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/buffer_head.h> /* for fsync_super() */
32 #include <linux/mount.h>
33 #include <linux/security.h>
34 #include <linux/syscalls.h>
35 #include <linux/vfs.h>
36 #include <linux/writeback.h> /* for the emergency remount stuff */
37 #include <linux/idr.h>
38 #include <linux/kobject.h>
39 #include <linux/mutex.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(struct file_system_type *type)
58 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
59 static struct super_operations default_op;
62 if (security_sb_alloc(s)) {
67 INIT_LIST_HEAD(&s->s_dirty);
68 INIT_LIST_HEAD(&s->s_io);
69 INIT_LIST_HEAD(&s->s_files);
70 INIT_LIST_HEAD(&s->s_instances);
71 INIT_HLIST_HEAD(&s->s_anon);
72 INIT_LIST_HEAD(&s->s_inodes);
73 init_rwsem(&s->s_umount);
74 mutex_init(&s->s_lock);
75 lockdep_set_class(&s->s_umount, &type->s_umount_key);
77 * The locking rules for s_lock are up to the
78 * filesystem. For example ext3fs has different
79 * lock ordering than usbfs:
81 lockdep_set_class(&s->s_lock, &type->s_lock_key);
82 down_write(&s->s_umount);
84 atomic_set(&s->s_active, 1);
85 mutex_init(&s->s_vfs_rename_mutex);
86 mutex_init(&s->s_dquot.dqio_mutex);
87 mutex_init(&s->s_dquot.dqonoff_mutex);
88 init_rwsem(&s->s_dquot.dqptr_sem);
89 init_waitqueue_head(&s->s_wait_unfrozen);
90 s->s_maxbytes = MAX_NON_LFS;
91 s->dq_op = sb_dquot_ops;
92 s->s_qcop = sb_quotactl_ops;
93 s->s_op = &default_op;
94 s->s_time_gran = 1000000000;
101 * destroy_super - frees a superblock
102 * @s: superblock to free
104 * Frees a superblock.
106 static inline void destroy_super(struct super_block *s)
112 /* Superblock refcounting */
115 * Drop a superblock's refcount. Returns non-zero if the superblock was
116 * destroyed. The caller must hold sb_lock.
118 int __put_super(struct super_block *sb)
122 if (!--sb->s_count) {
130 * Drop a superblock's refcount.
131 * Returns non-zero if the superblock is about to be destroyed and
132 * at least is already removed from super_blocks list, so if we are
133 * making a loop through super blocks then we need to restart.
134 * The caller must hold sb_lock.
136 int __put_super_and_need_restart(struct super_block *sb)
138 /* check for race with generic_shutdown_super() */
139 if (list_empty(&sb->s_list)) {
140 /* super block is removed, need to restart... */
144 /* can't be the last, since s_list is still in use */
146 BUG_ON(sb->s_count == 0);
151 * put_super - drop a temporary reference to superblock
152 * @sb: superblock in question
154 * Drops a temporary reference, frees superblock if there's no
157 static void put_super(struct super_block *sb)
161 spin_unlock(&sb_lock);
166 * deactivate_super - drop an active reference to superblock
167 * @s: superblock to deactivate
169 * Drops an active reference to superblock, acquiring a temprory one if
170 * there is no active references left. In that case we lock superblock,
171 * tell fs driver to shut it down and drop the temporary reference we
174 void deactivate_super(struct super_block *s)
176 struct file_system_type *fs = s->s_type;
177 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
178 s->s_count -= S_BIAS-1;
179 spin_unlock(&sb_lock);
181 down_write(&s->s_umount);
188 EXPORT_SYMBOL(deactivate_super);
191 * grab_super - acquire an active reference
192 * @s: reference we are trying to make active
194 * Tries to acquire an active reference. grab_super() is used when we
195 * had just found a superblock in super_blocks or fs_type->fs_supers
196 * and want to turn it into a full-blown active reference. grab_super()
197 * is called with sb_lock held and drops it. Returns 1 in case of
198 * success, 0 if we had failed (superblock contents was already dead or
199 * dying when grab_super() had been called).
201 static int grab_super(struct super_block *s)
204 spin_unlock(&sb_lock);
205 down_write(&s->s_umount);
208 if (s->s_count > S_BIAS) {
209 atomic_inc(&s->s_active);
211 spin_unlock(&sb_lock);
214 spin_unlock(&sb_lock);
216 up_write(&s->s_umount);
223 * generic_shutdown_super - common helper for ->kill_sb()
224 * @sb: superblock to kill
226 * generic_shutdown_super() does all fs-independent work on superblock
227 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
228 * that need destruction out of superblock, call generic_shutdown_super()
229 * and release aforementioned objects. Note: dentries and inodes _are_
230 * taken care of and do not need specific handling.
232 void generic_shutdown_super(struct super_block *sb)
234 struct dentry *root = sb->s_root;
235 struct super_operations *sop = sb->s_op;
239 shrink_dcache_parent(root);
240 shrink_dcache_sb(sb);
244 sb->s_flags &= ~MS_ACTIVE;
245 /* bad name - it should be evict_inodes() */
246 invalidate_inodes(sb);
249 if (sop->write_super && sb->s_dirt)
250 sop->write_super(sb);
254 /* Forget any remaining inodes */
255 if (invalidate_inodes(sb)) {
256 printk("VFS: Busy inodes after unmount of %s. "
257 "Self-destruct in 5 seconds. Have a nice day...\n",
265 /* should be initialized for __put_super_and_need_restart() */
266 list_del_init(&sb->s_list);
267 list_del(&sb->s_instances);
268 spin_unlock(&sb_lock);
269 up_write(&sb->s_umount);
272 EXPORT_SYMBOL(generic_shutdown_super);
275 * sget - find or create a superblock
276 * @type: filesystem type superblock should belong to
277 * @test: comparison callback
278 * @set: setup callback
279 * @data: argument to each of them
281 struct super_block *sget(struct file_system_type *type,
282 int (*test)(struct super_block *,void *),
283 int (*set)(struct super_block *,void *),
286 struct super_block *s = NULL;
292 if (test) list_for_each(p, &type->fs_supers) {
293 struct super_block *old;
294 old = list_entry(p, struct super_block, s_instances);
295 if (!test(old, data))
297 if (!grab_super(old))
304 spin_unlock(&sb_lock);
305 s = alloc_super(type);
307 return ERR_PTR(-ENOMEM);
313 spin_unlock(&sb_lock);
318 strlcpy(s->s_id, type->name, sizeof(s->s_id));
319 list_add_tail(&s->s_list, &super_blocks);
320 list_add(&s->s_instances, &type->fs_supers);
321 spin_unlock(&sb_lock);
322 get_filesystem(type);
328 void drop_super(struct super_block *sb)
330 up_read(&sb->s_umount);
334 EXPORT_SYMBOL(drop_super);
336 static inline void write_super(struct super_block *sb)
339 if (sb->s_root && sb->s_dirt)
340 if (sb->s_op->write_super)
341 sb->s_op->write_super(sb);
346 * Note: check the dirty flag before waiting, so we don't
347 * hold up the sync while mounting a device. (The newly
348 * mounted device won't need syncing.)
350 void sync_supers(void)
352 struct super_block *sb;
356 list_for_each_entry(sb, &super_blocks, s_list) {
359 spin_unlock(&sb_lock);
360 down_read(&sb->s_umount);
362 up_read(&sb->s_umount);
364 if (__put_super_and_need_restart(sb))
368 spin_unlock(&sb_lock);
372 * Call the ->sync_fs super_op against all filesytems which are r/w and
373 * which implement it.
375 * This operation is careful to avoid the livelock which could easily happen
376 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
377 * is used only here. We set it against all filesystems and then clear it as
378 * we sync them. So redirtied filesystems are skipped.
380 * But if process A is currently running sync_filesytems and then process B
381 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
382 * flags again, which will cause process A to resync everything. Fix that with
385 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
387 void sync_filesystems(int wait)
389 struct super_block *sb;
390 static DEFINE_MUTEX(mutex);
392 mutex_lock(&mutex); /* Could be down_interruptible */
394 list_for_each_entry(sb, &super_blocks, s_list) {
395 if (!sb->s_op->sync_fs)
397 if (sb->s_flags & MS_RDONLY)
399 sb->s_need_sync_fs = 1;
403 list_for_each_entry(sb, &super_blocks, s_list) {
404 if (!sb->s_need_sync_fs)
406 sb->s_need_sync_fs = 0;
407 if (sb->s_flags & MS_RDONLY)
408 continue; /* hm. Was remounted r/o meanwhile */
410 spin_unlock(&sb_lock);
411 down_read(&sb->s_umount);
412 if (sb->s_root && (wait || sb->s_dirt))
413 sb->s_op->sync_fs(sb, wait);
414 up_read(&sb->s_umount);
415 /* restart only when sb is no longer on the list */
417 if (__put_super_and_need_restart(sb))
420 spin_unlock(&sb_lock);
421 mutex_unlock(&mutex);
425 * get_super - get the superblock of a device
426 * @bdev: device to get the superblock for
428 * Scans the superblock list and finds the superblock of the file system
429 * mounted on the device given. %NULL is returned if no match is found.
432 struct super_block * get_super(struct block_device *bdev)
434 struct super_block *sb;
441 list_for_each_entry(sb, &super_blocks, s_list) {
442 if (sb->s_bdev == bdev) {
444 spin_unlock(&sb_lock);
445 down_read(&sb->s_umount);
448 up_read(&sb->s_umount);
449 /* restart only when sb is no longer on the list */
451 if (__put_super_and_need_restart(sb))
455 spin_unlock(&sb_lock);
459 EXPORT_SYMBOL(get_super);
461 struct super_block * user_get_super(dev_t dev)
463 struct super_block *sb;
467 list_for_each_entry(sb, &super_blocks, s_list) {
468 if (sb->s_dev == dev) {
470 spin_unlock(&sb_lock);
471 down_read(&sb->s_umount);
474 up_read(&sb->s_umount);
475 /* restart only when sb is no longer on the list */
477 if (__put_super_and_need_restart(sb))
481 spin_unlock(&sb_lock);
485 asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
487 struct super_block *s;
492 s = user_get_super(new_decode_dev(dev));
495 err = vfs_statfs(s->s_root, &sbuf);
500 memset(&tmp,0,sizeof(struct ustat));
501 tmp.f_tfree = sbuf.f_bfree;
502 tmp.f_tinode = sbuf.f_ffree;
504 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
511 * @sb: superblock in question
513 * All files are marked read/only. We don't care about pending
514 * delete files so this should be used in 'force' mode only
517 static void mark_files_ro(struct super_block *sb)
522 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
523 if (S_ISREG(f->f_dentry->d_inode->i_mode) && file_count(f))
524 f->f_mode &= ~FMODE_WRITE;
530 * do_remount_sb - asks filesystem to change mount options.
531 * @sb: superblock in question
532 * @flags: numeric part of options
533 * @data: the rest of options
534 * @force: whether or not to force the change
536 * Alters the mount options of a mounted file system.
538 int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
542 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
544 if (flags & MS_RDONLY)
546 shrink_dcache_sb(sb);
549 /* If we are remounting RDONLY and current sb is read/write,
550 make sure there are no rw files opened */
551 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
554 else if (!fs_may_remount_ro(sb))
558 if (sb->s_op->remount_fs) {
560 retval = sb->s_op->remount_fs(sb, &flags, data);
565 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
569 static void do_emergency_remount(unsigned long foo)
571 struct super_block *sb;
574 list_for_each_entry(sb, &super_blocks, s_list) {
576 spin_unlock(&sb_lock);
577 down_read(&sb->s_umount);
578 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
580 * ->remount_fs needs lock_kernel().
582 * What lock protects sb->s_flags??
585 do_remount_sb(sb, MS_RDONLY, NULL, 1);
591 spin_unlock(&sb_lock);
592 printk("Emergency Remount complete\n");
595 void emergency_remount(void)
597 pdflush_operation(do_emergency_remount, 0);
601 * Unnamed block devices are dummy devices used by virtual
602 * filesystems which don't use real block-devices. -- jrs
605 static struct idr unnamed_dev_idr;
606 static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
608 int set_anon_super(struct super_block *s, void *data)
614 if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0)
616 spin_lock(&unnamed_dev_lock);
617 error = idr_get_new(&unnamed_dev_idr, NULL, &dev);
618 spin_unlock(&unnamed_dev_lock);
619 if (error == -EAGAIN)
620 /* We raced and lost with another CPU. */
625 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
626 spin_lock(&unnamed_dev_lock);
627 idr_remove(&unnamed_dev_idr, dev);
628 spin_unlock(&unnamed_dev_lock);
631 s->s_dev = MKDEV(0, dev & MINORMASK);
635 EXPORT_SYMBOL(set_anon_super);
637 void kill_anon_super(struct super_block *sb)
639 int slot = MINOR(sb->s_dev);
641 generic_shutdown_super(sb);
642 spin_lock(&unnamed_dev_lock);
643 idr_remove(&unnamed_dev_idr, slot);
644 spin_unlock(&unnamed_dev_lock);
647 EXPORT_SYMBOL(kill_anon_super);
649 void __init unnamed_dev_init(void)
651 idr_init(&unnamed_dev_idr);
654 void kill_litter_super(struct super_block *sb)
657 d_genocide(sb->s_root);
661 EXPORT_SYMBOL(kill_litter_super);
663 static int set_bdev_super(struct super_block *s, void *data)
666 s->s_dev = s->s_bdev->bd_dev;
670 static int test_bdev_super(struct super_block *s, void *data)
672 return (void *)s->s_bdev == data;
675 static void bdev_uevent(struct block_device *bdev, enum kobject_action action)
679 kobject_uevent(&bdev->bd_part->kobj, action);
681 kobject_uevent(&bdev->bd_disk->kobj, action);
685 int get_sb_bdev(struct file_system_type *fs_type,
686 int flags, const char *dev_name, void *data,
687 int (*fill_super)(struct super_block *, void *, int),
688 struct vfsmount *mnt)
690 struct block_device *bdev;
691 struct super_block *s;
694 bdev = open_bdev_excl(dev_name, flags, fs_type);
696 return PTR_ERR(bdev);
699 * once the super is inserted into the list by sget, s_umount
700 * will protect the lockfs code from trying to start a snapshot
701 * while we are mounting
703 mutex_lock(&bdev->bd_mount_mutex);
704 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
705 mutex_unlock(&bdev->bd_mount_mutex);
710 if ((flags ^ s->s_flags) & MS_RDONLY) {
711 up_write(&s->s_umount);
717 close_bdev_excl(bdev);
719 char b[BDEVNAME_SIZE];
722 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
723 sb_set_blocksize(s, block_size(bdev));
724 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
726 up_write(&s->s_umount);
731 s->s_flags |= MS_ACTIVE;
732 bdev_uevent(bdev, KOBJ_MOUNT);
735 return simple_set_mnt(mnt, s);
740 close_bdev_excl(bdev);
745 EXPORT_SYMBOL(get_sb_bdev);
747 void kill_block_super(struct super_block *sb)
749 struct block_device *bdev = sb->s_bdev;
751 bdev_uevent(bdev, KOBJ_UMOUNT);
752 generic_shutdown_super(sb);
754 close_bdev_excl(bdev);
757 EXPORT_SYMBOL(kill_block_super);
759 int get_sb_nodev(struct file_system_type *fs_type,
760 int flags, void *data,
761 int (*fill_super)(struct super_block *, void *, int),
762 struct vfsmount *mnt)
765 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
772 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
774 up_write(&s->s_umount);
778 s->s_flags |= MS_ACTIVE;
779 return simple_set_mnt(mnt, s);
782 EXPORT_SYMBOL(get_sb_nodev);
784 static int compare_single(struct super_block *s, void *p)
789 int get_sb_single(struct file_system_type *fs_type,
790 int flags, void *data,
791 int (*fill_super)(struct super_block *, void *, int),
792 struct vfsmount *mnt)
794 struct super_block *s;
797 s = sget(fs_type, compare_single, set_anon_super, NULL);
802 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
804 up_write(&s->s_umount);
808 s->s_flags |= MS_ACTIVE;
810 do_remount_sb(s, flags, data, 0);
811 return simple_set_mnt(mnt, s);
814 EXPORT_SYMBOL(get_sb_single);
817 vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
819 struct vfsmount *mnt;
820 char *secdata = NULL;
824 return ERR_PTR(-ENODEV);
827 mnt = alloc_vfsmnt(name);
832 secdata = alloc_secdata();
836 error = security_sb_copy_data(type, data, secdata);
838 goto out_free_secdata;
841 error = type->get_sb(type, flags, name, data, mnt);
843 goto out_free_secdata;
845 error = security_sb_kern_mount(mnt->mnt_sb, secdata);
849 mnt->mnt_mountpoint = mnt->mnt_root;
850 mnt->mnt_parent = mnt;
851 up_write(&mnt->mnt_sb->s_umount);
852 free_secdata(secdata);
856 up_write(&mnt->mnt_sb->s_umount);
857 deactivate_super(mnt->mnt_sb);
859 free_secdata(secdata);
863 return ERR_PTR(error);
866 EXPORT_SYMBOL_GPL(vfs_kern_mount);
869 do_kern_mount(const char *fstype, int flags, const char *name, void *data)
871 struct file_system_type *type = get_fs_type(fstype);
872 struct vfsmount *mnt;
874 return ERR_PTR(-ENODEV);
875 mnt = vfs_kern_mount(type, flags, name, data);
876 put_filesystem(type);
880 struct vfsmount *kern_mount(struct file_system_type *type)
882 return vfs_kern_mount(type, 0, type->name, NULL);
885 EXPORT_SYMBOL(kern_mount);