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/config.h>
24 #include <linux/module.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/smp_lock.h>
28 #include <linux/acct.h>
29 #include <linux/blkdev.h>
30 #include <linux/quotaops.h>
31 #include <linux/namei.h>
32 #include <linux/buffer_head.h> /* for fsync_super() */
33 #include <linux/mount.h>
34 #include <linux/security.h>
35 #include <linux/vfs.h>
36 #include <linux/writeback.h> /* for the emergency remount stuff */
37 #include <linux/idr.h>
38 #include <asm/uaccess.h>
41 void get_filesystem(struct file_system_type
*fs
);
42 void put_filesystem(struct file_system_type
*fs
);
43 struct file_system_type
*get_fs_type(const char *name
);
45 LIST_HEAD(super_blocks
);
46 spinlock_t sb_lock
= SPIN_LOCK_UNLOCKED
;
49 * alloc_super - create new superblock
51 * Allocates and initializes a new &struct super_block. alloc_super()
52 * returns a pointer new superblock or %NULL if allocation had failed.
54 static struct super_block
*alloc_super(void)
56 struct super_block
*s
= kmalloc(sizeof(struct super_block
), GFP_USER
);
57 static struct super_operations default_op
;
60 memset(s
, 0, sizeof(struct super_block
));
61 if (security_sb_alloc(s
)) {
66 INIT_LIST_HEAD(&s
->s_dirty
);
67 INIT_LIST_HEAD(&s
->s_io
);
68 INIT_LIST_HEAD(&s
->s_files
);
69 INIT_LIST_HEAD(&s
->s_instances
);
70 INIT_HLIST_HEAD(&s
->s_anon
);
71 init_rwsem(&s
->s_umount
);
72 sema_init(&s
->s_lock
, 1);
73 down_write(&s
->s_umount
);
75 atomic_set(&s
->s_active
, 1);
76 sema_init(&s
->s_vfs_rename_sem
,1);
77 sema_init(&s
->s_dquot
.dqio_sem
, 1);
78 sema_init(&s
->s_dquot
.dqonoff_sem
, 1);
79 init_rwsem(&s
->s_dquot
.dqptr_sem
);
80 s
->s_maxbytes
= MAX_NON_LFS
;
81 s
->dq_op
= sb_dquot_ops
;
82 s
->s_qcop
= sb_quotactl_ops
;
83 s
->s_op
= &default_op
;
90 * destroy_super - frees a superblock
91 * @s: superblock to free
95 static inline void destroy_super(struct super_block
*s
)
101 /* Superblock refcounting */
104 * put_super - drop a temporary reference to superblock
105 * @s: superblock in question
107 * Drops a temporary reference, frees superblock if there's no
110 static inline void put_super(struct super_block
*s
)
115 spin_unlock(&sb_lock
);
119 * deactivate_super - drop an active reference to superblock
120 * @s: superblock to deactivate
122 * Drops an active reference to superblock, acquiring a temprory one if
123 * there is no active references left. In that case we lock superblock,
124 * tell fs driver to shut it down and drop the temporary reference we
127 void deactivate_super(struct super_block
*s
)
129 struct file_system_type
*fs
= s
->s_type
;
130 if (atomic_dec_and_lock(&s
->s_active
, &sb_lock
)) {
131 s
->s_count
-= S_BIAS
-1;
132 spin_unlock(&sb_lock
);
133 down_write(&s
->s_umount
);
140 EXPORT_SYMBOL(deactivate_super
);
143 * grab_super - acquire an active reference
144 * @s: reference we are trying to make active
146 * Tries to acquire an active reference. grab_super() is used when we
147 * had just found a superblock in super_blocks or fs_type->fs_supers
148 * and want to turn it into a full-blown active reference. grab_super()
149 * is called with sb_lock held and drops it. Returns 1 in case of
150 * success, 0 if we had failed (superblock contents was already dead or
151 * dying when grab_super() had been called).
153 static int grab_super(struct super_block
*s
)
156 spin_unlock(&sb_lock
);
157 down_write(&s
->s_umount
);
160 if (s
->s_count
> S_BIAS
) {
161 atomic_inc(&s
->s_active
);
163 spin_unlock(&sb_lock
);
166 spin_unlock(&sb_lock
);
168 up_write(&s
->s_umount
);
175 * generic_shutdown_super - common helper for ->kill_sb()
176 * @sb: superblock to kill
178 * generic_shutdown_super() does all fs-independent work on superblock
179 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
180 * that need destruction out of superblock, call generic_shutdown_super()
181 * and release aforementioned objects. Note: dentries and inodes _are_
182 * taken care of and do not need specific handling.
184 void generic_shutdown_super(struct super_block
*sb
)
186 struct dentry
*root
= sb
->s_root
;
187 struct super_operations
*sop
= sb
->s_op
;
191 shrink_dcache_parent(root
);
192 shrink_dcache_anon(&sb
->s_anon
);
197 sb
->s_flags
&= ~MS_ACTIVE
;
198 /* bad name - it should be evict_inodes() */
199 invalidate_inodes(sb
);
201 if (sop
->write_super
&& sb
->s_dirt
)
202 sop
->write_super(sb
);
206 /* Forget any remaining inodes */
207 if (invalidate_inodes(sb
)) {
208 printk("VFS: Busy inodes after unmount. "
209 "Self-destruct in 5 seconds. Have a nice day...\n");
216 list_del(&sb
->s_list
);
217 list_del(&sb
->s_instances
);
218 spin_unlock(&sb_lock
);
219 up_write(&sb
->s_umount
);
222 EXPORT_SYMBOL(generic_shutdown_super
);
225 * sget - find or create a superblock
226 * @type: filesystem type superblock should belong to
227 * @test: comparison callback
228 * @set: setup callback
229 * @data: argument to each of them
231 struct super_block
*sget(struct file_system_type
*type
,
232 int (*test
)(struct super_block
*,void *),
233 int (*set
)(struct super_block
*,void *),
236 struct super_block
*s
= NULL
;
242 if (test
) list_for_each(p
, &type
->fs_supers
) {
243 struct super_block
*old
;
244 old
= list_entry(p
, struct super_block
, s_instances
);
245 if (!test(old
, data
))
247 if (!grab_super(old
))
254 spin_unlock(&sb_lock
);
257 return ERR_PTR(-ENOMEM
);
263 spin_unlock(&sb_lock
);
268 list_add(&s
->s_list
, super_blocks
.prev
);
269 list_add(&s
->s_instances
, &type
->fs_supers
);
270 spin_unlock(&sb_lock
);
271 get_filesystem(type
);
277 void drop_super(struct super_block
*sb
)
279 up_read(&sb
->s_umount
);
283 EXPORT_SYMBOL(drop_super
);
285 static inline void write_super(struct super_block
*sb
)
288 if (sb
->s_root
&& sb
->s_dirt
)
289 if (sb
->s_op
->write_super
)
290 sb
->s_op
->write_super(sb
);
295 * Note: check the dirty flag before waiting, so we don't
296 * hold up the sync while mounting a device. (The newly
297 * mounted device won't need syncing.)
299 void sync_supers(void)
301 struct super_block
* sb
;
304 sb
= sb_entry(super_blocks
.next
);
305 while (sb
!= sb_entry(&super_blocks
))
308 spin_unlock(&sb_lock
);
309 down_read(&sb
->s_umount
);
314 sb
= sb_entry(sb
->s_list
.next
);
315 spin_unlock(&sb_lock
);
319 * Call the ->sync_fs super_op against all filesytems which are r/w and
320 * which implement it.
322 * This operation is careful to avoid the livelock which could easily happen
323 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
324 * is used only here. We set it against all filesystems and then clear it as
325 * we sync them. So redirtied filesystems are skipped.
327 * But if process A is currently running sync_filesytems and then process B
328 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
329 * flags again, which will cause process A to resync everything. Fix that with
332 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
334 void sync_filesystems(int wait
)
336 struct super_block
*sb
;
337 static DECLARE_MUTEX(mutex
);
339 down(&mutex
); /* Could be down_interruptible */
341 for (sb
= sb_entry(super_blocks
.next
); sb
!= sb_entry(&super_blocks
);
342 sb
= sb_entry(sb
->s_list
.next
)) {
343 if (!sb
->s_op
->sync_fs
)
345 if (sb
->s_flags
& MS_RDONLY
)
347 sb
->s_need_sync_fs
= 1;
349 spin_unlock(&sb_lock
);
353 for (sb
= sb_entry(super_blocks
.next
); sb
!= sb_entry(&super_blocks
);
354 sb
= sb_entry(sb
->s_list
.next
)) {
355 if (!sb
->s_need_sync_fs
)
357 sb
->s_need_sync_fs
= 0;
358 if (sb
->s_flags
& MS_RDONLY
)
359 continue; /* hm. Was remounted r/o meanwhile */
361 spin_unlock(&sb_lock
);
362 down_read(&sb
->s_umount
);
363 if (sb
->s_root
&& (wait
|| sb
->s_dirt
))
364 sb
->s_op
->sync_fs(sb
, wait
);
368 spin_unlock(&sb_lock
);
373 * get_super - get the superblock of a device
374 * @bdev: device to get the superblock for
376 * Scans the superblock list and finds the superblock of the file system
377 * mounted on the device given. %NULL is returned if no match is found.
380 struct super_block
* get_super(struct block_device
*bdev
)
387 list_for_each(p
, &super_blocks
) {
388 struct super_block
*s
= sb_entry(p
);
389 if (s
->s_bdev
== bdev
) {
391 spin_unlock(&sb_lock
);
392 down_read(&s
->s_umount
);
399 spin_unlock(&sb_lock
);
403 EXPORT_SYMBOL(get_super
);
405 struct super_block
* user_get_super(dev_t dev
)
411 list_for_each(p
, &super_blocks
) {
412 struct super_block
*s
= sb_entry(p
);
413 if (s
->s_dev
== dev
) {
415 spin_unlock(&sb_lock
);
416 down_read(&s
->s_umount
);
423 spin_unlock(&sb_lock
);
427 EXPORT_SYMBOL(user_get_super
);
429 asmlinkage
long sys_ustat(unsigned dev
, struct ustat __user
* ubuf
)
431 struct super_block
*s
;
436 s
= user_get_super(new_decode_dev(dev
));
439 err
= vfs_statfs(s
, &sbuf
);
444 memset(&tmp
,0,sizeof(struct ustat
));
445 tmp
.f_tfree
= sbuf
.f_bfree
;
446 tmp
.f_tinode
= sbuf
.f_ffree
;
448 err
= copy_to_user(ubuf
,&tmp
,sizeof(struct ustat
)) ? -EFAULT
: 0;
455 * @sb: superblock in question
457 * All files are marked read/only. We don't care about pending
458 * delete files so this should be used in 'force' mode only
461 static void mark_files_ro(struct super_block
*sb
)
466 list_for_each_entry(f
, &sb
->s_files
, f_list
) {
467 if (S_ISREG(f
->f_dentry
->d_inode
->i_mode
) && file_count(f
))
468 f
->f_mode
&= ~FMODE_WRITE
;
474 * do_remount_sb - asks filesystem to change mount options.
475 * @sb: superblock in question
476 * @flags: numeric part of options
477 * @data: the rest of options
478 * @force: whether or not to force the change
480 * Alters the mount options of a mounted file system.
482 int do_remount_sb(struct super_block
*sb
, int flags
, void *data
, int force
)
486 if (!(flags
& MS_RDONLY
) && bdev_read_only(sb
->s_bdev
))
488 if (flags
& MS_RDONLY
)
490 shrink_dcache_sb(sb
);
493 /* If we are remounting RDONLY and current sb is read/write,
494 make sure there are no rw files opened */
495 if ((flags
& MS_RDONLY
) && !(sb
->s_flags
& MS_RDONLY
)) {
498 else if (!fs_may_remount_ro(sb
))
502 if (sb
->s_op
->remount_fs
) {
504 retval
= sb
->s_op
->remount_fs(sb
, &flags
, data
);
509 sb
->s_flags
= (sb
->s_flags
& ~MS_RMT_MASK
) | (flags
& MS_RMT_MASK
);
513 static void do_emergency_remount(unsigned long foo
)
515 struct super_block
*sb
;
518 list_for_each_entry(sb
, &super_blocks
, s_list
) {
520 spin_unlock(&sb_lock
);
521 down_read(&sb
->s_umount
);
522 if (sb
->s_root
&& sb
->s_bdev
&& !(sb
->s_flags
& MS_RDONLY
)) {
524 * ->remount_fs needs lock_kernel().
526 * What lock protects sb->s_flags??
529 do_remount_sb(sb
, MS_RDONLY
, NULL
, 1);
535 spin_unlock(&sb_lock
);
536 printk("Emergency Remount complete\n");
539 void emergency_remount(void)
541 pdflush_operation(do_emergency_remount
, 0);
545 * Unnamed block devices are dummy devices used by virtual
546 * filesystems which don't use real block-devices. -- jrs
549 static struct idr unnamed_dev_idr
;
550 static spinlock_t unnamed_dev_lock
= SPIN_LOCK_UNLOCKED
;/* protects the above */
552 int set_anon_super(struct super_block
*s
, void *data
)
556 spin_lock(&unnamed_dev_lock
);
557 if (idr_pre_get(&unnamed_dev_idr
, GFP_ATOMIC
) == 0) {
558 spin_unlock(&unnamed_dev_lock
);
561 dev
= idr_get_new(&unnamed_dev_idr
, NULL
);
562 spin_unlock(&unnamed_dev_lock
);
564 if ((dev
& MAX_ID_MASK
) == (1 << MINORBITS
)) {
565 idr_remove(&unnamed_dev_idr
, dev
);
568 s
->s_dev
= MKDEV(0, dev
& MINORMASK
);
572 EXPORT_SYMBOL(set_anon_super
);
574 void kill_anon_super(struct super_block
*sb
)
576 int slot
= MINOR(sb
->s_dev
);
578 generic_shutdown_super(sb
);
579 spin_lock(&unnamed_dev_lock
);
580 idr_remove(&unnamed_dev_idr
, slot
);
581 spin_unlock(&unnamed_dev_lock
);
584 EXPORT_SYMBOL(kill_anon_super
);
586 void __init
unnamed_dev_init(void)
588 idr_init(&unnamed_dev_idr
);
591 void kill_litter_super(struct super_block
*sb
)
594 d_genocide(sb
->s_root
);
598 EXPORT_SYMBOL(kill_litter_super
);
600 static int set_bdev_super(struct super_block
*s
, void *data
)
603 s
->s_dev
= s
->s_bdev
->bd_dev
;
607 static int test_bdev_super(struct super_block
*s
, void *data
)
609 return (void *)s
->s_bdev
== data
;
612 struct super_block
*get_sb_bdev(struct file_system_type
*fs_type
,
613 int flags
, const char *dev_name
, void *data
,
614 int (*fill_super
)(struct super_block
*, void *, int))
616 struct block_device
*bdev
;
617 struct super_block
*s
;
620 bdev
= open_bdev_excl(dev_name
, flags
, fs_type
);
622 return (struct super_block
*)bdev
;
624 s
= sget(fs_type
, test_bdev_super
, set_bdev_super
, bdev
);
629 if ((flags
^ s
->s_flags
) & MS_RDONLY
) {
630 up_write(&s
->s_umount
);
636 char b
[BDEVNAME_SIZE
];
639 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
640 s
->s_old_blocksize
= block_size(bdev
);
641 sb_set_blocksize(s
, s
->s_old_blocksize
);
642 error
= fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
644 up_write(&s
->s_umount
);
648 s
->s_flags
|= MS_ACTIVE
;
654 close_bdev_excl(bdev
);
658 EXPORT_SYMBOL(get_sb_bdev
);
660 void kill_block_super(struct super_block
*sb
)
662 struct block_device
*bdev
= sb
->s_bdev
;
663 generic_shutdown_super(sb
);
664 set_blocksize(bdev
, sb
->s_old_blocksize
);
665 close_bdev_excl(bdev
);
668 EXPORT_SYMBOL(kill_block_super
);
670 struct super_block
*get_sb_nodev(struct file_system_type
*fs_type
,
671 int flags
, void *data
,
672 int (*fill_super
)(struct super_block
*, void *, int))
675 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
682 error
= fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
684 up_write(&s
->s_umount
);
686 return ERR_PTR(error
);
688 s
->s_flags
|= MS_ACTIVE
;
692 EXPORT_SYMBOL(get_sb_nodev
);
694 static int compare_single(struct super_block
*s
, void *p
)
699 struct super_block
*get_sb_single(struct file_system_type
*fs_type
,
700 int flags
, void *data
,
701 int (*fill_super
)(struct super_block
*, void *, int))
703 struct super_block
*s
;
706 s
= sget(fs_type
, compare_single
, set_anon_super
, NULL
);
711 error
= fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
713 up_write(&s
->s_umount
);
715 return ERR_PTR(error
);
717 s
->s_flags
|= MS_ACTIVE
;
719 do_remount_sb(s
, flags
, data
, 0);
723 EXPORT_SYMBOL(get_sb_single
);
726 do_kern_mount(const char *fstype
, int flags
, const char *name
, void *data
)
728 struct file_system_type
*type
= get_fs_type(fstype
);
729 struct super_block
*sb
= ERR_PTR(-ENOMEM
);
730 struct vfsmount
*mnt
;
732 char *secdata
= NULL
;
735 return ERR_PTR(-ENODEV
);
737 mnt
= alloc_vfsmnt(name
);
742 secdata
= alloc_secdata();
744 sb
= ERR_PTR(-ENOMEM
);
748 error
= security_sb_copy_data(type
, data
, secdata
);
751 goto out_free_secdata
;
755 sb
= type
->get_sb(type
, flags
, name
, data
);
757 goto out_free_secdata
;
758 error
= security_sb_kern_mount(sb
, secdata
);
762 mnt
->mnt_root
= dget(sb
->s_root
);
763 mnt
->mnt_mountpoint
= sb
->s_root
;
764 mnt
->mnt_parent
= mnt
;
765 up_write(&sb
->s_umount
);
766 put_filesystem(type
);
769 up_write(&sb
->s_umount
);
770 deactivate_super(sb
);
773 free_secdata(secdata
);
777 put_filesystem(type
);
778 return (struct vfsmount
*)sb
;
781 struct vfsmount
*kern_mount(struct file_system_type
*type
)
783 return do_kern_mount(type
->name
, 0, type
->name
, NULL
);
786 EXPORT_SYMBOL(kern_mount
);