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 <asm/uaccess.h>
43 void get_filesystem(struct file_system_type
*fs
);
44 void put_filesystem(struct file_system_type
*fs
);
45 struct file_system_type
*get_fs_type(const char *name
);
47 LIST_HEAD(super_blocks
);
48 DEFINE_SPINLOCK(sb_lock
);
51 * alloc_super - create new superblock
53 * Allocates and initializes a new &struct super_block. alloc_super()
54 * returns a pointer new superblock or %NULL if allocation had failed.
56 static struct super_block
*alloc_super(void)
58 struct super_block
*s
= kmalloc(sizeof(struct super_block
), GFP_USER
);
59 static struct super_operations default_op
;
62 memset(s
, 0, sizeof(struct super_block
));
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 sema_init(&s
->s_lock
, 1);
76 down_write(&s
->s_umount
);
78 atomic_set(&s
->s_active
, 1);
79 sema_init(&s
->s_vfs_rename_sem
,1);
80 sema_init(&s
->s_dquot
.dqio_sem
, 1);
81 sema_init(&s
->s_dquot
.dqonoff_sem
, 1);
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
);
174 down_write(&s
->s_umount
);
181 EXPORT_SYMBOL(deactivate_super
);
184 * grab_super - acquire an active reference
185 * @s: reference we are trying to make active
187 * Tries to acquire an active reference. grab_super() is used when we
188 * had just found a superblock in super_blocks or fs_type->fs_supers
189 * and want to turn it into a full-blown active reference. grab_super()
190 * is called with sb_lock held and drops it. Returns 1 in case of
191 * success, 0 if we had failed (superblock contents was already dead or
192 * dying when grab_super() had been called).
194 static int grab_super(struct super_block
*s
)
197 spin_unlock(&sb_lock
);
198 down_write(&s
->s_umount
);
201 if (s
->s_count
> S_BIAS
) {
202 atomic_inc(&s
->s_active
);
204 spin_unlock(&sb_lock
);
207 spin_unlock(&sb_lock
);
209 up_write(&s
->s_umount
);
216 * generic_shutdown_super - common helper for ->kill_sb()
217 * @sb: superblock to kill
219 * generic_shutdown_super() does all fs-independent work on superblock
220 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
221 * that need destruction out of superblock, call generic_shutdown_super()
222 * and release aforementioned objects. Note: dentries and inodes _are_
223 * taken care of and do not need specific handling.
225 void generic_shutdown_super(struct super_block
*sb
)
227 struct dentry
*root
= sb
->s_root
;
228 struct super_operations
*sop
= sb
->s_op
;
232 shrink_dcache_parent(root
);
233 shrink_dcache_anon(&sb
->s_anon
);
237 sb
->s_flags
&= ~MS_ACTIVE
;
238 /* bad name - it should be evict_inodes() */
239 invalidate_inodes(sb
);
242 if (sop
->write_super
&& sb
->s_dirt
)
243 sop
->write_super(sb
);
247 /* Forget any remaining inodes */
248 if (invalidate_inodes(sb
)) {
249 printk("VFS: Busy inodes after unmount. "
250 "Self-destruct in 5 seconds. Have a nice day...\n");
257 /* should be initialized for __put_super_and_need_restart() */
258 list_del_init(&sb
->s_list
);
259 list_del(&sb
->s_instances
);
260 spin_unlock(&sb_lock
);
261 up_write(&sb
->s_umount
);
264 EXPORT_SYMBOL(generic_shutdown_super
);
267 * sget - find or create a superblock
268 * @type: filesystem type superblock should belong to
269 * @test: comparison callback
270 * @set: setup callback
271 * @data: argument to each of them
273 struct super_block
*sget(struct file_system_type
*type
,
274 int (*test
)(struct super_block
*,void *),
275 int (*set
)(struct super_block
*,void *),
278 struct super_block
*s
= NULL
;
284 if (test
) list_for_each(p
, &type
->fs_supers
) {
285 struct super_block
*old
;
286 old
= list_entry(p
, struct super_block
, s_instances
);
287 if (!test(old
, data
))
289 if (!grab_super(old
))
296 spin_unlock(&sb_lock
);
299 return ERR_PTR(-ENOMEM
);
305 spin_unlock(&sb_lock
);
310 strlcpy(s
->s_id
, type
->name
, sizeof(s
->s_id
));
311 list_add_tail(&s
->s_list
, &super_blocks
);
312 list_add(&s
->s_instances
, &type
->fs_supers
);
313 spin_unlock(&sb_lock
);
314 get_filesystem(type
);
320 void drop_super(struct super_block
*sb
)
322 up_read(&sb
->s_umount
);
326 EXPORT_SYMBOL(drop_super
);
328 static inline void write_super(struct super_block
*sb
)
331 if (sb
->s_root
&& sb
->s_dirt
)
332 if (sb
->s_op
->write_super
)
333 sb
->s_op
->write_super(sb
);
338 * Note: check the dirty flag before waiting, so we don't
339 * hold up the sync while mounting a device. (The newly
340 * mounted device won't need syncing.)
342 void sync_supers(void)
344 struct super_block
*sb
;
348 list_for_each_entry(sb
, &super_blocks
, s_list
) {
351 spin_unlock(&sb_lock
);
352 down_read(&sb
->s_umount
);
354 up_read(&sb
->s_umount
);
356 if (__put_super_and_need_restart(sb
))
360 spin_unlock(&sb_lock
);
364 * Call the ->sync_fs super_op against all filesytems which are r/w and
365 * which implement it.
367 * This operation is careful to avoid the livelock which could easily happen
368 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
369 * is used only here. We set it against all filesystems and then clear it as
370 * we sync them. So redirtied filesystems are skipped.
372 * But if process A is currently running sync_filesytems and then process B
373 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
374 * flags again, which will cause process A to resync everything. Fix that with
377 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
379 void sync_filesystems(int wait
)
381 struct super_block
*sb
;
382 static DECLARE_MUTEX(mutex
);
384 down(&mutex
); /* Could be down_interruptible */
386 list_for_each_entry(sb
, &super_blocks
, s_list
) {
387 if (!sb
->s_op
->sync_fs
)
389 if (sb
->s_flags
& MS_RDONLY
)
391 sb
->s_need_sync_fs
= 1;
395 list_for_each_entry(sb
, &super_blocks
, s_list
) {
396 if (!sb
->s_need_sync_fs
)
398 sb
->s_need_sync_fs
= 0;
399 if (sb
->s_flags
& MS_RDONLY
)
400 continue; /* hm. Was remounted r/o meanwhile */
402 spin_unlock(&sb_lock
);
403 down_read(&sb
->s_umount
);
404 if (sb
->s_root
&& (wait
|| sb
->s_dirt
))
405 sb
->s_op
->sync_fs(sb
, wait
);
406 up_read(&sb
->s_umount
);
407 /* restart only when sb is no longer on the list */
409 if (__put_super_and_need_restart(sb
))
412 spin_unlock(&sb_lock
);
417 * get_super - get the superblock of a device
418 * @bdev: device to get the superblock for
420 * Scans the superblock list and finds the superblock of the file system
421 * mounted on the device given. %NULL is returned if no match is found.
424 struct super_block
* get_super(struct block_device
*bdev
)
426 struct super_block
*sb
;
433 list_for_each_entry(sb
, &super_blocks
, s_list
) {
434 if (sb
->s_bdev
== bdev
) {
436 spin_unlock(&sb_lock
);
437 down_read(&sb
->s_umount
);
440 up_read(&sb
->s_umount
);
441 /* restart only when sb is no longer on the list */
443 if (__put_super_and_need_restart(sb
))
447 spin_unlock(&sb_lock
);
451 EXPORT_SYMBOL(get_super
);
453 struct super_block
* user_get_super(dev_t dev
)
455 struct super_block
*sb
;
459 list_for_each_entry(sb
, &super_blocks
, s_list
) {
460 if (sb
->s_dev
== dev
) {
462 spin_unlock(&sb_lock
);
463 down_read(&sb
->s_umount
);
466 up_read(&sb
->s_umount
);
467 /* restart only when sb is no longer on the list */
469 if (__put_super_and_need_restart(sb
))
473 spin_unlock(&sb_lock
);
477 EXPORT_SYMBOL(user_get_super
);
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
, &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_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
, NULL
);
675 kobject_uevent(&bdev
->bd_disk
->kobj
, action
, NULL
);
679 struct super_block
*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))
683 struct block_device
*bdev
;
684 struct super_block
*s
;
687 bdev
= open_bdev_excl(dev_name
, flags
, fs_type
);
689 return (struct super_block
*)bdev
;
692 * once the super is inserted into the list by sget, s_umount
693 * will protect the lockfs code from trying to start a snapshot
694 * while we are mounting
696 down(&bdev
->bd_mount_sem
);
697 s
= sget(fs_type
, test_bdev_super
, set_bdev_super
, bdev
);
698 up(&bdev
->bd_mount_sem
);
703 if ((flags
^ s
->s_flags
) & MS_RDONLY
) {
704 up_write(&s
->s_umount
);
710 char b
[BDEVNAME_SIZE
];
713 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
714 s
->s_old_blocksize
= block_size(bdev
);
715 sb_set_blocksize(s
, s
->s_old_blocksize
);
716 error
= fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
718 up_write(&s
->s_umount
);
722 s
->s_flags
|= MS_ACTIVE
;
723 bdev_uevent(bdev
, KOBJ_MOUNT
);
730 close_bdev_excl(bdev
);
734 EXPORT_SYMBOL(get_sb_bdev
);
736 void kill_block_super(struct super_block
*sb
)
738 struct block_device
*bdev
= sb
->s_bdev
;
740 bdev_uevent(bdev
, KOBJ_UMOUNT
);
741 generic_shutdown_super(sb
);
743 close_bdev_excl(bdev
);
746 EXPORT_SYMBOL(kill_block_super
);
748 struct super_block
*get_sb_nodev(struct file_system_type
*fs_type
,
749 int flags
, void *data
,
750 int (*fill_super
)(struct super_block
*, void *, int))
753 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
760 error
= fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
762 up_write(&s
->s_umount
);
764 return ERR_PTR(error
);
766 s
->s_flags
|= MS_ACTIVE
;
770 EXPORT_SYMBOL(get_sb_nodev
);
772 static int compare_single(struct super_block
*s
, void *p
)
777 struct super_block
*get_sb_single(struct file_system_type
*fs_type
,
778 int flags
, void *data
,
779 int (*fill_super
)(struct super_block
*, void *, int))
781 struct super_block
*s
;
784 s
= sget(fs_type
, compare_single
, set_anon_super
, NULL
);
789 error
= fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
791 up_write(&s
->s_umount
);
793 return ERR_PTR(error
);
795 s
->s_flags
|= MS_ACTIVE
;
797 do_remount_sb(s
, flags
, data
, 0);
801 EXPORT_SYMBOL(get_sb_single
);
804 do_kern_mount(const char *fstype
, int flags
, const char *name
, void *data
)
806 struct file_system_type
*type
= get_fs_type(fstype
);
807 struct super_block
*sb
= ERR_PTR(-ENOMEM
);
808 struct vfsmount
*mnt
;
810 char *secdata
= NULL
;
813 return ERR_PTR(-ENODEV
);
815 mnt
= alloc_vfsmnt(name
);
820 secdata
= alloc_secdata();
822 sb
= ERR_PTR(-ENOMEM
);
826 error
= security_sb_copy_data(type
, data
, secdata
);
829 goto out_free_secdata
;
833 sb
= type
->get_sb(type
, flags
, name
, data
);
835 goto out_free_secdata
;
836 error
= security_sb_kern_mount(sb
, secdata
);
840 mnt
->mnt_root
= dget(sb
->s_root
);
841 mnt
->mnt_mountpoint
= sb
->s_root
;
842 mnt
->mnt_parent
= mnt
;
843 mnt
->mnt_namespace
= current
->namespace;
844 up_write(&sb
->s_umount
);
845 free_secdata(secdata
);
846 put_filesystem(type
);
849 up_write(&sb
->s_umount
);
850 deactivate_super(sb
);
853 free_secdata(secdata
);
857 put_filesystem(type
);
858 return (struct vfsmount
*)sb
;
861 EXPORT_SYMBOL_GPL(do_kern_mount
);
863 struct vfsmount
*kern_mount(struct file_system_type
*type
)
865 return do_kern_mount(type
->name
, 0, type
->name
, NULL
);
868 EXPORT_SYMBOL(kern_mount
);