blob: 6ce501447ada090cac2f5d8ec575030070356831 [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 Molnar7892f2f2006-01-09 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);
Peter Zijlstraada723d2009-02-18 14:48:30 -080085 /*
86 * sget() can have s_umount recursion.
87 *
88 * When it cannot find a suitable sb, it allocates a new
89 * one (this one), and tries again to find a suitable old
90 * one.
91 *
92 * In case that succeeds, it will acquire the s_umount
93 * lock of the old one. Since these are clearly distrinct
94 * locks, and this object isn't exposed yet, there's no
95 * risk of deadlocks.
96 *
97 * Annotate this by putting this lock in a different
98 * subclass.
99 */
100 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 s->s_count = S_BIAS;
102 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -0800103 mutex_init(&s->s_vfs_rename_mutex);
Ingo Molnard3be9152006-03-23 03:00:29 -0800104 mutex_init(&s->s_dquot.dqio_mutex);
105 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 init_rwsem(&s->s_dquot.dqptr_sem);
107 init_waitqueue_head(&s->s_wait_unfrozen);
108 s->s_maxbytes = MAX_NON_LFS;
109 s->dq_op = sb_dquot_ops;
110 s->s_qcop = sb_quotactl_ops;
111 s->s_op = &default_op;
112 s->s_time_gran = 1000000000;
113 }
114out:
115 return s;
116}
117
118/**
119 * destroy_super - frees a superblock
120 * @s: superblock to free
121 *
122 * Frees a superblock.
123 */
124static inline void destroy_super(struct super_block *s)
125{
126 security_sb_free(s);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700127 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800128 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 kfree(s);
130}
131
132/* Superblock refcounting */
133
134/*
135 * Drop a superblock's refcount. Returns non-zero if the superblock was
136 * destroyed. The caller must hold sb_lock.
137 */
Adrian Bunk6b09ae62008-04-29 00:58:54 -0700138static int __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139{
140 int ret = 0;
141
142 if (!--sb->s_count) {
143 destroy_super(sb);
144 ret = 1;
145 }
146 return ret;
147}
148
149/*
150 * Drop a superblock's refcount.
151 * Returns non-zero if the superblock is about to be destroyed and
152 * at least is already removed from super_blocks list, so if we are
153 * making a loop through super blocks then we need to restart.
154 * The caller must hold sb_lock.
155 */
156int __put_super_and_need_restart(struct super_block *sb)
157{
158 /* check for race with generic_shutdown_super() */
159 if (list_empty(&sb->s_list)) {
160 /* super block is removed, need to restart... */
161 __put_super(sb);
162 return 1;
163 }
164 /* can't be the last, since s_list is still in use */
165 sb->s_count--;
166 BUG_ON(sb->s_count == 0);
167 return 0;
168}
169
170/**
171 * put_super - drop a temporary reference to superblock
172 * @sb: superblock in question
173 *
174 * Drops a temporary reference, frees superblock if there's no
175 * references left.
176 */
177static void put_super(struct super_block *sb)
178{
179 spin_lock(&sb_lock);
180 __put_super(sb);
181 spin_unlock(&sb_lock);
182}
183
184
185/**
186 * deactivate_super - drop an active reference to superblock
187 * @s: superblock to deactivate
188 *
189 * Drops an active reference to superblock, acquiring a temprory one if
190 * there is no active references left. In that case we lock superblock,
191 * tell fs driver to shut it down and drop the temporary reference we
192 * had just acquired.
193 */
194void deactivate_super(struct super_block *s)
195{
196 struct file_system_type *fs = s->s_type;
197 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
198 s->s_count -= S_BIAS-1;
199 spin_unlock(&sb_lock);
Jan Kara0ff5af82008-04-28 02:14:33 -0700200 DQUOT_OFF(s, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 down_write(&s->s_umount);
202 fs->kill_sb(s);
203 put_filesystem(fs);
204 put_super(s);
205 }
206}
207
208EXPORT_SYMBOL(deactivate_super);
209
210/**
211 * grab_super - acquire an active reference
212 * @s: reference we are trying to make active
213 *
214 * Tries to acquire an active reference. grab_super() is used when we
215 * had just found a superblock in super_blocks or fs_type->fs_supers
216 * and want to turn it into a full-blown active reference. grab_super()
217 * is called with sb_lock held and drops it. Returns 1 in case of
218 * success, 0 if we had failed (superblock contents was already dead or
219 * dying when grab_super() had been called).
220 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700221static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222{
223 s->s_count++;
224 spin_unlock(&sb_lock);
225 down_write(&s->s_umount);
226 if (s->s_root) {
227 spin_lock(&sb_lock);
228 if (s->s_count > S_BIAS) {
229 atomic_inc(&s->s_active);
230 s->s_count--;
231 spin_unlock(&sb_lock);
232 return 1;
233 }
234 spin_unlock(&sb_lock);
235 }
236 up_write(&s->s_umount);
237 put_super(s);
238 yield();
239 return 0;
240}
241
David Howellscf9a2ae2006-08-29 19:05:54 +0100242/*
Al Viro914e2632006-10-18 13:55:46 -0400243 * Superblock locking. We really ought to get rid of these two.
244 */
245void lock_super(struct super_block * sb)
246{
247 get_fs_excl();
248 mutex_lock(&sb->s_lock);
249}
250
251void unlock_super(struct super_block * sb)
252{
253 put_fs_excl();
254 mutex_unlock(&sb->s_lock);
255}
256
257EXPORT_SYMBOL(lock_super);
258EXPORT_SYMBOL(unlock_super);
259
260/*
David Howellscf9a2ae2006-08-29 19:05:54 +0100261 * Write out and wait upon all dirty data associated with this
262 * superblock. Filesystem data as well as the underlying block
263 * device. Takes the superblock lock. Requires a second blkdev
264 * flush by the caller to complete the operation.
265 */
266void __fsync_super(struct super_block *sb)
267{
268 sync_inodes_sb(sb, 0);
269 DQUOT_SYNC(sb);
270 lock_super(sb);
271 if (sb->s_dirt && sb->s_op->write_super)
272 sb->s_op->write_super(sb);
273 unlock_super(sb);
274 if (sb->s_op->sync_fs)
275 sb->s_op->sync_fs(sb, 1);
276 sync_blockdev(sb->s_bdev);
277 sync_inodes_sb(sb, 1);
278}
279
280/*
281 * Write out and wait upon all dirty data associated with this
282 * superblock. Filesystem data as well as the underlying block
283 * device. Takes the superblock lock.
284 */
285int fsync_super(struct super_block *sb)
286{
287 __fsync_super(sb);
288 return sync_blockdev(sb->s_bdev);
289}
290
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291/**
292 * generic_shutdown_super - common helper for ->kill_sb()
293 * @sb: superblock to kill
294 *
295 * generic_shutdown_super() does all fs-independent work on superblock
296 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
297 * that need destruction out of superblock, call generic_shutdown_super()
298 * and release aforementioned objects. Note: dentries and inodes _are_
299 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700300 *
301 * Upon calling this function, the filesystem may no longer alter or
302 * rearrange the set of dentries belonging to this super_block, nor may it
303 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 */
305void generic_shutdown_super(struct super_block *sb)
306{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800307 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Arjan van de Venefaee192009-01-06 07:20:54 -0800309
David Howellsc636ebd2006-10-11 01:22:19 -0700310 if (sb->s_root) {
311 shrink_dcache_for_umount(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 fsync_super(sb);
313 lock_super(sb);
314 sb->s_flags &= ~MS_ACTIVE;
Arjan van de Venefaee192009-01-06 07:20:54 -0800315
316 /*
317 * wait for asynchronous fs operations to finish before going further
318 */
Cornelia Huck766ccb92009-01-20 15:31:31 +0100319 async_synchronize_full_domain(&sb->s_async_list);
Arjan van de Venefaee192009-01-06 07:20:54 -0800320
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 /* bad name - it should be evict_inodes() */
322 invalidate_inodes(sb);
323 lock_kernel();
324
325 if (sop->write_super && sb->s_dirt)
326 sop->write_super(sb);
327 if (sop->put_super)
328 sop->put_super(sb);
329
330 /* Forget any remaining inodes */
331 if (invalidate_inodes(sb)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800332 printk("VFS: Busy inodes after unmount of %s. "
333 "Self-destruct in 5 seconds. Have a nice day...\n",
334 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 }
336
337 unlock_kernel();
338 unlock_super(sb);
339 }
340 spin_lock(&sb_lock);
341 /* should be initialized for __put_super_and_need_restart() */
342 list_del_init(&sb->s_list);
343 list_del(&sb->s_instances);
344 spin_unlock(&sb_lock);
345 up_write(&sb->s_umount);
346}
347
348EXPORT_SYMBOL(generic_shutdown_super);
349
350/**
351 * sget - find or create a superblock
352 * @type: filesystem type superblock should belong to
353 * @test: comparison callback
354 * @set: setup callback
355 * @data: argument to each of them
356 */
357struct super_block *sget(struct file_system_type *type,
358 int (*test)(struct super_block *,void *),
359 int (*set)(struct super_block *,void *),
360 void *data)
361{
362 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700363 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 int err;
365
366retry:
367 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700368 if (test) {
369 list_for_each_entry(old, &type->fs_supers, s_instances) {
370 if (!test(old, data))
371 continue;
372 if (!grab_super(old))
373 goto retry;
Li Zefana3cfbb52009-03-12 14:31:29 -0700374 if (s) {
375 up_write(&s->s_umount);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700376 destroy_super(s);
Li Zefana3cfbb52009-03-12 14:31:29 -0700377 }
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700378 return old;
379 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 }
381 if (!s) {
382 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700383 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 if (!s)
385 return ERR_PTR(-ENOMEM);
386 goto retry;
387 }
388
389 err = set(s, data);
390 if (err) {
391 spin_unlock(&sb_lock);
Li Zefana3cfbb52009-03-12 14:31:29 -0700392 up_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 destroy_super(s);
394 return ERR_PTR(err);
395 }
396 s->s_type = type;
397 strlcpy(s->s_id, type->name, sizeof(s->s_id));
398 list_add_tail(&s->s_list, &super_blocks);
399 list_add(&s->s_instances, &type->fs_supers);
400 spin_unlock(&sb_lock);
401 get_filesystem(type);
402 return s;
403}
404
405EXPORT_SYMBOL(sget);
406
407void drop_super(struct super_block *sb)
408{
409 up_read(&sb->s_umount);
410 put_super(sb);
411}
412
413EXPORT_SYMBOL(drop_super);
414
415static inline void write_super(struct super_block *sb)
416{
417 lock_super(sb);
418 if (sb->s_root && sb->s_dirt)
419 if (sb->s_op->write_super)
420 sb->s_op->write_super(sb);
421 unlock_super(sb);
422}
423
424/*
425 * Note: check the dirty flag before waiting, so we don't
426 * hold up the sync while mounting a device. (The newly
427 * mounted device won't need syncing.)
428 */
429void sync_supers(void)
430{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700431 struct super_block *sb;
432
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700434restart:
435 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 if (sb->s_dirt) {
437 sb->s_count++;
438 spin_unlock(&sb_lock);
439 down_read(&sb->s_umount);
440 write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700441 up_read(&sb->s_umount);
442 spin_lock(&sb_lock);
443 if (__put_super_and_need_restart(sb))
444 goto restart;
445 }
446 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 spin_unlock(&sb_lock);
448}
449
450/*
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200451 * Call the ->sync_fs super_op against all filesystems which are r/w and
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 * which implement it.
453 *
454 * This operation is careful to avoid the livelock which could easily happen
455 * if two or more filesystems are being continuously dirtied. s_need_sync_fs
456 * is used only here. We set it against all filesystems and then clear it as
457 * we sync them. So redirtied filesystems are skipped.
458 *
Robert P. J. Day3a4fa0a2007-10-19 23:10:43 +0200459 * But if process A is currently running sync_filesystems and then process B
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 * calls sync_filesystems as well, process B will set all the s_need_sync_fs
461 * flags again, which will cause process A to resync everything. Fix that with
462 * a local mutex.
463 *
464 * (Fabian) Avoid sync_fs with clean fs & wait mode 0
465 */
466void sync_filesystems(int wait)
467{
468 struct super_block *sb;
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800469 static DEFINE_MUTEX(mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800471 mutex_lock(&mutex); /* Could be down_interruptible */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700473 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 if (!sb->s_op->sync_fs)
475 continue;
476 if (sb->s_flags & MS_RDONLY)
477 continue;
478 sb->s_need_sync_fs = 1;
479 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480
481restart:
Kirill Korotaev618f0632005-06-23 00:09:54 -0700482 list_for_each_entry(sb, &super_blocks, s_list) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 if (!sb->s_need_sync_fs)
484 continue;
485 sb->s_need_sync_fs = 0;
486 if (sb->s_flags & MS_RDONLY)
487 continue; /* hm. Was remounted r/o meanwhile */
488 sb->s_count++;
489 spin_unlock(&sb_lock);
490 down_read(&sb->s_umount);
Cornelia Huck766ccb92009-01-20 15:31:31 +0100491 async_synchronize_full_domain(&sb->s_async_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 if (sb->s_root && (wait || sb->s_dirt))
493 sb->s_op->sync_fs(sb, wait);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700494 up_read(&sb->s_umount);
495 /* restart only when sb is no longer on the list */
496 spin_lock(&sb_lock);
497 if (__put_super_and_need_restart(sb))
498 goto restart;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 }
500 spin_unlock(&sb_lock);
Ingo Molnar353ab6e2006-03-26 01:37:12 -0800501 mutex_unlock(&mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502}
503
504/**
505 * get_super - get the superblock of a device
506 * @bdev: device to get the superblock for
507 *
508 * Scans the superblock list and finds the superblock of the file system
509 * mounted on the device given. %NULL is returned if no match is found.
510 */
511
512struct super_block * get_super(struct block_device *bdev)
513{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700514 struct super_block *sb;
515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 if (!bdev)
517 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700518
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700520rescan:
521 list_for_each_entry(sb, &super_blocks, s_list) {
522 if (sb->s_bdev == bdev) {
523 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700525 down_read(&sb->s_umount);
526 if (sb->s_root)
527 return sb;
528 up_read(&sb->s_umount);
529 /* restart only when sb is no longer on the list */
530 spin_lock(&sb_lock);
531 if (__put_super_and_need_restart(sb))
532 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 }
534 }
535 spin_unlock(&sb_lock);
536 return NULL;
537}
538
539EXPORT_SYMBOL(get_super);
540
541struct super_block * user_get_super(dev_t dev)
542{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700543 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700546rescan:
547 list_for_each_entry(sb, &super_blocks, s_list) {
548 if (sb->s_dev == dev) {
549 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700551 down_read(&sb->s_umount);
552 if (sb->s_root)
553 return sb;
554 up_read(&sb->s_umount);
555 /* restart only when sb is no longer on the list */
556 spin_lock(&sb_lock);
557 if (__put_super_and_need_restart(sb))
558 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 }
560 }
561 spin_unlock(&sb_lock);
562 return NULL;
563}
564
Heiko Carstens257ac262009-01-14 14:14:13 +0100565SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566{
567 struct super_block *s;
568 struct ustat tmp;
569 struct kstatfs sbuf;
570 int err = -EINVAL;
571
572 s = user_get_super(new_decode_dev(dev));
573 if (s == NULL)
574 goto out;
David Howells726c3342006-06-23 02:02:58 -0700575 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 drop_super(s);
577 if (err)
578 goto out;
579
580 memset(&tmp,0,sizeof(struct ustat));
581 tmp.f_tfree = sbuf.f_bfree;
582 tmp.f_tinode = sbuf.f_ffree;
583
584 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
585out:
586 return err;
587}
588
589/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700590 * mark_files_ro - mark all files read-only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 * @sb: superblock in question
592 *
Randy Dunlapa6b91912008-03-19 17:01:00 -0700593 * All files are marked read-only. We don't care about pending
594 * delete files so this should be used in 'force' mode only.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 */
596
597static void mark_files_ro(struct super_block *sb)
598{
599 struct file *f;
600
Dave Hansen49e0d022008-02-15 14:37:32 -0800601retry:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 file_list_lock();
Eric Dumazet2f512012005-10-30 15:02:16 -0800603 list_for_each_entry(f, &sb->s_files, f_u.fu_list) {
Dave Hansen49e0d022008-02-15 14:37:32 -0800604 struct vfsmount *mnt;
605 if (!S_ISREG(f->f_path.dentry->d_inode->i_mode))
606 continue;
607 if (!file_count(f))
608 continue;
609 if (!(f->f_mode & FMODE_WRITE))
610 continue;
611 f->f_mode &= ~FMODE_WRITE;
Dave Hansenad775f52008-02-15 14:38:01 -0800612 if (file_check_writeable(f) != 0)
613 continue;
614 file_release_write(f);
Dave Hansen49e0d022008-02-15 14:37:32 -0800615 mnt = mntget(f->f_path.mnt);
616 file_list_unlock();
617 /*
618 * This can sleep, so we can't hold
619 * the file_list_lock() spinlock.
620 */
621 mnt_drop_write(mnt);
622 mntput(mnt);
623 goto retry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 }
625 file_list_unlock();
626}
627
628/**
629 * do_remount_sb - asks filesystem to change mount options.
630 * @sb: superblock in question
631 * @flags: numeric part of options
632 * @data: the rest of options
633 * @force: whether or not to force the change
634 *
635 * Alters the mount options of a mounted file system.
636 */
637int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
638{
639 int retval;
Jan Kara0ff5af82008-04-28 02:14:33 -0700640 int remount_rw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641
David Howells93614012006-09-30 20:45:40 +0200642#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
644 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200645#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 if (flags & MS_RDONLY)
647 acct_auto_close(sb);
648 shrink_dcache_sb(sb);
649 fsync_super(sb);
650
651 /* If we are remounting RDONLY and current sb is read/write,
652 make sure there are no rw files opened */
653 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
654 if (force)
655 mark_files_ro(sb);
656 else if (!fs_may_remount_ro(sb))
657 return -EBUSY;
Jan Kara0ff5af82008-04-28 02:14:33 -0700658 retval = DQUOT_OFF(sb, 1);
659 if (retval < 0 && retval != -ENOSYS)
660 return -EBUSY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 }
Jan Kara0ff5af82008-04-28 02:14:33 -0700662 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663
664 if (sb->s_op->remount_fs) {
665 lock_super(sb);
666 retval = sb->s_op->remount_fs(sb, &flags, data);
667 unlock_super(sb);
668 if (retval)
669 return retval;
670 }
671 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Jan Kara0ff5af82008-04-28 02:14:33 -0700672 if (remount_rw)
673 DQUOT_ON_REMOUNT(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 return 0;
675}
676
677static void do_emergency_remount(unsigned long foo)
678{
679 struct super_block *sb;
680
681 spin_lock(&sb_lock);
682 list_for_each_entry(sb, &super_blocks, s_list) {
683 sb->s_count++;
684 spin_unlock(&sb_lock);
685 down_read(&sb->s_umount);
686 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
687 /*
688 * ->remount_fs needs lock_kernel().
689 *
690 * What lock protects sb->s_flags??
691 */
692 lock_kernel();
693 do_remount_sb(sb, MS_RDONLY, NULL, 1);
694 unlock_kernel();
695 }
696 drop_super(sb);
697 spin_lock(&sb_lock);
698 }
699 spin_unlock(&sb_lock);
700 printk("Emergency Remount complete\n");
701}
702
703void emergency_remount(void)
704{
705 pdflush_operation(do_emergency_remount, 0);
706}
707
708/*
709 * Unnamed block devices are dummy devices used by virtual
710 * filesystems which don't use real block-devices. -- jrs
711 */
712
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400713static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
715
716int set_anon_super(struct super_block *s, void *data)
717{
718 int dev;
719 int error;
720
721 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400722 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 return -ENOMEM;
724 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400725 error = ida_get_new(&unnamed_dev_ida, &dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 spin_unlock(&unnamed_dev_lock);
727 if (error == -EAGAIN)
728 /* We raced and lost with another CPU. */
729 goto retry;
730 else if (error)
731 return -EAGAIN;
732
733 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
734 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400735 ida_remove(&unnamed_dev_ida, dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 spin_unlock(&unnamed_dev_lock);
737 return -EMFILE;
738 }
739 s->s_dev = MKDEV(0, dev & MINORMASK);
740 return 0;
741}
742
743EXPORT_SYMBOL(set_anon_super);
744
745void kill_anon_super(struct super_block *sb)
746{
747 int slot = MINOR(sb->s_dev);
748
749 generic_shutdown_super(sb);
750 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400751 ida_remove(&unnamed_dev_ida, slot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 spin_unlock(&unnamed_dev_lock);
753}
754
755EXPORT_SYMBOL(kill_anon_super);
756
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757void kill_litter_super(struct super_block *sb)
758{
759 if (sb->s_root)
760 d_genocide(sb->s_root);
761 kill_anon_super(sb);
762}
763
764EXPORT_SYMBOL(kill_litter_super);
765
David Howells93614012006-09-30 20:45:40 +0200766#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767static int set_bdev_super(struct super_block *s, void *data)
768{
769 s->s_bdev = data;
770 s->s_dev = s->s_bdev->bd_dev;
771 return 0;
772}
773
774static int test_bdev_super(struct super_block *s, void *data)
775{
776 return (void *)s->s_bdev == data;
777}
778
David Howells454e2392006-06-23 02:02:57 -0700779int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700781 int (*fill_super)(struct super_block *, void *, int),
782 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783{
784 struct block_device *bdev;
785 struct super_block *s;
Al Viro30c40d22008-02-22 19:50:45 -0500786 fmode_t mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 int error = 0;
788
Al Viro30c40d22008-02-22 19:50:45 -0500789 if (!(flags & MS_RDONLY))
790 mode |= FMODE_WRITE;
791
792 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700794 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
796 /*
797 * once the super is inserted into the list by sget, s_umount
798 * will protect the lockfs code from trying to start a snapshot
799 * while we are mounting
800 */
David Chinnerf73ca1b2007-01-10 23:15:41 -0800801 down(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
David Chinnerf73ca1b2007-01-10 23:15:41 -0800803 up(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700805 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806
807 if (s->s_root) {
808 if ((flags ^ s->s_flags) & MS_RDONLY) {
809 up_write(&s->s_umount);
810 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700811 error = -EBUSY;
812 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 }
David Howells454e2392006-06-23 02:02:57 -0700814
Al Viro30c40d22008-02-22 19:50:45 -0500815 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 } else {
817 char b[BDEVNAME_SIZE];
818
819 s->s_flags = flags;
Al Viro30c40d22008-02-22 19:50:45 -0500820 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800822 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800823 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 if (error) {
825 up_write(&s->s_umount);
826 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700827 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800828 }
David Howells454e2392006-06-23 02:02:57 -0700829
830 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -0500831 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 }
833
David Howells454e2392006-06-23 02:02:57 -0700834 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835
David Howells454e2392006-06-23 02:02:57 -0700836error_s:
837 error = PTR_ERR(s);
838error_bdev:
Al Viro30c40d22008-02-22 19:50:45 -0500839 close_bdev_exclusive(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -0700840error:
841 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842}
843
844EXPORT_SYMBOL(get_sb_bdev);
845
846void kill_block_super(struct super_block *sb)
847{
848 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -0500849 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850
Theodore Ts'o87d8fe12009-01-03 09:47:09 -0500851 bdev->bd_super = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 generic_shutdown_super(sb);
853 sync_blockdev(bdev);
Al Viro30c40d22008-02-22 19:50:45 -0500854 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855}
856
857EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200858#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859
David Howells454e2392006-06-23 02:02:57 -0700860int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700862 int (*fill_super)(struct super_block *, void *, int),
863 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864{
865 int error;
866 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
867
868 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700869 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870
871 s->s_flags = flags;
872
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800873 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 if (error) {
875 up_write(&s->s_umount);
876 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700877 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 }
879 s->s_flags |= MS_ACTIVE;
David Howells454e2392006-06-23 02:02:57 -0700880 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881}
882
883EXPORT_SYMBOL(get_sb_nodev);
884
885static int compare_single(struct super_block *s, void *p)
886{
887 return 1;
888}
889
David Howells454e2392006-06-23 02:02:57 -0700890int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700892 int (*fill_super)(struct super_block *, void *, int),
893 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894{
895 struct super_block *s;
896 int error;
897
898 s = sget(fs_type, compare_single, set_anon_super, NULL);
899 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700900 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 if (!s->s_root) {
902 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800903 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 if (error) {
905 up_write(&s->s_umount);
906 deactivate_super(s);
David Howells454e2392006-06-23 02:02:57 -0700907 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 }
909 s->s_flags |= MS_ACTIVE;
910 }
911 do_remount_sb(s, flags, data, 0);
David Howells454e2392006-06-23 02:02:57 -0700912 return simple_set_mnt(mnt, s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913}
914
915EXPORT_SYMBOL(get_sb_single);
916
917struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400918vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700922 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923
924 if (!type)
925 return ERR_PTR(-ENODEV);
926
David Howells454e2392006-06-23 02:02:57 -0700927 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 mnt = alloc_vfsmnt(name);
929 if (!mnt)
930 goto out;
931
Eric Parise0007522008-03-05 10:31:54 -0500932 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700934 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936
Eric Parise0007522008-03-05 10:31:54 -0500937 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700938 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 }
941
David Howells454e2392006-06-23 02:02:57 -0700942 error = type->get_sb(type, flags, name, data, mnt);
943 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 goto out_free_secdata;
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700945 BUG_ON(!mnt->mnt_sb);
David Howells454e2392006-06-23 02:02:57 -0700946
James Morris12204e22008-12-19 10:44:42 +1100947 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 if (error)
949 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700950
951 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700953 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700954 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 return mnt;
956out_sb:
David Howells454e2392006-06-23 02:02:57 -0700957 dput(mnt->mnt_root);
958 up_write(&mnt->mnt_sb->s_umount);
959 deactivate_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960out_free_secdata:
961 free_secdata(secdata);
962out_mnt:
963 free_vfsmnt(mnt);
964out:
David Howells454e2392006-06-23 02:02:57 -0700965 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966}
967
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400968EXPORT_SYMBOL_GPL(vfs_kern_mount);
969
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700970static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
971{
972 int err;
973 const char *subtype = strchr(fstype, '.');
974 if (subtype) {
975 subtype++;
976 err = -EINVAL;
977 if (!subtype[0])
978 goto err;
979 } else
980 subtype = "";
981
982 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
983 err = -ENOMEM;
984 if (!mnt->mnt_sb->s_subtype)
985 goto err;
986 return mnt;
987
988 err:
989 mntput(mnt);
990 return ERR_PTR(err);
991}
992
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400993struct vfsmount *
994do_kern_mount(const char *fstype, int flags, const char *name, void *data)
995{
996 struct file_system_type *type = get_fs_type(fstype);
997 struct vfsmount *mnt;
998 if (!type)
999 return ERR_PTR(-ENODEV);
1000 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -07001001 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
1002 !mnt->mnt_sb->s_subtype)
1003 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -04001004 put_filesystem(type);
1005 return mnt;
1006}
Al Viro8a4e98d2008-02-24 01:43:03 -05001007EXPORT_SYMBOL_GPL(do_kern_mount);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -04001008
Pavel Emelyanov8bf97252007-10-18 23:40:02 -07001009struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -07001011 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012}
1013
Pavel Emelyanov8bf97252007-10-18 23:40:02 -07001014EXPORT_SYMBOL_GPL(kern_mount_data);