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 dentry *dentry, struct kstatfs *buf)
  35 {
  36         int retval = -ENODEV;
  37
  38         if (dentry) {
  39                 retval = -ENOSYS;
  40                 if (dentry->d_sb->s_op->statfs) {
  41                         memset(buf, 0, sizeof(*buf));
  42                         retval = security_sb_statfs(dentry);
  43                         if (retval)
  44                                 return retval;
  45                         retval = dentry->d_sb->s_op->statfs(dentry, 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 dentry *dentry, struct statfs *buf)
  56 {
  57         struct kstatfs st;
  58         int retval;
  59
  60         retval = vfs_statfs(dentry, &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 dentry *dentry, struct statfs64 *buf)
  99 {
 100         struct kstatfs st;
 101         int retval;
 102
 103         retval = vfs_statfs(dentry, &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, &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, &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, &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, &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, ATTR_MTIME|ATTR_CTIME, 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         long ret = do_sys_ftruncate(fd, length, 1);
 335         /* avoid REGPARM breakage on x86: */
 336         prevent_tail_call(ret);
 337         return ret;
 338 }
 339
 340 /* LFS versions of truncate are only needed on 32 bit machines */
 341 #if BITS_PER_LONG == 32
 342 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
 343 {
 344         return do_sys_truncate(path, length);
 345 }
 346
 347 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
 348 {
 349         long ret = do_sys_ftruncate(fd, length, 0);
 350         /* avoid REGPARM breakage on x86: */
 351         prevent_tail_call(ret);
 352         return ret;
 353 }
 354 #endif
 355
 356 #ifdef __ARCH_WANT_SYS_UTIME
 357
 358 /*
 359  * sys_utime() can be implemented in user-level using sys_utimes().
 360  * Is this for backwards compatibility?  If so, why not move it
 361  * into the appropriate arch directory (for those architectures that
 362  * need it).
 363  */
 364
 365 /* If times==NULL, set access and modification to current time,
 366  * must be owner or have write permission.
 367  * Else, update from *times, must be owner or super user.
 368  */
 369 asmlinkage long sys_utime(char __user * filename, struct utimbuf __user * times)
 370 {
 371         int error;
 372         struct nameidata nd;
 373         struct inode * inode;
 374         struct iattr newattrs;
 375
 376         error = user_path_walk(filename, &nd);
 377         if (error)
 378                 goto out;
 379         inode = nd.dentry->d_inode;
 380
 381         error = -EROFS;
 382         if (IS_RDONLY(inode))
 383                 goto dput_and_out;
 384
 385         /* Don't worry, the checks are done in inode_change_ok() */
 386         newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
 387         if (times) {
 388                 error = -EPERM;
 389                 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
 390                         goto dput_and_out;
 391
 392                 error = get_user(newattrs.ia_atime.tv_sec, &times->actime);
 393                 newattrs.ia_atime.tv_nsec = 0;
 394                 if (!error)
 395                         error = get_user(newattrs.ia_mtime.tv_sec, &times->modtime);
 396                 newattrs.ia_mtime.tv_nsec = 0;
 397                 if (error)
 398                         goto dput_and_out;
 399
 400                 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
 401         } else {
 402                 error = -EACCES;
 403                 if (IS_IMMUTABLE(inode))
 404                         goto dput_and_out;
 405
 406                 if (current->fsuid != inode->i_uid &&
 407                     (error = vfs_permission(&nd, MAY_WRITE)) != 0)
 408                         goto dput_and_out;
 409         }
 410         mutex_lock(&inode->i_mutex);
 411         error = notify_change(nd.dentry, &newattrs);
 412         mutex_unlock(&inode->i_mutex);
 413 dput_and_out:
 414         path_release(&nd);
 415 out:
 416         return error;
 417 }
 418
 419 #endif
 420
 421 /* If times==NULL, set access and modification to current time,
 422  * must be owner or have write permission.
 423  * Else, update from *times, must be owner or super user.
 424  */
 425 long do_utimes(int dfd, char __user *filename, struct timeval *times)
 426 {
 427         int error;
 428         struct nameidata nd;
 429         struct inode * inode;
 430         struct iattr newattrs;
 431
 432         error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
 433
 434         if (error)
 435                 goto out;
 436         inode = nd.dentry->d_inode;
 437
 438         error = -EROFS;
 439         if (IS_RDONLY(inode))
 440                 goto dput_and_out;
 441
 442         /* Don't worry, the checks are done in inode_change_ok() */
 443         newattrs.ia_valid = ATTR_CTIME | ATTR_MTIME | ATTR_ATIME;
 444         if (times) {
 445                 error = -EPERM;
 446                 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
 447                         goto dput_and_out;
 448
 449                 newattrs.ia_atime.tv_sec = times[0].tv_sec;
 450                 newattrs.ia_atime.tv_nsec = times[0].tv_usec * 1000;
 451                 newattrs.ia_mtime.tv_sec = times[1].tv_sec;
 452                 newattrs.ia_mtime.tv_nsec = times[1].tv_usec * 1000;
 453                 newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET;
 454         } else {
 455                 error = -EACCES;
 456                 if (IS_IMMUTABLE(inode))
 457                         goto dput_and_out;
 458
 459                 if (current->fsuid != inode->i_uid &&
 460                     (error = vfs_permission(&nd, MAY_WRITE)) != 0)
 461                         goto dput_and_out;
 462         }
 463         mutex_lock(&inode->i_mutex);
 464         error = notify_change(nd.dentry, &newattrs);
 465         mutex_unlock(&inode->i_mutex);
 466 dput_and_out:
 467         path_release(&nd);
 468 out:
 469         return error;
 470 }
 471
 472 asmlinkage long sys_futimesat(int dfd, char __user *filename, struct timeval __user *utimes)
 473 {
 474         struct timeval times[2];
 475
 476         if (utimes && copy_from_user(&times, utimes, sizeof(times)))
 477                 return -EFAULT;
 478         return do_utimes(dfd, filename, utimes ? times : NULL);
 479 }
 480
 481 asmlinkage long sys_utimes(char __user *filename, struct timeval __user *utimes)
 482 {
 483         return sys_futimesat(AT_FDCWD, filename, utimes);
 484 }
 485
 486
 487 /*
 488  * access() needs to use the real uid/gid, not the effective uid/gid.
 489  * We do this by temporarily clearing all FS-related capabilities and
 490  * switching the fsuid/fsgid around to the real ones.
 491  */
 492 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
 493 {
 494         struct nameidata nd;
 495         int old_fsuid, old_fsgid;
 496         kernel_cap_t old_cap;
 497         int res;
 498
 499         if (mode & ~S_IRWXO)    /* where's F_OK, X_OK, W_OK, R_OK? */
 500                 return -EINVAL;
 501
 502         old_fsuid = current->fsuid;
 503         old_fsgid = current->fsgid;
 504         old_cap = current->cap_effective;
 505
 506         current->fsuid = current->uid;
 507         current->fsgid = current->gid;
 508
 509         /*
 510          * Clear the capabilities if we switch to a non-root user
 511          *
 512          * FIXME: There is a race here against sys_capset.  The
 513          * capabilities can change yet we will restore the old
 514          * value below.  We should hold task_capabilities_lock,
 515          * but we cannot because user_path_walk can sleep.
 516          */
 517         if (current->uid)
 518                 cap_clear(current->cap_effective);
 519         else
 520                 current->cap_effective = current->cap_permitted;
 521
 522         res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
 523         if (!res) {
 524                 res = vfs_permission(&nd, mode);
 525                 /* SuS v2 requires we report a read only fs too */
 526                 if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode)
 527                    && !special_file(nd.dentry->d_inode->i_mode))
 528                         res = -EROFS;
 529                 path_release(&nd);
 530         }
 531
 532         current->fsuid = old_fsuid;
 533         current->fsgid = old_fsgid;
 534         current->cap_effective = old_cap;
 535
 536         return res;
 537 }
 538
 539 asmlinkage long sys_access(const char __user *filename, int mode)
 540 {
 541         return sys_faccessat(AT_FDCWD, filename, mode);
 542 }
 543
 544 asmlinkage long sys_chdir(const char __user * filename)
 545 {
 546         struct nameidata nd;
 547         int error;
 548
 549         error = __user_walk(filename,
 550                             LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_CHDIR, &nd);
 551         if (error)
 552                 goto out;
 553
 554         error = vfs_permission(&nd, MAY_EXEC);
 555         if (error)
 556                 goto dput_and_out;
 557
 558         set_fs_pwd(current->fs, nd.mnt, nd.dentry);
 559
 560 dput_and_out:
 561         path_release(&nd);
 562 out:
 563         return error;
 564 }
 565
 566 asmlinkage long sys_fchdir(unsigned int fd)
 567 {
 568         struct file *file;
 569         struct dentry *dentry;
 570         struct inode *inode;
 571         struct vfsmount *mnt;
 572         int error;
 573
 574         error = -EBADF;
 575         file = fget(fd);
 576         if (!file)
 577                 goto out;
 578
 579         dentry = file->f_dentry;
 580         mnt = file->f_vfsmnt;
 581         inode = dentry->d_inode;
 582
 583         error = -ENOTDIR;
 584         if (!S_ISDIR(inode->i_mode))
 585                 goto out_putf;
 586
 587         error = file_permission(file, MAY_EXEC);
 588         if (!error)
 589                 set_fs_pwd(current->fs, mnt, dentry);
 590 out_putf:
 591         fput(file);
 592 out:
 593         return error;
 594 }
 595
 596 asmlinkage long sys_chroot(const char __user * filename)
 597 {
 598         struct nameidata nd;
 599         int error;
 600
 601         error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
 602         if (error)
 603                 goto out;
 604
 605         error = vfs_permission(&nd, MAY_EXEC);
 606         if (error)
 607                 goto dput_and_out;
 608
 609         error = -EPERM;
 610         if (!capable(CAP_SYS_CHROOT))
 611                 goto dput_and_out;
 612
 613         set_fs_root(current->fs, nd.mnt, nd.dentry);
 614         set_fs_altroot();
 615         error = 0;
 616 dput_and_out:
 617         path_release(&nd);
 618 out:
 619         return error;
 620 }
 621
 622 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
 623 {
 624         struct inode * inode;
 625         struct dentry * dentry;
 626         struct file * file;
 627         int err = -EBADF;
 628         struct iattr newattrs;
 629
 630         file = fget(fd);
 631         if (!file)
 632                 goto out;
 633
 634         dentry = file->f_dentry;
 635         inode = dentry->d_inode;
 636
 637         audit_inode(NULL, inode);
 638
 639         err = -EROFS;
 640         if (IS_RDONLY(inode))
 641                 goto out_putf;
 642         err = -EPERM;
 643         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 644                 goto out_putf;
 645         mutex_lock(&inode->i_mutex);
 646         if (mode == (mode_t) -1)
 647                 mode = inode->i_mode;
 648         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 649         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 650         err = notify_change(dentry, &newattrs);
 651         mutex_unlock(&inode->i_mutex);
 652
 653 out_putf:
 654         fput(file);
 655 out:
 656         return err;
 657 }
 658
 659 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
 660                              mode_t mode)
 661 {
 662         struct nameidata nd;
 663         struct inode * inode;
 664         int error;
 665         struct iattr newattrs;
 666
 667         error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
 668         if (error)
 669                 goto out;
 670         inode = nd.dentry->d_inode;
 671
 672         error = -EROFS;
 673         if (IS_RDONLY(inode))
 674                 goto dput_and_out;
 675
 676         error = -EPERM;
 677         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 678                 goto dput_and_out;
 679
 680         mutex_lock(&inode->i_mutex);
 681         if (mode == (mode_t) -1)
 682                 mode = inode->i_mode;
 683         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 684         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 685         error = notify_change(nd.dentry, &newattrs);
 686         mutex_unlock(&inode->i_mutex);
 687
 688 dput_and_out:
 689         path_release(&nd);
 690 out:
 691         return error;
 692 }
 693
 694 asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
 695 {
 696         return sys_fchmodat(AT_FDCWD, filename, mode);
 697 }
 698
 699 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
 700 {
 701         struct inode * inode;
 702         int error;
 703         struct iattr newattrs;
 704
 705         error = -ENOENT;
 706         if (!(inode = dentry->d_inode)) {
 707                 printk(KERN_ERR "chown_common: NULL inode\n");
 708                 goto out;
 709         }
 710         error = -EROFS;
 711         if (IS_RDONLY(inode))
 712                 goto out;
 713         error = -EPERM;
 714         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 715                 goto out;
 716         newattrs.ia_valid =  ATTR_CTIME;
 717         if (user != (uid_t) -1) {
 718                 newattrs.ia_valid |= ATTR_UID;
 719                 newattrs.ia_uid = user;
 720         }
 721         if (group != (gid_t) -1) {
 722                 newattrs.ia_valid |= ATTR_GID;
 723                 newattrs.ia_gid = group;
 724         }
 725         if (!S_ISDIR(inode->i_mode))
 726                 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
 727         mutex_lock(&inode->i_mutex);
 728         error = notify_change(dentry, &newattrs);
 729         mutex_unlock(&inode->i_mutex);
 730 out:
 731         return error;
 732 }
 733
 734 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
 735 {
 736         struct nameidata nd;
 737         int error;
 738
 739         error = user_path_walk(filename, &nd);
 740         if (!error) {
 741                 error = chown_common(nd.dentry, user, group);
 742                 path_release(&nd);
 743         }
 744         return error;
 745 }
 746
 747 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
 748                              gid_t group, int flag)
 749 {
 750         struct nameidata nd;
 751         int error = -EINVAL;
 752         int follow;
 753
 754         if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
 755                 goto out;
 756
 757         follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
 758         error = __user_walk_fd(dfd, filename, follow, &nd);
 759         if (!error) {
 760                 error = chown_common(nd.dentry, user, group);
 761                 path_release(&nd);
 762         }
 763 out:
 764         return error;
 765 }
 766
 767 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
 768 {
 769         struct nameidata nd;
 770         int error;
 771
 772         error = user_path_walk_link(filename, &nd);
 773         if (!error) {
 774                 error = chown_common(nd.dentry, user, group);
 775                 path_release(&nd);
 776         }
 777         return error;
 778 }
 779
 780
 781 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
 782 {
 783         struct file * file;
 784         int error = -EBADF;
 785
 786         file = fget(fd);
 787         if (file) {
 788                 struct dentry * dentry;
 789                 dentry = file->f_dentry;
 790                 audit_inode(NULL, dentry->d_inode);
 791                 error = chown_common(dentry, user, group);
 792                 fput(file);
 793         }
 794         return error;
 795 }
 796
 797 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
 798                                         int flags, struct file *f,
 799                                         int (*open)(struct inode *, struct file *))
 800 {
 801         struct inode *inode;
 802         int error;
 803
 804         f->f_flags = flags;
 805         f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
 806                                 FMODE_PREAD | FMODE_PWRITE;
 807         inode = dentry->d_inode;
 808         if (f->f_mode & FMODE_WRITE) {
 809                 error = get_write_access(inode);
 810                 if (error)
 811                         goto cleanup_file;
 812         }
 813
 814         f->f_mapping = inode->i_mapping;
 815         f->f_dentry = dentry;
 816         f->f_vfsmnt = mnt;
 817         f->f_pos = 0;
 818         f->f_op = fops_get(inode->i_fop);
 819         file_move(f, &inode->i_sb->s_files);
 820
 821         if (!open && f->f_op)
 822                 open = f->f_op->open;
 823         if (open) {
 824                 error = open(inode, f);
 825                 if (error)
 826                         goto cleanup_all;
 827         }
 828
 829         f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
 830
 831         file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
 832
 833         /* NB: we're sure to have correct a_ops only after f_op->open */
 834         if (f->f_flags & O_DIRECT) {
 835                 if (!f->f_mapping->a_ops ||
 836                     ((!f->f_mapping->a_ops->direct_IO) &&
 837                     (!f->f_mapping->a_ops->get_xip_page))) {
 838                         fput(f);
 839                         f = ERR_PTR(-EINVAL);
 840                 }
 841         }
 842
 843         return f;
 844
 845 cleanup_all:
 846         fops_put(f->f_op);
 847         if (f->f_mode & FMODE_WRITE)
 848                 put_write_access(inode);
 849         file_kill(f);
 850         f->f_dentry = NULL;
 851         f->f_vfsmnt = NULL;
 852 cleanup_file:
 853         put_filp(f);
 854         dput(dentry);
 855         mntput(mnt);
 856         return ERR_PTR(error);
 857 }
 858
 859 /*
 860  * Note that while the flag value (low two bits) for sys_open means:
 861  *      00 - read-only
 862  *      01 - write-only
 863  *      10 - read-write
 864  *      11 - special
 865  * it is changed into
 866  *      00 - no permissions needed
 867  *      01 - read-permission
 868  *      10 - write-permission
 869  *      11 - read-write
 870  * for the internal routines (ie open_namei()/follow_link() etc). 00 is
 871  * used by symlinks.
 872  */
 873 static struct file *do_filp_open(int dfd, const char *filename, int flags,
 874                                  int mode)
 875 {
 876         int namei_flags, error;
 877         struct nameidata nd;
 878
 879         namei_flags = flags;
 880         if ((namei_flags+1) & O_ACCMODE)
 881                 namei_flags++;
 882
 883         error = open_namei(dfd, filename, namei_flags, mode, &nd);
 884         if (!error)
 885                 return nameidata_to_filp(&nd, flags);
 886
 887         return ERR_PTR(error);
 888 }
 889
 890 struct file *filp_open(const char *filename, int flags, int mode)
 891 {
 892         return do_filp_open(AT_FDCWD, filename, flags, mode);
 893 }
 894 EXPORT_SYMBOL(filp_open);
 895
 896 /**
 897  * lookup_instantiate_filp - instantiates the open intent filp
 898  * @nd: pointer to nameidata
 899  * @dentry: pointer to dentry
 900  * @open: open callback
 901  *
 902  * Helper for filesystems that want to use lookup open intents and pass back
 903  * a fully instantiated struct file to the caller.
 904  * This function is meant to be called from within a filesystem's
 905  * lookup method.
 906  * Beware of calling it for non-regular files! Those ->open methods might block
 907  * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
 908  * leading to a deadlock, as nobody can open that fifo anymore, because
 909  * another process to open fifo will block on locked parent when doing lookup).
 910  * Note that in case of error, nd->intent.open.file is destroyed, but the
 911  * path information remains valid.
 912  * If the open callback is set to NULL, then the standard f_op->open()
 913  * filesystem callback is substituted.
 914  */
 915 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
 916                 int (*open)(struct inode *, struct file *))
 917 {
 918         if (IS_ERR(nd->intent.open.file))
 919                 goto out;
 920         if (IS_ERR(dentry))
 921                 goto out_err;
 922         nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
 923                                              nd->intent.open.flags - 1,
 924                                              nd->intent.open.file,
 925                                              open);
 926 out:
 927         return nd->intent.open.file;
 928 out_err:
 929         release_open_intent(nd);
 930         nd->intent.open.file = (struct file *)dentry;
 931         goto out;
 932 }
 933 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
 934
 935 /**
 936  * nameidata_to_filp - convert a nameidata to an open filp.
 937  * @nd: pointer to nameidata
 938  * @flags: open flags
 939  *
 940  * Note that this function destroys the original nameidata
 941  */
 942 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
 943 {
 944         struct file *filp;
 945
 946         /* Pick up the filp from the open intent */
 947         filp = nd->intent.open.file;
 948         /* Has the filesystem initialised the file for us? */
 949         if (filp->f_dentry == NULL)
 950                 filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
 951         else
 952                 path_release(nd);
 953         return filp;
 954 }
 955
 956 /*
 957  * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
 958  * error.
 959  */
 960 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
 961 {
 962         int error;
 963         struct file *f;
 964
 965         error = -ENFILE;
 966         f = get_empty_filp();
 967         if (f == NULL) {
 968                 dput(dentry);
 969                 mntput(mnt);
 970                 return ERR_PTR(error);
 971         }
 972
 973         return __dentry_open(dentry, mnt, flags, f, NULL);
 974 }
 975 EXPORT_SYMBOL(dentry_open);
 976
 977 /*
 978  * Find an empty file descriptor entry, and mark it busy.
 979  */
 980 int get_unused_fd(void)
 981 {
 982         struct files_struct * files = current->files;
 983         int fd, error;
 984         struct fdtable *fdt;
 985
 986         error = -EMFILE;
 987         spin_lock(&files->file_lock);
 988
 989 repeat:
 990         fdt = files_fdtable(files);
 991         fd = find_next_zero_bit(fdt->open_fds->fds_bits,
 992                                 fdt->max_fdset,
 993                                 files->next_fd);
 994
 995         /*
 996          * N.B. For clone tasks sharing a files structure, this test
 997          * will limit the total number of files that can be opened.
 998          */
 999         if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
1000                 goto out;
1001
1002         /* Do we need to expand the fd array or fd set?  */
1003         error = expand_files(files, fd);
1004         if (error < 0)
1005                 goto out;
1006
1007         if (error) {
1008                 /*
1009                  * If we needed to expand the fs array we
1010                  * might have blocked - try again.
1011                  */
1012                 error = -EMFILE;
1013                 goto repeat;
1014         }
1015
1016         FD_SET(fd, fdt->open_fds);
1017         FD_CLR(fd, fdt->close_on_exec);
1018         files->next_fd = fd + 1;
1019 #if 1
1020         /* Sanity check */
1021         if (fdt->fd[fd] != NULL) {
1022                 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
1023                 fdt->fd[fd] = NULL;
1024         }
1025 #endif
1026         error = fd;
1027
1028 out:
1029         spin_unlock(&files->file_lock);
1030         return error;
1031 }
1032
1033 EXPORT_SYMBOL(get_unused_fd);
1034
1035 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
1036 {
1037         struct fdtable *fdt = files_fdtable(files);
1038         __FD_CLR(fd, fdt->open_fds);
1039         if (fd < files->next_fd)
1040                 files->next_fd = fd;
1041 }
1042
1043 void fastcall put_unused_fd(unsigned int fd)
1044 {
1045         struct files_struct *files = current->files;
1046         spin_lock(&files->file_lock);
1047         __put_unused_fd(files, fd);
1048         spin_unlock(&files->file_lock);
1049 }
1050
1051 EXPORT_SYMBOL(put_unused_fd);
1052
1053 /*
1054  * Install a file pointer in the fd array.
1055  *
1056  * The VFS is full of places where we drop the files lock between
1057  * setting the open_fds bitmap and installing the file in the file
1058  * array.  At any such point, we are vulnerable to a dup2() race
1059  * installing a file in the array before us.  We need to detect this and
1060  * fput() the struct file we are about to overwrite in this case.
1061  *
1062  * It should never happen - if we allow dup2() do it, _really_ bad things
1063  * will follow.
1064  */
1065
1066 void fastcall fd_install(unsigned int fd, struct file * file)
1067 {
1068         struct files_struct *files = current->files;
1069         struct fdtable *fdt;
1070         spin_lock(&files->file_lock);
1071         fdt = files_fdtable(files);
1072         BUG_ON(fdt->fd[fd] != NULL);
1073         rcu_assign_pointer(fdt->fd[fd], file);
1074         spin_unlock(&files->file_lock);
1075 }
1076
1077 EXPORT_SYMBOL(fd_install);
1078
1079 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
1080 {
1081         char *tmp = getname(filename);
1082         int fd = PTR_ERR(tmp);
1083
1084         if (!IS_ERR(tmp)) {
1085                 fd = get_unused_fd();
1086                 if (fd >= 0) {
1087                         struct file *f = do_filp_open(dfd, tmp, flags, mode);
1088                         if (IS_ERR(f)) {
1089                                 put_unused_fd(fd);
1090                                 fd = PTR_ERR(f);
1091                         } else {
1092                                 fsnotify_open(f->f_dentry);
1093                                 fd_install(fd, f);
1094                         }
1095                 }
1096                 putname(tmp);
1097         }
1098         return fd;
1099 }
1100
1101 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1102 {
1103         long ret;
1104
1105         if (force_o_largefile())
1106                 flags |= O_LARGEFILE;
1107
1108         ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1109         /* avoid REGPARM breakage on x86: */
1110         prevent_tail_call(ret);
1111         return ret;
1112 }
1113 EXPORT_SYMBOL_GPL(sys_open);
1114
1115 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1116                            int mode)
1117 {
1118         long ret;
1119
1120         if (force_o_largefile())
1121                 flags |= O_LARGEFILE;
1122
1123         ret = do_sys_open(dfd, filename, flags, mode);
1124         /* avoid REGPARM breakage on x86: */
1125         prevent_tail_call(ret);
1126         return ret;
1127 }
1128
1129 #ifndef __alpha__
1130
1131 /*
1132  * For backward compatibility?  Maybe this should be moved
1133  * into arch/i386 instead?
1134  */
1135 asmlinkage long sys_creat(const char __user * pathname, int mode)
1136 {
1137         return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1138 }
1139
1140 #endif
1141
1142 /*
1143  * "id" is the POSIX thread ID. We use the
1144  * files pointer for this..
1145  */
1146 int filp_close(struct file *filp, fl_owner_t id)
1147 {
1148         int retval = 0;
1149
1150         if (!file_count(filp)) {
1151                 printk(KERN_ERR "VFS: Close: file count is 0\n");
1152                 return 0;
1153         }
1154
1155         if (filp->f_op && filp->f_op->flush)
1156                 retval = filp->f_op->flush(filp, id);
1157
1158         dnotify_flush(filp, id);
1159         locks_remove_posix(filp, id);
1160         fput(filp);
1161         return retval;
1162 }
1163
1164 EXPORT_SYMBOL(filp_close);
1165
1166 /*
1167  * Careful here! We test whether the file pointer is NULL before
1168  * releasing the fd. This ensures that one clone task can't release
1169  * an fd while another clone is opening it.
1170  */
1171 asmlinkage long sys_close(unsigned int fd)
1172 {
1173         struct file * filp;
1174         struct files_struct *files = current->files;
1175         struct fdtable *fdt;
1176         int retval;
1177
1178         spin_lock(&files->file_lock);
1179         fdt = files_fdtable(files);
1180         if (fd >= fdt->max_fds)
1181                 goto out_unlock;
1182         filp = fdt->fd[fd];
1183         if (!filp)
1184                 goto out_unlock;
1185         rcu_assign_pointer(fdt->fd[fd], NULL);
1186         FD_CLR(fd, fdt->close_on_exec);
1187         __put_unused_fd(files, fd);
1188         spin_unlock(&files->file_lock);
1189         retval = filp_close(filp, files);
1190
1191         /* can't restart close syscall because file table entry was cleared */
1192         if (unlikely(retval == -ERESTARTSYS ||
1193                      retval == -ERESTARTNOINTR ||
1194                      retval == -ERESTARTNOHAND ||
1195                      retval == -ERESTART_RESTARTBLOCK))
1196                 retval = -EINTR;
1197
1198         return retval;
1199
1200 out_unlock:
1201         spin_unlock(&files->file_lock);
1202         return -EBADF;
1203 }
1204
1205 EXPORT_SYMBOL(sys_close);
1206
1207 /*
1208  * This routine simulates a hangup on the tty, to arrange that users
1209  * are given clean terminals at login time.
1210  */
1211 asmlinkage long sys_vhangup(void)
1212 {
1213         if (capable(CAP_SYS_TTY_CONFIG)) {
1214                 tty_vhangup(current->signal->tty);
1215                 return 0;
1216         }
1217         return -EPERM;
1218 }
1219
1220 /*
1221  * Called when an inode is about to be open.
1222  * We use this to disallow opening large files on 32bit systems if
1223  * the caller didn't specify O_LARGEFILE.  On 64bit systems we force
1224  * on this flag in sys_open.
1225  */
1226 int generic_file_open(struct inode * inode, struct file * filp)
1227 {
1228         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1229                 return -EFBIG;
1230         return 0;
1231 }
1232
1233 EXPORT_SYMBOL(generic_file_open);
1234
1235 /*
1236  * This is used by subsystems that don't want seekable
1237  * file descriptors
1238  */
1239 int nonseekable_open(struct inode *inode, struct file *filp)
1240 {
1241         filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1242         return 0;
1243 }
1244
1245 EXPORT_SYMBOL(nonseekable_open);