blob: cb19fffc7681f9d5b96bfa22c8050df0728bcaca [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 Molnar7892f2f2006-01-09 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);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 lock_super(sb);
308 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);
312 lock_kernel();
313
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 if (sop->put_super)
315 sop->put_super(sb);
316
317 /* Forget any remaining inodes */
318 if (invalidate_inodes(sb)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800319 printk("VFS: Busy inodes after unmount of %s. "
320 "Self-destruct in 5 seconds. Have a nice day...\n",
321 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 }
323
324 unlock_kernel();
325 unlock_super(sb);
326 }
327 spin_lock(&sb_lock);
328 /* should be initialized for __put_super_and_need_restart() */
329 list_del_init(&sb->s_list);
330 list_del(&sb->s_instances);
331 spin_unlock(&sb_lock);
332 up_write(&sb->s_umount);
333}
334
335EXPORT_SYMBOL(generic_shutdown_super);
336
337/**
338 * sget - find or create a superblock
339 * @type: filesystem type superblock should belong to
340 * @test: comparison callback
341 * @set: setup callback
342 * @data: argument to each of them
343 */
344struct super_block *sget(struct file_system_type *type,
345 int (*test)(struct super_block *,void *),
346 int (*set)(struct super_block *,void *),
347 void *data)
348{
349 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700350 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 int err;
352
353retry:
354 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700355 if (test) {
356 list_for_each_entry(old, &type->fs_supers, s_instances) {
357 if (!test(old, data))
358 continue;
359 if (!grab_super(old))
360 goto retry;
Li Zefana3cfbb52009-03-12 14:31:29 -0700361 if (s) {
362 up_write(&s->s_umount);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700363 destroy_super(s);
Li Zefana3cfbb52009-03-12 14:31:29 -0700364 }
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700365 return old;
366 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 }
368 if (!s) {
369 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700370 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 if (!s)
372 return ERR_PTR(-ENOMEM);
373 goto retry;
374 }
375
376 err = set(s, data);
377 if (err) {
378 spin_unlock(&sb_lock);
Li Zefana3cfbb52009-03-12 14:31:29 -0700379 up_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 destroy_super(s);
381 return ERR_PTR(err);
382 }
383 s->s_type = type;
384 strlcpy(s->s_id, type->name, sizeof(s->s_id));
385 list_add_tail(&s->s_list, &super_blocks);
386 list_add(&s->s_instances, &type->fs_supers);
387 spin_unlock(&sb_lock);
388 get_filesystem(type);
389 return s;
390}
391
392EXPORT_SYMBOL(sget);
393
394void drop_super(struct super_block *sb)
395{
396 up_read(&sb->s_umount);
397 put_super(sb);
398}
399
400EXPORT_SYMBOL(drop_super);
401
Christoph Hellwige5004752009-05-05 16:08:56 +0200402/**
403 * sync_supers - helper for periodic superblock writeback
404 *
405 * Call the write_super method if present on all dirty superblocks in
406 * the system. This is for the periodic writeback used by most older
407 * filesystems. For data integrity superblock writeback use
408 * sync_filesystems() instead.
409 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 * Note: check the dirty flag before waiting, so we don't
411 * hold up the sync while mounting a device. (The newly
412 * mounted device won't need syncing.)
413 */
414void sync_supers(void)
415{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700416 struct super_block *sb;
417
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700419restart:
420 list_for_each_entry(sb, &super_blocks, s_list) {
Christoph Hellwige5004752009-05-05 16:08:56 +0200421 if (sb->s_op->write_super && sb->s_dirt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 sb->s_count++;
423 spin_unlock(&sb_lock);
Christoph Hellwige5004752009-05-05 16:08:56 +0200424
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 down_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200426 lock_super(sb);
427 if (sb->s_root && sb->s_dirt)
428 sb->s_op->write_super(sb);
429 unlock_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700430 up_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200431
Kirill Korotaev618f0632005-06-23 00:09:54 -0700432 spin_lock(&sb_lock);
433 if (__put_super_and_need_restart(sb))
434 goto restart;
435 }
436 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 spin_unlock(&sb_lock);
438}
439
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440/**
441 * get_super - get the superblock of a device
442 * @bdev: device to get the superblock for
443 *
444 * Scans the superblock list and finds the superblock of the file system
445 * mounted on the device given. %NULL is returned if no match is found.
446 */
447
448struct super_block * get_super(struct block_device *bdev)
449{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700450 struct super_block *sb;
451
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 if (!bdev)
453 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700454
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700456rescan:
457 list_for_each_entry(sb, &super_blocks, s_list) {
458 if (sb->s_bdev == bdev) {
459 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700461 down_read(&sb->s_umount);
462 if (sb->s_root)
463 return sb;
464 up_read(&sb->s_umount);
465 /* restart only when sb is no longer on the list */
466 spin_lock(&sb_lock);
467 if (__put_super_and_need_restart(sb))
468 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 }
470 }
471 spin_unlock(&sb_lock);
472 return NULL;
473}
474
475EXPORT_SYMBOL(get_super);
476
477struct super_block * user_get_super(dev_t dev)
478{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700479 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700482rescan:
483 list_for_each_entry(sb, &super_blocks, s_list) {
484 if (sb->s_dev == dev) {
485 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700487 down_read(&sb->s_umount);
488 if (sb->s_root)
489 return sb;
490 up_read(&sb->s_umount);
491 /* restart only when sb is no longer on the list */
492 spin_lock(&sb_lock);
493 if (__put_super_and_need_restart(sb))
494 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 }
496 }
497 spin_unlock(&sb_lock);
498 return NULL;
499}
500
Heiko Carstens257ac262009-01-14 14:14:13 +0100501SYSCALL_DEFINE2(ustat, unsigned, dev, struct ustat __user *, ubuf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502{
503 struct super_block *s;
504 struct ustat tmp;
505 struct kstatfs sbuf;
506 int err = -EINVAL;
507
508 s = user_get_super(new_decode_dev(dev));
509 if (s == NULL)
510 goto out;
David Howells726c3342006-06-23 02:02:58 -0700511 err = vfs_statfs(s->s_root, &sbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 drop_super(s);
513 if (err)
514 goto out;
515
516 memset(&tmp,0,sizeof(struct ustat));
517 tmp.f_tfree = sbuf.f_bfree;
518 tmp.f_tinode = sbuf.f_ffree;
519
520 err = copy_to_user(ubuf,&tmp,sizeof(struct ustat)) ? -EFAULT : 0;
521out:
522 return err;
523}
524
525/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 * do_remount_sb - asks filesystem to change mount options.
527 * @sb: superblock in question
528 * @flags: numeric part of options
529 * @data: the rest of options
530 * @force: whether or not to force the change
531 *
532 * Alters the mount options of a mounted file system.
533 */
534int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
535{
536 int retval;
Jan Kara0ff5af82008-04-28 02:14:33 -0700537 int remount_rw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
David Howells93614012006-09-30 20:45:40 +0200539#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
541 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200542#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 if (flags & MS_RDONLY)
544 acct_auto_close(sb);
545 shrink_dcache_sb(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200546 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547
548 /* If we are remounting RDONLY and current sb is read/write,
549 make sure there are no rw files opened */
550 if ((flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY)) {
551 if (force)
552 mark_files_ro(sb);
553 else if (!fs_may_remount_ro(sb))
554 return -EBUSY;
Jan Kara9e3509e2009-01-26 16:45:12 +0100555 retval = vfs_dq_off(sb, 1);
Jan Kara0ff5af82008-04-28 02:14:33 -0700556 if (retval < 0 && retval != -ENOSYS)
557 return -EBUSY;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 }
Jan Kara0ff5af82008-04-28 02:14:33 -0700559 remount_rw = !(flags & MS_RDONLY) && (sb->s_flags & MS_RDONLY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560
561 if (sb->s_op->remount_fs) {
562 lock_super(sb);
563 retval = sb->s_op->remount_fs(sb, &flags, data);
564 unlock_super(sb);
565 if (retval)
566 return retval;
567 }
568 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Jan Kara0ff5af82008-04-28 02:14:33 -0700569 if (remount_rw)
Jan Kara9e3509e2009-01-26 16:45:12 +0100570 vfs_dq_quota_on_remount(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 return 0;
572}
573
Jens Axboea2a95372009-03-17 09:38:40 +0100574static void do_emergency_remount(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575{
576 struct super_block *sb;
577
578 spin_lock(&sb_lock);
579 list_for_each_entry(sb, &super_blocks, s_list) {
580 sb->s_count++;
581 spin_unlock(&sb_lock);
582 down_read(&sb->s_umount);
583 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
584 /*
585 * ->remount_fs needs lock_kernel().
586 *
587 * What lock protects sb->s_flags??
588 */
589 lock_kernel();
590 do_remount_sb(sb, MS_RDONLY, NULL, 1);
591 unlock_kernel();
592 }
593 drop_super(sb);
594 spin_lock(&sb_lock);
595 }
596 spin_unlock(&sb_lock);
Jens Axboea2a95372009-03-17 09:38:40 +0100597 kfree(work);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 printk("Emergency Remount complete\n");
599}
600
601void emergency_remount(void)
602{
Jens Axboea2a95372009-03-17 09:38:40 +0100603 struct work_struct *work;
604
605 work = kmalloc(sizeof(*work), GFP_ATOMIC);
606 if (work) {
607 INIT_WORK(work, do_emergency_remount);
608 schedule_work(work);
609 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610}
611
612/*
613 * Unnamed block devices are dummy devices used by virtual
614 * filesystems which don't use real block-devices. -- jrs
615 */
616
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400617static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
619
620int set_anon_super(struct super_block *s, void *data)
621{
622 int dev;
623 int error;
624
625 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400626 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 return -ENOMEM;
628 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400629 error = ida_get_new(&unnamed_dev_ida, &dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 spin_unlock(&unnamed_dev_lock);
631 if (error == -EAGAIN)
632 /* We raced and lost with another CPU. */
633 goto retry;
634 else if (error)
635 return -EAGAIN;
636
637 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
638 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400639 ida_remove(&unnamed_dev_ida, dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 spin_unlock(&unnamed_dev_lock);
641 return -EMFILE;
642 }
643 s->s_dev = MKDEV(0, dev & MINORMASK);
644 return 0;
645}
646
647EXPORT_SYMBOL(set_anon_super);
648
649void kill_anon_super(struct super_block *sb)
650{
651 int slot = MINOR(sb->s_dev);
652
653 generic_shutdown_super(sb);
654 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400655 ida_remove(&unnamed_dev_ida, slot);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 spin_unlock(&unnamed_dev_lock);
657}
658
659EXPORT_SYMBOL(kill_anon_super);
660
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661void kill_litter_super(struct super_block *sb)
662{
663 if (sb->s_root)
664 d_genocide(sb->s_root);
665 kill_anon_super(sb);
666}
667
668EXPORT_SYMBOL(kill_litter_super);
669
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700670static int ns_test_super(struct super_block *sb, void *data)
671{
672 return sb->s_fs_info == data;
673}
674
675static int ns_set_super(struct super_block *sb, void *data)
676{
677 sb->s_fs_info = data;
678 return set_anon_super(sb, NULL);
679}
680
681int get_sb_ns(struct file_system_type *fs_type, int flags, void *data,
682 int (*fill_super)(struct super_block *, void *, int),
683 struct vfsmount *mnt)
684{
685 struct super_block *sb;
686
687 sb = sget(fs_type, ns_test_super, ns_set_super, data);
688 if (IS_ERR(sb))
689 return PTR_ERR(sb);
690
691 if (!sb->s_root) {
692 int err;
693 sb->s_flags = flags;
694 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
695 if (err) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400696 deactivate_locked_super(sb);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700697 return err;
698 }
699
700 sb->s_flags |= MS_ACTIVE;
701 }
702
703 simple_set_mnt(mnt, sb);
704 return 0;
705}
706
707EXPORT_SYMBOL(get_sb_ns);
708
David Howells93614012006-09-30 20:45:40 +0200709#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710static int set_bdev_super(struct super_block *s, void *data)
711{
712 s->s_bdev = data;
713 s->s_dev = s->s_bdev->bd_dev;
714 return 0;
715}
716
717static int test_bdev_super(struct super_block *s, void *data)
718{
719 return (void *)s->s_bdev == data;
720}
721
David Howells454e2392006-06-23 02:02:57 -0700722int get_sb_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 int flags, const char *dev_name, void *data,
David Howells454e2392006-06-23 02:02:57 -0700724 int (*fill_super)(struct super_block *, void *, int),
725 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726{
727 struct block_device *bdev;
728 struct super_block *s;
Al Viro30c40d22008-02-22 19:50:45 -0500729 fmode_t mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 int error = 0;
731
Al Viro30c40d22008-02-22 19:50:45 -0500732 if (!(flags & MS_RDONLY))
733 mode |= FMODE_WRITE;
734
735 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 if (IS_ERR(bdev))
David Howells454e2392006-06-23 02:02:57 -0700737 return PTR_ERR(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738
739 /*
740 * once the super is inserted into the list by sget, s_umount
741 * will protect the lockfs code from trying to start a snapshot
742 * while we are mounting
743 */
David Chinnerf73ca1b2007-01-10 23:15:41 -0800744 down(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
David Chinnerf73ca1b2007-01-10 23:15:41 -0800746 up(&bdev->bd_mount_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700748 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749
750 if (s->s_root) {
751 if ((flags ^ s->s_flags) & MS_RDONLY) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400752 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700753 error = -EBUSY;
754 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 }
David Howells454e2392006-06-23 02:02:57 -0700756
Al Viro30c40d22008-02-22 19:50:45 -0500757 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 } else {
759 char b[BDEVNAME_SIZE];
760
761 s->s_flags = flags;
Al Viro30c40d22008-02-22 19:50:45 -0500762 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800764 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800765 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400767 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700768 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800769 }
David Howells454e2392006-06-23 02:02:57 -0700770
771 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -0500772 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 }
774
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800775 simple_set_mnt(mnt, s);
776 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777
David Howells454e2392006-06-23 02:02:57 -0700778error_s:
779 error = PTR_ERR(s);
780error_bdev:
Al Viro30c40d22008-02-22 19:50:45 -0500781 close_bdev_exclusive(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -0700782error:
783 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784}
785
786EXPORT_SYMBOL(get_sb_bdev);
787
788void kill_block_super(struct super_block *sb)
789{
790 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -0500791 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
H Hartley Sweetenddbaaf32009-04-29 20:14:57 -0400793 bdev->bd_super = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 generic_shutdown_super(sb);
795 sync_blockdev(bdev);
Al Viro30c40d22008-02-22 19:50:45 -0500796 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797}
798
799EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200800#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801
David Howells454e2392006-06-23 02:02:57 -0700802int get_sb_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700804 int (*fill_super)(struct super_block *, void *, int),
805 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806{
807 int error;
808 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
809
810 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700811 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
813 s->s_flags = flags;
814
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800815 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400817 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700818 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 }
820 s->s_flags |= MS_ACTIVE;
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800821 simple_set_mnt(mnt, s);
822 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823}
824
825EXPORT_SYMBOL(get_sb_nodev);
826
827static int compare_single(struct super_block *s, void *p)
828{
829 return 1;
830}
831
David Howells454e2392006-06-23 02:02:57 -0700832int get_sb_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 int flags, void *data,
David Howells454e2392006-06-23 02:02:57 -0700834 int (*fill_super)(struct super_block *, void *, int),
835 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836{
837 struct super_block *s;
838 int error;
839
840 s = sget(fs_type, compare_single, set_anon_super, NULL);
841 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700842 return PTR_ERR(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 if (!s->s_root) {
844 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800845 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400847 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700848 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 }
850 s->s_flags |= MS_ACTIVE;
851 }
852 do_remount_sb(s, flags, data, 0);
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800853 simple_set_mnt(mnt, s);
854 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855}
856
857EXPORT_SYMBOL(get_sb_single);
858
859struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400860vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 struct vfsmount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700864 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
866 if (!type)
867 return ERR_PTR(-ENODEV);
868
David Howells454e2392006-06-23 02:02:57 -0700869 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 mnt = alloc_vfsmnt(name);
871 if (!mnt)
872 goto out;
873
Eric Parise0007522008-03-05 10:31:54 -0500874 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700876 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
Eric Parise0007522008-03-05 10:31:54 -0500879 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700880 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 }
883
David Howells454e2392006-06-23 02:02:57 -0700884 error = type->get_sb(type, flags, name, data, mnt);
885 if (error < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 goto out_free_secdata;
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700887 BUG_ON(!mnt->mnt_sb);
David Howells454e2392006-06-23 02:02:57 -0700888
James Morris12204e22008-12-19 10:44:42 +1100889 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 if (error)
891 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -0700892
893 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -0700895 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -0700896 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 return mnt;
898out_sb:
David Howells454e2392006-06-23 02:02:57 -0700899 dput(mnt->mnt_root);
Al Viro74dbbdd2009-05-06 01:07:50 -0400900 deactivate_locked_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901out_free_secdata:
902 free_secdata(secdata);
903out_mnt:
904 free_vfsmnt(mnt);
905out:
David Howells454e2392006-06-23 02:02:57 -0700906 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907}
908
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400909EXPORT_SYMBOL_GPL(vfs_kern_mount);
910
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700911static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
912{
913 int err;
914 const char *subtype = strchr(fstype, '.');
915 if (subtype) {
916 subtype++;
917 err = -EINVAL;
918 if (!subtype[0])
919 goto err;
920 } else
921 subtype = "";
922
923 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
924 err = -ENOMEM;
925 if (!mnt->mnt_sb->s_subtype)
926 goto err;
927 return mnt;
928
929 err:
930 mntput(mnt);
931 return ERR_PTR(err);
932}
933
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400934struct vfsmount *
935do_kern_mount(const char *fstype, int flags, const char *name, void *data)
936{
937 struct file_system_type *type = get_fs_type(fstype);
938 struct vfsmount *mnt;
939 if (!type)
940 return ERR_PTR(-ENODEV);
941 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700942 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
943 !mnt->mnt_sb->s_subtype)
944 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400945 put_filesystem(type);
946 return mnt;
947}
Al Viro8a4e98d2008-02-24 01:43:03 -0500948EXPORT_SYMBOL_GPL(do_kern_mount);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400949
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700950struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700952 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953}
954
Pavel Emelyanov8bf97252007-10-18 23:40:02 -0700955EXPORT_SYMBOL_GPL(kern_mount_data);