blob: 645e5403f2a0da0b4ed97cd6b6499791c732338d [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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:
Jan Engelhardt96de0e22007-10-19 23:21:04 +020018 * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996.
Linus Torvalds1da177e2005-04-16 15:20:36 -070019 * 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
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#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>
Ingo Molnar353ab6e2006-03-26 01:37:12 -080039#include <linux/mutex.h>
Dave Hansen49e0d022008-02-15 14:37:32 -080040#include <linux/file.h>
Arjan van de Venefaee192009-01-06 07:20:54 -080041#include <linux/async.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#include <asm/uaccess.h>
Al Viro6d59e7f2008-03-22 15:48:17 -040043#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070044
45
Linus Torvalds1da177e2005-04-16 15:20:36 -070046LIST_HEAD(super_blocks);
47DEFINE_SPINLOCK(sb_lock);
48
49/**
50 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -070051 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -070052 *
53 * Allocates and initializes a new &struct super_block. alloc_super()
54 * returns a pointer new superblock or %NULL if allocation had failed.
55 */
Ingo Molnarcf516242006-07-03 00:25:27 -070056static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -070057{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -080058 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 static struct super_operations default_op;
60
61 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070062 if (security_sb_alloc(s)) {
63 kfree(s);
64 s = NULL;
65 goto out;
66 }
67 INIT_LIST_HEAD(&s->s_dirty);
68 INIT_LIST_HEAD(&s->s_io);
Ken Chen0e0f4fc2007-10-16 23:30:38 -070069 INIT_LIST_HEAD(&s->s_more_io);
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 INIT_LIST_HEAD(&s->s_files);
71 INIT_LIST_HEAD(&s->s_instances);
72 INIT_HLIST_HEAD(&s->s_anon);
73 INIT_LIST_HEAD(&s->s_inodes);
Kentaro Makitada3bbdd2008-07-23 21:27:13 -070074 INIT_LIST_HEAD(&s->s_dentry_lru);
Arjan van de Venefaee192009-01-06 07:20:54 -080075 INIT_LIST_HEAD(&s->s_async_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f42006-01-09 15:59:25 -080077 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -070078 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -070079 /*
80 * The locking rules for s_lock are up to the
81 * filesystem. For example ext3fs has different
82 * lock ordering than usbfs:
83 */
84 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -070085 down_write(&s->s_umount);
86 s->s_count = S_BIAS;
87 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -080088 mutex_init(&s->s_vfs_rename_mutex);
Ingo Molnard3be9152006-03-23 03:00:29 -080089 mutex_init(&s->s_dquot.dqio_mutex);
90 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 init_rwsem(&s->s_dquot.dqptr_sem);
92 init_waitqueue_head(&s->s_wait_unfrozen);
93 s->s_maxbytes = MAX_NON_LFS;
94 s->dq_op = sb_dquot_ops;
95 s->s_qcop = sb_quotactl_ops;
96 s->s_op = &default_op;
97 s->s_time_gran = 1000000000;
98 }
99out:
100 return s;
101}
102
103/**
104 * destroy_super - frees a superblock
105 * @s: superblock to free
106 *
107 * Frees a superblock.
108 */
109static inline void destroy_super(struct super_block *s)
110{
111 security_sb_free(s);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700112 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800113 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114 kfree(s);
115}
116
117/* Superblock refcounting */
118
119/*
120 * Drop a superblock's refcount. Returns non-zero if the superblock was
121 * destroyed. The caller must hold sb_lock.
122 */
Adrian Bunk6b09ae62008-04-29 00:58:54 -0700123static int __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124{
125 int ret = 0;
126
127 if (!--sb->s_count) {
128 destroy_super(sb);
129 ret = 1;
130 }
131 return ret;
132}
133
134/*
135 * Drop a superblock's refcount.
136 * Returns non-zero if the superblock is about to be destroyed and
137 * at least is already removed from super_blocks list, so if we are
138 * making a loop through super blocks then we need to restart.
139 * The caller must hold sb_lock.
140 */
141int __put_super_and_need_restart(struct super_block *sb)
142{
143 /* check for race with generic_shutdown_super() */
144 if (list_empty(&sb->s_list)) {
145 /* super block is removed, need to restart... */
146 __put_super(sb);
147 return 1;
148 }
149 /* can't be the last, since s_list is still in use */
150 sb->s_count--;
151 BUG_ON(sb->s_count == 0);
152 return 0;
153}
154
155/**
156 * put_super - drop a temporary reference to superblock
157 * @sb: superblock in question
158 *
159 * Drops a temporary reference, frees superblock if there's no
160 * references left.
161 */
162static void put_super(struct super_block *sb)
163{
164 spin_lock(&sb_lock);
165 __put_super(sb);
166 spin_unlock(&sb_lock);
167}
168
169
170/**
171 * deactivate_super - drop an active reference to superblock
172 * @s: superblock to deactivate
173 *
174 * Drops an active reference to superblock, acquiring a temprory one if
175 * there is no active references left. In that case we lock superblock,
176 * tell fs driver to shut it down and drop the temporary reference we
177 * had just acquired.
178 */
179void deactivate_super(struct super_block *s)
180{
181 struct file_system_type *fs = s->s_type;
182 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
183 s->s_count -= S_BIAS-1;
184 spin_unlock(&sb_lock);
Jan Kara0ff5af82008-04-28 02:14:33 -0700185 DQUOT_OFF(s, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 down_write(&s->s_umount);
187 fs->kill_sb(s);
188 put_filesystem(fs);
189 put_super(s);
190 }
191}
192
193EXPORT_SYMBOL(deactivate_super);
194
195/**
196 * grab_super - acquire an active reference
197 * @s: reference we are trying to make active
198 *
199 * Tries to acquire an active reference. grab_super() is used when we
200 * had just found a superblock in super_blocks or fs_type->fs_supers
201 * and want to turn it into a full-blown active reference. grab_super()
202 * is called with sb_lock held and drops it. Returns 1 in case of
203 * success, 0 if we had failed (superblock contents was already dead or
204 * dying when grab_super() had been called).
205 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700206static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207{
208 s->s_count++;
209 spin_unlock(&sb_lock);
210 down_write(&s->s_umount);
211 if (s->s_root) {
212 spin_lock(&sb_lock);
213 if (s->s_count > S_BIAS) {
214 atomic_inc(&s->s_active);
215 s->s_count--;
216 spin_unlock(&sb_lock);
217 return 1;
218 }
219 spin_unlock(&sb_lock);
220 }
221 up_write(&s->s_umount);
222 put_super(s);
223 yield();
224 return 0;
225}
226
David Howellscf9a2ae2006-08-29 19:05:54 +0100227/*
Al Viro914e2632006-10-18 13:55:46 -0400228 * Superblock locking. We really ought to get rid of these two.
229 */
230void lock_super(struct super_block * sb)
231{
232 get_fs_excl();
233 mutex_lock(&sb->s_lock);
234}
235
236void unlock_super(struct super_block * sb)
237{
238 put_fs_excl();
239 mutex_unlock(&sb->s_lock);
240}
241
242EXPORT_SYMBOL(lock_super);
243EXPORT_SYMBOL(unlock_super);
244
245/*
David Howellscf9a2ae2006-08-29 19:05:54 +0100246 * Write out and wait upon all dirty data associated with this
247 * superblock. Filesystem data as well as the underlying block
248 * device. Takes the superblock lock. Requires a second blkdev
249 * flush by the caller to complete the operation.
250 */
251void __fsync_super(struct super_block *sb)
252{
253 sync_inodes_sb(sb, 0);
254 DQUOT_SYNC(sb);
255 lock_super(sb);
256 if (sb->s_dirt && sb->s_op->write_super)
257 sb->s_op->write_super(sb);
258 unlock_super(sb);
259 if (sb->s_op->sync_fs)
260 sb->s_op->sync_fs(sb, 1);
261 sync_blockdev(sb->s_bdev);
262 sync_inodes_sb(sb, 1);
263}
264
265/*
266 * Write out and wait upon all dirty data associated with this
267 * superblock. Filesystem data as well as the underlying block
268 * device. Takes the superblock lock.
269 */
270int fsync_super(struct super_block *sb)
271{
272 __fsync_super(sb);
273 return sync_blockdev(sb->s_bdev);
274}
275
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276/**
277 * generic_shutdown_super - common helper for ->kill_sb()
278 * @sb: superblock to kill
279 *
280 * generic_shutdown_super() does all fs-independent work on superblock
281 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
282 * that need destruction out of superblock, call generic_shutdown_super()
283 * and release aforementioned objects. Note: dentries and inodes _are_
284 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700285 *
286 * Upon calling this function, the filesystem may no longer alter or
287 * rearrange the set of dentries belonging to this super_block, nor may it
288 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 */
290void generic_shutdown_super(struct super_block *sb)
291{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800292 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293
Arjan van de Venefaee192009-01-06 07:20:54 -0800294
David Howellsc636ebd2006-10-11 01:22:19 -0700295 if (sb->s_root) {
296 shrink_dcache_for_umount(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 fsync_super(sb);
298 lock_super(sb);
299 sb->s_flags &= ~MS_ACTIVE;
Arjan van de Venefaee192009-01-06 07:20:54 -0800300
301 /*
302 * wait for asynchronous fs operations to finish before going further
303 */
304 async_synchronize_full_special(&sb->s_async_list);
305
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 /* bad name - it should be evict_inodes() */
307 invalidate_inodes(sb);
308 lock_kernel();
309
310 if (sop->write_super && sb->s_dirt)
311 sop->write_super(sb);
312 if (sop->put_super)
313 sop->put_super(sb);
314
315 /* Forget any remaining inodes */
316 if (invalidate_inodes(sb)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800317 printk("VFS: Busy inodes after unmount of %s. "
318 "Self-destruct in 5 seconds. Have a nice day...\n",
319 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 }
321
322 unlock_kernel();
323 unlock_super(sb);
324 }
325 spin_lock(&sb_lock);
326 /* should be initialized for __put_super_and_need_restart() */
327 list_del_init(&sb->s_list);
328 list_del(&sb->s_instances);
329 spin_unlock(&sb_lock);
330 up_write(&sb->s_umount);
331}
332
333EXPORT_SYMBOL(generic_shutdown_super);
334
335/**
336 * sget - find or create a superblock
337 * @type: filesystem type superblock should belong to
338 * @test: comparison callback
339 * @set: setup callback
340 * @data: argument to each of them
341 */
342struct super_block *sget(struct file_system_type *type,
343 int (*test)(struct super_block *,void *),
344 int (*set)(struct super_block *,void *),
345 void *data)
346{
347 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700348 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 int err;
350
351retry:
352 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700353 if (test) {
354 list_for_each_entry(old, &type->fs_supers, s_instances) {
355 if (!test(old, data))
356 continue;
357 if (!grab_super(old))
358 goto retry;
359 if (s)
360 destroy_super(s);
361 return old;
362 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 }
364 if (!s) {
365 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700366 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 if (!s)
368 return ERR_PTR(-ENOMEM);
369 goto retry;
370 }
371
372 err = set(s, data);
373 if (err) {
374 spin_unlock(&sb_lock);
375 destroy_super(s);
376 return ERR_PTR(err);
377 }
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;
385}
386
387EXPORT_SYMBOL(sget);
388
389void drop_super(struct super_block *sb)
390{
391 up_read(&sb->s_umount);
392 put_super(sb);
393}
394
395EXPORT_SYMBOL(drop_super);
396
397static inline void write_super(struct super_block *sb)
398{
399 lock_super(sb);
400 if (sb->s_root && sb->s_dirt)
401 if (sb->s_op->write_super)
402 sb->s_op->write_super(sb);
403 unlock_super(sb);
404}
405
406/*
407 * Note: check the dirty flag before waiting, so we don't
408 * hold up the sync while mounting a device. (The newly
409 * mounted device won't need syncing.)
410 */
411void sync_supers(void)
412{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700413 struct super_block *sb;
414
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700416restart:
417 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 if (sb->s_dirt) {
419 sb->s_count++;
420 spin_unlock(&sb_lock);
421 down_read(&sb->s_umount);
422 write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700423 up_read(&sb->s_umount);
424 spin_lock(&sb_lock);
425 if (__put_super_and_need_restart(sb))
426 goto restart;
427 }
428 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 spin_unlock(&sb_lock);
430}
431
432/*
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200433 * Call the ->sync_fs super_op against all filesystems which are r/w and
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 * which implement it.
435 *
436 * This operation is careful to avoid the livelock which could easily happen
437 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
438 * is used only here. We set it against all filesystems and then clear it as
439 * we sync them. So redirtied filesystems are skipped.
440 *
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200441 * But if process A is currently running sync_filesystems and then process B
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
443 * flags again, which will cause process A to resync everything. Fix that with
444 * a local mutex.
445 *
446 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
447 */
448void sync_filesystems(int wait)
449{
450 struct super_block *sb;
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800451 static DEFINE_MUTEX(mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800453 mutex_lock(&mutex); /* Could be down_interruptible */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700455 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 if (!sb->s_op->sync_fs)
457 continue;
458 if (sb->s_flags & MS_RDONLY)
459 continue;
460 sb->s_need_sync_fs = 1;
461 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462
463restart:
Kirill Korotaev618f0632005-06-23 00:09:54 -0700464 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 if (!sb->s_need_sync_fs)
466 continue;
467 sb->s_need_sync_fs = 0;
468 if (sb->s_flags & MS_RDONLY)
469 continue; /* hm. Was remounted r/o meanwhile */
470 sb->s_count++;
471 spin_unlock(&sb_lock);
472 down_read(&sb->s_umount);
Dave Kleikamp96777fe2009-01-08 09:46:31 -0600473 async_synchronize_full_special(&sb->s_async_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 if (sb->s_root && (wait || sb->s_dirt))
475 sb->s_op->sync_fs(sb, wait);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700476 up_read(&sb->s_umount);
477 /* restart only when sb is no longer on the list */
478 spin_lock(&sb_lock);
479 if (__put_super_and_need_restart(sb))
480 goto restart;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 }
482 spin_unlock(&sb_lock);
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800483 mutex_unlock(&mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484}
485
486/**
487 * get_super - get the superblock of a device
488 * @bdev: device to get the superblock for
489 *
490 * Scans the superblock list and finds the superblock of the file system
491 * mounted on the device given. %NULL is returned if no match is found.
492 */
493
494struct super_block * get_super(struct block_device *bdev)
495{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700496 struct super_block *sb;
497
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 if (!bdev)
499 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700500
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700502rescan:
503 list_for_each_entry(sb, &super_blocks, s_list) {
504 if (sb->s_bdev == bdev) {
505 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700507 down_read(&sb->s_umount);
508 if (sb->s_root)
509 return sb;
510 up_read(&sb->s_umount);
511 /* restart only when sb is no longer on the list */
512 spin_lock(&sb_lock);
513 if (__put_super_and_need_restart(sb))
514 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 }
516 }
517 spin_unlock(&sb_lock);
518 return NULL;
519}
520
521EXPORT_SYMBOL(get_super);
522
523struct super_block * user_get_super(dev_t dev)
524{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700525 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700528rescan:
529 list_for_each_entry(sb, &super_blocks, s_list) {
530 if (sb->s_dev == dev) {
531 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700533 down_read(&sb->s_umount);
534 if (sb->s_root)
535 return sb;
536 up_read(&sb->s_umount);
537 /* restart only when sb is no longer on the list */
538 spin_lock(&sb_lock);
539 if (__put_super_and_need_restart(sb))
540 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 }
542 }
543 spin_unlock(&sb_lock);
544 return NULL;
545}
546
Heiko Carstens257ac262009-01-14 14:14:13 +0100547SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548{
549 struct super_block *s;
550 struct ustat tmp;
551 struct kstatfs sbuf;
552 int err = -EINVAL;
553
554 s = user_get_super(new_decode_dev(dev));
555 if (s == NULL)
556 goto out;
David Howells726c3342006-06-23 02:02:58 -0700557 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 drop_super(s);
559 if (err)
560 goto out;
561
562 memset(&tmp,0,sizeof(struct ustat));
563 tmp.f_tfree = sbuf.f_bfree;
564 tmp.f_tinode = sbuf.f_ffree;
565
566 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
567out:
568 return err;
569}
570
571/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700572 * mark_files_ro - mark all files read-only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 * @sb: superblock in question
574 *
Randy Dunlapa6b91912008-03-19 17:01:00 -0700575 * All files are marked read-only. We don't care about pending
576 * delete files so this should be used in 'force' mode only.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 */
578
579static void mark_files_ro(struct super_block *sb)
580{
581 struct file *f;
582
Dave Hansen49e0d022008-02-15 14:37:32 -0800583retry:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800585 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
Dave Hansen49e0d022008-02-15 14:37:32 -0800586 struct vfsmount *mnt;
587 if (!S_ISREG(f->f_path.dentry->d_inode->i_mode))
588 continue;
589 if (!file_count(f))
590 continue;
591 if (!(f->f_mode & FMODE_WRITE))
592 continue;
593 f->f_mode &= ~FMODE_WRITE;
Dave Hansenad775f52008-02-15 14:38:01 -0800594 if (file_check_writeable(f) != 0)
595 continue;
596 file_release_write(f);
Dave Hansen49e0d022008-02-15 14:37:32 -0800597 mnt = mntget(f->f_path.mnt);
598 file_list_unlock();
599 /*
600 * This can sleep, so we can't hold
601 * the file_list_lock() spinlock.
602 */
603 mnt_drop_write(mnt);
604 mntput(mnt);
605 goto retry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 }
607 file_list_unlock();
608}
609
610/**
611 * do_remount_sb - asks filesystem to change mount options.
612 * @sb: superblock in question
613 * @flags: numeric part of options
614 * @data: the rest of options
615 * @force: whether or not to force the change
616 *
617 * Alters the mount options of a mounted file system.
618 */
619int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
620{
621 int retval;
Jan Kara0ff5af82008-04-28 02:14:33 -0700622 int remount_rw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623
David Howells93614012006-09-30 20:45:40 +0200624#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
626 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200627#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 if (flags & MS_RDONLY)
629 acct_auto_close(sb);
630 shrink_dcache_sb(sb);
631 fsync_super(sb);
632
633 /* If we are remounting RDONLY and current sb is read/write,
634 make sure there are no rw files opened */
635 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
636 if (force)
637 mark_files_ro(sb);
638 else if (!fs_may_remount_ro(sb))
639 return -EBUSY;
Jan Kara0ff5af82008-04-28 02:14:33 -0700640 retval = DQUOT_OFF(sb, 1);
641 if (retval < 0 && retval != -ENOSYS)
642 return -EBUSY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 }
Jan Kara0ff5af82008-04-28 02:14:33 -0700644 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645
646 if (sb->s_op->remount_fs) {
647 lock_super(sb);
648 retval = sb->s_op->remount_fs(sb, &flags, data);
649 unlock_super(sb);
650 if (retval)
651 return retval;
652 }
653 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Jan Kara0ff5af82008-04-28 02:14:33 -0700654 if (remount_rw)
655 DQUOT_ON_REMOUNT(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 return 0;
657}
658
659static void do_emergency_remount(unsigned long foo)
660{
661 struct super_block *sb;
662
663 spin_lock(&sb_lock);
664 list_for_each_entry(sb, &super_blocks, s_list) {
665 sb->s_count++;
666 spin_unlock(&sb_lock);
667 down_read(&sb->s_umount);
668 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
669 /*
670 * ->remount_fs needs lock_kernel().
671 *
672 * What lock protects sb->s_flags??
673 */
674 lock_kernel();
675 do_remount_sb(sb, MS_RDONLY, NULL, 1);
676 unlock_kernel();
677 }
678 drop_super(sb);
679 spin_lock(&sb_lock);
680 }
681 spin_unlock(&sb_lock);
682 printk("Emergency Remount complete\n");
683}
684
685void emergency_remount(void)
686{
687 pdflush_operation(do_emergency_remount, 0);
688}
689
690/*
691 * Unnamed block devices are dummy devices used by virtual
692 * filesystems which don't use real block-devices. -- jrs
693 */
694
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400695static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
697
698int set_anon_super(struct super_block *s, void *data)
699{
700 int dev;
701 int error;
702
703 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400704 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 return -ENOMEM;
706 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400707 error = ida_get_new(&unnamed_dev_ida, &dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 spin_unlock(&unnamed_dev_lock);
709 if (error == -EAGAIN)
710 /* We raced and lost with another CPU. */
711 goto retry;
712 else if (error)
713 return -EAGAIN;
714
715 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
716 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400717 ida_remove(&unnamed_dev_ida, dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 spin_unlock(&unnamed_dev_lock);
719 return -EMFILE;
720 }
721 s->s_dev = MKDEV(0, dev & MINORMASK);
722 return 0;
723}
724
725EXPORT_SYMBOL(set_anon_super);
726
727void kill_anon_super(struct super_block *sb)
728{
729 int slot = MINOR(sb->s_dev);
730
731 generic_shutdown_super(sb);
732 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400733 ida_remove(&unnamed_dev_ida, slot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 spin_unlock(&unnamed_dev_lock);
735}
736
737EXPORT_SYMBOL(kill_anon_super);
738
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739void kill_litter_super(struct super_block *sb)
740{
741 if (sb->s_root)
742 d_genocide(sb->s_root);
743 kill_anon_super(sb);
744}
745
746EXPORT_SYMBOL(kill_litter_super);
747
David Howells93614012006-09-30 20:45:40 +0200748#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749static int set_bdev_super(struct super_block *s, void *data)
750{
751 s->s_bdev = data;
752 s->s_dev = s->s_bdev->bd_dev;
753 return 0;
754}
755
756static int test_bdev_super(struct super_block *s, void *data)
757{
758 return (void *)s->s_bdev == data;
759}
760
David Howells454e2392006-06-23 02:02:57 -0700761int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700763 int (*fill_super)(struct super_block *, void *, int),
764 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765{
766 struct block_device *bdev;
767 struct super_block *s;
Al Viro30c40d22008-02-22 19:50:45 -0500768 fmode_t mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 int error = 0;
770
Al Viro30c40d22008-02-22 19:50:45 -0500771 if (!(flags & MS_RDONLY))
772 mode |= FMODE_WRITE;
773
774 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700776 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777
778 /*
779 * once the super is inserted into the list by sget, s_umount
780 * will protect the lockfs code from trying to start a snapshot
781 * while we are mounting
782 */
David Chinnerf73ca1b2007-01-10 23:15:41 -0800783 down(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
David Chinnerf73ca1b2007-01-10 23:15:41 -0800785 up(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700787 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
789 if (s->s_root) {
790 if ((flags ^ s->s_flags) & MS_RDONLY) {
791 up_write(&s->s_umount);
792 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700793 error = -EBUSY;
794 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 }
David Howells454e2392006-06-23 02:02:57 -0700796
Al Viro30c40d22008-02-22 19:50:45 -0500797 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 } else {
799 char b[BDEVNAME_SIZE];
800
801 s->s_flags = flags;
Al Viro30c40d22008-02-22 19:50:45 -0500802 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800804 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800805 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 if (error) {
807 up_write(&s->s_umount);
808 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700809 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800810 }
David Howells454e2392006-06-23 02:02:57 -0700811
812 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -0500813 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 }
815
David Howells454e2392006-06-23 02:02:57 -0700816 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817
David Howells454e2392006-06-23 02:02:57 -0700818error_s:
819 error = PTR_ERR(s);
820error_bdev:
Al Viro30c40d22008-02-22 19:50:45 -0500821 close_bdev_exclusive(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -0700822error:
823 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824}
825
826EXPORT_SYMBOL(get_sb_bdev);
827
828void kill_block_super(struct super_block *sb)
829{
830 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -0500831 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832
Theodore Ts'o87d8fe12009-01-03 09:47:09 -0500833 bdev->bd_super = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 generic_shutdown_super(sb);
835 sync_blockdev(bdev);
Al Viro30c40d22008-02-22 19:50:45 -0500836 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837}
838
839EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200840#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
David Howells454e2392006-06-23 02:02:57 -0700842int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700844 int (*fill_super)(struct super_block *, void *, int),
845 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846{
847 int error;
848 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
849
850 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700851 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852
853 s->s_flags = flags;
854
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800855 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 if (error) {
857 up_write(&s->s_umount);
858 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700859 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 }
861 s->s_flags |= MS_ACTIVE;
David Howells454e2392006-06-23 02:02:57 -0700862 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863}
864
865EXPORT_SYMBOL(get_sb_nodev);
866
867static int compare_single(struct super_block *s, void *p)
868{
869 return 1;
870}
871
David Howells454e2392006-06-23 02:02:57 -0700872int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700874 int (*fill_super)(struct super_block *, void *, int),
875 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876{
877 struct super_block *s;
878 int error;
879
880 s = sget(fs_type, compare_single, set_anon_super, NULL);
881 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700882 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 if (!s->s_root) {
884 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800885 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 if (error) {
887 up_write(&s->s_umount);
888 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700889 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 }
891 s->s_flags |= MS_ACTIVE;
892 }
893 do_remount_sb(s, flags, data, 0);
David Howells454e2392006-06-23 02:02:57 -0700894 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895}
896
897EXPORT_SYMBOL(get_sb_single);
898
899struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400900vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700904 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905
906 if (!type)
907 return ERR_PTR(-ENODEV);
908
David Howells454e2392006-06-23 02:02:57 -0700909 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 mnt = alloc_vfsmnt(name);
911 if (!mnt)
912 goto out;
913
Eric Parise0007522008-03-05 10:31:54 -0500914 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700916 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918
Eric Parise0007522008-03-05 10:31:54 -0500919 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700920 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 }
923
David Howells454e2392006-06-23 02:02:57 -0700924 error = type->get_sb(type, flags, name, data, mnt);
925 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 goto out_free_secdata;
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700927 BUG_ON(!mnt->mnt_sb);
David Howells454e2392006-06-23 02:02:57 -0700928
James Morris12204e22008-12-19 10:44:42 +1100929 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 if (error)
931 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700932
933 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700935 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700936 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 return mnt;
938out_sb:
David Howells454e2392006-06-23 02:02:57 -0700939 dput(mnt->mnt_root);
940 up_write(&mnt->mnt_sb->s_umount);
941 deactivate_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942out_free_secdata:
943 free_secdata(secdata);
944out_mnt:
945 free_vfsmnt(mnt);
946out:
David Howells454e2392006-06-23 02:02:57 -0700947 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948}
949
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400950EXPORT_SYMBOL_GPL(vfs_kern_mount);
951
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700952static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
953{
954 int err;
955 const char *subtype = strchr(fstype, '.');
956 if (subtype) {
957 subtype++;
958 err = -EINVAL;
959 if (!subtype[0])
960 goto err;
961 } else
962 subtype = "";
963
964 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
965 err = -ENOMEM;
966 if (!mnt->mnt_sb->s_subtype)
967 goto err;
968 return mnt;
969
970 err:
971 mntput(mnt);
972 return ERR_PTR(err);
973}
974
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400975struct vfsmount *
976do_kern_mount(const char *fstype, int flags, const char *name, void *data)
977{
978 struct file_system_type *type = get_fs_type(fstype);
979 struct vfsmount *mnt;
980 if (!type)
981 return ERR_PTR(-ENODEV);
982 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700983 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
984 !mnt->mnt_sb->s_subtype)
985 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400986 put_filesystem(type);
987 return mnt;
988}
Al Viro8a4e98d2008-02-24 01:43:03 -0500989EXPORT_SYMBOL_GPL(do_kern_mount);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400990
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700991struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700993 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994}
995
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700996EXPORT_SYMBOL_GPL(kern_mount_data);