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/acct.h>
26 #include <linux/blkdev.h>
27 #include <linux/mount.h>
28 #include <linux/security.h>
29 #include <linux/writeback.h> /* for the emergency remount stuff */
30 #include <linux/idr.h>
31 #include <linux/mutex.h>
32 #include <linux/backing-dev.h>
33 #include <linux/rculist_bl.h>
34 #include <linux/cleancache.h>
38 LIST_HEAD(super_blocks
);
39 DEFINE_SPINLOCK(sb_lock
);
42 * One thing we have to be careful of with a per-sb shrinker is that we don't
43 * drop the last active reference to the superblock from within the shrinker.
44 * If that happens we could trigger unregistering the shrinker from within the
45 * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we
46 * take a passive reference to the superblock to avoid this from occurring.
48 static int prune_super(struct shrinker
*shrink
, struct shrink_control
*sc
)
50 struct super_block
*sb
;
54 sb
= container_of(shrink
, struct super_block
, s_shrink
);
57 * Deadlock avoidance. We may hold various FS locks, and we don't want
58 * to recurse into the FS that called us in clear_inode() and friends..
60 if (sc
->nr_to_scan
&& !(sc
->gfp_mask
& __GFP_FS
))
63 if (!grab_super_passive(sb
))
64 return !sc
->nr_to_scan
? 0 : -1;
66 if (sb
->s_op
&& sb
->s_op
->nr_cached_objects
)
67 fs_objects
= sb
->s_op
->nr_cached_objects(sb
);
69 total_objects
= sb
->s_nr_dentry_unused
+
70 sb
->s_nr_inodes_unused
+ fs_objects
+ 1;
76 /* proportion the scan between the caches */
77 dentries
= (sc
->nr_to_scan
* sb
->s_nr_dentry_unused
) /
79 inodes
= (sc
->nr_to_scan
* sb
->s_nr_inodes_unused
) /
82 fs_objects
= (sc
->nr_to_scan
* fs_objects
) /
85 * prune the dcache first as the icache is pinned by it, then
86 * prune the icache, followed by the filesystem specific caches
88 prune_dcache_sb(sb
, dentries
);
89 prune_icache_sb(sb
, inodes
);
91 if (fs_objects
&& sb
->s_op
->free_cached_objects
) {
92 sb
->s_op
->free_cached_objects(sb
, fs_objects
);
93 fs_objects
= sb
->s_op
->nr_cached_objects(sb
);
95 total_objects
= sb
->s_nr_dentry_unused
+
96 sb
->s_nr_inodes_unused
+ fs_objects
;
99 total_objects
= (total_objects
/ 100) * sysctl_vfs_cache_pressure
;
101 return total_objects
;
105 * alloc_super - create new superblock
106 * @type: filesystem type superblock should belong to
108 * Allocates and initializes a new &struct super_block. alloc_super()
109 * returns a pointer new superblock or %NULL if allocation had failed.
111 static struct super_block
*alloc_super(struct file_system_type
*type
)
113 struct super_block
*s
= kzalloc(sizeof(struct super_block
), GFP_USER
);
114 static const struct super_operations default_op
;
117 if (security_sb_alloc(s
)) {
123 s
->s_files
= alloc_percpu(struct list_head
);
132 for_each_possible_cpu(i
)
133 INIT_LIST_HEAD(per_cpu_ptr(s
->s_files
, i
));
136 INIT_LIST_HEAD(&s
->s_files
);
138 s
->s_bdi
= &default_backing_dev_info
;
139 INIT_HLIST_NODE(&s
->s_instances
);
140 INIT_HLIST_BL_HEAD(&s
->s_anon
);
141 INIT_LIST_HEAD(&s
->s_inodes
);
142 INIT_LIST_HEAD(&s
->s_dentry_lru
);
143 INIT_LIST_HEAD(&s
->s_inode_lru
);
144 spin_lock_init(&s
->s_inode_lru_lock
);
145 init_rwsem(&s
->s_umount
);
146 mutex_init(&s
->s_lock
);
147 lockdep_set_class(&s
->s_umount
, &type
->s_umount_key
);
149 * The locking rules for s_lock are up to the
150 * filesystem. For example ext3fs has different
151 * lock ordering than usbfs:
153 lockdep_set_class(&s
->s_lock
, &type
->s_lock_key
);
155 * sget() can have s_umount recursion.
157 * When it cannot find a suitable sb, it allocates a new
158 * one (this one), and tries again to find a suitable old
161 * In case that succeeds, it will acquire the s_umount
162 * lock of the old one. Since these are clearly distrinct
163 * locks, and this object isn't exposed yet, there's no
166 * Annotate this by putting this lock in a different
169 down_write_nested(&s
->s_umount
, SINGLE_DEPTH_NESTING
);
171 atomic_set(&s
->s_active
, 1);
172 mutex_init(&s
->s_vfs_rename_mutex
);
173 lockdep_set_class(&s
->s_vfs_rename_mutex
, &type
->s_vfs_rename_key
);
174 mutex_init(&s
->s_dquot
.dqio_mutex
);
175 mutex_init(&s
->s_dquot
.dqonoff_mutex
);
176 init_rwsem(&s
->s_dquot
.dqptr_sem
);
177 init_waitqueue_head(&s
->s_wait_unfrozen
);
178 s
->s_maxbytes
= MAX_NON_LFS
;
179 s
->s_op
= &default_op
;
180 s
->s_time_gran
= 1000000000;
181 s
->cleancache_poolid
= -1;
183 s
->s_shrink
.seeks
= DEFAULT_SEEKS
;
184 s
->s_shrink
.shrink
= prune_super
;
185 s
->s_shrink
.batch
= 1024;
192 * destroy_super - frees a superblock
193 * @s: superblock to free
195 * Frees a superblock.
197 static inline void destroy_super(struct super_block
*s
)
200 free_percpu(s
->s_files
);
208 /* Superblock refcounting */
211 * Drop a superblock's refcount. The caller must hold sb_lock.
213 static void __put_super(struct super_block
*sb
)
215 if (!--sb
->s_count
) {
216 list_del_init(&sb
->s_list
);
222 * put_super - drop a temporary reference to superblock
223 * @sb: superblock in question
225 * Drops a temporary reference, frees superblock if there's no
228 static void put_super(struct super_block
*sb
)
232 spin_unlock(&sb_lock
);
237 * deactivate_locked_super - drop an active reference to superblock
238 * @s: superblock to deactivate
240 * Drops an active reference to superblock, converting it into a temprory
241 * one if there is no other active references left. In that case we
242 * tell fs driver to shut it down and drop the temporary reference we
245 * Caller holds exclusive lock on superblock; that lock is released.
247 void deactivate_locked_super(struct super_block
*s
)
249 struct file_system_type
*fs
= s
->s_type
;
250 if (atomic_dec_and_test(&s
->s_active
)) {
251 cleancache_flush_fs(s
);
254 /* caches are now gone, we can safely kill the shrinker now */
255 unregister_shrinker(&s
->s_shrink
);
258 * We need to call rcu_barrier so all the delayed rcu free
259 * inodes are flushed before we release the fs module.
265 up_write(&s
->s_umount
);
269 EXPORT_SYMBOL(deactivate_locked_super
);
272 * deactivate_super - drop an active reference to superblock
273 * @s: superblock to deactivate
275 * Variant of deactivate_locked_super(), except that superblock is *not*
276 * locked by caller. If we are going to drop the final active reference,
277 * lock will be acquired prior to that.
279 void deactivate_super(struct super_block
*s
)
281 if (!atomic_add_unless(&s
->s_active
, -1, 1)) {
282 down_write(&s
->s_umount
);
283 deactivate_locked_super(s
);
287 EXPORT_SYMBOL(deactivate_super
);
290 * grab_super - acquire an active reference
291 * @s: reference we are trying to make active
293 * Tries to acquire an active reference. grab_super() is used when we
294 * had just found a superblock in super_blocks or fs_type->fs_supers
295 * and want to turn it into a full-blown active reference. grab_super()
296 * is called with sb_lock held and drops it. Returns 1 in case of
297 * success, 0 if we had failed (superblock contents was already dead or
298 * dying when grab_super() had been called).
300 static int grab_super(struct super_block
*s
) __releases(sb_lock
)
302 if (atomic_inc_not_zero(&s
->s_active
)) {
303 spin_unlock(&sb_lock
);
306 /* it's going away */
308 spin_unlock(&sb_lock
);
309 /* wait for it to die */
310 down_write(&s
->s_umount
);
311 up_write(&s
->s_umount
);
317 * grab_super_passive - acquire a passive reference
318 * @s: reference we are trying to grab
320 * Tries to acquire a passive reference. This is used in places where we
321 * cannot take an active reference but we need to ensure that the
322 * superblock does not go away while we are working on it. It returns
323 * false if a reference was not gained, and returns true with the s_umount
324 * lock held in read mode if a reference is gained. On successful return,
325 * the caller must drop the s_umount lock and the passive reference when
328 bool grab_super_passive(struct super_block
*sb
)
331 if (hlist_unhashed(&sb
->s_instances
)) {
332 spin_unlock(&sb_lock
);
337 spin_unlock(&sb_lock
);
339 if (down_read_trylock(&sb
->s_umount
)) {
340 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
342 up_read(&sb
->s_umount
);
350 * Superblock locking. We really ought to get rid of these two.
352 void lock_super(struct super_block
* sb
)
354 mutex_lock(&sb
->s_lock
);
357 void unlock_super(struct super_block
* sb
)
359 mutex_unlock(&sb
->s_lock
);
362 EXPORT_SYMBOL(lock_super
);
363 EXPORT_SYMBOL(unlock_super
);
366 * generic_shutdown_super - common helper for ->kill_sb()
367 * @sb: superblock to kill
369 * generic_shutdown_super() does all fs-independent work on superblock
370 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
371 * that need destruction out of superblock, call generic_shutdown_super()
372 * and release aforementioned objects. Note: dentries and inodes _are_
373 * taken care of and do not need specific handling.
375 * Upon calling this function, the filesystem may no longer alter or
376 * rearrange the set of dentries belonging to this super_block, nor may it
377 * change the attachments of dentries to inodes.
379 void generic_shutdown_super(struct super_block
*sb
)
381 const struct super_operations
*sop
= sb
->s_op
;
384 shrink_dcache_for_umount(sb
);
386 sb
->s_flags
&= ~MS_ACTIVE
;
388 fsnotify_unmount_inodes(&sb
->s_inodes
);
395 if (!list_empty(&sb
->s_inodes
)) {
396 printk("VFS: Busy inodes after unmount of %s. "
397 "Self-destruct in 5 seconds. Have a nice day...\n",
402 /* should be initialized for __put_super_and_need_restart() */
403 hlist_del_init(&sb
->s_instances
);
404 spin_unlock(&sb_lock
);
405 up_write(&sb
->s_umount
);
408 EXPORT_SYMBOL(generic_shutdown_super
);
411 * sget - find or create a superblock
412 * @type: filesystem type superblock should belong to
413 * @test: comparison callback
414 * @set: setup callback
415 * @data: argument to each of them
417 struct super_block
*sget(struct file_system_type
*type
,
418 int (*test
)(struct super_block
*,void *),
419 int (*set
)(struct super_block
*,void *),
422 struct super_block
*s
= NULL
;
423 struct hlist_node
*node
;
424 struct super_block
*old
;
430 hlist_for_each_entry(old
, node
, &type
->fs_supers
, s_instances
) {
431 if (!test(old
, data
))
433 if (!grab_super(old
))
436 up_write(&s
->s_umount
);
440 down_write(&old
->s_umount
);
441 if (unlikely(!(old
->s_flags
& MS_BORN
))) {
442 deactivate_locked_super(old
);
449 spin_unlock(&sb_lock
);
450 s
= alloc_super(type
);
452 return ERR_PTR(-ENOMEM
);
458 spin_unlock(&sb_lock
);
459 up_write(&s
->s_umount
);
464 strlcpy(s
->s_id
, type
->name
, sizeof(s
->s_id
));
465 list_add_tail(&s
->s_list
, &super_blocks
);
466 hlist_add_head(&s
->s_instances
, &type
->fs_supers
);
467 spin_unlock(&sb_lock
);
468 get_filesystem(type
);
469 register_shrinker(&s
->s_shrink
);
475 void drop_super(struct super_block
*sb
)
477 up_read(&sb
->s_umount
);
481 EXPORT_SYMBOL(drop_super
);
484 * sync_supers - helper for periodic superblock writeback
486 * Call the write_super method if present on all dirty superblocks in
487 * the system. This is for the periodic writeback used by most older
488 * filesystems. For data integrity superblock writeback use
489 * sync_filesystems() instead.
491 * Note: check the dirty flag before waiting, so we don't
492 * hold up the sync while mounting a device. (The newly
493 * mounted device won't need syncing.)
495 void sync_supers(void)
497 struct super_block
*sb
, *p
= NULL
;
500 list_for_each_entry(sb
, &super_blocks
, s_list
) {
501 if (hlist_unhashed(&sb
->s_instances
))
503 if (sb
->s_op
->write_super
&& sb
->s_dirt
) {
505 spin_unlock(&sb_lock
);
507 down_read(&sb
->s_umount
);
508 if (sb
->s_root
&& sb
->s_dirt
&& (sb
->s_flags
& MS_BORN
))
509 sb
->s_op
->write_super(sb
);
510 up_read(&sb
->s_umount
);
520 spin_unlock(&sb_lock
);
524 * iterate_supers - call function for all active superblocks
525 * @f: function to call
526 * @arg: argument to pass to it
528 * Scans the superblock list and calls given function, passing it
529 * locked superblock and given argument.
531 void iterate_supers(void (*f
)(struct super_block
*, void *), void *arg
)
533 struct super_block
*sb
, *p
= NULL
;
536 list_for_each_entry(sb
, &super_blocks
, s_list
) {
537 if (hlist_unhashed(&sb
->s_instances
))
540 spin_unlock(&sb_lock
);
542 down_read(&sb
->s_umount
);
543 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
545 up_read(&sb
->s_umount
);
554 spin_unlock(&sb_lock
);
558 * iterate_supers_type - call function for superblocks of given type
560 * @f: function to call
561 * @arg: argument to pass to it
563 * Scans the superblock list and calls given function, passing it
564 * locked superblock and given argument.
566 void iterate_supers_type(struct file_system_type
*type
,
567 void (*f
)(struct super_block
*, void *), void *arg
)
569 struct super_block
*sb
, *p
= NULL
;
570 struct hlist_node
*node
;
573 hlist_for_each_entry(sb
, node
, &type
->fs_supers
, s_instances
) {
575 spin_unlock(&sb_lock
);
577 down_read(&sb
->s_umount
);
578 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
580 up_read(&sb
->s_umount
);
589 spin_unlock(&sb_lock
);
592 EXPORT_SYMBOL(iterate_supers_type
);
595 * get_super - get the superblock of a device
596 * @bdev: device to get the superblock for
598 * Scans the superblock list and finds the superblock of the file system
599 * mounted on the device given. %NULL is returned if no match is found.
602 struct super_block
*get_super(struct block_device
*bdev
)
604 struct super_block
*sb
;
611 list_for_each_entry(sb
, &super_blocks
, s_list
) {
612 if (hlist_unhashed(&sb
->s_instances
))
614 if (sb
->s_bdev
== bdev
) {
616 spin_unlock(&sb_lock
);
617 down_read(&sb
->s_umount
);
619 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
621 up_read(&sb
->s_umount
);
622 /* nope, got unmounted */
628 spin_unlock(&sb_lock
);
632 EXPORT_SYMBOL(get_super
);
635 * get_active_super - get an active reference to the superblock of a device
636 * @bdev: device to get the superblock for
638 * Scans the superblock list and finds the superblock of the file system
639 * mounted on the device given. Returns the superblock with an active
640 * reference or %NULL if none was found.
642 struct super_block
*get_active_super(struct block_device
*bdev
)
644 struct super_block
*sb
;
651 list_for_each_entry(sb
, &super_blocks
, s_list
) {
652 if (hlist_unhashed(&sb
->s_instances
))
654 if (sb
->s_bdev
== bdev
) {
655 if (grab_super(sb
)) /* drops sb_lock */
661 spin_unlock(&sb_lock
);
665 struct super_block
*user_get_super(dev_t dev
)
667 struct super_block
*sb
;
671 list_for_each_entry(sb
, &super_blocks
, s_list
) {
672 if (hlist_unhashed(&sb
->s_instances
))
674 if (sb
->s_dev
== dev
) {
676 spin_unlock(&sb_lock
);
677 down_read(&sb
->s_umount
);
679 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
681 up_read(&sb
->s_umount
);
682 /* nope, got unmounted */
688 spin_unlock(&sb_lock
);
693 * do_remount_sb - asks filesystem to change mount options.
694 * @sb: superblock in question
695 * @flags: numeric part of options
696 * @data: the rest of options
697 * @force: whether or not to force the change
699 * Alters the mount options of a mounted file system.
701 int do_remount_sb(struct super_block
*sb
, int flags
, void *data
, int force
)
706 if (sb
->s_frozen
!= SB_UNFROZEN
)
710 if (!(flags
& MS_RDONLY
) && bdev_read_only(sb
->s_bdev
))
714 if (flags
& MS_RDONLY
)
716 shrink_dcache_sb(sb
);
719 remount_ro
= (flags
& MS_RDONLY
) && !(sb
->s_flags
& MS_RDONLY
);
721 /* If we are remounting RDONLY and current sb is read/write,
722 make sure there are no rw files opened */
726 else if (!fs_may_remount_ro(sb
))
730 if (sb
->s_op
->remount_fs
) {
731 retval
= sb
->s_op
->remount_fs(sb
, &flags
, data
);
735 /* If forced remount, go ahead despite any errors */
736 WARN(1, "forced remount of a %s fs returned %i\n",
737 sb
->s_type
->name
, retval
);
740 sb
->s_flags
= (sb
->s_flags
& ~MS_RMT_MASK
) | (flags
& MS_RMT_MASK
);
743 * Some filesystems modify their metadata via some other path than the
744 * bdev buffer cache (eg. use a private mapping, or directories in
745 * pagecache, etc). Also file data modifications go via their own
746 * mappings. So If we try to mount readonly then copy the filesystem
747 * from bdev, we could get stale data, so invalidate it to give a best
748 * effort at coherency.
750 if (remount_ro
&& sb
->s_bdev
)
751 invalidate_bdev(sb
->s_bdev
);
755 static void do_emergency_remount(struct work_struct
*work
)
757 struct super_block
*sb
, *p
= NULL
;
760 list_for_each_entry(sb
, &super_blocks
, s_list
) {
761 if (hlist_unhashed(&sb
->s_instances
))
764 spin_unlock(&sb_lock
);
765 down_write(&sb
->s_umount
);
766 if (sb
->s_root
&& sb
->s_bdev
&& (sb
->s_flags
& MS_BORN
) &&
767 !(sb
->s_flags
& MS_RDONLY
)) {
769 * What lock protects sb->s_flags??
771 do_remount_sb(sb
, MS_RDONLY
, NULL
, 1);
773 up_write(&sb
->s_umount
);
781 spin_unlock(&sb_lock
);
783 printk("Emergency Remount complete\n");
786 void emergency_remount(void)
788 struct work_struct
*work
;
790 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
792 INIT_WORK(work
, do_emergency_remount
);
798 * Unnamed block devices are dummy devices used by virtual
799 * filesystems which don't use real block-devices. -- jrs
802 static DEFINE_IDA(unnamed_dev_ida
);
803 static DEFINE_SPINLOCK(unnamed_dev_lock
);/* protects the above */
804 static int unnamed_dev_start
= 0; /* don't bother trying below it */
806 int get_anon_bdev(dev_t
*p
)
812 if (ida_pre_get(&unnamed_dev_ida
, GFP_ATOMIC
) == 0)
814 spin_lock(&unnamed_dev_lock
);
815 error
= ida_get_new_above(&unnamed_dev_ida
, unnamed_dev_start
, &dev
);
817 unnamed_dev_start
= dev
+ 1;
818 spin_unlock(&unnamed_dev_lock
);
819 if (error
== -EAGAIN
)
820 /* We raced and lost with another CPU. */
825 if ((dev
& MAX_ID_MASK
) == (1 << MINORBITS
)) {
826 spin_lock(&unnamed_dev_lock
);
827 ida_remove(&unnamed_dev_ida
, dev
);
828 if (unnamed_dev_start
> dev
)
829 unnamed_dev_start
= dev
;
830 spin_unlock(&unnamed_dev_lock
);
833 *p
= MKDEV(0, dev
& MINORMASK
);
836 EXPORT_SYMBOL(get_anon_bdev
);
838 void free_anon_bdev(dev_t dev
)
840 int slot
= MINOR(dev
);
841 spin_lock(&unnamed_dev_lock
);
842 ida_remove(&unnamed_dev_ida
, slot
);
843 if (slot
< unnamed_dev_start
)
844 unnamed_dev_start
= slot
;
845 spin_unlock(&unnamed_dev_lock
);
847 EXPORT_SYMBOL(free_anon_bdev
);
849 int set_anon_super(struct super_block
*s
, void *data
)
851 int error
= get_anon_bdev(&s
->s_dev
);
853 s
->s_bdi
= &noop_backing_dev_info
;
857 EXPORT_SYMBOL(set_anon_super
);
859 void kill_anon_super(struct super_block
*sb
)
861 dev_t dev
= sb
->s_dev
;
862 generic_shutdown_super(sb
);
866 EXPORT_SYMBOL(kill_anon_super
);
868 void kill_litter_super(struct super_block
*sb
)
871 d_genocide(sb
->s_root
);
875 EXPORT_SYMBOL(kill_litter_super
);
877 static int ns_test_super(struct super_block
*sb
, void *data
)
879 return sb
->s_fs_info
== data
;
882 static int ns_set_super(struct super_block
*sb
, void *data
)
884 sb
->s_fs_info
= data
;
885 return set_anon_super(sb
, NULL
);
888 struct dentry
*mount_ns(struct file_system_type
*fs_type
, int flags
,
889 void *data
, int (*fill_super
)(struct super_block
*, void *, int))
891 struct super_block
*sb
;
893 sb
= sget(fs_type
, ns_test_super
, ns_set_super
, data
);
900 err
= fill_super(sb
, data
, flags
& MS_SILENT
? 1 : 0);
902 deactivate_locked_super(sb
);
906 sb
->s_flags
|= MS_ACTIVE
;
909 return dget(sb
->s_root
);
912 EXPORT_SYMBOL(mount_ns
);
915 static int set_bdev_super(struct super_block
*s
, void *data
)
918 s
->s_dev
= s
->s_bdev
->bd_dev
;
921 * We set the bdi here to the queue backing, file systems can
922 * overwrite this in ->fill_super()
924 s
->s_bdi
= &bdev_get_queue(s
->s_bdev
)->backing_dev_info
;
928 static int test_bdev_super(struct super_block
*s
, void *data
)
930 return (void *)s
->s_bdev
== data
;
933 struct dentry
*mount_bdev(struct file_system_type
*fs_type
,
934 int flags
, const char *dev_name
, void *data
,
935 int (*fill_super
)(struct super_block
*, void *, int))
937 struct block_device
*bdev
;
938 struct super_block
*s
;
939 fmode_t mode
= FMODE_READ
| FMODE_EXCL
;
942 if (!(flags
& MS_RDONLY
))
945 bdev
= blkdev_get_by_path(dev_name
, mode
, fs_type
);
947 return ERR_CAST(bdev
);
950 * once the super is inserted into the list by sget, s_umount
951 * will protect the lockfs code from trying to start a snapshot
952 * while we are mounting
954 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
955 if (bdev
->bd_fsfreeze_count
> 0) {
956 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
960 s
= sget(fs_type
, test_bdev_super
, set_bdev_super
, bdev
);
961 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
966 if ((flags
^ s
->s_flags
) & MS_RDONLY
) {
967 deactivate_locked_super(s
);
973 * s_umount nests inside bd_mutex during
974 * __invalidate_device(). blkdev_put() acquires
975 * bd_mutex and can't be called under s_umount. Drop
976 * s_umount temporarily. This is safe as we're
977 * holding an active reference.
979 up_write(&s
->s_umount
);
980 blkdev_put(bdev
, mode
);
981 down_write(&s
->s_umount
);
983 char b
[BDEVNAME_SIZE
];
985 s
->s_flags
= flags
| MS_NOSEC
;
987 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
988 sb_set_blocksize(s
, block_size(bdev
));
989 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
991 deactivate_locked_super(s
);
995 s
->s_flags
|= MS_ACTIVE
;
999 return dget(s
->s_root
);
1004 blkdev_put(bdev
, mode
);
1006 return ERR_PTR(error
);
1008 EXPORT_SYMBOL(mount_bdev
);
1010 void kill_block_super(struct super_block
*sb
)
1012 struct block_device
*bdev
= sb
->s_bdev
;
1013 fmode_t mode
= sb
->s_mode
;
1015 bdev
->bd_super
= NULL
;
1016 generic_shutdown_super(sb
);
1017 sync_blockdev(bdev
);
1018 WARN_ON_ONCE(!(mode
& FMODE_EXCL
));
1019 blkdev_put(bdev
, mode
| FMODE_EXCL
);
1022 EXPORT_SYMBOL(kill_block_super
);
1025 struct dentry
*mount_nodev(struct file_system_type
*fs_type
,
1026 int flags
, void *data
,
1027 int (*fill_super
)(struct super_block
*, void *, int))
1030 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
1037 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1039 deactivate_locked_super(s
);
1040 return ERR_PTR(error
);
1042 s
->s_flags
|= MS_ACTIVE
;
1043 return dget(s
->s_root
);
1045 EXPORT_SYMBOL(mount_nodev
);
1047 static int compare_single(struct super_block
*s
, void *p
)
1052 struct dentry
*mount_single(struct file_system_type
*fs_type
,
1053 int flags
, void *data
,
1054 int (*fill_super
)(struct super_block
*, void *, int))
1056 struct super_block
*s
;
1059 s
= sget(fs_type
, compare_single
, set_anon_super
, NULL
);
1064 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1066 deactivate_locked_super(s
);
1067 return ERR_PTR(error
);
1069 s
->s_flags
|= MS_ACTIVE
;
1071 do_remount_sb(s
, flags
, data
, 0);
1073 return dget(s
->s_root
);
1075 EXPORT_SYMBOL(mount_single
);
1078 mount_fs(struct file_system_type
*type
, int flags
, const char *name
, void *data
)
1080 struct dentry
*root
;
1081 struct super_block
*sb
;
1082 char *secdata
= NULL
;
1083 int error
= -ENOMEM
;
1085 if (data
&& !(type
->fs_flags
& FS_BINARY_MOUNTDATA
)) {
1086 secdata
= alloc_secdata();
1090 error
= security_sb_copy_data(data
, secdata
);
1092 goto out_free_secdata
;
1095 root
= type
->mount(type
, flags
, name
, data
);
1097 error
= PTR_ERR(root
);
1098 goto out_free_secdata
;
1102 WARN_ON(!sb
->s_bdi
);
1103 WARN_ON(sb
->s_bdi
== &default_backing_dev_info
);
1104 sb
->s_flags
|= MS_BORN
;
1106 error
= security_sb_kern_mount(sb
, flags
, secdata
);
1111 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
1112 * but s_maxbytes was an unsigned long long for many releases. Throw
1113 * this warning for a little while to try and catch filesystems that
1114 * violate this rule.
1116 WARN((sb
->s_maxbytes
< 0), "%s set sb->s_maxbytes to "
1117 "negative value (%lld)\n", type
->name
, sb
->s_maxbytes
);
1119 up_write(&sb
->s_umount
);
1120 free_secdata(secdata
);
1124 deactivate_locked_super(sb
);
1126 free_secdata(secdata
);
1128 return ERR_PTR(error
);
1132 * freeze_super - lock the filesystem and force it into a consistent state
1133 * @sb: the super to lock
1135 * Syncs the super to make sure the filesystem is consistent and calls the fs's
1136 * freeze_fs. Subsequent calls to this without first thawing the fs will return
1139 int freeze_super(struct super_block
*sb
)
1143 atomic_inc(&sb
->s_active
);
1144 down_write(&sb
->s_umount
);
1146 deactivate_locked_super(sb
);
1150 if (!(sb
->s_flags
& MS_BORN
)) {
1151 up_write(&sb
->s_umount
);
1152 return 0; /* sic - it's "nothing to do" */
1155 if (sb
->s_flags
& MS_RDONLY
) {
1156 sb
->s_frozen
= SB_FREEZE_TRANS
;
1158 up_write(&sb
->s_umount
);
1162 sb
->s_frozen
= SB_FREEZE_WRITE
;
1165 sync_filesystem(sb
);
1167 sb
->s_frozen
= SB_FREEZE_TRANS
;
1170 sync_blockdev(sb
->s_bdev
);
1171 if (sb
->s_op
->freeze_fs
) {
1172 ret
= sb
->s_op
->freeze_fs(sb
);
1175 "VFS:Filesystem freeze failed\n");
1176 sb
->s_frozen
= SB_UNFROZEN
;
1177 deactivate_locked_super(sb
);
1181 up_write(&sb
->s_umount
);
1184 EXPORT_SYMBOL(freeze_super
);
1187 * thaw_super -- unlock filesystem
1188 * @sb: the super to thaw
1190 * Unlocks the filesystem and marks it writeable again after freeze_super().
1192 int thaw_super(struct super_block
*sb
)
1196 down_write(&sb
->s_umount
);
1197 if (sb
->s_frozen
== SB_UNFROZEN
) {
1198 up_write(&sb
->s_umount
);
1202 if (sb
->s_flags
& MS_RDONLY
)
1205 if (sb
->s_op
->unfreeze_fs
) {
1206 error
= sb
->s_op
->unfreeze_fs(sb
);
1209 "VFS:Filesystem thaw failed\n");
1210 sb
->s_frozen
= SB_FREEZE_TRANS
;
1211 up_write(&sb
->s_umount
);
1217 sb
->s_frozen
= SB_UNFROZEN
;
1219 wake_up(&sb
->s_wait_unfrozen
);
1220 deactivate_locked_super(sb
);
1224 EXPORT_SYMBOL(thaw_super
);