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_LIST_HEAD(&s
->s_mounts
);
146 init_rwsem(&s
->s_umount
);
147 mutex_init(&s
->s_lock
);
148 lockdep_set_class(&s
->s_umount
, &type
->s_umount_key
);
150 * The locking rules for s_lock are up to the
151 * filesystem. For example ext3fs has different
152 * lock ordering than usbfs:
154 lockdep_set_class(&s
->s_lock
, &type
->s_lock_key
);
156 * sget() can have s_umount recursion.
158 * When it cannot find a suitable sb, it allocates a new
159 * one (this one), and tries again to find a suitable old
162 * In case that succeeds, it will acquire the s_umount
163 * lock of the old one. Since these are clearly distrinct
164 * locks, and this object isn't exposed yet, there's no
167 * Annotate this by putting this lock in a different
170 down_write_nested(&s
->s_umount
, SINGLE_DEPTH_NESTING
);
172 atomic_set(&s
->s_active
, 1);
173 mutex_init(&s
->s_vfs_rename_mutex
);
174 lockdep_set_class(&s
->s_vfs_rename_mutex
, &type
->s_vfs_rename_key
);
175 mutex_init(&s
->s_dquot
.dqio_mutex
);
176 mutex_init(&s
->s_dquot
.dqonoff_mutex
);
177 init_rwsem(&s
->s_dquot
.dqptr_sem
);
178 init_waitqueue_head(&s
->s_wait_unfrozen
);
179 s
->s_maxbytes
= MAX_NON_LFS
;
180 s
->s_op
= &default_op
;
181 s
->s_time_gran
= 1000000000;
182 s
->cleancache_poolid
= -1;
184 s
->s_shrink
.seeks
= DEFAULT_SEEKS
;
185 s
->s_shrink
.shrink
= prune_super
;
186 s
->s_shrink
.batch
= 1024;
193 * destroy_super - frees a superblock
194 * @s: superblock to free
196 * Frees a superblock.
198 static inline void destroy_super(struct super_block
*s
)
201 free_percpu(s
->s_files
);
204 WARN_ON(!list_empty(&s
->s_mounts
));
210 /* Superblock refcounting */
213 * Drop a superblock's refcount. The caller must hold sb_lock.
215 static void __put_super(struct super_block
*sb
)
217 if (!--sb
->s_count
) {
218 list_del_init(&sb
->s_list
);
224 * put_super - drop a temporary reference to superblock
225 * @sb: superblock in question
227 * Drops a temporary reference, frees superblock if there's no
230 static void put_super(struct super_block
*sb
)
234 spin_unlock(&sb_lock
);
239 * deactivate_locked_super - drop an active reference to superblock
240 * @s: superblock to deactivate
242 * Drops an active reference to superblock, converting it into a temprory
243 * one if there is no other active references left. In that case we
244 * tell fs driver to shut it down and drop the temporary reference we
247 * Caller holds exclusive lock on superblock; that lock is released.
249 void deactivate_locked_super(struct super_block
*s
)
251 struct file_system_type
*fs
= s
->s_type
;
252 if (atomic_dec_and_test(&s
->s_active
)) {
253 cleancache_flush_fs(s
);
256 /* caches are now gone, we can safely kill the shrinker now */
257 unregister_shrinker(&s
->s_shrink
);
260 * We need to call rcu_barrier so all the delayed rcu free
261 * inodes are flushed before we release the fs module.
267 up_write(&s
->s_umount
);
271 EXPORT_SYMBOL(deactivate_locked_super
);
274 * deactivate_super - drop an active reference to superblock
275 * @s: superblock to deactivate
277 * Variant of deactivate_locked_super(), except that superblock is *not*
278 * locked by caller. If we are going to drop the final active reference,
279 * lock will be acquired prior to that.
281 void deactivate_super(struct super_block
*s
)
283 if (!atomic_add_unless(&s
->s_active
, -1, 1)) {
284 down_write(&s
->s_umount
);
285 deactivate_locked_super(s
);
289 EXPORT_SYMBOL(deactivate_super
);
292 * grab_super - acquire an active reference
293 * @s: reference we are trying to make active
295 * Tries to acquire an active reference. grab_super() is used when we
296 * had just found a superblock in super_blocks or fs_type->fs_supers
297 * and want to turn it into a full-blown active reference. grab_super()
298 * is called with sb_lock held and drops it. Returns 1 in case of
299 * success, 0 if we had failed (superblock contents was already dead or
300 * dying when grab_super() had been called).
302 static int grab_super(struct super_block
*s
) __releases(sb_lock
)
304 if (atomic_inc_not_zero(&s
->s_active
)) {
305 spin_unlock(&sb_lock
);
308 /* it's going away */
310 spin_unlock(&sb_lock
);
311 /* wait for it to die */
312 down_write(&s
->s_umount
);
313 up_write(&s
->s_umount
);
319 * grab_super_passive - acquire a passive reference
320 * @s: reference we are trying to grab
322 * Tries to acquire a passive reference. This is used in places where we
323 * cannot take an active reference but we need to ensure that the
324 * superblock does not go away while we are working on it. It returns
325 * false if a reference was not gained, and returns true with the s_umount
326 * lock held in read mode if a reference is gained. On successful return,
327 * the caller must drop the s_umount lock and the passive reference when
330 bool grab_super_passive(struct super_block
*sb
)
333 if (hlist_unhashed(&sb
->s_instances
)) {
334 spin_unlock(&sb_lock
);
339 spin_unlock(&sb_lock
);
341 if (down_read_trylock(&sb
->s_umount
)) {
342 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
344 up_read(&sb
->s_umount
);
352 * Superblock locking. We really ought to get rid of these two.
354 void lock_super(struct super_block
* sb
)
356 mutex_lock(&sb
->s_lock
);
359 void unlock_super(struct super_block
* sb
)
361 mutex_unlock(&sb
->s_lock
);
364 EXPORT_SYMBOL(lock_super
);
365 EXPORT_SYMBOL(unlock_super
);
368 * generic_shutdown_super - common helper for ->kill_sb()
369 * @sb: superblock to kill
371 * generic_shutdown_super() does all fs-independent work on superblock
372 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
373 * that need destruction out of superblock, call generic_shutdown_super()
374 * and release aforementioned objects. Note: dentries and inodes _are_
375 * taken care of and do not need specific handling.
377 * Upon calling this function, the filesystem may no longer alter or
378 * rearrange the set of dentries belonging to this super_block, nor may it
379 * change the attachments of dentries to inodes.
381 void generic_shutdown_super(struct super_block
*sb
)
383 const struct super_operations
*sop
= sb
->s_op
;
386 shrink_dcache_for_umount(sb
);
388 sb
->s_flags
&= ~MS_ACTIVE
;
390 fsnotify_unmount_inodes(&sb
->s_inodes
);
397 if (!list_empty(&sb
->s_inodes
)) {
398 printk("VFS: Busy inodes after unmount of %s. "
399 "Self-destruct in 5 seconds. Have a nice day...\n",
404 /* should be initialized for __put_super_and_need_restart() */
405 hlist_del_init(&sb
->s_instances
);
406 spin_unlock(&sb_lock
);
407 up_write(&sb
->s_umount
);
410 EXPORT_SYMBOL(generic_shutdown_super
);
413 * sget - find or create a superblock
414 * @type: filesystem type superblock should belong to
415 * @test: comparison callback
416 * @set: setup callback
417 * @data: argument to each of them
419 struct super_block
*sget(struct file_system_type
*type
,
420 int (*test
)(struct super_block
*,void *),
421 int (*set
)(struct super_block
*,void *),
424 struct super_block
*s
= NULL
;
425 struct hlist_node
*node
;
426 struct super_block
*old
;
432 hlist_for_each_entry(old
, node
, &type
->fs_supers
, s_instances
) {
433 if (!test(old
, data
))
435 if (!grab_super(old
))
438 up_write(&s
->s_umount
);
442 down_write(&old
->s_umount
);
443 if (unlikely(!(old
->s_flags
& MS_BORN
))) {
444 deactivate_locked_super(old
);
451 spin_unlock(&sb_lock
);
452 s
= alloc_super(type
);
454 return ERR_PTR(-ENOMEM
);
460 spin_unlock(&sb_lock
);
461 up_write(&s
->s_umount
);
466 strlcpy(s
->s_id
, type
->name
, sizeof(s
->s_id
));
467 list_add_tail(&s
->s_list
, &super_blocks
);
468 hlist_add_head(&s
->s_instances
, &type
->fs_supers
);
469 spin_unlock(&sb_lock
);
470 get_filesystem(type
);
471 register_shrinker(&s
->s_shrink
);
477 void drop_super(struct super_block
*sb
)
479 up_read(&sb
->s_umount
);
483 EXPORT_SYMBOL(drop_super
);
486 * sync_supers - helper for periodic superblock writeback
488 * Call the write_super method if present on all dirty superblocks in
489 * the system. This is for the periodic writeback used by most older
490 * filesystems. For data integrity superblock writeback use
491 * sync_filesystems() instead.
493 * Note: check the dirty flag before waiting, so we don't
494 * hold up the sync while mounting a device. (The newly
495 * mounted device won't need syncing.)
497 void sync_supers(void)
499 struct super_block
*sb
, *p
= NULL
;
502 list_for_each_entry(sb
, &super_blocks
, s_list
) {
503 if (hlist_unhashed(&sb
->s_instances
))
505 if (sb
->s_op
->write_super
&& sb
->s_dirt
) {
507 spin_unlock(&sb_lock
);
509 down_read(&sb
->s_umount
);
510 if (sb
->s_root
&& sb
->s_dirt
&& (sb
->s_flags
& MS_BORN
))
511 sb
->s_op
->write_super(sb
);
512 up_read(&sb
->s_umount
);
522 spin_unlock(&sb_lock
);
526 * iterate_supers - call function for all active superblocks
527 * @f: function to call
528 * @arg: argument to pass to it
530 * Scans the superblock list and calls given function, passing it
531 * locked superblock and given argument.
533 void iterate_supers(void (*f
)(struct super_block
*, void *), void *arg
)
535 struct super_block
*sb
, *p
= NULL
;
538 list_for_each_entry(sb
, &super_blocks
, s_list
) {
539 if (hlist_unhashed(&sb
->s_instances
))
542 spin_unlock(&sb_lock
);
544 down_read(&sb
->s_umount
);
545 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
547 up_read(&sb
->s_umount
);
556 spin_unlock(&sb_lock
);
560 * iterate_supers_type - call function for superblocks of given type
562 * @f: function to call
563 * @arg: argument to pass to it
565 * Scans the superblock list and calls given function, passing it
566 * locked superblock and given argument.
568 void iterate_supers_type(struct file_system_type
*type
,
569 void (*f
)(struct super_block
*, void *), void *arg
)
571 struct super_block
*sb
, *p
= NULL
;
572 struct hlist_node
*node
;
575 hlist_for_each_entry(sb
, node
, &type
->fs_supers
, s_instances
) {
577 spin_unlock(&sb_lock
);
579 down_read(&sb
->s_umount
);
580 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
582 up_read(&sb
->s_umount
);
591 spin_unlock(&sb_lock
);
594 EXPORT_SYMBOL(iterate_supers_type
);
597 * get_super - get the superblock of a device
598 * @bdev: device to get the superblock for
600 * Scans the superblock list and finds the superblock of the file system
601 * mounted on the device given. %NULL is returned if no match is found.
604 struct super_block
*get_super(struct block_device
*bdev
)
606 struct super_block
*sb
;
613 list_for_each_entry(sb
, &super_blocks
, s_list
) {
614 if (hlist_unhashed(&sb
->s_instances
))
616 if (sb
->s_bdev
== bdev
) {
618 spin_unlock(&sb_lock
);
619 down_read(&sb
->s_umount
);
621 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
623 up_read(&sb
->s_umount
);
624 /* nope, got unmounted */
630 spin_unlock(&sb_lock
);
634 EXPORT_SYMBOL(get_super
);
637 * get_active_super - get an active reference to the superblock of a device
638 * @bdev: device to get the superblock for
640 * Scans the superblock list and finds the superblock of the file system
641 * mounted on the device given. Returns the superblock with an active
642 * reference or %NULL if none was found.
644 struct super_block
*get_active_super(struct block_device
*bdev
)
646 struct super_block
*sb
;
653 list_for_each_entry(sb
, &super_blocks
, s_list
) {
654 if (hlist_unhashed(&sb
->s_instances
))
656 if (sb
->s_bdev
== bdev
) {
657 if (grab_super(sb
)) /* drops sb_lock */
663 spin_unlock(&sb_lock
);
667 struct super_block
*user_get_super(dev_t dev
)
669 struct super_block
*sb
;
673 list_for_each_entry(sb
, &super_blocks
, s_list
) {
674 if (hlist_unhashed(&sb
->s_instances
))
676 if (sb
->s_dev
== dev
) {
678 spin_unlock(&sb_lock
);
679 down_read(&sb
->s_umount
);
681 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
683 up_read(&sb
->s_umount
);
684 /* nope, got unmounted */
690 spin_unlock(&sb_lock
);
695 * do_remount_sb - asks filesystem to change mount options.
696 * @sb: superblock in question
697 * @flags: numeric part of options
698 * @data: the rest of options
699 * @force: whether or not to force the change
701 * Alters the mount options of a mounted file system.
703 int do_remount_sb(struct super_block
*sb
, int flags
, void *data
, int force
)
708 if (sb
->s_frozen
!= SB_UNFROZEN
)
712 if (!(flags
& MS_RDONLY
) && bdev_read_only(sb
->s_bdev
))
716 if (flags
& MS_RDONLY
)
718 shrink_dcache_sb(sb
);
721 remount_ro
= (flags
& MS_RDONLY
) && !(sb
->s_flags
& MS_RDONLY
);
723 /* If we are remounting RDONLY and current sb is read/write,
724 make sure there are no rw files opened */
729 retval
= sb_prepare_remount_readonly(sb
);
735 if (sb
->s_op
->remount_fs
) {
736 retval
= sb
->s_op
->remount_fs(sb
, &flags
, data
);
739 goto cancel_readonly
;
740 /* If forced remount, go ahead despite any errors */
741 WARN(1, "forced remount of a %s fs returned %i\n",
742 sb
->s_type
->name
, retval
);
745 sb
->s_flags
= (sb
->s_flags
& ~MS_RMT_MASK
) | (flags
& MS_RMT_MASK
);
746 /* Needs to be ordered wrt mnt_is_readonly() */
748 sb
->s_readonly_remount
= 0;
751 * Some filesystems modify their metadata via some other path than the
752 * bdev buffer cache (eg. use a private mapping, or directories in
753 * pagecache, etc). Also file data modifications go via their own
754 * mappings. So If we try to mount readonly then copy the filesystem
755 * from bdev, we could get stale data, so invalidate it to give a best
756 * effort at coherency.
758 if (remount_ro
&& sb
->s_bdev
)
759 invalidate_bdev(sb
->s_bdev
);
763 sb
->s_readonly_remount
= 0;
767 static void do_emergency_remount(struct work_struct
*work
)
769 struct super_block
*sb
, *p
= NULL
;
772 list_for_each_entry(sb
, &super_blocks
, s_list
) {
773 if (hlist_unhashed(&sb
->s_instances
))
776 spin_unlock(&sb_lock
);
777 down_write(&sb
->s_umount
);
778 if (sb
->s_root
&& sb
->s_bdev
&& (sb
->s_flags
& MS_BORN
) &&
779 !(sb
->s_flags
& MS_RDONLY
)) {
781 * What lock protects sb->s_flags??
783 do_remount_sb(sb
, MS_RDONLY
, NULL
, 1);
785 up_write(&sb
->s_umount
);
793 spin_unlock(&sb_lock
);
795 printk("Emergency Remount complete\n");
798 void emergency_remount(void)
800 struct work_struct
*work
;
802 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
804 INIT_WORK(work
, do_emergency_remount
);
810 * Unnamed block devices are dummy devices used by virtual
811 * filesystems which don't use real block-devices. -- jrs
814 static DEFINE_IDA(unnamed_dev_ida
);
815 static DEFINE_SPINLOCK(unnamed_dev_lock
);/* protects the above */
816 static int unnamed_dev_start
= 0; /* don't bother trying below it */
818 int get_anon_bdev(dev_t
*p
)
824 if (ida_pre_get(&unnamed_dev_ida
, GFP_ATOMIC
) == 0)
826 spin_lock(&unnamed_dev_lock
);
827 error
= ida_get_new_above(&unnamed_dev_ida
, unnamed_dev_start
, &dev
);
829 unnamed_dev_start
= dev
+ 1;
830 spin_unlock(&unnamed_dev_lock
);
831 if (error
== -EAGAIN
)
832 /* We raced and lost with another CPU. */
837 if ((dev
& MAX_ID_MASK
) == (1 << MINORBITS
)) {
838 spin_lock(&unnamed_dev_lock
);
839 ida_remove(&unnamed_dev_ida
, dev
);
840 if (unnamed_dev_start
> dev
)
841 unnamed_dev_start
= dev
;
842 spin_unlock(&unnamed_dev_lock
);
845 *p
= MKDEV(0, dev
& MINORMASK
);
848 EXPORT_SYMBOL(get_anon_bdev
);
850 void free_anon_bdev(dev_t dev
)
852 int slot
= MINOR(dev
);
853 spin_lock(&unnamed_dev_lock
);
854 ida_remove(&unnamed_dev_ida
, slot
);
855 if (slot
< unnamed_dev_start
)
856 unnamed_dev_start
= slot
;
857 spin_unlock(&unnamed_dev_lock
);
859 EXPORT_SYMBOL(free_anon_bdev
);
861 int set_anon_super(struct super_block
*s
, void *data
)
863 int error
= get_anon_bdev(&s
->s_dev
);
865 s
->s_bdi
= &noop_backing_dev_info
;
869 EXPORT_SYMBOL(set_anon_super
);
871 void kill_anon_super(struct super_block
*sb
)
873 dev_t dev
= sb
->s_dev
;
874 generic_shutdown_super(sb
);
878 EXPORT_SYMBOL(kill_anon_super
);
880 void kill_litter_super(struct super_block
*sb
)
883 d_genocide(sb
->s_root
);
887 EXPORT_SYMBOL(kill_litter_super
);
889 static int ns_test_super(struct super_block
*sb
, void *data
)
891 return sb
->s_fs_info
== data
;
894 static int ns_set_super(struct super_block
*sb
, void *data
)
896 sb
->s_fs_info
= data
;
897 return set_anon_super(sb
, NULL
);
900 struct dentry
*mount_ns(struct file_system_type
*fs_type
, int flags
,
901 void *data
, int (*fill_super
)(struct super_block
*, void *, int))
903 struct super_block
*sb
;
905 sb
= sget(fs_type
, ns_test_super
, ns_set_super
, data
);
912 err
= fill_super(sb
, data
, flags
& MS_SILENT
? 1 : 0);
914 deactivate_locked_super(sb
);
918 sb
->s_flags
|= MS_ACTIVE
;
921 return dget(sb
->s_root
);
924 EXPORT_SYMBOL(mount_ns
);
927 static int set_bdev_super(struct super_block
*s
, void *data
)
930 s
->s_dev
= s
->s_bdev
->bd_dev
;
933 * We set the bdi here to the queue backing, file systems can
934 * overwrite this in ->fill_super()
936 s
->s_bdi
= &bdev_get_queue(s
->s_bdev
)->backing_dev_info
;
940 static int test_bdev_super(struct super_block
*s
, void *data
)
942 return (void *)s
->s_bdev
== data
;
945 struct dentry
*mount_bdev(struct file_system_type
*fs_type
,
946 int flags
, const char *dev_name
, void *data
,
947 int (*fill_super
)(struct super_block
*, void *, int))
949 struct block_device
*bdev
;
950 struct super_block
*s
;
951 fmode_t mode
= FMODE_READ
| FMODE_EXCL
;
954 if (!(flags
& MS_RDONLY
))
957 bdev
= blkdev_get_by_path(dev_name
, mode
, fs_type
);
959 return ERR_CAST(bdev
);
962 * once the super is inserted into the list by sget, s_umount
963 * will protect the lockfs code from trying to start a snapshot
964 * while we are mounting
966 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
967 if (bdev
->bd_fsfreeze_count
> 0) {
968 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
972 s
= sget(fs_type
, test_bdev_super
, set_bdev_super
, bdev
);
973 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
978 if ((flags
^ s
->s_flags
) & MS_RDONLY
) {
979 deactivate_locked_super(s
);
985 * s_umount nests inside bd_mutex during
986 * __invalidate_device(). blkdev_put() acquires
987 * bd_mutex and can't be called under s_umount. Drop
988 * s_umount temporarily. This is safe as we're
989 * holding an active reference.
991 up_write(&s
->s_umount
);
992 blkdev_put(bdev
, mode
);
993 down_write(&s
->s_umount
);
995 char b
[BDEVNAME_SIZE
];
997 s
->s_flags
= flags
| MS_NOSEC
;
999 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
1000 sb_set_blocksize(s
, block_size(bdev
));
1001 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1003 deactivate_locked_super(s
);
1007 s
->s_flags
|= MS_ACTIVE
;
1011 return dget(s
->s_root
);
1016 blkdev_put(bdev
, mode
);
1018 return ERR_PTR(error
);
1020 EXPORT_SYMBOL(mount_bdev
);
1022 void kill_block_super(struct super_block
*sb
)
1024 struct block_device
*bdev
= sb
->s_bdev
;
1025 fmode_t mode
= sb
->s_mode
;
1027 bdev
->bd_super
= NULL
;
1028 generic_shutdown_super(sb
);
1029 sync_blockdev(bdev
);
1030 WARN_ON_ONCE(!(mode
& FMODE_EXCL
));
1031 blkdev_put(bdev
, mode
| FMODE_EXCL
);
1034 EXPORT_SYMBOL(kill_block_super
);
1037 struct dentry
*mount_nodev(struct file_system_type
*fs_type
,
1038 int flags
, void *data
,
1039 int (*fill_super
)(struct super_block
*, void *, int))
1042 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
1049 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1051 deactivate_locked_super(s
);
1052 return ERR_PTR(error
);
1054 s
->s_flags
|= MS_ACTIVE
;
1055 return dget(s
->s_root
);
1057 EXPORT_SYMBOL(mount_nodev
);
1059 static int compare_single(struct super_block
*s
, void *p
)
1064 struct dentry
*mount_single(struct file_system_type
*fs_type
,
1065 int flags
, void *data
,
1066 int (*fill_super
)(struct super_block
*, void *, int))
1068 struct super_block
*s
;
1071 s
= sget(fs_type
, compare_single
, set_anon_super
, NULL
);
1076 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1078 deactivate_locked_super(s
);
1079 return ERR_PTR(error
);
1081 s
->s_flags
|= MS_ACTIVE
;
1083 do_remount_sb(s
, flags
, data
, 0);
1085 return dget(s
->s_root
);
1087 EXPORT_SYMBOL(mount_single
);
1090 mount_fs(struct file_system_type
*type
, int flags
, const char *name
, void *data
)
1092 struct dentry
*root
;
1093 struct super_block
*sb
;
1094 char *secdata
= NULL
;
1095 int error
= -ENOMEM
;
1097 if (data
&& !(type
->fs_flags
& FS_BINARY_MOUNTDATA
)) {
1098 secdata
= alloc_secdata();
1102 error
= security_sb_copy_data(data
, secdata
);
1104 goto out_free_secdata
;
1107 root
= type
->mount(type
, flags
, name
, data
);
1109 error
= PTR_ERR(root
);
1110 goto out_free_secdata
;
1114 WARN_ON(!sb
->s_bdi
);
1115 WARN_ON(sb
->s_bdi
== &default_backing_dev_info
);
1116 sb
->s_flags
|= MS_BORN
;
1118 error
= security_sb_kern_mount(sb
, flags
, secdata
);
1123 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
1124 * but s_maxbytes was an unsigned long long for many releases. Throw
1125 * this warning for a little while to try and catch filesystems that
1126 * violate this rule.
1128 WARN((sb
->s_maxbytes
< 0), "%s set sb->s_maxbytes to "
1129 "negative value (%lld)\n", type
->name
, sb
->s_maxbytes
);
1131 up_write(&sb
->s_umount
);
1132 free_secdata(secdata
);
1136 deactivate_locked_super(sb
);
1138 free_secdata(secdata
);
1140 return ERR_PTR(error
);
1144 * freeze_super - lock the filesystem and force it into a consistent state
1145 * @sb: the super to lock
1147 * Syncs the super to make sure the filesystem is consistent and calls the fs's
1148 * freeze_fs. Subsequent calls to this without first thawing the fs will return
1151 int freeze_super(struct super_block
*sb
)
1155 atomic_inc(&sb
->s_active
);
1156 down_write(&sb
->s_umount
);
1158 deactivate_locked_super(sb
);
1162 if (!(sb
->s_flags
& MS_BORN
)) {
1163 up_write(&sb
->s_umount
);
1164 return 0; /* sic - it's "nothing to do" */
1167 if (sb
->s_flags
& MS_RDONLY
) {
1168 sb
->s_frozen
= SB_FREEZE_TRANS
;
1170 up_write(&sb
->s_umount
);
1174 sb
->s_frozen
= SB_FREEZE_WRITE
;
1177 sync_filesystem(sb
);
1179 sb
->s_frozen
= SB_FREEZE_TRANS
;
1182 sync_blockdev(sb
->s_bdev
);
1183 if (sb
->s_op
->freeze_fs
) {
1184 ret
= sb
->s_op
->freeze_fs(sb
);
1187 "VFS:Filesystem freeze failed\n");
1188 sb
->s_frozen
= SB_UNFROZEN
;
1190 wake_up(&sb
->s_wait_unfrozen
);
1191 deactivate_locked_super(sb
);
1195 up_write(&sb
->s_umount
);
1198 EXPORT_SYMBOL(freeze_super
);
1201 * thaw_super -- unlock filesystem
1202 * @sb: the super to thaw
1204 * Unlocks the filesystem and marks it writeable again after freeze_super().
1206 int thaw_super(struct super_block
*sb
)
1210 down_write(&sb
->s_umount
);
1211 if (sb
->s_frozen
== SB_UNFROZEN
) {
1212 up_write(&sb
->s_umount
);
1216 if (sb
->s_flags
& MS_RDONLY
)
1219 if (sb
->s_op
->unfreeze_fs
) {
1220 error
= sb
->s_op
->unfreeze_fs(sb
);
1223 "VFS:Filesystem thaw failed\n");
1224 sb
->s_frozen
= SB_FREEZE_TRANS
;
1225 up_write(&sb
->s_umount
);
1231 sb
->s_frozen
= SB_UNFROZEN
;
1233 wake_up(&sb
->s_wait_unfrozen
);
1234 deactivate_locked_super(sb
);
1238 EXPORT_SYMBOL(thaw_super
);