]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/super.c
ext4: Fix s_dirty_blocks_counter if block allocation failed with nodelalloc
[net-next-2.6.git] / fs / super.c
CommitLineData
1da177e4
LT
1/*
2 * linux/fs/super.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
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
12 *
13 * GK 2/5/95 - Changed to support mounting the root fs via NFS
14 *
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:
96de0e25 18 * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996.
1da177e4
LT
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
21 */
22
1da177e4
LT
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>
353ab6e9 39#include <linux/mutex.h>
49e0d02c 40#include <linux/file.h>
1da177e4 41#include <asm/uaccess.h>
6d59e7f5 42#include "internal.h"
1da177e4
LT
43
44
1da177e4
LT
45LIST_HEAD(super_blocks);
46DEFINE_SPINLOCK(sb_lock);
47
48/**
49 * alloc_super - create new superblock
fe2bbc48 50 * @type: filesystem type superblock should belong to
1da177e4
LT
51 *
52 * Allocates and initializes a new &struct super_block. alloc_super()
53 * returns a pointer new superblock or %NULL if allocation had failed.
54 */
cf516249 55static struct super_block *alloc_super(struct file_system_type *type)
1da177e4 56{
11b0b5ab 57 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
1da177e4
LT
58 static struct super_operations default_op;
59
60 if (s) {
1da177e4
LT
61 if (security_sb_alloc(s)) {
62 kfree(s);
63 s = NULL;
64 goto out;
65 }
66 INIT_LIST_HEAD(&s->s_dirty);
67 INIT_LIST_HEAD(&s->s_io);
0e0f4fc2 68 INIT_LIST_HEAD(&s->s_more_io);
1da177e4
LT
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);
da3bbdd4 73 INIT_LIST_HEAD(&s->s_dentry_lru);
1da177e4 74 init_rwsem(&s->s_umount);
7892f2f4 75 mutex_init(&s->s_lock);
897c6ff9 76 lockdep_set_class(&s->s_umount, &type->s_umount_key);
cf516249
IM
77 /*
78 * The locking rules for s_lock are up to the
79 * filesystem. For example ext3fs has different
80 * lock ordering than usbfs:
81 */
82 lockdep_set_class(&s->s_lock, &type->s_lock_key);
1da177e4
LT
83 down_write(&s->s_umount);
84 s->s_count = S_BIAS;
85 atomic_set(&s->s_active, 1);
a11f3a05 86 mutex_init(&s->s_vfs_rename_mutex);
d3be915f
IM
87 mutex_init(&s->s_dquot.dqio_mutex);
88 mutex_init(&s->s_dquot.dqonoff_mutex);
1da177e4
LT
89 init_rwsem(&s->s_dquot.dqptr_sem);
90 init_waitqueue_head(&s->s_wait_unfrozen);
91 s->s_maxbytes = MAX_NON_LFS;
92 s->dq_op = sb_dquot_ops;
93 s->s_qcop = sb_quotactl_ops;
94 s->s_op = &default_op;
95 s->s_time_gran = 1000000000;
96 }
97out:
98 return s;
99}
100
101/**
102 * destroy_super - frees a superblock
103 * @s: superblock to free
104 *
105 * Frees a superblock.
106 */
107static inline void destroy_super(struct super_block *s)
108{
109 security_sb_free(s);
79c0b2df 110 kfree(s->s_subtype);
b3b304a2 111 kfree(s->s_options);
1da177e4
LT
112 kfree(s);
113}
114
115/* Superblock refcounting */
116
117/*
118 * Drop a superblock's refcount. Returns non-zero if the superblock was
119 * destroyed. The caller must hold sb_lock.
120 */
6b09ae66 121static int __put_super(struct super_block *sb)
1da177e4
LT
122{
123 int ret = 0;
124
125 if (!--sb->s_count) {
126 destroy_super(sb);
127 ret = 1;
128 }
129 return ret;
130}
131
132/*
133 * Drop a superblock's refcount.
134 * Returns non-zero if the superblock is about to be destroyed and
135 * at least is already removed from super_blocks list, so if we are
136 * making a loop through super blocks then we need to restart.
137 * The caller must hold sb_lock.
138 */
139int __put_super_and_need_restart(struct super_block *sb)
140{
141 /* check for race with generic_shutdown_super() */
142 if (list_empty(&sb->s_list)) {
143 /* super block is removed, need to restart... */
144 __put_super(sb);
145 return 1;
146 }
147 /* can't be the last, since s_list is still in use */
148 sb->s_count--;
149 BUG_ON(sb->s_count == 0);
150 return 0;
151}
152
153/**
154 * put_super - drop a temporary reference to superblock
155 * @sb: superblock in question
156 *
157 * Drops a temporary reference, frees superblock if there's no
158 * references left.
159 */
160static void put_super(struct super_block *sb)
161{
162 spin_lock(&sb_lock);
163 __put_super(sb);
164 spin_unlock(&sb_lock);
165}
166
167
168/**
169 * deactivate_super - drop an active reference to superblock
170 * @s: superblock to deactivate
171 *
172 * Drops an active reference to superblock, acquiring a temprory one if
173 * there is no active references left. In that case we lock superblock,
174 * tell fs driver to shut it down and drop the temporary reference we
175 * had just acquired.
176 */
177void deactivate_super(struct super_block *s)
178{
179 struct file_system_type *fs = s->s_type;
180 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
181 s->s_count -= S_BIAS-1;
182 spin_unlock(&sb_lock);
0ff5af83 183 DQUOT_OFF(s, 0);
1da177e4
LT
184 down_write(&s->s_umount);
185 fs->kill_sb(s);
186 put_filesystem(fs);
187 put_super(s);
188 }
189}
190
191EXPORT_SYMBOL(deactivate_super);
192
193/**
194 * grab_super - acquire an active reference
195 * @s: reference we are trying to make active
196 *
197 * Tries to acquire an active reference. grab_super() is used when we
198 * had just found a superblock in super_blocks or fs_type->fs_supers
199 * and want to turn it into a full-blown active reference. grab_super()
200 * is called with sb_lock held and drops it. Returns 1 in case of
201 * success, 0 if we had failed (superblock contents was already dead or
202 * dying when grab_super() had been called).
203 */
9c4dbee7 204static int grab_super(struct super_block *s) __releases(sb_lock)
1da177e4
LT
205{
206 s->s_count++;
207 spin_unlock(&sb_lock);
208 down_write(&s->s_umount);
209 if (s->s_root) {
210 spin_lock(&sb_lock);
211 if (s->s_count > S_BIAS) {
212 atomic_inc(&s->s_active);
213 s->s_count--;
214 spin_unlock(&sb_lock);
215 return 1;
216 }
217 spin_unlock(&sb_lock);
218 }
219 up_write(&s->s_umount);
220 put_super(s);
221 yield();
222 return 0;
223}
224
914e2637
AV
225/*
226 * Superblock locking. We really ought to get rid of these two.
227 */
228void lock_super(struct super_block * sb)
229{
230 get_fs_excl();
231 mutex_lock(&sb->s_lock);
232}
233
234void unlock_super(struct super_block * sb)
235{
236 put_fs_excl();
237 mutex_unlock(&sb->s_lock);
238}
239
240EXPORT_SYMBOL(lock_super);
241EXPORT_SYMBOL(unlock_super);
242
cf9a2ae8
DH
243/*
244 * Write out and wait upon all dirty data associated with this
245 * superblock. Filesystem data as well as the underlying block
246 * device. Takes the superblock lock. Requires a second blkdev
247 * flush by the caller to complete the operation.
248 */
249void __fsync_super(struct super_block *sb)
250{
251 sync_inodes_sb(sb, 0);
252 DQUOT_SYNC(sb);
253 lock_super(sb);
254 if (sb->s_dirt && sb->s_op->write_super)
255 sb->s_op->write_super(sb);
256 unlock_super(sb);
257 if (sb->s_op->sync_fs)
258 sb->s_op->sync_fs(sb, 1);
259 sync_blockdev(sb->s_bdev);
260 sync_inodes_sb(sb, 1);
261}
262
263/*
264 * Write out and wait upon all dirty data associated with this
265 * superblock. Filesystem data as well as the underlying block
266 * device. Takes the superblock lock.
267 */
268int fsync_super(struct super_block *sb)
269{
270 __fsync_super(sb);
271 return sync_blockdev(sb->s_bdev);
272}
273
1da177e4
LT
274/**
275 * generic_shutdown_super - common helper for ->kill_sb()
276 * @sb: superblock to kill
277 *
278 * generic_shutdown_super() does all fs-independent work on superblock
279 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
280 * that need destruction out of superblock, call generic_shutdown_super()
281 * and release aforementioned objects. Note: dentries and inodes _are_
282 * taken care of and do not need specific handling.
c636ebdb
DH
283 *
284 * Upon calling this function, the filesystem may no longer alter or
285 * rearrange the set of dentries belonging to this super_block, nor may it
286 * change the attachments of dentries to inodes.
1da177e4
LT
287 */
288void generic_shutdown_super(struct super_block *sb)
289{
ee9b6d61 290 const struct super_operations *sop = sb->s_op;
1da177e4 291
c636ebdb
DH
292 if (sb->s_root) {
293 shrink_dcache_for_umount(sb);
1da177e4
LT
294 fsync_super(sb);
295 lock_super(sb);
296 sb->s_flags &= ~MS_ACTIVE;
297 /* bad name - it should be evict_inodes() */
298 invalidate_inodes(sb);
299 lock_kernel();
300
301 if (sop->write_super && sb->s_dirt)
302 sop->write_super(sb);
303 if (sop->put_super)
304 sop->put_super(sb);
305
306 /* Forget any remaining inodes */
307 if (invalidate_inodes(sb)) {
7b4fe29e
DJ
308 printk("VFS: Busy inodes after unmount of %s. "
309 "Self-destruct in 5 seconds. Have a nice day...\n",
310 sb->s_id);
1da177e4
LT
311 }
312
313 unlock_kernel();
314 unlock_super(sb);
315 }
316 spin_lock(&sb_lock);
317 /* should be initialized for __put_super_and_need_restart() */
318 list_del_init(&sb->s_list);
319 list_del(&sb->s_instances);
320 spin_unlock(&sb_lock);
321 up_write(&sb->s_umount);
322}
323
324EXPORT_SYMBOL(generic_shutdown_super);
325
326/**
327 * sget - find or create a superblock
328 * @type: filesystem type superblock should belong to
329 * @test: comparison callback
330 * @set: setup callback
331 * @data: argument to each of them
332 */
333struct super_block *sget(struct file_system_type *type,
334 int (*test)(struct super_block *,void *),
335 int (*set)(struct super_block *,void *),
336 void *data)
337{
338 struct super_block *s = NULL;
d4730127 339 struct super_block *old;
1da177e4
LT
340 int err;
341
342retry:
343 spin_lock(&sb_lock);
d4730127
MK
344 if (test) {
345 list_for_each_entry(old, &type->fs_supers, s_instances) {
346 if (!test(old, data))
347 continue;
348 if (!grab_super(old))
349 goto retry;
350 if (s)
351 destroy_super(s);
352 return old;
353 }
1da177e4
LT
354 }
355 if (!s) {
356 spin_unlock(&sb_lock);
cf516249 357 s = alloc_super(type);
1da177e4
LT
358 if (!s)
359 return ERR_PTR(-ENOMEM);
360 goto retry;
361 }
362
363 err = set(s, data);
364 if (err) {
365 spin_unlock(&sb_lock);
366 destroy_super(s);
367 return ERR_PTR(err);
368 }
369 s->s_type = type;
370 strlcpy(s->s_id, type->name, sizeof(s->s_id));
371 list_add_tail(&s->s_list, &super_blocks);
372 list_add(&s->s_instances, &type->fs_supers);
373 spin_unlock(&sb_lock);
374 get_filesystem(type);
375 return s;
376}
377
378EXPORT_SYMBOL(sget);
379
380void drop_super(struct super_block *sb)
381{
382 up_read(&sb->s_umount);
383 put_super(sb);
384}
385
386EXPORT_SYMBOL(drop_super);
387
388static inline void write_super(struct super_block *sb)
389{
390 lock_super(sb);
391 if (sb->s_root && sb->s_dirt)
392 if (sb->s_op->write_super)
393 sb->s_op->write_super(sb);
394 unlock_super(sb);
395}
396
397/*
398 * Note: check the dirty flag before waiting, so we don't
399 * hold up the sync while mounting a device. (The newly
400 * mounted device won't need syncing.)
401 */
402void sync_supers(void)
403{
618f0636
KK
404 struct super_block *sb;
405
1da177e4 406 spin_lock(&sb_lock);
618f0636
KK
407restart:
408 list_for_each_entry(sb, &super_blocks, s_list) {
1da177e4
LT
409 if (sb->s_dirt) {
410 sb->s_count++;
411 spin_unlock(&sb_lock);
412 down_read(&sb->s_umount);
413 write_super(sb);
618f0636
KK
414 up_read(&sb->s_umount);
415 spin_lock(&sb_lock);
416 if (__put_super_and_need_restart(sb))
417 goto restart;
418 }
419 }
1da177e4
LT
420 spin_unlock(&sb_lock);
421}
422
423/*
3a4fa0a2 424 * Call the ->sync_fs super_op against all filesystems which are r/w and
1da177e4
LT
425 * which implement it.
426 *
427 * This operation is careful to avoid the livelock which could easily happen
428 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
429 * is used only here. We set it against all filesystems and then clear it as
430 * we sync them. So redirtied filesystems are skipped.
431 *
3a4fa0a2 432 * But if process A is currently running sync_filesystems and then process B
1da177e4
LT
433 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
434 * flags again, which will cause process A to resync everything. Fix that with
435 * a local mutex.
436 *
437 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
438 */
439void sync_filesystems(int wait)
440{
441 struct super_block *sb;
353ab6e9 442 static DEFINE_MUTEX(mutex);
1da177e4 443
353ab6e9 444 mutex_lock(&mutex); /* Could be down_interruptible */
1da177e4 445 spin_lock(&sb_lock);
618f0636 446 list_for_each_entry(sb, &super_blocks, s_list) {
1da177e4
LT
447 if (!sb->s_op->sync_fs)
448 continue;
449 if (sb->s_flags & MS_RDONLY)
450 continue;
451 sb->s_need_sync_fs = 1;
452 }
1da177e4
LT
453
454restart:
618f0636 455 list_for_each_entry(sb, &super_blocks, s_list) {
1da177e4
LT
456 if (!sb->s_need_sync_fs)
457 continue;
458 sb->s_need_sync_fs = 0;
459 if (sb->s_flags & MS_RDONLY)
460 continue; /* hm. Was remounted r/o meanwhile */
461 sb->s_count++;
462 spin_unlock(&sb_lock);
463 down_read(&sb->s_umount);
464 if (sb->s_root && (wait || sb->s_dirt))
465 sb->s_op->sync_fs(sb, wait);
618f0636
KK
466 up_read(&sb->s_umount);
467 /* restart only when sb is no longer on the list */
468 spin_lock(&sb_lock);
469 if (__put_super_and_need_restart(sb))
470 goto restart;
1da177e4
LT
471 }
472 spin_unlock(&sb_lock);
353ab6e9 473 mutex_unlock(&mutex);
1da177e4
LT
474}
475
476/**
477 * get_super - get the superblock of a device
478 * @bdev: device to get the superblock for
479 *
480 * Scans the superblock list and finds the superblock of the file system
481 * mounted on the device given. %NULL is returned if no match is found.
482 */
483
484struct super_block * get_super(struct block_device *bdev)
485{
618f0636
KK
486 struct super_block *sb;
487
1da177e4
LT
488 if (!bdev)
489 return NULL;
618f0636 490
1da177e4 491 spin_lock(&sb_lock);
618f0636
KK
492rescan:
493 list_for_each_entry(sb, &super_blocks, s_list) {
494 if (sb->s_bdev == bdev) {
495 sb->s_count++;
1da177e4 496 spin_unlock(&sb_lock);
618f0636
KK
497 down_read(&sb->s_umount);
498 if (sb->s_root)
499 return sb;
500 up_read(&sb->s_umount);
501 /* restart only when sb is no longer on the list */
502 spin_lock(&sb_lock);
503 if (__put_super_and_need_restart(sb))
504 goto rescan;
1da177e4
LT
505 }
506 }
507 spin_unlock(&sb_lock);
508 return NULL;
509}
510
511EXPORT_SYMBOL(get_super);
512
513struct super_block * user_get_super(dev_t dev)
514{
618f0636 515 struct super_block *sb;
1da177e4 516
1da177e4 517 spin_lock(&sb_lock);
618f0636
KK
518rescan:
519 list_for_each_entry(sb, &super_blocks, s_list) {
520 if (sb->s_dev == dev) {
521 sb->s_count++;
1da177e4 522 spin_unlock(&sb_lock);
618f0636
KK
523 down_read(&sb->s_umount);
524 if (sb->s_root)
525 return sb;
526 up_read(&sb->s_umount);
527 /* restart only when sb is no longer on the list */
528 spin_lock(&sb_lock);
529 if (__put_super_and_need_restart(sb))
530 goto rescan;
1da177e4
LT
531 }
532 }
533 spin_unlock(&sb_lock);
534 return NULL;
535}
536
1da177e4
LT
537asmlinkage long sys_ustat(unsigned dev, struct ustat __user * ubuf)
538{
539 struct super_block *s;
540 struct ustat tmp;
541 struct kstatfs sbuf;
542 int err = -EINVAL;
543
544 s = user_get_super(new_decode_dev(dev));
545 if (s == NULL)
546 goto out;
726c3342 547 err = vfs_statfs(s->s_root, &sbuf);
1da177e4
LT
548 drop_super(s);
549 if (err)
550 goto out;
551
552 memset(&tmp,0,sizeof(struct ustat));
553 tmp.f_tfree = sbuf.f_bfree;
554 tmp.f_tinode = sbuf.f_ffree;
555
556 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
557out:
558 return err;
559}
560
561/**
a6b91919 562 * mark_files_ro - mark all files read-only
1da177e4
LT
563 * @sb: superblock in question
564 *
a6b91919
RD
565 * All files are marked read-only. We don't care about pending
566 * delete files so this should be used in 'force' mode only.
1da177e4
LT
567 */
568
569static void mark_files_ro(struct super_block *sb)
570{
571 struct file *f;
572
49e0d02c 573retry:
1da177e4 574 file_list_lock();
2f512016 575 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
49e0d02c
DH
576 struct vfsmount *mnt;
577 if (!S_ISREG(f->f_path.dentry->d_inode->i_mode))
578 continue;
579 if (!file_count(f))
580 continue;
581 if (!(f->f_mode & FMODE_WRITE))
582 continue;
583 f->f_mode &= ~FMODE_WRITE;
ad775f5a
DH
584 if (file_check_writeable(f) != 0)
585 continue;
586 file_release_write(f);
49e0d02c
DH
587 mnt = mntget(f->f_path.mnt);
588 file_list_unlock();
589 /*
590 * This can sleep, so we can't hold
591 * the file_list_lock() spinlock.
592 */
593 mnt_drop_write(mnt);
594 mntput(mnt);
595 goto retry;
1da177e4
LT
596 }
597 file_list_unlock();
598}
599
600/**
601 * do_remount_sb - asks filesystem to change mount options.
602 * @sb: superblock in question
603 * @flags: numeric part of options
604 * @data: the rest of options
605 * @force: whether or not to force the change
606 *
607 * Alters the mount options of a mounted file system.
608 */
609int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
610{
611 int retval;
0ff5af83 612 int remount_rw;
1da177e4 613
9361401e 614#ifdef CONFIG_BLOCK
1da177e4
LT
615 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
616 return -EACCES;
9361401e 617#endif
1da177e4
LT
618 if (flags & MS_RDONLY)
619 acct_auto_close(sb);
620 shrink_dcache_sb(sb);
621 fsync_super(sb);
622
623 /* If we are remounting RDONLY and current sb is read/write,
624 make sure there are no rw files opened */
625 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
626 if (force)
627 mark_files_ro(sb);
628 else if (!fs_may_remount_ro(sb))
629 return -EBUSY;
0ff5af83
JK
630 retval = DQUOT_OFF(sb, 1);
631 if (retval < 0 && retval != -ENOSYS)
632 return -EBUSY;
1da177e4 633 }
0ff5af83 634 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
1da177e4
LT
635
636 if (sb->s_op->remount_fs) {
637 lock_super(sb);
638 retval = sb->s_op->remount_fs(sb, &flags, data);
639 unlock_super(sb);
640 if (retval)
641 return retval;
642 }
643 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
0ff5af83
JK
644 if (remount_rw)
645 DQUOT_ON_REMOUNT(sb);
1da177e4
LT
646 return 0;
647}
648
649static void do_emergency_remount(unsigned long foo)
650{
651 struct super_block *sb;
652
653 spin_lock(&sb_lock);
654 list_for_each_entry(sb, &super_blocks, s_list) {
655 sb->s_count++;
656 spin_unlock(&sb_lock);
657 down_read(&sb->s_umount);
658 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
659 /*
660 * ->remount_fs needs lock_kernel().
661 *
662 * What lock protects sb->s_flags??
663 */
664 lock_kernel();
665 do_remount_sb(sb, MS_RDONLY, NULL, 1);
666 unlock_kernel();
667 }
668 drop_super(sb);
669 spin_lock(&sb_lock);
670 }
671 spin_unlock(&sb_lock);
672 printk("Emergency Remount complete\n");
673}
674
675void emergency_remount(void)
676{
677 pdflush_operation(do_emergency_remount, 0);
678}
679
680/*
681 * Unnamed block devices are dummy devices used by virtual
682 * filesystems which don't use real block-devices. -- jrs
683 */
684
ad76cbc6 685static DEFINE_IDA(unnamed_dev_ida);
1da177e4
LT
686static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
687
688int set_anon_super(struct super_block *s, void *data)
689{
690 int dev;
691 int error;
692
693 retry:
ad76cbc6 694 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
1da177e4
LT
695 return -ENOMEM;
696 spin_lock(&unnamed_dev_lock);
ad76cbc6 697 error = ida_get_new(&unnamed_dev_ida, &dev);
1da177e4
LT
698 spin_unlock(&unnamed_dev_lock);
699 if (error == -EAGAIN)
700 /* We raced and lost with another CPU. */
701 goto retry;
702 else if (error)
703 return -EAGAIN;
704
705 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
706 spin_lock(&unnamed_dev_lock);
ad76cbc6 707 ida_remove(&unnamed_dev_ida, dev);
1da177e4
LT
708 spin_unlock(&unnamed_dev_lock);
709 return -EMFILE;
710 }
711 s->s_dev = MKDEV(0, dev & MINORMASK);
712 return 0;
713}
714
715EXPORT_SYMBOL(set_anon_super);
716
717void kill_anon_super(struct super_block *sb)
718{
719 int slot = MINOR(sb->s_dev);
720
721 generic_shutdown_super(sb);
722 spin_lock(&unnamed_dev_lock);
ad76cbc6 723 ida_remove(&unnamed_dev_ida, slot);
1da177e4
LT
724 spin_unlock(&unnamed_dev_lock);
725}
726
727EXPORT_SYMBOL(kill_anon_super);
728
1da177e4
LT
729void kill_litter_super(struct super_block *sb)
730{
731 if (sb->s_root)
732 d_genocide(sb->s_root);
733 kill_anon_super(sb);
734}
735
736EXPORT_SYMBOL(kill_litter_super);
737
9361401e 738#ifdef CONFIG_BLOCK
1da177e4
LT
739static int set_bdev_super(struct super_block *s, void *data)
740{
741 s->s_bdev = data;
742 s->s_dev = s->s_bdev->bd_dev;
743 return 0;
744}
745
746static int test_bdev_super(struct super_block *s, void *data)
747{
748 return (void *)s->s_bdev == data;
749}
750
454e2398 751int get_sb_bdev(struct file_system_type *fs_type,
1da177e4 752 int flags, const char *dev_name, void *data,
454e2398
DH
753 int (*fill_super)(struct super_block *, void *, int),
754 struct vfsmount *mnt)
1da177e4
LT
755{
756 struct block_device *bdev;
757 struct super_block *s;
30c40d2c 758 fmode_t mode = FMODE_READ;
1da177e4
LT
759 int error = 0;
760
30c40d2c
AV
761 if (!(flags & MS_RDONLY))
762 mode |= FMODE_WRITE;
763
764 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
1da177e4 765 if (IS_ERR(bdev))
454e2398 766 return PTR_ERR(bdev);
1da177e4
LT
767
768 /*
769 * once the super is inserted into the list by sget, s_umount
770 * will protect the lockfs code from trying to start a snapshot
771 * while we are mounting
772 */
f73ca1b7 773 down(&bdev->bd_mount_sem);
1da177e4 774 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
f73ca1b7 775 up(&bdev->bd_mount_sem);
1da177e4 776 if (IS_ERR(s))
454e2398 777 goto error_s;
1da177e4
LT
778
779 if (s->s_root) {
780 if ((flags ^ s->s_flags) & MS_RDONLY) {
781 up_write(&s->s_umount);
782 deactivate_super(s);
454e2398
DH
783 error = -EBUSY;
784 goto error_bdev;
1da177e4 785 }
454e2398 786
30c40d2c 787 close_bdev_exclusive(bdev, mode);
1da177e4
LT
788 } else {
789 char b[BDEVNAME_SIZE];
790
791 s->s_flags = flags;
30c40d2c 792 s->s_mode = mode;
1da177e4 793 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
e78c9a00 794 sb_set_blocksize(s, block_size(bdev));
9b04c997 795 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4
LT
796 if (error) {
797 up_write(&s->s_umount);
798 deactivate_super(s);
454e2398 799 goto error;
fa675765 800 }
454e2398
DH
801
802 s->s_flags |= MS_ACTIVE;
1da177e4
LT
803 }
804
454e2398 805 return simple_set_mnt(mnt, s);
1da177e4 806
454e2398
DH
807error_s:
808 error = PTR_ERR(s);
809error_bdev:
30c40d2c 810 close_bdev_exclusive(bdev, mode);
454e2398
DH
811error:
812 return error;
1da177e4
LT
813}
814
815EXPORT_SYMBOL(get_sb_bdev);
816
817void kill_block_super(struct super_block *sb)
818{
819 struct block_device *bdev = sb->s_bdev;
30c40d2c 820 fmode_t mode = sb->s_mode;
1da177e4 821
1da177e4
LT
822 generic_shutdown_super(sb);
823 sync_blockdev(bdev);
30c40d2c 824 close_bdev_exclusive(bdev, mode);
1da177e4
LT
825}
826
827EXPORT_SYMBOL(kill_block_super);
9361401e 828#endif
1da177e4 829
454e2398 830int get_sb_nodev(struct file_system_type *fs_type,
1da177e4 831 int flags, void *data,
454e2398
DH
832 int (*fill_super)(struct super_block *, void *, int),
833 struct vfsmount *mnt)
1da177e4
LT
834{
835 int error;
836 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
837
838 if (IS_ERR(s))
454e2398 839 return PTR_ERR(s);
1da177e4
LT
840
841 s->s_flags = flags;
842
9b04c997 843 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4
LT
844 if (error) {
845 up_write(&s->s_umount);
846 deactivate_super(s);
454e2398 847 return error;
1da177e4
LT
848 }
849 s->s_flags |= MS_ACTIVE;
454e2398 850 return simple_set_mnt(mnt, s);
1da177e4
LT
851}
852
853EXPORT_SYMBOL(get_sb_nodev);
854
855static int compare_single(struct super_block *s, void *p)
856{
857 return 1;
858}
859
454e2398 860int get_sb_single(struct file_system_type *fs_type,
1da177e4 861 int flags, void *data,
454e2398
DH
862 int (*fill_super)(struct super_block *, void *, int),
863 struct vfsmount *mnt)
1da177e4
LT
864{
865 struct super_block *s;
866 int error;
867
868 s = sget(fs_type, compare_single, set_anon_super, NULL);
869 if (IS_ERR(s))
454e2398 870 return PTR_ERR(s);
1da177e4
LT
871 if (!s->s_root) {
872 s->s_flags = flags;
9b04c997 873 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4
LT
874 if (error) {
875 up_write(&s->s_umount);
876 deactivate_super(s);
454e2398 877 return error;
1da177e4
LT
878 }
879 s->s_flags |= MS_ACTIVE;
880 }
881 do_remount_sb(s, flags, data, 0);
454e2398 882 return simple_set_mnt(mnt, s);
1da177e4
LT
883}
884
885EXPORT_SYMBOL(get_sb_single);
886
887struct vfsmount *
bb4a58bf 888vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
1da177e4 889{
1da177e4 890 struct vfsmount *mnt;
1da177e4 891 char *secdata = NULL;
454e2398 892 int error;
1da177e4
LT
893
894 if (!type)
895 return ERR_PTR(-ENODEV);
1da177e4 896
454e2398 897 error = -ENOMEM;
1da177e4
LT
898 mnt = alloc_vfsmnt(name);
899 if (!mnt)
900 goto out;
901
e0007529 902 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
1da177e4 903 secdata = alloc_secdata();
454e2398 904 if (!secdata)
1da177e4 905 goto out_mnt;
1da177e4 906
e0007529 907 error = security_sb_copy_data(data, secdata);
454e2398 908 if (error)
1da177e4 909 goto out_free_secdata;
1da177e4
LT
910 }
911
454e2398
DH
912 error = type->get_sb(type, flags, name, data, mnt);
913 if (error < 0)
1da177e4 914 goto out_free_secdata;
b4c07bce 915 BUG_ON(!mnt->mnt_sb);
454e2398 916
12204e24 917 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
1da177e4
LT
918 if (error)
919 goto out_sb;
454e2398
DH
920
921 mnt->mnt_mountpoint = mnt->mnt_root;
1da177e4 922 mnt->mnt_parent = mnt;
454e2398 923 up_write(&mnt->mnt_sb->s_umount);
8680e22f 924 free_secdata(secdata);
1da177e4
LT
925 return mnt;
926out_sb:
454e2398
DH
927 dput(mnt->mnt_root);
928 up_write(&mnt->mnt_sb->s_umount);
929 deactivate_super(mnt->mnt_sb);
1da177e4
LT
930out_free_secdata:
931 free_secdata(secdata);
932out_mnt:
933 free_vfsmnt(mnt);
934out:
454e2398 935 return ERR_PTR(error);
1da177e4
LT
936}
937
bb4a58bf
TM
938EXPORT_SYMBOL_GPL(vfs_kern_mount);
939
79c0b2df
MS
940static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
941{
942 int err;
943 const char *subtype = strchr(fstype, '.');
944 if (subtype) {
945 subtype++;
946 err = -EINVAL;
947 if (!subtype[0])
948 goto err;
949 } else
950 subtype = "";
951
952 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
953 err = -ENOMEM;
954 if (!mnt->mnt_sb->s_subtype)
955 goto err;
956 return mnt;
957
958 err:
959 mntput(mnt);
960 return ERR_PTR(err);
961}
962
bb4a58bf
TM
963struct vfsmount *
964do_kern_mount(const char *fstype, int flags, const char *name, void *data)
965{
966 struct file_system_type *type = get_fs_type(fstype);
967 struct vfsmount *mnt;
968 if (!type)
969 return ERR_PTR(-ENODEV);
970 mnt = vfs_kern_mount(type, flags, name, data);
79c0b2df
MS
971 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
972 !mnt->mnt_sb->s_subtype)
973 mnt = fs_set_subtype(mnt, fstype);
bb4a58bf
TM
974 put_filesystem(type);
975 return mnt;
976}
8a4e98d9 977EXPORT_SYMBOL_GPL(do_kern_mount);
bb4a58bf 978
8bf9725c 979struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
1da177e4 980{
8bf9725c 981 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
1da177e4
LT
982}
983
8bf9725c 984EXPORT_SYMBOL_GPL(kern_mount_data);