blob: b03fea8fbfb6e0c1d5276e59141ec620f0050db7 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/mount.h>
32#include <linux/security.h>
33#include <linux/syscalls.h>
34#include <linux/vfs.h>
35#include <linux/writeback.h> /* for the emergency remount stuff */
36#include <linux/idr.h>
37#include <linux/kobject.h>
Ingo Molnar353ab6e2006-03-26 01:37:12 -080038#include <linux/mutex.h>
Dave Hansen49e0d022008-02-15 14:37:32 -080039#include <linux/file.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070040#include <asm/uaccess.h>
Al Viro6d59e7f2008-03-22 15:48:17 -040041#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
43
Linus Torvalds1da177e2005-04-16 15:20:36 -070044LIST_HEAD(super_blocks);
45DEFINE_SPINLOCK(sb_lock);
46
47/**
48 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -070049 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -070050 *
51 * Allocates and initializes a new &struct super_block. alloc_super()
52 * returns a pointer new superblock or %NULL if allocation had failed.
53 */
Ingo Molnarcf516242006-07-03 00:25:27 -070054static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -070055{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -080056 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 static struct super_operations default_op;
58
59 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070060 if (security_sb_alloc(s)) {
61 kfree(s);
62 s = NULL;
63 goto out;
64 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 INIT_LIST_HEAD(&s->s_files);
66 INIT_LIST_HEAD(&s->s_instances);
67 INIT_HLIST_HEAD(&s->s_anon);
68 INIT_LIST_HEAD(&s->s_inodes);
Kentaro Makitada3bbdd2008-07-23 21:27:13 -070069 INIT_LIST_HEAD(&s->s_dentry_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f42006-01-09 15:59:25 -080071 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -070072 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -070073 /*
74 * The locking rules for s_lock are up to the
75 * filesystem. For example ext3fs has different
76 * lock ordering than usbfs:
77 */
78 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Peter Zijlstraada723d2009-02-18 14:48:30 -080079 /*
80 * sget() can have s_umount recursion.
81 *
82 * When it cannot find a suitable sb, it allocates a new
83 * one (this one), and tries again to find a suitable old
84 * one.
85 *
86 * In case that succeeds, it will acquire the s_umount
87 * lock of the old one. Since these are clearly distrinct
88 * locks, and this object isn't exposed yet, there's no
89 * risk of deadlocks.
90 *
91 * Annotate this by putting this lock in a different
92 * subclass.
93 */
94 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
Linus Torvalds1da177e2005-04-16 15:20:36 -070095 s->s_count = S_BIAS;
96 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -080097 mutex_init(&s->s_vfs_rename_mutex);
Ingo Molnard3be9152006-03-23 03:00:29 -080098 mutex_init(&s->s_dquot.dqio_mutex);
99 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 init_rwsem(&s->s_dquot.dqptr_sem);
101 init_waitqueue_head(&s->s_wait_unfrozen);
102 s->s_maxbytes = MAX_NON_LFS;
103 s->dq_op = sb_dquot_ops;
104 s->s_qcop = sb_quotactl_ops;
105 s->s_op = &default_op;
106 s->s_time_gran = 1000000000;
107 }
108out:
109 return s;
110}
111
112/**
113 * destroy_super - frees a superblock
114 * @s: superblock to free
115 *
116 * Frees a superblock.
117 */
118static inline void destroy_super(struct super_block *s)
119{
120 security_sb_free(s);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700121 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800122 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 kfree(s);
124}
125
126/* Superblock refcounting */
127
128/*
129 * Drop a superblock's refcount. Returns non-zero if the superblock was
130 * destroyed. The caller must hold sb_lock.
131 */
Adrian Bunk6b09ae62008-04-29 00:58:54 -0700132static int __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133{
134 int ret = 0;
135
136 if (!--sb->s_count) {
137 destroy_super(sb);
138 ret = 1;
139 }
140 return ret;
141}
142
143/*
144 * Drop a superblock's refcount.
145 * Returns non-zero if the superblock is about to be destroyed and
146 * at least is already removed from super_blocks list, so if we are
147 * making a loop through super blocks then we need to restart.
148 * The caller must hold sb_lock.
149 */
150int __put_super_and_need_restart(struct super_block *sb)
151{
152 /* check for race with generic_shutdown_super() */
153 if (list_empty(&sb->s_list)) {
154 /* super block is removed, need to restart... */
155 __put_super(sb);
156 return 1;
157 }
158 /* can't be the last, since s_list is still in use */
159 sb->s_count--;
160 BUG_ON(sb->s_count == 0);
161 return 0;
162}
163
164/**
165 * put_super - drop a temporary reference to superblock
166 * @sb: superblock in question
167 *
168 * Drops a temporary reference, frees superblock if there's no
169 * references left.
170 */
Jens Axboe03ba3782009-09-09 09:08:54 +0200171void put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172{
173 spin_lock(&sb_lock);
174 __put_super(sb);
175 spin_unlock(&sb_lock);
176}
177
178
179/**
180 * deactivate_super - drop an active reference to superblock
181 * @s: superblock to deactivate
182 *
183 * Drops an active reference to superblock, acquiring a temprory one if
184 * there is no active references left. In that case we lock superblock,
185 * tell fs driver to shut it down and drop the temporary reference we
186 * had just acquired.
187 */
188void deactivate_super(struct super_block *s)
189{
190 struct file_system_type *fs = s->s_type;
191 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
192 s->s_count -= S_BIAS-1;
193 spin_unlock(&sb_lock);
Jan Kara9e3509e2009-01-26 16:45:12 +0100194 vfs_dq_off(s, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 down_write(&s->s_umount);
196 fs->kill_sb(s);
197 put_filesystem(fs);
198 put_super(s);
199 }
200}
201
202EXPORT_SYMBOL(deactivate_super);
203
204/**
Al Viro74dbbdd2009-05-06 01:07:50 -0400205 * deactivate_locked_super - drop an active reference to superblock
206 * @s: superblock to deactivate
207 *
208 * Equivalent of up_write(&s->s_umount); deactivate_super(s);, except that
209 * it does not unlock it until it's all over. As the result, it's safe to
210 * use to dispose of new superblock on ->get_sb() failure exits - nobody
211 * will see the sucker until it's all over. Equivalent using up_write +
212 * deactivate_super is safe for that purpose only if superblock is either
213 * safe to use or has NULL ->s_root when we unlock.
214 */
215void deactivate_locked_super(struct super_block *s)
216{
217 struct file_system_type *fs = s->s_type;
218 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
219 s->s_count -= S_BIAS-1;
220 spin_unlock(&sb_lock);
221 vfs_dq_off(s, 0);
222 fs->kill_sb(s);
223 put_filesystem(fs);
224 put_super(s);
225 } else {
226 up_write(&s->s_umount);
227 }
228}
229
230EXPORT_SYMBOL(deactivate_locked_super);
231
232/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 * grab_super - acquire an active reference
234 * @s: reference we are trying to make active
235 *
236 * Tries to acquire an active reference. grab_super() is used when we
237 * had just found a superblock in super_blocks or fs_type->fs_supers
238 * and want to turn it into a full-blown active reference. grab_super()
239 * is called with sb_lock held and drops it. Returns 1 in case of
240 * success, 0 if we had failed (superblock contents was already dead or
241 * dying when grab_super() had been called).
242 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700243static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244{
245 s->s_count++;
246 spin_unlock(&sb_lock);
247 down_write(&s->s_umount);
248 if (s->s_root) {
249 spin_lock(&sb_lock);
250 if (s->s_count > S_BIAS) {
251 atomic_inc(&s->s_active);
252 s->s_count--;
253 spin_unlock(&sb_lock);
254 return 1;
255 }
256 spin_unlock(&sb_lock);
257 }
258 up_write(&s->s_umount);
259 put_super(s);
260 yield();
261 return 0;
262}
263
David Howellscf9a2ae2006-08-29 19:05:54 +0100264/*
Al Viro914e2632006-10-18 13:55:46 -0400265 * Superblock locking. We really ought to get rid of these two.
266 */
267void lock_super(struct super_block * sb)
268{
269 get_fs_excl();
270 mutex_lock(&sb->s_lock);
271}
272
273void unlock_super(struct super_block * sb)
274{
275 put_fs_excl();
276 mutex_unlock(&sb->s_lock);
277}
278
279EXPORT_SYMBOL(lock_super);
280EXPORT_SYMBOL(unlock_super);
281
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282/**
283 * generic_shutdown_super - common helper for ->kill_sb()
284 * @sb: superblock to kill
285 *
286 * generic_shutdown_super() does all fs-independent work on superblock
287 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
288 * that need destruction out of superblock, call generic_shutdown_super()
289 * and release aforementioned objects. Note: dentries and inodes _are_
290 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700291 *
292 * Upon calling this function, the filesystem may no longer alter or
293 * rearrange the set of dentries belonging to this super_block, nor may it
294 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295 */
296void generic_shutdown_super(struct super_block *sb)
297{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800298 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299
Arjan van de Venefaee192009-01-06 07:20:54 -0800300
David Howellsc636ebd2006-10-11 01:22:19 -0700301 if (sb->s_root) {
302 shrink_dcache_for_umount(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200303 sync_filesystem(sb);
Al Viroa9e220f2009-05-05 22:10:44 -0400304 get_fs_excl();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 sb->s_flags &= ~MS_ACTIVE;
Arjan van de Venefaee192009-01-06 07:20:54 -0800306
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 /* bad name - it should be evict_inodes() */
308 invalidate_inodes(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 if (sop->put_super)
311 sop->put_super(sb);
312
313 /* Forget any remaining inodes */
314 if (invalidate_inodes(sb)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800315 printk("VFS: Busy inodes after unmount of %s. "
316 "Self-destruct in 5 seconds. Have a nice day...\n",
317 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 }
Al Viroa9e220f2009-05-05 22:10:44 -0400319 put_fs_excl();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 }
321 spin_lock(&sb_lock);
322 /* should be initialized for __put_super_and_need_restart() */
323 list_del_init(&sb->s_list);
324 list_del(&sb->s_instances);
325 spin_unlock(&sb_lock);
326 up_write(&sb->s_umount);
327}
328
329EXPORT_SYMBOL(generic_shutdown_super);
330
331/**
332 * sget - find or create a superblock
333 * @type: filesystem type superblock should belong to
334 * @test: comparison callback
335 * @set: setup callback
336 * @data: argument to each of them
337 */
338struct super_block *sget(struct file_system_type *type,
339 int (*test)(struct super_block *,void *),
340 int (*set)(struct super_block *,void *),
341 void *data)
342{
343 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700344 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 int err;
346
347retry:
348 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700349 if (test) {
350 list_for_each_entry(old, &type->fs_supers, s_instances) {
351 if (!test(old, data))
352 continue;
353 if (!grab_super(old))
354 goto retry;
Li Zefana3cfbb52009-03-12 14:31:29 -0700355 if (s) {
356 up_write(&s->s_umount);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700357 destroy_super(s);
Li Zefana3cfbb52009-03-12 14:31:29 -0700358 }
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700359 return old;
360 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 }
362 if (!s) {
363 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700364 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 if (!s)
366 return ERR_PTR(-ENOMEM);
367 goto retry;
368 }
369
370 err = set(s, data);
371 if (err) {
372 spin_unlock(&sb_lock);
Li Zefana3cfbb52009-03-12 14:31:29 -0700373 up_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 destroy_super(s);
375 return ERR_PTR(err);
376 }
377 s->s_type = type;
378 strlcpy(s->s_id, type->name, sizeof(s->s_id));
379 list_add_tail(&s->s_list, &super_blocks);
380 list_add(&s->s_instances, &type->fs_supers);
381 spin_unlock(&sb_lock);
382 get_filesystem(type);
383 return s;
384}
385
386EXPORT_SYMBOL(sget);
387
388void drop_super(struct super_block *sb)
389{
390 up_read(&sb->s_umount);
391 put_super(sb);
392}
393
394EXPORT_SYMBOL(drop_super);
395
Christoph Hellwige5004752009-05-05 16:08:56 +0200396/**
397 * sync_supers - helper for periodic superblock writeback
398 *
399 * Call the write_super method if present on all dirty superblocks in
400 * the system. This is for the periodic writeback used by most older
401 * filesystems. For data integrity superblock writeback use
402 * sync_filesystems() instead.
403 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 * Note: check the dirty flag before waiting, so we don't
405 * hold up the sync while mounting a device. (The newly
406 * mounted device won't need syncing.)
407 */
408void sync_supers(void)
409{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700410 struct super_block *sb;
411
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700413restart:
414 list_for_each_entry(sb, &super_blocks, s_list) {
Christoph Hellwige5004752009-05-05 16:08:56 +0200415 if (sb->s_op->write_super && sb->s_dirt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 sb->s_count++;
417 spin_unlock(&sb_lock);
Christoph Hellwige5004752009-05-05 16:08:56 +0200418
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 down_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200420 if (sb->s_root && sb->s_dirt)
421 sb->s_op->write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700422 up_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200423
Kirill Korotaev618f0632005-06-23 00:09:54 -0700424 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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432/**
433 * get_super - get the superblock of a device
434 * @bdev: device to get the superblock for
435 *
436 * Scans the superblock list and finds the superblock of the file system
437 * mounted on the device given. %NULL is returned if no match is found.
438 */
439
440struct super_block * get_super(struct block_device *bdev)
441{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700442 struct super_block *sb;
443
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 if (!bdev)
445 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700446
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700448rescan:
449 list_for_each_entry(sb, &super_blocks, s_list) {
450 if (sb->s_bdev == bdev) {
451 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700453 down_read(&sb->s_umount);
454 if (sb->s_root)
455 return sb;
456 up_read(&sb->s_umount);
457 /* restart only when sb is no longer on the list */
458 spin_lock(&sb_lock);
459 if (__put_super_and_need_restart(sb))
460 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 }
462 }
463 spin_unlock(&sb_lock);
464 return NULL;
465}
466
467EXPORT_SYMBOL(get_super);
468
469struct super_block * user_get_super(dev_t dev)
470{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700471 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700474rescan:
475 list_for_each_entry(sb, &super_blocks, s_list) {
476 if (sb->s_dev == dev) {
477 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700479 down_read(&sb->s_umount);
480 if (sb->s_root)
481 return sb;
482 up_read(&sb->s_umount);
483 /* restart only when sb is no longer on the list */
484 spin_lock(&sb_lock);
485 if (__put_super_and_need_restart(sb))
486 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 }
488 }
489 spin_unlock(&sb_lock);
490 return NULL;
491}
492
Heiko Carstens257ac262009-01-14 14:14:13 +0100493SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494{
495 struct super_block *s;
496 struct ustat tmp;
497 struct kstatfs sbuf;
498 int err = -EINVAL;
499
500 s = user_get_super(new_decode_dev(dev));
501 if (s == NULL)
502 goto out;
David Howells726c3342006-06-23 02:02:58 -0700503 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 drop_super(s);
505 if (err)
506 goto out;
507
508 memset(&tmp,0,sizeof(struct ustat));
509 tmp.f_tfree = sbuf.f_bfree;
510 tmp.f_tinode = sbuf.f_ffree;
511
512 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
513out:
514 return err;
515}
516
517/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 * do_remount_sb - asks filesystem to change mount options.
519 * @sb: superblock in question
520 * @flags: numeric part of options
521 * @data: the rest of options
522 * @force: whether or not to force the change
523 *
524 * Alters the mount options of a mounted file system.
525 */
526int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
527{
528 int retval;
Jan Kara0ff5af82008-04-28 02:14:33 -0700529 int remount_rw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530
David Howells93614012006-09-30 20:45:40 +0200531#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
533 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200534#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 if (flags & MS_RDONLY)
536 acct_auto_close(sb);
537 shrink_dcache_sb(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200538 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
540 /* If we are remounting RDONLY and current sb is read/write,
541 make sure there are no rw files opened */
542 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
543 if (force)
544 mark_files_ro(sb);
J. R. Okajimab0895512009-06-17 01:16:50 +0900545 else if (!fs_may_remount_ro(sb))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 return -EBUSY;
Jan Kara9e3509e2009-01-26 16:45:12 +0100547 retval = vfs_dq_off(sb, 1);
J. R. Okajimab0895512009-06-17 01:16:50 +0900548 if (retval < 0 && retval != -ENOSYS)
Jan Kara0ff5af82008-04-28 02:14:33 -0700549 return -EBUSY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 }
Jan Kara0ff5af82008-04-28 02:14:33 -0700551 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552
553 if (sb->s_op->remount_fs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 retval = sb->s_op->remount_fs(sb, &flags, data);
J. R. Okajimab0895512009-06-17 01:16:50 +0900555 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 return retval;
557 }
558 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Jan Kara0ff5af82008-04-28 02:14:33 -0700559 if (remount_rw)
Jan Kara9e3509e2009-01-26 16:45:12 +0100560 vfs_dq_quota_on_remount(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 return 0;
562}
563
Jens Axboea2a95372009-03-17 09:38:40 +0100564static void do_emergency_remount(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565{
566 struct super_block *sb;
567
568 spin_lock(&sb_lock);
569 list_for_each_entry(sb, &super_blocks, s_list) {
570 sb->s_count++;
571 spin_unlock(&sb_lock);
Al Viro443b94b2009-05-05 23:48:50 -0400572 down_write(&sb->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
574 /*
575 * ->remount_fs needs lock_kernel().
576 *
577 * What lock protects sb->s_flags??
578 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 do_remount_sb(sb, MS_RDONLY, NULL, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 }
Al Viro443b94b2009-05-05 23:48:50 -0400581 up_write(&sb->s_umount);
582 put_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 spin_lock(&sb_lock);
584 }
585 spin_unlock(&sb_lock);
Jens Axboea2a95372009-03-17 09:38:40 +0100586 kfree(work);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 printk("Emergency Remount complete\n");
588}
589
590void emergency_remount(void)
591{
Jens Axboea2a95372009-03-17 09:38:40 +0100592 struct work_struct *work;
593
594 work = kmalloc(sizeof(*work), GFP_ATOMIC);
595 if (work) {
596 INIT_WORK(work, do_emergency_remount);
597 schedule_work(work);
598 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599}
600
601/*
602 * Unnamed block devices are dummy devices used by virtual
603 * filesystems which don't use real block-devices. -- jrs
604 */
605
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400606static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
Al Viroc63e09e2009-06-24 02:05:18 -0400608static int unnamed_dev_start = 0; /* don't bother trying below it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
610int set_anon_super(struct super_block *s, void *data)
611{
612 int dev;
613 int error;
614
615 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400616 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 return -ENOMEM;
618 spin_lock(&unnamed_dev_lock);
Al Viroc63e09e2009-06-24 02:05:18 -0400619 error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
Al Virof21f6222009-06-24 03:12:00 -0400620 if (!error)
621 unnamed_dev_start = dev + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 spin_unlock(&unnamed_dev_lock);
623 if (error == -EAGAIN)
624 /* We raced and lost with another CPU. */
625 goto retry;
626 else if (error)
627 return -EAGAIN;
628
629 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
630 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400631 ida_remove(&unnamed_dev_ida, dev);
Al Virof21f6222009-06-24 03:12:00 -0400632 if (unnamed_dev_start > dev)
633 unnamed_dev_start = dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 spin_unlock(&unnamed_dev_lock);
635 return -EMFILE;
636 }
637 s->s_dev = MKDEV(0, dev & MINORMASK);
638 return 0;
639}
640
641EXPORT_SYMBOL(set_anon_super);
642
643void kill_anon_super(struct super_block *sb)
644{
645 int slot = MINOR(sb->s_dev);
646
647 generic_shutdown_super(sb);
648 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400649 ida_remove(&unnamed_dev_ida, slot);
Al Viroc63e09e2009-06-24 02:05:18 -0400650 if (slot < unnamed_dev_start)
651 unnamed_dev_start = slot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 spin_unlock(&unnamed_dev_lock);
653}
654
655EXPORT_SYMBOL(kill_anon_super);
656
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657void kill_litter_super(struct super_block *sb)
658{
659 if (sb->s_root)
660 d_genocide(sb->s_root);
661 kill_anon_super(sb);
662}
663
664EXPORT_SYMBOL(kill_litter_super);
665
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700666static int ns_test_super(struct super_block *sb, void *data)
667{
668 return sb->s_fs_info == data;
669}
670
671static int ns_set_super(struct super_block *sb, void *data)
672{
673 sb->s_fs_info = data;
674 return set_anon_super(sb, NULL);
675}
676
677int get_sb_ns(struct file_system_type *fs_type, int flags, void *data,
678 int (*fill_super)(struct super_block *, void *, int),
679 struct vfsmount *mnt)
680{
681 struct super_block *sb;
682
683 sb = sget(fs_type, ns_test_super, ns_set_super, data);
684 if (IS_ERR(sb))
685 return PTR_ERR(sb);
686
687 if (!sb->s_root) {
688 int err;
689 sb->s_flags = flags;
690 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
691 if (err) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400692 deactivate_locked_super(sb);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700693 return err;
694 }
695
696 sb->s_flags |= MS_ACTIVE;
697 }
698
699 simple_set_mnt(mnt, sb);
700 return 0;
701}
702
703EXPORT_SYMBOL(get_sb_ns);
704
David Howells93614012006-09-30 20:45:40 +0200705#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706static int set_bdev_super(struct super_block *s, void *data)
707{
708 s->s_bdev = data;
709 s->s_dev = s->s_bdev->bd_dev;
Jens Axboe32a88aa2009-09-16 15:02:33 +0200710
711 /*
712 * We set the bdi here to the queue backing, file systems can
713 * overwrite this in ->fill_super()
714 */
715 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 return 0;
717}
718
719static int test_bdev_super(struct super_block *s, void *data)
720{
721 return (void *)s->s_bdev == data;
722}
723
David Howells454e2392006-06-23 02:02:57 -0700724int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700726 int (*fill_super)(struct super_block *, void *, int),
727 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728{
729 struct block_device *bdev;
730 struct super_block *s;
Al Viro30c40d22008-02-22 19:50:45 -0500731 fmode_t mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 int error = 0;
733
Al Viro30c40d22008-02-22 19:50:45 -0500734 if (!(flags & MS_RDONLY))
735 mode |= FMODE_WRITE;
736
737 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700739 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740
741 /*
742 * once the super is inserted into the list by sget, s_umount
743 * will protect the lockfs code from trying to start a snapshot
744 * while we are mounting
745 */
David Chinnerf73ca1b2007-01-10 23:15:41 -0800746 down(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
David Chinnerf73ca1b2007-01-10 23:15:41 -0800748 up(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700750 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751
752 if (s->s_root) {
753 if ((flags ^ s->s_flags) & MS_RDONLY) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400754 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700755 error = -EBUSY;
756 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 }
David Howells454e2392006-06-23 02:02:57 -0700758
Al Viro30c40d22008-02-22 19:50:45 -0500759 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 } else {
761 char b[BDEVNAME_SIZE];
762
763 s->s_flags = flags;
Al Viro30c40d22008-02-22 19:50:45 -0500764 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800766 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800767 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400769 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700770 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800771 }
David Howells454e2392006-06-23 02:02:57 -0700772
773 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -0500774 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 }
776
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800777 simple_set_mnt(mnt, s);
778 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779
David Howells454e2392006-06-23 02:02:57 -0700780error_s:
781 error = PTR_ERR(s);
782error_bdev:
Al Viro30c40d22008-02-22 19:50:45 -0500783 close_bdev_exclusive(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -0700784error:
785 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786}
787
788EXPORT_SYMBOL(get_sb_bdev);
789
790void kill_block_super(struct super_block *sb)
791{
792 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -0500793 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
H Hartley Sweetenddbaaf32009-04-29 20:14:57 -0400795 bdev->bd_super = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 generic_shutdown_super(sb);
797 sync_blockdev(bdev);
Al Viro30c40d22008-02-22 19:50:45 -0500798 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799}
800
801EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200802#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803
David Howells454e2392006-06-23 02:02:57 -0700804int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700806 int (*fill_super)(struct super_block *, void *, int),
807 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808{
809 int error;
810 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
811
812 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700813 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
815 s->s_flags = flags;
816
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800817 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400819 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700820 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 }
822 s->s_flags |= MS_ACTIVE;
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800823 simple_set_mnt(mnt, s);
824 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825}
826
827EXPORT_SYMBOL(get_sb_nodev);
828
829static int compare_single(struct super_block *s, void *p)
830{
831 return 1;
832}
833
David Howells454e2392006-06-23 02:02:57 -0700834int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700836 int (*fill_super)(struct super_block *, void *, int),
837 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838{
839 struct super_block *s;
840 int error;
841
842 s = sget(fs_type, compare_single, set_anon_super, NULL);
843 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700844 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 if (!s->s_root) {
846 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800847 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400849 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700850 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 }
852 s->s_flags |= MS_ACTIVE;
853 }
854 do_remount_sb(s, flags, data, 0);
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800855 simple_set_mnt(mnt, s);
856 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857}
858
859EXPORT_SYMBOL(get_sb_single);
860
861struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400862vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700866 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867
868 if (!type)
869 return ERR_PTR(-ENODEV);
870
David Howells454e2392006-06-23 02:02:57 -0700871 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 mnt = alloc_vfsmnt(name);
873 if (!mnt)
874 goto out;
875
Eric Parise0007522008-03-05 10:31:54 -0500876 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700878 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880
Eric Parise0007522008-03-05 10:31:54 -0500881 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700882 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 }
885
David Howells454e2392006-06-23 02:02:57 -0700886 error = type->get_sb(type, flags, name, data, mnt);
887 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 goto out_free_secdata;
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700889 BUG_ON(!mnt->mnt_sb);
David Howells454e2392006-06-23 02:02:57 -0700890
James Morris12204e22008-12-19 10:44:42 +1100891 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 if (error)
893 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700894
895 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700897 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700898 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 return mnt;
900out_sb:
David Howells454e2392006-06-23 02:02:57 -0700901 dput(mnt->mnt_root);
Al Viro74dbbdd2009-05-06 01:07:50 -0400902 deactivate_locked_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903out_free_secdata:
904 free_secdata(secdata);
905out_mnt:
906 free_vfsmnt(mnt);
907out:
David Howells454e2392006-06-23 02:02:57 -0700908 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909}
910
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400911EXPORT_SYMBOL_GPL(vfs_kern_mount);
912
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700913static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
914{
915 int err;
916 const char *subtype = strchr(fstype, '.');
917 if (subtype) {
918 subtype++;
919 err = -EINVAL;
920 if (!subtype[0])
921 goto err;
922 } else
923 subtype = "";
924
925 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
926 err = -ENOMEM;
927 if (!mnt->mnt_sb->s_subtype)
928 goto err;
929 return mnt;
930
931 err:
932 mntput(mnt);
933 return ERR_PTR(err);
934}
935
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400936struct vfsmount *
937do_kern_mount(const char *fstype, int flags, const char *name, void *data)
938{
939 struct file_system_type *type = get_fs_type(fstype);
940 struct vfsmount *mnt;
941 if (!type)
942 return ERR_PTR(-ENODEV);
943 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700944 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
945 !mnt->mnt_sb->s_subtype)
946 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400947 put_filesystem(type);
948 return mnt;
949}
Al Viro8a4e98d2008-02-24 01:43:03 -0500950EXPORT_SYMBOL_GPL(do_kern_mount);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400951
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700952struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700954 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955}
956
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700957EXPORT_SYMBOL_GPL(kern_mount_data);