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 * put_super - drop a temporary reference to superblock
106 * @s: superblock in question
108 * Drops a temporary reference, frees superblock if there's no
111 static inline void put_super(struct super_block *s)
116 spin_unlock(&sb_lock);
120 * deactivate_super - drop an active reference to superblock
121 * @s: superblock to deactivate
123 * Drops an active reference to superblock, acquiring a temprory one if
124 * there is no active references left. In that case we lock superblock,
125 * tell fs driver to shut it down and drop the temporary reference we
128 void deactivate_super(struct super_block *s)
130 struct file_system_type *fs = s->s_type;
131 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
132 s->s_count -= S_BIAS-1;
133 spin_unlock(&sb_lock);
134 down_write(&s->s_umount);
141 EXPORT_SYMBOL(deactivate_super);
144 * grab_super - acquire an active reference
145 * @s: reference we are trying to make active
147 * Tries to acquire an active reference. grab_super() is used when we
148 * had just found a superblock in super_blocks or fs_type->fs_supers
149 * and want to turn it into a full-blown active reference. grab_super()
150 * is called with sb_lock held and drops it. Returns 1 in case of
151 * success, 0 if we had failed (superblock contents was already dead or
152 * dying when grab_super() had been called).
154 static int grab_super(struct super_block *s)
157 spin_unlock(&sb_lock);
158 down_write(&s->s_umount);
161 if (s->s_count > S_BIAS) {
162 atomic_inc(&s->s_active);
164 spin_unlock(&sb_lock);
167 spin_unlock(&sb_lock);
169 up_write(&s->s_umount);
176 * generic_shutdown_super - common helper for ->kill_sb()
177 * @sb: superblock to kill
179 * generic_shutdown_super() does all fs-independent work on superblock
180 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
181 * that need destruction out of superblock, call generic_shutdown_super()
182 * and release aforementioned objects. Note: dentries and inodes _are_
183 * taken care of and do not need specific handling.
185 void generic_shutdown_super(struct super_block *sb)
187 struct dentry *root = sb->s_root;
188 struct super_operations *sop = sb->s_op;
192 shrink_dcache_parent(root);
193 shrink_dcache_anon(&sb->s_anon);
198 sb->s_flags &= ~MS_ACTIVE;
199 /* bad name - it should be evict_inodes() */
200 invalidate_inodes(sb);
202 if (sop->write_super && sb->s_dirt)
203 sop->write_super(sb);
207 /* Forget any remaining inodes */
208 if (invalidate_inodes(sb)) {
209 printk("VFS: Busy inodes after unmount. "
210 "Self-destruct in 5 seconds. Have a nice day...\n");
217 list_del(&sb->s_list);
218 list_del(&sb->s_instances);
219 spin_unlock(&sb_lock);
220 up_write(&sb->s_umount);
223 EXPORT_SYMBOL(generic_shutdown_super);
226 * sget - find or create a superblock
227 * @type: filesystem type superblock should belong to
228 * @test: comparison callback
229 * @set: setup callback
230 * @data: argument to each of them
232 struct super_block *sget(struct file_system_type *type,
233 int (*test)(struct super_block *,void *),
234 int (*set)(struct super_block *,void *),
237 struct super_block *s = NULL;
243 if (test) list_for_each(p, &type->fs_supers) {
244 struct super_block *old;
245 old = list_entry(p, struct super_block, s_instances);
246 if (!test(old, data))
248 if (!grab_super(old))
255 spin_unlock(&sb_lock);
258 return ERR_PTR(-ENOMEM);
264 spin_unlock(&sb_lock);
269 list_add(&s->s_list, super_blocks.prev);
270 list_add(&s->s_instances, &type->fs_supers);
271 spin_unlock(&sb_lock);
272 get_filesystem(type);
278 void drop_super(struct super_block *sb)
280 up_read(&sb->s_umount);
284 EXPORT_SYMBOL(drop_super);
286 static inline void write_super(struct super_block *sb)
289 if (sb->s_root && sb->s_dirt)
290 if (sb->s_op->write_super)
291 sb->s_op->write_super(sb);
296 * Note: check the dirty flag before waiting, so we don't
297 * hold up the sync while mounting a device. (The newly
298 * mounted device won't need syncing.)
300 void sync_supers(void)
302 struct super_block * sb;
305 sb = sb_entry(super_blocks.next);
306 while (sb != sb_entry(&super_blocks))
309 spin_unlock(&sb_lock);
310 down_read(&sb->s_umount);
315 sb = sb_entry(sb->s_list.next);
316 spin_unlock(&sb_lock);
320 * Call the ->sync_fs super_op against all filesytems which are r/w and
321 * which implement it.
323 * This operation is careful to avoid the livelock which could easily happen
324 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
325 * is used only here. We set it against all filesystems and then clear it as
326 * we sync them. So redirtied filesystems are skipped.
328 * But if process A is currently running sync_filesytems and then process B
329 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
330 * flags again, which will cause process A to resync everything. Fix that with
333 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
335 void sync_filesystems(int wait)
337 struct super_block *sb;
338 static DECLARE_MUTEX(mutex);
340 down(&mutex); /* Could be down_interruptible */
342 for (sb = sb_entry(super_blocks.next); sb != sb_entry(&super_blocks);
343 sb = sb_entry(sb->s_list.next)) {
344 if (!sb->s_op->sync_fs)
346 if (sb->s_flags & MS_RDONLY)
348 sb->s_need_sync_fs = 1;
350 spin_unlock(&sb_lock);
354 for (sb = sb_entry(super_blocks.next); sb != sb_entry(&super_blocks);
355 sb = sb_entry(sb->s_list.next)) {
356 if (!sb->s_need_sync_fs)
358 sb->s_need_sync_fs = 0;
359 if (sb->s_flags & MS_RDONLY)
360 continue; /* hm. Was remounted r/o meanwhile */
362 spin_unlock(&sb_lock);
363 down_read(&sb->s_umount);
364 if (sb->s_root && (wait || sb->s_dirt))
365 sb->s_op->sync_fs(sb, wait);
369 spin_unlock(&sb_lock);
374 * get_super - get the superblock of a device
375 * @bdev: device to get the superblock for
377 * Scans the superblock list and finds the superblock of the file system
378 * mounted on the device given. %NULL is returned if no match is found.
381 struct super_block * get_super(struct block_device *bdev)
388 list_for_each(p, &super_blocks) {
389 struct super_block *s = sb_entry(p);
390 if (s->s_bdev == bdev) {
392 spin_unlock(&sb_lock);
393 down_read(&s->s_umount);
400 spin_unlock(&sb_lock);
404 EXPORT_SYMBOL(get_super);
406 struct super_block * user_get_super(dev_t dev)
412 list_for_each(p, &super_blocks) {
413 struct super_block *s = sb_entry(p);
414 if (s->s_dev == dev) {
416 spin_unlock(&sb_lock);
417 down_read(&s->s_umount);
424 spin_unlock(&sb_lock);
428 EXPORT_SYMBOL(user_get_super);
430 asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
432 struct super_block *s;
437 s = user_get_super(new_decode_dev(dev));
440 err = vfs_statfs(s, &sbuf);
445 memset(&tmp,0,sizeof(struct ustat));
446 tmp.f_tfree = sbuf.f_bfree;
447 tmp.f_tinode = sbuf.f_ffree;
449 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
456 * @sb: superblock in question
458 * All files are marked read/only. We don't care about pending
459 * delete files so this should be used in 'force' mode only
462 static void mark_files_ro(struct super_block *sb)
467 list_for_each_entry(f, &sb->s_files, f_list) {
468 if (S_ISREG(f->f_dentry->d_inode->i_mode) && file_count(f))
469 f->f_mode &= ~FMODE_WRITE;
475 * do_remount_sb - asks filesystem to change mount options.
476 * @sb: superblock in question
477 * @flags: numeric part of options
478 * @data: the rest of options
479 * @force: whether or not to force the change
481 * Alters the mount options of a mounted file system.
483 int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
487 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
489 if (flags & MS_RDONLY)
491 shrink_dcache_sb(sb);
494 /* If we are remounting RDONLY and current sb is read/write,
495 make sure there are no rw files opened */
496 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
499 else if (!fs_may_remount_ro(sb))
503 if (sb->s_op->remount_fs) {
505 retval = sb->s_op->remount_fs(sb, &flags, data);
510 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
514 static void do_emergency_remount(unsigned long foo)
516 struct super_block *sb;
519 list_for_each_entry(sb, &super_blocks, s_list) {
521 spin_unlock(&sb_lock);
522 down_read(&sb->s_umount);
523 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
525 * ->remount_fs needs lock_kernel().
527 * What lock protects sb->s_flags??
530 do_remount_sb(sb, MS_RDONLY, NULL, 1);
536 spin_unlock(&sb_lock);
537 printk("Emergency Remount complete\n");
540 void emergency_remount(void)
542 pdflush_operation(do_emergency_remount, 0);
546 * Unnamed block devices are dummy devices used by virtual
547 * filesystems which don't use real block-devices. -- jrs
550 static struct idr unnamed_dev_idr;
551 static spinlock_t unnamed_dev_lock = SPIN_LOCK_UNLOCKED;/* protects the above */
553 int set_anon_super(struct super_block *s, void *data)
557 spin_lock(&unnamed_dev_lock);
558 if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0) {
559 spin_unlock(&unnamed_dev_lock);
562 dev = idr_get_new(&unnamed_dev_idr, NULL);
563 spin_unlock(&unnamed_dev_lock);
565 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
566 spin_lock(&unnamed_dev_lock);
567 idr_remove(&unnamed_dev_idr, dev);
568 spin_unlock(&unnamed_dev_lock);
571 s->s_dev = MKDEV(0, dev & MINORMASK);
575 EXPORT_SYMBOL(set_anon_super);
577 void kill_anon_super(struct super_block *sb)
579 int slot = MINOR(sb->s_dev);
581 generic_shutdown_super(sb);
582 spin_lock(&unnamed_dev_lock);
583 idr_remove(&unnamed_dev_idr, slot);
584 spin_unlock(&unnamed_dev_lock);
587 EXPORT_SYMBOL(kill_anon_super);
589 void __init unnamed_dev_init(void)
591 idr_init(&unnamed_dev_idr);
594 void kill_litter_super(struct super_block *sb)
597 d_genocide(sb->s_root);
601 EXPORT_SYMBOL(kill_litter_super);
603 static int set_bdev_super(struct super_block *s, void *data)
606 s->s_dev = s->s_bdev->bd_dev;
610 static int test_bdev_super(struct super_block *s, void *data)
612 return (void *)s->s_bdev == data;
615 struct super_block *get_sb_bdev(struct file_system_type *fs_type,
616 int flags, const char *dev_name, void *data,
617 int (*fill_super)(struct super_block *, void *, int))
619 struct block_device *bdev;
620 struct super_block *s;
623 bdev = open_bdev_excl(dev_name, flags, fs_type);
625 return (struct super_block *)bdev;
628 * once the super is inserted into the list by sget, s_umount
629 * will protect the lockfs code from trying to start a snapshot
630 * while we are mounting
632 down(&bdev->bd_mount_sem);
633 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
634 up(&bdev->bd_mount_sem);
639 if ((flags ^ s->s_flags) & MS_RDONLY) {
640 up_write(&s->s_umount);
646 char b[BDEVNAME_SIZE];
649 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
650 s->s_old_blocksize = block_size(bdev);
651 sb_set_blocksize(s, s->s_old_blocksize);
652 error = fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
654 up_write(&s->s_umount);
658 s->s_flags |= MS_ACTIVE;
664 close_bdev_excl(bdev);
668 EXPORT_SYMBOL(get_sb_bdev);
670 void kill_block_super(struct super_block *sb)
672 struct block_device *bdev = sb->s_bdev;
673 generic_shutdown_super(sb);
674 set_blocksize(bdev, sb->s_old_blocksize);
675 close_bdev_excl(bdev);
678 EXPORT_SYMBOL(kill_block_super);
680 struct super_block *get_sb_nodev(struct file_system_type *fs_type,
681 int flags, void *data,
682 int (*fill_super)(struct super_block *, void *, int))
685 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
692 error = fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
694 up_write(&s->s_umount);
696 return ERR_PTR(error);
698 s->s_flags |= MS_ACTIVE;
702 EXPORT_SYMBOL(get_sb_nodev);
704 static int compare_single(struct super_block *s, void *p)
709 struct super_block *get_sb_single(struct file_system_type *fs_type,
710 int flags, void *data,
711 int (*fill_super)(struct super_block *, void *, int))
713 struct super_block *s;
716 s = sget(fs_type, compare_single, set_anon_super, NULL);
721 error = fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
723 up_write(&s->s_umount);
725 return ERR_PTR(error);
727 s->s_flags |= MS_ACTIVE;
729 do_remount_sb(s, flags, data, 0);
733 EXPORT_SYMBOL(get_sb_single);
736 do_kern_mount(const char *fstype, int flags, const char *name, void *data)
738 struct file_system_type *type = get_fs_type(fstype);
739 struct super_block *sb = ERR_PTR(-ENOMEM);
740 struct vfsmount *mnt;
742 char *secdata = NULL;
745 return ERR_PTR(-ENODEV);
747 mnt = alloc_vfsmnt(name);
752 secdata = alloc_secdata();
754 sb = ERR_PTR(-ENOMEM);
758 error = security_sb_copy_data(type, data, secdata);
761 goto out_free_secdata;
765 sb = type->get_sb(type, flags, name, data);
767 goto out_free_secdata;
768 error = security_sb_kern_mount(sb, secdata);
772 mnt->mnt_root = dget(sb->s_root);
773 mnt->mnt_mountpoint = sb->s_root;
774 mnt->mnt_parent = mnt;
775 up_write(&sb->s_umount);
776 put_filesystem(type);
779 up_write(&sb->s_umount);
780 deactivate_super(sb);
783 free_secdata(secdata);
787 put_filesystem(type);
788 return (struct vfsmount *)sb;
791 struct vfsmount *kern_mount(struct file_system_type *type)
793 return do_kern_mount(type->name, 0, type->name, NULL);
796 EXPORT_SYMBOL(kern_mount);