4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * super.c contains code to handle: - mount structures
8 * - filesystem drivers list
10 * - umount system call
13 * GK 2/5/95 - Changed to support mounting the root fs via NFS
15 * Added kerneld support: Jacques Gelinas and Bjorn Ekwall
16 * Added change_root: Werner Almesberger & Hans Lermen, Feb '96
17 * Added options to /proc/mounts:
18 * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996.
19 * Added devfs support: Richard Gooch <rgooch@atnf.csiro.au>, 13-JAN-1998
20 * Heavily rewritten for 'one fs - one tree' dcache architecture. AV, Mar 2000
23 #include <linux/module.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/smp_lock.h>
27 #include <linux/acct.h>
28 #include <linux/blkdev.h>
29 #include <linux/quotaops.h>
30 #include <linux/namei.h>
31 #include <linux/buffer_head.h> /* for fsync_super() */
32 #include <linux/mount.h>
33 #include <linux/security.h>
34 #include <linux/syscalls.h>
35 #include <linux/vfs.h>
36 #include <linux/writeback.h> /* for the emergency remount stuff */
37 #include <linux/idr.h>
38 #include <linux/kobject.h>
39 #include <linux/mutex.h>
40 #include <asm/uaccess.h>
43 LIST_HEAD(super_blocks
);
44 DEFINE_SPINLOCK(sb_lock
);
47 * alloc_super - create new superblock
48 * @type: filesystem type superblock should belong to
50 * Allocates and initializes a new &struct super_block. alloc_super()
51 * returns a pointer new superblock or %NULL if allocation had failed.
53 static struct super_block
*alloc_super(struct file_system_type
*type
)
55 struct super_block
*s
= kzalloc(sizeof(struct super_block
), GFP_USER
);
56 static struct super_operations default_op
;
59 if (security_sb_alloc(s
)) {
64 INIT_LIST_HEAD(&s
->s_dirty
);
65 INIT_LIST_HEAD(&s
->s_io
);
66 INIT_LIST_HEAD(&s
->s_more_io
);
67 INIT_LIST_HEAD(&s
->s_files
);
68 INIT_LIST_HEAD(&s
->s_instances
);
69 INIT_HLIST_HEAD(&s
->s_anon
);
70 INIT_LIST_HEAD(&s
->s_inodes
);
71 init_rwsem(&s
->s_umount
);
72 mutex_init(&s
->s_lock
);
73 lockdep_set_class(&s
->s_umount
, &type
->s_umount_key
);
75 * The locking rules for s_lock are up to the
76 * filesystem. For example ext3fs has different
77 * lock ordering than usbfs:
79 lockdep_set_class(&s
->s_lock
, &type
->s_lock_key
);
80 down_write(&s
->s_umount
);
82 atomic_set(&s
->s_active
, 1);
83 mutex_init(&s
->s_vfs_rename_mutex
);
84 mutex_init(&s
->s_dquot
.dqio_mutex
);
85 mutex_init(&s
->s_dquot
.dqonoff_mutex
);
86 init_rwsem(&s
->s_dquot
.dqptr_sem
);
87 init_waitqueue_head(&s
->s_wait_unfrozen
);
88 s
->s_maxbytes
= MAX_NON_LFS
;
89 s
->dq_op
= sb_dquot_ops
;
90 s
->s_qcop
= sb_quotactl_ops
;
91 s
->s_op
= &default_op
;
92 s
->s_time_gran
= 1000000000;
99 * destroy_super - frees a superblock
100 * @s: superblock to free
102 * Frees a superblock.
104 static inline void destroy_super(struct super_block
*s
)
111 /* Superblock refcounting */
114 * Drop a superblock's refcount. Returns non-zero if the superblock was
115 * destroyed. The caller must hold sb_lock.
117 int __put_super(struct super_block
*sb
)
121 if (!--sb
->s_count
) {
129 * Drop a superblock's refcount.
130 * Returns non-zero if the superblock is about to be destroyed and
131 * at least is already removed from super_blocks list, so if we are
132 * making a loop through super blocks then we need to restart.
133 * The caller must hold sb_lock.
135 int __put_super_and_need_restart(struct super_block
*sb
)
137 /* check for race with generic_shutdown_super() */
138 if (list_empty(&sb
->s_list
)) {
139 /* super block is removed, need to restart... */
143 /* can't be the last, since s_list is still in use */
145 BUG_ON(sb
->s_count
== 0);
150 * put_super - drop a temporary reference to superblock
151 * @sb: superblock in question
153 * Drops a temporary reference, frees superblock if there's no
156 static void put_super(struct super_block
*sb
)
160 spin_unlock(&sb_lock
);
165 * deactivate_super - drop an active reference to superblock
166 * @s: superblock to deactivate
168 * Drops an active reference to superblock, acquiring a temprory one if
169 * there is no active references left. In that case we lock superblock,
170 * tell fs driver to shut it down and drop the temporary reference we
173 void deactivate_super(struct super_block
*s
)
175 struct file_system_type
*fs
= s
->s_type
;
176 if (atomic_dec_and_lock(&s
->s_active
, &sb_lock
)) {
177 s
->s_count
-= S_BIAS
-1;
178 spin_unlock(&sb_lock
);
180 down_write(&s
->s_umount
);
187 EXPORT_SYMBOL(deactivate_super
);
190 * grab_super - acquire an active reference
191 * @s: reference we are trying to make active
193 * Tries to acquire an active reference. grab_super() is used when we
194 * had just found a superblock in super_blocks or fs_type->fs_supers
195 * and want to turn it into a full-blown active reference. grab_super()
196 * is called with sb_lock held and drops it. Returns 1 in case of
197 * success, 0 if we had failed (superblock contents was already dead or
198 * dying when grab_super() had been called).
200 static int grab_super(struct super_block
*s
) __releases(sb_lock
)
203 spin_unlock(&sb_lock
);
204 down_write(&s
->s_umount
);
207 if (s
->s_count
> S_BIAS
) {
208 atomic_inc(&s
->s_active
);
210 spin_unlock(&sb_lock
);
213 spin_unlock(&sb_lock
);
215 up_write(&s
->s_umount
);
222 * Superblock locking. We really ought to get rid of these two.
224 void lock_super(struct super_block
* sb
)
227 mutex_lock(&sb
->s_lock
);
230 void unlock_super(struct super_block
* sb
)
233 mutex_unlock(&sb
->s_lock
);
236 EXPORT_SYMBOL(lock_super
);
237 EXPORT_SYMBOL(unlock_super
);
240 * Write out and wait upon all dirty data associated with this
241 * superblock. Filesystem data as well as the underlying block
242 * device. Takes the superblock lock. Requires a second blkdev
243 * flush by the caller to complete the operation.
245 void __fsync_super(struct super_block
*sb
)
247 sync_inodes_sb(sb
, 0);
250 if (sb
->s_dirt
&& sb
->s_op
->write_super
)
251 sb
->s_op
->write_super(sb
);
253 if (sb
->s_op
->sync_fs
)
254 sb
->s_op
->sync_fs(sb
, 1);
255 sync_blockdev(sb
->s_bdev
);
256 sync_inodes_sb(sb
, 1);
260 * Write out and wait upon all dirty data associated with this
261 * superblock. Filesystem data as well as the underlying block
262 * device. Takes the superblock lock.
264 int fsync_super(struct super_block
*sb
)
267 return sync_blockdev(sb
->s_bdev
);
271 * generic_shutdown_super - common helper for ->kill_sb()
272 * @sb: superblock to kill
274 * generic_shutdown_super() does all fs-independent work on superblock
275 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
276 * that need destruction out of superblock, call generic_shutdown_super()
277 * and release aforementioned objects. Note: dentries and inodes _are_
278 * taken care of and do not need specific handling.
280 * Upon calling this function, the filesystem may no longer alter or
281 * rearrange the set of dentries belonging to this super_block, nor may it
282 * change the attachments of dentries to inodes.
284 void generic_shutdown_super(struct super_block
*sb
)
286 const struct super_operations
*sop
= sb
->s_op
;
289 shrink_dcache_for_umount(sb
);
292 sb
->s_flags
&= ~MS_ACTIVE
;
293 /* bad name - it should be evict_inodes() */
294 invalidate_inodes(sb
);
297 if (sop
->write_super
&& sb
->s_dirt
)
298 sop
->write_super(sb
);
302 /* Forget any remaining inodes */
303 if (invalidate_inodes(sb
)) {
304 printk("VFS: Busy inodes after unmount of %s. "
305 "Self-destruct in 5 seconds. Have a nice day...\n",
313 /* should be initialized for __put_super_and_need_restart() */
314 list_del_init(&sb
->s_list
);
315 list_del(&sb
->s_instances
);
316 spin_unlock(&sb_lock
);
317 up_write(&sb
->s_umount
);
320 EXPORT_SYMBOL(generic_shutdown_super
);
323 * sget - find or create a superblock
324 * @type: filesystem type superblock should belong to
325 * @test: comparison callback
326 * @set: setup callback
327 * @data: argument to each of them
329 struct super_block
*sget(struct file_system_type
*type
,
330 int (*test
)(struct super_block
*,void *),
331 int (*set
)(struct super_block
*,void *),
334 struct super_block
*s
= NULL
;
335 struct super_block
*old
;
341 list_for_each_entry(old
, &type
->fs_supers
, s_instances
) {
342 if (!test(old
, data
))
344 if (!grab_super(old
))
352 spin_unlock(&sb_lock
);
353 s
= alloc_super(type
);
355 return ERR_PTR(-ENOMEM
);
361 spin_unlock(&sb_lock
);
366 strlcpy(s
->s_id
, type
->name
, sizeof(s
->s_id
));
367 list_add_tail(&s
->s_list
, &super_blocks
);
368 list_add(&s
->s_instances
, &type
->fs_supers
);
369 spin_unlock(&sb_lock
);
370 get_filesystem(type
);
376 void drop_super(struct super_block
*sb
)
378 up_read(&sb
->s_umount
);
382 EXPORT_SYMBOL(drop_super
);
384 static inline void write_super(struct super_block
*sb
)
387 if (sb
->s_root
&& sb
->s_dirt
)
388 if (sb
->s_op
->write_super
)
389 sb
->s_op
->write_super(sb
);
394 * Note: check the dirty flag before waiting, so we don't
395 * hold up the sync while mounting a device. (The newly
396 * mounted device won't need syncing.)
398 void sync_supers(void)
400 struct super_block
*sb
;
404 list_for_each_entry(sb
, &super_blocks
, s_list
) {
407 spin_unlock(&sb_lock
);
408 down_read(&sb
->s_umount
);
410 up_read(&sb
->s_umount
);
412 if (__put_super_and_need_restart(sb
))
416 spin_unlock(&sb_lock
);
420 * Call the ->sync_fs super_op against all filesytems which are r/w and
421 * which implement it.
423 * This operation is careful to avoid the livelock which could easily happen
424 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
425 * is used only here. We set it against all filesystems and then clear it as
426 * we sync them. So redirtied filesystems are skipped.
428 * But if process A is currently running sync_filesytems and then process B
429 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
430 * flags again, which will cause process A to resync everything. Fix that with
433 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
435 void sync_filesystems(int wait
)
437 struct super_block
*sb
;
438 static DEFINE_MUTEX(mutex
);
440 mutex_lock(&mutex
); /* Could be down_interruptible */
442 list_for_each_entry(sb
, &super_blocks
, s_list
) {
443 if (!sb
->s_op
->sync_fs
)
445 if (sb
->s_flags
& MS_RDONLY
)
447 sb
->s_need_sync_fs
= 1;
451 list_for_each_entry(sb
, &super_blocks
, s_list
) {
452 if (!sb
->s_need_sync_fs
)
454 sb
->s_need_sync_fs
= 0;
455 if (sb
->s_flags
& MS_RDONLY
)
456 continue; /* hm. Was remounted r/o meanwhile */
458 spin_unlock(&sb_lock
);
459 down_read(&sb
->s_umount
);
460 if (sb
->s_root
&& (wait
|| sb
->s_dirt
))
461 sb
->s_op
->sync_fs(sb
, wait
);
462 up_read(&sb
->s_umount
);
463 /* restart only when sb is no longer on the list */
465 if (__put_super_and_need_restart(sb
))
468 spin_unlock(&sb_lock
);
469 mutex_unlock(&mutex
);
473 * get_super - get the superblock of a device
474 * @bdev: device to get the superblock for
476 * Scans the superblock list and finds the superblock of the file system
477 * mounted on the device given. %NULL is returned if no match is found.
480 struct super_block
* get_super(struct block_device
*bdev
)
482 struct super_block
*sb
;
489 list_for_each_entry(sb
, &super_blocks
, s_list
) {
490 if (sb
->s_bdev
== bdev
) {
492 spin_unlock(&sb_lock
);
493 down_read(&sb
->s_umount
);
496 up_read(&sb
->s_umount
);
497 /* restart only when sb is no longer on the list */
499 if (__put_super_and_need_restart(sb
))
503 spin_unlock(&sb_lock
);
507 EXPORT_SYMBOL(get_super
);
509 struct super_block
* user_get_super(dev_t dev
)
511 struct super_block
*sb
;
515 list_for_each_entry(sb
, &super_blocks
, s_list
) {
516 if (sb
->s_dev
== dev
) {
518 spin_unlock(&sb_lock
);
519 down_read(&sb
->s_umount
);
522 up_read(&sb
->s_umount
);
523 /* restart only when sb is no longer on the list */
525 if (__put_super_and_need_restart(sb
))
529 spin_unlock(&sb_lock
);
533 asmlinkage
long sys_ustat(unsigned dev
, struct ustat __user
* ubuf
)
535 struct super_block
*s
;
540 s
= user_get_super(new_decode_dev(dev
));
543 err
= vfs_statfs(s
->s_root
, &sbuf
);
548 memset(&tmp
,0,sizeof(struct ustat
));
549 tmp
.f_tfree
= sbuf
.f_bfree
;
550 tmp
.f_tinode
= sbuf
.f_ffree
;
552 err
= copy_to_user(ubuf
,&tmp
,sizeof(struct ustat
)) ? -EFAULT
: 0;
559 * @sb: superblock in question
561 * All files are marked read/only. We don't care about pending
562 * delete files so this should be used in 'force' mode only
565 static void mark_files_ro(struct super_block
*sb
)
570 list_for_each_entry(f
, &sb
->s_files
, f_u
.fu_list
) {
571 if (S_ISREG(f
->f_path
.dentry
->d_inode
->i_mode
) && file_count(f
))
572 f
->f_mode
&= ~FMODE_WRITE
;
578 * do_remount_sb - asks filesystem to change mount options.
579 * @sb: superblock in question
580 * @flags: numeric part of options
581 * @data: the rest of options
582 * @force: whether or not to force the change
584 * Alters the mount options of a mounted file system.
586 int do_remount_sb(struct super_block
*sb
, int flags
, void *data
, int force
)
591 if (!(flags
& MS_RDONLY
) && bdev_read_only(sb
->s_bdev
))
594 if (flags
& MS_RDONLY
)
596 shrink_dcache_sb(sb
);
599 /* If we are remounting RDONLY and current sb is read/write,
600 make sure there are no rw files opened */
601 if ((flags
& MS_RDONLY
) && !(sb
->s_flags
& MS_RDONLY
)) {
604 else if (!fs_may_remount_ro(sb
))
608 if (sb
->s_op
->remount_fs
) {
610 retval
= sb
->s_op
->remount_fs(sb
, &flags
, data
);
615 sb
->s_flags
= (sb
->s_flags
& ~MS_RMT_MASK
) | (flags
& MS_RMT_MASK
);
619 static void do_emergency_remount(unsigned long foo
)
621 struct super_block
*sb
;
624 list_for_each_entry(sb
, &super_blocks
, s_list
) {
626 spin_unlock(&sb_lock
);
627 down_read(&sb
->s_umount
);
628 if (sb
->s_root
&& sb
->s_bdev
&& !(sb
->s_flags
& MS_RDONLY
)) {
630 * ->remount_fs needs lock_kernel().
632 * What lock protects sb->s_flags??
635 do_remount_sb(sb
, MS_RDONLY
, NULL
, 1);
641 spin_unlock(&sb_lock
);
642 printk("Emergency Remount complete\n");
645 void emergency_remount(void)
647 pdflush_operation(do_emergency_remount
, 0);
651 * Unnamed block devices are dummy devices used by virtual
652 * filesystems which don't use real block-devices. -- jrs
655 static struct idr unnamed_dev_idr
;
656 static DEFINE_SPINLOCK(unnamed_dev_lock
);/* protects the above */
658 int set_anon_super(struct super_block
*s
, void *data
)
664 if (idr_pre_get(&unnamed_dev_idr
, GFP_ATOMIC
) == 0)
666 spin_lock(&unnamed_dev_lock
);
667 error
= idr_get_new(&unnamed_dev_idr
, NULL
, &dev
);
668 spin_unlock(&unnamed_dev_lock
);
669 if (error
== -EAGAIN
)
670 /* We raced and lost with another CPU. */
675 if ((dev
& MAX_ID_MASK
) == (1 << MINORBITS
)) {
676 spin_lock(&unnamed_dev_lock
);
677 idr_remove(&unnamed_dev_idr
, dev
);
678 spin_unlock(&unnamed_dev_lock
);
681 s
->s_dev
= MKDEV(0, dev
& MINORMASK
);
685 EXPORT_SYMBOL(set_anon_super
);
687 void kill_anon_super(struct super_block
*sb
)
689 int slot
= MINOR(sb
->s_dev
);
691 generic_shutdown_super(sb
);
692 spin_lock(&unnamed_dev_lock
);
693 idr_remove(&unnamed_dev_idr
, slot
);
694 spin_unlock(&unnamed_dev_lock
);
697 EXPORT_SYMBOL(kill_anon_super
);
699 void __init
unnamed_dev_init(void)
701 idr_init(&unnamed_dev_idr
);
704 void kill_litter_super(struct super_block
*sb
)
707 d_genocide(sb
->s_root
);
711 EXPORT_SYMBOL(kill_litter_super
);
714 static int set_bdev_super(struct super_block
*s
, void *data
)
717 s
->s_dev
= s
->s_bdev
->bd_dev
;
721 static int test_bdev_super(struct super_block
*s
, void *data
)
723 return (void *)s
->s_bdev
== data
;
726 int get_sb_bdev(struct file_system_type
*fs_type
,
727 int flags
, const char *dev_name
, void *data
,
728 int (*fill_super
)(struct super_block
*, void *, int),
729 struct vfsmount
*mnt
)
731 struct block_device
*bdev
;
732 struct super_block
*s
;
735 bdev
= open_bdev_excl(dev_name
, flags
, fs_type
);
737 return PTR_ERR(bdev
);
740 * once the super is inserted into the list by sget, s_umount
741 * will protect the lockfs code from trying to start a snapshot
742 * while we are mounting
744 down(&bdev
->bd_mount_sem
);
745 s
= sget(fs_type
, test_bdev_super
, set_bdev_super
, bdev
);
746 up(&bdev
->bd_mount_sem
);
751 if ((flags
^ s
->s_flags
) & MS_RDONLY
) {
752 up_write(&s
->s_umount
);
758 close_bdev_excl(bdev
);
760 char b
[BDEVNAME_SIZE
];
763 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
764 sb_set_blocksize(s
, block_size(bdev
));
765 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
767 up_write(&s
->s_umount
);
772 s
->s_flags
|= MS_ACTIVE
;
775 return simple_set_mnt(mnt
, s
);
780 close_bdev_excl(bdev
);
785 EXPORT_SYMBOL(get_sb_bdev
);
787 void kill_block_super(struct super_block
*sb
)
789 struct block_device
*bdev
= sb
->s_bdev
;
791 generic_shutdown_super(sb
);
793 close_bdev_excl(bdev
);
796 EXPORT_SYMBOL(kill_block_super
);
799 int get_sb_nodev(struct file_system_type
*fs_type
,
800 int flags
, void *data
,
801 int (*fill_super
)(struct super_block
*, void *, int),
802 struct vfsmount
*mnt
)
805 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
812 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
814 up_write(&s
->s_umount
);
818 s
->s_flags
|= MS_ACTIVE
;
819 return simple_set_mnt(mnt
, s
);
822 EXPORT_SYMBOL(get_sb_nodev
);
824 static int compare_single(struct super_block
*s
, void *p
)
829 int get_sb_single(struct file_system_type
*fs_type
,
830 int flags
, void *data
,
831 int (*fill_super
)(struct super_block
*, void *, int),
832 struct vfsmount
*mnt
)
834 struct super_block
*s
;
837 s
= sget(fs_type
, compare_single
, set_anon_super
, NULL
);
842 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
844 up_write(&s
->s_umount
);
848 s
->s_flags
|= MS_ACTIVE
;
850 do_remount_sb(s
, flags
, data
, 0);
851 return simple_set_mnt(mnt
, s
);
854 EXPORT_SYMBOL(get_sb_single
);
857 vfs_kern_mount(struct file_system_type
*type
, int flags
, const char *name
, void *data
)
859 struct vfsmount
*mnt
;
860 char *secdata
= NULL
;
864 return ERR_PTR(-ENODEV
);
867 mnt
= alloc_vfsmnt(name
);
872 secdata
= alloc_secdata();
876 error
= security_sb_copy_data(type
, data
, secdata
);
878 goto out_free_secdata
;
881 error
= type
->get_sb(type
, flags
, name
, data
, mnt
);
883 goto out_free_secdata
;
884 BUG_ON(!mnt
->mnt_sb
);
886 error
= security_sb_kern_mount(mnt
->mnt_sb
, secdata
);
890 mnt
->mnt_mountpoint
= mnt
->mnt_root
;
891 mnt
->mnt_parent
= mnt
;
892 up_write(&mnt
->mnt_sb
->s_umount
);
893 free_secdata(secdata
);
897 up_write(&mnt
->mnt_sb
->s_umount
);
898 deactivate_super(mnt
->mnt_sb
);
900 free_secdata(secdata
);
904 return ERR_PTR(error
);
907 EXPORT_SYMBOL_GPL(vfs_kern_mount
);
909 static struct vfsmount
*fs_set_subtype(struct vfsmount
*mnt
, const char *fstype
)
912 const char *subtype
= strchr(fstype
, '.');
921 mnt
->mnt_sb
->s_subtype
= kstrdup(subtype
, GFP_KERNEL
);
923 if (!mnt
->mnt_sb
->s_subtype
)
933 do_kern_mount(const char *fstype
, int flags
, const char *name
, void *data
)
935 struct file_system_type
*type
= get_fs_type(fstype
);
936 struct vfsmount
*mnt
;
938 return ERR_PTR(-ENODEV
);
939 mnt
= vfs_kern_mount(type
, flags
, name
, data
);
940 if (!IS_ERR(mnt
) && (type
->fs_flags
& FS_HAS_SUBTYPE
) &&
941 !mnt
->mnt_sb
->s_subtype
)
942 mnt
= fs_set_subtype(mnt
, fstype
);
943 put_filesystem(type
);
947 struct vfsmount
*kern_mount_data(struct file_system_type
*type
, void *data
)
949 return vfs_kern_mount(type
, MS_KERNMOUNT
, type
->name
, data
);
952 EXPORT_SYMBOL_GPL(kern_mount_data
);