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 <linux/fcntl.h>
  24 #include <asm/uaccess.h>
  25 #include <linux/fs.h>
  26 #include <linux/personality.h>
  27 #include <linux/pagemap.h>
  28 #include <linux/syscalls.h>
  29 #include <linux/rcupdate.h>
  30 #include <linux/audit.h>
  31
  32 #include <asm/unistd.h>
  33
  34 int vfs_statfs(struct super_block *sb, struct kstatfs *buf)
  35 {
  36         int retval = -ENODEV;
  37
  38         if (sb) {
  39                 retval = -ENOSYS;
  40                 if (sb->s_op->statfs) {
  41                         memset(buf, 0, sizeof(*buf));
  42                         retval = security_sb_statfs(sb);
  43                         if (retval)
  44                                 return retval;
  45                         retval = sb->s_op->statfs(sb, buf);
  46                         if (retval == 0 && buf->f_frsize == 0)
  47                                 buf->f_frsize = buf->f_bsize;
  48                 }
  49         }
  50         return retval;
  51 }
  52
  53 EXPORT_SYMBOL(vfs_statfs);
  54
  55 static int vfs_statfs_native(struct super_block *sb, struct statfs *buf)
  56 {
  57         struct kstatfs st;
  58         int retval;
  59
  60         retval = vfs_statfs(sb, &st);
  61         if (retval)
  62                 return retval;
  63
  64         if (sizeof(*buf) == sizeof(st))
  65                 memcpy(buf, &st, sizeof(st));
  66         else {
  67                 if (sizeof buf->f_blocks == 4) {
  68                         if ((st.f_blocks | st.f_bfree | st.f_bavail) &
  69                             0xffffffff00000000ULL)
  70                                 return -EOVERFLOW;
  71                         /*
  72                          * f_files and f_ffree may be -1; it's okay to stuff
  73                          * that into 32 bits
  74                          */
  75                         if (st.f_files != -1 &&
  76                             (st.f_files & 0xffffffff00000000ULL))
  77                                 return -EOVERFLOW;
  78                         if (st.f_ffree != -1 &&
  79                             (st.f_ffree & 0xffffffff00000000ULL))
  80                                 return -EOVERFLOW;
  81                 }
  82
  83                 buf->f_type = st.f_type;
  84                 buf->f_bsize = st.f_bsize;
  85                 buf->f_blocks = st.f_blocks;
  86                 buf->f_bfree = st.f_bfree;
  87                 buf->f_bavail = st.f_bavail;
  88                 buf->f_files = st.f_files;
  89                 buf->f_ffree = st.f_ffree;
  90                 buf->f_fsid = st.f_fsid;
  91                 buf->f_namelen = st.f_namelen;
  92                 buf->f_frsize = st.f_frsize;
  93                 memset(buf->f_spare, 0, sizeof(buf->f_spare));
  94         }
  95         return 0;
  96 }
  97
  98 static int vfs_statfs64(struct super_block *sb, struct statfs64 *buf)
  99 {
 100         struct kstatfs st;
 101         int retval;
 102
 103         retval = vfs_statfs(sb, &st);
 104         if (retval)
 105                 return retval;
 106
 107         if (sizeof(*buf) == sizeof(st))
 108                 memcpy(buf, &st, sizeof(st));
 109         else {
 110                 buf->f_type = st.f_type;
 111                 buf->f_bsize = st.f_bsize;
 112                 buf->f_blocks = st.f_blocks;
 113                 buf->f_bfree = st.f_bfree;
 114                 buf->f_bavail = st.f_bavail;
 115                 buf->f_files = st.f_files;
 116                 buf->f_ffree = st.f_ffree;
 117                 buf->f_fsid = st.f_fsid;
 118                 buf->f_namelen = st.f_namelen;
 119                 buf->f_frsize = st.f_frsize;
 120                 memset(buf->f_spare, 0, sizeof(buf->f_spare));
 121         }
 122         return 0;
 123 }
 124
 125 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
 126 {
 127         struct nameidata nd;
 128         int error;
 129
 130         error = user_path_walk(path, &nd);
 131         if (!error) {
 132                 struct statfs tmp;
 133                 error = vfs_statfs_native(nd.dentry->d_inode->i_sb, &tmp);
 134                 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 135                         error = -EFAULT;
 136                 path_release(&nd);
 137         }
 138         return error;
 139 }
 140
 141
 142 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
 143 {
 144         struct nameidata nd;
 145         long error;
 146
 147         if (sz != sizeof(*buf))
 148                 return -EINVAL;
 149         error = user_path_walk(path, &nd);
 150         if (!error) {
 151                 struct statfs64 tmp;
 152                 error = vfs_statfs64(nd.dentry->d_inode->i_sb, &tmp);
 153                 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 154                         error = -EFAULT;
 155                 path_release(&nd);
 156         }
 157         return error;
 158 }
 159
 160
 161 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
 162 {
 163         struct file * file;
 164         struct statfs tmp;
 165         int error;
 166
 167         error = -EBADF;
 168         file = fget(fd);
 169         if (!file)
 170                 goto out;
 171         error = vfs_statfs_native(file->f_dentry->d_inode->i_sb, &tmp);
 172         if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 173                 error = -EFAULT;
 174         fput(file);
 175 out:
 176         return error;
 177 }
 178
 179 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
 180 {
 181         struct file * file;
 182         struct statfs64 tmp;
 183         int error;
 184
 185         if (sz != sizeof(*buf))
 186                 return -EINVAL;
 187
 188         error = -EBADF;
 189         file = fget(fd);
 190         if (!file)
 191                 goto out;
 192         error = vfs_statfs64(file->f_dentry->d_inode->i_sb, &tmp);
 193         if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 194                 error = -EFAULT;
 195         fput(file);
 196 out:
 197         return error;
 198 }
 199
 200 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
 201         struct file *filp)
 202 {
 203         int err;
 204         struct iattr newattrs;
 205
 206         /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
 207         if (length < 0)
 208                 return -EINVAL;
 209
 210         newattrs.ia_size = length;
 211         newattrs.ia_valid = ATTR_SIZE | time_attrs;
 212         if (filp) {
 213                 newattrs.ia_file = filp;
 214                 newattrs.ia_valid |= ATTR_FILE;
 215         }
 216
 217         mutex_lock(&dentry->d_inode->i_mutex);
 218         err = notify_change(dentry, &newattrs);
 219         mutex_unlock(&dentry->d_inode->i_mutex);
 220         return err;
 221 }
 222
 223 static long do_sys_truncate(const char __user * path, loff_t length)
 224 {
 225         struct nameidata nd;
 226         struct inode * inode;
 227         int error;
 228
 229         error = -EINVAL;
 230         if (length < 0) /* sorry, but loff_t says... */
 231                 goto out;
 232
 233         error = user_path_walk(path, &nd);
 234         if (error)
 235                 goto out;
 236         inode = nd.dentry->d_inode;
 237
 238         /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
 239         error = -EISDIR;
 240         if (S_ISDIR(inode->i_mode))
 241                 goto dput_and_out;
 242
 243         error = -EINVAL;
 244         if (!S_ISREG(inode->i_mode))
 245                 goto dput_and_out;
 246
 247         error = vfs_permission(&nd, MAY_WRITE);
 248         if (error)
 249                 goto dput_and_out;
 250
 251         error = -EROFS;
 252         if (IS_RDONLY(inode))
 253                 goto dput_and_out;
 254
 255         error = -EPERM;
 256         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 257                 goto dput_and_out;
 258
 259         /*
 260          * Make sure that there are no leases.
 261          */
 262         error = break_lease(inode, FMODE_WRITE);
 263         if (error)
 264                 goto dput_and_out;
 265
 266         error = get_write_access(inode);
 267         if (error)
 268                 goto dput_and_out;
 269
 270         error = locks_verify_truncate(inode, NULL, length);
 271         if (!error) {
 272                 DQUOT_INIT(inode);
 273                 error = do_truncate(nd.dentry, length, 0, NULL);
 274         }
 275         put_write_access(inode);
 276
 277 dput_and_out:
 278         path_release(&nd);
 279 out:
 280         return error;
 281 }
 282
 283 asmlinkage long sys_truncate(const char __user * path, unsigned long length)
 284 {
 285         /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
 286         return do_sys_truncate(path, (long)length);
 287 }
 288
 289 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
 290 {
 291         struct inode * inode;
 292         struct dentry *dentry;
 293         struct file * file;
 294         int error;
 295
 296         error = -EINVAL;
 297         if (length < 0)
 298                 goto out;
 299         error = -EBADF;
 300         file = fget(fd);
 301         if (!file)
 302                 goto out;
 303
 304         /* explicitly opened as large or we are on 64-bit box */
 305         if (file->f_flags & O_LARGEFILE)
 306                 small = 0;
 307
 308         dentry = file->f_dentry;
 309         inode = dentry->d_inode;
 310         error = -EINVAL;
 311         if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
 312                 goto out_putf;
 313
 314         error = -EINVAL;
 315         /* Cannot ftruncate over 2^31 bytes without large file support */
 316         if (small && length > MAX_NON_LFS)
 317                 goto out_putf;
 318
 319         error = -EPERM;
 320         if (IS_APPEND(inode))
 321                 goto out_putf;
 322
 323         error = locks_verify_truncate(inode, file, length);
 324         if (!error)
 325                 error = do_truncate(dentry, length, 0, file);
 326 out_putf:
 327         fput(file);
 328 out:
 329         return error;
 330 }
 331
 332 asmlinkage long sys_ftruncate(unsigned int fd, unsigned long length)
 333 {
 334         return do_sys_ftruncate(fd, length, 1);
 335 }
 336
 337 /* LFS versions of truncate are only needed on 32 bit machines */
 338 #if BITS_PER_LONG == 32
 339 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
 340 {
 341         return do_sys_truncate(path, length);
 342 }
 343
 344 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
 345 {
 346         return do_sys_ftruncate(fd, length, 0);
 347 }
 348 #endif
 349
 350 #ifdef __ARCH_WANT_SYS_UTIME
 351
 352 /*
 353  * sys_utime() can be implemented in user-level using sys_utimes().
 354  * Is this for backwards compatibility?  If so, why not move it
 355  * into the appropriate arch directory (for those architectures that
 356  * need it).
 357  */
 358
 359 /* If times==NULL, set access and modification to current time,
 360  * must be owner or have write permission.
 361  * Else, update from *times, must be owner or super user.
 362  */
 363 asmlinkage long sys_utime(char __user * filename, struct utimbuf __user * times)
 364 {
 365         int error;
 366         struct nameidata nd;
 367         struct inode * inode;
 368         struct iattr newattrs;
 369
 370         error = user_path_walk(filename, &nd);
 371         if (error)
 372                 goto out;
 373         inode = nd.dentry->d_inode;
 374
 375         error = -EROFS;
 376         if (IS_RDONLY(inode))
 377                 goto dput_and_out;
 378
 379         /* Don't worry, the checks are done in inode_change_ok() */
 380         newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
 381         if (times) {
 382                 error = -EPERM;
 383                 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
 384                         goto dput_and_out;
 385
 386                 error = get_user(newattrs.ia_atime.tv_sec, &times->actime);
 387                 newattrs.ia_atime.tv_nsec = 0;
 388                 if (!error)
 389                         error = get_user(newattrs.ia_mtime.tv_sec, &times->modtime);
 390                 newattrs.ia_mtime.tv_nsec = 0;
 391                 if (error)
 392                         goto dput_and_out;
 393
 394                 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
 395         } else {
 396                 error = -EACCES;
 397                 if (IS_IMMUTABLE(inode))
 398                         goto dput_and_out;
 399
 400                 if (current->fsuid != inode->i_uid &&
 401                     (error = vfs_permission(&nd, MAY_WRITE)) != 0)
 402                         goto dput_and_out;
 403         }
 404         mutex_lock(&inode->i_mutex);
 405         error = notify_change(nd.dentry, &newattrs);
 406         mutex_unlock(&inode->i_mutex);
 407 dput_and_out:
 408         path_release(&nd);
 409 out:
 410         return error;
 411 }
 412
 413 #endif
 414
 415 /* If times==NULL, set access and modification to current time,
 416  * must be owner or have write permission.
 417  * Else, update from *times, must be owner or super user.
 418  */
 419 long do_utimes(int dfd, char __user *filename, struct timeval *times)
 420 {
 421         int error;
 422         struct nameidata nd;
 423         struct inode * inode;
 424         struct iattr newattrs;
 425
 426         error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
 427
 428         if (error)
 429                 goto out;
 430         inode = nd.dentry->d_inode;
 431
 432         error = -EROFS;
 433         if (IS_RDONLY(inode))
 434                 goto dput_and_out;
 435
 436         /* Don't worry, the checks are done in inode_change_ok() */
 437         newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
 438         if (times) {
 439                 error = -EPERM;
 440                 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
 441                         goto dput_and_out;
 442
 443                 newattrs.ia_atime.tv_sec = times[0].tv_sec;
 444                 newattrs.ia_atime.tv_nsec = times[0].tv_usec * 1000;
 445                 newattrs.ia_mtime.tv_sec = times[1].tv_sec;
 446                 newattrs.ia_mtime.tv_nsec = times[1].tv_usec * 1000;
 447                 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
 448         } else {
 449                 error = -EACCES;
 450                 if (IS_IMMUTABLE(inode))
 451                         goto dput_and_out;
 452
 453                 if (current->fsuid != inode->i_uid &&
 454                     (error = vfs_permission(&nd, MAY_WRITE)) != 0)
 455                         goto dput_and_out;
 456         }
 457         mutex_lock(&inode->i_mutex);
 458         error = notify_change(nd.dentry, &newattrs);
 459         mutex_unlock(&inode->i_mutex);
 460 dput_and_out:
 461         path_release(&nd);
 462 out:
 463         return error;
 464 }
 465
 466 asmlinkage long sys_futimesat(int dfd, char __user *filename, struct timeval __user *utimes)
 467 {
 468         struct timeval times[2];
 469
 470         if (utimes && copy_from_user(&times, utimes, sizeof(times)))
 471                 return -EFAULT;
 472         return do_utimes(dfd, filename, utimes ? times : NULL);
 473 }
 474
 475 asmlinkage long sys_utimes(char __user *filename, struct timeval __user *utimes)
 476 {
 477         return sys_futimesat(AT_FDCWD, filename, utimes);
 478 }
 479
 480
 481 /*
 482  * access() needs to use the real uid/gid, not the effective uid/gid.
 483  * We do this by temporarily clearing all FS-related capabilities and
 484  * switching the fsuid/fsgid around to the real ones.
 485  */
 486 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
 487 {
 488         struct nameidata nd;
 489         int old_fsuid, old_fsgid;
 490         kernel_cap_t old_cap;
 491         int res;
 492
 493         if (mode & ~S_IRWXO)    /* where's F_OK, X_OK, W_OK, R_OK? */
 494                 return -EINVAL;
 495
 496         old_fsuid = current->fsuid;
 497         old_fsgid = current->fsgid;
 498         old_cap = current->cap_effective;
 499
 500         current->fsuid = current->uid;
 501         current->fsgid = current->gid;
 502
 503         /*
 504          * Clear the capabilities if we switch to a non-root user
 505          *
 506          * FIXME: There is a race here against sys_capset.  The
 507          * capabilities can change yet we will restore the old
 508          * value below.  We should hold task_capabilities_lock,
 509          * but we cannot because user_path_walk can sleep.
 510          */
 511         if (current->uid)
 512                 cap_clear(current->cap_effective);
 513         else
 514                 current->cap_effective = current->cap_permitted;
 515
 516         res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
 517         if (!res) {
 518                 res = vfs_permission(&nd, mode);
 519                 /* SuS v2 requires we report a read only fs too */
 520                 if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode)
 521                    && !special_file(nd.dentry->d_inode->i_mode))
 522                         res = -EROFS;
 523                 path_release(&nd);
 524         }
 525
 526         current->fsuid = old_fsuid;
 527         current->fsgid = old_fsgid;
 528         current->cap_effective = old_cap;
 529
 530         return res;
 531 }
 532
 533 asmlinkage long sys_access(const char __user *filename, int mode)
 534 {
 535         return sys_faccessat(AT_FDCWD, filename, mode);
 536 }
 537
 538 asmlinkage long sys_chdir(const char __user * filename)
 539 {
 540         struct nameidata nd;
 541         int error;
 542
 543         error = __user_walk(filename, LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &nd);
 544         if (error)
 545                 goto out;
 546
 547         error = vfs_permission(&nd, MAY_EXEC);
 548         if (error)
 549                 goto dput_and_out;
 550
 551         set_fs_pwd(current->fs, nd.mnt, nd.dentry);
 552
 553 dput_and_out:
 554         path_release(&nd);
 555 out:
 556         return error;
 557 }
 558
 559 asmlinkage long sys_fchdir(unsigned int fd)
 560 {
 561         struct file *file;
 562         struct dentry *dentry;
 563         struct inode *inode;
 564         struct vfsmount *mnt;
 565         int error;
 566
 567         error = -EBADF;
 568         file = fget(fd);
 569         if (!file)
 570                 goto out;
 571
 572         dentry = file->f_dentry;
 573         mnt = file->f_vfsmnt;
 574         inode = dentry->d_inode;
 575
 576         error = -ENOTDIR;
 577         if (!S_ISDIR(inode->i_mode))
 578                 goto out_putf;
 579
 580         error = file_permission(file, MAY_EXEC);
 581         if (!error)
 582                 set_fs_pwd(current->fs, mnt, dentry);
 583 out_putf:
 584         fput(file);
 585 out:
 586         return error;
 587 }
 588
 589 asmlinkage long sys_chroot(const char __user * filename)
 590 {
 591         struct nameidata nd;
 592         int error;
 593
 594         error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
 595         if (error)
 596                 goto out;
 597
 598         error = vfs_permission(&nd, MAY_EXEC);
 599         if (error)
 600                 goto dput_and_out;
 601
 602         error = -EPERM;
 603         if (!capable(CAP_SYS_CHROOT))
 604                 goto dput_and_out;
 605
 606         set_fs_root(current->fs, nd.mnt, nd.dentry);
 607         set_fs_altroot();
 608         error = 0;
 609 dput_and_out:
 610         path_release(&nd);
 611 out:
 612         return error;
 613 }
 614
 615 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
 616 {
 617         struct inode * inode;
 618         struct dentry * dentry;
 619         struct file * file;
 620         int err = -EBADF;
 621         struct iattr newattrs;
 622
 623         file = fget(fd);
 624         if (!file)
 625                 goto out;
 626
 627         dentry = file->f_dentry;
 628         inode = dentry->d_inode;
 629
 630         audit_inode(NULL, inode, 0);
 631
 632         err = -EROFS;
 633         if (IS_RDONLY(inode))
 634                 goto out_putf;
 635         err = -EPERM;
 636         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 637                 goto out_putf;
 638         mutex_lock(&inode->i_mutex);
 639         if (mode == (mode_t) -1)
 640                 mode = inode->i_mode;
 641         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 642         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 643         err = notify_change(dentry, &newattrs);
 644         mutex_unlock(&inode->i_mutex);
 645
 646 out_putf:
 647         fput(file);
 648 out:
 649         return err;
 650 }
 651
 652 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
 653                              mode_t mode)
 654 {
 655         struct nameidata nd;
 656         struct inode * inode;
 657         int error;
 658         struct iattr newattrs;
 659
 660         error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
 661         if (error)
 662                 goto out;
 663         inode = nd.dentry->d_inode;
 664
 665         error = -EROFS;
 666         if (IS_RDONLY(inode))
 667                 goto dput_and_out;
 668
 669         error = -EPERM;
 670         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 671                 goto dput_and_out;
 672
 673         mutex_lock(&inode->i_mutex);
 674         if (mode == (mode_t) -1)
 675                 mode = inode->i_mode;
 676         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 677         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 678         error = notify_change(nd.dentry, &newattrs);
 679         mutex_unlock(&inode->i_mutex);
 680
 681 dput_and_out:
 682         path_release(&nd);
 683 out:
 684         return error;
 685 }
 686
 687 asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
 688 {
 689         return sys_fchmodat(AT_FDCWD, filename, mode);
 690 }
 691
 692 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
 693 {
 694         struct inode * inode;
 695         int error;
 696         struct iattr newattrs;
 697
 698         error = -ENOENT;
 699         if (!(inode = dentry->d_inode)) {
 700                 printk(KERN_ERR "chown_common: NULL inode\n");
 701                 goto out;
 702         }
 703         error = -EROFS;
 704         if (IS_RDONLY(inode))
 705                 goto out;
 706         error = -EPERM;
 707         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 708                 goto out;
 709         newattrs.ia_valid =  ATTR_CTIME;
 710         if (user != (uid_t) -1) {
 711                 newattrs.ia_valid |= ATTR_UID;
 712                 newattrs.ia_uid = user;
 713         }
 714         if (group != (gid_t) -1) {
 715                 newattrs.ia_valid |= ATTR_GID;
 716                 newattrs.ia_gid = group;
 717         }
 718         if (!S_ISDIR(inode->i_mode))
 719                 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
 720         mutex_lock(&inode->i_mutex);
 721         error = notify_change(dentry, &newattrs);
 722         mutex_unlock(&inode->i_mutex);
 723 out:
 724         return error;
 725 }
 726
 727 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
 728 {
 729         struct nameidata nd;
 730         int error;
 731
 732         error = user_path_walk(filename, &nd);
 733         if (!error) {
 734                 error = chown_common(nd.dentry, user, group);
 735                 path_release(&nd);
 736         }
 737         return error;
 738 }
 739
 740 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
 741                              gid_t group, int flag)
 742 {
 743         struct nameidata nd;
 744         int error = -EINVAL;
 745         int follow;
 746
 747         if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
 748                 goto out;
 749
 750         follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
 751         error = __user_walk_fd(dfd, filename, follow, &nd);
 752         if (!error) {
 753                 error = chown_common(nd.dentry, user, group);
 754                 path_release(&nd);
 755         }
 756 out:
 757         return error;
 758 }
 759
 760 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
 761 {
 762         struct nameidata nd;
 763         int error;
 764
 765         error = user_path_walk_link(filename, &nd);
 766         if (!error) {
 767                 error = chown_common(nd.dentry, user, group);
 768                 path_release(&nd);
 769         }
 770         return error;
 771 }
 772
 773
 774 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
 775 {
 776         struct file * file;
 777         int error = -EBADF;
 778
 779         file = fget(fd);
 780         if (file) {
 781                 struct dentry * dentry;
 782                 dentry = file->f_dentry;
 783                 audit_inode(NULL, dentry->d_inode, 0);
 784                 error = chown_common(dentry, user, group);
 785                 fput(file);
 786         }
 787         return error;
 788 }
 789
 790 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
 791                                         int flags, struct file *f,
 792                                         int (*open)(struct inode *, struct file *))
 793 {
 794         struct inode *inode;
 795         int error;
 796
 797         f->f_flags = flags;
 798         f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
 799                                 FMODE_PREAD | FMODE_PWRITE;
 800         inode = dentry->d_inode;
 801         if (f->f_mode & FMODE_WRITE) {
 802                 error = get_write_access(inode);
 803                 if (error)
 804                         goto cleanup_file;
 805         }
 806
 807         f->f_mapping = inode->i_mapping;
 808         f->f_dentry = dentry;
 809         f->f_vfsmnt = mnt;
 810         f->f_pos = 0;
 811         f->f_op = fops_get(inode->i_fop);
 812         file_move(f, &inode->i_sb->s_files);
 813
 814         if (!open && f->f_op)
 815                 open = f->f_op->open;
 816         if (open) {
 817                 error = open(inode, f);
 818                 if (error)
 819                         goto cleanup_all;
 820         }
 821
 822         f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
 823
 824         file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
 825
 826         /* NB: we're sure to have correct a_ops only after f_op->open */
 827         if (f->f_flags & O_DIRECT) {
 828                 if (!f->f_mapping->a_ops ||
 829                     ((!f->f_mapping->a_ops->direct_IO) &&
 830                     (!f->f_mapping->a_ops->get_xip_page))) {
 831                         fput(f);
 832                         f = ERR_PTR(-EINVAL);
 833                 }
 834         }
 835
 836         return f;
 837
 838 cleanup_all:
 839         fops_put(f->f_op);
 840         if (f->f_mode & FMODE_WRITE)
 841                 put_write_access(inode);
 842         file_kill(f);
 843         f->f_dentry = NULL;
 844         f->f_vfsmnt = NULL;
 845 cleanup_file:
 846         put_filp(f);
 847         dput(dentry);
 848         mntput(mnt);
 849         return ERR_PTR(error);
 850 }
 851
 852 /*
 853  * Note that while the flag value (low two bits) for sys_open means:
 854  *      00 - read-only
 855  *      01 - write-only
 856  *      10 - read-write
 857  *      11 - special
 858  * it is changed into
 859  *      00 - no permissions needed
 860  *      01 - read-permission
 861  *      10 - write-permission
 862  *      11 - read-write
 863  * for the internal routines (ie open_namei()/follow_link() etc). 00 is
 864  * used by symlinks.
 865  */
 866 static struct file *do_filp_open(int dfd, const char *filename, int flags,
 867                                  int mode)
 868 {
 869         int namei_flags, error;
 870         struct nameidata nd;
 871
 872         namei_flags = flags;
 873         if ((namei_flags+1) & O_ACCMODE)
 874                 namei_flags++;
 875
 876         error = open_namei(dfd, filename, namei_flags, mode, &nd);
 877         if (!error)
 878                 return nameidata_to_filp(&nd, flags);
 879
 880         return ERR_PTR(error);
 881 }
 882
 883 struct file *filp_open(const char *filename, int flags, int mode)
 884 {
 885         return do_filp_open(AT_FDCWD, filename, flags, mode);
 886 }
 887 EXPORT_SYMBOL(filp_open);
 888
 889 /**
 890  * lookup_instantiate_filp - instantiates the open intent filp
 891  * @nd: pointer to nameidata
 892  * @dentry: pointer to dentry
 893  * @open: open callback
 894  *
 895  * Helper for filesystems that want to use lookup open intents and pass back
 896  * a fully instantiated struct file to the caller.
 897  * This function is meant to be called from within a filesystem's
 898  * lookup method.
 899  * Beware of calling it for non-regular files! Those ->open methods might block
 900  * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
 901  * leading to a deadlock, as nobody can open that fifo anymore, because
 902  * another process to open fifo will block on locked parent when doing lookup).
 903  * Note that in case of error, nd->intent.open.file is destroyed, but the
 904  * path information remains valid.
 905  * If the open callback is set to NULL, then the standard f_op->open()
 906  * filesystem callback is substituted.
 907  */
 908 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
 909                 int (*open)(struct inode *, struct file *))
 910 {
 911         if (IS_ERR(nd->intent.open.file))
 912                 goto out;
 913         if (IS_ERR(dentry))
 914                 goto out_err;
 915         nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
 916                                              nd->intent.open.flags - 1,
 917                                              nd->intent.open.file,
 918                                              open);
 919 out:
 920         return nd->intent.open.file;
 921 out_err:
 922         release_open_intent(nd);
 923         nd->intent.open.file = (struct file *)dentry;
 924         goto out;
 925 }
 926 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
 927
 928 /**
 929  * nameidata_to_filp - convert a nameidata to an open filp.
 930  * @nd: pointer to nameidata
 931  * @flags: open flags
 932  *
 933  * Note that this function destroys the original nameidata
 934  */
 935 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
 936 {
 937         struct file *filp;
 938
 939         /* Pick up the filp from the open intent */
 940         filp = nd->intent.open.file;
 941         /* Has the filesystem initialised the file for us? */
 942         if (filp->f_dentry == NULL)
 943                 filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
 944         else
 945                 path_release(nd);
 946         return filp;
 947 }
 948
 949 /*
 950  * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
 951  * error.
 952  */
 953 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
 954 {
 955         int error;
 956         struct file *f;
 957
 958         error = -ENFILE;
 959         f = get_empty_filp();
 960         if (f == NULL) {
 961                 dput(dentry);
 962                 mntput(mnt);
 963                 return ERR_PTR(error);
 964         }
 965
 966         return __dentry_open(dentry, mnt, flags, f, NULL);
 967 }
 968 EXPORT_SYMBOL(dentry_open);
 969
 970 /*
 971  * Find an empty file descriptor entry, and mark it busy.
 972  */
 973 int get_unused_fd(void)
 974 {
 975         struct files_struct * files = current->files;
 976         int fd, error;
 977         struct fdtable *fdt;
 978
 979         error = -EMFILE;
 980         spin_lock(&files->file_lock);
 981
 982 repeat:
 983         fdt = files_fdtable(files);
 984         fd = find_next_zero_bit(fdt->open_fds->fds_bits,
 985                                 fdt->max_fdset,
 986                                 files->next_fd);
 987
 988         /*
 989          * N.B. For clone tasks sharing a files structure, this test
 990          * will limit the total number of files that can be opened.
 991          */
 992         if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
 993                 goto out;
 994
 995         /* Do we need to expand the fd array or fd set?  */
 996         error = expand_files(files, fd);
 997         if (error < 0)
 998                 goto out;
 999
1000         if (error) {
1001                 /*
1002                  * If we needed to expand the fs array we
1003                  * might have blocked - try again.
1004                  */
1005                 error = -EMFILE;
1006                 goto repeat;
1007         }
1008
1009         FD_SET(fd, fdt->open_fds);
1010         FD_CLR(fd, fdt->close_on_exec);
1011         files->next_fd = fd + 1;
1012 #if 1
1013         /* Sanity check */
1014         if (fdt->fd[fd] != NULL) {
1015                 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
1016                 fdt->fd[fd] = NULL;
1017         }
1018 #endif
1019         error = fd;
1020
1021 out:
1022         spin_unlock(&files->file_lock);
1023         return error;
1024 }
1025
1026 EXPORT_SYMBOL(get_unused_fd);
1027
1028 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
1029 {
1030         struct fdtable *fdt = files_fdtable(files);
1031         __FD_CLR(fd, fdt->open_fds);
1032         if (fd < files->next_fd)
1033                 files->next_fd = fd;
1034 }
1035
1036 void fastcall put_unused_fd(unsigned int fd)
1037 {
1038         struct files_struct *files = current->files;
1039         spin_lock(&files->file_lock);
1040         __put_unused_fd(files, fd);
1041         spin_unlock(&files->file_lock);
1042 }
1043
1044 EXPORT_SYMBOL(put_unused_fd);
1045
1046 /*
1047  * Install a file pointer in the fd array.
1048  *
1049  * The VFS is full of places where we drop the files lock between
1050  * setting the open_fds bitmap and installing the file in the file
1051  * array.  At any such point, we are vulnerable to a dup2() race
1052  * installing a file in the array before us.  We need to detect this and
1053  * fput() the struct file we are about to overwrite in this case.
1054  *
1055  * It should never happen - if we allow dup2() do it, _really_ bad things
1056  * will follow.
1057  */
1058
1059 void fastcall fd_install(unsigned int fd, struct file * file)
1060 {
1061         struct files_struct *files = current->files;
1062         struct fdtable *fdt;
1063         spin_lock(&files->file_lock);
1064         fdt = files_fdtable(files);
1065         BUG_ON(fdt->fd[fd] != NULL);
1066         rcu_assign_pointer(fdt->fd[fd], file);
1067         spin_unlock(&files->file_lock);
1068 }
1069
1070 EXPORT_SYMBOL(fd_install);
1071
1072 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
1073 {
1074         char *tmp = getname(filename);
1075         int fd = PTR_ERR(tmp);
1076
1077         if (!IS_ERR(tmp)) {
1078                 fd = get_unused_fd();
1079                 if (fd >= 0) {
1080                         struct file *f = do_filp_open(dfd, tmp, flags, mode);
1081                         if (IS_ERR(f)) {
1082                                 put_unused_fd(fd);
1083                                 fd = PTR_ERR(f);
1084                         } else {
1085                                 fsnotify_open(f->f_dentry);
1086                                 fd_install(fd, f);
1087                         }
1088                 }
1089                 putname(tmp);
1090         }
1091         return fd;
1092 }
1093
1094 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1095 {
1096         if (force_o_largefile())
1097                 flags |= O_LARGEFILE;
1098
1099         return do_sys_open(AT_FDCWD, filename, flags, mode);
1100 }
1101 EXPORT_SYMBOL_GPL(sys_open);
1102
1103 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1104                            int mode)
1105 {
1106         if (force_o_largefile())
1107                 flags |= O_LARGEFILE;
1108
1109         return do_sys_open(dfd, filename, flags, mode);
1110 }
1111 EXPORT_SYMBOL_GPL(sys_openat);
1112
1113 #ifndef __alpha__
1114
1115 /*
1116  * For backward compatibility?  Maybe this should be moved
1117  * into arch/i386 instead?
1118  */
1119 asmlinkage long sys_creat(const char __user * pathname, int mode)
1120 {
1121         return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1122 }
1123
1124 #endif
1125
1126 /*
1127  * "id" is the POSIX thread ID. We use the
1128  * files pointer for this..
1129  */
1130 int filp_close(struct file *filp, fl_owner_t id)
1131 {
1132         int retval = 0;
1133
1134         if (!file_count(filp)) {
1135                 printk(KERN_ERR "VFS: Close: file count is 0\n");
1136                 return 0;
1137         }
1138
1139         if (filp->f_op && filp->f_op->flush)
1140                 retval = filp->f_op->flush(filp);
1141
1142         dnotify_flush(filp, id);
1143         locks_remove_posix(filp, id);
1144         fput(filp);
1145         return retval;
1146 }
1147
1148 EXPORT_SYMBOL(filp_close);
1149
1150 /*
1151  * Careful here! We test whether the file pointer is NULL before
1152  * releasing the fd. This ensures that one clone task can't release
1153  * an fd while another clone is opening it.
1154  */
1155 asmlinkage long sys_close(unsigned int fd)
1156 {
1157         struct file * filp;
1158         struct files_struct *files = current->files;
1159         struct fdtable *fdt;
1160
1161         spin_lock(&files->file_lock);
1162         fdt = files_fdtable(files);
1163         if (fd >= fdt->max_fds)
1164                 goto out_unlock;
1165         filp = fdt->fd[fd];
1166         if (!filp)
1167                 goto out_unlock;
1168         rcu_assign_pointer(fdt->fd[fd], NULL);
1169         FD_CLR(fd, fdt->close_on_exec);
1170         __put_unused_fd(files, fd);
1171         spin_unlock(&files->file_lock);
1172         return filp_close(filp, files);
1173
1174 out_unlock:
1175         spin_unlock(&files->file_lock);
1176         return -EBADF;
1177 }
1178
1179 EXPORT_SYMBOL(sys_close);
1180
1181 /*
1182  * This routine simulates a hangup on the tty, to arrange that users
1183  * are given clean terminals at login time.
1184  */
1185 asmlinkage long sys_vhangup(void)
1186 {
1187         if (capable(CAP_SYS_TTY_CONFIG)) {
1188                 tty_vhangup(current->signal->tty);
1189                 return 0;
1190         }
1191         return -EPERM;
1192 }
1193
1194 /*
1195  * Called when an inode is about to be open.
1196  * We use this to disallow opening large files on 32bit systems if
1197  * the caller didn't specify O_LARGEFILE.  On 64bit systems we force
1198  * on this flag in sys_open.
1199  */
1200 int generic_file_open(struct inode * inode, struct file * filp)
1201 {
1202         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1203                 return -EFBIG;
1204         return 0;
1205 }
1206
1207 EXPORT_SYMBOL(generic_file_open);
1208
1209 /*
1210  * This is used by subsystems that don't want seekable
1211  * file descriptors
1212  */
1213 int nonseekable_open(struct inode *inode, struct file *filp)
1214 {
1215         filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1216         return 0;
1217 }
1218
1219 EXPORT_SYMBOL(nonseekable_open);