On Tue, Nov 06, 2007 at 02:33:53AM -0800, akpm@linux-foundation.org wrote:
[mmotm.git] / fs / super.c
blobfb35c73dde2d30687824ed199bceed0890a29ddd
1 /*
2 * linux/fs/super.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * super.c contains code to handle: - mount structures
7 * - super-block tables
8 * - filesystem drivers list
9 * - mount system call
10 * - umount system call
11 * - ustat 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>
41 #include "internal.h"
44 LIST_HEAD(super_blocks);
45 DEFINE_SPINLOCK(sb_lock);
47 /**
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;
59 if (s) {
60 if (security_sb_alloc(s)) {
61 kfree(s);
62 s = NULL;
63 goto out;
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
84 * one.
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
89 * risk of deadlocks.
91 * Annotate this by putting this lock in a different
92 * subclass.
94 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
95 s->s_count = S_BIAS;
96 atomic_set(&s->s_active, 1);
97 mutex_init(&s->s_vfs_rename_mutex);
98 lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key);
99 mutex_init(&s->s_dquot.dqio_mutex);
100 mutex_init(&s->s_dquot.dqonoff_mutex);
101 init_rwsem(&s->s_dquot.dqptr_sem);
102 init_waitqueue_head(&s->s_wait_unfrozen);
103 s->s_maxbytes = MAX_NON_LFS;
104 s->dq_op = sb_dquot_ops;
105 s->s_qcop = sb_quotactl_ops;
106 s->s_op = &default_op;
107 s->s_time_gran = 1000000000;
109 out:
110 return s;
114 * destroy_super - frees a superblock
115 * @s: superblock to free
117 * Frees a superblock.
119 static inline void destroy_super(struct super_block *s)
121 security_sb_free(s);
122 kfree(s->s_subtype);
123 kfree(s->s_options);
124 kfree(s);
127 /* Superblock refcounting */
130 * Drop a superblock's refcount. Returns non-zero if the superblock was
131 * destroyed. The caller must hold sb_lock.
133 static int __put_super(struct super_block *sb)
135 int ret = 0;
137 if (!--sb->s_count) {
138 destroy_super(sb);
139 ret = 1;
141 return ret;
145 * Drop a superblock's refcount.
146 * Returns non-zero if the superblock is about to be destroyed and
147 * at least is already removed from super_blocks list, so if we are
148 * making a loop through super blocks then we need to restart.
149 * The caller must hold sb_lock.
151 int __put_super_and_need_restart(struct super_block *sb)
153 /* check for race with generic_shutdown_super() */
154 if (list_empty(&sb->s_list)) {
155 /* super block is removed, need to restart... */
156 __put_super(sb);
157 return 1;
159 /* can't be the last, since s_list is still in use */
160 sb->s_count--;
161 BUG_ON(sb->s_count == 0);
162 return 0;
166 * put_super - drop a temporary reference to superblock
167 * @sb: superblock in question
169 * Drops a temporary reference, frees superblock if there's no
170 * references left.
172 void put_super(struct super_block *sb)
174 spin_lock(&sb_lock);
175 __put_super(sb);
176 spin_unlock(&sb_lock);
181 * deactivate_super - drop an active reference to superblock
182 * @s: superblock to deactivate
184 * Drops an active reference to superblock, acquiring a temprory one if
185 * there is no active references left. In that case we lock superblock,
186 * tell fs driver to shut it down and drop the temporary reference we
187 * had just acquired.
189 void deactivate_super(struct super_block *s)
191 struct file_system_type *fs = s->s_type;
192 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
193 s->s_count -= S_BIAS-1;
194 spin_unlock(&sb_lock);
195 vfs_dq_off(s, 0);
196 down_write(&s->s_umount);
197 fs->kill_sb(s);
198 put_filesystem(fs);
199 put_super(s);
203 EXPORT_SYMBOL(deactivate_super);
206 * deactivate_locked_super - drop an active reference to superblock
207 * @s: superblock to deactivate
209 * Equivalent of up_write(&s->s_umount); deactivate_super(s);, except that
210 * it does not unlock it until it's all over. As the result, it's safe to
211 * use to dispose of new superblock on ->get_sb() failure exits - nobody
212 * will see the sucker until it's all over. Equivalent using up_write +
213 * deactivate_super is safe for that purpose only if superblock is either
214 * safe to use or has NULL ->s_root when we unlock.
216 void deactivate_locked_super(struct super_block *s)
218 struct file_system_type *fs = s->s_type;
219 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
220 s->s_count -= S_BIAS-1;
221 spin_unlock(&sb_lock);
222 vfs_dq_off(s, 0);
223 fs->kill_sb(s);
224 put_filesystem(fs);
225 put_super(s);
226 } else {
227 up_write(&s->s_umount);
231 EXPORT_SYMBOL(deactivate_locked_super);
234 * grab_super - acquire an active reference
235 * @s: reference we are trying to make active
237 * Tries to acquire an active reference. grab_super() is used when we
238 * had just found a superblock in super_blocks or fs_type->fs_supers
239 * and want to turn it into a full-blown active reference. grab_super()
240 * is called with sb_lock held and drops it. Returns 1 in case of
241 * success, 0 if we had failed (superblock contents was already dead or
242 * dying when grab_super() had been called).
244 static int grab_super(struct super_block *s) __releases(sb_lock)
246 s->s_count++;
247 spin_unlock(&sb_lock);
248 down_write(&s->s_umount);
249 if (s->s_root) {
250 spin_lock(&sb_lock);
251 if (s->s_count > S_BIAS) {
252 atomic_inc(&s->s_active);
253 s->s_count--;
254 spin_unlock(&sb_lock);
255 return 1;
257 spin_unlock(&sb_lock);
259 up_write(&s->s_umount);
260 put_super(s);
261 yield();
262 return 0;
266 * Superblock locking. We really ought to get rid of these two.
268 void lock_super(struct super_block * sb)
270 get_fs_excl();
271 mutex_lock(&sb->s_lock);
274 void unlock_super(struct super_block * sb)
276 put_fs_excl();
277 mutex_unlock(&sb->s_lock);
280 EXPORT_SYMBOL(lock_super);
281 EXPORT_SYMBOL(unlock_super);
284 * generic_shutdown_super - common helper for ->kill_sb()
285 * @sb: superblock to kill
287 * generic_shutdown_super() does all fs-independent work on superblock
288 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
289 * that need destruction out of superblock, call generic_shutdown_super()
290 * and release aforementioned objects. Note: dentries and inodes _are_
291 * taken care of and do not need specific handling.
293 * Upon calling this function, the filesystem may no longer alter or
294 * rearrange the set of dentries belonging to this super_block, nor may it
295 * change the attachments of dentries to inodes.
297 void generic_shutdown_super(struct super_block *sb)
299 const struct super_operations *sop = sb->s_op;
302 if (sb->s_root) {
303 shrink_dcache_for_umount(sb);
304 sync_filesystem(sb);
305 get_fs_excl();
306 sb->s_flags &= ~MS_ACTIVE;
308 /* bad name - it should be evict_inodes() */
309 invalidate_inodes(sb);
311 if (sop->put_super)
312 sop->put_super(sb);
314 /* Forget any remaining inodes */
315 if (invalidate_inodes(sb)) {
316 printk("VFS: Busy inodes after unmount of %s. "
317 "Self-destruct in 5 seconds. Have a nice day...\n",
318 sb->s_id);
320 put_fs_excl();
322 spin_lock(&sb_lock);
323 /* should be initialized for __put_super_and_need_restart() */
324 list_del_init(&sb->s_list);
325 list_del(&sb->s_instances);
326 spin_unlock(&sb_lock);
327 up_write(&sb->s_umount);
330 EXPORT_SYMBOL(generic_shutdown_super);
333 * sget - find or create a superblock
334 * @type: filesystem type superblock should belong to
335 * @test: comparison callback
336 * @set: setup callback
337 * @data: argument to each of them
339 struct super_block *sget(struct file_system_type *type,
340 int (*test)(struct super_block *,void *),
341 int (*set)(struct super_block *,void *),
342 void *data)
344 struct super_block *s = NULL;
345 struct super_block *old;
346 int err;
348 retry:
349 spin_lock(&sb_lock);
350 if (test) {
351 list_for_each_entry(old, &type->fs_supers, s_instances) {
352 if (!test(old, data))
353 continue;
354 if (!grab_super(old))
355 goto retry;
356 if (s) {
357 up_write(&s->s_umount);
358 destroy_super(s);
360 return old;
363 if (!s) {
364 spin_unlock(&sb_lock);
365 s = alloc_super(type);
366 if (!s)
367 return ERR_PTR(-ENOMEM);
368 goto retry;
371 err = set(s, data);
372 if (err) {
373 spin_unlock(&sb_lock);
374 up_write(&s->s_umount);
375 destroy_super(s);
376 return ERR_PTR(err);
378 s->s_type = type;
379 strlcpy(s->s_id, type->name, sizeof(s->s_id));
380 list_add_tail(&s->s_list, &super_blocks);
381 list_add(&s->s_instances, &type->fs_supers);
382 spin_unlock(&sb_lock);
383 get_filesystem(type);
384 return s;
387 EXPORT_SYMBOL(sget);
389 void drop_super(struct super_block *sb)
391 up_read(&sb->s_umount);
392 put_super(sb);
395 EXPORT_SYMBOL(drop_super);
398 * sync_supers - helper for periodic superblock writeback
400 * Call the write_super method if present on all dirty superblocks in
401 * the system. This is for the periodic writeback used by most older
402 * filesystems. For data integrity superblock writeback use
403 * sync_filesystems() instead.
405 * Note: check the dirty flag before waiting, so we don't
406 * hold up the sync while mounting a device. (The newly
407 * mounted device won't need syncing.)
409 void sync_supers(void)
411 struct super_block *sb;
413 spin_lock(&sb_lock);
414 restart:
415 list_for_each_entry(sb, &super_blocks, s_list) {
416 if (sb->s_op->write_super && sb->s_dirt) {
417 sb->s_count++;
418 spin_unlock(&sb_lock);
420 down_read(&sb->s_umount);
421 if (sb->s_root && sb->s_dirt)
422 sb->s_op->write_super(sb);
423 up_read(&sb->s_umount);
425 spin_lock(&sb_lock);
426 if (__put_super_and_need_restart(sb))
427 goto restart;
430 spin_unlock(&sb_lock);
434 * get_super - get the superblock of a device
435 * @bdev: device to get the superblock for
437 * Scans the superblock list and finds the superblock of the file system
438 * mounted on the device given. %NULL is returned if no match is found.
441 struct super_block * get_super(struct block_device *bdev)
443 struct super_block *sb;
445 if (!bdev)
446 return NULL;
448 spin_lock(&sb_lock);
449 rescan:
450 list_for_each_entry(sb, &super_blocks, s_list) {
451 if (sb->s_bdev == bdev) {
452 sb->s_count++;
453 spin_unlock(&sb_lock);
454 down_read(&sb->s_umount);
455 if (sb->s_root)
456 return sb;
457 up_read(&sb->s_umount);
458 /* restart only when sb is no longer on the list */
459 spin_lock(&sb_lock);
460 if (__put_super_and_need_restart(sb))
461 goto rescan;
464 spin_unlock(&sb_lock);
465 return NULL;
468 EXPORT_SYMBOL(get_super);
471 * get_active_super - get an active reference to the superblock of a device
472 * @bdev: device to get the superblock for
474 * Scans the superblock list and finds the superblock of the file system
475 * mounted on the device given. Returns the superblock with an active
476 * reference and s_umount held exclusively or %NULL if none was found.
478 struct super_block *get_active_super(struct block_device *bdev)
480 struct super_block *sb;
482 if (!bdev)
483 return NULL;
485 spin_lock(&sb_lock);
486 list_for_each_entry(sb, &super_blocks, s_list) {
487 if (sb->s_bdev != bdev)
488 continue;
490 sb->s_count++;
491 spin_unlock(&sb_lock);
492 down_write(&sb->s_umount);
493 if (sb->s_root) {
494 spin_lock(&sb_lock);
495 if (sb->s_count > S_BIAS) {
496 atomic_inc(&sb->s_active);
497 sb->s_count--;
498 spin_unlock(&sb_lock);
499 return sb;
501 spin_unlock(&sb_lock);
503 up_write(&sb->s_umount);
504 put_super(sb);
505 yield();
506 spin_lock(&sb_lock);
508 spin_unlock(&sb_lock);
509 return NULL;
512 struct super_block * user_get_super(dev_t dev)
514 struct super_block *sb;
516 spin_lock(&sb_lock);
517 rescan:
518 list_for_each_entry(sb, &super_blocks, s_list) {
519 if (sb->s_dev == dev) {
520 sb->s_count++;
521 spin_unlock(&sb_lock);
522 down_read(&sb->s_umount);
523 if (sb->s_root)
524 return sb;
525 up_read(&sb->s_umount);
526 /* restart only when sb is no longer on the list */
527 spin_lock(&sb_lock);
528 if (__put_super_and_need_restart(sb))
529 goto rescan;
532 spin_unlock(&sb_lock);
533 return NULL;
536 SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf)
538 struct super_block *s;
539 struct ustat tmp;
540 struct kstatfs sbuf;
541 int err = -EINVAL;
543 s = user_get_super(new_decode_dev(dev));
544 if (s == NULL)
545 goto out;
546 err = vfs_statfs(s->s_root, &sbuf);
547 drop_super(s);
548 if (err)
549 goto out;
551 memset(&tmp,0,sizeof(struct ustat));
552 tmp.f_tfree = sbuf.f_bfree;
553 tmp.f_tinode = sbuf.f_ffree;
555 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
556 out:
557 return err;
561 * do_remount_sb - asks filesystem to change mount options.
562 * @sb: superblock in question
563 * @flags: numeric part of options
564 * @data: the rest of options
565 * @force: whether or not to force the change
567 * Alters the mount options of a mounted file system.
569 int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
571 int retval;
572 int remount_rw, remount_ro;
574 if (sb->s_frozen != SB_UNFROZEN)
575 return -EBUSY;
577 #ifdef CONFIG_BLOCK
578 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
579 return -EACCES;
580 #endif
582 if (flags & MS_RDONLY)
583 acct_auto_close(sb);
584 shrink_dcache_sb(sb);
585 sync_filesystem(sb);
587 remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY);
588 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
590 /* If we are remounting RDONLY and current sb is read/write,
591 make sure there are no rw files opened */
592 if (remount_ro) {
593 if (force)
594 mark_files_ro(sb);
595 else if (!fs_may_remount_ro(sb))
596 return -EBUSY;
597 retval = vfs_dq_off(sb, 1);
598 if (retval < 0 && retval != -ENOSYS)
599 return -EBUSY;
602 if (sb->s_op->remount_fs) {
603 retval = sb->s_op->remount_fs(sb, &flags, data);
604 if (retval)
605 return retval;
607 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
608 if (remount_rw)
609 vfs_dq_quota_on_remount(sb);
611 * Some filesystems modify their metadata via some other path than the
612 * bdev buffer cache (eg. use a private mapping, or directories in
613 * pagecache, etc). Also file data modifications go via their own
614 * mappings. So If we try to mount readonly then copy the filesystem
615 * from bdev, we could get stale data, so invalidate it to give a best
616 * effort at coherency.
618 if (remount_ro && sb->s_bdev)
619 invalidate_bdev(sb->s_bdev);
620 return 0;
623 static void do_emergency_remount(struct work_struct *work)
625 struct super_block *sb;
627 spin_lock(&sb_lock);
628 list_for_each_entry(sb, &super_blocks, s_list) {
629 sb->s_count++;
630 spin_unlock(&sb_lock);
631 down_write(&sb->s_umount);
632 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
634 * ->remount_fs needs lock_kernel().
636 * What lock protects sb->s_flags??
638 do_remount_sb(sb, MS_RDONLY, NULL, 1);
640 up_write(&sb->s_umount);
641 put_super(sb);
642 spin_lock(&sb_lock);
644 spin_unlock(&sb_lock);
645 kfree(work);
646 printk("Emergency Remount complete\n");
649 void emergency_remount(void)
651 struct work_struct *work;
653 work = kmalloc(sizeof(*work), GFP_ATOMIC);
654 if (work) {
655 INIT_WORK(work, do_emergency_remount);
656 schedule_work(work);
661 * Unnamed block devices are dummy devices used by virtual
662 * filesystems which don't use real block-devices. -- jrs
665 static DEFINE_IDA(unnamed_dev_ida);
666 static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
667 static int unnamed_dev_start = 0; /* don't bother trying below it */
669 int set_anon_super(struct super_block *s, void *data)
671 int dev;
672 int error;
674 retry:
675 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
676 return -ENOMEM;
677 spin_lock(&unnamed_dev_lock);
678 error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
679 if (!error)
680 unnamed_dev_start = dev + 1;
681 spin_unlock(&unnamed_dev_lock);
682 if (error == -EAGAIN)
683 /* We raced and lost with another CPU. */
684 goto retry;
685 else if (error)
686 return -EAGAIN;
688 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
689 spin_lock(&unnamed_dev_lock);
690 ida_remove(&unnamed_dev_ida, dev);
691 if (unnamed_dev_start > dev)
692 unnamed_dev_start = dev;
693 spin_unlock(&unnamed_dev_lock);
694 return -EMFILE;
696 s->s_dev = MKDEV(0, dev & MINORMASK);
697 return 0;
700 EXPORT_SYMBOL(set_anon_super);
702 void kill_anon_super(struct super_block *sb)
704 int slot = MINOR(sb->s_dev);
706 generic_shutdown_super(sb);
707 spin_lock(&unnamed_dev_lock);
708 ida_remove(&unnamed_dev_ida, slot);
709 if (slot < unnamed_dev_start)
710 unnamed_dev_start = slot;
711 spin_unlock(&unnamed_dev_lock);
714 EXPORT_SYMBOL(kill_anon_super);
716 void kill_litter_super(struct super_block *sb)
718 if (sb->s_root)
719 d_genocide(sb->s_root);
720 kill_anon_super(sb);
723 EXPORT_SYMBOL(kill_litter_super);
725 static int ns_test_super(struct super_block *sb, void *data)
727 return sb->s_fs_info == data;
730 static int ns_set_super(struct super_block *sb, void *data)
732 sb->s_fs_info = data;
733 return set_anon_super(sb, NULL);
736 int get_sb_ns(struct file_system_type *fs_type, int flags, void *data,
737 int (*fill_super)(struct super_block *, void *, int),
738 struct vfsmount *mnt)
740 struct super_block *sb;
742 sb = sget(fs_type, ns_test_super, ns_set_super, data);
743 if (IS_ERR(sb))
744 return PTR_ERR(sb);
746 if (!sb->s_root) {
747 int err;
748 sb->s_flags = flags;
749 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
750 if (err) {
751 deactivate_locked_super(sb);
752 return err;
755 sb->s_flags |= MS_ACTIVE;
758 simple_set_mnt(mnt, sb);
759 return 0;
762 EXPORT_SYMBOL(get_sb_ns);
764 #ifdef CONFIG_BLOCK
765 static int set_bdev_super(struct super_block *s, void *data)
767 s->s_bdev = data;
768 s->s_dev = s->s_bdev->bd_dev;
771 * We set the bdi here to the queue backing, file systems can
772 * overwrite this in ->fill_super()
774 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
775 return 0;
778 static int test_bdev_super(struct super_block *s, void *data)
780 return (void *)s->s_bdev == data;
783 int get_sb_bdev(struct file_system_type *fs_type,
784 int flags, const char *dev_name, void *data,
785 int (*fill_super)(struct super_block *, void *, int),
786 struct vfsmount *mnt)
788 struct block_device *bdev;
789 struct super_block *s;
790 fmode_t mode = FMODE_READ;
791 int error = 0;
793 if (!(flags & MS_RDONLY))
794 mode |= FMODE_WRITE;
796 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
797 if (IS_ERR(bdev))
798 return PTR_ERR(bdev);
801 * once the super is inserted into the list by sget, s_umount
802 * will protect the lockfs code from trying to start a snapshot
803 * while we are mounting
805 mutex_lock(&bdev->bd_fsfreeze_mutex);
806 if (bdev->bd_fsfreeze_count > 0) {
807 mutex_unlock(&bdev->bd_fsfreeze_mutex);
808 error = -EBUSY;
809 goto error_bdev;
811 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
812 mutex_unlock(&bdev->bd_fsfreeze_mutex);
813 if (IS_ERR(s))
814 goto error_s;
816 if (s->s_root) {
817 if ((flags ^ s->s_flags) & MS_RDONLY) {
818 deactivate_locked_super(s);
819 error = -EBUSY;
820 goto error_bdev;
823 close_bdev_exclusive(bdev, mode);
824 } else {
825 char b[BDEVNAME_SIZE];
827 s->s_flags = flags;
828 s->s_mode = mode;
829 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
830 sb_set_blocksize(s, block_size(bdev));
831 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
832 if (error) {
833 deactivate_locked_super(s);
834 goto error;
837 s->s_flags |= MS_ACTIVE;
838 bdev->bd_super = s;
841 simple_set_mnt(mnt, s);
842 return 0;
844 error_s:
845 error = PTR_ERR(s);
846 error_bdev:
847 close_bdev_exclusive(bdev, mode);
848 error:
849 return error;
852 EXPORT_SYMBOL(get_sb_bdev);
854 void kill_block_super(struct super_block *sb)
856 struct block_device *bdev = sb->s_bdev;
857 fmode_t mode = sb->s_mode;
859 bdev->bd_super = NULL;
860 generic_shutdown_super(sb);
861 sync_blockdev(bdev);
862 close_bdev_exclusive(bdev, mode);
865 EXPORT_SYMBOL(kill_block_super);
866 #endif
868 int get_sb_nodev(struct file_system_type *fs_type,
869 int flags, void *data,
870 int (*fill_super)(struct super_block *, void *, int),
871 struct vfsmount *mnt)
873 int error;
874 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
876 if (IS_ERR(s))
877 return PTR_ERR(s);
879 s->s_flags = flags;
881 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
882 if (error) {
883 deactivate_locked_super(s);
884 return error;
886 s->s_flags |= MS_ACTIVE;
887 simple_set_mnt(mnt, s);
888 return 0;
891 EXPORT_SYMBOL(get_sb_nodev);
893 static int compare_single(struct super_block *s, void *p)
895 return 1;
898 int get_sb_single(struct file_system_type *fs_type,
899 int flags, void *data,
900 int (*fill_super)(struct super_block *, void *, int),
901 struct vfsmount *mnt)
903 struct super_block *s;
904 int error;
906 s = sget(fs_type, compare_single, set_anon_super, NULL);
907 if (IS_ERR(s))
908 return PTR_ERR(s);
909 if (!s->s_root) {
910 s->s_flags = flags;
911 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
912 if (error) {
913 deactivate_locked_super(s);
914 return error;
916 s->s_flags |= MS_ACTIVE;
918 do_remount_sb(s, flags, data, 0);
919 simple_set_mnt(mnt, s);
920 return 0;
923 EXPORT_SYMBOL(get_sb_single);
925 struct vfsmount *
926 vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
928 struct vfsmount *mnt;
929 char *secdata = NULL;
930 int error;
932 if (!type)
933 return ERR_PTR(-ENODEV);
935 error = -ENOMEM;
936 mnt = alloc_vfsmnt(name);
937 if (!mnt)
938 goto out;
940 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
941 secdata = alloc_secdata();
942 if (!secdata)
943 goto out_mnt;
945 error = security_sb_copy_data(data, secdata);
946 if (error)
947 goto out_free_secdata;
950 error = type->get_sb(type, flags, name, data, mnt);
951 if (error < 0)
952 goto out_free_secdata;
953 BUG_ON(!mnt->mnt_sb);
955 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
956 if (error)
957 goto out_sb;
960 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
961 * but s_maxbytes was an unsigned long long for many releases. Throw
962 * this warning for a little while to try and catch filesystems that
963 * violate this rule. This warning should be either removed or
964 * converted to a BUG() in 2.6.34.
966 WARN((mnt->mnt_sb->s_maxbytes < 0), "%s set sb->s_maxbytes to "
967 "negative value (%lld)\n", type->name, mnt->mnt_sb->s_maxbytes);
969 mnt->mnt_mountpoint = mnt->mnt_root;
970 mnt->mnt_parent = mnt;
971 up_write(&mnt->mnt_sb->s_umount);
972 free_secdata(secdata);
973 return mnt;
974 out_sb:
975 dput(mnt->mnt_root);
976 deactivate_locked_super(mnt->mnt_sb);
977 out_free_secdata:
978 free_secdata(secdata);
979 out_mnt:
980 free_vfsmnt(mnt);
981 out:
982 return ERR_PTR(error);
985 EXPORT_SYMBOL_GPL(vfs_kern_mount);
987 static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
989 int err;
990 const char *subtype = strchr(fstype, '.');
991 if (subtype) {
992 subtype++;
993 err = -EINVAL;
994 if (!subtype[0])
995 goto err;
996 } else
997 subtype = "";
999 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
1000 err = -ENOMEM;
1001 if (!mnt->mnt_sb->s_subtype)
1002 goto err;
1003 return mnt;
1005 err:
1006 mntput(mnt);
1007 return ERR_PTR(err);
1010 struct vfsmount *
1011 do_kern_mount(const char *fstype, int flags, const char *name, void *data)
1013 struct file_system_type *type = get_fs_type(fstype);
1014 struct vfsmount *mnt;
1015 if (!type)
1016 return ERR_PTR(-ENODEV);
1017 mnt = vfs_kern_mount(type, flags, name, data);
1018 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
1019 !mnt->mnt_sb->s_subtype)
1020 mnt = fs_set_subtype(mnt, fstype);
1021 put_filesystem(type);
1022 return mnt;
1024 EXPORT_SYMBOL_GPL(do_kern_mount);
1026 struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
1028 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
1031 EXPORT_SYMBOL_GPL(kern_mount_data);