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 * put_super - drop a temporary reference to superblock
121 * @s: superblock in question
123 * Drops a temporary reference, frees superblock if there's no
126 static void put_super(struct super_block *sb)
130 spin_unlock(&sb_lock);
135 * deactivate_super - drop an active reference to superblock
136 * @s: superblock to deactivate
138 * Drops an active reference to superblock, acquiring a temprory one if
139 * there is no active references left. In that case we lock superblock,
140 * tell fs driver to shut it down and drop the temporary reference we
143 void deactivate_super(struct super_block *s)
145 struct file_system_type *fs = s->s_type;
146 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
147 s->s_count -= S_BIAS-1;
148 spin_unlock(&sb_lock);
149 down_write(&s->s_umount);
156 EXPORT_SYMBOL(deactivate_super);
159 * grab_super - acquire an active reference
160 * @s: reference we are trying to make active
162 * Tries to acquire an active reference. grab_super() is used when we
163 * had just found a superblock in super_blocks or fs_type->fs_supers
164 * and want to turn it into a full-blown active reference. grab_super()
165 * is called with sb_lock held and drops it. Returns 1 in case of
166 * success, 0 if we had failed (superblock contents was already dead or
167 * dying when grab_super() had been called).
169 static int grab_super(struct super_block *s)
172 spin_unlock(&sb_lock);
173 down_write(&s->s_umount);
176 if (s->s_count > S_BIAS) {
177 atomic_inc(&s->s_active);
179 spin_unlock(&sb_lock);
182 spin_unlock(&sb_lock);
184 up_write(&s->s_umount);
191 * generic_shutdown_super - common helper for ->kill_sb()
192 * @sb: superblock to kill
194 * generic_shutdown_super() does all fs-independent work on superblock
195 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
196 * that need destruction out of superblock, call generic_shutdown_super()
197 * and release aforementioned objects. Note: dentries and inodes _are_
198 * taken care of and do not need specific handling.
200 void generic_shutdown_super(struct super_block *sb)
202 struct dentry *root = sb->s_root;
203 struct super_operations *sop = sb->s_op;
207 shrink_dcache_parent(root);
208 shrink_dcache_anon(&sb->s_anon);
213 sb->s_flags &= ~MS_ACTIVE;
214 /* bad name - it should be evict_inodes() */
215 invalidate_inodes(sb);
217 if (sop->write_super && sb->s_dirt)
218 sop->write_super(sb);
222 /* Forget any remaining inodes */
223 if (invalidate_inodes(sb)) {
224 printk("VFS: Busy inodes after unmount. "
225 "Self-destruct in 5 seconds. Have a nice day...\n");
232 list_del(&sb->s_list);
233 list_del(&sb->s_instances);
234 spin_unlock(&sb_lock);
235 up_write(&sb->s_umount);
238 EXPORT_SYMBOL(generic_shutdown_super);
241 * sget - find or create a superblock
242 * @type: filesystem type superblock should belong to
243 * @test: comparison callback
244 * @set: setup callback
245 * @data: argument to each of them
247 struct super_block *sget(struct file_system_type *type,
248 int (*test)(struct super_block *,void *),
249 int (*set)(struct super_block *,void *),
252 struct super_block *s = NULL;
258 if (test) list_for_each(p, &type->fs_supers) {
259 struct super_block *old;
260 old = list_entry(p, struct super_block, s_instances);
261 if (!test(old, data))
263 if (!grab_super(old))
270 spin_unlock(&sb_lock);
273 return ERR_PTR(-ENOMEM);
279 spin_unlock(&sb_lock);
284 strlcpy(s->s_id, type->name, sizeof(s->s_id));
285 list_add(&s->s_list, super_blocks.prev);
286 list_add(&s->s_instances, &type->fs_supers);
287 spin_unlock(&sb_lock);
288 get_filesystem(type);
294 void drop_super(struct super_block *sb)
296 up_read(&sb->s_umount);
300 EXPORT_SYMBOL(drop_super);
302 static inline void write_super(struct super_block *sb)
305 if (sb->s_root && sb->s_dirt)
306 if (sb->s_op->write_super)
307 sb->s_op->write_super(sb);
312 * Note: check the dirty flag before waiting, so we don't
313 * hold up the sync while mounting a device. (The newly
314 * mounted device won't need syncing.)
316 void sync_supers(void)
318 struct super_block * sb;
321 sb = sb_entry(super_blocks.next);
322 while (sb != sb_entry(&super_blocks))
325 spin_unlock(&sb_lock);
326 down_read(&sb->s_umount);
331 sb = sb_entry(sb->s_list.next);
332 spin_unlock(&sb_lock);
336 * Call the ->sync_fs super_op against all filesytems which are r/w and
337 * which implement it.
339 * This operation is careful to avoid the livelock which could easily happen
340 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
341 * is used only here. We set it against all filesystems and then clear it as
342 * we sync them. So redirtied filesystems are skipped.
344 * But if process A is currently running sync_filesytems and then process B
345 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
346 * flags again, which will cause process A to resync everything. Fix that with
349 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
351 void sync_filesystems(int wait)
353 struct super_block *sb;
354 static DECLARE_MUTEX(mutex);
356 down(&mutex); /* Could be down_interruptible */
358 for (sb = sb_entry(super_blocks.next); sb != sb_entry(&super_blocks);
359 sb = sb_entry(sb->s_list.next)) {
360 if (!sb->s_op->sync_fs)
362 if (sb->s_flags & MS_RDONLY)
364 sb->s_need_sync_fs = 1;
366 spin_unlock(&sb_lock);
370 for (sb = sb_entry(super_blocks.next); sb != sb_entry(&super_blocks);
371 sb = sb_entry(sb->s_list.next)) {
372 if (!sb->s_need_sync_fs)
374 sb->s_need_sync_fs = 0;
375 if (sb->s_flags & MS_RDONLY)
376 continue; /* hm. Was remounted r/o meanwhile */
378 spin_unlock(&sb_lock);
379 down_read(&sb->s_umount);
380 if (sb->s_root && (wait || sb->s_dirt))
381 sb->s_op->sync_fs(sb, wait);
385 spin_unlock(&sb_lock);
390 * get_super - get the superblock of a device
391 * @bdev: device to get the superblock for
393 * Scans the superblock list and finds the superblock of the file system
394 * mounted on the device given. %NULL is returned if no match is found.
397 struct super_block * get_super(struct block_device *bdev)
404 list_for_each(p, &super_blocks) {
405 struct super_block *s = sb_entry(p);
406 if (s->s_bdev == bdev) {
408 spin_unlock(&sb_lock);
409 down_read(&s->s_umount);
416 spin_unlock(&sb_lock);
420 EXPORT_SYMBOL(get_super);
422 struct super_block * user_get_super(dev_t dev)
428 list_for_each(p, &super_blocks) {
429 struct super_block *s = sb_entry(p);
430 if (s->s_dev == dev) {
432 spin_unlock(&sb_lock);
433 down_read(&s->s_umount);
440 spin_unlock(&sb_lock);
444 EXPORT_SYMBOL(user_get_super);
446 asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
448 struct super_block *s;
453 s = user_get_super(new_decode_dev(dev));
456 err = vfs_statfs(s, &sbuf);
461 memset(&tmp,0,sizeof(struct ustat));
462 tmp.f_tfree = sbuf.f_bfree;
463 tmp.f_tinode = sbuf.f_ffree;
465 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
472 * @sb: superblock in question
474 * All files are marked read/only. We don't care about pending
475 * delete files so this should be used in 'force' mode only
478 static void mark_files_ro(struct super_block *sb)
483 list_for_each_entry(f, &sb->s_files, f_list) {
484 if (S_ISREG(f->f_dentry->d_inode->i_mode) && file_count(f))
485 f->f_mode &= ~FMODE_WRITE;
491 * do_remount_sb - asks filesystem to change mount options.
492 * @sb: superblock in question
493 * @flags: numeric part of options
494 * @data: the rest of options
495 * @force: whether or not to force the change
497 * Alters the mount options of a mounted file system.
499 int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
503 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
505 if (flags & MS_RDONLY)
507 shrink_dcache_sb(sb);
510 /* If we are remounting RDONLY and current sb is read/write,
511 make sure there are no rw files opened */
512 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
515 else if (!fs_may_remount_ro(sb))
519 if (sb->s_op->remount_fs) {
521 retval = sb->s_op->remount_fs(sb, &flags, data);
526 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
530 static void do_emergency_remount(unsigned long foo)
532 struct super_block *sb;
535 list_for_each_entry(sb, &super_blocks, s_list) {
537 spin_unlock(&sb_lock);
538 down_read(&sb->s_umount);
539 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
541 * ->remount_fs needs lock_kernel().
543 * What lock protects sb->s_flags??
546 do_remount_sb(sb, MS_RDONLY, NULL, 1);
552 spin_unlock(&sb_lock);
553 printk("Emergency Remount complete\n");
556 void emergency_remount(void)
558 pdflush_operation(do_emergency_remount, 0);
562 * Unnamed block devices are dummy devices used by virtual
563 * filesystems which don't use real block-devices. -- jrs
566 static struct idr unnamed_dev_idr;
567 static spinlock_t unnamed_dev_lock = SPIN_LOCK_UNLOCKED;/* protects the above */
569 int set_anon_super(struct super_block *s, void *data)
575 if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0)
577 spin_lock(&unnamed_dev_lock);
578 error = idr_get_new(&unnamed_dev_idr, NULL, &dev);
579 spin_unlock(&unnamed_dev_lock);
580 if (error == -EAGAIN)
581 /* We raced and lost with another CPU. */
586 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
587 spin_lock(&unnamed_dev_lock);
588 idr_remove(&unnamed_dev_idr, dev);
589 spin_unlock(&unnamed_dev_lock);
592 s->s_dev = MKDEV(0, dev & MINORMASK);
596 EXPORT_SYMBOL(set_anon_super);
598 void kill_anon_super(struct super_block *sb)
600 int slot = MINOR(sb->s_dev);
602 generic_shutdown_super(sb);
603 spin_lock(&unnamed_dev_lock);
604 idr_remove(&unnamed_dev_idr, slot);
605 spin_unlock(&unnamed_dev_lock);
608 EXPORT_SYMBOL(kill_anon_super);
610 void __init unnamed_dev_init(void)
612 idr_init(&unnamed_dev_idr);
615 void kill_litter_super(struct super_block *sb)
618 d_genocide(sb->s_root);
622 EXPORT_SYMBOL(kill_litter_super);
624 static int set_bdev_super(struct super_block *s, void *data)
627 s->s_dev = s->s_bdev->bd_dev;
631 static int test_bdev_super(struct super_block *s, void *data)
633 return (void *)s->s_bdev == data;
636 struct super_block *get_sb_bdev(struct file_system_type *fs_type,
637 int flags, const char *dev_name, void *data,
638 int (*fill_super)(struct super_block *, void *, int))
640 struct block_device *bdev;
641 struct super_block *s;
644 bdev = open_bdev_excl(dev_name, flags, fs_type);
646 return (struct super_block *)bdev;
649 * once the super is inserted into the list by sget, s_umount
650 * will protect the lockfs code from trying to start a snapshot
651 * while we are mounting
653 down(&bdev->bd_mount_sem);
654 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
655 up(&bdev->bd_mount_sem);
660 if ((flags ^ s->s_flags) & MS_RDONLY) {
661 up_write(&s->s_umount);
667 char b[BDEVNAME_SIZE];
670 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
671 s->s_old_blocksize = block_size(bdev);
672 sb_set_blocksize(s, s->s_old_blocksize);
673 error = fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
675 up_write(&s->s_umount);
679 s->s_flags |= MS_ACTIVE;
685 close_bdev_excl(bdev);
689 EXPORT_SYMBOL(get_sb_bdev);
691 void kill_block_super(struct super_block *sb)
693 struct block_device *bdev = sb->s_bdev;
694 generic_shutdown_super(sb);
695 set_blocksize(bdev, sb->s_old_blocksize);
696 close_bdev_excl(bdev);
699 EXPORT_SYMBOL(kill_block_super);
701 struct super_block *get_sb_nodev(struct file_system_type *fs_type,
702 int flags, void *data,
703 int (*fill_super)(struct super_block *, void *, int))
706 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
713 error = fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
715 up_write(&s->s_umount);
717 return ERR_PTR(error);
719 s->s_flags |= MS_ACTIVE;
723 EXPORT_SYMBOL(get_sb_nodev);
725 static int compare_single(struct super_block *s, void *p)
730 struct super_block *get_sb_single(struct file_system_type *fs_type,
731 int flags, void *data,
732 int (*fill_super)(struct super_block *, void *, int))
734 struct super_block *s;
737 s = sget(fs_type, compare_single, set_anon_super, NULL);
742 error = fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
744 up_write(&s->s_umount);
746 return ERR_PTR(error);
748 s->s_flags |= MS_ACTIVE;
750 do_remount_sb(s, flags, data, 0);
754 EXPORT_SYMBOL(get_sb_single);
757 do_kern_mount(const char *fstype, int flags, const char *name, void *data)
759 struct file_system_type *type = get_fs_type(fstype);
760 struct super_block *sb = ERR_PTR(-ENOMEM);
761 struct vfsmount *mnt;
763 char *secdata = NULL;
766 return ERR_PTR(-ENODEV);
768 mnt = alloc_vfsmnt(name);
773 secdata = alloc_secdata();
775 sb = ERR_PTR(-ENOMEM);
779 error = security_sb_copy_data(type, data, secdata);
782 goto out_free_secdata;
786 sb = type->get_sb(type, flags, name, data);
788 goto out_free_secdata;
789 error = security_sb_kern_mount(sb, secdata);
793 mnt->mnt_root = dget(sb->s_root);
794 mnt->mnt_mountpoint = sb->s_root;
795 mnt->mnt_parent = mnt;
796 mnt->mnt_namespace = current->namespace;
797 up_write(&sb->s_umount);
798 put_filesystem(type);
801 up_write(&sb->s_umount);
802 deactivate_super(sb);
805 free_secdata(secdata);
809 put_filesystem(type);
810 return (struct vfsmount *)sb;
813 EXPORT_SYMBOL_GPL(do_kern_mount);
815 struct vfsmount *kern_mount(struct file_system_type *type)
817 return do_kern_mount(type->name, 0, type->name, NULL);
820 EXPORT_SYMBOL(kern_mount);