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/module.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/smp_lock.h>
27 #include <linux/acct.h>
28 #include <linux/blkdev.h>
29 #include <linux/quotaops.h>
30 #include <linux/namei.h>
31 #include <linux/mount.h>
32 #include <linux/security.h>
33 #include <linux/syscalls.h>
34 #include <linux/vfs.h>
35 #include <linux/writeback.h> /* for the emergency remount stuff */
36 #include <linux/idr.h>
37 #include <linux/kobject.h>
38 #include <linux/mutex.h>
39 #include <linux/file.h>
40 #include <asm/uaccess.h>
44 LIST_HEAD(super_blocks
);
45 DEFINE_SPINLOCK(sb_lock
);
48 * alloc_super - create new superblock
49 * @type: filesystem type superblock should belong to
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(struct file_system_type
*type
)
56 struct super_block
*s
= kzalloc(sizeof(struct super_block
), GFP_USER
);
57 static const struct super_operations default_op
;
60 if (security_sb_alloc(s
)) {
65 INIT_LIST_HEAD(&s
->s_files
);
66 INIT_LIST_HEAD(&s
->s_instances
);
67 INIT_HLIST_HEAD(&s
->s_anon
);
68 INIT_LIST_HEAD(&s
->s_inodes
);
69 INIT_LIST_HEAD(&s
->s_dentry_lru
);
70 init_rwsem(&s
->s_umount
);
71 mutex_init(&s
->s_lock
);
72 lockdep_set_class(&s
->s_umount
, &type
->s_umount_key
);
74 * The locking rules for s_lock are up to the
75 * filesystem. For example ext3fs has different
76 * lock ordering than usbfs:
78 lockdep_set_class(&s
->s_lock
, &type
->s_lock_key
);
80 * sget() can have s_umount recursion.
82 * When it cannot find a suitable sb, it allocates a new
83 * one (this one), and tries again to find a suitable old
86 * In case that succeeds, it will acquire the s_umount
87 * lock of the old one. Since these are clearly distrinct
88 * locks, and this object isn't exposed yet, there's no
91 * Annotate this by putting this lock in a different
94 down_write_nested(&s
->s_umount
, SINGLE_DEPTH_NESTING
);
96 atomic_set(&s
->s_active
, 1);
97 mutex_init(&s
->s_vfs_rename_mutex
);
98 mutex_init(&s
->s_dquot
.dqio_mutex
);
99 mutex_init(&s
->s_dquot
.dqonoff_mutex
);
100 init_rwsem(&s
->s_dquot
.dqptr_sem
);
101 init_waitqueue_head(&s
->s_wait_unfrozen
);
102 s
->s_maxbytes
= MAX_NON_LFS
;
103 s
->dq_op
= sb_dquot_ops
;
104 s
->s_qcop
= sb_quotactl_ops
;
105 s
->s_op
= &default_op
;
106 s
->s_time_gran
= 1000000000;
113 * destroy_super - frees a superblock
114 * @s: superblock to free
116 * Frees a superblock.
118 static inline void destroy_super(struct super_block
*s
)
126 /* Superblock refcounting */
129 * Drop a superblock's refcount. Returns non-zero if the superblock was
130 * destroyed. The caller must hold sb_lock.
132 static int __put_super(struct super_block
*sb
)
136 if (!--sb
->s_count
) {
144 * Drop a superblock's refcount.
145 * Returns non-zero if the superblock is about to be destroyed and
146 * at least is already removed from super_blocks list, so if we are
147 * making a loop through super blocks then we need to restart.
148 * The caller must hold sb_lock.
150 int __put_super_and_need_restart(struct super_block
*sb
)
152 /* check for race with generic_shutdown_super() */
153 if (list_empty(&sb
->s_list
)) {
154 /* super block is removed, need to restart... */
158 /* can't be the last, since s_list is still in use */
160 BUG_ON(sb
->s_count
== 0);
165 * put_super - drop a temporary reference to superblock
166 * @sb: superblock in question
168 * Drops a temporary reference, frees superblock if there's no
171 void put_super(struct super_block
*sb
)
175 spin_unlock(&sb_lock
);
180 * deactivate_super - drop an active reference to superblock
181 * @s: superblock to deactivate
183 * Drops an active reference to superblock, acquiring a temprory one if
184 * there is no active references left. In that case we lock superblock,
185 * tell fs driver to shut it down and drop the temporary reference we
188 void deactivate_super(struct super_block
*s
)
190 struct file_system_type
*fs
= s
->s_type
;
191 if (atomic_dec_and_lock(&s
->s_active
, &sb_lock
)) {
192 s
->s_count
-= S_BIAS
-1;
193 spin_unlock(&sb_lock
);
195 down_write(&s
->s_umount
);
202 EXPORT_SYMBOL(deactivate_super
);
205 * deactivate_locked_super - drop an active reference to superblock
206 * @s: superblock to deactivate
208 * Equivalent of up_write(&s->s_umount); deactivate_super(s);, except that
209 * it does not unlock it until it's all over. As the result, it's safe to
210 * use to dispose of new superblock on ->get_sb() failure exits - nobody
211 * will see the sucker until it's all over. Equivalent using up_write +
212 * deactivate_super is safe for that purpose only if superblock is either
213 * safe to use or has NULL ->s_root when we unlock.
215 void deactivate_locked_super(struct super_block
*s
)
217 struct file_system_type
*fs
= s
->s_type
;
218 if (atomic_dec_and_lock(&s
->s_active
, &sb_lock
)) {
219 s
->s_count
-= S_BIAS
-1;
220 spin_unlock(&sb_lock
);
226 up_write(&s
->s_umount
);
230 EXPORT_SYMBOL(deactivate_locked_super
);
233 * grab_super - acquire an active reference
234 * @s: reference we are trying to make active
236 * Tries to acquire an active reference. grab_super() is used when we
237 * had just found a superblock in super_blocks or fs_type->fs_supers
238 * and want to turn it into a full-blown active reference. grab_super()
239 * is called with sb_lock held and drops it. Returns 1 in case of
240 * success, 0 if we had failed (superblock contents was already dead or
241 * dying when grab_super() had been called).
243 static int grab_super(struct super_block
*s
) __releases(sb_lock
)
246 spin_unlock(&sb_lock
);
247 down_write(&s
->s_umount
);
250 if (s
->s_count
> S_BIAS
) {
251 atomic_inc(&s
->s_active
);
253 spin_unlock(&sb_lock
);
256 spin_unlock(&sb_lock
);
258 up_write(&s
->s_umount
);
265 * Superblock locking. We really ought to get rid of these two.
267 void lock_super(struct super_block
* sb
)
270 mutex_lock(&sb
->s_lock
);
273 void unlock_super(struct super_block
* sb
)
276 mutex_unlock(&sb
->s_lock
);
279 EXPORT_SYMBOL(lock_super
);
280 EXPORT_SYMBOL(unlock_super
);
283 * generic_shutdown_super - common helper for ->kill_sb()
284 * @sb: superblock to kill
286 * generic_shutdown_super() does all fs-independent work on superblock
287 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
288 * that need destruction out of superblock, call generic_shutdown_super()
289 * and release aforementioned objects. Note: dentries and inodes _are_
290 * taken care of and do not need specific handling.
292 * Upon calling this function, the filesystem may no longer alter or
293 * rearrange the set of dentries belonging to this super_block, nor may it
294 * change the attachments of dentries to inodes.
296 void generic_shutdown_super(struct super_block
*sb
)
298 const struct super_operations
*sop
= sb
->s_op
;
302 shrink_dcache_for_umount(sb
);
305 sb
->s_flags
&= ~MS_ACTIVE
;
307 /* bad name - it should be evict_inodes() */
308 invalidate_inodes(sb
);
313 /* Forget any remaining inodes */
314 if (invalidate_inodes(sb
)) {
315 printk("VFS: Busy inodes after unmount of %s. "
316 "Self-destruct in 5 seconds. Have a nice day...\n",
322 /* should be initialized for __put_super_and_need_restart() */
323 list_del_init(&sb
->s_list
);
324 list_del(&sb
->s_instances
);
325 spin_unlock(&sb_lock
);
326 up_write(&sb
->s_umount
);
329 EXPORT_SYMBOL(generic_shutdown_super
);
332 * sget - find or create a superblock
333 * @type: filesystem type superblock should belong to
334 * @test: comparison callback
335 * @set: setup callback
336 * @data: argument to each of them
338 struct super_block
*sget(struct file_system_type
*type
,
339 int (*test
)(struct super_block
*,void *),
340 int (*set
)(struct super_block
*,void *),
343 struct super_block
*s
= NULL
;
344 struct super_block
*old
;
350 list_for_each_entry(old
, &type
->fs_supers
, s_instances
) {
351 if (!test(old
, data
))
353 if (!grab_super(old
))
356 up_write(&s
->s_umount
);
363 spin_unlock(&sb_lock
);
364 s
= alloc_super(type
);
366 return ERR_PTR(-ENOMEM
);
372 spin_unlock(&sb_lock
);
373 up_write(&s
->s_umount
);
378 strlcpy(s
->s_id
, type
->name
, sizeof(s
->s_id
));
379 list_add_tail(&s
->s_list
, &super_blocks
);
380 list_add(&s
->s_instances
, &type
->fs_supers
);
381 spin_unlock(&sb_lock
);
382 get_filesystem(type
);
388 void drop_super(struct super_block
*sb
)
390 up_read(&sb
->s_umount
);
394 EXPORT_SYMBOL(drop_super
);
397 * sync_supers - helper for periodic superblock writeback
399 * Call the write_super method if present on all dirty superblocks in
400 * the system. This is for the periodic writeback used by most older
401 * filesystems. For data integrity superblock writeback use
402 * sync_filesystems() instead.
404 * Note: check the dirty flag before waiting, so we don't
405 * hold up the sync while mounting a device. (The newly
406 * mounted device won't need syncing.)
408 void sync_supers(void)
410 struct super_block
*sb
;
414 list_for_each_entry(sb
, &super_blocks
, s_list
) {
415 if (sb
->s_op
->write_super
&& sb
->s_dirt
) {
417 spin_unlock(&sb_lock
);
419 down_read(&sb
->s_umount
);
420 if (sb
->s_root
&& sb
->s_dirt
)
421 sb
->s_op
->write_super(sb
);
422 up_read(&sb
->s_umount
);
425 if (__put_super_and_need_restart(sb
))
429 spin_unlock(&sb_lock
);
433 * get_super - get the superblock of a device
434 * @bdev: device to get the superblock for
436 * Scans the superblock list and finds the superblock of the file system
437 * mounted on the device given. %NULL is returned if no match is found.
440 struct super_block
* get_super(struct block_device
*bdev
)
442 struct super_block
*sb
;
449 list_for_each_entry(sb
, &super_blocks
, s_list
) {
450 if (sb
->s_bdev
== bdev
) {
452 spin_unlock(&sb_lock
);
453 down_read(&sb
->s_umount
);
456 up_read(&sb
->s_umount
);
457 /* restart only when sb is no longer on the list */
459 if (__put_super_and_need_restart(sb
))
463 spin_unlock(&sb_lock
);
467 EXPORT_SYMBOL(get_super
);
469 struct super_block
* user_get_super(dev_t dev
)
471 struct super_block
*sb
;
475 list_for_each_entry(sb
, &super_blocks
, s_list
) {
476 if (sb
->s_dev
== dev
) {
478 spin_unlock(&sb_lock
);
479 down_read(&sb
->s_umount
);
482 up_read(&sb
->s_umount
);
483 /* restart only when sb is no longer on the list */
485 if (__put_super_and_need_restart(sb
))
489 spin_unlock(&sb_lock
);
493 SYSCALL_DEFINE2(ustat
, unsigned, dev
, struct ustat __user
*, ubuf
)
495 struct super_block
*s
;
500 s
= user_get_super(new_decode_dev(dev
));
503 err
= vfs_statfs(s
->s_root
, &sbuf
);
508 memset(&tmp
,0,sizeof(struct ustat
));
509 tmp
.f_tfree
= sbuf
.f_bfree
;
510 tmp
.f_tinode
= sbuf
.f_ffree
;
512 err
= copy_to_user(ubuf
,&tmp
,sizeof(struct ustat
)) ? -EFAULT
: 0;
518 * do_remount_sb - asks filesystem to change mount options.
519 * @sb: superblock in question
520 * @flags: numeric part of options
521 * @data: the rest of options
522 * @force: whether or not to force the change
524 * Alters the mount options of a mounted file system.
526 int do_remount_sb(struct super_block
*sb
, int flags
, void *data
, int force
)
532 if (!(flags
& MS_RDONLY
) && bdev_read_only(sb
->s_bdev
))
535 if (flags
& MS_RDONLY
)
537 shrink_dcache_sb(sb
);
540 /* If we are remounting RDONLY and current sb is read/write,
541 make sure there are no rw files opened */
542 if ((flags
& MS_RDONLY
) && !(sb
->s_flags
& MS_RDONLY
)) {
545 else if (!fs_may_remount_ro(sb
))
547 retval
= vfs_dq_off(sb
, 1);
548 if (retval
< 0 && retval
!= -ENOSYS
)
551 remount_rw
= !(flags
& MS_RDONLY
) && (sb
->s_flags
& MS_RDONLY
);
553 if (sb
->s_op
->remount_fs
) {
554 retval
= sb
->s_op
->remount_fs(sb
, &flags
, data
);
558 sb
->s_flags
= (sb
->s_flags
& ~MS_RMT_MASK
) | (flags
& MS_RMT_MASK
);
560 vfs_dq_quota_on_remount(sb
);
564 static void do_emergency_remount(struct work_struct
*work
)
566 struct super_block
*sb
;
569 list_for_each_entry(sb
, &super_blocks
, s_list
) {
571 spin_unlock(&sb_lock
);
572 down_write(&sb
->s_umount
);
573 if (sb
->s_root
&& sb
->s_bdev
&& !(sb
->s_flags
& MS_RDONLY
)) {
575 * ->remount_fs needs lock_kernel().
577 * What lock protects sb->s_flags??
579 do_remount_sb(sb
, MS_RDONLY
, NULL
, 1);
581 up_write(&sb
->s_umount
);
585 spin_unlock(&sb_lock
);
587 printk("Emergency Remount complete\n");
590 void emergency_remount(void)
592 struct work_struct
*work
;
594 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
596 INIT_WORK(work
, do_emergency_remount
);
602 * Unnamed block devices are dummy devices used by virtual
603 * filesystems which don't use real block-devices. -- jrs
606 static DEFINE_IDA(unnamed_dev_ida
);
607 static DEFINE_SPINLOCK(unnamed_dev_lock
);/* protects the above */
608 static int unnamed_dev_start
= 0; /* don't bother trying below it */
610 int set_anon_super(struct super_block
*s
, void *data
)
616 if (ida_pre_get(&unnamed_dev_ida
, GFP_ATOMIC
) == 0)
618 spin_lock(&unnamed_dev_lock
);
619 error
= ida_get_new_above(&unnamed_dev_ida
, unnamed_dev_start
, &dev
);
621 unnamed_dev_start
= dev
+ 1;
622 spin_unlock(&unnamed_dev_lock
);
623 if (error
== -EAGAIN
)
624 /* We raced and lost with another CPU. */
629 if ((dev
& MAX_ID_MASK
) == (1 << MINORBITS
)) {
630 spin_lock(&unnamed_dev_lock
);
631 ida_remove(&unnamed_dev_ida
, dev
);
632 if (unnamed_dev_start
> dev
)
633 unnamed_dev_start
= dev
;
634 spin_unlock(&unnamed_dev_lock
);
637 s
->s_dev
= MKDEV(0, dev
& MINORMASK
);
641 EXPORT_SYMBOL(set_anon_super
);
643 void kill_anon_super(struct super_block
*sb
)
645 int slot
= MINOR(sb
->s_dev
);
647 generic_shutdown_super(sb
);
648 spin_lock(&unnamed_dev_lock
);
649 ida_remove(&unnamed_dev_ida
, slot
);
650 if (slot
< unnamed_dev_start
)
651 unnamed_dev_start
= slot
;
652 spin_unlock(&unnamed_dev_lock
);
655 EXPORT_SYMBOL(kill_anon_super
);
657 void kill_litter_super(struct super_block
*sb
)
660 d_genocide(sb
->s_root
);
664 EXPORT_SYMBOL(kill_litter_super
);
666 static int ns_test_super(struct super_block
*sb
, void *data
)
668 return sb
->s_fs_info
== data
;
671 static int ns_set_super(struct super_block
*sb
, void *data
)
673 sb
->s_fs_info
= data
;
674 return set_anon_super(sb
, NULL
);
677 int get_sb_ns(struct file_system_type
*fs_type
, int flags
, void *data
,
678 int (*fill_super
)(struct super_block
*, void *, int),
679 struct vfsmount
*mnt
)
681 struct super_block
*sb
;
683 sb
= sget(fs_type
, ns_test_super
, ns_set_super
, data
);
690 err
= fill_super(sb
, data
, flags
& MS_SILENT
? 1 : 0);
692 deactivate_locked_super(sb
);
696 sb
->s_flags
|= MS_ACTIVE
;
699 simple_set_mnt(mnt
, sb
);
703 EXPORT_SYMBOL(get_sb_ns
);
706 static int set_bdev_super(struct super_block
*s
, void *data
)
709 s
->s_dev
= s
->s_bdev
->bd_dev
;
712 * We set the bdi here to the queue backing, file systems can
713 * overwrite this in ->fill_super()
715 s
->s_bdi
= &bdev_get_queue(s
->s_bdev
)->backing_dev_info
;
719 static int test_bdev_super(struct super_block
*s
, void *data
)
721 return (void *)s
->s_bdev
== data
;
724 int get_sb_bdev(struct file_system_type
*fs_type
,
725 int flags
, const char *dev_name
, void *data
,
726 int (*fill_super
)(struct super_block
*, void *, int),
727 struct vfsmount
*mnt
)
729 struct block_device
*bdev
;
730 struct super_block
*s
;
731 fmode_t mode
= FMODE_READ
;
734 if (!(flags
& MS_RDONLY
))
737 bdev
= open_bdev_exclusive(dev_name
, mode
, fs_type
);
739 return PTR_ERR(bdev
);
742 * once the super is inserted into the list by sget, s_umount
743 * will protect the lockfs code from trying to start a snapshot
744 * while we are mounting
746 down(&bdev
->bd_mount_sem
);
747 s
= sget(fs_type
, test_bdev_super
, set_bdev_super
, bdev
);
748 up(&bdev
->bd_mount_sem
);
753 if ((flags
^ s
->s_flags
) & MS_RDONLY
) {
754 deactivate_locked_super(s
);
759 close_bdev_exclusive(bdev
, mode
);
761 char b
[BDEVNAME_SIZE
];
765 strlcpy(s
->s_id
, bdevname(bdev
, b
), sizeof(s
->s_id
));
766 sb_set_blocksize(s
, block_size(bdev
));
767 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
769 deactivate_locked_super(s
);
773 s
->s_flags
|= MS_ACTIVE
;
777 simple_set_mnt(mnt
, s
);
783 close_bdev_exclusive(bdev
, mode
);
788 EXPORT_SYMBOL(get_sb_bdev
);
790 void kill_block_super(struct super_block
*sb
)
792 struct block_device
*bdev
= sb
->s_bdev
;
793 fmode_t mode
= sb
->s_mode
;
795 bdev
->bd_super
= NULL
;
796 generic_shutdown_super(sb
);
798 close_bdev_exclusive(bdev
, mode
);
801 EXPORT_SYMBOL(kill_block_super
);
804 int get_sb_nodev(struct file_system_type
*fs_type
,
805 int flags
, void *data
,
806 int (*fill_super
)(struct super_block
*, void *, int),
807 struct vfsmount
*mnt
)
810 struct super_block
*s
= sget(fs_type
, NULL
, set_anon_super
, NULL
);
817 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
819 deactivate_locked_super(s
);
822 s
->s_flags
|= MS_ACTIVE
;
823 simple_set_mnt(mnt
, s
);
827 EXPORT_SYMBOL(get_sb_nodev
);
829 static int compare_single(struct super_block
*s
, void *p
)
834 int get_sb_single(struct file_system_type
*fs_type
,
835 int flags
, void *data
,
836 int (*fill_super
)(struct super_block
*, void *, int),
837 struct vfsmount
*mnt
)
839 struct super_block
*s
;
842 s
= sget(fs_type
, compare_single
, set_anon_super
, NULL
);
847 error
= fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
849 deactivate_locked_super(s
);
852 s
->s_flags
|= MS_ACTIVE
;
854 do_remount_sb(s
, flags
, data
, 0);
855 simple_set_mnt(mnt
, s
);
859 EXPORT_SYMBOL(get_sb_single
);
862 vfs_kern_mount(struct file_system_type
*type
, int flags
, const char *name
, void *data
)
864 struct vfsmount
*mnt
;
865 char *secdata
= NULL
;
869 return ERR_PTR(-ENODEV
);
872 mnt
= alloc_vfsmnt(name
);
876 if (data
&& !(type
->fs_flags
& FS_BINARY_MOUNTDATA
)) {
877 secdata
= alloc_secdata();
881 error
= security_sb_copy_data(data
, secdata
);
883 goto out_free_secdata
;
886 error
= type
->get_sb(type
, flags
, name
, data
, mnt
);
888 goto out_free_secdata
;
889 BUG_ON(!mnt
->mnt_sb
);
891 error
= security_sb_kern_mount(mnt
->mnt_sb
, flags
, secdata
);
895 mnt
->mnt_mountpoint
= mnt
->mnt_root
;
896 mnt
->mnt_parent
= mnt
;
897 up_write(&mnt
->mnt_sb
->s_umount
);
898 free_secdata(secdata
);
902 deactivate_locked_super(mnt
->mnt_sb
);
904 free_secdata(secdata
);
908 return ERR_PTR(error
);
911 EXPORT_SYMBOL_GPL(vfs_kern_mount
);
913 static struct vfsmount
*fs_set_subtype(struct vfsmount
*mnt
, const char *fstype
)
916 const char *subtype
= strchr(fstype
, '.');
925 mnt
->mnt_sb
->s_subtype
= kstrdup(subtype
, GFP_KERNEL
);
927 if (!mnt
->mnt_sb
->s_subtype
)
937 do_kern_mount(const char *fstype
, int flags
, const char *name
, void *data
)
939 struct file_system_type
*type
= get_fs_type(fstype
);
940 struct vfsmount
*mnt
;
942 return ERR_PTR(-ENODEV
);
943 mnt
= vfs_kern_mount(type
, flags
, name
, data
);
944 if (!IS_ERR(mnt
) && (type
->fs_flags
& FS_HAS_SUBTYPE
) &&
945 !mnt
->mnt_sb
->s_subtype
)
946 mnt
= fs_set_subtype(mnt
, fstype
);
947 put_filesystem(type
);
950 EXPORT_SYMBOL_GPL(do_kern_mount
);
952 struct vfsmount
*kern_mount_data(struct file_system_type
*type
, void *data
)
954 return vfs_kern_mount(type
, MS_KERNMOUNT
, type
->name
, data
);
957 EXPORT_SYMBOL_GPL(kern_mount_data
);