hwmon: New driver sch5636
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / super.c
blob7943f04cb3a92dd32b0ee68893cf2dfc3b3127af
1 /*
2 * linux/fs/super.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * super.c contains code to handle: - mount structures
7 * - super-block tables
8 * - filesystem drivers list
9 * - mount system call
10 * - umount system call
11 * - ustat 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 "internal.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;
51 int fs_objects = 0;
52 int total_objects;
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))
61 return -1;
63 if (!grab_super_passive(sb))
64 return -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;
72 if (sc->nr_to_scan) {
73 int dentries;
74 int inodes;
76 /* proportion the scan between the caches */
77 dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) /
78 total_objects;
79 inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) /
80 total_objects;
81 if (fs_objects)
82 fs_objects = (sc->nr_to_scan * fs_objects) /
83 total_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;
100 drop_super(sb);
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;
116 if (s) {
117 if (security_sb_alloc(s)) {
118 kfree(s);
119 s = NULL;
120 goto out;
122 #ifdef CONFIG_SMP
123 s->s_files = alloc_percpu(struct list_head);
124 if (!s->s_files) {
125 security_sb_free(s);
126 kfree(s);
127 s = NULL;
128 goto out;
129 } else {
130 int i;
132 for_each_possible_cpu(i)
133 INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i));
135 #else
136 INIT_LIST_HEAD(&s->s_files);
137 #endif
138 s->s_bdi = &default_backing_dev_info;
139 INIT_LIST_HEAD(&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
159 * one.
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
164 * risk of deadlocks.
166 * Annotate this by putting this lock in a different
167 * subclass.
169 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
170 s->s_count = 1;
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;
187 out:
188 return s;
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)
199 #ifdef CONFIG_SMP
200 free_percpu(s->s_files);
201 #endif
202 security_sb_free(s);
203 kfree(s->s_subtype);
204 kfree(s->s_options);
205 kfree(s);
208 /* Superblock refcounting */
211 * Drop a superblock's refcount. The caller must hold sb_lock.
213 void __put_super(struct super_block *sb)
215 if (!--sb->s_count) {
216 list_del_init(&sb->s_list);
217 destroy_super(sb);
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
226 * references left.
228 void put_super(struct super_block *sb)
230 spin_lock(&sb_lock);
231 __put_super(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
243 * had just acquired.
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);
252 fs->kill_sb(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.
261 rcu_barrier();
262 put_filesystem(fs);
263 put_super(s);
264 } else {
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);
304 return 1;
306 /* it's going away */
307 s->s_count++;
308 spin_unlock(&sb_lock);
309 /* wait for it to die */
310 down_write(&s->s_umount);
311 up_write(&s->s_umount);
312 put_super(s);
313 return 0;
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
326 * done.
328 bool grab_super_passive(struct super_block *sb)
330 spin_lock(&sb_lock);
331 if (list_empty(&sb->s_instances)) {
332 spin_unlock(&sb_lock);
333 return false;
336 sb->s_count++;
337 spin_unlock(&sb_lock);
339 if (down_read_trylock(&sb->s_umount)) {
340 if (sb->s_root)
341 return true;
342 up_read(&sb->s_umount);
345 put_super(sb);
346 return false;
350 * Superblock locking. We really ought to get rid of these two.
352 void lock_super(struct super_block * sb)
354 get_fs_excl();
355 mutex_lock(&sb->s_lock);
358 void unlock_super(struct super_block * sb)
360 put_fs_excl();
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;
385 if (sb->s_root) {
386 shrink_dcache_for_umount(sb);
387 sync_filesystem(sb);
388 get_fs_excl();
389 sb->s_flags &= ~MS_ACTIVE;
391 fsnotify_unmount_inodes(&sb->s_inodes);
393 evict_inodes(sb);
395 if (sop->put_super)
396 sop->put_super(sb);
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",
401 sb->s_id);
403 put_fs_excl();
405 spin_lock(&sb_lock);
406 /* should be initialized for __put_super_and_need_restart() */
407 list_del_init(&sb->s_instances);
408 spin_unlock(&sb_lock);
409 up_write(&sb->s_umount);
412 EXPORT_SYMBOL(generic_shutdown_super);
415 * sget - find or create a superblock
416 * @type: filesystem type superblock should belong to
417 * @test: comparison callback
418 * @set: setup callback
419 * @data: argument to each of them
421 struct super_block *sget(struct file_system_type *type,
422 int (*test)(struct super_block *,void *),
423 int (*set)(struct super_block *,void *),
424 void *data)
426 struct super_block *s = NULL;
427 struct super_block *old;
428 int err;
430 retry:
431 spin_lock(&sb_lock);
432 if (test) {
433 list_for_each_entry(old, &type->fs_supers, s_instances) {
434 if (!test(old, data))
435 continue;
436 if (!grab_super(old))
437 goto retry;
438 if (s) {
439 up_write(&s->s_umount);
440 destroy_super(s);
441 s = NULL;
443 down_write(&old->s_umount);
444 if (unlikely(!(old->s_flags & MS_BORN))) {
445 deactivate_locked_super(old);
446 goto retry;
448 return old;
451 if (!s) {
452 spin_unlock(&sb_lock);
453 s = alloc_super(type);
454 if (!s)
455 return ERR_PTR(-ENOMEM);
456 goto retry;
459 err = set(s, data);
460 if (err) {
461 spin_unlock(&sb_lock);
462 up_write(&s->s_umount);
463 destroy_super(s);
464 return ERR_PTR(err);
466 s->s_type = type;
467 strlcpy(s->s_id, type->name, sizeof(s->s_id));
468 list_add_tail(&s->s_list, &super_blocks);
469 list_add(&s->s_instances, &type->fs_supers);
470 spin_unlock(&sb_lock);
471 get_filesystem(type);
472 register_shrinker(&s->s_shrink);
473 return s;
476 EXPORT_SYMBOL(sget);
478 void drop_super(struct super_block *sb)
480 up_read(&sb->s_umount);
481 put_super(sb);
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;
502 spin_lock(&sb_lock);
503 list_for_each_entry(sb, &super_blocks, s_list) {
504 if (list_empty(&sb->s_instances))
505 continue;
506 if (sb->s_op->write_super && sb->s_dirt) {
507 sb->s_count++;
508 spin_unlock(&sb_lock);
510 down_read(&sb->s_umount);
511 if (sb->s_root && sb->s_dirt)
512 sb->s_op->write_super(sb);
513 up_read(&sb->s_umount);
515 spin_lock(&sb_lock);
516 if (p)
517 __put_super(p);
518 p = sb;
521 if (p)
522 __put_super(p);
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;
538 spin_lock(&sb_lock);
539 list_for_each_entry(sb, &super_blocks, s_list) {
540 if (list_empty(&sb->s_instances))
541 continue;
542 sb->s_count++;
543 spin_unlock(&sb_lock);
545 down_read(&sb->s_umount);
546 if (sb->s_root)
547 f(sb, arg);
548 up_read(&sb->s_umount);
550 spin_lock(&sb_lock);
551 if (p)
552 __put_super(p);
553 p = sb;
555 if (p)
556 __put_super(p);
557 spin_unlock(&sb_lock);
561 * iterate_supers_type - call function for superblocks of given type
562 * @type: fs 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;
574 spin_lock(&sb_lock);
575 list_for_each_entry(sb, &type->fs_supers, s_instances) {
576 sb->s_count++;
577 spin_unlock(&sb_lock);
579 down_read(&sb->s_umount);
580 if (sb->s_root)
581 f(sb, arg);
582 up_read(&sb->s_umount);
584 spin_lock(&sb_lock);
585 if (p)
586 __put_super(p);
587 p = sb;
589 if (p)
590 __put_super(p);
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;
608 if (!bdev)
609 return NULL;
611 spin_lock(&sb_lock);
612 rescan:
613 list_for_each_entry(sb, &super_blocks, s_list) {
614 if (list_empty(&sb->s_instances))
615 continue;
616 if (sb->s_bdev == bdev) {
617 sb->s_count++;
618 spin_unlock(&sb_lock);
619 down_read(&sb->s_umount);
620 /* still alive? */
621 if (sb->s_root)
622 return sb;
623 up_read(&sb->s_umount);
624 /* nope, got unmounted */
625 spin_lock(&sb_lock);
626 __put_super(sb);
627 goto rescan;
630 spin_unlock(&sb_lock);
631 return NULL;
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;
648 if (!bdev)
649 return NULL;
651 restart:
652 spin_lock(&sb_lock);
653 list_for_each_entry(sb, &super_blocks, s_list) {
654 if (list_empty(&sb->s_instances))
655 continue;
656 if (sb->s_bdev == bdev) {
657 if (grab_super(sb)) /* drops sb_lock */
658 return sb;
659 else
660 goto restart;
663 spin_unlock(&sb_lock);
664 return NULL;
667 struct super_block *user_get_super(dev_t dev)
669 struct super_block *sb;
671 spin_lock(&sb_lock);
672 rescan:
673 list_for_each_entry(sb, &super_blocks, s_list) {
674 if (list_empty(&sb->s_instances))
675 continue;
676 if (sb->s_dev == dev) {
677 sb->s_count++;
678 spin_unlock(&sb_lock);
679 down_read(&sb->s_umount);
680 /* still alive? */
681 if (sb->s_root)
682 return sb;
683 up_read(&sb->s_umount);
684 /* nope, got unmounted */
685 spin_lock(&sb_lock);
686 __put_super(sb);
687 goto rescan;
690 spin_unlock(&sb_lock);
691 return NULL;
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)
705 int retval;
706 int remount_ro;
708 if (sb->s_frozen != SB_UNFROZEN)
709 return -EBUSY;
711 #ifdef CONFIG_BLOCK
712 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
713 return -EACCES;
714 #endif
716 if (flags & MS_RDONLY)
717 acct_auto_close(sb);
718 shrink_dcache_sb(sb);
719 sync_filesystem(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 */
725 if (remount_ro) {
726 if (force)
727 mark_files_ro(sb);
728 else if (!fs_may_remount_ro(sb))
729 return -EBUSY;
732 if (sb->s_op->remount_fs) {
733 retval = sb->s_op->remount_fs(sb, &flags, data);
734 if (retval)
735 return retval;
737 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
740 * Some filesystems modify their metadata via some other path than the
741 * bdev buffer cache (eg. use a private mapping, or directories in
742 * pagecache, etc). Also file data modifications go via their own
743 * mappings. So If we try to mount readonly then copy the filesystem
744 * from bdev, we could get stale data, so invalidate it to give a best
745 * effort at coherency.
747 if (remount_ro && sb->s_bdev)
748 invalidate_bdev(sb->s_bdev);
749 return 0;
752 static void do_emergency_remount(struct work_struct *work)
754 struct super_block *sb, *p = NULL;
756 spin_lock(&sb_lock);
757 list_for_each_entry(sb, &super_blocks, s_list) {
758 if (list_empty(&sb->s_instances))
759 continue;
760 sb->s_count++;
761 spin_unlock(&sb_lock);
762 down_write(&sb->s_umount);
763 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
765 * What lock protects sb->s_flags??
767 do_remount_sb(sb, MS_RDONLY, NULL, 1);
769 up_write(&sb->s_umount);
770 spin_lock(&sb_lock);
771 if (p)
772 __put_super(p);
773 p = sb;
775 if (p)
776 __put_super(p);
777 spin_unlock(&sb_lock);
778 kfree(work);
779 printk("Emergency Remount complete\n");
782 void emergency_remount(void)
784 struct work_struct *work;
786 work = kmalloc(sizeof(*work), GFP_ATOMIC);
787 if (work) {
788 INIT_WORK(work, do_emergency_remount);
789 schedule_work(work);
794 * Unnamed block devices are dummy devices used by virtual
795 * filesystems which don't use real block-devices. -- jrs
798 static DEFINE_IDA(unnamed_dev_ida);
799 static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
800 static int unnamed_dev_start = 0; /* don't bother trying below it */
802 int get_anon_bdev(dev_t *p)
804 int dev;
805 int error;
807 retry:
808 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
809 return -ENOMEM;
810 spin_lock(&unnamed_dev_lock);
811 error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
812 if (!error)
813 unnamed_dev_start = dev + 1;
814 spin_unlock(&unnamed_dev_lock);
815 if (error == -EAGAIN)
816 /* We raced and lost with another CPU. */
817 goto retry;
818 else if (error)
819 return -EAGAIN;
821 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
822 spin_lock(&unnamed_dev_lock);
823 ida_remove(&unnamed_dev_ida, dev);
824 if (unnamed_dev_start > dev)
825 unnamed_dev_start = dev;
826 spin_unlock(&unnamed_dev_lock);
827 return -EMFILE;
829 *p = MKDEV(0, dev & MINORMASK);
830 return 0;
832 EXPORT_SYMBOL(get_anon_bdev);
834 void free_anon_bdev(dev_t dev)
836 int slot = MINOR(dev);
837 spin_lock(&unnamed_dev_lock);
838 ida_remove(&unnamed_dev_ida, slot);
839 if (slot < unnamed_dev_start)
840 unnamed_dev_start = slot;
841 spin_unlock(&unnamed_dev_lock);
843 EXPORT_SYMBOL(free_anon_bdev);
845 int set_anon_super(struct super_block *s, void *data)
847 int error = get_anon_bdev(&s->s_dev);
848 if (!error)
849 s->s_bdi = &noop_backing_dev_info;
850 return error;
853 EXPORT_SYMBOL(set_anon_super);
855 void kill_anon_super(struct super_block *sb)
857 dev_t dev = sb->s_dev;
858 generic_shutdown_super(sb);
859 free_anon_bdev(dev);
862 EXPORT_SYMBOL(kill_anon_super);
864 void kill_litter_super(struct super_block *sb)
866 if (sb->s_root)
867 d_genocide(sb->s_root);
868 kill_anon_super(sb);
871 EXPORT_SYMBOL(kill_litter_super);
873 static int ns_test_super(struct super_block *sb, void *data)
875 return sb->s_fs_info == data;
878 static int ns_set_super(struct super_block *sb, void *data)
880 sb->s_fs_info = data;
881 return set_anon_super(sb, NULL);
884 struct dentry *mount_ns(struct file_system_type *fs_type, int flags,
885 void *data, int (*fill_super)(struct super_block *, void *, int))
887 struct super_block *sb;
889 sb = sget(fs_type, ns_test_super, ns_set_super, data);
890 if (IS_ERR(sb))
891 return ERR_CAST(sb);
893 if (!sb->s_root) {
894 int err;
895 sb->s_flags = flags;
896 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
897 if (err) {
898 deactivate_locked_super(sb);
899 return ERR_PTR(err);
902 sb->s_flags |= MS_ACTIVE;
905 return dget(sb->s_root);
908 EXPORT_SYMBOL(mount_ns);
910 #ifdef CONFIG_BLOCK
911 static int set_bdev_super(struct super_block *s, void *data)
913 s->s_bdev = data;
914 s->s_dev = s->s_bdev->bd_dev;
917 * We set the bdi here to the queue backing, file systems can
918 * overwrite this in ->fill_super()
920 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
921 return 0;
924 static int test_bdev_super(struct super_block *s, void *data)
926 return (void *)s->s_bdev == data;
929 struct dentry *mount_bdev(struct file_system_type *fs_type,
930 int flags, const char *dev_name, void *data,
931 int (*fill_super)(struct super_block *, void *, int))
933 struct block_device *bdev;
934 struct super_block *s;
935 fmode_t mode = FMODE_READ | FMODE_EXCL;
936 int error = 0;
938 if (!(flags & MS_RDONLY))
939 mode |= FMODE_WRITE;
941 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
942 if (IS_ERR(bdev))
943 return ERR_CAST(bdev);
946 * once the super is inserted into the list by sget, s_umount
947 * will protect the lockfs code from trying to start a snapshot
948 * while we are mounting
950 mutex_lock(&bdev->bd_fsfreeze_mutex);
951 if (bdev->bd_fsfreeze_count > 0) {
952 mutex_unlock(&bdev->bd_fsfreeze_mutex);
953 error = -EBUSY;
954 goto error_bdev;
956 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
957 mutex_unlock(&bdev->bd_fsfreeze_mutex);
958 if (IS_ERR(s))
959 goto error_s;
961 if (s->s_root) {
962 if ((flags ^ s->s_flags) & MS_RDONLY) {
963 deactivate_locked_super(s);
964 error = -EBUSY;
965 goto error_bdev;
969 * s_umount nests inside bd_mutex during
970 * __invalidate_device(). blkdev_put() acquires
971 * bd_mutex and can't be called under s_umount. Drop
972 * s_umount temporarily. This is safe as we're
973 * holding an active reference.
975 up_write(&s->s_umount);
976 blkdev_put(bdev, mode);
977 down_write(&s->s_umount);
978 } else {
979 char b[BDEVNAME_SIZE];
981 s->s_flags = flags | MS_NOSEC;
982 s->s_mode = mode;
983 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
984 sb_set_blocksize(s, block_size(bdev));
985 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
986 if (error) {
987 deactivate_locked_super(s);
988 goto error;
991 s->s_flags |= MS_ACTIVE;
992 bdev->bd_super = s;
995 return dget(s->s_root);
997 error_s:
998 error = PTR_ERR(s);
999 error_bdev:
1000 blkdev_put(bdev, mode);
1001 error:
1002 return ERR_PTR(error);
1004 EXPORT_SYMBOL(mount_bdev);
1006 void kill_block_super(struct super_block *sb)
1008 struct block_device *bdev = sb->s_bdev;
1009 fmode_t mode = sb->s_mode;
1011 bdev->bd_super = NULL;
1012 generic_shutdown_super(sb);
1013 sync_blockdev(bdev);
1014 WARN_ON_ONCE(!(mode & FMODE_EXCL));
1015 blkdev_put(bdev, mode | FMODE_EXCL);
1018 EXPORT_SYMBOL(kill_block_super);
1019 #endif
1021 struct dentry *mount_nodev(struct file_system_type *fs_type,
1022 int flags, void *data,
1023 int (*fill_super)(struct super_block *, void *, int))
1025 int error;
1026 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
1028 if (IS_ERR(s))
1029 return ERR_CAST(s);
1031 s->s_flags = flags;
1033 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1034 if (error) {
1035 deactivate_locked_super(s);
1036 return ERR_PTR(error);
1038 s->s_flags |= MS_ACTIVE;
1039 return dget(s->s_root);
1041 EXPORT_SYMBOL(mount_nodev);
1043 static int compare_single(struct super_block *s, void *p)
1045 return 1;
1048 struct dentry *mount_single(struct file_system_type *fs_type,
1049 int flags, void *data,
1050 int (*fill_super)(struct super_block *, void *, int))
1052 struct super_block *s;
1053 int error;
1055 s = sget(fs_type, compare_single, set_anon_super, NULL);
1056 if (IS_ERR(s))
1057 return ERR_CAST(s);
1058 if (!s->s_root) {
1059 s->s_flags = flags;
1060 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1061 if (error) {
1062 deactivate_locked_super(s);
1063 return ERR_PTR(error);
1065 s->s_flags |= MS_ACTIVE;
1066 } else {
1067 do_remount_sb(s, flags, data, 0);
1069 return dget(s->s_root);
1071 EXPORT_SYMBOL(mount_single);
1073 struct dentry *
1074 mount_fs(struct file_system_type *type, int flags, const char *name, void *data)
1076 struct dentry *root;
1077 struct super_block *sb;
1078 char *secdata = NULL;
1079 int error = -ENOMEM;
1081 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
1082 secdata = alloc_secdata();
1083 if (!secdata)
1084 goto out;
1086 error = security_sb_copy_data(data, secdata);
1087 if (error)
1088 goto out_free_secdata;
1091 root = type->mount(type, flags, name, data);
1092 if (IS_ERR(root)) {
1093 error = PTR_ERR(root);
1094 goto out_free_secdata;
1096 sb = root->d_sb;
1097 BUG_ON(!sb);
1098 WARN_ON(!sb->s_bdi);
1099 WARN_ON(sb->s_bdi == &default_backing_dev_info);
1100 sb->s_flags |= MS_BORN;
1102 error = security_sb_kern_mount(sb, flags, secdata);
1103 if (error)
1104 goto out_sb;
1107 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
1108 * but s_maxbytes was an unsigned long long for many releases. Throw
1109 * this warning for a little while to try and catch filesystems that
1110 * violate this rule.
1112 WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to "
1113 "negative value (%lld)\n", type->name, sb->s_maxbytes);
1115 up_write(&sb->s_umount);
1116 free_secdata(secdata);
1117 return root;
1118 out_sb:
1119 dput(root);
1120 deactivate_locked_super(sb);
1121 out_free_secdata:
1122 free_secdata(secdata);
1123 out:
1124 return ERR_PTR(error);
1128 * freeze_super - lock the filesystem and force it into a consistent state
1129 * @sb: the super to lock
1131 * Syncs the super to make sure the filesystem is consistent and calls the fs's
1132 * freeze_fs. Subsequent calls to this without first thawing the fs will return
1133 * -EBUSY.
1135 int freeze_super(struct super_block *sb)
1137 int ret;
1139 atomic_inc(&sb->s_active);
1140 down_write(&sb->s_umount);
1141 if (sb->s_frozen) {
1142 deactivate_locked_super(sb);
1143 return -EBUSY;
1146 if (sb->s_flags & MS_RDONLY) {
1147 sb->s_frozen = SB_FREEZE_TRANS;
1148 smp_wmb();
1149 up_write(&sb->s_umount);
1150 return 0;
1153 sb->s_frozen = SB_FREEZE_WRITE;
1154 smp_wmb();
1156 sync_filesystem(sb);
1158 sb->s_frozen = SB_FREEZE_TRANS;
1159 smp_wmb();
1161 sync_blockdev(sb->s_bdev);
1162 if (sb->s_op->freeze_fs) {
1163 ret = sb->s_op->freeze_fs(sb);
1164 if (ret) {
1165 printk(KERN_ERR
1166 "VFS:Filesystem freeze failed\n");
1167 sb->s_frozen = SB_UNFROZEN;
1168 deactivate_locked_super(sb);
1169 return ret;
1172 up_write(&sb->s_umount);
1173 return 0;
1175 EXPORT_SYMBOL(freeze_super);
1178 * thaw_super -- unlock filesystem
1179 * @sb: the super to thaw
1181 * Unlocks the filesystem and marks it writeable again after freeze_super().
1183 int thaw_super(struct super_block *sb)
1185 int error;
1187 down_write(&sb->s_umount);
1188 if (sb->s_frozen == SB_UNFROZEN) {
1189 up_write(&sb->s_umount);
1190 return -EINVAL;
1193 if (sb->s_flags & MS_RDONLY)
1194 goto out;
1196 if (sb->s_op->unfreeze_fs) {
1197 error = sb->s_op->unfreeze_fs(sb);
1198 if (error) {
1199 printk(KERN_ERR
1200 "VFS:Filesystem thaw failed\n");
1201 sb->s_frozen = SB_FREEZE_TRANS;
1202 up_write(&sb->s_umount);
1203 return error;
1207 out:
1208 sb->s_frozen = SB_UNFROZEN;
1209 smp_wmb();
1210 wake_up(&sb->s_wait_unfrozen);
1211 deactivate_locked_super(sb);
1213 return 0;
1215 EXPORT_SYMBOL(thaw_super);