blob: 7e9dd4cc2c01170d0d7a38035a430e554c90cc57 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/acct.h>
26#include <linux/blkdev.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/mount.h>
28#include <linux/security.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/writeback.h> /* for the emergency remount stuff */
30#include <linux/idr.h>
Ingo Molnar353ab6e2006-03-26 01:37:12 -080031#include <linux/mutex.h>
Jens Axboe5477d0f2010-04-29 20:33:35 +020032#include <linux/backing-dev.h>
Nick Pigginceb5bdc2011-01-07 17:50:05 +110033#include <linux/rculist_bl.h>
Al Viro6d59e7f2008-03-22 15:48:17 -040034#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
36
Linus Torvalds1da177e2005-04-16 15:20:36 -070037LIST_HEAD(super_blocks);
38DEFINE_SPINLOCK(sb_lock);
39
40/**
41 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -070042 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -070043 *
44 * Allocates and initializes a new &struct super_block. alloc_super()
45 * returns a pointer new superblock or %NULL if allocation had failed.
46 */
Ingo Molnarcf516242006-07-03 00:25:27 -070047static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -070048{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -080049 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Alexey Dobriyanb87221d2009-09-21 17:01:09 -070050 static const struct super_operations default_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -070051
52 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070053 if (security_sb_alloc(s)) {
54 kfree(s);
55 s = NULL;
56 goto out;
57 }
Nick Piggin6416ccb2010-08-18 04:37:38 +100058#ifdef CONFIG_SMP
59 s->s_files = alloc_percpu(struct list_head);
60 if (!s->s_files) {
61 security_sb_free(s);
62 kfree(s);
63 s = NULL;
64 goto out;
65 } else {
66 int i;
67
68 for_each_possible_cpu(i)
69 INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i));
70 }
71#else
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 INIT_LIST_HEAD(&s->s_files);
Nick Piggin6416ccb2010-08-18 04:37:38 +100073#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 INIT_LIST_HEAD(&s->s_instances);
Nick Pigginceb5bdc2011-01-07 17:50:05 +110075 INIT_HLIST_BL_HEAD(&s->s_anon);
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 INIT_LIST_HEAD(&s->s_inodes);
Kentaro Makitada3bbdd2008-07-23 21:27:13 -070077 INIT_LIST_HEAD(&s->s_dentry_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f2006-01-09 15:59:25 -080079 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -070080 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -070081 /*
82 * The locking rules for s_lock are up to the
83 * filesystem. For example ext3fs has different
84 * lock ordering than usbfs:
85 */
86 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Peter Zijlstraada723d2009-02-18 14:48:30 -080087 /*
88 * sget() can have s_umount recursion.
89 *
90 * When it cannot find a suitable sb, it allocates a new
91 * one (this one), and tries again to find a suitable old
92 * one.
93 *
94 * In case that succeeds, it will acquire the s_umount
95 * lock of the old one. Since these are clearly distrinct
96 * locks, and this object isn't exposed yet, there's no
97 * risk of deadlocks.
98 *
99 * Annotate this by putting this lock in a different
100 * subclass.
101 */
102 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
Al Virob20bd1a2010-03-22 08:53:19 -0400103 s->s_count = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -0800105 mutex_init(&s->s_vfs_rename_mutex);
Roland Dreier51ee0492010-04-27 14:23:57 -0700106 lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key);
Ingo Molnard3be9152006-03-23 03:00:29 -0800107 mutex_init(&s->s_dquot.dqio_mutex);
108 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109 init_rwsem(&s->s_dquot.dqptr_sem);
110 init_waitqueue_head(&s->s_wait_unfrozen);
111 s->s_maxbytes = MAX_NON_LFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112 s->s_op = &default_op;
113 s->s_time_gran = 1000000000;
114 }
115out:
116 return s;
117}
118
119/**
120 * destroy_super - frees a superblock
121 * @s: superblock to free
122 *
123 * Frees a superblock.
124 */
125static inline void destroy_super(struct super_block *s)
126{
Nick Piggin6416ccb2010-08-18 04:37:38 +1000127#ifdef CONFIG_SMP
128 free_percpu(s->s_files);
129#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 security_sb_free(s);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700131 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800132 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 kfree(s);
134}
135
136/* Superblock refcounting */
137
138/*
Al Viro35cf7ba2010-03-22 21:13:53 -0400139 * Drop a superblock's refcount. The caller must hold sb_lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 */
Al Viro35cf7ba2010-03-22 21:13:53 -0400141void __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 if (!--sb->s_count) {
Al Viro551de6f2010-03-22 19:36:35 -0400144 list_del_init(&sb->s_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 destroy_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147}
148
149/**
150 * put_super - drop a temporary reference to superblock
151 * @sb: superblock in question
152 *
153 * Drops a temporary reference, frees superblock if there's no
154 * references left.
155 */
Jens Axboe03ba3782009-09-09 09:08:54 +0200156void put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157{
158 spin_lock(&sb_lock);
159 __put_super(sb);
160 spin_unlock(&sb_lock);
161}
162
163
164/**
Al Viro74dbbdd2009-05-06 01:07:50 -0400165 * deactivate_locked_super - drop an active reference to superblock
166 * @s: superblock to deactivate
167 *
Al Viro1712ac82010-03-22 15:22:31 -0400168 * Drops an active reference to superblock, converting it into a temprory
169 * one if there is no other active references left. In that case we
170 * tell fs driver to shut it down and drop the temporary reference we
171 * had just acquired.
172 *
173 * Caller holds exclusive lock on superblock; that lock is released.
Al Viro74dbbdd2009-05-06 01:07:50 -0400174 */
175void deactivate_locked_super(struct super_block *s)
176{
177 struct file_system_type *fs = s->s_type;
Al Virob20bd1a2010-03-22 08:53:19 -0400178 if (atomic_dec_and_test(&s->s_active)) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400179 fs->kill_sb(s);
Boaz Harroshd863b502011-02-10 15:01:20 -0800180 /*
181 * We need to call rcu_barrier so all the delayed rcu free
182 * inodes are flushed before we release the fs module.
183 */
184 rcu_barrier();
Al Viro74dbbdd2009-05-06 01:07:50 -0400185 put_filesystem(fs);
186 put_super(s);
187 } else {
188 up_write(&s->s_umount);
189 }
190}
191
192EXPORT_SYMBOL(deactivate_locked_super);
193
194/**
Al Viro1712ac82010-03-22 15:22:31 -0400195 * deactivate_super - drop an active reference to superblock
196 * @s: superblock to deactivate
197 *
198 * Variant of deactivate_locked_super(), except that superblock is *not*
199 * locked by caller. If we are going to drop the final active reference,
200 * lock will be acquired prior to that.
201 */
202void deactivate_super(struct super_block *s)
203{
204 if (!atomic_add_unless(&s->s_active, -1, 1)) {
205 down_write(&s->s_umount);
206 deactivate_locked_super(s);
207 }
208}
209
210EXPORT_SYMBOL(deactivate_super);
211
212/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 * grab_super - acquire an active reference
214 * @s: reference we are trying to make active
215 *
216 * Tries to acquire an active reference. grab_super() is used when we
217 * had just found a superblock in super_blocks or fs_type->fs_supers
218 * and want to turn it into a full-blown active reference. grab_super()
219 * is called with sb_lock held and drops it. Returns 1 in case of
220 * success, 0 if we had failed (superblock contents was already dead or
221 * dying when grab_super() had been called).
222 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700223static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224{
Al Virob20bd1a2010-03-22 08:53:19 -0400225 if (atomic_inc_not_zero(&s->s_active)) {
226 spin_unlock(&sb_lock);
Al Virob20bd1a2010-03-22 08:53:19 -0400227 return 1;
228 }
229 /* it's going away */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 s->s_count++;
231 spin_unlock(&sb_lock);
Al Viro1712ac82010-03-22 15:22:31 -0400232 /* wait for it to die */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 up_write(&s->s_umount);
235 put_super(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 return 0;
237}
238
David Howellscf9a2ae2006-08-29 19:05:54 +0100239/*
Al Viro914e2632006-10-18 13:55:46 -0400240 * Superblock locking. We really ought to get rid of these two.
241 */
242void lock_super(struct super_block * sb)
243{
244 get_fs_excl();
245 mutex_lock(&sb->s_lock);
246}
247
248void unlock_super(struct super_block * sb)
249{
250 put_fs_excl();
251 mutex_unlock(&sb->s_lock);
252}
253
254EXPORT_SYMBOL(lock_super);
255EXPORT_SYMBOL(unlock_super);
256
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257/**
258 * generic_shutdown_super - common helper for ->kill_sb()
259 * @sb: superblock to kill
260 *
261 * generic_shutdown_super() does all fs-independent work on superblock
262 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
263 * that need destruction out of superblock, call generic_shutdown_super()
264 * and release aforementioned objects. Note: dentries and inodes _are_
265 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700266 *
267 * Upon calling this function, the filesystem may no longer alter or
268 * rearrange the set of dentries belonging to this super_block, nor may it
269 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 */
271void generic_shutdown_super(struct super_block *sb)
272{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800273 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274
Arjan van de Venefaee192009-01-06 07:20:54 -0800275
David Howellsc636ebd2006-10-11 01:22:19 -0700276 if (sb->s_root) {
277 shrink_dcache_for_umount(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200278 sync_filesystem(sb);
Al Viroa9e220f2009-05-05 22:10:44 -0400279 get_fs_excl();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 sb->s_flags &= ~MS_ACTIVE;
Arjan van de Venefaee192009-01-06 07:20:54 -0800281
Al Viro63997e92010-10-25 20:49:35 -0400282 fsnotify_unmount_inodes(&sb->s_inodes);
283
284 evict_inodes(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 if (sop->put_super)
287 sop->put_super(sb);
288
Al Viro63997e92010-10-25 20:49:35 -0400289 if (!list_empty(&sb->s_inodes)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800290 printk("VFS: Busy inodes after unmount of %s. "
291 "Self-destruct in 5 seconds. Have a nice day...\n",
292 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 }
Al Viroa9e220f2009-05-05 22:10:44 -0400294 put_fs_excl();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295 }
296 spin_lock(&sb_lock);
297 /* should be initialized for __put_super_and_need_restart() */
Al Viro551de6f2010-03-22 19:36:35 -0400298 list_del_init(&sb->s_instances);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 spin_unlock(&sb_lock);
300 up_write(&sb->s_umount);
301}
302
303EXPORT_SYMBOL(generic_shutdown_super);
304
305/**
306 * sget - find or create a superblock
307 * @type: filesystem type superblock should belong to
308 * @test: comparison callback
309 * @set: setup callback
310 * @data: argument to each of them
311 */
312struct super_block *sget(struct file_system_type *type,
313 int (*test)(struct super_block *,void *),
314 int (*set)(struct super_block *,void *),
315 void *data)
316{
317 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700318 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 int err;
320
321retry:
322 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700323 if (test) {
324 list_for_each_entry(old, &type->fs_supers, s_instances) {
325 if (!test(old, data))
326 continue;
327 if (!grab_super(old))
328 goto retry;
Li Zefana3cfbb52009-03-12 14:31:29 -0700329 if (s) {
330 up_write(&s->s_umount);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700331 destroy_super(s);
Al Viro7a4dec52010-08-09 12:05:43 -0400332 s = NULL;
Li Zefana3cfbb52009-03-12 14:31:29 -0700333 }
Al Virod3f21472010-03-23 11:11:05 -0400334 down_write(&old->s_umount);
Al Viro7a4dec52010-08-09 12:05:43 -0400335 if (unlikely(!(old->s_flags & MS_BORN))) {
336 deactivate_locked_super(old);
337 goto retry;
338 }
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700339 return old;
340 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 }
342 if (!s) {
343 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700344 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 if (!s)
346 return ERR_PTR(-ENOMEM);
347 goto retry;
348 }
349
350 err = set(s, data);
351 if (err) {
352 spin_unlock(&sb_lock);
Li Zefana3cfbb52009-03-12 14:31:29 -0700353 up_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 destroy_super(s);
355 return ERR_PTR(err);
356 }
357 s->s_type = type;
358 strlcpy(s->s_id, type->name, sizeof(s->s_id));
359 list_add_tail(&s->s_list, &super_blocks);
360 list_add(&s->s_instances, &type->fs_supers);
361 spin_unlock(&sb_lock);
362 get_filesystem(type);
363 return s;
364}
365
366EXPORT_SYMBOL(sget);
367
368void drop_super(struct super_block *sb)
369{
370 up_read(&sb->s_umount);
371 put_super(sb);
372}
373
374EXPORT_SYMBOL(drop_super);
375
Christoph Hellwige5004752009-05-05 16:08:56 +0200376/**
377 * sync_supers - helper for periodic superblock writeback
378 *
379 * Call the write_super method if present on all dirty superblocks in
380 * the system. This is for the periodic writeback used by most older
381 * filesystems. For data integrity superblock writeback use
382 * sync_filesystems() instead.
383 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 * Note: check the dirty flag before waiting, so we don't
385 * hold up the sync while mounting a device. (The newly
386 * mounted device won't need syncing.)
387 */
388void sync_supers(void)
389{
Al Virodca33252010-07-25 02:31:46 +0400390 struct super_block *sb, *p = NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700391
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400393 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro551de6f2010-03-22 19:36:35 -0400394 if (list_empty(&sb->s_instances))
395 continue;
Christoph Hellwige5004752009-05-05 16:08:56 +0200396 if (sb->s_op->write_super && sb->s_dirt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 sb->s_count++;
398 spin_unlock(&sb_lock);
Christoph Hellwige5004752009-05-05 16:08:56 +0200399
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 down_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200401 if (sb->s_root && sb->s_dirt)
402 sb->s_op->write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700403 up_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200404
Kirill Korotaev618f0632005-06-23 00:09:54 -0700405 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400406 if (p)
407 __put_super(p);
408 p = sb;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700409 }
410 }
Al Virodca33252010-07-25 02:31:46 +0400411 if (p)
412 __put_super(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 spin_unlock(&sb_lock);
414}
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416/**
Al Viro01a05b32010-03-23 06:06:58 -0400417 * iterate_supers - call function for all active superblocks
418 * @f: function to call
419 * @arg: argument to pass to it
420 *
421 * Scans the superblock list and calls given function, passing it
422 * locked superblock and given argument.
423 */
424void iterate_supers(void (*f)(struct super_block *, void *), void *arg)
425{
Al Virodca33252010-07-25 02:31:46 +0400426 struct super_block *sb, *p = NULL;
Al Viro01a05b32010-03-23 06:06:58 -0400427
428 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400429 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro01a05b32010-03-23 06:06:58 -0400430 if (list_empty(&sb->s_instances))
431 continue;
432 sb->s_count++;
433 spin_unlock(&sb_lock);
434
435 down_read(&sb->s_umount);
436 if (sb->s_root)
437 f(sb, arg);
438 up_read(&sb->s_umount);
439
440 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400441 if (p)
442 __put_super(p);
443 p = sb;
Al Viro01a05b32010-03-23 06:06:58 -0400444 }
Al Virodca33252010-07-25 02:31:46 +0400445 if (p)
446 __put_super(p);
Al Viro01a05b32010-03-23 06:06:58 -0400447 spin_unlock(&sb_lock);
448}
449
450/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 * get_super - get the superblock of a device
452 * @bdev: device to get the superblock for
453 *
454 * Scans the superblock list and finds the superblock of the file system
455 * mounted on the device given. %NULL is returned if no match is found.
456 */
457
Al Virodf40c012010-03-22 20:23:25 -0400458struct super_block *get_super(struct block_device *bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700460 struct super_block *sb;
461
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 if (!bdev)
463 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700464
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700466rescan:
467 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro551de6f2010-03-22 19:36:35 -0400468 if (list_empty(&sb->s_instances))
469 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700470 if (sb->s_bdev == bdev) {
471 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700473 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400474 /* still alive? */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700475 if (sb->s_root)
476 return sb;
477 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400478 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700479 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400480 __put_super(sb);
481 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 }
483 }
484 spin_unlock(&sb_lock);
485 return NULL;
486}
487
488EXPORT_SYMBOL(get_super);
Christoph Hellwig45042302009-08-03 23:28:35 +0200489
490/**
491 * get_active_super - get an active reference to the superblock of a device
492 * @bdev: device to get the superblock for
493 *
494 * Scans the superblock list and finds the superblock of the file system
495 * mounted on the device given. Returns the superblock with an active
Al Virod3f21472010-03-23 11:11:05 -0400496 * reference or %NULL if none was found.
Christoph Hellwig45042302009-08-03 23:28:35 +0200497 */
498struct super_block *get_active_super(struct block_device *bdev)
499{
500 struct super_block *sb;
501
502 if (!bdev)
503 return NULL;
504
Al Viro14945832010-03-22 20:15:33 -0400505restart:
Christoph Hellwig45042302009-08-03 23:28:35 +0200506 spin_lock(&sb_lock);
507 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro551de6f2010-03-22 19:36:35 -0400508 if (list_empty(&sb->s_instances))
509 continue;
Al Viro14945832010-03-22 20:15:33 -0400510 if (sb->s_bdev == bdev) {
511 if (grab_super(sb)) /* drops sb_lock */
512 return sb;
513 else
514 goto restart;
515 }
Christoph Hellwig45042302009-08-03 23:28:35 +0200516 }
517 spin_unlock(&sb_lock);
518 return NULL;
519}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520
Al Virodf40c012010-03-22 20:23:25 -0400521struct super_block *user_get_super(dev_t dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700523 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700526rescan:
527 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro551de6f2010-03-22 19:36:35 -0400528 if (list_empty(&sb->s_instances))
529 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700530 if (sb->s_dev == dev) {
531 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700533 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400534 /* still alive? */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700535 if (sb->s_root)
536 return sb;
537 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400538 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700539 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400540 __put_super(sb);
541 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 }
543 }
544 spin_unlock(&sb_lock);
545 return NULL;
546}
547
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 * do_remount_sb - asks filesystem to change mount options.
550 * @sb: superblock in question
551 * @flags: numeric part of options
552 * @data: the rest of options
553 * @force: whether or not to force the change
554 *
555 * Alters the mount options of a mounted file system.
556 */
557int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
558{
559 int retval;
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400560 int remount_ro;
Christoph Hellwig45042302009-08-03 23:28:35 +0200561
562 if (sb->s_frozen != SB_UNFROZEN)
563 return -EBUSY;
564
David Howells93614012006-09-30 20:45:40 +0200565#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
567 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200568#endif
Christoph Hellwig45042302009-08-03 23:28:35 +0200569
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 if (flags & MS_RDONLY)
571 acct_auto_close(sb);
572 shrink_dcache_sb(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200573 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574
Nick Piggind208bbd2009-12-21 16:28:53 -0800575 remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY);
Nick Piggind208bbd2009-12-21 16:28:53 -0800576
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 /* If we are remounting RDONLY and current sb is read/write,
578 make sure there are no rw files opened */
Nick Piggind208bbd2009-12-21 16:28:53 -0800579 if (remount_ro) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 if (force)
581 mark_files_ro(sb);
J. R. Okajimab0895512009-06-17 01:16:50 +0900582 else if (!fs_may_remount_ro(sb))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 return -EBUSY;
584 }
585
586 if (sb->s_op->remount_fs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 retval = sb->s_op->remount_fs(sb, &flags, data);
J. R. Okajimab0895512009-06-17 01:16:50 +0900588 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 return retval;
590 }
591 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400592
Nick Piggind208bbd2009-12-21 16:28:53 -0800593 /*
594 * Some filesystems modify their metadata via some other path than the
595 * bdev buffer cache (eg. use a private mapping, or directories in
596 * pagecache, etc). Also file data modifications go via their own
597 * mappings. So If we try to mount readonly then copy the filesystem
598 * from bdev, we could get stale data, so invalidate it to give a best
599 * effort at coherency.
600 */
601 if (remount_ro && sb->s_bdev)
602 invalidate_bdev(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 return 0;
604}
605
Jens Axboea2a95372009-03-17 09:38:40 +0100606static void do_emergency_remount(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607{
Al Virodca33252010-07-25 02:31:46 +0400608 struct super_block *sb, *p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
610 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400611 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro551de6f2010-03-22 19:36:35 -0400612 if (list_empty(&sb->s_instances))
613 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 sb->s_count++;
615 spin_unlock(&sb_lock);
Al Viro443b94b2009-05-05 23:48:50 -0400616 down_write(&sb->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
618 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 * What lock protects sb->s_flags??
620 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 do_remount_sb(sb, MS_RDONLY, NULL, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 }
Al Viro443b94b2009-05-05 23:48:50 -0400623 up_write(&sb->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400625 if (p)
626 __put_super(p);
627 p = sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 }
Al Virodca33252010-07-25 02:31:46 +0400629 if (p)
630 __put_super(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 spin_unlock(&sb_lock);
Jens Axboea2a95372009-03-17 09:38:40 +0100632 kfree(work);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 printk("Emergency Remount complete\n");
634}
635
636void emergency_remount(void)
637{
Jens Axboea2a95372009-03-17 09:38:40 +0100638 struct work_struct *work;
639
640 work = kmalloc(sizeof(*work), GFP_ATOMIC);
641 if (work) {
642 INIT_WORK(work, do_emergency_remount);
643 schedule_work(work);
644 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645}
646
647/*
648 * Unnamed block devices are dummy devices used by virtual
649 * filesystems which don't use real block-devices. -- jrs
650 */
651
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400652static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
Al Viroc63e09e2009-06-24 02:05:18 -0400654static int unnamed_dev_start = 0; /* don't bother trying below it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
656int set_anon_super(struct super_block *s, void *data)
657{
658 int dev;
659 int error;
660
661 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400662 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 return -ENOMEM;
664 spin_lock(&unnamed_dev_lock);
Al Viroc63e09e2009-06-24 02:05:18 -0400665 error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
Al Virof21f6222009-06-24 03:12:00 -0400666 if (!error)
667 unnamed_dev_start = dev + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 spin_unlock(&unnamed_dev_lock);
669 if (error == -EAGAIN)
670 /* We raced and lost with another CPU. */
671 goto retry;
672 else if (error)
673 return -EAGAIN;
674
675 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
676 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400677 ida_remove(&unnamed_dev_ida, dev);
Al Virof21f6222009-06-24 03:12:00 -0400678 if (unnamed_dev_start > dev)
679 unnamed_dev_start = dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 spin_unlock(&unnamed_dev_lock);
681 return -EMFILE;
682 }
683 s->s_dev = MKDEV(0, dev & MINORMASK);
Jörn Engel5129a462010-04-25 08:54:42 +0200684 s->s_bdi = &noop_backing_dev_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 return 0;
686}
687
688EXPORT_SYMBOL(set_anon_super);
689
690void kill_anon_super(struct super_block *sb)
691{
692 int slot = MINOR(sb->s_dev);
693
694 generic_shutdown_super(sb);
695 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400696 ida_remove(&unnamed_dev_ida, slot);
Al Viroc63e09e2009-06-24 02:05:18 -0400697 if (slot < unnamed_dev_start)
698 unnamed_dev_start = slot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 spin_unlock(&unnamed_dev_lock);
700}
701
702EXPORT_SYMBOL(kill_anon_super);
703
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704void kill_litter_super(struct super_block *sb)
705{
706 if (sb->s_root)
707 d_genocide(sb->s_root);
708 kill_anon_super(sb);
709}
710
711EXPORT_SYMBOL(kill_litter_super);
712
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700713static int ns_test_super(struct super_block *sb, void *data)
714{
715 return sb->s_fs_info == data;
716}
717
718static int ns_set_super(struct super_block *sb, void *data)
719{
720 sb->s_fs_info = data;
721 return set_anon_super(sb, NULL);
722}
723
Al Viroceefda62010-07-26 13:16:50 +0400724struct dentry *mount_ns(struct file_system_type *fs_type, int flags,
725 void *data, int (*fill_super)(struct super_block *, void *, int))
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700726{
727 struct super_block *sb;
728
729 sb = sget(fs_type, ns_test_super, ns_set_super, data);
730 if (IS_ERR(sb))
Al Viroceefda62010-07-26 13:16:50 +0400731 return ERR_CAST(sb);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700732
733 if (!sb->s_root) {
734 int err;
735 sb->s_flags = flags;
736 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
737 if (err) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400738 deactivate_locked_super(sb);
Al Viroceefda62010-07-26 13:16:50 +0400739 return ERR_PTR(err);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700740 }
741
742 sb->s_flags |= MS_ACTIVE;
743 }
744
Al Viroceefda62010-07-26 13:16:50 +0400745 return dget(sb->s_root);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700746}
747
Al Viroceefda62010-07-26 13:16:50 +0400748EXPORT_SYMBOL(mount_ns);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700749
David Howells93614012006-09-30 20:45:40 +0200750#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751static int set_bdev_super(struct super_block *s, void *data)
752{
753 s->s_bdev = data;
754 s->s_dev = s->s_bdev->bd_dev;
Jens Axboe32a88aa2009-09-16 15:02:33 +0200755
756 /*
757 * We set the bdi here to the queue backing, file systems can
758 * overwrite this in ->fill_super()
759 */
760 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 return 0;
762}
763
764static int test_bdev_super(struct super_block *s, void *data)
765{
766 return (void *)s->s_bdev == data;
767}
768
Al Viro152a0832010-07-25 00:46:55 +0400769struct dentry *mount_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 int flags, const char *dev_name, void *data,
Al Viro152a0832010-07-25 00:46:55 +0400771 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772{
773 struct block_device *bdev;
774 struct super_block *s;
Tejun Heod4d77622010-11-13 11:55:18 +0100775 fmode_t mode = FMODE_READ | FMODE_EXCL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 int error = 0;
777
Al Viro30c40d22008-02-22 19:50:45 -0500778 if (!(flags & MS_RDONLY))
779 mode |= FMODE_WRITE;
780
Tejun Heod4d77622010-11-13 11:55:18 +0100781 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 if (IS_ERR(bdev))
Al Viro152a0832010-07-25 00:46:55 +0400783 return ERR_CAST(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784
785 /*
786 * once the super is inserted into the list by sget, s_umount
787 * will protect the lockfs code from trying to start a snapshot
788 * while we are mounting
789 */
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200790 mutex_lock(&bdev->bd_fsfreeze_mutex);
791 if (bdev->bd_fsfreeze_count > 0) {
792 mutex_unlock(&bdev->bd_fsfreeze_mutex);
793 error = -EBUSY;
794 goto error_bdev;
795 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200797 mutex_unlock(&bdev->bd_fsfreeze_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700799 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
801 if (s->s_root) {
802 if ((flags ^ s->s_flags) & MS_RDONLY) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400803 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700804 error = -EBUSY;
805 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 }
David Howells454e2392006-06-23 02:02:57 -0700807
Tejun Heo4f331f02010-07-20 15:18:07 -0700808 /*
809 * s_umount nests inside bd_mutex during
Tejun Heoe525fd82010-11-13 11:55:17 +0100810 * __invalidate_device(). blkdev_put() acquires
811 * bd_mutex and can't be called under s_umount. Drop
812 * s_umount temporarily. This is safe as we're
813 * holding an active reference.
Tejun Heo4f331f02010-07-20 15:18:07 -0700814 */
815 up_write(&s->s_umount);
Tejun Heod4d77622010-11-13 11:55:18 +0100816 blkdev_put(bdev, mode);
Tejun Heo4f331f02010-07-20 15:18:07 -0700817 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 } else {
819 char b[BDEVNAME_SIZE];
820
821 s->s_flags = flags;
Al Viro30c40d22008-02-22 19:50:45 -0500822 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800824 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800825 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400827 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700828 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800829 }
David Howells454e2392006-06-23 02:02:57 -0700830
831 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -0500832 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 }
834
Al Viro152a0832010-07-25 00:46:55 +0400835 return dget(s->s_root);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836
David Howells454e2392006-06-23 02:02:57 -0700837error_s:
838 error = PTR_ERR(s);
839error_bdev:
Tejun Heod4d77622010-11-13 11:55:18 +0100840 blkdev_put(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -0700841error:
Al Viro152a0832010-07-25 00:46:55 +0400842 return ERR_PTR(error);
843}
844EXPORT_SYMBOL(mount_bdev);
845
846int get_sb_bdev(struct file_system_type *fs_type,
847 int flags, const char *dev_name, void *data,
848 int (*fill_super)(struct super_block *, void *, int),
849 struct vfsmount *mnt)
850{
851 struct dentry *root;
852
853 root = mount_bdev(fs_type, flags, dev_name, data, fill_super);
854 if (IS_ERR(root))
855 return PTR_ERR(root);
856 mnt->mnt_root = root;
857 mnt->mnt_sb = root->d_sb;
858 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859}
860
861EXPORT_SYMBOL(get_sb_bdev);
862
863void kill_block_super(struct super_block *sb)
864{
865 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -0500866 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867
H Hartley Sweetenddbaaf32009-04-29 20:14:57 -0400868 bdev->bd_super = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 generic_shutdown_super(sb);
870 sync_blockdev(bdev);
Tejun Heod4d77622010-11-13 11:55:18 +0100871 WARN_ON_ONCE(!(mode & FMODE_EXCL));
Tejun Heoe525fd82010-11-13 11:55:17 +0100872 blkdev_put(bdev, mode | FMODE_EXCL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873}
874
875EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200876#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877
Al Viro3c26ff62010-07-25 11:46:36 +0400878struct dentry *mount_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 int flags, void *data,
Al Viro3c26ff62010-07-25 11:46:36 +0400880 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881{
882 int error;
883 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
884
885 if (IS_ERR(s))
Al Viro3c26ff62010-07-25 11:46:36 +0400886 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887
888 s->s_flags = flags;
889
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800890 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400892 deactivate_locked_super(s);
Al Viro3c26ff62010-07-25 11:46:36 +0400893 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 }
895 s->s_flags |= MS_ACTIVE;
Al Viro3c26ff62010-07-25 11:46:36 +0400896 return dget(s->s_root);
897}
898EXPORT_SYMBOL(mount_nodev);
899
900int get_sb_nodev(struct file_system_type *fs_type,
901 int flags, void *data,
902 int (*fill_super)(struct super_block *, void *, int),
903 struct vfsmount *mnt)
904{
905 struct dentry *root;
906
907 root = mount_nodev(fs_type, flags, data, fill_super);
908 if (IS_ERR(root))
909 return PTR_ERR(root);
910 mnt->mnt_root = root;
911 mnt->mnt_sb = root->d_sb;
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800912 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914EXPORT_SYMBOL(get_sb_nodev);
915
916static int compare_single(struct super_block *s, void *p)
917{
918 return 1;
919}
920
Al Virofc14f2f2010-07-25 01:48:30 +0400921struct dentry *mount_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 int flags, void *data,
Al Virofc14f2f2010-07-25 01:48:30 +0400923 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924{
925 struct super_block *s;
926 int error;
927
928 s = sget(fs_type, compare_single, set_anon_super, NULL);
929 if (IS_ERR(s))
Al Virofc14f2f2010-07-25 01:48:30 +0400930 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 if (!s->s_root) {
932 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800933 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400935 deactivate_locked_super(s);
Al Virofc14f2f2010-07-25 01:48:30 +0400936 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 }
938 s->s_flags |= MS_ACTIVE;
Kay Sievers9329d1b2009-12-18 21:18:15 +0100939 } else {
940 do_remount_sb(s, flags, data, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 }
Al Virofc14f2f2010-07-25 01:48:30 +0400942 return dget(s->s_root);
943}
944EXPORT_SYMBOL(mount_single);
945
946int get_sb_single(struct file_system_type *fs_type,
947 int flags, void *data,
948 int (*fill_super)(struct super_block *, void *, int),
949 struct vfsmount *mnt)
950{
951 struct dentry *root;
952 root = mount_single(fs_type, flags, data, fill_super);
953 if (IS_ERR(root))
954 return PTR_ERR(root);
955 mnt->mnt_root = root;
956 mnt->mnt_sb = root->d_sb;
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800957 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958}
959
960EXPORT_SYMBOL(get_sb_single);
961
962struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400963vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 struct vfsmount *mnt;
Al Viroc96e41e2010-07-25 00:17:56 +0400966 struct dentry *root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700968 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969
970 if (!type)
971 return ERR_PTR(-ENODEV);
972
David Howells454e2392006-06-23 02:02:57 -0700973 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 mnt = alloc_vfsmnt(name);
975 if (!mnt)
976 goto out;
977
Al Viro80893522010-02-05 09:30:46 -0500978 if (flags & MS_KERNMOUNT)
979 mnt->mnt_flags = MNT_INTERNAL;
980
Eric Parise0007522008-03-05 10:31:54 -0500981 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700983 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985
Eric Parise0007522008-03-05 10:31:54 -0500986 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700987 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 }
990
Al Viroc96e41e2010-07-25 00:17:56 +0400991 if (type->mount) {
992 root = type->mount(type, flags, name, data);
993 if (IS_ERR(root)) {
994 error = PTR_ERR(root);
995 goto out_free_secdata;
996 }
997 mnt->mnt_root = root;
998 mnt->mnt_sb = root->d_sb;
999 } else {
1000 error = type->get_sb(type, flags, name, data, mnt);
1001 if (error < 0)
1002 goto out_free_secdata;
1003 }
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -07001004 BUG_ON(!mnt->mnt_sb);
Jörn Engel5129a462010-04-25 08:54:42 +02001005 WARN_ON(!mnt->mnt_sb->s_bdi);
Al Viro7a4dec52010-08-09 12:05:43 -04001006 mnt->mnt_sb->s_flags |= MS_BORN;
David Howells454e2392006-06-23 02:02:57 -07001007
Jörn Engel5129a462010-04-25 08:54:42 +02001008 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
1009 if (error)
1010 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -07001011
Jeff Layton42cb56a2009-09-18 13:05:53 -07001012 /*
1013 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
1014 * but s_maxbytes was an unsigned long long for many releases. Throw
1015 * this warning for a little while to try and catch filesystems that
1016 * violate this rule. This warning should be either removed or
1017 * converted to a BUG() in 2.6.34.
1018 */
1019 WARN((mnt->mnt_sb->s_maxbytes < 0), "%s set sb->s_maxbytes to "
1020 "negative value (%lld)\n", type->name, mnt->mnt_sb->s_maxbytes);
1021
David Howells454e2392006-06-23 02:02:57 -07001022 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -07001024 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -07001025 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 return mnt;
1027out_sb:
David Howells454e2392006-06-23 02:02:57 -07001028 dput(mnt->mnt_root);
Al Viro74dbbdd2009-05-06 01:07:50 -04001029 deactivate_locked_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030out_free_secdata:
1031 free_secdata(secdata);
1032out_mnt:
1033 free_vfsmnt(mnt);
1034out:
David Howells454e2392006-06-23 02:02:57 -07001035 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036}
1037
Trond Myklebustbb4a58b2006-06-09 09:34:15 -04001038EXPORT_SYMBOL_GPL(vfs_kern_mount);
1039
Josef Bacik18e9e512010-03-23 10:34:56 -04001040/**
Randy Dunlap7000d3c2010-05-24 22:22:34 -07001041 * freeze_super - lock the filesystem and force it into a consistent state
1042 * @sb: the super to lock
Josef Bacik18e9e512010-03-23 10:34:56 -04001043 *
1044 * Syncs the super to make sure the filesystem is consistent and calls the fs's
1045 * freeze_fs. Subsequent calls to this without first thawing the fs will return
1046 * -EBUSY.
1047 */
1048int freeze_super(struct super_block *sb)
1049{
1050 int ret;
1051
1052 atomic_inc(&sb->s_active);
1053 down_write(&sb->s_umount);
1054 if (sb->s_frozen) {
1055 deactivate_locked_super(sb);
1056 return -EBUSY;
1057 }
1058
1059 if (sb->s_flags & MS_RDONLY) {
1060 sb->s_frozen = SB_FREEZE_TRANS;
1061 smp_wmb();
1062 up_write(&sb->s_umount);
1063 return 0;
1064 }
1065
1066 sb->s_frozen = SB_FREEZE_WRITE;
1067 smp_wmb();
1068
1069 sync_filesystem(sb);
1070
1071 sb->s_frozen = SB_FREEZE_TRANS;
1072 smp_wmb();
1073
1074 sync_blockdev(sb->s_bdev);
1075 if (sb->s_op->freeze_fs) {
1076 ret = sb->s_op->freeze_fs(sb);
1077 if (ret) {
1078 printk(KERN_ERR
1079 "VFS:Filesystem freeze failed\n");
1080 sb->s_frozen = SB_UNFROZEN;
1081 deactivate_locked_super(sb);
1082 return ret;
1083 }
1084 }
1085 up_write(&sb->s_umount);
1086 return 0;
1087}
1088EXPORT_SYMBOL(freeze_super);
1089
1090/**
1091 * thaw_super -- unlock filesystem
1092 * @sb: the super to thaw
1093 *
1094 * Unlocks the filesystem and marks it writeable again after freeze_super().
1095 */
1096int thaw_super(struct super_block *sb)
1097{
1098 int error;
1099
1100 down_write(&sb->s_umount);
1101 if (sb->s_frozen == SB_UNFROZEN) {
1102 up_write(&sb->s_umount);
1103 return -EINVAL;
1104 }
1105
1106 if (sb->s_flags & MS_RDONLY)
1107 goto out;
1108
1109 if (sb->s_op->unfreeze_fs) {
1110 error = sb->s_op->unfreeze_fs(sb);
1111 if (error) {
1112 printk(KERN_ERR
1113 "VFS:Filesystem thaw failed\n");
1114 sb->s_frozen = SB_FREEZE_TRANS;
1115 up_write(&sb->s_umount);
1116 return error;
1117 }
1118 }
1119
1120out:
1121 sb->s_frozen = SB_UNFROZEN;
1122 smp_wmb();
1123 wake_up(&sb->s_wait_unfrozen);
1124 deactivate_locked_super(sb);
1125
1126 return 0;
1127}
1128EXPORT_SYMBOL(thaw_super);
1129
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -07001130static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
1131{
1132 int err;
1133 const char *subtype = strchr(fstype, '.');
1134 if (subtype) {
1135 subtype++;
1136 err = -EINVAL;
1137 if (!subtype[0])
1138 goto err;
1139 } else
1140 subtype = "";
1141
1142 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
1143 err = -ENOMEM;
1144 if (!mnt->mnt_sb->s_subtype)
1145 goto err;
1146 return mnt;
1147
1148 err:
Al Virof03c6592011-01-14 22:30:21 -05001149 mntput(mnt);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -07001150 return ERR_PTR(err);
1151}
1152
Trond Myklebustbb4a58b2006-06-09 09:34:15 -04001153struct vfsmount *
1154do_kern_mount(const char *fstype, int flags, const char *name, void *data)
1155{
1156 struct file_system_type *type = get_fs_type(fstype);
1157 struct vfsmount *mnt;
1158 if (!type)
1159 return ERR_PTR(-ENODEV);
1160 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -07001161 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
1162 !mnt->mnt_sb->s_subtype)
1163 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -04001164 put_filesystem(type);
1165 return mnt;
1166}
Al Viro8a4e98d2008-02-24 01:43:03 -05001167EXPORT_SYMBOL_GPL(do_kern_mount);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -04001168
Pavel Emelyanov8bf97252007-10-18 23:40:02 -07001169struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -07001171 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172}
1173
Pavel Emelyanov8bf97252007-10-18 23:40:02 -07001174EXPORT_SYMBOL_GPL(kern_mount_data);