fs/open.c

   1 /*
   2  *  linux/fs/open.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/string.h>
   8 #include <linux/mm.h>
   9 #include <linux/utime.h>
  10 #include <linux/file.h>
  11 #include <linux/smp_lock.h>
  12 #include <linux/quotaops.h>
  13 #include <linux/dnotify.h>
  14 #include <linux/module.h>
  15 #include <linux/slab.h>
  16 #include <linux/tty.h>
  17 #include <linux/namei.h>
  18 #include <linux/backing-dev.h>
  19 #include <linux/security.h>
  20 #include <linux/mount.h>
  21 #include <linux/vfs.h>
  22 #include <asm/uaccess.h>
  23 #include <linux/fs.h>
  24 #include <linux/pagemap.h>
  25 #include <linux/syscalls.h>
  26
  27 #include <asm/unistd.h>
  28
  29 int vfs_statfs(struct super_block *sb, struct kstatfs *buf)
  30 {
  31         int retval = -ENODEV;
  32
  33         if (sb) {
  34                 retval = -ENOSYS;
  35                 if (sb->s_op->statfs) {
  36                         memset(buf, 0, sizeof(*buf));
  37                         retval = security_sb_statfs(sb);
  38                         if (retval)
  39                                 return retval;
  40                         retval = sb->s_op->statfs(sb, buf);
  41                         if (retval == 0 && buf->f_frsize == 0)
  42                                 buf->f_frsize = buf->f_bsize;
  43                 }
  44         }
  45         return retval;
  46 }
  47
  48 EXPORT_SYMBOL(vfs_statfs);
  49
  50 static int vfs_statfs_native(struct super_block *sb, struct statfs *buf)
  51 {
  52         struct kstatfs st;
  53         int retval;
  54
  55         retval = vfs_statfs(sb, &st);
  56         if (retval)
  57                 return retval;
  58
  59         if (sizeof(*buf) == sizeof(st))
  60                 memcpy(buf, &st, sizeof(st));
  61         else {
  62                 if (sizeof buf->f_blocks == 4) {
  63                         if ((st.f_blocks | st.f_bfree | st.f_bavail) &
  64                             0xffffffff00000000ULL)
  65                                 return -EOVERFLOW;
  66                         /*
  67                          * f_files and f_ffree may be -1; it's okay to stuff
  68                          * that into 32 bits
  69                          */
  70                         if (st.f_files != -1 &&
  71                             (st.f_files & 0xffffffff00000000ULL))
  72                                 return -EOVERFLOW;
  73                         if (st.f_ffree != -1 &&
  74                             (st.f_ffree & 0xffffffff00000000ULL))
  75                                 return -EOVERFLOW;
  76                 }
  77
  78                 buf->f_type = st.f_type;
  79                 buf->f_bsize = st.f_bsize;
  80                 buf->f_blocks = st.f_blocks;
  81                 buf->f_bfree = st.f_bfree;
  82                 buf->f_bavail = st.f_bavail;
  83                 buf->f_files = st.f_files;
  84                 buf->f_ffree = st.f_ffree;
  85                 buf->f_fsid = st.f_fsid;
  86                 buf->f_namelen = st.f_namelen;
  87                 buf->f_frsize = st.f_frsize;
  88                 memset(buf->f_spare, 0, sizeof(buf->f_spare));
  89         }
  90         return 0;
  91 }
  92
  93 static int vfs_statfs64(struct super_block *sb, struct statfs64 *buf)
  94 {
  95         struct kstatfs st;
  96         int retval;
  97
  98         retval = vfs_statfs(sb, &st);
  99         if (retval)
 100                 return retval;
 101
 102         if (sizeof(*buf) == sizeof(st))
 103                 memcpy(buf, &st, sizeof(st));
 104         else {
 105                 buf->f_type = st.f_type;
 106                 buf->f_bsize = st.f_bsize;
 107                 buf->f_blocks = st.f_blocks;
 108                 buf->f_bfree = st.f_bfree;
 109                 buf->f_bavail = st.f_bavail;
 110                 buf->f_files = st.f_files;
 111                 buf->f_ffree = st.f_ffree;
 112                 buf->f_fsid = st.f_fsid;
 113                 buf->f_namelen = st.f_namelen;
 114                 buf->f_frsize = st.f_frsize;
 115                 memset(buf->f_spare, 0, sizeof(buf->f_spare));
 116         }
 117         return 0;
 118 }
 119
 120 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
 121 {
 122         struct nameidata nd;
 123         int error;
 124
 125         error = user_path_walk(path, &nd);
 126         if (!error) {
 127                 struct statfs tmp;
 128                 error = vfs_statfs_native(nd.dentry->d_inode->i_sb, &tmp);
 129                 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 130                         error = -EFAULT;
 131                 path_release(&nd);
 132         }
 133         return error;
 134 }
 135
 136
 137 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
 138 {
 139         struct nameidata nd;
 140         long error;
 141
 142         if (sz != sizeof(*buf))
 143                 return -EINVAL;
 144         error = user_path_walk(path, &nd);
 145         if (!error) {
 146                 struct statfs64 tmp;
 147                 error = vfs_statfs64(nd.dentry->d_inode->i_sb, &tmp);
 148                 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 149                         error = -EFAULT;
 150                 path_release(&nd);
 151         }
 152         return error;
 153 }
 154
 155
 156 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
 157 {
 158         struct file * file;
 159         struct statfs tmp;
 160         int error;
 161
 162         error = -EBADF;
 163         file = fget(fd);
 164         if (!file)
 165                 goto out;
 166         error = vfs_statfs_native(file->f_dentry->d_inode->i_sb, &tmp);
 167         if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 168                 error = -EFAULT;
 169         fput(file);
 170 out:
 171         return error;
 172 }
 173
 174 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
 175 {
 176         struct file * file;
 177         struct statfs64 tmp;
 178         int error;
 179
 180         if (sz != sizeof(*buf))
 181                 return -EINVAL;
 182
 183         error = -EBADF;
 184         file = fget(fd);
 185         if (!file)
 186                 goto out;
 187         error = vfs_statfs64(file->f_dentry->d_inode->i_sb, &tmp);
 188         if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 189                 error = -EFAULT;
 190         fput(file);
 191 out:
 192         return error;
 193 }
 194
 195 int do_truncate(struct dentry *dentry, loff_t length)
 196 {
 197         int err;
 198         struct iattr newattrs;
 199
 200         /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
 201         if (length < 0)
 202                 return -EINVAL;
 203
 204         newattrs.ia_size = length;
 205         newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME;
 206
 207         down(&dentry->d_inode->i_sem);
 208         err = notify_change(dentry, &newattrs);
 209         up(&dentry->d_inode->i_sem);
 210         return err;
 211 }
 212
 213 static inline long do_sys_truncate(const char __user * path, loff_t length)
 214 {
 215         struct nameidata nd;
 216         struct inode * inode;
 217         int error;
 218
 219         error = -EINVAL;
 220         if (length < 0) /* sorry, but loff_t says... */
 221                 goto out;
 222
 223         error = user_path_walk(path, &nd);
 224         if (error)
 225                 goto out;
 226         inode = nd.dentry->d_inode;
 227
 228         /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
 229         error = -EISDIR;
 230         if (S_ISDIR(inode->i_mode))
 231                 goto dput_and_out;
 232
 233         error = -EINVAL;
 234         if (!S_ISREG(inode->i_mode))
 235                 goto dput_and_out;
 236
 237         error = permission(inode,MAY_WRITE,&nd);
 238         if (error)
 239                 goto dput_and_out;
 240
 241         error = -EROFS;
 242         if (IS_RDONLY(inode))
 243                 goto dput_and_out;
 244
 245         error = -EPERM;
 246         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 247                 goto dput_and_out;
 248
 249         /*
 250          * Make sure that there are no leases.
 251          */
 252         error = break_lease(inode, FMODE_WRITE);
 253         if (error)
 254                 goto dput_and_out;
 255
 256         error = get_write_access(inode);
 257         if (error)
 258                 goto dput_and_out;
 259
 260         error = locks_verify_truncate(inode, NULL, length);
 261         if (!error) {
 262                 DQUOT_INIT(inode);
 263                 error = do_truncate(nd.dentry, length);
 264         }
 265         put_write_access(inode);
 266
 267 dput_and_out:
 268         path_release(&nd);
 269 out:
 270         return error;
 271 }
 272
 273 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
 274 {
 275         /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
 276         return do_sys_truncate(path, (long)length);
 277 }
 278
 279 static inline long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
 280 {
 281         struct inode * inode;
 282         struct dentry *dentry;
 283         struct file * file;
 284         int error;
 285
 286         error = -EINVAL;
 287         if (length < 0)
 288                 goto out;
 289         error = -EBADF;
 290         file = fget(fd);
 291         if (!file)
 292                 goto out;
 293
 294         /* explicitly opened as large or we are on 64-bit box */
 295         if (file->f_flags & O_LARGEFILE)
 296                 small = 0;
 297
 298         dentry = file->f_dentry;
 299         inode = dentry->d_inode;
 300         error = -EINVAL;
 301         if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
 302                 goto out_putf;
 303
 304         error = -EINVAL;
 305         /* Cannot ftruncate over 2^31 bytes without large file support */
 306         if (small && length > MAX_NON_LFS)
 307                 goto out_putf;
 308
 309         error = -EPERM;
 310         if (IS_APPEND(inode))
 311                 goto out_putf;
 312
 313         error = locks_verify_truncate(inode, file, length);
 314         if (!error)
 315                 error = do_truncate(dentry, length);
 316 out_putf:
 317         fput(file);
 318 out:
 319         return error;
 320 }
 321
 322 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
 323 {
 324         return do_sys_ftruncate(fd, length, 1);
 325 }
 326
 327 /* LFS versions of truncate are only needed on 32 bit machines */
 328 #if BITS_PER_LONG == 32
 329 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
 330 {
 331         return do_sys_truncate(path, length);
 332 }
 333
 334 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
 335 {
 336         return do_sys_ftruncate(fd, length, 0);
 337 }
 338 #endif
 339
 340 #ifdef __ARCH_WANT_SYS_UTIME
 341
 342 /*
 343  * sys_utime() can be implemented in user-level using sys_utimes().
 344  * Is this for backwards compatibility?  If so, why not move it
 345  * into the appropriate arch directory (for those architectures that
 346  * need it).
 347  */
 348
 349 /* If times==NULL, set access and modification to current time,
 350  * must be owner or have write permission.
 351  * Else, update from *times, must be owner or super user.
 352  */
 353 asmlinkage long sys_utime(char __user * filename, struct utimbuf __user * times)
 354 {
 355         int error;
 356         struct nameidata nd;
 357         struct inode * inode;
 358         struct iattr newattrs;
 359
 360         error = user_path_walk(filename, &nd);
 361         if (error)
 362                 goto out;
 363         inode = nd.dentry->d_inode;
 364
 365         error = -EROFS;
 366         if (IS_RDONLY(inode))
 367                 goto dput_and_out;
 368
 369         /* Don't worry, the checks are done in inode_change_ok() */
 370         newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
 371         if (times) {
 372                 error = -EPERM;
 373                 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
 374                         goto dput_and_out;
 375
 376                 error = get_user(newattrs.ia_atime.tv_sec, &times->actime);
 377                 newattrs.ia_atime.tv_nsec = 0;
 378                 if (!error)
 379                         error = get_user(newattrs.ia_mtime.tv_sec, &times->modtime);
 380                 newattrs.ia_mtime.tv_nsec = 0;
 381                 if (error)
 382                         goto dput_and_out;
 383
 384                 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
 385         } else {
 386                 error = -EACCES;
 387                 if (IS_IMMUTABLE(inode))
 388                         goto dput_and_out;
 389
 390                 if (current->fsuid != inode->i_uid &&
 391                     (error = permission(inode,MAY_WRITE,&nd)) != 0)
 392                         goto dput_and_out;
 393         }
 394         down(&inode->i_sem);
 395         error = notify_change(nd.dentry, &newattrs);
 396         up(&inode->i_sem);
 397 dput_and_out:
 398         path_release(&nd);
 399 out:
 400         return error;
 401 }
 402
 403 #endif
 404
 405 /* If times==NULL, set access and modification to current time,
 406  * must be owner or have write permission.
 407  * Else, update from *times, must be owner or super user.
 408  */
 409 long do_utimes(char __user * filename, struct timeval * times)
 410 {
 411         int error;
 412         struct nameidata nd;
 413         struct inode * inode;
 414         struct iattr newattrs;
 415
 416         error = user_path_walk(filename, &nd);
 417
 418         if (error)
 419                 goto out;
 420         inode = nd.dentry->d_inode;
 421
 422         error = -EROFS;
 423         if (IS_RDONLY(inode))
 424                 goto dput_and_out;
 425
 426         /* Don't worry, the checks are done in inode_change_ok() */
 427         newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
 428         if (times) {
 429                 error = -EPERM;
 430                 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
 431                         goto dput_and_out;
 432
 433                 newattrs.ia_atime.tv_sec = times[0].tv_sec;
 434                 newattrs.ia_atime.tv_nsec = times[0].tv_usec * 1000;
 435                 newattrs.ia_mtime.tv_sec = times[1].tv_sec;
 436                 newattrs.ia_mtime.tv_nsec = times[1].tv_usec * 1000;
 437                 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
 438         } else {
 439                 error = -EACCES;
 440                 if (IS_IMMUTABLE(inode))
 441                         goto dput_and_out;
 442
 443                 if (current->fsuid != inode->i_uid &&
 444                     (error = permission(inode,MAY_WRITE,&nd)) != 0)
 445                         goto dput_and_out;
 446         }
 447         down(&inode->i_sem);
 448         error = notify_change(nd.dentry, &newattrs);
 449         up(&inode->i_sem);
 450 dput_and_out:
 451         path_release(&nd);
 452 out:
 453         return error;
 454 }
 455
 456 asmlinkage long sys_utimes(char __user * filename, struct timeval __user * utimes)
 457 {
 458         struct timeval times[2];
 459
 460         if (utimes && copy_from_user(&times, utimes, sizeof(times)))
 461                 return -EFAULT;
 462         return do_utimes(filename, utimes ? times : NULL);
 463 }
 464
 465
 466 /*
 467  * access() needs to use the real uid/gid, not the effective uid/gid.
 468  * We do this by temporarily clearing all FS-related capabilities and
 469  * switching the fsuid/fsgid around to the real ones.
 470  */
 471 asmlinkage long sys_access(const char __user * filename, int mode)
 472 {
 473         struct nameidata nd;
 474         int old_fsuid, old_fsgid;
 475         kernel_cap_t old_cap;
 476         int res;
 477
 478         if (mode & ~S_IRWXO)    /* where's F_OK, X_OK, W_OK, R_OK? */
 479                 return -EINVAL;
 480
 481         old_fsuid = current->fsuid;
 482         old_fsgid = current->fsgid;
 483         old_cap = current->cap_effective;
 484
 485         current->fsuid = current->uid;
 486         current->fsgid = current->gid;
 487
 488         /*
 489          * Clear the capabilities if we switch to a non-root user
 490          *
 491          * FIXME: There is a race here against sys_capset.  The
 492          * capabilities can change yet we will restore the old
 493          * value below.  We should hold task_capabilities_lock,
 494          * but we cannot because user_path_walk can sleep.
 495          */
 496         if (current->uid)
 497                 cap_clear(current->cap_effective);
 498         else
 499                 current->cap_effective = current->cap_permitted;
 500
 501         res = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
 502         if (!res) {
 503                 res = permission(nd.dentry->d_inode, mode, &nd);
 504                 /* SuS v2 requires we report a read only fs too */
 505                 if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode)
 506                    && !special_file(nd.dentry->d_inode->i_mode))
 507                         res = -EROFS;
 508                 path_release(&nd);
 509         }
 510
 511         current->fsuid = old_fsuid;
 512         current->fsgid = old_fsgid;
 513         current->cap_effective = old_cap;
 514
 515         return res;
 516 }
 517
 518 asmlinkage long sys_chdir(const char __user * filename)
 519 {
 520         struct nameidata nd;
 521         int error;
 522
 523         error = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &nd);
 524         if (error)
 525                 goto out;
 526
 527         error = permission(nd.dentry->d_inode,MAY_EXEC,&nd);
 528         if (error)
 529                 goto dput_and_out;
 530
 531         set_fs_pwd(current->fs, nd.mnt, nd.dentry);
 532
 533 dput_and_out:
 534         path_release(&nd);
 535 out:
 536         return error;
 537 }
 538
 539 asmlinkage long sys_fchdir(unsigned int fd)
 540 {
 541         struct file *file;
 542         struct dentry *dentry;
 543         struct inode *inode;
 544         struct vfsmount *mnt;
 545         int error;
 546
 547         error = -EBADF;
 548         file = fget(fd);
 549         if (!file)
 550                 goto out;
 551
 552         dentry = file->f_dentry;
 553         mnt = file->f_vfsmnt;
 554         inode = dentry->d_inode;
 555
 556         error = -ENOTDIR;
 557         if (!S_ISDIR(inode->i_mode))
 558                 goto out_putf;
 559
 560         error = permission(inode, MAY_EXEC, NULL);
 561         if (!error)
 562                 set_fs_pwd(current->fs, mnt, dentry);
 563 out_putf:
 564         fput(file);
 565 out:
 566         return error;
 567 }
 568
 569 asmlinkage long sys_chroot(const char __user * filename)
 570 {
 571         struct nameidata nd;
 572         int error;
 573
 574         error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
 575         if (error)
 576                 goto out;
 577
 578         error = permission(nd.dentry->d_inode,MAY_EXEC,&nd);
 579         if (error)
 580                 goto dput_and_out;
 581
 582         error = -EPERM;
 583         if (!capable(CAP_SYS_CHROOT))
 584                 goto dput_and_out;
 585
 586         set_fs_root(current->fs, nd.mnt, nd.dentry);
 587         set_fs_altroot();
 588         error = 0;
 589 dput_and_out:
 590         path_release(&nd);
 591 out:
 592         return error;
 593 }
 594
 595 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
 596 {
 597         struct inode * inode;
 598         struct dentry * dentry;
 599         struct file * file;
 600         int err = -EBADF;
 601         struct iattr newattrs;
 602
 603         file = fget(fd);
 604         if (!file)
 605                 goto out;
 606
 607         dentry = file->f_dentry;
 608         inode = dentry->d_inode;
 609
 610         err = -EROFS;
 611         if (IS_RDONLY(inode))
 612                 goto out_putf;
 613         err = -EPERM;
 614         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 615                 goto out_putf;
 616         down(&inode->i_sem);
 617         if (mode == (mode_t) -1)
 618                 mode = inode->i_mode;
 619         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 620         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 621         err = notify_change(dentry, &newattrs);
 622         up(&inode->i_sem);
 623
 624 out_putf:
 625         fput(file);
 626 out:
 627         return err;
 628 }
 629
 630 asmlinkage long sys_chmod(const char __user * filename, mode_t mode)
 631 {
 632         struct nameidata nd;
 633         struct inode * inode;
 634         int error;
 635         struct iattr newattrs;
 636
 637         error = user_path_walk(filename, &nd);
 638         if (error)
 639                 goto out;
 640         inode = nd.dentry->d_inode;
 641
 642         error = -EROFS;
 643         if (IS_RDONLY(inode))
 644                 goto dput_and_out;
 645
 646         error = -EPERM;
 647         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 648                 goto dput_and_out;
 649
 650         down(&inode->i_sem);
 651         if (mode == (mode_t) -1)
 652                 mode = inode->i_mode;
 653         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 654         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 655         error = notify_change(nd.dentry, &newattrs);
 656         up(&inode->i_sem);
 657
 658 dput_and_out:
 659         path_release(&nd);
 660 out:
 661         return error;
 662 }
 663
 664 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
 665 {
 666         struct inode * inode;
 667         int error;
 668         struct iattr newattrs;
 669
 670         error = -ENOENT;
 671         if (!(inode = dentry->d_inode)) {
 672                 printk(KERN_ERR "chown_common: NULL inode\n");
 673                 goto out;
 674         }
 675         error = -EROFS;
 676         if (IS_RDONLY(inode))
 677                 goto out;
 678         error = -EPERM;
 679         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 680                 goto out;
 681         newattrs.ia_valid =  ATTR_CTIME;
 682         if (user != (uid_t) -1) {
 683                 newattrs.ia_valid |= ATTR_UID;
 684                 newattrs.ia_uid = user;
 685         }
 686         if (group != (gid_t) -1) {
 687                 newattrs.ia_valid |= ATTR_GID;
 688                 newattrs.ia_gid = group;
 689         }
 690         if (!S_ISDIR(inode->i_mode))
 691                 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
 692         down(&inode->i_sem);
 693         error = notify_change(dentry, &newattrs);
 694         up(&inode->i_sem);
 695 out:
 696         return error;
 697 }
 698
 699 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
 700 {
 701         struct nameidata nd;
 702         int error;
 703
 704         error = user_path_walk(filename, &nd);
 705         if (!error) {
 706                 error = chown_common(nd.dentry, user, group);
 707                 path_release(&nd);
 708         }
 709         return error;
 710 }
 711
 712 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
 713 {
 714         struct nameidata nd;
 715         int error;
 716
 717         error = user_path_walk_link(filename, &nd);
 718         if (!error) {
 719                 error = chown_common(nd.dentry, user, group);
 720                 path_release(&nd);
 721         }
 722         return error;
 723 }
 724
 725
 726 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
 727 {
 728         struct file * file;
 729         int error = -EBADF;
 730
 731         file = fget(fd);
 732         if (file) {
 733                 error = chown_common(file->f_dentry, user, group);
 734                 fput(file);
 735         }
 736         return error;
 737 }
 738
 739 /*
 740  * Note that while the flag value (low two bits) for sys_open means:
 741  *      00 - read-only
 742  *      01 - write-only
 743  *      10 - read-write
 744  *      11 - special
 745  * it is changed into
 746  *      00 - no permissions needed
 747  *      01 - read-permission
 748  *      10 - write-permission
 749  *      11 - read-write
 750  * for the internal routines (ie open_namei()/follow_link() etc). 00 is
 751  * used by symlinks.
 752  */
 753 struct file *filp_open(const char * filename, int flags, int mode)
 754 {
 755         int namei_flags, error;
 756         struct nameidata nd;
 757
 758         namei_flags = flags;
 759         if ((namei_flags+1) & O_ACCMODE)
 760                 namei_flags++;
 761         if (namei_flags & O_TRUNC)
 762                 namei_flags |= 2;
 763
 764         error = open_namei(filename, namei_flags, mode, &nd);
 765         if (!error)
 766                 return dentry_open(nd.dentry, nd.mnt, flags);
 767
 768         return ERR_PTR(error);
 769 }
 770
 771 EXPORT_SYMBOL(filp_open);
 772
 773 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
 774 {
 775         struct file * f;
 776         struct inode *inode;
 777         int error;
 778
 779         error = -ENFILE;
 780         f = get_empty_filp();
 781         if (!f)
 782                 goto cleanup_dentry;
 783         f->f_flags = flags;
 784         f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
 785         inode = dentry->d_inode;
 786         if (f->f_mode & FMODE_WRITE) {
 787                 error = get_write_access(inode);
 788                 if (error)
 789                         goto cleanup_file;
 790         }
 791
 792         f->f_mapping = inode->i_mapping;
 793         f->f_dentry = dentry;
 794         f->f_vfsmnt = mnt;
 795         f->f_pos = 0;
 796         f->f_op = fops_get(inode->i_fop);
 797         file_move(f, &inode->i_sb->s_files);
 798
 799         if (f->f_op && f->f_op->open) {
 800                 error = f->f_op->open(inode,f);
 801                 if (error)
 802                         goto cleanup_all;
 803         }
 804         f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
 805
 806         file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
 807
 808         /* NB: we're sure to have correct a_ops only after f_op->open */
 809         if (f->f_flags & O_DIRECT) {
 810                 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) {
 811                         fput(f);
 812                         f = ERR_PTR(-EINVAL);
 813                 }
 814         }
 815
 816         return f;
 817
 818 cleanup_all:
 819         fops_put(f->f_op);
 820         if (f->f_mode & FMODE_WRITE)
 821                 put_write_access(inode);
 822         file_kill(f);
 823         f->f_dentry = NULL;
 824         f->f_vfsmnt = NULL;
 825 cleanup_file:
 826         put_filp(f);
 827 cleanup_dentry:
 828         dput(dentry);
 829         mntput(mnt);
 830         return ERR_PTR(error);
 831 }
 832
 833 EXPORT_SYMBOL(dentry_open);
 834
 835 /*
 836  * Find an empty file descriptor entry, and mark it busy.
 837  */
 838 int get_unused_fd(void)
 839 {
 840         struct files_struct * files = current->files;
 841         int fd, error;
 842
 843         error = -EMFILE;
 844         spin_lock(&files->file_lock);
 845
 846 repeat:
 847         fd = find_next_zero_bit(files->open_fds->fds_bits,
 848                                 files->max_fdset,
 849                                 files->next_fd);
 850
 851         /*
 852          * N.B. For clone tasks sharing a files structure, this test
 853          * will limit the total number of files that can be opened.
 854          */
 855         if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
 856                 goto out;
 857
 858         /* Do we need to expand the fd array or fd set?  */
 859         error = expand_files(files, fd);
 860         if (error < 0)
 861                 goto out;
 862
 863         if (error) {
 864                 /*
 865                  * If we needed to expand the fs array we
 866                  * might have blocked - try again.
 867                  */
 868                 error = -EMFILE;
 869                 goto repeat;
 870         }
 871
 872         FD_SET(fd, files->open_fds);
 873         FD_CLR(fd, files->close_on_exec);
 874         files->next_fd = fd + 1;
 875 #if 1
 876         /* Sanity check */
 877         if (files->fd[fd] != NULL) {
 878                 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
 879                 files->fd[fd] = NULL;
 880         }
 881 #endif
 882         error = fd;
 883
 884 out:
 885         spin_unlock(&files->file_lock);
 886         return error;
 887 }
 888
 889 EXPORT_SYMBOL(get_unused_fd);
 890
 891 static inline void __put_unused_fd(struct files_struct *files, unsigned int fd)
 892 {
 893         __FD_CLR(fd, files->open_fds);
 894         if (fd < files->next_fd)
 895                 files->next_fd = fd;
 896 }
 897
 898 void fastcall put_unused_fd(unsigned int fd)
 899 {
 900         struct files_struct *files = current->files;
 901         spin_lock(&files->file_lock);
 902         __put_unused_fd(files, fd);
 903         spin_unlock(&files->file_lock);
 904 }
 905
 906 EXPORT_SYMBOL(put_unused_fd);
 907
 908 /*
 909  * Install a file pointer in the fd array.
 910  *
 911  * The VFS is full of places where we drop the files lock between
 912  * setting the open_fds bitmap and installing the file in the file
 913  * array.  At any such point, we are vulnerable to a dup2() race
 914  * installing a file in the array before us.  We need to detect this and
 915  * fput() the struct file we are about to overwrite in this case.
 916  *
 917  * It should never happen - if we allow dup2() do it, _really_ bad things
 918  * will follow.
 919  */
 920
 921 void fastcall fd_install(unsigned int fd, struct file * file)
 922 {
 923         struct files_struct *files = current->files;
 924         spin_lock(&files->file_lock);
 925         if (unlikely(files->fd[fd] != NULL))
 926                 BUG();
 927         files->fd[fd] = file;
 928         spin_unlock(&files->file_lock);
 929 }
 930
 931 EXPORT_SYMBOL(fd_install);
 932
 933 asmlinkage long sys_open(const char __user * filename, int flags, int mode)
 934 {
 935         char * tmp;
 936         int fd, error;
 937
 938 #if BITS_PER_LONG != 32
 939         flags |= O_LARGEFILE;
 940 #endif
 941         tmp = getname(filename);
 942         fd = PTR_ERR(tmp);
 943         if (!IS_ERR(tmp)) {
 944                 fd = get_unused_fd();
 945                 if (fd >= 0) {
 946                         struct file *f = filp_open(tmp, flags, mode);
 947                         error = PTR_ERR(f);
 948                         if (IS_ERR(f))
 949                                 goto out_error;
 950                         fd_install(fd, f);
 951                 }
 952 out:
 953                 putname(tmp);
 954         }
 955         return fd;
 956
 957 out_error:
 958         put_unused_fd(fd);
 959         fd = error;
 960         goto out;
 961 }
 962 EXPORT_SYMBOL_GPL(sys_open);
 963
 964 #ifndef __alpha__
 965
 966 /*
 967  * For backward compatibility?  Maybe this should be moved
 968  * into arch/i386 instead?
 969  */
 970 asmlinkage long sys_creat(const char __user * pathname, int mode)
 971 {
 972         return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
 973 }
 974
 975 #endif
 976
 977 /*
 978  * "id" is the POSIX thread ID. We use the
 979  * files pointer for this..
 980  */
 981 int filp_close(struct file *filp, fl_owner_t id)
 982 {
 983         int retval;
 984
 985         /* Report and clear outstanding errors */
 986         retval = filp->f_error;
 987         if (retval)
 988                 filp->f_error = 0;
 989
 990         if (!file_count(filp)) {
 991                 printk(KERN_ERR "VFS: Close: file count is 0\n");
 992                 return retval;
 993         }
 994
 995         if (filp->f_op && filp->f_op->flush) {
 996                 int err = filp->f_op->flush(filp);
 997                 if (!retval)
 998                         retval = err;
 999         }
1000
1001         dnotify_flush(filp, id);
1002         locks_remove_posix(filp, id);
1003         fput(filp);
1004         return retval;
1005 }
1006
1007 EXPORT_SYMBOL(filp_close);
1008
1009 /*
1010  * Careful here! We test whether the file pointer is NULL before
1011  * releasing the fd. This ensures that one clone task can't release
1012  * an fd while another clone is opening it.
1013  */
1014 asmlinkage long sys_close(unsigned int fd)
1015 {
1016         struct file * filp;
1017         struct files_struct *files = current->files;
1018
1019         spin_lock(&files->file_lock);
1020         if (fd >= files->max_fds)
1021                 goto out_unlock;
1022         filp = files->fd[fd];
1023         if (!filp)
1024                 goto out_unlock;
1025         files->fd[fd] = NULL;
1026         FD_CLR(fd, files->close_on_exec);
1027         __put_unused_fd(files, fd);
1028         spin_unlock(&files->file_lock);
1029         return filp_close(filp, files);
1030
1031 out_unlock:
1032         spin_unlock(&files->file_lock);
1033         return -EBADF;
1034 }
1035
1036 EXPORT_SYMBOL(sys_close);
1037
1038 /*
1039  * This routine simulates a hangup on the tty, to arrange that users
1040  * are given clean terminals at login time.
1041  */
1042 asmlinkage long sys_vhangup(void)
1043 {
1044         if (capable(CAP_SYS_TTY_CONFIG)) {
1045                 tty_vhangup(current->signal->tty);
1046                 return 0;
1047         }
1048         return -EPERM;
1049 }
1050
1051 /*
1052  * Called when an inode is about to be open.
1053  * We use this to disallow opening large files on 32bit systems if
1054  * the caller didn't specify O_LARGEFILE.  On 64bit systems we force
1055  * on this flag in sys_open.
1056  */
1057 int generic_file_open(struct inode * inode, struct file * filp)
1058 {
1059         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1060                 return -EFBIG;
1061         return 0;
1062 }
1063
1064 EXPORT_SYMBOL(generic_file_open);
1065
1066 /*
1067  * This is used by subsystems that don't want seekable
1068  * file descriptors
1069  */
1070 int nonseekable_open(struct inode *inode, struct file *filp)
1071 {
1072         filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1073         return 0;
1074 }
1075
1076 EXPORT_SYMBOL(nonseekable_open);