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>
35 #include <linux/fsnotify.h>
39 LIST_HEAD(super_blocks
);
40 DEFINE_SPINLOCK(sb_lock
);
43 * One thing we have to be careful of with a per-sb shrinker is that we don't
44 * drop the last active reference to the superblock from within the shrinker.
45 * If that happens we could trigger unregistering the shrinker from within the
46 * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we
47 * take a passive reference to the superblock to avoid this from occurring.
49 static int prune_super(struct shrinker
*shrink
, struct shrink_control
*sc
)
51 struct super_block
*sb
;
55 sb
= container_of(shrink
, struct super_block
, s_shrink
);
58 * Deadlock avoidance. We may hold various FS locks, and we don't want
59 * to recurse into the FS that called us in clear_inode() and friends..
61 if (sc
->nr_to_scan
&& !(sc
->gfp_mask
& __GFP_FS
))
64 if (!grab_super_passive(sb
))
65 return !sc
->nr_to_scan
? 0 : -1;
67 if (sb
->s_op
&& sb
->s_op
->nr_cached_objects
)
68 fs_objects
= sb
->s_op
->nr_cached_objects(sb
);
70 total_objects
= sb
->s_nr_dentry_unused
+
71 sb
->s_nr_inodes_unused
+ fs_objects
+ 1;
77 /* proportion the scan between the caches */
78 dentries
= (sc
->nr_to_scan
* sb
->s_nr_dentry_unused
) /
80 inodes
= (sc
->nr_to_scan
* sb
->s_nr_inodes_unused
) /
83 fs_objects
= (sc
->nr_to_scan
* fs_objects
) /
86 * prune the dcache first as the icache is pinned by it, then
87 * prune the icache, followed by the filesystem specific caches
89 prune_dcache_sb(sb
, dentries
);
90 prune_icache_sb(sb
, inodes
);
92 if (fs_objects
&& sb
->s_op
->free_cached_objects
) {
93 sb
->s_op
->free_cached_objects(sb
, fs_objects
);
94 fs_objects
= sb
->s_op
->nr_cached_objects(sb
);
96 total_objects
= sb
->s_nr_dentry_unused
+
97 sb
->s_nr_inodes_unused
+ fs_objects
;
100 total_objects
= (total_objects
/ 100) * sysctl_vfs_cache_pressure
;
102 return total_objects
;
106 * alloc_super - create new superblock
107 * @type: filesystem type superblock should belong to
109 * Allocates and initializes a new &struct super_block. alloc_super()
110 * returns a pointer new superblock or %NULL if allocation had failed.
112 static struct super_block
*alloc_super(struct file_system_type
*type
)
114 struct super_block
*s
= kzalloc(sizeof(struct super_block
), GFP_USER
);
115 static const struct super_operations default_op
;
118 if (security_sb_alloc(s
)) {
124 s
->s_files
= alloc_percpu(struct list_head
);
133 for_each_possible_cpu(i
)
134 INIT_LIST_HEAD(per_cpu_ptr(s
->s_files
, i
));
137 INIT_LIST_HEAD(&s
->s_files
);
139 s
->s_bdi
= &default_backing_dev_info
;
140 INIT_HLIST_NODE(&s
->s_instances
);
141 INIT_HLIST_BL_HEAD(&s
->s_anon
);
142 INIT_LIST_HEAD(&s
->s_inodes
);
143 INIT_LIST_HEAD(&s
->s_dentry_lru
);
144 INIT_LIST_HEAD(&s
->s_inode_lru
);
145 spin_lock_init(&s
->s_inode_lru_lock
);
146 INIT_LIST_HEAD(&s
->s_mounts
);
147 init_rwsem(&s
->s_umount
);
148 mutex_init(&s
->s_lock
);
149 lockdep_set_class(&s
->s_umount
, &type
->s_umount_key
);
151 * The locking rules for s_lock are up to the
152 * filesystem. For example ext3fs has different
153 * lock ordering than usbfs:
155 lockdep_set_class(&s
->s_lock
, &type
->s_lock_key
);
157 * sget() can have s_umount recursion.
159 * When it cannot find a suitable sb, it allocates a new
160 * one (this one), and tries again to find a suitable old
163 * In case that succeeds, it will acquire the s_umount
164 * lock of the old one. Since these are clearly distrinct
165 * locks, and this object isn't exposed yet, there's no
168 * Annotate this by putting this lock in a different
171 down_write_nested(&s
->s_umount
, SINGLE_DEPTH_NESTING
);
173 atomic_set(&s
->s_active
, 1);
174 mutex_init(&s
->s_vfs_rename_mutex
);
175 lockdep_set_class(&s
->s_vfs_rename_mutex
, &type
->s_vfs_rename_key
);
176 mutex_init(&s
->s_dquot
.dqio_mutex
);
177 mutex_init(&s
->s_dquot
.dqonoff_mutex
);
178 init_rwsem(&s
->s_dquot
.dqptr_sem
);
179 init_waitqueue_head(&s
->s_wait_unfrozen
);
180 s
->s_maxbytes
= MAX_NON_LFS
;
181 s
->s_op
= &default_op
;
182 s
->s_time_gran
= 1000000000;
183 s
->cleancache_poolid
= -1;
185 s
->s_shrink
.seeks
= DEFAULT_SEEKS
;
186 s
->s_shrink
.shrink
= prune_super
;
187 s
->s_shrink
.batch
= 1024;
194 * destroy_super - frees a superblock
195 * @s: superblock to free
197 * Frees a superblock.
199 static inline void destroy_super(struct super_block
*s
)
202 free_percpu(s
->s_files
);
205 WARN_ON(!list_empty(&s
->s_mounts
));
211 /* Superblock refcounting */
214 * Drop a superblock's refcount. The caller must hold sb_lock.
216 static void __put_super(struct super_block
*sb
)
218 if (!--sb
->s_count
) {
219 list_del_init(&sb
->s_list
);
225 * put_super - drop a temporary reference to superblock
226 * @sb: superblock in question
228 * Drops a temporary reference, frees superblock if there's no
231 static void put_super(struct super_block
*sb
)
235 spin_unlock(&sb_lock
);
240 * deactivate_locked_super - drop an active reference to superblock
241 * @s: superblock to deactivate
243 * Drops an active reference to superblock, converting it into a temprory
244 * one if there is no other active references left. In that case we
245 * tell fs driver to shut it down and drop the temporary reference we
248 * Caller holds exclusive lock on superblock; that lock is released.
250 void deactivate_locked_super(struct super_block
*s
)
252 struct file_system_type
*fs
= s
->s_type
;
253 if (atomic_dec_and_test(&s
->s_active
)) {
254 cleancache_flush_fs(s
);
257 /* caches are now gone, we can safely kill the shrinker now */
258 unregister_shrinker(&s
->s_shrink
);
261 * We need to call rcu_barrier so all the delayed rcu free
262 * inodes are flushed before we release the fs module.
268 up_write(&s
->s_umount
);
272 EXPORT_SYMBOL(deactivate_locked_super
);
275 * deactivate_super - drop an active reference to superblock
276 * @s: superblock to deactivate
278 * Variant of deactivate_locked_super(), except that superblock is *not*
279 * locked by caller. If we are going to drop the final active reference,
280 * lock will be acquired prior to that.
282 void deactivate_super(struct super_block
*s
)
284 if (!atomic_add_unless(&s
->s_active
, -1, 1)) {
285 down_write(&s
->s_umount
);
286 deactivate_locked_super(s
);
290 EXPORT_SYMBOL(deactivate_super
);
293 * grab_super - acquire an active reference
294 * @s: reference we are trying to make active
296 * Tries to acquire an active reference. grab_super() is used when we
297 * had just found a superblock in super_blocks or fs_type->fs_supers
298 * and want to turn it into a full-blown active reference. grab_super()
299 * is called with sb_lock held and drops it. Returns 1 in case of
300 * success, 0 if we had failed (superblock contents was already dead or
301 * dying when grab_super() had been called).
303 static int grab_super(struct super_block
*s
) __releases(sb_lock
)
305 if (atomic_inc_not_zero(&s
->s_active
)) {
306 spin_unlock(&sb_lock
);
309 /* it's going away */
311 spin_unlock(&sb_lock
);
312 /* wait for it to die */
313 down_write(&s
->s_umount
);
314 up_write(&s
->s_umount
);
320 * grab_super_passive - acquire a passive reference
321 * @s: reference we are trying to grab
323 * Tries to acquire a passive reference. This is used in places where we
324 * cannot take an active reference but we need to ensure that the
325 * superblock does not go away while we are working on it. It returns
326 * false if a reference was not gained, and returns true with the s_umount
327 * lock held in read mode if a reference is gained. On successful return,
328 * the caller must drop the s_umount lock and the passive reference when
331 bool grab_super_passive(struct super_block
*sb
)
334 if (hlist_unhashed(&sb
->s_instances
)) {
335 spin_unlock(&sb_lock
);
340 spin_unlock(&sb_lock
);
342 if (down_read_trylock(&sb
->s_umount
)) {
343 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
345 up_read(&sb
->s_umount
);
353 * Superblock locking. We really ought to get rid of these two.
355 void lock_super(struct super_block
* sb
)
357 mutex_lock(&sb
->s_lock
);
360 void unlock_super(struct super_block
* sb
)
362 mutex_unlock(&sb
->s_lock
);
365 EXPORT_SYMBOL(lock_super
);
366 EXPORT_SYMBOL(unlock_super
);
369 * generic_shutdown_super - common helper for ->kill_sb()
370 * @sb: superblock to kill
372 * generic_shutdown_super() does all fs-independent work on superblock
373 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
374 * that need destruction out of superblock, call generic_shutdown_super()
375 * and release aforementioned objects. Note: dentries and inodes _are_
376 * taken care of and do not need specific handling.
378 * Upon calling this function, the filesystem may no longer alter or
379 * rearrange the set of dentries belonging to this super_block, nor may it
380 * change the attachments of dentries to inodes.
382 void generic_shutdown_super(struct super_block
*sb
)
384 const struct super_operations
*sop
= sb
->s_op
;
387 shrink_dcache_for_umount(sb
);
389 sb
->s_flags
&= ~MS_ACTIVE
;
391 fsnotify_unmount_inodes(&sb
->s_inodes
);
398 if (!list_empty(&sb
->s_inodes
)) {
399 printk("VFS: Busy inodes after unmount of %s. "
400 "Self-destruct in 5 seconds. Have a nice day...\n",
405 /* should be initialized for __put_super_and_need_restart() */
406 hlist_del_init(&sb
->s_instances
);
407 spin_unlock(&sb_lock
);
408 up_write(&sb
->s_umount
);
411 EXPORT_SYMBOL(generic_shutdown_super
);
414 * sget - find or create a superblock
415 * @type: filesystem type superblock should belong to
416 * @test: comparison callback
417 * @set: setup callback
418 * @data: argument to each of them
420 struct super_block
*sget(struct file_system_type
*type
,
421 int (*test
)(struct super_block
*,void *),
422 int (*set
)(struct super_block
*,void *),
425 struct super_block
*s
= NULL
;
426 struct hlist_node
*node
;
427 struct super_block
*old
;
433 hlist_for_each_entry(old
, node
, &type
->fs_supers
, s_instances
) {
434 if (!test(old
, data
))
436 if (!grab_super(old
))
439 up_write(&s
->s_umount
);
443 down_write(&old
->s_umount
);
444 if (unlikely(!(old
->s_flags
& MS_BORN
))) {
445 deactivate_locked_super(old
);
452 spin_unlock(&sb_lock
);
453 s
= alloc_super(type
);
455 return ERR_PTR(-ENOMEM
);
461 spin_unlock(&sb_lock
);
462 up_write(&s
->s_umount
);
467 strlcpy(s
->s_id
, type
->name
, sizeof(s
->s_id
));
468 list_add_tail(&s
->s_list
, &super_blocks
);
469 hlist_add_head(&s
->s_instances
, &type
->fs_supers
);
470 spin_unlock(&sb_lock
);
471 get_filesystem(type
);
472 register_shrinker(&s
->s_shrink
);
478 void drop_super(struct super_block
*sb
)
480 up_read(&sb
->s_umount
);
484 EXPORT_SYMBOL(drop_super
);
487 * sync_supers - helper for periodic superblock writeback
489 * Call the write_super method if present on all dirty superblocks in
490 * the system. This is for the periodic writeback used by most older
491 * filesystems. For data integrity superblock writeback use
492 * sync_filesystems() instead.
494 * Note: check the dirty flag before waiting, so we don't
495 * hold up the sync while mounting a device. (The newly
496 * mounted device won't need syncing.)
498 void sync_supers(void)
500 struct super_block
*sb
, *p
= NULL
;
503 list_for_each_entry(sb
, &super_blocks
, s_list
) {
504 if (hlist_unhashed(&sb
->s_instances
))
506 if (sb
->s_op
->write_super
&& sb
->s_dirt
) {
508 spin_unlock(&sb_lock
);
510 down_read(&sb
->s_umount
);
511 if (sb
->s_root
&& sb
->s_dirt
&& (sb
->s_flags
& MS_BORN
))
512 sb
->s_op
->write_super(sb
);
513 up_read(&sb
->s_umount
);
523 spin_unlock(&sb_lock
);
527 * iterate_supers - call function for all active superblocks
528 * @f: function to call
529 * @arg: argument to pass to it
531 * Scans the superblock list and calls given function, passing it
532 * locked superblock and given argument.
534 void iterate_supers(void (*f
)(struct super_block
*, void *), void *arg
)
536 struct super_block
*sb
, *p
= NULL
;
539 list_for_each_entry(sb
, &super_blocks
, s_list
) {
540 if (hlist_unhashed(&sb
->s_instances
))
543 spin_unlock(&sb_lock
);
545 down_read(&sb
->s_umount
);
546 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
548 up_read(&sb
->s_umount
);
557 spin_unlock(&sb_lock
);
561 * iterate_supers_type - call function for superblocks of given type
563 * @f: function to call
564 * @arg: argument to pass to it
566 * Scans the superblock list and calls given function, passing it
567 * locked superblock and given argument.
569 void iterate_supers_type(struct file_system_type
*type
,
570 void (*f
)(struct super_block
*, void *), void *arg
)
572 struct super_block
*sb
, *p
= NULL
;
573 struct hlist_node
*node
;
576 hlist_for_each_entry(sb
, node
, &type
->fs_supers
, s_instances
) {
578 spin_unlock(&sb_lock
);
580 down_read(&sb
->s_umount
);
581 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
583 up_read(&sb
->s_umount
);
592 spin_unlock(&sb_lock
);
595 EXPORT_SYMBOL(iterate_supers_type
);
598 * get_super - get the superblock of a device
599 * @bdev: device to get the superblock for
601 * Scans the superblock list and finds the superblock of the file system
602 * mounted on the device given. %NULL is returned if no match is found.
605 struct super_block
*get_super(struct block_device
*bdev
)
607 struct super_block
*sb
;
614 list_for_each_entry(sb
, &super_blocks
, s_list
) {
615 if (hlist_unhashed(&sb
->s_instances
))
617 if (sb
->s_bdev
== bdev
) {
619 spin_unlock(&sb_lock
);
620 down_read(&sb
->s_umount
);
622 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
624 up_read(&sb
->s_umount
);
625 /* nope, got unmounted */
631 spin_unlock(&sb_lock
);
635 EXPORT_SYMBOL(get_super
);
638 * get_super_thawed - get thawed superblock of a device
639 * @bdev: device to get the superblock for
641 * Scans the superblock list and finds the superblock of the file system
642 * mounted on the device. The superblock is returned once it is thawed
643 * (or immediately if it was not frozen). %NULL is returned if no match
646 struct super_block
*get_super_thawed(struct block_device
*bdev
)
649 struct super_block
*s
= get_super(bdev
);
650 if (!s
|| s
->s_frozen
== SB_UNFROZEN
)
652 up_read(&s
->s_umount
);
653 vfs_check_frozen(s
, SB_FREEZE_WRITE
);
657 EXPORT_SYMBOL(get_super_thawed
);
660 * get_active_super - get an active reference to the superblock of a device
661 * @bdev: device to get the superblock for
663 * Scans the superblock list and finds the superblock of the file system
664 * mounted on the device given. Returns the superblock with an active
665 * reference or %NULL if none was found.
667 struct super_block
*get_active_super(struct block_device
*bdev
)
669 struct super_block
*sb
;
676 list_for_each_entry(sb
, &super_blocks
, s_list
) {
677 if (hlist_unhashed(&sb
->s_instances
))
679 if (sb
->s_bdev
== bdev
) {
680 if (grab_super(sb
)) /* drops sb_lock */
686 spin_unlock(&sb_lock
);
690 struct super_block
*user_get_super(dev_t dev
)
692 struct super_block
*sb
;
696 list_for_each_entry(sb
, &super_blocks
, s_list
) {
697 if (hlist_unhashed(&sb
->s_instances
))
699 if (sb
->s_dev
== dev
) {
701 spin_unlock(&sb_lock
);
702 down_read(&sb
->s_umount
);
704 if (sb
->s_root
&& (sb
->s_flags
& MS_BORN
))
706 up_read(&sb
->s_umount
);
707 /* nope, got unmounted */
713 spin_unlock(&sb_lock
);
718 * do_remount_sb - asks filesystem to change mount options.
719 * @sb: superblock in question
720 * @flags: numeric part of options
721 * @data: the rest of options
722 * @force: whether or not to force the change
724 * Alters the mount options of a mounted file system.
726 int do_remount_sb(struct super_block
*sb
, int flags
, void *data
, int force
)
731 if (sb
->s_frozen
!= SB_UNFROZEN
)
735 if (!(flags
& MS_RDONLY
) && bdev_read_only(sb
->s_bdev
))
739 if (flags
& MS_RDONLY
)
741 shrink_dcache_sb(sb
);
744 remount_ro
= (flags
& MS_RDONLY
) && !(sb
->s_flags
& MS_RDONLY
);
746 /* If we are remounting RDONLY and current sb is read/write,
747 make sure there are no rw files opened */
752 retval
= sb_prepare_remount_readonly(sb
);
758 if (sb
->s_op
->remount_fs
) {
759 retval
= sb
->s_op
->remount_fs(sb
, &flags
, data
);
762 goto cancel_readonly
;
763 /* If forced remount, go ahead despite any errors */
764 WARN(1, "forced remount of a %s fs returned %i\n",
765 sb
->s_type
->name
, retval
);
768 sb
->s_flags
= (sb
->s_flags
& ~MS_RMT_MASK
) | (flags
& MS_RMT_MASK
);
769 /* Needs to be ordered wrt mnt_is_readonly() */
771 sb
->s_readonly_remount
= 0;
774 * Some filesystems modify their metadata via some other path than the
775 * bdev buffer cache (eg. use a private mapping, or directories in
776 * pagecache, etc). Also file data modifications go via their own
777 * mappings. So If we try to mount readonly then copy the filesystem
778 * from bdev, we could get stale data, so invalidate it to give a best
779 * effort at coherency.
781 if (remount_ro
&& sb
->s_bdev
)
782 invalidate_bdev(sb
->s_bdev
);
786 sb
->s_readonly_remount
= 0;
790 static void do_emergency_remount(struct work_struct
*work
)
792 struct super_block
*sb
, *p
= NULL
;
795 list_for_each_entry(sb
, &super_blocks
, s_list
) {
796 if (hlist_unhashed(&sb
->s_instances
))
799 spin_unlock(&sb_lock
);
800 down_write(&sb
->s_umount
);
801 if (sb
->s_root
&& sb
->s_bdev
&& (sb
->s_flags
& MS_BORN
) &&
802 !(sb
->s_flags
& MS_RDONLY
)) {
804 * What lock protects sb->s_flags??
806 do_remount_sb(sb
, MS_RDONLY
, NULL
, 1);
808 up_write(&sb
->s_umount
);
816 spin_unlock(&sb_lock
);
818 printk("Emergency Remount complete\n");
821 void emergency_remount(void)
823 struct work_struct
*work
;
825 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
827 INIT_WORK(work
, do_emergency_remount
);
833 * Unnamed block devices are dummy devices used by virtual
834 * filesystems which don't use real block-devices. -- jrs
837 static DEFINE_IDA(unnamed_dev_ida
);
838 static DEFINE_SPINLOCK(unnamed_dev_lock
);/* protects the above */
839 static int unnamed_dev_start
= 0; /* don't bother trying below it */
841 int get_anon_bdev(dev_t
*p
)
847 if (ida_pre_get(&unnamed_dev_ida
, GFP_ATOMIC
) == 0)
849 spin_lock(&unnamed_dev_lock
);
850 error
= ida_get_new_above(&unnamed_dev_ida
, unnamed_dev_start
, &dev
);
852 unnamed_dev_start
= dev
+ 1;
853 spin_unlock(&unnamed_dev_lock
);
854 if (error
== -EAGAIN
)
855 /* We raced and lost with another CPU. */
860 if ((dev
& MAX_ID_MASK
) == (1 << MINORBITS
)) {
861 spin_lock(&unnamed_dev_lock
);
862 ida_remove(&unnamed_dev_ida
, dev
);
863 if (unnamed_dev_start
> dev
)
864 unnamed_dev_start
= dev
;
865 spin_unlock(&unnamed_dev_lock
);
868 *p
= MKDEV(0, dev
& MINORMASK
);
871 EXPORT_SYMBOL(get_anon_bdev
);
873 void free_anon_bdev(dev_t dev
)
875 int slot
= MINOR(dev
);
876 spin_lock(&unnamed_dev_lock
);
877 ida_remove(&unnamed_dev_ida
, slot
);
878 if (slot
< unnamed_dev_start
)
879 unnamed_dev_start
= slot
;
880 spin_unlock(&unnamed_dev_lock
);
882 EXPORT_SYMBOL(free_anon_bdev
);
884 int set_anon_super(struct super_block
*s
, void *data
)
886 int error
= get_anon_bdev(&s
->s_dev
);
888 s
->s_bdi
= &noop_backing_dev_info
;
892 EXPORT_SYMBOL(set_anon_super
);
894 void kill_anon_super(struct super_block
*sb
)
896 dev_t dev
= sb
->s_dev
;
897 generic_shutdown_super(sb
);
901 EXPORT_SYMBOL(kill_anon_super
);
903 void kill_litter_super(struct super_block
*sb
)
906 d_genocide(sb
->s_root
);
910 EXPORT_SYMBOL(kill_litter_super
);
912 static int ns_test_super(struct super_block
*sb
, void *data
)
914 return sb
->s_fs_info
== data
;
917 static int ns_set_super(struct super_block
*sb
, void *data
)
919 sb
->s_fs_info
= data
;
920 return set_anon_super(sb
, NULL
);
923 struct dentry
*mount_ns(struct file_system_type
*fs_type
, int flags
,
924 void *data
, int (*fill_super
)(struct super_block
*, void *, int))
926 struct super_block
*sb
;
928 sb
= sget(fs_type
, ns_test_super
, ns_set_super
, data
);
935 err
= fill_super(sb
, data
, flags
& MS_SILENT
? 1 : 0);
937 deactivate_locked_super(sb
);
941 sb
->s_flags
|= MS_ACTIVE
;
944 return dget(sb
->s_root
);
947 EXPORT_SYMBOL(mount_ns
);
950 static int set_bdev_super(struct super_block
*s
, void *data
)
953 s
->s_dev
= s
->s_bdev
->bd_dev
;
956 * We set the bdi here to the queue backing, file systems can
957 * overwrite this in ->fill_super()
959 s
->s_bdi
= &bdev_get_queue(s
->s_bdev
)->backing_dev_info
;
963 static int test_bdev_super(struct super_block
*s
, void *data
)
965 return (void *)s
->s_bdev
== data
;
968 struct dentry
*mount_bdev(struct file_system_type
*fs_type
,
969 int flags
, const char *dev_name
, void *data
,
970 int (*fill_super
)(struct super_block
*, void *, int))
972 struct block_device
*bdev
;
973 struct super_block
*s
;
974 fmode_t mode
= FMODE_READ
| FMODE_EXCL
;
977 if (!(flags
& MS_RDONLY
))
980 bdev
= blkdev_get_by_path(dev_name
, mode
, fs_type
);
982 return ERR_CAST(bdev
);
985 * once the super is inserted into the list by sget, s_umount
986 * will protect the lockfs code from trying to start a snapshot
987 * while we are mounting
989 mutex_lock(&bdev
->bd_fsfreeze_mutex
);
990 if (bdev
->bd_fsfreeze_count
> 0) {
991 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
995 s
= sget(fs_type
, test_bdev_super
, set_bdev_super
, bdev
);
996 mutex_unlock(&bdev
->bd_fsfreeze_mutex
);
1001 if ((flags
^ s
->s_flags
) & MS_RDONLY
) {
1002 deactivate_locked_super(s
);
1008 * s_umount nests inside bd_mutex during
1009 * __invalidate_device(). blkdev_put() acquires
1010 * bd_mutex and can't be called under s_umount. Drop
1011 * s_umount temporarily. This is safe as we're
1012 * holding an active reference.
1014 up_write(&s
->s_umount
);
1015 blkdev_put(bdev
, mode
);
1016 down_write(&s
->s_umount
);
1018 char b
[BDEVNAME_SIZE
];
1020 s
->s_flags
= flags
| MS_NOSEC
;
1022 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
1023 sb_set_blocksize(s
, block_size(bdev
));
1024 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1026 deactivate_locked_super(s
);
1030 s
->s_flags
|= MS_ACTIVE
;
1034 return dget(s
->s_root
);
1039 blkdev_put(bdev
, mode
);
1041 return ERR_PTR(error
);
1043 EXPORT_SYMBOL(mount_bdev
);
1045 void kill_block_super(struct super_block
*sb
)
1047 struct block_device
*bdev
= sb
->s_bdev
;
1048 fmode_t mode
= sb
->s_mode
;
1050 bdev
->bd_super
= NULL
;
1051 generic_shutdown_super(sb
);
1052 sync_blockdev(bdev
);
1053 WARN_ON_ONCE(!(mode
& FMODE_EXCL
));
1054 blkdev_put(bdev
, mode
| FMODE_EXCL
);
1057 EXPORT_SYMBOL(kill_block_super
);
1060 struct dentry
*mount_nodev(struct file_system_type
*fs_type
,
1061 int flags
, void *data
,
1062 int (*fill_super
)(struct super_block
*, void *, int))
1065 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
1072 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1074 deactivate_locked_super(s
);
1075 return ERR_PTR(error
);
1077 s
->s_flags
|= MS_ACTIVE
;
1078 return dget(s
->s_root
);
1080 EXPORT_SYMBOL(mount_nodev
);
1082 static int compare_single(struct super_block
*s
, void *p
)
1087 struct dentry
*mount_single(struct file_system_type
*fs_type
,
1088 int flags
, void *data
,
1089 int (*fill_super
)(struct super_block
*, void *, int))
1091 struct super_block
*s
;
1094 s
= sget(fs_type
, compare_single
, set_anon_super
, NULL
);
1099 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1101 deactivate_locked_super(s
);
1102 return ERR_PTR(error
);
1104 s
->s_flags
|= MS_ACTIVE
;
1106 do_remount_sb(s
, flags
, data
, 0);
1108 return dget(s
->s_root
);
1110 EXPORT_SYMBOL(mount_single
);
1113 mount_fs(struct file_system_type
*type
, int flags
, const char *name
, void *data
)
1115 struct dentry
*root
;
1116 struct super_block
*sb
;
1117 char *secdata
= NULL
;
1118 int error
= -ENOMEM
;
1120 if (data
&& !(type
->fs_flags
& FS_BINARY_MOUNTDATA
)) {
1121 secdata
= alloc_secdata();
1125 error
= security_sb_copy_data(data
, secdata
);
1127 goto out_free_secdata
;
1130 root
= type
->mount(type
, flags
, name
, data
);
1132 error
= PTR_ERR(root
);
1133 goto out_free_secdata
;
1137 WARN_ON(!sb
->s_bdi
);
1138 WARN_ON(sb
->s_bdi
== &default_backing_dev_info
);
1139 sb
->s_flags
|= MS_BORN
;
1141 error
= security_sb_kern_mount(sb
, flags
, secdata
);
1146 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
1147 * but s_maxbytes was an unsigned long long for many releases. Throw
1148 * this warning for a little while to try and catch filesystems that
1149 * violate this rule.
1151 WARN((sb
->s_maxbytes
< 0), "%s set sb->s_maxbytes to "
1152 "negative value (%lld)\n", type
->name
, sb
->s_maxbytes
);
1154 up_write(&sb
->s_umount
);
1155 free_secdata(secdata
);
1159 deactivate_locked_super(sb
);
1161 free_secdata(secdata
);
1163 return ERR_PTR(error
);
1167 * freeze_super - lock the filesystem and force it into a consistent state
1168 * @sb: the super to lock
1170 * Syncs the super to make sure the filesystem is consistent and calls the fs's
1171 * freeze_fs. Subsequent calls to this without first thawing the fs will return
1174 int freeze_super(struct super_block
*sb
)
1178 atomic_inc(&sb
->s_active
);
1179 down_write(&sb
->s_umount
);
1181 deactivate_locked_super(sb
);
1185 if (!(sb
->s_flags
& MS_BORN
)) {
1186 up_write(&sb
->s_umount
);
1187 return 0; /* sic - it's "nothing to do" */
1190 if (sb
->s_flags
& MS_RDONLY
) {
1191 sb
->s_frozen
= SB_FREEZE_TRANS
;
1193 up_write(&sb
->s_umount
);
1197 sb
->s_frozen
= SB_FREEZE_WRITE
;
1200 sync_filesystem(sb
);
1202 sb
->s_frozen
= SB_FREEZE_TRANS
;
1205 sync_blockdev(sb
->s_bdev
);
1206 if (sb
->s_op
->freeze_fs
) {
1207 ret
= sb
->s_op
->freeze_fs(sb
);
1210 "VFS:Filesystem freeze failed\n");
1211 sb
->s_frozen
= SB_UNFROZEN
;
1213 wake_up(&sb
->s_wait_unfrozen
);
1214 deactivate_locked_super(sb
);
1218 up_write(&sb
->s_umount
);
1221 EXPORT_SYMBOL(freeze_super
);
1224 * thaw_super -- unlock filesystem
1225 * @sb: the super to thaw
1227 * Unlocks the filesystem and marks it writeable again after freeze_super().
1229 int thaw_super(struct super_block
*sb
)
1233 down_write(&sb
->s_umount
);
1234 if (sb
->s_frozen
== SB_UNFROZEN
) {
1235 up_write(&sb
->s_umount
);
1239 if (sb
->s_flags
& MS_RDONLY
)
1242 if (sb
->s_op
->unfreeze_fs
) {
1243 error
= sb
->s_op
->unfreeze_fs(sb
);
1246 "VFS:Filesystem thaw failed\n");
1247 sb
->s_frozen
= SB_FREEZE_TRANS
;
1248 up_write(&sb
->s_umount
);
1254 sb
->s_frozen
= SB_UNFROZEN
;
1256 wake_up(&sb
->s_wait_unfrozen
);
1257 deactivate_locked_super(sb
);
1261 EXPORT_SYMBOL(thaw_super
);