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#include <net/busy_poll.h>
 108
 109#ifdef CONFIG_NET_RX_BUSY_POLL
 110unsigned int sysctl_net_busy_read __read_mostly;
 111unsigned int sysctl_net_busy_poll __read_mostly;
 112#endif
 113
 114static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
 115static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 116                         unsigned long nr_segs, loff_t pos);
 117static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 118                          unsigned long nr_segs, loff_t pos);
 119static int sock_mmap(struct file *file, struct vm_area_struct *vma);
 120
 121static int sock_close(struct inode *inode, struct file *file);
 122static unsigned int sock_poll(struct file *file,
 123                              struct poll_table_struct *wait);
 124static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 125#ifdef CONFIG_COMPAT
 126static long compat_sock_ioctl(struct file *file,
 127                              unsigned int cmd, unsigned long arg);
 128#endif
 129static int sock_fasync(int fd, struct file *filp, int on);
 130static ssize_t sock_sendpage(struct file *file, struct page *page,
 131                             int offset, size_t size, loff_t *ppos, int more);
 132static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 133                                struct pipe_inode_info *pipe, size_t len,
 134                                unsigned int flags);
 135
 136/*
 137 *      Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
 138 *      in the operation structures but are done directly via the socketcall() multiplexor.
 139 */
 140
 141static const struct file_operations socket_file_ops = {
 142        .owner =        THIS_MODULE,
 143        .llseek =       no_llseek,
 144        .aio_read =     sock_aio_read,
 145        .aio_write =    sock_aio_write,
 146        .poll =         sock_poll,
 147        .unlocked_ioctl = sock_ioctl,
 148#ifdef CONFIG_COMPAT
 149        .compat_ioctl = compat_sock_ioctl,
 150#endif
 151        .mmap =         sock_mmap,
 152        .open =         sock_no_open,   /* special open code to disallow open via /proc */
 153        .release =      sock_close,
 154        .fasync =       sock_fasync,
 155        .sendpage =     sock_sendpage,
 156        .splice_write = generic_splice_sendpage,
 157        .splice_read =  sock_splice_read,
 158};
 159
 160/*
 161 *      The protocol list. Each protocol is registered in here.
 162 */
 163
 164static DEFINE_SPINLOCK(net_family_lock);
 165static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
 166
 167/*
 168 *      Statistics counters of the socket lists
 169 */
 170
 171static DEFINE_PER_CPU(int, sockets_in_use);
 172
 173/*
 174 * Support routines.
 175 * Move socket addresses back and forth across the kernel/user
 176 * divide and look after the messy bits.
 177 */
 178
 179/**
 180 *      move_addr_to_kernel     -       copy a socket address into kernel space
 181 *      @uaddr: Address in user space
 182 *      @kaddr: Address in kernel space
 183 *      @ulen: Length in user space
 184 *
 185 *      The address is copied into kernel space. If the provided address is
 186 *      too long an error code of -EINVAL is returned. If the copy gives
 187 *      invalid addresses -EFAULT is returned. On a success 0 is returned.
 188 */
 189
 190int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
 191{
 192        if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
 193                return -EINVAL;
 194        if (ulen == 0)
 195                return 0;
 196        if (copy_from_user(kaddr, uaddr, ulen))
 197                return -EFAULT;
 198        return audit_sockaddr(ulen, kaddr);
 199}
 200
 201/**
 202 *      move_addr_to_user       -       copy an address to user space
 203 *      @kaddr: kernel space address
 204 *      @klen: length of address in kernel
 205 *      @uaddr: user space address
 206 *      @ulen: pointer to user length field
 207 *
 208 *      The value pointed to by ulen on entry is the buffer length available.
 209 *      This is overwritten with the buffer space used. -EINVAL is returned
 210 *      if an overlong buffer is specified or a negative buffer size. -EFAULT
 211 *      is returned if either the buffer or the length field are not
 212 *      accessible.
 213 *      After copying the data up to the limit the user specifies, the true
 214 *      length of the data is written over the length limit the user
 215 *      specified. Zero is returned for a success.
 216 */
 217
 218static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
 219                             void __user *uaddr, int __user *ulen)
 220{
 221        int err;
 222        int len;
 223
 224        err = get_user(len, ulen);
 225        if (err)
 226                return err;
 227        if (len > klen)
 228                len = klen;
 229        if (len < 0 || len > sizeof(struct sockaddr_storage))
 230                return -EINVAL;
 231        if (len) {
 232                if (audit_sockaddr(klen, kaddr))
 233                        return -ENOMEM;
 234                if (copy_to_user(uaddr, kaddr, len))
 235                        return -EFAULT;
 236        }
 237        /*
 238         *      "fromlen shall refer to the value before truncation.."
 239         *                      1003.1g
 240         */
 241        return __put_user(klen, ulen);
 242}
 243
 244static struct kmem_cache *sock_inode_cachep __read_mostly;
 245
 246static struct inode *sock_alloc_inode(struct super_block *sb)
 247{
 248        struct socket_alloc *ei;
 249        struct socket_wq *wq;
 250
 251        ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
 252        if (!ei)
 253                return NULL;
 254        wq = kmalloc(sizeof(*wq), GFP_KERNEL);
 255        if (!wq) {
 256                kmem_cache_free(sock_inode_cachep, ei);
 257                return NULL;
 258        }
 259        init_waitqueue_head(&wq->wait);
 260        wq->fasync_list = NULL;
 261        RCU_INIT_POINTER(ei->socket.wq, wq);
 262
 263        ei->socket.state = SS_UNCONNECTED;
 264        ei->socket.flags = 0;
 265        ei->socket.ops = NULL;
 266        ei->socket.sk = NULL;
 267        ei->socket.file = NULL;
 268
 269        return &ei->vfs_inode;
 270}
 271
 272static void sock_destroy_inode(struct inode *inode)
 273{
 274        struct socket_alloc *ei;
 275        struct socket_wq *wq;
 276
 277        ei = container_of(inode, struct socket_alloc, vfs_inode);
 278        wq = rcu_dereference_protected(ei->socket.wq, 1);
 279        kfree_rcu(wq, rcu);
 280        kmem_cache_free(sock_inode_cachep, ei);
 281}
 282
 283static void init_once(void *foo)
 284{
 285        struct socket_alloc *ei = (struct socket_alloc *)foo;
 286
 287        inode_init_once(&ei->vfs_inode);
 288}
 289
 290static int init_inodecache(void)
 291{
 292        sock_inode_cachep = kmem_cache_create("sock_inode_cache",
 293                                              sizeof(struct socket_alloc),
 294                                              0,
 295                                              (SLAB_HWCACHE_ALIGN |
 296                                               SLAB_RECLAIM_ACCOUNT |
 297                                               SLAB_MEM_SPREAD),
 298                                              init_once);
 299        if (sock_inode_cachep == NULL)
 300                return -ENOMEM;
 301        return 0;
 302}
 303
 304static const struct super_operations sockfs_ops = {
 305        .alloc_inode    = sock_alloc_inode,
 306        .destroy_inode  = sock_destroy_inode,
 307        .statfs         = simple_statfs,
 308};
 309
 310/*
 311 * sockfs_dname() is called from d_path().
 312 */
 313static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
 314{
 315        return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
 316                                dentry->d_inode->i_ino);
 317}
 318
 319static const struct dentry_operations sockfs_dentry_operations = {
 320        .d_dname  = sockfs_dname,
 321};
 322
 323static struct dentry *sockfs_mount(struct file_system_type *fs_type,
 324                         int flags, const char *dev_name, void *data)
 325{
 326        return mount_pseudo(fs_type, "socket:", &sockfs_ops,
 327                &sockfs_dentry_operations, SOCKFS_MAGIC);
 328}
 329
 330static struct vfsmount *sock_mnt __read_mostly;
 331
 332static struct file_system_type sock_fs_type = {
 333        .name =         "sockfs",
 334        .mount =        sockfs_mount,
 335        .kill_sb =      kill_anon_super,
 336};
 337
 338/*
 339 *      Obtains the first available file descriptor and sets it up for use.
 340 *
 341 *      These functions create file structures and maps them to fd space
 342 *      of the current process. On success it returns file descriptor
 343 *      and file struct implicitly stored in sock->file.
 344 *      Note that another thread may close file descriptor before we return
 345 *      from this function. We use the fact that now we do not refer
 346 *      to socket after mapping. If one day we will need it, this
 347 *      function will increment ref. count on file by 1.
 348 *
 349 *      In any case returned fd MAY BE not valid!
 350 *      This race condition is unavoidable
 351 *      with shared fd spaces, we cannot solve it inside kernel,
 352 *      but we take care of internal coherence yet.
 353 */
 354
 355struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
 356{
 357        struct qstr name = { .name = "" };
 358        struct path path;
 359        struct file *file;
 360
 361        if (dname) {
 362                name.name = dname;
 363                name.len = strlen(name.name);
 364        } else if (sock->sk) {
 365                name.name = sock->sk->sk_prot_creator->name;
 366                name.len = strlen(name.name);
 367        }
 368        path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
 369        if (unlikely(!path.dentry))
 370                return ERR_PTR(-ENOMEM);
 371        path.mnt = mntget(sock_mnt);
 372
 373        d_instantiate(path.dentry, SOCK_INODE(sock));
 374        SOCK_INODE(sock)->i_fop = &socket_file_ops;
 375
 376        file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
 377                  &socket_file_ops);
 378        if (unlikely(IS_ERR(file))) {
 379                /* drop dentry, keep inode */
 380                ihold(path.dentry->d_inode);
 381                path_put(&path);
 382                return file;
 383        }
 384
 385        sock->file = file;
 386        file->f_flags = O_RDWR | (flags & O_NONBLOCK);
 387        file->private_data = sock;
 388        return file;
 389}
 390EXPORT_SYMBOL(sock_alloc_file);
 391
 392static int sock_map_fd(struct socket *sock, int flags)
 393{
 394        struct file *newfile;
 395        int fd = get_unused_fd_flags(flags);
 396        if (unlikely(fd < 0))
 397                return fd;
 398
 399        newfile = sock_alloc_file(sock, flags, NULL);
 400        if (likely(!IS_ERR(newfile))) {
 401                fd_install(fd, newfile);
 402                return fd;
 403        }
 404
 405        put_unused_fd(fd);
 406        return PTR_ERR(newfile);
 407}
 408
 409struct socket *sock_from_file(struct file *file, int *err)
 410{
 411        if (file->f_op == &socket_file_ops)
 412                return file->private_data;      /* set in sock_map_fd */
 413
 414        *err = -ENOTSOCK;
 415        return NULL;
 416}
 417EXPORT_SYMBOL(sock_from_file);
 418
 419/**
 420 *      sockfd_lookup - Go from a file number to its socket slot
 421 *      @fd: file handle
 422 *      @err: pointer to an error code return
 423 *
 424 *      The file handle passed in is locked and the socket it is bound
 425 *      too is returned. If an error occurs the err pointer is overwritten
 426 *      with a negative errno code and NULL is returned. The function checks
 427 *      for both invalid handles and passing a handle which is not a socket.
 428 *
 429 *      On a success the socket object pointer is returned.
 430 */
 431
 432struct socket *sockfd_lookup(int fd, int *err)
 433{
 434        struct file *file;
 435        struct socket *sock;
 436
 437        file = fget(fd);
 438        if (!file) {
 439                *err = -EBADF;
 440                return NULL;
 441        }
 442
 443        sock = sock_from_file(file, err);
 444        if (!sock)
 445                fput(file);
 446        return sock;
 447}
 448EXPORT_SYMBOL(sockfd_lookup);
 449
 450static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 451{
 452        struct file *file;
 453        struct socket *sock;
 454
 455        *err = -EBADF;
 456        file = fget_light(fd, fput_needed);
 457        if (file) {
 458                sock = sock_from_file(file, err);
 459                if (sock)
 460                        return sock;
 461                fput_light(file, *fput_needed);
 462        }
 463        return NULL;
 464}
 465
 466#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
 467#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
 468#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
 469static ssize_t sockfs_getxattr(struct dentry *dentry,
 470                               const char *name, void *value, size_t size)
 471{
 472        const char *proto_name;
 473        size_t proto_size;
 474        int error;
 475
 476        error = -ENODATA;
 477        if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
 478                proto_name = dentry->d_name.name;
 479                proto_size = strlen(proto_name);
 480
 481                if (value) {
 482                        error = -ERANGE;
 483                        if (proto_size + 1 > size)
 484                                goto out;
 485
 486                        strncpy(value, proto_name, proto_size + 1);
 487                }
 488                error = proto_size + 1;
 489        }
 490
 491out:
 492        return error;
 493}
 494
 495static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
 496                                size_t size)
 497{
 498        ssize_t len;
 499        ssize_t used = 0;
 500
 501        len = security_inode_listsecurity(dentry->d_inode, buffer, size);
 502        if (len < 0)
 503                return len;
 504        used += len;
 505        if (buffer) {
 506                if (size < used)
 507                        return -ERANGE;
 508                buffer += len;
 509        }
 510
 511        len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
 512        used += len;
 513        if (buffer) {
 514                if (size < used)
 515                        return -ERANGE;
 516                memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
 517                buffer += len;
 518        }
 519
 520        return used;
 521}
 522
 523static const struct inode_operations sockfs_inode_ops = {
 524        .getxattr = sockfs_getxattr,
 525        .listxattr = sockfs_listxattr,
 526};
 527
 528/**
 529 *      sock_alloc      -       allocate a socket
 530 *
 531 *      Allocate a new inode and socket object. The two are bound together
 532 *      and initialised. The socket is then returned. If we are out of inodes
 533 *      NULL is returned.
 534 */
 535
 536static struct socket *sock_alloc(void)
 537{
 538        struct inode *inode;
 539        struct socket *sock;
 540
 541        inode = new_inode_pseudo(sock_mnt->mnt_sb);
 542        if (!inode)
 543                return NULL;
 544
 545        sock = SOCKET_I(inode);
 546
 547        kmemcheck_annotate_bitfield(sock, type);
 548        inode->i_ino = get_next_ino();
 549        inode->i_mode = S_IFSOCK | S_IRWXUGO;
 550        inode->i_uid = current_fsuid();
 551        inode->i_gid = current_fsgid();
 552        inode->i_op = &sockfs_inode_ops;
 553
 554        this_cpu_add(sockets_in_use, 1);
 555        return sock;
 556}
 557
 558/*
 559 *      In theory you can't get an open on this inode, but /proc provides
 560 *      a back door. Remember to keep it shut otherwise you'll let the
 561 *      creepy crawlies in.
 562 */
 563
 564static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
 565{
 566        return -ENXIO;
 567}
 568
 569const struct file_operations bad_sock_fops = {
 570        .owner = THIS_MODULE,
 571        .open = sock_no_open,
 572        .llseek = noop_llseek,
 573};
 574
 575/**
 576 *      sock_release    -       close a socket
 577 *      @sock: socket to close
 578 *
 579 *      The socket is released from the protocol stack if it has a release
 580 *      callback, and the inode is then released if the socket is bound to
 581 *      an inode not a file.
 582 */
 583
 584void sock_release(struct socket *sock)
 585{
 586        if (sock->ops) {
 587                struct module *owner = sock->ops->owner;
 588
 589                sock->ops->release(sock);
 590                sock->ops = NULL;
 591                module_put(owner);
 592        }
 593
 594        if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
 595                printk(KERN_ERR "sock_release: fasync list not empty!\n");
 596
 597        if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
 598                return;
 599
 600        this_cpu_sub(sockets_in_use, 1);
 601        if (!sock->file) {
 602                iput(SOCK_INODE(sock));
 603                return;
 604        }
 605        sock->file = NULL;
 606}
 607EXPORT_SYMBOL(sock_release);
 608
 609void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
 610{
 611        *tx_flags = 0;
 612        if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
 613                *tx_flags |= SKBTX_HW_TSTAMP;
 614        if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
 615                *tx_flags |= SKBTX_SW_TSTAMP;
 616        if (sock_flag(sk, SOCK_WIFI_STATUS))
 617                *tx_flags |= SKBTX_WIFI_STATUS;
 618}
 619EXPORT_SYMBOL(sock_tx_timestamp);
 620
 621static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
 622                                       struct msghdr *msg, size_t size)
 623{
 624        struct sock_iocb *si = kiocb_to_siocb(iocb);
 625
 626        si->sock = sock;
 627        si->scm = NULL;
 628        si->msg = msg;
 629        si->size = size;
 630
 631        return sock->ops->sendmsg(iocb, sock, msg, size);
 632}
 633
 634static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 635                                 struct msghdr *msg, size_t size)
 636{
 637        int err = security_socket_sendmsg(sock, msg, size);
 638
 639        return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
 640}
 641
 642int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 643{
 644        struct kiocb iocb;
 645        struct sock_iocb siocb;
 646        int ret;
 647
 648        init_sync_kiocb(&iocb, NULL);
 649        iocb.private = &siocb;
 650        ret = __sock_sendmsg(&iocb, sock, msg, size);
 651        if (-EIOCBQUEUED == ret)
 652                ret = wait_on_sync_kiocb(&iocb);
 653        return ret;
 654}
 655EXPORT_SYMBOL(sock_sendmsg);
 656
 657static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
 658{
 659        struct kiocb iocb;
 660        struct sock_iocb siocb;
 661        int ret;
 662
 663        init_sync_kiocb(&iocb, NULL);
 664        iocb.private = &siocb;
 665        ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
 666        if (-EIOCBQUEUED == ret)
 667                ret = wait_on_sync_kiocb(&iocb);
 668        return ret;
 669}
 670
 671int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 672                   struct kvec *vec, size_t num, size_t size)
 673{
 674        mm_segment_t oldfs = get_fs();
 675        int result;
 676
 677        set_fs(KERNEL_DS);
 678        /*
 679         * the following is safe, since for compiler definitions of kvec and
 680         * iovec are identical, yielding the same in-core layout and alignment
 681         */
 682        msg->msg_iov = (struct iovec *)vec;
 683        msg->msg_iovlen = num;
 684        result = sock_sendmsg(sock, msg, size);
 685        set_fs(oldfs);
 686        return result;
 687}
 688EXPORT_SYMBOL(kernel_sendmsg);
 689
 690/*
 691 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 692 */
 693void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 694        struct sk_buff *skb)
 695{
 696        int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 697        struct timespec ts[3];
 698        int empty = 1;
 699        struct skb_shared_hwtstamps *shhwtstamps =
 700                skb_hwtstamps(skb);
 701
 702        /* Race occurred between timestamp enabling and packet
 703           receiving.  Fill in the current time for now. */
 704        if (need_software_tstamp && skb->tstamp.tv64 == 0)
 705                __net_timestamp(skb);
 706
 707        if (need_software_tstamp) {
 708                if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 709                        struct timeval tv;
 710                        skb_get_timestamp(skb, &tv);
 711                        put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 712                                 sizeof(tv), &tv);
 713                } else {
 714                        skb_get_timestampns(skb, &ts[0]);
 715                        put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 716                                 sizeof(ts[0]), &ts[0]);
 717                }
 718        }
 719
 720
 721        memset(ts, 0, sizeof(ts));
 722        if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
 723            ktime_to_timespec_cond(skb->tstamp, ts + 0))
 724                empty = 0;
 725        if (shhwtstamps) {
 726                if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 727                    ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
 728                        empty = 0;
 729                if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 730                    ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
 731                        empty = 0;
 732        }
 733        if (!empty)
 734                put_cmsg(msg, SOL_SOCKET,
 735                         SCM_TIMESTAMPING, sizeof(ts), &ts);
 736}
 737EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 738
 739void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
 740        struct sk_buff *skb)
 741{
 742        int ack;
 743
 744        if (!sock_flag(sk, SOCK_WIFI_STATUS))
 745                return;
 746        if (!skb->wifi_acked_valid)
 747                return;
 748
 749        ack = skb->wifi_acked;
 750
 751        put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
 752}
 753EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
 754
 755static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 756                                   struct sk_buff *skb)
 757{
 758        if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 759                put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 760                        sizeof(__u32), &skb->dropcount);
 761}
 762
 763void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 764        struct sk_buff *skb)
 765{
 766        sock_recv_timestamp(msg, sk, skb);
 767        sock_recv_drops(msg, sk, skb);
 768}
 769EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 770
 771static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 772                                       struct msghdr *msg, size_t size, int flags)
 773{
 774        struct sock_iocb *si = kiocb_to_siocb(iocb);
 775
 776        si->sock = sock;
 777        si->scm = NULL;
 778        si->msg = msg;
 779        si->size = size;
 780        si->flags = flags;
 781
 782        return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 783}
 784
 785static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 786                                 struct msghdr *msg, size_t size, int flags)
 787{
 788        int err = security_socket_recvmsg(sock, msg, size, flags);
 789
 790        return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 791}
 792
 793int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 794                 size_t size, int flags)
 795{
 796        struct kiocb iocb;
 797        struct sock_iocb siocb;
 798        int ret;
 799
 800        init_sync_kiocb(&iocb, NULL);
 801        iocb.private = &siocb;
 802        ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 803        if (-EIOCBQUEUED == ret)
 804                ret = wait_on_sync_kiocb(&iocb);
 805        return ret;
 806}
 807EXPORT_SYMBOL(sock_recvmsg);
 808
 809static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 810                              size_t size, int flags)
 811{
 812        struct kiocb iocb;
 813        struct sock_iocb siocb;
 814        int ret;
 815
 816        init_sync_kiocb(&iocb, NULL);
 817        iocb.private = &siocb;
 818        ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 819        if (-EIOCBQUEUED == ret)
 820                ret = wait_on_sync_kiocb(&iocb);
 821        return ret;
 822}
 823
 824/**
 825 * kernel_recvmsg - Receive a message from a socket (kernel space)
 826 * @sock:       The socket to receive the message from
 827 * @msg:        Received message
 828 * @vec:        Input s/g array for message data
 829 * @num:        Size of input s/g array
 830 * @size:       Number of bytes to read
 831 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 832 *
 833 * On return the msg structure contains the scatter/gather array passed in the
 834 * vec argument. The array is modified so that it consists of the unfilled
 835 * portion of the original array.
 836 *
 837 * The returned value is the total number of bytes received, or an error.
 838 */
 839int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 840                   struct kvec *vec, size_t num, size_t size, int flags)
 841{
 842        mm_segment_t oldfs = get_fs();
 843        int result;
 844
 845        set_fs(KERNEL_DS);
 846        /*
 847         * the following is safe, since for compiler definitions of kvec and
 848         * iovec are identical, yielding the same in-core layout and alignment
 849         */
 850        msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 851        result = sock_recvmsg(sock, msg, size, flags);
 852        set_fs(oldfs);
 853        return result;
 854}
 855EXPORT_SYMBOL(kernel_recvmsg);
 856
 857static ssize_t sock_sendpage(struct file *file, struct page *page,
 858                             int offset, size_t size, loff_t *ppos, int more)
 859{
 860        struct socket *sock;
 861        int flags;
 862
 863        sock = file->private_data;
 864
 865        flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 866        /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
 867        flags |= more;
 868
 869        return kernel_sendpage(sock, page, offset, size, flags);
 870}
 871
 872static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 873                                struct pipe_inode_info *pipe, size_t len,
 874                                unsigned int flags)
 875{
 876        struct socket *sock = file->private_data;
 877
 878        if (unlikely(!sock->ops->splice_read))
 879                return -EINVAL;
 880
 881        return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 882}
 883
 884static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 885                                         struct sock_iocb *siocb)
 886{
 887        if (!is_sync_kiocb(iocb))
 888                BUG();
 889
 890        siocb->kiocb = iocb;
 891        iocb->private = siocb;
 892        return siocb;
 893}
 894
 895static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 896                struct file *file, const struct iovec *iov,
 897                unsigned long nr_segs)
 898{
 899        struct socket *sock = file->private_data;
 900        size_t size = 0;
 901        int i;
 902
 903        for (i = 0; i < nr_segs; i++)
 904                size += iov[i].iov_len;
 905
 906        msg->msg_name = NULL;
 907        msg->msg_namelen = 0;
 908        msg->msg_control = NULL;
 909        msg->msg_controllen = 0;
 910        msg->msg_iov = (struct iovec *)iov;
 911        msg->msg_iovlen = nr_segs;
 912        msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 913
 914        return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 915}
 916
 917static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 918                                unsigned long nr_segs, loff_t pos)
 919{
 920        struct sock_iocb siocb, *x;
 921
 922        if (pos != 0)
 923                return -ESPIPE;
 924
 925        if (iocb->ki_nbytes == 0)       /* Match SYS5 behaviour */
 926                return 0;
 927
 928
 929        x = alloc_sock_iocb(iocb, &siocb);
 930        if (!x)
 931                return -ENOMEM;
 932        return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 933}
 934
 935static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 936                        struct file *file, const struct iovec *iov,
 937                        unsigned long nr_segs)
 938{
 939        struct socket *sock = file->private_data;
 940        size_t size = 0;
 941        int i;
 942
 943        for (i = 0; i < nr_segs; i++)
 944                size += iov[i].iov_len;
 945
 946        msg->msg_name = NULL;
 947        msg->msg_namelen = 0;
 948        msg->msg_control = NULL;
 949        msg->msg_controllen = 0;
 950        msg->msg_iov = (struct iovec *)iov;
 951        msg->msg_iovlen = nr_segs;
 952        msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 953        if (sock->type == SOCK_SEQPACKET)
 954                msg->msg_flags |= MSG_EOR;
 955
 956        return __sock_sendmsg(iocb, sock, msg, size);
 957}
 958
 959static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 960                          unsigned long nr_segs, loff_t pos)
 961{
 962        struct sock_iocb siocb, *x;
 963
 964        if (pos != 0)
 965                return -ESPIPE;
 966
 967        x = alloc_sock_iocb(iocb, &siocb);
 968        if (!x)
 969                return -ENOMEM;
 970
 971        return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 972}
 973
 974/*
 975 * Atomic setting of ioctl hooks to avoid race
 976 * with module unload.
 977 */
 978
 979static DEFINE_MUTEX(br_ioctl_mutex);
 980static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 981
 982void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 983{
 984        mutex_lock(&br_ioctl_mutex);
 985        br_ioctl_hook = hook;
 986        mutex_unlock(&br_ioctl_mutex);
 987}
 988EXPORT_SYMBOL(brioctl_set);
 989
 990static DEFINE_MUTEX(vlan_ioctl_mutex);
 991static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 992
 993void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 994{
 995        mutex_lock(&vlan_ioctl_mutex);
 996        vlan_ioctl_hook = hook;
 997        mutex_unlock(&vlan_ioctl_mutex);
 998}
 999EXPORT_SYMBOL(vlan_ioctl_set);
1000
1001static DEFINE_MUTEX(dlci_ioctl_mutex);
1002static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1003
1004void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1005{
1006        mutex_lock(&dlci_ioctl_mutex);
1007        dlci_ioctl_hook = hook;
1008        mutex_unlock(&dlci_ioctl_mutex);
1009}
1010EXPORT_SYMBOL(dlci_ioctl_set);
1011
1012static long sock_do_ioctl(struct net *net, struct socket *sock,
1013                                 unsigned int cmd, unsigned long arg)
1014{
1015        int err;
1016        void __user *argp = (void __user *)arg;
1017
1018        err = sock->ops->ioctl(sock, cmd, arg);
1019
1020        /*
1021         * If this ioctl is unknown try to hand it down
1022         * to the NIC driver.
1023         */
1024        if (err == -ENOIOCTLCMD)
1025                err = dev_ioctl(net, cmd, argp);
1026
1027        return err;
1028}
1029
1030/*
1031 *      With an ioctl, arg may well be a user mode pointer, but we don't know
1032 *      what to do with it - that's up to the protocol still.
1033 */
1034
1035static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1036{
1037        struct socket *sock;
1038        struct sock *sk;
1039        void __user *argp = (void __user *)arg;
1040        int pid, err;
1041        struct net *net;
1042
1043        sock = file->private_data;
1044        sk = sock->sk;
1045        net = sock_net(sk);
1046        if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1047                err = dev_ioctl(net, cmd, argp);
1048        } else
1049#ifdef CONFIG_WEXT_CORE
1050        if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1051                err = dev_ioctl(net, cmd, argp);
1052        } else
1053#endif
1054                switch (cmd) {
1055                case FIOSETOWN:
1056                case SIOCSPGRP:
1057                        err = -EFAULT;
1058                        if (get_user(pid, (int __user *)argp))
1059                                break;
1060                        err = f_setown(sock->file, pid, 1);
1061                        break;
1062                case FIOGETOWN:
1063                case SIOCGPGRP:
1064                        err = put_user(f_getown(sock->file),
1065                                       (int __user *)argp);
1066                        break;
1067                case SIOCGIFBR:
1068                case SIOCSIFBR:
1069                case SIOCBRADDBR:
1070                case SIOCBRDELBR:
1071                        err = -ENOPKG;
1072                        if (!br_ioctl_hook)
1073                                request_module("bridge");
1074
1075                        mutex_lock(&br_ioctl_mutex);
1076                        if (br_ioctl_hook)
1077                                err = br_ioctl_hook(net, cmd, argp);
1078                        mutex_unlock(&br_ioctl_mutex);
1079                        break;
1080                case SIOCGIFVLAN:
1081                case SIOCSIFVLAN:
1082                        err = -ENOPKG;
1083                        if (!vlan_ioctl_hook)
1084                                request_module("8021q");
1085
1086                        mutex_lock(&vlan_ioctl_mutex);
1087                        if (vlan_ioctl_hook)
1088                                err = vlan_ioctl_hook(net, argp);
1089                        mutex_unlock(&vlan_ioctl_mutex);
1090                        break;
1091                case SIOCADDDLCI:
1092                case SIOCDELDLCI:
1093                        err = -ENOPKG;
1094                        if (!dlci_ioctl_hook)
1095                                request_module("dlci");
1096
1097                        mutex_lock(&dlci_ioctl_mutex);
1098                        if (dlci_ioctl_hook)
1099                                err = dlci_ioctl_hook(cmd, argp);
1100                        mutex_unlock(&dlci_ioctl_mutex);
1101                        break;
1102                default:
1103                        err = sock_do_ioctl(net, sock, cmd, arg);
1104                        break;
1105                }
1106        return err;
1107}
1108
1109int sock_create_lite(int family, int type, int protocol, struct socket **res)
1110{
1111        int err;
1112        struct socket *sock = NULL;
1113
1114        err = security_socket_create(family, type, protocol, 1);
1115        if (err)
1116                goto out;
1117
1118        sock = sock_alloc();
1119        if (!sock) {
1120                err = -ENOMEM;
1121                goto out;
1122        }
1123
1124        sock->type = type;
1125        err = security_socket_post_create(sock, family, type, protocol, 1);
1126        if (err)
1127                goto out_release;
1128
1129out:
1130        *res = sock;
1131        return err;
1132out_release:
1133        sock_release(sock);
1134        sock = NULL;
1135        goto out;
1136}
1137EXPORT_SYMBOL(sock_create_lite);
1138
1139/* No kernel lock held - perfect */
1140static unsigned int sock_poll(struct file *file, poll_table *wait)
1141{
1142        unsigned int busy_flag = 0;
1143        struct socket *sock;
1144
1145        /*
1146         *      We can't return errors to poll, so it's either yes or no.
1147         */
1148        sock = file->private_data;
1149
1150        if (sk_can_busy_loop(sock->sk)) {
1151                /* this socket can poll_ll so tell the system call */
1152                busy_flag = POLL_BUSY_LOOP;
1153
1154                /* once, only if requested by syscall */
1155                if (wait && (wait->_key & POLL_BUSY_LOOP))
1156                        sk_busy_loop(sock->sk, 1);
1157        }
1158
1159        return busy_flag | sock->ops->poll(file, sock, wait);
1160}
1161
1162static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1163{
1164        struct socket *sock = file->private_data;
1165
1166        return sock->ops->mmap(file, sock, vma);
1167}
1168
1169static int sock_close(struct inode *inode, struct file *filp)
1170{
1171        sock_release(SOCKET_I(inode));
1172        return 0;
1173}
1174
1175/*
1176 *      Update the socket async list
1177 *
1178 *      Fasync_list locking strategy.
1179 *
1180 *      1. fasync_list is modified only under process context socket lock
1181 *         i.e. under semaphore.
1182 *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1183 *         or under socket lock
1184 */
1185
1186static int sock_fasync(int fd, struct file *filp, int on)
1187{
1188        struct socket *sock = filp->private_data;
1189        struct sock *sk = sock->sk;
1190        struct socket_wq *wq;
1191
1192        if (sk == NULL)
1193                return -EINVAL;
1194
1195        lock_sock(sk);
1196        wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1197        fasync_helper(fd, filp, on, &wq->fasync_list);
1198
1199        if (!wq->fasync_list)
1200                sock_reset_flag(sk, SOCK_FASYNC);
1201        else
1202                sock_set_flag(sk, SOCK_FASYNC);
1203
1204        release_sock(sk);
1205        return 0;
1206}
1207
1208/* This function may be called only under socket lock or callback_lock or rcu_lock */
1209
1210int sock_wake_async(struct socket *sock, int how, int band)
1211{
1212        struct socket_wq *wq;
1213
1214        if (!sock)
1215                return -1;
1216        rcu_read_lock();
1217        wq = rcu_dereference(sock->wq);
1218        if (!wq || !wq->fasync_list) {
1219                rcu_read_unlock();
1220                return -1;
1221        }
1222        switch (how) {
1223        case SOCK_WAKE_WAITD:
1224                if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1225                        break;
1226                goto call_kill;
1227        case SOCK_WAKE_SPACE:
1228                if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1229                        break;
1230                /* fall through */
1231        case SOCK_WAKE_IO:
1232call_kill:
1233                kill_fasync(&wq->fasync_list, SIGIO, band);
1234                break;
1235        case SOCK_WAKE_URG:
1236                kill_fasync(&wq->fasync_list, SIGURG, band);
1237        }
1238        rcu_read_unlock();
1239        return 0;
1240}
1241EXPORT_SYMBOL(sock_wake_async);
1242
1243int __sock_create(struct net *net, int family, int type, int protocol,
1244                         struct socket **res, int kern)
1245{
1246        int err;
1247        struct socket *sock;
1248        const struct net_proto_family *pf;
1249
1250        /*
1251         *      Check protocol is in range
1252         */
1253        if (family < 0 || family >= NPROTO)
1254                return -EAFNOSUPPORT;
1255        if (type < 0 || type >= SOCK_MAX)
1256                return -EINVAL;
1257
1258        /* Compatibility.
1259
1260           This uglymoron is moved from INET layer to here to avoid
1261           deadlock in module load.
1262         */
1263        if (family == PF_INET && type == SOCK_PACKET) {
1264                static int warned;
1265                if (!warned) {
1266                        warned = 1;
1267                        printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1268                               current->comm);
1269                }
1270                family = PF_PACKET;
1271        }
1272
1273        err = security_socket_create(family, type, protocol, kern);
1274        if (err)
1275                return err;
1276
1277        /*
1278         *      Allocate the socket and allow the family to set things up. if
1279         *      the protocol is 0, the family is instructed to select an appropriate
1280         *      default.
1281         */
1282        sock = sock_alloc();
1283        if (!sock) {
1284                net_warn_ratelimited("socket: no more sockets\n");
1285                return -ENFILE; /* Not exactly a match, but its the
1286                                   closest posix thing */
1287        }
1288
1289        sock->type = type;
1290
1291#ifdef CONFIG_MODULES
1292        /* Attempt to load a protocol module if the find failed.
1293         *
1294         * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1295         * requested real, full-featured networking support upon configuration.
1296         * Otherwise module support will break!
1297         */
1298        if (rcu_access_pointer(net_families[family]) == NULL)
1299                request_module("net-pf-%d", family);
1300#endif
1301
1302        rcu_read_lock();
1303        pf = rcu_dereference(net_families[family]);
1304        err = -EAFNOSUPPORT;
1305        if (!pf)
1306                goto out_release;
1307
1308        /*
1309         * We will call the ->create function, that possibly is in a loadable
1310         * module, so we have to bump that loadable module refcnt first.
1311         */
1312        if (!try_module_get(pf->owner))
1313                goto out_release;
1314
1315        /* Now protected by module ref count */
1316        rcu_read_unlock();
1317
1318        err = pf->create(net, sock, protocol, kern);
1319        if (err < 0)
1320                goto out_module_put;
1321
1322        /*
1323         * Now to bump the refcnt of the [loadable] module that owns this
1324         * socket at sock_release time we decrement its refcnt.
1325         */
1326        if (!try_module_get(sock->ops->owner))
1327                goto out_module_busy;
1328
1329        /*
1330         * Now that we're done with the ->create function, the [loadable]
1331         * module can have its refcnt decremented
1332         */
1333        module_put(pf->owner);
1334        err = security_socket_post_create(sock, family, type, protocol, kern);
1335        if (err)
1336                goto out_sock_release;
1337        *res = sock;
1338
1339        return 0;
1340
1341out_module_busy:
1342        err = -EAFNOSUPPORT;
1343out_module_put:
1344        sock->ops = NULL;
1345        module_put(pf->owner);
1346out_sock_release:
1347        sock_release(sock);
1348        return err;
1349
1350out_release:
1351        rcu_read_unlock();
1352        goto out_sock_release;
1353}
1354EXPORT_SYMBOL(__sock_create);
1355
1356int sock_create(int family, int type, int protocol, struct socket **res)
1357{
1358        return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1359}
1360EXPORT_SYMBOL(sock_create);
1361
1362int sock_create_kern(int family, int type, int protocol, struct socket **res)
1363{
1364        return __sock_create(&init_net, family, type, protocol, res, 1);
1365}
1366EXPORT_SYMBOL(sock_create_kern);
1367
1368SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1369{
1370        int retval;
1371        struct socket *sock;
1372        int flags;
1373
1374        /* Check the SOCK_* constants for consistency.  */
1375        BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1376        BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1377        BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1378        BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1379
1380        flags = type & ~SOCK_TYPE_MASK;
1381        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1382                return -EINVAL;
1383        type &= SOCK_TYPE_MASK;
1384
1385        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1386                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1387
1388        retval = sock_create(family, type, protocol, &sock);
1389        if (retval < 0)
1390                goto out;
1391
1392        retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1393        if (retval < 0)
1394                goto out_release;
1395
1396out:
1397        /* It may be already another descriptor 8) Not kernel problem. */
1398        return retval;
1399
1400out_release:
1401        sock_release(sock);
1402        return retval;
1403}
1404
1405/*
1406 *      Create a pair of connected sockets.
1407 */
1408
1409SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1410                int __user *, usockvec)
1411{
1412        struct socket *sock1, *sock2;
1413        int fd1, fd2, err;
1414        struct file *newfile1, *newfile2;
1415        int flags;
1416
1417        flags = type & ~SOCK_TYPE_MASK;
1418        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1419                return -EINVAL;
1420        type &= SOCK_TYPE_MASK;
1421
1422        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1423                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1424
1425        /*
1426         * Obtain the first socket and check if the underlying protocol
1427         * supports the socketpair call.
1428         */
1429
1430        err = sock_create(family, type, protocol, &sock1);
1431        if (err < 0)
1432                goto out;
1433
1434        err = sock_create(family, type, protocol, &sock2);
1435        if (err < 0)
1436                goto out_release_1;
1437
1438        err = sock1->ops->socketpair(sock1, sock2);
1439        if (err < 0)
1440                goto out_release_both;
1441
1442        fd1 = get_unused_fd_flags(flags);
1443        if (unlikely(fd1 < 0)) {
1444                err = fd1;
1445                goto out_release_both;
1446        }
1447        fd2 = get_unused_fd_flags(flags);
1448        if (unlikely(fd2 < 0)) {
1449                err = fd2;
1450                put_unused_fd(fd1);
1451                goto out_release_both;
1452        }
1453
1454        newfile1 = sock_alloc_file(sock1, flags, NULL);
1455        if (unlikely(IS_ERR(newfile1))) {
1456                err = PTR_ERR(newfile1);
1457                put_unused_fd(fd1);
1458                put_unused_fd(fd2);
1459                goto out_release_both;
1460        }
1461
1462        newfile2 = sock_alloc_file(sock2, flags, NULL);
1463        if (IS_ERR(newfile2)) {
1464                err = PTR_ERR(newfile2);
1465                fput(newfile1);
1466                put_unused_fd(fd1);
1467                put_unused_fd(fd2);
1468                sock_release(sock2);
1469                goto out;
1470        }
1471
1472        audit_fd_pair(fd1, fd2);
1473        fd_install(fd1, newfile1);
1474        fd_install(fd2, newfile2);
1475        /* fd1 and fd2 may be already another descriptors.
1476         * Not kernel problem.
1477         */
1478
1479        err = put_user(fd1, &usockvec[0]);
1480        if (!err)
1481                err = put_user(fd2, &usockvec[1]);
1482        if (!err)
1483                return 0;
1484
1485        sys_close(fd2);
1486        sys_close(fd1);
1487        return err;
1488
1489out_release_both:
1490        sock_release(sock2);
1491out_release_1:
1492        sock_release(sock1);
1493out:
1494        return err;
1495}
1496
1497/*
1498 *      Bind a name to a socket. Nothing much to do here since it's
1499 *      the protocol's responsibility to handle the local address.
1500 *
1501 *      We move the socket address to kernel space before we call
1502 *      the protocol layer (having also checked the address is ok).
1503 */
1504
1505SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1506{
1507        struct socket *sock;
1508        struct sockaddr_storage address;
1509        int err, fput_needed;
1510
1511        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1512        if (sock) {
1513                err = move_addr_to_kernel(umyaddr, addrlen, &address);
1514                if (err >= 0) {
1515                        err = security_socket_bind(sock,
1516                                                   (struct sockaddr *)&address,
1517                                                   addrlen);
1518                        if (!err)
1519                                err = sock->ops->bind(sock,
1520                                                      (struct sockaddr *)
1521                                                      &address, addrlen);
1522                }
1523                fput_light(sock->file, fput_needed);
1524        }
1525        return err;
1526}
1527
1528/*
1529 *      Perform a listen. Basically, we allow the protocol to do anything
1530 *      necessary for a listen, and if that works, we mark the socket as
1531 *      ready for listening.
1532 */
1533
1534SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1535{
1536        struct socket *sock;
1537        int err, fput_needed;
1538        int somaxconn;
1539
1540        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1541        if (sock) {
1542                somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1543                if ((unsigned int)backlog > somaxconn)
1544                        backlog = somaxconn;
1545
1546                err = security_socket_listen(sock, backlog);
1547                if (!err)
1548                        err = sock->ops->listen(sock, backlog);
1549
1550                fput_light(sock->file, fput_needed);
1551        }
1552        return err;
1553}
1554
1555/*
1556 *      For accept, we attempt to create a new socket, set up the link
1557 *      with the client, wake up the client, then return the new
1558 *      connected fd. We collect the address of the connector in kernel
1559 *      space and move it to user at the very end. This is unclean because
1560 *      we open the socket then return an error.
1561 *
1562 *      1003.1g adds the ability to recvmsg() to query connection pending
1563 *      status to recvmsg. We need to add that support in a way thats
1564 *      clean when we restucture accept also.
1565 */
1566
1567SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1568                int __user *, upeer_addrlen, int, flags)
1569{
1570        struct socket *sock, *newsock;
1571        struct file *newfile;
1572        int err, len, newfd, fput_needed;
1573        struct sockaddr_storage address;
1574
1575        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1576                return -EINVAL;
1577
1578        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1579                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1580
1581        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1582        if (!sock)
1583                goto out;
1584
1585        err = -ENFILE;
1586        newsock = sock_alloc();
1587        if (!newsock)
1588                goto out_put;
1589
1590        newsock->type = sock->type;
1591        newsock->ops = sock->ops;
1592
1593        /*
1594         * We don't need try_module_get here, as the listening socket (sock)
1595         * has the protocol module (sock->ops->owner) held.
1596         */
1597        __module_get(newsock->ops->owner);
1598
1599        newfd = get_unused_fd_flags(flags);
1600        if (unlikely(newfd < 0)) {
1601                err = newfd;
1602                sock_release(newsock);
1603                goto out_put;
1604        }
1605        newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1606        if (unlikely(IS_ERR(newfile))) {
1607                err = PTR_ERR(newfile);
1608                put_unused_fd(newfd);
1609                sock_release(newsock);
1610                goto out_put;
1611        }
1612
1613        err = security_socket_accept(sock, newsock);
1614        if (err)
1615                goto out_fd;
1616
1617        err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1618        if (err < 0)
1619                goto out_fd;
1620
1621        if (upeer_sockaddr) {
1622                if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1623                                          &len, 2) < 0) {
1624                        err = -ECONNABORTED;
1625                        goto out_fd;
1626                }
1627                err = move_addr_to_user(&address,
1628                                        len, upeer_sockaddr, upeer_addrlen);
1629                if (err < 0)
1630                        goto out_fd;
1631        }
1632
1633        /* File flags are not inherited via accept() unlike another OSes. */
1634
1635        fd_install(newfd, newfile);
1636        err = newfd;
1637
1638out_put:
1639        fput_light(sock->file, fput_needed);
1640out:
1641        return err;
1642out_fd:
1643        fput(newfile);
1644        put_unused_fd(newfd);
1645        goto out_put;
1646}
1647
1648SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1649                int __user *, upeer_addrlen)
1650{
1651        return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1652}
1653
1654/*
1655 *      Attempt to connect to a socket with the server address.  The address
1656 *      is in user space so we verify it is OK and move it to kernel space.
1657 *
1658 *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1659 *      break bindings
1660 *
1661 *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1662 *      other SEQPACKET protocols that take time to connect() as it doesn't
1663 *      include the -EINPROGRESS status for such sockets.
1664 */
1665
1666SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1667                int, addrlen)
1668{
1669        struct socket *sock;
1670        struct sockaddr_storage address;
1671        int err, fput_needed;
1672
1673        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1674        if (!sock)
1675                goto out;
1676        err = move_addr_to_kernel(uservaddr, addrlen, &address);
1677        if (err < 0)
1678                goto out_put;
1679
1680        err =
1681            security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1682        if (err)
1683                goto out_put;
1684
1685        err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1686                                 sock->file->f_flags);
1687out_put:
1688        fput_light(sock->file, fput_needed);
1689out:
1690        return err;
1691}
1692
1693/*
1694 *      Get the local address ('name') of a socket object. Move the obtained
1695 *      name to user space.
1696 */
1697
1698SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1699                int __user *, usockaddr_len)
1700{
1701        struct socket *sock;
1702        struct sockaddr_storage address;
1703        int len, err, fput_needed;
1704
1705        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1706        if (!sock)
1707                goto out;
1708
1709        err = security_socket_getsockname(sock);
1710        if (err)
1711                goto out_put;
1712
1713        err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1714        if (err)
1715                goto out_put;
1716        err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1717
1718out_put:
1719        fput_light(sock->file, fput_needed);
1720out:
1721        return err;
1722}
1723
1724/*
1725 *      Get the remote address ('name') of a socket object. Move the obtained
1726 *      name to user space.
1727 */
1728
1729SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1730                int __user *, usockaddr_len)
1731{
1732        struct socket *sock;
1733        struct sockaddr_storage address;
1734        int len, err, fput_needed;
1735
1736        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1737        if (sock != NULL) {
1738                err = security_socket_getpeername(sock);
1739                if (err) {
1740                        fput_light(sock->file, fput_needed);
1741                        return err;
1742                }
1743
1744                err =
1745                    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1746                                       1);
1747                if (!err)
1748                        err = move_addr_to_user(&address, len, usockaddr,
1749                                                usockaddr_len);
1750                fput_light(sock->file, fput_needed);
1751        }
1752        return err;
1753}
1754
1755/*
1756 *      Send a datagram to a given address. We move the address into kernel
1757 *      space and check the user space data area is readable before invoking
1758 *      the protocol.
1759 */
1760
1761SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1762                unsigned int, flags, struct sockaddr __user *, addr,
1763                int, addr_len)
1764{
1765        struct socket *sock;
1766        struct sockaddr_storage address;
1767        int err;
1768        struct msghdr msg;
1769        struct iovec iov;
1770        int fput_needed;
1771
1772        if (len > INT_MAX)
1773                len = INT_MAX;
1774        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1775        if (!sock)
1776                goto out;
1777
1778        iov.iov_base = buff;
1779        iov.iov_len = len;
1780        msg.msg_name = NULL;
1781        msg.msg_iov = &iov;
1782        msg.msg_iovlen = 1;
1783        msg.msg_control = NULL;
1784        msg.msg_controllen = 0;
1785        msg.msg_namelen = 0;
1786        if (addr) {
1787                err = move_addr_to_kernel(addr, addr_len, &address);
1788                if (err < 0)
1789                        goto out_put;
1790                msg.msg_name = (struct sockaddr *)&address;
1791                msg.msg_namelen = addr_len;
1792        }
1793        if (sock->file->f_flags & O_NONBLOCK)
1794                flags |= MSG_DONTWAIT;
1795        msg.msg_flags = flags;
1796        err = sock_sendmsg(sock, &msg, len);
1797
1798out_put:
1799        fput_light(sock->file, fput_needed);
1800out:
1801        return err;
1802}
1803
1804/*
1805 *      Send a datagram down a socket.
1806 */
1807
1808SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1809                unsigned int, flags)
1810{
1811        return sys_sendto(fd, buff, len, flags, NULL, 0);
1812}
1813
1814/*
1815 *      Receive a frame from the socket and optionally record the address of the
1816 *      sender. We verify the buffers are writable and if needed move the
1817 *      sender address from kernel to user space.
1818 */
1819
1820SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1821                unsigned int, flags, struct sockaddr __user *, addr,
1822                int __user *, addr_len)
1823{
1824        struct socket *sock;
1825        struct iovec iov;
1826        struct msghdr msg;
1827        struct sockaddr_storage address;
1828        int err, err2;
1829        int fput_needed;
1830
1831        if (size > INT_MAX)
1832                size = INT_MAX;
1833        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1834        if (!sock)
1835                goto out;
1836
1837        msg.msg_control = NULL;
1838        msg.msg_controllen = 0;
1839        msg.msg_iovlen = 1;
1840        msg.msg_iov = &iov;
1841        iov.iov_len = size;
1842        iov.iov_base = ubuf;
1843        msg.msg_name = (struct sockaddr *)&address;
1844        msg.msg_namelen = sizeof(address);
1845        if (sock->file->f_flags & O_NONBLOCK)
1846                flags |= MSG_DONTWAIT;
1847        err = sock_recvmsg(sock, &msg, size, flags);
1848
1849        if (err >= 0 && addr != NULL) {
1850                err2 = move_addr_to_user(&address,
1851                                         msg.msg_namelen, addr, addr_len);
1852                if (err2 < 0)
1853                        err = err2;
1854        }
1855
1856        fput_light(sock->file, fput_needed);
1857out:
1858        return err;
1859}
1860
1861/*
1862 *      Receive a datagram from a socket.
1863 */
1864
1865asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1866                         unsigned int flags)
1867{
1868        return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1869}
1870
1871/*
1872 *      Set a socket option. Because we don't know the option lengths we have
1873 *      to pass the user mode parameter for the protocols to sort out.
1874 */
1875
1876SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1877                char __user *, optval, int, optlen)
1878{
1879        int err, fput_needed;
1880        struct socket *sock;
1881
1882        if (optlen < 0)
1883                return -EINVAL;
1884
1885        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1886        if (sock != NULL) {
1887                err = security_socket_setsockopt(sock, level, optname);
1888                if (err)
1889                        goto out_put;
1890
1891                if (level == SOL_SOCKET)
1892                        err =
1893                            sock_setsockopt(sock, level, optname, optval,
1894                                            optlen);
1895                else
1896                        err =
1897                            sock->ops->setsockopt(sock, level, optname, optval,
1898                                                  optlen);
1899out_put:
1900                fput_light(sock->file, fput_needed);
1901        }
1902        return err;
1903}
1904
1905/*
1906 *      Get a socket option. Because we don't know the option lengths we have
1907 *      to pass a user mode parameter for the protocols to sort out.
1908 */
1909
1910SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1911                char __user *, optval, int __user *, optlen)
1912{
1913        int err, fput_needed;
1914        struct socket *sock;
1915
1916        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1917        if (sock != NULL) {
1918                err = security_socket_getsockopt(sock, level, optname);
1919                if (err)
1920                        goto out_put;
1921
1922                if (level == SOL_SOCKET)
1923                        err =
1924                            sock_getsockopt(sock, level, optname, optval,
1925                                            optlen);
1926                else
1927                        err =
1928                            sock->ops->getsockopt(sock, level, optname, optval,
1929                                                  optlen);
1930out_put:
1931                fput_light(sock->file, fput_needed);
1932        }
1933        return err;
1934}
1935
1936/*
1937 *      Shutdown a socket.
1938 */
1939
1940SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1941{
1942        int err, fput_needed;
1943        struct socket *sock;
1944
1945        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1946        if (sock != NULL) {
1947                err = security_socket_shutdown(sock, how);
1948                if (!err)
1949                        err = sock->ops->shutdown(sock, how);
1950                fput_light(sock->file, fput_needed);
1951        }
1952        return err;
1953}
1954
1955/* A couple of helpful macros for getting the address of the 32/64 bit
1956 * fields which are the same type (int / unsigned) on our platforms.
1957 */
1958#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1959#define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1960#define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1961
1962struct used_address {
1963        struct sockaddr_storage name;
1964        unsigned int name_len;
1965};
1966
1967static int copy_msghdr_from_user(struct msghdr *kmsg,
1968                                 struct msghdr __user *umsg)
1969{
1970        if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1971                return -EFAULT;
1972        if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1973                return -EINVAL;
1974        return 0;
1975}
1976
1977static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1978                         struct msghdr *msg_sys, unsigned int flags,
1979                         struct used_address *used_address)
1980{
1981        struct compat_msghdr __user *msg_compat =
1982            (struct compat_msghdr __user *)msg;
1983        struct sockaddr_storage address;
1984        struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1985        unsigned char ctl[sizeof(struct cmsghdr) + 20]
1986            __attribute__ ((aligned(sizeof(__kernel_size_t))));
1987        /* 20 is size of ipv6_pktinfo */
1988        unsigned char *ctl_buf = ctl;
1989        int err, ctl_len, total_len;
1990
1991        err = -EFAULT;
1992        if (MSG_CMSG_COMPAT & flags) {
1993                if (get_compat_msghdr(msg_sys, msg_compat))
1994                        return -EFAULT;
1995        } else {
1996                err = copy_msghdr_from_user(msg_sys, msg);
1997                if (err)
1998                        return err;
1999        }
2000
2001        if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2002                err = -EMSGSIZE;
2003                if (msg_sys->msg_iovlen > UIO_MAXIOV)
2004                        goto out;
2005                err = -ENOMEM;
2006                iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2007                              GFP_KERNEL);
2008                if (!iov)
2009                        goto out;
2010        }
2011
2012        /* This will also move the address data into kernel space */
2013        if (MSG_CMSG_COMPAT & flags) {
2014                err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2015        } else
2016                err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2017        if (err < 0)
2018                goto out_freeiov;
2019        total_len = err;
2020
2021        err = -ENOBUFS;
2022
2023        if (msg_sys->msg_controllen > INT_MAX)
2024                goto out_freeiov;
2025        ctl_len = msg_sys->msg_controllen;
2026        if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2027                err =
2028                    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2029                                                     sizeof(ctl));
2030                if (err)
2031                        goto out_freeiov;
2032                ctl_buf = msg_sys->msg_control;
2033                ctl_len = msg_sys->msg_controllen;
2034        } else if (ctl_len) {
2035                if (ctl_len > sizeof(ctl)) {
2036                        ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2037                        if (ctl_buf == NULL)
2038                                goto out_freeiov;
2039                }
2040                err = -EFAULT;
2041                /*
2042                 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2043                 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2044                 * checking falls down on this.
2045                 */
2046                if (copy_from_user(ctl_buf,
2047                                   (void __user __force *)msg_sys->msg_control,
2048                                   ctl_len))
2049                        goto out_freectl;
2050                msg_sys->msg_control = ctl_buf;
2051        }
2052        msg_sys->msg_flags = flags;
2053
2054        if (sock->file->f_flags & O_NONBLOCK)
2055                msg_sys->msg_flags |= MSG_DONTWAIT;
2056        /*
2057         * If this is sendmmsg() and current destination address is same as
2058         * previously succeeded address, omit asking LSM's decision.
2059         * used_address->name_len is initialized to UINT_MAX so that the first
2060         * destination address never matches.
2061         */
2062        if (used_address && msg_sys->msg_name &&
2063            used_address->name_len == msg_sys->msg_namelen &&
2064            !memcmp(&used_address->name, msg_sys->msg_name,
2065                    used_address->name_len)) {
2066                err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2067                goto out_freectl;
2068        }
2069        err = sock_sendmsg(sock, msg_sys, total_len);
2070        /*
2071         * If this is sendmmsg() and sending to current destination address was
2072         * successful, remember it.
2073         */
2074        if (used_address && err >= 0) {
2075                used_address->name_len = msg_sys->msg_namelen;
2076                if (msg_sys->msg_name)
2077                        memcpy(&used_address->name, msg_sys->msg_name,
2078                               used_address->name_len);
2079        }
2080
2081out_freectl:
2082        if (ctl_buf != ctl)
2083                sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2084out_freeiov:
2085        if (iov != iovstack)
2086                kfree(iov);
2087out:
2088        return err;
2089}
2090
2091/*
2092 *      BSD sendmsg interface
2093 */
2094
2095long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2096{
2097        int fput_needed, err;
2098        struct msghdr msg_sys;
2099        struct socket *sock;
2100
2101        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2102        if (!sock)
2103                goto out;
2104
2105        err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2106
2107        fput_light(sock->file, fput_needed);
2108out:
2109        return err;
2110}
2111
2112SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2113{
2114        if (flags & MSG_CMSG_COMPAT)
2115                return -EINVAL;
2116        return __sys_sendmsg(fd, msg, flags);
2117}
2118
2119/*
2120 *      Linux sendmmsg interface
2121 */
2122
2123int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2124                   unsigned int flags)
2125{
2126        int fput_needed, err, datagrams;
2127        struct socket *sock;
2128        struct mmsghdr __user *entry;
2129        struct compat_mmsghdr __user *compat_entry;
2130        struct msghdr msg_sys;
2131        struct used_address used_address;
2132
2133        if (vlen > UIO_MAXIOV)
2134                vlen = UIO_MAXIOV;
2135
2136        datagrams = 0;
2137
2138        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2139        if (!sock)
2140                return err;
2141
2142        used_address.name_len = UINT_MAX;
2143        entry = mmsg;
2144        compat_entry = (struct compat_mmsghdr __user *)mmsg;
2145        err = 0;
2146
2147        while (datagrams < vlen) {
2148                if (MSG_CMSG_COMPAT & flags) {
2149                        err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2150                                             &msg_sys, flags, &used_address);
2151                        if (err < 0)
2152                                break;
2153                        err = __put_user(err, &compat_entry->msg_len);
2154                        ++compat_entry;
2155                } else {
2156                        err = ___sys_sendmsg(sock,
2157                                             (struct msghdr __user *)entry,
2158                                             &msg_sys, flags, &used_address);
2159                        if (err < 0)
2160                                break;
2161                        err = put_user(err, &entry->msg_len);
2162                        ++entry;
2163                }
2164
2165                if (err)
2166                        break;
2167                ++datagrams;
2168        }
2169
2170        fput_light(sock->file, fput_needed);
2171
2172        /* We only return an error if no datagrams were able to be sent */
2173        if (datagrams != 0)
2174                return datagrams;
2175
2176        return err;
2177}
2178
2179SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2180                unsigned int, vlen, unsigned int, flags)
2181{
2182        if (flags & MSG_CMSG_COMPAT)
2183                return -EINVAL;
2184        return __sys_sendmmsg(fd, mmsg, vlen, flags);
2185}
2186
2187static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2188                         struct msghdr *msg_sys, unsigned int flags, int nosec)
2189{
2190        struct compat_msghdr __user *msg_compat =
2191            (struct compat_msghdr __user *)msg;
2192        struct iovec iovstack[UIO_FASTIOV];
2193        struct iovec *iov = iovstack;
2194        unsigned long cmsg_ptr;
2195        int err, total_len, len;
2196
2197        /* kernel mode address */
2198        struct sockaddr_storage addr;
2199
2200        /* user mode address pointers */
2201        struct sockaddr __user *uaddr;
2202        int __user *uaddr_len;
2203
2204        if (MSG_CMSG_COMPAT & flags) {
2205                if (get_compat_msghdr(msg_sys, msg_compat))
2206                        return -EFAULT;
2207        } else {
2208                err = copy_msghdr_from_user(msg_sys, msg);
2209                if (err)
2210                        return err;
2211        }
2212
2213        if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2214                err = -EMSGSIZE;
2215                if (msg_sys->msg_iovlen > UIO_MAXIOV)
2216                        goto out;
2217                err = -ENOMEM;
2218                iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2219                              GFP_KERNEL);
2220                if (!iov)
2221                        goto out;
2222        }
2223
2224        /*
2225         *      Save the user-mode address (verify_iovec will change the
2226         *      kernel msghdr to use the kernel address space)
2227         */
2228
2229        uaddr = (__force void __user *)msg_sys->msg_name;
2230        uaddr_len = COMPAT_NAMELEN(msg);
2231        if (MSG_CMSG_COMPAT & flags) {
2232                err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2233        } else
2234                err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2235        if (err < 0)
2236                goto out_freeiov;
2237        total_len = err;
2238
2239        cmsg_ptr = (unsigned long)msg_sys->msg_control;
2240        msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2241
2242        if (sock->file->f_flags & O_NONBLOCK)
2243                flags |= MSG_DONTWAIT;
2244        err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2245                                                          total_len, flags);
2246        if (err < 0)
2247                goto out_freeiov;
2248        len = err;
2249
2250        if (uaddr != NULL) {
2251                err = move_addr_to_user(&addr,
2252                                        msg_sys->msg_namelen, uaddr,
2253                                        uaddr_len);
2254                if (err < 0)
2255                        goto out_freeiov;
2256        }
2257        err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2258                         COMPAT_FLAGS(msg));
2259        if (err)
2260                goto out_freeiov;
2261        if (MSG_CMSG_COMPAT & flags)
2262                err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2263                                 &msg_compat->msg_controllen);
2264        else
2265                err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2266                                 &msg->msg_controllen);
2267        if (err)
2268                goto out_freeiov;
2269        err = len;
2270
2271out_freeiov:
2272        if (iov != iovstack)
2273                kfree(iov);
2274out:
2275        return err;
2276}
2277
2278/*
2279 *      BSD recvmsg interface
2280 */
2281
2282long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2283{
2284        int fput_needed, err;
2285        struct msghdr msg_sys;
2286        struct socket *sock;
2287
2288        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2289        if (!sock)
2290                goto out;
2291
2292        err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2293
2294        fput_light(sock->file, fput_needed);
2295out:
2296        return err;
2297}
2298
2299SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2300                unsigned int, flags)
2301{
2302        if (flags & MSG_CMSG_COMPAT)
2303                return -EINVAL;
2304        return __sys_recvmsg(fd, msg, flags);
2305}
2306
2307/*
2308 *     Linux recvmmsg interface
2309 */
2310
2311int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2312                   unsigned int flags, struct timespec *timeout)
2313{
2314        int fput_needed, err, datagrams;
2315        struct socket *sock;
2316        struct mmsghdr __user *entry;
2317        struct compat_mmsghdr __user *compat_entry;
2318        struct msghdr msg_sys;
2319        struct timespec end_time;
2320
2321        if (timeout &&
2322            poll_select_set_timeout(&end_time, timeout->tv_sec,
2323                                    timeout->tv_nsec))
2324                return -EINVAL;
2325
2326        datagrams = 0;
2327
2328        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2329        if (!sock)
2330                return err;
2331
2332        err = sock_error(sock->sk);
2333        if (err)
2334                goto out_put;
2335
2336        entry = mmsg;
2337        compat_entry = (struct compat_mmsghdr __user *)mmsg;
2338
2339        while (datagrams < vlen) {
2340                /*
2341                 * No need to ask LSM for more than the first datagram.
2342                 */
2343                if (MSG_CMSG_COMPAT & flags) {
2344                        err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2345                                             &msg_sys, flags & ~MSG_WAITFORONE,
2346                                             datagrams);
2347                        if (err < 0)
2348                                break;
2349                        err = __put_user(err, &compat_entry->msg_len);
2350                        ++compat_entry;
2351                } else {
2352                        err = ___sys_recvmsg(sock,
2353                                             (struct msghdr __user *)entry,
2354                                             &msg_sys, flags & ~MSG_WAITFORONE,
2355                                             datagrams);
2356                        if (err < 0)
2357                                break;
2358                        err = put_user(err, &entry->msg_len);
2359                        ++entry;
2360                }
2361
2362                if (err)
2363                        break;
2364                ++datagrams;
2365
2366                /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2367                if (flags & MSG_WAITFORONE)
2368                        flags |= MSG_DONTWAIT;
2369
2370                if (timeout) {
2371                        ktime_get_ts(timeout);
2372                        *timeout = timespec_sub(end_time, *timeout);
2373                        if (timeout->tv_sec < 0) {
2374                                timeout->tv_sec = timeout->tv_nsec = 0;
2375                                break;
2376                        }
2377
2378                        /* Timeout, return less than vlen datagrams */
2379                        if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2380                                break;
2381                }
2382
2383                /* Out of band data, return right away */
2384                if (msg_sys.msg_flags & MSG_OOB)
2385                        break;
2386        }
2387
2388out_put:
2389        fput_light(sock->file, fput_needed);
2390
2391        if (err == 0)
2392                return datagrams;
2393
2394        if (datagrams != 0) {
2395                /*
2396                 * We may return less entries than requested (vlen) if the
2397                 * sock is non block and there aren't enough datagrams...
2398                 */
2399                if (err != -EAGAIN) {
2400                        /*
2401                         * ... or  if recvmsg returns an error after we
2402                         * received some datagrams, where we record the
2403                         * error to return on the next call or if the
2404                         * app asks about it using getsockopt(SO_ERROR).
2405                         */
2406                        sock->sk->sk_err = -err;
2407                }
2408
2409                return datagrams;
2410        }
2411
2412        return err;
2413}
2414
2415SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2416                unsigned int, vlen, unsigned int, flags,
2417                struct timespec __user *, timeout)
2418{
2419        int datagrams;
2420        struct timespec timeout_sys;
2421
2422        if (flags & MSG_CMSG_COMPAT)
2423                return -EINVAL;
2424
2425        if (!timeout)
2426                return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2427
2428        if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2429                return -EFAULT;
2430
2431        datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2432
2433        if (datagrams > 0 &&
2434            copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2435                datagrams = -EFAULT;
2436
2437        return datagrams;
2438}
2439
2440#ifdef __ARCH_WANT_SYS_SOCKETCALL
2441/* Argument list sizes for sys_socketcall */
2442#define AL(x) ((x) * sizeof(unsigned long))
2443static const unsigned char nargs[21] = {
2444        AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2445        AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2446        AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2447        AL(4), AL(5), AL(4)
2448};
2449
2450#undef AL
2451
2452/*
2453 *      System call vectors.
2454 *
2455 *      Argument checking cleaned up. Saved 20% in size.
2456 *  This function doesn't need to set the kernel lock because
2457 *  it is set by the callees.
2458 */
2459
2460SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2461{
2462        unsigned long a[AUDITSC_ARGS];
2463        unsigned long a0, a1;
2464        int err;
2465        unsigned int len;
2466
2467        if (call < 1 || call > SYS_SENDMMSG)
2468                return -EINVAL;
2469
2470        len = nargs[call];
2471        if (len > sizeof(a))
2472                return -EINVAL;
2473
2474        /* copy_from_user should be SMP safe. */
2475        if (copy_from_user(a, args, len))
2476                return -EFAULT;
2477
2478        err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2479        if (err)
2480                return err;
2481
2482        a0 = a[0];
2483        a1 = a[1];
2484
2485        switch (call) {
2486        case SYS_SOCKET:
2487                err = sys_socket(a0, a1, a[2]);
2488                break;
2489        case SYS_BIND:
2490                err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2491                break;
2492        case SYS_CONNECT:
2493                err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2494                break;
2495        case SYS_LISTEN:
2496                err = sys_listen(a0, a1);
2497                break;
2498        case SYS_ACCEPT:
2499                err = sys_accept4(a0, (struct sockaddr __user *)a1,
2500                                  (int __user *)a[2], 0);
2501                break;
2502        case SYS_GETSOCKNAME:
2503                err =
2504                    sys_getsockname(a0, (struct sockaddr __user *)a1,
2505                                    (int __user *)a[2]);
2506                break;
2507        case SYS_GETPEERNAME:
2508                err =
2509                    sys_getpeername(a0, (struct sockaddr __user *)a1,
2510                                    (int __user *)a[2]);
2511                break;
2512        case SYS_SOCKETPAIR:
2513                err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2514                break;
2515        case SYS_SEND:
2516                err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2517                break;
2518        case SYS_SENDTO:
2519                err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2520                                 (struct sockaddr __user *)a[4], a[5]);
2521                break;
2522        case SYS_RECV:
2523                err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2524                break;
2525        case SYS_RECVFROM:
2526                err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2527                                   (struct sockaddr __user *)a[4],
2528                                   (int __user *)a[5]);
2529                break;
2530        case SYS_SHUTDOWN:
2531                err = sys_shutdown(a0, a1);
2532                break;
2533        case SYS_SETSOCKOPT:
2534                err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2535                break;
2536        case SYS_GETSOCKOPT:
2537                err =
2538                    sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2539                                   (int __user *)a[4]);
2540                break;
2541        case SYS_SENDMSG:
2542                err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2543                break;
2544        case SYS_SENDMMSG:
2545                err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2546                break;
2547        case SYS_RECVMSG:
2548                err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2549                break;
2550        case SYS_RECVMMSG:
2551                err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2552                                   (struct timespec __user *)a[4]);
2553                break;
2554        case SYS_ACCEPT4:
2555                err = sys_accept4(a0, (struct sockaddr __user *)a1,
2556                                  (int __user *)a[2], a[3]);
2557                break;
2558        default:
2559                err = -EINVAL;
2560                break;
2561        }
2562        return err;
2563}
2564
2565#endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2566
2567/**
2568 *      sock_register - add a socket protocol handler
2569 *      @ops: description of protocol
2570 *
2571 *      This function is called by a protocol handler that wants to
2572 *      advertise its address family, and have it linked into the
2573 *      socket interface. The value ops->family coresponds to the
2574 *      socket system call protocol family.
2575 */
2576int sock_register(const struct net_proto_family *ops)
2577{
2578        int err;
2579
2580        if (ops->family >= NPROTO) {
2581                printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2582                       NPROTO);
2583                return -ENOBUFS;
2584        }
2585
2586        spin_lock(&net_family_lock);
2587        if (rcu_dereference_protected(net_families[ops->family],
2588                                      lockdep_is_held(&net_family_lock)))
2589                err = -EEXIST;
2590        else {
2591                rcu_assign_pointer(net_families[ops->family], ops);
2592                err = 0;
2593        }
2594        spin_unlock(&net_family_lock);
2595
2596        printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2597        return err;
2598}
2599EXPORT_SYMBOL(sock_register);
2600
2601/**
2602 *      sock_unregister - remove a protocol handler
2603 *      @family: protocol family to remove
2604 *
2605 *      This function is called by a protocol handler that wants to
2606 *      remove its address family, and have it unlinked from the
2607 *      new socket creation.
2608 *
2609 *      If protocol handler is a module, then it can use module reference
2610 *      counts to protect against new references. If protocol handler is not
2611 *      a module then it needs to provide its own protection in
2612 *      the ops->create routine.
2613 */
2614void sock_unregister(int family)
2615{
2616        BUG_ON(family < 0 || family >= NPROTO);
2617
2618        spin_lock(&net_family_lock);
2619        RCU_INIT_POINTER(net_families[family], NULL);
2620        spin_unlock(&net_family_lock);
2621
2622        synchronize_rcu();
2623
2624        printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2625}
2626EXPORT_SYMBOL(sock_unregister);
2627
2628static int __init sock_init(void)
2629{
2630        int err;
2631        /*
2632         *      Initialize the network sysctl infrastructure.
2633         */
2634        err = net_sysctl_init();
2635        if (err)
2636                goto out;
2637
2638        /*
2639         *      Initialize skbuff SLAB cache
2640         */
2641        skb_init();
2642
2643        /*
2644         *      Initialize the protocols module.
2645         */
2646
2647        init_inodecache();
2648
2649        err = register_filesystem(&sock_fs_type);
2650        if (err)
2651                goto out_fs;
2652        sock_mnt = kern_mount(&sock_fs_type);
2653        if (IS_ERR(sock_mnt)) {
2654                err = PTR_ERR(sock_mnt);
2655                goto out_mount;
2656        }
2657
2658        /* The real protocol initialization is performed in later initcalls.
2659         */
2660
2661#ifdef CONFIG_NETFILTER
2662        err = netfilter_init();
2663        if (err)
2664                goto out;
2665#endif
2666
2667#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2668        skb_timestamping_init();
2669#endif
2670
2671out:
2672        return err;
2673
2674out_mount:
2675        unregister_filesystem(&sock_fs_type);
2676out_fs:
2677        goto out;
2678}
2679
2680core_initcall(sock_init);       /* early initcall */
2681
2682#ifdef CONFIG_PROC_FS
2683void socket_seq_show(struct seq_file *seq)
2684{
2685        int cpu;
2686        int counter = 0;
2687
2688        for_each_possible_cpu(cpu)
2689            counter += per_cpu(sockets_in_use, cpu);
2690
2691        /* It can be negative, by the way. 8) */
2692        if (counter < 0)
2693                counter = 0;
2694
2695        seq_printf(seq, "sockets: used %d\n", counter);
2696}
2697#endif                          /* CONFIG_PROC_FS */
2698
2699#ifdef CONFIG_COMPAT
2700static int do_siocgstamp(struct net *net, struct socket *sock,
2701                         unsigned int cmd, void __user *up)
2702{
2703        mm_segment_t old_fs = get_fs();
2704        struct timeval ktv;
2705        int err;
2706
2707        set_fs(KERNEL_DS);
2708        err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2709        set_fs(old_fs);
2710        if (!err)
2711                err = compat_put_timeval(&ktv, up);
2712
2713        return err;
2714}
2715
2716static int do_siocgstampns(struct net *net, struct socket *sock,
2717                           unsigned int cmd, void __user *up)
2718{
2719        mm_segment_t old_fs = get_fs();
2720        struct timespec kts;
2721        int err;
2722
2723        set_fs(KERNEL_DS);
2724        err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2725        set_fs(old_fs);
2726        if (!err)
2727                err = compat_put_timespec(&kts, up);
2728
2729        return err;
2730}
2731
2732static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2733{
2734        struct ifreq __user *uifr;
2735        int err;
2736
2737        uifr = compat_alloc_user_space(sizeof(struct ifreq));
2738        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2739                return -EFAULT;
2740
2741        err = dev_ioctl(net, SIOCGIFNAME, uifr);
2742        if (err)
2743                return err;
2744
2745        if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2746                return -EFAULT;
2747
2748        return 0;
2749}
2750
2751static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2752{
2753        struct compat_ifconf ifc32;
2754        struct ifconf ifc;
2755        struct ifconf __user *uifc;
2756        struct compat_ifreq __user *ifr32;
2757        struct ifreq __user *ifr;
2758        unsigned int i, j;
2759        int err;
2760
2761        if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2762                return -EFAULT;
2763
2764        memset(&ifc, 0, sizeof(ifc));
2765        if (ifc32.ifcbuf == 0) {
2766                ifc32.ifc_len = 0;
2767                ifc.ifc_len = 0;
2768                ifc.ifc_req = NULL;
2769                uifc = compat_alloc_user_space(sizeof(struct ifconf));
2770        } else {
2771                size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2772                        sizeof(struct ifreq);
2773                uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2774                ifc.ifc_len = len;
2775                ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2776                ifr32 = compat_ptr(ifc32.ifcbuf);
2777                for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2778                        if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2779                                return -EFAULT;
2780                        ifr++;
2781                        ifr32++;
2782                }
2783        }
2784        if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2785                return -EFAULT;
2786
2787        err = dev_ioctl(net, SIOCGIFCONF, uifc);
2788        if (err)
2789                return err;
2790
2791        if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2792                return -EFAULT;
2793
2794        ifr = ifc.ifc_req;
2795        ifr32 = compat_ptr(ifc32.ifcbuf);
2796        for (i = 0, j = 0;
2797             i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2798             i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2799                if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2800                        return -EFAULT;
2801                ifr32++;
2802                ifr++;
2803        }
2804
2805        if (ifc32.ifcbuf == 0) {
2806                /* Translate from 64-bit structure multiple to
2807                 * a 32-bit one.
2808                 */
2809                i = ifc.ifc_len;
2810                i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2811                ifc32.ifc_len = i;
2812        } else {
2813                ifc32.ifc_len = i;
2814        }
2815        if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2816                return -EFAULT;
2817
2818        return 0;
2819}
2820
2821static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2822{
2823        struct compat_ethtool_rxnfc __user *compat_rxnfc;
2824        bool convert_in = false, convert_out = false;
2825        size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2826        struct ethtool_rxnfc __user *rxnfc;
2827        struct ifreq __user *ifr;
2828        u32 rule_cnt = 0, actual_rule_cnt;
2829        u32 ethcmd;
2830        u32 data;
2831        int ret;
2832
2833        if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2834                return -EFAULT;
2835
2836        compat_rxnfc = compat_ptr(data);
2837
2838        if (get_user(ethcmd, &compat_rxnfc->cmd))
2839                return -EFAULT;
2840
2841        /* Most ethtool structures are defined without padding.
2842         * Unfortunately struct ethtool_rxnfc is an exception.
2843         */
2844        switch (ethcmd) {
2845        default:
2846                break;
2847        case ETHTOOL_GRXCLSRLALL:
2848                /* Buffer size is variable */
2849                if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2850                        return -EFAULT;
2851                if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2852                        return -ENOMEM;
2853                buf_size += rule_cnt * sizeof(u32);
2854                /* fall through */
2855        case ETHTOOL_GRXRINGS:
2856        case ETHTOOL_GRXCLSRLCNT:
2857        case ETHTOOL_GRXCLSRULE:
2858        case ETHTOOL_SRXCLSRLINS:
2859                convert_out = true;
2860                /* fall through */
2861        case ETHTOOL_SRXCLSRLDEL:
2862                buf_size += sizeof(struct ethtool_rxnfc);
2863                convert_in = true;
2864                break;
2865        }
2866
2867        ifr = compat_alloc_user_space(buf_size);
2868        rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2869
2870        if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2871                return -EFAULT;
2872
2873        if (put_user(convert_in ? rxnfc : compat_ptr(data),
2874                     &ifr->ifr_ifru.ifru_data))
2875                return -EFAULT;
2876
2877        if (convert_in) {
2878                /* We expect there to be holes between fs.m_ext and
2879                 * fs.ring_cookie and at the end of fs, but nowhere else.
2880                 */
2881                BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2882                             sizeof(compat_rxnfc->fs.m_ext) !=
2883                             offsetof(struct ethtool_rxnfc, fs.m_ext) +
2884                             sizeof(rxnfc->fs.m_ext));
2885                BUILD_BUG_ON(
2886                        offsetof(struct compat_ethtool_rxnfc, fs.location) -
2887                        offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2888                        offsetof(struct ethtool_rxnfc, fs.location) -
2889                        offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2890
2891                if (copy_in_user(rxnfc, compat_rxnfc,
2892                                 (void __user *)(&rxnfc->fs.m_ext + 1) -
2893                                 (void __user *)rxnfc) ||
2894                    copy_in_user(&rxnfc->fs.ring_cookie,
2895                                 &compat_rxnfc->fs.ring_cookie,
2896                                 (void __user *)(&rxnfc->fs.location + 1) -
2897                                 (void __user *)&rxnfc->fs.ring_cookie) ||
2898                    copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2899                                 sizeof(rxnfc->rule_cnt)))
2900                        return -EFAULT;
2901        }
2902
2903        ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2904        if (ret)
2905                return ret;
2906
2907        if (convert_out) {
2908                if (copy_in_user(compat_rxnfc, rxnfc,
2909                                 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2910                                 (const void __user *)rxnfc) ||
2911                    copy_in_user(&compat_rxnfc->fs.ring_cookie,
2912                                 &rxnfc->fs.ring_cookie,
2913                                 (const void __user *)(&rxnfc->fs.location + 1) -
2914                                 (const void __user *)&rxnfc->fs.ring_cookie) ||
2915                    copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2916                                 sizeof(rxnfc->rule_cnt)))
2917                        return -EFAULT;
2918
2919                if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2920                        /* As an optimisation, we only copy the actual
2921                         * number of rules that the underlying
2922                         * function returned.  Since Mallory might
2923                         * change the rule count in user memory, we
2924                         * check that it is less than the rule count
2925                         * originally given (as the user buffer size),
2926                         * which has been range-checked.
2927                         */
2928                        if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2929                                return -EFAULT;
2930                        if (actual_rule_cnt < rule_cnt)
2931                                rule_cnt = actual_rule_cnt;
2932                        if (copy_in_user(&compat_rxnfc->rule_locs[0],
2933                                         &rxnfc->rule_locs[0],
2934                                         rule_cnt * sizeof(u32)))
2935                                return -EFAULT;
2936                }
2937        }
2938
2939        return 0;
2940}
2941
2942static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2943{
2944        void __user *uptr;
2945        compat_uptr_t uptr32;
2946        struct ifreq __user *uifr;
2947
2948        uifr = compat_alloc_user_space(sizeof(*uifr));
2949        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2950                return -EFAULT;
2951
2952        if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2953                return -EFAULT;
2954
2955        uptr = compat_ptr(uptr32);
2956
2957        if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2958                return -EFAULT;
2959
2960        return dev_ioctl(net, SIOCWANDEV, uifr);
2961}
2962
2963static int bond_ioctl(struct net *net, unsigned int cmd,
2964                         struct compat_ifreq __user *ifr32)
2965{
2966        struct ifreq kifr;
2967        struct ifreq __user *uifr;
2968        mm_segment_t old_fs;
2969        int err;
2970        u32 data;
2971        void __user *datap;
2972
2973        switch (cmd) {
2974        case SIOCBONDENSLAVE:
2975        case SIOCBONDRELEASE:
2976        case SIOCBONDSETHWADDR:
2977        case SIOCBONDCHANGEACTIVE:
2978                if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2979                        return -EFAULT;
2980
2981                old_fs = get_fs();
2982                set_fs(KERNEL_DS);
2983                err = dev_ioctl(net, cmd,
2984                                (struct ifreq __user __force *) &kifr);
2985                set_fs(old_fs);
2986
2987                return err;
2988        case SIOCBONDSLAVEINFOQUERY:
2989        case SIOCBONDINFOQUERY:
2990                uifr = compat_alloc_user_space(sizeof(*uifr));
2991                if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2992                        return -EFAULT;
2993
2994                if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2995                        return -EFAULT;
2996
2997                datap = compat_ptr(data);
2998                if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2999                        return -EFAULT;
3000
3001                return dev_ioctl(net, cmd, uifr);
3002        default:
3003                return -ENOIOCTLCMD;
3004        }
3005}
3006
3007static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
3008                                 struct compat_ifreq __user *u_ifreq32)
3009{
3010        struct ifreq __user *u_ifreq64;
3011        char tmp_buf[IFNAMSIZ];
3012        void __user *data64;
3013        u32 data32;
3014
3015        if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3016                           IFNAMSIZ))
3017                return -EFAULT;
3018        if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3019                return -EFAULT;
3020        data64 = compat_ptr(data32);
3021
3022        u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3023
3024        /* Don't check these user accesses, just let that get trapped
3025         * in the ioctl handler instead.
3026         */
3027        if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3028                         IFNAMSIZ))
3029                return -EFAULT;
3030        if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3031                return -EFAULT;
3032
3033        return dev_ioctl(net, cmd, u_ifreq64);
3034}
3035
3036static int dev_ifsioc(struct net *net, struct socket *sock,
3037                         unsigned int cmd, struct compat_ifreq __user *uifr32)
3038{
3039        struct ifreq __user *uifr;
3040        int err;
3041
3042        uifr = compat_alloc_user_space(sizeof(*uifr));
3043        if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3044                return -EFAULT;
3045
3046        err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3047
3048        if (!err) {
3049                switch (cmd) {
3050                case SIOCGIFFLAGS:
3051                case SIOCGIFMETRIC:
3052                case SIOCGIFMTU:
3053                case SIOCGIFMEM:
3054                case SIOCGIFHWADDR:
3055                case SIOCGIFINDEX:
3056                case SIOCGIFADDR:
3057                case SIOCGIFBRDADDR:
3058                case SIOCGIFDSTADDR:
3059                case SIOCGIFNETMASK:
3060                case SIOCGIFPFLAGS:
3061                case SIOCGIFTXQLEN:
3062                case SIOCGMIIPHY:
3063                case SIOCGMIIREG:
3064                        if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3065                                err = -EFAULT;
3066                        break;
3067                }
3068        }
3069        return err;
3070}
3071
3072static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3073                        struct compat_ifreq __user *uifr32)
3074{
3075        struct ifreq ifr;
3076        struct compat_ifmap __user *uifmap32;
3077        mm_segment_t old_fs;
3078        int err;
3079
3080        uifmap32 = &uifr32->ifr_ifru.ifru_map;
3081        err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3082        err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3083        err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3084        err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3085        err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3086        err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3087        err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3088        if (err)
3089                return -EFAULT;
3090
3091        old_fs = get_fs();
3092        set_fs(KERNEL_DS);
3093        err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
3094        set_fs(old_fs);
3095
3096        if (cmd == SIOCGIFMAP && !err) {
3097                err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3098                err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3099                err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3100                err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3101                err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3102                err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3103                err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3104                if (err)
3105                        err = -EFAULT;
3106        }
3107        return err;
3108}
3109
3110static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3111{
3112        void __user *uptr;
3113        compat_uptr_t uptr32;
3114        struct ifreq __user *uifr;
3115
3116        uifr = compat_alloc_user_space(sizeof(*uifr));
3117        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3118                return -EFAULT;
3119
3120        if (get_user(uptr32, &uifr32->ifr_data))
3121                return -EFAULT;
3122
3123        uptr = compat_ptr(uptr32);
3124
3125        if (put_user(uptr, &uifr->ifr_data))
3126                return -EFAULT;
3127
3128        return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3129}
3130
3131struct rtentry32 {
3132        u32             rt_pad1;
3133        struct sockaddr rt_dst;         /* target address               */
3134        struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3135        struct sockaddr rt_genmask;     /* target network mask (IP)     */
3136        unsigned short  rt_flags;
3137        short           rt_pad2;
3138        u32             rt_pad3;
3139        unsigned char   rt_tos;
3140        unsigned char   rt_class;
3141        short           rt_pad4;
3142        short           rt_metric;      /* +1 for binary compatibility! */
3143        /* char * */ u32 rt_dev;        /* forcing the device at add    */
3144        u32             rt_mtu;         /* per route MTU/Window         */
3145        u32             rt_window;      /* Window clamping              */
3146        unsigned short  rt_irtt;        /* Initial RTT                  */
3147};
3148
3149struct in6_rtmsg32 {
3150        struct in6_addr         rtmsg_dst;
3151        struct in6_addr         rtmsg_src;
3152        struct in6_addr         rtmsg_gateway;
3153        u32                     rtmsg_type;
3154        u16                     rtmsg_dst_len;
3155        u16                     rtmsg_src_len;
3156        u32                     rtmsg_metric;
3157        u32                     rtmsg_info;
3158        u32                     rtmsg_flags;
3159        s32                     rtmsg_ifindex;
3160};
3161
3162static int routing_ioctl(struct net *net, struct socket *sock,
3163                         unsigned int cmd, void __user *argp)
3164{
3165        int ret;
3166        void *r = NULL;
3167        struct in6_rtmsg r6;
3168        struct rtentry r4;
3169        char devname[16];
3170        u32 rtdev;
3171        mm_segment_t old_fs = get_fs();
3172
3173        if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3174                struct in6_rtmsg32 __user *ur6 = argp;
3175                ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3176                        3 * sizeof(struct in6_addr));
3177                ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3178                ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3179                ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3180                ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3181                ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3182                ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3183                ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3184
3185                r = (void *) &r6;
3186        } else { /* ipv4 */
3187                struct rtentry32 __user *ur4 = argp;
3188                ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3189                                        3 * sizeof(struct sockaddr));
3190                ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3191                ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3192                ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3193                ret |= get_user(r4.rt_window, &(ur4->rt_window));
3194                ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3195                ret |= get_user(rtdev, &(ur4->rt_dev));
3196                if (rtdev) {
3197                        ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3198                        r4.rt_dev = (char __user __force *)devname;
3199                        devname[15] = 0;
3200                } else
3201                        r4.rt_dev = NULL;
3202
3203                r = (void *) &r4;
3204        }
3205
3206        if (ret) {
3207                ret = -EFAULT;
3208                goto out;
3209        }
3210
3211        set_fs(KERNEL_DS);
3212        ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3213        set_fs(old_fs);
3214
3215out:
3216        return ret;
3217}
3218
3219/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3220 * for some operations; this forces use of the newer bridge-utils that
3221 * use compatible ioctls
3222 */
3223static int old_bridge_ioctl(compat_ulong_t __user *argp)
3224{
3225        compat_ulong_t tmp;
3226
3227        if (get_user(tmp, argp))
3228                return -EFAULT;
3229        if (tmp == BRCTL_GET_VERSION)
3230                return BRCTL_VERSION + 1;
3231        return -EINVAL;
3232}
3233
3234static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3235                         unsigned int cmd, unsigned long arg)
3236{
3237        void __user *argp = compat_ptr(arg);
3238        struct sock *sk = sock->sk;
3239        struct net *net = sock_net(sk);
3240
3241        if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3242                return siocdevprivate_ioctl(net, cmd, argp);
3243
3244        switch (cmd) {
3245        case SIOCSIFBR:
3246        case SIOCGIFBR:
3247                return old_bridge_ioctl(argp);
3248        case SIOCGIFNAME:
3249                return dev_ifname32(net, argp);
3250        case SIOCGIFCONF:
3251                return dev_ifconf(net, argp);
3252        case SIOCETHTOOL:
3253                return ethtool_ioctl(net, argp);
3254        case SIOCWANDEV:
3255                return compat_siocwandev(net, argp);
3256        case SIOCGIFMAP:
3257        case SIOCSIFMAP:
3258                return compat_sioc_ifmap(net, cmd, argp);
3259        case SIOCBONDENSLAVE:
3260        case SIOCBONDRELEASE:
3261        case SIOCBONDSETHWADDR:
3262        case SIOCBONDSLAVEINFOQUERY:
3263        case SIOCBONDINFOQUERY:
3264        case SIOCBONDCHANGEACTIVE:
3265                return bond_ioctl(net, cmd, argp);
3266        case SIOCADDRT:
3267        case SIOCDELRT:
3268                return routing_ioctl(net, sock, cmd, argp);
3269        case SIOCGSTAMP:
3270                return do_siocgstamp(net, sock, cmd, argp);
3271        case SIOCGSTAMPNS:
3272                return do_siocgstampns(net, sock, cmd, argp);
3273        case SIOCSHWTSTAMP:
3274                return compat_siocshwtstamp(net, argp);
3275
3276        case FIOSETOWN:
3277        case SIOCSPGRP:
3278        case FIOGETOWN:
3279        case SIOCGPGRP:
3280        case SIOCBRADDBR:
3281        case SIOCBRDELBR:
3282        case SIOCGIFVLAN:
3283        case SIOCSIFVLAN:
3284        case SIOCADDDLCI:
3285        case SIOCDELDLCI:
3286                return sock_ioctl(file, cmd, arg);
3287
3288        case SIOCGIFFLAGS:
3289        case SIOCSIFFLAGS:
3290        case SIOCGIFMETRIC:
3291        case SIOCSIFMETRIC:
3292        case SIOCGIFMTU:
3293        case SIOCSIFMTU:
3294        case SIOCGIFMEM:
3295        case SIOCSIFMEM:
3296        case SIOCGIFHWADDR:
3297        case SIOCSIFHWADDR:
3298        case SIOCADDMULTI:
3299        case SIOCDELMULTI:
3300        case SIOCGIFINDEX:
3301        case SIOCGIFADDR:
3302        case SIOCSIFADDR:
3303        case SIOCSIFHWBROADCAST:
3304        case SIOCDIFADDR:
3305        case SIOCGIFBRDADDR:
3306        case SIOCSIFBRDADDR:
3307        case SIOCGIFDSTADDR:
3308        case SIOCSIFDSTADDR:
3309        case SIOCGIFNETMASK:
3310        case SIOCSIFNETMASK:
3311        case SIOCSIFPFLAGS:
3312        case SIOCGIFPFLAGS:
3313        case SIOCGIFTXQLEN:
3314        case SIOCSIFTXQLEN:
3315        case SIOCBRADDIF:
3316        case SIOCBRDELIF:
3317        case SIOCSIFNAME:
3318        case SIOCGMIIPHY:
3319        case SIOCGMIIREG:
3320        case SIOCSMIIREG:
3321                return dev_ifsioc(net, sock, cmd, argp);
3322
3323        case SIOCSARP:
3324        case SIOCGARP:
3325        case SIOCDARP:
3326        case SIOCATMARK:
3327                return sock_do_ioctl(net, sock, cmd, arg);
3328        }
3329
3330        return -ENOIOCTLCMD;
3331}
3332
3333static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3334                              unsigned long arg)
3335{
3336        struct socket *sock = file->private_data;
3337        int ret = -ENOIOCTLCMD;
3338        struct sock *sk;
3339        struct net *net;
3340
3341        sk = sock->sk;
3342        net = sock_net(sk);
3343
3344        if (sock->ops->compat_ioctl)
3345                ret = sock->ops->compat_ioctl(sock, cmd, arg);
3346
3347        if (ret == -ENOIOCTLCMD &&
3348            (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3349                ret = compat_wext_handle_ioctl(net, cmd, arg);
3350
3351        if (ret == -ENOIOCTLCMD)
3352                ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3353
3354        return ret;
3355}
3356#endif
3357
3358int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3359{
3360        return sock->ops->bind(sock, addr, addrlen);
3361}
3362EXPORT_SYMBOL(kernel_bind);
3363
3364int kernel_listen(struct socket *sock, int backlog)
3365{
3366        return sock->ops->listen(sock, backlog);
3367}
3368EXPORT_SYMBOL(kernel_listen);
3369
3370int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3371{
3372        struct sock *sk = sock->sk;
3373        int err;
3374
3375        err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3376                               newsock);
3377        if (err < 0)
3378                goto done;
3379
3380        err = sock->ops->accept(sock, *newsock, flags);
3381        if (err < 0) {
3382                sock_release(*newsock);
3383                *newsock = NULL;
3384                goto done;
3385        }
3386
3387        (*newsock)->ops = sock->ops;
3388        __module_get((*newsock)->ops->owner);
3389
3390done:
3391        return err;
3392}
3393EXPORT_SYMBOL(kernel_accept);
3394
3395int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3396                   int flags)
3397{
3398        return sock->ops->connect(sock, addr, addrlen, flags);
3399}
3400EXPORT_SYMBOL(kernel_connect);
3401
3402int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3403                         int *addrlen)
3404{
3405        return sock->ops->getname(sock, addr, addrlen, 0);
3406}
3407EXPORT_SYMBOL(kernel_getsockname);
3408
3409int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3410                         int *addrlen)
3411{
3412        return sock->ops->getname(sock, addr, addrlen, 1);
3413}
3414EXPORT_SYMBOL(kernel_getpeername);
3415
3416int kernel_getsockopt(struct socket *sock, int level, int optname,
3417                        char *optval, int *optlen)
3418{
3419        mm_segment_t oldfs = get_fs();
3420        char __user *uoptval;
3421        int __user *uoptlen;
3422        int err;
3423
3424        uoptval = (char __user __force *) optval;
3425        uoptlen = (int __user __force *) optlen;
3426
3427        set_fs(KERNEL_DS);
3428        if (level == SOL_SOCKET)
3429                err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3430        else
3431                err = sock->ops->getsockopt(sock, level, optname, uoptval,
3432                                            uoptlen);
3433        set_fs(oldfs);
3434        return err;
3435}
3436EXPORT_SYMBOL(kernel_getsockopt);
3437
3438int kernel_setsockopt(struct socket *sock, int level, int optname,
3439                        char *optval, unsigned int optlen)
3440{
3441        mm_segment_t oldfs = get_fs();
3442        char __user *uoptval;
3443        int err;
3444
3445        uoptval = (char __user __force *) optval;
3446
3447        set_fs(KERNEL_DS);
3448        if (level == SOL_SOCKET)
3449                err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3450        else
3451                err = sock->ops->setsockopt(sock, level, optname, uoptval,
3452                                            optlen);
3453        set_fs(oldfs);
3454        return err;
3455}
3456EXPORT_SYMBOL(kernel_setsockopt);
3457
3458int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3459                    size_t size, int flags)
3460{
3461        if (sock->ops->sendpage)
3462                return sock->ops->sendpage(sock, page, offset, size, flags);
3463
3464        return sock_no_sendpage(sock, page, offset, size, flags);
3465}
3466EXPORT_SYMBOL(kernel_sendpage);
3467
3468int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3469{
3470        mm_segment_t oldfs = get_fs();
3471        int err;
3472
3473        set_fs(KERNEL_DS);
3474        err = sock->ops->ioctl(sock, cmd, arg);
3475        set_fs(oldfs);
3476
3477        return err;
3478}
3479EXPORT_SYMBOL(kernel_sock_ioctl);
3480
3481int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3482{
3483        return sock->ops->shutdown(sock, how);
3484}
3485EXPORT_SYMBOL(kernel_sock_shutdown);
3486
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.