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 init_waitqueue_head(&s
->s_wait_unfrozen
);
81 s
->s_maxbytes
= MAX_NON_LFS
;
82 s
->dq_op
= sb_dquot_ops
;
83 s
->s_qcop
= sb_quotactl_ops
;
84 s
->s_op
= &default_op
;
91 * destroy_super - frees a superblock
92 * @s: superblock to free
96 static inline void destroy_super(struct super_block
*s
)
102 /* Superblock refcounting */
105 * Drop a superblock's refcount. Returns non-zero if the superblock was
106 * destroyed. The caller must hold sb_lock.
108 int __put_super(struct super_block
*sb
)
112 if (!--sb
->s_count
) {
120 * Drop a superblock's refcount.
121 * Returns non-zero if the superblock is about to be destroyed and
122 * at least is already removed from super_blocks list, so if we are
123 * making a loop through super blocks then we need to restart.
124 * The caller must hold sb_lock.
126 int __put_super_and_need_restart(struct super_block
*sb
)
128 /* check for race with generic_shutdown_super() */
129 if (list_empty(&sb
->s_list
)) {
130 /* super block is removed, need to restart... */
134 /* can't be the last, since s_list is still in use */
136 BUG_ON(sb
->s_count
== 0);
141 * put_super - drop a temporary reference to superblock
142 * @s: superblock in question
144 * Drops a temporary reference, frees superblock if there's no
147 static void put_super(struct super_block
*sb
)
151 spin_unlock(&sb_lock
);
156 * deactivate_super - drop an active reference to superblock
157 * @s: superblock to deactivate
159 * Drops an active reference to superblock, acquiring a temprory one if
160 * there is no active references left. In that case we lock superblock,
161 * tell fs driver to shut it down and drop the temporary reference we
164 void deactivate_super(struct super_block
*s
)
166 struct file_system_type
*fs
= s
->s_type
;
167 if (atomic_dec_and_lock(&s
->s_active
, &sb_lock
)) {
168 s
->s_count
-= S_BIAS
-1;
169 spin_unlock(&sb_lock
);
170 down_write(&s
->s_umount
);
177 EXPORT_SYMBOL(deactivate_super
);
180 * grab_super - acquire an active reference
181 * @s: reference we are trying to make active
183 * Tries to acquire an active reference. grab_super() is used when we
184 * had just found a superblock in super_blocks or fs_type->fs_supers
185 * and want to turn it into a full-blown active reference. grab_super()
186 * is called with sb_lock held and drops it. Returns 1 in case of
187 * success, 0 if we had failed (superblock contents was already dead or
188 * dying when grab_super() had been called).
190 static int grab_super(struct super_block
*s
)
193 spin_unlock(&sb_lock
);
194 down_write(&s
->s_umount
);
197 if (s
->s_count
> S_BIAS
) {
198 atomic_inc(&s
->s_active
);
200 spin_unlock(&sb_lock
);
203 spin_unlock(&sb_lock
);
205 up_write(&s
->s_umount
);
212 * generic_shutdown_super - common helper for ->kill_sb()
213 * @sb: superblock to kill
215 * generic_shutdown_super() does all fs-independent work on superblock
216 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
217 * that need destruction out of superblock, call generic_shutdown_super()
218 * and release aforementioned objects. Note: dentries and inodes _are_
219 * taken care of and do not need specific handling.
221 void generic_shutdown_super(struct super_block
*sb
)
223 struct dentry
*root
= sb
->s_root
;
224 struct super_operations
*sop
= sb
->s_op
;
228 shrink_dcache_parent(root
);
229 shrink_dcache_anon(&sb
->s_anon
);
234 sb
->s_flags
&= ~MS_ACTIVE
;
235 /* bad name - it should be evict_inodes() */
236 invalidate_inodes(sb
);
238 if (sop
->write_super
&& sb
->s_dirt
)
239 sop
->write_super(sb
);
243 /* Forget any remaining inodes */
244 if (invalidate_inodes(sb
)) {
245 printk("VFS: Busy inodes after unmount. "
246 "Self-destruct in 5 seconds. Have a nice day...\n");
253 /* should be initialized for __put_super_and_need_restart() */
254 list_del_init(&sb
->s_list
);
255 list_del(&sb
->s_instances
);
256 spin_unlock(&sb_lock
);
257 up_write(&sb
->s_umount
);
260 EXPORT_SYMBOL(generic_shutdown_super
);
263 * sget - find or create a superblock
264 * @type: filesystem type superblock should belong to
265 * @test: comparison callback
266 * @set: setup callback
267 * @data: argument to each of them
269 struct super_block
*sget(struct file_system_type
*type
,
270 int (*test
)(struct super_block
*,void *),
271 int (*set
)(struct super_block
*,void *),
274 struct super_block
*s
= NULL
;
280 if (test
) list_for_each(p
, &type
->fs_supers
) {
281 struct super_block
*old
;
282 old
= list_entry(p
, struct super_block
, s_instances
);
283 if (!test(old
, data
))
285 if (!grab_super(old
))
292 spin_unlock(&sb_lock
);
295 return ERR_PTR(-ENOMEM
);
301 spin_unlock(&sb_lock
);
306 strlcpy(s
->s_id
, type
->name
, sizeof(s
->s_id
));
307 list_add_tail(&s
->s_list
, &super_blocks
);
308 list_add(&s
->s_instances
, &type
->fs_supers
);
309 spin_unlock(&sb_lock
);
310 get_filesystem(type
);
316 void drop_super(struct super_block
*sb
)
318 up_read(&sb
->s_umount
);
322 EXPORT_SYMBOL(drop_super
);
324 static inline void write_super(struct super_block
*sb
)
327 if (sb
->s_root
&& sb
->s_dirt
)
328 if (sb
->s_op
->write_super
)
329 sb
->s_op
->write_super(sb
);
334 * Note: check the dirty flag before waiting, so we don't
335 * hold up the sync while mounting a device. (The newly
336 * mounted device won't need syncing.)
338 void sync_supers(void)
340 struct super_block
* sb
;
343 sb
= sb_entry(super_blocks
.next
);
344 while (sb
!= sb_entry(&super_blocks
))
347 spin_unlock(&sb_lock
);
348 down_read(&sb
->s_umount
);
353 sb
= sb_entry(sb
->s_list
.next
);
354 spin_unlock(&sb_lock
);
358 * Call the ->sync_fs super_op against all filesytems which are r/w and
359 * which implement it.
361 * This operation is careful to avoid the livelock which could easily happen
362 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
363 * is used only here. We set it against all filesystems and then clear it as
364 * we sync them. So redirtied filesystems are skipped.
366 * But if process A is currently running sync_filesytems and then process B
367 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
368 * flags again, which will cause process A to resync everything. Fix that with
371 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
373 void sync_filesystems(int wait
)
375 struct super_block
*sb
;
376 static DECLARE_MUTEX(mutex
);
378 down(&mutex
); /* Could be down_interruptible */
380 for (sb
= sb_entry(super_blocks
.next
); sb
!= sb_entry(&super_blocks
);
381 sb
= sb_entry(sb
->s_list
.next
)) {
382 if (!sb
->s_op
->sync_fs
)
384 if (sb
->s_flags
& MS_RDONLY
)
386 sb
->s_need_sync_fs
= 1;
388 spin_unlock(&sb_lock
);
392 for (sb
= sb_entry(super_blocks
.next
); sb
!= sb_entry(&super_blocks
);
393 sb
= sb_entry(sb
->s_list
.next
)) {
394 if (!sb
->s_need_sync_fs
)
396 sb
->s_need_sync_fs
= 0;
397 if (sb
->s_flags
& MS_RDONLY
)
398 continue; /* hm. Was remounted r/o meanwhile */
400 spin_unlock(&sb_lock
);
401 down_read(&sb
->s_umount
);
402 if (sb
->s_root
&& (wait
|| sb
->s_dirt
))
403 sb
->s_op
->sync_fs(sb
, wait
);
407 spin_unlock(&sb_lock
);
412 * get_super - get the superblock of a device
413 * @bdev: device to get the superblock for
415 * Scans the superblock list and finds the superblock of the file system
416 * mounted on the device given. %NULL is returned if no match is found.
419 struct super_block
* get_super(struct block_device
*bdev
)
426 list_for_each(p
, &super_blocks
) {
427 struct super_block
*s
= sb_entry(p
);
428 if (s
->s_bdev
== bdev
) {
430 spin_unlock(&sb_lock
);
431 down_read(&s
->s_umount
);
438 spin_unlock(&sb_lock
);
442 EXPORT_SYMBOL(get_super
);
444 struct super_block
* user_get_super(dev_t dev
)
450 list_for_each(p
, &super_blocks
) {
451 struct super_block
*s
= sb_entry(p
);
452 if (s
->s_dev
== dev
) {
454 spin_unlock(&sb_lock
);
455 down_read(&s
->s_umount
);
462 spin_unlock(&sb_lock
);
466 EXPORT_SYMBOL(user_get_super
);
468 asmlinkage
long sys_ustat(unsigned dev
, struct ustat __user
* ubuf
)
470 struct super_block
*s
;
475 s
= user_get_super(new_decode_dev(dev
));
478 err
= vfs_statfs(s
, &sbuf
);
483 memset(&tmp
,0,sizeof(struct ustat
));
484 tmp
.f_tfree
= sbuf
.f_bfree
;
485 tmp
.f_tinode
= sbuf
.f_ffree
;
487 err
= copy_to_user(ubuf
,&tmp
,sizeof(struct ustat
)) ? -EFAULT
: 0;
494 * @sb: superblock in question
496 * All files are marked read/only. We don't care about pending
497 * delete files so this should be used in 'force' mode only
500 static void mark_files_ro(struct super_block
*sb
)
505 list_for_each_entry(f
, &sb
->s_files
, f_list
) {
506 if (S_ISREG(f
->f_dentry
->d_inode
->i_mode
) && file_count(f
))
507 f
->f_mode
&= ~FMODE_WRITE
;
513 * do_remount_sb - asks filesystem to change mount options.
514 * @sb: superblock in question
515 * @flags: numeric part of options
516 * @data: the rest of options
517 * @force: whether or not to force the change
519 * Alters the mount options of a mounted file system.
521 int do_remount_sb(struct super_block
*sb
, int flags
, void *data
, int force
)
525 if (!(flags
& MS_RDONLY
) && bdev_read_only(sb
->s_bdev
))
527 if (flags
& MS_RDONLY
)
529 shrink_dcache_sb(sb
);
532 /* If we are remounting RDONLY and current sb is read/write,
533 make sure there are no rw files opened */
534 if ((flags
& MS_RDONLY
) && !(sb
->s_flags
& MS_RDONLY
)) {
537 else if (!fs_may_remount_ro(sb
))
541 if (sb
->s_op
->remount_fs
) {
543 retval
= sb
->s_op
->remount_fs(sb
, &flags
, data
);
548 sb
->s_flags
= (sb
->s_flags
& ~MS_RMT_MASK
) | (flags
& MS_RMT_MASK
);
552 static void do_emergency_remount(unsigned long foo
)
554 struct super_block
*sb
;
557 list_for_each_entry(sb
, &super_blocks
, s_list
) {
559 spin_unlock(&sb_lock
);
560 down_read(&sb
->s_umount
);
561 if (sb
->s_root
&& sb
->s_bdev
&& !(sb
->s_flags
& MS_RDONLY
)) {
563 * ->remount_fs needs lock_kernel().
565 * What lock protects sb->s_flags??
568 do_remount_sb(sb
, MS_RDONLY
, NULL
, 1);
574 spin_unlock(&sb_lock
);
575 printk("Emergency Remount complete\n");
578 void emergency_remount(void)
580 pdflush_operation(do_emergency_remount
, 0);
584 * Unnamed block devices are dummy devices used by virtual
585 * filesystems which don't use real block-devices. -- jrs
588 static struct idr unnamed_dev_idr
;
589 static spinlock_t unnamed_dev_lock
= SPIN_LOCK_UNLOCKED
;/* protects the above */
591 int set_anon_super(struct super_block
*s
, void *data
)
597 if (idr_pre_get(&unnamed_dev_idr
, GFP_ATOMIC
) == 0)
599 spin_lock(&unnamed_dev_lock
);
600 error
= idr_get_new(&unnamed_dev_idr
, NULL
, &dev
);
601 spin_unlock(&unnamed_dev_lock
);
602 if (error
== -EAGAIN
)
603 /* We raced and lost with another CPU. */
608 if ((dev
& MAX_ID_MASK
) == (1 << MINORBITS
)) {
609 spin_lock(&unnamed_dev_lock
);
610 idr_remove(&unnamed_dev_idr
, dev
);
611 spin_unlock(&unnamed_dev_lock
);
614 s
->s_dev
= MKDEV(0, dev
& MINORMASK
);
618 EXPORT_SYMBOL(set_anon_super
);
620 void kill_anon_super(struct super_block
*sb
)
622 int slot
= MINOR(sb
->s_dev
);
624 generic_shutdown_super(sb
);
625 spin_lock(&unnamed_dev_lock
);
626 idr_remove(&unnamed_dev_idr
, slot
);
627 spin_unlock(&unnamed_dev_lock
);
630 EXPORT_SYMBOL(kill_anon_super
);
632 void __init
unnamed_dev_init(void)
634 idr_init(&unnamed_dev_idr
);
637 void kill_litter_super(struct super_block
*sb
)
640 d_genocide(sb
->s_root
);
644 EXPORT_SYMBOL(kill_litter_super
);
646 static int set_bdev_super(struct super_block
*s
, void *data
)
649 s
->s_dev
= s
->s_bdev
->bd_dev
;
653 static int test_bdev_super(struct super_block
*s
, void *data
)
655 return (void *)s
->s_bdev
== data
;
658 struct super_block
*get_sb_bdev(struct file_system_type
*fs_type
,
659 int flags
, const char *dev_name
, void *data
,
660 int (*fill_super
)(struct super_block
*, void *, int))
662 struct block_device
*bdev
;
663 struct super_block
*s
;
666 bdev
= open_bdev_excl(dev_name
, flags
, fs_type
);
668 return (struct super_block
*)bdev
;
671 * once the super is inserted into the list by sget, s_umount
672 * will protect the lockfs code from trying to start a snapshot
673 * while we are mounting
675 down(&bdev
->bd_mount_sem
);
676 s
= sget(fs_type
, test_bdev_super
, set_bdev_super
, bdev
);
677 up(&bdev
->bd_mount_sem
);
682 if ((flags
^ s
->s_flags
) & MS_RDONLY
) {
683 up_write(&s
->s_umount
);
689 char b
[BDEVNAME_SIZE
];
692 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
693 s
->s_old_blocksize
= block_size(bdev
);
694 sb_set_blocksize(s
, s
->s_old_blocksize
);
695 error
= fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
697 up_write(&s
->s_umount
);
701 s
->s_flags
|= MS_ACTIVE
;
707 close_bdev_excl(bdev
);
711 EXPORT_SYMBOL(get_sb_bdev
);
713 void kill_block_super(struct super_block
*sb
)
715 struct block_device
*bdev
= sb
->s_bdev
;
716 generic_shutdown_super(sb
);
717 set_blocksize(bdev
, sb
->s_old_blocksize
);
718 close_bdev_excl(bdev
);
721 EXPORT_SYMBOL(kill_block_super
);
723 struct super_block
*get_sb_nodev(struct file_system_type
*fs_type
,
724 int flags
, void *data
,
725 int (*fill_super
)(struct super_block
*, void *, int))
728 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
735 error
= fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
737 up_write(&s
->s_umount
);
739 return ERR_PTR(error
);
741 s
->s_flags
|= MS_ACTIVE
;
745 EXPORT_SYMBOL(get_sb_nodev
);
747 static int compare_single(struct super_block
*s
, void *p
)
752 struct super_block
*get_sb_single(struct file_system_type
*fs_type
,
753 int flags
, void *data
,
754 int (*fill_super
)(struct super_block
*, void *, int))
756 struct super_block
*s
;
759 s
= sget(fs_type
, compare_single
, set_anon_super
, NULL
);
764 error
= fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
766 up_write(&s
->s_umount
);
768 return ERR_PTR(error
);
770 s
->s_flags
|= MS_ACTIVE
;
772 do_remount_sb(s
, flags
, data
, 0);
776 EXPORT_SYMBOL(get_sb_single
);
779 do_kern_mount(const char *fstype
, int flags
, const char *name
, void *data
)
781 struct file_system_type
*type
= get_fs_type(fstype
);
782 struct super_block
*sb
= ERR_PTR(-ENOMEM
);
783 struct vfsmount
*mnt
;
785 char *secdata
= NULL
;
788 return ERR_PTR(-ENODEV
);
790 mnt
= alloc_vfsmnt(name
);
795 secdata
= alloc_secdata();
797 sb
= ERR_PTR(-ENOMEM
);
801 error
= security_sb_copy_data(type
, data
, secdata
);
804 goto out_free_secdata
;
808 sb
= type
->get_sb(type
, flags
, name
, data
);
810 goto out_free_secdata
;
811 error
= security_sb_kern_mount(sb
, secdata
);
815 mnt
->mnt_root
= dget(sb
->s_root
);
816 mnt
->mnt_mountpoint
= sb
->s_root
;
817 mnt
->mnt_parent
= mnt
;
818 mnt
->mnt_namespace
= current
->namespace;
819 up_write(&sb
->s_umount
);
820 put_filesystem(type
);
823 up_write(&sb
->s_umount
);
824 deactivate_super(sb
);
827 free_secdata(secdata
);
831 put_filesystem(type
);
832 return (struct vfsmount
*)sb
;
835 EXPORT_SYMBOL_GPL(do_kern_mount
);
837 struct vfsmount
*kern_mount(struct file_system_type
*type
)
839 return do_kern_mount(type
->name
, 0, type
->name
, NULL
);
842 EXPORT_SYMBOL(kern_mount
);