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/quotaops.h>
12 #include <linux/fsnotify.h>
13 #include <linux/module.h>
14 #include <linux/slab.h>
15 #include <linux/tty.h>
16 #include <linux/namei.h>
17 #include <linux/backing-dev.h>
18 #include <linux/capability.h>
19 #include <linux/securebits.h>
20 #include <linux/security.h>
21 #include <linux/mount.h>
22 #include <linux/vfs.h>
23 #include <linux/fcntl.h>
24 #include <asm/uaccess.h>
26 #include <linux/personality.h>
27 #include <linux/pagemap.h>
28 #include <linux/syscalls.h>
29 #include <linux/rcupdate.h>
30 #include <linux/audit.h>
31 #include <linux/falloc.h>
33 int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
39 if (dentry->d_sb->s_op->statfs) {
40 memset(buf, 0, sizeof(*buf));
41 retval = security_sb_statfs(dentry);
44 retval = dentry->d_sb->s_op->statfs(dentry, buf);
45 if (retval == 0 && buf->f_frsize == 0)
46 buf->f_frsize = buf->f_bsize;
52 EXPORT_SYMBOL(vfs_statfs);
54 static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
59 retval = vfs_statfs(dentry, &st);
63 if (sizeof(*buf) == sizeof(st))
64 memcpy(buf, &st, sizeof(st));
66 if (sizeof buf->f_blocks == 4) {
67 if ((st.f_blocks | st.f_bfree | st.f_bavail |
68 st.f_bsize | st.f_frsize) &
69 0xffffffff00000000ULL)
72 * f_files and f_ffree may be -1; it's okay to stuff
75 if (st.f_files != -1 &&
76 (st.f_files & 0xffffffff00000000ULL))
78 if (st.f_ffree != -1 &&
79 (st.f_ffree & 0xffffffff00000000ULL))
83 buf->f_type = st.f_type;
84 buf->f_bsize = st.f_bsize;
85 buf->f_blocks = st.f_blocks;
86 buf->f_bfree = st.f_bfree;
87 buf->f_bavail = st.f_bavail;
88 buf->f_files = st.f_files;
89 buf->f_ffree = st.f_ffree;
90 buf->f_fsid = st.f_fsid;
91 buf->f_namelen = st.f_namelen;
92 buf->f_frsize = st.f_frsize;
93 memset(buf->f_spare, 0, sizeof(buf->f_spare));
98 static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
103 retval = vfs_statfs(dentry, &st);
107 if (sizeof(*buf) == sizeof(st))
108 memcpy(buf, &st, sizeof(st));
110 buf->f_type = st.f_type;
111 buf->f_bsize = st.f_bsize;
112 buf->f_blocks = st.f_blocks;
113 buf->f_bfree = st.f_bfree;
114 buf->f_bavail = st.f_bavail;
115 buf->f_files = st.f_files;
116 buf->f_ffree = st.f_ffree;
117 buf->f_fsid = st.f_fsid;
118 buf->f_namelen = st.f_namelen;
119 buf->f_frsize = st.f_frsize;
120 memset(buf->f_spare, 0, sizeof(buf->f_spare));
125 SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct statfs __user *, buf)
130 error = user_path(pathname, &path);
133 error = vfs_statfs_native(path.dentry, &tmp);
134 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
141 SYSCALL_DEFINE3(statfs64, const char __user *, pathname, size_t, sz, struct statfs64 __user *, buf)
146 if (sz != sizeof(*buf))
148 error = user_path(pathname, &path);
151 error = vfs_statfs64(path.dentry, &tmp);
152 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
159 SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct statfs __user *, buf)
169 error = vfs_statfs_native(file->f_path.dentry, &tmp);
170 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
177 SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, size_t, sz, struct statfs64 __user *, buf)
183 if (sz != sizeof(*buf))
190 error = vfs_statfs64(file->f_path.dentry, &tmp);
191 if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
198 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
202 struct iattr newattrs;
204 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
208 newattrs.ia_size = length;
209 newattrs.ia_valid = ATTR_SIZE | time_attrs;
211 newattrs.ia_file = filp;
212 newattrs.ia_valid |= ATTR_FILE;
215 /* Remove suid/sgid on truncate too */
216 newattrs.ia_valid |= should_remove_suid(dentry);
218 mutex_lock(&dentry->d_inode->i_mutex);
219 err = notify_change(dentry, &newattrs);
220 mutex_unlock(&dentry->d_inode->i_mutex);
224 static long do_sys_truncate(const char __user *pathname, loff_t length)
231 if (length < 0) /* sorry, but loff_t says... */
234 error = user_path(pathname, &path);
237 inode = path.dentry->d_inode;
239 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
241 if (S_ISDIR(inode->i_mode))
245 if (!S_ISREG(inode->i_mode))
248 error = mnt_want_write(path.mnt);
252 error = inode_permission(inode, MAY_WRITE);
254 goto mnt_drop_write_and_out;
257 if (IS_APPEND(inode))
258 goto mnt_drop_write_and_out;
260 error = get_write_access(inode);
262 goto mnt_drop_write_and_out;
265 * Make sure that there are no leases. get_write_access() protects
266 * against the truncate racing with a lease-granting setlease().
268 error = break_lease(inode, FMODE_WRITE);
270 goto put_write_and_out;
272 error = locks_verify_truncate(inode, NULL, length);
274 error = security_path_truncate(&path, length, 0);
277 error = do_truncate(path.dentry, length, 0, NULL);
281 put_write_access(inode);
282 mnt_drop_write_and_out:
283 mnt_drop_write(path.mnt);
290 SYSCALL_DEFINE2(truncate, const char __user *, path, unsigned long, length)
292 /* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
293 return do_sys_truncate(path, (long)length);
296 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
298 struct inode * inode;
299 struct dentry *dentry;
311 /* explicitly opened as large or we are on 64-bit box */
312 if (file->f_flags & O_LARGEFILE)
315 dentry = file->f_path.dentry;
316 inode = dentry->d_inode;
318 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
322 /* Cannot ftruncate over 2^31 bytes without large file support */
323 if (small && length > MAX_NON_LFS)
327 if (IS_APPEND(inode))
330 error = locks_verify_truncate(inode, file, length);
332 error = security_path_truncate(&file->f_path, length,
333 ATTR_MTIME|ATTR_CTIME);
335 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
342 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
344 long ret = do_sys_ftruncate(fd, length, 1);
345 /* avoid REGPARM breakage on x86: */
346 asmlinkage_protect(2, ret, fd, length);
350 /* LFS versions of truncate are only needed on 32 bit machines */
351 #if BITS_PER_LONG == 32
352 SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
354 return do_sys_truncate(path, length);
356 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
357 asmlinkage long SyS_truncate64(long path, loff_t length)
359 return SYSC_truncate64((const char __user *) path, length);
361 SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
364 SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
366 long ret = do_sys_ftruncate(fd, length, 0);
367 /* avoid REGPARM breakage on x86: */
368 asmlinkage_protect(2, ret, fd, length);
371 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
372 asmlinkage long SyS_ftruncate64(long fd, loff_t length)
374 return SYSC_ftruncate64((unsigned int) fd, length);
376 SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
378 #endif /* BITS_PER_LONG == 32 */
380 SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
386 if (offset < 0 || len <= 0)
389 /* Return error if mode is not supported */
391 if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
398 if (!(file->f_mode & FMODE_WRITE))
401 * Revalidate the write permissions, in case security policy has
402 * changed since the files were opened.
404 ret = security_file_permission(file, MAY_WRITE);
408 inode = file->f_path.dentry->d_inode;
411 if (S_ISFIFO(inode->i_mode))
416 * Let individual file system decide if it supports preallocation
417 * for directories or not.
419 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
423 /* Check for wrap through zero too */
424 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
427 if (inode->i_op->fallocate)
428 ret = inode->i_op->fallocate(inode, mode, offset, len);
437 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
438 asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
440 return SYSC_fallocate((int)fd, (int)mode, offset, len);
442 SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
446 * access() needs to use the real uid/gid, not the effective uid/gid.
447 * We do this by temporarily clearing all FS-related capabilities and
448 * switching the fsuid/fsgid around to the real ones.
450 asmlinkage long sys_faccessat(int dfd, const char __user *filename, int mode)
452 const struct cred *old_cred;
453 struct cred *override_cred;
458 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
461 override_cred = prepare_creds();
465 override_cred->fsuid = override_cred->uid;
466 override_cred->fsgid = override_cred->gid;
468 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
469 /* Clear the capabilities if we switch to a non-root user */
470 if (override_cred->uid)
471 cap_clear(override_cred->cap_effective);
473 override_cred->cap_effective =
474 override_cred->cap_permitted;
477 old_cred = override_creds(override_cred);
479 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
483 inode = path.dentry->d_inode;
485 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
487 * MAY_EXEC on regular files is denied if the fs is mounted
488 * with the "noexec" flag.
491 if (path.mnt->mnt_flags & MNT_NOEXEC)
492 goto out_path_release;
495 res = inode_permission(inode, mode | MAY_ACCESS);
496 /* SuS v2 requires we report a read only fs too */
497 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
498 goto out_path_release;
500 * This is a rare case where using __mnt_is_readonly()
501 * is OK without a mnt_want/drop_write() pair. Since
502 * no actual write to the fs is performed here, we do
503 * not need to telegraph to that to anyone.
505 * By doing this, we accept that this access is
506 * inherently racy and know that the fs may change
507 * state before we even see this result.
509 if (__mnt_is_readonly(path.mnt))
515 revert_creds(old_cred);
516 put_cred(override_cred);
520 asmlinkage long sys_access(const char __user *filename, int mode)
522 return sys_faccessat(AT_FDCWD, filename, mode);
525 asmlinkage long sys_chdir(const char __user * filename)
530 error = user_path_dir(filename, &path);
534 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
538 set_fs_pwd(current->fs, &path);
546 asmlinkage long sys_fchdir(unsigned int fd)
557 inode = file->f_path.dentry->d_inode;
560 if (!S_ISDIR(inode->i_mode))
563 error = inode_permission(inode, MAY_EXEC | MAY_ACCESS);
565 set_fs_pwd(current->fs, &file->f_path);
572 SYSCALL_DEFINE1(chroot, const char __user *, filename)
577 error = user_path_dir(filename, &path);
581 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
586 if (!capable(CAP_SYS_CHROOT))
589 set_fs_root(current->fs, &path);
597 SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
599 struct inode * inode;
600 struct dentry * dentry;
603 struct iattr newattrs;
609 dentry = file->f_path.dentry;
610 inode = dentry->d_inode;
612 audit_inode(NULL, dentry);
614 err = mnt_want_write(file->f_path.mnt);
617 mutex_lock(&inode->i_mutex);
618 if (mode == (mode_t) -1)
619 mode = inode->i_mode;
620 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
621 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
622 err = notify_change(dentry, &newattrs);
623 mutex_unlock(&inode->i_mutex);
624 mnt_drop_write(file->f_path.mnt);
631 asmlinkage long sys_fchmodat(int dfd, const char __user *filename,
637 struct iattr newattrs;
639 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
642 inode = path.dentry->d_inode;
644 error = mnt_want_write(path.mnt);
647 mutex_lock(&inode->i_mutex);
648 if (mode == (mode_t) -1)
649 mode = inode->i_mode;
650 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
651 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
652 error = notify_change(path.dentry, &newattrs);
653 mutex_unlock(&inode->i_mutex);
654 mnt_drop_write(path.mnt);
661 SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
663 return sys_fchmodat(AT_FDCWD, filename, mode);
666 static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
668 struct inode *inode = dentry->d_inode;
670 struct iattr newattrs;
672 newattrs.ia_valid = ATTR_CTIME;
673 if (user != (uid_t) -1) {
674 newattrs.ia_valid |= ATTR_UID;
675 newattrs.ia_uid = user;
677 if (group != (gid_t) -1) {
678 newattrs.ia_valid |= ATTR_GID;
679 newattrs.ia_gid = group;
681 if (!S_ISDIR(inode->i_mode))
683 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
684 mutex_lock(&inode->i_mutex);
685 error = notify_change(dentry, &newattrs);
686 mutex_unlock(&inode->i_mutex);
691 asmlinkage long sys_chown(const char __user * filename, uid_t user, gid_t group)
696 error = user_path(filename, &path);
699 error = mnt_want_write(path.mnt);
702 error = chown_common(path.dentry, user, group);
703 mnt_drop_write(path.mnt);
710 asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
711 gid_t group, int flag)
717 if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
720 follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
721 error = user_path_at(dfd, filename, follow, &path);
724 error = mnt_want_write(path.mnt);
727 error = chown_common(path.dentry, user, group);
728 mnt_drop_write(path.mnt);
735 asmlinkage long sys_lchown(const char __user * filename, uid_t user, gid_t group)
740 error = user_lpath(filename, &path);
743 error = mnt_want_write(path.mnt);
746 error = chown_common(path.dentry, user, group);
747 mnt_drop_write(path.mnt);
755 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group)
759 struct dentry * dentry;
765 error = mnt_want_write(file->f_path.mnt);
768 dentry = file->f_path.dentry;
769 audit_inode(NULL, dentry);
770 error = chown_common(dentry, user, group);
771 mnt_drop_write(file->f_path.mnt);
779 * You have to be very careful that these write
780 * counts get cleaned up in error cases and
781 * upon __fput(). This should probably never
782 * be called outside of __dentry_open().
784 static inline int __get_file_write_access(struct inode *inode,
785 struct vfsmount *mnt)
788 error = get_write_access(inode);
792 * Do not take mount writer counts on
793 * special files since no writes to
794 * the mount itself will occur.
796 if (!special_file(inode->i_mode)) {
798 * Balanced in __fput()
800 error = mnt_want_write(mnt);
802 put_write_access(inode);
807 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
808 int flags, struct file *f,
809 int (*open)(struct inode *, struct file *),
810 const struct cred *cred)
816 f->f_mode = (__force fmode_t)((flags+1) & O_ACCMODE) | FMODE_LSEEK |
817 FMODE_PREAD | FMODE_PWRITE;
818 inode = dentry->d_inode;
819 if (f->f_mode & FMODE_WRITE) {
820 error = __get_file_write_access(inode, mnt);
823 if (!special_file(inode->i_mode))
827 f->f_mapping = inode->i_mapping;
828 f->f_path.dentry = dentry;
831 f->f_op = fops_get(inode->i_fop);
832 file_move(f, &inode->i_sb->s_files);
834 error = security_dentry_open(f, cred);
838 if (!open && f->f_op)
839 open = f->f_op->open;
841 error = open(inode, f);
846 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
848 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
850 /* NB: we're sure to have correct a_ops only after f_op->open */
851 if (f->f_flags & O_DIRECT) {
852 if (!f->f_mapping->a_ops ||
853 ((!f->f_mapping->a_ops->direct_IO) &&
854 (!f->f_mapping->a_ops->get_xip_mem))) {
856 f = ERR_PTR(-EINVAL);
864 if (f->f_mode & FMODE_WRITE) {
865 put_write_access(inode);
866 if (!special_file(inode->i_mode)) {
868 * We don't consider this a real
869 * mnt_want/drop_write() pair
870 * because it all happenend right
871 * here, so just reset the state.
878 f->f_path.dentry = NULL;
879 f->f_path.mnt = NULL;
884 return ERR_PTR(error);
888 * lookup_instantiate_filp - instantiates the open intent filp
889 * @nd: pointer to nameidata
890 * @dentry: pointer to dentry
891 * @open: open callback
893 * Helper for filesystems that want to use lookup open intents and pass back
894 * a fully instantiated struct file to the caller.
895 * This function is meant to be called from within a filesystem's
897 * Beware of calling it for non-regular files! Those ->open methods might block
898 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
899 * leading to a deadlock, as nobody can open that fifo anymore, because
900 * another process to open fifo will block on locked parent when doing lookup).
901 * Note that in case of error, nd->intent.open.file is destroyed, but the
902 * path information remains valid.
903 * If the open callback is set to NULL, then the standard f_op->open()
904 * filesystem callback is substituted.
906 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
907 int (*open)(struct inode *, struct file *))
909 const struct cred *cred = current_cred();
911 if (IS_ERR(nd->intent.open.file))
915 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
916 nd->intent.open.flags - 1,
917 nd->intent.open.file,
920 return nd->intent.open.file;
922 release_open_intent(nd);
923 nd->intent.open.file = (struct file *)dentry;
926 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
929 * nameidata_to_filp - convert a nameidata to an open filp.
930 * @nd: pointer to nameidata
933 * Note that this function destroys the original nameidata
935 struct file *nameidata_to_filp(struct nameidata *nd, int flags)
937 const struct cred *cred = current_cred();
940 /* Pick up the filp from the open intent */
941 filp = nd->intent.open.file;
942 /* Has the filesystem initialised the file for us? */
943 if (filp->f_path.dentry == NULL)
944 filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
952 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
955 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
956 const struct cred *cred)
962 * We must always pass in a valid mount pointer. Historically
963 * callers got away with not passing it, but we must enforce this at
964 * the earliest possible point now to avoid strange problems deep in the
968 printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
970 return ERR_PTR(-EINVAL);
974 f = get_empty_filp();
978 return ERR_PTR(error);
981 return __dentry_open(dentry, mnt, flags, f, NULL, cred);
983 EXPORT_SYMBOL(dentry_open);
985 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
987 struct fdtable *fdt = files_fdtable(files);
988 __FD_CLR(fd, fdt->open_fds);
989 if (fd < files->next_fd)
993 void put_unused_fd(unsigned int fd)
995 struct files_struct *files = current->files;
996 spin_lock(&files->file_lock);
997 __put_unused_fd(files, fd);
998 spin_unlock(&files->file_lock);
1001 EXPORT_SYMBOL(put_unused_fd);
1004 * Install a file pointer in the fd array.
1006 * The VFS is full of places where we drop the files lock between
1007 * setting the open_fds bitmap and installing the file in the file
1008 * array. At any such point, we are vulnerable to a dup2() race
1009 * installing a file in the array before us. We need to detect this and
1010 * fput() the struct file we are about to overwrite in this case.
1012 * It should never happen - if we allow dup2() do it, _really_ bad things
1016 void fd_install(unsigned int fd, struct file *file)
1018 struct files_struct *files = current->files;
1019 struct fdtable *fdt;
1020 spin_lock(&files->file_lock);
1021 fdt = files_fdtable(files);
1022 BUG_ON(fdt->fd[fd] != NULL);
1023 rcu_assign_pointer(fdt->fd[fd], file);
1024 spin_unlock(&files->file_lock);
1027 EXPORT_SYMBOL(fd_install);
1029 long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
1031 char *tmp = getname(filename);
1032 int fd = PTR_ERR(tmp);
1035 fd = get_unused_fd_flags(flags);
1037 struct file *f = do_filp_open(dfd, tmp, flags, mode);
1042 fsnotify_open(f->f_path.dentry);
1051 asmlinkage long sys_open(const char __user *filename, int flags, int mode)
1055 if (force_o_largefile())
1056 flags |= O_LARGEFILE;
1058 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1059 /* avoid REGPARM breakage on x86: */
1060 asmlinkage_protect(3, ret, filename, flags, mode);
1064 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
1069 if (force_o_largefile())
1070 flags |= O_LARGEFILE;
1072 ret = do_sys_open(dfd, filename, flags, mode);
1073 /* avoid REGPARM breakage on x86: */
1074 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
1081 * For backward compatibility? Maybe this should be moved
1082 * into arch/i386 instead?
1084 asmlinkage long sys_creat(const char __user * pathname, int mode)
1086 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1092 * "id" is the POSIX thread ID. We use the
1093 * files pointer for this..
1095 int filp_close(struct file *filp, fl_owner_t id)
1099 if (!file_count(filp)) {
1100 printk(KERN_ERR "VFS: Close: file count is 0\n");
1104 if (filp->f_op && filp->f_op->flush)
1105 retval = filp->f_op->flush(filp, id);
1107 dnotify_flush(filp, id);
1108 locks_remove_posix(filp, id);
1113 EXPORT_SYMBOL(filp_close);
1116 * Careful here! We test whether the file pointer is NULL before
1117 * releasing the fd. This ensures that one clone task can't release
1118 * an fd while another clone is opening it.
1120 asmlinkage long sys_close(unsigned int fd)
1123 struct files_struct *files = current->files;
1124 struct fdtable *fdt;
1127 spin_lock(&files->file_lock);
1128 fdt = files_fdtable(files);
1129 if (fd >= fdt->max_fds)
1134 rcu_assign_pointer(fdt->fd[fd], NULL);
1135 FD_CLR(fd, fdt->close_on_exec);
1136 __put_unused_fd(files, fd);
1137 spin_unlock(&files->file_lock);
1138 retval = filp_close(filp, files);
1140 /* can't restart close syscall because file table entry was cleared */
1141 if (unlikely(retval == -ERESTARTSYS ||
1142 retval == -ERESTARTNOINTR ||
1143 retval == -ERESTARTNOHAND ||
1144 retval == -ERESTART_RESTARTBLOCK))
1150 spin_unlock(&files->file_lock);
1154 EXPORT_SYMBOL(sys_close);
1157 * This routine simulates a hangup on the tty, to arrange that users
1158 * are given clean terminals at login time.
1160 asmlinkage long sys_vhangup(void)
1162 if (capable(CAP_SYS_TTY_CONFIG)) {
1170 * Called when an inode is about to be open.
1171 * We use this to disallow opening large files on 32bit systems if
1172 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1173 * on this flag in sys_open.
1175 int generic_file_open(struct inode * inode, struct file * filp)
1177 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1182 EXPORT_SYMBOL(generic_file_open);
1185 * This is used by subsystems that don't want seekable
1188 int nonseekable_open(struct inode *inode, struct file *filp)
1190 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1194 EXPORT_SYMBOL(nonseekable_open);