfect */
1062static unsigned int sock_poll(struct file *file, poll_table *wait)
1063{
1064        struct socket *sock;
1065
1066        /*
1067         *      We can't return errors to poll, so it's either yes or no.
1068         */
1069        sock = file->private_data;
1070        return sock->ops->poll(file, sock, wait);
1071}
1072
1073static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1074{
1075        struct socket *sock = file->private_data;
1076
1077        return sock->ops->mmap(file, sock, vma);
1078}
1079
1080static int sock_close(struct inode *inode, struct file *filp)
1081{
1082        /*
1083         *      It was possible the inode is NULL we were
1084         *      closing an unfinished socket.
1085         */
1086
1087        if (!inode) {
1088                printk(KERN_DEBUG "sock_close: NULL inode\n");
1089                return 0;
1090        }
1091        sock_release(SOCKET_I(inode));
1092        return 0;
1093}
1094
1095/*
1096 *      Update the socket async list
1097 *
1098 *      Fasync_list locking strategy.
1099 *
1100 *      1. fasync_list is modified only under process context socket lock
1101 *         i.e. under semaphore.
1102 *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1103 *         or under socket lock
1104 */
1105
1106static int sock_fasync(int fd, struct file *filp, int on)
1107{
1108        struct socket *sock = filp->private_data;
1109        struct sock *sk = sock->sk;
1110
1111        if (sk == NULL)
1112                return -EINVAL;
1113
1114        lock_sock(sk);
1115
1116        fasync_helper(fd, filp, on, &sock->wq->fasync_list);
1117
1118        if (!sock->wq->fasync_list)
1119                sock_reset_flag(sk, SOCK_FASYNC);
1120        else
1121                sock_set_flag(sk, SOCK_FASYNC);
1122
1123        release_sock(sk);
1124        return 0;
1125}
1126
1127/* This function may be called only under socket lock or callback_lock or rcu_lock */
1128
1129int sock_wake_async(struct socket *sock, int how, int band)
1130{
1131        struct socket_wq *wq;
1132
1133        if (!sock)
1134                return -1;
1135        rcu_read_lock();
1136        wq = rcu_dereference(sock->wq);
1137        if (!wq || !wq->fasync_list) {
1138                rcu_read_unlock();
1139                return -1;
1140        }
1141        switch (how) {
1142        case SOCK_WAKE_WAITD:
1143                if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1144                        break;
1145                goto call_kill;
1146        case SOCK_WAKE_SPACE:
1147                if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1148                        break;
1149                /* fall through */
1150        case SOCK_WAKE_IO:
1151call_kill:
1152                kill_fasync(&wq->fasync_list, SIGIO, band);
1153                break;
1154        case SOCK_WAKE_URG:
1155                kill_fasync(&wq->fasync_list, SIGURG, band);
1156        }
1157        rcu_read_unlock();
1158        return 0;
1159}
1160EXPORT_SYMBOL(sock_wake_async);
1161
1162int __sock_create(struct net *net, int family, int type, int protocol,
1163                         struct socket **res, int kern)
1164{
1165        int err;
1166        struct socket *sock;
1167        const struct net_proto_family *pf;
1168
1169        /*
1170         *      Check protocol is in range
1171         */
1172        if (family < 0 || family >= NPROTO)
1173                return -EAFNOSUPPORT;
1174        if (type < 0 || type >= SOCK_MAX)
1175                return -EINVAL;
1176
1177        /* Compatibility.
1178
1179           This uglymoron is moved from INET layer to here to avoid
1180           deadlock in module load.
1181         */
1182        if (family == PF_INET && type == SOCK_PACKET) {
1183                static int warned;
1184                if (!warned) {
1185                        warned = 1;
1186                        printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1187                               current->comm);
1188                }
1189                family = PF_PACKET;
1190        }
1191
1192        err = security_socket_create(family, type, protocol, kern);
1193        if (err)
1194                return err;
1195
1196        /*
1197         *      Allocate the socket and allow the family to set things up. if
1198         *      the protocol is 0, the family is instructed to select an appropriate
1199         *      default.
1200         */
1201        sock = sock_alloc();
1202        if (!sock) {
1203                if (net_ratelimit())
1204                        printk(KERN_WARNING "socket: no more sockets\n");
1205                return -ENFILE; /* Not exactly a match, but its the
1206                                   closest posix thing */
1207        }
1208
1209        sock->type = type;
1210
1211#ifdef CONFIG_MODULES
1212        /* Attempt to load a protocol module if the find failed.
1213         *
1214         * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1215         * requested real, full-featured networking support upon configuration.
1216         * Otherwise module support will break!
1217         */
1218        if (net_families[family] == NULL)
1219                request_module("net-pf-%d", family);
1220#endif
1221
1222        rcu_read_lock();
1223        pf = rcu_dereference(net_families[family]);
1224        err = -EAFNOSUPPORT;
1225        if (!pf)
1226                goto out_release;
1227
1228        /*
1229         * We will call the ->create function, that possibly is in a loadable
1230         * module, so we have to bump that loadable module refcnt first.
1231         */
1232        if (!try_module_get(pf->owner))
1233                goto out_release;
1234
1235        /* Now protected by module ref count */
1236        rcu_read_unlock();
1237
1238        err = pf->create(net, sock, protocol, kern);
1239        if (err < 0)
1240                goto out_module_put;
1241
1242        /*
1243         * Now to bump the refcnt of the [loadable] module that owns this
1244         * socket at sock_release time we decrement its refcnt.
1245         */
1246        if (!try_module_get(sock->ops->owner))
1247                goto out_module_busy;
1248
1249        /*
1250         * Now that we're done with the ->create function, the [loadable]
1251         * module can have its refcnt decremented
1252         */
1253        module_put(pf->owner);
1254        err = security_socket_post_create(sock, family, type, protocol, kern);
1255        if (err)
1256                goto out_sock_release;
1257        *res = sock;
1258
1259        return 0;
1260
1261out_module_busy:
1262        err = -EAFNOSUPPORT;
1263out_module_put:
1264        sock->ops = NULL;
1265        module_put(pf->owner);
1266out_sock_release:
1267        sock_release(sock);
1268        return err;
1269
1270out_release:
1271        rcu_read_unlock();
1272        goto out_sock_release;
1273}
1274EXPORT_SYMBOL(__sock_create);
1275
1276int sock_create(int family, int type, int protocol, struct socket **res)
1277{
1278        return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1279}
1280EXPORT_SYMBOL(sock_create);
1281
1282int sock_create_kern(int family, int type, int protocol, struct socket **res)
1283{
1284        return __sock_create(&init_net, family, type, protocol, res, 1);
1285}
1286EXPORT_SYMBOL(sock_create_kern);
1287
1288SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1289{
1290        int retval;
1291        struct socket *sock;
1292        int flags;
1293
1294        /* Check the SOCK_* constants for consistency.  */
1295        BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1296        BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1297        BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1298        BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1299
1300        flags = type & ~SOCK_TYPE_MASK;
1301        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1302                return -EINVAL;
1303        type &= SOCK_TYPE_MASK;
1304
1305        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1306                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1307
1308        retval = sock_create(family, type, protocol, &sock);
1309        if (retval < 0)
1310                goto out;
1311
1312        retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1313        if (retval < 0)
1314                goto out_release;
1315
1316out:
1317        /* It may be already another descriptor 8) Not kernel problem. */
1318        return retval;
1319
1320out_release:
1321        sock_release(sock);
1322        return retval;
1323}
1324
1325/*
1326 *      Create a pair of connected sockets.
1327 */
1328
1329SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1330                int __user *, usockvec)
1331{
1332        struct socket *sock1, *sock2;
1333        int fd1, fd2, err;
1334        struct file *newfile1, *newfile2;
1335        int flags;
1336
1337        flags = type & ~SOCK_TYPE_MASK;
1338        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1339                return -EINVAL;
1340        type &= SOCK_TYPE_MASK;
1341
1342        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1343                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1344
1345        /*
1346         * Obtain the first socket and check if the underlying protocol
1347         * supports the socketpair call.
1348         */
1349
1350        err = sock_create(family, type, protocol, &sock1);
1351        if (err < 0)
1352                goto out;
1353
1354        err = sock_create(family, type, protocol, &sock2);
1355        if (err < 0)
1356                goto out_release_1;
1357
1358        err = sock1->ops->socketpair(sock1, sock2);
1359        if (err < 0)
1360                goto out_release_both;
1361
1362        fd1 = sock_alloc_file(sock1, &newfile1, flags);
1363        if (unlikely(fd1 < 0)) {
1364                err = fd1;
1365                goto out_release_both;
1366        }
1367
1368        fd2 = sock_alloc_file(sock2, &newfile2, flags);
1369        if (unlikely(fd2 < 0)) {
1370                err = fd2;
1371                fput(newfile1);
1372                put_unused_fd(fd1);
1373                sock_release(sock2);
1374                goto out;
1375        }
1376
1377        audit_fd_pair(fd1, fd2);
1378        fd_install(fd1, newfile1);
1379        fd_install(fd2, newfile2);
1380        /* fd1 and fd2 may be already another descriptors.
1381         * Not kernel problem.
1382         */
1383
1384        err = put_user(fd1, &usockvec[0]);
1385        if (!err)
1386                err = put_user(fd2, &usockvec[1]);
1387        if (!err)
1388                return 0;
1389
1390        sys_close(fd2);
1391        sys_close(fd1);
1392        return err;
1393
1394out_release_both:
1395        sock_release(sock2);
1396out_release_1:
1397        sock_release(sock1);
1398out:
1399        return err;
1400}
1401
1402/*
1403 *      Bind a name to a socket. Nothing much to do here since it's
1404 *      the protocol's responsibility to handle the local address.
1405 *
1406 *      We move the socket address to kernel space before we call
1407 *      the protocol layer (having also checked the address is ok).
1408 */
1409
1410SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1411{
1412        struct socket *sock;
1413        struct sockaddr_storage address;
1414        int err, fput_needed;
1415
1416        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1417        if (sock) {
1418                err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1419                if (err >= 0) {
1420                        err = security_socket_bind(sock,
1421                                                   (struct sockaddr *)&address,
1422                                                   addrlen);
1423                        if (!err)
1424                                err = sock->ops->bind(sock,
1425                                                      (struct sockaddr *)
1426                                                      &address, addrlen);
1427                }
1428                fput_light(sock->file, fput_needed);
1429        }
1430        return err;
1431}
1432
1433/*
1434 *      Perform a listen. Basically, we allow the protocol to do anything
1435 *      necessary for a listen, and if that works, we mark the socket as
1436 *      ready for listening.
1437 */
1438
1439SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1440{
1441        struct socket *sock;
1442        int err, fput_needed;
1443        int somaxconn;
1444
1445        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1446        if (sock) {
1447                somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1448                if ((unsigned)backlog > somaxconn)
1449                        backlog = somaxconn;
1450
1451                err = security_socket_listen(sock, backlog);
1452                if (!err)
1453                        err = sock->ops->listen(sock, backlog);
1454
1455                fput_light(sock->file, fput_needed);
1456        }
1457        return err;
1458}
1459
1460/*
1461 *      For accept, we attempt to create a new socket, set up the link
1462 *      with the client, wake up the client, then return the new
1463 *      connected fd. We collect the address of the connector in kernel
1464 *      space and move it to user at the very end. This is unclean because
1465 *      we open the socket then return an error.
1466 *
1467 *      1003.1g adds the ability to recvmsg() to query connection pending
1468 *      status to recvmsg. We need to add that support in a way thats
1469 *      clean when we restucture accept also.
1470 */
1471
1472SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1473                int __user *, upeer_addrlen, int, flags)
1474{
1475        struct socket *sock, *newsock;
1476        struct file *newfile;
1477        int err, len, newfd, fput_needed;
1478        struct sockaddr_storage address;
1479
1480        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1481                return -EINVAL;
1482
1483        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1484                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1485
1486        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1487        if (!sock)
1488                goto out;
1489
1490        err = -ENFILE;
1491        newsock = sock_alloc();
1492        if (!newsock)
1493                goto out_put;
1494
1495        newsock->type = sock->type;
1496        newsock->ops = sock->ops;
1497
1498        /*
1499         * We don't need try_module_get here, as the listening socket (sock)
1500         * has the protocol module (sock->ops->owner) held.
1501         */
1502        __module_get(newsock->ops->owner);
1503
1504        newfd = sock_alloc_file(newsock, &newfile, flags);
1505        if (unlikely(newfd < 0)) {
1506                err = newfd;
1507                sock_release(newsock);
1508                goto out_put;
1509        }
1510
1511        err = security_socket_accept(sock, newsock);
1512        if (err)
1513                goto out_fd;
1514
1515        err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1516        if (err < 0)
1517                goto out_fd;
1518
1519        if (upeer_sockaddr) {
1520                if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1521                                          &len, 2) < 0) {
1522                        err = -ECONNABORTED;
1523                        goto out_fd;
1524                }
1525                err = move_addr_to_user((struct sockaddr *)&address,
1526                                        len, upeer_sockaddr, upeer_addrlen);
1527                if (err < 0)
1528                        goto out_fd;
1529        }
1530
1531        /* File flags are not inherited via accept() unlike another OSes. */
1532
1533        fd_install(newfd, newfile);
1534        err = newfd;
1535
1536out_put:
1537        fput_light(sock->file, fput_needed);
1538out:
1539        return err;
1540out_fd:
1541        fput(newfile);
1542        put_unused_fd(newfd);
1543        goto out_put;
1544}
1545
1546SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1547                int __user *, upeer_addrlen)
1548{
1549        return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1550}
1551
1552/*
1553 *      Attempt to connect to a socket with the server address.  The address
1554 *      is in user space so we verify it is OK and move it to kernel space.
1555 *
1556 *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1557 *      break bindings
1558 *
1559 *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1560 *      other SEQPACKET protocols that take time to connect() as it doesn't
1561 *      include the -EINPROGRESS status for such sockets.
1562 */
1563
1564SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1565                int, addrlen)
1566{
1567        struct socket *sock;
1568        struct sockaddr_storage address;
1569        int err, fput_needed;
1570
1571        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1572        if (!sock)
1573                goto out;
1574        err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1575        if (err < 0)
1576                goto out_put;
1577
1578        err =
1579            security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1580        if (err)
1581                goto out_put;
1582
1583        err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1584                                 sock->file->f_flags);
1585out_put:
1586        fput_light(sock->file, fput_needed);
1587out:
1588        return err;
1589}
1590
1591/*
1592 *      Get the local address ('name') of a socket object. Move the obtained
1593 *      name to user space.
1594 */
1595
1596SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1597                int __user *, usockaddr_len)
1598{
1599        struct socket *sock;
1600        struct sockaddr_storage address;
1601        int len, err, fput_needed;
1602
1603        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1604        if (!sock)
1605                goto out;
1606
1607        err = security_socket_getsockname(sock);
1608        if (err)
1609                goto out_put;
1610
1611        err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1612        if (err)
1613                goto out_put;
1614        err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1615
1616out_put:
1617        fput_light(sock->file, fput_needed);
1618out:
1619        return err;
1620}
1621
1622/*
1623 *      Get the remote address ('name') of a socket object. Move the obtained
1624 *      name to user space.
1625 */
1626
1627SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1628                int __user *, usockaddr_len)
1629{
1630        struct socket *sock;
1631        struct sockaddr_storage address;
1632        int len, err, fput_needed;
1633
1634        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1635        if (sock != NULL) {
1636                err = security_socket_getpeername(sock);
1637                if (err) {
1638                        fput_light(sock->file, fput_needed);
1639                        return err;
1640                }
1641
1642                err =
1643                    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1644                                       1);
1645                if (!err)
1646                        err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1647                                                usockaddr_len);
1648                fput_light(sock->file, fput_needed);
1649        }
1650        return err;
1651}
1652
1653/*
1654 *      Send a datagram to a given address. We move the address into kernel
1655 *      space and check the user space data area is readable before invoking
1656 *      the protocol.
1657 */
1658
1659SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1660                unsigned, flags, struct sockaddr __user *, addr,
1661                int, addr_len)
1662{
1663        struct socket *sock;
1664        struct sockaddr_storage address;
1665        int err;
1666        struct msghdr msg;
1667        struct iovec iov;
1668        int fput_needed;
1669
1670        if (len > INT_MAX)
1671                len = INT_MAX;
1672        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1673        if (!sock)
1674                goto out;
1675
1676        iov.iov_base = buff;
1677        iov.iov_len = len;
1678        msg.msg_name = NULL;
1679        msg.msg_iov = &iov;
1680        msg.msg_iovlen = 1;
1681        msg.msg_control = NULL;
1682        msg.msg_controllen = 0;
1683        msg.msg_namelen = 0;
1684        if (addr) {
1685                err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1686                if (err < 0)
1687                        goto out_put;
1688                msg.msg_name = (struct sockaddr *)&address;
1689                msg.msg_namelen = addr_len;
1690        }
1691        if (sock->file->f_flags & O_NONBLOCK)
1692                flags |= MSG_DONTWAIT;
1693        msg.msg_flags = flags;
1694        err = sock_sendmsg(sock, &msg, len);
1695
1696out_put:
1697        fput_light(sock->file, fput_needed);
1698out:
1699        return err;
1700}
1701
1702/*
1703 *      Send a datagram down a socket.
1704 */
1705
1706SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1707                unsigned, flags)
1708{
1709        return sys_sendto(fd, buff, len, flags, NULL, 0);
1710}
1711
1712/*
1713 *      Receive a frame from the socket and optionally record the address of the
1714 *      sender. We verify the buffers are writable and if needed move the
1715 *      sender address from kernel to user space.
1716 */
1717
1718SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1719                unsigned, flags, struct sockaddr __user *, addr,
1720                int __user *, addr_len)
1721{
1722        struct socket *sock;
1723        struct iovec iov;
1724        struct msghdr msg;
1725        struct sockaddr_storage address;
1726        int err, err2;
1727        int fput_needed;
1728
1729        if (size > INT_MAX)
1730                size = INT_MAX;
1731        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1732        if (!sock)
1733                goto out;
1734
1735        msg.msg_control = NULL;
1736        msg.msg_controllen = 0;
1737        msg.msg_iovlen = 1;
1738        msg.msg_iov = &iov;
1739        iov.iov_len = size;
1740        iov.iov_base = ubuf;
1741        msg.msg_name = (struct sockaddr *)&address;
1742        msg.msg_namelen = sizeof(address);
1743        if (sock->file->f_flags & O_NONBLOCK)
1744                flags |= MSG_DONTWAIT;
1745        err = sock_recvmsg(sock, &msg, size, flags);
1746
1747        if (err >= 0 && addr != NULL) {
1748                err2 = move_addr_to_user((struct sockaddr *)&address,
1749                                         msg.msg_namelen, addr, addr_len);
1750                if (err2 < 0)
1751                        err = err2;
1752        }
1753
1754        fput_light(sock->file, fput_needed);
1755out:
1756        return err;
1757}
1758
1759/*
1760 *      Receive a datagram from a socket.
1761 */
1762
1763asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1764                         unsigned flags)
1765{
1766        return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1767}
1768
1769/*
1770 *      Set a socket option. Because we don't know the option lengths we have
1771 *      to pass the user mode parameter for the protocols to sort out.
1772 */
1773
1774SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1775                char __user *, optval, int, optlen)
1776{
1777        int err, fput_needed;
1778        struct socket *sock;
1779
1780        if (optlen < 0)
1781                return -EINVAL;
1782
1783        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1784        if (sock != NULL) {
1785                err = security_socket_setsockopt(sock, level, optname);
1786                if (err)
1787                        goto out_put;
1788
1789                if (level == SOL_SOCKET)
1790                        err =
1791                            sock_setsockopt(sock, level, optname, optval,
1792                                            optlen);
1793                else
1794                        err =
1795                            sock->ops->setsockopt(sock, level, optname, optval,
1796                                                  optlen);
1797out_put:
1798                fput_light(sock->file, fput_needed);
1799        }
1800        return err;
1801}
1802
1803/*
1804 *      Get a socket option. Because we don't know the option lengths we have
1805 *      to pass a user mode parameter for the protocols to sort out.
1806 */
1807
1808SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1809                char __user *, optval, int __user *, optlen)
1810{
1811        int err, fput_needed;
1812        struct socket *sock;
1813
1814        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1815        if (sock != NULL) {
1816                err = security_socket_getsockopt(sock, level, optname);
1817                if (err)
1818                        goto out_put;
1819
1820                if (level == SOL_SOCKET)
1821                        err =
1822                            sock_getsockopt(sock, level, optname, optval,
1823                                            optlen);
1824                else
1825                        err =
1826                            sock->ops->getsockopt(sock, level, optname, optval,
1827                                                  optlen);
1828out_put:
1829                fput_light(sock->file, fput_needed);
1830        }
1831        return err;
1832}
1833
1834/*
1835 *      Shutdown a socket.
1836 */
1837
1838SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1839{
1840        int err, fput_needed;
1841        struct socket *sock;
1842
1843        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1844        if (sock != NULL) {
1845                err = security_socket_shutdown(sock, how);
1846                if (!err)
1847                        err = sock->ops->shutdown(sock, how);
1848                fput_light(sock->file, fput_needed);
1849        }
1850        return err;
1851}
1852
1853/* A couple of helpful macros for getting the address of the 32/64 bit
1854 * fields which are the same type (int / unsigned) on our platforms.
1855 */
1856#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1857#define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1858#define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1859
1860/*
1861 *      BSD sendmsg interface
1862 */
1863
1864SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1865{
1866        struct compat_msghdr __user *msg_compat =
1867            (struct compat_msghdr __user *)msg;
1868        struct socket *sock;
1869        struct sockaddr_storage address;
1870        struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1871        unsigned char ctl[sizeof(struct cmsghdr) + 20]
1872            __attribute__ ((aligned(sizeof(__kernel_size_t))));
1873        /* 20 is size of ipv6_pktinfo */
1874        unsigned char *ctl_buf = ctl;
1875        struct msghdr msg_sys;
1876        int err, ctl_len, iov_size, total_len;
1877        int fput_needed;
1878
1879        err = -EFAULT;
1880        if (MSG_CMSG_COMPAT & flags) {
1881                if (get_compat_msghdr(&msg_sys, msg_compat))
1882                        return -EFAULT;
1883        } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1884                return -EFAULT;
1885
1886        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1887        if (!sock)
1888                goto out;
1889
1890        /* do not move before msg_sys is valid */
1891        err = -EMSGSIZE;
1892        if (msg_sys.msg_iovlen > UIO_MAXIOV)
1893                goto out_put;
1894
1895        /* Check whether to allocate the iovec area */
1896        err = -ENOMEM;
1897        iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1898        if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1899                iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1900                if (!iov)
1901                        goto out_put;
1902        }
1903
1904        /* This will also move the address data into kernel space */
1905        if (MSG_CMSG_COMPAT & flags) {
1906                err = verify_compat_iovec(&msg_sys, iov,
1907                                          (struct sockaddr *)&address,
1908                                          VERIFY_READ);
1909        } else
1910                err = verify_iovec(&msg_sys, iov,
1911                                   (struct sockaddr *)&address,
1912                                   VERIFY_READ);
1913        if (err < 0)
1914                goto out_freeiov;
1915        total_len = err;
1916
1917        err = -ENOBUFS;
1918
1919        if (msg_sys.msg_controllen > INT_MAX)
1920                goto out_freeiov;
1921        ctl_len = msg_sys.msg_controllen;
1922        if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1923                err =
1924                    cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1925                                                     sizeof(ctl));
1926                if (err)
1927                        goto out_freeiov;
1928                ctl_buf = msg_sys.msg_control;
1929                ctl_len = msg_sys.msg_controllen;
1930        } else if (ctl_len) {
1931                if (ctl_len > sizeof(ctl)) {
1932                        ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1933                        if (ctl_buf == NULL)
1934                                goto out_freeiov;
1935                }
1936                err = -EFAULT;
1937                /*
1938                 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1939                 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1940                 * checking falls down on this.
1941                 */
1942                if (copy_from_user(ctl_buf,
1943                                   (void __user __force *)msg_sys.msg_control,
1944                                   ctl_len))
1945                        goto out_freectl;
1946                msg_sys.msg_control = ctl_buf;
1947        }
1948        msg_sys.msg_flags = flags;
1949
1950        if (sock->file->f_flags & O_NONBLOCK)
1951                msg_sys.msg_flags |= MSG_DONTWAIT;
1952        err = sock_sendmsg(sock, &msg_sys, total_len);
1953
1954out_freectl:
1955        if (ctl_buf != ctl)
1956                sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1957out_freeiov:
1958        if (iov != iovstack)
1959                sock_kfree_s(sock->sk, iov, iov_size);
1960out_put:
1961        fput_light(sock->file, fput_needed);
1962out:
1963        return err;
1964}
1965
1966static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1967                         struct msghdr *msg_sys, unsigned flags, int nosec)
1968{
1969        struct compat_msghdr __user *msg_compat =
1970            (struct compat_msghdr __user *)msg;
1971        struct iovec iovstack[UIO_FASTIOV];
1972        struct iovec *iov = iovstack;
1973        unsigned long cmsg_ptr;
1974        int err, iov_size, total_len, len;
1975
1976        /* kernel mode address */
1977        struct sockaddr_storage addr;
1978
1979        /* user mode address pointers */
1980        struct sockaddr __user *uaddr;
1981        int __user *uaddr_len;
1982
1983        if (MSG_CMSG_COMPAT & flags) {
1984                if (get_compat_msghdr(msg_sys, msg_compat))
1985                        return -EFAULT;
1986        } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1987                return -EFAULT;
1988
1989        err = -EMSGSIZE;
1990        if (msg_sys->msg_iovlen > UIO_MAXIOV)
1991                goto out;
1992
1993        /* Check whether to allocate the iovec area */
1994        err = -ENOMEM;
1995        iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1996        if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1997                iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1998                if (!iov)
1999                        goto out;
2000        }
2001
2002        /*
2003         *      Save the user-mode address (verify_iovec will change the
2004         *      kernel msghdr to use the kernel address space)
2005         */
2006
2007        uaddr = (__force void __user *)msg_sys->msg_name;
2008        uaddr_len = COMPAT_NAMELEN(msg);
2009        if (MSG_CMSG_COMPAT & flags) {
2010                err = verify_compat_iovec(msg_sys, iov,
2011                                          (struct sockaddr *)&addr,
2012                                          VERIFY_WRITE);
2013        } else
2014                err = verify_iovec(msg_sys, iov,
2015                                   (struct sockaddr *)&addr,
2016                                   VERIFY_WRITE);
2017        if (err < 0)
2018                goto out_freeiov;
2019        total_len = err;
2020
2021        cmsg_ptr = (unsigned long)msg_sys->msg_control;
2022        msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2023
2024        if (sock->file->f_flags & O_NONBLOCK)
2025                flags |= MSG_DONTWAIT;
2026        err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2027                                                          total_len, flags);
2028        if (err < 0)
2029                goto out_freeiov;
2030        len = err;
2031
2032        if (uaddr != NULL) {
2033                err = move_addr_to_user((struct sockaddr *)&addr,
2034                                        msg_sys->msg_namelen, uaddr,
2035                                        uaddr_len);
2036                if (err < 0)
2037                        goto out_freeiov;
2038        }
2039        err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2040                         COMPAT_FLAGS(msg));
2041        if (err)
2042                goto out_freeiov;
2043        if (MSG_CMSG_COMPAT & flags)
2044                err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2045                                 &msg_compat->msg_controllen);
2046        else
2047                err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2048                                 &msg->msg_controllen);
2049        if (err)
2050                goto out_freeiov;
2051        err = len;
2052
2053out_freeiov:
2054        if (iov != iovstack)
2055                sock_kfree_s(sock->sk, iov, iov_size);
2056out:
2057        return err;
2058}
2059
2060/*
2061 *      BSD recvmsg interface
2062 */
2063
2064SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2065                unsigned int, flags)
2066{
2067        int fput_needed, err;
2068        struct msghdr msg_sys;
2069        struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2070
2071        if (!sock)
2072                goto out;
2073
2074        err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2075
2076        fput_light(sock->file, fput_needed);
2077out:
2078        return err;
2079}
2080
2081/*
2082 *     Linux recvmmsg interface
2083 */
2084
2085int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2086                   unsigned int flags, struct timespec *timeout)
2087{
2088        int fput_needed, err, datagrams;
2089        struct socket *sock;
2090        struct mmsghdr __user *entry;
2091        struct compat_mmsghdr __user *compat_entry;
2092        struct msghdr msg_sys;
2093        struct timespec end_time;
2094
2095        if (timeout &&
2096            poll_select_set_timeout(&end_time, timeout->tv_sec,
2097                                    timeout->tv_nsec))
2098                return -EINVAL;
2099
2100        datagrams = 0;
2101
2102        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2103        if (!sock)
2104                return err;
2105
2106        err = sock_error(sock->sk);
2107        if (err)
2108                goto out_put;
2109
2110        entry = mmsg;
2111        compat_entry = (struct compat_mmsghdr __user *)mmsg;
2112
2113        while (datagrams < vlen) {
2114                /*
2115                 * No need to ask LSM for more than the first datagram.
2116                 */
2117                if (MSG_CMSG_COMPAT & flags) {
2118                        err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2119                                            &msg_sys, flags, datagrams);
2120                        if (err < 0)
2121                                break;
2122                        err = __put_user(err, &compat_entry->msg_len);
2123                        ++compat_entry;
2124                } else {
2125                        err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2126                                            &msg_sys, flags, datagrams);
2127                        if (err < 0)
2128                                break;
2129                        err = put_user(err, &entry->msg_len);
2130                        ++entry;
2131                }
2132
2133                if (err)
2134                        break;
2135                ++datagrams;
2136
2137                /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2138                if (flags & MSG_WAITFORONE)
2139                        flags |= MSG_DONTWAIT;
2140
2141                if (timeout) {
2142                        ktime_get_ts(timeout);
2143                        *timeout = timespec_sub(end_time, *timeout);
2144                        if (timeout->tv_sec < 0) {
2145                                timeout->tv_sec = timeout->tv_nsec = 0;
2146                                break;
2147                        }
2148
2149                        /* Timeout, return less than vlen datagrams */
2150                        if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2151                                break;
2152                }
2153
2154                /* Out of band data, return right away */
2155                if (msg_sys.msg_flags & MSG_OOB)
2156                        break;
2157        }
2158
2159out_put:
2160        fput_light(sock->file, fput_needed);
2161
2162        if (err == 0)
2163                return datagrams;
2164
2165        if (datagrams != 0) {
2166                /*
2167                 * We may return less entries than requested (vlen) if the
2168                 * sock is non block and there aren't enough datagrams...
2169                 */
2170                if (err != -EAGAIN) {
2171                        /*
2172                         * ... or  if recvmsg returns an error after we
2173                         * received some datagrams, where we record the
2174                         * error to return on the next call or if the
2175                         * app asks about it using getsockopt(SO_ERROR).
2176                         */
2177                        sock->sk->sk_err = -err;
2178                }
2179
2180                return datagrams;
2181        }
2182
2183        return err;
2184}
2185
2186SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2187                unsigned int, vlen, unsigned int, flags,
2188                struct timespec __user *, timeout)
2189{
2190        int datagrams;
2191        struct timespec timeout_sys;
2192
2193        if (!timeout)
2194                return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2195
2196        if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2197                return -EFAULT;
2198
2199        datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2200
2201        if (datagrams > 0 &&
2202            copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2203                datagrams = -EFAULT;
2204
2205        return datagrams;
2206}
2207
2208#ifdef __ARCH_WANT_SYS_SOCKETCALL
2209/* Argument list sizes for sys_socketcall */
2210#define AL(x) ((x) * sizeof(unsigned long))
2211static const unsigned char nargs[20] = {
2212        AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2213        AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2214        AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2215        AL(4), AL(5)
2216};
2217
2218#undef AL
2219
2220/*
2221 *      System call vectors.
2222 *
2223 *      Argument checking cleaned up. Saved 20% in size.
2224 *  This function doesn't need to set the kernel lock because
2225 *  it is set by the callees.
2226 */
2227
2228SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2229{
2230        unsigned long a[6];
2231        unsigned long a0, a1;
2232        int err;
2233        unsigned int len;
2234
2235        if (call < 1 || call > SYS_RECVMMSG)
2236                return -EINVAL;
2237
2238        len = nargs[call];
2239        if (len > sizeof(a))
2240                return -EINVAL;
2241
2242        /* copy_from_user should be SMP safe. */
2243        if (copy_from_user(a, args, len))
2244                return -EFAULT;
2245
2246        audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2247
2248        a0 = a[0];
2249        a1 = a[1];
2250
2251        switch (call) {
2252        case SYS_SOCKET:
2253                err = sys_socket(a0, a1, a[2]);
2254                break;
2255        case SYS_BIND:
2256                err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2257                break;
2258        case SYS_CONNECT:
2259                err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2260                break;
2261        case SYS_LISTEN:
2262                err = sys_listen(a0, a1);
2263                break;
2264        case SYS_ACCEPT:
2265                err = sys_accept4(a0, (struct sockaddr __user *)a1,
2266                                  (int __user *)a[2], 0);
2267                break;
2268        case SYS_GETSOCKNAME:
2269                err =
2270                    sys_getsockname(a0, (struct sockaddr __user *)a1,
2271                                    (int __user *)a[2]);
2272                break;
2273        case SYS_GETPEERNAME:
2274                err =
2275                    sys_getpeername(a0, (struct sockaddr __user *)a1,
2276                                    (int __user *)a[2]);
2277                break;
2278        case SYS_SOCKETPAIR:
2279                err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2280                break;
2281        case SYS_SEND:
2282                err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2283                break;
2284        case SYS_SENDTO:
2285                err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2286                                 (struct sockaddr __user *)a[4], a[5]);
2287                break;
2288        case SYS_RECV:
2289                err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2290                break;
2291        case SYS_RECVFROM:
2292                err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2293                                   (struct sockaddr __user *)a[4],
2294                                   (int __user *)a[5]);
2295                break;
2296        case SYS_SHUTDOWN:
2297                err = sys_shutdown(a0, a1);
2298                break;
2299        case SYS_SETSOCKOPT:
2300                err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2301                break;
2302        case SYS_GETSOCKOPT:
2303                err =
2304                    sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2305                                   (int __user *)a[4]);
2306                break;
2307        case SYS_SENDMSG:
2308                err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2309                break;
2310        case SYS_RECVMSG:
2311                err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2312                break;
2313        case SYS_RECVMMSG:
2314                err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2315                                   (struct timespec __user *)a[4]);
2316                break;
2317        case SYS_ACCEPT4:
2318                err = sys_accept4(a0, (struct sockaddr __user *)a1,
2319                                  (int __user *)a[2], a[3]);
2320                break;
2321        default:
2322                err = -EINVAL;
2323                break;
2324        }
2325        return err;
2326}
2327
2328#endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2329
2330/**
2331 *      sock_register - add a socket protocol handler
2332 *      @ops: description of protocol
2333 *
2334 *      This function is called by a protocol handler that wants to
2335 *      advertise its address family, and have it linked into the
2336 *      socket interface. The value ops->family coresponds to the
2337 *      socket system call protocol family.
2338 */
2339int sock_register(const struct net_proto_family *ops)
2340{
2341        int err;
2342
2343        if (ops->family >= NPROTO) {
2344                printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2345                       NPROTO);
2346                return -ENOBUFS;
2347        }
2348
2349        spin_lock(&net_family_lock);
2350        if (net_families[ops->family])
2351                err = -EEXIST;
2352        else {
2353                net_families[ops->family] = ops;
2354                err = 0;
2355        }
2356        spin_unlock(&net_family_lock);
2357
2358        printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2359        return err;
2360}
2361EXPORT_SYMBOL(sock_register);
2362
2363/**
2364 *      sock_unregister - remove a protocol handler
2365 *      @family: protocol family to remove
2366 *
2367 *      This function is called by a protocol handler that wants to
2368 *      remove its address family, and have it unlinked from the
2369 *      new socket creation.
2370 *
2371 *      If protocol handler is a module, then it can use module reference
2372 *      counts to protect against new references. If protocol handler is not
2373 *      a module then it needs to provide its own protection in
2374 *      the ops->create routine.
2375 */
2376void sock_unregister(int family)
2377{
2378        BUG_ON(family < 0 || family >= NPROTO);
2379
2380        spin_lock(&net_family_lock);
2381        net_families[family] = NULL;
2382        spin_unlock(&net_family_lock);
2383
2384        synchronize_rcu();
2385
2386        printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2387}
2388EXPORT_SYMBOL(sock_unregister);
2389
2390static int __init sock_init(void)
2391{
2392        /*
2393         *      Initialize sock SLAB cache.
2394         */
2395
2396        sk_init();
2397
2398        /*
2399         *      Initialize skbuff SLAB cache
2400         */
2401        skb_init();
2402
2403        /*
2404         *      Initialize the protocols module.
2405         */
2406
2407        init_inodecache();
2408        register_filesystem(&sock_fs_type);
2409        sock_mnt = kern_mount(&sock_fs_type);
2410
2411        /* The real protocol initialization is performed in later initcalls.
2412         */
2413
2414#ifdef CONFIG_NETFILTER
2415        netfilter_init();
2416#endif
2417
2418#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2419        skb_timestamping_init();
2420#endif
2421
2422        return 0;
2423}
2424
2425core_initcall(sock_init);       /* early initcall */
2426
2427#ifdef CONFIG_PROC_FS
2428void socket_seq_show(struct seq_file *seq)
2429{
2430        int cpu;
2431        int counter = 0;
2432
2433        for_each_possible_cpu(cpu)
2434            counter += per_cpu(sockets_in_use, cpu);
2435
2436        /* It can be negative, by the way. 8) */
2437        if (counter < 0)
2438                counter = 0;
2439
2440        seq_printf(seq, "sockets: used %d\n", counter);
2441}
2442#endif                          /* CONFIG_PROC_FS */
2443
2444#ifdef CONFIG_COMPAT
2445static int do_siocgstamp(struct net *net, struct socket *sock,
2446                         unsigned int cmd, struct compat_timeval __user *up)
2447{
2448        mm_segment_t old_fs = get_fs();
2449        struct timeval ktv;
2450        int err;
2451
2452        set_fs(KERNEL_DS);
2453        err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2454        set_fs(old_fs);
2455        if (!err) {
2456                err = put_user(ktv.tv_sec, &up->tv_sec);
2457                err |= __put_user(ktv.tv_usec, &up->tv_usec);
2458        }
2459        return err;
2460}
2461
2462static int do_siocgstampns(struct net *net, struct socket *sock,
2463                         unsigned int cmd, struct compat_timespec __user *up)
2464{
2465        mm_segment_t old_fs = get_fs();
2466        struct timespec kts;
2467        int err;
2468
2469        set_fs(KERNEL_DS);
2470        err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2471        set_fs(old_fs);
2472        if (!err) {
2473                err = put_user(kts.tv_sec, &up->tv_sec);
2474                err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2475        }
2476        return err;
2477}
2478
2479static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2480{
2481        struct ifreq __user *uifr;
2482        int err;
2483
2484        uifr = compat_alloc_user_space(sizeof(struct ifreq));
2485        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2486                return -EFAULT;
2487
2488        err = dev_ioctl(net, SIOCGIFNAME, uifr);
2489        if (err)
2490                return err;
2491
2492        if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2493                return -EFAULT;
2494
2495        return 0;
2496}
2497
2498static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2499{
2500        struct compat_ifconf ifc32;
2501        struct ifconf ifc;
2502        struct ifconf __user *uifc;
2503        struct compat_ifreq __user *ifr32;
2504        struct ifreq __user *ifr;
2505        unsigned int i, j;
2506        int err;
2507
2508        if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2509                return -EFAULT;
2510
2511        if (ifc32.ifcbuf == 0) {
2512                ifc32.ifc_len = 0;
2513                ifc.ifc_len = 0;
2514                ifc.ifc_req = NULL;
2515                uifc = compat_alloc_user_space(sizeof(struct ifconf));
2516        } else {
2517                size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2518                        sizeof(struct ifreq);
2519                uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2520                ifc.ifc_len = len;
2521                ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2522                ifr32 = compat_ptr(ifc32.ifcbuf);
2523                for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2524                        if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2525                                return -EFAULT;
2526                        ifr++;
2527                        ifr32++;
2528                }
2529        }
2530        if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2531                return -EFAULT;
2532
2533        err = dev_ioctl(net, SIOCGIFCONF, uifc);
2534        if (err)
2535                return err;
2536
2537        if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2538                return -EFAULT;
2539
2540        ifr = ifc.ifc_req;
2541        ifr32 = compat_ptr(ifc32.ifcbuf);
2542        for (i = 0, j = 0;
2543             i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2544             i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2545                if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2546                        return -EFAULT;
2547                ifr32++;
2548                ifr++;
2549        }
2550
2551        if (ifc32.ifcbuf == 0) {
2552                /* Translate from 64-bit structure multiple to
2553                 * a 32-bit one.
2554                 */
2555                i = ifc.ifc_len;
2556                i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2557                ifc32.ifc_len = i;
2558        } else {
2559                ifc32.ifc_len = i;
2560        }
2561        if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2562                return -EFAULT;
2563
2564        return 0;
2565}
2566
2567static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2568{
2569        struct ifreq __user *ifr;
2570        u32 data;
2571        void __user *datap;
2572
2573        ifr = compat_alloc_user_space(sizeof(*ifr));
2574
2575        if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2576                return -EFAULT;
2577
2578        if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2579                return -EFAULT;
2580
2581        datap = compat_ptr(data);
2582        if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2583                return -EFAULT;
2584
2585        return dev_ioctl(net, SIOCETHTOOL, ifr);
2586}
2587
2588static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2589{
2590        void __user *uptr;
2591        compat_uptr_t uptr32;
2592        struct ifreq __user *uifr;
2593
2594        uifr = compat_alloc_user_space(sizeof(*uifr));
2595        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2596                return -EFAULT;
2597
2598        if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2599                return -EFAULT;
2600
2601        uptr = compat_ptr(uptr32);
2602
2603        if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2604                return -EFAULT;
2605
2606        return dev_ioctl(net, SIOCWANDEV, uifr);
2607}
2608
2609static int bond_ioctl(struct net *net, unsigned int cmd,
2610                         struct compat_ifreq __user *ifr32)
2611{
2612        struct ifreq kifr;
2613        struct ifreq __user *uifr;
2614        mm_segment_t old_fs;
2615        int err;
2616        u32 data;
2617        void __user *datap;
2618
2619        switch (cmd) {
2620        case SIOCBONDENSLAVE:
2621        case SIOCBONDRELEASE:
2622        case SIOCBONDSETHWADDR:
2623        case SIOCBONDCHANGEACTIVE:
2624                if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2625                        return -EFAULT;
2626
2627                old_fs = get_fs();
2628                set_fs(KERNEL_DS);
2629                err = dev_ioctl(net, cmd, &kifr);
2630                set_fs(old_fs);
2631
2632                return err;
2633        case SIOCBONDSLAVEINFOQUERY:
2634        case SIOCBONDINFOQUERY:
2635                uifr = compat_alloc_user_space(sizeof(*uifr));
2636                if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2637                        return -EFAULT;
2638
2639                if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2640                        return -EFAULT;
2641
2642                datap = compat_ptr(data);
2643                if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2644                        return -EFAULT;
2645
2646                return dev_ioctl(net, cmd, uifr);
2647        default:
2648                return -EINVAL;
2649        }
2650}
2651
2652static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2653                                 struct compat_ifreq __user *u_ifreq32)
2654{
2655        struct ifreq __user *u_ifreq64;
2656        char tmp_buf[IFNAMSIZ];
2657        void __user *data64;
2658        u32 data32;
2659
2660        if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2661                           IFNAMSIZ))
2662                return -EFAULT;
2663        if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2664                return -EFAULT;
2665        data64 = compat_ptr(data32);
2666
2667        u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2668
2669        /* Don't check these user accesses, just let that get trapped
2670         * in the ioctl handler instead.
2671         */
2672        if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2673                         IFNAMSIZ))
2674                return -EFAULT;
2675        if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2676                return -EFAULT;
2677
2678        return dev_ioctl(net, cmd, u_ifreq64);
2679}
2680
2681static int dev_ifsioc(struct net *net, struct socket *sock,
2682                         unsigned int cmd, struct compat_ifreq __user *uifr32)
2683{
2684        struct ifreq __user *uifr;
2685        int err;
2686
2687        uifr = compat_alloc_user_space(sizeof(*uifr));
2688        if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2689                return -EFAULT;
2690
2691        err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2692
2693        if (!err) {
2694                switch (cmd) {
2695                case SIOCGIFFLAGS:
2696                case SIOCGIFMETRIC:
2697                case SIOCGIFMTU:
2698                case SIOCGIFMEM:
2699                case SIOCGIFHWADDR:
2700                case SIOCGIFINDEX:
2701                case SIOCGIFADDR:
2702                case SIOCGIFBRDADDR:
2703                case SIOCGIFDSTADDR:
2704                case SIOCGIFNETMASK:
2705                case SIOCGIFPFLAGS:
2706                case SIOCGIFTXQLEN:
2707                case SIOCGMIIPHY:
2708                case SIOCGMIIREG:
2709                        if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2710                                err = -EFAULT;
2711                        break;
2712                }
2713        }
2714        return err;
2715}
2716
2717static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2718                        struct compat_ifreq __user *uifr32)
2719{
2720        struct ifreq ifr;
2721        struct compat_ifmap __user *uifmap32;
2722        mm_segment_t old_fs;
2723        int err;
2724
2725        uifmap32 = &uifr32->ifr_ifru.ifru_map;
2726        err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2727        err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2728        err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2729        err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2730        err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2731        err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2732        err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2733        if (err)
2734                return -EFAULT;
2735
2736        old_fs = get_fs();
2737        set_fs(KERNEL_DS);
2738        err = dev_ioctl(net, cmd, (void __user *)&ifr);
2739        set_fs(old_fs);
2740
2741        if (cmd == SIOCGIFMAP && !err) {
2742                err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2743                err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2744                err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2745                err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2746                err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2747                err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2748                err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2749                if (err)
2750                        err = -EFAULT;
2751        }
2752        return err;
2753}
2754
2755static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2756{
2757        void __user *uptr;
2758        compat_uptr_t uptr32;
2759        struct ifreq __user *uifr;
2760
2761        uifr = compat_alloc_user_space(sizeof(*uifr));
2762        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2763                return -EFAULT;
2764
2765        if (get_user(uptr32, &uifr32->ifr_data))
2766                return -EFAULT;
2767
2768        uptr = compat_ptr(uptr32);
2769
2770        if (put_user(uptr, &uifr->ifr_data))
2771                return -EFAULT;
2772
2773        return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2774}
2775
2776struct rtentry32 {
2777        u32             rt_pad1;
2778        struct sockaddr rt_dst;         /* target address               */
2779        struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
2780        struct sockaddr rt_genmask;     /* target network mask (IP)     */
2781        unsigned short  rt_flags;
2782        short           rt_pad2;
2783        u32             rt_pad3;
2784        unsigned char   rt_tos;
2785        unsigned char   rt_class;
2786        short           rt_pad4;
2787        short           rt_metric;      /* +1 for binary compatibility! */
2788        /* char * */ u32 rt_dev;        /* forcing the device at add    */
2789        u32             rt_mtu;         /* per route MTU/Window         */
2790        u32             rt_window;      /* Window clamping              */
2791        unsigned short  rt_irtt;        /* Initial RTT                  */
2792};
2793
2794struct in6_rtmsg32 {
2795        struct in6_addr         rtmsg_dst;
2796        struct in6_addr         rtmsg_src;
2797        struct in6_addr         rtmsg_gateway;
2798        u32                     rtmsg_type;
2799        u16                     rtmsg_dst_len;
2800        u16                     rtmsg_src_len;
2801        u32                     rtmsg_metric;
2802        u32                     rtmsg_info;
2803        u32                     rtmsg_flags;
2804        s32                     rtmsg_ifindex;
2805};
2806
2807static int routing_ioctl(struct net *net, struct socket *sock,
2808                         unsigned int cmd, void __user *argp)
2809{
2810        int ret;
2811        void *r = NULL;
2812        struct in6_rtmsg r6;
2813        struct rtentry r4;
2814        char devname[16];
2815        u32 rtdev;
2816        mm_segment_t old_fs = get_fs();
2817
2818        if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2819                struct in6_rtmsg32 __user *ur6 = argp;
2820                ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2821                        3 * sizeof(struct in6_addr));
2822                ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2823                ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2824                ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2825                ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2826                ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2827                ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2828                ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2829
2830                r = (void *) &r6;
2831        } else { /* ipv4 */
2832                struct rtentry32 __user *ur4 = argp;
2833                ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
2834                                        3 * sizeof(struct sockaddr));
2835                ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
2836                ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
2837                ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
2838                ret |= __get_user(r4.rt_window, &(ur4->rt_window));
2839                ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
2840                ret |= __get_user(rtdev, &(ur4->rt_dev));
2841                if (rtdev) {
2842                        ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
2843                        r4.rt_dev = devname; devname[15] = 0;
2844                } else
2845                        r4.rt_dev = NULL;
2846
2847                r = (void *) &r4;
2848        }
2849
2850        if (ret) {
2851                ret = -EFAULT;
2852                goto out;
2853        }
2854
2855        set_fs(KERNEL_DS);
2856        ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
2857        set_fs(old_fs);
2858
2859out:
2860        return ret;
2861}
2862
2863/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2864 * for some operations; this forces use of the newer bridge-utils that
2865 * use compatiable ioctls
2866 */
2867static int old_bridge_ioctl(compat_ulong_t __user *argp)
2868{
2869        compat_ulong_t tmp;
2870
2871        if (get_user(tmp, argp))
2872                return -EFAULT;
2873        if (tmp == BRCTL_GET_VERSION)
2874                return BRCTL_VERSION + 1;
2875        return -EINVAL;
2876}
2877
2878static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2879                         unsigned int cmd, unsigned long arg)
2880{
2881        void __user *argp = compat_ptr(arg);
2882        struct sock *sk = sock->sk;
2883        struct net *net = sock_net(sk);
2884
2885        if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2886                return siocdevprivate_ioctl(net, cmd, argp);
2887
2888        switch (cmd) {
2889        case SIOCSIFBR:
2890        case SIOCGIFBR:
2891                return old_bridge_ioctl(argp);
2892        case SIOCGIFNAME:
2893                return dev_ifname32(net, argp);
2894        case SIOCGIFCONF:
2895                return dev_ifconf(net, argp);
2896        case SIOCETHTOOL:
2897                return ethtool_ioctl(net, argp);
2898        case SIOCWANDEV:
2899                return compat_siocwandev(net, argp);
2900        case SIOCGIFMAP:
2901        case SIOCSIFMAP:
2902                return compat_sioc_ifmap(net, cmd, argp);
2903        case SIOCBONDENSLAVE:
2904        case SIOCBONDRELEASE:
2905        case SIOCBONDSETHWADDR:
2906        case SIOCBONDSLAVEINFOQUERY:
2907        case SIOCBONDINFOQUERY:
2908        case SIOCBONDCHANGEACTIVE:
2909                return bond_ioctl(net, cmd, argp);
2910        case SIOCADDRT:
2911        case SIOCDELRT:
2912                return routing_ioctl(net, sock, cmd, argp);
2913        case SIOCGSTAMP:
2914                return do_siocgstamp(net, sock, cmd, argp);
2915        case SIOCGSTAMPNS:
2916                return do_siocgstampns(net, sock, cmd, argp);
2917        case SIOCSHWTSTAMP:
2918                return compat_siocshwtstamp(net, argp);
2919
2920        case FIOSETOWN:
2921        case SIOCSPGRP:
2922        case FIOGETOWN:
2923        case SIOCGPGRP:
2924        case SIOCBRADDBR:
2925        case SIOCBRDELBR:
2926        case SIOCGIFVLAN:
2927        case SIOCSIFVLAN:
2928        case SIOCADDDLCI:
2929        case SIOCDELDLCI:
2930                return sock_ioctl(file, cmd, arg);
2931
2932        case SIOCGIFFLAGS:
2933        case SIOCSIFFLAGS:
2934        case SIOCGIFMETRIC:
2935        case SIOCSIFMETRIC:
2936        case SIOCGIFMTU:
2937        case SIOCSIFMTU:
2938        case SIOCGIFMEM:
2939        case SIOCSIFMEM:
2940        case SIOCGIFHWADDR:
2941        case SIOCSIFHWADDR:
2942        case SIOCADDMULTI:
2943        case SIOCDELMULTI:
2944        case SIOCGIFINDEX:
2945        case SIOCGIFADDR:
2946        case SIOCSIFADDR:
2947        case SIOCSIFHWBROADCAST:
2948        case SIOCDIFADDR:
2949        case SIOCGIFBRDADDR:
2950        case SIOCSIFBRDADDR:
2951        case SIOCGIFDSTADDR:
2952        case SIOCSIFDSTADDR:
2953        case SIOCGIFNETMASK:
2954        case SIOCSIFNETMASK:
2955        case SIOCSIFPFLAGS:
2956        case SIOCGIFPFLAGS:
2957        case SIOCGIFTXQLEN:
2958        case SIOCSIFTXQLEN:
2959        case SIOCBRADDIF:
2960        case SIOCBRDELIF:
2961        case SIOCSIFNAME:
2962        case SIOCGMIIPHY:
2963        case SIOCGMIIREG:
2964        case SIOCSMIIREG:
2965                return dev_ifsioc(net, sock, cmd, argp);
2966
2967        case SIOCSARP:
2968        case SIOCGARP:
2969        case SIOCDARP:
2970        case SIOCATMARK:
2971                return sock_do_ioctl(net, sock, cmd, arg);
2972        }
2973
2974        /* Prevent warning from compat_sys_ioctl, these always
2975         * result in -EINVAL in the native case anyway. */
2976        switch (cmd) {
2977        case SIOCRTMSG:
2978        case SIOCGIFCOUNT:
2979        case SIOCSRARP:
2980        case SIOCGRARP:
2981        case SIOCDRARP:
2982        case SIOCSIFLINK:
2983        case SIOCGIFSLAVE:
2984        case SIOCSIFSLAVE:
2985                return -EINVAL;
2986        }
2987
2988        return -ENOIOCTLCMD;
2989}
2990
2991static long compat_sock_ioctl(struct file *file, unsigned cmd,
2992                              unsigned long arg)
2993{
2994        struct socket *sock = file->private_data;
2995        int ret = -ENOIOCTLCMD;
2996        struct sock *sk;
2997        struct net *net;
2998
2999        sk = sock->sk;
3000        net = sock_net(sk);
3001
3002        if (sock->ops->compat_ioctl)
3003                ret = sock->ops->compat_ioctl(sock, cmd, arg);
3004
3005        if (ret == -ENOIOCTLCMD &&
3006            (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3007                ret = compat_wext_handle_ioctl(net, cmd, arg);
3008
3009        if (ret == -ENOIOCTLCMD)
3010                ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3011
3012        return ret;
3013}
3014#endif
3015
3016int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3017{
3018        return sock->ops->bind(sock, addr, addrlen);
3019}
3020EXPORT_SYMBOL(kernel_bind);
3021
3022int kernel_listen(struct socket *sock, int backlog)
3023{
3024        return sock->ops->listen(sock, backlog);
3025}
3026EXPORT_SYMBOL(kernel_listen);
3027
3028int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3029{
3030        struct sock *sk = sock->sk;
3031        int err;
3032
3033        err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3034                               newsock);
3035        if (err < 0)
3036                goto done;
3037
3038        err = sock->ops->accept(sock, *newsock, flags);
3039        if (err < 0) {
3040                sock_release(*newsock);
3041                *newsock = NULL;
3042                goto done;
3043        }
3044
3045        (*newsock)->ops = sock->ops;
3046        __module_get((*newsock)->ops->owner);
3047
3048done:
3049        return err;
3050}
3051EXPORT_SYMBOL(kernel_accept);
3052
3053int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3054                   int flags)
3055{
3056        return sock->ops->connect(sock, addr, addrlen, flags);
3057}
3058EXPORT_SYMBOL(kernel_connect);
3059
3060int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3061                         int *addrlen)
3062{
3063        return sock->ops->getname(sock, addr, addrlen, 0);
3064}
3065EXPORT_SYMBOL(kernel_getsockname);
3066
3067int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3068                         int *addrlen)
3069{
3070        return sock->ops->getname(sock, addr, addrlen, 1);
3071}
3072EXPORT_SYMBOL(kernel_getpeername);
3073
3074int kernel_getsockopt(struct socket *sock, int level, int optname,
3075                        char *optval, int *optlen)
3076{
3077        mm_segment_t oldfs = get_fs();
3078        char __user *uoptval;
3079        int __user *uoptlen;
3080        int err;
3081
3082        uoptval = (char __user __force *) optval;
3083        uoptlen = (int __user __force *) optlen;
3084
3085        set_fs(KERNEL_DS);
3086        if (level == SOL_SOCKET)
3087                err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3088        else
3089                err = sock->ops->getsockopt(sock, level, optname, uoptval,
3090                                            uoptlen);
3091        set_fs(oldfs);
3092        return err;
3093}
3094EXPORT_SYMBOL(kernel_getsockopt);
3095
3096int kernel_setsockopt(struct socket *sock, int level, int optname,
3097                        char *optval, unsigned int optlen)
3098{
3099        mm_segment_t oldfs = get_fs();
3100        char __user *uoptval;
3101        int err;
3102
3103        uoptval = (char __user __force *) optval;
3104
3105        set_fs(KERNEL_DS);
3106        if (level == SOL_SOCKET)
3107                err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3108        else
3109                err = sock->ops->setsockopt(sock, level, optname, uoptval,
3110                                            optlen);
3111        set_fs(oldfs);
3112        return err;
3113}
3114EXPORT_SYMBOL(kernel_setsockopt);
3115
3116int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3117                    size_t size, int flags)
3118{
3119        sock_update_classid(sock->sk);
3120
3121        if (sock->ops->sendpage)
3122                return sock->ops->sendpage(sock, page, offset, size, flags);
3123
3124        return sock_no_sendpage(sock, page, offset, size, flags);
3125}
3126EXPORT_SYMBOL(kernel_sendpage);
3127
3128int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3129{
3130        mm_segment_t oldfs = get_fs();
3131        int err;
3132
3133        set_fs(KERNEL_DS);
3134        err = sock->ops->ioctl(sock, cmd, arg);
3135        set_fs(oldfs);
3136
3137        return err;
3138}
3139EXPORT_SYMBOL(kernel_sock_ioctl);
3140
3141int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3142{
3143        return sock->ops->shutdown(sock, how);
3144}
3145EXPORT_SYMBOL(kernel_sock_shutdown);
3146