blob: 6277ec6cb60a65e71b2c590260c67a057a1748ca [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>
Dan Magenheimerc515e1f2011-05-26 10:01:43 -060034#include <linux/cleancache.h>
Al Viro6d59e7f2008-03-22 15:48:17 -040035#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070036
37
Linus Torvalds1da177e2005-04-16 15:20:36 -070038LIST_HEAD(super_blocks);
39DEFINE_SPINLOCK(sb_lock);
40
Dave Chinnerb0d40c92011-07-08 14:14:42 +100041/*
42 * One thing we have to be careful of with a per-sb shrinker is that we don't
43 * drop the last active reference to the superblock from within the shrinker.
44 * If that happens we could trigger unregistering the shrinker from within the
45 * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we
46 * take a passive reference to the superblock to avoid this from occurring.
47 */
48static int prune_super(struct shrinker *shrink, struct shrink_control *sc)
49{
50 struct super_block *sb;
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100051 int fs_objects = 0;
52 int total_objects;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100053
54 sb = container_of(shrink, struct super_block, s_shrink);
55
56 /*
57 * Deadlock avoidance. We may hold various FS locks, and we don't want
58 * to recurse into the FS that called us in clear_inode() and friends..
59 */
60 if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS))
61 return -1;
62
63 if (!grab_super_passive(sb))
Mikulas Patocka09f363c2011-10-31 17:08:57 -070064 return !sc->nr_to_scan ? 0 : -1;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100065
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100066 if (sb->s_op && sb->s_op->nr_cached_objects)
67 fs_objects = sb->s_op->nr_cached_objects(sb);
68
69 total_objects = sb->s_nr_dentry_unused +
70 sb->s_nr_inodes_unused + fs_objects + 1;
71
Dave Chinnerb0d40c92011-07-08 14:14:42 +100072 if (sc->nr_to_scan) {
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100073 int dentries;
74 int inodes;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100075
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100076 /* proportion the scan between the caches */
77 dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) /
78 total_objects;
79 inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) /
80 total_objects;
81 if (fs_objects)
82 fs_objects = (sc->nr_to_scan * fs_objects) /
83 total_objects;
84 /*
85 * prune the dcache first as the icache is pinned by it, then
86 * prune the icache, followed by the filesystem specific caches
87 */
88 prune_dcache_sb(sb, dentries);
89 prune_icache_sb(sb, inodes);
Dave Chinnerb0d40c92011-07-08 14:14:42 +100090
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100091 if (fs_objects && sb->s_op->free_cached_objects) {
92 sb->s_op->free_cached_objects(sb, fs_objects);
93 fs_objects = sb->s_op->nr_cached_objects(sb);
94 }
95 total_objects = sb->s_nr_dentry_unused +
96 sb->s_nr_inodes_unused + fs_objects;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100097 }
98
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100099 total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000100 drop_super(sb);
Dave Chinner0e1fdaf2011-07-08 14:14:44 +1000101 return total_objects;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000102}
103
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104/**
105 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -0700106 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 *
108 * Allocates and initializes a new &struct super_block. alloc_super()
109 * returns a pointer new superblock or %NULL if allocation had failed.
110 */
Ingo Molnarcf516242006-07-03 00:25:27 -0700111static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -0800113 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700114 static const struct super_operations default_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115
116 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 if (security_sb_alloc(s)) {
118 kfree(s);
119 s = NULL;
120 goto out;
121 }
Nick Piggin6416ccb2010-08-18 04:37:38 +1000122#ifdef CONFIG_SMP
123 s->s_files = alloc_percpu(struct list_head);
124 if (!s->s_files) {
125 security_sb_free(s);
126 kfree(s);
127 s = NULL;
128 goto out;
129 } else {
130 int i;
131
132 for_each_possible_cpu(i)
133 INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i));
134 }
135#else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 INIT_LIST_HEAD(&s->s_files);
Nick Piggin6416ccb2010-08-18 04:37:38 +1000137#endif
Jens Axboe95f28602011-03-17 11:13:12 +0100138 s->s_bdi = &default_backing_dev_info;
Al Viroa5166162011-12-12 22:53:00 -0500139 INIT_HLIST_NODE(&s->s_instances);
Nick Pigginceb5bdc2011-01-07 17:50:05 +1100140 INIT_HLIST_BL_HEAD(&s->s_anon);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 INIT_LIST_HEAD(&s->s_inodes);
Kentaro Makitada3bbdd2008-07-23 21:27:13 -0700142 INIT_LIST_HEAD(&s->s_dentry_lru);
Dave Chinner98b745c2011-07-08 14:14:39 +1000143 INIT_LIST_HEAD(&s->s_inode_lru);
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000144 spin_lock_init(&s->s_inode_lru_lock);
Miklos Szeredi39f7c4d2011-11-21 12:11:30 +0100145 INIT_LIST_HEAD(&s->s_mounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f2006-01-09 15:59:25 -0800147 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -0700148 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -0700149 /*
150 * The locking rules for s_lock are up to the
151 * filesystem. For example ext3fs has different
152 * lock ordering than usbfs:
153 */
154 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Peter Zijlstraada723d2009-02-18 14:48:30 -0800155 /*
156 * sget() can have s_umount recursion.
157 *
158 * When it cannot find a suitable sb, it allocates a new
159 * one (this one), and tries again to find a suitable old
160 * one.
161 *
162 * In case that succeeds, it will acquire the s_umount
163 * lock of the old one. Since these are clearly distrinct
164 * locks, and this object isn't exposed yet, there's no
165 * risk of deadlocks.
166 *
167 * Annotate this by putting this lock in a different
168 * subclass.
169 */
170 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
Al Virob20bd1a2010-03-22 08:53:19 -0400171 s->s_count = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -0800173 mutex_init(&s->s_vfs_rename_mutex);
Roland Dreier51ee0492010-04-27 14:23:57 -0700174 lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key);
Ingo Molnard3be9152006-03-23 03:00:29 -0800175 mutex_init(&s->s_dquot.dqio_mutex);
176 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 init_rwsem(&s->s_dquot.dqptr_sem);
178 init_waitqueue_head(&s->s_wait_unfrozen);
179 s->s_maxbytes = MAX_NON_LFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 s->s_op = &default_op;
181 s->s_time_gran = 1000000000;
Dan Magenheimerc515e1f2011-05-26 10:01:43 -0600182 s->cleancache_poolid = -1;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000183
184 s->s_shrink.seeks = DEFAULT_SEEKS;
185 s->s_shrink.shrink = prune_super;
Dave Chinner8ab47662011-07-08 14:14:45 +1000186 s->s_shrink.batch = 1024;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 }
188out:
189 return s;
190}
191
192/**
193 * destroy_super - frees a superblock
194 * @s: superblock to free
195 *
196 * Frees a superblock.
197 */
198static inline void destroy_super(struct super_block *s)
199{
Nick Piggin6416ccb2010-08-18 04:37:38 +1000200#ifdef CONFIG_SMP
201 free_percpu(s->s_files);
202#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 security_sb_free(s);
Miklos Szeredi39f7c4d2011-11-21 12:11:30 +0100204 WARN_ON(!list_empty(&s->s_mounts));
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700205 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800206 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 kfree(s);
208}
209
210/* Superblock refcounting */
211
212/*
Al Viro35cf7ba2010-03-22 21:13:53 -0400213 * Drop a superblock's refcount. The caller must hold sb_lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 */
Al Virof47ec3f2011-11-21 21:15:42 -0500215static void __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 if (!--sb->s_count) {
Al Viro551de6f2010-03-22 19:36:35 -0400218 list_del_init(&sb->s_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 destroy_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221}
222
223/**
224 * put_super - drop a temporary reference to superblock
225 * @sb: superblock in question
226 *
227 * Drops a temporary reference, frees superblock if there's no
228 * references left.
229 */
Al Virof47ec3f2011-11-21 21:15:42 -0500230static void put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231{
232 spin_lock(&sb_lock);
233 __put_super(sb);
234 spin_unlock(&sb_lock);
235}
236
237
238/**
Al Viro74dbbdd2009-05-06 01:07:50 -0400239 * deactivate_locked_super - drop an active reference to superblock
240 * @s: superblock to deactivate
241 *
Al Viro1712ac82010-03-22 15:22:31 -0400242 * Drops an active reference to superblock, converting it into a temprory
243 * one if there is no other active references left. In that case we
244 * tell fs driver to shut it down and drop the temporary reference we
245 * had just acquired.
246 *
247 * Caller holds exclusive lock on superblock; that lock is released.
Al Viro74dbbdd2009-05-06 01:07:50 -0400248 */
249void deactivate_locked_super(struct super_block *s)
250{
251 struct file_system_type *fs = s->s_type;
Al Virob20bd1a2010-03-22 08:53:19 -0400252 if (atomic_dec_and_test(&s->s_active)) {
Dan Magenheimerc515e1f2011-05-26 10:01:43 -0600253 cleancache_flush_fs(s);
Al Viro74dbbdd2009-05-06 01:07:50 -0400254 fs->kill_sb(s);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000255
256 /* caches are now gone, we can safely kill the shrinker now */
257 unregister_shrinker(&s->s_shrink);
258
Boaz Harroshd863b502011-02-10 15:01:20 -0800259 /*
260 * We need to call rcu_barrier so all the delayed rcu free
261 * inodes are flushed before we release the fs module.
262 */
263 rcu_barrier();
Al Viro74dbbdd2009-05-06 01:07:50 -0400264 put_filesystem(fs);
265 put_super(s);
266 } else {
267 up_write(&s->s_umount);
268 }
269}
270
271EXPORT_SYMBOL(deactivate_locked_super);
272
273/**
Al Viro1712ac82010-03-22 15:22:31 -0400274 * deactivate_super - drop an active reference to superblock
275 * @s: superblock to deactivate
276 *
277 * Variant of deactivate_locked_super(), except that superblock is *not*
278 * locked by caller. If we are going to drop the final active reference,
279 * lock will be acquired prior to that.
280 */
281void deactivate_super(struct super_block *s)
282{
283 if (!atomic_add_unless(&s->s_active, -1, 1)) {
284 down_write(&s->s_umount);
285 deactivate_locked_super(s);
286 }
287}
288
289EXPORT_SYMBOL(deactivate_super);
290
291/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 * grab_super - acquire an active reference
293 * @s: reference we are trying to make active
294 *
295 * Tries to acquire an active reference. grab_super() is used when we
296 * had just found a superblock in super_blocks or fs_type->fs_supers
297 * and want to turn it into a full-blown active reference. grab_super()
298 * is called with sb_lock held and drops it. Returns 1 in case of
299 * success, 0 if we had failed (superblock contents was already dead or
300 * dying when grab_super() had been called).
301 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700302static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303{
Al Virob20bd1a2010-03-22 08:53:19 -0400304 if (atomic_inc_not_zero(&s->s_active)) {
305 spin_unlock(&sb_lock);
Al Virob20bd1a2010-03-22 08:53:19 -0400306 return 1;
307 }
308 /* it's going away */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 s->s_count++;
310 spin_unlock(&sb_lock);
Al Viro1712ac82010-03-22 15:22:31 -0400311 /* wait for it to die */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 up_write(&s->s_umount);
314 put_super(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 return 0;
316}
317
David Howellscf9a2ae2006-08-29 19:05:54 +0100318/*
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000319 * grab_super_passive - acquire a passive reference
320 * @s: reference we are trying to grab
321 *
322 * Tries to acquire a passive reference. This is used in places where we
323 * cannot take an active reference but we need to ensure that the
324 * superblock does not go away while we are working on it. It returns
325 * false if a reference was not gained, and returns true with the s_umount
326 * lock held in read mode if a reference is gained. On successful return,
327 * the caller must drop the s_umount lock and the passive reference when
328 * done.
329 */
330bool grab_super_passive(struct super_block *sb)
331{
332 spin_lock(&sb_lock);
Al Viroa5166162011-12-12 22:53:00 -0500333 if (hlist_unhashed(&sb->s_instances)) {
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000334 spin_unlock(&sb_lock);
335 return false;
336 }
337
338 sb->s_count++;
339 spin_unlock(&sb_lock);
340
341 if (down_read_trylock(&sb->s_umount)) {
Al Virodabe0dc2012-01-03 21:01:29 -0500342 if (sb->s_root && (sb->s_flags & MS_BORN))
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000343 return true;
344 up_read(&sb->s_umount);
345 }
346
347 put_super(sb);
348 return false;
349}
350
351/*
Al Viro914e2632006-10-18 13:55:46 -0400352 * Superblock locking. We really ought to get rid of these two.
353 */
354void lock_super(struct super_block * sb)
355{
Al Viro914e2632006-10-18 13:55:46 -0400356 mutex_lock(&sb->s_lock);
357}
358
359void unlock_super(struct super_block * sb)
360{
Al Viro914e2632006-10-18 13:55:46 -0400361 mutex_unlock(&sb->s_lock);
362}
363
364EXPORT_SYMBOL(lock_super);
365EXPORT_SYMBOL(unlock_super);
366
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367/**
368 * generic_shutdown_super - common helper for ->kill_sb()
369 * @sb: superblock to kill
370 *
371 * generic_shutdown_super() does all fs-independent work on superblock
372 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
373 * that need destruction out of superblock, call generic_shutdown_super()
374 * and release aforementioned objects. Note: dentries and inodes _are_
375 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700376 *
377 * Upon calling this function, the filesystem may no longer alter or
378 * rearrange the set of dentries belonging to this super_block, nor may it
379 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 */
381void generic_shutdown_super(struct super_block *sb)
382{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800383 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384
David Howellsc636ebd2006-10-11 01:22:19 -0700385 if (sb->s_root) {
386 shrink_dcache_for_umount(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200387 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 sb->s_flags &= ~MS_ACTIVE;
Arjan van de Venefaee192009-01-06 07:20:54 -0800389
Al Viro63997e92010-10-25 20:49:35 -0400390 fsnotify_unmount_inodes(&sb->s_inodes);
391
392 evict_inodes(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 if (sop->put_super)
395 sop->put_super(sb);
396
Al Viro63997e92010-10-25 20:49:35 -0400397 if (!list_empty(&sb->s_inodes)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800398 printk("VFS: Busy inodes after unmount of %s. "
399 "Self-destruct in 5 seconds. Have a nice day...\n",
400 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 }
403 spin_lock(&sb_lock);
404 /* should be initialized for __put_super_and_need_restart() */
Al Viroa5166162011-12-12 22:53:00 -0500405 hlist_del_init(&sb->s_instances);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 spin_unlock(&sb_lock);
407 up_write(&sb->s_umount);
408}
409
410EXPORT_SYMBOL(generic_shutdown_super);
411
412/**
413 * sget - find or create a superblock
414 * @type: filesystem type superblock should belong to
415 * @test: comparison callback
416 * @set: setup callback
417 * @data: argument to each of them
418 */
419struct super_block *sget(struct file_system_type *type,
420 int (*test)(struct super_block *,void *),
421 int (*set)(struct super_block *,void *),
422 void *data)
423{
424 struct super_block *s = NULL;
Al Viroa5166162011-12-12 22:53:00 -0500425 struct hlist_node *node;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700426 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 int err;
428
429retry:
430 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700431 if (test) {
Al Viroa5166162011-12-12 22:53:00 -0500432 hlist_for_each_entry(old, node, &type->fs_supers, s_instances) {
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700433 if (!test(old, data))
434 continue;
435 if (!grab_super(old))
436 goto retry;
Li Zefana3cfbb52009-03-12 14:31:29 -0700437 if (s) {
438 up_write(&s->s_umount);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700439 destroy_super(s);
Al Viro7a4dec52010-08-09 12:05:43 -0400440 s = NULL;
Li Zefana3cfbb52009-03-12 14:31:29 -0700441 }
Al Virod3f21472010-03-23 11:11:05 -0400442 down_write(&old->s_umount);
Al Viro7a4dec52010-08-09 12:05:43 -0400443 if (unlikely(!(old->s_flags & MS_BORN))) {
444 deactivate_locked_super(old);
445 goto retry;
446 }
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700447 return old;
448 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 }
450 if (!s) {
451 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700452 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 if (!s)
454 return ERR_PTR(-ENOMEM);
455 goto retry;
456 }
457
458 err = set(s, data);
459 if (err) {
460 spin_unlock(&sb_lock);
Li Zefana3cfbb52009-03-12 14:31:29 -0700461 up_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 destroy_super(s);
463 return ERR_PTR(err);
464 }
465 s->s_type = type;
466 strlcpy(s->s_id, type->name, sizeof(s->s_id));
467 list_add_tail(&s->s_list, &super_blocks);
Al Viroa5166162011-12-12 22:53:00 -0500468 hlist_add_head(&s->s_instances, &type->fs_supers);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 spin_unlock(&sb_lock);
470 get_filesystem(type);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000471 register_shrinker(&s->s_shrink);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 return s;
473}
474
475EXPORT_SYMBOL(sget);
476
477void drop_super(struct super_block *sb)
478{
479 up_read(&sb->s_umount);
480 put_super(sb);
481}
482
483EXPORT_SYMBOL(drop_super);
484
Christoph Hellwige5004752009-05-05 16:08:56 +0200485/**
486 * sync_supers - helper for periodic superblock writeback
487 *
488 * Call the write_super method if present on all dirty superblocks in
489 * the system. This is for the periodic writeback used by most older
490 * filesystems. For data integrity superblock writeback use
491 * sync_filesystems() instead.
492 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 * Note: check the dirty flag before waiting, so we don't
494 * hold up the sync while mounting a device. (The newly
495 * mounted device won't need syncing.)
496 */
497void sync_supers(void)
498{
Al Virodca33252010-07-25 02:31:46 +0400499 struct super_block *sb, *p = NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700500
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400502 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500503 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400504 continue;
Christoph Hellwige5004752009-05-05 16:08:56 +0200505 if (sb->s_op->write_super && sb->s_dirt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 sb->s_count++;
507 spin_unlock(&sb_lock);
Christoph Hellwige5004752009-05-05 16:08:56 +0200508
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 down_read(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500510 if (sb->s_root && sb->s_dirt && (sb->s_flags & MS_BORN))
Christoph Hellwige5004752009-05-05 16:08:56 +0200511 sb->s_op->write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700512 up_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200513
Kirill Korotaev618f0632005-06-23 00:09:54 -0700514 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400515 if (p)
516 __put_super(p);
517 p = sb;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700518 }
519 }
Al Virodca33252010-07-25 02:31:46 +0400520 if (p)
521 __put_super(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 spin_unlock(&sb_lock);
523}
524
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525/**
Al Viro01a05b32010-03-23 06:06:58 -0400526 * iterate_supers - call function for all active superblocks
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(void (*f)(struct super_block *, void *), void *arg)
534{
Al Virodca33252010-07-25 02:31:46 +0400535 struct super_block *sb, *p = NULL;
Al Viro01a05b32010-03-23 06:06:58 -0400536
537 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400538 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500539 if (hlist_unhashed(&sb->s_instances))
Al Viro01a05b32010-03-23 06:06:58 -0400540 continue;
541 sb->s_count++;
542 spin_unlock(&sb_lock);
543
544 down_read(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500545 if (sb->s_root && (sb->s_flags & MS_BORN))
Al Viro01a05b32010-03-23 06:06:58 -0400546 f(sb, arg);
547 up_read(&sb->s_umount);
548
549 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400550 if (p)
551 __put_super(p);
552 p = sb;
Al Viro01a05b32010-03-23 06:06:58 -0400553 }
Al Virodca33252010-07-25 02:31:46 +0400554 if (p)
555 __put_super(p);
Al Viro01a05b32010-03-23 06:06:58 -0400556 spin_unlock(&sb_lock);
557}
558
559/**
Al Viro43e15cd2011-06-03 20:16:57 -0400560 * iterate_supers_type - call function for superblocks of given type
561 * @type: fs type
562 * @f: function to call
563 * @arg: argument to pass to it
564 *
565 * Scans the superblock list and calls given function, passing it
566 * locked superblock and given argument.
567 */
568void iterate_supers_type(struct file_system_type *type,
569 void (*f)(struct super_block *, void *), void *arg)
570{
571 struct super_block *sb, *p = NULL;
Al Viroa5166162011-12-12 22:53:00 -0500572 struct hlist_node *node;
Al Viro43e15cd2011-06-03 20:16:57 -0400573
574 spin_lock(&sb_lock);
Al Viroa5166162011-12-12 22:53:00 -0500575 hlist_for_each_entry(sb, node, &type->fs_supers, s_instances) {
Al Viro43e15cd2011-06-03 20:16:57 -0400576 sb->s_count++;
577 spin_unlock(&sb_lock);
578
579 down_read(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500580 if (sb->s_root && (sb->s_flags & MS_BORN))
Al Viro43e15cd2011-06-03 20:16:57 -0400581 f(sb, arg);
582 up_read(&sb->s_umount);
583
584 spin_lock(&sb_lock);
585 if (p)
586 __put_super(p);
587 p = sb;
588 }
589 if (p)
590 __put_super(p);
591 spin_unlock(&sb_lock);
592}
593
594EXPORT_SYMBOL(iterate_supers_type);
595
596/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 * get_super - get the superblock of a device
598 * @bdev: device to get the superblock for
599 *
600 * Scans the superblock list and finds the superblock of the file system
601 * mounted on the device given. %NULL is returned if no match is found.
602 */
603
Al Virodf40c012010-03-22 20:23:25 -0400604struct super_block *get_super(struct block_device *bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700606 struct super_block *sb;
607
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 if (!bdev)
609 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700610
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700612rescan:
613 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500614 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400615 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700616 if (sb->s_bdev == bdev) {
617 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700619 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400620 /* still alive? */
Al Virodabe0dc2012-01-03 21:01:29 -0500621 if (sb->s_root && (sb->s_flags & MS_BORN))
Kirill Korotaev618f0632005-06-23 00:09:54 -0700622 return sb;
623 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400624 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700625 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400626 __put_super(sb);
627 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 }
629 }
630 spin_unlock(&sb_lock);
631 return NULL;
632}
633
634EXPORT_SYMBOL(get_super);
Christoph Hellwig45042302009-08-03 23:28:35 +0200635
636/**
Jan Kara6b6dc832012-02-10 11:03:00 +0100637 * get_super_thawed - get thawed superblock of a device
638 * @bdev: device to get the superblock for
639 *
640 * Scans the superblock list and finds the superblock of the file system
641 * mounted on the device. The superblock is returned once it is thawed
642 * (or immediately if it was not frozen). %NULL is returned if no match
643 * is found.
644 */
645struct super_block *get_super_thawed(struct block_device *bdev)
646{
647 while (1) {
648 struct super_block *s = get_super(bdev);
649 if (!s || s->s_frozen == SB_UNFROZEN)
650 return s;
651 up_read(&s->s_umount);
652 vfs_check_frozen(s, SB_FREEZE_WRITE);
653 put_super(s);
654 }
655}
656EXPORT_SYMBOL(get_super_thawed);
657
658/**
Christoph Hellwig45042302009-08-03 23:28:35 +0200659 * get_active_super - get an active reference to the superblock of a device
660 * @bdev: device to get the superblock for
661 *
662 * Scans the superblock list and finds the superblock of the file system
663 * mounted on the device given. Returns the superblock with an active
Al Virod3f21472010-03-23 11:11:05 -0400664 * reference or %NULL if none was found.
Christoph Hellwig45042302009-08-03 23:28:35 +0200665 */
666struct super_block *get_active_super(struct block_device *bdev)
667{
668 struct super_block *sb;
669
670 if (!bdev)
671 return NULL;
672
Al Viro14945832010-03-22 20:15:33 -0400673restart:
Christoph Hellwig45042302009-08-03 23:28:35 +0200674 spin_lock(&sb_lock);
675 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500676 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400677 continue;
Al Viro14945832010-03-22 20:15:33 -0400678 if (sb->s_bdev == bdev) {
679 if (grab_super(sb)) /* drops sb_lock */
680 return sb;
681 else
682 goto restart;
683 }
Christoph Hellwig45042302009-08-03 23:28:35 +0200684 }
685 spin_unlock(&sb_lock);
686 return NULL;
687}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
Al Virodf40c012010-03-22 20:23:25 -0400689struct super_block *user_get_super(dev_t dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700691 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700694rescan:
695 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500696 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400697 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700698 if (sb->s_dev == dev) {
699 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700701 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400702 /* still alive? */
Al Virodabe0dc2012-01-03 21:01:29 -0500703 if (sb->s_root && (sb->s_flags & MS_BORN))
Kirill Korotaev618f0632005-06-23 00:09:54 -0700704 return sb;
705 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400706 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700707 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400708 __put_super(sb);
709 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 }
711 }
712 spin_unlock(&sb_lock);
713 return NULL;
714}
715
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 * do_remount_sb - asks filesystem to change mount options.
718 * @sb: superblock in question
719 * @flags: numeric part of options
720 * @data: the rest of options
721 * @force: whether or not to force the change
722 *
723 * Alters the mount options of a mounted file system.
724 */
725int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
726{
727 int retval;
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400728 int remount_ro;
Christoph Hellwig45042302009-08-03 23:28:35 +0200729
730 if (sb->s_frozen != SB_UNFROZEN)
731 return -EBUSY;
732
David Howells93614012006-09-30 20:45:40 +0200733#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
735 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200736#endif
Christoph Hellwig45042302009-08-03 23:28:35 +0200737
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 if (flags & MS_RDONLY)
739 acct_auto_close(sb);
740 shrink_dcache_sb(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200741 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742
Nick Piggind208bbd2009-12-21 16:28:53 -0800743 remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY);
Nick Piggind208bbd2009-12-21 16:28:53 -0800744
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 /* If we are remounting RDONLY and current sb is read/write,
746 make sure there are no rw files opened */
Nick Piggind208bbd2009-12-21 16:28:53 -0800747 if (remount_ro) {
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100748 if (force) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 mark_files_ro(sb);
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100750 } else {
751 retval = sb_prepare_remount_readonly(sb);
752 if (retval)
753 return retval;
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100754 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 }
756
757 if (sb->s_op->remount_fs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 retval = sb->s_op->remount_fs(sb, &flags, data);
Miklos Szeredi2833eb22011-10-28 14:13:20 +0200759 if (retval) {
760 if (!force)
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100761 goto cancel_readonly;
Miklos Szeredi2833eb22011-10-28 14:13:20 +0200762 /* If forced remount, go ahead despite any errors */
763 WARN(1, "forced remount of a %s fs returned %i\n",
764 sb->s_type->name, retval);
765 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 }
767 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100768 /* Needs to be ordered wrt mnt_is_readonly() */
769 smp_wmb();
770 sb->s_readonly_remount = 0;
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400771
Nick Piggind208bbd2009-12-21 16:28:53 -0800772 /*
773 * Some filesystems modify their metadata via some other path than the
774 * bdev buffer cache (eg. use a private mapping, or directories in
775 * pagecache, etc). Also file data modifications go via their own
776 * mappings. So If we try to mount readonly then copy the filesystem
777 * from bdev, we could get stale data, so invalidate it to give a best
778 * effort at coherency.
779 */
780 if (remount_ro && sb->s_bdev)
781 invalidate_bdev(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 return 0;
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100783
784cancel_readonly:
785 sb->s_readonly_remount = 0;
786 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787}
788
Jens Axboea2a95372009-03-17 09:38:40 +0100789static void do_emergency_remount(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790{
Al Virodca33252010-07-25 02:31:46 +0400791 struct super_block *sb, *p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
793 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400794 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500795 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400796 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 sb->s_count++;
798 spin_unlock(&sb_lock);
Al Viro443b94b2009-05-05 23:48:50 -0400799 down_write(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500800 if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) &&
801 !(sb->s_flags & MS_RDONLY)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 * What lock protects sb->s_flags??
804 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 do_remount_sb(sb, MS_RDONLY, NULL, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 }
Al Viro443b94b2009-05-05 23:48:50 -0400807 up_write(&sb->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400809 if (p)
810 __put_super(p);
811 p = sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 }
Al Virodca33252010-07-25 02:31:46 +0400813 if (p)
814 __put_super(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 spin_unlock(&sb_lock);
Jens Axboea2a95372009-03-17 09:38:40 +0100816 kfree(work);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 printk("Emergency Remount complete\n");
818}
819
820void emergency_remount(void)
821{
Jens Axboea2a95372009-03-17 09:38:40 +0100822 struct work_struct *work;
823
824 work = kmalloc(sizeof(*work), GFP_ATOMIC);
825 if (work) {
826 INIT_WORK(work, do_emergency_remount);
827 schedule_work(work);
828 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829}
830
831/*
832 * Unnamed block devices are dummy devices used by virtual
833 * filesystems which don't use real block-devices. -- jrs
834 */
835
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400836static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
Al Viroc63e09e2009-06-24 02:05:18 -0400838static int unnamed_dev_start = 0; /* don't bother trying below it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839
Al Viro0ee5dc62011-07-07 15:44:25 -0400840int get_anon_bdev(dev_t *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841{
842 int dev;
843 int error;
844
845 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400846 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 return -ENOMEM;
848 spin_lock(&unnamed_dev_lock);
Al Viroc63e09e2009-06-24 02:05:18 -0400849 error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
Al Virof21f6222009-06-24 03:12:00 -0400850 if (!error)
851 unnamed_dev_start = dev + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 spin_unlock(&unnamed_dev_lock);
853 if (error == -EAGAIN)
854 /* We raced and lost with another CPU. */
855 goto retry;
856 else if (error)
857 return -EAGAIN;
858
859 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
860 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400861 ida_remove(&unnamed_dev_ida, dev);
Al Virof21f6222009-06-24 03:12:00 -0400862 if (unnamed_dev_start > dev)
863 unnamed_dev_start = dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 spin_unlock(&unnamed_dev_lock);
865 return -EMFILE;
866 }
Al Viro0ee5dc62011-07-07 15:44:25 -0400867 *p = MKDEV(0, dev & MINORMASK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 return 0;
869}
Al Viro0ee5dc62011-07-07 15:44:25 -0400870EXPORT_SYMBOL(get_anon_bdev);
871
872void free_anon_bdev(dev_t dev)
873{
874 int slot = MINOR(dev);
875 spin_lock(&unnamed_dev_lock);
876 ida_remove(&unnamed_dev_ida, slot);
877 if (slot < unnamed_dev_start)
878 unnamed_dev_start = slot;
879 spin_unlock(&unnamed_dev_lock);
880}
881EXPORT_SYMBOL(free_anon_bdev);
882
883int set_anon_super(struct super_block *s, void *data)
884{
885 int error = get_anon_bdev(&s->s_dev);
886 if (!error)
887 s->s_bdi = &noop_backing_dev_info;
888 return error;
889}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
891EXPORT_SYMBOL(set_anon_super);
892
893void kill_anon_super(struct super_block *sb)
894{
Al Viro0ee5dc62011-07-07 15:44:25 -0400895 dev_t dev = sb->s_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 generic_shutdown_super(sb);
Al Viro0ee5dc62011-07-07 15:44:25 -0400897 free_anon_bdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898}
899
900EXPORT_SYMBOL(kill_anon_super);
901
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902void kill_litter_super(struct super_block *sb)
903{
904 if (sb->s_root)
905 d_genocide(sb->s_root);
906 kill_anon_super(sb);
907}
908
909EXPORT_SYMBOL(kill_litter_super);
910
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700911static int ns_test_super(struct super_block *sb, void *data)
912{
913 return sb->s_fs_info == data;
914}
915
916static int ns_set_super(struct super_block *sb, void *data)
917{
918 sb->s_fs_info = data;
919 return set_anon_super(sb, NULL);
920}
921
Al Viroceefda62010-07-26 13:16:50 +0400922struct dentry *mount_ns(struct file_system_type *fs_type, int flags,
923 void *data, int (*fill_super)(struct super_block *, void *, int))
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700924{
925 struct super_block *sb;
926
927 sb = sget(fs_type, ns_test_super, ns_set_super, data);
928 if (IS_ERR(sb))
Al Viroceefda62010-07-26 13:16:50 +0400929 return ERR_CAST(sb);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700930
931 if (!sb->s_root) {
932 int err;
933 sb->s_flags = flags;
934 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
935 if (err) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400936 deactivate_locked_super(sb);
Al Viroceefda62010-07-26 13:16:50 +0400937 return ERR_PTR(err);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700938 }
939
940 sb->s_flags |= MS_ACTIVE;
941 }
942
Al Viroceefda62010-07-26 13:16:50 +0400943 return dget(sb->s_root);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700944}
945
Al Viroceefda62010-07-26 13:16:50 +0400946EXPORT_SYMBOL(mount_ns);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700947
David Howells93614012006-09-30 20:45:40 +0200948#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949static int set_bdev_super(struct super_block *s, void *data)
950{
951 s->s_bdev = data;
952 s->s_dev = s->s_bdev->bd_dev;
Jens Axboe32a88aa2009-09-16 15:02:33 +0200953
954 /*
955 * We set the bdi here to the queue backing, file systems can
956 * overwrite this in ->fill_super()
957 */
958 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 return 0;
960}
961
962static int test_bdev_super(struct super_block *s, void *data)
963{
964 return (void *)s->s_bdev == data;
965}
966
Al Viro152a0832010-07-25 00:46:55 +0400967struct dentry *mount_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 int flags, const char *dev_name, void *data,
Al Viro152a0832010-07-25 00:46:55 +0400969 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970{
971 struct block_device *bdev;
972 struct super_block *s;
Tejun Heod4d77622010-11-13 11:55:18 +0100973 fmode_t mode = FMODE_READ | FMODE_EXCL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 int error = 0;
975
Al Viro30c40d22008-02-22 19:50:45 -0500976 if (!(flags & MS_RDONLY))
977 mode |= FMODE_WRITE;
978
Tejun Heod4d77622010-11-13 11:55:18 +0100979 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 if (IS_ERR(bdev))
Al Viro152a0832010-07-25 00:46:55 +0400981 return ERR_CAST(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
983 /*
984 * once the super is inserted into the list by sget, s_umount
985 * will protect the lockfs code from trying to start a snapshot
986 * while we are mounting
987 */
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200988 mutex_lock(&bdev->bd_fsfreeze_mutex);
989 if (bdev->bd_fsfreeze_count > 0) {
990 mutex_unlock(&bdev->bd_fsfreeze_mutex);
991 error = -EBUSY;
992 goto error_bdev;
993 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200995 mutex_unlock(&bdev->bd_fsfreeze_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700997 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998
999 if (s->s_root) {
1000 if ((flags ^ s->s_flags) & MS_RDONLY) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001001 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -07001002 error = -EBUSY;
1003 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 }
David Howells454e2392006-06-23 02:02:57 -07001005
Tejun Heo4f331f02010-07-20 15:18:07 -07001006 /*
1007 * s_umount nests inside bd_mutex during
Tejun Heoe525fd82010-11-13 11:55:17 +01001008 * __invalidate_device(). blkdev_put() acquires
1009 * bd_mutex and can't be called under s_umount. Drop
1010 * s_umount temporarily. This is safe as we're
1011 * holding an active reference.
Tejun Heo4f331f02010-07-20 15:18:07 -07001012 */
1013 up_write(&s->s_umount);
Tejun Heod4d77622010-11-13 11:55:18 +01001014 blkdev_put(bdev, mode);
Tejun Heo4f331f02010-07-20 15:18:07 -07001015 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 } else {
1017 char b[BDEVNAME_SIZE];
1018
Al Viro9e1f1de2011-06-03 18:24:58 -04001019 s->s_flags = flags | MS_NOSEC;
Al Viro30c40d22008-02-22 19:50:45 -05001020 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -08001022 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001023 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001025 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -07001026 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -08001027 }
David Howells454e2392006-06-23 02:02:57 -07001028
1029 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001030 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 }
1032
Al Viro152a0832010-07-25 00:46:55 +04001033 return dget(s->s_root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034
David Howells454e2392006-06-23 02:02:57 -07001035error_s:
1036 error = PTR_ERR(s);
1037error_bdev:
Tejun Heod4d77622010-11-13 11:55:18 +01001038 blkdev_put(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -07001039error:
Al Viro152a0832010-07-25 00:46:55 +04001040 return ERR_PTR(error);
1041}
1042EXPORT_SYMBOL(mount_bdev);
1043
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044void kill_block_super(struct super_block *sb)
1045{
1046 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -05001047 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048
H Hartley Sweetenddbaaf32009-04-29 20:14:57 -04001049 bdev->bd_super = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 generic_shutdown_super(sb);
1051 sync_blockdev(bdev);
Tejun Heod4d77622010-11-13 11:55:18 +01001052 WARN_ON_ONCE(!(mode & FMODE_EXCL));
Tejun Heoe525fd82010-11-13 11:55:17 +01001053 blkdev_put(bdev, mode | FMODE_EXCL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054}
1055
1056EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +02001057#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058
Al Viro3c26ff62010-07-25 11:46:36 +04001059struct dentry *mount_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 int flags, void *data,
Al Viro3c26ff62010-07-25 11:46:36 +04001061 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062{
1063 int error;
1064 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
1065
1066 if (IS_ERR(s))
Al Viro3c26ff62010-07-25 11:46:36 +04001067 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068
1069 s->s_flags = flags;
1070
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001071 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001073 deactivate_locked_super(s);
Al Viro3c26ff62010-07-25 11:46:36 +04001074 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 }
1076 s->s_flags |= MS_ACTIVE;
Al Viro3c26ff62010-07-25 11:46:36 +04001077 return dget(s->s_root);
1078}
1079EXPORT_SYMBOL(mount_nodev);
1080
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081static int compare_single(struct super_block *s, void *p)
1082{
1083 return 1;
1084}
1085
Al Virofc14f2f2010-07-25 01:48:30 +04001086struct dentry *mount_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 int flags, void *data,
Al Virofc14f2f2010-07-25 01:48:30 +04001088 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089{
1090 struct super_block *s;
1091 int error;
1092
1093 s = sget(fs_type, compare_single, set_anon_super, NULL);
1094 if (IS_ERR(s))
Al Virofc14f2f2010-07-25 01:48:30 +04001095 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 if (!s->s_root) {
1097 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001098 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001100 deactivate_locked_super(s);
Al Virofc14f2f2010-07-25 01:48:30 +04001101 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 }
1103 s->s_flags |= MS_ACTIVE;
Kay Sievers9329d1b2009-12-18 21:18:15 +01001104 } else {
1105 do_remount_sb(s, flags, data, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 }
Al Virofc14f2f2010-07-25 01:48:30 +04001107 return dget(s->s_root);
1108}
1109EXPORT_SYMBOL(mount_single);
1110
Al Viro9d412a42011-03-17 22:08:28 -04001111struct dentry *
1112mount_fs(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113{
Al Viroc96e41e2010-07-25 00:17:56 +04001114 struct dentry *root;
Al Viro9d412a42011-03-17 22:08:28 -04001115 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 char *secdata = NULL;
Al Viro9d412a42011-03-17 22:08:28 -04001117 int error = -ENOMEM;
Al Viro80893522010-02-05 09:30:46 -05001118
Eric Parise0007522008-03-05 10:31:54 -05001119 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -07001121 if (!secdata)
Al Viro9d412a42011-03-17 22:08:28 -04001122 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123
Eric Parise0007522008-03-05 10:31:54 -05001124 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -07001125 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 }
1128
Al Viro1a102ff2011-03-16 09:07:58 -04001129 root = type->mount(type, flags, name, data);
1130 if (IS_ERR(root)) {
1131 error = PTR_ERR(root);
1132 goto out_free_secdata;
Al Viroc96e41e2010-07-25 00:17:56 +04001133 }
Al Viro9d412a42011-03-17 22:08:28 -04001134 sb = root->d_sb;
1135 BUG_ON(!sb);
1136 WARN_ON(!sb->s_bdi);
Linus Torvalds6c510382011-03-24 10:16:26 -07001137 WARN_ON(sb->s_bdi == &default_backing_dev_info);
Al Viro9d412a42011-03-17 22:08:28 -04001138 sb->s_flags |= MS_BORN;
David Howells454e2392006-06-23 02:02:57 -07001139
Al Viro9d412a42011-03-17 22:08:28 -04001140 error = security_sb_kern_mount(sb, flags, secdata);
Jörn Engel5129a462010-04-25 08:54:42 +02001141 if (error)
1142 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -07001143
Jeff Layton42cb56a2009-09-18 13:05:53 -07001144 /*
1145 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
1146 * but s_maxbytes was an unsigned long long for many releases. Throw
1147 * this warning for a little while to try and catch filesystems that
Jeff Layton4358b562011-03-29 09:33:31 -04001148 * violate this rule.
Jeff Layton42cb56a2009-09-18 13:05:53 -07001149 */
Al Viro9d412a42011-03-17 22:08:28 -04001150 WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to "
1151 "negative value (%lld)\n", type->name, sb->s_maxbytes);
Jeff Layton42cb56a2009-09-18 13:05:53 -07001152
Al Viro9d412a42011-03-17 22:08:28 -04001153 up_write(&sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -07001154 free_secdata(secdata);
Al Viro9d412a42011-03-17 22:08:28 -04001155 return root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156out_sb:
Al Viro9d412a42011-03-17 22:08:28 -04001157 dput(root);
1158 deactivate_locked_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159out_free_secdata:
1160 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161out:
David Howells454e2392006-06-23 02:02:57 -07001162 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163}
1164
Josef Bacik18e9e512010-03-23 10:34:56 -04001165/**
Randy Dunlap7000d3c2010-05-24 22:22:34 -07001166 * freeze_super - lock the filesystem and force it into a consistent state
1167 * @sb: the super to lock
Josef Bacik18e9e512010-03-23 10:34:56 -04001168 *
1169 * Syncs the super to make sure the filesystem is consistent and calls the fs's
1170 * freeze_fs. Subsequent calls to this without first thawing the fs will return
1171 * -EBUSY.
1172 */
1173int freeze_super(struct super_block *sb)
1174{
1175 int ret;
1176
1177 atomic_inc(&sb->s_active);
1178 down_write(&sb->s_umount);
1179 if (sb->s_frozen) {
1180 deactivate_locked_super(sb);
1181 return -EBUSY;
1182 }
1183
Al Virodabe0dc2012-01-03 21:01:29 -05001184 if (!(sb->s_flags & MS_BORN)) {
1185 up_write(&sb->s_umount);
1186 return 0; /* sic - it's "nothing to do" */
1187 }
1188
Josef Bacik18e9e512010-03-23 10:34:56 -04001189 if (sb->s_flags & MS_RDONLY) {
1190 sb->s_frozen = SB_FREEZE_TRANS;
1191 smp_wmb();
1192 up_write(&sb->s_umount);
1193 return 0;
1194 }
1195
1196 sb->s_frozen = SB_FREEZE_WRITE;
1197 smp_wmb();
1198
1199 sync_filesystem(sb);
1200
1201 sb->s_frozen = SB_FREEZE_TRANS;
1202 smp_wmb();
1203
1204 sync_blockdev(sb->s_bdev);
1205 if (sb->s_op->freeze_fs) {
1206 ret = sb->s_op->freeze_fs(sb);
1207 if (ret) {
1208 printk(KERN_ERR
1209 "VFS:Filesystem freeze failed\n");
1210 sb->s_frozen = SB_UNFROZEN;
Kazuya Mioe1616302011-12-01 16:51:07 +09001211 smp_wmb();
1212 wake_up(&sb->s_wait_unfrozen);
Josef Bacik18e9e512010-03-23 10:34:56 -04001213 deactivate_locked_super(sb);
1214 return ret;
1215 }
1216 }
1217 up_write(&sb->s_umount);
1218 return 0;
1219}
1220EXPORT_SYMBOL(freeze_super);
1221
1222/**
1223 * thaw_super -- unlock filesystem
1224 * @sb: the super to thaw
1225 *
1226 * Unlocks the filesystem and marks it writeable again after freeze_super().
1227 */
1228int thaw_super(struct super_block *sb)
1229{
1230 int error;
1231
1232 down_write(&sb->s_umount);
1233 if (sb->s_frozen == SB_UNFROZEN) {
1234 up_write(&sb->s_umount);
1235 return -EINVAL;
1236 }
1237
1238 if (sb->s_flags & MS_RDONLY)
1239 goto out;
1240
1241 if (sb->s_op->unfreeze_fs) {
1242 error = sb->s_op->unfreeze_fs(sb);
1243 if (error) {
1244 printk(KERN_ERR
1245 "VFS:Filesystem thaw failed\n");
1246 sb->s_frozen = SB_FREEZE_TRANS;
1247 up_write(&sb->s_umount);
1248 return error;
1249 }
1250 }
1251
1252out:
1253 sb->s_frozen = SB_UNFROZEN;
1254 smp_wmb();
1255 wake_up(&sb->s_wait_unfrozen);
1256 deactivate_locked_super(sb);
1257
1258 return 0;
1259}
1260EXPORT_SYMBOL(thaw_super);