4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/string.h>
9 #include <linux/file.h>
10 #include <linux/fdtable.h>
11 #include <linux/fsnotify.h>
12 #include <linux/module.h>
13 #include <linux/tty.h>
14 #include <linux/namei.h>
15 #include <linux/backing-dev.h>
16 #include <linux/capability.h>
17 #include <linux/securebits.h>
18 #include <linux/security.h>
19 #include <linux/mount.h>
20 #include <linux/fcntl.h>
21 #include <linux/slab.h>
22 #include <asm/uaccess.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 #include <linux/falloc.h>
30 #include <linux/fs_struct.h>
31 #include <linux/ima.h>
32 #include <linux/dnotify.h>
33 #include <linux/compat.h>
37 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
41 struct iattr newattrs;
43 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
47 newattrs.ia_size = length;
48 newattrs.ia_valid = ATTR_SIZE | time_attrs;
50 newattrs.ia_file = filp;
51 newattrs.ia_valid |= ATTR_FILE;
54 /* Remove suid, sgid, and file capabilities on truncate too */
55 ret = dentry_needs_remove_privs(dentry);
59 newattrs.ia_valid |= ret | ATTR_FORCE;
61 inode_lock(dentry->d_inode);
62 /* Note any delegations or leases have already been broken: */
63 ret = notify_change(dentry, &newattrs, NULL);
64 inode_unlock(dentry->d_inode);
68 long vfs_truncate(const struct path *path, loff_t length)
71 struct dentry *upperdentry;
74 inode = path->dentry->d_inode;
76 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
77 if (S_ISDIR(inode->i_mode))
79 if (!S_ISREG(inode->i_mode))
82 error = mnt_want_write(path->mnt);
86 error = inode_permission(inode, MAY_WRITE);
88 goto mnt_drop_write_and_out;
92 goto mnt_drop_write_and_out;
95 * If this is an overlayfs then do as if opening the file so we get
96 * write access on the upper inode, not on the overlay inode. For
97 * non-overlay filesystems d_real() is an identity function.
99 upperdentry = d_real(path->dentry, NULL, O_WRONLY);
100 error = PTR_ERR(upperdentry);
101 if (IS_ERR(upperdentry))
102 goto mnt_drop_write_and_out;
104 error = get_write_access(upperdentry->d_inode);
106 goto mnt_drop_write_and_out;
109 * Make sure that there are no leases. get_write_access() protects
110 * against the truncate racing with a lease-granting setlease().
112 error = break_lease(inode, O_WRONLY);
114 goto put_write_and_out;
116 error = locks_verify_truncate(inode, NULL, length);
118 error = security_path_truncate(path);
120 error = do_truncate(path->dentry, length, 0, NULL);
123 put_write_access(upperdentry->d_inode);
124 mnt_drop_write_and_out:
125 mnt_drop_write(path->mnt);
129 EXPORT_SYMBOL_GPL(vfs_truncate);
131 static long do_sys_truncate(const char __user *pathname, loff_t length)
133 unsigned int lookup_flags = LOOKUP_FOLLOW;
137 if (length < 0) /* sorry, but loff_t says... */
141 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
143 error = vfs_truncate(&path, length);
146 if (retry_estale(error, lookup_flags)) {
147 lookup_flags |= LOOKUP_REVAL;
153 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
155 return do_sys_truncate(path, length);
159 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
161 return do_sys_truncate(path, length);
165 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
168 struct dentry *dentry;
180 /* explicitly opened as large or we are on 64-bit box */
181 if (f.file->f_flags & O_LARGEFILE)
184 dentry = f.file->f_path.dentry;
185 inode = dentry->d_inode;
187 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
191 /* Cannot ftruncate over 2^31 bytes without large file support */
192 if (small && length > MAX_NON_LFS)
196 if (IS_APPEND(inode))
199 sb_start_write(inode->i_sb);
200 error = locks_verify_truncate(inode, f.file, length);
202 error = security_path_truncate(&f.file->f_path);
204 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
205 sb_end_write(inode->i_sb);
212 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
214 return do_sys_ftruncate(fd, length, 1);
218 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
220 return do_sys_ftruncate(fd, length, 1);
224 /* LFS versions of truncate are only needed on 32 bit machines */
225 #if BITS_PER_LONG == 32
226 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
228 return do_sys_truncate(path, length);
231 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
233 return do_sys_ftruncate(fd, length, 0);
235 #endif /* BITS_PER_LONG == 32 */
238 int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
240 struct inode *inode = file_inode(file);
243 if (offset < 0 || len <= 0)
246 /* Return error if mode is not supported */
247 if (mode & ~FALLOC_FL_SUPPORTED_MASK)
250 /* Punch hole and zero range are mutually exclusive */
251 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
252 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
255 /* Punch hole must have keep size set */
256 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
257 !(mode & FALLOC_FL_KEEP_SIZE))
260 /* Collapse range should only be used exclusively. */
261 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
262 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
265 /* Insert range should only be used exclusively. */
266 if ((mode & FALLOC_FL_INSERT_RANGE) &&
267 (mode & ~FALLOC_FL_INSERT_RANGE))
270 /* Unshare range should only be used with allocate mode. */
271 if ((mode & FALLOC_FL_UNSHARE_RANGE) &&
272 (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE)))
275 if (!(file->f_mode & FMODE_WRITE))
279 * We can only allow pure fallocate on append only files
281 if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
284 if (IS_IMMUTABLE(inode))
288 * We cannot allow any fallocate operation on an active swapfile
290 if (IS_SWAPFILE(inode))
294 * Revalidate the write permissions, in case security policy has
295 * changed since the files were opened.
297 ret = security_file_permission(file, MAY_WRITE);
301 if (S_ISFIFO(inode->i_mode))
305 * Let individual file system decide if it supports preallocation
306 * for directories or not.
308 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode) &&
309 !S_ISBLK(inode->i_mode))
312 /* Check for wrap through zero too */
313 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
316 if (!file->f_op->fallocate)
319 sb_start_write(inode->i_sb);
320 ret = file->f_op->fallocate(file, mode, offset, len);
323 * Create inotify and fanotify events.
325 * To keep the logic simple always create events if fallocate succeeds.
326 * This implies that events are even created if the file size remains
327 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
330 fsnotify_modify(file);
332 sb_end_write(inode->i_sb);
335 EXPORT_SYMBOL_GPL(vfs_fallocate);
337 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
339 struct fd f = fdget(fd);
343 error = vfs_fallocate(f.file, mode, offset, len);
350 * access() needs to use the real uid/gid, not the effective uid/gid.
351 * We do this by temporarily clearing all FS-related capabilities and
352 * switching the fsuid/fsgid around to the real ones.
354 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
356 const struct cred *old_cred;
357 struct cred *override_cred;
361 unsigned int lookup_flags = LOOKUP_FOLLOW;
363 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
366 override_cred = prepare_creds();
370 override_cred->fsuid = override_cred->uid;
371 override_cred->fsgid = override_cred->gid;
373 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
374 /* Clear the capabilities if we switch to a non-root user */
375 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
376 if (!uid_eq(override_cred->uid, root_uid))
377 cap_clear(override_cred->cap_effective);
379 override_cred->cap_effective =
380 override_cred->cap_permitted;
383 old_cred = override_creds(override_cred);
385 res = user_path_at(dfd, filename, lookup_flags, &path);
389 inode = d_backing_inode(path.dentry);
391 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
393 * MAY_EXEC on regular files is denied if the fs is mounted
394 * with the "noexec" flag.
397 if (path_noexec(&path))
398 goto out_path_release;
401 res = inode_permission(inode, mode | MAY_ACCESS);
402 /* SuS v2 requires we report a read only fs too */
403 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
404 goto out_path_release;
406 * This is a rare case where using __mnt_is_readonly()
407 * is OK without a mnt_want/drop_write() pair. Since
408 * no actual write to the fs is performed here, we do
409 * not need to telegraph to that to anyone.
411 * By doing this, we accept that this access is
412 * inherently racy and know that the fs may change
413 * state before we even see this result.
415 if (__mnt_is_readonly(path.mnt))
420 if (retry_estale(res, lookup_flags)) {
421 lookup_flags |= LOOKUP_REVAL;
425 revert_creds(old_cred);
426 put_cred(override_cred);
430 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
432 return sys_faccessat(AT_FDCWD, filename, mode);
435 SYSCALL_DEFINE1(chdir, const char __user *, filename)
439 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
441 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
445 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
449 set_fs_pwd(current->fs, &path);
453 if (retry_estale(error, lookup_flags)) {
454 lookup_flags |= LOOKUP_REVAL;
461 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
463 struct fd f = fdget_raw(fd);
471 inode = file_inode(f.file);
474 if (!S_ISDIR(inode->i_mode))
477 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
479 set_fs_pwd(current->fs, &f.file->f_path);
486 SYSCALL_DEFINE1(chroot, const char __user *, filename)
490 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
492 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
496 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
501 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
503 error = security_path_chroot(&path);
507 set_fs_root(current->fs, &path);
511 if (retry_estale(error, lookup_flags)) {
512 lookup_flags |= LOOKUP_REVAL;
519 static int chmod_common(const struct path *path, umode_t mode)
521 struct inode *inode = path->dentry->d_inode;
522 struct inode *delegated_inode = NULL;
523 struct iattr newattrs;
526 error = mnt_want_write(path->mnt);
531 error = security_path_chmod(path, mode);
534 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
535 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
536 error = notify_change(path->dentry, &newattrs, &delegated_inode);
539 if (delegated_inode) {
540 error = break_deleg_wait(&delegated_inode);
544 mnt_drop_write(path->mnt);
548 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
550 struct fd f = fdget(fd);
555 err = chmod_common(&f.file->f_path, mode);
561 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
565 unsigned int lookup_flags = LOOKUP_FOLLOW;
567 error = user_path_at(dfd, filename, lookup_flags, &path);
569 error = chmod_common(&path, mode);
571 if (retry_estale(error, lookup_flags)) {
572 lookup_flags |= LOOKUP_REVAL;
579 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
581 return sys_fchmodat(AT_FDCWD, filename, mode);
584 static int chown_common(const struct path *path, uid_t user, gid_t group)
586 struct inode *inode = path->dentry->d_inode;
587 struct inode *delegated_inode = NULL;
589 struct iattr newattrs;
593 uid = make_kuid(current_user_ns(), user);
594 gid = make_kgid(current_user_ns(), group);
597 newattrs.ia_valid = ATTR_CTIME;
598 if (user != (uid_t) -1) {
601 newattrs.ia_valid |= ATTR_UID;
602 newattrs.ia_uid = uid;
604 if (group != (gid_t) -1) {
607 newattrs.ia_valid |= ATTR_GID;
608 newattrs.ia_gid = gid;
610 if (!S_ISDIR(inode->i_mode))
612 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
614 error = security_path_chown(path, uid, gid);
616 error = notify_change(path->dentry, &newattrs, &delegated_inode);
618 if (delegated_inode) {
619 error = break_deleg_wait(&delegated_inode);
626 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
627 gid_t, group, int, flag)
633 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
636 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
637 if (flag & AT_EMPTY_PATH)
638 lookup_flags |= LOOKUP_EMPTY;
640 error = user_path_at(dfd, filename, lookup_flags, &path);
643 error = mnt_want_write(path.mnt);
646 error = chown_common(&path, user, group);
647 mnt_drop_write(path.mnt);
650 if (retry_estale(error, lookup_flags)) {
651 lookup_flags |= LOOKUP_REVAL;
658 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
660 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
663 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
665 return sys_fchownat(AT_FDCWD, filename, user, group,
666 AT_SYMLINK_NOFOLLOW);
669 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
671 struct fd f = fdget(fd);
677 error = mnt_want_write_file(f.file);
681 error = chown_common(&f.file->f_path, user, group);
682 mnt_drop_write_file(f.file);
689 int open_check_o_direct(struct file *f)
691 /* NB: we're sure to have correct a_ops only after f_op->open */
692 if (f->f_flags & O_DIRECT) {
693 if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO)
699 static int do_dentry_open(struct file *f,
701 int (*open)(struct inode *, struct file *),
702 const struct cred *cred)
704 static const struct file_operations empty_fops = {};
707 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
708 FMODE_PREAD | FMODE_PWRITE;
710 path_get(&f->f_path);
712 f->f_mapping = inode->i_mapping;
714 if (unlikely(f->f_flags & O_PATH)) {
715 f->f_mode = FMODE_PATH;
716 f->f_op = &empty_fops;
720 if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
721 error = get_write_access(inode);
724 error = __mnt_want_write(f->f_path.mnt);
725 if (unlikely(error)) {
726 put_write_access(inode);
729 f->f_mode |= FMODE_WRITER;
732 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
733 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
734 f->f_mode |= FMODE_ATOMIC_POS;
736 f->f_op = fops_get(inode->i_fop);
737 if (unlikely(WARN_ON(!f->f_op))) {
742 error = security_file_open(f, cred);
746 error = break_lease(locks_inode(f), f->f_flags);
751 open = f->f_op->open;
753 error = open(inode, f);
757 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
758 i_readcount_inc(inode);
759 if ((f->f_mode & FMODE_READ) &&
760 likely(f->f_op->read || f->f_op->read_iter))
761 f->f_mode |= FMODE_CAN_READ;
762 if ((f->f_mode & FMODE_WRITE) &&
763 likely(f->f_op->write || f->f_op->write_iter))
764 f->f_mode |= FMODE_CAN_WRITE;
766 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
768 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
774 if (f->f_mode & FMODE_WRITER) {
775 put_write_access(inode);
776 __mnt_drop_write(f->f_path.mnt);
779 path_put(&f->f_path);
780 f->f_path.mnt = NULL;
781 f->f_path.dentry = NULL;
787 * finish_open - finish opening a file
788 * @file: file pointer
789 * @dentry: pointer to dentry
790 * @open: open callback
791 * @opened: state of open
793 * This can be used to finish opening a file passed to i_op->atomic_open().
795 * If the open callback is set to NULL, then the standard f_op->open()
796 * filesystem callback is substituted.
798 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
799 * the return value of d_splice_alias(), then the caller needs to perform dput()
800 * on it after finish_open().
802 * On successful return @file is a fully instantiated open file. After this, if
803 * an error occurs in ->atomic_open(), it needs to clean up with fput().
805 * Returns zero on success or -errno if the open failed.
807 int finish_open(struct file *file, struct dentry *dentry,
808 int (*open)(struct inode *, struct file *),
812 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
814 file->f_path.dentry = dentry;
815 error = do_dentry_open(file, d_backing_inode(dentry), open,
818 *opened |= FILE_OPENED;
822 EXPORT_SYMBOL(finish_open);
825 * finish_no_open - finish ->atomic_open() without opening the file
827 * @file: file pointer
828 * @dentry: dentry or NULL (as returned from ->lookup())
830 * This can be used to set the result of a successful lookup in ->atomic_open().
832 * NB: unlike finish_open() this function does consume the dentry reference and
833 * the caller need not dput() it.
835 * Returns "1" which must be the return value of ->atomic_open() after having
836 * called this function.
838 int finish_no_open(struct file *file, struct dentry *dentry)
840 file->f_path.dentry = dentry;
843 EXPORT_SYMBOL(finish_no_open);
845 char *file_path(struct file *filp, char *buf, int buflen)
847 return d_path(&filp->f_path, buf, buflen);
849 EXPORT_SYMBOL(file_path);
852 * vfs_open - open the file at the given path
853 * @path: path to open
854 * @file: newly allocated file with f_flag initialized
855 * @cred: credentials to use
857 int vfs_open(const struct path *path, struct file *file,
858 const struct cred *cred)
860 struct dentry *dentry = d_real(path->dentry, NULL, file->f_flags);
863 return PTR_ERR(dentry);
865 file->f_path = *path;
866 return do_dentry_open(file, d_backing_inode(dentry), NULL, cred);
869 struct file *dentry_open(const struct path *path, int flags,
870 const struct cred *cred)
875 validate_creds(cred);
877 /* We must always pass in a valid mount pointer. */
880 f = get_empty_filp();
883 error = vfs_open(path, f, cred);
885 /* from now on we need fput() to dispose of f */
886 error = open_check_o_direct(f);
898 EXPORT_SYMBOL(dentry_open);
900 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
902 int lookup_flags = 0;
903 int acc_mode = ACC_MODE(flags);
905 if (flags & (O_CREAT | __O_TMPFILE))
906 op->mode = (mode & S_IALLUGO) | S_IFREG;
910 /* Must never be set by userspace */
911 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
914 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
915 * check for O_DSYNC if the need any syncing at all we enforce it's
916 * always set instead of having to deal with possibly weird behaviour
917 * for malicious applications setting only __O_SYNC.
919 if (flags & __O_SYNC)
922 if (flags & __O_TMPFILE) {
923 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
925 if (!(acc_mode & MAY_WRITE))
927 } else if (flags & O_PATH) {
929 * If we have O_PATH in the open flag. Then we
930 * cannot have anything other than the below set of flags
932 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
936 op->open_flag = flags;
938 /* O_TRUNC implies we need access checks for write permissions */
940 acc_mode |= MAY_WRITE;
942 /* Allow the LSM permission hook to distinguish append
943 access from general write access. */
944 if (flags & O_APPEND)
945 acc_mode |= MAY_APPEND;
947 op->acc_mode = acc_mode;
949 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
951 if (flags & O_CREAT) {
952 op->intent |= LOOKUP_CREATE;
954 op->intent |= LOOKUP_EXCL;
957 if (flags & O_DIRECTORY)
958 lookup_flags |= LOOKUP_DIRECTORY;
959 if (!(flags & O_NOFOLLOW))
960 lookup_flags |= LOOKUP_FOLLOW;
961 op->lookup_flags = lookup_flags;
966 * file_open_name - open file and return file pointer
968 * @name: struct filename containing path to open
969 * @flags: open flags as per the open(2) second argument
970 * @mode: mode for the new file if O_CREAT is set, else ignored
972 * This is the helper to open a file from kernelspace if you really
973 * have to. But in generally you should not do this, so please move
974 * along, nothing to see here..
976 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
978 struct open_flags op;
979 int err = build_open_flags(flags, mode, &op);
980 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
984 * filp_open - open file and return file pointer
986 * @filename: path to open
987 * @flags: open flags as per the open(2) second argument
988 * @mode: mode for the new file if O_CREAT is set, else ignored
990 * This is the helper to open a file from kernelspace if you really
991 * have to. But in generally you should not do this, so please move
992 * along, nothing to see here..
994 struct file *filp_open(const char *filename, int flags, umode_t mode)
996 struct filename *name = getname_kernel(filename);
997 struct file *file = ERR_CAST(name);
1000 file = file_open_name(name, flags, mode);
1005 EXPORT_SYMBOL(filp_open);
1007 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
1008 const char *filename, int flags, umode_t mode)
1010 struct open_flags op;
1011 int err = build_open_flags(flags, mode, &op);
1013 return ERR_PTR(err);
1014 return do_file_open_root(dentry, mnt, filename, &op);
1016 EXPORT_SYMBOL(file_open_root);
1018 struct file *filp_clone_open(struct file *oldfile)
1023 file = get_empty_filp();
1027 file->f_flags = oldfile->f_flags;
1028 retval = vfs_open(&oldfile->f_path, file, oldfile->f_cred);
1031 return ERR_PTR(retval);
1036 EXPORT_SYMBOL(filp_clone_open);
1038 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1040 struct open_flags op;
1041 int fd = build_open_flags(flags, mode, &op);
1042 struct filename *tmp;
1047 tmp = getname(filename);
1049 return PTR_ERR(tmp);
1051 fd = get_unused_fd_flags(flags);
1053 struct file *f = do_filp_open(dfd, tmp, &op);
1066 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1068 if (force_o_largefile())
1069 flags |= O_LARGEFILE;
1071 return do_sys_open(AT_FDCWD, filename, flags, mode);
1074 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1077 if (force_o_largefile())
1078 flags |= O_LARGEFILE;
1080 return do_sys_open(dfd, filename, flags, mode);
1086 * For backward compatibility? Maybe this should be moved
1087 * into arch/i386 instead?
1089 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1091 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1097 * "id" is the POSIX thread ID. We use the
1098 * files pointer for this..
1100 int filp_close(struct file *filp, fl_owner_t id)
1104 if (!file_count(filp)) {
1105 printk(KERN_ERR "VFS: Close: file count is 0\n");
1109 if (filp->f_op->flush)
1110 retval = filp->f_op->flush(filp, id);
1112 if (likely(!(filp->f_mode & FMODE_PATH))) {
1113 dnotify_flush(filp, id);
1114 locks_remove_posix(filp, id);
1120 EXPORT_SYMBOL(filp_close);
1123 * Careful here! We test whether the file pointer is NULL before
1124 * releasing the fd. This ensures that one clone task can't release
1125 * an fd while another clone is opening it.
1127 SYSCALL_DEFINE1(close, unsigned int, fd)
1129 int retval = __close_fd(current->files, fd);
1131 /* can't restart close syscall because file table entry was cleared */
1132 if (unlikely(retval == -ERESTARTSYS ||
1133 retval == -ERESTARTNOINTR ||
1134 retval == -ERESTARTNOHAND ||
1135 retval == -ERESTART_RESTARTBLOCK))
1140 EXPORT_SYMBOL(sys_close);
1143 * This routine simulates a hangup on the tty, to arrange that users
1144 * are given clean terminals at login time.
1146 SYSCALL_DEFINE0(vhangup)
1148 if (capable(CAP_SYS_TTY_CONFIG)) {
1156 * Called when an inode is about to be open.
1157 * We use this to disallow opening large files on 32bit systems if
1158 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1159 * on this flag in sys_open.
1161 int generic_file_open(struct inode * inode, struct file * filp)
1163 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1168 EXPORT_SYMBOL(generic_file_open);
1171 * This is used by subsystems that don't want seekable
1172 * file descriptors. The function is not supposed to ever fail, the only
1173 * reason it returns an 'int' and not 'void' is so that it can be plugged
1174 * directly into file_operations structure.
1176 int nonseekable_open(struct inode *inode, struct file *filp)
1178 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1182 EXPORT_SYMBOL(nonseekable_open);