blob: 83b47416d0069bba99e738212a0d673d2b98564e [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 }
65 INIT_LIST_HEAD(&s->s_dirty);
66 INIT_LIST_HEAD(&s->s_io);
Ken Chen0e0f4fc2007-10-16 23:30:38 -070067 INIT_LIST_HEAD(&s->s_more_io);
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 INIT_LIST_HEAD(&s->s_files);
69 INIT_LIST_HEAD(&s->s_instances);
70 INIT_HLIST_HEAD(&s->s_anon);
71 INIT_LIST_HEAD(&s->s_inodes);
Kentaro Makitada3bbdd2008-07-23 21:27:13 -070072 INIT_LIST_HEAD(&s->s_dentry_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f42006-01-09 15:59:25 -080074 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -070075 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -070076 /*
77 * The locking rules for s_lock are up to the
78 * filesystem. For example ext3fs has different
79 * lock ordering than usbfs:
80 */
81 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Peter Zijlstraada723d2009-02-18 14:48:30 -080082 /*
83 * sget() can have s_umount recursion.
84 *
85 * When it cannot find a suitable sb, it allocates a new
86 * one (this one), and tries again to find a suitable old
87 * one.
88 *
89 * In case that succeeds, it will acquire the s_umount
90 * lock of the old one. Since these are clearly distrinct
91 * locks, and this object isn't exposed yet, there's no
92 * risk of deadlocks.
93 *
94 * Annotate this by putting this lock in a different
95 * subclass.
96 */
97 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
Linus Torvalds1da177e2005-04-16 15:20:36 -070098 s->s_count = S_BIAS;
99 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -0800100 mutex_init(&s->s_vfs_rename_mutex);
Ingo Molnard3be9152006-03-23 03:00:29 -0800101 mutex_init(&s->s_dquot.dqio_mutex);
102 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 init_rwsem(&s->s_dquot.dqptr_sem);
104 init_waitqueue_head(&s->s_wait_unfrozen);
105 s->s_maxbytes = MAX_NON_LFS;
106 s->dq_op = sb_dquot_ops;
107 s->s_qcop = sb_quotactl_ops;
108 s->s_op = &default_op;
109 s->s_time_gran = 1000000000;
110 }
111out:
112 return s;
113}
114
115/**
116 * destroy_super - frees a superblock
117 * @s: superblock to free
118 *
119 * Frees a superblock.
120 */
121static inline void destroy_super(struct super_block *s)
122{
123 security_sb_free(s);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700124 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800125 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126 kfree(s);
127}
128
129/* Superblock refcounting */
130
131/*
132 * Drop a superblock's refcount. Returns non-zero if the superblock was
133 * destroyed. The caller must hold sb_lock.
134 */
Adrian Bunk6b09ae62008-04-29 00:58:54 -0700135static int __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136{
137 int ret = 0;
138
139 if (!--sb->s_count) {
140 destroy_super(sb);
141 ret = 1;
142 }
143 return ret;
144}
145
146/*
147 * Drop a superblock's refcount.
148 * Returns non-zero if the superblock is about to be destroyed and
149 * at least is already removed from super_blocks list, so if we are
150 * making a loop through super blocks then we need to restart.
151 * The caller must hold sb_lock.
152 */
153int __put_super_and_need_restart(struct super_block *sb)
154{
155 /* check for race with generic_shutdown_super() */
156 if (list_empty(&sb->s_list)) {
157 /* super block is removed, need to restart... */
158 __put_super(sb);
159 return 1;
160 }
161 /* can't be the last, since s_list is still in use */
162 sb->s_count--;
163 BUG_ON(sb->s_count == 0);
164 return 0;
165}
166
167/**
168 * put_super - drop a temporary reference to superblock
169 * @sb: superblock in question
170 *
171 * Drops a temporary reference, frees superblock if there's no
172 * references left.
173 */
174static void put_super(struct super_block *sb)
175{
176 spin_lock(&sb_lock);
177 __put_super(sb);
178 spin_unlock(&sb_lock);
179}
180
181
182/**
183 * deactivate_super - drop an active reference to superblock
184 * @s: superblock to deactivate
185 *
186 * Drops an active reference to superblock, acquiring a temprory one if
187 * there is no active references left. In that case we lock superblock,
188 * tell fs driver to shut it down and drop the temporary reference we
189 * had just acquired.
190 */
191void deactivate_super(struct super_block *s)
192{
193 struct file_system_type *fs = s->s_type;
194 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
195 s->s_count -= S_BIAS-1;
196 spin_unlock(&sb_lock);
Jan Kara9e3509e2009-01-26 16:45:12 +0100197 vfs_dq_off(s, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 down_write(&s->s_umount);
199 fs->kill_sb(s);
200 put_filesystem(fs);
201 put_super(s);
202 }
203}
204
205EXPORT_SYMBOL(deactivate_super);
206
207/**
Al Viro74dbbdd2009-05-06 01:07:50 -0400208 * deactivate_locked_super - drop an active reference to superblock
209 * @s: superblock to deactivate
210 *
211 * Equivalent of up_write(&s->s_umount); deactivate_super(s);, except that
212 * it does not unlock it until it's all over. As the result, it's safe to
213 * use to dispose of new superblock on ->get_sb() failure exits - nobody
214 * will see the sucker until it's all over. Equivalent using up_write +
215 * deactivate_super is safe for that purpose only if superblock is either
216 * safe to use or has NULL ->s_root when we unlock.
217 */
218void deactivate_locked_super(struct super_block *s)
219{
220 struct file_system_type *fs = s->s_type;
221 if (atomic_dec_and_lock(&s->s_active, &sb_lock)) {
222 s->s_count -= S_BIAS-1;
223 spin_unlock(&sb_lock);
224 vfs_dq_off(s, 0);
225 fs->kill_sb(s);
226 put_filesystem(fs);
227 put_super(s);
228 } else {
229 up_write(&s->s_umount);
230 }
231}
232
233EXPORT_SYMBOL(deactivate_locked_super);
234
235/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 * grab_super - acquire an active reference
237 * @s: reference we are trying to make active
238 *
239 * Tries to acquire an active reference. grab_super() is used when we
240 * had just found a superblock in super_blocks or fs_type->fs_supers
241 * and want to turn it into a full-blown active reference. grab_super()
242 * is called with sb_lock held and drops it. Returns 1 in case of
243 * success, 0 if we had failed (superblock contents was already dead or
244 * dying when grab_super() had been called).
245 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700246static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247{
248 s->s_count++;
249 spin_unlock(&sb_lock);
250 down_write(&s->s_umount);
251 if (s->s_root) {
252 spin_lock(&sb_lock);
253 if (s->s_count > S_BIAS) {
254 atomic_inc(&s->s_active);
255 s->s_count--;
256 spin_unlock(&sb_lock);
257 return 1;
258 }
259 spin_unlock(&sb_lock);
260 }
261 up_write(&s->s_umount);
262 put_super(s);
263 yield();
264 return 0;
265}
266
David Howellscf9a2ae2006-08-29 19:05:54 +0100267/*
Al Viro914e2632006-10-18 13:55:46 -0400268 * Superblock locking. We really ought to get rid of these two.
269 */
270void lock_super(struct super_block * sb)
271{
272 get_fs_excl();
273 mutex_lock(&sb->s_lock);
274}
275
276void unlock_super(struct super_block * sb)
277{
278 put_fs_excl();
279 mutex_unlock(&sb->s_lock);
280}
281
282EXPORT_SYMBOL(lock_super);
283EXPORT_SYMBOL(unlock_super);
284
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285/**
286 * generic_shutdown_super - common helper for ->kill_sb()
287 * @sb: superblock to kill
288 *
289 * generic_shutdown_super() does all fs-independent work on superblock
290 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
291 * that need destruction out of superblock, call generic_shutdown_super()
292 * and release aforementioned objects. Note: dentries and inodes _are_
293 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700294 *
295 * Upon calling this function, the filesystem may no longer alter or
296 * rearrange the set of dentries belonging to this super_block, nor may it
297 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 */
299void generic_shutdown_super(struct super_block *sb)
300{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800301 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
Arjan van de Venefaee192009-01-06 07:20:54 -0800303
David Howellsc636ebd2006-10-11 01:22:19 -0700304 if (sb->s_root) {
305 shrink_dcache_for_umount(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200306 sync_filesystem(sb);
Al Viroa9e220f2009-05-05 22:10:44 -0400307 get_fs_excl();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 sb->s_flags &= ~MS_ACTIVE;
Arjan van de Venefaee192009-01-06 07:20:54 -0800309
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 /* bad name - it should be evict_inodes() */
311 invalidate_inodes(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 if (sop->put_super)
314 sop->put_super(sb);
315
316 /* Forget any remaining inodes */
317 if (invalidate_inodes(sb)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800318 printk("VFS: Busy inodes after unmount of %s. "
319 "Self-destruct in 5 seconds. Have a nice day...\n",
320 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 }
Al Viroa9e220f2009-05-05 22:10:44 -0400322 put_fs_excl();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 }
324 spin_lock(&sb_lock);
325 /* should be initialized for __put_super_and_need_restart() */
326 list_del_init(&sb->s_list);
327 list_del(&sb->s_instances);
328 spin_unlock(&sb_lock);
329 up_write(&sb->s_umount);
330}
331
332EXPORT_SYMBOL(generic_shutdown_super);
333
334/**
335 * sget - find or create a superblock
336 * @type: filesystem type superblock should belong to
337 * @test: comparison callback
338 * @set: setup callback
339 * @data: argument to each of them
340 */
341struct super_block *sget(struct file_system_type *type,
342 int (*test)(struct super_block *,void *),
343 int (*set)(struct super_block *,void *),
344 void *data)
345{
346 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700347 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 int err;
349
350retry:
351 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700352 if (test) {
353 list_for_each_entry(old, &type->fs_supers, s_instances) {
354 if (!test(old, data))
355 continue;
356 if (!grab_super(old))
357 goto retry;
Li Zefana3cfbb52009-03-12 14:31:29 -0700358 if (s) {
359 up_write(&s->s_umount);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700360 destroy_super(s);
Li Zefana3cfbb52009-03-12 14:31:29 -0700361 }
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700362 return old;
363 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 }
365 if (!s) {
366 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700367 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 if (!s)
369 return ERR_PTR(-ENOMEM);
370 goto retry;
371 }
372
373 err = set(s, data);
374 if (err) {
375 spin_unlock(&sb_lock);
Li Zefana3cfbb52009-03-12 14:31:29 -0700376 up_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 destroy_super(s);
378 return ERR_PTR(err);
379 }
380 s->s_type = type;
381 strlcpy(s->s_id, type->name, sizeof(s->s_id));
382 list_add_tail(&s->s_list, &super_blocks);
383 list_add(&s->s_instances, &type->fs_supers);
384 spin_unlock(&sb_lock);
385 get_filesystem(type);
386 return s;
387}
388
389EXPORT_SYMBOL(sget);
390
391void drop_super(struct super_block *sb)
392{
393 up_read(&sb->s_umount);
394 put_super(sb);
395}
396
397EXPORT_SYMBOL(drop_super);
398
Christoph Hellwige5004752009-05-05 16:08:56 +0200399/**
400 * sync_supers - helper for periodic superblock writeback
401 *
402 * Call the write_super method if present on all dirty superblocks in
403 * the system. This is for the periodic writeback used by most older
404 * filesystems. For data integrity superblock writeback use
405 * sync_filesystems() instead.
406 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 * Note: check the dirty flag before waiting, so we don't
408 * hold up the sync while mounting a device. (The newly
409 * mounted device won't need syncing.)
410 */
411void sync_supers(void)
412{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700413 struct super_block *sb;
414
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700416restart:
417 list_for_each_entry(sb, &super_blocks, s_list) {
Christoph Hellwige5004752009-05-05 16:08:56 +0200418 if (sb->s_op->write_super && sb->s_dirt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 sb->s_count++;
420 spin_unlock(&sb_lock);
Christoph Hellwige5004752009-05-05 16:08:56 +0200421
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 down_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200423 if (sb->s_root && sb->s_dirt)
424 sb->s_op->write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700425 up_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200426
Kirill Korotaev618f0632005-06-23 00:09:54 -0700427 spin_lock(&sb_lock);
428 if (__put_super_and_need_restart(sb))
429 goto restart;
430 }
431 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 spin_unlock(&sb_lock);
433}
434
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435/**
436 * get_super - get the superblock of a device
437 * @bdev: device to get the superblock for
438 *
439 * Scans the superblock list and finds the superblock of the file system
440 * mounted on the device given. %NULL is returned if no match is found.
441 */
442
443struct super_block * get_super(struct block_device *bdev)
444{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700445 struct super_block *sb;
446
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 if (!bdev)
448 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700449
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700451rescan:
452 list_for_each_entry(sb, &super_blocks, s_list) {
453 if (sb->s_bdev == bdev) {
454 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700456 down_read(&sb->s_umount);
457 if (sb->s_root)
458 return sb;
459 up_read(&sb->s_umount);
460 /* restart only when sb is no longer on the list */
461 spin_lock(&sb_lock);
462 if (__put_super_and_need_restart(sb))
463 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 }
465 }
466 spin_unlock(&sb_lock);
467 return NULL;
468}
469
470EXPORT_SYMBOL(get_super);
471
472struct super_block * user_get_super(dev_t dev)
473{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700474 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700477rescan:
478 list_for_each_entry(sb, &super_blocks, s_list) {
479 if (sb->s_dev == dev) {
480 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700482 down_read(&sb->s_umount);
483 if (sb->s_root)
484 return sb;
485 up_read(&sb->s_umount);
486 /* restart only when sb is no longer on the list */
487 spin_lock(&sb_lock);
488 if (__put_super_and_need_restart(sb))
489 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 }
491 }
492 spin_unlock(&sb_lock);
493 return NULL;
494}
495
Heiko Carstens257ac262009-01-14 14:14:13 +0100496SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497{
498 struct super_block *s;
499 struct ustat tmp;
500 struct kstatfs sbuf;
501 int err = -EINVAL;
502
503 s = user_get_super(new_decode_dev(dev));
504 if (s == NULL)
505 goto out;
David Howells726c3342006-06-23 02:02:58 -0700506 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 drop_super(s);
508 if (err)
509 goto out;
510
511 memset(&tmp,0,sizeof(struct ustat));
512 tmp.f_tfree = sbuf.f_bfree;
513 tmp.f_tinode = sbuf.f_ffree;
514
515 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
516out:
517 return err;
518}
519
520/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 * do_remount_sb - asks filesystem to change mount options.
522 * @sb: superblock in question
523 * @flags: numeric part of options
524 * @data: the rest of options
525 * @force: whether or not to force the change
526 *
527 * Alters the mount options of a mounted file system.
528 */
529int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
530{
531 int retval;
Jan Kara0ff5af82008-04-28 02:14:33 -0700532 int remount_rw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
David Howells93614012006-09-30 20:45:40 +0200534#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
536 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200537#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 if (flags & MS_RDONLY)
539 acct_auto_close(sb);
540 shrink_dcache_sb(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200541 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542
543 /* If we are remounting RDONLY and current sb is read/write,
544 make sure there are no rw files opened */
545 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
546 if (force)
547 mark_files_ro(sb);
Al Viro4aa98cf2009-05-08 13:36:58 -0400548 else if (!fs_may_remount_ro(sb)) {
549 unlock_kernel();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 return -EBUSY;
Al Viro4aa98cf2009-05-08 13:36:58 -0400551 }
Jan Kara9e3509e2009-01-26 16:45:12 +0100552 retval = vfs_dq_off(sb, 1);
Al Viro4aa98cf2009-05-08 13:36:58 -0400553 if (retval < 0 && retval != -ENOSYS) {
554 unlock_kernel();
Jan Kara0ff5af82008-04-28 02:14:33 -0700555 return -EBUSY;
Al Viro4aa98cf2009-05-08 13:36:58 -0400556 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 }
Jan Kara0ff5af82008-04-28 02:14:33 -0700558 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559
560 if (sb->s_op->remount_fs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 retval = sb->s_op->remount_fs(sb, &flags, data);
Al Viro4aa98cf2009-05-08 13:36:58 -0400562 if (retval) {
563 unlock_kernel();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 return retval;
Al Viro4aa98cf2009-05-08 13:36:58 -0400565 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 }
567 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Jan Kara0ff5af82008-04-28 02:14:33 -0700568 if (remount_rw)
Jan Kara9e3509e2009-01-26 16:45:12 +0100569 vfs_dq_quota_on_remount(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 return 0;
571}
572
Jens Axboea2a95372009-03-17 09:38:40 +0100573static void do_emergency_remount(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574{
575 struct super_block *sb;
576
577 spin_lock(&sb_lock);
578 list_for_each_entry(sb, &super_blocks, s_list) {
579 sb->s_count++;
580 spin_unlock(&sb_lock);
Al Viro443b94b2009-05-05 23:48:50 -0400581 down_write(&sb->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
583 /*
584 * ->remount_fs needs lock_kernel().
585 *
586 * What lock protects sb->s_flags??
587 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 do_remount_sb(sb, MS_RDONLY, NULL, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 }
Al Viro443b94b2009-05-05 23:48:50 -0400590 up_write(&sb->s_umount);
591 put_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 spin_lock(&sb_lock);
593 }
594 spin_unlock(&sb_lock);
Jens Axboea2a95372009-03-17 09:38:40 +0100595 kfree(work);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 printk("Emergency Remount complete\n");
597}
598
599void emergency_remount(void)
600{
Jens Axboea2a95372009-03-17 09:38:40 +0100601 struct work_struct *work;
602
603 work = kmalloc(sizeof(*work), GFP_ATOMIC);
604 if (work) {
605 INIT_WORK(work, do_emergency_remount);
606 schedule_work(work);
607 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608}
609
610/*
611 * Unnamed block devices are dummy devices used by virtual
612 * filesystems which don't use real block-devices. -- jrs
613 */
614
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400615static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
617
618int set_anon_super(struct super_block *s, void *data)
619{
620 int dev;
621 int error;
622
623 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400624 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 return -ENOMEM;
626 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400627 error = ida_get_new(&unnamed_dev_ida, &dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 spin_unlock(&unnamed_dev_lock);
629 if (error == -EAGAIN)
630 /* We raced and lost with another CPU. */
631 goto retry;
632 else if (error)
633 return -EAGAIN;
634
635 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
636 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400637 ida_remove(&unnamed_dev_ida, dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 spin_unlock(&unnamed_dev_lock);
639 return -EMFILE;
640 }
641 s->s_dev = MKDEV(0, dev & MINORMASK);
642 return 0;
643}
644
645EXPORT_SYMBOL(set_anon_super);
646
647void kill_anon_super(struct super_block *sb)
648{
649 int slot = MINOR(sb->s_dev);
650
651 generic_shutdown_super(sb);
652 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400653 ida_remove(&unnamed_dev_ida, slot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 spin_unlock(&unnamed_dev_lock);
655}
656
657EXPORT_SYMBOL(kill_anon_super);
658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659void kill_litter_super(struct super_block *sb)
660{
661 if (sb->s_root)
662 d_genocide(sb->s_root);
663 kill_anon_super(sb);
664}
665
666EXPORT_SYMBOL(kill_litter_super);
667
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700668static int ns_test_super(struct super_block *sb, void *data)
669{
670 return sb->s_fs_info == data;
671}
672
673static int ns_set_super(struct super_block *sb, void *data)
674{
675 sb->s_fs_info = data;
676 return set_anon_super(sb, NULL);
677}
678
679int get_sb_ns(struct file_system_type *fs_type, int flags, void *data,
680 int (*fill_super)(struct super_block *, void *, int),
681 struct vfsmount *mnt)
682{
683 struct super_block *sb;
684
685 sb = sget(fs_type, ns_test_super, ns_set_super, data);
686 if (IS_ERR(sb))
687 return PTR_ERR(sb);
688
689 if (!sb->s_root) {
690 int err;
691 sb->s_flags = flags;
692 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
693 if (err) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400694 deactivate_locked_super(sb);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700695 return err;
696 }
697
698 sb->s_flags |= MS_ACTIVE;
699 }
700
701 simple_set_mnt(mnt, sb);
702 return 0;
703}
704
705EXPORT_SYMBOL(get_sb_ns);
706
David Howells93614012006-09-30 20:45:40 +0200707#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708static int set_bdev_super(struct super_block *s, void *data)
709{
710 s->s_bdev = data;
711 s->s_dev = s->s_bdev->bd_dev;
712 return 0;
713}
714
715static int test_bdev_super(struct super_block *s, void *data)
716{
717 return (void *)s->s_bdev == data;
718}
719
David Howells454e2392006-06-23 02:02:57 -0700720int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700722 int (*fill_super)(struct super_block *, void *, int),
723 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724{
725 struct block_device *bdev;
726 struct super_block *s;
Al Viro30c40d22008-02-22 19:50:45 -0500727 fmode_t mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 int error = 0;
729
Al Viro30c40d22008-02-22 19:50:45 -0500730 if (!(flags & MS_RDONLY))
731 mode |= FMODE_WRITE;
732
733 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700735 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736
737 /*
738 * once the super is inserted into the list by sget, s_umount
739 * will protect the lockfs code from trying to start a snapshot
740 * while we are mounting
741 */
David Chinnerf73ca1b2007-01-10 23:15:41 -0800742 down(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
David Chinnerf73ca1b2007-01-10 23:15:41 -0800744 up(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700746 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
748 if (s->s_root) {
749 if ((flags ^ s->s_flags) & MS_RDONLY) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400750 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700751 error = -EBUSY;
752 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 }
David Howells454e2392006-06-23 02:02:57 -0700754
Al Viro30c40d22008-02-22 19:50:45 -0500755 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 } else {
757 char b[BDEVNAME_SIZE];
758
759 s->s_flags = flags;
Al Viro30c40d22008-02-22 19:50:45 -0500760 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800762 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800763 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400765 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700766 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800767 }
David Howells454e2392006-06-23 02:02:57 -0700768
769 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -0500770 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 }
772
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800773 simple_set_mnt(mnt, s);
774 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
David Howells454e2392006-06-23 02:02:57 -0700776error_s:
777 error = PTR_ERR(s);
778error_bdev:
Al Viro30c40d22008-02-22 19:50:45 -0500779 close_bdev_exclusive(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -0700780error:
781 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782}
783
784EXPORT_SYMBOL(get_sb_bdev);
785
786void kill_block_super(struct super_block *sb)
787{
788 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -0500789 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790
H Hartley Sweetenddbaaf32009-04-29 20:14:57 -0400791 bdev->bd_super = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 generic_shutdown_super(sb);
793 sync_blockdev(bdev);
Al Viro30c40d22008-02-22 19:50:45 -0500794 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795}
796
797EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200798#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799
David Howells454e2392006-06-23 02:02:57 -0700800int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700802 int (*fill_super)(struct super_block *, void *, int),
803 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804{
805 int error;
806 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
807
808 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700809 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
811 s->s_flags = flags;
812
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800813 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400815 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700816 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 }
818 s->s_flags |= MS_ACTIVE;
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800819 simple_set_mnt(mnt, s);
820 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821}
822
823EXPORT_SYMBOL(get_sb_nodev);
824
825static int compare_single(struct super_block *s, void *p)
826{
827 return 1;
828}
829
David Howells454e2392006-06-23 02:02:57 -0700830int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700832 int (*fill_super)(struct super_block *, void *, int),
833 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834{
835 struct super_block *s;
836 int error;
837
838 s = sget(fs_type, compare_single, set_anon_super, NULL);
839 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700840 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 if (!s->s_root) {
842 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800843 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400845 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700846 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 }
848 s->s_flags |= MS_ACTIVE;
849 }
850 do_remount_sb(s, flags, data, 0);
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800851 simple_set_mnt(mnt, s);
852 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853}
854
855EXPORT_SYMBOL(get_sb_single);
856
857struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400858vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700862 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
864 if (!type)
865 return ERR_PTR(-ENODEV);
866
David Howells454e2392006-06-23 02:02:57 -0700867 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 mnt = alloc_vfsmnt(name);
869 if (!mnt)
870 goto out;
871
Eric Parise0007522008-03-05 10:31:54 -0500872 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700874 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876
Eric Parise0007522008-03-05 10:31:54 -0500877 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700878 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 }
881
David Howells454e2392006-06-23 02:02:57 -0700882 error = type->get_sb(type, flags, name, data, mnt);
883 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 goto out_free_secdata;
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700885 BUG_ON(!mnt->mnt_sb);
David Howells454e2392006-06-23 02:02:57 -0700886
James Morris12204e22008-12-19 10:44:42 +1100887 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 if (error)
889 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700890
891 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700893 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700894 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 return mnt;
896out_sb:
David Howells454e2392006-06-23 02:02:57 -0700897 dput(mnt->mnt_root);
Al Viro74dbbdd2009-05-06 01:07:50 -0400898 deactivate_locked_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899out_free_secdata:
900 free_secdata(secdata);
901out_mnt:
902 free_vfsmnt(mnt);
903out:
David Howells454e2392006-06-23 02:02:57 -0700904 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905}
906
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400907EXPORT_SYMBOL_GPL(vfs_kern_mount);
908
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700909static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
910{
911 int err;
912 const char *subtype = strchr(fstype, '.');
913 if (subtype) {
914 subtype++;
915 err = -EINVAL;
916 if (!subtype[0])
917 goto err;
918 } else
919 subtype = "";
920
921 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
922 err = -ENOMEM;
923 if (!mnt->mnt_sb->s_subtype)
924 goto err;
925 return mnt;
926
927 err:
928 mntput(mnt);
929 return ERR_PTR(err);
930}
931
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400932struct vfsmount *
933do_kern_mount(const char *fstype, int flags, const char *name, void *data)
934{
935 struct file_system_type *type = get_fs_type(fstype);
936 struct vfsmount *mnt;
937 if (!type)
938 return ERR_PTR(-ENODEV);
939 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700940 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
941 !mnt->mnt_sb->s_subtype)
942 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400943 put_filesystem(type);
944 return mnt;
945}
Al Viro8a4e98d2008-02-24 01:43:03 -0500946EXPORT_SYMBOL_GPL(do_kern_mount);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400947
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700948struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700950 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951}
952
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700953EXPORT_SYMBOL_GPL(kern_mount_data);