blob: 12f123712161c2cb10d20eae673e4b53de965d31 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/super.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * super.c contains code to handle: - mount structures
7 * - super-block tables
8 * - filesystem drivers list
9 * - mount system call
10 * - umount system call
11 * - ustat system call
12 *
13 * GK 2/5/95 - Changed to support mounting the root fs via NFS
14 *
15 * Added kerneld support: Jacques Gelinas and Bjorn Ekwall
16 * Added change_root: Werner Almesberger & Hans Lermen, Feb '96
17 * Added options to /proc/mounts:
Jan Engelhardt96de0e22007-10-19 23:21:04 +020018 * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996.
Linus Torvalds1da177e2005-04-16 15:20:36 -070019 * Added devfs support: Richard Gooch <rgooch@atnf.csiro.au>, 13-JAN-1998
20 * Heavily rewritten for 'one fs - one tree' dcache architecture. AV, Mar 2000
21 */
22
Paul Gortmaker630d9c42011-11-16 23:57:37 -050023#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/acct.h>
26#include <linux/blkdev.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/mount.h>
28#include <linux/security.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/writeback.h> /* for the emergency remount stuff */
30#include <linux/idr.h>
Ingo Molnar353ab6e2006-03-26 01:37:12 -080031#include <linux/mutex.h>
Jens Axboe5477d0f2010-04-29 20:33:35 +020032#include <linux/backing-dev.h>
Nick Pigginceb5bdc2011-01-07 17:50:05 +110033#include <linux/rculist_bl.h>
Dan Magenheimerc515e1f2011-05-26 10:01:43 -060034#include <linux/cleancache.h>
Al Viro40401532012-02-13 03:58:52 +000035#include <linux/fsnotify.h>
Jan Kara5accdf82012-06-12 16:20:34 +020036#include <linux/lockdep.h>
Al Viro6d59e7f2008-03-22 15:48:17 -040037#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
39
Linus Torvalds1da177e2005-04-16 15:20:36 -070040LIST_HEAD(super_blocks);
41DEFINE_SPINLOCK(sb_lock);
42
Jan Kara5accdf82012-06-12 16:20:34 +020043static char *sb_writers_name[SB_FREEZE_LEVELS] = {
44 "sb_writers",
45 "sb_pagefaults",
46 "sb_internal",
47};
48
Dave Chinnerb0d40c92011-07-08 14:14:42 +100049/*
50 * One thing we have to be careful of with a per-sb shrinker is that we don't
51 * drop the last active reference to the superblock from within the shrinker.
52 * If that happens we could trigger unregistering the shrinker from within the
53 * shrinker path and that leads to deadlock on the shrinker_rwsem. Hence we
54 * take a passive reference to the superblock to avoid this from occurring.
55 */
56static int prune_super(struct shrinker *shrink, struct shrink_control *sc)
57{
58 struct super_block *sb;
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100059 int fs_objects = 0;
60 int total_objects;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100061
62 sb = container_of(shrink, struct super_block, s_shrink);
63
64 /*
65 * Deadlock avoidance. We may hold various FS locks, and we don't want
66 * to recurse into the FS that called us in clear_inode() and friends..
67 */
68 if (sc->nr_to_scan && !(sc->gfp_mask & __GFP_FS))
69 return -1;
70
71 if (!grab_super_passive(sb))
Minchan Kim8e125cd2012-07-31 16:43:16 -070072 return -1;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100073
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100074 if (sb->s_op && sb->s_op->nr_cached_objects)
75 fs_objects = sb->s_op->nr_cached_objects(sb);
76
77 total_objects = sb->s_nr_dentry_unused +
78 sb->s_nr_inodes_unused + fs_objects + 1;
79
Dave Chinnerb0d40c92011-07-08 14:14:42 +100080 if (sc->nr_to_scan) {
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100081 int dentries;
82 int inodes;
Dave Chinnerb0d40c92011-07-08 14:14:42 +100083
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100084 /* proportion the scan between the caches */
85 dentries = (sc->nr_to_scan * sb->s_nr_dentry_unused) /
86 total_objects;
87 inodes = (sc->nr_to_scan * sb->s_nr_inodes_unused) /
88 total_objects;
89 if (fs_objects)
90 fs_objects = (sc->nr_to_scan * fs_objects) /
91 total_objects;
92 /*
93 * prune the dcache first as the icache is pinned by it, then
94 * prune the icache, followed by the filesystem specific caches
95 */
96 prune_dcache_sb(sb, dentries);
97 prune_icache_sb(sb, inodes);
Dave Chinnerb0d40c92011-07-08 14:14:42 +100098
Dave Chinner0e1fdaf2011-07-08 14:14:44 +100099 if (fs_objects && sb->s_op->free_cached_objects) {
100 sb->s_op->free_cached_objects(sb, fs_objects);
101 fs_objects = sb->s_op->nr_cached_objects(sb);
102 }
103 total_objects = sb->s_nr_dentry_unused +
104 sb->s_nr_inodes_unused + fs_objects;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000105 }
106
Dave Chinner0e1fdaf2011-07-08 14:14:44 +1000107 total_objects = (total_objects / 100) * sysctl_vfs_cache_pressure;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000108 drop_super(sb);
Dave Chinner0e1fdaf2011-07-08 14:14:44 +1000109 return total_objects;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000110}
111
Jan Kara5accdf82012-06-12 16:20:34 +0200112static int init_sb_writers(struct super_block *s, struct file_system_type *type)
113{
114 int err;
115 int i;
116
117 for (i = 0; i < SB_FREEZE_LEVELS; i++) {
118 err = percpu_counter_init(&s->s_writers.counter[i], 0);
119 if (err < 0)
120 goto err_out;
121 lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i],
122 &type->s_writers_key[i], 0);
123 }
124 init_waitqueue_head(&s->s_writers.wait);
125 init_waitqueue_head(&s->s_writers.wait_unfrozen);
126 return 0;
127err_out:
128 while (--i >= 0)
129 percpu_counter_destroy(&s->s_writers.counter[i]);
130 return err;
131}
132
133static void destroy_sb_writers(struct super_block *s)
134{
135 int i;
136
137 for (i = 0; i < SB_FREEZE_LEVELS; i++)
138 percpu_counter_destroy(&s->s_writers.counter[i]);
139}
140
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141/**
142 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -0700143 * @type: filesystem type superblock should belong to
David Howells9249e172012-06-25 12:55:37 +0100144 * @flags: the mount flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 *
146 * Allocates and initializes a new &struct super_block. alloc_super()
147 * returns a pointer new superblock or %NULL if allocation had failed.
148 */
David Howells9249e172012-06-25 12:55:37 +0100149static struct super_block *alloc_super(struct file_system_type *type, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -0800151 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700152 static const struct super_operations default_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153
154 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 if (security_sb_alloc(s)) {
Jan Kara5accdf82012-06-12 16:20:34 +0200156 /*
157 * We cannot call security_sb_free() without
158 * security_sb_alloc() succeeding. So bail out manually
159 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 kfree(s);
161 s = NULL;
162 goto out;
163 }
Nick Piggin6416ccb2010-08-18 04:37:38 +1000164#ifdef CONFIG_SMP
165 s->s_files = alloc_percpu(struct list_head);
Jan Kara5accdf82012-06-12 16:20:34 +0200166 if (!s->s_files)
167 goto err_out;
168 else {
Nick Piggin6416ccb2010-08-18 04:37:38 +1000169 int i;
170
171 for_each_possible_cpu(i)
172 INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i));
173 }
174#else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 INIT_LIST_HEAD(&s->s_files);
Nick Piggin6416ccb2010-08-18 04:37:38 +1000176#endif
Jan Kara5accdf82012-06-12 16:20:34 +0200177 if (init_sb_writers(s, type))
178 goto err_out;
David Howells9249e172012-06-25 12:55:37 +0100179 s->s_flags = flags;
Jens Axboe95f28602011-03-17 11:13:12 +0100180 s->s_bdi = &default_backing_dev_info;
Al Viroa5166162011-12-12 22:53:00 -0500181 INIT_HLIST_NODE(&s->s_instances);
Nick Pigginceb5bdc2011-01-07 17:50:05 +1100182 INIT_HLIST_BL_HEAD(&s->s_anon);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 INIT_LIST_HEAD(&s->s_inodes);
Kentaro Makitada3bbdd2008-07-23 21:27:13 -0700184 INIT_LIST_HEAD(&s->s_dentry_lru);
Dave Chinner98b745c2011-07-08 14:14:39 +1000185 INIT_LIST_HEAD(&s->s_inode_lru);
Dave Chinner09cc9fc2011-07-08 14:14:40 +1000186 spin_lock_init(&s->s_inode_lru_lock);
Miklos Szeredi39f7c4d2011-11-21 12:11:30 +0100187 INIT_LIST_HEAD(&s->s_mounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 init_rwsem(&s->s_umount);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -0700189 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -0700190 /*
Peter Zijlstraada723d2009-02-18 14:48:30 -0800191 * sget() can have s_umount recursion.
192 *
193 * When it cannot find a suitable sb, it allocates a new
194 * one (this one), and tries again to find a suitable old
195 * one.
196 *
197 * In case that succeeds, it will acquire the s_umount
198 * lock of the old one. Since these are clearly distrinct
199 * locks, and this object isn't exposed yet, there's no
200 * risk of deadlocks.
201 *
202 * Annotate this by putting this lock in a different
203 * subclass.
204 */
205 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
Al Virob20bd1a2010-03-22 08:53:19 -0400206 s->s_count = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -0800208 mutex_init(&s->s_vfs_rename_mutex);
Roland Dreier51ee0492010-04-27 14:23:57 -0700209 lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key);
Ingo Molnard3be9152006-03-23 03:00:29 -0800210 mutex_init(&s->s_dquot.dqio_mutex);
211 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 init_rwsem(&s->s_dquot.dqptr_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 s->s_maxbytes = MAX_NON_LFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 s->s_op = &default_op;
215 s->s_time_gran = 1000000000;
Dan Magenheimerc515e1f2011-05-26 10:01:43 -0600216 s->cleancache_poolid = -1;
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000217
218 s->s_shrink.seeks = DEFAULT_SEEKS;
219 s->s_shrink.shrink = prune_super;
Dave Chinner8ab47662011-07-08 14:14:45 +1000220 s->s_shrink.batch = 1024;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 }
222out:
223 return s;
Jan Kara5accdf82012-06-12 16:20:34 +0200224err_out:
225 security_sb_free(s);
226#ifdef CONFIG_SMP
227 if (s->s_files)
228 free_percpu(s->s_files);
229#endif
230 destroy_sb_writers(s);
231 kfree(s);
232 s = NULL;
233 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234}
235
236/**
237 * destroy_super - frees a superblock
238 * @s: superblock to free
239 *
240 * Frees a superblock.
241 */
242static inline void destroy_super(struct super_block *s)
243{
Nick Piggin6416ccb2010-08-18 04:37:38 +1000244#ifdef CONFIG_SMP
245 free_percpu(s->s_files);
246#endif
Jan Kara5accdf82012-06-12 16:20:34 +0200247 destroy_sb_writers(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 security_sb_free(s);
Miklos Szeredi39f7c4d2011-11-21 12:11:30 +0100249 WARN_ON(!list_empty(&s->s_mounts));
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700250 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800251 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 kfree(s);
253}
254
255/* Superblock refcounting */
256
257/*
Al Viro35cf7ba2010-03-22 21:13:53 -0400258 * Drop a superblock's refcount. The caller must hold sb_lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 */
Al Virof47ec3f2011-11-21 21:15:42 -0500260static void __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 if (!--sb->s_count) {
Al Viro551de6f2010-03-22 19:36:35 -0400263 list_del_init(&sb->s_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 destroy_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266}
267
268/**
269 * put_super - drop a temporary reference to superblock
270 * @sb: superblock in question
271 *
272 * Drops a temporary reference, frees superblock if there's no
273 * references left.
274 */
Al Virof47ec3f2011-11-21 21:15:42 -0500275static void put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276{
277 spin_lock(&sb_lock);
278 __put_super(sb);
279 spin_unlock(&sb_lock);
280}
281
282
283/**
Al Viro74dbbdd2009-05-06 01:07:50 -0400284 * deactivate_locked_super - drop an active reference to superblock
285 * @s: superblock to deactivate
286 *
Al Viro1712ac82010-03-22 15:22:31 -0400287 * Drops an active reference to superblock, converting it into a temprory
288 * one if there is no other active references left. In that case we
289 * tell fs driver to shut it down and drop the temporary reference we
290 * had just acquired.
291 *
292 * Caller holds exclusive lock on superblock; that lock is released.
Al Viro74dbbdd2009-05-06 01:07:50 -0400293 */
294void deactivate_locked_super(struct super_block *s)
295{
296 struct file_system_type *fs = s->s_type;
Al Virob20bd1a2010-03-22 08:53:19 -0400297 if (atomic_dec_and_test(&s->s_active)) {
Dan Magenheimer31677602011-09-21 11:56:28 -0400298 cleancache_invalidate_fs(s);
Al Viro74dbbdd2009-05-06 01:07:50 -0400299 fs->kill_sb(s);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000300
301 /* caches are now gone, we can safely kill the shrinker now */
302 unregister_shrinker(&s->s_shrink);
Al Viro74dbbdd2009-05-06 01:07:50 -0400303 put_filesystem(fs);
304 put_super(s);
305 } else {
306 up_write(&s->s_umount);
307 }
308}
309
310EXPORT_SYMBOL(deactivate_locked_super);
311
312/**
Al Viro1712ac82010-03-22 15:22:31 -0400313 * deactivate_super - drop an active reference to superblock
314 * @s: superblock to deactivate
315 *
316 * Variant of deactivate_locked_super(), except that superblock is *not*
317 * locked by caller. If we are going to drop the final active reference,
318 * lock will be acquired prior to that.
319 */
320void deactivate_super(struct super_block *s)
321{
322 if (!atomic_add_unless(&s->s_active, -1, 1)) {
323 down_write(&s->s_umount);
324 deactivate_locked_super(s);
325 }
326}
327
328EXPORT_SYMBOL(deactivate_super);
329
330/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 * grab_super - acquire an active reference
332 * @s: reference we are trying to make active
333 *
334 * Tries to acquire an active reference. grab_super() is used when we
335 * had just found a superblock in super_blocks or fs_type->fs_supers
336 * and want to turn it into a full-blown active reference. grab_super()
337 * is called with sb_lock held and drops it. Returns 1 in case of
338 * success, 0 if we had failed (superblock contents was already dead or
339 * dying when grab_super() had been called).
340 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700341static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342{
Al Virob20bd1a2010-03-22 08:53:19 -0400343 if (atomic_inc_not_zero(&s->s_active)) {
344 spin_unlock(&sb_lock);
Al Virob20bd1a2010-03-22 08:53:19 -0400345 return 1;
346 }
347 /* it's going away */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 s->s_count++;
349 spin_unlock(&sb_lock);
Al Viro1712ac82010-03-22 15:22:31 -0400350 /* wait for it to die */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 up_write(&s->s_umount);
353 put_super(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 return 0;
355}
356
David Howellscf9a2ae2006-08-29 19:05:54 +0100357/*
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000358 * grab_super_passive - acquire a passive reference
Wanpeng Li331cbde2012-06-09 11:10:55 +0800359 * @sb: reference we are trying to grab
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000360 *
361 * Tries to acquire a passive reference. This is used in places where we
362 * cannot take an active reference but we need to ensure that the
363 * superblock does not go away while we are working on it. It returns
364 * false if a reference was not gained, and returns true with the s_umount
365 * lock held in read mode if a reference is gained. On successful return,
366 * the caller must drop the s_umount lock and the passive reference when
367 * done.
368 */
369bool grab_super_passive(struct super_block *sb)
370{
371 spin_lock(&sb_lock);
Al Viroa5166162011-12-12 22:53:00 -0500372 if (hlist_unhashed(&sb->s_instances)) {
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000373 spin_unlock(&sb_lock);
374 return false;
375 }
376
377 sb->s_count++;
378 spin_unlock(&sb_lock);
379
380 if (down_read_trylock(&sb->s_umount)) {
Al Virodabe0dc2012-01-03 21:01:29 -0500381 if (sb->s_root && (sb->s_flags & MS_BORN))
Dave Chinner12ad3ab2011-07-08 14:14:41 +1000382 return true;
383 up_read(&sb->s_umount);
384 }
385
386 put_super(sb);
387 return false;
388}
389
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390/**
391 * generic_shutdown_super - common helper for ->kill_sb()
392 * @sb: superblock to kill
393 *
394 * generic_shutdown_super() does all fs-independent work on superblock
395 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
396 * that need destruction out of superblock, call generic_shutdown_super()
397 * and release aforementioned objects. Note: dentries and inodes _are_
398 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700399 *
400 * Upon calling this function, the filesystem may no longer alter or
401 * rearrange the set of dentries belonging to this super_block, nor may it
402 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 */
404void generic_shutdown_super(struct super_block *sb)
405{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800406 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
David Howellsc636ebd2006-10-11 01:22:19 -0700408 if (sb->s_root) {
409 shrink_dcache_for_umount(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200410 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 sb->s_flags &= ~MS_ACTIVE;
Arjan van de Venefaee192009-01-06 07:20:54 -0800412
Al Viro63997e92010-10-25 20:49:35 -0400413 fsnotify_unmount_inodes(&sb->s_inodes);
414
415 evict_inodes(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 if (sop->put_super)
418 sop->put_super(sb);
419
Al Viro63997e92010-10-25 20:49:35 -0400420 if (!list_empty(&sb->s_inodes)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800421 printk("VFS: Busy inodes after unmount of %s. "
422 "Self-destruct in 5 seconds. Have a nice day...\n",
423 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 }
426 spin_lock(&sb_lock);
427 /* should be initialized for __put_super_and_need_restart() */
Al Viroa5166162011-12-12 22:53:00 -0500428 hlist_del_init(&sb->s_instances);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 spin_unlock(&sb_lock);
430 up_write(&sb->s_umount);
431}
432
433EXPORT_SYMBOL(generic_shutdown_super);
434
435/**
436 * sget - find or create a superblock
437 * @type: filesystem type superblock should belong to
438 * @test: comparison callback
439 * @set: setup callback
David Howells9249e172012-06-25 12:55:37 +0100440 * @flags: mount flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 * @data: argument to each of them
442 */
443struct super_block *sget(struct file_system_type *type,
444 int (*test)(struct super_block *,void *),
445 int (*set)(struct super_block *,void *),
David Howells9249e172012-06-25 12:55:37 +0100446 int flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 void *data)
448{
449 struct super_block *s = NULL;
Al Viroa5166162011-12-12 22:53:00 -0500450 struct hlist_node *node;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700451 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 int err;
453
454retry:
455 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700456 if (test) {
Al Viroa5166162011-12-12 22:53:00 -0500457 hlist_for_each_entry(old, node, &type->fs_supers, s_instances) {
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700458 if (!test(old, data))
459 continue;
460 if (!grab_super(old))
461 goto retry;
Li Zefana3cfbb52009-03-12 14:31:29 -0700462 if (s) {
463 up_write(&s->s_umount);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700464 destroy_super(s);
Al Viro7a4dec52010-08-09 12:05:43 -0400465 s = NULL;
Li Zefana3cfbb52009-03-12 14:31:29 -0700466 }
Al Virod3f21472010-03-23 11:11:05 -0400467 down_write(&old->s_umount);
Al Viro7a4dec52010-08-09 12:05:43 -0400468 if (unlikely(!(old->s_flags & MS_BORN))) {
469 deactivate_locked_super(old);
470 goto retry;
471 }
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700472 return old;
473 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 }
475 if (!s) {
476 spin_unlock(&sb_lock);
David Howells9249e172012-06-25 12:55:37 +0100477 s = alloc_super(type, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 if (!s)
479 return ERR_PTR(-ENOMEM);
480 goto retry;
481 }
482
483 err = set(s, data);
484 if (err) {
485 spin_unlock(&sb_lock);
Li Zefana3cfbb52009-03-12 14:31:29 -0700486 up_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 destroy_super(s);
488 return ERR_PTR(err);
489 }
490 s->s_type = type;
491 strlcpy(s->s_id, type->name, sizeof(s->s_id));
492 list_add_tail(&s->s_list, &super_blocks);
Al Viroa5166162011-12-12 22:53:00 -0500493 hlist_add_head(&s->s_instances, &type->fs_supers);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 spin_unlock(&sb_lock);
495 get_filesystem(type);
Dave Chinnerb0d40c92011-07-08 14:14:42 +1000496 register_shrinker(&s->s_shrink);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 return s;
498}
499
500EXPORT_SYMBOL(sget);
501
502void drop_super(struct super_block *sb)
503{
504 up_read(&sb->s_umount);
505 put_super(sb);
506}
507
508EXPORT_SYMBOL(drop_super);
509
Christoph Hellwige5004752009-05-05 16:08:56 +0200510/**
Al Viro01a05b32010-03-23 06:06:58 -0400511 * iterate_supers - call function for all active superblocks
512 * @f: function to call
513 * @arg: argument to pass to it
514 *
515 * Scans the superblock list and calls given function, passing it
516 * locked superblock and given argument.
517 */
518void iterate_supers(void (*f)(struct super_block *, void *), void *arg)
519{
Al Virodca33252010-07-25 02:31:46 +0400520 struct super_block *sb, *p = NULL;
Al Viro01a05b32010-03-23 06:06:58 -0400521
522 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400523 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500524 if (hlist_unhashed(&sb->s_instances))
Al Viro01a05b32010-03-23 06:06:58 -0400525 continue;
526 sb->s_count++;
527 spin_unlock(&sb_lock);
528
529 down_read(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500530 if (sb->s_root && (sb->s_flags & MS_BORN))
Al Viro01a05b32010-03-23 06:06:58 -0400531 f(sb, arg);
532 up_read(&sb->s_umount);
533
534 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400535 if (p)
536 __put_super(p);
537 p = sb;
Al Viro01a05b32010-03-23 06:06:58 -0400538 }
Al Virodca33252010-07-25 02:31:46 +0400539 if (p)
540 __put_super(p);
Al Viro01a05b32010-03-23 06:06:58 -0400541 spin_unlock(&sb_lock);
542}
543
544/**
Al Viro43e15cd2011-06-03 20:16:57 -0400545 * iterate_supers_type - call function for superblocks of given type
546 * @type: fs type
547 * @f: function to call
548 * @arg: argument to pass to it
549 *
550 * Scans the superblock list and calls given function, passing it
551 * locked superblock and given argument.
552 */
553void iterate_supers_type(struct file_system_type *type,
554 void (*f)(struct super_block *, void *), void *arg)
555{
556 struct super_block *sb, *p = NULL;
Al Viroa5166162011-12-12 22:53:00 -0500557 struct hlist_node *node;
Al Viro43e15cd2011-06-03 20:16:57 -0400558
559 spin_lock(&sb_lock);
Al Viroa5166162011-12-12 22:53:00 -0500560 hlist_for_each_entry(sb, node, &type->fs_supers, s_instances) {
Al Viro43e15cd2011-06-03 20:16:57 -0400561 sb->s_count++;
562 spin_unlock(&sb_lock);
563
564 down_read(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500565 if (sb->s_root && (sb->s_flags & MS_BORN))
Al Viro43e15cd2011-06-03 20:16:57 -0400566 f(sb, arg);
567 up_read(&sb->s_umount);
568
569 spin_lock(&sb_lock);
570 if (p)
571 __put_super(p);
572 p = sb;
573 }
574 if (p)
575 __put_super(p);
576 spin_unlock(&sb_lock);
577}
578
579EXPORT_SYMBOL(iterate_supers_type);
580
581/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 * get_super - get the superblock of a device
583 * @bdev: device to get the superblock for
584 *
585 * Scans the superblock list and finds the superblock of the file system
586 * mounted on the device given. %NULL is returned if no match is found.
587 */
588
Al Virodf40c012010-03-22 20:23:25 -0400589struct super_block *get_super(struct block_device *bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700591 struct super_block *sb;
592
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 if (!bdev)
594 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700595
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700597rescan:
598 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500599 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400600 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700601 if (sb->s_bdev == bdev) {
602 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700604 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400605 /* still alive? */
Al Virodabe0dc2012-01-03 21:01:29 -0500606 if (sb->s_root && (sb->s_flags & MS_BORN))
Kirill Korotaev618f0632005-06-23 00:09:54 -0700607 return sb;
608 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400609 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700610 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400611 __put_super(sb);
612 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613 }
614 }
615 spin_unlock(&sb_lock);
616 return NULL;
617}
618
619EXPORT_SYMBOL(get_super);
Christoph Hellwig45042302009-08-03 23:28:35 +0200620
621/**
Jan Kara6b6dc832012-02-10 11:03:00 +0100622 * get_super_thawed - get thawed superblock of a device
623 * @bdev: device to get the superblock for
624 *
625 * Scans the superblock list and finds the superblock of the file system
626 * mounted on the device. The superblock is returned once it is thawed
627 * (or immediately if it was not frozen). %NULL is returned if no match
628 * is found.
629 */
630struct super_block *get_super_thawed(struct block_device *bdev)
631{
632 while (1) {
633 struct super_block *s = get_super(bdev);
Jan Kara5accdf82012-06-12 16:20:34 +0200634 if (!s || s->s_writers.frozen == SB_UNFROZEN)
Jan Kara6b6dc832012-02-10 11:03:00 +0100635 return s;
636 up_read(&s->s_umount);
Jan Kara5accdf82012-06-12 16:20:34 +0200637 wait_event(s->s_writers.wait_unfrozen,
638 s->s_writers.frozen == SB_UNFROZEN);
Jan Kara6b6dc832012-02-10 11:03:00 +0100639 put_super(s);
640 }
641}
642EXPORT_SYMBOL(get_super_thawed);
643
644/**
Christoph Hellwig45042302009-08-03 23:28:35 +0200645 * get_active_super - get an active reference to the superblock of a device
646 * @bdev: device to get the superblock for
647 *
648 * Scans the superblock list and finds the superblock of the file system
649 * mounted on the device given. Returns the superblock with an active
Al Virod3f21472010-03-23 11:11:05 -0400650 * reference or %NULL if none was found.
Christoph Hellwig45042302009-08-03 23:28:35 +0200651 */
652struct super_block *get_active_super(struct block_device *bdev)
653{
654 struct super_block *sb;
655
656 if (!bdev)
657 return NULL;
658
Al Viro14945832010-03-22 20:15:33 -0400659restart:
Christoph Hellwig45042302009-08-03 23:28:35 +0200660 spin_lock(&sb_lock);
661 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500662 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400663 continue;
Al Viro14945832010-03-22 20:15:33 -0400664 if (sb->s_bdev == bdev) {
665 if (grab_super(sb)) /* drops sb_lock */
666 return sb;
667 else
668 goto restart;
669 }
Christoph Hellwig45042302009-08-03 23:28:35 +0200670 }
671 spin_unlock(&sb_lock);
672 return NULL;
673}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674
Al Virodf40c012010-03-22 20:23:25 -0400675struct super_block *user_get_super(dev_t dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700677 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700680rescan:
681 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500682 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400683 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700684 if (sb->s_dev == dev) {
685 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700687 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400688 /* still alive? */
Al Virodabe0dc2012-01-03 21:01:29 -0500689 if (sb->s_root && (sb->s_flags & MS_BORN))
Kirill Korotaev618f0632005-06-23 00:09:54 -0700690 return sb;
691 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400692 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700693 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400694 __put_super(sb);
695 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 }
697 }
698 spin_unlock(&sb_lock);
699 return NULL;
700}
701
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 * do_remount_sb - asks filesystem to change mount options.
704 * @sb: superblock in question
705 * @flags: numeric part of options
706 * @data: the rest of options
707 * @force: whether or not to force the change
708 *
709 * Alters the mount options of a mounted file system.
710 */
711int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
712{
713 int retval;
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400714 int remount_ro;
Christoph Hellwig45042302009-08-03 23:28:35 +0200715
Jan Kara5accdf82012-06-12 16:20:34 +0200716 if (sb->s_writers.frozen != SB_UNFROZEN)
Christoph Hellwig45042302009-08-03 23:28:35 +0200717 return -EBUSY;
718
David Howells93614012006-09-30 20:45:40 +0200719#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
721 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200722#endif
Christoph Hellwig45042302009-08-03 23:28:35 +0200723
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 if (flags & MS_RDONLY)
725 acct_auto_close(sb);
726 shrink_dcache_sb(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200727 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728
Nick Piggind208bbd2009-12-21 16:28:53 -0800729 remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY);
Nick Piggind208bbd2009-12-21 16:28:53 -0800730
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 /* If we are remounting RDONLY and current sb is read/write,
732 make sure there are no rw files opened */
Nick Piggind208bbd2009-12-21 16:28:53 -0800733 if (remount_ro) {
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100734 if (force) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 mark_files_ro(sb);
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100736 } else {
737 retval = sb_prepare_remount_readonly(sb);
738 if (retval)
739 return retval;
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100740 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 }
742
743 if (sb->s_op->remount_fs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 retval = sb->s_op->remount_fs(sb, &flags, data);
Miklos Szeredi2833eb22011-10-28 14:13:20 +0200745 if (retval) {
746 if (!force)
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100747 goto cancel_readonly;
Miklos Szeredi2833eb22011-10-28 14:13:20 +0200748 /* If forced remount, go ahead despite any errors */
749 WARN(1, "forced remount of a %s fs returned %i\n",
750 sb->s_type->name, retval);
751 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 }
753 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100754 /* Needs to be ordered wrt mnt_is_readonly() */
755 smp_wmb();
756 sb->s_readonly_remount = 0;
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400757
Nick Piggind208bbd2009-12-21 16:28:53 -0800758 /*
759 * Some filesystems modify their metadata via some other path than the
760 * bdev buffer cache (eg. use a private mapping, or directories in
761 * pagecache, etc). Also file data modifications go via their own
762 * mappings. So If we try to mount readonly then copy the filesystem
763 * from bdev, we could get stale data, so invalidate it to give a best
764 * effort at coherency.
765 */
766 if (remount_ro && sb->s_bdev)
767 invalidate_bdev(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 return 0;
Miklos Szeredi4ed5e822011-11-21 12:11:31 +0100769
770cancel_readonly:
771 sb->s_readonly_remount = 0;
772 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773}
774
Jens Axboea2a95372009-03-17 09:38:40 +0100775static void do_emergency_remount(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776{
Al Virodca33252010-07-25 02:31:46 +0400777 struct super_block *sb, *p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
779 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400780 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viroa5166162011-12-12 22:53:00 -0500781 if (hlist_unhashed(&sb->s_instances))
Al Viro551de6f2010-03-22 19:36:35 -0400782 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 sb->s_count++;
784 spin_unlock(&sb_lock);
Al Viro443b94b2009-05-05 23:48:50 -0400785 down_write(&sb->s_umount);
Al Virodabe0dc2012-01-03 21:01:29 -0500786 if (sb->s_root && sb->s_bdev && (sb->s_flags & MS_BORN) &&
787 !(sb->s_flags & MS_RDONLY)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 * What lock protects sb->s_flags??
790 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 do_remount_sb(sb, MS_RDONLY, NULL, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 }
Al Viro443b94b2009-05-05 23:48:50 -0400793 up_write(&sb->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400795 if (p)
796 __put_super(p);
797 p = sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 }
Al Virodca33252010-07-25 02:31:46 +0400799 if (p)
800 __put_super(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 spin_unlock(&sb_lock);
Jens Axboea2a95372009-03-17 09:38:40 +0100802 kfree(work);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 printk("Emergency Remount complete\n");
804}
805
806void emergency_remount(void)
807{
Jens Axboea2a95372009-03-17 09:38:40 +0100808 struct work_struct *work;
809
810 work = kmalloc(sizeof(*work), GFP_ATOMIC);
811 if (work) {
812 INIT_WORK(work, do_emergency_remount);
813 schedule_work(work);
814 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815}
816
817/*
818 * Unnamed block devices are dummy devices used by virtual
819 * filesystems which don't use real block-devices. -- jrs
820 */
821
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400822static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
Al Viroc63e09e2009-06-24 02:05:18 -0400824static int unnamed_dev_start = 0; /* don't bother trying below it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825
Al Viro0ee5dc62011-07-07 15:44:25 -0400826int get_anon_bdev(dev_t *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827{
828 int dev;
829 int error;
830
831 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400832 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 return -ENOMEM;
834 spin_lock(&unnamed_dev_lock);
Al Viroc63e09e2009-06-24 02:05:18 -0400835 error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
Al Virof21f6222009-06-24 03:12:00 -0400836 if (!error)
837 unnamed_dev_start = dev + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 spin_unlock(&unnamed_dev_lock);
839 if (error == -EAGAIN)
840 /* We raced and lost with another CPU. */
841 goto retry;
842 else if (error)
843 return -EAGAIN;
844
Fengguang Wu125c4c72012-10-04 17:13:15 -0700845 if ((dev & MAX_IDR_MASK) == (1 << MINORBITS)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400847 ida_remove(&unnamed_dev_ida, dev);
Al Virof21f6222009-06-24 03:12:00 -0400848 if (unnamed_dev_start > dev)
849 unnamed_dev_start = dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 spin_unlock(&unnamed_dev_lock);
851 return -EMFILE;
852 }
Al Viro0ee5dc62011-07-07 15:44:25 -0400853 *p = MKDEV(0, dev & MINORMASK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 return 0;
855}
Al Viro0ee5dc62011-07-07 15:44:25 -0400856EXPORT_SYMBOL(get_anon_bdev);
857
858void free_anon_bdev(dev_t dev)
859{
860 int slot = MINOR(dev);
861 spin_lock(&unnamed_dev_lock);
862 ida_remove(&unnamed_dev_ida, slot);
863 if (slot < unnamed_dev_start)
864 unnamed_dev_start = slot;
865 spin_unlock(&unnamed_dev_lock);
866}
867EXPORT_SYMBOL(free_anon_bdev);
868
869int set_anon_super(struct super_block *s, void *data)
870{
871 int error = get_anon_bdev(&s->s_dev);
872 if (!error)
873 s->s_bdi = &noop_backing_dev_info;
874 return error;
875}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876
877EXPORT_SYMBOL(set_anon_super);
878
879void kill_anon_super(struct super_block *sb)
880{
Al Viro0ee5dc62011-07-07 15:44:25 -0400881 dev_t dev = sb->s_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 generic_shutdown_super(sb);
Al Viro0ee5dc62011-07-07 15:44:25 -0400883 free_anon_bdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884}
885
886EXPORT_SYMBOL(kill_anon_super);
887
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888void kill_litter_super(struct super_block *sb)
889{
890 if (sb->s_root)
891 d_genocide(sb->s_root);
892 kill_anon_super(sb);
893}
894
895EXPORT_SYMBOL(kill_litter_super);
896
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700897static int ns_test_super(struct super_block *sb, void *data)
898{
899 return sb->s_fs_info == data;
900}
901
902static int ns_set_super(struct super_block *sb, void *data)
903{
904 sb->s_fs_info = data;
905 return set_anon_super(sb, NULL);
906}
907
Al Viroceefda62010-07-26 13:16:50 +0400908struct dentry *mount_ns(struct file_system_type *fs_type, int flags,
909 void *data, int (*fill_super)(struct super_block *, void *, int))
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700910{
911 struct super_block *sb;
912
David Howells9249e172012-06-25 12:55:37 +0100913 sb = sget(fs_type, ns_test_super, ns_set_super, flags, data);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700914 if (IS_ERR(sb))
Al Viroceefda62010-07-26 13:16:50 +0400915 return ERR_CAST(sb);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700916
917 if (!sb->s_root) {
918 int err;
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700919 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
920 if (err) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400921 deactivate_locked_super(sb);
Al Viroceefda62010-07-26 13:16:50 +0400922 return ERR_PTR(err);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700923 }
924
925 sb->s_flags |= MS_ACTIVE;
926 }
927
Al Viroceefda62010-07-26 13:16:50 +0400928 return dget(sb->s_root);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700929}
930
Al Viroceefda62010-07-26 13:16:50 +0400931EXPORT_SYMBOL(mount_ns);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700932
David Howells93614012006-09-30 20:45:40 +0200933#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934static int set_bdev_super(struct super_block *s, void *data)
935{
936 s->s_bdev = data;
937 s->s_dev = s->s_bdev->bd_dev;
Jens Axboe32a88aa2009-09-16 15:02:33 +0200938
939 /*
940 * We set the bdi here to the queue backing, file systems can
941 * overwrite this in ->fill_super()
942 */
943 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 return 0;
945}
946
947static int test_bdev_super(struct super_block *s, void *data)
948{
949 return (void *)s->s_bdev == data;
950}
951
Al Viro152a0832010-07-25 00:46:55 +0400952struct dentry *mount_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 int flags, const char *dev_name, void *data,
Al Viro152a0832010-07-25 00:46:55 +0400954 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955{
956 struct block_device *bdev;
957 struct super_block *s;
Tejun Heod4d77622010-11-13 11:55:18 +0100958 fmode_t mode = FMODE_READ | FMODE_EXCL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 int error = 0;
960
Al Viro30c40d22008-02-22 19:50:45 -0500961 if (!(flags & MS_RDONLY))
962 mode |= FMODE_WRITE;
963
Tejun Heod4d77622010-11-13 11:55:18 +0100964 bdev = blkdev_get_by_path(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 if (IS_ERR(bdev))
Al Viro152a0832010-07-25 00:46:55 +0400966 return ERR_CAST(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967
968 /*
969 * once the super is inserted into the list by sget, s_umount
970 * will protect the lockfs code from trying to start a snapshot
971 * while we are mounting
972 */
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200973 mutex_lock(&bdev->bd_fsfreeze_mutex);
974 if (bdev->bd_fsfreeze_count > 0) {
975 mutex_unlock(&bdev->bd_fsfreeze_mutex);
976 error = -EBUSY;
977 goto error_bdev;
978 }
David Howells9249e172012-06-25 12:55:37 +0100979 s = sget(fs_type, test_bdev_super, set_bdev_super, flags | MS_NOSEC,
980 bdev);
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200981 mutex_unlock(&bdev->bd_fsfreeze_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700983 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984
985 if (s->s_root) {
986 if ((flags ^ s->s_flags) & MS_RDONLY) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400987 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700988 error = -EBUSY;
989 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 }
David Howells454e2392006-06-23 02:02:57 -0700991
Tejun Heo4f331f02010-07-20 15:18:07 -0700992 /*
993 * s_umount nests inside bd_mutex during
Tejun Heoe525fd82010-11-13 11:55:17 +0100994 * __invalidate_device(). blkdev_put() acquires
995 * bd_mutex and can't be called under s_umount. Drop
996 * s_umount temporarily. This is safe as we're
997 * holding an active reference.
Tejun Heo4f331f02010-07-20 15:18:07 -0700998 */
999 up_write(&s->s_umount);
Tejun Heod4d77622010-11-13 11:55:18 +01001000 blkdev_put(bdev, mode);
Tejun Heo4f331f02010-07-20 15:18:07 -07001001 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 } else {
1003 char b[BDEVNAME_SIZE];
1004
Al Viro30c40d22008-02-22 19:50:45 -05001005 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -08001007 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001008 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001010 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -07001011 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -08001012 }
David Howells454e2392006-06-23 02:02:57 -07001013
1014 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -05001015 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 }
1017
Al Viro152a0832010-07-25 00:46:55 +04001018 return dget(s->s_root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
David Howells454e2392006-06-23 02:02:57 -07001020error_s:
1021 error = PTR_ERR(s);
1022error_bdev:
Tejun Heod4d77622010-11-13 11:55:18 +01001023 blkdev_put(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -07001024error:
Al Viro152a0832010-07-25 00:46:55 +04001025 return ERR_PTR(error);
1026}
1027EXPORT_SYMBOL(mount_bdev);
1028
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029void kill_block_super(struct super_block *sb)
1030{
1031 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -05001032 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033
H Hartley Sweetenddbaaf32009-04-29 20:14:57 -04001034 bdev->bd_super = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 generic_shutdown_super(sb);
1036 sync_blockdev(bdev);
Tejun Heod4d77622010-11-13 11:55:18 +01001037 WARN_ON_ONCE(!(mode & FMODE_EXCL));
Tejun Heoe525fd82010-11-13 11:55:17 +01001038 blkdev_put(bdev, mode | FMODE_EXCL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039}
1040
1041EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +02001042#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043
Al Viro3c26ff62010-07-25 11:46:36 +04001044struct dentry *mount_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 int flags, void *data,
Al Viro3c26ff62010-07-25 11:46:36 +04001046 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047{
1048 int error;
David Howells9249e172012-06-25 12:55:37 +01001049 struct super_block *s = sget(fs_type, NULL, set_anon_super, flags, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050
1051 if (IS_ERR(s))
Al Viro3c26ff62010-07-25 11:46:36 +04001052 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001054 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001056 deactivate_locked_super(s);
Al Viro3c26ff62010-07-25 11:46:36 +04001057 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 }
1059 s->s_flags |= MS_ACTIVE;
Al Viro3c26ff62010-07-25 11:46:36 +04001060 return dget(s->s_root);
1061}
1062EXPORT_SYMBOL(mount_nodev);
1063
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064static int compare_single(struct super_block *s, void *p)
1065{
1066 return 1;
1067}
1068
Al Virofc14f2f2010-07-25 01:48:30 +04001069struct dentry *mount_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 int flags, void *data,
Al Virofc14f2f2010-07-25 01:48:30 +04001071 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072{
1073 struct super_block *s;
1074 int error;
1075
David Howells9249e172012-06-25 12:55:37 +01001076 s = sget(fs_type, compare_single, set_anon_super, flags, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 if (IS_ERR(s))
Al Virofc14f2f2010-07-25 01:48:30 +04001078 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 if (!s->s_root) {
Theodore Ts'o9b04c992006-03-24 03:15:10 -08001080 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -04001082 deactivate_locked_super(s);
Al Virofc14f2f2010-07-25 01:48:30 +04001083 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 }
1085 s->s_flags |= MS_ACTIVE;
Kay Sievers9329d1b2009-12-18 21:18:15 +01001086 } else {
1087 do_remount_sb(s, flags, data, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 }
Al Virofc14f2f2010-07-25 01:48:30 +04001089 return dget(s->s_root);
1090}
1091EXPORT_SYMBOL(mount_single);
1092
Al Viro9d412a42011-03-17 22:08:28 -04001093struct dentry *
1094mount_fs(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095{
Al Viroc96e41e2010-07-25 00:17:56 +04001096 struct dentry *root;
Al Viro9d412a42011-03-17 22:08:28 -04001097 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 char *secdata = NULL;
Al Viro9d412a42011-03-17 22:08:28 -04001099 int error = -ENOMEM;
Al Viro80893522010-02-05 09:30:46 -05001100
Eric Parise0007522008-03-05 10:31:54 -05001101 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -07001103 if (!secdata)
Al Viro9d412a42011-03-17 22:08:28 -04001104 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105
Eric Parise0007522008-03-05 10:31:54 -05001106 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -07001107 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 }
1110
Al Viro1a102ff2011-03-16 09:07:58 -04001111 root = type->mount(type, flags, name, data);
1112 if (IS_ERR(root)) {
1113 error = PTR_ERR(root);
1114 goto out_free_secdata;
Al Viroc96e41e2010-07-25 00:17:56 +04001115 }
Al Viro9d412a42011-03-17 22:08:28 -04001116 sb = root->d_sb;
1117 BUG_ON(!sb);
1118 WARN_ON(!sb->s_bdi);
Linus Torvalds6c510382011-03-24 10:16:26 -07001119 WARN_ON(sb->s_bdi == &default_backing_dev_info);
Al Viro9d412a42011-03-17 22:08:28 -04001120 sb->s_flags |= MS_BORN;
David Howells454e2392006-06-23 02:02:57 -07001121
Al Viro9d412a42011-03-17 22:08:28 -04001122 error = security_sb_kern_mount(sb, flags, secdata);
Jörn Engel5129a462010-04-25 08:54:42 +02001123 if (error)
1124 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -07001125
Jeff Layton42cb56a2009-09-18 13:05:53 -07001126 /*
1127 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
1128 * but s_maxbytes was an unsigned long long for many releases. Throw
1129 * this warning for a little while to try and catch filesystems that
Jeff Layton4358b562011-03-29 09:33:31 -04001130 * violate this rule.
Jeff Layton42cb56a2009-09-18 13:05:53 -07001131 */
Al Viro9d412a42011-03-17 22:08:28 -04001132 WARN((sb->s_maxbytes < 0), "%s set sb->s_maxbytes to "
1133 "negative value (%lld)\n", type->name, sb->s_maxbytes);
Jeff Layton42cb56a2009-09-18 13:05:53 -07001134
Al Viro9d412a42011-03-17 22:08:28 -04001135 up_write(&sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -07001136 free_secdata(secdata);
Al Viro9d412a42011-03-17 22:08:28 -04001137 return root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138out_sb:
Al Viro9d412a42011-03-17 22:08:28 -04001139 dput(root);
1140 deactivate_locked_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141out_free_secdata:
1142 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143out:
David Howells454e2392006-06-23 02:02:57 -07001144 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145}
1146
Jan Kara5accdf82012-06-12 16:20:34 +02001147/*
1148 * This is an internal function, please use sb_end_{write,pagefault,intwrite}
1149 * instead.
1150 */
1151void __sb_end_write(struct super_block *sb, int level)
1152{
1153 percpu_counter_dec(&sb->s_writers.counter[level-1]);
1154 /*
1155 * Make sure s_writers are updated before we wake up waiters in
1156 * freeze_super().
1157 */
1158 smp_mb();
1159 if (waitqueue_active(&sb->s_writers.wait))
1160 wake_up(&sb->s_writers.wait);
1161 rwsem_release(&sb->s_writers.lock_map[level-1], 1, _RET_IP_);
1162}
1163EXPORT_SYMBOL(__sb_end_write);
1164
1165#ifdef CONFIG_LOCKDEP
1166/*
1167 * We want lockdep to tell us about possible deadlocks with freezing but
1168 * it's it bit tricky to properly instrument it. Getting a freeze protection
1169 * works as getting a read lock but there are subtle problems. XFS for example
1170 * gets freeze protection on internal level twice in some cases, which is OK
1171 * only because we already hold a freeze protection also on higher level. Due
1172 * to these cases we have to tell lockdep we are doing trylock when we
1173 * already hold a freeze protection for a higher freeze level.
1174 */
1175static void acquire_freeze_lock(struct super_block *sb, int level, bool trylock,
1176 unsigned long ip)
1177{
1178 int i;
1179
1180 if (!trylock) {
1181 for (i = 0; i < level - 1; i++)
1182 if (lock_is_held(&sb->s_writers.lock_map[i])) {
1183 trylock = true;
1184 break;
1185 }
1186 }
1187 rwsem_acquire_read(&sb->s_writers.lock_map[level-1], 0, trylock, ip);
1188}
1189#endif
1190
1191/*
1192 * This is an internal function, please use sb_start_{write,pagefault,intwrite}
1193 * instead.
1194 */
1195int __sb_start_write(struct super_block *sb, int level, bool wait)
1196{
1197retry:
1198 if (unlikely(sb->s_writers.frozen >= level)) {
1199 if (!wait)
1200 return 0;
1201 wait_event(sb->s_writers.wait_unfrozen,
1202 sb->s_writers.frozen < level);
1203 }
1204
1205#ifdef CONFIG_LOCKDEP
1206 acquire_freeze_lock(sb, level, !wait, _RET_IP_);
1207#endif
1208 percpu_counter_inc(&sb->s_writers.counter[level-1]);
1209 /*
1210 * Make sure counter is updated before we check for frozen.
1211 * freeze_super() first sets frozen and then checks the counter.
1212 */
1213 smp_mb();
1214 if (unlikely(sb->s_writers.frozen >= level)) {
1215 __sb_end_write(sb, level);
1216 goto retry;
1217 }
1218 return 1;
1219}
1220EXPORT_SYMBOL(__sb_start_write);
1221
1222/**
1223 * sb_wait_write - wait until all writers to given file system finish
1224 * @sb: the super for which we wait
1225 * @level: type of writers we wait for (normal vs page fault)
1226 *
1227 * This function waits until there are no writers of given type to given file
1228 * system. Caller of this function should make sure there can be no new writers
1229 * of type @level before calling this function. Otherwise this function can
1230 * livelock.
1231 */
1232static void sb_wait_write(struct super_block *sb, int level)
1233{
1234 s64 writers;
1235
1236 /*
1237 * We just cycle-through lockdep here so that it does not complain
1238 * about returning with lock to userspace
1239 */
1240 rwsem_acquire(&sb->s_writers.lock_map[level-1], 0, 0, _THIS_IP_);
1241 rwsem_release(&sb->s_writers.lock_map[level-1], 1, _THIS_IP_);
1242
1243 do {
1244 DEFINE_WAIT(wait);
1245
1246 /*
1247 * We use a barrier in prepare_to_wait() to separate setting
1248 * of frozen and checking of the counter
1249 */
1250 prepare_to_wait(&sb->s_writers.wait, &wait,
1251 TASK_UNINTERRUPTIBLE);
1252
1253 writers = percpu_counter_sum(&sb->s_writers.counter[level-1]);
1254 if (writers)
1255 schedule();
1256
1257 finish_wait(&sb->s_writers.wait, &wait);
1258 } while (writers);
1259}
1260
Josef Bacik18e9e512010-03-23 10:34:56 -04001261/**
Randy Dunlap7000d3c2010-05-24 22:22:34 -07001262 * freeze_super - lock the filesystem and force it into a consistent state
1263 * @sb: the super to lock
Josef Bacik18e9e512010-03-23 10:34:56 -04001264 *
1265 * Syncs the super to make sure the filesystem is consistent and calls the fs's
1266 * freeze_fs. Subsequent calls to this without first thawing the fs will return
1267 * -EBUSY.
Jan Kara5accdf82012-06-12 16:20:34 +02001268 *
1269 * During this function, sb->s_writers.frozen goes through these values:
1270 *
1271 * SB_UNFROZEN: File system is normal, all writes progress as usual.
1272 *
1273 * SB_FREEZE_WRITE: The file system is in the process of being frozen. New
1274 * writes should be blocked, though page faults are still allowed. We wait for
1275 * all writes to complete and then proceed to the next stage.
1276 *
1277 * SB_FREEZE_PAGEFAULT: Freezing continues. Now also page faults are blocked
1278 * but internal fs threads can still modify the filesystem (although they
1279 * should not dirty new pages or inodes), writeback can run etc. After waiting
1280 * for all running page faults we sync the filesystem which will clean all
1281 * dirty pages and inodes (no new dirty pages or inodes can be created when
1282 * sync is running).
1283 *
1284 * SB_FREEZE_FS: The file system is frozen. Now all internal sources of fs
1285 * modification are blocked (e.g. XFS preallocation truncation on inode
1286 * reclaim). This is usually implemented by blocking new transactions for
1287 * filesystems that have them and need this additional guard. After all
1288 * internal writers are finished we call ->freeze_fs() to finish filesystem
1289 * freezing. Then we transition to SB_FREEZE_COMPLETE state. This state is
1290 * mostly auxiliary for filesystems to verify they do not modify frozen fs.
1291 *
1292 * sb->s_writers.frozen is protected by sb->s_umount.
Josef Bacik18e9e512010-03-23 10:34:56 -04001293 */
1294int freeze_super(struct super_block *sb)
1295{
1296 int ret;
1297
1298 atomic_inc(&sb->s_active);
1299 down_write(&sb->s_umount);
Jan Kara5accdf82012-06-12 16:20:34 +02001300 if (sb->s_writers.frozen != SB_UNFROZEN) {
Josef Bacik18e9e512010-03-23 10:34:56 -04001301 deactivate_locked_super(sb);
1302 return -EBUSY;
1303 }
1304
Al Virodabe0dc2012-01-03 21:01:29 -05001305 if (!(sb->s_flags & MS_BORN)) {
1306 up_write(&sb->s_umount);
1307 return 0; /* sic - it's "nothing to do" */
1308 }
1309
Josef Bacik18e9e512010-03-23 10:34:56 -04001310 if (sb->s_flags & MS_RDONLY) {
Jan Kara5accdf82012-06-12 16:20:34 +02001311 /* Nothing to do really... */
1312 sb->s_writers.frozen = SB_FREEZE_COMPLETE;
Josef Bacik18e9e512010-03-23 10:34:56 -04001313 up_write(&sb->s_umount);
1314 return 0;
1315 }
1316
Jan Kara5accdf82012-06-12 16:20:34 +02001317 /* From now on, no new normal writers can start */
1318 sb->s_writers.frozen = SB_FREEZE_WRITE;
Josef Bacik18e9e512010-03-23 10:34:56 -04001319 smp_wmb();
1320
Jan Kara5accdf82012-06-12 16:20:34 +02001321 /* Release s_umount to preserve sb_start_write -> s_umount ordering */
1322 up_write(&sb->s_umount);
1323
1324 sb_wait_write(sb, SB_FREEZE_WRITE);
1325
1326 /* Now we go and block page faults... */
1327 down_write(&sb->s_umount);
1328 sb->s_writers.frozen = SB_FREEZE_PAGEFAULT;
1329 smp_wmb();
1330
1331 sb_wait_write(sb, SB_FREEZE_PAGEFAULT);
1332
1333 /* All writers are done so after syncing there won't be dirty data */
Josef Bacik18e9e512010-03-23 10:34:56 -04001334 sync_filesystem(sb);
1335
Jan Kara5accdf82012-06-12 16:20:34 +02001336 /* Now wait for internal filesystem counter */
1337 sb->s_writers.frozen = SB_FREEZE_FS;
Josef Bacik18e9e512010-03-23 10:34:56 -04001338 smp_wmb();
Jan Kara5accdf82012-06-12 16:20:34 +02001339 sb_wait_write(sb, SB_FREEZE_FS);
Josef Bacik18e9e512010-03-23 10:34:56 -04001340
Josef Bacik18e9e512010-03-23 10:34:56 -04001341 if (sb->s_op->freeze_fs) {
1342 ret = sb->s_op->freeze_fs(sb);
1343 if (ret) {
1344 printk(KERN_ERR
1345 "VFS:Filesystem freeze failed\n");
Jan Kara5accdf82012-06-12 16:20:34 +02001346 sb->s_writers.frozen = SB_UNFROZEN;
Kazuya Mioe1616302011-12-01 16:51:07 +09001347 smp_wmb();
Jan Kara5accdf82012-06-12 16:20:34 +02001348 wake_up(&sb->s_writers.wait_unfrozen);
Josef Bacik18e9e512010-03-23 10:34:56 -04001349 deactivate_locked_super(sb);
1350 return ret;
1351 }
1352 }
Jan Kara5accdf82012-06-12 16:20:34 +02001353 /*
1354 * This is just for debugging purposes so that fs can warn if it
1355 * sees write activity when frozen is set to SB_FREEZE_COMPLETE.
1356 */
1357 sb->s_writers.frozen = SB_FREEZE_COMPLETE;
Josef Bacik18e9e512010-03-23 10:34:56 -04001358 up_write(&sb->s_umount);
1359 return 0;
1360}
1361EXPORT_SYMBOL(freeze_super);
1362
1363/**
1364 * thaw_super -- unlock filesystem
1365 * @sb: the super to thaw
1366 *
1367 * Unlocks the filesystem and marks it writeable again after freeze_super().
1368 */
1369int thaw_super(struct super_block *sb)
1370{
1371 int error;
1372
1373 down_write(&sb->s_umount);
Jan Kara5accdf82012-06-12 16:20:34 +02001374 if (sb->s_writers.frozen == SB_UNFROZEN) {
Josef Bacik18e9e512010-03-23 10:34:56 -04001375 up_write(&sb->s_umount);
1376 return -EINVAL;
1377 }
1378
1379 if (sb->s_flags & MS_RDONLY)
1380 goto out;
1381
1382 if (sb->s_op->unfreeze_fs) {
1383 error = sb->s_op->unfreeze_fs(sb);
1384 if (error) {
1385 printk(KERN_ERR
1386 "VFS:Filesystem thaw failed\n");
Josef Bacik18e9e512010-03-23 10:34:56 -04001387 up_write(&sb->s_umount);
1388 return error;
1389 }
1390 }
1391
1392out:
Jan Kara5accdf82012-06-12 16:20:34 +02001393 sb->s_writers.frozen = SB_UNFROZEN;
Josef Bacik18e9e512010-03-23 10:34:56 -04001394 smp_wmb();
Jan Kara5accdf82012-06-12 16:20:34 +02001395 wake_up(&sb->s_writers.wait_unfrozen);
Josef Bacik18e9e512010-03-23 10:34:56 -04001396 deactivate_locked_super(sb);
1397
1398 return 0;
1399}
1400EXPORT_SYMBOL(thaw_super);