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/if_bridge.h>
  73#include <linux/if_frad.h>
  74#include <linux/if_vlan.h>
  75#include <linux/init.h>
  76#include <linux/poll.h>
  77#include <linux/cache.h>
  78#include <linux/module.h>
  79#include <linux/highmem.h>
  80#include <linux/mount.h>
  81#include <linux/security.h>
  82#include <linux/syscalls.h>
  83#include <linux/compat.h>
  84#include <linux/kmod.h>
  85#include <linux/audit.h>
  86#include <linux/wireless.h>
  87#include <linux/nsproxy.h>
  88#include <linux/magic.h>
  89#include <linux/slab.h>
  90#include <linux/xattr.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 __rcu *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_storage *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_storage *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        struct socket_wq *wq;
 244
 245        ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
 246        if (!ei)
 247                return NULL;
 248        wq = kmalloc(sizeof(*wq), GFP_KERNEL);
 249        if (!wq) {
 250                kmem_cache_free(sock_inode_cachep, ei);
 251                return NULL;
 252        }
 253        init_waitqueue_head(&wq->wait);
 254        wq->fasync_list = NULL;
 255        RCU_INIT_POINTER(ei->socket.wq, wq);
 256
 257        ei->socket.state = SS_UNCONNECTED;
 258        ei->socket.flags = 0;
 259        ei->socket.ops = NULL;
 260        ei->socket.sk = NULL;
 261        ei->socket.file = NULL;
 262
 263        return &ei->vfs_inode;
 264}
 265
 266static void sock_destroy_inode(struct inode *inode)
 267{
 268        struct socket_alloc *ei;
 269        struct socket_wq *wq;
 270
 271        ei = container_of(inode, struct socket_alloc, vfs_inode);
 272        wq = rcu_dereference_protected(ei->socket.wq, 1);
 273        kfree_rcu(wq, rcu);
 274        kmem_cache_free(sock_inode_cachep, ei);
 275}
 276
 277static void init_once(void *foo)
 278{
 279        struct socket_alloc *ei = (struct socket_alloc *)foo;
 280
 281        inode_init_once(&ei->vfs_inode);
 282}
 283
 284static int init_inodecache(void)
 285{
 286        sock_inode_cachep = kmem_cache_create("sock_inode_cache",
 287                                              sizeof(struct socket_alloc),
 288                                              0,
 289                                              (SLAB_HWCACHE_ALIGN |
 290                                               SLAB_RECLAIM_ACCOUNT |
 291                                               SLAB_MEM_SPREAD),
 292                                              init_once);
 293        if (sock_inode_cachep == NULL)
 294                return -ENOMEM;
 295        return 0;
 296}
 297
 298static const struct super_operations sockfs_ops = {
 299        .alloc_inode    = sock_alloc_inode,
 300        .destroy_inode  = sock_destroy_inode,
 301        .statfs         = simple_statfs,
 302};
 303
 304/*
 305 * sockfs_dname() is called from d_path().
 306 */
 307static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
 308{
 309        return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
 310                                dentry->d_inode->i_ino);
 311}
 312
 313static const struct dentry_operations sockfs_dentry_operations = {
 314        .d_dname  = sockfs_dname,
 315};
 316
 317static struct dentry *sockfs_mount(struct file_system_type *fs_type,
 318                         int flags, const char *dev_name, void *data)
 319{
 320        return mount_pseudo(fs_type, "socket:", &sockfs_ops,
 321                &sockfs_dentry_operations, SOCKFS_MAGIC);
 322}
 323
 324static struct vfsmount *sock_mnt __read_mostly;
 325
 326static struct file_system_type sock_fs_type = {
 327        .name =         "sockfs",
 328        .mount =        sockfs_mount,
 329        .kill_sb =      kill_anon_super,
 330};
 331
 332/*
 333 *      Obtains the first available file descriptor and sets it up for use.
 334 *
 335 *      These functions create file structures and maps them to fd space
 336 *      of the current process. On success it returns file descriptor
 337 *      and file struct implicitly stored in sock->file.
 338 *      Note that another thread may close file descriptor before we return
 339 *      from this function. We use the fact that now we do not refer
 340 *      to socket after mapping. If one day we will need it, this
 341 *      function will increment ref. count on file by 1.
 342 *
 343 *      In any case returned fd MAY BE not valid!
 344 *      This race condition is unavoidable
 345 *      with shared fd spaces, we cannot solve it inside kernel,
 346 *      but we take care of internal coherence yet.
 347 */
 348
 349struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
 350{
 351        struct qstr name = { .name = "" };
 352        struct path path;
 353        struct file *file;
 354
 355        if (dname) {
 356                name.name = dname;
 357                name.len = strlen(name.name);
 358        } else if (sock->sk) {
 359                name.name = sock->sk->sk_prot_creator->name;
 360                name.len = strlen(name.name);
 361        }
 362        path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
 363        if (unlikely(!path.dentry))
 364                return ERR_PTR(-ENOMEM);
 365        path.mnt = mntget(sock_mnt);
 366
 367        d_instantiate(path.dentry, SOCK_INODE(sock));
 368        SOCK_INODE(sock)->i_fop = &socket_file_ops;
 369
 370        file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
 371                  &socket_file_ops);
 372        if (unlikely(IS_ERR(file))) {
 373                /* drop dentry, keep inode */
 374                ihold(path.dentry->d_inode);
 375                path_put(&path);
 376                return file;
 377        }
 378
 379        sock->file = file;
 380        file->f_flags = O_RDWR | (flags & O_NONBLOCK);
 381        file->private_data = sock;
 382        return file;
 383}
 384EXPORT_SYMBOL(sock_alloc_file);
 385
 386static int sock_map_fd(struct socket *sock, int flags)
 387{
 388        struct file *newfile;
 389        int fd = get_unused_fd_flags(flags);
 390        if (unlikely(fd < 0))
 391                return fd;
 392
 393        newfile = sock_alloc_file(sock, flags, NULL);
 394        if (likely(!IS_ERR(newfile))) {
 395                fd_install(fd, newfile);
 396                return fd;
 397        }
 398
 399        put_unused_fd(fd);
 400        return PTR_ERR(newfile);
 401}
 402
 403struct socket *sock_from_file(struct file *file, int *err)
 404{
 405        if (file->f_op == &socket_file_ops)
 406                return file->private_data;      /* set in sock_map_fd */
 407
 408        *err = -ENOTSOCK;
 409        return NULL;
 410}
 411EXPORT_SYMBOL(sock_from_file);
 412
 413/**
 414 *      sockfd_lookup - Go from a file number to its socket slot
 415 *      @fd: file handle
 416 *      @err: pointer to an error code return
 417 *
 418 *      The file handle passed in is locked and the socket it is bound
 419 *      too is returned. If an error occurs the err pointer is overwritten
 420 *      with a negative errno code and NULL is returned. The function checks
 421 *      for both invalid handles and passing a handle which is not a socket.
 422 *
 423 *      On a success the socket object pointer is returned.
 424 */
 425
 426struct socket *sockfd_lookup(int fd, int *err)
 427{
 428        struct file *file;
 429        struct socket *sock;
 430
 431        file = fget(fd);
 432        if (!file) {
 433                *err = -EBADF;
 434                return NULL;
 435        }
 436
 437        sock = sock_from_file(file, err);
 438        if (!sock)
 439                fput(file);
 440        return sock;
 441}
 442EXPORT_SYMBOL(sockfd_lookup);
 443
 444static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 445{
 446        struct file *file;
 447        struct socket *sock;
 448
 449        *err = -EBADF;
 450        file = fget_light(fd, fput_needed);
 451        if (file) {
 452                sock = sock_from_file(file, err);
 453                if (sock)
 454                        return sock;
 455                fput_light(file, *fput_needed);
 456        }
 457        return NULL;
 458}
 459
 460#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
 461#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
 462#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
 463static ssize_t sockfs_getxattr(struct dentry *dentry,
 464                               const char *name, void *value, size_t size)
 465{
 466        const char *proto_name;
 467        size_t proto_size;
 468        int error;
 469
 470        error = -ENODATA;
 471        if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
 472                proto_name = dentry->d_name.name;
 473                proto_size = strlen(proto_name);
 474
 475                if (value) {
 476                        error = -ERANGE;
 477                        if (proto_size + 1 > size)
 478                                goto out;
 479
 480                        strncpy(value, proto_name, proto_size + 1);
 481                }
 482                error = proto_size + 1;
 483        }
 484
 485out:
 486        return error;
 487}
 488
 489static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
 490                                size_t size)
 491{
 492        ssize_t len;
 493        ssize_t used = 0;
 494
 495        len = security_inode_listsecurity(dentry->d_inode, buffer, size);
 496        if (len < 0)
 497                return len;
 498        used += len;
 499        if (buffer) {
 500                if (size < used)
 501                        return -ERANGE;
 502                buffer += len;
 503        }
 504
 505        len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
 506        used += len;
 507        if (buffer) {
 508                if (size < used)
 509                        return -ERANGE;
 510                memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
 511                buffer += len;
 512        }
 513
 514        return used;
 515}
 516
 517static const struct inode_operations sockfs_inode_ops = {
 518        .getxattr = sockfs_getxattr,
 519        .listxattr = sockfs_listxattr,
 520};
 521
 522/**
 523 *      sock_alloc      -       allocate a socket
 524 *
 525 *      Allocate a new inode and socket object. The two are bound together
 526 *      and initialised. The socket is then returned. If we are out of inodes
 527 *      NULL is returned.
 528 */
 529
 530static struct socket *sock_alloc(void)
 531{
 532        struct inode *inode;
 533        struct socket *sock;
 534
 535        inode = new_inode_pseudo(sock_mnt->mnt_sb);
 536        if (!inode)
 537                return NULL;
 538
 539        sock = SOCKET_I(inode);
 540
 541        kmemcheck_annotate_bitfield(sock, type);
 542        inode->i_ino = get_next_ino();
 543        inode->i_mode = S_IFSOCK | S_IRWXUGO;
 544        inode->i_uid = current_fsuid();
 545        inode->i_gid = current_fsgid();
 546        inode->i_op = &sockfs_inode_ops;
 547
 548        this_cpu_add(sockets_in_use, 1);
 549        return sock;
 550}
 551
 552/*
 553 *      In theory you can't get an open on this inode, but /proc provides
 554 *      a back door. Remember to keep it shut otherwise you'll let the
 555 *      creepy crawlies in.
 556 */
 557
 558static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
 559{
 560        return -ENXIO;
 561}
 562
 563const struct file_operations bad_sock_fops = {
 564        .owner = THIS_MODULE,
 565        .open = sock_no_open,
 566        .llseek = noop_llseek,
 567};
 568
 569/**
 570 *      sock_release    -       close a socket
 571 *      @sock: socket to close
 572 *
 573 *      The socket is released from the protocol stack if it has a release
 574 *      callback, and the inode is then released if the socket is bound to
 575 *      an inode not a file.
 576 */
 577
 578void sock_release(struct socket *sock)
 579{
 580        if (sock->ops) {
 581                struct module *owner = sock->ops->owner;
 582
 583                sock->ops->release(sock);
 584                sock->ops = NULL;
 585                module_put(owner);
 586        }
 587
 588        if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
 589                printk(KERN_ERR "sock_release: fasync list not empty!\n");
 590
 591        if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
 592                return;
 593
 594        this_cpu_sub(sockets_in_use, 1);
 595        if (!sock->file) {
 596                iput(SOCK_INODE(sock));
 597                return;
 598        }
 599        sock->file = NULL;
 600}
 601EXPORT_SYMBOL(sock_release);
 602
 603void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
 604{
 605        *tx_flags = 0;
 606        if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
 607                *tx_flags |= SKBTX_HW_TSTAMP;
 608        if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
 609                *tx_flags |= SKBTX_SW_TSTAMP;
 610        if (sock_flag(sk, SOCK_WIFI_STATUS))
 611                *tx_flags |= SKBTX_WIFI_STATUS;
 612}
 613EXPORT_SYMBOL(sock_tx_timestamp);
 614
 615static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
 616                                       struct msghdr *msg, size_t size)
 617{
 618        struct sock_iocb *si = kiocb_to_siocb(iocb);
 619
 620        si->sock = sock;
 621        si->scm = NULL;
 622        si->msg = msg;
 623        si->size = size;
 624
 625        return sock->ops->sendmsg(iocb, sock, msg, size);
 626}
 627
 628static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 629                                 struct msghdr *msg, size_t size)
 630{
 631        int err = security_socket_sendmsg(sock, msg, size);
 632
 633        return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
 634}
 635
 636int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 637{
 638        struct kiocb iocb;
 639        struct sock_iocb siocb;
 640        int ret;
 641
 642        init_sync_kiocb(&iocb, NULL);
 643        iocb.private = &siocb;
 644        ret = __sock_sendmsg(&iocb, sock, msg, size);
 645        if (-EIOCBQUEUED == ret)
 646                ret = wait_on_sync_kiocb(&iocb);
 647        return ret;
 648}
 649EXPORT_SYMBOL(sock_sendmsg);
 650
 651static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
 652{
 653        struct kiocb iocb;
 654        struct sock_iocb siocb;
 655        int ret;
 656
 657        init_sync_kiocb(&iocb, NULL);
 658        iocb.private = &siocb;
 659        ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
 660        if (-EIOCBQUEUED == ret)
 661                ret = wait_on_sync_kiocb(&iocb);
 662        return ret;
 663}
 664
 665int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 666                   struct kvec *vec, size_t num, size_t size)
 667{
 668        mm_segment_t oldfs = get_fs();
 669        int result;
 670
 671        set_fs(KERNEL_DS);
 672        /*
 673         * the following is safe, since for compiler definitions of kvec and
 674         * iovec are identical, yielding the same in-core layout and alignment
 675         */
 676        msg->msg_iov = (struct iovec *)vec;
 677        msg->msg_iovlen = num;
 678        result = sock_sendmsg(sock, msg, size);
 679        set_fs(oldfs);
 680        return result;
 681}
 682EXPORT_SYMBOL(kernel_sendmsg);
 683
 684/*
 685 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 686 */
 687void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 688        struct sk_buff *skb)
 689{
 690        int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 691        struct timespec ts[3];
 692        int empty = 1;
 693        struct skb_shared_hwtstamps *shhwtstamps =
 694                skb_hwtstamps(skb);
 695
 696        /* Race occurred between timestamp enabling and packet
 697           receiving.  Fill in the current time for now. */
 698        if (need_software_tstamp && skb->tstamp.tv64 == 0)
 699                __net_timestamp(skb);
 700
 701        if (need_software_tstamp) {
 702                if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 703                        struct timeval tv;
 704                        skb_get_timestamp(skb, &tv);
 705                        put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 706                                 sizeof(tv), &tv);
 707                } else {
 708                        skb_get_timestampns(skb, &ts[0]);
 709                        put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 710                                 sizeof(ts[0]), &ts[0]);
 711                }
 712        }
 713
 714
 715        memset(ts, 0, sizeof(ts));
 716        if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
 717            ktime_to_timespec_cond(skb->tstamp, ts + 0))
 718                empty = 0;
 719        if (shhwtstamps) {
 720                if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 721                    ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
 722                        empty = 0;
 723                if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 724                    ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
 725                        empty = 0;
 726        }
 727        if (!empty)
 728                put_cmsg(msg, SOL_SOCKET,
 729                         SCM_TIMESTAMPING, sizeof(ts), &ts);
 730}
 731EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 732
 733void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
 734        struct sk_buff *skb)
 735{
 736        int ack;
 737
 738        if (!sock_flag(sk, SOCK_WIFI_STATUS))
 739                return;
 740        if (!skb->wifi_acked_valid)
 741                return;
 742
 743        ack = skb->wifi_acked;
 744
 745        put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
 746}
 747EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
 748
 749static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 750                                   struct sk_buff *skb)
 751{
 752        if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 753                put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 754                        sizeof(__u32), &skb->dropcount);
 755}
 756
 757void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 758        struct sk_buff *skb)
 759{
 760        sock_recv_timestamp(msg, sk, skb);
 761        sock_recv_drops(msg, sk, skb);
 762}
 763EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 764
 765static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 766                                       struct msghdr *msg, size_t size, int flags)
 767{
 768        struct sock_iocb *si = kiocb_to_siocb(iocb);
 769
 770        si->sock = sock;
 771        si->scm = NULL;
 772        si->msg = msg;
 773        si->size = size;
 774        si->flags = flags;
 775
 776        return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 777}
 778
 779static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 780                                 struct msghdr *msg, size_t size, int flags)
 781{
 782        int err = security_socket_recvmsg(sock, msg, size, flags);
 783
 784        return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 785}
 786
 787int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 788                 size_t size, int flags)
 789{
 790        struct kiocb iocb;
 791        struct sock_iocb siocb;
 792        int ret;
 793
 794        init_sync_kiocb(&iocb, NULL);
 795        iocb.private = &siocb;
 796        ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 797        if (-EIOCBQUEUED == ret)
 798                ret = wait_on_sync_kiocb(&iocb);
 799        return ret;
 800}
 801EXPORT_SYMBOL(sock_recvmsg);
 802
 803static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 804                              size_t size, int flags)
 805{
 806        struct kiocb iocb;
 807        struct sock_iocb siocb;
 808        int ret;
 809
 810        init_sync_kiocb(&iocb, NULL);
 811        iocb.private = &siocb;
 812        ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 813        if (-EIOCBQUEUED == ret)
 814                ret = wait_on_sync_kiocb(&iocb);
 815        return ret;
 816}
 817
 818/**
 819 * kernel_recvmsg - Receive a message from a socket (kernel space)
 820 * @sock:       The socket to receive the message from
 821 * @msg:        Received message
 822 * @vec:        Input s/g array for message data
 823 * @num:        Size of input s/g array
 824 * @size:       Number of bytes to read
 825 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 826 *
 827 * On return the msg structure contains the scatter/gather array passed in the
 828 * vec argument. The array is modified so that it consists of the unfilled
 829 * portion of the original array.
 830 *
 831 * The returned value is the total number of bytes received, or an error.
 832 */
 833int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 834                   struct kvec *vec, size_t num, size_t size, int flags)
 835{
 836        mm_segment_t oldfs = get_fs();
 837        int result;
 838
 839        set_fs(KERNEL_DS);
 840        /*
 841         * the following is safe, since for compiler definitions of kvec and
 842         * iovec are identical, yielding the same in-core layout and alignment
 843         */
 844        msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 845        result = sock_recvmsg(sock, msg, size, flags);
 846        set_fs(oldfs);
 847        return result;
 848}
 849EXPORT_SYMBOL(kernel_recvmsg);
 850
 851static void sock_aio_dtor(struct kiocb *iocb)
 852{
 853        kfree(iocb->private);
 854}
 855
 856static ssize_t sock_sendpage(struct file *file, struct page *page,
 857                             int offset, size_t size, loff_t *ppos, int more)
 858{
 859        struct socket *sock;
 860        int flags;
 861
 862        sock = file->private_data;
 863
 864        flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 865        /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
 866        flags |= more;
 867
 868        return kernel_sendpage(sock, page, offset, size, flags);
 869}
 870
 871static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 872                                struct pipe_inode_info *pipe, size_t len,
 873                                unsigned int flags)
 874{
 875        struct socket *sock = file->private_data;
 876
 877        if (unlikely(!sock->ops->splice_read))
 878                return -EINVAL;
 879
 880        return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 881}
 882
 883static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 884                                         struct sock_iocb *siocb)
 885{
 886        if (!is_sync_kiocb(iocb)) {
 887                siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
 888                if (!siocb)
 889                        return NULL;
 890                iocb->ki_dtor = sock_aio_dtor;
 891        }
 892
 893        siocb->kiocb = iocb;
 894        iocb->private = siocb;
 895        return siocb;
 896}
 897
 898static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 899                struct file *file, const struct iovec *iov,
 900                unsigned long nr_segs)
 901{
 902        struct socket *sock = file->private_data;
 903        size_t size = 0;
 904        int i;
 905
 906        for (i = 0; i < nr_segs; i++)
 907                size += iov[i].iov_len;
 908
 909        msg->msg_name = NULL;
 910        msg->msg_namelen = 0;
 911        msg->msg_control = NULL;
 912        msg->msg_controllen = 0;
 913        msg->msg_iov = (struct iovec *)iov;
 914        msg->msg_iovlen = nr_segs;
 915        msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 916
 917        return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 918}
 919
 920static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 921                                unsigned long nr_segs, loff_t pos)
 922{
 923        struct sock_iocb siocb, *x;
 924
 925        if (pos != 0)
 926                return -ESPIPE;
 927
 928        if (iocb->ki_left == 0) /* Match SYS5 behaviour */
 929                return 0;
 930
 931
 932        x = alloc_sock_iocb(iocb, &siocb);
 933        if (!x)
 934                return -ENOMEM;
 935        return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 936}
 937
 938static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 939                        struct file *file, const struct iovec *iov,
 940                        unsigned long nr_segs)
 941{
 942        struct socket *sock = file->private_data;
 943        size_t size = 0;
 944        int i;
 945
 946        for (i = 0; i < nr_segs; i++)
 947                size += iov[i].iov_len;
 948
 949        msg->msg_name = NULL;
 950        msg->msg_namelen = 0;
 951        msg->msg_control = NULL;
 952        msg->msg_controllen = 0;
 953        msg->msg_iov = (struct iovec *)iov;
 954        msg->msg_iovlen = nr_segs;
 955        msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 956        if (sock->type == SOCK_SEQPACKET)
 957                msg->msg_flags |= MSG_EOR;
 958
 959        return __sock_sendmsg(iocb, sock, msg, size);
 960}
 961
 962static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 963                          unsigned long nr_segs, loff_t pos)
 964{
 965        struct sock_iocb siocb, *x;
 966
 967        if (pos != 0)
 968                return -ESPIPE;
 969
 970        x = alloc_sock_iocb(iocb, &siocb);
 971        if (!x)
 972                return -ENOMEM;
 973
 974        return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 975}
 976
 977/*
 978 * Atomic setting of ioctl hooks to avoid race
 979 * with module unload.
 980 */
 981
 982static DEFINE_MUTEX(br_ioctl_mutex);
 983static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 984
 985void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 986{
 987        mutex_lock(&br_ioctl_mutex);
 988        br_ioctl_hook = hook;
 989        mutex_unlock(&br_ioctl_mutex);
 990}
 991EXPORT_SYMBOL(brioctl_set);
 992
 993static DEFINE_MUTEX(vlan_ioctl_mutex);
 994static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 995
 996void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 997{
 998        mutex_lock(&vlan_ioctl_mutex);
 999        vlan_ioctl_hook = hook;
1000        mutex_unlock(&vlan_ioctl_mutex);
1001}
1002EXPORT_SYMBOL(vlan_ioctl_set);
1003
1004static DEFINE_MUTEX(dlci_ioctl_mutex);
1005static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1006
1007void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1008{
1009        mutex_lock(&dlci_ioctl_mutex);
1010        dlci_ioctl_hook = hook;
1011        mutex_unlock(&dlci_ioctl_mutex);
1012}
1013EXPORT_SYMBOL(dlci_ioctl_set);
1014
1015static long sock_do_ioctl(struct net *net, struct socket *sock,
1016                                 unsigned int cmd, unsigned long arg)
1017{
1018        int err;
1019        void __user *argp = (void __user *)arg;
1020
1021        err = sock->ops->ioctl(sock, cmd, arg);
1022
1023        /*
1024         * If this ioctl is unknown try to hand it down
1025         * to the NIC driver.
1026         */
1027        if (err == -ENOIOCTLCMD)
1028                err = dev_ioctl(net, cmd, argp);
1029
1030        return err;
1031}
1032
1033/*
1034 *      With an ioctl, arg may well be a user mode pointer, but we don't know
1035 *      what to do with it - that's up to the protocol still.
1036 */
1037
1038static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1039{
1040        struct socket *sock;
1041        struct sock *sk;
1042        void __user *argp = (void __user *)arg;
1043        int pid, err;
1044        struct net *net;
1045
1046        sock = file->private_data;
1047        sk = sock->sk;
1048        net = sock_net(sk);
1049        if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1050                err = dev_ioctl(net, cmd, argp);
1051        } else
1052#ifdef CONFIG_WEXT_CORE
1053        if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1054                err = dev_ioctl(net, cmd, argp);
1055        } else
1056#endif
1057                switch (cmd) {
1058                case FIOSETOWN:
1059                case SIOCSPGRP:
1060                        err = -EFAULT;
1061                        if (get_user(pid, (int __user *)argp))
1062                                break;
1063                        err = f_setown(sock->file, pid, 1);
1064                        break;
1065                case FIOGETOWN:
1066                case SIOCGPGRP:
1067                        err = put_user(f_getown(sock->file),
1068                                       (int __user *)argp);
1069                        break;
1070                case SIOCGIFBR:
1071                case SIOCSIFBR:
1072                case SIOCBRADDBR:
1073                case SIOCBRDELBR:
1074                        err = -ENOPKG;
1075                        if (!br_ioctl_hook)
1076                                request_module("bridge");
1077
1078                        mutex_lock(&br_ioctl_mutex);
1079                        if (br_ioctl_hook)
1080                                err = br_ioctl_hook(net, cmd, argp);
1081                        mutex_unlock(&br_ioctl_mutex);
1082                        break;
1083                case SIOCGIFVLAN:
1084                case SIOCSIFVLAN:
1085                        err = -ENOPKG;
1086                        if (!vlan_ioctl_hook)
1087                                request_module("8021q");
1088
1089                        mutex_lock(&vlan_ioctl_mutex);
1090                        if (vlan_ioctl_hook)
1091                                err = vlan_ioctl_hook(net, argp);
1092                        mutex_unlock(&vlan_ioctl_mutex);
1093                        break;
1094                case SIOCADDDLCI:
1095                case SIOCDELDLCI:
1096                        err = -ENOPKG;
1097                        if (!dlci_ioctl_hook)
1098                                request_module("dlci");
1099
1100                        mutex_lock(&dlci_ioctl_mutex);
1101                        if (dlci_ioctl_hook)
1102                                err = dlci_ioctl_hook(cmd, argp);
1103                        mutex_unlock(&dlci_ioctl_mutex);
1104                        break;
1105                default:
1106                        err = sock_do_ioctl(net, sock, cmd, arg);
1107                        break;
1108                }
1109        return err;
1110}
1111
1112int sock_create_lite(int family, int type, int protocol, struct socket **res)
1113{
1114        int err;
1115        struct socket *sock = NULL;
1116
1117        err = security_socket_create(family, type, protocol, 1);
1118        if (err)
1119                goto out;
1120
1121        sock = sock_alloc();
1122        if (!sock) {
1123                err = -ENOMEM;
1124                goto out;
1125        }
1126
1127        sock->type = type;
1128        err = security_socket_post_create(sock, family, type, protocol, 1);
1129        if (err)
1130                goto out_release;
1131
1132out:
1133        *res = sock;
1134        return err;
1135out_release:
1136        sock_release(sock);
1137        sock = NULL;
1138        goto out;
1139}
1140EXPORT_SYMBOL(sock_create_lite);
1141
1142/* No kernel lock held - perfect */
1143static unsigned int sock_poll(struct file *file, poll_table *wait)
1144{
1145        struct socket *sock;
1146
1147        /*
1148         *      We can't return errors to poll, so it's either yes or no.
1149         */
1150        sock = file->private_data;
1151        return sock->ops->poll(file, sock, wait);
1152}
1153
1154static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1155{
1156        struct socket *sock = file->private_data;
1157
1158        return sock->ops->mmap(file, sock, vma);
1159}
1160
1161static int sock_close(struct inode *inode, struct file *filp)
1162{
1163        sock_release(SOCKET_I(inode));
1164        return 0;
1165}
1166
1167/*
1168 *      Update the socket async list
1169 *
1170 *      Fasync_list locking strategy.
1171 *
1172 *      1. fasync_list is modified only under process context socket lock
1173 *         i.e. under semaphore.
1174 *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1175 *         or under socket lock
1176 */
1177
1178static int sock_fasync(int fd, struct file *filp, int on)
1179{
1180        struct socket *sock = filp->private_data;
1181        struct sock *sk = sock->sk;
1182        struct socket_wq *wq;
1183
1184        if (sk == NULL)
1185                return -EINVAL;
1186
1187        lock_sock(sk);
1188        wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1189        fasync_helper(fd, filp, on, &wq->fasync_list);
1190
1191        if (!wq->fasync_list)
1192                sock_reset_flag(sk, SOCK_FASYNC);
1193        else
1194                sock_set_flag(sk, SOCK_FASYNC);
1195
1196        release_sock(sk);
1197        return 0;
1198}
1199
1200/* This function may be called only under socket lock or callback_lock or rcu_lock */
1201
1202int sock_wake_async(struct socket *sock, int how, int band)
1203{
1204        struct socket_wq *wq;
1205
1206        if (!sock)
1207                return -1;
1208        rcu_read_lock();
1209        wq = rcu_dereference(sock->wq);
1210        if (!wq || !wq->fasync_list) {
1211                rcu_read_unlock();
1212                return -1;
1213        }
1214        switch (how) {
1215        case SOCK_WAKE_WAITD:
1216                if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1217                        break;
1218                goto call_kill;
1219        case SOCK_WAKE_SPACE:
1220                if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1221                        break;
1222                /* fall through */
1223        case SOCK_WAKE_IO:
1224call_kill:
1225                kill_fasync(&wq->fasync_list, SIGIO, band);
1226                break;
1227        case SOCK_WAKE_URG:
1228                kill_fasync(&wq->fasync_list, SIGURG, band);
1229        }
1230        rcu_read_unlock();
1231        return 0;
1232}
1233EXPORT_SYMBOL(sock_wake_async);
1234
1235int __sock_create(struct net *net, int family, int type, int protocol,
1236                         struct socket **res, int kern)
1237{
1238        int err;
1239        struct socket *sock;
1240        const struct net_proto_family *pf;
1241
1242        /*
1243         *      Check protocol is in range
1244         */
1245        if (family < 0 || family >= NPROTO)
1246                return -EAFNOSUPPORT;
1247        if (type < 0 || type >= SOCK_MAX)
1248                return -EINVAL;
1249
1250        /* Compatibility.
1251
1252           This uglymoron is moved from INET layer to here to avoid
1253           deadlock in module load.
1254         */
1255        if (family == PF_INET && type == SOCK_PACKET) {
1256                static int warned;
1257                if (!warned) {
1258                        warned = 1;
1259                        printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1260                               current->comm);
1261                }
1262                family = PF_PACKET;
1263        }
1264
1265        err = security_socket_create(family, type, protocol, kern);
1266        if (err)
1267                return err;
1268
1269        /*
1270         *      Allocate the socket and allow the family to set things up. if
1271         *      the protocol is 0, the family is instructed to select an appropriate
1272         *      default.
1273         */
1274        sock = sock_alloc();
1275        if (!sock) {
1276                net_warn_ratelimited("socket: no more sockets\n");
1277                return -ENFILE; /* Not exactly a match, but its the
1278                                   closest posix thing */
1279        }
1280
1281        sock->type = type;
1282
1283#ifdef CONFIG_MODULES
1284        /* Attempt to load a protocol module if the find failed.
1285         *
1286         * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1287         * requested real, full-featured networking support upon configuration.
1288         * Otherwise module support will break!
1289         */
1290        if (rcu_access_pointer(net_families[family]) == NULL)
1291                request_module("net-pf-%d", family);
1292#endif
1293
1294        rcu_read_lock();
1295        pf = rcu_dereference(net_families[family]);
1296        err = -EAFNOSUPPORT;
1297        if (!pf)
1298                goto out_release;
1299
1300        /*
1301         * We will call the ->create function, that possibly is in a loadable
1302         * module, so we have to bump that loadable module refcnt first.
1303         */
1304        if (!try_module_get(pf->owner))
1305                goto out_release;
1306
1307        /* Now protected by module ref count */
1308        rcu_read_unlock();
1309
1310        err = pf->create(net, sock, protocol, kern);
1311        if (err < 0)
1312                goto out_module_put;
1313
1314        /*
1315         * Now to bump the refcnt of the [loadable] module that owns this
1316         * socket at sock_release time we decrement its refcnt.
1317         */
1318        if (!try_module_get(sock->ops->owner))
1319                goto out_module_busy;
1320
1321        /*
1322         * Now that we're done with the ->create function, the [loadable]
1323         * module can have its refcnt decremented
1324         */
1325        module_put(pf->owner);
1326        err = security_socket_post_create(sock, family, type, protocol, kern);
1327        if (err)
1328                goto out_sock_release;
1329        *res = sock;
1330
1331        return 0;
1332
1333out_module_busy:
1334        err = -EAFNOSUPPORT;
1335out_module_put:
1336        sock->ops = NULL;
1337        module_put(pf->owner);
1338out_sock_release:
1339        sock_release(sock);
1340        return err;
1341
1342out_release:
1343        rcu_read_unlock();
1344        goto out_sock_release;
1345}
1346EXPORT_SYMBOL(__sock_create);
1347
1348int sock_create(int family, int type, int protocol, struct socket **res)
1349{
1350        return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1351}
1352EXPORT_SYMBOL(sock_create);
1353
1354int sock_create_kern(int family, int type, int protocol, struct socket **res)
1355{
1356        return __sock_create(&init_net, family, type, protocol, res, 1);
1357}
1358EXPORT_SYMBOL(sock_create_kern);
1359
1360SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1361{
1362        int retval;
1363        struct socket *sock;
1364        int flags;
1365
1366        /* Check the SOCK_* constants for consistency.  */
1367        BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1368        BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1369        BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1370        BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1371
1372        flags = type & ~SOCK_TYPE_MASK;
1373        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1374                return -EINVAL;
1375        type &= SOCK_TYPE_MASK;
1376
1377        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1378                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1379
1380        retval = sock_create(family, type, protocol, &sock);
1381        if (retval < 0)
1382                goto out;
1383
1384        retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1385        if (retval < 0)
1386                goto out_release;
1387
1388out:
1389        /* It may be already another descriptor 8) Not kernel problem. */
1390        return retval;
1391
1392out_release:
1393        sock_release(sock);
1394        return retval;
1395}
1396
1397/*
1398 *      Create a pair of connected sockets.
1399 */
1400
1401SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1402                int __user *, usockvec)
1403{
1404        struct socket *sock1, *sock2;
1405        int fd1, fd2, err;
1406        struct file *newfile1, *newfile2;
1407        int flags;
1408
1409        flags = type & ~SOCK_TYPE_MASK;
1410        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1411                return -EINVAL;
1412        type &= SOCK_TYPE_MASK;
1413
1414        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1415                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1416
1417        /*
1418         * Obtain the first socket and check if the underlying protocol
1419         * supports the socketpair call.
1420         */
1421
1422        err = sock_create(family, type, protocol, &sock1);
1423        if (err < 0)
1424                goto out;
1425
1426        err = sock_create(family, type, protocol, &sock2);
1427        if (err < 0)
1428                goto out_release_1;
1429
1430        err = sock1->ops->socketpair(sock1, sock2);
1431        if (err < 0)
1432                goto out_release_both;
1433
1434        fd1 = get_unused_fd_flags(flags);
1435        if (unlikely(fd1 < 0)) {
1436                err = fd1;
1437                goto out_release_both;
1438        }
1439        fd2 = get_unused_fd_flags(flags);
1440        if (unlikely(fd2 < 0)) {
1441                err = fd2;
1442                put_unused_fd(fd1);
1443                goto out_release_both;
1444        }
1445
1446        newfile1 = sock_alloc_file(sock1, flags, NULL);
1447        if (unlikely(IS_ERR(newfile1))) {
1448                err = PTR_ERR(newfile1);
1449                put_unused_fd(fd1);
1450                put_unused_fd(fd2);
1451                goto out_release_both;
1452        }
1453
1454        newfile2 = sock_alloc_file(sock2, flags, NULL);
1455        if (IS_ERR(newfile2)) {
1456                err = PTR_ERR(newfile2);
1457                fput(newfile1);
1458                put_unused_fd(fd1);
1459                put_unused_fd(fd2);
1460                sock_release(sock2);
1461                goto out;
1462        }
1463
1464        audit_fd_pair(fd1, fd2);
1465        fd_install(fd1, newfile1);
1466        fd_install(fd2, newfile2);
1467        /* fd1 and fd2 may be already another descriptors.
1468         * Not kernel problem.
1469         */
1470
1471        err = put_user(fd1, &usockvec[0]);
1472        if (!err)
1473                err = put_user(fd2, &usockvec[1]);
1474        if (!err)
1475                return 0;
1476
1477        sys_close(fd2);
1478        sys_close(fd1);
1479        return err;
1480
1481out_release_both:
1482        sock_release(sock2);
1483out_release_1:
1484        sock_release(sock1);
1485out:
1486        return err;
1487}
1488
1489/*
1490 *      Bind a name to a socket. Nothing much to do here since it's
1491 *      the protocol's responsibility to handle the local address.
1492 *
1493 *      We move the socket address to kernel space before we call
1494 *      the protocol layer (having also checked the address is ok).
1495 */
1496
1497SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1498{
1499        struct socket *sock;
1500        struct sockaddr_storage address;
1501        int err, fput_needed;
1502
1503        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1504        if (sock) {
1505                err = move_addr_to_kernel(umyaddr, addrlen, &address);
1506                if (err >= 0) {
1507                        err = security_socket_bind(sock,
1508                                                   (struct sockaddr *)&address,
1509                                                   addrlen);
1510                        if (!err)
1511                                err = sock->ops->bind(sock,
1512                                                      (struct sockaddr *)
1513                                                      &address, addrlen);
1514                }
1515                fput_light(sock->file, fput_needed);
1516        }
1517        return err;
1518}
1519
1520/*
1521 *      Perform a listen. Basically, we allow the protocol to do anything
1522 *      necessary for a listen, and if that works, we mark the socket as
1523 *      ready for listening.
1524 */
1525
1526SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1527{
1528        struct socket *sock;
1529        int err, fput_needed;
1530        int somaxconn;
1531
1532        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1533        if (sock) {
1534                somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1535                if ((unsigned int)backlog > somaxconn)
1536                        backlog = somaxconn;
1537
1538                err = security_socket_listen(sock, backlog);
1539                if (!err)
1540                        err = sock->ops->listen(sock, backlog);
1541
1542                fput_light(sock->file, fput_needed);
1543        }
1544        return err;
1545}
1546
1547/*
1548 *      For accept, we attempt to create a new socket, set up the link
1549 *      with the client, wake up the client, then return the new
1550 *      connected fd. We collect the address of the connector in kernel
1551 *      space and move it to user at the very end. This is unclean because
1552 *      we open the socket then return an error.
1553 *
1554 *      1003.1g adds the ability to recvmsg() to query connection pending
1555 *      status to recvmsg. We need to add that support in a way thats
1556 *      clean when we restucture accept also.
1557 */
1558
1559SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1560                int __user *, upeer_addrlen, int, flags)
1561{
1562        struct socket *sock, *newsock;
1563        struct file *newfile;
1564        int err, len, newfd, fput_needed;
1565        struct sockaddr_storage address;
1566
1567        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1568                return -EINVAL;
1569
1570        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1571                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1572
1573        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1574        if (!sock)
1575                goto out;
1576
1577        err = -ENFILE;
1578        newsock = sock_alloc();
1579        if (!newsock)
1580                goto out_put;
1581
1582        newsock->type = sock->type;
1583        newsock->ops = sock->ops;
1584
1585        /*
1586         * We don't need try_module_get here, as the listening socket (sock)
1587         * has the protocol module (sock->ops->owner) held.
1588         */
1589        __module_get(newsock->ops->owner);
1590
1591        newfd = get_unused_fd_flags(flags);
1592        if (unlikely(newfd < 0)) {
1593                err = newfd;
1594                sock_release(newsock);
1595                goto out_put;
1596        }
1597        newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1598        if (unlikely(IS_ERR(newfile))) {
1599                err = PTR_ERR(newfile);
1600                put_unused_fd(newfd);
1601                sock_release(newsock);
1602                goto out_put;
1603        }
1604
1605        err = security_socket_accept(sock, newsock);
1606        if (err)
1607                goto out_fd;
1608
1609        err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1610        if (err < 0)
1611                goto out_fd;
1612
1613        if (upeer_sockaddr) {
1614                if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1615                                          &len, 2) < 0) {
1616                        err = -ECONNABORTED;
1617                        goto out_fd;
1618                }
1619                err = move_addr_to_user(&address,
1620                                        len, upeer_sockaddr, upeer_addrlen);
1621                if (err < 0)
1622                        goto out_fd;
1623        }
1624
1625        /* File flags are not inherited via accept() unlike another OSes. */
1626
1627        fd_install(newfd, newfile);
1628        err = newfd;
1629
1630out_put:
1631        fput_light(sock->file, fput_needed);
1632out:
1633        return err;
1634out_fd:
1635        fput(newfile);
1636        put_unused_fd(newfd);
1637        goto out_put;
1638}
1639
1640SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1641                int __user *, upeer_addrlen)
1642{
1643        return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1644}
1645
1646/*
1647 *      Attempt to connect to a socket with the server address.  The address
1648 *      is in user space so we verify it is OK and move it to kernel space.
1649 *
1650 *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1651 *      break bindings
1652 *
1653 *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1654 *      other SEQPACKET protocols that take time to connect() as it doesn't
1655 *      include the -EINPROGRESS status for such sockets.
1656 */
1657
1658SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1659                int, addrlen)
1660{
1661        struct socket *sock;
1662        struct sockaddr_storage address;
1663        int err, fput_needed;
1664
1665        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1666        if (!sock)
1667                goto out;
1668        err = move_addr_to_kernel(uservaddr, addrlen, &address);
1669        if (err < 0)
1670                goto out_put;
1671
1672        err =
1673            security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1674        if (err)
1675                goto out_put;
1676
1677        err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1678                                 sock->file->f_flags);
1679out_put:
1680        fput_light(sock->file, fput_needed);
1681out:
1682        return err;
1683}
1684
1685/*
1686 *      Get the local address ('name') of a socket object. Move the obtained
1687 *      name to user space.
1688 */
1689
1690SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1691                int __user *, usockaddr_len)
1692{
1693        struct socket *sock;
1694        struct sockaddr_storage address;
1695        int len, err, fput_needed;
1696
1697        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1698        if (!sock)
1699                goto out;
1700
1701        err = security_socket_getsockname(sock);
1702        if (err)
1703                goto out_put;
1704
1705        err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1706        if (err)
1707                goto out_put;
1708        err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1709
1710out_put:
1711        fput_light(sock->file, fput_needed);
1712out:
1713        return err;
1714}
1715
1716/*
1717 *      Get the remote address ('name') of a socket object. Move the obtained
1718 *      name to user space.
1719 */
1720
1721SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1722                int __user *, usockaddr_len)
1723{
1724        struct socket *sock;
1725        struct sockaddr_storage address;
1726        int len, err, fput_needed;
1727
1728        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1729        if (sock != NULL) {
1730                err = security_socket_getpeername(sock);
1731                if (err) {
1732                        fput_light(sock->file, fput_needed);
1733                        return err;
1734                }
1735
1736                err =
1737                    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1738                                       1);
1739                if (!err)
1740                        err = move_addr_to_user(&address, len, usockaddr,
1741                                                usockaddr_len);
1742                fput_light(sock->file, fput_needed);
1743        }
1744        return err;
1745}
1746
1747/*
1748 *      Send a datagram to a given address. We move the address into kernel
1749 *      space and check the user space data area is readable before invoking
1750 *      the protocol.
1751 */
1752
1753SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1754                unsigned int, flags, struct sockaddr __user *, addr,
1755                int, addr_len)
1756{
1757        struct socket *sock;
1758        struct sockaddr_storage address;
1759        int err;
1760        struct msghdr msg;
1761        struct iovec iov;
1762        int fput_needed;
1763
1764        if (len > INT_MAX)
1765                len = INT_MAX;
1766        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1767        if (!sock)
1768                goto out;
1769
1770        iov.iov_base = buff;
1771        iov.iov_len = len;
1772        msg.msg_name = NULL;
1773        msg.msg_iov = &iov;
1774        msg.msg_iovlen = 1;
1775        msg.msg_control = NULL;
1776        msg.msg_controllen = 0;
1777        msg.msg_namelen = 0;
1778        if (addr) {
1779                err = move_addr_to_kernel(addr, addr_len, &address);
1780                if (err < 0)
1781                        goto out_put;
1782                msg.msg_name = (struct sockaddr *)&address;
1783                msg.msg_namelen = addr_len;
1784        }
1785        if (sock->file->f_flags & O_NONBLOCK)
1786                flags |= MSG_DONTWAIT;
1787        msg.msg_flags = flags;
1788        err = sock_sendmsg(sock, &msg, len);
1789
1790out_put:
1791        fput_light(sock->file, fput_needed);
1792out:
1793        return err;
1794}
1795
1796/*
1797 *      Send a datagram down a socket.
1798 */
1799
1800SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1801                unsigned int, flags)
1802{
1803        return sys_sendto(fd, buff, len, flags, NULL, 0);
1804}
1805
1806/*
1807 *      Receive a frame from the socket and optionally record the address of the
1808 *      sender. We verify the buffers are writable and if needed move the
1809 *      sender address from kernel to user space.
1810 */
1811
1812SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1813                unsigned int, flags, struct sockaddr __user *, addr,
1814                int __user *, addr_len)
1815{
1816        struct socket *sock;
1817        struct iovec iov;
1818        struct msghdr msg;
1819        struct sockaddr_storage address;
1820        int err, err2;
1821        int fput_needed;
1822
1823        if (size > INT_MAX)
1824                size = INT_MAX;
1825        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1826        if (!sock)
1827                goto out;
1828
1829        msg.msg_control = NULL;
1830        msg.msg_controllen = 0;
1831        msg.msg_iovlen = 1;
1832        msg.msg_iov = &iov;
1833        iov.iov_len = size;
1834        iov.iov_base = ubuf;
1835        msg.msg_name = (struct sockaddr *)&address;
1836        msg.msg_namelen = sizeof(address);
1837        if (sock->file->f_flags & O_NONBLOCK)
1838                flags |= MSG_DONTWAIT;
1839        err = sock_recvmsg(sock, &msg, size, flags);
1840
1841        if (err >= 0 && addr != NULL) {
1842                err2 = move_addr_to_user(&address,
1843                                         msg.msg_namelen, addr, addr_len);
1844                if (err2 < 0)
1845                        err = err2;
1846        }
1847
1848        fput_light(sock->file, fput_needed);
1849out:
1850        return err;
1851}
1852
1853/*
1854 *      Receive a datagram from a socket.
1855 */
1856
1857asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1858                         unsigned int flags)
1859{
1860        return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1861}
1862
1863/*
1864 *      Set a socket option. Because we don't know the option lengths we have
1865 *      to pass the user mode parameter for the protocols to sort out.
1866 */
1867
1868SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1869                char __user *, optval, int, optlen)
1870{
1871        int err, fput_needed;
1872        struct socket *sock;
1873
1874        if (optlen < 0)
1875                return -EINVAL;
1876
1877        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1878        if (sock != NULL) {
1879                err = security_socket_setsockopt(sock, level, optname);
1880                if (err)
1881                        goto out_put;
1882
1883                if (level == SOL_SOCKET)
1884                        err =
1885                            sock_setsockopt(sock, level, optname, optval,
1886                                            optlen);
1887                else
1888                        err =
1889                            sock->ops->setsockopt(sock, level, optname, optval,
1890                                                  optlen);
1891out_put:
1892                fput_light(sock->file, fput_needed);
1893        }
1894        return err;
1895}
1896
1897/*
1898 *      Get a socket option. Because we don't know the option lengths we have
1899 *      to pass a user mode parameter for the protocols to sort out.
1900 */
1901
1902SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1903                char __user *, optval, int __user *, optlen)
1904{
1905        int err, fput_needed;
1906        struct socket *sock;
1907
1908        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1909        if (sock != NULL) {
1910                err = security_socket_getsockopt(sock, level, optname);
1911                if (err)
1912                        goto out_put;
1913
1914                if (level == SOL_SOCKET)
1915                        err =
1916                            sock_getsockopt(sock, level, optname, optval,
1917                                            optlen);
1918                else
1919                        err =
1920                            sock->ops->getsockopt(sock, level, optname, optval,
1921                                                  optlen);
1922out_put:
1923                fput_light(sock->file, fput_needed);
1924        }
1925        return err;
1926}
1927
1928/*
1929 *      Shutdown a socket.
1930 */
1931
1932SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1933{
1934        int err, fput_needed;
1935        struct socket *sock;
1936
1937        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1938        if (sock != NULL) {
1939                err = security_socket_shutdown(sock, how);
1940                if (!err)
1941                        err = sock->ops->shutdown(sock, how);
1942                fput_light(sock->file, fput_needed);
1943        }
1944        return err;
1945}
1946
1947/* A couple of helpful macros for getting the address of the 32/64 bit
1948 * fields which are the same type (int / unsigned) on our platforms.
1949 */
1950#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1951#define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1952#define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1953
1954struct used_address {
1955        struct sockaddr_storage name;
1956        unsigned int name_len;
1957};
1958
1959static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1960                         struct msghdr *msg_sys, unsigned int flags,
1961                         struct used_address *used_address)
1962{
1963        struct compat_msghdr __user *msg_compat =
1964            (struct compat_msghdr __user *)msg;
1965        struct sockaddr_storage address;
1966        struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1967        unsigned char ctl[sizeof(struct cmsghdr) + 20]
1968            __attribute__ ((aligned(sizeof(__kernel_size_t))));
1969        /* 20 is size of ipv6_pktinfo */
1970        unsigned char *ctl_buf = ctl;
1971        int err, ctl_len, total_len;
1972
1973        err = -EFAULT;
1974        if (MSG_CMSG_COMPAT & flags) {
1975                if (get_compat_msghdr(msg_sys, msg_compat))
1976                        return -EFAULT;
1977        } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1978                return -EFAULT;
1979
1980        if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1981                err = -EMSGSIZE;
1982                if (msg_sys->msg_iovlen > UIO_MAXIOV)
1983                        goto out;
1984                err = -ENOMEM;
1985                iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
1986                              GFP_KERNEL);
1987                if (!iov)
1988                        goto out;
1989        }
1990
1991        /* This will also move the address data into kernel space */
1992        if (MSG_CMSG_COMPAT & flags) {
1993                err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
1994        } else
1995                err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
1996        if (err < 0)
1997                goto out_freeiov;
1998        total_len = err;
1999
2000        err = -ENOBUFS;
2001
2002        if (msg_sys->msg_controllen > INT_MAX)
2003                goto out_freeiov;
2004        ctl_len = msg_sys->msg_controllen;
2005        if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2006                err =
2007                    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2008                                                     sizeof(ctl));
2009                if (err)
2010                        goto out_freeiov;
2011                ctl_buf = msg_sys->msg_control;
2012                ctl_len = msg_sys->msg_controllen;
2013        } else if (ctl_len) {
2014                if (ctl_len > sizeof(ctl)) {
2015                        ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2016                        if (ctl_buf == NULL)
2017                                goto out_freeiov;
2018                }
2019                err = -EFAULT;
2020                /*
2021                 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2022                 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2023                 * checking falls down on this.
2024                 */
2025                if (copy_from_user(ctl_buf,
2026                                   (void __user __force *)msg_sys->msg_control,
2027                                   ctl_len))
2028                        goto out_freectl;
2029                msg_sys->msg_control = ctl_buf;
2030        }
2031        msg_sys->msg_flags = flags;
2032
2033        if (sock->file->f_flags & O_NONBLOCK)
2034                msg_sys->msg_flags |= MSG_DONTWAIT;
2035        /*
2036         * If this is sendmmsg() and current destination address is same as
2037         * previously succeeded address, omit asking LSM's decision.
2038         * used_address->name_len is initialized to UINT_MAX so that the first
2039         * destination address never matches.
2040         */
2041        if (used_address && msg_sys->msg_name &&
2042            used_address->name_len == msg_sys->msg_namelen &&
2043            !memcmp(&used_address->name, msg_sys->msg_name,
2044                    used_address->name_len)) {
2045                err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2046                goto out_freectl;
2047        }
2048        err = sock_sendmsg(sock, msg_sys, total_len);
2049        /*
2050         * If this is sendmmsg() and sending to current destination address was
2051         * successful, remember it.
2052         */
2053        if (used_address && err >= 0) {
2054                used_address->name_len = msg_sys->msg_namelen;
2055                if (msg_sys->msg_name)
2056                        memcpy(&used_address->name, msg_sys->msg_name,
2057                               used_address->name_len);
2058        }
2059
2060out_freectl:
2061        if (ctl_buf != ctl)
2062                sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2063out_freeiov:
2064        if (iov != iovstack)
2065                kfree(iov);
2066out:
2067        return err;
2068}
2069
2070/*
2071 *      BSD sendmsg interface
2072 */
2073
2074long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2075{
2076        int fput_needed, err;
2077        struct msghdr msg_sys;
2078        struct socket *sock;
2079
2080        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2081        if (!sock)
2082                goto out;
2083
2084        err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2085
2086        fput_light(sock->file, fput_needed);
2087out:
2088        return err;
2089}
2090
2091SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2092{
2093        if (flags & MSG_CMSG_COMPAT)
2094                return -EINVAL;
2095        return __sys_sendmsg(fd, msg, flags);
2096}
2097
2098/*
2099 *      Linux sendmmsg interface
2100 */
2101
2102int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2103                   unsigned int flags)
2104{
2105        int fput_needed, err, datagrams;
2106        struct socket *sock;
2107        struct mmsghdr __user *entry;
2108        struct compat_mmsghdr __user *compat_entry;
2109        struct msghdr msg_sys;
2110        struct used_address used_address;
2111
2112        if (vlen > UIO_MAXIOV)
2113                vlen = UIO_MAXIOV;
2114
2115        datagrams = 0;
2116
2117        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2118        if (!sock)
2119                return err;
2120
2121        used_address.name_len = UINT_MAX;
2122        entry = mmsg;
2123        compat_entry = (struct compat_mmsghdr __user *)mmsg;
2124        err = 0;
2125
2126        while (datagrams < vlen) {
2127                if (MSG_CMSG_COMPAT & flags) {
2128                        err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2129                                             &msg_sys, flags, &used_address);
2130                        if (err < 0)
2131                                break;
2132                        err = __put_user(err, &compat_entry->msg_len);
2133                        ++compat_entry;
2134                } else {
2135                        err = ___sys_sendmsg(sock,
2136                                             (struct msghdr __user *)entry,
2137                                             &msg_sys, flags, &used_address);
2138                        if (err < 0)
2139                                break;
2140                        err = put_user(err, &entry->msg_len);
2141                        ++entry;
2142                }
2143
2144                if (err)
2145                        break;
2146                ++datagrams;
2147        }
2148
2149        fput_light(sock->file, fput_needed);
2150
2151        /* We only return an error if no datagrams were able to be sent */
2152        if (datagrams != 0)
2153                return datagrams;
2154
2155        return err;
2156}
2157
2158SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2159                unsigned int, vlen, unsigned int, flags)
2160{
2161        if (flags & MSG_CMSG_COMPAT)
2162                return -EINVAL;
2163        return __sys_sendmmsg(fd, mmsg, vlen, flags);
2164}
2165
2166static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2167                         struct msghdr *msg_sys, unsigned int flags, int nosec)
2168{
2169        struct compat_msghdr __user *msg_compat =
2170            (struct compat_msghdr __user *)msg;
2171        struct iovec iovstack[UIO_FASTIOV];
2172        struct iovec *iov = iovstack;
2173        unsigned long cmsg_ptr;
2174        int err, total_len, len;
2175
2176        /* kernel mode address */
2177        struct sockaddr_storage addr;
2178
2179        /* user mode address pointers */
2180        struct sockaddr __user *uaddr;
2181        int __user *uaddr_len;
2182
2183        if (MSG_CMSG_COMPAT & flags) {
2184                if (get_compat_msghdr(msg_sys, msg_compat))
2185                        return -EFAULT;
2186        } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
2187                return -EFAULT;
2188
2189        if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2190                err = -EMSGSIZE;
2191                if (msg_sys->msg_iovlen > UIO_MAXIOV)
2192                        goto out;
2193                err = -ENOMEM;
2194                iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2195                              GFP_KERNEL);
2196                if (!iov)
2197                        goto out;
2198        }
2199
2200        /*
2201         *      Save the user-mode address (verify_iovec will change the
2202         *      kernel msghdr to use the kernel address space)
2203         */
2204
2205        uaddr = (__force void __user *)msg_sys->msg_name;
2206        uaddr_len = COMPAT_NAMELEN(msg);
2207        if (MSG_CMSG_COMPAT & flags) {
2208                err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2209        } else
2210                err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2211        if (err < 0)
2212                goto out_freeiov;
2213        total_len = err;
2214
2215        cmsg_ptr = (unsigned long)msg_sys->msg_control;
2216        msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2217
2218        if (sock->file->f_flags & O_NONBLOCK)
2219                flags |= MSG_DONTWAIT;
2220        err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2221                                                          total_len, flags);
2222        if (err < 0)
2223                goto out_freeiov;
2224        len = err;
2225
2226        if (uaddr != NULL) {
2227                err = move_addr_to_user(&addr,
2228                                        msg_sys->msg_namelen, uaddr,
2229                                        uaddr_len);
2230                if (err < 0)
2231                        goto out_freeiov;
2232        }
2233        err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2234                         COMPAT_FLAGS(msg));
2235        if (err)
2236                goto out_freeiov;
2237        if (MSG_CMSG_COMPAT & flags)
2238                err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2239                                 &msg_compat->msg_controllen);
2240        else
2241                err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2242                                 &msg->msg_controllen);
2243        if (err)
2244                goto out_freeiov;
2245        err = len;
2246
2247out_freeiov:
2248        if (iov != iovstack)
2249                kfree(iov);
2250out:
2251        return err;
2252}
2253
2254/*
2255 *      BSD recvmsg interface
2256 */
2257
2258long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2259{
2260        int fput_needed, err;
2261        struct msghdr msg_sys;
2262        struct socket *sock;
2263
2264        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2265        if (!sock)
2266                goto out;
2267
2268        err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2269
2270        fput_light(sock->file, fput_needed);
2271out:
2272        return err;
2273}
2274
2275SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2276                unsigned int, flags)
2277{
2278        if (flags & MSG_CMSG_COMPAT)
2279                return -EINVAL;
2280        return __sys_recvmsg(fd, msg, flags);
2281}
2282
2283/*
2284 *     Linux recvmmsg interface
2285 */
2286
2287int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2288                   unsigned int flags, struct timespec *timeout)
2289{
2290        int fput_needed, err, datagrams;
2291        struct socket *sock;
2292        struct mmsghdr __user *entry;
2293        struct compat_mmsghdr __user *compat_entry;
2294        struct msghdr msg_sys;
2295        struct timespec end_time;
2296
2297        if (timeout &&
2298            poll_select_set_timeout(&end_time, timeout->tv_sec,
2299                                    timeout->tv_nsec))
2300                return -EINVAL;
2301
2302        datagrams = 0;
2303
2304        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2305        if (!sock)
2306                return err;
2307
2308        err = sock_error(sock->sk);
2309        if (err)
2310                goto out_put;
2311
2312        entry = mmsg;
2313        compat_entry = (struct compat_mmsghdr __user *)mmsg;
2314
2315        while (datagrams < vlen) {
2316                /*
2317                 * No need to ask LSM for more than the first datagram.
2318                 */
2319                if (MSG_CMSG_COMPAT & flags) {
2320                        err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2321                                             &msg_sys, flags & ~MSG_WAITFORONE,
2322                                             datagrams);
2323                        if (err < 0)
2324                                break;
2325                        err = __put_user(err, &compat_entry->msg_len);
2326                        ++compat_entry;
2327                } else {
2328                        err = ___sys_recvmsg(sock,
2329                                             (struct msghdr __user *)entry,
2330                                             &msg_sys, flags & ~MSG_WAITFORONE,
2331                                             datagrams);
2332                        if (err < 0)
2333                                break;
2334                        err = put_user(err, &entry->msg_len);
2335                        ++entry;
2336                }
2337
2338                if (err)
2339                        break;
2340                ++datagrams;
2341
2342                /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2343                if (flags & MSG_WAITFORONE)
2344                        flags |= MSG_DONTWAIT;
2345
2346                if (timeout) {
2347                        ktime_get_ts(timeout);
2348                        *timeout = timespec_sub(end_time, *timeout);
2349                        if (timeout->tv_sec < 0) {
2350                                timeout->tv_sec = timeout->tv_nsec = 0;
2351                                break;
2352                        }
2353
2354                        /* Timeout, return less than vlen datagrams */
2355                        if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2356                                break;
2357                }
2358
2359                /* Out of band data, return right away */
2360                if (msg_sys.msg_flags & MSG_OOB)
2361                        break;
2362        }
2363
2364out_put:
2365        fput_light(sock->file, fput_needed);
2366
2367        if (err == 0)
2368                return datagrams;
2369
2370        if (datagrams != 0) {
2371                /*
2372                 * We may return less entries than requested (vlen) if the
2373                 * sock is non block and there aren't enough datagrams...
2374                 */
2375                if (err != -EAGAIN) {
2376                        /*
2377                         * ... or  if recvmsg returns an error after we
2378                         * received some datagrams, where we record the
2379                         * error to return on the next call or if the
2380                         * app asks about it using getsockopt(SO_ERROR).
2381                         */
2382                        sock->sk->sk_err = -err;
2383                }
2384
2385                return datagrams;
2386        }
2387
2388        return err;
2389}
2390
2391SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2392                unsigned int, vlen, unsigned int, flags,
2393                struct timespec __user *, timeout)
2394{
2395        int datagrams;
2396        struct timespec timeout_sys;
2397
2398        if (flags & MSG_CMSG_COMPAT)
2399                return -EINVAL;
2400
2401        if (!timeout)
2402                return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2403
2404        if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2405                return -EFAULT;
2406
2407        datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2408
2409        if (datagrams > 0 &&
2410            copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2411                datagrams = -EFAULT;
2412
2413        return datagrams;
2414}
2415
2416#ifdef __ARCH_WANT_SYS_SOCKETCALL
2417/* Argument list sizes for sys_socketcall */
2418#define AL(x) ((x) * sizeof(unsigned long))
2419static const unsigned char nargs[21] = {
2420        AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2421        AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2422        AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2423        AL(4), AL(5), AL(4)
2424};
2425
2426#undef AL
2427
2428/*
2429 *      System call vectors.
2430 *
2431 *      Argument checking cleaned up. Saved 20% in size.
2432 *  This function doesn't need to set the kernel lock because
2433 *  it is set by the callees.
2434 */
2435
2436SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2437{
2438        unsigned long a[AUDITSC_ARGS];
2439        unsigned long a0, a1;
2440        int err;
2441        unsigned int len;
2442
2443        if (call < 1 || call > SYS_SENDMMSG)
2444                return -EINVAL;
2445
2446        len = nargs[call];
2447        if (len > sizeof(a))
2448                return -EINVAL;
2449
2450        /* copy_from_user should be SMP safe. */
2451        if (copy_from_user(a, args, len))
2452                return -EFAULT;
2453
2454        err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2455        if (err)
2456                return err;
2457
2458        a0 = a[0];
2459        a1 = a[1];
2460
2461        switch (call) {
2462        case SYS_SOCKET:
2463                err = sys_socket(a0, a1, a[2]);
2464                break;
2465        case SYS_BIND:
2466                err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2467                break;
2468        case SYS_CONNECT:
2469                err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2470                break;
2471        case SYS_LISTEN:
2472                err = sys_listen(a0, a1);
2473                break;
2474        case SYS_ACCEPT:
2475                err = sys_accept4(a0, (struct sockaddr __user *)a1,
2476                                  (int __user *)a[2], 0);
2477                break;
2478        case SYS_GETSOCKNAME:
2479                err =
2480                    sys_getsockname(a0, (struct sockaddr __user *)a1,
2481                                    (int __user *)a[2]);
2482                break;
2483        case SYS_GETPEERNAME:
2484                err =
2485                    sys_getpeername(a0, (struct sockaddr __user *)a1,
2486                                    (int __user *)a[2]);
2487                break;
2488        case SYS_SOCKETPAIR:
2489                err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2490                break;
2491        case SYS_SEND:
2492                err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2493                break;
2494        case SYS_SENDTO:
2495                err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2496                                 (struct sockaddr __user *)a[4], a[5]);
2497                break;
2498        case SYS_RECV:
2499                err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2500                break;
2501        case SYS_RECVFROM:
2502                err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2503                                   (struct sockaddr __user *)a[4],
2504                                   (int __user *)a[5]);
2505                break;
2506        case SYS_SHUTDOWN:
2507                err = sys_shutdown(a0, a1);
2508                break;
2509        case SYS_SETSOCKOPT:
2510                err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2511                break;
2512        case SYS_GETSOCKOPT:
2513                err =
2514                    sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2515                                   (int __user *)a[4]);
2516                break;
2517        case SYS_SENDMSG:
2518                err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2519                break;
2520        case SYS_SENDMMSG:
2521                err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2522                break;
2523        case SYS_RECVMSG:
2524                err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2525                break;
2526        case SYS_RECVMMSG:
2527                err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2528                                   (struct timespec __user *)a[4]);
2529                break;
2530        case SYS_ACCEPT4:
2531                err = sys_accept4(a0, (struct sockaddr __user *)a1,
2532                                  (int __user *)a[2], a[3]);
2533                break;
2534        default:
2535                err = -EINVAL;
2536                break;
2537        }
2538        return err;
2539}
2540
2541#endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2542
2543/**
2544 *      sock_register - add a socket protocol handler
2545 *      @ops: description of protocol
2546 *
2547 *      This function is called by a protocol handler that wants to
2548 *      advertise its address family, and have it linked into the
2549 *      socket interface. The value ops->family coresponds to the
2550 *      socket system call protocol family.
2551 */
2552int sock_register(const struct net_proto_family *ops)
2553{
2554        int err;
2555
2556        if (ops->family >= NPROTO) {
2557                printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2558                       NPROTO);
2559                return -ENOBUFS;
2560        }
2561
2562        spin_lock(&net_family_lock);
2563        if (rcu_dereference_protected(net_families[ops->family],
2564                                      lockdep_is_held(&net_family_lock)))
2565                err = -EEXIST;
2566        else {
2567                rcu_assign_pointer(net_families[ops->family], ops);
2568                err = 0;
2569        }
2570        spin_unlock(&net_family_lock);
2571
2572        printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2573        return err;
2574}
2575EXPORT_SYMBOL(sock_register);
2576
2577/**
2578 *      sock_unregister - remove a protocol handler
2579 *      @family: protocol family to remove
2580 *
2581 *      This function is called by a protocol handler that wants to
2582 *      remove its address family, and have it unlinked from the
2583 *      new socket creation.
2584 *
2585 *      If protocol handler is a module, then it can use module reference
2586 *      counts to protect against new references. If protocol handler is not
2587 *      a module then it needs to provide its own protection in
2588 *      the ops->create routine.
2589 */
2590void sock_unregister(int family)
2591{
2592        BUG_ON(family < 0 || family >= NPROTO);
2593
2594        spin_lock(&net_family_lock);
2595        RCU_INIT_POINTER(net_families[family], NULL);
2596        spin_unlock(&net_family_lock);
2597
2598        synchronize_rcu();
2599
2600        printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2601}
2602EXPORT_SYMBOL(sock_unregister);
2603
2604static int __init sock_init(void)
2605{
2606        int err;
2607        /*
2608         *      Initialize the network sysctl infrastructure.
2609         */
2610        err = net_sysctl_init();
2611        if (err)
2612                goto out;
2613
2614        /*
2615         *      Initialize skbuff SLAB cache
2616         */
2617        skb_init();
2618
2619        /*
2620         *      Initialize the protocols module.
2621         */
2622
2623        init_inodecache();
2624
2625        err = register_filesystem(&sock_fs_type);
2626        if (err)
2627                goto out_fs;
2628        sock_mnt = kern_mount(&sock_fs_type);
2629        if (IS_ERR(sock_mnt)) {
2630                err = PTR_ERR(sock_mnt);
2631                goto out_mount;
2632        }
2633
2634        /* The real protocol initialization is performed in later initcalls.
2635         */
2636
2637#ifdef CONFIG_NETFILTER
2638        netfilter_init();
2639#endif
2640
2641#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2642        skb_timestamping_init();
2643#endif
2644
2645out:
2646        return err;
2647
2648out_mount:
2649        unregister_filesystem(&sock_fs_type);
2650out_fs:
2651        goto out;
2652}
2653
2654core_initcall(sock_init);       /* early initcall */
2655
2656#ifdef CONFIG_PROC_FS
2657void socket_seq_show(struct seq_file *seq)
2658{
2659        int cpu;
2660        int counter = 0;
2661
2662        for_each_possible_cpu(cpu)
2663            counter += per_cpu(sockets_in_use, cpu);
2664
2665        /* It can be negative, by the way. 8) */
2666        if (counter < 0)
2667                counter = 0;
2668
2669        seq_printf(seq, "sockets: used %d\n", counter);
2670}
2671#endif                          /* CONFIG_PROC_FS */
2672
2673#ifdef CONFIG_COMPAT
2674static int do_siocgstamp(struct net *net, struct socket *sock,
2675                         unsigned int cmd, void __user *up)
2676{
2677        mm_segment_t old_fs = get_fs();
2678        struct timeval ktv;
2679        int err;
2680
2681        set_fs(KERNEL_DS);
2682        err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2683        set_fs(old_fs);
2684        if (!err)
2685                err = compat_put_timeval(&ktv, up);
2686
2687        return err;
2688}
2689
2690static int do_siocgstampns(struct net *net, struct socket *sock,
2691                           unsigned int cmd, void __user *up)
2692{
2693        mm_segment_t old_fs = get_fs();
2694        struct timespec kts;
2695        int err;
2696
2697        set_fs(KERNEL_DS);
2698        err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2699        set_fs(old_fs);
2700        if (!err)
2701                err = compat_put_timespec(&kts, up);
2702
2703        return err;
2704}
2705
2706static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2707{
2708        struct ifreq __user *uifr;
2709        int err;
2710
2711        uifr = compat_alloc_user_space(sizeof(struct ifreq));
2712        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2713                return -EFAULT;
2714
2715        err = dev_ioctl(net, SIOCGIFNAME, uifr);
2716        if (err)
2717                return err;
2718
2719        if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2720                return -EFAULT;
2721
2722        return 0;
2723}
2724
2725static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2726{
2727        struct compat_ifconf ifc32;
2728        struct ifconf ifc;
2729        struct ifconf __user *uifc;
2730        struct compat_ifreq __user *ifr32;
2731        struct ifreq __user *ifr;
2732        unsigned int i, j;
2733        int err;
2734
2735        if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2736                return -EFAULT;
2737
2738        memset(&ifc, 0, sizeof(ifc));
2739        if (ifc32.ifcbuf == 0) {
2740                ifc32.ifc_len = 0;
2741                ifc.ifc_len = 0;
2742                ifc.ifc_req = NULL;
2743                uifc = compat_alloc_user_space(sizeof(struct ifconf));
2744        } else {
2745                size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2746                        sizeof(struct ifreq);
2747                uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2748                ifc.ifc_len = len;
2749                ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2750                ifr32 = compat_ptr(ifc32.ifcbuf);
2751                for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2752                        if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2753                                return -EFAULT;
2754                        ifr++;
2755                        ifr32++;
2756                }
2757        }
2758        if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2759                return -EFAULT;
2760
2761        err = dev_ioctl(net, SIOCGIFCONF, uifc);
2762        if (err)
2763                return err;
2764
2765        if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2766                return -EFAULT;
2767
2768        ifr = ifc.ifc_req;
2769        ifr32 = compat_ptr(ifc32.ifcbuf);
2770        for (i = 0, j = 0;
2771             i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2772             i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2773                if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2774                        return -EFAULT;
2775                ifr32++;
2776                ifr++;
2777        }
2778
2779        if (ifc32.ifcbuf == 0) {
2780                /* Translate from 64-bit structure multiple to
2781                 * a 32-bit one.
2782                 */
2783                i = ifc.ifc_len;
2784                i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2785                ifc32.ifc_len = i;
2786        } else {
2787                ifc32.ifc_len = i;
2788        }
2789        if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2790                return -EFAULT;
2791
2792        return 0;
2793}
2794
2795static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2796{
2797        struct compat_ethtool_rxnfc __user *compat_rxnfc;
2798        bool convert_in = false, convert_out = false;
2799        size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2800        struct ethtool_rxnfc __user *rxnfc;
2801        struct ifreq __user *ifr;
2802        u32 rule_cnt = 0, actual_rule_cnt;
2803        u32 ethcmd;
2804        u32 data;
2805        int ret;
2806
2807        if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2808                return -EFAULT;
2809
2810        compat_rxnfc = compat_ptr(data);
2811
2812        if (get_user(ethcmd, &compat_rxnfc->cmd))
2813                return -EFAULT;
2814
2815        /* Most ethtool structures are defined without padding.
2816         * Unfortunately struct ethtool_rxnfc is an exception.
2817         */
2818        switch (ethcmd) {
2819        default:
2820                break;
2821        case ETHTOOL_GRXCLSRLALL:
2822                /* Buffer size is variable */
2823                if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2824                        return -EFAULT;
2825                if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2826                        return -ENOMEM;
2827                buf_size += rule_cnt * sizeof(u32);
2828                /* fall through */
2829        case ETHTOOL_GRXRINGS:
2830        case ETHTOOL_GRXCLSRLCNT:
2831        case ETHTOOL_GRXCLSRULE:
2832        case ETHTOOL_SRXCLSRLINS:
2833                convert_out = true;
2834                /* fall through */
2835        case ETHTOOL_SRXCLSRLDEL:
2836                buf_size += sizeof(struct ethtool_rxnfc);
2837                convert_in = true;
2838                break;
2839        }
2840
2841        ifr = compat_alloc_user_space(buf_size);
2842        rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2843
2844        if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2845                return -EFAULT;
2846
2847        if (put_user(convert_in ? rxnfc : compat_ptr(data),
2848                     &ifr->ifr_ifru.ifru_data))
2849                return -EFAULT;
2850
2851        if (convert_in) {
2852                /* We expect there to be holes between fs.m_ext and
2853                 * fs.ring_cookie and at the end of fs, but nowhere else.
2854                 */
2855                BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2856                             sizeof(compat_rxnfc->fs.m_ext) !=
2857                             offsetof(struct ethtool_rxnfc, fs.m_ext) +
2858                             sizeof(rxnfc->fs.m_ext));
2859                BUILD_BUG_ON(
2860                        offsetof(struct compat_ethtool_rxnfc, fs.location) -
2861                        offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2862                        offsetof(struct ethtool_rxnfc, fs.location) -
2863                        offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2864
2865                if (copy_in_user(rxnfc, compat_rxnfc,
2866                                 (void __user *)(&rxnfc->fs.m_ext + 1) -
2867                                 (void __user *)rxnfc) ||
2868                    copy_in_user(&rxnfc->fs.ring_cookie,
2869                                 &compat_rxnfc->fs.ring_cookie,
2870                                 (void __user *)(&rxnfc->fs.location + 1) -
2871                                 (void __user *)&rxnfc->fs.ring_cookie) ||
2872                    copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2873                                 sizeof(rxnfc->rule_cnt)))
2874                        return -EFAULT;
2875        }
2876
2877        ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2878        if (ret)
2879                return ret;
2880
2881        if (convert_out) {
2882                if (copy_in_user(compat_rxnfc, rxnfc,
2883                                 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2884                                 (const void __user *)rxnfc) ||
2885                    copy_in_user(&compat_rxnfc->fs.ring_cookie,
2886                                 &rxnfc->fs.ring_cookie,
2887                                 (const void __user *)(&rxnfc->fs.location + 1) -
2888                                 (const void __user *)&rxnfc->fs.ring_cookie) ||
2889                    copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2890                                 sizeof(rxnfc->rule_cnt)))
2891                        return -EFAULT;
2892
2893                if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2894                        /* As an optimisation, we only copy the actual
2895                         * number of rules that the underlying
2896                         * function returned.  Since Mallory might
2897                         * change the rule count in user memory, we
2898                         * check that it is less than the rule count
2899                         * originally given (as the user buffer size),
2900                         * which has been range-checked.
2901                         */
2902                        if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2903                                return -EFAULT;
2904                        if (actual_rule_cnt < rule_cnt)
2905                                rule_cnt = actual_rule_cnt;
2906                        if (copy_in_user(&compat_rxnfc->rule_locs[0],
2907                                         &rxnfc->rule_locs[0],
2908                                         rule_cnt * sizeof(u32)))
2909                                return -EFAULT;
2910                }
2911        }
2912
2913        return 0;
2914}
2915
2916static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2917{
2918        void __user *uptr;
2919        compat_uptr_t uptr32;
2920        struct ifreq __user *uifr;
2921
2922        uifr = compat_alloc_user_space(sizeof(*uifr));
2923        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2924                return -EFAULT;
2925
2926        if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2927                return -EFAULT;
2928
2929        uptr = compat_ptr(uptr32);
2930
2931        if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2932                return -EFAULT;
2933
2934        return dev_ioctl(net, SIOCWANDEV, uifr);
2935}
2936
2937static int bond_ioctl(struct net *net, unsigned int cmd,
2938                         struct compat_ifreq __user *ifr32)
2939{
2940        struct ifreq kifr;
2941        struct ifreq __user *uifr;
2942        mm_segment_t old_fs;
2943        int err;
2944        u32 data;
2945        void __user *datap;
2946
2947        switch (cmd) {
2948        case SIOCBONDENSLAVE:
2949        case SIOCBONDRELEASE:
2950        case SIOCBONDSETHWADDR:
2951        case SIOCBONDCHANGEACTIVE:
2952                if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2953                        return -EFAULT;
2954
2955                old_fs = get_fs();
2956                set_fs(KERNEL_DS);
2957                err = dev_ioctl(net, cmd,
2958                                (struct ifreq __user __force *) &kifr);
2959                set_fs(old_fs);
2960
2961                return err;
2962        case SIOCBONDSLAVEINFOQUERY:
2963        case SIOCBONDINFOQUERY:
2964                uifr = compat_alloc_user_space(sizeof(*uifr));
2965                if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2966                        return -EFAULT;
2967
2968                if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2969                        return -EFAULT;
2970
2971                datap = compat_ptr(data);
2972                if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2973                        return -EFAULT;
2974
2975                return dev_ioctl(net, cmd, uifr);
2976        default:
2977                return -ENOIOCTLCMD;
2978        }
2979}
2980
2981static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2982                                 struct compat_ifreq __user *u_ifreq32)
2983{
2984        struct ifreq __user *u_ifreq64;
2985        char tmp_buf[IFNAMSIZ];
2986        void __user *data64;
2987        u32 data32;
2988
2989        if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2990                           IFNAMSIZ))
2991                return -EFAULT;
2992        if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2993                return -EFAULT;
2994        data64 = compat_ptr(data32);
2995
2996        u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2997
2998        /* Don't check these user accesses, just let that get trapped
2999         * in the ioctl handler instead.
3000         */
3001        if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3002                         IFNAMSIZ))
3003                return -EFAULT;
3004        if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3005                return -EFAULT;
3006
3007        return dev_ioctl(net, cmd, u_ifreq64);
3008}
3009
3010static int dev_ifsioc(struct net *net, struct socket *sock,
3011                         unsigned int cmd, struct compat_ifreq __user *uifr32)
3012{
3013        struct ifreq __user *uifr;
3014        int err;
3015
3016        uifr = compat_alloc_user_space(sizeof(*uifr));
3017        if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3018                return -EFAULT;
3019
3020        err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3021
3022        if (!err) {
3023                switch (cmd) {
3024                case SIOCGIFFLAGS:
3025                case SIOCGIFMETRIC:
3026                case SIOCGIFMTU:
3027                case SIOCGIFMEM:
3028                case SIOCGIFHWADDR:
3029                case SIOCGIFINDEX:
3030                case SIOCGIFADDR:
3031                case SIOCGIFBRDADDR:
3032                case SIOCGIFDSTADDR:
3033                case SIOCGIFNETMASK:
3034                case SIOCGIFPFLAGS:
3035                case SIOCGIFTXQLEN:
3036                case SIOCGMIIPHY:
3037                case SIOCGMIIREG:
3038                        if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3039                                err = -EFAULT;
3040                        break;
3041                }
3042        }
3043        return err;
3044}
3045
3046static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3047                        struct compat_ifreq __user *uifr32)
3048{
3049        struct ifreq ifr;
3050        struct compat_ifmap __user *uifmap32;
3051        mm_segment_t old_fs;
3052        int err;
3053
3054        uifmap32 = &uifr32->ifr_ifru.ifru_map;
3055        err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3056        err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3057        err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3058        err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3059        err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3060        err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3061        err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3062        if (err)
3063                return -EFAULT;
3064
3065        old_fs = get_fs();
3066        set_fs(KERNEL_DS);
3067        err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
3068        set_fs(old_fs);
3069
3070        if (cmd == SIOCGIFMAP && !err) {
3071                err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3072                err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3073                err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3074                err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3075                err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3076                err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3077                err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3078                if (err)
3079                        err = -EFAULT;
3080        }
3081        return err;
3082}
3083
3084static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3085{
3086        void __user *uptr;
3087        compat_uptr_t uptr32;
3088        struct ifreq __user *uifr;
3089
3090        uifr = compat_alloc_user_space(sizeof(*uifr));
3091        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3092                return -EFAULT;
3093
3094        if (get_user(uptr32, &uifr32->ifr_data))
3095                return -EFAULT;
3096
3097        uptr = compat_ptr(uptr32);
3098
3099        if (put_user(uptr, &uifr->ifr_data))
3100                return -EFAULT;
3101
3102        return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3103}
3104
3105struct rtentry32 {
3106        u32             rt_pad1;
3107        struct sockaddr rt_dst;         /* target address               */
3108        struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3109        struct sockaddr rt_genmask;     /* target network mask (IP)     */
3110        unsigned short  rt_flags;
3111        short           rt_pad2;
3112        u32             rt_pad3;
3113        unsigned char   rt_tos;
3114        unsigned char   rt_class;
3115        short           rt_pad4;
3116        short           rt_metric;      /* +1 for binary compatibility! */
3117        /* char * */ u32 rt_dev;        /* forcing the device at add    */
3118        u32             rt_mtu;         /* per route MTU/Window         */
3119        u32             rt_window;      /* Window clamping              */
3120        unsigned short  rt_irtt;        /* Initial RTT                  */
3121};
3122
3123struct in6_rtmsg32 {
3124        struct in6_addr         rtmsg_dst;
3125        struct in6_addr         rtmsg_src;
3126        struct in6_addr         rtmsg_gateway;
3127        u32                     rtmsg_type;
3128        u16                     rtmsg_dst_len;
3129        u16                     rtmsg_src_len;
3130        u32                     rtmsg_metric;
3131        u32                     rtmsg_info;
3132        u32                     rtmsg_flags;
3133        s32                     rtmsg_ifindex;
3134};
3135
3136static int routing_ioctl(struct net *net, struct socket *sock,
3137                         unsigned int cmd, void __user *argp)
3138{
3139        int ret;
3140        void *r = NULL;
3141        struct in6_rtmsg r6;
3142        struct rtentry r4;
3143        char devname[16];
3144        u32 rtdev;
3145        mm_segment_t old_fs = get_fs();
3146
3147        if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3148                struct in6_rtmsg32 __user *ur6 = argp;
3149                ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3150                        3 * sizeof(struct in6_addr));
3151                ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3152                ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3153                ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3154                ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3155                ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3156                ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3157                ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3158
3159                r = (void *) &r6;
3160        } else { /* ipv4 */
3161                struct rtentry32 __user *ur4 = argp;
3162                ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3163                                        3 * sizeof(struct sockaddr));
3164                ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3165                ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3166                ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3167                ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3168                ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3169                ret |= __get_user(rtdev, &(ur4->rt_dev));
3170                if (rtdev) {
3171                        ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3172                        r4.rt_dev = (char __user __force *)devname;
3173                        devname[15] = 0;
3174                } else
3175                        r4.rt_dev = NULL;
3176
3177                r = (void *) &r4;
3178        }
3179
3180        if (ret) {
3181                ret = -EFAULT;
3182                goto out;
3183        }
3184
3185        set_fs(KERNEL_DS);
3186        ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3187        set_fs(old_fs);
3188
3189out:
3190        return ret;
3191}
3192
3193/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3194 * for some operations; this forces use of the newer bridge-utils that
3195 * use compatible ioctls
3196 */
3197static int old_bridge_ioctl(compat_ulong_t __user *argp)
3198{
3199        compat_ulong_t tmp;
3200
3201        if (get_user(tmp, argp))
3202                return -EFAULT;
3203        if (tmp == BRCTL_GET_VERSION)
3204                return BRCTL_VERSION + 1;
3205        return -EINVAL;
3206}
3207
3208static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3209                         unsigned int cmd, unsigned long arg)
3210{
3211        void __user *argp = compat_ptr(arg);
3212        struct sock *sk = sock->sk;
3213        struct net *net = sock_net(sk);
3214
3215        if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3216                return siocdevprivate_ioctl(net, cmd, argp);
3217
3218        switch (cmd) {
3219        case SIOCSIFBR:
3220        case SIOCGIFBR:
3221                return old_bridge_ioctl(argp);
3222        case SIOCGIFNAME:
3223                return dev_ifname32(net, argp);
3224        case SIOCGIFCONF:
3225                return dev_ifconf(net, argp);
3226        case SIOCETHTOOL:
3227                return ethtool_ioctl(net, argp);
3228        case SIOCWANDEV:
3229                return compat_siocwandev(net, argp);
3230        case SIOCGIFMAP:
3231        case SIOCSIFMAP:
3232                return compat_sioc_ifmap(net, cmd, argp);
3233        case SIOCBONDENSLAVE:
3234        case SIOCBONDRELEASE:
3235        case SIOCBONDSETHWADDR:
3236        case SIOCBONDSLAVEINFOQUERY:
3237        case SIOCBONDINFOQUERY:
3238        case SIOCBONDCHANGEACTIVE:
3239                return bond_ioctl(net, cmd, argp);
3240        case SIOCADDRT:
3241        case SIOCDELRT:
3242                return routing_ioctl(net, sock, cmd, argp);
3243        case SIOCGSTAMP:
3244                return do_siocgstamp(net, sock, cmd, argp);
3245        case SIOCGSTAMPNS:
3246                return do_siocgstampns(net, sock, cmd, argp);
3247        case SIOCSHWTSTAMP:
3248                return compat_siocshwtstamp(net, argp);
3249
3250        case FIOSETOWN:
3251        case SIOCSPGRP:
3252        case FIOGETOWN:
3253        case SIOCGPGRP:
3254        case SIOCBRADDBR:
3255        case SIOCBRDELBR:
3256        case SIOCGIFVLAN:
3257        case SIOCSIFVLAN:
3258        case SIOCADDDLCI:
3259        case SIOCDELDLCI:
3260                return sock_ioctl(file, cmd, arg);
3261
3262        case SIOCGIFFLAGS:
3263        case SIOCSIFFLAGS:
3264        case SIOCGIFMETRIC:
3265        case SIOCSIFMETRIC:
3266        case SIOCGIFMTU:
3267        case SIOCSIFMTU:
3268        case SIOCGIFMEM:
3269        case SIOCSIFMEM:
3270        case SIOCGIFHWADDR:
3271        case SIOCSIFHWADDR:
3272        case SIOCADDMULTI:
3273        case SIOCDELMULTI:
3274        case SIOCGIFINDEX:
3275        case SIOCGIFADDR:
3276        case SIOCSIFADDR:
3277        case SIOCSIFHWBROADCAST:
3278        case SIOCDIFADDR:
3279        case SIOCGIFBRDADDR:
3280        case SIOCSIFBRDADDR:
3281        case SIOCGIFDSTADDR:
3282        case SIOCSIFDSTADDR:
3283        case SIOCGIFNETMASK:
3284        case SIOCSIFNETMASK:
3285        case SIOCSIFPFLAGS:
3286        case SIOCGIFPFLAGS:
3287        case SIOCGIFTXQLEN:
3288        case SIOCSIFTXQLEN:
3289        case SIOCBRADDIF:
3290        case SIOCBRDELIF:
3291        case SIOCSIFNAME:
3292        case SIOCGMIIPHY:
3293        case SIOCGMIIREG:
3294        case SIOCSMIIREG:
3295                return dev_ifsioc(net, sock, cmd, argp);
3296
3297        case SIOCSARP:
3298        case SIOCGARP:
3299        case SIOCDARP:
3300        case SIOCATMARK:
3301                return sock_do_ioctl(net, sock, cmd, arg);
3302        }
3303
3304        return -ENOIOCTLCMD;
3305}
3306
3307static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3308                              unsigned long arg)
3309{
3310        struct socket *sock = file->private_data;
3311        int ret = -ENOIOCTLCMD;
3312        struct sock *sk;
3313        struct net *net;
3314
3315        sk = sock->sk;
3316        net = sock_net(sk);
3317
3318        if (sock->ops->compat_ioctl)
3319                ret = sock->ops->compat_ioctl(sock, cmd, arg);
3320
3321        if (ret == -ENOIOCTLCMD &&
3322            (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3323                ret = compat_wext_handle_ioctl(net, cmd, arg);
3324
3325        if (ret == -ENOIOCTLCMD)
3326                ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3327
3328        return ret;
3329}
3330#endif
3331
3332int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3333{
3334        return sock->ops->bind(sock, addr, addrlen);
3335}
3336EXPORT_SYMBOL(kernel_bind);
3337
3338int kernel_listen(struct socket *sock, int backlog)
3339{
3340        return sock->ops->listen(sock, backlog);
3341}
3342EXPORT_SYMBOL(kernel_listen);
3343
3344int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3345{
3346        struct sock *sk = sock->sk;
3347        int err;
3348
3349        err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3350                               newsock);
3351        if (err < 0)
3352                goto done;
3353
3354        err = sock->ops->accept(sock, *newsock, flags);
3355        if (err < 0) {
3356                sock_release(*newsock);
3357                *newsock = NULL;
3358                goto done;
3359        }
3360
3361        (*newsock)->ops = sock->ops;
3362        __module_get((*newsock)->ops->owner);
3363
3364done:
3365        return err;
3366}
3367EXPORT_SYMBOL(kernel_accept);
3368
3369int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3370                   int flags)
3371{
3372        return sock->ops->connect(sock, addr, addrlen, flags);
3373}
3374EXPORT_SYMBOL(kernel_connect);
3375
3376int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3377                         int *addrlen)
3378{
3379        return sock->ops->getname(sock, addr, addrlen, 0);
3380}
3381EXPORT_SYMBOL(kernel_getsockname);
3382
3383int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3384                         int *addrlen)
3385{
3386        return sock->ops->getname(sock, addr, addrlen, 1);
3387}
3388EXPORT_SYMBOL(kernel_getpeername);
3389
3390int kernel_getsockopt(struct socket *sock, int level, int optname,
3391                        char *optval, int *optlen)
3392{
3393        mm_segment_t oldfs = get_fs();
3394        char __user *uoptval;
3395        int __user *uoptlen;
3396        int err;
3397
3398        uoptval = (char __user __force *) optval;
3399        uoptlen = (int __user __force *) optlen;
3400
3401        set_fs(KERNEL_DS);
3402        if (level == SOL_SOCKET)
3403                err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3404        else
3405                err = sock->ops->getsockopt(sock, level, optname, uoptval,
3406                                            uoptlen);
3407        set_fs(oldfs);
3408        return err;
3409}
3410EXPORT_SYMBOL(kernel_getsockopt);
3411
3412int kernel_setsockopt(struct socket *sock, int level, int optname,
3413                        char *optval, unsigned int optlen)
3414{
3415        mm_segment_t oldfs = get_fs();
3416        char __user *uoptval;
3417        int err;
3418
3419        uoptval = (char __user __force *) optval;
3420
3421        set_fs(KERNEL_DS);
3422        if (level == SOL_SOCKET)
3423                err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3424        else
3425                err = sock->ops->setsockopt(sock, level, optname, uoptval,
3426                                            optlen);
3427        set_fs(oldfs);
3428        return err;
3429}
3430EXPORT_SYMBOL(kernel_setsockopt);
3431
3432int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3433                    size_t size, int flags)
3434{
3435        if (sock->ops->sendpage)
3436                return sock->ops->sendpage(sock, page, offset, size, flags);
3437
3438        return sock_no_sendpage(sock, page, offset, size, flags);
3439}
3440EXPORT_SYMBOL(kernel_sendpage);
3441
3442int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3443{
3444        mm_segment_t oldfs = get_fs();
3445        int err;
3446
3447        set_fs(KERNEL_DS);
3448        err = sock->ops->ioctl(sock, cmd, arg);
3449        set_fs(oldfs);
3450
3451        return err;
3452}
3453EXPORT_SYMBOL(kernel_sock_ioctl);
3454
3455int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3456{
3457        return sock->ops->shutdown(sock, how);
3458}
3459EXPORT_SYMBOL(kernel_sock_shutdown);
3460
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.