blob: 80d5cf2ca76518a450a31cb7cad340794a525ed8 [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
Paul Gortmaker630d9c42011-11-16 23:57:37 -050023#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#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>
Dan Magenheimerc515e1f2011-05-26 10:01:43 -060034#include <linux/cleancache.h>
Al Viro40401532012-02-13 03:58:52 +000035#include <linux/fsnotify.h>
Jan Kara5accdf82012-06-12 16:20:34 +020036#include <linux/lockdep.h>
Al Viro6d59e7f2008-03-22 15:48:17 -040037#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
39
Linus Torvalds1da177e2005-04-16 15:20:36 -070040LIST_HEAD(super_blocks);
41DEFINE_SPINLOCK(sb_lock);
42
Jan Kara5accdf82012-06-12 16:20:34 +020043static char *sb_writers_name[SB_FREEZE_LEVELS] = {
44 "sb_writers",
45 "sb_pagefaults",
46 "sb_internal",
47};
48
Dave Chinnerb0d40c92011-07-08 14:14:42 +100049/*
50 * One thing we have to be careful of with a per-sb shrinker is that we don't
51 * drop the last active reference to the superblock from within the shrinker.
52 * If that happens we could trigger unregistering the shrinker from within the
53 * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we
54 * take a passive reference to the superblock to avoid this from occurring.
55 */
Dave Chinner0a234c62013-08-28 10:17:57 +100056static unsigned long super_cache_scan(struct shrinker *shrink,
57 struct shrink_control *sc)
Dave Chinnerb0d40c92011-07-08 14:14:42 +100058{
59 struct super_block *sb;
Dave Chinner0a234c62013-08-28 10:17:57 +100060 long fs_objects = 0;
61 long total_objects;
62 long freed = 0;
63 long dentries;
64 long inodes;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100065
66 sb = container_of(shrink, struct super_block, s_shrink);
67
68 /*
69 * Deadlock avoidance. We may hold various FS locks, and we don't want
70 * to recurse into the FS that called us in clear_inode() and friends..
71 */
Dave Chinner0a234c62013-08-28 10:17:57 +100072 if (!(sc->gfp_mask & __GFP_FS))
73 return SHRINK_STOP;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100074
75 if (!grab_super_passive(sb))
Dave Chinner0a234c62013-08-28 10:17:57 +100076 return SHRINK_STOP;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100077
Al Virod0407902013-07-19 21:12:31 +040078 if (sb->s_op->nr_cached_objects)
Dave Chinner9b17c622013-08-28 10:18:05 +100079 fs_objects = sb->s_op->nr_cached_objects(sb, sc->nid);
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100080
Dave Chinner9b17c622013-08-28 10:18:05 +100081 inodes = list_lru_count_node(&sb->s_inode_lru, sc->nid);
82 dentries = list_lru_count_node(&sb->s_dentry_lru, sc->nid);
Dave Chinnerf6041562013-08-28 10:18:00 +100083 total_objects = dentries + inodes + fs_objects + 1;
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100084
Dave Chinner0a234c62013-08-28 10:17:57 +100085 /* proportion the scan between the caches */
Dave Chinnerf6041562013-08-28 10:18:00 +100086 dentries = mult_frac(sc->nr_to_scan, dentries, total_objects);
Dave Chinnerbc3b14c2013-08-28 10:17:58 +100087 inodes = mult_frac(sc->nr_to_scan, inodes, total_objects);
Dave Chinnerb0d40c92011-07-08 14:14:42 +100088
Dave Chinner0a234c62013-08-28 10:17:57 +100089 /*
90 * prune the dcache first as the icache is pinned by it, then
91 * prune the icache, followed by the filesystem specific caches
92 */
Dave Chinner9b17c622013-08-28 10:18:05 +100093 freed = prune_dcache_sb(sb, dentries, sc->nid);
94 freed += prune_icache_sb(sb, inodes, sc->nid);
Dave Chinnerb0d40c92011-07-08 14:14:42 +100095
Dave Chinner0a234c62013-08-28 10:17:57 +100096 if (fs_objects) {
97 fs_objects = mult_frac(sc->nr_to_scan, fs_objects,
98 total_objects);
Dave Chinner9b17c622013-08-28 10:18:05 +100099 freed += sb->s_op->free_cached_objects(sb, fs_objects,
100 sc->nid);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000101 }
102
Dave Chinner0a234c62013-08-28 10:17:57 +1000103 drop_super(sb);
104 return freed;
105}
106
107static unsigned long super_cache_count(struct shrinker *shrink,
108 struct shrink_control *sc)
109{
110 struct super_block *sb;
111 long total_objects = 0;
112
113 sb = container_of(shrink, struct super_block, s_shrink);
114
115 if (!grab_super_passive(sb))
116 return 0;
117
118 if (sb->s_op && sb->s_op->nr_cached_objects)
Dave Chinner9b17c622013-08-28 10:18:05 +1000119 total_objects = sb->s_op->nr_cached_objects(sb,
120 sc->nid);
Dave Chinner0a234c62013-08-28 10:17:57 +1000121
Dave Chinner9b17c622013-08-28 10:18:05 +1000122 total_objects += list_lru_count_node(&sb->s_dentry_lru,
123 sc->nid);
124 total_objects += list_lru_count_node(&sb->s_inode_lru,
125 sc->nid);
Dave Chinner0a234c62013-08-28 10:17:57 +1000126
Glauber Costa55f841c2013-08-28 10:17:53 +1000127 total_objects = vfs_pressure_ratio(total_objects);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000128 drop_super(sb);
Dave Chinner0e1fdaf2011-07-08 14:14:44 +1000129 return total_objects;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000130}
131
Al Viro7eb5e882013-10-01 15:09:58 -0400132/**
133 * destroy_super - frees a superblock
134 * @s: superblock to free
135 *
136 * Frees a superblock.
137 */
138static void destroy_super(struct super_block *s)
Jan Kara5accdf82012-06-12 16:20:34 +0200139{
140 int i;
Al Viro7eb5e882013-10-01 15:09:58 -0400141 list_lru_destroy(&s->s_dentry_lru);
142 list_lru_destroy(&s->s_inode_lru);
Jan Kara5accdf82012-06-12 16:20:34 +0200143 for (i = 0; i < SB_FREEZE_LEVELS; i++)
144 percpu_counter_destroy(&s->s_writers.counter[i]);
Al Viro7eb5e882013-10-01 15:09:58 -0400145 security_sb_free(s);
146 WARN_ON(!list_empty(&s->s_mounts));
147 kfree(s->s_subtype);
148 kfree(s->s_options);
Al Viroe2fec7c2013-10-04 17:06:56 -0400149 kfree_rcu(s, rcu);
Jan Kara5accdf82012-06-12 16:20:34 +0200150}
151
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152/**
153 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -0700154 * @type: filesystem type superblock should belong to
David Howells9249e172012-06-25 12:55:37 +0100155 * @flags: the mount flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 *
157 * Allocates and initializes a new &struct super_block. alloc_super()
158 * returns a pointer new superblock or %NULL if allocation had failed.
159 */
David Howells9249e172012-06-25 12:55:37 +0100160static struct super_block *alloc_super(struct file_system_type *type, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -0800162 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700163 static const struct super_operations default_op;
Al Viro7eb5e882013-10-01 15:09:58 -0400164 int i;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165
Al Viro7eb5e882013-10-01 15:09:58 -0400166 if (!s)
167 return NULL;
168
Vladimir Davydovb5bd8562014-01-21 15:48:45 -0800169 INIT_LIST_HEAD(&s->s_mounts);
170
Al Viro7eb5e882013-10-01 15:09:58 -0400171 if (security_sb_alloc(s))
172 goto fail;
Christoph Hellwig7b7a8662013-09-04 15:04:39 +0200173
Al Viro7eb5e882013-10-01 15:09:58 -0400174 for (i = 0; i < SB_FREEZE_LEVELS; i++) {
175 if (percpu_counter_init(&s->s_writers.counter[i], 0) < 0)
176 goto fail;
177 lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i],
178 &type->s_writers_key[i], 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 }
Al Viro7eb5e882013-10-01 15:09:58 -0400180 init_waitqueue_head(&s->s_writers.wait);
181 init_waitqueue_head(&s->s_writers.wait_unfrozen);
182 s->s_flags = flags;
183 s->s_bdi = &default_backing_dev_info;
184 INIT_HLIST_NODE(&s->s_instances);
185 INIT_HLIST_BL_HEAD(&s->s_anon);
186 INIT_LIST_HEAD(&s->s_inodes);
187
188 if (list_lru_init(&s->s_dentry_lru))
189 goto fail;
190 if (list_lru_init(&s->s_inode_lru))
191 goto fail;
192
Al Viro7eb5e882013-10-01 15:09:58 -0400193 init_rwsem(&s->s_umount);
194 lockdep_set_class(&s->s_umount, &type->s_umount_key);
195 /*
196 * sget() can have s_umount recursion.
197 *
198 * When it cannot find a suitable sb, it allocates a new
199 * one (this one), and tries again to find a suitable old
200 * one.
201 *
202 * In case that succeeds, it will acquire the s_umount
203 * lock of the old one. Since these are clearly distrinct
204 * locks, and this object isn't exposed yet, there's no
205 * risk of deadlocks.
206 *
207 * Annotate this by putting this lock in a different
208 * subclass.
209 */
210 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
211 s->s_count = 1;
212 atomic_set(&s->s_active, 1);
213 mutex_init(&s->s_vfs_rename_mutex);
214 lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key);
215 mutex_init(&s->s_dquot.dqio_mutex);
216 mutex_init(&s->s_dquot.dqonoff_mutex);
217 init_rwsem(&s->s_dquot.dqptr_sem);
218 s->s_maxbytes = MAX_NON_LFS;
219 s->s_op = &default_op;
220 s->s_time_gran = 1000000000;
221 s->cleancache_poolid = -1;
222
223 s->s_shrink.seeks = DEFAULT_SEEKS;
224 s->s_shrink.scan_objects = super_cache_scan;
225 s->s_shrink.count_objects = super_cache_count;
226 s->s_shrink.batch = 1024;
227 s->s_shrink.flags = SHRINKER_NUMA_AWARE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 return s;
Glauber Costa5ca302c2013-08-28 10:18:18 +1000229
Al Viro7eb5e882013-10-01 15:09:58 -0400230fail:
231 destroy_super(s);
232 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233}
234
235/* Superblock refcounting */
236
237/*
Al Viro35cf7ba2010-03-22 21:13:53 -0400238 * Drop a superblock's refcount. The caller must hold sb_lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 */
Al Virof47ec3f2011-11-21 21:15:42 -0500240static void __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 if (!--sb->s_count) {
Al Viro551de6f2010-03-22 19:36:35 -0400243 list_del_init(&sb->s_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 destroy_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246}
247
248/**
249 * put_super - drop a temporary reference to superblock
250 * @sb: superblock in question
251 *
252 * Drops a temporary reference, frees superblock if there's no
253 * references left.
254 */
Al Virof47ec3f2011-11-21 21:15:42 -0500255static void put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256{
257 spin_lock(&sb_lock);
258 __put_super(sb);
259 spin_unlock(&sb_lock);
260}
261
262
263/**
Al Viro74dbbdd2009-05-06 01:07:50 -0400264 * deactivate_locked_super - drop an active reference to superblock
265 * @s: superblock to deactivate
266 *
Al Viro1712ac82010-03-22 15:22:31 -0400267 * Drops an active reference to superblock, converting it into a temprory
268 * one if there is no other active references left. In that case we
269 * tell fs driver to shut it down and drop the temporary reference we
270 * had just acquired.
271 *
272 * Caller holds exclusive lock on superblock; that lock is released.
Al Viro74dbbdd2009-05-06 01:07:50 -0400273 */
274void deactivate_locked_super(struct super_block *s)
275{
276 struct file_system_type *fs = s->s_type;
Al Virob20bd1a2010-03-22 08:53:19 -0400277 if (atomic_dec_and_test(&s->s_active)) {
Dan Magenheimer31677602011-09-21 11:56:28 -0400278 cleancache_invalidate_fs(s);
Al Viro74dbbdd2009-05-06 01:07:50 -0400279 fs->kill_sb(s);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000280
281 /* caches are now gone, we can safely kill the shrinker now */
282 unregister_shrinker(&s->s_shrink);
Glauber Costaf5e1dd32013-08-28 10:18:18 +1000283
Al Viro74dbbdd2009-05-06 01:07:50 -0400284 put_filesystem(fs);
285 put_super(s);
286 } else {
287 up_write(&s->s_umount);
288 }
289}
290
291EXPORT_SYMBOL(deactivate_locked_super);
292
293/**
Al Viro1712ac82010-03-22 15:22:31 -0400294 * deactivate_super - drop an active reference to superblock
295 * @s: superblock to deactivate
296 *
297 * Variant of deactivate_locked_super(), except that superblock is *not*
298 * locked by caller. If we are going to drop the final active reference,
299 * lock will be acquired prior to that.
300 */
301void deactivate_super(struct super_block *s)
302{
303 if (!atomic_add_unless(&s->s_active, -1, 1)) {
304 down_write(&s->s_umount);
305 deactivate_locked_super(s);
306 }
307}
308
309EXPORT_SYMBOL(deactivate_super);
310
311/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 * grab_super - acquire an active reference
313 * @s: reference we are trying to make active
314 *
315 * Tries to acquire an active reference. grab_super() is used when we
316 * had just found a superblock in super_blocks or fs_type->fs_supers
317 * and want to turn it into a full-blown active reference. grab_super()
318 * is called with sb_lock held and drops it. Returns 1 in case of
319 * success, 0 if we had failed (superblock contents was already dead or
Al Viroacfec9a2013-07-20 03:13:55 +0400320 * dying when grab_super() had been called). Note that this is only
321 * called for superblocks not in rundown mode (== ones still on ->fs_supers
322 * of their type), so increment of ->s_count is OK here.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700324static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325{
326 s->s_count++;
327 spin_unlock(&sb_lock);
328 down_write(&s->s_umount);
Al Viroacfec9a2013-07-20 03:13:55 +0400329 if ((s->s_flags & MS_BORN) && atomic_inc_not_zero(&s->s_active)) {
330 put_super(s);
331 return 1;
332 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 up_write(&s->s_umount);
334 put_super(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 return 0;
336}
337
David Howellscf9a2ae2006-08-29 19:05:54 +0100338/*
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000339 * grab_super_passive - acquire a passive reference
Wanpeng Li331cbde2012-06-09 11:10:55 +0800340 * @sb: reference we are trying to grab
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000341 *
342 * Tries to acquire a passive reference. This is used in places where we
343 * cannot take an active reference but we need to ensure that the
344 * superblock does not go away while we are working on it. It returns
345 * false if a reference was not gained, and returns true with the s_umount
346 * lock held in read mode if a reference is gained. On successful return,
347 * the caller must drop the s_umount lock and the passive reference when
348 * done.
349 */
350bool grab_super_passive(struct super_block *sb)
351{
352 spin_lock(&sb_lock);
Al Viroa5166162011-12-12 22:53:00 -0500353 if (hlist_unhashed(&sb->s_instances)) {
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000354 spin_unlock(&sb_lock);
355 return false;
356 }
357
358 sb->s_count++;
359 spin_unlock(&sb_lock);
360
361 if (down_read_trylock(&sb->s_umount)) {
Al Virodabe0dc2012-01-03 21:01:29 -0500362 if (sb->s_root && (sb->s_flags & MS_BORN))
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000363 return true;
364 up_read(&sb->s_umount);
365 }
366
367 put_super(sb);
368 return false;
369}
370
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371/**
372 * generic_shutdown_super - common helper for ->kill_sb()
373 * @sb: superblock to kill
374 *
375 * generic_shutdown_super() does all fs-independent work on superblock
376 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
377 * that need destruction out of superblock, call generic_shutdown_super()
378 * and release aforementioned objects. Note: dentries and inodes _are_
379 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700380 *
381 * Upon calling this function, the filesystem may no longer alter or
382 * rearrange the set of dentries belonging to this super_block, nor may it
383 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 */
385void generic_shutdown_super(struct super_block *sb)
386{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800387 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
David Howellsc636ebd2006-10-11 01:22:19 -0700389 if (sb->s_root) {
390 shrink_dcache_for_umount(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200391 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 sb->s_flags &= ~MS_ACTIVE;
Arjan van de Venefaee192009-01-06 07:20:54 -0800393
Al Viro63997e92010-10-25 20:49:35 -0400394 fsnotify_unmount_inodes(&sb->s_inodes);
395
396 evict_inodes(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397
Christoph Hellwig7b7a8662013-09-04 15:04:39 +0200398 if (sb->s_dio_done_wq) {
399 destroy_workqueue(sb->s_dio_done_wq);
400 sb->s_dio_done_wq = NULL;
401 }
402
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 if (sop->put_super)
404 sop->put_super(sb);
405
Al Viro63997e92010-10-25 20:49:35 -0400406 if (!list_empty(&sb->s_inodes)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800407 printk("VFS: Busy inodes after unmount of %s. "
408 "Self-destruct in 5 seconds. Have a nice day...\n",
409 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 }
412 spin_lock(&sb_lock);
413 /* should be initialized for __put_super_and_need_restart() */
Al Viroa5166162011-12-12 22:53:00 -0500414 hlist_del_init(&sb->s_instances);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 spin_unlock(&sb_lock);
416 up_write(&sb->s_umount);
417}
418
419EXPORT_SYMBOL(generic_shutdown_super);
420
421/**
422 * sget - find or create a superblock
423 * @type: filesystem type superblock should belong to
424 * @test: comparison callback
425 * @set: setup callback
David Howells9249e172012-06-25 12:55:37 +0100426 * @flags: mount flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 * @data: argument to each of them
428 */
429struct super_block *sget(struct file_system_type *type,
430 int (*test)(struct super_block *,void *),
431 int (*set)(struct super_block *,void *),
David Howells9249e172012-06-25 12:55:37 +0100432 int flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 void *data)
434{
435 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700436 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 int err;
438
439retry:
440 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700441 if (test) {
Sasha Levinb67bfe02013-02-27 17:06:00 -0800442 hlist_for_each_entry(old, &type->fs_supers, s_instances) {
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700443 if (!test(old, data))
444 continue;
445 if (!grab_super(old))
446 goto retry;
Li Zefana3cfbb52009-03-12 14:31:29 -0700447 if (s) {
448 up_write(&s->s_umount);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700449 destroy_super(s);
Al Viro7a4dec52010-08-09 12:05:43 -0400450 s = NULL;
Li Zefana3cfbb52009-03-12 14:31:29 -0700451 }
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700452 return old;
453 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 }
455 if (!s) {
456 spin_unlock(&sb_lock);
David Howells9249e172012-06-25 12:55:37 +0100457 s = alloc_super(type, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 if (!s)
459 return ERR_PTR(-ENOMEM);
460 goto retry;
461 }
462
463 err = set(s, data);
464 if (err) {
465 spin_unlock(&sb_lock);
Li Zefana3cfbb52009-03-12 14:31:29 -0700466 up_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 destroy_super(s);
468 return ERR_PTR(err);
469 }
470 s->s_type = type;
471 strlcpy(s->s_id, type->name, sizeof(s->s_id));
472 list_add_tail(&s->s_list, &super_blocks);
Al Viroa5166162011-12-12 22:53:00 -0500473 hlist_add_head(&s->s_instances, &type->fs_supers);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 spin_unlock(&sb_lock);
475 get_filesystem(type);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000476 register_shrinker(&s->s_shrink);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 return s;
478}
479
480EXPORT_SYMBOL(sget);
481
482void drop_super(struct super_block *sb)
483{
484 up_read(&sb->s_umount);
485 put_super(sb);
486}
487
488EXPORT_SYMBOL(drop_super);
489
Christoph Hellwige5004752009-05-05 16:08:56 +0200490/**
Al Viro01a05b32010-03-23 06:06:58 -0400491 * iterate_supers - call function for all active superblocks
492 * @f: function to call
493 * @arg: argument to pass to it
494 *
495 * Scans the superblock list and calls given function, passing it
496 * locked superblock and given argument.
497 */
498void iterate_supers(void (*f)(struct super_block *, void *), void *arg)
499{
Al Virodca33252010-07-25 02:31:46 +0400500 struct super_block *sb, *p = NULL;
Al Viro01a05b32010-03-23 06:06:58 -0400501
502 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400503 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500504 if (hlist_unhashed(&sb->s_instances))
Al Viro01a05b32010-03-23 06:06:58 -0400505 continue;
506 sb->s_count++;
507 spin_unlock(&sb_lock);
508
509 down_read(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500510 if (sb->s_root && (sb->s_flags & MS_BORN))
Al Viro01a05b32010-03-23 06:06:58 -0400511 f(sb, arg);
512 up_read(&sb->s_umount);
513
514 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400515 if (p)
516 __put_super(p);
517 p = sb;
Al Viro01a05b32010-03-23 06:06:58 -0400518 }
Al Virodca33252010-07-25 02:31:46 +0400519 if (p)
520 __put_super(p);
Al Viro01a05b32010-03-23 06:06:58 -0400521 spin_unlock(&sb_lock);
522}
523
524/**
Al Viro43e15cd2011-06-03 20:16:57 -0400525 * iterate_supers_type - call function for superblocks of given type
526 * @type: fs type
527 * @f: function to call
528 * @arg: argument to pass to it
529 *
530 * Scans the superblock list and calls given function, passing it
531 * locked superblock and given argument.
532 */
533void iterate_supers_type(struct file_system_type *type,
534 void (*f)(struct super_block *, void *), void *arg)
535{
536 struct super_block *sb, *p = NULL;
537
538 spin_lock(&sb_lock);
Sasha Levinb67bfe02013-02-27 17:06:00 -0800539 hlist_for_each_entry(sb, &type->fs_supers, s_instances) {
Al Viro43e15cd2011-06-03 20:16:57 -0400540 sb->s_count++;
541 spin_unlock(&sb_lock);
542
543 down_read(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500544 if (sb->s_root && (sb->s_flags & MS_BORN))
Al Viro43e15cd2011-06-03 20:16:57 -0400545 f(sb, arg);
546 up_read(&sb->s_umount);
547
548 spin_lock(&sb_lock);
549 if (p)
550 __put_super(p);
551 p = sb;
552 }
553 if (p)
554 __put_super(p);
555 spin_unlock(&sb_lock);
556}
557
558EXPORT_SYMBOL(iterate_supers_type);
559
560/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 * get_super - get the superblock of a device
562 * @bdev: device to get the superblock for
563 *
564 * Scans the superblock list and finds the superblock of the file system
565 * mounted on the device given. %NULL is returned if no match is found.
566 */
567
Al Virodf40c012010-03-22 20:23:25 -0400568struct super_block *get_super(struct block_device *bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700570 struct super_block *sb;
571
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 if (!bdev)
573 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700574
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700576rescan:
577 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500578 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400579 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700580 if (sb->s_bdev == bdev) {
581 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700583 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400584 /* still alive? */
Al Virodabe0dc2012-01-03 21:01:29 -0500585 if (sb->s_root && (sb->s_flags & MS_BORN))
Kirill Korotaev618f0632005-06-23 00:09:54 -0700586 return sb;
587 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400588 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700589 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400590 __put_super(sb);
591 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 }
593 }
594 spin_unlock(&sb_lock);
595 return NULL;
596}
597
598EXPORT_SYMBOL(get_super);
Christoph Hellwig45042302009-08-03 23:28:35 +0200599
600/**
Jan Kara6b6dc832012-02-10 11:03:00 +0100601 * get_super_thawed - get thawed superblock of a device
602 * @bdev: device to get the superblock for
603 *
604 * Scans the superblock list and finds the superblock of the file system
605 * mounted on the device. The superblock is returned once it is thawed
606 * (or immediately if it was not frozen). %NULL is returned if no match
607 * is found.
608 */
609struct super_block *get_super_thawed(struct block_device *bdev)
610{
611 while (1) {
612 struct super_block *s = get_super(bdev);
Jan Kara5accdf82012-06-12 16:20:34 +0200613 if (!s || s->s_writers.frozen == SB_UNFROZEN)
Jan Kara6b6dc832012-02-10 11:03:00 +0100614 return s;
615 up_read(&s->s_umount);
Jan Kara5accdf82012-06-12 16:20:34 +0200616 wait_event(s->s_writers.wait_unfrozen,
617 s->s_writers.frozen == SB_UNFROZEN);
Jan Kara6b6dc832012-02-10 11:03:00 +0100618 put_super(s);
619 }
620}
621EXPORT_SYMBOL(get_super_thawed);
622
623/**
Christoph Hellwig45042302009-08-03 23:28:35 +0200624 * get_active_super - get an active reference to the superblock of a device
625 * @bdev: device to get the superblock for
626 *
627 * Scans the superblock list and finds the superblock of the file system
628 * mounted on the device given. Returns the superblock with an active
Al Virod3f21472010-03-23 11:11:05 -0400629 * reference or %NULL if none was found.
Christoph Hellwig45042302009-08-03 23:28:35 +0200630 */
631struct super_block *get_active_super(struct block_device *bdev)
632{
633 struct super_block *sb;
634
635 if (!bdev)
636 return NULL;
637
Al Viro14945832010-03-22 20:15:33 -0400638restart:
Christoph Hellwig45042302009-08-03 23:28:35 +0200639 spin_lock(&sb_lock);
640 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500641 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400642 continue;
Al Viro14945832010-03-22 20:15:33 -0400643 if (sb->s_bdev == bdev) {
Al Viroacfec9a2013-07-20 03:13:55 +0400644 if (!grab_super(sb))
Al Viro14945832010-03-22 20:15:33 -0400645 goto restart;
Al Viroacfec9a2013-07-20 03:13:55 +0400646 up_write(&sb->s_umount);
647 return sb;
Al Viro14945832010-03-22 20:15:33 -0400648 }
Christoph Hellwig45042302009-08-03 23:28:35 +0200649 }
650 spin_unlock(&sb_lock);
651 return NULL;
652}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Al Virodf40c012010-03-22 20:23:25 -0400654struct super_block *user_get_super(dev_t dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700656 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700659rescan:
660 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500661 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400662 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700663 if (sb->s_dev == dev) {
664 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700666 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400667 /* still alive? */
Al Virodabe0dc2012-01-03 21:01:29 -0500668 if (sb->s_root && (sb->s_flags & MS_BORN))
Kirill Korotaev618f0632005-06-23 00:09:54 -0700669 return sb;
670 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400671 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700672 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400673 __put_super(sb);
674 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 }
676 }
677 spin_unlock(&sb_lock);
678 return NULL;
679}
680
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 * do_remount_sb - asks filesystem to change mount options.
683 * @sb: superblock in question
684 * @flags: numeric part of options
685 * @data: the rest of options
686 * @force: whether or not to force the change
687 *
688 * Alters the mount options of a mounted file system.
689 */
690int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
691{
692 int retval;
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400693 int remount_ro;
Christoph Hellwig45042302009-08-03 23:28:35 +0200694
Jan Kara5accdf82012-06-12 16:20:34 +0200695 if (sb->s_writers.frozen != SB_UNFROZEN)
Christoph Hellwig45042302009-08-03 23:28:35 +0200696 return -EBUSY;
697
David Howells93614012006-09-30 20:45:40 +0200698#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
700 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200701#endif
Christoph Hellwig45042302009-08-03 23:28:35 +0200702
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 if (flags & MS_RDONLY)
704 acct_auto_close(sb);
705 shrink_dcache_sb(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706
Nick Piggind208bbd2009-12-21 16:28:53 -0800707 remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY);
Nick Piggind208bbd2009-12-21 16:28:53 -0800708
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 /* If we are remounting RDONLY and current sb is read/write,
710 make sure there are no rw files opened */
Nick Piggind208bbd2009-12-21 16:28:53 -0800711 if (remount_ro) {
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100712 if (force) {
Al Viroeee5cc22013-10-04 11:06:42 -0400713 sb->s_readonly_remount = 1;
714 smp_wmb();
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100715 } else {
716 retval = sb_prepare_remount_readonly(sb);
717 if (retval)
718 return retval;
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100719 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 }
721
Andrew Ruder807612d2014-01-30 09:26:54 -0600722 sync_filesystem(sb);
723
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 if (sb->s_op->remount_fs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 retval = sb->s_op->remount_fs(sb, &flags, data);
Miklos Szeredi2833eb22011-10-28 14:13:20 +0200726 if (retval) {
727 if (!force)
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100728 goto cancel_readonly;
Miklos Szeredi2833eb22011-10-28 14:13:20 +0200729 /* If forced remount, go ahead despite any errors */
730 WARN(1, "forced remount of a %s fs returned %i\n",
731 sb->s_type->name, retval);
732 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 }
734 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100735 /* Needs to be ordered wrt mnt_is_readonly() */
736 smp_wmb();
737 sb->s_readonly_remount = 0;
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400738
Nick Piggind208bbd2009-12-21 16:28:53 -0800739 /*
740 * Some filesystems modify their metadata via some other path than the
741 * bdev buffer cache (eg. use a private mapping, or directories in
742 * pagecache, etc). Also file data modifications go via their own
743 * mappings. So If we try to mount readonly then copy the filesystem
744 * from bdev, we could get stale data, so invalidate it to give a best
745 * effort at coherency.
746 */
747 if (remount_ro && sb->s_bdev)
748 invalidate_bdev(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 return 0;
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100750
751cancel_readonly:
752 sb->s_readonly_remount = 0;
753 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754}
755
Jens Axboea2a95372009-03-17 09:38:40 +0100756static void do_emergency_remount(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757{
Al Virodca33252010-07-25 02:31:46 +0400758 struct super_block *sb, *p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
760 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400761 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500762 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400763 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 sb->s_count++;
765 spin_unlock(&sb_lock);
Al Viro443b94b2009-05-05 23:48:50 -0400766 down_write(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500767 if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) &&
768 !(sb->s_flags & MS_RDONLY)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 * What lock protects sb->s_flags??
771 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772 do_remount_sb(sb, MS_RDONLY, NULL, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 }
Al Viro443b94b2009-05-05 23:48:50 -0400774 up_write(&sb->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400776 if (p)
777 __put_super(p);
778 p = sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 }
Al Virodca33252010-07-25 02:31:46 +0400780 if (p)
781 __put_super(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 spin_unlock(&sb_lock);
Jens Axboea2a95372009-03-17 09:38:40 +0100783 kfree(work);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 printk("Emergency Remount complete\n");
785}
786
787void emergency_remount(void)
788{
Jens Axboea2a95372009-03-17 09:38:40 +0100789 struct work_struct *work;
790
791 work = kmalloc(sizeof(*work), GFP_ATOMIC);
792 if (work) {
793 INIT_WORK(work, do_emergency_remount);
794 schedule_work(work);
795 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796}
797
798/*
799 * Unnamed block devices are dummy devices used by virtual
800 * filesystems which don't use real block-devices. -- jrs
801 */
802
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400803static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
Al Viroc63e09e2009-06-24 02:05:18 -0400805static int unnamed_dev_start = 0; /* don't bother trying below it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806
Al Viro0ee5dc62011-07-07 15:44:25 -0400807int get_anon_bdev(dev_t *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808{
809 int dev;
810 int error;
811
812 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400813 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 return -ENOMEM;
815 spin_lock(&unnamed_dev_lock);
Al Viroc63e09e2009-06-24 02:05:18 -0400816 error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
Al Virof21f6222009-06-24 03:12:00 -0400817 if (!error)
818 unnamed_dev_start = dev + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 spin_unlock(&unnamed_dev_lock);
820 if (error == -EAGAIN)
821 /* We raced and lost with another CPU. */
822 goto retry;
823 else if (error)
824 return -EAGAIN;
825
Tejun Heoe8c8d1b2013-02-27 17:05:04 -0800826 if (dev == (1 << MINORBITS)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400828 ida_remove(&unnamed_dev_ida, dev);
Al Virof21f6222009-06-24 03:12:00 -0400829 if (unnamed_dev_start > dev)
830 unnamed_dev_start = dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 spin_unlock(&unnamed_dev_lock);
832 return -EMFILE;
833 }
Al Viro0ee5dc62011-07-07 15:44:25 -0400834 *p = MKDEV(0, dev & MINORMASK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 return 0;
836}
Al Viro0ee5dc62011-07-07 15:44:25 -0400837EXPORT_SYMBOL(get_anon_bdev);
838
839void free_anon_bdev(dev_t dev)
840{
841 int slot = MINOR(dev);
842 spin_lock(&unnamed_dev_lock);
843 ida_remove(&unnamed_dev_ida, slot);
844 if (slot < unnamed_dev_start)
845 unnamed_dev_start = slot;
846 spin_unlock(&unnamed_dev_lock);
847}
848EXPORT_SYMBOL(free_anon_bdev);
849
850int set_anon_super(struct super_block *s, void *data)
851{
852 int error = get_anon_bdev(&s->s_dev);
853 if (!error)
854 s->s_bdi = &noop_backing_dev_info;
855 return error;
856}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857
858EXPORT_SYMBOL(set_anon_super);
859
860void kill_anon_super(struct super_block *sb)
861{
Al Viro0ee5dc62011-07-07 15:44:25 -0400862 dev_t dev = sb->s_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 generic_shutdown_super(sb);
Al Viro0ee5dc62011-07-07 15:44:25 -0400864 free_anon_bdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865}
866
867EXPORT_SYMBOL(kill_anon_super);
868
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869void kill_litter_super(struct super_block *sb)
870{
871 if (sb->s_root)
872 d_genocide(sb->s_root);
873 kill_anon_super(sb);
874}
875
876EXPORT_SYMBOL(kill_litter_super);
877
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700878static int ns_test_super(struct super_block *sb, void *data)
879{
880 return sb->s_fs_info == data;
881}
882
883static int ns_set_super(struct super_block *sb, void *data)
884{
885 sb->s_fs_info = data;
886 return set_anon_super(sb, NULL);
887}
888
Al Viroceefda62010-07-26 13:16:50 +0400889struct dentry *mount_ns(struct file_system_type *fs_type, int flags,
890 void *data, int (*fill_super)(struct super_block *, void *, int))
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700891{
892 struct super_block *sb;
893
David Howells9249e172012-06-25 12:55:37 +0100894 sb = sget(fs_type, ns_test_super, ns_set_super, flags, data);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700895 if (IS_ERR(sb))
Al Viroceefda62010-07-26 13:16:50 +0400896 return ERR_CAST(sb);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700897
898 if (!sb->s_root) {
899 int err;
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700900 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
901 if (err) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400902 deactivate_locked_super(sb);
Al Viroceefda62010-07-26 13:16:50 +0400903 return ERR_PTR(err);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700904 }
905
906 sb->s_flags |= MS_ACTIVE;
907 }
908
Al Viroceefda62010-07-26 13:16:50 +0400909 return dget(sb->s_root);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700910}
911
Al Viroceefda62010-07-26 13:16:50 +0400912EXPORT_SYMBOL(mount_ns);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700913
David Howells93614012006-09-30 20:45:40 +0200914#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915static int set_bdev_super(struct super_block *s, void *data)
916{
917 s->s_bdev = data;
918 s->s_dev = s->s_bdev->bd_dev;
Jens Axboe32a88aa2009-09-16 15:02:33 +0200919
920 /*
921 * We set the bdi here to the queue backing, file systems can
922 * overwrite this in ->fill_super()
923 */
924 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 return 0;
926}
927
928static int test_bdev_super(struct super_block *s, void *data)
929{
930 return (void *)s->s_bdev == data;
931}
932
Al Viro152a0832010-07-25 00:46:55 +0400933struct dentry *mount_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 int flags, const char *dev_name, void *data,
Al Viro152a0832010-07-25 00:46:55 +0400935 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936{
937 struct block_device *bdev;
938 struct super_block *s;
Tejun Heod4d77622010-11-13 11:55:18 +0100939 fmode_t mode = FMODE_READ | FMODE_EXCL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 int error = 0;
941
Al Viro30c40d22008-02-22 19:50:45 -0500942 if (!(flags & MS_RDONLY))
943 mode |= FMODE_WRITE;
944
Tejun Heod4d77622010-11-13 11:55:18 +0100945 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 if (IS_ERR(bdev))
Al Viro152a0832010-07-25 00:46:55 +0400947 return ERR_CAST(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948
949 /*
950 * once the super is inserted into the list by sget, s_umount
951 * will protect the lockfs code from trying to start a snapshot
952 * while we are mounting
953 */
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200954 mutex_lock(&bdev->bd_fsfreeze_mutex);
955 if (bdev->bd_fsfreeze_count > 0) {
956 mutex_unlock(&bdev->bd_fsfreeze_mutex);
957 error = -EBUSY;
958 goto error_bdev;
959 }
David Howells9249e172012-06-25 12:55:37 +0100960 s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC,
961 bdev);
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200962 mutex_unlock(&bdev->bd_fsfreeze_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700964 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965
966 if (s->s_root) {
967 if ((flags ^ s->s_flags) & MS_RDONLY) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400968 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700969 error = -EBUSY;
970 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 }
David Howells454e2392006-06-23 02:02:57 -0700972
Tejun Heo4f331f02010-07-20 15:18:07 -0700973 /*
974 * s_umount nests inside bd_mutex during
Tejun Heoe525fd82010-11-13 11:55:17 +0100975 * __invalidate_device(). blkdev_put() acquires
976 * bd_mutex and can't be called under s_umount. Drop
977 * s_umount temporarily. This is safe as we're
978 * holding an active reference.
Tejun Heo4f331f02010-07-20 15:18:07 -0700979 */
980 up_write(&s->s_umount);
Tejun Heod4d77622010-11-13 11:55:18 +0100981 blkdev_put(bdev, mode);
Tejun Heo4f331f02010-07-20 15:18:07 -0700982 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 } else {
984 char b[BDEVNAME_SIZE];
985
Al Viro30c40d22008-02-22 19:50:45 -0500986 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800988 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800989 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400991 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700992 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800993 }
David Howells454e2392006-06-23 02:02:57 -0700994
995 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -0500996 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 }
998
Al Viro152a0832010-07-25 00:46:55 +0400999 return dget(s->s_root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000
David Howells454e2392006-06-23 02:02:57 -07001001error_s:
1002 error = PTR_ERR(s);
1003error_bdev:
Tejun Heod4d77622010-11-13 11:55:18 +01001004 blkdev_put(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -07001005error:
Al Viro152a0832010-07-25 00:46:55 +04001006 return ERR_PTR(error);
1007}
1008EXPORT_SYMBOL(mount_bdev);
1009
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010void kill_block_super(struct super_block *sb)
1011{
1012 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -05001013 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
H Hartley Sweetenddbaaf32009-04-29 20:14:57 -04001015 bdev->bd_super = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 generic_shutdown_super(sb);
1017 sync_blockdev(bdev);
Tejun Heod4d77622010-11-13 11:55:18 +01001018 WARN_ON_ONCE(!(mode & FMODE_EXCL));
Tejun Heoe525fd82010-11-13 11:55:17 +01001019 blkdev_put(bdev, mode | FMODE_EXCL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020}
1021
1022EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +02001023#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024
Al Viro3c26ff62010-07-25 11:46:36 +04001025struct dentry *mount_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 int flags, void *data,
Al Viro3c26ff62010-07-25 11:46:36 +04001027 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028{
1029 int error;
David Howells9249e172012-06-25 12:55:37 +01001030 struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031
1032 if (IS_ERR(s))
Al Viro3c26ff62010-07-25 11:46:36 +04001033 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001035 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001037 deactivate_locked_super(s);
Al Viro3c26ff62010-07-25 11:46:36 +04001038 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 }
1040 s->s_flags |= MS_ACTIVE;
Al Viro3c26ff62010-07-25 11:46:36 +04001041 return dget(s->s_root);
1042}
1043EXPORT_SYMBOL(mount_nodev);
1044
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045static int compare_single(struct super_block *s, void *p)
1046{
1047 return 1;
1048}
1049
Al Virofc14f2f2010-07-25 01:48:30 +04001050struct dentry *mount_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 int flags, void *data,
Al Virofc14f2f2010-07-25 01:48:30 +04001052 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053{
1054 struct super_block *s;
1055 int error;
1056
David Howells9249e172012-06-25 12:55:37 +01001057 s = sget(fs_type, compare_single, set_anon_super, flags, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 if (IS_ERR(s))
Al Virofc14f2f2010-07-25 01:48:30 +04001059 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 if (!s->s_root) {
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001061 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001063 deactivate_locked_super(s);
Al Virofc14f2f2010-07-25 01:48:30 +04001064 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 }
1066 s->s_flags |= MS_ACTIVE;
Kay Sievers9329d1b2009-12-18 21:18:15 +01001067 } else {
1068 do_remount_sb(s, flags, data, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 }
Al Virofc14f2f2010-07-25 01:48:30 +04001070 return dget(s->s_root);
1071}
1072EXPORT_SYMBOL(mount_single);
1073
Al Viro9d412a42011-03-17 22:08:28 -04001074struct dentry *
1075mount_fs(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076{
Al Viroc96e41e2010-07-25 00:17:56 +04001077 struct dentry *root;
Al Viro9d412a42011-03-17 22:08:28 -04001078 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 char *secdata = NULL;
Al Viro9d412a42011-03-17 22:08:28 -04001080 int error = -ENOMEM;
Al Viro80893522010-02-05 09:30:46 -05001081
Eric Parise0007522008-03-05 10:31:54 -05001082 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -07001084 if (!secdata)
Al Viro9d412a42011-03-17 22:08:28 -04001085 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086
Eric Parise0007522008-03-05 10:31:54 -05001087 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -07001088 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 }
1091
Al Viro1a102ff2011-03-16 09:07:58 -04001092 root = type->mount(type, flags, name, data);
1093 if (IS_ERR(root)) {
1094 error = PTR_ERR(root);
1095 goto out_free_secdata;
Al Viroc96e41e2010-07-25 00:17:56 +04001096 }
Al Viro9d412a42011-03-17 22:08:28 -04001097 sb = root->d_sb;
1098 BUG_ON(!sb);
1099 WARN_ON(!sb->s_bdi);
Linus Torvalds6c510382011-03-24 10:16:26 -07001100 WARN_ON(sb->s_bdi == &default_backing_dev_info);
Al Viro9d412a42011-03-17 22:08:28 -04001101 sb->s_flags |= MS_BORN;
David Howells454e2392006-06-23 02:02:57 -07001102
Al Viro9d412a42011-03-17 22:08:28 -04001103 error = security_sb_kern_mount(sb, flags, secdata);
Jörn Engel5129a462010-04-25 08:54:42 +02001104 if (error)
1105 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -07001106
Jeff Layton42cb56a2009-09-18 13:05:53 -07001107 /*
1108 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
1109 * but s_maxbytes was an unsigned long long for many releases. Throw
1110 * this warning for a little while to try and catch filesystems that
Jeff Layton4358b562011-03-29 09:33:31 -04001111 * violate this rule.
Jeff Layton42cb56a2009-09-18 13:05:53 -07001112 */
Al Viro9d412a42011-03-17 22:08:28 -04001113 WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to "
1114 "negative value (%lld)\n", type->name, sb->s_maxbytes);
Jeff Layton42cb56a2009-09-18 13:05:53 -07001115
Al Viro9d412a42011-03-17 22:08:28 -04001116 up_write(&sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -07001117 free_secdata(secdata);
Al Viro9d412a42011-03-17 22:08:28 -04001118 return root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119out_sb:
Al Viro9d412a42011-03-17 22:08:28 -04001120 dput(root);
1121 deactivate_locked_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122out_free_secdata:
1123 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124out:
David Howells454e2392006-06-23 02:02:57 -07001125 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126}
1127
Jan Kara5accdf82012-06-12 16:20:34 +02001128/*
1129 * This is an internal function, please use sb_end_{write,pagefault,intwrite}
1130 * instead.
1131 */
1132void __sb_end_write(struct super_block *sb, int level)
1133{
1134 percpu_counter_dec(&sb->s_writers.counter[level-1]);
1135 /*
1136 * Make sure s_writers are updated before we wake up waiters in
1137 * freeze_super().
1138 */
1139 smp_mb();
1140 if (waitqueue_active(&sb->s_writers.wait))
1141 wake_up(&sb->s_writers.wait);
1142 rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_);
1143}
1144EXPORT_SYMBOL(__sb_end_write);
1145
1146#ifdef CONFIG_LOCKDEP
1147/*
1148 * We want lockdep to tell us about possible deadlocks with freezing but
1149 * it's it bit tricky to properly instrument it. Getting a freeze protection
1150 * works as getting a read lock but there are subtle problems. XFS for example
1151 * gets freeze protection on internal level twice in some cases, which is OK
1152 * only because we already hold a freeze protection also on higher level. Due
1153 * to these cases we have to tell lockdep we are doing trylock when we
1154 * already hold a freeze protection for a higher freeze level.
1155 */
1156static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock,
1157 unsigned long ip)
1158{
1159 int i;
1160
1161 if (!trylock) {
1162 for (i = 0; i < level - 1; i++)
1163 if (lock_is_held(&sb->s_writers.lock_map[i])) {
1164 trylock = true;
1165 break;
1166 }
1167 }
1168 rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip);
1169}
1170#endif
1171
1172/*
1173 * This is an internal function, please use sb_start_{write,pagefault,intwrite}
1174 * instead.
1175 */
1176int __sb_start_write(struct super_block *sb, int level, bool wait)
1177{
1178retry:
1179 if (unlikely(sb->s_writers.frozen >= level)) {
1180 if (!wait)
1181 return 0;
1182 wait_event(sb->s_writers.wait_unfrozen,
1183 sb->s_writers.frozen < level);
1184 }
1185
1186#ifdef CONFIG_LOCKDEP
1187 acquire_freeze_lock(sb, level, !wait, _RET_IP_);
1188#endif
1189 percpu_counter_inc(&sb->s_writers.counter[level-1]);
1190 /*
1191 * Make sure counter is updated before we check for frozen.
1192 * freeze_super() first sets frozen and then checks the counter.
1193 */
1194 smp_mb();
1195 if (unlikely(sb->s_writers.frozen >= level)) {
1196 __sb_end_write(sb, level);
1197 goto retry;
1198 }
1199 return 1;
1200}
1201EXPORT_SYMBOL(__sb_start_write);
1202
1203/**
1204 * sb_wait_write - wait until all writers to given file system finish
1205 * @sb: the super for which we wait
1206 * @level: type of writers we wait for (normal vs page fault)
1207 *
1208 * This function waits until there are no writers of given type to given file
1209 * system. Caller of this function should make sure there can be no new writers
1210 * of type @level before calling this function. Otherwise this function can
1211 * livelock.
1212 */
1213static void sb_wait_write(struct super_block *sb, int level)
1214{
1215 s64 writers;
1216
1217 /*
1218 * We just cycle-through lockdep here so that it does not complain
1219 * about returning with lock to userspace
1220 */
1221 rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_);
1222 rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_);
1223
1224 do {
1225 DEFINE_WAIT(wait);
1226
1227 /*
1228 * We use a barrier in prepare_to_wait() to separate setting
1229 * of frozen and checking of the counter
1230 */
1231 prepare_to_wait(&sb->s_writers.wait, &wait,
1232 TASK_UNINTERRUPTIBLE);
1233
1234 writers = percpu_counter_sum(&sb->s_writers.counter[level-1]);
1235 if (writers)
1236 schedule();
1237
1238 finish_wait(&sb->s_writers.wait, &wait);
1239 } while (writers);
1240}
1241
Josef Bacik18e9e512010-03-23 10:34:56 -04001242/**
Randy Dunlap7000d3c2010-05-24 22:22:34 -07001243 * freeze_super - lock the filesystem and force it into a consistent state
1244 * @sb: the super to lock
Josef Bacik18e9e512010-03-23 10:34:56 -04001245 *
1246 * Syncs the super to make sure the filesystem is consistent and calls the fs's
1247 * freeze_fs. Subsequent calls to this without first thawing the fs will return
1248 * -EBUSY.
Jan Kara5accdf82012-06-12 16:20:34 +02001249 *
1250 * During this function, sb->s_writers.frozen goes through these values:
1251 *
1252 * SB_UNFROZEN: File system is normal, all writes progress as usual.
1253 *
1254 * SB_FREEZE_WRITE: The file system is in the process of being frozen. New
1255 * writes should be blocked, though page faults are still allowed. We wait for
1256 * all writes to complete and then proceed to the next stage.
1257 *
1258 * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked
1259 * but internal fs threads can still modify the filesystem (although they
1260 * should not dirty new pages or inodes), writeback can run etc. After waiting
1261 * for all running page faults we sync the filesystem which will clean all
1262 * dirty pages and inodes (no new dirty pages or inodes can be created when
1263 * sync is running).
1264 *
1265 * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs
1266 * modification are blocked (e.g. XFS preallocation truncation on inode
1267 * reclaim). This is usually implemented by blocking new transactions for
1268 * filesystems that have them and need this additional guard. After all
1269 * internal writers are finished we call ->freeze_fs() to finish filesystem
1270 * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is
1271 * mostly auxiliary for filesystems to verify they do not modify frozen fs.
1272 *
1273 * sb->s_writers.frozen is protected by sb->s_umount.
Josef Bacik18e9e512010-03-23 10:34:56 -04001274 */
1275int freeze_super(struct super_block *sb)
1276{
1277 int ret;
1278
1279 atomic_inc(&sb->s_active);
1280 down_write(&sb->s_umount);
Jan Kara5accdf82012-06-12 16:20:34 +02001281 if (sb->s_writers.frozen != SB_UNFROZEN) {
Josef Bacik18e9e512010-03-23 10:34:56 -04001282 deactivate_locked_super(sb);
1283 return -EBUSY;
1284 }
1285
Al Virodabe0dc2012-01-03 21:01:29 -05001286 if (!(sb->s_flags & MS_BORN)) {
1287 up_write(&sb->s_umount);
1288 return 0; /* sic - it's "nothing to do" */
1289 }
1290
Josef Bacik18e9e512010-03-23 10:34:56 -04001291 if (sb->s_flags & MS_RDONLY) {
Jan Kara5accdf82012-06-12 16:20:34 +02001292 /* Nothing to do really... */
1293 sb->s_writers.frozen = SB_FREEZE_COMPLETE;
Josef Bacik18e9e512010-03-23 10:34:56 -04001294 up_write(&sb->s_umount);
1295 return 0;
1296 }
1297
Jan Kara5accdf82012-06-12 16:20:34 +02001298 /* From now on, no new normal writers can start */
1299 sb->s_writers.frozen = SB_FREEZE_WRITE;
Josef Bacik18e9e512010-03-23 10:34:56 -04001300 smp_wmb();
1301
Jan Kara5accdf82012-06-12 16:20:34 +02001302 /* Release s_umount to preserve sb_start_write -> s_umount ordering */
1303 up_write(&sb->s_umount);
1304
1305 sb_wait_write(sb, SB_FREEZE_WRITE);
1306
1307 /* Now we go and block page faults... */
1308 down_write(&sb->s_umount);
1309 sb->s_writers.frozen = SB_FREEZE_PAGEFAULT;
1310 smp_wmb();
1311
1312 sb_wait_write(sb, SB_FREEZE_PAGEFAULT);
1313
1314 /* All writers are done so after syncing there won't be dirty data */
Josef Bacik18e9e512010-03-23 10:34:56 -04001315 sync_filesystem(sb);
1316
Jan Kara5accdf82012-06-12 16:20:34 +02001317 /* Now wait for internal filesystem counter */
1318 sb->s_writers.frozen = SB_FREEZE_FS;
Josef Bacik18e9e512010-03-23 10:34:56 -04001319 smp_wmb();
Jan Kara5accdf82012-06-12 16:20:34 +02001320 sb_wait_write(sb, SB_FREEZE_FS);
Josef Bacik18e9e512010-03-23 10:34:56 -04001321
Josef Bacik18e9e512010-03-23 10:34:56 -04001322 if (sb->s_op->freeze_fs) {
1323 ret = sb->s_op->freeze_fs(sb);
1324 if (ret) {
1325 printk(KERN_ERR
1326 "VFS:Filesystem freeze failed\n");
Jan Kara5accdf82012-06-12 16:20:34 +02001327 sb->s_writers.frozen = SB_UNFROZEN;
Kazuya Mioe1616302011-12-01 16:51:07 +09001328 smp_wmb();
Jan Kara5accdf82012-06-12 16:20:34 +02001329 wake_up(&sb->s_writers.wait_unfrozen);
Josef Bacik18e9e512010-03-23 10:34:56 -04001330 deactivate_locked_super(sb);
1331 return ret;
1332 }
1333 }
Jan Kara5accdf82012-06-12 16:20:34 +02001334 /*
1335 * This is just for debugging purposes so that fs can warn if it
1336 * sees write activity when frozen is set to SB_FREEZE_COMPLETE.
1337 */
1338 sb->s_writers.frozen = SB_FREEZE_COMPLETE;
Josef Bacik18e9e512010-03-23 10:34:56 -04001339 up_write(&sb->s_umount);
1340 return 0;
1341}
1342EXPORT_SYMBOL(freeze_super);
1343
1344/**
1345 * thaw_super -- unlock filesystem
1346 * @sb: the super to thaw
1347 *
1348 * Unlocks the filesystem and marks it writeable again after freeze_super().
1349 */
1350int thaw_super(struct super_block *sb)
1351{
1352 int error;
1353
1354 down_write(&sb->s_umount);
Jan Kara5accdf82012-06-12 16:20:34 +02001355 if (sb->s_writers.frozen == SB_UNFROZEN) {
Josef Bacik18e9e512010-03-23 10:34:56 -04001356 up_write(&sb->s_umount);
1357 return -EINVAL;
1358 }
1359
1360 if (sb->s_flags & MS_RDONLY)
1361 goto out;
1362
1363 if (sb->s_op->unfreeze_fs) {
1364 error = sb->s_op->unfreeze_fs(sb);
1365 if (error) {
1366 printk(KERN_ERR
1367 "VFS:Filesystem thaw failed\n");
Josef Bacik18e9e512010-03-23 10:34:56 -04001368 up_write(&sb->s_umount);
1369 return error;
1370 }
1371 }
1372
1373out:
Jan Kara5accdf82012-06-12 16:20:34 +02001374 sb->s_writers.frozen = SB_UNFROZEN;
Josef Bacik18e9e512010-03-23 10:34:56 -04001375 smp_wmb();
Jan Kara5accdf82012-06-12 16:20:34 +02001376 wake_up(&sb->s_writers.wait_unfrozen);
Josef Bacik18e9e512010-03-23 10:34:56 -04001377 deactivate_locked_super(sb);
1378
1379 return 0;
1380}
1381EXPORT_SYMBOL(thaw_super);