blob: cf001775617f5781ac4f43f5695ad7802c6b4524 [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>
Al Viro6d59e7f2008-03-22 15:48:17 -040036#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
38
Linus Torvalds1da177e2005-04-16 15:20:36 -070039LIST_HEAD(super_blocks);
40DEFINE_SPINLOCK(sb_lock);
41
Dave Chinnerb0d40c92011-07-08 14:14:42 +100042/*
43 * One thing we have to be careful of with a per-sb shrinker is that we don't
44 * drop the last active reference to the superblock from within the shrinker.
45 * If that happens we could trigger unregistering the shrinker from within the
46 * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we
47 * take a passive reference to the superblock to avoid this from occurring.
48 */
49static int prune_super(struct shrinker *shrink, struct shrink_control *sc)
50{
51 struct super_block *sb;
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100052 int fs_objects = 0;
53 int total_objects;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100054
55 sb = container_of(shrink, struct super_block, s_shrink);
56
57 /*
58 * Deadlock avoidance. We may hold various FS locks, and we don't want
59 * to recurse into the FS that called us in clear_inode() and friends..
60 */
61 if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS))
62 return -1;
63
64 if (!grab_super_passive(sb))
Mikulas Patocka09f363c2011-10-31 17:08:57 -070065 return !sc->nr_to_scan ? 0 : -1;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100066
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100067 if (sb->s_op && sb->s_op->nr_cached_objects)
68 fs_objects = sb->s_op->nr_cached_objects(sb);
69
70 total_objects = sb->s_nr_dentry_unused +
71 sb->s_nr_inodes_unused + fs_objects + 1;
72
Dave Chinnerb0d40c92011-07-08 14:14:42 +100073 if (sc->nr_to_scan) {
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100074 int dentries;
75 int inodes;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100076
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100077 /* proportion the scan between the caches */
78 dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) /
79 total_objects;
80 inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) /
81 total_objects;
82 if (fs_objects)
83 fs_objects = (sc->nr_to_scan * fs_objects) /
84 total_objects;
85 /*
86 * prune the dcache first as the icache is pinned by it, then
87 * prune the icache, followed by the filesystem specific caches
88 */
89 prune_dcache_sb(sb, dentries);
90 prune_icache_sb(sb, inodes);
Dave Chinnerb0d40c92011-07-08 14:14:42 +100091
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100092 if (fs_objects && sb->s_op->free_cached_objects) {
93 sb->s_op->free_cached_objects(sb, fs_objects);
94 fs_objects = sb->s_op->nr_cached_objects(sb);
95 }
96 total_objects = sb->s_nr_dentry_unused +
97 sb->s_nr_inodes_unused + fs_objects;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100098 }
99
Dave Chinner0e1fdaf2011-07-08 14:14:44 +1000100 total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000101 drop_super(sb);
Dave Chinner0e1fdaf2011-07-08 14:14:44 +1000102 return total_objects;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000103}
104
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105/**
106 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -0700107 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 *
109 * Allocates and initializes a new &struct super_block. alloc_super()
110 * returns a pointer new superblock or %NULL if allocation had failed.
111 */
Ingo Molnarcf516242006-07-03 00:25:27 -0700112static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -0800114 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700115 static const struct super_operations default_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
117 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118 if (security_sb_alloc(s)) {
119 kfree(s);
120 s = NULL;
121 goto out;
122 }
Nick Piggin6416ccb2010-08-18 04:37:38 +1000123#ifdef CONFIG_SMP
124 s->s_files = alloc_percpu(struct list_head);
125 if (!s->s_files) {
126 security_sb_free(s);
127 kfree(s);
128 s = NULL;
129 goto out;
130 } else {
131 int i;
132
133 for_each_possible_cpu(i)
134 INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i));
135 }
136#else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 INIT_LIST_HEAD(&s->s_files);
Nick Piggin6416ccb2010-08-18 04:37:38 +1000138#endif
Jens Axboe95f28602011-03-17 11:13:12 +0100139 s->s_bdi = &default_backing_dev_info;
Al Viroa5166162011-12-12 22:53:00 -0500140 INIT_HLIST_NODE(&s->s_instances);
Nick Pigginceb5bdc2011-01-07 17:50:05 +1100141 INIT_HLIST_BL_HEAD(&s->s_anon);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 INIT_LIST_HEAD(&s->s_inodes);
Kentaro Makitada3bbdd2008-07-23 21:27:13 -0700143 INIT_LIST_HEAD(&s->s_dentry_lru);
Dave Chinner98b745c2011-07-08 14:14:39 +1000144 INIT_LIST_HEAD(&s->s_inode_lru);
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000145 spin_lock_init(&s->s_inode_lru_lock);
Miklos Szeredi39f7c4d2011-11-21 12:11:30 +0100146 INIT_LIST_HEAD(&s->s_mounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f42006-01-09 15:59:25 -0800148 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -0700149 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -0700150 /*
151 * The locking rules for s_lock are up to the
152 * filesystem. For example ext3fs has different
153 * lock ordering than usbfs:
154 */
155 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Peter Zijlstraada723d2009-02-18 14:48:30 -0800156 /*
157 * sget() can have s_umount recursion.
158 *
159 * When it cannot find a suitable sb, it allocates a new
160 * one (this one), and tries again to find a suitable old
161 * one.
162 *
163 * In case that succeeds, it will acquire the s_umount
164 * lock of the old one. Since these are clearly distrinct
165 * locks, and this object isn't exposed yet, there's no
166 * risk of deadlocks.
167 *
168 * Annotate this by putting this lock in a different
169 * subclass.
170 */
171 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
Al Virob20bd1a2010-03-22 08:53:19 -0400172 s->s_count = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -0800174 mutex_init(&s->s_vfs_rename_mutex);
Roland Dreier51ee0492010-04-27 14:23:57 -0700175 lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key);
Ingo Molnard3be9152006-03-23 03:00:29 -0800176 mutex_init(&s->s_dquot.dqio_mutex);
177 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 init_rwsem(&s->s_dquot.dqptr_sem);
179 init_waitqueue_head(&s->s_wait_unfrozen);
180 s->s_maxbytes = MAX_NON_LFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 s->s_op = &default_op;
182 s->s_time_gran = 1000000000;
Dan Magenheimerc515e1f2011-05-26 10:01:43 -0600183 s->cleancache_poolid = -1;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000184
185 s->s_shrink.seeks = DEFAULT_SEEKS;
186 s->s_shrink.shrink = prune_super;
Dave Chinner8ab47662011-07-08 14:14:45 +1000187 s->s_shrink.batch = 1024;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 }
189out:
190 return s;
191}
192
193/**
194 * destroy_super - frees a superblock
195 * @s: superblock to free
196 *
197 * Frees a superblock.
198 */
199static inline void destroy_super(struct super_block *s)
200{
Nick Piggin6416ccb2010-08-18 04:37:38 +1000201#ifdef CONFIG_SMP
202 free_percpu(s->s_files);
203#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 security_sb_free(s);
Miklos Szeredi39f7c4d2011-11-21 12:11:30 +0100205 WARN_ON(!list_empty(&s->s_mounts));
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700206 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800207 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 kfree(s);
209}
210
211/* Superblock refcounting */
212
213/*
Al Viro35cf7ba2010-03-22 21:13:53 -0400214 * Drop a superblock's refcount. The caller must hold sb_lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 */
Al Virof47ec3f2011-11-21 21:15:42 -0500216static void __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 if (!--sb->s_count) {
Al Viro551de6f2010-03-22 19:36:35 -0400219 list_del_init(&sb->s_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 destroy_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222}
223
224/**
225 * put_super - drop a temporary reference to superblock
226 * @sb: superblock in question
227 *
228 * Drops a temporary reference, frees superblock if there's no
229 * references left.
230 */
Al Virof47ec3f2011-11-21 21:15:42 -0500231static void put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232{
233 spin_lock(&sb_lock);
234 __put_super(sb);
235 spin_unlock(&sb_lock);
236}
237
238
239/**
Al Viro74dbbdd2009-05-06 01:07:50 -0400240 * deactivate_locked_super - drop an active reference to superblock
241 * @s: superblock to deactivate
242 *
Al Viro1712ac82010-03-22 15:22:31 -0400243 * Drops an active reference to superblock, converting it into a temprory
244 * one if there is no other active references left. In that case we
245 * tell fs driver to shut it down and drop the temporary reference we
246 * had just acquired.
247 *
248 * Caller holds exclusive lock on superblock; that lock is released.
Al Viro74dbbdd2009-05-06 01:07:50 -0400249 */
250void deactivate_locked_super(struct super_block *s)
251{
252 struct file_system_type *fs = s->s_type;
Al Virob20bd1a2010-03-22 08:53:19 -0400253 if (atomic_dec_and_test(&s->s_active)) {
Dan Magenheimer31677602011-09-21 11:56:28 -0400254 cleancache_invalidate_fs(s);
Al Viro74dbbdd2009-05-06 01:07:50 -0400255 fs->kill_sb(s);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000256
257 /* caches are now gone, we can safely kill the shrinker now */
258 unregister_shrinker(&s->s_shrink);
259
Boaz Harroshd863b502011-02-10 15:01:20 -0800260 /*
261 * We need to call rcu_barrier so all the delayed rcu free
262 * inodes are flushed before we release the fs module.
263 */
264 rcu_barrier();
Al Viro74dbbdd2009-05-06 01:07:50 -0400265 put_filesystem(fs);
266 put_super(s);
267 } else {
268 up_write(&s->s_umount);
269 }
270}
271
272EXPORT_SYMBOL(deactivate_locked_super);
273
274/**
Al Viro1712ac82010-03-22 15:22:31 -0400275 * deactivate_super - drop an active reference to superblock
276 * @s: superblock to deactivate
277 *
278 * Variant of deactivate_locked_super(), except that superblock is *not*
279 * locked by caller. If we are going to drop the final active reference,
280 * lock will be acquired prior to that.
281 */
282void deactivate_super(struct super_block *s)
283{
284 if (!atomic_add_unless(&s->s_active, -1, 1)) {
285 down_write(&s->s_umount);
286 deactivate_locked_super(s);
287 }
288}
289
290EXPORT_SYMBOL(deactivate_super);
291
292/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 * grab_super - acquire an active reference
294 * @s: reference we are trying to make active
295 *
296 * Tries to acquire an active reference. grab_super() is used when we
297 * had just found a superblock in super_blocks or fs_type->fs_supers
298 * and want to turn it into a full-blown active reference. grab_super()
299 * is called with sb_lock held and drops it. Returns 1 in case of
300 * success, 0 if we had failed (superblock contents was already dead or
301 * dying when grab_super() had been called).
302 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700303static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304{
Al Virob20bd1a2010-03-22 08:53:19 -0400305 if (atomic_inc_not_zero(&s->s_active)) {
306 spin_unlock(&sb_lock);
Al Virob20bd1a2010-03-22 08:53:19 -0400307 return 1;
308 }
309 /* it's going away */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 s->s_count++;
311 spin_unlock(&sb_lock);
Al Viro1712ac82010-03-22 15:22:31 -0400312 /* wait for it to die */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 up_write(&s->s_umount);
315 put_super(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 return 0;
317}
318
David Howellscf9a2ae2006-08-29 19:05:54 +0100319/*
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000320 * grab_super_passive - acquire a passive reference
321 * @s: reference we are trying to grab
322 *
323 * Tries to acquire a passive reference. This is used in places where we
324 * cannot take an active reference but we need to ensure that the
325 * superblock does not go away while we are working on it. It returns
326 * false if a reference was not gained, and returns true with the s_umount
327 * lock held in read mode if a reference is gained. On successful return,
328 * the caller must drop the s_umount lock and the passive reference when
329 * done.
330 */
331bool grab_super_passive(struct super_block *sb)
332{
333 spin_lock(&sb_lock);
Al Viroa5166162011-12-12 22:53:00 -0500334 if (hlist_unhashed(&sb->s_instances)) {
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000335 spin_unlock(&sb_lock);
336 return false;
337 }
338
339 sb->s_count++;
340 spin_unlock(&sb_lock);
341
342 if (down_read_trylock(&sb->s_umount)) {
Al Virodabe0dc2012-01-03 21:01:29 -0500343 if (sb->s_root && (sb->s_flags & MS_BORN))
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000344 return true;
345 up_read(&sb->s_umount);
346 }
347
348 put_super(sb);
349 return false;
350}
351
352/*
Al Viro914e2632006-10-18 13:55:46 -0400353 * Superblock locking. We really ought to get rid of these two.
354 */
355void lock_super(struct super_block * sb)
356{
Al Viro914e2632006-10-18 13:55:46 -0400357 mutex_lock(&sb->s_lock);
358}
359
360void unlock_super(struct super_block * sb)
361{
Al Viro914e2632006-10-18 13:55:46 -0400362 mutex_unlock(&sb->s_lock);
363}
364
365EXPORT_SYMBOL(lock_super);
366EXPORT_SYMBOL(unlock_super);
367
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368/**
369 * generic_shutdown_super - common helper for ->kill_sb()
370 * @sb: superblock to kill
371 *
372 * generic_shutdown_super() does all fs-independent work on superblock
373 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
374 * that need destruction out of superblock, call generic_shutdown_super()
375 * and release aforementioned objects. Note: dentries and inodes _are_
376 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700377 *
378 * Upon calling this function, the filesystem may no longer alter or
379 * rearrange the set of dentries belonging to this super_block, nor may it
380 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 */
382void generic_shutdown_super(struct super_block *sb)
383{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800384 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385
David Howellsc636ebd2006-10-11 01:22:19 -0700386 if (sb->s_root) {
387 shrink_dcache_for_umount(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200388 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 sb->s_flags &= ~MS_ACTIVE;
Arjan van de Venefaee192009-01-06 07:20:54 -0800390
Al Viro63997e92010-10-25 20:49:35 -0400391 fsnotify_unmount_inodes(&sb->s_inodes);
392
393 evict_inodes(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 if (sop->put_super)
396 sop->put_super(sb);
397
Al Viro63997e92010-10-25 20:49:35 -0400398 if (!list_empty(&sb->s_inodes)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800399 printk("VFS: Busy inodes after unmount of %s. "
400 "Self-destruct in 5 seconds. Have a nice day...\n",
401 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 }
404 spin_lock(&sb_lock);
405 /* should be initialized for __put_super_and_need_restart() */
Al Viroa5166162011-12-12 22:53:00 -0500406 hlist_del_init(&sb->s_instances);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 spin_unlock(&sb_lock);
408 up_write(&sb->s_umount);
409}
410
411EXPORT_SYMBOL(generic_shutdown_super);
412
413/**
414 * sget - find or create a superblock
415 * @type: filesystem type superblock should belong to
416 * @test: comparison callback
417 * @set: setup callback
418 * @data: argument to each of them
419 */
420struct super_block *sget(struct file_system_type *type,
421 int (*test)(struct super_block *,void *),
422 int (*set)(struct super_block *,void *),
423 void *data)
424{
425 struct super_block *s = NULL;
Al Viroa5166162011-12-12 22:53:00 -0500426 struct hlist_node *node;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700427 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 int err;
429
430retry:
431 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700432 if (test) {
Al Viroa5166162011-12-12 22:53:00 -0500433 hlist_for_each_entry(old, node, &type->fs_supers, s_instances) {
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700434 if (!test(old, data))
435 continue;
436 if (!grab_super(old))
437 goto retry;
Li Zefana3cfbb52009-03-12 14:31:29 -0700438 if (s) {
439 up_write(&s->s_umount);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700440 destroy_super(s);
Al Viro7a4dec52010-08-09 12:05:43 -0400441 s = NULL;
Li Zefana3cfbb52009-03-12 14:31:29 -0700442 }
Al Virod3f21472010-03-23 11:11:05 -0400443 down_write(&old->s_umount);
Al Viro7a4dec52010-08-09 12:05:43 -0400444 if (unlikely(!(old->s_flags & MS_BORN))) {
445 deactivate_locked_super(old);
446 goto retry;
447 }
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700448 return old;
449 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 }
451 if (!s) {
452 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700453 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 if (!s)
455 return ERR_PTR(-ENOMEM);
456 goto retry;
457 }
458
459 err = set(s, data);
460 if (err) {
461 spin_unlock(&sb_lock);
Li Zefana3cfbb52009-03-12 14:31:29 -0700462 up_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 destroy_super(s);
464 return ERR_PTR(err);
465 }
466 s->s_type = type;
467 strlcpy(s->s_id, type->name, sizeof(s->s_id));
468 list_add_tail(&s->s_list, &super_blocks);
Al Viroa5166162011-12-12 22:53:00 -0500469 hlist_add_head(&s->s_instances, &type->fs_supers);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 spin_unlock(&sb_lock);
471 get_filesystem(type);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000472 register_shrinker(&s->s_shrink);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 return s;
474}
475
476EXPORT_SYMBOL(sget);
477
478void drop_super(struct super_block *sb)
479{
480 up_read(&sb->s_umount);
481 put_super(sb);
482}
483
484EXPORT_SYMBOL(drop_super);
485
Christoph Hellwige5004752009-05-05 16:08:56 +0200486/**
487 * sync_supers - helper for periodic superblock writeback
488 *
489 * Call the write_super method if present on all dirty superblocks in
490 * the system. This is for the periodic writeback used by most older
491 * filesystems. For data integrity superblock writeback use
492 * sync_filesystems() instead.
493 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 * Note: check the dirty flag before waiting, so we don't
495 * hold up the sync while mounting a device. (The newly
496 * mounted device won't need syncing.)
497 */
498void sync_supers(void)
499{
Al Virodca33252010-07-25 02:31:46 +0400500 struct super_block *sb, *p = NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700501
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 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 Viro551de6f2010-03-22 19:36:35 -0400505 continue;
Christoph Hellwige5004752009-05-05 16:08:56 +0200506 if (sb->s_op->write_super && sb->s_dirt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 sb->s_count++;
508 spin_unlock(&sb_lock);
Christoph Hellwige5004752009-05-05 16:08:56 +0200509
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 down_read(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500511 if (sb->s_root && sb->s_dirt && (sb->s_flags & MS_BORN))
Christoph Hellwige5004752009-05-05 16:08:56 +0200512 sb->s_op->write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700513 up_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200514
Kirill Korotaev618f0632005-06-23 00:09:54 -0700515 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400516 if (p)
517 __put_super(p);
518 p = sb;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700519 }
520 }
Al Virodca33252010-07-25 02:31:46 +0400521 if (p)
522 __put_super(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 spin_unlock(&sb_lock);
524}
525
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526/**
Al Viro01a05b32010-03-23 06:06:58 -0400527 * iterate_supers - call function for all active superblocks
528 * @f: function to call
529 * @arg: argument to pass to it
530 *
531 * Scans the superblock list and calls given function, passing it
532 * locked superblock and given argument.
533 */
534void iterate_supers(void (*f)(struct super_block *, void *), void *arg)
535{
Al Virodca33252010-07-25 02:31:46 +0400536 struct super_block *sb, *p = NULL;
Al Viro01a05b32010-03-23 06:06:58 -0400537
538 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400539 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500540 if (hlist_unhashed(&sb->s_instances))
Al Viro01a05b32010-03-23 06:06:58 -0400541 continue;
542 sb->s_count++;
543 spin_unlock(&sb_lock);
544
545 down_read(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500546 if (sb->s_root && (sb->s_flags & MS_BORN))
Al Viro01a05b32010-03-23 06:06:58 -0400547 f(sb, arg);
548 up_read(&sb->s_umount);
549
550 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400551 if (p)
552 __put_super(p);
553 p = sb;
Al Viro01a05b32010-03-23 06:06:58 -0400554 }
Al Virodca33252010-07-25 02:31:46 +0400555 if (p)
556 __put_super(p);
Al Viro01a05b32010-03-23 06:06:58 -0400557 spin_unlock(&sb_lock);
558}
559
560/**
Al Viro43e15cd2011-06-03 20:16:57 -0400561 * iterate_supers_type - call function for superblocks of given type
562 * @type: fs type
563 * @f: function to call
564 * @arg: argument to pass to it
565 *
566 * Scans the superblock list and calls given function, passing it
567 * locked superblock and given argument.
568 */
569void iterate_supers_type(struct file_system_type *type,
570 void (*f)(struct super_block *, void *), void *arg)
571{
572 struct super_block *sb, *p = NULL;
Al Viroa5166162011-12-12 22:53:00 -0500573 struct hlist_node *node;
Al Viro43e15cd2011-06-03 20:16:57 -0400574
575 spin_lock(&sb_lock);
Al Viroa5166162011-12-12 22:53:00 -0500576 hlist_for_each_entry(sb, node, &type->fs_supers, s_instances) {
Al Viro43e15cd2011-06-03 20:16:57 -0400577 sb->s_count++;
578 spin_unlock(&sb_lock);
579
580 down_read(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500581 if (sb->s_root && (sb->s_flags & MS_BORN))
Al Viro43e15cd2011-06-03 20:16:57 -0400582 f(sb, arg);
583 up_read(&sb->s_umount);
584
585 spin_lock(&sb_lock);
586 if (p)
587 __put_super(p);
588 p = sb;
589 }
590 if (p)
591 __put_super(p);
592 spin_unlock(&sb_lock);
593}
594
595EXPORT_SYMBOL(iterate_supers_type);
596
597/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 * get_super - get the superblock of a device
599 * @bdev: device to get the superblock for
600 *
601 * Scans the superblock list and finds the superblock of the file system
602 * mounted on the device given. %NULL is returned if no match is found.
603 */
604
Al Virodf40c012010-03-22 20:23:25 -0400605struct super_block *get_super(struct block_device *bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700607 struct super_block *sb;
608
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 if (!bdev)
610 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700611
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700613rescan:
614 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500615 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400616 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700617 if (sb->s_bdev == bdev) {
618 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700620 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400621 /* still alive? */
Al Virodabe0dc2012-01-03 21:01:29 -0500622 if (sb->s_root && (sb->s_flags & MS_BORN))
Kirill Korotaev618f0632005-06-23 00:09:54 -0700623 return sb;
624 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400625 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700626 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400627 __put_super(sb);
628 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 }
630 }
631 spin_unlock(&sb_lock);
632 return NULL;
633}
634
635EXPORT_SYMBOL(get_super);
Christoph Hellwig45042302009-08-03 23:28:35 +0200636
637/**
Jan Kara6b6dc832012-02-10 11:03:00 +0100638 * get_super_thawed - get thawed superblock of a device
639 * @bdev: device to get the superblock for
640 *
641 * Scans the superblock list and finds the superblock of the file system
642 * mounted on the device. The superblock is returned once it is thawed
643 * (or immediately if it was not frozen). %NULL is returned if no match
644 * is found.
645 */
646struct super_block *get_super_thawed(struct block_device *bdev)
647{
648 while (1) {
649 struct super_block *s = get_super(bdev);
650 if (!s || s->s_frozen == SB_UNFROZEN)
651 return s;
652 up_read(&s->s_umount);
653 vfs_check_frozen(s, SB_FREEZE_WRITE);
654 put_super(s);
655 }
656}
657EXPORT_SYMBOL(get_super_thawed);
658
659/**
Christoph Hellwig45042302009-08-03 23:28:35 +0200660 * get_active_super - get an active reference to the superblock of a device
661 * @bdev: device to get the superblock for
662 *
663 * Scans the superblock list and finds the superblock of the file system
664 * mounted on the device given. Returns the superblock with an active
Al Virod3f21472010-03-23 11:11:05 -0400665 * reference or %NULL if none was found.
Christoph Hellwig45042302009-08-03 23:28:35 +0200666 */
667struct super_block *get_active_super(struct block_device *bdev)
668{
669 struct super_block *sb;
670
671 if (!bdev)
672 return NULL;
673
Al Viro14945832010-03-22 20:15:33 -0400674restart:
Christoph Hellwig45042302009-08-03 23:28:35 +0200675 spin_lock(&sb_lock);
676 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500677 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400678 continue;
Al Viro14945832010-03-22 20:15:33 -0400679 if (sb->s_bdev == bdev) {
680 if (grab_super(sb)) /* drops sb_lock */
681 return sb;
682 else
683 goto restart;
684 }
Christoph Hellwig45042302009-08-03 23:28:35 +0200685 }
686 spin_unlock(&sb_lock);
687 return NULL;
688}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689
Al Virodf40c012010-03-22 20:23:25 -0400690struct super_block *user_get_super(dev_t dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700692 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700695rescan:
696 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500697 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400698 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700699 if (sb->s_dev == dev) {
700 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700702 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400703 /* still alive? */
Al Virodabe0dc2012-01-03 21:01:29 -0500704 if (sb->s_root && (sb->s_flags & MS_BORN))
Kirill Korotaev618f0632005-06-23 00:09:54 -0700705 return sb;
706 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400707 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700708 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400709 __put_super(sb);
710 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 }
712 }
713 spin_unlock(&sb_lock);
714 return NULL;
715}
716
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 * do_remount_sb - asks filesystem to change mount options.
719 * @sb: superblock in question
720 * @flags: numeric part of options
721 * @data: the rest of options
722 * @force: whether or not to force the change
723 *
724 * Alters the mount options of a mounted file system.
725 */
726int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
727{
728 int retval;
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400729 int remount_ro;
Christoph Hellwig45042302009-08-03 23:28:35 +0200730
731 if (sb->s_frozen != SB_UNFROZEN)
732 return -EBUSY;
733
David Howells93614012006-09-30 20:45:40 +0200734#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
736 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200737#endif
Christoph Hellwig45042302009-08-03 23:28:35 +0200738
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 if (flags & MS_RDONLY)
740 acct_auto_close(sb);
741 shrink_dcache_sb(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200742 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
Nick Piggind208bbd2009-12-21 16:28:53 -0800744 remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY);
Nick Piggind208bbd2009-12-21 16:28:53 -0800745
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 /* If we are remounting RDONLY and current sb is read/write,
747 make sure there are no rw files opened */
Nick Piggind208bbd2009-12-21 16:28:53 -0800748 if (remount_ro) {
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100749 if (force) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 mark_files_ro(sb);
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100751 } else {
752 retval = sb_prepare_remount_readonly(sb);
753 if (retval)
754 return retval;
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100755 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 }
757
758 if (sb->s_op->remount_fs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 retval = sb->s_op->remount_fs(sb, &flags, data);
Miklos Szeredi2833eb22011-10-28 14:13:20 +0200760 if (retval) {
761 if (!force)
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100762 goto cancel_readonly;
Miklos Szeredi2833eb22011-10-28 14:13:20 +0200763 /* If forced remount, go ahead despite any errors */
764 WARN(1, "forced remount of a %s fs returned %i\n",
765 sb->s_type->name, retval);
766 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 }
768 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100769 /* Needs to be ordered wrt mnt_is_readonly() */
770 smp_wmb();
771 sb->s_readonly_remount = 0;
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400772
Nick Piggind208bbd2009-12-21 16:28:53 -0800773 /*
774 * Some filesystems modify their metadata via some other path than the
775 * bdev buffer cache (eg. use a private mapping, or directories in
776 * pagecache, etc). Also file data modifications go via their own
777 * mappings. So If we try to mount readonly then copy the filesystem
778 * from bdev, we could get stale data, so invalidate it to give a best
779 * effort at coherency.
780 */
781 if (remount_ro && sb->s_bdev)
782 invalidate_bdev(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 return 0;
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100784
785cancel_readonly:
786 sb->s_readonly_remount = 0;
787 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788}
789
Jens Axboea2a95372009-03-17 09:38:40 +0100790static void do_emergency_remount(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791{
Al Virodca33252010-07-25 02:31:46 +0400792 struct super_block *sb, *p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
794 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400795 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500796 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400797 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 sb->s_count++;
799 spin_unlock(&sb_lock);
Al Viro443b94b2009-05-05 23:48:50 -0400800 down_write(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500801 if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) &&
802 !(sb->s_flags & MS_RDONLY)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 * What lock protects sb->s_flags??
805 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 do_remount_sb(sb, MS_RDONLY, NULL, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 }
Al Viro443b94b2009-05-05 23:48:50 -0400808 up_write(&sb->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400810 if (p)
811 __put_super(p);
812 p = sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 }
Al Virodca33252010-07-25 02:31:46 +0400814 if (p)
815 __put_super(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 spin_unlock(&sb_lock);
Jens Axboea2a95372009-03-17 09:38:40 +0100817 kfree(work);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 printk("Emergency Remount complete\n");
819}
820
821void emergency_remount(void)
822{
Jens Axboea2a95372009-03-17 09:38:40 +0100823 struct work_struct *work;
824
825 work = kmalloc(sizeof(*work), GFP_ATOMIC);
826 if (work) {
827 INIT_WORK(work, do_emergency_remount);
828 schedule_work(work);
829 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830}
831
832/*
833 * Unnamed block devices are dummy devices used by virtual
834 * filesystems which don't use real block-devices. -- jrs
835 */
836
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400837static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
Al Viroc63e09e2009-06-24 02:05:18 -0400839static int unnamed_dev_start = 0; /* don't bother trying below it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840
Al Viro0ee5dc62011-07-07 15:44:25 -0400841int get_anon_bdev(dev_t *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842{
843 int dev;
844 int error;
845
846 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400847 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 return -ENOMEM;
849 spin_lock(&unnamed_dev_lock);
Al Viroc63e09e2009-06-24 02:05:18 -0400850 error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
Al Virof21f6222009-06-24 03:12:00 -0400851 if (!error)
852 unnamed_dev_start = dev + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 spin_unlock(&unnamed_dev_lock);
854 if (error == -EAGAIN)
855 /* We raced and lost with another CPU. */
856 goto retry;
857 else if (error)
858 return -EAGAIN;
859
860 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
861 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400862 ida_remove(&unnamed_dev_ida, dev);
Al Virof21f6222009-06-24 03:12:00 -0400863 if (unnamed_dev_start > dev)
864 unnamed_dev_start = dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 spin_unlock(&unnamed_dev_lock);
866 return -EMFILE;
867 }
Al Viro0ee5dc62011-07-07 15:44:25 -0400868 *p = MKDEV(0, dev & MINORMASK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 return 0;
870}
Al Viro0ee5dc62011-07-07 15:44:25 -0400871EXPORT_SYMBOL(get_anon_bdev);
872
873void free_anon_bdev(dev_t dev)
874{
875 int slot = MINOR(dev);
876 spin_lock(&unnamed_dev_lock);
877 ida_remove(&unnamed_dev_ida, slot);
878 if (slot < unnamed_dev_start)
879 unnamed_dev_start = slot;
880 spin_unlock(&unnamed_dev_lock);
881}
882EXPORT_SYMBOL(free_anon_bdev);
883
884int set_anon_super(struct super_block *s, void *data)
885{
886 int error = get_anon_bdev(&s->s_dev);
887 if (!error)
888 s->s_bdi = &noop_backing_dev_info;
889 return error;
890}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891
892EXPORT_SYMBOL(set_anon_super);
893
894void kill_anon_super(struct super_block *sb)
895{
Al Viro0ee5dc62011-07-07 15:44:25 -0400896 dev_t dev = sb->s_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 generic_shutdown_super(sb);
Al Viro0ee5dc62011-07-07 15:44:25 -0400898 free_anon_bdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899}
900
901EXPORT_SYMBOL(kill_anon_super);
902
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903void kill_litter_super(struct super_block *sb)
904{
905 if (sb->s_root)
906 d_genocide(sb->s_root);
907 kill_anon_super(sb);
908}
909
910EXPORT_SYMBOL(kill_litter_super);
911
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700912static int ns_test_super(struct super_block *sb, void *data)
913{
914 return sb->s_fs_info == data;
915}
916
917static int ns_set_super(struct super_block *sb, void *data)
918{
919 sb->s_fs_info = data;
920 return set_anon_super(sb, NULL);
921}
922
Al Viroceefda62010-07-26 13:16:50 +0400923struct dentry *mount_ns(struct file_system_type *fs_type, int flags,
924 void *data, int (*fill_super)(struct super_block *, void *, int))
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700925{
926 struct super_block *sb;
927
928 sb = sget(fs_type, ns_test_super, ns_set_super, data);
929 if (IS_ERR(sb))
Al Viroceefda62010-07-26 13:16:50 +0400930 return ERR_CAST(sb);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700931
932 if (!sb->s_root) {
933 int err;
934 sb->s_flags = flags;
935 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
936 if (err) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400937 deactivate_locked_super(sb);
Al Viroceefda62010-07-26 13:16:50 +0400938 return ERR_PTR(err);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700939 }
940
941 sb->s_flags |= MS_ACTIVE;
942 }
943
Al Viroceefda62010-07-26 13:16:50 +0400944 return dget(sb->s_root);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700945}
946
Al Viroceefda62010-07-26 13:16:50 +0400947EXPORT_SYMBOL(mount_ns);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700948
David Howells93614012006-09-30 20:45:40 +0200949#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950static int set_bdev_super(struct super_block *s, void *data)
951{
952 s->s_bdev = data;
953 s->s_dev = s->s_bdev->bd_dev;
Jens Axboe32a88aa2009-09-16 15:02:33 +0200954
955 /*
956 * We set the bdi here to the queue backing, file systems can
957 * overwrite this in ->fill_super()
958 */
959 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 return 0;
961}
962
963static int test_bdev_super(struct super_block *s, void *data)
964{
965 return (void *)s->s_bdev == data;
966}
967
Al Viro152a0832010-07-25 00:46:55 +0400968struct dentry *mount_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 int flags, const char *dev_name, void *data,
Al Viro152a0832010-07-25 00:46:55 +0400970 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971{
972 struct block_device *bdev;
973 struct super_block *s;
Tejun Heod4d77622010-11-13 11:55:18 +0100974 fmode_t mode = FMODE_READ | FMODE_EXCL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 int error = 0;
976
Al Viro30c40d22008-02-22 19:50:45 -0500977 if (!(flags & MS_RDONLY))
978 mode |= FMODE_WRITE;
979
Tejun Heod4d77622010-11-13 11:55:18 +0100980 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 if (IS_ERR(bdev))
Al Viro152a0832010-07-25 00:46:55 +0400982 return ERR_CAST(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983
984 /*
985 * once the super is inserted into the list by sget, s_umount
986 * will protect the lockfs code from trying to start a snapshot
987 * while we are mounting
988 */
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200989 mutex_lock(&bdev->bd_fsfreeze_mutex);
990 if (bdev->bd_fsfreeze_count > 0) {
991 mutex_unlock(&bdev->bd_fsfreeze_mutex);
992 error = -EBUSY;
993 goto error_bdev;
994 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200996 mutex_unlock(&bdev->bd_fsfreeze_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700998 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
1000 if (s->s_root) {
1001 if ((flags ^ s->s_flags) & MS_RDONLY) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001002 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -07001003 error = -EBUSY;
1004 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 }
David Howells454e2392006-06-23 02:02:57 -07001006
Tejun Heo4f331f02010-07-20 15:18:07 -07001007 /*
1008 * s_umount nests inside bd_mutex during
Tejun Heoe525fd82010-11-13 11:55:17 +01001009 * __invalidate_device(). blkdev_put() acquires
1010 * bd_mutex and can't be called under s_umount. Drop
1011 * s_umount temporarily. This is safe as we're
1012 * holding an active reference.
Tejun Heo4f331f02010-07-20 15:18:07 -07001013 */
1014 up_write(&s->s_umount);
Tejun Heod4d77622010-11-13 11:55:18 +01001015 blkdev_put(bdev, mode);
Tejun Heo4f331f02010-07-20 15:18:07 -07001016 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 } else {
1018 char b[BDEVNAME_SIZE];
1019
Al Viro9e1f1de2011-06-03 18:24:58 -04001020 s->s_flags = flags | MS_NOSEC;
Al Viro30c40d22008-02-22 19:50:45 -05001021 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -08001023 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001024 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001026 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -07001027 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -08001028 }
David Howells454e2392006-06-23 02:02:57 -07001029
1030 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001031 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 }
1033
Al Viro152a0832010-07-25 00:46:55 +04001034 return dget(s->s_root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035
David Howells454e2392006-06-23 02:02:57 -07001036error_s:
1037 error = PTR_ERR(s);
1038error_bdev:
Tejun Heod4d77622010-11-13 11:55:18 +01001039 blkdev_put(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -07001040error:
Al Viro152a0832010-07-25 00:46:55 +04001041 return ERR_PTR(error);
1042}
1043EXPORT_SYMBOL(mount_bdev);
1044
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045void kill_block_super(struct super_block *sb)
1046{
1047 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -05001048 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049
H Hartley Sweetenddbaaf32009-04-29 20:14:57 -04001050 bdev->bd_super = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 generic_shutdown_super(sb);
1052 sync_blockdev(bdev);
Tejun Heod4d77622010-11-13 11:55:18 +01001053 WARN_ON_ONCE(!(mode & FMODE_EXCL));
Tejun Heoe525fd82010-11-13 11:55:17 +01001054 blkdev_put(bdev, mode | FMODE_EXCL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055}
1056
1057EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +02001058#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059
Al Viro3c26ff62010-07-25 11:46:36 +04001060struct dentry *mount_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 int flags, void *data,
Al Viro3c26ff62010-07-25 11:46:36 +04001062 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063{
1064 int error;
1065 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
1066
1067 if (IS_ERR(s))
Al Viro3c26ff62010-07-25 11:46:36 +04001068 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069
1070 s->s_flags = flags;
1071
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001072 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001074 deactivate_locked_super(s);
Al Viro3c26ff62010-07-25 11:46:36 +04001075 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 }
1077 s->s_flags |= MS_ACTIVE;
Al Viro3c26ff62010-07-25 11:46:36 +04001078 return dget(s->s_root);
1079}
1080EXPORT_SYMBOL(mount_nodev);
1081
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082static int compare_single(struct super_block *s, void *p)
1083{
1084 return 1;
1085}
1086
Al Virofc14f2f2010-07-25 01:48:30 +04001087struct dentry *mount_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 int flags, void *data,
Al Virofc14f2f2010-07-25 01:48:30 +04001089 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090{
1091 struct super_block *s;
1092 int error;
1093
1094 s = sget(fs_type, compare_single, set_anon_super, NULL);
1095 if (IS_ERR(s))
Al Virofc14f2f2010-07-25 01:48:30 +04001096 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 if (!s->s_root) {
1098 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001099 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001101 deactivate_locked_super(s);
Al Virofc14f2f2010-07-25 01:48:30 +04001102 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 }
1104 s->s_flags |= MS_ACTIVE;
Kay Sievers9329d1b2009-12-18 21:18:15 +01001105 } else {
1106 do_remount_sb(s, flags, data, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 }
Al Virofc14f2f2010-07-25 01:48:30 +04001108 return dget(s->s_root);
1109}
1110EXPORT_SYMBOL(mount_single);
1111
Al Viro9d412a42011-03-17 22:08:28 -04001112struct dentry *
1113mount_fs(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114{
Al Viroc96e41e2010-07-25 00:17:56 +04001115 struct dentry *root;
Al Viro9d412a42011-03-17 22:08:28 -04001116 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 char *secdata = NULL;
Al Viro9d412a42011-03-17 22:08:28 -04001118 int error = -ENOMEM;
Al Viro80893522010-02-05 09:30:46 -05001119
Eric Parise0007522008-03-05 10:31:54 -05001120 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -07001122 if (!secdata)
Al Viro9d412a42011-03-17 22:08:28 -04001123 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
Eric Parise0007522008-03-05 10:31:54 -05001125 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -07001126 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 }
1129
Al Viro1a102ff2011-03-16 09:07:58 -04001130 root = type->mount(type, flags, name, data);
1131 if (IS_ERR(root)) {
1132 error = PTR_ERR(root);
1133 goto out_free_secdata;
Al Viroc96e41e2010-07-25 00:17:56 +04001134 }
Al Viro9d412a42011-03-17 22:08:28 -04001135 sb = root->d_sb;
1136 BUG_ON(!sb);
1137 WARN_ON(!sb->s_bdi);
Linus Torvalds6c510382011-03-24 10:16:26 -07001138 WARN_ON(sb->s_bdi == &default_backing_dev_info);
Al Viro9d412a42011-03-17 22:08:28 -04001139 sb->s_flags |= MS_BORN;
David Howells454e2392006-06-23 02:02:57 -07001140
Al Viro9d412a42011-03-17 22:08:28 -04001141 error = security_sb_kern_mount(sb, flags, secdata);
Jörn Engel5129a462010-04-25 08:54:42 +02001142 if (error)
1143 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -07001144
Jeff Layton42cb56a2009-09-18 13:05:53 -07001145 /*
1146 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
1147 * but s_maxbytes was an unsigned long long for many releases. Throw
1148 * this warning for a little while to try and catch filesystems that
Jeff Layton4358b562011-03-29 09:33:31 -04001149 * violate this rule.
Jeff Layton42cb56a2009-09-18 13:05:53 -07001150 */
Al Viro9d412a42011-03-17 22:08:28 -04001151 WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to "
1152 "negative value (%lld)\n", type->name, sb->s_maxbytes);
Jeff Layton42cb56a2009-09-18 13:05:53 -07001153
Al Viro9d412a42011-03-17 22:08:28 -04001154 up_write(&sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -07001155 free_secdata(secdata);
Al Viro9d412a42011-03-17 22:08:28 -04001156 return root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157out_sb:
Al Viro9d412a42011-03-17 22:08:28 -04001158 dput(root);
1159 deactivate_locked_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160out_free_secdata:
1161 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162out:
David Howells454e2392006-06-23 02:02:57 -07001163 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164}
1165
Josef Bacik18e9e512010-03-23 10:34:56 -04001166/**
Randy Dunlap7000d3c2010-05-24 22:22:34 -07001167 * freeze_super - lock the filesystem and force it into a consistent state
1168 * @sb: the super to lock
Josef Bacik18e9e512010-03-23 10:34:56 -04001169 *
1170 * Syncs the super to make sure the filesystem is consistent and calls the fs's
1171 * freeze_fs. Subsequent calls to this without first thawing the fs will return
1172 * -EBUSY.
1173 */
1174int freeze_super(struct super_block *sb)
1175{
1176 int ret;
1177
1178 atomic_inc(&sb->s_active);
1179 down_write(&sb->s_umount);
1180 if (sb->s_frozen) {
1181 deactivate_locked_super(sb);
1182 return -EBUSY;
1183 }
1184
Al Virodabe0dc2012-01-03 21:01:29 -05001185 if (!(sb->s_flags & MS_BORN)) {
1186 up_write(&sb->s_umount);
1187 return 0; /* sic - it's "nothing to do" */
1188 }
1189
Josef Bacik18e9e512010-03-23 10:34:56 -04001190 if (sb->s_flags & MS_RDONLY) {
1191 sb->s_frozen = SB_FREEZE_TRANS;
1192 smp_wmb();
1193 up_write(&sb->s_umount);
1194 return 0;
1195 }
1196
1197 sb->s_frozen = SB_FREEZE_WRITE;
1198 smp_wmb();
1199
1200 sync_filesystem(sb);
1201
1202 sb->s_frozen = SB_FREEZE_TRANS;
1203 smp_wmb();
1204
1205 sync_blockdev(sb->s_bdev);
1206 if (sb->s_op->freeze_fs) {
1207 ret = sb->s_op->freeze_fs(sb);
1208 if (ret) {
1209 printk(KERN_ERR
1210 "VFS:Filesystem freeze failed\n");
1211 sb->s_frozen = SB_UNFROZEN;
Kazuya Mioe1616302011-12-01 16:51:07 +09001212 smp_wmb();
1213 wake_up(&sb->s_wait_unfrozen);
Josef Bacik18e9e512010-03-23 10:34:56 -04001214 deactivate_locked_super(sb);
1215 return ret;
1216 }
1217 }
1218 up_write(&sb->s_umount);
1219 return 0;
1220}
1221EXPORT_SYMBOL(freeze_super);
1222
1223/**
1224 * thaw_super -- unlock filesystem
1225 * @sb: the super to thaw
1226 *
1227 * Unlocks the filesystem and marks it writeable again after freeze_super().
1228 */
1229int thaw_super(struct super_block *sb)
1230{
1231 int error;
1232
1233 down_write(&sb->s_umount);
1234 if (sb->s_frozen == SB_UNFROZEN) {
1235 up_write(&sb->s_umount);
1236 return -EINVAL;
1237 }
1238
1239 if (sb->s_flags & MS_RDONLY)
1240 goto out;
1241
1242 if (sb->s_op->unfreeze_fs) {
1243 error = sb->s_op->unfreeze_fs(sb);
1244 if (error) {
1245 printk(KERN_ERR
1246 "VFS:Filesystem thaw failed\n");
1247 sb->s_frozen = SB_FREEZE_TRANS;
1248 up_write(&sb->s_umount);
1249 return error;
1250 }
1251 }
1252
1253out:
1254 sb->s_frozen = SB_UNFROZEN;
1255 smp_wmb();
1256 wake_up(&sb->s_wait_unfrozen);
1257 deactivate_locked_super(sb);
1258
1259 return 0;
1260}
1261EXPORT_SYMBOL(thaw_super);