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/syscalls.h>
36 #include <linux/vfs.h>
37 #include <linux/writeback.h> /* for the emergency remount stuff */
38 #include <linux/idr.h>
39 #include <linux/kobject.h>
40 #include <linux/mutex.h>
41 #include <asm/uaccess.h>
44 void get_filesystem(struct file_system_type
*fs
);
45 void put_filesystem(struct file_system_type
*fs
);
46 struct file_system_type
*get_fs_type(const char *name
);
48 LIST_HEAD(super_blocks
);
49 DEFINE_SPINLOCK(sb_lock
);
52 * alloc_super - create new superblock
54 * Allocates and initializes a new &struct super_block. alloc_super()
55 * returns a pointer new superblock or %NULL if allocation had failed.
57 static struct super_block
*alloc_super(void)
59 struct super_block
*s
= kzalloc(sizeof(struct super_block
), GFP_USER
);
60 static struct super_operations default_op
;
63 if (security_sb_alloc(s
)) {
68 INIT_LIST_HEAD(&s
->s_dirty
);
69 INIT_LIST_HEAD(&s
->s_io
);
70 INIT_LIST_HEAD(&s
->s_files
);
71 INIT_LIST_HEAD(&s
->s_instances
);
72 INIT_HLIST_HEAD(&s
->s_anon
);
73 INIT_LIST_HEAD(&s
->s_inodes
);
74 init_rwsem(&s
->s_umount
);
75 mutex_init(&s
->s_lock
);
76 down_write(&s
->s_umount
);
78 atomic_set(&s
->s_active
, 1);
79 mutex_init(&s
->s_vfs_rename_mutex
);
80 mutex_init(&s
->s_dquot
.dqio_mutex
);
81 mutex_init(&s
->s_dquot
.dqonoff_mutex
);
82 init_rwsem(&s
->s_dquot
.dqptr_sem
);
83 init_waitqueue_head(&s
->s_wait_unfrozen
);
84 s
->s_maxbytes
= MAX_NON_LFS
;
85 s
->dq_op
= sb_dquot_ops
;
86 s
->s_qcop
= sb_quotactl_ops
;
87 s
->s_op
= &default_op
;
88 s
->s_time_gran
= 1000000000;
95 * destroy_super - frees a superblock
96 * @s: superblock to free
100 static inline void destroy_super(struct super_block
*s
)
106 /* Superblock refcounting */
109 * Drop a superblock's refcount. Returns non-zero if the superblock was
110 * destroyed. The caller must hold sb_lock.
112 int __put_super(struct super_block
*sb
)
116 if (!--sb
->s_count
) {
124 * Drop a superblock's refcount.
125 * Returns non-zero if the superblock is about to be destroyed and
126 * at least is already removed from super_blocks list, so if we are
127 * making a loop through super blocks then we need to restart.
128 * The caller must hold sb_lock.
130 int __put_super_and_need_restart(struct super_block
*sb
)
132 /* check for race with generic_shutdown_super() */
133 if (list_empty(&sb
->s_list
)) {
134 /* super block is removed, need to restart... */
138 /* can't be the last, since s_list is still in use */
140 BUG_ON(sb
->s_count
== 0);
145 * put_super - drop a temporary reference to superblock
146 * @sb: superblock in question
148 * Drops a temporary reference, frees superblock if there's no
151 static void put_super(struct super_block
*sb
)
155 spin_unlock(&sb_lock
);
160 * deactivate_super - drop an active reference to superblock
161 * @s: superblock to deactivate
163 * Drops an active reference to superblock, acquiring a temprory one if
164 * there is no active references left. In that case we lock superblock,
165 * tell fs driver to shut it down and drop the temporary reference we
168 void deactivate_super(struct super_block
*s
)
170 struct file_system_type
*fs
= s
->s_type
;
171 if (atomic_dec_and_lock(&s
->s_active
, &sb_lock
)) {
172 s
->s_count
-= S_BIAS
-1;
173 spin_unlock(&sb_lock
);
175 down_write(&s
->s_umount
);
182 EXPORT_SYMBOL(deactivate_super
);
185 * grab_super - acquire an active reference
186 * @s: reference we are trying to make active
188 * Tries to acquire an active reference. grab_super() is used when we
189 * had just found a superblock in super_blocks or fs_type->fs_supers
190 * and want to turn it into a full-blown active reference. grab_super()
191 * is called with sb_lock held and drops it. Returns 1 in case of
192 * success, 0 if we had failed (superblock contents was already dead or
193 * dying when grab_super() had been called).
195 static int grab_super(struct super_block
*s
)
198 spin_unlock(&sb_lock
);
199 down_write(&s
->s_umount
);
202 if (s
->s_count
> S_BIAS
) {
203 atomic_inc(&s
->s_active
);
205 spin_unlock(&sb_lock
);
208 spin_unlock(&sb_lock
);
210 up_write(&s
->s_umount
);
217 * generic_shutdown_super - common helper for ->kill_sb()
218 * @sb: superblock to kill
220 * generic_shutdown_super() does all fs-independent work on superblock
221 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
222 * that need destruction out of superblock, call generic_shutdown_super()
223 * and release aforementioned objects. Note: dentries and inodes _are_
224 * taken care of and do not need specific handling.
226 void generic_shutdown_super(struct super_block
*sb
)
228 struct dentry
*root
= sb
->s_root
;
229 struct super_operations
*sop
= sb
->s_op
;
233 shrink_dcache_parent(root
);
234 shrink_dcache_sb(sb
);
238 sb
->s_flags
&= ~MS_ACTIVE
;
239 /* bad name - it should be evict_inodes() */
240 invalidate_inodes(sb
);
243 if (sop
->write_super
&& sb
->s_dirt
)
244 sop
->write_super(sb
);
248 /* Forget any remaining inodes */
249 if (invalidate_inodes(sb
)) {
250 printk("VFS: Busy inodes after unmount of %s. "
251 "Self-destruct in 5 seconds. Have a nice day...\n",
259 /* should be initialized for __put_super_and_need_restart() */
260 list_del_init(&sb
->s_list
);
261 list_del(&sb
->s_instances
);
262 spin_unlock(&sb_lock
);
263 up_write(&sb
->s_umount
);
266 EXPORT_SYMBOL(generic_shutdown_super
);
269 * sget - find or create a superblock
270 * @type: filesystem type superblock should belong to
271 * @test: comparison callback
272 * @set: setup callback
273 * @data: argument to each of them
275 struct super_block
*sget(struct file_system_type
*type
,
276 int (*test
)(struct super_block
*,void *),
277 int (*set
)(struct super_block
*,void *),
280 struct super_block
*s
= NULL
;
286 if (test
) list_for_each(p
, &type
->fs_supers
) {
287 struct super_block
*old
;
288 old
= list_entry(p
, struct super_block
, s_instances
);
289 if (!test(old
, data
))
291 if (!grab_super(old
))
298 spin_unlock(&sb_lock
);
301 return ERR_PTR(-ENOMEM
);
307 spin_unlock(&sb_lock
);
312 strlcpy(s
->s_id
, type
->name
, sizeof(s
->s_id
));
313 list_add_tail(&s
->s_list
, &super_blocks
);
314 list_add(&s
->s_instances
, &type
->fs_supers
);
315 spin_unlock(&sb_lock
);
316 get_filesystem(type
);
322 void drop_super(struct super_block
*sb
)
324 up_read(&sb
->s_umount
);
328 EXPORT_SYMBOL(drop_super
);
330 static inline void write_super(struct super_block
*sb
)
333 if (sb
->s_root
&& sb
->s_dirt
)
334 if (sb
->s_op
->write_super
)
335 sb
->s_op
->write_super(sb
);
340 * Note: check the dirty flag before waiting, so we don't
341 * hold up the sync while mounting a device. (The newly
342 * mounted device won't need syncing.)
344 void sync_supers(void)
346 struct super_block
*sb
;
350 list_for_each_entry(sb
, &super_blocks
, s_list
) {
353 spin_unlock(&sb_lock
);
354 down_read(&sb
->s_umount
);
356 up_read(&sb
->s_umount
);
358 if (__put_super_and_need_restart(sb
))
362 spin_unlock(&sb_lock
);
366 * Call the ->sync_fs super_op against all filesytems which are r/w and
367 * which implement it.
369 * This operation is careful to avoid the livelock which could easily happen
370 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
371 * is used only here. We set it against all filesystems and then clear it as
372 * we sync them. So redirtied filesystems are skipped.
374 * But if process A is currently running sync_filesytems and then process B
375 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
376 * flags again, which will cause process A to resync everything. Fix that with
379 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
381 void sync_filesystems(int wait
)
383 struct super_block
*sb
;
384 static DEFINE_MUTEX(mutex
);
386 mutex_lock(&mutex
); /* Could be down_interruptible */
388 list_for_each_entry(sb
, &super_blocks
, s_list
) {
389 if (!sb
->s_op
->sync_fs
)
391 if (sb
->s_flags
& MS_RDONLY
)
393 sb
->s_need_sync_fs
= 1;
397 list_for_each_entry(sb
, &super_blocks
, s_list
) {
398 if (!sb
->s_need_sync_fs
)
400 sb
->s_need_sync_fs
= 0;
401 if (sb
->s_flags
& MS_RDONLY
)
402 continue; /* hm. Was remounted r/o meanwhile */
404 spin_unlock(&sb_lock
);
405 down_read(&sb
->s_umount
);
406 if (sb
->s_root
&& (wait
|| sb
->s_dirt
))
407 sb
->s_op
->sync_fs(sb
, wait
);
408 up_read(&sb
->s_umount
);
409 /* restart only when sb is no longer on the list */
411 if (__put_super_and_need_restart(sb
))
414 spin_unlock(&sb_lock
);
415 mutex_unlock(&mutex
);
419 * get_super - get the superblock of a device
420 * @bdev: device to get the superblock for
422 * Scans the superblock list and finds the superblock of the file system
423 * mounted on the device given. %NULL is returned if no match is found.
426 struct super_block
* get_super(struct block_device
*bdev
)
428 struct super_block
*sb
;
435 list_for_each_entry(sb
, &super_blocks
, s_list
) {
436 if (sb
->s_bdev
== bdev
) {
438 spin_unlock(&sb_lock
);
439 down_read(&sb
->s_umount
);
442 up_read(&sb
->s_umount
);
443 /* restart only when sb is no longer on the list */
445 if (__put_super_and_need_restart(sb
))
449 spin_unlock(&sb_lock
);
453 EXPORT_SYMBOL(get_super
);
455 struct super_block
* user_get_super(dev_t dev
)
457 struct super_block
*sb
;
461 list_for_each_entry(sb
, &super_blocks
, s_list
) {
462 if (sb
->s_dev
== dev
) {
464 spin_unlock(&sb_lock
);
465 down_read(&sb
->s_umount
);
468 up_read(&sb
->s_umount
);
469 /* restart only when sb is no longer on the list */
471 if (__put_super_and_need_restart(sb
))
475 spin_unlock(&sb_lock
);
479 asmlinkage
long sys_ustat(unsigned dev
, struct ustat __user
* ubuf
)
481 struct super_block
*s
;
486 s
= user_get_super(new_decode_dev(dev
));
489 err
= vfs_statfs(s
->s_root
, &sbuf
);
494 memset(&tmp
,0,sizeof(struct ustat
));
495 tmp
.f_tfree
= sbuf
.f_bfree
;
496 tmp
.f_tinode
= sbuf
.f_ffree
;
498 err
= copy_to_user(ubuf
,&tmp
,sizeof(struct ustat
)) ? -EFAULT
: 0;
505 * @sb: superblock in question
507 * All files are marked read/only. We don't care about pending
508 * delete files so this should be used in 'force' mode only
511 static void mark_files_ro(struct super_block
*sb
)
516 list_for_each_entry(f
, &sb
->s_files
, f_u
.fu_list
) {
517 if (S_ISREG(f
->f_dentry
->d_inode
->i_mode
) && file_count(f
))
518 f
->f_mode
&= ~FMODE_WRITE
;
524 * do_remount_sb - asks filesystem to change mount options.
525 * @sb: superblock in question
526 * @flags: numeric part of options
527 * @data: the rest of options
528 * @force: whether or not to force the change
530 * Alters the mount options of a mounted file system.
532 int do_remount_sb(struct super_block
*sb
, int flags
, void *data
, int force
)
536 if (!(flags
& MS_RDONLY
) && bdev_read_only(sb
->s_bdev
))
538 if (flags
& MS_RDONLY
)
540 shrink_dcache_sb(sb
);
543 /* If we are remounting RDONLY and current sb is read/write,
544 make sure there are no rw files opened */
545 if ((flags
& MS_RDONLY
) && !(sb
->s_flags
& MS_RDONLY
)) {
548 else if (!fs_may_remount_ro(sb
))
552 if (sb
->s_op
->remount_fs
) {
554 retval
= sb
->s_op
->remount_fs(sb
, &flags
, data
);
559 sb
->s_flags
= (sb
->s_flags
& ~MS_RMT_MASK
) | (flags
& MS_RMT_MASK
);
563 static void do_emergency_remount(unsigned long foo
)
565 struct super_block
*sb
;
568 list_for_each_entry(sb
, &super_blocks
, s_list
) {
570 spin_unlock(&sb_lock
);
571 down_read(&sb
->s_umount
);
572 if (sb
->s_root
&& sb
->s_bdev
&& !(sb
->s_flags
& MS_RDONLY
)) {
574 * ->remount_fs needs lock_kernel().
576 * What lock protects sb->s_flags??
579 do_remount_sb(sb
, MS_RDONLY
, NULL
, 1);
585 spin_unlock(&sb_lock
);
586 printk("Emergency Remount complete\n");
589 void emergency_remount(void)
591 pdflush_operation(do_emergency_remount
, 0);
595 * Unnamed block devices are dummy devices used by virtual
596 * filesystems which don't use real block-devices. -- jrs
599 static struct idr unnamed_dev_idr
;
600 static DEFINE_SPINLOCK(unnamed_dev_lock
);/* protects the above */
602 int set_anon_super(struct super_block
*s
, void *data
)
608 if (idr_pre_get(&unnamed_dev_idr
, GFP_ATOMIC
) == 0)
610 spin_lock(&unnamed_dev_lock
);
611 error
= idr_get_new(&unnamed_dev_idr
, NULL
, &dev
);
612 spin_unlock(&unnamed_dev_lock
);
613 if (error
== -EAGAIN
)
614 /* We raced and lost with another CPU. */
619 if ((dev
& MAX_ID_MASK
) == (1 << MINORBITS
)) {
620 spin_lock(&unnamed_dev_lock
);
621 idr_remove(&unnamed_dev_idr
, dev
);
622 spin_unlock(&unnamed_dev_lock
);
625 s
->s_dev
= MKDEV(0, dev
& MINORMASK
);
629 EXPORT_SYMBOL(set_anon_super
);
631 void kill_anon_super(struct super_block
*sb
)
633 int slot
= MINOR(sb
->s_dev
);
635 generic_shutdown_super(sb
);
636 spin_lock(&unnamed_dev_lock
);
637 idr_remove(&unnamed_dev_idr
, slot
);
638 spin_unlock(&unnamed_dev_lock
);
641 EXPORT_SYMBOL(kill_anon_super
);
643 void __init
unnamed_dev_init(void)
645 idr_init(&unnamed_dev_idr
);
648 void kill_litter_super(struct super_block
*sb
)
651 d_genocide(sb
->s_root
);
655 EXPORT_SYMBOL(kill_litter_super
);
657 static int set_bdev_super(struct super_block
*s
, void *data
)
660 s
->s_dev
= s
->s_bdev
->bd_dev
;
664 static int test_bdev_super(struct super_block
*s
, void *data
)
666 return (void *)s
->s_bdev
== data
;
669 static void bdev_uevent(struct block_device
*bdev
, enum kobject_action action
)
673 kobject_uevent(&bdev
->bd_part
->kobj
, action
);
675 kobject_uevent(&bdev
->bd_disk
->kobj
, action
);
679 int get_sb_bdev(struct file_system_type
*fs_type
,
680 int flags
, const char *dev_name
, void *data
,
681 int (*fill_super
)(struct super_block
*, void *, int),
682 struct vfsmount
*mnt
)
684 struct block_device
*bdev
;
685 struct super_block
*s
;
688 bdev
= open_bdev_excl(dev_name
, flags
, fs_type
);
690 return PTR_ERR(bdev
);
693 * once the super is inserted into the list by sget, s_umount
694 * will protect the lockfs code from trying to start a snapshot
695 * while we are mounting
697 mutex_lock(&bdev
->bd_mount_mutex
);
698 s
= sget(fs_type
, test_bdev_super
, set_bdev_super
, bdev
);
699 mutex_unlock(&bdev
->bd_mount_mutex
);
704 if ((flags
^ s
->s_flags
) & MS_RDONLY
) {
705 up_write(&s
->s_umount
);
711 close_bdev_excl(bdev
);
713 char b
[BDEVNAME_SIZE
];
716 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
717 sb_set_blocksize(s
, block_size(bdev
));
718 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
720 up_write(&s
->s_umount
);
725 s
->s_flags
|= MS_ACTIVE
;
726 bdev_uevent(bdev
, KOBJ_MOUNT
);
729 return simple_set_mnt(mnt
, s
);
734 close_bdev_excl(bdev
);
739 EXPORT_SYMBOL(get_sb_bdev
);
741 void kill_block_super(struct super_block
*sb
)
743 struct block_device
*bdev
= sb
->s_bdev
;
745 bdev_uevent(bdev
, KOBJ_UMOUNT
);
746 generic_shutdown_super(sb
);
748 close_bdev_excl(bdev
);
751 EXPORT_SYMBOL(kill_block_super
);
753 int get_sb_nodev(struct file_system_type
*fs_type
,
754 int flags
, void *data
,
755 int (*fill_super
)(struct super_block
*, void *, int),
756 struct vfsmount
*mnt
)
759 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
766 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
768 up_write(&s
->s_umount
);
772 s
->s_flags
|= MS_ACTIVE
;
773 return simple_set_mnt(mnt
, s
);
776 EXPORT_SYMBOL(get_sb_nodev
);
778 static int compare_single(struct super_block
*s
, void *p
)
783 int get_sb_single(struct file_system_type
*fs_type
,
784 int flags
, void *data
,
785 int (*fill_super
)(struct super_block
*, void *, int),
786 struct vfsmount
*mnt
)
788 struct super_block
*s
;
791 s
= sget(fs_type
, compare_single
, set_anon_super
, NULL
);
796 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
798 up_write(&s
->s_umount
);
802 s
->s_flags
|= MS_ACTIVE
;
804 do_remount_sb(s
, flags
, data
, 0);
805 return simple_set_mnt(mnt
, s
);
808 EXPORT_SYMBOL(get_sb_single
);
811 vfs_kern_mount(struct file_system_type
*type
, int flags
, const char *name
, void *data
)
813 struct vfsmount
*mnt
;
814 char *secdata
= NULL
;
818 return ERR_PTR(-ENODEV
);
821 mnt
= alloc_vfsmnt(name
);
826 secdata
= alloc_secdata();
830 error
= security_sb_copy_data(type
, data
, secdata
);
832 goto out_free_secdata
;
835 error
= type
->get_sb(type
, flags
, name
, data
, mnt
);
837 goto out_free_secdata
;
839 error
= security_sb_kern_mount(mnt
->mnt_sb
, secdata
);
843 mnt
->mnt_mountpoint
= mnt
->mnt_root
;
844 mnt
->mnt_parent
= mnt
;
845 up_write(&mnt
->mnt_sb
->s_umount
);
846 free_secdata(secdata
);
850 up_write(&mnt
->mnt_sb
->s_umount
);
851 deactivate_super(mnt
->mnt_sb
);
853 free_secdata(secdata
);
857 return ERR_PTR(error
);
860 EXPORT_SYMBOL_GPL(vfs_kern_mount
);
863 do_kern_mount(const char *fstype
, int flags
, const char *name
, void *data
)
865 struct file_system_type
*type
= get_fs_type(fstype
);
866 struct vfsmount
*mnt
;
868 return ERR_PTR(-ENODEV
);
869 mnt
= vfs_kern_mount(type
, flags
, name
, data
);
870 put_filesystem(type
);
874 EXPORT_SYMBOL_GPL(do_kern_mount
);
876 struct vfsmount
*kern_mount(struct file_system_type
*type
)
878 return vfs_kern_mount(type
, 0, type
->name
, NULL
);
881 EXPORT_SYMBOL(kern_mount
);