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 on truncate too */
55 ret = should_remove_suid(dentry);
57 newattrs.ia_valid |= ret | ATTR_FORCE;
59 mutex_lock(&dentry->d_inode->i_mutex);
60 /* Note any delegations or leases have already been broken: */
61 ret = notify_change(dentry, &newattrs, NULL);
62 mutex_unlock(&dentry->d_inode->i_mutex);
66 long vfs_truncate(struct path *path, loff_t length)
71 inode = path->dentry->d_inode;
73 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
74 if (S_ISDIR(inode->i_mode))
76 if (!S_ISREG(inode->i_mode))
79 error = mnt_want_write(path->mnt);
83 error = inode_permission(inode, MAY_WRITE);
85 goto mnt_drop_write_and_out;
89 goto mnt_drop_write_and_out;
91 error = get_write_access(inode);
93 goto mnt_drop_write_and_out;
96 * Make sure that there are no leases. get_write_access() protects
97 * against the truncate racing with a lease-granting setlease().
99 error = break_lease(inode, O_WRONLY);
101 goto put_write_and_out;
103 error = locks_verify_truncate(inode, NULL, length);
105 error = security_path_truncate(path);
107 error = do_truncate(path->dentry, length, 0, NULL);
110 put_write_access(inode);
111 mnt_drop_write_and_out:
112 mnt_drop_write(path->mnt);
116 EXPORT_SYMBOL_GPL(vfs_truncate);
118 static long do_sys_truncate(const char __user *pathname, loff_t length)
120 unsigned int lookup_flags = LOOKUP_FOLLOW;
124 if (length < 0) /* sorry, but loff_t says... */
128 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
130 error = vfs_truncate(&path, length);
133 if (retry_estale(error, lookup_flags)) {
134 lookup_flags |= LOOKUP_REVAL;
140 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
142 return do_sys_truncate(path, length);
146 COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
148 return do_sys_truncate(path, length);
152 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
155 struct dentry *dentry;
167 /* explicitly opened as large or we are on 64-bit box */
168 if (f.file->f_flags & O_LARGEFILE)
171 dentry = f.file->f_path.dentry;
172 inode = dentry->d_inode;
174 if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
178 /* Cannot ftruncate over 2^31 bytes without large file support */
179 if (small && length > MAX_NON_LFS)
183 if (IS_APPEND(inode))
186 sb_start_write(inode->i_sb);
187 error = locks_verify_truncate(inode, f.file, length);
189 error = security_path_truncate(&f.file->f_path);
191 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
192 sb_end_write(inode->i_sb);
199 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
201 return do_sys_ftruncate(fd, length, 1);
205 COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
207 return do_sys_ftruncate(fd, length, 1);
211 /* LFS versions of truncate are only needed on 32 bit machines */
212 #if BITS_PER_LONG == 32
213 SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
215 return do_sys_truncate(path, length);
218 SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
220 return do_sys_ftruncate(fd, length, 0);
222 #endif /* BITS_PER_LONG == 32 */
225 int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
227 struct inode *inode = file_inode(file);
230 if (offset < 0 || len <= 0)
233 /* Return error if mode is not supported */
234 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
235 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE))
238 /* Punch hole and zero range are mutually exclusive */
239 if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
240 (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
243 /* Punch hole must have keep size set */
244 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
245 !(mode & FALLOC_FL_KEEP_SIZE))
248 /* Collapse range should only be used exclusively. */
249 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
250 (mode & ~FALLOC_FL_COLLAPSE_RANGE))
253 if (!(file->f_mode & FMODE_WRITE))
257 * It's not possible to punch hole or perform collapse range
258 * on append only file
260 if (mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE)
264 if (IS_IMMUTABLE(inode))
268 * Revalidate the write permissions, in case security policy has
269 * changed since the files were opened.
271 ret = security_file_permission(file, MAY_WRITE);
275 if (S_ISFIFO(inode->i_mode))
279 * Let individual file system decide if it supports preallocation
280 * for directories or not.
282 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
285 /* Check for wrap through zero too */
286 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
290 * There is no need to overlap collapse range with EOF, in which case
291 * it is effectively a truncate operation
293 if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
294 (offset + len >= i_size_read(inode)))
297 if (!file->f_op->fallocate)
300 sb_start_write(inode->i_sb);
301 ret = file->f_op->fallocate(file, mode, offset, len);
302 sb_end_write(inode->i_sb);
306 SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
308 struct fd f = fdget(fd);
312 error = do_fallocate(f.file, mode, offset, len);
319 * access() needs to use the real uid/gid, not the effective uid/gid.
320 * We do this by temporarily clearing all FS-related capabilities and
321 * switching the fsuid/fsgid around to the real ones.
323 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
325 const struct cred *old_cred;
326 struct cred *override_cred;
330 unsigned int lookup_flags = LOOKUP_FOLLOW;
332 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
335 override_cred = prepare_creds();
339 override_cred->fsuid = override_cred->uid;
340 override_cred->fsgid = override_cred->gid;
342 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
343 /* Clear the capabilities if we switch to a non-root user */
344 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
345 if (!uid_eq(override_cred->uid, root_uid))
346 cap_clear(override_cred->cap_effective);
348 override_cred->cap_effective =
349 override_cred->cap_permitted;
352 old_cred = override_creds(override_cred);
354 res = user_path_at(dfd, filename, lookup_flags, &path);
358 inode = path.dentry->d_inode;
360 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
362 * MAY_EXEC on regular files is denied if the fs is mounted
363 * with the "noexec" flag.
366 if (path.mnt->mnt_flags & MNT_NOEXEC)
367 goto out_path_release;
370 res = inode_permission(inode, mode | MAY_ACCESS);
371 /* SuS v2 requires we report a read only fs too */
372 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
373 goto out_path_release;
375 * This is a rare case where using __mnt_is_readonly()
376 * is OK without a mnt_want/drop_write() pair. Since
377 * no actual write to the fs is performed here, we do
378 * not need to telegraph to that to anyone.
380 * By doing this, we accept that this access is
381 * inherently racy and know that the fs may change
382 * state before we even see this result.
384 if (__mnt_is_readonly(path.mnt))
389 if (retry_estale(res, lookup_flags)) {
390 lookup_flags |= LOOKUP_REVAL;
394 revert_creds(old_cred);
395 put_cred(override_cred);
399 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
401 return sys_faccessat(AT_FDCWD, filename, mode);
404 SYSCALL_DEFINE1(chdir, const char __user *, filename)
408 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
410 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
414 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
418 set_fs_pwd(current->fs, &path);
422 if (retry_estale(error, lookup_flags)) {
423 lookup_flags |= LOOKUP_REVAL;
430 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
432 struct fd f = fdget_raw(fd);
440 inode = file_inode(f.file);
443 if (!S_ISDIR(inode->i_mode))
446 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
448 set_fs_pwd(current->fs, &f.file->f_path);
455 SYSCALL_DEFINE1(chroot, const char __user *, filename)
459 unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
461 error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
465 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
470 if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
472 error = security_path_chroot(&path);
476 set_fs_root(current->fs, &path);
480 if (retry_estale(error, lookup_flags)) {
481 lookup_flags |= LOOKUP_REVAL;
488 static int chmod_common(struct path *path, umode_t mode)
490 struct inode *inode = path->dentry->d_inode;
491 struct inode *delegated_inode = NULL;
492 struct iattr newattrs;
495 error = mnt_want_write(path->mnt);
499 mutex_lock(&inode->i_mutex);
500 error = security_path_chmod(path, mode);
503 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
504 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
505 error = notify_change(path->dentry, &newattrs, &delegated_inode);
507 mutex_unlock(&inode->i_mutex);
508 if (delegated_inode) {
509 error = break_deleg_wait(&delegated_inode);
513 mnt_drop_write(path->mnt);
517 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
519 struct fd f = fdget(fd);
523 audit_inode(NULL, f.file->f_path.dentry, 0);
524 err = chmod_common(&f.file->f_path, mode);
530 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
534 unsigned int lookup_flags = LOOKUP_FOLLOW;
536 error = user_path_at(dfd, filename, lookup_flags, &path);
538 error = chmod_common(&path, mode);
540 if (retry_estale(error, lookup_flags)) {
541 lookup_flags |= LOOKUP_REVAL;
548 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
550 return sys_fchmodat(AT_FDCWD, filename, mode);
553 static int chown_common(struct path *path, uid_t user, gid_t group)
555 struct inode *inode = path->dentry->d_inode;
556 struct inode *delegated_inode = NULL;
558 struct iattr newattrs;
562 uid = make_kuid(current_user_ns(), user);
563 gid = make_kgid(current_user_ns(), group);
565 newattrs.ia_valid = ATTR_CTIME;
566 if (user != (uid_t) -1) {
569 newattrs.ia_valid |= ATTR_UID;
570 newattrs.ia_uid = uid;
572 if (group != (gid_t) -1) {
575 newattrs.ia_valid |= ATTR_GID;
576 newattrs.ia_gid = gid;
578 if (!S_ISDIR(inode->i_mode))
580 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
582 mutex_lock(&inode->i_mutex);
583 error = security_path_chown(path, uid, gid);
585 error = notify_change(path->dentry, &newattrs, &delegated_inode);
586 mutex_unlock(&inode->i_mutex);
587 if (delegated_inode) {
588 error = break_deleg_wait(&delegated_inode);
595 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
596 gid_t, group, int, flag)
602 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
605 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
606 if (flag & AT_EMPTY_PATH)
607 lookup_flags |= LOOKUP_EMPTY;
609 error = user_path_at(dfd, filename, lookup_flags, &path);
612 error = mnt_want_write(path.mnt);
615 error = chown_common(&path, user, group);
616 mnt_drop_write(path.mnt);
619 if (retry_estale(error, lookup_flags)) {
620 lookup_flags |= LOOKUP_REVAL;
627 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
629 return sys_fchownat(AT_FDCWD, filename, user, group, 0);
632 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
634 return sys_fchownat(AT_FDCWD, filename, user, group,
635 AT_SYMLINK_NOFOLLOW);
638 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
640 struct fd f = fdget(fd);
646 error = mnt_want_write_file(f.file);
649 audit_inode(NULL, f.file->f_path.dentry, 0);
650 error = chown_common(&f.file->f_path, user, group);
651 mnt_drop_write_file(f.file);
659 * You have to be very careful that these write
660 * counts get cleaned up in error cases and
661 * upon __fput(). This should probably never
662 * be called outside of __dentry_open().
664 static inline int __get_file_write_access(struct inode *inode,
665 struct vfsmount *mnt)
668 error = get_write_access(inode);
672 * Do not take mount writer counts on
673 * special files since no writes to
674 * the mount itself will occur.
676 if (!special_file(inode->i_mode)) {
678 * Balanced in __fput()
680 error = __mnt_want_write(mnt);
682 put_write_access(inode);
687 int open_check_o_direct(struct file *f)
689 /* NB: we're sure to have correct a_ops only after f_op->open */
690 if (f->f_flags & O_DIRECT) {
691 if (!f->f_mapping->a_ops ||
692 ((!f->f_mapping->a_ops->direct_IO) &&
693 (!f->f_mapping->a_ops->get_xip_mem))) {
700 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 = {};
708 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
709 FMODE_PREAD | FMODE_PWRITE;
711 if (unlikely(f->f_flags & O_PATH))
712 f->f_mode = FMODE_PATH;
714 path_get(&f->f_path);
715 inode = f->f_inode = f->f_path.dentry->d_inode;
716 if (f->f_mode & FMODE_WRITE) {
717 error = __get_file_write_access(inode, f->f_path.mnt);
720 if (!special_file(inode->i_mode))
724 f->f_mapping = inode->i_mapping;
726 if (unlikely(f->f_mode & FMODE_PATH)) {
727 f->f_op = &empty_fops;
731 /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
732 if (S_ISREG(inode->i_mode))
733 f->f_mode |= FMODE_ATOMIC_POS;
735 f->f_op = fops_get(inode->i_fop);
736 if (unlikely(WARN_ON(!f->f_op))) {
741 error = security_file_open(f, cred);
745 error = break_lease(inode, f->f_flags);
750 open = f->f_op->open;
752 error = open(inode, f);
756 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
757 i_readcount_inc(inode);
759 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
761 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
767 if (f->f_mode & FMODE_WRITE) {
768 put_write_access(inode);
769 if (!special_file(inode->i_mode)) {
771 * We don't consider this a real
772 * mnt_want/drop_write() pair
773 * because it all happenend right
774 * here, so just reset the state.
777 __mnt_drop_write(f->f_path.mnt);
781 path_put(&f->f_path);
782 f->f_path.mnt = NULL;
783 f->f_path.dentry = NULL;
789 * finish_open - finish opening a file
790 * @file: file pointer
791 * @dentry: pointer to dentry
792 * @open: open callback
793 * @opened: state of open
795 * This can be used to finish opening a file passed to i_op->atomic_open().
797 * If the open callback is set to NULL, then the standard f_op->open()
798 * filesystem callback is substituted.
800 * NB: the dentry reference is _not_ consumed. If, for example, the dentry is
801 * the return value of d_splice_alias(), then the caller needs to perform dput()
802 * on it after finish_open().
804 * On successful return @file is a fully instantiated open file. After this, if
805 * an error occurs in ->atomic_open(), it needs to clean up with fput().
807 * Returns zero on success or -errno if the open failed.
809 int finish_open(struct file *file, struct dentry *dentry,
810 int (*open)(struct inode *, struct file *),
814 BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
816 file->f_path.dentry = dentry;
817 error = do_dentry_open(file, open, current_cred());
819 *opened |= FILE_OPENED;
823 EXPORT_SYMBOL(finish_open);
826 * finish_no_open - finish ->atomic_open() without opening the file
828 * @file: file pointer
829 * @dentry: dentry or NULL (as returned from ->lookup())
831 * This can be used to set the result of a successful lookup in ->atomic_open().
833 * NB: unlike finish_open() this function does consume the dentry reference and
834 * the caller need not dput() it.
836 * Returns "1" which must be the return value of ->atomic_open() after having
837 * called this function.
839 int finish_no_open(struct file *file, struct dentry *dentry)
841 file->f_path.dentry = dentry;
844 EXPORT_SYMBOL(finish_no_open);
846 struct file *dentry_open(const struct path *path, int flags,
847 const struct cred *cred)
852 validate_creds(cred);
854 /* We must always pass in a valid mount pointer. */
857 f = get_empty_filp();
861 error = do_dentry_open(f, NULL, cred);
863 /* from now on we need fput() to dispose of f */
864 error = open_check_o_direct(f);
876 EXPORT_SYMBOL(dentry_open);
878 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
880 int lookup_flags = 0;
883 if (flags & (O_CREAT | __O_TMPFILE))
884 op->mode = (mode & S_IALLUGO) | S_IFREG;
888 /* Must never be set by userspace */
889 flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
892 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
893 * check for O_DSYNC if the need any syncing at all we enforce it's
894 * always set instead of having to deal with possibly weird behaviour
895 * for malicious applications setting only __O_SYNC.
897 if (flags & __O_SYNC)
900 if (flags & __O_TMPFILE) {
901 if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
903 acc_mode = MAY_OPEN | ACC_MODE(flags);
904 if (!(acc_mode & MAY_WRITE))
906 } else if (flags & O_PATH) {
908 * If we have O_PATH in the open flag. Then we
909 * cannot have anything other than the below set of flags
911 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
914 acc_mode = MAY_OPEN | ACC_MODE(flags);
917 op->open_flag = flags;
919 /* O_TRUNC implies we need access checks for write permissions */
921 acc_mode |= MAY_WRITE;
923 /* Allow the LSM permission hook to distinguish append
924 access from general write access. */
925 if (flags & O_APPEND)
926 acc_mode |= MAY_APPEND;
928 op->acc_mode = acc_mode;
930 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
932 if (flags & O_CREAT) {
933 op->intent |= LOOKUP_CREATE;
935 op->intent |= LOOKUP_EXCL;
938 if (flags & O_DIRECTORY)
939 lookup_flags |= LOOKUP_DIRECTORY;
940 if (!(flags & O_NOFOLLOW))
941 lookup_flags |= LOOKUP_FOLLOW;
942 op->lookup_flags = lookup_flags;
947 * file_open_name - open file and return file pointer
949 * @name: struct filename containing path to open
950 * @flags: open flags as per the open(2) second argument
951 * @mode: mode for the new file if O_CREAT is set, else ignored
953 * This is the helper to open a file from kernelspace if you really
954 * have to. But in generally you should not do this, so please move
955 * along, nothing to see here..
957 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
959 struct open_flags op;
960 int err = build_open_flags(flags, mode, &op);
961 return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
965 * filp_open - open file and return file pointer
967 * @filename: path to open
968 * @flags: open flags as per the open(2) second argument
969 * @mode: mode for the new file if O_CREAT is set, else ignored
971 * This is the helper to open a file from kernelspace if you really
972 * have to. But in generally you should not do this, so please move
973 * along, nothing to see here..
975 struct file *filp_open(const char *filename, int flags, umode_t mode)
977 struct filename name = {.name = filename};
978 return file_open_name(&name, flags, mode);
980 EXPORT_SYMBOL(filp_open);
982 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
983 const char *filename, int flags)
985 struct open_flags op;
986 int err = build_open_flags(flags, 0, &op);
990 return ERR_PTR(-EINVAL);
991 if (!filename && (flags & O_DIRECTORY))
992 if (!dentry->d_inode->i_op->lookup)
993 return ERR_PTR(-ENOTDIR);
994 return do_file_open_root(dentry, mnt, filename, &op);
996 EXPORT_SYMBOL(file_open_root);
998 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
1000 struct open_flags op;
1001 int fd = build_open_flags(flags, mode, &op);
1002 struct filename *tmp;
1007 tmp = getname(filename);
1009 return PTR_ERR(tmp);
1011 fd = get_unused_fd_flags(flags);
1013 struct file *f = do_filp_open(dfd, tmp, &op);
1026 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1028 if (force_o_largefile())
1029 flags |= O_LARGEFILE;
1031 return do_sys_open(AT_FDCWD, filename, flags, mode);
1034 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1037 if (force_o_largefile())
1038 flags |= O_LARGEFILE;
1040 return do_sys_open(dfd, filename, flags, mode);
1046 * For backward compatibility? Maybe this should be moved
1047 * into arch/i386 instead?
1049 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1051 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1057 * "id" is the POSIX thread ID. We use the
1058 * files pointer for this..
1060 int filp_close(struct file *filp, fl_owner_t id)
1064 if (!file_count(filp)) {
1065 printk(KERN_ERR "VFS: Close: file count is 0\n");
1069 if (filp->f_op->flush)
1070 retval = filp->f_op->flush(filp, id);
1072 if (likely(!(filp->f_mode & FMODE_PATH))) {
1073 dnotify_flush(filp, id);
1074 locks_remove_posix(filp, id);
1080 EXPORT_SYMBOL(filp_close);
1083 * Careful here! We test whether the file pointer is NULL before
1084 * releasing the fd. This ensures that one clone task can't release
1085 * an fd while another clone is opening it.
1087 SYSCALL_DEFINE1(close, unsigned int, fd)
1089 int retval = __close_fd(current->files, fd);
1091 /* can't restart close syscall because file table entry was cleared */
1092 if (unlikely(retval == -ERESTARTSYS ||
1093 retval == -ERESTARTNOINTR ||
1094 retval == -ERESTARTNOHAND ||
1095 retval == -ERESTART_RESTARTBLOCK))
1100 EXPORT_SYMBOL(sys_close);
1103 * This routine simulates a hangup on the tty, to arrange that users
1104 * are given clean terminals at login time.
1106 SYSCALL_DEFINE0(vhangup)
1108 if (capable(CAP_SYS_TTY_CONFIG)) {
1116 * Called when an inode is about to be open.
1117 * We use this to disallow opening large files on 32bit systems if
1118 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1119 * on this flag in sys_open.
1121 int generic_file_open(struct inode * inode, struct file * filp)
1123 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1128 EXPORT_SYMBOL(generic_file_open);
1131 * This is used by subsystems that don't want seekable
1132 * file descriptors. The function is not supposed to ever fail, the only
1133 * reason it returns an 'int' and not 'void' is so that it can be plugged
1134 * directly into file_operations structure.
1136 int nonseekable_open(struct inode *inode, struct file *filp)
1138 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1142 EXPORT_SYMBOL(nonseekable_open);