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 <linux/mutex.h>
41 #include <linux/devpts_fs.h>
42 #include <linux/proc_fs.h>
43 #include <asm/uaccess.h>
46 void get_filesystem(struct file_system_type *fs);
47 void put_filesystem(struct file_system_type *fs);
48 struct file_system_type *get_fs_type(const char *name);
50 LIST_HEAD(super_blocks);
51 DEFINE_SPINLOCK(sb_lock);
54 * alloc_super - create new superblock
56 * Allocates and initializes a new &struct super_block. alloc_super()
57 * returns a pointer new superblock or %NULL if allocation had failed.
59 static struct super_block *alloc_super(void)
61 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
62 static struct super_operations default_op;
65 if (security_sb_alloc(s)) {
70 INIT_LIST_HEAD(&s->s_dirty);
71 INIT_LIST_HEAD(&s->s_io);
72 INIT_LIST_HEAD(&s->s_files);
73 INIT_LIST_HEAD(&s->s_instances);
74 INIT_HLIST_HEAD(&s->s_anon);
75 INIT_LIST_HEAD(&s->s_inodes);
76 init_rwsem(&s->s_umount);
77 mutex_init(&s->s_lock);
78 down_write(&s->s_umount);
80 atomic_set(&s->s_active, 1);
81 mutex_init(&s->s_vfs_rename_mutex);
82 mutex_init(&s->s_dquot.dqio_mutex);
83 mutex_init(&s->s_dquot.dqonoff_mutex);
84 init_rwsem(&s->s_dquot.dqptr_sem);
85 init_waitqueue_head(&s->s_wait_unfrozen);
86 s->s_maxbytes = MAX_NON_LFS;
87 s->dq_op = sb_dquot_ops;
88 s->s_qcop = sb_quotactl_ops;
89 s->s_op = &default_op;
90 s->s_time_gran = 1000000000;
97 * destroy_super - frees a superblock
98 * @s: superblock to free
100 * Frees a superblock.
102 static inline void destroy_super(struct super_block *s)
108 /* Superblock refcounting */
111 * Drop a superblock's refcount. Returns non-zero if the superblock was
112 * destroyed. The caller must hold sb_lock.
114 int __put_super(struct super_block *sb)
118 if (!--sb->s_count) {
126 * Drop a superblock's refcount.
127 * Returns non-zero if the superblock is about to be destroyed and
128 * at least is already removed from super_blocks list, so if we are
129 * making a loop through super blocks then we need to restart.
130 * The caller must hold sb_lock.
132 int __put_super_and_need_restart(struct super_block *sb)
134 /* check for race with generic_shutdown_super() */
135 if (list_empty(&sb->s_list)) {
136 /* super block is removed, need to restart... */
140 /* can't be the last, since s_list is still in use */
142 BUG_ON(sb->s_count == 0);
147 * put_super - drop a temporary reference to superblock
148 * @sb: superblock in question
150 * Drops a temporary reference, frees superblock if there's no
153 static void put_super(struct super_block *sb)
157 spin_unlock(&sb_lock);
162 * deactivate_super - drop an active reference to superblock
163 * @s: superblock to deactivate
165 * Drops an active reference to superblock, acquiring a temprory one if
166 * there is no active references left. In that case we lock superblock,
167 * tell fs driver to shut it down and drop the temporary reference we
170 void deactivate_super(struct super_block *s)
172 struct file_system_type *fs = s->s_type;
173 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
174 s->s_count -= S_BIAS-1;
175 spin_unlock(&sb_lock);
177 down_write(&s->s_umount);
184 EXPORT_SYMBOL(deactivate_super);
187 * grab_super - acquire an active reference
188 * @s: reference we are trying to make active
190 * Tries to acquire an active reference. grab_super() is used when we
191 * had just found a superblock in super_blocks or fs_type->fs_supers
192 * and want to turn it into a full-blown active reference. grab_super()
193 * is called with sb_lock held and drops it. Returns 1 in case of
194 * success, 0 if we had failed (superblock contents was already dead or
195 * dying when grab_super() had been called).
197 static int grab_super(struct super_block *s)
200 spin_unlock(&sb_lock);
201 down_write(&s->s_umount);
204 if (s->s_count > S_BIAS) {
205 atomic_inc(&s->s_active);
207 spin_unlock(&sb_lock);
210 spin_unlock(&sb_lock);
212 up_write(&s->s_umount);
219 * generic_shutdown_super - common helper for ->kill_sb()
220 * @sb: superblock to kill
222 * generic_shutdown_super() does all fs-independent work on superblock
223 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
224 * that need destruction out of superblock, call generic_shutdown_super()
225 * and release aforementioned objects. Note: dentries and inodes _are_
226 * taken care of and do not need specific handling.
228 void generic_shutdown_super(struct super_block *sb)
230 struct dentry *root = sb->s_root;
231 struct super_operations *sop = sb->s_op;
235 shrink_dcache_parent(root);
236 shrink_dcache_anon(&sb->s_anon);
240 sb->s_flags &= ~MS_ACTIVE;
241 /* bad name - it should be evict_inodes() */
242 invalidate_inodes(sb);
245 if (sop->write_super && sb->s_dirt)
246 sop->write_super(sb);
250 /* Forget any remaining inodes */
251 if (invalidate_inodes(sb)) {
252 printk("VFS: Busy inodes after unmount of %s. "
253 "Self-destruct in 5 seconds. Have a nice day...\n",
261 /* should be initialized for __put_super_and_need_restart() */
262 list_del_init(&sb->s_list);
263 list_del(&sb->s_instances);
264 spin_unlock(&sb_lock);
265 up_write(&sb->s_umount);
268 EXPORT_SYMBOL(generic_shutdown_super);
271 * sget - find or create a superblock
272 * @type: filesystem type superblock should belong to
273 * @test: comparison callback
274 * @set: setup callback
275 * @data: argument to each of them
277 struct super_block *sget(struct file_system_type *type,
278 int (*test)(struct super_block *,void *),
279 int (*set)(struct super_block *,void *),
282 struct super_block *s = NULL;
288 if (test) list_for_each(p, &type->fs_supers) {
289 struct super_block *old;
290 old = list_entry(p, struct super_block, s_instances);
291 if (!test(old, data))
293 if (!grab_super(old))
300 spin_unlock(&sb_lock);
303 return ERR_PTR(-ENOMEM);
309 spin_unlock(&sb_lock);
314 strlcpy(s->s_id, type->name, sizeof(s->s_id));
315 list_add_tail(&s->s_list, &super_blocks);
316 list_add(&s->s_instances, &type->fs_supers);
317 spin_unlock(&sb_lock);
318 get_filesystem(type);
324 void drop_super(struct super_block *sb)
326 up_read(&sb->s_umount);
330 EXPORT_SYMBOL(drop_super);
332 static inline void write_super(struct super_block *sb)
335 if (sb->s_root && sb->s_dirt)
336 if (sb->s_op->write_super)
337 sb->s_op->write_super(sb);
342 * Note: check the dirty flag before waiting, so we don't
343 * hold up the sync while mounting a device. (The newly
344 * mounted device won't need syncing.)
346 void sync_supers(void)
348 struct super_block *sb;
352 list_for_each_entry(sb, &super_blocks, s_list) {
355 spin_unlock(&sb_lock);
356 down_read(&sb->s_umount);
358 up_read(&sb->s_umount);
360 if (__put_super_and_need_restart(sb))
364 spin_unlock(&sb_lock);
368 * Call the ->sync_fs super_op against all filesytems which are r/w and
369 * which implement it.
371 * This operation is careful to avoid the livelock which could easily happen
372 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
373 * is used only here. We set it against all filesystems and then clear it as
374 * we sync them. So redirtied filesystems are skipped.
376 * But if process A is currently running sync_filesytems and then process B
377 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
378 * flags again, which will cause process A to resync everything. Fix that with
381 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
383 void sync_filesystems(int wait)
385 struct super_block *sb;
386 static DEFINE_MUTEX(mutex);
388 mutex_lock(&mutex); /* Could be down_interruptible */
390 list_for_each_entry(sb, &super_blocks, s_list) {
391 if (!sb->s_op->sync_fs)
393 if (sb->s_flags & MS_RDONLY)
395 sb->s_need_sync_fs = 1;
399 list_for_each_entry(sb, &super_blocks, s_list) {
400 if (!sb->s_need_sync_fs)
402 sb->s_need_sync_fs = 0;
403 if (sb->s_flags & MS_RDONLY)
404 continue; /* hm. Was remounted r/o meanwhile */
406 spin_unlock(&sb_lock);
407 down_read(&sb->s_umount);
408 if (sb->s_root && (wait || sb->s_dirt))
409 sb->s_op->sync_fs(sb, wait);
410 up_read(&sb->s_umount);
411 /* restart only when sb is no longer on the list */
413 if (__put_super_and_need_restart(sb))
416 spin_unlock(&sb_lock);
417 mutex_unlock(&mutex);
421 * get_super - get the superblock of a device
422 * @bdev: device to get the superblock for
424 * Scans the superblock list and finds the superblock of the file system
425 * mounted on the device given. %NULL is returned if no match is found.
428 struct super_block * get_super(struct block_device *bdev)
430 struct super_block *sb;
437 list_for_each_entry(sb, &super_blocks, s_list) {
438 if (sb->s_bdev == bdev) {
440 spin_unlock(&sb_lock);
441 down_read(&sb->s_umount);
444 up_read(&sb->s_umount);
445 /* restart only when sb is no longer on the list */
447 if (__put_super_and_need_restart(sb))
451 spin_unlock(&sb_lock);
455 EXPORT_SYMBOL(get_super);
457 struct super_block * user_get_super(dev_t dev)
459 struct super_block *sb;
463 list_for_each_entry(sb, &super_blocks, s_list) {
464 if (sb->s_dev == dev) {
466 spin_unlock(&sb_lock);
467 down_read(&sb->s_umount);
470 up_read(&sb->s_umount);
471 /* restart only when sb is no longer on the list */
473 if (__put_super_and_need_restart(sb))
477 spin_unlock(&sb_lock);
481 asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
483 struct super_block *s;
488 s = user_get_super(new_decode_dev(dev));
491 err = vfs_statfs(s, &sbuf);
496 memset(&tmp,0,sizeof(struct ustat));
497 tmp.f_tfree = sbuf.f_bfree;
498 tmp.f_tinode = sbuf.f_ffree;
500 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
507 * @sb: superblock in question
509 * All files are marked read/only. We don't care about pending
510 * delete files so this should be used in 'force' mode only
513 static void mark_files_ro(struct super_block *sb)
518 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
519 if (S_ISREG(f->f_dentry->d_inode->i_mode) && file_count(f))
520 f->f_mode &= ~FMODE_WRITE;
526 * do_remount_sb - asks filesystem to change mount options.
527 * @sb: superblock in question
528 * @flags: numeric part of options
529 * @data: the rest of options
530 * @force: whether or not to force the change
532 * Alters the mount options of a mounted file system.
534 int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
538 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
540 if (flags & MS_RDONLY)
542 shrink_dcache_sb(sb);
545 /* If we are remounting RDONLY and current sb is read/write,
546 make sure there are no rw files opened */
547 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
550 else if (!fs_may_remount_ro(sb))
554 if (sb->s_op->remount_fs) {
556 retval = sb->s_op->remount_fs(sb, &flags, data);
561 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
565 static void do_emergency_remount(unsigned long foo)
567 struct super_block *sb;
570 list_for_each_entry(sb, &super_blocks, s_list) {
572 spin_unlock(&sb_lock);
573 down_read(&sb->s_umount);
574 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
576 * ->remount_fs needs lock_kernel().
578 * What lock protects sb->s_flags??
581 do_remount_sb(sb, MS_RDONLY, NULL, 1);
587 spin_unlock(&sb_lock);
588 printk("Emergency Remount complete\n");
591 void emergency_remount(void)
593 pdflush_operation(do_emergency_remount, 0);
597 * Unnamed block devices are dummy devices used by virtual
598 * filesystems which don't use real block-devices. -- jrs
601 static struct idr unnamed_dev_idr;
602 static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
604 int set_anon_super(struct super_block *s, void *data)
610 if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0)
612 spin_lock(&unnamed_dev_lock);
613 error = idr_get_new(&unnamed_dev_idr, NULL, &dev);
614 spin_unlock(&unnamed_dev_lock);
615 if (error == -EAGAIN)
616 /* We raced and lost with another CPU. */
621 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
622 spin_lock(&unnamed_dev_lock);
623 idr_remove(&unnamed_dev_idr, dev);
624 spin_unlock(&unnamed_dev_lock);
627 s->s_dev = MKDEV(0, dev & MINORMASK);
631 EXPORT_SYMBOL(set_anon_super);
633 void kill_anon_super(struct super_block *sb)
635 int slot = MINOR(sb->s_dev);
637 generic_shutdown_super(sb);
638 spin_lock(&unnamed_dev_lock);
639 idr_remove(&unnamed_dev_idr, slot);
640 spin_unlock(&unnamed_dev_lock);
643 EXPORT_SYMBOL(kill_anon_super);
645 void __init unnamed_dev_init(void)
647 idr_init(&unnamed_dev_idr);
650 void kill_litter_super(struct super_block *sb)
653 d_genocide(sb->s_root);
657 EXPORT_SYMBOL(kill_litter_super);
659 static int set_bdev_super(struct super_block *s, void *data)
662 s->s_dev = s->s_bdev->bd_dev;
666 static int test_bdev_super(struct super_block *s, void *data)
668 return (void *)s->s_bdev == data;
671 static void bdev_uevent(struct block_device *bdev, enum kobject_action action)
675 kobject_uevent(&bdev->bd_part->kobj, action);
677 kobject_uevent(&bdev->bd_disk->kobj, action);
681 struct super_block *get_sb_bdev(struct file_system_type *fs_type,
682 int flags, const char *dev_name, void *data,
683 int (*fill_super)(struct super_block *, void *, int))
685 struct block_device *bdev;
686 struct super_block *s;
689 bdev = open_bdev_excl(dev_name, flags, fs_type);
691 return (struct super_block *)bdev;
694 * once the super is inserted into the list by sget, s_umount
695 * will protect the lockfs code from trying to start a snapshot
696 * while we are mounting
698 mutex_lock(&bdev->bd_mount_mutex);
699 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
700 mutex_unlock(&bdev->bd_mount_mutex);
705 if ((flags ^ s->s_flags) & MS_RDONLY) {
706 up_write(&s->s_umount);
712 char b[BDEVNAME_SIZE];
715 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
716 sb_set_blocksize(s, block_size(bdev));
717 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
719 up_write(&s->s_umount);
723 s->s_flags |= MS_ACTIVE;
724 bdev_uevent(bdev, KOBJ_MOUNT);
731 close_bdev_excl(bdev);
735 EXPORT_SYMBOL(get_sb_bdev);
737 void kill_block_super(struct super_block *sb)
739 struct block_device *bdev = sb->s_bdev;
741 bdev_uevent(bdev, KOBJ_UMOUNT);
742 generic_shutdown_super(sb);
744 close_bdev_excl(bdev);
747 EXPORT_SYMBOL(kill_block_super);
749 struct super_block *get_sb_nodev(struct file_system_type *fs_type,
750 int flags, void *data,
751 int (*fill_super)(struct super_block *, void *, int))
754 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
761 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
763 up_write(&s->s_umount);
765 return ERR_PTR(error);
767 s->s_flags |= MS_ACTIVE;
771 EXPORT_SYMBOL(get_sb_nodev);
773 static int compare_single(struct super_block *s, void *p)
778 struct super_block *get_sb_single(struct file_system_type *fs_type,
779 int flags, void *data,
780 int (*fill_super)(struct super_block *, void *, int))
782 struct super_block *s;
785 s = sget(fs_type, compare_single, set_anon_super, NULL);
790 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
792 up_write(&s->s_umount);
794 return ERR_PTR(error);
796 s->s_flags |= MS_ACTIVE;
798 do_remount_sb(s, flags, data, 0);
802 EXPORT_SYMBOL(get_sb_single);
805 do_kern_mount(const char *fstype, int flags, const char *name, void *data)
807 struct file_system_type *type = get_fs_type(fstype);
808 struct super_block *sb;
809 struct vfsmount *mnt;
811 char *secdata = NULL;
814 return ERR_PTR(-ENODEV);
816 sb = ERR_PTR(-EPERM);
817 if ((type->fs_flags & FS_BINARY_MOUNTDATA) &&
818 !vx_capable(CAP_SYS_ADMIN, VXC_BINARY_MOUNT))
821 sb = ERR_PTR(-ENOMEM);
822 mnt = alloc_vfsmnt(name);
827 secdata = alloc_secdata();
829 sb = ERR_PTR(-ENOMEM);
833 error = security_sb_copy_data(type, data, secdata);
836 goto out_free_secdata;
840 sb = type->get_sb(type, flags, name, data);
842 goto out_free_secdata;
845 if (!capable(CAP_SYS_ADMIN) && !sb->s_bdev &&
846 (sb->s_magic != PROC_SUPER_MAGIC) &&
847 (sb->s_magic != DEVPTS_SUPER_MAGIC))
850 error = security_sb_kern_mount(sb, secdata);
854 mnt->mnt_root = dget(sb->s_root);
855 mnt->mnt_mountpoint = sb->s_root;
856 mnt->mnt_parent = mnt;
857 up_write(&sb->s_umount);
858 free_secdata(secdata);
859 put_filesystem(type);
862 up_write(&sb->s_umount);
863 deactivate_super(sb);
866 free_secdata(secdata);
870 put_filesystem(type);
871 return (struct vfsmount *)sb;
874 EXPORT_SYMBOL_GPL(do_kern_mount);
876 struct vfsmount *kern_mount(struct file_system_type *type)
878 return do_kern_mount(type->name, 0, type->name, NULL);
881 EXPORT_SYMBOL(kern_mount);