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        BUG_ON(klen > sizeof(struct sockaddr_storage));
 225        err = get_user(len, ulen);
 226        if (err)
 227                return err;
 228        if (len > klen)
 229                len = klen;
 230        if (len < 0)
 231                return -EINVAL;
 232        if (len) {
 233                if (audit_sockaddr(klen, kaddr))
 234                        return -ENOMEM;
 235                if (copy_to_user(uaddr, kaddr, len))
 236                        return -EFAULT;
 237        }
 238        /*
 239         *      "fromlen shall refer to the value before truncation.."
 240         *                      1003.1g
 241         */
 242        return __put_user(klen, ulen);
 243}
 244
 245static struct kmem_cache *sock_inode_cachep __read_mostly;
 246
 247static struct inode *sock_alloc_inode(struct super_block *sb)
 248{
 249        struct socket_alloc *ei;
 250        struct socket_wq *wq;
 251
 252        ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
 253        if (!ei)
 254                return NULL;
 255        wq = kmalloc(sizeof(*wq), GFP_KERNEL);
 256        if (!wq) {
 257                kmem_cache_free(sock_inode_cachep, ei);
 258                return NULL;
 259        }
 260        init_waitqueue_head(&wq->wait);
 261        wq->fasync_list = NULL;
 262        RCU_INIT_POINTER(ei->socket.wq, wq);
 263
 264        ei->socket.state = SS_UNCONNECTED;
 265        ei->socket.flags = 0;
 266        ei->socket.ops = NULL;
 267        ei->socket.sk = NULL;
 268        ei->socket.file = NULL;
 269
 270        return &ei->vfs_inode;
 271}
 272
 273static void sock_destroy_inode(struct inode *inode)
 274{
 275        struct socket_alloc *ei;
 276        struct socket_wq *wq;
 277
 278        ei = container_of(inode, struct socket_alloc, vfs_inode);
 279        wq = rcu_dereference_protected(ei->socket.wq, 1);
 280        kfree_rcu(wq, rcu);
 281        kmem_cache_free(sock_inode_cachep, ei);
 282}
 283
 284static void init_once(void *foo)
 285{
 286        struct socket_alloc *ei = (struct socket_alloc *)foo;
 287
 288        inode_init_once(&ei->vfs_inode);
 289}
 290
 291static int init_inodecache(void)
 292{
 293        sock_inode_cachep = kmem_cache_create("sock_inode_cache",
 294                                              sizeof(struct socket_alloc),
 295                                              0,
 296                                              (SLAB_HWCACHE_ALIGN |
 297                                               SLAB_RECLAIM_ACCOUNT |
 298                                               SLAB_MEM_SPREAD),
 299                                              init_once);
 300        if (sock_inode_cachep == NULL)
 301                return -ENOMEM;
 302        return 0;
 303}
 304
 305static const struct super_operations sockfs_ops = {
 306        .alloc_inode    = sock_alloc_inode,
 307        .destroy_inode  = sock_destroy_inode,
 308        .statfs         = simple_statfs,
 309};
 310
 311/*
 312 * sockfs_dname() is called from d_path().
 313 */
 314static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
 315{
 316        return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
 317                                dentry->d_inode->i_ino);
 318}
 319
 320static const struct dentry_operations sockfs_dentry_operations = {
 321        .d_dname  = sockfs_dname,
 322};
 323
 324static struct dentry *sockfs_mount(struct file_system_type *fs_type,
 325                         int flags, const char *dev_name, void *data)
 326{
 327        return mount_pseudo(fs_type, "socket:", &sockfs_ops,
 328                &sockfs_dentry_operations, SOCKFS_MAGIC);
 329}
 330
 331static struct vfsmount *sock_mnt __read_mostly;
 332
 333static struct file_system_type sock_fs_type = {
 334        .name =         "sockfs",
 335        .mount =        sockfs_mount,
 336        .kill_sb =      kill_anon_super,
 337};
 338
 339/*
 340 *      Obtains the first available file descriptor and sets it up for use.
 341 *
 342 *      These functions create file structures and maps them to fd space
 343 *      of the current process. On success it returns file descriptor
 344 *      and file struct implicitly stored in sock->file.
 345 *      Note that another thread may close file descriptor before we return
 346 *      from this function. We use the fact that now we do not refer
 347 *      to socket after mapping. If one day we will need it, this
 348 *      function will increment ref. count on file by 1.
 349 *
 350 *      In any case returned fd MAY BE not valid!
 351 *      This race condition is unavoidable
 352 *      with shared fd spaces, we cannot solve it inside kernel,
 353 *      but we take care of internal coherence yet.
 354 */
 355
 356struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
 357{
 358        struct qstr name = { .name = "" };
 359        struct path path;
 360        struct file *file;
 361
 362        if (dname) {
 363                name.name = dname;
 364                name.len = strlen(name.name);
 365        } else if (sock->sk) {
 366                name.name = sock->sk->sk_prot_creator->name;
 367                name.len = strlen(name.name);
 368        }
 369        path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
 370        if (unlikely(!path.dentry))
 371                return ERR_PTR(-ENOMEM);
 372        path.mnt = mntget(sock_mnt);
 373
 374        d_instantiate(path.dentry, SOCK_INODE(sock));
 375        SOCK_INODE(sock)->i_fop = &socket_file_ops;
 376
 377        file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
 378                  &socket_file_ops);
 379        if (unlikely(IS_ERR(file))) {
 380                /* drop dentry, keep inode */
 381                ihold(path.dentry->d_inode);
 382                path_put(&path);
 383                return file;
 384        }
 385
 386        sock->file = file;
 387        file->f_flags = O_RDWR | (flags & O_NONBLOCK);
 388        file->private_data = sock;
 389        return file;
 390}
 391EXPORT_SYMBOL(sock_alloc_file);
 392
 393static int sock_map_fd(struct socket *sock, int flags)
 394{
 395        struct file *newfile;
 396        int fd = get_unused_fd_flags(flags);
 397        if (unlikely(fd < 0))
 398                return fd;
 399
 400        newfile = sock_alloc_file(sock, flags, NULL);
 401        if (likely(!IS_ERR(newfile))) {
 402                fd_install(fd, newfile);
 403                return fd;
 404        }
 405
 406        put_unused_fd(fd);
 407        return PTR_ERR(newfile);
 408}
 409
 410struct socket *sock_from_file(struct file *file, int *err)
 411{
 412        if (file->f_op == &socket_file_ops)
 413                return file->private_data;      /* set in sock_map_fd */
 414
 415        *err = -ENOTSOCK;
 416        return NULL;
 417}
 418EXPORT_SYMBOL(sock_from_file);
 419
 420/**
 421 *      sockfd_lookup - Go from a file number to its socket slot
 422 *      @fd: file handle
 423 *      @err: pointer to an error code return
 424 *
 425 *      The file handle passed in is locked and the socket it is bound
 426 *      too is returned. If an error occurs the err pointer is overwritten
 427 *      with a negative errno code and NULL is returned. The function checks
 428 *      for both invalid handles and passing a handle which is not a socket.
 429 *
 430 *      On a success the socket object pointer is returned.
 431 */
 432
 433struct socket *sockfd_lookup(int fd, int *err)
 434{
 435        struct file *file;
 436        struct socket *sock;
 437
 438        file = fget(fd);
 439        if (!file) {
 440                *err = -EBADF;
 441                return NULL;
 442        }
 443
 444        sock = sock_from_file(file, err);
 445        if (!sock)
 446                fput(file);
 447        return sock;
 448}
 449EXPORT_SYMBOL(sockfd_lookup);
 450
 451static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 452{
 453        struct fd f = fdget(fd);
 454        struct socket *sock;
 455
 456        *err = -EBADF;
 457        if (f.file) {
 458                sock = sock_from_file(f.file, err);
 459                if (likely(sock)) {
 460                        *fput_needed = f.flags;
 461                        return sock;
 462                }
 463                fdput(f);
 464        }
 465        return NULL;
 466}
 467
 468#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
 469#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
 470#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
 471static ssize_t sockfs_getxattr(struct dentry *dentry,
 472                               const char *name, void *value, size_t size)
 473{
 474        const char *proto_name;
 475        size_t proto_size;
 476        int error;
 477
 478        error = -ENODATA;
 479        if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
 480                proto_name = dentry->d_name.name;
 481                proto_size = strlen(proto_name);
 482
 483                if (value) {
 484                        error = -ERANGE;
 485                        if (proto_size + 1 > size)
 486                                goto out;
 487
 488                        strncpy(value, proto_name, proto_size + 1);
 489                }
 490                error = proto_size + 1;
 491        }
 492
 493out:
 494        return error;
 495}
 496
 497static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
 498                                size_t size)
 499{
 500        ssize_t len;
 501        ssize_t used = 0;
 502
 503        len = security_inode_listsecurity(dentry->d_inode, buffer, size);
 504        if (len < 0)
 505                return len;
 506        used += len;
 507        if (buffer) {
 508                if (size < used)
 509                        return -ERANGE;
 510                buffer += len;
 511        }
 512
 513        len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
 514        used += len;
 515        if (buffer) {
 516                if (size < used)
 517                        return -ERANGE;
 518                memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
 519                buffer += len;
 520        }
 521
 522        return used;
 523}
 524
 525static const struct inode_operations sockfs_inode_ops = {
 526        .getxattr = sockfs_getxattr,
 527        .listxattr = sockfs_listxattr,
 528};
 529
 530/**
 531 *      sock_alloc      -       allocate a socket
 532 *
 533 *      Allocate a new inode and socket object. The two are bound together
 534 *      and initialised. The socket is then returned. If we are out of inodes
 535 *      NULL is returned.
 536 */
 537
 538static struct socket *sock_alloc(void)
 539{
 540        struct inode *inode;
 541        struct socket *sock;
 542
 543        inode = new_inode_pseudo(sock_mnt->mnt_sb);
 544        if (!inode)
 545                return NULL;
 546
 547        sock = SOCKET_I(inode);
 548
 549        kmemcheck_annotate_bitfield(sock, type);
 550        inode->i_ino = get_next_ino();
 551        inode->i_mode = S_IFSOCK | S_IRWXUGO;
 552        inode->i_uid = current_fsuid();
 553        inode->i_gid = current_fsgid();
 554        inode->i_op = &sockfs_inode_ops;
 555
 556        this_cpu_add(sockets_in_use, 1);
 557        return sock;
 558}
 559
 560/*
 561 *      In theory you can't get an open on this inode, but /proc provides
 562 *      a back door. Remember to keep it shut otherwise you'll let the
 563 *      creepy crawlies in.
 564 */
 565
 566static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
 567{
 568        return -ENXIO;
 569}
 570
 571const struct file_operations bad_sock_fops = {
 572        .owner = THIS_MODULE,
 573        .open = sock_no_open,
 574        .llseek = noop_llseek,
 575};
 576
 577/**
 578 *      sock_release    -       close a socket
 579 *      @sock: socket to close
 580 *
 581 *      The socket is released from the protocol stack if it has a release
 582 *      callback, and the inode is then released if the socket is bound to
 583 *      an inode not a file.
 584 */
 585
 586void sock_release(struct socket *sock)
 587{
 588        if (sock->ops) {
 589                struct module *owner = sock->ops->owner;
 590
 591                sock->ops->release(sock);
 592                sock->ops = NULL;
 593                module_put(owner);
 594        }
 595
 596        if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
 597                printk(KERN_ERR "sock_release: fasync list not empty!\n");
 598
 599        if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
 600                return;
 601
 602        this_cpu_sub(sockets_in_use, 1);
 603        if (!sock->file) {
 604                iput(SOCK_INODE(sock));
 605                return;
 606        }
 607        sock->file = NULL;
 608}
 609EXPORT_SYMBOL(sock_release);
 610
 611void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
 612{
 613        *tx_flags = 0;
 614        if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
 615                *tx_flags |= SKBTX_HW_TSTAMP;
 616        if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
 617                *tx_flags |= SKBTX_SW_TSTAMP;
 618        if (sock_flag(sk, SOCK_WIFI_STATUS))
 619                *tx_flags |= SKBTX_WIFI_STATUS;
 620}
 621EXPORT_SYMBOL(sock_tx_timestamp);
 622
 623static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
 624                                       struct msghdr *msg, size_t size)
 625{
 626        struct sock_iocb *si = kiocb_to_siocb(iocb);
 627
 628        si->sock = sock;
 629        si->scm = NULL;
 630        si->msg = msg;
 631        si->size = size;
 632
 633        return sock->ops->sendmsg(iocb, sock, msg, size);
 634}
 635
 636static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 637                                 struct msghdr *msg, size_t size)
 638{
 639        int err = security_socket_sendmsg(sock, msg, size);
 640
 641        return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
 642}
 643
 644int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 645{
 646        struct kiocb iocb;
 647        struct sock_iocb siocb;
 648        int ret;
 649
 650        init_sync_kiocb(&iocb, NULL);
 651        iocb.private = &siocb;
 652        ret = __sock_sendmsg(&iocb, sock, msg, size);
 653        if (-EIOCBQUEUED == ret)
 654                ret = wait_on_sync_kiocb(&iocb);
 655        return ret;
 656}
 657EXPORT_SYMBOL(sock_sendmsg);
 658
 659static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
 660{
 661        struct kiocb iocb;
 662        struct sock_iocb siocb;
 663        int ret;
 664
 665        init_sync_kiocb(&iocb, NULL);
 666        iocb.private = &siocb;
 667        ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
 668        if (-EIOCBQUEUED == ret)
 669                ret = wait_on_sync_kiocb(&iocb);
 670        return ret;
 671}
 672
 673int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 674                   struct kvec *vec, size_t num, size_t size)
 675{
 676        mm_segment_t oldfs = get_fs();
 677        int result;
 678
 679        set_fs(KERNEL_DS);
 680        /*
 681         * the following is safe, since for compiler definitions of kvec and
 682         * iovec are identical, yielding the same in-core layout and alignment
 683         */
 684        msg->msg_iov = (struct iovec *)vec;
 685        msg->msg_iovlen = num;
 686        result = sock_sendmsg(sock, msg, size);
 687        set_fs(oldfs);
 688        return result;
 689}
 690EXPORT_SYMBOL(kernel_sendmsg);
 691
 692/*
 693 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 694 */
 695void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 696        struct sk_buff *skb)
 697{
 698        int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 699        struct timespec ts[3];
 700        int empty = 1;
 701        struct skb_shared_hwtstamps *shhwtstamps =
 702                skb_hwtstamps(skb);
 703
 704        /* Race occurred between timestamp enabling and packet
 705           receiving.  Fill in the current time for now. */
 706        if (need_software_tstamp && skb->tstamp.tv64 == 0)
 707                __net_timestamp(skb);
 708
 709        if (need_software_tstamp) {
 710                if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 711                        struct timeval tv;
 712                        skb_get_timestamp(skb, &tv);
 713                        put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 714                                 sizeof(tv), &tv);
 715                } else {
 716                        skb_get_timestampns(skb, &ts[0]);
 717                        put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 718                                 sizeof(ts[0]), &ts[0]);
 719                }
 720        }
 721
 722
 723        memset(ts, 0, sizeof(ts));
 724        if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
 725            ktime_to_timespec_cond(skb->tstamp, ts + 0))
 726                empty = 0;
 727        if (shhwtstamps) {
 728                if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 729                    ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
 730                        empty = 0;
 731                if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 732                    ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
 733                        empty = 0;
 734        }
 735        if (!empty)
 736                put_cmsg(msg, SOL_SOCKET,
 737                         SCM_TIMESTAMPING, sizeof(ts), &ts);
 738}
 739EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 740
 741void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
 742        struct sk_buff *skb)
 743{
 744        int ack;
 745
 746        if (!sock_flag(sk, SOCK_WIFI_STATUS))
 747                return;
 748        if (!skb->wifi_acked_valid)
 749                return;
 750
 751        ack = skb->wifi_acked;
 752
 753        put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
 754}
 755EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
 756
 757static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 758                                   struct sk_buff *skb)
 759{
 760        if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 761                put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 762                        sizeof(__u32), &skb->dropcount);
 763}
 764
 765void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 766        struct sk_buff *skb)
 767{
 768        sock_recv_timestamp(msg, sk, skb);
 769        sock_recv_drops(msg, sk, skb);
 770}
 771EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 772
 773static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 774                                       struct msghdr *msg, size_t size, int flags)
 775{
 776        struct sock_iocb *si = kiocb_to_siocb(iocb);
 777
 778        si->sock = sock;
 779        si->scm = NULL;
 780        si->msg = msg;
 781        si->size = size;
 782        si->flags = flags;
 783
 784        return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 785}
 786
 787static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 788                                 struct msghdr *msg, size_t size, int flags)
 789{
 790        int err = security_socket_recvmsg(sock, msg, size, flags);
 791
 792        return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 793}
 794
 795int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 796                 size_t size, int flags)
 797{
 798        struct kiocb iocb;
 799        struct sock_iocb siocb;
 800        int ret;
 801
 802        init_sync_kiocb(&iocb, NULL);
 803        iocb.private = &siocb;
 804        ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 805        if (-EIOCBQUEUED == ret)
 806                ret = wait_on_sync_kiocb(&iocb);
 807        return ret;
 808}
 809EXPORT_SYMBOL(sock_recvmsg);
 810
 811static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 812                              size_t size, int flags)
 813{
 814        struct kiocb iocb;
 815        struct sock_iocb siocb;
 816        int ret;
 817
 818        init_sync_kiocb(&iocb, NULL);
 819        iocb.private = &siocb;
 820        ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 821        if (-EIOCBQUEUED == ret)
 822                ret = wait_on_sync_kiocb(&iocb);
 823        return ret;
 824}
 825
 826/**
 827 * kernel_recvmsg - Receive a message from a socket (kernel space)
 828 * @sock:       The socket to receive the message from
 829 * @msg:        Received message
 830 * @vec:        Input s/g array for message data
 831 * @num:        Size of input s/g array
 832 * @size:       Number of bytes to read
 833 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 834 *
 835 * On return the msg structure contains the scatter/gather array passed in the
 836 * vec argument. The array is modified so that it consists of the unfilled
 837 * portion of the original array.
 838 *
 839 * The returned value is the total number of bytes received, or an error.
 840 */
 841int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 842                   struct kvec *vec, size_t num, size_t size, int flags)
 843{
 844        mm_segment_t oldfs = get_fs();
 845        int result;
 846
 847        set_fs(KERNEL_DS);
 848        /*
 849         * the following is safe, since for compiler definitions of kvec and
 850         * iovec are identical, yielding the same in-core layout and alignment
 851         */
 852        msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 853        result = sock_recvmsg(sock, msg, size, flags);
 854        set_fs(oldfs);
 855        return result;
 856}
 857EXPORT_SYMBOL(kernel_recvmsg);
 858
 859static ssize_t sock_sendpage(struct file *file, struct page *page,
 860                             int offset, size_t size, loff_t *ppos, int more)
 861{
 862        struct socket *sock;
 863        int flags;
 864
 865        sock = file->private_data;
 866
 867        flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 868        /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
 869        flags |= more;
 870
 871        return kernel_sendpage(sock, page, offset, size, flags);
 872}
 873
 874static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 875                                struct pipe_inode_info *pipe, size_t len,
 876                                unsigned int flags)
 877{
 878        struct socket *sock = file->private_data;
 879
 880        if (unlikely(!sock->ops->splice_read))
 881                return -EINVAL;
 882
 883        return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 884}
 885
 886static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 887                                         struct sock_iocb *siocb)
 888{
 889        if (!is_sync_kiocb(iocb))
 890                BUG();
 891
 892        siocb->kiocb = iocb;
 893        iocb->private = siocb;
 894        return siocb;
 895}
 896
 897static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 898                struct file *file, const struct iovec *iov,
 899                unsigned long nr_segs)
 900{
 901        struct socket *sock = file->private_data;
 902        size_t size = 0;
 903        int i;
 904
 905        for (i = 0; i < nr_segs; i++)
 906                size += iov[i].iov_len;
 907
 908        msg->msg_name = NULL;
 909        msg->msg_namelen = 0;
 910        msg->msg_control = NULL;
 911        msg->msg_controllen = 0;
 912        msg->msg_iov = (struct iovec *)iov;
 913        msg->msg_iovlen = nr_segs;
 914        msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 915
 916        return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 917}
 918
 919static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 920                                unsigned long nr_segs, loff_t pos)
 921{
 922        struct sock_iocb siocb, *x;
 923
 924        if (pos != 0)
 925                return -ESPIPE;
 926
 927        if (iocb->ki_nbytes == 0)       /* Match SYS5 behaviour */
 928                return 0;
 929
 930
 931        x = alloc_sock_iocb(iocb, &siocb);
 932        if (!x)
 933                return -ENOMEM;
 934        return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 935}
 936
 937static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 938                        struct file *file, const struct iovec *iov,
 939                        unsigned long nr_segs)
 940{
 941        struct socket *sock = file->private_data;
 942        size_t size = 0;
 943        int i;
 944
 945        for (i = 0; i < nr_segs; i++)
 946                size += iov[i].iov_len;
 947
 948        msg->msg_name = NULL;
 949        msg->msg_namelen = 0;
 950        msg->msg_control = NULL;
 951        msg->msg_controllen = 0;
 952        msg->msg_iov = (struct iovec *)iov;
 953        msg->msg_iovlen = nr_segs;
 954        msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 955        if (sock->type == SOCK_SEQPACKET)
 956                msg->msg_flags |= MSG_EOR;
 957
 958        return __sock_sendmsg(iocb, sock, msg, size);
 959}
 960
 961static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 962                          unsigned long nr_segs, loff_t pos)
 963{
 964        struct sock_iocb siocb, *x;
 965
 966        if (pos != 0)
 967                return -ESPIPE;
 968
 969        x = alloc_sock_iocb(iocb, &siocb);
 970        if (!x)
 971                return -ENOMEM;
 972
 973        return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 974}
 975
 976/*
 977 * Atomic setting of ioctl hooks to avoid race
 978 * with module unload.
 979 */
 980
 981static DEFINE_MUTEX(br_ioctl_mutex);
 982static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 983
 984void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 985{
 986        mutex_lock(&br_ioctl_mutex);
 987        br_ioctl_hook = hook;
 988        mutex_unlock(&br_ioctl_mutex);
 989}
 990EXPORT_SYMBOL(brioctl_set);
 991
 992static DEFINE_MUTEX(vlan_ioctl_mutex);
 993static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 994
 995void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 996{
 997        mutex_lock(&vlan_ioctl_mutex);
 998        vlan_ioctl_hook = hook;
 999        mutex_unlock(&vlan_ioctl_mutex);
1000}
1001EXPORT_SYMBOL(vlan_ioctl_set);
1002
1003static DEFINE_MUTEX(dlci_ioctl_mutex);
1004static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1005
1006void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1007{
1008        mutex_lock(&dlci_ioctl_mutex);
1009        dlci_ioctl_hook = hook;
1010        mutex_unlock(&dlci_ioctl_mutex);
1011}
1012EXPORT_SYMBOL(dlci_ioctl_set);
1013
1014static long sock_do_ioctl(struct net *net, struct socket *sock,
1015                                 unsigned int cmd, unsigned long arg)
1016{
1017        int err;
1018        void __user *argp = (void __user *)arg;
1019
1020        err = sock->ops->ioctl(sock, cmd, arg);
1021
1022        /*
1023         * If this ioctl is unknown try to hand it down
1024         * to the NIC driver.
1025         */
1026        if (err == -ENOIOCTLCMD)
1027                err = dev_ioctl(net, cmd, argp);
1028
1029        return err;
1030}
1031
1032/*
1033 *      With an ioctl, arg may well be a user mode pointer, but we don't know
1034 *      what to do with it - that's up to the protocol still.
1035 */
1036
1037static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1038{
1039        struct socket *sock;
1040        struct sock *sk;
1041        void __user *argp = (void __user *)arg;
1042        int pid, err;
1043        struct net *net;
1044
1045        sock = file->private_data;
1046        sk = sock->sk;
1047        net = sock_net(sk);
1048        if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1049                err = dev_ioctl(net, cmd, argp);
1050        } else
1051#ifdef CONFIG_WEXT_CORE
1052        if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1053                err = dev_ioctl(net, cmd, argp);
1054        } else
1055#endif
1056                switch (cmd) {
1057                case FIOSETOWN:
1058                case SIOCSPGRP:
1059                        err = -EFAULT;
1060                        if (get_user(pid, (int __user *)argp))
1061                                break;
1062                        err = f_setown(sock->file, pid, 1);
1063                        break;
1064                case FIOGETOWN:
1065                case SIOCGPGRP:
1066                        err = put_user(f_getown(sock->file),
1067                                       (int __user *)argp);
1068                        break;
1069                case SIOCGIFBR:
1070                case SIOCSIFBR:
1071                case SIOCBRADDBR:
1072                case SIOCBRDELBR:
1073                        err = -ENOPKG;
1074                        if (!br_ioctl_hook)
1075                                request_module("bridge");
1076
1077                        mutex_lock(&br_ioctl_mutex);
1078                        if (br_ioctl_hook)
1079                                err = br_ioctl_hook(net, cmd, argp);
1080                        mutex_unlock(&br_ioctl_mutex);
1081                        break;
1082                case SIOCGIFVLAN:
1083                case SIOCSIFVLAN:
1084                        err = -ENOPKG;
1085                        if (!vlan_ioctl_hook)
1086                                request_module("8021q");
1087
1088                        mutex_lock(&vlan_ioctl_mutex);
1089                        if (vlan_ioctl_hook)
1090                                err = vlan_ioctl_hook(net, argp);
1091                        mutex_unlock(&vlan_ioctl_mutex);
1092                        break;
1093                case SIOCADDDLCI:
1094                case SIOCDELDLCI:
1095                        err = -ENOPKG;
1096                        if (!dlci_ioctl_hook)
1097                                request_module("dlci");
1098
1099                        mutex_lock(&dlci_ioctl_mutex);
1100                        if (dlci_ioctl_hook)
1101                                err = dlci_ioctl_hook(cmd, argp);
1102                        mutex_unlock(&dlci_ioctl_mutex);
1103                        break;
1104                default:
1105                        err = sock_do_ioctl(net, sock, cmd, arg);
1106                        break;
1107                }
1108        return err;
1109}
1110
1111int sock_create_lite(int family, int type, int protocol, struct socket **res)
1112{
1113        int err;
1114        struct socket *sock = NULL;
1115
1116        err = security_socket_create(family, type, protocol, 1);
1117        if (err)
1118                goto out;
1119
1120        sock = sock_alloc();
1121        if (!sock) {
1122                err = -ENOMEM;
1123                goto out;
1124        }
1125
1126        sock->type = type;
1127        err = security_socket_post_create(sock, family, type, protocol, 1);
1128        if (err)
1129                goto out_release;
1130
1131out:
1132        *res = sock;
1133        return err;
1134out_release:
1135        sock_release(sock);
1136        sock = NULL;
1137        goto out;
1138}
1139EXPORT_SYMBOL(sock_create_lite);
1140
1141/* No kernel lock held - perfect */
1142static unsigned int sock_poll(struct file *file, poll_table *wait)
1143{
1144        unsigned int busy_flag = 0;
1145        struct socket *sock;
1146
1147        /*
1148         *      We can't return errors to poll, so it's either yes or no.
1149         */
1150        sock = file->private_data;
1151
1152        if (sk_can_busy_loop(sock->sk)) {
1153                /* this socket can poll_ll so tell the system call */
1154                busy_flag = POLL_BUSY_LOOP;
1155
1156                /* once, only if requested by syscall */
1157                if (wait && (wait->_key & POLL_BUSY_LOOP))
1158                        sk_busy_loop(sock->sk, 1);
1159        }
1160
1161        return busy_flag | sock->ops->poll(file, sock, wait);
1162}
1163
1164static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1165{
1166        struct socket *sock = file->private_data;
1167
1168        return sock->ops->mmap(file, sock, vma);
1169}
1170
1171static int sock_close(struct inode *inode, struct file *filp)
1172{
1173        sock_release(SOCKET_I(inode));
1174        return 0;
1175}
1176
1177/*
1178 *      Update the socket async list
1179 *
1180 *      Fasync_list locking strategy.
1181 *
1182 *      1. fasync_list is modified only under process context socket lock
1183 *         i.e. under semaphore.
1184 *      2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1185 *         or under socket lock
1186 */
1187
1188static int sock_fasync(int fd, struct file *filp, int on)
1189{
1190        struct socket *sock = filp->private_data;
1191        struct sock *sk = sock->sk;
1192        struct socket_wq *wq;
1193
1194        if (sk == NULL)
1195                return -EINVAL;
1196
1197        lock_sock(sk);
1198        wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1199        fasync_helper(fd, filp, on, &wq->fasync_list);
1200
1201        if (!wq->fasync_list)
1202                sock_reset_flag(sk, SOCK_FASYNC);
1203        else
1204                sock_set_flag(sk, SOCK_FASYNC);
1205
1206        release_sock(sk);
1207        return 0;
1208}
1209
1210/* This function may be called only under socket lock or callback_lock or rcu_lock */
1211
1212int sock_wake_async(struct socket *sock, int how, int band)
1213{
1214        struct socket_wq *wq;
1215
1216        if (!sock)
1217                return -1;
1218        rcu_read_lock();
1219        wq = rcu_dereference(sock->wq);
1220        if (!wq || !wq->fasync_list) {
1221                rcu_read_unlock();
1222                return -1;
1223        }
1224        switch (how) {
1225        case SOCK_WAKE_WAITD:
1226                if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1227                        break;
1228                goto call_kill;
1229        case SOCK_WAKE_SPACE:
1230                if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1231                        break;
1232                /* fall through */
1233        case SOCK_WAKE_IO:
1234call_kill:
1235                kill_fasync(&wq->fasync_list, SIGIO, band);
1236                break;
1237        case SOCK_WAKE_URG:
1238                kill_fasync(&wq->fasync_list, SIGURG, band);
1239        }
1240        rcu_read_unlock();
1241        return 0;
1242}
1243EXPORT_SYMBOL(sock_wake_async);
1244
1245int __sock_create(struct net *net, int family, int type, int protocol,
1246                         struct socket **res, int kern)
1247{
1248        int err;
1249        struct socket *sock;
1250        const struct net_proto_family *pf;
1251
1252        /*
1253         *      Check protocol is in range
1254         */
1255        if (family < 0 || family >= NPROTO)
1256                return -EAFNOSUPPORT;
1257        if (type < 0 || type >= SOCK_MAX)
1258                return -EINVAL;
1259
1260        /* Compatibility.
1261
1262           This uglymoron is moved from INET layer to here to avoid
1263           deadlock in module load.
1264         */
1265        if (family == PF_INET && type == SOCK_PACKET) {
1266                static int warned;
1267                if (!warned) {
1268                        warned = 1;
1269                        printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1270                               current->comm);
1271                }
1272                family = PF_PACKET;
1273        }
1274
1275        err = security_socket_create(family, type, protocol, kern);
1276        if (err)
1277                return err;
1278
1279        /*
1280         *      Allocate the socket and allow the family to set things up. if
1281         *      the protocol is 0, the family is instructed to select an appropriate
1282         *      default.
1283         */
1284        sock = sock_alloc();
1285        if (!sock) {
1286                net_warn_ratelimited("socket: no more sockets\n");
1287                return -ENFILE; /* Not exactly a match, but its the
1288                                   closest posix thing */
1289        }
1290
1291        sock->type = type;
1292
1293#ifdef CONFIG_MODULES
1294        /* Attempt to load a protocol module if the find failed.
1295         *
1296         * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1297         * requested real, full-featured networking support upon configuration.
1298         * Otherwise module support will break!
1299         */
1300        if (rcu_access_pointer(net_families[family]) == NULL)
1301                request_module("net-pf-%d", family);
1302#endif
1303
1304        rcu_read_lock();
1305        pf = rcu_dereference(net_families[family]);
1306        err = -EAFNOSUPPORT;
1307        if (!pf)
1308                goto out_release;
1309
1310        /*
1311         * We will call the ->create function, that possibly is in a loadable
1312         * module, so we have to bump that loadable module refcnt first.
1313         */
1314        if (!try_module_get(pf->owner))
1315                goto out_release;
1316
1317        /* Now protected by module ref count */
1318        rcu_read_unlock();
1319
1320        err = pf->create(net, sock, protocol, kern);
1321        if (err < 0)
1322                goto out_module_put;
1323
1324        /*
1325         * Now to bump the refcnt of the [loadable] module that owns this
1326         * socket at sock_release time we decrement its refcnt.
1327         */
1328        if (!try_module_get(sock->ops->owner))
1329                goto out_module_busy;
1330
1331        /*
1332         * Now that we're done with the ->create function, the [loadable]
1333         * module can have its refcnt decremented
1334         */
1335        module_put(pf->owner);
1336        err = security_socket_post_create(sock, family, type, protocol, kern);
1337        if (err)
1338                goto out_sock_release;
1339        *res = sock;
1340
1341        return 0;
1342
1343out_module_busy:
1344        err = -EAFNOSUPPORT;
1345out_module_put:
1346        sock->ops = NULL;
1347        module_put(pf->owner);
1348out_sock_release:
1349        sock_release(sock);
1350        return err;
1351
1352out_release:
1353        rcu_read_unlock();
1354        goto out_sock_release;
1355}
1356EXPORT_SYMBOL(__sock_create);
1357
1358int sock_create(int family, int type, int protocol, struct socket **res)
1359{
1360        return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1361}
1362EXPORT_SYMBOL(sock_create);
1363
1364int sock_create_kern(int family, int type, int protocol, struct socket **res)
1365{
1366        return __sock_create(&init_net, family, type, protocol, res, 1);
1367}
1368EXPORT_SYMBOL(sock_create_kern);
1369
1370SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1371{
1372        int retval;
1373        struct socket *sock;
1374        int flags;
1375
1376        /* Check the SOCK_* constants for consistency.  */
1377        BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1378        BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1379        BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1380        BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1381
1382        flags = type & ~SOCK_TYPE_MASK;
1383        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1384                return -EINVAL;
1385        type &= SOCK_TYPE_MASK;
1386
1387        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1388                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1389
1390        retval = sock_create(family, type, protocol, &sock);
1391        if (retval < 0)
1392                goto out;
1393
1394        retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1395        if (retval < 0)
1396                goto out_release;
1397
1398out:
1399        /* It may be already another descriptor 8) Not kernel problem. */
1400        return retval;
1401
1402out_release:
1403        sock_release(sock);
1404        return retval;
1405}
1406
1407/*
1408 *      Create a pair of connected sockets.
1409 */
1410
1411SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1412                int __user *, usockvec)
1413{
1414        struct socket *sock1, *sock2;
1415        int fd1, fd2, err;
1416        struct file *newfile1, *newfile2;
1417        int flags;
1418
1419        flags = type & ~SOCK_TYPE_MASK;
1420        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1421                return -EINVAL;
1422        type &= SOCK_TYPE_MASK;
1423
1424        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1425                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1426
1427        /*
1428         * Obtain the first socket and check if the underlying protocol
1429         * supports the socketpair call.
1430         */
1431
1432        err = sock_create(family, type, protocol, &sock1);
1433        if (err < 0)
1434                goto out;
1435
1436        err = sock_create(family, type, protocol, &sock2);
1437        if (err < 0)
1438                goto out_release_1;
1439
1440        err = sock1->ops->socketpair(sock1, sock2);
1441        if (err < 0)
1442                goto out_release_both;
1443
1444        fd1 = get_unused_fd_flags(flags);
1445        if (unlikely(fd1 < 0)) {
1446                err = fd1;
1447                goto out_release_both;
1448        }
1449
1450        fd2 = get_unused_fd_flags(flags);
1451        if (unlikely(fd2 < 0)) {
1452                err = fd2;
1453                goto out_put_unused_1;
1454        }
1455
1456        newfile1 = sock_alloc_file(sock1, flags, NULL);
1457        if (unlikely(IS_ERR(newfile1))) {
1458                err = PTR_ERR(newfile1);
1459                goto out_put_unused_both;
1460        }
1461
1462        newfile2 = sock_alloc_file(sock2, flags, NULL);
1463        if (IS_ERR(newfile2)) {
1464                err = PTR_ERR(newfile2);
1465                goto out_fput_1;
1466        }
1467
1468        err = put_user(fd1, &usockvec[0]);
1469        if (err)
1470                goto out_fput_both;
1471
1472        err = put_user(fd2, &usockvec[1]);
1473        if (err)
1474                goto out_fput_both;
1475
1476        audit_fd_pair(fd1, fd2);
1477
1478        fd_install(fd1, newfile1);
1479        fd_install(fd2, newfile2);
1480        /* fd1 and fd2 may be already another descriptors.
1481         * Not kernel problem.
1482         */
1483
1484        return 0;
1485
1486out_fput_both:
1487        fput(newfile2);
1488        fput(newfile1);
1489        put_unused_fd(fd2);
1490        put_unused_fd(fd1);
1491        goto out;
1492
1493out_fput_1:
1494        fput(newfile1);
1495        put_unused_fd(fd2);
1496        put_unused_fd(fd1);
1497        sock_release(sock2);
1498        goto out;
1499
1500out_put_unused_both:
1501        put_unused_fd(fd2);
1502out_put_unused_1:
1503        put_unused_fd(fd1);
1504out_release_both:
1505        sock_release(sock2);
1506out_release_1:
1507        sock_release(sock1);
1508out:
1509        return err;
1510}
1511
1512/*
1513 *      Bind a name to a socket. Nothing much to do here since it's
1514 *      the protocol's responsibility to handle the local address.
1515 *
1516 *      We move the socket address to kernel space before we call
1517 *      the protocol layer (having also checked the address is ok).
1518 */
1519
1520SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1521{
1522        struct socket *sock;
1523        struct sockaddr_storage address;
1524        int err, fput_needed;
1525
1526        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1527        if (sock) {
1528                err = move_addr_to_kernel(umyaddr, addrlen, &address);
1529                if (err >= 0) {
1530                        err = security_socket_bind(sock,
1531                                                   (struct sockaddr *)&address,
1532                                                   addrlen);
1533                        if (!err)
1534                                err = sock->ops->bind(sock,
1535                                                      (struct sockaddr *)
1536                                                      &address, addrlen);
1537                }
1538                fput_light(sock->file, fput_needed);
1539        }
1540        return err;
1541}
1542
1543/*
1544 *      Perform a listen. Basically, we allow the protocol to do anything
1545 *      necessary for a listen, and if that works, we mark the socket as
1546 *      ready for listening.
1547 */
1548
1549SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1550{
1551        struct socket *sock;
1552        int err, fput_needed;
1553        int somaxconn;
1554
1555        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1556        if (sock) {
1557                somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1558                if ((unsigned int)backlog > somaxconn)
1559                        backlog = somaxconn;
1560
1561                err = security_socket_listen(sock, backlog);
1562                if (!err)
1563                        err = sock->ops->listen(sock, backlog);
1564
1565                fput_light(sock->file, fput_needed);
1566        }
1567        return err;
1568}
1569
1570/*
1571 *      For accept, we attempt to create a new socket, set up the link
1572 *      with the client, wake up the client, then return the new
1573 *      connected fd. We collect the address of the connector in kernel
1574 *      space and move it to user at the very end. This is unclean because
1575 *      we open the socket then return an error.
1576 *
1577 *      1003.1g adds the ability to recvmsg() to query connection pending
1578 *      status to recvmsg. We need to add that support in a way thats
1579 *      clean when we restucture accept also.
1580 */
1581
1582SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1583                int __user *, upeer_addrlen, int, flags)
1584{
1585        struct socket *sock, *newsock;
1586        struct file *newfile;
1587        int err, len, newfd, fput_needed;
1588        struct sockaddr_storage address;
1589
1590        if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1591                return -EINVAL;
1592
1593        if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1594                flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1595
1596        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1597        if (!sock)
1598                goto out;
1599
1600        err = -ENFILE;
1601        newsock = sock_alloc();
1602        if (!newsock)
1603                goto out_put;
1604
1605        newsock->type = sock->type;
1606        newsock->ops = sock->ops;
1607
1608        /*
1609         * We don't need try_module_get here, as the listening socket (sock)
1610         * has the protocol module (sock->ops->owner) held.
1611         */
1612        __module_get(newsock->ops->owner);
1613
1614        newfd = get_unused_fd_flags(flags);
1615        if (unlikely(newfd < 0)) {
1616                err = newfd;
1617                sock_release(newsock);
1618                goto out_put;
1619        }
1620        newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1621        if (unlikely(IS_ERR(newfile))) {
1622                err = PTR_ERR(newfile);
1623                put_unused_fd(newfd);
1624                sock_release(newsock);
1625                goto out_put;
1626        }
1627
1628        err = security_socket_accept(sock, newsock);
1629        if (err)
1630                goto out_fd;
1631
1632        err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1633        if (err < 0)
1634                goto out_fd;
1635
1636        if (upeer_sockaddr) {
1637                if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1638                                          &len, 2) < 0) {
1639                        err = -ECONNABORTED;
1640                        goto out_fd;
1641                }
1642                err = move_addr_to_user(&address,
1643                                        len, upeer_sockaddr, upeer_addrlen);
1644                if (err < 0)
1645                        goto out_fd;
1646        }
1647
1648        /* File flags are not inherited via accept() unlike another OSes. */
1649
1650        fd_install(newfd, newfile);
1651        err = newfd;
1652
1653out_put:
1654        fput_light(sock->file, fput_needed);
1655out:
1656        return err;
1657out_fd:
1658        fput(newfile);
1659        put_unused_fd(newfd);
1660        goto out_put;
1661}
1662
1663SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1664                int __user *, upeer_addrlen)
1665{
1666        return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1667}
1668
1669/*
1670 *      Attempt to connect to a socket with the server address.  The address
1671 *      is in user space so we verify it is OK and move it to kernel space.
1672 *
1673 *      For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1674 *      break bindings
1675 *
1676 *      NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1677 *      other SEQPACKET protocols that take time to connect() as it doesn't
1678 *      include the -EINPROGRESS status for such sockets.
1679 */
1680
1681SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1682                int, addrlen)
1683{
1684        struct socket *sock;
1685        struct sockaddr_storage address;
1686        int err, fput_needed;
1687
1688        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1689        if (!sock)
1690                goto out;
1691        err = move_addr_to_kernel(uservaddr, addrlen, &address);
1692        if (err < 0)
1693                goto out_put;
1694
1695        err =
1696            security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1697        if (err)
1698                goto out_put;
1699
1700        err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1701                                 sock->file->f_flags);
1702out_put:
1703        fput_light(sock->file, fput_needed);
1704out:
1705        return err;
1706}
1707
1708/*
1709 *      Get the local address ('name') of a socket object. Move the obtained
1710 *      name to user space.
1711 */
1712
1713SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1714                int __user *, usockaddr_len)
1715{
1716        struct socket *sock;
1717        struct sockaddr_storage address;
1718        int len, err, fput_needed;
1719
1720        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1721        if (!sock)
1722                goto out;
1723
1724        err = security_socket_getsockname(sock);
1725        if (err)
1726                goto out_put;
1727
1728        err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1729        if (err)
1730                goto out_put;
1731        err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1732
1733out_put:
1734        fput_light(sock->file, fput_needed);
1735out:
1736        return err;
1737}
1738
1739/*
1740 *      Get the remote address ('name') of a socket object. Move the obtained
1741 *      name to user space.
1742 */
1743
1744SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1745                int __user *, usockaddr_len)
1746{
1747        struct socket *sock;
1748        struct sockaddr_storage address;
1749        int len, err, fput_needed;
1750
1751        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1752        if (sock != NULL) {
1753                err = security_socket_getpeername(sock);
1754                if (err) {
1755                        fput_light(sock->file, fput_needed);
1756                        return err;
1757                }
1758
1759                err =
1760                    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1761                                       1);
1762                if (!err)
1763                        err = move_addr_to_user(&address, len, usockaddr,
1764                                                usockaddr_len);
1765                fput_light(sock->file, fput_needed);
1766        }
1767        return err;
1768}
1769
1770/*
1771 *      Send a datagram to a given address. We move the address into kernel
1772 *      space and check the user space data area is readable before invoking
1773 *      the protocol.
1774 */
1775
1776SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1777                unsigned int, flags, struct sockaddr __user *, addr,
1778                int, addr_len)
1779{
1780        struct socket *sock;
1781        struct sockaddr_storage address;
1782        int err;
1783        struct msghdr msg;
1784        struct iovec iov;
1785        int fput_needed;
1786
1787        if (len > INT_MAX)
1788                len = INT_MAX;
1789        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1790        if (!sock)
1791                goto out;
1792
1793        iov.iov_base = buff;
1794        iov.iov_len = len;
1795        msg.msg_name = NULL;
1796        msg.msg_iov = &iov;
1797        msg.msg_iovlen = 1;
1798        msg.msg_control = NULL;
1799        msg.msg_controllen = 0;
1800        msg.msg_namelen = 0;
1801        if (addr) {
1802                err = move_addr_to_kernel(addr, addr_len, &address);
1803                if (err < 0)
1804                        goto out_put;
1805                msg.msg_name = (struct sockaddr *)&address;
1806                msg.msg_namelen = addr_len;
1807        }
1808        if (sock->file->f_flags & O_NONBLOCK)
1809                flags |= MSG_DONTWAIT;
1810        msg.msg_flags = flags;
1811        err = sock_sendmsg(sock, &msg, len);
1812
1813out_put:
1814        fput_light(sock->file, fput_needed);
1815out:
1816        return err;
1817}
1818
1819/*
1820 *      Send a datagram down a socket.
1821 */
1822
1823SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1824                unsigned int, flags)
1825{
1826        return sys_sendto(fd, buff, len, flags, NULL, 0);
1827}
1828
1829/*
1830 *      Receive a frame from the socket and optionally record the address of the
1831 *      sender. We verify the buffers are writable and if needed move the
1832 *      sender address from kernel to user space.
1833 */
1834
1835SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1836                unsigned int, flags, struct sockaddr __user *, addr,
1837                int __user *, addr_len)
1838{
1839        struct socket *sock;
1840        struct iovec iov;
1841        struct msghdr msg;
1842        struct sockaddr_storage address;
1843        int err, err2;
1844        int fput_needed;
1845
1846        if (size > INT_MAX)
1847                size = INT_MAX;
1848        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1849        if (!sock)
1850                goto out;
1851
1852        msg.msg_control = NULL;
1853        msg.msg_controllen = 0;
1854        msg.msg_iovlen = 1;
1855        msg.msg_iov = &iov;
1856        iov.iov_len = size;
1857        iov.iov_base = ubuf;
1858        /* Save some cycles and don't copy the address if not needed */
1859        msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1860        /* We assume all kernel code knows the size of sockaddr_storage */
1861        msg.msg_namelen = 0;
1862        if (sock->file->f_flags & O_NONBLOCK)
1863                flags |= MSG_DONTWAIT;
1864        err = sock_recvmsg(sock, &msg, size, flags);
1865
1866        if (err >= 0 && addr != NULL) {
1867                err2 = move_addr_to_user(&address,
1868                                         msg.msg_namelen, addr, addr_len);
1869                if (err2 < 0)
1870                        err = err2;
1871        }
1872
1873        fput_light(sock->file, fput_needed);
1874out:
1875        return err;
1876}
1877
1878/*
1879 *      Receive a datagram from a socket.
1880 */
1881
1882asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1883                         unsigned int flags)
1884{
1885        return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1886}
1887
1888/*
1889 *      Set a socket option. Because we don't know the option lengths we have
1890 *      to pass the user mode parameter for the protocols to sort out.
1891 */
1892
1893SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1894                char __user *, optval, int, optlen)
1895{
1896        int err, fput_needed;
1897        struct socket *sock;
1898
1899        if (optlen < 0)
1900                return -EINVAL;
1901
1902        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1903        if (sock != NULL) {
1904                err = security_socket_setsockopt(sock, level, optname);
1905                if (err)
1906                        goto out_put;
1907
1908                if (level == SOL_SOCKET)
1909                        err =
1910                            sock_setsockopt(sock, level, optname, optval,
1911                                            optlen);
1912                else
1913                        err =
1914                            sock->ops->setsockopt(sock, level, optname, optval,
1915                                                  optlen);
1916out_put:
1917                fput_light(sock->file, fput_needed);
1918        }
1919        return err;
1920}
1921
1922/*
1923 *      Get a socket option. Because we don't know the option lengths we have
1924 *      to pass a user mode parameter for the protocols to sort out.
1925 */
1926
1927SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1928                char __user *, optval, int __user *, optlen)
1929{
1930        int err, fput_needed;
1931        struct socket *sock;
1932
1933        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1934        if (sock != NULL) {
1935                err = security_socket_getsockopt(sock, level, optname);
1936                if (err)
1937                        goto out_put;
1938
1939                if (level == SOL_SOCKET)
1940                        err =
1941                            sock_getsockopt(sock, level, optname, optval,
1942                                            optlen);
1943                else
1944                        err =
1945                            sock->ops->getsockopt(sock, level, optname, optval,
1946                                                  optlen);
1947out_put:
1948                fput_light(sock->file, fput_needed);
1949        }
1950        return err;
1951}
1952
1953/*
1954 *      Shutdown a socket.
1955 */
1956
1957SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1958{
1959        int err, fput_needed;
1960        struct socket *sock;
1961
1962        sock = sockfd_lookup_light(fd, &err, &fput_needed);
1963        if (sock != NULL) {
1964                err = security_socket_shutdown(sock, how);
1965                if (!err)
1966                        err = sock->ops->shutdown(sock, how);
1967                fput_light(sock->file, fput_needed);
1968        }
1969        return err;
1970}
1971
1972/* A couple of helpful macros for getting the address of the 32/64 bit
1973 * fields which are the same type (int / unsigned) on our platforms.
1974 */
1975#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1976#define COMPAT_NAMELEN(msg)     COMPAT_MSG(msg, msg_namelen)
1977#define COMPAT_FLAGS(msg)       COMPAT_MSG(msg, msg_flags)
1978
1979struct used_address {
1980        struct sockaddr_storage name;
1981        unsigned int name_len;
1982};
1983
1984static int copy_msghdr_from_user(struct msghdr *kmsg,
1985                                 struct msghdr __user *umsg)
1986{
1987        if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1988                return -EFAULT;
1989
1990        if (kmsg->msg_namelen < 0)
1991                return -EINVAL;
1992
1993        if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1994                kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1995        return 0;
1996}
1997
1998static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1999                         struct msghdr *msg_sys, unsigned int flags,
2000                         struct used_address *used_address)
2001{
2002        struct compat_msghdr __user *msg_compat =
2003            (struct compat_msghdr __user *)msg;
2004        struct sockaddr_storage address;
2005        struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2006        unsigned char ctl[sizeof(struct cmsghdr) + 20]
2007            __attribute__ ((aligned(sizeof(__kernel_size_t))));
2008        /* 20 is size of ipv6_pktinfo */
2009        unsigned char *ctl_buf = ctl;
2010        int err, ctl_len, total_len;
2011
2012        err = -EFAULT;
2013        if (MSG_CMSG_COMPAT & flags) {
2014                if (get_compat_msghdr(msg_sys, msg_compat))
2015                        return -EFAULT;
2016        } else {
2017                err = copy_msghdr_from_user(msg_sys, msg);
2018                if (err)
2019                        return err;
2020        }
2021
2022        if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2023                err = -EMSGSIZE;
2024                if (msg_sys->msg_iovlen > UIO_MAXIOV)
2025                        goto out;
2026                err = -ENOMEM;
2027                iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2028                              GFP_KERNEL);
2029                if (!iov)
2030                        goto out;
2031        }
2032
2033        /* This will also move the address data into kernel space */
2034        if (MSG_CMSG_COMPAT & flags) {
2035                err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2036        } else
2037                err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2038        if (err < 0)
2039                goto out_freeiov;
2040        total_len = err;
2041
2042        err = -ENOBUFS;
2043
2044        if (msg_sys->msg_controllen > INT_MAX)
2045                goto out_freeiov;
2046        ctl_len = msg_sys->msg_controllen;
2047        if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2048                err =
2049                    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2050                                                     sizeof(ctl));
2051                if (err)
2052                        goto out_freeiov;
2053                ctl_buf = msg_sys->msg_control;
2054                ctl_len = msg_sys->msg_controllen;
2055        } else if (ctl_len) {
2056                if (ctl_len > sizeof(ctl)) {
2057                        ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2058                        if (ctl_buf == NULL)
2059                                goto out_freeiov;
2060                }
2061                err = -EFAULT;
2062                /*
2063                 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2064                 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2065                 * checking falls down on this.
2066                 */
2067                if (copy_from_user(ctl_buf,
2068                                   (void __user __force *)msg_sys->msg_control,
2069                                   ctl_len))
2070                        goto out_freectl;
2071                msg_sys->msg_control = ctl_buf;
2072        }
2073        msg_sys->msg_flags = flags;
2074
2075        if (sock->file->f_flags & O_NONBLOCK)
2076                msg_sys->msg_flags |= MSG_DONTWAIT;
2077        /*
2078         * If this is sendmmsg() and current destination address is same as
2079         * previously succeeded address, omit asking LSM's decision.
2080         * used_address->name_len is initialized to UINT_MAX so that the first
2081         * destination address never matches.
2082         */
2083        if (used_address && msg_sys->msg_name &&
2084            used_address->name_len == msg_sys->msg_namelen &&
2085            !memcmp(&used_address->name, msg_sys->msg_name,
2086                    used_address->name_len)) {
2087                err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2088                goto out_freectl;
2089        }
2090        err = sock_sendmsg(sock, msg_sys, total_len);
2091        /*
2092         * If this is sendmmsg() and sending to current destination address was
2093         * successful, remember it.
2094         */
2095        if (used_address && err >= 0) {
2096                used_address->name_len = msg_sys->msg_namelen;
2097                if (msg_sys->msg_name)
2098                        memcpy(&used_address->name, msg_sys->msg_name,
2099                               used_address->name_len);
2100        }
2101
2102out_freectl:
2103        if (ctl_buf != ctl)
2104                sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2105out_freeiov:
2106        if (iov != iovstack)
2107                kfree(iov);
2108out:
2109        return err;
2110}
2111
2112/*
2113 *      BSD sendmsg interface
2114 */
2115
2116long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2117{
2118        int fput_needed, err;
2119        struct msghdr msg_sys;
2120        struct socket *sock;
2121
2122        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2123        if (!sock)
2124                goto out;
2125
2126        err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2127
2128        fput_light(sock->file, fput_needed);
2129out:
2130        return err;
2131}
2132
2133SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2134{
2135        if (flags & MSG_CMSG_COMPAT)
2136                return -EINVAL;
2137        return __sys_sendmsg(fd, msg, flags);
2138}
2139
2140/*
2141 *      Linux sendmmsg interface
2142 */
2143
2144int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2145                   unsigned int flags)
2146{
2147        int fput_needed, err, datagrams;
2148        struct socket *sock;
2149        struct mmsghdr __user *entry;
2150        struct compat_mmsghdr __user *compat_entry;
2151        struct msghdr msg_sys;
2152        struct used_address used_address;
2153
2154        if (vlen > UIO_MAXIOV)
2155                vlen = UIO_MAXIOV;
2156
2157        datagrams = 0;
2158
2159        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2160        if (!sock)
2161                return err;
2162
2163        used_address.name_len = UINT_MAX;
2164        entry = mmsg;
2165        compat_entry = (struct compat_mmsghdr __user *)mmsg;
2166        err = 0;
2167
2168        while (datagrams < vlen) {
2169                if (MSG_CMSG_COMPAT & flags) {
2170                        err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2171                                             &msg_sys, flags, &used_address);
2172                        if (err < 0)
2173                                break;
2174                        err = __put_user(err, &compat_entry->msg_len);
2175                        ++compat_entry;
2176                } else {
2177                        err = ___sys_sendmsg(sock,
2178                                             (struct msghdr __user *)entry,
2179                                             &msg_sys, flags, &used_address);
2180                        if (err < 0)
2181                                break;
2182                        err = put_user(err, &entry->msg_len);
2183                        ++entry;
2184                }
2185
2186                if (err)
2187                        break;
2188                ++datagrams;
2189        }
2190
2191        fput_light(sock->file, fput_needed);
2192
2193        /* We only return an error if no datagrams were able to be sent */
2194        if (datagrams != 0)
2195                return datagrams;
2196
2197        return err;
2198}
2199
2200SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2201                unsigned int, vlen, unsigned int, flags)
2202{
2203        if (flags & MSG_CMSG_COMPAT)
2204                return -EINVAL;
2205        return __sys_sendmmsg(fd, mmsg, vlen, flags);
2206}
2207
2208static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2209                         struct msghdr *msg_sys, unsigned int flags, int nosec)
2210{
2211        struct compat_msghdr __user *msg_compat =
2212            (struct compat_msghdr __user *)msg;
2213        struct iovec iovstack[UIO_FASTIOV];
2214        struct iovec *iov = iovstack;
2215        unsigned long cmsg_ptr;
2216        int err, total_len, len;
2217
2218        /* kernel mode address */
2219        struct sockaddr_storage addr;
2220
2221        /* user mode address pointers */
2222        struct sockaddr __user *uaddr;
2223        int __user *uaddr_len;
2224
2225        if (MSG_CMSG_COMPAT & flags) {
2226                if (get_compat_msghdr(msg_sys, msg_compat))
2227                        return -EFAULT;
2228        } else {
2229                err = copy_msghdr_from_user(msg_sys, msg);
2230                if (err)
2231                        return err;
2232        }
2233
2234        if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2235                err = -EMSGSIZE;
2236                if (msg_sys->msg_iovlen > UIO_MAXIOV)
2237                        goto out;
2238                err = -ENOMEM;
2239                iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2240                              GFP_KERNEL);
2241                if (!iov)
2242                        goto out;
2243        }
2244
2245        /* Save the user-mode address (verify_iovec will change the
2246         * kernel msghdr to use the kernel address space)
2247         */
2248        uaddr = (__force void __user *)msg_sys->msg_name;
2249        uaddr_len = COMPAT_NAMELEN(msg);
2250        if (MSG_CMSG_COMPAT & flags)
2251                err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2252        else
2253                err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2254        if (err < 0)
2255                goto out_freeiov;
2256        total_len = err;
2257
2258        cmsg_ptr = (unsigned long)msg_sys->msg_control;
2259        msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2260
2261        /* We assume all kernel code knows the size of sockaddr_storage */
2262        msg_sys->msg_namelen = 0;
2263
2264        if (sock->file->f_flags & O_NONBLOCK)
2265                flags |= MSG_DONTWAIT;
2266        err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2267                                                          total_len, flags);
2268        if (err < 0)
2269                goto out_freeiov;
2270        len = err;
2271
2272        if (uaddr != NULL) {
2273                err = move_addr_to_user(&addr,
2274                                        msg_sys->msg_namelen, uaddr,
2275                                        uaddr_len);
2276                if (err < 0)
2277                        goto out_freeiov;
2278        }
2279        err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2280                         COMPAT_FLAGS(msg));
2281        if (err)
2282                goto out_freeiov;
2283        if (MSG_CMSG_COMPAT & flags)
2284                err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2285                                 &msg_compat->msg_controllen);
2286        else
2287                err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2288                                 &msg->msg_controllen);
2289        if (err)
2290                goto out_freeiov;
2291        err = len;
2292
2293out_freeiov:
2294        if (iov != iovstack)
2295                kfree(iov);
2296out:
2297        return err;
2298}
2299
2300/*
2301 *      BSD recvmsg interface
2302 */
2303
2304long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2305{
2306        int fput_needed, err;
2307        struct msghdr msg_sys;
2308        struct socket *sock;
2309
2310        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2311        if (!sock)
2312                goto out;
2313
2314        err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2315
2316        fput_light(sock->file, fput_needed);
2317out:
2318        return err;
2319}
2320
2321SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2322                unsigned int, flags)
2323{
2324        if (flags & MSG_CMSG_COMPAT)
2325                return -EINVAL;
2326        return __sys_recvmsg(fd, msg, flags);
2327}
2328
2329/*
2330 *     Linux recvmmsg interface
2331 */
2332
2333int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2334                   unsigned int flags, struct timespec *timeout)
2335{
2336        int fput_needed, err, datagrams;
2337        struct socket *sock;
2338        struct mmsghdr __user *entry;
2339        struct compat_mmsghdr __user *compat_entry;
2340        struct msghdr msg_sys;
2341        struct timespec end_time;
2342
2343        if (timeout &&
2344            poll_select_set_timeout(&end_time, timeout->tv_sec,
2345                                    timeout->tv_nsec))
2346                return -EINVAL;
2347
2348        datagrams = 0;
2349
2350        sock = sockfd_lookup_light(fd, &err, &fput_needed);
2351        if (!sock)
2352                return err;
2353
2354        err = sock_error(sock->sk);
2355        if (err)
2356                goto out_put;
2357
2358        entry = mmsg;
2359        compat_entry = (struct compat_mmsghdr __user *)mmsg;
2360
2361        while (datagrams < vlen) {
2362                /*
2363                 * No need to ask LSM for more than the first datagram.
2364                 */
2365                if (MSG_CMSG_COMPAT & flags) {
2366                        err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2367                                             &msg_sys, flags & ~MSG_WAITFORONE,
2368                                             datagrams);
2369                        if (err < 0)
2370                                break;
2371                        err = __put_user(err, &compat_entry->msg_len);
2372                        ++compat_entry;
2373                } else {
2374                        err = ___sys_recvmsg(sock,
2375                                             (struct msghdr __user *)entry,
2376                                             &msg_sys, flags & ~MSG_WAITFORONE,
2377                                             datagrams);
2378                        if (err < 0)
2379                                break;
2380                        err = put_user(err, &entry->msg_len);
2381                        ++entry;
2382                }
2383
2384                if (err)
2385                        break;
2386                ++datagrams;
2387
2388                /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2389                if (flags & MSG_WAITFORONE)
2390                        flags |= MSG_DONTWAIT;
2391
2392                if (timeout) {
2393                        ktime_get_ts(timeout);
2394                        *timeout = timespec_sub(end_time, *timeout);
2395                        if (timeout->tv_sec < 0) {
2396                                timeout->tv_sec = timeout->tv_nsec = 0;
2397                                break;
2398                        }
2399
2400                        /* Timeout, return less than vlen datagrams */
2401                        if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2402                                break;
2403                }
2404
2405                /* Out of band data, return right away */
2406                if (msg_sys.msg_flags & MSG_OOB)
2407                        break;
2408        }
2409
2410out_put:
2411        fput_light(sock->file, fput_needed);
2412
2413        if (err == 0)
2414                return datagrams;
2415
2416        if (datagrams != 0) {
2417                /*
2418                 * We may return less entries than requested (vlen) if the
2419                 * sock is non block and there aren't enough datagrams...
2420                 */
2421                if (err != -EAGAIN) {
2422                        /*
2423                         * ... or  if recvmsg returns an error after we
2424                         * received some datagrams, where we record the
2425                         * error to return on the next call or if the
2426                         * app asks about it using getsockopt(SO_ERROR).
2427                         */
2428                        sock->sk->sk_err = -err;
2429                }
2430
2431                return datagrams;
2432        }
2433
2434        return err;
2435}
2436
2437SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2438                unsigned int, vlen, unsigned int, flags,
2439                struct timespec __user *, timeout)
2440{
2441        int datagrams;
2442        struct timespec timeout_sys;
2443
2444        if (flags & MSG_CMSG_COMPAT)
2445                return -EINVAL;
2446
2447        if (!timeout)
2448                return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2449
2450        if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2451                return -EFAULT;
2452
2453        datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2454
2455        if (datagrams > 0 &&
2456            copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2457                datagrams = -EFAULT;
2458
2459        return datagrams;
2460}
2461
2462#ifdef __ARCH_WANT_SYS_SOCKETCALL
2463/* Argument list sizes for sys_socketcall */
2464#define AL(x) ((x) * sizeof(unsigned long))
2465static const unsigned char nargs[21] = {
2466        AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2467        AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2468        AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2469        AL(4), AL(5), AL(4)
2470};
2471
2472#undef AL
2473
2474/*
2475 *      System call vectors.
2476 *
2477 *      Argument checking cleaned up. Saved 20% in size.
2478 *  This function doesn't need to set the kernel lock because
2479 *  it is set by the callees.
2480 */
2481
2482SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2483{
2484        unsigned long a[AUDITSC_ARGS];
2485        unsigned long a0, a1;
2486        int err;
2487        unsigned int len;
2488
2489        if (call < 1 || call > SYS_SENDMMSG)
2490                return -EINVAL;
2491
2492        len = nargs[call];
2493        if (len > sizeof(a))
2494                return -EINVAL;
2495
2496        /* copy_from_user should be SMP safe. */
2497        if (copy_from_user(a, args, len))
2498                return -EFAULT;
2499
2500        err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2501        if (err)
2502                return err;
2503
2504        a0 = a[0];
2505        a1 = a[1];
2506
2507        switch (call) {
2508        case SYS_SOCKET:
2509                err = sys_socket(a0, a1, a[2]);
2510                break;
2511        case SYS_BIND:
2512                err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2513                break;
2514        case SYS_CONNECT:
2515                err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2516                break;
2517        case SYS_LISTEN:
2518                err = sys_listen(a0, a1);
2519                break;
2520        case SYS_ACCEPT:
2521                err = sys_accept4(a0, (struct sockaddr __user *)a1,
2522                                  (int __user *)a[2], 0);
2523                break;
2524        case SYS_GETSOCKNAME:
2525                err =
2526                    sys_getsockname(a0, (struct sockaddr __user *)a1,
2527                                    (int __user *)a[2]);
2528                break;
2529        case SYS_GETPEERNAME:
2530                err =
2531                    sys_getpeername(a0, (struct sockaddr __user *)a1,
2532                                    (int __user *)a[2]);
2533                break;
2534        case SYS_SOCKETPAIR:
2535                err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2536                break;
2537        case SYS_SEND:
2538                err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2539                break;
2540        case SYS_SENDTO:
2541                err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2542                                 (struct sockaddr __user *)a[4], a[5]);
2543                break;
2544        case SYS_RECV:
2545                err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2546                break;
2547        case SYS_RECVFROM:
2548                err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2549                                   (struct sockaddr __user *)a[4],
2550                                   (int __user *)a[5]);
2551                break;
2552        case SYS_SHUTDOWN:
2553                err = sys_shutdown(a0, a1);
2554                break;
2555        case SYS_SETSOCKOPT:
2556                err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2557                break;
2558        case SYS_GETSOCKOPT:
2559                err =
2560                    sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2561                                   (int __user *)a[4]);
2562                break;
2563        case SYS_SENDMSG:
2564                err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2565                break;
2566        case SYS_SENDMMSG:
2567                err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2568                break;
2569        case SYS_RECVMSG:
2570                err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2571                break;
2572        case SYS_RECVMMSG:
2573                err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2574                                   (struct timespec __user *)a[4]);
2575                break;
2576        case SYS_ACCEPT4:
2577                err = sys_accept4(a0, (struct sockaddr __user *)a1,
2578                                  (int __user *)a[2], a[3]);
2579                break;
2580        default:
2581                err = -EINVAL;
2582                break;
2583        }
2584        return err;
2585}
2586
2587#endif                          /* __ARCH_WANT_SYS_SOCKETCALL */
2588
2589/**
2590 *      sock_register - add a socket protocol handler
2591 *      @ops: description of protocol
2592 *
2593 *      This function is called by a protocol handler that wants to
2594 *      advertise its address family, and have it linked into the
2595 *      socket interface. The value ops->family coresponds to the
2596 *      socket system call protocol family.
2597 */
2598int sock_register(const struct net_proto_family *ops)
2599{
2600        int err;
2601
2602        if (ops->family >= NPROTO) {
2603                printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2604                       NPROTO);
2605                return -ENOBUFS;
2606        }
2607
2608        spin_lock(&net_family_lock);
2609        if (rcu_dereference_protected(net_families[ops->family],
2610                                      lockdep_is_held(&net_family_lock)))
2611                err = -EEXIST;
2612        else {
2613                rcu_assign_pointer(net_families[ops->family], ops);
2614                err = 0;
2615        }
2616        spin_unlock(&net_family_lock);
2617
2618        printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2619        return err;
2620}
2621EXPORT_SYMBOL(sock_register);
2622
2623/**
2624 *      sock_unregister - remove a protocol handler
2625 *      @family: protocol family to remove
2626 *
2627 *      This function is called by a protocol handler that wants to
2628 *      remove its address family, and have it unlinked from the
2629 *      new socket creation.
2630 *
2631 *      If protocol handler is a module, then it can use module reference
2632 *      counts to protect against new references. If protocol handler is not
2633 *      a module then it needs to provide its own protection in
2634 *      the ops->create routine.
2635 */
2636void sock_unregister(int family)
2637{
2638        BUG_ON(family < 0 || family >= NPROTO);
2639
2640        spin_lock(&net_family_lock);
2641        RCU_INIT_POINTER(net_families[family], NULL);
2642        spin_unlock(&net_family_lock);
2643
2644        synchronize_rcu();
2645
2646        printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2647}
2648EXPORT_SYMBOL(sock_unregister);
2649
2650static int __init sock_init(void)
2651{
2652        int err;
2653        /*
2654         *      Initialize the network sysctl infrastructure.
2655         */
2656        err = net_sysctl_init();
2657        if (err)
2658                goto out;
2659
2660        /*
2661         *      Initialize skbuff SLAB cache
2662         */
2663        skb_init();
2664
2665        /*
2666         *      Initialize the protocols module.
2667         */
2668
2669        init_inodecache();
2670
2671        err = register_filesystem(&sock_fs_type);
2672        if (err)
2673                goto out_fs;
2674        sock_mnt = kern_mount(&sock_fs_type);
2675        if (IS_ERR(sock_mnt)) {
2676                err = PTR_ERR(sock_mnt);
2677                goto out_mount;
2678        }
2679
2680        /* The real protocol initialization is performed in later initcalls.
2681         */
2682
2683#ifdef CONFIG_NETFILTER
2684        err = netfilter_init();
2685        if (err)
2686                goto out;
2687#endif
2688
2689#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2690        skb_timestamping_init();
2691#endif
2692
2693out:
2694        return err;
2695
2696out_mount:
2697        unregister_filesystem(&sock_fs_type);
2698out_fs:
2699        goto out;
2700}
2701
2702core_initcall(sock_init);       /* early initcall */
2703
2704#ifdef CONFIG_PROC_FS
2705void socket_seq_show(struct seq_file *seq)
2706{
2707        int cpu;
2708        int counter = 0;
2709
2710        for_each_possible_cpu(cpu)
2711            counter += per_cpu(sockets_in_use, cpu);
2712
2713        /* It can be negative, by the way. 8) */
2714        if (counter < 0)
2715                counter = 0;
2716
2717        seq_printf(seq, "sockets: used %d\n", counter);
2718}
2719#endif                          /* CONFIG_PROC_FS */
2720
2721#ifdef CONFIG_COMPAT
2722static int do_siocgstamp(struct net *net, struct socket *sock,
2723                         unsigned int cmd, void __user *up)
2724{
2725        mm_segment_t old_fs = get_fs();
2726        struct timeval ktv;
2727        int err;
2728
2729        set_fs(KERNEL_DS);
2730        err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2731        set_fs(old_fs);
2732        if (!err)
2733                err = compat_put_timeval(&ktv, up);
2734
2735        return err;
2736}
2737
2738static int do_siocgstampns(struct net *net, struct socket *sock,
2739                           unsigned int cmd, void __user *up)
2740{
2741        mm_segment_t old_fs = get_fs();
2742        struct timespec kts;
2743        int err;
2744
2745        set_fs(KERNEL_DS);
2746        err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2747        set_fs(old_fs);
2748        if (!err)
2749                err = compat_put_timespec(&kts, up);
2750
2751        return err;
2752}
2753
2754static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2755{
2756        struct ifreq __user *uifr;
2757        int err;
2758
2759        uifr = compat_alloc_user_space(sizeof(struct ifreq));
2760        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2761                return -EFAULT;
2762
2763        err = dev_ioctl(net, SIOCGIFNAME, uifr);
2764        if (err)
2765                return err;
2766
2767        if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2768                return -EFAULT;
2769
2770        return 0;
2771}
2772
2773static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2774{
2775        struct compat_ifconf ifc32;
2776        struct ifconf ifc;
2777        struct ifconf __user *uifc;
2778        struct compat_ifreq __user *ifr32;
2779        struct ifreq __user *ifr;
2780        unsigned int i, j;
2781        int err;
2782
2783        if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2784                return -EFAULT;
2785
2786        memset(&ifc, 0, sizeof(ifc));
2787        if (ifc32.ifcbuf == 0) {
2788                ifc32.ifc_len = 0;
2789                ifc.ifc_len = 0;
2790                ifc.ifc_req = NULL;
2791                uifc = compat_alloc_user_space(sizeof(struct ifconf));
2792        } else {
2793                size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2794                        sizeof(struct ifreq);
2795                uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2796                ifc.ifc_len = len;
2797                ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2798                ifr32 = compat_ptr(ifc32.ifcbuf);
2799                for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2800                        if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2801                                return -EFAULT;
2802                        ifr++;
2803                        ifr32++;
2804                }
2805        }
2806        if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2807                return -EFAULT;
2808
2809        err = dev_ioctl(net, SIOCGIFCONF, uifc);
2810        if (err)
2811                return err;
2812
2813        if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2814                return -EFAULT;
2815
2816        ifr = ifc.ifc_req;
2817        ifr32 = compat_ptr(ifc32.ifcbuf);
2818        for (i = 0, j = 0;
2819             i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2820             i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2821                if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2822                        return -EFAULT;
2823                ifr32++;
2824                ifr++;
2825        }
2826
2827        if (ifc32.ifcbuf == 0) {
2828                /* Translate from 64-bit structure multiple to
2829                 * a 32-bit one.
2830                 */
2831                i = ifc.ifc_len;
2832                i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2833                ifc32.ifc_len = i;
2834        } else {
2835                ifc32.ifc_len = i;
2836        }
2837        if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2838                return -EFAULT;
2839
2840        return 0;
2841}
2842
2843static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2844{
2845        struct compat_ethtool_rxnfc __user *compat_rxnfc;
2846        bool convert_in = false, convert_out = false;
2847        size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2848        struct ethtool_rxnfc __user *rxnfc;
2849        struct ifreq __user *ifr;
2850        u32 rule_cnt = 0, actual_rule_cnt;
2851        u32 ethcmd;
2852        u32 data;
2853        int ret;
2854
2855        if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2856                return -EFAULT;
2857
2858        compat_rxnfc = compat_ptr(data);
2859
2860        if (get_user(ethcmd, &compat_rxnfc->cmd))
2861                return -EFAULT;
2862
2863        /* Most ethtool structures are defined without padding.
2864         * Unfortunately struct ethtool_rxnfc is an exception.
2865         */
2866        switch (ethcmd) {
2867        default:
2868                break;
2869        case ETHTOOL_GRXCLSRLALL:
2870                /* Buffer size is variable */
2871                if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2872                        return -EFAULT;
2873                if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2874                        return -ENOMEM;
2875                buf_size += rule_cnt * sizeof(u32);
2876                /* fall through */
2877        case ETHTOOL_GRXRINGS:
2878        case ETHTOOL_GRXCLSRLCNT:
2879        case ETHTOOL_GRXCLSRULE:
2880        case ETHTOOL_SRXCLSRLINS:
2881                convert_out = true;
2882                /* fall through */
2883        case ETHTOOL_SRXCLSRLDEL:
2884                buf_size += sizeof(struct ethtool_rxnfc);
2885                convert_in = true;
2886                break;
2887        }
2888
2889        ifr = compat_alloc_user_space(buf_size);
2890        rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2891
2892        if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2893                return -EFAULT;
2894
2895        if (put_user(convert_in ? rxnfc : compat_ptr(data),
2896                     &ifr->ifr_ifru.ifru_data))
2897                return -EFAULT;
2898
2899        if (convert_in) {
2900                /* We expect there to be holes between fs.m_ext and
2901                 * fs.ring_cookie and at the end of fs, but nowhere else.
2902                 */
2903                BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2904                             sizeof(compat_rxnfc->fs.m_ext) !=
2905                             offsetof(struct ethtool_rxnfc, fs.m_ext) +
2906                             sizeof(rxnfc->fs.m_ext));
2907                BUILD_BUG_ON(
2908                        offsetof(struct compat_ethtool_rxnfc, fs.location) -
2909                        offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2910                        offsetof(struct ethtool_rxnfc, fs.location) -
2911                        offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2912
2913                if (copy_in_user(rxnfc, compat_rxnfc,
2914                                 (void __user *)(&rxnfc->fs.m_ext + 1) -
2915                                 (void __user *)rxnfc) ||
2916                    copy_in_user(&rxnfc->fs.ring_cookie,
2917                                 &compat_rxnfc->fs.ring_cookie,
2918                                 (void __user *)(&rxnfc->fs.location + 1) -
2919                                 (void __user *)&rxnfc->fs.ring_cookie) ||
2920                    copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2921                                 sizeof(rxnfc->rule_cnt)))
2922                        return -EFAULT;
2923        }
2924
2925        ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2926        if (ret)
2927                return ret;
2928
2929        if (convert_out) {
2930                if (copy_in_user(compat_rxnfc, rxnfc,
2931                                 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2932                                 (const void __user *)rxnfc) ||
2933                    copy_in_user(&compat_rxnfc->fs.ring_cookie,
2934                                 &rxnfc->fs.ring_cookie,
2935                                 (const void __user *)(&rxnfc->fs.location + 1) -
2936                                 (const void __user *)&rxnfc->fs.ring_cookie) ||
2937                    copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2938                                 sizeof(rxnfc->rule_cnt)))
2939                        return -EFAULT;
2940
2941                if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2942                        /* As an optimisation, we only copy the actual
2943                         * number of rules that the underlying
2944                         * function returned.  Since Mallory might
2945                         * change the rule count in user memory, we
2946                         * check that it is less than the rule count
2947                         * originally given (as the user buffer size),
2948                         * which has been range-checked.
2949                         */
2950                        if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2951                                return -EFAULT;
2952                        if (actual_rule_cnt < rule_cnt)
2953                                rule_cnt = actual_rule_cnt;
2954                        if (copy_in_user(&compat_rxnfc->rule_locs[0],
2955                                         &rxnfc->rule_locs[0],
2956                                         rule_cnt * sizeof(u32)))
2957                                return -EFAULT;
2958                }
2959        }
2960
2961        return 0;
2962}
2963
2964static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2965{
2966        void __user *uptr;
2967        compat_uptr_t uptr32;
2968        struct ifreq __user *uifr;
2969
2970        uifr = compat_alloc_user_space(sizeof(*uifr));
2971        if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2972                return -EFAULT;
2973
2974        if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2975                return -EFAULT;
2976
2977        uptr = compat_ptr(uptr32);
2978
2979        if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2980                return -EFAULT;
2981
2982        return dev_ioctl(net, SIOCWANDEV, uifr);
2983}
2984
2985static int bond_ioctl(struct net *net, unsigned int cmd,
2986                         struct compat_ifreq __user *ifr32)
2987{
2988        struct ifreq kifr;
2989        mm_segment_t old_fs;
2990        int err;
2991
2992        switch (cmd) {
2993        case SIOCBONDENSLAVE:
2994        case SIOCBONDRELEASE:
2995        case SIOCBONDSETHWADDR:
2996        case SIOCBONDCHANGEACTIVE:
2997                if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2998                        return -EFAULT;
2999
3000                old_fs = get_fs();
3001                set_fs(KERNEL_DS);
3002                err = dev_ioctl(net, cmd,
3003                                (struct ifreq __user __force *) &kifr);
3004                set_fs(old_fs);
3005
3006                return err;
3007        default:
3008                return -ENOIOCTLCMD;
3009        }
3010}
3011
3012/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3013static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3014                                 struct compat_ifreq __user *u_ifreq32)
3015{
3016        struct ifreq __user *u_ifreq64;
3017        char tmp_buf[IFNAMSIZ];
3018        void __user *data64;
3019        u32 data32;
3020
3021        if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3022                           IFNAMSIZ))
3023                return -EFAULT;
3024        if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3025                return -EFAULT;
3026        data64 = compat_ptr(data32);
3027
3028        u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3029
3030        if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3031                         IFNAMSIZ))
3032                return -EFAULT;
3033        if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3034                return -EFAULT;
3035
3036        return dev_ioctl(net, cmd, u_ifreq64);
3037}
3038
3039static int dev_ifsioc(struct net *net, struct socket *sock,
3040                         unsigned int cmd, struct compat_ifreq __user *uifr32)
3041{
3042        struct ifreq __user *uifr;
3043        int err;
3044
3045        uifr = compat_alloc_user_space(sizeof(*uifr));
3046        if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3047                return -EFAULT;
3048
3049        err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3050
3051        if (!err) {
3052                switch (cmd) {
3053                case SIOCGIFFLAGS:
3054                case SIOCGIFMETRIC:
3055                case SIOCGIFMTU:
3056                case SIOCGIFMEM:
3057                case SIOCGIFHWADDR:
3058                case SIOCGIFINDEX:
3059                case SIOCGIFADDR:
3060                case SIOCGIFBRDADDR:
3061                case SIOCGIFDSTADDR:
3062                case SIOCGIFNETMASK:
3063                case SIOCGIFPFLAGS:
3064                case SIOCGIFTXQLEN:
3065                case SIOCGMIIPHY:
3066                case SIOCGMIIREG:
3067                        if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3068                                err = -EFAULT;
3069                        break;
3070                }
3071        }
3072        return err;
3073}
3074
3075static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3076                        struct compat_ifreq __user *uifr32)
3077{
3078        struct ifreq ifr;
3079        struct compat_ifmap __user *uifmap32;
3080        mm_segment_t old_fs;
3081        int err;
3082
3083        uifmap32 = &uifr32->ifr_ifru.ifru_map;
3084        err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3085        err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3086        err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3087        err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3088        err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3089        err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3090        err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3091        if (err)
3092                return -EFAULT;
3093
3094        old_fs = get_fs();
3095        set_fs(KERNEL_DS);
3096        err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
3097        set_fs(old_fs);
3098
3099        if (cmd == SIOCGIFMAP && !err) {
3100                err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3101                err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3102                err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3103                err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3104                err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3105                err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3106                err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3107                if (err)
3108                        err = -EFAULT;
3109        }
3110        return err;
3111}
3112
3113struct rtentry32 {
3114        u32             rt_pad1;
3115        struct sockaddr rt_dst;         /* target address               */
3116        struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3117        struct sockaddr rt_genmask;     /* target network mask (IP)     */
3118        unsigned short  rt_flags;
3119        short           rt_pad2;
3120        u32             rt_pad3;
3121        unsigned char   rt_tos;
3122        unsigned char   rt_class;
3123        short           rt_pad4;
3124        short           rt_metric;      /* +1 for binary compatibility! */
3125        /* char * */ u32 rt_dev;        /* forcing the device at add    */
3126        u32             rt_mtu;         /* per route MTU/Window         */
3127        u32             rt_window;      /* Window clamping              */
3128        unsigned short  rt_irtt;        /* Initial RTT                  */
3129};
3130
3131struct in6_rtmsg32 {
3132        struct in6_addr         rtmsg_dst;
3133        struct in6_addr         rtmsg_src;
3134        struct in6_addr         rtmsg_gateway;
3135        u32                     rtmsg_type;
3136        u16                     rtmsg_dst_len;
3137        u16                     rtmsg_src_len;
3138        u32                     rtmsg_metric;
3139        u32                     rtmsg_info;
3140        u32                     rtmsg_flags;
3141        s32                     rtmsg_ifindex;
3142};
3143
3144static int routing_ioctl(struct net *net, struct socket *sock,
3145                         unsigned int cmd, void __user *argp)
3146{
3147        int ret;
3148        void *r = NULL;
3149        struct in6_rtmsg r6;
3150        struct rtentry r4;
3151        char devname[16];
3152        u32 rtdev;
3153        mm_segment_t old_fs = get_fs();
3154
3155        if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3156                struct in6_rtmsg32 __user *ur6 = argp;
3157                ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3158                        3 * sizeof(struct in6_addr));
3159                ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3160                ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3161                ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3162                ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3163                ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3164                ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3165                ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3166
3167                r = (void *) &r6;
3168        } else { /* ipv4 */
3169                struct rtentry32 __user *ur4 = argp;
3170                ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3171                                        3 * sizeof(struct sockaddr));
3172                ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3173                ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3174                ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3175                ret |= get_user(r4.rt_window, &(ur4->rt_window));
3176                ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3177                ret |= get_user(rtdev, &(ur4->rt_dev));
3178                if (rtdev) {
3179                        ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3180                        r4.rt_dev = (char __user __force *)devname;
3181                        devname[15] = 0;
3182                } else
3183                        r4.rt_dev = NULL;
3184
3185                r = (void *) &r4;
3186        }
3187
3188        if (ret) {
3189                ret = -EFAULT;
3190                goto out;
3191        }
3192
3193        set_fs(KERNEL_DS);
3194        ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3195        set_fs(old_fs);
3196
3197out:
3198        return ret;
3199}
3200
3201/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3202 * for some operations; this forces use of the newer bridge-utils that
3203 * use compatible ioctls
3204 */
3205static int old_bridge_ioctl(compat_ulong_t __user *argp)
3206{
3207        compat_ulong_t tmp;
3208
3209        if (get_user(tmp, argp))
3210                return -EFAULT;
3211        if (tmp == BRCTL_GET_VERSION)
3212                return BRCTL_VERSION + 1;
3213        return -EINVAL;
3214}
3215
3216static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3217                         unsigned int cmd, unsigned long arg)
3218{
3219        void __user *argp = compat_ptr(arg);
3220        struct sock *sk = sock->sk;
3221        struct net *net = sock_net(sk);
3222
3223        if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3224                return compat_ifr_data_ioctl(net, cmd, argp);
3225
3226        switch (cmd) {
3227        case SIOCSIFBR:
3228        case SIOCGIFBR:
3229                return old_bridge_ioctl(argp);
3230        case SIOCGIFNAME:
3231                return dev_ifname32(net, argp);
3232        case SIOCGIFCONF:
3233                return dev_ifconf(net, argp);
3234        case SIOCETHTOOL:
3235                return ethtool_ioctl(net, argp);
3236        case SIOCWANDEV:
3237                return compat_siocwandev(net, argp);
3238        case SIOCGIFMAP:
3239        case SIOCSIFMAP:
3240                return compat_sioc_ifmap(net, cmd, argp);
3241        case SIOCBONDENSLAVE:
3242        case SIOCBONDRELEASE:
3243        case SIOCBONDSETHWADDR:
3244        case SIOCBONDCHANGEACTIVE:
3245                return bond_ioctl(net, cmd, argp);
3246        case SIOCADDRT:
3247        case SIOCDELRT:
3248                return routing_ioctl(net, sock, cmd, argp);
3249        case SIOCGSTAMP:
3250                return do_siocgstamp(net, sock, cmd, argp);
3251        case SIOCGSTAMPNS:
3252                return do_siocgstampns(net, sock, cmd, argp);
3253        case SIOCBONDSLAVEINFOQUERY:
3254        case SIOCBONDINFOQUERY:
3255        case SIOCSHWTSTAMP:
3256        case SIOCGHWTSTAMP:
3257                return compat_ifr_data_ioctl(net, cmd, argp);
3258
3259        case FIOSETOWN:
3260        case SIOCSPGRP:
3261        case FIOGETOWN:
3262        case SIOCGPGRP:
3263        case SIOCBRADDBR:
3264        case SIOCBRDELBR:
3265        case SIOCGIFVLAN:
3266        case SIOCSIFVLAN:
3267        case SIOCADDDLCI:
3268        case SIOCDELDLCI:
3269                return sock_ioctl(file, cmd, arg);
3270
3271        case SIOCGIFFLAGS:
3272        case SIOCSIFFLAGS:
3273        case SIOCGIFMETRIC:
3274        case SIOCSIFMETRIC:
3275        case SIOCGIFMTU:
3276        case SIOCSIFMTU:
3277        case SIOCGIFMEM:
3278        case SIOCSIFMEM:
3279        case SIOCGIFHWADDR:
3280        case SIOCSIFHWADDR:
3281        case SIOCADDMULTI:
3282        case SIOCDELMULTI:
3283        case SIOCGIFINDEX:
3284        case SIOCGIFADDR:
3285        case SIOCSIFADDR:
3286        case SIOCSIFHWBROADCAST:
3287        case SIOCDIFADDR:
3288        case SIOCGIFBRDADDR:
3289        case SIOCSIFBRDADDR:
3290        case SIOCGIFDSTADDR:
3291        case SIOCSIFDSTADDR:
3292        case SIOCGIFNETMASK:
3293        case SIOCSIFNETMASK:
3294        case SIOCSIFPFLAGS:
3295        case SIOCGIFPFLAGS:
3296        case SIOCGIFTXQLEN:
3297        case SIOCSIFTXQLEN:
3298        case SIOCBRADDIF:
3299        case SIOCBRDELIF:
3300        case SIOCSIFNAME:
3301        case SIOCGMIIPHY:
3302        case SIOCGMIIREG:
3303        case SIOCSMIIREG:
3304                return dev_ifsioc(net, sock, cmd, argp);
3305
3306        case SIOCSARP:
3307        case SIOCGARP:
3308        case SIOCDARP:
3309        case SIOCATMARK:
3310                return sock_do_ioctl(net, sock, cmd, arg);
3311        }
3312
3313        return -ENOIOCTLCMD;
3314}
3315
3316static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3317                              unsigned long arg)
3318{
3319        struct socket *sock = file->private_data;
3320        int ret = -ENOIOCTLCMD;
3321        struct sock *sk;
3322        struct net *net;
3323
3324        sk = sock->sk;
3325        net = sock_net(sk);
3326
3327        if (sock->ops->compat_ioctl)
3328                ret = sock->ops->compat_ioctl(sock, cmd, arg);
3329
3330        if (ret == -ENOIOCTLCMD &&
3331            (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3332                ret = compat_wext_handle_ioctl(net, cmd, arg);
3333
3334        if (ret == -ENOIOCTLCMD)
3335                ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3336
3337        return ret;
3338}
3339#endif
3340
3341int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3342{
3343        return sock->ops->bind(sock, addr, addrlen);
3344}
3345EXPORT_SYMBOL(kernel_bind);
3346
3347int kernel_listen(struct socket *sock, int backlog)
3348{
3349        return sock->ops->listen(sock, backlog);
3350}
3351EXPORT_SYMBOL(kernel_listen);
3352
3353int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3354{
3355        struct sock *sk = sock->sk;
3356        int err;
3357
3358        err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3359                               newsock);
3360        if (err < 0)
3361                goto done;
3362
3363        err = sock->ops->accept(sock, *newsock, flags);
3364        if (err < 0) {
3365                sock_release(*newsock);
3366                *newsock = NULL;
3367                goto done;
3368        }
3369
3370        (*newsock)->ops = sock->ops;
3371        __module_get((*newsock)->ops->owner);
3372
3373done:
3374        return err;
3375}
3376EXPORT_SYMBOL(kernel_accept);
3377
3378int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3379                   int flags)
3380{
3381        return sock->ops->connect(sock, addr, addrlen, flags);
3382}
3383EXPORT_SYMBOL(kernel_connect);
3384
3385int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3386                         int *addrlen)
3387{
3388        return sock->ops->getname(sock, addr, addrlen, 0);
3389}
3390EXPORT_SYMBOL(kernel_getsockname);
3391
3392int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3393                         int *addrlen)
3394{
3395        return sock->ops->getname(sock, addr, addrlen, 1);
3396}
3397EXPORT_SYMBOL(kernel_getpeername);
3398
3399int kernel_getsockopt(struct socket *sock, int level, int optname,
3400                        char *optval, int *optlen)
3401{
3402        mm_segment_t oldfs = get_fs();
3403        char __user *uoptval;
3404        int __user *uoptlen;
3405        int err;
3406
3407        uoptval = (char __user __force *) optval;
3408        uoptlen = (int __user __force *) optlen;
3409
3410        set_fs(KERNEL_DS);
3411        if (level == SOL_SOCKET)
3412                err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3413        else
3414                err = sock->ops->getsockopt(sock, level, optname, uoptval,
3415                                            uoptlen);
3416        set_fs(oldfs);
3417        return err;
3418}
3419EXPORT_SYMBOL(kernel_getsockopt);
3420
3421int kernel_setsockopt(struct socket *sock, int level, int optname,
3422                        char *optval, unsigned int optlen)
3423{
3424        mm_segment_t oldfs = get_fs();
3425        char __user *uoptval;
3426        int err;
3427
3428        uoptval = (char __user __force *) optval;
3429
3430        set_fs(KERNEL_DS);
3431        if (level == SOL_SOCKET)
3432                err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3433        else
3434                err = sock->ops->setsockopt(sock, level, optname, uoptval,
3435                                            optlen);
3436        set_fs(oldfs);
3437        return err;
3438}
3439EXPORT_SYMBOL(kernel_setsockopt);
3440
3441int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3442                    size_t size, int flags)
3443{
3444        if (sock->ops->sendpage)
3445                return sock->ops->sendpage(sock, page, offset, size, flags);
3446
3447        return sock_no_sendpage(sock, page, offset, size, flags);
3448}
3449EXPORT_SYMBOL(kernel_sendpage);
3450
3451int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3452{
3453        mm_segment_t oldfs = get_fs();
3454        int err;
3455
3456        set_fs(KERNEL_DS);
3457        err = sock->ops->ioctl(sock, cmd, arg);
3458        set_fs(oldfs);
3459
3460        return err;
3461}
3462EXPORT_SYMBOL(kernel_sock_ioctl);
3463
3464int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3465{
3466        return sock->ops->shutdown(sock, how);
3467}
3468EXPORT_SYMBOL(kernel_sock_shutdown);
3469
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.