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/fsnotify.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/capability.h>
  20 #include <linux/security.h>
  21 #include <linux/mount.h>
  22 #include <linux/vfs.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/fs.h>
  25 #include <linux/personality.h>
  26 #include <linux/pagemap.h>
  27 #include <linux/syscalls.h>
  28 #include <linux/rcupdate.h>
  29
  30 #include <asm/unistd.h>
  31
  32 int vfs_statfs(struct super_block *sb, struct kstatfs *buf)
  33 {
  34         int retval = -ENODEV;
  35
  36         if (sb) {
  37                 retval = -ENOSYS;
  38                 if (sb->s_op->statfs) {
  39                         memset(buf, 0, sizeof(*buf));
  40                         retval = security_sb_statfs(sb);
  41                         if (retval)
  42                                 return retval;
  43                         retval = sb->s_op->statfs(sb, buf);
  44                         if (retval == 0 && buf->f_frsize == 0)
  45                                 buf->f_frsize = buf->f_bsize;
  46                 }
  47         }
  48         return retval;
  49 }
  50
  51 EXPORT_SYMBOL(vfs_statfs);
  52
  53 static int vfs_statfs_native(struct super_block *sb, struct statfs *buf)
  54 {
  55         struct kstatfs st;
  56         int retval;
  57
  58         retval = vfs_statfs(sb, &st);
  59         if (retval)
  60                 return retval;
  61
  62         if (sizeof(*buf) == sizeof(st))
  63                 memcpy(buf, &st, sizeof(st));
  64         else {
  65                 if (sizeof buf->f_blocks == 4) {
  66                         if ((st.f_blocks | st.f_bfree | st.f_bavail) &
  67                             0xffffffff00000000ULL)
  68                                 return -EOVERFLOW;
  69                         /*
  70                          * f_files and f_ffree may be -1; it's okay to stuff
  71                          * that into 32 bits
  72                          */
  73                         if (st.f_files != -1 &&
  74                             (st.f_files & 0xffffffff00000000ULL))
  75                                 return -EOVERFLOW;
  76                         if (st.f_ffree != -1 &&
  77                             (st.f_ffree & 0xffffffff00000000ULL))
  78                                 return -EOVERFLOW;
  79                 }
  80
  81                 buf->f_type = st.f_type;
  82                 buf->f_bsize = st.f_bsize;
  83                 buf->f_blocks = st.f_blocks;
  84                 buf->f_bfree = st.f_bfree;
  85                 buf->f_bavail = st.f_bavail;
  86                 buf->f_files = st.f_files;
  87                 buf->f_ffree = st.f_ffree;
  88                 buf->f_fsid = st.f_fsid;
  89                 buf->f_namelen = st.f_namelen;
  90                 buf->f_frsize = st.f_frsize;
  91                 memset(buf->f_spare, 0, sizeof(buf->f_spare));
  92         }
  93         return 0;
  94 }
  95
  96 static int vfs_statfs64(struct super_block *sb, struct statfs64 *buf)
  97 {
  98         struct kstatfs st;
  99         int retval;
 100
 101         retval = vfs_statfs(sb, &st);
 102         if (retval)
 103                 return retval;
 104
 105         if (sizeof(*buf) == sizeof(st))
 106                 memcpy(buf, &st, sizeof(st));
 107         else {
 108                 buf->f_type = st.f_type;
 109                 buf->f_bsize = st.f_bsize;
 110                 buf->f_blocks = st.f_blocks;
 111                 buf->f_bfree = st.f_bfree;
 112                 buf->f_bavail = st.f_bavail;
 113                 buf->f_files = st.f_files;
 114                 buf->f_ffree = st.f_ffree;
 115                 buf->f_fsid = st.f_fsid;
 116                 buf->f_namelen = st.f_namelen;
 117                 buf->f_frsize = st.f_frsize;
 118                 memset(buf->f_spare, 0, sizeof(buf->f_spare));
 119         }
 120         return 0;
 121 }
 122
 123 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
 124 {
 125         struct nameidata nd;
 126         int error;
 127
 128         error = user_path_walk(path, &nd);
 129         if (!error) {
 130                 struct statfs tmp;
 131                 error = vfs_statfs_native(nd.dentry->d_inode->i_sb, &tmp);
 132                 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 133                         error = -EFAULT;
 134                 path_release(&nd);
 135         }
 136         return error;
 137 }
 138
 139
 140 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
 141 {
 142         struct nameidata nd;
 143         long error;
 144
 145         if (sz != sizeof(*buf))
 146                 return -EINVAL;
 147         error = user_path_walk(path, &nd);
 148         if (!error) {
 149                 struct statfs64 tmp;
 150                 error = vfs_statfs64(nd.dentry->d_inode->i_sb, &tmp);
 151                 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 152                         error = -EFAULT;
 153                 path_release(&nd);
 154         }
 155         return error;
 156 }
 157
 158
 159 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
 160 {
 161         struct file * file;
 162         struct statfs tmp;
 163         int error;
 164
 165         error = -EBADF;
 166         file = fget(fd);
 167         if (!file)
 168                 goto out;
 169         error = vfs_statfs_native(file->f_dentry->d_inode->i_sb, &tmp);
 170         if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 171                 error = -EFAULT;
 172         fput(file);
 173 out:
 174         return error;
 175 }
 176
 177 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
 178 {
 179         struct file * file;
 180         struct statfs64 tmp;
 181         int error;
 182
 183         if (sz != sizeof(*buf))
 184                 return -EINVAL;
 185
 186         error = -EBADF;
 187         file = fget(fd);
 188         if (!file)
 189                 goto out;
 190         error = vfs_statfs64(file->f_dentry->d_inode->i_sb, &tmp);
 191         if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 192                 error = -EFAULT;
 193         fput(file);
 194 out:
 195         return error;
 196 }
 197
 198 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
 199         struct file *filp)
 200 {
 201         int err;
 202         struct iattr newattrs;
 203
 204         /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
 205         if (length < 0)
 206                 return -EINVAL;
 207
 208         newattrs.ia_size = length;
 209         newattrs.ia_valid = ATTR_SIZE | time_attrs;
 210         if (filp) {
 211                 newattrs.ia_file = filp;
 212                 newattrs.ia_valid |= ATTR_FILE;
 213         }
 214
 215         mutex_lock(&dentry->d_inode->i_mutex);
 216         err = notify_change(dentry, &newattrs);
 217         mutex_unlock(&dentry->d_inode->i_mutex);
 218         return err;
 219 }
 220
 221 static long do_sys_truncate(const char __user * path, loff_t length)
 222 {
 223         struct nameidata nd;
 224         struct inode * inode;
 225         int error;
 226
 227         error = -EINVAL;
 228         if (length < 0) /* sorry, but loff_t says... */
 229                 goto out;
 230
 231         error = user_path_walk(path, &nd);
 232         if (error)
 233                 goto out;
 234         inode = nd.dentry->d_inode;
 235
 236         /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
 237         error = -EISDIR;
 238         if (S_ISDIR(inode->i_mode))
 239                 goto dput_and_out;
 240
 241         error = -EINVAL;
 242         if (!S_ISREG(inode->i_mode))
 243                 goto dput_and_out;
 244
 245         error = vfs_permission(&nd, MAY_WRITE);
 246         if (error)
 247                 goto dput_and_out;
 248
 249         error = -EROFS;
 250         if (IS_RDONLY(inode))
 251                 goto dput_and_out;
 252
 253         error = -EPERM;
 254         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 255                 goto dput_and_out;
 256
 257         /*
 258          * Make sure that there are no leases.
 259          */
 260         error = break_lease(inode, FMODE_WRITE);
 261         if (error)
 262                 goto dput_and_out;
 263
 264         error = get_write_access(inode);
 265         if (error)
 266                 goto dput_and_out;
 267
 268         error = locks_verify_truncate(inode, NULL, length);
 269         if (!error) {
 270                 DQUOT_INIT(inode);
 271                 error = do_truncate(nd.dentry, length, 0, NULL);
 272         }
 273         put_write_access(inode);
 274
 275 dput_and_out:
 276         path_release(&nd);
 277 out:
 278         return error;
 279 }
 280
 281 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
 282 {
 283         /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
 284         return do_sys_truncate(path, (long)length);
 285 }
 286
 287 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
 288 {
 289         struct inode * inode;
 290         struct dentry *dentry;
 291         struct file * file;
 292         int error;
 293
 294         error = -EINVAL;
 295         if (length < 0)
 296                 goto out;
 297         error = -EBADF;
 298         file = fget(fd);
 299         if (!file)
 300                 goto out;
 301
 302         /* explicitly opened as large or we are on 64-bit box */
 303         if (file->f_flags & O_LARGEFILE)
 304                 small = 0;
 305
 306         dentry = file->f_dentry;
 307         inode = dentry->d_inode;
 308         error = -EINVAL;
 309         if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
 310                 goto out_putf;
 311
 312         error = -EINVAL;
 313         /* Cannot ftruncate over 2^31 bytes without large file support */
 314         if (small && length > MAX_NON_LFS)
 315                 goto out_putf;
 316
 317         error = -EPERM;
 318         if (IS_APPEND(inode))
 319                 goto out_putf;
 320
 321         error = locks_verify_truncate(inode, file, length);
 322         if (!error)
 323                 error = do_truncate(dentry, length, 0, file);
 324 out_putf:
 325         fput(file);
 326 out:
 327         return error;
 328 }
 329
 330 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
 331 {
 332         return do_sys_ftruncate(fd, length, 1);
 333 }
 334
 335 /* LFS versions of truncate are only needed on 32 bit machines */
 336 #if BITS_PER_LONG == 32
 337 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
 338 {
 339         return do_sys_truncate(path, length);
 340 }
 341
 342 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
 343 {
 344         return do_sys_ftruncate(fd, length, 0);
 345 }
 346 #endif
 347
 348 #ifdef __ARCH_WANT_SYS_UTIME
 349
 350 /*
 351  * sys_utime() can be implemented in user-level using sys_utimes().
 352  * Is this for backwards compatibility?  If so, why not move it
 353  * into the appropriate arch directory (for those architectures that
 354  * need it).
 355  */
 356
 357 /* If times==NULL, set access and modification to current time,
 358  * must be owner or have write permission.
 359  * Else, update from *times, must be owner or super user.
 360  */
 361 asmlinkage long sys_utime(char __user * filename, struct utimbuf __user * times)
 362 {
 363         int error;
 364         struct nameidata nd;
 365         struct inode * inode;
 366         struct iattr newattrs;
 367
 368         error = user_path_walk(filename, &nd);
 369         if (error)
 370                 goto out;
 371         inode = nd.dentry->d_inode;
 372
 373         error = -EROFS;
 374         if (IS_RDONLY(inode))
 375                 goto dput_and_out;
 376
 377         /* Don't worry, the checks are done in inode_change_ok() */
 378         newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
 379         if (times) {
 380                 error = -EPERM;
 381                 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
 382                         goto dput_and_out;
 383
 384                 error = get_user(newattrs.ia_atime.tv_sec, &times->actime);
 385                 newattrs.ia_atime.tv_nsec = 0;
 386                 if (!error)
 387                         error = get_user(newattrs.ia_mtime.tv_sec, &times->modtime);
 388                 newattrs.ia_mtime.tv_nsec = 0;
 389                 if (error)
 390                         goto dput_and_out;
 391
 392                 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
 393         } else {
 394                 error = -EACCES;
 395                 if (IS_IMMUTABLE(inode))
 396                         goto dput_and_out;
 397
 398                 if (current->fsuid != inode->i_uid &&
 399                     (error = vfs_permission(&nd, MAY_WRITE)) != 0)
 400                         goto dput_and_out;
 401         }
 402         mutex_lock(&inode->i_mutex);
 403         error = notify_change(nd.dentry, &newattrs);
 404         mutex_unlock(&inode->i_mutex);
 405 dput_and_out:
 406         path_release(&nd);
 407 out:
 408         return error;
 409 }
 410
 411 #endif
 412
 413 /* If times==NULL, set access and modification to current time,
 414  * must be owner or have write permission.
 415  * Else, update from *times, must be owner or super user.
 416  */
 417 long do_utimes(char __user * filename, struct timeval * times)
 418 {
 419         int error;
 420         struct nameidata nd;
 421         struct inode * inode;
 422         struct iattr newattrs;
 423
 424         error = user_path_walk(filename, &nd);
 425
 426         if (error)
 427                 goto out;
 428         inode = nd.dentry->d_inode;
 429
 430         error = -EROFS;
 431         if (IS_RDONLY(inode))
 432                 goto dput_and_out;
 433
 434         /* Don't worry, the checks are done in inode_change_ok() */
 435         newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
 436         if (times) {
 437                 error = -EPERM;
 438                 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
 439                         goto dput_and_out;
 440
 441                 newattrs.ia_atime.tv_sec = times[0].tv_sec;
 442                 newattrs.ia_atime.tv_nsec = times[0].tv_usec * 1000;
 443                 newattrs.ia_mtime.tv_sec = times[1].tv_sec;
 444                 newattrs.ia_mtime.tv_nsec = times[1].tv_usec * 1000;
 445                 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
 446         } else {
 447                 error = -EACCES;
 448                 if (IS_IMMUTABLE(inode))
 449                         goto dput_and_out;
 450
 451                 if (current->fsuid != inode->i_uid &&
 452                     (error = vfs_permission(&nd, MAY_WRITE)) != 0)
 453                         goto dput_and_out;
 454         }
 455         mutex_lock(&inode->i_mutex);
 456         error = notify_change(nd.dentry, &newattrs);
 457         mutex_unlock(&inode->i_mutex);
 458 dput_and_out:
 459         path_release(&nd);
 460 out:
 461         return error;
 462 }
 463
 464 asmlinkage long sys_utimes(char __user * filename, struct timeval __user * utimes)
 465 {
 466         struct timeval times[2];
 467
 468         if (utimes && copy_from_user(&times, utimes, sizeof(times)))
 469                 return -EFAULT;
 470         return do_utimes(filename, utimes ? times : NULL);
 471 }
 472
 473
 474 /*
 475  * access() needs to use the real uid/gid, not the effective uid/gid.
 476  * We do this by temporarily clearing all FS-related capabilities and
 477  * switching the fsuid/fsgid around to the real ones.
 478  */
 479 asmlinkage long sys_access(const char __user * filename, int mode)
 480 {
 481         struct nameidata nd;
 482         int old_fsuid, old_fsgid;
 483         kernel_cap_t old_cap;
 484         int res;
 485
 486         if (mode & ~S_IRWXO)    /* where's F_OK, X_OK, W_OK, R_OK? */
 487                 return -EINVAL;
 488
 489         old_fsuid = current->fsuid;
 490         old_fsgid = current->fsgid;
 491         old_cap = current->cap_effective;
 492
 493         current->fsuid = current->uid;
 494         current->fsgid = current->gid;
 495
 496         /*
 497          * Clear the capabilities if we switch to a non-root user
 498          *
 499          * FIXME: There is a race here against sys_capset.  The
 500          * capabilities can change yet we will restore the old
 501          * value below.  We should hold task_capabilities_lock,
 502          * but we cannot because user_path_walk can sleep.
 503          */
 504         if (current->uid)
 505                 cap_clear(current->cap_effective);
 506         else
 507                 current->cap_effective = current->cap_permitted;
 508
 509         res = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
 510         if (!res) {
 511                 res = vfs_permission(&nd, mode);
 512                 /* SuS v2 requires we report a read only fs too */
 513                 if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode)
 514                    && !special_file(nd.dentry->d_inode->i_mode))
 515                         res = -EROFS;
 516                 path_release(&nd);
 517         }
 518
 519         current->fsuid = old_fsuid;
 520         current->fsgid = old_fsgid;
 521         current->cap_effective = old_cap;
 522
 523         return res;
 524 }
 525
 526 asmlinkage long sys_chdir(const char __user * filename)
 527 {
 528         struct nameidata nd;
 529         int error;
 530
 531         error = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &nd);
 532         if (error)
 533                 goto out;
 534
 535         error = vfs_permission(&nd, MAY_EXEC);
 536         if (error)
 537                 goto dput_and_out;
 538
 539         set_fs_pwd(current->fs, nd.mnt, nd.dentry);
 540
 541 dput_and_out:
 542         path_release(&nd);
 543 out:
 544         return error;
 545 }
 546
 547 asmlinkage long sys_fchdir(unsigned int fd)
 548 {
 549         struct file *file;
 550         struct dentry *dentry;
 551         struct inode *inode;
 552         struct vfsmount *mnt;
 553         int error;
 554
 555         error = -EBADF;
 556         file = fget(fd);
 557         if (!file)
 558                 goto out;
 559
 560         dentry = file->f_dentry;
 561         mnt = file->f_vfsmnt;
 562         inode = dentry->d_inode;
 563
 564         error = -ENOTDIR;
 565         if (!S_ISDIR(inode->i_mode))
 566                 goto out_putf;
 567
 568         error = file_permission(file, MAY_EXEC);
 569         if (!error)
 570                 set_fs_pwd(current->fs, mnt, dentry);
 571 out_putf:
 572         fput(file);
 573 out:
 574         return error;
 575 }
 576
 577 asmlinkage long sys_chroot(const char __user * filename)
 578 {
 579         struct nameidata nd;
 580         int error;
 581
 582         error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
 583         if (error)
 584                 goto out;
 585
 586         error = vfs_permission(&nd, MAY_EXEC);
 587         if (error)
 588                 goto dput_and_out;
 589
 590         error = -EPERM;
 591         if (!capable(CAP_SYS_CHROOT))
 592                 goto dput_and_out;
 593
 594         set_fs_root(current->fs, nd.mnt, nd.dentry);
 595         set_fs_altroot();
 596         error = 0;
 597 dput_and_out:
 598         path_release(&nd);
 599 out:
 600         return error;
 601 }
 602
 603 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
 604 {
 605         struct inode * inode;
 606         struct dentry * dentry;
 607         struct file * file;
 608         int err = -EBADF;
 609         struct iattr newattrs;
 610
 611         file = fget(fd);
 612         if (!file)
 613                 goto out;
 614
 615         dentry = file->f_dentry;
 616         inode = dentry->d_inode;
 617
 618         err = -EROFS;
 619         if (IS_RDONLY(inode))
 620                 goto out_putf;
 621         err = -EPERM;
 622         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 623                 goto out_putf;
 624         mutex_lock(&inode->i_mutex);
 625         if (mode == (mode_t) -1)
 626                 mode = inode->i_mode;
 627         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 628         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 629         err = notify_change(dentry, &newattrs);
 630         mutex_unlock(&inode->i_mutex);
 631
 632 out_putf:
 633         fput(file);
 634 out:
 635         return err;
 636 }
 637
 638 asmlinkage long sys_chmod(const char __user * filename, mode_t mode)
 639 {
 640         struct nameidata nd;
 641         struct inode * inode;
 642         int error;
 643         struct iattr newattrs;
 644
 645         error = user_path_walk(filename, &nd);
 646         if (error)
 647                 goto out;
 648         inode = nd.dentry->d_inode;
 649
 650         error = -EROFS;
 651         if (IS_RDONLY(inode))
 652                 goto dput_and_out;
 653
 654         error = -EPERM;
 655         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 656                 goto dput_and_out;
 657
 658         mutex_lock(&inode->i_mutex);
 659         if (mode == (mode_t) -1)
 660                 mode = inode->i_mode;
 661         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 662         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 663         error = notify_change(nd.dentry, &newattrs);
 664         mutex_unlock(&inode->i_mutex);
 665
 666 dput_and_out:
 667         path_release(&nd);
 668 out:
 669         return error;
 670 }
 671
 672 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
 673 {
 674         struct inode * inode;
 675         int error;
 676         struct iattr newattrs;
 677
 678         error = -ENOENT;
 679         if (!(inode = dentry->d_inode)) {
 680                 printk(KERN_ERR "chown_common: NULL inode\n");
 681                 goto out;
 682         }
 683         error = -EROFS;
 684         if (IS_RDONLY(inode))
 685                 goto out;
 686         error = -EPERM;
 687         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 688                 goto out;
 689         newattrs.ia_valid =  ATTR_CTIME;
 690         if (user != (uid_t) -1) {
 691                 newattrs.ia_valid |= ATTR_UID;
 692                 newattrs.ia_uid = user;
 693         }
 694         if (group != (gid_t) -1) {
 695                 newattrs.ia_valid |= ATTR_GID;
 696                 newattrs.ia_gid = group;
 697         }
 698         if (!S_ISDIR(inode->i_mode))
 699                 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
 700         mutex_lock(&inode->i_mutex);
 701         error = notify_change(dentry, &newattrs);
 702         mutex_unlock(&inode->i_mutex);
 703 out:
 704         return error;
 705 }
 706
 707 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
 708 {
 709         struct nameidata nd;
 710         int error;
 711
 712         error = user_path_walk(filename, &nd);
 713         if (!error) {
 714                 error = chown_common(nd.dentry, user, group);
 715                 path_release(&nd);
 716         }
 717         return error;
 718 }
 719
 720 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
 721 {
 722         struct nameidata nd;
 723         int error;
 724
 725         error = user_path_walk_link(filename, &nd);
 726         if (!error) {
 727                 error = chown_common(nd.dentry, user, group);
 728                 path_release(&nd);
 729         }
 730         return error;
 731 }
 732
 733
 734 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
 735 {
 736         struct file * file;
 737         int error = -EBADF;
 738
 739         file = fget(fd);
 740         if (file) {
 741                 error = chown_common(file->f_dentry, user, group);
 742                 fput(file);
 743         }
 744         return error;
 745 }
 746
 747 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
 748                                         int flags, struct file *f,
 749                                         int (*open)(struct inode *, struct file *))
 750 {
 751         struct inode *inode;
 752         int error;
 753
 754         f->f_flags = flags;
 755         f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
 756                                 FMODE_PREAD | FMODE_PWRITE;
 757         inode = dentry->d_inode;
 758         if (f->f_mode & FMODE_WRITE) {
 759                 error = get_write_access(inode);
 760                 if (error)
 761                         goto cleanup_file;
 762         }
 763
 764         f->f_mapping = inode->i_mapping;
 765         f->f_dentry = dentry;
 766         f->f_vfsmnt = mnt;
 767         f->f_pos = 0;
 768         f->f_op = fops_get(inode->i_fop);
 769         file_move(f, &inode->i_sb->s_files);
 770
 771         if (!open && f->f_op)
 772                 open = f->f_op->open;
 773         if (open) {
 774                 error = open(inode, f);
 775                 if (error)
 776                         goto cleanup_all;
 777         }
 778
 779         f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
 780
 781         file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
 782
 783         /* NB: we're sure to have correct a_ops only after f_op->open */
 784         if (f->f_flags & O_DIRECT) {
 785                 if (!f->f_mapping->a_ops ||
 786                     ((!f->f_mapping->a_ops->direct_IO) &&
 787                     (!f->f_mapping->a_ops->get_xip_page))) {
 788                         fput(f);
 789                         f = ERR_PTR(-EINVAL);
 790                 }
 791         }
 792
 793         return f;
 794
 795 cleanup_all:
 796         fops_put(f->f_op);
 797         if (f->f_mode & FMODE_WRITE)
 798                 put_write_access(inode);
 799         file_kill(f);
 800         f->f_dentry = NULL;
 801         f->f_vfsmnt = NULL;
 802 cleanup_file:
 803         put_filp(f);
 804         dput(dentry);
 805         mntput(mnt);
 806         return ERR_PTR(error);
 807 }
 808
 809 /*
 810  * Note that while the flag value (low two bits) for sys_open means:
 811  *      00 - read-only
 812  *      01 - write-only
 813  *      10 - read-write
 814  *      11 - special
 815  * it is changed into
 816  *      00 - no permissions needed
 817  *      01 - read-permission
 818  *      10 - write-permission
 819  *      11 - read-write
 820  * for the internal routines (ie open_namei()/follow_link() etc). 00 is
 821  * used by symlinks.
 822  */
 823 struct file *filp_open(const char * filename, int flags, int mode)
 824 {
 825         int namei_flags, error;
 826         struct nameidata nd;
 827
 828         namei_flags = flags;
 829         if ((namei_flags+1) & O_ACCMODE)
 830                 namei_flags++;
 831
 832         error = open_namei(filename, namei_flags, mode, &nd);
 833         if (!error)
 834                 return nameidata_to_filp(&nd, flags);
 835
 836         return ERR_PTR(error);
 837 }
 838 EXPORT_SYMBOL(filp_open);
 839
 840 /**
 841  * lookup_instantiate_filp - instantiates the open intent filp
 842  * @nd: pointer to nameidata
 843  * @dentry: pointer to dentry
 844  * @open: open callback
 845  *
 846  * Helper for filesystems that want to use lookup open intents and pass back
 847  * a fully instantiated struct file to the caller.
 848  * This function is meant to be called from within a filesystem's
 849  * lookup method.
 850  * Note that in case of error, nd->intent.open.file is destroyed, but the
 851  * path information remains valid.
 852  * If the open callback is set to NULL, then the standard f_op->open()
 853  * filesystem callback is substituted.
 854  */
 855 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
 856                 int (*open)(struct inode *, struct file *))
 857 {
 858         if (IS_ERR(nd->intent.open.file))
 859                 goto out;
 860         if (IS_ERR(dentry))
 861                 goto out_err;
 862         nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
 863                                              nd->intent.open.flags - 1,
 864                                              nd->intent.open.file,
 865                                              open);
 866 out:
 867         return nd->intent.open.file;
 868 out_err:
 869         release_open_intent(nd);
 870         nd->intent.open.file = (struct file *)dentry;
 871         goto out;
 872 }
 873 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
 874
 875 /**
 876  * nameidata_to_filp - convert a nameidata to an open filp.
 877  * @nd: pointer to nameidata
 878  * @flags: open flags
 879  *
 880  * Note that this function destroys the original nameidata
 881  */
 882 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
 883 {
 884         struct file *filp;
 885
 886         /* Pick up the filp from the open intent */
 887         filp = nd->intent.open.file;
 888         /* Has the filesystem initialised the file for us? */
 889         if (filp->f_dentry == NULL)
 890                 filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
 891         else
 892                 path_release(nd);
 893         return filp;
 894 }
 895
 896 /*
 897  * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
 898  * error.
 899  */
 900 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
 901 {
 902         int error;
 903         struct file *f;
 904
 905         error = -ENFILE;
 906         f = get_empty_filp();
 907         if (f == NULL) {
 908                 dput(dentry);
 909                 mntput(mnt);
 910                 return ERR_PTR(error);
 911         }
 912
 913         return __dentry_open(dentry, mnt, flags, f, NULL);
 914 }
 915 EXPORT_SYMBOL(dentry_open);
 916
 917 /*
 918  * Find an empty file descriptor entry, and mark it busy.
 919  */
 920 int get_unused_fd(void)
 921 {
 922         struct files_struct * files = current->files;
 923         int fd, error;
 924         struct fdtable *fdt;
 925
 926         error = -EMFILE;
 927         spin_lock(&files->file_lock);
 928
 929 repeat:
 930         fdt = files_fdtable(files);
 931         fd = find_next_zero_bit(fdt->open_fds->fds_bits,
 932                                 fdt->max_fdset,
 933                                 fdt->next_fd);
 934
 935         /*
 936          * N.B. For clone tasks sharing a files structure, this test
 937          * will limit the total number of files that can be opened.
 938          */
 939         if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
 940                 goto out;
 941
 942         /* Do we need to expand the fd array or fd set?  */
 943         error = expand_files(files, fd);
 944         if (error < 0)
 945                 goto out;
 946
 947         if (error) {
 948                 /*
 949                  * If we needed to expand the fs array we
 950                  * might have blocked - try again.
 951                  */
 952                 error = -EMFILE;
 953                 goto repeat;
 954         }
 955
 956         FD_SET(fd, fdt->open_fds);
 957         FD_CLR(fd, fdt->close_on_exec);
 958         fdt->next_fd = fd + 1;
 959 #if 1
 960         /* Sanity check */
 961         if (fdt->fd[fd] != NULL) {
 962                 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
 963                 fdt->fd[fd] = NULL;
 964         }
 965 #endif
 966         error = fd;
 967
 968 out:
 969         spin_unlock(&files->file_lock);
 970         return error;
 971 }
 972
 973 EXPORT_SYMBOL(get_unused_fd);
 974
 975 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
 976 {
 977         struct fdtable *fdt = files_fdtable(files);
 978         __FD_CLR(fd, fdt->open_fds);
 979         if (fd < fdt->next_fd)
 980                 fdt->next_fd = fd;
 981 }
 982
 983 void fastcall put_unused_fd(unsigned int fd)
 984 {
 985         struct files_struct *files = current->files;
 986         spin_lock(&files->file_lock);
 987         __put_unused_fd(files, fd);
 988         spin_unlock(&files->file_lock);
 989 }
 990
 991 EXPORT_SYMBOL(put_unused_fd);
 992
 993 /*
 994  * Install a file pointer in the fd array.
 995  *
 996  * The VFS is full of places where we drop the files lock between
 997  * setting the open_fds bitmap and installing the file in the file
 998  * array.  At any such point, we are vulnerable to a dup2() race
 999  * installing a file in the array before us.  We need to detect this and
1000  * fput() the struct file we are about to overwrite in this case.
1001  *
1002  * It should never happen - if we allow dup2() do it, _really_ bad things
1003  * will follow.
1004  */
1005
1006 void fastcall fd_install(unsigned int fd, struct file * file)
1007 {
1008         struct files_struct *files = current->files;
1009         struct fdtable *fdt;
1010         spin_lock(&files->file_lock);
1011         fdt = files_fdtable(files);
1012         BUG_ON(fdt->fd[fd] != NULL);
1013         rcu_assign_pointer(fdt->fd[fd], file);
1014         spin_unlock(&files->file_lock);
1015 }
1016
1017 EXPORT_SYMBOL(fd_install);
1018
1019 long do_sys_open(const char __user *filename, int flags, int mode)
1020 {
1021         char *tmp = getname(filename);
1022         int fd = PTR_ERR(tmp);
1023
1024         if (!IS_ERR(tmp)) {
1025                 fd = get_unused_fd();
1026                 if (fd >= 0) {
1027                         struct file *f = filp_open(tmp, flags, mode);
1028                         if (IS_ERR(f)) {
1029                                 put_unused_fd(fd);
1030                                 fd = PTR_ERR(f);
1031                         } else {
1032                                 fsnotify_open(f->f_dentry);
1033                                 fd_install(fd, f);
1034                         }
1035                 }
1036                 putname(tmp);
1037         }
1038         return fd;
1039 }
1040
1041 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1042 {
1043         if (force_o_largefile())
1044                 flags |= O_LARGEFILE;
1045
1046         return do_sys_open(filename, flags, mode);
1047 }
1048 EXPORT_SYMBOL_GPL(sys_open);
1049
1050 #ifndef __alpha__
1051
1052 /*
1053  * For backward compatibility?  Maybe this should be moved
1054  * into arch/i386 instead?
1055  */
1056 asmlinkage long sys_creat(const char __user * pathname, int mode)
1057 {
1058         return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1059 }
1060
1061 #endif
1062
1063 /*
1064  * "id" is the POSIX thread ID. We use the
1065  * files pointer for this..
1066  */
1067 int filp_close(struct file *filp, fl_owner_t id)
1068 {
1069         int retval = 0;
1070
1071         if (!file_count(filp)) {
1072                 printk(KERN_ERR "VFS: Close: file count is 0\n");
1073                 return 0;
1074         }
1075
1076         if (filp->f_op && filp->f_op->flush)
1077                 retval = filp->f_op->flush(filp);
1078
1079         dnotify_flush(filp, id);
1080         locks_remove_posix(filp, id);
1081         fput(filp);
1082         return retval;
1083 }
1084
1085 EXPORT_SYMBOL(filp_close);
1086
1087 /*
1088  * Careful here! We test whether the file pointer is NULL before
1089  * releasing the fd. This ensures that one clone task can't release
1090  * an fd while another clone is opening it.
1091  */
1092 asmlinkage long sys_close(unsigned int fd)
1093 {
1094         struct file * filp;
1095         struct files_struct *files = current->files;
1096         struct fdtable *fdt;
1097
1098         spin_lock(&files->file_lock);
1099         fdt = files_fdtable(files);
1100         if (fd >= fdt->max_fds)
1101                 goto out_unlock;
1102         filp = fdt->fd[fd];
1103         if (!filp)
1104                 goto out_unlock;
1105         rcu_assign_pointer(fdt->fd[fd], NULL);
1106         FD_CLR(fd, fdt->close_on_exec);
1107         __put_unused_fd(files, fd);
1108         spin_unlock(&files->file_lock);
1109         return filp_close(filp, files);
1110
1111 out_unlock:
1112         spin_unlock(&files->file_lock);
1113         return -EBADF;
1114 }
1115
1116 EXPORT_SYMBOL(sys_close);
1117
1118 /*
1119  * This routine simulates a hangup on the tty, to arrange that users
1120  * are given clean terminals at login time.
1121  */
1122 asmlinkage long sys_vhangup(void)
1123 {
1124         if (capable(CAP_SYS_TTY_CONFIG)) {
1125                 tty_vhangup(current->signal->tty);
1126                 return 0;
1127         }
1128         return -EPERM;
1129 }
1130
1131 /*
1132  * Called when an inode is about to be open.
1133  * We use this to disallow opening large files on 32bit systems if
1134  * the caller didn't specify O_LARGEFILE.  On 64bit systems we force
1135  * on this flag in sys_open.
1136  */
1137 int generic_file_open(struct inode * inode, struct file * filp)
1138 {
1139         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1140                 return -EFBIG;
1141         return 0;
1142 }
1143
1144 EXPORT_SYMBOL(generic_file_open);
1145
1146 /*
1147  * This is used by subsystems that don't want seekable
1148  * file descriptors
1149  */
1150 int nonseekable_open(struct inode *inode, struct file *filp)
1151 {
1152         filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1153         return 0;
1154 }
1155
1156 EXPORT_SYMBOL(nonseekable_open);