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/file.h>
  10 #include <linux/quotaops.h>
  11 #include <linux/fsnotify.h>
  12 #include <linux/module.h>
  13 #include <linux/slab.h>
  14 #include <linux/tty.h>
  15 #include <linux/namei.h>
  16 #include <linux/backing-dev.h>
  17 #include <linux/capability.h>
  18 #include <linux/security.h>
  19 #include <linux/mount.h>
  20 #include <linux/vfs.h>
  21 #include <linux/fcntl.h>
  22 #include <asm/uaccess.h>
  23 #include <linux/fs.h>
  24 #include <linux/personality.h>
  25 #include <linux/pagemap.h>
  26 #include <linux/syscalls.h>
  27 #include <linux/rcupdate.h>
  28 #include <linux/audit.h>
  29
  30 int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
  31 {
  32         int retval = -ENODEV;
  33
  34         if (dentry) {
  35                 retval = -ENOSYS;
  36                 if (dentry->d_sb->s_op->statfs) {
  37                         memset(buf, 0, sizeof(*buf));
  38                         retval = security_sb_statfs(dentry);
  39                         if (retval)
  40                                 return retval;
  41                         retval = dentry->d_sb->s_op->statfs(dentry, buf);
  42                         if (retval == 0 && buf->f_frsize == 0)
  43                                 buf->f_frsize = buf->f_bsize;
  44                 }
  45         }
  46         return retval;
  47 }
  48
  49 EXPORT_SYMBOL(vfs_statfs);
  50
  51 static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
  52 {
  53         struct kstatfs st;
  54         int retval;
  55
  56         retval = vfs_statfs(dentry, &st);
  57         if (retval)
  58                 return retval;
  59
  60         if (sizeof(*buf) == sizeof(st))
  61                 memcpy(buf, &st, sizeof(st));
  62         else {
  63                 if (sizeof buf->f_blocks == 4) {
  64                         if ((st.f_blocks | st.f_bfree | st.f_bavail) &
  65                             0xffffffff00000000ULL)
  66                                 return -EOVERFLOW;
  67                         /*
  68                          * f_files and f_ffree may be -1; it's okay to stuff
  69                          * that into 32 bits
  70                          */
  71                         if (st.f_files != -1 &&
  72                             (st.f_files & 0xffffffff00000000ULL))
  73                                 return -EOVERFLOW;
  74                         if (st.f_ffree != -1 &&
  75                             (st.f_ffree & 0xffffffff00000000ULL))
  76                                 return -EOVERFLOW;
  77                 }
  78
  79                 buf->f_type = st.f_type;
  80                 buf->f_bsize = st.f_bsize;
  81                 buf->f_blocks = st.f_blocks;
  82                 buf->f_bfree = st.f_bfree;
  83                 buf->f_bavail = st.f_bavail;
  84                 buf->f_files = st.f_files;
  85                 buf->f_ffree = st.f_ffree;
  86                 buf->f_fsid = st.f_fsid;
  87                 buf->f_namelen = st.f_namelen;
  88                 buf->f_frsize = st.f_frsize;
  89                 memset(buf->f_spare, 0, sizeof(buf->f_spare));
  90         }
  91         return 0;
  92 }
  93
  94 static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
  95 {
  96         struct kstatfs st;
  97         int retval;
  98
  99         retval = vfs_statfs(dentry, &st);
 100         if (retval)
 101                 return retval;
 102
 103         if (sizeof(*buf) == sizeof(st))
 104                 memcpy(buf, &st, sizeof(st));
 105         else {
 106                 buf->f_type = st.f_type;
 107                 buf->f_bsize = st.f_bsize;
 108                 buf->f_blocks = st.f_blocks;
 109                 buf->f_bfree = st.f_bfree;
 110                 buf->f_bavail = st.f_bavail;
 111                 buf->f_files = st.f_files;
 112                 buf->f_ffree = st.f_ffree;
 113                 buf->f_fsid = st.f_fsid;
 114                 buf->f_namelen = st.f_namelen;
 115                 buf->f_frsize = st.f_frsize;
 116                 memset(buf->f_spare, 0, sizeof(buf->f_spare));
 117         }
 118         return 0;
 119 }
 120
 121 asmlinkage long sys_statfs(const char __user * path, struct statfs __user * buf)
 122 {
 123         struct nameidata nd;
 124         int error;
 125
 126         error = user_path_walk(path, &nd);
 127         if (!error) {
 128                 struct statfs tmp;
 129                 error = vfs_statfs_native(nd.dentry, &tmp);
 130                 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 131                         error = -EFAULT;
 132                 path_release(&nd);
 133         }
 134         return error;
 135 }
 136
 137
 138 asmlinkage long sys_statfs64(const char __user *path, size_t sz, struct statfs64 __user *buf)
 139 {
 140         struct nameidata nd;
 141         long error;
 142
 143         if (sz != sizeof(*buf))
 144                 return -EINVAL;
 145         error = user_path_walk(path, &nd);
 146         if (!error) {
 147                 struct statfs64 tmp;
 148                 error = vfs_statfs64(nd.dentry, &tmp);
 149                 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 150                         error = -EFAULT;
 151                 path_release(&nd);
 152         }
 153         return error;
 154 }
 155
 156
 157 asmlinkage long sys_fstatfs(unsigned int fd, struct statfs __user * buf)
 158 {
 159         struct file * file;
 160         struct statfs tmp;
 161         int error;
 162
 163         error = -EBADF;
 164         file = fget(fd);
 165         if (!file)
 166                 goto out;
 167         error = vfs_statfs_native(file->f_path.dentry, &tmp);
 168         if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 169                 error = -EFAULT;
 170         fput(file);
 171 out:
 172         return error;
 173 }
 174
 175 asmlinkage long sys_fstatfs64(unsigned int fd, size_t sz, struct statfs64 __user *buf)
 176 {
 177         struct file * file;
 178         struct statfs64 tmp;
 179         int error;
 180
 181         if (sz != sizeof(*buf))
 182                 return -EINVAL;
 183
 184         error = -EBADF;
 185         file = fget(fd);
 186         if (!file)
 187                 goto out;
 188         error = vfs_statfs64(file->f_path.dentry, &tmp);
 189         if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
 190                 error = -EFAULT;
 191         fput(file);
 192 out:
 193         return error;
 194 }
 195
 196 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
 197         struct file *filp)
 198 {
 199         int err;
 200         struct iattr newattrs;
 201
 202         /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
 203         if (length < 0)
 204                 return -EINVAL;
 205
 206         newattrs.ia_size = length;
 207         newattrs.ia_valid = ATTR_SIZE | time_attrs;
 208         if (filp) {
 209                 newattrs.ia_file = filp;
 210                 newattrs.ia_valid |= ATTR_FILE;
 211         }
 212
 213         /* Remove suid/sgid on truncate too */
 214         newattrs.ia_valid |= should_remove_suid(dentry);
 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_path.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, ATTR_MTIME|ATTR_CTIME, 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         long ret = do_sys_ftruncate(fd, length, 1);
 334         /* avoid REGPARM breakage on x86: */
 335         prevent_tail_call(ret);
 336         return ret;
 337 }
 338
 339 /* LFS versions of truncate are only needed on 32 bit machines */
 340 #if BITS_PER_LONG == 32
 341 asmlinkage long sys_truncate64(const char __user * path, loff_t length)
 342 {
 343         return do_sys_truncate(path, length);
 344 }
 345
 346 asmlinkage long sys_ftruncate64(unsigned int fd, loff_t length)
 347 {
 348         long ret = do_sys_ftruncate(fd, length, 0);
 349         /* avoid REGPARM breakage on x86: */
 350         prevent_tail_call(ret);
 351         return ret;
 352 }
 353 #endif
 354
 355 /*
 356  * access() needs to use the real uid/gid, not the effective uid/gid.
 357  * We do this by temporarily clearing all FS-related capabilities and
 358  * switching the fsuid/fsgid around to the real ones.
 359  */
 360 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
 361 {
 362         struct nameidata nd;
 363         int old_fsuid, old_fsgid;
 364         kernel_cap_t old_cap;
 365         int res;
 366
 367         if (mode & ~S_IRWXO)    /* where's F_OK, X_OK, W_OK, R_OK? */
 368                 return -EINVAL;
 369
 370         old_fsuid = current->fsuid;
 371         old_fsgid = current->fsgid;
 372         old_cap = current->cap_effective;
 373
 374         current->fsuid = current->uid;
 375         current->fsgid = current->gid;
 376
 377         /*
 378          * Clear the capabilities if we switch to a non-root user
 379          *
 380          * FIXME: There is a race here against sys_capset.  The
 381          * capabilities can change yet we will restore the old
 382          * value below.  We should hold task_capabilities_lock,
 383          * but we cannot because user_path_walk can sleep.
 384          */
 385         if (current->uid)
 386                 cap_clear(current->cap_effective);
 387         else
 388                 current->cap_effective = current->cap_permitted;
 389
 390         res = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW|LOOKUP_ACCESS, &nd);
 391         if (res)
 392                 goto out;
 393
 394         res = vfs_permission(&nd, mode);
 395         /* SuS v2 requires we report a read only fs too */
 396         if(res || !(mode & S_IWOTH) ||
 397            special_file(nd.dentry->d_inode->i_mode))
 398                 goto out_path_release;
 399
 400         if(IS_RDONLY(nd.dentry->d_inode))
 401                 res = -EROFS;
 402
 403 out_path_release:
 404         path_release(&nd);
 405 out:
 406         current->fsuid = old_fsuid;
 407         current->fsgid = old_fsgid;
 408         current->cap_effective = old_cap;
 409
 410         return res;
 411 }
 412
 413 asmlinkage long sys_access(const char __user *filename, int mode)
 414 {
 415         return sys_faccessat(AT_FDCWD, filename, mode);
 416 }
 417
 418 asmlinkage long sys_chdir(const char __user * filename)
 419 {
 420         struct nameidata nd;
 421         int error;
 422
 423         error = __user_walk(filename,
 424                             LOOKUP_FOLLOW|LOOKUP_DIRECTORY|LOOKUP_CHDIR, &nd);
 425         if (error)
 426                 goto out;
 427
 428         error = vfs_permission(&nd, MAY_EXEC);
 429         if (error)
 430                 goto dput_and_out;
 431
 432         set_fs_pwd(current->fs, nd.mnt, nd.dentry);
 433
 434 dput_and_out:
 435         path_release(&nd);
 436 out:
 437         return error;
 438 }
 439
 440 asmlinkage long sys_fchdir(unsigned int fd)
 441 {
 442         struct file *file;
 443         struct dentry *dentry;
 444         struct inode *inode;
 445         struct vfsmount *mnt;
 446         int error;
 447
 448         error = -EBADF;
 449         file = fget(fd);
 450         if (!file)
 451                 goto out;
 452
 453         dentry = file->f_path.dentry;
 454         mnt = file->f_path.mnt;
 455         inode = dentry->d_inode;
 456
 457         error = -ENOTDIR;
 458         if (!S_ISDIR(inode->i_mode))
 459                 goto out_putf;
 460
 461         error = file_permission(file, MAY_EXEC);
 462         if (!error)
 463                 set_fs_pwd(current->fs, mnt, dentry);
 464 out_putf:
 465         fput(file);
 466 out:
 467         return error;
 468 }
 469
 470 asmlinkage long sys_chroot(const char __user * filename)
 471 {
 472         struct nameidata nd;
 473         int error;
 474
 475         error = __user_walk(filename, LOOKUP_FOLLOW | LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
 476         if (error)
 477                 goto out;
 478
 479         error = vfs_permission(&nd, MAY_EXEC);
 480         if (error)
 481                 goto dput_and_out;
 482
 483         error = -EPERM;
 484         if (!capable(CAP_SYS_CHROOT))
 485                 goto dput_and_out;
 486
 487         set_fs_root(current->fs, nd.mnt, nd.dentry);
 488         set_fs_altroot();
 489         error = 0;
 490 dput_and_out:
 491         path_release(&nd);
 492 out:
 493         return error;
 494 }
 495
 496 asmlinkage long sys_fchmod(unsigned int fd, mode_t mode)
 497 {
 498         struct inode * inode;
 499         struct dentry * dentry;
 500         struct file * file;
 501         int err = -EBADF;
 502         struct iattr newattrs;
 503
 504         file = fget(fd);
 505         if (!file)
 506                 goto out;
 507
 508         dentry = file->f_path.dentry;
 509         inode = dentry->d_inode;
 510
 511         audit_inode(NULL, inode);
 512
 513         err = -EROFS;
 514         if (IS_RDONLY(inode))
 515                 goto out_putf;
 516         err = -EPERM;
 517         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 518                 goto out_putf;
 519         mutex_lock(&inode->i_mutex);
 520         if (mode == (mode_t) -1)
 521                 mode = inode->i_mode;
 522         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 523         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 524         err = notify_change(dentry, &newattrs);
 525         mutex_unlock(&inode->i_mutex);
 526
 527 out_putf:
 528         fput(file);
 529 out:
 530         return err;
 531 }
 532
 533 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
 534                              mode_t mode)
 535 {
 536         struct nameidata nd;
 537         struct inode * inode;
 538         int error;
 539         struct iattr newattrs;
 540
 541         error = __user_walk_fd(dfd, filename, LOOKUP_FOLLOW, &nd);
 542         if (error)
 543                 goto out;
 544         inode = nd.dentry->d_inode;
 545
 546         error = -EROFS;
 547         if (IS_RDONLY(inode))
 548                 goto dput_and_out;
 549
 550         error = -EPERM;
 551         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 552                 goto dput_and_out;
 553
 554         mutex_lock(&inode->i_mutex);
 555         if (mode == (mode_t) -1)
 556                 mode = inode->i_mode;
 557         newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
 558         newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
 559         error = notify_change(nd.dentry, &newattrs);
 560         mutex_unlock(&inode->i_mutex);
 561
 562 dput_and_out:
 563         path_release(&nd);
 564 out:
 565         return error;
 566 }
 567
 568 asmlinkage long sys_chmod(const char __user *filename, mode_t mode)
 569 {
 570         return sys_fchmodat(AT_FDCWD, filename, mode);
 571 }
 572
 573 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
 574 {
 575         struct inode * inode;
 576         int error;
 577         struct iattr newattrs;
 578
 579         error = -ENOENT;
 580         if (!(inode = dentry->d_inode)) {
 581                 printk(KERN_ERR "chown_common: NULL inode\n");
 582                 goto out;
 583         }
 584         error = -EROFS;
 585         if (IS_RDONLY(inode))
 586                 goto out;
 587         error = -EPERM;
 588         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 589                 goto out;
 590         newattrs.ia_valid =  ATTR_CTIME;
 591         if (user != (uid_t) -1) {
 592                 newattrs.ia_valid |= ATTR_UID;
 593                 newattrs.ia_uid = user;
 594         }
 595         if (group != (gid_t) -1) {
 596                 newattrs.ia_valid |= ATTR_GID;
 597                 newattrs.ia_gid = group;
 598         }
 599         if (!S_ISDIR(inode->i_mode))
 600                 newattrs.ia_valid |= ATTR_KILL_SUID|ATTR_KILL_SGID;
 601         mutex_lock(&inode->i_mutex);
 602         error = notify_change(dentry, &newattrs);
 603         mutex_unlock(&inode->i_mutex);
 604 out:
 605         return error;
 606 }
 607
 608 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
 609 {
 610         struct nameidata nd;
 611         int error;
 612
 613         error = user_path_walk(filename, &nd);
 614         if (error)
 615                 goto out;
 616         error = chown_common(nd.dentry, user, group);
 617         path_release(&nd);
 618 out:
 619         return error;
 620 }
 621
 622 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
 623                              gid_t group, int flag)
 624 {
 625         struct nameidata nd;
 626         int error = -EINVAL;
 627         int follow;
 628
 629         if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
 630                 goto out;
 631
 632         follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
 633         error = __user_walk_fd(dfd, filename, follow, &nd);
 634         if (error)
 635                 goto out;
 636         error = chown_common(nd.dentry, user, group);
 637         path_release(&nd);
 638 out:
 639         return error;
 640 }
 641
 642 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
 643 {
 644         struct nameidata nd;
 645         int error;
 646
 647         error = user_path_walk_link(filename, &nd);
 648         if (error)
 649                 goto out;
 650         error = chown_common(nd.dentry, user, group);
 651         path_release(&nd);
 652 out:
 653         return error;
 654 }
 655
 656
 657 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
 658 {
 659         struct file * file;
 660         int error = -EBADF;
 661         struct dentry * dentry;
 662
 663         file = fget(fd);
 664         if (!file)
 665                 goto out;
 666
 667         dentry = file->f_path.dentry;
 668         audit_inode(NULL, dentry->d_inode);
 669         error = chown_common(dentry, user, group);
 670         fput(file);
 671 out:
 672         return error;
 673 }
 674
 675 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
 676                                         int flags, struct file *f,
 677                                         int (*open)(struct inode *, struct file *))
 678 {
 679         struct inode *inode;
 680         int error;
 681
 682         f->f_flags = flags;
 683         f->f_mode = ((flags+1) & O_ACCMODE) | FMODE_LSEEK |
 684                                 FMODE_PREAD | FMODE_PWRITE;
 685         inode = dentry->d_inode;
 686         if (f->f_mode & FMODE_WRITE) {
 687                 error = get_write_access(inode);
 688                 if (error)
 689                         goto cleanup_file;
 690         }
 691
 692         f->f_mapping = inode->i_mapping;
 693         f->f_path.dentry = dentry;
 694         f->f_path.mnt = mnt;
 695         f->f_pos = 0;
 696         f->f_op = fops_get(inode->i_fop);
 697         file_move(f, &inode->i_sb->s_files);
 698
 699         if (!open && f->f_op)
 700                 open = f->f_op->open;
 701         if (open) {
 702                 error = open(inode, f);
 703                 if (error)
 704                         goto cleanup_all;
 705         }
 706
 707         f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
 708
 709         file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
 710
 711         /* NB: we're sure to have correct a_ops only after f_op->open */
 712         if (f->f_flags & O_DIRECT) {
 713                 if (!f->f_mapping->a_ops ||
 714                     ((!f->f_mapping->a_ops->direct_IO) &&
 715                     (!f->f_mapping->a_ops->get_xip_page))) {
 716                         fput(f);
 717                         f = ERR_PTR(-EINVAL);
 718                 }
 719         }
 720
 721         return f;
 722
 723 cleanup_all:
 724         fops_put(f->f_op);
 725         if (f->f_mode & FMODE_WRITE)
 726                 put_write_access(inode);
 727         file_kill(f);
 728         f->f_path.dentry = NULL;
 729         f->f_path.mnt = NULL;
 730 cleanup_file:
 731         put_filp(f);
 732         dput(dentry);
 733         mntput(mnt);
 734         return ERR_PTR(error);
 735 }
 736
 737 /*
 738  * Note that while the flag value (low two bits) for sys_open means:
 739  *      00 - read-only
 740  *      01 - write-only
 741  *      10 - read-write
 742  *      11 - special
 743  * it is changed into
 744  *      00 - no permissions needed
 745  *      01 - read-permission
 746  *      10 - write-permission
 747  *      11 - read-write
 748  * for the internal routines (ie open_namei()/follow_link() etc). 00 is
 749  * used by symlinks.
 750  */
 751 static struct file *do_filp_open(int dfd, const char *filename, int flags,
 752                                  int mode)
 753 {
 754         int namei_flags, error;
 755         struct nameidata nd;
 756
 757         namei_flags = flags;
 758         if ((namei_flags+1) & O_ACCMODE)
 759                 namei_flags++;
 760
 761         error = open_namei(dfd, filename, namei_flags, mode, &nd);
 762         if (!error)
 763                 return nameidata_to_filp(&nd, flags);
 764
 765         return ERR_PTR(error);
 766 }
 767
 768 struct file *filp_open(const char *filename, int flags, int mode)
 769 {
 770         return do_filp_open(AT_FDCWD, filename, flags, mode);
 771 }
 772 EXPORT_SYMBOL(filp_open);
 773
 774 /**
 775  * lookup_instantiate_filp - instantiates the open intent filp
 776  * @nd: pointer to nameidata
 777  * @dentry: pointer to dentry
 778  * @open: open callback
 779  *
 780  * Helper for filesystems that want to use lookup open intents and pass back
 781  * a fully instantiated struct file to the caller.
 782  * This function is meant to be called from within a filesystem's
 783  * lookup method.
 784  * Beware of calling it for non-regular files! Those ->open methods might block
 785  * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
 786  * leading to a deadlock, as nobody can open that fifo anymore, because
 787  * another process to open fifo will block on locked parent when doing lookup).
 788  * Note that in case of error, nd->intent.open.file is destroyed, but the
 789  * path information remains valid.
 790  * If the open callback is set to NULL, then the standard f_op->open()
 791  * filesystem callback is substituted.
 792  */
 793 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
 794                 int (*open)(struct inode *, struct file *))
 795 {
 796         if (IS_ERR(nd->intent.open.file))
 797                 goto out;
 798         if (IS_ERR(dentry))
 799                 goto out_err;
 800         nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->mnt),
 801                                              nd->intent.open.flags - 1,
 802                                              nd->intent.open.file,
 803                                              open);
 804 out:
 805         return nd->intent.open.file;
 806 out_err:
 807         release_open_intent(nd);
 808         nd->intent.open.file = (struct file *)dentry;
 809         goto out;
 810 }
 811 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
 812
 813 /**
 814  * nameidata_to_filp - convert a nameidata to an open filp.
 815  * @nd: pointer to nameidata
 816  * @flags: open flags
 817  *
 818  * Note that this function destroys the original nameidata
 819  */
 820 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
 821 {
 822         struct file *filp;
 823
 824         /* Pick up the filp from the open intent */
 825         filp = nd->intent.open.file;
 826         /* Has the filesystem initialised the file for us? */
 827         if (filp->f_path.dentry == NULL)
 828                 filp = __dentry_open(nd->dentry, nd->mnt, flags, filp, NULL);
 829         else
 830                 path_release(nd);
 831         return filp;
 832 }
 833
 834 /*
 835  * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
 836  * error.
 837  */
 838 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags)
 839 {
 840         int error;
 841         struct file *f;
 842
 843         error = -ENFILE;
 844         f = get_empty_filp();
 845         if (f == NULL) {
 846                 dput(dentry);
 847                 mntput(mnt);
 848                 return ERR_PTR(error);
 849         }
 850
 851         return __dentry_open(dentry, mnt, flags, f, NULL);
 852 }
 853 EXPORT_SYMBOL(dentry_open);
 854
 855 /*
 856  * Find an empty file descriptor entry, and mark it busy.
 857  */
 858 int get_unused_fd_flags(int flags)
 859 {
 860         struct files_struct * files = current->files;
 861         int fd, error;
 862         struct fdtable *fdt;
 863
 864         error = -EMFILE;
 865         spin_lock(&files->file_lock);
 866
 867 repeat:
 868         fdt = files_fdtable(files);
 869         fd = find_next_zero_bit(fdt->open_fds->fds_bits, fdt->max_fds,
 870                                 files->next_fd);
 871
 872         /*
 873          * N.B. For clone tasks sharing a files structure, this test
 874          * will limit the total number of files that can be opened.
 875          */
 876         if (fd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
 877                 goto out;
 878
 879         /* Do we need to expand the fd array or fd set?  */
 880         error = expand_files(files, fd);
 881         if (error < 0)
 882                 goto out;
 883
 884         if (error) {
 885                 /*
 886                  * If we needed to expand the fs array we
 887                  * might have blocked - try again.
 888                  */
 889                 error = -EMFILE;
 890                 goto repeat;
 891         }
 892
 893         FD_SET(fd, fdt->open_fds);
 894         if (flags & O_CLOEXEC)
 895                 FD_SET(fd, fdt->close_on_exec);
 896         else
 897                 FD_CLR(fd, fdt->close_on_exec);
 898         files->next_fd = fd + 1;
 899 #if 1
 900         /* Sanity check */
 901         if (fdt->fd[fd] != NULL) {
 902                 printk(KERN_WARNING "get_unused_fd: slot %d not NULL!\n", fd);
 903                 fdt->fd[fd] = NULL;
 904         }
 905 #endif
 906         error = fd;
 907
 908 out:
 909         spin_unlock(&files->file_lock);
 910         return error;
 911 }
 912
 913 int get_unused_fd(void)
 914 {
 915         return get_unused_fd_flags(0);
 916 }
 917
 918 EXPORT_SYMBOL(get_unused_fd);
 919
 920 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
 921 {
 922         struct fdtable *fdt = files_fdtable(files);
 923         __FD_CLR(fd, fdt->open_fds);
 924         if (fd < files->next_fd)
 925                 files->next_fd = fd;
 926 }
 927
 928 void fastcall put_unused_fd(unsigned int fd)
 929 {
 930         struct files_struct *files = current->files;
 931         spin_lock(&files->file_lock);
 932         __put_unused_fd(files, fd);
 933         spin_unlock(&files->file_lock);
 934 }
 935
 936 EXPORT_SYMBOL(put_unused_fd);
 937
 938 /*
 939  * Install a file pointer in the fd array.
 940  *
 941  * The VFS is full of places where we drop the files lock between
 942  * setting the open_fds bitmap and installing the file in the file
 943  * array.  At any such point, we are vulnerable to a dup2() race
 944  * installing a file in the array before us.  We need to detect this and
 945  * fput() the struct file we are about to overwrite in this case.
 946  *
 947  * It should never happen - if we allow dup2() do it, _really_ bad things
 948  * will follow.
 949  */
 950
 951 void fastcall fd_install(unsigned int fd, struct file * file)
 952 {
 953         struct files_struct *files = current->files;
 954         struct fdtable *fdt;
 955         spin_lock(&files->file_lock);
 956         fdt = files_fdtable(files);
 957         BUG_ON(fdt->fd[fd] != NULL);
 958         rcu_assign_pointer(fdt->fd[fd], file);
 959         spin_unlock(&files->file_lock);
 960 }
 961
 962 EXPORT_SYMBOL(fd_install);
 963
 964 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
 965 {
 966         char *tmp = getname(filename);
 967         int fd = PTR_ERR(tmp);
 968
 969         if (!IS_ERR(tmp)) {
 970                 fd = get_unused_fd_flags(flags);
 971                 if (fd >= 0) {
 972                         struct file *f = do_filp_open(dfd, tmp, flags, mode);
 973                         if (IS_ERR(f)) {
 974                                 put_unused_fd(fd);
 975                                 fd = PTR_ERR(f);
 976                         } else {
 977                                 fsnotify_open(f->f_path.dentry);
 978                                 fd_install(fd, f);
 979                         }
 980                 }
 981                 putname(tmp);
 982         }
 983         return fd;
 984 }
 985
 986 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
 987 {
 988         long ret;
 989
 990         if (force_o_largefile())
 991                 flags |= O_LARGEFILE;
 992
 993         ret = do_sys_open(AT_FDCWD, filename, flags, mode);
 994         /* avoid REGPARM breakage on x86: */
 995         prevent_tail_call(ret);
 996         return ret;
 997 }
 998 EXPORT_SYMBOL_GPL(sys_open);
 999
1000 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1001                            int mode)
1002 {
1003         long ret;
1004
1005         if (force_o_largefile())
1006                 flags |= O_LARGEFILE;
1007
1008         ret = do_sys_open(dfd, filename, flags, mode);
1009         /* avoid REGPARM breakage on x86: */
1010         prevent_tail_call(ret);
1011         return ret;
1012 }
1013
1014 #ifndef __alpha__
1015
1016 /*
1017  * For backward compatibility?  Maybe this should be moved
1018  * into arch/i386 instead?
1019  */
1020 asmlinkage long sys_creat(const char __user * pathname, int mode)
1021 {
1022         return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1023 }
1024
1025 #endif
1026
1027 /*
1028  * "id" is the POSIX thread ID. We use the
1029  * files pointer for this..
1030  */
1031 int filp_close(struct file *filp, fl_owner_t id)
1032 {
1033         int retval = 0;
1034
1035         if (!file_count(filp)) {
1036                 printk(KERN_ERR "VFS: Close: file count is 0\n");
1037                 return 0;
1038         }
1039
1040         if (filp->f_op && filp->f_op->flush)
1041                 retval = filp->f_op->flush(filp, id);
1042
1043         dnotify_flush(filp, id);
1044         locks_remove_posix(filp, id);
1045         fput(filp);
1046         return retval;
1047 }
1048
1049 EXPORT_SYMBOL(filp_close);
1050
1051 /*
1052  * Careful here! We test whether the file pointer is NULL before
1053  * releasing the fd. This ensures that one clone task can't release
1054  * an fd while another clone is opening it.
1055  */
1056 asmlinkage long sys_close(unsigned int fd)
1057 {
1058         struct file * filp;
1059         struct files_struct *files = current->files;
1060         struct fdtable *fdt;
1061         int retval;
1062
1063         spin_lock(&files->file_lock);
1064         fdt = files_fdtable(files);
1065         if (fd >= fdt->max_fds)
1066                 goto out_unlock;
1067         filp = fdt->fd[fd];
1068         if (!filp)
1069                 goto out_unlock;
1070         rcu_assign_pointer(fdt->fd[fd], NULL);
1071         FD_CLR(fd, fdt->close_on_exec);
1072         __put_unused_fd(files, fd);
1073         spin_unlock(&files->file_lock);
1074         retval = filp_close(filp, files);
1075
1076         /* can't restart close syscall because file table entry was cleared */
1077         if (unlikely(retval == -ERESTARTSYS ||
1078                      retval == -ERESTARTNOINTR ||
1079                      retval == -ERESTARTNOHAND ||
1080                      retval == -ERESTART_RESTARTBLOCK))
1081                 retval = -EINTR;
1082
1083         return retval;
1084
1085 out_unlock:
1086         spin_unlock(&files->file_lock);
1087         return -EBADF;
1088 }
1089
1090 EXPORT_SYMBOL(sys_close);
1091
1092 /*
1093  * This routine simulates a hangup on the tty, to arrange that users
1094  * are given clean terminals at login time.
1095  */
1096 asmlinkage long sys_vhangup(void)
1097 {
1098         if (capable(CAP_SYS_TTY_CONFIG)) {
1099                 /* XXX: this needs locking */
1100                 tty_vhangup(current->signal->tty);
1101                 return 0;
1102         }
1103         return -EPERM;
1104 }
1105
1106 /*
1107  * Called when an inode is about to be open.
1108  * We use this to disallow opening large files on 32bit systems if
1109  * the caller didn't specify O_LARGEFILE.  On 64bit systems we force
1110  * on this flag in sys_open.
1111  */
1112 int generic_file_open(struct inode * inode, struct file * filp)
1113 {
1114         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1115                 return -EFBIG;
1116         return 0;
1117 }
1118
1119 EXPORT_SYMBOL(generic_file_open);
1120
1121 /*
1122  * This is used by subsystems that don't want seekable
1123  * file descriptors
1124  */
1125 int nonseekable_open(struct inode *inode, struct file *filp)
1126 {
1127         filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1128         return 0;
1129 }
1130
1131 EXPORT_SYMBOL(nonseekable_open);