blob: ca696155cd9aea65d527af3e8d25a8bdc9cc40f0 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/super.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * super.c contains code to handle: - mount structures
7 * - super-block tables
8 * - filesystem drivers list
9 * - mount system call
10 * - umount system call
11 * - ustat system call
12 *
13 * GK 2/5/95 - Changed to support mounting the root fs via NFS
14 *
15 * Added kerneld support: Jacques Gelinas and Bjorn Ekwall
16 * Added change_root: Werner Almesberger & Hans Lermen, Feb '96
17 * Added options to /proc/mounts:
Jan Engelhardt96de0e22007-10-19 23:21:04 +020018 * Torbjörn Lindh (torbjorn.lindh@gopta.se), April 14, 1996.
Linus Torvalds1da177e2005-04-16 15:20:36 -070019 * Added devfs support: Richard Gooch <rgooch@atnf.csiro.au>, 13-JAN-1998
20 * Heavily rewritten for 'one fs - one tree' dcache architecture. AV, Mar 2000
21 */
22
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <linux/module.h>
24#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/acct.h>
26#include <linux/blkdev.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/mount.h>
28#include <linux/security.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/writeback.h> /* for the emergency remount stuff */
30#include <linux/idr.h>
Ingo Molnar353ab6e2006-03-26 01:37:12 -080031#include <linux/mutex.h>
Jens Axboe5477d0f2010-04-29 20:33:35 +020032#include <linux/backing-dev.h>
Al Viro6d59e7f2008-03-22 15:48:17 -040033#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
35
Linus Torvalds1da177e2005-04-16 15:20:36 -070036LIST_HEAD(super_blocks);
37DEFINE_SPINLOCK(sb_lock);
38
39/**
40 * alloc_super - create new superblock
Henrik Kretzschmarfe2bbc42006-09-06 00:03:41 -070041 * @type: filesystem type superblock should belong to
Linus Torvalds1da177e2005-04-16 15:20:36 -070042 *
43 * Allocates and initializes a new &struct super_block. alloc_super()
44 * returns a pointer new superblock or %NULL if allocation had failed.
45 */
Ingo Molnarcf516242006-07-03 00:25:27 -070046static struct super_block *alloc_super(struct file_system_type *type)
Linus Torvalds1da177e2005-04-16 15:20:36 -070047{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -080048 struct super_block *s = kzalloc(sizeof(struct super_block), GFP_USER);
Alexey Dobriyanb87221d2009-09-21 17:01:09 -070049 static const struct super_operations default_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -070050
51 if (s) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070052 if (security_sb_alloc(s)) {
53 kfree(s);
54 s = NULL;
55 goto out;
56 }
Nick Piggin6416ccb2010-08-18 04:37:38 +100057#ifdef CONFIG_SMP
58 s->s_files = alloc_percpu(struct list_head);
59 if (!s->s_files) {
60 security_sb_free(s);
61 kfree(s);
62 s = NULL;
63 goto out;
64 } else {
65 int i;
66
67 for_each_possible_cpu(i)
68 INIT_LIST_HEAD(per_cpu_ptr(s->s_files, i));
69 }
70#else
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 INIT_LIST_HEAD(&s->s_files);
Nick Piggin6416ccb2010-08-18 04:37:38 +100072#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 INIT_LIST_HEAD(&s->s_instances);
74 INIT_HLIST_HEAD(&s->s_anon);
75 INIT_LIST_HEAD(&s->s_inodes);
Kentaro Makitada3bbdd2008-07-23 21:27:13 -070076 INIT_LIST_HEAD(&s->s_dentry_lru);
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 init_rwsem(&s->s_umount);
Ingo Molnar7892f2f2006-01-09 15:59:25 -080078 mutex_init(&s->s_lock);
Arjan van de Ven897c6ff2006-07-03 00:25:28 -070079 lockdep_set_class(&s->s_umount, &type->s_umount_key);
Ingo Molnarcf516242006-07-03 00:25:27 -070080 /*
81 * The locking rules for s_lock are up to the
82 * filesystem. For example ext3fs has different
83 * lock ordering than usbfs:
84 */
85 lockdep_set_class(&s->s_lock, &type->s_lock_key);
Peter Zijlstraada723d2009-02-18 14:48:30 -080086 /*
87 * sget() can have s_umount recursion.
88 *
89 * When it cannot find a suitable sb, it allocates a new
90 * one (this one), and tries again to find a suitable old
91 * one.
92 *
93 * In case that succeeds, it will acquire the s_umount
94 * lock of the old one. Since these are clearly distrinct
95 * locks, and this object isn't exposed yet, there's no
96 * risk of deadlocks.
97 *
98 * Annotate this by putting this lock in a different
99 * subclass.
100 */
101 down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING);
Al Virob20bd1a2010-03-22 08:53:19 -0400102 s->s_count = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 atomic_set(&s->s_active, 1);
Arjan van de Vena11f3a02006-03-23 03:00:33 -0800104 mutex_init(&s->s_vfs_rename_mutex);
Roland Dreier51ee0492010-04-27 14:23:57 -0700105 lockdep_set_class(&s->s_vfs_rename_mutex, &type->s_vfs_rename_key);
Ingo Molnard3be9152006-03-23 03:00:29 -0800106 mutex_init(&s->s_dquot.dqio_mutex);
107 mutex_init(&s->s_dquot.dqonoff_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 init_rwsem(&s->s_dquot.dqptr_sem);
109 init_waitqueue_head(&s->s_wait_unfrozen);
110 s->s_maxbytes = MAX_NON_LFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 s->s_op = &default_op;
112 s->s_time_gran = 1000000000;
113 }
114out:
115 return s;
116}
117
118/**
119 * destroy_super - frees a superblock
120 * @s: superblock to free
121 *
122 * Frees a superblock.
123 */
124static inline void destroy_super(struct super_block *s)
125{
Nick Piggin6416ccb2010-08-18 04:37:38 +1000126#ifdef CONFIG_SMP
127 free_percpu(s->s_files);
128#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 security_sb_free(s);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700130 kfree(s->s_subtype);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800131 kfree(s->s_options);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 kfree(s);
133}
134
135/* Superblock refcounting */
136
137/*
Al Viro35cf7ba2010-03-22 21:13:53 -0400138 * Drop a superblock's refcount. The caller must hold sb_lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139 */
Al Viro35cf7ba2010-03-22 21:13:53 -0400140void __put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 if (!--sb->s_count) {
Al Viro551de6f2010-03-22 19:36:35 -0400143 list_del_init(&sb->s_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 destroy_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146}
147
148/**
149 * put_super - drop a temporary reference to superblock
150 * @sb: superblock in question
151 *
152 * Drops a temporary reference, frees superblock if there's no
153 * references left.
154 */
Jens Axboe03ba3782009-09-09 09:08:54 +0200155void put_super(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156{
157 spin_lock(&sb_lock);
158 __put_super(sb);
159 spin_unlock(&sb_lock);
160}
161
162
163/**
Al Viro74dbbdd2009-05-06 01:07:50 -0400164 * deactivate_locked_super - drop an active reference to superblock
165 * @s: superblock to deactivate
166 *
Al Viro1712ac82010-03-22 15:22:31 -0400167 * Drops an active reference to superblock, converting it into a temprory
168 * one if there is no other active references left. In that case we
169 * tell fs driver to shut it down and drop the temporary reference we
170 * had just acquired.
171 *
172 * Caller holds exclusive lock on superblock; that lock is released.
Al Viro74dbbdd2009-05-06 01:07:50 -0400173 */
174void deactivate_locked_super(struct super_block *s)
175{
176 struct file_system_type *fs = s->s_type;
Al Virob20bd1a2010-03-22 08:53:19 -0400177 if (atomic_dec_and_test(&s->s_active)) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400178 fs->kill_sb(s);
179 put_filesystem(fs);
180 put_super(s);
181 } else {
182 up_write(&s->s_umount);
183 }
184}
185
186EXPORT_SYMBOL(deactivate_locked_super);
187
188/**
Al Viro1712ac82010-03-22 15:22:31 -0400189 * deactivate_super - drop an active reference to superblock
190 * @s: superblock to deactivate
191 *
192 * Variant of deactivate_locked_super(), except that superblock is *not*
193 * locked by caller. If we are going to drop the final active reference,
194 * lock will be acquired prior to that.
195 */
196void deactivate_super(struct super_block *s)
197{
198 if (!atomic_add_unless(&s->s_active, -1, 1)) {
199 down_write(&s->s_umount);
200 deactivate_locked_super(s);
201 }
202}
203
204EXPORT_SYMBOL(deactivate_super);
205
206/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 * grab_super - acquire an active reference
208 * @s: reference we are trying to make active
209 *
210 * Tries to acquire an active reference. grab_super() is used when we
211 * had just found a superblock in super_blocks or fs_type->fs_supers
212 * and want to turn it into a full-blown active reference. grab_super()
213 * is called with sb_lock held and drops it. Returns 1 in case of
214 * success, 0 if we had failed (superblock contents was already dead or
215 * dying when grab_super() had been called).
216 */
Josh Triplett9c4dbee2006-09-29 01:59:29 -0700217static int grab_super(struct super_block *s) __releases(sb_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218{
Al Virob20bd1a2010-03-22 08:53:19 -0400219 if (atomic_inc_not_zero(&s->s_active)) {
220 spin_unlock(&sb_lock);
Al Virob20bd1a2010-03-22 08:53:19 -0400221 return 1;
222 }
223 /* it's going away */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 s->s_count++;
225 spin_unlock(&sb_lock);
Al Viro1712ac82010-03-22 15:22:31 -0400226 /* wait for it to die */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 up_write(&s->s_umount);
229 put_super(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 return 0;
231}
232
David Howellscf9a2ae2006-08-29 19:05:54 +0100233/*
Al Viro914e2632006-10-18 13:55:46 -0400234 * Superblock locking. We really ought to get rid of these two.
235 */
236void lock_super(struct super_block * sb)
237{
238 get_fs_excl();
239 mutex_lock(&sb->s_lock);
240}
241
242void unlock_super(struct super_block * sb)
243{
244 put_fs_excl();
245 mutex_unlock(&sb->s_lock);
246}
247
248EXPORT_SYMBOL(lock_super);
249EXPORT_SYMBOL(unlock_super);
250
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251/**
252 * generic_shutdown_super - common helper for ->kill_sb()
253 * @sb: superblock to kill
254 *
255 * generic_shutdown_super() does all fs-independent work on superblock
256 * shutdown. Typical ->kill_sb() should pick all fs-specific objects
257 * that need destruction out of superblock, call generic_shutdown_super()
258 * and release aforementioned objects. Note: dentries and inodes _are_
259 * taken care of and do not need specific handling.
David Howellsc636ebd2006-10-11 01:22:19 -0700260 *
261 * Upon calling this function, the filesystem may no longer alter or
262 * rearrange the set of dentries belonging to this super_block, nor may it
263 * change the attachments of dentries to inodes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 */
265void generic_shutdown_super(struct super_block *sb)
266{
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800267 const struct super_operations *sop = sb->s_op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
Arjan van de Venefaee192009-01-06 07:20:54 -0800269
David Howellsc636ebd2006-10-11 01:22:19 -0700270 if (sb->s_root) {
271 shrink_dcache_for_umount(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200272 sync_filesystem(sb);
Al Viroa9e220f2009-05-05 22:10:44 -0400273 get_fs_excl();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 sb->s_flags &= ~MS_ACTIVE;
Arjan van de Venefaee192009-01-06 07:20:54 -0800275
Al Viro63997e92010-10-25 20:49:35 -0400276 fsnotify_unmount_inodes(&sb->s_inodes);
277
278 evict_inodes(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 if (sop->put_super)
281 sop->put_super(sb);
282
Al Viro63997e92010-10-25 20:49:35 -0400283 if (!list_empty(&sb->s_inodes)) {
Dave Jones7b4fe292006-02-07 12:58:48 -0800284 printk("VFS: Busy inodes after unmount of %s. "
285 "Self-destruct in 5 seconds. Have a nice day...\n",
286 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 }
Al Viroa9e220f2009-05-05 22:10:44 -0400288 put_fs_excl();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 }
290 spin_lock(&sb_lock);
291 /* should be initialized for __put_super_and_need_restart() */
Al Viro551de6f2010-03-22 19:36:35 -0400292 list_del_init(&sb->s_instances);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 spin_unlock(&sb_lock);
294 up_write(&sb->s_umount);
295}
296
297EXPORT_SYMBOL(generic_shutdown_super);
298
299/**
300 * sget - find or create a superblock
301 * @type: filesystem type superblock should belong to
302 * @test: comparison callback
303 * @set: setup callback
304 * @data: argument to each of them
305 */
306struct super_block *sget(struct file_system_type *type,
307 int (*test)(struct super_block *,void *),
308 int (*set)(struct super_block *,void *),
309 void *data)
310{
311 struct super_block *s = NULL;
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700312 struct super_block *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 int err;
314
315retry:
316 spin_lock(&sb_lock);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700317 if (test) {
318 list_for_each_entry(old, &type->fs_supers, s_instances) {
319 if (!test(old, data))
320 continue;
321 if (!grab_super(old))
322 goto retry;
Li Zefana3cfbb52009-03-12 14:31:29 -0700323 if (s) {
324 up_write(&s->s_umount);
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700325 destroy_super(s);
Al Viro7a4dec52010-08-09 12:05:43 -0400326 s = NULL;
Li Zefana3cfbb52009-03-12 14:31:29 -0700327 }
Al Virod3f21472010-03-23 11:11:05 -0400328 down_write(&old->s_umount);
Al Viro7a4dec52010-08-09 12:05:43 -0400329 if (unlikely(!(old->s_flags & MS_BORN))) {
330 deactivate_locked_super(old);
331 goto retry;
332 }
Matthias Kaehlcked4730122007-10-18 23:39:57 -0700333 return old;
334 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 }
336 if (!s) {
337 spin_unlock(&sb_lock);
Ingo Molnarcf516242006-07-03 00:25:27 -0700338 s = alloc_super(type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 if (!s)
340 return ERR_PTR(-ENOMEM);
341 goto retry;
342 }
343
344 err = set(s, data);
345 if (err) {
346 spin_unlock(&sb_lock);
Li Zefana3cfbb52009-03-12 14:31:29 -0700347 up_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 destroy_super(s);
349 return ERR_PTR(err);
350 }
351 s->s_type = type;
352 strlcpy(s->s_id, type->name, sizeof(s->s_id));
353 list_add_tail(&s->s_list, &super_blocks);
354 list_add(&s->s_instances, &type->fs_supers);
355 spin_unlock(&sb_lock);
356 get_filesystem(type);
357 return s;
358}
359
360EXPORT_SYMBOL(sget);
361
362void drop_super(struct super_block *sb)
363{
364 up_read(&sb->s_umount);
365 put_super(sb);
366}
367
368EXPORT_SYMBOL(drop_super);
369
Christoph Hellwige5004752009-05-05 16:08:56 +0200370/**
371 * sync_supers - helper for periodic superblock writeback
372 *
373 * Call the write_super method if present on all dirty superblocks in
374 * the system. This is for the periodic writeback used by most older
375 * filesystems. For data integrity superblock writeback use
376 * sync_filesystems() instead.
377 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 * Note: check the dirty flag before waiting, so we don't
379 * hold up the sync while mounting a device. (The newly
380 * mounted device won't need syncing.)
381 */
382void sync_supers(void)
383{
Al Virodca33252010-07-25 02:31:46 +0400384 struct super_block *sb, *p = NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700385
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400387 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro551de6f2010-03-22 19:36:35 -0400388 if (list_empty(&sb->s_instances))
389 continue;
Christoph Hellwige5004752009-05-05 16:08:56 +0200390 if (sb->s_op->write_super && sb->s_dirt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 sb->s_count++;
392 spin_unlock(&sb_lock);
Christoph Hellwige5004752009-05-05 16:08:56 +0200393
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 down_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200395 if (sb->s_root && sb->s_dirt)
396 sb->s_op->write_super(sb);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700397 up_read(&sb->s_umount);
Christoph Hellwige5004752009-05-05 16:08:56 +0200398
Kirill Korotaev618f0632005-06-23 00:09:54 -0700399 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400400 if (p)
401 __put_super(p);
402 p = sb;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700403 }
404 }
Al Virodca33252010-07-25 02:31:46 +0400405 if (p)
406 __put_super(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 spin_unlock(&sb_lock);
408}
409
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410/**
Al Viro01a05b32010-03-23 06:06:58 -0400411 * iterate_supers - call function for all active superblocks
412 * @f: function to call
413 * @arg: argument to pass to it
414 *
415 * Scans the superblock list and calls given function, passing it
416 * locked superblock and given argument.
417 */
418void iterate_supers(void (*f)(struct super_block *, void *), void *arg)
419{
Al Virodca33252010-07-25 02:31:46 +0400420 struct super_block *sb, *p = NULL;
Al Viro01a05b32010-03-23 06:06:58 -0400421
422 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400423 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro01a05b32010-03-23 06:06:58 -0400424 if (list_empty(&sb->s_instances))
425 continue;
426 sb->s_count++;
427 spin_unlock(&sb_lock);
428
429 down_read(&sb->s_umount);
430 if (sb->s_root)
431 f(sb, arg);
432 up_read(&sb->s_umount);
433
434 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400435 if (p)
436 __put_super(p);
437 p = sb;
Al Viro01a05b32010-03-23 06:06:58 -0400438 }
Al Virodca33252010-07-25 02:31:46 +0400439 if (p)
440 __put_super(p);
Al Viro01a05b32010-03-23 06:06:58 -0400441 spin_unlock(&sb_lock);
442}
443
444/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 * get_super - get the superblock of a device
446 * @bdev: device to get the superblock for
447 *
448 * Scans the superblock list and finds the superblock of the file system
449 * mounted on the device given. %NULL is returned if no match is found.
450 */
451
Al Virodf40c012010-03-22 20:23:25 -0400452struct super_block *get_super(struct block_device *bdev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700454 struct super_block *sb;
455
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 if (!bdev)
457 return NULL;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700458
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700460rescan:
461 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro551de6f2010-03-22 19:36:35 -0400462 if (list_empty(&sb->s_instances))
463 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700464 if (sb->s_bdev == bdev) {
465 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700467 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400468 /* still alive? */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700469 if (sb->s_root)
470 return sb;
471 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400472 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700473 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400474 __put_super(sb);
475 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 }
477 }
478 spin_unlock(&sb_lock);
479 return NULL;
480}
481
482EXPORT_SYMBOL(get_super);
Christoph Hellwig45042302009-08-03 23:28:35 +0200483
484/**
485 * get_active_super - get an active reference to the superblock of a device
486 * @bdev: device to get the superblock for
487 *
488 * Scans the superblock list and finds the superblock of the file system
489 * mounted on the device given. Returns the superblock with an active
Al Virod3f21472010-03-23 11:11:05 -0400490 * reference or %NULL if none was found.
Christoph Hellwig45042302009-08-03 23:28:35 +0200491 */
492struct super_block *get_active_super(struct block_device *bdev)
493{
494 struct super_block *sb;
495
496 if (!bdev)
497 return NULL;
498
Al Viro14945832010-03-22 20:15:33 -0400499restart:
Christoph Hellwig45042302009-08-03 23:28:35 +0200500 spin_lock(&sb_lock);
501 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro551de6f2010-03-22 19:36:35 -0400502 if (list_empty(&sb->s_instances))
503 continue;
Al Viro14945832010-03-22 20:15:33 -0400504 if (sb->s_bdev == bdev) {
505 if (grab_super(sb)) /* drops sb_lock */
506 return sb;
507 else
508 goto restart;
509 }
Christoph Hellwig45042302009-08-03 23:28:35 +0200510 }
511 spin_unlock(&sb_lock);
512 return NULL;
513}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514
Al Virodf40c012010-03-22 20:23:25 -0400515struct super_block *user_get_super(dev_t dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516{
Kirill Korotaev618f0632005-06-23 00:09:54 -0700517 struct super_block *sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 spin_lock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700520rescan:
521 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro551de6f2010-03-22 19:36:35 -0400522 if (list_empty(&sb->s_instances))
523 continue;
Kirill Korotaev618f0632005-06-23 00:09:54 -0700524 if (sb->s_dev == dev) {
525 sb->s_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 spin_unlock(&sb_lock);
Kirill Korotaev618f0632005-06-23 00:09:54 -0700527 down_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400528 /* still alive? */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700529 if (sb->s_root)
530 return sb;
531 up_read(&sb->s_umount);
Al Virodf40c012010-03-22 20:23:25 -0400532 /* nope, got unmounted */
Kirill Korotaev618f0632005-06-23 00:09:54 -0700533 spin_lock(&sb_lock);
Al Virodf40c012010-03-22 20:23:25 -0400534 __put_super(sb);
535 goto rescan;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 }
537 }
538 spin_unlock(&sb_lock);
539 return NULL;
540}
541
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 * do_remount_sb - asks filesystem to change mount options.
544 * @sb: superblock in question
545 * @flags: numeric part of options
546 * @data: the rest of options
547 * @force: whether or not to force the change
548 *
549 * Alters the mount options of a mounted file system.
550 */
551int do_remount_sb(struct super_block *sb, int flags, void *data, int force)
552{
553 int retval;
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400554 int remount_ro;
Christoph Hellwig45042302009-08-03 23:28:35 +0200555
556 if (sb->s_frozen != SB_UNFROZEN)
557 return -EBUSY;
558
David Howells93614012006-09-30 20:45:40 +0200559#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 if (!(flags & MS_RDONLY) && bdev_read_only(sb->s_bdev))
561 return -EACCES;
David Howells93614012006-09-30 20:45:40 +0200562#endif
Christoph Hellwig45042302009-08-03 23:28:35 +0200563
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 if (flags & MS_RDONLY)
565 acct_auto_close(sb);
566 shrink_dcache_sb(sb);
Jan Kara60b06802009-04-27 16:43:53 +0200567 sync_filesystem(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568
Nick Piggind208bbd2009-12-21 16:28:53 -0800569 remount_ro = (flags & MS_RDONLY) && !(sb->s_flags & MS_RDONLY);
Nick Piggind208bbd2009-12-21 16:28:53 -0800570
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 /* If we are remounting RDONLY and current sb is read/write,
572 make sure there are no rw files opened */
Nick Piggind208bbd2009-12-21 16:28:53 -0800573 if (remount_ro) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 if (force)
575 mark_files_ro(sb);
J. R. Okajimab0895512009-06-17 01:16:50 +0900576 else if (!fs_may_remount_ro(sb))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 return -EBUSY;
578 }
579
580 if (sb->s_op->remount_fs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 retval = sb->s_op->remount_fs(sb, &flags, data);
J. R. Okajimab0895512009-06-17 01:16:50 +0900582 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 return retval;
584 }
585 sb->s_flags = (sb->s_flags & ~MS_RMT_MASK) | (flags & MS_RMT_MASK);
Christoph Hellwigc79d9672010-05-19 07:16:40 -0400586
Nick Piggind208bbd2009-12-21 16:28:53 -0800587 /*
588 * Some filesystems modify their metadata via some other path than the
589 * bdev buffer cache (eg. use a private mapping, or directories in
590 * pagecache, etc). Also file data modifications go via their own
591 * mappings. So If we try to mount readonly then copy the filesystem
592 * from bdev, we could get stale data, so invalidate it to give a best
593 * effort at coherency.
594 */
595 if (remount_ro && sb->s_bdev)
596 invalidate_bdev(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 return 0;
598}
599
Jens Axboea2a95372009-03-17 09:38:40 +0100600static void do_emergency_remount(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601{
Al Virodca33252010-07-25 02:31:46 +0400602 struct super_block *sb, *p = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603
604 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400605 list_for_each_entry(sb, &super_blocks, s_list) {
Al Viro551de6f2010-03-22 19:36:35 -0400606 if (list_empty(&sb->s_instances))
607 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 sb->s_count++;
609 spin_unlock(&sb_lock);
Al Viro443b94b2009-05-05 23:48:50 -0400610 down_write(&sb->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 if (sb->s_root && sb->s_bdev && !(sb->s_flags & MS_RDONLY)) {
612 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613 * What lock protects sb->s_flags??
614 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 do_remount_sb(sb, MS_RDONLY, NULL, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 }
Al Viro443b94b2009-05-05 23:48:50 -0400617 up_write(&sb->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 spin_lock(&sb_lock);
Al Virodca33252010-07-25 02:31:46 +0400619 if (p)
620 __put_super(p);
621 p = sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 }
Al Virodca33252010-07-25 02:31:46 +0400623 if (p)
624 __put_super(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 spin_unlock(&sb_lock);
Jens Axboea2a95372009-03-17 09:38:40 +0100626 kfree(work);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 printk("Emergency Remount complete\n");
628}
629
630void emergency_remount(void)
631{
Jens Axboea2a95372009-03-17 09:38:40 +0100632 struct work_struct *work;
633
634 work = kmalloc(sizeof(*work), GFP_ATOMIC);
635 if (work) {
636 INIT_WORK(work, do_emergency_remount);
637 schedule_work(work);
638 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639}
640
641/*
642 * Unnamed block devices are dummy devices used by virtual
643 * filesystems which don't use real block-devices. -- jrs
644 */
645
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400646static DEFINE_IDA(unnamed_dev_ida);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647static DEFINE_SPINLOCK(unnamed_dev_lock);/* protects the above */
Al Viroc63e09e2009-06-24 02:05:18 -0400648static int unnamed_dev_start = 0; /* don't bother trying below it */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649
650int set_anon_super(struct super_block *s, void *data)
651{
652 int dev;
653 int error;
654
655 retry:
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400656 if (ida_pre_get(&unnamed_dev_ida, GFP_ATOMIC) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 return -ENOMEM;
658 spin_lock(&unnamed_dev_lock);
Al Viroc63e09e2009-06-24 02:05:18 -0400659 error = ida_get_new_above(&unnamed_dev_ida, unnamed_dev_start, &dev);
Al Virof21f6222009-06-24 03:12:00 -0400660 if (!error)
661 unnamed_dev_start = dev + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 spin_unlock(&unnamed_dev_lock);
663 if (error == -EAGAIN)
664 /* We raced and lost with another CPU. */
665 goto retry;
666 else if (error)
667 return -EAGAIN;
668
669 if ((dev & MAX_ID_MASK) == (1 << MINORBITS)) {
670 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400671 ida_remove(&unnamed_dev_ida, dev);
Al Virof21f6222009-06-24 03:12:00 -0400672 if (unnamed_dev_start > dev)
673 unnamed_dev_start = dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 spin_unlock(&unnamed_dev_lock);
675 return -EMFILE;
676 }
677 s->s_dev = MKDEV(0, dev & MINORMASK);
Jörn Engel5129a462010-04-25 08:54:42 +0200678 s->s_bdi = &noop_backing_dev_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 return 0;
680}
681
682EXPORT_SYMBOL(set_anon_super);
683
684void kill_anon_super(struct super_block *sb)
685{
686 int slot = MINOR(sb->s_dev);
687
688 generic_shutdown_super(sb);
689 spin_lock(&unnamed_dev_lock);
Alexey Dobriyanad76cbc2008-08-28 06:26:23 +0400690 ida_remove(&unnamed_dev_ida, slot);
Al Viroc63e09e2009-06-24 02:05:18 -0400691 if (slot < unnamed_dev_start)
692 unnamed_dev_start = slot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 spin_unlock(&unnamed_dev_lock);
694}
695
696EXPORT_SYMBOL(kill_anon_super);
697
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698void kill_litter_super(struct super_block *sb)
699{
700 if (sb->s_root)
701 d_genocide(sb->s_root);
702 kill_anon_super(sb);
703}
704
705EXPORT_SYMBOL(kill_litter_super);
706
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700707static int ns_test_super(struct super_block *sb, void *data)
708{
709 return sb->s_fs_info == data;
710}
711
712static int ns_set_super(struct super_block *sb, void *data)
713{
714 sb->s_fs_info = data;
715 return set_anon_super(sb, NULL);
716}
717
Al Viroceefda62010-07-26 13:16:50 +0400718struct dentry *mount_ns(struct file_system_type *fs_type, int flags,
719 void *data, int (*fill_super)(struct super_block *, void *, int))
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700720{
721 struct super_block *sb;
722
723 sb = sget(fs_type, ns_test_super, ns_set_super, data);
724 if (IS_ERR(sb))
Al Viroceefda62010-07-26 13:16:50 +0400725 return ERR_CAST(sb);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700726
727 if (!sb->s_root) {
728 int err;
729 sb->s_flags = flags;
730 err = fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
731 if (err) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400732 deactivate_locked_super(sb);
Al Viroceefda62010-07-26 13:16:50 +0400733 return ERR_PTR(err);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700734 }
735
736 sb->s_flags |= MS_ACTIVE;
737 }
738
Al Viroceefda62010-07-26 13:16:50 +0400739 return dget(sb->s_root);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700740}
741
Al Viroceefda62010-07-26 13:16:50 +0400742EXPORT_SYMBOL(mount_ns);
Serge E. Hallyn909e6d92009-04-06 19:01:07 -0700743
David Howells93614012006-09-30 20:45:40 +0200744#ifdef CONFIG_BLOCK
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745static int set_bdev_super(struct super_block *s, void *data)
746{
747 s->s_bdev = data;
748 s->s_dev = s->s_bdev->bd_dev;
Jens Axboe32a88aa2009-09-16 15:02:33 +0200749
750 /*
751 * We set the bdi here to the queue backing, file systems can
752 * overwrite this in ->fill_super()
753 */
754 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 return 0;
756}
757
758static int test_bdev_super(struct super_block *s, void *data)
759{
760 return (void *)s->s_bdev == data;
761}
762
Al Viro152a0832010-07-25 00:46:55 +0400763struct dentry *mount_bdev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 int flags, const char *dev_name, void *data,
Al Viro152a0832010-07-25 00:46:55 +0400765 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766{
767 struct block_device *bdev;
768 struct super_block *s;
Al Viro30c40d22008-02-22 19:50:45 -0500769 fmode_t mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 int error = 0;
771
Al Viro30c40d22008-02-22 19:50:45 -0500772 if (!(flags & MS_RDONLY))
773 mode |= FMODE_WRITE;
774
775 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 if (IS_ERR(bdev))
Al Viro152a0832010-07-25 00:46:55 +0400777 return ERR_CAST(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
779 /*
780 * once the super is inserted into the list by sget, s_umount
781 * will protect the lockfs code from trying to start a snapshot
782 * while we are mounting
783 */
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200784 mutex_lock(&bdev->bd_fsfreeze_mutex);
785 if (bdev->bd_fsfreeze_count > 0) {
786 mutex_unlock(&bdev->bd_fsfreeze_mutex);
787 error = -EBUSY;
788 goto error_bdev;
789 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
Christoph Hellwig4fadd7b2009-08-03 23:28:06 +0200791 mutex_unlock(&bdev->bd_fsfreeze_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 if (IS_ERR(s))
David Howells454e2392006-06-23 02:02:57 -0700793 goto error_s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
795 if (s->s_root) {
796 if ((flags ^ s->s_flags) & MS_RDONLY) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400797 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700798 error = -EBUSY;
799 goto error_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 }
David Howells454e2392006-06-23 02:02:57 -0700801
Tejun Heo4f331f02010-07-20 15:18:07 -0700802 /*
803 * s_umount nests inside bd_mutex during
804 * __invalidate_device(). close_bdev_exclusive()
805 * acquires bd_mutex and can't be called under
806 * s_umount. Drop s_umount temporarily. This is safe
807 * as we're holding an active reference.
808 */
809 up_write(&s->s_umount);
Al Viro30c40d22008-02-22 19:50:45 -0500810 close_bdev_exclusive(bdev, mode);
Tejun Heo4f331f02010-07-20 15:18:07 -0700811 down_write(&s->s_umount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 } else {
813 char b[BDEVNAME_SIZE];
814
815 s->s_flags = flags;
Al Viro30c40d22008-02-22 19:50:45 -0500816 s->s_mode = mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
Pekka Enberge78c9a02006-01-08 01:03:39 -0800818 sb_set_blocksize(s, block_size(bdev));
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800819 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400821 deactivate_locked_super(s);
David Howells454e2392006-06-23 02:02:57 -0700822 goto error;
Greg Kroah-Hartmanfa675762006-02-22 09:39:02 -0800823 }
David Howells454e2392006-06-23 02:02:57 -0700824
825 s->s_flags |= MS_ACTIVE;
Theodore Ts'o87d8fe12009-01-03 09:47:09 -0500826 bdev->bd_super = s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 }
828
Al Viro152a0832010-07-25 00:46:55 +0400829 return dget(s->s_root);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830
David Howells454e2392006-06-23 02:02:57 -0700831error_s:
832 error = PTR_ERR(s);
833error_bdev:
Al Viro30c40d22008-02-22 19:50:45 -0500834 close_bdev_exclusive(bdev, mode);
David Howells454e2392006-06-23 02:02:57 -0700835error:
Al Viro152a0832010-07-25 00:46:55 +0400836 return ERR_PTR(error);
837}
838EXPORT_SYMBOL(mount_bdev);
839
840int get_sb_bdev(struct file_system_type *fs_type,
841 int flags, const char *dev_name, void *data,
842 int (*fill_super)(struct super_block *, void *, int),
843 struct vfsmount *mnt)
844{
845 struct dentry *root;
846
847 root = mount_bdev(fs_type, flags, dev_name, data, fill_super);
848 if (IS_ERR(root))
849 return PTR_ERR(root);
850 mnt->mnt_root = root;
851 mnt->mnt_sb = root->d_sb;
852 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853}
854
855EXPORT_SYMBOL(get_sb_bdev);
856
857void kill_block_super(struct super_block *sb)
858{
859 struct block_device *bdev = sb->s_bdev;
Al Viro30c40d22008-02-22 19:50:45 -0500860 fmode_t mode = sb->s_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861
H Hartley Sweetenddbaaf32009-04-29 20:14:57 -0400862 bdev->bd_super = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 generic_shutdown_super(sb);
864 sync_blockdev(bdev);
Al Viro30c40d22008-02-22 19:50:45 -0500865 close_bdev_exclusive(bdev, mode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866}
867
868EXPORT_SYMBOL(kill_block_super);
David Howells93614012006-09-30 20:45:40 +0200869#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870
Al Viro3c26ff62010-07-25 11:46:36 +0400871struct dentry *mount_nodev(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 int flags, void *data,
Al Viro3c26ff62010-07-25 11:46:36 +0400873 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874{
875 int error;
876 struct super_block *s = sget(fs_type, NULL, set_anon_super, NULL);
877
878 if (IS_ERR(s))
Al Viro3c26ff62010-07-25 11:46:36 +0400879 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880
881 s->s_flags = flags;
882
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800883 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400885 deactivate_locked_super(s);
Al Viro3c26ff62010-07-25 11:46:36 +0400886 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 }
888 s->s_flags |= MS_ACTIVE;
Al Viro3c26ff62010-07-25 11:46:36 +0400889 return dget(s->s_root);
890}
891EXPORT_SYMBOL(mount_nodev);
892
893int get_sb_nodev(struct file_system_type *fs_type,
894 int flags, void *data,
895 int (*fill_super)(struct super_block *, void *, int),
896 struct vfsmount *mnt)
897{
898 struct dentry *root;
899
900 root = mount_nodev(fs_type, flags, data, fill_super);
901 if (IS_ERR(root))
902 return PTR_ERR(root);
903 mnt->mnt_root = root;
904 mnt->mnt_sb = root->d_sb;
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800905 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907EXPORT_SYMBOL(get_sb_nodev);
908
909static int compare_single(struct super_block *s, void *p)
910{
911 return 1;
912}
913
Al Virofc14f2f2010-07-25 01:48:30 +0400914struct dentry *mount_single(struct file_system_type *fs_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 int flags, void *data,
Al Virofc14f2f2010-07-25 01:48:30 +0400916 int (*fill_super)(struct super_block *, void *, int))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917{
918 struct super_block *s;
919 int error;
920
921 s = sget(fs_type, compare_single, set_anon_super, NULL);
922 if (IS_ERR(s))
Al Virofc14f2f2010-07-25 01:48:30 +0400923 return ERR_CAST(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 if (!s->s_root) {
925 s->s_flags = flags;
Theodore Ts'o9b04c992006-03-24 03:15:10 -0800926 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 if (error) {
Al Viro74dbbdd2009-05-06 01:07:50 -0400928 deactivate_locked_super(s);
Al Virofc14f2f2010-07-25 01:48:30 +0400929 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 }
931 s->s_flags |= MS_ACTIVE;
Kay Sievers9329d1b2009-12-18 21:18:15 +0100932 } else {
933 do_remount_sb(s, flags, data, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 }
Al Virofc14f2f2010-07-25 01:48:30 +0400935 return dget(s->s_root);
936}
937EXPORT_SYMBOL(mount_single);
938
939int get_sb_single(struct file_system_type *fs_type,
940 int flags, void *data,
941 int (*fill_super)(struct super_block *, void *, int),
942 struct vfsmount *mnt)
943{
944 struct dentry *root;
945 root = mount_single(fs_type, flags, data, fill_super);
946 if (IS_ERR(root))
947 return PTR_ERR(root);
948 mnt->mnt_root = root;
949 mnt->mnt_sb = root->d_sb;
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800950 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951}
952
953EXPORT_SYMBOL(get_sb_single);
954
955struct vfsmount *
Trond Myklebustbb4a58b2006-06-09 09:34:15 -0400956vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 struct vfsmount *mnt;
Al Viroc96e41e2010-07-25 00:17:56 +0400959 struct dentry *root;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 char *secdata = NULL;
David Howells454e2392006-06-23 02:02:57 -0700961 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962
963 if (!type)
964 return ERR_PTR(-ENODEV);
965
David Howells454e2392006-06-23 02:02:57 -0700966 error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 mnt = alloc_vfsmnt(name);
968 if (!mnt)
969 goto out;
970
Al Viro80893522010-02-05 09:30:46 -0500971 if (flags & MS_KERNMOUNT)
972 mnt->mnt_flags = MNT_INTERNAL;
973
Eric Parise0007522008-03-05 10:31:54 -0500974 if (data && !(type->fs_flags & FS_BINARY_MOUNTDATA)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 secdata = alloc_secdata();
David Howells454e2392006-06-23 02:02:57 -0700976 if (!secdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 goto out_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978
Eric Parise0007522008-03-05 10:31:54 -0500979 error = security_sb_copy_data(data, secdata);
David Howells454e2392006-06-23 02:02:57 -0700980 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 goto out_free_secdata;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 }
983
Al Viroc96e41e2010-07-25 00:17:56 +0400984 if (type->mount) {
985 root = type->mount(type, flags, name, data);
986 if (IS_ERR(root)) {
987 error = PTR_ERR(root);
988 goto out_free_secdata;
989 }
990 mnt->mnt_root = root;
991 mnt->mnt_sb = root->d_sb;
992 } else {
993 error = type->get_sb(type, flags, name, data, mnt);
994 if (error < 0)
995 goto out_free_secdata;
996 }
Lee Schermerhornb4c07bc2007-07-15 23:40:54 -0700997 BUG_ON(!mnt->mnt_sb);
Jörn Engel5129a462010-04-25 08:54:42 +0200998 WARN_ON(!mnt->mnt_sb->s_bdi);
Al Viro7a4dec52010-08-09 12:05:43 -0400999 mnt->mnt_sb->s_flags |= MS_BORN;
David Howells454e2392006-06-23 02:02:57 -07001000
Jörn Engel5129a462010-04-25 08:54:42 +02001001 error = security_sb_kern_mount(mnt->mnt_sb, flags, secdata);
1002 if (error)
1003 goto out_sb;
David Howells454e2392006-06-23 02:02:57 -07001004
Jeff Layton42cb56a2009-09-18 13:05:53 -07001005 /*
1006 * filesystems should never set s_maxbytes larger than MAX_LFS_FILESIZE
1007 * but s_maxbytes was an unsigned long long for many releases. Throw
1008 * this warning for a little while to try and catch filesystems that
1009 * violate this rule. This warning should be either removed or
1010 * converted to a BUG() in 2.6.34.
1011 */
1012 WARN((mnt->mnt_sb->s_maxbytes < 0), "%s set sb->s_maxbytes to "
1013 "negative value (%lld)\n", type->name, mnt->mnt_sb->s_maxbytes);
1014
David Howells454e2392006-06-23 02:02:57 -07001015 mnt->mnt_mountpoint = mnt->mnt_root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 mnt->mnt_parent = mnt;
David Howells454e2392006-06-23 02:02:57 -07001017 up_write(&mnt->mnt_sb->s_umount);
Gerald Schaefer8680e222005-06-21 17:15:16 -07001018 free_secdata(secdata);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 return mnt;
1020out_sb:
David Howells454e2392006-06-23 02:02:57 -07001021 dput(mnt->mnt_root);
Al Viro74dbbdd2009-05-06 01:07:50 -04001022 deactivate_locked_super(mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023out_free_secdata:
1024 free_secdata(secdata);
1025out_mnt:
1026 free_vfsmnt(mnt);
1027out:
David Howells454e2392006-06-23 02:02:57 -07001028 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029}
1030
Trond Myklebustbb4a58b2006-06-09 09:34:15 -04001031EXPORT_SYMBOL_GPL(vfs_kern_mount);
1032
Josef Bacik18e9e512010-03-23 10:34:56 -04001033/**
Randy Dunlap7000d3c2010-05-24 22:22:34 -07001034 * freeze_super - lock the filesystem and force it into a consistent state
1035 * @sb: the super to lock
Josef Bacik18e9e512010-03-23 10:34:56 -04001036 *
1037 * Syncs the super to make sure the filesystem is consistent and calls the fs's
1038 * freeze_fs. Subsequent calls to this without first thawing the fs will return
1039 * -EBUSY.
1040 */
1041int freeze_super(struct super_block *sb)
1042{
1043 int ret;
1044
1045 atomic_inc(&sb->s_active);
1046 down_write(&sb->s_umount);
1047 if (sb->s_frozen) {
1048 deactivate_locked_super(sb);
1049 return -EBUSY;
1050 }
1051
1052 if (sb->s_flags & MS_RDONLY) {
1053 sb->s_frozen = SB_FREEZE_TRANS;
1054 smp_wmb();
1055 up_write(&sb->s_umount);
1056 return 0;
1057 }
1058
1059 sb->s_frozen = SB_FREEZE_WRITE;
1060 smp_wmb();
1061
1062 sync_filesystem(sb);
1063
1064 sb->s_frozen = SB_FREEZE_TRANS;
1065 smp_wmb();
1066
1067 sync_blockdev(sb->s_bdev);
1068 if (sb->s_op->freeze_fs) {
1069 ret = sb->s_op->freeze_fs(sb);
1070 if (ret) {
1071 printk(KERN_ERR
1072 "VFS:Filesystem freeze failed\n");
1073 sb->s_frozen = SB_UNFROZEN;
1074 deactivate_locked_super(sb);
1075 return ret;
1076 }
1077 }
1078 up_write(&sb->s_umount);
1079 return 0;
1080}
1081EXPORT_SYMBOL(freeze_super);
1082
1083/**
1084 * thaw_super -- unlock filesystem
1085 * @sb: the super to thaw
1086 *
1087 * Unlocks the filesystem and marks it writeable again after freeze_super().
1088 */
1089int thaw_super(struct super_block *sb)
1090{
1091 int error;
1092
1093 down_write(&sb->s_umount);
1094 if (sb->s_frozen == SB_UNFROZEN) {
1095 up_write(&sb->s_umount);
1096 return -EINVAL;
1097 }
1098
1099 if (sb->s_flags & MS_RDONLY)
1100 goto out;
1101
1102 if (sb->s_op->unfreeze_fs) {
1103 error = sb->s_op->unfreeze_fs(sb);
1104 if (error) {
1105 printk(KERN_ERR
1106 "VFS:Filesystem thaw failed\n");
1107 sb->s_frozen = SB_FREEZE_TRANS;
1108 up_write(&sb->s_umount);
1109 return error;
1110 }
1111 }
1112
1113out:
1114 sb->s_frozen = SB_UNFROZEN;
1115 smp_wmb();
1116 wake_up(&sb->s_wait_unfrozen);
1117 deactivate_locked_super(sb);
1118
1119 return 0;
1120}
1121EXPORT_SYMBOL(thaw_super);
1122
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -07001123static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
1124{
1125 int err;
1126 const char *subtype = strchr(fstype, '.');
1127 if (subtype) {
1128 subtype++;
1129 err = -EINVAL;
1130 if (!subtype[0])
1131 goto err;
1132 } else
1133 subtype = "";
1134
1135 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
1136 err = -ENOMEM;
1137 if (!mnt->mnt_sb->s_subtype)
1138 goto err;
1139 return mnt;
1140
1141 err:
1142 mntput(mnt);
1143 return ERR_PTR(err);
1144}
1145
Trond Myklebustbb4a58b2006-06-09 09:34:15 -04001146struct vfsmount *
1147do_kern_mount(const char *fstype, int flags, const char *name, void *data)
1148{
1149 struct file_system_type *type = get_fs_type(fstype);
1150 struct vfsmount *mnt;
1151 if (!type)
1152 return ERR_PTR(-ENODEV);
1153 mnt = vfs_kern_mount(type, flags, name, data);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -07001154 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
1155 !mnt->mnt_sb->s_subtype)
1156 mnt = fs_set_subtype(mnt, fstype);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -04001157 put_filesystem(type);
1158 return mnt;
1159}
Al Viro8a4e98d2008-02-24 01:43:03 -05001160EXPORT_SYMBOL_GPL(do_kern_mount);
Trond Myklebustbb4a58b2006-06-09 09:34:15 -04001161
Pavel Emelyanov8bf97252007-10-18 23:40:02 -07001162struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163{
Pavel Emelyanov8bf97252007-10-18 23:40:02 -07001164 return vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165}
1166
Pavel Emelyanov8bf97252007-10-18 23:40:02 -07001167EXPORT_SYMBOL_GPL(kern_mount_data);