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