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>
36 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
40 struct iattr newattrs;
42 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
46 newattrs.ia_size = length;
47 newattrs.ia_valid = ATTR_SIZE | time_attrs;
49 newattrs.ia_file = filp;
50 newattrs.ia_valid |= ATTR_FILE;
53 /* Remove suid/sgid on truncate too */
54 ret = should_remove_suid(dentry);
56 newattrs.ia_valid |= ret | ATTR_FORCE;
58 mutex_lock(&dentry->d_inode->i_mutex);
59 ret = notify_change(dentry, &newattrs);
60 mutex_unlock(&dentry->d_inode->i_mutex);
64 static long do_sys_truncate(const char __user *pathname, loff_t length)
71 if (length < 0) /* sorry, but loff_t says... */
74 error = user_path(pathname, &path);
77 inode = path.dentry->d_inode;
79 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
81 if (S_ISDIR(inode->i_mode))
85 if (!S_ISREG(inode->i_mode))
88 error = mnt_want_write(path.mnt);
92 error = inode_permission(inode, MAY_WRITE);
94 goto mnt_drop_write_and_out;
98 goto mnt_drop_write_and_out;
100 error = get_write_access(inode);
102 goto mnt_drop_write_and_out;
105 * Make sure that there are no leases. get_write_access() protects
106 * against the truncate racing with a lease-granting setlease().
108 error = break_lease(inode, O_WRONLY);
110 goto put_write_and_out;
112 error = locks_verify_truncate(inode, NULL, length);
114 error = security_path_truncate(&path);
116 error = do_truncate(path.dentry, length, 0, NULL);
119 put_write_access(inode);
120 mnt_drop_write_and_out:
121 mnt_drop_write(path.mnt);
128 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
130 return do_sys_truncate(path, length);
133 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
135 struct inode * inode;
136 struct dentry *dentry;
148 /* explicitly opened as large or we are on 64-bit box */
149 if (file->f_flags & O_LARGEFILE)
152 dentry = file->f_path.dentry;
153 inode = dentry->d_inode;
155 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
159 /* Cannot ftruncate over 2^31 bytes without large file support */
160 if (small && length > MAX_NON_LFS)
164 if (IS_APPEND(inode))
167 error = locks_verify_truncate(inode, file, length);
169 error = security_path_truncate(&file->f_path);
171 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
178 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
180 long ret = do_sys_ftruncate(fd, length, 1);
181 /* avoid REGPARM breakage on x86: */
182 asmlinkage_protect(2, ret, fd, length);
186 /* LFS versions of truncate are only needed on 32 bit machines */
187 #if BITS_PER_LONG == 32
188 SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
190 return do_sys_truncate(path, length);
192 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
193 asmlinkage long SyS_truncate64(long path, loff_t length)
195 return SYSC_truncate64((const char __user *) path, length);
197 SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
200 SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
202 long ret = do_sys_ftruncate(fd, length, 0);
203 /* avoid REGPARM breakage on x86: */
204 asmlinkage_protect(2, ret, fd, length);
207 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
208 asmlinkage long SyS_ftruncate64(long fd, loff_t length)
210 return SYSC_ftruncate64((unsigned int) fd, length);
212 SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
214 #endif /* BITS_PER_LONG == 32 */
217 int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
219 struct inode *inode = file->f_path.dentry->d_inode;
222 if (offset < 0 || len <= 0)
225 /* Return error if mode is not supported */
226 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
229 /* Punch hole must have keep size set */
230 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
231 !(mode & FALLOC_FL_KEEP_SIZE))
234 if (!(file->f_mode & FMODE_WRITE))
237 /* It's not possible punch hole on append only file */
238 if (mode & FALLOC_FL_PUNCH_HOLE && IS_APPEND(inode))
241 if (IS_IMMUTABLE(inode))
245 * Revalidate the write permissions, in case security policy has
246 * changed since the files were opened.
248 ret = security_file_permission(file, MAY_WRITE);
252 if (S_ISFIFO(inode->i_mode))
256 * Let individual file system decide if it supports preallocation
257 * for directories or not.
259 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
262 /* Check for wrap through zero too */
263 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
266 if (!file->f_op->fallocate)
269 return file->f_op->fallocate(file, mode, offset, len);
272 SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
279 error = do_fallocate(file, mode, offset, len);
286 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
287 asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
289 return SYSC_fallocate((int)fd, (int)mode, offset, len);
291 SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
295 * access() needs to use the real uid/gid, not the effective uid/gid.
296 * We do this by temporarily clearing all FS-related capabilities and
297 * switching the fsuid/fsgid around to the real ones.
299 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
301 const struct cred *old_cred;
302 struct cred *override_cred;
307 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
310 override_cred = prepare_creds();
314 override_cred->fsuid = override_cred->uid;
315 override_cred->fsgid = override_cred->gid;
317 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
318 /* Clear the capabilities if we switch to a non-root user */
319 kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
320 if (!uid_eq(override_cred->uid, root_uid))
321 cap_clear(override_cred->cap_effective);
323 override_cred->cap_effective =
324 override_cred->cap_permitted;
327 old_cred = override_creds(override_cred);
329 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
333 inode = path.dentry->d_inode;
335 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
337 * MAY_EXEC on regular files is denied if the fs is mounted
338 * with the "noexec" flag.
341 if (path.mnt->mnt_flags & MNT_NOEXEC)
342 goto out_path_release;
345 res = inode_permission(inode, mode | MAY_ACCESS);
346 /* SuS v2 requires we report a read only fs too */
347 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
348 goto out_path_release;
350 * This is a rare case where using __mnt_is_readonly()
351 * is OK without a mnt_want/drop_write() pair. Since
352 * no actual write to the fs is performed here, we do
353 * not need to telegraph to that to anyone.
355 * By doing this, we accept that this access is
356 * inherently racy and know that the fs may change
357 * state before we even see this result.
359 if (__mnt_is_readonly(path.mnt))
365 revert_creds(old_cred);
366 put_cred(override_cred);
370 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
372 return sys_faccessat(AT_FDCWD, filename, mode);
375 SYSCALL_DEFINE1(chdir, const char __user *, filename)
380 error = user_path_dir(filename, &path);
384 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
388 set_fs_pwd(current->fs, &path);
396 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
407 inode = file->f_path.dentry->d_inode;
410 if (!S_ISDIR(inode->i_mode))
413 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
415 set_fs_pwd(current->fs, &file->f_path);
422 SYSCALL_DEFINE1(chroot, const char __user *, filename)
427 error = user_path_dir(filename, &path);
431 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
436 if (!capable(CAP_SYS_CHROOT))
438 error = security_path_chroot(&path);
442 set_fs_root(current->fs, &path);
450 static int chmod_common(struct path *path, umode_t mode)
452 struct inode *inode = path->dentry->d_inode;
453 struct iattr newattrs;
456 error = mnt_want_write(path->mnt);
459 mutex_lock(&inode->i_mutex);
460 error = security_path_chmod(path, mode);
463 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
464 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
465 error = notify_change(path->dentry, &newattrs);
467 mutex_unlock(&inode->i_mutex);
468 mnt_drop_write(path->mnt);
472 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
479 audit_inode(NULL, file->f_path.dentry);
480 err = chmod_common(&file->f_path, mode);
486 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
491 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
493 error = chmod_common(&path, mode);
499 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
501 return sys_fchmodat(AT_FDCWD, filename, mode);
504 static int chown_common(struct path *path, uid_t user, gid_t group)
506 struct inode *inode = path->dentry->d_inode;
508 struct iattr newattrs;
512 uid = make_kuid(current_user_ns(), user);
513 gid = make_kgid(current_user_ns(), group);
515 newattrs.ia_valid = ATTR_CTIME;
516 if (user != (uid_t) -1) {
519 newattrs.ia_valid |= ATTR_UID;
520 newattrs.ia_uid = uid;
522 if (group != (gid_t) -1) {
525 newattrs.ia_valid |= ATTR_GID;
526 newattrs.ia_gid = gid;
528 if (!S_ISDIR(inode->i_mode))
530 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
531 mutex_lock(&inode->i_mutex);
532 error = security_path_chown(path, user, group);
534 error = notify_change(path->dentry, &newattrs);
535 mutex_unlock(&inode->i_mutex);
540 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
545 error = user_path(filename, &path);
548 error = mnt_want_write(path.mnt);
551 error = chown_common(&path, user, group);
552 mnt_drop_write(path.mnt);
559 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
560 gid_t, group, int, flag)
566 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
569 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
570 if (flag & AT_EMPTY_PATH)
571 lookup_flags |= LOOKUP_EMPTY;
572 error = user_path_at(dfd, filename, lookup_flags, &path);
575 error = mnt_want_write(path.mnt);
578 error = chown_common(&path, user, group);
579 mnt_drop_write(path.mnt);
586 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
591 error = user_lpath(filename, &path);
594 error = mnt_want_write(path.mnt);
597 error = chown_common(&path, user, group);
598 mnt_drop_write(path.mnt);
605 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
609 struct dentry * dentry;
615 error = mnt_want_write_file(file);
618 dentry = file->f_path.dentry;
619 audit_inode(NULL, dentry);
620 error = chown_common(&file->f_path, user, group);
621 mnt_drop_write_file(file);
629 * You have to be very careful that these write
630 * counts get cleaned up in error cases and
631 * upon __fput(). This should probably never
632 * be called outside of __dentry_open().
634 static inline int __get_file_write_access(struct inode *inode,
635 struct vfsmount *mnt)
638 error = get_write_access(inode);
642 * Do not take mount writer counts on
643 * special files since no writes to
644 * the mount itself will occur.
646 if (!special_file(inode->i_mode)) {
648 * Balanced in __fput()
650 error = mnt_want_write(mnt);
652 put_write_access(inode);
657 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
659 int (*open)(struct inode *, struct file *),
660 const struct cred *cred)
662 static const struct file_operations empty_fops = {};
666 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
667 FMODE_PREAD | FMODE_PWRITE;
669 if (unlikely(f->f_flags & O_PATH))
670 f->f_mode = FMODE_PATH;
672 inode = dentry->d_inode;
673 if (f->f_mode & FMODE_WRITE) {
674 error = __get_file_write_access(inode, mnt);
677 if (!special_file(inode->i_mode))
681 f->f_mapping = inode->i_mapping;
682 f->f_path.dentry = dentry;
685 file_sb_list_add(f, inode->i_sb);
687 if (unlikely(f->f_mode & FMODE_PATH)) {
688 f->f_op = &empty_fops;
692 f->f_op = fops_get(inode->i_fop);
694 error = security_file_open(f, cred);
698 error = break_lease(inode, f->f_flags);
702 if (!open && f->f_op)
703 open = f->f_op->open;
705 error = open(inode, f);
709 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
710 i_readcount_inc(inode);
712 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
714 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
716 /* NB: we're sure to have correct a_ops only after f_op->open */
717 if (f->f_flags & O_DIRECT) {
718 if (!f->f_mapping->a_ops ||
719 ((!f->f_mapping->a_ops->direct_IO) &&
720 (!f->f_mapping->a_ops->get_xip_mem))) {
722 f = ERR_PTR(-EINVAL);
730 if (f->f_mode & FMODE_WRITE) {
731 put_write_access(inode);
732 if (!special_file(inode->i_mode)) {
734 * We don't consider this a real
735 * mnt_want/drop_write() pair
736 * because it all happenend right
737 * here, so just reset the state.
744 f->f_path.dentry = NULL;
745 f->f_path.mnt = NULL;
750 return ERR_PTR(error);
754 * lookup_instantiate_filp - instantiates the open intent filp
755 * @nd: pointer to nameidata
756 * @dentry: pointer to dentry
757 * @open: open callback
759 * Helper for filesystems that want to use lookup open intents and pass back
760 * a fully instantiated struct file to the caller.
761 * This function is meant to be called from within a filesystem's
763 * Beware of calling it for non-regular files! Those ->open methods might block
764 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
765 * leading to a deadlock, as nobody can open that fifo anymore, because
766 * another process to open fifo will block on locked parent when doing lookup).
767 * Note that in case of error, nd->intent.open.file is destroyed, but the
768 * path information remains valid.
769 * If the open callback is set to NULL, then the standard f_op->open()
770 * filesystem callback is substituted.
772 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
773 int (*open)(struct inode *, struct file *))
775 const struct cred *cred = current_cred();
777 if (IS_ERR(nd->intent.open.file))
781 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
782 nd->intent.open.file,
785 return nd->intent.open.file;
787 release_open_intent(nd);
788 nd->intent.open.file = ERR_CAST(dentry);
791 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
794 * nameidata_to_filp - convert a nameidata to an open filp.
795 * @nd: pointer to nameidata
798 * Note that this function destroys the original nameidata
800 struct file *nameidata_to_filp(struct nameidata *nd)
802 const struct cred *cred = current_cred();
805 /* Pick up the filp from the open intent */
806 filp = nd->intent.open.file;
807 nd->intent.open.file = NULL;
809 /* Has the filesystem initialised the file for us? */
810 if (filp->f_path.dentry == NULL) {
812 filp = __dentry_open(nd->path.dentry, nd->path.mnt, filp,
819 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
822 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
823 const struct cred *cred)
828 validate_creds(cred);
830 /* We must always pass in a valid mount pointer. */
834 f = get_empty_filp();
838 return ERR_PTR(error);
842 return __dentry_open(dentry, mnt, f, NULL, cred);
844 EXPORT_SYMBOL(dentry_open);
846 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
848 struct fdtable *fdt = files_fdtable(files);
849 __clear_open_fd(fd, fdt);
850 if (fd < files->next_fd)
854 void put_unused_fd(unsigned int fd)
856 struct files_struct *files = current->files;
857 spin_lock(&files->file_lock);
858 __put_unused_fd(files, fd);
859 spin_unlock(&files->file_lock);
862 EXPORT_SYMBOL(put_unused_fd);
865 * Install a file pointer in the fd array.
867 * The VFS is full of places where we drop the files lock between
868 * setting the open_fds bitmap and installing the file in the file
869 * array. At any such point, we are vulnerable to a dup2() race
870 * installing a file in the array before us. We need to detect this and
871 * fput() the struct file we are about to overwrite in this case.
873 * It should never happen - if we allow dup2() do it, _really_ bad things
877 void fd_install(unsigned int fd, struct file *file)
879 struct files_struct *files = current->files;
881 spin_lock(&files->file_lock);
882 fdt = files_fdtable(files);
883 BUG_ON(fdt->fd[fd] != NULL);
884 rcu_assign_pointer(fdt->fd[fd], file);
885 spin_unlock(&files->file_lock);
888 EXPORT_SYMBOL(fd_install);
890 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
892 int lookup_flags = 0;
895 if (!(flags & O_CREAT))
899 /* Must never be set by userspace */
900 flags &= ~FMODE_NONOTIFY;
903 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
904 * check for O_DSYNC if the need any syncing at all we enforce it's
905 * always set instead of having to deal with possibly weird behaviour
906 * for malicious applications setting only __O_SYNC.
908 if (flags & __O_SYNC)
912 * If we have O_PATH in the open flag. Then we
913 * cannot have anything other than the below set of flags
915 if (flags & O_PATH) {
916 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
919 acc_mode = MAY_OPEN | ACC_MODE(flags);
922 op->open_flag = flags;
924 /* O_TRUNC implies we need access checks for write permissions */
926 acc_mode |= MAY_WRITE;
928 /* Allow the LSM permission hook to distinguish append
929 access from general write access. */
930 if (flags & O_APPEND)
931 acc_mode |= MAY_APPEND;
933 op->acc_mode = acc_mode;
935 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
937 if (flags & O_CREAT) {
938 op->intent |= LOOKUP_CREATE;
940 op->intent |= LOOKUP_EXCL;
943 if (flags & O_DIRECTORY)
944 lookup_flags |= LOOKUP_DIRECTORY;
945 if (!(flags & O_NOFOLLOW))
946 lookup_flags |= LOOKUP_FOLLOW;
951 * filp_open - open file and return file pointer
953 * @filename: path to open
954 * @flags: open flags as per the open(2) second argument
955 * @mode: mode for the new file if O_CREAT is set, else ignored
957 * This is the helper to open a file from kernelspace if you really
958 * have to. But in generally you should not do this, so please move
959 * along, nothing to see here..
961 struct file *filp_open(const char *filename, int flags, umode_t mode)
963 struct open_flags op;
964 int lookup = build_open_flags(flags, mode, &op);
965 return do_filp_open(AT_FDCWD, filename, &op, lookup);
967 EXPORT_SYMBOL(filp_open);
969 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
970 const char *filename, int flags)
972 struct open_flags op;
973 int lookup = build_open_flags(flags, 0, &op);
975 return ERR_PTR(-EINVAL);
976 if (!filename && (flags & O_DIRECTORY))
977 if (!dentry->d_inode->i_op->lookup)
978 return ERR_PTR(-ENOTDIR);
979 return do_file_open_root(dentry, mnt, filename, &op, lookup);
981 EXPORT_SYMBOL(file_open_root);
983 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
985 struct open_flags op;
986 int lookup = build_open_flags(flags, mode, &op);
987 char *tmp = getname(filename);
988 int fd = PTR_ERR(tmp);
991 fd = get_unused_fd_flags(flags);
993 struct file *f = do_filp_open(dfd, tmp, &op, lookup);
1007 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1011 if (force_o_largefile())
1012 flags |= O_LARGEFILE;
1014 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1015 /* avoid REGPARM breakage on x86: */
1016 asmlinkage_protect(3, ret, filename, flags, mode);
1020 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1025 if (force_o_largefile())
1026 flags |= O_LARGEFILE;
1028 ret = do_sys_open(dfd, filename, flags, mode);
1029 /* avoid REGPARM breakage on x86: */
1030 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
1037 * For backward compatibility? Maybe this should be moved
1038 * into arch/i386 instead?
1040 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1042 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1048 * "id" is the POSIX thread ID. We use the
1049 * files pointer for this..
1051 int filp_close(struct file *filp, fl_owner_t id)
1055 if (!file_count(filp)) {
1056 printk(KERN_ERR "VFS: Close: file count is 0\n");
1060 if (filp->f_op && filp->f_op->flush)
1061 retval = filp->f_op->flush(filp, id);
1063 if (likely(!(filp->f_mode & FMODE_PATH))) {
1064 dnotify_flush(filp, id);
1065 locks_remove_posix(filp, id);
1071 EXPORT_SYMBOL(filp_close);
1074 * Careful here! We test whether the file pointer is NULL before
1075 * releasing the fd. This ensures that one clone task can't release
1076 * an fd while another clone is opening it.
1078 SYSCALL_DEFINE1(close, unsigned int, fd)
1081 struct files_struct *files = current->files;
1082 struct fdtable *fdt;
1085 spin_lock(&files->file_lock);
1086 fdt = files_fdtable(files);
1087 if (fd >= fdt->max_fds)
1092 rcu_assign_pointer(fdt->fd[fd], NULL);
1093 __clear_close_on_exec(fd, fdt);
1094 __put_unused_fd(files, fd);
1095 spin_unlock(&files->file_lock);
1096 retval = filp_close(filp, files);
1098 /* can't restart close syscall because file table entry was cleared */
1099 if (unlikely(retval == -ERESTARTSYS ||
1100 retval == -ERESTARTNOINTR ||
1101 retval == -ERESTARTNOHAND ||
1102 retval == -ERESTART_RESTARTBLOCK))
1108 spin_unlock(&files->file_lock);
1111 EXPORT_SYMBOL(sys_close);
1114 * This routine simulates a hangup on the tty, to arrange that users
1115 * are given clean terminals at login time.
1117 SYSCALL_DEFINE0(vhangup)
1119 if (capable(CAP_SYS_TTY_CONFIG)) {
1127 * Called when an inode is about to be open.
1128 * We use this to disallow opening large files on 32bit systems if
1129 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1130 * on this flag in sys_open.
1132 int generic_file_open(struct inode * inode, struct file * filp)
1134 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1139 EXPORT_SYMBOL(generic_file_open);
1142 * This is used by subsystems that don't want seekable
1143 * file descriptors. The function is not supposed to ever fail, the only
1144 * reason it returns an 'int' and not 'void' is so that it can be plugged
1145 * directly into file_operations structure.
1147 int nonseekable_open(struct inode *inode, struct file *filp)
1149 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1153 EXPORT_SYMBOL(nonseekable_open);