blob: dd23bf927fbc2e52c96bf1111ece7242c0eb3c6a [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#include <asm/uaccess.h>
Al Viro6d59e7f2008-03-22 15:48:17 -040042#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
44
Linus Torvalds1da177e2005-04-16 15:20:36 -070045LIST_HEAD(super_blocks);
46DEFINE_SPINLOCK(sb_lock);
47
48/**
49 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -070050 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -070051 *
52 * Allocates and initializes a new &struct super_block. alloc_super()
53 * returns a pointer new superblock or %NULL if allocation had failed.
54 */
Ingo Molnarcf516242006-07-03 00:25:27 -070055static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -070056{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -080057 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Linus Torvalds1da177e2005-04-16 15:20:36 -070058 static struct super_operations default_op;
59
60 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 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);
Ken Chen0e0f4fc2007-10-16 23:30:38 -070068 INIT_LIST_HEAD(&s->s_more_io);
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 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);
Kentaro Makitada3bbdd2008-07-23 21:27:13 -070073 INIT_LIST_HEAD(&s->s_dentry_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f2006-01-09 15:59:25 -080075 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -070076 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -070077 /*
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);
Linus Torvalds1da177e2005-04-16 15:20:36 -070083 down_write(&s->s_umount);
84 s->s_count = S_BIAS;
85 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -080086 mutex_init(&s->s_vfs_rename_mutex);
Ingo Molnard3be9152006-03-23 03:00:29 -080087 mutex_init(&s->s_dquot.dqio_mutex);
88 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089 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);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700110 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800111 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 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 */
Adrian Bunk6b09ae62008-04-29 00:58:54 -0700121static int __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122{
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);
Jan Kara0ff5af82008-04-28 02:14:33 -0700183 DQUOT_OFF(s, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 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 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700204static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205{
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
David Howellscf9a2ae2006-08-29 19:05:54 +0100225/*
Al Viro914e2632006-10-18 13:55:46 -0400226 * 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
243/*
David Howellscf9a2ae2006-08-29 19:05:54 +0100244 * 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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274/**
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.
David Howellsc636ebd2006-10-11 01:22:19 -0700283 *
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.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 */
288void generic_shutdown_super(struct super_block *sb)
289{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800290 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291
David Howellsc636ebd2006-10-11 01:22:19 -0700292 if (sb->s_root) {
293 shrink_dcache_for_umount(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 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)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800308 printk("VFS: Busy inodes after unmount of %s. "
309 "Self-destruct in 5 seconds. Have a nice day...\n",
310 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 }
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;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700339 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 int err;
341
342retry:
343 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700344 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 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 }
355 if (!s) {
356 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700357 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 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{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700404 struct super_block *sb;
405
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700407restart:
408 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 if (sb->s_dirt) {
410 sb->s_count++;
411 spin_unlock(&sb_lock);
412 down_read(&sb->s_umount);
413 write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700414 up_read(&sb->s_umount);
415 spin_lock(&sb_lock);
416 if (__put_super_and_need_restart(sb))
417 goto restart;
418 }
419 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 spin_unlock(&sb_lock);
421}
422
423/*
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200424 * Call the ->sync_fs super_op against all filesystems which are r/w and
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 * 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 *
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200432 * But if process A is currently running sync_filesystems and then process B
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 * 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;
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800442 static DEFINE_MUTEX(mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800444 mutex_lock(&mutex); /* Could be down_interruptible */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700446 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 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 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453
454restart:
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_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);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700466 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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 }
472 spin_unlock(&sb_lock);
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800473 mutex_unlock(&mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474}
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{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700486 struct super_block *sb;
487
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 if (!bdev)
489 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700490
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700492rescan:
493 list_for_each_entry(sb, &super_blocks, s_list) {
494 if (sb->s_bdev == bdev) {
495 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700497 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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 }
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{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700515 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700518rescan:
519 list_for_each_entry(sb, &super_blocks, s_list) {
520 if (sb->s_dev == dev) {
521 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700523 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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 }
532 }
533 spin_unlock(&sb_lock);
534 return NULL;
535}
536
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537asmlinkage 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;
David Howells726c3342006-06-23 02:02:58 -0700547 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 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/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700562 * mark_files_ro - mark all files read-only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 * @sb: superblock in question
564 *
Randy Dunlapa6b91912008-03-19 17:01:00 -0700565 * All files are marked read-only. We don't care about pending
566 * delete files so this should be used in 'force' mode only.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 */
568
569static void mark_files_ro(struct super_block *sb)
570{
571 struct file *f;
572
Dave Hansen49e0d022008-02-15 14:37:32 -0800573retry:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800575 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
Dave Hansen49e0d022008-02-15 14:37:32 -0800576 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;
Dave Hansenad775f52008-02-15 14:38:01 -0800584 if (file_check_writeable(f) != 0)
585 continue;
586 file_release_write(f);
Dave Hansen49e0d022008-02-15 14:37:32 -0800587 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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 }
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;
Jan Kara0ff5af82008-04-28 02:14:33 -0700612 int remount_rw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
David Howells93614012006-09-30 20:45:40 +0200614#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
616 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200617#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 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;
Jan Kara0ff5af82008-04-28 02:14:33 -0700630 retval = DQUOT_OFF(sb, 1);
631 if (retval < 0 && retval != -ENOSYS)
632 return -EBUSY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 }
Jan Kara0ff5af82008-04-28 02:14:33 -0700634 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635
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);
Jan Kara0ff5af82008-04-28 02:14:33 -0700644 if (remount_rw)
645 DQUOT_ON_REMOUNT(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 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
Alexey Dobriyan6de24f02008-08-28 06:25:49 +0400685static DEFINE_IDR(unnamed_dev_idr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686static 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:
694 if (idr_pre_get(&unnamed_dev_idr, GFP_ATOMIC) == 0)
695 return -ENOMEM;
696 spin_lock(&unnamed_dev_lock);
697 error = idr_get_new(&unnamed_dev_idr, NULL, &dev);
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);
707 idr_remove(&unnamed_dev_idr, dev);
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);
723 idr_remove(&unnamed_dev_idr, slot);
724 spin_unlock(&unnamed_dev_lock);
725}
726
727EXPORT_SYMBOL(kill_anon_super);
728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729void 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
David Howells93614012006-09-30 20:45:40 +0200738#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739static 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
David Howells454e2392006-06-23 02:02:57 -0700751int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700753 int (*fill_super)(struct super_block *, void *, int),
754 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755{
756 struct block_device *bdev;
757 struct super_block *s;
758 int error = 0;
759
760 bdev = open_bdev_excl(dev_name, flags, fs_type);
761 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700762 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763
764 /*
765 * once the super is inserted into the list by sget, s_umount
766 * will protect the lockfs code from trying to start a snapshot
767 * while we are mounting
768 */
David Chinnerf73ca1b2007-01-10 23:15:41 -0800769 down(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
David Chinnerf73ca1b2007-01-10 23:15:41 -0800771 up(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700773 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774
775 if (s->s_root) {
776 if ((flags ^ s->s_flags) & MS_RDONLY) {
777 up_write(&s->s_umount);
778 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700779 error = -EBUSY;
780 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 }
David Howells454e2392006-06-23 02:02:57 -0700782
783 close_bdev_excl(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 } else {
785 char b[BDEVNAME_SIZE];
786
787 s->s_flags = flags;
788 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800789 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800790 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 if (error) {
792 up_write(&s->s_umount);
793 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700794 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800795 }
David Howells454e2392006-06-23 02:02:57 -0700796
797 s->s_flags |= MS_ACTIVE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 }
799
David Howells454e2392006-06-23 02:02:57 -0700800 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801
David Howells454e2392006-06-23 02:02:57 -0700802error_s:
803 error = PTR_ERR(s);
804error_bdev:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 close_bdev_excl(bdev);
David Howells454e2392006-06-23 02:02:57 -0700806error:
807 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808}
809
810EXPORT_SYMBOL(get_sb_bdev);
811
812void kill_block_super(struct super_block *sb)
813{
814 struct block_device *bdev = sb->s_bdev;
815
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 generic_shutdown_super(sb);
817 sync_blockdev(bdev);
818 close_bdev_excl(bdev);
819}
820
821EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200822#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823
David Howells454e2392006-06-23 02:02:57 -0700824int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700826 int (*fill_super)(struct super_block *, void *, int),
827 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828{
829 int error;
830 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
831
832 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700833 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834
835 s->s_flags = flags;
836
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800837 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 if (error) {
839 up_write(&s->s_umount);
840 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700841 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 }
843 s->s_flags |= MS_ACTIVE;
David Howells454e2392006-06-23 02:02:57 -0700844 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845}
846
847EXPORT_SYMBOL(get_sb_nodev);
848
849static int compare_single(struct super_block *s, void *p)
850{
851 return 1;
852}
853
David Howells454e2392006-06-23 02:02:57 -0700854int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700856 int (*fill_super)(struct super_block *, void *, int),
857 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858{
859 struct super_block *s;
860 int error;
861
862 s = sget(fs_type, compare_single, set_anon_super, NULL);
863 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700864 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 if (!s->s_root) {
866 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800867 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 if (error) {
869 up_write(&s->s_umount);
870 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700871 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 }
873 s->s_flags |= MS_ACTIVE;
874 }
875 do_remount_sb(s, flags, data, 0);
David Howells454e2392006-06-23 02:02:57 -0700876 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877}
878
879EXPORT_SYMBOL(get_sb_single);
880
881struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400882vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700886 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887
888 if (!type)
889 return ERR_PTR(-ENODEV);
890
David Howells454e2392006-06-23 02:02:57 -0700891 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 mnt = alloc_vfsmnt(name);
893 if (!mnt)
894 goto out;
895
Eric Parise0007522008-03-05 10:31:54 -0500896 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700898 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900
Eric Parise0007522008-03-05 10:31:54 -0500901 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700902 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 }
905
David Howells454e2392006-06-23 02:02:57 -0700906 error = type->get_sb(type, flags, name, data, mnt);
907 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 goto out_free_secdata;
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700909 BUG_ON(!mnt->mnt_sb);
David Howells454e2392006-06-23 02:02:57 -0700910
911 error = security_sb_kern_mount(mnt->mnt_sb, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 if (error)
913 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700914
915 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700917 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700918 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 return mnt;
920out_sb:
David Howells454e2392006-06-23 02:02:57 -0700921 dput(mnt->mnt_root);
922 up_write(&mnt->mnt_sb->s_umount);
923 deactivate_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924out_free_secdata:
925 free_secdata(secdata);
926out_mnt:
927 free_vfsmnt(mnt);
928out:
David Howells454e2392006-06-23 02:02:57 -0700929 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930}
931
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400932EXPORT_SYMBOL_GPL(vfs_kern_mount);
933
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700934static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
935{
936 int err;
937 const char *subtype = strchr(fstype, '.');
938 if (subtype) {
939 subtype++;
940 err = -EINVAL;
941 if (!subtype[0])
942 goto err;
943 } else
944 subtype = "";
945
946 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
947 err = -ENOMEM;
948 if (!mnt->mnt_sb->s_subtype)
949 goto err;
950 return mnt;
951
952 err:
953 mntput(mnt);
954 return ERR_PTR(err);
955}
956
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400957struct vfsmount *
958do_kern_mount(const char *fstype, int flags, const char *name, void *data)
959{
960 struct file_system_type *type = get_fs_type(fstype);
961 struct vfsmount *mnt;
962 if (!type)
963 return ERR_PTR(-ENODEV);
964 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700965 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
966 !mnt->mnt_sb->s_subtype)
967 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400968 put_filesystem(type);
969 return mnt;
970}
Al Viro8a4e98d2008-02-24 01:43:03 -0500971EXPORT_SYMBOL_GPL(do_kern_mount);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400972
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700973struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700975 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976}
977
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700978EXPORT_SYMBOL_GPL(kern_mount_data);